SAFE public projects git trees. - safe/jmp/linux-2.6/blob - 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
  73 #include "buffer_head_io.h"
  74
  75 static struct kmem_cache *ocfs2_inode_cachep = NULL;
  76 struct kmem_cache *ocfs2_dquot_cachep;
  77 struct kmem_cache *ocfs2_qf_chunk_cachep;
  78
  79 /* OCFS2 needs to schedule several differnt types of work which
  80  * require cluster locking, disk I/O, recovery waits, etc. Since these
  81  * types of work tend to be heavy we avoid using the kernel events
  82  * workqueue and schedule on our own. */
  83 struct workqueue_struct *ocfs2_wq = NULL;
  84
  85 static struct dentry *ocfs2_debugfs_root = NULL;
  86
  87 MODULE_AUTHOR("Oracle");
  88 MODULE_LICENSE("GPL");
  89
  90 struct mount_options
  91 {
  92         unsigned long   commit_interval;
  93         unsigned long   mount_opt;
  94         unsigned int    atime_quantum;
  95         signed short    slot;
  96         unsigned int    localalloc_opt;
  97         char            cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
  98 };
  99
 100 static int ocfs2_parse_options(struct super_block *sb, char *options,
 101                                struct mount_options *mopt,
 102                                int is_remount);
 103 static int ocfs2_check_set_options(struct super_block *sb,
 104                                    struct mount_options *options);
 105 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt);
 106 static void ocfs2_put_super(struct super_block *sb);
 107 static int ocfs2_mount_volume(struct super_block *sb);
 108 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
 109 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
 110 static int ocfs2_initialize_mem_caches(void);
 111 static void ocfs2_free_mem_caches(void);
 112 static void ocfs2_delete_osb(struct ocfs2_super *osb);
 113
 114 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
 115
 116 static int ocfs2_sync_fs(struct super_block *sb, int wait);
 117
 118 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
 119 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
 120 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
 121 static int ocfs2_check_volume(struct ocfs2_super *osb);
 122 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
 123                                struct buffer_head *bh,
 124                                u32 sectsize,
 125                                struct ocfs2_blockcheck_stats *stats);
 126 static int ocfs2_initialize_super(struct super_block *sb,
 127                                   struct buffer_head *bh,
 128                                   int sector_size,
 129                                   struct ocfs2_blockcheck_stats *stats);
 130 static int ocfs2_get_sector(struct super_block *sb,
 131                             struct buffer_head **bh,
 132                             int block,
 133                             int sect_size);
 134 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
 135 static void ocfs2_destroy_inode(struct inode *inode);
 136 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
 137 static int ocfs2_enable_quotas(struct ocfs2_super *osb);
 138 static void ocfs2_disable_quotas(struct ocfs2_super *osb);
 139
 140 static const struct super_operations ocfs2_sops = {
 141         .statfs         = ocfs2_statfs,
 142         .alloc_inode    = ocfs2_alloc_inode,
 143         .destroy_inode  = ocfs2_destroy_inode,
 144         .drop_inode     = ocfs2_drop_inode,
 145         .clear_inode    = ocfs2_clear_inode,
 146         .delete_inode   = ocfs2_delete_inode,
 147         .sync_fs        = ocfs2_sync_fs,
 148         .put_super      = ocfs2_put_super,
 149         .remount_fs     = ocfs2_remount,
 150         .show_options   = ocfs2_show_options,
 151         .quota_read     = ocfs2_quota_read,
 152         .quota_write    = ocfs2_quota_write,
 153 };
 154
 155 enum {
 156         Opt_barrier,
 157         Opt_err_panic,
 158         Opt_err_ro,
 159         Opt_intr,
 160         Opt_nointr,
 161         Opt_hb_none,
 162         Opt_hb_local,
 163         Opt_data_ordered,
 164         Opt_data_writeback,
 165         Opt_atime_quantum,
 166         Opt_slot,
 167         Opt_commit,
 168         Opt_localalloc,
 169         Opt_localflocks,
 170         Opt_stack,
 171         Opt_user_xattr,
 172         Opt_nouser_xattr,
 173         Opt_inode64,
 174         Opt_acl,
 175         Opt_noacl,
 176         Opt_usrquota,
 177         Opt_grpquota,
 178         Opt_err,
 179 };
 180
 181 static const match_table_t tokens = {
 182         {Opt_barrier, "barrier=%u"},
 183         {Opt_err_panic, "errors=panic"},
 184         {Opt_err_ro, "errors=remount-ro"},
 185         {Opt_intr, "intr"},
 186         {Opt_nointr, "nointr"},
 187         {Opt_hb_none, OCFS2_HB_NONE},
 188         {Opt_hb_local, OCFS2_HB_LOCAL},
 189         {Opt_data_ordered, "data=ordered"},
 190         {Opt_data_writeback, "data=writeback"},
 191         {Opt_atime_quantum, "atime_quantum=%u"},
 192         {Opt_slot, "preferred_slot=%u"},
 193         {Opt_commit, "commit=%u"},
 194         {Opt_localalloc, "localalloc=%d"},
 195         {Opt_localflocks, "localflocks"},
 196         {Opt_stack, "cluster_stack=%s"},
 197         {Opt_user_xattr, "user_xattr"},
 198         {Opt_nouser_xattr, "nouser_xattr"},
 199         {Opt_inode64, "inode64"},
 200         {Opt_acl, "acl"},
 201         {Opt_noacl, "noacl"},
 202         {Opt_usrquota, "usrquota"},
 203         {Opt_grpquota, "grpquota"},
 204         {Opt_err, NULL}
 205 };
 206
 207 #ifdef CONFIG_DEBUG_FS
 208 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
 209 {
 210         struct ocfs2_cluster_connection *cconn = osb->cconn;
 211         struct ocfs2_recovery_map *rm = osb->recovery_map;
 212         struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
 213         int i, out = 0;
 214
 215         out += snprintf(buf + out, len - out,
 216                         "%10s => Id: %-s  Uuid: %-s  Gen: 0x%X  Label: %-s\n",
 217                         "Device", osb->dev_str, osb->uuid_str,
 218                         osb->fs_generation, osb->vol_label);
 219
 220         out += snprintf(buf + out, len - out,
 221                         "%10s => State: %d  Flags: 0x%lX\n", "Volume",
 222                         atomic_read(&osb->vol_state), osb->osb_flags);
 223
 224         out += snprintf(buf + out, len - out,
 225                         "%10s => Block: %lu  Cluster: %d\n", "Sizes",
 226                         osb->sb->s_blocksize, osb->s_clustersize);
 227
 228         out += snprintf(buf + out, len - out,
 229                         "%10s => Compat: 0x%X  Incompat: 0x%X  "
 230                         "ROcompat: 0x%X\n",
 231                         "Features", osb->s_feature_compat,
 232                         osb->s_feature_incompat, osb->s_feature_ro_compat);
 233
 234         out += snprintf(buf + out, len - out,
 235                         "%10s => Opts: 0x%lX  AtimeQuanta: %u\n", "Mount",
 236                         osb->s_mount_opt, osb->s_atime_quantum);
 237
 238         if (cconn) {
 239                 out += snprintf(buf + out, len - out,
 240                                 "%10s => Stack: %s  Name: %*s  "
 241                                 "Version: %d.%d\n", "Cluster",
 242                                 (*osb->osb_cluster_stack == '\0' ?
 243                                  "o2cb" : osb->osb_cluster_stack),
 244                                 cconn->cc_namelen, cconn->cc_name,
 245                                 cconn->cc_version.pv_major,
 246                                 cconn->cc_version.pv_minor);
 247         }
 248
 249         spin_lock(&osb->dc_task_lock);
 250         out += snprintf(buf + out, len - out,
 251                         "%10s => Pid: %d  Count: %lu  WakeSeq: %lu  "
 252                         "WorkSeq: %lu\n", "DownCnvt",
 253                         (osb->dc_task ?  task_pid_nr(osb->dc_task) : -1),
 254                         osb->blocked_lock_count, osb->dc_wake_sequence,
 255                         osb->dc_work_sequence);
 256         spin_unlock(&osb->dc_task_lock);
 257
 258         spin_lock(&osb->osb_lock);
 259         out += snprintf(buf + out, len - out, "%10s => Pid: %d  Nodes:",
 260                         "Recovery",
 261                         (osb->recovery_thread_task ?
 262                          task_pid_nr(osb->recovery_thread_task) : -1));
 263         if (rm->rm_used == 0)
 264                 out += snprintf(buf + out, len - out, " None\n");
 265         else {
 266                 for (i = 0; i < rm->rm_used; i++)
 267                         out += snprintf(buf + out, len - out, " %d",
 268                                         rm->rm_entries[i]);
 269                 out += snprintf(buf + out, len - out, "\n");
 270         }
 271         spin_unlock(&osb->osb_lock);
 272
 273         out += snprintf(buf + out, len - out,
 274                         "%10s => Pid: %d  Interval: %lu  Needs: %d\n", "Commit",
 275                         (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
 276                         osb->osb_commit_interval,
 277                         atomic_read(&osb->needs_checkpoint));
 278
 279         out += snprintf(buf + out, len - out,
 280                         "%10s => State: %d  TxnId: %lu  NumTxns: %d\n",
 281                         "Journal", osb->journal->j_state,
 282                         osb->journal->j_trans_id,
 283                         atomic_read(&osb->journal->j_num_trans));
 284
 285         out += snprintf(buf + out, len - out,
 286                         "%10s => GlobalAllocs: %d  LocalAllocs: %d  "
 287                         "SubAllocs: %d  LAWinMoves: %d  SAExtends: %d\n",
 288                         "Stats",
 289                         atomic_read(&osb->alloc_stats.bitmap_data),
 290                         atomic_read(&osb->alloc_stats.local_data),
 291                         atomic_read(&osb->alloc_stats.bg_allocs),
 292                         atomic_read(&osb->alloc_stats.moves),
 293                         atomic_read(&osb->alloc_stats.bg_extends));
 294
 295         out += snprintf(buf + out, len - out,
 296                         "%10s => State: %u  Descriptor: %llu  Size: %u bits  "
 297                         "Default: %u bits\n",
 298                         "LocalAlloc", osb->local_alloc_state,
 299                         (unsigned long long)osb->la_last_gd,
 300                         osb->local_alloc_bits, osb->local_alloc_default_bits);
 301
 302         spin_lock(&osb->osb_lock);
 303         out += snprintf(buf + out, len - out,
 304                         "%10s => Slot: %d  NumStolen: %d\n", "Steal",
 305                         osb->s_inode_steal_slot,
 306                         atomic_read(&osb->s_num_inodes_stolen));
 307         spin_unlock(&osb->osb_lock);
 308
 309         out += snprintf(buf + out, len - out, "OrphanScan => ");
 310         out += snprintf(buf + out, len - out, "Local: %u  Global: %u ",
 311                         os->os_count, os->os_seqno);
 312         out += snprintf(buf + out, len - out, " Last Scan: ");
 313         if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
 314                 out += snprintf(buf + out, len - out, "Disabled\n");
 315         else
 316                 out += snprintf(buf + out, len - out, "%lu seconds ago\n",
 317                                 (get_seconds() - os->os_scantime.tv_sec));
 318
 319         out += snprintf(buf + out, len - out, "%10s => %3s  %10s\n",
 320                         "Slots", "Num", "RecoGen");
 321         for (i = 0; i < osb->max_slots; ++i) {
 322                 out += snprintf(buf + out, len - out,
 323                                 "%10s  %c %3d  %10d\n",
 324                                 " ",
 325                                 (i == osb->slot_num ? '*' : ' '),
 326                                 i, osb->slot_recovery_generations[i]);
 327         }
 328
 329         return out;
 330 }
 331
 332 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 333 {
 334         struct ocfs2_super *osb = inode->i_private;
 335         char *buf = NULL;
 336
 337         buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
 338         if (!buf)
 339                 goto bail;
 340
 341         i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
 342
 343         file->private_data = buf;
 344
 345         return 0;
 346 bail:
 347         return -ENOMEM;
 348 }
 349
 350 static int ocfs2_debug_release(struct inode *inode, struct file *file)
 351 {
 352         kfree(file->private_data);
 353         return 0;
 354 }
 355
 356 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 357                                 size_t nbytes, loff_t *ppos)
 358 {
 359         return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
 360                                        i_size_read(file->f_mapping->host));
 361 }
 362 #else
 363 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 364 {
 365         return 0;
 366 }
 367 static int ocfs2_debug_release(struct inode *inode, struct file *file)
 368 {
 369         return 0;
 370 }
 371 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 372                                 size_t nbytes, loff_t *ppos)
 373 {
 374         return 0;
 375 }
 376 #endif  /* CONFIG_DEBUG_FS */
 377
 378 static const struct file_operations ocfs2_osb_debug_fops = {
 379         .open =         ocfs2_osb_debug_open,
 380         .release =      ocfs2_debug_release,
 381         .read =         ocfs2_debug_read,
 382         .llseek =       generic_file_llseek,
 383 };
 384
 385 static int ocfs2_sync_fs(struct super_block *sb, int wait)
 386 {
 387         int status;
 388         tid_t target;
 389         struct ocfs2_super *osb = OCFS2_SB(sb);
 390
 391         if (ocfs2_is_hard_readonly(osb))
 392                 return -EROFS;
 393
 394         if (wait) {
 395                 status = ocfs2_flush_truncate_log(osb);
 396                 if (status < 0)
 397                         mlog_errno(status);
 398         } else {
 399                 ocfs2_schedule_truncate_log_flush(osb, 0);
 400         }
 401
 402         if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
 403                                       &target)) {
 404                 if (wait)
 405                         jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
 406                                              target);
 407         }
 408         return 0;
 409 }
 410
 411 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
 412 {
 413         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
 414             && (ino == USER_QUOTA_SYSTEM_INODE
 415                 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
 416                 return 0;
 417         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
 418             && (ino == GROUP_QUOTA_SYSTEM_INODE
 419                 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
 420                 return 0;
 421         return 1;
 422 }
 423
 424 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
 425 {
 426         struct inode *new = NULL;
 427         int status = 0;
 428         int i;
 429
 430         mlog_entry_void();
 431
 432         new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 433         if (IS_ERR(new)) {
 434                 status = PTR_ERR(new);
 435                 mlog_errno(status);
 436                 goto bail;
 437         }
 438         osb->root_inode = new;
 439
 440         new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 441         if (IS_ERR(new)) {
 442                 status = PTR_ERR(new);
 443                 mlog_errno(status);
 444                 goto bail;
 445         }
 446         osb->sys_root_inode = new;
 447
 448         for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
 449              i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
 450                 if (!ocfs2_need_system_inode(osb, i))
 451                         continue;
 452                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 453                 if (!new) {
 454                         ocfs2_release_system_inodes(osb);
 455                         status = -EINVAL;
 456                         mlog_errno(status);
 457                         /* FIXME: Should ERROR_RO_FS */
 458                         mlog(ML_ERROR, "Unable to load system inode %d, "
 459                              "possibly corrupt fs?", i);
 460                         goto bail;
 461                 }
 462                 // the array now has one ref, so drop this one
 463                 iput(new);
 464         }
 465
 466 bail:
 467         mlog_exit(status);
 468         return status;
 469 }
 470
 471 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
 472 {
 473         struct inode *new = NULL;
 474         int status = 0;
 475         int i;
 476
 477         mlog_entry_void();
 478
 479         for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
 480              i < NUM_SYSTEM_INODES;
 481              i++) {
 482                 if (!ocfs2_need_system_inode(osb, i))
 483                         continue;
 484                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 485                 if (!new) {
 486                         ocfs2_release_system_inodes(osb);
 487                         status = -EINVAL;
 488                         mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
 489                              status, i, osb->slot_num);
 490                         goto bail;
 491                 }
 492                 /* the array now has one ref, so drop this one */
 493                 iput(new);
 494         }
 495
 496 bail:
 497         mlog_exit(status);
 498         return status;
 499 }
 500
 501 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
 502 {
 503         int i;
 504         struct inode *inode;
 505
 506         mlog_entry_void();
 507
 508         for (i = 0; i < NUM_SYSTEM_INODES; i++) {
 509                 inode = osb->system_inodes[i];
 510                 if (inode) {
 511                         iput(inode);
 512                         osb->system_inodes[i] = NULL;
 513                 }
 514         }
 515
 516         inode = osb->sys_root_inode;
 517         if (inode) {
 518                 iput(inode);
 519                 osb->sys_root_inode = NULL;
 520         }
 521
 522         inode = osb->root_inode;
 523         if (inode) {
 524                 iput(inode);
 525                 osb->root_inode = NULL;
 526         }
 527
 528         mlog_exit(0);
 529 }
 530
 531 /* We're allocating fs objects, use GFP_NOFS */
 532 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
 533 {
 534         struct ocfs2_inode_info *oi;
 535
 536         oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
 537         if (!oi)
 538                 return NULL;
 539
 540         jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
 541         return &oi->vfs_inode;
 542 }
 543
 544 static void ocfs2_destroy_inode(struct inode *inode)
 545 {
 546         kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
 547 }
 548
 549 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
 550                                                 unsigned int cbits)
 551 {
 552         unsigned int bytes = 1 << cbits;
 553         unsigned int trim = bytes;
 554         unsigned int bitshift = 32;
 555
 556         /*
 557          * i_size and all block offsets in ocfs2 are always 64 bits
 558          * wide. i_clusters is 32 bits, in cluster-sized units. So on
 559          * 64 bit platforms, cluster size will be the limiting factor.
 560          */
 561
 562 #if BITS_PER_LONG == 32
 563 # if defined(CONFIG_LBDAF)
 564         BUILD_BUG_ON(sizeof(sector_t) != 8);
 565         /*
 566          * We might be limited by page cache size.
 567          */
 568         if (bytes > PAGE_CACHE_SIZE) {
 569                 bytes = PAGE_CACHE_SIZE;
 570                 trim = 1;
 571                 /*
 572                  * Shift by 31 here so that we don't get larger than
 573                  * MAX_LFS_FILESIZE
 574                  */
 575                 bitshift = 31;
 576         }
 577 # else
 578         /*
 579          * We are limited by the size of sector_t. Use block size, as
 580          * that's what we expose to the VFS.
 581          */
 582         bytes = 1 << bbits;
 583         trim = 1;
 584         bitshift = 31;
 585 # endif
 586 #endif
 587
 588         /*
 589          * Trim by a whole cluster when we can actually approach the
 590          * on-disk limits. Otherwise we can overflow i_clusters when
 591          * an extent start is at the max offset.
 592          */
 593         return (((unsigned long long)bytes) << bitshift) - trim;
 594 }
 595
 596 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
 597 {
 598         int incompat_features;
 599         int ret = 0;
 600         struct mount_options parsed_options;
 601         struct ocfs2_super *osb = OCFS2_SB(sb);
 602
 603         lock_kernel();
 604
 605         if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
 606             !ocfs2_check_set_options(sb, &parsed_options)) {
 607                 ret = -EINVAL;
 608                 goto out;
 609         }
 610
 611         if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) !=
 612             (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
 613                 ret = -EINVAL;
 614                 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
 615                 goto out;
 616         }
 617
 618         if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
 619             (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
 620                 ret = -EINVAL;
 621                 mlog(ML_ERROR, "Cannot change data mode on remount\n");
 622                 goto out;
 623         }
 624
 625         /* Probably don't want this on remount; it might
 626          * mess with other nodes */
 627         if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
 628             (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
 629                 ret = -EINVAL;
 630                 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
 631                 goto out;
 632         }
 633
 634         /* We're going to/from readonly mode. */
 635         if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
 636                 /* Disable quota accounting before remounting RO */
 637                 if (*flags & MS_RDONLY) {
 638                         ret = ocfs2_susp_quotas(osb, 0);
 639                         if (ret < 0)
 640                                 goto out;
 641                 }
 642                 /* Lock here so the check of HARD_RO and the potential
 643                  * setting of SOFT_RO is atomic. */
 644                 spin_lock(&osb->osb_lock);
 645                 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
 646                         mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
 647                         ret = -EROFS;
 648                         goto unlock_osb;
 649                 }
 650
 651                 if (*flags & MS_RDONLY) {
 652                         mlog(0, "Going to ro mode.\n");
 653                         sb->s_flags |= MS_RDONLY;
 654                         osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 655                 } else {
 656                         mlog(0, "Making ro filesystem writeable.\n");
 657
 658                         if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
 659                                 mlog(ML_ERROR, "Cannot remount RDWR "
 660                                      "filesystem due to previous errors.\n");
 661                                 ret = -EROFS;
 662                                 goto unlock_osb;
 663                         }
 664                         incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
 665                         if (incompat_features) {
 666                                 mlog(ML_ERROR, "Cannot remount RDWR because "
 667                                      "of unsupported optional features "
 668                                      "(%x).\n", incompat_features);
 669                                 ret = -EINVAL;
 670                                 goto unlock_osb;
 671                         }
 672                         sb->s_flags &= ~MS_RDONLY;
 673                         osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
 674                 }
 675 unlock_osb:
 676                 spin_unlock(&osb->osb_lock);
 677                 /* Enable quota accounting after remounting RW */
 678                 if (!ret && !(*flags & MS_RDONLY)) {
 679                         if (sb_any_quota_suspended(sb))
 680                                 ret = ocfs2_susp_quotas(osb, 1);
 681                         else
 682                                 ret = ocfs2_enable_quotas(osb);
 683                         if (ret < 0) {
 684                                 /* Return back changes... */
 685                                 spin_lock(&osb->osb_lock);
 686                                 sb->s_flags |= MS_RDONLY;
 687                                 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 688                                 spin_unlock(&osb->osb_lock);
 689                                 goto out;
 690                         }
 691                 }
 692         }
 693
 694         if (!ret) {
 695                 /* Only save off the new mount options in case of a successful
 696                  * remount. */
 697                 osb->s_mount_opt = parsed_options.mount_opt;
 698                 osb->s_atime_quantum = parsed_options.atime_quantum;
 699                 osb->preferred_slot = parsed_options.slot;
 700                 if (parsed_options.commit_interval)
 701                         osb->osb_commit_interval = parsed_options.commit_interval;
 702
 703                 if (!ocfs2_is_hard_readonly(osb))
 704                         ocfs2_set_journal_params(osb);
 705         }
 706 out:
 707         unlock_kernel();
 708         return ret;
 709 }
 710
 711 static int ocfs2_sb_probe(struct super_block *sb,
 712                           struct buffer_head **bh,
 713                           int *sector_size,
 714                           struct ocfs2_blockcheck_stats *stats)
 715 {
 716         int status, tmpstat;
 717         struct ocfs1_vol_disk_hdr *hdr;
 718         struct ocfs2_dinode *di;
 719         int blksize;
 720
 721         *bh = NULL;
 722
 723         /* may be > 512 */
 724         *sector_size = bdev_logical_block_size(sb->s_bdev);
 725         if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
 726                 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
 727                      *sector_size, OCFS2_MAX_BLOCKSIZE);
 728                 status = -EINVAL;
 729                 goto bail;
 730         }
 731
 732         /* Can this really happen? */
 733         if (*sector_size < OCFS2_MIN_BLOCKSIZE)
 734                 *sector_size = OCFS2_MIN_BLOCKSIZE;
 735
 736         /* check block zero for old format */
 737         status = ocfs2_get_sector(sb, bh, 0, *sector_size);
 738         if (status < 0) {
 739                 mlog_errno(status);
 740                 goto bail;
 741         }
 742         hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
 743         if (hdr->major_version == OCFS1_MAJOR_VERSION) {
 744                 mlog(ML_ERROR, "incompatible version: %u.%u\n",
 745                      hdr->major_version, hdr->minor_version);
 746                 status = -EINVAL;
 747         }
 748         if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
 749                    strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
 750                 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
 751                      hdr->signature);
 752                 status = -EINVAL;
 753         }
 754         brelse(*bh);
 755         *bh = NULL;
 756         if (status < 0) {
 757                 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
 758                      "upgraded before mounting with ocfs v2\n");
 759                 goto bail;
 760         }
 761
 762         /*
 763          * Now check at magic offset for 512, 1024, 2048, 4096
 764          * blocksizes.  4096 is the maximum blocksize because it is
 765          * the minimum clustersize.
 766          */
 767         status = -EINVAL;
 768         for (blksize = *sector_size;
 769              blksize <= OCFS2_MAX_BLOCKSIZE;
 770              blksize <<= 1) {
 771                 tmpstat = ocfs2_get_sector(sb, bh,
 772                                            OCFS2_SUPER_BLOCK_BLKNO,
 773                                            blksize);
 774                 if (tmpstat < 0) {
 775                         status = tmpstat;
 776                         mlog_errno(status);
 777                         break;
 778                 }
 779                 di = (struct ocfs2_dinode *) (*bh)->b_data;
 780                 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
 781                 spin_lock_init(&stats->b_lock);
 782                 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
 783                 if (tmpstat < 0) {
 784                         brelse(*bh);
 785                         *bh = NULL;
 786                 }
 787                 if (tmpstat != -EAGAIN) {
 788                         status = tmpstat;
 789                         break;
 790                 }
 791         }
 792
 793 bail:
 794         return status;
 795 }
 796
 797 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
 798 {
 799         if (ocfs2_mount_local(osb)) {
 800                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
 801                         mlog(ML_ERROR, "Cannot heartbeat on a locally "
 802                              "mounted device.\n");
 803                         return -EINVAL;
 804                 }
 805         }
 806
 807         if (ocfs2_userspace_stack(osb)) {
 808                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
 809                         mlog(ML_ERROR, "Userspace stack expected, but "
 810                              "o2cb heartbeat arguments passed to mount\n");
 811                         return -EINVAL;
 812                 }
 813         }
 814
 815         if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
 816                 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
 817                     !ocfs2_userspace_stack(osb)) {
 818                         mlog(ML_ERROR, "Heartbeat has to be started to mount "
 819                              "a read-write clustered device.\n");
 820                         return -EINVAL;
 821                 }
 822         }
 823
 824         return 0;
 825 }
 826
 827 /*
 828  * If we're using a userspace stack, mount should have passed
 829  * a name that matches the disk.  If not, mount should not
 830  * have passed a stack.
 831  */
 832 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
 833                                         struct mount_options *mopt)
 834 {
 835         if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
 836                 mlog(ML_ERROR,
 837                      "cluster stack passed to mount, but this filesystem "
 838                      "does not support it\n");
 839                 return -EINVAL;
 840         }
 841
 842         if (ocfs2_userspace_stack(osb) &&
 843             strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
 844                     OCFS2_STACK_LABEL_LEN)) {
 845                 mlog(ML_ERROR,
 846                      "cluster stack passed to mount (\"%s\") does not "
 847                      "match the filesystem (\"%s\")\n",
 848                      mopt->cluster_stack,
 849                      osb->osb_cluster_stack);
 850                 return -EINVAL;
 851         }
 852
 853         return 0;
 854 }
 855
 856 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
 857 {
 858         int type;
 859         struct super_block *sb = osb->sb;
 860         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 861                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 862         int status = 0;
 863
 864         for (type = 0; type < MAXQUOTAS; type++) {
 865                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 866                         continue;
 867                 if (unsuspend)
 868                         status = vfs_quota_enable(
 869                                         sb_dqopt(sb)->files[type],
 870                                         type, QFMT_OCFS2,
 871                                         DQUOT_SUSPENDED);
 872                 else
 873                         status = vfs_quota_disable(sb, type,
 874                                                    DQUOT_SUSPENDED);
 875                 if (status < 0)
 876                         break;
 877         }
 878         if (status < 0)
 879                 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
 880                      "remount (error = %d).\n", status);
 881         return status;
 882 }
 883
 884 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
 885 {
 886         struct inode *inode[MAXQUOTAS] = { NULL, NULL };
 887         struct super_block *sb = osb->sb;
 888         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 889                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 890         unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
 891                                         LOCAL_GROUP_QUOTA_SYSTEM_INODE };
 892         int status;
 893         int type;
 894
 895         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
 896         for (type = 0; type < MAXQUOTAS; type++) {
 897                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 898                         continue;
 899                 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
 900                                                         osb->slot_num);
 901                 if (!inode[type]) {
 902                         status = -ENOENT;
 903                         goto out_quota_off;
 904                 }
 905                 status = vfs_quota_enable(inode[type], type, QFMT_OCFS2,
 906                                                 DQUOT_USAGE_ENABLED);
 907                 if (status < 0)
 908                         goto out_quota_off;
 909         }
 910
 911         for (type = 0; type < MAXQUOTAS; type++)
 912                 iput(inode[type]);
 913         return 0;
 914 out_quota_off:
 915         ocfs2_disable_quotas(osb);
 916         for (type = 0; type < MAXQUOTAS; type++)
 917                 iput(inode[type]);
 918         mlog_errno(status);
 919         return status;
 920 }
 921
 922 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
 923 {
 924         int type;
 925         struct inode *inode;
 926         struct super_block *sb = osb->sb;
 927
 928         /* We mostly ignore errors in this function because there's not much
 929          * we can do when we see them */
 930         for (type = 0; type < MAXQUOTAS; type++) {
 931                 if (!sb_has_quota_loaded(sb, type))
 932                         continue;
 933                 inode = igrab(sb->s_dquot.files[type]);
 934                 /* Turn off quotas. This will remove all dquot structures from
 935                  * memory and so they will be automatically synced to global
 936                  * quota files */
 937                 vfs_quota_disable(sb, type, DQUOT_USAGE_ENABLED |
 938                                             DQUOT_LIMITS_ENABLED);
 939                 if (!inode)
 940                         continue;
 941                 iput(inode);
 942         }
 943 }
 944
 945 /* Handle quota on quotactl */
 946 static int ocfs2_quota_on(struct super_block *sb, int type, int format_id,
 947                           char *path, int remount)
 948 {
 949         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 950                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 951
 952         if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 953                 return -EINVAL;
 954
 955         if (remount)
 956                 return 0;       /* Just ignore it has been handled in
 957                                  * ocfs2_remount() */
 958         return vfs_quota_enable(sb_dqopt(sb)->files[type], type,
 959                                     format_id, DQUOT_LIMITS_ENABLED);
 960 }
 961
 962 /* Handle quota off quotactl */
 963 static int ocfs2_quota_off(struct super_block *sb, int type, int remount)
 964 {
 965         if (remount)
 966                 return 0;       /* Ignore now and handle later in
 967                                  * ocfs2_remount() */
 968         return vfs_quota_disable(sb, type, DQUOT_LIMITS_ENABLED);
 969 }
 970
 971 static const struct quotactl_ops ocfs2_quotactl_ops = {
 972         .quota_on       = ocfs2_quota_on,
 973         .quota_off      = ocfs2_quota_off,
 974         .quota_sync     = vfs_quota_sync,
 975         .get_info       = vfs_get_dqinfo,
 976         .set_info       = vfs_set_dqinfo,
 977         .get_dqblk      = vfs_get_dqblk,
 978         .set_dqblk      = vfs_set_dqblk,
 979 };
 980
 981 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
 982 {
 983         struct dentry *root;
 984         int status, sector_size;
 985         struct mount_options parsed_options;
 986         struct inode *inode = NULL;
 987         struct ocfs2_super *osb = NULL;
 988         struct buffer_head *bh = NULL;
 989         char nodestr[8];
 990         struct ocfs2_blockcheck_stats stats;
 991
 992         mlog_entry("%p, %p, %i", sb, data, silent);
 993
 994         if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
 995                 status = -EINVAL;
 996                 goto read_super_error;
 997         }
 998
 999         /* probe for superblock */
1000         status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
1001         if (status < 0) {
1002                 mlog(ML_ERROR, "superblock probe failed!\n");
1003                 goto read_super_error;
1004         }
1005
1006         status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1007         osb = OCFS2_SB(sb);
1008         if (status < 0) {
1009                 mlog_errno(status);
1010                 goto read_super_error;
1011         }
1012         brelse(bh);
1013         bh = NULL;
1014
1015         if (!ocfs2_check_set_options(sb, &parsed_options)) {
1016                 status = -EINVAL;
1017                 goto read_super_error;
1018         }
1019         osb->s_mount_opt = parsed_options.mount_opt;
1020         osb->s_atime_quantum = parsed_options.atime_quantum;
1021         osb->preferred_slot = parsed_options.slot;
1022         osb->osb_commit_interval = parsed_options.commit_interval;
1023         osb->local_alloc_default_bits = ocfs2_megabytes_to_clusters(sb, parsed_options.localalloc_opt);
1024         osb->local_alloc_bits = osb->local_alloc_default_bits;
1025
1026         status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1027         if (status)
1028                 goto read_super_error;
1029
1030         sb->s_magic = OCFS2_SUPER_MAGIC;
1031
1032         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1033                 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1034
1035         /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1036          * heartbeat=none */
1037         if (bdev_read_only(sb->s_bdev)) {
1038                 if (!(sb->s_flags & MS_RDONLY)) {
1039                         status = -EACCES;
1040                         mlog(ML_ERROR, "Readonly device detected but readonly "
1041                              "mount was not specified.\n");
1042                         goto read_super_error;
1043                 }
1044
1045                 /* You should not be able to start a local heartbeat
1046                  * on a readonly device. */
1047                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1048                         status = -EROFS;
1049                         mlog(ML_ERROR, "Local heartbeat specified on readonly "
1050                              "device.\n");
1051                         goto read_super_error;
1052                 }
1053
1054                 status = ocfs2_check_journals_nolocks(osb);
1055                 if (status < 0) {
1056                         if (status == -EROFS)
1057                                 mlog(ML_ERROR, "Recovery required on readonly "
1058                                      "file system, but write access is "
1059                                      "unavailable.\n");
1060                         else
1061                                 mlog_errno(status);
1062                         goto read_super_error;
1063                 }
1064
1065                 ocfs2_set_ro_flag(osb, 1);
1066
1067                 printk(KERN_NOTICE "Readonly device detected. No cluster "
1068                        "services will be utilized for this mount. Recovery "
1069                        "will be skipped.\n");
1070         }
1071
1072         if (!ocfs2_is_hard_readonly(osb)) {
1073                 if (sb->s_flags & MS_RDONLY)
1074                         ocfs2_set_ro_flag(osb, 0);
1075         }
1076
1077         status = ocfs2_verify_heartbeat(osb);
1078         if (status < 0) {
1079                 mlog_errno(status);
1080                 goto read_super_error;
1081         }
1082
1083         osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1084                                                  ocfs2_debugfs_root);
1085         if (!osb->osb_debug_root) {
1086                 status = -EINVAL;
1087                 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1088                 goto read_super_error;
1089         }
1090
1091         osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1092                                             osb->osb_debug_root,
1093                                             osb,
1094                                             &ocfs2_osb_debug_fops);
1095         if (!osb->osb_ctxt) {
1096                 status = -EINVAL;
1097                 mlog_errno(status);
1098                 goto read_super_error;
1099         }
1100
1101         if (ocfs2_meta_ecc(osb)) {
1102                 status = ocfs2_blockcheck_stats_debugfs_install(
1103                                                 &osb->osb_ecc_stats,
1104                                                 osb->osb_debug_root);
1105                 if (status) {
1106                         mlog(ML_ERROR,
1107                              "Unable to create blockcheck statistics "
1108                              "files\n");
1109                         goto read_super_error;
1110                 }
1111         }
1112
1113         status = ocfs2_mount_volume(sb);
1114         if (osb->root_inode)
1115                 inode = igrab(osb->root_inode);
1116
1117         if (status < 0)
1118                 goto read_super_error;
1119
1120         if (!inode) {
1121                 status = -EIO;
1122                 mlog_errno(status);
1123                 goto read_super_error;
1124         }
1125
1126         root = d_alloc_root(inode);
1127         if (!root) {
1128                 status = -ENOMEM;
1129                 mlog_errno(status);
1130                 goto read_super_error;
1131         }
1132
1133         sb->s_root = root;
1134
1135         ocfs2_complete_mount_recovery(osb);
1136
1137         if (ocfs2_mount_local(osb))
1138                 snprintf(nodestr, sizeof(nodestr), "local");
1139         else
1140                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1141
1142         printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1143                "with %s data mode.\n",
1144                osb->dev_str, nodestr, osb->slot_num,
1145                osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1146                "ordered");
1147
1148         atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1149         wake_up(&osb->osb_mount_event);
1150
1151         /* Now we can initialize quotas because we can afford to wait
1152          * for cluster locks recovery now. That also means that truncation
1153          * log recovery can happen but that waits for proper quota setup */
1154         if (!(sb->s_flags & MS_RDONLY)) {
1155                 status = ocfs2_enable_quotas(osb);
1156                 if (status < 0) {
1157                         /* We have to err-out specially here because
1158                          * s_root is already set */
1159                         mlog_errno(status);
1160                         atomic_set(&osb->vol_state, VOLUME_DISABLED);
1161                         wake_up(&osb->osb_mount_event);
1162                         mlog_exit(status);
1163                         return status;
1164                 }
1165         }
1166
1167         ocfs2_complete_quota_recovery(osb);
1168
1169         /* Now we wake up again for processes waiting for quotas */
1170         atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1171         wake_up(&osb->osb_mount_event);
1172
1173         /* Start this when the mount is almost sure of being successful */
1174         ocfs2_orphan_scan_start(osb);
1175
1176         mlog_exit(status);
1177         return status;
1178
1179 read_super_error:
1180         brelse(bh);
1181
1182         if (inode)
1183                 iput(inode);
1184
1185         if (osb) {
1186                 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1187                 wake_up(&osb->osb_mount_event);
1188                 ocfs2_dismount_volume(sb, 1);
1189         }
1190
1191         mlog_exit(status);
1192         return status;
1193 }
1194
1195 static int ocfs2_get_sb(struct file_system_type *fs_type,
1196                         int flags,
1197                         const char *dev_name,
1198                         void *data,
1199                         struct vfsmount *mnt)
1200 {
1201         return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super,
1202                            mnt);
1203 }
1204
1205 static void ocfs2_kill_sb(struct super_block *sb)
1206 {
1207         struct ocfs2_super *osb = OCFS2_SB(sb);
1208
1209         /* Failed mount? */
1210         if (!osb || atomic_read(&osb->vol_state) == VOLUME_DISABLED)
1211                 goto out;
1212
1213         /* Prevent further queueing of inode drop events */
1214         spin_lock(&dentry_list_lock);
1215         ocfs2_set_osb_flag(osb, OCFS2_OSB_DROP_DENTRY_LOCK_IMMED);
1216         spin_unlock(&dentry_list_lock);
1217         /* Wait for work to finish and/or remove it */
1218         cancel_work_sync(&osb->dentry_lock_work);
1219 out:
1220         kill_block_super(sb);
1221 }
1222
1223 static struct file_system_type ocfs2_fs_type = {
1224         .owner          = THIS_MODULE,
1225         .name           = "ocfs2",
1226         .get_sb         = ocfs2_get_sb, /* is this called when we mount
1227                                         * the fs? */
1228         .kill_sb        = ocfs2_kill_sb,
1229
1230         .fs_flags       = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1231         .next           = NULL
1232 };
1233
1234 static int ocfs2_check_set_options(struct super_block *sb,
1235                                    struct mount_options *options)
1236 {
1237         if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1238             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1239                                          OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1240                 mlog(ML_ERROR, "User quotas were requested, but this "
1241                      "filesystem does not have the feature enabled.\n");
1242                 return 0;
1243         }
1244         if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1245             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1246                                          OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1247                 mlog(ML_ERROR, "Group quotas were requested, but this "
1248                      "filesystem does not have the feature enabled.\n");
1249                 return 0;
1250         }
1251         if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1252             !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1253                 mlog(ML_ERROR, "ACL support requested but extended attributes "
1254                      "feature is not enabled\n");
1255                 return 0;
1256         }
1257         /* No ACL setting specified? Use XATTR feature... */
1258         if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1259                                     OCFS2_MOUNT_NO_POSIX_ACL))) {
1260                 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1261                         options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1262                 else
1263                         options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1264         }
1265         return 1;
1266 }
1267
1268 static int ocfs2_parse_options(struct super_block *sb,
1269                                char *options,
1270                                struct mount_options *mopt,
1271                                int is_remount)
1272 {
1273         int status;
1274         char *p;
1275
1276         mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
1277                    options ? options : "(none)");
1278
1279         mopt->commit_interval = 0;
1280         mopt->mount_opt = 0;
1281         mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1282         mopt->slot = OCFS2_INVALID_SLOT;
1283         mopt->localalloc_opt = OCFS2_DEFAULT_LOCAL_ALLOC_SIZE;
1284         mopt->cluster_stack[0] = '\0';
1285
1286         if (!options) {
1287                 status = 1;
1288                 goto bail;
1289         }
1290
1291         while ((p = strsep(&options, ",")) != NULL) {
1292                 int token, option;
1293                 substring_t args[MAX_OPT_ARGS];
1294
1295                 if (!*p)
1296                         continue;
1297
1298                 token = match_token(p, tokens, args);
1299                 switch (token) {
1300                 case Opt_hb_local:
1301                         mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1302                         break;
1303                 case Opt_hb_none:
1304                         mopt->mount_opt &= ~OCFS2_MOUNT_HB_LOCAL;
1305                         break;
1306                 case Opt_barrier:
1307                         if (match_int(&args[0], &option)) {
1308                                 status = 0;
1309                                 goto bail;
1310                         }
1311                         if (option)
1312                                 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1313                         else
1314                                 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1315                         break;
1316                 case Opt_intr:
1317                         mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1318                         break;
1319                 case Opt_nointr:
1320                         mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1321                         break;
1322                 case Opt_err_panic:
1323                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1324                         break;
1325                 case Opt_err_ro:
1326                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1327                         break;
1328                 case Opt_data_ordered:
1329                         mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1330                         break;
1331                 case Opt_data_writeback:
1332                         mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1333                         break;
1334                 case Opt_user_xattr:
1335                         mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1336                         break;
1337                 case Opt_nouser_xattr:
1338                         mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1339                         break;
1340                 case Opt_atime_quantum:
1341                         if (match_int(&args[0], &option)) {
1342                                 status = 0;
1343                                 goto bail;
1344                         }
1345                         if (option >= 0)
1346                                 mopt->atime_quantum = option;
1347                         break;
1348                 case Opt_slot:
1349                         option = 0;
1350                         if (match_int(&args[0], &option)) {
1351                                 status = 0;
1352                                 goto bail;
1353                         }
1354                         if (option)
1355                                 mopt->slot = (s16)option;
1356                         break;
1357                 case Opt_commit:
1358                         option = 0;
1359                         if (match_int(&args[0], &option)) {
1360                                 status = 0;
1361                                 goto bail;
1362                         }
1363                         if (option < 0)
1364                                 return 0;
1365                         if (option == 0)
1366                                 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1367                         mopt->commit_interval = HZ * option;
1368                         break;
1369                 case Opt_localalloc:
1370                         option = 0;
1371                         if (match_int(&args[0], &option)) {
1372                                 status = 0;
1373                                 goto bail;
1374                         }
1375                         if (option >= 0 && (option <= ocfs2_local_alloc_size(sb) * 8))
1376                                 mopt->localalloc_opt = option;
1377                         break;
1378                 case Opt_localflocks:
1379                         /*
1380                          * Changing this during remount could race
1381                          * flock() requests, or "unbalance" existing
1382                          * ones (e.g., a lock is taken in one mode but
1383                          * dropped in the other). If users care enough
1384                          * to flip locking modes during remount, we
1385                          * could add a "local" flag to individual
1386                          * flock structures for proper tracking of
1387                          * state.
1388                          */
1389                         if (!is_remount)
1390                                 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1391                         break;
1392                 case Opt_stack:
1393                         /* Check both that the option we were passed
1394                          * is of the right length and that it is a proper
1395                          * string of the right length.
1396                          */
1397                         if (((args[0].to - args[0].from) !=
1398                              OCFS2_STACK_LABEL_LEN) ||
1399                             (strnlen(args[0].from,
1400                                      OCFS2_STACK_LABEL_LEN) !=
1401                              OCFS2_STACK_LABEL_LEN)) {
1402                                 mlog(ML_ERROR,
1403                                      "Invalid cluster_stack option\n");
1404                                 status = 0;
1405                                 goto bail;
1406                         }
1407                         memcpy(mopt->cluster_stack, args[0].from,
1408                                OCFS2_STACK_LABEL_LEN);
1409                         mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1410                         break;
1411                 case Opt_inode64:
1412                         mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1413                         break;
1414                 case Opt_usrquota:
1415                         mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1416                         break;
1417                 case Opt_grpquota:
1418                         mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1419                         break;
1420                 case Opt_acl:
1421                         mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1422                         mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1423                         break;
1424                 case Opt_noacl:
1425                         mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1426                         mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1427                         break;
1428                 default:
1429                         mlog(ML_ERROR,
1430                              "Unrecognized mount option \"%s\" "
1431                              "or missing value\n", p);
1432                         status = 0;
1433                         goto bail;
1434                 }
1435         }
1436
1437         status = 1;
1438
1439 bail:
1440         mlog_exit(status);
1441         return status;
1442 }
1443
1444 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1445 {
1446         struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb);
1447         unsigned long opts = osb->s_mount_opt;
1448         unsigned int local_alloc_megs;
1449
1450         if (opts & OCFS2_MOUNT_HB_LOCAL)
1451                 seq_printf(s, ",_netdev,heartbeat=local");
1452         else
1453                 seq_printf(s, ",heartbeat=none");
1454
1455         if (opts & OCFS2_MOUNT_NOINTR)
1456                 seq_printf(s, ",nointr");
1457
1458         if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1459                 seq_printf(s, ",data=writeback");
1460         else
1461                 seq_printf(s, ",data=ordered");
1462
1463         if (opts & OCFS2_MOUNT_BARRIER)
1464                 seq_printf(s, ",barrier=1");
1465
1466         if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1467                 seq_printf(s, ",errors=panic");
1468         else
1469                 seq_printf(s, ",errors=remount-ro");
1470
1471         if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1472                 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1473
1474         if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM)
1475                 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1476
1477         if (osb->osb_commit_interval)
1478                 seq_printf(s, ",commit=%u",
1479                            (unsigned) (osb->osb_commit_interval / HZ));
1480
1481         local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1482         if (local_alloc_megs != OCFS2_DEFAULT_LOCAL_ALLOC_SIZE)
1483                 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1484
1485         if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1486                 seq_printf(s, ",localflocks,");
1487
1488         if (osb->osb_cluster_stack[0])
1489                 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1490                            osb->osb_cluster_stack);
1491         if (opts & OCFS2_MOUNT_USRQUOTA)
1492                 seq_printf(s, ",usrquota");
1493         if (opts & OCFS2_MOUNT_GRPQUOTA)
1494                 seq_printf(s, ",grpquota");
1495
1496         if (opts & OCFS2_MOUNT_NOUSERXATTR)
1497                 seq_printf(s, ",nouser_xattr");
1498         else
1499                 seq_printf(s, ",user_xattr");
1500
1501         if (opts & OCFS2_MOUNT_INODE64)
1502                 seq_printf(s, ",inode64");
1503
1504         if (opts & OCFS2_MOUNT_POSIX_ACL)
1505                 seq_printf(s, ",acl");
1506         else
1507                 seq_printf(s, ",noacl");
1508
1509         return 0;
1510 }
1511
1512 static int __init ocfs2_init(void)
1513 {
1514         int status;
1515
1516         mlog_entry_void();
1517
1518         ocfs2_print_version();
1519
1520         status = init_ocfs2_uptodate_cache();
1521         if (status < 0) {
1522                 mlog_errno(status);
1523                 goto leave;
1524         }
1525
1526         status = ocfs2_initialize_mem_caches();
1527         if (status < 0) {
1528                 mlog_errno(status);
1529                 goto leave;
1530         }
1531
1532         ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1533         if (!ocfs2_wq) {
1534                 status = -ENOMEM;
1535                 goto leave;
1536         }
1537
1538         ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1539         if (!ocfs2_debugfs_root) {
1540                 status = -EFAULT;
1541                 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1542         }
1543
1544         status = ocfs2_quota_setup();
1545         if (status)
1546                 goto leave;
1547
1548         ocfs2_set_locking_protocol();
1549
1550         status = register_quota_format(&ocfs2_quota_format);
1551 leave:
1552         if (status < 0) {
1553                 ocfs2_quota_shutdown();
1554                 ocfs2_free_mem_caches();
1555                 exit_ocfs2_uptodate_cache();
1556         }
1557
1558         mlog_exit(status);
1559
1560         if (status >= 0) {
1561                 return register_filesystem(&ocfs2_fs_type);
1562         } else
1563                 return -1;
1564 }
1565
1566 static void __exit ocfs2_exit(void)
1567 {
1568         mlog_entry_void();
1569
1570         ocfs2_quota_shutdown();
1571
1572         if (ocfs2_wq) {
1573                 flush_workqueue(ocfs2_wq);
1574                 destroy_workqueue(ocfs2_wq);
1575         }
1576
1577         unregister_quota_format(&ocfs2_quota_format);
1578
1579         debugfs_remove(ocfs2_debugfs_root);
1580
1581         ocfs2_free_mem_caches();
1582
1583         unregister_filesystem(&ocfs2_fs_type);
1584
1585         exit_ocfs2_uptodate_cache();
1586
1587         mlog_exit_void();
1588 }
1589
1590 static void ocfs2_put_super(struct super_block *sb)
1591 {
1592         mlog_entry("(0x%p)\n", sb);
1593
1594         lock_kernel();
1595
1596         ocfs2_sync_blockdev(sb);
1597         ocfs2_dismount_volume(sb, 0);
1598
1599         unlock_kernel();
1600
1601         mlog_exit_void();
1602 }
1603
1604 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1605 {
1606         struct ocfs2_super *osb;
1607         u32 numbits, freebits;
1608         int status;
1609         struct ocfs2_dinode *bm_lock;
1610         struct buffer_head *bh = NULL;
1611         struct inode *inode = NULL;
1612
1613         mlog_entry("(%p, %p)\n", dentry->d_sb, buf);
1614
1615         osb = OCFS2_SB(dentry->d_sb);
1616
1617         inode = ocfs2_get_system_file_inode(osb,
1618                                             GLOBAL_BITMAP_SYSTEM_INODE,
1619                                             OCFS2_INVALID_SLOT);
1620         if (!inode) {
1621                 mlog(ML_ERROR, "failed to get bitmap inode\n");
1622                 status = -EIO;
1623                 goto bail;
1624         }
1625
1626         status = ocfs2_inode_lock(inode, &bh, 0);
1627         if (status < 0) {
1628                 mlog_errno(status);
1629                 goto bail;
1630         }
1631
1632         bm_lock = (struct ocfs2_dinode *) bh->b_data;
1633
1634         numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1635         freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1636
1637         buf->f_type = OCFS2_SUPER_MAGIC;
1638         buf->f_bsize = dentry->d_sb->s_blocksize;
1639         buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1640         buf->f_blocks = ((sector_t) numbits) *
1641                         (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1642         buf->f_bfree = ((sector_t) freebits) *
1643                        (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1644         buf->f_bavail = buf->f_bfree;
1645         buf->f_files = numbits;
1646         buf->f_ffree = freebits;
1647
1648         brelse(bh);
1649
1650         ocfs2_inode_unlock(inode, 0);
1651         status = 0;
1652 bail:
1653         if (inode)
1654                 iput(inode);
1655
1656         mlog_exit(status);
1657
1658         return status;
1659 }
1660
1661 static void ocfs2_inode_init_once(void *data)
1662 {
1663         struct ocfs2_inode_info *oi = data;
1664
1665         oi->ip_flags = 0;
1666         oi->ip_open_count = 0;
1667         spin_lock_init(&oi->ip_lock);
1668         ocfs2_extent_map_init(&oi->vfs_inode);
1669         INIT_LIST_HEAD(&oi->ip_io_markers);
1670         oi->ip_dir_start_lookup = 0;
1671
1672         init_rwsem(&oi->ip_alloc_sem);
1673         init_rwsem(&oi->ip_xattr_sem);
1674         mutex_init(&oi->ip_io_mutex);
1675
1676         oi->ip_blkno = 0ULL;
1677         oi->ip_clusters = 0;
1678
1679         ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1680         ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1681         ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1682
1683         ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1684                                   &ocfs2_inode_caching_ops);
1685
1686         inode_init_once(&oi->vfs_inode);
1687 }
1688
1689 static int ocfs2_initialize_mem_caches(void)
1690 {
1691         ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1692                                        sizeof(struct ocfs2_inode_info),
1693                                        0,
1694                                        (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1695                                                 SLAB_MEM_SPREAD),
1696                                        ocfs2_inode_init_once);
1697         ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1698                                         sizeof(struct ocfs2_dquot),
1699                                         0,
1700                                         (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1701                                                 SLAB_MEM_SPREAD),
1702                                         NULL);
1703         ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1704                                         sizeof(struct ocfs2_quota_chunk),
1705                                         0,
1706                                         (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1707                                         NULL);
1708         if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1709             !ocfs2_qf_chunk_cachep) {
1710                 if (ocfs2_inode_cachep)
1711                         kmem_cache_destroy(ocfs2_inode_cachep);
1712                 if (ocfs2_dquot_cachep)
1713                         kmem_cache_destroy(ocfs2_dquot_cachep);
1714                 if (ocfs2_qf_chunk_cachep)
1715                         kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1716                 return -ENOMEM;
1717         }
1718
1719         return 0;
1720 }
1721
1722 static void ocfs2_free_mem_caches(void)
1723 {
1724         if (ocfs2_inode_cachep)
1725                 kmem_cache_destroy(ocfs2_inode_cachep);
1726         ocfs2_inode_cachep = NULL;
1727
1728         if (ocfs2_dquot_cachep)
1729                 kmem_cache_destroy(ocfs2_dquot_cachep);
1730         ocfs2_dquot_cachep = NULL;
1731
1732         if (ocfs2_qf_chunk_cachep)
1733                 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1734         ocfs2_qf_chunk_cachep = NULL;
1735 }
1736
1737 static int ocfs2_get_sector(struct super_block *sb,
1738                             struct buffer_head **bh,
1739                             int block,
1740                             int sect_size)
1741 {
1742         if (!sb_set_blocksize(sb, sect_size)) {
1743                 mlog(ML_ERROR, "unable to set blocksize\n");
1744                 return -EIO;
1745         }
1746
1747         *bh = sb_getblk(sb, block);
1748         if (!*bh) {
1749                 mlog_errno(-EIO);
1750                 return -EIO;
1751         }
1752         lock_buffer(*bh);
1753         if (!buffer_dirty(*bh))
1754                 clear_buffer_uptodate(*bh);
1755         unlock_buffer(*bh);
1756         ll_rw_block(READ, 1, bh);
1757         wait_on_buffer(*bh);
1758         if (!buffer_uptodate(*bh)) {
1759                 mlog_errno(-EIO);
1760                 brelse(*bh);
1761                 *bh = NULL;
1762                 return -EIO;
1763         }
1764
1765         return 0;
1766 }
1767
1768 static int ocfs2_mount_volume(struct super_block *sb)
1769 {
1770         int status = 0;
1771         int unlock_super = 0;
1772         struct ocfs2_super *osb = OCFS2_SB(sb);
1773
1774         mlog_entry_void();
1775
1776         if (ocfs2_is_hard_readonly(osb))
1777                 goto leave;
1778
1779         status = ocfs2_dlm_init(osb);
1780         if (status < 0) {
1781                 mlog_errno(status);
1782                 goto leave;
1783         }
1784
1785         status = ocfs2_super_lock(osb, 1);
1786         if (status < 0) {
1787                 mlog_errno(status);
1788                 goto leave;
1789         }
1790         unlock_super = 1;
1791
1792         /* This will load up the node map and add ourselves to it. */
1793         status = ocfs2_find_slot(osb);
1794         if (status < 0) {
1795                 mlog_errno(status);
1796                 goto leave;
1797         }
1798
1799         /* load all node-local system inodes */
1800         status = ocfs2_init_local_system_inodes(osb);
1801         if (status < 0) {
1802                 mlog_errno(status);
1803                 goto leave;
1804         }
1805
1806         status = ocfs2_check_volume(osb);
1807         if (status < 0) {
1808                 mlog_errno(status);
1809                 goto leave;
1810         }
1811
1812         status = ocfs2_truncate_log_init(osb);
1813         if (status < 0)
1814                 mlog_errno(status);
1815
1816 leave:
1817         if (unlock_super)
1818                 ocfs2_super_unlock(osb, 1);
1819
1820         mlog_exit(status);
1821         return status;
1822 }
1823
1824 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1825 {
1826         int tmp, hangup_needed = 0;
1827         struct ocfs2_super *osb = NULL;
1828         char nodestr[8];
1829
1830         mlog_entry("(0x%p)\n", sb);
1831
1832         BUG_ON(!sb);
1833         osb = OCFS2_SB(sb);
1834         BUG_ON(!osb);
1835
1836         debugfs_remove(osb->osb_ctxt);
1837
1838         /*
1839          * Flush inode dropping work queue so that deletes are
1840          * performed while the filesystem is still working
1841          */
1842         ocfs2_drop_all_dl_inodes(osb);
1843
1844         /* Orphan scan should be stopped as early as possible */
1845         ocfs2_orphan_scan_stop(osb);
1846
1847         ocfs2_disable_quotas(osb);
1848
1849         ocfs2_shutdown_local_alloc(osb);
1850
1851         ocfs2_truncate_log_shutdown(osb);
1852
1853         /* This will disable recovery and flush any recovery work. */
1854         ocfs2_recovery_exit(osb);
1855
1856         ocfs2_journal_shutdown(osb);
1857
1858         ocfs2_sync_blockdev(sb);
1859
1860         ocfs2_purge_refcount_trees(osb);
1861
1862         /* No cluster connection means we've failed during mount, so skip
1863          * all the steps which depended on that to complete. */
1864         if (osb->cconn) {
1865                 tmp = ocfs2_super_lock(osb, 1);
1866                 if (tmp < 0) {
1867                         mlog_errno(tmp);
1868                         return;
1869                 }
1870         }
1871
1872         if (osb->slot_num != OCFS2_INVALID_SLOT)
1873                 ocfs2_put_slot(osb);
1874
1875         if (osb->cconn)
1876                 ocfs2_super_unlock(osb, 1);
1877
1878         ocfs2_release_system_inodes(osb);
1879
1880         /*
1881          * If we're dismounting due to mount error, mount.ocfs2 will clean
1882          * up heartbeat.  If we're a local mount, there is no heartbeat.
1883          * If we failed before we got a uuid_str yet, we can't stop
1884          * heartbeat.  Otherwise, do it.
1885          */
1886         if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str)
1887                 hangup_needed = 1;
1888
1889         if (osb->cconn)
1890                 ocfs2_dlm_shutdown(osb, hangup_needed);
1891
1892         ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1893         debugfs_remove(osb->osb_debug_root);
1894
1895         if (hangup_needed)
1896                 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1897
1898         atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1899
1900         if (ocfs2_mount_local(osb))
1901                 snprintf(nodestr, sizeof(nodestr), "local");
1902         else
1903                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1904
1905         printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1906                osb->dev_str, nodestr);
1907
1908         ocfs2_delete_osb(osb);
1909         kfree(osb);
1910         sb->s_dev = 0;
1911         sb->s_fs_info = NULL;
1912 }
1913
1914 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1915                                 unsigned uuid_bytes)
1916 {
1917         int i, ret;
1918         char *ptr;
1919
1920         BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1921
1922         osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1923         if (osb->uuid_str == NULL)
1924                 return -ENOMEM;
1925
1926         for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1927                 /* print with null */
1928                 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1929                 if (ret != 2) /* drop super cleans up */
1930                         return -EINVAL;
1931                 /* then only advance past the last char */
1932                 ptr += 2;
1933         }
1934
1935         return 0;
1936 }
1937
1938 static int ocfs2_initialize_super(struct super_block *sb,
1939                                   struct buffer_head *bh,
1940                                   int sector_size,
1941                                   struct ocfs2_blockcheck_stats *stats)
1942 {
1943         int status;
1944         int i, cbits, bbits;
1945         struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
1946         struct inode *inode = NULL;
1947         struct ocfs2_journal *journal;
1948         __le32 uuid_net_key;
1949         struct ocfs2_super *osb;
1950
1951         mlog_entry_void();
1952
1953         osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
1954         if (!osb) {
1955                 status = -ENOMEM;
1956                 mlog_errno(status);
1957                 goto bail;
1958         }
1959
1960         sb->s_fs_info = osb;
1961         sb->s_op = &ocfs2_sops;
1962         sb->s_export_op = &ocfs2_export_ops;
1963         sb->s_qcop = &ocfs2_quotactl_ops;
1964         sb->dq_op = &ocfs2_quota_operations;
1965         sb->s_xattr = ocfs2_xattr_handlers;
1966         sb->s_time_gran = 1;
1967         sb->s_flags |= MS_NOATIME;
1968         /* this is needed to support O_LARGEFILE */
1969         cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
1970         bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
1971         sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
1972
1973         osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
1974
1975         for (i = 0; i < 3; i++)
1976                 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
1977         osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
1978
1979         osb->sb = sb;
1980         /* Save off for ocfs2_rw_direct */
1981         osb->s_sectsize_bits = blksize_bits(sector_size);
1982         BUG_ON(!osb->s_sectsize_bits);
1983
1984         spin_lock_init(&osb->dc_task_lock);
1985         init_waitqueue_head(&osb->dc_event);
1986         osb->dc_work_sequence = 0;
1987         osb->dc_wake_sequence = 0;
1988         INIT_LIST_HEAD(&osb->blocked_lock_list);
1989         osb->blocked_lock_count = 0;
1990         spin_lock_init(&osb->osb_lock);
1991         spin_lock_init(&osb->osb_xattr_lock);
1992         ocfs2_init_inode_steal_slot(osb);
1993
1994         atomic_set(&osb->alloc_stats.moves, 0);
1995         atomic_set(&osb->alloc_stats.local_data, 0);
1996         atomic_set(&osb->alloc_stats.bitmap_data, 0);
1997         atomic_set(&osb->alloc_stats.bg_allocs, 0);
1998         atomic_set(&osb->alloc_stats.bg_extends, 0);
1999
2000         /* Copy the blockcheck stats from the superblock probe */
2001         osb->osb_ecc_stats = *stats;
2002
2003         ocfs2_init_node_maps(osb);
2004
2005         snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2006                  MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2007
2008         ocfs2_orphan_scan_init(osb);
2009
2010         status = ocfs2_recovery_init(osb);
2011         if (status) {
2012                 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2013                 mlog_errno(status);
2014                 goto bail;
2015         }
2016
2017         init_waitqueue_head(&osb->checkpoint_event);
2018         atomic_set(&osb->needs_checkpoint, 0);
2019
2020         osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2021
2022         osb->slot_num = OCFS2_INVALID_SLOT;
2023
2024         osb->s_xattr_inline_size = le16_to_cpu(
2025                                         di->id2.i_super.s_xattr_inline_size);
2026
2027         osb->local_alloc_state = OCFS2_LA_UNUSED;
2028         osb->local_alloc_bh = NULL;
2029         INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2030
2031         init_waitqueue_head(&osb->osb_mount_event);
2032
2033         osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2034         if (!osb->vol_label) {
2035                 mlog(ML_ERROR, "unable to alloc vol label\n");
2036                 status = -ENOMEM;
2037                 goto bail;
2038         }
2039
2040         osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2041         if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2042                 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2043                      osb->max_slots);
2044                 status = -EINVAL;
2045                 goto bail;
2046         }
2047         mlog(0, "max_slots for this device: %u\n", osb->max_slots);
2048
2049         osb->slot_recovery_generations =
2050                 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2051                         GFP_KERNEL);
2052         if (!osb->slot_recovery_generations) {
2053                 status = -ENOMEM;
2054                 mlog_errno(status);
2055                 goto bail;
2056         }
2057
2058         init_waitqueue_head(&osb->osb_wipe_event);
2059         osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2060                                         sizeof(*osb->osb_orphan_wipes),
2061                                         GFP_KERNEL);
2062         if (!osb->osb_orphan_wipes) {
2063                 status = -ENOMEM;
2064                 mlog_errno(status);
2065                 goto bail;
2066         }
2067
2068         osb->osb_rf_lock_tree = RB_ROOT;
2069
2070         osb->s_feature_compat =
2071                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2072         osb->s_feature_ro_compat =
2073                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2074         osb->s_feature_incompat =
2075                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2076
2077         if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2078                 mlog(ML_ERROR, "couldn't mount because of unsupported "
2079                      "optional features (%x).\n", i);
2080                 status = -EINVAL;
2081                 goto bail;
2082         }
2083         if (!(osb->sb->s_flags & MS_RDONLY) &&
2084             (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2085                 mlog(ML_ERROR, "couldn't mount RDWR because of "
2086                      "unsupported optional features (%x).\n", i);
2087                 status = -EINVAL;
2088                 goto bail;
2089         }
2090
2091         if (ocfs2_userspace_stack(osb)) {
2092                 memcpy(osb->osb_cluster_stack,
2093                        OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2094                        OCFS2_STACK_LABEL_LEN);
2095                 osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
2096                 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2097                         mlog(ML_ERROR,
2098                              "couldn't mount because of an invalid "
2099                              "cluster stack label (%s) \n",
2100                              osb->osb_cluster_stack);
2101                         status = -EINVAL;
2102                         goto bail;
2103                 }
2104         } else {
2105                 /* The empty string is identical with classic tools that
2106                  * don't know about s_cluster_info. */
2107                 osb->osb_cluster_stack[0] = '\0';
2108         }
2109
2110         get_random_bytes(&osb->s_next_generation, sizeof(u32));
2111
2112         /* FIXME
2113          * This should be done in ocfs2_journal_init(), but unknown
2114          * ordering issues will cause the filesystem to crash.
2115          * If anyone wants to figure out what part of the code
2116          * refers to osb->journal before ocfs2_journal_init() is run,
2117          * be my guest.
2118          */
2119         /* initialize our journal structure */
2120
2121         journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2122         if (!journal) {
2123                 mlog(ML_ERROR, "unable to alloc journal\n");
2124                 status = -ENOMEM;
2125                 goto bail;
2126         }
2127         osb->journal = journal;
2128         journal->j_osb = osb;
2129
2130         atomic_set(&journal->j_num_trans, 0);
2131         init_rwsem(&journal->j_trans_barrier);
2132         init_waitqueue_head(&journal->j_checkpointed);
2133         spin_lock_init(&journal->j_lock);
2134         journal->j_trans_id = (unsigned long) 1;
2135         INIT_LIST_HEAD(&journal->j_la_cleanups);
2136         INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2137         journal->j_state = OCFS2_JOURNAL_FREE;
2138
2139         INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes);
2140         osb->dentry_lock_list = NULL;
2141
2142         /* get some pseudo constants for clustersize bits */
2143         osb->s_clustersize_bits =
2144                 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2145         osb->s_clustersize = 1 << osb->s_clustersize_bits;
2146         mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
2147
2148         if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2149             osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2150                 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2151                      osb->s_clustersize);
2152                 status = -EINVAL;
2153                 goto bail;
2154         }
2155
2156         if (ocfs2_clusters_to_blocks(osb->sb, le32_to_cpu(di->i_clusters) - 1)
2157             > (u32)~0UL) {
2158                 mlog(ML_ERROR, "Volume might try to write to blocks beyond "
2159                      "what jbd can address in 32 bits.\n");
2160                 status = -EINVAL;
2161                 goto bail;
2162         }
2163
2164         if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2165                                  sizeof(di->id2.i_super.s_uuid))) {
2166                 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2167                 status = -ENOMEM;
2168                 goto bail;
2169         }
2170
2171         memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
2172
2173         strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2174         osb->vol_label[63] = '\0';
2175         osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2176         osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2177         osb->first_cluster_group_blkno =
2178                 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2179         osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2180         osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2181         mlog(0, "vol_label: %s\n", osb->vol_label);
2182         mlog(0, "uuid: %s\n", osb->uuid_str);
2183         mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
2184              (unsigned long long)osb->root_blkno,
2185              (unsigned long long)osb->system_dir_blkno);
2186
2187         osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2188         if (!osb->osb_dlm_debug) {
2189                 status = -ENOMEM;
2190                 mlog_errno(status);
2191                 goto bail;
2192         }
2193
2194         atomic_set(&osb->vol_state, VOLUME_INIT);
2195
2196         /* load root, system_dir, and all global system inodes */
2197         status = ocfs2_init_global_system_inodes(osb);
2198         if (status < 0) {
2199                 mlog_errno(status);
2200                 goto bail;
2201         }
2202
2203         /*
2204          * global bitmap
2205          */
2206         inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2207                                             OCFS2_INVALID_SLOT);
2208         if (!inode) {
2209                 status = -EINVAL;
2210                 mlog_errno(status);
2211                 goto bail;
2212         }
2213
2214         osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2215         iput(inode);
2216
2217         osb->bitmap_cpg = ocfs2_group_bitmap_size(sb) * 8;
2218
2219         status = ocfs2_init_slot_info(osb);
2220         if (status < 0) {
2221                 mlog_errno(status);
2222                 goto bail;
2223         }
2224
2225 bail:
2226         mlog_exit(status);
2227         return status;
2228 }
2229
2230 /*
2231  * will return: -EAGAIN if it is ok to keep searching for superblocks
2232  *              -EINVAL if there is a bad superblock
2233  *              0 on success
2234  */
2235 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2236                                struct buffer_head *bh,
2237                                u32 blksz,
2238                                struct ocfs2_blockcheck_stats *stats)
2239 {
2240         int status = -EAGAIN;
2241
2242         mlog_entry_void();
2243
2244         if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2245                    strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2246                 /* We have to do a raw check of the feature here */
2247                 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2248                     OCFS2_FEATURE_INCOMPAT_META_ECC) {
2249                         status = ocfs2_block_check_validate(bh->b_data,
2250                                                             bh->b_size,
2251                                                             &di->i_check,
2252                                                             stats);
2253                         if (status)
2254                                 goto out;
2255                 }
2256                 status = -EINVAL;
2257                 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2258                         mlog(ML_ERROR, "found superblock with incorrect block "
2259                              "size: found %u, should be %u\n",
2260                              1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2261                                blksz);
2262                 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2263                            OCFS2_MAJOR_REV_LEVEL ||
2264                            le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2265                            OCFS2_MINOR_REV_LEVEL) {
2266                         mlog(ML_ERROR, "found superblock with bad version: "
2267                              "found %u.%u, should be %u.%u\n",
2268                              le16_to_cpu(di->id2.i_super.s_major_rev_level),
2269                              le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2270                              OCFS2_MAJOR_REV_LEVEL,
2271                              OCFS2_MINOR_REV_LEVEL);
2272                 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2273                         mlog(ML_ERROR, "bad block number on superblock: "
2274                              "found %llu, should be %llu\n",
2275                              (unsigned long long)le64_to_cpu(di->i_blkno),
2276                              (unsigned long long)bh->b_blocknr);
2277                 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2278                             le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2279                         mlog(ML_ERROR, "bad cluster size found: %u\n",
2280                              1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2281                 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2282                         mlog(ML_ERROR, "bad root_blkno: 0\n");
2283                 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2284                         mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2285                 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2286                         mlog(ML_ERROR,
2287                              "Superblock slots found greater than file system "
2288                              "maximum: found %u, max %u\n",
2289                              le16_to_cpu(di->id2.i_super.s_max_slots),
2290                              OCFS2_MAX_SLOTS);
2291                 } else {
2292                         /* found it! */
2293                         status = 0;
2294                 }
2295         }
2296
2297 out:
2298         mlog_exit(status);
2299         return status;
2300 }
2301
2302 static int ocfs2_check_volume(struct ocfs2_super *osb)
2303 {
2304         int status;
2305         int dirty;
2306         int local;
2307         struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2308                                                   * recover
2309                                                   * ourselves. */
2310
2311         mlog_entry_void();
2312
2313         /* Init our journal object. */
2314         status = ocfs2_journal_init(osb->journal, &dirty);
2315         if (status < 0) {
2316                 mlog(ML_ERROR, "Could not initialize journal!\n");
2317                 goto finally;
2318         }
2319
2320         /* If the journal was unmounted cleanly then we don't want to
2321          * recover anything. Otherwise, journal_load will do that
2322          * dirty work for us :) */
2323         if (!dirty) {
2324                 status = ocfs2_journal_wipe(osb->journal, 0);
2325                 if (status < 0) {
2326                         mlog_errno(status);
2327                         goto finally;
2328                 }
2329         } else {
2330                 mlog(ML_NOTICE, "File system was not unmounted cleanly, "
2331                      "recovering volume.\n");
2332         }
2333
2334         local = ocfs2_mount_local(osb);
2335
2336         /* will play back anything left in the journal. */
2337         status = ocfs2_journal_load(osb->journal, local, dirty);
2338         if (status < 0) {
2339                 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2340                 goto finally;
2341         }
2342
2343         if (dirty) {
2344                 /* recover my local alloc if we didn't unmount cleanly. */
2345                 status = ocfs2_begin_local_alloc_recovery(osb,
2346                                                           osb->slot_num,
2347                                                           &local_alloc);
2348                 if (status < 0) {
2349                         mlog_errno(status);
2350                         goto finally;
2351                 }
2352                 /* we complete the recovery process after we've marked
2353                  * ourselves as mounted. */
2354         }
2355
2356         mlog(0, "Journal loaded.\n");
2357
2358         status = ocfs2_load_local_alloc(osb);
2359         if (status < 0) {
2360                 mlog_errno(status);
2361                 goto finally;
2362         }
2363
2364         if (dirty) {
2365                 /* Recovery will be completed after we've mounted the
2366                  * rest of the volume. */
2367                 osb->dirty = 1;
2368                 osb->local_alloc_copy = local_alloc;
2369                 local_alloc = NULL;
2370         }
2371
2372         /* go through each journal, trylock it and if you get the
2373          * lock, and it's marked as dirty, set the bit in the recover
2374          * map and launch a recovery thread for it. */
2375         status = ocfs2_mark_dead_nodes(osb);
2376         if (status < 0) {
2377                 mlog_errno(status);
2378                 goto finally;
2379         }
2380
2381         status = ocfs2_compute_replay_slots(osb);
2382         if (status < 0)
2383                 mlog_errno(status);
2384
2385 finally:
2386         if (local_alloc)
2387                 kfree(local_alloc);
2388
2389         mlog_exit(status);
2390         return status;
2391 }
2392
2393 /*
2394  * The routine gets called from dismount or close whenever a dismount on
2395  * volume is requested and the osb open count becomes 1.
2396  * It will remove the osb from the global list and also free up all the
2397  * initialized resources and fileobject.
2398  */
2399 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2400 {
2401         mlog_entry_void();
2402
2403         /* This function assumes that the caller has the main osb resource */
2404
2405         ocfs2_free_slot_info(osb);
2406
2407         kfree(osb->osb_orphan_wipes);
2408         kfree(osb->slot_recovery_generations);
2409         /* FIXME
2410          * This belongs in journal shutdown, but because we have to
2411          * allocate osb->journal at the start of ocfs2_initalize_osb(),
2412          * we free it here.
2413          */
2414         kfree(osb->journal);
2415         if (osb->local_alloc_copy)
2416                 kfree(osb->local_alloc_copy);
2417         kfree(osb->uuid_str);
2418         ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2419         memset(osb, 0, sizeof(struct ocfs2_super));
2420
2421         mlog_exit_void();
2422 }
2423
2424 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2425  * panic(). We do not support continue-on-error operation. */
2426 static void ocfs2_handle_error(struct super_block *sb)
2427 {
2428         struct ocfs2_super *osb = OCFS2_SB(sb);
2429
2430         if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2431                 panic("OCFS2: (device %s): panic forced after error\n",
2432                       sb->s_id);
2433
2434         ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2435
2436         if (sb->s_flags & MS_RDONLY &&
2437             (ocfs2_is_soft_readonly(osb) ||
2438              ocfs2_is_hard_readonly(osb)))
2439                 return;
2440
2441         printk(KERN_CRIT "File system is now read-only due to the potential "
2442                "of on-disk corruption. Please run fsck.ocfs2 once the file "
2443                "system is unmounted.\n");
2444         sb->s_flags |= MS_RDONLY;
2445         ocfs2_set_ro_flag(osb, 0);
2446 }
2447
2448 static char error_buf[1024];
2449
2450 void __ocfs2_error(struct super_block *sb,
2451                    const char *function,
2452                    const char *fmt, ...)
2453 {
2454         va_list args;
2455
2456         va_start(args, fmt);
2457         vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2458         va_end(args);
2459
2460         /* Not using mlog here because we want to show the actual
2461          * function the error came from. */
2462         printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2463                sb->s_id, function, error_buf);
2464
2465         ocfs2_handle_error(sb);
2466 }
2467
2468 /* Handle critical errors. This is intentionally more drastic than
2469  * ocfs2_handle_error, so we only use for things like journal errors,
2470  * etc. */
2471 void __ocfs2_abort(struct super_block* sb,
2472                    const char *function,
2473                    const char *fmt, ...)
2474 {
2475         va_list args;
2476
2477         va_start(args, fmt);
2478         vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2479         va_end(args);
2480
2481         printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2482                sb->s_id, function, error_buf);
2483
2484         /* We don't have the cluster support yet to go straight to
2485          * hard readonly in here. Until then, we want to keep
2486          * ocfs2_abort() so that we can at least mark critical
2487          * errors.
2488          *
2489          * TODO: This should abort the journal and alert other nodes
2490          * that our slot needs recovery. */
2491
2492         /* Force a panic(). This stinks, but it's better than letting
2493          * things continue without having a proper hard readonly
2494          * here. */
2495         if (!ocfs2_mount_local(OCFS2_SB(sb)))
2496                 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2497         ocfs2_handle_error(sb);
2498 }
2499
2500 module_init(ocfs2_init);
2501 module_exit(ocfs2_exit);