fs/ocfs2/super.c

   1 /* -*- mode: c; c-basic-offset: 8; -*-
   2  * vim: noexpandtab sw=8 ts=8 sts=0:
   3  *
   4  * super.c
   5  *
   6  * load/unload driver, mount/dismount volumes
   7  *
   8  * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
   9  *
  10  * This program is free software; you can redistribute it and/or
  11  * modify it under the terms of the GNU General Public
  12  * License as published by the Free Software Foundation; either
  13  * version 2 of the License, or (at your option) any later version.
  14  *
  15  * This program is distributed in the hope that it will be useful,
  16  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  17  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  18  * General Public License for more details.
  19  *
  20  * You should have received a copy of the GNU General Public
  21  * License along with this program; if not, write to the
  22  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  23  * Boston, MA 021110-1307, USA.
  24  */
  25
  26 #include <linux/module.h>
  27 #include <linux/fs.h>
  28 #include <linux/types.h>
  29 #include <linux/slab.h>
  30 #include <linux/highmem.h>
  31 #include <linux/init.h>
  32 #include <linux/random.h>
  33 #include <linux/statfs.h>
  34 #include <linux/moduleparam.h>
  35 #include <linux/blkdev.h>
  36 #include <linux/socket.h>
  37 #include <linux/inet.h>
  38 #include <linux/parser.h>
  39 #include <linux/crc32.h>
  40 #include <linux/debugfs.h>
  41 #include <linux/mount.h>
  42 #include <linux/seq_file.h>
  43 #include <linux/quotaops.h>
  44 #include <linux/smp_lock.h>
  45
  46 #define MLOG_MASK_PREFIX ML_SUPER
  47 #include <cluster/masklog.h>
  48
  49 #include "ocfs2.h"
  50
  51 /* this should be the only file to include a version 1 header */
  52 #include "ocfs1_fs_compat.h"
  53
  54 #include "alloc.h"
  55 #include "blockcheck.h"
  56 #include "dlmglue.h"
  57 #include "export.h"
  58 #include "extent_map.h"
  59 #include "heartbeat.h"
  60 #include "inode.h"
  61 #include "journal.h"
  62 #include "localalloc.h"
  63 #include "namei.h"
  64 #include "slot_map.h"
  65 #include "super.h"
  66 #include "sysfile.h"
  67 #include "uptodate.h"
  68 #include "ver.h"
  69 #include "xattr.h"
  70 #include "quota.h"
  71 #include "refcounttree.h"
  72 #include "suballoc.h"
  73
  74 #include "buffer_head_io.h"
  75
  76 static struct kmem_cache *ocfs2_inode_cachep = NULL;
  77 struct kmem_cache *ocfs2_dquot_cachep;
  78 struct kmem_cache *ocfs2_qf_chunk_cachep;
  79
  80 /* OCFS2 needs to schedule several differnt types of work which
  81  * require cluster locking, disk I/O, recovery waits, etc. Since these
  82  * types of work tend to be heavy we avoid using the kernel events
  83  * workqueue and schedule on our own. */
  84 struct workqueue_struct *ocfs2_wq = NULL;
  85
  86 static struct dentry *ocfs2_debugfs_root = NULL;
  87
  88 MODULE_AUTHOR("Oracle");
  89 MODULE_LICENSE("GPL");
  90
  91 struct mount_options
  92 {
  93         unsigned long   commit_interval;
  94         unsigned long   mount_opt;
  95         unsigned int    atime_quantum;
  96         signed short    slot;
  97         int             localalloc_opt;
  98         unsigned int    resv_level;
  99         int             dir_resv_level;
 100         char            cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
 101 };
 102
 103 static int ocfs2_parse_options(struct super_block *sb, char *options,
 104                                struct mount_options *mopt,
 105                                int is_remount);
 106 static int ocfs2_check_set_options(struct super_block *sb,
 107                                    struct mount_options *options);
 108 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt);
 109 static void ocfs2_put_super(struct super_block *sb);
 110 static int ocfs2_mount_volume(struct super_block *sb);
 111 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
 112 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
 113 static int ocfs2_initialize_mem_caches(void);
 114 static void ocfs2_free_mem_caches(void);
 115 static void ocfs2_delete_osb(struct ocfs2_super *osb);
 116
 117 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
 118
 119 static int ocfs2_sync_fs(struct super_block *sb, int wait);
 120
 121 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
 122 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
 123 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
 124 static int ocfs2_check_volume(struct ocfs2_super *osb);
 125 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
 126                                struct buffer_head *bh,
 127                                u32 sectsize,
 128                                struct ocfs2_blockcheck_stats *stats);
 129 static int ocfs2_initialize_super(struct super_block *sb,
 130                                   struct buffer_head *bh,
 131                                   int sector_size,
 132                                   struct ocfs2_blockcheck_stats *stats);
 133 static int ocfs2_get_sector(struct super_block *sb,
 134                             struct buffer_head **bh,
 135                             int block,
 136                             int sect_size);
 137 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
 138 static void ocfs2_destroy_inode(struct inode *inode);
 139 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
 140 static int ocfs2_enable_quotas(struct ocfs2_super *osb);
 141 static void ocfs2_disable_quotas(struct ocfs2_super *osb);
 142
 143 static const struct super_operations ocfs2_sops = {
 144         .statfs         = ocfs2_statfs,
 145         .alloc_inode    = ocfs2_alloc_inode,
 146         .destroy_inode  = ocfs2_destroy_inode,
 147         .drop_inode     = ocfs2_drop_inode,
 148         .clear_inode    = ocfs2_clear_inode,
 149         .delete_inode   = ocfs2_delete_inode,
 150         .sync_fs        = ocfs2_sync_fs,
 151         .put_super      = ocfs2_put_super,
 152         .remount_fs     = ocfs2_remount,
 153         .show_options   = ocfs2_show_options,
 154         .quota_read     = ocfs2_quota_read,
 155         .quota_write    = ocfs2_quota_write,
 156 };
 157
 158 enum {
 159         Opt_barrier,
 160         Opt_err_panic,
 161         Opt_err_ro,
 162         Opt_intr,
 163         Opt_nointr,
 164         Opt_hb_none,
 165         Opt_hb_local,
 166         Opt_data_ordered,
 167         Opt_data_writeback,
 168         Opt_atime_quantum,
 169         Opt_slot,
 170         Opt_commit,
 171         Opt_localalloc,
 172         Opt_localflocks,
 173         Opt_stack,
 174         Opt_user_xattr,
 175         Opt_nouser_xattr,
 176         Opt_inode64,
 177         Opt_acl,
 178         Opt_noacl,
 179         Opt_usrquota,
 180         Opt_grpquota,
 181         Opt_resv_level,
 182         Opt_dir_resv_level,
 183         Opt_err,
 184 };
 185
 186 static const match_table_t tokens = {
 187         {Opt_barrier, "barrier=%u"},
 188         {Opt_err_panic, "errors=panic"},
 189         {Opt_err_ro, "errors=remount-ro"},
 190         {Opt_intr, "intr"},
 191         {Opt_nointr, "nointr"},
 192         {Opt_hb_none, OCFS2_HB_NONE},
 193         {Opt_hb_local, OCFS2_HB_LOCAL},
 194         {Opt_data_ordered, "data=ordered"},
 195         {Opt_data_writeback, "data=writeback"},
 196         {Opt_atime_quantum, "atime_quantum=%u"},
 197         {Opt_slot, "preferred_slot=%u"},
 198         {Opt_commit, "commit=%u"},
 199         {Opt_localalloc, "localalloc=%d"},
 200         {Opt_localflocks, "localflocks"},
 201         {Opt_stack, "cluster_stack=%s"},
 202         {Opt_user_xattr, "user_xattr"},
 203         {Opt_nouser_xattr, "nouser_xattr"},
 204         {Opt_inode64, "inode64"},
 205         {Opt_acl, "acl"},
 206         {Opt_noacl, "noacl"},
 207         {Opt_usrquota, "usrquota"},
 208         {Opt_grpquota, "grpquota"},
 209         {Opt_resv_level, "resv_level=%u"},
 210         {Opt_dir_resv_level, "dir_resv_level=%u"},
 211         {Opt_err, NULL}
 212 };
 213
 214 #ifdef CONFIG_DEBUG_FS
 215 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
 216 {
 217         struct ocfs2_cluster_connection *cconn = osb->cconn;
 218         struct ocfs2_recovery_map *rm = osb->recovery_map;
 219         struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
 220         int i, out = 0;
 221
 222         out += snprintf(buf + out, len - out,
 223                         "%10s => Id: %-s  Uuid: %-s  Gen: 0x%X  Label: %-s\n",
 224                         "Device", osb->dev_str, osb->uuid_str,
 225                         osb->fs_generation, osb->vol_label);
 226
 227         out += snprintf(buf + out, len - out,
 228                         "%10s => State: %d  Flags: 0x%lX\n", "Volume",
 229                         atomic_read(&osb->vol_state), osb->osb_flags);
 230
 231         out += snprintf(buf + out, len - out,
 232                         "%10s => Block: %lu  Cluster: %d\n", "Sizes",
 233                         osb->sb->s_blocksize, osb->s_clustersize);
 234
 235         out += snprintf(buf + out, len - out,
 236                         "%10s => Compat: 0x%X  Incompat: 0x%X  "
 237                         "ROcompat: 0x%X\n",
 238                         "Features", osb->s_feature_compat,
 239                         osb->s_feature_incompat, osb->s_feature_ro_compat);
 240
 241         out += snprintf(buf + out, len - out,
 242                         "%10s => Opts: 0x%lX  AtimeQuanta: %u\n", "Mount",
 243                         osb->s_mount_opt, osb->s_atime_quantum);
 244
 245         if (cconn) {
 246                 out += snprintf(buf + out, len - out,
 247                                 "%10s => Stack: %s  Name: %*s  "
 248                                 "Version: %d.%d\n", "Cluster",
 249                                 (*osb->osb_cluster_stack == '\0' ?
 250                                  "o2cb" : osb->osb_cluster_stack),
 251                                 cconn->cc_namelen, cconn->cc_name,
 252                                 cconn->cc_version.pv_major,
 253                                 cconn->cc_version.pv_minor);
 254         }
 255
 256         spin_lock(&osb->dc_task_lock);
 257         out += snprintf(buf + out, len - out,
 258                         "%10s => Pid: %d  Count: %lu  WakeSeq: %lu  "
 259                         "WorkSeq: %lu\n", "DownCnvt",
 260                         (osb->dc_task ?  task_pid_nr(osb->dc_task) : -1),
 261                         osb->blocked_lock_count, osb->dc_wake_sequence,
 262                         osb->dc_work_sequence);
 263         spin_unlock(&osb->dc_task_lock);
 264
 265         spin_lock(&osb->osb_lock);
 266         out += snprintf(buf + out, len - out, "%10s => Pid: %d  Nodes:",
 267                         "Recovery",
 268                         (osb->recovery_thread_task ?
 269                          task_pid_nr(osb->recovery_thread_task) : -1));
 270         if (rm->rm_used == 0)
 271                 out += snprintf(buf + out, len - out, " None\n");
 272         else {
 273                 for (i = 0; i < rm->rm_used; i++)
 274                         out += snprintf(buf + out, len - out, " %d",
 275                                         rm->rm_entries[i]);
 276                 out += snprintf(buf + out, len - out, "\n");
 277         }
 278         spin_unlock(&osb->osb_lock);
 279
 280         out += snprintf(buf + out, len - out,
 281                         "%10s => Pid: %d  Interval: %lu  Needs: %d\n", "Commit",
 282                         (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
 283                         osb->osb_commit_interval,
 284                         atomic_read(&osb->needs_checkpoint));
 285
 286         out += snprintf(buf + out, len - out,
 287                         "%10s => State: %d  TxnId: %lu  NumTxns: %d\n",
 288                         "Journal", osb->journal->j_state,
 289                         osb->journal->j_trans_id,
 290                         atomic_read(&osb->journal->j_num_trans));
 291
 292         out += snprintf(buf + out, len - out,
 293                         "%10s => GlobalAllocs: %d  LocalAllocs: %d  "
 294                         "SubAllocs: %d  LAWinMoves: %d  SAExtends: %d\n",
 295                         "Stats",
 296                         atomic_read(&osb->alloc_stats.bitmap_data),
 297                         atomic_read(&osb->alloc_stats.local_data),
 298                         atomic_read(&osb->alloc_stats.bg_allocs),
 299                         atomic_read(&osb->alloc_stats.moves),
 300                         atomic_read(&osb->alloc_stats.bg_extends));
 301
 302         out += snprintf(buf + out, len - out,
 303                         "%10s => State: %u  Descriptor: %llu  Size: %u bits  "
 304                         "Default: %u bits\n",
 305                         "LocalAlloc", osb->local_alloc_state,
 306                         (unsigned long long)osb->la_last_gd,
 307                         osb->local_alloc_bits, osb->local_alloc_default_bits);
 308
 309         spin_lock(&osb->osb_lock);
 310         out += snprintf(buf + out, len - out,
 311                         "%10s => InodeSlot: %d  StolenInodes: %d, "
 312                         "MetaSlot: %d  StolenMeta: %d\n", "Steal",
 313                         osb->s_inode_steal_slot,
 314                         atomic_read(&osb->s_num_inodes_stolen),
 315                         osb->s_meta_steal_slot,
 316                         atomic_read(&osb->s_num_meta_stolen));
 317         spin_unlock(&osb->osb_lock);
 318
 319         out += snprintf(buf + out, len - out, "OrphanScan => ");
 320         out += snprintf(buf + out, len - out, "Local: %u  Global: %u ",
 321                         os->os_count, os->os_seqno);
 322         out += snprintf(buf + out, len - out, " Last Scan: ");
 323         if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
 324                 out += snprintf(buf + out, len - out, "Disabled\n");
 325         else
 326                 out += snprintf(buf + out, len - out, "%lu seconds ago\n",
 327                                 (get_seconds() - os->os_scantime.tv_sec));
 328
 329         out += snprintf(buf + out, len - out, "%10s => %3s  %10s\n",
 330                         "Slots", "Num", "RecoGen");
 331         for (i = 0; i < osb->max_slots; ++i) {
 332                 out += snprintf(buf + out, len - out,
 333                                 "%10s  %c %3d  %10d\n",
 334                                 " ",
 335                                 (i == osb->slot_num ? '*' : ' '),
 336                                 i, osb->slot_recovery_generations[i]);
 337         }
 338
 339         return out;
 340 }
 341
 342 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 343 {
 344         struct ocfs2_super *osb = inode->i_private;
 345         char *buf = NULL;
 346
 347         buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
 348         if (!buf)
 349                 goto bail;
 350
 351         i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
 352
 353         file->private_data = buf;
 354
 355         return 0;
 356 bail:
 357         return -ENOMEM;
 358 }
 359
 360 static int ocfs2_debug_release(struct inode *inode, struct file *file)
 361 {
 362         kfree(file->private_data);
 363         return 0;
 364 }
 365
 366 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 367                                 size_t nbytes, loff_t *ppos)
 368 {
 369         return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
 370                                        i_size_read(file->f_mapping->host));
 371 }
 372 #else
 373 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 374 {
 375         return 0;
 376 }
 377 static int ocfs2_debug_release(struct inode *inode, struct file *file)
 378 {
 379         return 0;
 380 }
 381 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 382                                 size_t nbytes, loff_t *ppos)
 383 {
 384         return 0;
 385 }
 386 #endif  /* CONFIG_DEBUG_FS */
 387
 388 static const struct file_operations ocfs2_osb_debug_fops = {
 389         .open =         ocfs2_osb_debug_open,
 390         .release =      ocfs2_debug_release,
 391         .read =         ocfs2_debug_read,
 392         .llseek =       generic_file_llseek,
 393 };
 394
 395 static int ocfs2_sync_fs(struct super_block *sb, int wait)
 396 {
 397         int status;
 398         tid_t target;
 399         struct ocfs2_super *osb = OCFS2_SB(sb);
 400
 401         if (ocfs2_is_hard_readonly(osb))
 402                 return -EROFS;
 403
 404         if (wait) {
 405                 status = ocfs2_flush_truncate_log(osb);
 406                 if (status < 0)
 407                         mlog_errno(status);
 408         } else {
 409                 ocfs2_schedule_truncate_log_flush(osb, 0);
 410         }
 411
 412         if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
 413                                       &target)) {
 414                 if (wait)
 415                         jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
 416                                              target);
 417         }
 418         return 0;
 419 }
 420
 421 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
 422 {
 423         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
 424             && (ino == USER_QUOTA_SYSTEM_INODE
 425                 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
 426                 return 0;
 427         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
 428             && (ino == GROUP_QUOTA_SYSTEM_INODE
 429                 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
 430                 return 0;
 431         return 1;
 432 }
 433
 434 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
 435 {
 436         struct inode *new = NULL;
 437         int status = 0;
 438         int i;
 439
 440         mlog_entry_void();
 441
 442         new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 443         if (IS_ERR(new)) {
 444                 status = PTR_ERR(new);
 445                 mlog_errno(status);
 446                 goto bail;
 447         }
 448         osb->root_inode = new;
 449
 450         new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 451         if (IS_ERR(new)) {
 452                 status = PTR_ERR(new);
 453                 mlog_errno(status);
 454                 goto bail;
 455         }
 456         osb->sys_root_inode = new;
 457
 458         for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
 459              i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
 460                 if (!ocfs2_need_system_inode(osb, i))
 461                         continue;
 462                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 463                 if (!new) {
 464                         ocfs2_release_system_inodes(osb);
 465                         status = -EINVAL;
 466                         mlog_errno(status);
 467                         /* FIXME: Should ERROR_RO_FS */
 468                         mlog(ML_ERROR, "Unable to load system inode %d, "
 469                              "possibly corrupt fs?", i);
 470                         goto bail;
 471                 }
 472                 // the array now has one ref, so drop this one
 473                 iput(new);
 474         }
 475
 476 bail:
 477         mlog_exit(status);
 478         return status;
 479 }
 480
 481 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
 482 {
 483         struct inode *new = NULL;
 484         int status = 0;
 485         int i;
 486
 487         mlog_entry_void();
 488
 489         for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
 490              i < NUM_SYSTEM_INODES;
 491              i++) {
 492                 if (!ocfs2_need_system_inode(osb, i))
 493                         continue;
 494                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 495                 if (!new) {
 496                         ocfs2_release_system_inodes(osb);
 497                         status = -EINVAL;
 498                         mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
 499                              status, i, osb->slot_num);
 500                         goto bail;
 501                 }
 502                 /* the array now has one ref, so drop this one */
 503                 iput(new);
 504         }
 505
 506 bail:
 507         mlog_exit(status);
 508         return status;
 509 }
 510
 511 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
 512 {
 513         int i;
 514         struct inode *inode;
 515
 516         mlog_entry_void();
 517
 518         for (i = 0; i < NUM_SYSTEM_INODES; i++) {
 519                 inode = osb->system_inodes[i];
 520                 if (inode) {
 521                         iput(inode);
 522                         osb->system_inodes[i] = NULL;
 523                 }
 524         }
 525
 526         inode = osb->sys_root_inode;
 527         if (inode) {
 528                 iput(inode);
 529                 osb->sys_root_inode = NULL;
 530         }
 531
 532         inode = osb->root_inode;
 533         if (inode) {
 534                 iput(inode);
 535                 osb->root_inode = NULL;
 536         }
 537
 538         mlog_exit(0);
 539 }
 540
 541 /* We're allocating fs objects, use GFP_NOFS */
 542 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
 543 {
 544         struct ocfs2_inode_info *oi;
 545
 546         oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
 547         if (!oi)
 548                 return NULL;
 549
 550         jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
 551         return &oi->vfs_inode;
 552 }
 553
 554 static void ocfs2_destroy_inode(struct inode *inode)
 555 {
 556         kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
 557 }
 558
 559 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
 560                                                 unsigned int cbits)
 561 {
 562         unsigned int bytes = 1 << cbits;
 563         unsigned int trim = bytes;
 564         unsigned int bitshift = 32;
 565
 566         /*
 567          * i_size and all block offsets in ocfs2 are always 64 bits
 568          * wide. i_clusters is 32 bits, in cluster-sized units. So on
 569          * 64 bit platforms, cluster size will be the limiting factor.
 570          */
 571
 572 #if BITS_PER_LONG == 32
 573 # if defined(CONFIG_LBDAF)
 574         BUILD_BUG_ON(sizeof(sector_t) != 8);
 575         /*
 576          * We might be limited by page cache size.
 577          */
 578         if (bytes > PAGE_CACHE_SIZE) {
 579                 bytes = PAGE_CACHE_SIZE;
 580                 trim = 1;
 581                 /*
 582                  * Shift by 31 here so that we don't get larger than
 583                  * MAX_LFS_FILESIZE
 584                  */
 585                 bitshift = 31;
 586         }
 587 # else
 588         /*
 589          * We are limited by the size of sector_t. Use block size, as
 590          * that's what we expose to the VFS.
 591          */
 592         bytes = 1 << bbits;
 593         trim = 1;
 594         bitshift = 31;
 595 # endif
 596 #endif
 597
 598         /*
 599          * Trim by a whole cluster when we can actually approach the
 600          * on-disk limits. Otherwise we can overflow i_clusters when
 601          * an extent start is at the max offset.
 602          */
 603         return (((unsigned long long)bytes) << bitshift) - trim;
 604 }
 605
 606 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
 607 {
 608         int incompat_features;
 609         int ret = 0;
 610         struct mount_options parsed_options;
 611         struct ocfs2_super *osb = OCFS2_SB(sb);
 612
 613         lock_kernel();
 614
 615         if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
 616             !ocfs2_check_set_options(sb, &parsed_options)) {
 617                 ret = -EINVAL;
 618                 goto out;
 619         }
 620
 621         if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) !=
 622             (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
 623                 ret = -EINVAL;
 624                 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
 625                 goto out;
 626         }
 627
 628         if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
 629             (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
 630                 ret = -EINVAL;
 631                 mlog(ML_ERROR, "Cannot change data mode on remount\n");
 632                 goto out;
 633         }
 634
 635         /* Probably don't want this on remount; it might
 636          * mess with other nodes */
 637         if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
 638             (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
 639                 ret = -EINVAL;
 640                 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
 641                 goto out;
 642         }
 643
 644         /* We're going to/from readonly mode. */
 645         if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
 646                 /* Disable quota accounting before remounting RO */
 647                 if (*flags & MS_RDONLY) {
 648                         ret = ocfs2_susp_quotas(osb, 0);
 649                         if (ret < 0)
 650                                 goto out;
 651                 }
 652                 /* Lock here so the check of HARD_RO and the potential
 653                  * setting of SOFT_RO is atomic. */
 654                 spin_lock(&osb->osb_lock);
 655                 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
 656                         mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
 657                         ret = -EROFS;
 658                         goto unlock_osb;
 659                 }
 660
 661                 if (*flags & MS_RDONLY) {
 662                         mlog(0, "Going to ro mode.\n");
 663                         sb->s_flags |= MS_RDONLY;
 664                         osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 665                 } else {
 666                         mlog(0, "Making ro filesystem writeable.\n");
 667
 668                         if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
 669                                 mlog(ML_ERROR, "Cannot remount RDWR "
 670                                      "filesystem due to previous errors.\n");
 671                                 ret = -EROFS;
 672                                 goto unlock_osb;
 673                         }
 674                         incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
 675                         if (incompat_features) {
 676                                 mlog(ML_ERROR, "Cannot remount RDWR because "
 677                                      "of unsupported optional features "
 678                                      "(%x).\n", incompat_features);
 679                                 ret = -EINVAL;
 680                                 goto unlock_osb;
 681                         }
 682                         sb->s_flags &= ~MS_RDONLY;
 683                         osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
 684                 }
 685 unlock_osb:
 686                 spin_unlock(&osb->osb_lock);
 687                 /* Enable quota accounting after remounting RW */
 688                 if (!ret && !(*flags & MS_RDONLY)) {
 689                         if (sb_any_quota_suspended(sb))
 690                                 ret = ocfs2_susp_quotas(osb, 1);
 691                         else
 692                                 ret = ocfs2_enable_quotas(osb);
 693                         if (ret < 0) {
 694                                 /* Return back changes... */
 695                                 spin_lock(&osb->osb_lock);
 696                                 sb->s_flags |= MS_RDONLY;
 697                                 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 698                                 spin_unlock(&osb->osb_lock);
 699                                 goto out;
 700                         }
 701                 }
 702         }
 703
 704         if (!ret) {
 705                 /* Only save off the new mount options in case of a successful
 706                  * remount. */
 707                 osb->s_mount_opt = parsed_options.mount_opt;
 708                 osb->s_atime_quantum = parsed_options.atime_quantum;
 709                 osb->preferred_slot = parsed_options.slot;
 710                 if (parsed_options.commit_interval)
 711                         osb->osb_commit_interval = parsed_options.commit_interval;
 712
 713                 if (!ocfs2_is_hard_readonly(osb))
 714                         ocfs2_set_journal_params(osb);
 715
 716                 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 717                         ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
 718                                                         MS_POSIXACL : 0);
 719         }
 720 out:
 721         unlock_kernel();
 722         return ret;
 723 }
 724
 725 static int ocfs2_sb_probe(struct super_block *sb,
 726                           struct buffer_head **bh,
 727                           int *sector_size,
 728                           struct ocfs2_blockcheck_stats *stats)
 729 {
 730         int status, tmpstat;
 731         struct ocfs1_vol_disk_hdr *hdr;
 732         struct ocfs2_dinode *di;
 733         int blksize;
 734
 735         *bh = NULL;
 736
 737         /* may be > 512 */
 738         *sector_size = bdev_logical_block_size(sb->s_bdev);
 739         if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
 740                 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
 741                      *sector_size, OCFS2_MAX_BLOCKSIZE);
 742                 status = -EINVAL;
 743                 goto bail;
 744         }
 745
 746         /* Can this really happen? */
 747         if (*sector_size < OCFS2_MIN_BLOCKSIZE)
 748                 *sector_size = OCFS2_MIN_BLOCKSIZE;
 749
 750         /* check block zero for old format */
 751         status = ocfs2_get_sector(sb, bh, 0, *sector_size);
 752         if (status < 0) {
 753                 mlog_errno(status);
 754                 goto bail;
 755         }
 756         hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
 757         if (hdr->major_version == OCFS1_MAJOR_VERSION) {
 758                 mlog(ML_ERROR, "incompatible version: %u.%u\n",
 759                      hdr->major_version, hdr->minor_version);
 760                 status = -EINVAL;
 761         }
 762         if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
 763                    strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
 764                 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
 765                      hdr->signature);
 766                 status = -EINVAL;
 767         }
 768         brelse(*bh);
 769         *bh = NULL;
 770         if (status < 0) {
 771                 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
 772                      "upgraded before mounting with ocfs v2\n");
 773                 goto bail;
 774         }
 775
 776         /*
 777          * Now check at magic offset for 512, 1024, 2048, 4096
 778          * blocksizes.  4096 is the maximum blocksize because it is
 779          * the minimum clustersize.
 780          */
 781         status = -EINVAL;
 782         for (blksize = *sector_size;
 783              blksize <= OCFS2_MAX_BLOCKSIZE;
 784              blksize <<= 1) {
 785                 tmpstat = ocfs2_get_sector(sb, bh,
 786                                            OCFS2_SUPER_BLOCK_BLKNO,
 787                                            blksize);
 788                 if (tmpstat < 0) {
 789                         status = tmpstat;
 790                         mlog_errno(status);
 791                         break;
 792                 }
 793                 di = (struct ocfs2_dinode *) (*bh)->b_data;
 794                 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
 795                 spin_lock_init(&stats->b_lock);
 796                 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
 797                 if (tmpstat < 0) {
 798                         brelse(*bh);
 799                         *bh = NULL;
 800                 }
 801                 if (tmpstat != -EAGAIN) {
 802                         status = tmpstat;
 803                         break;
 804                 }
 805         }
 806
 807 bail:
 808         return status;
 809 }
 810
 811 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
 812 {
 813         if (ocfs2_mount_local(osb)) {
 814                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
 815                         mlog(ML_ERROR, "Cannot heartbeat on a locally "
 816                              "mounted device.\n");
 817                         return -EINVAL;
 818                 }
 819         }
 820
 821         if (ocfs2_userspace_stack(osb)) {
 822                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
 823                         mlog(ML_ERROR, "Userspace stack expected, but "
 824                              "o2cb heartbeat arguments passed to mount\n");
 825                         return -EINVAL;
 826                 }
 827         }
 828
 829         if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
 830                 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
 831                     !ocfs2_userspace_stack(osb)) {
 832                         mlog(ML_ERROR, "Heartbeat has to be started to mount "
 833                              "a read-write clustered device.\n");
 834                         return -EINVAL;
 835                 }
 836         }
 837
 838         return 0;
 839 }
 840
 841 /*
 842  * If we're using a userspace stack, mount should have passed
 843  * a name that matches the disk.  If not, mount should not
 844  * have passed a stack.
 845  */
 846 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
 847                                         struct mount_options *mopt)
 848 {
 849         if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
 850                 mlog(ML_ERROR,
 851                      "cluster stack passed to mount, but this filesystem "
 852                      "does not support it\n");
 853                 return -EINVAL;
 854         }
 855
 856         if (ocfs2_userspace_stack(osb) &&
 857             strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
 858                     OCFS2_STACK_LABEL_LEN)) {
 859                 mlog(ML_ERROR,
 860                      "cluster stack passed to mount (\"%s\") does not "
 861                      "match the filesystem (\"%s\")\n",
 862                      mopt->cluster_stack,
 863                      osb->osb_cluster_stack);
 864                 return -EINVAL;
 865         }
 866
 867         return 0;
 868 }
 869
 870 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
 871 {
 872         int type;
 873         struct super_block *sb = osb->sb;
 874         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 875                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 876         int status = 0;
 877
 878         for (type = 0; type < MAXQUOTAS; type++) {
 879                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 880                         continue;
 881                 if (unsuspend)
 882                         status = vfs_quota_enable(
 883                                         sb_dqopt(sb)->files[type],
 884                                         type, QFMT_OCFS2,
 885                                         DQUOT_SUSPENDED);
 886                 else
 887                         status = vfs_quota_disable(sb, type,
 888                                                    DQUOT_SUSPENDED);
 889                 if (status < 0)
 890                         break;
 891         }
 892         if (status < 0)
 893                 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
 894                      "remount (error = %d).\n", status);
 895         return status;
 896 }
 897
 898 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
 899 {
 900         struct inode *inode[MAXQUOTAS] = { NULL, NULL };
 901         struct super_block *sb = osb->sb;
 902         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 903                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 904         unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
 905                                         LOCAL_GROUP_QUOTA_SYSTEM_INODE };
 906         int status;
 907         int type;
 908
 909         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
 910         for (type = 0; type < MAXQUOTAS; type++) {
 911                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 912                         continue;
 913                 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
 914                                                         osb->slot_num);
 915                 if (!inode[type]) {
 916                         status = -ENOENT;
 917                         goto out_quota_off;
 918                 }
 919                 status = vfs_quota_enable(inode[type], type, QFMT_OCFS2,
 920                                                 DQUOT_USAGE_ENABLED);
 921                 if (status < 0)
 922                         goto out_quota_off;
 923         }
 924
 925         for (type = 0; type < MAXQUOTAS; type++)
 926                 iput(inode[type]);
 927         return 0;
 928 out_quota_off:
 929         ocfs2_disable_quotas(osb);
 930         for (type = 0; type < MAXQUOTAS; type++)
 931                 iput(inode[type]);
 932         mlog_errno(status);
 933         return status;
 934 }
 935
 936 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
 937 {
 938         int type;
 939         struct inode *inode;
 940         struct super_block *sb = osb->sb;
 941
 942         /* We mostly ignore errors in this function because there's not much
 943          * we can do when we see them */
 944         for (type = 0; type < MAXQUOTAS; type++) {
 945                 if (!sb_has_quota_loaded(sb, type))
 946                         continue;
 947                 inode = igrab(sb->s_dquot.files[type]);
 948                 /* Turn off quotas. This will remove all dquot structures from
 949                  * memory and so they will be automatically synced to global
 950                  * quota files */
 951                 vfs_quota_disable(sb, type, DQUOT_USAGE_ENABLED |
 952                                             DQUOT_LIMITS_ENABLED);
 953                 if (!inode)
 954                         continue;
 955                 iput(inode);
 956         }
 957 }
 958
 959 /* Handle quota on quotactl */
 960 static int ocfs2_quota_on(struct super_block *sb, int type, int format_id,
 961                           char *path, int remount)
 962 {
 963         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 964                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 965
 966         if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 967                 return -EINVAL;
 968
 969         if (remount)
 970                 return 0;       /* Just ignore it has been handled in
 971                                  * ocfs2_remount() */
 972         return vfs_quota_enable(sb_dqopt(sb)->files[type], type,
 973                                     format_id, DQUOT_LIMITS_ENABLED);
 974 }
 975
 976 /* Handle quota off quotactl */
 977 static int ocfs2_quota_off(struct super_block *sb, int type, int remount)
 978 {
 979         if (remount)
 980                 return 0;       /* Ignore now and handle later in
 981                                  * ocfs2_remount() */
 982         return vfs_quota_disable(sb, type, DQUOT_LIMITS_ENABLED);
 983 }
 984
 985 static const struct quotactl_ops ocfs2_quotactl_ops = {
 986         .quota_on       = ocfs2_quota_on,
 987         .quota_off      = ocfs2_quota_off,
 988         .quota_sync     = vfs_quota_sync,
 989         .get_info       = vfs_get_dqinfo,
 990         .set_info       = vfs_set_dqinfo,
 991         .get_dqblk      = vfs_get_dqblk,
 992         .set_dqblk      = vfs_set_dqblk,
 993 };
 994
 995 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
 996 {
 997         struct dentry *root;
 998         int status, sector_size;
 999         struct mount_options parsed_options;
1000         struct inode *inode = NULL;
1001         struct ocfs2_super *osb = NULL;
1002         struct buffer_head *bh = NULL;
1003         char nodestr[8];
1004         struct ocfs2_blockcheck_stats stats;
1005
1006         mlog_entry("%p, %p, %i", sb, data, silent);
1007
1008         if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
1009                 status = -EINVAL;
1010                 goto read_super_error;
1011         }
1012
1013         /* probe for superblock */
1014         status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
1015         if (status < 0) {
1016                 mlog(ML_ERROR, "superblock probe failed!\n");
1017                 goto read_super_error;
1018         }
1019
1020         status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1021         osb = OCFS2_SB(sb);
1022         if (status < 0) {
1023                 mlog_errno(status);
1024                 goto read_super_error;
1025         }
1026         brelse(bh);
1027         bh = NULL;
1028
1029         if (!ocfs2_check_set_options(sb, &parsed_options)) {
1030                 status = -EINVAL;
1031                 goto read_super_error;
1032         }
1033         osb->s_mount_opt = parsed_options.mount_opt;
1034         osb->s_atime_quantum = parsed_options.atime_quantum;
1035         osb->preferred_slot = parsed_options.slot;
1036         osb->osb_commit_interval = parsed_options.commit_interval;
1037
1038         ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1039         osb->osb_resv_level = parsed_options.resv_level;
1040         osb->osb_dir_resv_level = parsed_options.resv_level;
1041         if (parsed_options.dir_resv_level == -1)
1042                 osb->osb_dir_resv_level = parsed_options.resv_level;
1043         else
1044                 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1045
1046         status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1047         if (status)
1048                 goto read_super_error;
1049
1050         sb->s_magic = OCFS2_SUPER_MAGIC;
1051
1052         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1053                 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1054
1055         /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1056          * heartbeat=none */
1057         if (bdev_read_only(sb->s_bdev)) {
1058                 if (!(sb->s_flags & MS_RDONLY)) {
1059                         status = -EACCES;
1060                         mlog(ML_ERROR, "Readonly device detected but readonly "
1061                              "mount was not specified.\n");
1062                         goto read_super_error;
1063                 }
1064
1065                 /* You should not be able to start a local heartbeat
1066                  * on a readonly device. */
1067                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1068                         status = -EROFS;
1069                         mlog(ML_ERROR, "Local heartbeat specified on readonly "
1070                              "device.\n");
1071                         goto read_super_error;
1072                 }
1073
1074                 status = ocfs2_check_journals_nolocks(osb);
1075                 if (status < 0) {
1076                         if (status == -EROFS)
1077                                 mlog(ML_ERROR, "Recovery required on readonly "
1078                                      "file system, but write access is "
1079                                      "unavailable.\n");
1080                         else
1081                                 mlog_errno(status);
1082                         goto read_super_error;
1083                 }
1084
1085                 ocfs2_set_ro_flag(osb, 1);
1086
1087                 printk(KERN_NOTICE "Readonly device detected. No cluster "
1088                        "services will be utilized for this mount. Recovery "
1089                        "will be skipped.\n");
1090         }
1091
1092         if (!ocfs2_is_hard_readonly(osb)) {
1093                 if (sb->s_flags & MS_RDONLY)
1094                         ocfs2_set_ro_flag(osb, 0);
1095         }
1096
1097         status = ocfs2_verify_heartbeat(osb);
1098         if (status < 0) {
1099                 mlog_errno(status);
1100                 goto read_super_error;
1101         }
1102
1103         osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1104                                                  ocfs2_debugfs_root);
1105         if (!osb->osb_debug_root) {
1106                 status = -EINVAL;
1107                 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1108                 goto read_super_error;
1109         }
1110
1111         osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1112                                             osb->osb_debug_root,
1113                                             osb,
1114                                             &ocfs2_osb_debug_fops);
1115         if (!osb->osb_ctxt) {
1116                 status = -EINVAL;
1117                 mlog_errno(status);
1118                 goto read_super_error;
1119         }
1120
1121         if (ocfs2_meta_ecc(osb)) {
1122                 status = ocfs2_blockcheck_stats_debugfs_install(
1123                                                 &osb->osb_ecc_stats,
1124                                                 osb->osb_debug_root);
1125                 if (status) {
1126                         mlog(ML_ERROR,
1127                              "Unable to create blockcheck statistics "
1128                              "files\n");
1129                         goto read_super_error;
1130                 }
1131         }
1132
1133         status = ocfs2_mount_volume(sb);
1134         if (osb->root_inode)
1135                 inode = igrab(osb->root_inode);
1136
1137         if (status < 0)
1138                 goto read_super_error;
1139
1140         if (!inode) {
1141                 status = -EIO;
1142                 mlog_errno(status);
1143                 goto read_super_error;
1144         }
1145
1146         root = d_alloc_root(inode);
1147         if (!root) {
1148                 status = -ENOMEM;
1149                 mlog_errno(status);
1150                 goto read_super_error;
1151         }
1152
1153         sb->s_root = root;
1154
1155         ocfs2_complete_mount_recovery(osb);
1156
1157         if (ocfs2_mount_local(osb))
1158                 snprintf(nodestr, sizeof(nodestr), "local");
1159         else
1160                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1161
1162         printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1163                "with %s data mode.\n",
1164                osb->dev_str, nodestr, osb->slot_num,
1165                osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1166                "ordered");
1167
1168         atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1169         wake_up(&osb->osb_mount_event);
1170
1171         /* Now we can initialize quotas because we can afford to wait
1172          * for cluster locks recovery now. That also means that truncation
1173          * log recovery can happen but that waits for proper quota setup */
1174         if (!(sb->s_flags & MS_RDONLY)) {
1175                 status = ocfs2_enable_quotas(osb);
1176                 if (status < 0) {
1177                         /* We have to err-out specially here because
1178                          * s_root is already set */
1179                         mlog_errno(status);
1180                         atomic_set(&osb->vol_state, VOLUME_DISABLED);
1181                         wake_up(&osb->osb_mount_event);
1182                         mlog_exit(status);
1183                         return status;
1184                 }
1185         }
1186
1187         ocfs2_complete_quota_recovery(osb);
1188
1189         /* Now we wake up again for processes waiting for quotas */
1190         atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1191         wake_up(&osb->osb_mount_event);
1192
1193         /* Start this when the mount is almost sure of being successful */
1194         ocfs2_orphan_scan_start(osb);
1195
1196         mlog_exit(status);
1197         return status;
1198
1199 read_super_error:
1200         brelse(bh);
1201
1202         if (inode)
1203                 iput(inode);
1204
1205         if (osb) {
1206                 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1207                 wake_up(&osb->osb_mount_event);
1208                 ocfs2_dismount_volume(sb, 1);
1209         }
1210
1211         mlog_exit(status);
1212         return status;
1213 }
1214
1215 static int ocfs2_get_sb(struct file_system_type *fs_type,
1216                         int flags,
1217                         const char *dev_name,
1218                         void *data,
1219                         struct vfsmount *mnt)
1220 {
1221         return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super,
1222                            mnt);
1223 }
1224
1225 static void ocfs2_kill_sb(struct super_block *sb)
1226 {
1227         struct ocfs2_super *osb = OCFS2_SB(sb);
1228
1229         /* Failed mount? */
1230         if (!osb || atomic_read(&osb->vol_state) == VOLUME_DISABLED)
1231                 goto out;
1232
1233         /* Prevent further queueing of inode drop events */
1234         spin_lock(&dentry_list_lock);
1235         ocfs2_set_osb_flag(osb, OCFS2_OSB_DROP_DENTRY_LOCK_IMMED);
1236         spin_unlock(&dentry_list_lock);
1237         /* Wait for work to finish and/or remove it */
1238         cancel_work_sync(&osb->dentry_lock_work);
1239 out:
1240         kill_block_super(sb);
1241 }
1242
1243 static struct file_system_type ocfs2_fs_type = {
1244         .owner          = THIS_MODULE,
1245         .name           = "ocfs2",
1246         .get_sb         = ocfs2_get_sb, /* is this called when we mount
1247                                         * the fs? */
1248         .kill_sb        = ocfs2_kill_sb,
1249
1250         .fs_flags       = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1251         .next           = NULL
1252 };
1253
1254 static int ocfs2_check_set_options(struct super_block *sb,
1255                                    struct mount_options *options)
1256 {
1257         if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1258             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1259                                          OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1260                 mlog(ML_ERROR, "User quotas were requested, but this "
1261                      "filesystem does not have the feature enabled.\n");
1262                 return 0;
1263         }
1264         if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1265             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1266                                          OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1267                 mlog(ML_ERROR, "Group quotas were requested, but this "
1268                      "filesystem does not have the feature enabled.\n");
1269                 return 0;
1270         }
1271         if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1272             !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1273                 mlog(ML_ERROR, "ACL support requested but extended attributes "
1274                      "feature is not enabled\n");
1275                 return 0;
1276         }
1277         /* No ACL setting specified? Use XATTR feature... */
1278         if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1279                                     OCFS2_MOUNT_NO_POSIX_ACL))) {
1280                 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1281                         options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1282                 else
1283                         options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1284         }
1285         return 1;
1286 }
1287
1288 static int ocfs2_parse_options(struct super_block *sb,
1289                                char *options,
1290                                struct mount_options *mopt,
1291                                int is_remount)
1292 {
1293         int status;
1294         char *p;
1295
1296         mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
1297                    options ? options : "(none)");
1298
1299         mopt->commit_interval = 0;
1300         mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1301         mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1302         mopt->slot = OCFS2_INVALID_SLOT;
1303         mopt->localalloc_opt = -1;
1304         mopt->cluster_stack[0] = '\0';
1305         mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1306         mopt->dir_resv_level = -1;
1307
1308         if (!options) {
1309                 status = 1;
1310                 goto bail;
1311         }
1312
1313         while ((p = strsep(&options, ",")) != NULL) {
1314                 int token, option;
1315                 substring_t args[MAX_OPT_ARGS];
1316
1317                 if (!*p)
1318                         continue;
1319
1320                 token = match_token(p, tokens, args);
1321                 switch (token) {
1322                 case Opt_hb_local:
1323                         mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1324                         break;
1325                 case Opt_hb_none:
1326                         mopt->mount_opt &= ~OCFS2_MOUNT_HB_LOCAL;
1327                         break;
1328                 case Opt_barrier:
1329                         if (match_int(&args[0], &option)) {
1330                                 status = 0;
1331                                 goto bail;
1332                         }
1333                         if (option)
1334                                 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1335                         else
1336                                 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1337                         break;
1338                 case Opt_intr:
1339                         mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1340                         break;
1341                 case Opt_nointr:
1342                         mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1343                         break;
1344                 case Opt_err_panic:
1345                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1346                         break;
1347                 case Opt_err_ro:
1348                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1349                         break;
1350                 case Opt_data_ordered:
1351                         mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1352                         break;
1353                 case Opt_data_writeback:
1354                         mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1355                         break;
1356                 case Opt_user_xattr:
1357                         mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1358                         break;
1359                 case Opt_nouser_xattr:
1360                         mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1361                         break;
1362                 case Opt_atime_quantum:
1363                         if (match_int(&args[0], &option)) {
1364                                 status = 0;
1365                                 goto bail;
1366                         }
1367                         if (option >= 0)
1368                                 mopt->atime_quantum = option;
1369                         break;
1370                 case Opt_slot:
1371                         option = 0;
1372                         if (match_int(&args[0], &option)) {
1373                                 status = 0;
1374                                 goto bail;
1375                         }
1376                         if (option)
1377                                 mopt->slot = (s16)option;
1378                         break;
1379                 case Opt_commit:
1380                         option = 0;
1381                         if (match_int(&args[0], &option)) {
1382                                 status = 0;
1383                                 goto bail;
1384                         }
1385                         if (option < 0)
1386                                 return 0;
1387                         if (option == 0)
1388                                 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1389                         mopt->commit_interval = HZ * option;
1390                         break;
1391                 case Opt_localalloc:
1392                         option = 0;
1393                         if (match_int(&args[0], &option)) {
1394                                 status = 0;
1395                                 goto bail;
1396                         }
1397                         if (option >= 0)
1398                                 mopt->localalloc_opt = option;
1399                         break;
1400                 case Opt_localflocks:
1401                         /*
1402                          * Changing this during remount could race
1403                          * flock() requests, or "unbalance" existing
1404                          * ones (e.g., a lock is taken in one mode but
1405                          * dropped in the other). If users care enough
1406                          * to flip locking modes during remount, we
1407                          * could add a "local" flag to individual
1408                          * flock structures for proper tracking of
1409                          * state.
1410                          */
1411                         if (!is_remount)
1412                                 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1413                         break;
1414                 case Opt_stack:
1415                         /* Check both that the option we were passed
1416                          * is of the right length and that it is a proper
1417                          * string of the right length.
1418                          */
1419                         if (((args[0].to - args[0].from) !=
1420                              OCFS2_STACK_LABEL_LEN) ||
1421                             (strnlen(args[0].from,
1422                                      OCFS2_STACK_LABEL_LEN) !=
1423                              OCFS2_STACK_LABEL_LEN)) {
1424                                 mlog(ML_ERROR,
1425                                      "Invalid cluster_stack option\n");
1426                                 status = 0;
1427                                 goto bail;
1428                         }
1429                         memcpy(mopt->cluster_stack, args[0].from,
1430                                OCFS2_STACK_LABEL_LEN);
1431                         mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1432                         break;
1433                 case Opt_inode64:
1434                         mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1435                         break;
1436                 case Opt_usrquota:
1437                         mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1438                         break;
1439                 case Opt_grpquota:
1440                         mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1441                         break;
1442                 case Opt_acl:
1443                         mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1444                         mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1445                         break;
1446                 case Opt_noacl:
1447                         mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1448                         mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1449                         break;
1450                 case Opt_resv_level:
1451                         if (is_remount)
1452                                 break;
1453                         if (match_int(&args[0], &option)) {
1454                                 status = 0;
1455                                 goto bail;
1456                         }
1457                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1458                             option < OCFS2_MAX_RESV_LEVEL)
1459                                 mopt->resv_level = option;
1460                         break;
1461                 case Opt_dir_resv_level:
1462                         if (is_remount)
1463                                 break;
1464                         if (match_int(&args[0], &option)) {
1465                                 status = 0;
1466                                 goto bail;
1467                         }
1468                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1469                             option < OCFS2_MAX_RESV_LEVEL)
1470                                 mopt->dir_resv_level = option;
1471                         break;
1472                 default:
1473                         mlog(ML_ERROR,
1474                              "Unrecognized mount option \"%s\" "
1475                              "or missing value\n", p);
1476                         status = 0;
1477                         goto bail;
1478                 }
1479         }
1480
1481         status = 1;
1482
1483 bail:
1484         mlog_exit(status);
1485         return status;
1486 }
1487
1488 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1489 {
1490         struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb);
1491         unsigned long opts = osb->s_mount_opt;
1492         unsigned int local_alloc_megs;
1493
1494         if (opts & OCFS2_MOUNT_HB_LOCAL)
1495                 seq_printf(s, ",_netdev,heartbeat=local");
1496         else
1497                 seq_printf(s, ",heartbeat=none");
1498
1499         if (opts & OCFS2_MOUNT_NOINTR)
1500                 seq_printf(s, ",nointr");
1501
1502         if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1503                 seq_printf(s, ",data=writeback");
1504         else
1505                 seq_printf(s, ",data=ordered");
1506
1507         if (opts & OCFS2_MOUNT_BARRIER)
1508                 seq_printf(s, ",barrier=1");
1509
1510         if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1511                 seq_printf(s, ",errors=panic");
1512         else
1513                 seq_printf(s, ",errors=remount-ro");
1514
1515         if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1516                 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1517
1518         if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM)
1519                 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1520
1521         if (osb->osb_commit_interval)
1522                 seq_printf(s, ",commit=%u",
1523                            (unsigned) (osb->osb_commit_interval / HZ));
1524
1525         local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1526         if (local_alloc_megs != ocfs2_la_default_mb(osb))
1527                 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1528
1529         if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1530                 seq_printf(s, ",localflocks,");
1531
1532         if (osb->osb_cluster_stack[0])
1533                 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1534                            osb->osb_cluster_stack);
1535         if (opts & OCFS2_MOUNT_USRQUOTA)
1536                 seq_printf(s, ",usrquota");
1537         if (opts & OCFS2_MOUNT_GRPQUOTA)
1538                 seq_printf(s, ",grpquota");
1539
1540         if (opts & OCFS2_MOUNT_NOUSERXATTR)
1541                 seq_printf(s, ",nouser_xattr");
1542         else
1543                 seq_printf(s, ",user_xattr");
1544
1545         if (opts & OCFS2_MOUNT_INODE64)
1546                 seq_printf(s, ",inode64");
1547
1548         if (opts & OCFS2_MOUNT_POSIX_ACL)
1549                 seq_printf(s, ",acl");
1550         else
1551                 seq_printf(s, ",noacl");
1552
1553         if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1554                 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1555
1556         if (osb->osb_dir_resv_level != osb->osb_resv_level)
1557                 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1558
1559         return 0;
1560 }
1561
1562 static int __init ocfs2_init(void)
1563 {
1564         int status;
1565
1566         mlog_entry_void();
1567
1568         ocfs2_print_version();
1569
1570         status = init_ocfs2_uptodate_cache();
1571         if (status < 0) {
1572                 mlog_errno(status);
1573                 goto leave;
1574         }
1575
1576         status = ocfs2_initialize_mem_caches();
1577         if (status < 0) {
1578                 mlog_errno(status);
1579                 goto leave;
1580         }
1581
1582         ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1583         if (!ocfs2_wq) {
1584                 status = -ENOMEM;
1585                 goto leave;
1586         }
1587
1588         ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1589         if (!ocfs2_debugfs_root) {
1590                 status = -EFAULT;
1591                 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1592         }
1593
1594         status = ocfs2_quota_setup();
1595         if (status)
1596                 goto leave;
1597
1598         ocfs2_set_locking_protocol();
1599
1600         status = register_quota_format(&ocfs2_quota_format);
1601 leave:
1602         if (status < 0) {
1603                 ocfs2_quota_shutdown();
1604                 ocfs2_free_mem_caches();
1605                 exit_ocfs2_uptodate_cache();
1606         }
1607
1608         mlog_exit(status);
1609
1610         if (status >= 0) {
1611                 return register_filesystem(&ocfs2_fs_type);
1612         } else
1613                 return -1;
1614 }
1615
1616 static void __exit ocfs2_exit(void)
1617 {
1618         mlog_entry_void();
1619
1620         ocfs2_quota_shutdown();
1621
1622         if (ocfs2_wq) {
1623                 flush_workqueue(ocfs2_wq);
1624                 destroy_workqueue(ocfs2_wq);
1625         }
1626
1627         unregister_quota_format(&ocfs2_quota_format);
1628
1629         debugfs_remove(ocfs2_debugfs_root);
1630
1631         ocfs2_free_mem_caches();
1632
1633         unregister_filesystem(&ocfs2_fs_type);
1634
1635         exit_ocfs2_uptodate_cache();
1636
1637         mlog_exit_void();
1638 }
1639
1640 static void ocfs2_put_super(struct super_block *sb)
1641 {
1642         mlog_entry("(0x%p)\n", sb);
1643
1644         lock_kernel();
1645
1646         ocfs2_sync_blockdev(sb);
1647         ocfs2_dismount_volume(sb, 0);
1648
1649         unlock_kernel();
1650
1651         mlog_exit_void();
1652 }
1653
1654 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1655 {
1656         struct ocfs2_super *osb;
1657         u32 numbits, freebits;
1658         int status;
1659         struct ocfs2_dinode *bm_lock;
1660         struct buffer_head *bh = NULL;
1661         struct inode *inode = NULL;
1662
1663         mlog_entry("(%p, %p)\n", dentry->d_sb, buf);
1664
1665         osb = OCFS2_SB(dentry->d_sb);
1666
1667         inode = ocfs2_get_system_file_inode(osb,
1668                                             GLOBAL_BITMAP_SYSTEM_INODE,
1669                                             OCFS2_INVALID_SLOT);
1670         if (!inode) {
1671                 mlog(ML_ERROR, "failed to get bitmap inode\n");
1672                 status = -EIO;
1673                 goto bail;
1674         }
1675
1676         status = ocfs2_inode_lock(inode, &bh, 0);
1677         if (status < 0) {
1678                 mlog_errno(status);
1679                 goto bail;
1680         }
1681
1682         bm_lock = (struct ocfs2_dinode *) bh->b_data;
1683
1684         numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1685         freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1686
1687         buf->f_type = OCFS2_SUPER_MAGIC;
1688         buf->f_bsize = dentry->d_sb->s_blocksize;
1689         buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1690         buf->f_blocks = ((sector_t) numbits) *
1691                         (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1692         buf->f_bfree = ((sector_t) freebits) *
1693                        (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1694         buf->f_bavail = buf->f_bfree;
1695         buf->f_files = numbits;
1696         buf->f_ffree = freebits;
1697         buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1698                                 & 0xFFFFFFFFUL;
1699         buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1700                                 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1701
1702         brelse(bh);
1703
1704         ocfs2_inode_unlock(inode, 0);
1705         status = 0;
1706 bail:
1707         if (inode)
1708                 iput(inode);
1709
1710         mlog_exit(status);
1711
1712         return status;
1713 }
1714
1715 static void ocfs2_inode_init_once(void *data)
1716 {
1717         struct ocfs2_inode_info *oi = data;
1718
1719         oi->ip_flags = 0;
1720         oi->ip_open_count = 0;
1721         spin_lock_init(&oi->ip_lock);
1722         ocfs2_extent_map_init(&oi->vfs_inode);
1723         INIT_LIST_HEAD(&oi->ip_io_markers);
1724         oi->ip_dir_start_lookup = 0;
1725
1726         init_rwsem(&oi->ip_alloc_sem);
1727         init_rwsem(&oi->ip_xattr_sem);
1728         mutex_init(&oi->ip_io_mutex);
1729
1730         oi->ip_blkno = 0ULL;
1731         oi->ip_clusters = 0;
1732
1733         ocfs2_resv_init_once(&oi->ip_la_data_resv);
1734
1735         ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1736         ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1737         ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1738
1739         ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1740                                   &ocfs2_inode_caching_ops);
1741
1742         inode_init_once(&oi->vfs_inode);
1743 }
1744
1745 static int ocfs2_initialize_mem_caches(void)
1746 {
1747         ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1748                                        sizeof(struct ocfs2_inode_info),
1749                                        0,
1750                                        (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1751                                                 SLAB_MEM_SPREAD),
1752                                        ocfs2_inode_init_once);
1753         ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1754                                         sizeof(struct ocfs2_dquot),
1755                                         0,
1756                                         (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1757                                                 SLAB_MEM_SPREAD),
1758                                         NULL);
1759         ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1760                                         sizeof(struct ocfs2_quota_chunk),
1761                                         0,
1762                                         (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1763                                         NULL);
1764         if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1765             !ocfs2_qf_chunk_cachep) {
1766                 if (ocfs2_inode_cachep)
1767                         kmem_cache_destroy(ocfs2_inode_cachep);
1768                 if (ocfs2_dquot_cachep)
1769                         kmem_cache_destroy(ocfs2_dquot_cachep);
1770                 if (ocfs2_qf_chunk_cachep)
1771                         kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1772                 return -ENOMEM;
1773         }
1774
1775         return 0;
1776 }
1777
1778 static void ocfs2_free_mem_caches(void)
1779 {
1780         if (ocfs2_inode_cachep)
1781                 kmem_cache_destroy(ocfs2_inode_cachep);
1782         ocfs2_inode_cachep = NULL;
1783
1784         if (ocfs2_dquot_cachep)
1785                 kmem_cache_destroy(ocfs2_dquot_cachep);
1786         ocfs2_dquot_cachep = NULL;
1787
1788         if (ocfs2_qf_chunk_cachep)
1789                 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1790         ocfs2_qf_chunk_cachep = NULL;
1791 }
1792
1793 static int ocfs2_get_sector(struct super_block *sb,
1794                             struct buffer_head **bh,
1795                             int block,
1796                             int sect_size)
1797 {
1798         if (!sb_set_blocksize(sb, sect_size)) {
1799                 mlog(ML_ERROR, "unable to set blocksize\n");
1800                 return -EIO;
1801         }
1802
1803         *bh = sb_getblk(sb, block);
1804         if (!*bh) {
1805                 mlog_errno(-EIO);
1806                 return -EIO;
1807         }
1808         lock_buffer(*bh);
1809         if (!buffer_dirty(*bh))
1810                 clear_buffer_uptodate(*bh);
1811         unlock_buffer(*bh);
1812         ll_rw_block(READ, 1, bh);
1813         wait_on_buffer(*bh);
1814         if (!buffer_uptodate(*bh)) {
1815                 mlog_errno(-EIO);
1816                 brelse(*bh);
1817                 *bh = NULL;
1818                 return -EIO;
1819         }
1820
1821         return 0;
1822 }
1823
1824 static int ocfs2_mount_volume(struct super_block *sb)
1825 {
1826         int status = 0;
1827         int unlock_super = 0;
1828         struct ocfs2_super *osb = OCFS2_SB(sb);
1829
1830         mlog_entry_void();
1831
1832         if (ocfs2_is_hard_readonly(osb))
1833                 goto leave;
1834
1835         status = ocfs2_dlm_init(osb);
1836         if (status < 0) {
1837                 mlog_errno(status);
1838                 goto leave;
1839         }
1840
1841         status = ocfs2_super_lock(osb, 1);
1842         if (status < 0) {
1843                 mlog_errno(status);
1844                 goto leave;
1845         }
1846         unlock_super = 1;
1847
1848         /* This will load up the node map and add ourselves to it. */
1849         status = ocfs2_find_slot(osb);
1850         if (status < 0) {
1851                 mlog_errno(status);
1852                 goto leave;
1853         }
1854
1855         /* load all node-local system inodes */
1856         status = ocfs2_init_local_system_inodes(osb);
1857         if (status < 0) {
1858                 mlog_errno(status);
1859                 goto leave;
1860         }
1861
1862         status = ocfs2_check_volume(osb);
1863         if (status < 0) {
1864                 mlog_errno(status);
1865                 goto leave;
1866         }
1867
1868         status = ocfs2_truncate_log_init(osb);
1869         if (status < 0)
1870                 mlog_errno(status);
1871
1872 leave:
1873         if (unlock_super)
1874                 ocfs2_super_unlock(osb, 1);
1875
1876         mlog_exit(status);
1877         return status;
1878 }
1879
1880 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1881 {
1882         int tmp, hangup_needed = 0;
1883         struct ocfs2_super *osb = NULL;
1884         char nodestr[8];
1885
1886         mlog_entry("(0x%p)\n", sb);
1887
1888         BUG_ON(!sb);
1889         osb = OCFS2_SB(sb);
1890         BUG_ON(!osb);
1891
1892         debugfs_remove(osb->osb_ctxt);
1893
1894         /*
1895          * Flush inode dropping work queue so that deletes are
1896          * performed while the filesystem is still working
1897          */
1898         ocfs2_drop_all_dl_inodes(osb);
1899
1900         /* Orphan scan should be stopped as early as possible */
1901         ocfs2_orphan_scan_stop(osb);
1902
1903         ocfs2_disable_quotas(osb);
1904
1905         ocfs2_shutdown_local_alloc(osb);
1906
1907         ocfs2_truncate_log_shutdown(osb);
1908
1909         /* This will disable recovery and flush any recovery work. */
1910         ocfs2_recovery_exit(osb);
1911
1912         ocfs2_journal_shutdown(osb);
1913
1914         ocfs2_sync_blockdev(sb);
1915
1916         ocfs2_purge_refcount_trees(osb);
1917
1918         /* No cluster connection means we've failed during mount, so skip
1919          * all the steps which depended on that to complete. */
1920         if (osb->cconn) {
1921                 tmp = ocfs2_super_lock(osb, 1);
1922                 if (tmp < 0) {
1923                         mlog_errno(tmp);
1924                         return;
1925                 }
1926         }
1927
1928         if (osb->slot_num != OCFS2_INVALID_SLOT)
1929                 ocfs2_put_slot(osb);
1930
1931         if (osb->cconn)
1932                 ocfs2_super_unlock(osb, 1);
1933
1934         ocfs2_release_system_inodes(osb);
1935
1936         /*
1937          * If we're dismounting due to mount error, mount.ocfs2 will clean
1938          * up heartbeat.  If we're a local mount, there is no heartbeat.
1939          * If we failed before we got a uuid_str yet, we can't stop
1940          * heartbeat.  Otherwise, do it.
1941          */
1942         if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str)
1943                 hangup_needed = 1;
1944
1945         if (osb->cconn)
1946                 ocfs2_dlm_shutdown(osb, hangup_needed);
1947
1948         ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1949         debugfs_remove(osb->osb_debug_root);
1950
1951         if (hangup_needed)
1952                 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1953
1954         atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1955
1956         if (ocfs2_mount_local(osb))
1957                 snprintf(nodestr, sizeof(nodestr), "local");
1958         else
1959                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1960
1961         printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1962                osb->dev_str, nodestr);
1963
1964         ocfs2_delete_osb(osb);
1965         kfree(osb);
1966         sb->s_dev = 0;
1967         sb->s_fs_info = NULL;
1968 }
1969
1970 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1971                                 unsigned uuid_bytes)
1972 {
1973         int i, ret;
1974         char *ptr;
1975
1976         BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1977
1978         osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1979         if (osb->uuid_str == NULL)
1980                 return -ENOMEM;
1981
1982         for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1983                 /* print with null */
1984                 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1985                 if (ret != 2) /* drop super cleans up */
1986                         return -EINVAL;
1987                 /* then only advance past the last char */
1988                 ptr += 2;
1989         }
1990
1991         return 0;
1992 }
1993
1994 static int ocfs2_initialize_super(struct super_block *sb,
1995                                   struct buffer_head *bh,
1996                                   int sector_size,
1997                                   struct ocfs2_blockcheck_stats *stats)
1998 {
1999         int status;
2000         int i, cbits, bbits;
2001         struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2002         struct inode *inode = NULL;
2003         struct ocfs2_journal *journal;
2004         __le32 uuid_net_key;
2005         struct ocfs2_super *osb;
2006
2007         mlog_entry_void();
2008
2009         osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2010         if (!osb) {
2011                 status = -ENOMEM;
2012                 mlog_errno(status);
2013                 goto bail;
2014         }
2015
2016         sb->s_fs_info = osb;
2017         sb->s_op = &ocfs2_sops;
2018         sb->s_export_op = &ocfs2_export_ops;
2019         sb->s_qcop = &ocfs2_quotactl_ops;
2020         sb->dq_op = &ocfs2_quota_operations;
2021         sb->s_xattr = ocfs2_xattr_handlers;
2022         sb->s_time_gran = 1;
2023         sb->s_flags |= MS_NOATIME;
2024         /* this is needed to support O_LARGEFILE */
2025         cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2026         bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2027         sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2028
2029         osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2030
2031         for (i = 0; i < 3; i++)
2032                 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2033         osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2034
2035         osb->sb = sb;
2036         /* Save off for ocfs2_rw_direct */
2037         osb->s_sectsize_bits = blksize_bits(sector_size);
2038         BUG_ON(!osb->s_sectsize_bits);
2039
2040         spin_lock_init(&osb->dc_task_lock);
2041         init_waitqueue_head(&osb->dc_event);
2042         osb->dc_work_sequence = 0;
2043         osb->dc_wake_sequence = 0;
2044         INIT_LIST_HEAD(&osb->blocked_lock_list);
2045         osb->blocked_lock_count = 0;
2046         spin_lock_init(&osb->osb_lock);
2047         spin_lock_init(&osb->osb_xattr_lock);
2048         ocfs2_init_steal_slots(osb);
2049
2050         atomic_set(&osb->alloc_stats.moves, 0);
2051         atomic_set(&osb->alloc_stats.local_data, 0);
2052         atomic_set(&osb->alloc_stats.bitmap_data, 0);
2053         atomic_set(&osb->alloc_stats.bg_allocs, 0);
2054         atomic_set(&osb->alloc_stats.bg_extends, 0);
2055
2056         /* Copy the blockcheck stats from the superblock probe */
2057         osb->osb_ecc_stats = *stats;
2058
2059         ocfs2_init_node_maps(osb);
2060
2061         snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2062                  MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2063
2064         ocfs2_orphan_scan_init(osb);
2065
2066         status = ocfs2_recovery_init(osb);
2067         if (status) {
2068                 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2069                 mlog_errno(status);
2070                 goto bail;
2071         }
2072
2073         init_waitqueue_head(&osb->checkpoint_event);
2074         atomic_set(&osb->needs_checkpoint, 0);
2075
2076         osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2077
2078         osb->slot_num = OCFS2_INVALID_SLOT;
2079
2080         osb->s_xattr_inline_size = le16_to_cpu(
2081                                         di->id2.i_super.s_xattr_inline_size);
2082
2083         osb->local_alloc_state = OCFS2_LA_UNUSED;
2084         osb->local_alloc_bh = NULL;
2085         INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2086
2087         init_waitqueue_head(&osb->osb_mount_event);
2088
2089         status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2090         if (status) {
2091                 mlog_errno(status);
2092                 goto bail;
2093         }
2094
2095         osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2096         if (!osb->vol_label) {
2097                 mlog(ML_ERROR, "unable to alloc vol label\n");
2098                 status = -ENOMEM;
2099                 goto bail;
2100         }
2101
2102         osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2103         if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2104                 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2105                      osb->max_slots);
2106                 status = -EINVAL;
2107                 goto bail;
2108         }
2109         mlog(0, "max_slots for this device: %u\n", osb->max_slots);
2110
2111         osb->slot_recovery_generations =
2112                 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2113                         GFP_KERNEL);
2114         if (!osb->slot_recovery_generations) {
2115                 status = -ENOMEM;
2116                 mlog_errno(status);
2117                 goto bail;
2118         }
2119
2120         init_waitqueue_head(&osb->osb_wipe_event);
2121         osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2122                                         sizeof(*osb->osb_orphan_wipes),
2123                                         GFP_KERNEL);
2124         if (!osb->osb_orphan_wipes) {
2125                 status = -ENOMEM;
2126                 mlog_errno(status);
2127                 goto bail;
2128         }
2129
2130         osb->osb_rf_lock_tree = RB_ROOT;
2131
2132         osb->s_feature_compat =
2133                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2134         osb->s_feature_ro_compat =
2135                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2136         osb->s_feature_incompat =
2137                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2138
2139         if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2140                 mlog(ML_ERROR, "couldn't mount because of unsupported "
2141                      "optional features (%x).\n", i);
2142                 status = -EINVAL;
2143                 goto bail;
2144         }
2145         if (!(osb->sb->s_flags & MS_RDONLY) &&
2146             (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2147                 mlog(ML_ERROR, "couldn't mount RDWR because of "
2148                      "unsupported optional features (%x).\n", i);
2149                 status = -EINVAL;
2150                 goto bail;
2151         }
2152
2153         if (ocfs2_userspace_stack(osb)) {
2154                 memcpy(osb->osb_cluster_stack,
2155                        OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2156                        OCFS2_STACK_LABEL_LEN);
2157                 osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
2158                 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2159                         mlog(ML_ERROR,
2160                              "couldn't mount because of an invalid "
2161                              "cluster stack label (%s) \n",
2162                              osb->osb_cluster_stack);
2163                         status = -EINVAL;
2164                         goto bail;
2165                 }
2166         } else {
2167                 /* The empty string is identical with classic tools that
2168                  * don't know about s_cluster_info. */
2169                 osb->osb_cluster_stack[0] = '\0';
2170         }
2171
2172         get_random_bytes(&osb->s_next_generation, sizeof(u32));
2173
2174         /* FIXME
2175          * This should be done in ocfs2_journal_init(), but unknown
2176          * ordering issues will cause the filesystem to crash.
2177          * If anyone wants to figure out what part of the code
2178          * refers to osb->journal before ocfs2_journal_init() is run,
2179          * be my guest.
2180          */
2181         /* initialize our journal structure */
2182
2183         journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2184         if (!journal) {
2185                 mlog(ML_ERROR, "unable to alloc journal\n");
2186                 status = -ENOMEM;
2187                 goto bail;
2188         }
2189         osb->journal = journal;
2190         journal->j_osb = osb;
2191
2192         atomic_set(&journal->j_num_trans, 0);
2193         init_rwsem(&journal->j_trans_barrier);
2194         init_waitqueue_head(&journal->j_checkpointed);
2195         spin_lock_init(&journal->j_lock);
2196         journal->j_trans_id = (unsigned long) 1;
2197         INIT_LIST_HEAD(&journal->j_la_cleanups);
2198         INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2199         journal->j_state = OCFS2_JOURNAL_FREE;
2200
2201         INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes);
2202         osb->dentry_lock_list = NULL;
2203
2204         /* get some pseudo constants for clustersize bits */
2205         osb->s_clustersize_bits =
2206                 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2207         osb->s_clustersize = 1 << osb->s_clustersize_bits;
2208         mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
2209
2210         if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2211             osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2212                 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2213                      osb->s_clustersize);
2214                 status = -EINVAL;
2215                 goto bail;
2216         }
2217
2218         if (ocfs2_clusters_to_blocks(osb->sb, le32_to_cpu(di->i_clusters) - 1)
2219             > (u32)~0UL) {
2220                 mlog(ML_ERROR, "Volume might try to write to blocks beyond "
2221                      "what jbd can address in 32 bits.\n");
2222                 status = -EINVAL;
2223                 goto bail;
2224         }
2225
2226         if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2227                                  sizeof(di->id2.i_super.s_uuid))) {
2228                 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2229                 status = -ENOMEM;
2230                 goto bail;
2231         }
2232
2233         memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
2234
2235         strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2236         osb->vol_label[63] = '\0';
2237         osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2238         osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2239         osb->first_cluster_group_blkno =
2240                 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2241         osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2242         osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2243         mlog(0, "vol_label: %s\n", osb->vol_label);
2244         mlog(0, "uuid: %s\n", osb->uuid_str);
2245         mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
2246              (unsigned long long)osb->root_blkno,
2247              (unsigned long long)osb->system_dir_blkno);
2248
2249         osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2250         if (!osb->osb_dlm_debug) {
2251                 status = -ENOMEM;
2252                 mlog_errno(status);
2253                 goto bail;
2254         }
2255
2256         atomic_set(&osb->vol_state, VOLUME_INIT);
2257
2258         /* load root, system_dir, and all global system inodes */
2259         status = ocfs2_init_global_system_inodes(osb);
2260         if (status < 0) {
2261                 mlog_errno(status);
2262                 goto bail;
2263         }
2264
2265         /*
2266          * global bitmap
2267          */
2268         inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2269                                             OCFS2_INVALID_SLOT);
2270         if (!inode) {
2271                 status = -EINVAL;
2272                 mlog_errno(status);
2273                 goto bail;
2274         }
2275
2276         osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2277         osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2278         iput(inode);
2279
2280         osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2281                                  osb->s_feature_incompat) * 8;
2282
2283         status = ocfs2_init_slot_info(osb);
2284         if (status < 0) {
2285                 mlog_errno(status);
2286                 goto bail;
2287         }
2288
2289 bail:
2290         mlog_exit(status);
2291         return status;
2292 }
2293
2294 /*
2295  * will return: -EAGAIN if it is ok to keep searching for superblocks
2296  *              -EINVAL if there is a bad superblock
2297  *              0 on success
2298  */
2299 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2300                                struct buffer_head *bh,
2301                                u32 blksz,
2302                                struct ocfs2_blockcheck_stats *stats)
2303 {
2304         int status = -EAGAIN;
2305
2306         mlog_entry_void();
2307
2308         if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2309                    strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2310                 /* We have to do a raw check of the feature here */
2311                 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2312                     OCFS2_FEATURE_INCOMPAT_META_ECC) {
2313                         status = ocfs2_block_check_validate(bh->b_data,
2314                                                             bh->b_size,
2315                                                             &di->i_check,
2316                                                             stats);
2317                         if (status)
2318                                 goto out;
2319                 }
2320                 status = -EINVAL;
2321                 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2322                         mlog(ML_ERROR, "found superblock with incorrect block "
2323                              "size: found %u, should be %u\n",
2324                              1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2325                                blksz);
2326                 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2327                            OCFS2_MAJOR_REV_LEVEL ||
2328                            le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2329                            OCFS2_MINOR_REV_LEVEL) {
2330                         mlog(ML_ERROR, "found superblock with bad version: "
2331                              "found %u.%u, should be %u.%u\n",
2332                              le16_to_cpu(di->id2.i_super.s_major_rev_level),
2333                              le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2334                              OCFS2_MAJOR_REV_LEVEL,
2335                              OCFS2_MINOR_REV_LEVEL);
2336                 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2337                         mlog(ML_ERROR, "bad block number on superblock: "
2338                              "found %llu, should be %llu\n",
2339                              (unsigned long long)le64_to_cpu(di->i_blkno),
2340                              (unsigned long long)bh->b_blocknr);
2341                 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2342                             le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2343                         mlog(ML_ERROR, "bad cluster size found: %u\n",
2344                              1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2345                 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2346                         mlog(ML_ERROR, "bad root_blkno: 0\n");
2347                 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2348                         mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2349                 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2350                         mlog(ML_ERROR,
2351                              "Superblock slots found greater than file system "
2352                              "maximum: found %u, max %u\n",
2353                              le16_to_cpu(di->id2.i_super.s_max_slots),
2354                              OCFS2_MAX_SLOTS);
2355                 } else {
2356                         /* found it! */
2357                         status = 0;
2358                 }
2359         }
2360
2361 out:
2362         mlog_exit(status);
2363         return status;
2364 }
2365
2366 static int ocfs2_check_volume(struct ocfs2_super *osb)
2367 {
2368         int status;
2369         int dirty;
2370         int local;
2371         struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2372                                                   * recover
2373                                                   * ourselves. */
2374
2375         mlog_entry_void();
2376
2377         /* Init our journal object. */
2378         status = ocfs2_journal_init(osb->journal, &dirty);
2379         if (status < 0) {
2380                 mlog(ML_ERROR, "Could not initialize journal!\n");
2381                 goto finally;
2382         }
2383
2384         /* If the journal was unmounted cleanly then we don't want to
2385          * recover anything. Otherwise, journal_load will do that
2386          * dirty work for us :) */
2387         if (!dirty) {
2388                 status = ocfs2_journal_wipe(osb->journal, 0);
2389                 if (status < 0) {
2390                         mlog_errno(status);
2391                         goto finally;
2392                 }
2393         } else {
2394                 mlog(ML_NOTICE, "File system was not unmounted cleanly, "
2395                      "recovering volume.\n");
2396         }
2397
2398         local = ocfs2_mount_local(osb);
2399
2400         /* will play back anything left in the journal. */
2401         status = ocfs2_journal_load(osb->journal, local, dirty);
2402         if (status < 0) {
2403                 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2404                 goto finally;
2405         }
2406
2407         if (dirty) {
2408                 /* recover my local alloc if we didn't unmount cleanly. */
2409                 status = ocfs2_begin_local_alloc_recovery(osb,
2410                                                           osb->slot_num,
2411                                                           &local_alloc);
2412                 if (status < 0) {
2413                         mlog_errno(status);
2414                         goto finally;
2415                 }
2416                 /* we complete the recovery process after we've marked
2417                  * ourselves as mounted. */
2418         }
2419
2420         mlog(0, "Journal loaded.\n");
2421
2422         status = ocfs2_load_local_alloc(osb);
2423         if (status < 0) {
2424                 mlog_errno(status);
2425                 goto finally;
2426         }
2427
2428         if (dirty) {
2429                 /* Recovery will be completed after we've mounted the
2430                  * rest of the volume. */
2431                 osb->dirty = 1;
2432                 osb->local_alloc_copy = local_alloc;
2433                 local_alloc = NULL;
2434         }
2435
2436         /* go through each journal, trylock it and if you get the
2437          * lock, and it's marked as dirty, set the bit in the recover
2438          * map and launch a recovery thread for it. */
2439         status = ocfs2_mark_dead_nodes(osb);
2440         if (status < 0) {
2441                 mlog_errno(status);
2442                 goto finally;
2443         }
2444
2445         status = ocfs2_compute_replay_slots(osb);
2446         if (status < 0)
2447                 mlog_errno(status);
2448
2449 finally:
2450         if (local_alloc)
2451                 kfree(local_alloc);
2452
2453         mlog_exit(status);
2454         return status;
2455 }
2456
2457 /*
2458  * The routine gets called from dismount or close whenever a dismount on
2459  * volume is requested and the osb open count becomes 1.
2460  * It will remove the osb from the global list and also free up all the
2461  * initialized resources and fileobject.
2462  */
2463 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2464 {
2465         mlog_entry_void();
2466
2467         /* This function assumes that the caller has the main osb resource */
2468
2469         ocfs2_free_slot_info(osb);
2470
2471         kfree(osb->osb_orphan_wipes);
2472         kfree(osb->slot_recovery_generations);
2473         /* FIXME
2474          * This belongs in journal shutdown, but because we have to
2475          * allocate osb->journal at the start of ocfs2_initalize_osb(),
2476          * we free it here.
2477          */
2478         kfree(osb->journal);
2479         if (osb->local_alloc_copy)
2480                 kfree(osb->local_alloc_copy);
2481         kfree(osb->uuid_str);
2482         ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2483         memset(osb, 0, sizeof(struct ocfs2_super));
2484
2485         mlog_exit_void();
2486 }
2487
2488 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2489  * panic(). We do not support continue-on-error operation. */
2490 static void ocfs2_handle_error(struct super_block *sb)
2491 {
2492         struct ocfs2_super *osb = OCFS2_SB(sb);
2493
2494         if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2495                 panic("OCFS2: (device %s): panic forced after error\n",
2496                       sb->s_id);
2497
2498         ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2499
2500         if (sb->s_flags & MS_RDONLY &&
2501             (ocfs2_is_soft_readonly(osb) ||
2502              ocfs2_is_hard_readonly(osb)))
2503                 return;
2504
2505         printk(KERN_CRIT "File system is now read-only due to the potential "
2506                "of on-disk corruption. Please run fsck.ocfs2 once the file "
2507                "system is unmounted.\n");
2508         sb->s_flags |= MS_RDONLY;
2509         ocfs2_set_ro_flag(osb, 0);
2510 }
2511
2512 static char error_buf[1024];
2513
2514 void __ocfs2_error(struct super_block *sb,
2515                    const char *function,
2516                    const char *fmt, ...)
2517 {
2518         va_list args;
2519
2520         va_start(args, fmt);
2521         vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2522         va_end(args);
2523
2524         /* Not using mlog here because we want to show the actual
2525          * function the error came from. */
2526         printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2527                sb->s_id, function, error_buf);
2528
2529         ocfs2_handle_error(sb);
2530 }
2531
2532 /* Handle critical errors. This is intentionally more drastic than
2533  * ocfs2_handle_error, so we only use for things like journal errors,
2534  * etc. */
2535 void __ocfs2_abort(struct super_block* sb,
2536                    const char *function,
2537                    const char *fmt, ...)
2538 {
2539         va_list args;
2540
2541         va_start(args, fmt);
2542         vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2543         va_end(args);
2544
2545         printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2546                sb->s_id, function, error_buf);
2547
2548         /* We don't have the cluster support yet to go straight to
2549          * hard readonly in here. Until then, we want to keep
2550          * ocfs2_abort() so that we can at least mark critical
2551          * errors.
2552          *
2553          * TODO: This should abort the journal and alert other nodes
2554          * that our slot needs recovery. */
2555
2556         /* Force a panic(). This stinks, but it's better than letting
2557          * things continue without having a proper hard readonly
2558          * here. */
2559         if (!ocfs2_mount_local(OCFS2_SB(sb)))
2560                 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2561         ocfs2_handle_error(sb);
2562 }
2563
2564 /*
2565  * Void signal blockers, because in-kernel sigprocmask() only fails
2566  * when SIG_* is wrong.
2567  */
2568 void ocfs2_block_signals(sigset_t *oldset)
2569 {
2570         int rc;
2571         sigset_t blocked;
2572
2573         sigfillset(&blocked);
2574         rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2575         BUG_ON(rc);
2576 }
2577
2578 void ocfs2_unblock_signals(sigset_t *oldset)
2579 {
2580         int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2581         BUG_ON(rc);
2582 }
2583
2584 module_init(ocfs2_init);
2585 module_exit(ocfs2_exit);