fs/ocfs2/super.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  * super.c
   4  *
   5  * load/unload driver, mount/dismount volumes
   6  *
   7  * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
   8  */
   9
  10 #include <linux/module.h>
  11 #include <linux/fs.h>
  12 #include <linux/types.h>
  13 #include <linux/slab.h>
  14 #include <linux/highmem.h>
  15 #include <linux/init.h>
  16 #include <linux/random.h>
  17 #include <linux/statfs.h>
  18 #include <linux/moduleparam.h>
  19 #include <linux/blkdev.h>
  20 #include <linux/socket.h>
  21 #include <linux/inet.h>
  22 #include <linux/parser.h>
  23 #include <linux/crc32.h>
  24 #include <linux/debugfs.h>
  25 #include <linux/mount.h>
  26 #include <linux/seq_file.h>
  27 #include <linux/quotaops.h>
  28 #include <linux/signal.h>
  29
  30 #define CREATE_TRACE_POINTS
  31 #include "ocfs2_trace.h"
  32
  33 #include <cluster/masklog.h>
  34
  35 #include "ocfs2.h"
  36
  37 /* this should be the only file to include a version 1 header */
  38 #include "ocfs1_fs_compat.h"
  39
  40 #include "alloc.h"
  41 #include "aops.h"
  42 #include "blockcheck.h"
  43 #include "dlmglue.h"
  44 #include "export.h"
  45 #include "extent_map.h"
  46 #include "heartbeat.h"
  47 #include "inode.h"
  48 #include "journal.h"
  49 #include "localalloc.h"
  50 #include "namei.h"
  51 #include "slot_map.h"
  52 #include "super.h"
  53 #include "sysfile.h"
  54 #include "uptodate.h"
  55 #include "xattr.h"
  56 #include "quota.h"
  57 #include "refcounttree.h"
  58 #include "suballoc.h"
  59
  60 #include "buffer_head_io.h"
  61 #include "filecheck.h"
  62
  63 static struct kmem_cache *ocfs2_inode_cachep;
  64 struct kmem_cache *ocfs2_dquot_cachep;
  65 struct kmem_cache *ocfs2_qf_chunk_cachep;
  66
  67 static struct dentry *ocfs2_debugfs_root;
  68
  69 MODULE_AUTHOR("Oracle");
  70 MODULE_LICENSE("GPL");
  71 MODULE_DESCRIPTION("OCFS2 cluster file system");
  72
  73 struct mount_options
  74 {
  75         unsigned long   commit_interval;
  76         unsigned long   mount_opt;
  77         unsigned int    atime_quantum;
  78         unsigned short  slot;
  79         int             localalloc_opt;
  80         unsigned int    resv_level;
  81         int             dir_resv_level;
  82         char            cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
  83 };
  84
  85 static int ocfs2_parse_options(struct super_block *sb, char *options,
  86                                struct mount_options *mopt,
  87                                int is_remount);
  88 static int ocfs2_check_set_options(struct super_block *sb,
  89                                    struct mount_options *options);
  90 static int ocfs2_show_options(struct seq_file *s, struct dentry *root);
  91 static void ocfs2_put_super(struct super_block *sb);
  92 static int ocfs2_mount_volume(struct super_block *sb);
  93 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
  94 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
  95 static int ocfs2_initialize_mem_caches(void);
  96 static void ocfs2_free_mem_caches(void);
  97 static void ocfs2_delete_osb(struct ocfs2_super *osb);
  98
  99 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
 100
 101 static int ocfs2_sync_fs(struct super_block *sb, int wait);
 102
 103 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
 104 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
 105 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
 106 static int ocfs2_check_volume(struct ocfs2_super *osb);
 107 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
 108                                struct buffer_head *bh,
 109                                u32 sectsize,
 110                                struct ocfs2_blockcheck_stats *stats);
 111 static int ocfs2_initialize_super(struct super_block *sb,
 112                                   struct buffer_head *bh,
 113                                   int sector_size,
 114                                   struct ocfs2_blockcheck_stats *stats);
 115 static int ocfs2_get_sector(struct super_block *sb,
 116                             struct buffer_head **bh,
 117                             int block,
 118                             int sect_size);
 119 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
 120 static void ocfs2_free_inode(struct inode *inode);
 121 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
 122 static int ocfs2_enable_quotas(struct ocfs2_super *osb);
 123 static void ocfs2_disable_quotas(struct ocfs2_super *osb);
 124
 125 static struct dquot **ocfs2_get_dquots(struct inode *inode)
 126 {
 127         return OCFS2_I(inode)->i_dquot;
 128 }
 129
 130 static const struct super_operations ocfs2_sops = {
 131         .statfs         = ocfs2_statfs,
 132         .alloc_inode    = ocfs2_alloc_inode,
 133         .free_inode     = ocfs2_free_inode,
 134         .drop_inode     = ocfs2_drop_inode,
 135         .evict_inode    = ocfs2_evict_inode,
 136         .sync_fs        = ocfs2_sync_fs,
 137         .put_super      = ocfs2_put_super,
 138         .remount_fs     = ocfs2_remount,
 139         .show_options   = ocfs2_show_options,
 140         .quota_read     = ocfs2_quota_read,
 141         .quota_write    = ocfs2_quota_write,
 142         .get_dquots     = ocfs2_get_dquots,
 143 };
 144
 145 enum {
 146         Opt_barrier,
 147         Opt_err_panic,
 148         Opt_err_ro,
 149         Opt_intr,
 150         Opt_nointr,
 151         Opt_hb_none,
 152         Opt_hb_local,
 153         Opt_hb_global,
 154         Opt_data_ordered,
 155         Opt_data_writeback,
 156         Opt_atime_quantum,
 157         Opt_slot,
 158         Opt_commit,
 159         Opt_localalloc,
 160         Opt_localflocks,
 161         Opt_stack,
 162         Opt_user_xattr,
 163         Opt_nouser_xattr,
 164         Opt_inode64,
 165         Opt_acl,
 166         Opt_noacl,
 167         Opt_usrquota,
 168         Opt_grpquota,
 169         Opt_coherency_buffered,
 170         Opt_coherency_full,
 171         Opt_resv_level,
 172         Opt_dir_resv_level,
 173         Opt_journal_async_commit,
 174         Opt_err_cont,
 175         Opt_nocluster,
 176         Opt_err,
 177 };
 178
 179 static const match_table_t tokens = {
 180         {Opt_barrier, "barrier=%u"},
 181         {Opt_err_panic, "errors=panic"},
 182         {Opt_err_ro, "errors=remount-ro"},
 183         {Opt_intr, "intr"},
 184         {Opt_nointr, "nointr"},
 185         {Opt_hb_none, OCFS2_HB_NONE},
 186         {Opt_hb_local, OCFS2_HB_LOCAL},
 187         {Opt_hb_global, OCFS2_HB_GLOBAL},
 188         {Opt_data_ordered, "data=ordered"},
 189         {Opt_data_writeback, "data=writeback"},
 190         {Opt_atime_quantum, "atime_quantum=%u"},
 191         {Opt_slot, "preferred_slot=%u"},
 192         {Opt_commit, "commit=%u"},
 193         {Opt_localalloc, "localalloc=%d"},
 194         {Opt_localflocks, "localflocks"},
 195         {Opt_stack, "cluster_stack=%s"},
 196         {Opt_user_xattr, "user_xattr"},
 197         {Opt_nouser_xattr, "nouser_xattr"},
 198         {Opt_inode64, "inode64"},
 199         {Opt_acl, "acl"},
 200         {Opt_noacl, "noacl"},
 201         {Opt_usrquota, "usrquota"},
 202         {Opt_grpquota, "grpquota"},
 203         {Opt_coherency_buffered, "coherency=buffered"},
 204         {Opt_coherency_full, "coherency=full"},
 205         {Opt_resv_level, "resv_level=%u"},
 206         {Opt_dir_resv_level, "dir_resv_level=%u"},
 207         {Opt_journal_async_commit, "journal_async_commit"},
 208         {Opt_err_cont, "errors=continue"},
 209         {Opt_nocluster, "nocluster"},
 210         {Opt_err, NULL}
 211 };
 212
 213 #ifdef CONFIG_DEBUG_FS
 214 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
 215 {
 216         struct ocfs2_cluster_connection *cconn = osb->cconn;
 217         struct ocfs2_recovery_map *rm = osb->recovery_map;
 218         struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
 219         int i, out = 0;
 220         unsigned long flags;
 221
 222         out += scnprintf(buf + out, len - out,
 223                         "%10s => Id: %-s  Uuid: %-s  Gen: 0x%X  Label: %-s\n",
 224                         "Device", osb->dev_str, osb->uuid_str,
 225                         osb->fs_generation, osb->vol_label);
 226
 227         out += scnprintf(buf + out, len - out,
 228                         "%10s => State: %d  Flags: 0x%lX\n", "Volume",
 229                         atomic_read(&osb->vol_state), osb->osb_flags);
 230
 231         out += scnprintf(buf + out, len - out,
 232                         "%10s => Block: %lu  Cluster: %d\n", "Sizes",
 233                         osb->sb->s_blocksize, osb->s_clustersize);
 234
 235         out += scnprintf(buf + out, len - out,
 236                         "%10s => Compat: 0x%X  Incompat: 0x%X  "
 237                         "ROcompat: 0x%X\n",
 238                         "Features", osb->s_feature_compat,
 239                         osb->s_feature_incompat, osb->s_feature_ro_compat);
 240
 241         out += scnprintf(buf + out, len - out,
 242                         "%10s => Opts: 0x%lX  AtimeQuanta: %u\n", "Mount",
 243                         osb->s_mount_opt, osb->s_atime_quantum);
 244
 245         if (cconn) {
 246                 out += scnprintf(buf + out, len - out,
 247                                 "%10s => Stack: %s  Name: %*s  "
 248                                 "Version: %d.%d\n", "Cluster",
 249                                 (*osb->osb_cluster_stack == '\0' ?
 250                                  "o2cb" : osb->osb_cluster_stack),
 251                                 cconn->cc_namelen, cconn->cc_name,
 252                                 cconn->cc_version.pv_major,
 253                                 cconn->cc_version.pv_minor);
 254         }
 255
 256         spin_lock_irqsave(&osb->dc_task_lock, flags);
 257         out += scnprintf(buf + out, len - out,
 258                         "%10s => Pid: %d  Count: %lu  WakeSeq: %lu  "
 259                         "WorkSeq: %lu\n", "DownCnvt",
 260                         (osb->dc_task ?  task_pid_nr(osb->dc_task) : -1),
 261                         osb->blocked_lock_count, osb->dc_wake_sequence,
 262                         osb->dc_work_sequence);
 263         spin_unlock_irqrestore(&osb->dc_task_lock, flags);
 264
 265         spin_lock(&osb->osb_lock);
 266         out += scnprintf(buf + out, len - out, "%10s => Pid: %d  Nodes:",
 267                         "Recovery",
 268                         (osb->recovery_thread_task ?
 269                          task_pid_nr(osb->recovery_thread_task) : -1));
 270         if (rm->rm_used == 0)
 271                 out += scnprintf(buf + out, len - out, " None\n");
 272         else {
 273                 for (i = 0; i < rm->rm_used; i++)
 274                         out += scnprintf(buf + out, len - out, " %d",
 275                                         rm->rm_entries[i]);
 276                 out += scnprintf(buf + out, len - out, "\n");
 277         }
 278         spin_unlock(&osb->osb_lock);
 279
 280         out += scnprintf(buf + out, len - out,
 281                         "%10s => Pid: %d  Interval: %lu\n", "Commit",
 282                         (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
 283                         osb->osb_commit_interval);
 284
 285         out += scnprintf(buf + out, len - out,
 286                         "%10s => State: %d  TxnId: %lu  NumTxns: %d\n",
 287                         "Journal", osb->journal->j_state,
 288                         osb->journal->j_trans_id,
 289                         atomic_read(&osb->journal->j_num_trans));
 290
 291         out += scnprintf(buf + out, len - out,
 292                         "%10s => GlobalAllocs: %d  LocalAllocs: %d  "
 293                         "SubAllocs: %d  LAWinMoves: %d  SAExtends: %d\n",
 294                         "Stats",
 295                         atomic_read(&osb->alloc_stats.bitmap_data),
 296                         atomic_read(&osb->alloc_stats.local_data),
 297                         atomic_read(&osb->alloc_stats.bg_allocs),
 298                         atomic_read(&osb->alloc_stats.moves),
 299                         atomic_read(&osb->alloc_stats.bg_extends));
 300
 301         out += scnprintf(buf + out, len - out,
 302                         "%10s => State: %u  Descriptor: %llu  Size: %u bits  "
 303                         "Default: %u bits\n",
 304                         "LocalAlloc", osb->local_alloc_state,
 305                         (unsigned long long)osb->la_last_gd,
 306                         osb->local_alloc_bits, osb->local_alloc_default_bits);
 307
 308         spin_lock(&osb->osb_lock);
 309         out += scnprintf(buf + out, len - out,
 310                         "%10s => InodeSlot: %d  StolenInodes: %d, "
 311                         "MetaSlot: %d  StolenMeta: %d\n", "Steal",
 312                         osb->s_inode_steal_slot,
 313                         atomic_read(&osb->s_num_inodes_stolen),
 314                         osb->s_meta_steal_slot,
 315                         atomic_read(&osb->s_num_meta_stolen));
 316         spin_unlock(&osb->osb_lock);
 317
 318         out += scnprintf(buf + out, len - out, "OrphanScan => ");
 319         out += scnprintf(buf + out, len - out, "Local: %u  Global: %u ",
 320                         os->os_count, os->os_seqno);
 321         out += scnprintf(buf + out, len - out, " Last Scan: ");
 322         if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
 323                 out += scnprintf(buf + out, len - out, "Disabled\n");
 324         else
 325                 out += scnprintf(buf + out, len - out, "%lu seconds ago\n",
 326                                 (unsigned long)(ktime_get_seconds() - os->os_scantime));
 327
 328         out += scnprintf(buf + out, len - out, "%10s => %3s  %10s\n",
 329                         "Slots", "Num", "RecoGen");
 330         for (i = 0; i < osb->max_slots; ++i) {
 331                 out += scnprintf(buf + out, len - out,
 332                                 "%10s  %c %3d  %10d\n",
 333                                 " ",
 334                                 (i == osb->slot_num ? '*' : ' '),
 335                                 i, osb->slot_recovery_generations[i]);
 336         }
 337
 338         return out;
 339 }
 340
 341 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 342 {
 343         struct ocfs2_super *osb = inode->i_private;
 344         char *buf = NULL;
 345
 346         buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
 347         if (!buf)
 348                 goto bail;
 349
 350         i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
 351
 352         file->private_data = buf;
 353
 354         return 0;
 355 bail:
 356         return -ENOMEM;
 357 }
 358
 359 static int ocfs2_debug_release(struct inode *inode, struct file *file)
 360 {
 361         kfree(file->private_data);
 362         return 0;
 363 }
 364
 365 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 366                                 size_t nbytes, loff_t *ppos)
 367 {
 368         return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
 369                                        i_size_read(file->f_mapping->host));
 370 }
 371 #else
 372 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 373 {
 374         return 0;
 375 }
 376 static int ocfs2_debug_release(struct inode *inode, struct file *file)
 377 {
 378         return 0;
 379 }
 380 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 381                                 size_t nbytes, loff_t *ppos)
 382 {
 383         return 0;
 384 }
 385 #endif  /* CONFIG_DEBUG_FS */
 386
 387 static const struct file_operations ocfs2_osb_debug_fops = {
 388         .open =         ocfs2_osb_debug_open,
 389         .release =      ocfs2_debug_release,
 390         .read =         ocfs2_debug_read,
 391         .llseek =       generic_file_llseek,
 392 };
 393
 394 static int ocfs2_sync_fs(struct super_block *sb, int wait)
 395 {
 396         int status;
 397         tid_t target;
 398         struct ocfs2_super *osb = OCFS2_SB(sb);
 399
 400         if (ocfs2_is_hard_readonly(osb))
 401                 return -EROFS;
 402
 403         if (wait) {
 404                 status = ocfs2_flush_truncate_log(osb);
 405                 if (status < 0)
 406                         mlog_errno(status);
 407         } else {
 408                 ocfs2_schedule_truncate_log_flush(osb, 0);
 409         }
 410
 411         if (jbd2_journal_start_commit(osb->journal->j_journal,
 412                                       &target)) {
 413                 if (wait)
 414                         jbd2_log_wait_commit(osb->journal->j_journal,
 415                                              target);
 416         }
 417         return 0;
 418 }
 419
 420 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
 421 {
 422         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
 423             && (ino == USER_QUOTA_SYSTEM_INODE
 424                 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
 425                 return 0;
 426         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
 427             && (ino == GROUP_QUOTA_SYSTEM_INODE
 428                 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
 429                 return 0;
 430         return 1;
 431 }
 432
 433 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
 434 {
 435         struct inode *new = NULL;
 436         int status = 0;
 437         int i;
 438
 439         new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 440         if (IS_ERR(new)) {
 441                 status = PTR_ERR(new);
 442                 mlog_errno(status);
 443                 goto bail;
 444         }
 445         osb->root_inode = new;
 446
 447         new = ocfs2_iget(osb, osb->system_dir_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->sys_root_inode = new;
 454
 455         for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
 456              i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
 457                 if (!ocfs2_need_system_inode(osb, i))
 458                         continue;
 459                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 460                 if (!new) {
 461                         ocfs2_release_system_inodes(osb);
 462                         status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
 463                         mlog_errno(status);
 464                         mlog(ML_ERROR, "Unable to load system inode %d, "
 465                              "possibly corrupt fs?", i);
 466                         goto bail;
 467                 }
 468                 // the array now has one ref, so drop this one
 469                 iput(new);
 470         }
 471
 472 bail:
 473         if (status)
 474                 mlog_errno(status);
 475         return status;
 476 }
 477
 478 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
 479 {
 480         struct inode *new = NULL;
 481         int status = 0;
 482         int i;
 483
 484         for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
 485              i < NUM_SYSTEM_INODES;
 486              i++) {
 487                 if (!ocfs2_need_system_inode(osb, i))
 488                         continue;
 489                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 490                 if (!new) {
 491                         ocfs2_release_system_inodes(osb);
 492                         status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
 493                         mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
 494                              status, i, osb->slot_num);
 495                         goto bail;
 496                 }
 497                 /* the array now has one ref, so drop this one */
 498                 iput(new);
 499         }
 500
 501 bail:
 502         if (status)
 503                 mlog_errno(status);
 504         return status;
 505 }
 506
 507 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
 508 {
 509         int i;
 510         struct inode *inode;
 511
 512         for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
 513                 inode = osb->global_system_inodes[i];
 514                 if (inode) {
 515                         iput(inode);
 516                         osb->global_system_inodes[i] = NULL;
 517                 }
 518         }
 519
 520         inode = osb->sys_root_inode;
 521         if (inode) {
 522                 iput(inode);
 523                 osb->sys_root_inode = NULL;
 524         }
 525
 526         inode = osb->root_inode;
 527         if (inode) {
 528                 iput(inode);
 529                 osb->root_inode = NULL;
 530         }
 531
 532         if (!osb->local_system_inodes)
 533                 return;
 534
 535         for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
 536                 if (osb->local_system_inodes[i]) {
 537                         iput(osb->local_system_inodes[i]);
 538                         osb->local_system_inodes[i] = NULL;
 539                 }
 540         }
 541
 542         kfree(osb->local_system_inodes);
 543         osb->local_system_inodes = NULL;
 544 }
 545
 546 /* We're allocating fs objects, use GFP_NOFS */
 547 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
 548 {
 549         struct ocfs2_inode_info *oi;
 550
 551         oi = alloc_inode_sb(sb, ocfs2_inode_cachep, GFP_NOFS);
 552         if (!oi)
 553                 return NULL;
 554
 555         oi->i_sync_tid = 0;
 556         oi->i_datasync_tid = 0;
 557         memset(&oi->i_dquot, 0, sizeof(oi->i_dquot));
 558
 559         jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
 560         return &oi->vfs_inode;
 561 }
 562
 563 static void ocfs2_free_inode(struct inode *inode)
 564 {
 565         kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
 566 }
 567
 568 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
 569                                                 unsigned int cbits)
 570 {
 571         unsigned int bytes = 1 << cbits;
 572         unsigned int trim = bytes;
 573         unsigned int bitshift = 32;
 574
 575         /*
 576          * i_size and all block offsets in ocfs2 are always 64 bits
 577          * wide. i_clusters is 32 bits, in cluster-sized units. So on
 578          * 64 bit platforms, cluster size will be the limiting factor.
 579          */
 580
 581 #if BITS_PER_LONG == 32
 582         BUILD_BUG_ON(sizeof(sector_t) != 8);
 583         /*
 584          * We might be limited by page cache size.
 585          */
 586         if (bytes > PAGE_SIZE) {
 587                 bytes = PAGE_SIZE;
 588                 trim = 1;
 589                 /*
 590                  * Shift by 31 here so that we don't get larger than
 591                  * MAX_LFS_FILESIZE
 592                  */
 593                 bitshift = 31;
 594         }
 595 #endif
 596
 597         /*
 598          * Trim by a whole cluster when we can actually approach the
 599          * on-disk limits. Otherwise we can overflow i_clusters when
 600          * an extent start is at the max offset.
 601          */
 602         return (((unsigned long long)bytes) << bitshift) - trim;
 603 }
 604
 605 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
 606 {
 607         int incompat_features;
 608         int ret = 0;
 609         struct mount_options parsed_options;
 610         struct ocfs2_super *osb = OCFS2_SB(sb);
 611         u32 tmp;
 612
 613         sync_filesystem(sb);
 614
 615         if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
 616             !ocfs2_check_set_options(sb, &parsed_options)) {
 617                 ret = -EINVAL;
 618                 goto out;
 619         }
 620
 621         tmp = OCFS2_MOUNT_NOCLUSTER;
 622         if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
 623                 ret = -EINVAL;
 624                 mlog(ML_ERROR, "Cannot change nocluster option on remount\n");
 625                 goto out;
 626         }
 627
 628         tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
 629                 OCFS2_MOUNT_HB_NONE;
 630         if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
 631                 ret = -EINVAL;
 632                 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
 633                 goto out;
 634         }
 635
 636         if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
 637             (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
 638                 ret = -EINVAL;
 639                 mlog(ML_ERROR, "Cannot change data mode on remount\n");
 640                 goto out;
 641         }
 642
 643         /* Probably don't want this on remount; it might
 644          * mess with other nodes */
 645         if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
 646             (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
 647                 ret = -EINVAL;
 648                 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
 649                 goto out;
 650         }
 651
 652         /* We're going to/from readonly mode. */
 653         if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
 654                 /* Disable quota accounting before remounting RO */
 655                 if (*flags & SB_RDONLY) {
 656                         ret = ocfs2_susp_quotas(osb, 0);
 657                         if (ret < 0)
 658                                 goto out;
 659                 }
 660                 /* Lock here so the check of HARD_RO and the potential
 661                  * setting of SOFT_RO is atomic. */
 662                 spin_lock(&osb->osb_lock);
 663                 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
 664                         mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
 665                         ret = -EROFS;
 666                         goto unlock_osb;
 667                 }
 668
 669                 if (*flags & SB_RDONLY) {
 670                         sb->s_flags |= SB_RDONLY;
 671                         osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 672                 } else {
 673                         if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
 674                                 mlog(ML_ERROR, "Cannot remount RDWR "
 675                                      "filesystem due to previous errors.\n");
 676                                 ret = -EROFS;
 677                                 goto unlock_osb;
 678                         }
 679                         incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
 680                         if (incompat_features) {
 681                                 mlog(ML_ERROR, "Cannot remount RDWR because "
 682                                      "of unsupported optional features "
 683                                      "(%x).\n", incompat_features);
 684                                 ret = -EINVAL;
 685                                 goto unlock_osb;
 686                         }
 687                         sb->s_flags &= ~SB_RDONLY;
 688                         osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
 689                 }
 690                 trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags);
 691 unlock_osb:
 692                 spin_unlock(&osb->osb_lock);
 693                 /* Enable quota accounting after remounting RW */
 694                 if (!ret && !(*flags & SB_RDONLY)) {
 695                         if (sb_any_quota_suspended(sb))
 696                                 ret = ocfs2_susp_quotas(osb, 1);
 697                         else
 698                                 ret = ocfs2_enable_quotas(osb);
 699                         if (ret < 0) {
 700                                 /* Return back changes... */
 701                                 spin_lock(&osb->osb_lock);
 702                                 sb->s_flags |= SB_RDONLY;
 703                                 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 704                                 spin_unlock(&osb->osb_lock);
 705                                 goto out;
 706                         }
 707                 }
 708         }
 709
 710         if (!ret) {
 711                 /* Only save off the new mount options in case of a successful
 712                  * remount. */
 713                 osb->s_mount_opt = parsed_options.mount_opt;
 714                 osb->s_atime_quantum = parsed_options.atime_quantum;
 715                 osb->preferred_slot = parsed_options.slot;
 716                 if (parsed_options.commit_interval)
 717                         osb->osb_commit_interval = parsed_options.commit_interval;
 718
 719                 if (!ocfs2_is_hard_readonly(osb))
 720                         ocfs2_set_journal_params(osb);
 721
 722                 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
 723                         ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
 724                                                         SB_POSIXACL : 0);
 725         }
 726 out:
 727         return ret;
 728 }
 729
 730 static int ocfs2_sb_probe(struct super_block *sb,
 731                           struct buffer_head **bh,
 732                           int *sector_size,
 733                           struct ocfs2_blockcheck_stats *stats)
 734 {
 735         int status, tmpstat;
 736         struct ocfs1_vol_disk_hdr *hdr;
 737         struct ocfs2_dinode *di;
 738         int blksize;
 739
 740         *bh = NULL;
 741
 742         /* may be > 512 */
 743         *sector_size = bdev_logical_block_size(sb->s_bdev);
 744         if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
 745                 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
 746                      *sector_size, OCFS2_MAX_BLOCKSIZE);
 747                 status = -EINVAL;
 748                 goto bail;
 749         }
 750
 751         /* Can this really happen? */
 752         if (*sector_size < OCFS2_MIN_BLOCKSIZE)
 753                 *sector_size = OCFS2_MIN_BLOCKSIZE;
 754
 755         /* check block zero for old format */
 756         status = ocfs2_get_sector(sb, bh, 0, *sector_size);
 757         if (status < 0) {
 758                 mlog_errno(status);
 759                 goto bail;
 760         }
 761         hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
 762         if (hdr->major_version == OCFS1_MAJOR_VERSION) {
 763                 mlog(ML_ERROR, "incompatible version: %u.%u\n",
 764                      hdr->major_version, hdr->minor_version);
 765                 status = -EINVAL;
 766         }
 767         if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
 768                    strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
 769                 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
 770                      hdr->signature);
 771                 status = -EINVAL;
 772         }
 773         brelse(*bh);
 774         *bh = NULL;
 775         if (status < 0) {
 776                 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
 777                      "upgraded before mounting with ocfs v2\n");
 778                 goto bail;
 779         }
 780
 781         /*
 782          * Now check at magic offset for 512, 1024, 2048, 4096
 783          * blocksizes.  4096 is the maximum blocksize because it is
 784          * the minimum clustersize.
 785          */
 786         status = -EINVAL;
 787         for (blksize = *sector_size;
 788              blksize <= OCFS2_MAX_BLOCKSIZE;
 789              blksize <<= 1) {
 790                 tmpstat = ocfs2_get_sector(sb, bh,
 791                                            OCFS2_SUPER_BLOCK_BLKNO,
 792                                            blksize);
 793                 if (tmpstat < 0) {
 794                         status = tmpstat;
 795                         mlog_errno(status);
 796                         break;
 797                 }
 798                 di = (struct ocfs2_dinode *) (*bh)->b_data;
 799                 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
 800                 spin_lock_init(&stats->b_lock);
 801                 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
 802                 if (tmpstat < 0) {
 803                         brelse(*bh);
 804                         *bh = NULL;
 805                 }
 806                 if (tmpstat != -EAGAIN) {
 807                         status = tmpstat;
 808                         break;
 809                 }
 810         }
 811
 812 bail:
 813         return status;
 814 }
 815
 816 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
 817 {
 818         u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
 819
 820         if (osb->s_mount_opt & hb_enabled) {
 821                 if (ocfs2_mount_local(osb)) {
 822                         mlog(ML_ERROR, "Cannot heartbeat on a locally "
 823                              "mounted device.\n");
 824                         return -EINVAL;
 825                 }
 826                 if (ocfs2_userspace_stack(osb)) {
 827                         mlog(ML_ERROR, "Userspace stack expected, but "
 828                              "o2cb heartbeat arguments passed to mount\n");
 829                         return -EINVAL;
 830                 }
 831                 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
 832                      !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
 833                     ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
 834                      ocfs2_cluster_o2cb_global_heartbeat(osb))) {
 835                         mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
 836                         return -EINVAL;
 837                 }
 838         }
 839
 840         if (!(osb->s_mount_opt & hb_enabled)) {
 841                 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
 842                     !ocfs2_userspace_stack(osb)) {
 843                         mlog(ML_ERROR, "Heartbeat has to be started to mount "
 844                              "a read-write clustered device.\n");
 845                         return -EINVAL;
 846                 }
 847         }
 848
 849         return 0;
 850 }
 851
 852 /*
 853  * If we're using a userspace stack, mount should have passed
 854  * a name that matches the disk.  If not, mount should not
 855  * have passed a stack.
 856  */
 857 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
 858                                         struct mount_options *mopt)
 859 {
 860         if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
 861                 mlog(ML_ERROR,
 862                      "cluster stack passed to mount, but this filesystem "
 863                      "does not support it\n");
 864                 return -EINVAL;
 865         }
 866
 867         if (ocfs2_userspace_stack(osb) &&
 868             !(osb->s_mount_opt & OCFS2_MOUNT_NOCLUSTER) &&
 869             strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
 870                     OCFS2_STACK_LABEL_LEN)) {
 871                 mlog(ML_ERROR,
 872                      "cluster stack passed to mount (\"%s\") does not "
 873                      "match the filesystem (\"%s\")\n",
 874                      mopt->cluster_stack,
 875                      osb->osb_cluster_stack);
 876                 return -EINVAL;
 877         }
 878
 879         return 0;
 880 }
 881
 882 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
 883 {
 884         int type;
 885         struct super_block *sb = osb->sb;
 886         unsigned int feature[OCFS2_MAXQUOTAS] = {
 887                                         OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 888                                         OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 889         int status = 0;
 890
 891         for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
 892                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 893                         continue;
 894                 if (unsuspend)
 895                         status = dquot_resume(sb, type);
 896                 else {
 897                         struct ocfs2_mem_dqinfo *oinfo;
 898
 899                         /* Cancel periodic syncing before suspending */
 900                         oinfo = sb_dqinfo(sb, type)->dqi_priv;
 901                         cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 902                         status = dquot_suspend(sb, type);
 903                 }
 904                 if (status < 0)
 905                         break;
 906         }
 907         if (status < 0)
 908                 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
 909                      "remount (error = %d).\n", status);
 910         return status;
 911 }
 912
 913 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
 914 {
 915         struct inode *inode[OCFS2_MAXQUOTAS] = { NULL, NULL };
 916         struct super_block *sb = osb->sb;
 917         unsigned int feature[OCFS2_MAXQUOTAS] = {
 918                                         OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 919                                         OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 920         unsigned int ino[OCFS2_MAXQUOTAS] = {
 921                                         LOCAL_USER_QUOTA_SYSTEM_INODE,
 922                                         LOCAL_GROUP_QUOTA_SYSTEM_INODE };
 923         int status;
 924         int type;
 925
 926         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
 927         for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
 928                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 929                         continue;
 930                 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
 931                                                         osb->slot_num);
 932                 if (!inode[type]) {
 933                         status = -ENOENT;
 934                         goto out_quota_off;
 935                 }
 936                 status = dquot_load_quota_inode(inode[type], type, QFMT_OCFS2,
 937                                                 DQUOT_USAGE_ENABLED);
 938                 if (status < 0)
 939                         goto out_quota_off;
 940         }
 941
 942         for (type = 0; type < OCFS2_MAXQUOTAS; type++)
 943                 iput(inode[type]);
 944         return 0;
 945 out_quota_off:
 946         ocfs2_disable_quotas(osb);
 947         for (type = 0; type < OCFS2_MAXQUOTAS; type++)
 948                 iput(inode[type]);
 949         mlog_errno(status);
 950         return status;
 951 }
 952
 953 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
 954 {
 955         int type;
 956         struct inode *inode;
 957         struct super_block *sb = osb->sb;
 958         struct ocfs2_mem_dqinfo *oinfo;
 959
 960         /* We mostly ignore errors in this function because there's not much
 961          * we can do when we see them */
 962         for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
 963                 if (!sb_has_quota_loaded(sb, type))
 964                         continue;
 965                 oinfo = sb_dqinfo(sb, type)->dqi_priv;
 966                 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 967                 inode = igrab(sb->s_dquot.files[type]);
 968                 /* Turn off quotas. This will remove all dquot structures from
 969                  * memory and so they will be automatically synced to global
 970                  * quota files */
 971                 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
 972                                         DQUOT_LIMITS_ENABLED);
 973                 iput(inode);
 974         }
 975 }
 976
 977 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
 978 {
 979         struct dentry *root;
 980         int status, sector_size;
 981         struct mount_options parsed_options;
 982         struct inode *inode = NULL;
 983         struct ocfs2_super *osb = NULL;
 984         struct buffer_head *bh = NULL;
 985         char nodestr[12];
 986         struct ocfs2_blockcheck_stats stats;
 987
 988         trace_ocfs2_fill_super(sb, data, silent);
 989
 990         if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
 991                 status = -EINVAL;
 992                 goto out;
 993         }
 994
 995         /* probe for superblock */
 996         status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
 997         if (status < 0) {
 998                 mlog(ML_ERROR, "superblock probe failed!\n");
 999                 goto out;
1000         }
1001
1002         status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1003         brelse(bh);
1004         bh = NULL;
1005         if (status < 0)
1006                 goto out;
1007
1008         osb = OCFS2_SB(sb);
1009
1010         if (!ocfs2_check_set_options(sb, &parsed_options)) {
1011                 status = -EINVAL;
1012                 goto out_super;
1013         }
1014         osb->s_mount_opt = parsed_options.mount_opt;
1015         osb->s_atime_quantum = parsed_options.atime_quantum;
1016         osb->preferred_slot = parsed_options.slot;
1017         osb->osb_commit_interval = parsed_options.commit_interval;
1018
1019         ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1020         osb->osb_resv_level = parsed_options.resv_level;
1021         osb->osb_dir_resv_level = parsed_options.resv_level;
1022         if (parsed_options.dir_resv_level == -1)
1023                 osb->osb_dir_resv_level = parsed_options.resv_level;
1024         else
1025                 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1026
1027         status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1028         if (status)
1029                 goto out_super;
1030
1031         sb->s_magic = OCFS2_SUPER_MAGIC;
1032
1033         sb->s_flags = (sb->s_flags & ~(SB_POSIXACL | SB_NOSEC)) |
1034                 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? SB_POSIXACL : 0);
1035
1036         /* Hard readonly mode only if: bdev_read_only, SB_RDONLY,
1037          * heartbeat=none */
1038         if (bdev_read_only(sb->s_bdev)) {
1039                 if (!sb_rdonly(sb)) {
1040                         status = -EACCES;
1041                         mlog(ML_ERROR, "Readonly device detected but readonly "
1042                              "mount was not specified.\n");
1043                         goto out_super;
1044                 }
1045
1046                 /* You should not be able to start a local heartbeat
1047                  * on a readonly device. */
1048                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1049                         status = -EROFS;
1050                         mlog(ML_ERROR, "Local heartbeat specified on readonly "
1051                              "device.\n");
1052                         goto out_super;
1053                 }
1054
1055                 status = ocfs2_check_journals_nolocks(osb);
1056                 if (status < 0) {
1057                         if (status == -EROFS)
1058                                 mlog(ML_ERROR, "Recovery required on readonly "
1059                                      "file system, but write access is "
1060                                      "unavailable.\n");
1061                         goto out_super;
1062                 }
1063
1064                 ocfs2_set_ro_flag(osb, 1);
1065
1066                 printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. "
1067                        "Cluster services will not be used for this mount. "
1068                        "Recovery will be skipped.\n", osb->dev_str);
1069         }
1070
1071         if (!ocfs2_is_hard_readonly(osb)) {
1072                 if (sb_rdonly(sb))
1073                         ocfs2_set_ro_flag(osb, 0);
1074         }
1075
1076         status = ocfs2_verify_heartbeat(osb);
1077         if (status < 0)
1078                 goto out_super;
1079
1080         osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1081                                                  ocfs2_debugfs_root);
1082
1083         debugfs_create_file("fs_state", S_IFREG|S_IRUSR, osb->osb_debug_root,
1084                             osb, &ocfs2_osb_debug_fops);
1085
1086         if (ocfs2_meta_ecc(osb))
1087                 ocfs2_blockcheck_stats_debugfs_install( &osb->osb_ecc_stats,
1088                                                         osb->osb_debug_root);
1089
1090         status = ocfs2_mount_volume(sb);
1091         if (status < 0)
1092                 goto out_debugfs;
1093
1094         if (osb->root_inode)
1095                 inode = igrab(osb->root_inode);
1096
1097         if (!inode) {
1098                 status = -EIO;
1099                 goto out_dismount;
1100         }
1101
1102         osb->osb_dev_kset = kset_create_and_add(sb->s_id, NULL,
1103                                                 &ocfs2_kset->kobj);
1104         if (!osb->osb_dev_kset) {
1105                 status = -ENOMEM;
1106                 mlog(ML_ERROR, "Unable to create device kset %s.\n", sb->s_id);
1107                 goto out_dismount;
1108         }
1109
1110         /* Create filecheck sysfs related directories/files at
1111          * /sys/fs/ocfs2/<devname>/filecheck */
1112         if (ocfs2_filecheck_create_sysfs(osb)) {
1113                 status = -ENOMEM;
1114                 mlog(ML_ERROR, "Unable to create filecheck sysfs directory at "
1115                         "/sys/fs/ocfs2/%s/filecheck.\n", sb->s_id);
1116                 goto out_dismount;
1117         }
1118
1119         root = d_make_root(inode);
1120         if (!root) {
1121                 status = -ENOMEM;
1122                 goto out_dismount;
1123         }
1124
1125         sb->s_root = root;
1126
1127         ocfs2_complete_mount_recovery(osb);
1128
1129         if (ocfs2_mount_local(osb))
1130                 snprintf(nodestr, sizeof(nodestr), "local");
1131         else
1132                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1133
1134         printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1135                "with %s data mode.\n",
1136                osb->dev_str, nodestr, osb->slot_num,
1137                osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1138                "ordered");
1139
1140         if ((osb->s_mount_opt & OCFS2_MOUNT_NOCLUSTER) &&
1141            !(osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT))
1142                 printk(KERN_NOTICE "ocfs2: The shared device (%s) is mounted "
1143                        "without cluster aware mode.\n", osb->dev_str);
1144
1145         atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1146         wake_up(&osb->osb_mount_event);
1147
1148         /* Now we can initialize quotas because we can afford to wait
1149          * for cluster locks recovery now. That also means that truncation
1150          * log recovery can happen but that waits for proper quota setup */
1151         if (!sb_rdonly(sb)) {
1152                 status = ocfs2_enable_quotas(osb);
1153                 if (status < 0) {
1154                         /* We have to err-out specially here because
1155                          * s_root is already set */
1156                         mlog_errno(status);
1157                         atomic_set(&osb->vol_state, VOLUME_DISABLED);
1158                         wake_up(&osb->osb_mount_event);
1159                         return status;
1160                 }
1161         }
1162
1163         ocfs2_complete_quota_recovery(osb);
1164
1165         /* Now we wake up again for processes waiting for quotas */
1166         atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1167         wake_up(&osb->osb_mount_event);
1168
1169         /* Start this when the mount is almost sure of being successful */
1170         ocfs2_orphan_scan_start(osb);
1171
1172         return status;
1173
1174 out_dismount:
1175         atomic_set(&osb->vol_state, VOLUME_DISABLED);
1176         wake_up(&osb->osb_mount_event);
1177         ocfs2_dismount_volume(sb, 1);
1178         goto out;
1179
1180 out_debugfs:
1181         debugfs_remove_recursive(osb->osb_debug_root);
1182 out_super:
1183         ocfs2_release_system_inodes(osb);
1184         kfree(osb->recovery_map);
1185         ocfs2_delete_osb(osb);
1186         kfree(osb);
1187 out:
1188         mlog_errno(status);
1189
1190         return status;
1191 }
1192
1193 static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
1194                         int flags,
1195                         const char *dev_name,
1196                         void *data)
1197 {
1198         return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
1199 }
1200
1201 static struct file_system_type ocfs2_fs_type = {
1202         .owner          = THIS_MODULE,
1203         .name           = "ocfs2",
1204         .mount          = ocfs2_mount,
1205         .kill_sb        = kill_block_super,
1206         .fs_flags       = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1207         .next           = NULL
1208 };
1209 MODULE_ALIAS_FS("ocfs2");
1210
1211 static int ocfs2_check_set_options(struct super_block *sb,
1212                                    struct mount_options *options)
1213 {
1214         if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1215             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1216                                          OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1217                 mlog(ML_ERROR, "User quotas were requested, but this "
1218                      "filesystem does not have the feature enabled.\n");
1219                 return 0;
1220         }
1221         if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1222             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1223                                          OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1224                 mlog(ML_ERROR, "Group quotas were requested, but this "
1225                      "filesystem does not have the feature enabled.\n");
1226                 return 0;
1227         }
1228         if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1229             !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1230                 mlog(ML_ERROR, "ACL support requested but extended attributes "
1231                      "feature is not enabled\n");
1232                 return 0;
1233         }
1234         /* No ACL setting specified? Use XATTR feature... */
1235         if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1236                                     OCFS2_MOUNT_NO_POSIX_ACL))) {
1237                 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1238                         options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1239                 else
1240                         options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1241         }
1242         return 1;
1243 }
1244
1245 static int ocfs2_parse_options(struct super_block *sb,
1246                                char *options,
1247                                struct mount_options *mopt,
1248                                int is_remount)
1249 {
1250         int status, user_stack = 0;
1251         char *p;
1252         u32 tmp;
1253         int token, option;
1254         substring_t args[MAX_OPT_ARGS];
1255
1256         trace_ocfs2_parse_options(is_remount, options ? options : "(none)");
1257
1258         mopt->commit_interval = 0;
1259         mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1260         mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1261         mopt->slot = OCFS2_INVALID_SLOT;
1262         mopt->localalloc_opt = -1;
1263         mopt->cluster_stack[0] = '\0';
1264         mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1265         mopt->dir_resv_level = -1;
1266
1267         if (!options) {
1268                 status = 1;
1269                 goto bail;
1270         }
1271
1272         while ((p = strsep(&options, ",")) != NULL) {
1273                 if (!*p)
1274                         continue;
1275
1276                 token = match_token(p, tokens, args);
1277                 switch (token) {
1278                 case Opt_hb_local:
1279                         mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1280                         break;
1281                 case Opt_hb_none:
1282                         mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1283                         break;
1284                 case Opt_hb_global:
1285                         mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1286                         break;
1287                 case Opt_barrier:
1288                         if (match_int(&args[0], &option)) {
1289                                 status = 0;
1290                                 goto bail;
1291                         }
1292                         if (option)
1293                                 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1294                         else
1295                                 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1296                         break;
1297                 case Opt_intr:
1298                         mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1299                         break;
1300                 case Opt_nointr:
1301                         mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1302                         break;
1303                 case Opt_err_panic:
1304                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1305                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1306                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1307                         break;
1308                 case Opt_err_ro:
1309                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1310                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1311                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_ROFS;
1312                         break;
1313                 case Opt_err_cont:
1314                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1315                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1316                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_CONT;
1317                         break;
1318                 case Opt_data_ordered:
1319                         mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1320                         break;
1321                 case Opt_data_writeback:
1322                         mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1323                         break;
1324                 case Opt_user_xattr:
1325                         mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1326                         break;
1327                 case Opt_nouser_xattr:
1328                         mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1329                         break;
1330                 case Opt_atime_quantum:
1331                         if (match_int(&args[0], &option)) {
1332                                 status = 0;
1333                                 goto bail;
1334                         }
1335                         if (option >= 0)
1336                                 mopt->atime_quantum = option;
1337                         break;
1338                 case Opt_slot:
1339                         if (match_int(&args[0], &option)) {
1340                                 status = 0;
1341                                 goto bail;
1342                         }
1343                         if (option)
1344                                 mopt->slot = (u16)option;
1345                         break;
1346                 case Opt_commit:
1347                         if (match_int(&args[0], &option)) {
1348                                 status = 0;
1349                                 goto bail;
1350                         }
1351                         if (option < 0)
1352                                 return 0;
1353                         if (option == 0)
1354                                 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1355                         mopt->commit_interval = HZ * option;
1356                         break;
1357                 case Opt_localalloc:
1358                         if (match_int(&args[0], &option)) {
1359                                 status = 0;
1360                                 goto bail;
1361                         }
1362                         if (option >= 0)
1363                                 mopt->localalloc_opt = option;
1364                         break;
1365                 case Opt_localflocks:
1366                         /*
1367                          * Changing this during remount could race
1368                          * flock() requests, or "unbalance" existing
1369                          * ones (e.g., a lock is taken in one mode but
1370                          * dropped in the other). If users care enough
1371                          * to flip locking modes during remount, we
1372                          * could add a "local" flag to individual
1373                          * flock structures for proper tracking of
1374                          * state.
1375                          */
1376                         if (!is_remount)
1377                                 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1378                         break;
1379                 case Opt_stack:
1380                         /* Check both that the option we were passed
1381                          * is of the right length and that it is a proper
1382                          * string of the right length.
1383                          */
1384                         if (((args[0].to - args[0].from) !=
1385                              OCFS2_STACK_LABEL_LEN) ||
1386                             (strnlen(args[0].from,
1387                                      OCFS2_STACK_LABEL_LEN) !=
1388                              OCFS2_STACK_LABEL_LEN)) {
1389                                 mlog(ML_ERROR,
1390                                      "Invalid cluster_stack option\n");
1391                                 status = 0;
1392                                 goto bail;
1393                         }
1394                         memcpy(mopt->cluster_stack, args[0].from,
1395                                OCFS2_STACK_LABEL_LEN);
1396                         mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1397                         /*
1398                          * Open code the memcmp here as we don't have
1399                          * an osb to pass to
1400                          * ocfs2_userspace_stack().
1401                          */
1402                         if (memcmp(mopt->cluster_stack,
1403                                    OCFS2_CLASSIC_CLUSTER_STACK,
1404                                    OCFS2_STACK_LABEL_LEN))
1405                                 user_stack = 1;
1406                         break;
1407                 case Opt_inode64:
1408                         mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1409                         break;
1410                 case Opt_usrquota:
1411                         mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1412                         break;
1413                 case Opt_grpquota:
1414                         mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1415                         break;
1416                 case Opt_coherency_buffered:
1417                         mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1418                         break;
1419                 case Opt_coherency_full:
1420                         mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1421                         break;
1422                 case Opt_acl:
1423                         mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1424                         mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1425                         break;
1426                 case Opt_noacl:
1427                         mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1428                         mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1429                         break;
1430                 case Opt_resv_level:
1431                         if (is_remount)
1432                                 break;
1433                         if (match_int(&args[0], &option)) {
1434                                 status = 0;
1435                                 goto bail;
1436                         }
1437                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1438                             option < OCFS2_MAX_RESV_LEVEL)
1439                                 mopt->resv_level = option;
1440                         break;
1441                 case Opt_dir_resv_level:
1442                         if (is_remount)
1443                                 break;
1444                         if (match_int(&args[0], &option)) {
1445                                 status = 0;
1446                                 goto bail;
1447                         }
1448                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1449                             option < OCFS2_MAX_RESV_LEVEL)
1450                                 mopt->dir_resv_level = option;
1451                         break;
1452                 case Opt_journal_async_commit:
1453                         mopt->mount_opt |= OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT;
1454                         break;
1455                 case Opt_nocluster:
1456                         mopt->mount_opt |= OCFS2_MOUNT_NOCLUSTER;
1457                         break;
1458                 default:
1459                         mlog(ML_ERROR,
1460                              "Unrecognized mount option \"%s\" "
1461                              "or missing value\n", p);
1462                         status = 0;
1463                         goto bail;
1464                 }
1465         }
1466
1467         if (user_stack == 0) {
1468                 /* Ensure only one heartbeat mode */
1469                 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1470                                          OCFS2_MOUNT_HB_GLOBAL |
1471                                          OCFS2_MOUNT_HB_NONE);
1472                 if (hweight32(tmp) != 1) {
1473                         mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1474                         status = 0;
1475                         goto bail;
1476                 }
1477         }
1478
1479         status = 1;
1480
1481 bail:
1482         return status;
1483 }
1484
1485 static int ocfs2_show_options(struct seq_file *s, struct dentry *root)
1486 {
1487         struct ocfs2_super *osb = OCFS2_SB(root->d_sb);
1488         unsigned long opts = osb->s_mount_opt;
1489         unsigned int local_alloc_megs;
1490
1491         if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1492                 seq_printf(s, ",_netdev");
1493                 if (opts & OCFS2_MOUNT_HB_LOCAL)
1494                         seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1495                 else
1496                         seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1497         } else
1498                 seq_printf(s, ",%s", OCFS2_HB_NONE);
1499
1500         if (opts & OCFS2_MOUNT_NOINTR)
1501                 seq_printf(s, ",nointr");
1502
1503         if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1504                 seq_printf(s, ",data=writeback");
1505         else
1506                 seq_printf(s, ",data=ordered");
1507
1508         if (opts & OCFS2_MOUNT_BARRIER)
1509                 seq_printf(s, ",barrier=1");
1510
1511         if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1512                 seq_printf(s, ",errors=panic");
1513         else if (opts & OCFS2_MOUNT_ERRORS_CONT)
1514                 seq_printf(s, ",errors=continue");
1515         else
1516                 seq_printf(s, ",errors=remount-ro");
1517
1518         if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1519                 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1520
1521         seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1522
1523         if (osb->osb_commit_interval)
1524                 seq_printf(s, ",commit=%u",
1525                            (unsigned) (osb->osb_commit_interval / HZ));
1526
1527         local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1528         if (local_alloc_megs != ocfs2_la_default_mb(osb))
1529                 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1530
1531         if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1532                 seq_printf(s, ",localflocks,");
1533
1534         if (osb->osb_cluster_stack[0])
1535                 seq_show_option_n(s, "cluster_stack", osb->osb_cluster_stack,
1536                                   OCFS2_STACK_LABEL_LEN);
1537         if (opts & OCFS2_MOUNT_USRQUOTA)
1538                 seq_printf(s, ",usrquota");
1539         if (opts & OCFS2_MOUNT_GRPQUOTA)
1540                 seq_printf(s, ",grpquota");
1541
1542         if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1543                 seq_printf(s, ",coherency=buffered");
1544         else
1545                 seq_printf(s, ",coherency=full");
1546
1547         if (opts & OCFS2_MOUNT_NOUSERXATTR)
1548                 seq_printf(s, ",nouser_xattr");
1549         else
1550                 seq_printf(s, ",user_xattr");
1551
1552         if (opts & OCFS2_MOUNT_INODE64)
1553                 seq_printf(s, ",inode64");
1554
1555         if (opts & OCFS2_MOUNT_POSIX_ACL)
1556                 seq_printf(s, ",acl");
1557         else
1558                 seq_printf(s, ",noacl");
1559
1560         if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1561                 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1562
1563         if (osb->osb_dir_resv_level != osb->osb_resv_level)
1564                 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1565
1566         if (opts & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
1567                 seq_printf(s, ",journal_async_commit");
1568
1569         if (opts & OCFS2_MOUNT_NOCLUSTER)
1570                 seq_printf(s, ",nocluster");
1571
1572         return 0;
1573 }
1574
1575 static int __init ocfs2_init(void)
1576 {
1577         int status;
1578
1579         status = init_ocfs2_uptodate_cache();
1580         if (status < 0)
1581                 goto out1;
1582
1583         status = ocfs2_initialize_mem_caches();
1584         if (status < 0)
1585                 goto out2;
1586
1587         ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1588
1589         ocfs2_set_locking_protocol();
1590
1591         status = register_quota_format(&ocfs2_quota_format);
1592         if (status < 0)
1593                 goto out3;
1594         status = register_filesystem(&ocfs2_fs_type);
1595         if (!status)
1596                 return 0;
1597
1598         unregister_quota_format(&ocfs2_quota_format);
1599 out3:
1600         debugfs_remove(ocfs2_debugfs_root);
1601         ocfs2_free_mem_caches();
1602 out2:
1603         exit_ocfs2_uptodate_cache();
1604 out1:
1605         mlog_errno(status);
1606         return status;
1607 }
1608
1609 static void __exit ocfs2_exit(void)
1610 {
1611         unregister_quota_format(&ocfs2_quota_format);
1612
1613         debugfs_remove(ocfs2_debugfs_root);
1614
1615         ocfs2_free_mem_caches();
1616
1617         unregister_filesystem(&ocfs2_fs_type);
1618
1619         exit_ocfs2_uptodate_cache();
1620 }
1621
1622 static void ocfs2_put_super(struct super_block *sb)
1623 {
1624         trace_ocfs2_put_super(sb);
1625
1626         ocfs2_sync_blockdev(sb);
1627         ocfs2_dismount_volume(sb, 0);
1628 }
1629
1630 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1631 {
1632         struct ocfs2_super *osb;
1633         u32 numbits, freebits;
1634         int status;
1635         struct ocfs2_dinode *bm_lock;
1636         struct buffer_head *bh = NULL;
1637         struct inode *inode = NULL;
1638
1639         trace_ocfs2_statfs(dentry->d_sb, buf);
1640
1641         osb = OCFS2_SB(dentry->d_sb);
1642
1643         inode = ocfs2_get_system_file_inode(osb,
1644                                             GLOBAL_BITMAP_SYSTEM_INODE,
1645                                             OCFS2_INVALID_SLOT);
1646         if (!inode) {
1647                 mlog(ML_ERROR, "failed to get bitmap inode\n");
1648                 status = -EIO;
1649                 goto bail;
1650         }
1651
1652         status = ocfs2_inode_lock(inode, &bh, 0);
1653         if (status < 0) {
1654                 mlog_errno(status);
1655                 goto bail;
1656         }
1657
1658         bm_lock = (struct ocfs2_dinode *) bh->b_data;
1659
1660         numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1661         freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1662
1663         buf->f_type = OCFS2_SUPER_MAGIC;
1664         buf->f_bsize = dentry->d_sb->s_blocksize;
1665         buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1666         buf->f_blocks = ((sector_t) numbits) *
1667                         (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1668         buf->f_bfree = ((sector_t) freebits) *
1669                        (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1670         buf->f_bavail = buf->f_bfree;
1671         buf->f_files = numbits;
1672         buf->f_ffree = freebits;
1673         buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1674                                 & 0xFFFFFFFFUL;
1675         buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1676                                 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1677
1678         brelse(bh);
1679
1680         ocfs2_inode_unlock(inode, 0);
1681         status = 0;
1682 bail:
1683         iput(inode);
1684
1685         if (status)
1686                 mlog_errno(status);
1687
1688         return status;
1689 }
1690
1691 static void ocfs2_inode_init_once(void *data)
1692 {
1693         struct ocfs2_inode_info *oi = data;
1694
1695         oi->ip_flags = 0;
1696         oi->ip_open_count = 0;
1697         spin_lock_init(&oi->ip_lock);
1698         ocfs2_extent_map_init(&oi->vfs_inode);
1699         INIT_LIST_HEAD(&oi->ip_io_markers);
1700         INIT_LIST_HEAD(&oi->ip_unwritten_list);
1701         oi->ip_dir_start_lookup = 0;
1702         init_rwsem(&oi->ip_alloc_sem);
1703         init_rwsem(&oi->ip_xattr_sem);
1704         mutex_init(&oi->ip_io_mutex);
1705
1706         oi->ip_blkno = 0ULL;
1707         oi->ip_clusters = 0;
1708         oi->ip_next_orphan = NULL;
1709
1710         ocfs2_resv_init_once(&oi->ip_la_data_resv);
1711
1712         ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1713         ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1714         ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1715
1716         ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1717                                   &ocfs2_inode_caching_ops);
1718
1719         inode_init_once(&oi->vfs_inode);
1720 }
1721
1722 static int ocfs2_initialize_mem_caches(void)
1723 {
1724         ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1725                                        sizeof(struct ocfs2_inode_info),
1726                                        0,
1727                                        (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1728                                                 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1729                                        ocfs2_inode_init_once);
1730         ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1731                                         sizeof(struct ocfs2_dquot),
1732                                         0,
1733                                         (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1734                                                 SLAB_MEM_SPREAD),
1735                                         NULL);
1736         ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1737                                         sizeof(struct ocfs2_quota_chunk),
1738                                         0,
1739                                         (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1740                                         NULL);
1741         if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1742             !ocfs2_qf_chunk_cachep) {
1743                 kmem_cache_destroy(ocfs2_inode_cachep);
1744                 kmem_cache_destroy(ocfs2_dquot_cachep);
1745                 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1746                 return -ENOMEM;
1747         }
1748
1749         return 0;
1750 }
1751
1752 static void ocfs2_free_mem_caches(void)
1753 {
1754         /*
1755          * Make sure all delayed rcu free inodes are flushed before we
1756          * destroy cache.
1757          */
1758         rcu_barrier();
1759         kmem_cache_destroy(ocfs2_inode_cachep);
1760         ocfs2_inode_cachep = NULL;
1761
1762         kmem_cache_destroy(ocfs2_dquot_cachep);
1763         ocfs2_dquot_cachep = NULL;
1764
1765         kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1766         ocfs2_qf_chunk_cachep = NULL;
1767 }
1768
1769 static int ocfs2_get_sector(struct super_block *sb,
1770                             struct buffer_head **bh,
1771                             int block,
1772                             int sect_size)
1773 {
1774         if (!sb_set_blocksize(sb, sect_size)) {
1775                 mlog(ML_ERROR, "unable to set blocksize\n");
1776                 return -EIO;
1777         }
1778
1779         *bh = sb_getblk(sb, block);
1780         if (!*bh) {
1781                 mlog_errno(-ENOMEM);
1782                 return -ENOMEM;
1783         }
1784         lock_buffer(*bh);
1785         if (!buffer_dirty(*bh))
1786                 clear_buffer_uptodate(*bh);
1787         unlock_buffer(*bh);
1788         ll_rw_block(REQ_OP_READ, 0, 1, bh);
1789         wait_on_buffer(*bh);
1790         if (!buffer_uptodate(*bh)) {
1791                 mlog_errno(-EIO);
1792                 brelse(*bh);
1793                 *bh = NULL;
1794                 return -EIO;
1795         }
1796
1797         return 0;
1798 }
1799
1800 static int ocfs2_mount_volume(struct super_block *sb)
1801 {
1802         int status = 0;
1803         struct ocfs2_super *osb = OCFS2_SB(sb);
1804
1805         if (ocfs2_is_hard_readonly(osb))
1806                 goto out;
1807
1808         mutex_init(&osb->obs_trim_fs_mutex);
1809
1810         status = ocfs2_dlm_init(osb);
1811         if (status < 0) {
1812                 mlog_errno(status);
1813                 if (status == -EBADR && ocfs2_userspace_stack(osb))
1814                         mlog(ML_ERROR, "couldn't mount because cluster name on"
1815                         " disk does not match the running cluster name.\n");
1816                 goto out;
1817         }
1818
1819         status = ocfs2_super_lock(osb, 1);
1820         if (status < 0) {
1821                 mlog_errno(status);
1822                 goto out_dlm;
1823         }
1824
1825         /* This will load up the node map and add ourselves to it. */
1826         status = ocfs2_find_slot(osb);
1827         if (status < 0) {
1828                 mlog_errno(status);
1829                 goto out_super_lock;
1830         }
1831
1832         /* load all node-local system inodes */
1833         status = ocfs2_init_local_system_inodes(osb);
1834         if (status < 0) {
1835                 mlog_errno(status);
1836                 goto out_super_lock;
1837         }
1838
1839         status = ocfs2_check_volume(osb);
1840         if (status < 0) {
1841                 mlog_errno(status);
1842                 goto out_system_inodes;
1843         }
1844
1845         status = ocfs2_truncate_log_init(osb);
1846         if (status < 0) {
1847                 mlog_errno(status);
1848                 goto out_system_inodes;
1849         }
1850
1851         ocfs2_super_unlock(osb, 1);
1852         return 0;
1853
1854 out_system_inodes:
1855         if (osb->local_alloc_state == OCFS2_LA_ENABLED)
1856                 ocfs2_shutdown_local_alloc(osb);
1857         ocfs2_release_system_inodes(osb);
1858         /* before journal shutdown, we should release slot_info */
1859         ocfs2_free_slot_info(osb);
1860         ocfs2_journal_shutdown(osb);
1861 out_super_lock:
1862         ocfs2_super_unlock(osb, 1);
1863 out_dlm:
1864         ocfs2_dlm_shutdown(osb, 0);
1865 out:
1866         return status;
1867 }
1868
1869 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1870 {
1871         int tmp, hangup_needed = 0;
1872         struct ocfs2_super *osb = NULL;
1873         char nodestr[12];
1874
1875         trace_ocfs2_dismount_volume(sb);
1876
1877         BUG_ON(!sb);
1878         osb = OCFS2_SB(sb);
1879         BUG_ON(!osb);
1880
1881         /* Remove file check sysfs related directores/files,
1882          * and wait for the pending file check operations */
1883         ocfs2_filecheck_remove_sysfs(osb);
1884
1885         kset_unregister(osb->osb_dev_kset);
1886
1887         /* Orphan scan should be stopped as early as possible */
1888         ocfs2_orphan_scan_stop(osb);
1889
1890         ocfs2_disable_quotas(osb);
1891
1892         /* All dquots should be freed by now */
1893         WARN_ON(!llist_empty(&osb->dquot_drop_list));
1894         /* Wait for worker to be done with the work structure in osb */
1895         cancel_work_sync(&osb->dquot_drop_work);
1896
1897         ocfs2_shutdown_local_alloc(osb);
1898
1899         ocfs2_truncate_log_shutdown(osb);
1900
1901         /* This will disable recovery and flush any recovery work. */
1902         ocfs2_recovery_exit(osb);
1903
1904         ocfs2_sync_blockdev(sb);
1905
1906         ocfs2_purge_refcount_trees(osb);
1907
1908         /* No cluster connection means we've failed during mount, so skip
1909          * all the steps which depended on that to complete. */
1910         if (osb->cconn) {
1911                 tmp = ocfs2_super_lock(osb, 1);
1912                 if (tmp < 0) {
1913                         mlog_errno(tmp);
1914                         return;
1915                 }
1916         }
1917
1918         if (osb->slot_num != OCFS2_INVALID_SLOT)
1919                 ocfs2_put_slot(osb);
1920
1921         if (osb->cconn)
1922                 ocfs2_super_unlock(osb, 1);
1923
1924         ocfs2_release_system_inodes(osb);
1925
1926         ocfs2_journal_shutdown(osb);
1927
1928         /*
1929          * If we're dismounting due to mount error, mount.ocfs2 will clean
1930          * up heartbeat.  If we're a local mount, there is no heartbeat.
1931          * If we failed before we got a uuid_str yet, we can't stop
1932          * heartbeat.  Otherwise, do it.
1933          */
1934         if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str &&
1935             !ocfs2_is_hard_readonly(osb))
1936                 hangup_needed = 1;
1937
1938         if (osb->cconn)
1939                 ocfs2_dlm_shutdown(osb, hangup_needed);
1940
1941         ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1942         debugfs_remove_recursive(osb->osb_debug_root);
1943
1944         if (hangup_needed)
1945                 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1946
1947         atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1948
1949         if (ocfs2_mount_local(osb))
1950                 snprintf(nodestr, sizeof(nodestr), "local");
1951         else
1952                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1953
1954         printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1955                osb->dev_str, nodestr);
1956
1957         ocfs2_delete_osb(osb);
1958         kfree(osb);
1959         sb->s_dev = 0;
1960         sb->s_fs_info = NULL;
1961 }
1962
1963 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1964                                 unsigned uuid_bytes)
1965 {
1966         int i, ret;
1967         char *ptr;
1968
1969         BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1970
1971         osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1972         if (osb->uuid_str == NULL)
1973                 return -ENOMEM;
1974
1975         for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1976                 /* print with null */
1977                 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1978                 if (ret != 2) /* drop super cleans up */
1979                         return -EINVAL;
1980                 /* then only advance past the last char */
1981                 ptr += 2;
1982         }
1983
1984         return 0;
1985 }
1986
1987 /* Make sure entire volume is addressable by our journal.  Requires
1988    osb_clusters_at_boot to be valid and for the journal to have been
1989    initialized by ocfs2_journal_init(). */
1990 static int ocfs2_journal_addressable(struct ocfs2_super *osb)
1991 {
1992         int status = 0;
1993         u64 max_block =
1994                 ocfs2_clusters_to_blocks(osb->sb,
1995                                          osb->osb_clusters_at_boot) - 1;
1996
1997         /* 32-bit block number is always OK. */
1998         if (max_block <= (u32)~0ULL)
1999                 goto out;
2000
2001         /* Volume is "huge", so see if our journal is new enough to
2002            support it. */
2003         if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2004                                        OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2005               jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2006                                                JBD2_FEATURE_INCOMPAT_64BIT))) {
2007                 mlog(ML_ERROR, "The journal cannot address the entire volume. "
2008                      "Enable the 'block64' journal option with tunefs.ocfs2");
2009                 status = -EFBIG;
2010                 goto out;
2011         }
2012
2013  out:
2014         return status;
2015 }
2016
2017 static int ocfs2_initialize_super(struct super_block *sb,
2018                                   struct buffer_head *bh,
2019                                   int sector_size,
2020                                   struct ocfs2_blockcheck_stats *stats)
2021 {
2022         int status;
2023         int i, cbits, bbits;
2024         struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2025         struct inode *inode = NULL;
2026         struct ocfs2_super *osb;
2027         u64 total_blocks;
2028
2029         osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2030         if (!osb) {
2031                 status = -ENOMEM;
2032                 mlog_errno(status);
2033                 goto out;
2034         }
2035
2036         sb->s_fs_info = osb;
2037         sb->s_op = &ocfs2_sops;
2038         sb->s_d_op = &ocfs2_dentry_ops;
2039         sb->s_export_op = &ocfs2_export_ops;
2040         sb->s_qcop = &dquot_quotactl_sysfile_ops;
2041         sb->dq_op = &ocfs2_quota_operations;
2042         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
2043         sb->s_xattr = ocfs2_xattr_handlers;
2044         sb->s_time_gran = 1;
2045         sb->s_flags |= SB_NOATIME;
2046         /* this is needed to support O_LARGEFILE */
2047         cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2048         bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2049         sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2050         memcpy(&sb->s_uuid, di->id2.i_super.s_uuid,
2051                sizeof(di->id2.i_super.s_uuid));
2052
2053         osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2054
2055         for (i = 0; i < 3; i++)
2056                 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2057         osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2058
2059         osb->sb = sb;
2060         osb->s_sectsize_bits = blksize_bits(sector_size);
2061         BUG_ON(!osb->s_sectsize_bits);
2062
2063         spin_lock_init(&osb->dc_task_lock);
2064         init_waitqueue_head(&osb->dc_event);
2065         osb->dc_work_sequence = 0;
2066         osb->dc_wake_sequence = 0;
2067         INIT_LIST_HEAD(&osb->blocked_lock_list);
2068         osb->blocked_lock_count = 0;
2069         spin_lock_init(&osb->osb_lock);
2070         spin_lock_init(&osb->osb_xattr_lock);
2071         ocfs2_init_steal_slots(osb);
2072
2073         mutex_init(&osb->system_file_mutex);
2074
2075         atomic_set(&osb->alloc_stats.moves, 0);
2076         atomic_set(&osb->alloc_stats.local_data, 0);
2077         atomic_set(&osb->alloc_stats.bitmap_data, 0);
2078         atomic_set(&osb->alloc_stats.bg_allocs, 0);
2079         atomic_set(&osb->alloc_stats.bg_extends, 0);
2080
2081         /* Copy the blockcheck stats from the superblock probe */
2082         osb->osb_ecc_stats = *stats;
2083
2084         ocfs2_init_node_maps(osb);
2085
2086         snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2087                  MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2088
2089         osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2090         if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2091                 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2092                      osb->max_slots);
2093                 status = -EINVAL;
2094                 goto out;
2095         }
2096
2097         ocfs2_orphan_scan_init(osb);
2098
2099         status = ocfs2_recovery_init(osb);
2100         if (status) {
2101                 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2102                 mlog_errno(status);
2103                 goto out;
2104         }
2105
2106         init_waitqueue_head(&osb->checkpoint_event);
2107
2108         osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2109
2110         osb->slot_num = OCFS2_INVALID_SLOT;
2111
2112         osb->s_xattr_inline_size = le16_to_cpu(
2113                                         di->id2.i_super.s_xattr_inline_size);
2114
2115         osb->local_alloc_state = OCFS2_LA_UNUSED;
2116         osb->local_alloc_bh = NULL;
2117         INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2118
2119         init_waitqueue_head(&osb->osb_mount_event);
2120
2121         ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2122
2123         osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2124         if (!osb->vol_label) {
2125                 mlog(ML_ERROR, "unable to alloc vol label\n");
2126                 status = -ENOMEM;
2127                 goto out_recovery_map;
2128         }
2129
2130         osb->slot_recovery_generations =
2131                 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2132                         GFP_KERNEL);
2133         if (!osb->slot_recovery_generations) {
2134                 status = -ENOMEM;
2135                 mlog_errno(status);
2136                 goto out_vol_label;
2137         }
2138
2139         init_waitqueue_head(&osb->osb_wipe_event);
2140         osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2141                                         sizeof(*osb->osb_orphan_wipes),
2142                                         GFP_KERNEL);
2143         if (!osb->osb_orphan_wipes) {
2144                 status = -ENOMEM;
2145                 mlog_errno(status);
2146                 goto out_slot_recovery_gen;
2147         }
2148
2149         osb->osb_rf_lock_tree = RB_ROOT;
2150
2151         osb->s_feature_compat =
2152                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2153         osb->s_feature_ro_compat =
2154                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2155         osb->s_feature_incompat =
2156                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2157
2158         if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2159                 mlog(ML_ERROR, "couldn't mount because of unsupported "
2160                      "optional features (%x).\n", i);
2161                 status = -EINVAL;
2162                 goto out_orphan_wipes;
2163         }
2164         if (!sb_rdonly(osb->sb) && (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2165                 mlog(ML_ERROR, "couldn't mount RDWR because of "
2166                      "unsupported optional features (%x).\n", i);
2167                 status = -EINVAL;
2168                 goto out_orphan_wipes;
2169         }
2170
2171         if (ocfs2_clusterinfo_valid(osb)) {
2172                 /*
2173                  * ci_stack and ci_cluster in ocfs2_cluster_info may not be null
2174                  * terminated, so make sure no overflow happens here by using
2175                  * memcpy. Destination strings will always be null terminated
2176                  * because osb is allocated using kzalloc.
2177                  */
2178                 osb->osb_stackflags =
2179                         OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2180                 memcpy(osb->osb_cluster_stack,
2181                        OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2182                        OCFS2_STACK_LABEL_LEN);
2183                 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2184                         mlog(ML_ERROR,
2185                              "couldn't mount because of an invalid "
2186                              "cluster stack label (%s) \n",
2187                              osb->osb_cluster_stack);
2188                         status = -EINVAL;
2189                         goto out_orphan_wipes;
2190                 }
2191                 memcpy(osb->osb_cluster_name,
2192                         OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster,
2193                         OCFS2_CLUSTER_NAME_LEN);
2194         } else {
2195                 /* The empty string is identical with classic tools that
2196                  * don't know about s_cluster_info. */
2197                 osb->osb_cluster_stack[0] = '\0';
2198         }
2199
2200         get_random_bytes(&osb->s_next_generation, sizeof(u32));
2201
2202         /*
2203          * FIXME
2204          * This should be done in ocfs2_journal_init(), but any inode
2205          * writes back operation will cause the filesystem to crash.
2206          */
2207         status = ocfs2_journal_alloc(osb);
2208         if (status < 0)
2209                 goto out_orphan_wipes;
2210
2211         INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs);
2212         init_llist_head(&osb->dquot_drop_list);
2213
2214         /* get some pseudo constants for clustersize bits */
2215         osb->s_clustersize_bits =
2216                 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2217         osb->s_clustersize = 1 << osb->s_clustersize_bits;
2218
2219         if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2220             osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2221                 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2222                      osb->s_clustersize);
2223                 status = -EINVAL;
2224                 goto out_journal;
2225         }
2226
2227         total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2228                                                 le32_to_cpu(di->i_clusters));
2229
2230         status = generic_check_addressable(osb->sb->s_blocksize_bits,
2231                                            total_blocks);
2232         if (status) {
2233                 mlog(ML_ERROR, "Volume too large "
2234                      "to mount safely on this system");
2235                 status = -EFBIG;
2236                 goto out_journal;
2237         }
2238
2239         if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2240                                  sizeof(di->id2.i_super.s_uuid))) {
2241                 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2242                 status = -ENOMEM;
2243                 goto out_journal;
2244         }
2245
2246         strlcpy(osb->vol_label, di->id2.i_super.s_label,
2247                 OCFS2_MAX_VOL_LABEL_LEN);
2248         osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2249         osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2250         osb->first_cluster_group_blkno =
2251                 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2252         osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2253         osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2254         trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str,
2255                                      (unsigned long long)osb->root_blkno,
2256                                      (unsigned long long)osb->system_dir_blkno,
2257                                      osb->s_clustersize_bits);
2258
2259         osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2260         if (!osb->osb_dlm_debug) {
2261                 status = -ENOMEM;
2262                 mlog_errno(status);
2263                 goto out_uuid_str;
2264         }
2265
2266         atomic_set(&osb->vol_state, VOLUME_INIT);
2267
2268         /* load root, system_dir, and all global system inodes */
2269         status = ocfs2_init_global_system_inodes(osb);
2270         if (status < 0) {
2271                 mlog_errno(status);
2272                 goto out_dlm_out;
2273         }
2274
2275         /*
2276          * global bitmap
2277          */
2278         inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2279                                             OCFS2_INVALID_SLOT);
2280         if (!inode) {
2281                 status = -EINVAL;
2282                 mlog_errno(status);
2283                 goto out_system_inodes;
2284         }
2285
2286         osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2287         osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2288         iput(inode);
2289
2290         osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2291                                  osb->s_feature_incompat) * 8;
2292
2293         status = ocfs2_init_slot_info(osb);
2294         if (status < 0) {
2295                 mlog_errno(status);
2296                 goto out_system_inodes;
2297         }
2298
2299         osb->ocfs2_wq = alloc_ordered_workqueue("ocfs2_wq", WQ_MEM_RECLAIM);
2300         if (!osb->ocfs2_wq) {
2301                 status = -ENOMEM;
2302                 mlog_errno(status);
2303                 goto out_slot_info;
2304         }
2305
2306         return status;
2307
2308 out_slot_info:
2309         ocfs2_free_slot_info(osb);
2310 out_system_inodes:
2311         ocfs2_release_system_inodes(osb);
2312 out_dlm_out:
2313         ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2314 out_uuid_str:
2315         kfree(osb->uuid_str);
2316 out_journal:
2317         kfree(osb->journal);
2318 out_orphan_wipes:
2319         kfree(osb->osb_orphan_wipes);
2320 out_slot_recovery_gen:
2321         kfree(osb->slot_recovery_generations);
2322 out_vol_label:
2323         kfree(osb->vol_label);
2324 out_recovery_map:
2325         kfree(osb->recovery_map);
2326 out:
2327         kfree(osb);
2328         sb->s_fs_info = NULL;
2329         return status;
2330 }
2331
2332 /*
2333  * will return: -EAGAIN if it is ok to keep searching for superblocks
2334  *              -EINVAL if there is a bad superblock
2335  *              0 on success
2336  */
2337 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2338                                struct buffer_head *bh,
2339                                u32 blksz,
2340                                struct ocfs2_blockcheck_stats *stats)
2341 {
2342         int status = -EAGAIN;
2343
2344         if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2345                    strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2346                 /* We have to do a raw check of the feature here */
2347                 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2348                     OCFS2_FEATURE_INCOMPAT_META_ECC) {
2349                         status = ocfs2_block_check_validate(bh->b_data,
2350                                                             bh->b_size,
2351                                                             &di->i_check,
2352                                                             stats);
2353                         if (status)
2354                                 goto out;
2355                 }
2356                 status = -EINVAL;
2357                 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2358                         mlog(ML_ERROR, "found superblock with incorrect block "
2359                              "size: found %u, should be %u\n",
2360                              1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2361                                blksz);
2362                 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2363                            OCFS2_MAJOR_REV_LEVEL ||
2364                            le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2365                            OCFS2_MINOR_REV_LEVEL) {
2366                         mlog(ML_ERROR, "found superblock with bad version: "
2367                              "found %u.%u, should be %u.%u\n",
2368                              le16_to_cpu(di->id2.i_super.s_major_rev_level),
2369                              le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2370                              OCFS2_MAJOR_REV_LEVEL,
2371                              OCFS2_MINOR_REV_LEVEL);
2372                 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2373                         mlog(ML_ERROR, "bad block number on superblock: "
2374                              "found %llu, should be %llu\n",
2375                              (unsigned long long)le64_to_cpu(di->i_blkno),
2376                              (unsigned long long)bh->b_blocknr);
2377                 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2378                             le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2379                         mlog(ML_ERROR, "bad cluster size found: %u\n",
2380                              1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2381                 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2382                         mlog(ML_ERROR, "bad root_blkno: 0\n");
2383                 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2384                         mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2385                 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2386                         mlog(ML_ERROR,
2387                              "Superblock slots found greater than file system "
2388                              "maximum: found %u, max %u\n",
2389                              le16_to_cpu(di->id2.i_super.s_max_slots),
2390                              OCFS2_MAX_SLOTS);
2391                 } else {
2392                         /* found it! */
2393                         status = 0;
2394                 }
2395         }
2396
2397 out:
2398         if (status && status != -EAGAIN)
2399                 mlog_errno(status);
2400         return status;
2401 }
2402
2403 static int ocfs2_check_volume(struct ocfs2_super *osb)
2404 {
2405         int status;
2406         int dirty;
2407         int local;
2408         struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2409                                                   * recover
2410                                                   * ourselves. */
2411
2412         /* Init our journal object. */
2413         status = ocfs2_journal_init(osb, &dirty);
2414         if (status < 0) {
2415                 mlog(ML_ERROR, "Could not initialize journal!\n");
2416                 goto finally;
2417         }
2418
2419         /* Now that journal has been initialized, check to make sure
2420            entire volume is addressable. */
2421         status = ocfs2_journal_addressable(osb);
2422         if (status)
2423                 goto finally;
2424
2425         /* If the journal was unmounted cleanly then we don't want to
2426          * recover anything. Otherwise, journal_load will do that
2427          * dirty work for us :) */
2428         if (!dirty) {
2429                 status = ocfs2_journal_wipe(osb->journal, 0);
2430                 if (status < 0) {
2431                         mlog_errno(status);
2432                         goto finally;
2433                 }
2434         } else {
2435                 printk(KERN_NOTICE "ocfs2: File system on device (%s) was not "
2436                        "unmounted cleanly, recovering it.\n", osb->dev_str);
2437         }
2438
2439         local = ocfs2_mount_local(osb);
2440
2441         /* will play back anything left in the journal. */
2442         status = ocfs2_journal_load(osb->journal, local, dirty);
2443         if (status < 0) {
2444                 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2445                 goto finally;
2446         }
2447
2448         if (osb->s_mount_opt & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
2449                 jbd2_journal_set_features(osb->journal->j_journal,
2450                                 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2451                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2452         else
2453                 jbd2_journal_clear_features(osb->journal->j_journal,
2454                                 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2455                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2456
2457         if (dirty) {
2458                 /* recover my local alloc if we didn't unmount cleanly. */
2459                 status = ocfs2_begin_local_alloc_recovery(osb,
2460                                                           osb->slot_num,
2461                                                           &local_alloc);
2462                 if (status < 0) {
2463                         mlog_errno(status);
2464                         goto finally;
2465                 }
2466                 /* we complete the recovery process after we've marked
2467                  * ourselves as mounted. */
2468         }
2469
2470         status = ocfs2_load_local_alloc(osb);
2471         if (status < 0) {
2472                 mlog_errno(status);
2473                 goto finally;
2474         }
2475
2476         if (dirty) {
2477                 /* Recovery will be completed after we've mounted the
2478                  * rest of the volume. */
2479                 osb->local_alloc_copy = local_alloc;
2480                 local_alloc = NULL;
2481         }
2482
2483         /* go through each journal, trylock it and if you get the
2484          * lock, and it's marked as dirty, set the bit in the recover
2485          * map and launch a recovery thread for it. */
2486         status = ocfs2_mark_dead_nodes(osb);
2487         if (status < 0) {
2488                 mlog_errno(status);
2489                 goto finally;
2490         }
2491
2492         status = ocfs2_compute_replay_slots(osb);
2493         if (status < 0)
2494                 mlog_errno(status);
2495
2496 finally:
2497         kfree(local_alloc);
2498
2499         if (status)
2500                 mlog_errno(status);
2501         return status;
2502 }
2503
2504 /*
2505  * The routine gets called from dismount or close whenever a dismount on
2506  * volume is requested and the osb open count becomes 1.
2507  * It will remove the osb from the global list and also free up all the
2508  * initialized resources and fileobject.
2509  */
2510 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2511 {
2512         /* This function assumes that the caller has the main osb resource */
2513
2514         /* ocfs2_initializer_super have already created this workqueue */
2515         if (osb->ocfs2_wq)
2516                 destroy_workqueue(osb->ocfs2_wq);
2517
2518         ocfs2_free_slot_info(osb);
2519
2520         kfree(osb->osb_orphan_wipes);
2521         kfree(osb->slot_recovery_generations);
2522         /* FIXME
2523          * This belongs in journal shutdown, but because we have to
2524          * allocate osb->journal at the middle of ocfs2_initialize_super(),
2525          * we free it here.
2526          */
2527         kfree(osb->journal);
2528         kfree(osb->local_alloc_copy);
2529         kfree(osb->uuid_str);
2530         kfree(osb->vol_label);
2531         ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2532         memset(osb, 0, sizeof(struct ocfs2_super));
2533 }
2534
2535 /* Depending on the mount option passed, perform one of the following:
2536  * Put OCFS2 into a readonly state (default)
2537  * Return EIO so that only the process errs
2538  * Fix the error as if fsck.ocfs2 -y
2539  * panic
2540  */
2541 static int ocfs2_handle_error(struct super_block *sb)
2542 {
2543         struct ocfs2_super *osb = OCFS2_SB(sb);
2544         int rv = 0;
2545
2546         ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2547         pr_crit("On-disk corruption discovered. "
2548                 "Please run fsck.ocfs2 once the filesystem is unmounted.\n");
2549
2550         if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC) {
2551                 panic("OCFS2: (device %s): panic forced after error\n",
2552                       sb->s_id);
2553         } else if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_CONT) {
2554                 pr_crit("OCFS2: Returning error to the calling process.\n");
2555                 rv = -EIO;
2556         } else { /* default option */
2557                 rv = -EROFS;
2558                 if (sb_rdonly(sb) && (ocfs2_is_soft_readonly(osb) || ocfs2_is_hard_readonly(osb)))
2559                         return rv;
2560
2561                 pr_crit("OCFS2: File system is now read-only.\n");
2562                 sb->s_flags |= SB_RDONLY;
2563                 ocfs2_set_ro_flag(osb, 0);
2564         }
2565
2566         return rv;
2567 }
2568
2569 int __ocfs2_error(struct super_block *sb, const char *function,
2570                   const char *fmt, ...)
2571 {
2572         struct va_format vaf;
2573         va_list args;
2574
2575         va_start(args, fmt);
2576         vaf.fmt = fmt;
2577         vaf.va = &args;
2578
2579         /* Not using mlog here because we want to show the actual
2580          * function the error came from. */
2581         printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %pV",
2582                sb->s_id, function, &vaf);
2583
2584         va_end(args);
2585
2586         return ocfs2_handle_error(sb);
2587 }
2588
2589 /* Handle critical errors. This is intentionally more drastic than
2590  * ocfs2_handle_error, so we only use for things like journal errors,
2591  * etc. */
2592 void __ocfs2_abort(struct super_block *sb, const char *function,
2593                    const char *fmt, ...)
2594 {
2595         struct va_format vaf;
2596         va_list args;
2597
2598         va_start(args, fmt);
2599
2600         vaf.fmt = fmt;
2601         vaf.va = &args;
2602
2603         printk(KERN_CRIT "OCFS2: abort (device %s): %s: %pV",
2604                sb->s_id, function, &vaf);
2605
2606         va_end(args);
2607
2608         /* We don't have the cluster support yet to go straight to
2609          * hard readonly in here. Until then, we want to keep
2610          * ocfs2_abort() so that we can at least mark critical
2611          * errors.
2612          *
2613          * TODO: This should abort the journal and alert other nodes
2614          * that our slot needs recovery. */
2615
2616         /* Force a panic(). This stinks, but it's better than letting
2617          * things continue without having a proper hard readonly
2618          * here. */
2619         if (!ocfs2_mount_local(OCFS2_SB(sb)))
2620                 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2621         ocfs2_handle_error(sb);
2622 }
2623
2624 /*
2625  * Void signal blockers, because in-kernel sigprocmask() only fails
2626  * when SIG_* is wrong.
2627  */
2628 void ocfs2_block_signals(sigset_t *oldset)
2629 {
2630         int rc;
2631         sigset_t blocked;
2632
2633         sigfillset(&blocked);
2634         rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2635         BUG_ON(rc);
2636 }
2637
2638 void ocfs2_unblock_signals(sigset_t *oldset)
2639 {
2640         int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2641         BUG_ON(rc);
2642 }
2643
2644 module_init(ocfs2_init);
2645 module_exit(ocfs2_exit);