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"
70 #include "buffer_head_io.h"
72 static struct kmem_cache
*ocfs2_inode_cachep
= NULL
;
73 struct kmem_cache
*ocfs2_dquot_cachep
;
74 struct kmem_cache
*ocfs2_qf_chunk_cachep
;
76 /* OCFS2 needs to schedule several differnt types of work which
77 * require cluster locking, disk I/O, recovery waits, etc. Since these
78 * types of work tend to be heavy we avoid using the kernel events
79 * workqueue and schedule on our own. */
80 struct workqueue_struct
*ocfs2_wq
= NULL
;
82 static struct dentry
*ocfs2_debugfs_root
= NULL
;
84 MODULE_AUTHOR("Oracle");
85 MODULE_LICENSE("GPL");
89 unsigned long commit_interval
;
90 unsigned long mount_opt
;
91 unsigned int atime_quantum
;
93 unsigned int localalloc_opt
;
94 char cluster_stack
[OCFS2_STACK_LABEL_LEN
+ 1];
97 static int ocfs2_parse_options(struct super_block
*sb
, char *options
,
98 struct mount_options
*mopt
,
100 static int ocfs2_show_options(struct seq_file
*s
, struct vfsmount
*mnt
);
101 static void ocfs2_put_super(struct super_block
*sb
);
102 static int ocfs2_mount_volume(struct super_block
*sb
);
103 static int ocfs2_remount(struct super_block
*sb
, int *flags
, char *data
);
104 static void ocfs2_dismount_volume(struct super_block
*sb
, int mnt_err
);
105 static int ocfs2_initialize_mem_caches(void);
106 static void ocfs2_free_mem_caches(void);
107 static void ocfs2_delete_osb(struct ocfs2_super
*osb
);
109 static int ocfs2_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
111 static int ocfs2_sync_fs(struct super_block
*sb
, int wait
);
113 static int ocfs2_init_global_system_inodes(struct ocfs2_super
*osb
);
114 static int ocfs2_init_local_system_inodes(struct ocfs2_super
*osb
);
115 static void ocfs2_release_system_inodes(struct ocfs2_super
*osb
);
116 static int ocfs2_check_volume(struct ocfs2_super
*osb
);
117 static int ocfs2_verify_volume(struct ocfs2_dinode
*di
,
118 struct buffer_head
*bh
,
120 static int ocfs2_initialize_super(struct super_block
*sb
,
121 struct buffer_head
*bh
,
123 static int ocfs2_get_sector(struct super_block
*sb
,
124 struct buffer_head
**bh
,
127 static void ocfs2_write_super(struct super_block
*sb
);
128 static struct inode
*ocfs2_alloc_inode(struct super_block
*sb
);
129 static void ocfs2_destroy_inode(struct inode
*inode
);
131 static const struct super_operations ocfs2_sops
= {
132 .statfs
= ocfs2_statfs
,
133 .alloc_inode
= ocfs2_alloc_inode
,
134 .destroy_inode
= ocfs2_destroy_inode
,
135 .drop_inode
= ocfs2_drop_inode
,
136 .clear_inode
= ocfs2_clear_inode
,
137 .delete_inode
= ocfs2_delete_inode
,
138 .sync_fs
= ocfs2_sync_fs
,
139 .write_super
= ocfs2_write_super
,
140 .put_super
= ocfs2_put_super
,
141 .remount_fs
= ocfs2_remount
,
142 .show_options
= ocfs2_show_options
,
143 .quota_read
= ocfs2_quota_read
,
144 .quota_write
= ocfs2_quota_write
,
171 static const match_table_t tokens
= {
172 {Opt_barrier
, "barrier=%u"},
173 {Opt_err_panic
, "errors=panic"},
174 {Opt_err_ro
, "errors=remount-ro"},
176 {Opt_nointr
, "nointr"},
177 {Opt_hb_none
, OCFS2_HB_NONE
},
178 {Opt_hb_local
, OCFS2_HB_LOCAL
},
179 {Opt_data_ordered
, "data=ordered"},
180 {Opt_data_writeback
, "data=writeback"},
181 {Opt_atime_quantum
, "atime_quantum=%u"},
182 {Opt_slot
, "preferred_slot=%u"},
183 {Opt_commit
, "commit=%u"},
184 {Opt_localalloc
, "localalloc=%d"},
185 {Opt_localflocks
, "localflocks"},
186 {Opt_stack
, "cluster_stack=%s"},
187 {Opt_user_xattr
, "user_xattr"},
188 {Opt_nouser_xattr
, "nouser_xattr"},
189 {Opt_inode64
, "inode64"},
191 {Opt_noacl
, "noacl"},
196 * write_super and sync_fs ripped right out of ext3.
198 static void ocfs2_write_super(struct super_block
*sb
)
200 if (mutex_trylock(&sb
->s_lock
) != 0)
205 static int ocfs2_sync_fs(struct super_block
*sb
, int wait
)
209 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
213 if (ocfs2_is_hard_readonly(osb
))
217 status
= ocfs2_flush_truncate_log(osb
);
221 ocfs2_schedule_truncate_log_flush(osb
, 0);
224 if (jbd2_journal_start_commit(OCFS2_SB(sb
)->journal
->j_journal
,
227 jbd2_log_wait_commit(OCFS2_SB(sb
)->journal
->j_journal
,
233 static int ocfs2_need_system_inode(struct ocfs2_super
*osb
, int ino
)
235 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb
->sb
, OCFS2_FEATURE_RO_COMPAT_USRQUOTA
)
236 && (ino
== USER_QUOTA_SYSTEM_INODE
237 || ino
== LOCAL_USER_QUOTA_SYSTEM_INODE
))
239 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb
->sb
, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA
)
240 && (ino
== GROUP_QUOTA_SYSTEM_INODE
241 || ino
== LOCAL_GROUP_QUOTA_SYSTEM_INODE
))
246 static int ocfs2_init_global_system_inodes(struct ocfs2_super
*osb
)
248 struct inode
*new = NULL
;
254 new = ocfs2_iget(osb
, osb
->root_blkno
, OCFS2_FI_FLAG_SYSFILE
, 0);
256 status
= PTR_ERR(new);
260 osb
->root_inode
= new;
262 new = ocfs2_iget(osb
, osb
->system_dir_blkno
, OCFS2_FI_FLAG_SYSFILE
, 0);
264 status
= PTR_ERR(new);
268 osb
->sys_root_inode
= new;
270 for (i
= OCFS2_FIRST_ONLINE_SYSTEM_INODE
;
271 i
<= OCFS2_LAST_GLOBAL_SYSTEM_INODE
; i
++) {
272 if (!ocfs2_need_system_inode(osb
, i
))
274 new = ocfs2_get_system_file_inode(osb
, i
, osb
->slot_num
);
276 ocfs2_release_system_inodes(osb
);
279 /* FIXME: Should ERROR_RO_FS */
280 mlog(ML_ERROR
, "Unable to load system inode %d, "
281 "possibly corrupt fs?", i
);
284 // the array now has one ref, so drop this one
293 static int ocfs2_init_local_system_inodes(struct ocfs2_super
*osb
)
295 struct inode
*new = NULL
;
301 for (i
= OCFS2_LAST_GLOBAL_SYSTEM_INODE
+ 1;
302 i
< NUM_SYSTEM_INODES
;
304 if (!ocfs2_need_system_inode(osb
, i
))
306 new = ocfs2_get_system_file_inode(osb
, i
, osb
->slot_num
);
308 ocfs2_release_system_inodes(osb
);
310 mlog(ML_ERROR
, "status=%d, sysfile=%d, slot=%d\n",
311 status
, i
, osb
->slot_num
);
314 /* the array now has one ref, so drop this one */
323 static void ocfs2_release_system_inodes(struct ocfs2_super
*osb
)
330 for (i
= 0; i
< NUM_SYSTEM_INODES
; i
++) {
331 inode
= osb
->system_inodes
[i
];
334 osb
->system_inodes
[i
] = NULL
;
338 inode
= osb
->sys_root_inode
;
341 osb
->sys_root_inode
= NULL
;
344 inode
= osb
->root_inode
;
347 osb
->root_inode
= NULL
;
353 /* We're allocating fs objects, use GFP_NOFS */
354 static struct inode
*ocfs2_alloc_inode(struct super_block
*sb
)
356 struct ocfs2_inode_info
*oi
;
358 oi
= kmem_cache_alloc(ocfs2_inode_cachep
, GFP_NOFS
);
362 jbd2_journal_init_jbd_inode(&oi
->ip_jinode
, &oi
->vfs_inode
);
363 return &oi
->vfs_inode
;
366 static void ocfs2_destroy_inode(struct inode
*inode
)
368 kmem_cache_free(ocfs2_inode_cachep
, OCFS2_I(inode
));
371 static unsigned long long ocfs2_max_file_offset(unsigned int bbits
,
374 unsigned int bytes
= 1 << cbits
;
375 unsigned int trim
= bytes
;
376 unsigned int bitshift
= 32;
379 * i_size and all block offsets in ocfs2 are always 64 bits
380 * wide. i_clusters is 32 bits, in cluster-sized units. So on
381 * 64 bit platforms, cluster size will be the limiting factor.
384 #if BITS_PER_LONG == 32
385 # if defined(CONFIG_LBD)
386 BUILD_BUG_ON(sizeof(sector_t
) != 8);
388 * We might be limited by page cache size.
390 if (bytes
> PAGE_CACHE_SIZE
) {
391 bytes
= PAGE_CACHE_SIZE
;
394 * Shift by 31 here so that we don't get larger than
401 * We are limited by the size of sector_t. Use block size, as
402 * that's what we expose to the VFS.
411 * Trim by a whole cluster when we can actually approach the
412 * on-disk limits. Otherwise we can overflow i_clusters when
413 * an extent start is at the max offset.
415 return (((unsigned long long)bytes
) << bitshift
) - trim
;
418 static int ocfs2_remount(struct super_block
*sb
, int *flags
, char *data
)
420 int incompat_features
;
422 struct mount_options parsed_options
;
423 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
425 if (!ocfs2_parse_options(sb
, data
, &parsed_options
, 1)) {
430 if ((osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) !=
431 (parsed_options
.mount_opt
& OCFS2_MOUNT_HB_LOCAL
)) {
433 mlog(ML_ERROR
, "Cannot change heartbeat mode on remount\n");
437 if ((osb
->s_mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
) !=
438 (parsed_options
.mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
)) {
440 mlog(ML_ERROR
, "Cannot change data mode on remount\n");
444 /* Probably don't want this on remount; it might
445 * mess with other nodes */
446 if (!(osb
->s_mount_opt
& OCFS2_MOUNT_INODE64
) &&
447 (parsed_options
.mount_opt
& OCFS2_MOUNT_INODE64
)) {
449 mlog(ML_ERROR
, "Cannot enable inode64 on remount\n");
453 /* We're going to/from readonly mode. */
454 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
)) {
455 /* Lock here so the check of HARD_RO and the potential
456 * setting of SOFT_RO is atomic. */
457 spin_lock(&osb
->osb_lock
);
458 if (osb
->osb_flags
& OCFS2_OSB_HARD_RO
) {
459 mlog(ML_ERROR
, "Remount on readonly device is forbidden.\n");
464 if (*flags
& MS_RDONLY
) {
465 mlog(0, "Going to ro mode.\n");
466 sb
->s_flags
|= MS_RDONLY
;
467 osb
->osb_flags
|= OCFS2_OSB_SOFT_RO
;
469 mlog(0, "Making ro filesystem writeable.\n");
471 if (osb
->osb_flags
& OCFS2_OSB_ERROR_FS
) {
472 mlog(ML_ERROR
, "Cannot remount RDWR "
473 "filesystem due to previous errors.\n");
477 incompat_features
= OCFS2_HAS_RO_COMPAT_FEATURE(sb
, ~OCFS2_FEATURE_RO_COMPAT_SUPP
);
478 if (incompat_features
) {
479 mlog(ML_ERROR
, "Cannot remount RDWR because "
480 "of unsupported optional features "
481 "(%x).\n", incompat_features
);
485 sb
->s_flags
&= ~MS_RDONLY
;
486 osb
->osb_flags
&= ~OCFS2_OSB_SOFT_RO
;
489 spin_unlock(&osb
->osb_lock
);
493 /* Only save off the new mount options in case of a successful
495 if (!(osb
->s_feature_incompat
& OCFS2_FEATURE_INCOMPAT_XATTR
))
496 parsed_options
.mount_opt
&= ~OCFS2_MOUNT_POSIX_ACL
;
497 osb
->s_mount_opt
= parsed_options
.mount_opt
;
498 osb
->s_atime_quantum
= parsed_options
.atime_quantum
;
499 osb
->preferred_slot
= parsed_options
.slot
;
500 if (parsed_options
.commit_interval
)
501 osb
->osb_commit_interval
= parsed_options
.commit_interval
;
503 if (!ocfs2_is_hard_readonly(osb
))
504 ocfs2_set_journal_params(osb
);
510 static int ocfs2_sb_probe(struct super_block
*sb
,
511 struct buffer_head
**bh
,
515 struct ocfs1_vol_disk_hdr
*hdr
;
516 struct ocfs2_dinode
*di
;
522 *sector_size
= bdev_hardsect_size(sb
->s_bdev
);
523 if (*sector_size
> OCFS2_MAX_BLOCKSIZE
) {
524 mlog(ML_ERROR
, "Hardware sector size too large: %d (max=%d)\n",
525 *sector_size
, OCFS2_MAX_BLOCKSIZE
);
530 /* Can this really happen? */
531 if (*sector_size
< OCFS2_MIN_BLOCKSIZE
)
532 *sector_size
= OCFS2_MIN_BLOCKSIZE
;
534 /* check block zero for old format */
535 status
= ocfs2_get_sector(sb
, bh
, 0, *sector_size
);
540 hdr
= (struct ocfs1_vol_disk_hdr
*) (*bh
)->b_data
;
541 if (hdr
->major_version
== OCFS1_MAJOR_VERSION
) {
542 mlog(ML_ERROR
, "incompatible version: %u.%u\n",
543 hdr
->major_version
, hdr
->minor_version
);
546 if (memcmp(hdr
->signature
, OCFS1_VOLUME_SIGNATURE
,
547 strlen(OCFS1_VOLUME_SIGNATURE
)) == 0) {
548 mlog(ML_ERROR
, "incompatible volume signature: %8s\n",
555 mlog(ML_ERROR
, "This is an ocfs v1 filesystem which must be "
556 "upgraded before mounting with ocfs v2\n");
561 * Now check at magic offset for 512, 1024, 2048, 4096
562 * blocksizes. 4096 is the maximum blocksize because it is
563 * the minimum clustersize.
566 for (blksize
= *sector_size
;
567 blksize
<= OCFS2_MAX_BLOCKSIZE
;
569 tmpstat
= ocfs2_get_sector(sb
, bh
,
570 OCFS2_SUPER_BLOCK_BLKNO
,
577 di
= (struct ocfs2_dinode
*) (*bh
)->b_data
;
578 status
= ocfs2_verify_volume(di
, *bh
, blksize
);
583 if (status
!= -EAGAIN
)
591 static int ocfs2_verify_heartbeat(struct ocfs2_super
*osb
)
593 if (ocfs2_mount_local(osb
)) {
594 if (osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) {
595 mlog(ML_ERROR
, "Cannot heartbeat on a locally "
596 "mounted device.\n");
601 if (ocfs2_userspace_stack(osb
)) {
602 if (osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) {
603 mlog(ML_ERROR
, "Userspace stack expected, but "
604 "o2cb heartbeat arguments passed to mount\n");
609 if (!(osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
)) {
610 if (!ocfs2_mount_local(osb
) && !ocfs2_is_hard_readonly(osb
) &&
611 !ocfs2_userspace_stack(osb
)) {
612 mlog(ML_ERROR
, "Heartbeat has to be started to mount "
613 "a read-write clustered device.\n");
622 * If we're using a userspace stack, mount should have passed
623 * a name that matches the disk. If not, mount should not
624 * have passed a stack.
626 static int ocfs2_verify_userspace_stack(struct ocfs2_super
*osb
,
627 struct mount_options
*mopt
)
629 if (!ocfs2_userspace_stack(osb
) && mopt
->cluster_stack
[0]) {
631 "cluster stack passed to mount, but this filesystem "
632 "does not support it\n");
636 if (ocfs2_userspace_stack(osb
) &&
637 strncmp(osb
->osb_cluster_stack
, mopt
->cluster_stack
,
638 OCFS2_STACK_LABEL_LEN
)) {
640 "cluster stack passed to mount (\"%s\") does not "
641 "match the filesystem (\"%s\")\n",
643 osb
->osb_cluster_stack
);
650 static int ocfs2_fill_super(struct super_block
*sb
, void *data
, int silent
)
653 int status
, sector_size
;
654 struct mount_options parsed_options
;
655 struct inode
*inode
= NULL
;
656 struct ocfs2_super
*osb
= NULL
;
657 struct buffer_head
*bh
= NULL
;
660 mlog_entry("%p, %p, %i", sb
, data
, silent
);
662 if (!ocfs2_parse_options(sb
, data
, &parsed_options
, 0)) {
664 goto read_super_error
;
667 /* probe for superblock */
668 status
= ocfs2_sb_probe(sb
, &bh
, §or_size
);
670 mlog(ML_ERROR
, "superblock probe failed!\n");
671 goto read_super_error
;
674 status
= ocfs2_initialize_super(sb
, bh
, sector_size
);
678 goto read_super_error
;
683 if (!(osb
->s_feature_incompat
& OCFS2_FEATURE_INCOMPAT_XATTR
))
684 parsed_options
.mount_opt
&= ~OCFS2_MOUNT_POSIX_ACL
;
686 osb
->s_mount_opt
= parsed_options
.mount_opt
;
687 osb
->s_atime_quantum
= parsed_options
.atime_quantum
;
688 osb
->preferred_slot
= parsed_options
.slot
;
689 osb
->osb_commit_interval
= parsed_options
.commit_interval
;
690 osb
->local_alloc_default_bits
= ocfs2_megabytes_to_clusters(sb
, parsed_options
.localalloc_opt
);
691 osb
->local_alloc_bits
= osb
->local_alloc_default_bits
;
693 status
= ocfs2_verify_userspace_stack(osb
, &parsed_options
);
695 goto read_super_error
;
697 sb
->s_magic
= OCFS2_SUPER_MAGIC
;
699 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
700 ((osb
->s_mount_opt
& OCFS2_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
702 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
704 if (bdev_read_only(sb
->s_bdev
)) {
705 if (!(sb
->s_flags
& MS_RDONLY
)) {
707 mlog(ML_ERROR
, "Readonly device detected but readonly "
708 "mount was not specified.\n");
709 goto read_super_error
;
712 /* You should not be able to start a local heartbeat
713 * on a readonly device. */
714 if (osb
->s_mount_opt
& OCFS2_MOUNT_HB_LOCAL
) {
716 mlog(ML_ERROR
, "Local heartbeat specified on readonly "
718 goto read_super_error
;
721 status
= ocfs2_check_journals_nolocks(osb
);
723 if (status
== -EROFS
)
724 mlog(ML_ERROR
, "Recovery required on readonly "
725 "file system, but write access is "
729 goto read_super_error
;
732 ocfs2_set_ro_flag(osb
, 1);
734 printk(KERN_NOTICE
"Readonly device detected. No cluster "
735 "services will be utilized for this mount. Recovery "
736 "will be skipped.\n");
739 if (!ocfs2_is_hard_readonly(osb
)) {
740 if (sb
->s_flags
& MS_RDONLY
)
741 ocfs2_set_ro_flag(osb
, 0);
744 status
= ocfs2_verify_heartbeat(osb
);
747 goto read_super_error
;
750 osb
->osb_debug_root
= debugfs_create_dir(osb
->uuid_str
,
752 if (!osb
->osb_debug_root
) {
754 mlog(ML_ERROR
, "Unable to create per-mount debugfs root.\n");
755 goto read_super_error
;
758 status
= ocfs2_mount_volume(sb
);
760 inode
= igrab(osb
->root_inode
);
763 goto read_super_error
;
768 goto read_super_error
;
771 root
= d_alloc_root(inode
);
775 goto read_super_error
;
780 ocfs2_complete_mount_recovery(osb
);
782 if (ocfs2_mount_local(osb
))
783 snprintf(nodestr
, sizeof(nodestr
), "local");
785 snprintf(nodestr
, sizeof(nodestr
), "%u", osb
->node_num
);
787 printk(KERN_INFO
"ocfs2: Mounting device (%s) on (node %s, slot %d) "
788 "with %s data mode.\n",
789 osb
->dev_str
, nodestr
, osb
->slot_num
,
790 osb
->s_mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
? "writeback" :
793 atomic_set(&osb
->vol_state
, VOLUME_MOUNTED
);
794 wake_up(&osb
->osb_mount_event
);
806 atomic_set(&osb
->vol_state
, VOLUME_DISABLED
);
807 wake_up(&osb
->osb_mount_event
);
808 ocfs2_dismount_volume(sb
, 1);
815 static int ocfs2_get_sb(struct file_system_type
*fs_type
,
817 const char *dev_name
,
819 struct vfsmount
*mnt
)
821 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, ocfs2_fill_super
,
825 static struct file_system_type ocfs2_fs_type
= {
826 .owner
= THIS_MODULE
,
828 .get_sb
= ocfs2_get_sb
, /* is this called when we mount
830 .kill_sb
= kill_block_super
, /* set to the generic one
831 * right now, but do we
832 * need to change that? */
833 .fs_flags
= FS_REQUIRES_DEV
|FS_RENAME_DOES_D_MOVE
,
837 static int ocfs2_parse_options(struct super_block
*sb
,
839 struct mount_options
*mopt
,
845 mlog_entry("remount: %d, options: \"%s\"\n", is_remount
,
846 options
? options
: "(none)");
848 mopt
->commit_interval
= 0;
850 mopt
->atime_quantum
= OCFS2_DEFAULT_ATIME_QUANTUM
;
851 mopt
->slot
= OCFS2_INVALID_SLOT
;
852 mopt
->localalloc_opt
= OCFS2_DEFAULT_LOCAL_ALLOC_SIZE
;
853 mopt
->cluster_stack
[0] = '\0';
860 while ((p
= strsep(&options
, ",")) != NULL
) {
862 substring_t args
[MAX_OPT_ARGS
];
867 token
= match_token(p
, tokens
, args
);
870 mopt
->mount_opt
|= OCFS2_MOUNT_HB_LOCAL
;
873 mopt
->mount_opt
&= ~OCFS2_MOUNT_HB_LOCAL
;
876 if (match_int(&args
[0], &option
)) {
881 mopt
->mount_opt
|= OCFS2_MOUNT_BARRIER
;
883 mopt
->mount_opt
&= ~OCFS2_MOUNT_BARRIER
;
886 mopt
->mount_opt
&= ~OCFS2_MOUNT_NOINTR
;
889 mopt
->mount_opt
|= OCFS2_MOUNT_NOINTR
;
892 mopt
->mount_opt
|= OCFS2_MOUNT_ERRORS_PANIC
;
895 mopt
->mount_opt
&= ~OCFS2_MOUNT_ERRORS_PANIC
;
897 case Opt_data_ordered
:
898 mopt
->mount_opt
&= ~OCFS2_MOUNT_DATA_WRITEBACK
;
900 case Opt_data_writeback
:
901 mopt
->mount_opt
|= OCFS2_MOUNT_DATA_WRITEBACK
;
904 mopt
->mount_opt
&= ~OCFS2_MOUNT_NOUSERXATTR
;
906 case Opt_nouser_xattr
:
907 mopt
->mount_opt
|= OCFS2_MOUNT_NOUSERXATTR
;
909 case Opt_atime_quantum
:
910 if (match_int(&args
[0], &option
)) {
915 mopt
->atime_quantum
= option
;
919 if (match_int(&args
[0], &option
)) {
924 mopt
->slot
= (s16
)option
;
928 if (match_int(&args
[0], &option
)) {
935 option
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
936 mopt
->commit_interval
= HZ
* option
;
940 if (match_int(&args
[0], &option
)) {
944 if (option
>= 0 && (option
<= ocfs2_local_alloc_size(sb
) * 8))
945 mopt
->localalloc_opt
= option
;
947 case Opt_localflocks
:
949 * Changing this during remount could race
950 * flock() requests, or "unbalance" existing
951 * ones (e.g., a lock is taken in one mode but
952 * dropped in the other). If users care enough
953 * to flip locking modes during remount, we
954 * could add a "local" flag to individual
955 * flock structures for proper tracking of
959 mopt
->mount_opt
|= OCFS2_MOUNT_LOCALFLOCKS
;
962 /* Check both that the option we were passed
963 * is of the right length and that it is a proper
964 * string of the right length.
966 if (((args
[0].to
- args
[0].from
) !=
967 OCFS2_STACK_LABEL_LEN
) ||
968 (strnlen(args
[0].from
,
969 OCFS2_STACK_LABEL_LEN
) !=
970 OCFS2_STACK_LABEL_LEN
)) {
972 "Invalid cluster_stack option\n");
976 memcpy(mopt
->cluster_stack
, args
[0].from
,
977 OCFS2_STACK_LABEL_LEN
);
978 mopt
->cluster_stack
[OCFS2_STACK_LABEL_LEN
] = '\0';
981 mopt
->mount_opt
|= OCFS2_MOUNT_INODE64
;
983 #ifdef CONFIG_OCFS2_FS_POSIX_ACL
985 mopt
->mount_opt
|= OCFS2_MOUNT_POSIX_ACL
;
988 mopt
->mount_opt
&= ~OCFS2_MOUNT_POSIX_ACL
;
993 printk(KERN_INFO
"ocfs2 (no)acl options not supported\n");
998 "Unrecognized mount option \"%s\" "
999 "or missing value\n", p
);
1012 static int ocfs2_show_options(struct seq_file
*s
, struct vfsmount
*mnt
)
1014 struct ocfs2_super
*osb
= OCFS2_SB(mnt
->mnt_sb
);
1015 unsigned long opts
= osb
->s_mount_opt
;
1016 unsigned int local_alloc_megs
;
1018 if (opts
& OCFS2_MOUNT_HB_LOCAL
)
1019 seq_printf(s
, ",_netdev,heartbeat=local");
1021 seq_printf(s
, ",heartbeat=none");
1023 if (opts
& OCFS2_MOUNT_NOINTR
)
1024 seq_printf(s
, ",nointr");
1026 if (opts
& OCFS2_MOUNT_DATA_WRITEBACK
)
1027 seq_printf(s
, ",data=writeback");
1029 seq_printf(s
, ",data=ordered");
1031 if (opts
& OCFS2_MOUNT_BARRIER
)
1032 seq_printf(s
, ",barrier=1");
1034 if (opts
& OCFS2_MOUNT_ERRORS_PANIC
)
1035 seq_printf(s
, ",errors=panic");
1037 seq_printf(s
, ",errors=remount-ro");
1039 if (osb
->preferred_slot
!= OCFS2_INVALID_SLOT
)
1040 seq_printf(s
, ",preferred_slot=%d", osb
->preferred_slot
);
1042 if (osb
->s_atime_quantum
!= OCFS2_DEFAULT_ATIME_QUANTUM
)
1043 seq_printf(s
, ",atime_quantum=%u", osb
->s_atime_quantum
);
1045 if (osb
->osb_commit_interval
)
1046 seq_printf(s
, ",commit=%u",
1047 (unsigned) (osb
->osb_commit_interval
/ HZ
));
1049 local_alloc_megs
= osb
->local_alloc_bits
>> (20 - osb
->s_clustersize_bits
);
1050 if (local_alloc_megs
!= OCFS2_DEFAULT_LOCAL_ALLOC_SIZE
)
1051 seq_printf(s
, ",localalloc=%d", local_alloc_megs
);
1053 if (opts
& OCFS2_MOUNT_LOCALFLOCKS
)
1054 seq_printf(s
, ",localflocks,");
1056 if (osb
->osb_cluster_stack
[0])
1057 seq_printf(s
, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN
,
1058 osb
->osb_cluster_stack
);
1060 if (opts
& OCFS2_MOUNT_NOUSERXATTR
)
1061 seq_printf(s
, ",nouser_xattr");
1063 seq_printf(s
, ",user_xattr");
1065 if (opts
& OCFS2_MOUNT_INODE64
)
1066 seq_printf(s
, ",inode64");
1068 #ifdef CONFIG_OCFS2_FS_POSIX_ACL
1069 if (opts
& OCFS2_MOUNT_POSIX_ACL
)
1070 seq_printf(s
, ",acl");
1072 seq_printf(s
, ",noacl");
1078 static int __init
ocfs2_init(void)
1084 ocfs2_print_version();
1086 status
= init_ocfs2_uptodate_cache();
1092 status
= ocfs2_initialize_mem_caches();
1098 ocfs2_wq
= create_singlethread_workqueue("ocfs2_wq");
1104 ocfs2_debugfs_root
= debugfs_create_dir("ocfs2", NULL
);
1105 if (!ocfs2_debugfs_root
) {
1107 mlog(ML_ERROR
, "Unable to create ocfs2 debugfs root.\n");
1110 ocfs2_set_locking_protocol();
1112 status
= register_quota_format(&ocfs2_quota_format
);
1115 ocfs2_free_mem_caches();
1116 exit_ocfs2_uptodate_cache();
1122 return register_filesystem(&ocfs2_fs_type
);
1127 static void __exit
ocfs2_exit(void)
1132 flush_workqueue(ocfs2_wq
);
1133 destroy_workqueue(ocfs2_wq
);
1136 unregister_quota_format(&ocfs2_quota_format
);
1138 debugfs_remove(ocfs2_debugfs_root
);
1140 ocfs2_free_mem_caches();
1142 unregister_filesystem(&ocfs2_fs_type
);
1144 exit_ocfs2_uptodate_cache();
1149 static void ocfs2_put_super(struct super_block
*sb
)
1151 mlog_entry("(0x%p)\n", sb
);
1153 ocfs2_sync_blockdev(sb
);
1154 ocfs2_dismount_volume(sb
, 0);
1159 static int ocfs2_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
1161 struct ocfs2_super
*osb
;
1162 u32 numbits
, freebits
;
1164 struct ocfs2_dinode
*bm_lock
;
1165 struct buffer_head
*bh
= NULL
;
1166 struct inode
*inode
= NULL
;
1168 mlog_entry("(%p, %p)\n", dentry
->d_sb
, buf
);
1170 osb
= OCFS2_SB(dentry
->d_sb
);
1172 inode
= ocfs2_get_system_file_inode(osb
,
1173 GLOBAL_BITMAP_SYSTEM_INODE
,
1174 OCFS2_INVALID_SLOT
);
1176 mlog(ML_ERROR
, "failed to get bitmap inode\n");
1181 status
= ocfs2_inode_lock(inode
, &bh
, 0);
1187 bm_lock
= (struct ocfs2_dinode
*) bh
->b_data
;
1189 numbits
= le32_to_cpu(bm_lock
->id1
.bitmap1
.i_total
);
1190 freebits
= numbits
- le32_to_cpu(bm_lock
->id1
.bitmap1
.i_used
);
1192 buf
->f_type
= OCFS2_SUPER_MAGIC
;
1193 buf
->f_bsize
= dentry
->d_sb
->s_blocksize
;
1194 buf
->f_namelen
= OCFS2_MAX_FILENAME_LEN
;
1195 buf
->f_blocks
= ((sector_t
) numbits
) *
1196 (osb
->s_clustersize
>> osb
->sb
->s_blocksize_bits
);
1197 buf
->f_bfree
= ((sector_t
) freebits
) *
1198 (osb
->s_clustersize
>> osb
->sb
->s_blocksize_bits
);
1199 buf
->f_bavail
= buf
->f_bfree
;
1200 buf
->f_files
= numbits
;
1201 buf
->f_ffree
= freebits
;
1205 ocfs2_inode_unlock(inode
, 0);
1216 static void ocfs2_inode_init_once(void *data
)
1218 struct ocfs2_inode_info
*oi
= data
;
1221 oi
->ip_open_count
= 0;
1222 spin_lock_init(&oi
->ip_lock
);
1223 ocfs2_extent_map_init(&oi
->vfs_inode
);
1224 INIT_LIST_HEAD(&oi
->ip_io_markers
);
1225 oi
->ip_created_trans
= 0;
1226 oi
->ip_last_trans
= 0;
1227 oi
->ip_dir_start_lookup
= 0;
1229 init_rwsem(&oi
->ip_alloc_sem
);
1230 init_rwsem(&oi
->ip_xattr_sem
);
1231 mutex_init(&oi
->ip_io_mutex
);
1233 oi
->ip_blkno
= 0ULL;
1234 oi
->ip_clusters
= 0;
1236 ocfs2_lock_res_init_once(&oi
->ip_rw_lockres
);
1237 ocfs2_lock_res_init_once(&oi
->ip_inode_lockres
);
1238 ocfs2_lock_res_init_once(&oi
->ip_open_lockres
);
1240 ocfs2_metadata_cache_init(&oi
->vfs_inode
);
1242 inode_init_once(&oi
->vfs_inode
);
1245 static int ocfs2_initialize_mem_caches(void)
1247 ocfs2_inode_cachep
= kmem_cache_create("ocfs2_inode_cache",
1248 sizeof(struct ocfs2_inode_info
),
1250 (SLAB_HWCACHE_ALIGN
|SLAB_RECLAIM_ACCOUNT
|
1252 ocfs2_inode_init_once
);
1253 ocfs2_dquot_cachep
= kmem_cache_create("ocfs2_dquot_cache",
1254 sizeof(struct ocfs2_dquot
),
1256 (SLAB_HWCACHE_ALIGN
|SLAB_RECLAIM_ACCOUNT
|
1259 ocfs2_qf_chunk_cachep
= kmem_cache_create("ocfs2_qf_chunk_cache",
1260 sizeof(struct ocfs2_quota_chunk
),
1262 (SLAB_RECLAIM_ACCOUNT
|SLAB_MEM_SPREAD
),
1264 if (!ocfs2_inode_cachep
|| !ocfs2_dquot_cachep
||
1265 !ocfs2_qf_chunk_cachep
) {
1266 if (ocfs2_inode_cachep
)
1267 kmem_cache_destroy(ocfs2_inode_cachep
);
1268 if (ocfs2_dquot_cachep
)
1269 kmem_cache_destroy(ocfs2_dquot_cachep
);
1270 if (ocfs2_qf_chunk_cachep
)
1271 kmem_cache_destroy(ocfs2_qf_chunk_cachep
);
1278 static void ocfs2_free_mem_caches(void)
1280 if (ocfs2_inode_cachep
)
1281 kmem_cache_destroy(ocfs2_inode_cachep
);
1282 ocfs2_inode_cachep
= NULL
;
1284 if (ocfs2_dquot_cachep
)
1285 kmem_cache_destroy(ocfs2_dquot_cachep
);
1286 ocfs2_dquot_cachep
= NULL
;
1288 if (ocfs2_qf_chunk_cachep
)
1289 kmem_cache_destroy(ocfs2_qf_chunk_cachep
);
1290 ocfs2_qf_chunk_cachep
= NULL
;
1293 static int ocfs2_get_sector(struct super_block
*sb
,
1294 struct buffer_head
**bh
,
1298 if (!sb_set_blocksize(sb
, sect_size
)) {
1299 mlog(ML_ERROR
, "unable to set blocksize\n");
1303 *bh
= sb_getblk(sb
, block
);
1309 if (!buffer_dirty(*bh
))
1310 clear_buffer_uptodate(*bh
);
1312 ll_rw_block(READ
, 1, bh
);
1313 wait_on_buffer(*bh
);
1317 static int ocfs2_mount_volume(struct super_block
*sb
)
1320 int unlock_super
= 0;
1321 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
1325 if (ocfs2_is_hard_readonly(osb
))
1328 status
= ocfs2_dlm_init(osb
);
1334 status
= ocfs2_super_lock(osb
, 1);
1341 /* This will load up the node map and add ourselves to it. */
1342 status
= ocfs2_find_slot(osb
);
1348 /* load all node-local system inodes */
1349 status
= ocfs2_init_local_system_inodes(osb
);
1355 status
= ocfs2_check_volume(osb
);
1361 status
= ocfs2_truncate_log_init(osb
);
1367 if (ocfs2_mount_local(osb
))
1372 ocfs2_super_unlock(osb
, 1);
1378 static void ocfs2_dismount_volume(struct super_block
*sb
, int mnt_err
)
1380 int tmp
, hangup_needed
= 0;
1381 struct ocfs2_super
*osb
= NULL
;
1384 mlog_entry("(0x%p)\n", sb
);
1390 ocfs2_shutdown_local_alloc(osb
);
1392 ocfs2_truncate_log_shutdown(osb
);
1394 /* This will disable recovery and flush any recovery work. */
1395 ocfs2_recovery_exit(osb
);
1397 ocfs2_journal_shutdown(osb
);
1399 ocfs2_sync_blockdev(sb
);
1401 /* No cluster connection means we've failed during mount, so skip
1402 * all the steps which depended on that to complete. */
1404 tmp
= ocfs2_super_lock(osb
, 1);
1411 if (osb
->slot_num
!= OCFS2_INVALID_SLOT
)
1412 ocfs2_put_slot(osb
);
1415 ocfs2_super_unlock(osb
, 1);
1417 ocfs2_release_system_inodes(osb
);
1420 * If we're dismounting due to mount error, mount.ocfs2 will clean
1421 * up heartbeat. If we're a local mount, there is no heartbeat.
1422 * If we failed before we got a uuid_str yet, we can't stop
1423 * heartbeat. Otherwise, do it.
1425 if (!mnt_err
&& !ocfs2_mount_local(osb
) && osb
->uuid_str
)
1429 ocfs2_dlm_shutdown(osb
, hangup_needed
);
1431 debugfs_remove(osb
->osb_debug_root
);
1434 ocfs2_cluster_hangup(osb
->uuid_str
, strlen(osb
->uuid_str
));
1436 atomic_set(&osb
->vol_state
, VOLUME_DISMOUNTED
);
1438 if (ocfs2_mount_local(osb
))
1439 snprintf(nodestr
, sizeof(nodestr
), "local");
1441 snprintf(nodestr
, sizeof(nodestr
), "%u", osb
->node_num
);
1443 printk(KERN_INFO
"ocfs2: Unmounting device (%s) on (node %s)\n",
1444 osb
->dev_str
, nodestr
);
1446 ocfs2_delete_osb(osb
);
1449 sb
->s_fs_info
= NULL
;
1452 static int ocfs2_setup_osb_uuid(struct ocfs2_super
*osb
, const unsigned char *uuid
,
1453 unsigned uuid_bytes
)
1458 BUG_ON(uuid_bytes
!= OCFS2_VOL_UUID_LEN
);
1460 osb
->uuid_str
= kzalloc(OCFS2_VOL_UUID_LEN
* 2 + 1, GFP_KERNEL
);
1461 if (osb
->uuid_str
== NULL
)
1464 for (i
= 0, ptr
= osb
->uuid_str
; i
< OCFS2_VOL_UUID_LEN
; i
++) {
1465 /* print with null */
1466 ret
= snprintf(ptr
, 3, "%02X", uuid
[i
]);
1467 if (ret
!= 2) /* drop super cleans up */
1469 /* then only advance past the last char */
1476 static int ocfs2_initialize_super(struct super_block
*sb
,
1477 struct buffer_head
*bh
,
1481 int i
, cbits
, bbits
;
1482 struct ocfs2_dinode
*di
= (struct ocfs2_dinode
*)bh
->b_data
;
1483 struct inode
*inode
= NULL
;
1484 struct ocfs2_journal
*journal
;
1485 __le32 uuid_net_key
;
1486 struct ocfs2_super
*osb
;
1490 osb
= kzalloc(sizeof(struct ocfs2_super
), GFP_KERNEL
);
1497 sb
->s_fs_info
= osb
;
1498 sb
->s_op
= &ocfs2_sops
;
1499 sb
->s_export_op
= &ocfs2_export_ops
;
1500 sb
->s_xattr
= ocfs2_xattr_handlers
;
1501 sb
->s_time_gran
= 1;
1502 sb
->s_flags
|= MS_NOATIME
;
1503 /* this is needed to support O_LARGEFILE */
1504 cbits
= le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
);
1505 bbits
= le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
);
1506 sb
->s_maxbytes
= ocfs2_max_file_offset(bbits
, cbits
);
1509 /* Save off for ocfs2_rw_direct */
1510 osb
->s_sectsize_bits
= blksize_bits(sector_size
);
1511 BUG_ON(!osb
->s_sectsize_bits
);
1513 spin_lock_init(&osb
->dc_task_lock
);
1514 init_waitqueue_head(&osb
->dc_event
);
1515 osb
->dc_work_sequence
= 0;
1516 osb
->dc_wake_sequence
= 0;
1517 INIT_LIST_HEAD(&osb
->blocked_lock_list
);
1518 osb
->blocked_lock_count
= 0;
1519 spin_lock_init(&osb
->osb_lock
);
1520 ocfs2_init_inode_steal_slot(osb
);
1522 atomic_set(&osb
->alloc_stats
.moves
, 0);
1523 atomic_set(&osb
->alloc_stats
.local_data
, 0);
1524 atomic_set(&osb
->alloc_stats
.bitmap_data
, 0);
1525 atomic_set(&osb
->alloc_stats
.bg_allocs
, 0);
1526 atomic_set(&osb
->alloc_stats
.bg_extends
, 0);
1528 ocfs2_init_node_maps(osb
);
1530 snprintf(osb
->dev_str
, sizeof(osb
->dev_str
), "%u,%u",
1531 MAJOR(osb
->sb
->s_dev
), MINOR(osb
->sb
->s_dev
));
1533 status
= ocfs2_recovery_init(osb
);
1535 mlog(ML_ERROR
, "Unable to initialize recovery state\n");
1540 init_waitqueue_head(&osb
->checkpoint_event
);
1541 atomic_set(&osb
->needs_checkpoint
, 0);
1543 osb
->s_atime_quantum
= OCFS2_DEFAULT_ATIME_QUANTUM
;
1545 osb
->slot_num
= OCFS2_INVALID_SLOT
;
1547 osb
->s_xattr_inline_size
= le16_to_cpu(
1548 di
->id2
.i_super
.s_xattr_inline_size
);
1550 osb
->local_alloc_state
= OCFS2_LA_UNUSED
;
1551 osb
->local_alloc_bh
= NULL
;
1552 INIT_DELAYED_WORK(&osb
->la_enable_wq
, ocfs2_la_enable_worker
);
1554 init_waitqueue_head(&osb
->osb_mount_event
);
1556 osb
->vol_label
= kmalloc(OCFS2_MAX_VOL_LABEL_LEN
, GFP_KERNEL
);
1557 if (!osb
->vol_label
) {
1558 mlog(ML_ERROR
, "unable to alloc vol label\n");
1563 osb
->max_slots
= le16_to_cpu(di
->id2
.i_super
.s_max_slots
);
1564 if (osb
->max_slots
> OCFS2_MAX_SLOTS
|| osb
->max_slots
== 0) {
1565 mlog(ML_ERROR
, "Invalid number of node slots (%u)\n",
1570 mlog(0, "max_slots for this device: %u\n", osb
->max_slots
);
1572 osb
->slot_recovery_generations
=
1573 kcalloc(osb
->max_slots
, sizeof(*osb
->slot_recovery_generations
),
1575 if (!osb
->slot_recovery_generations
) {
1581 init_waitqueue_head(&osb
->osb_wipe_event
);
1582 osb
->osb_orphan_wipes
= kcalloc(osb
->max_slots
,
1583 sizeof(*osb
->osb_orphan_wipes
),
1585 if (!osb
->osb_orphan_wipes
) {
1591 osb
->s_feature_compat
=
1592 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_compat
);
1593 osb
->s_feature_ro_compat
=
1594 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_ro_compat
);
1595 osb
->s_feature_incompat
=
1596 le32_to_cpu(OCFS2_RAW_SB(di
)->s_feature_incompat
);
1598 if ((i
= OCFS2_HAS_INCOMPAT_FEATURE(osb
->sb
, ~OCFS2_FEATURE_INCOMPAT_SUPP
))) {
1599 mlog(ML_ERROR
, "couldn't mount because of unsupported "
1600 "optional features (%x).\n", i
);
1604 if (!(osb
->sb
->s_flags
& MS_RDONLY
) &&
1605 (i
= OCFS2_HAS_RO_COMPAT_FEATURE(osb
->sb
, ~OCFS2_FEATURE_RO_COMPAT_SUPP
))) {
1606 mlog(ML_ERROR
, "couldn't mount RDWR because of "
1607 "unsupported optional features (%x).\n", i
);
1612 if (ocfs2_userspace_stack(osb
)) {
1613 memcpy(osb
->osb_cluster_stack
,
1614 OCFS2_RAW_SB(di
)->s_cluster_info
.ci_stack
,
1615 OCFS2_STACK_LABEL_LEN
);
1616 osb
->osb_cluster_stack
[OCFS2_STACK_LABEL_LEN
] = '\0';
1617 if (strlen(osb
->osb_cluster_stack
) != OCFS2_STACK_LABEL_LEN
) {
1619 "couldn't mount because of an invalid "
1620 "cluster stack label (%s) \n",
1621 osb
->osb_cluster_stack
);
1626 /* The empty string is identical with classic tools that
1627 * don't know about s_cluster_info. */
1628 osb
->osb_cluster_stack
[0] = '\0';
1631 get_random_bytes(&osb
->s_next_generation
, sizeof(u32
));
1634 * This should be done in ocfs2_journal_init(), but unknown
1635 * ordering issues will cause the filesystem to crash.
1636 * If anyone wants to figure out what part of the code
1637 * refers to osb->journal before ocfs2_journal_init() is run,
1640 /* initialize our journal structure */
1642 journal
= kzalloc(sizeof(struct ocfs2_journal
), GFP_KERNEL
);
1644 mlog(ML_ERROR
, "unable to alloc journal\n");
1648 osb
->journal
= journal
;
1649 journal
->j_osb
= osb
;
1651 atomic_set(&journal
->j_num_trans
, 0);
1652 init_rwsem(&journal
->j_trans_barrier
);
1653 init_waitqueue_head(&journal
->j_checkpointed
);
1654 spin_lock_init(&journal
->j_lock
);
1655 journal
->j_trans_id
= (unsigned long) 1;
1656 INIT_LIST_HEAD(&journal
->j_la_cleanups
);
1657 INIT_WORK(&journal
->j_recovery_work
, ocfs2_complete_recovery
);
1658 journal
->j_state
= OCFS2_JOURNAL_FREE
;
1660 /* get some pseudo constants for clustersize bits */
1661 osb
->s_clustersize_bits
=
1662 le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
);
1663 osb
->s_clustersize
= 1 << osb
->s_clustersize_bits
;
1664 mlog(0, "clusterbits=%d\n", osb
->s_clustersize_bits
);
1666 if (osb
->s_clustersize
< OCFS2_MIN_CLUSTERSIZE
||
1667 osb
->s_clustersize
> OCFS2_MAX_CLUSTERSIZE
) {
1668 mlog(ML_ERROR
, "Volume has invalid cluster size (%d)\n",
1669 osb
->s_clustersize
);
1674 if (ocfs2_clusters_to_blocks(osb
->sb
, le32_to_cpu(di
->i_clusters
) - 1)
1676 mlog(ML_ERROR
, "Volume might try to write to blocks beyond "
1677 "what jbd can address in 32 bits.\n");
1682 if (ocfs2_setup_osb_uuid(osb
, di
->id2
.i_super
.s_uuid
,
1683 sizeof(di
->id2
.i_super
.s_uuid
))) {
1684 mlog(ML_ERROR
, "Out of memory trying to setup our uuid.\n");
1689 memcpy(&uuid_net_key
, di
->id2
.i_super
.s_uuid
, sizeof(uuid_net_key
));
1691 strncpy(osb
->vol_label
, di
->id2
.i_super
.s_label
, 63);
1692 osb
->vol_label
[63] = '\0';
1693 osb
->root_blkno
= le64_to_cpu(di
->id2
.i_super
.s_root_blkno
);
1694 osb
->system_dir_blkno
= le64_to_cpu(di
->id2
.i_super
.s_system_dir_blkno
);
1695 osb
->first_cluster_group_blkno
=
1696 le64_to_cpu(di
->id2
.i_super
.s_first_cluster_group
);
1697 osb
->fs_generation
= le32_to_cpu(di
->i_fs_generation
);
1698 osb
->uuid_hash
= le32_to_cpu(di
->id2
.i_super
.s_uuid_hash
);
1699 mlog(0, "vol_label: %s\n", osb
->vol_label
);
1700 mlog(0, "uuid: %s\n", osb
->uuid_str
);
1701 mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
1702 (unsigned long long)osb
->root_blkno
,
1703 (unsigned long long)osb
->system_dir_blkno
);
1705 osb
->osb_dlm_debug
= ocfs2_new_dlm_debug();
1706 if (!osb
->osb_dlm_debug
) {
1712 atomic_set(&osb
->vol_state
, VOLUME_INIT
);
1714 /* load root, system_dir, and all global system inodes */
1715 status
= ocfs2_init_global_system_inodes(osb
);
1724 inode
= ocfs2_get_system_file_inode(osb
, GLOBAL_BITMAP_SYSTEM_INODE
,
1725 OCFS2_INVALID_SLOT
);
1732 osb
->bitmap_blkno
= OCFS2_I(inode
)->ip_blkno
;
1735 osb
->bitmap_cpg
= ocfs2_group_bitmap_size(sb
) * 8;
1737 status
= ocfs2_init_slot_info(osb
);
1749 * will return: -EAGAIN if it is ok to keep searching for superblocks
1750 * -EINVAL if there is a bad superblock
1753 static int ocfs2_verify_volume(struct ocfs2_dinode
*di
,
1754 struct buffer_head
*bh
,
1757 int status
= -EAGAIN
;
1761 if (memcmp(di
->i_signature
, OCFS2_SUPER_BLOCK_SIGNATURE
,
1762 strlen(OCFS2_SUPER_BLOCK_SIGNATURE
)) == 0) {
1764 if ((1 << le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
)) != blksz
) {
1765 mlog(ML_ERROR
, "found superblock with incorrect block "
1766 "size: found %u, should be %u\n",
1767 1 << le32_to_cpu(di
->id2
.i_super
.s_blocksize_bits
),
1769 } else if (le16_to_cpu(di
->id2
.i_super
.s_major_rev_level
) !=
1770 OCFS2_MAJOR_REV_LEVEL
||
1771 le16_to_cpu(di
->id2
.i_super
.s_minor_rev_level
) !=
1772 OCFS2_MINOR_REV_LEVEL
) {
1773 mlog(ML_ERROR
, "found superblock with bad version: "
1774 "found %u.%u, should be %u.%u\n",
1775 le16_to_cpu(di
->id2
.i_super
.s_major_rev_level
),
1776 le16_to_cpu(di
->id2
.i_super
.s_minor_rev_level
),
1777 OCFS2_MAJOR_REV_LEVEL
,
1778 OCFS2_MINOR_REV_LEVEL
);
1779 } else if (bh
->b_blocknr
!= le64_to_cpu(di
->i_blkno
)) {
1780 mlog(ML_ERROR
, "bad block number on superblock: "
1781 "found %llu, should be %llu\n",
1782 (unsigned long long)le64_to_cpu(di
->i_blkno
),
1783 (unsigned long long)bh
->b_blocknr
);
1784 } else if (le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
) < 12 ||
1785 le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
) > 20) {
1786 mlog(ML_ERROR
, "bad cluster size found: %u\n",
1787 1 << le32_to_cpu(di
->id2
.i_super
.s_clustersize_bits
));
1788 } else if (!le64_to_cpu(di
->id2
.i_super
.s_root_blkno
)) {
1789 mlog(ML_ERROR
, "bad root_blkno: 0\n");
1790 } else if (!le64_to_cpu(di
->id2
.i_super
.s_system_dir_blkno
)) {
1791 mlog(ML_ERROR
, "bad system_dir_blkno: 0\n");
1792 } else if (le16_to_cpu(di
->id2
.i_super
.s_max_slots
) > OCFS2_MAX_SLOTS
) {
1794 "Superblock slots found greater than file system "
1795 "maximum: found %u, max %u\n",
1796 le16_to_cpu(di
->id2
.i_super
.s_max_slots
),
1808 static int ocfs2_check_volume(struct ocfs2_super
*osb
)
1813 struct ocfs2_dinode
*local_alloc
= NULL
; /* only used if we
1819 /* Init our journal object. */
1820 status
= ocfs2_journal_init(osb
->journal
, &dirty
);
1822 mlog(ML_ERROR
, "Could not initialize journal!\n");
1826 /* If the journal was unmounted cleanly then we don't want to
1827 * recover anything. Otherwise, journal_load will do that
1828 * dirty work for us :) */
1830 status
= ocfs2_journal_wipe(osb
->journal
, 0);
1836 mlog(ML_NOTICE
, "File system was not unmounted cleanly, "
1837 "recovering volume.\n");
1840 local
= ocfs2_mount_local(osb
);
1842 /* will play back anything left in the journal. */
1843 status
= ocfs2_journal_load(osb
->journal
, local
, dirty
);
1845 mlog(ML_ERROR
, "ocfs2 journal load failed! %d\n", status
);
1850 /* recover my local alloc if we didn't unmount cleanly. */
1851 status
= ocfs2_begin_local_alloc_recovery(osb
,
1858 /* we complete the recovery process after we've marked
1859 * ourselves as mounted. */
1862 mlog(0, "Journal loaded.\n");
1864 status
= ocfs2_load_local_alloc(osb
);
1871 /* Recovery will be completed after we've mounted the
1872 * rest of the volume. */
1874 osb
->local_alloc_copy
= local_alloc
;
1878 /* go through each journal, trylock it and if you get the
1879 * lock, and it's marked as dirty, set the bit in the recover
1880 * map and launch a recovery thread for it. */
1881 status
= ocfs2_mark_dead_nodes(osb
);
1894 * The routine gets called from dismount or close whenever a dismount on
1895 * volume is requested and the osb open count becomes 1.
1896 * It will remove the osb from the global list and also free up all the
1897 * initialized resources and fileobject.
1899 static void ocfs2_delete_osb(struct ocfs2_super
*osb
)
1903 /* This function assumes that the caller has the main osb resource */
1905 ocfs2_free_slot_info(osb
);
1907 kfree(osb
->osb_orphan_wipes
);
1908 kfree(osb
->slot_recovery_generations
);
1910 * This belongs in journal shutdown, but because we have to
1911 * allocate osb->journal at the start of ocfs2_initalize_osb(),
1914 kfree(osb
->journal
);
1915 if (osb
->local_alloc_copy
)
1916 kfree(osb
->local_alloc_copy
);
1917 kfree(osb
->uuid_str
);
1918 ocfs2_put_dlm_debug(osb
->osb_dlm_debug
);
1919 memset(osb
, 0, sizeof(struct ocfs2_super
));
1924 /* Put OCFS2 into a readonly state, or (if the user specifies it),
1925 * panic(). We do not support continue-on-error operation. */
1926 static void ocfs2_handle_error(struct super_block
*sb
)
1928 struct ocfs2_super
*osb
= OCFS2_SB(sb
);
1930 if (osb
->s_mount_opt
& OCFS2_MOUNT_ERRORS_PANIC
)
1931 panic("OCFS2: (device %s): panic forced after error\n",
1934 ocfs2_set_osb_flag(osb
, OCFS2_OSB_ERROR_FS
);
1936 if (sb
->s_flags
& MS_RDONLY
&&
1937 (ocfs2_is_soft_readonly(osb
) ||
1938 ocfs2_is_hard_readonly(osb
)))
1941 printk(KERN_CRIT
"File system is now read-only due to the potential "
1942 "of on-disk corruption. Please run fsck.ocfs2 once the file "
1943 "system is unmounted.\n");
1944 sb
->s_flags
|= MS_RDONLY
;
1945 ocfs2_set_ro_flag(osb
, 0);
1948 static char error_buf
[1024];
1950 void __ocfs2_error(struct super_block
*sb
,
1951 const char *function
,
1952 const char *fmt
, ...)
1956 va_start(args
, fmt
);
1957 vsnprintf(error_buf
, sizeof(error_buf
), fmt
, args
);
1960 /* Not using mlog here because we want to show the actual
1961 * function the error came from. */
1962 printk(KERN_CRIT
"OCFS2: ERROR (device %s): %s: %s\n",
1963 sb
->s_id
, function
, error_buf
);
1965 ocfs2_handle_error(sb
);
1968 /* Handle critical errors. This is intentionally more drastic than
1969 * ocfs2_handle_error, so we only use for things like journal errors,
1971 void __ocfs2_abort(struct super_block
* sb
,
1972 const char *function
,
1973 const char *fmt
, ...)
1977 va_start(args
, fmt
);
1978 vsnprintf(error_buf
, sizeof(error_buf
), fmt
, args
);
1981 printk(KERN_CRIT
"OCFS2: abort (device %s): %s: %s\n",
1982 sb
->s_id
, function
, error_buf
);
1984 /* We don't have the cluster support yet to go straight to
1985 * hard readonly in here. Until then, we want to keep
1986 * ocfs2_abort() so that we can at least mark critical
1989 * TODO: This should abort the journal and alert other nodes
1990 * that our slot needs recovery. */
1992 /* Force a panic(). This stinks, but it's better than letting
1993 * things continue without having a proper hard readonly
1995 OCFS2_SB(sb
)->s_mount_opt
|= OCFS2_MOUNT_ERRORS_PANIC
;
1996 ocfs2_handle_error(sb
);
1999 module_init(ocfs2_init
);
2000 module_exit(ocfs2_exit
);