1 /* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
6 * load/unload driver, mount/dismount volumes
8 * Copyright (C) 2002, 2004 Oracle. All rights reserved.
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.
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.
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.
26 #include <linux/module.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>
47 #define MLOG_MASK_PREFIX ML_SUPER
48 #include <cluster/masklog.h>
52 /* this should be the only file to include a version 1 header */
53 #include "ocfs1_fs_compat.h"
56 #include "blockcheck.h"
59 #include "extent_map.h"
60 #include "heartbeat.h"
63 #include "localalloc.h"
72 #include "refcounttree.h"
74 #include "buffer_head_io.h"
76 static struct kmem_cache
*ocfs2_inode_cachep
= NULL
;
77 struct kmem_cache
*ocfs2_dquot_cachep
;
78 struct kmem_cache
*ocfs2_qf_chunk_cachep
;
80 /* OCFS2 needs to schedule several differnt types of work which
81 * require cluster locking, disk I/O, recovery waits, etc. Since these
82 * types of work tend to be heavy we avoid using the kernel events
83 * workqueue and schedule on our own. */
84 struct workqueue_struct
*ocfs2_wq
= NULL
;
86 static struct dentry
*ocfs2_debugfs_root
= NULL
;
88 MODULE_AUTHOR("Oracle");
89 MODULE_LICENSE("GPL");
93 unsigned long commit_interval
;
94 unsigned long mount_opt
;
95 unsigned int atime_quantum
;
97 unsigned int localalloc_opt
;
98 char cluster_stack
[OCFS2_STACK_LABEL_LEN
+ 1];
101 static int ocfs2_parse_options(struct super_block
*sb
, char *options
,
102 struct mount_options
*mopt
,
104 static int ocfs2_show_options(struct seq_file
*s
, struct vfsmount
*mnt
);
105 static void ocfs2_put_super(struct super_block
*sb
);
106 static int ocfs2_mount_volume(struct super_block
*sb
);
107 static int ocfs2_remount(struct super_block
*sb
, int *flags
, char *data
);
108 static void ocfs2_dismount_volume(struct super_block
*sb
, int mnt_err
);
109 static int ocfs2_initialize_mem_caches(void);
110 static void ocfs2_free_mem_caches(void);
111 static void ocfs2_delete_osb(struct ocfs2_super
*osb
);
113 static int ocfs2_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
115 static int ocfs2_sync_fs(struct super_block
*sb
, int wait
);
117 static int ocfs2_init_global_system_inodes(struct ocfs2_super
*osb
);
118 static int ocfs2_init_local_system_inodes(struct ocfs2_super
*osb
);
119 static void ocfs2_release_system_inodes(struct ocfs2_super
*osb
);
120 static int ocfs2_check_volume(struct ocfs2_super
*osb
);
121 static int ocfs2_verify_volume(struct ocfs2_dinode
*di
,
122 struct buffer_head
*bh
,
124 struct ocfs2_blockcheck_stats
*stats
);
125 static int ocfs2_initialize_super(struct super_block
*sb
,
126 struct buffer_head
*bh
,
128 struct ocfs2_blockcheck_stats
*stats
);
129 static int ocfs2_get_sector(struct super_block
*sb
,
130 struct buffer_head
**bh
,
133 static struct inode
*ocfs2_alloc_inode(struct super_block
*sb
);
134 static void ocfs2_destroy_inode(struct inode
*inode
);
135 static int ocfs2_susp_quotas(struct ocfs2_super
*osb
, int unsuspend
);
136 static int ocfs2_enable_quotas(struct ocfs2_super
*osb
);
137 static void ocfs2_disable_quotas(struct ocfs2_super
*osb
);
139 static const struct super_operations ocfs2_sops
= {
140 .statfs
= ocfs2_statfs
,
141 .alloc_inode
= ocfs2_alloc_inode
,
142 .destroy_inode
= ocfs2_destroy_inode
,
143 .drop_inode
= ocfs2_drop_inode
,
144 .clear_inode
= ocfs2_clear_inode
,
145 .delete_inode
= ocfs2_delete_inode
,
146 .sync_fs
= ocfs2_sync_fs
,
147 .put_super
= ocfs2_put_super
,
148 .remount_fs
= ocfs2_remount
,
149 .show_options
= ocfs2_show_options
,
150 .quota_read
= ocfs2_quota_read
,
151 .quota_write
= ocfs2_quota_write
,
180 static const match_table_t tokens
= {
181 {Opt_barrier
, "barrier=%u"},
182 {Opt_err_panic
, "errors=panic"},
183 {Opt_err_ro
, "errors=remount-ro"},
185 {Opt_nointr
, "nointr"},
186 {Opt_hb_none
, OCFS2_HB_NONE
},
187 {Opt_hb_local
, OCFS2_HB_LOCAL
},
188 {Opt_data_ordered
, "data=ordered"},
189 {Opt_data_writeback
, "data=writeback"},
190 {Opt_atime_quantum
, "atime_quantum=%u"},
191 {Opt_slot
, "preferred_slot=%u"},
192 {Opt_commit
, "commit=%u"},
193 {Opt_localalloc
, "localalloc=%d"},
194 {Opt_localflocks
, "localflocks"},
195 {Opt_stack
, "cluster_stack=%s"},
196 {Opt_user_xattr
, "user_xattr"},
197 {Opt_nouser_xattr
, "nouser_xattr"},
198 {Opt_inode64
, "inode64"},
200 {Opt_noacl
, "noacl"},
201 {Opt_usrquota
, "usrquota"},
202 {Opt_grpquota
, "grpquota"},
206 #ifdef CONFIG_DEBUG_FS
207 static int ocfs2_osb_dump(struct ocfs2_super
*osb
, char *buf
, int len
)
209 struct ocfs2_cluster_connection
*cconn
= osb
->cconn
;
210 struct ocfs2_recovery_map
*rm
= osb
->recovery_map
;
211 struct ocfs2_orphan_scan
*os
= &osb
->osb_orphan_scan
;
214 out
+= snprintf(buf
+ out
, len
- out
,
215 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n",
216 "Device", osb
->dev_str
, osb
->uuid_str
,
217 osb
->fs_generation
, osb
->vol_label
);
219 out
+= snprintf(buf
+ out
, len
- out
,
220 "%10s => State: %d Flags: 0x%lX\n", "Volume",
221 atomic_read(&osb
->vol_state
), osb
->osb_flags
);
223 out
+= snprintf(buf
+ out
, len
- out
,
224 "%10s => Block: %lu Cluster: %d\n", "Sizes",
225 osb
->sb
->s_blocksize
, osb
->s_clustersize
);
227 out
+= snprintf(buf
+ out
, len
- out
,
228 "%10s => Compat: 0x%X Incompat: 0x%X "
230 "Features", osb
->s_feature_compat
,
231 osb
->s_feature_incompat
, osb
->s_feature_ro_compat
);
233 out
+= snprintf(buf
+ out
, len
- out
,
234 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount",
235 osb
->s_mount_opt
, osb
->s_atime_quantum
);
238 out
+= snprintf(buf
+ out
, len
- out
,
239 "%10s => Stack: %s Name: %*s "
240 "Version: %d.%d\n", "Cluster",
241 (*osb
->osb_cluster_stack
== '\0' ?
242 "o2cb" : osb
->osb_cluster_stack
),
243 cconn
->cc_namelen
, cconn
->cc_name
,
244 cconn
->cc_version
.pv_major
,
245 cconn
->cc_version
.pv_minor
);
248 spin_lock(&osb
->dc_task_lock
);
249 out
+= snprintf(buf
+ out
, len
- out
,
250 "%10s => Pid: %d Count: %lu WakeSeq: %lu "
251 "WorkSeq: %lu\n", "DownCnvt",
252 (osb
->dc_task
? task_pid_nr(osb
->dc_task
) : -1),
253 osb
->blocked_lock_count
, osb
->dc_wake_sequence
,
254 osb
->dc_work_sequence
);
255 spin_unlock(&osb
->dc_task_lock
);
257 spin_lock(&osb
->osb_lock
);
258 out
+= snprintf(buf
+ out
, len
- out
, "%10s => Pid: %d Nodes:",
260 (osb
->recovery_thread_task
?
261 task_pid_nr(osb
->recovery_thread_task
) : -1));
262 if (rm
->rm_used
== 0)
263 out
+= snprintf(buf
+ out
, len
- out
, " None\n");
265 for (i
= 0; i
< rm
->rm_used
; i
++)
266 out
+= snprintf(buf
+ out
, len
- out
, " %d",
268 out
+= snprintf(buf
+ out
, len
- out
, "\n");
270 spin_unlock(&osb
->osb_lock
);
272 out
+= snprintf(buf
+ out
, len
- out
,
273 "%10s => Pid: %d Interval: %lu Needs: %d\n", "Commit",
274 (osb
->commit_task
? task_pid_nr(osb
->commit_task
) : -1),
275 osb
->osb_commit_interval
,
276 atomic_read(&osb
->needs_checkpoint
));
278 out
+= snprintf(buf
+ out
, len
- out
,
279 "%10s => State: %d TxnId: %lu NumTxns: %d\n",
280 "Journal", osb
->journal
->j_state
,
281 osb
->journal
->j_trans_id
,
282 atomic_read(&osb
->journal
->j_num_trans
));
284 out
+= snprintf(buf
+ out
, len
- out
,
285 "%10s => GlobalAllocs: %d LocalAllocs: %d "
286 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
288 atomic_read(&osb
->alloc_stats
.bitmap_data
),
289 atomic_read(&osb
->alloc_stats
.local_data
),
290 atomic_read(&osb
->alloc_stats
.bg_allocs
),
291 atomic_read(&osb
->alloc_stats
.moves
),
292 atomic_read(&osb
->alloc_stats
.bg_extends
));
294 out
+= snprintf(buf
+ out
, len
- out
,
295 "%10s => State: %u Descriptor: %llu Size: %u bits "
296 "Default: %u bits\n",
297 "LocalAlloc", osb
->local_alloc_state
,
298 (unsigned long long)osb
->la_last_gd
,
299 osb
->local_alloc_bits
, osb
->local_alloc_default_bits
);
301 spin_lock(&osb
->osb_lock
);
302 out
+= snprintf(buf
+ out
, len
- out
,
303 "%10s => Slot: %d NumStolen: %d\n", "Steal",
304 osb
->s_inode_steal_slot
,
305 atomic_read(&osb
->s_num_inodes_stolen
));
306 spin_unlock(&osb
->osb_lock
);
308 out
+= snprintf(buf
+ out
, len
- out
, "OrphanScan => ");
309 out
+= snprintf(buf
+ out
, len
- out
, "Local: %u Global: %u ",
310 os
->os_count
, os
->os_seqno
);
311 out
+= snprintf(buf
+ out
, len
- out
, " Last Scan: ");
312 if (atomic_read(&os
->os_state
) == ORPHAN_SCAN_INACTIVE
)
313 out
+= snprintf(buf
+ out
, len
- out
, "Disabled\n");
315 out
+= snprintf(buf
+ out
, len
- out
, "%lu seconds ago\n",
316 (get_seconds() - os
->os_scantime
.tv_sec
));
318 out
+= snprintf(buf
+ out
, len
- out
, "%10s => %3s %10s\n",
319 "Slots", "Num", "RecoGen");
320 for (i
= 0; i
< osb
->max_slots
; ++i
) {
321 out
+= snprintf(buf
+ out
, len
- out
,
322 "%10s %c %3d %10d\n",
324 (i
== osb
->slot_num
? '*' : ' '),
325 i
, osb
->slot_recovery_generations
[i
]);
331 static int ocfs2_osb_debug_open(struct inode
*inode
, struct file
*file
)
333 struct ocfs2_super
*osb
= inode
->i_private
;
336 buf
= kmalloc(PAGE_SIZE
, GFP_KERNEL
);
340 i_size_write(inode
, ocfs2_osb_dump(osb
, buf
, PAGE_SIZE
));
342 file
->private_data
= buf
;
349 static int ocfs2_debug_release(struct inode
*inode
, struct file
*file
)
351 kfree(file
->private_data
);
355 static ssize_t
ocfs2_debug_read(struct file
*file
, char __user
*buf
,
356 size_t nbytes
, loff_t
*ppos
)
358 return simple_read_from_buffer(buf
, nbytes
, ppos
, file
->private_data
,
359 i_size_read(file
->f_mapping
->host
));
362 static int ocfs2_osb_debug_open(struct inode
*inode
, struct file
*file
)
366 static int ocfs2_debug_release(struct inode
*inode
, struct file
*file
)
370 static ssize_t
ocfs2_debug_read(struct file
*file
, char __user
*buf
,
371 size_t nbytes
, loff_t
*ppos
)
375 #endif /* CONFIG_DEBUG_FS */
377 static struct file_operations ocfs2_osb_debug_fops
= {
378 .open
= ocfs2_osb_debug_open
,
379 .release
= ocfs2_debug_release
,
380 .read
= ocfs2_debug_read
,
381 .llseek
= generic_file_llseek
,
384 static int ocfs2_sync_fs(struct super_block
*sb
, int wait
)
388 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
390 if (ocfs2_is_hard_readonly(osb
))
394 status
= ocfs2_flush_truncate_log(osb
);
398 ocfs2_schedule_truncate_log_flush(osb
, 0);
401 if (jbd2_journal_start_commit(OCFS2_SB(sb
)->journal
->j_journal
,
404 jbd2_log_wait_commit(OCFS2_SB(sb
)->journal
->j_journal
,
410 static int ocfs2_need_system_inode(struct ocfs2_super
*osb
, int ino
)
412 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb
->sb
, OCFS2_FEATURE_RO_COMPAT_USRQUOTA
)
413 && (ino
== USER_QUOTA_SYSTEM_INODE
414 || ino
== LOCAL_USER_QUOTA_SYSTEM_INODE
))
416 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb
->sb
, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
)
417 && (ino
== GROUP_QUOTA_SYSTEM_INODE
418 || ino
== LOCAL_GROUP_QUOTA_SYSTEM_INODE
))
423 static int ocfs2_init_global_system_inodes(struct ocfs2_super
*osb
)
425 struct inode
*new = NULL
;
431 new = ocfs2_iget(osb
, osb
->root_blkno
, OCFS2_FI_FLAG_SYSFILE
, 0);
433 status
= PTR_ERR(new);
437 osb
->root_inode
= new;
439 new = ocfs2_iget(osb
, osb
->system_dir_blkno
, OCFS2_FI_FLAG_SYSFILE
, 0);
441 status
= PTR_ERR(new);
445 osb
->sys_root_inode
= new;
447 for (i
= OCFS2_FIRST_ONLINE_SYSTEM_INODE
;
448 i
<= OCFS2_LAST_GLOBAL_SYSTEM_INODE
; i
++) {
449 if (!ocfs2_need_system_inode(osb
, i
))
451 new = ocfs2_get_system_file_inode(osb
, i
, osb
->slot_num
);
453 ocfs2_release_system_inodes(osb
);
456 /* FIXME: Should ERROR_RO_FS */
457 mlog(ML_ERROR
, "Unable to load system inode %d, "
458 "possibly corrupt fs?", i
);
461 // the array now has one ref, so drop this one
470 static int ocfs2_init_local_system_inodes(struct ocfs2_super
*osb
)
472 struct inode
*new = NULL
;
478 for (i
= OCFS2_LAST_GLOBAL_SYSTEM_INODE
+ 1;
479 i
< NUM_SYSTEM_INODES
;
481 if (!ocfs2_need_system_inode(osb
, i
))
483 new = ocfs2_get_system_file_inode(osb
, i
, osb
->slot_num
);
485 ocfs2_release_system_inodes(osb
);
487 mlog(ML_ERROR
, "status=%d, sysfile=%d, slot=%d\n",
488 status
, i
, osb
->slot_num
);
491 /* the array now has one ref, so drop this one */
500 static void ocfs2_release_system_inodes(struct ocfs2_super
*osb
)
507 for (i
= 0; i
< NUM_SYSTEM_INODES
; i
++) {
508 inode
= osb
->system_inodes
[i
];
511 osb
->system_inodes
[i
] = NULL
;
515 inode
= osb
->sys_root_inode
;
518 osb
->sys_root_inode
= NULL
;
521 inode
= osb
->root_inode
;
524 osb
->root_inode
= NULL
;
530 /* We're allocating fs objects, use GFP_NOFS */
531 static struct inode
*ocfs2_alloc_inode(struct super_block
*sb
)
533 struct ocfs2_inode_info
*oi
;
535 oi
= kmem_cache_alloc(ocfs2_inode_cachep
, GFP_NOFS
);
539 jbd2_journal_init_jbd_inode(&oi
->ip_jinode
, &oi
->vfs_inode
);
540 return &oi
->vfs_inode
;
543 static void ocfs2_destroy_inode(struct inode
*inode
)
545 kmem_cache_free(ocfs2_inode_cachep
, OCFS2_I(inode
));
548 static unsigned long long ocfs2_max_file_offset(unsigned int bbits
,
551 unsigned int bytes
= 1 << cbits
;
552 unsigned int trim
= bytes
;
553 unsigned int bitshift
= 32;
556 * i_size and all block offsets in ocfs2 are always 64 bits
557 * wide. i_clusters is 32 bits, in cluster-sized units. So on
558 * 64 bit platforms, cluster size will be the limiting factor.
561 #if BITS_PER_LONG == 32
562 # if defined(CONFIG_LBDAF)
563 BUILD_BUG_ON(sizeof(sector_t
) != 8);
565 * We might be limited by page cache size.
567 if (bytes
> PAGE_CACHE_SIZE
) {
568 bytes
= PAGE_CACHE_SIZE
;
571 * Shift by 31 here so that we don't get larger than
578 * We are limited by the size of sector_t. Use block size, as
579 * that's what we expose to the VFS.
588 * Trim by a whole cluster when we can actually approach the
589 * on-disk limits. Otherwise we can overflow i_clusters when
590 * an extent start is at the max offset.
592 return (((unsigned long long)bytes
) << bitshift
) - trim
;
595 static int ocfs2_remount(struct super_block
*sb
, int *flags
, char *data
)
597 int incompat_features
;
599 struct mount_options parsed_options
;
600 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
604 if (!ocfs2_parse_options(sb
, data
, &parsed_options
, 1)) {
609 if ((osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) !=
610 (parsed_options
.mount_opt
& OCFS2_MOUNT_HB_LOCAL
)) {
612 mlog(ML_ERROR
, "Cannot change heartbeat mode on remount\n");
616 if ((osb
->s_mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
) !=
617 (parsed_options
.mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
)) {
619 mlog(ML_ERROR
, "Cannot change data mode on remount\n");
623 /* Probably don't want this on remount; it might
624 * mess with other nodes */
625 if (!(osb
->s_mount_opt
& OCFS2_MOUNT_INODE64
) &&
626 (parsed_options
.mount_opt
& OCFS2_MOUNT_INODE64
)) {
628 mlog(ML_ERROR
, "Cannot enable inode64 on remount\n");
632 /* We're going to/from readonly mode. */
633 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
)) {
634 /* Disable quota accounting before remounting RO */
635 if (*flags
& MS_RDONLY
) {
636 ret
= ocfs2_susp_quotas(osb
, 0);
640 /* Lock here so the check of HARD_RO and the potential
641 * setting of SOFT_RO is atomic. */
642 spin_lock(&osb
->osb_lock
);
643 if (osb
->osb_flags
& OCFS2_OSB_HARD_RO
) {
644 mlog(ML_ERROR
, "Remount on readonly device is forbidden.\n");
649 if (*flags
& MS_RDONLY
) {
650 mlog(0, "Going to ro mode.\n");
651 sb
->s_flags
|= MS_RDONLY
;
652 osb
->osb_flags
|= OCFS2_OSB_SOFT_RO
;
654 mlog(0, "Making ro filesystem writeable.\n");
656 if (osb
->osb_flags
& OCFS2_OSB_ERROR_FS
) {
657 mlog(ML_ERROR
, "Cannot remount RDWR "
658 "filesystem due to previous errors.\n");
662 incompat_features
= OCFS2_HAS_RO_COMPAT_FEATURE(sb
, ~OCFS2_FEATURE_RO_COMPAT_SUPP
);
663 if (incompat_features
) {
664 mlog(ML_ERROR
, "Cannot remount RDWR because "
665 "of unsupported optional features "
666 "(%x).\n", incompat_features
);
670 sb
->s_flags
&= ~MS_RDONLY
;
671 osb
->osb_flags
&= ~OCFS2_OSB_SOFT_RO
;
674 spin_unlock(&osb
->osb_lock
);
675 /* Enable quota accounting after remounting RW */
676 if (!ret
&& !(*flags
& MS_RDONLY
)) {
677 if (sb_any_quota_suspended(sb
))
678 ret
= ocfs2_susp_quotas(osb
, 1);
680 ret
= ocfs2_enable_quotas(osb
);
682 /* Return back changes... */
683 spin_lock(&osb
->osb_lock
);
684 sb
->s_flags
|= MS_RDONLY
;
685 osb
->osb_flags
|= OCFS2_OSB_SOFT_RO
;
686 spin_unlock(&osb
->osb_lock
);
693 /* Only save off the new mount options in case of a successful
695 if (!(osb
->s_feature_incompat
& OCFS2_FEATURE_INCOMPAT_XATTR
))
696 parsed_options
.mount_opt
&= ~OCFS2_MOUNT_POSIX_ACL
;
697 osb
->s_mount_opt
= parsed_options
.mount_opt
;
698 osb
->s_atime_quantum
= parsed_options
.atime_quantum
;
699 osb
->preferred_slot
= parsed_options
.slot
;
700 if (parsed_options
.commit_interval
)
701 osb
->osb_commit_interval
= parsed_options
.commit_interval
;
703 if (!ocfs2_is_hard_readonly(osb
))
704 ocfs2_set_journal_params(osb
);
711 static int ocfs2_sb_probe(struct super_block
*sb
,
712 struct buffer_head
**bh
,
714 struct ocfs2_blockcheck_stats
*stats
)
717 struct ocfs1_vol_disk_hdr
*hdr
;
718 struct ocfs2_dinode
*di
;
724 *sector_size
= bdev_logical_block_size(sb
->s_bdev
);
725 if (*sector_size
> OCFS2_MAX_BLOCKSIZE
) {
726 mlog(ML_ERROR
, "Hardware sector size too large: %d (max=%d)\n",
727 *sector_size
, OCFS2_MAX_BLOCKSIZE
);
732 /* Can this really happen? */
733 if (*sector_size
< OCFS2_MIN_BLOCKSIZE
)
734 *sector_size
= OCFS2_MIN_BLOCKSIZE
;
736 /* check block zero for old format */
737 status
= ocfs2_get_sector(sb
, bh
, 0, *sector_size
);
742 hdr
= (struct ocfs1_vol_disk_hdr
*) (*bh
)->b_data
;
743 if (hdr
->major_version
== OCFS1_MAJOR_VERSION
) {
744 mlog(ML_ERROR
, "incompatible version: %u.%u\n",
745 hdr
->major_version
, hdr
->minor_version
);
748 if (memcmp(hdr
->signature
, OCFS1_VOLUME_SIGNATURE
,
749 strlen(OCFS1_VOLUME_SIGNATURE
)) == 0) {
750 mlog(ML_ERROR
, "incompatible volume signature: %8s\n",
757 mlog(ML_ERROR
, "This is an ocfs v1 filesystem which must be "
758 "upgraded before mounting with ocfs v2\n");
763 * Now check at magic offset for 512, 1024, 2048, 4096
764 * blocksizes. 4096 is the maximum blocksize because it is
765 * the minimum clustersize.
768 for (blksize
= *sector_size
;
769 blksize
<= OCFS2_MAX_BLOCKSIZE
;
771 tmpstat
= ocfs2_get_sector(sb
, bh
,
772 OCFS2_SUPER_BLOCK_BLKNO
,
779 di
= (struct ocfs2_dinode
*) (*bh
)->b_data
;
780 memset(stats
, 0, sizeof(struct ocfs2_blockcheck_stats
));
781 spin_lock_init(&stats
->b_lock
);
782 status
= ocfs2_verify_volume(di
, *bh
, blksize
, stats
);
787 if (status
!= -EAGAIN
)
795 static int ocfs2_verify_heartbeat(struct ocfs2_super
*osb
)
797 if (ocfs2_mount_local(osb
)) {
798 if (osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) {
799 mlog(ML_ERROR
, "Cannot heartbeat on a locally "
800 "mounted device.\n");
805 if (ocfs2_userspace_stack(osb
)) {
806 if (osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) {
807 mlog(ML_ERROR
, "Userspace stack expected, but "
808 "o2cb heartbeat arguments passed to mount\n");
813 if (!(osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
)) {
814 if (!ocfs2_mount_local(osb
) && !ocfs2_is_hard_readonly(osb
) &&
815 !ocfs2_userspace_stack(osb
)) {
816 mlog(ML_ERROR
, "Heartbeat has to be started to mount "
817 "a read-write clustered device.\n");
826 * If we're using a userspace stack, mount should have passed
827 * a name that matches the disk. If not, mount should not
828 * have passed a stack.
830 static int ocfs2_verify_userspace_stack(struct ocfs2_super
*osb
,
831 struct mount_options
*mopt
)
833 if (!ocfs2_userspace_stack(osb
) && mopt
->cluster_stack
[0]) {
835 "cluster stack passed to mount, but this filesystem "
836 "does not support it\n");
840 if (ocfs2_userspace_stack(osb
) &&
841 strncmp(osb
->osb_cluster_stack
, mopt
->cluster_stack
,
842 OCFS2_STACK_LABEL_LEN
)) {
844 "cluster stack passed to mount (\"%s\") does not "
845 "match the filesystem (\"%s\")\n",
847 osb
->osb_cluster_stack
);
854 static int ocfs2_susp_quotas(struct ocfs2_super
*osb
, int unsuspend
)
857 struct super_block
*sb
= osb
->sb
;
858 unsigned int feature
[MAXQUOTAS
] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA
,
859 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
};
862 for (type
= 0; type
< MAXQUOTAS
; type
++) {
863 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb
, feature
[type
]))
866 status
= vfs_quota_enable(
867 sb_dqopt(sb
)->files
[type
],
871 status
= vfs_quota_disable(sb
, type
,
877 mlog(ML_ERROR
, "Failed to suspend/unsuspend quotas on "
878 "remount (error = %d).\n", status
);
882 static int ocfs2_enable_quotas(struct ocfs2_super
*osb
)
884 struct inode
*inode
[MAXQUOTAS
] = { NULL
, NULL
};
885 struct super_block
*sb
= osb
->sb
;
886 unsigned int feature
[MAXQUOTAS
] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA
,
887 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
};
888 unsigned int ino
[MAXQUOTAS
] = { LOCAL_USER_QUOTA_SYSTEM_INODE
,
889 LOCAL_GROUP_QUOTA_SYSTEM_INODE
};
893 sb_dqopt(sb
)->flags
|= DQUOT_QUOTA_SYS_FILE
| DQUOT_NEGATIVE_USAGE
;
894 for (type
= 0; type
< MAXQUOTAS
; type
++) {
895 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb
, feature
[type
]))
897 inode
[type
] = ocfs2_get_system_file_inode(osb
, ino
[type
],
903 status
= vfs_quota_enable(inode
[type
], type
, QFMT_OCFS2
,
904 DQUOT_USAGE_ENABLED
);
909 for (type
= 0; type
< MAXQUOTAS
; type
++)
913 ocfs2_disable_quotas(osb
);
914 for (type
= 0; type
< MAXQUOTAS
; type
++)
920 static void ocfs2_disable_quotas(struct ocfs2_super
*osb
)
924 struct super_block
*sb
= osb
->sb
;
926 /* We mostly ignore errors in this function because there's not much
927 * we can do when we see them */
928 for (type
= 0; type
< MAXQUOTAS
; type
++) {
929 if (!sb_has_quota_loaded(sb
, type
))
931 inode
= igrab(sb
->s_dquot
.files
[type
]);
932 /* Turn off quotas. This will remove all dquot structures from
933 * memory and so they will be automatically synced to global
935 vfs_quota_disable(sb
, type
, DQUOT_USAGE_ENABLED
|
936 DQUOT_LIMITS_ENABLED
);
943 /* Handle quota on quotactl */
944 static int ocfs2_quota_on(struct super_block
*sb
, int type
, int format_id
,
945 char *path
, int remount
)
947 unsigned int feature
[MAXQUOTAS
] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA
,
948 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
};
950 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb
, feature
[type
]))
954 return 0; /* Just ignore it has been handled in
956 return vfs_quota_enable(sb_dqopt(sb
)->files
[type
], type
,
957 format_id
, DQUOT_LIMITS_ENABLED
);
960 /* Handle quota off quotactl */
961 static int ocfs2_quota_off(struct super_block
*sb
, int type
, int remount
)
964 return 0; /* Ignore now and handle later in
966 return vfs_quota_disable(sb
, type
, DQUOT_LIMITS_ENABLED
);
969 static struct quotactl_ops ocfs2_quotactl_ops
= {
970 .quota_on
= ocfs2_quota_on
,
971 .quota_off
= ocfs2_quota_off
,
972 .quota_sync
= vfs_quota_sync
,
973 .get_info
= vfs_get_dqinfo
,
974 .set_info
= vfs_set_dqinfo
,
975 .get_dqblk
= vfs_get_dqblk
,
976 .set_dqblk
= vfs_set_dqblk
,
979 static int ocfs2_fill_super(struct super_block
*sb
, void *data
, int silent
)
982 int status
, sector_size
;
983 struct mount_options parsed_options
;
984 struct inode
*inode
= NULL
;
985 struct ocfs2_super
*osb
= NULL
;
986 struct buffer_head
*bh
= NULL
;
988 struct ocfs2_blockcheck_stats stats
;
990 mlog_entry("%p, %p, %i", sb
, data
, silent
);
992 if (!ocfs2_parse_options(sb
, data
, &parsed_options
, 0)) {
994 goto read_super_error
;
997 /* probe for superblock */
998 status
= ocfs2_sb_probe(sb
, &bh
, §or_size
, &stats
);
1000 mlog(ML_ERROR
, "superblock probe failed!\n");
1001 goto read_super_error
;
1004 status
= ocfs2_initialize_super(sb
, bh
, sector_size
, &stats
);
1008 goto read_super_error
;
1013 if (!(osb
->s_feature_incompat
& OCFS2_FEATURE_INCOMPAT_XATTR
))
1014 parsed_options
.mount_opt
&= ~OCFS2_MOUNT_POSIX_ACL
;
1016 osb
->s_mount_opt
= parsed_options
.mount_opt
;
1017 osb
->s_atime_quantum
= parsed_options
.atime_quantum
;
1018 osb
->preferred_slot
= parsed_options
.slot
;
1019 osb
->osb_commit_interval
= parsed_options
.commit_interval
;
1020 osb
->local_alloc_default_bits
= ocfs2_megabytes_to_clusters(sb
, parsed_options
.localalloc_opt
);
1021 osb
->local_alloc_bits
= osb
->local_alloc_default_bits
;
1022 if (osb
->s_mount_opt
& OCFS2_MOUNT_USRQUOTA
&&
1023 !OCFS2_HAS_RO_COMPAT_FEATURE(sb
,
1024 OCFS2_FEATURE_RO_COMPAT_USRQUOTA
)) {
1026 mlog(ML_ERROR
, "User quotas were requested, but this "
1027 "filesystem does not have the feature enabled.\n");
1028 goto read_super_error
;
1030 if (osb
->s_mount_opt
& OCFS2_MOUNT_GRPQUOTA
&&
1031 !OCFS2_HAS_RO_COMPAT_FEATURE(sb
,
1032 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
)) {
1034 mlog(ML_ERROR
, "Group quotas were requested, but this "
1035 "filesystem does not have the feature enabled.\n");
1036 goto read_super_error
;
1039 status
= ocfs2_verify_userspace_stack(osb
, &parsed_options
);
1041 goto read_super_error
;
1043 sb
->s_magic
= OCFS2_SUPER_MAGIC
;
1045 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
1046 ((osb
->s_mount_opt
& OCFS2_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
1048 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1050 if (bdev_read_only(sb
->s_bdev
)) {
1051 if (!(sb
->s_flags
& MS_RDONLY
)) {
1053 mlog(ML_ERROR
, "Readonly device detected but readonly "
1054 "mount was not specified.\n");
1055 goto read_super_error
;
1058 /* You should not be able to start a local heartbeat
1059 * on a readonly device. */
1060 if (osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) {
1062 mlog(ML_ERROR
, "Local heartbeat specified on readonly "
1064 goto read_super_error
;
1067 status
= ocfs2_check_journals_nolocks(osb
);
1069 if (status
== -EROFS
)
1070 mlog(ML_ERROR
, "Recovery required on readonly "
1071 "file system, but write access is "
1075 goto read_super_error
;
1078 ocfs2_set_ro_flag(osb
, 1);
1080 printk(KERN_NOTICE
"Readonly device detected. No cluster "
1081 "services will be utilized for this mount. Recovery "
1082 "will be skipped.\n");
1085 if (!ocfs2_is_hard_readonly(osb
)) {
1086 if (sb
->s_flags
& MS_RDONLY
)
1087 ocfs2_set_ro_flag(osb
, 0);
1090 status
= ocfs2_verify_heartbeat(osb
);
1093 goto read_super_error
;
1096 osb
->osb_debug_root
= debugfs_create_dir(osb
->uuid_str
,
1097 ocfs2_debugfs_root
);
1098 if (!osb
->osb_debug_root
) {
1100 mlog(ML_ERROR
, "Unable to create per-mount debugfs root.\n");
1101 goto read_super_error
;
1104 osb
->osb_ctxt
= debugfs_create_file("fs_state", S_IFREG
|S_IRUSR
,
1105 osb
->osb_debug_root
,
1107 &ocfs2_osb_debug_fops
);
1108 if (!osb
->osb_ctxt
) {
1111 goto read_super_error
;
1114 if (ocfs2_meta_ecc(osb
)) {
1115 status
= ocfs2_blockcheck_stats_debugfs_install(
1116 &osb
->osb_ecc_stats
,
1117 osb
->osb_debug_root
);
1120 "Unable to create blockcheck statistics "
1122 goto read_super_error
;
1126 status
= ocfs2_mount_volume(sb
);
1127 if (osb
->root_inode
)
1128 inode
= igrab(osb
->root_inode
);
1131 goto read_super_error
;
1136 goto read_super_error
;
1139 root
= d_alloc_root(inode
);
1143 goto read_super_error
;
1148 ocfs2_complete_mount_recovery(osb
);
1150 if (ocfs2_mount_local(osb
))
1151 snprintf(nodestr
, sizeof(nodestr
), "local");
1153 snprintf(nodestr
, sizeof(nodestr
), "%u", osb
->node_num
);
1155 printk(KERN_INFO
"ocfs2: Mounting device (%s) on (node %s, slot %d) "
1156 "with %s data mode.\n",
1157 osb
->dev_str
, nodestr
, osb
->slot_num
,
1158 osb
->s_mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
? "writeback" :
1161 atomic_set(&osb
->vol_state
, VOLUME_MOUNTED
);
1162 wake_up(&osb
->osb_mount_event
);
1164 /* Now we can initialize quotas because we can afford to wait
1165 * for cluster locks recovery now. That also means that truncation
1166 * log recovery can happen but that waits for proper quota setup */
1167 if (!(sb
->s_flags
& MS_RDONLY
)) {
1168 status
= ocfs2_enable_quotas(osb
);
1170 /* We have to err-out specially here because
1171 * s_root is already set */
1173 atomic_set(&osb
->vol_state
, VOLUME_DISABLED
);
1174 wake_up(&osb
->osb_mount_event
);
1180 ocfs2_complete_quota_recovery(osb
);
1182 /* Now we wake up again for processes waiting for quotas */
1183 atomic_set(&osb
->vol_state
, VOLUME_MOUNTED_QUOTAS
);
1184 wake_up(&osb
->osb_mount_event
);
1186 /* Start this when the mount is almost sure of being successful */
1187 ocfs2_orphan_scan_start(osb
);
1199 atomic_set(&osb
->vol_state
, VOLUME_DISABLED
);
1200 wake_up(&osb
->osb_mount_event
);
1201 ocfs2_dismount_volume(sb
, 1);
1208 static int ocfs2_get_sb(struct file_system_type
*fs_type
,
1210 const char *dev_name
,
1212 struct vfsmount
*mnt
)
1214 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, ocfs2_fill_super
,
1218 static void ocfs2_kill_sb(struct super_block
*sb
)
1220 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
1223 if (!osb
|| atomic_read(&osb
->vol_state
) == VOLUME_DISABLED
)
1226 /* Prevent further queueing of inode drop events */
1227 spin_lock(&dentry_list_lock
);
1228 ocfs2_set_osb_flag(osb
, OCFS2_OSB_DROP_DENTRY_LOCK_IMMED
);
1229 spin_unlock(&dentry_list_lock
);
1230 /* Wait for work to finish and/or remove it */
1231 cancel_work_sync(&osb
->dentry_lock_work
);
1233 kill_block_super(sb
);
1236 static struct file_system_type ocfs2_fs_type
= {
1237 .owner
= THIS_MODULE
,
1239 .get_sb
= ocfs2_get_sb
, /* is this called when we mount
1241 .kill_sb
= ocfs2_kill_sb
,
1243 .fs_flags
= FS_REQUIRES_DEV
|FS_RENAME_DOES_D_MOVE
,
1247 static int ocfs2_parse_options(struct super_block
*sb
,
1249 struct mount_options
*mopt
,
1255 mlog_entry("remount: %d, options: \"%s\"\n", is_remount
,
1256 options
? options
: "(none)");
1258 mopt
->commit_interval
= 0;
1259 mopt
->mount_opt
= 0;
1260 mopt
->atime_quantum
= OCFS2_DEFAULT_ATIME_QUANTUM
;
1261 mopt
->slot
= OCFS2_INVALID_SLOT
;
1262 mopt
->localalloc_opt
= OCFS2_DEFAULT_LOCAL_ALLOC_SIZE
;
1263 mopt
->cluster_stack
[0] = '\0';
1270 while ((p
= strsep(&options
, ",")) != NULL
) {
1272 substring_t args
[MAX_OPT_ARGS
];
1277 token
= match_token(p
, tokens
, args
);
1280 mopt
->mount_opt
|= OCFS2_MOUNT_HB_LOCAL
;
1283 mopt
->mount_opt
&= ~OCFS2_MOUNT_HB_LOCAL
;
1286 if (match_int(&args
[0], &option
)) {
1291 mopt
->mount_opt
|= OCFS2_MOUNT_BARRIER
;
1293 mopt
->mount_opt
&= ~OCFS2_MOUNT_BARRIER
;
1296 mopt
->mount_opt
&= ~OCFS2_MOUNT_NOINTR
;
1299 mopt
->mount_opt
|= OCFS2_MOUNT_NOINTR
;
1302 mopt
->mount_opt
|= OCFS2_MOUNT_ERRORS_PANIC
;
1305 mopt
->mount_opt
&= ~OCFS2_MOUNT_ERRORS_PANIC
;
1307 case Opt_data_ordered
:
1308 mopt
->mount_opt
&= ~OCFS2_MOUNT_DATA_WRITEBACK
;
1310 case Opt_data_writeback
:
1311 mopt
->mount_opt
|= OCFS2_MOUNT_DATA_WRITEBACK
;
1313 case Opt_user_xattr
:
1314 mopt
->mount_opt
&= ~OCFS2_MOUNT_NOUSERXATTR
;
1316 case Opt_nouser_xattr
:
1317 mopt
->mount_opt
|= OCFS2_MOUNT_NOUSERXATTR
;
1319 case Opt_atime_quantum
:
1320 if (match_int(&args
[0], &option
)) {
1325 mopt
->atime_quantum
= option
;
1329 if (match_int(&args
[0], &option
)) {
1334 mopt
->slot
= (s16
)option
;
1338 if (match_int(&args
[0], &option
)) {
1345 option
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1346 mopt
->commit_interval
= HZ
* option
;
1348 case Opt_localalloc
:
1350 if (match_int(&args
[0], &option
)) {
1354 if (option
>= 0 && (option
<= ocfs2_local_alloc_size(sb
) * 8))
1355 mopt
->localalloc_opt
= option
;
1357 case Opt_localflocks
:
1359 * Changing this during remount could race
1360 * flock() requests, or "unbalance" existing
1361 * ones (e.g., a lock is taken in one mode but
1362 * dropped in the other). If users care enough
1363 * to flip locking modes during remount, we
1364 * could add a "local" flag to individual
1365 * flock structures for proper tracking of
1369 mopt
->mount_opt
|= OCFS2_MOUNT_LOCALFLOCKS
;
1372 /* Check both that the option we were passed
1373 * is of the right length and that it is a proper
1374 * string of the right length.
1376 if (((args
[0].to
- args
[0].from
) !=
1377 OCFS2_STACK_LABEL_LEN
) ||
1378 (strnlen(args
[0].from
,
1379 OCFS2_STACK_LABEL_LEN
) !=
1380 OCFS2_STACK_LABEL_LEN
)) {
1382 "Invalid cluster_stack option\n");
1386 memcpy(mopt
->cluster_stack
, args
[0].from
,
1387 OCFS2_STACK_LABEL_LEN
);
1388 mopt
->cluster_stack
[OCFS2_STACK_LABEL_LEN
] = '\0';
1391 mopt
->mount_opt
|= OCFS2_MOUNT_INODE64
;
1394 /* We check only on remount, otherwise features
1395 * aren't yet initialized. */
1396 if (is_remount
&& !OCFS2_HAS_RO_COMPAT_FEATURE(sb
,
1397 OCFS2_FEATURE_RO_COMPAT_USRQUOTA
)) {
1398 mlog(ML_ERROR
, "User quota requested but "
1399 "filesystem feature is not set\n");
1403 mopt
->mount_opt
|= OCFS2_MOUNT_USRQUOTA
;
1406 if (is_remount
&& !OCFS2_HAS_RO_COMPAT_FEATURE(sb
,
1407 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
)) {
1408 mlog(ML_ERROR
, "Group quota requested but "
1409 "filesystem feature is not set\n");
1413 mopt
->mount_opt
|= OCFS2_MOUNT_GRPQUOTA
;
1415 #ifdef CONFIG_OCFS2_FS_POSIX_ACL
1417 mopt
->mount_opt
|= OCFS2_MOUNT_POSIX_ACL
;
1420 mopt
->mount_opt
&= ~OCFS2_MOUNT_POSIX_ACL
;
1425 printk(KERN_INFO
"ocfs2 (no)acl options not supported\n");
1430 "Unrecognized mount option \"%s\" "
1431 "or missing value\n", p
);
1444 static int ocfs2_show_options(struct seq_file
*s
, struct vfsmount
*mnt
)
1446 struct ocfs2_super
*osb
= OCFS2_SB(mnt
->mnt_sb
);
1447 unsigned long opts
= osb
->s_mount_opt
;
1448 unsigned int local_alloc_megs
;
1450 if (opts
& OCFS2_MOUNT_HB_LOCAL
)
1451 seq_printf(s
, ",_netdev,heartbeat=local");
1453 seq_printf(s
, ",heartbeat=none");
1455 if (opts
& OCFS2_MOUNT_NOINTR
)
1456 seq_printf(s
, ",nointr");
1458 if (opts
& OCFS2_MOUNT_DATA_WRITEBACK
)
1459 seq_printf(s
, ",data=writeback");
1461 seq_printf(s
, ",data=ordered");
1463 if (opts
& OCFS2_MOUNT_BARRIER
)
1464 seq_printf(s
, ",barrier=1");
1466 if (opts
& OCFS2_MOUNT_ERRORS_PANIC
)
1467 seq_printf(s
, ",errors=panic");
1469 seq_printf(s
, ",errors=remount-ro");
1471 if (osb
->preferred_slot
!= OCFS2_INVALID_SLOT
)
1472 seq_printf(s
, ",preferred_slot=%d", osb
->preferred_slot
);
1474 if (osb
->s_atime_quantum
!= OCFS2_DEFAULT_ATIME_QUANTUM
)
1475 seq_printf(s
, ",atime_quantum=%u", osb
->s_atime_quantum
);
1477 if (osb
->osb_commit_interval
)
1478 seq_printf(s
, ",commit=%u",
1479 (unsigned) (osb
->osb_commit_interval
/ HZ
));
1481 local_alloc_megs
= osb
->local_alloc_bits
>> (20 - osb
->s_clustersize_bits
);
1482 if (local_alloc_megs
!= OCFS2_DEFAULT_LOCAL_ALLOC_SIZE
)
1483 seq_printf(s
, ",localalloc=%d", local_alloc_megs
);
1485 if (opts
& OCFS2_MOUNT_LOCALFLOCKS
)
1486 seq_printf(s
, ",localflocks,");
1488 if (osb
->osb_cluster_stack
[0])
1489 seq_printf(s
, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN
,
1490 osb
->osb_cluster_stack
);
1491 if (opts
& OCFS2_MOUNT_USRQUOTA
)
1492 seq_printf(s
, ",usrquota");
1493 if (opts
& OCFS2_MOUNT_GRPQUOTA
)
1494 seq_printf(s
, ",grpquota");
1496 if (opts
& OCFS2_MOUNT_NOUSERXATTR
)
1497 seq_printf(s
, ",nouser_xattr");
1499 seq_printf(s
, ",user_xattr");
1501 if (opts
& OCFS2_MOUNT_INODE64
)
1502 seq_printf(s
, ",inode64");
1504 #ifdef CONFIG_OCFS2_FS_POSIX_ACL
1505 if (opts
& OCFS2_MOUNT_POSIX_ACL
)
1506 seq_printf(s
, ",acl");
1508 seq_printf(s
, ",noacl");
1514 static int __init
ocfs2_init(void)
1520 ocfs2_print_version();
1522 status
= init_ocfs2_uptodate_cache();
1528 status
= ocfs2_initialize_mem_caches();
1534 ocfs2_wq
= create_singlethread_workqueue("ocfs2_wq");
1540 ocfs2_debugfs_root
= debugfs_create_dir("ocfs2", NULL
);
1541 if (!ocfs2_debugfs_root
) {
1543 mlog(ML_ERROR
, "Unable to create ocfs2 debugfs root.\n");
1546 status
= ocfs2_quota_setup();
1550 ocfs2_set_locking_protocol();
1552 status
= register_quota_format(&ocfs2_quota_format
);
1555 ocfs2_quota_shutdown();
1556 ocfs2_free_mem_caches();
1557 exit_ocfs2_uptodate_cache();
1563 return register_filesystem(&ocfs2_fs_type
);
1568 static void __exit
ocfs2_exit(void)
1572 ocfs2_quota_shutdown();
1575 flush_workqueue(ocfs2_wq
);
1576 destroy_workqueue(ocfs2_wq
);
1579 unregister_quota_format(&ocfs2_quota_format
);
1581 debugfs_remove(ocfs2_debugfs_root
);
1583 ocfs2_free_mem_caches();
1585 unregister_filesystem(&ocfs2_fs_type
);
1587 exit_ocfs2_uptodate_cache();
1592 static void ocfs2_put_super(struct super_block
*sb
)
1594 mlog_entry("(0x%p)\n", sb
);
1598 ocfs2_sync_blockdev(sb
);
1599 ocfs2_dismount_volume(sb
, 0);
1606 static int ocfs2_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
1608 struct ocfs2_super
*osb
;
1609 u32 numbits
, freebits
;
1611 struct ocfs2_dinode
*bm_lock
;
1612 struct buffer_head
*bh
= NULL
;
1613 struct inode
*inode
= NULL
;
1615 mlog_entry("(%p, %p)\n", dentry
->d_sb
, buf
);
1617 osb
= OCFS2_SB(dentry
->d_sb
);
1619 inode
= ocfs2_get_system_file_inode(osb
,
1620 GLOBAL_BITMAP_SYSTEM_INODE
,
1621 OCFS2_INVALID_SLOT
);
1623 mlog(ML_ERROR
, "failed to get bitmap inode\n");
1628 status
= ocfs2_inode_lock(inode
, &bh
, 0);
1634 bm_lock
= (struct ocfs2_dinode
*) bh
->b_data
;
1636 numbits
= le32_to_cpu(bm_lock
->id1
.bitmap1
.i_total
);
1637 freebits
= numbits
- le32_to_cpu(bm_lock
->id1
.bitmap1
.i_used
);
1639 buf
->f_type
= OCFS2_SUPER_MAGIC
;
1640 buf
->f_bsize
= dentry
->d_sb
->s_blocksize
;
1641 buf
->f_namelen
= OCFS2_MAX_FILENAME_LEN
;
1642 buf
->f_blocks
= ((sector_t
) numbits
) *
1643 (osb
->s_clustersize
>> osb
->sb
->s_blocksize_bits
);
1644 buf
->f_bfree
= ((sector_t
) freebits
) *
1645 (osb
->s_clustersize
>> osb
->sb
->s_blocksize_bits
);
1646 buf
->f_bavail
= buf
->f_bfree
;
1647 buf
->f_files
= numbits
;
1648 buf
->f_ffree
= freebits
;
1652 ocfs2_inode_unlock(inode
, 0);
1663 static void ocfs2_inode_init_once(void *data
)
1665 struct ocfs2_inode_info
*oi
= data
;
1668 oi
->ip_open_count
= 0;
1669 spin_lock_init(&oi
->ip_lock
);
1670 ocfs2_extent_map_init(&oi
->vfs_inode
);
1671 INIT_LIST_HEAD(&oi
->ip_io_markers
);
1672 oi
->ip_dir_start_lookup
= 0;
1674 init_rwsem(&oi
->ip_alloc_sem
);
1675 init_rwsem(&oi
->ip_xattr_sem
);
1676 mutex_init(&oi
->ip_io_mutex
);
1678 oi
->ip_blkno
= 0ULL;
1679 oi
->ip_clusters
= 0;
1681 ocfs2_lock_res_init_once(&oi
->ip_rw_lockres
);
1682 ocfs2_lock_res_init_once(&oi
->ip_inode_lockres
);
1683 ocfs2_lock_res_init_once(&oi
->ip_open_lockres
);
1685 ocfs2_metadata_cache_init(INODE_CACHE(&oi
->vfs_inode
),
1686 &ocfs2_inode_caching_ops
);
1688 inode_init_once(&oi
->vfs_inode
);
1691 static int ocfs2_initialize_mem_caches(void)
1693 ocfs2_inode_cachep
= kmem_cache_create("ocfs2_inode_cache",
1694 sizeof(struct ocfs2_inode_info
),
1696 (SLAB_HWCACHE_ALIGN
|SLAB_RECLAIM_ACCOUNT
|
1698 ocfs2_inode_init_once
);
1699 ocfs2_dquot_cachep
= kmem_cache_create("ocfs2_dquot_cache",
1700 sizeof(struct ocfs2_dquot
),
1702 (SLAB_HWCACHE_ALIGN
|SLAB_RECLAIM_ACCOUNT
|
1705 ocfs2_qf_chunk_cachep
= kmem_cache_create("ocfs2_qf_chunk_cache",
1706 sizeof(struct ocfs2_quota_chunk
),
1708 (SLAB_RECLAIM_ACCOUNT
|SLAB_MEM_SPREAD
),
1710 if (!ocfs2_inode_cachep
|| !ocfs2_dquot_cachep
||
1711 !ocfs2_qf_chunk_cachep
) {
1712 if (ocfs2_inode_cachep
)
1713 kmem_cache_destroy(ocfs2_inode_cachep
);
1714 if (ocfs2_dquot_cachep
)
1715 kmem_cache_destroy(ocfs2_dquot_cachep
);
1716 if (ocfs2_qf_chunk_cachep
)
1717 kmem_cache_destroy(ocfs2_qf_chunk_cachep
);
1724 static void ocfs2_free_mem_caches(void)
1726 if (ocfs2_inode_cachep
)
1727 kmem_cache_destroy(ocfs2_inode_cachep
);
1728 ocfs2_inode_cachep
= NULL
;
1730 if (ocfs2_dquot_cachep
)
1731 kmem_cache_destroy(ocfs2_dquot_cachep
);
1732 ocfs2_dquot_cachep
= NULL
;
1734 if (ocfs2_qf_chunk_cachep
)
1735 kmem_cache_destroy(ocfs2_qf_chunk_cachep
);
1736 ocfs2_qf_chunk_cachep
= NULL
;
1739 static int ocfs2_get_sector(struct super_block
*sb
,
1740 struct buffer_head
**bh
,
1744 if (!sb_set_blocksize(sb
, sect_size
)) {
1745 mlog(ML_ERROR
, "unable to set blocksize\n");
1749 *bh
= sb_getblk(sb
, block
);
1755 if (!buffer_dirty(*bh
))
1756 clear_buffer_uptodate(*bh
);
1758 ll_rw_block(READ
, 1, bh
);
1759 wait_on_buffer(*bh
);
1760 if (!buffer_uptodate(*bh
)) {
1770 static int ocfs2_mount_volume(struct super_block
*sb
)
1773 int unlock_super
= 0;
1774 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
1778 if (ocfs2_is_hard_readonly(osb
))
1781 status
= ocfs2_dlm_init(osb
);
1787 status
= ocfs2_super_lock(osb
, 1);
1794 /* This will load up the node map and add ourselves to it. */
1795 status
= ocfs2_find_slot(osb
);
1801 /* load all node-local system inodes */
1802 status
= ocfs2_init_local_system_inodes(osb
);
1808 status
= ocfs2_check_volume(osb
);
1814 status
= ocfs2_truncate_log_init(osb
);
1820 ocfs2_super_unlock(osb
, 1);
1826 static void ocfs2_dismount_volume(struct super_block
*sb
, int mnt_err
)
1828 int tmp
, hangup_needed
= 0;
1829 struct ocfs2_super
*osb
= NULL
;
1832 mlog_entry("(0x%p)\n", sb
);
1838 debugfs_remove(osb
->osb_ctxt
);
1841 * Flush inode dropping work queue so that deletes are
1842 * performed while the filesystem is still working
1844 ocfs2_drop_all_dl_inodes(osb
);
1846 /* Orphan scan should be stopped as early as possible */
1847 ocfs2_orphan_scan_stop(osb
);
1849 ocfs2_disable_quotas(osb
);
1851 ocfs2_shutdown_local_alloc(osb
);
1853 ocfs2_truncate_log_shutdown(osb
);
1855 /* This will disable recovery and flush any recovery work. */
1856 ocfs2_recovery_exit(osb
);
1858 ocfs2_journal_shutdown(osb
);
1860 ocfs2_sync_blockdev(sb
);
1862 ocfs2_purge_refcount_trees(osb
);
1864 /* No cluster connection means we've failed during mount, so skip
1865 * all the steps which depended on that to complete. */
1867 tmp
= ocfs2_super_lock(osb
, 1);
1874 if (osb
->slot_num
!= OCFS2_INVALID_SLOT
)
1875 ocfs2_put_slot(osb
);
1878 ocfs2_super_unlock(osb
, 1);
1880 ocfs2_release_system_inodes(osb
);
1883 * If we're dismounting due to mount error, mount.ocfs2 will clean
1884 * up heartbeat. If we're a local mount, there is no heartbeat.
1885 * If we failed before we got a uuid_str yet, we can't stop
1886 * heartbeat. Otherwise, do it.
1888 if (!mnt_err
&& !ocfs2_mount_local(osb
) && osb
->uuid_str
)
1892 ocfs2_dlm_shutdown(osb
, hangup_needed
);
1894 ocfs2_blockcheck_stats_debugfs_remove(&osb
->osb_ecc_stats
);
1895 debugfs_remove(osb
->osb_debug_root
);
1898 ocfs2_cluster_hangup(osb
->uuid_str
, strlen(osb
->uuid_str
));
1900 atomic_set(&osb
->vol_state
, VOLUME_DISMOUNTED
);
1902 if (ocfs2_mount_local(osb
))
1903 snprintf(nodestr
, sizeof(nodestr
), "local");
1905 snprintf(nodestr
, sizeof(nodestr
), "%u", osb
->node_num
);
1907 printk(KERN_INFO
"ocfs2: Unmounting device (%s) on (node %s)\n",
1908 osb
->dev_str
, nodestr
);
1910 ocfs2_delete_osb(osb
);
1913 sb
->s_fs_info
= NULL
;
1916 static int ocfs2_setup_osb_uuid(struct ocfs2_super
*osb
, const unsigned char *uuid
,
1917 unsigned uuid_bytes
)
1922 BUG_ON(uuid_bytes
!= OCFS2_VOL_UUID_LEN
);
1924 osb
->uuid_str
= kzalloc(OCFS2_VOL_UUID_LEN
* 2 + 1, GFP_KERNEL
);
1925 if (osb
->uuid_str
== NULL
)
1928 for (i
= 0, ptr
= osb
->uuid_str
; i
< OCFS2_VOL_UUID_LEN
; i
++) {
1929 /* print with null */
1930 ret
= snprintf(ptr
, 3, "%02X", uuid
[i
]);
1931 if (ret
!= 2) /* drop super cleans up */
1933 /* then only advance past the last char */
1940 static int ocfs2_initialize_super(struct super_block
*sb
,
1941 struct buffer_head
*bh
,
1943 struct ocfs2_blockcheck_stats
*stats
)
1946 int i
, cbits
, bbits
;
1947 struct ocfs2_dinode
*di
= (struct ocfs2_dinode
*)bh
->b_data
;
1948 struct inode
*inode
= NULL
;
1949 struct ocfs2_journal
*journal
;
1950 __le32 uuid_net_key
;
1951 struct ocfs2_super
*osb
;
1955 osb
= kzalloc(sizeof(struct ocfs2_super
), GFP_KERNEL
);
1962 sb
->s_fs_info
= osb
;
1963 sb
->s_op
= &ocfs2_sops
;
1964 sb
->s_export_op
= &ocfs2_export_ops
;
1965 sb
->s_qcop
= &ocfs2_quotactl_ops
;
1966 sb
->dq_op
= &ocfs2_quota_operations
;
1967 sb
->s_xattr
= ocfs2_xattr_handlers
;
1968 sb
->s_time_gran
= 1;
1969 sb
->s_flags
|= MS_NOATIME
;
1970 /* this is needed to support O_LARGEFILE */
1971 cbits
= le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
);
1972 bbits
= le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
);
1973 sb
->s_maxbytes
= ocfs2_max_file_offset(bbits
, cbits
);
1975 osb
->osb_dx_mask
= (1 << (cbits
- bbits
)) - 1;
1977 for (i
= 0; i
< 3; i
++)
1978 osb
->osb_dx_seed
[i
] = le32_to_cpu(di
->id2
.i_super
.s_dx_seed
[i
]);
1979 osb
->osb_dx_seed
[3] = le32_to_cpu(di
->id2
.i_super
.s_uuid_hash
);
1982 /* Save off for ocfs2_rw_direct */
1983 osb
->s_sectsize_bits
= blksize_bits(sector_size
);
1984 BUG_ON(!osb
->s_sectsize_bits
);
1986 spin_lock_init(&osb
->dc_task_lock
);
1987 init_waitqueue_head(&osb
->dc_event
);
1988 osb
->dc_work_sequence
= 0;
1989 osb
->dc_wake_sequence
= 0;
1990 INIT_LIST_HEAD(&osb
->blocked_lock_list
);
1991 osb
->blocked_lock_count
= 0;
1992 spin_lock_init(&osb
->osb_lock
);
1993 spin_lock_init(&osb
->osb_xattr_lock
);
1994 ocfs2_init_inode_steal_slot(osb
);
1996 atomic_set(&osb
->alloc_stats
.moves
, 0);
1997 atomic_set(&osb
->alloc_stats
.local_data
, 0);
1998 atomic_set(&osb
->alloc_stats
.bitmap_data
, 0);
1999 atomic_set(&osb
->alloc_stats
.bg_allocs
, 0);
2000 atomic_set(&osb
->alloc_stats
.bg_extends
, 0);
2002 /* Copy the blockcheck stats from the superblock probe */
2003 osb
->osb_ecc_stats
= *stats
;
2005 ocfs2_init_node_maps(osb
);
2007 snprintf(osb
->dev_str
, sizeof(osb
->dev_str
), "%u,%u",
2008 MAJOR(osb
->sb
->s_dev
), MINOR(osb
->sb
->s_dev
));
2010 ocfs2_orphan_scan_init(osb
);
2012 status
= ocfs2_recovery_init(osb
);
2014 mlog(ML_ERROR
, "Unable to initialize recovery state\n");
2019 init_waitqueue_head(&osb
->checkpoint_event
);
2020 atomic_set(&osb
->needs_checkpoint
, 0);
2022 osb
->s_atime_quantum
= OCFS2_DEFAULT_ATIME_QUANTUM
;
2024 osb
->slot_num
= OCFS2_INVALID_SLOT
;
2026 osb
->s_xattr_inline_size
= le16_to_cpu(
2027 di
->id2
.i_super
.s_xattr_inline_size
);
2029 osb
->local_alloc_state
= OCFS2_LA_UNUSED
;
2030 osb
->local_alloc_bh
= NULL
;
2031 INIT_DELAYED_WORK(&osb
->la_enable_wq
, ocfs2_la_enable_worker
);
2033 init_waitqueue_head(&osb
->osb_mount_event
);
2035 osb
->vol_label
= kmalloc(OCFS2_MAX_VOL_LABEL_LEN
, GFP_KERNEL
);
2036 if (!osb
->vol_label
) {
2037 mlog(ML_ERROR
, "unable to alloc vol label\n");
2042 osb
->max_slots
= le16_to_cpu(di
->id2
.i_super
.s_max_slots
);
2043 if (osb
->max_slots
> OCFS2_MAX_SLOTS
|| osb
->max_slots
== 0) {
2044 mlog(ML_ERROR
, "Invalid number of node slots (%u)\n",
2049 mlog(0, "max_slots for this device: %u\n", osb
->max_slots
);
2051 osb
->slot_recovery_generations
=
2052 kcalloc(osb
->max_slots
, sizeof(*osb
->slot_recovery_generations
),
2054 if (!osb
->slot_recovery_generations
) {
2060 init_waitqueue_head(&osb
->osb_wipe_event
);
2061 osb
->osb_orphan_wipes
= kcalloc(osb
->max_slots
,
2062 sizeof(*osb
->osb_orphan_wipes
),
2064 if (!osb
->osb_orphan_wipes
) {
2070 osb
->osb_rf_lock_tree
= RB_ROOT
;
2072 osb
->s_feature_compat
=
2073 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_compat
);
2074 osb
->s_feature_ro_compat
=
2075 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_ro_compat
);
2076 osb
->s_feature_incompat
=
2077 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_incompat
);
2079 if ((i
= OCFS2_HAS_INCOMPAT_FEATURE(osb
->sb
, ~OCFS2_FEATURE_INCOMPAT_SUPP
))) {
2080 mlog(ML_ERROR
, "couldn't mount because of unsupported "
2081 "optional features (%x).\n", i
);
2085 if (!(osb
->sb
->s_flags
& MS_RDONLY
) &&
2086 (i
= OCFS2_HAS_RO_COMPAT_FEATURE(osb
->sb
, ~OCFS2_FEATURE_RO_COMPAT_SUPP
))) {
2087 mlog(ML_ERROR
, "couldn't mount RDWR because of "
2088 "unsupported optional features (%x).\n", i
);
2093 if (ocfs2_userspace_stack(osb
)) {
2094 memcpy(osb
->osb_cluster_stack
,
2095 OCFS2_RAW_SB(di
)->s_cluster_info
.ci_stack
,
2096 OCFS2_STACK_LABEL_LEN
);
2097 osb
->osb_cluster_stack
[OCFS2_STACK_LABEL_LEN
] = '\0';
2098 if (strlen(osb
->osb_cluster_stack
) != OCFS2_STACK_LABEL_LEN
) {
2100 "couldn't mount because of an invalid "
2101 "cluster stack label (%s) \n",
2102 osb
->osb_cluster_stack
);
2107 /* The empty string is identical with classic tools that
2108 * don't know about s_cluster_info. */
2109 osb
->osb_cluster_stack
[0] = '\0';
2112 get_random_bytes(&osb
->s_next_generation
, sizeof(u32
));
2115 * This should be done in ocfs2_journal_init(), but unknown
2116 * ordering issues will cause the filesystem to crash.
2117 * If anyone wants to figure out what part of the code
2118 * refers to osb->journal before ocfs2_journal_init() is run,
2121 /* initialize our journal structure */
2123 journal
= kzalloc(sizeof(struct ocfs2_journal
), GFP_KERNEL
);
2125 mlog(ML_ERROR
, "unable to alloc journal\n");
2129 osb
->journal
= journal
;
2130 journal
->j_osb
= osb
;
2132 atomic_set(&journal
->j_num_trans
, 0);
2133 init_rwsem(&journal
->j_trans_barrier
);
2134 init_waitqueue_head(&journal
->j_checkpointed
);
2135 spin_lock_init(&journal
->j_lock
);
2136 journal
->j_trans_id
= (unsigned long) 1;
2137 INIT_LIST_HEAD(&journal
->j_la_cleanups
);
2138 INIT_WORK(&journal
->j_recovery_work
, ocfs2_complete_recovery
);
2139 journal
->j_state
= OCFS2_JOURNAL_FREE
;
2141 INIT_WORK(&osb
->dentry_lock_work
, ocfs2_drop_dl_inodes
);
2142 osb
->dentry_lock_list
= NULL
;
2144 /* get some pseudo constants for clustersize bits */
2145 osb
->s_clustersize_bits
=
2146 le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
);
2147 osb
->s_clustersize
= 1 << osb
->s_clustersize_bits
;
2148 mlog(0, "clusterbits=%d\n", osb
->s_clustersize_bits
);
2150 if (osb
->s_clustersize
< OCFS2_MIN_CLUSTERSIZE
||
2151 osb
->s_clustersize
> OCFS2_MAX_CLUSTERSIZE
) {
2152 mlog(ML_ERROR
, "Volume has invalid cluster size (%d)\n",
2153 osb
->s_clustersize
);
2158 if (ocfs2_clusters_to_blocks(osb
->sb
, le32_to_cpu(di
->i_clusters
) - 1)
2160 mlog(ML_ERROR
, "Volume might try to write to blocks beyond "
2161 "what jbd can address in 32 bits.\n");
2166 if (ocfs2_setup_osb_uuid(osb
, di
->id2
.i_super
.s_uuid
,
2167 sizeof(di
->id2
.i_super
.s_uuid
))) {
2168 mlog(ML_ERROR
, "Out of memory trying to setup our uuid.\n");
2173 memcpy(&uuid_net_key
, di
->id2
.i_super
.s_uuid
, sizeof(uuid_net_key
));
2175 strncpy(osb
->vol_label
, di
->id2
.i_super
.s_label
, 63);
2176 osb
->vol_label
[63] = '\0';
2177 osb
->root_blkno
= le64_to_cpu(di
->id2
.i_super
.s_root_blkno
);
2178 osb
->system_dir_blkno
= le64_to_cpu(di
->id2
.i_super
.s_system_dir_blkno
);
2179 osb
->first_cluster_group_blkno
=
2180 le64_to_cpu(di
->id2
.i_super
.s_first_cluster_group
);
2181 osb
->fs_generation
= le32_to_cpu(di
->i_fs_generation
);
2182 osb
->uuid_hash
= le32_to_cpu(di
->id2
.i_super
.s_uuid_hash
);
2183 mlog(0, "vol_label: %s\n", osb
->vol_label
);
2184 mlog(0, "uuid: %s\n", osb
->uuid_str
);
2185 mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
2186 (unsigned long long)osb
->root_blkno
,
2187 (unsigned long long)osb
->system_dir_blkno
);
2189 osb
->osb_dlm_debug
= ocfs2_new_dlm_debug();
2190 if (!osb
->osb_dlm_debug
) {
2196 atomic_set(&osb
->vol_state
, VOLUME_INIT
);
2198 /* load root, system_dir, and all global system inodes */
2199 status
= ocfs2_init_global_system_inodes(osb
);
2208 inode
= ocfs2_get_system_file_inode(osb
, GLOBAL_BITMAP_SYSTEM_INODE
,
2209 OCFS2_INVALID_SLOT
);
2216 osb
->bitmap_blkno
= OCFS2_I(inode
)->ip_blkno
;
2219 osb
->bitmap_cpg
= ocfs2_group_bitmap_size(sb
) * 8;
2221 status
= ocfs2_init_slot_info(osb
);
2233 * will return: -EAGAIN if it is ok to keep searching for superblocks
2234 * -EINVAL if there is a bad superblock
2237 static int ocfs2_verify_volume(struct ocfs2_dinode
*di
,
2238 struct buffer_head
*bh
,
2240 struct ocfs2_blockcheck_stats
*stats
)
2242 int status
= -EAGAIN
;
2246 if (memcmp(di
->i_signature
, OCFS2_SUPER_BLOCK_SIGNATURE
,
2247 strlen(OCFS2_SUPER_BLOCK_SIGNATURE
)) == 0) {
2248 /* We have to do a raw check of the feature here */
2249 if (le32_to_cpu(di
->id2
.i_super
.s_feature_incompat
) &
2250 OCFS2_FEATURE_INCOMPAT_META_ECC
) {
2251 status
= ocfs2_block_check_validate(bh
->b_data
,
2259 if ((1 << le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
)) != blksz
) {
2260 mlog(ML_ERROR
, "found superblock with incorrect block "
2261 "size: found %u, should be %u\n",
2262 1 << le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
),
2264 } else if (le16_to_cpu(di
->id2
.i_super
.s_major_rev_level
) !=
2265 OCFS2_MAJOR_REV_LEVEL
||
2266 le16_to_cpu(di
->id2
.i_super
.s_minor_rev_level
) !=
2267 OCFS2_MINOR_REV_LEVEL
) {
2268 mlog(ML_ERROR
, "found superblock with bad version: "
2269 "found %u.%u, should be %u.%u\n",
2270 le16_to_cpu(di
->id2
.i_super
.s_major_rev_level
),
2271 le16_to_cpu(di
->id2
.i_super
.s_minor_rev_level
),
2272 OCFS2_MAJOR_REV_LEVEL
,
2273 OCFS2_MINOR_REV_LEVEL
);
2274 } else if (bh
->b_blocknr
!= le64_to_cpu(di
->i_blkno
)) {
2275 mlog(ML_ERROR
, "bad block number on superblock: "
2276 "found %llu, should be %llu\n",
2277 (unsigned long long)le64_to_cpu(di
->i_blkno
),
2278 (unsigned long long)bh
->b_blocknr
);
2279 } else if (le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
) < 12 ||
2280 le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
) > 20) {
2281 mlog(ML_ERROR
, "bad cluster size found: %u\n",
2282 1 << le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
));
2283 } else if (!le64_to_cpu(di
->id2
.i_super
.s_root_blkno
)) {
2284 mlog(ML_ERROR
, "bad root_blkno: 0\n");
2285 } else if (!le64_to_cpu(di
->id2
.i_super
.s_system_dir_blkno
)) {
2286 mlog(ML_ERROR
, "bad system_dir_blkno: 0\n");
2287 } else if (le16_to_cpu(di
->id2
.i_super
.s_max_slots
) > OCFS2_MAX_SLOTS
) {
2289 "Superblock slots found greater than file system "
2290 "maximum: found %u, max %u\n",
2291 le16_to_cpu(di
->id2
.i_super
.s_max_slots
),
2304 static int ocfs2_check_volume(struct ocfs2_super
*osb
)
2309 struct ocfs2_dinode
*local_alloc
= NULL
; /* only used if we
2315 /* Init our journal object. */
2316 status
= ocfs2_journal_init(osb
->journal
, &dirty
);
2318 mlog(ML_ERROR
, "Could not initialize journal!\n");
2322 /* If the journal was unmounted cleanly then we don't want to
2323 * recover anything. Otherwise, journal_load will do that
2324 * dirty work for us :) */
2326 status
= ocfs2_journal_wipe(osb
->journal
, 0);
2332 mlog(ML_NOTICE
, "File system was not unmounted cleanly, "
2333 "recovering volume.\n");
2336 local
= ocfs2_mount_local(osb
);
2338 /* will play back anything left in the journal. */
2339 status
= ocfs2_journal_load(osb
->journal
, local
, dirty
);
2341 mlog(ML_ERROR
, "ocfs2 journal load failed! %d\n", status
);
2346 /* recover my local alloc if we didn't unmount cleanly. */
2347 status
= ocfs2_begin_local_alloc_recovery(osb
,
2354 /* we complete the recovery process after we've marked
2355 * ourselves as mounted. */
2358 mlog(0, "Journal loaded.\n");
2360 status
= ocfs2_load_local_alloc(osb
);
2367 /* Recovery will be completed after we've mounted the
2368 * rest of the volume. */
2370 osb
->local_alloc_copy
= local_alloc
;
2374 /* go through each journal, trylock it and if you get the
2375 * lock, and it's marked as dirty, set the bit in the recover
2376 * map and launch a recovery thread for it. */
2377 status
= ocfs2_mark_dead_nodes(osb
);
2383 status
= ocfs2_compute_replay_slots(osb
);
2396 * The routine gets called from dismount or close whenever a dismount on
2397 * volume is requested and the osb open count becomes 1.
2398 * It will remove the osb from the global list and also free up all the
2399 * initialized resources and fileobject.
2401 static void ocfs2_delete_osb(struct ocfs2_super
*osb
)
2405 /* This function assumes that the caller has the main osb resource */
2407 ocfs2_free_slot_info(osb
);
2409 kfree(osb
->osb_orphan_wipes
);
2410 kfree(osb
->slot_recovery_generations
);
2412 * This belongs in journal shutdown, but because we have to
2413 * allocate osb->journal at the start of ocfs2_initalize_osb(),
2416 kfree(osb
->journal
);
2417 if (osb
->local_alloc_copy
)
2418 kfree(osb
->local_alloc_copy
);
2419 kfree(osb
->uuid_str
);
2420 ocfs2_put_dlm_debug(osb
->osb_dlm_debug
);
2421 memset(osb
, 0, sizeof(struct ocfs2_super
));
2426 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2427 * panic(). We do not support continue-on-error operation. */
2428 static void ocfs2_handle_error(struct super_block
*sb
)
2430 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
2432 if (osb
->s_mount_opt
& OCFS2_MOUNT_ERRORS_PANIC
)
2433 panic("OCFS2: (device %s): panic forced after error\n",
2436 ocfs2_set_osb_flag(osb
, OCFS2_OSB_ERROR_FS
);
2438 if (sb
->s_flags
& MS_RDONLY
&&
2439 (ocfs2_is_soft_readonly(osb
) ||
2440 ocfs2_is_hard_readonly(osb
)))
2443 printk(KERN_CRIT
"File system is now read-only due to the potential "
2444 "of on-disk corruption. Please run fsck.ocfs2 once the file "
2445 "system is unmounted.\n");
2446 sb
->s_flags
|= MS_RDONLY
;
2447 ocfs2_set_ro_flag(osb
, 0);
2450 static char error_buf
[1024];
2452 void __ocfs2_error(struct super_block
*sb
,
2453 const char *function
,
2454 const char *fmt
, ...)
2458 va_start(args
, fmt
);
2459 vsnprintf(error_buf
, sizeof(error_buf
), fmt
, args
);
2462 /* Not using mlog here because we want to show the actual
2463 * function the error came from. */
2464 printk(KERN_CRIT
"OCFS2: ERROR (device %s): %s: %s\n",
2465 sb
->s_id
, function
, error_buf
);
2467 ocfs2_handle_error(sb
);
2470 /* Handle critical errors. This is intentionally more drastic than
2471 * ocfs2_handle_error, so we only use for things like journal errors,
2473 void __ocfs2_abort(struct super_block
* sb
,
2474 const char *function
,
2475 const char *fmt
, ...)
2479 va_start(args
, fmt
);
2480 vsnprintf(error_buf
, sizeof(error_buf
), fmt
, args
);
2483 printk(KERN_CRIT
"OCFS2: abort (device %s): %s: %s\n",
2484 sb
->s_id
, function
, error_buf
);
2486 /* We don't have the cluster support yet to go straight to
2487 * hard readonly in here. Until then, we want to keep
2488 * ocfs2_abort() so that we can at least mark critical
2491 * TODO: This should abort the journal and alert other nodes
2492 * that our slot needs recovery. */
2494 /* Force a panic(). This stinks, but it's better than letting
2495 * things continue without having a proper hard readonly
2497 OCFS2_SB(sb
)->s_mount_opt
|= OCFS2_MOUNT_ERRORS_PANIC
;
2498 ocfs2_handle_error(sb
);
2501 module_init(ocfs2_init
);
2502 module_exit(ocfs2_exit
);