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>
45 #define MLOG_MASK_PREFIX ML_SUPER
46 #include <cluster/masklog.h>
50 /* this should be the only file to include a version 1 header */
51 #include "ocfs1_fs_compat.h"
54 #include "blockcheck.h"
57 #include "extent_map.h"
58 #include "heartbeat.h"
61 #include "localalloc.h"
70 #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
;
97 unsigned int resv_level
;
99 char cluster_stack
[OCFS2_STACK_LABEL_LEN
+ 1];
102 static int ocfs2_parse_options(struct super_block
*sb
, char *options
,
103 struct mount_options
*mopt
,
105 static int ocfs2_check_set_options(struct super_block
*sb
,
106 struct mount_options
*options
);
107 static int ocfs2_show_options(struct seq_file
*s
, struct vfsmount
*mnt
);
108 static void ocfs2_put_super(struct super_block
*sb
);
109 static int ocfs2_mount_volume(struct super_block
*sb
);
110 static int ocfs2_remount(struct super_block
*sb
, int *flags
, char *data
);
111 static void ocfs2_dismount_volume(struct super_block
*sb
, int mnt_err
);
112 static int ocfs2_initialize_mem_caches(void);
113 static void ocfs2_free_mem_caches(void);
114 static void ocfs2_delete_osb(struct ocfs2_super
*osb
);
116 static int ocfs2_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
118 static int ocfs2_sync_fs(struct super_block
*sb
, int wait
);
120 static int ocfs2_init_global_system_inodes(struct ocfs2_super
*osb
);
121 static int ocfs2_init_local_system_inodes(struct ocfs2_super
*osb
);
122 static void ocfs2_release_system_inodes(struct ocfs2_super
*osb
);
123 static int ocfs2_check_volume(struct ocfs2_super
*osb
);
124 static int ocfs2_verify_volume(struct ocfs2_dinode
*di
,
125 struct buffer_head
*bh
,
127 struct ocfs2_blockcheck_stats
*stats
);
128 static int ocfs2_initialize_super(struct super_block
*sb
,
129 struct buffer_head
*bh
,
131 struct ocfs2_blockcheck_stats
*stats
);
132 static int ocfs2_get_sector(struct super_block
*sb
,
133 struct buffer_head
**bh
,
136 static struct inode
*ocfs2_alloc_inode(struct super_block
*sb
);
137 static void ocfs2_destroy_inode(struct inode
*inode
);
138 static int ocfs2_susp_quotas(struct ocfs2_super
*osb
, int unsuspend
);
139 static int ocfs2_enable_quotas(struct ocfs2_super
*osb
);
140 static void ocfs2_disable_quotas(struct ocfs2_super
*osb
);
142 static const struct super_operations ocfs2_sops
= {
143 .statfs
= ocfs2_statfs
,
144 .alloc_inode
= ocfs2_alloc_inode
,
145 .destroy_inode
= ocfs2_destroy_inode
,
146 .drop_inode
= ocfs2_drop_inode
,
147 .evict_inode
= ocfs2_evict_inode
,
148 .sync_fs
= ocfs2_sync_fs
,
149 .put_super
= ocfs2_put_super
,
150 .remount_fs
= ocfs2_remount
,
151 .show_options
= ocfs2_show_options
,
152 .quota_read
= ocfs2_quota_read
,
153 .quota_write
= ocfs2_quota_write
,
180 Opt_coherency_buffered
,
187 static const match_table_t tokens
= {
188 {Opt_barrier
, "barrier=%u"},
189 {Opt_err_panic
, "errors=panic"},
190 {Opt_err_ro
, "errors=remount-ro"},
192 {Opt_nointr
, "nointr"},
193 {Opt_hb_none
, OCFS2_HB_NONE
},
194 {Opt_hb_local
, OCFS2_HB_LOCAL
},
195 {Opt_hb_global
, OCFS2_HB_GLOBAL
},
196 {Opt_data_ordered
, "data=ordered"},
197 {Opt_data_writeback
, "data=writeback"},
198 {Opt_atime_quantum
, "atime_quantum=%u"},
199 {Opt_slot
, "preferred_slot=%u"},
200 {Opt_commit
, "commit=%u"},
201 {Opt_localalloc
, "localalloc=%d"},
202 {Opt_localflocks
, "localflocks"},
203 {Opt_stack
, "cluster_stack=%s"},
204 {Opt_user_xattr
, "user_xattr"},
205 {Opt_nouser_xattr
, "nouser_xattr"},
206 {Opt_inode64
, "inode64"},
208 {Opt_noacl
, "noacl"},
209 {Opt_usrquota
, "usrquota"},
210 {Opt_grpquota
, "grpquota"},
211 {Opt_coherency_buffered
, "coherency=buffered"},
212 {Opt_coherency_full
, "coherency=full"},
213 {Opt_resv_level
, "resv_level=%u"},
214 {Opt_dir_resv_level
, "dir_resv_level=%u"},
218 #ifdef CONFIG_DEBUG_FS
219 static int ocfs2_osb_dump(struct ocfs2_super
*osb
, char *buf
, int len
)
221 struct ocfs2_cluster_connection
*cconn
= osb
->cconn
;
222 struct ocfs2_recovery_map
*rm
= osb
->recovery_map
;
223 struct ocfs2_orphan_scan
*os
= &osb
->osb_orphan_scan
;
226 out
+= snprintf(buf
+ out
, len
- out
,
227 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n",
228 "Device", osb
->dev_str
, osb
->uuid_str
,
229 osb
->fs_generation
, osb
->vol_label
);
231 out
+= snprintf(buf
+ out
, len
- out
,
232 "%10s => State: %d Flags: 0x%lX\n", "Volume",
233 atomic_read(&osb
->vol_state
), osb
->osb_flags
);
235 out
+= snprintf(buf
+ out
, len
- out
,
236 "%10s => Block: %lu Cluster: %d\n", "Sizes",
237 osb
->sb
->s_blocksize
, osb
->s_clustersize
);
239 out
+= snprintf(buf
+ out
, len
- out
,
240 "%10s => Compat: 0x%X Incompat: 0x%X "
242 "Features", osb
->s_feature_compat
,
243 osb
->s_feature_incompat
, osb
->s_feature_ro_compat
);
245 out
+= snprintf(buf
+ out
, len
- out
,
246 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount",
247 osb
->s_mount_opt
, osb
->s_atime_quantum
);
250 out
+= snprintf(buf
+ out
, len
- out
,
251 "%10s => Stack: %s Name: %*s "
252 "Version: %d.%d\n", "Cluster",
253 (*osb
->osb_cluster_stack
== '\0' ?
254 "o2cb" : osb
->osb_cluster_stack
),
255 cconn
->cc_namelen
, cconn
->cc_name
,
256 cconn
->cc_version
.pv_major
,
257 cconn
->cc_version
.pv_minor
);
260 spin_lock(&osb
->dc_task_lock
);
261 out
+= snprintf(buf
+ out
, len
- out
,
262 "%10s => Pid: %d Count: %lu WakeSeq: %lu "
263 "WorkSeq: %lu\n", "DownCnvt",
264 (osb
->dc_task
? task_pid_nr(osb
->dc_task
) : -1),
265 osb
->blocked_lock_count
, osb
->dc_wake_sequence
,
266 osb
->dc_work_sequence
);
267 spin_unlock(&osb
->dc_task_lock
);
269 spin_lock(&osb
->osb_lock
);
270 out
+= snprintf(buf
+ out
, len
- out
, "%10s => Pid: %d Nodes:",
272 (osb
->recovery_thread_task
?
273 task_pid_nr(osb
->recovery_thread_task
) : -1));
274 if (rm
->rm_used
== 0)
275 out
+= snprintf(buf
+ out
, len
- out
, " None\n");
277 for (i
= 0; i
< rm
->rm_used
; i
++)
278 out
+= snprintf(buf
+ out
, len
- out
, " %d",
280 out
+= snprintf(buf
+ out
, len
- out
, "\n");
282 spin_unlock(&osb
->osb_lock
);
284 out
+= snprintf(buf
+ out
, len
- out
,
285 "%10s => Pid: %d Interval: %lu Needs: %d\n", "Commit",
286 (osb
->commit_task
? task_pid_nr(osb
->commit_task
) : -1),
287 osb
->osb_commit_interval
,
288 atomic_read(&osb
->needs_checkpoint
));
290 out
+= snprintf(buf
+ out
, len
- out
,
291 "%10s => State: %d TxnId: %lu NumTxns: %d\n",
292 "Journal", osb
->journal
->j_state
,
293 osb
->journal
->j_trans_id
,
294 atomic_read(&osb
->journal
->j_num_trans
));
296 out
+= snprintf(buf
+ out
, len
- out
,
297 "%10s => GlobalAllocs: %d LocalAllocs: %d "
298 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
300 atomic_read(&osb
->alloc_stats
.bitmap_data
),
301 atomic_read(&osb
->alloc_stats
.local_data
),
302 atomic_read(&osb
->alloc_stats
.bg_allocs
),
303 atomic_read(&osb
->alloc_stats
.moves
),
304 atomic_read(&osb
->alloc_stats
.bg_extends
));
306 out
+= snprintf(buf
+ out
, len
- out
,
307 "%10s => State: %u Descriptor: %llu Size: %u bits "
308 "Default: %u bits\n",
309 "LocalAlloc", osb
->local_alloc_state
,
310 (unsigned long long)osb
->la_last_gd
,
311 osb
->local_alloc_bits
, osb
->local_alloc_default_bits
);
313 spin_lock(&osb
->osb_lock
);
314 out
+= snprintf(buf
+ out
, len
- out
,
315 "%10s => InodeSlot: %d StolenInodes: %d, "
316 "MetaSlot: %d StolenMeta: %d\n", "Steal",
317 osb
->s_inode_steal_slot
,
318 atomic_read(&osb
->s_num_inodes_stolen
),
319 osb
->s_meta_steal_slot
,
320 atomic_read(&osb
->s_num_meta_stolen
));
321 spin_unlock(&osb
->osb_lock
);
323 out
+= snprintf(buf
+ out
, len
- out
, "OrphanScan => ");
324 out
+= snprintf(buf
+ out
, len
- out
, "Local: %u Global: %u ",
325 os
->os_count
, os
->os_seqno
);
326 out
+= snprintf(buf
+ out
, len
- out
, " Last Scan: ");
327 if (atomic_read(&os
->os_state
) == ORPHAN_SCAN_INACTIVE
)
328 out
+= snprintf(buf
+ out
, len
- out
, "Disabled\n");
330 out
+= snprintf(buf
+ out
, len
- out
, "%lu seconds ago\n",
331 (get_seconds() - os
->os_scantime
.tv_sec
));
333 out
+= snprintf(buf
+ out
, len
- out
, "%10s => %3s %10s\n",
334 "Slots", "Num", "RecoGen");
335 for (i
= 0; i
< osb
->max_slots
; ++i
) {
336 out
+= snprintf(buf
+ out
, len
- out
,
337 "%10s %c %3d %10d\n",
339 (i
== osb
->slot_num
? '*' : ' '),
340 i
, osb
->slot_recovery_generations
[i
]);
346 static int ocfs2_osb_debug_open(struct inode
*inode
, struct file
*file
)
348 struct ocfs2_super
*osb
= inode
->i_private
;
351 buf
= kmalloc(PAGE_SIZE
, GFP_KERNEL
);
355 i_size_write(inode
, ocfs2_osb_dump(osb
, buf
, PAGE_SIZE
));
357 file
->private_data
= buf
;
364 static int ocfs2_debug_release(struct inode
*inode
, struct file
*file
)
366 kfree(file
->private_data
);
370 static ssize_t
ocfs2_debug_read(struct file
*file
, char __user
*buf
,
371 size_t nbytes
, loff_t
*ppos
)
373 return simple_read_from_buffer(buf
, nbytes
, ppos
, file
->private_data
,
374 i_size_read(file
->f_mapping
->host
));
377 static int ocfs2_osb_debug_open(struct inode
*inode
, struct file
*file
)
381 static int ocfs2_debug_release(struct inode
*inode
, struct file
*file
)
385 static ssize_t
ocfs2_debug_read(struct file
*file
, char __user
*buf
,
386 size_t nbytes
, loff_t
*ppos
)
390 #endif /* CONFIG_DEBUG_FS */
392 static const struct file_operations ocfs2_osb_debug_fops
= {
393 .open
= ocfs2_osb_debug_open
,
394 .release
= ocfs2_debug_release
,
395 .read
= ocfs2_debug_read
,
396 .llseek
= generic_file_llseek
,
399 static int ocfs2_sync_fs(struct super_block
*sb
, int wait
)
403 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
405 if (ocfs2_is_hard_readonly(osb
))
409 status
= ocfs2_flush_truncate_log(osb
);
413 ocfs2_schedule_truncate_log_flush(osb
, 0);
416 if (jbd2_journal_start_commit(OCFS2_SB(sb
)->journal
->j_journal
,
419 jbd2_log_wait_commit(OCFS2_SB(sb
)->journal
->j_journal
,
425 static int ocfs2_need_system_inode(struct ocfs2_super
*osb
, int ino
)
427 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb
->sb
, OCFS2_FEATURE_RO_COMPAT_USRQUOTA
)
428 && (ino
== USER_QUOTA_SYSTEM_INODE
429 || ino
== LOCAL_USER_QUOTA_SYSTEM_INODE
))
431 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb
->sb
, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
)
432 && (ino
== GROUP_QUOTA_SYSTEM_INODE
433 || ino
== LOCAL_GROUP_QUOTA_SYSTEM_INODE
))
438 static int ocfs2_init_global_system_inodes(struct ocfs2_super
*osb
)
440 struct inode
*new = NULL
;
446 new = ocfs2_iget(osb
, osb
->root_blkno
, OCFS2_FI_FLAG_SYSFILE
, 0);
448 status
= PTR_ERR(new);
452 osb
->root_inode
= new;
454 new = ocfs2_iget(osb
, osb
->system_dir_blkno
, OCFS2_FI_FLAG_SYSFILE
, 0);
456 status
= PTR_ERR(new);
460 osb
->sys_root_inode
= new;
462 for (i
= OCFS2_FIRST_ONLINE_SYSTEM_INODE
;
463 i
<= OCFS2_LAST_GLOBAL_SYSTEM_INODE
; i
++) {
464 if (!ocfs2_need_system_inode(osb
, i
))
466 new = ocfs2_get_system_file_inode(osb
, i
, osb
->slot_num
);
468 ocfs2_release_system_inodes(osb
);
471 /* FIXME: Should ERROR_RO_FS */
472 mlog(ML_ERROR
, "Unable to load system inode %d, "
473 "possibly corrupt fs?", i
);
476 // the array now has one ref, so drop this one
485 static int ocfs2_init_local_system_inodes(struct ocfs2_super
*osb
)
487 struct inode
*new = NULL
;
493 for (i
= OCFS2_LAST_GLOBAL_SYSTEM_INODE
+ 1;
494 i
< NUM_SYSTEM_INODES
;
496 if (!ocfs2_need_system_inode(osb
, i
))
498 new = ocfs2_get_system_file_inode(osb
, i
, osb
->slot_num
);
500 ocfs2_release_system_inodes(osb
);
502 mlog(ML_ERROR
, "status=%d, sysfile=%d, slot=%d\n",
503 status
, i
, osb
->slot_num
);
506 /* the array now has one ref, so drop this one */
515 static void ocfs2_release_system_inodes(struct ocfs2_super
*osb
)
522 for (i
= 0; i
< NUM_GLOBAL_SYSTEM_INODES
; i
++) {
523 inode
= osb
->global_system_inodes
[i
];
526 osb
->global_system_inodes
[i
] = NULL
;
530 inode
= osb
->sys_root_inode
;
533 osb
->sys_root_inode
= NULL
;
536 inode
= osb
->root_inode
;
539 osb
->root_inode
= NULL
;
542 if (!osb
->local_system_inodes
)
545 for (i
= 0; i
< NUM_LOCAL_SYSTEM_INODES
* osb
->max_slots
; i
++) {
546 if (osb
->local_system_inodes
[i
]) {
547 iput(osb
->local_system_inodes
[i
]);
548 osb
->local_system_inodes
[i
] = NULL
;
552 kfree(osb
->local_system_inodes
);
553 osb
->local_system_inodes
= NULL
;
559 /* We're allocating fs objects, use GFP_NOFS */
560 static struct inode
*ocfs2_alloc_inode(struct super_block
*sb
)
562 struct ocfs2_inode_info
*oi
;
564 oi
= kmem_cache_alloc(ocfs2_inode_cachep
, GFP_NOFS
);
568 jbd2_journal_init_jbd_inode(&oi
->ip_jinode
, &oi
->vfs_inode
);
569 return &oi
->vfs_inode
;
572 static void ocfs2_i_callback(struct rcu_head
*head
)
574 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
575 INIT_LIST_HEAD(&inode
->i_dentry
);
576 kmem_cache_free(ocfs2_inode_cachep
, OCFS2_I(inode
));
579 static void ocfs2_destroy_inode(struct inode
*inode
)
581 call_rcu(&inode
->i_rcu
, ocfs2_i_callback
);
584 static unsigned long long ocfs2_max_file_offset(unsigned int bbits
,
587 unsigned int bytes
= 1 << cbits
;
588 unsigned int trim
= bytes
;
589 unsigned int bitshift
= 32;
592 * i_size and all block offsets in ocfs2 are always 64 bits
593 * wide. i_clusters is 32 bits, in cluster-sized units. So on
594 * 64 bit platforms, cluster size will be the limiting factor.
597 #if BITS_PER_LONG == 32
598 # if defined(CONFIG_LBDAF)
599 BUILD_BUG_ON(sizeof(sector_t
) != 8);
601 * We might be limited by page cache size.
603 if (bytes
> PAGE_CACHE_SIZE
) {
604 bytes
= PAGE_CACHE_SIZE
;
607 * Shift by 31 here so that we don't get larger than
614 * We are limited by the size of sector_t. Use block size, as
615 * that's what we expose to the VFS.
624 * Trim by a whole cluster when we can actually approach the
625 * on-disk limits. Otherwise we can overflow i_clusters when
626 * an extent start is at the max offset.
628 return (((unsigned long long)bytes
) << bitshift
) - trim
;
631 static int ocfs2_remount(struct super_block
*sb
, int *flags
, char *data
)
633 int incompat_features
;
635 struct mount_options parsed_options
;
636 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
639 if (!ocfs2_parse_options(sb
, data
, &parsed_options
, 1) ||
640 !ocfs2_check_set_options(sb
, &parsed_options
)) {
645 tmp
= OCFS2_MOUNT_HB_LOCAL
| OCFS2_MOUNT_HB_GLOBAL
|
647 if ((osb
->s_mount_opt
& tmp
) != (parsed_options
.mount_opt
& tmp
)) {
649 mlog(ML_ERROR
, "Cannot change heartbeat mode on remount\n");
653 if ((osb
->s_mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
) !=
654 (parsed_options
.mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
)) {
656 mlog(ML_ERROR
, "Cannot change data mode on remount\n");
660 /* Probably don't want this on remount; it might
661 * mess with other nodes */
662 if (!(osb
->s_mount_opt
& OCFS2_MOUNT_INODE64
) &&
663 (parsed_options
.mount_opt
& OCFS2_MOUNT_INODE64
)) {
665 mlog(ML_ERROR
, "Cannot enable inode64 on remount\n");
669 /* We're going to/from readonly mode. */
670 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
)) {
671 /* Disable quota accounting before remounting RO */
672 if (*flags
& MS_RDONLY
) {
673 ret
= ocfs2_susp_quotas(osb
, 0);
677 /* Lock here so the check of HARD_RO and the potential
678 * setting of SOFT_RO is atomic. */
679 spin_lock(&osb
->osb_lock
);
680 if (osb
->osb_flags
& OCFS2_OSB_HARD_RO
) {
681 mlog(ML_ERROR
, "Remount on readonly device is forbidden.\n");
686 if (*flags
& MS_RDONLY
) {
687 mlog(0, "Going to ro mode.\n");
688 sb
->s_flags
|= MS_RDONLY
;
689 osb
->osb_flags
|= OCFS2_OSB_SOFT_RO
;
691 mlog(0, "Making ro filesystem writeable.\n");
693 if (osb
->osb_flags
& OCFS2_OSB_ERROR_FS
) {
694 mlog(ML_ERROR
, "Cannot remount RDWR "
695 "filesystem due to previous errors.\n");
699 incompat_features
= OCFS2_HAS_RO_COMPAT_FEATURE(sb
, ~OCFS2_FEATURE_RO_COMPAT_SUPP
);
700 if (incompat_features
) {
701 mlog(ML_ERROR
, "Cannot remount RDWR because "
702 "of unsupported optional features "
703 "(%x).\n", incompat_features
);
707 sb
->s_flags
&= ~MS_RDONLY
;
708 osb
->osb_flags
&= ~OCFS2_OSB_SOFT_RO
;
711 spin_unlock(&osb
->osb_lock
);
712 /* Enable quota accounting after remounting RW */
713 if (!ret
&& !(*flags
& MS_RDONLY
)) {
714 if (sb_any_quota_suspended(sb
))
715 ret
= ocfs2_susp_quotas(osb
, 1);
717 ret
= ocfs2_enable_quotas(osb
);
719 /* Return back changes... */
720 spin_lock(&osb
->osb_lock
);
721 sb
->s_flags
|= MS_RDONLY
;
722 osb
->osb_flags
|= OCFS2_OSB_SOFT_RO
;
723 spin_unlock(&osb
->osb_lock
);
730 /* Only save off the new mount options in case of a successful
732 osb
->s_mount_opt
= parsed_options
.mount_opt
;
733 osb
->s_atime_quantum
= parsed_options
.atime_quantum
;
734 osb
->preferred_slot
= parsed_options
.slot
;
735 if (parsed_options
.commit_interval
)
736 osb
->osb_commit_interval
= parsed_options
.commit_interval
;
738 if (!ocfs2_is_hard_readonly(osb
))
739 ocfs2_set_journal_params(osb
);
741 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
742 ((osb
->s_mount_opt
& OCFS2_MOUNT_POSIX_ACL
) ?
749 static int ocfs2_sb_probe(struct super_block
*sb
,
750 struct buffer_head
**bh
,
752 struct ocfs2_blockcheck_stats
*stats
)
755 struct ocfs1_vol_disk_hdr
*hdr
;
756 struct ocfs2_dinode
*di
;
762 *sector_size
= bdev_logical_block_size(sb
->s_bdev
);
763 if (*sector_size
> OCFS2_MAX_BLOCKSIZE
) {
764 mlog(ML_ERROR
, "Hardware sector size too large: %d (max=%d)\n",
765 *sector_size
, OCFS2_MAX_BLOCKSIZE
);
770 /* Can this really happen? */
771 if (*sector_size
< OCFS2_MIN_BLOCKSIZE
)
772 *sector_size
= OCFS2_MIN_BLOCKSIZE
;
774 /* check block zero for old format */
775 status
= ocfs2_get_sector(sb
, bh
, 0, *sector_size
);
780 hdr
= (struct ocfs1_vol_disk_hdr
*) (*bh
)->b_data
;
781 if (hdr
->major_version
== OCFS1_MAJOR_VERSION
) {
782 mlog(ML_ERROR
, "incompatible version: %u.%u\n",
783 hdr
->major_version
, hdr
->minor_version
);
786 if (memcmp(hdr
->signature
, OCFS1_VOLUME_SIGNATURE
,
787 strlen(OCFS1_VOLUME_SIGNATURE
)) == 0) {
788 mlog(ML_ERROR
, "incompatible volume signature: %8s\n",
795 mlog(ML_ERROR
, "This is an ocfs v1 filesystem which must be "
796 "upgraded before mounting with ocfs v2\n");
801 * Now check at magic offset for 512, 1024, 2048, 4096
802 * blocksizes. 4096 is the maximum blocksize because it is
803 * the minimum clustersize.
806 for (blksize
= *sector_size
;
807 blksize
<= OCFS2_MAX_BLOCKSIZE
;
809 tmpstat
= ocfs2_get_sector(sb
, bh
,
810 OCFS2_SUPER_BLOCK_BLKNO
,
817 di
= (struct ocfs2_dinode
*) (*bh
)->b_data
;
818 memset(stats
, 0, sizeof(struct ocfs2_blockcheck_stats
));
819 spin_lock_init(&stats
->b_lock
);
820 tmpstat
= ocfs2_verify_volume(di
, *bh
, blksize
, stats
);
825 if (tmpstat
!= -EAGAIN
) {
835 static int ocfs2_verify_heartbeat(struct ocfs2_super
*osb
)
837 u32 hb_enabled
= OCFS2_MOUNT_HB_LOCAL
| OCFS2_MOUNT_HB_GLOBAL
;
839 if (osb
->s_mount_opt
& hb_enabled
) {
840 if (ocfs2_mount_local(osb
)) {
841 mlog(ML_ERROR
, "Cannot heartbeat on a locally "
842 "mounted device.\n");
845 if (ocfs2_userspace_stack(osb
)) {
846 mlog(ML_ERROR
, "Userspace stack expected, but "
847 "o2cb heartbeat arguments passed to mount\n");
850 if (((osb
->s_mount_opt
& OCFS2_MOUNT_HB_GLOBAL
) &&
851 !ocfs2_cluster_o2cb_global_heartbeat(osb
)) ||
852 ((osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) &&
853 ocfs2_cluster_o2cb_global_heartbeat(osb
))) {
854 mlog(ML_ERROR
, "Mismatching o2cb heartbeat modes\n");
859 if (!(osb
->s_mount_opt
& hb_enabled
)) {
860 if (!ocfs2_mount_local(osb
) && !ocfs2_is_hard_readonly(osb
) &&
861 !ocfs2_userspace_stack(osb
)) {
862 mlog(ML_ERROR
, "Heartbeat has to be started to mount "
863 "a read-write clustered device.\n");
872 * If we're using a userspace stack, mount should have passed
873 * a name that matches the disk. If not, mount should not
874 * have passed a stack.
876 static int ocfs2_verify_userspace_stack(struct ocfs2_super
*osb
,
877 struct mount_options
*mopt
)
879 if (!ocfs2_userspace_stack(osb
) && mopt
->cluster_stack
[0]) {
881 "cluster stack passed to mount, but this filesystem "
882 "does not support it\n");
886 if (ocfs2_userspace_stack(osb
) &&
887 strncmp(osb
->osb_cluster_stack
, mopt
->cluster_stack
,
888 OCFS2_STACK_LABEL_LEN
)) {
890 "cluster stack passed to mount (\"%s\") does not "
891 "match the filesystem (\"%s\")\n",
893 osb
->osb_cluster_stack
);
900 static int ocfs2_susp_quotas(struct ocfs2_super
*osb
, int unsuspend
)
903 struct super_block
*sb
= osb
->sb
;
904 unsigned int feature
[MAXQUOTAS
] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA
,
905 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
};
908 for (type
= 0; type
< MAXQUOTAS
; type
++) {
909 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb
, feature
[type
]))
912 status
= dquot_resume(sb
, type
);
914 struct ocfs2_mem_dqinfo
*oinfo
;
916 /* Cancel periodic syncing before suspending */
917 oinfo
= sb_dqinfo(sb
, type
)->dqi_priv
;
918 cancel_delayed_work_sync(&oinfo
->dqi_sync_work
);
919 status
= dquot_suspend(sb
, type
);
925 mlog(ML_ERROR
, "Failed to suspend/unsuspend quotas on "
926 "remount (error = %d).\n", status
);
930 static int ocfs2_enable_quotas(struct ocfs2_super
*osb
)
932 struct inode
*inode
[MAXQUOTAS
] = { NULL
, NULL
};
933 struct super_block
*sb
= osb
->sb
;
934 unsigned int feature
[MAXQUOTAS
] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA
,
935 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
};
936 unsigned int ino
[MAXQUOTAS
] = { LOCAL_USER_QUOTA_SYSTEM_INODE
,
937 LOCAL_GROUP_QUOTA_SYSTEM_INODE
};
941 sb_dqopt(sb
)->flags
|= DQUOT_QUOTA_SYS_FILE
| DQUOT_NEGATIVE_USAGE
;
942 for (type
= 0; type
< MAXQUOTAS
; type
++) {
943 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb
, feature
[type
]))
945 inode
[type
] = ocfs2_get_system_file_inode(osb
, ino
[type
],
951 status
= dquot_enable(inode
[type
], type
, QFMT_OCFS2
,
952 DQUOT_USAGE_ENABLED
);
957 for (type
= 0; type
< MAXQUOTAS
; type
++)
961 ocfs2_disable_quotas(osb
);
962 for (type
= 0; type
< MAXQUOTAS
; type
++)
968 static void ocfs2_disable_quotas(struct ocfs2_super
*osb
)
972 struct super_block
*sb
= osb
->sb
;
973 struct ocfs2_mem_dqinfo
*oinfo
;
975 /* We mostly ignore errors in this function because there's not much
976 * we can do when we see them */
977 for (type
= 0; type
< MAXQUOTAS
; type
++) {
978 if (!sb_has_quota_loaded(sb
, type
))
980 /* Cancel periodic syncing before we grab dqonoff_mutex */
981 oinfo
= sb_dqinfo(sb
, type
)->dqi_priv
;
982 cancel_delayed_work_sync(&oinfo
->dqi_sync_work
);
983 inode
= igrab(sb
->s_dquot
.files
[type
]);
984 /* Turn off quotas. This will remove all dquot structures from
985 * memory and so they will be automatically synced to global
987 dquot_disable(sb
, type
, DQUOT_USAGE_ENABLED
|
988 DQUOT_LIMITS_ENABLED
);
995 /* Handle quota on quotactl */
996 static int ocfs2_quota_on(struct super_block
*sb
, int type
, int format_id
)
998 unsigned int feature
[MAXQUOTAS
] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA
,
999 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
};
1001 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb
, feature
[type
]))
1004 return dquot_enable(sb_dqopt(sb
)->files
[type
], type
,
1005 format_id
, DQUOT_LIMITS_ENABLED
);
1008 /* Handle quota off quotactl */
1009 static int ocfs2_quota_off(struct super_block
*sb
, int type
)
1011 return dquot_disable(sb
, type
, DQUOT_LIMITS_ENABLED
);
1014 static const struct quotactl_ops ocfs2_quotactl_ops
= {
1015 .quota_on_meta
= ocfs2_quota_on
,
1016 .quota_off
= ocfs2_quota_off
,
1017 .quota_sync
= dquot_quota_sync
,
1018 .get_info
= dquot_get_dqinfo
,
1019 .set_info
= dquot_set_dqinfo
,
1020 .get_dqblk
= dquot_get_dqblk
,
1021 .set_dqblk
= dquot_set_dqblk
,
1024 static int ocfs2_fill_super(struct super_block
*sb
, void *data
, int silent
)
1026 struct dentry
*root
;
1027 int status
, sector_size
;
1028 struct mount_options parsed_options
;
1029 struct inode
*inode
= NULL
;
1030 struct ocfs2_super
*osb
= NULL
;
1031 struct buffer_head
*bh
= NULL
;
1033 struct ocfs2_blockcheck_stats stats
;
1035 mlog_entry("%p, %p, %i", sb
, data
, silent
);
1037 if (!ocfs2_parse_options(sb
, data
, &parsed_options
, 0)) {
1039 goto read_super_error
;
1042 /* probe for superblock */
1043 status
= ocfs2_sb_probe(sb
, &bh
, §or_size
, &stats
);
1045 mlog(ML_ERROR
, "superblock probe failed!\n");
1046 goto read_super_error
;
1049 status
= ocfs2_initialize_super(sb
, bh
, sector_size
, &stats
);
1053 goto read_super_error
;
1058 if (!ocfs2_check_set_options(sb
, &parsed_options
)) {
1060 goto read_super_error
;
1062 osb
->s_mount_opt
= parsed_options
.mount_opt
;
1063 osb
->s_atime_quantum
= parsed_options
.atime_quantum
;
1064 osb
->preferred_slot
= parsed_options
.slot
;
1065 osb
->osb_commit_interval
= parsed_options
.commit_interval
;
1067 ocfs2_la_set_sizes(osb
, parsed_options
.localalloc_opt
);
1068 osb
->osb_resv_level
= parsed_options
.resv_level
;
1069 osb
->osb_dir_resv_level
= parsed_options
.resv_level
;
1070 if (parsed_options
.dir_resv_level
== -1)
1071 osb
->osb_dir_resv_level
= parsed_options
.resv_level
;
1073 osb
->osb_dir_resv_level
= parsed_options
.dir_resv_level
;
1075 status
= ocfs2_verify_userspace_stack(osb
, &parsed_options
);
1077 goto read_super_error
;
1079 sb
->s_magic
= OCFS2_SUPER_MAGIC
;
1081 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
1082 ((osb
->s_mount_opt
& OCFS2_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
1084 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1086 if (bdev_read_only(sb
->s_bdev
)) {
1087 if (!(sb
->s_flags
& MS_RDONLY
)) {
1089 mlog(ML_ERROR
, "Readonly device detected but readonly "
1090 "mount was not specified.\n");
1091 goto read_super_error
;
1094 /* You should not be able to start a local heartbeat
1095 * on a readonly device. */
1096 if (osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) {
1098 mlog(ML_ERROR
, "Local heartbeat specified on readonly "
1100 goto read_super_error
;
1103 status
= ocfs2_check_journals_nolocks(osb
);
1105 if (status
== -EROFS
)
1106 mlog(ML_ERROR
, "Recovery required on readonly "
1107 "file system, but write access is "
1111 goto read_super_error
;
1114 ocfs2_set_ro_flag(osb
, 1);
1116 printk(KERN_NOTICE
"Readonly device detected. No cluster "
1117 "services will be utilized for this mount. Recovery "
1118 "will be skipped.\n");
1121 if (!ocfs2_is_hard_readonly(osb
)) {
1122 if (sb
->s_flags
& MS_RDONLY
)
1123 ocfs2_set_ro_flag(osb
, 0);
1126 status
= ocfs2_verify_heartbeat(osb
);
1129 goto read_super_error
;
1132 osb
->osb_debug_root
= debugfs_create_dir(osb
->uuid_str
,
1133 ocfs2_debugfs_root
);
1134 if (!osb
->osb_debug_root
) {
1136 mlog(ML_ERROR
, "Unable to create per-mount debugfs root.\n");
1137 goto read_super_error
;
1140 osb
->osb_ctxt
= debugfs_create_file("fs_state", S_IFREG
|S_IRUSR
,
1141 osb
->osb_debug_root
,
1143 &ocfs2_osb_debug_fops
);
1144 if (!osb
->osb_ctxt
) {
1147 goto read_super_error
;
1150 if (ocfs2_meta_ecc(osb
)) {
1151 status
= ocfs2_blockcheck_stats_debugfs_install(
1152 &osb
->osb_ecc_stats
,
1153 osb
->osb_debug_root
);
1156 "Unable to create blockcheck statistics "
1158 goto read_super_error
;
1162 status
= ocfs2_mount_volume(sb
);
1163 if (osb
->root_inode
)
1164 inode
= igrab(osb
->root_inode
);
1167 goto read_super_error
;
1172 goto read_super_error
;
1175 root
= d_alloc_root(inode
);
1179 goto read_super_error
;
1184 ocfs2_complete_mount_recovery(osb
);
1186 if (ocfs2_mount_local(osb
))
1187 snprintf(nodestr
, sizeof(nodestr
), "local");
1189 snprintf(nodestr
, sizeof(nodestr
), "%u", osb
->node_num
);
1191 printk(KERN_INFO
"ocfs2: Mounting device (%s) on (node %s, slot %d) "
1192 "with %s data mode.\n",
1193 osb
->dev_str
, nodestr
, osb
->slot_num
,
1194 osb
->s_mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
? "writeback" :
1197 atomic_set(&osb
->vol_state
, VOLUME_MOUNTED
);
1198 wake_up(&osb
->osb_mount_event
);
1200 /* Now we can initialize quotas because we can afford to wait
1201 * for cluster locks recovery now. That also means that truncation
1202 * log recovery can happen but that waits for proper quota setup */
1203 if (!(sb
->s_flags
& MS_RDONLY
)) {
1204 status
= ocfs2_enable_quotas(osb
);
1206 /* We have to err-out specially here because
1207 * s_root is already set */
1209 atomic_set(&osb
->vol_state
, VOLUME_DISABLED
);
1210 wake_up(&osb
->osb_mount_event
);
1216 ocfs2_complete_quota_recovery(osb
);
1218 /* Now we wake up again for processes waiting for quotas */
1219 atomic_set(&osb
->vol_state
, VOLUME_MOUNTED_QUOTAS
);
1220 wake_up(&osb
->osb_mount_event
);
1222 /* Start this when the mount is almost sure of being successful */
1223 ocfs2_orphan_scan_start(osb
);
1235 atomic_set(&osb
->vol_state
, VOLUME_DISABLED
);
1236 wake_up(&osb
->osb_mount_event
);
1237 ocfs2_dismount_volume(sb
, 1);
1244 static struct dentry
*ocfs2_mount(struct file_system_type
*fs_type
,
1246 const char *dev_name
,
1249 return mount_bdev(fs_type
, flags
, dev_name
, data
, ocfs2_fill_super
);
1252 static void ocfs2_kill_sb(struct super_block
*sb
)
1254 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
1257 if (!osb
|| atomic_read(&osb
->vol_state
) == VOLUME_DISABLED
)
1260 /* Prevent further queueing of inode drop events */
1261 spin_lock(&dentry_list_lock
);
1262 ocfs2_set_osb_flag(osb
, OCFS2_OSB_DROP_DENTRY_LOCK_IMMED
);
1263 spin_unlock(&dentry_list_lock
);
1264 /* Wait for work to finish and/or remove it */
1265 cancel_work_sync(&osb
->dentry_lock_work
);
1267 kill_block_super(sb
);
1270 static struct file_system_type ocfs2_fs_type
= {
1271 .owner
= THIS_MODULE
,
1273 .mount
= ocfs2_mount
,
1274 .kill_sb
= ocfs2_kill_sb
,
1276 .fs_flags
= FS_REQUIRES_DEV
|FS_RENAME_DOES_D_MOVE
,
1280 static int ocfs2_check_set_options(struct super_block
*sb
,
1281 struct mount_options
*options
)
1283 if (options
->mount_opt
& OCFS2_MOUNT_USRQUOTA
&&
1284 !OCFS2_HAS_RO_COMPAT_FEATURE(sb
,
1285 OCFS2_FEATURE_RO_COMPAT_USRQUOTA
)) {
1286 mlog(ML_ERROR
, "User quotas were requested, but this "
1287 "filesystem does not have the feature enabled.\n");
1290 if (options
->mount_opt
& OCFS2_MOUNT_GRPQUOTA
&&
1291 !OCFS2_HAS_RO_COMPAT_FEATURE(sb
,
1292 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
)) {
1293 mlog(ML_ERROR
, "Group quotas were requested, but this "
1294 "filesystem does not have the feature enabled.\n");
1297 if (options
->mount_opt
& OCFS2_MOUNT_POSIX_ACL
&&
1298 !OCFS2_HAS_INCOMPAT_FEATURE(sb
, OCFS2_FEATURE_INCOMPAT_XATTR
)) {
1299 mlog(ML_ERROR
, "ACL support requested but extended attributes "
1300 "feature is not enabled\n");
1303 /* No ACL setting specified? Use XATTR feature... */
1304 if (!(options
->mount_opt
& (OCFS2_MOUNT_POSIX_ACL
|
1305 OCFS2_MOUNT_NO_POSIX_ACL
))) {
1306 if (OCFS2_HAS_INCOMPAT_FEATURE(sb
, OCFS2_FEATURE_INCOMPAT_XATTR
))
1307 options
->mount_opt
|= OCFS2_MOUNT_POSIX_ACL
;
1309 options
->mount_opt
|= OCFS2_MOUNT_NO_POSIX_ACL
;
1314 static int ocfs2_parse_options(struct super_block
*sb
,
1316 struct mount_options
*mopt
,
1319 int status
, user_stack
= 0;
1323 mlog_entry("remount: %d, options: \"%s\"\n", is_remount
,
1324 options
? options
: "(none)");
1326 mopt
->commit_interval
= 0;
1327 mopt
->mount_opt
= OCFS2_MOUNT_NOINTR
;
1328 mopt
->atime_quantum
= OCFS2_DEFAULT_ATIME_QUANTUM
;
1329 mopt
->slot
= OCFS2_INVALID_SLOT
;
1330 mopt
->localalloc_opt
= -1;
1331 mopt
->cluster_stack
[0] = '\0';
1332 mopt
->resv_level
= OCFS2_DEFAULT_RESV_LEVEL
;
1333 mopt
->dir_resv_level
= -1;
1340 while ((p
= strsep(&options
, ",")) != NULL
) {
1342 substring_t args
[MAX_OPT_ARGS
];
1347 token
= match_token(p
, tokens
, args
);
1350 mopt
->mount_opt
|= OCFS2_MOUNT_HB_LOCAL
;
1353 mopt
->mount_opt
|= OCFS2_MOUNT_HB_NONE
;
1356 mopt
->mount_opt
|= OCFS2_MOUNT_HB_GLOBAL
;
1359 if (match_int(&args
[0], &option
)) {
1364 mopt
->mount_opt
|= OCFS2_MOUNT_BARRIER
;
1366 mopt
->mount_opt
&= ~OCFS2_MOUNT_BARRIER
;
1369 mopt
->mount_opt
&= ~OCFS2_MOUNT_NOINTR
;
1372 mopt
->mount_opt
|= OCFS2_MOUNT_NOINTR
;
1375 mopt
->mount_opt
|= OCFS2_MOUNT_ERRORS_PANIC
;
1378 mopt
->mount_opt
&= ~OCFS2_MOUNT_ERRORS_PANIC
;
1380 case Opt_data_ordered
:
1381 mopt
->mount_opt
&= ~OCFS2_MOUNT_DATA_WRITEBACK
;
1383 case Opt_data_writeback
:
1384 mopt
->mount_opt
|= OCFS2_MOUNT_DATA_WRITEBACK
;
1386 case Opt_user_xattr
:
1387 mopt
->mount_opt
&= ~OCFS2_MOUNT_NOUSERXATTR
;
1389 case Opt_nouser_xattr
:
1390 mopt
->mount_opt
|= OCFS2_MOUNT_NOUSERXATTR
;
1392 case Opt_atime_quantum
:
1393 if (match_int(&args
[0], &option
)) {
1398 mopt
->atime_quantum
= option
;
1402 if (match_int(&args
[0], &option
)) {
1407 mopt
->slot
= (s16
)option
;
1411 if (match_int(&args
[0], &option
)) {
1418 option
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1419 mopt
->commit_interval
= HZ
* option
;
1421 case Opt_localalloc
:
1423 if (match_int(&args
[0], &option
)) {
1428 mopt
->localalloc_opt
= option
;
1430 case Opt_localflocks
:
1432 * Changing this during remount could race
1433 * flock() requests, or "unbalance" existing
1434 * ones (e.g., a lock is taken in one mode but
1435 * dropped in the other). If users care enough
1436 * to flip locking modes during remount, we
1437 * could add a "local" flag to individual
1438 * flock structures for proper tracking of
1442 mopt
->mount_opt
|= OCFS2_MOUNT_LOCALFLOCKS
;
1445 /* Check both that the option we were passed
1446 * is of the right length and that it is a proper
1447 * string of the right length.
1449 if (((args
[0].to
- args
[0].from
) !=
1450 OCFS2_STACK_LABEL_LEN
) ||
1451 (strnlen(args
[0].from
,
1452 OCFS2_STACK_LABEL_LEN
) !=
1453 OCFS2_STACK_LABEL_LEN
)) {
1455 "Invalid cluster_stack option\n");
1459 memcpy(mopt
->cluster_stack
, args
[0].from
,
1460 OCFS2_STACK_LABEL_LEN
);
1461 mopt
->cluster_stack
[OCFS2_STACK_LABEL_LEN
] = '\0';
1463 * Open code the memcmp here as we don't have
1465 * ocfs2_userspace_stack().
1467 if (memcmp(mopt
->cluster_stack
,
1468 OCFS2_CLASSIC_CLUSTER_STACK
,
1469 OCFS2_STACK_LABEL_LEN
))
1473 mopt
->mount_opt
|= OCFS2_MOUNT_INODE64
;
1476 mopt
->mount_opt
|= OCFS2_MOUNT_USRQUOTA
;
1479 mopt
->mount_opt
|= OCFS2_MOUNT_GRPQUOTA
;
1481 case Opt_coherency_buffered
:
1482 mopt
->mount_opt
|= OCFS2_MOUNT_COHERENCY_BUFFERED
;
1484 case Opt_coherency_full
:
1485 mopt
->mount_opt
&= ~OCFS2_MOUNT_COHERENCY_BUFFERED
;
1488 mopt
->mount_opt
|= OCFS2_MOUNT_POSIX_ACL
;
1489 mopt
->mount_opt
&= ~OCFS2_MOUNT_NO_POSIX_ACL
;
1492 mopt
->mount_opt
|= OCFS2_MOUNT_NO_POSIX_ACL
;
1493 mopt
->mount_opt
&= ~OCFS2_MOUNT_POSIX_ACL
;
1495 case Opt_resv_level
:
1498 if (match_int(&args
[0], &option
)) {
1502 if (option
>= OCFS2_MIN_RESV_LEVEL
&&
1503 option
< OCFS2_MAX_RESV_LEVEL
)
1504 mopt
->resv_level
= option
;
1506 case Opt_dir_resv_level
:
1509 if (match_int(&args
[0], &option
)) {
1513 if (option
>= OCFS2_MIN_RESV_LEVEL
&&
1514 option
< OCFS2_MAX_RESV_LEVEL
)
1515 mopt
->dir_resv_level
= option
;
1519 "Unrecognized mount option \"%s\" "
1520 "or missing value\n", p
);
1526 if (user_stack
== 0) {
1527 /* Ensure only one heartbeat mode */
1528 tmp
= mopt
->mount_opt
& (OCFS2_MOUNT_HB_LOCAL
|
1529 OCFS2_MOUNT_HB_GLOBAL
|
1530 OCFS2_MOUNT_HB_NONE
);
1531 if (hweight32(tmp
) != 1) {
1532 mlog(ML_ERROR
, "Invalid heartbeat mount options\n");
1545 static int ocfs2_show_options(struct seq_file
*s
, struct vfsmount
*mnt
)
1547 struct ocfs2_super
*osb
= OCFS2_SB(mnt
->mnt_sb
);
1548 unsigned long opts
= osb
->s_mount_opt
;
1549 unsigned int local_alloc_megs
;
1551 if (opts
& (OCFS2_MOUNT_HB_LOCAL
| OCFS2_MOUNT_HB_GLOBAL
)) {
1552 seq_printf(s
, ",_netdev");
1553 if (opts
& OCFS2_MOUNT_HB_LOCAL
)
1554 seq_printf(s
, ",%s", OCFS2_HB_LOCAL
);
1556 seq_printf(s
, ",%s", OCFS2_HB_GLOBAL
);
1558 seq_printf(s
, ",%s", OCFS2_HB_NONE
);
1560 if (opts
& OCFS2_MOUNT_NOINTR
)
1561 seq_printf(s
, ",nointr");
1563 if (opts
& OCFS2_MOUNT_DATA_WRITEBACK
)
1564 seq_printf(s
, ",data=writeback");
1566 seq_printf(s
, ",data=ordered");
1568 if (opts
& OCFS2_MOUNT_BARRIER
)
1569 seq_printf(s
, ",barrier=1");
1571 if (opts
& OCFS2_MOUNT_ERRORS_PANIC
)
1572 seq_printf(s
, ",errors=panic");
1574 seq_printf(s
, ",errors=remount-ro");
1576 if (osb
->preferred_slot
!= OCFS2_INVALID_SLOT
)
1577 seq_printf(s
, ",preferred_slot=%d", osb
->preferred_slot
);
1579 if (osb
->s_atime_quantum
!= OCFS2_DEFAULT_ATIME_QUANTUM
)
1580 seq_printf(s
, ",atime_quantum=%u", osb
->s_atime_quantum
);
1582 if (osb
->osb_commit_interval
)
1583 seq_printf(s
, ",commit=%u",
1584 (unsigned) (osb
->osb_commit_interval
/ HZ
));
1586 local_alloc_megs
= osb
->local_alloc_bits
>> (20 - osb
->s_clustersize_bits
);
1587 if (local_alloc_megs
!= ocfs2_la_default_mb(osb
))
1588 seq_printf(s
, ",localalloc=%d", local_alloc_megs
);
1590 if (opts
& OCFS2_MOUNT_LOCALFLOCKS
)
1591 seq_printf(s
, ",localflocks,");
1593 if (osb
->osb_cluster_stack
[0])
1594 seq_printf(s
, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN
,
1595 osb
->osb_cluster_stack
);
1596 if (opts
& OCFS2_MOUNT_USRQUOTA
)
1597 seq_printf(s
, ",usrquota");
1598 if (opts
& OCFS2_MOUNT_GRPQUOTA
)
1599 seq_printf(s
, ",grpquota");
1601 if (opts
& OCFS2_MOUNT_COHERENCY_BUFFERED
)
1602 seq_printf(s
, ",coherency=buffered");
1604 seq_printf(s
, ",coherency=full");
1606 if (opts
& OCFS2_MOUNT_NOUSERXATTR
)
1607 seq_printf(s
, ",nouser_xattr");
1609 seq_printf(s
, ",user_xattr");
1611 if (opts
& OCFS2_MOUNT_INODE64
)
1612 seq_printf(s
, ",inode64");
1614 if (opts
& OCFS2_MOUNT_POSIX_ACL
)
1615 seq_printf(s
, ",acl");
1617 seq_printf(s
, ",noacl");
1619 if (osb
->osb_resv_level
!= OCFS2_DEFAULT_RESV_LEVEL
)
1620 seq_printf(s
, ",resv_level=%d", osb
->osb_resv_level
);
1622 if (osb
->osb_dir_resv_level
!= osb
->osb_resv_level
)
1623 seq_printf(s
, ",dir_resv_level=%d", osb
->osb_resv_level
);
1628 static int __init
ocfs2_init(void)
1634 ocfs2_print_version();
1636 status
= init_ocfs2_uptodate_cache();
1642 status
= ocfs2_initialize_mem_caches();
1648 ocfs2_wq
= create_singlethread_workqueue("ocfs2_wq");
1654 ocfs2_debugfs_root
= debugfs_create_dir("ocfs2", NULL
);
1655 if (!ocfs2_debugfs_root
) {
1657 mlog(ML_ERROR
, "Unable to create ocfs2 debugfs root.\n");
1660 status
= ocfs2_quota_setup();
1664 ocfs2_set_locking_protocol();
1666 status
= register_quota_format(&ocfs2_quota_format
);
1669 ocfs2_quota_shutdown();
1670 ocfs2_free_mem_caches();
1671 exit_ocfs2_uptodate_cache();
1677 return register_filesystem(&ocfs2_fs_type
);
1682 static void __exit
ocfs2_exit(void)
1686 ocfs2_quota_shutdown();
1689 flush_workqueue(ocfs2_wq
);
1690 destroy_workqueue(ocfs2_wq
);
1693 unregister_quota_format(&ocfs2_quota_format
);
1695 debugfs_remove(ocfs2_debugfs_root
);
1697 ocfs2_free_mem_caches();
1699 unregister_filesystem(&ocfs2_fs_type
);
1701 exit_ocfs2_uptodate_cache();
1706 static void ocfs2_put_super(struct super_block
*sb
)
1708 mlog_entry("(0x%p)\n", sb
);
1710 ocfs2_sync_blockdev(sb
);
1711 ocfs2_dismount_volume(sb
, 0);
1716 static int ocfs2_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
1718 struct ocfs2_super
*osb
;
1719 u32 numbits
, freebits
;
1721 struct ocfs2_dinode
*bm_lock
;
1722 struct buffer_head
*bh
= NULL
;
1723 struct inode
*inode
= NULL
;
1725 mlog_entry("(%p, %p)\n", dentry
->d_sb
, buf
);
1727 osb
= OCFS2_SB(dentry
->d_sb
);
1729 inode
= ocfs2_get_system_file_inode(osb
,
1730 GLOBAL_BITMAP_SYSTEM_INODE
,
1731 OCFS2_INVALID_SLOT
);
1733 mlog(ML_ERROR
, "failed to get bitmap inode\n");
1738 status
= ocfs2_inode_lock(inode
, &bh
, 0);
1744 bm_lock
= (struct ocfs2_dinode
*) bh
->b_data
;
1746 numbits
= le32_to_cpu(bm_lock
->id1
.bitmap1
.i_total
);
1747 freebits
= numbits
- le32_to_cpu(bm_lock
->id1
.bitmap1
.i_used
);
1749 buf
->f_type
= OCFS2_SUPER_MAGIC
;
1750 buf
->f_bsize
= dentry
->d_sb
->s_blocksize
;
1751 buf
->f_namelen
= OCFS2_MAX_FILENAME_LEN
;
1752 buf
->f_blocks
= ((sector_t
) numbits
) *
1753 (osb
->s_clustersize
>> osb
->sb
->s_blocksize_bits
);
1754 buf
->f_bfree
= ((sector_t
) freebits
) *
1755 (osb
->s_clustersize
>> osb
->sb
->s_blocksize_bits
);
1756 buf
->f_bavail
= buf
->f_bfree
;
1757 buf
->f_files
= numbits
;
1758 buf
->f_ffree
= freebits
;
1759 buf
->f_fsid
.val
[0] = crc32_le(0, osb
->uuid_str
, OCFS2_VOL_UUID_LEN
)
1761 buf
->f_fsid
.val
[1] = crc32_le(0, osb
->uuid_str
+ OCFS2_VOL_UUID_LEN
,
1762 OCFS2_VOL_UUID_LEN
) & 0xFFFFFFFFUL
;
1766 ocfs2_inode_unlock(inode
, 0);
1777 static void ocfs2_inode_init_once(void *data
)
1779 struct ocfs2_inode_info
*oi
= data
;
1782 oi
->ip_open_count
= 0;
1783 spin_lock_init(&oi
->ip_lock
);
1784 ocfs2_extent_map_init(&oi
->vfs_inode
);
1785 INIT_LIST_HEAD(&oi
->ip_io_markers
);
1786 oi
->ip_dir_start_lookup
= 0;
1788 init_rwsem(&oi
->ip_alloc_sem
);
1789 init_rwsem(&oi
->ip_xattr_sem
);
1790 mutex_init(&oi
->ip_io_mutex
);
1792 oi
->ip_blkno
= 0ULL;
1793 oi
->ip_clusters
= 0;
1795 ocfs2_resv_init_once(&oi
->ip_la_data_resv
);
1797 ocfs2_lock_res_init_once(&oi
->ip_rw_lockres
);
1798 ocfs2_lock_res_init_once(&oi
->ip_inode_lockres
);
1799 ocfs2_lock_res_init_once(&oi
->ip_open_lockres
);
1801 ocfs2_metadata_cache_init(INODE_CACHE(&oi
->vfs_inode
),
1802 &ocfs2_inode_caching_ops
);
1804 inode_init_once(&oi
->vfs_inode
);
1807 static int ocfs2_initialize_mem_caches(void)
1809 ocfs2_inode_cachep
= kmem_cache_create("ocfs2_inode_cache",
1810 sizeof(struct ocfs2_inode_info
),
1812 (SLAB_HWCACHE_ALIGN
|SLAB_RECLAIM_ACCOUNT
|
1814 ocfs2_inode_init_once
);
1815 ocfs2_dquot_cachep
= kmem_cache_create("ocfs2_dquot_cache",
1816 sizeof(struct ocfs2_dquot
),
1818 (SLAB_HWCACHE_ALIGN
|SLAB_RECLAIM_ACCOUNT
|
1821 ocfs2_qf_chunk_cachep
= kmem_cache_create("ocfs2_qf_chunk_cache",
1822 sizeof(struct ocfs2_quota_chunk
),
1824 (SLAB_RECLAIM_ACCOUNT
|SLAB_MEM_SPREAD
),
1826 if (!ocfs2_inode_cachep
|| !ocfs2_dquot_cachep
||
1827 !ocfs2_qf_chunk_cachep
) {
1828 if (ocfs2_inode_cachep
)
1829 kmem_cache_destroy(ocfs2_inode_cachep
);
1830 if (ocfs2_dquot_cachep
)
1831 kmem_cache_destroy(ocfs2_dquot_cachep
);
1832 if (ocfs2_qf_chunk_cachep
)
1833 kmem_cache_destroy(ocfs2_qf_chunk_cachep
);
1840 static void ocfs2_free_mem_caches(void)
1842 if (ocfs2_inode_cachep
)
1843 kmem_cache_destroy(ocfs2_inode_cachep
);
1844 ocfs2_inode_cachep
= NULL
;
1846 if (ocfs2_dquot_cachep
)
1847 kmem_cache_destroy(ocfs2_dquot_cachep
);
1848 ocfs2_dquot_cachep
= NULL
;
1850 if (ocfs2_qf_chunk_cachep
)
1851 kmem_cache_destroy(ocfs2_qf_chunk_cachep
);
1852 ocfs2_qf_chunk_cachep
= NULL
;
1855 static int ocfs2_get_sector(struct super_block
*sb
,
1856 struct buffer_head
**bh
,
1860 if (!sb_set_blocksize(sb
, sect_size
)) {
1861 mlog(ML_ERROR
, "unable to set blocksize\n");
1865 *bh
= sb_getblk(sb
, block
);
1871 if (!buffer_dirty(*bh
))
1872 clear_buffer_uptodate(*bh
);
1874 ll_rw_block(READ
, 1, bh
);
1875 wait_on_buffer(*bh
);
1876 if (!buffer_uptodate(*bh
)) {
1886 static int ocfs2_mount_volume(struct super_block
*sb
)
1889 int unlock_super
= 0;
1890 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
1894 if (ocfs2_is_hard_readonly(osb
))
1897 status
= ocfs2_dlm_init(osb
);
1903 status
= ocfs2_super_lock(osb
, 1);
1910 /* This will load up the node map and add ourselves to it. */
1911 status
= ocfs2_find_slot(osb
);
1917 /* load all node-local system inodes */
1918 status
= ocfs2_init_local_system_inodes(osb
);
1924 status
= ocfs2_check_volume(osb
);
1930 status
= ocfs2_truncate_log_init(osb
);
1936 ocfs2_super_unlock(osb
, 1);
1942 static void ocfs2_dismount_volume(struct super_block
*sb
, int mnt_err
)
1944 int tmp
, hangup_needed
= 0;
1945 struct ocfs2_super
*osb
= NULL
;
1948 mlog_entry("(0x%p)\n", sb
);
1954 debugfs_remove(osb
->osb_ctxt
);
1957 * Flush inode dropping work queue so that deletes are
1958 * performed while the filesystem is still working
1960 ocfs2_drop_all_dl_inodes(osb
);
1962 /* Orphan scan should be stopped as early as possible */
1963 ocfs2_orphan_scan_stop(osb
);
1965 ocfs2_disable_quotas(osb
);
1967 ocfs2_shutdown_local_alloc(osb
);
1969 ocfs2_truncate_log_shutdown(osb
);
1971 /* This will disable recovery and flush any recovery work. */
1972 ocfs2_recovery_exit(osb
);
1974 ocfs2_journal_shutdown(osb
);
1976 ocfs2_sync_blockdev(sb
);
1978 ocfs2_purge_refcount_trees(osb
);
1980 /* No cluster connection means we've failed during mount, so skip
1981 * all the steps which depended on that to complete. */
1983 tmp
= ocfs2_super_lock(osb
, 1);
1990 if (osb
->slot_num
!= OCFS2_INVALID_SLOT
)
1991 ocfs2_put_slot(osb
);
1994 ocfs2_super_unlock(osb
, 1);
1996 ocfs2_release_system_inodes(osb
);
1999 * If we're dismounting due to mount error, mount.ocfs2 will clean
2000 * up heartbeat. If we're a local mount, there is no heartbeat.
2001 * If we failed before we got a uuid_str yet, we can't stop
2002 * heartbeat. Otherwise, do it.
2004 if (!mnt_err
&& !ocfs2_mount_local(osb
) && osb
->uuid_str
)
2008 ocfs2_dlm_shutdown(osb
, hangup_needed
);
2010 ocfs2_blockcheck_stats_debugfs_remove(&osb
->osb_ecc_stats
);
2011 debugfs_remove(osb
->osb_debug_root
);
2014 ocfs2_cluster_hangup(osb
->uuid_str
, strlen(osb
->uuid_str
));
2016 atomic_set(&osb
->vol_state
, VOLUME_DISMOUNTED
);
2018 if (ocfs2_mount_local(osb
))
2019 snprintf(nodestr
, sizeof(nodestr
), "local");
2021 snprintf(nodestr
, sizeof(nodestr
), "%u", osb
->node_num
);
2023 printk(KERN_INFO
"ocfs2: Unmounting device (%s) on (node %s)\n",
2024 osb
->dev_str
, nodestr
);
2026 ocfs2_delete_osb(osb
);
2029 sb
->s_fs_info
= NULL
;
2032 static int ocfs2_setup_osb_uuid(struct ocfs2_super
*osb
, const unsigned char *uuid
,
2033 unsigned uuid_bytes
)
2038 BUG_ON(uuid_bytes
!= OCFS2_VOL_UUID_LEN
);
2040 osb
->uuid_str
= kzalloc(OCFS2_VOL_UUID_LEN
* 2 + 1, GFP_KERNEL
);
2041 if (osb
->uuid_str
== NULL
)
2044 for (i
= 0, ptr
= osb
->uuid_str
; i
< OCFS2_VOL_UUID_LEN
; i
++) {
2045 /* print with null */
2046 ret
= snprintf(ptr
, 3, "%02X", uuid
[i
]);
2047 if (ret
!= 2) /* drop super cleans up */
2049 /* then only advance past the last char */
2056 /* Make sure entire volume is addressable by our journal. Requires
2057 osb_clusters_at_boot to be valid and for the journal to have been
2058 initialized by ocfs2_journal_init(). */
2059 static int ocfs2_journal_addressable(struct ocfs2_super
*osb
)
2063 ocfs2_clusters_to_blocks(osb
->sb
,
2064 osb
->osb_clusters_at_boot
) - 1;
2066 /* 32-bit block number is always OK. */
2067 if (max_block
<= (u32
)~0ULL)
2070 /* Volume is "huge", so see if our journal is new enough to
2072 if (!(OCFS2_HAS_COMPAT_FEATURE(osb
->sb
,
2073 OCFS2_FEATURE_COMPAT_JBD2_SB
) &&
2074 jbd2_journal_check_used_features(osb
->journal
->j_journal
, 0, 0,
2075 JBD2_FEATURE_INCOMPAT_64BIT
))) {
2076 mlog(ML_ERROR
, "The journal cannot address the entire volume. "
2077 "Enable the 'block64' journal option with tunefs.ocfs2");
2086 static int ocfs2_initialize_super(struct super_block
*sb
,
2087 struct buffer_head
*bh
,
2089 struct ocfs2_blockcheck_stats
*stats
)
2092 int i
, cbits
, bbits
;
2093 struct ocfs2_dinode
*di
= (struct ocfs2_dinode
*)bh
->b_data
;
2094 struct inode
*inode
= NULL
;
2095 struct ocfs2_journal
*journal
;
2096 __le32 uuid_net_key
;
2097 struct ocfs2_super
*osb
;
2102 osb
= kzalloc(sizeof(struct ocfs2_super
), GFP_KERNEL
);
2109 sb
->s_fs_info
= osb
;
2110 sb
->s_op
= &ocfs2_sops
;
2111 sb
->s_d_op
= &ocfs2_dentry_ops
;
2112 sb
->s_export_op
= &ocfs2_export_ops
;
2113 sb
->s_qcop
= &ocfs2_quotactl_ops
;
2114 sb
->dq_op
= &ocfs2_quota_operations
;
2115 sb
->s_xattr
= ocfs2_xattr_handlers
;
2116 sb
->s_time_gran
= 1;
2117 sb
->s_flags
|= MS_NOATIME
;
2118 /* this is needed to support O_LARGEFILE */
2119 cbits
= le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
);
2120 bbits
= le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
);
2121 sb
->s_maxbytes
= ocfs2_max_file_offset(bbits
, cbits
);
2123 osb
->osb_dx_mask
= (1 << (cbits
- bbits
)) - 1;
2125 for (i
= 0; i
< 3; i
++)
2126 osb
->osb_dx_seed
[i
] = le32_to_cpu(di
->id2
.i_super
.s_dx_seed
[i
]);
2127 osb
->osb_dx_seed
[3] = le32_to_cpu(di
->id2
.i_super
.s_uuid_hash
);
2130 /* Save off for ocfs2_rw_direct */
2131 osb
->s_sectsize_bits
= blksize_bits(sector_size
);
2132 BUG_ON(!osb
->s_sectsize_bits
);
2134 spin_lock_init(&osb
->dc_task_lock
);
2135 init_waitqueue_head(&osb
->dc_event
);
2136 osb
->dc_work_sequence
= 0;
2137 osb
->dc_wake_sequence
= 0;
2138 INIT_LIST_HEAD(&osb
->blocked_lock_list
);
2139 osb
->blocked_lock_count
= 0;
2140 spin_lock_init(&osb
->osb_lock
);
2141 spin_lock_init(&osb
->osb_xattr_lock
);
2142 ocfs2_init_steal_slots(osb
);
2144 atomic_set(&osb
->alloc_stats
.moves
, 0);
2145 atomic_set(&osb
->alloc_stats
.local_data
, 0);
2146 atomic_set(&osb
->alloc_stats
.bitmap_data
, 0);
2147 atomic_set(&osb
->alloc_stats
.bg_allocs
, 0);
2148 atomic_set(&osb
->alloc_stats
.bg_extends
, 0);
2150 /* Copy the blockcheck stats from the superblock probe */
2151 osb
->osb_ecc_stats
= *stats
;
2153 ocfs2_init_node_maps(osb
);
2155 snprintf(osb
->dev_str
, sizeof(osb
->dev_str
), "%u,%u",
2156 MAJOR(osb
->sb
->s_dev
), MINOR(osb
->sb
->s_dev
));
2158 osb
->max_slots
= le16_to_cpu(di
->id2
.i_super
.s_max_slots
);
2159 if (osb
->max_slots
> OCFS2_MAX_SLOTS
|| osb
->max_slots
== 0) {
2160 mlog(ML_ERROR
, "Invalid number of node slots (%u)\n",
2165 mlog(0, "max_slots for this device: %u\n", osb
->max_slots
);
2167 ocfs2_orphan_scan_init(osb
);
2169 status
= ocfs2_recovery_init(osb
);
2171 mlog(ML_ERROR
, "Unable to initialize recovery state\n");
2176 init_waitqueue_head(&osb
->checkpoint_event
);
2177 atomic_set(&osb
->needs_checkpoint
, 0);
2179 osb
->s_atime_quantum
= OCFS2_DEFAULT_ATIME_QUANTUM
;
2181 osb
->slot_num
= OCFS2_INVALID_SLOT
;
2183 osb
->s_xattr_inline_size
= le16_to_cpu(
2184 di
->id2
.i_super
.s_xattr_inline_size
);
2186 osb
->local_alloc_state
= OCFS2_LA_UNUSED
;
2187 osb
->local_alloc_bh
= NULL
;
2188 INIT_DELAYED_WORK(&osb
->la_enable_wq
, ocfs2_la_enable_worker
);
2190 init_waitqueue_head(&osb
->osb_mount_event
);
2192 status
= ocfs2_resmap_init(osb
, &osb
->osb_la_resmap
);
2198 osb
->vol_label
= kmalloc(OCFS2_MAX_VOL_LABEL_LEN
, GFP_KERNEL
);
2199 if (!osb
->vol_label
) {
2200 mlog(ML_ERROR
, "unable to alloc vol label\n");
2205 osb
->slot_recovery_generations
=
2206 kcalloc(osb
->max_slots
, sizeof(*osb
->slot_recovery_generations
),
2208 if (!osb
->slot_recovery_generations
) {
2214 init_waitqueue_head(&osb
->osb_wipe_event
);
2215 osb
->osb_orphan_wipes
= kcalloc(osb
->max_slots
,
2216 sizeof(*osb
->osb_orphan_wipes
),
2218 if (!osb
->osb_orphan_wipes
) {
2224 osb
->osb_rf_lock_tree
= RB_ROOT
;
2226 osb
->s_feature_compat
=
2227 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_compat
);
2228 osb
->s_feature_ro_compat
=
2229 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_ro_compat
);
2230 osb
->s_feature_incompat
=
2231 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_incompat
);
2233 if ((i
= OCFS2_HAS_INCOMPAT_FEATURE(osb
->sb
, ~OCFS2_FEATURE_INCOMPAT_SUPP
))) {
2234 mlog(ML_ERROR
, "couldn't mount because of unsupported "
2235 "optional features (%x).\n", i
);
2239 if (!(osb
->sb
->s_flags
& MS_RDONLY
) &&
2240 (i
= OCFS2_HAS_RO_COMPAT_FEATURE(osb
->sb
, ~OCFS2_FEATURE_RO_COMPAT_SUPP
))) {
2241 mlog(ML_ERROR
, "couldn't mount RDWR because of "
2242 "unsupported optional features (%x).\n", i
);
2247 if (ocfs2_clusterinfo_valid(osb
)) {
2248 osb
->osb_stackflags
=
2249 OCFS2_RAW_SB(di
)->s_cluster_info
.ci_stackflags
;
2250 memcpy(osb
->osb_cluster_stack
,
2251 OCFS2_RAW_SB(di
)->s_cluster_info
.ci_stack
,
2252 OCFS2_STACK_LABEL_LEN
);
2253 osb
->osb_cluster_stack
[OCFS2_STACK_LABEL_LEN
] = '\0';
2254 if (strlen(osb
->osb_cluster_stack
) != OCFS2_STACK_LABEL_LEN
) {
2256 "couldn't mount because of an invalid "
2257 "cluster stack label (%s) \n",
2258 osb
->osb_cluster_stack
);
2263 /* The empty string is identical with classic tools that
2264 * don't know about s_cluster_info. */
2265 osb
->osb_cluster_stack
[0] = '\0';
2268 get_random_bytes(&osb
->s_next_generation
, sizeof(u32
));
2271 * This should be done in ocfs2_journal_init(), but unknown
2272 * ordering issues will cause the filesystem to crash.
2273 * If anyone wants to figure out what part of the code
2274 * refers to osb->journal before ocfs2_journal_init() is run,
2277 /* initialize our journal structure */
2279 journal
= kzalloc(sizeof(struct ocfs2_journal
), GFP_KERNEL
);
2281 mlog(ML_ERROR
, "unable to alloc journal\n");
2285 osb
->journal
= journal
;
2286 journal
->j_osb
= osb
;
2288 atomic_set(&journal
->j_num_trans
, 0);
2289 init_rwsem(&journal
->j_trans_barrier
);
2290 init_waitqueue_head(&journal
->j_checkpointed
);
2291 spin_lock_init(&journal
->j_lock
);
2292 journal
->j_trans_id
= (unsigned long) 1;
2293 INIT_LIST_HEAD(&journal
->j_la_cleanups
);
2294 INIT_WORK(&journal
->j_recovery_work
, ocfs2_complete_recovery
);
2295 journal
->j_state
= OCFS2_JOURNAL_FREE
;
2297 INIT_WORK(&osb
->dentry_lock_work
, ocfs2_drop_dl_inodes
);
2298 osb
->dentry_lock_list
= NULL
;
2300 /* get some pseudo constants for clustersize bits */
2301 osb
->s_clustersize_bits
=
2302 le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
);
2303 osb
->s_clustersize
= 1 << osb
->s_clustersize_bits
;
2304 mlog(0, "clusterbits=%d\n", osb
->s_clustersize_bits
);
2306 if (osb
->s_clustersize
< OCFS2_MIN_CLUSTERSIZE
||
2307 osb
->s_clustersize
> OCFS2_MAX_CLUSTERSIZE
) {
2308 mlog(ML_ERROR
, "Volume has invalid cluster size (%d)\n",
2309 osb
->s_clustersize
);
2314 total_blocks
= ocfs2_clusters_to_blocks(osb
->sb
,
2315 le32_to_cpu(di
->i_clusters
));
2317 status
= generic_check_addressable(osb
->sb
->s_blocksize_bits
,
2320 mlog(ML_ERROR
, "Volume too large "
2321 "to mount safely on this system");
2326 if (ocfs2_setup_osb_uuid(osb
, di
->id2
.i_super
.s_uuid
,
2327 sizeof(di
->id2
.i_super
.s_uuid
))) {
2328 mlog(ML_ERROR
, "Out of memory trying to setup our uuid.\n");
2333 memcpy(&uuid_net_key
, di
->id2
.i_super
.s_uuid
, sizeof(uuid_net_key
));
2335 strncpy(osb
->vol_label
, di
->id2
.i_super
.s_label
, 63);
2336 osb
->vol_label
[63] = '\0';
2337 osb
->root_blkno
= le64_to_cpu(di
->id2
.i_super
.s_root_blkno
);
2338 osb
->system_dir_blkno
= le64_to_cpu(di
->id2
.i_super
.s_system_dir_blkno
);
2339 osb
->first_cluster_group_blkno
=
2340 le64_to_cpu(di
->id2
.i_super
.s_first_cluster_group
);
2341 osb
->fs_generation
= le32_to_cpu(di
->i_fs_generation
);
2342 osb
->uuid_hash
= le32_to_cpu(di
->id2
.i_super
.s_uuid_hash
);
2343 mlog(0, "vol_label: %s\n", osb
->vol_label
);
2344 mlog(0, "uuid: %s\n", osb
->uuid_str
);
2345 mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
2346 (unsigned long long)osb
->root_blkno
,
2347 (unsigned long long)osb
->system_dir_blkno
);
2349 osb
->osb_dlm_debug
= ocfs2_new_dlm_debug();
2350 if (!osb
->osb_dlm_debug
) {
2356 atomic_set(&osb
->vol_state
, VOLUME_INIT
);
2358 /* load root, system_dir, and all global system inodes */
2359 status
= ocfs2_init_global_system_inodes(osb
);
2368 inode
= ocfs2_get_system_file_inode(osb
, GLOBAL_BITMAP_SYSTEM_INODE
,
2369 OCFS2_INVALID_SLOT
);
2376 osb
->bitmap_blkno
= OCFS2_I(inode
)->ip_blkno
;
2377 osb
->osb_clusters_at_boot
= OCFS2_I(inode
)->ip_clusters
;
2380 osb
->bitmap_cpg
= ocfs2_group_bitmap_size(sb
, 0,
2381 osb
->s_feature_incompat
) * 8;
2383 status
= ocfs2_init_slot_info(osb
);
2395 * will return: -EAGAIN if it is ok to keep searching for superblocks
2396 * -EINVAL if there is a bad superblock
2399 static int ocfs2_verify_volume(struct ocfs2_dinode
*di
,
2400 struct buffer_head
*bh
,
2402 struct ocfs2_blockcheck_stats
*stats
)
2404 int status
= -EAGAIN
;
2408 if (memcmp(di
->i_signature
, OCFS2_SUPER_BLOCK_SIGNATURE
,
2409 strlen(OCFS2_SUPER_BLOCK_SIGNATURE
)) == 0) {
2410 /* We have to do a raw check of the feature here */
2411 if (le32_to_cpu(di
->id2
.i_super
.s_feature_incompat
) &
2412 OCFS2_FEATURE_INCOMPAT_META_ECC
) {
2413 status
= ocfs2_block_check_validate(bh
->b_data
,
2421 if ((1 << le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
)) != blksz
) {
2422 mlog(ML_ERROR
, "found superblock with incorrect block "
2423 "size: found %u, should be %u\n",
2424 1 << le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
),
2426 } else if (le16_to_cpu(di
->id2
.i_super
.s_major_rev_level
) !=
2427 OCFS2_MAJOR_REV_LEVEL
||
2428 le16_to_cpu(di
->id2
.i_super
.s_minor_rev_level
) !=
2429 OCFS2_MINOR_REV_LEVEL
) {
2430 mlog(ML_ERROR
, "found superblock with bad version: "
2431 "found %u.%u, should be %u.%u\n",
2432 le16_to_cpu(di
->id2
.i_super
.s_major_rev_level
),
2433 le16_to_cpu(di
->id2
.i_super
.s_minor_rev_level
),
2434 OCFS2_MAJOR_REV_LEVEL
,
2435 OCFS2_MINOR_REV_LEVEL
);
2436 } else if (bh
->b_blocknr
!= le64_to_cpu(di
->i_blkno
)) {
2437 mlog(ML_ERROR
, "bad block number on superblock: "
2438 "found %llu, should be %llu\n",
2439 (unsigned long long)le64_to_cpu(di
->i_blkno
),
2440 (unsigned long long)bh
->b_blocknr
);
2441 } else if (le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
) < 12 ||
2442 le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
) > 20) {
2443 mlog(ML_ERROR
, "bad cluster size found: %u\n",
2444 1 << le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
));
2445 } else if (!le64_to_cpu(di
->id2
.i_super
.s_root_blkno
)) {
2446 mlog(ML_ERROR
, "bad root_blkno: 0\n");
2447 } else if (!le64_to_cpu(di
->id2
.i_super
.s_system_dir_blkno
)) {
2448 mlog(ML_ERROR
, "bad system_dir_blkno: 0\n");
2449 } else if (le16_to_cpu(di
->id2
.i_super
.s_max_slots
) > OCFS2_MAX_SLOTS
) {
2451 "Superblock slots found greater than file system "
2452 "maximum: found %u, max %u\n",
2453 le16_to_cpu(di
->id2
.i_super
.s_max_slots
),
2466 static int ocfs2_check_volume(struct ocfs2_super
*osb
)
2471 struct ocfs2_dinode
*local_alloc
= NULL
; /* only used if we
2477 /* Init our journal object. */
2478 status
= ocfs2_journal_init(osb
->journal
, &dirty
);
2480 mlog(ML_ERROR
, "Could not initialize journal!\n");
2484 /* Now that journal has been initialized, check to make sure
2485 entire volume is addressable. */
2486 status
= ocfs2_journal_addressable(osb
);
2490 /* If the journal was unmounted cleanly then we don't want to
2491 * recover anything. Otherwise, journal_load will do that
2492 * dirty work for us :) */
2494 status
= ocfs2_journal_wipe(osb
->journal
, 0);
2500 mlog(ML_NOTICE
, "File system was not unmounted cleanly, "
2501 "recovering volume.\n");
2504 local
= ocfs2_mount_local(osb
);
2506 /* will play back anything left in the journal. */
2507 status
= ocfs2_journal_load(osb
->journal
, local
, dirty
);
2509 mlog(ML_ERROR
, "ocfs2 journal load failed! %d\n", status
);
2514 /* recover my local alloc if we didn't unmount cleanly. */
2515 status
= ocfs2_begin_local_alloc_recovery(osb
,
2522 /* we complete the recovery process after we've marked
2523 * ourselves as mounted. */
2526 mlog(0, "Journal loaded.\n");
2528 status
= ocfs2_load_local_alloc(osb
);
2535 /* Recovery will be completed after we've mounted the
2536 * rest of the volume. */
2538 osb
->local_alloc_copy
= local_alloc
;
2542 /* go through each journal, trylock it and if you get the
2543 * lock, and it's marked as dirty, set the bit in the recover
2544 * map and launch a recovery thread for it. */
2545 status
= ocfs2_mark_dead_nodes(osb
);
2551 status
= ocfs2_compute_replay_slots(osb
);
2564 * The routine gets called from dismount or close whenever a dismount on
2565 * volume is requested and the osb open count becomes 1.
2566 * It will remove the osb from the global list and also free up all the
2567 * initialized resources and fileobject.
2569 static void ocfs2_delete_osb(struct ocfs2_super
*osb
)
2573 /* This function assumes that the caller has the main osb resource */
2575 ocfs2_free_slot_info(osb
);
2577 kfree(osb
->osb_orphan_wipes
);
2578 kfree(osb
->slot_recovery_generations
);
2580 * This belongs in journal shutdown, but because we have to
2581 * allocate osb->journal at the start of ocfs2_initialize_osb(),
2584 kfree(osb
->journal
);
2585 if (osb
->local_alloc_copy
)
2586 kfree(osb
->local_alloc_copy
);
2587 kfree(osb
->uuid_str
);
2588 ocfs2_put_dlm_debug(osb
->osb_dlm_debug
);
2589 memset(osb
, 0, sizeof(struct ocfs2_super
));
2594 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2595 * panic(). We do not support continue-on-error operation. */
2596 static void ocfs2_handle_error(struct super_block
*sb
)
2598 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
2600 if (osb
->s_mount_opt
& OCFS2_MOUNT_ERRORS_PANIC
)
2601 panic("OCFS2: (device %s): panic forced after error\n",
2604 ocfs2_set_osb_flag(osb
, OCFS2_OSB_ERROR_FS
);
2606 if (sb
->s_flags
& MS_RDONLY
&&
2607 (ocfs2_is_soft_readonly(osb
) ||
2608 ocfs2_is_hard_readonly(osb
)))
2611 printk(KERN_CRIT
"File system is now read-only due to the potential "
2612 "of on-disk corruption. Please run fsck.ocfs2 once the file "
2613 "system is unmounted.\n");
2614 sb
->s_flags
|= MS_RDONLY
;
2615 ocfs2_set_ro_flag(osb
, 0);
2618 static char error_buf
[1024];
2620 void __ocfs2_error(struct super_block
*sb
,
2621 const char *function
,
2622 const char *fmt
, ...)
2626 va_start(args
, fmt
);
2627 vsnprintf(error_buf
, sizeof(error_buf
), fmt
, args
);
2630 /* Not using mlog here because we want to show the actual
2631 * function the error came from. */
2632 printk(KERN_CRIT
"OCFS2: ERROR (device %s): %s: %s\n",
2633 sb
->s_id
, function
, error_buf
);
2635 ocfs2_handle_error(sb
);
2638 /* Handle critical errors. This is intentionally more drastic than
2639 * ocfs2_handle_error, so we only use for things like journal errors,
2641 void __ocfs2_abort(struct super_block
* sb
,
2642 const char *function
,
2643 const char *fmt
, ...)
2647 va_start(args
, fmt
);
2648 vsnprintf(error_buf
, sizeof(error_buf
), fmt
, args
);
2651 printk(KERN_CRIT
"OCFS2: abort (device %s): %s: %s\n",
2652 sb
->s_id
, function
, error_buf
);
2654 /* We don't have the cluster support yet to go straight to
2655 * hard readonly in here. Until then, we want to keep
2656 * ocfs2_abort() so that we can at least mark critical
2659 * TODO: This should abort the journal and alert other nodes
2660 * that our slot needs recovery. */
2662 /* Force a panic(). This stinks, but it's better than letting
2663 * things continue without having a proper hard readonly
2665 if (!ocfs2_mount_local(OCFS2_SB(sb
)))
2666 OCFS2_SB(sb
)->s_mount_opt
|= OCFS2_MOUNT_ERRORS_PANIC
;
2667 ocfs2_handle_error(sb
);
2671 * Void signal blockers, because in-kernel sigprocmask() only fails
2672 * when SIG_* is wrong.
2674 void ocfs2_block_signals(sigset_t
*oldset
)
2679 sigfillset(&blocked
);
2680 rc
= sigprocmask(SIG_BLOCK
, &blocked
, oldset
);
2684 void ocfs2_unblock_signals(sigset_t
*oldset
)
2686 int rc
= sigprocmask(SIG_SETMASK
, oldset
, NULL
);
2690 module_init(ocfs2_init
);
2691 module_exit(ocfs2_exit
);