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/signal.h>
30 #define CREATE_TRACE_POINTS
31 #include "ocfs2_trace.h"
33 #include <cluster/masklog.h>
37 /* this should be the only file to include a version 1 header */
38 #include "ocfs1_fs_compat.h"
42 #include "blockcheck.h"
45 #include "extent_map.h"
46 #include "heartbeat.h"
49 #include "localalloc.h"
57 #include "refcounttree.h"
60 #include "buffer_head_io.h"
61 #include "filecheck.h"
63 static struct kmem_cache
*ocfs2_inode_cachep
;
64 struct kmem_cache
*ocfs2_dquot_cachep
;
65 struct kmem_cache
*ocfs2_qf_chunk_cachep
;
67 static struct dentry
*ocfs2_debugfs_root
;
69 MODULE_AUTHOR("Oracle");
70 MODULE_LICENSE("GPL");
71 MODULE_DESCRIPTION("OCFS2 cluster file system");
75 unsigned long commit_interval
;
76 unsigned long mount_opt
;
77 unsigned int atime_quantum
;
80 unsigned int resv_level
;
82 char cluster_stack
[OCFS2_STACK_LABEL_LEN
+ 1];
85 static int ocfs2_parse_options(struct super_block
*sb
, char *options
,
86 struct mount_options
*mopt
,
88 static int ocfs2_check_set_options(struct super_block
*sb
,
89 struct mount_options
*options
);
90 static int ocfs2_show_options(struct seq_file
*s
, struct dentry
*root
);
91 static void ocfs2_put_super(struct super_block
*sb
);
92 static int ocfs2_mount_volume(struct super_block
*sb
);
93 static int ocfs2_remount(struct super_block
*sb
, int *flags
, char *data
);
94 static void ocfs2_dismount_volume(struct super_block
*sb
, int mnt_err
);
95 static int ocfs2_initialize_mem_caches(void);
96 static void ocfs2_free_mem_caches(void);
97 static void ocfs2_delete_osb(struct ocfs2_super
*osb
);
99 static int ocfs2_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
101 static int ocfs2_sync_fs(struct super_block
*sb
, int wait
);
103 static int ocfs2_init_global_system_inodes(struct ocfs2_super
*osb
);
104 static int ocfs2_init_local_system_inodes(struct ocfs2_super
*osb
);
105 static void ocfs2_release_system_inodes(struct ocfs2_super
*osb
);
106 static int ocfs2_check_volume(struct ocfs2_super
*osb
);
107 static int ocfs2_verify_volume(struct ocfs2_dinode
*di
,
108 struct buffer_head
*bh
,
110 struct ocfs2_blockcheck_stats
*stats
);
111 static int ocfs2_initialize_super(struct super_block
*sb
,
112 struct buffer_head
*bh
,
114 struct ocfs2_blockcheck_stats
*stats
);
115 static int ocfs2_get_sector(struct super_block
*sb
,
116 struct buffer_head
**bh
,
119 static struct inode
*ocfs2_alloc_inode(struct super_block
*sb
);
120 static void ocfs2_free_inode(struct inode
*inode
);
121 static int ocfs2_susp_quotas(struct ocfs2_super
*osb
, int unsuspend
);
122 static int ocfs2_enable_quotas(struct ocfs2_super
*osb
);
123 static void ocfs2_disable_quotas(struct ocfs2_super
*osb
);
125 static struct dquot
**ocfs2_get_dquots(struct inode
*inode
)
127 return OCFS2_I(inode
)->i_dquot
;
130 static const struct super_operations ocfs2_sops
= {
131 .statfs
= ocfs2_statfs
,
132 .alloc_inode
= ocfs2_alloc_inode
,
133 .free_inode
= ocfs2_free_inode
,
134 .drop_inode
= ocfs2_drop_inode
,
135 .evict_inode
= ocfs2_evict_inode
,
136 .sync_fs
= ocfs2_sync_fs
,
137 .put_super
= ocfs2_put_super
,
138 .remount_fs
= ocfs2_remount
,
139 .show_options
= ocfs2_show_options
,
140 .quota_read
= ocfs2_quota_read
,
141 .quota_write
= ocfs2_quota_write
,
142 .get_dquots
= ocfs2_get_dquots
,
169 Opt_coherency_buffered
,
173 Opt_journal_async_commit
,
178 static const match_table_t tokens
= {
179 {Opt_barrier
, "barrier=%u"},
180 {Opt_err_panic
, "errors=panic"},
181 {Opt_err_ro
, "errors=remount-ro"},
183 {Opt_nointr
, "nointr"},
184 {Opt_hb_none
, OCFS2_HB_NONE
},
185 {Opt_hb_local
, OCFS2_HB_LOCAL
},
186 {Opt_hb_global
, OCFS2_HB_GLOBAL
},
187 {Opt_data_ordered
, "data=ordered"},
188 {Opt_data_writeback
, "data=writeback"},
189 {Opt_atime_quantum
, "atime_quantum=%u"},
190 {Opt_slot
, "preferred_slot=%u"},
191 {Opt_commit
, "commit=%u"},
192 {Opt_localalloc
, "localalloc=%d"},
193 {Opt_localflocks
, "localflocks"},
194 {Opt_stack
, "cluster_stack=%s"},
195 {Opt_user_xattr
, "user_xattr"},
196 {Opt_nouser_xattr
, "nouser_xattr"},
197 {Opt_inode64
, "inode64"},
199 {Opt_noacl
, "noacl"},
200 {Opt_usrquota
, "usrquota"},
201 {Opt_grpquota
, "grpquota"},
202 {Opt_coherency_buffered
, "coherency=buffered"},
203 {Opt_coherency_full
, "coherency=full"},
204 {Opt_resv_level
, "resv_level=%u"},
205 {Opt_dir_resv_level
, "dir_resv_level=%u"},
206 {Opt_journal_async_commit
, "journal_async_commit"},
207 {Opt_err_cont
, "errors=continue"},
211 #ifdef CONFIG_DEBUG_FS
212 static int ocfs2_osb_dump(struct ocfs2_super
*osb
, char *buf
, int len
)
214 struct ocfs2_cluster_connection
*cconn
= osb
->cconn
;
215 struct ocfs2_recovery_map
*rm
= osb
->recovery_map
;
216 struct ocfs2_orphan_scan
*os
= &osb
->osb_orphan_scan
;
220 out
+= scnprintf(buf
+ out
, len
- out
,
221 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n",
222 "Device", osb
->dev_str
, osb
->uuid_str
,
223 osb
->fs_generation
, osb
->vol_label
);
225 out
+= scnprintf(buf
+ out
, len
- out
,
226 "%10s => State: %d Flags: 0x%lX\n", "Volume",
227 atomic_read(&osb
->vol_state
), osb
->osb_flags
);
229 out
+= scnprintf(buf
+ out
, len
- out
,
230 "%10s => Block: %lu Cluster: %d\n", "Sizes",
231 osb
->sb
->s_blocksize
, osb
->s_clustersize
);
233 out
+= scnprintf(buf
+ out
, len
- out
,
234 "%10s => Compat: 0x%X Incompat: 0x%X "
236 "Features", osb
->s_feature_compat
,
237 osb
->s_feature_incompat
, osb
->s_feature_ro_compat
);
239 out
+= scnprintf(buf
+ out
, len
- out
,
240 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount",
241 osb
->s_mount_opt
, osb
->s_atime_quantum
);
244 out
+= scnprintf(buf
+ out
, len
- out
,
245 "%10s => Stack: %s Name: %*s "
246 "Version: %d.%d\n", "Cluster",
247 (*osb
->osb_cluster_stack
== '\0' ?
248 "o2cb" : osb
->osb_cluster_stack
),
249 cconn
->cc_namelen
, cconn
->cc_name
,
250 cconn
->cc_version
.pv_major
,
251 cconn
->cc_version
.pv_minor
);
254 spin_lock_irqsave(&osb
->dc_task_lock
, flags
);
255 out
+= scnprintf(buf
+ out
, len
- out
,
256 "%10s => Pid: %d Count: %lu WakeSeq: %lu "
257 "WorkSeq: %lu\n", "DownCnvt",
258 (osb
->dc_task
? task_pid_nr(osb
->dc_task
) : -1),
259 osb
->blocked_lock_count
, osb
->dc_wake_sequence
,
260 osb
->dc_work_sequence
);
261 spin_unlock_irqrestore(&osb
->dc_task_lock
, flags
);
263 spin_lock(&osb
->osb_lock
);
264 out
+= scnprintf(buf
+ out
, len
- out
, "%10s => Pid: %d Nodes:",
266 (osb
->recovery_thread_task
?
267 task_pid_nr(osb
->recovery_thread_task
) : -1));
268 if (rm
->rm_used
== 0)
269 out
+= scnprintf(buf
+ out
, len
- out
, " None\n");
271 for (i
= 0; i
< rm
->rm_used
; i
++)
272 out
+= scnprintf(buf
+ out
, len
- out
, " %d",
274 out
+= scnprintf(buf
+ out
, len
- out
, "\n");
276 spin_unlock(&osb
->osb_lock
);
278 out
+= scnprintf(buf
+ out
, len
- out
,
279 "%10s => Pid: %d Interval: %lu\n", "Commit",
280 (osb
->commit_task
? task_pid_nr(osb
->commit_task
) : -1),
281 osb
->osb_commit_interval
);
283 out
+= scnprintf(buf
+ out
, len
- out
,
284 "%10s => State: %d TxnId: %lu NumTxns: %d\n",
285 "Journal", osb
->journal
->j_state
,
286 osb
->journal
->j_trans_id
,
287 atomic_read(&osb
->journal
->j_num_trans
));
289 out
+= scnprintf(buf
+ out
, len
- out
,
290 "%10s => GlobalAllocs: %d LocalAllocs: %d "
291 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
293 atomic_read(&osb
->alloc_stats
.bitmap_data
),
294 atomic_read(&osb
->alloc_stats
.local_data
),
295 atomic_read(&osb
->alloc_stats
.bg_allocs
),
296 atomic_read(&osb
->alloc_stats
.moves
),
297 atomic_read(&osb
->alloc_stats
.bg_extends
));
299 out
+= scnprintf(buf
+ out
, len
- out
,
300 "%10s => State: %u Descriptor: %llu Size: %u bits "
301 "Default: %u bits\n",
302 "LocalAlloc", osb
->local_alloc_state
,
303 (unsigned long long)osb
->la_last_gd
,
304 osb
->local_alloc_bits
, osb
->local_alloc_default_bits
);
306 spin_lock(&osb
->osb_lock
);
307 out
+= scnprintf(buf
+ out
, len
- out
,
308 "%10s => InodeSlot: %d StolenInodes: %d, "
309 "MetaSlot: %d StolenMeta: %d\n", "Steal",
310 osb
->s_inode_steal_slot
,
311 atomic_read(&osb
->s_num_inodes_stolen
),
312 osb
->s_meta_steal_slot
,
313 atomic_read(&osb
->s_num_meta_stolen
));
314 spin_unlock(&osb
->osb_lock
);
316 out
+= scnprintf(buf
+ out
, len
- out
, "OrphanScan => ");
317 out
+= scnprintf(buf
+ out
, len
- out
, "Local: %u Global: %u ",
318 os
->os_count
, os
->os_seqno
);
319 out
+= scnprintf(buf
+ out
, len
- out
, " Last Scan: ");
320 if (atomic_read(&os
->os_state
) == ORPHAN_SCAN_INACTIVE
)
321 out
+= scnprintf(buf
+ out
, len
- out
, "Disabled\n");
323 out
+= scnprintf(buf
+ out
, len
- out
, "%lu seconds ago\n",
324 (unsigned long)(ktime_get_seconds() - os
->os_scantime
));
326 out
+= scnprintf(buf
+ out
, len
- out
, "%10s => %3s %10s\n",
327 "Slots", "Num", "RecoGen");
328 for (i
= 0; i
< osb
->max_slots
; ++i
) {
329 out
+= scnprintf(buf
+ out
, len
- out
,
330 "%10s %c %3d %10d\n",
332 (i
== osb
->slot_num
? '*' : ' '),
333 i
, osb
->slot_recovery_generations
[i
]);
339 static int ocfs2_osb_debug_open(struct inode
*inode
, struct file
*file
)
341 struct ocfs2_super
*osb
= inode
->i_private
;
344 buf
= kmalloc(PAGE_SIZE
, GFP_KERNEL
);
348 i_size_write(inode
, ocfs2_osb_dump(osb
, buf
, PAGE_SIZE
));
350 file
->private_data
= buf
;
357 static int ocfs2_debug_release(struct inode
*inode
, struct file
*file
)
359 kfree(file
->private_data
);
363 static ssize_t
ocfs2_debug_read(struct file
*file
, char __user
*buf
,
364 size_t nbytes
, loff_t
*ppos
)
366 return simple_read_from_buffer(buf
, nbytes
, ppos
, file
->private_data
,
367 i_size_read(file
->f_mapping
->host
));
370 static int ocfs2_osb_debug_open(struct inode
*inode
, struct file
*file
)
374 static int ocfs2_debug_release(struct inode
*inode
, struct file
*file
)
378 static ssize_t
ocfs2_debug_read(struct file
*file
, char __user
*buf
,
379 size_t nbytes
, loff_t
*ppos
)
383 #endif /* CONFIG_DEBUG_FS */
385 static const struct file_operations ocfs2_osb_debug_fops
= {
386 .open
= ocfs2_osb_debug_open
,
387 .release
= ocfs2_debug_release
,
388 .read
= ocfs2_debug_read
,
389 .llseek
= generic_file_llseek
,
392 static int ocfs2_sync_fs(struct super_block
*sb
, int wait
)
396 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
398 if (ocfs2_is_hard_readonly(osb
))
402 status
= ocfs2_flush_truncate_log(osb
);
406 ocfs2_schedule_truncate_log_flush(osb
, 0);
409 if (jbd2_journal_start_commit(osb
->journal
->j_journal
,
412 jbd2_log_wait_commit(osb
->journal
->j_journal
,
418 static int ocfs2_need_system_inode(struct ocfs2_super
*osb
, int ino
)
420 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb
->sb
, OCFS2_FEATURE_RO_COMPAT_USRQUOTA
)
421 && (ino
== USER_QUOTA_SYSTEM_INODE
422 || ino
== LOCAL_USER_QUOTA_SYSTEM_INODE
))
424 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb
->sb
, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
)
425 && (ino
== GROUP_QUOTA_SYSTEM_INODE
426 || ino
== LOCAL_GROUP_QUOTA_SYSTEM_INODE
))
431 static int ocfs2_init_global_system_inodes(struct ocfs2_super
*osb
)
433 struct inode
*new = NULL
;
437 new = ocfs2_iget(osb
, osb
->root_blkno
, OCFS2_FI_FLAG_SYSFILE
, 0);
439 status
= PTR_ERR(new);
443 osb
->root_inode
= new;
445 new = ocfs2_iget(osb
, osb
->system_dir_blkno
, OCFS2_FI_FLAG_SYSFILE
, 0);
447 status
= PTR_ERR(new);
451 osb
->sys_root_inode
= new;
453 for (i
= OCFS2_FIRST_ONLINE_SYSTEM_INODE
;
454 i
<= OCFS2_LAST_GLOBAL_SYSTEM_INODE
; i
++) {
455 if (!ocfs2_need_system_inode(osb
, i
))
457 new = ocfs2_get_system_file_inode(osb
, i
, osb
->slot_num
);
459 ocfs2_release_system_inodes(osb
);
460 status
= ocfs2_is_soft_readonly(osb
) ? -EROFS
: -EINVAL
;
462 mlog(ML_ERROR
, "Unable to load system inode %d, "
463 "possibly corrupt fs?", i
);
466 // the array now has one ref, so drop this one
476 static int ocfs2_init_local_system_inodes(struct ocfs2_super
*osb
)
478 struct inode
*new = NULL
;
482 for (i
= OCFS2_LAST_GLOBAL_SYSTEM_INODE
+ 1;
483 i
< NUM_SYSTEM_INODES
;
485 if (!ocfs2_need_system_inode(osb
, i
))
487 new = ocfs2_get_system_file_inode(osb
, i
, osb
->slot_num
);
489 ocfs2_release_system_inodes(osb
);
490 status
= ocfs2_is_soft_readonly(osb
) ? -EROFS
: -EINVAL
;
491 mlog(ML_ERROR
, "status=%d, sysfile=%d, slot=%d\n",
492 status
, i
, osb
->slot_num
);
495 /* the array now has one ref, so drop this one */
505 static void ocfs2_release_system_inodes(struct ocfs2_super
*osb
)
510 for (i
= 0; i
< NUM_GLOBAL_SYSTEM_INODES
; i
++) {
511 inode
= osb
->global_system_inodes
[i
];
514 osb
->global_system_inodes
[i
] = NULL
;
518 inode
= osb
->sys_root_inode
;
521 osb
->sys_root_inode
= NULL
;
524 inode
= osb
->root_inode
;
527 osb
->root_inode
= NULL
;
530 if (!osb
->local_system_inodes
)
533 for (i
= 0; i
< NUM_LOCAL_SYSTEM_INODES
* osb
->max_slots
; i
++) {
534 if (osb
->local_system_inodes
[i
]) {
535 iput(osb
->local_system_inodes
[i
]);
536 osb
->local_system_inodes
[i
] = NULL
;
540 kfree(osb
->local_system_inodes
);
541 osb
->local_system_inodes
= NULL
;
544 /* We're allocating fs objects, use GFP_NOFS */
545 static struct inode
*ocfs2_alloc_inode(struct super_block
*sb
)
547 struct ocfs2_inode_info
*oi
;
549 oi
= alloc_inode_sb(sb
, ocfs2_inode_cachep
, GFP_NOFS
);
554 oi
->i_datasync_tid
= 0;
555 memset(&oi
->i_dquot
, 0, sizeof(oi
->i_dquot
));
557 jbd2_journal_init_jbd_inode(&oi
->ip_jinode
, &oi
->vfs_inode
);
558 return &oi
->vfs_inode
;
561 static void ocfs2_free_inode(struct inode
*inode
)
563 kmem_cache_free(ocfs2_inode_cachep
, OCFS2_I(inode
));
566 static unsigned long long ocfs2_max_file_offset(unsigned int bbits
,
569 unsigned int bytes
= 1 << cbits
;
570 unsigned int trim
= bytes
;
571 unsigned int bitshift
= 32;
574 * i_size and all block offsets in ocfs2 are always 64 bits
575 * wide. i_clusters is 32 bits, in cluster-sized units. So on
576 * 64 bit platforms, cluster size will be the limiting factor.
579 #if BITS_PER_LONG == 32
580 BUILD_BUG_ON(sizeof(sector_t
) != 8);
582 * We might be limited by page cache size.
584 if (bytes
> PAGE_SIZE
) {
588 * Shift by 31 here so that we don't get larger than
596 * Trim by a whole cluster when we can actually approach the
597 * on-disk limits. Otherwise we can overflow i_clusters when
598 * an extent start is at the max offset.
600 return (((unsigned long long)bytes
) << bitshift
) - trim
;
603 static int ocfs2_remount(struct super_block
*sb
, int *flags
, char *data
)
605 int incompat_features
;
607 struct mount_options parsed_options
;
608 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
613 if (!ocfs2_parse_options(sb
, data
, &parsed_options
, 1) ||
614 !ocfs2_check_set_options(sb
, &parsed_options
)) {
619 tmp
= OCFS2_MOUNT_HB_LOCAL
| OCFS2_MOUNT_HB_GLOBAL
|
621 if ((osb
->s_mount_opt
& tmp
) != (parsed_options
.mount_opt
& tmp
)) {
623 mlog(ML_ERROR
, "Cannot change heartbeat mode on remount\n");
627 if ((osb
->s_mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
) !=
628 (parsed_options
.mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
)) {
630 mlog(ML_ERROR
, "Cannot change data mode on remount\n");
634 /* Probably don't want this on remount; it might
635 * mess with other nodes */
636 if (!(osb
->s_mount_opt
& OCFS2_MOUNT_INODE64
) &&
637 (parsed_options
.mount_opt
& OCFS2_MOUNT_INODE64
)) {
639 mlog(ML_ERROR
, "Cannot enable inode64 on remount\n");
643 /* We're going to/from readonly mode. */
644 if ((bool)(*flags
& SB_RDONLY
) != sb_rdonly(sb
)) {
645 /* Disable quota accounting before remounting RO */
646 if (*flags
& SB_RDONLY
) {
647 ret
= ocfs2_susp_quotas(osb
, 0);
651 /* Lock here so the check of HARD_RO and the potential
652 * setting of SOFT_RO is atomic. */
653 spin_lock(&osb
->osb_lock
);
654 if (osb
->osb_flags
& OCFS2_OSB_HARD_RO
) {
655 mlog(ML_ERROR
, "Remount on readonly device is forbidden.\n");
660 if (*flags
& SB_RDONLY
) {
661 sb
->s_flags
|= SB_RDONLY
;
662 osb
->osb_flags
|= OCFS2_OSB_SOFT_RO
;
664 if (osb
->osb_flags
& OCFS2_OSB_ERROR_FS
) {
665 mlog(ML_ERROR
, "Cannot remount RDWR "
666 "filesystem due to previous errors.\n");
670 incompat_features
= OCFS2_HAS_RO_COMPAT_FEATURE(sb
, ~OCFS2_FEATURE_RO_COMPAT_SUPP
);
671 if (incompat_features
) {
672 mlog(ML_ERROR
, "Cannot remount RDWR because "
673 "of unsupported optional features "
674 "(%x).\n", incompat_features
);
678 sb
->s_flags
&= ~SB_RDONLY
;
679 osb
->osb_flags
&= ~OCFS2_OSB_SOFT_RO
;
681 trace_ocfs2_remount(sb
->s_flags
, osb
->osb_flags
, *flags
);
683 spin_unlock(&osb
->osb_lock
);
684 /* Enable quota accounting after remounting RW */
685 if (!ret
&& !(*flags
& SB_RDONLY
)) {
686 if (sb_any_quota_suspended(sb
))
687 ret
= ocfs2_susp_quotas(osb
, 1);
689 ret
= ocfs2_enable_quotas(osb
);
691 /* Return back changes... */
692 spin_lock(&osb
->osb_lock
);
693 sb
->s_flags
|= SB_RDONLY
;
694 osb
->osb_flags
|= OCFS2_OSB_SOFT_RO
;
695 spin_unlock(&osb
->osb_lock
);
702 /* Only save off the new mount options in case of a successful
704 osb
->s_mount_opt
= parsed_options
.mount_opt
;
705 osb
->s_atime_quantum
= parsed_options
.atime_quantum
;
706 osb
->preferred_slot
= parsed_options
.slot
;
707 if (parsed_options
.commit_interval
)
708 osb
->osb_commit_interval
= parsed_options
.commit_interval
;
710 if (!ocfs2_is_hard_readonly(osb
))
711 ocfs2_set_journal_params(osb
);
713 sb
->s_flags
= (sb
->s_flags
& ~SB_POSIXACL
) |
714 ((osb
->s_mount_opt
& OCFS2_MOUNT_POSIX_ACL
) ?
721 static int ocfs2_sb_probe(struct super_block
*sb
,
722 struct buffer_head
**bh
,
724 struct ocfs2_blockcheck_stats
*stats
)
727 struct ocfs1_vol_disk_hdr
*hdr
;
728 struct ocfs2_dinode
*di
;
734 *sector_size
= bdev_logical_block_size(sb
->s_bdev
);
735 if (*sector_size
> OCFS2_MAX_BLOCKSIZE
) {
736 mlog(ML_ERROR
, "Hardware sector size too large: %d (max=%d)\n",
737 *sector_size
, OCFS2_MAX_BLOCKSIZE
);
742 /* Can this really happen? */
743 if (*sector_size
< OCFS2_MIN_BLOCKSIZE
)
744 *sector_size
= OCFS2_MIN_BLOCKSIZE
;
746 /* check block zero for old format */
747 status
= ocfs2_get_sector(sb
, bh
, 0, *sector_size
);
752 hdr
= (struct ocfs1_vol_disk_hdr
*) (*bh
)->b_data
;
753 if (hdr
->major_version
== OCFS1_MAJOR_VERSION
) {
754 mlog(ML_ERROR
, "incompatible version: %u.%u\n",
755 hdr
->major_version
, hdr
->minor_version
);
758 if (memcmp(hdr
->signature
, OCFS1_VOLUME_SIGNATURE
,
759 strlen(OCFS1_VOLUME_SIGNATURE
)) == 0) {
760 mlog(ML_ERROR
, "incompatible volume signature: %8s\n",
767 mlog(ML_ERROR
, "This is an ocfs v1 filesystem which must be "
768 "upgraded before mounting with ocfs v2\n");
773 * Now check at magic offset for 512, 1024, 2048, 4096
774 * blocksizes. 4096 is the maximum blocksize because it is
775 * the minimum clustersize.
778 for (blksize
= *sector_size
;
779 blksize
<= OCFS2_MAX_BLOCKSIZE
;
781 tmpstat
= ocfs2_get_sector(sb
, bh
,
782 OCFS2_SUPER_BLOCK_BLKNO
,
789 di
= (struct ocfs2_dinode
*) (*bh
)->b_data
;
790 memset(stats
, 0, sizeof(struct ocfs2_blockcheck_stats
));
791 spin_lock_init(&stats
->b_lock
);
792 tmpstat
= ocfs2_verify_volume(di
, *bh
, blksize
, stats
);
797 if (tmpstat
!= -EAGAIN
) {
807 static int ocfs2_verify_heartbeat(struct ocfs2_super
*osb
)
809 u32 hb_enabled
= OCFS2_MOUNT_HB_LOCAL
| OCFS2_MOUNT_HB_GLOBAL
;
811 if (osb
->s_mount_opt
& hb_enabled
) {
812 if (ocfs2_mount_local(osb
)) {
813 mlog(ML_ERROR
, "Cannot heartbeat on a locally "
814 "mounted device.\n");
817 if (ocfs2_userspace_stack(osb
)) {
818 mlog(ML_ERROR
, "Userspace stack expected, but "
819 "o2cb heartbeat arguments passed to mount\n");
822 if (((osb
->s_mount_opt
& OCFS2_MOUNT_HB_GLOBAL
) &&
823 !ocfs2_cluster_o2cb_global_heartbeat(osb
)) ||
824 ((osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) &&
825 ocfs2_cluster_o2cb_global_heartbeat(osb
))) {
826 mlog(ML_ERROR
, "Mismatching o2cb heartbeat modes\n");
831 if (!(osb
->s_mount_opt
& hb_enabled
)) {
832 if (!ocfs2_mount_local(osb
) && !ocfs2_is_hard_readonly(osb
) &&
833 !ocfs2_userspace_stack(osb
)) {
834 mlog(ML_ERROR
, "Heartbeat has to be started to mount "
835 "a read-write clustered device.\n");
844 * If we're using a userspace stack, mount should have passed
845 * a name that matches the disk. If not, mount should not
846 * have passed a stack.
848 static int ocfs2_verify_userspace_stack(struct ocfs2_super
*osb
,
849 struct mount_options
*mopt
)
851 if (!ocfs2_userspace_stack(osb
) && mopt
->cluster_stack
[0]) {
853 "cluster stack passed to mount, but this filesystem "
854 "does not support it\n");
858 if (ocfs2_userspace_stack(osb
) &&
859 strncmp(osb
->osb_cluster_stack
, mopt
->cluster_stack
,
860 OCFS2_STACK_LABEL_LEN
)) {
862 "cluster stack passed to mount (\"%s\") does not "
863 "match the filesystem (\"%s\")\n",
865 osb
->osb_cluster_stack
);
872 static int ocfs2_susp_quotas(struct ocfs2_super
*osb
, int unsuspend
)
875 struct super_block
*sb
= osb
->sb
;
876 unsigned int feature
[OCFS2_MAXQUOTAS
] = {
877 OCFS2_FEATURE_RO_COMPAT_USRQUOTA
,
878 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
};
881 for (type
= 0; type
< OCFS2_MAXQUOTAS
; type
++) {
882 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb
, feature
[type
]))
885 status
= dquot_resume(sb
, type
);
887 struct ocfs2_mem_dqinfo
*oinfo
;
889 /* Cancel periodic syncing before suspending */
890 oinfo
= sb_dqinfo(sb
, type
)->dqi_priv
;
891 cancel_delayed_work_sync(&oinfo
->dqi_sync_work
);
892 status
= dquot_suspend(sb
, type
);
898 mlog(ML_ERROR
, "Failed to suspend/unsuspend quotas on "
899 "remount (error = %d).\n", status
);
903 static int ocfs2_enable_quotas(struct ocfs2_super
*osb
)
905 struct inode
*inode
[OCFS2_MAXQUOTAS
] = { NULL
, NULL
};
906 struct super_block
*sb
= osb
->sb
;
907 unsigned int feature
[OCFS2_MAXQUOTAS
] = {
908 OCFS2_FEATURE_RO_COMPAT_USRQUOTA
,
909 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
};
910 unsigned int ino
[OCFS2_MAXQUOTAS
] = {
911 LOCAL_USER_QUOTA_SYSTEM_INODE
,
912 LOCAL_GROUP_QUOTA_SYSTEM_INODE
};
916 sb_dqopt(sb
)->flags
|= DQUOT_QUOTA_SYS_FILE
| DQUOT_NEGATIVE_USAGE
;
917 for (type
= 0; type
< OCFS2_MAXQUOTAS
; type
++) {
918 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb
, feature
[type
]))
920 inode
[type
] = ocfs2_get_system_file_inode(osb
, ino
[type
],
926 status
= dquot_load_quota_inode(inode
[type
], type
, QFMT_OCFS2
,
927 DQUOT_USAGE_ENABLED
);
932 for (type
= 0; type
< OCFS2_MAXQUOTAS
; type
++)
936 ocfs2_disable_quotas(osb
);
937 for (type
= 0; type
< OCFS2_MAXQUOTAS
; type
++)
943 static void ocfs2_disable_quotas(struct ocfs2_super
*osb
)
947 struct super_block
*sb
= osb
->sb
;
948 struct ocfs2_mem_dqinfo
*oinfo
;
950 /* We mostly ignore errors in this function because there's not much
951 * we can do when we see them */
952 for (type
= 0; type
< OCFS2_MAXQUOTAS
; type
++) {
953 if (!sb_has_quota_loaded(sb
, type
))
955 if (!sb_has_quota_suspended(sb
, type
)) {
956 oinfo
= sb_dqinfo(sb
, type
)->dqi_priv
;
957 cancel_delayed_work_sync(&oinfo
->dqi_sync_work
);
959 inode
= igrab(sb
->s_dquot
.files
[type
]);
960 /* Turn off quotas. This will remove all dquot structures from
961 * memory and so they will be automatically synced to global
963 dquot_disable(sb
, type
, DQUOT_USAGE_ENABLED
|
964 DQUOT_LIMITS_ENABLED
);
969 static int ocfs2_fill_super(struct super_block
*sb
, void *data
, int silent
)
972 int status
, sector_size
;
973 struct mount_options parsed_options
;
974 struct inode
*inode
= NULL
;
975 struct ocfs2_super
*osb
= NULL
;
976 struct buffer_head
*bh
= NULL
;
978 struct ocfs2_blockcheck_stats stats
;
980 trace_ocfs2_fill_super(sb
, data
, silent
);
982 if (!ocfs2_parse_options(sb
, data
, &parsed_options
, 0)) {
987 /* probe for superblock */
988 status
= ocfs2_sb_probe(sb
, &bh
, §or_size
, &stats
);
990 mlog(ML_ERROR
, "superblock probe failed!\n");
994 status
= ocfs2_initialize_super(sb
, bh
, sector_size
, &stats
);
1002 if (!ocfs2_check_set_options(sb
, &parsed_options
)) {
1006 osb
->s_mount_opt
= parsed_options
.mount_opt
;
1007 osb
->s_atime_quantum
= parsed_options
.atime_quantum
;
1008 osb
->preferred_slot
= parsed_options
.slot
;
1009 osb
->osb_commit_interval
= parsed_options
.commit_interval
;
1011 ocfs2_la_set_sizes(osb
, parsed_options
.localalloc_opt
);
1012 osb
->osb_resv_level
= parsed_options
.resv_level
;
1013 osb
->osb_dir_resv_level
= parsed_options
.resv_level
;
1014 if (parsed_options
.dir_resv_level
== -1)
1015 osb
->osb_dir_resv_level
= parsed_options
.resv_level
;
1017 osb
->osb_dir_resv_level
= parsed_options
.dir_resv_level
;
1019 status
= ocfs2_verify_userspace_stack(osb
, &parsed_options
);
1023 sb
->s_magic
= OCFS2_SUPER_MAGIC
;
1025 sb
->s_flags
= (sb
->s_flags
& ~(SB_POSIXACL
| SB_NOSEC
)) |
1026 ((osb
->s_mount_opt
& OCFS2_MOUNT_POSIX_ACL
) ? SB_POSIXACL
: 0);
1028 /* Hard readonly mode only if: bdev_read_only, SB_RDONLY,
1030 if (bdev_read_only(sb
->s_bdev
)) {
1031 if (!sb_rdonly(sb
)) {
1033 mlog(ML_ERROR
, "Readonly device detected but readonly "
1034 "mount was not specified.\n");
1038 /* You should not be able to start a local heartbeat
1039 * on a readonly device. */
1040 if (osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) {
1042 mlog(ML_ERROR
, "Local heartbeat specified on readonly "
1047 status
= ocfs2_check_journals_nolocks(osb
);
1049 if (status
== -EROFS
)
1050 mlog(ML_ERROR
, "Recovery required on readonly "
1051 "file system, but write access is "
1056 ocfs2_set_ro_flag(osb
, 1);
1058 printk(KERN_NOTICE
"ocfs2: Readonly device (%s) detected. "
1059 "Cluster services will not be used for this mount. "
1060 "Recovery will be skipped.\n", osb
->dev_str
);
1063 if (!ocfs2_is_hard_readonly(osb
)) {
1065 ocfs2_set_ro_flag(osb
, 0);
1068 status
= ocfs2_verify_heartbeat(osb
);
1072 osb
->osb_debug_root
= debugfs_create_dir(osb
->uuid_str
,
1073 ocfs2_debugfs_root
);
1075 debugfs_create_file("fs_state", S_IFREG
|S_IRUSR
, osb
->osb_debug_root
,
1076 osb
, &ocfs2_osb_debug_fops
);
1078 if (ocfs2_meta_ecc(osb
))
1079 ocfs2_blockcheck_stats_debugfs_install( &osb
->osb_ecc_stats
,
1080 osb
->osb_debug_root
);
1082 status
= ocfs2_mount_volume(sb
);
1086 if (osb
->root_inode
)
1087 inode
= igrab(osb
->root_inode
);
1094 osb
->osb_dev_kset
= kset_create_and_add(sb
->s_id
, NULL
,
1096 if (!osb
->osb_dev_kset
) {
1098 mlog(ML_ERROR
, "Unable to create device kset %s.\n", sb
->s_id
);
1102 /* Create filecheck sysfs related directories/files at
1103 * /sys/fs/ocfs2/<devname>/filecheck */
1104 if (ocfs2_filecheck_create_sysfs(osb
)) {
1106 mlog(ML_ERROR
, "Unable to create filecheck sysfs directory at "
1107 "/sys/fs/ocfs2/%s/filecheck.\n", sb
->s_id
);
1111 root
= d_make_root(inode
);
1119 ocfs2_complete_mount_recovery(osb
);
1121 if (ocfs2_mount_local(osb
))
1122 snprintf(nodestr
, sizeof(nodestr
), "local");
1124 snprintf(nodestr
, sizeof(nodestr
), "%u", osb
->node_num
);
1126 printk(KERN_INFO
"ocfs2: Mounting device (%s) on (node %s, slot %d) "
1127 "with %s data mode.\n",
1128 osb
->dev_str
, nodestr
, osb
->slot_num
,
1129 osb
->s_mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
? "writeback" :
1132 atomic_set(&osb
->vol_state
, VOLUME_MOUNTED
);
1133 wake_up(&osb
->osb_mount_event
);
1135 /* Now we can initialize quotas because we can afford to wait
1136 * for cluster locks recovery now. That also means that truncation
1137 * log recovery can happen but that waits for proper quota setup */
1138 if (!sb_rdonly(sb
)) {
1139 status
= ocfs2_enable_quotas(osb
);
1141 /* We have to err-out specially here because
1142 * s_root is already set */
1144 atomic_set(&osb
->vol_state
, VOLUME_DISABLED
);
1145 wake_up(&osb
->osb_mount_event
);
1150 ocfs2_complete_quota_recovery(osb
);
1152 /* Now we wake up again for processes waiting for quotas */
1153 atomic_set(&osb
->vol_state
, VOLUME_MOUNTED_QUOTAS
);
1154 wake_up(&osb
->osb_mount_event
);
1156 /* Start this when the mount is almost sure of being successful */
1157 ocfs2_orphan_scan_start(osb
);
1162 atomic_set(&osb
->vol_state
, VOLUME_DISABLED
);
1163 wake_up(&osb
->osb_mount_event
);
1164 ocfs2_free_replay_slots(osb
);
1165 ocfs2_dismount_volume(sb
, 1);
1169 debugfs_remove_recursive(osb
->osb_debug_root
);
1171 ocfs2_release_system_inodes(osb
);
1172 kfree(osb
->recovery_map
);
1173 ocfs2_delete_osb(osb
);
1181 static struct dentry
*ocfs2_mount(struct file_system_type
*fs_type
,
1183 const char *dev_name
,
1186 return mount_bdev(fs_type
, flags
, dev_name
, data
, ocfs2_fill_super
);
1189 static struct file_system_type ocfs2_fs_type
= {
1190 .owner
= THIS_MODULE
,
1192 .mount
= ocfs2_mount
,
1193 .kill_sb
= kill_block_super
,
1194 .fs_flags
= FS_REQUIRES_DEV
|FS_RENAME_DOES_D_MOVE
,
1197 MODULE_ALIAS_FS("ocfs2");
1199 static int ocfs2_check_set_options(struct super_block
*sb
,
1200 struct mount_options
*options
)
1202 if (options
->mount_opt
& OCFS2_MOUNT_USRQUOTA
&&
1203 !OCFS2_HAS_RO_COMPAT_FEATURE(sb
,
1204 OCFS2_FEATURE_RO_COMPAT_USRQUOTA
)) {
1205 mlog(ML_ERROR
, "User quotas were requested, but this "
1206 "filesystem does not have the feature enabled.\n");
1209 if (options
->mount_opt
& OCFS2_MOUNT_GRPQUOTA
&&
1210 !OCFS2_HAS_RO_COMPAT_FEATURE(sb
,
1211 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
)) {
1212 mlog(ML_ERROR
, "Group quotas were requested, but this "
1213 "filesystem does not have the feature enabled.\n");
1216 if (options
->mount_opt
& OCFS2_MOUNT_POSIX_ACL
&&
1217 !OCFS2_HAS_INCOMPAT_FEATURE(sb
, OCFS2_FEATURE_INCOMPAT_XATTR
)) {
1218 mlog(ML_ERROR
, "ACL support requested but extended attributes "
1219 "feature is not enabled\n");
1222 /* No ACL setting specified? Use XATTR feature... */
1223 if (!(options
->mount_opt
& (OCFS2_MOUNT_POSIX_ACL
|
1224 OCFS2_MOUNT_NO_POSIX_ACL
))) {
1225 if (OCFS2_HAS_INCOMPAT_FEATURE(sb
, OCFS2_FEATURE_INCOMPAT_XATTR
))
1226 options
->mount_opt
|= OCFS2_MOUNT_POSIX_ACL
;
1228 options
->mount_opt
|= OCFS2_MOUNT_NO_POSIX_ACL
;
1233 static int ocfs2_parse_options(struct super_block
*sb
,
1235 struct mount_options
*mopt
,
1238 int status
, user_stack
= 0;
1242 substring_t args
[MAX_OPT_ARGS
];
1244 trace_ocfs2_parse_options(is_remount
, options
? options
: "(none)");
1246 mopt
->commit_interval
= 0;
1247 mopt
->mount_opt
= OCFS2_MOUNT_NOINTR
;
1248 mopt
->atime_quantum
= OCFS2_DEFAULT_ATIME_QUANTUM
;
1249 mopt
->slot
= OCFS2_INVALID_SLOT
;
1250 mopt
->localalloc_opt
= -1;
1251 mopt
->cluster_stack
[0] = '\0';
1252 mopt
->resv_level
= OCFS2_DEFAULT_RESV_LEVEL
;
1253 mopt
->dir_resv_level
= -1;
1260 while ((p
= strsep(&options
, ",")) != NULL
) {
1264 token
= match_token(p
, tokens
, args
);
1267 mopt
->mount_opt
|= OCFS2_MOUNT_HB_LOCAL
;
1270 mopt
->mount_opt
|= OCFS2_MOUNT_HB_NONE
;
1273 mopt
->mount_opt
|= OCFS2_MOUNT_HB_GLOBAL
;
1276 if (match_int(&args
[0], &option
)) {
1281 mopt
->mount_opt
|= OCFS2_MOUNT_BARRIER
;
1283 mopt
->mount_opt
&= ~OCFS2_MOUNT_BARRIER
;
1286 mopt
->mount_opt
&= ~OCFS2_MOUNT_NOINTR
;
1289 mopt
->mount_opt
|= OCFS2_MOUNT_NOINTR
;
1292 mopt
->mount_opt
&= ~OCFS2_MOUNT_ERRORS_CONT
;
1293 mopt
->mount_opt
&= ~OCFS2_MOUNT_ERRORS_ROFS
;
1294 mopt
->mount_opt
|= OCFS2_MOUNT_ERRORS_PANIC
;
1297 mopt
->mount_opt
&= ~OCFS2_MOUNT_ERRORS_CONT
;
1298 mopt
->mount_opt
&= ~OCFS2_MOUNT_ERRORS_PANIC
;
1299 mopt
->mount_opt
|= OCFS2_MOUNT_ERRORS_ROFS
;
1302 mopt
->mount_opt
&= ~OCFS2_MOUNT_ERRORS_ROFS
;
1303 mopt
->mount_opt
&= ~OCFS2_MOUNT_ERRORS_PANIC
;
1304 mopt
->mount_opt
|= OCFS2_MOUNT_ERRORS_CONT
;
1306 case Opt_data_ordered
:
1307 mopt
->mount_opt
&= ~OCFS2_MOUNT_DATA_WRITEBACK
;
1309 case Opt_data_writeback
:
1310 mopt
->mount_opt
|= OCFS2_MOUNT_DATA_WRITEBACK
;
1312 case Opt_user_xattr
:
1313 mopt
->mount_opt
&= ~OCFS2_MOUNT_NOUSERXATTR
;
1315 case Opt_nouser_xattr
:
1316 mopt
->mount_opt
|= OCFS2_MOUNT_NOUSERXATTR
;
1318 case Opt_atime_quantum
:
1319 if (match_int(&args
[0], &option
)) {
1324 mopt
->atime_quantum
= option
;
1327 if (match_int(&args
[0], &option
)) {
1332 mopt
->slot
= (u16
)option
;
1335 if (match_int(&args
[0], &option
)) {
1342 option
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1343 mopt
->commit_interval
= HZ
* option
;
1345 case Opt_localalloc
:
1346 if (match_int(&args
[0], &option
)) {
1351 mopt
->localalloc_opt
= option
;
1353 case Opt_localflocks
:
1355 * Changing this during remount could race
1356 * flock() requests, or "unbalance" existing
1357 * ones (e.g., a lock is taken in one mode but
1358 * dropped in the other). If users care enough
1359 * to flip locking modes during remount, we
1360 * could add a "local" flag to individual
1361 * flock structures for proper tracking of
1365 mopt
->mount_opt
|= OCFS2_MOUNT_LOCALFLOCKS
;
1368 /* Check both that the option we were passed
1369 * is of the right length and that it is a proper
1370 * string of the right length.
1372 if (((args
[0].to
- args
[0].from
) !=
1373 OCFS2_STACK_LABEL_LEN
) ||
1374 (strnlen(args
[0].from
,
1375 OCFS2_STACK_LABEL_LEN
) !=
1376 OCFS2_STACK_LABEL_LEN
)) {
1378 "Invalid cluster_stack option\n");
1382 memcpy(mopt
->cluster_stack
, args
[0].from
,
1383 OCFS2_STACK_LABEL_LEN
);
1384 mopt
->cluster_stack
[OCFS2_STACK_LABEL_LEN
] = '\0';
1386 * Open code the memcmp here as we don't have
1388 * ocfs2_userspace_stack().
1390 if (memcmp(mopt
->cluster_stack
,
1391 OCFS2_CLASSIC_CLUSTER_STACK
,
1392 OCFS2_STACK_LABEL_LEN
))
1396 mopt
->mount_opt
|= OCFS2_MOUNT_INODE64
;
1399 mopt
->mount_opt
|= OCFS2_MOUNT_USRQUOTA
;
1402 mopt
->mount_opt
|= OCFS2_MOUNT_GRPQUOTA
;
1404 case Opt_coherency_buffered
:
1405 mopt
->mount_opt
|= OCFS2_MOUNT_COHERENCY_BUFFERED
;
1407 case Opt_coherency_full
:
1408 mopt
->mount_opt
&= ~OCFS2_MOUNT_COHERENCY_BUFFERED
;
1411 mopt
->mount_opt
|= OCFS2_MOUNT_POSIX_ACL
;
1412 mopt
->mount_opt
&= ~OCFS2_MOUNT_NO_POSIX_ACL
;
1415 mopt
->mount_opt
|= OCFS2_MOUNT_NO_POSIX_ACL
;
1416 mopt
->mount_opt
&= ~OCFS2_MOUNT_POSIX_ACL
;
1418 case Opt_resv_level
:
1421 if (match_int(&args
[0], &option
)) {
1425 if (option
>= OCFS2_MIN_RESV_LEVEL
&&
1426 option
< OCFS2_MAX_RESV_LEVEL
)
1427 mopt
->resv_level
= option
;
1429 case Opt_dir_resv_level
:
1432 if (match_int(&args
[0], &option
)) {
1436 if (option
>= OCFS2_MIN_RESV_LEVEL
&&
1437 option
< OCFS2_MAX_RESV_LEVEL
)
1438 mopt
->dir_resv_level
= option
;
1440 case Opt_journal_async_commit
:
1441 mopt
->mount_opt
|= OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT
;
1445 "Unrecognized mount option \"%s\" "
1446 "or missing value\n", p
);
1452 if (user_stack
== 0) {
1453 /* Ensure only one heartbeat mode */
1454 tmp
= mopt
->mount_opt
& (OCFS2_MOUNT_HB_LOCAL
|
1455 OCFS2_MOUNT_HB_GLOBAL
|
1456 OCFS2_MOUNT_HB_NONE
);
1457 if (hweight32(tmp
) != 1) {
1458 mlog(ML_ERROR
, "Invalid heartbeat mount options\n");
1470 static int ocfs2_show_options(struct seq_file
*s
, struct dentry
*root
)
1472 struct ocfs2_super
*osb
= OCFS2_SB(root
->d_sb
);
1473 unsigned long opts
= osb
->s_mount_opt
;
1474 unsigned int local_alloc_megs
;
1476 if (opts
& (OCFS2_MOUNT_HB_LOCAL
| OCFS2_MOUNT_HB_GLOBAL
)) {
1477 seq_printf(s
, ",_netdev");
1478 if (opts
& OCFS2_MOUNT_HB_LOCAL
)
1479 seq_printf(s
, ",%s", OCFS2_HB_LOCAL
);
1481 seq_printf(s
, ",%s", OCFS2_HB_GLOBAL
);
1483 seq_printf(s
, ",%s", OCFS2_HB_NONE
);
1485 if (opts
& OCFS2_MOUNT_NOINTR
)
1486 seq_printf(s
, ",nointr");
1488 if (opts
& OCFS2_MOUNT_DATA_WRITEBACK
)
1489 seq_printf(s
, ",data=writeback");
1491 seq_printf(s
, ",data=ordered");
1493 if (opts
& OCFS2_MOUNT_BARRIER
)
1494 seq_printf(s
, ",barrier=1");
1496 if (opts
& OCFS2_MOUNT_ERRORS_PANIC
)
1497 seq_printf(s
, ",errors=panic");
1498 else if (opts
& OCFS2_MOUNT_ERRORS_CONT
)
1499 seq_printf(s
, ",errors=continue");
1501 seq_printf(s
, ",errors=remount-ro");
1503 if (osb
->preferred_slot
!= OCFS2_INVALID_SLOT
)
1504 seq_printf(s
, ",preferred_slot=%d", osb
->preferred_slot
);
1506 seq_printf(s
, ",atime_quantum=%u", osb
->s_atime_quantum
);
1508 if (osb
->osb_commit_interval
)
1509 seq_printf(s
, ",commit=%u",
1510 (unsigned) (osb
->osb_commit_interval
/ HZ
));
1512 local_alloc_megs
= osb
->local_alloc_bits
>> (20 - osb
->s_clustersize_bits
);
1513 if (local_alloc_megs
!= ocfs2_la_default_mb(osb
))
1514 seq_printf(s
, ",localalloc=%d", local_alloc_megs
);
1516 if (opts
& OCFS2_MOUNT_LOCALFLOCKS
)
1517 seq_printf(s
, ",localflocks,");
1519 if (osb
->osb_cluster_stack
[0])
1520 seq_show_option(s
, "cluster_stack", osb
->osb_cluster_stack
);
1521 if (opts
& OCFS2_MOUNT_USRQUOTA
)
1522 seq_printf(s
, ",usrquota");
1523 if (opts
& OCFS2_MOUNT_GRPQUOTA
)
1524 seq_printf(s
, ",grpquota");
1526 if (opts
& OCFS2_MOUNT_COHERENCY_BUFFERED
)
1527 seq_printf(s
, ",coherency=buffered");
1529 seq_printf(s
, ",coherency=full");
1531 if (opts
& OCFS2_MOUNT_NOUSERXATTR
)
1532 seq_printf(s
, ",nouser_xattr");
1534 seq_printf(s
, ",user_xattr");
1536 if (opts
& OCFS2_MOUNT_INODE64
)
1537 seq_printf(s
, ",inode64");
1539 if (opts
& OCFS2_MOUNT_POSIX_ACL
)
1540 seq_printf(s
, ",acl");
1542 seq_printf(s
, ",noacl");
1544 if (osb
->osb_resv_level
!= OCFS2_DEFAULT_RESV_LEVEL
)
1545 seq_printf(s
, ",resv_level=%d", osb
->osb_resv_level
);
1547 if (osb
->osb_dir_resv_level
!= osb
->osb_resv_level
)
1548 seq_printf(s
, ",dir_resv_level=%d", osb
->osb_resv_level
);
1550 if (opts
& OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT
)
1551 seq_printf(s
, ",journal_async_commit");
1556 static int __init
ocfs2_init(void)
1560 status
= init_ocfs2_uptodate_cache();
1564 status
= ocfs2_initialize_mem_caches();
1568 ocfs2_debugfs_root
= debugfs_create_dir("ocfs2", NULL
);
1570 ocfs2_set_locking_protocol();
1572 status
= register_quota_format(&ocfs2_quota_format
);
1575 status
= register_filesystem(&ocfs2_fs_type
);
1579 unregister_quota_format(&ocfs2_quota_format
);
1581 debugfs_remove(ocfs2_debugfs_root
);
1582 ocfs2_free_mem_caches();
1584 exit_ocfs2_uptodate_cache();
1590 static void __exit
ocfs2_exit(void)
1592 unregister_quota_format(&ocfs2_quota_format
);
1594 debugfs_remove(ocfs2_debugfs_root
);
1596 ocfs2_free_mem_caches();
1598 unregister_filesystem(&ocfs2_fs_type
);
1600 exit_ocfs2_uptodate_cache();
1603 static void ocfs2_put_super(struct super_block
*sb
)
1605 trace_ocfs2_put_super(sb
);
1607 ocfs2_sync_blockdev(sb
);
1608 ocfs2_dismount_volume(sb
, 0);
1611 static int ocfs2_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
1613 struct ocfs2_super
*osb
;
1614 u32 numbits
, freebits
;
1616 struct ocfs2_dinode
*bm_lock
;
1617 struct buffer_head
*bh
= NULL
;
1618 struct inode
*inode
= NULL
;
1620 trace_ocfs2_statfs(dentry
->d_sb
, buf
);
1622 osb
= OCFS2_SB(dentry
->d_sb
);
1624 inode
= ocfs2_get_system_file_inode(osb
,
1625 GLOBAL_BITMAP_SYSTEM_INODE
,
1626 OCFS2_INVALID_SLOT
);
1628 mlog(ML_ERROR
, "failed to get bitmap inode\n");
1633 status
= ocfs2_inode_lock(inode
, &bh
, 0);
1639 bm_lock
= (struct ocfs2_dinode
*) bh
->b_data
;
1641 numbits
= le32_to_cpu(bm_lock
->id1
.bitmap1
.i_total
);
1642 freebits
= numbits
- le32_to_cpu(bm_lock
->id1
.bitmap1
.i_used
);
1644 buf
->f_type
= OCFS2_SUPER_MAGIC
;
1645 buf
->f_bsize
= dentry
->d_sb
->s_blocksize
;
1646 buf
->f_namelen
= OCFS2_MAX_FILENAME_LEN
;
1647 buf
->f_blocks
= ((sector_t
) numbits
) *
1648 (osb
->s_clustersize
>> osb
->sb
->s_blocksize_bits
);
1649 buf
->f_bfree
= ((sector_t
) freebits
) *
1650 (osb
->s_clustersize
>> osb
->sb
->s_blocksize_bits
);
1651 buf
->f_bavail
= buf
->f_bfree
;
1652 buf
->f_files
= numbits
;
1653 buf
->f_ffree
= freebits
;
1654 buf
->f_fsid
.val
[0] = crc32_le(0, osb
->uuid_str
, OCFS2_VOL_UUID_LEN
)
1656 buf
->f_fsid
.val
[1] = crc32_le(0, osb
->uuid_str
+ OCFS2_VOL_UUID_LEN
,
1657 OCFS2_VOL_UUID_LEN
) & 0xFFFFFFFFUL
;
1661 ocfs2_inode_unlock(inode
, 0);
1672 static void ocfs2_inode_init_once(void *data
)
1674 struct ocfs2_inode_info
*oi
= data
;
1677 oi
->ip_open_count
= 0;
1678 spin_lock_init(&oi
->ip_lock
);
1679 ocfs2_extent_map_init(&oi
->vfs_inode
);
1680 INIT_LIST_HEAD(&oi
->ip_io_markers
);
1681 INIT_LIST_HEAD(&oi
->ip_unwritten_list
);
1682 oi
->ip_dir_start_lookup
= 0;
1683 init_rwsem(&oi
->ip_alloc_sem
);
1684 init_rwsem(&oi
->ip_xattr_sem
);
1685 mutex_init(&oi
->ip_io_mutex
);
1687 oi
->ip_blkno
= 0ULL;
1688 oi
->ip_clusters
= 0;
1689 oi
->ip_next_orphan
= NULL
;
1691 ocfs2_resv_init_once(&oi
->ip_la_data_resv
);
1693 ocfs2_lock_res_init_once(&oi
->ip_rw_lockres
);
1694 ocfs2_lock_res_init_once(&oi
->ip_inode_lockres
);
1695 ocfs2_lock_res_init_once(&oi
->ip_open_lockres
);
1697 ocfs2_metadata_cache_init(INODE_CACHE(&oi
->vfs_inode
),
1698 &ocfs2_inode_caching_ops
);
1700 inode_init_once(&oi
->vfs_inode
);
1703 static int ocfs2_initialize_mem_caches(void)
1705 ocfs2_inode_cachep
= kmem_cache_create("ocfs2_inode_cache",
1706 sizeof(struct ocfs2_inode_info
),
1708 (SLAB_HWCACHE_ALIGN
|SLAB_RECLAIM_ACCOUNT
|
1709 SLAB_MEM_SPREAD
|SLAB_ACCOUNT
),
1710 ocfs2_inode_init_once
);
1711 ocfs2_dquot_cachep
= kmem_cache_create("ocfs2_dquot_cache",
1712 sizeof(struct ocfs2_dquot
),
1714 (SLAB_HWCACHE_ALIGN
|SLAB_RECLAIM_ACCOUNT
|
1717 ocfs2_qf_chunk_cachep
= kmem_cache_create("ocfs2_qf_chunk_cache",
1718 sizeof(struct ocfs2_quota_chunk
),
1720 (SLAB_RECLAIM_ACCOUNT
|SLAB_MEM_SPREAD
),
1722 if (!ocfs2_inode_cachep
|| !ocfs2_dquot_cachep
||
1723 !ocfs2_qf_chunk_cachep
) {
1724 kmem_cache_destroy(ocfs2_inode_cachep
);
1725 kmem_cache_destroy(ocfs2_dquot_cachep
);
1726 kmem_cache_destroy(ocfs2_qf_chunk_cachep
);
1733 static void ocfs2_free_mem_caches(void)
1736 * Make sure all delayed rcu free inodes are flushed before we
1740 kmem_cache_destroy(ocfs2_inode_cachep
);
1741 ocfs2_inode_cachep
= NULL
;
1743 kmem_cache_destroy(ocfs2_dquot_cachep
);
1744 ocfs2_dquot_cachep
= NULL
;
1746 kmem_cache_destroy(ocfs2_qf_chunk_cachep
);
1747 ocfs2_qf_chunk_cachep
= NULL
;
1750 static int ocfs2_get_sector(struct super_block
*sb
,
1751 struct buffer_head
**bh
,
1755 if (!sb_set_blocksize(sb
, sect_size
)) {
1756 mlog(ML_ERROR
, "unable to set blocksize\n");
1760 *bh
= sb_getblk(sb
, block
);
1762 mlog_errno(-ENOMEM
);
1766 if (!buffer_dirty(*bh
))
1767 clear_buffer_uptodate(*bh
);
1769 if (bh_read(*bh
, 0) < 0) {
1779 static int ocfs2_mount_volume(struct super_block
*sb
)
1782 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
1784 if (ocfs2_is_hard_readonly(osb
))
1787 mutex_init(&osb
->obs_trim_fs_mutex
);
1789 status
= ocfs2_dlm_init(osb
);
1792 if (status
== -EBADR
&& ocfs2_userspace_stack(osb
))
1793 mlog(ML_ERROR
, "couldn't mount because cluster name on"
1794 " disk does not match the running cluster name.\n");
1798 status
= ocfs2_super_lock(osb
, 1);
1804 /* This will load up the node map and add ourselves to it. */
1805 status
= ocfs2_find_slot(osb
);
1808 goto out_super_lock
;
1811 /* load all node-local system inodes */
1812 status
= ocfs2_init_local_system_inodes(osb
);
1815 goto out_super_lock
;
1818 status
= ocfs2_check_volume(osb
);
1821 goto out_system_inodes
;
1824 status
= ocfs2_truncate_log_init(osb
);
1827 goto out_check_volume
;
1830 ocfs2_super_unlock(osb
, 1);
1834 ocfs2_free_replay_slots(osb
);
1836 if (osb
->local_alloc_state
== OCFS2_LA_ENABLED
)
1837 ocfs2_shutdown_local_alloc(osb
);
1838 ocfs2_release_system_inodes(osb
);
1839 /* before journal shutdown, we should release slot_info */
1840 ocfs2_free_slot_info(osb
);
1841 ocfs2_journal_shutdown(osb
);
1843 ocfs2_super_unlock(osb
, 1);
1845 ocfs2_dlm_shutdown(osb
, 0);
1850 static void ocfs2_dismount_volume(struct super_block
*sb
, int mnt_err
)
1852 int tmp
, hangup_needed
= 0;
1853 struct ocfs2_super
*osb
= NULL
;
1856 trace_ocfs2_dismount_volume(sb
);
1862 /* Remove file check sysfs related directores/files,
1863 * and wait for the pending file check operations */
1864 ocfs2_filecheck_remove_sysfs(osb
);
1866 kset_unregister(osb
->osb_dev_kset
);
1868 /* Orphan scan should be stopped as early as possible */
1869 ocfs2_orphan_scan_stop(osb
);
1871 ocfs2_disable_quotas(osb
);
1873 /* All dquots should be freed by now */
1874 WARN_ON(!llist_empty(&osb
->dquot_drop_list
));
1875 /* Wait for worker to be done with the work structure in osb */
1876 cancel_work_sync(&osb
->dquot_drop_work
);
1878 ocfs2_shutdown_local_alloc(osb
);
1880 ocfs2_truncate_log_shutdown(osb
);
1882 /* This will disable recovery and flush any recovery work. */
1883 ocfs2_recovery_exit(osb
);
1885 ocfs2_sync_blockdev(sb
);
1887 ocfs2_purge_refcount_trees(osb
);
1889 /* No cluster connection means we've failed during mount, so skip
1890 * all the steps which depended on that to complete. */
1892 tmp
= ocfs2_super_lock(osb
, 1);
1899 if (osb
->slot_num
!= OCFS2_INVALID_SLOT
)
1900 ocfs2_put_slot(osb
);
1903 ocfs2_super_unlock(osb
, 1);
1905 ocfs2_release_system_inodes(osb
);
1907 ocfs2_journal_shutdown(osb
);
1910 * If we're dismounting due to mount error, mount.ocfs2 will clean
1911 * up heartbeat. If we're a local mount, there is no heartbeat.
1912 * If we failed before we got a uuid_str yet, we can't stop
1913 * heartbeat. Otherwise, do it.
1915 if (!mnt_err
&& !ocfs2_mount_local(osb
) && osb
->uuid_str
&&
1916 !ocfs2_is_hard_readonly(osb
))
1919 ocfs2_dlm_shutdown(osb
, hangup_needed
);
1921 ocfs2_blockcheck_stats_debugfs_remove(&osb
->osb_ecc_stats
);
1922 debugfs_remove_recursive(osb
->osb_debug_root
);
1925 ocfs2_cluster_hangup(osb
->uuid_str
, strlen(osb
->uuid_str
));
1927 atomic_set(&osb
->vol_state
, VOLUME_DISMOUNTED
);
1929 if (ocfs2_mount_local(osb
))
1930 snprintf(nodestr
, sizeof(nodestr
), "local");
1932 snprintf(nodestr
, sizeof(nodestr
), "%u", osb
->node_num
);
1934 printk(KERN_INFO
"ocfs2: Unmounting device (%s) on (node %s)\n",
1935 osb
->dev_str
, nodestr
);
1937 ocfs2_delete_osb(osb
);
1940 sb
->s_fs_info
= NULL
;
1943 static int ocfs2_setup_osb_uuid(struct ocfs2_super
*osb
, const unsigned char *uuid
,
1944 unsigned uuid_bytes
)
1949 BUG_ON(uuid_bytes
!= OCFS2_VOL_UUID_LEN
);
1951 osb
->uuid_str
= kzalloc(OCFS2_VOL_UUID_LEN
* 2 + 1, GFP_KERNEL
);
1952 if (osb
->uuid_str
== NULL
)
1955 for (i
= 0, ptr
= osb
->uuid_str
; i
< OCFS2_VOL_UUID_LEN
; i
++) {
1956 /* print with null */
1957 ret
= snprintf(ptr
, 3, "%02X", uuid
[i
]);
1958 if (ret
!= 2) /* drop super cleans up */
1960 /* then only advance past the last char */
1967 /* Make sure entire volume is addressable by our journal. Requires
1968 osb_clusters_at_boot to be valid and for the journal to have been
1969 initialized by ocfs2_journal_init(). */
1970 static int ocfs2_journal_addressable(struct ocfs2_super
*osb
)
1974 ocfs2_clusters_to_blocks(osb
->sb
,
1975 osb
->osb_clusters_at_boot
) - 1;
1977 /* 32-bit block number is always OK. */
1978 if (max_block
<= (u32
)~0ULL)
1981 /* Volume is "huge", so see if our journal is new enough to
1983 if (!(OCFS2_HAS_COMPAT_FEATURE(osb
->sb
,
1984 OCFS2_FEATURE_COMPAT_JBD2_SB
) &&
1985 jbd2_journal_check_used_features(osb
->journal
->j_journal
, 0, 0,
1986 JBD2_FEATURE_INCOMPAT_64BIT
))) {
1987 mlog(ML_ERROR
, "The journal cannot address the entire volume. "
1988 "Enable the 'block64' journal option with tunefs.ocfs2");
1997 static int ocfs2_initialize_super(struct super_block
*sb
,
1998 struct buffer_head
*bh
,
2000 struct ocfs2_blockcheck_stats
*stats
)
2003 int i
, cbits
, bbits
;
2004 struct ocfs2_dinode
*di
= (struct ocfs2_dinode
*)bh
->b_data
;
2005 struct inode
*inode
= NULL
;
2006 struct ocfs2_super
*osb
;
2009 osb
= kzalloc(sizeof(struct ocfs2_super
), GFP_KERNEL
);
2016 sb
->s_fs_info
= osb
;
2017 sb
->s_op
= &ocfs2_sops
;
2018 sb
->s_d_op
= &ocfs2_dentry_ops
;
2019 sb
->s_export_op
= &ocfs2_export_ops
;
2020 sb
->s_qcop
= &dquot_quotactl_sysfile_ops
;
2021 sb
->dq_op
= &ocfs2_quota_operations
;
2022 sb
->s_quota_types
= QTYPE_MASK_USR
| QTYPE_MASK_GRP
;
2023 sb
->s_xattr
= ocfs2_xattr_handlers
;
2024 sb
->s_time_gran
= 1;
2025 sb
->s_flags
|= SB_NOATIME
;
2026 /* this is needed to support O_LARGEFILE */
2027 cbits
= le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
);
2028 bbits
= le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
);
2029 sb
->s_maxbytes
= ocfs2_max_file_offset(bbits
, cbits
);
2030 super_set_uuid(sb
, di
->id2
.i_super
.s_uuid
,
2031 sizeof(di
->id2
.i_super
.s_uuid
));
2033 osb
->osb_dx_mask
= (1 << (cbits
- bbits
)) - 1;
2035 for (i
= 0; i
< 3; i
++)
2036 osb
->osb_dx_seed
[i
] = le32_to_cpu(di
->id2
.i_super
.s_dx_seed
[i
]);
2037 osb
->osb_dx_seed
[3] = le32_to_cpu(di
->id2
.i_super
.s_uuid_hash
);
2040 osb
->s_sectsize_bits
= blksize_bits(sector_size
);
2041 BUG_ON(!osb
->s_sectsize_bits
);
2043 spin_lock_init(&osb
->dc_task_lock
);
2044 init_waitqueue_head(&osb
->dc_event
);
2045 osb
->dc_work_sequence
= 0;
2046 osb
->dc_wake_sequence
= 0;
2047 INIT_LIST_HEAD(&osb
->blocked_lock_list
);
2048 osb
->blocked_lock_count
= 0;
2049 spin_lock_init(&osb
->osb_lock
);
2050 spin_lock_init(&osb
->osb_xattr_lock
);
2051 ocfs2_init_steal_slots(osb
);
2053 mutex_init(&osb
->system_file_mutex
);
2055 atomic_set(&osb
->alloc_stats
.moves
, 0);
2056 atomic_set(&osb
->alloc_stats
.local_data
, 0);
2057 atomic_set(&osb
->alloc_stats
.bitmap_data
, 0);
2058 atomic_set(&osb
->alloc_stats
.bg_allocs
, 0);
2059 atomic_set(&osb
->alloc_stats
.bg_extends
, 0);
2061 /* Copy the blockcheck stats from the superblock probe */
2062 osb
->osb_ecc_stats
= *stats
;
2064 ocfs2_init_node_maps(osb
);
2066 snprintf(osb
->dev_str
, sizeof(osb
->dev_str
), "%u,%u",
2067 MAJOR(osb
->sb
->s_dev
), MINOR(osb
->sb
->s_dev
));
2069 osb
->max_slots
= le16_to_cpu(di
->id2
.i_super
.s_max_slots
);
2070 if (osb
->max_slots
> OCFS2_MAX_SLOTS
|| osb
->max_slots
== 0) {
2071 mlog(ML_ERROR
, "Invalid number of node slots (%u)\n",
2077 ocfs2_orphan_scan_init(osb
);
2079 status
= ocfs2_recovery_init(osb
);
2081 mlog(ML_ERROR
, "Unable to initialize recovery state\n");
2086 init_waitqueue_head(&osb
->checkpoint_event
);
2088 osb
->s_atime_quantum
= OCFS2_DEFAULT_ATIME_QUANTUM
;
2090 osb
->slot_num
= OCFS2_INVALID_SLOT
;
2092 osb
->s_xattr_inline_size
= le16_to_cpu(
2093 di
->id2
.i_super
.s_xattr_inline_size
);
2095 osb
->local_alloc_state
= OCFS2_LA_UNUSED
;
2096 osb
->local_alloc_bh
= NULL
;
2097 INIT_DELAYED_WORK(&osb
->la_enable_wq
, ocfs2_la_enable_worker
);
2099 init_waitqueue_head(&osb
->osb_mount_event
);
2101 ocfs2_resmap_init(osb
, &osb
->osb_la_resmap
);
2103 osb
->vol_label
= kmalloc(OCFS2_MAX_VOL_LABEL_LEN
, GFP_KERNEL
);
2104 if (!osb
->vol_label
) {
2105 mlog(ML_ERROR
, "unable to alloc vol label\n");
2107 goto out_recovery_map
;
2110 osb
->slot_recovery_generations
=
2111 kcalloc(osb
->max_slots
, sizeof(*osb
->slot_recovery_generations
),
2113 if (!osb
->slot_recovery_generations
) {
2119 init_waitqueue_head(&osb
->osb_wipe_event
);
2120 osb
->osb_orphan_wipes
= kcalloc(osb
->max_slots
,
2121 sizeof(*osb
->osb_orphan_wipes
),
2123 if (!osb
->osb_orphan_wipes
) {
2126 goto out_slot_recovery_gen
;
2129 osb
->osb_rf_lock_tree
= RB_ROOT
;
2131 osb
->s_feature_compat
=
2132 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_compat
);
2133 osb
->s_feature_ro_compat
=
2134 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_ro_compat
);
2135 osb
->s_feature_incompat
=
2136 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_incompat
);
2138 if ((i
= OCFS2_HAS_INCOMPAT_FEATURE(osb
->sb
, ~OCFS2_FEATURE_INCOMPAT_SUPP
))) {
2139 mlog(ML_ERROR
, "couldn't mount because of unsupported "
2140 "optional features (%x).\n", i
);
2142 goto out_orphan_wipes
;
2144 if (!sb_rdonly(osb
->sb
) && (i
= OCFS2_HAS_RO_COMPAT_FEATURE(osb
->sb
, ~OCFS2_FEATURE_RO_COMPAT_SUPP
))) {
2145 mlog(ML_ERROR
, "couldn't mount RDWR because of "
2146 "unsupported optional features (%x).\n", i
);
2148 goto out_orphan_wipes
;
2151 if (ocfs2_clusterinfo_valid(osb
)) {
2153 * ci_stack and ci_cluster in ocfs2_cluster_info may not be null
2154 * terminated, so make sure no overflow happens here by using
2155 * memcpy. Destination strings will always be null terminated
2156 * because osb is allocated using kzalloc.
2158 osb
->osb_stackflags
=
2159 OCFS2_RAW_SB(di
)->s_cluster_info
.ci_stackflags
;
2160 memcpy(osb
->osb_cluster_stack
,
2161 OCFS2_RAW_SB(di
)->s_cluster_info
.ci_stack
,
2162 OCFS2_STACK_LABEL_LEN
);
2163 if (strlen(osb
->osb_cluster_stack
) != OCFS2_STACK_LABEL_LEN
) {
2165 "couldn't mount because of an invalid "
2166 "cluster stack label (%s) \n",
2167 osb
->osb_cluster_stack
);
2169 goto out_orphan_wipes
;
2171 memcpy(osb
->osb_cluster_name
,
2172 OCFS2_RAW_SB(di
)->s_cluster_info
.ci_cluster
,
2173 OCFS2_CLUSTER_NAME_LEN
);
2175 /* The empty string is identical with classic tools that
2176 * don't know about s_cluster_info. */
2177 osb
->osb_cluster_stack
[0] = '\0';
2180 get_random_bytes(&osb
->s_next_generation
, sizeof(u32
));
2184 * This should be done in ocfs2_journal_init(), but any inode
2185 * writes back operation will cause the filesystem to crash.
2187 status
= ocfs2_journal_alloc(osb
);
2189 goto out_orphan_wipes
;
2191 INIT_WORK(&osb
->dquot_drop_work
, ocfs2_drop_dquot_refs
);
2192 init_llist_head(&osb
->dquot_drop_list
);
2194 /* get some pseudo constants for clustersize bits */
2195 osb
->s_clustersize_bits
=
2196 le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
);
2197 osb
->s_clustersize
= 1 << osb
->s_clustersize_bits
;
2199 if (osb
->s_clustersize
< OCFS2_MIN_CLUSTERSIZE
||
2200 osb
->s_clustersize
> OCFS2_MAX_CLUSTERSIZE
) {
2201 mlog(ML_ERROR
, "Volume has invalid cluster size (%d)\n",
2202 osb
->s_clustersize
);
2207 total_blocks
= ocfs2_clusters_to_blocks(osb
->sb
,
2208 le32_to_cpu(di
->i_clusters
));
2210 status
= generic_check_addressable(osb
->sb
->s_blocksize_bits
,
2213 mlog(ML_ERROR
, "Volume too large "
2214 "to mount safely on this system");
2219 if (ocfs2_setup_osb_uuid(osb
, di
->id2
.i_super
.s_uuid
,
2220 sizeof(di
->id2
.i_super
.s_uuid
))) {
2221 mlog(ML_ERROR
, "Out of memory trying to setup our uuid.\n");
2226 strscpy(osb
->vol_label
, di
->id2
.i_super
.s_label
,
2227 OCFS2_MAX_VOL_LABEL_LEN
);
2228 osb
->root_blkno
= le64_to_cpu(di
->id2
.i_super
.s_root_blkno
);
2229 osb
->system_dir_blkno
= le64_to_cpu(di
->id2
.i_super
.s_system_dir_blkno
);
2230 osb
->first_cluster_group_blkno
=
2231 le64_to_cpu(di
->id2
.i_super
.s_first_cluster_group
);
2232 osb
->fs_generation
= le32_to_cpu(di
->i_fs_generation
);
2233 osb
->uuid_hash
= le32_to_cpu(di
->id2
.i_super
.s_uuid_hash
);
2234 trace_ocfs2_initialize_super(osb
->vol_label
, osb
->uuid_str
,
2235 (unsigned long long)osb
->root_blkno
,
2236 (unsigned long long)osb
->system_dir_blkno
,
2237 osb
->s_clustersize_bits
);
2239 osb
->osb_dlm_debug
= ocfs2_new_dlm_debug();
2240 if (!osb
->osb_dlm_debug
) {
2246 atomic_set(&osb
->vol_state
, VOLUME_INIT
);
2248 /* load root, system_dir, and all global system inodes */
2249 status
= ocfs2_init_global_system_inodes(osb
);
2258 inode
= ocfs2_get_system_file_inode(osb
, GLOBAL_BITMAP_SYSTEM_INODE
,
2259 OCFS2_INVALID_SLOT
);
2263 goto out_system_inodes
;
2266 osb
->bitmap_blkno
= OCFS2_I(inode
)->ip_blkno
;
2267 osb
->osb_clusters_at_boot
= OCFS2_I(inode
)->ip_clusters
;
2270 osb
->bitmap_cpg
= ocfs2_group_bitmap_size(sb
, 0,
2271 osb
->s_feature_incompat
) * 8;
2273 status
= ocfs2_init_slot_info(osb
);
2276 goto out_system_inodes
;
2279 osb
->ocfs2_wq
= alloc_ordered_workqueue("ocfs2_wq", WQ_MEM_RECLAIM
);
2280 if (!osb
->ocfs2_wq
) {
2289 ocfs2_free_slot_info(osb
);
2291 ocfs2_release_system_inodes(osb
);
2293 ocfs2_put_dlm_debug(osb
->osb_dlm_debug
);
2295 kfree(osb
->uuid_str
);
2297 kfree(osb
->journal
);
2299 kfree(osb
->osb_orphan_wipes
);
2300 out_slot_recovery_gen
:
2301 kfree(osb
->slot_recovery_generations
);
2303 kfree(osb
->vol_label
);
2305 kfree(osb
->recovery_map
);
2308 sb
->s_fs_info
= NULL
;
2313 * will return: -EAGAIN if it is ok to keep searching for superblocks
2314 * -EINVAL if there is a bad superblock
2317 static int ocfs2_verify_volume(struct ocfs2_dinode
*di
,
2318 struct buffer_head
*bh
,
2320 struct ocfs2_blockcheck_stats
*stats
)
2322 int status
= -EAGAIN
;
2324 if (memcmp(di
->i_signature
, OCFS2_SUPER_BLOCK_SIGNATURE
,
2325 strlen(OCFS2_SUPER_BLOCK_SIGNATURE
)) == 0) {
2326 /* We have to do a raw check of the feature here */
2327 if (le32_to_cpu(di
->id2
.i_super
.s_feature_incompat
) &
2328 OCFS2_FEATURE_INCOMPAT_META_ECC
) {
2329 status
= ocfs2_block_check_validate(bh
->b_data
,
2337 if ((1 << le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
)) != blksz
) {
2338 mlog(ML_ERROR
, "found superblock with incorrect block "
2339 "size: found %u, should be %u\n",
2340 1 << le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
),
2342 } else if (le16_to_cpu(di
->id2
.i_super
.s_major_rev_level
) !=
2343 OCFS2_MAJOR_REV_LEVEL
||
2344 le16_to_cpu(di
->id2
.i_super
.s_minor_rev_level
) !=
2345 OCFS2_MINOR_REV_LEVEL
) {
2346 mlog(ML_ERROR
, "found superblock with bad version: "
2347 "found %u.%u, should be %u.%u\n",
2348 le16_to_cpu(di
->id2
.i_super
.s_major_rev_level
),
2349 le16_to_cpu(di
->id2
.i_super
.s_minor_rev_level
),
2350 OCFS2_MAJOR_REV_LEVEL
,
2351 OCFS2_MINOR_REV_LEVEL
);
2352 } else if (bh
->b_blocknr
!= le64_to_cpu(di
->i_blkno
)) {
2353 mlog(ML_ERROR
, "bad block number on superblock: "
2354 "found %llu, should be %llu\n",
2355 (unsigned long long)le64_to_cpu(di
->i_blkno
),
2356 (unsigned long long)bh
->b_blocknr
);
2357 } else if (le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
) < 12 ||
2358 le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
) > 20) {
2359 mlog(ML_ERROR
, "bad cluster size found: %u\n",
2360 1 << le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
));
2361 } else if (!le64_to_cpu(di
->id2
.i_super
.s_root_blkno
)) {
2362 mlog(ML_ERROR
, "bad root_blkno: 0\n");
2363 } else if (!le64_to_cpu(di
->id2
.i_super
.s_system_dir_blkno
)) {
2364 mlog(ML_ERROR
, "bad system_dir_blkno: 0\n");
2365 } else if (le16_to_cpu(di
->id2
.i_super
.s_max_slots
) > OCFS2_MAX_SLOTS
) {
2367 "Superblock slots found greater than file system "
2368 "maximum: found %u, max %u\n",
2369 le16_to_cpu(di
->id2
.i_super
.s_max_slots
),
2378 if (status
&& status
!= -EAGAIN
)
2383 static int ocfs2_check_volume(struct ocfs2_super
*osb
)
2388 struct ocfs2_dinode
*local_alloc
= NULL
; /* only used if we
2392 /* Init our journal object. */
2393 status
= ocfs2_journal_init(osb
, &dirty
);
2395 mlog(ML_ERROR
, "Could not initialize journal!\n");
2399 /* Now that journal has been initialized, check to make sure
2400 entire volume is addressable. */
2401 status
= ocfs2_journal_addressable(osb
);
2405 /* If the journal was unmounted cleanly then we don't want to
2406 * recover anything. Otherwise, journal_load will do that
2407 * dirty work for us :) */
2409 status
= ocfs2_journal_wipe(osb
->journal
, 0);
2415 printk(KERN_NOTICE
"ocfs2: File system on device (%s) was not "
2416 "unmounted cleanly, recovering it.\n", osb
->dev_str
);
2419 local
= ocfs2_mount_local(osb
);
2421 /* will play back anything left in the journal. */
2422 status
= ocfs2_journal_load(osb
->journal
, local
, dirty
);
2424 mlog(ML_ERROR
, "ocfs2 journal load failed! %d\n", status
);
2428 if (osb
->s_mount_opt
& OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT
)
2429 jbd2_journal_set_features(osb
->journal
->j_journal
,
2430 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2431 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2433 jbd2_journal_clear_features(osb
->journal
->j_journal
,
2434 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2435 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2438 /* recover my local alloc if we didn't unmount cleanly. */
2439 status
= ocfs2_begin_local_alloc_recovery(osb
,
2446 /* we complete the recovery process after we've marked
2447 * ourselves as mounted. */
2450 status
= ocfs2_load_local_alloc(osb
);
2457 /* Recovery will be completed after we've mounted the
2458 * rest of the volume. */
2459 osb
->local_alloc_copy
= local_alloc
;
2463 /* go through each journal, trylock it and if you get the
2464 * lock, and it's marked as dirty, set the bit in the recover
2465 * map and launch a recovery thread for it. */
2466 status
= ocfs2_mark_dead_nodes(osb
);
2472 status
= ocfs2_compute_replay_slots(osb
);
2485 * The routine gets called from dismount or close whenever a dismount on
2486 * volume is requested and the osb open count becomes 1.
2487 * It will remove the osb from the global list and also free up all the
2488 * initialized resources and fileobject.
2490 static void ocfs2_delete_osb(struct ocfs2_super
*osb
)
2492 /* This function assumes that the caller has the main osb resource */
2494 /* ocfs2_initializer_super have already created this workqueue */
2496 destroy_workqueue(osb
->ocfs2_wq
);
2498 ocfs2_free_slot_info(osb
);
2500 kfree(osb
->osb_orphan_wipes
);
2501 kfree(osb
->slot_recovery_generations
);
2503 * This belongs in journal shutdown, but because we have to
2504 * allocate osb->journal at the middle of ocfs2_initialize_super(),
2507 kfree(osb
->journal
);
2508 kfree(osb
->local_alloc_copy
);
2509 kfree(osb
->uuid_str
);
2510 kfree(osb
->vol_label
);
2511 ocfs2_put_dlm_debug(osb
->osb_dlm_debug
);
2512 memset(osb
, 0, sizeof(struct ocfs2_super
));
2515 /* Depending on the mount option passed, perform one of the following:
2516 * Put OCFS2 into a readonly state (default)
2517 * Return EIO so that only the process errs
2518 * Fix the error as if fsck.ocfs2 -y
2521 static int ocfs2_handle_error(struct super_block
*sb
)
2523 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
2526 ocfs2_set_osb_flag(osb
, OCFS2_OSB_ERROR_FS
);
2527 pr_crit("On-disk corruption discovered. "
2528 "Please run fsck.ocfs2 once the filesystem is unmounted.\n");
2530 if (osb
->s_mount_opt
& OCFS2_MOUNT_ERRORS_PANIC
) {
2531 panic("OCFS2: (device %s): panic forced after error\n",
2533 } else if (osb
->s_mount_opt
& OCFS2_MOUNT_ERRORS_CONT
) {
2534 pr_crit("OCFS2: Returning error to the calling process.\n");
2536 } else { /* default option */
2538 if (sb_rdonly(sb
) && (ocfs2_is_soft_readonly(osb
) || ocfs2_is_hard_readonly(osb
)))
2541 pr_crit("OCFS2: File system is now read-only.\n");
2542 sb
->s_flags
|= SB_RDONLY
;
2543 ocfs2_set_ro_flag(osb
, 0);
2549 int __ocfs2_error(struct super_block
*sb
, const char *function
,
2550 const char *fmt
, ...)
2552 struct va_format vaf
;
2555 va_start(args
, fmt
);
2559 /* Not using mlog here because we want to show the actual
2560 * function the error came from. */
2561 printk(KERN_CRIT
"OCFS2: ERROR (device %s): %s: %pV",
2562 sb
->s_id
, function
, &vaf
);
2566 return ocfs2_handle_error(sb
);
2569 /* Handle critical errors. This is intentionally more drastic than
2570 * ocfs2_handle_error, so we only use for things like journal errors,
2572 void __ocfs2_abort(struct super_block
*sb
, const char *function
,
2573 const char *fmt
, ...)
2575 struct va_format vaf
;
2578 va_start(args
, fmt
);
2583 printk(KERN_CRIT
"OCFS2: abort (device %s): %s: %pV",
2584 sb
->s_id
, function
, &vaf
);
2588 /* We don't have the cluster support yet to go straight to
2589 * hard readonly in here. Until then, we want to keep
2590 * ocfs2_abort() so that we can at least mark critical
2593 * TODO: This should abort the journal and alert other nodes
2594 * that our slot needs recovery. */
2596 /* Force a panic(). This stinks, but it's better than letting
2597 * things continue without having a proper hard readonly
2599 if (!ocfs2_mount_local(OCFS2_SB(sb
)))
2600 OCFS2_SB(sb
)->s_mount_opt
|= OCFS2_MOUNT_ERRORS_PANIC
;
2601 ocfs2_handle_error(sb
);
2605 * Void signal blockers, because in-kernel sigprocmask() only fails
2606 * when SIG_* is wrong.
2608 void ocfs2_block_signals(sigset_t
*oldset
)
2613 sigfillset(&blocked
);
2614 rc
= sigprocmask(SIG_BLOCK
, &blocked
, oldset
);
2618 void ocfs2_unblock_signals(sigset_t
*oldset
)
2620 int rc
= sigprocmask(SIG_SETMASK
, oldset
, NULL
);
2624 module_init(ocfs2_init
);
2625 module_exit(ocfs2_exit
);