Merge branch 'drivers-platform' into release
[linux-2.6] / 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
46 #define MLOG_MASK_PREFIX ML_SUPER
47 #include <cluster/masklog.h>
48
49 #include "ocfs2.h"
50
51 /* this should be the only file to include a version 1 header */
52 #include "ocfs1_fs_compat.h"
53
54 #include "alloc.h"
55 #include "blockcheck.h"
56 #include "dlmglue.h"
57 #include "export.h"
58 #include "extent_map.h"
59 #include "heartbeat.h"
60 #include "inode.h"
61 #include "journal.h"
62 #include "localalloc.h"
63 #include "namei.h"
64 #include "slot_map.h"
65 #include "super.h"
66 #include "sysfile.h"
67 #include "uptodate.h"
68 #include "ver.h"
69 #include "xattr.h"
70 #include "quota.h"
71
72 #include "buffer_head_io.h"
73
74 static struct kmem_cache *ocfs2_inode_cachep = NULL;
75 struct kmem_cache *ocfs2_dquot_cachep;
76 struct kmem_cache *ocfs2_qf_chunk_cachep;
77
78 /* OCFS2 needs to schedule several differnt types of work which
79  * require cluster locking, disk I/O, recovery waits, etc. Since these
80  * types of work tend to be heavy we avoid using the kernel events
81  * workqueue and schedule on our own. */
82 struct workqueue_struct *ocfs2_wq = NULL;
83
84 static struct dentry *ocfs2_debugfs_root = NULL;
85
86 MODULE_AUTHOR("Oracle");
87 MODULE_LICENSE("GPL");
88
89 struct mount_options
90 {
91         unsigned long   commit_interval;
92         unsigned long   mount_opt;
93         unsigned int    atime_quantum;
94         signed short    slot;
95         unsigned int    localalloc_opt;
96         char            cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
97 };
98
99 static int ocfs2_parse_options(struct super_block *sb, char *options,
100                                struct mount_options *mopt,
101                                int is_remount);
102 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt);
103 static void ocfs2_put_super(struct super_block *sb);
104 static int ocfs2_mount_volume(struct super_block *sb);
105 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
106 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
107 static int ocfs2_initialize_mem_caches(void);
108 static void ocfs2_free_mem_caches(void);
109 static void ocfs2_delete_osb(struct ocfs2_super *osb);
110
111 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
112
113 static int ocfs2_sync_fs(struct super_block *sb, int wait);
114
115 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
116 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
117 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
118 static int ocfs2_check_volume(struct ocfs2_super *osb);
119 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
120                                struct buffer_head *bh,
121                                u32 sectsize);
122 static int ocfs2_initialize_super(struct super_block *sb,
123                                   struct buffer_head *bh,
124                                   int sector_size);
125 static int ocfs2_get_sector(struct super_block *sb,
126                             struct buffer_head **bh,
127                             int block,
128                             int sect_size);
129 static void ocfs2_write_super(struct super_block *sb);
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         .write_super    = ocfs2_write_super,
145         .put_super      = ocfs2_put_super,
146         .remount_fs     = ocfs2_remount,
147         .show_options   = ocfs2_show_options,
148         .quota_read     = ocfs2_quota_read,
149         .quota_write    = ocfs2_quota_write,
150 };
151
152 enum {
153         Opt_barrier,
154         Opt_err_panic,
155         Opt_err_ro,
156         Opt_intr,
157         Opt_nointr,
158         Opt_hb_none,
159         Opt_hb_local,
160         Opt_data_ordered,
161         Opt_data_writeback,
162         Opt_atime_quantum,
163         Opt_slot,
164         Opt_commit,
165         Opt_localalloc,
166         Opt_localflocks,
167         Opt_stack,
168         Opt_user_xattr,
169         Opt_nouser_xattr,
170         Opt_inode64,
171         Opt_acl,
172         Opt_noacl,
173         Opt_usrquota,
174         Opt_grpquota,
175         Opt_err,
176 };
177
178 static const match_table_t tokens = {
179         {Opt_barrier, "barrier=%u"},
180         {Opt_err_panic, "errors=panic"},
181         {Opt_err_ro, "errors=remount-ro"},
182         {Opt_intr, "intr"},
183         {Opt_nointr, "nointr"},
184         {Opt_hb_none, OCFS2_HB_NONE},
185         {Opt_hb_local, OCFS2_HB_LOCAL},
186         {Opt_data_ordered, "data=ordered"},
187         {Opt_data_writeback, "data=writeback"},
188         {Opt_atime_quantum, "atime_quantum=%u"},
189         {Opt_slot, "preferred_slot=%u"},
190         {Opt_commit, "commit=%u"},
191         {Opt_localalloc, "localalloc=%d"},
192         {Opt_localflocks, "localflocks"},
193         {Opt_stack, "cluster_stack=%s"},
194         {Opt_user_xattr, "user_xattr"},
195         {Opt_nouser_xattr, "nouser_xattr"},
196         {Opt_inode64, "inode64"},
197         {Opt_acl, "acl"},
198         {Opt_noacl, "noacl"},
199         {Opt_usrquota, "usrquota"},
200         {Opt_grpquota, "grpquota"},
201         {Opt_err, NULL}
202 };
203
204 /*
205  * write_super and sync_fs ripped right out of ext3.
206  */
207 static void ocfs2_write_super(struct super_block *sb)
208 {
209         if (mutex_trylock(&sb->s_lock) != 0)
210                 BUG();
211         sb->s_dirt = 0;
212 }
213
214 static int ocfs2_sync_fs(struct super_block *sb, int wait)
215 {
216         int status;
217         tid_t target;
218         struct ocfs2_super *osb = OCFS2_SB(sb);
219
220         sb->s_dirt = 0;
221
222         if (ocfs2_is_hard_readonly(osb))
223                 return -EROFS;
224
225         if (wait) {
226                 status = ocfs2_flush_truncate_log(osb);
227                 if (status < 0)
228                         mlog_errno(status);
229         } else {
230                 ocfs2_schedule_truncate_log_flush(osb, 0);
231         }
232
233         if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
234                                       &target)) {
235                 if (wait)
236                         jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
237                                              target);
238         }
239         return 0;
240 }
241
242 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
243 {
244         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
245             && (ino == USER_QUOTA_SYSTEM_INODE
246                 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
247                 return 0;
248         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
249             && (ino == GROUP_QUOTA_SYSTEM_INODE
250                 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
251                 return 0;
252         return 1;
253 }
254
255 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
256 {
257         struct inode *new = NULL;
258         int status = 0;
259         int i;
260
261         mlog_entry_void();
262
263         new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
264         if (IS_ERR(new)) {
265                 status = PTR_ERR(new);
266                 mlog_errno(status);
267                 goto bail;
268         }
269         osb->root_inode = new;
270
271         new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
272         if (IS_ERR(new)) {
273                 status = PTR_ERR(new);
274                 mlog_errno(status);
275                 goto bail;
276         }
277         osb->sys_root_inode = new;
278
279         for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
280              i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
281                 if (!ocfs2_need_system_inode(osb, i))
282                         continue;
283                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
284                 if (!new) {
285                         ocfs2_release_system_inodes(osb);
286                         status = -EINVAL;
287                         mlog_errno(status);
288                         /* FIXME: Should ERROR_RO_FS */
289                         mlog(ML_ERROR, "Unable to load system inode %d, "
290                              "possibly corrupt fs?", i);
291                         goto bail;
292                 }
293                 // the array now has one ref, so drop this one
294                 iput(new);
295         }
296
297 bail:
298         mlog_exit(status);
299         return status;
300 }
301
302 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
303 {
304         struct inode *new = NULL;
305         int status = 0;
306         int i;
307
308         mlog_entry_void();
309
310         for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
311              i < NUM_SYSTEM_INODES;
312              i++) {
313                 if (!ocfs2_need_system_inode(osb, i))
314                         continue;
315                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
316                 if (!new) {
317                         ocfs2_release_system_inodes(osb);
318                         status = -EINVAL;
319                         mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
320                              status, i, osb->slot_num);
321                         goto bail;
322                 }
323                 /* the array now has one ref, so drop this one */
324                 iput(new);
325         }
326
327 bail:
328         mlog_exit(status);
329         return status;
330 }
331
332 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
333 {
334         int i;
335         struct inode *inode;
336
337         mlog_entry_void();
338
339         for (i = 0; i < NUM_SYSTEM_INODES; i++) {
340                 inode = osb->system_inodes[i];
341                 if (inode) {
342                         iput(inode);
343                         osb->system_inodes[i] = NULL;
344                 }
345         }
346
347         inode = osb->sys_root_inode;
348         if (inode) {
349                 iput(inode);
350                 osb->sys_root_inode = NULL;
351         }
352
353         inode = osb->root_inode;
354         if (inode) {
355                 iput(inode);
356                 osb->root_inode = NULL;
357         }
358
359         mlog_exit(0);
360 }
361
362 /* We're allocating fs objects, use GFP_NOFS */
363 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
364 {
365         struct ocfs2_inode_info *oi;
366
367         oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
368         if (!oi)
369                 return NULL;
370
371         jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
372         return &oi->vfs_inode;
373 }
374
375 static void ocfs2_destroy_inode(struct inode *inode)
376 {
377         kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
378 }
379
380 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
381                                                 unsigned int cbits)
382 {
383         unsigned int bytes = 1 << cbits;
384         unsigned int trim = bytes;
385         unsigned int bitshift = 32;
386
387         /*
388          * i_size and all block offsets in ocfs2 are always 64 bits
389          * wide. i_clusters is 32 bits, in cluster-sized units. So on
390          * 64 bit platforms, cluster size will be the limiting factor.
391          */
392
393 #if BITS_PER_LONG == 32
394 # if defined(CONFIG_LBD)
395         BUILD_BUG_ON(sizeof(sector_t) != 8);
396         /*
397          * We might be limited by page cache size.
398          */
399         if (bytes > PAGE_CACHE_SIZE) {
400                 bytes = PAGE_CACHE_SIZE;
401                 trim = 1;
402                 /*
403                  * Shift by 31 here so that we don't get larger than
404                  * MAX_LFS_FILESIZE
405                  */
406                 bitshift = 31;
407         }
408 # else
409         /*
410          * We are limited by the size of sector_t. Use block size, as
411          * that's what we expose to the VFS.
412          */
413         bytes = 1 << bbits;
414         trim = 1;
415         bitshift = 31;
416 # endif
417 #endif
418
419         /*
420          * Trim by a whole cluster when we can actually approach the
421          * on-disk limits. Otherwise we can overflow i_clusters when
422          * an extent start is at the max offset.
423          */
424         return (((unsigned long long)bytes) << bitshift) - trim;
425 }
426
427 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
428 {
429         int incompat_features;
430         int ret = 0;
431         struct mount_options parsed_options;
432         struct ocfs2_super *osb = OCFS2_SB(sb);
433
434         if (!ocfs2_parse_options(sb, data, &parsed_options, 1)) {
435                 ret = -EINVAL;
436                 goto out;
437         }
438
439         if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) !=
440             (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
441                 ret = -EINVAL;
442                 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
443                 goto out;
444         }
445
446         if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
447             (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
448                 ret = -EINVAL;
449                 mlog(ML_ERROR, "Cannot change data mode on remount\n");
450                 goto out;
451         }
452
453         /* Probably don't want this on remount; it might
454          * mess with other nodes */
455         if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
456             (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
457                 ret = -EINVAL;
458                 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
459                 goto out;
460         }
461
462         /* We're going to/from readonly mode. */
463         if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
464                 /* Disable quota accounting before remounting RO */
465                 if (*flags & MS_RDONLY) {
466                         ret = ocfs2_susp_quotas(osb, 0);
467                         if (ret < 0)
468                                 goto out;
469                 }
470                 /* Lock here so the check of HARD_RO and the potential
471                  * setting of SOFT_RO is atomic. */
472                 spin_lock(&osb->osb_lock);
473                 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
474                         mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
475                         ret = -EROFS;
476                         goto unlock_osb;
477                 }
478
479                 if (*flags & MS_RDONLY) {
480                         mlog(0, "Going to ro mode.\n");
481                         sb->s_flags |= MS_RDONLY;
482                         osb->osb_flags |= OCFS2_OSB_SOFT_RO;
483                 } else {
484                         mlog(0, "Making ro filesystem writeable.\n");
485
486                         if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
487                                 mlog(ML_ERROR, "Cannot remount RDWR "
488                                      "filesystem due to previous errors.\n");
489                                 ret = -EROFS;
490                                 goto unlock_osb;
491                         }
492                         incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
493                         if (incompat_features) {
494                                 mlog(ML_ERROR, "Cannot remount RDWR because "
495                                      "of unsupported optional features "
496                                      "(%x).\n", incompat_features);
497                                 ret = -EINVAL;
498                                 goto unlock_osb;
499                         }
500                         sb->s_flags &= ~MS_RDONLY;
501                         osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
502                 }
503 unlock_osb:
504                 spin_unlock(&osb->osb_lock);
505                 /* Enable quota accounting after remounting RW */
506                 if (!ret && !(*flags & MS_RDONLY)) {
507                         if (sb_any_quota_suspended(sb))
508                                 ret = ocfs2_susp_quotas(osb, 1);
509                         else
510                                 ret = ocfs2_enable_quotas(osb);
511                         if (ret < 0) {
512                                 /* Return back changes... */
513                                 spin_lock(&osb->osb_lock);
514                                 sb->s_flags |= MS_RDONLY;
515                                 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
516                                 spin_unlock(&osb->osb_lock);
517                                 goto out;
518                         }
519                 }
520         }
521
522         if (!ret) {
523                 /* Only save off the new mount options in case of a successful
524                  * remount. */
525                 if (!(osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_XATTR))
526                         parsed_options.mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
527                 osb->s_mount_opt = parsed_options.mount_opt;
528                 osb->s_atime_quantum = parsed_options.atime_quantum;
529                 osb->preferred_slot = parsed_options.slot;
530                 if (parsed_options.commit_interval)
531                         osb->osb_commit_interval = parsed_options.commit_interval;
532
533                 if (!ocfs2_is_hard_readonly(osb))
534                         ocfs2_set_journal_params(osb);
535         }
536 out:
537         return ret;
538 }
539
540 static int ocfs2_sb_probe(struct super_block *sb,
541                           struct buffer_head **bh,
542                           int *sector_size)
543 {
544         int status, tmpstat;
545         struct ocfs1_vol_disk_hdr *hdr;
546         struct ocfs2_dinode *di;
547         int blksize;
548
549         *bh = NULL;
550
551         /* may be > 512 */
552         *sector_size = bdev_hardsect_size(sb->s_bdev);
553         if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
554                 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
555                      *sector_size, OCFS2_MAX_BLOCKSIZE);
556                 status = -EINVAL;
557                 goto bail;
558         }
559
560         /* Can this really happen? */
561         if (*sector_size < OCFS2_MIN_BLOCKSIZE)
562                 *sector_size = OCFS2_MIN_BLOCKSIZE;
563
564         /* check block zero for old format */
565         status = ocfs2_get_sector(sb, bh, 0, *sector_size);
566         if (status < 0) {
567                 mlog_errno(status);
568                 goto bail;
569         }
570         hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
571         if (hdr->major_version == OCFS1_MAJOR_VERSION) {
572                 mlog(ML_ERROR, "incompatible version: %u.%u\n",
573                      hdr->major_version, hdr->minor_version);
574                 status = -EINVAL;
575         }
576         if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
577                    strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
578                 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
579                      hdr->signature);
580                 status = -EINVAL;
581         }
582         brelse(*bh);
583         *bh = NULL;
584         if (status < 0) {
585                 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
586                      "upgraded before mounting with ocfs v2\n");
587                 goto bail;
588         }
589
590         /*
591          * Now check at magic offset for 512, 1024, 2048, 4096
592          * blocksizes.  4096 is the maximum blocksize because it is
593          * the minimum clustersize.
594          */
595         status = -EINVAL;
596         for (blksize = *sector_size;
597              blksize <= OCFS2_MAX_BLOCKSIZE;
598              blksize <<= 1) {
599                 tmpstat = ocfs2_get_sector(sb, bh,
600                                            OCFS2_SUPER_BLOCK_BLKNO,
601                                            blksize);
602                 if (tmpstat < 0) {
603                         status = tmpstat;
604                         mlog_errno(status);
605                         goto bail;
606                 }
607                 di = (struct ocfs2_dinode *) (*bh)->b_data;
608                 status = ocfs2_verify_volume(di, *bh, blksize);
609                 if (status >= 0)
610                         goto bail;
611                 brelse(*bh);
612                 *bh = NULL;
613                 if (status != -EAGAIN)
614                         break;
615         }
616
617 bail:
618         return status;
619 }
620
621 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
622 {
623         if (ocfs2_mount_local(osb)) {
624                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
625                         mlog(ML_ERROR, "Cannot heartbeat on a locally "
626                              "mounted device.\n");
627                         return -EINVAL;
628                 }
629         }
630
631         if (ocfs2_userspace_stack(osb)) {
632                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
633                         mlog(ML_ERROR, "Userspace stack expected, but "
634                              "o2cb heartbeat arguments passed to mount\n");
635                         return -EINVAL;
636                 }
637         }
638
639         if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
640                 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
641                     !ocfs2_userspace_stack(osb)) {
642                         mlog(ML_ERROR, "Heartbeat has to be started to mount "
643                              "a read-write clustered device.\n");
644                         return -EINVAL;
645                 }
646         }
647
648         return 0;
649 }
650
651 /*
652  * If we're using a userspace stack, mount should have passed
653  * a name that matches the disk.  If not, mount should not
654  * have passed a stack.
655  */
656 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
657                                         struct mount_options *mopt)
658 {
659         if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
660                 mlog(ML_ERROR,
661                      "cluster stack passed to mount, but this filesystem "
662                      "does not support it\n");
663                 return -EINVAL;
664         }
665
666         if (ocfs2_userspace_stack(osb) &&
667             strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
668                     OCFS2_STACK_LABEL_LEN)) {
669                 mlog(ML_ERROR,
670                      "cluster stack passed to mount (\"%s\") does not "
671                      "match the filesystem (\"%s\")\n",
672                      mopt->cluster_stack,
673                      osb->osb_cluster_stack);
674                 return -EINVAL;
675         }
676
677         return 0;
678 }
679
680 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
681 {
682         int type;
683         struct super_block *sb = osb->sb;
684         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
685                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
686         int status = 0;
687
688         for (type = 0; type < MAXQUOTAS; type++) {
689                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
690                         continue;
691                 if (unsuspend)
692                         status = vfs_quota_enable(
693                                         sb_dqopt(sb)->files[type],
694                                         type, QFMT_OCFS2,
695                                         DQUOT_SUSPENDED);
696                 else
697                         status = vfs_quota_disable(sb, type,
698                                                    DQUOT_SUSPENDED);
699                 if (status < 0)
700                         break;
701         }
702         if (status < 0)
703                 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
704                      "remount (error = %d).\n", status);
705         return status;
706 }
707
708 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
709 {
710         struct inode *inode[MAXQUOTAS] = { NULL, NULL };
711         struct super_block *sb = osb->sb;
712         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
713                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
714         unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
715                                         LOCAL_GROUP_QUOTA_SYSTEM_INODE };
716         int status;
717         int type;
718
719         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
720         for (type = 0; type < MAXQUOTAS; type++) {
721                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
722                         continue;
723                 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
724                                                         osb->slot_num);
725                 if (!inode[type]) {
726                         status = -ENOENT;
727                         goto out_quota_off;
728                 }
729                 status = vfs_quota_enable(inode[type], type, QFMT_OCFS2,
730                                                 DQUOT_USAGE_ENABLED);
731                 if (status < 0)
732                         goto out_quota_off;
733         }
734
735         for (type = 0; type < MAXQUOTAS; type++)
736                 iput(inode[type]);
737         return 0;
738 out_quota_off:
739         ocfs2_disable_quotas(osb);
740         for (type = 0; type < MAXQUOTAS; type++)
741                 iput(inode[type]);
742         mlog_errno(status);
743         return status;
744 }
745
746 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
747 {
748         int type;
749         struct inode *inode;
750         struct super_block *sb = osb->sb;
751
752         /* We mostly ignore errors in this function because there's not much
753          * we can do when we see them */
754         for (type = 0; type < MAXQUOTAS; type++) {
755                 if (!sb_has_quota_loaded(sb, type))
756                         continue;
757                 inode = igrab(sb->s_dquot.files[type]);
758                 /* Turn off quotas. This will remove all dquot structures from
759                  * memory and so they will be automatically synced to global
760                  * quota files */
761                 vfs_quota_disable(sb, type, DQUOT_USAGE_ENABLED |
762                                             DQUOT_LIMITS_ENABLED);
763                 if (!inode)
764                         continue;
765                 iput(inode);
766         }
767 }
768
769 /* Handle quota on quotactl */
770 static int ocfs2_quota_on(struct super_block *sb, int type, int format_id,
771                           char *path, int remount)
772 {
773         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
774                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
775
776         if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
777                 return -EINVAL;
778
779         if (remount)
780                 return 0;       /* Just ignore it has been handled in
781                                  * ocfs2_remount() */
782         return vfs_quota_enable(sb_dqopt(sb)->files[type], type,
783                                     format_id, DQUOT_LIMITS_ENABLED);
784 }
785
786 /* Handle quota off quotactl */
787 static int ocfs2_quota_off(struct super_block *sb, int type, int remount)
788 {
789         if (remount)
790                 return 0;       /* Ignore now and handle later in
791                                  * ocfs2_remount() */
792         return vfs_quota_disable(sb, type, DQUOT_LIMITS_ENABLED);
793 }
794
795 static struct quotactl_ops ocfs2_quotactl_ops = {
796         .quota_on       = ocfs2_quota_on,
797         .quota_off      = ocfs2_quota_off,
798         .quota_sync     = vfs_quota_sync,
799         .get_info       = vfs_get_dqinfo,
800         .set_info       = vfs_set_dqinfo,
801         .get_dqblk      = vfs_get_dqblk,
802         .set_dqblk      = vfs_set_dqblk,
803 };
804
805 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
806 {
807         struct dentry *root;
808         int status, sector_size;
809         struct mount_options parsed_options;
810         struct inode *inode = NULL;
811         struct ocfs2_super *osb = NULL;
812         struct buffer_head *bh = NULL;
813         char nodestr[8];
814
815         mlog_entry("%p, %p, %i", sb, data, silent);
816
817         if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
818                 status = -EINVAL;
819                 goto read_super_error;
820         }
821
822         /* probe for superblock */
823         status = ocfs2_sb_probe(sb, &bh, &sector_size);
824         if (status < 0) {
825                 mlog(ML_ERROR, "superblock probe failed!\n");
826                 goto read_super_error;
827         }
828
829         status = ocfs2_initialize_super(sb, bh, sector_size);
830         osb = OCFS2_SB(sb);
831         if (status < 0) {
832                 mlog_errno(status);
833                 goto read_super_error;
834         }
835         brelse(bh);
836         bh = NULL;
837
838         if (!(osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_XATTR))
839                 parsed_options.mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
840
841         osb->s_mount_opt = parsed_options.mount_opt;
842         osb->s_atime_quantum = parsed_options.atime_quantum;
843         osb->preferred_slot = parsed_options.slot;
844         osb->osb_commit_interval = parsed_options.commit_interval;
845         osb->local_alloc_default_bits = ocfs2_megabytes_to_clusters(sb, parsed_options.localalloc_opt);
846         osb->local_alloc_bits = osb->local_alloc_default_bits;
847         if (osb->s_mount_opt & OCFS2_MOUNT_USRQUOTA &&
848             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
849                                          OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
850                 status = -EINVAL;
851                 mlog(ML_ERROR, "User quotas were requested, but this "
852                      "filesystem does not have the feature enabled.\n");
853                 goto read_super_error;
854         }
855         if (osb->s_mount_opt & OCFS2_MOUNT_GRPQUOTA &&
856             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
857                                          OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
858                 status = -EINVAL;
859                 mlog(ML_ERROR, "Group quotas were requested, but this "
860                      "filesystem does not have the feature enabled.\n");
861                 goto read_super_error;
862         }
863
864         status = ocfs2_verify_userspace_stack(osb, &parsed_options);
865         if (status)
866                 goto read_super_error;
867
868         sb->s_magic = OCFS2_SUPER_MAGIC;
869
870         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
871                 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
872
873         /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
874          * heartbeat=none */
875         if (bdev_read_only(sb->s_bdev)) {
876                 if (!(sb->s_flags & MS_RDONLY)) {
877                         status = -EACCES;
878                         mlog(ML_ERROR, "Readonly device detected but readonly "
879                              "mount was not specified.\n");
880                         goto read_super_error;
881                 }
882
883                 /* You should not be able to start a local heartbeat
884                  * on a readonly device. */
885                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
886                         status = -EROFS;
887                         mlog(ML_ERROR, "Local heartbeat specified on readonly "
888                              "device.\n");
889                         goto read_super_error;
890                 }
891
892                 status = ocfs2_check_journals_nolocks(osb);
893                 if (status < 0) {
894                         if (status == -EROFS)
895                                 mlog(ML_ERROR, "Recovery required on readonly "
896                                      "file system, but write access is "
897                                      "unavailable.\n");
898                         else
899                                 mlog_errno(status);                     
900                         goto read_super_error;
901                 }
902
903                 ocfs2_set_ro_flag(osb, 1);
904
905                 printk(KERN_NOTICE "Readonly device detected. No cluster "
906                        "services will be utilized for this mount. Recovery "
907                        "will be skipped.\n");
908         }
909
910         if (!ocfs2_is_hard_readonly(osb)) {
911                 if (sb->s_flags & MS_RDONLY)
912                         ocfs2_set_ro_flag(osb, 0);
913         }
914
915         status = ocfs2_verify_heartbeat(osb);
916         if (status < 0) {
917                 mlog_errno(status);
918                 goto read_super_error;
919         }
920
921         osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
922                                                  ocfs2_debugfs_root);
923         if (!osb->osb_debug_root) {
924                 status = -EINVAL;
925                 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
926                 goto read_super_error;
927         }
928
929         status = ocfs2_mount_volume(sb);
930         if (osb->root_inode)
931                 inode = igrab(osb->root_inode);
932
933         if (status < 0)
934                 goto read_super_error;
935
936         if (!inode) {
937                 status = -EIO;
938                 mlog_errno(status);
939                 goto read_super_error;
940         }
941
942         root = d_alloc_root(inode);
943         if (!root) {
944                 status = -ENOMEM;
945                 mlog_errno(status);
946                 goto read_super_error;
947         }
948
949         sb->s_root = root;
950
951         ocfs2_complete_mount_recovery(osb);
952
953         if (ocfs2_mount_local(osb))
954                 snprintf(nodestr, sizeof(nodestr), "local");
955         else
956                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
957
958         printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
959                "with %s data mode.\n",
960                osb->dev_str, nodestr, osb->slot_num,
961                osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
962                "ordered");
963
964         atomic_set(&osb->vol_state, VOLUME_MOUNTED);
965         wake_up(&osb->osb_mount_event);
966
967         /* Now we can initialize quotas because we can afford to wait
968          * for cluster locks recovery now. That also means that truncation
969          * log recovery can happen but that waits for proper quota setup */
970         if (!(sb->s_flags & MS_RDONLY)) {
971                 status = ocfs2_enable_quotas(osb);
972                 if (status < 0) {
973                         /* We have to err-out specially here because
974                          * s_root is already set */
975                         mlog_errno(status);
976                         atomic_set(&osb->vol_state, VOLUME_DISABLED);
977                         wake_up(&osb->osb_mount_event);
978                         mlog_exit(status);
979                         return status;
980                 }
981         }
982
983         ocfs2_complete_quota_recovery(osb);
984
985         /* Now we wake up again for processes waiting for quotas */
986         atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
987         wake_up(&osb->osb_mount_event);
988
989         mlog_exit(status);
990         return status;
991
992 read_super_error:
993         brelse(bh);
994
995         if (inode)
996                 iput(inode);
997
998         if (osb) {
999                 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1000                 wake_up(&osb->osb_mount_event);
1001                 ocfs2_dismount_volume(sb, 1);
1002         }
1003
1004         mlog_exit(status);
1005         return status;
1006 }
1007
1008 static int ocfs2_get_sb(struct file_system_type *fs_type,
1009                         int flags,
1010                         const char *dev_name,
1011                         void *data,
1012                         struct vfsmount *mnt)
1013 {
1014         return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super,
1015                            mnt);
1016 }
1017
1018 static struct file_system_type ocfs2_fs_type = {
1019         .owner          = THIS_MODULE,
1020         .name           = "ocfs2",
1021         .get_sb         = ocfs2_get_sb, /* is this called when we mount
1022                                         * the fs? */
1023         .kill_sb        = kill_block_super, /* set to the generic one
1024                                              * right now, but do we
1025                                              * need to change that? */
1026         .fs_flags       = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1027         .next           = NULL
1028 };
1029
1030 static int ocfs2_parse_options(struct super_block *sb,
1031                                char *options,
1032                                struct mount_options *mopt,
1033                                int is_remount)
1034 {
1035         int status;
1036         char *p;
1037
1038         mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
1039                    options ? options : "(none)");
1040
1041         mopt->commit_interval = 0;
1042         mopt->mount_opt = 0;
1043         mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1044         mopt->slot = OCFS2_INVALID_SLOT;
1045         mopt->localalloc_opt = OCFS2_DEFAULT_LOCAL_ALLOC_SIZE;
1046         mopt->cluster_stack[0] = '\0';
1047
1048         if (!options) {
1049                 status = 1;
1050                 goto bail;
1051         }
1052
1053         while ((p = strsep(&options, ",")) != NULL) {
1054                 int token, option;
1055                 substring_t args[MAX_OPT_ARGS];
1056
1057                 if (!*p)
1058                         continue;
1059
1060                 token = match_token(p, tokens, args);
1061                 switch (token) {
1062                 case Opt_hb_local:
1063                         mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1064                         break;
1065                 case Opt_hb_none:
1066                         mopt->mount_opt &= ~OCFS2_MOUNT_HB_LOCAL;
1067                         break;
1068                 case Opt_barrier:
1069                         if (match_int(&args[0], &option)) {
1070                                 status = 0;
1071                                 goto bail;
1072                         }
1073                         if (option)
1074                                 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1075                         else
1076                                 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1077                         break;
1078                 case Opt_intr:
1079                         mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1080                         break;
1081                 case Opt_nointr:
1082                         mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1083                         break;
1084                 case Opt_err_panic:
1085                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1086                         break;
1087                 case Opt_err_ro:
1088                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1089                         break;
1090                 case Opt_data_ordered:
1091                         mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1092                         break;
1093                 case Opt_data_writeback:
1094                         mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1095                         break;
1096                 case Opt_user_xattr:
1097                         mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1098                         break;
1099                 case Opt_nouser_xattr:
1100                         mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1101                         break;
1102                 case Opt_atime_quantum:
1103                         if (match_int(&args[0], &option)) {
1104                                 status = 0;
1105                                 goto bail;
1106                         }
1107                         if (option >= 0)
1108                                 mopt->atime_quantum = option;
1109                         break;
1110                 case Opt_slot:
1111                         option = 0;
1112                         if (match_int(&args[0], &option)) {
1113                                 status = 0;
1114                                 goto bail;
1115                         }
1116                         if (option)
1117                                 mopt->slot = (s16)option;
1118                         break;
1119                 case Opt_commit:
1120                         option = 0;
1121                         if (match_int(&args[0], &option)) {
1122                                 status = 0;
1123                                 goto bail;
1124                         }
1125                         if (option < 0)
1126                                 return 0;
1127                         if (option == 0)
1128                                 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1129                         mopt->commit_interval = HZ * option;
1130                         break;
1131                 case Opt_localalloc:
1132                         option = 0;
1133                         if (match_int(&args[0], &option)) {
1134                                 status = 0;
1135                                 goto bail;
1136                         }
1137                         if (option >= 0 && (option <= ocfs2_local_alloc_size(sb) * 8))
1138                                 mopt->localalloc_opt = option;
1139                         break;
1140                 case Opt_localflocks:
1141                         /*
1142                          * Changing this during remount could race
1143                          * flock() requests, or "unbalance" existing
1144                          * ones (e.g., a lock is taken in one mode but
1145                          * dropped in the other). If users care enough
1146                          * to flip locking modes during remount, we
1147                          * could add a "local" flag to individual
1148                          * flock structures for proper tracking of
1149                          * state.
1150                          */
1151                         if (!is_remount)
1152                                 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1153                         break;
1154                 case Opt_stack:
1155                         /* Check both that the option we were passed
1156                          * is of the right length and that it is a proper
1157                          * string of the right length.
1158                          */
1159                         if (((args[0].to - args[0].from) !=
1160                              OCFS2_STACK_LABEL_LEN) ||
1161                             (strnlen(args[0].from,
1162                                      OCFS2_STACK_LABEL_LEN) !=
1163                              OCFS2_STACK_LABEL_LEN)) {
1164                                 mlog(ML_ERROR,
1165                                      "Invalid cluster_stack option\n");
1166                                 status = 0;
1167                                 goto bail;
1168                         }
1169                         memcpy(mopt->cluster_stack, args[0].from,
1170                                OCFS2_STACK_LABEL_LEN);
1171                         mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1172                         break;
1173                 case Opt_inode64:
1174                         mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1175                         break;
1176                 case Opt_usrquota:
1177                         /* We check only on remount, otherwise features
1178                          * aren't yet initialized. */
1179                         if (is_remount && !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1180                             OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1181                                 mlog(ML_ERROR, "User quota requested but "
1182                                      "filesystem feature is not set\n");
1183                                 status = 0;
1184                                 goto bail;
1185                         }
1186                         mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1187                         break;
1188                 case Opt_grpquota:
1189                         if (is_remount && !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1190                             OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1191                                 mlog(ML_ERROR, "Group quota requested but "
1192                                      "filesystem feature is not set\n");
1193                                 status = 0;
1194                                 goto bail;
1195                         }
1196                         mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1197                         break;
1198 #ifdef CONFIG_OCFS2_FS_POSIX_ACL
1199                 case Opt_acl:
1200                         mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1201                         break;
1202                 case Opt_noacl:
1203                         mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1204                         break;
1205 #else
1206                 case Opt_acl:
1207                 case Opt_noacl:
1208                         printk(KERN_INFO "ocfs2 (no)acl options not supported\n");
1209                         break;
1210 #endif
1211                 default:
1212                         mlog(ML_ERROR,
1213                              "Unrecognized mount option \"%s\" "
1214                              "or missing value\n", p);
1215                         status = 0;
1216                         goto bail;
1217                 }
1218         }
1219
1220         status = 1;
1221
1222 bail:
1223         mlog_exit(status);
1224         return status;
1225 }
1226
1227 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1228 {
1229         struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb);
1230         unsigned long opts = osb->s_mount_opt;
1231         unsigned int local_alloc_megs;
1232
1233         if (opts & OCFS2_MOUNT_HB_LOCAL)
1234                 seq_printf(s, ",_netdev,heartbeat=local");
1235         else
1236                 seq_printf(s, ",heartbeat=none");
1237
1238         if (opts & OCFS2_MOUNT_NOINTR)
1239                 seq_printf(s, ",nointr");
1240
1241         if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1242                 seq_printf(s, ",data=writeback");
1243         else
1244                 seq_printf(s, ",data=ordered");
1245
1246         if (opts & OCFS2_MOUNT_BARRIER)
1247                 seq_printf(s, ",barrier=1");
1248
1249         if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1250                 seq_printf(s, ",errors=panic");
1251         else
1252                 seq_printf(s, ",errors=remount-ro");
1253
1254         if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1255                 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1256
1257         if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM)
1258                 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1259
1260         if (osb->osb_commit_interval)
1261                 seq_printf(s, ",commit=%u",
1262                            (unsigned) (osb->osb_commit_interval / HZ));
1263
1264         local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1265         if (local_alloc_megs != OCFS2_DEFAULT_LOCAL_ALLOC_SIZE)
1266                 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1267
1268         if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1269                 seq_printf(s, ",localflocks,");
1270
1271         if (osb->osb_cluster_stack[0])
1272                 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1273                            osb->osb_cluster_stack);
1274         if (opts & OCFS2_MOUNT_USRQUOTA)
1275                 seq_printf(s, ",usrquota");
1276         if (opts & OCFS2_MOUNT_GRPQUOTA)
1277                 seq_printf(s, ",grpquota");
1278
1279         if (opts & OCFS2_MOUNT_NOUSERXATTR)
1280                 seq_printf(s, ",nouser_xattr");
1281         else
1282                 seq_printf(s, ",user_xattr");
1283
1284         if (opts & OCFS2_MOUNT_INODE64)
1285                 seq_printf(s, ",inode64");
1286
1287 #ifdef CONFIG_OCFS2_FS_POSIX_ACL
1288         if (opts & OCFS2_MOUNT_POSIX_ACL)
1289                 seq_printf(s, ",acl");
1290         else
1291                 seq_printf(s, ",noacl");
1292 #endif
1293
1294         return 0;
1295 }
1296
1297 static int __init ocfs2_init(void)
1298 {
1299         int status;
1300
1301         mlog_entry_void();
1302
1303         ocfs2_print_version();
1304
1305         status = init_ocfs2_uptodate_cache();
1306         if (status < 0) {
1307                 mlog_errno(status);
1308                 goto leave;
1309         }
1310
1311         status = ocfs2_initialize_mem_caches();
1312         if (status < 0) {
1313                 mlog_errno(status);
1314                 goto leave;
1315         }
1316
1317         ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1318         if (!ocfs2_wq) {
1319                 status = -ENOMEM;
1320                 goto leave;
1321         }
1322
1323         ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1324         if (!ocfs2_debugfs_root) {
1325                 status = -EFAULT;
1326                 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1327         }
1328
1329         status = ocfs2_quota_setup();
1330         if (status)
1331                 goto leave;
1332
1333         ocfs2_set_locking_protocol();
1334
1335         status = register_quota_format(&ocfs2_quota_format);
1336 leave:
1337         if (status < 0) {
1338                 ocfs2_quota_shutdown();
1339                 ocfs2_free_mem_caches();
1340                 exit_ocfs2_uptodate_cache();
1341         }
1342
1343         mlog_exit(status);
1344
1345         if (status >= 0) {
1346                 return register_filesystem(&ocfs2_fs_type);
1347         } else
1348                 return -1;
1349 }
1350
1351 static void __exit ocfs2_exit(void)
1352 {
1353         mlog_entry_void();
1354
1355         ocfs2_quota_shutdown();
1356
1357         if (ocfs2_wq) {
1358                 flush_workqueue(ocfs2_wq);
1359                 destroy_workqueue(ocfs2_wq);
1360         }
1361
1362         unregister_quota_format(&ocfs2_quota_format);
1363
1364         debugfs_remove(ocfs2_debugfs_root);
1365
1366         ocfs2_free_mem_caches();
1367
1368         unregister_filesystem(&ocfs2_fs_type);
1369
1370         exit_ocfs2_uptodate_cache();
1371
1372         mlog_exit_void();
1373 }
1374
1375 static void ocfs2_put_super(struct super_block *sb)
1376 {
1377         mlog_entry("(0x%p)\n", sb);
1378
1379         ocfs2_sync_blockdev(sb);
1380         ocfs2_dismount_volume(sb, 0);
1381
1382         mlog_exit_void();
1383 }
1384
1385 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1386 {
1387         struct ocfs2_super *osb;
1388         u32 numbits, freebits;
1389         int status;
1390         struct ocfs2_dinode *bm_lock;
1391         struct buffer_head *bh = NULL;
1392         struct inode *inode = NULL;
1393
1394         mlog_entry("(%p, %p)\n", dentry->d_sb, buf);
1395
1396         osb = OCFS2_SB(dentry->d_sb);
1397
1398         inode = ocfs2_get_system_file_inode(osb,
1399                                             GLOBAL_BITMAP_SYSTEM_INODE,
1400                                             OCFS2_INVALID_SLOT);
1401         if (!inode) {
1402                 mlog(ML_ERROR, "failed to get bitmap inode\n");
1403                 status = -EIO;
1404                 goto bail;
1405         }
1406
1407         status = ocfs2_inode_lock(inode, &bh, 0);
1408         if (status < 0) {
1409                 mlog_errno(status);
1410                 goto bail;
1411         }
1412
1413         bm_lock = (struct ocfs2_dinode *) bh->b_data;
1414
1415         numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1416         freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1417
1418         buf->f_type = OCFS2_SUPER_MAGIC;
1419         buf->f_bsize = dentry->d_sb->s_blocksize;
1420         buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1421         buf->f_blocks = ((sector_t) numbits) *
1422                         (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1423         buf->f_bfree = ((sector_t) freebits) *
1424                        (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1425         buf->f_bavail = buf->f_bfree;
1426         buf->f_files = numbits;
1427         buf->f_ffree = freebits;
1428
1429         brelse(bh);
1430
1431         ocfs2_inode_unlock(inode, 0);
1432         status = 0;
1433 bail:
1434         if (inode)
1435                 iput(inode);
1436
1437         mlog_exit(status);
1438
1439         return status;
1440 }
1441
1442 static void ocfs2_inode_init_once(void *data)
1443 {
1444         struct ocfs2_inode_info *oi = data;
1445
1446         oi->ip_flags = 0;
1447         oi->ip_open_count = 0;
1448         spin_lock_init(&oi->ip_lock);
1449         ocfs2_extent_map_init(&oi->vfs_inode);
1450         INIT_LIST_HEAD(&oi->ip_io_markers);
1451         oi->ip_created_trans = 0;
1452         oi->ip_last_trans = 0;
1453         oi->ip_dir_start_lookup = 0;
1454
1455         init_rwsem(&oi->ip_alloc_sem);
1456         init_rwsem(&oi->ip_xattr_sem);
1457         mutex_init(&oi->ip_io_mutex);
1458
1459         oi->ip_blkno = 0ULL;
1460         oi->ip_clusters = 0;
1461
1462         ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1463         ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1464         ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1465
1466         ocfs2_metadata_cache_init(&oi->vfs_inode);
1467
1468         inode_init_once(&oi->vfs_inode);
1469 }
1470
1471 static int ocfs2_initialize_mem_caches(void)
1472 {
1473         ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1474                                        sizeof(struct ocfs2_inode_info),
1475                                        0,
1476                                        (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1477                                                 SLAB_MEM_SPREAD),
1478                                        ocfs2_inode_init_once);
1479         ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1480                                         sizeof(struct ocfs2_dquot),
1481                                         0,
1482                                         (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1483                                                 SLAB_MEM_SPREAD),
1484                                         NULL);
1485         ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1486                                         sizeof(struct ocfs2_quota_chunk),
1487                                         0,
1488                                         (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1489                                         NULL);
1490         if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1491             !ocfs2_qf_chunk_cachep) {
1492                 if (ocfs2_inode_cachep)
1493                         kmem_cache_destroy(ocfs2_inode_cachep);
1494                 if (ocfs2_dquot_cachep)
1495                         kmem_cache_destroy(ocfs2_dquot_cachep);
1496                 if (ocfs2_qf_chunk_cachep)
1497                         kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1498                 return -ENOMEM;
1499         }
1500
1501         return 0;
1502 }
1503
1504 static void ocfs2_free_mem_caches(void)
1505 {
1506         if (ocfs2_inode_cachep)
1507                 kmem_cache_destroy(ocfs2_inode_cachep);
1508         ocfs2_inode_cachep = NULL;
1509
1510         if (ocfs2_dquot_cachep)
1511                 kmem_cache_destroy(ocfs2_dquot_cachep);
1512         ocfs2_dquot_cachep = NULL;
1513
1514         if (ocfs2_qf_chunk_cachep)
1515                 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1516         ocfs2_qf_chunk_cachep = NULL;
1517 }
1518
1519 static int ocfs2_get_sector(struct super_block *sb,
1520                             struct buffer_head **bh,
1521                             int block,
1522                             int sect_size)
1523 {
1524         if (!sb_set_blocksize(sb, sect_size)) {
1525                 mlog(ML_ERROR, "unable to set blocksize\n");
1526                 return -EIO;
1527         }
1528
1529         *bh = sb_getblk(sb, block);
1530         if (!*bh) {
1531                 mlog_errno(-EIO);
1532                 return -EIO;
1533         }
1534         lock_buffer(*bh);
1535         if (!buffer_dirty(*bh))
1536                 clear_buffer_uptodate(*bh);
1537         unlock_buffer(*bh);
1538         ll_rw_block(READ, 1, bh);
1539         wait_on_buffer(*bh);
1540         return 0;
1541 }
1542
1543 static int ocfs2_mount_volume(struct super_block *sb)
1544 {
1545         int status = 0;
1546         int unlock_super = 0;
1547         struct ocfs2_super *osb = OCFS2_SB(sb);
1548
1549         mlog_entry_void();
1550
1551         if (ocfs2_is_hard_readonly(osb))
1552                 goto leave;
1553
1554         status = ocfs2_dlm_init(osb);
1555         if (status < 0) {
1556                 mlog_errno(status);
1557                 goto leave;
1558         }
1559
1560         status = ocfs2_super_lock(osb, 1);
1561         if (status < 0) {
1562                 mlog_errno(status);
1563                 goto leave;
1564         }
1565         unlock_super = 1;
1566
1567         /* This will load up the node map and add ourselves to it. */
1568         status = ocfs2_find_slot(osb);
1569         if (status < 0) {
1570                 mlog_errno(status);
1571                 goto leave;
1572         }
1573
1574         /* load all node-local system inodes */
1575         status = ocfs2_init_local_system_inodes(osb);
1576         if (status < 0) {
1577                 mlog_errno(status);
1578                 goto leave;
1579         }
1580
1581         status = ocfs2_check_volume(osb);
1582         if (status < 0) {
1583                 mlog_errno(status);
1584                 goto leave;
1585         }
1586
1587         status = ocfs2_truncate_log_init(osb);
1588         if (status < 0) {
1589                 mlog_errno(status);
1590                 goto leave;
1591         }
1592
1593         if (ocfs2_mount_local(osb))
1594                 goto leave;
1595
1596 leave:
1597         if (unlock_super)
1598                 ocfs2_super_unlock(osb, 1);
1599
1600         mlog_exit(status);
1601         return status;
1602 }
1603
1604 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1605 {
1606         int tmp, hangup_needed = 0;
1607         struct ocfs2_super *osb = NULL;
1608         char nodestr[8];
1609
1610         mlog_entry("(0x%p)\n", sb);
1611
1612         BUG_ON(!sb);
1613         osb = OCFS2_SB(sb);
1614         BUG_ON(!osb);
1615
1616         ocfs2_disable_quotas(osb);
1617
1618         ocfs2_shutdown_local_alloc(osb);
1619
1620         ocfs2_truncate_log_shutdown(osb);
1621
1622         /* This will disable recovery and flush any recovery work. */
1623         ocfs2_recovery_exit(osb);
1624
1625         ocfs2_journal_shutdown(osb);
1626
1627         ocfs2_sync_blockdev(sb);
1628
1629         /* No cluster connection means we've failed during mount, so skip
1630          * all the steps which depended on that to complete. */
1631         if (osb->cconn) {
1632                 tmp = ocfs2_super_lock(osb, 1);
1633                 if (tmp < 0) {
1634                         mlog_errno(tmp);
1635                         return;
1636                 }
1637         }
1638
1639         if (osb->slot_num != OCFS2_INVALID_SLOT)
1640                 ocfs2_put_slot(osb);
1641
1642         if (osb->cconn)
1643                 ocfs2_super_unlock(osb, 1);
1644
1645         ocfs2_release_system_inodes(osb);
1646
1647         /*
1648          * If we're dismounting due to mount error, mount.ocfs2 will clean
1649          * up heartbeat.  If we're a local mount, there is no heartbeat.
1650          * If we failed before we got a uuid_str yet, we can't stop
1651          * heartbeat.  Otherwise, do it.
1652          */
1653         if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str)
1654                 hangup_needed = 1;
1655
1656         if (osb->cconn)
1657                 ocfs2_dlm_shutdown(osb, hangup_needed);
1658
1659         debugfs_remove(osb->osb_debug_root);
1660
1661         if (hangup_needed)
1662                 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1663
1664         atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1665
1666         if (ocfs2_mount_local(osb))
1667                 snprintf(nodestr, sizeof(nodestr), "local");
1668         else
1669                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1670
1671         printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1672                osb->dev_str, nodestr);
1673
1674         ocfs2_delete_osb(osb);
1675         kfree(osb);
1676         sb->s_dev = 0;
1677         sb->s_fs_info = NULL;
1678 }
1679
1680 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1681                                 unsigned uuid_bytes)
1682 {
1683         int i, ret;
1684         char *ptr;
1685
1686         BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1687
1688         osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1689         if (osb->uuid_str == NULL)
1690                 return -ENOMEM;
1691
1692         for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1693                 /* print with null */
1694                 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1695                 if (ret != 2) /* drop super cleans up */
1696                         return -EINVAL;
1697                 /* then only advance past the last char */
1698                 ptr += 2;
1699         }
1700
1701         return 0;
1702 }
1703
1704 static int ocfs2_initialize_super(struct super_block *sb,
1705                                   struct buffer_head *bh,
1706                                   int sector_size)
1707 {
1708         int status;
1709         int i, cbits, bbits;
1710         struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
1711         struct inode *inode = NULL;
1712         struct ocfs2_journal *journal;
1713         __le32 uuid_net_key;
1714         struct ocfs2_super *osb;
1715
1716         mlog_entry_void();
1717
1718         osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
1719         if (!osb) {
1720                 status = -ENOMEM;
1721                 mlog_errno(status);
1722                 goto bail;
1723         }
1724
1725         sb->s_fs_info = osb;
1726         sb->s_op = &ocfs2_sops;
1727         sb->s_export_op = &ocfs2_export_ops;
1728         sb->s_qcop = &ocfs2_quotactl_ops;
1729         sb->dq_op = &ocfs2_quota_operations;
1730         sb->s_xattr = ocfs2_xattr_handlers;
1731         sb->s_time_gran = 1;
1732         sb->s_flags |= MS_NOATIME;
1733         /* this is needed to support O_LARGEFILE */
1734         cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
1735         bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
1736         sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
1737
1738         osb->sb = sb;
1739         /* Save off for ocfs2_rw_direct */
1740         osb->s_sectsize_bits = blksize_bits(sector_size);
1741         BUG_ON(!osb->s_sectsize_bits);
1742
1743         spin_lock_init(&osb->dc_task_lock);
1744         init_waitqueue_head(&osb->dc_event);
1745         osb->dc_work_sequence = 0;
1746         osb->dc_wake_sequence = 0;
1747         INIT_LIST_HEAD(&osb->blocked_lock_list);
1748         osb->blocked_lock_count = 0;
1749         spin_lock_init(&osb->osb_lock);
1750         ocfs2_init_inode_steal_slot(osb);
1751
1752         atomic_set(&osb->alloc_stats.moves, 0);
1753         atomic_set(&osb->alloc_stats.local_data, 0);
1754         atomic_set(&osb->alloc_stats.bitmap_data, 0);
1755         atomic_set(&osb->alloc_stats.bg_allocs, 0);
1756         atomic_set(&osb->alloc_stats.bg_extends, 0);
1757
1758         ocfs2_init_node_maps(osb);
1759
1760         snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
1761                  MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
1762
1763         status = ocfs2_recovery_init(osb);
1764         if (status) {
1765                 mlog(ML_ERROR, "Unable to initialize recovery state\n");
1766                 mlog_errno(status);
1767                 goto bail;
1768         }
1769
1770         init_waitqueue_head(&osb->checkpoint_event);
1771         atomic_set(&osb->needs_checkpoint, 0);
1772
1773         osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1774
1775         osb->slot_num = OCFS2_INVALID_SLOT;
1776
1777         osb->s_xattr_inline_size = le16_to_cpu(
1778                                         di->id2.i_super.s_xattr_inline_size);
1779
1780         osb->local_alloc_state = OCFS2_LA_UNUSED;
1781         osb->local_alloc_bh = NULL;
1782         INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
1783
1784         init_waitqueue_head(&osb->osb_mount_event);
1785
1786         osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
1787         if (!osb->vol_label) {
1788                 mlog(ML_ERROR, "unable to alloc vol label\n");
1789                 status = -ENOMEM;
1790                 goto bail;
1791         }
1792
1793         osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
1794         if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
1795                 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
1796                      osb->max_slots);
1797                 status = -EINVAL;
1798                 goto bail;
1799         }
1800         mlog(0, "max_slots for this device: %u\n", osb->max_slots);
1801
1802         osb->slot_recovery_generations =
1803                 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
1804                         GFP_KERNEL);
1805         if (!osb->slot_recovery_generations) {
1806                 status = -ENOMEM;
1807                 mlog_errno(status);
1808                 goto bail;
1809         }
1810
1811         init_waitqueue_head(&osb->osb_wipe_event);
1812         osb->osb_orphan_wipes = kcalloc(osb->max_slots,
1813                                         sizeof(*osb->osb_orphan_wipes),
1814                                         GFP_KERNEL);
1815         if (!osb->osb_orphan_wipes) {
1816                 status = -ENOMEM;
1817                 mlog_errno(status);
1818                 goto bail;
1819         }
1820
1821         osb->s_feature_compat =
1822                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
1823         osb->s_feature_ro_compat =
1824                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
1825         osb->s_feature_incompat =
1826                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
1827
1828         if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
1829                 mlog(ML_ERROR, "couldn't mount because of unsupported "
1830                      "optional features (%x).\n", i);
1831                 status = -EINVAL;
1832                 goto bail;
1833         }
1834         if (!(osb->sb->s_flags & MS_RDONLY) &&
1835             (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
1836                 mlog(ML_ERROR, "couldn't mount RDWR because of "
1837                      "unsupported optional features (%x).\n", i);
1838                 status = -EINVAL;
1839                 goto bail;
1840         }
1841
1842         if (ocfs2_userspace_stack(osb)) {
1843                 memcpy(osb->osb_cluster_stack,
1844                        OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
1845                        OCFS2_STACK_LABEL_LEN);
1846                 osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1847                 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
1848                         mlog(ML_ERROR,
1849                              "couldn't mount because of an invalid "
1850                              "cluster stack label (%s) \n",
1851                              osb->osb_cluster_stack);
1852                         status = -EINVAL;
1853                         goto bail;
1854                 }
1855         } else {
1856                 /* The empty string is identical with classic tools that
1857                  * don't know about s_cluster_info. */
1858                 osb->osb_cluster_stack[0] = '\0';
1859         }
1860
1861         get_random_bytes(&osb->s_next_generation, sizeof(u32));
1862
1863         /* FIXME
1864          * This should be done in ocfs2_journal_init(), but unknown
1865          * ordering issues will cause the filesystem to crash.
1866          * If anyone wants to figure out what part of the code
1867          * refers to osb->journal before ocfs2_journal_init() is run,
1868          * be my guest.
1869          */
1870         /* initialize our journal structure */
1871
1872         journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
1873         if (!journal) {
1874                 mlog(ML_ERROR, "unable to alloc journal\n");
1875                 status = -ENOMEM;
1876                 goto bail;
1877         }
1878         osb->journal = journal;
1879         journal->j_osb = osb;
1880
1881         atomic_set(&journal->j_num_trans, 0);
1882         init_rwsem(&journal->j_trans_barrier);
1883         init_waitqueue_head(&journal->j_checkpointed);
1884         spin_lock_init(&journal->j_lock);
1885         journal->j_trans_id = (unsigned long) 1;
1886         INIT_LIST_HEAD(&journal->j_la_cleanups);
1887         INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
1888         journal->j_state = OCFS2_JOURNAL_FREE;
1889
1890         /* get some pseudo constants for clustersize bits */
1891         osb->s_clustersize_bits =
1892                 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
1893         osb->s_clustersize = 1 << osb->s_clustersize_bits;
1894         mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
1895
1896         if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
1897             osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
1898                 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
1899                      osb->s_clustersize);
1900                 status = -EINVAL;
1901                 goto bail;
1902         }
1903
1904         if (ocfs2_clusters_to_blocks(osb->sb, le32_to_cpu(di->i_clusters) - 1)
1905             > (u32)~0UL) {
1906                 mlog(ML_ERROR, "Volume might try to write to blocks beyond "
1907                      "what jbd can address in 32 bits.\n");
1908                 status = -EINVAL;
1909                 goto bail;
1910         }
1911
1912         if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
1913                                  sizeof(di->id2.i_super.s_uuid))) {
1914                 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
1915                 status = -ENOMEM;
1916                 goto bail;
1917         }
1918
1919         memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
1920
1921         strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
1922         osb->vol_label[63] = '\0';
1923         osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
1924         osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
1925         osb->first_cluster_group_blkno =
1926                 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
1927         osb->fs_generation = le32_to_cpu(di->i_fs_generation);
1928         osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
1929         mlog(0, "vol_label: %s\n", osb->vol_label);
1930         mlog(0, "uuid: %s\n", osb->uuid_str);
1931         mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
1932              (unsigned long long)osb->root_blkno,
1933              (unsigned long long)osb->system_dir_blkno);
1934
1935         osb->osb_dlm_debug = ocfs2_new_dlm_debug();
1936         if (!osb->osb_dlm_debug) {
1937                 status = -ENOMEM;
1938                 mlog_errno(status);
1939                 goto bail;
1940         }
1941
1942         atomic_set(&osb->vol_state, VOLUME_INIT);
1943
1944         /* load root, system_dir, and all global system inodes */
1945         status = ocfs2_init_global_system_inodes(osb);
1946         if (status < 0) {
1947                 mlog_errno(status);
1948                 goto bail;
1949         }
1950
1951         /*
1952          * global bitmap
1953          */
1954         inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
1955                                             OCFS2_INVALID_SLOT);
1956         if (!inode) {
1957                 status = -EINVAL;
1958                 mlog_errno(status);
1959                 goto bail;
1960         }
1961
1962         osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
1963         iput(inode);
1964
1965         osb->bitmap_cpg = ocfs2_group_bitmap_size(sb) * 8;
1966
1967         status = ocfs2_init_slot_info(osb);
1968         if (status < 0) {
1969                 mlog_errno(status);
1970                 goto bail;
1971         }
1972
1973 bail:
1974         mlog_exit(status);
1975         return status;
1976 }
1977
1978 /*
1979  * will return: -EAGAIN if it is ok to keep searching for superblocks
1980  *              -EINVAL if there is a bad superblock
1981  *              0 on success
1982  */
1983 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
1984                                struct buffer_head *bh,
1985                                u32 blksz)
1986 {
1987         int status = -EAGAIN;
1988
1989         mlog_entry_void();
1990
1991         if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
1992                    strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
1993                 /* We have to do a raw check of the feature here */
1994                 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
1995                     OCFS2_FEATURE_INCOMPAT_META_ECC) {
1996                         status = ocfs2_block_check_validate(bh->b_data,
1997                                                             bh->b_size,
1998                                                             &di->i_check);
1999                         if (status)
2000                                 goto out;
2001                 }
2002                 status = -EINVAL;
2003                 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2004                         mlog(ML_ERROR, "found superblock with incorrect block "
2005                              "size: found %u, should be %u\n",
2006                              1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2007                                blksz);
2008                 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2009                            OCFS2_MAJOR_REV_LEVEL ||
2010                            le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2011                            OCFS2_MINOR_REV_LEVEL) {
2012                         mlog(ML_ERROR, "found superblock with bad version: "
2013                              "found %u.%u, should be %u.%u\n",
2014                              le16_to_cpu(di->id2.i_super.s_major_rev_level),
2015                              le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2016                              OCFS2_MAJOR_REV_LEVEL,
2017                              OCFS2_MINOR_REV_LEVEL);
2018                 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2019                         mlog(ML_ERROR, "bad block number on superblock: "
2020                              "found %llu, should be %llu\n",
2021                              (unsigned long long)le64_to_cpu(di->i_blkno),
2022                              (unsigned long long)bh->b_blocknr);
2023                 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2024                             le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2025                         mlog(ML_ERROR, "bad cluster size found: %u\n",
2026                              1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2027                 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2028                         mlog(ML_ERROR, "bad root_blkno: 0\n");
2029                 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2030                         mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2031                 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2032                         mlog(ML_ERROR,
2033                              "Superblock slots found greater than file system "
2034                              "maximum: found %u, max %u\n",
2035                              le16_to_cpu(di->id2.i_super.s_max_slots),
2036                              OCFS2_MAX_SLOTS);
2037                 } else {
2038                         /* found it! */
2039                         status = 0;
2040                 }
2041         }
2042
2043 out:
2044         mlog_exit(status);
2045         return status;
2046 }
2047
2048 static int ocfs2_check_volume(struct ocfs2_super *osb)
2049 {
2050         int status;
2051         int dirty;
2052         int local;
2053         struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2054                                                   * recover
2055                                                   * ourselves. */
2056
2057         mlog_entry_void();
2058
2059         /* Init our journal object. */
2060         status = ocfs2_journal_init(osb->journal, &dirty);
2061         if (status < 0) {
2062                 mlog(ML_ERROR, "Could not initialize journal!\n");
2063                 goto finally;
2064         }
2065
2066         /* If the journal was unmounted cleanly then we don't want to
2067          * recover anything. Otherwise, journal_load will do that
2068          * dirty work for us :) */
2069         if (!dirty) {
2070                 status = ocfs2_journal_wipe(osb->journal, 0);
2071                 if (status < 0) {
2072                         mlog_errno(status);
2073                         goto finally;
2074                 }
2075         } else {
2076                 mlog(ML_NOTICE, "File system was not unmounted cleanly, "
2077                      "recovering volume.\n");
2078         }
2079
2080         local = ocfs2_mount_local(osb);
2081
2082         /* will play back anything left in the journal. */
2083         status = ocfs2_journal_load(osb->journal, local, dirty);
2084         if (status < 0) {
2085                 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2086                 goto finally;
2087         }
2088
2089         if (dirty) {
2090                 /* recover my local alloc if we didn't unmount cleanly. */
2091                 status = ocfs2_begin_local_alloc_recovery(osb,
2092                                                           osb->slot_num,
2093                                                           &local_alloc);
2094                 if (status < 0) {
2095                         mlog_errno(status);
2096                         goto finally;
2097                 }
2098                 /* we complete the recovery process after we've marked
2099                  * ourselves as mounted. */
2100         }
2101
2102         mlog(0, "Journal loaded.\n");
2103
2104         status = ocfs2_load_local_alloc(osb);
2105         if (status < 0) {
2106                 mlog_errno(status);
2107                 goto finally;
2108         }
2109
2110         if (dirty) {
2111                 /* Recovery will be completed after we've mounted the
2112                  * rest of the volume. */
2113                 osb->dirty = 1;
2114                 osb->local_alloc_copy = local_alloc;
2115                 local_alloc = NULL;
2116         }
2117
2118         /* go through each journal, trylock it and if you get the
2119          * lock, and it's marked as dirty, set the bit in the recover
2120          * map and launch a recovery thread for it. */
2121         status = ocfs2_mark_dead_nodes(osb);
2122         if (status < 0)
2123                 mlog_errno(status);
2124
2125 finally:
2126         if (local_alloc)
2127                 kfree(local_alloc);
2128
2129         mlog_exit(status);
2130         return status;
2131 }
2132
2133 /*
2134  * The routine gets called from dismount or close whenever a dismount on
2135  * volume is requested and the osb open count becomes 1.
2136  * It will remove the osb from the global list and also free up all the
2137  * initialized resources and fileobject.
2138  */
2139 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2140 {
2141         mlog_entry_void();
2142
2143         /* This function assumes that the caller has the main osb resource */
2144
2145         ocfs2_free_slot_info(osb);
2146
2147         kfree(osb->osb_orphan_wipes);
2148         kfree(osb->slot_recovery_generations);
2149         /* FIXME
2150          * This belongs in journal shutdown, but because we have to
2151          * allocate osb->journal at the start of ocfs2_initalize_osb(),
2152          * we free it here.
2153          */
2154         kfree(osb->journal);
2155         if (osb->local_alloc_copy)
2156                 kfree(osb->local_alloc_copy);
2157         kfree(osb->uuid_str);
2158         ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2159         memset(osb, 0, sizeof(struct ocfs2_super));
2160
2161         mlog_exit_void();
2162 }
2163
2164 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2165  * panic(). We do not support continue-on-error operation. */
2166 static void ocfs2_handle_error(struct super_block *sb)
2167 {
2168         struct ocfs2_super *osb = OCFS2_SB(sb);
2169
2170         if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2171                 panic("OCFS2: (device %s): panic forced after error\n",
2172                       sb->s_id);
2173
2174         ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2175
2176         if (sb->s_flags & MS_RDONLY &&
2177             (ocfs2_is_soft_readonly(osb) ||
2178              ocfs2_is_hard_readonly(osb)))
2179                 return;
2180
2181         printk(KERN_CRIT "File system is now read-only due to the potential "
2182                "of on-disk corruption. Please run fsck.ocfs2 once the file "
2183                "system is unmounted.\n");
2184         sb->s_flags |= MS_RDONLY;
2185         ocfs2_set_ro_flag(osb, 0);
2186 }
2187
2188 static char error_buf[1024];
2189
2190 void __ocfs2_error(struct super_block *sb,
2191                    const char *function,
2192                    const char *fmt, ...)
2193 {
2194         va_list args;
2195
2196         va_start(args, fmt);
2197         vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2198         va_end(args);
2199
2200         /* Not using mlog here because we want to show the actual
2201          * function the error came from. */
2202         printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2203                sb->s_id, function, error_buf);
2204
2205         ocfs2_handle_error(sb);
2206 }
2207
2208 /* Handle critical errors. This is intentionally more drastic than
2209  * ocfs2_handle_error, so we only use for things like journal errors,
2210  * etc. */
2211 void __ocfs2_abort(struct super_block* sb,
2212                    const char *function,
2213                    const char *fmt, ...)
2214 {
2215         va_list args;
2216
2217         va_start(args, fmt);
2218         vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2219         va_end(args);
2220
2221         printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2222                sb->s_id, function, error_buf);
2223
2224         /* We don't have the cluster support yet to go straight to
2225          * hard readonly in here. Until then, we want to keep
2226          * ocfs2_abort() so that we can at least mark critical
2227          * errors.
2228          *
2229          * TODO: This should abort the journal and alert other nodes
2230          * that our slot needs recovery. */
2231
2232         /* Force a panic(). This stinks, but it's better than letting
2233          * things continue without having a proper hard readonly
2234          * here. */
2235         OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2236         ocfs2_handle_error(sb);
2237 }
2238
2239 module_init(ocfs2_init);
2240 module_exit(ocfs2_exit);