~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/fs/ocfs2/super.c

Version: ~ [ linux-5.13-rc5 ] ~ [ linux-5.12.9 ] ~ [ linux-5.11.22 ] ~ [ linux-5.10.42 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.124 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.193 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.235 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.271 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.271 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.18.140 ] ~ [ linux-3.16.85 ] ~ [ linux-3.14.79 ] ~ [ linux-3.12.74 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 /* -*- mode: c; c-basic-offset: 8; -*-
  2  * vim: noexpandtab sw=8 ts=8 sts=0:
  3  *
  4  * super.c
  5  *
  6  * load/unload driver, mount/dismount volumes
  7  *
  8  * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
  9  *
 10  * This program is free software; you can redistribute it and/or
 11  * modify it under the terms of the GNU General Public
 12  * License as published by the Free Software Foundation; either
 13  * version 2 of the License, or (at your option) any later version.
 14  *
 15  * This program is distributed in the hope that it will be useful,
 16  * but WITHOUT ANY WARRANTY; without even the implied warranty of
 17  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 18  * General Public License for more details.
 19  *
 20  * You should have received a copy of the GNU General Public
 21  * License along with this program; if not, write to the
 22  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
 23  * Boston, MA 021110-1307, USA.
 24  */
 25 
 26 #include <linux/module.h>
 27 #include <linux/fs.h>
 28 #include <linux/types.h>
 29 #include <linux/slab.h>
 30 #include <linux/highmem.h>
 31 #include <linux/init.h>
 32 #include <linux/random.h>
 33 #include <linux/statfs.h>
 34 #include <linux/moduleparam.h>
 35 #include <linux/blkdev.h>
 36 #include <linux/socket.h>
 37 #include <linux/inet.h>
 38 #include <linux/parser.h>
 39 #include <linux/crc32.h>
 40 #include <linux/debugfs.h>
 41 #include <linux/mount.h>
 42 #include <linux/seq_file.h>
 43 #include <linux/quotaops.h>
 44 #include <linux/cleancache.h>
 45 #include <linux/signal.h>
 46 
 47 #define CREATE_TRACE_POINTS
 48 #include "ocfs2_trace.h"
 49 
 50 #include <cluster/masklog.h>
 51 
 52 #include "ocfs2.h"
 53 
 54 /* this should be the only file to include a version 1 header */
 55 #include "ocfs1_fs_compat.h"
 56 
 57 #include "alloc.h"
 58 #include "aops.h"
 59 #include "blockcheck.h"
 60 #include "dlmglue.h"
 61 #include "export.h"
 62 #include "extent_map.h"
 63 #include "heartbeat.h"
 64 #include "inode.h"
 65 #include "journal.h"
 66 #include "localalloc.h"
 67 #include "namei.h"
 68 #include "slot_map.h"
 69 #include "super.h"
 70 #include "sysfile.h"
 71 #include "uptodate.h"
 72 #include "xattr.h"
 73 #include "quota.h"
 74 #include "refcounttree.h"
 75 #include "suballoc.h"
 76 
 77 #include "buffer_head_io.h"
 78 #include "filecheck.h"
 79 
 80 static struct kmem_cache *ocfs2_inode_cachep;
 81 struct kmem_cache *ocfs2_dquot_cachep;
 82 struct kmem_cache *ocfs2_qf_chunk_cachep;
 83 
 84 static struct dentry *ocfs2_debugfs_root;
 85 
 86 MODULE_AUTHOR("Oracle");
 87 MODULE_LICENSE("GPL");
 88 MODULE_DESCRIPTION("OCFS2 cluster file system");
 89 
 90 struct mount_options
 91 {
 92         unsigned long   commit_interval;
 93         unsigned long   mount_opt;
 94         unsigned int    atime_quantum;
 95         signed short    slot;
 96         int             localalloc_opt;
 97         unsigned int    resv_level;
 98         int             dir_resv_level;
 99         char            cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
100 };
101 
102 static int ocfs2_parse_options(struct super_block *sb, char *options,
103                                struct mount_options *mopt,
104                                int is_remount);
105 static int ocfs2_check_set_options(struct super_block *sb,
106                                    struct mount_options *options);
107 static int ocfs2_show_options(struct seq_file *s, struct dentry *root);
108 static void ocfs2_put_super(struct super_block *sb);
109 static int ocfs2_mount_volume(struct super_block *sb);
110 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
111 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
112 static int ocfs2_initialize_mem_caches(void);
113 static void ocfs2_free_mem_caches(void);
114 static void ocfs2_delete_osb(struct ocfs2_super *osb);
115 
116 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
117 
118 static int ocfs2_sync_fs(struct super_block *sb, int wait);
119 
120 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
121 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
122 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
123 static int ocfs2_check_volume(struct ocfs2_super *osb);
124 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
125                                struct buffer_head *bh,
126                                u32 sectsize,
127                                struct ocfs2_blockcheck_stats *stats);
128 static int ocfs2_initialize_super(struct super_block *sb,
129                                   struct buffer_head *bh,
130                                   int sector_size,
131                                   struct ocfs2_blockcheck_stats *stats);
132 static int ocfs2_get_sector(struct super_block *sb,
133                             struct buffer_head **bh,
134                             int block,
135                             int sect_size);
136 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
137 static void ocfs2_destroy_inode(struct inode *inode);
138 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
139 static int ocfs2_enable_quotas(struct ocfs2_super *osb);
140 static void ocfs2_disable_quotas(struct ocfs2_super *osb);
141 
142 static struct dquot **ocfs2_get_dquots(struct inode *inode)
143 {
144         return OCFS2_I(inode)->i_dquot;
145 }
146 
147 static const struct super_operations ocfs2_sops = {
148         .statfs         = ocfs2_statfs,
149         .alloc_inode    = ocfs2_alloc_inode,
150         .destroy_inode  = ocfs2_destroy_inode,
151         .drop_inode     = ocfs2_drop_inode,
152         .evict_inode    = ocfs2_evict_inode,
153         .sync_fs        = ocfs2_sync_fs,
154         .put_super      = ocfs2_put_super,
155         .remount_fs     = ocfs2_remount,
156         .show_options   = ocfs2_show_options,
157         .quota_read     = ocfs2_quota_read,
158         .quota_write    = ocfs2_quota_write,
159         .get_dquots     = ocfs2_get_dquots,
160 };
161 
162 enum {
163         Opt_barrier,
164         Opt_err_panic,
165         Opt_err_ro,
166         Opt_intr,
167         Opt_nointr,
168         Opt_hb_none,
169         Opt_hb_local,
170         Opt_hb_global,
171         Opt_data_ordered,
172         Opt_data_writeback,
173         Opt_atime_quantum,
174         Opt_slot,
175         Opt_commit,
176         Opt_localalloc,
177         Opt_localflocks,
178         Opt_stack,
179         Opt_user_xattr,
180         Opt_nouser_xattr,
181         Opt_inode64,
182         Opt_acl,
183         Opt_noacl,
184         Opt_usrquota,
185         Opt_grpquota,
186         Opt_coherency_buffered,
187         Opt_coherency_full,
188         Opt_resv_level,
189         Opt_dir_resv_level,
190         Opt_journal_async_commit,
191         Opt_err_cont,
192         Opt_err,
193 };
194 
195 static const match_table_t tokens = {
196         {Opt_barrier, "barrier=%u"},
197         {Opt_err_panic, "errors=panic"},
198         {Opt_err_ro, "errors=remount-ro"},
199         {Opt_intr, "intr"},
200         {Opt_nointr, "nointr"},
201         {Opt_hb_none, OCFS2_HB_NONE},
202         {Opt_hb_local, OCFS2_HB_LOCAL},
203         {Opt_hb_global, OCFS2_HB_GLOBAL},
204         {Opt_data_ordered, "data=ordered"},
205         {Opt_data_writeback, "data=writeback"},
206         {Opt_atime_quantum, "atime_quantum=%u"},
207         {Opt_slot, "preferred_slot=%u"},
208         {Opt_commit, "commit=%u"},
209         {Opt_localalloc, "localalloc=%d"},
210         {Opt_localflocks, "localflocks"},
211         {Opt_stack, "cluster_stack=%s"},
212         {Opt_user_xattr, "user_xattr"},
213         {Opt_nouser_xattr, "nouser_xattr"},
214         {Opt_inode64, "inode64"},
215         {Opt_acl, "acl"},
216         {Opt_noacl, "noacl"},
217         {Opt_usrquota, "usrquota"},
218         {Opt_grpquota, "grpquota"},
219         {Opt_coherency_buffered, "coherency=buffered"},
220         {Opt_coherency_full, "coherency=full"},
221         {Opt_resv_level, "resv_level=%u"},
222         {Opt_dir_resv_level, "dir_resv_level=%u"},
223         {Opt_journal_async_commit, "journal_async_commit"},
224         {Opt_err_cont, "errors=continue"},
225         {Opt_err, NULL}
226 };
227 
228 #ifdef CONFIG_DEBUG_FS
229 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
230 {
231         struct ocfs2_cluster_connection *cconn = osb->cconn;
232         struct ocfs2_recovery_map *rm = osb->recovery_map;
233         struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
234         int i, out = 0;
235         unsigned long flags;
236 
237         out += snprintf(buf + out, len - out,
238                         "%10s => Id: %-s  Uuid: %-s  Gen: 0x%X  Label: %-s\n",
239                         "Device", osb->dev_str, osb->uuid_str,
240                         osb->fs_generation, osb->vol_label);
241 
242         out += snprintf(buf + out, len - out,
243                         "%10s => State: %d  Flags: 0x%lX\n", "Volume",
244                         atomic_read(&osb->vol_state), osb->osb_flags);
245 
246         out += snprintf(buf + out, len - out,
247                         "%10s => Block: %lu  Cluster: %d\n", "Sizes",
248                         osb->sb->s_blocksize, osb->s_clustersize);
249 
250         out += snprintf(buf + out, len - out,
251                         "%10s => Compat: 0x%X  Incompat: 0x%X  "
252                         "ROcompat: 0x%X\n",
253                         "Features", osb->s_feature_compat,
254                         osb->s_feature_incompat, osb->s_feature_ro_compat);
255 
256         out += snprintf(buf + out, len - out,
257                         "%10s => Opts: 0x%lX  AtimeQuanta: %u\n", "Mount",
258                         osb->s_mount_opt, osb->s_atime_quantum);
259 
260         if (cconn) {
261                 out += snprintf(buf + out, len - out,
262                                 "%10s => Stack: %s  Name: %*s  "
263                                 "Version: %d.%d\n", "Cluster",
264                                 (*osb->osb_cluster_stack == '\0' ?
265                                  "o2cb" : osb->osb_cluster_stack),
266                                 cconn->cc_namelen, cconn->cc_name,
267                                 cconn->cc_version.pv_major,
268                                 cconn->cc_version.pv_minor);
269         }
270 
271         spin_lock_irqsave(&osb->dc_task_lock, flags);
272         out += snprintf(buf + out, len - out,
273                         "%10s => Pid: %d  Count: %lu  WakeSeq: %lu  "
274                         "WorkSeq: %lu\n", "DownCnvt",
275                         (osb->dc_task ?  task_pid_nr(osb->dc_task) : -1),
276                         osb->blocked_lock_count, osb->dc_wake_sequence,
277                         osb->dc_work_sequence);
278         spin_unlock_irqrestore(&osb->dc_task_lock, flags);
279 
280         spin_lock(&osb->osb_lock);
281         out += snprintf(buf + out, len - out, "%10s => Pid: %d  Nodes:",
282                         "Recovery",
283                         (osb->recovery_thread_task ?
284                          task_pid_nr(osb->recovery_thread_task) : -1));
285         if (rm->rm_used == 0)
286                 out += snprintf(buf + out, len - out, " None\n");
287         else {
288                 for (i = 0; i < rm->rm_used; i++)
289                         out += snprintf(buf + out, len - out, " %d",
290                                         rm->rm_entries[i]);
291                 out += snprintf(buf + out, len - out, "\n");
292         }
293         spin_unlock(&osb->osb_lock);
294 
295         out += snprintf(buf + out, len - out,
296                         "%10s => Pid: %d  Interval: %lu\n", "Commit",
297                         (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
298                         osb->osb_commit_interval);
299 
300         out += snprintf(buf + out, len - out,
301                         "%10s => State: %d  TxnId: %lu  NumTxns: %d\n",
302                         "Journal", osb->journal->j_state,
303                         osb->journal->j_trans_id,
304                         atomic_read(&osb->journal->j_num_trans));
305 
306         out += snprintf(buf + out, len - out,
307                         "%10s => GlobalAllocs: %d  LocalAllocs: %d  "
308                         "SubAllocs: %d  LAWinMoves: %d  SAExtends: %d\n",
309                         "Stats",
310                         atomic_read(&osb->alloc_stats.bitmap_data),
311                         atomic_read(&osb->alloc_stats.local_data),
312                         atomic_read(&osb->alloc_stats.bg_allocs),
313                         atomic_read(&osb->alloc_stats.moves),
314                         atomic_read(&osb->alloc_stats.bg_extends));
315 
316         out += snprintf(buf + out, len - out,
317                         "%10s => State: %u  Descriptor: %llu  Size: %u bits  "
318                         "Default: %u bits\n",
319                         "LocalAlloc", osb->local_alloc_state,
320                         (unsigned long long)osb->la_last_gd,
321                         osb->local_alloc_bits, osb->local_alloc_default_bits);
322 
323         spin_lock(&osb->osb_lock);
324         out += snprintf(buf + out, len - out,
325                         "%10s => InodeSlot: %d  StolenInodes: %d, "
326                         "MetaSlot: %d  StolenMeta: %d\n", "Steal",
327                         osb->s_inode_steal_slot,
328                         atomic_read(&osb->s_num_inodes_stolen),
329                         osb->s_meta_steal_slot,
330                         atomic_read(&osb->s_num_meta_stolen));
331         spin_unlock(&osb->osb_lock);
332 
333         out += snprintf(buf + out, len - out, "OrphanScan => ");
334         out += snprintf(buf + out, len - out, "Local: %u  Global: %u ",
335                         os->os_count, os->os_seqno);
336         out += snprintf(buf + out, len - out, " Last Scan: ");
337         if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
338                 out += snprintf(buf + out, len - out, "Disabled\n");
339         else
340                 out += snprintf(buf + out, len - out, "%lu seconds ago\n",
341                                 (unsigned long)(ktime_get_seconds() - os->os_scantime));
342 
343         out += snprintf(buf + out, len - out, "%10s => %3s  %10s\n",
344                         "Slots", "Num", "RecoGen");
345         for (i = 0; i < osb->max_slots; ++i) {
346                 out += snprintf(buf + out, len - out,
347                                 "%10s  %c %3d  %10d\n",
348                                 " ",
349                                 (i == osb->slot_num ? '*' : ' '),
350                                 i, osb->slot_recovery_generations[i]);
351         }
352 
353         return out;
354 }
355 
356 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
357 {
358         struct ocfs2_super *osb = inode->i_private;
359         char *buf = NULL;
360 
361         buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
362         if (!buf)
363                 goto bail;
364 
365         i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
366 
367         file->private_data = buf;
368 
369         return 0;
370 bail:
371         return -ENOMEM;
372 }
373 
374 static int ocfs2_debug_release(struct inode *inode, struct file *file)
375 {
376         kfree(file->private_data);
377         return 0;
378 }
379 
380 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
381                                 size_t nbytes, loff_t *ppos)
382 {
383         return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
384                                        i_size_read(file->f_mapping->host));
385 }
386 #else
387 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
388 {
389         return 0;
390 }
391 static int ocfs2_debug_release(struct inode *inode, struct file *file)
392 {
393         return 0;
394 }
395 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
396                                 size_t nbytes, loff_t *ppos)
397 {
398         return 0;
399 }
400 #endif  /* CONFIG_DEBUG_FS */
401 
402 static const struct file_operations ocfs2_osb_debug_fops = {
403         .open =         ocfs2_osb_debug_open,
404         .release =      ocfs2_debug_release,
405         .read =         ocfs2_debug_read,
406         .llseek =       generic_file_llseek,
407 };
408 
409 static int ocfs2_sync_fs(struct super_block *sb, int wait)
410 {
411         int status;
412         tid_t target;
413         struct ocfs2_super *osb = OCFS2_SB(sb);
414 
415         if (ocfs2_is_hard_readonly(osb))
416                 return -EROFS;
417 
418         if (wait) {
419                 status = ocfs2_flush_truncate_log(osb);
420                 if (status < 0)
421                         mlog_errno(status);
422         } else {
423                 ocfs2_schedule_truncate_log_flush(osb, 0);
424         }
425 
426         if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
427                                       &target)) {
428                 if (wait)
429                         jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
430                                              target);
431         }
432         return 0;
433 }
434 
435 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
436 {
437         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
438             && (ino == USER_QUOTA_SYSTEM_INODE
439                 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
440                 return 0;
441         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
442             && (ino == GROUP_QUOTA_SYSTEM_INODE
443                 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
444                 return 0;
445         return 1;
446 }
447 
448 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
449 {
450         struct inode *new = NULL;
451         int status = 0;
452         int i;
453 
454         new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
455         if (IS_ERR(new)) {
456                 status = PTR_ERR(new);
457                 mlog_errno(status);
458                 goto bail;
459         }
460         osb->root_inode = new;
461 
462         new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
463         if (IS_ERR(new)) {
464                 status = PTR_ERR(new);
465                 mlog_errno(status);
466                 goto bail;
467         }
468         osb->sys_root_inode = new;
469 
470         for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
471              i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
472                 if (!ocfs2_need_system_inode(osb, i))
473                         continue;
474                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
475                 if (!new) {
476                         ocfs2_release_system_inodes(osb);
477                         status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
478                         mlog_errno(status);
479                         mlog(ML_ERROR, "Unable to load system inode %d, "
480                              "possibly corrupt fs?", i);
481                         goto bail;
482                 }
483                 // the array now has one ref, so drop this one
484                 iput(new);
485         }
486 
487 bail:
488         if (status)
489                 mlog_errno(status);
490         return status;
491 }
492 
493 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
494 {
495         struct inode *new = NULL;
496         int status = 0;
497         int i;
498 
499         for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
500              i < NUM_SYSTEM_INODES;
501              i++) {
502                 if (!ocfs2_need_system_inode(osb, i))
503                         continue;
504                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
505                 if (!new) {
506                         ocfs2_release_system_inodes(osb);
507                         status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
508                         mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
509                              status, i, osb->slot_num);
510                         goto bail;
511                 }
512                 /* the array now has one ref, so drop this one */
513                 iput(new);
514         }
515 
516 bail:
517         if (status)
518                 mlog_errno(status);
519         return status;
520 }
521 
522 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
523 {
524         int i;
525         struct inode *inode;
526 
527         for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
528                 inode = osb->global_system_inodes[i];
529                 if (inode) {
530                         iput(inode);
531                         osb->global_system_inodes[i] = NULL;
532                 }
533         }
534 
535         inode = osb->sys_root_inode;
536         if (inode) {
537                 iput(inode);
538                 osb->sys_root_inode = NULL;
539         }
540 
541         inode = osb->root_inode;
542         if (inode) {
543                 iput(inode);
544                 osb->root_inode = NULL;
545         }
546 
547         if (!osb->local_system_inodes)
548                 return;
549 
550         for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
551                 if (osb->local_system_inodes[i]) {
552                         iput(osb->local_system_inodes[i]);
553                         osb->local_system_inodes[i] = NULL;
554                 }
555         }
556 
557         kfree(osb->local_system_inodes);
558         osb->local_system_inodes = NULL;
559 }
560 
561 /* We're allocating fs objects, use GFP_NOFS */
562 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
563 {
564         struct ocfs2_inode_info *oi;
565 
566         oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
567         if (!oi)
568                 return NULL;
569 
570         oi->i_sync_tid = 0;
571         oi->i_datasync_tid = 0;
572         memset(&oi->i_dquot, 0, sizeof(oi->i_dquot));
573 
574         jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
575         return &oi->vfs_inode;
576 }
577 
578 static void ocfs2_i_callback(struct rcu_head *head)
579 {
580         struct inode *inode = container_of(head, struct inode, i_rcu);
581         kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
582 }
583 
584 static void ocfs2_destroy_inode(struct inode *inode)
585 {
586         call_rcu(&inode->i_rcu, ocfs2_i_callback);
587 }
588 
589 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
590                                                 unsigned int cbits)
591 {
592         unsigned int bytes = 1 << cbits;
593         unsigned int trim = bytes;
594         unsigned int bitshift = 32;
595 
596         /*
597          * i_size and all block offsets in ocfs2 are always 64 bits
598          * wide. i_clusters is 32 bits, in cluster-sized units. So on
599          * 64 bit platforms, cluster size will be the limiting factor.
600          */
601 
602 #if BITS_PER_LONG == 32
603 # if defined(CONFIG_LBDAF)
604         BUILD_BUG_ON(sizeof(sector_t) != 8);
605         /*
606          * We might be limited by page cache size.
607          */
608         if (bytes > PAGE_SIZE) {
609                 bytes = PAGE_SIZE;
610                 trim = 1;
611                 /*
612                  * Shift by 31 here so that we don't get larger than
613                  * MAX_LFS_FILESIZE
614                  */
615                 bitshift = 31;
616         }
617 # else
618         /*
619          * We are limited by the size of sector_t. Use block size, as
620          * that's what we expose to the VFS.
621          */
622         bytes = 1 << bbits;
623         trim = 1;
624         bitshift = 31;
625 # endif
626 #endif
627 
628         /*
629          * Trim by a whole cluster when we can actually approach the
630          * on-disk limits. Otherwise we can overflow i_clusters when
631          * an extent start is at the max offset.
632          */
633         return (((unsigned long long)bytes) << bitshift) - trim;
634 }
635 
636 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
637 {
638         int incompat_features;
639         int ret = 0;
640         struct mount_options parsed_options;
641         struct ocfs2_super *osb = OCFS2_SB(sb);
642         u32 tmp;
643 
644         sync_filesystem(sb);
645 
646         if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
647             !ocfs2_check_set_options(sb, &parsed_options)) {
648                 ret = -EINVAL;
649                 goto out;
650         }
651 
652         tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
653                 OCFS2_MOUNT_HB_NONE;
654         if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
655                 ret = -EINVAL;
656                 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
657                 goto out;
658         }
659 
660         if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
661             (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
662                 ret = -EINVAL;
663                 mlog(ML_ERROR, "Cannot change data mode on remount\n");
664                 goto out;
665         }
666 
667         /* Probably don't want this on remount; it might
668          * mess with other nodes */
669         if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
670             (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
671                 ret = -EINVAL;
672                 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
673                 goto out;
674         }
675 
676         /* We're going to/from readonly mode. */
677         if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
678                 /* Disable quota accounting before remounting RO */
679                 if (*flags & SB_RDONLY) {
680                         ret = ocfs2_susp_quotas(osb, 0);
681                         if (ret < 0)
682                                 goto out;
683                 }
684                 /* Lock here so the check of HARD_RO and the potential
685                  * setting of SOFT_RO is atomic. */
686                 spin_lock(&osb->osb_lock);
687                 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
688                         mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
689                         ret = -EROFS;
690                         goto unlock_osb;
691                 }
692 
693                 if (*flags & SB_RDONLY) {
694                         sb->s_flags |= SB_RDONLY;
695                         osb->osb_flags |= OCFS2_OSB_SOFT_RO;
696                 } else {
697                         if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
698                                 mlog(ML_ERROR, "Cannot remount RDWR "
699                                      "filesystem due to previous errors.\n");
700                                 ret = -EROFS;
701                                 goto unlock_osb;
702                         }
703                         incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
704                         if (incompat_features) {
705                                 mlog(ML_ERROR, "Cannot remount RDWR because "
706                                      "of unsupported optional features "
707                                      "(%x).\n", incompat_features);
708                                 ret = -EINVAL;
709                                 goto unlock_osb;
710                         }
711                         sb->s_flags &= ~SB_RDONLY;
712                         osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
713                 }
714                 trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags);
715 unlock_osb:
716                 spin_unlock(&osb->osb_lock);
717                 /* Enable quota accounting after remounting RW */
718                 if (!ret && !(*flags & SB_RDONLY)) {
719                         if (sb_any_quota_suspended(sb))
720                                 ret = ocfs2_susp_quotas(osb, 1);
721                         else
722                                 ret = ocfs2_enable_quotas(osb);
723                         if (ret < 0) {
724                                 /* Return back changes... */
725                                 spin_lock(&osb->osb_lock);
726                                 sb->s_flags |= SB_RDONLY;
727                                 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
728                                 spin_unlock(&osb->osb_lock);
729                                 goto out;
730                         }
731                 }
732         }
733 
734         if (!ret) {
735                 /* Only save off the new mount options in case of a successful
736                  * remount. */
737                 osb->s_mount_opt = parsed_options.mount_opt;
738                 osb->s_atime_quantum = parsed_options.atime_quantum;
739                 osb->preferred_slot = parsed_options.slot;
740                 if (parsed_options.commit_interval)
741                         osb->osb_commit_interval = parsed_options.commit_interval;
742 
743                 if (!ocfs2_is_hard_readonly(osb))
744                         ocfs2_set_journal_params(osb);
745 
746                 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
747                         ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
748                                                         SB_POSIXACL : 0);
749         }
750 out:
751         return ret;
752 }
753 
754 static int ocfs2_sb_probe(struct super_block *sb,
755                           struct buffer_head **bh,
756                           int *sector_size,
757                           struct ocfs2_blockcheck_stats *stats)
758 {
759         int status, tmpstat;
760         struct ocfs1_vol_disk_hdr *hdr;
761         struct ocfs2_dinode *di;
762         int blksize;
763 
764         *bh = NULL;
765 
766         /* may be > 512 */
767         *sector_size = bdev_logical_block_size(sb->s_bdev);
768         if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
769                 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
770                      *sector_size, OCFS2_MAX_BLOCKSIZE);
771                 status = -EINVAL;
772                 goto bail;
773         }
774 
775         /* Can this really happen? */
776         if (*sector_size < OCFS2_MIN_BLOCKSIZE)
777                 *sector_size = OCFS2_MIN_BLOCKSIZE;
778 
779         /* check block zero for old format */
780         status = ocfs2_get_sector(sb, bh, 0, *sector_size);
781         if (status < 0) {
782                 mlog_errno(status);
783                 goto bail;
784         }
785         hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
786         if (hdr->major_version == OCFS1_MAJOR_VERSION) {
787                 mlog(ML_ERROR, "incompatible version: %u.%u\n",
788                      hdr->major_version, hdr->minor_version);
789                 status = -EINVAL;
790         }
791         if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
792                    strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
793                 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
794                      hdr->signature);
795                 status = -EINVAL;
796         }
797         brelse(*bh);
798         *bh = NULL;
799         if (status < 0) {
800                 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
801                      "upgraded before mounting with ocfs v2\n");
802                 goto bail;
803         }
804 
805         /*
806          * Now check at magic offset for 512, 1024, 2048, 4096
807          * blocksizes.  4096 is the maximum blocksize because it is
808          * the minimum clustersize.
809          */
810         status = -EINVAL;
811         for (blksize = *sector_size;
812              blksize <= OCFS2_MAX_BLOCKSIZE;
813              blksize <<= 1) {
814                 tmpstat = ocfs2_get_sector(sb, bh,
815                                            OCFS2_SUPER_BLOCK_BLKNO,
816                                            blksize);
817                 if (tmpstat < 0) {
818                         status = tmpstat;
819                         mlog_errno(status);
820                         break;
821                 }
822                 di = (struct ocfs2_dinode *) (*bh)->b_data;
823                 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
824                 spin_lock_init(&stats->b_lock);
825                 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
826                 if (tmpstat < 0) {
827                         brelse(*bh);
828                         *bh = NULL;
829                 }
830                 if (tmpstat != -EAGAIN) {
831                         status = tmpstat;
832                         break;
833                 }
834         }
835 
836 bail:
837         return status;
838 }
839 
840 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
841 {
842         u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
843 
844         if (osb->s_mount_opt & hb_enabled) {
845                 if (ocfs2_mount_local(osb)) {
846                         mlog(ML_ERROR, "Cannot heartbeat on a locally "
847                              "mounted device.\n");
848                         return -EINVAL;
849                 }
850                 if (ocfs2_userspace_stack(osb)) {
851                         mlog(ML_ERROR, "Userspace stack expected, but "
852                              "o2cb heartbeat arguments passed to mount\n");
853                         return -EINVAL;
854                 }
855                 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
856                      !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
857                     ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
858                      ocfs2_cluster_o2cb_global_heartbeat(osb))) {
859                         mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
860                         return -EINVAL;
861                 }
862         }
863 
864         if (!(osb->s_mount_opt & hb_enabled)) {
865                 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
866                     !ocfs2_userspace_stack(osb)) {
867                         mlog(ML_ERROR, "Heartbeat has to be started to mount "
868                              "a read-write clustered device.\n");
869                         return -EINVAL;
870                 }
871         }
872 
873         return 0;
874 }
875 
876 /*
877  * If we're using a userspace stack, mount should have passed
878  * a name that matches the disk.  If not, mount should not
879  * have passed a stack.
880  */
881 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
882                                         struct mount_options *mopt)
883 {
884         if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
885                 mlog(ML_ERROR,
886                      "cluster stack passed to mount, but this filesystem "
887                      "does not support it\n");
888                 return -EINVAL;
889         }
890 
891         if (ocfs2_userspace_stack(osb) &&
892             strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
893                     OCFS2_STACK_LABEL_LEN)) {
894                 mlog(ML_ERROR,
895                      "cluster stack passed to mount (\"%s\") does not "
896                      "match the filesystem (\"%s\")\n",
897                      mopt->cluster_stack,
898                      osb->osb_cluster_stack);
899                 return -EINVAL;
900         }
901 
902         return 0;
903 }
904 
905 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
906 {
907         int type;
908         struct super_block *sb = osb->sb;
909         unsigned int feature[OCFS2_MAXQUOTAS] = {
910                                         OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
911                                         OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
912         int status = 0;
913 
914         for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
915                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
916                         continue;
917                 if (unsuspend)
918                         status = dquot_resume(sb, type);
919                 else {
920                         struct ocfs2_mem_dqinfo *oinfo;
921 
922                         /* Cancel periodic syncing before suspending */
923                         oinfo = sb_dqinfo(sb, type)->dqi_priv;
924                         cancel_delayed_work_sync(&oinfo->dqi_sync_work);
925                         status = dquot_suspend(sb, type);
926                 }
927                 if (status < 0)
928                         break;
929         }
930         if (status < 0)
931                 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
932                      "remount (error = %d).\n", status);
933         return status;
934 }
935 
936 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
937 {
938         struct inode *inode[OCFS2_MAXQUOTAS] = { NULL, NULL };
939         struct super_block *sb = osb->sb;
940         unsigned int feature[OCFS2_MAXQUOTAS] = {
941                                         OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
942                                         OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
943         unsigned int ino[OCFS2_MAXQUOTAS] = {
944                                         LOCAL_USER_QUOTA_SYSTEM_INODE,
945                                         LOCAL_GROUP_QUOTA_SYSTEM_INODE };
946         int status;
947         int type;
948 
949         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
950         for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
951                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
952                         continue;
953                 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
954                                                         osb->slot_num);
955                 if (!inode[type]) {
956                         status = -ENOENT;
957                         goto out_quota_off;
958                 }
959                 status = dquot_enable(inode[type], type, QFMT_OCFS2,
960                                       DQUOT_USAGE_ENABLED);
961                 if (status < 0)
962                         goto out_quota_off;
963         }
964 
965         for (type = 0; type < OCFS2_MAXQUOTAS; type++)
966                 iput(inode[type]);
967         return 0;
968 out_quota_off:
969         ocfs2_disable_quotas(osb);
970         for (type = 0; type < OCFS2_MAXQUOTAS; type++)
971                 iput(inode[type]);
972         mlog_errno(status);
973         return status;
974 }
975 
976 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
977 {
978         int type;
979         struct inode *inode;
980         struct super_block *sb = osb->sb;
981         struct ocfs2_mem_dqinfo *oinfo;
982 
983         /* We mostly ignore errors in this function because there's not much
984          * we can do when we see them */
985         for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
986                 if (!sb_has_quota_loaded(sb, type))
987                         continue;
988                 oinfo = sb_dqinfo(sb, type)->dqi_priv;
989                 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
990                 inode = igrab(sb->s_dquot.files[type]);
991                 /* Turn off quotas. This will remove all dquot structures from
992                  * memory and so they will be automatically synced to global
993                  * quota files */
994                 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
995                                         DQUOT_LIMITS_ENABLED);
996                 if (!inode)
997                         continue;
998                 iput(inode);
999         }
1000 }
1001 
1002 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
1003 {
1004         struct dentry *root;
1005         int status, sector_size;
1006         struct mount_options parsed_options;
1007         struct inode *inode = NULL;
1008         struct ocfs2_super *osb = NULL;
1009         struct buffer_head *bh = NULL;
1010         char nodestr[12];
1011         struct ocfs2_blockcheck_stats stats;
1012 
1013         trace_ocfs2_fill_super(sb, data, silent);
1014 
1015         if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
1016                 status = -EINVAL;
1017                 goto read_super_error;
1018         }
1019 
1020         /* probe for superblock */
1021         status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
1022         if (status < 0) {
1023                 mlog(ML_ERROR, "superblock probe failed!\n");
1024                 goto read_super_error;
1025         }
1026 
1027         status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1028         osb = OCFS2_SB(sb);
1029         if (status < 0) {
1030                 mlog_errno(status);
1031                 goto read_super_error;
1032         }
1033         brelse(bh);
1034         bh = NULL;
1035 
1036         if (!ocfs2_check_set_options(sb, &parsed_options)) {
1037                 status = -EINVAL;
1038                 goto read_super_error;
1039         }
1040         osb->s_mount_opt = parsed_options.mount_opt;
1041         osb->s_atime_quantum = parsed_options.atime_quantum;
1042         osb->preferred_slot = parsed_options.slot;
1043         osb->osb_commit_interval = parsed_options.commit_interval;
1044 
1045         ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1046         osb->osb_resv_level = parsed_options.resv_level;
1047         osb->osb_dir_resv_level = parsed_options.resv_level;
1048         if (parsed_options.dir_resv_level == -1)
1049                 osb->osb_dir_resv_level = parsed_options.resv_level;
1050         else
1051                 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1052 
1053         status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1054         if (status)
1055                 goto read_super_error;
1056 
1057         sb->s_magic = OCFS2_SUPER_MAGIC;
1058 
1059         sb->s_flags = (sb->s_flags & ~(SB_POSIXACL | SB_NOSEC)) |
1060                 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? SB_POSIXACL : 0);
1061 
1062         /* Hard readonly mode only if: bdev_read_only, SB_RDONLY,
1063          * heartbeat=none */
1064         if (bdev_read_only(sb->s_bdev)) {
1065                 if (!sb_rdonly(sb)) {
1066                         status = -EACCES;
1067                         mlog(ML_ERROR, "Readonly device detected but readonly "
1068                              "mount was not specified.\n");
1069                         goto read_super_error;
1070                 }
1071 
1072                 /* You should not be able to start a local heartbeat
1073                  * on a readonly device. */
1074                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1075                         status = -EROFS;
1076                         mlog(ML_ERROR, "Local heartbeat specified on readonly "
1077                              "device.\n");
1078                         goto read_super_error;
1079                 }
1080 
1081                 status = ocfs2_check_journals_nolocks(osb);
1082                 if (status < 0) {
1083                         if (status == -EROFS)
1084                                 mlog(ML_ERROR, "Recovery required on readonly "
1085                                      "file system, but write access is "
1086                                      "unavailable.\n");
1087                         else
1088                                 mlog_errno(status);
1089                         goto read_super_error;
1090                 }
1091 
1092                 ocfs2_set_ro_flag(osb, 1);
1093 
1094                 printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. "
1095                        "Cluster services will not be used for this mount. "
1096                        "Recovery will be skipped.\n", osb->dev_str);
1097         }
1098 
1099         if (!ocfs2_is_hard_readonly(osb)) {
1100                 if (sb_rdonly(sb))
1101                         ocfs2_set_ro_flag(osb, 0);
1102         }
1103 
1104         status = ocfs2_verify_heartbeat(osb);
1105         if (status < 0) {
1106                 mlog_errno(status);
1107                 goto read_super_error;
1108         }
1109 
1110         osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1111                                                  ocfs2_debugfs_root);
1112         if (!osb->osb_debug_root) {
1113                 status = -EINVAL;
1114                 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1115                 goto read_super_error;
1116         }
1117 
1118         osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1119                                             osb->osb_debug_root,
1120                                             osb,
1121                                             &ocfs2_osb_debug_fops);
1122         if (!osb->osb_ctxt) {
1123                 status = -EINVAL;
1124                 mlog_errno(status);
1125                 goto read_super_error;
1126         }
1127 
1128         if (ocfs2_meta_ecc(osb)) {
1129                 status = ocfs2_blockcheck_stats_debugfs_install(
1130                                                 &osb->osb_ecc_stats,
1131                                                 osb->osb_debug_root);
1132                 if (status) {
1133                         mlog(ML_ERROR,
1134                              "Unable to create blockcheck statistics "
1135                              "files\n");
1136                         goto read_super_error;
1137                 }
1138         }
1139 
1140         status = ocfs2_mount_volume(sb);
1141         if (status < 0)
1142                 goto read_super_error;
1143 
1144         if (osb->root_inode)
1145                 inode = igrab(osb->root_inode);
1146 
1147         if (!inode) {
1148                 status = -EIO;
1149                 mlog_errno(status);
1150                 goto read_super_error;
1151         }
1152 
1153         root = d_make_root(inode);
1154         if (!root) {
1155                 status = -ENOMEM;
1156                 mlog_errno(status);
1157                 goto read_super_error;
1158         }
1159 
1160         sb->s_root = root;
1161 
1162         ocfs2_complete_mount_recovery(osb);
1163 
1164         if (ocfs2_mount_local(osb))
1165                 snprintf(nodestr, sizeof(nodestr), "local");
1166         else
1167                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1168 
1169         printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1170                "with %s data mode.\n",
1171                osb->dev_str, nodestr, osb->slot_num,
1172                osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1173                "ordered");
1174 
1175         atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1176         wake_up(&osb->osb_mount_event);
1177 
1178         /* Now we can initialize quotas because we can afford to wait
1179          * for cluster locks recovery now. That also means that truncation
1180          * log recovery can happen but that waits for proper quota setup */
1181         if (!sb_rdonly(sb)) {
1182                 status = ocfs2_enable_quotas(osb);
1183                 if (status < 0) {
1184                         /* We have to err-out specially here because
1185                          * s_root is already set */
1186                         mlog_errno(status);
1187                         atomic_set(&osb->vol_state, VOLUME_DISABLED);
1188                         wake_up(&osb->osb_mount_event);
1189                         return status;
1190                 }
1191         }
1192 
1193         ocfs2_complete_quota_recovery(osb);
1194 
1195         /* Now we wake up again for processes waiting for quotas */
1196         atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1197         wake_up(&osb->osb_mount_event);
1198 
1199         /* Start this when the mount is almost sure of being successful */
1200         ocfs2_orphan_scan_start(osb);
1201 
1202         /* Create filecheck sysfile /sys/fs/ocfs2/<devname>/filecheck */
1203         ocfs2_filecheck_create_sysfs(sb);
1204 
1205         return status;
1206 
1207 read_super_error:
1208         brelse(bh);
1209 
1210         if (status)
1211                 mlog_errno(status);
1212 
1213         if (osb) {
1214                 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1215                 wake_up(&osb->osb_mount_event);
1216                 ocfs2_dismount_volume(sb, 1);
1217         }
1218 
1219         return status;
1220 }
1221 
1222 static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
1223                         int flags,
1224                         const char *dev_name,
1225                         void *data)
1226 {
1227         return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
1228 }
1229 
1230 static struct file_system_type ocfs2_fs_type = {
1231         .owner          = THIS_MODULE,
1232         .name           = "ocfs2",
1233         .mount          = ocfs2_mount,
1234         .kill_sb        = kill_block_super,
1235         .fs_flags       = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1236         .next           = NULL
1237 };
1238 MODULE_ALIAS_FS("ocfs2");
1239 
1240 static int ocfs2_check_set_options(struct super_block *sb,
1241                                    struct mount_options *options)
1242 {
1243         if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1244             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1245                                          OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1246                 mlog(ML_ERROR, "User quotas were requested, but this "
1247                      "filesystem does not have the feature enabled.\n");
1248                 return 0;
1249         }
1250         if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1251             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1252                                          OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1253                 mlog(ML_ERROR, "Group quotas were requested, but this "
1254                      "filesystem does not have the feature enabled.\n");
1255                 return 0;
1256         }
1257         if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1258             !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1259                 mlog(ML_ERROR, "ACL support requested but extended attributes "
1260                      "feature is not enabled\n");
1261                 return 0;
1262         }
1263         /* No ACL setting specified? Use XATTR feature... */
1264         if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1265                                     OCFS2_MOUNT_NO_POSIX_ACL))) {
1266                 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1267                         options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1268                 else
1269                         options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1270         }
1271         return 1;
1272 }
1273 
1274 static int ocfs2_parse_options(struct super_block *sb,
1275                                char *options,
1276                                struct mount_options *mopt,
1277                                int is_remount)
1278 {
1279         int status, user_stack = 0;
1280         char *p;
1281         u32 tmp;
1282         int token, option;
1283         substring_t args[MAX_OPT_ARGS];
1284 
1285         trace_ocfs2_parse_options(is_remount, options ? options : "(none)");
1286 
1287         mopt->commit_interval = 0;
1288         mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1289         mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1290         mopt->slot = OCFS2_INVALID_SLOT;
1291         mopt->localalloc_opt = -1;
1292         mopt->cluster_stack[0] = '\0';
1293         mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1294         mopt->dir_resv_level = -1;
1295 
1296         if (!options) {
1297                 status = 1;
1298                 goto bail;
1299         }
1300 
1301         while ((p = strsep(&options, ",")) != NULL) {
1302                 if (!*p)
1303                         continue;
1304 
1305                 token = match_token(p, tokens, args);
1306                 switch (token) {
1307                 case Opt_hb_local:
1308                         mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1309                         break;
1310                 case Opt_hb_none:
1311                         mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1312                         break;
1313                 case Opt_hb_global:
1314                         mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1315                         break;
1316                 case Opt_barrier:
1317                         if (match_int(&args[0], &option)) {
1318                                 status = 0;
1319                                 goto bail;
1320                         }
1321                         if (option)
1322                                 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1323                         else
1324                                 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1325                         break;
1326                 case Opt_intr:
1327                         mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1328                         break;
1329                 case Opt_nointr:
1330                         mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1331                         break;
1332                 case Opt_err_panic:
1333                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1334                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1335                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1336                         break;
1337                 case Opt_err_ro:
1338                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1339                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1340                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_ROFS;
1341                         break;
1342                 case Opt_err_cont:
1343                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1344                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1345                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_CONT;
1346                         break;
1347                 case Opt_data_ordered:
1348                         mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1349                         break;
1350                 case Opt_data_writeback:
1351                         mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1352                         break;
1353                 case Opt_user_xattr:
1354                         mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1355                         break;
1356                 case Opt_nouser_xattr:
1357                         mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1358                         break;
1359                 case Opt_atime_quantum:
1360                         if (match_int(&args[0], &option)) {
1361                                 status = 0;
1362                                 goto bail;
1363                         }
1364                         if (option >= 0)
1365                                 mopt->atime_quantum = option;
1366                         break;
1367                 case Opt_slot:
1368                         if (match_int(&args[0], &option)) {
1369                                 status = 0;
1370                                 goto bail;
1371                         }
1372                         if (option)
1373                                 mopt->slot = (s16)option;
1374                         break;
1375                 case Opt_commit:
1376                         if (match_int(&args[0], &option)) {
1377                                 status = 0;
1378                                 goto bail;
1379                         }
1380                         if (option < 0)
1381                                 return 0;
1382                         if (option == 0)
1383                                 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1384                         mopt->commit_interval = HZ * option;
1385                         break;
1386                 case Opt_localalloc:
1387                         if (match_int(&args[0], &option)) {
1388                                 status = 0;
1389                                 goto bail;
1390                         }
1391                         if (option >= 0)
1392                                 mopt->localalloc_opt = option;
1393                         break;
1394                 case Opt_localflocks:
1395                         /*
1396                          * Changing this during remount could race
1397                          * flock() requests, or "unbalance" existing
1398                          * ones (e.g., a lock is taken in one mode but
1399                          * dropped in the other). If users care enough
1400                          * to flip locking modes during remount, we
1401                          * could add a "local" flag to individual
1402                          * flock structures for proper tracking of
1403                          * state.
1404                          */
1405                         if (!is_remount)
1406                                 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1407                         break;
1408                 case Opt_stack:
1409                         /* Check both that the option we were passed
1410                          * is of the right length and that it is a proper
1411                          * string of the right length.
1412                          */
1413                         if (((args[0].to - args[0].from) !=
1414                              OCFS2_STACK_LABEL_LEN) ||
1415                             (strnlen(args[0].from,
1416                                      OCFS2_STACK_LABEL_LEN) !=
1417                              OCFS2_STACK_LABEL_LEN)) {
1418                                 mlog(ML_ERROR,
1419                                      "Invalid cluster_stack option\n");
1420                                 status = 0;
1421                                 goto bail;
1422                         }
1423                         memcpy(mopt->cluster_stack, args[0].from,
1424                                OCFS2_STACK_LABEL_LEN);
1425                         mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1426                         /*
1427                          * Open code the memcmp here as we don't have
1428                          * an osb to pass to
1429                          * ocfs2_userspace_stack().
1430                          */
1431                         if (memcmp(mopt->cluster_stack,
1432                                    OCFS2_CLASSIC_CLUSTER_STACK,
1433                                    OCFS2_STACK_LABEL_LEN))
1434                                 user_stack = 1;
1435                         break;
1436                 case Opt_inode64:
1437                         mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1438                         break;
1439                 case Opt_usrquota:
1440                         mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1441                         break;
1442                 case Opt_grpquota:
1443                         mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1444                         break;
1445                 case Opt_coherency_buffered:
1446                         mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1447                         break;
1448                 case Opt_coherency_full:
1449                         mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1450                         break;
1451                 case Opt_acl:
1452                         mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1453                         mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1454                         break;
1455                 case Opt_noacl:
1456                         mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1457                         mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1458                         break;
1459                 case Opt_resv_level:
1460                         if (is_remount)
1461                                 break;
1462                         if (match_int(&args[0], &option)) {
1463                                 status = 0;
1464                                 goto bail;
1465                         }
1466                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1467                             option < OCFS2_MAX_RESV_LEVEL)
1468                                 mopt->resv_level = option;
1469                         break;
1470                 case Opt_dir_resv_level:
1471                         if (is_remount)
1472                                 break;
1473                         if (match_int(&args[0], &option)) {
1474                                 status = 0;
1475                                 goto bail;
1476                         }
1477                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1478                             option < OCFS2_MAX_RESV_LEVEL)
1479                                 mopt->dir_resv_level = option;
1480                         break;
1481                 case Opt_journal_async_commit:
1482                         mopt->mount_opt |= OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT;
1483                         break;
1484                 default:
1485                         mlog(ML_ERROR,
1486                              "Unrecognized mount option \"%s\" "
1487                              "or missing value\n", p);
1488                         status = 0;
1489                         goto bail;
1490                 }
1491         }
1492 
1493         if (user_stack == 0) {
1494                 /* Ensure only one heartbeat mode */
1495                 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1496                                          OCFS2_MOUNT_HB_GLOBAL |
1497                                          OCFS2_MOUNT_HB_NONE);
1498                 if (hweight32(tmp) != 1) {
1499                         mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1500                         status = 0;
1501                         goto bail;
1502                 }
1503         }
1504 
1505         status = 1;
1506 
1507 bail:
1508         return status;
1509 }
1510 
1511 static int ocfs2_show_options(struct seq_file *s, struct dentry *root)
1512 {
1513         struct ocfs2_super *osb = OCFS2_SB(root->d_sb);
1514         unsigned long opts = osb->s_mount_opt;
1515         unsigned int local_alloc_megs;
1516 
1517         if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1518                 seq_printf(s, ",_netdev");
1519                 if (opts & OCFS2_MOUNT_HB_LOCAL)
1520                         seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1521                 else
1522                         seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1523         } else
1524                 seq_printf(s, ",%s", OCFS2_HB_NONE);
1525 
1526         if (opts & OCFS2_MOUNT_NOINTR)
1527                 seq_printf(s, ",nointr");
1528 
1529         if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1530                 seq_printf(s, ",data=writeback");
1531         else
1532                 seq_printf(s, ",data=ordered");
1533 
1534         if (opts & OCFS2_MOUNT_BARRIER)
1535                 seq_printf(s, ",barrier=1");
1536 
1537         if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1538                 seq_printf(s, ",errors=panic");
1539         else if (opts & OCFS2_MOUNT_ERRORS_CONT)
1540                 seq_printf(s, ",errors=continue");
1541         else
1542                 seq_printf(s, ",errors=remount-ro");
1543 
1544         if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1545                 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1546 
1547         seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1548 
1549         if (osb->osb_commit_interval)
1550                 seq_printf(s, ",commit=%u",
1551                            (unsigned) (osb->osb_commit_interval / HZ));
1552 
1553         local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1554         if (local_alloc_megs != ocfs2_la_default_mb(osb))
1555                 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1556 
1557         if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1558                 seq_printf(s, ",localflocks,");
1559 
1560         if (osb->osb_cluster_stack[0])
1561                 seq_show_option_n(s, "cluster_stack", osb->osb_cluster_stack,
1562                                   OCFS2_STACK_LABEL_LEN);
1563         if (opts & OCFS2_MOUNT_USRQUOTA)
1564                 seq_printf(s, ",usrquota");
1565         if (opts & OCFS2_MOUNT_GRPQUOTA)
1566                 seq_printf(s, ",grpquota");
1567 
1568         if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1569                 seq_printf(s, ",coherency=buffered");
1570         else
1571                 seq_printf(s, ",coherency=full");
1572 
1573         if (opts & OCFS2_MOUNT_NOUSERXATTR)
1574                 seq_printf(s, ",nouser_xattr");
1575         else
1576                 seq_printf(s, ",user_xattr");
1577 
1578         if (opts & OCFS2_MOUNT_INODE64)
1579                 seq_printf(s, ",inode64");
1580 
1581         if (opts & OCFS2_MOUNT_POSIX_ACL)
1582                 seq_printf(s, ",acl");
1583         else
1584                 seq_printf(s, ",noacl");
1585 
1586         if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1587                 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1588 
1589         if (osb->osb_dir_resv_level != osb->osb_resv_level)
1590                 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1591 
1592         if (opts & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
1593                 seq_printf(s, ",journal_async_commit");
1594 
1595         return 0;
1596 }
1597 
1598 static int __init ocfs2_init(void)
1599 {
1600         int status;
1601 
1602         status = init_ocfs2_uptodate_cache();
1603         if (status < 0)
1604                 goto out1;
1605 
1606         status = ocfs2_initialize_mem_caches();
1607         if (status < 0)
1608                 goto out2;
1609 
1610         ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1611         if (!ocfs2_debugfs_root) {
1612                 status = -ENOMEM;
1613                 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1614                 goto out3;
1615         }
1616 
1617         ocfs2_set_locking_protocol();
1618 
1619         status = register_quota_format(&ocfs2_quota_format);
1620         if (status < 0)
1621                 goto out3;
1622         status = register_filesystem(&ocfs2_fs_type);
1623         if (!status)
1624                 return 0;
1625 
1626         unregister_quota_format(&ocfs2_quota_format);
1627 out3:
1628         debugfs_remove(ocfs2_debugfs_root);
1629         ocfs2_free_mem_caches();
1630 out2:
1631         exit_ocfs2_uptodate_cache();
1632 out1:
1633         mlog_errno(status);
1634         return status;
1635 }
1636 
1637 static void __exit ocfs2_exit(void)
1638 {
1639         unregister_quota_format(&ocfs2_quota_format);
1640 
1641         debugfs_remove(ocfs2_debugfs_root);
1642 
1643         ocfs2_free_mem_caches();
1644 
1645         unregister_filesystem(&ocfs2_fs_type);
1646 
1647         exit_ocfs2_uptodate_cache();
1648 }
1649 
1650 static void ocfs2_put_super(struct super_block *sb)
1651 {
1652         trace_ocfs2_put_super(sb);
1653 
1654         ocfs2_sync_blockdev(sb);
1655         ocfs2_dismount_volume(sb, 0);
1656         ocfs2_filecheck_remove_sysfs(sb);
1657 }
1658 
1659 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1660 {
1661         struct ocfs2_super *osb;
1662         u32 numbits, freebits;
1663         int status;
1664         struct ocfs2_dinode *bm_lock;
1665         struct buffer_head *bh = NULL;
1666         struct inode *inode = NULL;
1667 
1668         trace_ocfs2_statfs(dentry->d_sb, buf);
1669 
1670         osb = OCFS2_SB(dentry->d_sb);
1671 
1672         inode = ocfs2_get_system_file_inode(osb,
1673                                             GLOBAL_BITMAP_SYSTEM_INODE,
1674                                             OCFS2_INVALID_SLOT);
1675         if (!inode) {
1676                 mlog(ML_ERROR, "failed to get bitmap inode\n");
1677                 status = -EIO;
1678                 goto bail;
1679         }
1680 
1681         status = ocfs2_inode_lock(inode, &bh, 0);
1682         if (status < 0) {
1683                 mlog_errno(status);
1684                 goto bail;
1685         }
1686 
1687         bm_lock = (struct ocfs2_dinode *) bh->b_data;
1688 
1689         numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1690         freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1691 
1692         buf->f_type = OCFS2_SUPER_MAGIC;
1693         buf->f_bsize = dentry->d_sb->s_blocksize;
1694         buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1695         buf->f_blocks = ((sector_t) numbits) *
1696                         (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1697         buf->f_bfree = ((sector_t) freebits) *
1698                        (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1699         buf->f_bavail = buf->f_bfree;
1700         buf->f_files = numbits;
1701         buf->f_ffree = freebits;
1702         buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1703                                 & 0xFFFFFFFFUL;
1704         buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1705                                 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1706 
1707         brelse(bh);
1708 
1709         ocfs2_inode_unlock(inode, 0);
1710         status = 0;
1711 bail:
1712         iput(inode);
1713 
1714         if (status)
1715                 mlog_errno(status);
1716 
1717         return status;
1718 }
1719 
1720 static void ocfs2_inode_init_once(void *data)
1721 {
1722         struct ocfs2_inode_info *oi = data;
1723 
1724         oi->ip_flags = 0;
1725         oi->ip_open_count = 0;
1726         spin_lock_init(&oi->ip_lock);
1727         ocfs2_extent_map_init(&oi->vfs_inode);
1728         INIT_LIST_HEAD(&oi->ip_io_markers);
1729         INIT_LIST_HEAD(&oi->ip_unwritten_list);
1730         oi->ip_dir_start_lookup = 0;
1731         init_rwsem(&oi->ip_alloc_sem);
1732         init_rwsem(&oi->ip_xattr_sem);
1733         mutex_init(&oi->ip_io_mutex);
1734 
1735         oi->ip_blkno = 0ULL;
1736         oi->ip_clusters = 0;
1737 
1738         ocfs2_resv_init_once(&oi->ip_la_data_resv);
1739 
1740         ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1741         ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1742         ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1743 
1744         ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1745                                   &ocfs2_inode_caching_ops);
1746 
1747         inode_init_once(&oi->vfs_inode);
1748 }
1749 
1750 static int ocfs2_initialize_mem_caches(void)
1751 {
1752         ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1753                                        sizeof(struct ocfs2_inode_info),
1754                                        0,
1755                                        (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1756                                                 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1757                                        ocfs2_inode_init_once);
1758         ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1759                                         sizeof(struct ocfs2_dquot),
1760                                         0,
1761                                         (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1762                                                 SLAB_MEM_SPREAD),
1763                                         NULL);
1764         ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1765                                         sizeof(struct ocfs2_quota_chunk),
1766                                         0,
1767                                         (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1768                                         NULL);
1769         if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1770             !ocfs2_qf_chunk_cachep) {
1771                 if (ocfs2_inode_cachep)
1772                         kmem_cache_destroy(ocfs2_inode_cachep);
1773                 if (ocfs2_dquot_cachep)
1774                         kmem_cache_destroy(ocfs2_dquot_cachep);
1775                 if (ocfs2_qf_chunk_cachep)
1776                         kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1777                 return -ENOMEM;
1778         }
1779 
1780         return 0;
1781 }
1782 
1783 static void ocfs2_free_mem_caches(void)
1784 {
1785         /*
1786          * Make sure all delayed rcu free inodes are flushed before we
1787          * destroy cache.
1788          */
1789         rcu_barrier();
1790         if (ocfs2_inode_cachep)
1791                 kmem_cache_destroy(ocfs2_inode_cachep);
1792         ocfs2_inode_cachep = NULL;
1793 
1794         if (ocfs2_dquot_cachep)
1795                 kmem_cache_destroy(ocfs2_dquot_cachep);
1796         ocfs2_dquot_cachep = NULL;
1797 
1798         if (ocfs2_qf_chunk_cachep)
1799                 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1800         ocfs2_qf_chunk_cachep = NULL;
1801 }
1802 
1803 static int ocfs2_get_sector(struct super_block *sb,
1804                             struct buffer_head **bh,
1805                             int block,
1806                             int sect_size)
1807 {
1808         if (!sb_set_blocksize(sb, sect_size)) {
1809                 mlog(ML_ERROR, "unable to set blocksize\n");
1810                 return -EIO;
1811         }
1812 
1813         *bh = sb_getblk(sb, block);
1814         if (!*bh) {
1815                 mlog_errno(-ENOMEM);
1816                 return -ENOMEM;
1817         }
1818         lock_buffer(*bh);
1819         if (!buffer_dirty(*bh))
1820                 clear_buffer_uptodate(*bh);
1821         unlock_buffer(*bh);
1822         ll_rw_block(REQ_OP_READ, 0, 1, bh);
1823         wait_on_buffer(*bh);
1824         if (!buffer_uptodate(*bh)) {
1825                 mlog_errno(-EIO);
1826                 brelse(*bh);
1827                 *bh = NULL;
1828                 return -EIO;
1829         }
1830 
1831         return 0;
1832 }
1833 
1834 static int ocfs2_mount_volume(struct super_block *sb)
1835 {
1836         int status = 0;
1837         int unlock_super = 0;
1838         struct ocfs2_super *osb = OCFS2_SB(sb);
1839 
1840         if (ocfs2_is_hard_readonly(osb))
1841                 goto leave;
1842 
1843         status = ocfs2_dlm_init(osb);
1844         if (status < 0) {
1845                 mlog_errno(status);
1846                 if (status == -EBADR && ocfs2_userspace_stack(osb))
1847                         mlog(ML_ERROR, "couldn't mount because cluster name on"
1848                         " disk does not match the running cluster name.\n");
1849                 goto leave;
1850         }
1851 
1852         status = ocfs2_super_lock(osb, 1);
1853         if (status < 0) {
1854                 mlog_errno(status);
1855                 goto leave;
1856         }
1857         unlock_super = 1;
1858 
1859         /* This will load up the node map and add ourselves to it. */
1860         status = ocfs2_find_slot(osb);
1861         if (status < 0) {
1862                 mlog_errno(status);
1863                 goto leave;
1864         }
1865 
1866         /* load all node-local system inodes */
1867         status = ocfs2_init_local_system_inodes(osb);
1868         if (status < 0) {
1869                 mlog_errno(status);
1870                 goto leave;
1871         }
1872 
1873         status = ocfs2_check_volume(osb);
1874         if (status < 0) {
1875                 mlog_errno(status);
1876                 goto leave;
1877         }
1878 
1879         status = ocfs2_truncate_log_init(osb);
1880         if (status < 0)
1881                 mlog_errno(status);
1882 
1883 leave:
1884         if (unlock_super)
1885                 ocfs2_super_unlock(osb, 1);
1886 
1887         return status;
1888 }
1889 
1890 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1891 {
1892         int tmp, hangup_needed = 0;
1893         struct ocfs2_super *osb = NULL;
1894         char nodestr[12];
1895 
1896         trace_ocfs2_dismount_volume(sb);
1897 
1898         BUG_ON(!sb);
1899         osb = OCFS2_SB(sb);
1900         BUG_ON(!osb);
1901 
1902         debugfs_remove(osb->osb_ctxt);
1903 
1904         /* Orphan scan should be stopped as early as possible */
1905         ocfs2_orphan_scan_stop(osb);
1906 
1907         ocfs2_disable_quotas(osb);
1908 
1909         /* All dquots should be freed by now */
1910         WARN_ON(!llist_empty(&osb->dquot_drop_list));
1911         /* Wait for worker to be done with the work structure in osb */
1912         cancel_work_sync(&osb->dquot_drop_work);
1913 
1914         ocfs2_shutdown_local_alloc(osb);
1915 
1916         ocfs2_truncate_log_shutdown(osb);
1917 
1918         /* This will disable recovery and flush any recovery work. */
1919         ocfs2_recovery_exit(osb);
1920 
1921         ocfs2_journal_shutdown(osb);
1922 
1923         ocfs2_sync_blockdev(sb);
1924 
1925         ocfs2_purge_refcount_trees(osb);
1926 
1927         /* No cluster connection means we've failed during mount, so skip
1928          * all the steps which depended on that to complete. */
1929         if (osb->cconn) {
1930                 tmp = ocfs2_super_lock(osb, 1);
1931                 if (tmp < 0) {
1932                         mlog_errno(tmp);
1933                         return;
1934                 }
1935         }
1936 
1937         if (osb->slot_num != OCFS2_INVALID_SLOT)
1938                 ocfs2_put_slot(osb);
1939 
1940         if (osb->cconn)
1941                 ocfs2_super_unlock(osb, 1);
1942 
1943         ocfs2_release_system_inodes(osb);
1944 
1945         /*
1946          * If we're dismounting due to mount error, mount.ocfs2 will clean
1947          * up heartbeat.  If we're a local mount, there is no heartbeat.
1948          * If we failed before we got a uuid_str yet, we can't stop
1949          * heartbeat.  Otherwise, do it.
1950          */
1951         if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str &&
1952             !ocfs2_is_hard_readonly(osb))
1953                 hangup_needed = 1;
1954 
1955         if (osb->cconn)
1956                 ocfs2_dlm_shutdown(osb, hangup_needed);
1957 
1958         ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1959         debugfs_remove(osb->osb_debug_root);
1960 
1961         if (hangup_needed)
1962                 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1963 
1964         atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1965 
1966         if (ocfs2_mount_local(osb))
1967                 snprintf(nodestr, sizeof(nodestr), "local");
1968         else
1969                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1970 
1971         printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1972                osb->dev_str, nodestr);
1973 
1974         ocfs2_delete_osb(osb);
1975         kfree(osb);
1976         sb->s_dev = 0;
1977         sb->s_fs_info = NULL;
1978 }
1979 
1980 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1981                                 unsigned uuid_bytes)
1982 {
1983         int i, ret;
1984         char *ptr;
1985 
1986         BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1987 
1988         osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1989         if (osb->uuid_str == NULL)
1990                 return -ENOMEM;
1991 
1992         for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1993                 /* print with null */
1994                 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1995                 if (ret != 2) /* drop super cleans up */
1996                         return -EINVAL;
1997                 /* then only advance past the last char */
1998                 ptr += 2;
1999         }
2000 
2001         return 0;
2002 }
2003 
2004 /* Make sure entire volume is addressable by our journal.  Requires
2005    osb_clusters_at_boot to be valid and for the journal to have been
2006    initialized by ocfs2_journal_init(). */
2007 static int ocfs2_journal_addressable(struct ocfs2_super *osb)
2008 {
2009         int status = 0;
2010         u64 max_block =
2011                 ocfs2_clusters_to_blocks(osb->sb,
2012                                          osb->osb_clusters_at_boot) - 1;
2013 
2014         /* 32-bit block number is always OK. */
2015         if (max_block <= (u32)~0ULL)
2016                 goto out;
2017 
2018         /* Volume is "huge", so see if our journal is new enough to
2019            support it. */
2020         if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2021                                        OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2022               jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2023                                                JBD2_FEATURE_INCOMPAT_64BIT))) {
2024                 mlog(ML_ERROR, "The journal cannot address the entire volume. "
2025                      "Enable the 'block64' journal option with tunefs.ocfs2");
2026                 status = -EFBIG;
2027                 goto out;
2028         }
2029 
2030  out:
2031         return status;
2032 }
2033 
2034 static int ocfs2_initialize_super(struct super_block *sb,
2035                                   struct buffer_head *bh,
2036                                   int sector_size,
2037                                   struct ocfs2_blockcheck_stats *stats)
2038 {
2039         int status;
2040         int i, cbits, bbits;
2041         struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2042         struct inode *inode = NULL;
2043         struct ocfs2_journal *journal;
2044         struct ocfs2_super *osb;
2045         u64 total_blocks;
2046 
2047         osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2048         if (!osb) {
2049                 status = -ENOMEM;
2050                 mlog_errno(status);
2051                 goto bail;
2052         }
2053 
2054         sb->s_fs_info = osb;
2055         sb->s_op = &ocfs2_sops;
2056         sb->s_d_op = &ocfs2_dentry_ops;
2057         sb->s_export_op = &ocfs2_export_ops;
2058         sb->s_qcop = &dquot_quotactl_sysfile_ops;
2059         sb->dq_op = &ocfs2_quota_operations;
2060         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
2061         sb->s_xattr = ocfs2_xattr_handlers;
2062         sb->s_time_gran = 1;
2063         sb->s_flags |= SB_NOATIME;
2064         /* this is needed to support O_LARGEFILE */
2065         cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2066         bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2067         sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2068         memcpy(&sb->s_uuid, di->id2.i_super.s_uuid,
2069                sizeof(di->id2.i_super.s_uuid));
2070 
2071         osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2072 
2073         for (i = 0; i < 3; i++)
2074                 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2075         osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2076 
2077         osb->sb = sb;
2078         osb->s_sectsize_bits = blksize_bits(sector_size);
2079         BUG_ON(!osb->s_sectsize_bits);
2080 
2081         spin_lock_init(&osb->dc_task_lock);
2082         init_waitqueue_head(&osb->dc_event);
2083         osb->dc_work_sequence = 0;
2084         osb->dc_wake_sequence = 0;
2085         INIT_LIST_HEAD(&osb->blocked_lock_list);
2086         osb->blocked_lock_count = 0;
2087         spin_lock_init(&osb->osb_lock);
2088         spin_lock_init(&osb->osb_xattr_lock);
2089         ocfs2_init_steal_slots(osb);
2090 
2091         mutex_init(&osb->system_file_mutex);
2092 
2093         atomic_set(&osb->alloc_stats.moves, 0);
2094         atomic_set(&osb->alloc_stats.local_data, 0);
2095         atomic_set(&osb->alloc_stats.bitmap_data, 0);
2096         atomic_set(&osb->alloc_stats.bg_allocs, 0);
2097         atomic_set(&osb->alloc_stats.bg_extends, 0);
2098 
2099         /* Copy the blockcheck stats from the superblock probe */
2100         osb->osb_ecc_stats = *stats;
2101 
2102         ocfs2_init_node_maps(osb);
2103 
2104         snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2105                  MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2106 
2107         osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2108         if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2109                 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2110                      osb->max_slots);
2111                 status = -EINVAL;
2112                 goto bail;
2113         }
2114 
2115         ocfs2_orphan_scan_init(osb);
2116 
2117         status = ocfs2_recovery_init(osb);
2118         if (status) {
2119                 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2120                 mlog_errno(status);
2121                 goto bail;
2122         }
2123 
2124         init_waitqueue_head(&osb->checkpoint_event);
2125 
2126         osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2127 
2128         osb->slot_num = OCFS2_INVALID_SLOT;
2129 
2130         osb->s_xattr_inline_size = le16_to_cpu(
2131                                         di->id2.i_super.s_xattr_inline_size);
2132 
2133         osb->local_alloc_state = OCFS2_LA_UNUSED;
2134         osb->local_alloc_bh = NULL;
2135         INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2136 
2137         init_waitqueue_head(&osb->osb_mount_event);
2138 
2139         status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2140         if (status) {
2141                 mlog_errno(status);
2142                 goto bail;
2143         }
2144 
2145         osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2146         if (!osb->vol_label) {
2147                 mlog(ML_ERROR, "unable to alloc vol label\n");
2148                 status = -ENOMEM;
2149                 goto bail;
2150         }
2151 
2152         osb->slot_recovery_generations =
2153                 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2154                         GFP_KERNEL);
2155         if (!osb->slot_recovery_generations) {
2156                 status = -ENOMEM;
2157                 mlog_errno(status);
2158                 goto bail;
2159         }
2160 
2161         init_waitqueue_head(&osb->osb_wipe_event);
2162         osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2163                                         sizeof(*osb->osb_orphan_wipes),
2164                                         GFP_KERNEL);
2165         if (!osb->osb_orphan_wipes) {
2166                 status = -ENOMEM;
2167                 mlog_errno(status);
2168                 goto bail;
2169         }
2170 
2171         osb->osb_rf_lock_tree = RB_ROOT;
2172 
2173         osb->s_feature_compat =
2174                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2175         osb->s_feature_ro_compat =
2176                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2177         osb->s_feature_incompat =
2178                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2179 
2180         if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2181                 mlog(ML_ERROR, "couldn't mount because of unsupported "
2182                      "optional features (%x).\n", i);
2183                 status = -EINVAL;
2184                 goto bail;
2185         }
2186         if (!sb_rdonly(osb->sb) && (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2187                 mlog(ML_ERROR, "couldn't mount RDWR because of "
2188                      "unsupported optional features (%x).\n", i);
2189                 status = -EINVAL;
2190                 goto bail;
2191         }
2192 
2193         if (ocfs2_clusterinfo_valid(osb)) {
2194                 osb->osb_stackflags =
2195                         OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2196                 strlcpy(osb->osb_cluster_stack,
2197                        OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2198                        OCFS2_STACK_LABEL_LEN + 1);
2199                 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2200                         mlog(ML_ERROR,
2201                              "couldn't mount because of an invalid "
2202                              "cluster stack label (%s) \n",
2203                              osb->osb_cluster_stack);
2204                         status = -EINVAL;
2205                         goto bail;
2206                 }
2207                 strlcpy(osb->osb_cluster_name,
2208                         OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster,
2209                         OCFS2_CLUSTER_NAME_LEN + 1);
2210         } else {
2211                 /* The empty string is identical with classic tools that
2212                  * don't know about s_cluster_info. */
2213                 osb->osb_cluster_stack[0] = '\0';
2214         }
2215 
2216         get_random_bytes(&osb->s_next_generation, sizeof(u32));
2217 
2218         /* FIXME
2219          * This should be done in ocfs2_journal_init(), but unknown
2220          * ordering issues will cause the filesystem to crash.
2221          * If anyone wants to figure out what part of the code
2222          * refers to osb->journal before ocfs2_journal_init() is run,
2223          * be my guest.
2224          */
2225         /* initialize our journal structure */
2226 
2227         journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2228         if (!journal) {
2229                 mlog(ML_ERROR, "unable to alloc journal\n");
2230                 status = -ENOMEM;
2231                 goto bail;
2232         }
2233         osb->journal = journal;
2234         journal->j_osb = osb;
2235 
2236         atomic_set(&journal->j_num_trans, 0);
2237         init_rwsem(&journal->j_trans_barrier);
2238         init_waitqueue_head(&journal->j_checkpointed);
2239         spin_lock_init(&journal->j_lock);
2240         journal->j_trans_id = (unsigned long) 1;
2241         INIT_LIST_HEAD(&journal->j_la_cleanups);
2242         INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2243         journal->j_state = OCFS2_JOURNAL_FREE;
2244 
2245         INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs);
2246         init_llist_head(&osb->dquot_drop_list);
2247 
2248         /* get some pseudo constants for clustersize bits */
2249         osb->s_clustersize_bits =
2250                 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2251         osb->s_clustersize = 1 << osb->s_clustersize_bits;
2252 
2253         if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2254             osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2255                 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2256                      osb->s_clustersize);
2257                 status = -EINVAL;
2258                 goto bail;
2259         }
2260 
2261         total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2262                                                 le32_to_cpu(di->i_clusters));
2263 
2264         status = generic_check_addressable(osb->sb->s_blocksize_bits,
2265                                            total_blocks);
2266         if (status) {
2267                 mlog(ML_ERROR, "Volume too large "
2268                      "to mount safely on this system");
2269                 status = -EFBIG;
2270                 goto bail;
2271         }
2272 
2273         if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2274                                  sizeof(di->id2.i_super.s_uuid))) {
2275                 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2276                 status = -ENOMEM;
2277                 goto bail;
2278         }
2279 
2280         strlcpy(osb->vol_label, di->id2.i_super.s_label,
2281                 OCFS2_MAX_VOL_LABEL_LEN);
2282         osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2283         osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2284         osb->first_cluster_group_blkno =
2285                 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2286         osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2287         osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2288         trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str,
2289                                      (unsigned long long)osb->root_blkno,
2290                                      (unsigned long long)osb->system_dir_blkno,
2291                                      osb->s_clustersize_bits);
2292 
2293         osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2294         if (!osb->osb_dlm_debug) {
2295                 status = -ENOMEM;
2296                 mlog_errno(status);
2297                 goto bail;
2298         }
2299 
2300         atomic_set(&osb->vol_state, VOLUME_INIT);
2301 
2302         /* load root, system_dir, and all global system inodes */
2303         status = ocfs2_init_global_system_inodes(osb);
2304         if (status < 0) {
2305                 mlog_errno(status);
2306                 goto bail;
2307         }
2308 
2309         /*
2310          * global bitmap
2311          */
2312         inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2313                                             OCFS2_INVALID_SLOT);
2314         if (!inode) {
2315                 status = -EINVAL;
2316                 mlog_errno(status);
2317                 goto bail;
2318         }
2319 
2320         osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2321         osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2322         iput(inode);
2323 
2324         osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2325                                  osb->s_feature_incompat) * 8;
2326 
2327         status = ocfs2_init_slot_info(osb);
2328         if (status < 0) {
2329                 mlog_errno(status);
2330                 goto bail;
2331         }
2332         cleancache_init_shared_fs(sb);
2333 
2334         osb->ocfs2_wq = alloc_ordered_workqueue("ocfs2_wq", WQ_MEM_RECLAIM);
2335         if (!osb->ocfs2_wq) {
2336                 status = -ENOMEM;
2337                 mlog_errno(status);
2338         }
2339 
2340 bail:
2341         return status;
2342 }
2343 
2344 /*
2345  * will return: -EAGAIN if it is ok to keep searching for superblocks
2346  *              -EINVAL if there is a bad superblock
2347  *              0 on success
2348  */
2349 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2350                                struct buffer_head *bh,
2351                                u32 blksz,
2352                                struct ocfs2_blockcheck_stats *stats)
2353 {
2354         int status = -EAGAIN;
2355 
2356         if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2357                    strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2358                 /* We have to do a raw check of the feature here */
2359                 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2360                     OCFS2_FEATURE_INCOMPAT_META_ECC) {
2361                         status = ocfs2_block_check_validate(bh->b_data,
2362                                                             bh->b_size,
2363                                                             &di->i_check,
2364                                                             stats);
2365                         if (status)
2366                                 goto out;
2367                 }
2368                 status = -EINVAL;
2369                 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2370                         mlog(ML_ERROR, "found superblock with incorrect block "
2371                              "size: found %u, should be %u\n",
2372                              1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2373                                blksz);
2374                 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2375                            OCFS2_MAJOR_REV_LEVEL ||
2376                            le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2377                            OCFS2_MINOR_REV_LEVEL) {
2378                         mlog(ML_ERROR, "found superblock with bad version: "
2379                              "found %u.%u, should be %u.%u\n",
2380                              le16_to_cpu(di->id2.i_super.s_major_rev_level),
2381                              le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2382                              OCFS2_MAJOR_REV_LEVEL,
2383                              OCFS2_MINOR_REV_LEVEL);
2384                 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2385                         mlog(ML_ERROR, "bad block number on superblock: "
2386                              "found %llu, should be %llu\n",
2387                              (unsigned long long)le64_to_cpu(di->i_blkno),
2388                              (unsigned long long)bh->b_blocknr);
2389                 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2390                             le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2391                         mlog(ML_ERROR, "bad cluster size found: %u\n",
2392                              1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2393                 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2394                         mlog(ML_ERROR, "bad root_blkno: 0\n");
2395                 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2396                         mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2397                 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2398                         mlog(ML_ERROR,
2399                              "Superblock slots found greater than file system "
2400                              "maximum: found %u, max %u\n",
2401                              le16_to_cpu(di->id2.i_super.s_max_slots),
2402                              OCFS2_MAX_SLOTS);
2403                 } else {
2404                         /* found it! */
2405                         status = 0;
2406                 }
2407         }
2408 
2409 out:
2410         if (status && status != -EAGAIN)
2411                 mlog_errno(status);
2412         return status;
2413 }
2414 
2415 static int ocfs2_check_volume(struct ocfs2_super *osb)
2416 {
2417         int status;
2418         int dirty;
2419         int local;
2420         struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2421                                                   * recover
2422                                                   * ourselves. */
2423 
2424         /* Init our journal object. */
2425         status = ocfs2_journal_init(osb->journal, &dirty);
2426         if (status < 0) {
2427                 mlog(ML_ERROR, "Could not initialize journal!\n");
2428                 goto finally;
2429         }
2430 
2431         /* Now that journal has been initialized, check to make sure
2432            entire volume is addressable. */
2433         status = ocfs2_journal_addressable(osb);
2434         if (status)
2435                 goto finally;
2436 
2437         /* If the journal was unmounted cleanly then we don't want to
2438          * recover anything. Otherwise, journal_load will do that
2439          * dirty work for us :) */
2440         if (!dirty) {
2441                 status = ocfs2_journal_wipe(osb->journal, 0);
2442                 if (status < 0) {
2443                         mlog_errno(status);
2444                         goto finally;
2445                 }
2446         } else {
2447                 printk(KERN_NOTICE "ocfs2: File system on device (%s) was not "
2448                        "unmounted cleanly, recovering it.\n", osb->dev_str);
2449         }
2450 
2451         local = ocfs2_mount_local(osb);
2452 
2453         /* will play back anything left in the journal. */
2454         status = ocfs2_journal_load(osb->journal, local, dirty);
2455         if (status < 0) {
2456                 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2457                 goto finally;
2458         }
2459 
2460         if (osb->s_mount_opt & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
2461                 jbd2_journal_set_features(osb->journal->j_journal,
2462                                 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2463                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2464         else
2465                 jbd2_journal_clear_features(osb->journal->j_journal,
2466                                 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2467                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2468 
2469         if (dirty) {
2470                 /* recover my local alloc if we didn't unmount cleanly. */
2471                 status = ocfs2_begin_local_alloc_recovery(osb,
2472                                                           osb->slot_num,
2473                                                           &local_alloc);
2474                 if (status < 0) {
2475                         mlog_errno(status);
2476                         goto finally;
2477                 }
2478                 /* we complete the recovery process after we've marked
2479                  * ourselves as mounted. */
2480         }
2481 
2482         status = ocfs2_load_local_alloc(osb);
2483         if (status < 0) {
2484                 mlog_errno(status);
2485                 goto finally;
2486         }
2487 
2488         if (dirty) {
2489                 /* Recovery will be completed after we've mounted the
2490                  * rest of the volume. */
2491                 osb->local_alloc_copy = local_alloc;
2492                 local_alloc = NULL;
2493         }
2494 
2495         /* go through each journal, trylock it and if you get the
2496          * lock, and it's marked as dirty, set the bit in the recover
2497          * map and launch a recovery thread for it. */
2498         status = ocfs2_mark_dead_nodes(osb);
2499         if (status < 0) {
2500                 mlog_errno(status);
2501                 goto finally;
2502         }
2503 
2504         status = ocfs2_compute_replay_slots(osb);
2505         if (status < 0)
2506                 mlog_errno(status);
2507 
2508 finally:
2509         kfree(local_alloc);
2510 
2511         if (status)
2512                 mlog_errno(status);
2513         return status;
2514 }
2515 
2516 /*
2517  * The routine gets called from dismount or close whenever a dismount on
2518  * volume is requested and the osb open count becomes 1.
2519  * It will remove the osb from the global list and also free up all the
2520  * initialized resources and fileobject.
2521  */
2522 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2523 {
2524         /* This function assumes that the caller has the main osb resource */
2525 
2526         /* ocfs2_initializer_super have already created this workqueue */
2527         if (osb->ocfs2_wq)
2528                 destroy_workqueue(osb->ocfs2_wq);
2529 
2530         ocfs2_free_slot_info(osb);
2531 
2532         kfree(osb->osb_orphan_wipes);
2533         kfree(osb->slot_recovery_generations);
2534         /* FIXME
2535          * This belongs in journal shutdown, but because we have to
2536          * allocate osb->journal at the start of ocfs2_initialize_osb(),
2537          * we free it here.
2538          */
2539         kfree(osb->journal);
2540         kfree(osb->local_alloc_copy);
2541         kfree(osb->uuid_str);
2542         kfree(osb->vol_label);
2543         ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2544         memset(osb, 0, sizeof(struct ocfs2_super));
2545 }
2546 
2547 /* Depending on the mount option passed, perform one of the following:
2548  * Put OCFS2 into a readonly state (default)
2549  * Return EIO so that only the process errs
2550  * Fix the error as if fsck.ocfs2 -y
2551  * panic
2552  */
2553 static int ocfs2_handle_error(struct super_block *sb)
2554 {
2555         struct ocfs2_super *osb = OCFS2_SB(sb);
2556         int rv = 0;
2557 
2558         ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2559         pr_crit("On-disk corruption discovered. "
2560                 "Please run fsck.ocfs2 once the filesystem is unmounted.\n");
2561 
2562         if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC) {
2563                 panic("OCFS2: (device %s): panic forced after error\n",
2564                       sb->s_id);
2565         } else if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_CONT) {
2566                 pr_crit("OCFS2: Returning error to the calling process.\n");
2567                 rv = -EIO;
2568         } else { /* default option */
2569                 rv = -EROFS;
2570                 if (sb_rdonly(sb) && (ocfs2_is_soft_readonly(osb) || ocfs2_is_hard_readonly(osb)))
2571                         return rv;
2572 
2573                 pr_crit("OCFS2: File system is now read-only.\n");
2574                 sb->s_flags |= SB_RDONLY;
2575                 ocfs2_set_ro_flag(osb, 0);
2576         }
2577 
2578         return rv;
2579 }
2580 
2581 int __ocfs2_error(struct super_block *sb, const char *function,
2582                   const char *fmt, ...)
2583 {
2584         struct va_format vaf;
2585         va_list args;
2586 
2587         va_start(args, fmt);
2588         vaf.fmt = fmt;
2589         vaf.va = &args;
2590 
2591         /* Not using mlog here because we want to show the actual
2592          * function the error came from. */
2593         printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %pV",
2594                sb->s_id, function, &vaf);
2595 
2596         va_end(args);
2597 
2598         return ocfs2_handle_error(sb);
2599 }
2600 
2601 /* Handle critical errors. This is intentionally more drastic than
2602  * ocfs2_handle_error, so we only use for things like journal errors,
2603  * etc. */
2604 void __ocfs2_abort(struct super_block *sb, const char *function,
2605                    const char *fmt, ...)
2606 {
2607         struct va_format vaf;
2608         va_list args;
2609 
2610         va_start(args, fmt);
2611 
2612         vaf.fmt = fmt;
2613         vaf.va = &args;
2614 
2615         printk(KERN_CRIT "OCFS2: abort (device %s): %s: %pV",
2616                sb->s_id, function, &vaf);
2617 
2618         va_end(args);
2619 
2620         /* We don't have the cluster support yet to go straight to
2621          * hard readonly in here. Until then, we want to keep
2622          * ocfs2_abort() so that we can at least mark critical
2623          * errors.
2624          *
2625          * TODO: This should abort the journal and alert other nodes
2626          * that our slot needs recovery. */
2627 
2628         /* Force a panic(). This stinks, but it's better than letting
2629          * things continue without having a proper hard readonly
2630          * here. */
2631         if (!ocfs2_mount_local(OCFS2_SB(sb)))
2632                 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2633         ocfs2_handle_error(sb);
2634 }
2635 
2636 /*
2637  * Void signal blockers, because in-kernel sigprocmask() only fails
2638  * when SIG_* is wrong.
2639  */
2640 void ocfs2_block_signals(sigset_t *oldset)
2641 {
2642         int rc;
2643         sigset_t blocked;
2644 
2645         sigfillset(&blocked);
2646         rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2647         BUG_ON(rc);
2648 }
2649 
2650 void ocfs2_unblock_signals(sigset_t *oldset)
2651 {
2652         int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2653         BUG_ON(rc);
2654 }
2655 
2656 module_init(ocfs2_init);
2657 module_exit(ocfs2_exit);
2658 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp