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
,
194 Opt_journal_async_commit
,
199 static const match_table_t tokens
= {
200 {Opt_barrier
, "barrier=%u"},
201 {Opt_err_panic
, "errors=panic"},
202 {Opt_err_ro
, "errors=remount-ro"},
204 {Opt_nointr
, "nointr"},
205 {Opt_hb_none
, OCFS2_HB_NONE
},
206 {Opt_hb_local
, OCFS2_HB_LOCAL
},
207 {Opt_hb_global
, OCFS2_HB_GLOBAL
},
208 {Opt_data_ordered
, "data=ordered"},
209 {Opt_data_writeback
, "data=writeback"},
210 {Opt_atime_quantum
, "atime_quantum=%u"},
211 {Opt_slot
, "preferred_slot=%u"},
212 {Opt_commit
, "commit=%u"},
213 {Opt_localalloc
, "localalloc=%d"},
214 {Opt_localflocks
, "localflocks"},
215 {Opt_stack
, "cluster_stack=%s"},
216 {Opt_user_xattr
, "user_xattr"},
217 {Opt_nouser_xattr
, "nouser_xattr"},
218 {Opt_inode64
, "inode64"},
220 {Opt_noacl
, "noacl"},
221 {Opt_usrquota
, "usrquota"},
222 {Opt_grpquota
, "grpquota"},
223 {Opt_coherency_buffered
, "coherency=buffered"},
224 {Opt_coherency_full
, "coherency=full"},
225 {Opt_resv_level
, "resv_level=%u"},
226 {Opt_dir_resv_level
, "dir_resv_level=%u"},
227 {Opt_journal_async_commit
, "journal_async_commit"},
228 {Opt_err_cont
, "errors=continue"},
232 #ifdef CONFIG_DEBUG_FS
233 static int ocfs2_osb_dump(struct ocfs2_super
*osb
, char *buf
, int len
)
235 struct ocfs2_cluster_connection
*cconn
= osb
->cconn
;
236 struct ocfs2_recovery_map
*rm
= osb
->recovery_map
;
237 struct ocfs2_orphan_scan
*os
= &osb
->osb_orphan_scan
;
240 out
+= snprintf(buf
+ out
, len
- out
,
241 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n",
242 "Device", osb
->dev_str
, osb
->uuid_str
,
243 osb
->fs_generation
, osb
->vol_label
);
245 out
+= snprintf(buf
+ out
, len
- out
,
246 "%10s => State: %d Flags: 0x%lX\n", "Volume",
247 atomic_read(&osb
->vol_state
), osb
->osb_flags
);
249 out
+= snprintf(buf
+ out
, len
- out
,
250 "%10s => Block: %lu Cluster: %d\n", "Sizes",
251 osb
->sb
->s_blocksize
, osb
->s_clustersize
);
253 out
+= snprintf(buf
+ out
, len
- out
,
254 "%10s => Compat: 0x%X Incompat: 0x%X "
256 "Features", osb
->s_feature_compat
,
257 osb
->s_feature_incompat
, osb
->s_feature_ro_compat
);
259 out
+= snprintf(buf
+ out
, len
- out
,
260 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount",
261 osb
->s_mount_opt
, osb
->s_atime_quantum
);
264 out
+= snprintf(buf
+ out
, len
- out
,
265 "%10s => Stack: %s Name: %*s "
266 "Version: %d.%d\n", "Cluster",
267 (*osb
->osb_cluster_stack
== '\0' ?
268 "o2cb" : osb
->osb_cluster_stack
),
269 cconn
->cc_namelen
, cconn
->cc_name
,
270 cconn
->cc_version
.pv_major
,
271 cconn
->cc_version
.pv_minor
);
274 spin_lock(&osb
->dc_task_lock
);
275 out
+= snprintf(buf
+ out
, len
- out
,
276 "%10s => Pid: %d Count: %lu WakeSeq: %lu "
277 "WorkSeq: %lu\n", "DownCnvt",
278 (osb
->dc_task
? task_pid_nr(osb
->dc_task
) : -1),
279 osb
->blocked_lock_count
, osb
->dc_wake_sequence
,
280 osb
->dc_work_sequence
);
281 spin_unlock(&osb
->dc_task_lock
);
283 spin_lock(&osb
->osb_lock
);
284 out
+= snprintf(buf
+ out
, len
- out
, "%10s => Pid: %d Nodes:",
286 (osb
->recovery_thread_task
?
287 task_pid_nr(osb
->recovery_thread_task
) : -1));
288 if (rm
->rm_used
== 0)
289 out
+= snprintf(buf
+ out
, len
- out
, " None\n");
291 for (i
= 0; i
< rm
->rm_used
; i
++)
292 out
+= snprintf(buf
+ out
, len
- out
, " %d",
294 out
+= snprintf(buf
+ out
, len
- out
, "\n");
296 spin_unlock(&osb
->osb_lock
);
298 out
+= snprintf(buf
+ out
, len
- out
,
299 "%10s => Pid: %d Interval: %lu\n", "Commit",
300 (osb
->commit_task
? task_pid_nr(osb
->commit_task
) : -1),
301 osb
->osb_commit_interval
);
303 out
+= snprintf(buf
+ out
, len
- out
,
304 "%10s => State: %d TxnId: %lu NumTxns: %d\n",
305 "Journal", osb
->journal
->j_state
,
306 osb
->journal
->j_trans_id
,
307 atomic_read(&osb
->journal
->j_num_trans
));
309 out
+= snprintf(buf
+ out
, len
- out
,
310 "%10s => GlobalAllocs: %d LocalAllocs: %d "
311 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
313 atomic_read(&osb
->alloc_stats
.bitmap_data
),
314 atomic_read(&osb
->alloc_stats
.local_data
),
315 atomic_read(&osb
->alloc_stats
.bg_allocs
),
316 atomic_read(&osb
->alloc_stats
.moves
),
317 atomic_read(&osb
->alloc_stats
.bg_extends
));
319 out
+= snprintf(buf
+ out
, len
- out
,
320 "%10s => State: %u Descriptor: %llu Size: %u bits "
321 "Default: %u bits\n",
322 "LocalAlloc", osb
->local_alloc_state
,
323 (unsigned long long)osb
->la_last_gd
,
324 osb
->local_alloc_bits
, osb
->local_alloc_default_bits
);
326 spin_lock(&osb
->osb_lock
);
327 out
+= snprintf(buf
+ out
, len
- out
,
328 "%10s => InodeSlot: %d StolenInodes: %d, "
329 "MetaSlot: %d StolenMeta: %d\n", "Steal",
330 osb
->s_inode_steal_slot
,
331 atomic_read(&osb
->s_num_inodes_stolen
),
332 osb
->s_meta_steal_slot
,
333 atomic_read(&osb
->s_num_meta_stolen
));
334 spin_unlock(&osb
->osb_lock
);
336 out
+= snprintf(buf
+ out
, len
- out
, "OrphanScan => ");
337 out
+= snprintf(buf
+ out
, len
- out
, "Local: %u Global: %u ",
338 os
->os_count
, os
->os_seqno
);
339 out
+= snprintf(buf
+ out
, len
- out
, " Last Scan: ");
340 if (atomic_read(&os
->os_state
) == ORPHAN_SCAN_INACTIVE
)
341 out
+= snprintf(buf
+ out
, len
- out
, "Disabled\n");
343 out
+= snprintf(buf
+ out
, len
- out
, "%lu seconds ago\n",
344 (get_seconds() - os
->os_scantime
.tv_sec
));
346 out
+= snprintf(buf
+ out
, len
- out
, "%10s => %3s %10s\n",
347 "Slots", "Num", "RecoGen");
348 for (i
= 0; i
< osb
->max_slots
; ++i
) {
349 out
+= snprintf(buf
+ out
, len
- out
,
350 "%10s %c %3d %10d\n",
352 (i
== osb
->slot_num
? '*' : ' '),
353 i
, osb
->slot_recovery_generations
[i
]);
359 static int ocfs2_osb_debug_open(struct inode
*inode
, struct file
*file
)
361 struct ocfs2_super
*osb
= inode
->i_private
;
364 buf
= kmalloc(PAGE_SIZE
, GFP_KERNEL
);
368 i_size_write(inode
, ocfs2_osb_dump(osb
, buf
, PAGE_SIZE
));
370 file
->private_data
= buf
;
377 static int ocfs2_debug_release(struct inode
*inode
, struct file
*file
)
379 kfree(file
->private_data
);
383 static ssize_t
ocfs2_debug_read(struct file
*file
, char __user
*buf
,
384 size_t nbytes
, loff_t
*ppos
)
386 return simple_read_from_buffer(buf
, nbytes
, ppos
, file
->private_data
,
387 i_size_read(file
->f_mapping
->host
));
390 static int ocfs2_osb_debug_open(struct inode
*inode
, struct file
*file
)
394 static int ocfs2_debug_release(struct inode
*inode
, struct file
*file
)
398 static ssize_t
ocfs2_debug_read(struct file
*file
, char __user
*buf
,
399 size_t nbytes
, loff_t
*ppos
)
403 #endif /* CONFIG_DEBUG_FS */
405 static const struct file_operations ocfs2_osb_debug_fops
= {
406 .open
= ocfs2_osb_debug_open
,
407 .release
= ocfs2_debug_release
,
408 .read
= ocfs2_debug_read
,
409 .llseek
= generic_file_llseek
,
412 static int ocfs2_sync_fs(struct super_block
*sb
, int wait
)
416 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
418 if (ocfs2_is_hard_readonly(osb
))
422 status
= ocfs2_flush_truncate_log(osb
);
426 ocfs2_schedule_truncate_log_flush(osb
, 0);
429 if (jbd2_journal_start_commit(OCFS2_SB(sb
)->journal
->j_journal
,
432 jbd2_log_wait_commit(OCFS2_SB(sb
)->journal
->j_journal
,
438 static int ocfs2_need_system_inode(struct ocfs2_super
*osb
, int ino
)
440 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb
->sb
, OCFS2_FEATURE_RO_COMPAT_USRQUOTA
)
441 && (ino
== USER_QUOTA_SYSTEM_INODE
442 || ino
== LOCAL_USER_QUOTA_SYSTEM_INODE
))
444 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb
->sb
, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
)
445 && (ino
== GROUP_QUOTA_SYSTEM_INODE
446 || ino
== LOCAL_GROUP_QUOTA_SYSTEM_INODE
))
451 static int ocfs2_init_global_system_inodes(struct ocfs2_super
*osb
)
453 struct inode
*new = NULL
;
457 new = ocfs2_iget(osb
, osb
->root_blkno
, OCFS2_FI_FLAG_SYSFILE
, 0);
459 status
= PTR_ERR(new);
463 osb
->root_inode
= new;
465 new = ocfs2_iget(osb
, osb
->system_dir_blkno
, OCFS2_FI_FLAG_SYSFILE
, 0);
467 status
= PTR_ERR(new);
471 osb
->sys_root_inode
= new;
473 for (i
= OCFS2_FIRST_ONLINE_SYSTEM_INODE
;
474 i
<= OCFS2_LAST_GLOBAL_SYSTEM_INODE
; i
++) {
475 if (!ocfs2_need_system_inode(osb
, i
))
477 new = ocfs2_get_system_file_inode(osb
, i
, osb
->slot_num
);
479 ocfs2_release_system_inodes(osb
);
482 /* FIXME: Should ERROR_RO_FS */
483 mlog(ML_ERROR
, "Unable to load system inode %d, "
484 "possibly corrupt fs?", i
);
487 // the array now has one ref, so drop this one
497 static int ocfs2_init_local_system_inodes(struct ocfs2_super
*osb
)
499 struct inode
*new = NULL
;
503 for (i
= OCFS2_LAST_GLOBAL_SYSTEM_INODE
+ 1;
504 i
< NUM_SYSTEM_INODES
;
506 if (!ocfs2_need_system_inode(osb
, i
))
508 new = ocfs2_get_system_file_inode(osb
, i
, osb
->slot_num
);
510 ocfs2_release_system_inodes(osb
);
512 mlog(ML_ERROR
, "status=%d, sysfile=%d, slot=%d\n",
513 status
, i
, osb
->slot_num
);
516 /* the array now has one ref, so drop this one */
526 static void ocfs2_release_system_inodes(struct ocfs2_super
*osb
)
531 for (i
= 0; i
< NUM_GLOBAL_SYSTEM_INODES
; i
++) {
532 inode
= osb
->global_system_inodes
[i
];
535 osb
->global_system_inodes
[i
] = NULL
;
539 inode
= osb
->sys_root_inode
;
542 osb
->sys_root_inode
= NULL
;
545 inode
= osb
->root_inode
;
548 osb
->root_inode
= NULL
;
551 if (!osb
->local_system_inodes
)
554 for (i
= 0; i
< NUM_LOCAL_SYSTEM_INODES
* osb
->max_slots
; i
++) {
555 if (osb
->local_system_inodes
[i
]) {
556 iput(osb
->local_system_inodes
[i
]);
557 osb
->local_system_inodes
[i
] = NULL
;
561 kfree(osb
->local_system_inodes
);
562 osb
->local_system_inodes
= NULL
;
565 /* We're allocating fs objects, use GFP_NOFS */
566 static struct inode
*ocfs2_alloc_inode(struct super_block
*sb
)
568 struct ocfs2_inode_info
*oi
;
570 oi
= kmem_cache_alloc(ocfs2_inode_cachep
, GFP_NOFS
);
575 oi
->i_datasync_tid
= 0;
576 memset(&oi
->i_dquot
, 0, sizeof(oi
->i_dquot
));
578 jbd2_journal_init_jbd_inode(&oi
->ip_jinode
, &oi
->vfs_inode
);
579 return &oi
->vfs_inode
;
582 static void ocfs2_i_callback(struct rcu_head
*head
)
584 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
585 kmem_cache_free(ocfs2_inode_cachep
, OCFS2_I(inode
));
588 static void ocfs2_destroy_inode(struct inode
*inode
)
590 call_rcu(&inode
->i_rcu
, ocfs2_i_callback
);
593 static unsigned long long ocfs2_max_file_offset(unsigned int bbits
,
596 unsigned int bytes
= 1 << cbits
;
597 unsigned int trim
= bytes
;
598 unsigned int bitshift
= 32;
601 * i_size and all block offsets in ocfs2 are always 64 bits
602 * wide. i_clusters is 32 bits, in cluster-sized units. So on
603 * 64 bit platforms, cluster size will be the limiting factor.
606 #if BITS_PER_LONG == 32
607 # if defined(CONFIG_LBDAF)
608 BUILD_BUG_ON(sizeof(sector_t
) != 8);
610 * We might be limited by page cache size.
612 if (bytes
> PAGE_CACHE_SIZE
) {
613 bytes
= PAGE_CACHE_SIZE
;
616 * Shift by 31 here so that we don't get larger than
623 * We are limited by the size of sector_t. Use block size, as
624 * that's what we expose to the VFS.
633 * Trim by a whole cluster when we can actually approach the
634 * on-disk limits. Otherwise we can overflow i_clusters when
635 * an extent start is at the max offset.
637 return (((unsigned long long)bytes
) << bitshift
) - trim
;
640 static int ocfs2_remount(struct super_block
*sb
, int *flags
, char *data
)
642 int incompat_features
;
644 struct mount_options parsed_options
;
645 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
650 if (!ocfs2_parse_options(sb
, data
, &parsed_options
, 1) ||
651 !ocfs2_check_set_options(sb
, &parsed_options
)) {
656 tmp
= OCFS2_MOUNT_HB_LOCAL
| OCFS2_MOUNT_HB_GLOBAL
|
658 if ((osb
->s_mount_opt
& tmp
) != (parsed_options
.mount_opt
& tmp
)) {
660 mlog(ML_ERROR
, "Cannot change heartbeat mode on remount\n");
664 if ((osb
->s_mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
) !=
665 (parsed_options
.mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
)) {
667 mlog(ML_ERROR
, "Cannot change data mode on remount\n");
671 /* Probably don't want this on remount; it might
672 * mess with other nodes */
673 if (!(osb
->s_mount_opt
& OCFS2_MOUNT_INODE64
) &&
674 (parsed_options
.mount_opt
& OCFS2_MOUNT_INODE64
)) {
676 mlog(ML_ERROR
, "Cannot enable inode64 on remount\n");
680 /* We're going to/from readonly mode. */
681 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
)) {
682 /* Disable quota accounting before remounting RO */
683 if (*flags
& MS_RDONLY
) {
684 ret
= ocfs2_susp_quotas(osb
, 0);
688 /* Lock here so the check of HARD_RO and the potential
689 * setting of SOFT_RO is atomic. */
690 spin_lock(&osb
->osb_lock
);
691 if (osb
->osb_flags
& OCFS2_OSB_HARD_RO
) {
692 mlog(ML_ERROR
, "Remount on readonly device is forbidden.\n");
697 if (*flags
& MS_RDONLY
) {
698 sb
->s_flags
|= MS_RDONLY
;
699 osb
->osb_flags
|= OCFS2_OSB_SOFT_RO
;
701 if (osb
->osb_flags
& OCFS2_OSB_ERROR_FS
) {
702 mlog(ML_ERROR
, "Cannot remount RDWR "
703 "filesystem due to previous errors.\n");
707 incompat_features
= OCFS2_HAS_RO_COMPAT_FEATURE(sb
, ~OCFS2_FEATURE_RO_COMPAT_SUPP
);
708 if (incompat_features
) {
709 mlog(ML_ERROR
, "Cannot remount RDWR because "
710 "of unsupported optional features "
711 "(%x).\n", incompat_features
);
715 sb
->s_flags
&= ~MS_RDONLY
;
716 osb
->osb_flags
&= ~OCFS2_OSB_SOFT_RO
;
718 trace_ocfs2_remount(sb
->s_flags
, osb
->osb_flags
, *flags
);
720 spin_unlock(&osb
->osb_lock
);
721 /* Enable quota accounting after remounting RW */
722 if (!ret
&& !(*flags
& MS_RDONLY
)) {
723 if (sb_any_quota_suspended(sb
))
724 ret
= ocfs2_susp_quotas(osb
, 1);
726 ret
= ocfs2_enable_quotas(osb
);
728 /* Return back changes... */
729 spin_lock(&osb
->osb_lock
);
730 sb
->s_flags
|= MS_RDONLY
;
731 osb
->osb_flags
|= OCFS2_OSB_SOFT_RO
;
732 spin_unlock(&osb
->osb_lock
);
739 /* Only save off the new mount options in case of a successful
741 osb
->s_mount_opt
= parsed_options
.mount_opt
;
742 osb
->s_atime_quantum
= parsed_options
.atime_quantum
;
743 osb
->preferred_slot
= parsed_options
.slot
;
744 if (parsed_options
.commit_interval
)
745 osb
->osb_commit_interval
= parsed_options
.commit_interval
;
747 if (!ocfs2_is_hard_readonly(osb
))
748 ocfs2_set_journal_params(osb
);
750 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
751 ((osb
->s_mount_opt
& OCFS2_MOUNT_POSIX_ACL
) ?
758 static int ocfs2_sb_probe(struct super_block
*sb
,
759 struct buffer_head
**bh
,
761 struct ocfs2_blockcheck_stats
*stats
)
764 struct ocfs1_vol_disk_hdr
*hdr
;
765 struct ocfs2_dinode
*di
;
771 *sector_size
= bdev_logical_block_size(sb
->s_bdev
);
772 if (*sector_size
> OCFS2_MAX_BLOCKSIZE
) {
773 mlog(ML_ERROR
, "Hardware sector size too large: %d (max=%d)\n",
774 *sector_size
, OCFS2_MAX_BLOCKSIZE
);
779 /* Can this really happen? */
780 if (*sector_size
< OCFS2_MIN_BLOCKSIZE
)
781 *sector_size
= OCFS2_MIN_BLOCKSIZE
;
783 /* check block zero for old format */
784 status
= ocfs2_get_sector(sb
, bh
, 0, *sector_size
);
789 hdr
= (struct ocfs1_vol_disk_hdr
*) (*bh
)->b_data
;
790 if (hdr
->major_version
== OCFS1_MAJOR_VERSION
) {
791 mlog(ML_ERROR
, "incompatible version: %u.%u\n",
792 hdr
->major_version
, hdr
->minor_version
);
795 if (memcmp(hdr
->signature
, OCFS1_VOLUME_SIGNATURE
,
796 strlen(OCFS1_VOLUME_SIGNATURE
)) == 0) {
797 mlog(ML_ERROR
, "incompatible volume signature: %8s\n",
804 mlog(ML_ERROR
, "This is an ocfs v1 filesystem which must be "
805 "upgraded before mounting with ocfs v2\n");
810 * Now check at magic offset for 512, 1024, 2048, 4096
811 * blocksizes. 4096 is the maximum blocksize because it is
812 * the minimum clustersize.
815 for (blksize
= *sector_size
;
816 blksize
<= OCFS2_MAX_BLOCKSIZE
;
818 tmpstat
= ocfs2_get_sector(sb
, bh
,
819 OCFS2_SUPER_BLOCK_BLKNO
,
826 di
= (struct ocfs2_dinode
*) (*bh
)->b_data
;
827 memset(stats
, 0, sizeof(struct ocfs2_blockcheck_stats
));
828 spin_lock_init(&stats
->b_lock
);
829 tmpstat
= ocfs2_verify_volume(di
, *bh
, blksize
, stats
);
834 if (tmpstat
!= -EAGAIN
) {
844 static int ocfs2_verify_heartbeat(struct ocfs2_super
*osb
)
846 u32 hb_enabled
= OCFS2_MOUNT_HB_LOCAL
| OCFS2_MOUNT_HB_GLOBAL
;
848 if (osb
->s_mount_opt
& hb_enabled
) {
849 if (ocfs2_mount_local(osb
)) {
850 mlog(ML_ERROR
, "Cannot heartbeat on a locally "
851 "mounted device.\n");
854 if (ocfs2_userspace_stack(osb
)) {
855 mlog(ML_ERROR
, "Userspace stack expected, but "
856 "o2cb heartbeat arguments passed to mount\n");
859 if (((osb
->s_mount_opt
& OCFS2_MOUNT_HB_GLOBAL
) &&
860 !ocfs2_cluster_o2cb_global_heartbeat(osb
)) ||
861 ((osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) &&
862 ocfs2_cluster_o2cb_global_heartbeat(osb
))) {
863 mlog(ML_ERROR
, "Mismatching o2cb heartbeat modes\n");
868 if (!(osb
->s_mount_opt
& hb_enabled
)) {
869 if (!ocfs2_mount_local(osb
) && !ocfs2_is_hard_readonly(osb
) &&
870 !ocfs2_userspace_stack(osb
)) {
871 mlog(ML_ERROR
, "Heartbeat has to be started to mount "
872 "a read-write clustered device.\n");
881 * If we're using a userspace stack, mount should have passed
882 * a name that matches the disk. If not, mount should not
883 * have passed a stack.
885 static int ocfs2_verify_userspace_stack(struct ocfs2_super
*osb
,
886 struct mount_options
*mopt
)
888 if (!ocfs2_userspace_stack(osb
) && mopt
->cluster_stack
[0]) {
890 "cluster stack passed to mount, but this filesystem "
891 "does not support it\n");
895 if (ocfs2_userspace_stack(osb
) &&
896 strncmp(osb
->osb_cluster_stack
, mopt
->cluster_stack
,
897 OCFS2_STACK_LABEL_LEN
)) {
899 "cluster stack passed to mount (\"%s\") does not "
900 "match the filesystem (\"%s\")\n",
902 osb
->osb_cluster_stack
);
909 static int ocfs2_susp_quotas(struct ocfs2_super
*osb
, int unsuspend
)
912 struct super_block
*sb
= osb
->sb
;
913 unsigned int feature
[OCFS2_MAXQUOTAS
] = {
914 OCFS2_FEATURE_RO_COMPAT_USRQUOTA
,
915 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
};
918 for (type
= 0; type
< OCFS2_MAXQUOTAS
; type
++) {
919 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb
, feature
[type
]))
922 status
= dquot_resume(sb
, type
);
924 struct ocfs2_mem_dqinfo
*oinfo
;
926 /* Cancel periodic syncing before suspending */
927 oinfo
= sb_dqinfo(sb
, type
)->dqi_priv
;
928 cancel_delayed_work_sync(&oinfo
->dqi_sync_work
);
929 status
= dquot_suspend(sb
, type
);
935 mlog(ML_ERROR
, "Failed to suspend/unsuspend quotas on "
936 "remount (error = %d).\n", status
);
940 static int ocfs2_enable_quotas(struct ocfs2_super
*osb
)
942 struct inode
*inode
[OCFS2_MAXQUOTAS
] = { NULL
, NULL
};
943 struct super_block
*sb
= osb
->sb
;
944 unsigned int feature
[OCFS2_MAXQUOTAS
] = {
945 OCFS2_FEATURE_RO_COMPAT_USRQUOTA
,
946 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
};
947 unsigned int ino
[OCFS2_MAXQUOTAS
] = {
948 LOCAL_USER_QUOTA_SYSTEM_INODE
,
949 LOCAL_GROUP_QUOTA_SYSTEM_INODE
};
953 sb_dqopt(sb
)->flags
|= DQUOT_QUOTA_SYS_FILE
| DQUOT_NEGATIVE_USAGE
;
954 for (type
= 0; type
< OCFS2_MAXQUOTAS
; type
++) {
955 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb
, feature
[type
]))
957 inode
[type
] = ocfs2_get_system_file_inode(osb
, ino
[type
],
963 status
= dquot_enable(inode
[type
], type
, QFMT_OCFS2
,
964 DQUOT_USAGE_ENABLED
);
969 for (type
= 0; type
< OCFS2_MAXQUOTAS
; type
++)
973 ocfs2_disable_quotas(osb
);
974 for (type
= 0; type
< OCFS2_MAXQUOTAS
; type
++)
980 static void ocfs2_disable_quotas(struct ocfs2_super
*osb
)
984 struct super_block
*sb
= osb
->sb
;
985 struct ocfs2_mem_dqinfo
*oinfo
;
987 /* We mostly ignore errors in this function because there's not much
988 * we can do when we see them */
989 for (type
= 0; type
< OCFS2_MAXQUOTAS
; type
++) {
990 if (!sb_has_quota_loaded(sb
, type
))
992 /* Cancel periodic syncing before we grab dqonoff_mutex */
993 oinfo
= sb_dqinfo(sb
, type
)->dqi_priv
;
994 cancel_delayed_work_sync(&oinfo
->dqi_sync_work
);
995 inode
= igrab(sb
->s_dquot
.files
[type
]);
996 /* Turn off quotas. This will remove all dquot structures from
997 * memory and so they will be automatically synced to global
999 dquot_disable(sb
, type
, DQUOT_USAGE_ENABLED
|
1000 DQUOT_LIMITS_ENABLED
);
1007 static int ocfs2_fill_super(struct super_block
*sb
, void *data
, int silent
)
1009 struct dentry
*root
;
1010 int status
, sector_size
;
1011 struct mount_options parsed_options
;
1012 struct inode
*inode
= NULL
;
1013 struct ocfs2_super
*osb
= NULL
;
1014 struct buffer_head
*bh
= NULL
;
1016 struct ocfs2_blockcheck_stats stats
;
1018 trace_ocfs2_fill_super(sb
, data
, silent
);
1020 if (!ocfs2_parse_options(sb
, data
, &parsed_options
, 0)) {
1022 goto read_super_error
;
1025 /* probe for superblock */
1026 status
= ocfs2_sb_probe(sb
, &bh
, §or_size
, &stats
);
1028 mlog(ML_ERROR
, "superblock probe failed!\n");
1029 goto read_super_error
;
1032 status
= ocfs2_initialize_super(sb
, bh
, sector_size
, &stats
);
1036 goto read_super_error
;
1041 if (!ocfs2_check_set_options(sb
, &parsed_options
)) {
1043 goto read_super_error
;
1045 osb
->s_mount_opt
= parsed_options
.mount_opt
;
1046 osb
->s_atime_quantum
= parsed_options
.atime_quantum
;
1047 osb
->preferred_slot
= parsed_options
.slot
;
1048 osb
->osb_commit_interval
= parsed_options
.commit_interval
;
1050 ocfs2_la_set_sizes(osb
, parsed_options
.localalloc_opt
);
1051 osb
->osb_resv_level
= parsed_options
.resv_level
;
1052 osb
->osb_dir_resv_level
= parsed_options
.resv_level
;
1053 if (parsed_options
.dir_resv_level
== -1)
1054 osb
->osb_dir_resv_level
= parsed_options
.resv_level
;
1056 osb
->osb_dir_resv_level
= parsed_options
.dir_resv_level
;
1058 status
= ocfs2_verify_userspace_stack(osb
, &parsed_options
);
1060 goto read_super_error
;
1062 sb
->s_magic
= OCFS2_SUPER_MAGIC
;
1064 sb
->s_flags
= (sb
->s_flags
& ~(MS_POSIXACL
| MS_NOSEC
)) |
1065 ((osb
->s_mount_opt
& OCFS2_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
1067 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1069 if (bdev_read_only(sb
->s_bdev
)) {
1070 if (!(sb
->s_flags
& MS_RDONLY
)) {
1072 mlog(ML_ERROR
, "Readonly device detected but readonly "
1073 "mount was not specified.\n");
1074 goto read_super_error
;
1077 /* You should not be able to start a local heartbeat
1078 * on a readonly device. */
1079 if (osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) {
1081 mlog(ML_ERROR
, "Local heartbeat specified on readonly "
1083 goto read_super_error
;
1086 status
= ocfs2_check_journals_nolocks(osb
);
1088 if (status
== -EROFS
)
1089 mlog(ML_ERROR
, "Recovery required on readonly "
1090 "file system, but write access is "
1094 goto read_super_error
;
1097 ocfs2_set_ro_flag(osb
, 1);
1099 printk(KERN_NOTICE
"ocfs2: Readonly device (%s) detected. "
1100 "Cluster services will not be used for this mount. "
1101 "Recovery will be skipped.\n", osb
->dev_str
);
1104 if (!ocfs2_is_hard_readonly(osb
)) {
1105 if (sb
->s_flags
& MS_RDONLY
)
1106 ocfs2_set_ro_flag(osb
, 0);
1109 status
= ocfs2_verify_heartbeat(osb
);
1112 goto read_super_error
;
1115 osb
->osb_debug_root
= debugfs_create_dir(osb
->uuid_str
,
1116 ocfs2_debugfs_root
);
1117 if (!osb
->osb_debug_root
) {
1119 mlog(ML_ERROR
, "Unable to create per-mount debugfs root.\n");
1120 goto read_super_error
;
1123 osb
->osb_ctxt
= debugfs_create_file("fs_state", S_IFREG
|S_IRUSR
,
1124 osb
->osb_debug_root
,
1126 &ocfs2_osb_debug_fops
);
1127 if (!osb
->osb_ctxt
) {
1130 goto read_super_error
;
1133 if (ocfs2_meta_ecc(osb
)) {
1134 status
= ocfs2_blockcheck_stats_debugfs_install(
1135 &osb
->osb_ecc_stats
,
1136 osb
->osb_debug_root
);
1139 "Unable to create blockcheck statistics "
1141 goto read_super_error
;
1145 status
= ocfs2_mount_volume(sb
);
1147 goto read_super_error
;
1149 if (osb
->root_inode
)
1150 inode
= igrab(osb
->root_inode
);
1155 goto read_super_error
;
1158 root
= d_make_root(inode
);
1162 goto read_super_error
;
1167 ocfs2_complete_mount_recovery(osb
);
1169 if (ocfs2_mount_local(osb
))
1170 snprintf(nodestr
, sizeof(nodestr
), "local");
1172 snprintf(nodestr
, sizeof(nodestr
), "%u", osb
->node_num
);
1174 printk(KERN_INFO
"ocfs2: Mounting device (%s) on (node %s, slot %d) "
1175 "with %s data mode.\n",
1176 osb
->dev_str
, nodestr
, osb
->slot_num
,
1177 osb
->s_mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
? "writeback" :
1180 atomic_set(&osb
->vol_state
, VOLUME_MOUNTED
);
1181 wake_up(&osb
->osb_mount_event
);
1183 /* Now we can initialize quotas because we can afford to wait
1184 * for cluster locks recovery now. That also means that truncation
1185 * log recovery can happen but that waits for proper quota setup */
1186 if (!(sb
->s_flags
& MS_RDONLY
)) {
1187 status
= ocfs2_enable_quotas(osb
);
1189 /* We have to err-out specially here because
1190 * s_root is already set */
1192 atomic_set(&osb
->vol_state
, VOLUME_DISABLED
);
1193 wake_up(&osb
->osb_mount_event
);
1198 ocfs2_complete_quota_recovery(osb
);
1200 /* Now we wake up again for processes waiting for quotas */
1201 atomic_set(&osb
->vol_state
, VOLUME_MOUNTED_QUOTAS
);
1202 wake_up(&osb
->osb_mount_event
);
1204 /* Start this when the mount is almost sure of being successful */
1205 ocfs2_orphan_scan_start(osb
);
1213 atomic_set(&osb
->vol_state
, VOLUME_DISABLED
);
1214 wake_up(&osb
->osb_mount_event
);
1215 ocfs2_dismount_volume(sb
, 1);
1223 static struct dentry
*ocfs2_mount(struct file_system_type
*fs_type
,
1225 const char *dev_name
,
1228 return mount_bdev(fs_type
, flags
, dev_name
, data
, ocfs2_fill_super
);
1231 static struct file_system_type ocfs2_fs_type
= {
1232 .owner
= THIS_MODULE
,
1234 .mount
= ocfs2_mount
,
1235 .kill_sb
= kill_block_super
,
1236 .fs_flags
= FS_REQUIRES_DEV
|FS_RENAME_DOES_D_MOVE
,
1239 MODULE_ALIAS_FS("ocfs2");
1241 static int ocfs2_check_set_options(struct super_block
*sb
,
1242 struct mount_options
*options
)
1244 if (options
->mount_opt
& OCFS2_MOUNT_USRQUOTA
&&
1245 !OCFS2_HAS_RO_COMPAT_FEATURE(sb
,
1246 OCFS2_FEATURE_RO_COMPAT_USRQUOTA
)) {
1247 mlog(ML_ERROR
, "User quotas were requested, but this "
1248 "filesystem does not have the feature enabled.\n");
1251 if (options
->mount_opt
& OCFS2_MOUNT_GRPQUOTA
&&
1252 !OCFS2_HAS_RO_COMPAT_FEATURE(sb
,
1253 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
)) {
1254 mlog(ML_ERROR
, "Group quotas were requested, but this "
1255 "filesystem does not have the feature enabled.\n");
1258 if (options
->mount_opt
& OCFS2_MOUNT_POSIX_ACL
&&
1259 !OCFS2_HAS_INCOMPAT_FEATURE(sb
, OCFS2_FEATURE_INCOMPAT_XATTR
)) {
1260 mlog(ML_ERROR
, "ACL support requested but extended attributes "
1261 "feature is not enabled\n");
1264 /* No ACL setting specified? Use XATTR feature... */
1265 if (!(options
->mount_opt
& (OCFS2_MOUNT_POSIX_ACL
|
1266 OCFS2_MOUNT_NO_POSIX_ACL
))) {
1267 if (OCFS2_HAS_INCOMPAT_FEATURE(sb
, OCFS2_FEATURE_INCOMPAT_XATTR
))
1268 options
->mount_opt
|= OCFS2_MOUNT_POSIX_ACL
;
1270 options
->mount_opt
|= OCFS2_MOUNT_NO_POSIX_ACL
;
1275 static int ocfs2_parse_options(struct super_block
*sb
,
1277 struct mount_options
*mopt
,
1280 int status
, user_stack
= 0;
1284 substring_t args
[MAX_OPT_ARGS
];
1286 trace_ocfs2_parse_options(is_remount
, options
? options
: "(none)");
1288 mopt
->commit_interval
= 0;
1289 mopt
->mount_opt
= OCFS2_MOUNT_NOINTR
;
1290 mopt
->atime_quantum
= OCFS2_DEFAULT_ATIME_QUANTUM
;
1291 mopt
->slot
= OCFS2_INVALID_SLOT
;
1292 mopt
->localalloc_opt
= -1;
1293 mopt
->cluster_stack
[0] = '\0';
1294 mopt
->resv_level
= OCFS2_DEFAULT_RESV_LEVEL
;
1295 mopt
->dir_resv_level
= -1;
1302 while ((p
= strsep(&options
, ",")) != NULL
) {
1306 token
= match_token(p
, tokens
, args
);
1309 mopt
->mount_opt
|= OCFS2_MOUNT_HB_LOCAL
;
1312 mopt
->mount_opt
|= OCFS2_MOUNT_HB_NONE
;
1315 mopt
->mount_opt
|= OCFS2_MOUNT_HB_GLOBAL
;
1318 if (match_int(&args
[0], &option
)) {
1323 mopt
->mount_opt
|= OCFS2_MOUNT_BARRIER
;
1325 mopt
->mount_opt
&= ~OCFS2_MOUNT_BARRIER
;
1328 mopt
->mount_opt
&= ~OCFS2_MOUNT_NOINTR
;
1331 mopt
->mount_opt
|= OCFS2_MOUNT_NOINTR
;
1334 mopt
->mount_opt
&= ~OCFS2_MOUNT_ERRORS_CONT
;
1335 mopt
->mount_opt
&= ~OCFS2_MOUNT_ERRORS_ROFS
;
1336 mopt
->mount_opt
|= OCFS2_MOUNT_ERRORS_PANIC
;
1339 mopt
->mount_opt
&= ~OCFS2_MOUNT_ERRORS_CONT
;
1340 mopt
->mount_opt
&= ~OCFS2_MOUNT_ERRORS_PANIC
;
1341 mopt
->mount_opt
|= OCFS2_MOUNT_ERRORS_ROFS
;
1344 mopt
->mount_opt
&= ~OCFS2_MOUNT_ERRORS_ROFS
;
1345 mopt
->mount_opt
&= ~OCFS2_MOUNT_ERRORS_PANIC
;
1346 mopt
->mount_opt
|= OCFS2_MOUNT_ERRORS_CONT
;
1348 case Opt_data_ordered
:
1349 mopt
->mount_opt
&= ~OCFS2_MOUNT_DATA_WRITEBACK
;
1351 case Opt_data_writeback
:
1352 mopt
->mount_opt
|= OCFS2_MOUNT_DATA_WRITEBACK
;
1354 case Opt_user_xattr
:
1355 mopt
->mount_opt
&= ~OCFS2_MOUNT_NOUSERXATTR
;
1357 case Opt_nouser_xattr
:
1358 mopt
->mount_opt
|= OCFS2_MOUNT_NOUSERXATTR
;
1360 case Opt_atime_quantum
:
1361 if (match_int(&args
[0], &option
)) {
1366 mopt
->atime_quantum
= option
;
1369 if (match_int(&args
[0], &option
)) {
1374 mopt
->slot
= (s16
)option
;
1377 if (match_int(&args
[0], &option
)) {
1384 option
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1385 mopt
->commit_interval
= HZ
* option
;
1387 case Opt_localalloc
:
1388 if (match_int(&args
[0], &option
)) {
1393 mopt
->localalloc_opt
= option
;
1395 case Opt_localflocks
:
1397 * Changing this during remount could race
1398 * flock() requests, or "unbalance" existing
1399 * ones (e.g., a lock is taken in one mode but
1400 * dropped in the other). If users care enough
1401 * to flip locking modes during remount, we
1402 * could add a "local" flag to individual
1403 * flock structures for proper tracking of
1407 mopt
->mount_opt
|= OCFS2_MOUNT_LOCALFLOCKS
;
1410 /* Check both that the option we were passed
1411 * is of the right length and that it is a proper
1412 * string of the right length.
1414 if (((args
[0].to
- args
[0].from
) !=
1415 OCFS2_STACK_LABEL_LEN
) ||
1416 (strnlen(args
[0].from
,
1417 OCFS2_STACK_LABEL_LEN
) !=
1418 OCFS2_STACK_LABEL_LEN
)) {
1420 "Invalid cluster_stack option\n");
1424 memcpy(mopt
->cluster_stack
, args
[0].from
,
1425 OCFS2_STACK_LABEL_LEN
);
1426 mopt
->cluster_stack
[OCFS2_STACK_LABEL_LEN
] = '\0';
1428 * Open code the memcmp here as we don't have
1430 * ocfs2_userspace_stack().
1432 if (memcmp(mopt
->cluster_stack
,
1433 OCFS2_CLASSIC_CLUSTER_STACK
,
1434 OCFS2_STACK_LABEL_LEN
))
1438 mopt
->mount_opt
|= OCFS2_MOUNT_INODE64
;
1441 mopt
->mount_opt
|= OCFS2_MOUNT_USRQUOTA
;
1444 mopt
->mount_opt
|= OCFS2_MOUNT_GRPQUOTA
;
1446 case Opt_coherency_buffered
:
1447 mopt
->mount_opt
|= OCFS2_MOUNT_COHERENCY_BUFFERED
;
1449 case Opt_coherency_full
:
1450 mopt
->mount_opt
&= ~OCFS2_MOUNT_COHERENCY_BUFFERED
;
1453 mopt
->mount_opt
|= OCFS2_MOUNT_POSIX_ACL
;
1454 mopt
->mount_opt
&= ~OCFS2_MOUNT_NO_POSIX_ACL
;
1457 mopt
->mount_opt
|= OCFS2_MOUNT_NO_POSIX_ACL
;
1458 mopt
->mount_opt
&= ~OCFS2_MOUNT_POSIX_ACL
;
1460 case Opt_resv_level
:
1463 if (match_int(&args
[0], &option
)) {
1467 if (option
>= OCFS2_MIN_RESV_LEVEL
&&
1468 option
< OCFS2_MAX_RESV_LEVEL
)
1469 mopt
->resv_level
= option
;
1471 case Opt_dir_resv_level
:
1474 if (match_int(&args
[0], &option
)) {
1478 if (option
>= OCFS2_MIN_RESV_LEVEL
&&
1479 option
< OCFS2_MAX_RESV_LEVEL
)
1480 mopt
->dir_resv_level
= option
;
1482 case Opt_journal_async_commit
:
1483 mopt
->mount_opt
|= OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT
;
1487 "Unrecognized mount option \"%s\" "
1488 "or missing value\n", p
);
1494 if (user_stack
== 0) {
1495 /* Ensure only one heartbeat mode */
1496 tmp
= mopt
->mount_opt
& (OCFS2_MOUNT_HB_LOCAL
|
1497 OCFS2_MOUNT_HB_GLOBAL
|
1498 OCFS2_MOUNT_HB_NONE
);
1499 if (hweight32(tmp
) != 1) {
1500 mlog(ML_ERROR
, "Invalid heartbeat mount options\n");
1512 static int ocfs2_show_options(struct seq_file
*s
, struct dentry
*root
)
1514 struct ocfs2_super
*osb
= OCFS2_SB(root
->d_sb
);
1515 unsigned long opts
= osb
->s_mount_opt
;
1516 unsigned int local_alloc_megs
;
1518 if (opts
& (OCFS2_MOUNT_HB_LOCAL
| OCFS2_MOUNT_HB_GLOBAL
)) {
1519 seq_printf(s
, ",_netdev");
1520 if (opts
& OCFS2_MOUNT_HB_LOCAL
)
1521 seq_printf(s
, ",%s", OCFS2_HB_LOCAL
);
1523 seq_printf(s
, ",%s", OCFS2_HB_GLOBAL
);
1525 seq_printf(s
, ",%s", OCFS2_HB_NONE
);
1527 if (opts
& OCFS2_MOUNT_NOINTR
)
1528 seq_printf(s
, ",nointr");
1530 if (opts
& OCFS2_MOUNT_DATA_WRITEBACK
)
1531 seq_printf(s
, ",data=writeback");
1533 seq_printf(s
, ",data=ordered");
1535 if (opts
& OCFS2_MOUNT_BARRIER
)
1536 seq_printf(s
, ",barrier=1");
1538 if (opts
& OCFS2_MOUNT_ERRORS_PANIC
)
1539 seq_printf(s
, ",errors=panic");
1540 else if (opts
& OCFS2_MOUNT_ERRORS_CONT
)
1541 seq_printf(s
, ",errors=continue");
1543 seq_printf(s
, ",errors=remount-ro");
1545 if (osb
->preferred_slot
!= OCFS2_INVALID_SLOT
)
1546 seq_printf(s
, ",preferred_slot=%d", osb
->preferred_slot
);
1548 seq_printf(s
, ",atime_quantum=%u", osb
->s_atime_quantum
);
1550 if (osb
->osb_commit_interval
)
1551 seq_printf(s
, ",commit=%u",
1552 (unsigned) (osb
->osb_commit_interval
/ HZ
));
1554 local_alloc_megs
= osb
->local_alloc_bits
>> (20 - osb
->s_clustersize_bits
);
1555 if (local_alloc_megs
!= ocfs2_la_default_mb(osb
))
1556 seq_printf(s
, ",localalloc=%d", local_alloc_megs
);
1558 if (opts
& OCFS2_MOUNT_LOCALFLOCKS
)
1559 seq_printf(s
, ",localflocks,");
1561 if (osb
->osb_cluster_stack
[0])
1562 seq_show_option_n(s
, "cluster_stack", osb
->osb_cluster_stack
,
1563 OCFS2_STACK_LABEL_LEN
);
1564 if (opts
& OCFS2_MOUNT_USRQUOTA
)
1565 seq_printf(s
, ",usrquota");
1566 if (opts
& OCFS2_MOUNT_GRPQUOTA
)
1567 seq_printf(s
, ",grpquota");
1569 if (opts
& OCFS2_MOUNT_COHERENCY_BUFFERED
)
1570 seq_printf(s
, ",coherency=buffered");
1572 seq_printf(s
, ",coherency=full");
1574 if (opts
& OCFS2_MOUNT_NOUSERXATTR
)
1575 seq_printf(s
, ",nouser_xattr");
1577 seq_printf(s
, ",user_xattr");
1579 if (opts
& OCFS2_MOUNT_INODE64
)
1580 seq_printf(s
, ",inode64");
1582 if (opts
& OCFS2_MOUNT_POSIX_ACL
)
1583 seq_printf(s
, ",acl");
1585 seq_printf(s
, ",noacl");
1587 if (osb
->osb_resv_level
!= OCFS2_DEFAULT_RESV_LEVEL
)
1588 seq_printf(s
, ",resv_level=%d", osb
->osb_resv_level
);
1590 if (osb
->osb_dir_resv_level
!= osb
->osb_resv_level
)
1591 seq_printf(s
, ",dir_resv_level=%d", osb
->osb_resv_level
);
1593 if (opts
& OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT
)
1594 seq_printf(s
, ",journal_async_commit");
1599 static int __init
ocfs2_init(void)
1603 status
= init_ocfs2_uptodate_cache();
1607 status
= ocfs2_initialize_mem_caches();
1611 ocfs2_wq
= create_singlethread_workqueue("ocfs2_wq");
1617 ocfs2_debugfs_root
= debugfs_create_dir("ocfs2", NULL
);
1618 if (!ocfs2_debugfs_root
) {
1620 mlog(ML_ERROR
, "Unable to create ocfs2 debugfs root.\n");
1624 ocfs2_set_locking_protocol();
1626 status
= register_quota_format(&ocfs2_quota_format
);
1629 status
= register_filesystem(&ocfs2_fs_type
);
1633 unregister_quota_format(&ocfs2_quota_format
);
1635 destroy_workqueue(ocfs2_wq
);
1636 debugfs_remove(ocfs2_debugfs_root
);
1638 ocfs2_free_mem_caches();
1640 exit_ocfs2_uptodate_cache();
1646 static void __exit
ocfs2_exit(void)
1649 flush_workqueue(ocfs2_wq
);
1650 destroy_workqueue(ocfs2_wq
);
1653 unregister_quota_format(&ocfs2_quota_format
);
1655 debugfs_remove(ocfs2_debugfs_root
);
1657 ocfs2_free_mem_caches();
1659 unregister_filesystem(&ocfs2_fs_type
);
1661 exit_ocfs2_uptodate_cache();
1664 static void ocfs2_put_super(struct super_block
*sb
)
1666 trace_ocfs2_put_super(sb
);
1668 ocfs2_sync_blockdev(sb
);
1669 ocfs2_dismount_volume(sb
, 0);
1672 static int ocfs2_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
1674 struct ocfs2_super
*osb
;
1675 u32 numbits
, freebits
;
1677 struct ocfs2_dinode
*bm_lock
;
1678 struct buffer_head
*bh
= NULL
;
1679 struct inode
*inode
= NULL
;
1681 trace_ocfs2_statfs(dentry
->d_sb
, buf
);
1683 osb
= OCFS2_SB(dentry
->d_sb
);
1685 inode
= ocfs2_get_system_file_inode(osb
,
1686 GLOBAL_BITMAP_SYSTEM_INODE
,
1687 OCFS2_INVALID_SLOT
);
1689 mlog(ML_ERROR
, "failed to get bitmap inode\n");
1694 status
= ocfs2_inode_lock(inode
, &bh
, 0);
1700 bm_lock
= (struct ocfs2_dinode
*) bh
->b_data
;
1702 numbits
= le32_to_cpu(bm_lock
->id1
.bitmap1
.i_total
);
1703 freebits
= numbits
- le32_to_cpu(bm_lock
->id1
.bitmap1
.i_used
);
1705 buf
->f_type
= OCFS2_SUPER_MAGIC
;
1706 buf
->f_bsize
= dentry
->d_sb
->s_blocksize
;
1707 buf
->f_namelen
= OCFS2_MAX_FILENAME_LEN
;
1708 buf
->f_blocks
= ((sector_t
) numbits
) *
1709 (osb
->s_clustersize
>> osb
->sb
->s_blocksize_bits
);
1710 buf
->f_bfree
= ((sector_t
) freebits
) *
1711 (osb
->s_clustersize
>> osb
->sb
->s_blocksize_bits
);
1712 buf
->f_bavail
= buf
->f_bfree
;
1713 buf
->f_files
= numbits
;
1714 buf
->f_ffree
= freebits
;
1715 buf
->f_fsid
.val
[0] = crc32_le(0, osb
->uuid_str
, OCFS2_VOL_UUID_LEN
)
1717 buf
->f_fsid
.val
[1] = crc32_le(0, osb
->uuid_str
+ OCFS2_VOL_UUID_LEN
,
1718 OCFS2_VOL_UUID_LEN
) & 0xFFFFFFFFUL
;
1722 ocfs2_inode_unlock(inode
, 0);
1733 static void ocfs2_inode_init_once(void *data
)
1735 struct ocfs2_inode_info
*oi
= data
;
1738 oi
->ip_open_count
= 0;
1739 spin_lock_init(&oi
->ip_lock
);
1740 ocfs2_extent_map_init(&oi
->vfs_inode
);
1741 INIT_LIST_HEAD(&oi
->ip_io_markers
);
1742 oi
->ip_dir_start_lookup
= 0;
1743 mutex_init(&oi
->ip_unaligned_aio
);
1744 init_rwsem(&oi
->ip_alloc_sem
);
1745 init_rwsem(&oi
->ip_xattr_sem
);
1746 mutex_init(&oi
->ip_io_mutex
);
1748 oi
->ip_blkno
= 0ULL;
1749 oi
->ip_clusters
= 0;
1751 ocfs2_resv_init_once(&oi
->ip_la_data_resv
);
1753 ocfs2_lock_res_init_once(&oi
->ip_rw_lockres
);
1754 ocfs2_lock_res_init_once(&oi
->ip_inode_lockres
);
1755 ocfs2_lock_res_init_once(&oi
->ip_open_lockres
);
1757 ocfs2_metadata_cache_init(INODE_CACHE(&oi
->vfs_inode
),
1758 &ocfs2_inode_caching_ops
);
1760 inode_init_once(&oi
->vfs_inode
);
1763 static int ocfs2_initialize_mem_caches(void)
1765 ocfs2_inode_cachep
= kmem_cache_create("ocfs2_inode_cache",
1766 sizeof(struct ocfs2_inode_info
),
1768 (SLAB_HWCACHE_ALIGN
|SLAB_RECLAIM_ACCOUNT
|
1769 SLAB_MEM_SPREAD
|SLAB_ACCOUNT
),
1770 ocfs2_inode_init_once
);
1771 ocfs2_dquot_cachep
= kmem_cache_create("ocfs2_dquot_cache",
1772 sizeof(struct ocfs2_dquot
),
1774 (SLAB_HWCACHE_ALIGN
|SLAB_RECLAIM_ACCOUNT
|
1777 ocfs2_qf_chunk_cachep
= kmem_cache_create("ocfs2_qf_chunk_cache",
1778 sizeof(struct ocfs2_quota_chunk
),
1780 (SLAB_RECLAIM_ACCOUNT
|SLAB_MEM_SPREAD
),
1782 if (!ocfs2_inode_cachep
|| !ocfs2_dquot_cachep
||
1783 !ocfs2_qf_chunk_cachep
) {
1784 if (ocfs2_inode_cachep
)
1785 kmem_cache_destroy(ocfs2_inode_cachep
);
1786 if (ocfs2_dquot_cachep
)
1787 kmem_cache_destroy(ocfs2_dquot_cachep
);
1788 if (ocfs2_qf_chunk_cachep
)
1789 kmem_cache_destroy(ocfs2_qf_chunk_cachep
);
1796 static void ocfs2_free_mem_caches(void)
1799 * Make sure all delayed rcu free inodes are flushed before we
1803 if (ocfs2_inode_cachep
)
1804 kmem_cache_destroy(ocfs2_inode_cachep
);
1805 ocfs2_inode_cachep
= NULL
;
1807 if (ocfs2_dquot_cachep
)
1808 kmem_cache_destroy(ocfs2_dquot_cachep
);
1809 ocfs2_dquot_cachep
= NULL
;
1811 if (ocfs2_qf_chunk_cachep
)
1812 kmem_cache_destroy(ocfs2_qf_chunk_cachep
);
1813 ocfs2_qf_chunk_cachep
= NULL
;
1816 static int ocfs2_get_sector(struct super_block
*sb
,
1817 struct buffer_head
**bh
,
1821 if (!sb_set_blocksize(sb
, sect_size
)) {
1822 mlog(ML_ERROR
, "unable to set blocksize\n");
1826 *bh
= sb_getblk(sb
, block
);
1828 mlog_errno(-ENOMEM
);
1832 if (!buffer_dirty(*bh
))
1833 clear_buffer_uptodate(*bh
);
1835 ll_rw_block(READ
, 1, bh
);
1836 wait_on_buffer(*bh
);
1837 if (!buffer_uptodate(*bh
)) {
1847 static int ocfs2_mount_volume(struct super_block
*sb
)
1850 int unlock_super
= 0;
1851 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
1853 if (ocfs2_is_hard_readonly(osb
))
1856 status
= ocfs2_dlm_init(osb
);
1862 status
= ocfs2_super_lock(osb
, 1);
1869 /* This will load up the node map and add ourselves to it. */
1870 status
= ocfs2_find_slot(osb
);
1876 /* load all node-local system inodes */
1877 status
= ocfs2_init_local_system_inodes(osb
);
1883 status
= ocfs2_check_volume(osb
);
1889 status
= ocfs2_truncate_log_init(osb
);
1895 ocfs2_super_unlock(osb
, 1);
1900 static void ocfs2_dismount_volume(struct super_block
*sb
, int mnt_err
)
1902 int tmp
, hangup_needed
= 0;
1903 struct ocfs2_super
*osb
= NULL
;
1906 trace_ocfs2_dismount_volume(sb
);
1912 debugfs_remove(osb
->osb_ctxt
);
1914 /* Orphan scan should be stopped as early as possible */
1915 ocfs2_orphan_scan_stop(osb
);
1917 ocfs2_disable_quotas(osb
);
1919 /* All dquots should be freed by now */
1920 WARN_ON(!llist_empty(&osb
->dquot_drop_list
));
1921 /* Wait for worker to be done with the work structure in osb */
1922 cancel_work_sync(&osb
->dquot_drop_work
);
1924 ocfs2_shutdown_local_alloc(osb
);
1926 ocfs2_truncate_log_shutdown(osb
);
1928 /* This will disable recovery and flush any recovery work. */
1929 ocfs2_recovery_exit(osb
);
1931 ocfs2_journal_shutdown(osb
);
1933 ocfs2_sync_blockdev(sb
);
1935 ocfs2_purge_refcount_trees(osb
);
1937 /* No cluster connection means we've failed during mount, so skip
1938 * all the steps which depended on that to complete. */
1940 tmp
= ocfs2_super_lock(osb
, 1);
1947 if (osb
->slot_num
!= OCFS2_INVALID_SLOT
)
1948 ocfs2_put_slot(osb
);
1951 ocfs2_super_unlock(osb
, 1);
1953 ocfs2_release_system_inodes(osb
);
1956 * If we're dismounting due to mount error, mount.ocfs2 will clean
1957 * up heartbeat. If we're a local mount, there is no heartbeat.
1958 * If we failed before we got a uuid_str yet, we can't stop
1959 * heartbeat. Otherwise, do it.
1961 if (!mnt_err
&& !ocfs2_mount_local(osb
) && osb
->uuid_str
&&
1962 !ocfs2_is_hard_readonly(osb
))
1966 ocfs2_dlm_shutdown(osb
, hangup_needed
);
1968 ocfs2_blockcheck_stats_debugfs_remove(&osb
->osb_ecc_stats
);
1969 debugfs_remove(osb
->osb_debug_root
);
1972 ocfs2_cluster_hangup(osb
->uuid_str
, strlen(osb
->uuid_str
));
1974 atomic_set(&osb
->vol_state
, VOLUME_DISMOUNTED
);
1976 if (ocfs2_mount_local(osb
))
1977 snprintf(nodestr
, sizeof(nodestr
), "local");
1979 snprintf(nodestr
, sizeof(nodestr
), "%u", osb
->node_num
);
1981 printk(KERN_INFO
"ocfs2: Unmounting device (%s) on (node %s)\n",
1982 osb
->dev_str
, nodestr
);
1984 ocfs2_delete_osb(osb
);
1987 sb
->s_fs_info
= NULL
;
1990 static int ocfs2_setup_osb_uuid(struct ocfs2_super
*osb
, const unsigned char *uuid
,
1991 unsigned uuid_bytes
)
1996 BUG_ON(uuid_bytes
!= OCFS2_VOL_UUID_LEN
);
1998 osb
->uuid_str
= kzalloc(OCFS2_VOL_UUID_LEN
* 2 + 1, GFP_KERNEL
);
1999 if (osb
->uuid_str
== NULL
)
2002 for (i
= 0, ptr
= osb
->uuid_str
; i
< OCFS2_VOL_UUID_LEN
; i
++) {
2003 /* print with null */
2004 ret
= snprintf(ptr
, 3, "%02X", uuid
[i
]);
2005 if (ret
!= 2) /* drop super cleans up */
2007 /* then only advance past the last char */
2014 /* Make sure entire volume is addressable by our journal. Requires
2015 osb_clusters_at_boot to be valid and for the journal to have been
2016 initialized by ocfs2_journal_init(). */
2017 static int ocfs2_journal_addressable(struct ocfs2_super
*osb
)
2021 ocfs2_clusters_to_blocks(osb
->sb
,
2022 osb
->osb_clusters_at_boot
) - 1;
2024 /* 32-bit block number is always OK. */
2025 if (max_block
<= (u32
)~0ULL)
2028 /* Volume is "huge", so see if our journal is new enough to
2030 if (!(OCFS2_HAS_COMPAT_FEATURE(osb
->sb
,
2031 OCFS2_FEATURE_COMPAT_JBD2_SB
) &&
2032 jbd2_journal_check_used_features(osb
->journal
->j_journal
, 0, 0,
2033 JBD2_FEATURE_INCOMPAT_64BIT
))) {
2034 mlog(ML_ERROR
, "The journal cannot address the entire volume. "
2035 "Enable the 'block64' journal option with tunefs.ocfs2");
2044 static int ocfs2_initialize_super(struct super_block
*sb
,
2045 struct buffer_head
*bh
,
2047 struct ocfs2_blockcheck_stats
*stats
)
2050 int i
, cbits
, bbits
;
2051 struct ocfs2_dinode
*di
= (struct ocfs2_dinode
*)bh
->b_data
;
2052 struct inode
*inode
= NULL
;
2053 struct ocfs2_journal
*journal
;
2054 struct ocfs2_super
*osb
;
2057 osb
= kzalloc(sizeof(struct ocfs2_super
), GFP_KERNEL
);
2064 sb
->s_fs_info
= osb
;
2065 sb
->s_op
= &ocfs2_sops
;
2066 sb
->s_d_op
= &ocfs2_dentry_ops
;
2067 sb
->s_export_op
= &ocfs2_export_ops
;
2068 sb
->s_qcop
= &dquot_quotactl_sysfile_ops
;
2069 sb
->dq_op
= &ocfs2_quota_operations
;
2070 sb
->s_quota_types
= QTYPE_MASK_USR
| QTYPE_MASK_GRP
;
2071 sb
->s_xattr
= ocfs2_xattr_handlers
;
2072 sb
->s_time_gran
= 1;
2073 sb
->s_flags
|= MS_NOATIME
;
2074 /* this is needed to support O_LARGEFILE */
2075 cbits
= le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
);
2076 bbits
= le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
);
2077 sb
->s_maxbytes
= ocfs2_max_file_offset(bbits
, cbits
);
2078 memcpy(sb
->s_uuid
, di
->id2
.i_super
.s_uuid
,
2079 sizeof(di
->id2
.i_super
.s_uuid
));
2081 osb
->osb_dx_mask
= (1 << (cbits
- bbits
)) - 1;
2083 for (i
= 0; i
< 3; i
++)
2084 osb
->osb_dx_seed
[i
] = le32_to_cpu(di
->id2
.i_super
.s_dx_seed
[i
]);
2085 osb
->osb_dx_seed
[3] = le32_to_cpu(di
->id2
.i_super
.s_uuid_hash
);
2088 /* Save off for ocfs2_rw_direct */
2089 osb
->s_sectsize_bits
= blksize_bits(sector_size
);
2090 BUG_ON(!osb
->s_sectsize_bits
);
2092 spin_lock_init(&osb
->dc_task_lock
);
2093 init_waitqueue_head(&osb
->dc_event
);
2094 osb
->dc_work_sequence
= 0;
2095 osb
->dc_wake_sequence
= 0;
2096 INIT_LIST_HEAD(&osb
->blocked_lock_list
);
2097 osb
->blocked_lock_count
= 0;
2098 spin_lock_init(&osb
->osb_lock
);
2099 spin_lock_init(&osb
->osb_xattr_lock
);
2100 ocfs2_init_steal_slots(osb
);
2102 mutex_init(&osb
->system_file_mutex
);
2104 atomic_set(&osb
->alloc_stats
.moves
, 0);
2105 atomic_set(&osb
->alloc_stats
.local_data
, 0);
2106 atomic_set(&osb
->alloc_stats
.bitmap_data
, 0);
2107 atomic_set(&osb
->alloc_stats
.bg_allocs
, 0);
2108 atomic_set(&osb
->alloc_stats
.bg_extends
, 0);
2110 /* Copy the blockcheck stats from the superblock probe */
2111 osb
->osb_ecc_stats
= *stats
;
2113 ocfs2_init_node_maps(osb
);
2115 snprintf(osb
->dev_str
, sizeof(osb
->dev_str
), "%u,%u",
2116 MAJOR(osb
->sb
->s_dev
), MINOR(osb
->sb
->s_dev
));
2118 osb
->max_slots
= le16_to_cpu(di
->id2
.i_super
.s_max_slots
);
2119 if (osb
->max_slots
> OCFS2_MAX_SLOTS
|| osb
->max_slots
== 0) {
2120 mlog(ML_ERROR
, "Invalid number of node slots (%u)\n",
2126 ocfs2_orphan_scan_init(osb
);
2128 status
= ocfs2_recovery_init(osb
);
2130 mlog(ML_ERROR
, "Unable to initialize recovery state\n");
2135 init_waitqueue_head(&osb
->checkpoint_event
);
2137 osb
->s_atime_quantum
= OCFS2_DEFAULT_ATIME_QUANTUM
;
2139 osb
->slot_num
= OCFS2_INVALID_SLOT
;
2141 osb
->s_xattr_inline_size
= le16_to_cpu(
2142 di
->id2
.i_super
.s_xattr_inline_size
);
2144 osb
->local_alloc_state
= OCFS2_LA_UNUSED
;
2145 osb
->local_alloc_bh
= NULL
;
2146 INIT_DELAYED_WORK(&osb
->la_enable_wq
, ocfs2_la_enable_worker
);
2148 init_waitqueue_head(&osb
->osb_mount_event
);
2150 status
= ocfs2_resmap_init(osb
, &osb
->osb_la_resmap
);
2156 osb
->vol_label
= kmalloc(OCFS2_MAX_VOL_LABEL_LEN
, GFP_KERNEL
);
2157 if (!osb
->vol_label
) {
2158 mlog(ML_ERROR
, "unable to alloc vol label\n");
2163 osb
->slot_recovery_generations
=
2164 kcalloc(osb
->max_slots
, sizeof(*osb
->slot_recovery_generations
),
2166 if (!osb
->slot_recovery_generations
) {
2172 init_waitqueue_head(&osb
->osb_wipe_event
);
2173 osb
->osb_orphan_wipes
= kcalloc(osb
->max_slots
,
2174 sizeof(*osb
->osb_orphan_wipes
),
2176 if (!osb
->osb_orphan_wipes
) {
2182 osb
->osb_rf_lock_tree
= RB_ROOT
;
2184 osb
->s_feature_compat
=
2185 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_compat
);
2186 osb
->s_feature_ro_compat
=
2187 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_ro_compat
);
2188 osb
->s_feature_incompat
=
2189 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_incompat
);
2191 if ((i
= OCFS2_HAS_INCOMPAT_FEATURE(osb
->sb
, ~OCFS2_FEATURE_INCOMPAT_SUPP
))) {
2192 mlog(ML_ERROR
, "couldn't mount because of unsupported "
2193 "optional features (%x).\n", i
);
2197 if (!(osb
->sb
->s_flags
& MS_RDONLY
) &&
2198 (i
= OCFS2_HAS_RO_COMPAT_FEATURE(osb
->sb
, ~OCFS2_FEATURE_RO_COMPAT_SUPP
))) {
2199 mlog(ML_ERROR
, "couldn't mount RDWR because of "
2200 "unsupported optional features (%x).\n", i
);
2205 if (ocfs2_clusterinfo_valid(osb
)) {
2206 osb
->osb_stackflags
=
2207 OCFS2_RAW_SB(di
)->s_cluster_info
.ci_stackflags
;
2208 strlcpy(osb
->osb_cluster_stack
,
2209 OCFS2_RAW_SB(di
)->s_cluster_info
.ci_stack
,
2210 OCFS2_STACK_LABEL_LEN
+ 1);
2211 if (strlen(osb
->osb_cluster_stack
) != OCFS2_STACK_LABEL_LEN
) {
2213 "couldn't mount because of an invalid "
2214 "cluster stack label (%s) \n",
2215 osb
->osb_cluster_stack
);
2219 strlcpy(osb
->osb_cluster_name
,
2220 OCFS2_RAW_SB(di
)->s_cluster_info
.ci_cluster
,
2221 OCFS2_CLUSTER_NAME_LEN
+ 1);
2223 /* The empty string is identical with classic tools that
2224 * don't know about s_cluster_info. */
2225 osb
->osb_cluster_stack
[0] = '\0';
2228 get_random_bytes(&osb
->s_next_generation
, sizeof(u32
));
2231 * This should be done in ocfs2_journal_init(), but unknown
2232 * ordering issues will cause the filesystem to crash.
2233 * If anyone wants to figure out what part of the code
2234 * refers to osb->journal before ocfs2_journal_init() is run,
2237 /* initialize our journal structure */
2239 journal
= kzalloc(sizeof(struct ocfs2_journal
), GFP_KERNEL
);
2241 mlog(ML_ERROR
, "unable to alloc journal\n");
2245 osb
->journal
= journal
;
2246 journal
->j_osb
= osb
;
2248 atomic_set(&journal
->j_num_trans
, 0);
2249 init_rwsem(&journal
->j_trans_barrier
);
2250 init_waitqueue_head(&journal
->j_checkpointed
);
2251 spin_lock_init(&journal
->j_lock
);
2252 journal
->j_trans_id
= (unsigned long) 1;
2253 INIT_LIST_HEAD(&journal
->j_la_cleanups
);
2254 INIT_WORK(&journal
->j_recovery_work
, ocfs2_complete_recovery
);
2255 journal
->j_state
= OCFS2_JOURNAL_FREE
;
2257 INIT_WORK(&osb
->dquot_drop_work
, ocfs2_drop_dquot_refs
);
2258 init_llist_head(&osb
->dquot_drop_list
);
2260 /* get some pseudo constants for clustersize bits */
2261 osb
->s_clustersize_bits
=
2262 le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
);
2263 osb
->s_clustersize
= 1 << osb
->s_clustersize_bits
;
2265 if (osb
->s_clustersize
< OCFS2_MIN_CLUSTERSIZE
||
2266 osb
->s_clustersize
> OCFS2_MAX_CLUSTERSIZE
) {
2267 mlog(ML_ERROR
, "Volume has invalid cluster size (%d)\n",
2268 osb
->s_clustersize
);
2273 total_blocks
= ocfs2_clusters_to_blocks(osb
->sb
,
2274 le32_to_cpu(di
->i_clusters
));
2276 status
= generic_check_addressable(osb
->sb
->s_blocksize_bits
,
2279 mlog(ML_ERROR
, "Volume too large "
2280 "to mount safely on this system");
2285 if (ocfs2_setup_osb_uuid(osb
, di
->id2
.i_super
.s_uuid
,
2286 sizeof(di
->id2
.i_super
.s_uuid
))) {
2287 mlog(ML_ERROR
, "Out of memory trying to setup our uuid.\n");
2292 strlcpy(osb
->vol_label
, di
->id2
.i_super
.s_label
,
2293 OCFS2_MAX_VOL_LABEL_LEN
);
2294 osb
->root_blkno
= le64_to_cpu(di
->id2
.i_super
.s_root_blkno
);
2295 osb
->system_dir_blkno
= le64_to_cpu(di
->id2
.i_super
.s_system_dir_blkno
);
2296 osb
->first_cluster_group_blkno
=
2297 le64_to_cpu(di
->id2
.i_super
.s_first_cluster_group
);
2298 osb
->fs_generation
= le32_to_cpu(di
->i_fs_generation
);
2299 osb
->uuid_hash
= le32_to_cpu(di
->id2
.i_super
.s_uuid_hash
);
2300 trace_ocfs2_initialize_super(osb
->vol_label
, osb
->uuid_str
,
2301 (unsigned long long)osb
->root_blkno
,
2302 (unsigned long long)osb
->system_dir_blkno
,
2303 osb
->s_clustersize_bits
);
2305 osb
->osb_dlm_debug
= ocfs2_new_dlm_debug();
2306 if (!osb
->osb_dlm_debug
) {
2312 atomic_set(&osb
->vol_state
, VOLUME_INIT
);
2314 /* load root, system_dir, and all global system inodes */
2315 status
= ocfs2_init_global_system_inodes(osb
);
2324 inode
= ocfs2_get_system_file_inode(osb
, GLOBAL_BITMAP_SYSTEM_INODE
,
2325 OCFS2_INVALID_SLOT
);
2332 osb
->bitmap_blkno
= OCFS2_I(inode
)->ip_blkno
;
2333 osb
->osb_clusters_at_boot
= OCFS2_I(inode
)->ip_clusters
;
2336 osb
->bitmap_cpg
= ocfs2_group_bitmap_size(sb
, 0,
2337 osb
->s_feature_incompat
) * 8;
2339 status
= ocfs2_init_slot_info(osb
);
2344 cleancache_init_shared_fs(sb
);
2351 * will return: -EAGAIN if it is ok to keep searching for superblocks
2352 * -EINVAL if there is a bad superblock
2355 static int ocfs2_verify_volume(struct ocfs2_dinode
*di
,
2356 struct buffer_head
*bh
,
2358 struct ocfs2_blockcheck_stats
*stats
)
2360 int status
= -EAGAIN
;
2362 if (memcmp(di
->i_signature
, OCFS2_SUPER_BLOCK_SIGNATURE
,
2363 strlen(OCFS2_SUPER_BLOCK_SIGNATURE
)) == 0) {
2364 /* We have to do a raw check of the feature here */
2365 if (le32_to_cpu(di
->id2
.i_super
.s_feature_incompat
) &
2366 OCFS2_FEATURE_INCOMPAT_META_ECC
) {
2367 status
= ocfs2_block_check_validate(bh
->b_data
,
2375 if ((1 << le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
)) != blksz
) {
2376 mlog(ML_ERROR
, "found superblock with incorrect block "
2377 "size: found %u, should be %u\n",
2378 1 << le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
),
2380 } else if (le16_to_cpu(di
->id2
.i_super
.s_major_rev_level
) !=
2381 OCFS2_MAJOR_REV_LEVEL
||
2382 le16_to_cpu(di
->id2
.i_super
.s_minor_rev_level
) !=
2383 OCFS2_MINOR_REV_LEVEL
) {
2384 mlog(ML_ERROR
, "found superblock with bad version: "
2385 "found %u.%u, should be %u.%u\n",
2386 le16_to_cpu(di
->id2
.i_super
.s_major_rev_level
),
2387 le16_to_cpu(di
->id2
.i_super
.s_minor_rev_level
),
2388 OCFS2_MAJOR_REV_LEVEL
,
2389 OCFS2_MINOR_REV_LEVEL
);
2390 } else if (bh
->b_blocknr
!= le64_to_cpu(di
->i_blkno
)) {
2391 mlog(ML_ERROR
, "bad block number on superblock: "
2392 "found %llu, should be %llu\n",
2393 (unsigned long long)le64_to_cpu(di
->i_blkno
),
2394 (unsigned long long)bh
->b_blocknr
);
2395 } else if (le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
) < 12 ||
2396 le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
) > 20) {
2397 mlog(ML_ERROR
, "bad cluster size found: %u\n",
2398 1 << le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
));
2399 } else if (!le64_to_cpu(di
->id2
.i_super
.s_root_blkno
)) {
2400 mlog(ML_ERROR
, "bad root_blkno: 0\n");
2401 } else if (!le64_to_cpu(di
->id2
.i_super
.s_system_dir_blkno
)) {
2402 mlog(ML_ERROR
, "bad system_dir_blkno: 0\n");
2403 } else if (le16_to_cpu(di
->id2
.i_super
.s_max_slots
) > OCFS2_MAX_SLOTS
) {
2405 "Superblock slots found greater than file system "
2406 "maximum: found %u, max %u\n",
2407 le16_to_cpu(di
->id2
.i_super
.s_max_slots
),
2416 if (status
&& status
!= -EAGAIN
)
2421 static int ocfs2_check_volume(struct ocfs2_super
*osb
)
2426 struct ocfs2_dinode
*local_alloc
= NULL
; /* only used if we
2430 /* Init our journal object. */
2431 status
= ocfs2_journal_init(osb
->journal
, &dirty
);
2433 mlog(ML_ERROR
, "Could not initialize journal!\n");
2437 /* Now that journal has been initialized, check to make sure
2438 entire volume is addressable. */
2439 status
= ocfs2_journal_addressable(osb
);
2443 /* If the journal was unmounted cleanly then we don't want to
2444 * recover anything. Otherwise, journal_load will do that
2445 * dirty work for us :) */
2447 status
= ocfs2_journal_wipe(osb
->journal
, 0);
2453 printk(KERN_NOTICE
"ocfs2: File system on device (%s) was not "
2454 "unmounted cleanly, recovering it.\n", osb
->dev_str
);
2457 local
= ocfs2_mount_local(osb
);
2459 /* will play back anything left in the journal. */
2460 status
= ocfs2_journal_load(osb
->journal
, local
, dirty
);
2462 mlog(ML_ERROR
, "ocfs2 journal load failed! %d\n", status
);
2466 if (osb
->s_mount_opt
& OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT
)
2467 jbd2_journal_set_features(osb
->journal
->j_journal
,
2468 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2469 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2471 jbd2_journal_clear_features(osb
->journal
->j_journal
,
2472 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2473 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2476 /* recover my local alloc if we didn't unmount cleanly. */
2477 status
= ocfs2_begin_local_alloc_recovery(osb
,
2484 /* we complete the recovery process after we've marked
2485 * ourselves as mounted. */
2488 status
= ocfs2_load_local_alloc(osb
);
2495 /* Recovery will be completed after we've mounted the
2496 * rest of the volume. */
2498 osb
->local_alloc_copy
= local_alloc
;
2502 /* go through each journal, trylock it and if you get the
2503 * lock, and it's marked as dirty, set the bit in the recover
2504 * map and launch a recovery thread for it. */
2505 status
= ocfs2_mark_dead_nodes(osb
);
2511 status
= ocfs2_compute_replay_slots(osb
);
2524 * The routine gets called from dismount or close whenever a dismount on
2525 * volume is requested and the osb open count becomes 1.
2526 * It will remove the osb from the global list and also free up all the
2527 * initialized resources and fileobject.
2529 static void ocfs2_delete_osb(struct ocfs2_super
*osb
)
2531 /* This function assumes that the caller has the main osb resource */
2533 ocfs2_free_slot_info(osb
);
2535 kfree(osb
->osb_orphan_wipes
);
2536 kfree(osb
->slot_recovery_generations
);
2538 * This belongs in journal shutdown, but because we have to
2539 * allocate osb->journal at the start of ocfs2_initialize_osb(),
2542 kfree(osb
->journal
);
2543 kfree(osb
->local_alloc_copy
);
2544 kfree(osb
->uuid_str
);
2545 kfree(osb
->vol_label
);
2546 ocfs2_put_dlm_debug(osb
->osb_dlm_debug
);
2547 memset(osb
, 0, sizeof(struct ocfs2_super
));
2550 /* Depending on the mount option passed, perform one of the following:
2551 * Put OCFS2 into a readonly state (default)
2552 * Return EIO so that only the process errs
2553 * Fix the error as if fsck.ocfs2 -y
2556 static int ocfs2_handle_error(struct super_block
*sb
)
2558 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
2561 ocfs2_set_osb_flag(osb
, OCFS2_OSB_ERROR_FS
);
2562 pr_crit("On-disk corruption discovered. "
2563 "Please run fsck.ocfs2 once the filesystem is unmounted.\n");
2565 if (osb
->s_mount_opt
& OCFS2_MOUNT_ERRORS_PANIC
) {
2566 panic("OCFS2: (device %s): panic forced after error\n",
2568 } else if (osb
->s_mount_opt
& OCFS2_MOUNT_ERRORS_CONT
) {
2569 pr_crit("OCFS2: Returning error to the calling process.\n");
2571 } else { /* default option */
2573 if (sb
->s_flags
& MS_RDONLY
&&
2574 (ocfs2_is_soft_readonly(osb
) ||
2575 ocfs2_is_hard_readonly(osb
)))
2578 pr_crit("OCFS2: File system is now read-only.\n");
2579 sb
->s_flags
|= MS_RDONLY
;
2580 ocfs2_set_ro_flag(osb
, 0);
2586 int __ocfs2_error(struct super_block
*sb
, const char *function
,
2587 const char *fmt
, ...)
2589 struct va_format vaf
;
2592 va_start(args
, fmt
);
2596 /* Not using mlog here because we want to show the actual
2597 * function the error came from. */
2598 printk(KERN_CRIT
"OCFS2: ERROR (device %s): %s: %pV",
2599 sb
->s_id
, function
, &vaf
);
2603 return ocfs2_handle_error(sb
);
2606 /* Handle critical errors. This is intentionally more drastic than
2607 * ocfs2_handle_error, so we only use for things like journal errors,
2609 void __ocfs2_abort(struct super_block
*sb
, const char *function
,
2610 const char *fmt
, ...)
2612 struct va_format vaf
;
2615 va_start(args
, fmt
);
2620 printk(KERN_CRIT
"OCFS2: abort (device %s): %s: %pV",
2621 sb
->s_id
, function
, &vaf
);
2625 /* We don't have the cluster support yet to go straight to
2626 * hard readonly in here. Until then, we want to keep
2627 * ocfs2_abort() so that we can at least mark critical
2630 * TODO: This should abort the journal and alert other nodes
2631 * that our slot needs recovery. */
2633 /* Force a panic(). This stinks, but it's better than letting
2634 * things continue without having a proper hard readonly
2636 if (!ocfs2_mount_local(OCFS2_SB(sb
)))
2637 OCFS2_SB(sb
)->s_mount_opt
|= OCFS2_MOUNT_ERRORS_PANIC
;
2638 ocfs2_handle_error(sb
);
2642 * Void signal blockers, because in-kernel sigprocmask() only fails
2643 * when SIG_* is wrong.
2645 void ocfs2_block_signals(sigset_t
*oldset
)
2650 sigfillset(&blocked
);
2651 rc
= sigprocmask(SIG_BLOCK
, &blocked
, oldset
);
2655 void ocfs2_unblock_signals(sigset_t
*oldset
)
2657 int rc
= sigprocmask(SIG_SETMASK
, oldset
, NULL
);
2661 module_init(ocfs2_init
);
2662 module_exit(ocfs2_exit
);