1 // SPDX-License-Identifier: GPL-2.0-or-later
5 * load/unload driver, mount/dismount volumes
7 * Copyright (C) 2002, 2004 Oracle. All rights reserved.
10 #include <linux/module.h>
12 #include <linux/types.h>
13 #include <linux/slab.h>
14 #include <linux/highmem.h>
15 #include <linux/init.h>
16 #include <linux/random.h>
17 #include <linux/statfs.h>
18 #include <linux/moduleparam.h>
19 #include <linux/blkdev.h>
20 #include <linux/socket.h>
21 #include <linux/inet.h>
22 #include <linux/parser.h>
23 #include <linux/crc32.h>
24 #include <linux/debugfs.h>
25 #include <linux/mount.h>
26 #include <linux/seq_file.h>
27 #include <linux/quotaops.h>
28 #include <linux/cleancache.h>
29 #include <linux/signal.h>
31 #define CREATE_TRACE_POINTS
32 #include "ocfs2_trace.h"
34 #include <cluster/masklog.h>
38 /* this should be the only file to include a version 1 header */
39 #include "ocfs1_fs_compat.h"
43 #include "blockcheck.h"
46 #include "extent_map.h"
47 #include "heartbeat.h"
50 #include "localalloc.h"
58 #include "refcounttree.h"
61 #include "buffer_head_io.h"
62 #include "filecheck.h"
64 static struct kmem_cache
*ocfs2_inode_cachep
;
65 struct kmem_cache
*ocfs2_dquot_cachep
;
66 struct kmem_cache
*ocfs2_qf_chunk_cachep
;
68 static struct dentry
*ocfs2_debugfs_root
;
70 MODULE_AUTHOR("Oracle");
71 MODULE_LICENSE("GPL");
72 MODULE_DESCRIPTION("OCFS2 cluster file system");
76 unsigned long commit_interval
;
77 unsigned long mount_opt
;
78 unsigned int atime_quantum
;
81 unsigned int resv_level
;
83 char cluster_stack
[OCFS2_STACK_LABEL_LEN
+ 1];
86 static int ocfs2_parse_options(struct super_block
*sb
, char *options
,
87 struct mount_options
*mopt
,
89 static int ocfs2_check_set_options(struct super_block
*sb
,
90 struct mount_options
*options
);
91 static int ocfs2_show_options(struct seq_file
*s
, struct dentry
*root
);
92 static void ocfs2_put_super(struct super_block
*sb
);
93 static int ocfs2_mount_volume(struct super_block
*sb
);
94 static int ocfs2_remount(struct super_block
*sb
, int *flags
, char *data
);
95 static void ocfs2_dismount_volume(struct super_block
*sb
, int mnt_err
);
96 static int ocfs2_initialize_mem_caches(void);
97 static void ocfs2_free_mem_caches(void);
98 static void ocfs2_delete_osb(struct ocfs2_super
*osb
);
100 static int ocfs2_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
102 static int ocfs2_sync_fs(struct super_block
*sb
, int wait
);
104 static int ocfs2_init_global_system_inodes(struct ocfs2_super
*osb
);
105 static int ocfs2_init_local_system_inodes(struct ocfs2_super
*osb
);
106 static void ocfs2_release_system_inodes(struct ocfs2_super
*osb
);
107 static int ocfs2_check_volume(struct ocfs2_super
*osb
);
108 static int ocfs2_verify_volume(struct ocfs2_dinode
*di
,
109 struct buffer_head
*bh
,
111 struct ocfs2_blockcheck_stats
*stats
);
112 static int ocfs2_initialize_super(struct super_block
*sb
,
113 struct buffer_head
*bh
,
115 struct ocfs2_blockcheck_stats
*stats
);
116 static int ocfs2_get_sector(struct super_block
*sb
,
117 struct buffer_head
**bh
,
120 static struct inode
*ocfs2_alloc_inode(struct super_block
*sb
);
121 static void ocfs2_free_inode(struct inode
*inode
);
122 static int ocfs2_susp_quotas(struct ocfs2_super
*osb
, int unsuspend
);
123 static int ocfs2_enable_quotas(struct ocfs2_super
*osb
);
124 static void ocfs2_disable_quotas(struct ocfs2_super
*osb
);
126 static struct dquot
**ocfs2_get_dquots(struct inode
*inode
)
128 return OCFS2_I(inode
)->i_dquot
;
131 static const struct super_operations ocfs2_sops
= {
132 .statfs
= ocfs2_statfs
,
133 .alloc_inode
= ocfs2_alloc_inode
,
134 .free_inode
= ocfs2_free_inode
,
135 .drop_inode
= ocfs2_drop_inode
,
136 .evict_inode
= ocfs2_evict_inode
,
137 .sync_fs
= ocfs2_sync_fs
,
138 .put_super
= ocfs2_put_super
,
139 .remount_fs
= ocfs2_remount
,
140 .show_options
= ocfs2_show_options
,
141 .quota_read
= ocfs2_quota_read
,
142 .quota_write
= ocfs2_quota_write
,
143 .get_dquots
= ocfs2_get_dquots
,
170 Opt_coherency_buffered
,
174 Opt_journal_async_commit
,
180 static const match_table_t tokens
= {
181 {Opt_barrier
, "barrier=%u"},
182 {Opt_err_panic
, "errors=panic"},
183 {Opt_err_ro
, "errors=remount-ro"},
185 {Opt_nointr
, "nointr"},
186 {Opt_hb_none
, OCFS2_HB_NONE
},
187 {Opt_hb_local
, OCFS2_HB_LOCAL
},
188 {Opt_hb_global
, OCFS2_HB_GLOBAL
},
189 {Opt_data_ordered
, "data=ordered"},
190 {Opt_data_writeback
, "data=writeback"},
191 {Opt_atime_quantum
, "atime_quantum=%u"},
192 {Opt_slot
, "preferred_slot=%u"},
193 {Opt_commit
, "commit=%u"},
194 {Opt_localalloc
, "localalloc=%d"},
195 {Opt_localflocks
, "localflocks"},
196 {Opt_stack
, "cluster_stack=%s"},
197 {Opt_user_xattr
, "user_xattr"},
198 {Opt_nouser_xattr
, "nouser_xattr"},
199 {Opt_inode64
, "inode64"},
201 {Opt_noacl
, "noacl"},
202 {Opt_usrquota
, "usrquota"},
203 {Opt_grpquota
, "grpquota"},
204 {Opt_coherency_buffered
, "coherency=buffered"},
205 {Opt_coherency_full
, "coherency=full"},
206 {Opt_resv_level
, "resv_level=%u"},
207 {Opt_dir_resv_level
, "dir_resv_level=%u"},
208 {Opt_journal_async_commit
, "journal_async_commit"},
209 {Opt_err_cont
, "errors=continue"},
210 {Opt_nocluster
, "nocluster"},
214 #ifdef CONFIG_DEBUG_FS
215 static int ocfs2_osb_dump(struct ocfs2_super
*osb
, char *buf
, int len
)
217 struct ocfs2_cluster_connection
*cconn
= osb
->cconn
;
218 struct ocfs2_recovery_map
*rm
= osb
->recovery_map
;
219 struct ocfs2_orphan_scan
*os
= &osb
->osb_orphan_scan
;
223 out
+= scnprintf(buf
+ out
, len
- out
,
224 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n",
225 "Device", osb
->dev_str
, osb
->uuid_str
,
226 osb
->fs_generation
, osb
->vol_label
);
228 out
+= scnprintf(buf
+ out
, len
- out
,
229 "%10s => State: %d Flags: 0x%lX\n", "Volume",
230 atomic_read(&osb
->vol_state
), osb
->osb_flags
);
232 out
+= scnprintf(buf
+ out
, len
- out
,
233 "%10s => Block: %lu Cluster: %d\n", "Sizes",
234 osb
->sb
->s_blocksize
, osb
->s_clustersize
);
236 out
+= scnprintf(buf
+ out
, len
- out
,
237 "%10s => Compat: 0x%X Incompat: 0x%X "
239 "Features", osb
->s_feature_compat
,
240 osb
->s_feature_incompat
, osb
->s_feature_ro_compat
);
242 out
+= scnprintf(buf
+ out
, len
- out
,
243 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount",
244 osb
->s_mount_opt
, osb
->s_atime_quantum
);
247 out
+= scnprintf(buf
+ out
, len
- out
,
248 "%10s => Stack: %s Name: %*s "
249 "Version: %d.%d\n", "Cluster",
250 (*osb
->osb_cluster_stack
== '\0' ?
251 "o2cb" : osb
->osb_cluster_stack
),
252 cconn
->cc_namelen
, cconn
->cc_name
,
253 cconn
->cc_version
.pv_major
,
254 cconn
->cc_version
.pv_minor
);
257 spin_lock_irqsave(&osb
->dc_task_lock
, flags
);
258 out
+= scnprintf(buf
+ out
, len
- out
,
259 "%10s => Pid: %d Count: %lu WakeSeq: %lu "
260 "WorkSeq: %lu\n", "DownCnvt",
261 (osb
->dc_task
? task_pid_nr(osb
->dc_task
) : -1),
262 osb
->blocked_lock_count
, osb
->dc_wake_sequence
,
263 osb
->dc_work_sequence
);
264 spin_unlock_irqrestore(&osb
->dc_task_lock
, flags
);
266 spin_lock(&osb
->osb_lock
);
267 out
+= scnprintf(buf
+ out
, len
- out
, "%10s => Pid: %d Nodes:",
269 (osb
->recovery_thread_task
?
270 task_pid_nr(osb
->recovery_thread_task
) : -1));
271 if (rm
->rm_used
== 0)
272 out
+= scnprintf(buf
+ out
, len
- out
, " None\n");
274 for (i
= 0; i
< rm
->rm_used
; i
++)
275 out
+= scnprintf(buf
+ out
, len
- out
, " %d",
277 out
+= scnprintf(buf
+ out
, len
- out
, "\n");
279 spin_unlock(&osb
->osb_lock
);
281 out
+= scnprintf(buf
+ out
, len
- out
,
282 "%10s => Pid: %d Interval: %lu\n", "Commit",
283 (osb
->commit_task
? task_pid_nr(osb
->commit_task
) : -1),
284 osb
->osb_commit_interval
);
286 out
+= scnprintf(buf
+ out
, len
- out
,
287 "%10s => State: %d TxnId: %lu NumTxns: %d\n",
288 "Journal", osb
->journal
->j_state
,
289 osb
->journal
->j_trans_id
,
290 atomic_read(&osb
->journal
->j_num_trans
));
292 out
+= scnprintf(buf
+ out
, len
- out
,
293 "%10s => GlobalAllocs: %d LocalAllocs: %d "
294 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
296 atomic_read(&osb
->alloc_stats
.bitmap_data
),
297 atomic_read(&osb
->alloc_stats
.local_data
),
298 atomic_read(&osb
->alloc_stats
.bg_allocs
),
299 atomic_read(&osb
->alloc_stats
.moves
),
300 atomic_read(&osb
->alloc_stats
.bg_extends
));
302 out
+= scnprintf(buf
+ out
, len
- out
,
303 "%10s => State: %u Descriptor: %llu Size: %u bits "
304 "Default: %u bits\n",
305 "LocalAlloc", osb
->local_alloc_state
,
306 (unsigned long long)osb
->la_last_gd
,
307 osb
->local_alloc_bits
, osb
->local_alloc_default_bits
);
309 spin_lock(&osb
->osb_lock
);
310 out
+= scnprintf(buf
+ out
, len
- out
,
311 "%10s => InodeSlot: %d StolenInodes: %d, "
312 "MetaSlot: %d StolenMeta: %d\n", "Steal",
313 osb
->s_inode_steal_slot
,
314 atomic_read(&osb
->s_num_inodes_stolen
),
315 osb
->s_meta_steal_slot
,
316 atomic_read(&osb
->s_num_meta_stolen
));
317 spin_unlock(&osb
->osb_lock
);
319 out
+= scnprintf(buf
+ out
, len
- out
, "OrphanScan => ");
320 out
+= scnprintf(buf
+ out
, len
- out
, "Local: %u Global: %u ",
321 os
->os_count
, os
->os_seqno
);
322 out
+= scnprintf(buf
+ out
, len
- out
, " Last Scan: ");
323 if (atomic_read(&os
->os_state
) == ORPHAN_SCAN_INACTIVE
)
324 out
+= scnprintf(buf
+ out
, len
- out
, "Disabled\n");
326 out
+= scnprintf(buf
+ out
, len
- out
, "%lu seconds ago\n",
327 (unsigned long)(ktime_get_seconds() - os
->os_scantime
));
329 out
+= scnprintf(buf
+ out
, len
- out
, "%10s => %3s %10s\n",
330 "Slots", "Num", "RecoGen");
331 for (i
= 0; i
< osb
->max_slots
; ++i
) {
332 out
+= scnprintf(buf
+ out
, len
- out
,
333 "%10s %c %3d %10d\n",
335 (i
== osb
->slot_num
? '*' : ' '),
336 i
, osb
->slot_recovery_generations
[i
]);
342 static int ocfs2_osb_debug_open(struct inode
*inode
, struct file
*file
)
344 struct ocfs2_super
*osb
= inode
->i_private
;
347 buf
= kmalloc(PAGE_SIZE
, GFP_KERNEL
);
351 i_size_write(inode
, ocfs2_osb_dump(osb
, buf
, PAGE_SIZE
));
353 file
->private_data
= buf
;
360 static int ocfs2_debug_release(struct inode
*inode
, struct file
*file
)
362 kfree(file
->private_data
);
366 static ssize_t
ocfs2_debug_read(struct file
*file
, char __user
*buf
,
367 size_t nbytes
, loff_t
*ppos
)
369 return simple_read_from_buffer(buf
, nbytes
, ppos
, file
->private_data
,
370 i_size_read(file
->f_mapping
->host
));
373 static int ocfs2_osb_debug_open(struct inode
*inode
, struct file
*file
)
377 static int ocfs2_debug_release(struct inode
*inode
, struct file
*file
)
381 static ssize_t
ocfs2_debug_read(struct file
*file
, char __user
*buf
,
382 size_t nbytes
, loff_t
*ppos
)
386 #endif /* CONFIG_DEBUG_FS */
388 static const struct file_operations ocfs2_osb_debug_fops
= {
389 .open
= ocfs2_osb_debug_open
,
390 .release
= ocfs2_debug_release
,
391 .read
= ocfs2_debug_read
,
392 .llseek
= generic_file_llseek
,
395 static int ocfs2_sync_fs(struct super_block
*sb
, int wait
)
399 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
401 if (ocfs2_is_hard_readonly(osb
))
405 status
= ocfs2_flush_truncate_log(osb
);
409 ocfs2_schedule_truncate_log_flush(osb
, 0);
412 if (jbd2_journal_start_commit(osb
->journal
->j_journal
,
415 jbd2_log_wait_commit(osb
->journal
->j_journal
,
421 static int ocfs2_need_system_inode(struct ocfs2_super
*osb
, int ino
)
423 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb
->sb
, OCFS2_FEATURE_RO_COMPAT_USRQUOTA
)
424 && (ino
== USER_QUOTA_SYSTEM_INODE
425 || ino
== LOCAL_USER_QUOTA_SYSTEM_INODE
))
427 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb
->sb
, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
)
428 && (ino
== GROUP_QUOTA_SYSTEM_INODE
429 || ino
== LOCAL_GROUP_QUOTA_SYSTEM_INODE
))
434 static int ocfs2_init_global_system_inodes(struct ocfs2_super
*osb
)
436 struct inode
*new = NULL
;
440 new = ocfs2_iget(osb
, osb
->root_blkno
, OCFS2_FI_FLAG_SYSFILE
, 0);
442 status
= PTR_ERR(new);
446 osb
->root_inode
= new;
448 new = ocfs2_iget(osb
, osb
->system_dir_blkno
, OCFS2_FI_FLAG_SYSFILE
, 0);
450 status
= PTR_ERR(new);
454 osb
->sys_root_inode
= new;
456 for (i
= OCFS2_FIRST_ONLINE_SYSTEM_INODE
;
457 i
<= OCFS2_LAST_GLOBAL_SYSTEM_INODE
; i
++) {
458 if (!ocfs2_need_system_inode(osb
, i
))
460 new = ocfs2_get_system_file_inode(osb
, i
, osb
->slot_num
);
462 ocfs2_release_system_inodes(osb
);
463 status
= ocfs2_is_soft_readonly(osb
) ? -EROFS
: -EINVAL
;
465 mlog(ML_ERROR
, "Unable to load system inode %d, "
466 "possibly corrupt fs?", i
);
469 // the array now has one ref, so drop this one
479 static int ocfs2_init_local_system_inodes(struct ocfs2_super
*osb
)
481 struct inode
*new = NULL
;
485 for (i
= OCFS2_LAST_GLOBAL_SYSTEM_INODE
+ 1;
486 i
< NUM_SYSTEM_INODES
;
488 if (!ocfs2_need_system_inode(osb
, i
))
490 new = ocfs2_get_system_file_inode(osb
, i
, osb
->slot_num
);
492 ocfs2_release_system_inodes(osb
);
493 status
= ocfs2_is_soft_readonly(osb
) ? -EROFS
: -EINVAL
;
494 mlog(ML_ERROR
, "status=%d, sysfile=%d, slot=%d\n",
495 status
, i
, osb
->slot_num
);
498 /* the array now has one ref, so drop this one */
508 static void ocfs2_release_system_inodes(struct ocfs2_super
*osb
)
513 for (i
= 0; i
< NUM_GLOBAL_SYSTEM_INODES
; i
++) {
514 inode
= osb
->global_system_inodes
[i
];
517 osb
->global_system_inodes
[i
] = NULL
;
521 inode
= osb
->sys_root_inode
;
524 osb
->sys_root_inode
= NULL
;
527 inode
= osb
->root_inode
;
530 osb
->root_inode
= NULL
;
533 if (!osb
->local_system_inodes
)
536 for (i
= 0; i
< NUM_LOCAL_SYSTEM_INODES
* osb
->max_slots
; i
++) {
537 if (osb
->local_system_inodes
[i
]) {
538 iput(osb
->local_system_inodes
[i
]);
539 osb
->local_system_inodes
[i
] = NULL
;
543 kfree(osb
->local_system_inodes
);
544 osb
->local_system_inodes
= NULL
;
547 /* We're allocating fs objects, use GFP_NOFS */
548 static struct inode
*ocfs2_alloc_inode(struct super_block
*sb
)
550 struct ocfs2_inode_info
*oi
;
552 oi
= kmem_cache_alloc(ocfs2_inode_cachep
, GFP_NOFS
);
557 oi
->i_datasync_tid
= 0;
558 memset(&oi
->i_dquot
, 0, sizeof(oi
->i_dquot
));
560 jbd2_journal_init_jbd_inode(&oi
->ip_jinode
, &oi
->vfs_inode
);
561 return &oi
->vfs_inode
;
564 static void ocfs2_free_inode(struct inode
*inode
)
566 kmem_cache_free(ocfs2_inode_cachep
, OCFS2_I(inode
));
569 static unsigned long long ocfs2_max_file_offset(unsigned int bbits
,
572 unsigned int bytes
= 1 << cbits
;
573 unsigned int trim
= bytes
;
574 unsigned int bitshift
= 32;
577 * i_size and all block offsets in ocfs2 are always 64 bits
578 * wide. i_clusters is 32 bits, in cluster-sized units. So on
579 * 64 bit platforms, cluster size will be the limiting factor.
582 #if BITS_PER_LONG == 32
583 BUILD_BUG_ON(sizeof(sector_t
) != 8);
585 * We might be limited by page cache size.
587 if (bytes
> PAGE_SIZE
) {
591 * Shift by 31 here so that we don't get larger than
599 * Trim by a whole cluster when we can actually approach the
600 * on-disk limits. Otherwise we can overflow i_clusters when
601 * an extent start is at the max offset.
603 return (((unsigned long long)bytes
) << bitshift
) - trim
;
606 static int ocfs2_remount(struct super_block
*sb
, int *flags
, char *data
)
608 int incompat_features
;
610 struct mount_options parsed_options
;
611 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
616 if (!ocfs2_parse_options(sb
, data
, &parsed_options
, 1) ||
617 !ocfs2_check_set_options(sb
, &parsed_options
)) {
622 tmp
= OCFS2_MOUNT_NOCLUSTER
;
623 if ((osb
->s_mount_opt
& tmp
) != (parsed_options
.mount_opt
& tmp
)) {
625 mlog(ML_ERROR
, "Cannot change nocluster option on remount\n");
629 tmp
= OCFS2_MOUNT_HB_LOCAL
| OCFS2_MOUNT_HB_GLOBAL
|
631 if ((osb
->s_mount_opt
& tmp
) != (parsed_options
.mount_opt
& tmp
)) {
633 mlog(ML_ERROR
, "Cannot change heartbeat mode on remount\n");
637 if ((osb
->s_mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
) !=
638 (parsed_options
.mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
)) {
640 mlog(ML_ERROR
, "Cannot change data mode on remount\n");
644 /* Probably don't want this on remount; it might
645 * mess with other nodes */
646 if (!(osb
->s_mount_opt
& OCFS2_MOUNT_INODE64
) &&
647 (parsed_options
.mount_opt
& OCFS2_MOUNT_INODE64
)) {
649 mlog(ML_ERROR
, "Cannot enable inode64 on remount\n");
653 /* We're going to/from readonly mode. */
654 if ((bool)(*flags
& SB_RDONLY
) != sb_rdonly(sb
)) {
655 /* Disable quota accounting before remounting RO */
656 if (*flags
& SB_RDONLY
) {
657 ret
= ocfs2_susp_quotas(osb
, 0);
661 /* Lock here so the check of HARD_RO and the potential
662 * setting of SOFT_RO is atomic. */
663 spin_lock(&osb
->osb_lock
);
664 if (osb
->osb_flags
& OCFS2_OSB_HARD_RO
) {
665 mlog(ML_ERROR
, "Remount on readonly device is forbidden.\n");
670 if (*flags
& SB_RDONLY
) {
671 sb
->s_flags
|= SB_RDONLY
;
672 osb
->osb_flags
|= OCFS2_OSB_SOFT_RO
;
674 if (osb
->osb_flags
& OCFS2_OSB_ERROR_FS
) {
675 mlog(ML_ERROR
, "Cannot remount RDWR "
676 "filesystem due to previous errors.\n");
680 incompat_features
= OCFS2_HAS_RO_COMPAT_FEATURE(sb
, ~OCFS2_FEATURE_RO_COMPAT_SUPP
);
681 if (incompat_features
) {
682 mlog(ML_ERROR
, "Cannot remount RDWR because "
683 "of unsupported optional features "
684 "(%x).\n", incompat_features
);
688 sb
->s_flags
&= ~SB_RDONLY
;
689 osb
->osb_flags
&= ~OCFS2_OSB_SOFT_RO
;
691 trace_ocfs2_remount(sb
->s_flags
, osb
->osb_flags
, *flags
);
693 spin_unlock(&osb
->osb_lock
);
694 /* Enable quota accounting after remounting RW */
695 if (!ret
&& !(*flags
& SB_RDONLY
)) {
696 if (sb_any_quota_suspended(sb
))
697 ret
= ocfs2_susp_quotas(osb
, 1);
699 ret
= ocfs2_enable_quotas(osb
);
701 /* Return back changes... */
702 spin_lock(&osb
->osb_lock
);
703 sb
->s_flags
|= SB_RDONLY
;
704 osb
->osb_flags
|= OCFS2_OSB_SOFT_RO
;
705 spin_unlock(&osb
->osb_lock
);
712 /* Only save off the new mount options in case of a successful
714 osb
->s_mount_opt
= parsed_options
.mount_opt
;
715 osb
->s_atime_quantum
= parsed_options
.atime_quantum
;
716 osb
->preferred_slot
= parsed_options
.slot
;
717 if (parsed_options
.commit_interval
)
718 osb
->osb_commit_interval
= parsed_options
.commit_interval
;
720 if (!ocfs2_is_hard_readonly(osb
))
721 ocfs2_set_journal_params(osb
);
723 sb
->s_flags
= (sb
->s_flags
& ~SB_POSIXACL
) |
724 ((osb
->s_mount_opt
& OCFS2_MOUNT_POSIX_ACL
) ?
731 static int ocfs2_sb_probe(struct super_block
*sb
,
732 struct buffer_head
**bh
,
734 struct ocfs2_blockcheck_stats
*stats
)
737 struct ocfs1_vol_disk_hdr
*hdr
;
738 struct ocfs2_dinode
*di
;
744 *sector_size
= bdev_logical_block_size(sb
->s_bdev
);
745 if (*sector_size
> OCFS2_MAX_BLOCKSIZE
) {
746 mlog(ML_ERROR
, "Hardware sector size too large: %d (max=%d)\n",
747 *sector_size
, OCFS2_MAX_BLOCKSIZE
);
752 /* Can this really happen? */
753 if (*sector_size
< OCFS2_MIN_BLOCKSIZE
)
754 *sector_size
= OCFS2_MIN_BLOCKSIZE
;
756 /* check block zero for old format */
757 status
= ocfs2_get_sector(sb
, bh
, 0, *sector_size
);
762 hdr
= (struct ocfs1_vol_disk_hdr
*) (*bh
)->b_data
;
763 if (hdr
->major_version
== OCFS1_MAJOR_VERSION
) {
764 mlog(ML_ERROR
, "incompatible version: %u.%u\n",
765 hdr
->major_version
, hdr
->minor_version
);
768 if (memcmp(hdr
->signature
, OCFS1_VOLUME_SIGNATURE
,
769 strlen(OCFS1_VOLUME_SIGNATURE
)) == 0) {
770 mlog(ML_ERROR
, "incompatible volume signature: %8s\n",
777 mlog(ML_ERROR
, "This is an ocfs v1 filesystem which must be "
778 "upgraded before mounting with ocfs v2\n");
783 * Now check at magic offset for 512, 1024, 2048, 4096
784 * blocksizes. 4096 is the maximum blocksize because it is
785 * the minimum clustersize.
788 for (blksize
= *sector_size
;
789 blksize
<= OCFS2_MAX_BLOCKSIZE
;
791 tmpstat
= ocfs2_get_sector(sb
, bh
,
792 OCFS2_SUPER_BLOCK_BLKNO
,
799 di
= (struct ocfs2_dinode
*) (*bh
)->b_data
;
800 memset(stats
, 0, sizeof(struct ocfs2_blockcheck_stats
));
801 spin_lock_init(&stats
->b_lock
);
802 tmpstat
= ocfs2_verify_volume(di
, *bh
, blksize
, stats
);
807 if (tmpstat
!= -EAGAIN
) {
817 static int ocfs2_verify_heartbeat(struct ocfs2_super
*osb
)
819 u32 hb_enabled
= OCFS2_MOUNT_HB_LOCAL
| OCFS2_MOUNT_HB_GLOBAL
;
821 if (osb
->s_mount_opt
& hb_enabled
) {
822 if (ocfs2_mount_local(osb
)) {
823 mlog(ML_ERROR
, "Cannot heartbeat on a locally "
824 "mounted device.\n");
827 if (ocfs2_userspace_stack(osb
)) {
828 mlog(ML_ERROR
, "Userspace stack expected, but "
829 "o2cb heartbeat arguments passed to mount\n");
832 if (((osb
->s_mount_opt
& OCFS2_MOUNT_HB_GLOBAL
) &&
833 !ocfs2_cluster_o2cb_global_heartbeat(osb
)) ||
834 ((osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) &&
835 ocfs2_cluster_o2cb_global_heartbeat(osb
))) {
836 mlog(ML_ERROR
, "Mismatching o2cb heartbeat modes\n");
841 if (!(osb
->s_mount_opt
& hb_enabled
)) {
842 if (!ocfs2_mount_local(osb
) && !ocfs2_is_hard_readonly(osb
) &&
843 !ocfs2_userspace_stack(osb
)) {
844 mlog(ML_ERROR
, "Heartbeat has to be started to mount "
845 "a read-write clustered device.\n");
854 * If we're using a userspace stack, mount should have passed
855 * a name that matches the disk. If not, mount should not
856 * have passed a stack.
858 static int ocfs2_verify_userspace_stack(struct ocfs2_super
*osb
,
859 struct mount_options
*mopt
)
861 if (!ocfs2_userspace_stack(osb
) && mopt
->cluster_stack
[0]) {
863 "cluster stack passed to mount, but this filesystem "
864 "does not support it\n");
868 if (ocfs2_userspace_stack(osb
) &&
869 !(osb
->s_mount_opt
& OCFS2_MOUNT_NOCLUSTER
) &&
870 strncmp(osb
->osb_cluster_stack
, mopt
->cluster_stack
,
871 OCFS2_STACK_LABEL_LEN
)) {
873 "cluster stack passed to mount (\"%s\") does not "
874 "match the filesystem (\"%s\")\n",
876 osb
->osb_cluster_stack
);
883 static int ocfs2_susp_quotas(struct ocfs2_super
*osb
, int unsuspend
)
886 struct super_block
*sb
= osb
->sb
;
887 unsigned int feature
[OCFS2_MAXQUOTAS
] = {
888 OCFS2_FEATURE_RO_COMPAT_USRQUOTA
,
889 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
};
892 for (type
= 0; type
< OCFS2_MAXQUOTAS
; type
++) {
893 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb
, feature
[type
]))
896 status
= dquot_resume(sb
, type
);
898 struct ocfs2_mem_dqinfo
*oinfo
;
900 /* Cancel periodic syncing before suspending */
901 oinfo
= sb_dqinfo(sb
, type
)->dqi_priv
;
902 cancel_delayed_work_sync(&oinfo
->dqi_sync_work
);
903 status
= dquot_suspend(sb
, type
);
909 mlog(ML_ERROR
, "Failed to suspend/unsuspend quotas on "
910 "remount (error = %d).\n", status
);
914 static int ocfs2_enable_quotas(struct ocfs2_super
*osb
)
916 struct inode
*inode
[OCFS2_MAXQUOTAS
] = { NULL
, NULL
};
917 struct super_block
*sb
= osb
->sb
;
918 unsigned int feature
[OCFS2_MAXQUOTAS
] = {
919 OCFS2_FEATURE_RO_COMPAT_USRQUOTA
,
920 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
};
921 unsigned int ino
[OCFS2_MAXQUOTAS
] = {
922 LOCAL_USER_QUOTA_SYSTEM_INODE
,
923 LOCAL_GROUP_QUOTA_SYSTEM_INODE
};
927 sb_dqopt(sb
)->flags
|= DQUOT_QUOTA_SYS_FILE
| DQUOT_NEGATIVE_USAGE
;
928 for (type
= 0; type
< OCFS2_MAXQUOTAS
; type
++) {
929 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb
, feature
[type
]))
931 inode
[type
] = ocfs2_get_system_file_inode(osb
, ino
[type
],
937 status
= dquot_load_quota_inode(inode
[type
], type
, QFMT_OCFS2
,
938 DQUOT_USAGE_ENABLED
);
943 for (type
= 0; type
< OCFS2_MAXQUOTAS
; type
++)
947 ocfs2_disable_quotas(osb
);
948 for (type
= 0; type
< OCFS2_MAXQUOTAS
; type
++)
954 static void ocfs2_disable_quotas(struct ocfs2_super
*osb
)
958 struct super_block
*sb
= osb
->sb
;
959 struct ocfs2_mem_dqinfo
*oinfo
;
961 /* We mostly ignore errors in this function because there's not much
962 * we can do when we see them */
963 for (type
= 0; type
< OCFS2_MAXQUOTAS
; type
++) {
964 if (!sb_has_quota_loaded(sb
, type
))
966 oinfo
= sb_dqinfo(sb
, type
)->dqi_priv
;
967 cancel_delayed_work_sync(&oinfo
->dqi_sync_work
);
968 inode
= igrab(sb
->s_dquot
.files
[type
]);
969 /* Turn off quotas. This will remove all dquot structures from
970 * memory and so they will be automatically synced to global
972 dquot_disable(sb
, type
, DQUOT_USAGE_ENABLED
|
973 DQUOT_LIMITS_ENABLED
);
978 static int ocfs2_fill_super(struct super_block
*sb
, void *data
, int silent
)
981 int status
, sector_size
;
982 struct mount_options parsed_options
;
983 struct inode
*inode
= NULL
;
984 struct ocfs2_super
*osb
= NULL
;
985 struct buffer_head
*bh
= NULL
;
987 struct ocfs2_blockcheck_stats stats
;
989 trace_ocfs2_fill_super(sb
, data
, silent
);
991 if (!ocfs2_parse_options(sb
, data
, &parsed_options
, 0)) {
993 goto read_super_error
;
996 /* probe for superblock */
997 status
= ocfs2_sb_probe(sb
, &bh
, §or_size
, &stats
);
999 mlog(ML_ERROR
, "superblock probe failed!\n");
1000 goto read_super_error
;
1003 status
= ocfs2_initialize_super(sb
, bh
, sector_size
, &stats
);
1007 goto read_super_error
;
1012 if (!ocfs2_check_set_options(sb
, &parsed_options
)) {
1014 goto read_super_error
;
1016 osb
->s_mount_opt
= parsed_options
.mount_opt
;
1017 osb
->s_atime_quantum
= parsed_options
.atime_quantum
;
1018 osb
->preferred_slot
= parsed_options
.slot
;
1019 osb
->osb_commit_interval
= parsed_options
.commit_interval
;
1021 ocfs2_la_set_sizes(osb
, parsed_options
.localalloc_opt
);
1022 osb
->osb_resv_level
= parsed_options
.resv_level
;
1023 osb
->osb_dir_resv_level
= parsed_options
.resv_level
;
1024 if (parsed_options
.dir_resv_level
== -1)
1025 osb
->osb_dir_resv_level
= parsed_options
.resv_level
;
1027 osb
->osb_dir_resv_level
= parsed_options
.dir_resv_level
;
1029 status
= ocfs2_verify_userspace_stack(osb
, &parsed_options
);
1031 goto read_super_error
;
1033 sb
->s_magic
= OCFS2_SUPER_MAGIC
;
1035 sb
->s_flags
= (sb
->s_flags
& ~(SB_POSIXACL
| SB_NOSEC
)) |
1036 ((osb
->s_mount_opt
& OCFS2_MOUNT_POSIX_ACL
) ? SB_POSIXACL
: 0);
1038 /* Hard readonly mode only if: bdev_read_only, SB_RDONLY,
1040 if (bdev_read_only(sb
->s_bdev
)) {
1041 if (!sb_rdonly(sb
)) {
1043 mlog(ML_ERROR
, "Readonly device detected but readonly "
1044 "mount was not specified.\n");
1045 goto read_super_error
;
1048 /* You should not be able to start a local heartbeat
1049 * on a readonly device. */
1050 if (osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) {
1052 mlog(ML_ERROR
, "Local heartbeat specified on readonly "
1054 goto read_super_error
;
1057 status
= ocfs2_check_journals_nolocks(osb
);
1059 if (status
== -EROFS
)
1060 mlog(ML_ERROR
, "Recovery required on readonly "
1061 "file system, but write access is "
1065 goto read_super_error
;
1068 ocfs2_set_ro_flag(osb
, 1);
1070 printk(KERN_NOTICE
"ocfs2: Readonly device (%s) detected. "
1071 "Cluster services will not be used for this mount. "
1072 "Recovery will be skipped.\n", osb
->dev_str
);
1075 if (!ocfs2_is_hard_readonly(osb
)) {
1077 ocfs2_set_ro_flag(osb
, 0);
1080 status
= ocfs2_verify_heartbeat(osb
);
1083 goto read_super_error
;
1086 osb
->osb_debug_root
= debugfs_create_dir(osb
->uuid_str
,
1087 ocfs2_debugfs_root
);
1089 debugfs_create_file("fs_state", S_IFREG
|S_IRUSR
, osb
->osb_debug_root
,
1090 osb
, &ocfs2_osb_debug_fops
);
1092 if (ocfs2_meta_ecc(osb
))
1093 ocfs2_blockcheck_stats_debugfs_install( &osb
->osb_ecc_stats
,
1094 osb
->osb_debug_root
);
1096 status
= ocfs2_mount_volume(sb
);
1098 goto read_super_error
;
1100 if (osb
->root_inode
)
1101 inode
= igrab(osb
->root_inode
);
1106 goto read_super_error
;
1109 root
= d_make_root(inode
);
1113 goto read_super_error
;
1118 ocfs2_complete_mount_recovery(osb
);
1120 osb
->osb_dev_kset
= kset_create_and_add(sb
->s_id
, NULL
,
1122 if (!osb
->osb_dev_kset
) {
1124 mlog(ML_ERROR
, "Unable to create device kset %s.\n", sb
->s_id
);
1125 goto read_super_error
;
1128 /* Create filecheck sysfs related directories/files at
1129 * /sys/fs/ocfs2/<devname>/filecheck */
1130 if (ocfs2_filecheck_create_sysfs(osb
)) {
1132 mlog(ML_ERROR
, "Unable to create filecheck sysfs directory at "
1133 "/sys/fs/ocfs2/%s/filecheck.\n", sb
->s_id
);
1134 goto read_super_error
;
1137 if (ocfs2_mount_local(osb
))
1138 snprintf(nodestr
, sizeof(nodestr
), "local");
1140 snprintf(nodestr
, sizeof(nodestr
), "%u", osb
->node_num
);
1142 printk(KERN_INFO
"ocfs2: Mounting device (%s) on (node %s, slot %d) "
1143 "with %s data mode.\n",
1144 osb
->dev_str
, nodestr
, osb
->slot_num
,
1145 osb
->s_mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
? "writeback" :
1148 if ((osb
->s_mount_opt
& OCFS2_MOUNT_NOCLUSTER
) &&
1149 !(osb
->s_feature_incompat
& OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT
))
1150 printk(KERN_NOTICE
"ocfs2: The shared device (%s) is mounted "
1151 "without cluster aware mode.\n", osb
->dev_str
);
1153 atomic_set(&osb
->vol_state
, VOLUME_MOUNTED
);
1154 wake_up(&osb
->osb_mount_event
);
1156 /* Now we can initialize quotas because we can afford to wait
1157 * for cluster locks recovery now. That also means that truncation
1158 * log recovery can happen but that waits for proper quota setup */
1159 if (!sb_rdonly(sb
)) {
1160 status
= ocfs2_enable_quotas(osb
);
1162 /* We have to err-out specially here because
1163 * s_root is already set */
1165 atomic_set(&osb
->vol_state
, VOLUME_DISABLED
);
1166 wake_up(&osb
->osb_mount_event
);
1171 ocfs2_complete_quota_recovery(osb
);
1173 /* Now we wake up again for processes waiting for quotas */
1174 atomic_set(&osb
->vol_state
, VOLUME_MOUNTED_QUOTAS
);
1175 wake_up(&osb
->osb_mount_event
);
1177 /* Start this when the mount is almost sure of being successful */
1178 ocfs2_orphan_scan_start(osb
);
1189 atomic_set(&osb
->vol_state
, VOLUME_DISABLED
);
1190 wake_up(&osb
->osb_mount_event
);
1191 ocfs2_dismount_volume(sb
, 1);
1197 static struct dentry
*ocfs2_mount(struct file_system_type
*fs_type
,
1199 const char *dev_name
,
1202 return mount_bdev(fs_type
, flags
, dev_name
, data
, ocfs2_fill_super
);
1205 static struct file_system_type ocfs2_fs_type
= {
1206 .owner
= THIS_MODULE
,
1208 .mount
= ocfs2_mount
,
1209 .kill_sb
= kill_block_super
,
1210 .fs_flags
= FS_REQUIRES_DEV
|FS_RENAME_DOES_D_MOVE
,
1213 MODULE_ALIAS_FS("ocfs2");
1215 static int ocfs2_check_set_options(struct super_block
*sb
,
1216 struct mount_options
*options
)
1218 if (options
->mount_opt
& OCFS2_MOUNT_USRQUOTA
&&
1219 !OCFS2_HAS_RO_COMPAT_FEATURE(sb
,
1220 OCFS2_FEATURE_RO_COMPAT_USRQUOTA
)) {
1221 mlog(ML_ERROR
, "User quotas were requested, but this "
1222 "filesystem does not have the feature enabled.\n");
1225 if (options
->mount_opt
& OCFS2_MOUNT_GRPQUOTA
&&
1226 !OCFS2_HAS_RO_COMPAT_FEATURE(sb
,
1227 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
)) {
1228 mlog(ML_ERROR
, "Group quotas were requested, but this "
1229 "filesystem does not have the feature enabled.\n");
1232 if (options
->mount_opt
& OCFS2_MOUNT_POSIX_ACL
&&
1233 !OCFS2_HAS_INCOMPAT_FEATURE(sb
, OCFS2_FEATURE_INCOMPAT_XATTR
)) {
1234 mlog(ML_ERROR
, "ACL support requested but extended attributes "
1235 "feature is not enabled\n");
1238 /* No ACL setting specified? Use XATTR feature... */
1239 if (!(options
->mount_opt
& (OCFS2_MOUNT_POSIX_ACL
|
1240 OCFS2_MOUNT_NO_POSIX_ACL
))) {
1241 if (OCFS2_HAS_INCOMPAT_FEATURE(sb
, OCFS2_FEATURE_INCOMPAT_XATTR
))
1242 options
->mount_opt
|= OCFS2_MOUNT_POSIX_ACL
;
1244 options
->mount_opt
|= OCFS2_MOUNT_NO_POSIX_ACL
;
1249 static int ocfs2_parse_options(struct super_block
*sb
,
1251 struct mount_options
*mopt
,
1254 int status
, user_stack
= 0;
1258 substring_t args
[MAX_OPT_ARGS
];
1260 trace_ocfs2_parse_options(is_remount
, options
? options
: "(none)");
1262 mopt
->commit_interval
= 0;
1263 mopt
->mount_opt
= OCFS2_MOUNT_NOINTR
;
1264 mopt
->atime_quantum
= OCFS2_DEFAULT_ATIME_QUANTUM
;
1265 mopt
->slot
= OCFS2_INVALID_SLOT
;
1266 mopt
->localalloc_opt
= -1;
1267 mopt
->cluster_stack
[0] = '\0';
1268 mopt
->resv_level
= OCFS2_DEFAULT_RESV_LEVEL
;
1269 mopt
->dir_resv_level
= -1;
1276 while ((p
= strsep(&options
, ",")) != NULL
) {
1280 token
= match_token(p
, tokens
, args
);
1283 mopt
->mount_opt
|= OCFS2_MOUNT_HB_LOCAL
;
1286 mopt
->mount_opt
|= OCFS2_MOUNT_HB_NONE
;
1289 mopt
->mount_opt
|= OCFS2_MOUNT_HB_GLOBAL
;
1292 if (match_int(&args
[0], &option
)) {
1297 mopt
->mount_opt
|= OCFS2_MOUNT_BARRIER
;
1299 mopt
->mount_opt
&= ~OCFS2_MOUNT_BARRIER
;
1302 mopt
->mount_opt
&= ~OCFS2_MOUNT_NOINTR
;
1305 mopt
->mount_opt
|= OCFS2_MOUNT_NOINTR
;
1308 mopt
->mount_opt
&= ~OCFS2_MOUNT_ERRORS_CONT
;
1309 mopt
->mount_opt
&= ~OCFS2_MOUNT_ERRORS_ROFS
;
1310 mopt
->mount_opt
|= OCFS2_MOUNT_ERRORS_PANIC
;
1313 mopt
->mount_opt
&= ~OCFS2_MOUNT_ERRORS_CONT
;
1314 mopt
->mount_opt
&= ~OCFS2_MOUNT_ERRORS_PANIC
;
1315 mopt
->mount_opt
|= OCFS2_MOUNT_ERRORS_ROFS
;
1318 mopt
->mount_opt
&= ~OCFS2_MOUNT_ERRORS_ROFS
;
1319 mopt
->mount_opt
&= ~OCFS2_MOUNT_ERRORS_PANIC
;
1320 mopt
->mount_opt
|= OCFS2_MOUNT_ERRORS_CONT
;
1322 case Opt_data_ordered
:
1323 mopt
->mount_opt
&= ~OCFS2_MOUNT_DATA_WRITEBACK
;
1325 case Opt_data_writeback
:
1326 mopt
->mount_opt
|= OCFS2_MOUNT_DATA_WRITEBACK
;
1328 case Opt_user_xattr
:
1329 mopt
->mount_opt
&= ~OCFS2_MOUNT_NOUSERXATTR
;
1331 case Opt_nouser_xattr
:
1332 mopt
->mount_opt
|= OCFS2_MOUNT_NOUSERXATTR
;
1334 case Opt_atime_quantum
:
1335 if (match_int(&args
[0], &option
)) {
1340 mopt
->atime_quantum
= option
;
1343 if (match_int(&args
[0], &option
)) {
1348 mopt
->slot
= (u16
)option
;
1351 if (match_int(&args
[0], &option
)) {
1358 option
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1359 mopt
->commit_interval
= HZ
* option
;
1361 case Opt_localalloc
:
1362 if (match_int(&args
[0], &option
)) {
1367 mopt
->localalloc_opt
= option
;
1369 case Opt_localflocks
:
1371 * Changing this during remount could race
1372 * flock() requests, or "unbalance" existing
1373 * ones (e.g., a lock is taken in one mode but
1374 * dropped in the other). If users care enough
1375 * to flip locking modes during remount, we
1376 * could add a "local" flag to individual
1377 * flock structures for proper tracking of
1381 mopt
->mount_opt
|= OCFS2_MOUNT_LOCALFLOCKS
;
1384 /* Check both that the option we were passed
1385 * is of the right length and that it is a proper
1386 * string of the right length.
1388 if (((args
[0].to
- args
[0].from
) !=
1389 OCFS2_STACK_LABEL_LEN
) ||
1390 (strnlen(args
[0].from
,
1391 OCFS2_STACK_LABEL_LEN
) !=
1392 OCFS2_STACK_LABEL_LEN
)) {
1394 "Invalid cluster_stack option\n");
1398 memcpy(mopt
->cluster_stack
, args
[0].from
,
1399 OCFS2_STACK_LABEL_LEN
);
1400 mopt
->cluster_stack
[OCFS2_STACK_LABEL_LEN
] = '\0';
1402 * Open code the memcmp here as we don't have
1404 * ocfs2_userspace_stack().
1406 if (memcmp(mopt
->cluster_stack
,
1407 OCFS2_CLASSIC_CLUSTER_STACK
,
1408 OCFS2_STACK_LABEL_LEN
))
1412 mopt
->mount_opt
|= OCFS2_MOUNT_INODE64
;
1415 mopt
->mount_opt
|= OCFS2_MOUNT_USRQUOTA
;
1418 mopt
->mount_opt
|= OCFS2_MOUNT_GRPQUOTA
;
1420 case Opt_coherency_buffered
:
1421 mopt
->mount_opt
|= OCFS2_MOUNT_COHERENCY_BUFFERED
;
1423 case Opt_coherency_full
:
1424 mopt
->mount_opt
&= ~OCFS2_MOUNT_COHERENCY_BUFFERED
;
1427 mopt
->mount_opt
|= OCFS2_MOUNT_POSIX_ACL
;
1428 mopt
->mount_opt
&= ~OCFS2_MOUNT_NO_POSIX_ACL
;
1431 mopt
->mount_opt
|= OCFS2_MOUNT_NO_POSIX_ACL
;
1432 mopt
->mount_opt
&= ~OCFS2_MOUNT_POSIX_ACL
;
1434 case Opt_resv_level
:
1437 if (match_int(&args
[0], &option
)) {
1441 if (option
>= OCFS2_MIN_RESV_LEVEL
&&
1442 option
< OCFS2_MAX_RESV_LEVEL
)
1443 mopt
->resv_level
= option
;
1445 case Opt_dir_resv_level
:
1448 if (match_int(&args
[0], &option
)) {
1452 if (option
>= OCFS2_MIN_RESV_LEVEL
&&
1453 option
< OCFS2_MAX_RESV_LEVEL
)
1454 mopt
->dir_resv_level
= option
;
1456 case Opt_journal_async_commit
:
1457 mopt
->mount_opt
|= OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT
;
1460 mopt
->mount_opt
|= OCFS2_MOUNT_NOCLUSTER
;
1464 "Unrecognized mount option \"%s\" "
1465 "or missing value\n", p
);
1471 if (user_stack
== 0) {
1472 /* Ensure only one heartbeat mode */
1473 tmp
= mopt
->mount_opt
& (OCFS2_MOUNT_HB_LOCAL
|
1474 OCFS2_MOUNT_HB_GLOBAL
|
1475 OCFS2_MOUNT_HB_NONE
);
1476 if (hweight32(tmp
) != 1) {
1477 mlog(ML_ERROR
, "Invalid heartbeat mount options\n");
1489 static int ocfs2_show_options(struct seq_file
*s
, struct dentry
*root
)
1491 struct ocfs2_super
*osb
= OCFS2_SB(root
->d_sb
);
1492 unsigned long opts
= osb
->s_mount_opt
;
1493 unsigned int local_alloc_megs
;
1495 if (opts
& (OCFS2_MOUNT_HB_LOCAL
| OCFS2_MOUNT_HB_GLOBAL
)) {
1496 seq_printf(s
, ",_netdev");
1497 if (opts
& OCFS2_MOUNT_HB_LOCAL
)
1498 seq_printf(s
, ",%s", OCFS2_HB_LOCAL
);
1500 seq_printf(s
, ",%s", OCFS2_HB_GLOBAL
);
1502 seq_printf(s
, ",%s", OCFS2_HB_NONE
);
1504 if (opts
& OCFS2_MOUNT_NOINTR
)
1505 seq_printf(s
, ",nointr");
1507 if (opts
& OCFS2_MOUNT_DATA_WRITEBACK
)
1508 seq_printf(s
, ",data=writeback");
1510 seq_printf(s
, ",data=ordered");
1512 if (opts
& OCFS2_MOUNT_BARRIER
)
1513 seq_printf(s
, ",barrier=1");
1515 if (opts
& OCFS2_MOUNT_ERRORS_PANIC
)
1516 seq_printf(s
, ",errors=panic");
1517 else if (opts
& OCFS2_MOUNT_ERRORS_CONT
)
1518 seq_printf(s
, ",errors=continue");
1520 seq_printf(s
, ",errors=remount-ro");
1522 if (osb
->preferred_slot
!= OCFS2_INVALID_SLOT
)
1523 seq_printf(s
, ",preferred_slot=%d", osb
->preferred_slot
);
1525 seq_printf(s
, ",atime_quantum=%u", osb
->s_atime_quantum
);
1527 if (osb
->osb_commit_interval
)
1528 seq_printf(s
, ",commit=%u",
1529 (unsigned) (osb
->osb_commit_interval
/ HZ
));
1531 local_alloc_megs
= osb
->local_alloc_bits
>> (20 - osb
->s_clustersize_bits
);
1532 if (local_alloc_megs
!= ocfs2_la_default_mb(osb
))
1533 seq_printf(s
, ",localalloc=%d", local_alloc_megs
);
1535 if (opts
& OCFS2_MOUNT_LOCALFLOCKS
)
1536 seq_printf(s
, ",localflocks,");
1538 if (osb
->osb_cluster_stack
[0])
1539 seq_show_option_n(s
, "cluster_stack", osb
->osb_cluster_stack
,
1540 OCFS2_STACK_LABEL_LEN
);
1541 if (opts
& OCFS2_MOUNT_USRQUOTA
)
1542 seq_printf(s
, ",usrquota");
1543 if (opts
& OCFS2_MOUNT_GRPQUOTA
)
1544 seq_printf(s
, ",grpquota");
1546 if (opts
& OCFS2_MOUNT_COHERENCY_BUFFERED
)
1547 seq_printf(s
, ",coherency=buffered");
1549 seq_printf(s
, ",coherency=full");
1551 if (opts
& OCFS2_MOUNT_NOUSERXATTR
)
1552 seq_printf(s
, ",nouser_xattr");
1554 seq_printf(s
, ",user_xattr");
1556 if (opts
& OCFS2_MOUNT_INODE64
)
1557 seq_printf(s
, ",inode64");
1559 if (opts
& OCFS2_MOUNT_POSIX_ACL
)
1560 seq_printf(s
, ",acl");
1562 seq_printf(s
, ",noacl");
1564 if (osb
->osb_resv_level
!= OCFS2_DEFAULT_RESV_LEVEL
)
1565 seq_printf(s
, ",resv_level=%d", osb
->osb_resv_level
);
1567 if (osb
->osb_dir_resv_level
!= osb
->osb_resv_level
)
1568 seq_printf(s
, ",dir_resv_level=%d", osb
->osb_resv_level
);
1570 if (opts
& OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT
)
1571 seq_printf(s
, ",journal_async_commit");
1573 if (opts
& OCFS2_MOUNT_NOCLUSTER
)
1574 seq_printf(s
, ",nocluster");
1579 static int __init
ocfs2_init(void)
1583 status
= init_ocfs2_uptodate_cache();
1587 status
= ocfs2_initialize_mem_caches();
1591 ocfs2_debugfs_root
= debugfs_create_dir("ocfs2", NULL
);
1593 ocfs2_set_locking_protocol();
1595 status
= register_quota_format(&ocfs2_quota_format
);
1598 status
= register_filesystem(&ocfs2_fs_type
);
1602 unregister_quota_format(&ocfs2_quota_format
);
1604 debugfs_remove(ocfs2_debugfs_root
);
1605 ocfs2_free_mem_caches();
1607 exit_ocfs2_uptodate_cache();
1613 static void __exit
ocfs2_exit(void)
1615 unregister_quota_format(&ocfs2_quota_format
);
1617 debugfs_remove(ocfs2_debugfs_root
);
1619 ocfs2_free_mem_caches();
1621 unregister_filesystem(&ocfs2_fs_type
);
1623 exit_ocfs2_uptodate_cache();
1626 static void ocfs2_put_super(struct super_block
*sb
)
1628 trace_ocfs2_put_super(sb
);
1630 ocfs2_sync_blockdev(sb
);
1631 ocfs2_dismount_volume(sb
, 0);
1634 static int ocfs2_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
1636 struct ocfs2_super
*osb
;
1637 u32 numbits
, freebits
;
1639 struct ocfs2_dinode
*bm_lock
;
1640 struct buffer_head
*bh
= NULL
;
1641 struct inode
*inode
= NULL
;
1643 trace_ocfs2_statfs(dentry
->d_sb
, buf
);
1645 osb
= OCFS2_SB(dentry
->d_sb
);
1647 inode
= ocfs2_get_system_file_inode(osb
,
1648 GLOBAL_BITMAP_SYSTEM_INODE
,
1649 OCFS2_INVALID_SLOT
);
1651 mlog(ML_ERROR
, "failed to get bitmap inode\n");
1656 status
= ocfs2_inode_lock(inode
, &bh
, 0);
1662 bm_lock
= (struct ocfs2_dinode
*) bh
->b_data
;
1664 numbits
= le32_to_cpu(bm_lock
->id1
.bitmap1
.i_total
);
1665 freebits
= numbits
- le32_to_cpu(bm_lock
->id1
.bitmap1
.i_used
);
1667 buf
->f_type
= OCFS2_SUPER_MAGIC
;
1668 buf
->f_bsize
= dentry
->d_sb
->s_blocksize
;
1669 buf
->f_namelen
= OCFS2_MAX_FILENAME_LEN
;
1670 buf
->f_blocks
= ((sector_t
) numbits
) *
1671 (osb
->s_clustersize
>> osb
->sb
->s_blocksize_bits
);
1672 buf
->f_bfree
= ((sector_t
) freebits
) *
1673 (osb
->s_clustersize
>> osb
->sb
->s_blocksize_bits
);
1674 buf
->f_bavail
= buf
->f_bfree
;
1675 buf
->f_files
= numbits
;
1676 buf
->f_ffree
= freebits
;
1677 buf
->f_fsid
.val
[0] = crc32_le(0, osb
->uuid_str
, OCFS2_VOL_UUID_LEN
)
1679 buf
->f_fsid
.val
[1] = crc32_le(0, osb
->uuid_str
+ OCFS2_VOL_UUID_LEN
,
1680 OCFS2_VOL_UUID_LEN
) & 0xFFFFFFFFUL
;
1684 ocfs2_inode_unlock(inode
, 0);
1695 static void ocfs2_inode_init_once(void *data
)
1697 struct ocfs2_inode_info
*oi
= data
;
1700 oi
->ip_open_count
= 0;
1701 spin_lock_init(&oi
->ip_lock
);
1702 ocfs2_extent_map_init(&oi
->vfs_inode
);
1703 INIT_LIST_HEAD(&oi
->ip_io_markers
);
1704 INIT_LIST_HEAD(&oi
->ip_unwritten_list
);
1705 oi
->ip_dir_start_lookup
= 0;
1706 init_rwsem(&oi
->ip_alloc_sem
);
1707 init_rwsem(&oi
->ip_xattr_sem
);
1708 mutex_init(&oi
->ip_io_mutex
);
1710 oi
->ip_blkno
= 0ULL;
1711 oi
->ip_clusters
= 0;
1712 oi
->ip_next_orphan
= NULL
;
1714 ocfs2_resv_init_once(&oi
->ip_la_data_resv
);
1716 ocfs2_lock_res_init_once(&oi
->ip_rw_lockres
);
1717 ocfs2_lock_res_init_once(&oi
->ip_inode_lockres
);
1718 ocfs2_lock_res_init_once(&oi
->ip_open_lockres
);
1720 ocfs2_metadata_cache_init(INODE_CACHE(&oi
->vfs_inode
),
1721 &ocfs2_inode_caching_ops
);
1723 inode_init_once(&oi
->vfs_inode
);
1726 static int ocfs2_initialize_mem_caches(void)
1728 ocfs2_inode_cachep
= kmem_cache_create("ocfs2_inode_cache",
1729 sizeof(struct ocfs2_inode_info
),
1731 (SLAB_HWCACHE_ALIGN
|SLAB_RECLAIM_ACCOUNT
|
1732 SLAB_MEM_SPREAD
|SLAB_ACCOUNT
),
1733 ocfs2_inode_init_once
);
1734 ocfs2_dquot_cachep
= kmem_cache_create("ocfs2_dquot_cache",
1735 sizeof(struct ocfs2_dquot
),
1737 (SLAB_HWCACHE_ALIGN
|SLAB_RECLAIM_ACCOUNT
|
1740 ocfs2_qf_chunk_cachep
= kmem_cache_create("ocfs2_qf_chunk_cache",
1741 sizeof(struct ocfs2_quota_chunk
),
1743 (SLAB_RECLAIM_ACCOUNT
|SLAB_MEM_SPREAD
),
1745 if (!ocfs2_inode_cachep
|| !ocfs2_dquot_cachep
||
1746 !ocfs2_qf_chunk_cachep
) {
1747 kmem_cache_destroy(ocfs2_inode_cachep
);
1748 kmem_cache_destroy(ocfs2_dquot_cachep
);
1749 kmem_cache_destroy(ocfs2_qf_chunk_cachep
);
1756 static void ocfs2_free_mem_caches(void)
1759 * Make sure all delayed rcu free inodes are flushed before we
1763 kmem_cache_destroy(ocfs2_inode_cachep
);
1764 ocfs2_inode_cachep
= NULL
;
1766 kmem_cache_destroy(ocfs2_dquot_cachep
);
1767 ocfs2_dquot_cachep
= NULL
;
1769 kmem_cache_destroy(ocfs2_qf_chunk_cachep
);
1770 ocfs2_qf_chunk_cachep
= NULL
;
1773 static int ocfs2_get_sector(struct super_block
*sb
,
1774 struct buffer_head
**bh
,
1778 if (!sb_set_blocksize(sb
, sect_size
)) {
1779 mlog(ML_ERROR
, "unable to set blocksize\n");
1783 *bh
= sb_getblk(sb
, block
);
1785 mlog_errno(-ENOMEM
);
1789 if (!buffer_dirty(*bh
))
1790 clear_buffer_uptodate(*bh
);
1792 ll_rw_block(REQ_OP_READ
, 0, 1, bh
);
1793 wait_on_buffer(*bh
);
1794 if (!buffer_uptodate(*bh
)) {
1804 static int ocfs2_mount_volume(struct super_block
*sb
)
1807 int unlock_super
= 0;
1808 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
1810 if (ocfs2_is_hard_readonly(osb
))
1813 mutex_init(&osb
->obs_trim_fs_mutex
);
1815 status
= ocfs2_dlm_init(osb
);
1818 if (status
== -EBADR
&& ocfs2_userspace_stack(osb
))
1819 mlog(ML_ERROR
, "couldn't mount because cluster name on"
1820 " disk does not match the running cluster name.\n");
1824 status
= ocfs2_super_lock(osb
, 1);
1831 /* This will load up the node map and add ourselves to it. */
1832 status
= ocfs2_find_slot(osb
);
1838 /* load all node-local system inodes */
1839 status
= ocfs2_init_local_system_inodes(osb
);
1845 status
= ocfs2_check_volume(osb
);
1851 status
= ocfs2_truncate_log_init(osb
);
1857 ocfs2_super_unlock(osb
, 1);
1862 static void ocfs2_dismount_volume(struct super_block
*sb
, int mnt_err
)
1864 int tmp
, hangup_needed
= 0;
1865 struct ocfs2_super
*osb
= NULL
;
1868 trace_ocfs2_dismount_volume(sb
);
1874 /* Remove file check sysfs related directores/files,
1875 * and wait for the pending file check operations */
1876 ocfs2_filecheck_remove_sysfs(osb
);
1878 kset_unregister(osb
->osb_dev_kset
);
1880 /* Orphan scan should be stopped as early as possible */
1881 ocfs2_orphan_scan_stop(osb
);
1883 ocfs2_disable_quotas(osb
);
1885 /* All dquots should be freed by now */
1886 WARN_ON(!llist_empty(&osb
->dquot_drop_list
));
1887 /* Wait for worker to be done with the work structure in osb */
1888 cancel_work_sync(&osb
->dquot_drop_work
);
1890 ocfs2_shutdown_local_alloc(osb
);
1892 ocfs2_truncate_log_shutdown(osb
);
1894 /* This will disable recovery and flush any recovery work. */
1895 ocfs2_recovery_exit(osb
);
1897 ocfs2_journal_shutdown(osb
);
1899 ocfs2_sync_blockdev(sb
);
1901 ocfs2_purge_refcount_trees(osb
);
1903 /* No cluster connection means we've failed during mount, so skip
1904 * all the steps which depended on that to complete. */
1906 tmp
= ocfs2_super_lock(osb
, 1);
1913 if (osb
->slot_num
!= OCFS2_INVALID_SLOT
)
1914 ocfs2_put_slot(osb
);
1917 ocfs2_super_unlock(osb
, 1);
1919 ocfs2_release_system_inodes(osb
);
1922 * If we're dismounting due to mount error, mount.ocfs2 will clean
1923 * up heartbeat. If we're a local mount, there is no heartbeat.
1924 * If we failed before we got a uuid_str yet, we can't stop
1925 * heartbeat. Otherwise, do it.
1927 if (!mnt_err
&& !ocfs2_mount_local(osb
) && osb
->uuid_str
&&
1928 !ocfs2_is_hard_readonly(osb
))
1932 ocfs2_dlm_shutdown(osb
, hangup_needed
);
1934 ocfs2_blockcheck_stats_debugfs_remove(&osb
->osb_ecc_stats
);
1935 debugfs_remove_recursive(osb
->osb_debug_root
);
1938 ocfs2_cluster_hangup(osb
->uuid_str
, strlen(osb
->uuid_str
));
1940 atomic_set(&osb
->vol_state
, VOLUME_DISMOUNTED
);
1942 if (ocfs2_mount_local(osb
))
1943 snprintf(nodestr
, sizeof(nodestr
), "local");
1945 snprintf(nodestr
, sizeof(nodestr
), "%u", osb
->node_num
);
1947 printk(KERN_INFO
"ocfs2: Unmounting device (%s) on (node %s)\n",
1948 osb
->dev_str
, nodestr
);
1950 ocfs2_delete_osb(osb
);
1953 sb
->s_fs_info
= NULL
;
1956 static int ocfs2_setup_osb_uuid(struct ocfs2_super
*osb
, const unsigned char *uuid
,
1957 unsigned uuid_bytes
)
1962 BUG_ON(uuid_bytes
!= OCFS2_VOL_UUID_LEN
);
1964 osb
->uuid_str
= kzalloc(OCFS2_VOL_UUID_LEN
* 2 + 1, GFP_KERNEL
);
1965 if (osb
->uuid_str
== NULL
)
1968 for (i
= 0, ptr
= osb
->uuid_str
; i
< OCFS2_VOL_UUID_LEN
; i
++) {
1969 /* print with null */
1970 ret
= snprintf(ptr
, 3, "%02X", uuid
[i
]);
1971 if (ret
!= 2) /* drop super cleans up */
1973 /* then only advance past the last char */
1980 /* Make sure entire volume is addressable by our journal. Requires
1981 osb_clusters_at_boot to be valid and for the journal to have been
1982 initialized by ocfs2_journal_init(). */
1983 static int ocfs2_journal_addressable(struct ocfs2_super
*osb
)
1987 ocfs2_clusters_to_blocks(osb
->sb
,
1988 osb
->osb_clusters_at_boot
) - 1;
1990 /* 32-bit block number is always OK. */
1991 if (max_block
<= (u32
)~0ULL)
1994 /* Volume is "huge", so see if our journal is new enough to
1996 if (!(OCFS2_HAS_COMPAT_FEATURE(osb
->sb
,
1997 OCFS2_FEATURE_COMPAT_JBD2_SB
) &&
1998 jbd2_journal_check_used_features(osb
->journal
->j_journal
, 0, 0,
1999 JBD2_FEATURE_INCOMPAT_64BIT
))) {
2000 mlog(ML_ERROR
, "The journal cannot address the entire volume. "
2001 "Enable the 'block64' journal option with tunefs.ocfs2");
2010 static int ocfs2_initialize_super(struct super_block
*sb
,
2011 struct buffer_head
*bh
,
2013 struct ocfs2_blockcheck_stats
*stats
)
2016 int i
, cbits
, bbits
;
2017 struct ocfs2_dinode
*di
= (struct ocfs2_dinode
*)bh
->b_data
;
2018 struct inode
*inode
= NULL
;
2019 struct ocfs2_journal
*journal
;
2020 struct ocfs2_super
*osb
;
2023 osb
= kzalloc(sizeof(struct ocfs2_super
), GFP_KERNEL
);
2030 sb
->s_fs_info
= osb
;
2031 sb
->s_op
= &ocfs2_sops
;
2032 sb
->s_d_op
= &ocfs2_dentry_ops
;
2033 sb
->s_export_op
= &ocfs2_export_ops
;
2034 sb
->s_qcop
= &dquot_quotactl_sysfile_ops
;
2035 sb
->dq_op
= &ocfs2_quota_operations
;
2036 sb
->s_quota_types
= QTYPE_MASK_USR
| QTYPE_MASK_GRP
;
2037 sb
->s_xattr
= ocfs2_xattr_handlers
;
2038 sb
->s_time_gran
= 1;
2039 sb
->s_flags
|= SB_NOATIME
;
2040 /* this is needed to support O_LARGEFILE */
2041 cbits
= le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
);
2042 bbits
= le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
);
2043 sb
->s_maxbytes
= ocfs2_max_file_offset(bbits
, cbits
);
2044 memcpy(&sb
->s_uuid
, di
->id2
.i_super
.s_uuid
,
2045 sizeof(di
->id2
.i_super
.s_uuid
));
2047 osb
->osb_dx_mask
= (1 << (cbits
- bbits
)) - 1;
2049 for (i
= 0; i
< 3; i
++)
2050 osb
->osb_dx_seed
[i
] = le32_to_cpu(di
->id2
.i_super
.s_dx_seed
[i
]);
2051 osb
->osb_dx_seed
[3] = le32_to_cpu(di
->id2
.i_super
.s_uuid_hash
);
2054 osb
->s_sectsize_bits
= blksize_bits(sector_size
);
2055 BUG_ON(!osb
->s_sectsize_bits
);
2057 spin_lock_init(&osb
->dc_task_lock
);
2058 init_waitqueue_head(&osb
->dc_event
);
2059 osb
->dc_work_sequence
= 0;
2060 osb
->dc_wake_sequence
= 0;
2061 INIT_LIST_HEAD(&osb
->blocked_lock_list
);
2062 osb
->blocked_lock_count
= 0;
2063 spin_lock_init(&osb
->osb_lock
);
2064 spin_lock_init(&osb
->osb_xattr_lock
);
2065 ocfs2_init_steal_slots(osb
);
2067 mutex_init(&osb
->system_file_mutex
);
2069 atomic_set(&osb
->alloc_stats
.moves
, 0);
2070 atomic_set(&osb
->alloc_stats
.local_data
, 0);
2071 atomic_set(&osb
->alloc_stats
.bitmap_data
, 0);
2072 atomic_set(&osb
->alloc_stats
.bg_allocs
, 0);
2073 atomic_set(&osb
->alloc_stats
.bg_extends
, 0);
2075 /* Copy the blockcheck stats from the superblock probe */
2076 osb
->osb_ecc_stats
= *stats
;
2078 ocfs2_init_node_maps(osb
);
2080 snprintf(osb
->dev_str
, sizeof(osb
->dev_str
), "%u,%u",
2081 MAJOR(osb
->sb
->s_dev
), MINOR(osb
->sb
->s_dev
));
2083 osb
->max_slots
= le16_to_cpu(di
->id2
.i_super
.s_max_slots
);
2084 if (osb
->max_slots
> OCFS2_MAX_SLOTS
|| osb
->max_slots
== 0) {
2085 mlog(ML_ERROR
, "Invalid number of node slots (%u)\n",
2091 ocfs2_orphan_scan_init(osb
);
2093 status
= ocfs2_recovery_init(osb
);
2095 mlog(ML_ERROR
, "Unable to initialize recovery state\n");
2100 init_waitqueue_head(&osb
->checkpoint_event
);
2102 osb
->s_atime_quantum
= OCFS2_DEFAULT_ATIME_QUANTUM
;
2104 osb
->slot_num
= OCFS2_INVALID_SLOT
;
2106 osb
->s_xattr_inline_size
= le16_to_cpu(
2107 di
->id2
.i_super
.s_xattr_inline_size
);
2109 osb
->local_alloc_state
= OCFS2_LA_UNUSED
;
2110 osb
->local_alloc_bh
= NULL
;
2111 INIT_DELAYED_WORK(&osb
->la_enable_wq
, ocfs2_la_enable_worker
);
2113 init_waitqueue_head(&osb
->osb_mount_event
);
2115 status
= ocfs2_resmap_init(osb
, &osb
->osb_la_resmap
);
2121 osb
->vol_label
= kmalloc(OCFS2_MAX_VOL_LABEL_LEN
, GFP_KERNEL
);
2122 if (!osb
->vol_label
) {
2123 mlog(ML_ERROR
, "unable to alloc vol label\n");
2128 osb
->slot_recovery_generations
=
2129 kcalloc(osb
->max_slots
, sizeof(*osb
->slot_recovery_generations
),
2131 if (!osb
->slot_recovery_generations
) {
2137 init_waitqueue_head(&osb
->osb_wipe_event
);
2138 osb
->osb_orphan_wipes
= kcalloc(osb
->max_slots
,
2139 sizeof(*osb
->osb_orphan_wipes
),
2141 if (!osb
->osb_orphan_wipes
) {
2147 osb
->osb_rf_lock_tree
= RB_ROOT
;
2149 osb
->s_feature_compat
=
2150 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_compat
);
2151 osb
->s_feature_ro_compat
=
2152 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_ro_compat
);
2153 osb
->s_feature_incompat
=
2154 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_incompat
);
2156 if ((i
= OCFS2_HAS_INCOMPAT_FEATURE(osb
->sb
, ~OCFS2_FEATURE_INCOMPAT_SUPP
))) {
2157 mlog(ML_ERROR
, "couldn't mount because of unsupported "
2158 "optional features (%x).\n", i
);
2162 if (!sb_rdonly(osb
->sb
) && (i
= OCFS2_HAS_RO_COMPAT_FEATURE(osb
->sb
, ~OCFS2_FEATURE_RO_COMPAT_SUPP
))) {
2163 mlog(ML_ERROR
, "couldn't mount RDWR because of "
2164 "unsupported optional features (%x).\n", i
);
2169 if (ocfs2_clusterinfo_valid(osb
)) {
2171 * ci_stack and ci_cluster in ocfs2_cluster_info may not be null
2172 * terminated, so make sure no overflow happens here by using
2173 * memcpy. Destination strings will always be null terminated
2174 * because osb is allocated using kzalloc.
2176 osb
->osb_stackflags
=
2177 OCFS2_RAW_SB(di
)->s_cluster_info
.ci_stackflags
;
2178 memcpy(osb
->osb_cluster_stack
,
2179 OCFS2_RAW_SB(di
)->s_cluster_info
.ci_stack
,
2180 OCFS2_STACK_LABEL_LEN
);
2181 if (strlen(osb
->osb_cluster_stack
) != OCFS2_STACK_LABEL_LEN
) {
2183 "couldn't mount because of an invalid "
2184 "cluster stack label (%s) \n",
2185 osb
->osb_cluster_stack
);
2189 memcpy(osb
->osb_cluster_name
,
2190 OCFS2_RAW_SB(di
)->s_cluster_info
.ci_cluster
,
2191 OCFS2_CLUSTER_NAME_LEN
);
2193 /* The empty string is identical with classic tools that
2194 * don't know about s_cluster_info. */
2195 osb
->osb_cluster_stack
[0] = '\0';
2198 get_random_bytes(&osb
->s_next_generation
, sizeof(u32
));
2201 * This should be done in ocfs2_journal_init(), but unknown
2202 * ordering issues will cause the filesystem to crash.
2203 * If anyone wants to figure out what part of the code
2204 * refers to osb->journal before ocfs2_journal_init() is run,
2207 /* initialize our journal structure */
2209 journal
= kzalloc(sizeof(struct ocfs2_journal
), GFP_KERNEL
);
2211 mlog(ML_ERROR
, "unable to alloc journal\n");
2215 osb
->journal
= journal
;
2216 journal
->j_osb
= osb
;
2218 atomic_set(&journal
->j_num_trans
, 0);
2219 init_rwsem(&journal
->j_trans_barrier
);
2220 init_waitqueue_head(&journal
->j_checkpointed
);
2221 spin_lock_init(&journal
->j_lock
);
2222 journal
->j_trans_id
= (unsigned long) 1;
2223 INIT_LIST_HEAD(&journal
->j_la_cleanups
);
2224 INIT_WORK(&journal
->j_recovery_work
, ocfs2_complete_recovery
);
2225 journal
->j_state
= OCFS2_JOURNAL_FREE
;
2227 INIT_WORK(&osb
->dquot_drop_work
, ocfs2_drop_dquot_refs
);
2228 init_llist_head(&osb
->dquot_drop_list
);
2230 /* get some pseudo constants for clustersize bits */
2231 osb
->s_clustersize_bits
=
2232 le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
);
2233 osb
->s_clustersize
= 1 << osb
->s_clustersize_bits
;
2235 if (osb
->s_clustersize
< OCFS2_MIN_CLUSTERSIZE
||
2236 osb
->s_clustersize
> OCFS2_MAX_CLUSTERSIZE
) {
2237 mlog(ML_ERROR
, "Volume has invalid cluster size (%d)\n",
2238 osb
->s_clustersize
);
2243 total_blocks
= ocfs2_clusters_to_blocks(osb
->sb
,
2244 le32_to_cpu(di
->i_clusters
));
2246 status
= generic_check_addressable(osb
->sb
->s_blocksize_bits
,
2249 mlog(ML_ERROR
, "Volume too large "
2250 "to mount safely on this system");
2255 if (ocfs2_setup_osb_uuid(osb
, di
->id2
.i_super
.s_uuid
,
2256 sizeof(di
->id2
.i_super
.s_uuid
))) {
2257 mlog(ML_ERROR
, "Out of memory trying to setup our uuid.\n");
2262 strlcpy(osb
->vol_label
, di
->id2
.i_super
.s_label
,
2263 OCFS2_MAX_VOL_LABEL_LEN
);
2264 osb
->root_blkno
= le64_to_cpu(di
->id2
.i_super
.s_root_blkno
);
2265 osb
->system_dir_blkno
= le64_to_cpu(di
->id2
.i_super
.s_system_dir_blkno
);
2266 osb
->first_cluster_group_blkno
=
2267 le64_to_cpu(di
->id2
.i_super
.s_first_cluster_group
);
2268 osb
->fs_generation
= le32_to_cpu(di
->i_fs_generation
);
2269 osb
->uuid_hash
= le32_to_cpu(di
->id2
.i_super
.s_uuid_hash
);
2270 trace_ocfs2_initialize_super(osb
->vol_label
, osb
->uuid_str
,
2271 (unsigned long long)osb
->root_blkno
,
2272 (unsigned long long)osb
->system_dir_blkno
,
2273 osb
->s_clustersize_bits
);
2275 osb
->osb_dlm_debug
= ocfs2_new_dlm_debug();
2276 if (!osb
->osb_dlm_debug
) {
2282 atomic_set(&osb
->vol_state
, VOLUME_INIT
);
2284 /* load root, system_dir, and all global system inodes */
2285 status
= ocfs2_init_global_system_inodes(osb
);
2294 inode
= ocfs2_get_system_file_inode(osb
, GLOBAL_BITMAP_SYSTEM_INODE
,
2295 OCFS2_INVALID_SLOT
);
2302 osb
->bitmap_blkno
= OCFS2_I(inode
)->ip_blkno
;
2303 osb
->osb_clusters_at_boot
= OCFS2_I(inode
)->ip_clusters
;
2306 osb
->bitmap_cpg
= ocfs2_group_bitmap_size(sb
, 0,
2307 osb
->s_feature_incompat
) * 8;
2309 status
= ocfs2_init_slot_info(osb
);
2314 cleancache_init_shared_fs(sb
);
2316 osb
->ocfs2_wq
= alloc_ordered_workqueue("ocfs2_wq", WQ_MEM_RECLAIM
);
2317 if (!osb
->ocfs2_wq
) {
2327 * will return: -EAGAIN if it is ok to keep searching for superblocks
2328 * -EINVAL if there is a bad superblock
2331 static int ocfs2_verify_volume(struct ocfs2_dinode
*di
,
2332 struct buffer_head
*bh
,
2334 struct ocfs2_blockcheck_stats
*stats
)
2336 int status
= -EAGAIN
;
2338 if (memcmp(di
->i_signature
, OCFS2_SUPER_BLOCK_SIGNATURE
,
2339 strlen(OCFS2_SUPER_BLOCK_SIGNATURE
)) == 0) {
2340 /* We have to do a raw check of the feature here */
2341 if (le32_to_cpu(di
->id2
.i_super
.s_feature_incompat
) &
2342 OCFS2_FEATURE_INCOMPAT_META_ECC
) {
2343 status
= ocfs2_block_check_validate(bh
->b_data
,
2351 if ((1 << le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
)) != blksz
) {
2352 mlog(ML_ERROR
, "found superblock with incorrect block "
2353 "size: found %u, should be %u\n",
2354 1 << le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
),
2356 } else if (le16_to_cpu(di
->id2
.i_super
.s_major_rev_level
) !=
2357 OCFS2_MAJOR_REV_LEVEL
||
2358 le16_to_cpu(di
->id2
.i_super
.s_minor_rev_level
) !=
2359 OCFS2_MINOR_REV_LEVEL
) {
2360 mlog(ML_ERROR
, "found superblock with bad version: "
2361 "found %u.%u, should be %u.%u\n",
2362 le16_to_cpu(di
->id2
.i_super
.s_major_rev_level
),
2363 le16_to_cpu(di
->id2
.i_super
.s_minor_rev_level
),
2364 OCFS2_MAJOR_REV_LEVEL
,
2365 OCFS2_MINOR_REV_LEVEL
);
2366 } else if (bh
->b_blocknr
!= le64_to_cpu(di
->i_blkno
)) {
2367 mlog(ML_ERROR
, "bad block number on superblock: "
2368 "found %llu, should be %llu\n",
2369 (unsigned long long)le64_to_cpu(di
->i_blkno
),
2370 (unsigned long long)bh
->b_blocknr
);
2371 } else if (le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
) < 12 ||
2372 le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
) > 20) {
2373 mlog(ML_ERROR
, "bad cluster size found: %u\n",
2374 1 << le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
));
2375 } else if (!le64_to_cpu(di
->id2
.i_super
.s_root_blkno
)) {
2376 mlog(ML_ERROR
, "bad root_blkno: 0\n");
2377 } else if (!le64_to_cpu(di
->id2
.i_super
.s_system_dir_blkno
)) {
2378 mlog(ML_ERROR
, "bad system_dir_blkno: 0\n");
2379 } else if (le16_to_cpu(di
->id2
.i_super
.s_max_slots
) > OCFS2_MAX_SLOTS
) {
2381 "Superblock slots found greater than file system "
2382 "maximum: found %u, max %u\n",
2383 le16_to_cpu(di
->id2
.i_super
.s_max_slots
),
2392 if (status
&& status
!= -EAGAIN
)
2397 static int ocfs2_check_volume(struct ocfs2_super
*osb
)
2402 struct ocfs2_dinode
*local_alloc
= NULL
; /* only used if we
2406 /* Init our journal object. */
2407 status
= ocfs2_journal_init(osb
->journal
, &dirty
);
2409 mlog(ML_ERROR
, "Could not initialize journal!\n");
2413 /* Now that journal has been initialized, check to make sure
2414 entire volume is addressable. */
2415 status
= ocfs2_journal_addressable(osb
);
2419 /* If the journal was unmounted cleanly then we don't want to
2420 * recover anything. Otherwise, journal_load will do that
2421 * dirty work for us :) */
2423 status
= ocfs2_journal_wipe(osb
->journal
, 0);
2429 printk(KERN_NOTICE
"ocfs2: File system on device (%s) was not "
2430 "unmounted cleanly, recovering it.\n", osb
->dev_str
);
2433 local
= ocfs2_mount_local(osb
);
2435 /* will play back anything left in the journal. */
2436 status
= ocfs2_journal_load(osb
->journal
, local
, dirty
);
2438 mlog(ML_ERROR
, "ocfs2 journal load failed! %d\n", status
);
2442 if (osb
->s_mount_opt
& OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT
)
2443 jbd2_journal_set_features(osb
->journal
->j_journal
,
2444 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2445 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2447 jbd2_journal_clear_features(osb
->journal
->j_journal
,
2448 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2449 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2452 /* recover my local alloc if we didn't unmount cleanly. */
2453 status
= ocfs2_begin_local_alloc_recovery(osb
,
2460 /* we complete the recovery process after we've marked
2461 * ourselves as mounted. */
2464 status
= ocfs2_load_local_alloc(osb
);
2471 /* Recovery will be completed after we've mounted the
2472 * rest of the volume. */
2473 osb
->local_alloc_copy
= local_alloc
;
2477 /* go through each journal, trylock it and if you get the
2478 * lock, and it's marked as dirty, set the bit in the recover
2479 * map and launch a recovery thread for it. */
2480 status
= ocfs2_mark_dead_nodes(osb
);
2486 status
= ocfs2_compute_replay_slots(osb
);
2499 * The routine gets called from dismount or close whenever a dismount on
2500 * volume is requested and the osb open count becomes 1.
2501 * It will remove the osb from the global list and also free up all the
2502 * initialized resources and fileobject.
2504 static void ocfs2_delete_osb(struct ocfs2_super
*osb
)
2506 /* This function assumes that the caller has the main osb resource */
2508 /* ocfs2_initializer_super have already created this workqueue */
2510 destroy_workqueue(osb
->ocfs2_wq
);
2512 ocfs2_free_slot_info(osb
);
2514 kfree(osb
->osb_orphan_wipes
);
2515 kfree(osb
->slot_recovery_generations
);
2517 * This belongs in journal shutdown, but because we have to
2518 * allocate osb->journal at the start of ocfs2_initialize_osb(),
2521 kfree(osb
->journal
);
2522 kfree(osb
->local_alloc_copy
);
2523 kfree(osb
->uuid_str
);
2524 kfree(osb
->vol_label
);
2525 ocfs2_put_dlm_debug(osb
->osb_dlm_debug
);
2526 memset(osb
, 0, sizeof(struct ocfs2_super
));
2529 /* Depending on the mount option passed, perform one of the following:
2530 * Put OCFS2 into a readonly state (default)
2531 * Return EIO so that only the process errs
2532 * Fix the error as if fsck.ocfs2 -y
2535 static int ocfs2_handle_error(struct super_block
*sb
)
2537 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
2540 ocfs2_set_osb_flag(osb
, OCFS2_OSB_ERROR_FS
);
2541 pr_crit("On-disk corruption discovered. "
2542 "Please run fsck.ocfs2 once the filesystem is unmounted.\n");
2544 if (osb
->s_mount_opt
& OCFS2_MOUNT_ERRORS_PANIC
) {
2545 panic("OCFS2: (device %s): panic forced after error\n",
2547 } else if (osb
->s_mount_opt
& OCFS2_MOUNT_ERRORS_CONT
) {
2548 pr_crit("OCFS2: Returning error to the calling process.\n");
2550 } else { /* default option */
2552 if (sb_rdonly(sb
) && (ocfs2_is_soft_readonly(osb
) || ocfs2_is_hard_readonly(osb
)))
2555 pr_crit("OCFS2: File system is now read-only.\n");
2556 sb
->s_flags
|= SB_RDONLY
;
2557 ocfs2_set_ro_flag(osb
, 0);
2563 int __ocfs2_error(struct super_block
*sb
, const char *function
,
2564 const char *fmt
, ...)
2566 struct va_format vaf
;
2569 va_start(args
, fmt
);
2573 /* Not using mlog here because we want to show the actual
2574 * function the error came from. */
2575 printk(KERN_CRIT
"OCFS2: ERROR (device %s): %s: %pV",
2576 sb
->s_id
, function
, &vaf
);
2580 return ocfs2_handle_error(sb
);
2583 /* Handle critical errors. This is intentionally more drastic than
2584 * ocfs2_handle_error, so we only use for things like journal errors,
2586 void __ocfs2_abort(struct super_block
*sb
, const char *function
,
2587 const char *fmt
, ...)
2589 struct va_format vaf
;
2592 va_start(args
, fmt
);
2597 printk(KERN_CRIT
"OCFS2: abort (device %s): %s: %pV",
2598 sb
->s_id
, function
, &vaf
);
2602 /* We don't have the cluster support yet to go straight to
2603 * hard readonly in here. Until then, we want to keep
2604 * ocfs2_abort() so that we can at least mark critical
2607 * TODO: This should abort the journal and alert other nodes
2608 * that our slot needs recovery. */
2610 /* Force a panic(). This stinks, but it's better than letting
2611 * things continue without having a proper hard readonly
2613 if (!ocfs2_mount_local(OCFS2_SB(sb
)))
2614 OCFS2_SB(sb
)->s_mount_opt
|= OCFS2_MOUNT_ERRORS_PANIC
;
2615 ocfs2_handle_error(sb
);
2619 * Void signal blockers, because in-kernel sigprocmask() only fails
2620 * when SIG_* is wrong.
2622 void ocfs2_block_signals(sigset_t
*oldset
)
2627 sigfillset(&blocked
);
2628 rc
= sigprocmask(SIG_BLOCK
, &blocked
, oldset
);
2632 void ocfs2_unblock_signals(sigset_t
*oldset
)
2634 int rc
= sigprocmask(SIG_SETMASK
, oldset
, NULL
);
2638 module_init(ocfs2_init
);
2639 module_exit(ocfs2_exit
);