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"
68 #include "buffer_head_io.h"
70 static struct kmem_cache *ocfs2_inode_cachep = NULL;
72 /* OCFS2 needs to schedule several differnt types of work which
73 * require cluster locking, disk I/O, recovery waits, etc. Since these
74 * types of work tend to be heavy we avoid using the kernel events
75 * workqueue and schedule on our own. */
76 struct workqueue_struct *ocfs2_wq = NULL;
78 static struct dentry *ocfs2_debugfs_root = NULL;
80 MODULE_AUTHOR("Oracle");
81 MODULE_LICENSE("GPL");
85 unsigned long commit_interval;
86 unsigned long mount_opt;
87 unsigned int atime_quantum;
89 unsigned int localalloc_opt;
90 char cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
93 static int ocfs2_parse_options(struct super_block *sb, char *options,
94 struct mount_options *mopt,
96 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt);
97 static void ocfs2_put_super(struct super_block *sb);
98 static int ocfs2_mount_volume(struct super_block *sb);
99 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
100 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
101 static int ocfs2_initialize_mem_caches(void);
102 static void ocfs2_free_mem_caches(void);
103 static void ocfs2_delete_osb(struct ocfs2_super *osb);
105 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
107 static int ocfs2_sync_fs(struct super_block *sb, int wait);
109 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
110 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
111 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
112 static int ocfs2_check_volume(struct ocfs2_super *osb);
113 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
114 struct buffer_head *bh,
116 static int ocfs2_initialize_super(struct super_block *sb,
117 struct buffer_head *bh,
119 static int ocfs2_get_sector(struct super_block *sb,
120 struct buffer_head **bh,
123 static void ocfs2_write_super(struct super_block *sb);
124 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
125 static void ocfs2_destroy_inode(struct inode *inode);
127 static const struct super_operations ocfs2_sops = {
128 .statfs = ocfs2_statfs,
129 .alloc_inode = ocfs2_alloc_inode,
130 .destroy_inode = ocfs2_destroy_inode,
131 .drop_inode = ocfs2_drop_inode,
132 .clear_inode = ocfs2_clear_inode,
133 .delete_inode = ocfs2_delete_inode,
134 .sync_fs = ocfs2_sync_fs,
135 .write_super = ocfs2_write_super,
136 .put_super = ocfs2_put_super,
137 .remount_fs = ocfs2_remount,
138 .show_options = ocfs2_show_options,
160 static const match_table_t tokens = {
161 {Opt_barrier, "barrier=%u"},
162 {Opt_err_panic, "errors=panic"},
163 {Opt_err_ro, "errors=remount-ro"},
165 {Opt_nointr, "nointr"},
166 {Opt_hb_none, OCFS2_HB_NONE},
167 {Opt_hb_local, OCFS2_HB_LOCAL},
168 {Opt_data_ordered, "data=ordered"},
169 {Opt_data_writeback, "data=writeback"},
170 {Opt_atime_quantum, "atime_quantum=%u"},
171 {Opt_slot, "preferred_slot=%u"},
172 {Opt_commit, "commit=%u"},
173 {Opt_localalloc, "localalloc=%d"},
174 {Opt_localflocks, "localflocks"},
175 {Opt_stack, "cluster_stack=%s"},
180 * write_super and sync_fs ripped right out of ext3.
182 static void ocfs2_write_super(struct super_block *sb)
184 if (mutex_trylock(&sb->s_lock) != 0)
189 static int ocfs2_sync_fs(struct super_block *sb, int wait)
193 struct ocfs2_super *osb = OCFS2_SB(sb);
197 if (ocfs2_is_hard_readonly(osb))
201 status = ocfs2_flush_truncate_log(osb);
205 ocfs2_schedule_truncate_log_flush(osb, 0);
208 if (journal_start_commit(OCFS2_SB(sb)->journal->j_journal, &target)) {
210 log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
216 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
218 struct inode *new = NULL;
224 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
226 status = PTR_ERR(new);
230 osb->root_inode = new;
232 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
234 status = PTR_ERR(new);
238 osb->sys_root_inode = new;
240 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
241 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
242 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
244 ocfs2_release_system_inodes(osb);
247 /* FIXME: Should ERROR_RO_FS */
248 mlog(ML_ERROR, "Unable to load system inode %d, "
249 "possibly corrupt fs?", i);
252 // the array now has one ref, so drop this one
261 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
263 struct inode *new = NULL;
269 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
270 i < NUM_SYSTEM_INODES;
272 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
274 ocfs2_release_system_inodes(osb);
276 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
277 status, i, osb->slot_num);
280 /* the array now has one ref, so drop this one */
289 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
296 for (i = 0; i < NUM_SYSTEM_INODES; i++) {
297 inode = osb->system_inodes[i];
300 osb->system_inodes[i] = NULL;
304 inode = osb->sys_root_inode;
307 osb->sys_root_inode = NULL;
310 inode = osb->root_inode;
313 osb->root_inode = NULL;
319 /* We're allocating fs objects, use GFP_NOFS */
320 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
322 struct ocfs2_inode_info *oi;
324 oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
328 return &oi->vfs_inode;
331 static void ocfs2_destroy_inode(struct inode *inode)
333 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
336 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
339 unsigned int bytes = 1 << cbits;
340 unsigned int trim = bytes;
341 unsigned int bitshift = 32;
344 * i_size and all block offsets in ocfs2 are always 64 bits
345 * wide. i_clusters is 32 bits, in cluster-sized units. So on
346 * 64 bit platforms, cluster size will be the limiting factor.
349 #if BITS_PER_LONG == 32
350 # if defined(CONFIG_LBD)
351 BUILD_BUG_ON(sizeof(sector_t) != 8);
353 * We might be limited by page cache size.
355 if (bytes > PAGE_CACHE_SIZE) {
356 bytes = PAGE_CACHE_SIZE;
359 * Shift by 31 here so that we don't get larger than
366 * We are limited by the size of sector_t. Use block size, as
367 * that's what we expose to the VFS.
376 * Trim by a whole cluster when we can actually approach the
377 * on-disk limits. Otherwise we can overflow i_clusters when
378 * an extent start is at the max offset.
380 return (((unsigned long long)bytes) << bitshift) - trim;
383 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
385 int incompat_features;
387 struct mount_options parsed_options;
388 struct ocfs2_super *osb = OCFS2_SB(sb);
390 if (!ocfs2_parse_options(sb, data, &parsed_options, 1)) {
395 if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) !=
396 (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
398 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
402 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
403 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
405 mlog(ML_ERROR, "Cannot change data mode on remount\n");
409 /* We're going to/from readonly mode. */
410 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
411 /* Lock here so the check of HARD_RO and the potential
412 * setting of SOFT_RO is atomic. */
413 spin_lock(&osb->osb_lock);
414 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
415 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
420 if (*flags & MS_RDONLY) {
421 mlog(0, "Going to ro mode.\n");
422 sb->s_flags |= MS_RDONLY;
423 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
425 mlog(0, "Making ro filesystem writeable.\n");
427 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
428 mlog(ML_ERROR, "Cannot remount RDWR "
429 "filesystem due to previous errors.\n");
433 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
434 if (incompat_features) {
435 mlog(ML_ERROR, "Cannot remount RDWR because "
436 "of unsupported optional features "
437 "(%x).\n", incompat_features);
441 sb->s_flags &= ~MS_RDONLY;
442 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
445 spin_unlock(&osb->osb_lock);
449 /* Only save off the new mount options in case of a successful
451 osb->s_mount_opt = parsed_options.mount_opt;
452 osb->s_atime_quantum = parsed_options.atime_quantum;
453 osb->preferred_slot = parsed_options.slot;
454 if (parsed_options.commit_interval)
455 osb->osb_commit_interval = parsed_options.commit_interval;
457 if (!ocfs2_is_hard_readonly(osb))
458 ocfs2_set_journal_params(osb);
464 static int ocfs2_sb_probe(struct super_block *sb,
465 struct buffer_head **bh,
469 struct ocfs1_vol_disk_hdr *hdr;
470 struct ocfs2_dinode *di;
476 *sector_size = bdev_hardsect_size(sb->s_bdev);
477 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
478 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
479 *sector_size, OCFS2_MAX_BLOCKSIZE);
484 /* Can this really happen? */
485 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
486 *sector_size = OCFS2_MIN_BLOCKSIZE;
488 /* check block zero for old format */
489 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
494 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
495 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
496 mlog(ML_ERROR, "incompatible version: %u.%u\n",
497 hdr->major_version, hdr->minor_version);
500 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
501 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
502 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
509 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
510 "upgraded before mounting with ocfs v2\n");
515 * Now check at magic offset for 512, 1024, 2048, 4096
516 * blocksizes. 4096 is the maximum blocksize because it is
517 * the minimum clustersize.
520 for (blksize = *sector_size;
521 blksize <= OCFS2_MAX_BLOCKSIZE;
523 tmpstat = ocfs2_get_sector(sb, bh,
524 OCFS2_SUPER_BLOCK_BLKNO,
531 di = (struct ocfs2_dinode *) (*bh)->b_data;
532 status = ocfs2_verify_volume(di, *bh, blksize);
537 if (status != -EAGAIN)
545 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
547 if (ocfs2_mount_local(osb)) {
548 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
549 mlog(ML_ERROR, "Cannot heartbeat on a locally "
550 "mounted device.\n");
555 if (ocfs2_userspace_stack(osb)) {
556 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
557 mlog(ML_ERROR, "Userspace stack expected, but "
558 "o2cb heartbeat arguments passed to mount\n");
563 if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
564 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
565 !ocfs2_userspace_stack(osb)) {
566 mlog(ML_ERROR, "Heartbeat has to be started to mount "
567 "a read-write clustered device.\n");
576 * If we're using a userspace stack, mount should have passed
577 * a name that matches the disk. If not, mount should not
578 * have passed a stack.
580 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
581 struct mount_options *mopt)
583 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
585 "cluster stack passed to mount, but this filesystem "
586 "does not support it\n");
590 if (ocfs2_userspace_stack(osb) &&
591 strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
592 OCFS2_STACK_LABEL_LEN)) {
594 "cluster stack passed to mount (\"%s\") does not "
595 "match the filesystem (\"%s\")\n",
597 osb->osb_cluster_stack);
604 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
607 int status, sector_size;
608 struct mount_options parsed_options;
609 struct inode *inode = NULL;
610 struct ocfs2_super *osb = NULL;
611 struct buffer_head *bh = NULL;
614 mlog_entry("%p, %p, %i", sb, data, silent);
616 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
618 goto read_super_error;
621 /* probe for superblock */
622 status = ocfs2_sb_probe(sb, &bh, §or_size);
624 mlog(ML_ERROR, "superblock probe failed!\n");
625 goto read_super_error;
628 status = ocfs2_initialize_super(sb, bh, sector_size);
632 goto read_super_error;
636 osb->s_mount_opt = parsed_options.mount_opt;
637 osb->s_atime_quantum = parsed_options.atime_quantum;
638 osb->preferred_slot = parsed_options.slot;
639 osb->osb_commit_interval = parsed_options.commit_interval;
640 osb->local_alloc_default_bits = ocfs2_megabytes_to_clusters(sb, parsed_options.localalloc_opt);
641 osb->local_alloc_bits = osb->local_alloc_default_bits;
643 status = ocfs2_verify_userspace_stack(osb, &parsed_options);
645 goto read_super_error;
647 sb->s_magic = OCFS2_SUPER_MAGIC;
649 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
651 if (bdev_read_only(sb->s_bdev)) {
652 if (!(sb->s_flags & MS_RDONLY)) {
654 mlog(ML_ERROR, "Readonly device detected but readonly "
655 "mount was not specified.\n");
656 goto read_super_error;
659 /* You should not be able to start a local heartbeat
660 * on a readonly device. */
661 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
663 mlog(ML_ERROR, "Local heartbeat specified on readonly "
665 goto read_super_error;
668 status = ocfs2_check_journals_nolocks(osb);
670 if (status == -EROFS)
671 mlog(ML_ERROR, "Recovery required on readonly "
672 "file system, but write access is "
676 goto read_super_error;
679 ocfs2_set_ro_flag(osb, 1);
681 printk(KERN_NOTICE "Readonly device detected. No cluster "
682 "services will be utilized for this mount. Recovery "
683 "will be skipped.\n");
686 if (!ocfs2_is_hard_readonly(osb)) {
687 if (sb->s_flags & MS_RDONLY)
688 ocfs2_set_ro_flag(osb, 0);
691 status = ocfs2_verify_heartbeat(osb);
694 goto read_super_error;
697 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
699 if (!osb->osb_debug_root) {
701 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
702 goto read_super_error;
705 status = ocfs2_mount_volume(sb);
707 inode = igrab(osb->root_inode);
710 goto read_super_error;
715 goto read_super_error;
718 root = d_alloc_root(inode);
722 goto read_super_error;
727 ocfs2_complete_mount_recovery(osb);
729 if (ocfs2_mount_local(osb))
730 snprintf(nodestr, sizeof(nodestr), "local");
732 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
734 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
735 "with %s data mode.\n",
736 osb->dev_str, nodestr, osb->slot_num,
737 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
740 atomic_set(&osb->vol_state, VOLUME_MOUNTED);
741 wake_up(&osb->osb_mount_event);
754 atomic_set(&osb->vol_state, VOLUME_DISABLED);
755 wake_up(&osb->osb_mount_event);
756 ocfs2_dismount_volume(sb, 1);
763 static int ocfs2_get_sb(struct file_system_type *fs_type,
765 const char *dev_name,
767 struct vfsmount *mnt)
769 return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super,
773 static struct file_system_type ocfs2_fs_type = {
774 .owner = THIS_MODULE,
776 .get_sb = ocfs2_get_sb, /* is this called when we mount
778 .kill_sb = kill_block_super, /* set to the generic one
779 * right now, but do we
780 * need to change that? */
781 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
785 static int ocfs2_parse_options(struct super_block *sb,
787 struct mount_options *mopt,
793 mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
794 options ? options : "(none)");
796 mopt->commit_interval = 0;
798 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
799 mopt->slot = OCFS2_INVALID_SLOT;
800 mopt->localalloc_opt = OCFS2_DEFAULT_LOCAL_ALLOC_SIZE;
801 mopt->cluster_stack[0] = '\0';
808 while ((p = strsep(&options, ",")) != NULL) {
810 substring_t args[MAX_OPT_ARGS];
815 token = match_token(p, tokens, args);
818 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
821 mopt->mount_opt &= ~OCFS2_MOUNT_HB_LOCAL;
824 if (match_int(&args[0], &option)) {
829 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
831 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
834 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
837 mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
840 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
843 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
845 case Opt_data_ordered:
846 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
848 case Opt_data_writeback:
849 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
851 case Opt_atime_quantum:
852 if (match_int(&args[0], &option)) {
857 mopt->atime_quantum = option;
861 if (match_int(&args[0], &option)) {
866 mopt->slot = (s16)option;
870 if (match_int(&args[0], &option)) {
877 option = JBD_DEFAULT_MAX_COMMIT_AGE;
878 mopt->commit_interval = HZ * option;
882 if (match_int(&args[0], &option)) {
886 if (option >= 0 && (option <= ocfs2_local_alloc_size(sb) * 8))
887 mopt->localalloc_opt = option;
889 case Opt_localflocks:
891 * Changing this during remount could race
892 * flock() requests, or "unbalance" existing
893 * ones (e.g., a lock is taken in one mode but
894 * dropped in the other). If users care enough
895 * to flip locking modes during remount, we
896 * could add a "local" flag to individual
897 * flock structures for proper tracking of
901 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
904 /* Check both that the option we were passed
905 * is of the right length and that it is a proper
906 * string of the right length.
908 if (((args[0].to - args[0].from) !=
909 OCFS2_STACK_LABEL_LEN) ||
910 (strnlen(args[0].from,
911 OCFS2_STACK_LABEL_LEN) !=
912 OCFS2_STACK_LABEL_LEN)) {
914 "Invalid cluster_stack option\n");
918 memcpy(mopt->cluster_stack, args[0].from,
919 OCFS2_STACK_LABEL_LEN);
920 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
924 "Unrecognized mount option \"%s\" "
925 "or missing value\n", p);
938 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
940 struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb);
941 unsigned long opts = osb->s_mount_opt;
942 unsigned int local_alloc_megs;
944 if (opts & OCFS2_MOUNT_HB_LOCAL)
945 seq_printf(s, ",_netdev,heartbeat=local");
947 seq_printf(s, ",heartbeat=none");
949 if (opts & OCFS2_MOUNT_NOINTR)
950 seq_printf(s, ",nointr");
952 if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
953 seq_printf(s, ",data=writeback");
955 seq_printf(s, ",data=ordered");
957 if (opts & OCFS2_MOUNT_BARRIER)
958 seq_printf(s, ",barrier=1");
960 if (opts & OCFS2_MOUNT_ERRORS_PANIC)
961 seq_printf(s, ",errors=panic");
963 seq_printf(s, ",errors=remount-ro");
965 if (osb->preferred_slot != OCFS2_INVALID_SLOT)
966 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
968 if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM)
969 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
971 if (osb->osb_commit_interval)
972 seq_printf(s, ",commit=%u",
973 (unsigned) (osb->osb_commit_interval / HZ));
975 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
976 if (local_alloc_megs != OCFS2_DEFAULT_LOCAL_ALLOC_SIZE)
977 seq_printf(s, ",localalloc=%d", local_alloc_megs);
979 if (opts & OCFS2_MOUNT_LOCALFLOCKS)
980 seq_printf(s, ",localflocks,");
982 if (osb->osb_cluster_stack[0])
983 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
984 osb->osb_cluster_stack);
989 static int __init ocfs2_init(void)
995 ocfs2_print_version();
997 status = init_ocfs2_uptodate_cache();
1003 status = ocfs2_initialize_mem_caches();
1009 ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1015 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1016 if (!ocfs2_debugfs_root) {
1018 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1021 ocfs2_set_locking_protocol();
1025 ocfs2_free_mem_caches();
1026 exit_ocfs2_uptodate_cache();
1032 return register_filesystem(&ocfs2_fs_type);
1037 static void __exit ocfs2_exit(void)
1042 flush_workqueue(ocfs2_wq);
1043 destroy_workqueue(ocfs2_wq);
1046 debugfs_remove(ocfs2_debugfs_root);
1048 ocfs2_free_mem_caches();
1050 unregister_filesystem(&ocfs2_fs_type);
1052 exit_ocfs2_uptodate_cache();
1057 static void ocfs2_put_super(struct super_block *sb)
1059 mlog_entry("(0x%p)\n", sb);
1061 ocfs2_sync_blockdev(sb);
1062 ocfs2_dismount_volume(sb, 0);
1067 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1069 struct ocfs2_super *osb;
1070 u32 numbits, freebits;
1072 struct ocfs2_dinode *bm_lock;
1073 struct buffer_head *bh = NULL;
1074 struct inode *inode = NULL;
1076 mlog_entry("(%p, %p)\n", dentry->d_sb, buf);
1078 osb = OCFS2_SB(dentry->d_sb);
1080 inode = ocfs2_get_system_file_inode(osb,
1081 GLOBAL_BITMAP_SYSTEM_INODE,
1082 OCFS2_INVALID_SLOT);
1084 mlog(ML_ERROR, "failed to get bitmap inode\n");
1089 status = ocfs2_inode_lock(inode, &bh, 0);
1095 bm_lock = (struct ocfs2_dinode *) bh->b_data;
1097 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1098 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1100 buf->f_type = OCFS2_SUPER_MAGIC;
1101 buf->f_bsize = dentry->d_sb->s_blocksize;
1102 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1103 buf->f_blocks = ((sector_t) numbits) *
1104 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1105 buf->f_bfree = ((sector_t) freebits) *
1106 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1107 buf->f_bavail = buf->f_bfree;
1108 buf->f_files = numbits;
1109 buf->f_ffree = freebits;
1113 ocfs2_inode_unlock(inode, 0);
1124 static void ocfs2_inode_init_once(void *data)
1126 struct ocfs2_inode_info *oi = data;
1129 oi->ip_open_count = 0;
1130 spin_lock_init(&oi->ip_lock);
1131 ocfs2_extent_map_init(&oi->vfs_inode);
1132 INIT_LIST_HEAD(&oi->ip_io_markers);
1133 oi->ip_created_trans = 0;
1134 oi->ip_last_trans = 0;
1135 oi->ip_dir_start_lookup = 0;
1137 init_rwsem(&oi->ip_alloc_sem);
1138 mutex_init(&oi->ip_io_mutex);
1140 oi->ip_blkno = 0ULL;
1141 oi->ip_clusters = 0;
1143 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1144 ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1145 ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1147 ocfs2_metadata_cache_init(&oi->vfs_inode);
1149 inode_init_once(&oi->vfs_inode);
1152 static int ocfs2_initialize_mem_caches(void)
1154 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1155 sizeof(struct ocfs2_inode_info),
1157 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1159 ocfs2_inode_init_once);
1160 if (!ocfs2_inode_cachep)
1166 static void ocfs2_free_mem_caches(void)
1168 if (ocfs2_inode_cachep)
1169 kmem_cache_destroy(ocfs2_inode_cachep);
1171 ocfs2_inode_cachep = NULL;
1174 static int ocfs2_get_sector(struct super_block *sb,
1175 struct buffer_head **bh,
1179 if (!sb_set_blocksize(sb, sect_size)) {
1180 mlog(ML_ERROR, "unable to set blocksize\n");
1184 *bh = sb_getblk(sb, block);
1190 if (!buffer_dirty(*bh))
1191 clear_buffer_uptodate(*bh);
1193 ll_rw_block(READ, 1, bh);
1194 wait_on_buffer(*bh);
1198 static int ocfs2_mount_volume(struct super_block *sb)
1201 int unlock_super = 0;
1202 struct ocfs2_super *osb = OCFS2_SB(sb);
1206 if (ocfs2_is_hard_readonly(osb))
1209 status = ocfs2_dlm_init(osb);
1215 status = ocfs2_super_lock(osb, 1);
1222 /* This will load up the node map and add ourselves to it. */
1223 status = ocfs2_find_slot(osb);
1229 /* load all node-local system inodes */
1230 status = ocfs2_init_local_system_inodes(osb);
1236 status = ocfs2_check_volume(osb);
1242 status = ocfs2_truncate_log_init(osb);
1248 if (ocfs2_mount_local(osb))
1253 ocfs2_super_unlock(osb, 1);
1259 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1261 int tmp, hangup_needed = 0;
1262 struct ocfs2_super *osb = NULL;
1265 mlog_entry("(0x%p)\n", sb);
1271 ocfs2_shutdown_local_alloc(osb);
1273 ocfs2_truncate_log_shutdown(osb);
1275 /* This will disable recovery and flush any recovery work. */
1276 ocfs2_recovery_exit(osb);
1278 ocfs2_journal_shutdown(osb);
1280 ocfs2_sync_blockdev(sb);
1282 /* No cluster connection means we've failed during mount, so skip
1283 * all the steps which depended on that to complete. */
1285 tmp = ocfs2_super_lock(osb, 1);
1292 if (osb->slot_num != OCFS2_INVALID_SLOT)
1293 ocfs2_put_slot(osb);
1296 ocfs2_super_unlock(osb, 1);
1298 ocfs2_release_system_inodes(osb);
1301 * If we're dismounting due to mount error, mount.ocfs2 will clean
1302 * up heartbeat. If we're a local mount, there is no heartbeat.
1303 * If we failed before we got a uuid_str yet, we can't stop
1304 * heartbeat. Otherwise, do it.
1306 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str)
1310 ocfs2_dlm_shutdown(osb, hangup_needed);
1312 debugfs_remove(osb->osb_debug_root);
1315 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1317 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1319 if (ocfs2_mount_local(osb))
1320 snprintf(nodestr, sizeof(nodestr), "local");
1322 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1324 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1325 osb->dev_str, nodestr);
1327 ocfs2_delete_osb(osb);
1330 sb->s_fs_info = NULL;
1333 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1334 unsigned uuid_bytes)
1339 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1341 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1342 if (osb->uuid_str == NULL)
1345 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1346 /* print with null */
1347 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1348 if (ret != 2) /* drop super cleans up */
1350 /* then only advance past the last char */
1357 static int ocfs2_initialize_super(struct super_block *sb,
1358 struct buffer_head *bh,
1362 int i, cbits, bbits;
1363 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
1364 struct inode *inode = NULL;
1365 struct ocfs2_journal *journal;
1366 __le32 uuid_net_key;
1367 struct ocfs2_super *osb;
1371 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
1378 sb->s_fs_info = osb;
1379 sb->s_op = &ocfs2_sops;
1380 sb->s_export_op = &ocfs2_export_ops;
1381 sb->s_time_gran = 1;
1382 sb->s_flags |= MS_NOATIME;
1383 /* this is needed to support O_LARGEFILE */
1384 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
1385 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
1386 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
1389 /* Save off for ocfs2_rw_direct */
1390 osb->s_sectsize_bits = blksize_bits(sector_size);
1391 BUG_ON(!osb->s_sectsize_bits);
1393 spin_lock_init(&osb->dc_task_lock);
1394 init_waitqueue_head(&osb->dc_event);
1395 osb->dc_work_sequence = 0;
1396 osb->dc_wake_sequence = 0;
1397 INIT_LIST_HEAD(&osb->blocked_lock_list);
1398 osb->blocked_lock_count = 0;
1399 spin_lock_init(&osb->osb_lock);
1400 ocfs2_init_inode_steal_slot(osb);
1402 atomic_set(&osb->alloc_stats.moves, 0);
1403 atomic_set(&osb->alloc_stats.local_data, 0);
1404 atomic_set(&osb->alloc_stats.bitmap_data, 0);
1405 atomic_set(&osb->alloc_stats.bg_allocs, 0);
1406 atomic_set(&osb->alloc_stats.bg_extends, 0);
1408 ocfs2_init_node_maps(osb);
1410 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
1411 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
1413 status = ocfs2_recovery_init(osb);
1415 mlog(ML_ERROR, "Unable to initialize recovery state\n");
1420 init_waitqueue_head(&osb->checkpoint_event);
1421 atomic_set(&osb->needs_checkpoint, 0);
1423 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1425 osb->slot_num = OCFS2_INVALID_SLOT;
1427 osb->local_alloc_state = OCFS2_LA_UNUSED;
1428 osb->local_alloc_bh = NULL;
1429 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
1431 init_waitqueue_head(&osb->osb_mount_event);
1433 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
1434 if (!osb->vol_label) {
1435 mlog(ML_ERROR, "unable to alloc vol label\n");
1440 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
1441 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
1442 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
1447 mlog(0, "max_slots for this device: %u\n", osb->max_slots);
1449 osb->slot_recovery_generations =
1450 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
1452 if (!osb->slot_recovery_generations) {
1458 init_waitqueue_head(&osb->osb_wipe_event);
1459 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
1460 sizeof(*osb->osb_orphan_wipes),
1462 if (!osb->osb_orphan_wipes) {
1468 osb->s_feature_compat =
1469 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
1470 osb->s_feature_ro_compat =
1471 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
1472 osb->s_feature_incompat =
1473 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
1475 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
1476 mlog(ML_ERROR, "couldn't mount because of unsupported "
1477 "optional features (%x).\n", i);
1481 if (!(osb->sb->s_flags & MS_RDONLY) &&
1482 (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
1483 mlog(ML_ERROR, "couldn't mount RDWR because of "
1484 "unsupported optional features (%x).\n", i);
1489 if (ocfs2_userspace_stack(osb)) {
1490 memcpy(osb->osb_cluster_stack,
1491 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
1492 OCFS2_STACK_LABEL_LEN);
1493 osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1494 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
1496 "couldn't mount because of an invalid "
1497 "cluster stack label (%s) \n",
1498 osb->osb_cluster_stack);
1503 /* The empty string is identical with classic tools that
1504 * don't know about s_cluster_info. */
1505 osb->osb_cluster_stack[0] = '\0';
1508 get_random_bytes(&osb->s_next_generation, sizeof(u32));
1511 * This should be done in ocfs2_journal_init(), but unknown
1512 * ordering issues will cause the filesystem to crash.
1513 * If anyone wants to figure out what part of the code
1514 * refers to osb->journal before ocfs2_journal_init() is run,
1517 /* initialize our journal structure */
1519 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
1521 mlog(ML_ERROR, "unable to alloc journal\n");
1525 osb->journal = journal;
1526 journal->j_osb = osb;
1528 atomic_set(&journal->j_num_trans, 0);
1529 init_rwsem(&journal->j_trans_barrier);
1530 init_waitqueue_head(&journal->j_checkpointed);
1531 spin_lock_init(&journal->j_lock);
1532 journal->j_trans_id = (unsigned long) 1;
1533 INIT_LIST_HEAD(&journal->j_la_cleanups);
1534 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
1535 journal->j_state = OCFS2_JOURNAL_FREE;
1537 /* get some pseudo constants for clustersize bits */
1538 osb->s_clustersize_bits =
1539 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
1540 osb->s_clustersize = 1 << osb->s_clustersize_bits;
1541 mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
1543 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
1544 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
1545 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
1546 osb->s_clustersize);
1551 if (ocfs2_clusters_to_blocks(osb->sb, le32_to_cpu(di->i_clusters) - 1)
1553 mlog(ML_ERROR, "Volume might try to write to blocks beyond "
1554 "what jbd can address in 32 bits.\n");
1559 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
1560 sizeof(di->id2.i_super.s_uuid))) {
1561 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
1566 memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
1568 strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
1569 osb->vol_label[63] = '\0';
1570 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
1571 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
1572 osb->first_cluster_group_blkno =
1573 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
1574 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
1575 mlog(0, "vol_label: %s\n", osb->vol_label);
1576 mlog(0, "uuid: %s\n", osb->uuid_str);
1577 mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
1578 (unsigned long long)osb->root_blkno,
1579 (unsigned long long)osb->system_dir_blkno);
1581 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
1582 if (!osb->osb_dlm_debug) {
1588 atomic_set(&osb->vol_state, VOLUME_INIT);
1590 /* load root, system_dir, and all global system inodes */
1591 status = ocfs2_init_global_system_inodes(osb);
1600 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
1601 OCFS2_INVALID_SLOT);
1608 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
1611 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb) * 8;
1613 status = ocfs2_init_slot_info(osb);
1625 * will return: -EAGAIN if it is ok to keep searching for superblocks
1626 * -EINVAL if there is a bad superblock
1629 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
1630 struct buffer_head *bh,
1633 int status = -EAGAIN;
1637 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
1638 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
1640 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
1641 mlog(ML_ERROR, "found superblock with incorrect block "
1642 "size: found %u, should be %u\n",
1643 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
1645 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
1646 OCFS2_MAJOR_REV_LEVEL ||
1647 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
1648 OCFS2_MINOR_REV_LEVEL) {
1649 mlog(ML_ERROR, "found superblock with bad version: "
1650 "found %u.%u, should be %u.%u\n",
1651 le16_to_cpu(di->id2.i_super.s_major_rev_level),
1652 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
1653 OCFS2_MAJOR_REV_LEVEL,
1654 OCFS2_MINOR_REV_LEVEL);
1655 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
1656 mlog(ML_ERROR, "bad block number on superblock: "
1657 "found %llu, should be %llu\n",
1658 (unsigned long long)le64_to_cpu(di->i_blkno),
1659 (unsigned long long)bh->b_blocknr);
1660 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
1661 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
1662 mlog(ML_ERROR, "bad cluster size found: %u\n",
1663 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
1664 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
1665 mlog(ML_ERROR, "bad root_blkno: 0\n");
1666 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
1667 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
1668 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
1670 "Superblock slots found greater than file system "
1671 "maximum: found %u, max %u\n",
1672 le16_to_cpu(di->id2.i_super.s_max_slots),
1684 static int ocfs2_check_volume(struct ocfs2_super *osb)
1689 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
1695 /* Init our journal object. */
1696 status = ocfs2_journal_init(osb->journal, &dirty);
1698 mlog(ML_ERROR, "Could not initialize journal!\n");
1702 /* If the journal was unmounted cleanly then we don't want to
1703 * recover anything. Otherwise, journal_load will do that
1704 * dirty work for us :) */
1706 status = ocfs2_journal_wipe(osb->journal, 0);
1712 mlog(ML_NOTICE, "File system was not unmounted cleanly, "
1713 "recovering volume.\n");
1716 local = ocfs2_mount_local(osb);
1718 /* will play back anything left in the journal. */
1719 status = ocfs2_journal_load(osb->journal, local, dirty);
1721 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
1726 /* recover my local alloc if we didn't unmount cleanly. */
1727 status = ocfs2_begin_local_alloc_recovery(osb,
1734 /* we complete the recovery process after we've marked
1735 * ourselves as mounted. */
1738 mlog(0, "Journal loaded.\n");
1740 status = ocfs2_load_local_alloc(osb);
1747 /* Recovery will be completed after we've mounted the
1748 * rest of the volume. */
1750 osb->local_alloc_copy = local_alloc;
1754 /* go through each journal, trylock it and if you get the
1755 * lock, and it's marked as dirty, set the bit in the recover
1756 * map and launch a recovery thread for it. */
1757 status = ocfs2_mark_dead_nodes(osb);
1770 * The routine gets called from dismount or close whenever a dismount on
1771 * volume is requested and the osb open count becomes 1.
1772 * It will remove the osb from the global list and also free up all the
1773 * initialized resources and fileobject.
1775 static void ocfs2_delete_osb(struct ocfs2_super *osb)
1779 /* This function assumes that the caller has the main osb resource */
1781 ocfs2_free_slot_info(osb);
1783 kfree(osb->osb_orphan_wipes);
1784 kfree(osb->slot_recovery_generations);
1786 * This belongs in journal shutdown, but because we have to
1787 * allocate osb->journal at the start of ocfs2_initalize_osb(),
1790 kfree(osb->journal);
1791 if (osb->local_alloc_copy)
1792 kfree(osb->local_alloc_copy);
1793 kfree(osb->uuid_str);
1794 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
1795 memset(osb, 0, sizeof(struct ocfs2_super));
1800 /* Put OCFS2 into a readonly state, or (if the user specifies it),
1801 * panic(). We do not support continue-on-error operation. */
1802 static void ocfs2_handle_error(struct super_block *sb)
1804 struct ocfs2_super *osb = OCFS2_SB(sb);
1806 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
1807 panic("OCFS2: (device %s): panic forced after error\n",
1810 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
1812 if (sb->s_flags & MS_RDONLY &&
1813 (ocfs2_is_soft_readonly(osb) ||
1814 ocfs2_is_hard_readonly(osb)))
1817 printk(KERN_CRIT "File system is now read-only due to the potential "
1818 "of on-disk corruption. Please run fsck.ocfs2 once the file "
1819 "system is unmounted.\n");
1820 sb->s_flags |= MS_RDONLY;
1821 ocfs2_set_ro_flag(osb, 0);
1824 static char error_buf[1024];
1826 void __ocfs2_error(struct super_block *sb,
1827 const char *function,
1828 const char *fmt, ...)
1832 va_start(args, fmt);
1833 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
1836 /* Not using mlog here because we want to show the actual
1837 * function the error came from. */
1838 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
1839 sb->s_id, function, error_buf);
1841 ocfs2_handle_error(sb);
1844 /* Handle critical errors. This is intentionally more drastic than
1845 * ocfs2_handle_error, so we only use for things like journal errors,
1847 void __ocfs2_abort(struct super_block* sb,
1848 const char *function,
1849 const char *fmt, ...)
1853 va_start(args, fmt);
1854 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
1857 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
1858 sb->s_id, function, error_buf);
1860 /* We don't have the cluster support yet to go straight to
1861 * hard readonly in here. Until then, we want to keep
1862 * ocfs2_abort() so that we can at least mark critical
1865 * TODO: This should abort the journal and alert other nodes
1866 * that our slot needs recovery. */
1868 /* Force a panic(). This stinks, but it's better than letting
1869 * things continue without having a proper hard readonly
1871 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1872 ocfs2_handle_error(sb);
1875 module_init(ocfs2_init);
1876 module_exit(ocfs2_exit);