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/cleancache.h>
46 #define CREATE_TRACE_POINTS
47 #include "ocfs2_trace.h"
49 #include <cluster/masklog.h>
53 /* this should be the only file to include a version 1 header */
54 #include "ocfs1_fs_compat.h"
58 #include "blockcheck.h"
61 #include "extent_map.h"
62 #include "heartbeat.h"
65 #include "localalloc.h"
73 #include "refcounttree.h"
76 #include "buffer_head_io.h"
78 static struct kmem_cache
*ocfs2_inode_cachep
;
79 struct kmem_cache
*ocfs2_dquot_cachep
;
80 struct kmem_cache
*ocfs2_qf_chunk_cachep
;
82 /* OCFS2 needs to schedule several different types of work which
83 * require cluster locking, disk I/O, recovery waits, etc. Since these
84 * types of work tend to be heavy we avoid using the kernel events
85 * workqueue and schedule on our own. */
86 struct workqueue_struct
*ocfs2_wq
= NULL
;
88 static struct dentry
*ocfs2_debugfs_root
;
90 MODULE_AUTHOR("Oracle");
91 MODULE_LICENSE("GPL");
92 MODULE_DESCRIPTION("OCFS2 cluster file system");
96 unsigned long commit_interval
;
97 unsigned long mount_opt
;
98 unsigned int atime_quantum
;
101 unsigned int resv_level
;
103 char cluster_stack
[OCFS2_STACK_LABEL_LEN
+ 1];
106 static int ocfs2_parse_options(struct super_block
*sb
, char *options
,
107 struct mount_options
*mopt
,
109 static int ocfs2_check_set_options(struct super_block
*sb
,
110 struct mount_options
*options
);
111 static int ocfs2_show_options(struct seq_file
*s
, struct dentry
*root
);
112 static void ocfs2_put_super(struct super_block
*sb
);
113 static int ocfs2_mount_volume(struct super_block
*sb
);
114 static int ocfs2_remount(struct super_block
*sb
, int *flags
, char *data
);
115 static void ocfs2_dismount_volume(struct super_block
*sb
, int mnt_err
);
116 static int ocfs2_initialize_mem_caches(void);
117 static void ocfs2_free_mem_caches(void);
118 static void ocfs2_delete_osb(struct ocfs2_super
*osb
);
120 static int ocfs2_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
122 static int ocfs2_sync_fs(struct super_block
*sb
, int wait
);
124 static int ocfs2_init_global_system_inodes(struct ocfs2_super
*osb
);
125 static int ocfs2_init_local_system_inodes(struct ocfs2_super
*osb
);
126 static void ocfs2_release_system_inodes(struct ocfs2_super
*osb
);
127 static int ocfs2_check_volume(struct ocfs2_super
*osb
);
128 static int ocfs2_verify_volume(struct ocfs2_dinode
*di
,
129 struct buffer_head
*bh
,
131 struct ocfs2_blockcheck_stats
*stats
);
132 static int ocfs2_initialize_super(struct super_block
*sb
,
133 struct buffer_head
*bh
,
135 struct ocfs2_blockcheck_stats
*stats
);
136 static int ocfs2_get_sector(struct super_block
*sb
,
137 struct buffer_head
**bh
,
140 static struct inode
*ocfs2_alloc_inode(struct super_block
*sb
);
141 static void ocfs2_destroy_inode(struct inode
*inode
);
142 static int ocfs2_susp_quotas(struct ocfs2_super
*osb
, int unsuspend
);
143 static int ocfs2_enable_quotas(struct ocfs2_super
*osb
);
144 static void ocfs2_disable_quotas(struct ocfs2_super
*osb
);
146 static struct dquot
**ocfs2_get_dquots(struct inode
*inode
)
148 return OCFS2_I(inode
)->i_dquot
;
151 static const struct super_operations ocfs2_sops
= {
152 .statfs
= ocfs2_statfs
,
153 .alloc_inode
= ocfs2_alloc_inode
,
154 .destroy_inode
= ocfs2_destroy_inode
,
155 .drop_inode
= ocfs2_drop_inode
,
156 .evict_inode
= ocfs2_evict_inode
,
157 .sync_fs
= ocfs2_sync_fs
,
158 .put_super
= ocfs2_put_super
,
159 .remount_fs
= ocfs2_remount
,
160 .show_options
= ocfs2_show_options
,
161 .quota_read
= ocfs2_quota_read
,
162 .quota_write
= ocfs2_quota_write
,
163 .get_dquots
= ocfs2_get_dquots
,
190 Opt_coherency_buffered
,
197 static const match_table_t tokens
= {
198 {Opt_barrier
, "barrier=%u"},
199 {Opt_err_panic
, "errors=panic"},
200 {Opt_err_ro
, "errors=remount-ro"},
202 {Opt_nointr
, "nointr"},
203 {Opt_hb_none
, OCFS2_HB_NONE
},
204 {Opt_hb_local
, OCFS2_HB_LOCAL
},
205 {Opt_hb_global
, OCFS2_HB_GLOBAL
},
206 {Opt_data_ordered
, "data=ordered"},
207 {Opt_data_writeback
, "data=writeback"},
208 {Opt_atime_quantum
, "atime_quantum=%u"},
209 {Opt_slot
, "preferred_slot=%u"},
210 {Opt_commit
, "commit=%u"},
211 {Opt_localalloc
, "localalloc=%d"},
212 {Opt_localflocks
, "localflocks"},
213 {Opt_stack
, "cluster_stack=%s"},
214 {Opt_user_xattr
, "user_xattr"},
215 {Opt_nouser_xattr
, "nouser_xattr"},
216 {Opt_inode64
, "inode64"},
218 {Opt_noacl
, "noacl"},
219 {Opt_usrquota
, "usrquota"},
220 {Opt_grpquota
, "grpquota"},
221 {Opt_coherency_buffered
, "coherency=buffered"},
222 {Opt_coherency_full
, "coherency=full"},
223 {Opt_resv_level
, "resv_level=%u"},
224 {Opt_dir_resv_level
, "dir_resv_level=%u"},
228 #ifdef CONFIG_DEBUG_FS
229 static int ocfs2_osb_dump(struct ocfs2_super
*osb
, char *buf
, int len
)
231 struct ocfs2_cluster_connection
*cconn
= osb
->cconn
;
232 struct ocfs2_recovery_map
*rm
= osb
->recovery_map
;
233 struct ocfs2_orphan_scan
*os
= &osb
->osb_orphan_scan
;
236 out
+= snprintf(buf
+ out
, len
- out
,
237 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n",
238 "Device", osb
->dev_str
, osb
->uuid_str
,
239 osb
->fs_generation
, osb
->vol_label
);
241 out
+= snprintf(buf
+ out
, len
- out
,
242 "%10s => State: %d Flags: 0x%lX\n", "Volume",
243 atomic_read(&osb
->vol_state
), osb
->osb_flags
);
245 out
+= snprintf(buf
+ out
, len
- out
,
246 "%10s => Block: %lu Cluster: %d\n", "Sizes",
247 osb
->sb
->s_blocksize
, osb
->s_clustersize
);
249 out
+= snprintf(buf
+ out
, len
- out
,
250 "%10s => Compat: 0x%X Incompat: 0x%X "
252 "Features", osb
->s_feature_compat
,
253 osb
->s_feature_incompat
, osb
->s_feature_ro_compat
);
255 out
+= snprintf(buf
+ out
, len
- out
,
256 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount",
257 osb
->s_mount_opt
, osb
->s_atime_quantum
);
260 out
+= snprintf(buf
+ out
, len
- out
,
261 "%10s => Stack: %s Name: %*s "
262 "Version: %d.%d\n", "Cluster",
263 (*osb
->osb_cluster_stack
== '\0' ?
264 "o2cb" : osb
->osb_cluster_stack
),
265 cconn
->cc_namelen
, cconn
->cc_name
,
266 cconn
->cc_version
.pv_major
,
267 cconn
->cc_version
.pv_minor
);
270 spin_lock(&osb
->dc_task_lock
);
271 out
+= snprintf(buf
+ out
, len
- out
,
272 "%10s => Pid: %d Count: %lu WakeSeq: %lu "
273 "WorkSeq: %lu\n", "DownCnvt",
274 (osb
->dc_task
? task_pid_nr(osb
->dc_task
) : -1),
275 osb
->blocked_lock_count
, osb
->dc_wake_sequence
,
276 osb
->dc_work_sequence
);
277 spin_unlock(&osb
->dc_task_lock
);
279 spin_lock(&osb
->osb_lock
);
280 out
+= snprintf(buf
+ out
, len
- out
, "%10s => Pid: %d Nodes:",
282 (osb
->recovery_thread_task
?
283 task_pid_nr(osb
->recovery_thread_task
) : -1));
284 if (rm
->rm_used
== 0)
285 out
+= snprintf(buf
+ out
, len
- out
, " None\n");
287 for (i
= 0; i
< rm
->rm_used
; i
++)
288 out
+= snprintf(buf
+ out
, len
- out
, " %d",
290 out
+= snprintf(buf
+ out
, len
- out
, "\n");
292 spin_unlock(&osb
->osb_lock
);
294 out
+= snprintf(buf
+ out
, len
- out
,
295 "%10s => Pid: %d Interval: %lu\n", "Commit",
296 (osb
->commit_task
? task_pid_nr(osb
->commit_task
) : -1),
297 osb
->osb_commit_interval
);
299 out
+= snprintf(buf
+ out
, len
- out
,
300 "%10s => State: %d TxnId: %lu NumTxns: %d\n",
301 "Journal", osb
->journal
->j_state
,
302 osb
->journal
->j_trans_id
,
303 atomic_read(&osb
->journal
->j_num_trans
));
305 out
+= snprintf(buf
+ out
, len
- out
,
306 "%10s => GlobalAllocs: %d LocalAllocs: %d "
307 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
309 atomic_read(&osb
->alloc_stats
.bitmap_data
),
310 atomic_read(&osb
->alloc_stats
.local_data
),
311 atomic_read(&osb
->alloc_stats
.bg_allocs
),
312 atomic_read(&osb
->alloc_stats
.moves
),
313 atomic_read(&osb
->alloc_stats
.bg_extends
));
315 out
+= snprintf(buf
+ out
, len
- out
,
316 "%10s => State: %u Descriptor: %llu Size: %u bits "
317 "Default: %u bits\n",
318 "LocalAlloc", osb
->local_alloc_state
,
319 (unsigned long long)osb
->la_last_gd
,
320 osb
->local_alloc_bits
, osb
->local_alloc_default_bits
);
322 spin_lock(&osb
->osb_lock
);
323 out
+= snprintf(buf
+ out
, len
- out
,
324 "%10s => InodeSlot: %d StolenInodes: %d, "
325 "MetaSlot: %d StolenMeta: %d\n", "Steal",
326 osb
->s_inode_steal_slot
,
327 atomic_read(&osb
->s_num_inodes_stolen
),
328 osb
->s_meta_steal_slot
,
329 atomic_read(&osb
->s_num_meta_stolen
));
330 spin_unlock(&osb
->osb_lock
);
332 out
+= snprintf(buf
+ out
, len
- out
, "OrphanScan => ");
333 out
+= snprintf(buf
+ out
, len
- out
, "Local: %u Global: %u ",
334 os
->os_count
, os
->os_seqno
);
335 out
+= snprintf(buf
+ out
, len
- out
, " Last Scan: ");
336 if (atomic_read(&os
->os_state
) == ORPHAN_SCAN_INACTIVE
)
337 out
+= snprintf(buf
+ out
, len
- out
, "Disabled\n");
339 out
+= snprintf(buf
+ out
, len
- out
, "%lu seconds ago\n",
340 (get_seconds() - os
->os_scantime
.tv_sec
));
342 out
+= snprintf(buf
+ out
, len
- out
, "%10s => %3s %10s\n",
343 "Slots", "Num", "RecoGen");
344 for (i
= 0; i
< osb
->max_slots
; ++i
) {
345 out
+= snprintf(buf
+ out
, len
- out
,
346 "%10s %c %3d %10d\n",
348 (i
== osb
->slot_num
? '*' : ' '),
349 i
, osb
->slot_recovery_generations
[i
]);
355 static int ocfs2_osb_debug_open(struct inode
*inode
, struct file
*file
)
357 struct ocfs2_super
*osb
= inode
->i_private
;
360 buf
= kmalloc(PAGE_SIZE
, GFP_KERNEL
);
364 i_size_write(inode
, ocfs2_osb_dump(osb
, buf
, PAGE_SIZE
));
366 file
->private_data
= buf
;
373 static int ocfs2_debug_release(struct inode
*inode
, struct file
*file
)
375 kfree(file
->private_data
);
379 static ssize_t
ocfs2_debug_read(struct file
*file
, char __user
*buf
,
380 size_t nbytes
, loff_t
*ppos
)
382 return simple_read_from_buffer(buf
, nbytes
, ppos
, file
->private_data
,
383 i_size_read(file
->f_mapping
->host
));
386 static int ocfs2_osb_debug_open(struct inode
*inode
, struct file
*file
)
390 static int ocfs2_debug_release(struct inode
*inode
, struct file
*file
)
394 static ssize_t
ocfs2_debug_read(struct file
*file
, char __user
*buf
,
395 size_t nbytes
, loff_t
*ppos
)
399 #endif /* CONFIG_DEBUG_FS */
401 static const struct file_operations ocfs2_osb_debug_fops
= {
402 .open
= ocfs2_osb_debug_open
,
403 .release
= ocfs2_debug_release
,
404 .read
= ocfs2_debug_read
,
405 .llseek
= generic_file_llseek
,
408 static int ocfs2_sync_fs(struct super_block
*sb
, int wait
)
412 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
414 if (ocfs2_is_hard_readonly(osb
))
418 status
= ocfs2_flush_truncate_log(osb
);
422 ocfs2_schedule_truncate_log_flush(osb
, 0);
425 if (jbd2_journal_start_commit(OCFS2_SB(sb
)->journal
->j_journal
,
428 jbd2_log_wait_commit(OCFS2_SB(sb
)->journal
->j_journal
,
434 static int ocfs2_need_system_inode(struct ocfs2_super
*osb
, int ino
)
436 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb
->sb
, OCFS2_FEATURE_RO_COMPAT_USRQUOTA
)
437 && (ino
== USER_QUOTA_SYSTEM_INODE
438 || ino
== LOCAL_USER_QUOTA_SYSTEM_INODE
))
440 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb
->sb
, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
)
441 && (ino
== GROUP_QUOTA_SYSTEM_INODE
442 || ino
== LOCAL_GROUP_QUOTA_SYSTEM_INODE
))
447 static int ocfs2_init_global_system_inodes(struct ocfs2_super
*osb
)
449 struct inode
*new = NULL
;
453 new = ocfs2_iget(osb
, osb
->root_blkno
, OCFS2_FI_FLAG_SYSFILE
, 0);
455 status
= PTR_ERR(new);
459 osb
->root_inode
= new;
461 new = ocfs2_iget(osb
, osb
->system_dir_blkno
, OCFS2_FI_FLAG_SYSFILE
, 0);
463 status
= PTR_ERR(new);
467 osb
->sys_root_inode
= new;
469 for (i
= OCFS2_FIRST_ONLINE_SYSTEM_INODE
;
470 i
<= OCFS2_LAST_GLOBAL_SYSTEM_INODE
; i
++) {
471 if (!ocfs2_need_system_inode(osb
, i
))
473 new = ocfs2_get_system_file_inode(osb
, i
, osb
->slot_num
);
475 ocfs2_release_system_inodes(osb
);
478 /* FIXME: Should ERROR_RO_FS */
479 mlog(ML_ERROR
, "Unable to load system inode %d, "
480 "possibly corrupt fs?", i
);
483 // the array now has one ref, so drop this one
493 static int ocfs2_init_local_system_inodes(struct ocfs2_super
*osb
)
495 struct inode
*new = NULL
;
499 for (i
= OCFS2_LAST_GLOBAL_SYSTEM_INODE
+ 1;
500 i
< NUM_SYSTEM_INODES
;
502 if (!ocfs2_need_system_inode(osb
, i
))
504 new = ocfs2_get_system_file_inode(osb
, i
, osb
->slot_num
);
506 ocfs2_release_system_inodes(osb
);
508 mlog(ML_ERROR
, "status=%d, sysfile=%d, slot=%d\n",
509 status
, i
, osb
->slot_num
);
512 /* the array now has one ref, so drop this one */
522 static void ocfs2_release_system_inodes(struct ocfs2_super
*osb
)
527 for (i
= 0; i
< NUM_GLOBAL_SYSTEM_INODES
; i
++) {
528 inode
= osb
->global_system_inodes
[i
];
531 osb
->global_system_inodes
[i
] = NULL
;
535 inode
= osb
->sys_root_inode
;
538 osb
->sys_root_inode
= NULL
;
541 inode
= osb
->root_inode
;
544 osb
->root_inode
= NULL
;
547 if (!osb
->local_system_inodes
)
550 for (i
= 0; i
< NUM_LOCAL_SYSTEM_INODES
* osb
->max_slots
; i
++) {
551 if (osb
->local_system_inodes
[i
]) {
552 iput(osb
->local_system_inodes
[i
]);
553 osb
->local_system_inodes
[i
] = NULL
;
557 kfree(osb
->local_system_inodes
);
558 osb
->local_system_inodes
= NULL
;
561 /* We're allocating fs objects, use GFP_NOFS */
562 static struct inode
*ocfs2_alloc_inode(struct super_block
*sb
)
564 struct ocfs2_inode_info
*oi
;
566 oi
= kmem_cache_alloc(ocfs2_inode_cachep
, GFP_NOFS
);
571 oi
->i_datasync_tid
= 0;
572 memset(&oi
->i_dquot
, 0, sizeof(oi
->i_dquot
));
574 jbd2_journal_init_jbd_inode(&oi
->ip_jinode
, &oi
->vfs_inode
);
575 return &oi
->vfs_inode
;
578 static void ocfs2_i_callback(struct rcu_head
*head
)
580 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
581 kmem_cache_free(ocfs2_inode_cachep
, OCFS2_I(inode
));
584 static void ocfs2_destroy_inode(struct inode
*inode
)
586 call_rcu(&inode
->i_rcu
, ocfs2_i_callback
);
589 static unsigned long long ocfs2_max_file_offset(unsigned int bbits
,
592 unsigned int bytes
= 1 << cbits
;
593 unsigned int trim
= bytes
;
594 unsigned int bitshift
= 32;
597 * i_size and all block offsets in ocfs2 are always 64 bits
598 * wide. i_clusters is 32 bits, in cluster-sized units. So on
599 * 64 bit platforms, cluster size will be the limiting factor.
602 #if BITS_PER_LONG == 32
603 # if defined(CONFIG_LBDAF)
604 BUILD_BUG_ON(sizeof(sector_t
) != 8);
606 * We might be limited by page cache size.
608 if (bytes
> PAGE_CACHE_SIZE
) {
609 bytes
= PAGE_CACHE_SIZE
;
612 * Shift by 31 here so that we don't get larger than
619 * We are limited by the size of sector_t. Use block size, as
620 * that's what we expose to the VFS.
629 * Trim by a whole cluster when we can actually approach the
630 * on-disk limits. Otherwise we can overflow i_clusters when
631 * an extent start is at the max offset.
633 return (((unsigned long long)bytes
) << bitshift
) - trim
;
636 static int ocfs2_remount(struct super_block
*sb
, int *flags
, char *data
)
638 int incompat_features
;
640 struct mount_options parsed_options
;
641 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
646 if (!ocfs2_parse_options(sb
, data
, &parsed_options
, 1) ||
647 !ocfs2_check_set_options(sb
, &parsed_options
)) {
652 tmp
= OCFS2_MOUNT_HB_LOCAL
| OCFS2_MOUNT_HB_GLOBAL
|
654 if ((osb
->s_mount_opt
& tmp
) != (parsed_options
.mount_opt
& tmp
)) {
656 mlog(ML_ERROR
, "Cannot change heartbeat mode on remount\n");
660 if ((osb
->s_mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
) !=
661 (parsed_options
.mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
)) {
663 mlog(ML_ERROR
, "Cannot change data mode on remount\n");
667 /* Probably don't want this on remount; it might
668 * mess with other nodes */
669 if (!(osb
->s_mount_opt
& OCFS2_MOUNT_INODE64
) &&
670 (parsed_options
.mount_opt
& OCFS2_MOUNT_INODE64
)) {
672 mlog(ML_ERROR
, "Cannot enable inode64 on remount\n");
676 /* We're going to/from readonly mode. */
677 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
)) {
678 /* Disable quota accounting before remounting RO */
679 if (*flags
& MS_RDONLY
) {
680 ret
= ocfs2_susp_quotas(osb
, 0);
684 /* Lock here so the check of HARD_RO and the potential
685 * setting of SOFT_RO is atomic. */
686 spin_lock(&osb
->osb_lock
);
687 if (osb
->osb_flags
& OCFS2_OSB_HARD_RO
) {
688 mlog(ML_ERROR
, "Remount on readonly device is forbidden.\n");
693 if (*flags
& MS_RDONLY
) {
694 sb
->s_flags
|= MS_RDONLY
;
695 osb
->osb_flags
|= OCFS2_OSB_SOFT_RO
;
697 if (osb
->osb_flags
& OCFS2_OSB_ERROR_FS
) {
698 mlog(ML_ERROR
, "Cannot remount RDWR "
699 "filesystem due to previous errors.\n");
703 incompat_features
= OCFS2_HAS_RO_COMPAT_FEATURE(sb
, ~OCFS2_FEATURE_RO_COMPAT_SUPP
);
704 if (incompat_features
) {
705 mlog(ML_ERROR
, "Cannot remount RDWR because "
706 "of unsupported optional features "
707 "(%x).\n", incompat_features
);
711 sb
->s_flags
&= ~MS_RDONLY
;
712 osb
->osb_flags
&= ~OCFS2_OSB_SOFT_RO
;
714 trace_ocfs2_remount(sb
->s_flags
, osb
->osb_flags
, *flags
);
716 spin_unlock(&osb
->osb_lock
);
717 /* Enable quota accounting after remounting RW */
718 if (!ret
&& !(*flags
& MS_RDONLY
)) {
719 if (sb_any_quota_suspended(sb
))
720 ret
= ocfs2_susp_quotas(osb
, 1);
722 ret
= ocfs2_enable_quotas(osb
);
724 /* Return back changes... */
725 spin_lock(&osb
->osb_lock
);
726 sb
->s_flags
|= MS_RDONLY
;
727 osb
->osb_flags
|= OCFS2_OSB_SOFT_RO
;
728 spin_unlock(&osb
->osb_lock
);
735 /* Only save off the new mount options in case of a successful
737 osb
->s_mount_opt
= parsed_options
.mount_opt
;
738 osb
->s_atime_quantum
= parsed_options
.atime_quantum
;
739 osb
->preferred_slot
= parsed_options
.slot
;
740 if (parsed_options
.commit_interval
)
741 osb
->osb_commit_interval
= parsed_options
.commit_interval
;
743 if (!ocfs2_is_hard_readonly(osb
))
744 ocfs2_set_journal_params(osb
);
746 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
747 ((osb
->s_mount_opt
& OCFS2_MOUNT_POSIX_ACL
) ?
754 static int ocfs2_sb_probe(struct super_block
*sb
,
755 struct buffer_head
**bh
,
757 struct ocfs2_blockcheck_stats
*stats
)
760 struct ocfs1_vol_disk_hdr
*hdr
;
761 struct ocfs2_dinode
*di
;
767 *sector_size
= bdev_logical_block_size(sb
->s_bdev
);
768 if (*sector_size
> OCFS2_MAX_BLOCKSIZE
) {
769 mlog(ML_ERROR
, "Hardware sector size too large: %d (max=%d)\n",
770 *sector_size
, OCFS2_MAX_BLOCKSIZE
);
775 /* Can this really happen? */
776 if (*sector_size
< OCFS2_MIN_BLOCKSIZE
)
777 *sector_size
= OCFS2_MIN_BLOCKSIZE
;
779 /* check block zero for old format */
780 status
= ocfs2_get_sector(sb
, bh
, 0, *sector_size
);
785 hdr
= (struct ocfs1_vol_disk_hdr
*) (*bh
)->b_data
;
786 if (hdr
->major_version
== OCFS1_MAJOR_VERSION
) {
787 mlog(ML_ERROR
, "incompatible version: %u.%u\n",
788 hdr
->major_version
, hdr
->minor_version
);
791 if (memcmp(hdr
->signature
, OCFS1_VOLUME_SIGNATURE
,
792 strlen(OCFS1_VOLUME_SIGNATURE
)) == 0) {
793 mlog(ML_ERROR
, "incompatible volume signature: %8s\n",
800 mlog(ML_ERROR
, "This is an ocfs v1 filesystem which must be "
801 "upgraded before mounting with ocfs v2\n");
806 * Now check at magic offset for 512, 1024, 2048, 4096
807 * blocksizes. 4096 is the maximum blocksize because it is
808 * the minimum clustersize.
811 for (blksize
= *sector_size
;
812 blksize
<= OCFS2_MAX_BLOCKSIZE
;
814 tmpstat
= ocfs2_get_sector(sb
, bh
,
815 OCFS2_SUPER_BLOCK_BLKNO
,
822 di
= (struct ocfs2_dinode
*) (*bh
)->b_data
;
823 memset(stats
, 0, sizeof(struct ocfs2_blockcheck_stats
));
824 spin_lock_init(&stats
->b_lock
);
825 tmpstat
= ocfs2_verify_volume(di
, *bh
, blksize
, stats
);
830 if (tmpstat
!= -EAGAIN
) {
840 static int ocfs2_verify_heartbeat(struct ocfs2_super
*osb
)
842 u32 hb_enabled
= OCFS2_MOUNT_HB_LOCAL
| OCFS2_MOUNT_HB_GLOBAL
;
844 if (osb
->s_mount_opt
& hb_enabled
) {
845 if (ocfs2_mount_local(osb
)) {
846 mlog(ML_ERROR
, "Cannot heartbeat on a locally "
847 "mounted device.\n");
850 if (ocfs2_userspace_stack(osb
)) {
851 mlog(ML_ERROR
, "Userspace stack expected, but "
852 "o2cb heartbeat arguments passed to mount\n");
855 if (((osb
->s_mount_opt
& OCFS2_MOUNT_HB_GLOBAL
) &&
856 !ocfs2_cluster_o2cb_global_heartbeat(osb
)) ||
857 ((osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) &&
858 ocfs2_cluster_o2cb_global_heartbeat(osb
))) {
859 mlog(ML_ERROR
, "Mismatching o2cb heartbeat modes\n");
864 if (!(osb
->s_mount_opt
& hb_enabled
)) {
865 if (!ocfs2_mount_local(osb
) && !ocfs2_is_hard_readonly(osb
) &&
866 !ocfs2_userspace_stack(osb
)) {
867 mlog(ML_ERROR
, "Heartbeat has to be started to mount "
868 "a read-write clustered device.\n");
877 * If we're using a userspace stack, mount should have passed
878 * a name that matches the disk. If not, mount should not
879 * have passed a stack.
881 static int ocfs2_verify_userspace_stack(struct ocfs2_super
*osb
,
882 struct mount_options
*mopt
)
884 if (!ocfs2_userspace_stack(osb
) && mopt
->cluster_stack
[0]) {
886 "cluster stack passed to mount, but this filesystem "
887 "does not support it\n");
891 if (ocfs2_userspace_stack(osb
) &&
892 strncmp(osb
->osb_cluster_stack
, mopt
->cluster_stack
,
893 OCFS2_STACK_LABEL_LEN
)) {
895 "cluster stack passed to mount (\"%s\") does not "
896 "match the filesystem (\"%s\")\n",
898 osb
->osb_cluster_stack
);
905 static int ocfs2_susp_quotas(struct ocfs2_super
*osb
, int unsuspend
)
908 struct super_block
*sb
= osb
->sb
;
909 unsigned int feature
[OCFS2_MAXQUOTAS
] = {
910 OCFS2_FEATURE_RO_COMPAT_USRQUOTA
,
911 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
};
914 for (type
= 0; type
< OCFS2_MAXQUOTAS
; type
++) {
915 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb
, feature
[type
]))
918 status
= dquot_resume(sb
, type
);
920 struct ocfs2_mem_dqinfo
*oinfo
;
922 /* Cancel periodic syncing before suspending */
923 oinfo
= sb_dqinfo(sb
, type
)->dqi_priv
;
924 cancel_delayed_work_sync(&oinfo
->dqi_sync_work
);
925 status
= dquot_suspend(sb
, type
);
931 mlog(ML_ERROR
, "Failed to suspend/unsuspend quotas on "
932 "remount (error = %d).\n", status
);
936 static int ocfs2_enable_quotas(struct ocfs2_super
*osb
)
938 struct inode
*inode
[OCFS2_MAXQUOTAS
] = { NULL
, NULL
};
939 struct super_block
*sb
= osb
->sb
;
940 unsigned int feature
[OCFS2_MAXQUOTAS
] = {
941 OCFS2_FEATURE_RO_COMPAT_USRQUOTA
,
942 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
};
943 unsigned int ino
[OCFS2_MAXQUOTAS
] = {
944 LOCAL_USER_QUOTA_SYSTEM_INODE
,
945 LOCAL_GROUP_QUOTA_SYSTEM_INODE
};
949 sb_dqopt(sb
)->flags
|= DQUOT_QUOTA_SYS_FILE
| DQUOT_NEGATIVE_USAGE
;
950 for (type
= 0; type
< OCFS2_MAXQUOTAS
; type
++) {
951 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb
, feature
[type
]))
953 inode
[type
] = ocfs2_get_system_file_inode(osb
, ino
[type
],
959 status
= dquot_enable(inode
[type
], type
, QFMT_OCFS2
,
960 DQUOT_USAGE_ENABLED
);
965 for (type
= 0; type
< OCFS2_MAXQUOTAS
; type
++)
969 ocfs2_disable_quotas(osb
);
970 for (type
= 0; type
< OCFS2_MAXQUOTAS
; type
++)
976 static void ocfs2_disable_quotas(struct ocfs2_super
*osb
)
980 struct super_block
*sb
= osb
->sb
;
981 struct ocfs2_mem_dqinfo
*oinfo
;
983 /* We mostly ignore errors in this function because there's not much
984 * we can do when we see them */
985 for (type
= 0; type
< OCFS2_MAXQUOTAS
; type
++) {
986 if (!sb_has_quota_loaded(sb
, type
))
988 /* Cancel periodic syncing before we grab dqonoff_mutex */
989 oinfo
= sb_dqinfo(sb
, type
)->dqi_priv
;
990 cancel_delayed_work_sync(&oinfo
->dqi_sync_work
);
991 inode
= igrab(sb
->s_dquot
.files
[type
]);
992 /* Turn off quotas. This will remove all dquot structures from
993 * memory and so they will be automatically synced to global
995 dquot_disable(sb
, type
, DQUOT_USAGE_ENABLED
|
996 DQUOT_LIMITS_ENABLED
);
1003 /* Handle quota on quotactl */
1004 static int ocfs2_quota_on(struct super_block
*sb
, int type
, int format_id
)
1006 unsigned int feature
[OCFS2_MAXQUOTAS
] = {
1007 OCFS2_FEATURE_RO_COMPAT_USRQUOTA
,
1008 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
};
1010 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb
, feature
[type
]))
1013 return dquot_enable(sb_dqopt(sb
)->files
[type
], type
,
1014 format_id
, DQUOT_LIMITS_ENABLED
);
1017 /* Handle quota off quotactl */
1018 static int ocfs2_quota_off(struct super_block
*sb
, int type
)
1020 return dquot_disable(sb
, type
, DQUOT_LIMITS_ENABLED
);
1023 static const struct quotactl_ops ocfs2_quotactl_ops
= {
1024 .quota_on_meta
= ocfs2_quota_on
,
1025 .quota_off
= ocfs2_quota_off
,
1026 .quota_sync
= dquot_quota_sync
,
1027 .get_info
= dquot_get_dqinfo
,
1028 .set_info
= dquot_set_dqinfo
,
1029 .get_dqblk
= dquot_get_dqblk
,
1030 .set_dqblk
= dquot_set_dqblk
,
1033 static int ocfs2_fill_super(struct super_block
*sb
, void *data
, int silent
)
1035 struct dentry
*root
;
1036 int status
, sector_size
;
1037 struct mount_options parsed_options
;
1038 struct inode
*inode
= NULL
;
1039 struct ocfs2_super
*osb
= NULL
;
1040 struct buffer_head
*bh
= NULL
;
1042 struct ocfs2_blockcheck_stats stats
;
1044 trace_ocfs2_fill_super(sb
, data
, silent
);
1046 if (!ocfs2_parse_options(sb
, data
, &parsed_options
, 0)) {
1048 goto read_super_error
;
1051 /* probe for superblock */
1052 status
= ocfs2_sb_probe(sb
, &bh
, §or_size
, &stats
);
1054 mlog(ML_ERROR
, "superblock probe failed!\n");
1055 goto read_super_error
;
1058 status
= ocfs2_initialize_super(sb
, bh
, sector_size
, &stats
);
1062 goto read_super_error
;
1067 if (!ocfs2_check_set_options(sb
, &parsed_options
)) {
1069 goto read_super_error
;
1071 osb
->s_mount_opt
= parsed_options
.mount_opt
;
1072 osb
->s_atime_quantum
= parsed_options
.atime_quantum
;
1073 osb
->preferred_slot
= parsed_options
.slot
;
1074 osb
->osb_commit_interval
= parsed_options
.commit_interval
;
1076 ocfs2_la_set_sizes(osb
, parsed_options
.localalloc_opt
);
1077 osb
->osb_resv_level
= parsed_options
.resv_level
;
1078 osb
->osb_dir_resv_level
= parsed_options
.resv_level
;
1079 if (parsed_options
.dir_resv_level
== -1)
1080 osb
->osb_dir_resv_level
= parsed_options
.resv_level
;
1082 osb
->osb_dir_resv_level
= parsed_options
.dir_resv_level
;
1084 status
= ocfs2_verify_userspace_stack(osb
, &parsed_options
);
1086 goto read_super_error
;
1088 sb
->s_magic
= OCFS2_SUPER_MAGIC
;
1090 sb
->s_flags
= (sb
->s_flags
& ~(MS_POSIXACL
| MS_NOSEC
)) |
1091 ((osb
->s_mount_opt
& OCFS2_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
1093 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1095 if (bdev_read_only(sb
->s_bdev
)) {
1096 if (!(sb
->s_flags
& MS_RDONLY
)) {
1098 mlog(ML_ERROR
, "Readonly device detected but readonly "
1099 "mount was not specified.\n");
1100 goto read_super_error
;
1103 /* You should not be able to start a local heartbeat
1104 * on a readonly device. */
1105 if (osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) {
1107 mlog(ML_ERROR
, "Local heartbeat specified on readonly "
1109 goto read_super_error
;
1112 status
= ocfs2_check_journals_nolocks(osb
);
1114 if (status
== -EROFS
)
1115 mlog(ML_ERROR
, "Recovery required on readonly "
1116 "file system, but write access is "
1120 goto read_super_error
;
1123 ocfs2_set_ro_flag(osb
, 1);
1125 printk(KERN_NOTICE
"ocfs2: Readonly device (%s) detected. "
1126 "Cluster services will not be used for this mount. "
1127 "Recovery will be skipped.\n", osb
->dev_str
);
1130 if (!ocfs2_is_hard_readonly(osb
)) {
1131 if (sb
->s_flags
& MS_RDONLY
)
1132 ocfs2_set_ro_flag(osb
, 0);
1135 status
= ocfs2_verify_heartbeat(osb
);
1138 goto read_super_error
;
1141 osb
->osb_debug_root
= debugfs_create_dir(osb
->uuid_str
,
1142 ocfs2_debugfs_root
);
1143 if (!osb
->osb_debug_root
) {
1145 mlog(ML_ERROR
, "Unable to create per-mount debugfs root.\n");
1146 goto read_super_error
;
1149 osb
->osb_ctxt
= debugfs_create_file("fs_state", S_IFREG
|S_IRUSR
,
1150 osb
->osb_debug_root
,
1152 &ocfs2_osb_debug_fops
);
1153 if (!osb
->osb_ctxt
) {
1156 goto read_super_error
;
1159 if (ocfs2_meta_ecc(osb
)) {
1160 status
= ocfs2_blockcheck_stats_debugfs_install(
1161 &osb
->osb_ecc_stats
,
1162 osb
->osb_debug_root
);
1165 "Unable to create blockcheck statistics "
1167 goto read_super_error
;
1171 status
= ocfs2_mount_volume(sb
);
1173 goto read_super_error
;
1175 if (osb
->root_inode
)
1176 inode
= igrab(osb
->root_inode
);
1181 goto read_super_error
;
1184 root
= d_make_root(inode
);
1188 goto read_super_error
;
1193 ocfs2_complete_mount_recovery(osb
);
1195 if (ocfs2_mount_local(osb
))
1196 snprintf(nodestr
, sizeof(nodestr
), "local");
1198 snprintf(nodestr
, sizeof(nodestr
), "%u", osb
->node_num
);
1200 printk(KERN_INFO
"ocfs2: Mounting device (%s) on (node %s, slot %d) "
1201 "with %s data mode.\n",
1202 osb
->dev_str
, nodestr
, osb
->slot_num
,
1203 osb
->s_mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
? "writeback" :
1206 atomic_set(&osb
->vol_state
, VOLUME_MOUNTED
);
1207 wake_up(&osb
->osb_mount_event
);
1209 /* Now we can initialize quotas because we can afford to wait
1210 * for cluster locks recovery now. That also means that truncation
1211 * log recovery can happen but that waits for proper quota setup */
1212 if (!(sb
->s_flags
& MS_RDONLY
)) {
1213 status
= ocfs2_enable_quotas(osb
);
1215 /* We have to err-out specially here because
1216 * s_root is already set */
1218 atomic_set(&osb
->vol_state
, VOLUME_DISABLED
);
1219 wake_up(&osb
->osb_mount_event
);
1224 ocfs2_complete_quota_recovery(osb
);
1226 /* Now we wake up again for processes waiting for quotas */
1227 atomic_set(&osb
->vol_state
, VOLUME_MOUNTED_QUOTAS
);
1228 wake_up(&osb
->osb_mount_event
);
1230 /* Start this when the mount is almost sure of being successful */
1231 ocfs2_orphan_scan_start(osb
);
1239 atomic_set(&osb
->vol_state
, VOLUME_DISABLED
);
1240 wake_up(&osb
->osb_mount_event
);
1241 ocfs2_dismount_volume(sb
, 1);
1249 static struct dentry
*ocfs2_mount(struct file_system_type
*fs_type
,
1251 const char *dev_name
,
1254 return mount_bdev(fs_type
, flags
, dev_name
, data
, ocfs2_fill_super
);
1257 static struct file_system_type ocfs2_fs_type
= {
1258 .owner
= THIS_MODULE
,
1260 .mount
= ocfs2_mount
,
1261 .kill_sb
= kill_block_super
,
1262 .fs_flags
= FS_REQUIRES_DEV
|FS_RENAME_DOES_D_MOVE
,
1265 MODULE_ALIAS_FS("ocfs2");
1267 static int ocfs2_check_set_options(struct super_block
*sb
,
1268 struct mount_options
*options
)
1270 if (options
->mount_opt
& OCFS2_MOUNT_USRQUOTA
&&
1271 !OCFS2_HAS_RO_COMPAT_FEATURE(sb
,
1272 OCFS2_FEATURE_RO_COMPAT_USRQUOTA
)) {
1273 mlog(ML_ERROR
, "User quotas were requested, but this "
1274 "filesystem does not have the feature enabled.\n");
1277 if (options
->mount_opt
& OCFS2_MOUNT_GRPQUOTA
&&
1278 !OCFS2_HAS_RO_COMPAT_FEATURE(sb
,
1279 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
)) {
1280 mlog(ML_ERROR
, "Group quotas were requested, but this "
1281 "filesystem does not have the feature enabled.\n");
1284 if (options
->mount_opt
& OCFS2_MOUNT_POSIX_ACL
&&
1285 !OCFS2_HAS_INCOMPAT_FEATURE(sb
, OCFS2_FEATURE_INCOMPAT_XATTR
)) {
1286 mlog(ML_ERROR
, "ACL support requested but extended attributes "
1287 "feature is not enabled\n");
1290 /* No ACL setting specified? Use XATTR feature... */
1291 if (!(options
->mount_opt
& (OCFS2_MOUNT_POSIX_ACL
|
1292 OCFS2_MOUNT_NO_POSIX_ACL
))) {
1293 if (OCFS2_HAS_INCOMPAT_FEATURE(sb
, OCFS2_FEATURE_INCOMPAT_XATTR
))
1294 options
->mount_opt
|= OCFS2_MOUNT_POSIX_ACL
;
1296 options
->mount_opt
|= OCFS2_MOUNT_NO_POSIX_ACL
;
1301 static int ocfs2_parse_options(struct super_block
*sb
,
1303 struct mount_options
*mopt
,
1306 int status
, user_stack
= 0;
1310 trace_ocfs2_parse_options(is_remount
, options
? options
: "(none)");
1312 mopt
->commit_interval
= 0;
1313 mopt
->mount_opt
= OCFS2_MOUNT_NOINTR
;
1314 mopt
->atime_quantum
= OCFS2_DEFAULT_ATIME_QUANTUM
;
1315 mopt
->slot
= OCFS2_INVALID_SLOT
;
1316 mopt
->localalloc_opt
= -1;
1317 mopt
->cluster_stack
[0] = '\0';
1318 mopt
->resv_level
= OCFS2_DEFAULT_RESV_LEVEL
;
1319 mopt
->dir_resv_level
= -1;
1326 while ((p
= strsep(&options
, ",")) != NULL
) {
1328 substring_t args
[MAX_OPT_ARGS
];
1333 token
= match_token(p
, tokens
, args
);
1336 mopt
->mount_opt
|= OCFS2_MOUNT_HB_LOCAL
;
1339 mopt
->mount_opt
|= OCFS2_MOUNT_HB_NONE
;
1342 mopt
->mount_opt
|= OCFS2_MOUNT_HB_GLOBAL
;
1345 if (match_int(&args
[0], &option
)) {
1350 mopt
->mount_opt
|= OCFS2_MOUNT_BARRIER
;
1352 mopt
->mount_opt
&= ~OCFS2_MOUNT_BARRIER
;
1355 mopt
->mount_opt
&= ~OCFS2_MOUNT_NOINTR
;
1358 mopt
->mount_opt
|= OCFS2_MOUNT_NOINTR
;
1361 mopt
->mount_opt
|= OCFS2_MOUNT_ERRORS_PANIC
;
1364 mopt
->mount_opt
&= ~OCFS2_MOUNT_ERRORS_PANIC
;
1366 case Opt_data_ordered
:
1367 mopt
->mount_opt
&= ~OCFS2_MOUNT_DATA_WRITEBACK
;
1369 case Opt_data_writeback
:
1370 mopt
->mount_opt
|= OCFS2_MOUNT_DATA_WRITEBACK
;
1372 case Opt_user_xattr
:
1373 mopt
->mount_opt
&= ~OCFS2_MOUNT_NOUSERXATTR
;
1375 case Opt_nouser_xattr
:
1376 mopt
->mount_opt
|= OCFS2_MOUNT_NOUSERXATTR
;
1378 case Opt_atime_quantum
:
1379 if (match_int(&args
[0], &option
)) {
1384 mopt
->atime_quantum
= option
;
1388 if (match_int(&args
[0], &option
)) {
1393 mopt
->slot
= (s16
)option
;
1397 if (match_int(&args
[0], &option
)) {
1404 option
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1405 mopt
->commit_interval
= HZ
* option
;
1407 case Opt_localalloc
:
1409 if (match_int(&args
[0], &option
)) {
1414 mopt
->localalloc_opt
= option
;
1416 case Opt_localflocks
:
1418 * Changing this during remount could race
1419 * flock() requests, or "unbalance" existing
1420 * ones (e.g., a lock is taken in one mode but
1421 * dropped in the other). If users care enough
1422 * to flip locking modes during remount, we
1423 * could add a "local" flag to individual
1424 * flock structures for proper tracking of
1428 mopt
->mount_opt
|= OCFS2_MOUNT_LOCALFLOCKS
;
1431 /* Check both that the option we were passed
1432 * is of the right length and that it is a proper
1433 * string of the right length.
1435 if (((args
[0].to
- args
[0].from
) !=
1436 OCFS2_STACK_LABEL_LEN
) ||
1437 (strnlen(args
[0].from
,
1438 OCFS2_STACK_LABEL_LEN
) !=
1439 OCFS2_STACK_LABEL_LEN
)) {
1441 "Invalid cluster_stack option\n");
1445 memcpy(mopt
->cluster_stack
, args
[0].from
,
1446 OCFS2_STACK_LABEL_LEN
);
1447 mopt
->cluster_stack
[OCFS2_STACK_LABEL_LEN
] = '\0';
1449 * Open code the memcmp here as we don't have
1451 * ocfs2_userspace_stack().
1453 if (memcmp(mopt
->cluster_stack
,
1454 OCFS2_CLASSIC_CLUSTER_STACK
,
1455 OCFS2_STACK_LABEL_LEN
))
1459 mopt
->mount_opt
|= OCFS2_MOUNT_INODE64
;
1462 mopt
->mount_opt
|= OCFS2_MOUNT_USRQUOTA
;
1465 mopt
->mount_opt
|= OCFS2_MOUNT_GRPQUOTA
;
1467 case Opt_coherency_buffered
:
1468 mopt
->mount_opt
|= OCFS2_MOUNT_COHERENCY_BUFFERED
;
1470 case Opt_coherency_full
:
1471 mopt
->mount_opt
&= ~OCFS2_MOUNT_COHERENCY_BUFFERED
;
1474 mopt
->mount_opt
|= OCFS2_MOUNT_POSIX_ACL
;
1475 mopt
->mount_opt
&= ~OCFS2_MOUNT_NO_POSIX_ACL
;
1478 mopt
->mount_opt
|= OCFS2_MOUNT_NO_POSIX_ACL
;
1479 mopt
->mount_opt
&= ~OCFS2_MOUNT_POSIX_ACL
;
1481 case Opt_resv_level
:
1484 if (match_int(&args
[0], &option
)) {
1488 if (option
>= OCFS2_MIN_RESV_LEVEL
&&
1489 option
< OCFS2_MAX_RESV_LEVEL
)
1490 mopt
->resv_level
= option
;
1492 case Opt_dir_resv_level
:
1495 if (match_int(&args
[0], &option
)) {
1499 if (option
>= OCFS2_MIN_RESV_LEVEL
&&
1500 option
< OCFS2_MAX_RESV_LEVEL
)
1501 mopt
->dir_resv_level
= option
;
1505 "Unrecognized mount option \"%s\" "
1506 "or missing value\n", p
);
1512 if (user_stack
== 0) {
1513 /* Ensure only one heartbeat mode */
1514 tmp
= mopt
->mount_opt
& (OCFS2_MOUNT_HB_LOCAL
|
1515 OCFS2_MOUNT_HB_GLOBAL
|
1516 OCFS2_MOUNT_HB_NONE
);
1517 if (hweight32(tmp
) != 1) {
1518 mlog(ML_ERROR
, "Invalid heartbeat mount options\n");
1530 static int ocfs2_show_options(struct seq_file
*s
, struct dentry
*root
)
1532 struct ocfs2_super
*osb
= OCFS2_SB(root
->d_sb
);
1533 unsigned long opts
= osb
->s_mount_opt
;
1534 unsigned int local_alloc_megs
;
1536 if (opts
& (OCFS2_MOUNT_HB_LOCAL
| OCFS2_MOUNT_HB_GLOBAL
)) {
1537 seq_printf(s
, ",_netdev");
1538 if (opts
& OCFS2_MOUNT_HB_LOCAL
)
1539 seq_printf(s
, ",%s", OCFS2_HB_LOCAL
);
1541 seq_printf(s
, ",%s", OCFS2_HB_GLOBAL
);
1543 seq_printf(s
, ",%s", OCFS2_HB_NONE
);
1545 if (opts
& OCFS2_MOUNT_NOINTR
)
1546 seq_printf(s
, ",nointr");
1548 if (opts
& OCFS2_MOUNT_DATA_WRITEBACK
)
1549 seq_printf(s
, ",data=writeback");
1551 seq_printf(s
, ",data=ordered");
1553 if (opts
& OCFS2_MOUNT_BARRIER
)
1554 seq_printf(s
, ",barrier=1");
1556 if (opts
& OCFS2_MOUNT_ERRORS_PANIC
)
1557 seq_printf(s
, ",errors=panic");
1559 seq_printf(s
, ",errors=remount-ro");
1561 if (osb
->preferred_slot
!= OCFS2_INVALID_SLOT
)
1562 seq_printf(s
, ",preferred_slot=%d", osb
->preferred_slot
);
1564 seq_printf(s
, ",atime_quantum=%u", osb
->s_atime_quantum
);
1566 if (osb
->osb_commit_interval
)
1567 seq_printf(s
, ",commit=%u",
1568 (unsigned) (osb
->osb_commit_interval
/ HZ
));
1570 local_alloc_megs
= osb
->local_alloc_bits
>> (20 - osb
->s_clustersize_bits
);
1571 if (local_alloc_megs
!= ocfs2_la_default_mb(osb
))
1572 seq_printf(s
, ",localalloc=%d", local_alloc_megs
);
1574 if (opts
& OCFS2_MOUNT_LOCALFLOCKS
)
1575 seq_printf(s
, ",localflocks,");
1577 if (osb
->osb_cluster_stack
[0])
1578 seq_printf(s
, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN
,
1579 osb
->osb_cluster_stack
);
1580 if (opts
& OCFS2_MOUNT_USRQUOTA
)
1581 seq_printf(s
, ",usrquota");
1582 if (opts
& OCFS2_MOUNT_GRPQUOTA
)
1583 seq_printf(s
, ",grpquota");
1585 if (opts
& OCFS2_MOUNT_COHERENCY_BUFFERED
)
1586 seq_printf(s
, ",coherency=buffered");
1588 seq_printf(s
, ",coherency=full");
1590 if (opts
& OCFS2_MOUNT_NOUSERXATTR
)
1591 seq_printf(s
, ",nouser_xattr");
1593 seq_printf(s
, ",user_xattr");
1595 if (opts
& OCFS2_MOUNT_INODE64
)
1596 seq_printf(s
, ",inode64");
1598 if (opts
& OCFS2_MOUNT_POSIX_ACL
)
1599 seq_printf(s
, ",acl");
1601 seq_printf(s
, ",noacl");
1603 if (osb
->osb_resv_level
!= OCFS2_DEFAULT_RESV_LEVEL
)
1604 seq_printf(s
, ",resv_level=%d", osb
->osb_resv_level
);
1606 if (osb
->osb_dir_resv_level
!= osb
->osb_resv_level
)
1607 seq_printf(s
, ",dir_resv_level=%d", osb
->osb_resv_level
);
1612 static int __init
ocfs2_init(void)
1616 status
= init_ocfs2_uptodate_cache();
1620 status
= ocfs2_initialize_mem_caches();
1624 ocfs2_wq
= create_singlethread_workqueue("ocfs2_wq");
1630 ocfs2_debugfs_root
= debugfs_create_dir("ocfs2", NULL
);
1631 if (!ocfs2_debugfs_root
) {
1633 mlog(ML_ERROR
, "Unable to create ocfs2 debugfs root.\n");
1637 ocfs2_set_locking_protocol();
1639 status
= register_quota_format(&ocfs2_quota_format
);
1642 status
= register_filesystem(&ocfs2_fs_type
);
1646 unregister_quota_format(&ocfs2_quota_format
);
1648 destroy_workqueue(ocfs2_wq
);
1649 debugfs_remove(ocfs2_debugfs_root
);
1651 ocfs2_free_mem_caches();
1653 exit_ocfs2_uptodate_cache();
1659 static void __exit
ocfs2_exit(void)
1662 flush_workqueue(ocfs2_wq
);
1663 destroy_workqueue(ocfs2_wq
);
1666 unregister_quota_format(&ocfs2_quota_format
);
1668 debugfs_remove(ocfs2_debugfs_root
);
1670 ocfs2_free_mem_caches();
1672 unregister_filesystem(&ocfs2_fs_type
);
1674 exit_ocfs2_uptodate_cache();
1677 static void ocfs2_put_super(struct super_block
*sb
)
1679 trace_ocfs2_put_super(sb
);
1681 ocfs2_sync_blockdev(sb
);
1682 ocfs2_dismount_volume(sb
, 0);
1685 static int ocfs2_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
1687 struct ocfs2_super
*osb
;
1688 u32 numbits
, freebits
;
1690 struct ocfs2_dinode
*bm_lock
;
1691 struct buffer_head
*bh
= NULL
;
1692 struct inode
*inode
= NULL
;
1694 trace_ocfs2_statfs(dentry
->d_sb
, buf
);
1696 osb
= OCFS2_SB(dentry
->d_sb
);
1698 inode
= ocfs2_get_system_file_inode(osb
,
1699 GLOBAL_BITMAP_SYSTEM_INODE
,
1700 OCFS2_INVALID_SLOT
);
1702 mlog(ML_ERROR
, "failed to get bitmap inode\n");
1707 status
= ocfs2_inode_lock(inode
, &bh
, 0);
1713 bm_lock
= (struct ocfs2_dinode
*) bh
->b_data
;
1715 numbits
= le32_to_cpu(bm_lock
->id1
.bitmap1
.i_total
);
1716 freebits
= numbits
- le32_to_cpu(bm_lock
->id1
.bitmap1
.i_used
);
1718 buf
->f_type
= OCFS2_SUPER_MAGIC
;
1719 buf
->f_bsize
= dentry
->d_sb
->s_blocksize
;
1720 buf
->f_namelen
= OCFS2_MAX_FILENAME_LEN
;
1721 buf
->f_blocks
= ((sector_t
) numbits
) *
1722 (osb
->s_clustersize
>> osb
->sb
->s_blocksize_bits
);
1723 buf
->f_bfree
= ((sector_t
) freebits
) *
1724 (osb
->s_clustersize
>> osb
->sb
->s_blocksize_bits
);
1725 buf
->f_bavail
= buf
->f_bfree
;
1726 buf
->f_files
= numbits
;
1727 buf
->f_ffree
= freebits
;
1728 buf
->f_fsid
.val
[0] = crc32_le(0, osb
->uuid_str
, OCFS2_VOL_UUID_LEN
)
1730 buf
->f_fsid
.val
[1] = crc32_le(0, osb
->uuid_str
+ OCFS2_VOL_UUID_LEN
,
1731 OCFS2_VOL_UUID_LEN
) & 0xFFFFFFFFUL
;
1735 ocfs2_inode_unlock(inode
, 0);
1747 static void ocfs2_inode_init_once(void *data
)
1749 struct ocfs2_inode_info
*oi
= data
;
1752 oi
->ip_open_count
= 0;
1753 spin_lock_init(&oi
->ip_lock
);
1754 ocfs2_extent_map_init(&oi
->vfs_inode
);
1755 INIT_LIST_HEAD(&oi
->ip_io_markers
);
1756 oi
->ip_dir_start_lookup
= 0;
1757 mutex_init(&oi
->ip_unaligned_aio
);
1758 init_rwsem(&oi
->ip_alloc_sem
);
1759 init_rwsem(&oi
->ip_xattr_sem
);
1760 mutex_init(&oi
->ip_io_mutex
);
1762 oi
->ip_blkno
= 0ULL;
1763 oi
->ip_clusters
= 0;
1765 ocfs2_resv_init_once(&oi
->ip_la_data_resv
);
1767 ocfs2_lock_res_init_once(&oi
->ip_rw_lockres
);
1768 ocfs2_lock_res_init_once(&oi
->ip_inode_lockres
);
1769 ocfs2_lock_res_init_once(&oi
->ip_open_lockres
);
1771 ocfs2_metadata_cache_init(INODE_CACHE(&oi
->vfs_inode
),
1772 &ocfs2_inode_caching_ops
);
1774 inode_init_once(&oi
->vfs_inode
);
1777 static int ocfs2_initialize_mem_caches(void)
1779 ocfs2_inode_cachep
= kmem_cache_create("ocfs2_inode_cache",
1780 sizeof(struct ocfs2_inode_info
),
1782 (SLAB_HWCACHE_ALIGN
|SLAB_RECLAIM_ACCOUNT
|
1784 ocfs2_inode_init_once
);
1785 ocfs2_dquot_cachep
= kmem_cache_create("ocfs2_dquot_cache",
1786 sizeof(struct ocfs2_dquot
),
1788 (SLAB_HWCACHE_ALIGN
|SLAB_RECLAIM_ACCOUNT
|
1791 ocfs2_qf_chunk_cachep
= kmem_cache_create("ocfs2_qf_chunk_cache",
1792 sizeof(struct ocfs2_quota_chunk
),
1794 (SLAB_RECLAIM_ACCOUNT
|SLAB_MEM_SPREAD
),
1796 if (!ocfs2_inode_cachep
|| !ocfs2_dquot_cachep
||
1797 !ocfs2_qf_chunk_cachep
) {
1798 if (ocfs2_inode_cachep
)
1799 kmem_cache_destroy(ocfs2_inode_cachep
);
1800 if (ocfs2_dquot_cachep
)
1801 kmem_cache_destroy(ocfs2_dquot_cachep
);
1802 if (ocfs2_qf_chunk_cachep
)
1803 kmem_cache_destroy(ocfs2_qf_chunk_cachep
);
1810 static void ocfs2_free_mem_caches(void)
1813 * Make sure all delayed rcu free inodes are flushed before we
1817 if (ocfs2_inode_cachep
)
1818 kmem_cache_destroy(ocfs2_inode_cachep
);
1819 ocfs2_inode_cachep
= NULL
;
1821 if (ocfs2_dquot_cachep
)
1822 kmem_cache_destroy(ocfs2_dquot_cachep
);
1823 ocfs2_dquot_cachep
= NULL
;
1825 if (ocfs2_qf_chunk_cachep
)
1826 kmem_cache_destroy(ocfs2_qf_chunk_cachep
);
1827 ocfs2_qf_chunk_cachep
= NULL
;
1830 static int ocfs2_get_sector(struct super_block
*sb
,
1831 struct buffer_head
**bh
,
1835 if (!sb_set_blocksize(sb
, sect_size
)) {
1836 mlog(ML_ERROR
, "unable to set blocksize\n");
1840 *bh
= sb_getblk(sb
, block
);
1842 mlog_errno(-ENOMEM
);
1846 if (!buffer_dirty(*bh
))
1847 clear_buffer_uptodate(*bh
);
1849 ll_rw_block(READ
, 1, bh
);
1850 wait_on_buffer(*bh
);
1851 if (!buffer_uptodate(*bh
)) {
1861 static int ocfs2_mount_volume(struct super_block
*sb
)
1864 int unlock_super
= 0;
1865 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
1867 if (ocfs2_is_hard_readonly(osb
))
1870 status
= ocfs2_dlm_init(osb
);
1876 status
= ocfs2_super_lock(osb
, 1);
1883 /* This will load up the node map and add ourselves to it. */
1884 status
= ocfs2_find_slot(osb
);
1890 /* load all node-local system inodes */
1891 status
= ocfs2_init_local_system_inodes(osb
);
1897 status
= ocfs2_check_volume(osb
);
1903 status
= ocfs2_truncate_log_init(osb
);
1909 ocfs2_super_unlock(osb
, 1);
1914 static void ocfs2_dismount_volume(struct super_block
*sb
, int mnt_err
)
1916 int tmp
, hangup_needed
= 0;
1917 struct ocfs2_super
*osb
= NULL
;
1920 trace_ocfs2_dismount_volume(sb
);
1926 debugfs_remove(osb
->osb_ctxt
);
1928 /* Orphan scan should be stopped as early as possible */
1929 ocfs2_orphan_scan_stop(osb
);
1931 ocfs2_disable_quotas(osb
);
1933 /* All dquots should be freed by now */
1934 WARN_ON(!llist_empty(&osb
->dquot_drop_list
));
1935 /* Wait for worker to be done with the work structure in osb */
1936 cancel_work_sync(&osb
->dquot_drop_work
);
1938 ocfs2_shutdown_local_alloc(osb
);
1940 ocfs2_truncate_log_shutdown(osb
);
1942 /* This will disable recovery and flush any recovery work. */
1943 ocfs2_recovery_exit(osb
);
1945 ocfs2_journal_shutdown(osb
);
1947 ocfs2_sync_blockdev(sb
);
1949 ocfs2_purge_refcount_trees(osb
);
1951 /* No cluster connection means we've failed during mount, so skip
1952 * all the steps which depended on that to complete. */
1954 tmp
= ocfs2_super_lock(osb
, 1);
1961 if (osb
->slot_num
!= OCFS2_INVALID_SLOT
)
1962 ocfs2_put_slot(osb
);
1965 ocfs2_super_unlock(osb
, 1);
1967 ocfs2_release_system_inodes(osb
);
1970 * If we're dismounting due to mount error, mount.ocfs2 will clean
1971 * up heartbeat. If we're a local mount, there is no heartbeat.
1972 * If we failed before we got a uuid_str yet, we can't stop
1973 * heartbeat. Otherwise, do it.
1975 if (!mnt_err
&& !ocfs2_mount_local(osb
) && osb
->uuid_str
&&
1976 !ocfs2_is_hard_readonly(osb
))
1980 ocfs2_dlm_shutdown(osb
, hangup_needed
);
1982 ocfs2_blockcheck_stats_debugfs_remove(&osb
->osb_ecc_stats
);
1983 debugfs_remove(osb
->osb_debug_root
);
1986 ocfs2_cluster_hangup(osb
->uuid_str
, strlen(osb
->uuid_str
));
1988 atomic_set(&osb
->vol_state
, VOLUME_DISMOUNTED
);
1990 if (ocfs2_mount_local(osb
))
1991 snprintf(nodestr
, sizeof(nodestr
), "local");
1993 snprintf(nodestr
, sizeof(nodestr
), "%u", osb
->node_num
);
1995 printk(KERN_INFO
"ocfs2: Unmounting device (%s) on (node %s)\n",
1996 osb
->dev_str
, nodestr
);
1998 ocfs2_delete_osb(osb
);
2001 sb
->s_fs_info
= NULL
;
2004 static int ocfs2_setup_osb_uuid(struct ocfs2_super
*osb
, const unsigned char *uuid
,
2005 unsigned uuid_bytes
)
2010 BUG_ON(uuid_bytes
!= OCFS2_VOL_UUID_LEN
);
2012 osb
->uuid_str
= kzalloc(OCFS2_VOL_UUID_LEN
* 2 + 1, GFP_KERNEL
);
2013 if (osb
->uuid_str
== NULL
)
2016 for (i
= 0, ptr
= osb
->uuid_str
; i
< OCFS2_VOL_UUID_LEN
; i
++) {
2017 /* print with null */
2018 ret
= snprintf(ptr
, 3, "%02X", uuid
[i
]);
2019 if (ret
!= 2) /* drop super cleans up */
2021 /* then only advance past the last char */
2028 /* Make sure entire volume is addressable by our journal. Requires
2029 osb_clusters_at_boot to be valid and for the journal to have been
2030 initialized by ocfs2_journal_init(). */
2031 static int ocfs2_journal_addressable(struct ocfs2_super
*osb
)
2035 ocfs2_clusters_to_blocks(osb
->sb
,
2036 osb
->osb_clusters_at_boot
) - 1;
2038 /* 32-bit block number is always OK. */
2039 if (max_block
<= (u32
)~0ULL)
2042 /* Volume is "huge", so see if our journal is new enough to
2044 if (!(OCFS2_HAS_COMPAT_FEATURE(osb
->sb
,
2045 OCFS2_FEATURE_COMPAT_JBD2_SB
) &&
2046 jbd2_journal_check_used_features(osb
->journal
->j_journal
, 0, 0,
2047 JBD2_FEATURE_INCOMPAT_64BIT
))) {
2048 mlog(ML_ERROR
, "The journal cannot address the entire volume. "
2049 "Enable the 'block64' journal option with tunefs.ocfs2");
2058 static int ocfs2_initialize_super(struct super_block
*sb
,
2059 struct buffer_head
*bh
,
2061 struct ocfs2_blockcheck_stats
*stats
)
2064 int i
, cbits
, bbits
;
2065 struct ocfs2_dinode
*di
= (struct ocfs2_dinode
*)bh
->b_data
;
2066 struct inode
*inode
= NULL
;
2067 struct ocfs2_journal
*journal
;
2068 struct ocfs2_super
*osb
;
2071 osb
= kzalloc(sizeof(struct ocfs2_super
), GFP_KERNEL
);
2078 sb
->s_fs_info
= osb
;
2079 sb
->s_op
= &ocfs2_sops
;
2080 sb
->s_d_op
= &ocfs2_dentry_ops
;
2081 sb
->s_export_op
= &ocfs2_export_ops
;
2082 sb
->s_qcop
= &ocfs2_quotactl_ops
;
2083 sb
->dq_op
= &ocfs2_quota_operations
;
2084 sb
->s_quota_types
= QTYPE_MASK_USR
| QTYPE_MASK_GRP
;
2085 sb
->s_xattr
= ocfs2_xattr_handlers
;
2086 sb
->s_time_gran
= 1;
2087 sb
->s_flags
|= MS_NOATIME
;
2088 /* this is needed to support O_LARGEFILE */
2089 cbits
= le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
);
2090 bbits
= le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
);
2091 sb
->s_maxbytes
= ocfs2_max_file_offset(bbits
, cbits
);
2093 osb
->osb_dx_mask
= (1 << (cbits
- bbits
)) - 1;
2095 for (i
= 0; i
< 3; i
++)
2096 osb
->osb_dx_seed
[i
] = le32_to_cpu(di
->id2
.i_super
.s_dx_seed
[i
]);
2097 osb
->osb_dx_seed
[3] = le32_to_cpu(di
->id2
.i_super
.s_uuid_hash
);
2100 /* Save off for ocfs2_rw_direct */
2101 osb
->s_sectsize_bits
= blksize_bits(sector_size
);
2102 BUG_ON(!osb
->s_sectsize_bits
);
2104 spin_lock_init(&osb
->dc_task_lock
);
2105 init_waitqueue_head(&osb
->dc_event
);
2106 osb
->dc_work_sequence
= 0;
2107 osb
->dc_wake_sequence
= 0;
2108 INIT_LIST_HEAD(&osb
->blocked_lock_list
);
2109 osb
->blocked_lock_count
= 0;
2110 spin_lock_init(&osb
->osb_lock
);
2111 spin_lock_init(&osb
->osb_xattr_lock
);
2112 ocfs2_init_steal_slots(osb
);
2114 mutex_init(&osb
->system_file_mutex
);
2116 atomic_set(&osb
->alloc_stats
.moves
, 0);
2117 atomic_set(&osb
->alloc_stats
.local_data
, 0);
2118 atomic_set(&osb
->alloc_stats
.bitmap_data
, 0);
2119 atomic_set(&osb
->alloc_stats
.bg_allocs
, 0);
2120 atomic_set(&osb
->alloc_stats
.bg_extends
, 0);
2122 /* Copy the blockcheck stats from the superblock probe */
2123 osb
->osb_ecc_stats
= *stats
;
2125 ocfs2_init_node_maps(osb
);
2127 snprintf(osb
->dev_str
, sizeof(osb
->dev_str
), "%u,%u",
2128 MAJOR(osb
->sb
->s_dev
), MINOR(osb
->sb
->s_dev
));
2130 osb
->max_slots
= le16_to_cpu(di
->id2
.i_super
.s_max_slots
);
2131 if (osb
->max_slots
> OCFS2_MAX_SLOTS
|| osb
->max_slots
== 0) {
2132 mlog(ML_ERROR
, "Invalid number of node slots (%u)\n",
2138 ocfs2_orphan_scan_init(osb
);
2140 status
= ocfs2_recovery_init(osb
);
2142 mlog(ML_ERROR
, "Unable to initialize recovery state\n");
2147 init_waitqueue_head(&osb
->checkpoint_event
);
2149 osb
->s_atime_quantum
= OCFS2_DEFAULT_ATIME_QUANTUM
;
2151 osb
->slot_num
= OCFS2_INVALID_SLOT
;
2153 osb
->s_xattr_inline_size
= le16_to_cpu(
2154 di
->id2
.i_super
.s_xattr_inline_size
);
2156 osb
->local_alloc_state
= OCFS2_LA_UNUSED
;
2157 osb
->local_alloc_bh
= NULL
;
2158 INIT_DELAYED_WORK(&osb
->la_enable_wq
, ocfs2_la_enable_worker
);
2160 init_waitqueue_head(&osb
->osb_mount_event
);
2162 status
= ocfs2_resmap_init(osb
, &osb
->osb_la_resmap
);
2168 osb
->vol_label
= kmalloc(OCFS2_MAX_VOL_LABEL_LEN
, GFP_KERNEL
);
2169 if (!osb
->vol_label
) {
2170 mlog(ML_ERROR
, "unable to alloc vol label\n");
2175 osb
->slot_recovery_generations
=
2176 kcalloc(osb
->max_slots
, sizeof(*osb
->slot_recovery_generations
),
2178 if (!osb
->slot_recovery_generations
) {
2184 init_waitqueue_head(&osb
->osb_wipe_event
);
2185 osb
->osb_orphan_wipes
= kcalloc(osb
->max_slots
,
2186 sizeof(*osb
->osb_orphan_wipes
),
2188 if (!osb
->osb_orphan_wipes
) {
2194 osb
->osb_rf_lock_tree
= RB_ROOT
;
2196 osb
->s_feature_compat
=
2197 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_compat
);
2198 osb
->s_feature_ro_compat
=
2199 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_ro_compat
);
2200 osb
->s_feature_incompat
=
2201 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_incompat
);
2203 if ((i
= OCFS2_HAS_INCOMPAT_FEATURE(osb
->sb
, ~OCFS2_FEATURE_INCOMPAT_SUPP
))) {
2204 mlog(ML_ERROR
, "couldn't mount because of unsupported "
2205 "optional features (%x).\n", i
);
2209 if (!(osb
->sb
->s_flags
& MS_RDONLY
) &&
2210 (i
= OCFS2_HAS_RO_COMPAT_FEATURE(osb
->sb
, ~OCFS2_FEATURE_RO_COMPAT_SUPP
))) {
2211 mlog(ML_ERROR
, "couldn't mount RDWR because of "
2212 "unsupported optional features (%x).\n", i
);
2217 if (ocfs2_clusterinfo_valid(osb
)) {
2218 osb
->osb_stackflags
=
2219 OCFS2_RAW_SB(di
)->s_cluster_info
.ci_stackflags
;
2220 strlcpy(osb
->osb_cluster_stack
,
2221 OCFS2_RAW_SB(di
)->s_cluster_info
.ci_stack
,
2222 OCFS2_STACK_LABEL_LEN
+ 1);
2223 if (strlen(osb
->osb_cluster_stack
) != OCFS2_STACK_LABEL_LEN
) {
2225 "couldn't mount because of an invalid "
2226 "cluster stack label (%s) \n",
2227 osb
->osb_cluster_stack
);
2231 strlcpy(osb
->osb_cluster_name
,
2232 OCFS2_RAW_SB(di
)->s_cluster_info
.ci_cluster
,
2233 OCFS2_CLUSTER_NAME_LEN
+ 1);
2235 /* The empty string is identical with classic tools that
2236 * don't know about s_cluster_info. */
2237 osb
->osb_cluster_stack
[0] = '\0';
2240 get_random_bytes(&osb
->s_next_generation
, sizeof(u32
));
2243 * This should be done in ocfs2_journal_init(), but unknown
2244 * ordering issues will cause the filesystem to crash.
2245 * If anyone wants to figure out what part of the code
2246 * refers to osb->journal before ocfs2_journal_init() is run,
2249 /* initialize our journal structure */
2251 journal
= kzalloc(sizeof(struct ocfs2_journal
), GFP_KERNEL
);
2253 mlog(ML_ERROR
, "unable to alloc journal\n");
2257 osb
->journal
= journal
;
2258 journal
->j_osb
= osb
;
2260 atomic_set(&journal
->j_num_trans
, 0);
2261 init_rwsem(&journal
->j_trans_barrier
);
2262 init_waitqueue_head(&journal
->j_checkpointed
);
2263 spin_lock_init(&journal
->j_lock
);
2264 journal
->j_trans_id
= (unsigned long) 1;
2265 INIT_LIST_HEAD(&journal
->j_la_cleanups
);
2266 INIT_WORK(&journal
->j_recovery_work
, ocfs2_complete_recovery
);
2267 journal
->j_state
= OCFS2_JOURNAL_FREE
;
2269 INIT_WORK(&osb
->dquot_drop_work
, ocfs2_drop_dquot_refs
);
2270 init_llist_head(&osb
->dquot_drop_list
);
2272 /* get some pseudo constants for clustersize bits */
2273 osb
->s_clustersize_bits
=
2274 le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
);
2275 osb
->s_clustersize
= 1 << osb
->s_clustersize_bits
;
2277 if (osb
->s_clustersize
< OCFS2_MIN_CLUSTERSIZE
||
2278 osb
->s_clustersize
> OCFS2_MAX_CLUSTERSIZE
) {
2279 mlog(ML_ERROR
, "Volume has invalid cluster size (%d)\n",
2280 osb
->s_clustersize
);
2285 total_blocks
= ocfs2_clusters_to_blocks(osb
->sb
,
2286 le32_to_cpu(di
->i_clusters
));
2288 status
= generic_check_addressable(osb
->sb
->s_blocksize_bits
,
2291 mlog(ML_ERROR
, "Volume too large "
2292 "to mount safely on this system");
2297 if (ocfs2_setup_osb_uuid(osb
, di
->id2
.i_super
.s_uuid
,
2298 sizeof(di
->id2
.i_super
.s_uuid
))) {
2299 mlog(ML_ERROR
, "Out of memory trying to setup our uuid.\n");
2304 strlcpy(osb
->vol_label
, di
->id2
.i_super
.s_label
,
2305 OCFS2_MAX_VOL_LABEL_LEN
);
2306 osb
->root_blkno
= le64_to_cpu(di
->id2
.i_super
.s_root_blkno
);
2307 osb
->system_dir_blkno
= le64_to_cpu(di
->id2
.i_super
.s_system_dir_blkno
);
2308 osb
->first_cluster_group_blkno
=
2309 le64_to_cpu(di
->id2
.i_super
.s_first_cluster_group
);
2310 osb
->fs_generation
= le32_to_cpu(di
->i_fs_generation
);
2311 osb
->uuid_hash
= le32_to_cpu(di
->id2
.i_super
.s_uuid_hash
);
2312 trace_ocfs2_initialize_super(osb
->vol_label
, osb
->uuid_str
,
2313 (unsigned long long)osb
->root_blkno
,
2314 (unsigned long long)osb
->system_dir_blkno
,
2315 osb
->s_clustersize_bits
);
2317 osb
->osb_dlm_debug
= ocfs2_new_dlm_debug();
2318 if (!osb
->osb_dlm_debug
) {
2324 atomic_set(&osb
->vol_state
, VOLUME_INIT
);
2326 /* load root, system_dir, and all global system inodes */
2327 status
= ocfs2_init_global_system_inodes(osb
);
2336 inode
= ocfs2_get_system_file_inode(osb
, GLOBAL_BITMAP_SYSTEM_INODE
,
2337 OCFS2_INVALID_SLOT
);
2344 osb
->bitmap_blkno
= OCFS2_I(inode
)->ip_blkno
;
2345 osb
->osb_clusters_at_boot
= OCFS2_I(inode
)->ip_clusters
;
2348 osb
->bitmap_cpg
= ocfs2_group_bitmap_size(sb
, 0,
2349 osb
->s_feature_incompat
) * 8;
2351 status
= ocfs2_init_slot_info(osb
);
2356 cleancache_init_shared_fs((char *)&di
->id2
.i_super
.s_uuid
, sb
);
2363 * will return: -EAGAIN if it is ok to keep searching for superblocks
2364 * -EINVAL if there is a bad superblock
2367 static int ocfs2_verify_volume(struct ocfs2_dinode
*di
,
2368 struct buffer_head
*bh
,
2370 struct ocfs2_blockcheck_stats
*stats
)
2372 int status
= -EAGAIN
;
2374 if (memcmp(di
->i_signature
, OCFS2_SUPER_BLOCK_SIGNATURE
,
2375 strlen(OCFS2_SUPER_BLOCK_SIGNATURE
)) == 0) {
2376 /* We have to do a raw check of the feature here */
2377 if (le32_to_cpu(di
->id2
.i_super
.s_feature_incompat
) &
2378 OCFS2_FEATURE_INCOMPAT_META_ECC
) {
2379 status
= ocfs2_block_check_validate(bh
->b_data
,
2387 if ((1 << le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
)) != blksz
) {
2388 mlog(ML_ERROR
, "found superblock with incorrect block "
2389 "size: found %u, should be %u\n",
2390 1 << le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
),
2392 } else if (le16_to_cpu(di
->id2
.i_super
.s_major_rev_level
) !=
2393 OCFS2_MAJOR_REV_LEVEL
||
2394 le16_to_cpu(di
->id2
.i_super
.s_minor_rev_level
) !=
2395 OCFS2_MINOR_REV_LEVEL
) {
2396 mlog(ML_ERROR
, "found superblock with bad version: "
2397 "found %u.%u, should be %u.%u\n",
2398 le16_to_cpu(di
->id2
.i_super
.s_major_rev_level
),
2399 le16_to_cpu(di
->id2
.i_super
.s_minor_rev_level
),
2400 OCFS2_MAJOR_REV_LEVEL
,
2401 OCFS2_MINOR_REV_LEVEL
);
2402 } else if (bh
->b_blocknr
!= le64_to_cpu(di
->i_blkno
)) {
2403 mlog(ML_ERROR
, "bad block number on superblock: "
2404 "found %llu, should be %llu\n",
2405 (unsigned long long)le64_to_cpu(di
->i_blkno
),
2406 (unsigned long long)bh
->b_blocknr
);
2407 } else if (le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
) < 12 ||
2408 le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
) > 20) {
2409 mlog(ML_ERROR
, "bad cluster size found: %u\n",
2410 1 << le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
));
2411 } else if (!le64_to_cpu(di
->id2
.i_super
.s_root_blkno
)) {
2412 mlog(ML_ERROR
, "bad root_blkno: 0\n");
2413 } else if (!le64_to_cpu(di
->id2
.i_super
.s_system_dir_blkno
)) {
2414 mlog(ML_ERROR
, "bad system_dir_blkno: 0\n");
2415 } else if (le16_to_cpu(di
->id2
.i_super
.s_max_slots
) > OCFS2_MAX_SLOTS
) {
2417 "Superblock slots found greater than file system "
2418 "maximum: found %u, max %u\n",
2419 le16_to_cpu(di
->id2
.i_super
.s_max_slots
),
2428 if (status
&& status
!= -EAGAIN
)
2433 static int ocfs2_check_volume(struct ocfs2_super
*osb
)
2438 struct ocfs2_dinode
*local_alloc
= NULL
; /* only used if we
2442 /* Init our journal object. */
2443 status
= ocfs2_journal_init(osb
->journal
, &dirty
);
2445 mlog(ML_ERROR
, "Could not initialize journal!\n");
2449 /* Now that journal has been initialized, check to make sure
2450 entire volume is addressable. */
2451 status
= ocfs2_journal_addressable(osb
);
2455 /* If the journal was unmounted cleanly then we don't want to
2456 * recover anything. Otherwise, journal_load will do that
2457 * dirty work for us :) */
2459 status
= ocfs2_journal_wipe(osb
->journal
, 0);
2465 printk(KERN_NOTICE
"ocfs2: File system on device (%s) was not "
2466 "unmounted cleanly, recovering it.\n", osb
->dev_str
);
2469 local
= ocfs2_mount_local(osb
);
2471 /* will play back anything left in the journal. */
2472 status
= ocfs2_journal_load(osb
->journal
, local
, dirty
);
2474 mlog(ML_ERROR
, "ocfs2 journal load failed! %d\n", status
);
2479 /* recover my local alloc if we didn't unmount cleanly. */
2480 status
= ocfs2_begin_local_alloc_recovery(osb
,
2487 /* we complete the recovery process after we've marked
2488 * ourselves as mounted. */
2491 status
= ocfs2_load_local_alloc(osb
);
2498 /* Recovery will be completed after we've mounted the
2499 * rest of the volume. */
2501 osb
->local_alloc_copy
= local_alloc
;
2505 /* go through each journal, trylock it and if you get the
2506 * lock, and it's marked as dirty, set the bit in the recover
2507 * map and launch a recovery thread for it. */
2508 status
= ocfs2_mark_dead_nodes(osb
);
2514 status
= ocfs2_compute_replay_slots(osb
);
2527 * The routine gets called from dismount or close whenever a dismount on
2528 * volume is requested and the osb open count becomes 1.
2529 * It will remove the osb from the global list and also free up all the
2530 * initialized resources and fileobject.
2532 static void ocfs2_delete_osb(struct ocfs2_super
*osb
)
2534 /* This function assumes that the caller has the main osb resource */
2536 ocfs2_free_slot_info(osb
);
2538 kfree(osb
->osb_orphan_wipes
);
2539 kfree(osb
->slot_recovery_generations
);
2541 * This belongs in journal shutdown, but because we have to
2542 * allocate osb->journal at the start of ocfs2_initialize_osb(),
2545 kfree(osb
->journal
);
2546 kfree(osb
->local_alloc_copy
);
2547 kfree(osb
->uuid_str
);
2548 kfree(osb
->vol_label
);
2549 ocfs2_put_dlm_debug(osb
->osb_dlm_debug
);
2550 memset(osb
, 0, sizeof(struct ocfs2_super
));
2553 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2554 * panic(). We do not support continue-on-error operation. */
2555 static void ocfs2_handle_error(struct super_block
*sb
)
2557 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
2559 if (osb
->s_mount_opt
& OCFS2_MOUNT_ERRORS_PANIC
)
2560 panic("OCFS2: (device %s): panic forced after error\n",
2563 ocfs2_set_osb_flag(osb
, OCFS2_OSB_ERROR_FS
);
2565 if (sb
->s_flags
& MS_RDONLY
&&
2566 (ocfs2_is_soft_readonly(osb
) ||
2567 ocfs2_is_hard_readonly(osb
)))
2570 printk(KERN_CRIT
"File system is now read-only due to the potential "
2571 "of on-disk corruption. Please run fsck.ocfs2 once the file "
2572 "system is unmounted.\n");
2573 sb
->s_flags
|= MS_RDONLY
;
2574 ocfs2_set_ro_flag(osb
, 0);
2577 static char error_buf
[1024];
2579 void __ocfs2_error(struct super_block
*sb
,
2580 const char *function
,
2581 const char *fmt
, ...)
2585 va_start(args
, fmt
);
2586 vsnprintf(error_buf
, sizeof(error_buf
), fmt
, args
);
2589 /* Not using mlog here because we want to show the actual
2590 * function the error came from. */
2591 printk(KERN_CRIT
"OCFS2: ERROR (device %s): %s: %s\n",
2592 sb
->s_id
, function
, error_buf
);
2594 ocfs2_handle_error(sb
);
2597 /* Handle critical errors. This is intentionally more drastic than
2598 * ocfs2_handle_error, so we only use for things like journal errors,
2600 void __ocfs2_abort(struct super_block
* sb
,
2601 const char *function
,
2602 const char *fmt
, ...)
2606 va_start(args
, fmt
);
2607 vsnprintf(error_buf
, sizeof(error_buf
), fmt
, args
);
2610 printk(KERN_CRIT
"OCFS2: abort (device %s): %s: %s\n",
2611 sb
->s_id
, function
, error_buf
);
2613 /* We don't have the cluster support yet to go straight to
2614 * hard readonly in here. Until then, we want to keep
2615 * ocfs2_abort() so that we can at least mark critical
2618 * TODO: This should abort the journal and alert other nodes
2619 * that our slot needs recovery. */
2621 /* Force a panic(). This stinks, but it's better than letting
2622 * things continue without having a proper hard readonly
2624 if (!ocfs2_mount_local(OCFS2_SB(sb
)))
2625 OCFS2_SB(sb
)->s_mount_opt
|= OCFS2_MOUNT_ERRORS_PANIC
;
2626 ocfs2_handle_error(sb
);
2630 * Void signal blockers, because in-kernel sigprocmask() only fails
2631 * when SIG_* is wrong.
2633 void ocfs2_block_signals(sigset_t
*oldset
)
2638 sigfillset(&blocked
);
2639 rc
= sigprocmask(SIG_BLOCK
, &blocked
, oldset
);
2643 void ocfs2_unblock_signals(sigset_t
*oldset
)
2645 int rc
= sigprocmask(SIG_SETMASK
, oldset
, NULL
);
2649 module_init(ocfs2_init
);
2650 module_exit(ocfs2_exit
);