fs/ocfs2/super.c

   1 /* -*- mode: c; c-basic-offset: 8; -*-
   2  * vim: noexpandtab sw=8 ts=8 sts=0:
   3  *
   4  * super.c
   5  *
   6  * load/unload driver, mount/dismount volumes
   7  *
   8  * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
   9  *
  10  * This program is free software; you can redistribute it and/or
  11  * modify it under the terms of the GNU General Public
  12  * License as published by the Free Software Foundation; either
  13  * version 2 of the License, or (at your option) any later version.
  14  *
  15  * This program is distributed in the hope that it will be useful,
  16  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  17  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  18  * General Public License for more details.
  19  *
  20  * You should have received a copy of the GNU General Public
  21  * License along with this program; if not, write to the
  22  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  23  * Boston, MA 021110-1307, USA.
  24  */
  25
  26 #include <linux/module.h>
  27 #include <linux/fs.h>
  28 #include <linux/types.h>
  29 #include <linux/slab.h>
  30 #include <linux/highmem.h>
  31 #include <linux/init.h>
  32 #include <linux/random.h>
  33 #include <linux/statfs.h>
  34 #include <linux/moduleparam.h>
  35 #include <linux/blkdev.h>
  36 #include <linux/socket.h>
  37 #include <linux/inet.h>
  38 #include <linux/parser.h>
  39 #include <linux/crc32.h>
  40 #include <linux/debugfs.h>
  41 #include <linux/mount.h>
  42 #include <linux/seq_file.h>
  43 #include <linux/quotaops.h>
  44 #include <linux/smp_lock.h>
  45
  46 #define MLOG_MASK_PREFIX ML_SUPER
  47 #include <cluster/masklog.h>
  48
  49 #include "ocfs2.h"
  50
  51 /* this should be the only file to include a version 1 header */
  52 #include "ocfs1_fs_compat.h"
  53
  54 #include "alloc.h"
  55 #include "blockcheck.h"
  56 #include "dlmglue.h"
  57 #include "export.h"
  58 #include "extent_map.h"
  59 #include "heartbeat.h"
  60 #include "inode.h"
  61 #include "journal.h"
  62 #include "localalloc.h"
  63 #include "namei.h"
  64 #include "slot_map.h"
  65 #include "super.h"
  66 #include "sysfile.h"
  67 #include "uptodate.h"
  68 #include "ver.h"
  69 #include "xattr.h"
  70 #include "quota.h"
  71 #include "refcounttree.h"
  72 #include "suballoc.h"
  73
  74 #include "buffer_head_io.h"
  75
  76 static struct kmem_cache *ocfs2_inode_cachep = NULL;
  77 struct kmem_cache *ocfs2_dquot_cachep;
  78 struct kmem_cache *ocfs2_qf_chunk_cachep;
  79
  80 /* OCFS2 needs to schedule several differnt types of work which
  81  * require cluster locking, disk I/O, recovery waits, etc. Since these
  82  * types of work tend to be heavy we avoid using the kernel events
  83  * workqueue and schedule on our own. */
  84 struct workqueue_struct *ocfs2_wq = NULL;
  85
  86 static struct dentry *ocfs2_debugfs_root = NULL;
  87
  88 MODULE_AUTHOR("Oracle");
  89 MODULE_LICENSE("GPL");
  90
  91 struct mount_options
  92 {
  93         unsigned long   commit_interval;
  94         unsigned long   mount_opt;
  95         unsigned int    atime_quantum;
  96         signed short    slot;
  97         int             localalloc_opt;
  98         unsigned int    resv_level;
  99         int             dir_resv_level;
 100         char            cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
 101 };
 102
 103 static int ocfs2_parse_options(struct super_block *sb, char *options,
 104                                struct mount_options *mopt,
 105                                int is_remount);
 106 static int ocfs2_check_set_options(struct super_block *sb,
 107                                    struct mount_options *options);
 108 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt);
 109 static void ocfs2_put_super(struct super_block *sb);
 110 static int ocfs2_mount_volume(struct super_block *sb);
 111 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
 112 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
 113 static int ocfs2_initialize_mem_caches(void);
 114 static void ocfs2_free_mem_caches(void);
 115 static void ocfs2_delete_osb(struct ocfs2_super *osb);
 116
 117 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
 118
 119 static int ocfs2_sync_fs(struct super_block *sb, int wait);
 120
 121 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
 122 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
 123 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
 124 static int ocfs2_check_volume(struct ocfs2_super *osb);
 125 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
 126                                struct buffer_head *bh,
 127                                u32 sectsize,
 128                                struct ocfs2_blockcheck_stats *stats);
 129 static int ocfs2_initialize_super(struct super_block *sb,
 130                                   struct buffer_head *bh,
 131                                   int sector_size,
 132                                   struct ocfs2_blockcheck_stats *stats);
 133 static int ocfs2_get_sector(struct super_block *sb,
 134                             struct buffer_head **bh,
 135                             int block,
 136                             int sect_size);
 137 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
 138 static void ocfs2_destroy_inode(struct inode *inode);
 139 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
 140 static int ocfs2_enable_quotas(struct ocfs2_super *osb);
 141 static void ocfs2_disable_quotas(struct ocfs2_super *osb);
 142
 143 static const struct super_operations ocfs2_sops = {
 144         .statfs         = ocfs2_statfs,
 145         .alloc_inode    = ocfs2_alloc_inode,
 146         .destroy_inode  = ocfs2_destroy_inode,
 147         .drop_inode     = ocfs2_drop_inode,
 148         .evict_inode    = ocfs2_evict_inode,
 149         .sync_fs        = ocfs2_sync_fs,
 150         .put_super      = ocfs2_put_super,
 151         .remount_fs     = ocfs2_remount,
 152         .show_options   = ocfs2_show_options,
 153         .quota_read     = ocfs2_quota_read,
 154         .quota_write    = ocfs2_quota_write,
 155 };
 156
 157 enum {
 158         Opt_barrier,
 159         Opt_err_panic,
 160         Opt_err_ro,
 161         Opt_intr,
 162         Opt_nointr,
 163         Opt_hb_none,
 164         Opt_hb_local,
 165         Opt_hb_global,
 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_coherency_buffered,
 182         Opt_coherency_full,
 183         Opt_resv_level,
 184         Opt_dir_resv_level,
 185         Opt_err,
 186 };
 187
 188 static const match_table_t tokens = {
 189         {Opt_barrier, "barrier=%u"},
 190         {Opt_err_panic, "errors=panic"},
 191         {Opt_err_ro, "errors=remount-ro"},
 192         {Opt_intr, "intr"},
 193         {Opt_nointr, "nointr"},
 194         {Opt_hb_none, OCFS2_HB_NONE},
 195         {Opt_hb_local, OCFS2_HB_LOCAL},
 196         {Opt_hb_global, OCFS2_HB_GLOBAL},
 197         {Opt_data_ordered, "data=ordered"},
 198         {Opt_data_writeback, "data=writeback"},
 199         {Opt_atime_quantum, "atime_quantum=%u"},
 200         {Opt_slot, "preferred_slot=%u"},
 201         {Opt_commit, "commit=%u"},
 202         {Opt_localalloc, "localalloc=%d"},
 203         {Opt_localflocks, "localflocks"},
 204         {Opt_stack, "cluster_stack=%s"},
 205         {Opt_user_xattr, "user_xattr"},
 206         {Opt_nouser_xattr, "nouser_xattr"},
 207         {Opt_inode64, "inode64"},
 208         {Opt_acl, "acl"},
 209         {Opt_noacl, "noacl"},
 210         {Opt_usrquota, "usrquota"},
 211         {Opt_grpquota, "grpquota"},
 212         {Opt_coherency_buffered, "coherency=buffered"},
 213         {Opt_coherency_full, "coherency=full"},
 214         {Opt_resv_level, "resv_level=%u"},
 215         {Opt_dir_resv_level, "dir_resv_level=%u"},
 216         {Opt_err, NULL}
 217 };
 218
 219 #ifdef CONFIG_DEBUG_FS
 220 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
 221 {
 222         struct ocfs2_cluster_connection *cconn = osb->cconn;
 223         struct ocfs2_recovery_map *rm = osb->recovery_map;
 224         struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
 225         int i, out = 0;
 226
 227         out += snprintf(buf + out, len - out,
 228                         "%10s => Id: %-s  Uuid: %-s  Gen: 0x%X  Label: %-s\n",
 229                         "Device", osb->dev_str, osb->uuid_str,
 230                         osb->fs_generation, osb->vol_label);
 231
 232         out += snprintf(buf + out, len - out,
 233                         "%10s => State: %d  Flags: 0x%lX\n", "Volume",
 234                         atomic_read(&osb->vol_state), osb->osb_flags);
 235
 236         out += snprintf(buf + out, len - out,
 237                         "%10s => Block: %lu  Cluster: %d\n", "Sizes",
 238                         osb->sb->s_blocksize, osb->s_clustersize);
 239
 240         out += snprintf(buf + out, len - out,
 241                         "%10s => Compat: 0x%X  Incompat: 0x%X  "
 242                         "ROcompat: 0x%X\n",
 243                         "Features", osb->s_feature_compat,
 244                         osb->s_feature_incompat, osb->s_feature_ro_compat);
 245
 246         out += snprintf(buf + out, len - out,
 247                         "%10s => Opts: 0x%lX  AtimeQuanta: %u\n", "Mount",
 248                         osb->s_mount_opt, osb->s_atime_quantum);
 249
 250         if (cconn) {
 251                 out += snprintf(buf + out, len - out,
 252                                 "%10s => Stack: %s  Name: %*s  "
 253                                 "Version: %d.%d\n", "Cluster",
 254                                 (*osb->osb_cluster_stack == '\0' ?
 255                                  "o2cb" : osb->osb_cluster_stack),
 256                                 cconn->cc_namelen, cconn->cc_name,
 257                                 cconn->cc_version.pv_major,
 258                                 cconn->cc_version.pv_minor);
 259         }
 260
 261         spin_lock(&osb->dc_task_lock);
 262         out += snprintf(buf + out, len - out,
 263                         "%10s => Pid: %d  Count: %lu  WakeSeq: %lu  "
 264                         "WorkSeq: %lu\n", "DownCnvt",
 265                         (osb->dc_task ?  task_pid_nr(osb->dc_task) : -1),
 266                         osb->blocked_lock_count, osb->dc_wake_sequence,
 267                         osb->dc_work_sequence);
 268         spin_unlock(&osb->dc_task_lock);
 269
 270         spin_lock(&osb->osb_lock);
 271         out += snprintf(buf + out, len - out, "%10s => Pid: %d  Nodes:",
 272                         "Recovery",
 273                         (osb->recovery_thread_task ?
 274                          task_pid_nr(osb->recovery_thread_task) : -1));
 275         if (rm->rm_used == 0)
 276                 out += snprintf(buf + out, len - out, " None\n");
 277         else {
 278                 for (i = 0; i < rm->rm_used; i++)
 279                         out += snprintf(buf + out, len - out, " %d",
 280                                         rm->rm_entries[i]);
 281                 out += snprintf(buf + out, len - out, "\n");
 282         }
 283         spin_unlock(&osb->osb_lock);
 284
 285         out += snprintf(buf + out, len - out,
 286                         "%10s => Pid: %d  Interval: %lu  Needs: %d\n", "Commit",
 287                         (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
 288                         osb->osb_commit_interval,
 289                         atomic_read(&osb->needs_checkpoint));
 290
 291         out += snprintf(buf + out, len - out,
 292                         "%10s => State: %d  TxnId: %lu  NumTxns: %d\n",
 293                         "Journal", osb->journal->j_state,
 294                         osb->journal->j_trans_id,
 295                         atomic_read(&osb->journal->j_num_trans));
 296
 297         out += snprintf(buf + out, len - out,
 298                         "%10s => GlobalAllocs: %d  LocalAllocs: %d  "
 299                         "SubAllocs: %d  LAWinMoves: %d  SAExtends: %d\n",
 300                         "Stats",
 301                         atomic_read(&osb->alloc_stats.bitmap_data),
 302                         atomic_read(&osb->alloc_stats.local_data),
 303                         atomic_read(&osb->alloc_stats.bg_allocs),
 304                         atomic_read(&osb->alloc_stats.moves),
 305                         atomic_read(&osb->alloc_stats.bg_extends));
 306
 307         out += snprintf(buf + out, len - out,
 308                         "%10s => State: %u  Descriptor: %llu  Size: %u bits  "
 309                         "Default: %u bits\n",
 310                         "LocalAlloc", osb->local_alloc_state,
 311                         (unsigned long long)osb->la_last_gd,
 312                         osb->local_alloc_bits, osb->local_alloc_default_bits);
 313
 314         spin_lock(&osb->osb_lock);
 315         out += snprintf(buf + out, len - out,
 316                         "%10s => InodeSlot: %d  StolenInodes: %d, "
 317                         "MetaSlot: %d  StolenMeta: %d\n", "Steal",
 318                         osb->s_inode_steal_slot,
 319                         atomic_read(&osb->s_num_inodes_stolen),
 320                         osb->s_meta_steal_slot,
 321                         atomic_read(&osb->s_num_meta_stolen));
 322         spin_unlock(&osb->osb_lock);
 323
 324         out += snprintf(buf + out, len - out, "OrphanScan => ");
 325         out += snprintf(buf + out, len - out, "Local: %u  Global: %u ",
 326                         os->os_count, os->os_seqno);
 327         out += snprintf(buf + out, len - out, " Last Scan: ");
 328         if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
 329                 out += snprintf(buf + out, len - out, "Disabled\n");
 330         else
 331                 out += snprintf(buf + out, len - out, "%lu seconds ago\n",
 332                                 (get_seconds() - os->os_scantime.tv_sec));
 333
 334         out += snprintf(buf + out, len - out, "%10s => %3s  %10s\n",
 335                         "Slots", "Num", "RecoGen");
 336         for (i = 0; i < osb->max_slots; ++i) {
 337                 out += snprintf(buf + out, len - out,
 338                                 "%10s  %c %3d  %10d\n",
 339                                 " ",
 340                                 (i == osb->slot_num ? '*' : ' '),
 341                                 i, osb->slot_recovery_generations[i]);
 342         }
 343
 344         return out;
 345 }
 346
 347 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 348 {
 349         struct ocfs2_super *osb = inode->i_private;
 350         char *buf = NULL;
 351
 352         buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
 353         if (!buf)
 354                 goto bail;
 355
 356         i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
 357
 358         file->private_data = buf;
 359
 360         return 0;
 361 bail:
 362         return -ENOMEM;
 363 }
 364
 365 static int ocfs2_debug_release(struct inode *inode, struct file *file)
 366 {
 367         kfree(file->private_data);
 368         return 0;
 369 }
 370
 371 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 372                                 size_t nbytes, loff_t *ppos)
 373 {
 374         return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
 375                                        i_size_read(file->f_mapping->host));
 376 }
 377 #else
 378 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 379 {
 380         return 0;
 381 }
 382 static int ocfs2_debug_release(struct inode *inode, struct file *file)
 383 {
 384         return 0;
 385 }
 386 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 387                                 size_t nbytes, loff_t *ppos)
 388 {
 389         return 0;
 390 }
 391 #endif  /* CONFIG_DEBUG_FS */
 392
 393 static const struct file_operations ocfs2_osb_debug_fops = {
 394         .open =         ocfs2_osb_debug_open,
 395         .release =      ocfs2_debug_release,
 396         .read =         ocfs2_debug_read,
 397         .llseek =       generic_file_llseek,
 398 };
 399
 400 static int ocfs2_sync_fs(struct super_block *sb, int wait)
 401 {
 402         int status;
 403         tid_t target;
 404         struct ocfs2_super *osb = OCFS2_SB(sb);
 405
 406         if (ocfs2_is_hard_readonly(osb))
 407                 return -EROFS;
 408
 409         if (wait) {
 410                 status = ocfs2_flush_truncate_log(osb);
 411                 if (status < 0)
 412                         mlog_errno(status);
 413         } else {
 414                 ocfs2_schedule_truncate_log_flush(osb, 0);
 415         }
 416
 417         if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
 418                                       &target)) {
 419                 if (wait)
 420                         jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
 421                                              target);
 422         }
 423         return 0;
 424 }
 425
 426 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
 427 {
 428         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
 429             && (ino == USER_QUOTA_SYSTEM_INODE
 430                 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
 431                 return 0;
 432         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
 433             && (ino == GROUP_QUOTA_SYSTEM_INODE
 434                 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
 435                 return 0;
 436         return 1;
 437 }
 438
 439 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
 440 {
 441         struct inode *new = NULL;
 442         int status = 0;
 443         int i;
 444
 445         mlog_entry_void();
 446
 447         new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 448         if (IS_ERR(new)) {
 449                 status = PTR_ERR(new);
 450                 mlog_errno(status);
 451                 goto bail;
 452         }
 453         osb->root_inode = new;
 454
 455         new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 456         if (IS_ERR(new)) {
 457                 status = PTR_ERR(new);
 458                 mlog_errno(status);
 459                 goto bail;
 460         }
 461         osb->sys_root_inode = new;
 462
 463         for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
 464              i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
 465                 if (!ocfs2_need_system_inode(osb, i))
 466                         continue;
 467                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 468                 if (!new) {
 469                         ocfs2_release_system_inodes(osb);
 470                         status = -EINVAL;
 471                         mlog_errno(status);
 472                         /* FIXME: Should ERROR_RO_FS */
 473                         mlog(ML_ERROR, "Unable to load system inode %d, "
 474                              "possibly corrupt fs?", i);
 475                         goto bail;
 476                 }
 477                 // the array now has one ref, so drop this one
 478                 iput(new);
 479         }
 480
 481 bail:
 482         mlog_exit(status);
 483         return status;
 484 }
 485
 486 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
 487 {
 488         struct inode *new = NULL;
 489         int status = 0;
 490         int i;
 491
 492         mlog_entry_void();
 493
 494         for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
 495              i < NUM_SYSTEM_INODES;
 496              i++) {
 497                 if (!ocfs2_need_system_inode(osb, i))
 498                         continue;
 499                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 500                 if (!new) {
 501                         ocfs2_release_system_inodes(osb);
 502                         status = -EINVAL;
 503                         mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
 504                              status, i, osb->slot_num);
 505                         goto bail;
 506                 }
 507                 /* the array now has one ref, so drop this one */
 508                 iput(new);
 509         }
 510
 511 bail:
 512         mlog_exit(status);
 513         return status;
 514 }
 515
 516 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
 517 {
 518         int i;
 519         struct inode *inode;
 520
 521         mlog_entry_void();
 522
 523         for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
 524                 inode = osb->global_system_inodes[i];
 525                 if (inode) {
 526                         iput(inode);
 527                         osb->global_system_inodes[i] = NULL;
 528                 }
 529         }
 530
 531         inode = osb->sys_root_inode;
 532         if (inode) {
 533                 iput(inode);
 534                 osb->sys_root_inode = NULL;
 535         }
 536
 537         inode = osb->root_inode;
 538         if (inode) {
 539                 iput(inode);
 540                 osb->root_inode = NULL;
 541         }
 542
 543         if (!osb->local_system_inodes)
 544                 goto out;
 545
 546         for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
 547                 if (osb->local_system_inodes[i]) {
 548                         iput(osb->local_system_inodes[i]);
 549                         osb->local_system_inodes[i] = NULL;
 550                 }
 551         }
 552
 553         kfree(osb->local_system_inodes);
 554         osb->local_system_inodes = NULL;
 555
 556 out:
 557         mlog_exit(0);
 558 }
 559
 560 /* We're allocating fs objects, use GFP_NOFS */
 561 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
 562 {
 563         struct ocfs2_inode_info *oi;
 564
 565         oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
 566         if (!oi)
 567                 return NULL;
 568
 569         jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
 570         return &oi->vfs_inode;
 571 }
 572
 573 static void ocfs2_destroy_inode(struct inode *inode)
 574 {
 575         kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
 576 }
 577
 578 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
 579                                                 unsigned int cbits)
 580 {
 581         unsigned int bytes = 1 << cbits;
 582         unsigned int trim = bytes;
 583         unsigned int bitshift = 32;
 584
 585         /*
 586          * i_size and all block offsets in ocfs2 are always 64 bits
 587          * wide. i_clusters is 32 bits, in cluster-sized units. So on
 588          * 64 bit platforms, cluster size will be the limiting factor.
 589          */
 590
 591 #if BITS_PER_LONG == 32
 592 # if defined(CONFIG_LBDAF)
 593         BUILD_BUG_ON(sizeof(sector_t) != 8);
 594         /*
 595          * We might be limited by page cache size.
 596          */
 597         if (bytes > PAGE_CACHE_SIZE) {
 598                 bytes = PAGE_CACHE_SIZE;
 599                 trim = 1;
 600                 /*
 601                  * Shift by 31 here so that we don't get larger than
 602                  * MAX_LFS_FILESIZE
 603                  */
 604                 bitshift = 31;
 605         }
 606 # else
 607         /*
 608          * We are limited by the size of sector_t. Use block size, as
 609          * that's what we expose to the VFS.
 610          */
 611         bytes = 1 << bbits;
 612         trim = 1;
 613         bitshift = 31;
 614 # endif
 615 #endif
 616
 617         /*
 618          * Trim by a whole cluster when we can actually approach the
 619          * on-disk limits. Otherwise we can overflow i_clusters when
 620          * an extent start is at the max offset.
 621          */
 622         return (((unsigned long long)bytes) << bitshift) - trim;
 623 }
 624
 625 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
 626 {
 627         int incompat_features;
 628         int ret = 0;
 629         struct mount_options parsed_options;
 630         struct ocfs2_super *osb = OCFS2_SB(sb);
 631         u32 tmp;
 632
 633         if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
 634             !ocfs2_check_set_options(sb, &parsed_options)) {
 635                 ret = -EINVAL;
 636                 goto out;
 637         }
 638
 639         tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
 640                 OCFS2_MOUNT_HB_NONE;
 641         if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
 642                 ret = -EINVAL;
 643                 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
 644                 goto out;
 645         }
 646
 647         if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
 648             (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
 649                 ret = -EINVAL;
 650                 mlog(ML_ERROR, "Cannot change data mode on remount\n");
 651                 goto out;
 652         }
 653
 654         /* Probably don't want this on remount; it might
 655          * mess with other nodes */
 656         if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
 657             (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
 658                 ret = -EINVAL;
 659                 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
 660                 goto out;
 661         }
 662
 663         /* We're going to/from readonly mode. */
 664         if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
 665                 /* Disable quota accounting before remounting RO */
 666                 if (*flags & MS_RDONLY) {
 667                         ret = ocfs2_susp_quotas(osb, 0);
 668                         if (ret < 0)
 669                                 goto out;
 670                 }
 671                 /* Lock here so the check of HARD_RO and the potential
 672                  * setting of SOFT_RO is atomic. */
 673                 spin_lock(&osb->osb_lock);
 674                 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
 675                         mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
 676                         ret = -EROFS;
 677                         goto unlock_osb;
 678                 }
 679
 680                 if (*flags & MS_RDONLY) {
 681                         mlog(0, "Going to ro mode.\n");
 682                         sb->s_flags |= MS_RDONLY;
 683                         osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 684                 } else {
 685                         mlog(0, "Making ro filesystem writeable.\n");
 686
 687                         if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
 688                                 mlog(ML_ERROR, "Cannot remount RDWR "
 689                                      "filesystem due to previous errors.\n");
 690                                 ret = -EROFS;
 691                                 goto unlock_osb;
 692                         }
 693                         incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
 694                         if (incompat_features) {
 695                                 mlog(ML_ERROR, "Cannot remount RDWR because "
 696                                      "of unsupported optional features "
 697                                      "(%x).\n", incompat_features);
 698                                 ret = -EINVAL;
 699                                 goto unlock_osb;
 700                         }
 701                         sb->s_flags &= ~MS_RDONLY;
 702                         osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
 703                 }
 704 unlock_osb:
 705                 spin_unlock(&osb->osb_lock);
 706                 /* Enable quota accounting after remounting RW */
 707                 if (!ret && !(*flags & MS_RDONLY)) {
 708                         if (sb_any_quota_suspended(sb))
 709                                 ret = ocfs2_susp_quotas(osb, 1);
 710                         else
 711                                 ret = ocfs2_enable_quotas(osb);
 712                         if (ret < 0) {
 713                                 /* Return back changes... */
 714                                 spin_lock(&osb->osb_lock);
 715                                 sb->s_flags |= MS_RDONLY;
 716                                 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 717                                 spin_unlock(&osb->osb_lock);
 718                                 goto out;
 719                         }
 720                 }
 721         }
 722
 723         if (!ret) {
 724                 /* Only save off the new mount options in case of a successful
 725                  * remount. */
 726                 osb->s_mount_opt = parsed_options.mount_opt;
 727                 osb->s_atime_quantum = parsed_options.atime_quantum;
 728                 osb->preferred_slot = parsed_options.slot;
 729                 if (parsed_options.commit_interval)
 730                         osb->osb_commit_interval = parsed_options.commit_interval;
 731
 732                 if (!ocfs2_is_hard_readonly(osb))
 733                         ocfs2_set_journal_params(osb);
 734
 735                 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 736                         ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
 737                                                         MS_POSIXACL : 0);
 738         }
 739 out:
 740         return ret;
 741 }
 742
 743 static int ocfs2_sb_probe(struct super_block *sb,
 744                           struct buffer_head **bh,
 745                           int *sector_size,
 746                           struct ocfs2_blockcheck_stats *stats)
 747 {
 748         int status, tmpstat;
 749         struct ocfs1_vol_disk_hdr *hdr;
 750         struct ocfs2_dinode *di;
 751         int blksize;
 752
 753         *bh = NULL;
 754
 755         /* may be > 512 */
 756         *sector_size = bdev_logical_block_size(sb->s_bdev);
 757         if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
 758                 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
 759                      *sector_size, OCFS2_MAX_BLOCKSIZE);
 760                 status = -EINVAL;
 761                 goto bail;
 762         }
 763
 764         /* Can this really happen? */
 765         if (*sector_size < OCFS2_MIN_BLOCKSIZE)
 766                 *sector_size = OCFS2_MIN_BLOCKSIZE;
 767
 768         /* check block zero for old format */
 769         status = ocfs2_get_sector(sb, bh, 0, *sector_size);
 770         if (status < 0) {
 771                 mlog_errno(status);
 772                 goto bail;
 773         }
 774         hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
 775         if (hdr->major_version == OCFS1_MAJOR_VERSION) {
 776                 mlog(ML_ERROR, "incompatible version: %u.%u\n",
 777                      hdr->major_version, hdr->minor_version);
 778                 status = -EINVAL;
 779         }
 780         if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
 781                    strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
 782                 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
 783                      hdr->signature);
 784                 status = -EINVAL;
 785         }
 786         brelse(*bh);
 787         *bh = NULL;
 788         if (status < 0) {
 789                 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
 790                      "upgraded before mounting with ocfs v2\n");
 791                 goto bail;
 792         }
 793
 794         /*
 795          * Now check at magic offset for 512, 1024, 2048, 4096
 796          * blocksizes.  4096 is the maximum blocksize because it is
 797          * the minimum clustersize.
 798          */
 799         status = -EINVAL;
 800         for (blksize = *sector_size;
 801              blksize <= OCFS2_MAX_BLOCKSIZE;
 802              blksize <<= 1) {
 803                 tmpstat = ocfs2_get_sector(sb, bh,
 804                                            OCFS2_SUPER_BLOCK_BLKNO,
 805                                            blksize);
 806                 if (tmpstat < 0) {
 807                         status = tmpstat;
 808                         mlog_errno(status);
 809                         break;
 810                 }
 811                 di = (struct ocfs2_dinode *) (*bh)->b_data;
 812                 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
 813                 spin_lock_init(&stats->b_lock);
 814                 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
 815                 if (tmpstat < 0) {
 816                         brelse(*bh);
 817                         *bh = NULL;
 818                 }
 819                 if (tmpstat != -EAGAIN) {
 820                         status = tmpstat;
 821                         break;
 822                 }
 823         }
 824
 825 bail:
 826         return status;
 827 }
 828
 829 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
 830 {
 831         u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
 832
 833         if (osb->s_mount_opt & hb_enabled) {
 834                 if (ocfs2_mount_local(osb)) {
 835                         mlog(ML_ERROR, "Cannot heartbeat on a locally "
 836                              "mounted device.\n");
 837                         return -EINVAL;
 838                 }
 839                 if (ocfs2_userspace_stack(osb)) {
 840                         mlog(ML_ERROR, "Userspace stack expected, but "
 841                              "o2cb heartbeat arguments passed to mount\n");
 842                         return -EINVAL;
 843                 }
 844                 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
 845                      !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
 846                     ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
 847                      ocfs2_cluster_o2cb_global_heartbeat(osb))) {
 848                         mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
 849                         return -EINVAL;
 850                 }
 851         }
 852
 853         if (!(osb->s_mount_opt & hb_enabled)) {
 854                 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
 855                     !ocfs2_userspace_stack(osb)) {
 856                         mlog(ML_ERROR, "Heartbeat has to be started to mount "
 857                              "a read-write clustered device.\n");
 858                         return -EINVAL;
 859                 }
 860         }
 861
 862         return 0;
 863 }
 864
 865 /*
 866  * If we're using a userspace stack, mount should have passed
 867  * a name that matches the disk.  If not, mount should not
 868  * have passed a stack.
 869  */
 870 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
 871                                         struct mount_options *mopt)
 872 {
 873         if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
 874                 mlog(ML_ERROR,
 875                      "cluster stack passed to mount, but this filesystem "
 876                      "does not support it\n");
 877                 return -EINVAL;
 878         }
 879
 880         if (ocfs2_userspace_stack(osb) &&
 881             strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
 882                     OCFS2_STACK_LABEL_LEN)) {
 883                 mlog(ML_ERROR,
 884                      "cluster stack passed to mount (\"%s\") does not "
 885                      "match the filesystem (\"%s\")\n",
 886                      mopt->cluster_stack,
 887                      osb->osb_cluster_stack);
 888                 return -EINVAL;
 889         }
 890
 891         return 0;
 892 }
 893
 894 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
 895 {
 896         int type;
 897         struct super_block *sb = osb->sb;
 898         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 899                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 900         int status = 0;
 901
 902         for (type = 0; type < MAXQUOTAS; type++) {
 903                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 904                         continue;
 905                 if (unsuspend)
 906                         status = dquot_resume(sb, type);
 907                 else {
 908                         struct ocfs2_mem_dqinfo *oinfo;
 909
 910                         /* Cancel periodic syncing before suspending */
 911                         oinfo = sb_dqinfo(sb, type)->dqi_priv;
 912                         cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 913                         status = dquot_suspend(sb, type);
 914                 }
 915                 if (status < 0)
 916                         break;
 917         }
 918         if (status < 0)
 919                 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
 920                      "remount (error = %d).\n", status);
 921         return status;
 922 }
 923
 924 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
 925 {
 926         struct inode *inode[MAXQUOTAS] = { NULL, NULL };
 927         struct super_block *sb = osb->sb;
 928         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 929                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 930         unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
 931                                         LOCAL_GROUP_QUOTA_SYSTEM_INODE };
 932         int status;
 933         int type;
 934
 935         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
 936         for (type = 0; type < MAXQUOTAS; type++) {
 937                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 938                         continue;
 939                 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
 940                                                         osb->slot_num);
 941                 if (!inode[type]) {
 942                         status = -ENOENT;
 943                         goto out_quota_off;
 944                 }
 945                 status = dquot_enable(inode[type], type, QFMT_OCFS2,
 946                                       DQUOT_USAGE_ENABLED);
 947                 if (status < 0)
 948                         goto out_quota_off;
 949         }
 950
 951         for (type = 0; type < MAXQUOTAS; type++)
 952                 iput(inode[type]);
 953         return 0;
 954 out_quota_off:
 955         ocfs2_disable_quotas(osb);
 956         for (type = 0; type < MAXQUOTAS; type++)
 957                 iput(inode[type]);
 958         mlog_errno(status);
 959         return status;
 960 }
 961
 962 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
 963 {
 964         int type;
 965         struct inode *inode;
 966         struct super_block *sb = osb->sb;
 967         struct ocfs2_mem_dqinfo *oinfo;
 968
 969         /* We mostly ignore errors in this function because there's not much
 970          * we can do when we see them */
 971         for (type = 0; type < MAXQUOTAS; type++) {
 972                 if (!sb_has_quota_loaded(sb, type))
 973                         continue;
 974                 /* Cancel periodic syncing before we grab dqonoff_mutex */
 975                 oinfo = sb_dqinfo(sb, type)->dqi_priv;
 976                 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 977                 inode = igrab(sb->s_dquot.files[type]);
 978                 /* Turn off quotas. This will remove all dquot structures from
 979                  * memory and so they will be automatically synced to global
 980                  * quota files */
 981                 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
 982                                         DQUOT_LIMITS_ENABLED);
 983                 if (!inode)
 984                         continue;
 985                 iput(inode);
 986         }
 987 }
 988
 989 /* Handle quota on quotactl */
 990 static int ocfs2_quota_on(struct super_block *sb, int type, int format_id,
 991                           char *path)
 992 {
 993         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 994                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 995
 996         if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 997                 return -EINVAL;
 998
 999         return dquot_enable(sb_dqopt(sb)->files[type], type,
1000                             format_id, DQUOT_LIMITS_ENABLED);
1001 }
1002
1003 /* Handle quota off quotactl */
1004 static int ocfs2_quota_off(struct super_block *sb, int type)
1005 {
1006         return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
1007 }
1008
1009 static const struct quotactl_ops ocfs2_quotactl_ops = {
1010         .quota_on       = ocfs2_quota_on,
1011         .quota_off      = ocfs2_quota_off,
1012         .quota_sync     = dquot_quota_sync,
1013         .get_info       = dquot_get_dqinfo,
1014         .set_info       = dquot_set_dqinfo,
1015         .get_dqblk      = dquot_get_dqblk,
1016         .set_dqblk      = dquot_set_dqblk,
1017 };
1018
1019 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
1020 {
1021         struct dentry *root;
1022         int status, sector_size;
1023         struct mount_options parsed_options;
1024         struct inode *inode = NULL;
1025         struct ocfs2_super *osb = NULL;
1026         struct buffer_head *bh = NULL;
1027         char nodestr[8];
1028         struct ocfs2_blockcheck_stats stats;
1029
1030         mlog_entry("%p, %p, %i", sb, data, silent);
1031
1032         if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
1033                 status = -EINVAL;
1034                 goto read_super_error;
1035         }
1036
1037         /* probe for superblock */
1038         status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
1039         if (status < 0) {
1040                 mlog(ML_ERROR, "superblock probe failed!\n");
1041                 goto read_super_error;
1042         }
1043
1044         status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1045         osb = OCFS2_SB(sb);
1046         if (status < 0) {
1047                 mlog_errno(status);
1048                 goto read_super_error;
1049         }
1050         brelse(bh);
1051         bh = NULL;
1052
1053         if (!ocfs2_check_set_options(sb, &parsed_options)) {
1054                 status = -EINVAL;
1055                 goto read_super_error;
1056         }
1057         osb->s_mount_opt = parsed_options.mount_opt;
1058         osb->s_atime_quantum = parsed_options.atime_quantum;
1059         osb->preferred_slot = parsed_options.slot;
1060         osb->osb_commit_interval = parsed_options.commit_interval;
1061
1062         ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1063         osb->osb_resv_level = parsed_options.resv_level;
1064         osb->osb_dir_resv_level = parsed_options.resv_level;
1065         if (parsed_options.dir_resv_level == -1)
1066                 osb->osb_dir_resv_level = parsed_options.resv_level;
1067         else
1068                 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1069
1070         status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1071         if (status)
1072                 goto read_super_error;
1073
1074         sb->s_magic = OCFS2_SUPER_MAGIC;
1075
1076         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1077                 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1078
1079         /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1080          * heartbeat=none */
1081         if (bdev_read_only(sb->s_bdev)) {
1082                 if (!(sb->s_flags & MS_RDONLY)) {
1083                         status = -EACCES;
1084                         mlog(ML_ERROR, "Readonly device detected but readonly "
1085                              "mount was not specified.\n");
1086                         goto read_super_error;
1087                 }
1088
1089                 /* You should not be able to start a local heartbeat
1090                  * on a readonly device. */
1091                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1092                         status = -EROFS;
1093                         mlog(ML_ERROR, "Local heartbeat specified on readonly "
1094                              "device.\n");
1095                         goto read_super_error;
1096                 }
1097
1098                 status = ocfs2_check_journals_nolocks(osb);
1099                 if (status < 0) {
1100                         if (status == -EROFS)
1101                                 mlog(ML_ERROR, "Recovery required on readonly "
1102                                      "file system, but write access is "
1103                                      "unavailable.\n");
1104                         else
1105                                 mlog_errno(status);
1106                         goto read_super_error;
1107                 }
1108
1109                 ocfs2_set_ro_flag(osb, 1);
1110
1111                 printk(KERN_NOTICE "Readonly device detected. No cluster "
1112                        "services will be utilized for this mount. Recovery "
1113                        "will be skipped.\n");
1114         }
1115
1116         if (!ocfs2_is_hard_readonly(osb)) {
1117                 if (sb->s_flags & MS_RDONLY)
1118                         ocfs2_set_ro_flag(osb, 0);
1119         }
1120
1121         status = ocfs2_verify_heartbeat(osb);
1122         if (status < 0) {
1123                 mlog_errno(status);
1124                 goto read_super_error;
1125         }
1126
1127         osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1128                                                  ocfs2_debugfs_root);
1129         if (!osb->osb_debug_root) {
1130                 status = -EINVAL;
1131                 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1132                 goto read_super_error;
1133         }
1134
1135         osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1136                                             osb->osb_debug_root,
1137                                             osb,
1138                                             &ocfs2_osb_debug_fops);
1139         if (!osb->osb_ctxt) {
1140                 status = -EINVAL;
1141                 mlog_errno(status);
1142                 goto read_super_error;
1143         }
1144
1145         if (ocfs2_meta_ecc(osb)) {
1146                 status = ocfs2_blockcheck_stats_debugfs_install(
1147                                                 &osb->osb_ecc_stats,
1148                                                 osb->osb_debug_root);
1149                 if (status) {
1150                         mlog(ML_ERROR,
1151                              "Unable to create blockcheck statistics "
1152                              "files\n");
1153                         goto read_super_error;
1154                 }
1155         }
1156
1157         status = ocfs2_mount_volume(sb);
1158         if (osb->root_inode)
1159                 inode = igrab(osb->root_inode);
1160
1161         if (status < 0)
1162                 goto read_super_error;
1163
1164         if (!inode) {
1165                 status = -EIO;
1166                 mlog_errno(status);
1167                 goto read_super_error;
1168         }
1169
1170         root = d_alloc_root(inode);
1171         if (!root) {
1172                 status = -ENOMEM;
1173                 mlog_errno(status);
1174                 goto read_super_error;
1175         }
1176
1177         sb->s_root = root;
1178
1179         ocfs2_complete_mount_recovery(osb);
1180
1181         if (ocfs2_mount_local(osb))
1182                 snprintf(nodestr, sizeof(nodestr), "local");
1183         else
1184                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1185
1186         printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1187                "with %s data mode.\n",
1188                osb->dev_str, nodestr, osb->slot_num,
1189                osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1190                "ordered");
1191
1192         atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1193         wake_up(&osb->osb_mount_event);
1194
1195         /* Now we can initialize quotas because we can afford to wait
1196          * for cluster locks recovery now. That also means that truncation
1197          * log recovery can happen but that waits for proper quota setup */
1198         if (!(sb->s_flags & MS_RDONLY)) {
1199                 status = ocfs2_enable_quotas(osb);
1200                 if (status < 0) {
1201                         /* We have to err-out specially here because
1202                          * s_root is already set */
1203                         mlog_errno(status);
1204                         atomic_set(&osb->vol_state, VOLUME_DISABLED);
1205                         wake_up(&osb->osb_mount_event);
1206                         mlog_exit(status);
1207                         return status;
1208                 }
1209         }
1210
1211         ocfs2_complete_quota_recovery(osb);
1212
1213         /* Now we wake up again for processes waiting for quotas */
1214         atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1215         wake_up(&osb->osb_mount_event);
1216
1217         /* Start this when the mount is almost sure of being successful */
1218         ocfs2_orphan_scan_start(osb);
1219
1220         mlog_exit(status);
1221         return status;
1222
1223 read_super_error:
1224         brelse(bh);
1225
1226         if (inode)
1227                 iput(inode);
1228
1229         if (osb) {
1230                 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1231                 wake_up(&osb->osb_mount_event);
1232                 ocfs2_dismount_volume(sb, 1);
1233         }
1234
1235         mlog_exit(status);
1236         return status;
1237 }
1238
1239 static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
1240                         int flags,
1241                         const char *dev_name,
1242                         void *data)
1243 {
1244         return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
1245 }
1246
1247 static void ocfs2_kill_sb(struct super_block *sb)
1248 {
1249         struct ocfs2_super *osb = OCFS2_SB(sb);
1250
1251         /* Failed mount? */
1252         if (!osb || atomic_read(&osb->vol_state) == VOLUME_DISABLED)
1253                 goto out;
1254
1255         /* Prevent further queueing of inode drop events */
1256         spin_lock(&dentry_list_lock);
1257         ocfs2_set_osb_flag(osb, OCFS2_OSB_DROP_DENTRY_LOCK_IMMED);
1258         spin_unlock(&dentry_list_lock);
1259         /* Wait for work to finish and/or remove it */
1260         cancel_work_sync(&osb->dentry_lock_work);
1261 out:
1262         kill_block_super(sb);
1263 }
1264
1265 static struct file_system_type ocfs2_fs_type = {
1266         .owner          = THIS_MODULE,
1267         .name           = "ocfs2",
1268         .mount          = ocfs2_mount,
1269         .kill_sb        = ocfs2_kill_sb,
1270
1271         .fs_flags       = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1272         .next           = NULL
1273 };
1274
1275 static int ocfs2_check_set_options(struct super_block *sb,
1276                                    struct mount_options *options)
1277 {
1278         if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1279             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1280                                          OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1281                 mlog(ML_ERROR, "User quotas were requested, but this "
1282                      "filesystem does not have the feature enabled.\n");
1283                 return 0;
1284         }
1285         if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1286             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1287                                          OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1288                 mlog(ML_ERROR, "Group quotas were requested, but this "
1289                      "filesystem does not have the feature enabled.\n");
1290                 return 0;
1291         }
1292         if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1293             !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1294                 mlog(ML_ERROR, "ACL support requested but extended attributes "
1295                      "feature is not enabled\n");
1296                 return 0;
1297         }
1298         /* No ACL setting specified? Use XATTR feature... */
1299         if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1300                                     OCFS2_MOUNT_NO_POSIX_ACL))) {
1301                 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1302                         options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1303                 else
1304                         options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1305         }
1306         return 1;
1307 }
1308
1309 static int ocfs2_parse_options(struct super_block *sb,
1310                                char *options,
1311                                struct mount_options *mopt,
1312                                int is_remount)
1313 {
1314         int status;
1315         char *p;
1316         u32 tmp;
1317
1318         mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
1319                    options ? options : "(none)");
1320
1321         mopt->commit_interval = 0;
1322         mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1323         mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1324         mopt->slot = OCFS2_INVALID_SLOT;
1325         mopt->localalloc_opt = -1;
1326         mopt->cluster_stack[0] = '\0';
1327         mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1328         mopt->dir_resv_level = -1;
1329
1330         if (!options) {
1331                 status = 1;
1332                 goto bail;
1333         }
1334
1335         while ((p = strsep(&options, ",")) != NULL) {
1336                 int token, option;
1337                 substring_t args[MAX_OPT_ARGS];
1338
1339                 if (!*p)
1340                         continue;
1341
1342                 token = match_token(p, tokens, args);
1343                 switch (token) {
1344                 case Opt_hb_local:
1345                         mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1346                         break;
1347                 case Opt_hb_none:
1348                         mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1349                         break;
1350                 case Opt_hb_global:
1351                         mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1352                         break;
1353                 case Opt_barrier:
1354                         if (match_int(&args[0], &option)) {
1355                                 status = 0;
1356                                 goto bail;
1357                         }
1358                         if (option)
1359                                 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1360                         else
1361                                 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1362                         break;
1363                 case Opt_intr:
1364                         mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1365                         break;
1366                 case Opt_nointr:
1367                         mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1368                         break;
1369                 case Opt_err_panic:
1370                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1371                         break;
1372                 case Opt_err_ro:
1373                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1374                         break;
1375                 case Opt_data_ordered:
1376                         mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1377                         break;
1378                 case Opt_data_writeback:
1379                         mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1380                         break;
1381                 case Opt_user_xattr:
1382                         mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1383                         break;
1384                 case Opt_nouser_xattr:
1385                         mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1386                         break;
1387                 case Opt_atime_quantum:
1388                         if (match_int(&args[0], &option)) {
1389                                 status = 0;
1390                                 goto bail;
1391                         }
1392                         if (option >= 0)
1393                                 mopt->atime_quantum = option;
1394                         break;
1395                 case Opt_slot:
1396                         option = 0;
1397                         if (match_int(&args[0], &option)) {
1398                                 status = 0;
1399                                 goto bail;
1400                         }
1401                         if (option)
1402                                 mopt->slot = (s16)option;
1403                         break;
1404                 case Opt_commit:
1405                         option = 0;
1406                         if (match_int(&args[0], &option)) {
1407                                 status = 0;
1408                                 goto bail;
1409                         }
1410                         if (option < 0)
1411                                 return 0;
1412                         if (option == 0)
1413                                 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1414                         mopt->commit_interval = HZ * option;
1415                         break;
1416                 case Opt_localalloc:
1417                         option = 0;
1418                         if (match_int(&args[0], &option)) {
1419                                 status = 0;
1420                                 goto bail;
1421                         }
1422                         if (option >= 0)
1423                                 mopt->localalloc_opt = option;
1424                         break;
1425                 case Opt_localflocks:
1426                         /*
1427                          * Changing this during remount could race
1428                          * flock() requests, or "unbalance" existing
1429                          * ones (e.g., a lock is taken in one mode but
1430                          * dropped in the other). If users care enough
1431                          * to flip locking modes during remount, we
1432                          * could add a "local" flag to individual
1433                          * flock structures for proper tracking of
1434                          * state.
1435                          */
1436                         if (!is_remount)
1437                                 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1438                         break;
1439                 case Opt_stack:
1440                         /* Check both that the option we were passed
1441                          * is of the right length and that it is a proper
1442                          * string of the right length.
1443                          */
1444                         if (((args[0].to - args[0].from) !=
1445                              OCFS2_STACK_LABEL_LEN) ||
1446                             (strnlen(args[0].from,
1447                                      OCFS2_STACK_LABEL_LEN) !=
1448                              OCFS2_STACK_LABEL_LEN)) {
1449                                 mlog(ML_ERROR,
1450                                      "Invalid cluster_stack option\n");
1451                                 status = 0;
1452                                 goto bail;
1453                         }
1454                         memcpy(mopt->cluster_stack, args[0].from,
1455                                OCFS2_STACK_LABEL_LEN);
1456                         mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1457                         break;
1458                 case Opt_inode64:
1459                         mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1460                         break;
1461                 case Opt_usrquota:
1462                         mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1463                         break;
1464                 case Opt_grpquota:
1465                         mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1466                         break;
1467                 case Opt_coherency_buffered:
1468                         mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1469                         break;
1470                 case Opt_coherency_full:
1471                         mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1472                         break;
1473                 case Opt_acl:
1474                         mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1475                         mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1476                         break;
1477                 case Opt_noacl:
1478                         mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1479                         mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1480                         break;
1481                 case Opt_resv_level:
1482                         if (is_remount)
1483                                 break;
1484                         if (match_int(&args[0], &option)) {
1485                                 status = 0;
1486                                 goto bail;
1487                         }
1488                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1489                             option < OCFS2_MAX_RESV_LEVEL)
1490                                 mopt->resv_level = option;
1491                         break;
1492                 case Opt_dir_resv_level:
1493                         if (is_remount)
1494                                 break;
1495                         if (match_int(&args[0], &option)) {
1496                                 status = 0;
1497                                 goto bail;
1498                         }
1499                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1500                             option < OCFS2_MAX_RESV_LEVEL)
1501                                 mopt->dir_resv_level = option;
1502                         break;
1503                 default:
1504                         mlog(ML_ERROR,
1505                              "Unrecognized mount option \"%s\" "
1506                              "or missing value\n", p);
1507                         status = 0;
1508                         goto bail;
1509                 }
1510         }
1511
1512         /* Ensure only one heartbeat mode */
1513         tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
1514                                  OCFS2_MOUNT_HB_NONE);
1515         if (hweight32(tmp) != 1) {
1516                 mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1517                 status = 0;
1518                 goto bail;
1519         }
1520
1521         status = 1;
1522
1523 bail:
1524         mlog_exit(status);
1525         return status;
1526 }
1527
1528 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1529 {
1530         struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb);
1531         unsigned long opts = osb->s_mount_opt;
1532         unsigned int local_alloc_megs;
1533
1534         if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1535                 seq_printf(s, ",_netdev");
1536                 if (opts & OCFS2_MOUNT_HB_LOCAL)
1537                         seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1538                 else
1539                         seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1540         } else
1541                 seq_printf(s, ",%s", OCFS2_HB_NONE);
1542
1543         if (opts & OCFS2_MOUNT_NOINTR)
1544                 seq_printf(s, ",nointr");
1545
1546         if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1547                 seq_printf(s, ",data=writeback");
1548         else
1549                 seq_printf(s, ",data=ordered");
1550
1551         if (opts & OCFS2_MOUNT_BARRIER)
1552                 seq_printf(s, ",barrier=1");
1553
1554         if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1555                 seq_printf(s, ",errors=panic");
1556         else
1557                 seq_printf(s, ",errors=remount-ro");
1558
1559         if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1560                 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1561
1562         if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM)
1563                 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1564
1565         if (osb->osb_commit_interval)
1566                 seq_printf(s, ",commit=%u",
1567                            (unsigned) (osb->osb_commit_interval / HZ));
1568
1569         local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1570         if (local_alloc_megs != ocfs2_la_default_mb(osb))
1571                 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1572
1573         if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1574                 seq_printf(s, ",localflocks,");
1575
1576         if (osb->osb_cluster_stack[0])
1577                 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1578                            osb->osb_cluster_stack);
1579         if (opts & OCFS2_MOUNT_USRQUOTA)
1580                 seq_printf(s, ",usrquota");
1581         if (opts & OCFS2_MOUNT_GRPQUOTA)
1582                 seq_printf(s, ",grpquota");
1583
1584         if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1585                 seq_printf(s, ",coherency=buffered");
1586         else
1587                 seq_printf(s, ",coherency=full");
1588
1589         if (opts & OCFS2_MOUNT_NOUSERXATTR)
1590                 seq_printf(s, ",nouser_xattr");
1591         else
1592                 seq_printf(s, ",user_xattr");
1593
1594         if (opts & OCFS2_MOUNT_INODE64)
1595                 seq_printf(s, ",inode64");
1596
1597         if (opts & OCFS2_MOUNT_POSIX_ACL)
1598                 seq_printf(s, ",acl");
1599         else
1600                 seq_printf(s, ",noacl");
1601
1602         if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1603                 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1604
1605         if (osb->osb_dir_resv_level != osb->osb_resv_level)
1606                 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1607
1608         return 0;
1609 }
1610
1611 static int __init ocfs2_init(void)
1612 {
1613         int status;
1614
1615         mlog_entry_void();
1616
1617         ocfs2_print_version();
1618
1619         status = init_ocfs2_uptodate_cache();
1620         if (status < 0) {
1621                 mlog_errno(status);
1622                 goto leave;
1623         }
1624
1625         status = ocfs2_initialize_mem_caches();
1626         if (status < 0) {
1627                 mlog_errno(status);
1628                 goto leave;
1629         }
1630
1631         ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1632         if (!ocfs2_wq) {
1633                 status = -ENOMEM;
1634                 goto leave;
1635         }
1636
1637         ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1638         if (!ocfs2_debugfs_root) {
1639                 status = -EFAULT;
1640                 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1641         }
1642
1643         status = ocfs2_quota_setup();
1644         if (status)
1645                 goto leave;
1646
1647         ocfs2_set_locking_protocol();
1648
1649         status = register_quota_format(&ocfs2_quota_format);
1650 leave:
1651         if (status < 0) {
1652                 ocfs2_quota_shutdown();
1653                 ocfs2_free_mem_caches();
1654                 exit_ocfs2_uptodate_cache();
1655         }
1656
1657         mlog_exit(status);
1658
1659         if (status >= 0) {
1660                 return register_filesystem(&ocfs2_fs_type);
1661         } else
1662                 return -1;
1663 }
1664
1665 static void __exit ocfs2_exit(void)
1666 {
1667         mlog_entry_void();
1668
1669         ocfs2_quota_shutdown();
1670
1671         if (ocfs2_wq) {
1672                 flush_workqueue(ocfs2_wq);
1673                 destroy_workqueue(ocfs2_wq);
1674         }
1675
1676         unregister_quota_format(&ocfs2_quota_format);
1677
1678         debugfs_remove(ocfs2_debugfs_root);
1679
1680         ocfs2_free_mem_caches();
1681
1682         unregister_filesystem(&ocfs2_fs_type);
1683
1684         exit_ocfs2_uptodate_cache();
1685
1686         mlog_exit_void();
1687 }
1688
1689 static void ocfs2_put_super(struct super_block *sb)
1690 {
1691         mlog_entry("(0x%p)\n", sb);
1692
1693         ocfs2_sync_blockdev(sb);
1694         ocfs2_dismount_volume(sb, 0);
1695
1696         mlog_exit_void();
1697 }
1698
1699 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1700 {
1701         struct ocfs2_super *osb;
1702         u32 numbits, freebits;
1703         int status;
1704         struct ocfs2_dinode *bm_lock;
1705         struct buffer_head *bh = NULL;
1706         struct inode *inode = NULL;
1707
1708         mlog_entry("(%p, %p)\n", dentry->d_sb, buf);
1709
1710         osb = OCFS2_SB(dentry->d_sb);
1711
1712         inode = ocfs2_get_system_file_inode(osb,
1713                                             GLOBAL_BITMAP_SYSTEM_INODE,
1714                                             OCFS2_INVALID_SLOT);
1715         if (!inode) {
1716                 mlog(ML_ERROR, "failed to get bitmap inode\n");
1717                 status = -EIO;
1718                 goto bail;
1719         }
1720
1721         status = ocfs2_inode_lock(inode, &bh, 0);
1722         if (status < 0) {
1723                 mlog_errno(status);
1724                 goto bail;
1725         }
1726
1727         bm_lock = (struct ocfs2_dinode *) bh->b_data;
1728
1729         numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1730         freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1731
1732         buf->f_type = OCFS2_SUPER_MAGIC;
1733         buf->f_bsize = dentry->d_sb->s_blocksize;
1734         buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1735         buf->f_blocks = ((sector_t) numbits) *
1736                         (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1737         buf->f_bfree = ((sector_t) freebits) *
1738                        (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1739         buf->f_bavail = buf->f_bfree;
1740         buf->f_files = numbits;
1741         buf->f_ffree = freebits;
1742         buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1743                                 & 0xFFFFFFFFUL;
1744         buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1745                                 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1746
1747         brelse(bh);
1748
1749         ocfs2_inode_unlock(inode, 0);
1750         status = 0;
1751 bail:
1752         if (inode)
1753                 iput(inode);
1754
1755         mlog_exit(status);
1756
1757         return status;
1758 }
1759
1760 static void ocfs2_inode_init_once(void *data)
1761 {
1762         struct ocfs2_inode_info *oi = data;
1763
1764         oi->ip_flags = 0;
1765         oi->ip_open_count = 0;
1766         spin_lock_init(&oi->ip_lock);
1767         ocfs2_extent_map_init(&oi->vfs_inode);
1768         INIT_LIST_HEAD(&oi->ip_io_markers);
1769         oi->ip_dir_start_lookup = 0;
1770
1771         init_rwsem(&oi->ip_alloc_sem);
1772         init_rwsem(&oi->ip_xattr_sem);
1773         mutex_init(&oi->ip_io_mutex);
1774
1775         oi->ip_blkno = 0ULL;
1776         oi->ip_clusters = 0;
1777
1778         ocfs2_resv_init_once(&oi->ip_la_data_resv);
1779
1780         ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1781         ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1782         ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1783
1784         ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1785                                   &ocfs2_inode_caching_ops);
1786
1787         inode_init_once(&oi->vfs_inode);
1788 }
1789
1790 static int ocfs2_initialize_mem_caches(void)
1791 {
1792         ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1793                                        sizeof(struct ocfs2_inode_info),
1794                                        0,
1795                                        (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1796                                                 SLAB_MEM_SPREAD),
1797                                        ocfs2_inode_init_once);
1798         ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1799                                         sizeof(struct ocfs2_dquot),
1800                                         0,
1801                                         (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1802                                                 SLAB_MEM_SPREAD),
1803                                         NULL);
1804         ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1805                                         sizeof(struct ocfs2_quota_chunk),
1806                                         0,
1807                                         (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1808                                         NULL);
1809         if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1810             !ocfs2_qf_chunk_cachep) {
1811                 if (ocfs2_inode_cachep)
1812                         kmem_cache_destroy(ocfs2_inode_cachep);
1813                 if (ocfs2_dquot_cachep)
1814                         kmem_cache_destroy(ocfs2_dquot_cachep);
1815                 if (ocfs2_qf_chunk_cachep)
1816                         kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1817                 return -ENOMEM;
1818         }
1819
1820         return 0;
1821 }
1822
1823 static void ocfs2_free_mem_caches(void)
1824 {
1825         if (ocfs2_inode_cachep)
1826                 kmem_cache_destroy(ocfs2_inode_cachep);
1827         ocfs2_inode_cachep = NULL;
1828
1829         if (ocfs2_dquot_cachep)
1830                 kmem_cache_destroy(ocfs2_dquot_cachep);
1831         ocfs2_dquot_cachep = NULL;
1832
1833         if (ocfs2_qf_chunk_cachep)
1834                 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1835         ocfs2_qf_chunk_cachep = NULL;
1836 }
1837
1838 static int ocfs2_get_sector(struct super_block *sb,
1839                             struct buffer_head **bh,
1840                             int block,
1841                             int sect_size)
1842 {
1843         if (!sb_set_blocksize(sb, sect_size)) {
1844                 mlog(ML_ERROR, "unable to set blocksize\n");
1845                 return -EIO;
1846         }
1847
1848         *bh = sb_getblk(sb, block);
1849         if (!*bh) {
1850                 mlog_errno(-EIO);
1851                 return -EIO;
1852         }
1853         lock_buffer(*bh);
1854         if (!buffer_dirty(*bh))
1855                 clear_buffer_uptodate(*bh);
1856         unlock_buffer(*bh);
1857         ll_rw_block(READ, 1, bh);
1858         wait_on_buffer(*bh);
1859         if (!buffer_uptodate(*bh)) {
1860                 mlog_errno(-EIO);
1861                 brelse(*bh);
1862                 *bh = NULL;
1863                 return -EIO;
1864         }
1865
1866         return 0;
1867 }
1868
1869 static int ocfs2_mount_volume(struct super_block *sb)
1870 {
1871         int status = 0;
1872         int unlock_super = 0;
1873         struct ocfs2_super *osb = OCFS2_SB(sb);
1874
1875         mlog_entry_void();
1876
1877         if (ocfs2_is_hard_readonly(osb))
1878                 goto leave;
1879
1880         status = ocfs2_dlm_init(osb);
1881         if (status < 0) {
1882                 mlog_errno(status);
1883                 goto leave;
1884         }
1885
1886         status = ocfs2_super_lock(osb, 1);
1887         if (status < 0) {
1888                 mlog_errno(status);
1889                 goto leave;
1890         }
1891         unlock_super = 1;
1892
1893         /* This will load up the node map and add ourselves to it. */
1894         status = ocfs2_find_slot(osb);
1895         if (status < 0) {
1896                 mlog_errno(status);
1897                 goto leave;
1898         }
1899
1900         /* load all node-local system inodes */
1901         status = ocfs2_init_local_system_inodes(osb);
1902         if (status < 0) {
1903                 mlog_errno(status);
1904                 goto leave;
1905         }
1906
1907         status = ocfs2_check_volume(osb);
1908         if (status < 0) {
1909                 mlog_errno(status);
1910                 goto leave;
1911         }
1912
1913         status = ocfs2_truncate_log_init(osb);
1914         if (status < 0)
1915                 mlog_errno(status);
1916
1917 leave:
1918         if (unlock_super)
1919                 ocfs2_super_unlock(osb, 1);
1920
1921         mlog_exit(status);
1922         return status;
1923 }
1924
1925 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1926 {
1927         int tmp, hangup_needed = 0;
1928         struct ocfs2_super *osb = NULL;
1929         char nodestr[8];
1930
1931         mlog_entry("(0x%p)\n", sb);
1932
1933         BUG_ON(!sb);
1934         osb = OCFS2_SB(sb);
1935         BUG_ON(!osb);
1936
1937         debugfs_remove(osb->osb_ctxt);
1938
1939         /*
1940          * Flush inode dropping work queue so that deletes are
1941          * performed while the filesystem is still working
1942          */
1943         ocfs2_drop_all_dl_inodes(osb);
1944
1945         /* Orphan scan should be stopped as early as possible */
1946         ocfs2_orphan_scan_stop(osb);
1947
1948         ocfs2_disable_quotas(osb);
1949
1950         ocfs2_shutdown_local_alloc(osb);
1951
1952         ocfs2_truncate_log_shutdown(osb);
1953
1954         /* This will disable recovery and flush any recovery work. */
1955         ocfs2_recovery_exit(osb);
1956
1957         ocfs2_journal_shutdown(osb);
1958
1959         ocfs2_sync_blockdev(sb);
1960
1961         ocfs2_purge_refcount_trees(osb);
1962
1963         /* No cluster connection means we've failed during mount, so skip
1964          * all the steps which depended on that to complete. */
1965         if (osb->cconn) {
1966                 tmp = ocfs2_super_lock(osb, 1);
1967                 if (tmp < 0) {
1968                         mlog_errno(tmp);
1969                         return;
1970                 }
1971         }
1972
1973         if (osb->slot_num != OCFS2_INVALID_SLOT)
1974                 ocfs2_put_slot(osb);
1975
1976         if (osb->cconn)
1977                 ocfs2_super_unlock(osb, 1);
1978
1979         ocfs2_release_system_inodes(osb);
1980
1981         /*
1982          * If we're dismounting due to mount error, mount.ocfs2 will clean
1983          * up heartbeat.  If we're a local mount, there is no heartbeat.
1984          * If we failed before we got a uuid_str yet, we can't stop
1985          * heartbeat.  Otherwise, do it.
1986          */
1987         if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str)
1988                 hangup_needed = 1;
1989
1990         if (osb->cconn)
1991                 ocfs2_dlm_shutdown(osb, hangup_needed);
1992
1993         ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1994         debugfs_remove(osb->osb_debug_root);
1995
1996         if (hangup_needed)
1997                 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1998
1999         atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
2000
2001         if (ocfs2_mount_local(osb))
2002                 snprintf(nodestr, sizeof(nodestr), "local");
2003         else
2004                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
2005
2006         printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
2007                osb->dev_str, nodestr);
2008
2009         ocfs2_delete_osb(osb);
2010         kfree(osb);
2011         sb->s_dev = 0;
2012         sb->s_fs_info = NULL;
2013 }
2014
2015 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
2016                                 unsigned uuid_bytes)
2017 {
2018         int i, ret;
2019         char *ptr;
2020
2021         BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
2022
2023         osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
2024         if (osb->uuid_str == NULL)
2025                 return -ENOMEM;
2026
2027         for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
2028                 /* print with null */
2029                 ret = snprintf(ptr, 3, "%02X", uuid[i]);
2030                 if (ret != 2) /* drop super cleans up */
2031                         return -EINVAL;
2032                 /* then only advance past the last char */
2033                 ptr += 2;
2034         }
2035
2036         return 0;
2037 }
2038
2039 /* Make sure entire volume is addressable by our journal.  Requires
2040    osb_clusters_at_boot to be valid and for the journal to have been
2041    initialized by ocfs2_journal_init(). */
2042 static int ocfs2_journal_addressable(struct ocfs2_super *osb)
2043 {
2044         int status = 0;
2045         u64 max_block =
2046                 ocfs2_clusters_to_blocks(osb->sb,
2047                                          osb->osb_clusters_at_boot) - 1;
2048
2049         /* 32-bit block number is always OK. */
2050         if (max_block <= (u32)~0ULL)
2051                 goto out;
2052
2053         /* Volume is "huge", so see if our journal is new enough to
2054            support it. */
2055         if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2056                                        OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2057               jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2058                                                JBD2_FEATURE_INCOMPAT_64BIT))) {
2059                 mlog(ML_ERROR, "The journal cannot address the entire volume. "
2060                      "Enable the 'block64' journal option with tunefs.ocfs2");
2061                 status = -EFBIG;
2062                 goto out;
2063         }
2064
2065  out:
2066         return status;
2067 }
2068
2069 static int ocfs2_initialize_super(struct super_block *sb,
2070                                   struct buffer_head *bh,
2071                                   int sector_size,
2072                                   struct ocfs2_blockcheck_stats *stats)
2073 {
2074         int status;
2075         int i, cbits, bbits;
2076         struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2077         struct inode *inode = NULL;
2078         struct ocfs2_journal *journal;
2079         __le32 uuid_net_key;
2080         struct ocfs2_super *osb;
2081         u64 total_blocks;
2082
2083         mlog_entry_void();
2084
2085         osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2086         if (!osb) {
2087                 status = -ENOMEM;
2088                 mlog_errno(status);
2089                 goto bail;
2090         }
2091
2092         sb->s_fs_info = osb;
2093         sb->s_op = &ocfs2_sops;
2094         sb->s_export_op = &ocfs2_export_ops;
2095         sb->s_qcop = &ocfs2_quotactl_ops;
2096         sb->dq_op = &ocfs2_quota_operations;
2097         sb->s_xattr = ocfs2_xattr_handlers;
2098         sb->s_time_gran = 1;
2099         sb->s_flags |= MS_NOATIME;
2100         /* this is needed to support O_LARGEFILE */
2101         cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2102         bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2103         sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2104
2105         osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2106
2107         for (i = 0; i < 3; i++)
2108                 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2109         osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2110
2111         osb->sb = sb;
2112         /* Save off for ocfs2_rw_direct */
2113         osb->s_sectsize_bits = blksize_bits(sector_size);
2114         BUG_ON(!osb->s_sectsize_bits);
2115
2116         spin_lock_init(&osb->dc_task_lock);
2117         init_waitqueue_head(&osb->dc_event);
2118         osb->dc_work_sequence = 0;
2119         osb->dc_wake_sequence = 0;
2120         INIT_LIST_HEAD(&osb->blocked_lock_list);
2121         osb->blocked_lock_count = 0;
2122         spin_lock_init(&osb->osb_lock);
2123         spin_lock_init(&osb->osb_xattr_lock);
2124         ocfs2_init_steal_slots(osb);
2125
2126         atomic_set(&osb->alloc_stats.moves, 0);
2127         atomic_set(&osb->alloc_stats.local_data, 0);
2128         atomic_set(&osb->alloc_stats.bitmap_data, 0);
2129         atomic_set(&osb->alloc_stats.bg_allocs, 0);
2130         atomic_set(&osb->alloc_stats.bg_extends, 0);
2131
2132         /* Copy the blockcheck stats from the superblock probe */
2133         osb->osb_ecc_stats = *stats;
2134
2135         ocfs2_init_node_maps(osb);
2136
2137         snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2138                  MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2139
2140         osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2141         if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2142                 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2143                      osb->max_slots);
2144                 status = -EINVAL;
2145                 goto bail;
2146         }
2147         mlog(0, "max_slots for this device: %u\n", osb->max_slots);
2148
2149         ocfs2_orphan_scan_init(osb);
2150
2151         status = ocfs2_recovery_init(osb);
2152         if (status) {
2153                 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2154                 mlog_errno(status);
2155                 goto bail;
2156         }
2157
2158         init_waitqueue_head(&osb->checkpoint_event);
2159         atomic_set(&osb->needs_checkpoint, 0);
2160
2161         osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2162
2163         osb->slot_num = OCFS2_INVALID_SLOT;
2164
2165         osb->s_xattr_inline_size = le16_to_cpu(
2166                                         di->id2.i_super.s_xattr_inline_size);
2167
2168         osb->local_alloc_state = OCFS2_LA_UNUSED;
2169         osb->local_alloc_bh = NULL;
2170         INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2171
2172         init_waitqueue_head(&osb->osb_mount_event);
2173
2174         status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2175         if (status) {
2176                 mlog_errno(status);
2177                 goto bail;
2178         }
2179
2180         osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2181         if (!osb->vol_label) {
2182                 mlog(ML_ERROR, "unable to alloc vol label\n");
2183                 status = -ENOMEM;
2184                 goto bail;
2185         }
2186
2187         osb->slot_recovery_generations =
2188                 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2189                         GFP_KERNEL);
2190         if (!osb->slot_recovery_generations) {
2191                 status = -ENOMEM;
2192                 mlog_errno(status);
2193                 goto bail;
2194         }
2195
2196         init_waitqueue_head(&osb->osb_wipe_event);
2197         osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2198                                         sizeof(*osb->osb_orphan_wipes),
2199                                         GFP_KERNEL);
2200         if (!osb->osb_orphan_wipes) {
2201                 status = -ENOMEM;
2202                 mlog_errno(status);
2203                 goto bail;
2204         }
2205
2206         osb->osb_rf_lock_tree = RB_ROOT;
2207
2208         osb->s_feature_compat =
2209                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2210         osb->s_feature_ro_compat =
2211                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2212         osb->s_feature_incompat =
2213                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2214
2215         if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2216                 mlog(ML_ERROR, "couldn't mount because of unsupported "
2217                      "optional features (%x).\n", i);
2218                 status = -EINVAL;
2219                 goto bail;
2220         }
2221         if (!(osb->sb->s_flags & MS_RDONLY) &&
2222             (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2223                 mlog(ML_ERROR, "couldn't mount RDWR because of "
2224                      "unsupported optional features (%x).\n", i);
2225                 status = -EINVAL;
2226                 goto bail;
2227         }
2228
2229         if (ocfs2_clusterinfo_valid(osb)) {
2230                 osb->osb_stackflags =
2231                         OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2232                 memcpy(osb->osb_cluster_stack,
2233                        OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2234                        OCFS2_STACK_LABEL_LEN);
2235                 osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
2236                 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2237                         mlog(ML_ERROR,
2238                              "couldn't mount because of an invalid "
2239                              "cluster stack label (%s) \n",
2240                              osb->osb_cluster_stack);
2241                         status = -EINVAL;
2242                         goto bail;
2243                 }
2244         } else {
2245                 /* The empty string is identical with classic tools that
2246                  * don't know about s_cluster_info. */
2247                 osb->osb_cluster_stack[0] = '\0';
2248         }
2249
2250         get_random_bytes(&osb->s_next_generation, sizeof(u32));
2251
2252         /* FIXME
2253          * This should be done in ocfs2_journal_init(), but unknown
2254          * ordering issues will cause the filesystem to crash.
2255          * If anyone wants to figure out what part of the code
2256          * refers to osb->journal before ocfs2_journal_init() is run,
2257          * be my guest.
2258          */
2259         /* initialize our journal structure */
2260
2261         journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2262         if (!journal) {
2263                 mlog(ML_ERROR, "unable to alloc journal\n");
2264                 status = -ENOMEM;
2265                 goto bail;
2266         }
2267         osb->journal = journal;
2268         journal->j_osb = osb;
2269
2270         atomic_set(&journal->j_num_trans, 0);
2271         init_rwsem(&journal->j_trans_barrier);
2272         init_waitqueue_head(&journal->j_checkpointed);
2273         spin_lock_init(&journal->j_lock);
2274         journal->j_trans_id = (unsigned long) 1;
2275         INIT_LIST_HEAD(&journal->j_la_cleanups);
2276         INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2277         journal->j_state = OCFS2_JOURNAL_FREE;
2278
2279         INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes);
2280         osb->dentry_lock_list = NULL;
2281
2282         /* get some pseudo constants for clustersize bits */
2283         osb->s_clustersize_bits =
2284                 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2285         osb->s_clustersize = 1 << osb->s_clustersize_bits;
2286         mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
2287
2288         if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2289             osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2290                 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2291                      osb->s_clustersize);
2292                 status = -EINVAL;
2293                 goto bail;
2294         }
2295
2296         total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2297                                                 le32_to_cpu(di->i_clusters));
2298
2299         status = generic_check_addressable(osb->sb->s_blocksize_bits,
2300                                            total_blocks);
2301         if (status) {
2302                 mlog(ML_ERROR, "Volume too large "
2303                      "to mount safely on this system");
2304                 status = -EFBIG;
2305                 goto bail;
2306         }
2307
2308         if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2309                                  sizeof(di->id2.i_super.s_uuid))) {
2310                 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2311                 status = -ENOMEM;
2312                 goto bail;
2313         }
2314
2315         memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
2316
2317         strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2318         osb->vol_label[63] = '\0';
2319         osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2320         osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2321         osb->first_cluster_group_blkno =
2322                 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2323         osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2324         osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2325         mlog(0, "vol_label: %s\n", osb->vol_label);
2326         mlog(0, "uuid: %s\n", osb->uuid_str);
2327         mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
2328              (unsigned long long)osb->root_blkno,
2329              (unsigned long long)osb->system_dir_blkno);
2330
2331         osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2332         if (!osb->osb_dlm_debug) {
2333                 status = -ENOMEM;
2334                 mlog_errno(status);
2335                 goto bail;
2336         }
2337
2338         atomic_set(&osb->vol_state, VOLUME_INIT);
2339
2340         /* load root, system_dir, and all global system inodes */
2341         status = ocfs2_init_global_system_inodes(osb);
2342         if (status < 0) {
2343                 mlog_errno(status);
2344                 goto bail;
2345         }
2346
2347         /*
2348          * global bitmap
2349          */
2350         inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2351                                             OCFS2_INVALID_SLOT);
2352         if (!inode) {
2353                 status = -EINVAL;
2354                 mlog_errno(status);
2355                 goto bail;
2356         }
2357
2358         osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2359         osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2360         iput(inode);
2361
2362         osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2363                                  osb->s_feature_incompat) * 8;
2364
2365         status = ocfs2_init_slot_info(osb);
2366         if (status < 0) {
2367                 mlog_errno(status);
2368                 goto bail;
2369         }
2370
2371 bail:
2372         mlog_exit(status);
2373         return status;
2374 }
2375
2376 /*
2377  * will return: -EAGAIN if it is ok to keep searching for superblocks
2378  *              -EINVAL if there is a bad superblock
2379  *              0 on success
2380  */
2381 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2382                                struct buffer_head *bh,
2383                                u32 blksz,
2384                                struct ocfs2_blockcheck_stats *stats)
2385 {
2386         int status = -EAGAIN;
2387
2388         mlog_entry_void();
2389
2390         if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2391                    strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2392                 /* We have to do a raw check of the feature here */
2393                 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2394                     OCFS2_FEATURE_INCOMPAT_META_ECC) {
2395                         status = ocfs2_block_check_validate(bh->b_data,
2396                                                             bh->b_size,
2397                                                             &di->i_check,
2398                                                             stats);
2399                         if (status)
2400                                 goto out;
2401                 }
2402                 status = -EINVAL;
2403                 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2404                         mlog(ML_ERROR, "found superblock with incorrect block "
2405                              "size: found %u, should be %u\n",
2406                              1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2407                                blksz);
2408                 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2409                            OCFS2_MAJOR_REV_LEVEL ||
2410                            le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2411                            OCFS2_MINOR_REV_LEVEL) {
2412                         mlog(ML_ERROR, "found superblock with bad version: "
2413                              "found %u.%u, should be %u.%u\n",
2414                              le16_to_cpu(di->id2.i_super.s_major_rev_level),
2415                              le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2416                              OCFS2_MAJOR_REV_LEVEL,
2417                              OCFS2_MINOR_REV_LEVEL);
2418                 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2419                         mlog(ML_ERROR, "bad block number on superblock: "
2420                              "found %llu, should be %llu\n",
2421                              (unsigned long long)le64_to_cpu(di->i_blkno),
2422                              (unsigned long long)bh->b_blocknr);
2423                 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2424                             le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2425                         mlog(ML_ERROR, "bad cluster size found: %u\n",
2426                              1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2427                 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2428                         mlog(ML_ERROR, "bad root_blkno: 0\n");
2429                 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2430                         mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2431                 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2432                         mlog(ML_ERROR,
2433                              "Superblock slots found greater than file system "
2434                              "maximum: found %u, max %u\n",
2435                              le16_to_cpu(di->id2.i_super.s_max_slots),
2436                              OCFS2_MAX_SLOTS);
2437                 } else {
2438                         /* found it! */
2439                         status = 0;
2440                 }
2441         }
2442
2443 out:
2444         mlog_exit(status);
2445         return status;
2446 }
2447
2448 static int ocfs2_check_volume(struct ocfs2_super *osb)
2449 {
2450         int status;
2451         int dirty;
2452         int local;
2453         struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2454                                                   * recover
2455                                                   * ourselves. */
2456
2457         mlog_entry_void();
2458
2459         /* Init our journal object. */
2460         status = ocfs2_journal_init(osb->journal, &dirty);
2461         if (status < 0) {
2462                 mlog(ML_ERROR, "Could not initialize journal!\n");
2463                 goto finally;
2464         }
2465
2466         /* Now that journal has been initialized, check to make sure
2467            entire volume is addressable. */
2468         status = ocfs2_journal_addressable(osb);
2469         if (status)
2470                 goto finally;
2471
2472         /* If the journal was unmounted cleanly then we don't want to
2473          * recover anything. Otherwise, journal_load will do that
2474          * dirty work for us :) */
2475         if (!dirty) {
2476                 status = ocfs2_journal_wipe(osb->journal, 0);
2477                 if (status < 0) {
2478                         mlog_errno(status);
2479                         goto finally;
2480                 }
2481         } else {
2482                 mlog(ML_NOTICE, "File system was not unmounted cleanly, "
2483                      "recovering volume.\n");
2484         }
2485
2486         local = ocfs2_mount_local(osb);
2487
2488         /* will play back anything left in the journal. */
2489         status = ocfs2_journal_load(osb->journal, local, dirty);
2490         if (status < 0) {
2491                 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2492                 goto finally;
2493         }
2494
2495         if (dirty) {
2496                 /* recover my local alloc if we didn't unmount cleanly. */
2497                 status = ocfs2_begin_local_alloc_recovery(osb,
2498                                                           osb->slot_num,
2499                                                           &local_alloc);
2500                 if (status < 0) {
2501                         mlog_errno(status);
2502                         goto finally;
2503                 }
2504                 /* we complete the recovery process after we've marked
2505                  * ourselves as mounted. */
2506         }
2507
2508         mlog(0, "Journal loaded.\n");
2509
2510         status = ocfs2_load_local_alloc(osb);
2511         if (status < 0) {
2512                 mlog_errno(status);
2513                 goto finally;
2514         }
2515
2516         if (dirty) {
2517                 /* Recovery will be completed after we've mounted the
2518                  * rest of the volume. */
2519                 osb->dirty = 1;
2520                 osb->local_alloc_copy = local_alloc;
2521                 local_alloc = NULL;
2522         }
2523
2524         /* go through each journal, trylock it and if you get the
2525          * lock, and it's marked as dirty, set the bit in the recover
2526          * map and launch a recovery thread for it. */
2527         status = ocfs2_mark_dead_nodes(osb);
2528         if (status < 0) {
2529                 mlog_errno(status);
2530                 goto finally;
2531         }
2532
2533         status = ocfs2_compute_replay_slots(osb);
2534         if (status < 0)
2535                 mlog_errno(status);
2536
2537 finally:
2538         if (local_alloc)
2539                 kfree(local_alloc);
2540
2541         mlog_exit(status);
2542         return status;
2543 }
2544
2545 /*
2546  * The routine gets called from dismount or close whenever a dismount on
2547  * volume is requested and the osb open count becomes 1.
2548  * It will remove the osb from the global list and also free up all the
2549  * initialized resources and fileobject.
2550  */
2551 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2552 {
2553         mlog_entry_void();
2554
2555         /* This function assumes that the caller has the main osb resource */
2556
2557         ocfs2_free_slot_info(osb);
2558
2559         kfree(osb->osb_orphan_wipes);
2560         kfree(osb->slot_recovery_generations);
2561         /* FIXME
2562          * This belongs in journal shutdown, but because we have to
2563          * allocate osb->journal at the start of ocfs2_initialize_osb(),
2564          * we free it here.
2565          */
2566         kfree(osb->journal);
2567         if (osb->local_alloc_copy)
2568                 kfree(osb->local_alloc_copy);
2569         kfree(osb->uuid_str);
2570         ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2571         memset(osb, 0, sizeof(struct ocfs2_super));
2572
2573         mlog_exit_void();
2574 }
2575
2576 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2577  * panic(). We do not support continue-on-error operation. */
2578 static void ocfs2_handle_error(struct super_block *sb)
2579 {
2580         struct ocfs2_super *osb = OCFS2_SB(sb);
2581
2582         if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2583                 panic("OCFS2: (device %s): panic forced after error\n",
2584                       sb->s_id);
2585
2586         ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2587
2588         if (sb->s_flags & MS_RDONLY &&
2589             (ocfs2_is_soft_readonly(osb) ||
2590              ocfs2_is_hard_readonly(osb)))
2591                 return;
2592
2593         printk(KERN_CRIT "File system is now read-only due to the potential "
2594                "of on-disk corruption. Please run fsck.ocfs2 once the file "
2595                "system is unmounted.\n");
2596         sb->s_flags |= MS_RDONLY;
2597         ocfs2_set_ro_flag(osb, 0);
2598 }
2599
2600 static char error_buf[1024];
2601
2602 void __ocfs2_error(struct super_block *sb,
2603                    const char *function,
2604                    const char *fmt, ...)
2605 {
2606         va_list args;
2607
2608         va_start(args, fmt);
2609         vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2610         va_end(args);
2611
2612         /* Not using mlog here because we want to show the actual
2613          * function the error came from. */
2614         printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2615                sb->s_id, function, error_buf);
2616
2617         ocfs2_handle_error(sb);
2618 }
2619
2620 /* Handle critical errors. This is intentionally more drastic than
2621  * ocfs2_handle_error, so we only use for things like journal errors,
2622  * etc. */
2623 void __ocfs2_abort(struct super_block* sb,
2624                    const char *function,
2625                    const char *fmt, ...)
2626 {
2627         va_list args;
2628
2629         va_start(args, fmt);
2630         vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2631         va_end(args);
2632
2633         printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2634                sb->s_id, function, error_buf);
2635
2636         /* We don't have the cluster support yet to go straight to
2637          * hard readonly in here. Until then, we want to keep
2638          * ocfs2_abort() so that we can at least mark critical
2639          * errors.
2640          *
2641          * TODO: This should abort the journal and alert other nodes
2642          * that our slot needs recovery. */
2643
2644         /* Force a panic(). This stinks, but it's better than letting
2645          * things continue without having a proper hard readonly
2646          * here. */
2647         if (!ocfs2_mount_local(OCFS2_SB(sb)))
2648                 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2649         ocfs2_handle_error(sb);
2650 }
2651
2652 /*
2653  * Void signal blockers, because in-kernel sigprocmask() only fails
2654  * when SIG_* is wrong.
2655  */
2656 void ocfs2_block_signals(sigset_t *oldset)
2657 {
2658         int rc;
2659         sigset_t blocked;
2660
2661         sigfillset(&blocked);
2662         rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2663         BUG_ON(rc);
2664 }
2665
2666 void ocfs2_unblock_signals(sigset_t *oldset)
2667 {
2668         int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2669         BUG_ON(rc);
2670 }
2671
2672 module_init(ocfs2_init);
2673 module_exit(ocfs2_exit);