fs/ocfs2/super.c

   1 /* -*- mode: c; c-basic-offset: 8; -*-
   2  * vim: noexpandtab sw=8 ts=8 sts=0:
   3  *
   4  * super.c
   5  *
   6  * load/unload driver, mount/dismount volumes
   7  *
   8  * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
   9  *
  10  * This program is free software; you can redistribute it and/or
  11  * modify it under the terms of the GNU General Public
  12  * License as published by the Free Software Foundation; either
  13  * version 2 of the License, or (at your option) any later version.
  14  *
  15  * This program is distributed in the hope that it will be useful,
  16  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  17  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  18  * General Public License for more details.
  19  *
  20  * You should have received a copy of the GNU General Public
  21  * License along with this program; if not, write to the
  22  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  23  * Boston, MA 021110-1307, USA.
  24  */
  25
  26 #include <linux/module.h>
  27 #include <linux/fs.h>
  28 #include <linux/types.h>
  29 #include <linux/slab.h>
  30 #include <linux/highmem.h>
  31 #include <linux/init.h>
  32 #include <linux/random.h>
  33 #include <linux/statfs.h>
  34 #include <linux/moduleparam.h>
  35 #include <linux/blkdev.h>
  36 #include <linux/socket.h>
  37 #include <linux/inet.h>
  38 #include <linux/parser.h>
  39 #include <linux/crc32.h>
  40 #include <linux/debugfs.h>
  41 #include <linux/mount.h>
  42 #include <linux/seq_file.h>
  43 #include <linux/quotaops.h>
  44 #include <linux/cleancache.h>
  45
  46 #define CREATE_TRACE_POINTS
  47 #include "ocfs2_trace.h"
  48
  49 #include <cluster/masklog.h>
  50
  51 #include "ocfs2.h"
  52
  53 /* this should be the only file to include a version 1 header */
  54 #include "ocfs1_fs_compat.h"
  55
  56 #include "alloc.h"
  57 #include "aops.h"
  58 #include "blockcheck.h"
  59 #include "dlmglue.h"
  60 #include "export.h"
  61 #include "extent_map.h"
  62 #include "heartbeat.h"
  63 #include "inode.h"
  64 #include "journal.h"
  65 #include "localalloc.h"
  66 #include "namei.h"
  67 #include "slot_map.h"
  68 #include "super.h"
  69 #include "sysfile.h"
  70 #include "uptodate.h"
  71 #include "ver.h"
  72 #include "xattr.h"
  73 #include "quota.h"
  74 #include "refcounttree.h"
  75 #include "suballoc.h"
  76
  77 #include "buffer_head_io.h"
  78
  79 static struct kmem_cache *ocfs2_inode_cachep = NULL;
  80 struct kmem_cache *ocfs2_dquot_cachep;
  81 struct kmem_cache *ocfs2_qf_chunk_cachep;
  82
  83 /* OCFS2 needs to schedule several different types of work which
  84  * require cluster locking, disk I/O, recovery waits, etc. Since these
  85  * types of work tend to be heavy we avoid using the kernel events
  86  * workqueue and schedule on our own. */
  87 struct workqueue_struct *ocfs2_wq = NULL;
  88
  89 static struct dentry *ocfs2_debugfs_root = NULL;
  90
  91 MODULE_AUTHOR("Oracle");
  92 MODULE_LICENSE("GPL");
  93
  94 struct mount_options
  95 {
  96         unsigned long   commit_interval;
  97         unsigned long   mount_opt;
  98         unsigned int    atime_quantum;
  99         signed short    slot;
 100         int             localalloc_opt;
 101         unsigned int    resv_level;
 102         int             dir_resv_level;
 103         char            cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
 104 };
 105
 106 static int ocfs2_parse_options(struct super_block *sb, char *options,
 107                                struct mount_options *mopt,
 108                                int is_remount);
 109 static int ocfs2_check_set_options(struct super_block *sb,
 110                                    struct mount_options *options);
 111 static int ocfs2_show_options(struct seq_file *s, struct dentry *root);
 112 static void ocfs2_put_super(struct super_block *sb);
 113 static int ocfs2_mount_volume(struct super_block *sb);
 114 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
 115 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
 116 static int ocfs2_initialize_mem_caches(void);
 117 static void ocfs2_free_mem_caches(void);
 118 static void ocfs2_delete_osb(struct ocfs2_super *osb);
 119
 120 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
 121
 122 static int ocfs2_sync_fs(struct super_block *sb, int wait);
 123
 124 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
 125 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
 126 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
 127 static int ocfs2_check_volume(struct ocfs2_super *osb);
 128 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
 129                                struct buffer_head *bh,
 130                                u32 sectsize,
 131                                struct ocfs2_blockcheck_stats *stats);
 132 static int ocfs2_initialize_super(struct super_block *sb,
 133                                   struct buffer_head *bh,
 134                                   int sector_size,
 135                                   struct ocfs2_blockcheck_stats *stats);
 136 static int ocfs2_get_sector(struct super_block *sb,
 137                             struct buffer_head **bh,
 138                             int block,
 139                             int sect_size);
 140 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
 141 static void ocfs2_destroy_inode(struct inode *inode);
 142 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
 143 static int ocfs2_enable_quotas(struct ocfs2_super *osb);
 144 static void ocfs2_disable_quotas(struct ocfs2_super *osb);
 145
 146 static const struct super_operations ocfs2_sops = {
 147         .statfs         = ocfs2_statfs,
 148         .alloc_inode    = ocfs2_alloc_inode,
 149         .destroy_inode  = ocfs2_destroy_inode,
 150         .drop_inode     = ocfs2_drop_inode,
 151         .evict_inode    = ocfs2_evict_inode,
 152         .sync_fs        = ocfs2_sync_fs,
 153         .put_super      = ocfs2_put_super,
 154         .remount_fs     = ocfs2_remount,
 155         .show_options   = ocfs2_show_options,
 156         .quota_read     = ocfs2_quota_read,
 157         .quota_write    = ocfs2_quota_write,
 158 };
 159
 160 enum {
 161         Opt_barrier,
 162         Opt_err_panic,
 163         Opt_err_ro,
 164         Opt_intr,
 165         Opt_nointr,
 166         Opt_hb_none,
 167         Opt_hb_local,
 168         Opt_hb_global,
 169         Opt_data_ordered,
 170         Opt_data_writeback,
 171         Opt_atime_quantum,
 172         Opt_slot,
 173         Opt_commit,
 174         Opt_localalloc,
 175         Opt_localflocks,
 176         Opt_stack,
 177         Opt_user_xattr,
 178         Opt_nouser_xattr,
 179         Opt_inode64,
 180         Opt_acl,
 181         Opt_noacl,
 182         Opt_usrquota,
 183         Opt_grpquota,
 184         Opt_coherency_buffered,
 185         Opt_coherency_full,
 186         Opt_resv_level,
 187         Opt_dir_resv_level,
 188         Opt_err,
 189 };
 190
 191 static const match_table_t tokens = {
 192         {Opt_barrier, "barrier=%u"},
 193         {Opt_err_panic, "errors=panic"},
 194         {Opt_err_ro, "errors=remount-ro"},
 195         {Opt_intr, "intr"},
 196         {Opt_nointr, "nointr"},
 197         {Opt_hb_none, OCFS2_HB_NONE},
 198         {Opt_hb_local, OCFS2_HB_LOCAL},
 199         {Opt_hb_global, OCFS2_HB_GLOBAL},
 200         {Opt_data_ordered, "data=ordered"},
 201         {Opt_data_writeback, "data=writeback"},
 202         {Opt_atime_quantum, "atime_quantum=%u"},
 203         {Opt_slot, "preferred_slot=%u"},
 204         {Opt_commit, "commit=%u"},
 205         {Opt_localalloc, "localalloc=%d"},
 206         {Opt_localflocks, "localflocks"},
 207         {Opt_stack, "cluster_stack=%s"},
 208         {Opt_user_xattr, "user_xattr"},
 209         {Opt_nouser_xattr, "nouser_xattr"},
 210         {Opt_inode64, "inode64"},
 211         {Opt_acl, "acl"},
 212         {Opt_noacl, "noacl"},
 213         {Opt_usrquota, "usrquota"},
 214         {Opt_grpquota, "grpquota"},
 215         {Opt_coherency_buffered, "coherency=buffered"},
 216         {Opt_coherency_full, "coherency=full"},
 217         {Opt_resv_level, "resv_level=%u"},
 218         {Opt_dir_resv_level, "dir_resv_level=%u"},
 219         {Opt_err, NULL}
 220 };
 221
 222 #ifdef CONFIG_DEBUG_FS
 223 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
 224 {
 225         struct ocfs2_cluster_connection *cconn = osb->cconn;
 226         struct ocfs2_recovery_map *rm = osb->recovery_map;
 227         struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
 228         int i, out = 0;
 229
 230         out += snprintf(buf + out, len - out,
 231                         "%10s => Id: %-s  Uuid: %-s  Gen: 0x%X  Label: %-s\n",
 232                         "Device", osb->dev_str, osb->uuid_str,
 233                         osb->fs_generation, osb->vol_label);
 234
 235         out += snprintf(buf + out, len - out,
 236                         "%10s => State: %d  Flags: 0x%lX\n", "Volume",
 237                         atomic_read(&osb->vol_state), osb->osb_flags);
 238
 239         out += snprintf(buf + out, len - out,
 240                         "%10s => Block: %lu  Cluster: %d\n", "Sizes",
 241                         osb->sb->s_blocksize, osb->s_clustersize);
 242
 243         out += snprintf(buf + out, len - out,
 244                         "%10s => Compat: 0x%X  Incompat: 0x%X  "
 245                         "ROcompat: 0x%X\n",
 246                         "Features", osb->s_feature_compat,
 247                         osb->s_feature_incompat, osb->s_feature_ro_compat);
 248
 249         out += snprintf(buf + out, len - out,
 250                         "%10s => Opts: 0x%lX  AtimeQuanta: %u\n", "Mount",
 251                         osb->s_mount_opt, osb->s_atime_quantum);
 252
 253         if (cconn) {
 254                 out += snprintf(buf + out, len - out,
 255                                 "%10s => Stack: %s  Name: %*s  "
 256                                 "Version: %d.%d\n", "Cluster",
 257                                 (*osb->osb_cluster_stack == '\0' ?
 258                                  "o2cb" : osb->osb_cluster_stack),
 259                                 cconn->cc_namelen, cconn->cc_name,
 260                                 cconn->cc_version.pv_major,
 261                                 cconn->cc_version.pv_minor);
 262         }
 263
 264         spin_lock(&osb->dc_task_lock);
 265         out += snprintf(buf + out, len - out,
 266                         "%10s => Pid: %d  Count: %lu  WakeSeq: %lu  "
 267                         "WorkSeq: %lu\n", "DownCnvt",
 268                         (osb->dc_task ?  task_pid_nr(osb->dc_task) : -1),
 269                         osb->blocked_lock_count, osb->dc_wake_sequence,
 270                         osb->dc_work_sequence);
 271         spin_unlock(&osb->dc_task_lock);
 272
 273         spin_lock(&osb->osb_lock);
 274         out += snprintf(buf + out, len - out, "%10s => Pid: %d  Nodes:",
 275                         "Recovery",
 276                         (osb->recovery_thread_task ?
 277                          task_pid_nr(osb->recovery_thread_task) : -1));
 278         if (rm->rm_used == 0)
 279                 out += snprintf(buf + out, len - out, " None\n");
 280         else {
 281                 for (i = 0; i < rm->rm_used; i++)
 282                         out += snprintf(buf + out, len - out, " %d",
 283                                         rm->rm_entries[i]);
 284                 out += snprintf(buf + out, len - out, "\n");
 285         }
 286         spin_unlock(&osb->osb_lock);
 287
 288         out += snprintf(buf + out, len - out,
 289                         "%10s => Pid: %d  Interval: %lu  Needs: %d\n", "Commit",
 290                         (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
 291                         osb->osb_commit_interval,
 292                         atomic_read(&osb->needs_checkpoint));
 293
 294         out += snprintf(buf + out, len - out,
 295                         "%10s => State: %d  TxnId: %lu  NumTxns: %d\n",
 296                         "Journal", osb->journal->j_state,
 297                         osb->journal->j_trans_id,
 298                         atomic_read(&osb->journal->j_num_trans));
 299
 300         out += snprintf(buf + out, len - out,
 301                         "%10s => GlobalAllocs: %d  LocalAllocs: %d  "
 302                         "SubAllocs: %d  LAWinMoves: %d  SAExtends: %d\n",
 303                         "Stats",
 304                         atomic_read(&osb->alloc_stats.bitmap_data),
 305                         atomic_read(&osb->alloc_stats.local_data),
 306                         atomic_read(&osb->alloc_stats.bg_allocs),
 307                         atomic_read(&osb->alloc_stats.moves),
 308                         atomic_read(&osb->alloc_stats.bg_extends));
 309
 310         out += snprintf(buf + out, len - out,
 311                         "%10s => State: %u  Descriptor: %llu  Size: %u bits  "
 312                         "Default: %u bits\n",
 313                         "LocalAlloc", osb->local_alloc_state,
 314                         (unsigned long long)osb->la_last_gd,
 315                         osb->local_alloc_bits, osb->local_alloc_default_bits);
 316
 317         spin_lock(&osb->osb_lock);
 318         out += snprintf(buf + out, len - out,
 319                         "%10s => InodeSlot: %d  StolenInodes: %d, "
 320                         "MetaSlot: %d  StolenMeta: %d\n", "Steal",
 321                         osb->s_inode_steal_slot,
 322                         atomic_read(&osb->s_num_inodes_stolen),
 323                         osb->s_meta_steal_slot,
 324                         atomic_read(&osb->s_num_meta_stolen));
 325         spin_unlock(&osb->osb_lock);
 326
 327         out += snprintf(buf + out, len - out, "OrphanScan => ");
 328         out += snprintf(buf + out, len - out, "Local: %u  Global: %u ",
 329                         os->os_count, os->os_seqno);
 330         out += snprintf(buf + out, len - out, " Last Scan: ");
 331         if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
 332                 out += snprintf(buf + out, len - out, "Disabled\n");
 333         else
 334                 out += snprintf(buf + out, len - out, "%lu seconds ago\n",
 335                                 (get_seconds() - os->os_scantime.tv_sec));
 336
 337         out += snprintf(buf + out, len - out, "%10s => %3s  %10s\n",
 338                         "Slots", "Num", "RecoGen");
 339         for (i = 0; i < osb->max_slots; ++i) {
 340                 out += snprintf(buf + out, len - out,
 341                                 "%10s  %c %3d  %10d\n",
 342                                 " ",
 343                                 (i == osb->slot_num ? '*' : ' '),
 344                                 i, osb->slot_recovery_generations[i]);
 345         }
 346
 347         return out;
 348 }
 349
 350 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 351 {
 352         struct ocfs2_super *osb = inode->i_private;
 353         char *buf = NULL;
 354
 355         buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
 356         if (!buf)
 357                 goto bail;
 358
 359         i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
 360
 361         file->private_data = buf;
 362
 363         return 0;
 364 bail:
 365         return -ENOMEM;
 366 }
 367
 368 static int ocfs2_debug_release(struct inode *inode, struct file *file)
 369 {
 370         kfree(file->private_data);
 371         return 0;
 372 }
 373
 374 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 375                                 size_t nbytes, loff_t *ppos)
 376 {
 377         return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
 378                                        i_size_read(file->f_mapping->host));
 379 }
 380 #else
 381 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 382 {
 383         return 0;
 384 }
 385 static int ocfs2_debug_release(struct inode *inode, struct file *file)
 386 {
 387         return 0;
 388 }
 389 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 390                                 size_t nbytes, loff_t *ppos)
 391 {
 392         return 0;
 393 }
 394 #endif  /* CONFIG_DEBUG_FS */
 395
 396 static const struct file_operations ocfs2_osb_debug_fops = {
 397         .open =         ocfs2_osb_debug_open,
 398         .release =      ocfs2_debug_release,
 399         .read =         ocfs2_debug_read,
 400         .llseek =       generic_file_llseek,
 401 };
 402
 403 static int ocfs2_sync_fs(struct super_block *sb, int wait)
 404 {
 405         int status;
 406         tid_t target;
 407         struct ocfs2_super *osb = OCFS2_SB(sb);
 408
 409         if (ocfs2_is_hard_readonly(osb))
 410                 return -EROFS;
 411
 412         if (wait) {
 413                 status = ocfs2_flush_truncate_log(osb);
 414                 if (status < 0)
 415                         mlog_errno(status);
 416         } else {
 417                 ocfs2_schedule_truncate_log_flush(osb, 0);
 418         }
 419
 420         if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
 421                                       &target)) {
 422                 if (wait)
 423                         jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
 424                                              target);
 425         }
 426         return 0;
 427 }
 428
 429 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
 430 {
 431         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
 432             && (ino == USER_QUOTA_SYSTEM_INODE
 433                 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
 434                 return 0;
 435         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
 436             && (ino == GROUP_QUOTA_SYSTEM_INODE
 437                 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
 438                 return 0;
 439         return 1;
 440 }
 441
 442 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
 443 {
 444         struct inode *new = NULL;
 445         int status = 0;
 446         int i;
 447
 448         new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 449         if (IS_ERR(new)) {
 450                 status = PTR_ERR(new);
 451                 mlog_errno(status);
 452                 goto bail;
 453         }
 454         osb->root_inode = new;
 455
 456         new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 457         if (IS_ERR(new)) {
 458                 status = PTR_ERR(new);
 459                 mlog_errno(status);
 460                 goto bail;
 461         }
 462         osb->sys_root_inode = new;
 463
 464         for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
 465              i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
 466                 if (!ocfs2_need_system_inode(osb, i))
 467                         continue;
 468                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 469                 if (!new) {
 470                         ocfs2_release_system_inodes(osb);
 471                         status = -EINVAL;
 472                         mlog_errno(status);
 473                         /* FIXME: Should ERROR_RO_FS */
 474                         mlog(ML_ERROR, "Unable to load system inode %d, "
 475                              "possibly corrupt fs?", i);
 476                         goto bail;
 477                 }
 478                 // the array now has one ref, so drop this one
 479                 iput(new);
 480         }
 481
 482 bail:
 483         if (status)
 484                 mlog_errno(status);
 485         return status;
 486 }
 487
 488 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
 489 {
 490         struct inode *new = NULL;
 491         int status = 0;
 492         int i;
 493
 494         for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
 495              i < NUM_SYSTEM_INODES;
 496              i++) {
 497                 if (!ocfs2_need_system_inode(osb, i))
 498                         continue;
 499                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 500                 if (!new) {
 501                         ocfs2_release_system_inodes(osb);
 502                         status = -EINVAL;
 503                         mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
 504                              status, i, osb->slot_num);
 505                         goto bail;
 506                 }
 507                 /* the array now has one ref, so drop this one */
 508                 iput(new);
 509         }
 510
 511 bail:
 512         if (status)
 513                 mlog_errno(status);
 514         return status;
 515 }
 516
 517 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
 518 {
 519         int i;
 520         struct inode *inode;
 521
 522         for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
 523                 inode = osb->global_system_inodes[i];
 524                 if (inode) {
 525                         iput(inode);
 526                         osb->global_system_inodes[i] = NULL;
 527                 }
 528         }
 529
 530         inode = osb->sys_root_inode;
 531         if (inode) {
 532                 iput(inode);
 533                 osb->sys_root_inode = NULL;
 534         }
 535
 536         inode = osb->root_inode;
 537         if (inode) {
 538                 iput(inode);
 539                 osb->root_inode = NULL;
 540         }
 541
 542         if (!osb->local_system_inodes)
 543                 return;
 544
 545         for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
 546                 if (osb->local_system_inodes[i]) {
 547                         iput(osb->local_system_inodes[i]);
 548                         osb->local_system_inodes[i] = NULL;
 549                 }
 550         }
 551
 552         kfree(osb->local_system_inodes);
 553         osb->local_system_inodes = NULL;
 554 }
 555
 556 /* We're allocating fs objects, use GFP_NOFS */
 557 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
 558 {
 559         struct ocfs2_inode_info *oi;
 560
 561         oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
 562         if (!oi)
 563                 return NULL;
 564
 565         jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
 566         return &oi->vfs_inode;
 567 }
 568
 569 static void ocfs2_i_callback(struct rcu_head *head)
 570 {
 571         struct inode *inode = container_of(head, struct inode, i_rcu);
 572         kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
 573 }
 574
 575 static void ocfs2_destroy_inode(struct inode *inode)
 576 {
 577         call_rcu(&inode->i_rcu, ocfs2_i_callback);
 578 }
 579
 580 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
 581                                                 unsigned int cbits)
 582 {
 583         unsigned int bytes = 1 << cbits;
 584         unsigned int trim = bytes;
 585         unsigned int bitshift = 32;
 586
 587         /*
 588          * i_size and all block offsets in ocfs2 are always 64 bits
 589          * wide. i_clusters is 32 bits, in cluster-sized units. So on
 590          * 64 bit platforms, cluster size will be the limiting factor.
 591          */
 592
 593 #if BITS_PER_LONG == 32
 594 # if defined(CONFIG_LBDAF)
 595         BUILD_BUG_ON(sizeof(sector_t) != 8);
 596         /*
 597          * We might be limited by page cache size.
 598          */
 599         if (bytes > PAGE_CACHE_SIZE) {
 600                 bytes = PAGE_CACHE_SIZE;
 601                 trim = 1;
 602                 /*
 603                  * Shift by 31 here so that we don't get larger than
 604                  * MAX_LFS_FILESIZE
 605                  */
 606                 bitshift = 31;
 607         }
 608 # else
 609         /*
 610          * We are limited by the size of sector_t. Use block size, as
 611          * that's what we expose to the VFS.
 612          */
 613         bytes = 1 << bbits;
 614         trim = 1;
 615         bitshift = 31;
 616 # endif
 617 #endif
 618
 619         /*
 620          * Trim by a whole cluster when we can actually approach the
 621          * on-disk limits. Otherwise we can overflow i_clusters when
 622          * an extent start is at the max offset.
 623          */
 624         return (((unsigned long long)bytes) << bitshift) - trim;
 625 }
 626
 627 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
 628 {
 629         int incompat_features;
 630         int ret = 0;
 631         struct mount_options parsed_options;
 632         struct ocfs2_super *osb = OCFS2_SB(sb);
 633         u32 tmp;
 634
 635         if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
 636             !ocfs2_check_set_options(sb, &parsed_options)) {
 637                 ret = -EINVAL;
 638                 goto out;
 639         }
 640
 641         tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
 642                 OCFS2_MOUNT_HB_NONE;
 643         if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
 644                 ret = -EINVAL;
 645                 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
 646                 goto out;
 647         }
 648
 649         if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
 650             (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
 651                 ret = -EINVAL;
 652                 mlog(ML_ERROR, "Cannot change data mode on remount\n");
 653                 goto out;
 654         }
 655
 656         /* Probably don't want this on remount; it might
 657          * mess with other nodes */
 658         if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
 659             (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
 660                 ret = -EINVAL;
 661                 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
 662                 goto out;
 663         }
 664
 665         /* We're going to/from readonly mode. */
 666         if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
 667                 /* Disable quota accounting before remounting RO */
 668                 if (*flags & MS_RDONLY) {
 669                         ret = ocfs2_susp_quotas(osb, 0);
 670                         if (ret < 0)
 671                                 goto out;
 672                 }
 673                 /* Lock here so the check of HARD_RO and the potential
 674                  * setting of SOFT_RO is atomic. */
 675                 spin_lock(&osb->osb_lock);
 676                 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
 677                         mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
 678                         ret = -EROFS;
 679                         goto unlock_osb;
 680                 }
 681
 682                 if (*flags & MS_RDONLY) {
 683                         sb->s_flags |= MS_RDONLY;
 684                         osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 685                 } else {
 686                         if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
 687                                 mlog(ML_ERROR, "Cannot remount RDWR "
 688                                      "filesystem due to previous errors.\n");
 689                                 ret = -EROFS;
 690                                 goto unlock_osb;
 691                         }
 692                         incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
 693                         if (incompat_features) {
 694                                 mlog(ML_ERROR, "Cannot remount RDWR because "
 695                                      "of unsupported optional features "
 696                                      "(%x).\n", incompat_features);
 697                                 ret = -EINVAL;
 698                                 goto unlock_osb;
 699                         }
 700                         sb->s_flags &= ~MS_RDONLY;
 701                         osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
 702                 }
 703                 trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags);
 704 unlock_osb:
 705                 spin_unlock(&osb->osb_lock);
 706                 /* Enable quota accounting after remounting RW */
 707                 if (!ret && !(*flags & MS_RDONLY)) {
 708                         if (sb_any_quota_suspended(sb))
 709                                 ret = ocfs2_susp_quotas(osb, 1);
 710                         else
 711                                 ret = ocfs2_enable_quotas(osb);
 712                         if (ret < 0) {
 713                                 /* Return back changes... */
 714                                 spin_lock(&osb->osb_lock);
 715                                 sb->s_flags |= MS_RDONLY;
 716                                 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 717                                 spin_unlock(&osb->osb_lock);
 718                                 goto out;
 719                         }
 720                 }
 721         }
 722
 723         if (!ret) {
 724                 /* Only save off the new mount options in case of a successful
 725                  * remount. */
 726                 osb->s_mount_opt = parsed_options.mount_opt;
 727                 osb->s_atime_quantum = parsed_options.atime_quantum;
 728                 osb->preferred_slot = parsed_options.slot;
 729                 if (parsed_options.commit_interval)
 730                         osb->osb_commit_interval = parsed_options.commit_interval;
 731
 732                 if (!ocfs2_is_hard_readonly(osb))
 733                         ocfs2_set_journal_params(osb);
 734
 735                 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 736                         ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
 737                                                         MS_POSIXACL : 0);
 738         }
 739 out:
 740         return ret;
 741 }
 742
 743 static int ocfs2_sb_probe(struct super_block *sb,
 744                           struct buffer_head **bh,
 745                           int *sector_size,
 746                           struct ocfs2_blockcheck_stats *stats)
 747 {
 748         int status, tmpstat;
 749         struct ocfs1_vol_disk_hdr *hdr;
 750         struct ocfs2_dinode *di;
 751         int blksize;
 752
 753         *bh = NULL;
 754
 755         /* may be > 512 */
 756         *sector_size = bdev_logical_block_size(sb->s_bdev);
 757         if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
 758                 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
 759                      *sector_size, OCFS2_MAX_BLOCKSIZE);
 760                 status = -EINVAL;
 761                 goto bail;
 762         }
 763
 764         /* Can this really happen? */
 765         if (*sector_size < OCFS2_MIN_BLOCKSIZE)
 766                 *sector_size = OCFS2_MIN_BLOCKSIZE;
 767
 768         /* check block zero for old format */
 769         status = ocfs2_get_sector(sb, bh, 0, *sector_size);
 770         if (status < 0) {
 771                 mlog_errno(status);
 772                 goto bail;
 773         }
 774         hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
 775         if (hdr->major_version == OCFS1_MAJOR_VERSION) {
 776                 mlog(ML_ERROR, "incompatible version: %u.%u\n",
 777                      hdr->major_version, hdr->minor_version);
 778                 status = -EINVAL;
 779         }
 780         if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
 781                    strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
 782                 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
 783                      hdr->signature);
 784                 status = -EINVAL;
 785         }
 786         brelse(*bh);
 787         *bh = NULL;
 788         if (status < 0) {
 789                 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
 790                      "upgraded before mounting with ocfs v2\n");
 791                 goto bail;
 792         }
 793
 794         /*
 795          * Now check at magic offset for 512, 1024, 2048, 4096
 796          * blocksizes.  4096 is the maximum blocksize because it is
 797          * the minimum clustersize.
 798          */
 799         status = -EINVAL;
 800         for (blksize = *sector_size;
 801              blksize <= OCFS2_MAX_BLOCKSIZE;
 802              blksize <<= 1) {
 803                 tmpstat = ocfs2_get_sector(sb, bh,
 804                                            OCFS2_SUPER_BLOCK_BLKNO,
 805                                            blksize);
 806                 if (tmpstat < 0) {
 807                         status = tmpstat;
 808                         mlog_errno(status);
 809                         break;
 810                 }
 811                 di = (struct ocfs2_dinode *) (*bh)->b_data;
 812                 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
 813                 spin_lock_init(&stats->b_lock);
 814                 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
 815                 if (tmpstat < 0) {
 816                         brelse(*bh);
 817                         *bh = NULL;
 818                 }
 819                 if (tmpstat != -EAGAIN) {
 820                         status = tmpstat;
 821                         break;
 822                 }
 823         }
 824
 825 bail:
 826         return status;
 827 }
 828
 829 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
 830 {
 831         u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
 832
 833         if (osb->s_mount_opt & hb_enabled) {
 834                 if (ocfs2_mount_local(osb)) {
 835                         mlog(ML_ERROR, "Cannot heartbeat on a locally "
 836                              "mounted device.\n");
 837                         return -EINVAL;
 838                 }
 839                 if (ocfs2_userspace_stack(osb)) {
 840                         mlog(ML_ERROR, "Userspace stack expected, but "
 841                              "o2cb heartbeat arguments passed to mount\n");
 842                         return -EINVAL;
 843                 }
 844                 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
 845                      !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
 846                     ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
 847                      ocfs2_cluster_o2cb_global_heartbeat(osb))) {
 848                         mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
 849                         return -EINVAL;
 850                 }
 851         }
 852
 853         if (!(osb->s_mount_opt & hb_enabled)) {
 854                 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
 855                     !ocfs2_userspace_stack(osb)) {
 856                         mlog(ML_ERROR, "Heartbeat has to be started to mount "
 857                              "a read-write clustered device.\n");
 858                         return -EINVAL;
 859                 }
 860         }
 861
 862         return 0;
 863 }
 864
 865 /*
 866  * If we're using a userspace stack, mount should have passed
 867  * a name that matches the disk.  If not, mount should not
 868  * have passed a stack.
 869  */
 870 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
 871                                         struct mount_options *mopt)
 872 {
 873         if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
 874                 mlog(ML_ERROR,
 875                      "cluster stack passed to mount, but this filesystem "
 876                      "does not support it\n");
 877                 return -EINVAL;
 878         }
 879
 880         if (ocfs2_userspace_stack(osb) &&
 881             strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
 882                     OCFS2_STACK_LABEL_LEN)) {
 883                 mlog(ML_ERROR,
 884                      "cluster stack passed to mount (\"%s\") does not "
 885                      "match the filesystem (\"%s\")\n",
 886                      mopt->cluster_stack,
 887                      osb->osb_cluster_stack);
 888                 return -EINVAL;
 889         }
 890
 891         return 0;
 892 }
 893
 894 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
 895 {
 896         int type;
 897         struct super_block *sb = osb->sb;
 898         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 899                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 900         int status = 0;
 901
 902         for (type = 0; type < MAXQUOTAS; type++) {
 903                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 904                         continue;
 905                 if (unsuspend)
 906                         status = dquot_resume(sb, type);
 907                 else {
 908                         struct ocfs2_mem_dqinfo *oinfo;
 909
 910                         /* Cancel periodic syncing before suspending */
 911                         oinfo = sb_dqinfo(sb, type)->dqi_priv;
 912                         cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 913                         status = dquot_suspend(sb, type);
 914                 }
 915                 if (status < 0)
 916                         break;
 917         }
 918         if (status < 0)
 919                 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
 920                      "remount (error = %d).\n", status);
 921         return status;
 922 }
 923
 924 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
 925 {
 926         struct inode *inode[MAXQUOTAS] = { NULL, NULL };
 927         struct super_block *sb = osb->sb;
 928         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 929                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 930         unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
 931                                         LOCAL_GROUP_QUOTA_SYSTEM_INODE };
 932         int status;
 933         int type;
 934
 935         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
 936         for (type = 0; type < MAXQUOTAS; type++) {
 937                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 938                         continue;
 939                 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
 940                                                         osb->slot_num);
 941                 if (!inode[type]) {
 942                         status = -ENOENT;
 943                         goto out_quota_off;
 944                 }
 945                 status = dquot_enable(inode[type], type, QFMT_OCFS2,
 946                                       DQUOT_USAGE_ENABLED);
 947                 if (status < 0)
 948                         goto out_quota_off;
 949         }
 950
 951         for (type = 0; type < MAXQUOTAS; type++)
 952                 iput(inode[type]);
 953         return 0;
 954 out_quota_off:
 955         ocfs2_disable_quotas(osb);
 956         for (type = 0; type < MAXQUOTAS; type++)
 957                 iput(inode[type]);
 958         mlog_errno(status);
 959         return status;
 960 }
 961
 962 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
 963 {
 964         int type;
 965         struct inode *inode;
 966         struct super_block *sb = osb->sb;
 967         struct ocfs2_mem_dqinfo *oinfo;
 968
 969         /* We mostly ignore errors in this function because there's not much
 970          * we can do when we see them */
 971         for (type = 0; type < MAXQUOTAS; type++) {
 972                 if (!sb_has_quota_loaded(sb, type))
 973                         continue;
 974                 /* Cancel periodic syncing before we grab dqonoff_mutex */
 975                 oinfo = sb_dqinfo(sb, type)->dqi_priv;
 976                 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 977                 inode = igrab(sb->s_dquot.files[type]);
 978                 /* Turn off quotas. This will remove all dquot structures from
 979                  * memory and so they will be automatically synced to global
 980                  * quota files */
 981                 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
 982                                         DQUOT_LIMITS_ENABLED);
 983                 if (!inode)
 984                         continue;
 985                 iput(inode);
 986         }
 987 }
 988
 989 /* Handle quota on quotactl */
 990 static int ocfs2_quota_on(struct super_block *sb, int type, int format_id)
 991 {
 992         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 993                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 994
 995         if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 996                 return -EINVAL;
 997
 998         return dquot_enable(sb_dqopt(sb)->files[type], type,
 999                             format_id, DQUOT_LIMITS_ENABLED);
1000 }
1001
1002 /* Handle quota off quotactl */
1003 static int ocfs2_quota_off(struct super_block *sb, int type)
1004 {
1005         return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
1006 }
1007
1008 static const struct quotactl_ops ocfs2_quotactl_ops = {
1009         .quota_on_meta  = ocfs2_quota_on,
1010         .quota_off      = ocfs2_quota_off,
1011         .quota_sync     = dquot_quota_sync,
1012         .get_info       = dquot_get_dqinfo,
1013         .set_info       = dquot_set_dqinfo,
1014         .get_dqblk      = dquot_get_dqblk,
1015         .set_dqblk      = dquot_set_dqblk,
1016 };
1017
1018 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
1019 {
1020         struct dentry *root;
1021         int status, sector_size;
1022         struct mount_options parsed_options;
1023         struct inode *inode = NULL;
1024         struct ocfs2_super *osb = NULL;
1025         struct buffer_head *bh = NULL;
1026         char nodestr[8];
1027         struct ocfs2_blockcheck_stats stats;
1028
1029         trace_ocfs2_fill_super(sb, data, silent);
1030
1031         if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
1032                 status = -EINVAL;
1033                 goto read_super_error;
1034         }
1035
1036         /* probe for superblock */
1037         status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
1038         if (status < 0) {
1039                 mlog(ML_ERROR, "superblock probe failed!\n");
1040                 goto read_super_error;
1041         }
1042
1043         status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1044         osb = OCFS2_SB(sb);
1045         if (status < 0) {
1046                 mlog_errno(status);
1047                 goto read_super_error;
1048         }
1049         brelse(bh);
1050         bh = NULL;
1051
1052         if (!ocfs2_check_set_options(sb, &parsed_options)) {
1053                 status = -EINVAL;
1054                 goto read_super_error;
1055         }
1056         osb->s_mount_opt = parsed_options.mount_opt;
1057         osb->s_atime_quantum = parsed_options.atime_quantum;
1058         osb->preferred_slot = parsed_options.slot;
1059         osb->osb_commit_interval = parsed_options.commit_interval;
1060
1061         ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1062         osb->osb_resv_level = parsed_options.resv_level;
1063         osb->osb_dir_resv_level = parsed_options.resv_level;
1064         if (parsed_options.dir_resv_level == -1)
1065                 osb->osb_dir_resv_level = parsed_options.resv_level;
1066         else
1067                 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1068
1069         status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1070         if (status)
1071                 goto read_super_error;
1072
1073         sb->s_magic = OCFS2_SUPER_MAGIC;
1074
1075         sb->s_flags = (sb->s_flags & ~(MS_POSIXACL | MS_NOSEC)) |
1076                 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1077
1078         /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1079          * heartbeat=none */
1080         if (bdev_read_only(sb->s_bdev)) {
1081                 if (!(sb->s_flags & MS_RDONLY)) {
1082                         status = -EACCES;
1083                         mlog(ML_ERROR, "Readonly device detected but readonly "
1084                              "mount was not specified.\n");
1085                         goto read_super_error;
1086                 }
1087
1088                 /* You should not be able to start a local heartbeat
1089                  * on a readonly device. */
1090                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1091                         status = -EROFS;
1092                         mlog(ML_ERROR, "Local heartbeat specified on readonly "
1093                              "device.\n");
1094                         goto read_super_error;
1095                 }
1096
1097                 status = ocfs2_check_journals_nolocks(osb);
1098                 if (status < 0) {
1099                         if (status == -EROFS)
1100                                 mlog(ML_ERROR, "Recovery required on readonly "
1101                                      "file system, but write access is "
1102                                      "unavailable.\n");
1103                         else
1104                                 mlog_errno(status);
1105                         goto read_super_error;
1106                 }
1107
1108                 ocfs2_set_ro_flag(osb, 1);
1109
1110                 printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. "
1111                        "Cluster services will not be used for this mount. "
1112                        "Recovery will be skipped.\n", osb->dev_str);
1113         }
1114
1115         if (!ocfs2_is_hard_readonly(osb)) {
1116                 if (sb->s_flags & MS_RDONLY)
1117                         ocfs2_set_ro_flag(osb, 0);
1118         }
1119
1120         status = ocfs2_verify_heartbeat(osb);
1121         if (status < 0) {
1122                 mlog_errno(status);
1123                 goto read_super_error;
1124         }
1125
1126         osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1127                                                  ocfs2_debugfs_root);
1128         if (!osb->osb_debug_root) {
1129                 status = -EINVAL;
1130                 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1131                 goto read_super_error;
1132         }
1133
1134         osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1135                                             osb->osb_debug_root,
1136                                             osb,
1137                                             &ocfs2_osb_debug_fops);
1138         if (!osb->osb_ctxt) {
1139                 status = -EINVAL;
1140                 mlog_errno(status);
1141                 goto read_super_error;
1142         }
1143
1144         if (ocfs2_meta_ecc(osb)) {
1145                 status = ocfs2_blockcheck_stats_debugfs_install(
1146                                                 &osb->osb_ecc_stats,
1147                                                 osb->osb_debug_root);
1148                 if (status) {
1149                         mlog(ML_ERROR,
1150                              "Unable to create blockcheck statistics "
1151                              "files\n");
1152                         goto read_super_error;
1153                 }
1154         }
1155
1156         status = ocfs2_mount_volume(sb);
1157         if (osb->root_inode)
1158                 inode = igrab(osb->root_inode);
1159
1160         if (status < 0)
1161                 goto read_super_error;
1162
1163         if (!inode) {
1164                 status = -EIO;
1165                 mlog_errno(status);
1166                 goto read_super_error;
1167         }
1168
1169         root = d_alloc_root(inode);
1170         if (!root) {
1171                 status = -ENOMEM;
1172                 mlog_errno(status);
1173                 goto read_super_error;
1174         }
1175
1176         sb->s_root = root;
1177
1178         ocfs2_complete_mount_recovery(osb);
1179
1180         if (ocfs2_mount_local(osb))
1181                 snprintf(nodestr, sizeof(nodestr), "local");
1182         else
1183                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1184
1185         printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1186                "with %s data mode.\n",
1187                osb->dev_str, nodestr, osb->slot_num,
1188                osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1189                "ordered");
1190
1191         atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1192         wake_up(&osb->osb_mount_event);
1193
1194         /* Now we can initialize quotas because we can afford to wait
1195          * for cluster locks recovery now. That also means that truncation
1196          * log recovery can happen but that waits for proper quota setup */
1197         if (!(sb->s_flags & MS_RDONLY)) {
1198                 status = ocfs2_enable_quotas(osb);
1199                 if (status < 0) {
1200                         /* We have to err-out specially here because
1201                          * s_root is already set */
1202                         mlog_errno(status);
1203                         atomic_set(&osb->vol_state, VOLUME_DISABLED);
1204                         wake_up(&osb->osb_mount_event);
1205                         return status;
1206                 }
1207         }
1208
1209         ocfs2_complete_quota_recovery(osb);
1210
1211         /* Now we wake up again for processes waiting for quotas */
1212         atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1213         wake_up(&osb->osb_mount_event);
1214
1215         /* Start this when the mount is almost sure of being successful */
1216         ocfs2_orphan_scan_start(osb);
1217
1218         return status;
1219
1220 read_super_error:
1221         brelse(bh);
1222
1223         if (inode)
1224                 iput(inode);
1225
1226         if (osb) {
1227                 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1228                 wake_up(&osb->osb_mount_event);
1229                 ocfs2_dismount_volume(sb, 1);
1230         }
1231
1232         if (status)
1233                 mlog_errno(status);
1234         return status;
1235 }
1236
1237 static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
1238                         int flags,
1239                         const char *dev_name,
1240                         void *data)
1241 {
1242         return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
1243 }
1244
1245 static void ocfs2_kill_sb(struct super_block *sb)
1246 {
1247         struct ocfs2_super *osb = OCFS2_SB(sb);
1248
1249         /* Failed mount? */
1250         if (!osb || atomic_read(&osb->vol_state) == VOLUME_DISABLED)
1251                 goto out;
1252
1253         /* Prevent further queueing of inode drop events */
1254         spin_lock(&dentry_list_lock);
1255         ocfs2_set_osb_flag(osb, OCFS2_OSB_DROP_DENTRY_LOCK_IMMED);
1256         spin_unlock(&dentry_list_lock);
1257         /* Wait for work to finish and/or remove it */
1258         cancel_work_sync(&osb->dentry_lock_work);
1259 out:
1260         kill_block_super(sb);
1261 }
1262
1263 static struct file_system_type ocfs2_fs_type = {
1264         .owner          = THIS_MODULE,
1265         .name           = "ocfs2",
1266         .mount          = ocfs2_mount,
1267         .kill_sb        = ocfs2_kill_sb,
1268
1269         .fs_flags       = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1270         .next           = NULL
1271 };
1272
1273 static int ocfs2_check_set_options(struct super_block *sb,
1274                                    struct mount_options *options)
1275 {
1276         if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1277             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1278                                          OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1279                 mlog(ML_ERROR, "User quotas were requested, but this "
1280                      "filesystem does not have the feature enabled.\n");
1281                 return 0;
1282         }
1283         if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1284             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1285                                          OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1286                 mlog(ML_ERROR, "Group quotas were requested, but this "
1287                      "filesystem does not have the feature enabled.\n");
1288                 return 0;
1289         }
1290         if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1291             !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1292                 mlog(ML_ERROR, "ACL support requested but extended attributes "
1293                      "feature is not enabled\n");
1294                 return 0;
1295         }
1296         /* No ACL setting specified? Use XATTR feature... */
1297         if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1298                                     OCFS2_MOUNT_NO_POSIX_ACL))) {
1299                 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1300                         options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1301                 else
1302                         options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1303         }
1304         return 1;
1305 }
1306
1307 static int ocfs2_parse_options(struct super_block *sb,
1308                                char *options,
1309                                struct mount_options *mopt,
1310                                int is_remount)
1311 {
1312         int status, user_stack = 0;
1313         char *p;
1314         u32 tmp;
1315
1316         trace_ocfs2_parse_options(is_remount, options ? options : "(none)");
1317
1318         mopt->commit_interval = 0;
1319         mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1320         mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1321         mopt->slot = OCFS2_INVALID_SLOT;
1322         mopt->localalloc_opt = -1;
1323         mopt->cluster_stack[0] = '\0';
1324         mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1325         mopt->dir_resv_level = -1;
1326
1327         if (!options) {
1328                 status = 1;
1329                 goto bail;
1330         }
1331
1332         while ((p = strsep(&options, ",")) != NULL) {
1333                 int token, option;
1334                 substring_t args[MAX_OPT_ARGS];
1335
1336                 if (!*p)
1337                         continue;
1338
1339                 token = match_token(p, tokens, args);
1340                 switch (token) {
1341                 case Opt_hb_local:
1342                         mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1343                         break;
1344                 case Opt_hb_none:
1345                         mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1346                         break;
1347                 case Opt_hb_global:
1348                         mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1349                         break;
1350                 case Opt_barrier:
1351                         if (match_int(&args[0], &option)) {
1352                                 status = 0;
1353                                 goto bail;
1354                         }
1355                         if (option)
1356                                 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1357                         else
1358                                 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1359                         break;
1360                 case Opt_intr:
1361                         mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1362                         break;
1363                 case Opt_nointr:
1364                         mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1365                         break;
1366                 case Opt_err_panic:
1367                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1368                         break;
1369                 case Opt_err_ro:
1370                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1371                         break;
1372                 case Opt_data_ordered:
1373                         mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1374                         break;
1375                 case Opt_data_writeback:
1376                         mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1377                         break;
1378                 case Opt_user_xattr:
1379                         mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1380                         break;
1381                 case Opt_nouser_xattr:
1382                         mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1383                         break;
1384                 case Opt_atime_quantum:
1385                         if (match_int(&args[0], &option)) {
1386                                 status = 0;
1387                                 goto bail;
1388                         }
1389                         if (option >= 0)
1390                                 mopt->atime_quantum = option;
1391                         break;
1392                 case Opt_slot:
1393                         option = 0;
1394                         if (match_int(&args[0], &option)) {
1395                                 status = 0;
1396                                 goto bail;
1397                         }
1398                         if (option)
1399                                 mopt->slot = (s16)option;
1400                         break;
1401                 case Opt_commit:
1402                         option = 0;
1403                         if (match_int(&args[0], &option)) {
1404                                 status = 0;
1405                                 goto bail;
1406                         }
1407                         if (option < 0)
1408                                 return 0;
1409                         if (option == 0)
1410                                 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1411                         mopt->commit_interval = HZ * option;
1412                         break;
1413                 case Opt_localalloc:
1414                         option = 0;
1415                         if (match_int(&args[0], &option)) {
1416                                 status = 0;
1417                                 goto bail;
1418                         }
1419                         if (option >= 0)
1420                                 mopt->localalloc_opt = option;
1421                         break;
1422                 case Opt_localflocks:
1423                         /*
1424                          * Changing this during remount could race
1425                          * flock() requests, or "unbalance" existing
1426                          * ones (e.g., a lock is taken in one mode but
1427                          * dropped in the other). If users care enough
1428                          * to flip locking modes during remount, we
1429                          * could add a "local" flag to individual
1430                          * flock structures for proper tracking of
1431                          * state.
1432                          */
1433                         if (!is_remount)
1434                                 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1435                         break;
1436                 case Opt_stack:
1437                         /* Check both that the option we were passed
1438                          * is of the right length and that it is a proper
1439                          * string of the right length.
1440                          */
1441                         if (((args[0].to - args[0].from) !=
1442                              OCFS2_STACK_LABEL_LEN) ||
1443                             (strnlen(args[0].from,
1444                                      OCFS2_STACK_LABEL_LEN) !=
1445                              OCFS2_STACK_LABEL_LEN)) {
1446                                 mlog(ML_ERROR,
1447                                      "Invalid cluster_stack option\n");
1448                                 status = 0;
1449                                 goto bail;
1450                         }
1451                         memcpy(mopt->cluster_stack, args[0].from,
1452                                OCFS2_STACK_LABEL_LEN);
1453                         mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1454                         /*
1455                          * Open code the memcmp here as we don't have
1456                          * an osb to pass to
1457                          * ocfs2_userspace_stack().
1458                          */
1459                         if (memcmp(mopt->cluster_stack,
1460                                    OCFS2_CLASSIC_CLUSTER_STACK,
1461                                    OCFS2_STACK_LABEL_LEN))
1462                                 user_stack = 1;
1463                         break;
1464                 case Opt_inode64:
1465                         mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1466                         break;
1467                 case Opt_usrquota:
1468                         mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1469                         break;
1470                 case Opt_grpquota:
1471                         mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1472                         break;
1473                 case Opt_coherency_buffered:
1474                         mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1475                         break;
1476                 case Opt_coherency_full:
1477                         mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1478                         break;
1479                 case Opt_acl:
1480                         mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1481                         mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1482                         break;
1483                 case Opt_noacl:
1484                         mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1485                         mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1486                         break;
1487                 case Opt_resv_level:
1488                         if (is_remount)
1489                                 break;
1490                         if (match_int(&args[0], &option)) {
1491                                 status = 0;
1492                                 goto bail;
1493                         }
1494                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1495                             option < OCFS2_MAX_RESV_LEVEL)
1496                                 mopt->resv_level = option;
1497                         break;
1498                 case Opt_dir_resv_level:
1499                         if (is_remount)
1500                                 break;
1501                         if (match_int(&args[0], &option)) {
1502                                 status = 0;
1503                                 goto bail;
1504                         }
1505                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1506                             option < OCFS2_MAX_RESV_LEVEL)
1507                                 mopt->dir_resv_level = option;
1508                         break;
1509                 default:
1510                         mlog(ML_ERROR,
1511                              "Unrecognized mount option \"%s\" "
1512                              "or missing value\n", p);
1513                         status = 0;
1514                         goto bail;
1515                 }
1516         }
1517
1518         if (user_stack == 0) {
1519                 /* Ensure only one heartbeat mode */
1520                 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1521                                          OCFS2_MOUNT_HB_GLOBAL |
1522                                          OCFS2_MOUNT_HB_NONE);
1523                 if (hweight32(tmp) != 1) {
1524                         mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1525                         status = 0;
1526                         goto bail;
1527                 }
1528         }
1529
1530         status = 1;
1531
1532 bail:
1533         return status;
1534 }
1535
1536 static int ocfs2_show_options(struct seq_file *s, struct dentry *root)
1537 {
1538         struct ocfs2_super *osb = OCFS2_SB(root->d_sb);
1539         unsigned long opts = osb->s_mount_opt;
1540         unsigned int local_alloc_megs;
1541
1542         if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1543                 seq_printf(s, ",_netdev");
1544                 if (opts & OCFS2_MOUNT_HB_LOCAL)
1545                         seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1546                 else
1547                         seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1548         } else
1549                 seq_printf(s, ",%s", OCFS2_HB_NONE);
1550
1551         if (opts & OCFS2_MOUNT_NOINTR)
1552                 seq_printf(s, ",nointr");
1553
1554         if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1555                 seq_printf(s, ",data=writeback");
1556         else
1557                 seq_printf(s, ",data=ordered");
1558
1559         if (opts & OCFS2_MOUNT_BARRIER)
1560                 seq_printf(s, ",barrier=1");
1561
1562         if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1563                 seq_printf(s, ",errors=panic");
1564         else
1565                 seq_printf(s, ",errors=remount-ro");
1566
1567         if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1568                 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1569
1570         seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1571
1572         if (osb->osb_commit_interval)
1573                 seq_printf(s, ",commit=%u",
1574                            (unsigned) (osb->osb_commit_interval / HZ));
1575
1576         local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1577         if (local_alloc_megs != ocfs2_la_default_mb(osb))
1578                 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1579
1580         if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1581                 seq_printf(s, ",localflocks,");
1582
1583         if (osb->osb_cluster_stack[0])
1584                 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1585                            osb->osb_cluster_stack);
1586         if (opts & OCFS2_MOUNT_USRQUOTA)
1587                 seq_printf(s, ",usrquota");
1588         if (opts & OCFS2_MOUNT_GRPQUOTA)
1589                 seq_printf(s, ",grpquota");
1590
1591         if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1592                 seq_printf(s, ",coherency=buffered");
1593         else
1594                 seq_printf(s, ",coherency=full");
1595
1596         if (opts & OCFS2_MOUNT_NOUSERXATTR)
1597                 seq_printf(s, ",nouser_xattr");
1598         else
1599                 seq_printf(s, ",user_xattr");
1600
1601         if (opts & OCFS2_MOUNT_INODE64)
1602                 seq_printf(s, ",inode64");
1603
1604         if (opts & OCFS2_MOUNT_POSIX_ACL)
1605                 seq_printf(s, ",acl");
1606         else
1607                 seq_printf(s, ",noacl");
1608
1609         if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1610                 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1611
1612         if (osb->osb_dir_resv_level != osb->osb_resv_level)
1613                 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1614
1615         return 0;
1616 }
1617
1618 wait_queue_head_t ocfs2__ioend_wq[OCFS2_IOEND_WQ_HASH_SZ];
1619
1620 static int __init ocfs2_init(void)
1621 {
1622         int status, i;
1623
1624         ocfs2_print_version();
1625
1626         for (i = 0; i < OCFS2_IOEND_WQ_HASH_SZ; i++)
1627                 init_waitqueue_head(&ocfs2__ioend_wq[i]);
1628
1629         status = init_ocfs2_uptodate_cache();
1630         if (status < 0) {
1631                 mlog_errno(status);
1632                 goto leave;
1633         }
1634
1635         status = ocfs2_initialize_mem_caches();
1636         if (status < 0) {
1637                 mlog_errno(status);
1638                 goto leave;
1639         }
1640
1641         ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1642         if (!ocfs2_wq) {
1643                 status = -ENOMEM;
1644                 goto leave;
1645         }
1646
1647         ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1648         if (!ocfs2_debugfs_root) {
1649                 status = -EFAULT;
1650                 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1651         }
1652
1653         ocfs2_set_locking_protocol();
1654
1655         status = register_quota_format(&ocfs2_quota_format);
1656 leave:
1657         if (status < 0) {
1658                 ocfs2_free_mem_caches();
1659                 exit_ocfs2_uptodate_cache();
1660                 mlog_errno(status);
1661         }
1662
1663         if (status >= 0) {
1664                 return register_filesystem(&ocfs2_fs_type);
1665         } else
1666                 return -1;
1667 }
1668
1669 static void __exit ocfs2_exit(void)
1670 {
1671         if (ocfs2_wq) {
1672                 flush_workqueue(ocfs2_wq);
1673                 destroy_workqueue(ocfs2_wq);
1674         }
1675
1676         unregister_quota_format(&ocfs2_quota_format);
1677
1678         debugfs_remove(ocfs2_debugfs_root);
1679
1680         ocfs2_free_mem_caches();
1681
1682         unregister_filesystem(&ocfs2_fs_type);
1683
1684         exit_ocfs2_uptodate_cache();
1685 }
1686
1687 static void ocfs2_put_super(struct super_block *sb)
1688 {
1689         trace_ocfs2_put_super(sb);
1690
1691         ocfs2_sync_blockdev(sb);
1692         ocfs2_dismount_volume(sb, 0);
1693 }
1694
1695 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1696 {
1697         struct ocfs2_super *osb;
1698         u32 numbits, freebits;
1699         int status;
1700         struct ocfs2_dinode *bm_lock;
1701         struct buffer_head *bh = NULL;
1702         struct inode *inode = NULL;
1703
1704         trace_ocfs2_statfs(dentry->d_sb, buf);
1705
1706         osb = OCFS2_SB(dentry->d_sb);
1707
1708         inode = ocfs2_get_system_file_inode(osb,
1709                                             GLOBAL_BITMAP_SYSTEM_INODE,
1710                                             OCFS2_INVALID_SLOT);
1711         if (!inode) {
1712                 mlog(ML_ERROR, "failed to get bitmap inode\n");
1713                 status = -EIO;
1714                 goto bail;
1715         }
1716
1717         status = ocfs2_inode_lock(inode, &bh, 0);
1718         if (status < 0) {
1719                 mlog_errno(status);
1720                 goto bail;
1721         }
1722
1723         bm_lock = (struct ocfs2_dinode *) bh->b_data;
1724
1725         numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1726         freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1727
1728         buf->f_type = OCFS2_SUPER_MAGIC;
1729         buf->f_bsize = dentry->d_sb->s_blocksize;
1730         buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1731         buf->f_blocks = ((sector_t) numbits) *
1732                         (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1733         buf->f_bfree = ((sector_t) freebits) *
1734                        (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1735         buf->f_bavail = buf->f_bfree;
1736         buf->f_files = numbits;
1737         buf->f_ffree = freebits;
1738         buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1739                                 & 0xFFFFFFFFUL;
1740         buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1741                                 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1742
1743         brelse(bh);
1744
1745         ocfs2_inode_unlock(inode, 0);
1746         status = 0;
1747 bail:
1748         if (inode)
1749                 iput(inode);
1750
1751         if (status)
1752                 mlog_errno(status);
1753
1754         return status;
1755 }
1756
1757 static void ocfs2_inode_init_once(void *data)
1758 {
1759         struct ocfs2_inode_info *oi = data;
1760
1761         oi->ip_flags = 0;
1762         oi->ip_open_count = 0;
1763         spin_lock_init(&oi->ip_lock);
1764         ocfs2_extent_map_init(&oi->vfs_inode);
1765         INIT_LIST_HEAD(&oi->ip_io_markers);
1766         oi->ip_dir_start_lookup = 0;
1767         atomic_set(&oi->ip_unaligned_aio, 0);
1768         init_rwsem(&oi->ip_alloc_sem);
1769         init_rwsem(&oi->ip_xattr_sem);
1770         mutex_init(&oi->ip_io_mutex);
1771
1772         oi->ip_blkno = 0ULL;
1773         oi->ip_clusters = 0;
1774
1775         ocfs2_resv_init_once(&oi->ip_la_data_resv);
1776
1777         ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1778         ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1779         ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1780
1781         ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1782                                   &ocfs2_inode_caching_ops);
1783
1784         inode_init_once(&oi->vfs_inode);
1785 }
1786
1787 static int ocfs2_initialize_mem_caches(void)
1788 {
1789         ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1790                                        sizeof(struct ocfs2_inode_info),
1791                                        0,
1792                                        (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1793                                                 SLAB_MEM_SPREAD),
1794                                        ocfs2_inode_init_once);
1795         ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1796                                         sizeof(struct ocfs2_dquot),
1797                                         0,
1798                                         (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1799                                                 SLAB_MEM_SPREAD),
1800                                         NULL);
1801         ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1802                                         sizeof(struct ocfs2_quota_chunk),
1803                                         0,
1804                                         (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1805                                         NULL);
1806         if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1807             !ocfs2_qf_chunk_cachep) {
1808                 if (ocfs2_inode_cachep)
1809                         kmem_cache_destroy(ocfs2_inode_cachep);
1810                 if (ocfs2_dquot_cachep)
1811                         kmem_cache_destroy(ocfs2_dquot_cachep);
1812                 if (ocfs2_qf_chunk_cachep)
1813                         kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1814                 return -ENOMEM;
1815         }
1816
1817         return 0;
1818 }
1819
1820 static void ocfs2_free_mem_caches(void)
1821 {
1822         if (ocfs2_inode_cachep)
1823                 kmem_cache_destroy(ocfs2_inode_cachep);
1824         ocfs2_inode_cachep = NULL;
1825
1826         if (ocfs2_dquot_cachep)
1827                 kmem_cache_destroy(ocfs2_dquot_cachep);
1828         ocfs2_dquot_cachep = NULL;
1829
1830         if (ocfs2_qf_chunk_cachep)
1831                 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1832         ocfs2_qf_chunk_cachep = NULL;
1833 }
1834
1835 static int ocfs2_get_sector(struct super_block *sb,
1836                             struct buffer_head **bh,
1837                             int block,
1838                             int sect_size)
1839 {
1840         if (!sb_set_blocksize(sb, sect_size)) {
1841                 mlog(ML_ERROR, "unable to set blocksize\n");
1842                 return -EIO;
1843         }
1844
1845         *bh = sb_getblk(sb, block);
1846         if (!*bh) {
1847                 mlog_errno(-EIO);
1848                 return -EIO;
1849         }
1850         lock_buffer(*bh);
1851         if (!buffer_dirty(*bh))
1852                 clear_buffer_uptodate(*bh);
1853         unlock_buffer(*bh);
1854         ll_rw_block(READ, 1, bh);
1855         wait_on_buffer(*bh);
1856         if (!buffer_uptodate(*bh)) {
1857                 mlog_errno(-EIO);
1858                 brelse(*bh);
1859                 *bh = NULL;
1860                 return -EIO;
1861         }
1862
1863         return 0;
1864 }
1865
1866 static int ocfs2_mount_volume(struct super_block *sb)
1867 {
1868         int status = 0;
1869         int unlock_super = 0;
1870         struct ocfs2_super *osb = OCFS2_SB(sb);
1871
1872         if (ocfs2_is_hard_readonly(osb))
1873                 goto leave;
1874
1875         status = ocfs2_dlm_init(osb);
1876         if (status < 0) {
1877                 mlog_errno(status);
1878                 goto leave;
1879         }
1880
1881         status = ocfs2_super_lock(osb, 1);
1882         if (status < 0) {
1883                 mlog_errno(status);
1884                 goto leave;
1885         }
1886         unlock_super = 1;
1887
1888         /* This will load up the node map and add ourselves to it. */
1889         status = ocfs2_find_slot(osb);
1890         if (status < 0) {
1891                 mlog_errno(status);
1892                 goto leave;
1893         }
1894
1895         /* load all node-local system inodes */
1896         status = ocfs2_init_local_system_inodes(osb);
1897         if (status < 0) {
1898                 mlog_errno(status);
1899                 goto leave;
1900         }
1901
1902         status = ocfs2_check_volume(osb);
1903         if (status < 0) {
1904                 mlog_errno(status);
1905                 goto leave;
1906         }
1907
1908         status = ocfs2_truncate_log_init(osb);
1909         if (status < 0)
1910                 mlog_errno(status);
1911
1912 leave:
1913         if (unlock_super)
1914                 ocfs2_super_unlock(osb, 1);
1915
1916         return status;
1917 }
1918
1919 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1920 {
1921         int tmp, hangup_needed = 0;
1922         struct ocfs2_super *osb = NULL;
1923         char nodestr[8];
1924
1925         trace_ocfs2_dismount_volume(sb);
1926
1927         BUG_ON(!sb);
1928         osb = OCFS2_SB(sb);
1929         BUG_ON(!osb);
1930
1931         debugfs_remove(osb->osb_ctxt);
1932
1933         /*
1934          * Flush inode dropping work queue so that deletes are
1935          * performed while the filesystem is still working
1936          */
1937         ocfs2_drop_all_dl_inodes(osb);
1938
1939         /* Orphan scan should be stopped as early as possible */
1940         ocfs2_orphan_scan_stop(osb);
1941
1942         ocfs2_disable_quotas(osb);
1943
1944         ocfs2_shutdown_local_alloc(osb);
1945
1946         ocfs2_truncate_log_shutdown(osb);
1947
1948         /* This will disable recovery and flush any recovery work. */
1949         ocfs2_recovery_exit(osb);
1950
1951         ocfs2_journal_shutdown(osb);
1952
1953         ocfs2_sync_blockdev(sb);
1954
1955         ocfs2_purge_refcount_trees(osb);
1956
1957         /* No cluster connection means we've failed during mount, so skip
1958          * all the steps which depended on that to complete. */
1959         if (osb->cconn) {
1960                 tmp = ocfs2_super_lock(osb, 1);
1961                 if (tmp < 0) {
1962                         mlog_errno(tmp);
1963                         return;
1964                 }
1965         }
1966
1967         if (osb->slot_num != OCFS2_INVALID_SLOT)
1968                 ocfs2_put_slot(osb);
1969
1970         if (osb->cconn)
1971                 ocfs2_super_unlock(osb, 1);
1972
1973         ocfs2_release_system_inodes(osb);
1974
1975         /*
1976          * If we're dismounting due to mount error, mount.ocfs2 will clean
1977          * up heartbeat.  If we're a local mount, there is no heartbeat.
1978          * If we failed before we got a uuid_str yet, we can't stop
1979          * heartbeat.  Otherwise, do it.
1980          */
1981         if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str &&
1982             !ocfs2_is_hard_readonly(osb))
1983                 hangup_needed = 1;
1984
1985         if (osb->cconn)
1986                 ocfs2_dlm_shutdown(osb, hangup_needed);
1987
1988         ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1989         debugfs_remove(osb->osb_debug_root);
1990
1991         if (hangup_needed)
1992                 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1993
1994         atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1995
1996         if (ocfs2_mount_local(osb))
1997                 snprintf(nodestr, sizeof(nodestr), "local");
1998         else
1999                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
2000
2001         printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
2002                osb->dev_str, nodestr);
2003
2004         ocfs2_delete_osb(osb);
2005         kfree(osb);
2006         sb->s_dev = 0;
2007         sb->s_fs_info = NULL;
2008 }
2009
2010 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
2011                                 unsigned uuid_bytes)
2012 {
2013         int i, ret;
2014         char *ptr;
2015
2016         BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
2017
2018         osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
2019         if (osb->uuid_str == NULL)
2020                 return -ENOMEM;
2021
2022         for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
2023                 /* print with null */
2024                 ret = snprintf(ptr, 3, "%02X", uuid[i]);
2025                 if (ret != 2) /* drop super cleans up */
2026                         return -EINVAL;
2027                 /* then only advance past the last char */
2028                 ptr += 2;
2029         }
2030
2031         return 0;
2032 }
2033
2034 /* Make sure entire volume is addressable by our journal.  Requires
2035    osb_clusters_at_boot to be valid and for the journal to have been
2036    initialized by ocfs2_journal_init(). */
2037 static int ocfs2_journal_addressable(struct ocfs2_super *osb)
2038 {
2039         int status = 0;
2040         u64 max_block =
2041                 ocfs2_clusters_to_blocks(osb->sb,
2042                                          osb->osb_clusters_at_boot) - 1;
2043
2044         /* 32-bit block number is always OK. */
2045         if (max_block <= (u32)~0ULL)
2046                 goto out;
2047
2048         /* Volume is "huge", so see if our journal is new enough to
2049            support it. */
2050         if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2051                                        OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2052               jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2053                                                JBD2_FEATURE_INCOMPAT_64BIT))) {
2054                 mlog(ML_ERROR, "The journal cannot address the entire volume. "
2055                      "Enable the 'block64' journal option with tunefs.ocfs2");
2056                 status = -EFBIG;
2057                 goto out;
2058         }
2059
2060  out:
2061         return status;
2062 }
2063
2064 static int ocfs2_initialize_super(struct super_block *sb,
2065                                   struct buffer_head *bh,
2066                                   int sector_size,
2067                                   struct ocfs2_blockcheck_stats *stats)
2068 {
2069         int status;
2070         int i, cbits, bbits;
2071         struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2072         struct inode *inode = NULL;
2073         struct ocfs2_journal *journal;
2074         __le32 uuid_net_key;
2075         struct ocfs2_super *osb;
2076         u64 total_blocks;
2077
2078         osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2079         if (!osb) {
2080                 status = -ENOMEM;
2081                 mlog_errno(status);
2082                 goto bail;
2083         }
2084
2085         sb->s_fs_info = osb;
2086         sb->s_op = &ocfs2_sops;
2087         sb->s_d_op = &ocfs2_dentry_ops;
2088         sb->s_export_op = &ocfs2_export_ops;
2089         sb->s_qcop = &ocfs2_quotactl_ops;
2090         sb->dq_op = &ocfs2_quota_operations;
2091         sb->s_xattr = ocfs2_xattr_handlers;
2092         sb->s_time_gran = 1;
2093         sb->s_flags |= MS_NOATIME;
2094         /* this is needed to support O_LARGEFILE */
2095         cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2096         bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2097         sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2098
2099         osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2100
2101         for (i = 0; i < 3; i++)
2102                 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2103         osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2104
2105         osb->sb = sb;
2106         /* Save off for ocfs2_rw_direct */
2107         osb->s_sectsize_bits = blksize_bits(sector_size);
2108         BUG_ON(!osb->s_sectsize_bits);
2109
2110         spin_lock_init(&osb->dc_task_lock);
2111         init_waitqueue_head(&osb->dc_event);
2112         osb->dc_work_sequence = 0;
2113         osb->dc_wake_sequence = 0;
2114         INIT_LIST_HEAD(&osb->blocked_lock_list);
2115         osb->blocked_lock_count = 0;
2116         spin_lock_init(&osb->osb_lock);
2117         spin_lock_init(&osb->osb_xattr_lock);
2118         ocfs2_init_steal_slots(osb);
2119
2120         atomic_set(&osb->alloc_stats.moves, 0);
2121         atomic_set(&osb->alloc_stats.local_data, 0);
2122         atomic_set(&osb->alloc_stats.bitmap_data, 0);
2123         atomic_set(&osb->alloc_stats.bg_allocs, 0);
2124         atomic_set(&osb->alloc_stats.bg_extends, 0);
2125
2126         /* Copy the blockcheck stats from the superblock probe */
2127         osb->osb_ecc_stats = *stats;
2128
2129         ocfs2_init_node_maps(osb);
2130
2131         snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2132                  MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2133
2134         osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2135         if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2136                 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2137                      osb->max_slots);
2138                 status = -EINVAL;
2139                 goto bail;
2140         }
2141
2142         ocfs2_orphan_scan_init(osb);
2143
2144         status = ocfs2_recovery_init(osb);
2145         if (status) {
2146                 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2147                 mlog_errno(status);
2148                 goto bail;
2149         }
2150
2151         init_waitqueue_head(&osb->checkpoint_event);
2152         atomic_set(&osb->needs_checkpoint, 0);
2153
2154         osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2155
2156         osb->slot_num = OCFS2_INVALID_SLOT;
2157
2158         osb->s_xattr_inline_size = le16_to_cpu(
2159                                         di->id2.i_super.s_xattr_inline_size);
2160
2161         osb->local_alloc_state = OCFS2_LA_UNUSED;
2162         osb->local_alloc_bh = NULL;
2163         INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2164
2165         init_waitqueue_head(&osb->osb_mount_event);
2166
2167         status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2168         if (status) {
2169                 mlog_errno(status);
2170                 goto bail;
2171         }
2172
2173         osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2174         if (!osb->vol_label) {
2175                 mlog(ML_ERROR, "unable to alloc vol label\n");
2176                 status = -ENOMEM;
2177                 goto bail;
2178         }
2179
2180         osb->slot_recovery_generations =
2181                 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2182                         GFP_KERNEL);
2183         if (!osb->slot_recovery_generations) {
2184                 status = -ENOMEM;
2185                 mlog_errno(status);
2186                 goto bail;
2187         }
2188
2189         init_waitqueue_head(&osb->osb_wipe_event);
2190         osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2191                                         sizeof(*osb->osb_orphan_wipes),
2192                                         GFP_KERNEL);
2193         if (!osb->osb_orphan_wipes) {
2194                 status = -ENOMEM;
2195                 mlog_errno(status);
2196                 goto bail;
2197         }
2198
2199         osb->osb_rf_lock_tree = RB_ROOT;
2200
2201         osb->s_feature_compat =
2202                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2203         osb->s_feature_ro_compat =
2204                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2205         osb->s_feature_incompat =
2206                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2207
2208         if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2209                 mlog(ML_ERROR, "couldn't mount because of unsupported "
2210                      "optional features (%x).\n", i);
2211                 status = -EINVAL;
2212                 goto bail;
2213         }
2214         if (!(osb->sb->s_flags & MS_RDONLY) &&
2215             (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2216                 mlog(ML_ERROR, "couldn't mount RDWR because of "
2217                      "unsupported optional features (%x).\n", i);
2218                 status = -EINVAL;
2219                 goto bail;
2220         }
2221
2222         if (ocfs2_clusterinfo_valid(osb)) {
2223                 osb->osb_stackflags =
2224                         OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2225                 memcpy(osb->osb_cluster_stack,
2226                        OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2227                        OCFS2_STACK_LABEL_LEN);
2228                 osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
2229                 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2230                         mlog(ML_ERROR,
2231                              "couldn't mount because of an invalid "
2232                              "cluster stack label (%s) \n",
2233                              osb->osb_cluster_stack);
2234                         status = -EINVAL;
2235                         goto bail;
2236                 }
2237         } else {
2238                 /* The empty string is identical with classic tools that
2239                  * don't know about s_cluster_info. */
2240                 osb->osb_cluster_stack[0] = '\0';
2241         }
2242
2243         get_random_bytes(&osb->s_next_generation, sizeof(u32));
2244
2245         /* FIXME
2246          * This should be done in ocfs2_journal_init(), but unknown
2247          * ordering issues will cause the filesystem to crash.
2248          * If anyone wants to figure out what part of the code
2249          * refers to osb->journal before ocfs2_journal_init() is run,
2250          * be my guest.
2251          */
2252         /* initialize our journal structure */
2253
2254         journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2255         if (!journal) {
2256                 mlog(ML_ERROR, "unable to alloc journal\n");
2257                 status = -ENOMEM;
2258                 goto bail;
2259         }
2260         osb->journal = journal;
2261         journal->j_osb = osb;
2262
2263         atomic_set(&journal->j_num_trans, 0);
2264         init_rwsem(&journal->j_trans_barrier);
2265         init_waitqueue_head(&journal->j_checkpointed);
2266         spin_lock_init(&journal->j_lock);
2267         journal->j_trans_id = (unsigned long) 1;
2268         INIT_LIST_HEAD(&journal->j_la_cleanups);
2269         INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2270         journal->j_state = OCFS2_JOURNAL_FREE;
2271
2272         INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes);
2273         osb->dentry_lock_list = NULL;
2274
2275         /* get some pseudo constants for clustersize bits */
2276         osb->s_clustersize_bits =
2277                 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2278         osb->s_clustersize = 1 << osb->s_clustersize_bits;
2279
2280         if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2281             osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2282                 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2283                      osb->s_clustersize);
2284                 status = -EINVAL;
2285                 goto bail;
2286         }
2287
2288         total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2289                                                 le32_to_cpu(di->i_clusters));
2290
2291         status = generic_check_addressable(osb->sb->s_blocksize_bits,
2292                                            total_blocks);
2293         if (status) {
2294                 mlog(ML_ERROR, "Volume too large "
2295                      "to mount safely on this system");
2296                 status = -EFBIG;
2297                 goto bail;
2298         }
2299
2300         if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2301                                  sizeof(di->id2.i_super.s_uuid))) {
2302                 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2303                 status = -ENOMEM;
2304                 goto bail;
2305         }
2306
2307         memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
2308
2309         strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2310         osb->vol_label[63] = '\0';
2311         osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2312         osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2313         osb->first_cluster_group_blkno =
2314                 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2315         osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2316         osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2317         trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str,
2318                                      (unsigned long long)osb->root_blkno,
2319                                      (unsigned long long)osb->system_dir_blkno,
2320                                      osb->s_clustersize_bits);
2321
2322         osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2323         if (!osb->osb_dlm_debug) {
2324                 status = -ENOMEM;
2325                 mlog_errno(status);
2326                 goto bail;
2327         }
2328
2329         atomic_set(&osb->vol_state, VOLUME_INIT);
2330
2331         /* load root, system_dir, and all global system inodes */
2332         status = ocfs2_init_global_system_inodes(osb);
2333         if (status < 0) {
2334                 mlog_errno(status);
2335                 goto bail;
2336         }
2337
2338         /*
2339          * global bitmap
2340          */
2341         inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2342                                             OCFS2_INVALID_SLOT);
2343         if (!inode) {
2344                 status = -EINVAL;
2345                 mlog_errno(status);
2346                 goto bail;
2347         }
2348
2349         osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2350         osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2351         iput(inode);
2352
2353         osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2354                                  osb->s_feature_incompat) * 8;
2355
2356         status = ocfs2_init_slot_info(osb);
2357         if (status < 0) {
2358                 mlog_errno(status);
2359                 goto bail;
2360         }
2361         cleancache_init_shared_fs((char *)&di->id2.i_super.s_uuid, sb);
2362
2363 bail:
2364         return status;
2365 }
2366
2367 /*
2368  * will return: -EAGAIN if it is ok to keep searching for superblocks
2369  *              -EINVAL if there is a bad superblock
2370  *              0 on success
2371  */
2372 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2373                                struct buffer_head *bh,
2374                                u32 blksz,
2375                                struct ocfs2_blockcheck_stats *stats)
2376 {
2377         int status = -EAGAIN;
2378
2379         if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2380                    strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2381                 /* We have to do a raw check of the feature here */
2382                 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2383                     OCFS2_FEATURE_INCOMPAT_META_ECC) {
2384                         status = ocfs2_block_check_validate(bh->b_data,
2385                                                             bh->b_size,
2386                                                             &di->i_check,
2387                                                             stats);
2388                         if (status)
2389                                 goto out;
2390                 }
2391                 status = -EINVAL;
2392                 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2393                         mlog(ML_ERROR, "found superblock with incorrect block "
2394                              "size: found %u, should be %u\n",
2395                              1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2396                                blksz);
2397                 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2398                            OCFS2_MAJOR_REV_LEVEL ||
2399                            le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2400                            OCFS2_MINOR_REV_LEVEL) {
2401                         mlog(ML_ERROR, "found superblock with bad version: "
2402                              "found %u.%u, should be %u.%u\n",
2403                              le16_to_cpu(di->id2.i_super.s_major_rev_level),
2404                              le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2405                              OCFS2_MAJOR_REV_LEVEL,
2406                              OCFS2_MINOR_REV_LEVEL);
2407                 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2408                         mlog(ML_ERROR, "bad block number on superblock: "
2409                              "found %llu, should be %llu\n",
2410                              (unsigned long long)le64_to_cpu(di->i_blkno),
2411                              (unsigned long long)bh->b_blocknr);
2412                 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2413                             le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2414                         mlog(ML_ERROR, "bad cluster size found: %u\n",
2415                              1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2416                 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2417                         mlog(ML_ERROR, "bad root_blkno: 0\n");
2418                 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2419                         mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2420                 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2421                         mlog(ML_ERROR,
2422                              "Superblock slots found greater than file system "
2423                              "maximum: found %u, max %u\n",
2424                              le16_to_cpu(di->id2.i_super.s_max_slots),
2425                              OCFS2_MAX_SLOTS);
2426                 } else {
2427                         /* found it! */
2428                         status = 0;
2429                 }
2430         }
2431
2432 out:
2433         if (status && status != -EAGAIN)
2434                 mlog_errno(status);
2435         return status;
2436 }
2437
2438 static int ocfs2_check_volume(struct ocfs2_super *osb)
2439 {
2440         int status;
2441         int dirty;
2442         int local;
2443         struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2444                                                   * recover
2445                                                   * ourselves. */
2446
2447         /* Init our journal object. */
2448         status = ocfs2_journal_init(osb->journal, &dirty);
2449         if (status < 0) {
2450                 mlog(ML_ERROR, "Could not initialize journal!\n");
2451                 goto finally;
2452         }
2453
2454         /* Now that journal has been initialized, check to make sure
2455            entire volume is addressable. */
2456         status = ocfs2_journal_addressable(osb);
2457         if (status)
2458                 goto finally;
2459
2460         /* If the journal was unmounted cleanly then we don't want to
2461          * recover anything. Otherwise, journal_load will do that
2462          * dirty work for us :) */
2463         if (!dirty) {
2464                 status = ocfs2_journal_wipe(osb->journal, 0);
2465                 if (status < 0) {
2466                         mlog_errno(status);
2467                         goto finally;
2468                 }
2469         } else {
2470                 printk(KERN_NOTICE "ocfs2: File system on device (%s) was not "
2471                        "unmounted cleanly, recovering it.\n", osb->dev_str);
2472         }
2473
2474         local = ocfs2_mount_local(osb);
2475
2476         /* will play back anything left in the journal. */
2477         status = ocfs2_journal_load(osb->journal, local, dirty);
2478         if (status < 0) {
2479                 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2480                 goto finally;
2481         }
2482
2483         if (dirty) {
2484                 /* recover my local alloc if we didn't unmount cleanly. */
2485                 status = ocfs2_begin_local_alloc_recovery(osb,
2486                                                           osb->slot_num,
2487                                                           &local_alloc);
2488                 if (status < 0) {
2489                         mlog_errno(status);
2490                         goto finally;
2491                 }
2492                 /* we complete the recovery process after we've marked
2493                  * ourselves as mounted. */
2494         }
2495
2496         status = ocfs2_load_local_alloc(osb);
2497         if (status < 0) {
2498                 mlog_errno(status);
2499                 goto finally;
2500         }
2501
2502         if (dirty) {
2503                 /* Recovery will be completed after we've mounted the
2504                  * rest of the volume. */
2505                 osb->dirty = 1;
2506                 osb->local_alloc_copy = local_alloc;
2507                 local_alloc = NULL;
2508         }
2509
2510         /* go through each journal, trylock it and if you get the
2511          * lock, and it's marked as dirty, set the bit in the recover
2512          * map and launch a recovery thread for it. */
2513         status = ocfs2_mark_dead_nodes(osb);
2514         if (status < 0) {
2515                 mlog_errno(status);
2516                 goto finally;
2517         }
2518
2519         status = ocfs2_compute_replay_slots(osb);
2520         if (status < 0)
2521                 mlog_errno(status);
2522
2523 finally:
2524         if (local_alloc)
2525                 kfree(local_alloc);
2526
2527         if (status)
2528                 mlog_errno(status);
2529         return status;
2530 }
2531
2532 /*
2533  * The routine gets called from dismount or close whenever a dismount on
2534  * volume is requested and the osb open count becomes 1.
2535  * It will remove the osb from the global list and also free up all the
2536  * initialized resources and fileobject.
2537  */
2538 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2539 {
2540         /* This function assumes that the caller has the main osb resource */
2541
2542         ocfs2_free_slot_info(osb);
2543
2544         kfree(osb->osb_orphan_wipes);
2545         kfree(osb->slot_recovery_generations);
2546         /* FIXME
2547          * This belongs in journal shutdown, but because we have to
2548          * allocate osb->journal at the start of ocfs2_initialize_osb(),
2549          * we free it here.
2550          */
2551         kfree(osb->journal);
2552         if (osb->local_alloc_copy)
2553                 kfree(osb->local_alloc_copy);
2554         kfree(osb->uuid_str);
2555         ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2556         memset(osb, 0, sizeof(struct ocfs2_super));
2557 }
2558
2559 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2560  * panic(). We do not support continue-on-error operation. */
2561 static void ocfs2_handle_error(struct super_block *sb)
2562 {
2563         struct ocfs2_super *osb = OCFS2_SB(sb);
2564
2565         if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2566                 panic("OCFS2: (device %s): panic forced after error\n",
2567                       sb->s_id);
2568
2569         ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2570
2571         if (sb->s_flags & MS_RDONLY &&
2572             (ocfs2_is_soft_readonly(osb) ||
2573              ocfs2_is_hard_readonly(osb)))
2574                 return;
2575
2576         printk(KERN_CRIT "File system is now read-only due to the potential "
2577                "of on-disk corruption. Please run fsck.ocfs2 once the file "
2578                "system is unmounted.\n");
2579         sb->s_flags |= MS_RDONLY;
2580         ocfs2_set_ro_flag(osb, 0);
2581 }
2582
2583 static char error_buf[1024];
2584
2585 void __ocfs2_error(struct super_block *sb,
2586                    const char *function,
2587                    const char *fmt, ...)
2588 {
2589         va_list args;
2590
2591         va_start(args, fmt);
2592         vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2593         va_end(args);
2594
2595         /* Not using mlog here because we want to show the actual
2596          * function the error came from. */
2597         printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2598                sb->s_id, function, error_buf);
2599
2600         ocfs2_handle_error(sb);
2601 }
2602
2603 /* Handle critical errors. This is intentionally more drastic than
2604  * ocfs2_handle_error, so we only use for things like journal errors,
2605  * etc. */
2606 void __ocfs2_abort(struct super_block* sb,
2607                    const char *function,
2608                    const char *fmt, ...)
2609 {
2610         va_list args;
2611
2612         va_start(args, fmt);
2613         vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2614         va_end(args);
2615
2616         printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2617                sb->s_id, function, error_buf);
2618
2619         /* We don't have the cluster support yet to go straight to
2620          * hard readonly in here. Until then, we want to keep
2621          * ocfs2_abort() so that we can at least mark critical
2622          * errors.
2623          *
2624          * TODO: This should abort the journal and alert other nodes
2625          * that our slot needs recovery. */
2626
2627         /* Force a panic(). This stinks, but it's better than letting
2628          * things continue without having a proper hard readonly
2629          * here. */
2630         if (!ocfs2_mount_local(OCFS2_SB(sb)))
2631                 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2632         ocfs2_handle_error(sb);
2633 }
2634
2635 /*
2636  * Void signal blockers, because in-kernel sigprocmask() only fails
2637  * when SIG_* is wrong.
2638  */
2639 void ocfs2_block_signals(sigset_t *oldset)
2640 {
2641         int rc;
2642         sigset_t blocked;
2643
2644         sigfillset(&blocked);
2645         rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2646         BUG_ON(rc);
2647 }
2648
2649 void ocfs2_unblock_signals(sigset_t *oldset)
2650 {
2651         int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2652         BUG_ON(rc);
2653 }
2654
2655 module_init(ocfs2_init);
2656 module_exit(ocfs2_exit);