1 /* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
6 * load/unload driver, mount/dismount volumes
8 * Copyright (C) 2002, 2004 Oracle. All rights reserved.
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public
12 * License as published by the Free Software Foundation; either
13 * version 2 of the License, or (at your option) any later version.
15 * This program is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * General Public License for more details.
20 * You should have received a copy of the GNU General Public
21 * License along with this program; if not, write to the
22 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
23 * Boston, MA 021110-1307, USA.
26 #include <linux/module.h>
28 #include <linux/types.h>
29 #include <linux/slab.h>
30 #include <linux/highmem.h>
31 #include <linux/utsname.h>
32 #include <linux/init.h>
33 #include <linux/random.h>
34 #include <linux/statfs.h>
35 #include <linux/moduleparam.h>
36 #include <linux/blkdev.h>
37 #include <linux/socket.h>
38 #include <linux/inet.h>
39 #include <linux/parser.h>
40 #include <linux/crc32.h>
41 #include <linux/debugfs.h>
42 #include <linux/mount.h>
43 #include <linux/seq_file.h>
45 #define MLOG_MASK_PREFIX ML_SUPER
46 #include <cluster/masklog.h>
50 /* this should be the only file to include a version 1 header */
51 #include "ocfs1_fs_compat.h"
56 #include "extent_map.h"
57 #include "heartbeat.h"
60 #include "localalloc.h"
69 #include "buffer_head_io.h"
71 static struct kmem_cache
*ocfs2_inode_cachep
= NULL
;
73 /* OCFS2 needs to schedule several differnt types of work which
74 * require cluster locking, disk I/O, recovery waits, etc. Since these
75 * types of work tend to be heavy we avoid using the kernel events
76 * workqueue and schedule on our own. */
77 struct workqueue_struct
*ocfs2_wq
= NULL
;
79 static struct dentry
*ocfs2_debugfs_root
= NULL
;
81 MODULE_AUTHOR("Oracle");
82 MODULE_LICENSE("GPL");
86 unsigned long commit_interval
;
87 unsigned long mount_opt
;
88 unsigned int atime_quantum
;
90 unsigned int localalloc_opt
;
91 char cluster_stack
[OCFS2_STACK_LABEL_LEN
+ 1];
94 static int ocfs2_parse_options(struct super_block
*sb
, char *options
,
95 struct mount_options
*mopt
,
97 static int ocfs2_show_options(struct seq_file
*s
, struct vfsmount
*mnt
);
98 static void ocfs2_put_super(struct super_block
*sb
);
99 static int ocfs2_mount_volume(struct super_block
*sb
);
100 static int ocfs2_remount(struct super_block
*sb
, int *flags
, char *data
);
101 static void ocfs2_dismount_volume(struct super_block
*sb
, int mnt_err
);
102 static int ocfs2_initialize_mem_caches(void);
103 static void ocfs2_free_mem_caches(void);
104 static void ocfs2_delete_osb(struct ocfs2_super
*osb
);
106 static int ocfs2_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
108 static int ocfs2_sync_fs(struct super_block
*sb
, int wait
);
110 static int ocfs2_init_global_system_inodes(struct ocfs2_super
*osb
);
111 static int ocfs2_init_local_system_inodes(struct ocfs2_super
*osb
);
112 static void ocfs2_release_system_inodes(struct ocfs2_super
*osb
);
113 static int ocfs2_check_volume(struct ocfs2_super
*osb
);
114 static int ocfs2_verify_volume(struct ocfs2_dinode
*di
,
115 struct buffer_head
*bh
,
117 static int ocfs2_initialize_super(struct super_block
*sb
,
118 struct buffer_head
*bh
,
120 static int ocfs2_get_sector(struct super_block
*sb
,
121 struct buffer_head
**bh
,
124 static void ocfs2_write_super(struct super_block
*sb
);
125 static struct inode
*ocfs2_alloc_inode(struct super_block
*sb
);
126 static void ocfs2_destroy_inode(struct inode
*inode
);
128 static const struct super_operations ocfs2_sops
= {
129 .statfs
= ocfs2_statfs
,
130 .alloc_inode
= ocfs2_alloc_inode
,
131 .destroy_inode
= ocfs2_destroy_inode
,
132 .drop_inode
= ocfs2_drop_inode
,
133 .clear_inode
= ocfs2_clear_inode
,
134 .delete_inode
= ocfs2_delete_inode
,
135 .sync_fs
= ocfs2_sync_fs
,
136 .write_super
= ocfs2_write_super
,
137 .put_super
= ocfs2_put_super
,
138 .remount_fs
= ocfs2_remount
,
139 .show_options
= ocfs2_show_options
,
163 static const match_table_t tokens
= {
164 {Opt_barrier
, "barrier=%u"},
165 {Opt_err_panic
, "errors=panic"},
166 {Opt_err_ro
, "errors=remount-ro"},
168 {Opt_nointr
, "nointr"},
169 {Opt_hb_none
, OCFS2_HB_NONE
},
170 {Opt_hb_local
, OCFS2_HB_LOCAL
},
171 {Opt_data_ordered
, "data=ordered"},
172 {Opt_data_writeback
, "data=writeback"},
173 {Opt_atime_quantum
, "atime_quantum=%u"},
174 {Opt_slot
, "preferred_slot=%u"},
175 {Opt_commit
, "commit=%u"},
176 {Opt_localalloc
, "localalloc=%d"},
177 {Opt_localflocks
, "localflocks"},
178 {Opt_stack
, "cluster_stack=%s"},
179 {Opt_user_xattr
, "user_xattr"},
180 {Opt_nouser_xattr
, "nouser_xattr"},
185 * write_super and sync_fs ripped right out of ext3.
187 static void ocfs2_write_super(struct super_block
*sb
)
189 if (mutex_trylock(&sb
->s_lock
) != 0)
194 static int ocfs2_sync_fs(struct super_block
*sb
, int wait
)
198 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
202 if (ocfs2_is_hard_readonly(osb
))
206 status
= ocfs2_flush_truncate_log(osb
);
210 ocfs2_schedule_truncate_log_flush(osb
, 0);
213 if (journal_start_commit(OCFS2_SB(sb
)->journal
->j_journal
, &target
)) {
215 log_wait_commit(OCFS2_SB(sb
)->journal
->j_journal
,
221 static int ocfs2_init_global_system_inodes(struct ocfs2_super
*osb
)
223 struct inode
*new = NULL
;
229 new = ocfs2_iget(osb
, osb
->root_blkno
, OCFS2_FI_FLAG_SYSFILE
, 0);
231 status
= PTR_ERR(new);
235 osb
->root_inode
= new;
237 new = ocfs2_iget(osb
, osb
->system_dir_blkno
, OCFS2_FI_FLAG_SYSFILE
, 0);
239 status
= PTR_ERR(new);
243 osb
->sys_root_inode
= new;
245 for (i
= OCFS2_FIRST_ONLINE_SYSTEM_INODE
;
246 i
<= OCFS2_LAST_GLOBAL_SYSTEM_INODE
; i
++) {
247 new = ocfs2_get_system_file_inode(osb
, i
, osb
->slot_num
);
249 ocfs2_release_system_inodes(osb
);
252 /* FIXME: Should ERROR_RO_FS */
253 mlog(ML_ERROR
, "Unable to load system inode %d, "
254 "possibly corrupt fs?", i
);
257 // the array now has one ref, so drop this one
266 static int ocfs2_init_local_system_inodes(struct ocfs2_super
*osb
)
268 struct inode
*new = NULL
;
274 for (i
= OCFS2_LAST_GLOBAL_SYSTEM_INODE
+ 1;
275 i
< NUM_SYSTEM_INODES
;
277 new = ocfs2_get_system_file_inode(osb
, i
, osb
->slot_num
);
279 ocfs2_release_system_inodes(osb
);
281 mlog(ML_ERROR
, "status=%d, sysfile=%d, slot=%d\n",
282 status
, i
, osb
->slot_num
);
285 /* the array now has one ref, so drop this one */
294 static void ocfs2_release_system_inodes(struct ocfs2_super
*osb
)
301 for (i
= 0; i
< NUM_SYSTEM_INODES
; i
++) {
302 inode
= osb
->system_inodes
[i
];
305 osb
->system_inodes
[i
] = NULL
;
309 inode
= osb
->sys_root_inode
;
312 osb
->sys_root_inode
= NULL
;
315 inode
= osb
->root_inode
;
318 osb
->root_inode
= NULL
;
324 /* We're allocating fs objects, use GFP_NOFS */
325 static struct inode
*ocfs2_alloc_inode(struct super_block
*sb
)
327 struct ocfs2_inode_info
*oi
;
329 oi
= kmem_cache_alloc(ocfs2_inode_cachep
, GFP_NOFS
);
333 return &oi
->vfs_inode
;
336 static void ocfs2_destroy_inode(struct inode
*inode
)
338 kmem_cache_free(ocfs2_inode_cachep
, OCFS2_I(inode
));
341 static unsigned long long ocfs2_max_file_offset(unsigned int bbits
,
344 unsigned int bytes
= 1 << cbits
;
345 unsigned int trim
= bytes
;
346 unsigned int bitshift
= 32;
349 * i_size and all block offsets in ocfs2 are always 64 bits
350 * wide. i_clusters is 32 bits, in cluster-sized units. So on
351 * 64 bit platforms, cluster size will be the limiting factor.
354 #if BITS_PER_LONG == 32
355 # if defined(CONFIG_LBD)
356 BUILD_BUG_ON(sizeof(sector_t
) != 8);
358 * We might be limited by page cache size.
360 if (bytes
> PAGE_CACHE_SIZE
) {
361 bytes
= PAGE_CACHE_SIZE
;
364 * Shift by 31 here so that we don't get larger than
371 * We are limited by the size of sector_t. Use block size, as
372 * that's what we expose to the VFS.
381 * Trim by a whole cluster when we can actually approach the
382 * on-disk limits. Otherwise we can overflow i_clusters when
383 * an extent start is at the max offset.
385 return (((unsigned long long)bytes
) << bitshift
) - trim
;
388 static int ocfs2_remount(struct super_block
*sb
, int *flags
, char *data
)
390 int incompat_features
;
392 struct mount_options parsed_options
;
393 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
395 if (!ocfs2_parse_options(sb
, data
, &parsed_options
, 1)) {
400 if ((osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) !=
401 (parsed_options
.mount_opt
& OCFS2_MOUNT_HB_LOCAL
)) {
403 mlog(ML_ERROR
, "Cannot change heartbeat mode on remount\n");
407 if ((osb
->s_mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
) !=
408 (parsed_options
.mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
)) {
410 mlog(ML_ERROR
, "Cannot change data mode on remount\n");
414 /* We're going to/from readonly mode. */
415 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
)) {
416 /* Lock here so the check of HARD_RO and the potential
417 * setting of SOFT_RO is atomic. */
418 spin_lock(&osb
->osb_lock
);
419 if (osb
->osb_flags
& OCFS2_OSB_HARD_RO
) {
420 mlog(ML_ERROR
, "Remount on readonly device is forbidden.\n");
425 if (*flags
& MS_RDONLY
) {
426 mlog(0, "Going to ro mode.\n");
427 sb
->s_flags
|= MS_RDONLY
;
428 osb
->osb_flags
|= OCFS2_OSB_SOFT_RO
;
430 mlog(0, "Making ro filesystem writeable.\n");
432 if (osb
->osb_flags
& OCFS2_OSB_ERROR_FS
) {
433 mlog(ML_ERROR
, "Cannot remount RDWR "
434 "filesystem due to previous errors.\n");
438 incompat_features
= OCFS2_HAS_RO_COMPAT_FEATURE(sb
, ~OCFS2_FEATURE_RO_COMPAT_SUPP
);
439 if (incompat_features
) {
440 mlog(ML_ERROR
, "Cannot remount RDWR because "
441 "of unsupported optional features "
442 "(%x).\n", incompat_features
);
446 sb
->s_flags
&= ~MS_RDONLY
;
447 osb
->osb_flags
&= ~OCFS2_OSB_SOFT_RO
;
450 spin_unlock(&osb
->osb_lock
);
454 /* Only save off the new mount options in case of a successful
456 osb
->s_mount_opt
= parsed_options
.mount_opt
;
457 osb
->s_atime_quantum
= parsed_options
.atime_quantum
;
458 osb
->preferred_slot
= parsed_options
.slot
;
459 if (parsed_options
.commit_interval
)
460 osb
->osb_commit_interval
= parsed_options
.commit_interval
;
462 if (!ocfs2_is_hard_readonly(osb
))
463 ocfs2_set_journal_params(osb
);
469 static int ocfs2_sb_probe(struct super_block
*sb
,
470 struct buffer_head
**bh
,
474 struct ocfs1_vol_disk_hdr
*hdr
;
475 struct ocfs2_dinode
*di
;
481 *sector_size
= bdev_hardsect_size(sb
->s_bdev
);
482 if (*sector_size
> OCFS2_MAX_BLOCKSIZE
) {
483 mlog(ML_ERROR
, "Hardware sector size too large: %d (max=%d)\n",
484 *sector_size
, OCFS2_MAX_BLOCKSIZE
);
489 /* Can this really happen? */
490 if (*sector_size
< OCFS2_MIN_BLOCKSIZE
)
491 *sector_size
= OCFS2_MIN_BLOCKSIZE
;
493 /* check block zero for old format */
494 status
= ocfs2_get_sector(sb
, bh
, 0, *sector_size
);
499 hdr
= (struct ocfs1_vol_disk_hdr
*) (*bh
)->b_data
;
500 if (hdr
->major_version
== OCFS1_MAJOR_VERSION
) {
501 mlog(ML_ERROR
, "incompatible version: %u.%u\n",
502 hdr
->major_version
, hdr
->minor_version
);
505 if (memcmp(hdr
->signature
, OCFS1_VOLUME_SIGNATURE
,
506 strlen(OCFS1_VOLUME_SIGNATURE
)) == 0) {
507 mlog(ML_ERROR
, "incompatible volume signature: %8s\n",
514 mlog(ML_ERROR
, "This is an ocfs v1 filesystem which must be "
515 "upgraded before mounting with ocfs v2\n");
520 * Now check at magic offset for 512, 1024, 2048, 4096
521 * blocksizes. 4096 is the maximum blocksize because it is
522 * the minimum clustersize.
525 for (blksize
= *sector_size
;
526 blksize
<= OCFS2_MAX_BLOCKSIZE
;
528 tmpstat
= ocfs2_get_sector(sb
, bh
,
529 OCFS2_SUPER_BLOCK_BLKNO
,
536 di
= (struct ocfs2_dinode
*) (*bh
)->b_data
;
537 status
= ocfs2_verify_volume(di
, *bh
, blksize
);
542 if (status
!= -EAGAIN
)
550 static int ocfs2_verify_heartbeat(struct ocfs2_super
*osb
)
552 if (ocfs2_mount_local(osb
)) {
553 if (osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) {
554 mlog(ML_ERROR
, "Cannot heartbeat on a locally "
555 "mounted device.\n");
560 if (ocfs2_userspace_stack(osb
)) {
561 if (osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) {
562 mlog(ML_ERROR
, "Userspace stack expected, but "
563 "o2cb heartbeat arguments passed to mount\n");
568 if (!(osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
)) {
569 if (!ocfs2_mount_local(osb
) && !ocfs2_is_hard_readonly(osb
) &&
570 !ocfs2_userspace_stack(osb
)) {
571 mlog(ML_ERROR
, "Heartbeat has to be started to mount "
572 "a read-write clustered device.\n");
581 * If we're using a userspace stack, mount should have passed
582 * a name that matches the disk. If not, mount should not
583 * have passed a stack.
585 static int ocfs2_verify_userspace_stack(struct ocfs2_super
*osb
,
586 struct mount_options
*mopt
)
588 if (!ocfs2_userspace_stack(osb
) && mopt
->cluster_stack
[0]) {
590 "cluster stack passed to mount, but this filesystem "
591 "does not support it\n");
595 if (ocfs2_userspace_stack(osb
) &&
596 strncmp(osb
->osb_cluster_stack
, mopt
->cluster_stack
,
597 OCFS2_STACK_LABEL_LEN
)) {
599 "cluster stack passed to mount (\"%s\") does not "
600 "match the filesystem (\"%s\")\n",
602 osb
->osb_cluster_stack
);
609 static int ocfs2_fill_super(struct super_block
*sb
, void *data
, int silent
)
612 int status
, sector_size
;
613 struct mount_options parsed_options
;
614 struct inode
*inode
= NULL
;
615 struct ocfs2_super
*osb
= NULL
;
616 struct buffer_head
*bh
= NULL
;
619 mlog_entry("%p, %p, %i", sb
, data
, silent
);
621 if (!ocfs2_parse_options(sb
, data
, &parsed_options
, 0)) {
623 goto read_super_error
;
626 /* probe for superblock */
627 status
= ocfs2_sb_probe(sb
, &bh
, §or_size
);
629 mlog(ML_ERROR
, "superblock probe failed!\n");
630 goto read_super_error
;
633 status
= ocfs2_initialize_super(sb
, bh
, sector_size
);
637 goto read_super_error
;
641 osb
->s_mount_opt
= parsed_options
.mount_opt
;
642 osb
->s_atime_quantum
= parsed_options
.atime_quantum
;
643 osb
->preferred_slot
= parsed_options
.slot
;
644 osb
->osb_commit_interval
= parsed_options
.commit_interval
;
645 osb
->local_alloc_default_bits
= ocfs2_megabytes_to_clusters(sb
, parsed_options
.localalloc_opt
);
646 osb
->local_alloc_bits
= osb
->local_alloc_default_bits
;
648 status
= ocfs2_verify_userspace_stack(osb
, &parsed_options
);
650 goto read_super_error
;
652 sb
->s_magic
= OCFS2_SUPER_MAGIC
;
654 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
656 if (bdev_read_only(sb
->s_bdev
)) {
657 if (!(sb
->s_flags
& MS_RDONLY
)) {
659 mlog(ML_ERROR
, "Readonly device detected but readonly "
660 "mount was not specified.\n");
661 goto read_super_error
;
664 /* You should not be able to start a local heartbeat
665 * on a readonly device. */
666 if (osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) {
668 mlog(ML_ERROR
, "Local heartbeat specified on readonly "
670 goto read_super_error
;
673 status
= ocfs2_check_journals_nolocks(osb
);
675 if (status
== -EROFS
)
676 mlog(ML_ERROR
, "Recovery required on readonly "
677 "file system, but write access is "
681 goto read_super_error
;
684 ocfs2_set_ro_flag(osb
, 1);
686 printk(KERN_NOTICE
"Readonly device detected. No cluster "
687 "services will be utilized for this mount. Recovery "
688 "will be skipped.\n");
691 if (!ocfs2_is_hard_readonly(osb
)) {
692 if (sb
->s_flags
& MS_RDONLY
)
693 ocfs2_set_ro_flag(osb
, 0);
696 status
= ocfs2_verify_heartbeat(osb
);
699 goto read_super_error
;
702 osb
->osb_debug_root
= debugfs_create_dir(osb
->uuid_str
,
704 if (!osb
->osb_debug_root
) {
706 mlog(ML_ERROR
, "Unable to create per-mount debugfs root.\n");
707 goto read_super_error
;
710 status
= ocfs2_mount_volume(sb
);
712 inode
= igrab(osb
->root_inode
);
715 goto read_super_error
;
720 goto read_super_error
;
723 root
= d_alloc_root(inode
);
727 goto read_super_error
;
732 ocfs2_complete_mount_recovery(osb
);
734 if (ocfs2_mount_local(osb
))
735 snprintf(nodestr
, sizeof(nodestr
), "local");
737 snprintf(nodestr
, sizeof(nodestr
), "%u", osb
->node_num
);
739 printk(KERN_INFO
"ocfs2: Mounting device (%s) on (node %s, slot %d) "
740 "with %s data mode.\n",
741 osb
->dev_str
, nodestr
, osb
->slot_num
,
742 osb
->s_mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
? "writeback" :
745 atomic_set(&osb
->vol_state
, VOLUME_MOUNTED
);
746 wake_up(&osb
->osb_mount_event
);
759 atomic_set(&osb
->vol_state
, VOLUME_DISABLED
);
760 wake_up(&osb
->osb_mount_event
);
761 ocfs2_dismount_volume(sb
, 1);
768 static int ocfs2_get_sb(struct file_system_type
*fs_type
,
770 const char *dev_name
,
772 struct vfsmount
*mnt
)
774 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, ocfs2_fill_super
,
778 static struct file_system_type ocfs2_fs_type
= {
779 .owner
= THIS_MODULE
,
781 .get_sb
= ocfs2_get_sb
, /* is this called when we mount
783 .kill_sb
= kill_block_super
, /* set to the generic one
784 * right now, but do we
785 * need to change that? */
786 .fs_flags
= FS_REQUIRES_DEV
|FS_RENAME_DOES_D_MOVE
,
790 static int ocfs2_parse_options(struct super_block
*sb
,
792 struct mount_options
*mopt
,
798 mlog_entry("remount: %d, options: \"%s\"\n", is_remount
,
799 options
? options
: "(none)");
801 mopt
->commit_interval
= 0;
803 mopt
->atime_quantum
= OCFS2_DEFAULT_ATIME_QUANTUM
;
804 mopt
->slot
= OCFS2_INVALID_SLOT
;
805 mopt
->localalloc_opt
= OCFS2_DEFAULT_LOCAL_ALLOC_SIZE
;
806 mopt
->cluster_stack
[0] = '\0';
813 while ((p
= strsep(&options
, ",")) != NULL
) {
815 substring_t args
[MAX_OPT_ARGS
];
820 token
= match_token(p
, tokens
, args
);
823 mopt
->mount_opt
|= OCFS2_MOUNT_HB_LOCAL
;
826 mopt
->mount_opt
&= ~OCFS2_MOUNT_HB_LOCAL
;
829 if (match_int(&args
[0], &option
)) {
834 mopt
->mount_opt
|= OCFS2_MOUNT_BARRIER
;
836 mopt
->mount_opt
&= ~OCFS2_MOUNT_BARRIER
;
839 mopt
->mount_opt
&= ~OCFS2_MOUNT_NOINTR
;
842 mopt
->mount_opt
|= OCFS2_MOUNT_NOINTR
;
845 mopt
->mount_opt
|= OCFS2_MOUNT_ERRORS_PANIC
;
848 mopt
->mount_opt
&= ~OCFS2_MOUNT_ERRORS_PANIC
;
850 case Opt_data_ordered
:
851 mopt
->mount_opt
&= ~OCFS2_MOUNT_DATA_WRITEBACK
;
853 case Opt_data_writeback
:
854 mopt
->mount_opt
|= OCFS2_MOUNT_DATA_WRITEBACK
;
857 mopt
->mount_opt
&= ~OCFS2_MOUNT_NOUSERXATTR
;
859 case Opt_nouser_xattr
:
860 mopt
->mount_opt
|= OCFS2_MOUNT_NOUSERXATTR
;
862 case Opt_atime_quantum
:
863 if (match_int(&args
[0], &option
)) {
868 mopt
->atime_quantum
= option
;
872 if (match_int(&args
[0], &option
)) {
877 mopt
->slot
= (s16
)option
;
881 if (match_int(&args
[0], &option
)) {
888 option
= JBD_DEFAULT_MAX_COMMIT_AGE
;
889 mopt
->commit_interval
= HZ
* option
;
893 if (match_int(&args
[0], &option
)) {
897 if (option
>= 0 && (option
<= ocfs2_local_alloc_size(sb
) * 8))
898 mopt
->localalloc_opt
= option
;
900 case Opt_localflocks
:
902 * Changing this during remount could race
903 * flock() requests, or "unbalance" existing
904 * ones (e.g., a lock is taken in one mode but
905 * dropped in the other). If users care enough
906 * to flip locking modes during remount, we
907 * could add a "local" flag to individual
908 * flock structures for proper tracking of
912 mopt
->mount_opt
|= OCFS2_MOUNT_LOCALFLOCKS
;
915 /* Check both that the option we were passed
916 * is of the right length and that it is a proper
917 * string of the right length.
919 if (((args
[0].to
- args
[0].from
) !=
920 OCFS2_STACK_LABEL_LEN
) ||
921 (strnlen(args
[0].from
,
922 OCFS2_STACK_LABEL_LEN
) !=
923 OCFS2_STACK_LABEL_LEN
)) {
925 "Invalid cluster_stack option\n");
929 memcpy(mopt
->cluster_stack
, args
[0].from
,
930 OCFS2_STACK_LABEL_LEN
);
931 mopt
->cluster_stack
[OCFS2_STACK_LABEL_LEN
] = '\0';
935 "Unrecognized mount option \"%s\" "
936 "or missing value\n", p
);
949 static int ocfs2_show_options(struct seq_file
*s
, struct vfsmount
*mnt
)
951 struct ocfs2_super
*osb
= OCFS2_SB(mnt
->mnt_sb
);
952 unsigned long opts
= osb
->s_mount_opt
;
953 unsigned int local_alloc_megs
;
955 if (opts
& OCFS2_MOUNT_HB_LOCAL
)
956 seq_printf(s
, ",_netdev,heartbeat=local");
958 seq_printf(s
, ",heartbeat=none");
960 if (opts
& OCFS2_MOUNT_NOINTR
)
961 seq_printf(s
, ",nointr");
963 if (opts
& OCFS2_MOUNT_DATA_WRITEBACK
)
964 seq_printf(s
, ",data=writeback");
966 seq_printf(s
, ",data=ordered");
968 if (opts
& OCFS2_MOUNT_BARRIER
)
969 seq_printf(s
, ",barrier=1");
971 if (opts
& OCFS2_MOUNT_ERRORS_PANIC
)
972 seq_printf(s
, ",errors=panic");
974 seq_printf(s
, ",errors=remount-ro");
976 if (osb
->preferred_slot
!= OCFS2_INVALID_SLOT
)
977 seq_printf(s
, ",preferred_slot=%d", osb
->preferred_slot
);
979 if (osb
->s_atime_quantum
!= OCFS2_DEFAULT_ATIME_QUANTUM
)
980 seq_printf(s
, ",atime_quantum=%u", osb
->s_atime_quantum
);
982 if (osb
->osb_commit_interval
)
983 seq_printf(s
, ",commit=%u",
984 (unsigned) (osb
->osb_commit_interval
/ HZ
));
986 local_alloc_megs
= osb
->local_alloc_bits
>> (20 - osb
->s_clustersize_bits
);
987 if (local_alloc_megs
!= OCFS2_DEFAULT_LOCAL_ALLOC_SIZE
)
988 seq_printf(s
, ",localalloc=%d", local_alloc_megs
);
990 if (opts
& OCFS2_MOUNT_LOCALFLOCKS
)
991 seq_printf(s
, ",localflocks,");
993 if (osb
->osb_cluster_stack
[0])
994 seq_printf(s
, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN
,
995 osb
->osb_cluster_stack
);
1000 static int __init
ocfs2_init(void)
1006 ocfs2_print_version();
1008 status
= init_ocfs2_uptodate_cache();
1014 status
= ocfs2_initialize_mem_caches();
1020 ocfs2_wq
= create_singlethread_workqueue("ocfs2_wq");
1026 ocfs2_debugfs_root
= debugfs_create_dir("ocfs2", NULL
);
1027 if (!ocfs2_debugfs_root
) {
1029 mlog(ML_ERROR
, "Unable to create ocfs2 debugfs root.\n");
1032 ocfs2_set_locking_protocol();
1036 ocfs2_free_mem_caches();
1037 exit_ocfs2_uptodate_cache();
1043 return register_filesystem(&ocfs2_fs_type
);
1048 static void __exit
ocfs2_exit(void)
1053 flush_workqueue(ocfs2_wq
);
1054 destroy_workqueue(ocfs2_wq
);
1057 debugfs_remove(ocfs2_debugfs_root
);
1059 ocfs2_free_mem_caches();
1061 unregister_filesystem(&ocfs2_fs_type
);
1063 exit_ocfs2_uptodate_cache();
1068 static void ocfs2_put_super(struct super_block
*sb
)
1070 mlog_entry("(0x%p)\n", sb
);
1072 ocfs2_sync_blockdev(sb
);
1073 ocfs2_dismount_volume(sb
, 0);
1078 static int ocfs2_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
1080 struct ocfs2_super
*osb
;
1081 u32 numbits
, freebits
;
1083 struct ocfs2_dinode
*bm_lock
;
1084 struct buffer_head
*bh
= NULL
;
1085 struct inode
*inode
= NULL
;
1087 mlog_entry("(%p, %p)\n", dentry
->d_sb
, buf
);
1089 osb
= OCFS2_SB(dentry
->d_sb
);
1091 inode
= ocfs2_get_system_file_inode(osb
,
1092 GLOBAL_BITMAP_SYSTEM_INODE
,
1093 OCFS2_INVALID_SLOT
);
1095 mlog(ML_ERROR
, "failed to get bitmap inode\n");
1100 status
= ocfs2_inode_lock(inode
, &bh
, 0);
1106 bm_lock
= (struct ocfs2_dinode
*) bh
->b_data
;
1108 numbits
= le32_to_cpu(bm_lock
->id1
.bitmap1
.i_total
);
1109 freebits
= numbits
- le32_to_cpu(bm_lock
->id1
.bitmap1
.i_used
);
1111 buf
->f_type
= OCFS2_SUPER_MAGIC
;
1112 buf
->f_bsize
= dentry
->d_sb
->s_blocksize
;
1113 buf
->f_namelen
= OCFS2_MAX_FILENAME_LEN
;
1114 buf
->f_blocks
= ((sector_t
) numbits
) *
1115 (osb
->s_clustersize
>> osb
->sb
->s_blocksize_bits
);
1116 buf
->f_bfree
= ((sector_t
) freebits
) *
1117 (osb
->s_clustersize
>> osb
->sb
->s_blocksize_bits
);
1118 buf
->f_bavail
= buf
->f_bfree
;
1119 buf
->f_files
= numbits
;
1120 buf
->f_ffree
= freebits
;
1124 ocfs2_inode_unlock(inode
, 0);
1135 static void ocfs2_inode_init_once(void *data
)
1137 struct ocfs2_inode_info
*oi
= data
;
1140 oi
->ip_open_count
= 0;
1141 spin_lock_init(&oi
->ip_lock
);
1142 ocfs2_extent_map_init(&oi
->vfs_inode
);
1143 INIT_LIST_HEAD(&oi
->ip_io_markers
);
1144 oi
->ip_created_trans
= 0;
1145 oi
->ip_last_trans
= 0;
1146 oi
->ip_dir_start_lookup
= 0;
1148 init_rwsem(&oi
->ip_alloc_sem
);
1149 init_rwsem(&oi
->ip_xattr_sem
);
1150 mutex_init(&oi
->ip_io_mutex
);
1152 oi
->ip_blkno
= 0ULL;
1153 oi
->ip_clusters
= 0;
1155 ocfs2_lock_res_init_once(&oi
->ip_rw_lockres
);
1156 ocfs2_lock_res_init_once(&oi
->ip_inode_lockres
);
1157 ocfs2_lock_res_init_once(&oi
->ip_open_lockres
);
1159 ocfs2_metadata_cache_init(&oi
->vfs_inode
);
1161 inode_init_once(&oi
->vfs_inode
);
1164 static int ocfs2_initialize_mem_caches(void)
1166 ocfs2_inode_cachep
= kmem_cache_create("ocfs2_inode_cache",
1167 sizeof(struct ocfs2_inode_info
),
1169 (SLAB_HWCACHE_ALIGN
|SLAB_RECLAIM_ACCOUNT
|
1171 ocfs2_inode_init_once
);
1172 if (!ocfs2_inode_cachep
)
1178 static void ocfs2_free_mem_caches(void)
1180 if (ocfs2_inode_cachep
)
1181 kmem_cache_destroy(ocfs2_inode_cachep
);
1183 ocfs2_inode_cachep
= NULL
;
1186 static int ocfs2_get_sector(struct super_block
*sb
,
1187 struct buffer_head
**bh
,
1191 if (!sb_set_blocksize(sb
, sect_size
)) {
1192 mlog(ML_ERROR
, "unable to set blocksize\n");
1196 *bh
= sb_getblk(sb
, block
);
1202 if (!buffer_dirty(*bh
))
1203 clear_buffer_uptodate(*bh
);
1205 ll_rw_block(READ
, 1, bh
);
1206 wait_on_buffer(*bh
);
1210 static int ocfs2_mount_volume(struct super_block
*sb
)
1213 int unlock_super
= 0;
1214 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
1218 if (ocfs2_is_hard_readonly(osb
))
1221 status
= ocfs2_dlm_init(osb
);
1227 status
= ocfs2_super_lock(osb
, 1);
1234 /* This will load up the node map and add ourselves to it. */
1235 status
= ocfs2_find_slot(osb
);
1241 /* load all node-local system inodes */
1242 status
= ocfs2_init_local_system_inodes(osb
);
1248 status
= ocfs2_check_volume(osb
);
1254 status
= ocfs2_truncate_log_init(osb
);
1260 if (ocfs2_mount_local(osb
))
1265 ocfs2_super_unlock(osb
, 1);
1271 static void ocfs2_dismount_volume(struct super_block
*sb
, int mnt_err
)
1273 int tmp
, hangup_needed
= 0;
1274 struct ocfs2_super
*osb
= NULL
;
1277 mlog_entry("(0x%p)\n", sb
);
1283 ocfs2_shutdown_local_alloc(osb
);
1285 ocfs2_truncate_log_shutdown(osb
);
1287 /* This will disable recovery and flush any recovery work. */
1288 ocfs2_recovery_exit(osb
);
1290 ocfs2_journal_shutdown(osb
);
1292 ocfs2_sync_blockdev(sb
);
1294 /* No cluster connection means we've failed during mount, so skip
1295 * all the steps which depended on that to complete. */
1297 tmp
= ocfs2_super_lock(osb
, 1);
1304 if (osb
->slot_num
!= OCFS2_INVALID_SLOT
)
1305 ocfs2_put_slot(osb
);
1308 ocfs2_super_unlock(osb
, 1);
1310 ocfs2_release_system_inodes(osb
);
1313 * If we're dismounting due to mount error, mount.ocfs2 will clean
1314 * up heartbeat. If we're a local mount, there is no heartbeat.
1315 * If we failed before we got a uuid_str yet, we can't stop
1316 * heartbeat. Otherwise, do it.
1318 if (!mnt_err
&& !ocfs2_mount_local(osb
) && osb
->uuid_str
)
1322 ocfs2_dlm_shutdown(osb
, hangup_needed
);
1324 debugfs_remove(osb
->osb_debug_root
);
1327 ocfs2_cluster_hangup(osb
->uuid_str
, strlen(osb
->uuid_str
));
1329 atomic_set(&osb
->vol_state
, VOLUME_DISMOUNTED
);
1331 if (ocfs2_mount_local(osb
))
1332 snprintf(nodestr
, sizeof(nodestr
), "local");
1334 snprintf(nodestr
, sizeof(nodestr
), "%u", osb
->node_num
);
1336 printk(KERN_INFO
"ocfs2: Unmounting device (%s) on (node %s)\n",
1337 osb
->dev_str
, nodestr
);
1339 ocfs2_delete_osb(osb
);
1342 sb
->s_fs_info
= NULL
;
1345 static int ocfs2_setup_osb_uuid(struct ocfs2_super
*osb
, const unsigned char *uuid
,
1346 unsigned uuid_bytes
)
1351 BUG_ON(uuid_bytes
!= OCFS2_VOL_UUID_LEN
);
1353 osb
->uuid_str
= kzalloc(OCFS2_VOL_UUID_LEN
* 2 + 1, GFP_KERNEL
);
1354 if (osb
->uuid_str
== NULL
)
1357 for (i
= 0, ptr
= osb
->uuid_str
; i
< OCFS2_VOL_UUID_LEN
; i
++) {
1358 /* print with null */
1359 ret
= snprintf(ptr
, 3, "%02X", uuid
[i
]);
1360 if (ret
!= 2) /* drop super cleans up */
1362 /* then only advance past the last char */
1369 static int ocfs2_initialize_super(struct super_block
*sb
,
1370 struct buffer_head
*bh
,
1374 int i
, cbits
, bbits
;
1375 struct ocfs2_dinode
*di
= (struct ocfs2_dinode
*)bh
->b_data
;
1376 struct inode
*inode
= NULL
;
1377 struct ocfs2_journal
*journal
;
1378 __le32 uuid_net_key
;
1379 struct ocfs2_super
*osb
;
1383 osb
= kzalloc(sizeof(struct ocfs2_super
), GFP_KERNEL
);
1390 sb
->s_fs_info
= osb
;
1391 sb
->s_op
= &ocfs2_sops
;
1392 sb
->s_export_op
= &ocfs2_export_ops
;
1393 sb
->s_xattr
= ocfs2_xattr_handlers
;
1394 sb
->s_time_gran
= 1;
1395 sb
->s_flags
|= MS_NOATIME
;
1396 /* this is needed to support O_LARGEFILE */
1397 cbits
= le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
);
1398 bbits
= le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
);
1399 sb
->s_maxbytes
= ocfs2_max_file_offset(bbits
, cbits
);
1402 /* Save off for ocfs2_rw_direct */
1403 osb
->s_sectsize_bits
= blksize_bits(sector_size
);
1404 BUG_ON(!osb
->s_sectsize_bits
);
1406 spin_lock_init(&osb
->dc_task_lock
);
1407 init_waitqueue_head(&osb
->dc_event
);
1408 osb
->dc_work_sequence
= 0;
1409 osb
->dc_wake_sequence
= 0;
1410 INIT_LIST_HEAD(&osb
->blocked_lock_list
);
1411 osb
->blocked_lock_count
= 0;
1412 spin_lock_init(&osb
->osb_lock
);
1413 ocfs2_init_inode_steal_slot(osb
);
1415 atomic_set(&osb
->alloc_stats
.moves
, 0);
1416 atomic_set(&osb
->alloc_stats
.local_data
, 0);
1417 atomic_set(&osb
->alloc_stats
.bitmap_data
, 0);
1418 atomic_set(&osb
->alloc_stats
.bg_allocs
, 0);
1419 atomic_set(&osb
->alloc_stats
.bg_extends
, 0);
1421 ocfs2_init_node_maps(osb
);
1423 snprintf(osb
->dev_str
, sizeof(osb
->dev_str
), "%u,%u",
1424 MAJOR(osb
->sb
->s_dev
), MINOR(osb
->sb
->s_dev
));
1426 status
= ocfs2_recovery_init(osb
);
1428 mlog(ML_ERROR
, "Unable to initialize recovery state\n");
1433 init_waitqueue_head(&osb
->checkpoint_event
);
1434 atomic_set(&osb
->needs_checkpoint
, 0);
1436 osb
->s_atime_quantum
= OCFS2_DEFAULT_ATIME_QUANTUM
;
1438 osb
->slot_num
= OCFS2_INVALID_SLOT
;
1440 osb
->s_xattr_inline_size
= OCFS2_MIN_XATTR_INLINE_SIZE
;
1442 osb
->local_alloc_state
= OCFS2_LA_UNUSED
;
1443 osb
->local_alloc_bh
= NULL
;
1444 INIT_DELAYED_WORK(&osb
->la_enable_wq
, ocfs2_la_enable_worker
);
1446 init_waitqueue_head(&osb
->osb_mount_event
);
1448 osb
->vol_label
= kmalloc(OCFS2_MAX_VOL_LABEL_LEN
, GFP_KERNEL
);
1449 if (!osb
->vol_label
) {
1450 mlog(ML_ERROR
, "unable to alloc vol label\n");
1455 osb
->max_slots
= le16_to_cpu(di
->id2
.i_super
.s_max_slots
);
1456 if (osb
->max_slots
> OCFS2_MAX_SLOTS
|| osb
->max_slots
== 0) {
1457 mlog(ML_ERROR
, "Invalid number of node slots (%u)\n",
1462 mlog(0, "max_slots for this device: %u\n", osb
->max_slots
);
1464 osb
->slot_recovery_generations
=
1465 kcalloc(osb
->max_slots
, sizeof(*osb
->slot_recovery_generations
),
1467 if (!osb
->slot_recovery_generations
) {
1473 init_waitqueue_head(&osb
->osb_wipe_event
);
1474 osb
->osb_orphan_wipes
= kcalloc(osb
->max_slots
,
1475 sizeof(*osb
->osb_orphan_wipes
),
1477 if (!osb
->osb_orphan_wipes
) {
1483 osb
->s_feature_compat
=
1484 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_compat
);
1485 osb
->s_feature_ro_compat
=
1486 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_ro_compat
);
1487 osb
->s_feature_incompat
=
1488 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_incompat
);
1490 if ((i
= OCFS2_HAS_INCOMPAT_FEATURE(osb
->sb
, ~OCFS2_FEATURE_INCOMPAT_SUPP
))) {
1491 mlog(ML_ERROR
, "couldn't mount because of unsupported "
1492 "optional features (%x).\n", i
);
1496 if (!(osb
->sb
->s_flags
& MS_RDONLY
) &&
1497 (i
= OCFS2_HAS_RO_COMPAT_FEATURE(osb
->sb
, ~OCFS2_FEATURE_RO_COMPAT_SUPP
))) {
1498 mlog(ML_ERROR
, "couldn't mount RDWR because of "
1499 "unsupported optional features (%x).\n", i
);
1504 if (ocfs2_userspace_stack(osb
)) {
1505 memcpy(osb
->osb_cluster_stack
,
1506 OCFS2_RAW_SB(di
)->s_cluster_info
.ci_stack
,
1507 OCFS2_STACK_LABEL_LEN
);
1508 osb
->osb_cluster_stack
[OCFS2_STACK_LABEL_LEN
] = '\0';
1509 if (strlen(osb
->osb_cluster_stack
) != OCFS2_STACK_LABEL_LEN
) {
1511 "couldn't mount because of an invalid "
1512 "cluster stack label (%s) \n",
1513 osb
->osb_cluster_stack
);
1518 /* The empty string is identical with classic tools that
1519 * don't know about s_cluster_info. */
1520 osb
->osb_cluster_stack
[0] = '\0';
1523 get_random_bytes(&osb
->s_next_generation
, sizeof(u32
));
1526 * This should be done in ocfs2_journal_init(), but unknown
1527 * ordering issues will cause the filesystem to crash.
1528 * If anyone wants to figure out what part of the code
1529 * refers to osb->journal before ocfs2_journal_init() is run,
1532 /* initialize our journal structure */
1534 journal
= kzalloc(sizeof(struct ocfs2_journal
), GFP_KERNEL
);
1536 mlog(ML_ERROR
, "unable to alloc journal\n");
1540 osb
->journal
= journal
;
1541 journal
->j_osb
= osb
;
1543 atomic_set(&journal
->j_num_trans
, 0);
1544 init_rwsem(&journal
->j_trans_barrier
);
1545 init_waitqueue_head(&journal
->j_checkpointed
);
1546 spin_lock_init(&journal
->j_lock
);
1547 journal
->j_trans_id
= (unsigned long) 1;
1548 INIT_LIST_HEAD(&journal
->j_la_cleanups
);
1549 INIT_WORK(&journal
->j_recovery_work
, ocfs2_complete_recovery
);
1550 journal
->j_state
= OCFS2_JOURNAL_FREE
;
1552 /* get some pseudo constants for clustersize bits */
1553 osb
->s_clustersize_bits
=
1554 le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
);
1555 osb
->s_clustersize
= 1 << osb
->s_clustersize_bits
;
1556 mlog(0, "clusterbits=%d\n", osb
->s_clustersize_bits
);
1558 if (osb
->s_clustersize
< OCFS2_MIN_CLUSTERSIZE
||
1559 osb
->s_clustersize
> OCFS2_MAX_CLUSTERSIZE
) {
1560 mlog(ML_ERROR
, "Volume has invalid cluster size (%d)\n",
1561 osb
->s_clustersize
);
1566 if (ocfs2_clusters_to_blocks(osb
->sb
, le32_to_cpu(di
->i_clusters
) - 1)
1568 mlog(ML_ERROR
, "Volume might try to write to blocks beyond "
1569 "what jbd can address in 32 bits.\n");
1574 if (ocfs2_setup_osb_uuid(osb
, di
->id2
.i_super
.s_uuid
,
1575 sizeof(di
->id2
.i_super
.s_uuid
))) {
1576 mlog(ML_ERROR
, "Out of memory trying to setup our uuid.\n");
1581 memcpy(&uuid_net_key
, di
->id2
.i_super
.s_uuid
, sizeof(uuid_net_key
));
1583 strncpy(osb
->vol_label
, di
->id2
.i_super
.s_label
, 63);
1584 osb
->vol_label
[63] = '\0';
1585 osb
->root_blkno
= le64_to_cpu(di
->id2
.i_super
.s_root_blkno
);
1586 osb
->system_dir_blkno
= le64_to_cpu(di
->id2
.i_super
.s_system_dir_blkno
);
1587 osb
->first_cluster_group_blkno
=
1588 le64_to_cpu(di
->id2
.i_super
.s_first_cluster_group
);
1589 osb
->fs_generation
= le32_to_cpu(di
->i_fs_generation
);
1590 osb
->uuid_hash
= le32_to_cpu(di
->id2
.i_super
.s_uuid_hash
);
1591 mlog(0, "vol_label: %s\n", osb
->vol_label
);
1592 mlog(0, "uuid: %s\n", osb
->uuid_str
);
1593 mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
1594 (unsigned long long)osb
->root_blkno
,
1595 (unsigned long long)osb
->system_dir_blkno
);
1597 osb
->osb_dlm_debug
= ocfs2_new_dlm_debug();
1598 if (!osb
->osb_dlm_debug
) {
1604 atomic_set(&osb
->vol_state
, VOLUME_INIT
);
1606 /* load root, system_dir, and all global system inodes */
1607 status
= ocfs2_init_global_system_inodes(osb
);
1616 inode
= ocfs2_get_system_file_inode(osb
, GLOBAL_BITMAP_SYSTEM_INODE
,
1617 OCFS2_INVALID_SLOT
);
1624 osb
->bitmap_blkno
= OCFS2_I(inode
)->ip_blkno
;
1627 osb
->bitmap_cpg
= ocfs2_group_bitmap_size(sb
) * 8;
1629 status
= ocfs2_init_slot_info(osb
);
1641 * will return: -EAGAIN if it is ok to keep searching for superblocks
1642 * -EINVAL if there is a bad superblock
1645 static int ocfs2_verify_volume(struct ocfs2_dinode
*di
,
1646 struct buffer_head
*bh
,
1649 int status
= -EAGAIN
;
1653 if (memcmp(di
->i_signature
, OCFS2_SUPER_BLOCK_SIGNATURE
,
1654 strlen(OCFS2_SUPER_BLOCK_SIGNATURE
)) == 0) {
1656 if ((1 << le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
)) != blksz
) {
1657 mlog(ML_ERROR
, "found superblock with incorrect block "
1658 "size: found %u, should be %u\n",
1659 1 << le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
),
1661 } else if (le16_to_cpu(di
->id2
.i_super
.s_major_rev_level
) !=
1662 OCFS2_MAJOR_REV_LEVEL
||
1663 le16_to_cpu(di
->id2
.i_super
.s_minor_rev_level
) !=
1664 OCFS2_MINOR_REV_LEVEL
) {
1665 mlog(ML_ERROR
, "found superblock with bad version: "
1666 "found %u.%u, should be %u.%u\n",
1667 le16_to_cpu(di
->id2
.i_super
.s_major_rev_level
),
1668 le16_to_cpu(di
->id2
.i_super
.s_minor_rev_level
),
1669 OCFS2_MAJOR_REV_LEVEL
,
1670 OCFS2_MINOR_REV_LEVEL
);
1671 } else if (bh
->b_blocknr
!= le64_to_cpu(di
->i_blkno
)) {
1672 mlog(ML_ERROR
, "bad block number on superblock: "
1673 "found %llu, should be %llu\n",
1674 (unsigned long long)le64_to_cpu(di
->i_blkno
),
1675 (unsigned long long)bh
->b_blocknr
);
1676 } else if (le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
) < 12 ||
1677 le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
) > 20) {
1678 mlog(ML_ERROR
, "bad cluster size found: %u\n",
1679 1 << le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
));
1680 } else if (!le64_to_cpu(di
->id2
.i_super
.s_root_blkno
)) {
1681 mlog(ML_ERROR
, "bad root_blkno: 0\n");
1682 } else if (!le64_to_cpu(di
->id2
.i_super
.s_system_dir_blkno
)) {
1683 mlog(ML_ERROR
, "bad system_dir_blkno: 0\n");
1684 } else if (le16_to_cpu(di
->id2
.i_super
.s_max_slots
) > OCFS2_MAX_SLOTS
) {
1686 "Superblock slots found greater than file system "
1687 "maximum: found %u, max %u\n",
1688 le16_to_cpu(di
->id2
.i_super
.s_max_slots
),
1700 static int ocfs2_check_volume(struct ocfs2_super
*osb
)
1705 struct ocfs2_dinode
*local_alloc
= NULL
; /* only used if we
1711 /* Init our journal object. */
1712 status
= ocfs2_journal_init(osb
->journal
, &dirty
);
1714 mlog(ML_ERROR
, "Could not initialize journal!\n");
1718 /* If the journal was unmounted cleanly then we don't want to
1719 * recover anything. Otherwise, journal_load will do that
1720 * dirty work for us :) */
1722 status
= ocfs2_journal_wipe(osb
->journal
, 0);
1728 mlog(ML_NOTICE
, "File system was not unmounted cleanly, "
1729 "recovering volume.\n");
1732 local
= ocfs2_mount_local(osb
);
1734 /* will play back anything left in the journal. */
1735 status
= ocfs2_journal_load(osb
->journal
, local
, dirty
);
1737 mlog(ML_ERROR
, "ocfs2 journal load failed! %d\n", status
);
1742 /* recover my local alloc if we didn't unmount cleanly. */
1743 status
= ocfs2_begin_local_alloc_recovery(osb
,
1750 /* we complete the recovery process after we've marked
1751 * ourselves as mounted. */
1754 mlog(0, "Journal loaded.\n");
1756 status
= ocfs2_load_local_alloc(osb
);
1763 /* Recovery will be completed after we've mounted the
1764 * rest of the volume. */
1766 osb
->local_alloc_copy
= local_alloc
;
1770 /* go through each journal, trylock it and if you get the
1771 * lock, and it's marked as dirty, set the bit in the recover
1772 * map and launch a recovery thread for it. */
1773 status
= ocfs2_mark_dead_nodes(osb
);
1786 * The routine gets called from dismount or close whenever a dismount on
1787 * volume is requested and the osb open count becomes 1.
1788 * It will remove the osb from the global list and also free up all the
1789 * initialized resources and fileobject.
1791 static void ocfs2_delete_osb(struct ocfs2_super
*osb
)
1795 /* This function assumes that the caller has the main osb resource */
1797 ocfs2_free_slot_info(osb
);
1799 kfree(osb
->osb_orphan_wipes
);
1800 kfree(osb
->slot_recovery_generations
);
1802 * This belongs in journal shutdown, but because we have to
1803 * allocate osb->journal at the start of ocfs2_initalize_osb(),
1806 kfree(osb
->journal
);
1807 if (osb
->local_alloc_copy
)
1808 kfree(osb
->local_alloc_copy
);
1809 kfree(osb
->uuid_str
);
1810 ocfs2_put_dlm_debug(osb
->osb_dlm_debug
);
1811 memset(osb
, 0, sizeof(struct ocfs2_super
));
1816 /* Put OCFS2 into a readonly state, or (if the user specifies it),
1817 * panic(). We do not support continue-on-error operation. */
1818 static void ocfs2_handle_error(struct super_block
*sb
)
1820 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
1822 if (osb
->s_mount_opt
& OCFS2_MOUNT_ERRORS_PANIC
)
1823 panic("OCFS2: (device %s): panic forced after error\n",
1826 ocfs2_set_osb_flag(osb
, OCFS2_OSB_ERROR_FS
);
1828 if (sb
->s_flags
& MS_RDONLY
&&
1829 (ocfs2_is_soft_readonly(osb
) ||
1830 ocfs2_is_hard_readonly(osb
)))
1833 printk(KERN_CRIT
"File system is now read-only due to the potential "
1834 "of on-disk corruption. Please run fsck.ocfs2 once the file "
1835 "system is unmounted.\n");
1836 sb
->s_flags
|= MS_RDONLY
;
1837 ocfs2_set_ro_flag(osb
, 0);
1840 static char error_buf
[1024];
1842 void __ocfs2_error(struct super_block
*sb
,
1843 const char *function
,
1844 const char *fmt
, ...)
1848 va_start(args
, fmt
);
1849 vsnprintf(error_buf
, sizeof(error_buf
), fmt
, args
);
1852 /* Not using mlog here because we want to show the actual
1853 * function the error came from. */
1854 printk(KERN_CRIT
"OCFS2: ERROR (device %s): %s: %s\n",
1855 sb
->s_id
, function
, error_buf
);
1857 ocfs2_handle_error(sb
);
1860 /* Handle critical errors. This is intentionally more drastic than
1861 * ocfs2_handle_error, so we only use for things like journal errors,
1863 void __ocfs2_abort(struct super_block
* sb
,
1864 const char *function
,
1865 const char *fmt
, ...)
1869 va_start(args
, fmt
);
1870 vsnprintf(error_buf
, sizeof(error_buf
), fmt
, args
);
1873 printk(KERN_CRIT
"OCFS2: abort (device %s): %s: %s\n",
1874 sb
->s_id
, function
, error_buf
);
1876 /* We don't have the cluster support yet to go straight to
1877 * hard readonly in here. Until then, we want to keep
1878 * ocfs2_abort() so that we can at least mark critical
1881 * TODO: This should abort the journal and alert other nodes
1882 * that our slot needs recovery. */
1884 /* Force a panic(). This stinks, but it's better than letting
1885 * things continue without having a proper hard readonly
1887 OCFS2_SB(sb
)->s_mount_opt
|= OCFS2_MOUNT_ERRORS_PANIC
;
1888 ocfs2_handle_error(sb
);
1891 module_init(ocfs2_init
);
1892 module_exit(ocfs2_exit
);