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
,
166 static const match_table_t tokens
= {
167 {Opt_barrier
, "barrier=%u"},
168 {Opt_err_panic
, "errors=panic"},
169 {Opt_err_ro
, "errors=remount-ro"},
171 {Opt_nointr
, "nointr"},
172 {Opt_hb_none
, OCFS2_HB_NONE
},
173 {Opt_hb_local
, OCFS2_HB_LOCAL
},
174 {Opt_data_ordered
, "data=ordered"},
175 {Opt_data_writeback
, "data=writeback"},
176 {Opt_atime_quantum
, "atime_quantum=%u"},
177 {Opt_slot
, "preferred_slot=%u"},
178 {Opt_commit
, "commit=%u"},
179 {Opt_localalloc
, "localalloc=%d"},
180 {Opt_localflocks
, "localflocks"},
181 {Opt_stack
, "cluster_stack=%s"},
182 {Opt_user_xattr
, "user_xattr"},
183 {Opt_nouser_xattr
, "nouser_xattr"},
184 {Opt_inode64
, "inode64"},
186 {Opt_noacl
, "noacl"},
191 * write_super and sync_fs ripped right out of ext3.
193 static void ocfs2_write_super(struct super_block
*sb
)
195 if (mutex_trylock(&sb
->s_lock
) != 0)
200 static int ocfs2_sync_fs(struct super_block
*sb
, int wait
)
204 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
208 if (ocfs2_is_hard_readonly(osb
))
212 status
= ocfs2_flush_truncate_log(osb
);
216 ocfs2_schedule_truncate_log_flush(osb
, 0);
219 if (jbd2_journal_start_commit(OCFS2_SB(sb
)->journal
->j_journal
,
222 jbd2_log_wait_commit(OCFS2_SB(sb
)->journal
->j_journal
,
228 static int ocfs2_init_global_system_inodes(struct ocfs2_super
*osb
)
230 struct inode
*new = NULL
;
236 new = ocfs2_iget(osb
, osb
->root_blkno
, OCFS2_FI_FLAG_SYSFILE
, 0);
238 status
= PTR_ERR(new);
242 osb
->root_inode
= new;
244 new = ocfs2_iget(osb
, osb
->system_dir_blkno
, OCFS2_FI_FLAG_SYSFILE
, 0);
246 status
= PTR_ERR(new);
250 osb
->sys_root_inode
= new;
252 for (i
= OCFS2_FIRST_ONLINE_SYSTEM_INODE
;
253 i
<= OCFS2_LAST_GLOBAL_SYSTEM_INODE
; i
++) {
254 new = ocfs2_get_system_file_inode(osb
, i
, osb
->slot_num
);
256 ocfs2_release_system_inodes(osb
);
259 /* FIXME: Should ERROR_RO_FS */
260 mlog(ML_ERROR
, "Unable to load system inode %d, "
261 "possibly corrupt fs?", i
);
264 // the array now has one ref, so drop this one
273 static int ocfs2_init_local_system_inodes(struct ocfs2_super
*osb
)
275 struct inode
*new = NULL
;
281 for (i
= OCFS2_LAST_GLOBAL_SYSTEM_INODE
+ 1;
282 i
< NUM_SYSTEM_INODES
;
284 new = ocfs2_get_system_file_inode(osb
, i
, osb
->slot_num
);
286 ocfs2_release_system_inodes(osb
);
288 mlog(ML_ERROR
, "status=%d, sysfile=%d, slot=%d\n",
289 status
, i
, osb
->slot_num
);
292 /* the array now has one ref, so drop this one */
301 static void ocfs2_release_system_inodes(struct ocfs2_super
*osb
)
308 for (i
= 0; i
< NUM_SYSTEM_INODES
; i
++) {
309 inode
= osb
->system_inodes
[i
];
312 osb
->system_inodes
[i
] = NULL
;
316 inode
= osb
->sys_root_inode
;
319 osb
->sys_root_inode
= NULL
;
322 inode
= osb
->root_inode
;
325 osb
->root_inode
= NULL
;
331 /* We're allocating fs objects, use GFP_NOFS */
332 static struct inode
*ocfs2_alloc_inode(struct super_block
*sb
)
334 struct ocfs2_inode_info
*oi
;
336 oi
= kmem_cache_alloc(ocfs2_inode_cachep
, GFP_NOFS
);
340 jbd2_journal_init_jbd_inode(&oi
->ip_jinode
, &oi
->vfs_inode
);
341 return &oi
->vfs_inode
;
344 static void ocfs2_destroy_inode(struct inode
*inode
)
346 kmem_cache_free(ocfs2_inode_cachep
, OCFS2_I(inode
));
349 static unsigned long long ocfs2_max_file_offset(unsigned int bbits
,
352 unsigned int bytes
= 1 << cbits
;
353 unsigned int trim
= bytes
;
354 unsigned int bitshift
= 32;
357 * i_size and all block offsets in ocfs2 are always 64 bits
358 * wide. i_clusters is 32 bits, in cluster-sized units. So on
359 * 64 bit platforms, cluster size will be the limiting factor.
362 #if BITS_PER_LONG == 32
363 # if defined(CONFIG_LBD)
364 BUILD_BUG_ON(sizeof(sector_t
) != 8);
366 * We might be limited by page cache size.
368 if (bytes
> PAGE_CACHE_SIZE
) {
369 bytes
= PAGE_CACHE_SIZE
;
372 * Shift by 31 here so that we don't get larger than
379 * We are limited by the size of sector_t. Use block size, as
380 * that's what we expose to the VFS.
389 * Trim by a whole cluster when we can actually approach the
390 * on-disk limits. Otherwise we can overflow i_clusters when
391 * an extent start is at the max offset.
393 return (((unsigned long long)bytes
) << bitshift
) - trim
;
396 static int ocfs2_remount(struct super_block
*sb
, int *flags
, char *data
)
398 int incompat_features
;
400 struct mount_options parsed_options
;
401 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
403 if (!ocfs2_parse_options(sb
, data
, &parsed_options
, 1)) {
408 if ((osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) !=
409 (parsed_options
.mount_opt
& OCFS2_MOUNT_HB_LOCAL
)) {
411 mlog(ML_ERROR
, "Cannot change heartbeat mode on remount\n");
415 if ((osb
->s_mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
) !=
416 (parsed_options
.mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
)) {
418 mlog(ML_ERROR
, "Cannot change data mode on remount\n");
422 /* Probably don't want this on remount; it might
423 * mess with other nodes */
424 if (!(osb
->s_mount_opt
& OCFS2_MOUNT_INODE64
) &&
425 (parsed_options
.mount_opt
& OCFS2_MOUNT_INODE64
)) {
427 mlog(ML_ERROR
, "Cannot enable inode64 on remount\n");
431 /* We're going to/from readonly mode. */
432 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
)) {
433 /* Lock here so the check of HARD_RO and the potential
434 * setting of SOFT_RO is atomic. */
435 spin_lock(&osb
->osb_lock
);
436 if (osb
->osb_flags
& OCFS2_OSB_HARD_RO
) {
437 mlog(ML_ERROR
, "Remount on readonly device is forbidden.\n");
442 if (*flags
& MS_RDONLY
) {
443 mlog(0, "Going to ro mode.\n");
444 sb
->s_flags
|= MS_RDONLY
;
445 osb
->osb_flags
|= OCFS2_OSB_SOFT_RO
;
447 mlog(0, "Making ro filesystem writeable.\n");
449 if (osb
->osb_flags
& OCFS2_OSB_ERROR_FS
) {
450 mlog(ML_ERROR
, "Cannot remount RDWR "
451 "filesystem due to previous errors.\n");
455 incompat_features
= OCFS2_HAS_RO_COMPAT_FEATURE(sb
, ~OCFS2_FEATURE_RO_COMPAT_SUPP
);
456 if (incompat_features
) {
457 mlog(ML_ERROR
, "Cannot remount RDWR because "
458 "of unsupported optional features "
459 "(%x).\n", incompat_features
);
463 sb
->s_flags
&= ~MS_RDONLY
;
464 osb
->osb_flags
&= ~OCFS2_OSB_SOFT_RO
;
467 spin_unlock(&osb
->osb_lock
);
471 /* Only save off the new mount options in case of a successful
473 if (!(osb
->s_feature_incompat
& OCFS2_FEATURE_INCOMPAT_XATTR
))
474 parsed_options
.mount_opt
&= ~OCFS2_MOUNT_POSIX_ACL
;
475 osb
->s_mount_opt
= parsed_options
.mount_opt
;
476 osb
->s_atime_quantum
= parsed_options
.atime_quantum
;
477 osb
->preferred_slot
= parsed_options
.slot
;
478 if (parsed_options
.commit_interval
)
479 osb
->osb_commit_interval
= parsed_options
.commit_interval
;
481 if (!ocfs2_is_hard_readonly(osb
))
482 ocfs2_set_journal_params(osb
);
488 static int ocfs2_sb_probe(struct super_block
*sb
,
489 struct buffer_head
**bh
,
493 struct ocfs1_vol_disk_hdr
*hdr
;
494 struct ocfs2_dinode
*di
;
500 *sector_size
= bdev_hardsect_size(sb
->s_bdev
);
501 if (*sector_size
> OCFS2_MAX_BLOCKSIZE
) {
502 mlog(ML_ERROR
, "Hardware sector size too large: %d (max=%d)\n",
503 *sector_size
, OCFS2_MAX_BLOCKSIZE
);
508 /* Can this really happen? */
509 if (*sector_size
< OCFS2_MIN_BLOCKSIZE
)
510 *sector_size
= OCFS2_MIN_BLOCKSIZE
;
512 /* check block zero for old format */
513 status
= ocfs2_get_sector(sb
, bh
, 0, *sector_size
);
518 hdr
= (struct ocfs1_vol_disk_hdr
*) (*bh
)->b_data
;
519 if (hdr
->major_version
== OCFS1_MAJOR_VERSION
) {
520 mlog(ML_ERROR
, "incompatible version: %u.%u\n",
521 hdr
->major_version
, hdr
->minor_version
);
524 if (memcmp(hdr
->signature
, OCFS1_VOLUME_SIGNATURE
,
525 strlen(OCFS1_VOLUME_SIGNATURE
)) == 0) {
526 mlog(ML_ERROR
, "incompatible volume signature: %8s\n",
533 mlog(ML_ERROR
, "This is an ocfs v1 filesystem which must be "
534 "upgraded before mounting with ocfs v2\n");
539 * Now check at magic offset for 512, 1024, 2048, 4096
540 * blocksizes. 4096 is the maximum blocksize because it is
541 * the minimum clustersize.
544 for (blksize
= *sector_size
;
545 blksize
<= OCFS2_MAX_BLOCKSIZE
;
547 tmpstat
= ocfs2_get_sector(sb
, bh
,
548 OCFS2_SUPER_BLOCK_BLKNO
,
555 di
= (struct ocfs2_dinode
*) (*bh
)->b_data
;
556 status
= ocfs2_verify_volume(di
, *bh
, blksize
);
561 if (status
!= -EAGAIN
)
569 static int ocfs2_verify_heartbeat(struct ocfs2_super
*osb
)
571 if (ocfs2_mount_local(osb
)) {
572 if (osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) {
573 mlog(ML_ERROR
, "Cannot heartbeat on a locally "
574 "mounted device.\n");
579 if (ocfs2_userspace_stack(osb
)) {
580 if (osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) {
581 mlog(ML_ERROR
, "Userspace stack expected, but "
582 "o2cb heartbeat arguments passed to mount\n");
587 if (!(osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
)) {
588 if (!ocfs2_mount_local(osb
) && !ocfs2_is_hard_readonly(osb
) &&
589 !ocfs2_userspace_stack(osb
)) {
590 mlog(ML_ERROR
, "Heartbeat has to be started to mount "
591 "a read-write clustered device.\n");
600 * If we're using a userspace stack, mount should have passed
601 * a name that matches the disk. If not, mount should not
602 * have passed a stack.
604 static int ocfs2_verify_userspace_stack(struct ocfs2_super
*osb
,
605 struct mount_options
*mopt
)
607 if (!ocfs2_userspace_stack(osb
) && mopt
->cluster_stack
[0]) {
609 "cluster stack passed to mount, but this filesystem "
610 "does not support it\n");
614 if (ocfs2_userspace_stack(osb
) &&
615 strncmp(osb
->osb_cluster_stack
, mopt
->cluster_stack
,
616 OCFS2_STACK_LABEL_LEN
)) {
618 "cluster stack passed to mount (\"%s\") does not "
619 "match the filesystem (\"%s\")\n",
621 osb
->osb_cluster_stack
);
628 static int ocfs2_fill_super(struct super_block
*sb
, void *data
, int silent
)
631 int status
, sector_size
;
632 struct mount_options parsed_options
;
633 struct inode
*inode
= NULL
;
634 struct ocfs2_super
*osb
= NULL
;
635 struct buffer_head
*bh
= NULL
;
638 mlog_entry("%p, %p, %i", sb
, data
, silent
);
640 if (!ocfs2_parse_options(sb
, data
, &parsed_options
, 0)) {
642 goto read_super_error
;
645 /* probe for superblock */
646 status
= ocfs2_sb_probe(sb
, &bh
, §or_size
);
648 mlog(ML_ERROR
, "superblock probe failed!\n");
649 goto read_super_error
;
652 status
= ocfs2_initialize_super(sb
, bh
, sector_size
);
656 goto read_super_error
;
661 if (!(osb
->s_feature_incompat
& OCFS2_FEATURE_INCOMPAT_XATTR
))
662 parsed_options
.mount_opt
&= ~OCFS2_MOUNT_POSIX_ACL
;
664 osb
->s_mount_opt
= parsed_options
.mount_opt
;
665 osb
->s_atime_quantum
= parsed_options
.atime_quantum
;
666 osb
->preferred_slot
= parsed_options
.slot
;
667 osb
->osb_commit_interval
= parsed_options
.commit_interval
;
668 osb
->local_alloc_default_bits
= ocfs2_megabytes_to_clusters(sb
, parsed_options
.localalloc_opt
);
669 osb
->local_alloc_bits
= osb
->local_alloc_default_bits
;
671 status
= ocfs2_verify_userspace_stack(osb
, &parsed_options
);
673 goto read_super_error
;
675 sb
->s_magic
= OCFS2_SUPER_MAGIC
;
677 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
678 ((osb
->s_mount_opt
& OCFS2_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
680 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
682 if (bdev_read_only(sb
->s_bdev
)) {
683 if (!(sb
->s_flags
& MS_RDONLY
)) {
685 mlog(ML_ERROR
, "Readonly device detected but readonly "
686 "mount was not specified.\n");
687 goto read_super_error
;
690 /* You should not be able to start a local heartbeat
691 * on a readonly device. */
692 if (osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) {
694 mlog(ML_ERROR
, "Local heartbeat specified on readonly "
696 goto read_super_error
;
699 status
= ocfs2_check_journals_nolocks(osb
);
701 if (status
== -EROFS
)
702 mlog(ML_ERROR
, "Recovery required on readonly "
703 "file system, but write access is "
707 goto read_super_error
;
710 ocfs2_set_ro_flag(osb
, 1);
712 printk(KERN_NOTICE
"Readonly device detected. No cluster "
713 "services will be utilized for this mount. Recovery "
714 "will be skipped.\n");
717 if (!ocfs2_is_hard_readonly(osb
)) {
718 if (sb
->s_flags
& MS_RDONLY
)
719 ocfs2_set_ro_flag(osb
, 0);
722 status
= ocfs2_verify_heartbeat(osb
);
725 goto read_super_error
;
728 osb
->osb_debug_root
= debugfs_create_dir(osb
->uuid_str
,
730 if (!osb
->osb_debug_root
) {
732 mlog(ML_ERROR
, "Unable to create per-mount debugfs root.\n");
733 goto read_super_error
;
736 status
= ocfs2_mount_volume(sb
);
738 inode
= igrab(osb
->root_inode
);
741 goto read_super_error
;
746 goto read_super_error
;
749 root
= d_alloc_root(inode
);
753 goto read_super_error
;
758 ocfs2_complete_mount_recovery(osb
);
760 if (ocfs2_mount_local(osb
))
761 snprintf(nodestr
, sizeof(nodestr
), "local");
763 snprintf(nodestr
, sizeof(nodestr
), "%u", osb
->node_num
);
765 printk(KERN_INFO
"ocfs2: Mounting device (%s) on (node %s, slot %d) "
766 "with %s data mode.\n",
767 osb
->dev_str
, nodestr
, osb
->slot_num
,
768 osb
->s_mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
? "writeback" :
771 atomic_set(&osb
->vol_state
, VOLUME_MOUNTED
);
772 wake_up(&osb
->osb_mount_event
);
784 atomic_set(&osb
->vol_state
, VOLUME_DISABLED
);
785 wake_up(&osb
->osb_mount_event
);
786 ocfs2_dismount_volume(sb
, 1);
793 static int ocfs2_get_sb(struct file_system_type
*fs_type
,
795 const char *dev_name
,
797 struct vfsmount
*mnt
)
799 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, ocfs2_fill_super
,
803 static struct file_system_type ocfs2_fs_type
= {
804 .owner
= THIS_MODULE
,
806 .get_sb
= ocfs2_get_sb
, /* is this called when we mount
808 .kill_sb
= kill_block_super
, /* set to the generic one
809 * right now, but do we
810 * need to change that? */
811 .fs_flags
= FS_REQUIRES_DEV
|FS_RENAME_DOES_D_MOVE
,
815 static int ocfs2_parse_options(struct super_block
*sb
,
817 struct mount_options
*mopt
,
823 mlog_entry("remount: %d, options: \"%s\"\n", is_remount
,
824 options
? options
: "(none)");
826 mopt
->commit_interval
= 0;
828 mopt
->atime_quantum
= OCFS2_DEFAULT_ATIME_QUANTUM
;
829 mopt
->slot
= OCFS2_INVALID_SLOT
;
830 mopt
->localalloc_opt
= OCFS2_DEFAULT_LOCAL_ALLOC_SIZE
;
831 mopt
->cluster_stack
[0] = '\0';
838 while ((p
= strsep(&options
, ",")) != NULL
) {
840 substring_t args
[MAX_OPT_ARGS
];
845 token
= match_token(p
, tokens
, args
);
848 mopt
->mount_opt
|= OCFS2_MOUNT_HB_LOCAL
;
851 mopt
->mount_opt
&= ~OCFS2_MOUNT_HB_LOCAL
;
854 if (match_int(&args
[0], &option
)) {
859 mopt
->mount_opt
|= OCFS2_MOUNT_BARRIER
;
861 mopt
->mount_opt
&= ~OCFS2_MOUNT_BARRIER
;
864 mopt
->mount_opt
&= ~OCFS2_MOUNT_NOINTR
;
867 mopt
->mount_opt
|= OCFS2_MOUNT_NOINTR
;
870 mopt
->mount_opt
|= OCFS2_MOUNT_ERRORS_PANIC
;
873 mopt
->mount_opt
&= ~OCFS2_MOUNT_ERRORS_PANIC
;
875 case Opt_data_ordered
:
876 mopt
->mount_opt
&= ~OCFS2_MOUNT_DATA_WRITEBACK
;
878 case Opt_data_writeback
:
879 mopt
->mount_opt
|= OCFS2_MOUNT_DATA_WRITEBACK
;
882 mopt
->mount_opt
&= ~OCFS2_MOUNT_NOUSERXATTR
;
884 case Opt_nouser_xattr
:
885 mopt
->mount_opt
|= OCFS2_MOUNT_NOUSERXATTR
;
887 case Opt_atime_quantum
:
888 if (match_int(&args
[0], &option
)) {
893 mopt
->atime_quantum
= option
;
897 if (match_int(&args
[0], &option
)) {
902 mopt
->slot
= (s16
)option
;
906 if (match_int(&args
[0], &option
)) {
913 option
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
914 mopt
->commit_interval
= HZ
* option
;
918 if (match_int(&args
[0], &option
)) {
922 if (option
>= 0 && (option
<= ocfs2_local_alloc_size(sb
) * 8))
923 mopt
->localalloc_opt
= option
;
925 case Opt_localflocks
:
927 * Changing this during remount could race
928 * flock() requests, or "unbalance" existing
929 * ones (e.g., a lock is taken in one mode but
930 * dropped in the other). If users care enough
931 * to flip locking modes during remount, we
932 * could add a "local" flag to individual
933 * flock structures for proper tracking of
937 mopt
->mount_opt
|= OCFS2_MOUNT_LOCALFLOCKS
;
940 /* Check both that the option we were passed
941 * is of the right length and that it is a proper
942 * string of the right length.
944 if (((args
[0].to
- args
[0].from
) !=
945 OCFS2_STACK_LABEL_LEN
) ||
946 (strnlen(args
[0].from
,
947 OCFS2_STACK_LABEL_LEN
) !=
948 OCFS2_STACK_LABEL_LEN
)) {
950 "Invalid cluster_stack option\n");
954 memcpy(mopt
->cluster_stack
, args
[0].from
,
955 OCFS2_STACK_LABEL_LEN
);
956 mopt
->cluster_stack
[OCFS2_STACK_LABEL_LEN
] = '\0';
959 mopt
->mount_opt
|= OCFS2_MOUNT_INODE64
;
961 #ifdef CONFIG_OCFS2_FS_POSIX_ACL
963 mopt
->mount_opt
|= OCFS2_MOUNT_POSIX_ACL
;
966 mopt
->mount_opt
&= ~OCFS2_MOUNT_POSIX_ACL
;
971 printk(KERN_INFO
"ocfs2 (no)acl options not supported\n");
976 "Unrecognized mount option \"%s\" "
977 "or missing value\n", p
);
990 static int ocfs2_show_options(struct seq_file
*s
, struct vfsmount
*mnt
)
992 struct ocfs2_super
*osb
= OCFS2_SB(mnt
->mnt_sb
);
993 unsigned long opts
= osb
->s_mount_opt
;
994 unsigned int local_alloc_megs
;
996 if (opts
& OCFS2_MOUNT_HB_LOCAL
)
997 seq_printf(s
, ",_netdev,heartbeat=local");
999 seq_printf(s
, ",heartbeat=none");
1001 if (opts
& OCFS2_MOUNT_NOINTR
)
1002 seq_printf(s
, ",nointr");
1004 if (opts
& OCFS2_MOUNT_DATA_WRITEBACK
)
1005 seq_printf(s
, ",data=writeback");
1007 seq_printf(s
, ",data=ordered");
1009 if (opts
& OCFS2_MOUNT_BARRIER
)
1010 seq_printf(s
, ",barrier=1");
1012 if (opts
& OCFS2_MOUNT_ERRORS_PANIC
)
1013 seq_printf(s
, ",errors=panic");
1015 seq_printf(s
, ",errors=remount-ro");
1017 if (osb
->preferred_slot
!= OCFS2_INVALID_SLOT
)
1018 seq_printf(s
, ",preferred_slot=%d", osb
->preferred_slot
);
1020 if (osb
->s_atime_quantum
!= OCFS2_DEFAULT_ATIME_QUANTUM
)
1021 seq_printf(s
, ",atime_quantum=%u", osb
->s_atime_quantum
);
1023 if (osb
->osb_commit_interval
)
1024 seq_printf(s
, ",commit=%u",
1025 (unsigned) (osb
->osb_commit_interval
/ HZ
));
1027 local_alloc_megs
= osb
->local_alloc_bits
>> (20 - osb
->s_clustersize_bits
);
1028 if (local_alloc_megs
!= OCFS2_DEFAULT_LOCAL_ALLOC_SIZE
)
1029 seq_printf(s
, ",localalloc=%d", local_alloc_megs
);
1031 if (opts
& OCFS2_MOUNT_LOCALFLOCKS
)
1032 seq_printf(s
, ",localflocks,");
1034 if (osb
->osb_cluster_stack
[0])
1035 seq_printf(s
, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN
,
1036 osb
->osb_cluster_stack
);
1038 if (opts
& OCFS2_MOUNT_NOUSERXATTR
)
1039 seq_printf(s
, ",nouser_xattr");
1041 seq_printf(s
, ",user_xattr");
1043 if (opts
& OCFS2_MOUNT_INODE64
)
1044 seq_printf(s
, ",inode64");
1046 #ifdef CONFIG_OCFS2_FS_POSIX_ACL
1047 if (opts
& OCFS2_MOUNT_POSIX_ACL
)
1048 seq_printf(s
, ",acl");
1050 seq_printf(s
, ",noacl");
1056 static int __init
ocfs2_init(void)
1062 ocfs2_print_version();
1064 status
= init_ocfs2_uptodate_cache();
1070 status
= ocfs2_initialize_mem_caches();
1076 ocfs2_wq
= create_singlethread_workqueue("ocfs2_wq");
1082 ocfs2_debugfs_root
= debugfs_create_dir("ocfs2", NULL
);
1083 if (!ocfs2_debugfs_root
) {
1085 mlog(ML_ERROR
, "Unable to create ocfs2 debugfs root.\n");
1088 ocfs2_set_locking_protocol();
1092 ocfs2_free_mem_caches();
1093 exit_ocfs2_uptodate_cache();
1099 return register_filesystem(&ocfs2_fs_type
);
1104 static void __exit
ocfs2_exit(void)
1109 flush_workqueue(ocfs2_wq
);
1110 destroy_workqueue(ocfs2_wq
);
1113 debugfs_remove(ocfs2_debugfs_root
);
1115 ocfs2_free_mem_caches();
1117 unregister_filesystem(&ocfs2_fs_type
);
1119 exit_ocfs2_uptodate_cache();
1124 static void ocfs2_put_super(struct super_block
*sb
)
1126 mlog_entry("(0x%p)\n", sb
);
1128 ocfs2_sync_blockdev(sb
);
1129 ocfs2_dismount_volume(sb
, 0);
1134 static int ocfs2_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
1136 struct ocfs2_super
*osb
;
1137 u32 numbits
, freebits
;
1139 struct ocfs2_dinode
*bm_lock
;
1140 struct buffer_head
*bh
= NULL
;
1141 struct inode
*inode
= NULL
;
1143 mlog_entry("(%p, %p)\n", dentry
->d_sb
, buf
);
1145 osb
= OCFS2_SB(dentry
->d_sb
);
1147 inode
= ocfs2_get_system_file_inode(osb
,
1148 GLOBAL_BITMAP_SYSTEM_INODE
,
1149 OCFS2_INVALID_SLOT
);
1151 mlog(ML_ERROR
, "failed to get bitmap inode\n");
1156 status
= ocfs2_inode_lock(inode
, &bh
, 0);
1162 bm_lock
= (struct ocfs2_dinode
*) bh
->b_data
;
1164 numbits
= le32_to_cpu(bm_lock
->id1
.bitmap1
.i_total
);
1165 freebits
= numbits
- le32_to_cpu(bm_lock
->id1
.bitmap1
.i_used
);
1167 buf
->f_type
= OCFS2_SUPER_MAGIC
;
1168 buf
->f_bsize
= dentry
->d_sb
->s_blocksize
;
1169 buf
->f_namelen
= OCFS2_MAX_FILENAME_LEN
;
1170 buf
->f_blocks
= ((sector_t
) numbits
) *
1171 (osb
->s_clustersize
>> osb
->sb
->s_blocksize_bits
);
1172 buf
->f_bfree
= ((sector_t
) freebits
) *
1173 (osb
->s_clustersize
>> osb
->sb
->s_blocksize_bits
);
1174 buf
->f_bavail
= buf
->f_bfree
;
1175 buf
->f_files
= numbits
;
1176 buf
->f_ffree
= freebits
;
1180 ocfs2_inode_unlock(inode
, 0);
1191 static void ocfs2_inode_init_once(void *data
)
1193 struct ocfs2_inode_info
*oi
= data
;
1196 oi
->ip_open_count
= 0;
1197 spin_lock_init(&oi
->ip_lock
);
1198 ocfs2_extent_map_init(&oi
->vfs_inode
);
1199 INIT_LIST_HEAD(&oi
->ip_io_markers
);
1200 oi
->ip_created_trans
= 0;
1201 oi
->ip_last_trans
= 0;
1202 oi
->ip_dir_start_lookup
= 0;
1204 init_rwsem(&oi
->ip_alloc_sem
);
1205 init_rwsem(&oi
->ip_xattr_sem
);
1206 mutex_init(&oi
->ip_io_mutex
);
1208 oi
->ip_blkno
= 0ULL;
1209 oi
->ip_clusters
= 0;
1211 ocfs2_lock_res_init_once(&oi
->ip_rw_lockres
);
1212 ocfs2_lock_res_init_once(&oi
->ip_inode_lockres
);
1213 ocfs2_lock_res_init_once(&oi
->ip_open_lockres
);
1215 ocfs2_metadata_cache_init(&oi
->vfs_inode
);
1217 inode_init_once(&oi
->vfs_inode
);
1220 static int ocfs2_initialize_mem_caches(void)
1222 ocfs2_inode_cachep
= kmem_cache_create("ocfs2_inode_cache",
1223 sizeof(struct ocfs2_inode_info
),
1225 (SLAB_HWCACHE_ALIGN
|SLAB_RECLAIM_ACCOUNT
|
1227 ocfs2_inode_init_once
);
1228 if (!ocfs2_inode_cachep
)
1234 static void ocfs2_free_mem_caches(void)
1236 if (ocfs2_inode_cachep
)
1237 kmem_cache_destroy(ocfs2_inode_cachep
);
1239 ocfs2_inode_cachep
= NULL
;
1242 static int ocfs2_get_sector(struct super_block
*sb
,
1243 struct buffer_head
**bh
,
1247 if (!sb_set_blocksize(sb
, sect_size
)) {
1248 mlog(ML_ERROR
, "unable to set blocksize\n");
1252 *bh
= sb_getblk(sb
, block
);
1258 if (!buffer_dirty(*bh
))
1259 clear_buffer_uptodate(*bh
);
1261 ll_rw_block(READ
, 1, bh
);
1262 wait_on_buffer(*bh
);
1266 static int ocfs2_mount_volume(struct super_block
*sb
)
1269 int unlock_super
= 0;
1270 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
1274 if (ocfs2_is_hard_readonly(osb
))
1277 status
= ocfs2_dlm_init(osb
);
1283 status
= ocfs2_super_lock(osb
, 1);
1290 /* This will load up the node map and add ourselves to it. */
1291 status
= ocfs2_find_slot(osb
);
1297 /* load all node-local system inodes */
1298 status
= ocfs2_init_local_system_inodes(osb
);
1304 status
= ocfs2_check_volume(osb
);
1310 status
= ocfs2_truncate_log_init(osb
);
1316 if (ocfs2_mount_local(osb
))
1321 ocfs2_super_unlock(osb
, 1);
1327 static void ocfs2_dismount_volume(struct super_block
*sb
, int mnt_err
)
1329 int tmp
, hangup_needed
= 0;
1330 struct ocfs2_super
*osb
= NULL
;
1333 mlog_entry("(0x%p)\n", sb
);
1339 ocfs2_shutdown_local_alloc(osb
);
1341 ocfs2_truncate_log_shutdown(osb
);
1343 /* This will disable recovery and flush any recovery work. */
1344 ocfs2_recovery_exit(osb
);
1346 ocfs2_journal_shutdown(osb
);
1348 ocfs2_sync_blockdev(sb
);
1350 /* No cluster connection means we've failed during mount, so skip
1351 * all the steps which depended on that to complete. */
1353 tmp
= ocfs2_super_lock(osb
, 1);
1360 if (osb
->slot_num
!= OCFS2_INVALID_SLOT
)
1361 ocfs2_put_slot(osb
);
1364 ocfs2_super_unlock(osb
, 1);
1366 ocfs2_release_system_inodes(osb
);
1369 * If we're dismounting due to mount error, mount.ocfs2 will clean
1370 * up heartbeat. If we're a local mount, there is no heartbeat.
1371 * If we failed before we got a uuid_str yet, we can't stop
1372 * heartbeat. Otherwise, do it.
1374 if (!mnt_err
&& !ocfs2_mount_local(osb
) && osb
->uuid_str
)
1378 ocfs2_dlm_shutdown(osb
, hangup_needed
);
1380 debugfs_remove(osb
->osb_debug_root
);
1383 ocfs2_cluster_hangup(osb
->uuid_str
, strlen(osb
->uuid_str
));
1385 atomic_set(&osb
->vol_state
, VOLUME_DISMOUNTED
);
1387 if (ocfs2_mount_local(osb
))
1388 snprintf(nodestr
, sizeof(nodestr
), "local");
1390 snprintf(nodestr
, sizeof(nodestr
), "%u", osb
->node_num
);
1392 printk(KERN_INFO
"ocfs2: Unmounting device (%s) on (node %s)\n",
1393 osb
->dev_str
, nodestr
);
1395 ocfs2_delete_osb(osb
);
1398 sb
->s_fs_info
= NULL
;
1401 static int ocfs2_setup_osb_uuid(struct ocfs2_super
*osb
, const unsigned char *uuid
,
1402 unsigned uuid_bytes
)
1407 BUG_ON(uuid_bytes
!= OCFS2_VOL_UUID_LEN
);
1409 osb
->uuid_str
= kzalloc(OCFS2_VOL_UUID_LEN
* 2 + 1, GFP_KERNEL
);
1410 if (osb
->uuid_str
== NULL
)
1413 for (i
= 0, ptr
= osb
->uuid_str
; i
< OCFS2_VOL_UUID_LEN
; i
++) {
1414 /* print with null */
1415 ret
= snprintf(ptr
, 3, "%02X", uuid
[i
]);
1416 if (ret
!= 2) /* drop super cleans up */
1418 /* then only advance past the last char */
1425 static int ocfs2_initialize_super(struct super_block
*sb
,
1426 struct buffer_head
*bh
,
1430 int i
, cbits
, bbits
;
1431 struct ocfs2_dinode
*di
= (struct ocfs2_dinode
*)bh
->b_data
;
1432 struct inode
*inode
= NULL
;
1433 struct ocfs2_journal
*journal
;
1434 __le32 uuid_net_key
;
1435 struct ocfs2_super
*osb
;
1439 osb
= kzalloc(sizeof(struct ocfs2_super
), GFP_KERNEL
);
1446 sb
->s_fs_info
= osb
;
1447 sb
->s_op
= &ocfs2_sops
;
1448 sb
->s_export_op
= &ocfs2_export_ops
;
1449 sb
->s_xattr
= ocfs2_xattr_handlers
;
1450 sb
->s_time_gran
= 1;
1451 sb
->s_flags
|= MS_NOATIME
;
1452 /* this is needed to support O_LARGEFILE */
1453 cbits
= le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
);
1454 bbits
= le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
);
1455 sb
->s_maxbytes
= ocfs2_max_file_offset(bbits
, cbits
);
1458 /* Save off for ocfs2_rw_direct */
1459 osb
->s_sectsize_bits
= blksize_bits(sector_size
);
1460 BUG_ON(!osb
->s_sectsize_bits
);
1462 spin_lock_init(&osb
->dc_task_lock
);
1463 init_waitqueue_head(&osb
->dc_event
);
1464 osb
->dc_work_sequence
= 0;
1465 osb
->dc_wake_sequence
= 0;
1466 INIT_LIST_HEAD(&osb
->blocked_lock_list
);
1467 osb
->blocked_lock_count
= 0;
1468 spin_lock_init(&osb
->osb_lock
);
1469 ocfs2_init_inode_steal_slot(osb
);
1471 atomic_set(&osb
->alloc_stats
.moves
, 0);
1472 atomic_set(&osb
->alloc_stats
.local_data
, 0);
1473 atomic_set(&osb
->alloc_stats
.bitmap_data
, 0);
1474 atomic_set(&osb
->alloc_stats
.bg_allocs
, 0);
1475 atomic_set(&osb
->alloc_stats
.bg_extends
, 0);
1477 ocfs2_init_node_maps(osb
);
1479 snprintf(osb
->dev_str
, sizeof(osb
->dev_str
), "%u,%u",
1480 MAJOR(osb
->sb
->s_dev
), MINOR(osb
->sb
->s_dev
));
1482 status
= ocfs2_recovery_init(osb
);
1484 mlog(ML_ERROR
, "Unable to initialize recovery state\n");
1489 init_waitqueue_head(&osb
->checkpoint_event
);
1490 atomic_set(&osb
->needs_checkpoint
, 0);
1492 osb
->s_atime_quantum
= OCFS2_DEFAULT_ATIME_QUANTUM
;
1494 osb
->slot_num
= OCFS2_INVALID_SLOT
;
1496 osb
->s_xattr_inline_size
= le16_to_cpu(
1497 di
->id2
.i_super
.s_xattr_inline_size
);
1499 osb
->local_alloc_state
= OCFS2_LA_UNUSED
;
1500 osb
->local_alloc_bh
= NULL
;
1501 INIT_DELAYED_WORK(&osb
->la_enable_wq
, ocfs2_la_enable_worker
);
1503 init_waitqueue_head(&osb
->osb_mount_event
);
1505 osb
->vol_label
= kmalloc(OCFS2_MAX_VOL_LABEL_LEN
, GFP_KERNEL
);
1506 if (!osb
->vol_label
) {
1507 mlog(ML_ERROR
, "unable to alloc vol label\n");
1512 osb
->max_slots
= le16_to_cpu(di
->id2
.i_super
.s_max_slots
);
1513 if (osb
->max_slots
> OCFS2_MAX_SLOTS
|| osb
->max_slots
== 0) {
1514 mlog(ML_ERROR
, "Invalid number of node slots (%u)\n",
1519 mlog(0, "max_slots for this device: %u\n", osb
->max_slots
);
1521 osb
->slot_recovery_generations
=
1522 kcalloc(osb
->max_slots
, sizeof(*osb
->slot_recovery_generations
),
1524 if (!osb
->slot_recovery_generations
) {
1530 init_waitqueue_head(&osb
->osb_wipe_event
);
1531 osb
->osb_orphan_wipes
= kcalloc(osb
->max_slots
,
1532 sizeof(*osb
->osb_orphan_wipes
),
1534 if (!osb
->osb_orphan_wipes
) {
1540 osb
->s_feature_compat
=
1541 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_compat
);
1542 osb
->s_feature_ro_compat
=
1543 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_ro_compat
);
1544 osb
->s_feature_incompat
=
1545 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_incompat
);
1547 if ((i
= OCFS2_HAS_INCOMPAT_FEATURE(osb
->sb
, ~OCFS2_FEATURE_INCOMPAT_SUPP
))) {
1548 mlog(ML_ERROR
, "couldn't mount because of unsupported "
1549 "optional features (%x).\n", i
);
1553 if (!(osb
->sb
->s_flags
& MS_RDONLY
) &&
1554 (i
= OCFS2_HAS_RO_COMPAT_FEATURE(osb
->sb
, ~OCFS2_FEATURE_RO_COMPAT_SUPP
))) {
1555 mlog(ML_ERROR
, "couldn't mount RDWR because of "
1556 "unsupported optional features (%x).\n", i
);
1561 if (ocfs2_userspace_stack(osb
)) {
1562 memcpy(osb
->osb_cluster_stack
,
1563 OCFS2_RAW_SB(di
)->s_cluster_info
.ci_stack
,
1564 OCFS2_STACK_LABEL_LEN
);
1565 osb
->osb_cluster_stack
[OCFS2_STACK_LABEL_LEN
] = '\0';
1566 if (strlen(osb
->osb_cluster_stack
) != OCFS2_STACK_LABEL_LEN
) {
1568 "couldn't mount because of an invalid "
1569 "cluster stack label (%s) \n",
1570 osb
->osb_cluster_stack
);
1575 /* The empty string is identical with classic tools that
1576 * don't know about s_cluster_info. */
1577 osb
->osb_cluster_stack
[0] = '\0';
1580 get_random_bytes(&osb
->s_next_generation
, sizeof(u32
));
1583 * This should be done in ocfs2_journal_init(), but unknown
1584 * ordering issues will cause the filesystem to crash.
1585 * If anyone wants to figure out what part of the code
1586 * refers to osb->journal before ocfs2_journal_init() is run,
1589 /* initialize our journal structure */
1591 journal
= kzalloc(sizeof(struct ocfs2_journal
), GFP_KERNEL
);
1593 mlog(ML_ERROR
, "unable to alloc journal\n");
1597 osb
->journal
= journal
;
1598 journal
->j_osb
= osb
;
1600 atomic_set(&journal
->j_num_trans
, 0);
1601 init_rwsem(&journal
->j_trans_barrier
);
1602 init_waitqueue_head(&journal
->j_checkpointed
);
1603 spin_lock_init(&journal
->j_lock
);
1604 journal
->j_trans_id
= (unsigned long) 1;
1605 INIT_LIST_HEAD(&journal
->j_la_cleanups
);
1606 INIT_WORK(&journal
->j_recovery_work
, ocfs2_complete_recovery
);
1607 journal
->j_state
= OCFS2_JOURNAL_FREE
;
1609 /* get some pseudo constants for clustersize bits */
1610 osb
->s_clustersize_bits
=
1611 le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
);
1612 osb
->s_clustersize
= 1 << osb
->s_clustersize_bits
;
1613 mlog(0, "clusterbits=%d\n", osb
->s_clustersize_bits
);
1615 if (osb
->s_clustersize
< OCFS2_MIN_CLUSTERSIZE
||
1616 osb
->s_clustersize
> OCFS2_MAX_CLUSTERSIZE
) {
1617 mlog(ML_ERROR
, "Volume has invalid cluster size (%d)\n",
1618 osb
->s_clustersize
);
1623 if (ocfs2_clusters_to_blocks(osb
->sb
, le32_to_cpu(di
->i_clusters
) - 1)
1625 mlog(ML_ERROR
, "Volume might try to write to blocks beyond "
1626 "what jbd can address in 32 bits.\n");
1631 if (ocfs2_setup_osb_uuid(osb
, di
->id2
.i_super
.s_uuid
,
1632 sizeof(di
->id2
.i_super
.s_uuid
))) {
1633 mlog(ML_ERROR
, "Out of memory trying to setup our uuid.\n");
1638 memcpy(&uuid_net_key
, di
->id2
.i_super
.s_uuid
, sizeof(uuid_net_key
));
1640 strncpy(osb
->vol_label
, di
->id2
.i_super
.s_label
, 63);
1641 osb
->vol_label
[63] = '\0';
1642 osb
->root_blkno
= le64_to_cpu(di
->id2
.i_super
.s_root_blkno
);
1643 osb
->system_dir_blkno
= le64_to_cpu(di
->id2
.i_super
.s_system_dir_blkno
);
1644 osb
->first_cluster_group_blkno
=
1645 le64_to_cpu(di
->id2
.i_super
.s_first_cluster_group
);
1646 osb
->fs_generation
= le32_to_cpu(di
->i_fs_generation
);
1647 osb
->uuid_hash
= le32_to_cpu(di
->id2
.i_super
.s_uuid_hash
);
1648 mlog(0, "vol_label: %s\n", osb
->vol_label
);
1649 mlog(0, "uuid: %s\n", osb
->uuid_str
);
1650 mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
1651 (unsigned long long)osb
->root_blkno
,
1652 (unsigned long long)osb
->system_dir_blkno
);
1654 osb
->osb_dlm_debug
= ocfs2_new_dlm_debug();
1655 if (!osb
->osb_dlm_debug
) {
1661 atomic_set(&osb
->vol_state
, VOLUME_INIT
);
1663 /* load root, system_dir, and all global system inodes */
1664 status
= ocfs2_init_global_system_inodes(osb
);
1673 inode
= ocfs2_get_system_file_inode(osb
, GLOBAL_BITMAP_SYSTEM_INODE
,
1674 OCFS2_INVALID_SLOT
);
1681 osb
->bitmap_blkno
= OCFS2_I(inode
)->ip_blkno
;
1684 osb
->bitmap_cpg
= ocfs2_group_bitmap_size(sb
) * 8;
1686 status
= ocfs2_init_slot_info(osb
);
1698 * will return: -EAGAIN if it is ok to keep searching for superblocks
1699 * -EINVAL if there is a bad superblock
1702 static int ocfs2_verify_volume(struct ocfs2_dinode
*di
,
1703 struct buffer_head
*bh
,
1706 int status
= -EAGAIN
;
1710 if (memcmp(di
->i_signature
, OCFS2_SUPER_BLOCK_SIGNATURE
,
1711 strlen(OCFS2_SUPER_BLOCK_SIGNATURE
)) == 0) {
1713 if ((1 << le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
)) != blksz
) {
1714 mlog(ML_ERROR
, "found superblock with incorrect block "
1715 "size: found %u, should be %u\n",
1716 1 << le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
),
1718 } else if (le16_to_cpu(di
->id2
.i_super
.s_major_rev_level
) !=
1719 OCFS2_MAJOR_REV_LEVEL
||
1720 le16_to_cpu(di
->id2
.i_super
.s_minor_rev_level
) !=
1721 OCFS2_MINOR_REV_LEVEL
) {
1722 mlog(ML_ERROR
, "found superblock with bad version: "
1723 "found %u.%u, should be %u.%u\n",
1724 le16_to_cpu(di
->id2
.i_super
.s_major_rev_level
),
1725 le16_to_cpu(di
->id2
.i_super
.s_minor_rev_level
),
1726 OCFS2_MAJOR_REV_LEVEL
,
1727 OCFS2_MINOR_REV_LEVEL
);
1728 } else if (bh
->b_blocknr
!= le64_to_cpu(di
->i_blkno
)) {
1729 mlog(ML_ERROR
, "bad block number on superblock: "
1730 "found %llu, should be %llu\n",
1731 (unsigned long long)le64_to_cpu(di
->i_blkno
),
1732 (unsigned long long)bh
->b_blocknr
);
1733 } else if (le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
) < 12 ||
1734 le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
) > 20) {
1735 mlog(ML_ERROR
, "bad cluster size found: %u\n",
1736 1 << le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
));
1737 } else if (!le64_to_cpu(di
->id2
.i_super
.s_root_blkno
)) {
1738 mlog(ML_ERROR
, "bad root_blkno: 0\n");
1739 } else if (!le64_to_cpu(di
->id2
.i_super
.s_system_dir_blkno
)) {
1740 mlog(ML_ERROR
, "bad system_dir_blkno: 0\n");
1741 } else if (le16_to_cpu(di
->id2
.i_super
.s_max_slots
) > OCFS2_MAX_SLOTS
) {
1743 "Superblock slots found greater than file system "
1744 "maximum: found %u, max %u\n",
1745 le16_to_cpu(di
->id2
.i_super
.s_max_slots
),
1757 static int ocfs2_check_volume(struct ocfs2_super
*osb
)
1762 struct ocfs2_dinode
*local_alloc
= NULL
; /* only used if we
1768 /* Init our journal object. */
1769 status
= ocfs2_journal_init(osb
->journal
, &dirty
);
1771 mlog(ML_ERROR
, "Could not initialize journal!\n");
1775 /* If the journal was unmounted cleanly then we don't want to
1776 * recover anything. Otherwise, journal_load will do that
1777 * dirty work for us :) */
1779 status
= ocfs2_journal_wipe(osb
->journal
, 0);
1785 mlog(ML_NOTICE
, "File system was not unmounted cleanly, "
1786 "recovering volume.\n");
1789 local
= ocfs2_mount_local(osb
);
1791 /* will play back anything left in the journal. */
1792 status
= ocfs2_journal_load(osb
->journal
, local
, dirty
);
1794 mlog(ML_ERROR
, "ocfs2 journal load failed! %d\n", status
);
1799 /* recover my local alloc if we didn't unmount cleanly. */
1800 status
= ocfs2_begin_local_alloc_recovery(osb
,
1807 /* we complete the recovery process after we've marked
1808 * ourselves as mounted. */
1811 mlog(0, "Journal loaded.\n");
1813 status
= ocfs2_load_local_alloc(osb
);
1820 /* Recovery will be completed after we've mounted the
1821 * rest of the volume. */
1823 osb
->local_alloc_copy
= local_alloc
;
1827 /* go through each journal, trylock it and if you get the
1828 * lock, and it's marked as dirty, set the bit in the recover
1829 * map and launch a recovery thread for it. */
1830 status
= ocfs2_mark_dead_nodes(osb
);
1843 * The routine gets called from dismount or close whenever a dismount on
1844 * volume is requested and the osb open count becomes 1.
1845 * It will remove the osb from the global list and also free up all the
1846 * initialized resources and fileobject.
1848 static void ocfs2_delete_osb(struct ocfs2_super
*osb
)
1852 /* This function assumes that the caller has the main osb resource */
1854 ocfs2_free_slot_info(osb
);
1856 kfree(osb
->osb_orphan_wipes
);
1857 kfree(osb
->slot_recovery_generations
);
1859 * This belongs in journal shutdown, but because we have to
1860 * allocate osb->journal at the start of ocfs2_initalize_osb(),
1863 kfree(osb
->journal
);
1864 if (osb
->local_alloc_copy
)
1865 kfree(osb
->local_alloc_copy
);
1866 kfree(osb
->uuid_str
);
1867 ocfs2_put_dlm_debug(osb
->osb_dlm_debug
);
1868 memset(osb
, 0, sizeof(struct ocfs2_super
));
1873 /* Put OCFS2 into a readonly state, or (if the user specifies it),
1874 * panic(). We do not support continue-on-error operation. */
1875 static void ocfs2_handle_error(struct super_block
*sb
)
1877 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
1879 if (osb
->s_mount_opt
& OCFS2_MOUNT_ERRORS_PANIC
)
1880 panic("OCFS2: (device %s): panic forced after error\n",
1883 ocfs2_set_osb_flag(osb
, OCFS2_OSB_ERROR_FS
);
1885 if (sb
->s_flags
& MS_RDONLY
&&
1886 (ocfs2_is_soft_readonly(osb
) ||
1887 ocfs2_is_hard_readonly(osb
)))
1890 printk(KERN_CRIT
"File system is now read-only due to the potential "
1891 "of on-disk corruption. Please run fsck.ocfs2 once the file "
1892 "system is unmounted.\n");
1893 sb
->s_flags
|= MS_RDONLY
;
1894 ocfs2_set_ro_flag(osb
, 0);
1897 static char error_buf
[1024];
1899 void __ocfs2_error(struct super_block
*sb
,
1900 const char *function
,
1901 const char *fmt
, ...)
1905 va_start(args
, fmt
);
1906 vsnprintf(error_buf
, sizeof(error_buf
), fmt
, args
);
1909 /* Not using mlog here because we want to show the actual
1910 * function the error came from. */
1911 printk(KERN_CRIT
"OCFS2: ERROR (device %s): %s: %s\n",
1912 sb
->s_id
, function
, error_buf
);
1914 ocfs2_handle_error(sb
);
1917 /* Handle critical errors. This is intentionally more drastic than
1918 * ocfs2_handle_error, so we only use for things like journal errors,
1920 void __ocfs2_abort(struct super_block
* sb
,
1921 const char *function
,
1922 const char *fmt
, ...)
1926 va_start(args
, fmt
);
1927 vsnprintf(error_buf
, sizeof(error_buf
), fmt
, args
);
1930 printk(KERN_CRIT
"OCFS2: abort (device %s): %s: %s\n",
1931 sb
->s_id
, function
, error_buf
);
1933 /* We don't have the cluster support yet to go straight to
1934 * hard readonly in here. Until then, we want to keep
1935 * ocfs2_abort() so that we can at least mark critical
1938 * TODO: This should abort the journal and alert other nodes
1939 * that our slot needs recovery. */
1941 /* Force a panic(). This stinks, but it's better than letting
1942 * things continue without having a proper hard readonly
1944 OCFS2_SB(sb
)->s_mount_opt
|= OCFS2_MOUNT_ERRORS_PANIC
;
1945 ocfs2_handle_error(sb
);
1948 module_init(ocfs2_init
);
1949 module_exit(ocfs2_exit
);