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/init.h>
32 #include <linux/random.h>
33 #include <linux/statfs.h>
34 #include <linux/moduleparam.h>
35 #include <linux/blkdev.h>
36 #include <linux/socket.h>
37 #include <linux/inet.h>
38 #include <linux/parser.h>
39 #include <linux/crc32.h>
40 #include <linux/debugfs.h>
41 #include <linux/mount.h>
42 #include <linux/seq_file.h>
43 #include <linux/quotaops.h>
44 #include <linux/smp_lock.h>
46 #define MLOG_MASK_PREFIX ML_SUPER
47 #include <cluster/masklog.h>
51 /* this should be the only file to include a version 1 header */
52 #include "ocfs1_fs_compat.h"
55 #include "blockcheck.h"
58 #include "extent_map.h"
59 #include "heartbeat.h"
62 #include "localalloc.h"
71 #include "refcounttree.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_check_set_options(struct super_block
*sb
,
104 struct mount_options
*options
);
105 static int ocfs2_show_options(struct seq_file
*s
, struct vfsmount
*mnt
);
106 static void ocfs2_put_super(struct super_block
*sb
);
107 static int ocfs2_mount_volume(struct super_block
*sb
);
108 static int ocfs2_remount(struct super_block
*sb
, int *flags
, char *data
);
109 static void ocfs2_dismount_volume(struct super_block
*sb
, int mnt_err
);
110 static int ocfs2_initialize_mem_caches(void);
111 static void ocfs2_free_mem_caches(void);
112 static void ocfs2_delete_osb(struct ocfs2_super
*osb
);
114 static int ocfs2_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
116 static int ocfs2_sync_fs(struct super_block
*sb
, int wait
);
118 static int ocfs2_init_global_system_inodes(struct ocfs2_super
*osb
);
119 static int ocfs2_init_local_system_inodes(struct ocfs2_super
*osb
);
120 static void ocfs2_release_system_inodes(struct ocfs2_super
*osb
);
121 static int ocfs2_check_volume(struct ocfs2_super
*osb
);
122 static int ocfs2_verify_volume(struct ocfs2_dinode
*di
,
123 struct buffer_head
*bh
,
125 struct ocfs2_blockcheck_stats
*stats
);
126 static int ocfs2_initialize_super(struct super_block
*sb
,
127 struct buffer_head
*bh
,
129 struct ocfs2_blockcheck_stats
*stats
);
130 static int ocfs2_get_sector(struct super_block
*sb
,
131 struct buffer_head
**bh
,
134 static struct inode
*ocfs2_alloc_inode(struct super_block
*sb
);
135 static void ocfs2_destroy_inode(struct inode
*inode
);
136 static int ocfs2_susp_quotas(struct ocfs2_super
*osb
, int unsuspend
);
137 static int ocfs2_enable_quotas(struct ocfs2_super
*osb
);
138 static void ocfs2_disable_quotas(struct ocfs2_super
*osb
);
140 static const struct super_operations ocfs2_sops
= {
141 .statfs
= ocfs2_statfs
,
142 .alloc_inode
= ocfs2_alloc_inode
,
143 .destroy_inode
= ocfs2_destroy_inode
,
144 .drop_inode
= ocfs2_drop_inode
,
145 .clear_inode
= ocfs2_clear_inode
,
146 .delete_inode
= ocfs2_delete_inode
,
147 .sync_fs
= ocfs2_sync_fs
,
148 .put_super
= ocfs2_put_super
,
149 .remount_fs
= ocfs2_remount
,
150 .show_options
= ocfs2_show_options
,
151 .quota_read
= ocfs2_quota_read
,
152 .quota_write
= ocfs2_quota_write
,
181 static const match_table_t tokens
= {
182 {Opt_barrier
, "barrier=%u"},
183 {Opt_err_panic
, "errors=panic"},
184 {Opt_err_ro
, "errors=remount-ro"},
186 {Opt_nointr
, "nointr"},
187 {Opt_hb_none
, OCFS2_HB_NONE
},
188 {Opt_hb_local
, OCFS2_HB_LOCAL
},
189 {Opt_data_ordered
, "data=ordered"},
190 {Opt_data_writeback
, "data=writeback"},
191 {Opt_atime_quantum
, "atime_quantum=%u"},
192 {Opt_slot
, "preferred_slot=%u"},
193 {Opt_commit
, "commit=%u"},
194 {Opt_localalloc
, "localalloc=%d"},
195 {Opt_localflocks
, "localflocks"},
196 {Opt_stack
, "cluster_stack=%s"},
197 {Opt_user_xattr
, "user_xattr"},
198 {Opt_nouser_xattr
, "nouser_xattr"},
199 {Opt_inode64
, "inode64"},
201 {Opt_noacl
, "noacl"},
202 {Opt_usrquota
, "usrquota"},
203 {Opt_grpquota
, "grpquota"},
207 #ifdef CONFIG_DEBUG_FS
208 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
= &osb
->osb_orphan_scan
;
215 out
+= snprintf(buf
+ out
, len
- out
,
216 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n",
217 "Device", osb
->dev_str
, osb
->uuid_str
,
218 osb
->fs_generation
, osb
->vol_label
);
220 out
+= snprintf(buf
+ out
, len
- out
,
221 "%10s => State: %d Flags: 0x%lX\n", "Volume",
222 atomic_read(&osb
->vol_state
), osb
->osb_flags
);
224 out
+= snprintf(buf
+ out
, len
- out
,
225 "%10s => Block: %lu Cluster: %d\n", "Sizes",
226 osb
->sb
->s_blocksize
, osb
->s_clustersize
);
228 out
+= snprintf(buf
+ out
, len
- out
,
229 "%10s => Compat: 0x%X Incompat: 0x%X "
231 "Features", osb
->s_feature_compat
,
232 osb
->s_feature_incompat
, osb
->s_feature_ro_compat
);
234 out
+= snprintf(buf
+ out
, len
- out
,
235 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount",
236 osb
->s_mount_opt
, osb
->s_atime_quantum
);
239 out
+= snprintf(buf
+ out
, len
- out
,
240 "%10s => Stack: %s Name: %*s "
241 "Version: %d.%d\n", "Cluster",
242 (*osb
->osb_cluster_stack
== '\0' ?
243 "o2cb" : osb
->osb_cluster_stack
),
244 cconn
->cc_namelen
, cconn
->cc_name
,
245 cconn
->cc_version
.pv_major
,
246 cconn
->cc_version
.pv_minor
);
249 spin_lock(&osb
->dc_task_lock
);
250 out
+= snprintf(buf
+ out
, len
- out
,
251 "%10s => Pid: %d Count: %lu WakeSeq: %lu "
252 "WorkSeq: %lu\n", "DownCnvt",
253 (osb
->dc_task
? task_pid_nr(osb
->dc_task
) : -1),
254 osb
->blocked_lock_count
, osb
->dc_wake_sequence
,
255 osb
->dc_work_sequence
);
256 spin_unlock(&osb
->dc_task_lock
);
258 spin_lock(&osb
->osb_lock
);
259 out
+= snprintf(buf
+ out
, len
- out
, "%10s => Pid: %d Nodes:",
261 (osb
->recovery_thread_task
?
262 task_pid_nr(osb
->recovery_thread_task
) : -1));
263 if (rm
->rm_used
== 0)
264 out
+= snprintf(buf
+ out
, len
- out
, " None\n");
266 for (i
= 0; i
< rm
->rm_used
; i
++)
267 out
+= snprintf(buf
+ out
, len
- out
, " %d",
269 out
+= snprintf(buf
+ out
, len
- out
, "\n");
271 spin_unlock(&osb
->osb_lock
);
273 out
+= snprintf(buf
+ out
, len
- out
,
274 "%10s => Pid: %d Interval: %lu Needs: %d\n", "Commit",
275 (osb
->commit_task
? task_pid_nr(osb
->commit_task
) : -1),
276 osb
->osb_commit_interval
,
277 atomic_read(&osb
->needs_checkpoint
));
279 out
+= snprintf(buf
+ out
, len
- out
,
280 "%10s => State: %d TxnId: %lu NumTxns: %d\n",
281 "Journal", osb
->journal
->j_state
,
282 osb
->journal
->j_trans_id
,
283 atomic_read(&osb
->journal
->j_num_trans
));
285 out
+= snprintf(buf
+ out
, len
- out
,
286 "%10s => GlobalAllocs: %d LocalAllocs: %d "
287 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
289 atomic_read(&osb
->alloc_stats
.bitmap_data
),
290 atomic_read(&osb
->alloc_stats
.local_data
),
291 atomic_read(&osb
->alloc_stats
.bg_allocs
),
292 atomic_read(&osb
->alloc_stats
.moves
),
293 atomic_read(&osb
->alloc_stats
.bg_extends
));
295 out
+= snprintf(buf
+ out
, len
- out
,
296 "%10s => State: %u Descriptor: %llu Size: %u bits "
297 "Default: %u bits\n",
298 "LocalAlloc", osb
->local_alloc_state
,
299 (unsigned long long)osb
->la_last_gd
,
300 osb
->local_alloc_bits
, osb
->local_alloc_default_bits
);
302 spin_lock(&osb
->osb_lock
);
303 out
+= snprintf(buf
+ out
, len
- out
,
304 "%10s => Slot: %d NumStolen: %d\n", "Steal",
305 osb
->s_inode_steal_slot
,
306 atomic_read(&osb
->s_num_inodes_stolen
));
307 spin_unlock(&osb
->osb_lock
);
309 out
+= snprintf(buf
+ out
, len
- out
, "OrphanScan => ");
310 out
+= snprintf(buf
+ out
, len
- out
, "Local: %u Global: %u ",
311 os
->os_count
, os
->os_seqno
);
312 out
+= snprintf(buf
+ out
, len
- out
, " Last Scan: ");
313 if (atomic_read(&os
->os_state
) == ORPHAN_SCAN_INACTIVE
)
314 out
+= snprintf(buf
+ out
, len
- out
, "Disabled\n");
316 out
+= snprintf(buf
+ out
, len
- out
, "%lu seconds ago\n",
317 (get_seconds() - os
->os_scantime
.tv_sec
));
319 out
+= snprintf(buf
+ out
, len
- out
, "%10s => %3s %10s\n",
320 "Slots", "Num", "RecoGen");
321 for (i
= 0; i
< osb
->max_slots
; ++i
) {
322 out
+= snprintf(buf
+ out
, len
- out
,
323 "%10s %c %3d %10d\n",
325 (i
== osb
->slot_num
? '*' : ' '),
326 i
, osb
->slot_recovery_generations
[i
]);
332 static int ocfs2_osb_debug_open(struct inode
*inode
, struct file
*file
)
334 struct ocfs2_super
*osb
= inode
->i_private
;
337 buf
= kmalloc(PAGE_SIZE
, GFP_KERNEL
);
341 i_size_write(inode
, ocfs2_osb_dump(osb
, buf
, PAGE_SIZE
));
343 file
->private_data
= buf
;
350 static int ocfs2_debug_release(struct inode
*inode
, struct file
*file
)
352 kfree(file
->private_data
);
356 static ssize_t
ocfs2_debug_read(struct file
*file
, char __user
*buf
,
357 size_t nbytes
, loff_t
*ppos
)
359 return simple_read_from_buffer(buf
, nbytes
, ppos
, file
->private_data
,
360 i_size_read(file
->f_mapping
->host
));
363 static int ocfs2_osb_debug_open(struct inode
*inode
, struct file
*file
)
367 static int ocfs2_debug_release(struct inode
*inode
, struct file
*file
)
371 static ssize_t
ocfs2_debug_read(struct file
*file
, char __user
*buf
,
372 size_t nbytes
, loff_t
*ppos
)
376 #endif /* CONFIG_DEBUG_FS */
378 static const struct file_operations ocfs2_osb_debug_fops
= {
379 .open
= ocfs2_osb_debug_open
,
380 .release
= ocfs2_debug_release
,
381 .read
= ocfs2_debug_read
,
382 .llseek
= generic_file_llseek
,
385 static int ocfs2_sync_fs(struct super_block
*sb
, int wait
)
389 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
391 if (ocfs2_is_hard_readonly(osb
))
395 status
= ocfs2_flush_truncate_log(osb
);
399 ocfs2_schedule_truncate_log_flush(osb
, 0);
402 if (jbd2_journal_start_commit(OCFS2_SB(sb
)->journal
->j_journal
,
405 jbd2_log_wait_commit(OCFS2_SB(sb
)->journal
->j_journal
,
411 static int ocfs2_need_system_inode(struct ocfs2_super
*osb
, int ino
)
413 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb
->sb
, OCFS2_FEATURE_RO_COMPAT_USRQUOTA
)
414 && (ino
== USER_QUOTA_SYSTEM_INODE
415 || ino
== LOCAL_USER_QUOTA_SYSTEM_INODE
))
417 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb
->sb
, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
)
418 && (ino
== GROUP_QUOTA_SYSTEM_INODE
419 || ino
== LOCAL_GROUP_QUOTA_SYSTEM_INODE
))
424 static int ocfs2_init_global_system_inodes(struct ocfs2_super
*osb
)
426 struct inode
*new = NULL
;
432 new = ocfs2_iget(osb
, osb
->root_blkno
, OCFS2_FI_FLAG_SYSFILE
, 0);
434 status
= PTR_ERR(new);
438 osb
->root_inode
= new;
440 new = ocfs2_iget(osb
, osb
->system_dir_blkno
, OCFS2_FI_FLAG_SYSFILE
, 0);
442 status
= PTR_ERR(new);
446 osb
->sys_root_inode
= new;
448 for (i
= OCFS2_FIRST_ONLINE_SYSTEM_INODE
;
449 i
<= OCFS2_LAST_GLOBAL_SYSTEM_INODE
; i
++) {
450 if (!ocfs2_need_system_inode(osb
, i
))
452 new = ocfs2_get_system_file_inode(osb
, i
, osb
->slot_num
);
454 ocfs2_release_system_inodes(osb
);
457 /* FIXME: Should ERROR_RO_FS */
458 mlog(ML_ERROR
, "Unable to load system inode %d, "
459 "possibly corrupt fs?", i
);
462 // the array now has one ref, so drop this one
471 static int ocfs2_init_local_system_inodes(struct ocfs2_super
*osb
)
473 struct inode
*new = NULL
;
479 for (i
= OCFS2_LAST_GLOBAL_SYSTEM_INODE
+ 1;
480 i
< NUM_SYSTEM_INODES
;
482 if (!ocfs2_need_system_inode(osb
, i
))
484 new = ocfs2_get_system_file_inode(osb
, i
, osb
->slot_num
);
486 ocfs2_release_system_inodes(osb
);
488 mlog(ML_ERROR
, "status=%d, sysfile=%d, slot=%d\n",
489 status
, i
, osb
->slot_num
);
492 /* the array now has one ref, so drop this one */
501 static void ocfs2_release_system_inodes(struct ocfs2_super
*osb
)
508 for (i
= 0; i
< NUM_SYSTEM_INODES
; i
++) {
509 inode
= osb
->system_inodes
[i
];
512 osb
->system_inodes
[i
] = NULL
;
516 inode
= osb
->sys_root_inode
;
519 osb
->sys_root_inode
= NULL
;
522 inode
= osb
->root_inode
;
525 osb
->root_inode
= NULL
;
531 /* We're allocating fs objects, use GFP_NOFS */
532 static struct inode
*ocfs2_alloc_inode(struct super_block
*sb
)
534 struct ocfs2_inode_info
*oi
;
536 oi
= kmem_cache_alloc(ocfs2_inode_cachep
, GFP_NOFS
);
540 jbd2_journal_init_jbd_inode(&oi
->ip_jinode
, &oi
->vfs_inode
);
541 return &oi
->vfs_inode
;
544 static void ocfs2_destroy_inode(struct inode
*inode
)
546 kmem_cache_free(ocfs2_inode_cachep
, OCFS2_I(inode
));
549 static unsigned long long ocfs2_max_file_offset(unsigned int bbits
,
552 unsigned int bytes
= 1 << cbits
;
553 unsigned int trim
= bytes
;
554 unsigned int bitshift
= 32;
557 * i_size and all block offsets in ocfs2 are always 64 bits
558 * wide. i_clusters is 32 bits, in cluster-sized units. So on
559 * 64 bit platforms, cluster size will be the limiting factor.
562 #if BITS_PER_LONG == 32
563 # if defined(CONFIG_LBDAF)
564 BUILD_BUG_ON(sizeof(sector_t
) != 8);
566 * We might be limited by page cache size.
568 if (bytes
> PAGE_CACHE_SIZE
) {
569 bytes
= PAGE_CACHE_SIZE
;
572 * Shift by 31 here so that we don't get larger than
579 * We are limited by the size of sector_t. Use block size, as
580 * that's what we expose to the VFS.
589 * Trim by a whole cluster when we can actually approach the
590 * on-disk limits. Otherwise we can overflow i_clusters when
591 * an extent start is at the max offset.
593 return (((unsigned long long)bytes
) << bitshift
) - trim
;
596 static int ocfs2_remount(struct super_block
*sb
, int *flags
, char *data
)
598 int incompat_features
;
600 struct mount_options parsed_options
;
601 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
605 if (!ocfs2_parse_options(sb
, data
, &parsed_options
, 1) ||
606 !ocfs2_check_set_options(sb
, &parsed_options
)) {
611 if ((osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) !=
612 (parsed_options
.mount_opt
& OCFS2_MOUNT_HB_LOCAL
)) {
614 mlog(ML_ERROR
, "Cannot change heartbeat mode on remount\n");
618 if ((osb
->s_mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
) !=
619 (parsed_options
.mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
)) {
621 mlog(ML_ERROR
, "Cannot change data mode on remount\n");
625 /* Probably don't want this on remount; it might
626 * mess with other nodes */
627 if (!(osb
->s_mount_opt
& OCFS2_MOUNT_INODE64
) &&
628 (parsed_options
.mount_opt
& OCFS2_MOUNT_INODE64
)) {
630 mlog(ML_ERROR
, "Cannot enable inode64 on remount\n");
634 /* We're going to/from readonly mode. */
635 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
)) {
636 /* Disable quota accounting before remounting RO */
637 if (*flags
& MS_RDONLY
) {
638 ret
= ocfs2_susp_quotas(osb
, 0);
642 /* Lock here so the check of HARD_RO and the potential
643 * setting of SOFT_RO is atomic. */
644 spin_lock(&osb
->osb_lock
);
645 if (osb
->osb_flags
& OCFS2_OSB_HARD_RO
) {
646 mlog(ML_ERROR
, "Remount on readonly device is forbidden.\n");
651 if (*flags
& MS_RDONLY
) {
652 mlog(0, "Going to ro mode.\n");
653 sb
->s_flags
|= MS_RDONLY
;
654 osb
->osb_flags
|= OCFS2_OSB_SOFT_RO
;
656 mlog(0, "Making ro filesystem writeable.\n");
658 if (osb
->osb_flags
& OCFS2_OSB_ERROR_FS
) {
659 mlog(ML_ERROR
, "Cannot remount RDWR "
660 "filesystem due to previous errors.\n");
664 incompat_features
= OCFS2_HAS_RO_COMPAT_FEATURE(sb
, ~OCFS2_FEATURE_RO_COMPAT_SUPP
);
665 if (incompat_features
) {
666 mlog(ML_ERROR
, "Cannot remount RDWR because "
667 "of unsupported optional features "
668 "(%x).\n", incompat_features
);
672 sb
->s_flags
&= ~MS_RDONLY
;
673 osb
->osb_flags
&= ~OCFS2_OSB_SOFT_RO
;
676 spin_unlock(&osb
->osb_lock
);
677 /* Enable quota accounting after remounting RW */
678 if (!ret
&& !(*flags
& MS_RDONLY
)) {
679 if (sb_any_quota_suspended(sb
))
680 ret
= ocfs2_susp_quotas(osb
, 1);
682 ret
= ocfs2_enable_quotas(osb
);
684 /* Return back changes... */
685 spin_lock(&osb
->osb_lock
);
686 sb
->s_flags
|= MS_RDONLY
;
687 osb
->osb_flags
|= OCFS2_OSB_SOFT_RO
;
688 spin_unlock(&osb
->osb_lock
);
695 /* Only save off the new mount options in case of a successful
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
);
706 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
707 ((osb
->s_mount_opt
& OCFS2_MOUNT_POSIX_ACL
) ?
715 static int ocfs2_sb_probe(struct super_block
*sb
,
716 struct buffer_head
**bh
,
718 struct ocfs2_blockcheck_stats
*stats
)
721 struct ocfs1_vol_disk_hdr
*hdr
;
722 struct ocfs2_dinode
*di
;
728 *sector_size
= bdev_logical_block_size(sb
->s_bdev
);
729 if (*sector_size
> OCFS2_MAX_BLOCKSIZE
) {
730 mlog(ML_ERROR
, "Hardware sector size too large: %d (max=%d)\n",
731 *sector_size
, OCFS2_MAX_BLOCKSIZE
);
736 /* Can this really happen? */
737 if (*sector_size
< OCFS2_MIN_BLOCKSIZE
)
738 *sector_size
= OCFS2_MIN_BLOCKSIZE
;
740 /* check block zero for old format */
741 status
= ocfs2_get_sector(sb
, bh
, 0, *sector_size
);
746 hdr
= (struct ocfs1_vol_disk_hdr
*) (*bh
)->b_data
;
747 if (hdr
->major_version
== OCFS1_MAJOR_VERSION
) {
748 mlog(ML_ERROR
, "incompatible version: %u.%u\n",
749 hdr
->major_version
, hdr
->minor_version
);
752 if (memcmp(hdr
->signature
, OCFS1_VOLUME_SIGNATURE
,
753 strlen(OCFS1_VOLUME_SIGNATURE
)) == 0) {
754 mlog(ML_ERROR
, "incompatible volume signature: %8s\n",
761 mlog(ML_ERROR
, "This is an ocfs v1 filesystem which must be "
762 "upgraded before mounting with ocfs v2\n");
767 * Now check at magic offset for 512, 1024, 2048, 4096
768 * blocksizes. 4096 is the maximum blocksize because it is
769 * the minimum clustersize.
772 for (blksize
= *sector_size
;
773 blksize
<= OCFS2_MAX_BLOCKSIZE
;
775 tmpstat
= ocfs2_get_sector(sb
, bh
,
776 OCFS2_SUPER_BLOCK_BLKNO
,
783 di
= (struct ocfs2_dinode
*) (*bh
)->b_data
;
784 memset(stats
, 0, sizeof(struct ocfs2_blockcheck_stats
));
785 spin_lock_init(&stats
->b_lock
);
786 tmpstat
= ocfs2_verify_volume(di
, *bh
, blksize
, stats
);
791 if (tmpstat
!= -EAGAIN
) {
801 static int ocfs2_verify_heartbeat(struct ocfs2_super
*osb
)
803 if (ocfs2_mount_local(osb
)) {
804 if (osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) {
805 mlog(ML_ERROR
, "Cannot heartbeat on a locally "
806 "mounted device.\n");
811 if (ocfs2_userspace_stack(osb
)) {
812 if (osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) {
813 mlog(ML_ERROR
, "Userspace stack expected, but "
814 "o2cb heartbeat arguments passed to mount\n");
819 if (!(osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
)) {
820 if (!ocfs2_mount_local(osb
) && !ocfs2_is_hard_readonly(osb
) &&
821 !ocfs2_userspace_stack(osb
)) {
822 mlog(ML_ERROR
, "Heartbeat has to be started to mount "
823 "a read-write clustered device.\n");
832 * If we're using a userspace stack, mount should have passed
833 * a name that matches the disk. If not, mount should not
834 * have passed a stack.
836 static int ocfs2_verify_userspace_stack(struct ocfs2_super
*osb
,
837 struct mount_options
*mopt
)
839 if (!ocfs2_userspace_stack(osb
) && mopt
->cluster_stack
[0]) {
841 "cluster stack passed to mount, but this filesystem "
842 "does not support it\n");
846 if (ocfs2_userspace_stack(osb
) &&
847 strncmp(osb
->osb_cluster_stack
, mopt
->cluster_stack
,
848 OCFS2_STACK_LABEL_LEN
)) {
850 "cluster stack passed to mount (\"%s\") does not "
851 "match the filesystem (\"%s\")\n",
853 osb
->osb_cluster_stack
);
860 static int ocfs2_susp_quotas(struct ocfs2_super
*osb
, int unsuspend
)
863 struct super_block
*sb
= osb
->sb
;
864 unsigned int feature
[MAXQUOTAS
] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA
,
865 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
};
868 for (type
= 0; type
< MAXQUOTAS
; type
++) {
869 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb
, feature
[type
]))
872 status
= vfs_quota_enable(
873 sb_dqopt(sb
)->files
[type
],
877 status
= vfs_quota_disable(sb
, type
,
883 mlog(ML_ERROR
, "Failed to suspend/unsuspend quotas on "
884 "remount (error = %d).\n", status
);
888 static int ocfs2_enable_quotas(struct ocfs2_super
*osb
)
890 struct inode
*inode
[MAXQUOTAS
] = { NULL
, NULL
};
891 struct super_block
*sb
= osb
->sb
;
892 unsigned int feature
[MAXQUOTAS
] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA
,
893 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
};
894 unsigned int ino
[MAXQUOTAS
] = { LOCAL_USER_QUOTA_SYSTEM_INODE
,
895 LOCAL_GROUP_QUOTA_SYSTEM_INODE
};
899 sb_dqopt(sb
)->flags
|= DQUOT_QUOTA_SYS_FILE
| DQUOT_NEGATIVE_USAGE
;
900 for (type
= 0; type
< MAXQUOTAS
; type
++) {
901 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb
, feature
[type
]))
903 inode
[type
] = ocfs2_get_system_file_inode(osb
, ino
[type
],
909 status
= vfs_quota_enable(inode
[type
], type
, QFMT_OCFS2
,
910 DQUOT_USAGE_ENABLED
);
915 for (type
= 0; type
< MAXQUOTAS
; type
++)
919 ocfs2_disable_quotas(osb
);
920 for (type
= 0; type
< MAXQUOTAS
; type
++)
926 static void ocfs2_disable_quotas(struct ocfs2_super
*osb
)
930 struct super_block
*sb
= osb
->sb
;
932 /* We mostly ignore errors in this function because there's not much
933 * we can do when we see them */
934 for (type
= 0; type
< MAXQUOTAS
; type
++) {
935 if (!sb_has_quota_loaded(sb
, type
))
937 inode
= igrab(sb
->s_dquot
.files
[type
]);
938 /* Turn off quotas. This will remove all dquot structures from
939 * memory and so they will be automatically synced to global
941 vfs_quota_disable(sb
, type
, DQUOT_USAGE_ENABLED
|
942 DQUOT_LIMITS_ENABLED
);
949 /* Handle quota on quotactl */
950 static int ocfs2_quota_on(struct super_block
*sb
, int type
, int format_id
,
951 char *path
, int remount
)
953 unsigned int feature
[MAXQUOTAS
] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA
,
954 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
};
956 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb
, feature
[type
]))
960 return 0; /* Just ignore it has been handled in
962 return vfs_quota_enable(sb_dqopt(sb
)->files
[type
], type
,
963 format_id
, DQUOT_LIMITS_ENABLED
);
966 /* Handle quota off quotactl */
967 static int ocfs2_quota_off(struct super_block
*sb
, int type
, int remount
)
970 return 0; /* Ignore now and handle later in
972 return vfs_quota_disable(sb
, type
, DQUOT_LIMITS_ENABLED
);
975 static const struct quotactl_ops ocfs2_quotactl_ops
= {
976 .quota_on
= ocfs2_quota_on
,
977 .quota_off
= ocfs2_quota_off
,
978 .quota_sync
= vfs_quota_sync
,
979 .get_info
= vfs_get_dqinfo
,
980 .set_info
= vfs_set_dqinfo
,
981 .get_dqblk
= vfs_get_dqblk
,
982 .set_dqblk
= vfs_set_dqblk
,
985 static int ocfs2_fill_super(struct super_block
*sb
, void *data
, int silent
)
988 int status
, sector_size
;
989 struct mount_options parsed_options
;
990 struct inode
*inode
= NULL
;
991 struct ocfs2_super
*osb
= NULL
;
992 struct buffer_head
*bh
= NULL
;
994 struct ocfs2_blockcheck_stats stats
;
996 mlog_entry("%p, %p, %i", sb
, data
, silent
);
998 if (!ocfs2_parse_options(sb
, data
, &parsed_options
, 0)) {
1000 goto read_super_error
;
1003 /* probe for superblock */
1004 status
= ocfs2_sb_probe(sb
, &bh
, §or_size
, &stats
);
1006 mlog(ML_ERROR
, "superblock probe failed!\n");
1007 goto read_super_error
;
1010 status
= ocfs2_initialize_super(sb
, bh
, sector_size
, &stats
);
1014 goto read_super_error
;
1019 if (!ocfs2_check_set_options(sb
, &parsed_options
)) {
1021 goto read_super_error
;
1023 osb
->s_mount_opt
= parsed_options
.mount_opt
;
1024 osb
->s_atime_quantum
= parsed_options
.atime_quantum
;
1025 osb
->preferred_slot
= parsed_options
.slot
;
1026 osb
->osb_commit_interval
= parsed_options
.commit_interval
;
1027 osb
->local_alloc_default_bits
= ocfs2_megabytes_to_clusters(sb
, parsed_options
.localalloc_opt
);
1028 osb
->local_alloc_bits
= osb
->local_alloc_default_bits
;
1030 status
= ocfs2_verify_userspace_stack(osb
, &parsed_options
);
1032 goto read_super_error
;
1034 sb
->s_magic
= OCFS2_SUPER_MAGIC
;
1036 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
1037 ((osb
->s_mount_opt
& OCFS2_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
1039 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1041 if (bdev_read_only(sb
->s_bdev
)) {
1042 if (!(sb
->s_flags
& MS_RDONLY
)) {
1044 mlog(ML_ERROR
, "Readonly device detected but readonly "
1045 "mount was not specified.\n");
1046 goto read_super_error
;
1049 /* You should not be able to start a local heartbeat
1050 * on a readonly device. */
1051 if (osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) {
1053 mlog(ML_ERROR
, "Local heartbeat specified on readonly "
1055 goto read_super_error
;
1058 status
= ocfs2_check_journals_nolocks(osb
);
1060 if (status
== -EROFS
)
1061 mlog(ML_ERROR
, "Recovery required on readonly "
1062 "file system, but write access is "
1066 goto read_super_error
;
1069 ocfs2_set_ro_flag(osb
, 1);
1071 printk(KERN_NOTICE
"Readonly device detected. No cluster "
1072 "services will be utilized for this mount. Recovery "
1073 "will be skipped.\n");
1076 if (!ocfs2_is_hard_readonly(osb
)) {
1077 if (sb
->s_flags
& MS_RDONLY
)
1078 ocfs2_set_ro_flag(osb
, 0);
1081 status
= ocfs2_verify_heartbeat(osb
);
1084 goto read_super_error
;
1087 osb
->osb_debug_root
= debugfs_create_dir(osb
->uuid_str
,
1088 ocfs2_debugfs_root
);
1089 if (!osb
->osb_debug_root
) {
1091 mlog(ML_ERROR
, "Unable to create per-mount debugfs root.\n");
1092 goto read_super_error
;
1095 osb
->osb_ctxt
= debugfs_create_file("fs_state", S_IFREG
|S_IRUSR
,
1096 osb
->osb_debug_root
,
1098 &ocfs2_osb_debug_fops
);
1099 if (!osb
->osb_ctxt
) {
1102 goto read_super_error
;
1105 if (ocfs2_meta_ecc(osb
)) {
1106 status
= ocfs2_blockcheck_stats_debugfs_install(
1107 &osb
->osb_ecc_stats
,
1108 osb
->osb_debug_root
);
1111 "Unable to create blockcheck statistics "
1113 goto read_super_error
;
1117 status
= ocfs2_mount_volume(sb
);
1118 if (osb
->root_inode
)
1119 inode
= igrab(osb
->root_inode
);
1122 goto read_super_error
;
1127 goto read_super_error
;
1130 root
= d_alloc_root(inode
);
1134 goto read_super_error
;
1139 ocfs2_complete_mount_recovery(osb
);
1141 if (ocfs2_mount_local(osb
))
1142 snprintf(nodestr
, sizeof(nodestr
), "local");
1144 snprintf(nodestr
, sizeof(nodestr
), "%u", osb
->node_num
);
1146 printk(KERN_INFO
"ocfs2: Mounting device (%s) on (node %s, slot %d) "
1147 "with %s data mode.\n",
1148 osb
->dev_str
, nodestr
, osb
->slot_num
,
1149 osb
->s_mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
? "writeback" :
1152 atomic_set(&osb
->vol_state
, VOLUME_MOUNTED
);
1153 wake_up(&osb
->osb_mount_event
);
1155 /* Now we can initialize quotas because we can afford to wait
1156 * for cluster locks recovery now. That also means that truncation
1157 * log recovery can happen but that waits for proper quota setup */
1158 if (!(sb
->s_flags
& MS_RDONLY
)) {
1159 status
= ocfs2_enable_quotas(osb
);
1161 /* We have to err-out specially here because
1162 * s_root is already set */
1164 atomic_set(&osb
->vol_state
, VOLUME_DISABLED
);
1165 wake_up(&osb
->osb_mount_event
);
1171 ocfs2_complete_quota_recovery(osb
);
1173 /* Now we wake up again for processes waiting for quotas */
1174 atomic_set(&osb
->vol_state
, VOLUME_MOUNTED_QUOTAS
);
1175 wake_up(&osb
->osb_mount_event
);
1177 /* Start this when the mount is almost sure of being successful */
1178 ocfs2_orphan_scan_start(osb
);
1190 atomic_set(&osb
->vol_state
, VOLUME_DISABLED
);
1191 wake_up(&osb
->osb_mount_event
);
1192 ocfs2_dismount_volume(sb
, 1);
1199 static int ocfs2_get_sb(struct file_system_type
*fs_type
,
1201 const char *dev_name
,
1203 struct vfsmount
*mnt
)
1205 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, ocfs2_fill_super
,
1209 static void ocfs2_kill_sb(struct super_block
*sb
)
1211 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
1214 if (!osb
|| atomic_read(&osb
->vol_state
) == VOLUME_DISABLED
)
1217 /* Prevent further queueing of inode drop events */
1218 spin_lock(&dentry_list_lock
);
1219 ocfs2_set_osb_flag(osb
, OCFS2_OSB_DROP_DENTRY_LOCK_IMMED
);
1220 spin_unlock(&dentry_list_lock
);
1221 /* Wait for work to finish and/or remove it */
1222 cancel_work_sync(&osb
->dentry_lock_work
);
1224 kill_block_super(sb
);
1227 static struct file_system_type ocfs2_fs_type
= {
1228 .owner
= THIS_MODULE
,
1230 .get_sb
= ocfs2_get_sb
, /* is this called when we mount
1232 .kill_sb
= ocfs2_kill_sb
,
1234 .fs_flags
= FS_REQUIRES_DEV
|FS_RENAME_DOES_D_MOVE
,
1238 static int ocfs2_check_set_options(struct super_block
*sb
,
1239 struct mount_options
*options
)
1241 if (options
->mount_opt
& OCFS2_MOUNT_USRQUOTA
&&
1242 !OCFS2_HAS_RO_COMPAT_FEATURE(sb
,
1243 OCFS2_FEATURE_RO_COMPAT_USRQUOTA
)) {
1244 mlog(ML_ERROR
, "User quotas were requested, but this "
1245 "filesystem does not have the feature enabled.\n");
1248 if (options
->mount_opt
& OCFS2_MOUNT_GRPQUOTA
&&
1249 !OCFS2_HAS_RO_COMPAT_FEATURE(sb
,
1250 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
)) {
1251 mlog(ML_ERROR
, "Group quotas were requested, but this "
1252 "filesystem does not have the feature enabled.\n");
1255 if (options
->mount_opt
& OCFS2_MOUNT_POSIX_ACL
&&
1256 !OCFS2_HAS_INCOMPAT_FEATURE(sb
, OCFS2_FEATURE_INCOMPAT_XATTR
)) {
1257 mlog(ML_ERROR
, "ACL support requested but extended attributes "
1258 "feature is not enabled\n");
1261 /* No ACL setting specified? Use XATTR feature... */
1262 if (!(options
->mount_opt
& (OCFS2_MOUNT_POSIX_ACL
|
1263 OCFS2_MOUNT_NO_POSIX_ACL
))) {
1264 if (OCFS2_HAS_INCOMPAT_FEATURE(sb
, OCFS2_FEATURE_INCOMPAT_XATTR
))
1265 options
->mount_opt
|= OCFS2_MOUNT_POSIX_ACL
;
1267 options
->mount_opt
|= OCFS2_MOUNT_NO_POSIX_ACL
;
1272 static int ocfs2_parse_options(struct super_block
*sb
,
1274 struct mount_options
*mopt
,
1280 mlog_entry("remount: %d, options: \"%s\"\n", is_remount
,
1281 options
? options
: "(none)");
1283 mopt
->commit_interval
= 0;
1284 mopt
->mount_opt
= 0;
1285 mopt
->atime_quantum
= OCFS2_DEFAULT_ATIME_QUANTUM
;
1286 mopt
->slot
= OCFS2_INVALID_SLOT
;
1287 mopt
->localalloc_opt
= OCFS2_DEFAULT_LOCAL_ALLOC_SIZE
;
1288 mopt
->cluster_stack
[0] = '\0';
1295 while ((p
= strsep(&options
, ",")) != NULL
) {
1297 substring_t args
[MAX_OPT_ARGS
];
1302 token
= match_token(p
, tokens
, args
);
1305 mopt
->mount_opt
|= OCFS2_MOUNT_HB_LOCAL
;
1308 mopt
->mount_opt
&= ~OCFS2_MOUNT_HB_LOCAL
;
1311 if (match_int(&args
[0], &option
)) {
1316 mopt
->mount_opt
|= OCFS2_MOUNT_BARRIER
;
1318 mopt
->mount_opt
&= ~OCFS2_MOUNT_BARRIER
;
1321 mopt
->mount_opt
&= ~OCFS2_MOUNT_NOINTR
;
1324 mopt
->mount_opt
|= OCFS2_MOUNT_NOINTR
;
1327 mopt
->mount_opt
|= OCFS2_MOUNT_ERRORS_PANIC
;
1330 mopt
->mount_opt
&= ~OCFS2_MOUNT_ERRORS_PANIC
;
1332 case Opt_data_ordered
:
1333 mopt
->mount_opt
&= ~OCFS2_MOUNT_DATA_WRITEBACK
;
1335 case Opt_data_writeback
:
1336 mopt
->mount_opt
|= OCFS2_MOUNT_DATA_WRITEBACK
;
1338 case Opt_user_xattr
:
1339 mopt
->mount_opt
&= ~OCFS2_MOUNT_NOUSERXATTR
;
1341 case Opt_nouser_xattr
:
1342 mopt
->mount_opt
|= OCFS2_MOUNT_NOUSERXATTR
;
1344 case Opt_atime_quantum
:
1345 if (match_int(&args
[0], &option
)) {
1350 mopt
->atime_quantum
= option
;
1354 if (match_int(&args
[0], &option
)) {
1359 mopt
->slot
= (s16
)option
;
1363 if (match_int(&args
[0], &option
)) {
1370 option
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1371 mopt
->commit_interval
= HZ
* option
;
1373 case Opt_localalloc
:
1375 if (match_int(&args
[0], &option
)) {
1379 if (option
>= 0 && (option
<= ocfs2_local_alloc_size(sb
) * 8))
1380 mopt
->localalloc_opt
= option
;
1382 case Opt_localflocks
:
1384 * Changing this during remount could race
1385 * flock() requests, or "unbalance" existing
1386 * ones (e.g., a lock is taken in one mode but
1387 * dropped in the other). If users care enough
1388 * to flip locking modes during remount, we
1389 * could add a "local" flag to individual
1390 * flock structures for proper tracking of
1394 mopt
->mount_opt
|= OCFS2_MOUNT_LOCALFLOCKS
;
1397 /* Check both that the option we were passed
1398 * is of the right length and that it is a proper
1399 * string of the right length.
1401 if (((args
[0].to
- args
[0].from
) !=
1402 OCFS2_STACK_LABEL_LEN
) ||
1403 (strnlen(args
[0].from
,
1404 OCFS2_STACK_LABEL_LEN
) !=
1405 OCFS2_STACK_LABEL_LEN
)) {
1407 "Invalid cluster_stack option\n");
1411 memcpy(mopt
->cluster_stack
, args
[0].from
,
1412 OCFS2_STACK_LABEL_LEN
);
1413 mopt
->cluster_stack
[OCFS2_STACK_LABEL_LEN
] = '\0';
1416 mopt
->mount_opt
|= OCFS2_MOUNT_INODE64
;
1419 mopt
->mount_opt
|= OCFS2_MOUNT_USRQUOTA
;
1422 mopt
->mount_opt
|= OCFS2_MOUNT_GRPQUOTA
;
1425 mopt
->mount_opt
|= OCFS2_MOUNT_POSIX_ACL
;
1426 mopt
->mount_opt
&= ~OCFS2_MOUNT_NO_POSIX_ACL
;
1429 mopt
->mount_opt
|= OCFS2_MOUNT_NO_POSIX_ACL
;
1430 mopt
->mount_opt
&= ~OCFS2_MOUNT_POSIX_ACL
;
1434 "Unrecognized mount option \"%s\" "
1435 "or missing value\n", p
);
1448 static int ocfs2_show_options(struct seq_file
*s
, struct vfsmount
*mnt
)
1450 struct ocfs2_super
*osb
= OCFS2_SB(mnt
->mnt_sb
);
1451 unsigned long opts
= osb
->s_mount_opt
;
1452 unsigned int local_alloc_megs
;
1454 if (opts
& OCFS2_MOUNT_HB_LOCAL
)
1455 seq_printf(s
, ",_netdev,heartbeat=local");
1457 seq_printf(s
, ",heartbeat=none");
1459 if (opts
& OCFS2_MOUNT_NOINTR
)
1460 seq_printf(s
, ",nointr");
1462 if (opts
& OCFS2_MOUNT_DATA_WRITEBACK
)
1463 seq_printf(s
, ",data=writeback");
1465 seq_printf(s
, ",data=ordered");
1467 if (opts
& OCFS2_MOUNT_BARRIER
)
1468 seq_printf(s
, ",barrier=1");
1470 if (opts
& OCFS2_MOUNT_ERRORS_PANIC
)
1471 seq_printf(s
, ",errors=panic");
1473 seq_printf(s
, ",errors=remount-ro");
1475 if (osb
->preferred_slot
!= OCFS2_INVALID_SLOT
)
1476 seq_printf(s
, ",preferred_slot=%d", osb
->preferred_slot
);
1478 if (osb
->s_atime_quantum
!= OCFS2_DEFAULT_ATIME_QUANTUM
)
1479 seq_printf(s
, ",atime_quantum=%u", osb
->s_atime_quantum
);
1481 if (osb
->osb_commit_interval
)
1482 seq_printf(s
, ",commit=%u",
1483 (unsigned) (osb
->osb_commit_interval
/ HZ
));
1485 local_alloc_megs
= osb
->local_alloc_bits
>> (20 - osb
->s_clustersize_bits
);
1486 if (local_alloc_megs
!= OCFS2_DEFAULT_LOCAL_ALLOC_SIZE
)
1487 seq_printf(s
, ",localalloc=%d", local_alloc_megs
);
1489 if (opts
& OCFS2_MOUNT_LOCALFLOCKS
)
1490 seq_printf(s
, ",localflocks,");
1492 if (osb
->osb_cluster_stack
[0])
1493 seq_printf(s
, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN
,
1494 osb
->osb_cluster_stack
);
1495 if (opts
& OCFS2_MOUNT_USRQUOTA
)
1496 seq_printf(s
, ",usrquota");
1497 if (opts
& OCFS2_MOUNT_GRPQUOTA
)
1498 seq_printf(s
, ",grpquota");
1500 if (opts
& OCFS2_MOUNT_NOUSERXATTR
)
1501 seq_printf(s
, ",nouser_xattr");
1503 seq_printf(s
, ",user_xattr");
1505 if (opts
& OCFS2_MOUNT_INODE64
)
1506 seq_printf(s
, ",inode64");
1508 if (opts
& OCFS2_MOUNT_POSIX_ACL
)
1509 seq_printf(s
, ",acl");
1511 seq_printf(s
, ",noacl");
1516 static int __init
ocfs2_init(void)
1522 ocfs2_print_version();
1524 status
= init_ocfs2_uptodate_cache();
1530 status
= ocfs2_initialize_mem_caches();
1536 ocfs2_wq
= create_singlethread_workqueue("ocfs2_wq");
1542 ocfs2_debugfs_root
= debugfs_create_dir("ocfs2", NULL
);
1543 if (!ocfs2_debugfs_root
) {
1545 mlog(ML_ERROR
, "Unable to create ocfs2 debugfs root.\n");
1548 status
= ocfs2_quota_setup();
1552 ocfs2_set_locking_protocol();
1554 status
= register_quota_format(&ocfs2_quota_format
);
1557 ocfs2_quota_shutdown();
1558 ocfs2_free_mem_caches();
1559 exit_ocfs2_uptodate_cache();
1565 return register_filesystem(&ocfs2_fs_type
);
1570 static void __exit
ocfs2_exit(void)
1574 ocfs2_quota_shutdown();
1577 flush_workqueue(ocfs2_wq
);
1578 destroy_workqueue(ocfs2_wq
);
1581 unregister_quota_format(&ocfs2_quota_format
);
1583 debugfs_remove(ocfs2_debugfs_root
);
1585 ocfs2_free_mem_caches();
1587 unregister_filesystem(&ocfs2_fs_type
);
1589 exit_ocfs2_uptodate_cache();
1594 static void ocfs2_put_super(struct super_block
*sb
)
1596 mlog_entry("(0x%p)\n", sb
);
1600 ocfs2_sync_blockdev(sb
);
1601 ocfs2_dismount_volume(sb
, 0);
1608 static int ocfs2_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
1610 struct ocfs2_super
*osb
;
1611 u32 numbits
, freebits
;
1613 struct ocfs2_dinode
*bm_lock
;
1614 struct buffer_head
*bh
= NULL
;
1615 struct inode
*inode
= NULL
;
1617 mlog_entry("(%p, %p)\n", dentry
->d_sb
, buf
);
1619 osb
= OCFS2_SB(dentry
->d_sb
);
1621 inode
= ocfs2_get_system_file_inode(osb
,
1622 GLOBAL_BITMAP_SYSTEM_INODE
,
1623 OCFS2_INVALID_SLOT
);
1625 mlog(ML_ERROR
, "failed to get bitmap inode\n");
1630 status
= ocfs2_inode_lock(inode
, &bh
, 0);
1636 bm_lock
= (struct ocfs2_dinode
*) bh
->b_data
;
1638 numbits
= le32_to_cpu(bm_lock
->id1
.bitmap1
.i_total
);
1639 freebits
= numbits
- le32_to_cpu(bm_lock
->id1
.bitmap1
.i_used
);
1641 buf
->f_type
= OCFS2_SUPER_MAGIC
;
1642 buf
->f_bsize
= dentry
->d_sb
->s_blocksize
;
1643 buf
->f_namelen
= OCFS2_MAX_FILENAME_LEN
;
1644 buf
->f_blocks
= ((sector_t
) numbits
) *
1645 (osb
->s_clustersize
>> osb
->sb
->s_blocksize_bits
);
1646 buf
->f_bfree
= ((sector_t
) freebits
) *
1647 (osb
->s_clustersize
>> osb
->sb
->s_blocksize_bits
);
1648 buf
->f_bavail
= buf
->f_bfree
;
1649 buf
->f_files
= numbits
;
1650 buf
->f_ffree
= freebits
;
1654 ocfs2_inode_unlock(inode
, 0);
1665 static void ocfs2_inode_init_once(void *data
)
1667 struct ocfs2_inode_info
*oi
= data
;
1670 oi
->ip_open_count
= 0;
1671 spin_lock_init(&oi
->ip_lock
);
1672 ocfs2_extent_map_init(&oi
->vfs_inode
);
1673 INIT_LIST_HEAD(&oi
->ip_io_markers
);
1674 oi
->ip_dir_start_lookup
= 0;
1676 init_rwsem(&oi
->ip_alloc_sem
);
1677 init_rwsem(&oi
->ip_xattr_sem
);
1678 mutex_init(&oi
->ip_io_mutex
);
1680 oi
->ip_blkno
= 0ULL;
1681 oi
->ip_clusters
= 0;
1683 ocfs2_lock_res_init_once(&oi
->ip_rw_lockres
);
1684 ocfs2_lock_res_init_once(&oi
->ip_inode_lockres
);
1685 ocfs2_lock_res_init_once(&oi
->ip_open_lockres
);
1687 ocfs2_metadata_cache_init(INODE_CACHE(&oi
->vfs_inode
),
1688 &ocfs2_inode_caching_ops
);
1690 inode_init_once(&oi
->vfs_inode
);
1693 static int ocfs2_initialize_mem_caches(void)
1695 ocfs2_inode_cachep
= kmem_cache_create("ocfs2_inode_cache",
1696 sizeof(struct ocfs2_inode_info
),
1698 (SLAB_HWCACHE_ALIGN
|SLAB_RECLAIM_ACCOUNT
|
1700 ocfs2_inode_init_once
);
1701 ocfs2_dquot_cachep
= kmem_cache_create("ocfs2_dquot_cache",
1702 sizeof(struct ocfs2_dquot
),
1704 (SLAB_HWCACHE_ALIGN
|SLAB_RECLAIM_ACCOUNT
|
1707 ocfs2_qf_chunk_cachep
= kmem_cache_create("ocfs2_qf_chunk_cache",
1708 sizeof(struct ocfs2_quota_chunk
),
1710 (SLAB_RECLAIM_ACCOUNT
|SLAB_MEM_SPREAD
),
1712 if (!ocfs2_inode_cachep
|| !ocfs2_dquot_cachep
||
1713 !ocfs2_qf_chunk_cachep
) {
1714 if (ocfs2_inode_cachep
)
1715 kmem_cache_destroy(ocfs2_inode_cachep
);
1716 if (ocfs2_dquot_cachep
)
1717 kmem_cache_destroy(ocfs2_dquot_cachep
);
1718 if (ocfs2_qf_chunk_cachep
)
1719 kmem_cache_destroy(ocfs2_qf_chunk_cachep
);
1726 static void ocfs2_free_mem_caches(void)
1728 if (ocfs2_inode_cachep
)
1729 kmem_cache_destroy(ocfs2_inode_cachep
);
1730 ocfs2_inode_cachep
= NULL
;
1732 if (ocfs2_dquot_cachep
)
1733 kmem_cache_destroy(ocfs2_dquot_cachep
);
1734 ocfs2_dquot_cachep
= NULL
;
1736 if (ocfs2_qf_chunk_cachep
)
1737 kmem_cache_destroy(ocfs2_qf_chunk_cachep
);
1738 ocfs2_qf_chunk_cachep
= NULL
;
1741 static int ocfs2_get_sector(struct super_block
*sb
,
1742 struct buffer_head
**bh
,
1746 if (!sb_set_blocksize(sb
, sect_size
)) {
1747 mlog(ML_ERROR
, "unable to set blocksize\n");
1751 *bh
= sb_getblk(sb
, block
);
1757 if (!buffer_dirty(*bh
))
1758 clear_buffer_uptodate(*bh
);
1760 ll_rw_block(READ
, 1, bh
);
1761 wait_on_buffer(*bh
);
1762 if (!buffer_uptodate(*bh
)) {
1772 static int ocfs2_mount_volume(struct super_block
*sb
)
1775 int unlock_super
= 0;
1776 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
1780 if (ocfs2_is_hard_readonly(osb
))
1783 status
= ocfs2_dlm_init(osb
);
1789 status
= ocfs2_super_lock(osb
, 1);
1796 /* This will load up the node map and add ourselves to it. */
1797 status
= ocfs2_find_slot(osb
);
1803 /* load all node-local system inodes */
1804 status
= ocfs2_init_local_system_inodes(osb
);
1810 status
= ocfs2_check_volume(osb
);
1816 status
= ocfs2_truncate_log_init(osb
);
1822 ocfs2_super_unlock(osb
, 1);
1828 static void ocfs2_dismount_volume(struct super_block
*sb
, int mnt_err
)
1830 int tmp
, hangup_needed
= 0;
1831 struct ocfs2_super
*osb
= NULL
;
1834 mlog_entry("(0x%p)\n", sb
);
1840 debugfs_remove(osb
->osb_ctxt
);
1843 * Flush inode dropping work queue so that deletes are
1844 * performed while the filesystem is still working
1846 ocfs2_drop_all_dl_inodes(osb
);
1848 /* Orphan scan should be stopped as early as possible */
1849 ocfs2_orphan_scan_stop(osb
);
1851 ocfs2_disable_quotas(osb
);
1853 ocfs2_shutdown_local_alloc(osb
);
1855 ocfs2_truncate_log_shutdown(osb
);
1857 /* This will disable recovery and flush any recovery work. */
1858 ocfs2_recovery_exit(osb
);
1860 ocfs2_journal_shutdown(osb
);
1862 ocfs2_sync_blockdev(sb
);
1864 ocfs2_purge_refcount_trees(osb
);
1866 /* No cluster connection means we've failed during mount, so skip
1867 * all the steps which depended on that to complete. */
1869 tmp
= ocfs2_super_lock(osb
, 1);
1876 if (osb
->slot_num
!= OCFS2_INVALID_SLOT
)
1877 ocfs2_put_slot(osb
);
1880 ocfs2_super_unlock(osb
, 1);
1882 ocfs2_release_system_inodes(osb
);
1885 * If we're dismounting due to mount error, mount.ocfs2 will clean
1886 * up heartbeat. If we're a local mount, there is no heartbeat.
1887 * If we failed before we got a uuid_str yet, we can't stop
1888 * heartbeat. Otherwise, do it.
1890 if (!mnt_err
&& !ocfs2_mount_local(osb
) && osb
->uuid_str
)
1894 ocfs2_dlm_shutdown(osb
, hangup_needed
);
1896 ocfs2_blockcheck_stats_debugfs_remove(&osb
->osb_ecc_stats
);
1897 debugfs_remove(osb
->osb_debug_root
);
1900 ocfs2_cluster_hangup(osb
->uuid_str
, strlen(osb
->uuid_str
));
1902 atomic_set(&osb
->vol_state
, VOLUME_DISMOUNTED
);
1904 if (ocfs2_mount_local(osb
))
1905 snprintf(nodestr
, sizeof(nodestr
), "local");
1907 snprintf(nodestr
, sizeof(nodestr
), "%u", osb
->node_num
);
1909 printk(KERN_INFO
"ocfs2: Unmounting device (%s) on (node %s)\n",
1910 osb
->dev_str
, nodestr
);
1912 ocfs2_delete_osb(osb
);
1915 sb
->s_fs_info
= NULL
;
1918 static int ocfs2_setup_osb_uuid(struct ocfs2_super
*osb
, const unsigned char *uuid
,
1919 unsigned uuid_bytes
)
1924 BUG_ON(uuid_bytes
!= OCFS2_VOL_UUID_LEN
);
1926 osb
->uuid_str
= kzalloc(OCFS2_VOL_UUID_LEN
* 2 + 1, GFP_KERNEL
);
1927 if (osb
->uuid_str
== NULL
)
1930 for (i
= 0, ptr
= osb
->uuid_str
; i
< OCFS2_VOL_UUID_LEN
; i
++) {
1931 /* print with null */
1932 ret
= snprintf(ptr
, 3, "%02X", uuid
[i
]);
1933 if (ret
!= 2) /* drop super cleans up */
1935 /* then only advance past the last char */
1942 static int ocfs2_initialize_super(struct super_block
*sb
,
1943 struct buffer_head
*bh
,
1945 struct ocfs2_blockcheck_stats
*stats
)
1948 int i
, cbits
, bbits
;
1949 struct ocfs2_dinode
*di
= (struct ocfs2_dinode
*)bh
->b_data
;
1950 struct inode
*inode
= NULL
;
1951 struct ocfs2_journal
*journal
;
1952 __le32 uuid_net_key
;
1953 struct ocfs2_super
*osb
;
1957 osb
= kzalloc(sizeof(struct ocfs2_super
), GFP_KERNEL
);
1964 sb
->s_fs_info
= osb
;
1965 sb
->s_op
= &ocfs2_sops
;
1966 sb
->s_export_op
= &ocfs2_export_ops
;
1967 sb
->s_qcop
= &ocfs2_quotactl_ops
;
1968 sb
->dq_op
= &ocfs2_quota_operations
;
1969 sb
->s_xattr
= ocfs2_xattr_handlers
;
1970 sb
->s_time_gran
= 1;
1971 sb
->s_flags
|= MS_NOATIME
;
1972 /* this is needed to support O_LARGEFILE */
1973 cbits
= le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
);
1974 bbits
= le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
);
1975 sb
->s_maxbytes
= ocfs2_max_file_offset(bbits
, cbits
);
1977 osb
->osb_dx_mask
= (1 << (cbits
- bbits
)) - 1;
1979 for (i
= 0; i
< 3; i
++)
1980 osb
->osb_dx_seed
[i
] = le32_to_cpu(di
->id2
.i_super
.s_dx_seed
[i
]);
1981 osb
->osb_dx_seed
[3] = le32_to_cpu(di
->id2
.i_super
.s_uuid_hash
);
1984 /* Save off for ocfs2_rw_direct */
1985 osb
->s_sectsize_bits
= blksize_bits(sector_size
);
1986 BUG_ON(!osb
->s_sectsize_bits
);
1988 spin_lock_init(&osb
->dc_task_lock
);
1989 init_waitqueue_head(&osb
->dc_event
);
1990 osb
->dc_work_sequence
= 0;
1991 osb
->dc_wake_sequence
= 0;
1992 INIT_LIST_HEAD(&osb
->blocked_lock_list
);
1993 osb
->blocked_lock_count
= 0;
1994 spin_lock_init(&osb
->osb_lock
);
1995 spin_lock_init(&osb
->osb_xattr_lock
);
1996 ocfs2_init_inode_steal_slot(osb
);
1998 atomic_set(&osb
->alloc_stats
.moves
, 0);
1999 atomic_set(&osb
->alloc_stats
.local_data
, 0);
2000 atomic_set(&osb
->alloc_stats
.bitmap_data
, 0);
2001 atomic_set(&osb
->alloc_stats
.bg_allocs
, 0);
2002 atomic_set(&osb
->alloc_stats
.bg_extends
, 0);
2004 /* Copy the blockcheck stats from the superblock probe */
2005 osb
->osb_ecc_stats
= *stats
;
2007 ocfs2_init_node_maps(osb
);
2009 snprintf(osb
->dev_str
, sizeof(osb
->dev_str
), "%u,%u",
2010 MAJOR(osb
->sb
->s_dev
), MINOR(osb
->sb
->s_dev
));
2012 ocfs2_orphan_scan_init(osb
);
2014 status
= ocfs2_recovery_init(osb
);
2016 mlog(ML_ERROR
, "Unable to initialize recovery state\n");
2021 init_waitqueue_head(&osb
->checkpoint_event
);
2022 atomic_set(&osb
->needs_checkpoint
, 0);
2024 osb
->s_atime_quantum
= OCFS2_DEFAULT_ATIME_QUANTUM
;
2026 osb
->slot_num
= OCFS2_INVALID_SLOT
;
2028 osb
->s_xattr_inline_size
= le16_to_cpu(
2029 di
->id2
.i_super
.s_xattr_inline_size
);
2031 osb
->local_alloc_state
= OCFS2_LA_UNUSED
;
2032 osb
->local_alloc_bh
= NULL
;
2033 INIT_DELAYED_WORK(&osb
->la_enable_wq
, ocfs2_la_enable_worker
);
2035 init_waitqueue_head(&osb
->osb_mount_event
);
2037 osb
->vol_label
= kmalloc(OCFS2_MAX_VOL_LABEL_LEN
, GFP_KERNEL
);
2038 if (!osb
->vol_label
) {
2039 mlog(ML_ERROR
, "unable to alloc vol label\n");
2044 osb
->max_slots
= le16_to_cpu(di
->id2
.i_super
.s_max_slots
);
2045 if (osb
->max_slots
> OCFS2_MAX_SLOTS
|| osb
->max_slots
== 0) {
2046 mlog(ML_ERROR
, "Invalid number of node slots (%u)\n",
2051 mlog(0, "max_slots for this device: %u\n", osb
->max_slots
);
2053 osb
->slot_recovery_generations
=
2054 kcalloc(osb
->max_slots
, sizeof(*osb
->slot_recovery_generations
),
2056 if (!osb
->slot_recovery_generations
) {
2062 init_waitqueue_head(&osb
->osb_wipe_event
);
2063 osb
->osb_orphan_wipes
= kcalloc(osb
->max_slots
,
2064 sizeof(*osb
->osb_orphan_wipes
),
2066 if (!osb
->osb_orphan_wipes
) {
2072 osb
->osb_rf_lock_tree
= RB_ROOT
;
2074 osb
->s_feature_compat
=
2075 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_compat
);
2076 osb
->s_feature_ro_compat
=
2077 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_ro_compat
);
2078 osb
->s_feature_incompat
=
2079 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_incompat
);
2081 if ((i
= OCFS2_HAS_INCOMPAT_FEATURE(osb
->sb
, ~OCFS2_FEATURE_INCOMPAT_SUPP
))) {
2082 mlog(ML_ERROR
, "couldn't mount because of unsupported "
2083 "optional features (%x).\n", i
);
2087 if (!(osb
->sb
->s_flags
& MS_RDONLY
) &&
2088 (i
= OCFS2_HAS_RO_COMPAT_FEATURE(osb
->sb
, ~OCFS2_FEATURE_RO_COMPAT_SUPP
))) {
2089 mlog(ML_ERROR
, "couldn't mount RDWR because of "
2090 "unsupported optional features (%x).\n", i
);
2095 if (ocfs2_userspace_stack(osb
)) {
2096 memcpy(osb
->osb_cluster_stack
,
2097 OCFS2_RAW_SB(di
)->s_cluster_info
.ci_stack
,
2098 OCFS2_STACK_LABEL_LEN
);
2099 osb
->osb_cluster_stack
[OCFS2_STACK_LABEL_LEN
] = '\0';
2100 if (strlen(osb
->osb_cluster_stack
) != OCFS2_STACK_LABEL_LEN
) {
2102 "couldn't mount because of an invalid "
2103 "cluster stack label (%s) \n",
2104 osb
->osb_cluster_stack
);
2109 /* The empty string is identical with classic tools that
2110 * don't know about s_cluster_info. */
2111 osb
->osb_cluster_stack
[0] = '\0';
2114 get_random_bytes(&osb
->s_next_generation
, sizeof(u32
));
2117 * This should be done in ocfs2_journal_init(), but unknown
2118 * ordering issues will cause the filesystem to crash.
2119 * If anyone wants to figure out what part of the code
2120 * refers to osb->journal before ocfs2_journal_init() is run,
2123 /* initialize our journal structure */
2125 journal
= kzalloc(sizeof(struct ocfs2_journal
), GFP_KERNEL
);
2127 mlog(ML_ERROR
, "unable to alloc journal\n");
2131 osb
->journal
= journal
;
2132 journal
->j_osb
= osb
;
2134 atomic_set(&journal
->j_num_trans
, 0);
2135 init_rwsem(&journal
->j_trans_barrier
);
2136 init_waitqueue_head(&journal
->j_checkpointed
);
2137 spin_lock_init(&journal
->j_lock
);
2138 journal
->j_trans_id
= (unsigned long) 1;
2139 INIT_LIST_HEAD(&journal
->j_la_cleanups
);
2140 INIT_WORK(&journal
->j_recovery_work
, ocfs2_complete_recovery
);
2141 journal
->j_state
= OCFS2_JOURNAL_FREE
;
2143 INIT_WORK(&osb
->dentry_lock_work
, ocfs2_drop_dl_inodes
);
2144 osb
->dentry_lock_list
= NULL
;
2146 /* get some pseudo constants for clustersize bits */
2147 osb
->s_clustersize_bits
=
2148 le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
);
2149 osb
->s_clustersize
= 1 << osb
->s_clustersize_bits
;
2150 mlog(0, "clusterbits=%d\n", osb
->s_clustersize_bits
);
2152 if (osb
->s_clustersize
< OCFS2_MIN_CLUSTERSIZE
||
2153 osb
->s_clustersize
> OCFS2_MAX_CLUSTERSIZE
) {
2154 mlog(ML_ERROR
, "Volume has invalid cluster size (%d)\n",
2155 osb
->s_clustersize
);
2160 if (ocfs2_clusters_to_blocks(osb
->sb
, le32_to_cpu(di
->i_clusters
) - 1)
2162 mlog(ML_ERROR
, "Volume might try to write to blocks beyond "
2163 "what jbd can address in 32 bits.\n");
2168 if (ocfs2_setup_osb_uuid(osb
, di
->id2
.i_super
.s_uuid
,
2169 sizeof(di
->id2
.i_super
.s_uuid
))) {
2170 mlog(ML_ERROR
, "Out of memory trying to setup our uuid.\n");
2175 memcpy(&uuid_net_key
, di
->id2
.i_super
.s_uuid
, sizeof(uuid_net_key
));
2177 strncpy(osb
->vol_label
, di
->id2
.i_super
.s_label
, 63);
2178 osb
->vol_label
[63] = '\0';
2179 osb
->root_blkno
= le64_to_cpu(di
->id2
.i_super
.s_root_blkno
);
2180 osb
->system_dir_blkno
= le64_to_cpu(di
->id2
.i_super
.s_system_dir_blkno
);
2181 osb
->first_cluster_group_blkno
=
2182 le64_to_cpu(di
->id2
.i_super
.s_first_cluster_group
);
2183 osb
->fs_generation
= le32_to_cpu(di
->i_fs_generation
);
2184 osb
->uuid_hash
= le32_to_cpu(di
->id2
.i_super
.s_uuid_hash
);
2185 mlog(0, "vol_label: %s\n", osb
->vol_label
);
2186 mlog(0, "uuid: %s\n", osb
->uuid_str
);
2187 mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
2188 (unsigned long long)osb
->root_blkno
,
2189 (unsigned long long)osb
->system_dir_blkno
);
2191 osb
->osb_dlm_debug
= ocfs2_new_dlm_debug();
2192 if (!osb
->osb_dlm_debug
) {
2198 atomic_set(&osb
->vol_state
, VOLUME_INIT
);
2200 /* load root, system_dir, and all global system inodes */
2201 status
= ocfs2_init_global_system_inodes(osb
);
2210 inode
= ocfs2_get_system_file_inode(osb
, GLOBAL_BITMAP_SYSTEM_INODE
,
2211 OCFS2_INVALID_SLOT
);
2218 osb
->bitmap_blkno
= OCFS2_I(inode
)->ip_blkno
;
2221 osb
->bitmap_cpg
= ocfs2_group_bitmap_size(sb
) * 8;
2223 status
= ocfs2_init_slot_info(osb
);
2235 * will return: -EAGAIN if it is ok to keep searching for superblocks
2236 * -EINVAL if there is a bad superblock
2239 static int ocfs2_verify_volume(struct ocfs2_dinode
*di
,
2240 struct buffer_head
*bh
,
2242 struct ocfs2_blockcheck_stats
*stats
)
2244 int status
= -EAGAIN
;
2248 if (memcmp(di
->i_signature
, OCFS2_SUPER_BLOCK_SIGNATURE
,
2249 strlen(OCFS2_SUPER_BLOCK_SIGNATURE
)) == 0) {
2250 /* We have to do a raw check of the feature here */
2251 if (le32_to_cpu(di
->id2
.i_super
.s_feature_incompat
) &
2252 OCFS2_FEATURE_INCOMPAT_META_ECC
) {
2253 status
= ocfs2_block_check_validate(bh
->b_data
,
2261 if ((1 << le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
)) != blksz
) {
2262 mlog(ML_ERROR
, "found superblock with incorrect block "
2263 "size: found %u, should be %u\n",
2264 1 << le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
),
2266 } else if (le16_to_cpu(di
->id2
.i_super
.s_major_rev_level
) !=
2267 OCFS2_MAJOR_REV_LEVEL
||
2268 le16_to_cpu(di
->id2
.i_super
.s_minor_rev_level
) !=
2269 OCFS2_MINOR_REV_LEVEL
) {
2270 mlog(ML_ERROR
, "found superblock with bad version: "
2271 "found %u.%u, should be %u.%u\n",
2272 le16_to_cpu(di
->id2
.i_super
.s_major_rev_level
),
2273 le16_to_cpu(di
->id2
.i_super
.s_minor_rev_level
),
2274 OCFS2_MAJOR_REV_LEVEL
,
2275 OCFS2_MINOR_REV_LEVEL
);
2276 } else if (bh
->b_blocknr
!= le64_to_cpu(di
->i_blkno
)) {
2277 mlog(ML_ERROR
, "bad block number on superblock: "
2278 "found %llu, should be %llu\n",
2279 (unsigned long long)le64_to_cpu(di
->i_blkno
),
2280 (unsigned long long)bh
->b_blocknr
);
2281 } else if (le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
) < 12 ||
2282 le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
) > 20) {
2283 mlog(ML_ERROR
, "bad cluster size found: %u\n",
2284 1 << le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
));
2285 } else if (!le64_to_cpu(di
->id2
.i_super
.s_root_blkno
)) {
2286 mlog(ML_ERROR
, "bad root_blkno: 0\n");
2287 } else if (!le64_to_cpu(di
->id2
.i_super
.s_system_dir_blkno
)) {
2288 mlog(ML_ERROR
, "bad system_dir_blkno: 0\n");
2289 } else if (le16_to_cpu(di
->id2
.i_super
.s_max_slots
) > OCFS2_MAX_SLOTS
) {
2291 "Superblock slots found greater than file system "
2292 "maximum: found %u, max %u\n",
2293 le16_to_cpu(di
->id2
.i_super
.s_max_slots
),
2306 static int ocfs2_check_volume(struct ocfs2_super
*osb
)
2311 struct ocfs2_dinode
*local_alloc
= NULL
; /* only used if we
2317 /* Init our journal object. */
2318 status
= ocfs2_journal_init(osb
->journal
, &dirty
);
2320 mlog(ML_ERROR
, "Could not initialize journal!\n");
2324 /* If the journal was unmounted cleanly then we don't want to
2325 * recover anything. Otherwise, journal_load will do that
2326 * dirty work for us :) */
2328 status
= ocfs2_journal_wipe(osb
->journal
, 0);
2334 mlog(ML_NOTICE
, "File system was not unmounted cleanly, "
2335 "recovering volume.\n");
2338 local
= ocfs2_mount_local(osb
);
2340 /* will play back anything left in the journal. */
2341 status
= ocfs2_journal_load(osb
->journal
, local
, dirty
);
2343 mlog(ML_ERROR
, "ocfs2 journal load failed! %d\n", status
);
2348 /* recover my local alloc if we didn't unmount cleanly. */
2349 status
= ocfs2_begin_local_alloc_recovery(osb
,
2356 /* we complete the recovery process after we've marked
2357 * ourselves as mounted. */
2360 mlog(0, "Journal loaded.\n");
2362 status
= ocfs2_load_local_alloc(osb
);
2369 /* Recovery will be completed after we've mounted the
2370 * rest of the volume. */
2372 osb
->local_alloc_copy
= local_alloc
;
2376 /* go through each journal, trylock it and if you get the
2377 * lock, and it's marked as dirty, set the bit in the recover
2378 * map and launch a recovery thread for it. */
2379 status
= ocfs2_mark_dead_nodes(osb
);
2385 status
= ocfs2_compute_replay_slots(osb
);
2398 * The routine gets called from dismount or close whenever a dismount on
2399 * volume is requested and the osb open count becomes 1.
2400 * It will remove the osb from the global list and also free up all the
2401 * initialized resources and fileobject.
2403 static void ocfs2_delete_osb(struct ocfs2_super
*osb
)
2407 /* This function assumes that the caller has the main osb resource */
2409 ocfs2_free_slot_info(osb
);
2411 kfree(osb
->osb_orphan_wipes
);
2412 kfree(osb
->slot_recovery_generations
);
2414 * This belongs in journal shutdown, but because we have to
2415 * allocate osb->journal at the start of ocfs2_initalize_osb(),
2418 kfree(osb
->journal
);
2419 if (osb
->local_alloc_copy
)
2420 kfree(osb
->local_alloc_copy
);
2421 kfree(osb
->uuid_str
);
2422 ocfs2_put_dlm_debug(osb
->osb_dlm_debug
);
2423 memset(osb
, 0, sizeof(struct ocfs2_super
));
2428 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2429 * panic(). We do not support continue-on-error operation. */
2430 static void ocfs2_handle_error(struct super_block
*sb
)
2432 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
2434 if (osb
->s_mount_opt
& OCFS2_MOUNT_ERRORS_PANIC
)
2435 panic("OCFS2: (device %s): panic forced after error\n",
2438 ocfs2_set_osb_flag(osb
, OCFS2_OSB_ERROR_FS
);
2440 if (sb
->s_flags
& MS_RDONLY
&&
2441 (ocfs2_is_soft_readonly(osb
) ||
2442 ocfs2_is_hard_readonly(osb
)))
2445 printk(KERN_CRIT
"File system is now read-only due to the potential "
2446 "of on-disk corruption. Please run fsck.ocfs2 once the file "
2447 "system is unmounted.\n");
2448 sb
->s_flags
|= MS_RDONLY
;
2449 ocfs2_set_ro_flag(osb
, 0);
2452 static char error_buf
[1024];
2454 void __ocfs2_error(struct super_block
*sb
,
2455 const char *function
,
2456 const char *fmt
, ...)
2460 va_start(args
, fmt
);
2461 vsnprintf(error_buf
, sizeof(error_buf
), fmt
, args
);
2464 /* Not using mlog here because we want to show the actual
2465 * function the error came from. */
2466 printk(KERN_CRIT
"OCFS2: ERROR (device %s): %s: %s\n",
2467 sb
->s_id
, function
, error_buf
);
2469 ocfs2_handle_error(sb
);
2472 /* Handle critical errors. This is intentionally more drastic than
2473 * ocfs2_handle_error, so we only use for things like journal errors,
2475 void __ocfs2_abort(struct super_block
* sb
,
2476 const char *function
,
2477 const char *fmt
, ...)
2481 va_start(args
, fmt
);
2482 vsnprintf(error_buf
, sizeof(error_buf
), fmt
, args
);
2485 printk(KERN_CRIT
"OCFS2: abort (device %s): %s: %s\n",
2486 sb
->s_id
, function
, error_buf
);
2488 /* We don't have the cluster support yet to go straight to
2489 * hard readonly in here. Until then, we want to keep
2490 * ocfs2_abort() so that we can at least mark critical
2493 * TODO: This should abort the journal and alert other nodes
2494 * that our slot needs recovery. */
2496 /* Force a panic(). This stinks, but it's better than letting
2497 * things continue without having a proper hard readonly
2499 if (!ocfs2_mount_local(OCFS2_SB(sb
)))
2500 OCFS2_SB(sb
)->s_mount_opt
|= OCFS2_MOUNT_ERRORS_PANIC
;
2501 ocfs2_handle_error(sb
);
2504 module_init(ocfs2_init
);
2505 module_exit(ocfs2_exit
);