Push BKL down into ->remount_fs()
[linux-2.6.git] / fs / ocfs2 / super.c
1 /* -*- mode: c; c-basic-offset: 8; -*-
2  * vim: noexpandtab sw=8 ts=8 sts=0:
3  *
4  * super.c
5  *
6  * load/unload driver, mount/dismount volumes
7  *
8  * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
9  *
10  * This program is free software; you can redistribute it and/or
11  * modify it under the terms of the GNU General Public
12  * License as published by the Free Software Foundation; either
13  * version 2 of the License, or (at your option) any later version.
14  *
15  * This program is distributed in the hope that it will be useful,
16  * but WITHOUT ANY WARRANTY; without even the implied warranty of
17  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
18  * General Public License for more details.
19  *
20  * You should have received a copy of the GNU General Public
21  * License along with this program; if not, write to the
22  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
23  * Boston, MA 021110-1307, USA.
24  */
25
26 #include <linux/module.h>
27 #include <linux/fs.h>
28 #include <linux/types.h>
29 #include <linux/slab.h>
30 #include <linux/highmem.h>
31 #include <linux/utsname.h>
32 #include <linux/init.h>
33 #include <linux/random.h>
34 #include <linux/statfs.h>
35 #include <linux/moduleparam.h>
36 #include <linux/blkdev.h>
37 #include <linux/socket.h>
38 #include <linux/inet.h>
39 #include <linux/parser.h>
40 #include <linux/crc32.h>
41 #include <linux/debugfs.h>
42 #include <linux/mount.h>
43 #include <linux/seq_file.h>
44 #include <linux/quotaops.h>
45 #include <linux/smp_lock.h>
46
47 #define MLOG_MASK_PREFIX ML_SUPER
48 #include <cluster/masklog.h>
49
50 #include "ocfs2.h"
51
52 /* this should be the only file to include a version 1 header */
53 #include "ocfs1_fs_compat.h"
54
55 #include "alloc.h"
56 #include "blockcheck.h"
57 #include "dlmglue.h"
58 #include "export.h"
59 #include "extent_map.h"
60 #include "heartbeat.h"
61 #include "inode.h"
62 #include "journal.h"
63 #include "localalloc.h"
64 #include "namei.h"
65 #include "slot_map.h"
66 #include "super.h"
67 #include "sysfile.h"
68 #include "uptodate.h"
69 #include "ver.h"
70 #include "xattr.h"
71 #include "quota.h"
72
73 #include "buffer_head_io.h"
74
75 static struct kmem_cache *ocfs2_inode_cachep = NULL;
76 struct kmem_cache *ocfs2_dquot_cachep;
77 struct kmem_cache *ocfs2_qf_chunk_cachep;
78
79 /* OCFS2 needs to schedule several differnt types of work which
80  * require cluster locking, disk I/O, recovery waits, etc. Since these
81  * types of work tend to be heavy we avoid using the kernel events
82  * workqueue and schedule on our own. */
83 struct workqueue_struct *ocfs2_wq = NULL;
84
85 static struct dentry *ocfs2_debugfs_root = NULL;
86
87 MODULE_AUTHOR("Oracle");
88 MODULE_LICENSE("GPL");
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         unsigned int    localalloc_opt;
97         char            cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
98 };
99
100 static int ocfs2_parse_options(struct super_block *sb, char *options,
101                                struct mount_options *mopt,
102                                int is_remount);
103 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt);
104 static void ocfs2_put_super(struct super_block *sb);
105 static int ocfs2_mount_volume(struct super_block *sb);
106 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
107 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
108 static int ocfs2_initialize_mem_caches(void);
109 static void ocfs2_free_mem_caches(void);
110 static void ocfs2_delete_osb(struct ocfs2_super *osb);
111
112 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
113
114 static int ocfs2_sync_fs(struct super_block *sb, int wait);
115
116 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
117 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
118 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
119 static int ocfs2_check_volume(struct ocfs2_super *osb);
120 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
121                                struct buffer_head *bh,
122                                u32 sectsize);
123 static int ocfs2_initialize_super(struct super_block *sb,
124                                   struct buffer_head *bh,
125                                   int sector_size);
126 static int ocfs2_get_sector(struct super_block *sb,
127                             struct buffer_head **bh,
128                             int block,
129                             int sect_size);
130 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
131 static void ocfs2_destroy_inode(struct inode *inode);
132 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
133 static int ocfs2_enable_quotas(struct ocfs2_super *osb);
134 static void ocfs2_disable_quotas(struct ocfs2_super *osb);
135
136 static const struct super_operations ocfs2_sops = {
137         .statfs         = ocfs2_statfs,
138         .alloc_inode    = ocfs2_alloc_inode,
139         .destroy_inode  = ocfs2_destroy_inode,
140         .drop_inode     = ocfs2_drop_inode,
141         .clear_inode    = ocfs2_clear_inode,
142         .delete_inode   = ocfs2_delete_inode,
143         .sync_fs        = ocfs2_sync_fs,
144         .put_super      = ocfs2_put_super,
145         .remount_fs     = ocfs2_remount,
146         .show_options   = ocfs2_show_options,
147         .quota_read     = ocfs2_quota_read,
148         .quota_write    = ocfs2_quota_write,
149 };
150
151 enum {
152         Opt_barrier,
153         Opt_err_panic,
154         Opt_err_ro,
155         Opt_intr,
156         Opt_nointr,
157         Opt_hb_none,
158         Opt_hb_local,
159         Opt_data_ordered,
160         Opt_data_writeback,
161         Opt_atime_quantum,
162         Opt_slot,
163         Opt_commit,
164         Opt_localalloc,
165         Opt_localflocks,
166         Opt_stack,
167         Opt_user_xattr,
168         Opt_nouser_xattr,
169         Opt_inode64,
170         Opt_acl,
171         Opt_noacl,
172         Opt_usrquota,
173         Opt_grpquota,
174         Opt_err,
175 };
176
177 static const match_table_t tokens = {
178         {Opt_barrier, "barrier=%u"},
179         {Opt_err_panic, "errors=panic"},
180         {Opt_err_ro, "errors=remount-ro"},
181         {Opt_intr, "intr"},
182         {Opt_nointr, "nointr"},
183         {Opt_hb_none, OCFS2_HB_NONE},
184         {Opt_hb_local, OCFS2_HB_LOCAL},
185         {Opt_data_ordered, "data=ordered"},
186         {Opt_data_writeback, "data=writeback"},
187         {Opt_atime_quantum, "atime_quantum=%u"},
188         {Opt_slot, "preferred_slot=%u"},
189         {Opt_commit, "commit=%u"},
190         {Opt_localalloc, "localalloc=%d"},
191         {Opt_localflocks, "localflocks"},
192         {Opt_stack, "cluster_stack=%s"},
193         {Opt_user_xattr, "user_xattr"},
194         {Opt_nouser_xattr, "nouser_xattr"},
195         {Opt_inode64, "inode64"},
196         {Opt_acl, "acl"},
197         {Opt_noacl, "noacl"},
198         {Opt_usrquota, "usrquota"},
199         {Opt_grpquota, "grpquota"},
200         {Opt_err, NULL}
201 };
202
203 #ifdef CONFIG_DEBUG_FS
204 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
205 {
206         int out = 0;
207         int i;
208         struct ocfs2_cluster_connection *cconn = osb->cconn;
209         struct ocfs2_recovery_map *rm = osb->recovery_map;
210
211         out += snprintf(buf + out, len - out,
212                         "%10s => Id: %-s  Uuid: %-s  Gen: 0x%X  Label: %-s\n",
213                         "Device", osb->dev_str, osb->uuid_str,
214                         osb->fs_generation, osb->vol_label);
215
216         out += snprintf(buf + out, len - out,
217                         "%10s => State: %d  Flags: 0x%lX\n", "Volume",
218                         atomic_read(&osb->vol_state), osb->osb_flags);
219
220         out += snprintf(buf + out, len - out,
221                         "%10s => Block: %lu  Cluster: %d\n", "Sizes",
222                         osb->sb->s_blocksize, osb->s_clustersize);
223
224         out += snprintf(buf + out, len - out,
225                         "%10s => Compat: 0x%X  Incompat: 0x%X  "
226                         "ROcompat: 0x%X\n",
227                         "Features", osb->s_feature_compat,
228                         osb->s_feature_incompat, osb->s_feature_ro_compat);
229
230         out += snprintf(buf + out, len - out,
231                         "%10s => Opts: 0x%lX  AtimeQuanta: %u\n", "Mount",
232                         osb->s_mount_opt, osb->s_atime_quantum);
233
234         out += snprintf(buf + out, len - out,
235                         "%10s => Stack: %s  Name: %*s  Version: %d.%d\n",
236                         "Cluster",
237                         (*osb->osb_cluster_stack == '\0' ?
238                          "o2cb" : osb->osb_cluster_stack),
239                         cconn->cc_namelen, cconn->cc_name,
240                         cconn->cc_version.pv_major, cconn->cc_version.pv_minor);
241
242         spin_lock(&osb->dc_task_lock);
243         out += snprintf(buf + out, len - out,
244                         "%10s => Pid: %d  Count: %lu  WakeSeq: %lu  "
245                         "WorkSeq: %lu\n", "DownCnvt",
246                         task_pid_nr(osb->dc_task), osb->blocked_lock_count,
247                         osb->dc_wake_sequence, osb->dc_work_sequence);
248         spin_unlock(&osb->dc_task_lock);
249
250         spin_lock(&osb->osb_lock);
251         out += snprintf(buf + out, len - out, "%10s => Pid: %d  Nodes:",
252                         "Recovery",
253                         (osb->recovery_thread_task ?
254                          task_pid_nr(osb->recovery_thread_task) : -1));
255         if (rm->rm_used == 0)
256                 out += snprintf(buf + out, len - out, " None\n");
257         else {
258                 for (i = 0; i < rm->rm_used; i++)
259                         out += snprintf(buf + out, len - out, " %d",
260                                         rm->rm_entries[i]);
261                 out += snprintf(buf + out, len - out, "\n");
262         }
263         spin_unlock(&osb->osb_lock);
264
265         out += snprintf(buf + out, len - out,
266                         "%10s => Pid: %d  Interval: %lu  Needs: %d\n", "Commit",
267                         task_pid_nr(osb->commit_task), osb->osb_commit_interval,
268                         atomic_read(&osb->needs_checkpoint));
269
270         out += snprintf(buf + out, len - out,
271                         "%10s => State: %d  NumTxns: %d  TxnId: %lu\n",
272                         "Journal", osb->journal->j_state,
273                         atomic_read(&osb->journal->j_num_trans),
274                         osb->journal->j_trans_id);
275
276         out += snprintf(buf + out, len - out,
277                         "%10s => GlobalAllocs: %d  LocalAllocs: %d  "
278                         "SubAllocs: %d  LAWinMoves: %d  SAExtends: %d\n",
279                         "Stats",
280                         atomic_read(&osb->alloc_stats.bitmap_data),
281                         atomic_read(&osb->alloc_stats.local_data),
282                         atomic_read(&osb->alloc_stats.bg_allocs),
283                         atomic_read(&osb->alloc_stats.moves),
284                         atomic_read(&osb->alloc_stats.bg_extends));
285
286         out += snprintf(buf + out, len - out,
287                         "%10s => State: %u  Descriptor: %llu  Size: %u bits  "
288                         "Default: %u bits\n",
289                         "LocalAlloc", osb->local_alloc_state,
290                         (unsigned long long)osb->la_last_gd,
291                         osb->local_alloc_bits, osb->local_alloc_default_bits);
292
293         spin_lock(&osb->osb_lock);
294         out += snprintf(buf + out, len - out,
295                         "%10s => Slot: %d  NumStolen: %d\n", "Steal",
296                         osb->s_inode_steal_slot,
297                         atomic_read(&osb->s_num_inodes_stolen));
298         spin_unlock(&osb->osb_lock);
299
300         out += snprintf(buf + out, len - out, "%10s => %3s  %10s\n",
301                         "Slots", "Num", "RecoGen");
302
303         for (i = 0; i < osb->max_slots; ++i) {
304                 out += snprintf(buf + out, len - out,
305                                 "%10s  %c %3d  %10d\n",
306                                 " ",
307                                 (i == osb->slot_num ? '*' : ' '),
308                                 i, osb->slot_recovery_generations[i]);
309         }
310
311         return out;
312 }
313
314 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
315 {
316         struct ocfs2_super *osb = inode->i_private;
317         char *buf = NULL;
318
319         buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
320         if (!buf)
321                 goto bail;
322
323         i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
324
325         file->private_data = buf;
326
327         return 0;
328 bail:
329         return -ENOMEM;
330 }
331
332 static int ocfs2_debug_release(struct inode *inode, struct file *file)
333 {
334         kfree(file->private_data);
335         return 0;
336 }
337
338 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
339                                 size_t nbytes, loff_t *ppos)
340 {
341         return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
342                                        i_size_read(file->f_mapping->host));
343 }
344 #else
345 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
346 {
347         return 0;
348 }
349 static int ocfs2_debug_release(struct inode *inode, struct file *file)
350 {
351         return 0;
352 }
353 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
354                                 size_t nbytes, loff_t *ppos)
355 {
356         return 0;
357 }
358 #endif  /* CONFIG_DEBUG_FS */
359
360 static struct file_operations ocfs2_osb_debug_fops = {
361         .open =         ocfs2_osb_debug_open,
362         .release =      ocfs2_debug_release,
363         .read =         ocfs2_debug_read,
364         .llseek =       generic_file_llseek,
365 };
366
367 static int ocfs2_sync_fs(struct super_block *sb, int wait)
368 {
369         int status;
370         tid_t target;
371         struct ocfs2_super *osb = OCFS2_SB(sb);
372
373         if (ocfs2_is_hard_readonly(osb))
374                 return -EROFS;
375
376         if (wait) {
377                 status = ocfs2_flush_truncate_log(osb);
378                 if (status < 0)
379                         mlog_errno(status);
380         } else {
381                 ocfs2_schedule_truncate_log_flush(osb, 0);
382         }
383
384         if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
385                                       &target)) {
386                 if (wait)
387                         jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
388                                              target);
389         }
390         return 0;
391 }
392
393 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
394 {
395         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
396             && (ino == USER_QUOTA_SYSTEM_INODE
397                 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
398                 return 0;
399         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
400             && (ino == GROUP_QUOTA_SYSTEM_INODE
401                 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
402                 return 0;
403         return 1;
404 }
405
406 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
407 {
408         struct inode *new = NULL;
409         int status = 0;
410         int i;
411
412         mlog_entry_void();
413
414         new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
415         if (IS_ERR(new)) {
416                 status = PTR_ERR(new);
417                 mlog_errno(status);
418                 goto bail;
419         }
420         osb->root_inode = new;
421
422         new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
423         if (IS_ERR(new)) {
424                 status = PTR_ERR(new);
425                 mlog_errno(status);
426                 goto bail;
427         }
428         osb->sys_root_inode = new;
429
430         for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
431              i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
432                 if (!ocfs2_need_system_inode(osb, i))
433                         continue;
434                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
435                 if (!new) {
436                         ocfs2_release_system_inodes(osb);
437                         status = -EINVAL;
438                         mlog_errno(status);
439                         /* FIXME: Should ERROR_RO_FS */
440                         mlog(ML_ERROR, "Unable to load system inode %d, "
441                              "possibly corrupt fs?", i);
442                         goto bail;
443                 }
444                 // the array now has one ref, so drop this one
445                 iput(new);
446         }
447
448 bail:
449         mlog_exit(status);
450         return status;
451 }
452
453 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
454 {
455         struct inode *new = NULL;
456         int status = 0;
457         int i;
458
459         mlog_entry_void();
460
461         for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
462              i < NUM_SYSTEM_INODES;
463              i++) {
464                 if (!ocfs2_need_system_inode(osb, i))
465                         continue;
466                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
467                 if (!new) {
468                         ocfs2_release_system_inodes(osb);
469                         status = -EINVAL;
470                         mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
471                              status, i, osb->slot_num);
472                         goto bail;
473                 }
474                 /* the array now has one ref, so drop this one */
475                 iput(new);
476         }
477
478 bail:
479         mlog_exit(status);
480         return status;
481 }
482
483 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
484 {
485         int i;
486         struct inode *inode;
487
488         mlog_entry_void();
489
490         for (i = 0; i < NUM_SYSTEM_INODES; i++) {
491                 inode = osb->system_inodes[i];
492                 if (inode) {
493                         iput(inode);
494                         osb->system_inodes[i] = NULL;
495                 }
496         }
497
498         inode = osb->sys_root_inode;
499         if (inode) {
500                 iput(inode);
501                 osb->sys_root_inode = NULL;
502         }
503
504         inode = osb->root_inode;
505         if (inode) {
506                 iput(inode);
507                 osb->root_inode = NULL;
508         }
509
510         mlog_exit(0);
511 }
512
513 /* We're allocating fs objects, use GFP_NOFS */
514 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
515 {
516         struct ocfs2_inode_info *oi;
517
518         oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
519         if (!oi)
520                 return NULL;
521
522         jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
523         return &oi->vfs_inode;
524 }
525
526 static void ocfs2_destroy_inode(struct inode *inode)
527 {
528         kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
529 }
530
531 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
532                                                 unsigned int cbits)
533 {
534         unsigned int bytes = 1 << cbits;
535         unsigned int trim = bytes;
536         unsigned int bitshift = 32;
537
538         /*
539          * i_size and all block offsets in ocfs2 are always 64 bits
540          * wide. i_clusters is 32 bits, in cluster-sized units. So on
541          * 64 bit platforms, cluster size will be the limiting factor.
542          */
543
544 #if BITS_PER_LONG == 32
545 # if defined(CONFIG_LBD)
546         BUILD_BUG_ON(sizeof(sector_t) != 8);
547         /*
548          * We might be limited by page cache size.
549          */
550         if (bytes > PAGE_CACHE_SIZE) {
551                 bytes = PAGE_CACHE_SIZE;
552                 trim = 1;
553                 /*
554                  * Shift by 31 here so that we don't get larger than
555                  * MAX_LFS_FILESIZE
556                  */
557                 bitshift = 31;
558         }
559 # else
560         /*
561          * We are limited by the size of sector_t. Use block size, as
562          * that's what we expose to the VFS.
563          */
564         bytes = 1 << bbits;
565         trim = 1;
566         bitshift = 31;
567 # endif
568 #endif
569
570         /*
571          * Trim by a whole cluster when we can actually approach the
572          * on-disk limits. Otherwise we can overflow i_clusters when
573          * an extent start is at the max offset.
574          */
575         return (((unsigned long long)bytes) << bitshift) - trim;
576 }
577
578 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
579 {
580         int incompat_features;
581         int ret = 0;
582         struct mount_options parsed_options;
583         struct ocfs2_super *osb = OCFS2_SB(sb);
584
585         lock_kernel();
586
587         if (!ocfs2_parse_options(sb, data, &parsed_options, 1)) {
588                 ret = -EINVAL;
589                 goto out;
590         }
591
592         if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) !=
593             (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
594                 ret = -EINVAL;
595                 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
596                 goto out;
597         }
598
599         if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
600             (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
601                 ret = -EINVAL;
602                 mlog(ML_ERROR, "Cannot change data mode on remount\n");
603                 goto out;
604         }
605
606         /* Probably don't want this on remount; it might
607          * mess with other nodes */
608         if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
609             (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
610                 ret = -EINVAL;
611                 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
612                 goto out;
613         }
614
615         /* We're going to/from readonly mode. */
616         if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
617                 /* Disable quota accounting before remounting RO */
618                 if (*flags & MS_RDONLY) {
619                         ret = ocfs2_susp_quotas(osb, 0);
620                         if (ret < 0)
621                                 goto out;
622                 }
623                 /* Lock here so the check of HARD_RO and the potential
624                  * setting of SOFT_RO is atomic. */
625                 spin_lock(&osb->osb_lock);
626                 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
627                         mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
628                         ret = -EROFS;
629                         goto unlock_osb;
630                 }
631
632                 if (*flags & MS_RDONLY) {
633                         mlog(0, "Going to ro mode.\n");
634                         sb->s_flags |= MS_RDONLY;
635                         osb->osb_flags |= OCFS2_OSB_SOFT_RO;
636                 } else {
637                         mlog(0, "Making ro filesystem writeable.\n");
638
639                         if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
640                                 mlog(ML_ERROR, "Cannot remount RDWR "
641                                      "filesystem due to previous errors.\n");
642                                 ret = -EROFS;
643                                 goto unlock_osb;
644                         }
645                         incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
646                         if (incompat_features) {
647                                 mlog(ML_ERROR, "Cannot remount RDWR because "
648                                      "of unsupported optional features "
649                                      "(%x).\n", incompat_features);
650                                 ret = -EINVAL;
651                                 goto unlock_osb;
652                         }
653                         sb->s_flags &= ~MS_RDONLY;
654                         osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
655                 }
656 unlock_osb:
657                 spin_unlock(&osb->osb_lock);
658                 /* Enable quota accounting after remounting RW */
659                 if (!ret && !(*flags & MS_RDONLY)) {
660                         if (sb_any_quota_suspended(sb))
661                                 ret = ocfs2_susp_quotas(osb, 1);
662                         else
663                                 ret = ocfs2_enable_quotas(osb);
664                         if (ret < 0) {
665                                 /* Return back changes... */
666                                 spin_lock(&osb->osb_lock);
667                                 sb->s_flags |= MS_RDONLY;
668                                 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
669                                 spin_unlock(&osb->osb_lock);
670                                 goto out;
671                         }
672                 }
673         }
674
675         if (!ret) {
676                 /* Only save off the new mount options in case of a successful
677                  * remount. */
678                 if (!(osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_XATTR))
679                         parsed_options.mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
680                 osb->s_mount_opt = parsed_options.mount_opt;
681                 osb->s_atime_quantum = parsed_options.atime_quantum;
682                 osb->preferred_slot = parsed_options.slot;
683                 if (parsed_options.commit_interval)
684                         osb->osb_commit_interval = parsed_options.commit_interval;
685
686                 if (!ocfs2_is_hard_readonly(osb))
687                         ocfs2_set_journal_params(osb);
688         }
689 out:
690         unlock_kernel();
691         return ret;
692 }
693
694 static int ocfs2_sb_probe(struct super_block *sb,
695                           struct buffer_head **bh,
696                           int *sector_size)
697 {
698         int status, tmpstat;
699         struct ocfs1_vol_disk_hdr *hdr;
700         struct ocfs2_dinode *di;
701         int blksize;
702
703         *bh = NULL;
704
705         /* may be > 512 */
706         *sector_size = bdev_logical_block_size(sb->s_bdev);
707         if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
708                 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
709                      *sector_size, OCFS2_MAX_BLOCKSIZE);
710                 status = -EINVAL;
711                 goto bail;
712         }
713
714         /* Can this really happen? */
715         if (*sector_size < OCFS2_MIN_BLOCKSIZE)
716                 *sector_size = OCFS2_MIN_BLOCKSIZE;
717
718         /* check block zero for old format */
719         status = ocfs2_get_sector(sb, bh, 0, *sector_size);
720         if (status < 0) {
721                 mlog_errno(status);
722                 goto bail;
723         }
724         hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
725         if (hdr->major_version == OCFS1_MAJOR_VERSION) {
726                 mlog(ML_ERROR, "incompatible version: %u.%u\n",
727                      hdr->major_version, hdr->minor_version);
728                 status = -EINVAL;
729         }
730         if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
731                    strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
732                 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
733                      hdr->signature);
734                 status = -EINVAL;
735         }
736         brelse(*bh);
737         *bh = NULL;
738         if (status < 0) {
739                 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
740                      "upgraded before mounting with ocfs v2\n");
741                 goto bail;
742         }
743
744         /*
745          * Now check at magic offset for 512, 1024, 2048, 4096
746          * blocksizes.  4096 is the maximum blocksize because it is
747          * the minimum clustersize.
748          */
749         status = -EINVAL;
750         for (blksize = *sector_size;
751              blksize <= OCFS2_MAX_BLOCKSIZE;
752              blksize <<= 1) {
753                 tmpstat = ocfs2_get_sector(sb, bh,
754                                            OCFS2_SUPER_BLOCK_BLKNO,
755                                            blksize);
756                 if (tmpstat < 0) {
757                         status = tmpstat;
758                         mlog_errno(status);
759                         goto bail;
760                 }
761                 di = (struct ocfs2_dinode *) (*bh)->b_data;
762                 status = ocfs2_verify_volume(di, *bh, blksize);
763                 if (status >= 0)
764                         goto bail;
765                 brelse(*bh);
766                 *bh = NULL;
767                 if (status != -EAGAIN)
768                         break;
769         }
770
771 bail:
772         return status;
773 }
774
775 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
776 {
777         if (ocfs2_mount_local(osb)) {
778                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
779                         mlog(ML_ERROR, "Cannot heartbeat on a locally "
780                              "mounted device.\n");
781                         return -EINVAL;
782                 }
783         }
784
785         if (ocfs2_userspace_stack(osb)) {
786                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
787                         mlog(ML_ERROR, "Userspace stack expected, but "
788                              "o2cb heartbeat arguments passed to mount\n");
789                         return -EINVAL;
790                 }
791         }
792
793         if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
794                 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
795                     !ocfs2_userspace_stack(osb)) {
796                         mlog(ML_ERROR, "Heartbeat has to be started to mount "
797                              "a read-write clustered device.\n");
798                         return -EINVAL;
799                 }
800         }
801
802         return 0;
803 }
804
805 /*
806  * If we're using a userspace stack, mount should have passed
807  * a name that matches the disk.  If not, mount should not
808  * have passed a stack.
809  */
810 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
811                                         struct mount_options *mopt)
812 {
813         if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
814                 mlog(ML_ERROR,
815                      "cluster stack passed to mount, but this filesystem "
816                      "does not support it\n");
817                 return -EINVAL;
818         }
819
820         if (ocfs2_userspace_stack(osb) &&
821             strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
822                     OCFS2_STACK_LABEL_LEN)) {
823                 mlog(ML_ERROR,
824                      "cluster stack passed to mount (\"%s\") does not "
825                      "match the filesystem (\"%s\")\n",
826                      mopt->cluster_stack,
827                      osb->osb_cluster_stack);
828                 return -EINVAL;
829         }
830
831         return 0;
832 }
833
834 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
835 {
836         int type;
837         struct super_block *sb = osb->sb;
838         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
839                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
840         int status = 0;
841
842         for (type = 0; type < MAXQUOTAS; type++) {
843                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
844                         continue;
845                 if (unsuspend)
846                         status = vfs_quota_enable(
847                                         sb_dqopt(sb)->files[type],
848                                         type, QFMT_OCFS2,
849                                         DQUOT_SUSPENDED);
850                 else
851                         status = vfs_quota_disable(sb, type,
852                                                    DQUOT_SUSPENDED);
853                 if (status < 0)
854                         break;
855         }
856         if (status < 0)
857                 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
858                      "remount (error = %d).\n", status);
859         return status;
860 }
861
862 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
863 {
864         struct inode *inode[MAXQUOTAS] = { NULL, NULL };
865         struct super_block *sb = osb->sb;
866         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
867                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
868         unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
869                                         LOCAL_GROUP_QUOTA_SYSTEM_INODE };
870         int status;
871         int type;
872
873         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
874         for (type = 0; type < MAXQUOTAS; type++) {
875                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
876                         continue;
877                 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
878                                                         osb->slot_num);
879                 if (!inode[type]) {
880                         status = -ENOENT;
881                         goto out_quota_off;
882                 }
883                 status = vfs_quota_enable(inode[type], type, QFMT_OCFS2,
884                                                 DQUOT_USAGE_ENABLED);
885                 if (status < 0)
886                         goto out_quota_off;
887         }
888
889         for (type = 0; type < MAXQUOTAS; type++)
890                 iput(inode[type]);
891         return 0;
892 out_quota_off:
893         ocfs2_disable_quotas(osb);
894         for (type = 0; type < MAXQUOTAS; type++)
895                 iput(inode[type]);
896         mlog_errno(status);
897         return status;
898 }
899
900 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
901 {
902         int type;
903         struct inode *inode;
904         struct super_block *sb = osb->sb;
905
906         /* We mostly ignore errors in this function because there's not much
907          * we can do when we see them */
908         for (type = 0; type < MAXQUOTAS; type++) {
909                 if (!sb_has_quota_loaded(sb, type))
910                         continue;
911                 inode = igrab(sb->s_dquot.files[type]);
912                 /* Turn off quotas. This will remove all dquot structures from
913                  * memory and so they will be automatically synced to global
914                  * quota files */
915                 vfs_quota_disable(sb, type, DQUOT_USAGE_ENABLED |
916                                             DQUOT_LIMITS_ENABLED);
917                 if (!inode)
918                         continue;
919                 iput(inode);
920         }
921 }
922
923 /* Handle quota on quotactl */
924 static int ocfs2_quota_on(struct super_block *sb, int type, int format_id,
925                           char *path, int remount)
926 {
927         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
928                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
929
930         if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
931                 return -EINVAL;
932
933         if (remount)
934                 return 0;       /* Just ignore it has been handled in
935                                  * ocfs2_remount() */
936         return vfs_quota_enable(sb_dqopt(sb)->files[type], type,
937                                     format_id, DQUOT_LIMITS_ENABLED);
938 }
939
940 /* Handle quota off quotactl */
941 static int ocfs2_quota_off(struct super_block *sb, int type, int remount)
942 {
943         if (remount)
944                 return 0;       /* Ignore now and handle later in
945                                  * ocfs2_remount() */
946         return vfs_quota_disable(sb, type, DQUOT_LIMITS_ENABLED);
947 }
948
949 static struct quotactl_ops ocfs2_quotactl_ops = {
950         .quota_on       = ocfs2_quota_on,
951         .quota_off      = ocfs2_quota_off,
952         .quota_sync     = vfs_quota_sync,
953         .get_info       = vfs_get_dqinfo,
954         .set_info       = vfs_set_dqinfo,
955         .get_dqblk      = vfs_get_dqblk,
956         .set_dqblk      = vfs_set_dqblk,
957 };
958
959 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
960 {
961         struct dentry *root;
962         int status, sector_size;
963         struct mount_options parsed_options;
964         struct inode *inode = NULL;
965         struct ocfs2_super *osb = NULL;
966         struct buffer_head *bh = NULL;
967         char nodestr[8];
968
969         mlog_entry("%p, %p, %i", sb, data, silent);
970
971         if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
972                 status = -EINVAL;
973                 goto read_super_error;
974         }
975
976         /* probe for superblock */
977         status = ocfs2_sb_probe(sb, &bh, &sector_size);
978         if (status < 0) {
979                 mlog(ML_ERROR, "superblock probe failed!\n");
980                 goto read_super_error;
981         }
982
983         status = ocfs2_initialize_super(sb, bh, sector_size);
984         osb = OCFS2_SB(sb);
985         if (status < 0) {
986                 mlog_errno(status);
987                 goto read_super_error;
988         }
989         brelse(bh);
990         bh = NULL;
991
992         if (!(osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_XATTR))
993                 parsed_options.mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
994
995         osb->s_mount_opt = parsed_options.mount_opt;
996         osb->s_atime_quantum = parsed_options.atime_quantum;
997         osb->preferred_slot = parsed_options.slot;
998         osb->osb_commit_interval = parsed_options.commit_interval;
999         osb->local_alloc_default_bits = ocfs2_megabytes_to_clusters(sb, parsed_options.localalloc_opt);
1000         osb->local_alloc_bits = osb->local_alloc_default_bits;
1001         if (osb->s_mount_opt & OCFS2_MOUNT_USRQUOTA &&
1002             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1003                                          OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1004                 status = -EINVAL;
1005                 mlog(ML_ERROR, "User quotas were requested, but this "
1006                      "filesystem does not have the feature enabled.\n");
1007                 goto read_super_error;
1008         }
1009         if (osb->s_mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1010             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1011                                          OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1012                 status = -EINVAL;
1013                 mlog(ML_ERROR, "Group quotas were requested, but this "
1014                      "filesystem does not have the feature enabled.\n");
1015                 goto read_super_error;
1016         }
1017
1018         status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1019         if (status)
1020                 goto read_super_error;
1021
1022         sb->s_magic = OCFS2_SUPER_MAGIC;
1023
1024         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1025                 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1026
1027         /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1028          * heartbeat=none */
1029         if (bdev_read_only(sb->s_bdev)) {
1030                 if (!(sb->s_flags & MS_RDONLY)) {
1031                         status = -EACCES;
1032                         mlog(ML_ERROR, "Readonly device detected but readonly "
1033                              "mount was not specified.\n");
1034                         goto read_super_error;
1035                 }
1036
1037                 /* You should not be able to start a local heartbeat
1038                  * on a readonly device. */
1039                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1040                         status = -EROFS;
1041                         mlog(ML_ERROR, "Local heartbeat specified on readonly "
1042                              "device.\n");
1043                         goto read_super_error;
1044                 }
1045
1046                 status = ocfs2_check_journals_nolocks(osb);
1047                 if (status < 0) {
1048                         if (status == -EROFS)
1049                                 mlog(ML_ERROR, "Recovery required on readonly "
1050                                      "file system, but write access is "
1051                                      "unavailable.\n");
1052                         else
1053                                 mlog_errno(status);                     
1054                         goto read_super_error;
1055                 }
1056
1057                 ocfs2_set_ro_flag(osb, 1);
1058
1059                 printk(KERN_NOTICE "Readonly device detected. No cluster "
1060                        "services will be utilized for this mount. Recovery "
1061                        "will be skipped.\n");
1062         }
1063
1064         if (!ocfs2_is_hard_readonly(osb)) {
1065                 if (sb->s_flags & MS_RDONLY)
1066                         ocfs2_set_ro_flag(osb, 0);
1067         }
1068
1069         status = ocfs2_verify_heartbeat(osb);
1070         if (status < 0) {
1071                 mlog_errno(status);
1072                 goto read_super_error;
1073         }
1074
1075         osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1076                                                  ocfs2_debugfs_root);
1077         if (!osb->osb_debug_root) {
1078                 status = -EINVAL;
1079                 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1080                 goto read_super_error;
1081         }
1082
1083         osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1084                                             osb->osb_debug_root,
1085                                             osb,
1086                                             &ocfs2_osb_debug_fops);
1087         if (!osb->osb_ctxt) {
1088                 status = -EINVAL;
1089                 mlog_errno(status);
1090                 goto read_super_error;
1091         }
1092
1093         status = ocfs2_mount_volume(sb);
1094         if (osb->root_inode)
1095                 inode = igrab(osb->root_inode);
1096
1097         if (status < 0)
1098                 goto read_super_error;
1099
1100         if (!inode) {
1101                 status = -EIO;
1102                 mlog_errno(status);
1103                 goto read_super_error;
1104         }
1105
1106         root = d_alloc_root(inode);
1107         if (!root) {
1108                 status = -ENOMEM;
1109                 mlog_errno(status);
1110                 goto read_super_error;
1111         }
1112
1113         sb->s_root = root;
1114
1115         ocfs2_complete_mount_recovery(osb);
1116
1117         if (ocfs2_mount_local(osb))
1118                 snprintf(nodestr, sizeof(nodestr), "local");
1119         else
1120                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1121
1122         printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1123                "with %s data mode.\n",
1124                osb->dev_str, nodestr, osb->slot_num,
1125                osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1126                "ordered");
1127
1128         atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1129         wake_up(&osb->osb_mount_event);
1130
1131         /* Now we can initialize quotas because we can afford to wait
1132          * for cluster locks recovery now. That also means that truncation
1133          * log recovery can happen but that waits for proper quota setup */
1134         if (!(sb->s_flags & MS_RDONLY)) {
1135                 status = ocfs2_enable_quotas(osb);
1136                 if (status < 0) {
1137                         /* We have to err-out specially here because
1138                          * s_root is already set */
1139                         mlog_errno(status);
1140                         atomic_set(&osb->vol_state, VOLUME_DISABLED);
1141                         wake_up(&osb->osb_mount_event);
1142                         mlog_exit(status);
1143                         return status;
1144                 }
1145         }
1146
1147         ocfs2_complete_quota_recovery(osb);
1148
1149         /* Now we wake up again for processes waiting for quotas */
1150         atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1151         wake_up(&osb->osb_mount_event);
1152
1153         mlog_exit(status);
1154         return status;
1155
1156 read_super_error:
1157         brelse(bh);
1158
1159         if (inode)
1160                 iput(inode);
1161
1162         if (osb) {
1163                 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1164                 wake_up(&osb->osb_mount_event);
1165                 ocfs2_dismount_volume(sb, 1);
1166         }
1167
1168         mlog_exit(status);
1169         return status;
1170 }
1171
1172 static int ocfs2_get_sb(struct file_system_type *fs_type,
1173                         int flags,
1174                         const char *dev_name,
1175                         void *data,
1176                         struct vfsmount *mnt)
1177 {
1178         return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super,
1179                            mnt);
1180 }
1181
1182 static struct file_system_type ocfs2_fs_type = {
1183         .owner          = THIS_MODULE,
1184         .name           = "ocfs2",
1185         .get_sb         = ocfs2_get_sb, /* is this called when we mount
1186                                         * the fs? */
1187         .kill_sb        = kill_block_super, /* set to the generic one
1188                                              * right now, but do we
1189                                              * need to change that? */
1190         .fs_flags       = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1191         .next           = NULL
1192 };
1193
1194 static int ocfs2_parse_options(struct super_block *sb,
1195                                char *options,
1196                                struct mount_options *mopt,
1197                                int is_remount)
1198 {
1199         int status;
1200         char *p;
1201
1202         mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
1203                    options ? options : "(none)");
1204
1205         mopt->commit_interval = 0;
1206         mopt->mount_opt = 0;
1207         mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1208         mopt->slot = OCFS2_INVALID_SLOT;
1209         mopt->localalloc_opt = OCFS2_DEFAULT_LOCAL_ALLOC_SIZE;
1210         mopt->cluster_stack[0] = '\0';
1211
1212         if (!options) {
1213                 status = 1;
1214                 goto bail;
1215         }
1216
1217         while ((p = strsep(&options, ",")) != NULL) {
1218                 int token, option;
1219                 substring_t args[MAX_OPT_ARGS];
1220
1221                 if (!*p)
1222                         continue;
1223
1224                 token = match_token(p, tokens, args);
1225                 switch (token) {
1226                 case Opt_hb_local:
1227                         mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1228                         break;
1229                 case Opt_hb_none:
1230                         mopt->mount_opt &= ~OCFS2_MOUNT_HB_LOCAL;
1231                         break;
1232                 case Opt_barrier:
1233                         if (match_int(&args[0], &option)) {
1234                                 status = 0;
1235                                 goto bail;
1236                         }
1237                         if (option)
1238                                 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1239                         else
1240                                 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1241                         break;
1242                 case Opt_intr:
1243                         mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1244                         break;
1245                 case Opt_nointr:
1246                         mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1247                         break;
1248                 case Opt_err_panic:
1249                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1250                         break;
1251                 case Opt_err_ro:
1252                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1253                         break;
1254                 case Opt_data_ordered:
1255                         mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1256                         break;
1257                 case Opt_data_writeback:
1258                         mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1259                         break;
1260                 case Opt_user_xattr:
1261                         mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1262                         break;
1263                 case Opt_nouser_xattr:
1264                         mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1265                         break;
1266                 case Opt_atime_quantum:
1267                         if (match_int(&args[0], &option)) {
1268                                 status = 0;
1269                                 goto bail;
1270                         }
1271                         if (option >= 0)
1272                                 mopt->atime_quantum = option;
1273                         break;
1274                 case Opt_slot:
1275                         option = 0;
1276                         if (match_int(&args[0], &option)) {
1277                                 status = 0;
1278                                 goto bail;
1279                         }
1280                         if (option)
1281                                 mopt->slot = (s16)option;
1282                         break;
1283                 case Opt_commit:
1284                         option = 0;
1285                         if (match_int(&args[0], &option)) {
1286                                 status = 0;
1287                                 goto bail;
1288                         }
1289                         if (option < 0)
1290                                 return 0;
1291                         if (option == 0)
1292                                 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1293                         mopt->commit_interval = HZ * option;
1294                         break;
1295                 case Opt_localalloc:
1296                         option = 0;
1297                         if (match_int(&args[0], &option)) {
1298                                 status = 0;
1299                                 goto bail;
1300                         }
1301                         if (option >= 0 && (option <= ocfs2_local_alloc_size(sb) * 8))
1302                                 mopt->localalloc_opt = option;
1303                         break;
1304                 case Opt_localflocks:
1305                         /*
1306                          * Changing this during remount could race
1307                          * flock() requests, or "unbalance" existing
1308                          * ones (e.g., a lock is taken in one mode but
1309                          * dropped in the other). If users care enough
1310                          * to flip locking modes during remount, we
1311                          * could add a "local" flag to individual
1312                          * flock structures for proper tracking of
1313                          * state.
1314                          */
1315                         if (!is_remount)
1316                                 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1317                         break;
1318                 case Opt_stack:
1319                         /* Check both that the option we were passed
1320                          * is of the right length and that it is a proper
1321                          * string of the right length.
1322                          */
1323                         if (((args[0].to - args[0].from) !=
1324                              OCFS2_STACK_LABEL_LEN) ||
1325                             (strnlen(args[0].from,
1326                                      OCFS2_STACK_LABEL_LEN) !=
1327                              OCFS2_STACK_LABEL_LEN)) {
1328                                 mlog(ML_ERROR,
1329                                      "Invalid cluster_stack option\n");
1330                                 status = 0;
1331                                 goto bail;
1332                         }
1333                         memcpy(mopt->cluster_stack, args[0].from,
1334                                OCFS2_STACK_LABEL_LEN);
1335                         mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1336                         break;
1337                 case Opt_inode64:
1338                         mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1339                         break;
1340                 case Opt_usrquota:
1341                         /* We check only on remount, otherwise features
1342                          * aren't yet initialized. */
1343                         if (is_remount && !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1344                             OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1345                                 mlog(ML_ERROR, "User quota requested but "
1346                                      "filesystem feature is not set\n");
1347                                 status = 0;
1348                                 goto bail;
1349                         }
1350                         mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1351                         break;
1352                 case Opt_grpquota:
1353                         if (is_remount && !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1354                             OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1355                                 mlog(ML_ERROR, "Group quota requested but "
1356                                      "filesystem feature is not set\n");
1357                                 status = 0;
1358                                 goto bail;
1359                         }
1360                         mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1361                         break;
1362 #ifdef CONFIG_OCFS2_FS_POSIX_ACL
1363                 case Opt_acl:
1364                         mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1365                         break;
1366                 case Opt_noacl:
1367                         mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1368                         break;
1369 #else
1370                 case Opt_acl:
1371                 case Opt_noacl:
1372                         printk(KERN_INFO "ocfs2 (no)acl options not supported\n");
1373                         break;
1374 #endif
1375                 default:
1376                         mlog(ML_ERROR,
1377                              "Unrecognized mount option \"%s\" "
1378                              "or missing value\n", p);
1379                         status = 0;
1380                         goto bail;
1381                 }
1382         }
1383
1384         status = 1;
1385
1386 bail:
1387         mlog_exit(status);
1388         return status;
1389 }
1390
1391 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1392 {
1393         struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb);
1394         unsigned long opts = osb->s_mount_opt;
1395         unsigned int local_alloc_megs;
1396
1397         if (opts & OCFS2_MOUNT_HB_LOCAL)
1398                 seq_printf(s, ",_netdev,heartbeat=local");
1399         else
1400                 seq_printf(s, ",heartbeat=none");
1401
1402         if (opts & OCFS2_MOUNT_NOINTR)
1403                 seq_printf(s, ",nointr");
1404
1405         if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1406                 seq_printf(s, ",data=writeback");
1407         else
1408                 seq_printf(s, ",data=ordered");
1409
1410         if (opts & OCFS2_MOUNT_BARRIER)
1411                 seq_printf(s, ",barrier=1");
1412
1413         if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1414                 seq_printf(s, ",errors=panic");
1415         else
1416                 seq_printf(s, ",errors=remount-ro");
1417
1418         if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1419                 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1420
1421         if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM)
1422                 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1423
1424         if (osb->osb_commit_interval)
1425                 seq_printf(s, ",commit=%u",
1426                            (unsigned) (osb->osb_commit_interval / HZ));
1427
1428         local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1429         if (local_alloc_megs != OCFS2_DEFAULT_LOCAL_ALLOC_SIZE)
1430                 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1431
1432         if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1433                 seq_printf(s, ",localflocks,");
1434
1435         if (osb->osb_cluster_stack[0])
1436                 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1437                            osb->osb_cluster_stack);
1438         if (opts & OCFS2_MOUNT_USRQUOTA)
1439                 seq_printf(s, ",usrquota");
1440         if (opts & OCFS2_MOUNT_GRPQUOTA)
1441                 seq_printf(s, ",grpquota");
1442
1443         if (opts & OCFS2_MOUNT_NOUSERXATTR)
1444                 seq_printf(s, ",nouser_xattr");
1445         else
1446                 seq_printf(s, ",user_xattr");
1447
1448         if (opts & OCFS2_MOUNT_INODE64)
1449                 seq_printf(s, ",inode64");
1450
1451 #ifdef CONFIG_OCFS2_FS_POSIX_ACL
1452         if (opts & OCFS2_MOUNT_POSIX_ACL)
1453                 seq_printf(s, ",acl");
1454         else
1455                 seq_printf(s, ",noacl");
1456 #endif
1457
1458         return 0;
1459 }
1460
1461 static int __init ocfs2_init(void)
1462 {
1463         int status;
1464
1465         mlog_entry_void();
1466
1467         ocfs2_print_version();
1468
1469         status = init_ocfs2_uptodate_cache();
1470         if (status < 0) {
1471                 mlog_errno(status);
1472                 goto leave;
1473         }
1474
1475         status = ocfs2_initialize_mem_caches();
1476         if (status < 0) {
1477                 mlog_errno(status);
1478                 goto leave;
1479         }
1480
1481         ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1482         if (!ocfs2_wq) {
1483                 status = -ENOMEM;
1484                 goto leave;
1485         }
1486
1487         ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1488         if (!ocfs2_debugfs_root) {
1489                 status = -EFAULT;
1490                 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1491         }
1492
1493         status = ocfs2_quota_setup();
1494         if (status)
1495                 goto leave;
1496
1497         ocfs2_set_locking_protocol();
1498
1499         status = register_quota_format(&ocfs2_quota_format);
1500 leave:
1501         if (status < 0) {
1502                 ocfs2_quota_shutdown();
1503                 ocfs2_free_mem_caches();
1504                 exit_ocfs2_uptodate_cache();
1505         }
1506
1507         mlog_exit(status);
1508
1509         if (status >= 0) {
1510                 return register_filesystem(&ocfs2_fs_type);
1511         } else
1512                 return -1;
1513 }
1514
1515 static void __exit ocfs2_exit(void)
1516 {
1517         mlog_entry_void();
1518
1519         ocfs2_quota_shutdown();
1520
1521         if (ocfs2_wq) {
1522                 flush_workqueue(ocfs2_wq);
1523                 destroy_workqueue(ocfs2_wq);
1524         }
1525
1526         unregister_quota_format(&ocfs2_quota_format);
1527
1528         debugfs_remove(ocfs2_debugfs_root);
1529
1530         ocfs2_free_mem_caches();
1531
1532         unregister_filesystem(&ocfs2_fs_type);
1533
1534         exit_ocfs2_uptodate_cache();
1535
1536         mlog_exit_void();
1537 }
1538
1539 static void ocfs2_put_super(struct super_block *sb)
1540 {
1541         mlog_entry("(0x%p)\n", sb);
1542
1543         lock_kernel();
1544
1545         ocfs2_sync_blockdev(sb);
1546         ocfs2_dismount_volume(sb, 0);
1547
1548         unlock_kernel();
1549
1550         mlog_exit_void();
1551 }
1552
1553 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1554 {
1555         struct ocfs2_super *osb;
1556         u32 numbits, freebits;
1557         int status;
1558         struct ocfs2_dinode *bm_lock;
1559         struct buffer_head *bh = NULL;
1560         struct inode *inode = NULL;
1561
1562         mlog_entry("(%p, %p)\n", dentry->d_sb, buf);
1563
1564         osb = OCFS2_SB(dentry->d_sb);
1565
1566         inode = ocfs2_get_system_file_inode(osb,
1567                                             GLOBAL_BITMAP_SYSTEM_INODE,
1568                                             OCFS2_INVALID_SLOT);
1569         if (!inode) {
1570                 mlog(ML_ERROR, "failed to get bitmap inode\n");
1571                 status = -EIO;
1572                 goto bail;
1573         }
1574
1575         status = ocfs2_inode_lock(inode, &bh, 0);
1576         if (status < 0) {
1577                 mlog_errno(status);
1578                 goto bail;
1579         }
1580
1581         bm_lock = (struct ocfs2_dinode *) bh->b_data;
1582
1583         numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1584         freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1585
1586         buf->f_type = OCFS2_SUPER_MAGIC;
1587         buf->f_bsize = dentry->d_sb->s_blocksize;
1588         buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1589         buf->f_blocks = ((sector_t) numbits) *
1590                         (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1591         buf->f_bfree = ((sector_t) freebits) *
1592                        (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1593         buf->f_bavail = buf->f_bfree;
1594         buf->f_files = numbits;
1595         buf->f_ffree = freebits;
1596
1597         brelse(bh);
1598
1599         ocfs2_inode_unlock(inode, 0);
1600         status = 0;
1601 bail:
1602         if (inode)
1603                 iput(inode);
1604
1605         mlog_exit(status);
1606
1607         return status;
1608 }
1609
1610 static void ocfs2_inode_init_once(void *data)
1611 {
1612         struct ocfs2_inode_info *oi = data;
1613
1614         oi->ip_flags = 0;
1615         oi->ip_open_count = 0;
1616         spin_lock_init(&oi->ip_lock);
1617         ocfs2_extent_map_init(&oi->vfs_inode);
1618         INIT_LIST_HEAD(&oi->ip_io_markers);
1619         oi->ip_created_trans = 0;
1620         oi->ip_last_trans = 0;
1621         oi->ip_dir_start_lookup = 0;
1622
1623         init_rwsem(&oi->ip_alloc_sem);
1624         init_rwsem(&oi->ip_xattr_sem);
1625         mutex_init(&oi->ip_io_mutex);
1626
1627         oi->ip_blkno = 0ULL;
1628         oi->ip_clusters = 0;
1629
1630         ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1631         ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1632         ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1633
1634         ocfs2_metadata_cache_init(&oi->vfs_inode);
1635
1636         inode_init_once(&oi->vfs_inode);
1637 }
1638
1639 static int ocfs2_initialize_mem_caches(void)
1640 {
1641         ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1642                                        sizeof(struct ocfs2_inode_info),
1643                                        0,
1644                                        (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1645                                                 SLAB_MEM_SPREAD),
1646                                        ocfs2_inode_init_once);
1647         ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1648                                         sizeof(struct ocfs2_dquot),
1649                                         0,
1650                                         (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1651                                                 SLAB_MEM_SPREAD),
1652                                         NULL);
1653         ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1654                                         sizeof(struct ocfs2_quota_chunk),
1655                                         0,
1656                                         (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1657                                         NULL);
1658         if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1659             !ocfs2_qf_chunk_cachep) {
1660                 if (ocfs2_inode_cachep)
1661                         kmem_cache_destroy(ocfs2_inode_cachep);
1662                 if (ocfs2_dquot_cachep)
1663                         kmem_cache_destroy(ocfs2_dquot_cachep);
1664                 if (ocfs2_qf_chunk_cachep)
1665                         kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1666                 return -ENOMEM;
1667         }
1668
1669         return 0;
1670 }
1671
1672 static void ocfs2_free_mem_caches(void)
1673 {
1674         if (ocfs2_inode_cachep)
1675                 kmem_cache_destroy(ocfs2_inode_cachep);
1676         ocfs2_inode_cachep = NULL;
1677
1678         if (ocfs2_dquot_cachep)
1679                 kmem_cache_destroy(ocfs2_dquot_cachep);
1680         ocfs2_dquot_cachep = NULL;
1681
1682         if (ocfs2_qf_chunk_cachep)
1683                 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1684         ocfs2_qf_chunk_cachep = NULL;
1685 }
1686
1687 static int ocfs2_get_sector(struct super_block *sb,
1688                             struct buffer_head **bh,
1689                             int block,
1690                             int sect_size)
1691 {
1692         if (!sb_set_blocksize(sb, sect_size)) {
1693                 mlog(ML_ERROR, "unable to set blocksize\n");
1694                 return -EIO;
1695         }
1696
1697         *bh = sb_getblk(sb, block);
1698         if (!*bh) {
1699                 mlog_errno(-EIO);
1700                 return -EIO;
1701         }
1702         lock_buffer(*bh);
1703         if (!buffer_dirty(*bh))
1704                 clear_buffer_uptodate(*bh);
1705         unlock_buffer(*bh);
1706         ll_rw_block(READ, 1, bh);
1707         wait_on_buffer(*bh);
1708         if (!buffer_uptodate(*bh)) {
1709                 mlog_errno(-EIO);
1710                 brelse(*bh);
1711                 *bh = NULL;
1712                 return -EIO;
1713         }
1714
1715         return 0;
1716 }
1717
1718 static int ocfs2_mount_volume(struct super_block *sb)
1719 {
1720         int status = 0;
1721         int unlock_super = 0;
1722         struct ocfs2_super *osb = OCFS2_SB(sb);
1723
1724         mlog_entry_void();
1725
1726         if (ocfs2_is_hard_readonly(osb))
1727                 goto leave;
1728
1729         status = ocfs2_dlm_init(osb);
1730         if (status < 0) {
1731                 mlog_errno(status);
1732                 goto leave;
1733         }
1734
1735         status = ocfs2_super_lock(osb, 1);
1736         if (status < 0) {
1737                 mlog_errno(status);
1738                 goto leave;
1739         }
1740         unlock_super = 1;
1741
1742         /* This will load up the node map and add ourselves to it. */
1743         status = ocfs2_find_slot(osb);
1744         if (status < 0) {
1745                 mlog_errno(status);
1746                 goto leave;
1747         }
1748
1749         /* load all node-local system inodes */
1750         status = ocfs2_init_local_system_inodes(osb);
1751         if (status < 0) {
1752                 mlog_errno(status);
1753                 goto leave;
1754         }
1755
1756         status = ocfs2_check_volume(osb);
1757         if (status < 0) {
1758                 mlog_errno(status);
1759                 goto leave;
1760         }
1761
1762         status = ocfs2_truncate_log_init(osb);
1763         if (status < 0) {
1764                 mlog_errno(status);
1765                 goto leave;
1766         }
1767
1768         if (ocfs2_mount_local(osb))
1769                 goto leave;
1770
1771 leave:
1772         if (unlock_super)
1773                 ocfs2_super_unlock(osb, 1);
1774
1775         mlog_exit(status);
1776         return status;
1777 }
1778
1779 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1780 {
1781         int tmp, hangup_needed = 0;
1782         struct ocfs2_super *osb = NULL;
1783         char nodestr[8];
1784
1785         mlog_entry("(0x%p)\n", sb);
1786
1787         BUG_ON(!sb);
1788         osb = OCFS2_SB(sb);
1789         BUG_ON(!osb);
1790
1791         debugfs_remove(osb->osb_ctxt);
1792
1793         ocfs2_disable_quotas(osb);
1794
1795         ocfs2_shutdown_local_alloc(osb);
1796
1797         ocfs2_truncate_log_shutdown(osb);
1798
1799         /* This will disable recovery and flush any recovery work. */
1800         ocfs2_recovery_exit(osb);
1801
1802         ocfs2_journal_shutdown(osb);
1803
1804         ocfs2_sync_blockdev(sb);
1805
1806         /* No cluster connection means we've failed during mount, so skip
1807          * all the steps which depended on that to complete. */
1808         if (osb->cconn) {
1809                 tmp = ocfs2_super_lock(osb, 1);
1810                 if (tmp < 0) {
1811                         mlog_errno(tmp);
1812                         return;
1813                 }
1814         }
1815
1816         if (osb->slot_num != OCFS2_INVALID_SLOT)
1817                 ocfs2_put_slot(osb);
1818
1819         if (osb->cconn)
1820                 ocfs2_super_unlock(osb, 1);
1821
1822         ocfs2_release_system_inodes(osb);
1823
1824         /*
1825          * If we're dismounting due to mount error, mount.ocfs2 will clean
1826          * up heartbeat.  If we're a local mount, there is no heartbeat.
1827          * If we failed before we got a uuid_str yet, we can't stop
1828          * heartbeat.  Otherwise, do it.
1829          */
1830         if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str)
1831                 hangup_needed = 1;
1832
1833         if (osb->cconn)
1834                 ocfs2_dlm_shutdown(osb, hangup_needed);
1835
1836         debugfs_remove(osb->osb_debug_root);
1837
1838         if (hangup_needed)
1839                 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1840
1841         atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1842
1843         if (ocfs2_mount_local(osb))
1844                 snprintf(nodestr, sizeof(nodestr), "local");
1845         else
1846                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1847
1848         printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1849                osb->dev_str, nodestr);
1850
1851         ocfs2_delete_osb(osb);
1852         kfree(osb);
1853         sb->s_dev = 0;
1854         sb->s_fs_info = NULL;
1855 }
1856
1857 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1858                                 unsigned uuid_bytes)
1859 {
1860         int i, ret;
1861         char *ptr;
1862
1863         BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1864
1865         osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1866         if (osb->uuid_str == NULL)
1867                 return -ENOMEM;
1868
1869         for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1870                 /* print with null */
1871                 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1872                 if (ret != 2) /* drop super cleans up */
1873                         return -EINVAL;
1874                 /* then only advance past the last char */
1875                 ptr += 2;
1876         }
1877
1878         return 0;
1879 }
1880
1881 static int ocfs2_initialize_super(struct super_block *sb,
1882                                   struct buffer_head *bh,
1883                                   int sector_size)
1884 {
1885         int status;
1886         int i, cbits, bbits;
1887         struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
1888         struct inode *inode = NULL;
1889         struct ocfs2_journal *journal;
1890         __le32 uuid_net_key;
1891         struct ocfs2_super *osb;
1892
1893         mlog_entry_void();
1894
1895         osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
1896         if (!osb) {
1897                 status = -ENOMEM;
1898                 mlog_errno(status);
1899                 goto bail;
1900         }
1901
1902         sb->s_fs_info = osb;
1903         sb->s_op = &ocfs2_sops;
1904         sb->s_export_op = &ocfs2_export_ops;
1905         sb->s_qcop = &ocfs2_quotactl_ops;
1906         sb->dq_op = &ocfs2_quota_operations;
1907         sb->s_xattr = ocfs2_xattr_handlers;
1908         sb->s_time_gran = 1;
1909         sb->s_flags |= MS_NOATIME;
1910         /* this is needed to support O_LARGEFILE */
1911         cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
1912         bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
1913         sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
1914
1915         osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
1916
1917         for (i = 0; i < 3; i++)
1918                 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
1919         osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
1920
1921         osb->sb = sb;
1922         /* Save off for ocfs2_rw_direct */
1923         osb->s_sectsize_bits = blksize_bits(sector_size);
1924         BUG_ON(!osb->s_sectsize_bits);
1925
1926         spin_lock_init(&osb->dc_task_lock);
1927         init_waitqueue_head(&osb->dc_event);
1928         osb->dc_work_sequence = 0;
1929         osb->dc_wake_sequence = 0;
1930         INIT_LIST_HEAD(&osb->blocked_lock_list);
1931         osb->blocked_lock_count = 0;
1932         spin_lock_init(&osb->osb_lock);
1933         spin_lock_init(&osb->osb_xattr_lock);
1934         ocfs2_init_inode_steal_slot(osb);
1935
1936         atomic_set(&osb->alloc_stats.moves, 0);
1937         atomic_set(&osb->alloc_stats.local_data, 0);
1938         atomic_set(&osb->alloc_stats.bitmap_data, 0);
1939         atomic_set(&osb->alloc_stats.bg_allocs, 0);
1940         atomic_set(&osb->alloc_stats.bg_extends, 0);
1941
1942         ocfs2_init_node_maps(osb);
1943
1944         snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
1945                  MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
1946
1947         status = ocfs2_recovery_init(osb);
1948         if (status) {
1949                 mlog(ML_ERROR, "Unable to initialize recovery state\n");
1950                 mlog_errno(status);
1951                 goto bail;
1952         }
1953
1954         init_waitqueue_head(&osb->checkpoint_event);
1955         atomic_set(&osb->needs_checkpoint, 0);
1956
1957         osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1958
1959         osb->slot_num = OCFS2_INVALID_SLOT;
1960
1961         osb->s_xattr_inline_size = le16_to_cpu(
1962                                         di->id2.i_super.s_xattr_inline_size);
1963
1964         osb->local_alloc_state = OCFS2_LA_UNUSED;
1965         osb->local_alloc_bh = NULL;
1966         INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
1967
1968         init_waitqueue_head(&osb->osb_mount_event);
1969
1970         osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
1971         if (!osb->vol_label) {
1972                 mlog(ML_ERROR, "unable to alloc vol label\n");
1973                 status = -ENOMEM;
1974                 goto bail;
1975         }
1976
1977         osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
1978         if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
1979                 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
1980                      osb->max_slots);
1981                 status = -EINVAL;
1982                 goto bail;
1983         }
1984         mlog(0, "max_slots for this device: %u\n", osb->max_slots);
1985
1986         osb->slot_recovery_generations =
1987                 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
1988                         GFP_KERNEL);
1989         if (!osb->slot_recovery_generations) {
1990                 status = -ENOMEM;
1991                 mlog_errno(status);
1992                 goto bail;
1993         }
1994
1995         init_waitqueue_head(&osb->osb_wipe_event);
1996         osb->osb_orphan_wipes = kcalloc(osb->max_slots,
1997                                         sizeof(*osb->osb_orphan_wipes),
1998                                         GFP_KERNEL);
1999         if (!osb->osb_orphan_wipes) {
2000                 status = -ENOMEM;
2001                 mlog_errno(status);
2002                 goto bail;
2003         }
2004
2005         osb->s_feature_compat =
2006                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2007         osb->s_feature_ro_compat =
2008                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2009         osb->s_feature_incompat =
2010                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2011
2012         if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2013                 mlog(ML_ERROR, "couldn't mount because of unsupported "
2014                      "optional features (%x).\n", i);
2015                 status = -EINVAL;
2016                 goto bail;
2017         }
2018         if (!(osb->sb->s_flags & MS_RDONLY) &&
2019             (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2020                 mlog(ML_ERROR, "couldn't mount RDWR because of "
2021                      "unsupported optional features (%x).\n", i);
2022                 status = -EINVAL;
2023                 goto bail;
2024         }
2025
2026         if (ocfs2_userspace_stack(osb)) {
2027                 memcpy(osb->osb_cluster_stack,
2028                        OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2029                        OCFS2_STACK_LABEL_LEN);
2030                 osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
2031                 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2032                         mlog(ML_ERROR,
2033                              "couldn't mount because of an invalid "
2034                              "cluster stack label (%s) \n",
2035                              osb->osb_cluster_stack);
2036                         status = -EINVAL;
2037                         goto bail;
2038                 }
2039         } else {
2040                 /* The empty string is identical with classic tools that
2041                  * don't know about s_cluster_info. */
2042                 osb->osb_cluster_stack[0] = '\0';
2043         }
2044
2045         get_random_bytes(&osb->s_next_generation, sizeof(u32));
2046
2047         /* FIXME
2048          * This should be done in ocfs2_journal_init(), but unknown
2049          * ordering issues will cause the filesystem to crash.
2050          * If anyone wants to figure out what part of the code
2051          * refers to osb->journal before ocfs2_journal_init() is run,
2052          * be my guest.
2053          */
2054         /* initialize our journal structure */
2055
2056         journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2057         if (!journal) {
2058                 mlog(ML_ERROR, "unable to alloc journal\n");
2059                 status = -ENOMEM;
2060                 goto bail;
2061         }
2062         osb->journal = journal;
2063         journal->j_osb = osb;
2064
2065         atomic_set(&journal->j_num_trans, 0);
2066         init_rwsem(&journal->j_trans_barrier);
2067         init_waitqueue_head(&journal->j_checkpointed);
2068         spin_lock_init(&journal->j_lock);
2069         journal->j_trans_id = (unsigned long) 1;
2070         INIT_LIST_HEAD(&journal->j_la_cleanups);
2071         INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2072         journal->j_state = OCFS2_JOURNAL_FREE;
2073
2074         INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes);
2075         osb->dentry_lock_list = NULL;
2076
2077         /* get some pseudo constants for clustersize bits */
2078         osb->s_clustersize_bits =
2079                 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2080         osb->s_clustersize = 1 << osb->s_clustersize_bits;
2081         mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
2082
2083         if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2084             osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2085                 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2086                      osb->s_clustersize);
2087                 status = -EINVAL;
2088                 goto bail;
2089         }
2090
2091         if (ocfs2_clusters_to_blocks(osb->sb, le32_to_cpu(di->i_clusters) - 1)
2092             > (u32)~0UL) {
2093                 mlog(ML_ERROR, "Volume might try to write to blocks beyond "
2094                      "what jbd can address in 32 bits.\n");
2095                 status = -EINVAL;
2096                 goto bail;
2097         }
2098
2099         if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2100                                  sizeof(di->id2.i_super.s_uuid))) {
2101                 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2102                 status = -ENOMEM;
2103                 goto bail;
2104         }
2105
2106         memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
2107
2108         strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2109         osb->vol_label[63] = '\0';
2110         osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2111         osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2112         osb->first_cluster_group_blkno =
2113                 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2114         osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2115         osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2116         mlog(0, "vol_label: %s\n", osb->vol_label);
2117         mlog(0, "uuid: %s\n", osb->uuid_str);
2118         mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
2119              (unsigned long long)osb->root_blkno,
2120              (unsigned long long)osb->system_dir_blkno);
2121
2122         osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2123         if (!osb->osb_dlm_debug) {
2124                 status = -ENOMEM;
2125                 mlog_errno(status);
2126                 goto bail;
2127         }
2128
2129         atomic_set(&osb->vol_state, VOLUME_INIT);
2130
2131         /* load root, system_dir, and all global system inodes */
2132         status = ocfs2_init_global_system_inodes(osb);
2133         if (status < 0) {
2134                 mlog_errno(status);
2135                 goto bail;
2136         }
2137
2138         /*
2139          * global bitmap
2140          */
2141         inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2142                                             OCFS2_INVALID_SLOT);
2143         if (!inode) {
2144                 status = -EINVAL;
2145                 mlog_errno(status);
2146                 goto bail;
2147         }
2148
2149         osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2150         iput(inode);
2151
2152         osb->bitmap_cpg = ocfs2_group_bitmap_size(sb) * 8;
2153
2154         status = ocfs2_init_slot_info(osb);
2155         if (status < 0) {
2156                 mlog_errno(status);
2157                 goto bail;
2158         }
2159
2160 bail:
2161         mlog_exit(status);
2162         return status;
2163 }
2164
2165 /*
2166  * will return: -EAGAIN if it is ok to keep searching for superblocks
2167  *              -EINVAL if there is a bad superblock
2168  *              0 on success
2169  */
2170 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2171                                struct buffer_head *bh,
2172                                u32 blksz)
2173 {
2174         int status = -EAGAIN;
2175
2176         mlog_entry_void();
2177
2178         if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2179                    strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2180                 /* We have to do a raw check of the feature here */
2181                 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2182                     OCFS2_FEATURE_INCOMPAT_META_ECC) {
2183                         status = ocfs2_block_check_validate(bh->b_data,
2184                                                             bh->b_size,
2185                                                             &di->i_check);
2186                         if (status)
2187                                 goto out;
2188                 }
2189                 status = -EINVAL;
2190                 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2191                         mlog(ML_ERROR, "found superblock with incorrect block "
2192                              "size: found %u, should be %u\n",
2193                              1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2194                                blksz);
2195                 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2196                            OCFS2_MAJOR_REV_LEVEL ||
2197                            le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2198                            OCFS2_MINOR_REV_LEVEL) {
2199                         mlog(ML_ERROR, "found superblock with bad version: "
2200                              "found %u.%u, should be %u.%u\n",
2201                              le16_to_cpu(di->id2.i_super.s_major_rev_level),
2202                              le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2203                              OCFS2_MAJOR_REV_LEVEL,
2204                              OCFS2_MINOR_REV_LEVEL);
2205                 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2206                         mlog(ML_ERROR, "bad block number on superblock: "
2207                              "found %llu, should be %llu\n",
2208                              (unsigned long long)le64_to_cpu(di->i_blkno),
2209                              (unsigned long long)bh->b_blocknr);
2210                 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2211                             le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2212                         mlog(ML_ERROR, "bad cluster size found: %u\n",
2213                              1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2214                 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2215                         mlog(ML_ERROR, "bad root_blkno: 0\n");
2216                 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2217                         mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2218                 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2219                         mlog(ML_ERROR,
2220                              "Superblock slots found greater than file system "
2221                              "maximum: found %u, max %u\n",
2222                              le16_to_cpu(di->id2.i_super.s_max_slots),
2223                              OCFS2_MAX_SLOTS);
2224                 } else {
2225                         /* found it! */
2226                         status = 0;
2227                 }
2228         }
2229
2230 out:
2231         mlog_exit(status);
2232         return status;
2233 }
2234
2235 static int ocfs2_check_volume(struct ocfs2_super *osb)
2236 {
2237         int status;
2238         int dirty;
2239         int local;
2240         struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2241                                                   * recover
2242                                                   * ourselves. */
2243
2244         mlog_entry_void();
2245
2246         /* Init our journal object. */
2247         status = ocfs2_journal_init(osb->journal, &dirty);
2248         if (status < 0) {
2249                 mlog(ML_ERROR, "Could not initialize journal!\n");
2250                 goto finally;
2251         }
2252
2253         /* If the journal was unmounted cleanly then we don't want to
2254          * recover anything. Otherwise, journal_load will do that
2255          * dirty work for us :) */
2256         if (!dirty) {
2257                 status = ocfs2_journal_wipe(osb->journal, 0);
2258                 if (status < 0) {
2259                         mlog_errno(status);
2260                         goto finally;
2261                 }
2262         } else {
2263                 mlog(ML_NOTICE, "File system was not unmounted cleanly, "
2264                      "recovering volume.\n");
2265         }
2266
2267         local = ocfs2_mount_local(osb);
2268
2269         /* will play back anything left in the journal. */
2270         status = ocfs2_journal_load(osb->journal, local, dirty);
2271         if (status < 0) {
2272                 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2273                 goto finally;
2274         }
2275
2276         if (dirty) {
2277                 /* recover my local alloc if we didn't unmount cleanly. */
2278                 status = ocfs2_begin_local_alloc_recovery(osb,
2279                                                           osb->slot_num,
2280                                                           &local_alloc);
2281                 if (status < 0) {
2282                         mlog_errno(status);
2283                         goto finally;
2284                 }
2285                 /* we complete the recovery process after we've marked
2286                  * ourselves as mounted. */
2287         }
2288
2289         mlog(0, "Journal loaded.\n");
2290
2291         status = ocfs2_load_local_alloc(osb);
2292         if (status < 0) {
2293                 mlog_errno(status);
2294                 goto finally;
2295         }
2296
2297         if (dirty) {
2298                 /* Recovery will be completed after we've mounted the
2299                  * rest of the volume. */
2300                 osb->dirty = 1;
2301                 osb->local_alloc_copy = local_alloc;
2302                 local_alloc = NULL;
2303         }
2304
2305         /* go through each journal, trylock it and if you get the
2306          * lock, and it's marked as dirty, set the bit in the recover
2307          * map and launch a recovery thread for it. */
2308         status = ocfs2_mark_dead_nodes(osb);
2309         if (status < 0) {
2310                 mlog_errno(status);
2311                 goto finally;
2312         }
2313
2314         status = ocfs2_compute_replay_slots(osb);
2315         if (status < 0)
2316                 mlog_errno(status);
2317
2318 finally:
2319         if (local_alloc)
2320                 kfree(local_alloc);
2321
2322         mlog_exit(status);
2323         return status;
2324 }
2325
2326 /*
2327  * The routine gets called from dismount or close whenever a dismount on
2328  * volume is requested and the osb open count becomes 1.
2329  * It will remove the osb from the global list and also free up all the
2330  * initialized resources and fileobject.
2331  */
2332 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2333 {
2334         mlog_entry_void();
2335
2336         /* This function assumes that the caller has the main osb resource */
2337
2338         ocfs2_free_slot_info(osb);
2339
2340         kfree(osb->osb_orphan_wipes);
2341         kfree(osb->slot_recovery_generations);
2342         /* FIXME
2343          * This belongs in journal shutdown, but because we have to
2344          * allocate osb->journal at the start of ocfs2_initalize_osb(),
2345          * we free it here.
2346          */
2347         kfree(osb->journal);
2348         if (osb->local_alloc_copy)
2349                 kfree(osb->local_alloc_copy);
2350         kfree(osb->uuid_str);
2351         ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2352         memset(osb, 0, sizeof(struct ocfs2_super));
2353
2354         mlog_exit_void();
2355 }
2356
2357 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2358  * panic(). We do not support continue-on-error operation. */
2359 static void ocfs2_handle_error(struct super_block *sb)
2360 {
2361         struct ocfs2_super *osb = OCFS2_SB(sb);
2362
2363         if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2364                 panic("OCFS2: (device %s): panic forced after error\n",
2365                       sb->s_id);
2366
2367         ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2368
2369         if (sb->s_flags & MS_RDONLY &&
2370             (ocfs2_is_soft_readonly(osb) ||
2371              ocfs2_is_hard_readonly(osb)))
2372                 return;
2373
2374         printk(KERN_CRIT "File system is now read-only due to the potential "
2375                "of on-disk corruption. Please run fsck.ocfs2 once the file "
2376                "system is unmounted.\n");
2377         sb->s_flags |= MS_RDONLY;
2378         ocfs2_set_ro_flag(osb, 0);
2379 }
2380
2381 static char error_buf[1024];
2382
2383 void __ocfs2_error(struct super_block *sb,
2384                    const char *function,
2385                    const char *fmt, ...)
2386 {
2387         va_list args;
2388
2389         va_start(args, fmt);
2390         vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2391         va_end(args);
2392
2393         /* Not using mlog here because we want to show the actual
2394          * function the error came from. */
2395         printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2396                sb->s_id, function, error_buf);
2397
2398         ocfs2_handle_error(sb);
2399 }
2400
2401 /* Handle critical errors. This is intentionally more drastic than
2402  * ocfs2_handle_error, so we only use for things like journal errors,
2403  * etc. */
2404 void __ocfs2_abort(struct super_block* sb,
2405                    const char *function,
2406                    const char *fmt, ...)
2407 {
2408         va_list args;
2409
2410         va_start(args, fmt);
2411         vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2412         va_end(args);
2413
2414         printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2415                sb->s_id, function, error_buf);
2416
2417         /* We don't have the cluster support yet to go straight to
2418          * hard readonly in here. Until then, we want to keep
2419          * ocfs2_abort() so that we can at least mark critical
2420          * errors.
2421          *
2422          * TODO: This should abort the journal and alert other nodes
2423          * that our slot needs recovery. */
2424
2425         /* Force a panic(). This stinks, but it's better than letting
2426          * things continue without having a proper hard readonly
2427          * here. */
2428         OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2429         ocfs2_handle_error(sb);
2430 }
2431
2432 module_init(ocfs2_init);
2433 module_exit(ocfs2_exit);