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/cleancache.h>
  29 #include <linux/signal.h>
  30
  31 #define CREATE_TRACE_POINTS
  32 #include "ocfs2_trace.h"
  33
  34 #include <cluster/masklog.h>
  35
  36 #include "ocfs2.h"
  37
  38 /* this should be the only file to include a version 1 header */
  39 #include "ocfs1_fs_compat.h"
  40
  41 #include "alloc.h"
  42 #include "aops.h"
  43 #include "blockcheck.h"
  44 #include "dlmglue.h"
  45 #include "export.h"
  46 #include "extent_map.h"
  47 #include "heartbeat.h"
  48 #include "inode.h"
  49 #include "journal.h"
  50 #include "localalloc.h"
  51 #include "namei.h"
  52 #include "slot_map.h"
  53 #include "super.h"
  54 #include "sysfile.h"
  55 #include "uptodate.h"
  56 #include "xattr.h"
  57 #include "quota.h"
  58 #include "refcounttree.h"
  59 #include "suballoc.h"
  60
  61 #include "buffer_head_io.h"
  62 #include "filecheck.h"
  63
  64 static struct kmem_cache *ocfs2_inode_cachep;
  65 struct kmem_cache *ocfs2_dquot_cachep;
  66 struct kmem_cache *ocfs2_qf_chunk_cachep;
  67
  68 static struct dentry *ocfs2_debugfs_root;
  69
  70 MODULE_AUTHOR("Oracle");
  71 MODULE_LICENSE("GPL");
  72 MODULE_DESCRIPTION("OCFS2 cluster file system");
  73
  74 struct mount_options
  75 {
  76         unsigned long   commit_interval;
  77         unsigned long   mount_opt;
  78         unsigned int    atime_quantum;
  79         unsigned short  slot;
  80         int             localalloc_opt;
  81         unsigned int    resv_level;
  82         int             dir_resv_level;
  83         char            cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
  84 };
  85
  86 static int ocfs2_parse_options(struct super_block *sb, char *options,
  87                                struct mount_options *mopt,
  88                                int is_remount);
  89 static int ocfs2_check_set_options(struct super_block *sb,
  90                                    struct mount_options *options);
  91 static int ocfs2_show_options(struct seq_file *s, struct dentry *root);
  92 static void ocfs2_put_super(struct super_block *sb);
  93 static int ocfs2_mount_volume(struct super_block *sb);
  94 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
  95 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
  96 static int ocfs2_initialize_mem_caches(void);
  97 static void ocfs2_free_mem_caches(void);
  98 static void ocfs2_delete_osb(struct ocfs2_super *osb);
  99
 100 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
 101
 102 static int ocfs2_sync_fs(struct super_block *sb, int wait);
 103
 104 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
 105 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
 106 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
 107 static int ocfs2_check_volume(struct ocfs2_super *osb);
 108 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
 109                                struct buffer_head *bh,
 110                                u32 sectsize,
 111                                struct ocfs2_blockcheck_stats *stats);
 112 static int ocfs2_initialize_super(struct super_block *sb,
 113                                   struct buffer_head *bh,
 114                                   int sector_size,
 115                                   struct ocfs2_blockcheck_stats *stats);
 116 static int ocfs2_get_sector(struct super_block *sb,
 117                             struct buffer_head **bh,
 118                             int block,
 119                             int sect_size);
 120 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
 121 static void ocfs2_free_inode(struct inode *inode);
 122 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
 123 static int ocfs2_enable_quotas(struct ocfs2_super *osb);
 124 static void ocfs2_disable_quotas(struct ocfs2_super *osb);
 125
 126 static struct dquot **ocfs2_get_dquots(struct inode *inode)
 127 {
 128         return OCFS2_I(inode)->i_dquot;
 129 }
 130
 131 static const struct super_operations ocfs2_sops = {
 132         .statfs         = ocfs2_statfs,
 133         .alloc_inode    = ocfs2_alloc_inode,
 134         .free_inode     = ocfs2_free_inode,
 135         .drop_inode     = ocfs2_drop_inode,
 136         .evict_inode    = ocfs2_evict_inode,
 137         .sync_fs        = ocfs2_sync_fs,
 138         .put_super      = ocfs2_put_super,
 139         .remount_fs     = ocfs2_remount,
 140         .show_options   = ocfs2_show_options,
 141         .quota_read     = ocfs2_quota_read,
 142         .quota_write    = ocfs2_quota_write,
 143         .get_dquots     = ocfs2_get_dquots,
 144 };
 145
 146 enum {
 147         Opt_barrier,
 148         Opt_err_panic,
 149         Opt_err_ro,
 150         Opt_intr,
 151         Opt_nointr,
 152         Opt_hb_none,
 153         Opt_hb_local,
 154         Opt_hb_global,
 155         Opt_data_ordered,
 156         Opt_data_writeback,
 157         Opt_atime_quantum,
 158         Opt_slot,
 159         Opt_commit,
 160         Opt_localalloc,
 161         Opt_localflocks,
 162         Opt_stack,
 163         Opt_user_xattr,
 164         Opt_nouser_xattr,
 165         Opt_inode64,
 166         Opt_acl,
 167         Opt_noacl,
 168         Opt_usrquota,
 169         Opt_grpquota,
 170         Opt_coherency_buffered,
 171         Opt_coherency_full,
 172         Opt_resv_level,
 173         Opt_dir_resv_level,
 174         Opt_journal_async_commit,
 175         Opt_err_cont,
 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_err, NULL}
 210 };
 211
 212 #ifdef CONFIG_DEBUG_FS
 213 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
 214 {
 215         struct ocfs2_cluster_connection *cconn = osb->cconn;
 216         struct ocfs2_recovery_map *rm = osb->recovery_map;
 217         struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
 218         int i, out = 0;
 219         unsigned long flags;
 220
 221         out += scnprintf(buf + out, len - out,
 222                         "%10s => Id: %-s  Uuid: %-s  Gen: 0x%X  Label: %-s\n",
 223                         "Device", osb->dev_str, osb->uuid_str,
 224                         osb->fs_generation, osb->vol_label);
 225
 226         out += scnprintf(buf + out, len - out,
 227                         "%10s => State: %d  Flags: 0x%lX\n", "Volume",
 228                         atomic_read(&osb->vol_state), osb->osb_flags);
 229
 230         out += scnprintf(buf + out, len - out,
 231                         "%10s => Block: %lu  Cluster: %d\n", "Sizes",
 232                         osb->sb->s_blocksize, osb->s_clustersize);
 233
 234         out += scnprintf(buf + out, len - out,
 235                         "%10s => Compat: 0x%X  Incompat: 0x%X  "
 236                         "ROcompat: 0x%X\n",
 237                         "Features", osb->s_feature_compat,
 238                         osb->s_feature_incompat, osb->s_feature_ro_compat);
 239
 240         out += scnprintf(buf + out, len - out,
 241                         "%10s => Opts: 0x%lX  AtimeQuanta: %u\n", "Mount",
 242                         osb->s_mount_opt, osb->s_atime_quantum);
 243
 244         if (cconn) {
 245                 out += scnprintf(buf + out, len - out,
 246                                 "%10s => Stack: %s  Name: %*s  "
 247                                 "Version: %d.%d\n", "Cluster",
 248                                 (*osb->osb_cluster_stack == '\0' ?
 249                                  "o2cb" : osb->osb_cluster_stack),
 250                                 cconn->cc_namelen, cconn->cc_name,
 251                                 cconn->cc_version.pv_major,
 252                                 cconn->cc_version.pv_minor);
 253         }
 254
 255         spin_lock_irqsave(&osb->dc_task_lock, flags);
 256         out += scnprintf(buf + out, len - out,
 257                         "%10s => Pid: %d  Count: %lu  WakeSeq: %lu  "
 258                         "WorkSeq: %lu\n", "DownCnvt",
 259                         (osb->dc_task ?  task_pid_nr(osb->dc_task) : -1),
 260                         osb->blocked_lock_count, osb->dc_wake_sequence,
 261                         osb->dc_work_sequence);
 262         spin_unlock_irqrestore(&osb->dc_task_lock, flags);
 263
 264         spin_lock(&osb->osb_lock);
 265         out += scnprintf(buf + out, len - out, "%10s => Pid: %d  Nodes:",
 266                         "Recovery",
 267                         (osb->recovery_thread_task ?
 268                          task_pid_nr(osb->recovery_thread_task) : -1));
 269         if (rm->rm_used == 0)
 270                 out += scnprintf(buf + out, len - out, " None\n");
 271         else {
 272                 for (i = 0; i < rm->rm_used; i++)
 273                         out += scnprintf(buf + out, len - out, " %d",
 274                                         rm->rm_entries[i]);
 275                 out += scnprintf(buf + out, len - out, "\n");
 276         }
 277         spin_unlock(&osb->osb_lock);
 278
 279         out += scnprintf(buf + out, len - out,
 280                         "%10s => Pid: %d  Interval: %lu\n", "Commit",
 281                         (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
 282                         osb->osb_commit_interval);
 283
 284         out += scnprintf(buf + out, len - out,
 285                         "%10s => State: %d  TxnId: %lu  NumTxns: %d\n",
 286                         "Journal", osb->journal->j_state,
 287                         osb->journal->j_trans_id,
 288                         atomic_read(&osb->journal->j_num_trans));
 289
 290         out += scnprintf(buf + out, len - out,
 291                         "%10s => GlobalAllocs: %d  LocalAllocs: %d  "
 292                         "SubAllocs: %d  LAWinMoves: %d  SAExtends: %d\n",
 293                         "Stats",
 294                         atomic_read(&osb->alloc_stats.bitmap_data),
 295                         atomic_read(&osb->alloc_stats.local_data),
 296                         atomic_read(&osb->alloc_stats.bg_allocs),
 297                         atomic_read(&osb->alloc_stats.moves),
 298                         atomic_read(&osb->alloc_stats.bg_extends));
 299
 300         out += scnprintf(buf + out, len - out,
 301                         "%10s => State: %u  Descriptor: %llu  Size: %u bits  "
 302                         "Default: %u bits\n",
 303                         "LocalAlloc", osb->local_alloc_state,
 304                         (unsigned long long)osb->la_last_gd,
 305                         osb->local_alloc_bits, osb->local_alloc_default_bits);
 306
 307         spin_lock(&osb->osb_lock);
 308         out += scnprintf(buf + out, len - out,
 309                         "%10s => InodeSlot: %d  StolenInodes: %d, "
 310                         "MetaSlot: %d  StolenMeta: %d\n", "Steal",
 311                         osb->s_inode_steal_slot,
 312                         atomic_read(&osb->s_num_inodes_stolen),
 313                         osb->s_meta_steal_slot,
 314                         atomic_read(&osb->s_num_meta_stolen));
 315         spin_unlock(&osb->osb_lock);
 316
 317         out += scnprintf(buf + out, len - out, "OrphanScan => ");
 318         out += scnprintf(buf + out, len - out, "Local: %u  Global: %u ",
 319                         os->os_count, os->os_seqno);
 320         out += scnprintf(buf + out, len - out, " Last Scan: ");
 321         if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
 322                 out += scnprintf(buf + out, len - out, "Disabled\n");
 323         else
 324                 out += scnprintf(buf + out, len - out, "%lu seconds ago\n",
 325                                 (unsigned long)(ktime_get_seconds() - os->os_scantime));
 326
 327         out += scnprintf(buf + out, len - out, "%10s => %3s  %10s\n",
 328                         "Slots", "Num", "RecoGen");
 329         for (i = 0; i < osb->max_slots; ++i) {
 330                 out += scnprintf(buf + out, len - out,
 331                                 "%10s  %c %3d  %10d\n",
 332                                 " ",
 333                                 (i == osb->slot_num ? '*' : ' '),
 334                                 i, osb->slot_recovery_generations[i]);
 335         }
 336
 337         return out;
 338 }
 339
 340 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 341 {
 342         struct ocfs2_super *osb = inode->i_private;
 343         char *buf = NULL;
 344
 345         buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
 346         if (!buf)
 347                 goto bail;
 348
 349         i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
 350
 351         file->private_data = buf;
 352
 353         return 0;
 354 bail:
 355         return -ENOMEM;
 356 }
 357
 358 static int ocfs2_debug_release(struct inode *inode, struct file *file)
 359 {
 360         kfree(file->private_data);
 361         return 0;
 362 }
 363
 364 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 365                                 size_t nbytes, loff_t *ppos)
 366 {
 367         return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
 368                                        i_size_read(file->f_mapping->host));
 369 }
 370 #else
 371 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 372 {
 373         return 0;
 374 }
 375 static int ocfs2_debug_release(struct inode *inode, struct file *file)
 376 {
 377         return 0;
 378 }
 379 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 380                                 size_t nbytes, loff_t *ppos)
 381 {
 382         return 0;
 383 }
 384 #endif  /* CONFIG_DEBUG_FS */
 385
 386 static const struct file_operations ocfs2_osb_debug_fops = {
 387         .open =         ocfs2_osb_debug_open,
 388         .release =      ocfs2_debug_release,
 389         .read =         ocfs2_debug_read,
 390         .llseek =       generic_file_llseek,
 391 };
 392
 393 static int ocfs2_sync_fs(struct super_block *sb, int wait)
 394 {
 395         int status;
 396         tid_t target;
 397         struct ocfs2_super *osb = OCFS2_SB(sb);
 398
 399         if (ocfs2_is_hard_readonly(osb))
 400                 return -EROFS;
 401
 402         if (wait) {
 403                 status = ocfs2_flush_truncate_log(osb);
 404                 if (status < 0)
 405                         mlog_errno(status);
 406         } else {
 407                 ocfs2_schedule_truncate_log_flush(osb, 0);
 408         }
 409
 410         if (jbd2_journal_start_commit(osb->journal->j_journal,
 411                                       &target)) {
 412                 if (wait)
 413                         jbd2_log_wait_commit(osb->journal->j_journal,
 414                                              target);
 415         }
 416         return 0;
 417 }
 418
 419 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
 420 {
 421         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
 422             && (ino == USER_QUOTA_SYSTEM_INODE
 423                 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
 424                 return 0;
 425         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
 426             && (ino == GROUP_QUOTA_SYSTEM_INODE
 427                 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
 428                 return 0;
 429         return 1;
 430 }
 431
 432 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
 433 {
 434         struct inode *new = NULL;
 435         int status = 0;
 436         int i;
 437
 438         new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 439         if (IS_ERR(new)) {
 440                 status = PTR_ERR(new);
 441                 mlog_errno(status);
 442                 goto bail;
 443         }
 444         osb->root_inode = new;
 445
 446         new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 447         if (IS_ERR(new)) {
 448                 status = PTR_ERR(new);
 449                 mlog_errno(status);
 450                 goto bail;
 451         }
 452         osb->sys_root_inode = new;
 453
 454         for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
 455              i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
 456                 if (!ocfs2_need_system_inode(osb, i))
 457                         continue;
 458                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 459                 if (!new) {
 460                         ocfs2_release_system_inodes(osb);
 461                         status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
 462                         mlog_errno(status);
 463                         mlog(ML_ERROR, "Unable to load system inode %d, "
 464                              "possibly corrupt fs?", i);
 465                         goto bail;
 466                 }
 467                 // the array now has one ref, so drop this one
 468                 iput(new);
 469         }
 470
 471 bail:
 472         if (status)
 473                 mlog_errno(status);
 474         return status;
 475 }
 476
 477 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
 478 {
 479         struct inode *new = NULL;
 480         int status = 0;
 481         int i;
 482
 483         for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
 484              i < NUM_SYSTEM_INODES;
 485              i++) {
 486                 if (!ocfs2_need_system_inode(osb, i))
 487                         continue;
 488                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 489                 if (!new) {
 490                         ocfs2_release_system_inodes(osb);
 491                         status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
 492                         mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
 493                              status, i, osb->slot_num);
 494                         goto bail;
 495                 }
 496                 /* the array now has one ref, so drop this one */
 497                 iput(new);
 498         }
 499
 500 bail:
 501         if (status)
 502                 mlog_errno(status);
 503         return status;
 504 }
 505
 506 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
 507 {
 508         int i;
 509         struct inode *inode;
 510
 511         for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
 512                 inode = osb->global_system_inodes[i];
 513                 if (inode) {
 514                         iput(inode);
 515                         osb->global_system_inodes[i] = NULL;
 516                 }
 517         }
 518
 519         inode = osb->sys_root_inode;
 520         if (inode) {
 521                 iput(inode);
 522                 osb->sys_root_inode = NULL;
 523         }
 524
 525         inode = osb->root_inode;
 526         if (inode) {
 527                 iput(inode);
 528                 osb->root_inode = NULL;
 529         }
 530
 531         if (!osb->local_system_inodes)
 532                 return;
 533
 534         for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
 535                 if (osb->local_system_inodes[i]) {
 536                         iput(osb->local_system_inodes[i]);
 537                         osb->local_system_inodes[i] = NULL;
 538                 }
 539         }
 540
 541         kfree(osb->local_system_inodes);
 542         osb->local_system_inodes = NULL;
 543 }
 544
 545 /* We're allocating fs objects, use GFP_NOFS */
 546 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
 547 {
 548         struct ocfs2_inode_info *oi;
 549
 550         oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
 551         if (!oi)
 552                 return NULL;
 553
 554         oi->i_sync_tid = 0;
 555         oi->i_datasync_tid = 0;
 556         memset(&oi->i_dquot, 0, sizeof(oi->i_dquot));
 557
 558         jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
 559         return &oi->vfs_inode;
 560 }
 561
 562 static void ocfs2_free_inode(struct inode *inode)
 563 {
 564         kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
 565 }
 566
 567 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
 568                                                 unsigned int cbits)
 569 {
 570         unsigned int bytes = 1 << cbits;
 571         unsigned int trim = bytes;
 572         unsigned int bitshift = 32;
 573
 574         /*
 575          * i_size and all block offsets in ocfs2 are always 64 bits
 576          * wide. i_clusters is 32 bits, in cluster-sized units. So on
 577          * 64 bit platforms, cluster size will be the limiting factor.
 578          */
 579
 580 #if BITS_PER_LONG == 32
 581         BUILD_BUG_ON(sizeof(sector_t) != 8);
 582         /*
 583          * We might be limited by page cache size.
 584          */
 585         if (bytes > PAGE_SIZE) {
 586                 bytes = PAGE_SIZE;
 587                 trim = 1;
 588                 /*
 589                  * Shift by 31 here so that we don't get larger than
 590                  * MAX_LFS_FILESIZE
 591                  */
 592                 bitshift = 31;
 593         }
 594 #endif
 595
 596         /*
 597          * Trim by a whole cluster when we can actually approach the
 598          * on-disk limits. Otherwise we can overflow i_clusters when
 599          * an extent start is at the max offset.
 600          */
 601         return (((unsigned long long)bytes) << bitshift) - trim;
 602 }
 603
 604 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
 605 {
 606         int incompat_features;
 607         int ret = 0;
 608         struct mount_options parsed_options;
 609         struct ocfs2_super *osb = OCFS2_SB(sb);
 610         u32 tmp;
 611
 612         sync_filesystem(sb);
 613
 614         if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
 615             !ocfs2_check_set_options(sb, &parsed_options)) {
 616                 ret = -EINVAL;
 617                 goto out;
 618         }
 619
 620         tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
 621                 OCFS2_MOUNT_HB_NONE;
 622         if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
 623                 ret = -EINVAL;
 624                 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
 625                 goto out;
 626         }
 627
 628         if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
 629             (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
 630                 ret = -EINVAL;
 631                 mlog(ML_ERROR, "Cannot change data mode on remount\n");
 632                 goto out;
 633         }
 634
 635         /* Probably don't want this on remount; it might
 636          * mess with other nodes */
 637         if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
 638             (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
 639                 ret = -EINVAL;
 640                 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
 641                 goto out;
 642         }
 643
 644         /* We're going to/from readonly mode. */
 645         if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
 646                 /* Disable quota accounting before remounting RO */
 647                 if (*flags & SB_RDONLY) {
 648                         ret = ocfs2_susp_quotas(osb, 0);
 649                         if (ret < 0)
 650                                 goto out;
 651                 }
 652                 /* Lock here so the check of HARD_RO and the potential
 653                  * setting of SOFT_RO is atomic. */
 654                 spin_lock(&osb->osb_lock);
 655                 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
 656                         mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
 657                         ret = -EROFS;
 658                         goto unlock_osb;
 659                 }
 660
 661                 if (*flags & SB_RDONLY) {
 662                         sb->s_flags |= SB_RDONLY;
 663                         osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 664                 } else {
 665                         if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
 666                                 mlog(ML_ERROR, "Cannot remount RDWR "
 667                                      "filesystem due to previous errors.\n");
 668                                 ret = -EROFS;
 669                                 goto unlock_osb;
 670                         }
 671                         incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
 672                         if (incompat_features) {
 673                                 mlog(ML_ERROR, "Cannot remount RDWR because "
 674                                      "of unsupported optional features "
 675                                      "(%x).\n", incompat_features);
 676                                 ret = -EINVAL;
 677                                 goto unlock_osb;
 678                         }
 679                         sb->s_flags &= ~SB_RDONLY;
 680                         osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
 681                 }
 682                 trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags);
 683 unlock_osb:
 684                 spin_unlock(&osb->osb_lock);
 685                 /* Enable quota accounting after remounting RW */
 686                 if (!ret && !(*flags & SB_RDONLY)) {
 687                         if (sb_any_quota_suspended(sb))
 688                                 ret = ocfs2_susp_quotas(osb, 1);
 689                         else
 690                                 ret = ocfs2_enable_quotas(osb);
 691                         if (ret < 0) {
 692                                 /* Return back changes... */
 693                                 spin_lock(&osb->osb_lock);
 694                                 sb->s_flags |= SB_RDONLY;
 695                                 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 696                                 spin_unlock(&osb->osb_lock);
 697                                 goto out;
 698                         }
 699                 }
 700         }
 701
 702         if (!ret) {
 703                 /* Only save off the new mount options in case of a successful
 704                  * remount. */
 705                 osb->s_mount_opt = parsed_options.mount_opt;
 706                 osb->s_atime_quantum = parsed_options.atime_quantum;
 707                 osb->preferred_slot = parsed_options.slot;
 708                 if (parsed_options.commit_interval)
 709                         osb->osb_commit_interval = parsed_options.commit_interval;
 710
 711                 if (!ocfs2_is_hard_readonly(osb))
 712                         ocfs2_set_journal_params(osb);
 713
 714                 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
 715                         ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
 716                                                         SB_POSIXACL : 0);
 717         }
 718 out:
 719         return ret;
 720 }
 721
 722 static int ocfs2_sb_probe(struct super_block *sb,
 723                           struct buffer_head **bh,
 724                           int *sector_size,
 725                           struct ocfs2_blockcheck_stats *stats)
 726 {
 727         int status, tmpstat;
 728         struct ocfs1_vol_disk_hdr *hdr;
 729         struct ocfs2_dinode *di;
 730         int blksize;
 731
 732         *bh = NULL;
 733
 734         /* may be > 512 */
 735         *sector_size = bdev_logical_block_size(sb->s_bdev);
 736         if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
 737                 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
 738                      *sector_size, OCFS2_MAX_BLOCKSIZE);
 739                 status = -EINVAL;
 740                 goto bail;
 741         }
 742
 743         /* Can this really happen? */
 744         if (*sector_size < OCFS2_MIN_BLOCKSIZE)
 745                 *sector_size = OCFS2_MIN_BLOCKSIZE;
 746
 747         /* check block zero for old format */
 748         status = ocfs2_get_sector(sb, bh, 0, *sector_size);
 749         if (status < 0) {
 750                 mlog_errno(status);
 751                 goto bail;
 752         }
 753         hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
 754         if (hdr->major_version == OCFS1_MAJOR_VERSION) {
 755                 mlog(ML_ERROR, "incompatible version: %u.%u\n",
 756                      hdr->major_version, hdr->minor_version);
 757                 status = -EINVAL;
 758         }
 759         if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
 760                    strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
 761                 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
 762                      hdr->signature);
 763                 status = -EINVAL;
 764         }
 765         brelse(*bh);
 766         *bh = NULL;
 767         if (status < 0) {
 768                 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
 769                      "upgraded before mounting with ocfs v2\n");
 770                 goto bail;
 771         }
 772
 773         /*
 774          * Now check at magic offset for 512, 1024, 2048, 4096
 775          * blocksizes.  4096 is the maximum blocksize because it is
 776          * the minimum clustersize.
 777          */
 778         status = -EINVAL;
 779         for (blksize = *sector_size;
 780              blksize <= OCFS2_MAX_BLOCKSIZE;
 781              blksize <<= 1) {
 782                 tmpstat = ocfs2_get_sector(sb, bh,
 783                                            OCFS2_SUPER_BLOCK_BLKNO,
 784                                            blksize);
 785                 if (tmpstat < 0) {
 786                         status = tmpstat;
 787                         mlog_errno(status);
 788                         break;
 789                 }
 790                 di = (struct ocfs2_dinode *) (*bh)->b_data;
 791                 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
 792                 spin_lock_init(&stats->b_lock);
 793                 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
 794                 if (tmpstat < 0) {
 795                         brelse(*bh);
 796                         *bh = NULL;
 797                 }
 798                 if (tmpstat != -EAGAIN) {
 799                         status = tmpstat;
 800                         break;
 801                 }
 802         }
 803
 804 bail:
 805         return status;
 806 }
 807
 808 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
 809 {
 810         u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
 811
 812         if (osb->s_mount_opt & hb_enabled) {
 813                 if (ocfs2_mount_local(osb)) {
 814                         mlog(ML_ERROR, "Cannot heartbeat on a locally "
 815                              "mounted device.\n");
 816                         return -EINVAL;
 817                 }
 818                 if (ocfs2_userspace_stack(osb)) {
 819                         mlog(ML_ERROR, "Userspace stack expected, but "
 820                              "o2cb heartbeat arguments passed to mount\n");
 821                         return -EINVAL;
 822                 }
 823                 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
 824                      !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
 825                     ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
 826                      ocfs2_cluster_o2cb_global_heartbeat(osb))) {
 827                         mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
 828                         return -EINVAL;
 829                 }
 830         }
 831
 832         if (!(osb->s_mount_opt & hb_enabled)) {
 833                 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
 834                     !ocfs2_userspace_stack(osb)) {
 835                         mlog(ML_ERROR, "Heartbeat has to be started to mount "
 836                              "a read-write clustered device.\n");
 837                         return -EINVAL;
 838                 }
 839         }
 840
 841         return 0;
 842 }
 843
 844 /*
 845  * If we're using a userspace stack, mount should have passed
 846  * a name that matches the disk.  If not, mount should not
 847  * have passed a stack.
 848  */
 849 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
 850                                         struct mount_options *mopt)
 851 {
 852         if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
 853                 mlog(ML_ERROR,
 854                      "cluster stack passed to mount, but this filesystem "
 855                      "does not support it\n");
 856                 return -EINVAL;
 857         }
 858
 859         if (ocfs2_userspace_stack(osb) &&
 860             strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
 861                     OCFS2_STACK_LABEL_LEN)) {
 862                 mlog(ML_ERROR,
 863                      "cluster stack passed to mount (\"%s\") does not "
 864                      "match the filesystem (\"%s\")\n",
 865                      mopt->cluster_stack,
 866                      osb->osb_cluster_stack);
 867                 return -EINVAL;
 868         }
 869
 870         return 0;
 871 }
 872
 873 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
 874 {
 875         int type;
 876         struct super_block *sb = osb->sb;
 877         unsigned int feature[OCFS2_MAXQUOTAS] = {
 878                                         OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 879                                         OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 880         int status = 0;
 881
 882         for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
 883                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 884                         continue;
 885                 if (unsuspend)
 886                         status = dquot_resume(sb, type);
 887                 else {
 888                         struct ocfs2_mem_dqinfo *oinfo;
 889
 890                         /* Cancel periodic syncing before suspending */
 891                         oinfo = sb_dqinfo(sb, type)->dqi_priv;
 892                         cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 893                         status = dquot_suspend(sb, type);
 894                 }
 895                 if (status < 0)
 896                         break;
 897         }
 898         if (status < 0)
 899                 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
 900                      "remount (error = %d).\n", status);
 901         return status;
 902 }
 903
 904 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
 905 {
 906         struct inode *inode[OCFS2_MAXQUOTAS] = { NULL, NULL };
 907         struct super_block *sb = osb->sb;
 908         unsigned int feature[OCFS2_MAXQUOTAS] = {
 909                                         OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 910                                         OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 911         unsigned int ino[OCFS2_MAXQUOTAS] = {
 912                                         LOCAL_USER_QUOTA_SYSTEM_INODE,
 913                                         LOCAL_GROUP_QUOTA_SYSTEM_INODE };
 914         int status;
 915         int type;
 916
 917         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
 918         for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
 919                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 920                         continue;
 921                 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
 922                                                         osb->slot_num);
 923                 if (!inode[type]) {
 924                         status = -ENOENT;
 925                         goto out_quota_off;
 926                 }
 927                 status = dquot_load_quota_inode(inode[type], type, QFMT_OCFS2,
 928                                                 DQUOT_USAGE_ENABLED);
 929                 if (status < 0)
 930                         goto out_quota_off;
 931         }
 932
 933         for (type = 0; type < OCFS2_MAXQUOTAS; type++)
 934                 iput(inode[type]);
 935         return 0;
 936 out_quota_off:
 937         ocfs2_disable_quotas(osb);
 938         for (type = 0; type < OCFS2_MAXQUOTAS; type++)
 939                 iput(inode[type]);
 940         mlog_errno(status);
 941         return status;
 942 }
 943
 944 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
 945 {
 946         int type;
 947         struct inode *inode;
 948         struct super_block *sb = osb->sb;
 949         struct ocfs2_mem_dqinfo *oinfo;
 950
 951         /* We mostly ignore errors in this function because there's not much
 952          * we can do when we see them */
 953         for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
 954                 if (!sb_has_quota_loaded(sb, type))
 955                         continue;
 956                 if (!sb_has_quota_suspended(sb, type)) {
 957                         oinfo = sb_dqinfo(sb, type)->dqi_priv;
 958                         cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 959                 }
 960                 inode = igrab(sb->s_dquot.files[type]);
 961                 /* Turn off quotas. This will remove all dquot structures from
 962                  * memory and so they will be automatically synced to global
 963                  * quota files */
 964                 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
 965                                         DQUOT_LIMITS_ENABLED);
 966                 iput(inode);
 967         }
 968 }
 969
 970 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
 971 {
 972         struct dentry *root;
 973         int status, sector_size;
 974         struct mount_options parsed_options;
 975         struct inode *inode = NULL;
 976         struct ocfs2_super *osb = NULL;
 977         struct buffer_head *bh = NULL;
 978         char nodestr[12];
 979         struct ocfs2_blockcheck_stats stats;
 980
 981         trace_ocfs2_fill_super(sb, data, silent);
 982
 983         if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
 984                 status = -EINVAL;
 985                 goto out;
 986         }
 987
 988         /* probe for superblock */
 989         status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
 990         if (status < 0) {
 991                 mlog(ML_ERROR, "superblock probe failed!\n");
 992                 goto out;
 993         }
 994
 995         status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
 996         brelse(bh);
 997         bh = NULL;
 998         if (status < 0)
 999                 goto out;
1000
1001         osb = OCFS2_SB(sb);
1002
1003         if (!ocfs2_check_set_options(sb, &parsed_options)) {
1004                 status = -EINVAL;
1005                 goto out_super;
1006         }
1007         osb->s_mount_opt = parsed_options.mount_opt;
1008         osb->s_atime_quantum = parsed_options.atime_quantum;
1009         osb->preferred_slot = parsed_options.slot;
1010         osb->osb_commit_interval = parsed_options.commit_interval;
1011
1012         ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1013         osb->osb_resv_level = parsed_options.resv_level;
1014         osb->osb_dir_resv_level = parsed_options.resv_level;
1015         if (parsed_options.dir_resv_level == -1)
1016                 osb->osb_dir_resv_level = parsed_options.resv_level;
1017         else
1018                 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1019
1020         status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1021         if (status)
1022                 goto out_super;
1023
1024         sb->s_magic = OCFS2_SUPER_MAGIC;
1025
1026         sb->s_flags = (sb->s_flags & ~(SB_POSIXACL | SB_NOSEC)) |
1027                 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? SB_POSIXACL : 0);
1028
1029         /* Hard readonly mode only if: bdev_read_only, SB_RDONLY,
1030          * heartbeat=none */
1031         if (bdev_read_only(sb->s_bdev)) {
1032                 if (!sb_rdonly(sb)) {
1033                         status = -EACCES;
1034                         mlog(ML_ERROR, "Readonly device detected but readonly "
1035                              "mount was not specified.\n");
1036                         goto out_super;
1037                 }
1038
1039                 /* You should not be able to start a local heartbeat
1040                  * on a readonly device. */
1041                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1042                         status = -EROFS;
1043                         mlog(ML_ERROR, "Local heartbeat specified on readonly "
1044                              "device.\n");
1045                         goto out_super;
1046                 }
1047
1048                 status = ocfs2_check_journals_nolocks(osb);
1049                 if (status < 0) {
1050                         if (status == -EROFS)
1051                                 mlog(ML_ERROR, "Recovery required on readonly "
1052                                      "file system, but write access is "
1053                                      "unavailable.\n");
1054                         goto out_super;
1055                 }
1056
1057                 ocfs2_set_ro_flag(osb, 1);
1058
1059                 printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. "
1060                        "Cluster services will not be used for this mount. "
1061                        "Recovery will be skipped.\n", osb->dev_str);
1062         }
1063
1064         if (!ocfs2_is_hard_readonly(osb)) {
1065                 if (sb_rdonly(sb))
1066                         ocfs2_set_ro_flag(osb, 0);
1067         }
1068
1069         status = ocfs2_verify_heartbeat(osb);
1070         if (status < 0)
1071                 goto out_super;
1072
1073         osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1074                                                  ocfs2_debugfs_root);
1075
1076         debugfs_create_file("fs_state", S_IFREG|S_IRUSR, osb->osb_debug_root,
1077                             osb, &ocfs2_osb_debug_fops);
1078
1079         if (ocfs2_meta_ecc(osb))
1080                 ocfs2_blockcheck_stats_debugfs_install( &osb->osb_ecc_stats,
1081                                                         osb->osb_debug_root);
1082
1083         status = ocfs2_mount_volume(sb);
1084         if (status < 0)
1085                 goto out_debugfs;
1086
1087         if (osb->root_inode)
1088                 inode = igrab(osb->root_inode);
1089
1090         if (!inode) {
1091                 status = -EIO;
1092                 goto out_dismount;
1093         }
1094
1095         osb->osb_dev_kset = kset_create_and_add(sb->s_id, NULL,
1096                                                 &ocfs2_kset->kobj);
1097         if (!osb->osb_dev_kset) {
1098                 status = -ENOMEM;
1099                 mlog(ML_ERROR, "Unable to create device kset %s.\n", sb->s_id);
1100                 goto out_dismount;
1101         }
1102
1103         /* Create filecheck sysfs related directories/files at
1104          * /sys/fs/ocfs2/<devname>/filecheck */
1105         if (ocfs2_filecheck_create_sysfs(osb)) {
1106                 status = -ENOMEM;
1107                 mlog(ML_ERROR, "Unable to create filecheck sysfs directory at "
1108                         "/sys/fs/ocfs2/%s/filecheck.\n", sb->s_id);
1109                 goto out_dismount;
1110         }
1111
1112         root = d_make_root(inode);
1113         if (!root) {
1114                 status = -ENOMEM;
1115                 goto out_dismount;
1116         }
1117
1118         sb->s_root = root;
1119
1120         ocfs2_complete_mount_recovery(osb);
1121
1122         if (ocfs2_mount_local(osb))
1123                 snprintf(nodestr, sizeof(nodestr), "local");
1124         else
1125                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1126
1127         printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1128                "with %s data mode.\n",
1129                osb->dev_str, nodestr, osb->slot_num,
1130                osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1131                "ordered");
1132
1133         atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1134         wake_up(&osb->osb_mount_event);
1135
1136         /* Now we can initialize quotas because we can afford to wait
1137          * for cluster locks recovery now. That also means that truncation
1138          * log recovery can happen but that waits for proper quota setup */
1139         if (!sb_rdonly(sb)) {
1140                 status = ocfs2_enable_quotas(osb);
1141                 if (status < 0) {
1142                         /* We have to err-out specially here because
1143                          * s_root is already set */
1144                         mlog_errno(status);
1145                         atomic_set(&osb->vol_state, VOLUME_DISABLED);
1146                         wake_up(&osb->osb_mount_event);
1147                         return status;
1148                 }
1149         }
1150
1151         ocfs2_complete_quota_recovery(osb);
1152
1153         /* Now we wake up again for processes waiting for quotas */
1154         atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1155         wake_up(&osb->osb_mount_event);
1156
1157         /* Start this when the mount is almost sure of being successful */
1158         ocfs2_orphan_scan_start(osb);
1159
1160         return status;
1161
1162 out_dismount:
1163         atomic_set(&osb->vol_state, VOLUME_DISABLED);
1164         wake_up(&osb->osb_mount_event);
1165         ocfs2_free_replay_slots(osb);
1166         ocfs2_dismount_volume(sb, 1);
1167         goto out;
1168
1169 out_debugfs:
1170         debugfs_remove_recursive(osb->osb_debug_root);
1171 out_super:
1172         ocfs2_release_system_inodes(osb);
1173         kfree(osb->recovery_map);
1174         ocfs2_delete_osb(osb);
1175         kfree(osb);
1176 out:
1177         mlog_errno(status);
1178
1179         return status;
1180 }
1181
1182 static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
1183                         int flags,
1184                         const char *dev_name,
1185                         void *data)
1186 {
1187         return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
1188 }
1189
1190 static struct file_system_type ocfs2_fs_type = {
1191         .owner          = THIS_MODULE,
1192         .name           = "ocfs2",
1193         .mount          = ocfs2_mount,
1194         .kill_sb        = kill_block_super,
1195         .fs_flags       = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1196         .next           = NULL
1197 };
1198 MODULE_ALIAS_FS("ocfs2");
1199
1200 static int ocfs2_check_set_options(struct super_block *sb,
1201                                    struct mount_options *options)
1202 {
1203         if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1204             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1205                                          OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1206                 mlog(ML_ERROR, "User quotas were requested, but this "
1207                      "filesystem does not have the feature enabled.\n");
1208                 return 0;
1209         }
1210         if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1211             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1212                                          OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1213                 mlog(ML_ERROR, "Group quotas were requested, but this "
1214                      "filesystem does not have the feature enabled.\n");
1215                 return 0;
1216         }
1217         if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1218             !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1219                 mlog(ML_ERROR, "ACL support requested but extended attributes "
1220                      "feature is not enabled\n");
1221                 return 0;
1222         }
1223         /* No ACL setting specified? Use XATTR feature... */
1224         if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1225                                     OCFS2_MOUNT_NO_POSIX_ACL))) {
1226                 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1227                         options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1228                 else
1229                         options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1230         }
1231         return 1;
1232 }
1233
1234 static int ocfs2_parse_options(struct super_block *sb,
1235                                char *options,
1236                                struct mount_options *mopt,
1237                                int is_remount)
1238 {
1239         int status, user_stack = 0;
1240         char *p;
1241         u32 tmp;
1242         int token, option;
1243         substring_t args[MAX_OPT_ARGS];
1244
1245         trace_ocfs2_parse_options(is_remount, options ? options : "(none)");
1246
1247         mopt->commit_interval = 0;
1248         mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1249         mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1250         mopt->slot = OCFS2_INVALID_SLOT;
1251         mopt->localalloc_opt = -1;
1252         mopt->cluster_stack[0] = '\0';
1253         mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1254         mopt->dir_resv_level = -1;
1255
1256         if (!options) {
1257                 status = 1;
1258                 goto bail;
1259         }
1260
1261         while ((p = strsep(&options, ",")) != NULL) {
1262                 if (!*p)
1263                         continue;
1264
1265                 token = match_token(p, tokens, args);
1266                 switch (token) {
1267                 case Opt_hb_local:
1268                         mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1269                         break;
1270                 case Opt_hb_none:
1271                         mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1272                         break;
1273                 case Opt_hb_global:
1274                         mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1275                         break;
1276                 case Opt_barrier:
1277                         if (match_int(&args[0], &option)) {
1278                                 status = 0;
1279                                 goto bail;
1280                         }
1281                         if (option)
1282                                 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1283                         else
1284                                 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1285                         break;
1286                 case Opt_intr:
1287                         mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1288                         break;
1289                 case Opt_nointr:
1290                         mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1291                         break;
1292                 case Opt_err_panic:
1293                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1294                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1295                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1296                         break;
1297                 case Opt_err_ro:
1298                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1299                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1300                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_ROFS;
1301                         break;
1302                 case Opt_err_cont:
1303                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1304                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1305                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_CONT;
1306                         break;
1307                 case Opt_data_ordered:
1308                         mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1309                         break;
1310                 case Opt_data_writeback:
1311                         mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1312                         break;
1313                 case Opt_user_xattr:
1314                         mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1315                         break;
1316                 case Opt_nouser_xattr:
1317                         mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1318                         break;
1319                 case Opt_atime_quantum:
1320                         if (match_int(&args[0], &option)) {
1321                                 status = 0;
1322                                 goto bail;
1323                         }
1324                         if (option >= 0)
1325                                 mopt->atime_quantum = option;
1326                         break;
1327                 case Opt_slot:
1328                         if (match_int(&args[0], &option)) {
1329                                 status = 0;
1330                                 goto bail;
1331                         }
1332                         if (option)
1333                                 mopt->slot = (u16)option;
1334                         break;
1335                 case Opt_commit:
1336                         if (match_int(&args[0], &option)) {
1337                                 status = 0;
1338                                 goto bail;
1339                         }
1340                         if (option < 0)
1341                                 return 0;
1342                         if (option == 0)
1343                                 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1344                         mopt->commit_interval = HZ * option;
1345                         break;
1346                 case Opt_localalloc:
1347                         if (match_int(&args[0], &option)) {
1348                                 status = 0;
1349                                 goto bail;
1350                         }
1351                         if (option >= 0)
1352                                 mopt->localalloc_opt = option;
1353                         break;
1354                 case Opt_localflocks:
1355                         /*
1356                          * Changing this during remount could race
1357                          * flock() requests, or "unbalance" existing
1358                          * ones (e.g., a lock is taken in one mode but
1359                          * dropped in the other). If users care enough
1360                          * to flip locking modes during remount, we
1361                          * could add a "local" flag to individual
1362                          * flock structures for proper tracking of
1363                          * state.
1364                          */
1365                         if (!is_remount)
1366                                 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1367                         break;
1368                 case Opt_stack:
1369                         /* Check both that the option we were passed
1370                          * is of the right length and that it is a proper
1371                          * string of the right length.
1372                          */
1373                         if (((args[0].to - args[0].from) !=
1374                              OCFS2_STACK_LABEL_LEN) ||
1375                             (strnlen(args[0].from,
1376                                      OCFS2_STACK_LABEL_LEN) !=
1377                              OCFS2_STACK_LABEL_LEN)) {
1378                                 mlog(ML_ERROR,
1379                                      "Invalid cluster_stack option\n");
1380                                 status = 0;
1381                                 goto bail;
1382                         }
1383                         memcpy(mopt->cluster_stack, args[0].from,
1384                                OCFS2_STACK_LABEL_LEN);
1385                         mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1386                         /*
1387                          * Open code the memcmp here as we don't have
1388                          * an osb to pass to
1389                          * ocfs2_userspace_stack().
1390                          */
1391                         if (memcmp(mopt->cluster_stack,
1392                                    OCFS2_CLASSIC_CLUSTER_STACK,
1393                                    OCFS2_STACK_LABEL_LEN))
1394                                 user_stack = 1;
1395                         break;
1396                 case Opt_inode64:
1397                         mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1398                         break;
1399                 case Opt_usrquota:
1400                         mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1401                         break;
1402                 case Opt_grpquota:
1403                         mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1404                         break;
1405                 case Opt_coherency_buffered:
1406                         mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1407                         break;
1408                 case Opt_coherency_full:
1409                         mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1410                         break;
1411                 case Opt_acl:
1412                         mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1413                         mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1414                         break;
1415                 case Opt_noacl:
1416                         mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1417                         mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1418                         break;
1419                 case Opt_resv_level:
1420                         if (is_remount)
1421                                 break;
1422                         if (match_int(&args[0], &option)) {
1423                                 status = 0;
1424                                 goto bail;
1425                         }
1426                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1427                             option < OCFS2_MAX_RESV_LEVEL)
1428                                 mopt->resv_level = option;
1429                         break;
1430                 case Opt_dir_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->dir_resv_level = option;
1440                         break;
1441                 case Opt_journal_async_commit:
1442                         mopt->mount_opt |= OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT;
1443                         break;
1444                 default:
1445                         mlog(ML_ERROR,
1446                              "Unrecognized mount option \"%s\" "
1447                              "or missing value\n", p);
1448                         status = 0;
1449                         goto bail;
1450                 }
1451         }
1452
1453         if (user_stack == 0) {
1454                 /* Ensure only one heartbeat mode */
1455                 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1456                                          OCFS2_MOUNT_HB_GLOBAL |
1457                                          OCFS2_MOUNT_HB_NONE);
1458                 if (hweight32(tmp) != 1) {
1459                         mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1460                         status = 0;
1461                         goto bail;
1462                 }
1463         }
1464
1465         status = 1;
1466
1467 bail:
1468         return status;
1469 }
1470
1471 static int ocfs2_show_options(struct seq_file *s, struct dentry *root)
1472 {
1473         struct ocfs2_super *osb = OCFS2_SB(root->d_sb);
1474         unsigned long opts = osb->s_mount_opt;
1475         unsigned int local_alloc_megs;
1476
1477         if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1478                 seq_printf(s, ",_netdev");
1479                 if (opts & OCFS2_MOUNT_HB_LOCAL)
1480                         seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1481                 else
1482                         seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1483         } else
1484                 seq_printf(s, ",%s", OCFS2_HB_NONE);
1485
1486         if (opts & OCFS2_MOUNT_NOINTR)
1487                 seq_printf(s, ",nointr");
1488
1489         if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1490                 seq_printf(s, ",data=writeback");
1491         else
1492                 seq_printf(s, ",data=ordered");
1493
1494         if (opts & OCFS2_MOUNT_BARRIER)
1495                 seq_printf(s, ",barrier=1");
1496
1497         if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1498                 seq_printf(s, ",errors=panic");
1499         else if (opts & OCFS2_MOUNT_ERRORS_CONT)
1500                 seq_printf(s, ",errors=continue");
1501         else
1502                 seq_printf(s, ",errors=remount-ro");
1503
1504         if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1505                 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1506
1507         seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1508
1509         if (osb->osb_commit_interval)
1510                 seq_printf(s, ",commit=%u",
1511                            (unsigned) (osb->osb_commit_interval / HZ));
1512
1513         local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1514         if (local_alloc_megs != ocfs2_la_default_mb(osb))
1515                 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1516
1517         if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1518                 seq_printf(s, ",localflocks,");
1519
1520         if (osb->osb_cluster_stack[0])
1521                 seq_show_option_n(s, "cluster_stack", osb->osb_cluster_stack,
1522                                   OCFS2_STACK_LABEL_LEN);
1523         if (opts & OCFS2_MOUNT_USRQUOTA)
1524                 seq_printf(s, ",usrquota");
1525         if (opts & OCFS2_MOUNT_GRPQUOTA)
1526                 seq_printf(s, ",grpquota");
1527
1528         if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1529                 seq_printf(s, ",coherency=buffered");
1530         else
1531                 seq_printf(s, ",coherency=full");
1532
1533         if (opts & OCFS2_MOUNT_NOUSERXATTR)
1534                 seq_printf(s, ",nouser_xattr");
1535         else
1536                 seq_printf(s, ",user_xattr");
1537
1538         if (opts & OCFS2_MOUNT_INODE64)
1539                 seq_printf(s, ",inode64");
1540
1541         if (opts & OCFS2_MOUNT_POSIX_ACL)
1542                 seq_printf(s, ",acl");
1543         else
1544                 seq_printf(s, ",noacl");
1545
1546         if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1547                 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1548
1549         if (osb->osb_dir_resv_level != osb->osb_resv_level)
1550                 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1551
1552         if (opts & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
1553                 seq_printf(s, ",journal_async_commit");
1554
1555         return 0;
1556 }
1557
1558 static int __init ocfs2_init(void)
1559 {
1560         int status;
1561
1562         status = init_ocfs2_uptodate_cache();
1563         if (status < 0)
1564                 goto out1;
1565
1566         status = ocfs2_initialize_mem_caches();
1567         if (status < 0)
1568                 goto out2;
1569
1570         ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1571
1572         ocfs2_set_locking_protocol();
1573
1574         status = register_quota_format(&ocfs2_quota_format);
1575         if (status < 0)
1576                 goto out3;
1577         status = register_filesystem(&ocfs2_fs_type);
1578         if (!status)
1579                 return 0;
1580
1581         unregister_quota_format(&ocfs2_quota_format);
1582 out3:
1583         debugfs_remove(ocfs2_debugfs_root);
1584         ocfs2_free_mem_caches();
1585 out2:
1586         exit_ocfs2_uptodate_cache();
1587 out1:
1588         mlog_errno(status);
1589         return status;
1590 }
1591
1592 static void __exit ocfs2_exit(void)
1593 {
1594         unregister_quota_format(&ocfs2_quota_format);
1595
1596         debugfs_remove(ocfs2_debugfs_root);
1597
1598         ocfs2_free_mem_caches();
1599
1600         unregister_filesystem(&ocfs2_fs_type);
1601
1602         exit_ocfs2_uptodate_cache();
1603 }
1604
1605 static void ocfs2_put_super(struct super_block *sb)
1606 {
1607         trace_ocfs2_put_super(sb);
1608
1609         ocfs2_sync_blockdev(sb);
1610         ocfs2_dismount_volume(sb, 0);
1611 }
1612
1613 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1614 {
1615         struct ocfs2_super *osb;
1616         u32 numbits, freebits;
1617         int status;
1618         struct ocfs2_dinode *bm_lock;
1619         struct buffer_head *bh = NULL;
1620         struct inode *inode = NULL;
1621
1622         trace_ocfs2_statfs(dentry->d_sb, buf);
1623
1624         osb = OCFS2_SB(dentry->d_sb);
1625
1626         inode = ocfs2_get_system_file_inode(osb,
1627                                             GLOBAL_BITMAP_SYSTEM_INODE,
1628                                             OCFS2_INVALID_SLOT);
1629         if (!inode) {
1630                 mlog(ML_ERROR, "failed to get bitmap inode\n");
1631                 status = -EIO;
1632                 goto bail;
1633         }
1634
1635         status = ocfs2_inode_lock(inode, &bh, 0);
1636         if (status < 0) {
1637                 mlog_errno(status);
1638                 goto bail;
1639         }
1640
1641         bm_lock = (struct ocfs2_dinode *) bh->b_data;
1642
1643         numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1644         freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1645
1646         buf->f_type = OCFS2_SUPER_MAGIC;
1647         buf->f_bsize = dentry->d_sb->s_blocksize;
1648         buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1649         buf->f_blocks = ((sector_t) numbits) *
1650                         (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1651         buf->f_bfree = ((sector_t) freebits) *
1652                        (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1653         buf->f_bavail = buf->f_bfree;
1654         buf->f_files = numbits;
1655         buf->f_ffree = freebits;
1656         buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1657                                 & 0xFFFFFFFFUL;
1658         buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1659                                 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1660
1661         brelse(bh);
1662
1663         ocfs2_inode_unlock(inode, 0);
1664         status = 0;
1665 bail:
1666         iput(inode);
1667
1668         if (status)
1669                 mlog_errno(status);
1670
1671         return status;
1672 }
1673
1674 static void ocfs2_inode_init_once(void *data)
1675 {
1676         struct ocfs2_inode_info *oi = data;
1677
1678         oi->ip_flags = 0;
1679         oi->ip_open_count = 0;
1680         spin_lock_init(&oi->ip_lock);
1681         ocfs2_extent_map_init(&oi->vfs_inode);
1682         INIT_LIST_HEAD(&oi->ip_io_markers);
1683         INIT_LIST_HEAD(&oi->ip_unwritten_list);
1684         oi->ip_dir_start_lookup = 0;
1685         init_rwsem(&oi->ip_alloc_sem);
1686         init_rwsem(&oi->ip_xattr_sem);
1687         mutex_init(&oi->ip_io_mutex);
1688
1689         oi->ip_blkno = 0ULL;
1690         oi->ip_clusters = 0;
1691         oi->ip_next_orphan = NULL;
1692
1693         ocfs2_resv_init_once(&oi->ip_la_data_resv);
1694
1695         ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1696         ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1697         ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1698
1699         ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1700                                   &ocfs2_inode_caching_ops);
1701
1702         inode_init_once(&oi->vfs_inode);
1703 }
1704
1705 static int ocfs2_initialize_mem_caches(void)
1706 {
1707         ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1708                                        sizeof(struct ocfs2_inode_info),
1709                                        0,
1710                                        (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1711                                                 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1712                                        ocfs2_inode_init_once);
1713         ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1714                                         sizeof(struct ocfs2_dquot),
1715                                         0,
1716                                         (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1717                                                 SLAB_MEM_SPREAD),
1718                                         NULL);
1719         ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1720                                         sizeof(struct ocfs2_quota_chunk),
1721                                         0,
1722                                         (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1723                                         NULL);
1724         if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1725             !ocfs2_qf_chunk_cachep) {
1726                 kmem_cache_destroy(ocfs2_inode_cachep);
1727                 kmem_cache_destroy(ocfs2_dquot_cachep);
1728                 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1729                 return -ENOMEM;
1730         }
1731
1732         return 0;
1733 }
1734
1735 static void ocfs2_free_mem_caches(void)
1736 {
1737         /*
1738          * Make sure all delayed rcu free inodes are flushed before we
1739          * destroy cache.
1740          */
1741         rcu_barrier();
1742         kmem_cache_destroy(ocfs2_inode_cachep);
1743         ocfs2_inode_cachep = NULL;
1744
1745         kmem_cache_destroy(ocfs2_dquot_cachep);
1746         ocfs2_dquot_cachep = NULL;
1747
1748         kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1749         ocfs2_qf_chunk_cachep = NULL;
1750 }
1751
1752 static int ocfs2_get_sector(struct super_block *sb,
1753                             struct buffer_head **bh,
1754                             int block,
1755                             int sect_size)
1756 {
1757         if (!sb_set_blocksize(sb, sect_size)) {
1758                 mlog(ML_ERROR, "unable to set blocksize\n");
1759                 return -EIO;
1760         }
1761
1762         *bh = sb_getblk(sb, block);
1763         if (!*bh) {
1764                 mlog_errno(-ENOMEM);
1765                 return -ENOMEM;
1766         }
1767         lock_buffer(*bh);
1768         if (!buffer_dirty(*bh))
1769                 clear_buffer_uptodate(*bh);
1770         unlock_buffer(*bh);
1771         ll_rw_block(REQ_OP_READ, 0, 1, bh);
1772         wait_on_buffer(*bh);
1773         if (!buffer_uptodate(*bh)) {
1774                 mlog_errno(-EIO);
1775                 brelse(*bh);
1776                 *bh = NULL;
1777                 return -EIO;
1778         }
1779
1780         return 0;
1781 }
1782
1783 static int ocfs2_mount_volume(struct super_block *sb)
1784 {
1785         int status = 0;
1786         struct ocfs2_super *osb = OCFS2_SB(sb);
1787
1788         if (ocfs2_is_hard_readonly(osb))
1789                 goto out;
1790
1791         mutex_init(&osb->obs_trim_fs_mutex);
1792
1793         status = ocfs2_dlm_init(osb);
1794         if (status < 0) {
1795                 mlog_errno(status);
1796                 if (status == -EBADR && ocfs2_userspace_stack(osb))
1797                         mlog(ML_ERROR, "couldn't mount because cluster name on"
1798                         " disk does not match the running cluster name.\n");
1799                 goto out;
1800         }
1801
1802         status = ocfs2_super_lock(osb, 1);
1803         if (status < 0) {
1804                 mlog_errno(status);
1805                 goto out_dlm;
1806         }
1807
1808         /* This will load up the node map and add ourselves to it. */
1809         status = ocfs2_find_slot(osb);
1810         if (status < 0) {
1811                 mlog_errno(status);
1812                 goto out_super_lock;
1813         }
1814
1815         /* load all node-local system inodes */
1816         status = ocfs2_init_local_system_inodes(osb);
1817         if (status < 0) {
1818                 mlog_errno(status);
1819                 goto out_super_lock;
1820         }
1821
1822         status = ocfs2_check_volume(osb);
1823         if (status < 0) {
1824                 mlog_errno(status);
1825                 goto out_system_inodes;
1826         }
1827
1828         status = ocfs2_truncate_log_init(osb);
1829         if (status < 0) {
1830                 mlog_errno(status);
1831                 goto out_check_volume;
1832         }
1833
1834         ocfs2_super_unlock(osb, 1);
1835         return 0;
1836
1837 out_check_volume:
1838         ocfs2_free_replay_slots(osb);
1839 out_system_inodes:
1840         if (osb->local_alloc_state == OCFS2_LA_ENABLED)
1841                 ocfs2_shutdown_local_alloc(osb);
1842         ocfs2_release_system_inodes(osb);
1843         /* before journal shutdown, we should release slot_info */
1844         ocfs2_free_slot_info(osb);
1845         ocfs2_journal_shutdown(osb);
1846 out_super_lock:
1847         ocfs2_super_unlock(osb, 1);
1848 out_dlm:
1849         ocfs2_dlm_shutdown(osb, 0);
1850 out:
1851         return status;
1852 }
1853
1854 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1855 {
1856         int tmp, hangup_needed = 0;
1857         struct ocfs2_super *osb = NULL;
1858         char nodestr[12];
1859
1860         trace_ocfs2_dismount_volume(sb);
1861
1862         BUG_ON(!sb);
1863         osb = OCFS2_SB(sb);
1864         BUG_ON(!osb);
1865
1866         /* Remove file check sysfs related directores/files,
1867          * and wait for the pending file check operations */
1868         ocfs2_filecheck_remove_sysfs(osb);
1869
1870         kset_unregister(osb->osb_dev_kset);
1871
1872         /* Orphan scan should be stopped as early as possible */
1873         ocfs2_orphan_scan_stop(osb);
1874
1875         ocfs2_disable_quotas(osb);
1876
1877         /* All dquots should be freed by now */
1878         WARN_ON(!llist_empty(&osb->dquot_drop_list));
1879         /* Wait for worker to be done with the work structure in osb */
1880         cancel_work_sync(&osb->dquot_drop_work);
1881
1882         ocfs2_shutdown_local_alloc(osb);
1883
1884         ocfs2_truncate_log_shutdown(osb);
1885
1886         /* This will disable recovery and flush any recovery work. */
1887         ocfs2_recovery_exit(osb);
1888
1889         ocfs2_journal_shutdown(osb);
1890
1891         ocfs2_sync_blockdev(sb);
1892
1893         ocfs2_purge_refcount_trees(osb);
1894
1895         /* No cluster connection means we've failed during mount, so skip
1896          * all the steps which depended on that to complete. */
1897         if (osb->cconn) {
1898                 tmp = ocfs2_super_lock(osb, 1);
1899                 if (tmp < 0) {
1900                         mlog_errno(tmp);
1901                         return;
1902                 }
1903         }
1904
1905         if (osb->slot_num != OCFS2_INVALID_SLOT)
1906                 ocfs2_put_slot(osb);
1907
1908         if (osb->cconn)
1909                 ocfs2_super_unlock(osb, 1);
1910
1911         ocfs2_release_system_inodes(osb);
1912
1913         /*
1914          * If we're dismounting due to mount error, mount.ocfs2 will clean
1915          * up heartbeat.  If we're a local mount, there is no heartbeat.
1916          * If we failed before we got a uuid_str yet, we can't stop
1917          * heartbeat.  Otherwise, do it.
1918          */
1919         if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str &&
1920             !ocfs2_is_hard_readonly(osb))
1921                 hangup_needed = 1;
1922
1923         ocfs2_dlm_shutdown(osb, hangup_needed);
1924
1925         ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1926         debugfs_remove_recursive(osb->osb_debug_root);
1927
1928         if (hangup_needed)
1929                 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1930
1931         atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1932
1933         if (ocfs2_mount_local(osb))
1934                 snprintf(nodestr, sizeof(nodestr), "local");
1935         else
1936                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1937
1938         printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1939                osb->dev_str, nodestr);
1940
1941         ocfs2_delete_osb(osb);
1942         kfree(osb);
1943         sb->s_dev = 0;
1944         sb->s_fs_info = NULL;
1945 }
1946
1947 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1948                                 unsigned uuid_bytes)
1949 {
1950         int i, ret;
1951         char *ptr;
1952
1953         BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1954
1955         osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1956         if (osb->uuid_str == NULL)
1957                 return -ENOMEM;
1958
1959         for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1960                 /* print with null */
1961                 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1962                 if (ret != 2) /* drop super cleans up */
1963                         return -EINVAL;
1964                 /* then only advance past the last char */
1965                 ptr += 2;
1966         }
1967
1968         return 0;
1969 }
1970
1971 /* Make sure entire volume is addressable by our journal.  Requires
1972    osb_clusters_at_boot to be valid and for the journal to have been
1973    initialized by ocfs2_journal_init(). */
1974 static int ocfs2_journal_addressable(struct ocfs2_super *osb)
1975 {
1976         int status = 0;
1977         u64 max_block =
1978                 ocfs2_clusters_to_blocks(osb->sb,
1979                                          osb->osb_clusters_at_boot) - 1;
1980
1981         /* 32-bit block number is always OK. */
1982         if (max_block <= (u32)~0ULL)
1983                 goto out;
1984
1985         /* Volume is "huge", so see if our journal is new enough to
1986            support it. */
1987         if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
1988                                        OCFS2_FEATURE_COMPAT_JBD2_SB) &&
1989               jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
1990                                                JBD2_FEATURE_INCOMPAT_64BIT))) {
1991                 mlog(ML_ERROR, "The journal cannot address the entire volume. "
1992                      "Enable the 'block64' journal option with tunefs.ocfs2");
1993                 status = -EFBIG;
1994                 goto out;
1995         }
1996
1997  out:
1998         return status;
1999 }
2000
2001 static int ocfs2_initialize_super(struct super_block *sb,
2002                                   struct buffer_head *bh,
2003                                   int sector_size,
2004                                   struct ocfs2_blockcheck_stats *stats)
2005 {
2006         int status;
2007         int i, cbits, bbits;
2008         struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2009         struct inode *inode = NULL;
2010         struct ocfs2_journal *journal;
2011         struct ocfs2_super *osb;
2012         u64 total_blocks;
2013
2014         osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2015         if (!osb) {
2016                 status = -ENOMEM;
2017                 mlog_errno(status);
2018                 goto bail;
2019         }
2020
2021         sb->s_fs_info = osb;
2022         sb->s_op = &ocfs2_sops;
2023         sb->s_d_op = &ocfs2_dentry_ops;
2024         sb->s_export_op = &ocfs2_export_ops;
2025         sb->s_qcop = &dquot_quotactl_sysfile_ops;
2026         sb->dq_op = &ocfs2_quota_operations;
2027         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
2028         sb->s_xattr = ocfs2_xattr_handlers;
2029         sb->s_time_gran = 1;
2030         sb->s_flags |= SB_NOATIME;
2031         /* this is needed to support O_LARGEFILE */
2032         cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2033         bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2034         sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2035         memcpy(&sb->s_uuid, di->id2.i_super.s_uuid,
2036                sizeof(di->id2.i_super.s_uuid));
2037
2038         osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2039
2040         for (i = 0; i < 3; i++)
2041                 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2042         osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2043
2044         osb->sb = sb;
2045         osb->s_sectsize_bits = blksize_bits(sector_size);
2046         BUG_ON(!osb->s_sectsize_bits);
2047
2048         spin_lock_init(&osb->dc_task_lock);
2049         init_waitqueue_head(&osb->dc_event);
2050         osb->dc_work_sequence = 0;
2051         osb->dc_wake_sequence = 0;
2052         INIT_LIST_HEAD(&osb->blocked_lock_list);
2053         osb->blocked_lock_count = 0;
2054         spin_lock_init(&osb->osb_lock);
2055         spin_lock_init(&osb->osb_xattr_lock);
2056         ocfs2_init_steal_slots(osb);
2057
2058         mutex_init(&osb->system_file_mutex);
2059
2060         atomic_set(&osb->alloc_stats.moves, 0);
2061         atomic_set(&osb->alloc_stats.local_data, 0);
2062         atomic_set(&osb->alloc_stats.bitmap_data, 0);
2063         atomic_set(&osb->alloc_stats.bg_allocs, 0);
2064         atomic_set(&osb->alloc_stats.bg_extends, 0);
2065
2066         /* Copy the blockcheck stats from the superblock probe */
2067         osb->osb_ecc_stats = *stats;
2068
2069         ocfs2_init_node_maps(osb);
2070
2071         snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2072                  MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2073
2074         osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2075         if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2076                 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2077                      osb->max_slots);
2078                 status = -EINVAL;
2079                 goto bail;
2080         }
2081
2082         ocfs2_orphan_scan_init(osb);
2083
2084         status = ocfs2_recovery_init(osb);
2085         if (status) {
2086                 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2087                 mlog_errno(status);
2088                 goto bail;
2089         }
2090
2091         init_waitqueue_head(&osb->checkpoint_event);
2092
2093         osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2094
2095         osb->slot_num = OCFS2_INVALID_SLOT;
2096
2097         osb->s_xattr_inline_size = le16_to_cpu(
2098                                         di->id2.i_super.s_xattr_inline_size);
2099
2100         osb->local_alloc_state = OCFS2_LA_UNUSED;
2101         osb->local_alloc_bh = NULL;
2102         INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2103
2104         init_waitqueue_head(&osb->osb_mount_event);
2105
2106         status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2107         if (status) {
2108                 mlog_errno(status);
2109                 goto bail;
2110         }
2111
2112         osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2113         if (!osb->vol_label) {
2114                 mlog(ML_ERROR, "unable to alloc vol label\n");
2115                 status = -ENOMEM;
2116                 goto bail;
2117         }
2118
2119         osb->slot_recovery_generations =
2120                 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2121                         GFP_KERNEL);
2122         if (!osb->slot_recovery_generations) {
2123                 status = -ENOMEM;
2124                 mlog_errno(status);
2125                 goto bail;
2126         }
2127
2128         init_waitqueue_head(&osb->osb_wipe_event);
2129         osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2130                                         sizeof(*osb->osb_orphan_wipes),
2131                                         GFP_KERNEL);
2132         if (!osb->osb_orphan_wipes) {
2133                 status = -ENOMEM;
2134                 mlog_errno(status);
2135                 goto bail;
2136         }
2137
2138         osb->osb_rf_lock_tree = RB_ROOT;
2139
2140         osb->s_feature_compat =
2141                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2142         osb->s_feature_ro_compat =
2143                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2144         osb->s_feature_incompat =
2145                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2146
2147         if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2148                 mlog(ML_ERROR, "couldn't mount because of unsupported "
2149                      "optional features (%x).\n", i);
2150                 status = -EINVAL;
2151                 goto bail;
2152         }
2153         if (!sb_rdonly(osb->sb) && (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2154                 mlog(ML_ERROR, "couldn't mount RDWR because of "
2155                      "unsupported optional features (%x).\n", i);
2156                 status = -EINVAL;
2157                 goto bail;
2158         }
2159
2160         if (ocfs2_clusterinfo_valid(osb)) {
2161                 /*
2162                  * ci_stack and ci_cluster in ocfs2_cluster_info may not be null
2163                  * terminated, so make sure no overflow happens here by using
2164                  * memcpy. Destination strings will always be null terminated
2165                  * because osb is allocated using kzalloc.
2166                  */
2167                 osb->osb_stackflags =
2168                         OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2169                 memcpy(osb->osb_cluster_stack,
2170                        OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2171                        OCFS2_STACK_LABEL_LEN);
2172                 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2173                         mlog(ML_ERROR,
2174                              "couldn't mount because of an invalid "
2175                              "cluster stack label (%s) \n",
2176                              osb->osb_cluster_stack);
2177                         status = -EINVAL;
2178                         goto bail;
2179                 }
2180                 memcpy(osb->osb_cluster_name,
2181                         OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster,
2182                         OCFS2_CLUSTER_NAME_LEN);
2183         } else {
2184                 /* The empty string is identical with classic tools that
2185                  * don't know about s_cluster_info. */
2186                 osb->osb_cluster_stack[0] = '\0';
2187         }
2188
2189         get_random_bytes(&osb->s_next_generation, sizeof(u32));
2190
2191         /* FIXME
2192          * This should be done in ocfs2_journal_init(), but unknown
2193          * ordering issues will cause the filesystem to crash.
2194          * If anyone wants to figure out what part of the code
2195          * refers to osb->journal before ocfs2_journal_init() is run,
2196          * be my guest.
2197          */
2198         /* initialize our journal structure */
2199
2200         journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2201         if (!journal) {
2202                 mlog(ML_ERROR, "unable to alloc journal\n");
2203                 status = -ENOMEM;
2204                 goto bail;
2205         }
2206         osb->journal = journal;
2207         journal->j_osb = osb;
2208
2209         atomic_set(&journal->j_num_trans, 0);
2210         init_rwsem(&journal->j_trans_barrier);
2211         init_waitqueue_head(&journal->j_checkpointed);
2212         spin_lock_init(&journal->j_lock);
2213         journal->j_trans_id = (unsigned long) 1;
2214         INIT_LIST_HEAD(&journal->j_la_cleanups);
2215         INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2216         journal->j_state = OCFS2_JOURNAL_FREE;
2217
2218         INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs);
2219         init_llist_head(&osb->dquot_drop_list);
2220
2221         /* get some pseudo constants for clustersize bits */
2222         osb->s_clustersize_bits =
2223                 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2224         osb->s_clustersize = 1 << osb->s_clustersize_bits;
2225
2226         if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2227             osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2228                 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2229                      osb->s_clustersize);
2230                 status = -EINVAL;
2231                 goto bail;
2232         }
2233
2234         total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2235                                                 le32_to_cpu(di->i_clusters));
2236
2237         status = generic_check_addressable(osb->sb->s_blocksize_bits,
2238                                            total_blocks);
2239         if (status) {
2240                 mlog(ML_ERROR, "Volume too large "
2241                      "to mount safely on this system");
2242                 status = -EFBIG;
2243                 goto bail;
2244         }
2245
2246         if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2247                                  sizeof(di->id2.i_super.s_uuid))) {
2248                 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2249                 status = -ENOMEM;
2250                 goto bail;
2251         }
2252
2253         strlcpy(osb->vol_label, di->id2.i_super.s_label,
2254                 OCFS2_MAX_VOL_LABEL_LEN);
2255         osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2256         osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2257         osb->first_cluster_group_blkno =
2258                 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2259         osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2260         osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2261         trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str,
2262                                      (unsigned long long)osb->root_blkno,
2263                                      (unsigned long long)osb->system_dir_blkno,
2264                                      osb->s_clustersize_bits);
2265
2266         osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2267         if (!osb->osb_dlm_debug) {
2268                 status = -ENOMEM;
2269                 mlog_errno(status);
2270                 goto bail;
2271         }
2272
2273         atomic_set(&osb->vol_state, VOLUME_INIT);
2274
2275         /* load root, system_dir, and all global system inodes */
2276         status = ocfs2_init_global_system_inodes(osb);
2277         if (status < 0) {
2278                 mlog_errno(status);
2279                 goto bail;
2280         }
2281
2282         /*
2283          * global bitmap
2284          */
2285         inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2286                                             OCFS2_INVALID_SLOT);
2287         if (!inode) {
2288                 status = -EINVAL;
2289                 mlog_errno(status);
2290                 goto bail;
2291         }
2292
2293         osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2294         osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2295         iput(inode);
2296
2297         osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2298                                  osb->s_feature_incompat) * 8;
2299
2300         status = ocfs2_init_slot_info(osb);
2301         if (status < 0) {
2302                 mlog_errno(status);
2303                 goto bail;
2304         }
2305         cleancache_init_shared_fs(sb);
2306
2307         osb->ocfs2_wq = alloc_ordered_workqueue("ocfs2_wq", WQ_MEM_RECLAIM);
2308         if (!osb->ocfs2_wq) {
2309                 status = -ENOMEM;
2310                 mlog_errno(status);
2311         }
2312
2313 bail:
2314         return status;
2315 }
2316
2317 /*
2318  * will return: -EAGAIN if it is ok to keep searching for superblocks
2319  *              -EINVAL if there is a bad superblock
2320  *              0 on success
2321  */
2322 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2323                                struct buffer_head *bh,
2324                                u32 blksz,
2325                                struct ocfs2_blockcheck_stats *stats)
2326 {
2327         int status = -EAGAIN;
2328
2329         if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2330                    strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2331                 /* We have to do a raw check of the feature here */
2332                 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2333                     OCFS2_FEATURE_INCOMPAT_META_ECC) {
2334                         status = ocfs2_block_check_validate(bh->b_data,
2335                                                             bh->b_size,
2336                                                             &di->i_check,
2337                                                             stats);
2338                         if (status)
2339                                 goto out;
2340                 }
2341                 status = -EINVAL;
2342                 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2343                         mlog(ML_ERROR, "found superblock with incorrect block "
2344                              "size: found %u, should be %u\n",
2345                              1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2346                                blksz);
2347                 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2348                            OCFS2_MAJOR_REV_LEVEL ||
2349                            le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2350                            OCFS2_MINOR_REV_LEVEL) {
2351                         mlog(ML_ERROR, "found superblock with bad version: "
2352                              "found %u.%u, should be %u.%u\n",
2353                              le16_to_cpu(di->id2.i_super.s_major_rev_level),
2354                              le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2355                              OCFS2_MAJOR_REV_LEVEL,
2356                              OCFS2_MINOR_REV_LEVEL);
2357                 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2358                         mlog(ML_ERROR, "bad block number on superblock: "
2359                              "found %llu, should be %llu\n",
2360                              (unsigned long long)le64_to_cpu(di->i_blkno),
2361                              (unsigned long long)bh->b_blocknr);
2362                 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2363                             le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2364                         mlog(ML_ERROR, "bad cluster size found: %u\n",
2365                              1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2366                 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2367                         mlog(ML_ERROR, "bad root_blkno: 0\n");
2368                 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2369                         mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2370                 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2371                         mlog(ML_ERROR,
2372                              "Superblock slots found greater than file system "
2373                              "maximum: found %u, max %u\n",
2374                              le16_to_cpu(di->id2.i_super.s_max_slots),
2375                              OCFS2_MAX_SLOTS);
2376                 } else {
2377                         /* found it! */
2378                         status = 0;
2379                 }
2380         }
2381
2382 out:
2383         if (status && status != -EAGAIN)
2384                 mlog_errno(status);
2385         return status;
2386 }
2387
2388 static int ocfs2_check_volume(struct ocfs2_super *osb)
2389 {
2390         int status;
2391         int dirty;
2392         int local;
2393         struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2394                                                   * recover
2395                                                   * ourselves. */
2396
2397         /* Init our journal object. */
2398         status = ocfs2_journal_init(osb->journal, &dirty);
2399         if (status < 0) {
2400                 mlog(ML_ERROR, "Could not initialize journal!\n");
2401                 goto finally;
2402         }
2403
2404         /* Now that journal has been initialized, check to make sure
2405            entire volume is addressable. */
2406         status = ocfs2_journal_addressable(osb);
2407         if (status)
2408                 goto finally;
2409
2410         /* If the journal was unmounted cleanly then we don't want to
2411          * recover anything. Otherwise, journal_load will do that
2412          * dirty work for us :) */
2413         if (!dirty) {
2414                 status = ocfs2_journal_wipe(osb->journal, 0);
2415                 if (status < 0) {
2416                         mlog_errno(status);
2417                         goto finally;
2418                 }
2419         } else {
2420                 printk(KERN_NOTICE "ocfs2: File system on device (%s) was not "
2421                        "unmounted cleanly, recovering it.\n", osb->dev_str);
2422         }
2423
2424         local = ocfs2_mount_local(osb);
2425
2426         /* will play back anything left in the journal. */
2427         status = ocfs2_journal_load(osb->journal, local, dirty);
2428         if (status < 0) {
2429                 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2430                 goto finally;
2431         }
2432
2433         if (osb->s_mount_opt & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
2434                 jbd2_journal_set_features(osb->journal->j_journal,
2435                                 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2436                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2437         else
2438                 jbd2_journal_clear_features(osb->journal->j_journal,
2439                                 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2440                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2441
2442         if (dirty) {
2443                 /* recover my local alloc if we didn't unmount cleanly. */
2444                 status = ocfs2_begin_local_alloc_recovery(osb,
2445                                                           osb->slot_num,
2446                                                           &local_alloc);
2447                 if (status < 0) {
2448                         mlog_errno(status);
2449                         goto finally;
2450                 }
2451                 /* we complete the recovery process after we've marked
2452                  * ourselves as mounted. */
2453         }
2454
2455         status = ocfs2_load_local_alloc(osb);
2456         if (status < 0) {
2457                 mlog_errno(status);
2458                 goto finally;
2459         }
2460
2461         if (dirty) {
2462                 /* Recovery will be completed after we've mounted the
2463                  * rest of the volume. */
2464                 osb->local_alloc_copy = local_alloc;
2465                 local_alloc = NULL;
2466         }
2467
2468         /* go through each journal, trylock it and if you get the
2469          * lock, and it's marked as dirty, set the bit in the recover
2470          * map and launch a recovery thread for it. */
2471         status = ocfs2_mark_dead_nodes(osb);
2472         if (status < 0) {
2473                 mlog_errno(status);
2474                 goto finally;
2475         }
2476
2477         status = ocfs2_compute_replay_slots(osb);
2478         if (status < 0)
2479                 mlog_errno(status);
2480
2481 finally:
2482         kfree(local_alloc);
2483
2484         if (status)
2485                 mlog_errno(status);
2486         return status;
2487 }
2488
2489 /*
2490  * The routine gets called from dismount or close whenever a dismount on
2491  * volume is requested and the osb open count becomes 1.
2492  * It will remove the osb from the global list and also free up all the
2493  * initialized resources and fileobject.
2494  */
2495 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2496 {
2497         /* This function assumes that the caller has the main osb resource */
2498
2499         /* ocfs2_initializer_super have already created this workqueue */
2500         if (osb->ocfs2_wq)
2501                 destroy_workqueue(osb->ocfs2_wq);
2502
2503         ocfs2_free_slot_info(osb);
2504
2505         kfree(osb->osb_orphan_wipes);
2506         kfree(osb->slot_recovery_generations);
2507         /* FIXME
2508          * This belongs in journal shutdown, but because we have to
2509          * allocate osb->journal at the start of ocfs2_initialize_osb(),
2510          * we free it here.
2511          */
2512         kfree(osb->journal);
2513         kfree(osb->local_alloc_copy);
2514         kfree(osb->uuid_str);
2515         kfree(osb->vol_label);
2516         ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2517         memset(osb, 0, sizeof(struct ocfs2_super));
2518 }
2519
2520 /* Depending on the mount option passed, perform one of the following:
2521  * Put OCFS2 into a readonly state (default)
2522  * Return EIO so that only the process errs
2523  * Fix the error as if fsck.ocfs2 -y
2524  * panic
2525  */
2526 static int ocfs2_handle_error(struct super_block *sb)
2527 {
2528         struct ocfs2_super *osb = OCFS2_SB(sb);
2529         int rv = 0;
2530
2531         ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2532         pr_crit("On-disk corruption discovered. "
2533                 "Please run fsck.ocfs2 once the filesystem is unmounted.\n");
2534
2535         if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC) {
2536                 panic("OCFS2: (device %s): panic forced after error\n",
2537                       sb->s_id);
2538         } else if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_CONT) {
2539                 pr_crit("OCFS2: Returning error to the calling process.\n");
2540                 rv = -EIO;
2541         } else { /* default option */
2542                 rv = -EROFS;
2543                 if (sb_rdonly(sb) && (ocfs2_is_soft_readonly(osb) || ocfs2_is_hard_readonly(osb)))
2544                         return rv;
2545
2546                 pr_crit("OCFS2: File system is now read-only.\n");
2547                 sb->s_flags |= SB_RDONLY;
2548                 ocfs2_set_ro_flag(osb, 0);
2549         }
2550
2551         return rv;
2552 }
2553
2554 int __ocfs2_error(struct super_block *sb, const char *function,
2555                   const char *fmt, ...)
2556 {
2557         struct va_format vaf;
2558         va_list args;
2559
2560         va_start(args, fmt);
2561         vaf.fmt = fmt;
2562         vaf.va = &args;
2563
2564         /* Not using mlog here because we want to show the actual
2565          * function the error came from. */
2566         printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %pV",
2567                sb->s_id, function, &vaf);
2568
2569         va_end(args);
2570
2571         return ocfs2_handle_error(sb);
2572 }
2573
2574 /* Handle critical errors. This is intentionally more drastic than
2575  * ocfs2_handle_error, so we only use for things like journal errors,
2576  * etc. */
2577 void __ocfs2_abort(struct super_block *sb, const char *function,
2578                    const char *fmt, ...)
2579 {
2580         struct va_format vaf;
2581         va_list args;
2582
2583         va_start(args, fmt);
2584
2585         vaf.fmt = fmt;
2586         vaf.va = &args;
2587
2588         printk(KERN_CRIT "OCFS2: abort (device %s): %s: %pV",
2589                sb->s_id, function, &vaf);
2590
2591         va_end(args);
2592
2593         /* We don't have the cluster support yet to go straight to
2594          * hard readonly in here. Until then, we want to keep
2595          * ocfs2_abort() so that we can at least mark critical
2596          * errors.
2597          *
2598          * TODO: This should abort the journal and alert other nodes
2599          * that our slot needs recovery. */
2600
2601         /* Force a panic(). This stinks, but it's better than letting
2602          * things continue without having a proper hard readonly
2603          * here. */
2604         if (!ocfs2_mount_local(OCFS2_SB(sb)))
2605                 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2606         ocfs2_handle_error(sb);
2607 }
2608
2609 /*
2610  * Void signal blockers, because in-kernel sigprocmask() only fails
2611  * when SIG_* is wrong.
2612  */
2613 void ocfs2_block_signals(sigset_t *oldset)
2614 {
2615         int rc;
2616         sigset_t blocked;
2617
2618         sigfillset(&blocked);
2619         rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2620         BUG_ON(rc);
2621 }
2622
2623 void ocfs2_unblock_signals(sigset_t *oldset)
2624 {
2625         int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2626         BUG_ON(rc);
2627 }
2628
2629 module_init(ocfs2_init);
2630 module_exit(ocfs2_exit);