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"
73 #include "buffer_head_io.h"
75 static struct kmem_cache
*ocfs2_inode_cachep
= NULL
;
76 struct kmem_cache
*ocfs2_dquot_cachep
;
77 struct kmem_cache
*ocfs2_qf_chunk_cachep
;
79 /* OCFS2 needs to schedule several differnt types of work which
80 * require cluster locking, disk I/O, recovery waits, etc. Since these
81 * types of work tend to be heavy we avoid using the kernel events
82 * workqueue and schedule on our own. */
83 struct workqueue_struct
*ocfs2_wq
= NULL
;
85 static struct dentry
*ocfs2_debugfs_root
= NULL
;
87 MODULE_AUTHOR("Oracle");
88 MODULE_LICENSE("GPL");
92 unsigned long commit_interval
;
93 unsigned long mount_opt
;
94 unsigned int atime_quantum
;
96 unsigned int localalloc_opt
;
97 char cluster_stack
[OCFS2_STACK_LABEL_LEN
+ 1];
100 static int ocfs2_parse_options(struct super_block
*sb
, char *options
,
101 struct mount_options
*mopt
,
103 static int ocfs2_show_options(struct seq_file
*s
, struct vfsmount
*mnt
);
104 static void ocfs2_put_super(struct super_block
*sb
);
105 static int ocfs2_mount_volume(struct super_block
*sb
);
106 static int ocfs2_remount(struct super_block
*sb
, int *flags
, char *data
);
107 static void ocfs2_dismount_volume(struct super_block
*sb
, int mnt_err
);
108 static int ocfs2_initialize_mem_caches(void);
109 static void ocfs2_free_mem_caches(void);
110 static void ocfs2_delete_osb(struct ocfs2_super
*osb
);
112 static int ocfs2_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
114 static int ocfs2_sync_fs(struct super_block
*sb
, int wait
);
116 static int ocfs2_init_global_system_inodes(struct ocfs2_super
*osb
);
117 static int ocfs2_init_local_system_inodes(struct ocfs2_super
*osb
);
118 static void ocfs2_release_system_inodes(struct ocfs2_super
*osb
);
119 static int ocfs2_check_volume(struct ocfs2_super
*osb
);
120 static int ocfs2_verify_volume(struct ocfs2_dinode
*di
,
121 struct buffer_head
*bh
,
123 struct ocfs2_blockcheck_stats
*stats
);
124 static int ocfs2_initialize_super(struct super_block
*sb
,
125 struct buffer_head
*bh
,
127 struct ocfs2_blockcheck_stats
*stats
);
128 static int ocfs2_get_sector(struct super_block
*sb
,
129 struct buffer_head
**bh
,
132 static struct inode
*ocfs2_alloc_inode(struct super_block
*sb
);
133 static void ocfs2_destroy_inode(struct inode
*inode
);
134 static int ocfs2_susp_quotas(struct ocfs2_super
*osb
, int unsuspend
);
135 static int ocfs2_enable_quotas(struct ocfs2_super
*osb
);
136 static void ocfs2_disable_quotas(struct ocfs2_super
*osb
);
138 static const struct super_operations ocfs2_sops
= {
139 .statfs
= ocfs2_statfs
,
140 .alloc_inode
= ocfs2_alloc_inode
,
141 .destroy_inode
= ocfs2_destroy_inode
,
142 .drop_inode
= ocfs2_drop_inode
,
143 .clear_inode
= ocfs2_clear_inode
,
144 .delete_inode
= ocfs2_delete_inode
,
145 .sync_fs
= ocfs2_sync_fs
,
146 .put_super
= ocfs2_put_super
,
147 .remount_fs
= ocfs2_remount
,
148 .show_options
= ocfs2_show_options
,
149 .quota_read
= ocfs2_quota_read
,
150 .quota_write
= ocfs2_quota_write
,
179 static const match_table_t tokens
= {
180 {Opt_barrier
, "barrier=%u"},
181 {Opt_err_panic
, "errors=panic"},
182 {Opt_err_ro
, "errors=remount-ro"},
184 {Opt_nointr
, "nointr"},
185 {Opt_hb_none
, OCFS2_HB_NONE
},
186 {Opt_hb_local
, OCFS2_HB_LOCAL
},
187 {Opt_data_ordered
, "data=ordered"},
188 {Opt_data_writeback
, "data=writeback"},
189 {Opt_atime_quantum
, "atime_quantum=%u"},
190 {Opt_slot
, "preferred_slot=%u"},
191 {Opt_commit
, "commit=%u"},
192 {Opt_localalloc
, "localalloc=%d"},
193 {Opt_localflocks
, "localflocks"},
194 {Opt_stack
, "cluster_stack=%s"},
195 {Opt_user_xattr
, "user_xattr"},
196 {Opt_nouser_xattr
, "nouser_xattr"},
197 {Opt_inode64
, "inode64"},
199 {Opt_noacl
, "noacl"},
200 {Opt_usrquota
, "usrquota"},
201 {Opt_grpquota
, "grpquota"},
205 #ifdef CONFIG_DEBUG_FS
206 static int ocfs2_osb_dump(struct ocfs2_super
*osb
, char *buf
, int len
)
210 struct ocfs2_cluster_connection
*cconn
= osb
->cconn
;
211 struct ocfs2_recovery_map
*rm
= osb
->recovery_map
;
212 struct ocfs2_orphan_scan
*os
;
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
);
237 out
+= snprintf(buf
+ out
, len
- out
,
238 "%10s => Stack: %s Name: %*s Version: %d.%d\n",
240 (*osb
->osb_cluster_stack
== '\0' ?
241 "o2cb" : osb
->osb_cluster_stack
),
242 cconn
->cc_namelen
, cconn
->cc_name
,
243 cconn
->cc_version
.pv_major
, cconn
->cc_version
.pv_minor
);
245 spin_lock(&osb
->dc_task_lock
);
246 out
+= snprintf(buf
+ out
, len
- out
,
247 "%10s => Pid: %d Count: %lu WakeSeq: %lu "
248 "WorkSeq: %lu\n", "DownCnvt",
249 task_pid_nr(osb
->dc_task
), osb
->blocked_lock_count
,
250 osb
->dc_wake_sequence
, osb
->dc_work_sequence
);
251 spin_unlock(&osb
->dc_task_lock
);
253 spin_lock(&osb
->osb_lock
);
254 out
+= snprintf(buf
+ out
, len
- out
, "%10s => Pid: %d Nodes:",
256 (osb
->recovery_thread_task
?
257 task_pid_nr(osb
->recovery_thread_task
) : -1));
258 if (rm
->rm_used
== 0)
259 out
+= snprintf(buf
+ out
, len
- out
, " None\n");
261 for (i
= 0; i
< rm
->rm_used
; i
++)
262 out
+= snprintf(buf
+ out
, len
- out
, " %d",
264 out
+= snprintf(buf
+ out
, len
- out
, "\n");
266 spin_unlock(&osb
->osb_lock
);
268 out
+= snprintf(buf
+ out
, len
- out
,
269 "%10s => Pid: %d Interval: %lu Needs: %d\n", "Commit",
270 task_pid_nr(osb
->commit_task
), osb
->osb_commit_interval
,
271 atomic_read(&osb
->needs_checkpoint
));
273 out
+= snprintf(buf
+ out
, len
- out
,
274 "%10s => State: %d NumTxns: %d TxnId: %lu\n",
275 "Journal", osb
->journal
->j_state
,
276 atomic_read(&osb
->journal
->j_num_trans
),
277 osb
->journal
->j_trans_id
);
279 out
+= snprintf(buf
+ out
, len
- out
,
280 "%10s => GlobalAllocs: %d LocalAllocs: %d "
281 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
283 atomic_read(&osb
->alloc_stats
.bitmap_data
),
284 atomic_read(&osb
->alloc_stats
.local_data
),
285 atomic_read(&osb
->alloc_stats
.bg_allocs
),
286 atomic_read(&osb
->alloc_stats
.moves
),
287 atomic_read(&osb
->alloc_stats
.bg_extends
));
289 out
+= snprintf(buf
+ out
, len
- out
,
290 "%10s => State: %u Descriptor: %llu Size: %u bits "
291 "Default: %u bits\n",
292 "LocalAlloc", osb
->local_alloc_state
,
293 (unsigned long long)osb
->la_last_gd
,
294 osb
->local_alloc_bits
, osb
->local_alloc_default_bits
);
296 spin_lock(&osb
->osb_lock
);
297 out
+= snprintf(buf
+ out
, len
- out
,
298 "%10s => Slot: %d NumStolen: %d\n", "Steal",
299 osb
->s_inode_steal_slot
,
300 atomic_read(&osb
->s_num_inodes_stolen
));
301 spin_unlock(&osb
->osb_lock
);
303 out
+= snprintf(buf
+ out
, len
- out
, "%10s => %3s %10s\n",
304 "Slots", "Num", "RecoGen");
306 for (i
= 0; i
< osb
->max_slots
; ++i
) {
307 out
+= snprintf(buf
+ out
, len
- out
,
308 "%10s %c %3d %10d\n",
310 (i
== osb
->slot_num
? '*' : ' '),
311 i
, osb
->slot_recovery_generations
[i
]);
314 os
= &osb
->osb_orphan_scan
;
315 out
+= snprintf(buf
+ out
, len
- out
, "Orphan Scan=> ");
316 out
+= snprintf(buf
+ out
, len
- out
, "Local: %u Global: %u ",
317 os
->os_count
, os
->os_seqno
);
318 out
+= snprintf(buf
+ out
, len
- out
, " Last Scan: %lu seconds ago\n",
319 (get_seconds() - os
->os_scantime
.tv_sec
));
324 static int ocfs2_osb_debug_open(struct inode
*inode
, struct file
*file
)
326 struct ocfs2_super
*osb
= inode
->i_private
;
329 buf
= kmalloc(PAGE_SIZE
, GFP_KERNEL
);
333 i_size_write(inode
, ocfs2_osb_dump(osb
, buf
, PAGE_SIZE
));
335 file
->private_data
= buf
;
342 static int ocfs2_debug_release(struct inode
*inode
, struct file
*file
)
344 kfree(file
->private_data
);
348 static ssize_t
ocfs2_debug_read(struct file
*file
, char __user
*buf
,
349 size_t nbytes
, loff_t
*ppos
)
351 return simple_read_from_buffer(buf
, nbytes
, ppos
, file
->private_data
,
352 i_size_read(file
->f_mapping
->host
));
355 static int ocfs2_osb_debug_open(struct inode
*inode
, struct file
*file
)
359 static int ocfs2_debug_release(struct inode
*inode
, struct file
*file
)
363 static ssize_t
ocfs2_debug_read(struct file
*file
, char __user
*buf
,
364 size_t nbytes
, loff_t
*ppos
)
368 #endif /* CONFIG_DEBUG_FS */
370 static struct file_operations ocfs2_osb_debug_fops
= {
371 .open
= ocfs2_osb_debug_open
,
372 .release
= ocfs2_debug_release
,
373 .read
= ocfs2_debug_read
,
374 .llseek
= generic_file_llseek
,
377 static int ocfs2_sync_fs(struct super_block
*sb
, int wait
)
381 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
383 if (ocfs2_is_hard_readonly(osb
))
387 status
= ocfs2_flush_truncate_log(osb
);
391 ocfs2_schedule_truncate_log_flush(osb
, 0);
394 if (jbd2_journal_start_commit(OCFS2_SB(sb
)->journal
->j_journal
,
397 jbd2_log_wait_commit(OCFS2_SB(sb
)->journal
->j_journal
,
403 static int ocfs2_need_system_inode(struct ocfs2_super
*osb
, int ino
)
405 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb
->sb
, OCFS2_FEATURE_RO_COMPAT_USRQUOTA
)
406 && (ino
== USER_QUOTA_SYSTEM_INODE
407 || ino
== LOCAL_USER_QUOTA_SYSTEM_INODE
))
409 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb
->sb
, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
)
410 && (ino
== GROUP_QUOTA_SYSTEM_INODE
411 || ino
== LOCAL_GROUP_QUOTA_SYSTEM_INODE
))
416 static int ocfs2_init_global_system_inodes(struct ocfs2_super
*osb
)
418 struct inode
*new = NULL
;
424 new = ocfs2_iget(osb
, osb
->root_blkno
, OCFS2_FI_FLAG_SYSFILE
, 0);
426 status
= PTR_ERR(new);
430 osb
->root_inode
= new;
432 new = ocfs2_iget(osb
, osb
->system_dir_blkno
, OCFS2_FI_FLAG_SYSFILE
, 0);
434 status
= PTR_ERR(new);
438 osb
->sys_root_inode
= new;
440 for (i
= OCFS2_FIRST_ONLINE_SYSTEM_INODE
;
441 i
<= OCFS2_LAST_GLOBAL_SYSTEM_INODE
; i
++) {
442 if (!ocfs2_need_system_inode(osb
, i
))
444 new = ocfs2_get_system_file_inode(osb
, i
, osb
->slot_num
);
446 ocfs2_release_system_inodes(osb
);
449 /* FIXME: Should ERROR_RO_FS */
450 mlog(ML_ERROR
, "Unable to load system inode %d, "
451 "possibly corrupt fs?", i
);
454 // the array now has one ref, so drop this one
463 static int ocfs2_init_local_system_inodes(struct ocfs2_super
*osb
)
465 struct inode
*new = NULL
;
471 for (i
= OCFS2_LAST_GLOBAL_SYSTEM_INODE
+ 1;
472 i
< NUM_SYSTEM_INODES
;
474 if (!ocfs2_need_system_inode(osb
, i
))
476 new = ocfs2_get_system_file_inode(osb
, i
, osb
->slot_num
);
478 ocfs2_release_system_inodes(osb
);
480 mlog(ML_ERROR
, "status=%d, sysfile=%d, slot=%d\n",
481 status
, i
, osb
->slot_num
);
484 /* the array now has one ref, so drop this one */
493 static void ocfs2_release_system_inodes(struct ocfs2_super
*osb
)
500 for (i
= 0; i
< NUM_SYSTEM_INODES
; i
++) {
501 inode
= osb
->system_inodes
[i
];
504 osb
->system_inodes
[i
] = NULL
;
508 inode
= osb
->sys_root_inode
;
511 osb
->sys_root_inode
= NULL
;
514 inode
= osb
->root_inode
;
517 osb
->root_inode
= NULL
;
523 /* We're allocating fs objects, use GFP_NOFS */
524 static struct inode
*ocfs2_alloc_inode(struct super_block
*sb
)
526 struct ocfs2_inode_info
*oi
;
528 oi
= kmem_cache_alloc(ocfs2_inode_cachep
, GFP_NOFS
);
532 jbd2_journal_init_jbd_inode(&oi
->ip_jinode
, &oi
->vfs_inode
);
533 return &oi
->vfs_inode
;
536 static void ocfs2_destroy_inode(struct inode
*inode
)
538 kmem_cache_free(ocfs2_inode_cachep
, OCFS2_I(inode
));
541 static unsigned long long ocfs2_max_file_offset(unsigned int bbits
,
544 unsigned int bytes
= 1 << cbits
;
545 unsigned int trim
= bytes
;
546 unsigned int bitshift
= 32;
549 * i_size and all block offsets in ocfs2 are always 64 bits
550 * wide. i_clusters is 32 bits, in cluster-sized units. So on
551 * 64 bit platforms, cluster size will be the limiting factor.
554 #if BITS_PER_LONG == 32
555 # if defined(CONFIG_LBD)
556 BUILD_BUG_ON(sizeof(sector_t
) != 8);
558 * We might be limited by page cache size.
560 if (bytes
> PAGE_CACHE_SIZE
) {
561 bytes
= PAGE_CACHE_SIZE
;
564 * Shift by 31 here so that we don't get larger than
571 * We are limited by the size of sector_t. Use block size, as
572 * that's what we expose to the VFS.
581 * Trim by a whole cluster when we can actually approach the
582 * on-disk limits. Otherwise we can overflow i_clusters when
583 * an extent start is at the max offset.
585 return (((unsigned long long)bytes
) << bitshift
) - trim
;
588 static int ocfs2_remount(struct super_block
*sb
, int *flags
, char *data
)
590 int incompat_features
;
592 struct mount_options parsed_options
;
593 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
597 if (!ocfs2_parse_options(sb
, data
, &parsed_options
, 1)) {
602 if ((osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) !=
603 (parsed_options
.mount_opt
& OCFS2_MOUNT_HB_LOCAL
)) {
605 mlog(ML_ERROR
, "Cannot change heartbeat mode on remount\n");
609 if ((osb
->s_mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
) !=
610 (parsed_options
.mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
)) {
612 mlog(ML_ERROR
, "Cannot change data mode on remount\n");
616 /* Probably don't want this on remount; it might
617 * mess with other nodes */
618 if (!(osb
->s_mount_opt
& OCFS2_MOUNT_INODE64
) &&
619 (parsed_options
.mount_opt
& OCFS2_MOUNT_INODE64
)) {
621 mlog(ML_ERROR
, "Cannot enable inode64 on remount\n");
625 /* We're going to/from readonly mode. */
626 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
)) {
627 /* Disable quota accounting before remounting RO */
628 if (*flags
& MS_RDONLY
) {
629 ret
= ocfs2_susp_quotas(osb
, 0);
633 /* Lock here so the check of HARD_RO and the potential
634 * setting of SOFT_RO is atomic. */
635 spin_lock(&osb
->osb_lock
);
636 if (osb
->osb_flags
& OCFS2_OSB_HARD_RO
) {
637 mlog(ML_ERROR
, "Remount on readonly device is forbidden.\n");
642 if (*flags
& MS_RDONLY
) {
643 mlog(0, "Going to ro mode.\n");
644 sb
->s_flags
|= MS_RDONLY
;
645 osb
->osb_flags
|= OCFS2_OSB_SOFT_RO
;
647 mlog(0, "Making ro filesystem writeable.\n");
649 if (osb
->osb_flags
& OCFS2_OSB_ERROR_FS
) {
650 mlog(ML_ERROR
, "Cannot remount RDWR "
651 "filesystem due to previous errors.\n");
655 incompat_features
= OCFS2_HAS_RO_COMPAT_FEATURE(sb
, ~OCFS2_FEATURE_RO_COMPAT_SUPP
);
656 if (incompat_features
) {
657 mlog(ML_ERROR
, "Cannot remount RDWR because "
658 "of unsupported optional features "
659 "(%x).\n", incompat_features
);
663 sb
->s_flags
&= ~MS_RDONLY
;
664 osb
->osb_flags
&= ~OCFS2_OSB_SOFT_RO
;
667 spin_unlock(&osb
->osb_lock
);
668 /* Enable quota accounting after remounting RW */
669 if (!ret
&& !(*flags
& MS_RDONLY
)) {
670 if (sb_any_quota_suspended(sb
))
671 ret
= ocfs2_susp_quotas(osb
, 1);
673 ret
= ocfs2_enable_quotas(osb
);
675 /* Return back changes... */
676 spin_lock(&osb
->osb_lock
);
677 sb
->s_flags
|= MS_RDONLY
;
678 osb
->osb_flags
|= OCFS2_OSB_SOFT_RO
;
679 spin_unlock(&osb
->osb_lock
);
686 /* Only save off the new mount options in case of a successful
688 if (!(osb
->s_feature_incompat
& OCFS2_FEATURE_INCOMPAT_XATTR
))
689 parsed_options
.mount_opt
&= ~OCFS2_MOUNT_POSIX_ACL
;
690 osb
->s_mount_opt
= parsed_options
.mount_opt
;
691 osb
->s_atime_quantum
= parsed_options
.atime_quantum
;
692 osb
->preferred_slot
= parsed_options
.slot
;
693 if (parsed_options
.commit_interval
)
694 osb
->osb_commit_interval
= parsed_options
.commit_interval
;
696 if (!ocfs2_is_hard_readonly(osb
))
697 ocfs2_set_journal_params(osb
);
704 static int ocfs2_sb_probe(struct super_block
*sb
,
705 struct buffer_head
**bh
,
707 struct ocfs2_blockcheck_stats
*stats
)
710 struct ocfs1_vol_disk_hdr
*hdr
;
711 struct ocfs2_dinode
*di
;
717 *sector_size
= bdev_logical_block_size(sb
->s_bdev
);
718 if (*sector_size
> OCFS2_MAX_BLOCKSIZE
) {
719 mlog(ML_ERROR
, "Hardware sector size too large: %d (max=%d)\n",
720 *sector_size
, OCFS2_MAX_BLOCKSIZE
);
725 /* Can this really happen? */
726 if (*sector_size
< OCFS2_MIN_BLOCKSIZE
)
727 *sector_size
= OCFS2_MIN_BLOCKSIZE
;
729 /* check block zero for old format */
730 status
= ocfs2_get_sector(sb
, bh
, 0, *sector_size
);
735 hdr
= (struct ocfs1_vol_disk_hdr
*) (*bh
)->b_data
;
736 if (hdr
->major_version
== OCFS1_MAJOR_VERSION
) {
737 mlog(ML_ERROR
, "incompatible version: %u.%u\n",
738 hdr
->major_version
, hdr
->minor_version
);
741 if (memcmp(hdr
->signature
, OCFS1_VOLUME_SIGNATURE
,
742 strlen(OCFS1_VOLUME_SIGNATURE
)) == 0) {
743 mlog(ML_ERROR
, "incompatible volume signature: %8s\n",
750 mlog(ML_ERROR
, "This is an ocfs v1 filesystem which must be "
751 "upgraded before mounting with ocfs v2\n");
756 * Now check at magic offset for 512, 1024, 2048, 4096
757 * blocksizes. 4096 is the maximum blocksize because it is
758 * the minimum clustersize.
761 for (blksize
= *sector_size
;
762 blksize
<= OCFS2_MAX_BLOCKSIZE
;
764 tmpstat
= ocfs2_get_sector(sb
, bh
,
765 OCFS2_SUPER_BLOCK_BLKNO
,
772 di
= (struct ocfs2_dinode
*) (*bh
)->b_data
;
773 memset(stats
, 0, sizeof(struct ocfs2_blockcheck_stats
));
774 status
= ocfs2_verify_volume(di
, *bh
, blksize
, stats
);
779 if (status
!= -EAGAIN
)
787 static int ocfs2_verify_heartbeat(struct ocfs2_super
*osb
)
789 if (ocfs2_mount_local(osb
)) {
790 if (osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) {
791 mlog(ML_ERROR
, "Cannot heartbeat on a locally "
792 "mounted device.\n");
797 if (ocfs2_userspace_stack(osb
)) {
798 if (osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) {
799 mlog(ML_ERROR
, "Userspace stack expected, but "
800 "o2cb heartbeat arguments passed to mount\n");
805 if (!(osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
)) {
806 if (!ocfs2_mount_local(osb
) && !ocfs2_is_hard_readonly(osb
) &&
807 !ocfs2_userspace_stack(osb
)) {
808 mlog(ML_ERROR
, "Heartbeat has to be started to mount "
809 "a read-write clustered device.\n");
818 * If we're using a userspace stack, mount should have passed
819 * a name that matches the disk. If not, mount should not
820 * have passed a stack.
822 static int ocfs2_verify_userspace_stack(struct ocfs2_super
*osb
,
823 struct mount_options
*mopt
)
825 if (!ocfs2_userspace_stack(osb
) && mopt
->cluster_stack
[0]) {
827 "cluster stack passed to mount, but this filesystem "
828 "does not support it\n");
832 if (ocfs2_userspace_stack(osb
) &&
833 strncmp(osb
->osb_cluster_stack
, mopt
->cluster_stack
,
834 OCFS2_STACK_LABEL_LEN
)) {
836 "cluster stack passed to mount (\"%s\") does not "
837 "match the filesystem (\"%s\")\n",
839 osb
->osb_cluster_stack
);
846 static int ocfs2_susp_quotas(struct ocfs2_super
*osb
, int unsuspend
)
849 struct super_block
*sb
= osb
->sb
;
850 unsigned int feature
[MAXQUOTAS
] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA
,
851 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
};
854 for (type
= 0; type
< MAXQUOTAS
; type
++) {
855 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb
, feature
[type
]))
858 status
= vfs_quota_enable(
859 sb_dqopt(sb
)->files
[type
],
863 status
= vfs_quota_disable(sb
, type
,
869 mlog(ML_ERROR
, "Failed to suspend/unsuspend quotas on "
870 "remount (error = %d).\n", status
);
874 static int ocfs2_enable_quotas(struct ocfs2_super
*osb
)
876 struct inode
*inode
[MAXQUOTAS
] = { NULL
, NULL
};
877 struct super_block
*sb
= osb
->sb
;
878 unsigned int feature
[MAXQUOTAS
] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA
,
879 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
};
880 unsigned int ino
[MAXQUOTAS
] = { LOCAL_USER_QUOTA_SYSTEM_INODE
,
881 LOCAL_GROUP_QUOTA_SYSTEM_INODE
};
885 sb_dqopt(sb
)->flags
|= DQUOT_QUOTA_SYS_FILE
| DQUOT_NEGATIVE_USAGE
;
886 for (type
= 0; type
< MAXQUOTAS
; type
++) {
887 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb
, feature
[type
]))
889 inode
[type
] = ocfs2_get_system_file_inode(osb
, ino
[type
],
895 status
= vfs_quota_enable(inode
[type
], type
, QFMT_OCFS2
,
896 DQUOT_USAGE_ENABLED
);
901 for (type
= 0; type
< MAXQUOTAS
; type
++)
905 ocfs2_disable_quotas(osb
);
906 for (type
= 0; type
< MAXQUOTAS
; type
++)
912 static void ocfs2_disable_quotas(struct ocfs2_super
*osb
)
916 struct super_block
*sb
= osb
->sb
;
918 /* We mostly ignore errors in this function because there's not much
919 * we can do when we see them */
920 for (type
= 0; type
< MAXQUOTAS
; type
++) {
921 if (!sb_has_quota_loaded(sb
, type
))
923 inode
= igrab(sb
->s_dquot
.files
[type
]);
924 /* Turn off quotas. This will remove all dquot structures from
925 * memory and so they will be automatically synced to global
927 vfs_quota_disable(sb
, type
, DQUOT_USAGE_ENABLED
|
928 DQUOT_LIMITS_ENABLED
);
935 /* Handle quota on quotactl */
936 static int ocfs2_quota_on(struct super_block
*sb
, int type
, int format_id
,
937 char *path
, int remount
)
939 unsigned int feature
[MAXQUOTAS
] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA
,
940 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
};
942 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb
, feature
[type
]))
946 return 0; /* Just ignore it has been handled in
948 return vfs_quota_enable(sb_dqopt(sb
)->files
[type
], type
,
949 format_id
, DQUOT_LIMITS_ENABLED
);
952 /* Handle quota off quotactl */
953 static int ocfs2_quota_off(struct super_block
*sb
, int type
, int remount
)
956 return 0; /* Ignore now and handle later in
958 return vfs_quota_disable(sb
, type
, DQUOT_LIMITS_ENABLED
);
961 static struct quotactl_ops ocfs2_quotactl_ops
= {
962 .quota_on
= ocfs2_quota_on
,
963 .quota_off
= ocfs2_quota_off
,
964 .quota_sync
= vfs_quota_sync
,
965 .get_info
= vfs_get_dqinfo
,
966 .set_info
= vfs_set_dqinfo
,
967 .get_dqblk
= vfs_get_dqblk
,
968 .set_dqblk
= vfs_set_dqblk
,
971 static int ocfs2_fill_super(struct super_block
*sb
, void *data
, int silent
)
974 int status
, sector_size
;
975 struct mount_options parsed_options
;
976 struct inode
*inode
= NULL
;
977 struct ocfs2_super
*osb
= NULL
;
978 struct buffer_head
*bh
= NULL
;
980 struct ocfs2_blockcheck_stats stats
;
982 mlog_entry("%p, %p, %i", sb
, data
, silent
);
984 if (!ocfs2_parse_options(sb
, data
, &parsed_options
, 0)) {
986 goto read_super_error
;
989 /* probe for superblock */
990 status
= ocfs2_sb_probe(sb
, &bh
, §or_size
, &stats
);
992 mlog(ML_ERROR
, "superblock probe failed!\n");
993 goto read_super_error
;
996 status
= ocfs2_initialize_super(sb
, bh
, sector_size
, &stats
);
1000 goto read_super_error
;
1005 if (!(osb
->s_feature_incompat
& OCFS2_FEATURE_INCOMPAT_XATTR
))
1006 parsed_options
.mount_opt
&= ~OCFS2_MOUNT_POSIX_ACL
;
1008 osb
->s_mount_opt
= parsed_options
.mount_opt
;
1009 osb
->s_atime_quantum
= parsed_options
.atime_quantum
;
1010 osb
->preferred_slot
= parsed_options
.slot
;
1011 osb
->osb_commit_interval
= parsed_options
.commit_interval
;
1012 osb
->local_alloc_default_bits
= ocfs2_megabytes_to_clusters(sb
, parsed_options
.localalloc_opt
);
1013 osb
->local_alloc_bits
= osb
->local_alloc_default_bits
;
1014 if (osb
->s_mount_opt
& OCFS2_MOUNT_USRQUOTA
&&
1015 !OCFS2_HAS_RO_COMPAT_FEATURE(sb
,
1016 OCFS2_FEATURE_RO_COMPAT_USRQUOTA
)) {
1018 mlog(ML_ERROR
, "User quotas were requested, but this "
1019 "filesystem does not have the feature enabled.\n");
1020 goto read_super_error
;
1022 if (osb
->s_mount_opt
& OCFS2_MOUNT_GRPQUOTA
&&
1023 !OCFS2_HAS_RO_COMPAT_FEATURE(sb
,
1024 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
)) {
1026 mlog(ML_ERROR
, "Group quotas were requested, but this "
1027 "filesystem does not have the feature enabled.\n");
1028 goto read_super_error
;
1031 status
= ocfs2_verify_userspace_stack(osb
, &parsed_options
);
1033 goto read_super_error
;
1035 sb
->s_magic
= OCFS2_SUPER_MAGIC
;
1037 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
1038 ((osb
->s_mount_opt
& OCFS2_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
1040 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1042 if (bdev_read_only(sb
->s_bdev
)) {
1043 if (!(sb
->s_flags
& MS_RDONLY
)) {
1045 mlog(ML_ERROR
, "Readonly device detected but readonly "
1046 "mount was not specified.\n");
1047 goto read_super_error
;
1050 /* You should not be able to start a local heartbeat
1051 * on a readonly device. */
1052 if (osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) {
1054 mlog(ML_ERROR
, "Local heartbeat specified on readonly "
1056 goto read_super_error
;
1059 status
= ocfs2_check_journals_nolocks(osb
);
1061 if (status
== -EROFS
)
1062 mlog(ML_ERROR
, "Recovery required on readonly "
1063 "file system, but write access is "
1067 goto read_super_error
;
1070 ocfs2_set_ro_flag(osb
, 1);
1072 printk(KERN_NOTICE
"Readonly device detected. No cluster "
1073 "services will be utilized for this mount. Recovery "
1074 "will be skipped.\n");
1077 if (!ocfs2_is_hard_readonly(osb
)) {
1078 if (sb
->s_flags
& MS_RDONLY
)
1079 ocfs2_set_ro_flag(osb
, 0);
1082 status
= ocfs2_verify_heartbeat(osb
);
1085 goto read_super_error
;
1088 osb
->osb_debug_root
= debugfs_create_dir(osb
->uuid_str
,
1089 ocfs2_debugfs_root
);
1090 if (!osb
->osb_debug_root
) {
1092 mlog(ML_ERROR
, "Unable to create per-mount debugfs root.\n");
1093 goto read_super_error
;
1096 osb
->osb_ctxt
= debugfs_create_file("fs_state", S_IFREG
|S_IRUSR
,
1097 osb
->osb_debug_root
,
1099 &ocfs2_osb_debug_fops
);
1100 if (!osb
->osb_ctxt
) {
1103 goto read_super_error
;
1106 if (ocfs2_meta_ecc(osb
)) {
1107 status
= ocfs2_blockcheck_stats_debugfs_install(
1108 &osb
->osb_ecc_stats
,
1109 osb
->osb_debug_root
);
1112 "Unable to create blockcheck statistics "
1114 goto read_super_error
;
1118 status
= ocfs2_mount_volume(sb
);
1119 if (osb
->root_inode
)
1120 inode
= igrab(osb
->root_inode
);
1123 goto read_super_error
;
1128 goto read_super_error
;
1131 root
= d_alloc_root(inode
);
1135 goto read_super_error
;
1140 ocfs2_complete_mount_recovery(osb
);
1142 if (ocfs2_mount_local(osb
))
1143 snprintf(nodestr
, sizeof(nodestr
), "local");
1145 snprintf(nodestr
, sizeof(nodestr
), "%u", osb
->node_num
);
1147 printk(KERN_INFO
"ocfs2: Mounting device (%s) on (node %s, slot %d) "
1148 "with %s data mode.\n",
1149 osb
->dev_str
, nodestr
, osb
->slot_num
,
1150 osb
->s_mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
? "writeback" :
1153 atomic_set(&osb
->vol_state
, VOLUME_MOUNTED
);
1154 wake_up(&osb
->osb_mount_event
);
1156 /* Now we can initialize quotas because we can afford to wait
1157 * for cluster locks recovery now. That also means that truncation
1158 * log recovery can happen but that waits for proper quota setup */
1159 if (!(sb
->s_flags
& MS_RDONLY
)) {
1160 status
= ocfs2_enable_quotas(osb
);
1162 /* We have to err-out specially here because
1163 * s_root is already set */
1165 atomic_set(&osb
->vol_state
, VOLUME_DISABLED
);
1166 wake_up(&osb
->osb_mount_event
);
1172 ocfs2_complete_quota_recovery(osb
);
1174 /* Now we wake up again for processes waiting for quotas */
1175 atomic_set(&osb
->vol_state
, VOLUME_MOUNTED_QUOTAS
);
1176 wake_up(&osb
->osb_mount_event
);
1188 atomic_set(&osb
->vol_state
, VOLUME_DISABLED
);
1189 wake_up(&osb
->osb_mount_event
);
1190 ocfs2_dismount_volume(sb
, 1);
1197 static int ocfs2_get_sb(struct file_system_type
*fs_type
,
1199 const char *dev_name
,
1201 struct vfsmount
*mnt
)
1203 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, ocfs2_fill_super
,
1207 static struct file_system_type ocfs2_fs_type
= {
1208 .owner
= THIS_MODULE
,
1210 .get_sb
= ocfs2_get_sb
, /* is this called when we mount
1212 .kill_sb
= kill_block_super
, /* set to the generic one
1213 * right now, but do we
1214 * need to change that? */
1215 .fs_flags
= FS_REQUIRES_DEV
|FS_RENAME_DOES_D_MOVE
,
1219 static int ocfs2_parse_options(struct super_block
*sb
,
1221 struct mount_options
*mopt
,
1227 mlog_entry("remount: %d, options: \"%s\"\n", is_remount
,
1228 options
? options
: "(none)");
1230 mopt
->commit_interval
= 0;
1231 mopt
->mount_opt
= 0;
1232 mopt
->atime_quantum
= OCFS2_DEFAULT_ATIME_QUANTUM
;
1233 mopt
->slot
= OCFS2_INVALID_SLOT
;
1234 mopt
->localalloc_opt
= OCFS2_DEFAULT_LOCAL_ALLOC_SIZE
;
1235 mopt
->cluster_stack
[0] = '\0';
1242 while ((p
= strsep(&options
, ",")) != NULL
) {
1244 substring_t args
[MAX_OPT_ARGS
];
1249 token
= match_token(p
, tokens
, args
);
1252 mopt
->mount_opt
|= OCFS2_MOUNT_HB_LOCAL
;
1255 mopt
->mount_opt
&= ~OCFS2_MOUNT_HB_LOCAL
;
1258 if (match_int(&args
[0], &option
)) {
1263 mopt
->mount_opt
|= OCFS2_MOUNT_BARRIER
;
1265 mopt
->mount_opt
&= ~OCFS2_MOUNT_BARRIER
;
1268 mopt
->mount_opt
&= ~OCFS2_MOUNT_NOINTR
;
1271 mopt
->mount_opt
|= OCFS2_MOUNT_NOINTR
;
1274 mopt
->mount_opt
|= OCFS2_MOUNT_ERRORS_PANIC
;
1277 mopt
->mount_opt
&= ~OCFS2_MOUNT_ERRORS_PANIC
;
1279 case Opt_data_ordered
:
1280 mopt
->mount_opt
&= ~OCFS2_MOUNT_DATA_WRITEBACK
;
1282 case Opt_data_writeback
:
1283 mopt
->mount_opt
|= OCFS2_MOUNT_DATA_WRITEBACK
;
1285 case Opt_user_xattr
:
1286 mopt
->mount_opt
&= ~OCFS2_MOUNT_NOUSERXATTR
;
1288 case Opt_nouser_xattr
:
1289 mopt
->mount_opt
|= OCFS2_MOUNT_NOUSERXATTR
;
1291 case Opt_atime_quantum
:
1292 if (match_int(&args
[0], &option
)) {
1297 mopt
->atime_quantum
= option
;
1301 if (match_int(&args
[0], &option
)) {
1306 mopt
->slot
= (s16
)option
;
1310 if (match_int(&args
[0], &option
)) {
1317 option
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1318 mopt
->commit_interval
= HZ
* option
;
1320 case Opt_localalloc
:
1322 if (match_int(&args
[0], &option
)) {
1326 if (option
>= 0 && (option
<= ocfs2_local_alloc_size(sb
) * 8))
1327 mopt
->localalloc_opt
= option
;
1329 case Opt_localflocks
:
1331 * Changing this during remount could race
1332 * flock() requests, or "unbalance" existing
1333 * ones (e.g., a lock is taken in one mode but
1334 * dropped in the other). If users care enough
1335 * to flip locking modes during remount, we
1336 * could add a "local" flag to individual
1337 * flock structures for proper tracking of
1341 mopt
->mount_opt
|= OCFS2_MOUNT_LOCALFLOCKS
;
1344 /* Check both that the option we were passed
1345 * is of the right length and that it is a proper
1346 * string of the right length.
1348 if (((args
[0].to
- args
[0].from
) !=
1349 OCFS2_STACK_LABEL_LEN
) ||
1350 (strnlen(args
[0].from
,
1351 OCFS2_STACK_LABEL_LEN
) !=
1352 OCFS2_STACK_LABEL_LEN
)) {
1354 "Invalid cluster_stack option\n");
1358 memcpy(mopt
->cluster_stack
, args
[0].from
,
1359 OCFS2_STACK_LABEL_LEN
);
1360 mopt
->cluster_stack
[OCFS2_STACK_LABEL_LEN
] = '\0';
1363 mopt
->mount_opt
|= OCFS2_MOUNT_INODE64
;
1366 /* We check only on remount, otherwise features
1367 * aren't yet initialized. */
1368 if (is_remount
&& !OCFS2_HAS_RO_COMPAT_FEATURE(sb
,
1369 OCFS2_FEATURE_RO_COMPAT_USRQUOTA
)) {
1370 mlog(ML_ERROR
, "User quota requested but "
1371 "filesystem feature is not set\n");
1375 mopt
->mount_opt
|= OCFS2_MOUNT_USRQUOTA
;
1378 if (is_remount
&& !OCFS2_HAS_RO_COMPAT_FEATURE(sb
,
1379 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
)) {
1380 mlog(ML_ERROR
, "Group quota requested but "
1381 "filesystem feature is not set\n");
1385 mopt
->mount_opt
|= OCFS2_MOUNT_GRPQUOTA
;
1387 #ifdef CONFIG_OCFS2_FS_POSIX_ACL
1389 mopt
->mount_opt
|= OCFS2_MOUNT_POSIX_ACL
;
1392 mopt
->mount_opt
&= ~OCFS2_MOUNT_POSIX_ACL
;
1397 printk(KERN_INFO
"ocfs2 (no)acl options not supported\n");
1402 "Unrecognized mount option \"%s\" "
1403 "or missing value\n", p
);
1416 static int ocfs2_show_options(struct seq_file
*s
, struct vfsmount
*mnt
)
1418 struct ocfs2_super
*osb
= OCFS2_SB(mnt
->mnt_sb
);
1419 unsigned long opts
= osb
->s_mount_opt
;
1420 unsigned int local_alloc_megs
;
1422 if (opts
& OCFS2_MOUNT_HB_LOCAL
)
1423 seq_printf(s
, ",_netdev,heartbeat=local");
1425 seq_printf(s
, ",heartbeat=none");
1427 if (opts
& OCFS2_MOUNT_NOINTR
)
1428 seq_printf(s
, ",nointr");
1430 if (opts
& OCFS2_MOUNT_DATA_WRITEBACK
)
1431 seq_printf(s
, ",data=writeback");
1433 seq_printf(s
, ",data=ordered");
1435 if (opts
& OCFS2_MOUNT_BARRIER
)
1436 seq_printf(s
, ",barrier=1");
1438 if (opts
& OCFS2_MOUNT_ERRORS_PANIC
)
1439 seq_printf(s
, ",errors=panic");
1441 seq_printf(s
, ",errors=remount-ro");
1443 if (osb
->preferred_slot
!= OCFS2_INVALID_SLOT
)
1444 seq_printf(s
, ",preferred_slot=%d", osb
->preferred_slot
);
1446 if (osb
->s_atime_quantum
!= OCFS2_DEFAULT_ATIME_QUANTUM
)
1447 seq_printf(s
, ",atime_quantum=%u", osb
->s_atime_quantum
);
1449 if (osb
->osb_commit_interval
)
1450 seq_printf(s
, ",commit=%u",
1451 (unsigned) (osb
->osb_commit_interval
/ HZ
));
1453 local_alloc_megs
= osb
->local_alloc_bits
>> (20 - osb
->s_clustersize_bits
);
1454 if (local_alloc_megs
!= OCFS2_DEFAULT_LOCAL_ALLOC_SIZE
)
1455 seq_printf(s
, ",localalloc=%d", local_alloc_megs
);
1457 if (opts
& OCFS2_MOUNT_LOCALFLOCKS
)
1458 seq_printf(s
, ",localflocks,");
1460 if (osb
->osb_cluster_stack
[0])
1461 seq_printf(s
, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN
,
1462 osb
->osb_cluster_stack
);
1463 if (opts
& OCFS2_MOUNT_USRQUOTA
)
1464 seq_printf(s
, ",usrquota");
1465 if (opts
& OCFS2_MOUNT_GRPQUOTA
)
1466 seq_printf(s
, ",grpquota");
1468 if (opts
& OCFS2_MOUNT_NOUSERXATTR
)
1469 seq_printf(s
, ",nouser_xattr");
1471 seq_printf(s
, ",user_xattr");
1473 if (opts
& OCFS2_MOUNT_INODE64
)
1474 seq_printf(s
, ",inode64");
1476 #ifdef CONFIG_OCFS2_FS_POSIX_ACL
1477 if (opts
& OCFS2_MOUNT_POSIX_ACL
)
1478 seq_printf(s
, ",acl");
1480 seq_printf(s
, ",noacl");
1486 static int __init
ocfs2_init(void)
1492 ocfs2_print_version();
1494 status
= init_ocfs2_uptodate_cache();
1500 status
= ocfs2_initialize_mem_caches();
1506 ocfs2_wq
= create_singlethread_workqueue("ocfs2_wq");
1512 ocfs2_debugfs_root
= debugfs_create_dir("ocfs2", NULL
);
1513 if (!ocfs2_debugfs_root
) {
1515 mlog(ML_ERROR
, "Unable to create ocfs2 debugfs root.\n");
1518 status
= ocfs2_quota_setup();
1522 ocfs2_set_locking_protocol();
1524 status
= register_quota_format(&ocfs2_quota_format
);
1527 ocfs2_quota_shutdown();
1528 ocfs2_free_mem_caches();
1529 exit_ocfs2_uptodate_cache();
1535 return register_filesystem(&ocfs2_fs_type
);
1540 static void __exit
ocfs2_exit(void)
1544 ocfs2_quota_shutdown();
1547 flush_workqueue(ocfs2_wq
);
1548 destroy_workqueue(ocfs2_wq
);
1551 unregister_quota_format(&ocfs2_quota_format
);
1553 debugfs_remove(ocfs2_debugfs_root
);
1555 ocfs2_free_mem_caches();
1557 unregister_filesystem(&ocfs2_fs_type
);
1559 exit_ocfs2_uptodate_cache();
1564 static void ocfs2_put_super(struct super_block
*sb
)
1566 mlog_entry("(0x%p)\n", sb
);
1570 ocfs2_sync_blockdev(sb
);
1571 ocfs2_dismount_volume(sb
, 0);
1578 static int ocfs2_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
1580 struct ocfs2_super
*osb
;
1581 u32 numbits
, freebits
;
1583 struct ocfs2_dinode
*bm_lock
;
1584 struct buffer_head
*bh
= NULL
;
1585 struct inode
*inode
= NULL
;
1587 mlog_entry("(%p, %p)\n", dentry
->d_sb
, buf
);
1589 osb
= OCFS2_SB(dentry
->d_sb
);
1591 inode
= ocfs2_get_system_file_inode(osb
,
1592 GLOBAL_BITMAP_SYSTEM_INODE
,
1593 OCFS2_INVALID_SLOT
);
1595 mlog(ML_ERROR
, "failed to get bitmap inode\n");
1600 status
= ocfs2_inode_lock(inode
, &bh
, 0);
1606 bm_lock
= (struct ocfs2_dinode
*) bh
->b_data
;
1608 numbits
= le32_to_cpu(bm_lock
->id1
.bitmap1
.i_total
);
1609 freebits
= numbits
- le32_to_cpu(bm_lock
->id1
.bitmap1
.i_used
);
1611 buf
->f_type
= OCFS2_SUPER_MAGIC
;
1612 buf
->f_bsize
= dentry
->d_sb
->s_blocksize
;
1613 buf
->f_namelen
= OCFS2_MAX_FILENAME_LEN
;
1614 buf
->f_blocks
= ((sector_t
) numbits
) *
1615 (osb
->s_clustersize
>> osb
->sb
->s_blocksize_bits
);
1616 buf
->f_bfree
= ((sector_t
) freebits
) *
1617 (osb
->s_clustersize
>> osb
->sb
->s_blocksize_bits
);
1618 buf
->f_bavail
= buf
->f_bfree
;
1619 buf
->f_files
= numbits
;
1620 buf
->f_ffree
= freebits
;
1624 ocfs2_inode_unlock(inode
, 0);
1635 static void ocfs2_inode_init_once(void *data
)
1637 struct ocfs2_inode_info
*oi
= data
;
1640 oi
->ip_open_count
= 0;
1641 spin_lock_init(&oi
->ip_lock
);
1642 ocfs2_extent_map_init(&oi
->vfs_inode
);
1643 INIT_LIST_HEAD(&oi
->ip_io_markers
);
1644 oi
->ip_created_trans
= 0;
1645 oi
->ip_last_trans
= 0;
1646 oi
->ip_dir_start_lookup
= 0;
1648 init_rwsem(&oi
->ip_alloc_sem
);
1649 init_rwsem(&oi
->ip_xattr_sem
);
1650 mutex_init(&oi
->ip_io_mutex
);
1652 oi
->ip_blkno
= 0ULL;
1653 oi
->ip_clusters
= 0;
1655 ocfs2_lock_res_init_once(&oi
->ip_rw_lockres
);
1656 ocfs2_lock_res_init_once(&oi
->ip_inode_lockres
);
1657 ocfs2_lock_res_init_once(&oi
->ip_open_lockres
);
1659 ocfs2_metadata_cache_init(&oi
->vfs_inode
);
1661 inode_init_once(&oi
->vfs_inode
);
1664 static int ocfs2_initialize_mem_caches(void)
1666 ocfs2_inode_cachep
= kmem_cache_create("ocfs2_inode_cache",
1667 sizeof(struct ocfs2_inode_info
),
1669 (SLAB_HWCACHE_ALIGN
|SLAB_RECLAIM_ACCOUNT
|
1671 ocfs2_inode_init_once
);
1672 ocfs2_dquot_cachep
= kmem_cache_create("ocfs2_dquot_cache",
1673 sizeof(struct ocfs2_dquot
),
1675 (SLAB_HWCACHE_ALIGN
|SLAB_RECLAIM_ACCOUNT
|
1678 ocfs2_qf_chunk_cachep
= kmem_cache_create("ocfs2_qf_chunk_cache",
1679 sizeof(struct ocfs2_quota_chunk
),
1681 (SLAB_RECLAIM_ACCOUNT
|SLAB_MEM_SPREAD
),
1683 if (!ocfs2_inode_cachep
|| !ocfs2_dquot_cachep
||
1684 !ocfs2_qf_chunk_cachep
) {
1685 if (ocfs2_inode_cachep
)
1686 kmem_cache_destroy(ocfs2_inode_cachep
);
1687 if (ocfs2_dquot_cachep
)
1688 kmem_cache_destroy(ocfs2_dquot_cachep
);
1689 if (ocfs2_qf_chunk_cachep
)
1690 kmem_cache_destroy(ocfs2_qf_chunk_cachep
);
1697 static void ocfs2_free_mem_caches(void)
1699 if (ocfs2_inode_cachep
)
1700 kmem_cache_destroy(ocfs2_inode_cachep
);
1701 ocfs2_inode_cachep
= NULL
;
1703 if (ocfs2_dquot_cachep
)
1704 kmem_cache_destroy(ocfs2_dquot_cachep
);
1705 ocfs2_dquot_cachep
= NULL
;
1707 if (ocfs2_qf_chunk_cachep
)
1708 kmem_cache_destroy(ocfs2_qf_chunk_cachep
);
1709 ocfs2_qf_chunk_cachep
= NULL
;
1712 static int ocfs2_get_sector(struct super_block
*sb
,
1713 struct buffer_head
**bh
,
1717 if (!sb_set_blocksize(sb
, sect_size
)) {
1718 mlog(ML_ERROR
, "unable to set blocksize\n");
1722 *bh
= sb_getblk(sb
, block
);
1728 if (!buffer_dirty(*bh
))
1729 clear_buffer_uptodate(*bh
);
1731 ll_rw_block(READ
, 1, bh
);
1732 wait_on_buffer(*bh
);
1733 if (!buffer_uptodate(*bh
)) {
1743 static int ocfs2_mount_volume(struct super_block
*sb
)
1746 int unlock_super
= 0;
1747 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
1751 if (ocfs2_is_hard_readonly(osb
))
1754 status
= ocfs2_dlm_init(osb
);
1760 status
= ocfs2_super_lock(osb
, 1);
1767 /* This will load up the node map and add ourselves to it. */
1768 status
= ocfs2_find_slot(osb
);
1774 /* load all node-local system inodes */
1775 status
= ocfs2_init_local_system_inodes(osb
);
1781 status
= ocfs2_check_volume(osb
);
1787 status
= ocfs2_truncate_log_init(osb
);
1793 ocfs2_super_unlock(osb
, 1);
1799 static void ocfs2_dismount_volume(struct super_block
*sb
, int mnt_err
)
1801 int tmp
, hangup_needed
= 0;
1802 struct ocfs2_super
*osb
= NULL
;
1805 mlog_entry("(0x%p)\n", sb
);
1811 debugfs_remove(osb
->osb_ctxt
);
1813 ocfs2_disable_quotas(osb
);
1815 ocfs2_shutdown_local_alloc(osb
);
1817 ocfs2_truncate_log_shutdown(osb
);
1819 ocfs2_orphan_scan_stop(osb
);
1821 /* This will disable recovery and flush any recovery work. */
1822 ocfs2_recovery_exit(osb
);
1824 ocfs2_journal_shutdown(osb
);
1826 ocfs2_sync_blockdev(sb
);
1828 /* No cluster connection means we've failed during mount, so skip
1829 * all the steps which depended on that to complete. */
1831 tmp
= ocfs2_super_lock(osb
, 1);
1838 if (osb
->slot_num
!= OCFS2_INVALID_SLOT
)
1839 ocfs2_put_slot(osb
);
1842 ocfs2_super_unlock(osb
, 1);
1844 ocfs2_release_system_inodes(osb
);
1847 * If we're dismounting due to mount error, mount.ocfs2 will clean
1848 * up heartbeat. If we're a local mount, there is no heartbeat.
1849 * If we failed before we got a uuid_str yet, we can't stop
1850 * heartbeat. Otherwise, do it.
1852 if (!mnt_err
&& !ocfs2_mount_local(osb
) && osb
->uuid_str
)
1856 ocfs2_dlm_shutdown(osb
, hangup_needed
);
1858 ocfs2_blockcheck_stats_debugfs_remove(&osb
->osb_ecc_stats
);
1859 debugfs_remove(osb
->osb_debug_root
);
1862 ocfs2_cluster_hangup(osb
->uuid_str
, strlen(osb
->uuid_str
));
1864 atomic_set(&osb
->vol_state
, VOLUME_DISMOUNTED
);
1866 if (ocfs2_mount_local(osb
))
1867 snprintf(nodestr
, sizeof(nodestr
), "local");
1869 snprintf(nodestr
, sizeof(nodestr
), "%u", osb
->node_num
);
1871 printk(KERN_INFO
"ocfs2: Unmounting device (%s) on (node %s)\n",
1872 osb
->dev_str
, nodestr
);
1874 ocfs2_delete_osb(osb
);
1877 sb
->s_fs_info
= NULL
;
1880 static int ocfs2_setup_osb_uuid(struct ocfs2_super
*osb
, const unsigned char *uuid
,
1881 unsigned uuid_bytes
)
1886 BUG_ON(uuid_bytes
!= OCFS2_VOL_UUID_LEN
);
1888 osb
->uuid_str
= kzalloc(OCFS2_VOL_UUID_LEN
* 2 + 1, GFP_KERNEL
);
1889 if (osb
->uuid_str
== NULL
)
1892 for (i
= 0, ptr
= osb
->uuid_str
; i
< OCFS2_VOL_UUID_LEN
; i
++) {
1893 /* print with null */
1894 ret
= snprintf(ptr
, 3, "%02X", uuid
[i
]);
1895 if (ret
!= 2) /* drop super cleans up */
1897 /* then only advance past the last char */
1904 static int ocfs2_initialize_super(struct super_block
*sb
,
1905 struct buffer_head
*bh
,
1907 struct ocfs2_blockcheck_stats
*stats
)
1910 int i
, cbits
, bbits
;
1911 struct ocfs2_dinode
*di
= (struct ocfs2_dinode
*)bh
->b_data
;
1912 struct inode
*inode
= NULL
;
1913 struct ocfs2_journal
*journal
;
1914 __le32 uuid_net_key
;
1915 struct ocfs2_super
*osb
;
1919 osb
= kzalloc(sizeof(struct ocfs2_super
), GFP_KERNEL
);
1926 sb
->s_fs_info
= osb
;
1927 sb
->s_op
= &ocfs2_sops
;
1928 sb
->s_export_op
= &ocfs2_export_ops
;
1929 sb
->s_qcop
= &ocfs2_quotactl_ops
;
1930 sb
->dq_op
= &ocfs2_quota_operations
;
1931 sb
->s_xattr
= ocfs2_xattr_handlers
;
1932 sb
->s_time_gran
= 1;
1933 sb
->s_flags
|= MS_NOATIME
;
1934 /* this is needed to support O_LARGEFILE */
1935 cbits
= le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
);
1936 bbits
= le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
);
1937 sb
->s_maxbytes
= ocfs2_max_file_offset(bbits
, cbits
);
1939 osb
->osb_dx_mask
= (1 << (cbits
- bbits
)) - 1;
1941 for (i
= 0; i
< 3; i
++)
1942 osb
->osb_dx_seed
[i
] = le32_to_cpu(di
->id2
.i_super
.s_dx_seed
[i
]);
1943 osb
->osb_dx_seed
[3] = le32_to_cpu(di
->id2
.i_super
.s_uuid_hash
);
1946 /* Save off for ocfs2_rw_direct */
1947 osb
->s_sectsize_bits
= blksize_bits(sector_size
);
1948 BUG_ON(!osb
->s_sectsize_bits
);
1950 spin_lock_init(&osb
->dc_task_lock
);
1951 init_waitqueue_head(&osb
->dc_event
);
1952 osb
->dc_work_sequence
= 0;
1953 osb
->dc_wake_sequence
= 0;
1954 INIT_LIST_HEAD(&osb
->blocked_lock_list
);
1955 osb
->blocked_lock_count
= 0;
1956 spin_lock_init(&osb
->osb_lock
);
1957 spin_lock_init(&osb
->osb_xattr_lock
);
1958 ocfs2_init_inode_steal_slot(osb
);
1960 atomic_set(&osb
->alloc_stats
.moves
, 0);
1961 atomic_set(&osb
->alloc_stats
.local_data
, 0);
1962 atomic_set(&osb
->alloc_stats
.bitmap_data
, 0);
1963 atomic_set(&osb
->alloc_stats
.bg_allocs
, 0);
1964 atomic_set(&osb
->alloc_stats
.bg_extends
, 0);
1966 /* Copy the blockcheck stats from the superblock probe */
1967 osb
->osb_ecc_stats
= *stats
;
1969 ocfs2_init_node_maps(osb
);
1971 snprintf(osb
->dev_str
, sizeof(osb
->dev_str
), "%u,%u",
1972 MAJOR(osb
->sb
->s_dev
), MINOR(osb
->sb
->s_dev
));
1974 status
= ocfs2_recovery_init(osb
);
1976 mlog(ML_ERROR
, "Unable to initialize recovery state\n");
1981 status
= ocfs2_orphan_scan_init(osb
);
1983 mlog(ML_ERROR
, "Unable to initialize delayed orphan scan\n");
1988 init_waitqueue_head(&osb
->checkpoint_event
);
1989 atomic_set(&osb
->needs_checkpoint
, 0);
1991 osb
->s_atime_quantum
= OCFS2_DEFAULT_ATIME_QUANTUM
;
1993 osb
->slot_num
= OCFS2_INVALID_SLOT
;
1995 osb
->s_xattr_inline_size
= le16_to_cpu(
1996 di
->id2
.i_super
.s_xattr_inline_size
);
1998 osb
->local_alloc_state
= OCFS2_LA_UNUSED
;
1999 osb
->local_alloc_bh
= NULL
;
2000 INIT_DELAYED_WORK(&osb
->la_enable_wq
, ocfs2_la_enable_worker
);
2002 init_waitqueue_head(&osb
->osb_mount_event
);
2004 osb
->vol_label
= kmalloc(OCFS2_MAX_VOL_LABEL_LEN
, GFP_KERNEL
);
2005 if (!osb
->vol_label
) {
2006 mlog(ML_ERROR
, "unable to alloc vol label\n");
2011 osb
->max_slots
= le16_to_cpu(di
->id2
.i_super
.s_max_slots
);
2012 if (osb
->max_slots
> OCFS2_MAX_SLOTS
|| osb
->max_slots
== 0) {
2013 mlog(ML_ERROR
, "Invalid number of node slots (%u)\n",
2018 mlog(0, "max_slots for this device: %u\n", osb
->max_slots
);
2020 osb
->slot_recovery_generations
=
2021 kcalloc(osb
->max_slots
, sizeof(*osb
->slot_recovery_generations
),
2023 if (!osb
->slot_recovery_generations
) {
2029 init_waitqueue_head(&osb
->osb_wipe_event
);
2030 osb
->osb_orphan_wipes
= kcalloc(osb
->max_slots
,
2031 sizeof(*osb
->osb_orphan_wipes
),
2033 if (!osb
->osb_orphan_wipes
) {
2039 osb
->s_feature_compat
=
2040 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_compat
);
2041 osb
->s_feature_ro_compat
=
2042 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_ro_compat
);
2043 osb
->s_feature_incompat
=
2044 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_incompat
);
2046 if ((i
= OCFS2_HAS_INCOMPAT_FEATURE(osb
->sb
, ~OCFS2_FEATURE_INCOMPAT_SUPP
))) {
2047 mlog(ML_ERROR
, "couldn't mount because of unsupported "
2048 "optional features (%x).\n", i
);
2052 if (!(osb
->sb
->s_flags
& MS_RDONLY
) &&
2053 (i
= OCFS2_HAS_RO_COMPAT_FEATURE(osb
->sb
, ~OCFS2_FEATURE_RO_COMPAT_SUPP
))) {
2054 mlog(ML_ERROR
, "couldn't mount RDWR because of "
2055 "unsupported optional features (%x).\n", i
);
2060 if (ocfs2_userspace_stack(osb
)) {
2061 memcpy(osb
->osb_cluster_stack
,
2062 OCFS2_RAW_SB(di
)->s_cluster_info
.ci_stack
,
2063 OCFS2_STACK_LABEL_LEN
);
2064 osb
->osb_cluster_stack
[OCFS2_STACK_LABEL_LEN
] = '\0';
2065 if (strlen(osb
->osb_cluster_stack
) != OCFS2_STACK_LABEL_LEN
) {
2067 "couldn't mount because of an invalid "
2068 "cluster stack label (%s) \n",
2069 osb
->osb_cluster_stack
);
2074 /* The empty string is identical with classic tools that
2075 * don't know about s_cluster_info. */
2076 osb
->osb_cluster_stack
[0] = '\0';
2079 get_random_bytes(&osb
->s_next_generation
, sizeof(u32
));
2082 * This should be done in ocfs2_journal_init(), but unknown
2083 * ordering issues will cause the filesystem to crash.
2084 * If anyone wants to figure out what part of the code
2085 * refers to osb->journal before ocfs2_journal_init() is run,
2088 /* initialize our journal structure */
2090 journal
= kzalloc(sizeof(struct ocfs2_journal
), GFP_KERNEL
);
2092 mlog(ML_ERROR
, "unable to alloc journal\n");
2096 osb
->journal
= journal
;
2097 journal
->j_osb
= osb
;
2099 atomic_set(&journal
->j_num_trans
, 0);
2100 init_rwsem(&journal
->j_trans_barrier
);
2101 init_waitqueue_head(&journal
->j_checkpointed
);
2102 spin_lock_init(&journal
->j_lock
);
2103 journal
->j_trans_id
= (unsigned long) 1;
2104 INIT_LIST_HEAD(&journal
->j_la_cleanups
);
2105 INIT_WORK(&journal
->j_recovery_work
, ocfs2_complete_recovery
);
2106 journal
->j_state
= OCFS2_JOURNAL_FREE
;
2108 INIT_WORK(&osb
->dentry_lock_work
, ocfs2_drop_dl_inodes
);
2109 osb
->dentry_lock_list
= NULL
;
2111 /* get some pseudo constants for clustersize bits */
2112 osb
->s_clustersize_bits
=
2113 le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
);
2114 osb
->s_clustersize
= 1 << osb
->s_clustersize_bits
;
2115 mlog(0, "clusterbits=%d\n", osb
->s_clustersize_bits
);
2117 if (osb
->s_clustersize
< OCFS2_MIN_CLUSTERSIZE
||
2118 osb
->s_clustersize
> OCFS2_MAX_CLUSTERSIZE
) {
2119 mlog(ML_ERROR
, "Volume has invalid cluster size (%d)\n",
2120 osb
->s_clustersize
);
2125 if (ocfs2_clusters_to_blocks(osb
->sb
, le32_to_cpu(di
->i_clusters
) - 1)
2127 mlog(ML_ERROR
, "Volume might try to write to blocks beyond "
2128 "what jbd can address in 32 bits.\n");
2133 if (ocfs2_setup_osb_uuid(osb
, di
->id2
.i_super
.s_uuid
,
2134 sizeof(di
->id2
.i_super
.s_uuid
))) {
2135 mlog(ML_ERROR
, "Out of memory trying to setup our uuid.\n");
2140 memcpy(&uuid_net_key
, di
->id2
.i_super
.s_uuid
, sizeof(uuid_net_key
));
2142 strncpy(osb
->vol_label
, di
->id2
.i_super
.s_label
, 63);
2143 osb
->vol_label
[63] = '\0';
2144 osb
->root_blkno
= le64_to_cpu(di
->id2
.i_super
.s_root_blkno
);
2145 osb
->system_dir_blkno
= le64_to_cpu(di
->id2
.i_super
.s_system_dir_blkno
);
2146 osb
->first_cluster_group_blkno
=
2147 le64_to_cpu(di
->id2
.i_super
.s_first_cluster_group
);
2148 osb
->fs_generation
= le32_to_cpu(di
->i_fs_generation
);
2149 osb
->uuid_hash
= le32_to_cpu(di
->id2
.i_super
.s_uuid_hash
);
2150 mlog(0, "vol_label: %s\n", osb
->vol_label
);
2151 mlog(0, "uuid: %s\n", osb
->uuid_str
);
2152 mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
2153 (unsigned long long)osb
->root_blkno
,
2154 (unsigned long long)osb
->system_dir_blkno
);
2156 osb
->osb_dlm_debug
= ocfs2_new_dlm_debug();
2157 if (!osb
->osb_dlm_debug
) {
2163 atomic_set(&osb
->vol_state
, VOLUME_INIT
);
2165 /* load root, system_dir, and all global system inodes */
2166 status
= ocfs2_init_global_system_inodes(osb
);
2175 inode
= ocfs2_get_system_file_inode(osb
, GLOBAL_BITMAP_SYSTEM_INODE
,
2176 OCFS2_INVALID_SLOT
);
2183 osb
->bitmap_blkno
= OCFS2_I(inode
)->ip_blkno
;
2186 osb
->bitmap_cpg
= ocfs2_group_bitmap_size(sb
) * 8;
2188 status
= ocfs2_init_slot_info(osb
);
2200 * will return: -EAGAIN if it is ok to keep searching for superblocks
2201 * -EINVAL if there is a bad superblock
2204 static int ocfs2_verify_volume(struct ocfs2_dinode
*di
,
2205 struct buffer_head
*bh
,
2207 struct ocfs2_blockcheck_stats
*stats
)
2209 int status
= -EAGAIN
;
2213 if (memcmp(di
->i_signature
, OCFS2_SUPER_BLOCK_SIGNATURE
,
2214 strlen(OCFS2_SUPER_BLOCK_SIGNATURE
)) == 0) {
2215 /* We have to do a raw check of the feature here */
2216 if (le32_to_cpu(di
->id2
.i_super
.s_feature_incompat
) &
2217 OCFS2_FEATURE_INCOMPAT_META_ECC
) {
2218 status
= ocfs2_block_check_validate(bh
->b_data
,
2226 if ((1 << le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
)) != blksz
) {
2227 mlog(ML_ERROR
, "found superblock with incorrect block "
2228 "size: found %u, should be %u\n",
2229 1 << le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
),
2231 } else if (le16_to_cpu(di
->id2
.i_super
.s_major_rev_level
) !=
2232 OCFS2_MAJOR_REV_LEVEL
||
2233 le16_to_cpu(di
->id2
.i_super
.s_minor_rev_level
) !=
2234 OCFS2_MINOR_REV_LEVEL
) {
2235 mlog(ML_ERROR
, "found superblock with bad version: "
2236 "found %u.%u, should be %u.%u\n",
2237 le16_to_cpu(di
->id2
.i_super
.s_major_rev_level
),
2238 le16_to_cpu(di
->id2
.i_super
.s_minor_rev_level
),
2239 OCFS2_MAJOR_REV_LEVEL
,
2240 OCFS2_MINOR_REV_LEVEL
);
2241 } else if (bh
->b_blocknr
!= le64_to_cpu(di
->i_blkno
)) {
2242 mlog(ML_ERROR
, "bad block number on superblock: "
2243 "found %llu, should be %llu\n",
2244 (unsigned long long)le64_to_cpu(di
->i_blkno
),
2245 (unsigned long long)bh
->b_blocknr
);
2246 } else if (le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
) < 12 ||
2247 le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
) > 20) {
2248 mlog(ML_ERROR
, "bad cluster size found: %u\n",
2249 1 << le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
));
2250 } else if (!le64_to_cpu(di
->id2
.i_super
.s_root_blkno
)) {
2251 mlog(ML_ERROR
, "bad root_blkno: 0\n");
2252 } else if (!le64_to_cpu(di
->id2
.i_super
.s_system_dir_blkno
)) {
2253 mlog(ML_ERROR
, "bad system_dir_blkno: 0\n");
2254 } else if (le16_to_cpu(di
->id2
.i_super
.s_max_slots
) > OCFS2_MAX_SLOTS
) {
2256 "Superblock slots found greater than file system "
2257 "maximum: found %u, max %u\n",
2258 le16_to_cpu(di
->id2
.i_super
.s_max_slots
),
2271 static int ocfs2_check_volume(struct ocfs2_super
*osb
)
2276 struct ocfs2_dinode
*local_alloc
= NULL
; /* only used if we
2282 /* Init our journal object. */
2283 status
= ocfs2_journal_init(osb
->journal
, &dirty
);
2285 mlog(ML_ERROR
, "Could not initialize journal!\n");
2289 /* If the journal was unmounted cleanly then we don't want to
2290 * recover anything. Otherwise, journal_load will do that
2291 * dirty work for us :) */
2293 status
= ocfs2_journal_wipe(osb
->journal
, 0);
2299 mlog(ML_NOTICE
, "File system was not unmounted cleanly, "
2300 "recovering volume.\n");
2303 local
= ocfs2_mount_local(osb
);
2305 /* will play back anything left in the journal. */
2306 status
= ocfs2_journal_load(osb
->journal
, local
, dirty
);
2308 mlog(ML_ERROR
, "ocfs2 journal load failed! %d\n", status
);
2313 /* recover my local alloc if we didn't unmount cleanly. */
2314 status
= ocfs2_begin_local_alloc_recovery(osb
,
2321 /* we complete the recovery process after we've marked
2322 * ourselves as mounted. */
2325 mlog(0, "Journal loaded.\n");
2327 status
= ocfs2_load_local_alloc(osb
);
2334 /* Recovery will be completed after we've mounted the
2335 * rest of the volume. */
2337 osb
->local_alloc_copy
= local_alloc
;
2341 /* go through each journal, trylock it and if you get the
2342 * lock, and it's marked as dirty, set the bit in the recover
2343 * map and launch a recovery thread for it. */
2344 status
= ocfs2_mark_dead_nodes(osb
);
2350 status
= ocfs2_compute_replay_slots(osb
);
2363 * The routine gets called from dismount or close whenever a dismount on
2364 * volume is requested and the osb open count becomes 1.
2365 * It will remove the osb from the global list and also free up all the
2366 * initialized resources and fileobject.
2368 static void ocfs2_delete_osb(struct ocfs2_super
*osb
)
2372 /* This function assumes that the caller has the main osb resource */
2374 ocfs2_free_slot_info(osb
);
2376 kfree(osb
->osb_orphan_wipes
);
2377 kfree(osb
->slot_recovery_generations
);
2379 * This belongs in journal shutdown, but because we have to
2380 * allocate osb->journal at the start of ocfs2_initalize_osb(),
2383 kfree(osb
->journal
);
2384 if (osb
->local_alloc_copy
)
2385 kfree(osb
->local_alloc_copy
);
2386 kfree(osb
->uuid_str
);
2387 ocfs2_put_dlm_debug(osb
->osb_dlm_debug
);
2388 memset(osb
, 0, sizeof(struct ocfs2_super
));
2393 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2394 * panic(). We do not support continue-on-error operation. */
2395 static void ocfs2_handle_error(struct super_block
*sb
)
2397 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
2399 if (osb
->s_mount_opt
& OCFS2_MOUNT_ERRORS_PANIC
)
2400 panic("OCFS2: (device %s): panic forced after error\n",
2403 ocfs2_set_osb_flag(osb
, OCFS2_OSB_ERROR_FS
);
2405 if (sb
->s_flags
& MS_RDONLY
&&
2406 (ocfs2_is_soft_readonly(osb
) ||
2407 ocfs2_is_hard_readonly(osb
)))
2410 printk(KERN_CRIT
"File system is now read-only due to the potential "
2411 "of on-disk corruption. Please run fsck.ocfs2 once the file "
2412 "system is unmounted.\n");
2413 sb
->s_flags
|= MS_RDONLY
;
2414 ocfs2_set_ro_flag(osb
, 0);
2417 static char error_buf
[1024];
2419 void __ocfs2_error(struct super_block
*sb
,
2420 const char *function
,
2421 const char *fmt
, ...)
2425 va_start(args
, fmt
);
2426 vsnprintf(error_buf
, sizeof(error_buf
), fmt
, args
);
2429 /* Not using mlog here because we want to show the actual
2430 * function the error came from. */
2431 printk(KERN_CRIT
"OCFS2: ERROR (device %s): %s: %s\n",
2432 sb
->s_id
, function
, error_buf
);
2434 ocfs2_handle_error(sb
);
2437 /* Handle critical errors. This is intentionally more drastic than
2438 * ocfs2_handle_error, so we only use for things like journal errors,
2440 void __ocfs2_abort(struct super_block
* sb
,
2441 const char *function
,
2442 const char *fmt
, ...)
2446 va_start(args
, fmt
);
2447 vsnprintf(error_buf
, sizeof(error_buf
), fmt
, args
);
2450 printk(KERN_CRIT
"OCFS2: abort (device %s): %s: %s\n",
2451 sb
->s_id
, function
, error_buf
);
2453 /* We don't have the cluster support yet to go straight to
2454 * hard readonly in here. Until then, we want to keep
2455 * ocfs2_abort() so that we can at least mark critical
2458 * TODO: This should abort the journal and alert other nodes
2459 * that our slot needs recovery. */
2461 /* Force a panic(). This stinks, but it's better than letting
2462 * things continue without having a proper hard readonly
2464 OCFS2_SB(sb
)->s_mount_opt
|= OCFS2_MOUNT_ERRORS_PANIC
;
2465 ocfs2_handle_error(sb
);
2468 module_init(ocfs2_init
);
2469 module_exit(ocfs2_exit
);