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,
164 static const match_table_t tokens = {
165 {Opt_barrier, "barrier=%u"},
166 {Opt_err_panic, "errors=panic"},
167 {Opt_err_ro, "errors=remount-ro"},
169 {Opt_nointr, "nointr"},
170 {Opt_hb_none, OCFS2_HB_NONE},
171 {Opt_hb_local, OCFS2_HB_LOCAL},
172 {Opt_data_ordered, "data=ordered"},
173 {Opt_data_writeback, "data=writeback"},
174 {Opt_atime_quantum, "atime_quantum=%u"},
175 {Opt_slot, "preferred_slot=%u"},
176 {Opt_commit, "commit=%u"},
177 {Opt_localalloc, "localalloc=%d"},
178 {Opt_localflocks, "localflocks"},
179 {Opt_stack, "cluster_stack=%s"},
180 {Opt_user_xattr, "user_xattr"},
181 {Opt_nouser_xattr, "nouser_xattr"},
182 {Opt_inode64, "inode64"},
187 * write_super and sync_fs ripped right out of ext3.
189 static void ocfs2_write_super(struct super_block *sb)
191 if (mutex_trylock(&sb->s_lock) != 0)
196 static int ocfs2_sync_fs(struct super_block *sb, int wait)
200 struct ocfs2_super *osb = OCFS2_SB(sb);
204 if (ocfs2_is_hard_readonly(osb))
208 status = ocfs2_flush_truncate_log(osb);
212 ocfs2_schedule_truncate_log_flush(osb, 0);
215 if (journal_start_commit(OCFS2_SB(sb)->journal->j_journal, &target)) {
217 log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
223 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
225 struct inode *new = NULL;
231 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
233 status = PTR_ERR(new);
237 osb->root_inode = new;
239 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
241 status = PTR_ERR(new);
245 osb->sys_root_inode = new;
247 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
248 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
249 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
251 ocfs2_release_system_inodes(osb);
254 /* FIXME: Should ERROR_RO_FS */
255 mlog(ML_ERROR, "Unable to load system inode %d, "
256 "possibly corrupt fs?", i);
259 // the array now has one ref, so drop this one
268 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
270 struct inode *new = NULL;
276 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
277 i < NUM_SYSTEM_INODES;
279 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
281 ocfs2_release_system_inodes(osb);
283 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
284 status, i, osb->slot_num);
287 /* the array now has one ref, so drop this one */
296 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
303 for (i = 0; i < NUM_SYSTEM_INODES; i++) {
304 inode = osb->system_inodes[i];
307 osb->system_inodes[i] = NULL;
311 inode = osb->sys_root_inode;
314 osb->sys_root_inode = NULL;
317 inode = osb->root_inode;
320 osb->root_inode = NULL;
326 /* We're allocating fs objects, use GFP_NOFS */
327 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
329 struct ocfs2_inode_info *oi;
331 oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
335 return &oi->vfs_inode;
338 static void ocfs2_destroy_inode(struct inode *inode)
340 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
343 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
346 unsigned int bytes = 1 << cbits;
347 unsigned int trim = bytes;
348 unsigned int bitshift = 32;
351 * i_size and all block offsets in ocfs2 are always 64 bits
352 * wide. i_clusters is 32 bits, in cluster-sized units. So on
353 * 64 bit platforms, cluster size will be the limiting factor.
356 #if BITS_PER_LONG == 32
357 # if defined(CONFIG_LBD)
358 BUILD_BUG_ON(sizeof(sector_t) != 8);
360 * We might be limited by page cache size.
362 if (bytes > PAGE_CACHE_SIZE) {
363 bytes = PAGE_CACHE_SIZE;
366 * Shift by 31 here so that we don't get larger than
373 * We are limited by the size of sector_t. Use block size, as
374 * that's what we expose to the VFS.
383 * Trim by a whole cluster when we can actually approach the
384 * on-disk limits. Otherwise we can overflow i_clusters when
385 * an extent start is at the max offset.
387 return (((unsigned long long)bytes) << bitshift) - trim;
390 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
392 int incompat_features;
394 struct mount_options parsed_options;
395 struct ocfs2_super *osb = OCFS2_SB(sb);
397 if (!ocfs2_parse_options(sb, data, &parsed_options, 1)) {
402 if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) !=
403 (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
405 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
409 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
410 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
412 mlog(ML_ERROR, "Cannot change data mode on remount\n");
416 /* Probably don't want this on remount; it might
417 * mess with other nodes */
418 if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
419 (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
421 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
425 /* We're going to/from readonly mode. */
426 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
427 /* Lock here so the check of HARD_RO and the potential
428 * setting of SOFT_RO is atomic. */
429 spin_lock(&osb->osb_lock);
430 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
431 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
436 if (*flags & MS_RDONLY) {
437 mlog(0, "Going to ro mode.\n");
438 sb->s_flags |= MS_RDONLY;
439 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
441 mlog(0, "Making ro filesystem writeable.\n");
443 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
444 mlog(ML_ERROR, "Cannot remount RDWR "
445 "filesystem due to previous errors.\n");
449 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
450 if (incompat_features) {
451 mlog(ML_ERROR, "Cannot remount RDWR because "
452 "of unsupported optional features "
453 "(%x).\n", incompat_features);
457 sb->s_flags &= ~MS_RDONLY;
458 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
461 spin_unlock(&osb->osb_lock);
465 /* Only save off the new mount options in case of a successful
467 osb->s_mount_opt = parsed_options.mount_opt;
468 osb->s_atime_quantum = parsed_options.atime_quantum;
469 osb->preferred_slot = parsed_options.slot;
470 if (parsed_options.commit_interval)
471 osb->osb_commit_interval = parsed_options.commit_interval;
473 if (!ocfs2_is_hard_readonly(osb))
474 ocfs2_set_journal_params(osb);
480 static int ocfs2_sb_probe(struct super_block *sb,
481 struct buffer_head **bh,
485 struct ocfs1_vol_disk_hdr *hdr;
486 struct ocfs2_dinode *di;
492 *sector_size = bdev_hardsect_size(sb->s_bdev);
493 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
494 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
495 *sector_size, OCFS2_MAX_BLOCKSIZE);
500 /* Can this really happen? */
501 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
502 *sector_size = OCFS2_MIN_BLOCKSIZE;
504 /* check block zero for old format */
505 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
510 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
511 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
512 mlog(ML_ERROR, "incompatible version: %u.%u\n",
513 hdr->major_version, hdr->minor_version);
516 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
517 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
518 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
525 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
526 "upgraded before mounting with ocfs v2\n");
531 * Now check at magic offset for 512, 1024, 2048, 4096
532 * blocksizes. 4096 is the maximum blocksize because it is
533 * the minimum clustersize.
536 for (blksize = *sector_size;
537 blksize <= OCFS2_MAX_BLOCKSIZE;
539 tmpstat = ocfs2_get_sector(sb, bh,
540 OCFS2_SUPER_BLOCK_BLKNO,
547 di = (struct ocfs2_dinode *) (*bh)->b_data;
548 status = ocfs2_verify_volume(di, *bh, blksize);
553 if (status != -EAGAIN)
561 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
563 if (ocfs2_mount_local(osb)) {
564 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
565 mlog(ML_ERROR, "Cannot heartbeat on a locally "
566 "mounted device.\n");
571 if (ocfs2_userspace_stack(osb)) {
572 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
573 mlog(ML_ERROR, "Userspace stack expected, but "
574 "o2cb heartbeat arguments passed to mount\n");
579 if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
580 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
581 !ocfs2_userspace_stack(osb)) {
582 mlog(ML_ERROR, "Heartbeat has to be started to mount "
583 "a read-write clustered device.\n");
592 * If we're using a userspace stack, mount should have passed
593 * a name that matches the disk. If not, mount should not
594 * have passed a stack.
596 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
597 struct mount_options *mopt)
599 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
601 "cluster stack passed to mount, but this filesystem "
602 "does not support it\n");
606 if (ocfs2_userspace_stack(osb) &&
607 strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
608 OCFS2_STACK_LABEL_LEN)) {
610 "cluster stack passed to mount (\"%s\") does not "
611 "match the filesystem (\"%s\")\n",
613 osb->osb_cluster_stack);
620 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
623 int status, sector_size;
624 struct mount_options parsed_options;
625 struct inode *inode = NULL;
626 struct ocfs2_super *osb = NULL;
627 struct buffer_head *bh = NULL;
630 mlog_entry("%p, %p, %i", sb, data, silent);
632 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
634 goto read_super_error;
637 /* probe for superblock */
638 status = ocfs2_sb_probe(sb, &bh, §or_size);
640 mlog(ML_ERROR, "superblock probe failed!\n");
641 goto read_super_error;
644 status = ocfs2_initialize_super(sb, bh, sector_size);
648 goto read_super_error;
652 osb->s_mount_opt = parsed_options.mount_opt;
653 osb->s_atime_quantum = parsed_options.atime_quantum;
654 osb->preferred_slot = parsed_options.slot;
655 osb->osb_commit_interval = parsed_options.commit_interval;
656 osb->local_alloc_default_bits = ocfs2_megabytes_to_clusters(sb, parsed_options.localalloc_opt);
657 osb->local_alloc_bits = osb->local_alloc_default_bits;
659 status = ocfs2_verify_userspace_stack(osb, &parsed_options);
661 goto read_super_error;
663 sb->s_magic = OCFS2_SUPER_MAGIC;
665 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
667 if (bdev_read_only(sb->s_bdev)) {
668 if (!(sb->s_flags & MS_RDONLY)) {
670 mlog(ML_ERROR, "Readonly device detected but readonly "
671 "mount was not specified.\n");
672 goto read_super_error;
675 /* You should not be able to start a local heartbeat
676 * on a readonly device. */
677 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
679 mlog(ML_ERROR, "Local heartbeat specified on readonly "
681 goto read_super_error;
684 status = ocfs2_check_journals_nolocks(osb);
686 if (status == -EROFS)
687 mlog(ML_ERROR, "Recovery required on readonly "
688 "file system, but write access is "
692 goto read_super_error;
695 ocfs2_set_ro_flag(osb, 1);
697 printk(KERN_NOTICE "Readonly device detected. No cluster "
698 "services will be utilized for this mount. Recovery "
699 "will be skipped.\n");
702 if (!ocfs2_is_hard_readonly(osb)) {
703 if (sb->s_flags & MS_RDONLY)
704 ocfs2_set_ro_flag(osb, 0);
707 status = ocfs2_verify_heartbeat(osb);
710 goto read_super_error;
713 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
715 if (!osb->osb_debug_root) {
717 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
718 goto read_super_error;
721 status = ocfs2_mount_volume(sb);
723 inode = igrab(osb->root_inode);
726 goto read_super_error;
731 goto read_super_error;
734 root = d_alloc_root(inode);
738 goto read_super_error;
743 ocfs2_complete_mount_recovery(osb);
745 if (ocfs2_mount_local(osb))
746 snprintf(nodestr, sizeof(nodestr), "local");
748 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
750 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
751 "with %s data mode.\n",
752 osb->dev_str, nodestr, osb->slot_num,
753 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
756 atomic_set(&osb->vol_state, VOLUME_MOUNTED);
757 wake_up(&osb->osb_mount_event);
770 atomic_set(&osb->vol_state, VOLUME_DISABLED);
771 wake_up(&osb->osb_mount_event);
772 ocfs2_dismount_volume(sb, 1);
779 static int ocfs2_get_sb(struct file_system_type *fs_type,
781 const char *dev_name,
783 struct vfsmount *mnt)
785 return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super,
789 static struct file_system_type ocfs2_fs_type = {
790 .owner = THIS_MODULE,
792 .get_sb = ocfs2_get_sb, /* is this called when we mount
794 .kill_sb = kill_block_super, /* set to the generic one
795 * right now, but do we
796 * need to change that? */
797 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
801 static int ocfs2_parse_options(struct super_block *sb,
803 struct mount_options *mopt,
809 mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
810 options ? options : "(none)");
812 mopt->commit_interval = 0;
814 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
815 mopt->slot = OCFS2_INVALID_SLOT;
816 mopt->localalloc_opt = OCFS2_DEFAULT_LOCAL_ALLOC_SIZE;
817 mopt->cluster_stack[0] = '\0';
824 while ((p = strsep(&options, ",")) != NULL) {
826 substring_t args[MAX_OPT_ARGS];
831 token = match_token(p, tokens, args);
834 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
837 mopt->mount_opt &= ~OCFS2_MOUNT_HB_LOCAL;
840 if (match_int(&args[0], &option)) {
845 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
847 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
850 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
853 mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
856 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
859 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
861 case Opt_data_ordered:
862 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
864 case Opt_data_writeback:
865 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
868 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
870 case Opt_nouser_xattr:
871 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
873 case Opt_atime_quantum:
874 if (match_int(&args[0], &option)) {
879 mopt->atime_quantum = option;
883 if (match_int(&args[0], &option)) {
888 mopt->slot = (s16)option;
892 if (match_int(&args[0], &option)) {
899 option = JBD_DEFAULT_MAX_COMMIT_AGE;
900 mopt->commit_interval = HZ * option;
904 if (match_int(&args[0], &option)) {
908 if (option >= 0 && (option <= ocfs2_local_alloc_size(sb) * 8))
909 mopt->localalloc_opt = option;
911 case Opt_localflocks:
913 * Changing this during remount could race
914 * flock() requests, or "unbalance" existing
915 * ones (e.g., a lock is taken in one mode but
916 * dropped in the other). If users care enough
917 * to flip locking modes during remount, we
918 * could add a "local" flag to individual
919 * flock structures for proper tracking of
923 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
926 /* Check both that the option we were passed
927 * is of the right length and that it is a proper
928 * string of the right length.
930 if (((args[0].to - args[0].from) !=
931 OCFS2_STACK_LABEL_LEN) ||
932 (strnlen(args[0].from,
933 OCFS2_STACK_LABEL_LEN) !=
934 OCFS2_STACK_LABEL_LEN)) {
936 "Invalid cluster_stack option\n");
940 memcpy(mopt->cluster_stack, args[0].from,
941 OCFS2_STACK_LABEL_LEN);
942 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
945 mopt->mount_opt |= OCFS2_MOUNT_INODE64;
949 "Unrecognized mount option \"%s\" "
950 "or missing value\n", p);
963 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
965 struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb);
966 unsigned long opts = osb->s_mount_opt;
967 unsigned int local_alloc_megs;
969 if (opts & OCFS2_MOUNT_HB_LOCAL)
970 seq_printf(s, ",_netdev,heartbeat=local");
972 seq_printf(s, ",heartbeat=none");
974 if (opts & OCFS2_MOUNT_NOINTR)
975 seq_printf(s, ",nointr");
977 if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
978 seq_printf(s, ",data=writeback");
980 seq_printf(s, ",data=ordered");
982 if (opts & OCFS2_MOUNT_BARRIER)
983 seq_printf(s, ",barrier=1");
985 if (opts & OCFS2_MOUNT_ERRORS_PANIC)
986 seq_printf(s, ",errors=panic");
988 seq_printf(s, ",errors=remount-ro");
990 if (osb->preferred_slot != OCFS2_INVALID_SLOT)
991 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
993 if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM)
994 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
996 if (osb->osb_commit_interval)
997 seq_printf(s, ",commit=%u",
998 (unsigned) (osb->osb_commit_interval / HZ));
1000 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1001 if (local_alloc_megs != OCFS2_DEFAULT_LOCAL_ALLOC_SIZE)
1002 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1004 if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1005 seq_printf(s, ",localflocks,");
1007 if (osb->osb_cluster_stack[0])
1008 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1009 osb->osb_cluster_stack);
1011 if (opts & OCFS2_MOUNT_INODE64)
1012 seq_printf(s, ",inode64");
1017 static int __init ocfs2_init(void)
1023 ocfs2_print_version();
1025 status = init_ocfs2_uptodate_cache();
1031 status = ocfs2_initialize_mem_caches();
1037 ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1043 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1044 if (!ocfs2_debugfs_root) {
1046 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1049 ocfs2_set_locking_protocol();
1053 ocfs2_free_mem_caches();
1054 exit_ocfs2_uptodate_cache();
1060 return register_filesystem(&ocfs2_fs_type);
1065 static void __exit ocfs2_exit(void)
1070 flush_workqueue(ocfs2_wq);
1071 destroy_workqueue(ocfs2_wq);
1074 debugfs_remove(ocfs2_debugfs_root);
1076 ocfs2_free_mem_caches();
1078 unregister_filesystem(&ocfs2_fs_type);
1080 exit_ocfs2_uptodate_cache();
1085 static void ocfs2_put_super(struct super_block *sb)
1087 mlog_entry("(0x%p)\n", sb);
1089 ocfs2_sync_blockdev(sb);
1090 ocfs2_dismount_volume(sb, 0);
1095 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1097 struct ocfs2_super *osb;
1098 u32 numbits, freebits;
1100 struct ocfs2_dinode *bm_lock;
1101 struct buffer_head *bh = NULL;
1102 struct inode *inode = NULL;
1104 mlog_entry("(%p, %p)\n", dentry->d_sb, buf);
1106 osb = OCFS2_SB(dentry->d_sb);
1108 inode = ocfs2_get_system_file_inode(osb,
1109 GLOBAL_BITMAP_SYSTEM_INODE,
1110 OCFS2_INVALID_SLOT);
1112 mlog(ML_ERROR, "failed to get bitmap inode\n");
1117 status = ocfs2_inode_lock(inode, &bh, 0);
1123 bm_lock = (struct ocfs2_dinode *) bh->b_data;
1125 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1126 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1128 buf->f_type = OCFS2_SUPER_MAGIC;
1129 buf->f_bsize = dentry->d_sb->s_blocksize;
1130 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1131 buf->f_blocks = ((sector_t) numbits) *
1132 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1133 buf->f_bfree = ((sector_t) freebits) *
1134 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1135 buf->f_bavail = buf->f_bfree;
1136 buf->f_files = numbits;
1137 buf->f_ffree = freebits;
1141 ocfs2_inode_unlock(inode, 0);
1152 static void ocfs2_inode_init_once(void *data)
1154 struct ocfs2_inode_info *oi = data;
1157 oi->ip_open_count = 0;
1158 spin_lock_init(&oi->ip_lock);
1159 ocfs2_extent_map_init(&oi->vfs_inode);
1160 INIT_LIST_HEAD(&oi->ip_io_markers);
1161 oi->ip_created_trans = 0;
1162 oi->ip_last_trans = 0;
1163 oi->ip_dir_start_lookup = 0;
1165 init_rwsem(&oi->ip_alloc_sem);
1166 init_rwsem(&oi->ip_xattr_sem);
1167 mutex_init(&oi->ip_io_mutex);
1169 oi->ip_blkno = 0ULL;
1170 oi->ip_clusters = 0;
1172 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1173 ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1174 ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1176 ocfs2_metadata_cache_init(&oi->vfs_inode);
1178 inode_init_once(&oi->vfs_inode);
1181 static int ocfs2_initialize_mem_caches(void)
1183 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1184 sizeof(struct ocfs2_inode_info),
1186 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1188 ocfs2_inode_init_once);
1189 if (!ocfs2_inode_cachep)
1195 static void ocfs2_free_mem_caches(void)
1197 if (ocfs2_inode_cachep)
1198 kmem_cache_destroy(ocfs2_inode_cachep);
1200 ocfs2_inode_cachep = NULL;
1203 static int ocfs2_get_sector(struct super_block *sb,
1204 struct buffer_head **bh,
1208 if (!sb_set_blocksize(sb, sect_size)) {
1209 mlog(ML_ERROR, "unable to set blocksize\n");
1213 *bh = sb_getblk(sb, block);
1219 if (!buffer_dirty(*bh))
1220 clear_buffer_uptodate(*bh);
1222 ll_rw_block(READ, 1, bh);
1223 wait_on_buffer(*bh);
1227 static int ocfs2_mount_volume(struct super_block *sb)
1230 int unlock_super = 0;
1231 struct ocfs2_super *osb = OCFS2_SB(sb);
1235 if (ocfs2_is_hard_readonly(osb))
1238 status = ocfs2_dlm_init(osb);
1244 status = ocfs2_super_lock(osb, 1);
1251 /* This will load up the node map and add ourselves to it. */
1252 status = ocfs2_find_slot(osb);
1258 /* load all node-local system inodes */
1259 status = ocfs2_init_local_system_inodes(osb);
1265 status = ocfs2_check_volume(osb);
1271 status = ocfs2_truncate_log_init(osb);
1277 if (ocfs2_mount_local(osb))
1282 ocfs2_super_unlock(osb, 1);
1288 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1290 int tmp, hangup_needed = 0;
1291 struct ocfs2_super *osb = NULL;
1294 mlog_entry("(0x%p)\n", sb);
1300 ocfs2_shutdown_local_alloc(osb);
1302 ocfs2_truncate_log_shutdown(osb);
1304 /* This will disable recovery and flush any recovery work. */
1305 ocfs2_recovery_exit(osb);
1307 ocfs2_journal_shutdown(osb);
1309 ocfs2_sync_blockdev(sb);
1311 /* No cluster connection means we've failed during mount, so skip
1312 * all the steps which depended on that to complete. */
1314 tmp = ocfs2_super_lock(osb, 1);
1321 if (osb->slot_num != OCFS2_INVALID_SLOT)
1322 ocfs2_put_slot(osb);
1325 ocfs2_super_unlock(osb, 1);
1327 ocfs2_release_system_inodes(osb);
1330 * If we're dismounting due to mount error, mount.ocfs2 will clean
1331 * up heartbeat. If we're a local mount, there is no heartbeat.
1332 * If we failed before we got a uuid_str yet, we can't stop
1333 * heartbeat. Otherwise, do it.
1335 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str)
1339 ocfs2_dlm_shutdown(osb, hangup_needed);
1341 debugfs_remove(osb->osb_debug_root);
1344 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1346 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1348 if (ocfs2_mount_local(osb))
1349 snprintf(nodestr, sizeof(nodestr), "local");
1351 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1353 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1354 osb->dev_str, nodestr);
1356 ocfs2_delete_osb(osb);
1359 sb->s_fs_info = NULL;
1362 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1363 unsigned uuid_bytes)
1368 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1370 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1371 if (osb->uuid_str == NULL)
1374 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1375 /* print with null */
1376 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1377 if (ret != 2) /* drop super cleans up */
1379 /* then only advance past the last char */
1386 static int ocfs2_initialize_super(struct super_block *sb,
1387 struct buffer_head *bh,
1391 int i, cbits, bbits;
1392 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
1393 struct inode *inode = NULL;
1394 struct ocfs2_journal *journal;
1395 __le32 uuid_net_key;
1396 struct ocfs2_super *osb;
1400 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
1407 sb->s_fs_info = osb;
1408 sb->s_op = &ocfs2_sops;
1409 sb->s_export_op = &ocfs2_export_ops;
1410 sb->s_xattr = ocfs2_xattr_handlers;
1411 sb->s_time_gran = 1;
1412 sb->s_flags |= MS_NOATIME;
1413 /* this is needed to support O_LARGEFILE */
1414 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
1415 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
1416 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
1419 /* Save off for ocfs2_rw_direct */
1420 osb->s_sectsize_bits = blksize_bits(sector_size);
1421 BUG_ON(!osb->s_sectsize_bits);
1423 spin_lock_init(&osb->dc_task_lock);
1424 init_waitqueue_head(&osb->dc_event);
1425 osb->dc_work_sequence = 0;
1426 osb->dc_wake_sequence = 0;
1427 INIT_LIST_HEAD(&osb->blocked_lock_list);
1428 osb->blocked_lock_count = 0;
1429 spin_lock_init(&osb->osb_lock);
1430 ocfs2_init_inode_steal_slot(osb);
1432 atomic_set(&osb->alloc_stats.moves, 0);
1433 atomic_set(&osb->alloc_stats.local_data, 0);
1434 atomic_set(&osb->alloc_stats.bitmap_data, 0);
1435 atomic_set(&osb->alloc_stats.bg_allocs, 0);
1436 atomic_set(&osb->alloc_stats.bg_extends, 0);
1438 ocfs2_init_node_maps(osb);
1440 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
1441 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
1443 status = ocfs2_recovery_init(osb);
1445 mlog(ML_ERROR, "Unable to initialize recovery state\n");
1450 init_waitqueue_head(&osb->checkpoint_event);
1451 atomic_set(&osb->needs_checkpoint, 0);
1453 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1455 osb->slot_num = OCFS2_INVALID_SLOT;
1457 osb->s_xattr_inline_size = le16_to_cpu(
1458 di->id2.i_super.s_xattr_inline_size);
1460 osb->local_alloc_state = OCFS2_LA_UNUSED;
1461 osb->local_alloc_bh = NULL;
1462 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
1464 init_waitqueue_head(&osb->osb_mount_event);
1466 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
1467 if (!osb->vol_label) {
1468 mlog(ML_ERROR, "unable to alloc vol label\n");
1473 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
1474 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
1475 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
1480 mlog(0, "max_slots for this device: %u\n", osb->max_slots);
1482 osb->slot_recovery_generations =
1483 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
1485 if (!osb->slot_recovery_generations) {
1491 init_waitqueue_head(&osb->osb_wipe_event);
1492 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
1493 sizeof(*osb->osb_orphan_wipes),
1495 if (!osb->osb_orphan_wipes) {
1501 osb->s_feature_compat =
1502 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
1503 osb->s_feature_ro_compat =
1504 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
1505 osb->s_feature_incompat =
1506 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
1508 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
1509 mlog(ML_ERROR, "couldn't mount because of unsupported "
1510 "optional features (%x).\n", i);
1514 if (!(osb->sb->s_flags & MS_RDONLY) &&
1515 (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
1516 mlog(ML_ERROR, "couldn't mount RDWR because of "
1517 "unsupported optional features (%x).\n", i);
1522 if (ocfs2_userspace_stack(osb)) {
1523 memcpy(osb->osb_cluster_stack,
1524 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
1525 OCFS2_STACK_LABEL_LEN);
1526 osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1527 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
1529 "couldn't mount because of an invalid "
1530 "cluster stack label (%s) \n",
1531 osb->osb_cluster_stack);
1536 /* The empty string is identical with classic tools that
1537 * don't know about s_cluster_info. */
1538 osb->osb_cluster_stack[0] = '\0';
1541 get_random_bytes(&osb->s_next_generation, sizeof(u32));
1544 * This should be done in ocfs2_journal_init(), but unknown
1545 * ordering issues will cause the filesystem to crash.
1546 * If anyone wants to figure out what part of the code
1547 * refers to osb->journal before ocfs2_journal_init() is run,
1550 /* initialize our journal structure */
1552 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
1554 mlog(ML_ERROR, "unable to alloc journal\n");
1558 osb->journal = journal;
1559 journal->j_osb = osb;
1561 atomic_set(&journal->j_num_trans, 0);
1562 init_rwsem(&journal->j_trans_barrier);
1563 init_waitqueue_head(&journal->j_checkpointed);
1564 spin_lock_init(&journal->j_lock);
1565 journal->j_trans_id = (unsigned long) 1;
1566 INIT_LIST_HEAD(&journal->j_la_cleanups);
1567 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
1568 journal->j_state = OCFS2_JOURNAL_FREE;
1570 /* get some pseudo constants for clustersize bits */
1571 osb->s_clustersize_bits =
1572 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
1573 osb->s_clustersize = 1 << osb->s_clustersize_bits;
1574 mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
1576 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
1577 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
1578 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
1579 osb->s_clustersize);
1584 if (ocfs2_clusters_to_blocks(osb->sb, le32_to_cpu(di->i_clusters) - 1)
1586 mlog(ML_ERROR, "Volume might try to write to blocks beyond "
1587 "what jbd can address in 32 bits.\n");
1592 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
1593 sizeof(di->id2.i_super.s_uuid))) {
1594 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
1599 memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
1601 strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
1602 osb->vol_label[63] = '\0';
1603 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
1604 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
1605 osb->first_cluster_group_blkno =
1606 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
1607 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
1608 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
1609 mlog(0, "vol_label: %s\n", osb->vol_label);
1610 mlog(0, "uuid: %s\n", osb->uuid_str);
1611 mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
1612 (unsigned long long)osb->root_blkno,
1613 (unsigned long long)osb->system_dir_blkno);
1615 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
1616 if (!osb->osb_dlm_debug) {
1622 atomic_set(&osb->vol_state, VOLUME_INIT);
1624 /* load root, system_dir, and all global system inodes */
1625 status = ocfs2_init_global_system_inodes(osb);
1634 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
1635 OCFS2_INVALID_SLOT);
1642 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
1645 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb) * 8;
1647 status = ocfs2_init_slot_info(osb);
1659 * will return: -EAGAIN if it is ok to keep searching for superblocks
1660 * -EINVAL if there is a bad superblock
1663 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
1664 struct buffer_head *bh,
1667 int status = -EAGAIN;
1671 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
1672 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
1674 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
1675 mlog(ML_ERROR, "found superblock with incorrect block "
1676 "size: found %u, should be %u\n",
1677 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
1679 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
1680 OCFS2_MAJOR_REV_LEVEL ||
1681 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
1682 OCFS2_MINOR_REV_LEVEL) {
1683 mlog(ML_ERROR, "found superblock with bad version: "
1684 "found %u.%u, should be %u.%u\n",
1685 le16_to_cpu(di->id2.i_super.s_major_rev_level),
1686 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
1687 OCFS2_MAJOR_REV_LEVEL,
1688 OCFS2_MINOR_REV_LEVEL);
1689 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
1690 mlog(ML_ERROR, "bad block number on superblock: "
1691 "found %llu, should be %llu\n",
1692 (unsigned long long)le64_to_cpu(di->i_blkno),
1693 (unsigned long long)bh->b_blocknr);
1694 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
1695 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
1696 mlog(ML_ERROR, "bad cluster size found: %u\n",
1697 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
1698 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
1699 mlog(ML_ERROR, "bad root_blkno: 0\n");
1700 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
1701 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
1702 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
1704 "Superblock slots found greater than file system "
1705 "maximum: found %u, max %u\n",
1706 le16_to_cpu(di->id2.i_super.s_max_slots),
1718 static int ocfs2_check_volume(struct ocfs2_super *osb)
1723 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
1729 /* Init our journal object. */
1730 status = ocfs2_journal_init(osb->journal, &dirty);
1732 mlog(ML_ERROR, "Could not initialize journal!\n");
1736 /* If the journal was unmounted cleanly then we don't want to
1737 * recover anything. Otherwise, journal_load will do that
1738 * dirty work for us :) */
1740 status = ocfs2_journal_wipe(osb->journal, 0);
1746 mlog(ML_NOTICE, "File system was not unmounted cleanly, "
1747 "recovering volume.\n");
1750 local = ocfs2_mount_local(osb);
1752 /* will play back anything left in the journal. */
1753 status = ocfs2_journal_load(osb->journal, local, dirty);
1755 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
1760 /* recover my local alloc if we didn't unmount cleanly. */
1761 status = ocfs2_begin_local_alloc_recovery(osb,
1768 /* we complete the recovery process after we've marked
1769 * ourselves as mounted. */
1772 mlog(0, "Journal loaded.\n");
1774 status = ocfs2_load_local_alloc(osb);
1781 /* Recovery will be completed after we've mounted the
1782 * rest of the volume. */
1784 osb->local_alloc_copy = local_alloc;
1788 /* go through each journal, trylock it and if you get the
1789 * lock, and it's marked as dirty, set the bit in the recover
1790 * map and launch a recovery thread for it. */
1791 status = ocfs2_mark_dead_nodes(osb);
1804 * The routine gets called from dismount or close whenever a dismount on
1805 * volume is requested and the osb open count becomes 1.
1806 * It will remove the osb from the global list and also free up all the
1807 * initialized resources and fileobject.
1809 static void ocfs2_delete_osb(struct ocfs2_super *osb)
1813 /* This function assumes that the caller has the main osb resource */
1815 ocfs2_free_slot_info(osb);
1817 kfree(osb->osb_orphan_wipes);
1818 kfree(osb->slot_recovery_generations);
1820 * This belongs in journal shutdown, but because we have to
1821 * allocate osb->journal at the start of ocfs2_initalize_osb(),
1824 kfree(osb->journal);
1825 if (osb->local_alloc_copy)
1826 kfree(osb->local_alloc_copy);
1827 kfree(osb->uuid_str);
1828 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
1829 memset(osb, 0, sizeof(struct ocfs2_super));
1834 /* Put OCFS2 into a readonly state, or (if the user specifies it),
1835 * panic(). We do not support continue-on-error operation. */
1836 static void ocfs2_handle_error(struct super_block *sb)
1838 struct ocfs2_super *osb = OCFS2_SB(sb);
1840 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
1841 panic("OCFS2: (device %s): panic forced after error\n",
1844 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
1846 if (sb->s_flags & MS_RDONLY &&
1847 (ocfs2_is_soft_readonly(osb) ||
1848 ocfs2_is_hard_readonly(osb)))
1851 printk(KERN_CRIT "File system is now read-only due to the potential "
1852 "of on-disk corruption. Please run fsck.ocfs2 once the file "
1853 "system is unmounted.\n");
1854 sb->s_flags |= MS_RDONLY;
1855 ocfs2_set_ro_flag(osb, 0);
1858 static char error_buf[1024];
1860 void __ocfs2_error(struct super_block *sb,
1861 const char *function,
1862 const char *fmt, ...)
1866 va_start(args, fmt);
1867 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
1870 /* Not using mlog here because we want to show the actual
1871 * function the error came from. */
1872 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
1873 sb->s_id, function, error_buf);
1875 ocfs2_handle_error(sb);
1878 /* Handle critical errors. This is intentionally more drastic than
1879 * ocfs2_handle_error, so we only use for things like journal errors,
1881 void __ocfs2_abort(struct super_block* sb,
1882 const char *function,
1883 const char *fmt, ...)
1887 va_start(args, fmt);
1888 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
1891 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
1892 sb->s_id, function, error_buf);
1894 /* We don't have the cluster support yet to go straight to
1895 * hard readonly in here. Until then, we want to keep
1896 * ocfs2_abort() so that we can at least mark critical
1899 * TODO: This should abort the journal and alert other nodes
1900 * that our slot needs recovery. */
1902 /* Force a panic(). This stinks, but it's better than letting
1903 * things continue without having a proper hard readonly
1905 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1906 ocfs2_handle_error(sb);
1909 module_init(ocfs2_init);
1910 module_exit(ocfs2_exit);