Pileus Git - ~andy/linux/blob - 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
  45 #define CREATE_TRACE_POINTS
  46 #include "ocfs2_trace.h"
  47
  48 #define MLOG_MASK_PREFIX ML_SUPER
  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 "blockcheck.h"
  58 #include "dlmglue.h"
  59 #include "export.h"
  60 #include "extent_map.h"
  61 #include "heartbeat.h"
  62 #include "inode.h"
  63 #include "journal.h"
  64 #include "localalloc.h"
  65 #include "namei.h"
  66 #include "slot_map.h"
  67 #include "super.h"
  68 #include "sysfile.h"
  69 #include "uptodate.h"
  70 #include "ver.h"
  71 #include "xattr.h"
  72 #include "quota.h"
  73 #include "refcounttree.h"
  74 #include "suballoc.h"
  75
  76 #include "buffer_head_io.h"
  77
  78 static struct kmem_cache *ocfs2_inode_cachep = NULL;
  79 struct kmem_cache *ocfs2_dquot_cachep;
  80 struct kmem_cache *ocfs2_qf_chunk_cachep;
  81
  82 /* OCFS2 needs to schedule several differnt types of work which
  83  * require cluster locking, disk I/O, recovery waits, etc. Since these
  84  * types of work tend to be heavy we avoid using the kernel events
  85  * workqueue and schedule on our own. */
  86 struct workqueue_struct *ocfs2_wq = NULL;
  87
  88 static struct dentry *ocfs2_debugfs_root = NULL;
  89
  90 MODULE_AUTHOR("Oracle");
  91 MODULE_LICENSE("GPL");
  92
  93 struct mount_options
  94 {
  95         unsigned long   commit_interval;
  96         unsigned long   mount_opt;
  97         unsigned int    atime_quantum;
  98         signed short    slot;
  99         int             localalloc_opt;
 100         unsigned int    resv_level;
 101         int             dir_resv_level;
 102         char            cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
 103 };
 104
 105 static int ocfs2_parse_options(struct super_block *sb, char *options,
 106                                struct mount_options *mopt,
 107                                int is_remount);
 108 static int ocfs2_check_set_options(struct super_block *sb,
 109                                    struct mount_options *options);
 110 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt);
 111 static void ocfs2_put_super(struct super_block *sb);
 112 static int ocfs2_mount_volume(struct super_block *sb);
 113 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
 114 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
 115 static int ocfs2_initialize_mem_caches(void);
 116 static void ocfs2_free_mem_caches(void);
 117 static void ocfs2_delete_osb(struct ocfs2_super *osb);
 118
 119 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
 120
 121 static int ocfs2_sync_fs(struct super_block *sb, int wait);
 122
 123 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
 124 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
 125 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
 126 static int ocfs2_check_volume(struct ocfs2_super *osb);
 127 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
 128                                struct buffer_head *bh,
 129                                u32 sectsize,
 130                                struct ocfs2_blockcheck_stats *stats);
 131 static int ocfs2_initialize_super(struct super_block *sb,
 132                                   struct buffer_head *bh,
 133                                   int sector_size,
 134                                   struct ocfs2_blockcheck_stats *stats);
 135 static int ocfs2_get_sector(struct super_block *sb,
 136                             struct buffer_head **bh,
 137                             int block,
 138                             int sect_size);
 139 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
 140 static void ocfs2_destroy_inode(struct inode *inode);
 141 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
 142 static int ocfs2_enable_quotas(struct ocfs2_super *osb);
 143 static void ocfs2_disable_quotas(struct ocfs2_super *osb);
 144
 145 static const struct super_operations ocfs2_sops = {
 146         .statfs         = ocfs2_statfs,
 147         .alloc_inode    = ocfs2_alloc_inode,
 148         .destroy_inode  = ocfs2_destroy_inode,
 149         .drop_inode     = ocfs2_drop_inode,
 150         .evict_inode    = ocfs2_evict_inode,
 151         .sync_fs        = ocfs2_sync_fs,
 152         .put_super      = ocfs2_put_super,
 153         .remount_fs     = ocfs2_remount,
 154         .show_options   = ocfs2_show_options,
 155         .quota_read     = ocfs2_quota_read,
 156         .quota_write    = ocfs2_quota_write,
 157 };
 158
 159 enum {
 160         Opt_barrier,
 161         Opt_err_panic,
 162         Opt_err_ro,
 163         Opt_intr,
 164         Opt_nointr,
 165         Opt_hb_none,
 166         Opt_hb_local,
 167         Opt_hb_global,
 168         Opt_data_ordered,
 169         Opt_data_writeback,
 170         Opt_atime_quantum,
 171         Opt_slot,
 172         Opt_commit,
 173         Opt_localalloc,
 174         Opt_localflocks,
 175         Opt_stack,
 176         Opt_user_xattr,
 177         Opt_nouser_xattr,
 178         Opt_inode64,
 179         Opt_acl,
 180         Opt_noacl,
 181         Opt_usrquota,
 182         Opt_grpquota,
 183         Opt_coherency_buffered,
 184         Opt_coherency_full,
 185         Opt_resv_level,
 186         Opt_dir_resv_level,
 187         Opt_err,
 188 };
 189
 190 static const match_table_t tokens = {
 191         {Opt_barrier, "barrier=%u"},
 192         {Opt_err_panic, "errors=panic"},
 193         {Opt_err_ro, "errors=remount-ro"},
 194         {Opt_intr, "intr"},
 195         {Opt_nointr, "nointr"},
 196         {Opt_hb_none, OCFS2_HB_NONE},
 197         {Opt_hb_local, OCFS2_HB_LOCAL},
 198         {Opt_hb_global, OCFS2_HB_GLOBAL},
 199         {Opt_data_ordered, "data=ordered"},
 200         {Opt_data_writeback, "data=writeback"},
 201         {Opt_atime_quantum, "atime_quantum=%u"},
 202         {Opt_slot, "preferred_slot=%u"},
 203         {Opt_commit, "commit=%u"},
 204         {Opt_localalloc, "localalloc=%d"},
 205         {Opt_localflocks, "localflocks"},
 206         {Opt_stack, "cluster_stack=%s"},
 207         {Opt_user_xattr, "user_xattr"},
 208         {Opt_nouser_xattr, "nouser_xattr"},
 209         {Opt_inode64, "inode64"},
 210         {Opt_acl, "acl"},
 211         {Opt_noacl, "noacl"},
 212         {Opt_usrquota, "usrquota"},
 213         {Opt_grpquota, "grpquota"},
 214         {Opt_coherency_buffered, "coherency=buffered"},
 215         {Opt_coherency_full, "coherency=full"},
 216         {Opt_resv_level, "resv_level=%u"},
 217         {Opt_dir_resv_level, "dir_resv_level=%u"},
 218         {Opt_err, NULL}
 219 };
 220
 221 #ifdef CONFIG_DEBUG_FS
 222 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
 223 {
 224         struct ocfs2_cluster_connection *cconn = osb->cconn;
 225         struct ocfs2_recovery_map *rm = osb->recovery_map;
 226         struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
 227         int i, out = 0;
 228
 229         out += snprintf(buf + out, len - out,
 230                         "%10s => Id: %-s  Uuid: %-s  Gen: 0x%X  Label: %-s\n",
 231                         "Device", osb->dev_str, osb->uuid_str,
 232                         osb->fs_generation, osb->vol_label);
 233
 234         out += snprintf(buf + out, len - out,
 235                         "%10s => State: %d  Flags: 0x%lX\n", "Volume",
 236                         atomic_read(&osb->vol_state), osb->osb_flags);
 237
 238         out += snprintf(buf + out, len - out,
 239                         "%10s => Block: %lu  Cluster: %d\n", "Sizes",
 240                         osb->sb->s_blocksize, osb->s_clustersize);
 241
 242         out += snprintf(buf + out, len - out,
 243                         "%10s => Compat: 0x%X  Incompat: 0x%X  "
 244                         "ROcompat: 0x%X\n",
 245                         "Features", osb->s_feature_compat,
 246                         osb->s_feature_incompat, osb->s_feature_ro_compat);
 247
 248         out += snprintf(buf + out, len - out,
 249                         "%10s => Opts: 0x%lX  AtimeQuanta: %u\n", "Mount",
 250                         osb->s_mount_opt, osb->s_atime_quantum);
 251
 252         if (cconn) {
 253                 out += snprintf(buf + out, len - out,
 254                                 "%10s => Stack: %s  Name: %*s  "
 255                                 "Version: %d.%d\n", "Cluster",
 256                                 (*osb->osb_cluster_stack == '\0' ?
 257                                  "o2cb" : osb->osb_cluster_stack),
 258                                 cconn->cc_namelen, cconn->cc_name,
 259                                 cconn->cc_version.pv_major,
 260                                 cconn->cc_version.pv_minor);
 261         }
 262
 263         spin_lock(&osb->dc_task_lock);
 264         out += snprintf(buf + out, len - out,
 265                         "%10s => Pid: %d  Count: %lu  WakeSeq: %lu  "
 266                         "WorkSeq: %lu\n", "DownCnvt",
 267                         (osb->dc_task ?  task_pid_nr(osb->dc_task) : -1),
 268                         osb->blocked_lock_count, osb->dc_wake_sequence,
 269                         osb->dc_work_sequence);
 270         spin_unlock(&osb->dc_task_lock);
 271
 272         spin_lock(&osb->osb_lock);
 273         out += snprintf(buf + out, len - out, "%10s => Pid: %d  Nodes:",
 274                         "Recovery",
 275                         (osb->recovery_thread_task ?
 276                          task_pid_nr(osb->recovery_thread_task) : -1));
 277         if (rm->rm_used == 0)
 278                 out += snprintf(buf + out, len - out, " None\n");
 279         else {
 280                 for (i = 0; i < rm->rm_used; i++)
 281                         out += snprintf(buf + out, len - out, " %d",
 282                                         rm->rm_entries[i]);
 283                 out += snprintf(buf + out, len - out, "\n");
 284         }
 285         spin_unlock(&osb->osb_lock);
 286
 287         out += snprintf(buf + out, len - out,
 288                         "%10s => Pid: %d  Interval: %lu  Needs: %d\n", "Commit",
 289                         (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
 290                         osb->osb_commit_interval,
 291                         atomic_read(&osb->needs_checkpoint));
 292
 293         out += snprintf(buf + out, len - out,
 294                         "%10s => State: %d  TxnId: %lu  NumTxns: %d\n",
 295                         "Journal", osb->journal->j_state,
 296                         osb->journal->j_trans_id,
 297                         atomic_read(&osb->journal->j_num_trans));
 298
 299         out += snprintf(buf + out, len - out,
 300                         "%10s => GlobalAllocs: %d  LocalAllocs: %d  "
 301                         "SubAllocs: %d  LAWinMoves: %d  SAExtends: %d\n",
 302                         "Stats",
 303                         atomic_read(&osb->alloc_stats.bitmap_data),
 304                         atomic_read(&osb->alloc_stats.local_data),
 305                         atomic_read(&osb->alloc_stats.bg_allocs),
 306                         atomic_read(&osb->alloc_stats.moves),
 307                         atomic_read(&osb->alloc_stats.bg_extends));
 308
 309         out += snprintf(buf + out, len - out,
 310                         "%10s => State: %u  Descriptor: %llu  Size: %u bits  "
 311                         "Default: %u bits\n",
 312                         "LocalAlloc", osb->local_alloc_state,
 313                         (unsigned long long)osb->la_last_gd,
 314                         osb->local_alloc_bits, osb->local_alloc_default_bits);
 315
 316         spin_lock(&osb->osb_lock);
 317         out += snprintf(buf + out, len - out,
 318                         "%10s => InodeSlot: %d  StolenInodes: %d, "
 319                         "MetaSlot: %d  StolenMeta: %d\n", "Steal",
 320                         osb->s_inode_steal_slot,
 321                         atomic_read(&osb->s_num_inodes_stolen),
 322                         osb->s_meta_steal_slot,
 323                         atomic_read(&osb->s_num_meta_stolen));
 324         spin_unlock(&osb->osb_lock);
 325
 326         out += snprintf(buf + out, len - out, "OrphanScan => ");
 327         out += snprintf(buf + out, len - out, "Local: %u  Global: %u ",
 328                         os->os_count, os->os_seqno);
 329         out += snprintf(buf + out, len - out, " Last Scan: ");
 330         if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
 331                 out += snprintf(buf + out, len - out, "Disabled\n");
 332         else
 333                 out += snprintf(buf + out, len - out, "%lu seconds ago\n",
 334                                 (get_seconds() - os->os_scantime.tv_sec));
 335
 336         out += snprintf(buf + out, len - out, "%10s => %3s  %10s\n",
 337                         "Slots", "Num", "RecoGen");
 338         for (i = 0; i < osb->max_slots; ++i) {
 339                 out += snprintf(buf + out, len - out,
 340                                 "%10s  %c %3d  %10d\n",
 341                                 " ",
 342                                 (i == osb->slot_num ? '*' : ' '),
 343                                 i, osb->slot_recovery_generations[i]);
 344         }
 345
 346         return out;
 347 }
 348
 349 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 350 {
 351         struct ocfs2_super *osb = inode->i_private;
 352         char *buf = NULL;
 353
 354         buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
 355         if (!buf)
 356                 goto bail;
 357
 358         i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
 359
 360         file->private_data = buf;
 361
 362         return 0;
 363 bail:
 364         return -ENOMEM;
 365 }
 366
 367 static int ocfs2_debug_release(struct inode *inode, struct file *file)
 368 {
 369         kfree(file->private_data);
 370         return 0;
 371 }
 372
 373 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 374                                 size_t nbytes, loff_t *ppos)
 375 {
 376         return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
 377                                        i_size_read(file->f_mapping->host));
 378 }
 379 #else
 380 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 381 {
 382         return 0;
 383 }
 384 static int ocfs2_debug_release(struct inode *inode, struct file *file)
 385 {
 386         return 0;
 387 }
 388 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 389                                 size_t nbytes, loff_t *ppos)
 390 {
 391         return 0;
 392 }
 393 #endif  /* CONFIG_DEBUG_FS */
 394
 395 static const struct file_operations ocfs2_osb_debug_fops = {
 396         .open =         ocfs2_osb_debug_open,
 397         .release =      ocfs2_debug_release,
 398         .read =         ocfs2_debug_read,
 399         .llseek =       generic_file_llseek,
 400 };
 401
 402 static int ocfs2_sync_fs(struct super_block *sb, int wait)
 403 {
 404         int status;
 405         tid_t target;
 406         struct ocfs2_super *osb = OCFS2_SB(sb);
 407
 408         if (ocfs2_is_hard_readonly(osb))
 409                 return -EROFS;
 410
 411         if (wait) {
 412                 status = ocfs2_flush_truncate_log(osb);
 413                 if (status < 0)
 414                         mlog_errno(status);
 415         } else {
 416                 ocfs2_schedule_truncate_log_flush(osb, 0);
 417         }
 418
 419         if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
 420                                       &target)) {
 421                 if (wait)
 422                         jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
 423                                              target);
 424         }
 425         return 0;
 426 }
 427
 428 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
 429 {
 430         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
 431             && (ino == USER_QUOTA_SYSTEM_INODE
 432                 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
 433                 return 0;
 434         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
 435             && (ino == GROUP_QUOTA_SYSTEM_INODE
 436                 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
 437                 return 0;
 438         return 1;
 439 }
 440
 441 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
 442 {
 443         struct inode *new = NULL;
 444         int status = 0;
 445         int i;
 446
 447         new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 448         if (IS_ERR(new)) {
 449                 status = PTR_ERR(new);
 450                 mlog_errno(status);
 451                 goto bail;
 452         }
 453         osb->root_inode = new;
 454
 455         new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 456         if (IS_ERR(new)) {
 457                 status = PTR_ERR(new);
 458                 mlog_errno(status);
 459                 goto bail;
 460         }
 461         osb->sys_root_inode = new;
 462
 463         for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
 464              i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
 465                 if (!ocfs2_need_system_inode(osb, i))
 466                         continue;
 467                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 468                 if (!new) {
 469                         ocfs2_release_system_inodes(osb);
 470                         status = -EINVAL;
 471                         mlog_errno(status);
 472                         /* FIXME: Should ERROR_RO_FS */
 473                         mlog(ML_ERROR, "Unable to load system inode %d, "
 474                              "possibly corrupt fs?", i);
 475                         goto bail;
 476                 }
 477                 // the array now has one ref, so drop this one
 478                 iput(new);
 479         }
 480
 481 bail:
 482         if (status)
 483                 mlog_errno(status);
 484         return status;
 485 }
 486
 487 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
 488 {
 489         struct inode *new = NULL;
 490         int status = 0;
 491         int i;
 492
 493         for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
 494              i < NUM_SYSTEM_INODES;
 495              i++) {
 496                 if (!ocfs2_need_system_inode(osb, i))
 497                         continue;
 498                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 499                 if (!new) {
 500                         ocfs2_release_system_inodes(osb);
 501                         status = -EINVAL;
 502                         mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
 503                              status, i, osb->slot_num);
 504                         goto bail;
 505                 }
 506                 /* the array now has one ref, so drop this one */
 507                 iput(new);
 508         }
 509
 510 bail:
 511         if (status)
 512                 mlog_errno(status);
 513         return status;
 514 }
 515
 516 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
 517 {
 518         int i;
 519         struct inode *inode;
 520
 521         for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
 522                 inode = osb->global_system_inodes[i];
 523                 if (inode) {
 524                         iput(inode);
 525                         osb->global_system_inodes[i] = NULL;
 526                 }
 527         }
 528
 529         inode = osb->sys_root_inode;
 530         if (inode) {
 531                 iput(inode);
 532                 osb->sys_root_inode = NULL;
 533         }
 534
 535         inode = osb->root_inode;
 536         if (inode) {
 537                 iput(inode);
 538                 osb->root_inode = NULL;
 539         }
 540
 541         if (!osb->local_system_inodes)
 542                 return;
 543
 544         for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
 545                 if (osb->local_system_inodes[i]) {
 546                         iput(osb->local_system_inodes[i]);
 547                         osb->local_system_inodes[i] = NULL;
 548                 }
 549         }
 550
 551         kfree(osb->local_system_inodes);
 552         osb->local_system_inodes = NULL;
 553 }
 554
 555 /* We're allocating fs objects, use GFP_NOFS */
 556 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
 557 {
 558         struct ocfs2_inode_info *oi;
 559
 560         oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
 561         if (!oi)
 562                 return NULL;
 563
 564         jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
 565         return &oi->vfs_inode;
 566 }
 567
 568 static void ocfs2_i_callback(struct rcu_head *head)
 569 {
 570         struct inode *inode = container_of(head, struct inode, i_rcu);
 571         INIT_LIST_HEAD(&inode->i_dentry);
 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                         mlog(0, "Going to ro mode.\n");
 684                         sb->s_flags |= MS_RDONLY;
 685                         osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 686                 } else {
 687                         mlog(0, "Making ro filesystem writeable.\n");
 688
 689                         if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
 690                                 mlog(ML_ERROR, "Cannot remount RDWR "
 691                                      "filesystem due to previous errors.\n");
 692                                 ret = -EROFS;
 693                                 goto unlock_osb;
 694                         }
 695                         incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
 696                         if (incompat_features) {
 697                                 mlog(ML_ERROR, "Cannot remount RDWR because "
 698                                      "of unsupported optional features "
 699                                      "(%x).\n", incompat_features);
 700                                 ret = -EINVAL;
 701                                 goto unlock_osb;
 702                         }
 703                         sb->s_flags &= ~MS_RDONLY;
 704                         osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
 705                 }
 706 unlock_osb:
 707                 spin_unlock(&osb->osb_lock);
 708                 /* Enable quota accounting after remounting RW */
 709                 if (!ret && !(*flags & MS_RDONLY)) {
 710                         if (sb_any_quota_suspended(sb))
 711                                 ret = ocfs2_susp_quotas(osb, 1);
 712                         else
 713                                 ret = ocfs2_enable_quotas(osb);
 714                         if (ret < 0) {
 715                                 /* Return back changes... */
 716                                 spin_lock(&osb->osb_lock);
 717                                 sb->s_flags |= MS_RDONLY;
 718                                 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 719                                 spin_unlock(&osb->osb_lock);
 720                                 goto out;
 721                         }
 722                 }
 723         }
 724
 725         if (!ret) {
 726                 /* Only save off the new mount options in case of a successful
 727                  * remount. */
 728                 osb->s_mount_opt = parsed_options.mount_opt;
 729                 osb->s_atime_quantum = parsed_options.atime_quantum;
 730                 osb->preferred_slot = parsed_options.slot;
 731                 if (parsed_options.commit_interval)
 732                         osb->osb_commit_interval = parsed_options.commit_interval;
 733
 734                 if (!ocfs2_is_hard_readonly(osb))
 735                         ocfs2_set_journal_params(osb);
 736
 737                 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 738                         ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
 739                                                         MS_POSIXACL : 0);
 740         }
 741 out:
 742         return ret;
 743 }
 744
 745 static int ocfs2_sb_probe(struct super_block *sb,
 746                           struct buffer_head **bh,
 747                           int *sector_size,
 748                           struct ocfs2_blockcheck_stats *stats)
 749 {
 750         int status, tmpstat;
 751         struct ocfs1_vol_disk_hdr *hdr;
 752         struct ocfs2_dinode *di;
 753         int blksize;
 754
 755         *bh = NULL;
 756
 757         /* may be > 512 */
 758         *sector_size = bdev_logical_block_size(sb->s_bdev);
 759         if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
 760                 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
 761                      *sector_size, OCFS2_MAX_BLOCKSIZE);
 762                 status = -EINVAL;
 763                 goto bail;
 764         }
 765
 766         /* Can this really happen? */
 767         if (*sector_size < OCFS2_MIN_BLOCKSIZE)
 768                 *sector_size = OCFS2_MIN_BLOCKSIZE;
 769
 770         /* check block zero for old format */
 771         status = ocfs2_get_sector(sb, bh, 0, *sector_size);
 772         if (status < 0) {
 773                 mlog_errno(status);
 774                 goto bail;
 775         }
 776         hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
 777         if (hdr->major_version == OCFS1_MAJOR_VERSION) {
 778                 mlog(ML_ERROR, "incompatible version: %u.%u\n",
 779                      hdr->major_version, hdr->minor_version);
 780                 status = -EINVAL;
 781         }
 782         if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
 783                    strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
 784                 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
 785                      hdr->signature);
 786                 status = -EINVAL;
 787         }
 788         brelse(*bh);
 789         *bh = NULL;
 790         if (status < 0) {
 791                 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
 792                      "upgraded before mounting with ocfs v2\n");
 793                 goto bail;
 794         }
 795
 796         /*
 797          * Now check at magic offset for 512, 1024, 2048, 4096
 798          * blocksizes.  4096 is the maximum blocksize because it is
 799          * the minimum clustersize.
 800          */
 801         status = -EINVAL;
 802         for (blksize = *sector_size;
 803              blksize <= OCFS2_MAX_BLOCKSIZE;
 804              blksize <<= 1) {
 805                 tmpstat = ocfs2_get_sector(sb, bh,
 806                                            OCFS2_SUPER_BLOCK_BLKNO,
 807                                            blksize);
 808                 if (tmpstat < 0) {
 809                         status = tmpstat;
 810                         mlog_errno(status);
 811                         break;
 812                 }
 813                 di = (struct ocfs2_dinode *) (*bh)->b_data;
 814                 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
 815                 spin_lock_init(&stats->b_lock);
 816                 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
 817                 if (tmpstat < 0) {
 818                         brelse(*bh);
 819                         *bh = NULL;
 820                 }
 821                 if (tmpstat != -EAGAIN) {
 822                         status = tmpstat;
 823                         break;
 824                 }
 825         }
 826
 827 bail:
 828         return status;
 829 }
 830
 831 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
 832 {
 833         u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
 834
 835         if (osb->s_mount_opt & hb_enabled) {
 836                 if (ocfs2_mount_local(osb)) {
 837                         mlog(ML_ERROR, "Cannot heartbeat on a locally "
 838                              "mounted device.\n");
 839                         return -EINVAL;
 840                 }
 841                 if (ocfs2_userspace_stack(osb)) {
 842                         mlog(ML_ERROR, "Userspace stack expected, but "
 843                              "o2cb heartbeat arguments passed to mount\n");
 844                         return -EINVAL;
 845                 }
 846                 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
 847                      !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
 848                     ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
 849                      ocfs2_cluster_o2cb_global_heartbeat(osb))) {
 850                         mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
 851                         return -EINVAL;
 852                 }
 853         }
 854
 855         if (!(osb->s_mount_opt & hb_enabled)) {
 856                 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
 857                     !ocfs2_userspace_stack(osb)) {
 858                         mlog(ML_ERROR, "Heartbeat has to be started to mount "
 859                              "a read-write clustered device.\n");
 860                         return -EINVAL;
 861                 }
 862         }
 863
 864         return 0;
 865 }
 866
 867 /*
 868  * If we're using a userspace stack, mount should have passed
 869  * a name that matches the disk.  If not, mount should not
 870  * have passed a stack.
 871  */
 872 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
 873                                         struct mount_options *mopt)
 874 {
 875         if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
 876                 mlog(ML_ERROR,
 877                      "cluster stack passed to mount, but this filesystem "
 878                      "does not support it\n");
 879                 return -EINVAL;
 880         }
 881
 882         if (ocfs2_userspace_stack(osb) &&
 883             strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
 884                     OCFS2_STACK_LABEL_LEN)) {
 885                 mlog(ML_ERROR,
 886                      "cluster stack passed to mount (\"%s\") does not "
 887                      "match the filesystem (\"%s\")\n",
 888                      mopt->cluster_stack,
 889                      osb->osb_cluster_stack);
 890                 return -EINVAL;
 891         }
 892
 893         return 0;
 894 }
 895
 896 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
 897 {
 898         int type;
 899         struct super_block *sb = osb->sb;
 900         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 901                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 902         int status = 0;
 903
 904         for (type = 0; type < MAXQUOTAS; type++) {
 905                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 906                         continue;
 907                 if (unsuspend)
 908                         status = dquot_resume(sb, type);
 909                 else {
 910                         struct ocfs2_mem_dqinfo *oinfo;
 911
 912                         /* Cancel periodic syncing before suspending */
 913                         oinfo = sb_dqinfo(sb, type)->dqi_priv;
 914                         cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 915                         status = dquot_suspend(sb, type);
 916                 }
 917                 if (status < 0)
 918                         break;
 919         }
 920         if (status < 0)
 921                 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
 922                      "remount (error = %d).\n", status);
 923         return status;
 924 }
 925
 926 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
 927 {
 928         struct inode *inode[MAXQUOTAS] = { NULL, NULL };
 929         struct super_block *sb = osb->sb;
 930         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 931                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 932         unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
 933                                         LOCAL_GROUP_QUOTA_SYSTEM_INODE };
 934         int status;
 935         int type;
 936
 937         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
 938         for (type = 0; type < MAXQUOTAS; type++) {
 939                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 940                         continue;
 941                 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
 942                                                         osb->slot_num);
 943                 if (!inode[type]) {
 944                         status = -ENOENT;
 945                         goto out_quota_off;
 946                 }
 947                 status = dquot_enable(inode[type], type, QFMT_OCFS2,
 948                                       DQUOT_USAGE_ENABLED);
 949                 if (status < 0)
 950                         goto out_quota_off;
 951         }
 952
 953         for (type = 0; type < MAXQUOTAS; type++)
 954                 iput(inode[type]);
 955         return 0;
 956 out_quota_off:
 957         ocfs2_disable_quotas(osb);
 958         for (type = 0; type < MAXQUOTAS; type++)
 959                 iput(inode[type]);
 960         mlog_errno(status);
 961         return status;
 962 }
 963
 964 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
 965 {
 966         int type;
 967         struct inode *inode;
 968         struct super_block *sb = osb->sb;
 969         struct ocfs2_mem_dqinfo *oinfo;
 970
 971         /* We mostly ignore errors in this function because there's not much
 972          * we can do when we see them */
 973         for (type = 0; type < MAXQUOTAS; type++) {
 974                 if (!sb_has_quota_loaded(sb, type))
 975                         continue;
 976                 /* Cancel periodic syncing before we grab dqonoff_mutex */
 977                 oinfo = sb_dqinfo(sb, type)->dqi_priv;
 978                 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 979                 inode = igrab(sb->s_dquot.files[type]);
 980                 /* Turn off quotas. This will remove all dquot structures from
 981                  * memory and so they will be automatically synced to global
 982                  * quota files */
 983                 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
 984                                         DQUOT_LIMITS_ENABLED);
 985                 if (!inode)
 986                         continue;
 987                 iput(inode);
 988         }
 989 }
 990
 991 /* Handle quota on quotactl */
 992 static int ocfs2_quota_on(struct super_block *sb, int type, int format_id)
 993 {
 994         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 995                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 996
 997         if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 998                 return -EINVAL;
 999
1000         return dquot_enable(sb_dqopt(sb)->files[type], type,
1001                             format_id, DQUOT_LIMITS_ENABLED);
1002 }
1003
1004 /* Handle quota off quotactl */
1005 static int ocfs2_quota_off(struct super_block *sb, int type)
1006 {
1007         return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
1008 }
1009
1010 static const struct quotactl_ops ocfs2_quotactl_ops = {
1011         .quota_on_meta  = ocfs2_quota_on,
1012         .quota_off      = ocfs2_quota_off,
1013         .quota_sync     = dquot_quota_sync,
1014         .get_info       = dquot_get_dqinfo,
1015         .set_info       = dquot_set_dqinfo,
1016         .get_dqblk      = dquot_get_dqblk,
1017         .set_dqblk      = dquot_set_dqblk,
1018 };
1019
1020 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
1021 {
1022         struct dentry *root;
1023         int status, sector_size;
1024         struct mount_options parsed_options;
1025         struct inode *inode = NULL;
1026         struct ocfs2_super *osb = NULL;
1027         struct buffer_head *bh = NULL;
1028         char nodestr[8];
1029         struct ocfs2_blockcheck_stats stats;
1030
1031         mlog(0, "%p, %p, %i", sb, data, silent);
1032
1033         if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
1034                 status = -EINVAL;
1035                 goto read_super_error;
1036         }
1037
1038         /* probe for superblock */
1039         status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
1040         if (status < 0) {
1041                 mlog(ML_ERROR, "superblock probe failed!\n");
1042                 goto read_super_error;
1043         }
1044
1045         status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1046         osb = OCFS2_SB(sb);
1047         if (status < 0) {
1048                 mlog_errno(status);
1049                 goto read_super_error;
1050         }
1051         brelse(bh);
1052         bh = NULL;
1053
1054         if (!ocfs2_check_set_options(sb, &parsed_options)) {
1055                 status = -EINVAL;
1056                 goto read_super_error;
1057         }
1058         osb->s_mount_opt = parsed_options.mount_opt;
1059         osb->s_atime_quantum = parsed_options.atime_quantum;
1060         osb->preferred_slot = parsed_options.slot;
1061         osb->osb_commit_interval = parsed_options.commit_interval;
1062
1063         ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1064         osb->osb_resv_level = parsed_options.resv_level;
1065         osb->osb_dir_resv_level = parsed_options.resv_level;
1066         if (parsed_options.dir_resv_level == -1)
1067                 osb->osb_dir_resv_level = parsed_options.resv_level;
1068         else
1069                 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1070
1071         status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1072         if (status)
1073                 goto read_super_error;
1074
1075         sb->s_magic = OCFS2_SUPER_MAGIC;
1076
1077         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1078                 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1079
1080         /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1081          * heartbeat=none */
1082         if (bdev_read_only(sb->s_bdev)) {
1083                 if (!(sb->s_flags & MS_RDONLY)) {
1084                         status = -EACCES;
1085                         mlog(ML_ERROR, "Readonly device detected but readonly "
1086                              "mount was not specified.\n");
1087                         goto read_super_error;
1088                 }
1089
1090                 /* You should not be able to start a local heartbeat
1091                  * on a readonly device. */
1092                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1093                         status = -EROFS;
1094                         mlog(ML_ERROR, "Local heartbeat specified on readonly "
1095                              "device.\n");
1096                         goto read_super_error;
1097                 }
1098
1099                 status = ocfs2_check_journals_nolocks(osb);
1100                 if (status < 0) {
1101                         if (status == -EROFS)
1102                                 mlog(ML_ERROR, "Recovery required on readonly "
1103                                      "file system, but write access is "
1104                                      "unavailable.\n");
1105                         else
1106                                 mlog_errno(status);
1107                         goto read_super_error;
1108                 }
1109
1110                 ocfs2_set_ro_flag(osb, 1);
1111
1112                 printk(KERN_NOTICE "Readonly device detected. No cluster "
1113                        "services will be utilized for this mount. Recovery "
1114                        "will be skipped.\n");
1115         }
1116
1117         if (!ocfs2_is_hard_readonly(osb)) {
1118                 if (sb->s_flags & MS_RDONLY)
1119                         ocfs2_set_ro_flag(osb, 0);
1120         }
1121
1122         status = ocfs2_verify_heartbeat(osb);
1123         if (status < 0) {
1124                 mlog_errno(status);
1125                 goto read_super_error;
1126         }
1127
1128         osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1129                                                  ocfs2_debugfs_root);
1130         if (!osb->osb_debug_root) {
1131                 status = -EINVAL;
1132                 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1133                 goto read_super_error;
1134         }
1135
1136         osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1137                                             osb->osb_debug_root,
1138                                             osb,
1139                                             &ocfs2_osb_debug_fops);
1140         if (!osb->osb_ctxt) {
1141                 status = -EINVAL;
1142                 mlog_errno(status);
1143                 goto read_super_error;
1144         }
1145
1146         if (ocfs2_meta_ecc(osb)) {
1147                 status = ocfs2_blockcheck_stats_debugfs_install(
1148                                                 &osb->osb_ecc_stats,
1149                                                 osb->osb_debug_root);
1150                 if (status) {
1151                         mlog(ML_ERROR,
1152                              "Unable to create blockcheck statistics "
1153                              "files\n");
1154                         goto read_super_error;
1155                 }
1156         }
1157
1158         status = ocfs2_mount_volume(sb);
1159         if (osb->root_inode)
1160                 inode = igrab(osb->root_inode);
1161
1162         if (status < 0)
1163                 goto read_super_error;
1164
1165         if (!inode) {
1166                 status = -EIO;
1167                 mlog_errno(status);
1168                 goto read_super_error;
1169         }
1170
1171         root = d_alloc_root(inode);
1172         if (!root) {
1173                 status = -ENOMEM;
1174                 mlog_errno(status);
1175                 goto read_super_error;
1176         }
1177
1178         sb->s_root = root;
1179
1180         ocfs2_complete_mount_recovery(osb);
1181
1182         if (ocfs2_mount_local(osb))
1183                 snprintf(nodestr, sizeof(nodestr), "local");
1184         else
1185                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1186
1187         printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1188                "with %s data mode.\n",
1189                osb->dev_str, nodestr, osb->slot_num,
1190                osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1191                "ordered");
1192
1193         atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1194         wake_up(&osb->osb_mount_event);
1195
1196         /* Now we can initialize quotas because we can afford to wait
1197          * for cluster locks recovery now. That also means that truncation
1198          * log recovery can happen but that waits for proper quota setup */
1199         if (!(sb->s_flags & MS_RDONLY)) {
1200                 status = ocfs2_enable_quotas(osb);
1201                 if (status < 0) {
1202                         /* We have to err-out specially here because
1203                          * s_root is already set */
1204                         mlog_errno(status);
1205                         atomic_set(&osb->vol_state, VOLUME_DISABLED);
1206                         wake_up(&osb->osb_mount_event);
1207                         return status;
1208                 }
1209         }
1210
1211         ocfs2_complete_quota_recovery(osb);
1212
1213         /* Now we wake up again for processes waiting for quotas */
1214         atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1215         wake_up(&osb->osb_mount_event);
1216
1217         /* Start this when the mount is almost sure of being successful */
1218         ocfs2_orphan_scan_start(osb);
1219
1220         return status;
1221
1222 read_super_error:
1223         brelse(bh);
1224
1225         if (inode)
1226                 iput(inode);
1227
1228         if (osb) {
1229                 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1230                 wake_up(&osb->osb_mount_event);
1231                 ocfs2_dismount_volume(sb, 1);
1232         }
1233
1234         if (status)
1235                 mlog_errno(status);
1236         return status;
1237 }
1238
1239 static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
1240                         int flags,
1241                         const char *dev_name,
1242                         void *data)
1243 {
1244         return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
1245 }
1246
1247 static void ocfs2_kill_sb(struct super_block *sb)
1248 {
1249         struct ocfs2_super *osb = OCFS2_SB(sb);
1250
1251         /* Failed mount? */
1252         if (!osb || atomic_read(&osb->vol_state) == VOLUME_DISABLED)
1253                 goto out;
1254
1255         /* Prevent further queueing of inode drop events */
1256         spin_lock(&dentry_list_lock);
1257         ocfs2_set_osb_flag(osb, OCFS2_OSB_DROP_DENTRY_LOCK_IMMED);
1258         spin_unlock(&dentry_list_lock);
1259         /* Wait for work to finish and/or remove it */
1260         cancel_work_sync(&osb->dentry_lock_work);
1261 out:
1262         kill_block_super(sb);
1263 }
1264
1265 static struct file_system_type ocfs2_fs_type = {
1266         .owner          = THIS_MODULE,
1267         .name           = "ocfs2",
1268         .mount          = ocfs2_mount,
1269         .kill_sb        = ocfs2_kill_sb,
1270
1271         .fs_flags       = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1272         .next           = NULL
1273 };
1274
1275 static int ocfs2_check_set_options(struct super_block *sb,
1276                                    struct mount_options *options)
1277 {
1278         if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1279             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1280                                          OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1281                 mlog(ML_ERROR, "User quotas were requested, but this "
1282                      "filesystem does not have the feature enabled.\n");
1283                 return 0;
1284         }
1285         if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1286             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1287                                          OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1288                 mlog(ML_ERROR, "Group quotas were requested, but this "
1289                      "filesystem does not have the feature enabled.\n");
1290                 return 0;
1291         }
1292         if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1293             !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1294                 mlog(ML_ERROR, "ACL support requested but extended attributes "
1295                      "feature is not enabled\n");
1296                 return 0;
1297         }
1298         /* No ACL setting specified? Use XATTR feature... */
1299         if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1300                                     OCFS2_MOUNT_NO_POSIX_ACL))) {
1301                 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1302                         options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1303                 else
1304                         options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1305         }
1306         return 1;
1307 }
1308
1309 static int ocfs2_parse_options(struct super_block *sb,
1310                                char *options,
1311                                struct mount_options *mopt,
1312                                int is_remount)
1313 {
1314         int status, user_stack = 0;
1315         char *p;
1316         u32 tmp;
1317
1318         mlog(0, "remount: %d, options: \"%s\"\n", is_remount,
1319              options ? options : "(none)");
1320
1321         mopt->commit_interval = 0;
1322         mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1323         mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1324         mopt->slot = OCFS2_INVALID_SLOT;
1325         mopt->localalloc_opt = -1;
1326         mopt->cluster_stack[0] = '\0';
1327         mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1328         mopt->dir_resv_level = -1;
1329
1330         if (!options) {
1331                 status = 1;
1332                 goto bail;
1333         }
1334
1335         while ((p = strsep(&options, ",")) != NULL) {
1336                 int token, option;
1337                 substring_t args[MAX_OPT_ARGS];
1338
1339                 if (!*p)
1340                         continue;
1341
1342                 token = match_token(p, tokens, args);
1343                 switch (token) {
1344                 case Opt_hb_local:
1345                         mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1346                         break;
1347                 case Opt_hb_none:
1348                         mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1349                         break;
1350                 case Opt_hb_global:
1351                         mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1352                         break;
1353                 case Opt_barrier:
1354                         if (match_int(&args[0], &option)) {
1355                                 status = 0;
1356                                 goto bail;
1357                         }
1358                         if (option)
1359                                 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1360                         else
1361                                 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1362                         break;
1363                 case Opt_intr:
1364                         mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1365                         break;
1366                 case Opt_nointr:
1367                         mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1368                         break;
1369                 case Opt_err_panic:
1370                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1371                         break;
1372                 case Opt_err_ro:
1373                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1374                         break;
1375                 case Opt_data_ordered:
1376                         mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1377                         break;
1378                 case Opt_data_writeback:
1379                         mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1380                         break;
1381                 case Opt_user_xattr:
1382                         mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1383                         break;
1384                 case Opt_nouser_xattr:
1385                         mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1386                         break;
1387                 case Opt_atime_quantum:
1388                         if (match_int(&args[0], &option)) {
1389                                 status = 0;
1390                                 goto bail;
1391                         }
1392                         if (option >= 0)
1393                                 mopt->atime_quantum = option;
1394                         break;
1395                 case Opt_slot:
1396                         option = 0;
1397                         if (match_int(&args[0], &option)) {
1398                                 status = 0;
1399                                 goto bail;
1400                         }
1401                         if (option)
1402                                 mopt->slot = (s16)option;
1403                         break;
1404                 case Opt_commit:
1405                         option = 0;
1406                         if (match_int(&args[0], &option)) {
1407                                 status = 0;
1408                                 goto bail;
1409                         }
1410                         if (option < 0)
1411                                 return 0;
1412                         if (option == 0)
1413                                 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1414                         mopt->commit_interval = HZ * option;
1415                         break;
1416                 case Opt_localalloc:
1417                         option = 0;
1418                         if (match_int(&args[0], &option)) {
1419                                 status = 0;
1420                                 goto bail;
1421                         }
1422                         if (option >= 0)
1423                                 mopt->localalloc_opt = option;
1424                         break;
1425                 case Opt_localflocks:
1426                         /*
1427                          * Changing this during remount could race
1428                          * flock() requests, or "unbalance" existing
1429                          * ones (e.g., a lock is taken in one mode but
1430                          * dropped in the other). If users care enough
1431                          * to flip locking modes during remount, we
1432                          * could add a "local" flag to individual
1433                          * flock structures for proper tracking of
1434                          * state.
1435                          */
1436                         if (!is_remount)
1437                                 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1438                         break;
1439                 case Opt_stack:
1440                         /* Check both that the option we were passed
1441                          * is of the right length and that it is a proper
1442                          * string of the right length.
1443                          */
1444                         if (((args[0].to - args[0].from) !=
1445                              OCFS2_STACK_LABEL_LEN) ||
1446                             (strnlen(args[0].from,
1447                                      OCFS2_STACK_LABEL_LEN) !=
1448                              OCFS2_STACK_LABEL_LEN)) {
1449                                 mlog(ML_ERROR,
1450                                      "Invalid cluster_stack option\n");
1451                                 status = 0;
1452                                 goto bail;
1453                         }
1454                         memcpy(mopt->cluster_stack, args[0].from,
1455                                OCFS2_STACK_LABEL_LEN);
1456                         mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1457                         /*
1458                          * Open code the memcmp here as we don't have
1459                          * an osb to pass to
1460                          * ocfs2_userspace_stack().
1461                          */
1462                         if (memcmp(mopt->cluster_stack,
1463                                    OCFS2_CLASSIC_CLUSTER_STACK,
1464                                    OCFS2_STACK_LABEL_LEN))
1465                                 user_stack = 1;
1466                         break;
1467                 case Opt_inode64:
1468                         mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1469                         break;
1470                 case Opt_usrquota:
1471                         mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1472                         break;
1473                 case Opt_grpquota:
1474                         mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1475                         break;
1476                 case Opt_coherency_buffered:
1477                         mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1478                         break;
1479                 case Opt_coherency_full:
1480                         mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1481                         break;
1482                 case Opt_acl:
1483                         mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1484                         mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1485                         break;
1486                 case Opt_noacl:
1487                         mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1488                         mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1489                         break;
1490                 case Opt_resv_level:
1491                         if (is_remount)
1492                                 break;
1493                         if (match_int(&args[0], &option)) {
1494                                 status = 0;
1495                                 goto bail;
1496                         }
1497                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1498                             option < OCFS2_MAX_RESV_LEVEL)
1499                                 mopt->resv_level = option;
1500                         break;
1501                 case Opt_dir_resv_level:
1502                         if (is_remount)
1503                                 break;
1504                         if (match_int(&args[0], &option)) {
1505                                 status = 0;
1506                                 goto bail;
1507                         }
1508                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1509                             option < OCFS2_MAX_RESV_LEVEL)
1510                                 mopt->dir_resv_level = option;
1511                         break;
1512                 default:
1513                         mlog(ML_ERROR,
1514                              "Unrecognized mount option \"%s\" "
1515                              "or missing value\n", p);
1516                         status = 0;
1517                         goto bail;
1518                 }
1519         }
1520
1521         if (user_stack == 0) {
1522                 /* Ensure only one heartbeat mode */
1523                 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1524                                          OCFS2_MOUNT_HB_GLOBAL |
1525                                          OCFS2_MOUNT_HB_NONE);
1526                 if (hweight32(tmp) != 1) {
1527                         mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1528                         status = 0;
1529                         goto bail;
1530                 }
1531         }
1532
1533         status = 1;
1534
1535 bail:
1536         return status;
1537 }
1538
1539 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1540 {
1541         struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb);
1542         unsigned long opts = osb->s_mount_opt;
1543         unsigned int local_alloc_megs;
1544
1545         if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1546                 seq_printf(s, ",_netdev");
1547                 if (opts & OCFS2_MOUNT_HB_LOCAL)
1548                         seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1549                 else
1550                         seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1551         } else
1552                 seq_printf(s, ",%s", OCFS2_HB_NONE);
1553
1554         if (opts & OCFS2_MOUNT_NOINTR)
1555                 seq_printf(s, ",nointr");
1556
1557         if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1558                 seq_printf(s, ",data=writeback");
1559         else
1560                 seq_printf(s, ",data=ordered");
1561
1562         if (opts & OCFS2_MOUNT_BARRIER)
1563                 seq_printf(s, ",barrier=1");
1564
1565         if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1566                 seq_printf(s, ",errors=panic");
1567         else
1568                 seq_printf(s, ",errors=remount-ro");
1569
1570         if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1571                 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1572
1573         if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM)
1574                 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1575
1576         if (osb->osb_commit_interval)
1577                 seq_printf(s, ",commit=%u",
1578                            (unsigned) (osb->osb_commit_interval / HZ));
1579
1580         local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1581         if (local_alloc_megs != ocfs2_la_default_mb(osb))
1582                 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1583
1584         if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1585                 seq_printf(s, ",localflocks,");
1586
1587         if (osb->osb_cluster_stack[0])
1588                 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1589                            osb->osb_cluster_stack);
1590         if (opts & OCFS2_MOUNT_USRQUOTA)
1591                 seq_printf(s, ",usrquota");
1592         if (opts & OCFS2_MOUNT_GRPQUOTA)
1593                 seq_printf(s, ",grpquota");
1594
1595         if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1596                 seq_printf(s, ",coherency=buffered");
1597         else
1598                 seq_printf(s, ",coherency=full");
1599
1600         if (opts & OCFS2_MOUNT_NOUSERXATTR)
1601                 seq_printf(s, ",nouser_xattr");
1602         else
1603                 seq_printf(s, ",user_xattr");
1604
1605         if (opts & OCFS2_MOUNT_INODE64)
1606                 seq_printf(s, ",inode64");
1607
1608         if (opts & OCFS2_MOUNT_POSIX_ACL)
1609                 seq_printf(s, ",acl");
1610         else
1611                 seq_printf(s, ",noacl");
1612
1613         if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1614                 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1615
1616         if (osb->osb_dir_resv_level != osb->osb_resv_level)
1617                 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1618
1619         return 0;
1620 }
1621
1622 static int __init ocfs2_init(void)
1623 {
1624         int status;
1625
1626         ocfs2_print_version();
1627
1628         status = init_ocfs2_uptodate_cache();
1629         if (status < 0) {
1630                 mlog_errno(status);
1631                 goto leave;
1632         }
1633
1634         status = ocfs2_initialize_mem_caches();
1635         if (status < 0) {
1636                 mlog_errno(status);
1637                 goto leave;
1638         }
1639
1640         ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1641         if (!ocfs2_wq) {
1642                 status = -ENOMEM;
1643                 goto leave;
1644         }
1645
1646         ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1647         if (!ocfs2_debugfs_root) {
1648                 status = -EFAULT;
1649                 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1650         }
1651
1652         status = ocfs2_quota_setup();
1653         if (status)
1654                 goto leave;
1655
1656         ocfs2_set_locking_protocol();
1657
1658         status = register_quota_format(&ocfs2_quota_format);
1659 leave:
1660         if (status < 0) {
1661                 ocfs2_quota_shutdown();
1662                 ocfs2_free_mem_caches();
1663                 exit_ocfs2_uptodate_cache();
1664                 mlog_errno(status);
1665         }
1666
1667         if (status >= 0) {
1668                 return register_filesystem(&ocfs2_fs_type);
1669         } else
1670                 return -1;
1671 }
1672
1673 static void __exit ocfs2_exit(void)
1674 {
1675         ocfs2_quota_shutdown();
1676
1677         if (ocfs2_wq) {
1678                 flush_workqueue(ocfs2_wq);
1679                 destroy_workqueue(ocfs2_wq);
1680         }
1681
1682         unregister_quota_format(&ocfs2_quota_format);
1683
1684         debugfs_remove(ocfs2_debugfs_root);
1685
1686         ocfs2_free_mem_caches();
1687
1688         unregister_filesystem(&ocfs2_fs_type);
1689
1690         exit_ocfs2_uptodate_cache();
1691 }
1692
1693 static void ocfs2_put_super(struct super_block *sb)
1694 {
1695         mlog(0, "(0x%p)\n", sb);
1696
1697         ocfs2_sync_blockdev(sb);
1698         ocfs2_dismount_volume(sb, 0);
1699 }
1700
1701 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1702 {
1703         struct ocfs2_super *osb;
1704         u32 numbits, freebits;
1705         int status;
1706         struct ocfs2_dinode *bm_lock;
1707         struct buffer_head *bh = NULL;
1708         struct inode *inode = NULL;
1709
1710         mlog(0, "(%p, %p)\n", dentry->d_sb, buf);
1711
1712         osb = OCFS2_SB(dentry->d_sb);
1713
1714         inode = ocfs2_get_system_file_inode(osb,
1715                                             GLOBAL_BITMAP_SYSTEM_INODE,
1716                                             OCFS2_INVALID_SLOT);
1717         if (!inode) {
1718                 mlog(ML_ERROR, "failed to get bitmap inode\n");
1719                 status = -EIO;
1720                 goto bail;
1721         }
1722
1723         status = ocfs2_inode_lock(inode, &bh, 0);
1724         if (status < 0) {
1725                 mlog_errno(status);
1726                 goto bail;
1727         }
1728
1729         bm_lock = (struct ocfs2_dinode *) bh->b_data;
1730
1731         numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1732         freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1733
1734         buf->f_type = OCFS2_SUPER_MAGIC;
1735         buf->f_bsize = dentry->d_sb->s_blocksize;
1736         buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1737         buf->f_blocks = ((sector_t) numbits) *
1738                         (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1739         buf->f_bfree = ((sector_t) freebits) *
1740                        (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1741         buf->f_bavail = buf->f_bfree;
1742         buf->f_files = numbits;
1743         buf->f_ffree = freebits;
1744         buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1745                                 & 0xFFFFFFFFUL;
1746         buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1747                                 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1748
1749         brelse(bh);
1750
1751         ocfs2_inode_unlock(inode, 0);
1752         status = 0;
1753 bail:
1754         if (inode)
1755                 iput(inode);
1756
1757         if (status)
1758                 mlog_errno(status);
1759
1760         return status;
1761 }
1762
1763 static void ocfs2_inode_init_once(void *data)
1764 {
1765         struct ocfs2_inode_info *oi = data;
1766
1767         oi->ip_flags = 0;
1768         oi->ip_open_count = 0;
1769         spin_lock_init(&oi->ip_lock);
1770         ocfs2_extent_map_init(&oi->vfs_inode);
1771         INIT_LIST_HEAD(&oi->ip_io_markers);
1772         oi->ip_dir_start_lookup = 0;
1773
1774         init_rwsem(&oi->ip_alloc_sem);
1775         init_rwsem(&oi->ip_xattr_sem);
1776         mutex_init(&oi->ip_io_mutex);
1777
1778         oi->ip_blkno = 0ULL;
1779         oi->ip_clusters = 0;
1780
1781         ocfs2_resv_init_once(&oi->ip_la_data_resv);
1782
1783         ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1784         ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1785         ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1786
1787         ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1788                                   &ocfs2_inode_caching_ops);
1789
1790         inode_init_once(&oi->vfs_inode);
1791 }
1792
1793 static int ocfs2_initialize_mem_caches(void)
1794 {
1795         ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1796                                        sizeof(struct ocfs2_inode_info),
1797                                        0,
1798                                        (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1799                                                 SLAB_MEM_SPREAD),
1800                                        ocfs2_inode_init_once);
1801         ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1802                                         sizeof(struct ocfs2_dquot),
1803                                         0,
1804                                         (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1805                                                 SLAB_MEM_SPREAD),
1806                                         NULL);
1807         ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1808                                         sizeof(struct ocfs2_quota_chunk),
1809                                         0,
1810                                         (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1811                                         NULL);
1812         if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1813             !ocfs2_qf_chunk_cachep) {
1814                 if (ocfs2_inode_cachep)
1815                         kmem_cache_destroy(ocfs2_inode_cachep);
1816                 if (ocfs2_dquot_cachep)
1817                         kmem_cache_destroy(ocfs2_dquot_cachep);
1818                 if (ocfs2_qf_chunk_cachep)
1819                         kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1820                 return -ENOMEM;
1821         }
1822
1823         return 0;
1824 }
1825
1826 static void ocfs2_free_mem_caches(void)
1827 {
1828         if (ocfs2_inode_cachep)
1829                 kmem_cache_destroy(ocfs2_inode_cachep);
1830         ocfs2_inode_cachep = NULL;
1831
1832         if (ocfs2_dquot_cachep)
1833                 kmem_cache_destroy(ocfs2_dquot_cachep);
1834         ocfs2_dquot_cachep = NULL;
1835
1836         if (ocfs2_qf_chunk_cachep)
1837                 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1838         ocfs2_qf_chunk_cachep = NULL;
1839 }
1840
1841 static int ocfs2_get_sector(struct super_block *sb,
1842                             struct buffer_head **bh,
1843                             int block,
1844                             int sect_size)
1845 {
1846         if (!sb_set_blocksize(sb, sect_size)) {
1847                 mlog(ML_ERROR, "unable to set blocksize\n");
1848                 return -EIO;
1849         }
1850
1851         *bh = sb_getblk(sb, block);
1852         if (!*bh) {
1853                 mlog_errno(-EIO);
1854                 return -EIO;
1855         }
1856         lock_buffer(*bh);
1857         if (!buffer_dirty(*bh))
1858                 clear_buffer_uptodate(*bh);
1859         unlock_buffer(*bh);
1860         ll_rw_block(READ, 1, bh);
1861         wait_on_buffer(*bh);
1862         if (!buffer_uptodate(*bh)) {
1863                 mlog_errno(-EIO);
1864                 brelse(*bh);
1865                 *bh = NULL;
1866                 return -EIO;
1867         }
1868
1869         return 0;
1870 }
1871
1872 static int ocfs2_mount_volume(struct super_block *sb)
1873 {
1874         int status = 0;
1875         int unlock_super = 0;
1876         struct ocfs2_super *osb = OCFS2_SB(sb);
1877
1878         if (ocfs2_is_hard_readonly(osb))
1879                 goto leave;
1880
1881         status = ocfs2_dlm_init(osb);
1882         if (status < 0) {
1883                 mlog_errno(status);
1884                 goto leave;
1885         }
1886
1887         status = ocfs2_super_lock(osb, 1);
1888         if (status < 0) {
1889                 mlog_errno(status);
1890                 goto leave;
1891         }
1892         unlock_super = 1;
1893
1894         /* This will load up the node map and add ourselves to it. */
1895         status = ocfs2_find_slot(osb);
1896         if (status < 0) {
1897                 mlog_errno(status);
1898                 goto leave;
1899         }
1900
1901         /* load all node-local system inodes */
1902         status = ocfs2_init_local_system_inodes(osb);
1903         if (status < 0) {
1904                 mlog_errno(status);
1905                 goto leave;
1906         }
1907
1908         status = ocfs2_check_volume(osb);
1909         if (status < 0) {
1910                 mlog_errno(status);
1911                 goto leave;
1912         }
1913
1914         status = ocfs2_truncate_log_init(osb);
1915         if (status < 0)
1916                 mlog_errno(status);
1917
1918 leave:
1919         if (unlock_super)
1920                 ocfs2_super_unlock(osb, 1);
1921
1922         return status;
1923 }
1924
1925 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1926 {
1927         int tmp, hangup_needed = 0;
1928         struct ocfs2_super *osb = NULL;
1929         char nodestr[8];
1930
1931         mlog(0, "(0x%p)\n", sb);
1932
1933         BUG_ON(!sb);
1934         osb = OCFS2_SB(sb);
1935         BUG_ON(!osb);
1936
1937         debugfs_remove(osb->osb_ctxt);
1938
1939         /*
1940          * Flush inode dropping work queue so that deletes are
1941          * performed while the filesystem is still working
1942          */
1943         ocfs2_drop_all_dl_inodes(osb);
1944
1945         /* Orphan scan should be stopped as early as possible */
1946         ocfs2_orphan_scan_stop(osb);
1947
1948         ocfs2_disable_quotas(osb);
1949
1950         ocfs2_shutdown_local_alloc(osb);
1951
1952         ocfs2_truncate_log_shutdown(osb);
1953
1954         /* This will disable recovery and flush any recovery work. */
1955         ocfs2_recovery_exit(osb);
1956
1957         ocfs2_journal_shutdown(osb);
1958
1959         ocfs2_sync_blockdev(sb);
1960
1961         ocfs2_purge_refcount_trees(osb);
1962
1963         /* No cluster connection means we've failed during mount, so skip
1964          * all the steps which depended on that to complete. */
1965         if (osb->cconn) {
1966                 tmp = ocfs2_super_lock(osb, 1);
1967                 if (tmp < 0) {
1968                         mlog_errno(tmp);
1969                         return;
1970                 }
1971         }
1972
1973         if (osb->slot_num != OCFS2_INVALID_SLOT)
1974                 ocfs2_put_slot(osb);
1975
1976         if (osb->cconn)
1977                 ocfs2_super_unlock(osb, 1);
1978
1979         ocfs2_release_system_inodes(osb);
1980
1981         /*
1982          * If we're dismounting due to mount error, mount.ocfs2 will clean
1983          * up heartbeat.  If we're a local mount, there is no heartbeat.
1984          * If we failed before we got a uuid_str yet, we can't stop
1985          * heartbeat.  Otherwise, do it.
1986          */
1987         if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str)
1988                 hangup_needed = 1;
1989
1990         if (osb->cconn)
1991                 ocfs2_dlm_shutdown(osb, hangup_needed);
1992
1993         ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1994         debugfs_remove(osb->osb_debug_root);
1995
1996         if (hangup_needed)
1997                 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1998
1999         atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
2000
2001         if (ocfs2_mount_local(osb))
2002                 snprintf(nodestr, sizeof(nodestr), "local");
2003         else
2004                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
2005
2006         printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
2007                osb->dev_str, nodestr);
2008
2009         ocfs2_delete_osb(osb);
2010         kfree(osb);
2011         sb->s_dev = 0;
2012         sb->s_fs_info = NULL;
2013 }
2014
2015 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
2016                                 unsigned uuid_bytes)
2017 {
2018         int i, ret;
2019         char *ptr;
2020
2021         BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
2022
2023         osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
2024         if (osb->uuid_str == NULL)
2025                 return -ENOMEM;
2026
2027         for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
2028                 /* print with null */
2029                 ret = snprintf(ptr, 3, "%02X", uuid[i]);
2030                 if (ret != 2) /* drop super cleans up */
2031                         return -EINVAL;
2032                 /* then only advance past the last char */
2033                 ptr += 2;
2034         }
2035
2036         return 0;
2037 }
2038
2039 /* Make sure entire volume is addressable by our journal.  Requires
2040    osb_clusters_at_boot to be valid and for the journal to have been
2041    initialized by ocfs2_journal_init(). */
2042 static int ocfs2_journal_addressable(struct ocfs2_super *osb)
2043 {
2044         int status = 0;
2045         u64 max_block =
2046                 ocfs2_clusters_to_blocks(osb->sb,
2047                                          osb->osb_clusters_at_boot) - 1;
2048
2049         /* 32-bit block number is always OK. */
2050         if (max_block <= (u32)~0ULL)
2051                 goto out;
2052
2053         /* Volume is "huge", so see if our journal is new enough to
2054            support it. */
2055         if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2056                                        OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2057               jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2058                                                JBD2_FEATURE_INCOMPAT_64BIT))) {
2059                 mlog(ML_ERROR, "The journal cannot address the entire volume. "
2060                      "Enable the 'block64' journal option with tunefs.ocfs2");
2061                 status = -EFBIG;
2062                 goto out;
2063         }
2064
2065  out:
2066         return status;
2067 }
2068
2069 static int ocfs2_initialize_super(struct super_block *sb,
2070                                   struct buffer_head *bh,
2071                                   int sector_size,
2072                                   struct ocfs2_blockcheck_stats *stats)
2073 {
2074         int status;
2075         int i, cbits, bbits;
2076         struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2077         struct inode *inode = NULL;
2078         struct ocfs2_journal *journal;
2079         __le32 uuid_net_key;
2080         struct ocfs2_super *osb;
2081         u64 total_blocks;
2082
2083         osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2084         if (!osb) {
2085                 status = -ENOMEM;
2086                 mlog_errno(status);
2087                 goto bail;
2088         }
2089
2090         sb->s_fs_info = osb;
2091         sb->s_op = &ocfs2_sops;
2092         sb->s_d_op = &ocfs2_dentry_ops;
2093         sb->s_export_op = &ocfs2_export_ops;
2094         sb->s_qcop = &ocfs2_quotactl_ops;
2095         sb->dq_op = &ocfs2_quota_operations;
2096         sb->s_xattr = ocfs2_xattr_handlers;
2097         sb->s_time_gran = 1;
2098         sb->s_flags |= MS_NOATIME;
2099         /* this is needed to support O_LARGEFILE */
2100         cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2101         bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2102         sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2103
2104         osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2105
2106         for (i = 0; i < 3; i++)
2107                 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2108         osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2109
2110         osb->sb = sb;
2111         /* Save off for ocfs2_rw_direct */
2112         osb->s_sectsize_bits = blksize_bits(sector_size);
2113         BUG_ON(!osb->s_sectsize_bits);
2114
2115         spin_lock_init(&osb->dc_task_lock);
2116         init_waitqueue_head(&osb->dc_event);
2117         osb->dc_work_sequence = 0;
2118         osb->dc_wake_sequence = 0;
2119         INIT_LIST_HEAD(&osb->blocked_lock_list);
2120         osb->blocked_lock_count = 0;
2121         spin_lock_init(&osb->osb_lock);
2122         spin_lock_init(&osb->osb_xattr_lock);
2123         ocfs2_init_steal_slots(osb);
2124
2125         atomic_set(&osb->alloc_stats.moves, 0);
2126         atomic_set(&osb->alloc_stats.local_data, 0);
2127         atomic_set(&osb->alloc_stats.bitmap_data, 0);
2128         atomic_set(&osb->alloc_stats.bg_allocs, 0);
2129         atomic_set(&osb->alloc_stats.bg_extends, 0);
2130
2131         /* Copy the blockcheck stats from the superblock probe */
2132         osb->osb_ecc_stats = *stats;
2133
2134         ocfs2_init_node_maps(osb);
2135
2136         snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2137                  MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2138
2139         osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2140         if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2141                 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2142                      osb->max_slots);
2143                 status = -EINVAL;
2144                 goto bail;
2145         }
2146         mlog(0, "max_slots for this device: %u\n", osb->max_slots);
2147
2148         ocfs2_orphan_scan_init(osb);
2149
2150         status = ocfs2_recovery_init(osb);
2151         if (status) {
2152                 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2153                 mlog_errno(status);
2154                 goto bail;
2155         }
2156
2157         init_waitqueue_head(&osb->checkpoint_event);
2158         atomic_set(&osb->needs_checkpoint, 0);
2159
2160         osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2161
2162         osb->slot_num = OCFS2_INVALID_SLOT;
2163
2164         osb->s_xattr_inline_size = le16_to_cpu(
2165                                         di->id2.i_super.s_xattr_inline_size);
2166
2167         osb->local_alloc_state = OCFS2_LA_UNUSED;
2168         osb->local_alloc_bh = NULL;
2169         INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2170
2171         init_waitqueue_head(&osb->osb_mount_event);
2172
2173         status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2174         if (status) {
2175                 mlog_errno(status);
2176                 goto bail;
2177         }
2178
2179         osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2180         if (!osb->vol_label) {
2181                 mlog(ML_ERROR, "unable to alloc vol label\n");
2182                 status = -ENOMEM;
2183                 goto bail;
2184         }
2185
2186         osb->slot_recovery_generations =
2187                 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2188                         GFP_KERNEL);
2189         if (!osb->slot_recovery_generations) {
2190                 status = -ENOMEM;
2191                 mlog_errno(status);
2192                 goto bail;
2193         }
2194
2195         init_waitqueue_head(&osb->osb_wipe_event);
2196         osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2197                                         sizeof(*osb->osb_orphan_wipes),
2198                                         GFP_KERNEL);
2199         if (!osb->osb_orphan_wipes) {
2200                 status = -ENOMEM;
2201                 mlog_errno(status);
2202                 goto bail;
2203         }
2204
2205         osb->osb_rf_lock_tree = RB_ROOT;
2206
2207         osb->s_feature_compat =
2208                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2209         osb->s_feature_ro_compat =
2210                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2211         osb->s_feature_incompat =
2212                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2213
2214         if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2215                 mlog(ML_ERROR, "couldn't mount because of unsupported "
2216                      "optional features (%x).\n", i);
2217                 status = -EINVAL;
2218                 goto bail;
2219         }
2220         if (!(osb->sb->s_flags & MS_RDONLY) &&
2221             (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2222                 mlog(ML_ERROR, "couldn't mount RDWR because of "
2223                      "unsupported optional features (%x).\n", i);
2224                 status = -EINVAL;
2225                 goto bail;
2226         }
2227
2228         if (ocfs2_clusterinfo_valid(osb)) {
2229                 osb->osb_stackflags =
2230                         OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2231                 memcpy(osb->osb_cluster_stack,
2232                        OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2233                        OCFS2_STACK_LABEL_LEN);
2234                 osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
2235                 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2236                         mlog(ML_ERROR,
2237                              "couldn't mount because of an invalid "
2238                              "cluster stack label (%s) \n",
2239                              osb->osb_cluster_stack);
2240                         status = -EINVAL;
2241                         goto bail;
2242                 }
2243         } else {
2244                 /* The empty string is identical with classic tools that
2245                  * don't know about s_cluster_info. */
2246                 osb->osb_cluster_stack[0] = '\0';
2247         }
2248
2249         get_random_bytes(&osb->s_next_generation, sizeof(u32));
2250
2251         /* FIXME
2252          * This should be done in ocfs2_journal_init(), but unknown
2253          * ordering issues will cause the filesystem to crash.
2254          * If anyone wants to figure out what part of the code
2255          * refers to osb->journal before ocfs2_journal_init() is run,
2256          * be my guest.
2257          */
2258         /* initialize our journal structure */
2259
2260         journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2261         if (!journal) {
2262                 mlog(ML_ERROR, "unable to alloc journal\n");
2263                 status = -ENOMEM;
2264                 goto bail;
2265         }
2266         osb->journal = journal;
2267         journal->j_osb = osb;
2268
2269         atomic_set(&journal->j_num_trans, 0);
2270         init_rwsem(&journal->j_trans_barrier);
2271         init_waitqueue_head(&journal->j_checkpointed);
2272         spin_lock_init(&journal->j_lock);
2273         journal->j_trans_id = (unsigned long) 1;
2274         INIT_LIST_HEAD(&journal->j_la_cleanups);
2275         INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2276         journal->j_state = OCFS2_JOURNAL_FREE;
2277
2278         INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes);
2279         osb->dentry_lock_list = NULL;
2280
2281         /* get some pseudo constants for clustersize bits */
2282         osb->s_clustersize_bits =
2283                 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2284         osb->s_clustersize = 1 << osb->s_clustersize_bits;
2285         mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
2286
2287         if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2288             osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2289                 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2290                      osb->s_clustersize);
2291                 status = -EINVAL;
2292                 goto bail;
2293         }
2294
2295         total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2296                                                 le32_to_cpu(di->i_clusters));
2297
2298         status = generic_check_addressable(osb->sb->s_blocksize_bits,
2299                                            total_blocks);
2300         if (status) {
2301                 mlog(ML_ERROR, "Volume too large "
2302                      "to mount safely on this system");
2303                 status = -EFBIG;
2304                 goto bail;
2305         }
2306
2307         if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2308                                  sizeof(di->id2.i_super.s_uuid))) {
2309                 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2310                 status = -ENOMEM;
2311                 goto bail;
2312         }
2313
2314         memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
2315
2316         strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2317         osb->vol_label[63] = '\0';
2318         osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2319         osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2320         osb->first_cluster_group_blkno =
2321                 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2322         osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2323         osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2324         mlog(0, "vol_label: %s\n", osb->vol_label);
2325         mlog(0, "uuid: %s\n", osb->uuid_str);
2326         mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
2327              (unsigned long long)osb->root_blkno,
2328              (unsigned long long)osb->system_dir_blkno);
2329
2330         osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2331         if (!osb->osb_dlm_debug) {
2332                 status = -ENOMEM;
2333                 mlog_errno(status);
2334                 goto bail;
2335         }
2336
2337         atomic_set(&osb->vol_state, VOLUME_INIT);
2338
2339         /* load root, system_dir, and all global system inodes */
2340         status = ocfs2_init_global_system_inodes(osb);
2341         if (status < 0) {
2342                 mlog_errno(status);
2343                 goto bail;
2344         }
2345
2346         /*
2347          * global bitmap
2348          */
2349         inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2350                                             OCFS2_INVALID_SLOT);
2351         if (!inode) {
2352                 status = -EINVAL;
2353                 mlog_errno(status);
2354                 goto bail;
2355         }
2356
2357         osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2358         osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2359         iput(inode);
2360
2361         osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2362                                  osb->s_feature_incompat) * 8;
2363
2364         status = ocfs2_init_slot_info(osb);
2365         if (status < 0) {
2366                 mlog_errno(status);
2367                 goto bail;
2368         }
2369
2370 bail:
2371         return status;
2372 }
2373
2374 /*
2375  * will return: -EAGAIN if it is ok to keep searching for superblocks
2376  *              -EINVAL if there is a bad superblock
2377  *              0 on success
2378  */
2379 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2380                                struct buffer_head *bh,
2381                                u32 blksz,
2382                                struct ocfs2_blockcheck_stats *stats)
2383 {
2384         int status = -EAGAIN;
2385
2386         if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2387                    strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2388                 /* We have to do a raw check of the feature here */
2389                 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2390                     OCFS2_FEATURE_INCOMPAT_META_ECC) {
2391                         status = ocfs2_block_check_validate(bh->b_data,
2392                                                             bh->b_size,
2393                                                             &di->i_check,
2394                                                             stats);
2395                         if (status)
2396                                 goto out;
2397                 }
2398                 status = -EINVAL;
2399                 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2400                         mlog(ML_ERROR, "found superblock with incorrect block "
2401                              "size: found %u, should be %u\n",
2402                              1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2403                                blksz);
2404                 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2405                            OCFS2_MAJOR_REV_LEVEL ||
2406                            le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2407                            OCFS2_MINOR_REV_LEVEL) {
2408                         mlog(ML_ERROR, "found superblock with bad version: "
2409                              "found %u.%u, should be %u.%u\n",
2410                              le16_to_cpu(di->id2.i_super.s_major_rev_level),
2411                              le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2412                              OCFS2_MAJOR_REV_LEVEL,
2413                              OCFS2_MINOR_REV_LEVEL);
2414                 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2415                         mlog(ML_ERROR, "bad block number on superblock: "
2416                              "found %llu, should be %llu\n",
2417                              (unsigned long long)le64_to_cpu(di->i_blkno),
2418                              (unsigned long long)bh->b_blocknr);
2419                 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2420                             le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2421                         mlog(ML_ERROR, "bad cluster size found: %u\n",
2422                              1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2423                 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2424                         mlog(ML_ERROR, "bad root_blkno: 0\n");
2425                 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2426                         mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2427                 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2428                         mlog(ML_ERROR,
2429                              "Superblock slots found greater than file system "
2430                              "maximum: found %u, max %u\n",
2431                              le16_to_cpu(di->id2.i_super.s_max_slots),
2432                              OCFS2_MAX_SLOTS);
2433                 } else {
2434                         /* found it! */
2435                         status = 0;
2436                 }
2437         }
2438
2439 out:
2440         if (status && status != -EAGAIN)
2441                 mlog_errno(status);
2442         return status;
2443 }
2444
2445 static int ocfs2_check_volume(struct ocfs2_super *osb)
2446 {
2447         int status;
2448         int dirty;
2449         int local;
2450         struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2451                                                   * recover
2452                                                   * ourselves. */
2453
2454         /* Init our journal object. */
2455         status = ocfs2_journal_init(osb->journal, &dirty);
2456         if (status < 0) {
2457                 mlog(ML_ERROR, "Could not initialize journal!\n");
2458                 goto finally;
2459         }
2460
2461         /* Now that journal has been initialized, check to make sure
2462            entire volume is addressable. */
2463         status = ocfs2_journal_addressable(osb);
2464         if (status)
2465                 goto finally;
2466
2467         /* If the journal was unmounted cleanly then we don't want to
2468          * recover anything. Otherwise, journal_load will do that
2469          * dirty work for us :) */
2470         if (!dirty) {
2471                 status = ocfs2_journal_wipe(osb->journal, 0);
2472                 if (status < 0) {
2473                         mlog_errno(status);
2474                         goto finally;
2475                 }
2476         } else {
2477                 mlog(ML_NOTICE, "File system was not unmounted cleanly, "
2478                      "recovering volume.\n");
2479         }
2480
2481         local = ocfs2_mount_local(osb);
2482
2483         /* will play back anything left in the journal. */
2484         status = ocfs2_journal_load(osb->journal, local, dirty);
2485         if (status < 0) {
2486                 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2487                 goto finally;
2488         }
2489
2490         if (dirty) {
2491                 /* recover my local alloc if we didn't unmount cleanly. */
2492                 status = ocfs2_begin_local_alloc_recovery(osb,
2493                                                           osb->slot_num,
2494                                                           &local_alloc);
2495                 if (status < 0) {
2496                         mlog_errno(status);
2497                         goto finally;
2498                 }
2499                 /* we complete the recovery process after we've marked
2500                  * ourselves as mounted. */
2501         }
2502
2503         mlog(0, "Journal loaded.\n");
2504
2505         status = ocfs2_load_local_alloc(osb);
2506         if (status < 0) {
2507                 mlog_errno(status);
2508                 goto finally;
2509         }
2510
2511         if (dirty) {
2512                 /* Recovery will be completed after we've mounted the
2513                  * rest of the volume. */
2514                 osb->dirty = 1;
2515                 osb->local_alloc_copy = local_alloc;
2516                 local_alloc = NULL;
2517         }
2518
2519         /* go through each journal, trylock it and if you get the
2520          * lock, and it's marked as dirty, set the bit in the recover
2521          * map and launch a recovery thread for it. */
2522         status = ocfs2_mark_dead_nodes(osb);
2523         if (status < 0) {
2524                 mlog_errno(status);
2525                 goto finally;
2526         }
2527
2528         status = ocfs2_compute_replay_slots(osb);
2529         if (status < 0)
2530                 mlog_errno(status);
2531
2532 finally:
2533         if (local_alloc)
2534                 kfree(local_alloc);
2535
2536         if (status)
2537                 mlog_errno(status);
2538         return status;
2539 }
2540
2541 /*
2542  * The routine gets called from dismount or close whenever a dismount on
2543  * volume is requested and the osb open count becomes 1.
2544  * It will remove the osb from the global list and also free up all the
2545  * initialized resources and fileobject.
2546  */
2547 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2548 {
2549         /* This function assumes that the caller has the main osb resource */
2550
2551         ocfs2_free_slot_info(osb);
2552
2553         kfree(osb->osb_orphan_wipes);
2554         kfree(osb->slot_recovery_generations);
2555         /* FIXME
2556          * This belongs in journal shutdown, but because we have to
2557          * allocate osb->journal at the start of ocfs2_initialize_osb(),
2558          * we free it here.
2559          */
2560         kfree(osb->journal);
2561         if (osb->local_alloc_copy)
2562                 kfree(osb->local_alloc_copy);
2563         kfree(osb->uuid_str);
2564         ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2565         memset(osb, 0, sizeof(struct ocfs2_super));
2566 }
2567
2568 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2569  * panic(). We do not support continue-on-error operation. */
2570 static void ocfs2_handle_error(struct super_block *sb)
2571 {
2572         struct ocfs2_super *osb = OCFS2_SB(sb);
2573
2574         if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2575                 panic("OCFS2: (device %s): panic forced after error\n",
2576                       sb->s_id);
2577
2578         ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2579
2580         if (sb->s_flags & MS_RDONLY &&
2581             (ocfs2_is_soft_readonly(osb) ||
2582              ocfs2_is_hard_readonly(osb)))
2583                 return;
2584
2585         printk(KERN_CRIT "File system is now read-only due to the potential "
2586                "of on-disk corruption. Please run fsck.ocfs2 once the file "
2587                "system is unmounted.\n");
2588         sb->s_flags |= MS_RDONLY;
2589         ocfs2_set_ro_flag(osb, 0);
2590 }
2591
2592 static char error_buf[1024];
2593
2594 void __ocfs2_error(struct super_block *sb,
2595                    const char *function,
2596                    const char *fmt, ...)
2597 {
2598         va_list args;
2599
2600         va_start(args, fmt);
2601         vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2602         va_end(args);
2603
2604         /* Not using mlog here because we want to show the actual
2605          * function the error came from. */
2606         printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2607                sb->s_id, function, error_buf);
2608
2609         ocfs2_handle_error(sb);
2610 }
2611
2612 /* Handle critical errors. This is intentionally more drastic than
2613  * ocfs2_handle_error, so we only use for things like journal errors,
2614  * etc. */
2615 void __ocfs2_abort(struct super_block* sb,
2616                    const char *function,
2617                    const char *fmt, ...)
2618 {
2619         va_list args;
2620
2621         va_start(args, fmt);
2622         vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2623         va_end(args);
2624
2625         printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2626                sb->s_id, function, error_buf);
2627
2628         /* We don't have the cluster support yet to go straight to
2629          * hard readonly in here. Until then, we want to keep
2630          * ocfs2_abort() so that we can at least mark critical
2631          * errors.
2632          *
2633          * TODO: This should abort the journal and alert other nodes
2634          * that our slot needs recovery. */
2635
2636         /* Force a panic(). This stinks, but it's better than letting
2637          * things continue without having a proper hard readonly
2638          * here. */
2639         if (!ocfs2_mount_local(OCFS2_SB(sb)))
2640                 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2641         ocfs2_handle_error(sb);
2642 }
2643
2644 /*
2645  * Void signal blockers, because in-kernel sigprocmask() only fails
2646  * when SIG_* is wrong.
2647  */
2648 void ocfs2_block_signals(sigset_t *oldset)
2649 {
2650         int rc;
2651         sigset_t blocked;
2652
2653         sigfillset(&blocked);
2654         rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2655         BUG_ON(rc);
2656 }
2657
2658 void ocfs2_unblock_signals(sigset_t *oldset)
2659 {
2660         int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2661         BUG_ON(rc);
2662 }
2663
2664 module_init(ocfs2_init);
2665 module_exit(ocfs2_exit);