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/init.h>
32 #include <linux/random.h>
33 #include <linux/statfs.h>
34 #include <linux/moduleparam.h>
35 #include <linux/blkdev.h>
36 #include <linux/socket.h>
37 #include <linux/inet.h>
38 #include <linux/parser.h>
39 #include <linux/crc32.h>
40 #include <linux/debugfs.h>
41 #include <linux/mount.h>
42 #include <linux/seq_file.h>
43 #include <linux/quotaops.h>
44 #include <linux/smp_lock.h>
46 #define MLOG_MASK_PREFIX ML_SUPER
47 #include <cluster/masklog.h>
51 /* this should be the only file to include a version 1 header */
52 #include "ocfs1_fs_compat.h"
55 #include "blockcheck.h"
58 #include "extent_map.h"
59 #include "heartbeat.h"
62 #include "localalloc.h"
71 #include "refcounttree.h"
74 #include "buffer_head_io.h"
76 static struct kmem_cache *ocfs2_inode_cachep = NULL;
77 struct kmem_cache *ocfs2_dquot_cachep;
78 struct kmem_cache *ocfs2_qf_chunk_cachep;
80 /* OCFS2 needs to schedule several differnt types of work which
81 * require cluster locking, disk I/O, recovery waits, etc. Since these
82 * types of work tend to be heavy we avoid using the kernel events
83 * workqueue and schedule on our own. */
84 struct workqueue_struct *ocfs2_wq = NULL;
86 static struct dentry *ocfs2_debugfs_root = NULL;
88 MODULE_AUTHOR("Oracle");
89 MODULE_LICENSE("GPL");
93 unsigned long commit_interval;
94 unsigned long mount_opt;
95 unsigned int atime_quantum;
98 unsigned int resv_level;
99 char cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
102 static int ocfs2_parse_options(struct super_block *sb, char *options,
103 struct mount_options *mopt,
105 static int ocfs2_check_set_options(struct super_block *sb,
106 struct mount_options *options);
107 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt);
108 static void ocfs2_put_super(struct super_block *sb);
109 static int ocfs2_mount_volume(struct super_block *sb);
110 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
111 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
112 static int ocfs2_initialize_mem_caches(void);
113 static void ocfs2_free_mem_caches(void);
114 static void ocfs2_delete_osb(struct ocfs2_super *osb);
116 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
118 static int ocfs2_sync_fs(struct super_block *sb, int wait);
120 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
121 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
122 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
123 static int ocfs2_check_volume(struct ocfs2_super *osb);
124 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
125 struct buffer_head *bh,
127 struct ocfs2_blockcheck_stats *stats);
128 static int ocfs2_initialize_super(struct super_block *sb,
129 struct buffer_head *bh,
131 struct ocfs2_blockcheck_stats *stats);
132 static int ocfs2_get_sector(struct super_block *sb,
133 struct buffer_head **bh,
136 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
137 static void ocfs2_destroy_inode(struct inode *inode);
138 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
139 static int ocfs2_enable_quotas(struct ocfs2_super *osb);
140 static void ocfs2_disable_quotas(struct ocfs2_super *osb);
142 static const struct super_operations ocfs2_sops = {
143 .statfs = ocfs2_statfs,
144 .alloc_inode = ocfs2_alloc_inode,
145 .destroy_inode = ocfs2_destroy_inode,
146 .drop_inode = ocfs2_drop_inode,
147 .clear_inode = ocfs2_clear_inode,
148 .delete_inode = ocfs2_delete_inode,
149 .sync_fs = ocfs2_sync_fs,
150 .put_super = ocfs2_put_super,
151 .remount_fs = ocfs2_remount,
152 .show_options = ocfs2_show_options,
153 .quota_read = ocfs2_quota_read,
154 .quota_write = ocfs2_quota_write,
184 static const match_table_t tokens = {
185 {Opt_barrier, "barrier=%u"},
186 {Opt_err_panic, "errors=panic"},
187 {Opt_err_ro, "errors=remount-ro"},
189 {Opt_nointr, "nointr"},
190 {Opt_hb_none, OCFS2_HB_NONE},
191 {Opt_hb_local, OCFS2_HB_LOCAL},
192 {Opt_data_ordered, "data=ordered"},
193 {Opt_data_writeback, "data=writeback"},
194 {Opt_atime_quantum, "atime_quantum=%u"},
195 {Opt_slot, "preferred_slot=%u"},
196 {Opt_commit, "commit=%u"},
197 {Opt_localalloc, "localalloc=%d"},
198 {Opt_localflocks, "localflocks"},
199 {Opt_stack, "cluster_stack=%s"},
200 {Opt_user_xattr, "user_xattr"},
201 {Opt_nouser_xattr, "nouser_xattr"},
202 {Opt_inode64, "inode64"},
204 {Opt_noacl, "noacl"},
205 {Opt_usrquota, "usrquota"},
206 {Opt_grpquota, "grpquota"},
207 {Opt_resv_level, "resv_level=%u"},
211 #ifdef CONFIG_DEBUG_FS
212 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
214 struct ocfs2_cluster_connection *cconn = osb->cconn;
215 struct ocfs2_recovery_map *rm = osb->recovery_map;
216 struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
219 out += snprintf(buf + out, len - out,
220 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n",
221 "Device", osb->dev_str, osb->uuid_str,
222 osb->fs_generation, osb->vol_label);
224 out += snprintf(buf + out, len - out,
225 "%10s => State: %d Flags: 0x%lX\n", "Volume",
226 atomic_read(&osb->vol_state), osb->osb_flags);
228 out += snprintf(buf + out, len - out,
229 "%10s => Block: %lu Cluster: %d\n", "Sizes",
230 osb->sb->s_blocksize, osb->s_clustersize);
232 out += snprintf(buf + out, len - out,
233 "%10s => Compat: 0x%X Incompat: 0x%X "
235 "Features", osb->s_feature_compat,
236 osb->s_feature_incompat, osb->s_feature_ro_compat);
238 out += snprintf(buf + out, len - out,
239 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount",
240 osb->s_mount_opt, osb->s_atime_quantum);
243 out += snprintf(buf + out, len - out,
244 "%10s => Stack: %s Name: %*s "
245 "Version: %d.%d\n", "Cluster",
246 (*osb->osb_cluster_stack == '\0' ?
247 "o2cb" : osb->osb_cluster_stack),
248 cconn->cc_namelen, cconn->cc_name,
249 cconn->cc_version.pv_major,
250 cconn->cc_version.pv_minor);
253 spin_lock(&osb->dc_task_lock);
254 out += snprintf(buf + out, len - out,
255 "%10s => Pid: %d Count: %lu WakeSeq: %lu "
256 "WorkSeq: %lu\n", "DownCnvt",
257 (osb->dc_task ? task_pid_nr(osb->dc_task) : -1),
258 osb->blocked_lock_count, osb->dc_wake_sequence,
259 osb->dc_work_sequence);
260 spin_unlock(&osb->dc_task_lock);
262 spin_lock(&osb->osb_lock);
263 out += snprintf(buf + out, len - out, "%10s => Pid: %d Nodes:",
265 (osb->recovery_thread_task ?
266 task_pid_nr(osb->recovery_thread_task) : -1));
267 if (rm->rm_used == 0)
268 out += snprintf(buf + out, len - out, " None\n");
270 for (i = 0; i < rm->rm_used; i++)
271 out += snprintf(buf + out, len - out, " %d",
273 out += snprintf(buf + out, len - out, "\n");
275 spin_unlock(&osb->osb_lock);
277 out += snprintf(buf + out, len - out,
278 "%10s => Pid: %d Interval: %lu Needs: %d\n", "Commit",
279 (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
280 osb->osb_commit_interval,
281 atomic_read(&osb->needs_checkpoint));
283 out += snprintf(buf + out, len - out,
284 "%10s => State: %d TxnId: %lu NumTxns: %d\n",
285 "Journal", osb->journal->j_state,
286 osb->journal->j_trans_id,
287 atomic_read(&osb->journal->j_num_trans));
289 out += snprintf(buf + out, len - out,
290 "%10s => GlobalAllocs: %d LocalAllocs: %d "
291 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
293 atomic_read(&osb->alloc_stats.bitmap_data),
294 atomic_read(&osb->alloc_stats.local_data),
295 atomic_read(&osb->alloc_stats.bg_allocs),
296 atomic_read(&osb->alloc_stats.moves),
297 atomic_read(&osb->alloc_stats.bg_extends));
299 out += snprintf(buf + out, len - out,
300 "%10s => State: %u Descriptor: %llu Size: %u bits "
301 "Default: %u bits\n",
302 "LocalAlloc", osb->local_alloc_state,
303 (unsigned long long)osb->la_last_gd,
304 osb->local_alloc_bits, osb->local_alloc_default_bits);
306 spin_lock(&osb->osb_lock);
307 out += snprintf(buf + out, len - out,
308 "%10s => InodeSlot: %d StolenInodes: %d, "
309 "MetaSlot: %d StolenMeta: %d\n", "Steal",
310 osb->s_inode_steal_slot,
311 atomic_read(&osb->s_num_inodes_stolen),
312 osb->s_meta_steal_slot,
313 atomic_read(&osb->s_num_meta_stolen));
314 spin_unlock(&osb->osb_lock);
316 out += snprintf(buf + out, len - out, "OrphanScan => ");
317 out += snprintf(buf + out, len - out, "Local: %u Global: %u ",
318 os->os_count, os->os_seqno);
319 out += snprintf(buf + out, len - out, " Last Scan: ");
320 if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
321 out += snprintf(buf + out, len - out, "Disabled\n");
323 out += snprintf(buf + out, len - out, "%lu seconds ago\n",
324 (get_seconds() - os->os_scantime.tv_sec));
326 out += snprintf(buf + out, len - out, "%10s => %3s %10s\n",
327 "Slots", "Num", "RecoGen");
328 for (i = 0; i < osb->max_slots; ++i) {
329 out += snprintf(buf + out, len - out,
330 "%10s %c %3d %10d\n",
332 (i == osb->slot_num ? '*' : ' '),
333 i, osb->slot_recovery_generations[i]);
339 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
341 struct ocfs2_super *osb = inode->i_private;
344 buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
348 i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
350 file->private_data = buf;
357 static int ocfs2_debug_release(struct inode *inode, struct file *file)
359 kfree(file->private_data);
363 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
364 size_t nbytes, loff_t *ppos)
366 return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
367 i_size_read(file->f_mapping->host));
370 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
374 static int ocfs2_debug_release(struct inode *inode, struct file *file)
378 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
379 size_t nbytes, loff_t *ppos)
383 #endif /* CONFIG_DEBUG_FS */
385 static const struct file_operations ocfs2_osb_debug_fops = {
386 .open = ocfs2_osb_debug_open,
387 .release = ocfs2_debug_release,
388 .read = ocfs2_debug_read,
389 .llseek = generic_file_llseek,
392 static int ocfs2_sync_fs(struct super_block *sb, int wait)
396 struct ocfs2_super *osb = OCFS2_SB(sb);
398 if (ocfs2_is_hard_readonly(osb))
402 status = ocfs2_flush_truncate_log(osb);
406 ocfs2_schedule_truncate_log_flush(osb, 0);
409 if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
412 jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
418 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
420 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
421 && (ino == USER_QUOTA_SYSTEM_INODE
422 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
424 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
425 && (ino == GROUP_QUOTA_SYSTEM_INODE
426 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
431 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
433 struct inode *new = NULL;
439 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
441 status = PTR_ERR(new);
445 osb->root_inode = new;
447 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
449 status = PTR_ERR(new);
453 osb->sys_root_inode = new;
455 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
456 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
457 if (!ocfs2_need_system_inode(osb, i))
459 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
461 ocfs2_release_system_inodes(osb);
464 /* FIXME: Should ERROR_RO_FS */
465 mlog(ML_ERROR, "Unable to load system inode %d, "
466 "possibly corrupt fs?", i);
469 // the array now has one ref, so drop this one
478 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
480 struct inode *new = NULL;
486 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
487 i < NUM_SYSTEM_INODES;
489 if (!ocfs2_need_system_inode(osb, i))
491 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
493 ocfs2_release_system_inodes(osb);
495 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
496 status, i, osb->slot_num);
499 /* the array now has one ref, so drop this one */
508 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
515 for (i = 0; i < NUM_SYSTEM_INODES; i++) {
516 inode = osb->system_inodes[i];
519 osb->system_inodes[i] = NULL;
523 inode = osb->sys_root_inode;
526 osb->sys_root_inode = NULL;
529 inode = osb->root_inode;
532 osb->root_inode = NULL;
538 /* We're allocating fs objects, use GFP_NOFS */
539 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
541 struct ocfs2_inode_info *oi;
543 oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
547 jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
548 return &oi->vfs_inode;
551 static void ocfs2_destroy_inode(struct inode *inode)
553 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
556 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
559 unsigned int bytes = 1 << cbits;
560 unsigned int trim = bytes;
561 unsigned int bitshift = 32;
564 * i_size and all block offsets in ocfs2 are always 64 bits
565 * wide. i_clusters is 32 bits, in cluster-sized units. So on
566 * 64 bit platforms, cluster size will be the limiting factor.
569 #if BITS_PER_LONG == 32
570 # if defined(CONFIG_LBDAF)
571 BUILD_BUG_ON(sizeof(sector_t) != 8);
573 * We might be limited by page cache size.
575 if (bytes > PAGE_CACHE_SIZE) {
576 bytes = PAGE_CACHE_SIZE;
579 * Shift by 31 here so that we don't get larger than
586 * We are limited by the size of sector_t. Use block size, as
587 * that's what we expose to the VFS.
596 * Trim by a whole cluster when we can actually approach the
597 * on-disk limits. Otherwise we can overflow i_clusters when
598 * an extent start is at the max offset.
600 return (((unsigned long long)bytes) << bitshift) - trim;
603 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
605 int incompat_features;
607 struct mount_options parsed_options;
608 struct ocfs2_super *osb = OCFS2_SB(sb);
612 if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
613 !ocfs2_check_set_options(sb, &parsed_options)) {
618 if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) !=
619 (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
621 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
625 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
626 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
628 mlog(ML_ERROR, "Cannot change data mode on remount\n");
632 /* Probably don't want this on remount; it might
633 * mess with other nodes */
634 if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
635 (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
637 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
641 /* We're going to/from readonly mode. */
642 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
643 /* Disable quota accounting before remounting RO */
644 if (*flags & MS_RDONLY) {
645 ret = ocfs2_susp_quotas(osb, 0);
649 /* Lock here so the check of HARD_RO and the potential
650 * setting of SOFT_RO is atomic. */
651 spin_lock(&osb->osb_lock);
652 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
653 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
658 if (*flags & MS_RDONLY) {
659 mlog(0, "Going to ro mode.\n");
660 sb->s_flags |= MS_RDONLY;
661 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
663 mlog(0, "Making ro filesystem writeable.\n");
665 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
666 mlog(ML_ERROR, "Cannot remount RDWR "
667 "filesystem due to previous errors.\n");
671 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
672 if (incompat_features) {
673 mlog(ML_ERROR, "Cannot remount RDWR because "
674 "of unsupported optional features "
675 "(%x).\n", incompat_features);
679 sb->s_flags &= ~MS_RDONLY;
680 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
683 spin_unlock(&osb->osb_lock);
684 /* Enable quota accounting after remounting RW */
685 if (!ret && !(*flags & MS_RDONLY)) {
686 if (sb_any_quota_suspended(sb))
687 ret = ocfs2_susp_quotas(osb, 1);
689 ret = ocfs2_enable_quotas(osb);
691 /* Return back changes... */
692 spin_lock(&osb->osb_lock);
693 sb->s_flags |= MS_RDONLY;
694 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
695 spin_unlock(&osb->osb_lock);
702 /* Only save off the new mount options in case of a successful
704 osb->s_mount_opt = parsed_options.mount_opt;
705 osb->s_atime_quantum = parsed_options.atime_quantum;
706 osb->preferred_slot = parsed_options.slot;
707 if (parsed_options.commit_interval)
708 osb->osb_commit_interval = parsed_options.commit_interval;
710 if (!ocfs2_is_hard_readonly(osb))
711 ocfs2_set_journal_params(osb);
713 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
714 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
722 static int ocfs2_sb_probe(struct super_block *sb,
723 struct buffer_head **bh,
725 struct ocfs2_blockcheck_stats *stats)
728 struct ocfs1_vol_disk_hdr *hdr;
729 struct ocfs2_dinode *di;
735 *sector_size = bdev_logical_block_size(sb->s_bdev);
736 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
737 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
738 *sector_size, OCFS2_MAX_BLOCKSIZE);
743 /* Can this really happen? */
744 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
745 *sector_size = OCFS2_MIN_BLOCKSIZE;
747 /* check block zero for old format */
748 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
753 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
754 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
755 mlog(ML_ERROR, "incompatible version: %u.%u\n",
756 hdr->major_version, hdr->minor_version);
759 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
760 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
761 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
768 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
769 "upgraded before mounting with ocfs v2\n");
774 * Now check at magic offset for 512, 1024, 2048, 4096
775 * blocksizes. 4096 is the maximum blocksize because it is
776 * the minimum clustersize.
779 for (blksize = *sector_size;
780 blksize <= OCFS2_MAX_BLOCKSIZE;
782 tmpstat = ocfs2_get_sector(sb, bh,
783 OCFS2_SUPER_BLOCK_BLKNO,
790 di = (struct ocfs2_dinode *) (*bh)->b_data;
791 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
792 spin_lock_init(&stats->b_lock);
793 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
798 if (tmpstat != -EAGAIN) {
808 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
810 if (ocfs2_mount_local(osb)) {
811 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
812 mlog(ML_ERROR, "Cannot heartbeat on a locally "
813 "mounted device.\n");
818 if (ocfs2_userspace_stack(osb)) {
819 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
820 mlog(ML_ERROR, "Userspace stack expected, but "
821 "o2cb heartbeat arguments passed to mount\n");
826 if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
827 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
828 !ocfs2_userspace_stack(osb)) {
829 mlog(ML_ERROR, "Heartbeat has to be started to mount "
830 "a read-write clustered device.\n");
839 * If we're using a userspace stack, mount should have passed
840 * a name that matches the disk. If not, mount should not
841 * have passed a stack.
843 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
844 struct mount_options *mopt)
846 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
848 "cluster stack passed to mount, but this filesystem "
849 "does not support it\n");
853 if (ocfs2_userspace_stack(osb) &&
854 strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
855 OCFS2_STACK_LABEL_LEN)) {
857 "cluster stack passed to mount (\"%s\") does not "
858 "match the filesystem (\"%s\")\n",
860 osb->osb_cluster_stack);
867 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
870 struct super_block *sb = osb->sb;
871 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
872 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
875 for (type = 0; type < MAXQUOTAS; type++) {
876 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
879 status = vfs_quota_enable(
880 sb_dqopt(sb)->files[type],
884 status = vfs_quota_disable(sb, type,
890 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
891 "remount (error = %d).\n", status);
895 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
897 struct inode *inode[MAXQUOTAS] = { NULL, NULL };
898 struct super_block *sb = osb->sb;
899 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
900 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
901 unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
902 LOCAL_GROUP_QUOTA_SYSTEM_INODE };
906 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
907 for (type = 0; type < MAXQUOTAS; type++) {
908 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
910 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
916 status = vfs_quota_enable(inode[type], type, QFMT_OCFS2,
917 DQUOT_USAGE_ENABLED);
922 for (type = 0; type < MAXQUOTAS; type++)
926 ocfs2_disable_quotas(osb);
927 for (type = 0; type < MAXQUOTAS; type++)
933 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
937 struct super_block *sb = osb->sb;
939 /* We mostly ignore errors in this function because there's not much
940 * we can do when we see them */
941 for (type = 0; type < MAXQUOTAS; type++) {
942 if (!sb_has_quota_loaded(sb, type))
944 inode = igrab(sb->s_dquot.files[type]);
945 /* Turn off quotas. This will remove all dquot structures from
946 * memory and so they will be automatically synced to global
948 vfs_quota_disable(sb, type, DQUOT_USAGE_ENABLED |
949 DQUOT_LIMITS_ENABLED);
956 /* Handle quota on quotactl */
957 static int ocfs2_quota_on(struct super_block *sb, int type, int format_id,
958 char *path, int remount)
960 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
961 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
963 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
967 return 0; /* Just ignore it has been handled in
969 return vfs_quota_enable(sb_dqopt(sb)->files[type], type,
970 format_id, DQUOT_LIMITS_ENABLED);
973 /* Handle quota off quotactl */
974 static int ocfs2_quota_off(struct super_block *sb, int type, int remount)
977 return 0; /* Ignore now and handle later in
979 return vfs_quota_disable(sb, type, DQUOT_LIMITS_ENABLED);
982 static const struct quotactl_ops ocfs2_quotactl_ops = {
983 .quota_on = ocfs2_quota_on,
984 .quota_off = ocfs2_quota_off,
985 .quota_sync = vfs_quota_sync,
986 .get_info = vfs_get_dqinfo,
987 .set_info = vfs_set_dqinfo,
988 .get_dqblk = vfs_get_dqblk,
989 .set_dqblk = vfs_set_dqblk,
992 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
995 int status, sector_size;
996 struct mount_options parsed_options;
997 struct inode *inode = NULL;
998 struct ocfs2_super *osb = NULL;
999 struct buffer_head *bh = NULL;
1001 struct ocfs2_blockcheck_stats stats;
1003 mlog_entry("%p, %p, %i", sb, data, silent);
1005 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
1007 goto read_super_error;
1010 /* probe for superblock */
1011 status = ocfs2_sb_probe(sb, &bh, §or_size, &stats);
1013 mlog(ML_ERROR, "superblock probe failed!\n");
1014 goto read_super_error;
1017 status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1021 goto read_super_error;
1026 if (!ocfs2_check_set_options(sb, &parsed_options)) {
1028 goto read_super_error;
1030 osb->s_mount_opt = parsed_options.mount_opt;
1031 osb->s_atime_quantum = parsed_options.atime_quantum;
1032 osb->preferred_slot = parsed_options.slot;
1033 osb->osb_commit_interval = parsed_options.commit_interval;
1035 ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1036 osb->osb_resv_level = parsed_options.resv_level;
1038 status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1040 goto read_super_error;
1042 sb->s_magic = OCFS2_SUPER_MAGIC;
1044 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1045 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1047 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1049 if (bdev_read_only(sb->s_bdev)) {
1050 if (!(sb->s_flags & MS_RDONLY)) {
1052 mlog(ML_ERROR, "Readonly device detected but readonly "
1053 "mount was not specified.\n");
1054 goto read_super_error;
1057 /* You should not be able to start a local heartbeat
1058 * on a readonly device. */
1059 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1061 mlog(ML_ERROR, "Local heartbeat specified on readonly "
1063 goto read_super_error;
1066 status = ocfs2_check_journals_nolocks(osb);
1068 if (status == -EROFS)
1069 mlog(ML_ERROR, "Recovery required on readonly "
1070 "file system, but write access is "
1074 goto read_super_error;
1077 ocfs2_set_ro_flag(osb, 1);
1079 printk(KERN_NOTICE "Readonly device detected. No cluster "
1080 "services will be utilized for this mount. Recovery "
1081 "will be skipped.\n");
1084 if (!ocfs2_is_hard_readonly(osb)) {
1085 if (sb->s_flags & MS_RDONLY)
1086 ocfs2_set_ro_flag(osb, 0);
1089 status = ocfs2_verify_heartbeat(osb);
1092 goto read_super_error;
1095 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1096 ocfs2_debugfs_root);
1097 if (!osb->osb_debug_root) {
1099 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1100 goto read_super_error;
1103 osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1104 osb->osb_debug_root,
1106 &ocfs2_osb_debug_fops);
1107 if (!osb->osb_ctxt) {
1110 goto read_super_error;
1113 if (ocfs2_meta_ecc(osb)) {
1114 status = ocfs2_blockcheck_stats_debugfs_install(
1115 &osb->osb_ecc_stats,
1116 osb->osb_debug_root);
1119 "Unable to create blockcheck statistics "
1121 goto read_super_error;
1125 status = ocfs2_mount_volume(sb);
1126 if (osb->root_inode)
1127 inode = igrab(osb->root_inode);
1130 goto read_super_error;
1135 goto read_super_error;
1138 root = d_alloc_root(inode);
1142 goto read_super_error;
1147 ocfs2_complete_mount_recovery(osb);
1149 if (ocfs2_mount_local(osb))
1150 snprintf(nodestr, sizeof(nodestr), "local");
1152 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1154 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1155 "with %s data mode.\n",
1156 osb->dev_str, nodestr, osb->slot_num,
1157 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1160 atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1161 wake_up(&osb->osb_mount_event);
1163 /* Now we can initialize quotas because we can afford to wait
1164 * for cluster locks recovery now. That also means that truncation
1165 * log recovery can happen but that waits for proper quota setup */
1166 if (!(sb->s_flags & MS_RDONLY)) {
1167 status = ocfs2_enable_quotas(osb);
1169 /* We have to err-out specially here because
1170 * s_root is already set */
1172 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1173 wake_up(&osb->osb_mount_event);
1179 ocfs2_complete_quota_recovery(osb);
1181 /* Now we wake up again for processes waiting for quotas */
1182 atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1183 wake_up(&osb->osb_mount_event);
1185 /* Start this when the mount is almost sure of being successful */
1186 ocfs2_orphan_scan_start(osb);
1198 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1199 wake_up(&osb->osb_mount_event);
1200 ocfs2_dismount_volume(sb, 1);
1207 static int ocfs2_get_sb(struct file_system_type *fs_type,
1209 const char *dev_name,
1211 struct vfsmount *mnt)
1213 return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super,
1217 static void ocfs2_kill_sb(struct super_block *sb)
1219 struct ocfs2_super *osb = OCFS2_SB(sb);
1222 if (!osb || atomic_read(&osb->vol_state) == VOLUME_DISABLED)
1225 /* Prevent further queueing of inode drop events */
1226 spin_lock(&dentry_list_lock);
1227 ocfs2_set_osb_flag(osb, OCFS2_OSB_DROP_DENTRY_LOCK_IMMED);
1228 spin_unlock(&dentry_list_lock);
1229 /* Wait for work to finish and/or remove it */
1230 cancel_work_sync(&osb->dentry_lock_work);
1232 kill_block_super(sb);
1235 static struct file_system_type ocfs2_fs_type = {
1236 .owner = THIS_MODULE,
1238 .get_sb = ocfs2_get_sb, /* is this called when we mount
1240 .kill_sb = ocfs2_kill_sb,
1242 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1246 static int ocfs2_check_set_options(struct super_block *sb,
1247 struct mount_options *options)
1249 if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1250 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1251 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1252 mlog(ML_ERROR, "User quotas were requested, but this "
1253 "filesystem does not have the feature enabled.\n");
1256 if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1257 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1258 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1259 mlog(ML_ERROR, "Group quotas were requested, but this "
1260 "filesystem does not have the feature enabled.\n");
1263 if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1264 !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1265 mlog(ML_ERROR, "ACL support requested but extended attributes "
1266 "feature is not enabled\n");
1269 /* No ACL setting specified? Use XATTR feature... */
1270 if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1271 OCFS2_MOUNT_NO_POSIX_ACL))) {
1272 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1273 options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1275 options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1280 static int ocfs2_parse_options(struct super_block *sb,
1282 struct mount_options *mopt,
1288 mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
1289 options ? options : "(none)");
1291 mopt->commit_interval = 0;
1292 mopt->mount_opt = 0;
1293 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1294 mopt->slot = OCFS2_INVALID_SLOT;
1295 mopt->localalloc_opt = -1;
1296 mopt->cluster_stack[0] = '\0';
1297 mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1304 while ((p = strsep(&options, ",")) != NULL) {
1306 substring_t args[MAX_OPT_ARGS];
1311 token = match_token(p, tokens, args);
1314 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1317 mopt->mount_opt &= ~OCFS2_MOUNT_HB_LOCAL;
1320 if (match_int(&args[0], &option)) {
1325 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1327 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1330 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1333 mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1336 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1339 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1341 case Opt_data_ordered:
1342 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1344 case Opt_data_writeback:
1345 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1347 case Opt_user_xattr:
1348 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1350 case Opt_nouser_xattr:
1351 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1353 case Opt_atime_quantum:
1354 if (match_int(&args[0], &option)) {
1359 mopt->atime_quantum = option;
1363 if (match_int(&args[0], &option)) {
1368 mopt->slot = (s16)option;
1372 if (match_int(&args[0], &option)) {
1379 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1380 mopt->commit_interval = HZ * option;
1382 case Opt_localalloc:
1384 if (match_int(&args[0], &option)) {
1389 mopt->localalloc_opt = option;
1391 case Opt_localflocks:
1393 * Changing this during remount could race
1394 * flock() requests, or "unbalance" existing
1395 * ones (e.g., a lock is taken in one mode but
1396 * dropped in the other). If users care enough
1397 * to flip locking modes during remount, we
1398 * could add a "local" flag to individual
1399 * flock structures for proper tracking of
1403 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1406 /* Check both that the option we were passed
1407 * is of the right length and that it is a proper
1408 * string of the right length.
1410 if (((args[0].to - args[0].from) !=
1411 OCFS2_STACK_LABEL_LEN) ||
1412 (strnlen(args[0].from,
1413 OCFS2_STACK_LABEL_LEN) !=
1414 OCFS2_STACK_LABEL_LEN)) {
1416 "Invalid cluster_stack option\n");
1420 memcpy(mopt->cluster_stack, args[0].from,
1421 OCFS2_STACK_LABEL_LEN);
1422 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1425 mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1428 mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1431 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1434 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1435 mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1438 mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1439 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1441 case Opt_resv_level:
1444 if (match_int(&args[0], &option)) {
1448 if (option >= OCFS2_MIN_RESV_LEVEL &&
1449 option < OCFS2_MAX_RESV_LEVEL)
1450 mopt->resv_level = option;
1454 "Unrecognized mount option \"%s\" "
1455 "or missing value\n", p);
1468 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1470 struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb);
1471 unsigned long opts = osb->s_mount_opt;
1472 unsigned int local_alloc_megs;
1474 if (opts & OCFS2_MOUNT_HB_LOCAL)
1475 seq_printf(s, ",_netdev,heartbeat=local");
1477 seq_printf(s, ",heartbeat=none");
1479 if (opts & OCFS2_MOUNT_NOINTR)
1480 seq_printf(s, ",nointr");
1482 if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1483 seq_printf(s, ",data=writeback");
1485 seq_printf(s, ",data=ordered");
1487 if (opts & OCFS2_MOUNT_BARRIER)
1488 seq_printf(s, ",barrier=1");
1490 if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1491 seq_printf(s, ",errors=panic");
1493 seq_printf(s, ",errors=remount-ro");
1495 if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1496 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1498 if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM)
1499 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1501 if (osb->osb_commit_interval)
1502 seq_printf(s, ",commit=%u",
1503 (unsigned) (osb->osb_commit_interval / HZ));
1505 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1506 if (local_alloc_megs != ocfs2_la_default_mb(osb))
1507 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1509 if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1510 seq_printf(s, ",localflocks,");
1512 if (osb->osb_cluster_stack[0])
1513 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1514 osb->osb_cluster_stack);
1515 if (opts & OCFS2_MOUNT_USRQUOTA)
1516 seq_printf(s, ",usrquota");
1517 if (opts & OCFS2_MOUNT_GRPQUOTA)
1518 seq_printf(s, ",grpquota");
1520 if (opts & OCFS2_MOUNT_NOUSERXATTR)
1521 seq_printf(s, ",nouser_xattr");
1523 seq_printf(s, ",user_xattr");
1525 if (opts & OCFS2_MOUNT_INODE64)
1526 seq_printf(s, ",inode64");
1528 if (opts & OCFS2_MOUNT_POSIX_ACL)
1529 seq_printf(s, ",acl");
1531 seq_printf(s, ",noacl");
1533 if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1534 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1539 static int __init ocfs2_init(void)
1545 ocfs2_print_version();
1547 status = init_ocfs2_uptodate_cache();
1553 status = ocfs2_initialize_mem_caches();
1559 ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1565 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1566 if (!ocfs2_debugfs_root) {
1568 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1571 status = ocfs2_quota_setup();
1575 ocfs2_set_locking_protocol();
1577 status = register_quota_format(&ocfs2_quota_format);
1580 ocfs2_quota_shutdown();
1581 ocfs2_free_mem_caches();
1582 exit_ocfs2_uptodate_cache();
1588 return register_filesystem(&ocfs2_fs_type);
1593 static void __exit ocfs2_exit(void)
1597 ocfs2_quota_shutdown();
1600 flush_workqueue(ocfs2_wq);
1601 destroy_workqueue(ocfs2_wq);
1604 unregister_quota_format(&ocfs2_quota_format);
1606 debugfs_remove(ocfs2_debugfs_root);
1608 ocfs2_free_mem_caches();
1610 unregister_filesystem(&ocfs2_fs_type);
1612 exit_ocfs2_uptodate_cache();
1617 static void ocfs2_put_super(struct super_block *sb)
1619 mlog_entry("(0x%p)\n", sb);
1623 ocfs2_sync_blockdev(sb);
1624 ocfs2_dismount_volume(sb, 0);
1631 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1633 struct ocfs2_super *osb;
1634 u32 numbits, freebits;
1636 struct ocfs2_dinode *bm_lock;
1637 struct buffer_head *bh = NULL;
1638 struct inode *inode = NULL;
1640 mlog_entry("(%p, %p)\n", dentry->d_sb, buf);
1642 osb = OCFS2_SB(dentry->d_sb);
1644 inode = ocfs2_get_system_file_inode(osb,
1645 GLOBAL_BITMAP_SYSTEM_INODE,
1646 OCFS2_INVALID_SLOT);
1648 mlog(ML_ERROR, "failed to get bitmap inode\n");
1653 status = ocfs2_inode_lock(inode, &bh, 0);
1659 bm_lock = (struct ocfs2_dinode *) bh->b_data;
1661 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1662 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1664 buf->f_type = OCFS2_SUPER_MAGIC;
1665 buf->f_bsize = dentry->d_sb->s_blocksize;
1666 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1667 buf->f_blocks = ((sector_t) numbits) *
1668 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1669 buf->f_bfree = ((sector_t) freebits) *
1670 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1671 buf->f_bavail = buf->f_bfree;
1672 buf->f_files = numbits;
1673 buf->f_ffree = freebits;
1674 buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1676 buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1677 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1681 ocfs2_inode_unlock(inode, 0);
1692 static void ocfs2_inode_init_once(void *data)
1694 struct ocfs2_inode_info *oi = data;
1697 oi->ip_open_count = 0;
1698 spin_lock_init(&oi->ip_lock);
1699 ocfs2_extent_map_init(&oi->vfs_inode);
1700 INIT_LIST_HEAD(&oi->ip_io_markers);
1701 oi->ip_dir_start_lookup = 0;
1703 init_rwsem(&oi->ip_alloc_sem);
1704 init_rwsem(&oi->ip_xattr_sem);
1705 mutex_init(&oi->ip_io_mutex);
1707 oi->ip_blkno = 0ULL;
1708 oi->ip_clusters = 0;
1710 ocfs2_resv_init_once(&oi->ip_la_data_resv);
1712 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1713 ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1714 ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1716 ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1717 &ocfs2_inode_caching_ops);
1719 inode_init_once(&oi->vfs_inode);
1722 static int ocfs2_initialize_mem_caches(void)
1724 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1725 sizeof(struct ocfs2_inode_info),
1727 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1729 ocfs2_inode_init_once);
1730 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1731 sizeof(struct ocfs2_dquot),
1733 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1736 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1737 sizeof(struct ocfs2_quota_chunk),
1739 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1741 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1742 !ocfs2_qf_chunk_cachep) {
1743 if (ocfs2_inode_cachep)
1744 kmem_cache_destroy(ocfs2_inode_cachep);
1745 if (ocfs2_dquot_cachep)
1746 kmem_cache_destroy(ocfs2_dquot_cachep);
1747 if (ocfs2_qf_chunk_cachep)
1748 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1755 static void ocfs2_free_mem_caches(void)
1757 if (ocfs2_inode_cachep)
1758 kmem_cache_destroy(ocfs2_inode_cachep);
1759 ocfs2_inode_cachep = NULL;
1761 if (ocfs2_dquot_cachep)
1762 kmem_cache_destroy(ocfs2_dquot_cachep);
1763 ocfs2_dquot_cachep = NULL;
1765 if (ocfs2_qf_chunk_cachep)
1766 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1767 ocfs2_qf_chunk_cachep = NULL;
1770 static int ocfs2_get_sector(struct super_block *sb,
1771 struct buffer_head **bh,
1775 if (!sb_set_blocksize(sb, sect_size)) {
1776 mlog(ML_ERROR, "unable to set blocksize\n");
1780 *bh = sb_getblk(sb, block);
1786 if (!buffer_dirty(*bh))
1787 clear_buffer_uptodate(*bh);
1789 ll_rw_block(READ, 1, bh);
1790 wait_on_buffer(*bh);
1791 if (!buffer_uptodate(*bh)) {
1801 static int ocfs2_mount_volume(struct super_block *sb)
1804 int unlock_super = 0;
1805 struct ocfs2_super *osb = OCFS2_SB(sb);
1809 if (ocfs2_is_hard_readonly(osb))
1812 status = ocfs2_dlm_init(osb);
1818 status = ocfs2_super_lock(osb, 1);
1825 /* This will load up the node map and add ourselves to it. */
1826 status = ocfs2_find_slot(osb);
1832 /* load all node-local system inodes */
1833 status = ocfs2_init_local_system_inodes(osb);
1839 status = ocfs2_check_volume(osb);
1845 status = ocfs2_truncate_log_init(osb);
1851 ocfs2_super_unlock(osb, 1);
1857 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1859 int tmp, hangup_needed = 0;
1860 struct ocfs2_super *osb = NULL;
1863 mlog_entry("(0x%p)\n", sb);
1869 debugfs_remove(osb->osb_ctxt);
1872 * Flush inode dropping work queue so that deletes are
1873 * performed while the filesystem is still working
1875 ocfs2_drop_all_dl_inodes(osb);
1877 /* Orphan scan should be stopped as early as possible */
1878 ocfs2_orphan_scan_stop(osb);
1880 ocfs2_disable_quotas(osb);
1882 ocfs2_shutdown_local_alloc(osb);
1884 ocfs2_truncate_log_shutdown(osb);
1886 /* This will disable recovery and flush any recovery work. */
1887 ocfs2_recovery_exit(osb);
1889 ocfs2_journal_shutdown(osb);
1891 ocfs2_sync_blockdev(sb);
1893 ocfs2_purge_refcount_trees(osb);
1895 /* No cluster connection means we've failed during mount, so skip
1896 * all the steps which depended on that to complete. */
1898 tmp = ocfs2_super_lock(osb, 1);
1905 if (osb->slot_num != OCFS2_INVALID_SLOT)
1906 ocfs2_put_slot(osb);
1909 ocfs2_super_unlock(osb, 1);
1911 ocfs2_release_system_inodes(osb);
1914 * If we're dismounting due to mount error, mount.ocfs2 will clean
1915 * up heartbeat. If we're a local mount, there is no heartbeat.
1916 * If we failed before we got a uuid_str yet, we can't stop
1917 * heartbeat. Otherwise, do it.
1919 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str)
1923 ocfs2_dlm_shutdown(osb, hangup_needed);
1925 ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1926 debugfs_remove(osb->osb_debug_root);
1929 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1931 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1933 if (ocfs2_mount_local(osb))
1934 snprintf(nodestr, sizeof(nodestr), "local");
1936 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1938 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1939 osb->dev_str, nodestr);
1941 ocfs2_delete_osb(osb);
1944 sb->s_fs_info = NULL;
1947 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1948 unsigned uuid_bytes)
1953 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1955 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1956 if (osb->uuid_str == NULL)
1959 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1960 /* print with null */
1961 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1962 if (ret != 2) /* drop super cleans up */
1964 /* then only advance past the last char */
1971 static int ocfs2_initialize_super(struct super_block *sb,
1972 struct buffer_head *bh,
1974 struct ocfs2_blockcheck_stats *stats)
1977 int i, cbits, bbits;
1978 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
1979 struct inode *inode = NULL;
1980 struct ocfs2_journal *journal;
1981 __le32 uuid_net_key;
1982 struct ocfs2_super *osb;
1986 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
1993 sb->s_fs_info = osb;
1994 sb->s_op = &ocfs2_sops;
1995 sb->s_export_op = &ocfs2_export_ops;
1996 sb->s_qcop = &ocfs2_quotactl_ops;
1997 sb->dq_op = &ocfs2_quota_operations;
1998 sb->s_xattr = ocfs2_xattr_handlers;
1999 sb->s_time_gran = 1;
2000 sb->s_flags |= MS_NOATIME;
2001 /* this is needed to support O_LARGEFILE */
2002 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2003 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2004 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2006 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2008 for (i = 0; i < 3; i++)
2009 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2010 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2013 /* Save off for ocfs2_rw_direct */
2014 osb->s_sectsize_bits = blksize_bits(sector_size);
2015 BUG_ON(!osb->s_sectsize_bits);
2017 spin_lock_init(&osb->dc_task_lock);
2018 init_waitqueue_head(&osb->dc_event);
2019 osb->dc_work_sequence = 0;
2020 osb->dc_wake_sequence = 0;
2021 INIT_LIST_HEAD(&osb->blocked_lock_list);
2022 osb->blocked_lock_count = 0;
2023 spin_lock_init(&osb->osb_lock);
2024 spin_lock_init(&osb->osb_xattr_lock);
2025 ocfs2_init_steal_slots(osb);
2027 atomic_set(&osb->alloc_stats.moves, 0);
2028 atomic_set(&osb->alloc_stats.local_data, 0);
2029 atomic_set(&osb->alloc_stats.bitmap_data, 0);
2030 atomic_set(&osb->alloc_stats.bg_allocs, 0);
2031 atomic_set(&osb->alloc_stats.bg_extends, 0);
2033 /* Copy the blockcheck stats from the superblock probe */
2034 osb->osb_ecc_stats = *stats;
2036 ocfs2_init_node_maps(osb);
2038 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2039 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2041 ocfs2_orphan_scan_init(osb);
2043 status = ocfs2_recovery_init(osb);
2045 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2050 init_waitqueue_head(&osb->checkpoint_event);
2051 atomic_set(&osb->needs_checkpoint, 0);
2053 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2055 osb->slot_num = OCFS2_INVALID_SLOT;
2057 osb->s_xattr_inline_size = le16_to_cpu(
2058 di->id2.i_super.s_xattr_inline_size);
2060 osb->local_alloc_state = OCFS2_LA_UNUSED;
2061 osb->local_alloc_bh = NULL;
2062 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2064 init_waitqueue_head(&osb->osb_mount_event);
2066 status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2072 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2073 if (!osb->vol_label) {
2074 mlog(ML_ERROR, "unable to alloc vol label\n");
2079 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2080 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2081 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2086 mlog(0, "max_slots for this device: %u\n", osb->max_slots);
2088 osb->slot_recovery_generations =
2089 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2091 if (!osb->slot_recovery_generations) {
2097 init_waitqueue_head(&osb->osb_wipe_event);
2098 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2099 sizeof(*osb->osb_orphan_wipes),
2101 if (!osb->osb_orphan_wipes) {
2107 osb->osb_rf_lock_tree = RB_ROOT;
2109 osb->s_feature_compat =
2110 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2111 osb->s_feature_ro_compat =
2112 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2113 osb->s_feature_incompat =
2114 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2116 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2117 mlog(ML_ERROR, "couldn't mount because of unsupported "
2118 "optional features (%x).\n", i);
2122 if (!(osb->sb->s_flags & MS_RDONLY) &&
2123 (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2124 mlog(ML_ERROR, "couldn't mount RDWR because of "
2125 "unsupported optional features (%x).\n", i);
2130 if (ocfs2_userspace_stack(osb)) {
2131 memcpy(osb->osb_cluster_stack,
2132 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2133 OCFS2_STACK_LABEL_LEN);
2134 osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
2135 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2137 "couldn't mount because of an invalid "
2138 "cluster stack label (%s) \n",
2139 osb->osb_cluster_stack);
2144 /* The empty string is identical with classic tools that
2145 * don't know about s_cluster_info. */
2146 osb->osb_cluster_stack[0] = '\0';
2149 get_random_bytes(&osb->s_next_generation, sizeof(u32));
2152 * This should be done in ocfs2_journal_init(), but unknown
2153 * ordering issues will cause the filesystem to crash.
2154 * If anyone wants to figure out what part of the code
2155 * refers to osb->journal before ocfs2_journal_init() is run,
2158 /* initialize our journal structure */
2160 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2162 mlog(ML_ERROR, "unable to alloc journal\n");
2166 osb->journal = journal;
2167 journal->j_osb = osb;
2169 atomic_set(&journal->j_num_trans, 0);
2170 init_rwsem(&journal->j_trans_barrier);
2171 init_waitqueue_head(&journal->j_checkpointed);
2172 spin_lock_init(&journal->j_lock);
2173 journal->j_trans_id = (unsigned long) 1;
2174 INIT_LIST_HEAD(&journal->j_la_cleanups);
2175 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2176 journal->j_state = OCFS2_JOURNAL_FREE;
2178 INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes);
2179 osb->dentry_lock_list = NULL;
2181 /* get some pseudo constants for clustersize bits */
2182 osb->s_clustersize_bits =
2183 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2184 osb->s_clustersize = 1 << osb->s_clustersize_bits;
2185 mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
2187 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2188 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2189 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2190 osb->s_clustersize);
2195 if (ocfs2_clusters_to_blocks(osb->sb, le32_to_cpu(di->i_clusters) - 1)
2197 mlog(ML_ERROR, "Volume might try to write to blocks beyond "
2198 "what jbd can address in 32 bits.\n");
2203 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2204 sizeof(di->id2.i_super.s_uuid))) {
2205 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2210 memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
2212 strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2213 osb->vol_label[63] = '\0';
2214 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2215 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2216 osb->first_cluster_group_blkno =
2217 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2218 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2219 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2220 mlog(0, "vol_label: %s\n", osb->vol_label);
2221 mlog(0, "uuid: %s\n", osb->uuid_str);
2222 mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
2223 (unsigned long long)osb->root_blkno,
2224 (unsigned long long)osb->system_dir_blkno);
2226 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2227 if (!osb->osb_dlm_debug) {
2233 atomic_set(&osb->vol_state, VOLUME_INIT);
2235 /* load root, system_dir, and all global system inodes */
2236 status = ocfs2_init_global_system_inodes(osb);
2245 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2246 OCFS2_INVALID_SLOT);
2253 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2254 osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2257 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb) * 8;
2259 status = ocfs2_init_slot_info(osb);
2271 * will return: -EAGAIN if it is ok to keep searching for superblocks
2272 * -EINVAL if there is a bad superblock
2275 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2276 struct buffer_head *bh,
2278 struct ocfs2_blockcheck_stats *stats)
2280 int status = -EAGAIN;
2284 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2285 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2286 /* We have to do a raw check of the feature here */
2287 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2288 OCFS2_FEATURE_INCOMPAT_META_ECC) {
2289 status = ocfs2_block_check_validate(bh->b_data,
2297 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2298 mlog(ML_ERROR, "found superblock with incorrect block "
2299 "size: found %u, should be %u\n",
2300 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2302 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2303 OCFS2_MAJOR_REV_LEVEL ||
2304 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2305 OCFS2_MINOR_REV_LEVEL) {
2306 mlog(ML_ERROR, "found superblock with bad version: "
2307 "found %u.%u, should be %u.%u\n",
2308 le16_to_cpu(di->id2.i_super.s_major_rev_level),
2309 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2310 OCFS2_MAJOR_REV_LEVEL,
2311 OCFS2_MINOR_REV_LEVEL);
2312 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2313 mlog(ML_ERROR, "bad block number on superblock: "
2314 "found %llu, should be %llu\n",
2315 (unsigned long long)le64_to_cpu(di->i_blkno),
2316 (unsigned long long)bh->b_blocknr);
2317 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2318 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2319 mlog(ML_ERROR, "bad cluster size found: %u\n",
2320 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2321 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2322 mlog(ML_ERROR, "bad root_blkno: 0\n");
2323 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2324 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2325 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2327 "Superblock slots found greater than file system "
2328 "maximum: found %u, max %u\n",
2329 le16_to_cpu(di->id2.i_super.s_max_slots),
2342 static int ocfs2_check_volume(struct ocfs2_super *osb)
2347 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2353 /* Init our journal object. */
2354 status = ocfs2_journal_init(osb->journal, &dirty);
2356 mlog(ML_ERROR, "Could not initialize journal!\n");
2360 /* If the journal was unmounted cleanly then we don't want to
2361 * recover anything. Otherwise, journal_load will do that
2362 * dirty work for us :) */
2364 status = ocfs2_journal_wipe(osb->journal, 0);
2370 mlog(ML_NOTICE, "File system was not unmounted cleanly, "
2371 "recovering volume.\n");
2374 local = ocfs2_mount_local(osb);
2376 /* will play back anything left in the journal. */
2377 status = ocfs2_journal_load(osb->journal, local, dirty);
2379 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2384 /* recover my local alloc if we didn't unmount cleanly. */
2385 status = ocfs2_begin_local_alloc_recovery(osb,
2392 /* we complete the recovery process after we've marked
2393 * ourselves as mounted. */
2396 mlog(0, "Journal loaded.\n");
2398 status = ocfs2_load_local_alloc(osb);
2405 /* Recovery will be completed after we've mounted the
2406 * rest of the volume. */
2408 osb->local_alloc_copy = local_alloc;
2412 /* go through each journal, trylock it and if you get the
2413 * lock, and it's marked as dirty, set the bit in the recover
2414 * map and launch a recovery thread for it. */
2415 status = ocfs2_mark_dead_nodes(osb);
2421 status = ocfs2_compute_replay_slots(osb);
2434 * The routine gets called from dismount or close whenever a dismount on
2435 * volume is requested and the osb open count becomes 1.
2436 * It will remove the osb from the global list and also free up all the
2437 * initialized resources and fileobject.
2439 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2443 /* This function assumes that the caller has the main osb resource */
2445 ocfs2_free_slot_info(osb);
2447 kfree(osb->osb_orphan_wipes);
2448 kfree(osb->slot_recovery_generations);
2450 * This belongs in journal shutdown, but because we have to
2451 * allocate osb->journal at the start of ocfs2_initalize_osb(),
2454 kfree(osb->journal);
2455 if (osb->local_alloc_copy)
2456 kfree(osb->local_alloc_copy);
2457 kfree(osb->uuid_str);
2458 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2459 memset(osb, 0, sizeof(struct ocfs2_super));
2464 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2465 * panic(). We do not support continue-on-error operation. */
2466 static void ocfs2_handle_error(struct super_block *sb)
2468 struct ocfs2_super *osb = OCFS2_SB(sb);
2470 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2471 panic("OCFS2: (device %s): panic forced after error\n",
2474 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2476 if (sb->s_flags & MS_RDONLY &&
2477 (ocfs2_is_soft_readonly(osb) ||
2478 ocfs2_is_hard_readonly(osb)))
2481 printk(KERN_CRIT "File system is now read-only due to the potential "
2482 "of on-disk corruption. Please run fsck.ocfs2 once the file "
2483 "system is unmounted.\n");
2484 sb->s_flags |= MS_RDONLY;
2485 ocfs2_set_ro_flag(osb, 0);
2488 static char error_buf[1024];
2490 void __ocfs2_error(struct super_block *sb,
2491 const char *function,
2492 const char *fmt, ...)
2496 va_start(args, fmt);
2497 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2500 /* Not using mlog here because we want to show the actual
2501 * function the error came from. */
2502 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2503 sb->s_id, function, error_buf);
2505 ocfs2_handle_error(sb);
2508 /* Handle critical errors. This is intentionally more drastic than
2509 * ocfs2_handle_error, so we only use for things like journal errors,
2511 void __ocfs2_abort(struct super_block* sb,
2512 const char *function,
2513 const char *fmt, ...)
2517 va_start(args, fmt);
2518 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2521 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2522 sb->s_id, function, error_buf);
2524 /* We don't have the cluster support yet to go straight to
2525 * hard readonly in here. Until then, we want to keep
2526 * ocfs2_abort() so that we can at least mark critical
2529 * TODO: This should abort the journal and alert other nodes
2530 * that our slot needs recovery. */
2532 /* Force a panic(). This stinks, but it's better than letting
2533 * things continue without having a proper hard readonly
2535 if (!ocfs2_mount_local(OCFS2_SB(sb)))
2536 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2537 ocfs2_handle_error(sb);
2540 module_init(ocfs2_init);
2541 module_exit(ocfs2_exit);