Pileus Git - ~andy/linux/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
  45 #define MLOG_MASK_PREFIX ML_SUPER
  46 #include <cluster/masklog.h>
  47
  48 #include "ocfs2.h"
  49
  50 /* this should be the only file to include a version 1 header */
  51 #include "ocfs1_fs_compat.h"
  52
  53 #include "alloc.h"
  54 #include "blockcheck.h"
  55 #include "dlmglue.h"
  56 #include "export.h"
  57 #include "extent_map.h"
  58 #include "heartbeat.h"
  59 #include "inode.h"
  60 #include "journal.h"
  61 #include "localalloc.h"
  62 #include "namei.h"
  63 #include "slot_map.h"
  64 #include "super.h"
  65 #include "sysfile.h"
  66 #include "uptodate.h"
  67 #include "ver.h"
  68 #include "xattr.h"
  69 #include "quota.h"
  70 #include "refcounttree.h"
  71 #include "suballoc.h"
  72
  73 #include "buffer_head_io.h"
  74
  75 static struct kmem_cache *ocfs2_inode_cachep = NULL;
  76 struct kmem_cache *ocfs2_dquot_cachep;
  77 struct kmem_cache *ocfs2_qf_chunk_cachep;
  78
  79 /* OCFS2 needs to schedule several differnt types of work which
  80  * require cluster locking, disk I/O, recovery waits, etc. Since these
  81  * types of work tend to be heavy we avoid using the kernel events
  82  * workqueue and schedule on our own. */
  83 struct workqueue_struct *ocfs2_wq = NULL;
  84
  85 static struct dentry *ocfs2_debugfs_root = NULL;
  86
  87 MODULE_AUTHOR("Oracle");
  88 MODULE_LICENSE("GPL");
  89
  90 struct mount_options
  91 {
  92         unsigned long   commit_interval;
  93         unsigned long   mount_opt;
  94         unsigned int    atime_quantum;
  95         signed short    slot;
  96         int             localalloc_opt;
  97         unsigned int    resv_level;
  98         int             dir_resv_level;
  99         char            cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
 100 };
 101
 102 static int ocfs2_parse_options(struct super_block *sb, char *options,
 103                                struct mount_options *mopt,
 104                                int is_remount);
 105 static int ocfs2_check_set_options(struct super_block *sb,
 106                                    struct mount_options *options);
 107 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt);
 108 static void ocfs2_put_super(struct super_block *sb);
 109 static int ocfs2_mount_volume(struct super_block *sb);
 110 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
 111 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
 112 static int ocfs2_initialize_mem_caches(void);
 113 static void ocfs2_free_mem_caches(void);
 114 static void ocfs2_delete_osb(struct ocfs2_super *osb);
 115
 116 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
 117
 118 static int ocfs2_sync_fs(struct super_block *sb, int wait);
 119
 120 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
 121 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
 122 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
 123 static int ocfs2_check_volume(struct ocfs2_super *osb);
 124 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
 125                                struct buffer_head *bh,
 126                                u32 sectsize,
 127                                struct ocfs2_blockcheck_stats *stats);
 128 static int ocfs2_initialize_super(struct super_block *sb,
 129                                   struct buffer_head *bh,
 130                                   int sector_size,
 131                                   struct ocfs2_blockcheck_stats *stats);
 132 static int ocfs2_get_sector(struct super_block *sb,
 133                             struct buffer_head **bh,
 134                             int block,
 135                             int sect_size);
 136 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
 137 static void ocfs2_destroy_inode(struct inode *inode);
 138 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
 139 static int ocfs2_enable_quotas(struct ocfs2_super *osb);
 140 static void ocfs2_disable_quotas(struct ocfs2_super *osb);
 141
 142 static const struct super_operations ocfs2_sops = {
 143         .statfs         = ocfs2_statfs,
 144         .alloc_inode    = ocfs2_alloc_inode,
 145         .destroy_inode  = ocfs2_destroy_inode,
 146         .drop_inode     = ocfs2_drop_inode,
 147         .evict_inode    = ocfs2_evict_inode,
 148         .sync_fs        = ocfs2_sync_fs,
 149         .put_super      = ocfs2_put_super,
 150         .remount_fs     = ocfs2_remount,
 151         .show_options   = ocfs2_show_options,
 152         .quota_read     = ocfs2_quota_read,
 153         .quota_write    = ocfs2_quota_write,
 154 };
 155
 156 enum {
 157         Opt_barrier,
 158         Opt_err_panic,
 159         Opt_err_ro,
 160         Opt_intr,
 161         Opt_nointr,
 162         Opt_hb_none,
 163         Opt_hb_local,
 164         Opt_hb_global,
 165         Opt_data_ordered,
 166         Opt_data_writeback,
 167         Opt_atime_quantum,
 168         Opt_slot,
 169         Opt_commit,
 170         Opt_localalloc,
 171         Opt_localflocks,
 172         Opt_stack,
 173         Opt_user_xattr,
 174         Opt_nouser_xattr,
 175         Opt_inode64,
 176         Opt_acl,
 177         Opt_noacl,
 178         Opt_usrquota,
 179         Opt_grpquota,
 180         Opt_coherency_buffered,
 181         Opt_coherency_full,
 182         Opt_resv_level,
 183         Opt_dir_resv_level,
 184         Opt_err,
 185 };
 186
 187 static const match_table_t tokens = {
 188         {Opt_barrier, "barrier=%u"},
 189         {Opt_err_panic, "errors=panic"},
 190         {Opt_err_ro, "errors=remount-ro"},
 191         {Opt_intr, "intr"},
 192         {Opt_nointr, "nointr"},
 193         {Opt_hb_none, OCFS2_HB_NONE},
 194         {Opt_hb_local, OCFS2_HB_LOCAL},
 195         {Opt_hb_global, OCFS2_HB_GLOBAL},
 196         {Opt_data_ordered, "data=ordered"},
 197         {Opt_data_writeback, "data=writeback"},
 198         {Opt_atime_quantum, "atime_quantum=%u"},
 199         {Opt_slot, "preferred_slot=%u"},
 200         {Opt_commit, "commit=%u"},
 201         {Opt_localalloc, "localalloc=%d"},
 202         {Opt_localflocks, "localflocks"},
 203         {Opt_stack, "cluster_stack=%s"},
 204         {Opt_user_xattr, "user_xattr"},
 205         {Opt_nouser_xattr, "nouser_xattr"},
 206         {Opt_inode64, "inode64"},
 207         {Opt_acl, "acl"},
 208         {Opt_noacl, "noacl"},
 209         {Opt_usrquota, "usrquota"},
 210         {Opt_grpquota, "grpquota"},
 211         {Opt_coherency_buffered, "coherency=buffered"},
 212         {Opt_coherency_full, "coherency=full"},
 213         {Opt_resv_level, "resv_level=%u"},
 214         {Opt_dir_resv_level, "dir_resv_level=%u"},
 215         {Opt_err, NULL}
 216 };
 217
 218 #ifdef CONFIG_DEBUG_FS
 219 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
 220 {
 221         struct ocfs2_cluster_connection *cconn = osb->cconn;
 222         struct ocfs2_recovery_map *rm = osb->recovery_map;
 223         struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
 224         int i, out = 0;
 225
 226         out += snprintf(buf + out, len - out,
 227                         "%10s => Id: %-s  Uuid: %-s  Gen: 0x%X  Label: %-s\n",
 228                         "Device", osb->dev_str, osb->uuid_str,
 229                         osb->fs_generation, osb->vol_label);
 230
 231         out += snprintf(buf + out, len - out,
 232                         "%10s => State: %d  Flags: 0x%lX\n", "Volume",
 233                         atomic_read(&osb->vol_state), osb->osb_flags);
 234
 235         out += snprintf(buf + out, len - out,
 236                         "%10s => Block: %lu  Cluster: %d\n", "Sizes",
 237                         osb->sb->s_blocksize, osb->s_clustersize);
 238
 239         out += snprintf(buf + out, len - out,
 240                         "%10s => Compat: 0x%X  Incompat: 0x%X  "
 241                         "ROcompat: 0x%X\n",
 242                         "Features", osb->s_feature_compat,
 243                         osb->s_feature_incompat, osb->s_feature_ro_compat);
 244
 245         out += snprintf(buf + out, len - out,
 246                         "%10s => Opts: 0x%lX  AtimeQuanta: %u\n", "Mount",
 247                         osb->s_mount_opt, osb->s_atime_quantum);
 248
 249         if (cconn) {
 250                 out += snprintf(buf + out, len - out,
 251                                 "%10s => Stack: %s  Name: %*s  "
 252                                 "Version: %d.%d\n", "Cluster",
 253                                 (*osb->osb_cluster_stack == '\0' ?
 254                                  "o2cb" : osb->osb_cluster_stack),
 255                                 cconn->cc_namelen, cconn->cc_name,
 256                                 cconn->cc_version.pv_major,
 257                                 cconn->cc_version.pv_minor);
 258         }
 259
 260         spin_lock(&osb->dc_task_lock);
 261         out += snprintf(buf + out, len - out,
 262                         "%10s => Pid: %d  Count: %lu  WakeSeq: %lu  "
 263                         "WorkSeq: %lu\n", "DownCnvt",
 264                         (osb->dc_task ?  task_pid_nr(osb->dc_task) : -1),
 265                         osb->blocked_lock_count, osb->dc_wake_sequence,
 266                         osb->dc_work_sequence);
 267         spin_unlock(&osb->dc_task_lock);
 268
 269         spin_lock(&osb->osb_lock);
 270         out += snprintf(buf + out, len - out, "%10s => Pid: %d  Nodes:",
 271                         "Recovery",
 272                         (osb->recovery_thread_task ?
 273                          task_pid_nr(osb->recovery_thread_task) : -1));
 274         if (rm->rm_used == 0)
 275                 out += snprintf(buf + out, len - out, " None\n");
 276         else {
 277                 for (i = 0; i < rm->rm_used; i++)
 278                         out += snprintf(buf + out, len - out, " %d",
 279                                         rm->rm_entries[i]);
 280                 out += snprintf(buf + out, len - out, "\n");
 281         }
 282         spin_unlock(&osb->osb_lock);
 283
 284         out += snprintf(buf + out, len - out,
 285                         "%10s => Pid: %d  Interval: %lu  Needs: %d\n", "Commit",
 286                         (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
 287                         osb->osb_commit_interval,
 288                         atomic_read(&osb->needs_checkpoint));
 289
 290         out += snprintf(buf + out, len - out,
 291                         "%10s => State: %d  TxnId: %lu  NumTxns: %d\n",
 292                         "Journal", osb->journal->j_state,
 293                         osb->journal->j_trans_id,
 294                         atomic_read(&osb->journal->j_num_trans));
 295
 296         out += snprintf(buf + out, len - out,
 297                         "%10s => GlobalAllocs: %d  LocalAllocs: %d  "
 298                         "SubAllocs: %d  LAWinMoves: %d  SAExtends: %d\n",
 299                         "Stats",
 300                         atomic_read(&osb->alloc_stats.bitmap_data),
 301                         atomic_read(&osb->alloc_stats.local_data),
 302                         atomic_read(&osb->alloc_stats.bg_allocs),
 303                         atomic_read(&osb->alloc_stats.moves),
 304                         atomic_read(&osb->alloc_stats.bg_extends));
 305
 306         out += snprintf(buf + out, len - out,
 307                         "%10s => State: %u  Descriptor: %llu  Size: %u bits  "
 308                         "Default: %u bits\n",
 309                         "LocalAlloc", osb->local_alloc_state,
 310                         (unsigned long long)osb->la_last_gd,
 311                         osb->local_alloc_bits, osb->local_alloc_default_bits);
 312
 313         spin_lock(&osb->osb_lock);
 314         out += snprintf(buf + out, len - out,
 315                         "%10s => InodeSlot: %d  StolenInodes: %d, "
 316                         "MetaSlot: %d  StolenMeta: %d\n", "Steal",
 317                         osb->s_inode_steal_slot,
 318                         atomic_read(&osb->s_num_inodes_stolen),
 319                         osb->s_meta_steal_slot,
 320                         atomic_read(&osb->s_num_meta_stolen));
 321         spin_unlock(&osb->osb_lock);
 322
 323         out += snprintf(buf + out, len - out, "OrphanScan => ");
 324         out += snprintf(buf + out, len - out, "Local: %u  Global: %u ",
 325                         os->os_count, os->os_seqno);
 326         out += snprintf(buf + out, len - out, " Last Scan: ");
 327         if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
 328                 out += snprintf(buf + out, len - out, "Disabled\n");
 329         else
 330                 out += snprintf(buf + out, len - out, "%lu seconds ago\n",
 331                                 (get_seconds() - os->os_scantime.tv_sec));
 332
 333         out += snprintf(buf + out, len - out, "%10s => %3s  %10s\n",
 334                         "Slots", "Num", "RecoGen");
 335         for (i = 0; i < osb->max_slots; ++i) {
 336                 out += snprintf(buf + out, len - out,
 337                                 "%10s  %c %3d  %10d\n",
 338                                 " ",
 339                                 (i == osb->slot_num ? '*' : ' '),
 340                                 i, osb->slot_recovery_generations[i]);
 341         }
 342
 343         return out;
 344 }
 345
 346 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 347 {
 348         struct ocfs2_super *osb = inode->i_private;
 349         char *buf = NULL;
 350
 351         buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
 352         if (!buf)
 353                 goto bail;
 354
 355         i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
 356
 357         file->private_data = buf;
 358
 359         return 0;
 360 bail:
 361         return -ENOMEM;
 362 }
 363
 364 static int ocfs2_debug_release(struct inode *inode, struct file *file)
 365 {
 366         kfree(file->private_data);
 367         return 0;
 368 }
 369
 370 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 371                                 size_t nbytes, loff_t *ppos)
 372 {
 373         return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
 374                                        i_size_read(file->f_mapping->host));
 375 }
 376 #else
 377 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 378 {
 379         return 0;
 380 }
 381 static int ocfs2_debug_release(struct inode *inode, struct file *file)
 382 {
 383         return 0;
 384 }
 385 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 386                                 size_t nbytes, loff_t *ppos)
 387 {
 388         return 0;
 389 }
 390 #endif  /* CONFIG_DEBUG_FS */
 391
 392 static const struct file_operations ocfs2_osb_debug_fops = {
 393         .open =         ocfs2_osb_debug_open,
 394         .release =      ocfs2_debug_release,
 395         .read =         ocfs2_debug_read,
 396         .llseek =       generic_file_llseek,
 397 };
 398
 399 static int ocfs2_sync_fs(struct super_block *sb, int wait)
 400 {
 401         int status;
 402         tid_t target;
 403         struct ocfs2_super *osb = OCFS2_SB(sb);
 404
 405         if (ocfs2_is_hard_readonly(osb))
 406                 return -EROFS;
 407
 408         if (wait) {
 409                 status = ocfs2_flush_truncate_log(osb);
 410                 if (status < 0)
 411                         mlog_errno(status);
 412         } else {
 413                 ocfs2_schedule_truncate_log_flush(osb, 0);
 414         }
 415
 416         if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
 417                                       &target)) {
 418                 if (wait)
 419                         jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
 420                                              target);
 421         }
 422         return 0;
 423 }
 424
 425 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
 426 {
 427         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
 428             && (ino == USER_QUOTA_SYSTEM_INODE
 429                 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
 430                 return 0;
 431         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
 432             && (ino == GROUP_QUOTA_SYSTEM_INODE
 433                 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
 434                 return 0;
 435         return 1;
 436 }
 437
 438 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
 439 {
 440         struct inode *new = NULL;
 441         int status = 0;
 442         int i;
 443
 444         new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 445         if (IS_ERR(new)) {
 446                 status = PTR_ERR(new);
 447                 mlog_errno(status);
 448                 goto bail;
 449         }
 450         osb->root_inode = new;
 451
 452         new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 453         if (IS_ERR(new)) {
 454                 status = PTR_ERR(new);
 455                 mlog_errno(status);
 456                 goto bail;
 457         }
 458         osb->sys_root_inode = new;
 459
 460         for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
 461              i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
 462                 if (!ocfs2_need_system_inode(osb, i))
 463                         continue;
 464                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 465                 if (!new) {
 466                         ocfs2_release_system_inodes(osb);
 467                         status = -EINVAL;
 468                         mlog_errno(status);
 469                         /* FIXME: Should ERROR_RO_FS */
 470                         mlog(ML_ERROR, "Unable to load system inode %d, "
 471                              "possibly corrupt fs?", i);
 472                         goto bail;
 473                 }
 474                 // the array now has one ref, so drop this one
 475                 iput(new);
 476         }
 477
 478 bail:
 479         mlog_exit(status);
 480         return status;
 481 }
 482
 483 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
 484 {
 485         struct inode *new = NULL;
 486         int status = 0;
 487         int i;
 488
 489         for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
 490              i < NUM_SYSTEM_INODES;
 491              i++) {
 492                 if (!ocfs2_need_system_inode(osb, i))
 493                         continue;
 494                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 495                 if (!new) {
 496                         ocfs2_release_system_inodes(osb);
 497                         status = -EINVAL;
 498                         mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
 499                              status, i, osb->slot_num);
 500                         goto bail;
 501                 }
 502                 /* the array now has one ref, so drop this one */
 503                 iput(new);
 504         }
 505
 506 bail:
 507         mlog_exit(status);
 508         return status;
 509 }
 510
 511 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
 512 {
 513         int i;
 514         struct inode *inode;
 515
 516         for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
 517                 inode = osb->global_system_inodes[i];
 518                 if (inode) {
 519                         iput(inode);
 520                         osb->global_system_inodes[i] = NULL;
 521                 }
 522         }
 523
 524         inode = osb->sys_root_inode;
 525         if (inode) {
 526                 iput(inode);
 527                 osb->sys_root_inode = NULL;
 528         }
 529
 530         inode = osb->root_inode;
 531         if (inode) {
 532                 iput(inode);
 533                 osb->root_inode = NULL;
 534         }
 535
 536         if (!osb->local_system_inodes)
 537                 goto out;
 538
 539         for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
 540                 if (osb->local_system_inodes[i]) {
 541                         iput(osb->local_system_inodes[i]);
 542                         osb->local_system_inodes[i] = NULL;
 543                 }
 544         }
 545
 546         kfree(osb->local_system_inodes);
 547         osb->local_system_inodes = NULL;
 548
 549 out:
 550         mlog_exit(0);
 551 }
 552
 553 /* We're allocating fs objects, use GFP_NOFS */
 554 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
 555 {
 556         struct ocfs2_inode_info *oi;
 557
 558         oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
 559         if (!oi)
 560                 return NULL;
 561
 562         jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
 563         return &oi->vfs_inode;
 564 }
 565
 566 static void ocfs2_i_callback(struct rcu_head *head)
 567 {
 568         struct inode *inode = container_of(head, struct inode, i_rcu);
 569         INIT_LIST_HEAD(&inode->i_dentry);
 570         kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
 571 }
 572
 573 static void ocfs2_destroy_inode(struct inode *inode)
 574 {
 575         call_rcu(&inode->i_rcu, ocfs2_i_callback);
 576 }
 577
 578 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
 579                                                 unsigned int cbits)
 580 {
 581         unsigned int bytes = 1 << cbits;
 582         unsigned int trim = bytes;
 583         unsigned int bitshift = 32;
 584
 585         /*
 586          * i_size and all block offsets in ocfs2 are always 64 bits
 587          * wide. i_clusters is 32 bits, in cluster-sized units. So on
 588          * 64 bit platforms, cluster size will be the limiting factor.
 589          */
 590
 591 #if BITS_PER_LONG == 32
 592 # if defined(CONFIG_LBDAF)
 593         BUILD_BUG_ON(sizeof(sector_t) != 8);
 594         /*
 595          * We might be limited by page cache size.
 596          */
 597         if (bytes > PAGE_CACHE_SIZE) {
 598                 bytes = PAGE_CACHE_SIZE;
 599                 trim = 1;
 600                 /*
 601                  * Shift by 31 here so that we don't get larger than
 602                  * MAX_LFS_FILESIZE
 603                  */
 604                 bitshift = 31;
 605         }
 606 # else
 607         /*
 608          * We are limited by the size of sector_t. Use block size, as
 609          * that's what we expose to the VFS.
 610          */
 611         bytes = 1 << bbits;
 612         trim = 1;
 613         bitshift = 31;
 614 # endif
 615 #endif
 616
 617         /*
 618          * Trim by a whole cluster when we can actually approach the
 619          * on-disk limits. Otherwise we can overflow i_clusters when
 620          * an extent start is at the max offset.
 621          */
 622         return (((unsigned long long)bytes) << bitshift) - trim;
 623 }
 624
 625 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
 626 {
 627         int incompat_features;
 628         int ret = 0;
 629         struct mount_options parsed_options;
 630         struct ocfs2_super *osb = OCFS2_SB(sb);
 631         u32 tmp;
 632
 633         if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
 634             !ocfs2_check_set_options(sb, &parsed_options)) {
 635                 ret = -EINVAL;
 636                 goto out;
 637         }
 638
 639         tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
 640                 OCFS2_MOUNT_HB_NONE;
 641         if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
 642                 ret = -EINVAL;
 643                 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
 644                 goto out;
 645         }
 646
 647         if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
 648             (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
 649                 ret = -EINVAL;
 650                 mlog(ML_ERROR, "Cannot change data mode on remount\n");
 651                 goto out;
 652         }
 653
 654         /* Probably don't want this on remount; it might
 655          * mess with other nodes */
 656         if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
 657             (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
 658                 ret = -EINVAL;
 659                 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
 660                 goto out;
 661         }
 662
 663         /* We're going to/from readonly mode. */
 664         if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
 665                 /* Disable quota accounting before remounting RO */
 666                 if (*flags & MS_RDONLY) {
 667                         ret = ocfs2_susp_quotas(osb, 0);
 668                         if (ret < 0)
 669                                 goto out;
 670                 }
 671                 /* Lock here so the check of HARD_RO and the potential
 672                  * setting of SOFT_RO is atomic. */
 673                 spin_lock(&osb->osb_lock);
 674                 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
 675                         mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
 676                         ret = -EROFS;
 677                         goto unlock_osb;
 678                 }
 679
 680                 if (*flags & MS_RDONLY) {
 681                         mlog(0, "Going to ro mode.\n");
 682                         sb->s_flags |= MS_RDONLY;
 683                         osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 684                 } else {
 685                         mlog(0, "Making ro filesystem writeable.\n");
 686
 687                         if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
 688                                 mlog(ML_ERROR, "Cannot remount RDWR "
 689                                      "filesystem due to previous errors.\n");
 690                                 ret = -EROFS;
 691                                 goto unlock_osb;
 692                         }
 693                         incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
 694                         if (incompat_features) {
 695                                 mlog(ML_ERROR, "Cannot remount RDWR because "
 696                                      "of unsupported optional features "
 697                                      "(%x).\n", incompat_features);
 698                                 ret = -EINVAL;
 699                                 goto unlock_osb;
 700                         }
 701                         sb->s_flags &= ~MS_RDONLY;
 702                         osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
 703                 }
 704 unlock_osb:
 705                 spin_unlock(&osb->osb_lock);
 706                 /* Enable quota accounting after remounting RW */
 707                 if (!ret && !(*flags & MS_RDONLY)) {
 708                         if (sb_any_quota_suspended(sb))
 709                                 ret = ocfs2_susp_quotas(osb, 1);
 710                         else
 711                                 ret = ocfs2_enable_quotas(osb);
 712                         if (ret < 0) {
 713                                 /* Return back changes... */
 714                                 spin_lock(&osb->osb_lock);
 715                                 sb->s_flags |= MS_RDONLY;
 716                                 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 717                                 spin_unlock(&osb->osb_lock);
 718                                 goto out;
 719                         }
 720                 }
 721         }
 722
 723         if (!ret) {
 724                 /* Only save off the new mount options in case of a successful
 725                  * remount. */
 726                 osb->s_mount_opt = parsed_options.mount_opt;
 727                 osb->s_atime_quantum = parsed_options.atime_quantum;
 728                 osb->preferred_slot = parsed_options.slot;
 729                 if (parsed_options.commit_interval)
 730                         osb->osb_commit_interval = parsed_options.commit_interval;
 731
 732                 if (!ocfs2_is_hard_readonly(osb))
 733                         ocfs2_set_journal_params(osb);
 734
 735                 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 736                         ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
 737                                                         MS_POSIXACL : 0);
 738         }
 739 out:
 740         return ret;
 741 }
 742
 743 static int ocfs2_sb_probe(struct super_block *sb,
 744                           struct buffer_head **bh,
 745                           int *sector_size,
 746                           struct ocfs2_blockcheck_stats *stats)
 747 {
 748         int status, tmpstat;
 749         struct ocfs1_vol_disk_hdr *hdr;
 750         struct ocfs2_dinode *di;
 751         int blksize;
 752
 753         *bh = NULL;
 754
 755         /* may be > 512 */
 756         *sector_size = bdev_logical_block_size(sb->s_bdev);
 757         if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
 758                 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
 759                      *sector_size, OCFS2_MAX_BLOCKSIZE);
 760                 status = -EINVAL;
 761                 goto bail;
 762         }
 763
 764         /* Can this really happen? */
 765         if (*sector_size < OCFS2_MIN_BLOCKSIZE)
 766                 *sector_size = OCFS2_MIN_BLOCKSIZE;
 767
 768         /* check block zero for old format */
 769         status = ocfs2_get_sector(sb, bh, 0, *sector_size);
 770         if (status < 0) {
 771                 mlog_errno(status);
 772                 goto bail;
 773         }
 774         hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
 775         if (hdr->major_version == OCFS1_MAJOR_VERSION) {
 776                 mlog(ML_ERROR, "incompatible version: %u.%u\n",
 777                      hdr->major_version, hdr->minor_version);
 778                 status = -EINVAL;
 779         }
 780         if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
 781                    strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
 782                 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
 783                      hdr->signature);
 784                 status = -EINVAL;
 785         }
 786         brelse(*bh);
 787         *bh = NULL;
 788         if (status < 0) {
 789                 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
 790                      "upgraded before mounting with ocfs v2\n");
 791                 goto bail;
 792         }
 793
 794         /*
 795          * Now check at magic offset for 512, 1024, 2048, 4096
 796          * blocksizes.  4096 is the maximum blocksize because it is
 797          * the minimum clustersize.
 798          */
 799         status = -EINVAL;
 800         for (blksize = *sector_size;
 801              blksize <= OCFS2_MAX_BLOCKSIZE;
 802              blksize <<= 1) {
 803                 tmpstat = ocfs2_get_sector(sb, bh,
 804                                            OCFS2_SUPER_BLOCK_BLKNO,
 805                                            blksize);
 806                 if (tmpstat < 0) {
 807                         status = tmpstat;
 808                         mlog_errno(status);
 809                         break;
 810                 }
 811                 di = (struct ocfs2_dinode *) (*bh)->b_data;
 812                 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
 813                 spin_lock_init(&stats->b_lock);
 814                 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
 815                 if (tmpstat < 0) {
 816                         brelse(*bh);
 817                         *bh = NULL;
 818                 }
 819                 if (tmpstat != -EAGAIN) {
 820                         status = tmpstat;
 821                         break;
 822                 }
 823         }
 824
 825 bail:
 826         return status;
 827 }
 828
 829 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
 830 {
 831         u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
 832
 833         if (osb->s_mount_opt & hb_enabled) {
 834                 if (ocfs2_mount_local(osb)) {
 835                         mlog(ML_ERROR, "Cannot heartbeat on a locally "
 836                              "mounted device.\n");
 837                         return -EINVAL;
 838                 }
 839                 if (ocfs2_userspace_stack(osb)) {
 840                         mlog(ML_ERROR, "Userspace stack expected, but "
 841                              "o2cb heartbeat arguments passed to mount\n");
 842                         return -EINVAL;
 843                 }
 844                 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
 845                      !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
 846                     ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
 847                      ocfs2_cluster_o2cb_global_heartbeat(osb))) {
 848                         mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
 849                         return -EINVAL;
 850                 }
 851         }
 852
 853         if (!(osb->s_mount_opt & hb_enabled)) {
 854                 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
 855                     !ocfs2_userspace_stack(osb)) {
 856                         mlog(ML_ERROR, "Heartbeat has to be started to mount "
 857                              "a read-write clustered device.\n");
 858                         return -EINVAL;
 859                 }
 860         }
 861
 862         return 0;
 863 }
 864
 865 /*
 866  * If we're using a userspace stack, mount should have passed
 867  * a name that matches the disk.  If not, mount should not
 868  * have passed a stack.
 869  */
 870 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
 871                                         struct mount_options *mopt)
 872 {
 873         if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
 874                 mlog(ML_ERROR,
 875                      "cluster stack passed to mount, but this filesystem "
 876                      "does not support it\n");
 877                 return -EINVAL;
 878         }
 879
 880         if (ocfs2_userspace_stack(osb) &&
 881             strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
 882                     OCFS2_STACK_LABEL_LEN)) {
 883                 mlog(ML_ERROR,
 884                      "cluster stack passed to mount (\"%s\") does not "
 885                      "match the filesystem (\"%s\")\n",
 886                      mopt->cluster_stack,
 887                      osb->osb_cluster_stack);
 888                 return -EINVAL;
 889         }
 890
 891         return 0;
 892 }
 893
 894 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
 895 {
 896         int type;
 897         struct super_block *sb = osb->sb;
 898         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 899                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 900         int status = 0;
 901
 902         for (type = 0; type < MAXQUOTAS; type++) {
 903                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 904                         continue;
 905                 if (unsuspend)
 906                         status = dquot_resume(sb, type);
 907                 else {
 908                         struct ocfs2_mem_dqinfo *oinfo;
 909
 910                         /* Cancel periodic syncing before suspending */
 911                         oinfo = sb_dqinfo(sb, type)->dqi_priv;
 912                         cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 913                         status = dquot_suspend(sb, type);
 914                 }
 915                 if (status < 0)
 916                         break;
 917         }
 918         if (status < 0)
 919                 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
 920                      "remount (error = %d).\n", status);
 921         return status;
 922 }
 923
 924 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
 925 {
 926         struct inode *inode[MAXQUOTAS] = { NULL, NULL };
 927         struct super_block *sb = osb->sb;
 928         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 929                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 930         unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
 931                                         LOCAL_GROUP_QUOTA_SYSTEM_INODE };
 932         int status;
 933         int type;
 934
 935         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
 936         for (type = 0; type < MAXQUOTAS; type++) {
 937                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 938                         continue;
 939                 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
 940                                                         osb->slot_num);
 941                 if (!inode[type]) {
 942                         status = -ENOENT;
 943                         goto out_quota_off;
 944                 }
 945                 status = dquot_enable(inode[type], type, QFMT_OCFS2,
 946                                       DQUOT_USAGE_ENABLED);
 947                 if (status < 0)
 948                         goto out_quota_off;
 949         }
 950
 951         for (type = 0; type < MAXQUOTAS; type++)
 952                 iput(inode[type]);
 953         return 0;
 954 out_quota_off:
 955         ocfs2_disable_quotas(osb);
 956         for (type = 0; type < MAXQUOTAS; type++)
 957                 iput(inode[type]);
 958         mlog_errno(status);
 959         return status;
 960 }
 961
 962 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
 963 {
 964         int type;
 965         struct inode *inode;
 966         struct super_block *sb = osb->sb;
 967         struct ocfs2_mem_dqinfo *oinfo;
 968
 969         /* We mostly ignore errors in this function because there's not much
 970          * we can do when we see them */
 971         for (type = 0; type < MAXQUOTAS; type++) {
 972                 if (!sb_has_quota_loaded(sb, type))
 973                         continue;
 974                 /* Cancel periodic syncing before we grab dqonoff_mutex */
 975                 oinfo = sb_dqinfo(sb, type)->dqi_priv;
 976                 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 977                 inode = igrab(sb->s_dquot.files[type]);
 978                 /* Turn off quotas. This will remove all dquot structures from
 979                  * memory and so they will be automatically synced to global
 980                  * quota files */
 981                 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
 982                                         DQUOT_LIMITS_ENABLED);
 983                 if (!inode)
 984                         continue;
 985                 iput(inode);
 986         }
 987 }
 988
 989 /* Handle quota on quotactl */
 990 static int ocfs2_quota_on(struct super_block *sb, int type, int format_id)
 991 {
 992         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 993                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 994
 995         if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 996                 return -EINVAL;
 997
 998         return dquot_enable(sb_dqopt(sb)->files[type], type,
 999                             format_id, DQUOT_LIMITS_ENABLED);
1000 }
1001
1002 /* Handle quota off quotactl */
1003 static int ocfs2_quota_off(struct super_block *sb, int type)
1004 {
1005         return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
1006 }
1007
1008 static const struct quotactl_ops ocfs2_quotactl_ops = {
1009         .quota_on_meta  = ocfs2_quota_on,
1010         .quota_off      = ocfs2_quota_off,
1011         .quota_sync     = dquot_quota_sync,
1012         .get_info       = dquot_get_dqinfo,
1013         .set_info       = dquot_set_dqinfo,
1014         .get_dqblk      = dquot_get_dqblk,
1015         .set_dqblk      = dquot_set_dqblk,
1016 };
1017
1018 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
1019 {
1020         struct dentry *root;
1021         int status, sector_size;
1022         struct mount_options parsed_options;
1023         struct inode *inode = NULL;
1024         struct ocfs2_super *osb = NULL;
1025         struct buffer_head *bh = NULL;
1026         char nodestr[8];
1027         struct ocfs2_blockcheck_stats stats;
1028
1029         mlog(0, "%p, %p, %i", sb, data, silent);
1030
1031         if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
1032                 status = -EINVAL;
1033                 goto read_super_error;
1034         }
1035
1036         /* probe for superblock */
1037         status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
1038         if (status < 0) {
1039                 mlog(ML_ERROR, "superblock probe failed!\n");
1040                 goto read_super_error;
1041         }
1042
1043         status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1044         osb = OCFS2_SB(sb);
1045         if (status < 0) {
1046                 mlog_errno(status);
1047                 goto read_super_error;
1048         }
1049         brelse(bh);
1050         bh = NULL;
1051
1052         if (!ocfs2_check_set_options(sb, &parsed_options)) {
1053                 status = -EINVAL;
1054                 goto read_super_error;
1055         }
1056         osb->s_mount_opt = parsed_options.mount_opt;
1057         osb->s_atime_quantum = parsed_options.atime_quantum;
1058         osb->preferred_slot = parsed_options.slot;
1059         osb->osb_commit_interval = parsed_options.commit_interval;
1060
1061         ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1062         osb->osb_resv_level = parsed_options.resv_level;
1063         osb->osb_dir_resv_level = parsed_options.resv_level;
1064         if (parsed_options.dir_resv_level == -1)
1065                 osb->osb_dir_resv_level = parsed_options.resv_level;
1066         else
1067                 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1068
1069         status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1070         if (status)
1071                 goto read_super_error;
1072
1073         sb->s_magic = OCFS2_SUPER_MAGIC;
1074
1075         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1076                 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1077
1078         /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1079          * heartbeat=none */
1080         if (bdev_read_only(sb->s_bdev)) {
1081                 if (!(sb->s_flags & MS_RDONLY)) {
1082                         status = -EACCES;
1083                         mlog(ML_ERROR, "Readonly device detected but readonly "
1084                              "mount was not specified.\n");
1085                         goto read_super_error;
1086                 }
1087
1088                 /* You should not be able to start a local heartbeat
1089                  * on a readonly device. */
1090                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1091                         status = -EROFS;
1092                         mlog(ML_ERROR, "Local heartbeat specified on readonly "
1093                              "device.\n");
1094                         goto read_super_error;
1095                 }
1096
1097                 status = ocfs2_check_journals_nolocks(osb);
1098                 if (status < 0) {
1099                         if (status == -EROFS)
1100                                 mlog(ML_ERROR, "Recovery required on readonly "
1101                                      "file system, but write access is "
1102                                      "unavailable.\n");
1103                         else
1104                                 mlog_errno(status);
1105                         goto read_super_error;
1106                 }
1107
1108                 ocfs2_set_ro_flag(osb, 1);
1109
1110                 printk(KERN_NOTICE "Readonly device detected. No cluster "
1111                        "services will be utilized for this mount. Recovery "
1112                        "will be skipped.\n");
1113         }
1114
1115         if (!ocfs2_is_hard_readonly(osb)) {
1116                 if (sb->s_flags & MS_RDONLY)
1117                         ocfs2_set_ro_flag(osb, 0);
1118         }
1119
1120         status = ocfs2_verify_heartbeat(osb);
1121         if (status < 0) {
1122                 mlog_errno(status);
1123                 goto read_super_error;
1124         }
1125
1126         osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1127                                                  ocfs2_debugfs_root);
1128         if (!osb->osb_debug_root) {
1129                 status = -EINVAL;
1130                 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1131                 goto read_super_error;
1132         }
1133
1134         osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1135                                             osb->osb_debug_root,
1136                                             osb,
1137                                             &ocfs2_osb_debug_fops);
1138         if (!osb->osb_ctxt) {
1139                 status = -EINVAL;
1140                 mlog_errno(status);
1141                 goto read_super_error;
1142         }
1143
1144         if (ocfs2_meta_ecc(osb)) {
1145                 status = ocfs2_blockcheck_stats_debugfs_install(
1146                                                 &osb->osb_ecc_stats,
1147                                                 osb->osb_debug_root);
1148                 if (status) {
1149                         mlog(ML_ERROR,
1150                              "Unable to create blockcheck statistics "
1151                              "files\n");
1152                         goto read_super_error;
1153                 }
1154         }
1155
1156         status = ocfs2_mount_volume(sb);
1157         if (osb->root_inode)
1158                 inode = igrab(osb->root_inode);
1159
1160         if (status < 0)
1161                 goto read_super_error;
1162
1163         if (!inode) {
1164                 status = -EIO;
1165                 mlog_errno(status);
1166                 goto read_super_error;
1167         }
1168
1169         root = d_alloc_root(inode);
1170         if (!root) {
1171                 status = -ENOMEM;
1172                 mlog_errno(status);
1173                 goto read_super_error;
1174         }
1175
1176         sb->s_root = root;
1177
1178         ocfs2_complete_mount_recovery(osb);
1179
1180         if (ocfs2_mount_local(osb))
1181                 snprintf(nodestr, sizeof(nodestr), "local");
1182         else
1183                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1184
1185         printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1186                "with %s data mode.\n",
1187                osb->dev_str, nodestr, osb->slot_num,
1188                osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1189                "ordered");
1190
1191         atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1192         wake_up(&osb->osb_mount_event);
1193
1194         /* Now we can initialize quotas because we can afford to wait
1195          * for cluster locks recovery now. That also means that truncation
1196          * log recovery can happen but that waits for proper quota setup */
1197         if (!(sb->s_flags & MS_RDONLY)) {
1198                 status = ocfs2_enable_quotas(osb);
1199                 if (status < 0) {
1200                         /* We have to err-out specially here because
1201                          * s_root is already set */
1202                         mlog_errno(status);
1203                         atomic_set(&osb->vol_state, VOLUME_DISABLED);
1204                         wake_up(&osb->osb_mount_event);
1205                         mlog_exit(status);
1206                         return status;
1207                 }
1208         }
1209
1210         ocfs2_complete_quota_recovery(osb);
1211
1212         /* Now we wake up again for processes waiting for quotas */
1213         atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1214         wake_up(&osb->osb_mount_event);
1215
1216         /* Start this when the mount is almost sure of being successful */
1217         ocfs2_orphan_scan_start(osb);
1218
1219         mlog_exit(status);
1220         return status;
1221
1222 read_super_error:
1223         brelse(bh);
1224
1225         if (inode)
1226                 iput(inode);
1227
1228         if (osb) {
1229                 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1230                 wake_up(&osb->osb_mount_event);
1231                 ocfs2_dismount_volume(sb, 1);
1232         }
1233
1234         mlog_exit(status);
1235         return status;
1236 }
1237
1238 static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
1239                         int flags,
1240                         const char *dev_name,
1241                         void *data)
1242 {
1243         return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
1244 }
1245
1246 static void ocfs2_kill_sb(struct super_block *sb)
1247 {
1248         struct ocfs2_super *osb = OCFS2_SB(sb);
1249
1250         /* Failed mount? */
1251         if (!osb || atomic_read(&osb->vol_state) == VOLUME_DISABLED)
1252                 goto out;
1253
1254         /* Prevent further queueing of inode drop events */
1255         spin_lock(&dentry_list_lock);
1256         ocfs2_set_osb_flag(osb, OCFS2_OSB_DROP_DENTRY_LOCK_IMMED);
1257         spin_unlock(&dentry_list_lock);
1258         /* Wait for work to finish and/or remove it */
1259         cancel_work_sync(&osb->dentry_lock_work);
1260 out:
1261         kill_block_super(sb);
1262 }
1263
1264 static struct file_system_type ocfs2_fs_type = {
1265         .owner          = THIS_MODULE,
1266         .name           = "ocfs2",
1267         .mount          = ocfs2_mount,
1268         .kill_sb        = ocfs2_kill_sb,
1269
1270         .fs_flags       = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1271         .next           = NULL
1272 };
1273
1274 static int ocfs2_check_set_options(struct super_block *sb,
1275                                    struct mount_options *options)
1276 {
1277         if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1278             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1279                                          OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1280                 mlog(ML_ERROR, "User quotas were requested, but this "
1281                      "filesystem does not have the feature enabled.\n");
1282                 return 0;
1283         }
1284         if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1285             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1286                                          OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1287                 mlog(ML_ERROR, "Group quotas were requested, but this "
1288                      "filesystem does not have the feature enabled.\n");
1289                 return 0;
1290         }
1291         if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1292             !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1293                 mlog(ML_ERROR, "ACL support requested but extended attributes "
1294                      "feature is not enabled\n");
1295                 return 0;
1296         }
1297         /* No ACL setting specified? Use XATTR feature... */
1298         if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1299                                     OCFS2_MOUNT_NO_POSIX_ACL))) {
1300                 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1301                         options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1302                 else
1303                         options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1304         }
1305         return 1;
1306 }
1307
1308 static int ocfs2_parse_options(struct super_block *sb,
1309                                char *options,
1310                                struct mount_options *mopt,
1311                                int is_remount)
1312 {
1313         int status, user_stack = 0;
1314         char *p;
1315         u32 tmp;
1316
1317         mlog(0, "remount: %d, options: \"%s\"\n", is_remount,
1318              options ? options : "(none)");
1319
1320         mopt->commit_interval = 0;
1321         mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1322         mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1323         mopt->slot = OCFS2_INVALID_SLOT;
1324         mopt->localalloc_opt = -1;
1325         mopt->cluster_stack[0] = '\0';
1326         mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1327         mopt->dir_resv_level = -1;
1328
1329         if (!options) {
1330                 status = 1;
1331                 goto bail;
1332         }
1333
1334         while ((p = strsep(&options, ",")) != NULL) {
1335                 int token, option;
1336                 substring_t args[MAX_OPT_ARGS];
1337
1338                 if (!*p)
1339                         continue;
1340
1341                 token = match_token(p, tokens, args);
1342                 switch (token) {
1343                 case Opt_hb_local:
1344                         mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1345                         break;
1346                 case Opt_hb_none:
1347                         mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1348                         break;
1349                 case Opt_hb_global:
1350                         mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1351                         break;
1352                 case Opt_barrier:
1353                         if (match_int(&args[0], &option)) {
1354                                 status = 0;
1355                                 goto bail;
1356                         }
1357                         if (option)
1358                                 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1359                         else
1360                                 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1361                         break;
1362                 case Opt_intr:
1363                         mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1364                         break;
1365                 case Opt_nointr:
1366                         mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1367                         break;
1368                 case Opt_err_panic:
1369                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1370                         break;
1371                 case Opt_err_ro:
1372                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1373                         break;
1374                 case Opt_data_ordered:
1375                         mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1376                         break;
1377                 case Opt_data_writeback:
1378                         mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1379                         break;
1380                 case Opt_user_xattr:
1381                         mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1382                         break;
1383                 case Opt_nouser_xattr:
1384                         mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1385                         break;
1386                 case Opt_atime_quantum:
1387                         if (match_int(&args[0], &option)) {
1388                                 status = 0;
1389                                 goto bail;
1390                         }
1391                         if (option >= 0)
1392                                 mopt->atime_quantum = option;
1393                         break;
1394                 case Opt_slot:
1395                         option = 0;
1396                         if (match_int(&args[0], &option)) {
1397                                 status = 0;
1398                                 goto bail;
1399                         }
1400                         if (option)
1401                                 mopt->slot = (s16)option;
1402                         break;
1403                 case Opt_commit:
1404                         option = 0;
1405                         if (match_int(&args[0], &option)) {
1406                                 status = 0;
1407                                 goto bail;
1408                         }
1409                         if (option < 0)
1410                                 return 0;
1411                         if (option == 0)
1412                                 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1413                         mopt->commit_interval = HZ * option;
1414                         break;
1415                 case Opt_localalloc:
1416                         option = 0;
1417                         if (match_int(&args[0], &option)) {
1418                                 status = 0;
1419                                 goto bail;
1420                         }
1421                         if (option >= 0)
1422                                 mopt->localalloc_opt = option;
1423                         break;
1424                 case Opt_localflocks:
1425                         /*
1426                          * Changing this during remount could race
1427                          * flock() requests, or "unbalance" existing
1428                          * ones (e.g., a lock is taken in one mode but
1429                          * dropped in the other). If users care enough
1430                          * to flip locking modes during remount, we
1431                          * could add a "local" flag to individual
1432                          * flock structures for proper tracking of
1433                          * state.
1434                          */
1435                         if (!is_remount)
1436                                 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1437                         break;
1438                 case Opt_stack:
1439                         /* Check both that the option we were passed
1440                          * is of the right length and that it is a proper
1441                          * string of the right length.
1442                          */
1443                         if (((args[0].to - args[0].from) !=
1444                              OCFS2_STACK_LABEL_LEN) ||
1445                             (strnlen(args[0].from,
1446                                      OCFS2_STACK_LABEL_LEN) !=
1447                              OCFS2_STACK_LABEL_LEN)) {
1448                                 mlog(ML_ERROR,
1449                                      "Invalid cluster_stack option\n");
1450                                 status = 0;
1451                                 goto bail;
1452                         }
1453                         memcpy(mopt->cluster_stack, args[0].from,
1454                                OCFS2_STACK_LABEL_LEN);
1455                         mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1456                         /*
1457                          * Open code the memcmp here as we don't have
1458                          * an osb to pass to
1459                          * ocfs2_userspace_stack().
1460                          */
1461                         if (memcmp(mopt->cluster_stack,
1462                                    OCFS2_CLASSIC_CLUSTER_STACK,
1463                                    OCFS2_STACK_LABEL_LEN))
1464                                 user_stack = 1;
1465                         break;
1466                 case Opt_inode64:
1467                         mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1468                         break;
1469                 case Opt_usrquota:
1470                         mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1471                         break;
1472                 case Opt_grpquota:
1473                         mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1474                         break;
1475                 case Opt_coherency_buffered:
1476                         mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1477                         break;
1478                 case Opt_coherency_full:
1479                         mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1480                         break;
1481                 case Opt_acl:
1482                         mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1483                         mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1484                         break;
1485                 case Opt_noacl:
1486                         mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1487                         mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1488                         break;
1489                 case Opt_resv_level:
1490                         if (is_remount)
1491                                 break;
1492                         if (match_int(&args[0], &option)) {
1493                                 status = 0;
1494                                 goto bail;
1495                         }
1496                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1497                             option < OCFS2_MAX_RESV_LEVEL)
1498                                 mopt->resv_level = option;
1499                         break;
1500                 case Opt_dir_resv_level:
1501                         if (is_remount)
1502                                 break;
1503                         if (match_int(&args[0], &option)) {
1504                                 status = 0;
1505                                 goto bail;
1506                         }
1507                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1508                             option < OCFS2_MAX_RESV_LEVEL)
1509                                 mopt->dir_resv_level = option;
1510                         break;
1511                 default:
1512                         mlog(ML_ERROR,
1513                              "Unrecognized mount option \"%s\" "
1514                              "or missing value\n", p);
1515                         status = 0;
1516                         goto bail;
1517                 }
1518         }
1519
1520         if (user_stack == 0) {
1521                 /* Ensure only one heartbeat mode */
1522                 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1523                                          OCFS2_MOUNT_HB_GLOBAL |
1524                                          OCFS2_MOUNT_HB_NONE);
1525                 if (hweight32(tmp) != 1) {
1526                         mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1527                         status = 0;
1528                         goto bail;
1529                 }
1530         }
1531
1532         status = 1;
1533
1534 bail:
1535         mlog_exit(status);
1536         return status;
1537 }
1538
1539 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1540 {
1541         struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb);
1542         unsigned long opts = osb->s_mount_opt;
1543         unsigned int local_alloc_megs;
1544
1545         if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1546                 seq_printf(s, ",_netdev");
1547                 if (opts & OCFS2_MOUNT_HB_LOCAL)
1548                         seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1549                 else
1550                         seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1551         } else
1552                 seq_printf(s, ",%s", OCFS2_HB_NONE);
1553
1554         if (opts & OCFS2_MOUNT_NOINTR)
1555                 seq_printf(s, ",nointr");
1556
1557         if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1558                 seq_printf(s, ",data=writeback");
1559         else
1560                 seq_printf(s, ",data=ordered");
1561
1562         if (opts & OCFS2_MOUNT_BARRIER)
1563                 seq_printf(s, ",barrier=1");
1564
1565         if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1566                 seq_printf(s, ",errors=panic");
1567         else
1568                 seq_printf(s, ",errors=remount-ro");
1569
1570         if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1571                 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1572
1573         if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM)
1574                 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1575
1576         if (osb->osb_commit_interval)
1577                 seq_printf(s, ",commit=%u",
1578                            (unsigned) (osb->osb_commit_interval / HZ));
1579
1580         local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1581         if (local_alloc_megs != ocfs2_la_default_mb(osb))
1582                 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1583
1584         if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1585                 seq_printf(s, ",localflocks,");
1586
1587         if (osb->osb_cluster_stack[0])
1588                 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1589                            osb->osb_cluster_stack);
1590         if (opts & OCFS2_MOUNT_USRQUOTA)
1591                 seq_printf(s, ",usrquota");
1592         if (opts & OCFS2_MOUNT_GRPQUOTA)
1593                 seq_printf(s, ",grpquota");
1594
1595         if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1596                 seq_printf(s, ",coherency=buffered");
1597         else
1598                 seq_printf(s, ",coherency=full");
1599
1600         if (opts & OCFS2_MOUNT_NOUSERXATTR)
1601                 seq_printf(s, ",nouser_xattr");
1602         else
1603                 seq_printf(s, ",user_xattr");
1604
1605         if (opts & OCFS2_MOUNT_INODE64)
1606                 seq_printf(s, ",inode64");
1607
1608         if (opts & OCFS2_MOUNT_POSIX_ACL)
1609                 seq_printf(s, ",acl");
1610         else
1611                 seq_printf(s, ",noacl");
1612
1613         if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1614                 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1615
1616         if (osb->osb_dir_resv_level != osb->osb_resv_level)
1617                 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1618
1619         return 0;
1620 }
1621
1622 static int __init ocfs2_init(void)
1623 {
1624         int status;
1625
1626         ocfs2_print_version();
1627
1628         status = init_ocfs2_uptodate_cache();
1629         if (status < 0) {
1630                 mlog_errno(status);
1631                 goto leave;
1632         }
1633
1634         status = ocfs2_initialize_mem_caches();
1635         if (status < 0) {
1636                 mlog_errno(status);
1637                 goto leave;
1638         }
1639
1640         ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1641         if (!ocfs2_wq) {
1642                 status = -ENOMEM;
1643                 goto leave;
1644         }
1645
1646         ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1647         if (!ocfs2_debugfs_root) {
1648                 status = -EFAULT;
1649                 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1650         }
1651
1652         status = ocfs2_quota_setup();
1653         if (status)
1654                 goto leave;
1655
1656         ocfs2_set_locking_protocol();
1657
1658         status = register_quota_format(&ocfs2_quota_format);
1659 leave:
1660         if (status < 0) {
1661                 ocfs2_quota_shutdown();
1662                 ocfs2_free_mem_caches();
1663                 exit_ocfs2_uptodate_cache();
1664         }
1665
1666         mlog_exit(status);
1667
1668         if (status >= 0) {
1669                 return register_filesystem(&ocfs2_fs_type);
1670         } else
1671                 return -1;
1672 }
1673
1674 static void __exit ocfs2_exit(void)
1675 {
1676         ocfs2_quota_shutdown();
1677
1678         if (ocfs2_wq) {
1679                 flush_workqueue(ocfs2_wq);
1680                 destroy_workqueue(ocfs2_wq);
1681         }
1682
1683         unregister_quota_format(&ocfs2_quota_format);
1684
1685         debugfs_remove(ocfs2_debugfs_root);
1686
1687         ocfs2_free_mem_caches();
1688
1689         unregister_filesystem(&ocfs2_fs_type);
1690
1691         exit_ocfs2_uptodate_cache();
1692
1693         mlog_exit_void();
1694 }
1695
1696 static void ocfs2_put_super(struct super_block *sb)
1697 {
1698         mlog(0, "(0x%p)\n", sb);
1699
1700         ocfs2_sync_blockdev(sb);
1701         ocfs2_dismount_volume(sb, 0);
1702
1703         mlog_exit_void();
1704 }
1705
1706 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1707 {
1708         struct ocfs2_super *osb;
1709         u32 numbits, freebits;
1710         int status;
1711         struct ocfs2_dinode *bm_lock;
1712         struct buffer_head *bh = NULL;
1713         struct inode *inode = NULL;
1714
1715         mlog(0, "(%p, %p)\n", dentry->d_sb, buf);
1716
1717         osb = OCFS2_SB(dentry->d_sb);
1718
1719         inode = ocfs2_get_system_file_inode(osb,
1720                                             GLOBAL_BITMAP_SYSTEM_INODE,
1721                                             OCFS2_INVALID_SLOT);
1722         if (!inode) {
1723                 mlog(ML_ERROR, "failed to get bitmap inode\n");
1724                 status = -EIO;
1725                 goto bail;
1726         }
1727
1728         status = ocfs2_inode_lock(inode, &bh, 0);
1729         if (status < 0) {
1730                 mlog_errno(status);
1731                 goto bail;
1732         }
1733
1734         bm_lock = (struct ocfs2_dinode *) bh->b_data;
1735
1736         numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1737         freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1738
1739         buf->f_type = OCFS2_SUPER_MAGIC;
1740         buf->f_bsize = dentry->d_sb->s_blocksize;
1741         buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1742         buf->f_blocks = ((sector_t) numbits) *
1743                         (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1744         buf->f_bfree = ((sector_t) freebits) *
1745                        (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1746         buf->f_bavail = buf->f_bfree;
1747         buf->f_files = numbits;
1748         buf->f_ffree = freebits;
1749         buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1750                                 & 0xFFFFFFFFUL;
1751         buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1752                                 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1753
1754         brelse(bh);
1755
1756         ocfs2_inode_unlock(inode, 0);
1757         status = 0;
1758 bail:
1759         if (inode)
1760                 iput(inode);
1761
1762         mlog_exit(status);
1763
1764         return status;
1765 }
1766
1767 static void ocfs2_inode_init_once(void *data)
1768 {
1769         struct ocfs2_inode_info *oi = data;
1770
1771         oi->ip_flags = 0;
1772         oi->ip_open_count = 0;
1773         spin_lock_init(&oi->ip_lock);
1774         ocfs2_extent_map_init(&oi->vfs_inode);
1775         INIT_LIST_HEAD(&oi->ip_io_markers);
1776         oi->ip_dir_start_lookup = 0;
1777
1778         init_rwsem(&oi->ip_alloc_sem);
1779         init_rwsem(&oi->ip_xattr_sem);
1780         mutex_init(&oi->ip_io_mutex);
1781
1782         oi->ip_blkno = 0ULL;
1783         oi->ip_clusters = 0;
1784
1785         ocfs2_resv_init_once(&oi->ip_la_data_resv);
1786
1787         ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1788         ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1789         ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1790
1791         ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1792                                   &ocfs2_inode_caching_ops);
1793
1794         inode_init_once(&oi->vfs_inode);
1795 }
1796
1797 static int ocfs2_initialize_mem_caches(void)
1798 {
1799         ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1800                                        sizeof(struct ocfs2_inode_info),
1801                                        0,
1802                                        (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1803                                                 SLAB_MEM_SPREAD),
1804                                        ocfs2_inode_init_once);
1805         ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1806                                         sizeof(struct ocfs2_dquot),
1807                                         0,
1808                                         (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1809                                                 SLAB_MEM_SPREAD),
1810                                         NULL);
1811         ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1812                                         sizeof(struct ocfs2_quota_chunk),
1813                                         0,
1814                                         (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1815                                         NULL);
1816         if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1817             !ocfs2_qf_chunk_cachep) {
1818                 if (ocfs2_inode_cachep)
1819                         kmem_cache_destroy(ocfs2_inode_cachep);
1820                 if (ocfs2_dquot_cachep)
1821                         kmem_cache_destroy(ocfs2_dquot_cachep);
1822                 if (ocfs2_qf_chunk_cachep)
1823                         kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1824                 return -ENOMEM;
1825         }
1826
1827         return 0;
1828 }
1829
1830 static void ocfs2_free_mem_caches(void)
1831 {
1832         if (ocfs2_inode_cachep)
1833                 kmem_cache_destroy(ocfs2_inode_cachep);
1834         ocfs2_inode_cachep = NULL;
1835
1836         if (ocfs2_dquot_cachep)
1837                 kmem_cache_destroy(ocfs2_dquot_cachep);
1838         ocfs2_dquot_cachep = NULL;
1839
1840         if (ocfs2_qf_chunk_cachep)
1841                 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1842         ocfs2_qf_chunk_cachep = NULL;
1843 }
1844
1845 static int ocfs2_get_sector(struct super_block *sb,
1846                             struct buffer_head **bh,
1847                             int block,
1848                             int sect_size)
1849 {
1850         if (!sb_set_blocksize(sb, sect_size)) {
1851                 mlog(ML_ERROR, "unable to set blocksize\n");
1852                 return -EIO;
1853         }
1854
1855         *bh = sb_getblk(sb, block);
1856         if (!*bh) {
1857                 mlog_errno(-EIO);
1858                 return -EIO;
1859         }
1860         lock_buffer(*bh);
1861         if (!buffer_dirty(*bh))
1862                 clear_buffer_uptodate(*bh);
1863         unlock_buffer(*bh);
1864         ll_rw_block(READ, 1, bh);
1865         wait_on_buffer(*bh);
1866         if (!buffer_uptodate(*bh)) {
1867                 mlog_errno(-EIO);
1868                 brelse(*bh);
1869                 *bh = NULL;
1870                 return -EIO;
1871         }
1872
1873         return 0;
1874 }
1875
1876 static int ocfs2_mount_volume(struct super_block *sb)
1877 {
1878         int status = 0;
1879         int unlock_super = 0;
1880         struct ocfs2_super *osb = OCFS2_SB(sb);
1881
1882         if (ocfs2_is_hard_readonly(osb))
1883                 goto leave;
1884
1885         status = ocfs2_dlm_init(osb);
1886         if (status < 0) {
1887                 mlog_errno(status);
1888                 goto leave;
1889         }
1890
1891         status = ocfs2_super_lock(osb, 1);
1892         if (status < 0) {
1893                 mlog_errno(status);
1894                 goto leave;
1895         }
1896         unlock_super = 1;
1897
1898         /* This will load up the node map and add ourselves to it. */
1899         status = ocfs2_find_slot(osb);
1900         if (status < 0) {
1901                 mlog_errno(status);
1902                 goto leave;
1903         }
1904
1905         /* load all node-local system inodes */
1906         status = ocfs2_init_local_system_inodes(osb);
1907         if (status < 0) {
1908                 mlog_errno(status);
1909                 goto leave;
1910         }
1911
1912         status = ocfs2_check_volume(osb);
1913         if (status < 0) {
1914                 mlog_errno(status);
1915                 goto leave;
1916         }
1917
1918         status = ocfs2_truncate_log_init(osb);
1919         if (status < 0)
1920                 mlog_errno(status);
1921
1922 leave:
1923         if (unlock_super)
1924                 ocfs2_super_unlock(osb, 1);
1925
1926         mlog_exit(status);
1927         return status;
1928 }
1929
1930 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1931 {
1932         int tmp, hangup_needed = 0;
1933         struct ocfs2_super *osb = NULL;
1934         char nodestr[8];
1935
1936         mlog(0, "(0x%p)\n", sb);
1937
1938         BUG_ON(!sb);
1939         osb = OCFS2_SB(sb);
1940         BUG_ON(!osb);
1941
1942         debugfs_remove(osb->osb_ctxt);
1943
1944         /*
1945          * Flush inode dropping work queue so that deletes are
1946          * performed while the filesystem is still working
1947          */
1948         ocfs2_drop_all_dl_inodes(osb);
1949
1950         /* Orphan scan should be stopped as early as possible */
1951         ocfs2_orphan_scan_stop(osb);
1952
1953         ocfs2_disable_quotas(osb);
1954
1955         ocfs2_shutdown_local_alloc(osb);
1956
1957         ocfs2_truncate_log_shutdown(osb);
1958
1959         /* This will disable recovery and flush any recovery work. */
1960         ocfs2_recovery_exit(osb);
1961
1962         ocfs2_journal_shutdown(osb);
1963
1964         ocfs2_sync_blockdev(sb);
1965
1966         ocfs2_purge_refcount_trees(osb);
1967
1968         /* No cluster connection means we've failed during mount, so skip
1969          * all the steps which depended on that to complete. */
1970         if (osb->cconn) {
1971                 tmp = ocfs2_super_lock(osb, 1);
1972                 if (tmp < 0) {
1973                         mlog_errno(tmp);
1974                         return;
1975                 }
1976         }
1977
1978         if (osb->slot_num != OCFS2_INVALID_SLOT)
1979                 ocfs2_put_slot(osb);
1980
1981         if (osb->cconn)
1982                 ocfs2_super_unlock(osb, 1);
1983
1984         ocfs2_release_system_inodes(osb);
1985
1986         /*
1987          * If we're dismounting due to mount error, mount.ocfs2 will clean
1988          * up heartbeat.  If we're a local mount, there is no heartbeat.
1989          * If we failed before we got a uuid_str yet, we can't stop
1990          * heartbeat.  Otherwise, do it.
1991          */
1992         if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str)
1993                 hangup_needed = 1;
1994
1995         if (osb->cconn)
1996                 ocfs2_dlm_shutdown(osb, hangup_needed);
1997
1998         ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1999         debugfs_remove(osb->osb_debug_root);
2000
2001         if (hangup_needed)
2002                 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
2003
2004         atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
2005
2006         if (ocfs2_mount_local(osb))
2007                 snprintf(nodestr, sizeof(nodestr), "local");
2008         else
2009                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
2010
2011         printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
2012                osb->dev_str, nodestr);
2013
2014         ocfs2_delete_osb(osb);
2015         kfree(osb);
2016         sb->s_dev = 0;
2017         sb->s_fs_info = NULL;
2018 }
2019
2020 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
2021                                 unsigned uuid_bytes)
2022 {
2023         int i, ret;
2024         char *ptr;
2025
2026         BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
2027
2028         osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
2029         if (osb->uuid_str == NULL)
2030                 return -ENOMEM;
2031
2032         for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
2033                 /* print with null */
2034                 ret = snprintf(ptr, 3, "%02X", uuid[i]);
2035                 if (ret != 2) /* drop super cleans up */
2036                         return -EINVAL;
2037                 /* then only advance past the last char */
2038                 ptr += 2;
2039         }
2040
2041         return 0;
2042 }
2043
2044 /* Make sure entire volume is addressable by our journal.  Requires
2045    osb_clusters_at_boot to be valid and for the journal to have been
2046    initialized by ocfs2_journal_init(). */
2047 static int ocfs2_journal_addressable(struct ocfs2_super *osb)
2048 {
2049         int status = 0;
2050         u64 max_block =
2051                 ocfs2_clusters_to_blocks(osb->sb,
2052                                          osb->osb_clusters_at_boot) - 1;
2053
2054         /* 32-bit block number is always OK. */
2055         if (max_block <= (u32)~0ULL)
2056                 goto out;
2057
2058         /* Volume is "huge", so see if our journal is new enough to
2059            support it. */
2060         if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2061                                        OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2062               jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2063                                                JBD2_FEATURE_INCOMPAT_64BIT))) {
2064                 mlog(ML_ERROR, "The journal cannot address the entire volume. "
2065                      "Enable the 'block64' journal option with tunefs.ocfs2");
2066                 status = -EFBIG;
2067                 goto out;
2068         }
2069
2070  out:
2071         return status;
2072 }
2073
2074 static int ocfs2_initialize_super(struct super_block *sb,
2075                                   struct buffer_head *bh,
2076                                   int sector_size,
2077                                   struct ocfs2_blockcheck_stats *stats)
2078 {
2079         int status;
2080         int i, cbits, bbits;
2081         struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2082         struct inode *inode = NULL;
2083         struct ocfs2_journal *journal;
2084         __le32 uuid_net_key;
2085         struct ocfs2_super *osb;
2086         u64 total_blocks;
2087
2088         osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2089         if (!osb) {
2090                 status = -ENOMEM;
2091                 mlog_errno(status);
2092                 goto bail;
2093         }
2094
2095         sb->s_fs_info = osb;
2096         sb->s_op = &ocfs2_sops;
2097         sb->s_d_op = &ocfs2_dentry_ops;
2098         sb->s_export_op = &ocfs2_export_ops;
2099         sb->s_qcop = &ocfs2_quotactl_ops;
2100         sb->dq_op = &ocfs2_quota_operations;
2101         sb->s_xattr = ocfs2_xattr_handlers;
2102         sb->s_time_gran = 1;
2103         sb->s_flags |= MS_NOATIME;
2104         /* this is needed to support O_LARGEFILE */
2105         cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2106         bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2107         sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2108
2109         osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2110
2111         for (i = 0; i < 3; i++)
2112                 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2113         osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2114
2115         osb->sb = sb;
2116         /* Save off for ocfs2_rw_direct */
2117         osb->s_sectsize_bits = blksize_bits(sector_size);
2118         BUG_ON(!osb->s_sectsize_bits);
2119
2120         spin_lock_init(&osb->dc_task_lock);
2121         init_waitqueue_head(&osb->dc_event);
2122         osb->dc_work_sequence = 0;
2123         osb->dc_wake_sequence = 0;
2124         INIT_LIST_HEAD(&osb->blocked_lock_list);
2125         osb->blocked_lock_count = 0;
2126         spin_lock_init(&osb->osb_lock);
2127         spin_lock_init(&osb->osb_xattr_lock);
2128         ocfs2_init_steal_slots(osb);
2129
2130         atomic_set(&osb->alloc_stats.moves, 0);
2131         atomic_set(&osb->alloc_stats.local_data, 0);
2132         atomic_set(&osb->alloc_stats.bitmap_data, 0);
2133         atomic_set(&osb->alloc_stats.bg_allocs, 0);
2134         atomic_set(&osb->alloc_stats.bg_extends, 0);
2135
2136         /* Copy the blockcheck stats from the superblock probe */
2137         osb->osb_ecc_stats = *stats;
2138
2139         ocfs2_init_node_maps(osb);
2140
2141         snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2142                  MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2143
2144         osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2145         if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2146                 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2147                      osb->max_slots);
2148                 status = -EINVAL;
2149                 goto bail;
2150         }
2151         mlog(0, "max_slots for this device: %u\n", osb->max_slots);
2152
2153         ocfs2_orphan_scan_init(osb);
2154
2155         status = ocfs2_recovery_init(osb);
2156         if (status) {
2157                 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2158                 mlog_errno(status);
2159                 goto bail;
2160         }
2161
2162         init_waitqueue_head(&osb->checkpoint_event);
2163         atomic_set(&osb->needs_checkpoint, 0);
2164
2165         osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2166
2167         osb->slot_num = OCFS2_INVALID_SLOT;
2168
2169         osb->s_xattr_inline_size = le16_to_cpu(
2170                                         di->id2.i_super.s_xattr_inline_size);
2171
2172         osb->local_alloc_state = OCFS2_LA_UNUSED;
2173         osb->local_alloc_bh = NULL;
2174         INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2175
2176         init_waitqueue_head(&osb->osb_mount_event);
2177
2178         status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2179         if (status) {
2180                 mlog_errno(status);
2181                 goto bail;
2182         }
2183
2184         osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2185         if (!osb->vol_label) {
2186                 mlog(ML_ERROR, "unable to alloc vol label\n");
2187                 status = -ENOMEM;
2188                 goto bail;
2189         }
2190
2191         osb->slot_recovery_generations =
2192                 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2193                         GFP_KERNEL);
2194         if (!osb->slot_recovery_generations) {
2195                 status = -ENOMEM;
2196                 mlog_errno(status);
2197                 goto bail;
2198         }
2199
2200         init_waitqueue_head(&osb->osb_wipe_event);
2201         osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2202                                         sizeof(*osb->osb_orphan_wipes),
2203                                         GFP_KERNEL);
2204         if (!osb->osb_orphan_wipes) {
2205                 status = -ENOMEM;
2206                 mlog_errno(status);
2207                 goto bail;
2208         }
2209
2210         osb->osb_rf_lock_tree = RB_ROOT;
2211
2212         osb->s_feature_compat =
2213                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2214         osb->s_feature_ro_compat =
2215                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2216         osb->s_feature_incompat =
2217                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2218
2219         if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2220                 mlog(ML_ERROR, "couldn't mount because of unsupported "
2221                      "optional features (%x).\n", i);
2222                 status = -EINVAL;
2223                 goto bail;
2224         }
2225         if (!(osb->sb->s_flags & MS_RDONLY) &&
2226             (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2227                 mlog(ML_ERROR, "couldn't mount RDWR because of "
2228                      "unsupported optional features (%x).\n", i);
2229                 status = -EINVAL;
2230                 goto bail;
2231         }
2232
2233         if (ocfs2_clusterinfo_valid(osb)) {
2234                 osb->osb_stackflags =
2235                         OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2236                 memcpy(osb->osb_cluster_stack,
2237                        OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2238                        OCFS2_STACK_LABEL_LEN);
2239                 osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
2240                 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2241                         mlog(ML_ERROR,
2242                              "couldn't mount because of an invalid "
2243                              "cluster stack label (%s) \n",
2244                              osb->osb_cluster_stack);
2245                         status = -EINVAL;
2246                         goto bail;
2247                 }
2248         } else {
2249                 /* The empty string is identical with classic tools that
2250                  * don't know about s_cluster_info. */
2251                 osb->osb_cluster_stack[0] = '\0';
2252         }
2253
2254         get_random_bytes(&osb->s_next_generation, sizeof(u32));
2255
2256         /* FIXME
2257          * This should be done in ocfs2_journal_init(), but unknown
2258          * ordering issues will cause the filesystem to crash.
2259          * If anyone wants to figure out what part of the code
2260          * refers to osb->journal before ocfs2_journal_init() is run,
2261          * be my guest.
2262          */
2263         /* initialize our journal structure */
2264
2265         journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2266         if (!journal) {
2267                 mlog(ML_ERROR, "unable to alloc journal\n");
2268                 status = -ENOMEM;
2269                 goto bail;
2270         }
2271         osb->journal = journal;
2272         journal->j_osb = osb;
2273
2274         atomic_set(&journal->j_num_trans, 0);
2275         init_rwsem(&journal->j_trans_barrier);
2276         init_waitqueue_head(&journal->j_checkpointed);
2277         spin_lock_init(&journal->j_lock);
2278         journal->j_trans_id = (unsigned long) 1;
2279         INIT_LIST_HEAD(&journal->j_la_cleanups);
2280         INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2281         journal->j_state = OCFS2_JOURNAL_FREE;
2282
2283         INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes);
2284         osb->dentry_lock_list = NULL;
2285
2286         /* get some pseudo constants for clustersize bits */
2287         osb->s_clustersize_bits =
2288                 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2289         osb->s_clustersize = 1 << osb->s_clustersize_bits;
2290         mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
2291
2292         if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2293             osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2294                 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2295                      osb->s_clustersize);
2296                 status = -EINVAL;
2297                 goto bail;
2298         }
2299
2300         total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2301                                                 le32_to_cpu(di->i_clusters));
2302
2303         status = generic_check_addressable(osb->sb->s_blocksize_bits,
2304                                            total_blocks);
2305         if (status) {
2306                 mlog(ML_ERROR, "Volume too large "
2307                      "to mount safely on this system");
2308                 status = -EFBIG;
2309                 goto bail;
2310         }
2311
2312         if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2313                                  sizeof(di->id2.i_super.s_uuid))) {
2314                 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2315                 status = -ENOMEM;
2316                 goto bail;
2317         }
2318
2319         memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
2320
2321         strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2322         osb->vol_label[63] = '\0';
2323         osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2324         osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2325         osb->first_cluster_group_blkno =
2326                 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2327         osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2328         osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2329         mlog(0, "vol_label: %s\n", osb->vol_label);
2330         mlog(0, "uuid: %s\n", osb->uuid_str);
2331         mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
2332              (unsigned long long)osb->root_blkno,
2333              (unsigned long long)osb->system_dir_blkno);
2334
2335         osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2336         if (!osb->osb_dlm_debug) {
2337                 status = -ENOMEM;
2338                 mlog_errno(status);
2339                 goto bail;
2340         }
2341
2342         atomic_set(&osb->vol_state, VOLUME_INIT);
2343
2344         /* load root, system_dir, and all global system inodes */
2345         status = ocfs2_init_global_system_inodes(osb);
2346         if (status < 0) {
2347                 mlog_errno(status);
2348                 goto bail;
2349         }
2350
2351         /*
2352          * global bitmap
2353          */
2354         inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2355                                             OCFS2_INVALID_SLOT);
2356         if (!inode) {
2357                 status = -EINVAL;
2358                 mlog_errno(status);
2359                 goto bail;
2360         }
2361
2362         osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2363         osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2364         iput(inode);
2365
2366         osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2367                                  osb->s_feature_incompat) * 8;
2368
2369         status = ocfs2_init_slot_info(osb);
2370         if (status < 0) {
2371                 mlog_errno(status);
2372                 goto bail;
2373         }
2374
2375 bail:
2376         mlog_exit(status);
2377         return status;
2378 }
2379
2380 /*
2381  * will return: -EAGAIN if it is ok to keep searching for superblocks
2382  *              -EINVAL if there is a bad superblock
2383  *              0 on success
2384  */
2385 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2386                                struct buffer_head *bh,
2387                                u32 blksz,
2388                                struct ocfs2_blockcheck_stats *stats)
2389 {
2390         int status = -EAGAIN;
2391
2392         if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2393                    strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2394                 /* We have to do a raw check of the feature here */
2395                 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2396                     OCFS2_FEATURE_INCOMPAT_META_ECC) {
2397                         status = ocfs2_block_check_validate(bh->b_data,
2398                                                             bh->b_size,
2399                                                             &di->i_check,
2400                                                             stats);
2401                         if (status)
2402                                 goto out;
2403                 }
2404                 status = -EINVAL;
2405                 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2406                         mlog(ML_ERROR, "found superblock with incorrect block "
2407                              "size: found %u, should be %u\n",
2408                              1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2409                                blksz);
2410                 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2411                            OCFS2_MAJOR_REV_LEVEL ||
2412                            le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2413                            OCFS2_MINOR_REV_LEVEL) {
2414                         mlog(ML_ERROR, "found superblock with bad version: "
2415                              "found %u.%u, should be %u.%u\n",
2416                              le16_to_cpu(di->id2.i_super.s_major_rev_level),
2417                              le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2418                              OCFS2_MAJOR_REV_LEVEL,
2419                              OCFS2_MINOR_REV_LEVEL);
2420                 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2421                         mlog(ML_ERROR, "bad block number on superblock: "
2422                              "found %llu, should be %llu\n",
2423                              (unsigned long long)le64_to_cpu(di->i_blkno),
2424                              (unsigned long long)bh->b_blocknr);
2425                 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2426                             le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2427                         mlog(ML_ERROR, "bad cluster size found: %u\n",
2428                              1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2429                 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2430                         mlog(ML_ERROR, "bad root_blkno: 0\n");
2431                 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2432                         mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2433                 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2434                         mlog(ML_ERROR,
2435                              "Superblock slots found greater than file system "
2436                              "maximum: found %u, max %u\n",
2437                              le16_to_cpu(di->id2.i_super.s_max_slots),
2438                              OCFS2_MAX_SLOTS);
2439                 } else {
2440                         /* found it! */
2441                         status = 0;
2442                 }
2443         }
2444
2445 out:
2446         mlog_exit(status);
2447         return status;
2448 }
2449
2450 static int ocfs2_check_volume(struct ocfs2_super *osb)
2451 {
2452         int status;
2453         int dirty;
2454         int local;
2455         struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2456                                                   * recover
2457                                                   * ourselves. */
2458
2459         /* Init our journal object. */
2460         status = ocfs2_journal_init(osb->journal, &dirty);
2461         if (status < 0) {
2462                 mlog(ML_ERROR, "Could not initialize journal!\n");
2463                 goto finally;
2464         }
2465
2466         /* Now that journal has been initialized, check to make sure
2467            entire volume is addressable. */
2468         status = ocfs2_journal_addressable(osb);
2469         if (status)
2470                 goto finally;
2471
2472         /* If the journal was unmounted cleanly then we don't want to
2473          * recover anything. Otherwise, journal_load will do that
2474          * dirty work for us :) */
2475         if (!dirty) {
2476                 status = ocfs2_journal_wipe(osb->journal, 0);
2477                 if (status < 0) {
2478                         mlog_errno(status);
2479                         goto finally;
2480                 }
2481         } else {
2482                 mlog(ML_NOTICE, "File system was not unmounted cleanly, "
2483                      "recovering volume.\n");
2484         }
2485
2486         local = ocfs2_mount_local(osb);
2487
2488         /* will play back anything left in the journal. */
2489         status = ocfs2_journal_load(osb->journal, local, dirty);
2490         if (status < 0) {
2491                 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2492                 goto finally;
2493         }
2494
2495         if (dirty) {
2496                 /* recover my local alloc if we didn't unmount cleanly. */
2497                 status = ocfs2_begin_local_alloc_recovery(osb,
2498                                                           osb->slot_num,
2499                                                           &local_alloc);
2500                 if (status < 0) {
2501                         mlog_errno(status);
2502                         goto finally;
2503                 }
2504                 /* we complete the recovery process after we've marked
2505                  * ourselves as mounted. */
2506         }
2507
2508         mlog(0, "Journal loaded.\n");
2509
2510         status = ocfs2_load_local_alloc(osb);
2511         if (status < 0) {
2512                 mlog_errno(status);
2513                 goto finally;
2514         }
2515
2516         if (dirty) {
2517                 /* Recovery will be completed after we've mounted the
2518                  * rest of the volume. */
2519                 osb->dirty = 1;
2520                 osb->local_alloc_copy = local_alloc;
2521                 local_alloc = NULL;
2522         }
2523
2524         /* go through each journal, trylock it and if you get the
2525          * lock, and it's marked as dirty, set the bit in the recover
2526          * map and launch a recovery thread for it. */
2527         status = ocfs2_mark_dead_nodes(osb);
2528         if (status < 0) {
2529                 mlog_errno(status);
2530                 goto finally;
2531         }
2532
2533         status = ocfs2_compute_replay_slots(osb);
2534         if (status < 0)
2535                 mlog_errno(status);
2536
2537 finally:
2538         if (local_alloc)
2539                 kfree(local_alloc);
2540
2541         mlog_exit(status);
2542         return status;
2543 }
2544
2545 /*
2546  * The routine gets called from dismount or close whenever a dismount on
2547  * volume is requested and the osb open count becomes 1.
2548  * It will remove the osb from the global list and also free up all the
2549  * initialized resources and fileobject.
2550  */
2551 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2552 {
2553         /* This function assumes that the caller has the main osb resource */
2554
2555         ocfs2_free_slot_info(osb);
2556
2557         kfree(osb->osb_orphan_wipes);
2558         kfree(osb->slot_recovery_generations);
2559         /* FIXME
2560          * This belongs in journal shutdown, but because we have to
2561          * allocate osb->journal at the start of ocfs2_initialize_osb(),
2562          * we free it here.
2563          */
2564         kfree(osb->journal);
2565         if (osb->local_alloc_copy)
2566                 kfree(osb->local_alloc_copy);
2567         kfree(osb->uuid_str);
2568         ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2569         memset(osb, 0, sizeof(struct ocfs2_super));
2570
2571         mlog_exit_void();
2572 }
2573
2574 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2575  * panic(). We do not support continue-on-error operation. */
2576 static void ocfs2_handle_error(struct super_block *sb)
2577 {
2578         struct ocfs2_super *osb = OCFS2_SB(sb);
2579
2580         if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2581                 panic("OCFS2: (device %s): panic forced after error\n",
2582                       sb->s_id);
2583
2584         ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2585
2586         if (sb->s_flags & MS_RDONLY &&
2587             (ocfs2_is_soft_readonly(osb) ||
2588              ocfs2_is_hard_readonly(osb)))
2589                 return;
2590
2591         printk(KERN_CRIT "File system is now read-only due to the potential "
2592                "of on-disk corruption. Please run fsck.ocfs2 once the file "
2593                "system is unmounted.\n");
2594         sb->s_flags |= MS_RDONLY;
2595         ocfs2_set_ro_flag(osb, 0);
2596 }
2597
2598 static char error_buf[1024];
2599
2600 void __ocfs2_error(struct super_block *sb,
2601                    const char *function,
2602                    const char *fmt, ...)
2603 {
2604         va_list args;
2605
2606         va_start(args, fmt);
2607         vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2608         va_end(args);
2609
2610         /* Not using mlog here because we want to show the actual
2611          * function the error came from. */
2612         printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2613                sb->s_id, function, error_buf);
2614
2615         ocfs2_handle_error(sb);
2616 }
2617
2618 /* Handle critical errors. This is intentionally more drastic than
2619  * ocfs2_handle_error, so we only use for things like journal errors,
2620  * etc. */
2621 void __ocfs2_abort(struct super_block* sb,
2622                    const char *function,
2623                    const char *fmt, ...)
2624 {
2625         va_list args;
2626
2627         va_start(args, fmt);
2628         vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2629         va_end(args);
2630
2631         printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2632                sb->s_id, function, error_buf);
2633
2634         /* We don't have the cluster support yet to go straight to
2635          * hard readonly in here. Until then, we want to keep
2636          * ocfs2_abort() so that we can at least mark critical
2637          * errors.
2638          *
2639          * TODO: This should abort the journal and alert other nodes
2640          * that our slot needs recovery. */
2641
2642         /* Force a panic(). This stinks, but it's better than letting
2643          * things continue without having a proper hard readonly
2644          * here. */
2645         if (!ocfs2_mount_local(OCFS2_SB(sb)))
2646                 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2647         ocfs2_handle_error(sb);
2648 }
2649
2650 /*
2651  * Void signal blockers, because in-kernel sigprocmask() only fails
2652  * when SIG_* is wrong.
2653  */
2654 void ocfs2_block_signals(sigset_t *oldset)
2655 {
2656         int rc;
2657         sigset_t blocked;
2658
2659         sigfillset(&blocked);
2660         rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2661         BUG_ON(rc);
2662 }
2663
2664 void ocfs2_unblock_signals(sigset_t *oldset)
2665 {
2666         int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2667         BUG_ON(rc);
2668 }
2669
2670 module_init(ocfs2_init);
2671 module_exit(ocfs2_exit);