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"
73 #include "buffer_head_io.h"
75 static struct kmem_cache *ocfs2_inode_cachep = NULL;
76 struct kmem_cache *ocfs2_dquot_cachep;
77 struct kmem_cache *ocfs2_qf_chunk_cachep;
79 /* OCFS2 needs to schedule several differnt types of work which
80 * require cluster locking, disk I/O, recovery waits, etc. Since these
81 * types of work tend to be heavy we avoid using the kernel events
82 * workqueue and schedule on our own. */
83 struct workqueue_struct *ocfs2_wq = NULL;
85 static struct dentry *ocfs2_debugfs_root = NULL;
87 MODULE_AUTHOR("Oracle");
88 MODULE_LICENSE("GPL");
92 unsigned long commit_interval;
93 unsigned long mount_opt;
94 unsigned int atime_quantum;
96 unsigned int localalloc_opt;
97 char cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
100 static int ocfs2_parse_options(struct super_block *sb, char *options,
101 struct mount_options *mopt,
103 static int ocfs2_check_set_options(struct super_block *sb,
104 struct mount_options *options);
105 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt);
106 static void ocfs2_put_super(struct super_block *sb);
107 static int ocfs2_mount_volume(struct super_block *sb);
108 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
109 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
110 static int ocfs2_initialize_mem_caches(void);
111 static void ocfs2_free_mem_caches(void);
112 static void ocfs2_delete_osb(struct ocfs2_super *osb);
114 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
116 static int ocfs2_sync_fs(struct super_block *sb, int wait);
118 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
119 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
120 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
121 static int ocfs2_check_volume(struct ocfs2_super *osb);
122 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
123 struct buffer_head *bh,
125 struct ocfs2_blockcheck_stats *stats);
126 static int ocfs2_initialize_super(struct super_block *sb,
127 struct buffer_head *bh,
129 struct ocfs2_blockcheck_stats *stats);
130 static int ocfs2_get_sector(struct super_block *sb,
131 struct buffer_head **bh,
134 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
135 static void ocfs2_destroy_inode(struct inode *inode);
136 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
137 static int ocfs2_enable_quotas(struct ocfs2_super *osb);
138 static void ocfs2_disable_quotas(struct ocfs2_super *osb);
140 static const struct super_operations ocfs2_sops = {
141 .statfs = ocfs2_statfs,
142 .alloc_inode = ocfs2_alloc_inode,
143 .destroy_inode = ocfs2_destroy_inode,
144 .drop_inode = ocfs2_drop_inode,
145 .clear_inode = ocfs2_clear_inode,
146 .delete_inode = ocfs2_delete_inode,
147 .sync_fs = ocfs2_sync_fs,
148 .put_super = ocfs2_put_super,
149 .remount_fs = ocfs2_remount,
150 .show_options = ocfs2_show_options,
151 .quota_read = ocfs2_quota_read,
152 .quota_write = ocfs2_quota_write,
181 static const match_table_t tokens = {
182 {Opt_barrier, "barrier=%u"},
183 {Opt_err_panic, "errors=panic"},
184 {Opt_err_ro, "errors=remount-ro"},
186 {Opt_nointr, "nointr"},
187 {Opt_hb_none, OCFS2_HB_NONE},
188 {Opt_hb_local, OCFS2_HB_LOCAL},
189 {Opt_data_ordered, "data=ordered"},
190 {Opt_data_writeback, "data=writeback"},
191 {Opt_atime_quantum, "atime_quantum=%u"},
192 {Opt_slot, "preferred_slot=%u"},
193 {Opt_commit, "commit=%u"},
194 {Opt_localalloc, "localalloc=%d"},
195 {Opt_localflocks, "localflocks"},
196 {Opt_stack, "cluster_stack=%s"},
197 {Opt_user_xattr, "user_xattr"},
198 {Opt_nouser_xattr, "nouser_xattr"},
199 {Opt_inode64, "inode64"},
201 {Opt_noacl, "noacl"},
202 {Opt_usrquota, "usrquota"},
203 {Opt_grpquota, "grpquota"},
207 #ifdef CONFIG_DEBUG_FS
208 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
210 struct ocfs2_cluster_connection *cconn = osb->cconn;
211 struct ocfs2_recovery_map *rm = osb->recovery_map;
212 struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
215 out += snprintf(buf + out, len - out,
216 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n",
217 "Device", osb->dev_str, osb->uuid_str,
218 osb->fs_generation, osb->vol_label);
220 out += snprintf(buf + out, len - out,
221 "%10s => State: %d Flags: 0x%lX\n", "Volume",
222 atomic_read(&osb->vol_state), osb->osb_flags);
224 out += snprintf(buf + out, len - out,
225 "%10s => Block: %lu Cluster: %d\n", "Sizes",
226 osb->sb->s_blocksize, osb->s_clustersize);
228 out += snprintf(buf + out, len - out,
229 "%10s => Compat: 0x%X Incompat: 0x%X "
231 "Features", osb->s_feature_compat,
232 osb->s_feature_incompat, osb->s_feature_ro_compat);
234 out += snprintf(buf + out, len - out,
235 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount",
236 osb->s_mount_opt, osb->s_atime_quantum);
239 out += snprintf(buf + out, len - out,
240 "%10s => Stack: %s Name: %*s "
241 "Version: %d.%d\n", "Cluster",
242 (*osb->osb_cluster_stack == '\0' ?
243 "o2cb" : osb->osb_cluster_stack),
244 cconn->cc_namelen, cconn->cc_name,
245 cconn->cc_version.pv_major,
246 cconn->cc_version.pv_minor);
249 spin_lock(&osb->dc_task_lock);
250 out += snprintf(buf + out, len - out,
251 "%10s => Pid: %d Count: %lu WakeSeq: %lu "
252 "WorkSeq: %lu\n", "DownCnvt",
253 (osb->dc_task ? task_pid_nr(osb->dc_task) : -1),
254 osb->blocked_lock_count, osb->dc_wake_sequence,
255 osb->dc_work_sequence);
256 spin_unlock(&osb->dc_task_lock);
258 spin_lock(&osb->osb_lock);
259 out += snprintf(buf + out, len - out, "%10s => Pid: %d Nodes:",
261 (osb->recovery_thread_task ?
262 task_pid_nr(osb->recovery_thread_task) : -1));
263 if (rm->rm_used == 0)
264 out += snprintf(buf + out, len - out, " None\n");
266 for (i = 0; i < rm->rm_used; i++)
267 out += snprintf(buf + out, len - out, " %d",
269 out += snprintf(buf + out, len - out, "\n");
271 spin_unlock(&osb->osb_lock);
273 out += snprintf(buf + out, len - out,
274 "%10s => Pid: %d Interval: %lu Needs: %d\n", "Commit",
275 (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
276 osb->osb_commit_interval,
277 atomic_read(&osb->needs_checkpoint));
279 out += snprintf(buf + out, len - out,
280 "%10s => State: %d TxnId: %lu NumTxns: %d\n",
281 "Journal", osb->journal->j_state,
282 osb->journal->j_trans_id,
283 atomic_read(&osb->journal->j_num_trans));
285 out += snprintf(buf + out, len - out,
286 "%10s => GlobalAllocs: %d LocalAllocs: %d "
287 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
289 atomic_read(&osb->alloc_stats.bitmap_data),
290 atomic_read(&osb->alloc_stats.local_data),
291 atomic_read(&osb->alloc_stats.bg_allocs),
292 atomic_read(&osb->alloc_stats.moves),
293 atomic_read(&osb->alloc_stats.bg_extends));
295 out += snprintf(buf + out, len - out,
296 "%10s => State: %u Descriptor: %llu Size: %u bits "
297 "Default: %u bits\n",
298 "LocalAlloc", osb->local_alloc_state,
299 (unsigned long long)osb->la_last_gd,
300 osb->local_alloc_bits, osb->local_alloc_default_bits);
302 spin_lock(&osb->osb_lock);
303 out += snprintf(buf + out, len - out,
304 "%10s => Slot: %d NumStolen: %d\n", "Steal",
305 osb->s_inode_steal_slot,
306 atomic_read(&osb->s_num_inodes_stolen));
307 spin_unlock(&osb->osb_lock);
309 out += snprintf(buf + out, len - out, "OrphanScan => ");
310 out += snprintf(buf + out, len - out, "Local: %u Global: %u ",
311 os->os_count, os->os_seqno);
312 out += snprintf(buf + out, len - out, " Last Scan: ");
313 if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
314 out += snprintf(buf + out, len - out, "Disabled\n");
316 out += snprintf(buf + out, len - out, "%lu seconds ago\n",
317 (get_seconds() - os->os_scantime.tv_sec));
319 out += snprintf(buf + out, len - out, "%10s => %3s %10s\n",
320 "Slots", "Num", "RecoGen");
321 for (i = 0; i < osb->max_slots; ++i) {
322 out += snprintf(buf + out, len - out,
323 "%10s %c %3d %10d\n",
325 (i == osb->slot_num ? '*' : ' '),
326 i, osb->slot_recovery_generations[i]);
332 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
334 struct ocfs2_super *osb = inode->i_private;
337 buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
341 i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
343 file->private_data = buf;
350 static int ocfs2_debug_release(struct inode *inode, struct file *file)
352 kfree(file->private_data);
356 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
357 size_t nbytes, loff_t *ppos)
359 return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
360 i_size_read(file->f_mapping->host));
363 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
367 static int ocfs2_debug_release(struct inode *inode, struct file *file)
371 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
372 size_t nbytes, loff_t *ppos)
376 #endif /* CONFIG_DEBUG_FS */
378 static const struct file_operations ocfs2_osb_debug_fops = {
379 .open = ocfs2_osb_debug_open,
380 .release = ocfs2_debug_release,
381 .read = ocfs2_debug_read,
382 .llseek = generic_file_llseek,
385 static int ocfs2_sync_fs(struct super_block *sb, int wait)
389 struct ocfs2_super *osb = OCFS2_SB(sb);
391 if (ocfs2_is_hard_readonly(osb))
395 status = ocfs2_flush_truncate_log(osb);
399 ocfs2_schedule_truncate_log_flush(osb, 0);
402 if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
405 jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
411 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
413 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
414 && (ino == USER_QUOTA_SYSTEM_INODE
415 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
417 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
418 && (ino == GROUP_QUOTA_SYSTEM_INODE
419 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
424 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
426 struct inode *new = NULL;
432 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
434 status = PTR_ERR(new);
438 osb->root_inode = new;
440 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
442 status = PTR_ERR(new);
446 osb->sys_root_inode = new;
448 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
449 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
450 if (!ocfs2_need_system_inode(osb, i))
452 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
454 ocfs2_release_system_inodes(osb);
457 /* FIXME: Should ERROR_RO_FS */
458 mlog(ML_ERROR, "Unable to load system inode %d, "
459 "possibly corrupt fs?", i);
462 // the array now has one ref, so drop this one
471 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
473 struct inode *new = NULL;
479 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
480 i < NUM_SYSTEM_INODES;
482 if (!ocfs2_need_system_inode(osb, i))
484 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
486 ocfs2_release_system_inodes(osb);
488 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
489 status, i, osb->slot_num);
492 /* the array now has one ref, so drop this one */
501 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
508 for (i = 0; i < NUM_SYSTEM_INODES; i++) {
509 inode = osb->system_inodes[i];
512 osb->system_inodes[i] = NULL;
516 inode = osb->sys_root_inode;
519 osb->sys_root_inode = NULL;
522 inode = osb->root_inode;
525 osb->root_inode = NULL;
531 /* We're allocating fs objects, use GFP_NOFS */
532 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
534 struct ocfs2_inode_info *oi;
536 oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
540 jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
541 return &oi->vfs_inode;
544 static void ocfs2_destroy_inode(struct inode *inode)
546 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
549 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
552 unsigned int bytes = 1 << cbits;
553 unsigned int trim = bytes;
554 unsigned int bitshift = 32;
557 * i_size and all block offsets in ocfs2 are always 64 bits
558 * wide. i_clusters is 32 bits, in cluster-sized units. So on
559 * 64 bit platforms, cluster size will be the limiting factor.
562 #if BITS_PER_LONG == 32
563 # if defined(CONFIG_LBDAF)
564 BUILD_BUG_ON(sizeof(sector_t) != 8);
566 * We might be limited by page cache size.
568 if (bytes > PAGE_CACHE_SIZE) {
569 bytes = PAGE_CACHE_SIZE;
572 * Shift by 31 here so that we don't get larger than
579 * We are limited by the size of sector_t. Use block size, as
580 * that's what we expose to the VFS.
589 * Trim by a whole cluster when we can actually approach the
590 * on-disk limits. Otherwise we can overflow i_clusters when
591 * an extent start is at the max offset.
593 return (((unsigned long long)bytes) << bitshift) - trim;
596 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
598 int incompat_features;
600 struct mount_options parsed_options;
601 struct ocfs2_super *osb = OCFS2_SB(sb);
605 if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
606 !ocfs2_check_set_options(sb, &parsed_options)) {
611 if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) !=
612 (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
614 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
618 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
619 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
621 mlog(ML_ERROR, "Cannot change data mode on remount\n");
625 /* Probably don't want this on remount; it might
626 * mess with other nodes */
627 if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
628 (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
630 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
634 /* We're going to/from readonly mode. */
635 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
636 /* Disable quota accounting before remounting RO */
637 if (*flags & MS_RDONLY) {
638 ret = ocfs2_susp_quotas(osb, 0);
642 /* Lock here so the check of HARD_RO and the potential
643 * setting of SOFT_RO is atomic. */
644 spin_lock(&osb->osb_lock);
645 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
646 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
651 if (*flags & MS_RDONLY) {
652 mlog(0, "Going to ro mode.\n");
653 sb->s_flags |= MS_RDONLY;
654 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
656 mlog(0, "Making ro filesystem writeable.\n");
658 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
659 mlog(ML_ERROR, "Cannot remount RDWR "
660 "filesystem due to previous errors.\n");
664 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
665 if (incompat_features) {
666 mlog(ML_ERROR, "Cannot remount RDWR because "
667 "of unsupported optional features "
668 "(%x).\n", incompat_features);
672 sb->s_flags &= ~MS_RDONLY;
673 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
676 spin_unlock(&osb->osb_lock);
677 /* Enable quota accounting after remounting RW */
678 if (!ret && !(*flags & MS_RDONLY)) {
679 if (sb_any_quota_suspended(sb))
680 ret = ocfs2_susp_quotas(osb, 1);
682 ret = ocfs2_enable_quotas(osb);
684 /* Return back changes... */
685 spin_lock(&osb->osb_lock);
686 sb->s_flags |= MS_RDONLY;
687 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
688 spin_unlock(&osb->osb_lock);
695 /* Only save off the new mount options in case of a successful
697 osb->s_mount_opt = parsed_options.mount_opt;
698 osb->s_atime_quantum = parsed_options.atime_quantum;
699 osb->preferred_slot = parsed_options.slot;
700 if (parsed_options.commit_interval)
701 osb->osb_commit_interval = parsed_options.commit_interval;
703 if (!ocfs2_is_hard_readonly(osb))
704 ocfs2_set_journal_params(osb);
711 static int ocfs2_sb_probe(struct super_block *sb,
712 struct buffer_head **bh,
714 struct ocfs2_blockcheck_stats *stats)
717 struct ocfs1_vol_disk_hdr *hdr;
718 struct ocfs2_dinode *di;
724 *sector_size = bdev_logical_block_size(sb->s_bdev);
725 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
726 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
727 *sector_size, OCFS2_MAX_BLOCKSIZE);
732 /* Can this really happen? */
733 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
734 *sector_size = OCFS2_MIN_BLOCKSIZE;
736 /* check block zero for old format */
737 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
742 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
743 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
744 mlog(ML_ERROR, "incompatible version: %u.%u\n",
745 hdr->major_version, hdr->minor_version);
748 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
749 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
750 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
757 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
758 "upgraded before mounting with ocfs v2\n");
763 * Now check at magic offset for 512, 1024, 2048, 4096
764 * blocksizes. 4096 is the maximum blocksize because it is
765 * the minimum clustersize.
768 for (blksize = *sector_size;
769 blksize <= OCFS2_MAX_BLOCKSIZE;
771 tmpstat = ocfs2_get_sector(sb, bh,
772 OCFS2_SUPER_BLOCK_BLKNO,
779 di = (struct ocfs2_dinode *) (*bh)->b_data;
780 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
781 spin_lock_init(&stats->b_lock);
782 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
787 if (tmpstat != -EAGAIN) {
797 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
799 if (ocfs2_mount_local(osb)) {
800 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
801 mlog(ML_ERROR, "Cannot heartbeat on a locally "
802 "mounted device.\n");
807 if (ocfs2_userspace_stack(osb)) {
808 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
809 mlog(ML_ERROR, "Userspace stack expected, but "
810 "o2cb heartbeat arguments passed to mount\n");
815 if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
816 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
817 !ocfs2_userspace_stack(osb)) {
818 mlog(ML_ERROR, "Heartbeat has to be started to mount "
819 "a read-write clustered device.\n");
828 * If we're using a userspace stack, mount should have passed
829 * a name that matches the disk. If not, mount should not
830 * have passed a stack.
832 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
833 struct mount_options *mopt)
835 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
837 "cluster stack passed to mount, but this filesystem "
838 "does not support it\n");
842 if (ocfs2_userspace_stack(osb) &&
843 strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
844 OCFS2_STACK_LABEL_LEN)) {
846 "cluster stack passed to mount (\"%s\") does not "
847 "match the filesystem (\"%s\")\n",
849 osb->osb_cluster_stack);
856 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
859 struct super_block *sb = osb->sb;
860 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
861 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
864 for (type = 0; type < MAXQUOTAS; type++) {
865 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
868 status = vfs_quota_enable(
869 sb_dqopt(sb)->files[type],
873 status = vfs_quota_disable(sb, type,
879 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
880 "remount (error = %d).\n", status);
884 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
886 struct inode *inode[MAXQUOTAS] = { NULL, NULL };
887 struct super_block *sb = osb->sb;
888 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
889 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
890 unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
891 LOCAL_GROUP_QUOTA_SYSTEM_INODE };
895 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
896 for (type = 0; type < MAXQUOTAS; type++) {
897 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
899 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
905 status = vfs_quota_enable(inode[type], type, QFMT_OCFS2,
906 DQUOT_USAGE_ENABLED);
911 for (type = 0; type < MAXQUOTAS; type++)
915 ocfs2_disable_quotas(osb);
916 for (type = 0; type < MAXQUOTAS; type++)
922 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
926 struct super_block *sb = osb->sb;
928 /* We mostly ignore errors in this function because there's not much
929 * we can do when we see them */
930 for (type = 0; type < MAXQUOTAS; type++) {
931 if (!sb_has_quota_loaded(sb, type))
933 inode = igrab(sb->s_dquot.files[type]);
934 /* Turn off quotas. This will remove all dquot structures from
935 * memory and so they will be automatically synced to global
937 vfs_quota_disable(sb, type, DQUOT_USAGE_ENABLED |
938 DQUOT_LIMITS_ENABLED);
945 /* Handle quota on quotactl */
946 static int ocfs2_quota_on(struct super_block *sb, int type, int format_id,
947 char *path, int remount)
949 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
950 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
952 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
956 return 0; /* Just ignore it has been handled in
958 return vfs_quota_enable(sb_dqopt(sb)->files[type], type,
959 format_id, DQUOT_LIMITS_ENABLED);
962 /* Handle quota off quotactl */
963 static int ocfs2_quota_off(struct super_block *sb, int type, int remount)
966 return 0; /* Ignore now and handle later in
968 return vfs_quota_disable(sb, type, DQUOT_LIMITS_ENABLED);
971 static const struct quotactl_ops ocfs2_quotactl_ops = {
972 .quota_on = ocfs2_quota_on,
973 .quota_off = ocfs2_quota_off,
974 .quota_sync = vfs_quota_sync,
975 .get_info = vfs_get_dqinfo,
976 .set_info = vfs_set_dqinfo,
977 .get_dqblk = vfs_get_dqblk,
978 .set_dqblk = vfs_set_dqblk,
981 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
984 int status, sector_size;
985 struct mount_options parsed_options;
986 struct inode *inode = NULL;
987 struct ocfs2_super *osb = NULL;
988 struct buffer_head *bh = NULL;
990 struct ocfs2_blockcheck_stats stats;
992 mlog_entry("%p, %p, %i", sb, data, silent);
994 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
996 goto read_super_error;
999 /* probe for superblock */
1000 status = ocfs2_sb_probe(sb, &bh, §or_size, &stats);
1002 mlog(ML_ERROR, "superblock probe failed!\n");
1003 goto read_super_error;
1006 status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1010 goto read_super_error;
1015 if (!ocfs2_check_set_options(sb, &parsed_options)) {
1017 goto read_super_error;
1019 osb->s_mount_opt = parsed_options.mount_opt;
1020 osb->s_atime_quantum = parsed_options.atime_quantum;
1021 osb->preferred_slot = parsed_options.slot;
1022 osb->osb_commit_interval = parsed_options.commit_interval;
1023 osb->local_alloc_default_bits = ocfs2_megabytes_to_clusters(sb, parsed_options.localalloc_opt);
1024 osb->local_alloc_bits = osb->local_alloc_default_bits;
1026 status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1028 goto read_super_error;
1030 sb->s_magic = OCFS2_SUPER_MAGIC;
1032 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1033 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1035 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1037 if (bdev_read_only(sb->s_bdev)) {
1038 if (!(sb->s_flags & MS_RDONLY)) {
1040 mlog(ML_ERROR, "Readonly device detected but readonly "
1041 "mount was not specified.\n");
1042 goto read_super_error;
1045 /* You should not be able to start a local heartbeat
1046 * on a readonly device. */
1047 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1049 mlog(ML_ERROR, "Local heartbeat specified on readonly "
1051 goto read_super_error;
1054 status = ocfs2_check_journals_nolocks(osb);
1056 if (status == -EROFS)
1057 mlog(ML_ERROR, "Recovery required on readonly "
1058 "file system, but write access is "
1062 goto read_super_error;
1065 ocfs2_set_ro_flag(osb, 1);
1067 printk(KERN_NOTICE "Readonly device detected. No cluster "
1068 "services will be utilized for this mount. Recovery "
1069 "will be skipped.\n");
1072 if (!ocfs2_is_hard_readonly(osb)) {
1073 if (sb->s_flags & MS_RDONLY)
1074 ocfs2_set_ro_flag(osb, 0);
1077 status = ocfs2_verify_heartbeat(osb);
1080 goto read_super_error;
1083 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1084 ocfs2_debugfs_root);
1085 if (!osb->osb_debug_root) {
1087 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1088 goto read_super_error;
1091 osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1092 osb->osb_debug_root,
1094 &ocfs2_osb_debug_fops);
1095 if (!osb->osb_ctxt) {
1098 goto read_super_error;
1101 if (ocfs2_meta_ecc(osb)) {
1102 status = ocfs2_blockcheck_stats_debugfs_install(
1103 &osb->osb_ecc_stats,
1104 osb->osb_debug_root);
1107 "Unable to create blockcheck statistics "
1109 goto read_super_error;
1113 status = ocfs2_mount_volume(sb);
1114 if (osb->root_inode)
1115 inode = igrab(osb->root_inode);
1118 goto read_super_error;
1123 goto read_super_error;
1126 root = d_alloc_root(inode);
1130 goto read_super_error;
1135 ocfs2_complete_mount_recovery(osb);
1137 if (ocfs2_mount_local(osb))
1138 snprintf(nodestr, sizeof(nodestr), "local");
1140 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1142 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1143 "with %s data mode.\n",
1144 osb->dev_str, nodestr, osb->slot_num,
1145 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1148 atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1149 wake_up(&osb->osb_mount_event);
1151 /* Now we can initialize quotas because we can afford to wait
1152 * for cluster locks recovery now. That also means that truncation
1153 * log recovery can happen but that waits for proper quota setup */
1154 if (!(sb->s_flags & MS_RDONLY)) {
1155 status = ocfs2_enable_quotas(osb);
1157 /* We have to err-out specially here because
1158 * s_root is already set */
1160 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1161 wake_up(&osb->osb_mount_event);
1167 ocfs2_complete_quota_recovery(osb);
1169 /* Now we wake up again for processes waiting for quotas */
1170 atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1171 wake_up(&osb->osb_mount_event);
1173 /* Start this when the mount is almost sure of being successful */
1174 ocfs2_orphan_scan_start(osb);
1186 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1187 wake_up(&osb->osb_mount_event);
1188 ocfs2_dismount_volume(sb, 1);
1195 static int ocfs2_get_sb(struct file_system_type *fs_type,
1197 const char *dev_name,
1199 struct vfsmount *mnt)
1201 return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super,
1205 static void ocfs2_kill_sb(struct super_block *sb)
1207 struct ocfs2_super *osb = OCFS2_SB(sb);
1210 if (!osb || atomic_read(&osb->vol_state) == VOLUME_DISABLED)
1213 /* Prevent further queueing of inode drop events */
1214 spin_lock(&dentry_list_lock);
1215 ocfs2_set_osb_flag(osb, OCFS2_OSB_DROP_DENTRY_LOCK_IMMED);
1216 spin_unlock(&dentry_list_lock);
1217 /* Wait for work to finish and/or remove it */
1218 cancel_work_sync(&osb->dentry_lock_work);
1220 kill_block_super(sb);
1223 static struct file_system_type ocfs2_fs_type = {
1224 .owner = THIS_MODULE,
1226 .get_sb = ocfs2_get_sb, /* is this called when we mount
1228 .kill_sb = ocfs2_kill_sb,
1230 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1234 static int ocfs2_check_set_options(struct super_block *sb,
1235 struct mount_options *options)
1237 if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1238 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1239 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1240 mlog(ML_ERROR, "User quotas were requested, but this "
1241 "filesystem does not have the feature enabled.\n");
1244 if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1245 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1246 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1247 mlog(ML_ERROR, "Group quotas were requested, but this "
1248 "filesystem does not have the feature enabled.\n");
1251 if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1252 !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1253 mlog(ML_ERROR, "ACL support requested but extended attributes "
1254 "feature is not enabled\n");
1257 /* No ACL setting specified? Use XATTR feature... */
1258 if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1259 OCFS2_MOUNT_NO_POSIX_ACL))) {
1260 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1261 options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1263 options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1268 static int ocfs2_parse_options(struct super_block *sb,
1270 struct mount_options *mopt,
1276 mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
1277 options ? options : "(none)");
1279 mopt->commit_interval = 0;
1280 mopt->mount_opt = 0;
1281 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1282 mopt->slot = OCFS2_INVALID_SLOT;
1283 mopt->localalloc_opt = OCFS2_DEFAULT_LOCAL_ALLOC_SIZE;
1284 mopt->cluster_stack[0] = '\0';
1291 while ((p = strsep(&options, ",")) != NULL) {
1293 substring_t args[MAX_OPT_ARGS];
1298 token = match_token(p, tokens, args);
1301 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1304 mopt->mount_opt &= ~OCFS2_MOUNT_HB_LOCAL;
1307 if (match_int(&args[0], &option)) {
1312 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1314 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1317 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1320 mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1323 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1326 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1328 case Opt_data_ordered:
1329 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1331 case Opt_data_writeback:
1332 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1334 case Opt_user_xattr:
1335 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1337 case Opt_nouser_xattr:
1338 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1340 case Opt_atime_quantum:
1341 if (match_int(&args[0], &option)) {
1346 mopt->atime_quantum = option;
1350 if (match_int(&args[0], &option)) {
1355 mopt->slot = (s16)option;
1359 if (match_int(&args[0], &option)) {
1366 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1367 mopt->commit_interval = HZ * option;
1369 case Opt_localalloc:
1371 if (match_int(&args[0], &option)) {
1375 if (option >= 0 && (option <= ocfs2_local_alloc_size(sb) * 8))
1376 mopt->localalloc_opt = option;
1378 case Opt_localflocks:
1380 * Changing this during remount could race
1381 * flock() requests, or "unbalance" existing
1382 * ones (e.g., a lock is taken in one mode but
1383 * dropped in the other). If users care enough
1384 * to flip locking modes during remount, we
1385 * could add a "local" flag to individual
1386 * flock structures for proper tracking of
1390 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1393 /* Check both that the option we were passed
1394 * is of the right length and that it is a proper
1395 * string of the right length.
1397 if (((args[0].to - args[0].from) !=
1398 OCFS2_STACK_LABEL_LEN) ||
1399 (strnlen(args[0].from,
1400 OCFS2_STACK_LABEL_LEN) !=
1401 OCFS2_STACK_LABEL_LEN)) {
1403 "Invalid cluster_stack option\n");
1407 memcpy(mopt->cluster_stack, args[0].from,
1408 OCFS2_STACK_LABEL_LEN);
1409 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1412 mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1415 mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1418 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1421 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1422 mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1425 mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1426 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1430 "Unrecognized mount option \"%s\" "
1431 "or missing value\n", p);
1444 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1446 struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb);
1447 unsigned long opts = osb->s_mount_opt;
1448 unsigned int local_alloc_megs;
1450 if (opts & OCFS2_MOUNT_HB_LOCAL)
1451 seq_printf(s, ",_netdev,heartbeat=local");
1453 seq_printf(s, ",heartbeat=none");
1455 if (opts & OCFS2_MOUNT_NOINTR)
1456 seq_printf(s, ",nointr");
1458 if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1459 seq_printf(s, ",data=writeback");
1461 seq_printf(s, ",data=ordered");
1463 if (opts & OCFS2_MOUNT_BARRIER)
1464 seq_printf(s, ",barrier=1");
1466 if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1467 seq_printf(s, ",errors=panic");
1469 seq_printf(s, ",errors=remount-ro");
1471 if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1472 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1474 if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM)
1475 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1477 if (osb->osb_commit_interval)
1478 seq_printf(s, ",commit=%u",
1479 (unsigned) (osb->osb_commit_interval / HZ));
1481 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1482 if (local_alloc_megs != OCFS2_DEFAULT_LOCAL_ALLOC_SIZE)
1483 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1485 if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1486 seq_printf(s, ",localflocks,");
1488 if (osb->osb_cluster_stack[0])
1489 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1490 osb->osb_cluster_stack);
1491 if (opts & OCFS2_MOUNT_USRQUOTA)
1492 seq_printf(s, ",usrquota");
1493 if (opts & OCFS2_MOUNT_GRPQUOTA)
1494 seq_printf(s, ",grpquota");
1496 if (opts & OCFS2_MOUNT_NOUSERXATTR)
1497 seq_printf(s, ",nouser_xattr");
1499 seq_printf(s, ",user_xattr");
1501 if (opts & OCFS2_MOUNT_INODE64)
1502 seq_printf(s, ",inode64");
1504 if (opts & OCFS2_MOUNT_POSIX_ACL)
1505 seq_printf(s, ",acl");
1507 seq_printf(s, ",noacl");
1512 static int __init ocfs2_init(void)
1518 ocfs2_print_version();
1520 status = init_ocfs2_uptodate_cache();
1526 status = ocfs2_initialize_mem_caches();
1532 ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1538 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1539 if (!ocfs2_debugfs_root) {
1541 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1544 status = ocfs2_quota_setup();
1548 ocfs2_set_locking_protocol();
1550 status = register_quota_format(&ocfs2_quota_format);
1553 ocfs2_quota_shutdown();
1554 ocfs2_free_mem_caches();
1555 exit_ocfs2_uptodate_cache();
1561 return register_filesystem(&ocfs2_fs_type);
1566 static void __exit ocfs2_exit(void)
1570 ocfs2_quota_shutdown();
1573 flush_workqueue(ocfs2_wq);
1574 destroy_workqueue(ocfs2_wq);
1577 unregister_quota_format(&ocfs2_quota_format);
1579 debugfs_remove(ocfs2_debugfs_root);
1581 ocfs2_free_mem_caches();
1583 unregister_filesystem(&ocfs2_fs_type);
1585 exit_ocfs2_uptodate_cache();
1590 static void ocfs2_put_super(struct super_block *sb)
1592 mlog_entry("(0x%p)\n", sb);
1596 ocfs2_sync_blockdev(sb);
1597 ocfs2_dismount_volume(sb, 0);
1604 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1606 struct ocfs2_super *osb;
1607 u32 numbits, freebits;
1609 struct ocfs2_dinode *bm_lock;
1610 struct buffer_head *bh = NULL;
1611 struct inode *inode = NULL;
1613 mlog_entry("(%p, %p)\n", dentry->d_sb, buf);
1615 osb = OCFS2_SB(dentry->d_sb);
1617 inode = ocfs2_get_system_file_inode(osb,
1618 GLOBAL_BITMAP_SYSTEM_INODE,
1619 OCFS2_INVALID_SLOT);
1621 mlog(ML_ERROR, "failed to get bitmap inode\n");
1626 status = ocfs2_inode_lock(inode, &bh, 0);
1632 bm_lock = (struct ocfs2_dinode *) bh->b_data;
1634 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1635 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1637 buf->f_type = OCFS2_SUPER_MAGIC;
1638 buf->f_bsize = dentry->d_sb->s_blocksize;
1639 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1640 buf->f_blocks = ((sector_t) numbits) *
1641 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1642 buf->f_bfree = ((sector_t) freebits) *
1643 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1644 buf->f_bavail = buf->f_bfree;
1645 buf->f_files = numbits;
1646 buf->f_ffree = freebits;
1650 ocfs2_inode_unlock(inode, 0);
1661 static void ocfs2_inode_init_once(void *data)
1663 struct ocfs2_inode_info *oi = data;
1666 oi->ip_open_count = 0;
1667 spin_lock_init(&oi->ip_lock);
1668 ocfs2_extent_map_init(&oi->vfs_inode);
1669 INIT_LIST_HEAD(&oi->ip_io_markers);
1670 oi->ip_dir_start_lookup = 0;
1672 init_rwsem(&oi->ip_alloc_sem);
1673 init_rwsem(&oi->ip_xattr_sem);
1674 mutex_init(&oi->ip_io_mutex);
1676 oi->ip_blkno = 0ULL;
1677 oi->ip_clusters = 0;
1679 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1680 ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1681 ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1683 ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1684 &ocfs2_inode_caching_ops);
1686 inode_init_once(&oi->vfs_inode);
1689 static int ocfs2_initialize_mem_caches(void)
1691 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1692 sizeof(struct ocfs2_inode_info),
1694 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1696 ocfs2_inode_init_once);
1697 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1698 sizeof(struct ocfs2_dquot),
1700 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1703 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1704 sizeof(struct ocfs2_quota_chunk),
1706 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1708 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1709 !ocfs2_qf_chunk_cachep) {
1710 if (ocfs2_inode_cachep)
1711 kmem_cache_destroy(ocfs2_inode_cachep);
1712 if (ocfs2_dquot_cachep)
1713 kmem_cache_destroy(ocfs2_dquot_cachep);
1714 if (ocfs2_qf_chunk_cachep)
1715 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1722 static void ocfs2_free_mem_caches(void)
1724 if (ocfs2_inode_cachep)
1725 kmem_cache_destroy(ocfs2_inode_cachep);
1726 ocfs2_inode_cachep = NULL;
1728 if (ocfs2_dquot_cachep)
1729 kmem_cache_destroy(ocfs2_dquot_cachep);
1730 ocfs2_dquot_cachep = NULL;
1732 if (ocfs2_qf_chunk_cachep)
1733 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1734 ocfs2_qf_chunk_cachep = NULL;
1737 static int ocfs2_get_sector(struct super_block *sb,
1738 struct buffer_head **bh,
1742 if (!sb_set_blocksize(sb, sect_size)) {
1743 mlog(ML_ERROR, "unable to set blocksize\n");
1747 *bh = sb_getblk(sb, block);
1753 if (!buffer_dirty(*bh))
1754 clear_buffer_uptodate(*bh);
1756 ll_rw_block(READ, 1, bh);
1757 wait_on_buffer(*bh);
1758 if (!buffer_uptodate(*bh)) {
1768 static int ocfs2_mount_volume(struct super_block *sb)
1771 int unlock_super = 0;
1772 struct ocfs2_super *osb = OCFS2_SB(sb);
1776 if (ocfs2_is_hard_readonly(osb))
1779 status = ocfs2_dlm_init(osb);
1785 status = ocfs2_super_lock(osb, 1);
1792 /* This will load up the node map and add ourselves to it. */
1793 status = ocfs2_find_slot(osb);
1799 /* load all node-local system inodes */
1800 status = ocfs2_init_local_system_inodes(osb);
1806 status = ocfs2_check_volume(osb);
1812 status = ocfs2_truncate_log_init(osb);
1818 ocfs2_super_unlock(osb, 1);
1824 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1826 int tmp, hangup_needed = 0;
1827 struct ocfs2_super *osb = NULL;
1830 mlog_entry("(0x%p)\n", sb);
1836 debugfs_remove(osb->osb_ctxt);
1839 * Flush inode dropping work queue so that deletes are
1840 * performed while the filesystem is still working
1842 ocfs2_drop_all_dl_inodes(osb);
1844 /* Orphan scan should be stopped as early as possible */
1845 ocfs2_orphan_scan_stop(osb);
1847 ocfs2_disable_quotas(osb);
1849 ocfs2_shutdown_local_alloc(osb);
1851 ocfs2_truncate_log_shutdown(osb);
1853 /* This will disable recovery and flush any recovery work. */
1854 ocfs2_recovery_exit(osb);
1856 ocfs2_journal_shutdown(osb);
1858 ocfs2_sync_blockdev(sb);
1860 ocfs2_purge_refcount_trees(osb);
1862 /* No cluster connection means we've failed during mount, so skip
1863 * all the steps which depended on that to complete. */
1865 tmp = ocfs2_super_lock(osb, 1);
1872 if (osb->slot_num != OCFS2_INVALID_SLOT)
1873 ocfs2_put_slot(osb);
1876 ocfs2_super_unlock(osb, 1);
1878 ocfs2_release_system_inodes(osb);
1881 * If we're dismounting due to mount error, mount.ocfs2 will clean
1882 * up heartbeat. If we're a local mount, there is no heartbeat.
1883 * If we failed before we got a uuid_str yet, we can't stop
1884 * heartbeat. Otherwise, do it.
1886 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str)
1890 ocfs2_dlm_shutdown(osb, hangup_needed);
1892 ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1893 debugfs_remove(osb->osb_debug_root);
1896 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1898 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1900 if (ocfs2_mount_local(osb))
1901 snprintf(nodestr, sizeof(nodestr), "local");
1903 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1905 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1906 osb->dev_str, nodestr);
1908 ocfs2_delete_osb(osb);
1911 sb->s_fs_info = NULL;
1914 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1915 unsigned uuid_bytes)
1920 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1922 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1923 if (osb->uuid_str == NULL)
1926 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1927 /* print with null */
1928 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1929 if (ret != 2) /* drop super cleans up */
1931 /* then only advance past the last char */
1938 static int ocfs2_initialize_super(struct super_block *sb,
1939 struct buffer_head *bh,
1941 struct ocfs2_blockcheck_stats *stats)
1944 int i, cbits, bbits;
1945 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
1946 struct inode *inode = NULL;
1947 struct ocfs2_journal *journal;
1948 __le32 uuid_net_key;
1949 struct ocfs2_super *osb;
1953 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
1960 sb->s_fs_info = osb;
1961 sb->s_op = &ocfs2_sops;
1962 sb->s_export_op = &ocfs2_export_ops;
1963 sb->s_qcop = &ocfs2_quotactl_ops;
1964 sb->dq_op = &ocfs2_quota_operations;
1965 sb->s_xattr = ocfs2_xattr_handlers;
1966 sb->s_time_gran = 1;
1967 sb->s_flags |= MS_NOATIME;
1968 /* this is needed to support O_LARGEFILE */
1969 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
1970 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
1971 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
1973 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
1975 for (i = 0; i < 3; i++)
1976 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
1977 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
1980 /* Save off for ocfs2_rw_direct */
1981 osb->s_sectsize_bits = blksize_bits(sector_size);
1982 BUG_ON(!osb->s_sectsize_bits);
1984 spin_lock_init(&osb->dc_task_lock);
1985 init_waitqueue_head(&osb->dc_event);
1986 osb->dc_work_sequence = 0;
1987 osb->dc_wake_sequence = 0;
1988 INIT_LIST_HEAD(&osb->blocked_lock_list);
1989 osb->blocked_lock_count = 0;
1990 spin_lock_init(&osb->osb_lock);
1991 spin_lock_init(&osb->osb_xattr_lock);
1992 ocfs2_init_inode_steal_slot(osb);
1994 atomic_set(&osb->alloc_stats.moves, 0);
1995 atomic_set(&osb->alloc_stats.local_data, 0);
1996 atomic_set(&osb->alloc_stats.bitmap_data, 0);
1997 atomic_set(&osb->alloc_stats.bg_allocs, 0);
1998 atomic_set(&osb->alloc_stats.bg_extends, 0);
2000 /* Copy the blockcheck stats from the superblock probe */
2001 osb->osb_ecc_stats = *stats;
2003 ocfs2_init_node_maps(osb);
2005 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2006 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2008 ocfs2_orphan_scan_init(osb);
2010 status = ocfs2_recovery_init(osb);
2012 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2017 init_waitqueue_head(&osb->checkpoint_event);
2018 atomic_set(&osb->needs_checkpoint, 0);
2020 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2022 osb->slot_num = OCFS2_INVALID_SLOT;
2024 osb->s_xattr_inline_size = le16_to_cpu(
2025 di->id2.i_super.s_xattr_inline_size);
2027 osb->local_alloc_state = OCFS2_LA_UNUSED;
2028 osb->local_alloc_bh = NULL;
2029 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2031 init_waitqueue_head(&osb->osb_mount_event);
2033 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2034 if (!osb->vol_label) {
2035 mlog(ML_ERROR, "unable to alloc vol label\n");
2040 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2041 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2042 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2047 mlog(0, "max_slots for this device: %u\n", osb->max_slots);
2049 osb->slot_recovery_generations =
2050 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2052 if (!osb->slot_recovery_generations) {
2058 init_waitqueue_head(&osb->osb_wipe_event);
2059 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2060 sizeof(*osb->osb_orphan_wipes),
2062 if (!osb->osb_orphan_wipes) {
2068 osb->osb_rf_lock_tree = RB_ROOT;
2070 osb->s_feature_compat =
2071 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2072 osb->s_feature_ro_compat =
2073 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2074 osb->s_feature_incompat =
2075 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2077 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2078 mlog(ML_ERROR, "couldn't mount because of unsupported "
2079 "optional features (%x).\n", i);
2083 if (!(osb->sb->s_flags & MS_RDONLY) &&
2084 (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2085 mlog(ML_ERROR, "couldn't mount RDWR because of "
2086 "unsupported optional features (%x).\n", i);
2091 if (ocfs2_userspace_stack(osb)) {
2092 memcpy(osb->osb_cluster_stack,
2093 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2094 OCFS2_STACK_LABEL_LEN);
2095 osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
2096 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2098 "couldn't mount because of an invalid "
2099 "cluster stack label (%s) \n",
2100 osb->osb_cluster_stack);
2105 /* The empty string is identical with classic tools that
2106 * don't know about s_cluster_info. */
2107 osb->osb_cluster_stack[0] = '\0';
2110 get_random_bytes(&osb->s_next_generation, sizeof(u32));
2113 * This should be done in ocfs2_journal_init(), but unknown
2114 * ordering issues will cause the filesystem to crash.
2115 * If anyone wants to figure out what part of the code
2116 * refers to osb->journal before ocfs2_journal_init() is run,
2119 /* initialize our journal structure */
2121 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2123 mlog(ML_ERROR, "unable to alloc journal\n");
2127 osb->journal = journal;
2128 journal->j_osb = osb;
2130 atomic_set(&journal->j_num_trans, 0);
2131 init_rwsem(&journal->j_trans_barrier);
2132 init_waitqueue_head(&journal->j_checkpointed);
2133 spin_lock_init(&journal->j_lock);
2134 journal->j_trans_id = (unsigned long) 1;
2135 INIT_LIST_HEAD(&journal->j_la_cleanups);
2136 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2137 journal->j_state = OCFS2_JOURNAL_FREE;
2139 INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes);
2140 osb->dentry_lock_list = NULL;
2142 /* get some pseudo constants for clustersize bits */
2143 osb->s_clustersize_bits =
2144 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2145 osb->s_clustersize = 1 << osb->s_clustersize_bits;
2146 mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
2148 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2149 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2150 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2151 osb->s_clustersize);
2156 if (ocfs2_clusters_to_blocks(osb->sb, le32_to_cpu(di->i_clusters) - 1)
2158 mlog(ML_ERROR, "Volume might try to write to blocks beyond "
2159 "what jbd can address in 32 bits.\n");
2164 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2165 sizeof(di->id2.i_super.s_uuid))) {
2166 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2171 memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
2173 strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2174 osb->vol_label[63] = '\0';
2175 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2176 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2177 osb->first_cluster_group_blkno =
2178 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2179 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2180 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2181 mlog(0, "vol_label: %s\n", osb->vol_label);
2182 mlog(0, "uuid: %s\n", osb->uuid_str);
2183 mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
2184 (unsigned long long)osb->root_blkno,
2185 (unsigned long long)osb->system_dir_blkno);
2187 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2188 if (!osb->osb_dlm_debug) {
2194 atomic_set(&osb->vol_state, VOLUME_INIT);
2196 /* load root, system_dir, and all global system inodes */
2197 status = ocfs2_init_global_system_inodes(osb);
2206 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2207 OCFS2_INVALID_SLOT);
2214 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2217 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb) * 8;
2219 status = ocfs2_init_slot_info(osb);
2231 * will return: -EAGAIN if it is ok to keep searching for superblocks
2232 * -EINVAL if there is a bad superblock
2235 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2236 struct buffer_head *bh,
2238 struct ocfs2_blockcheck_stats *stats)
2240 int status = -EAGAIN;
2244 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2245 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2246 /* We have to do a raw check of the feature here */
2247 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2248 OCFS2_FEATURE_INCOMPAT_META_ECC) {
2249 status = ocfs2_block_check_validate(bh->b_data,
2257 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2258 mlog(ML_ERROR, "found superblock with incorrect block "
2259 "size: found %u, should be %u\n",
2260 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2262 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2263 OCFS2_MAJOR_REV_LEVEL ||
2264 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2265 OCFS2_MINOR_REV_LEVEL) {
2266 mlog(ML_ERROR, "found superblock with bad version: "
2267 "found %u.%u, should be %u.%u\n",
2268 le16_to_cpu(di->id2.i_super.s_major_rev_level),
2269 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2270 OCFS2_MAJOR_REV_LEVEL,
2271 OCFS2_MINOR_REV_LEVEL);
2272 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2273 mlog(ML_ERROR, "bad block number on superblock: "
2274 "found %llu, should be %llu\n",
2275 (unsigned long long)le64_to_cpu(di->i_blkno),
2276 (unsigned long long)bh->b_blocknr);
2277 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2278 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2279 mlog(ML_ERROR, "bad cluster size found: %u\n",
2280 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2281 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2282 mlog(ML_ERROR, "bad root_blkno: 0\n");
2283 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2284 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2285 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2287 "Superblock slots found greater than file system "
2288 "maximum: found %u, max %u\n",
2289 le16_to_cpu(di->id2.i_super.s_max_slots),
2302 static int ocfs2_check_volume(struct ocfs2_super *osb)
2307 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2313 /* Init our journal object. */
2314 status = ocfs2_journal_init(osb->journal, &dirty);
2316 mlog(ML_ERROR, "Could not initialize journal!\n");
2320 /* If the journal was unmounted cleanly then we don't want to
2321 * recover anything. Otherwise, journal_load will do that
2322 * dirty work for us :) */
2324 status = ocfs2_journal_wipe(osb->journal, 0);
2330 mlog(ML_NOTICE, "File system was not unmounted cleanly, "
2331 "recovering volume.\n");
2334 local = ocfs2_mount_local(osb);
2336 /* will play back anything left in the journal. */
2337 status = ocfs2_journal_load(osb->journal, local, dirty);
2339 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2344 /* recover my local alloc if we didn't unmount cleanly. */
2345 status = ocfs2_begin_local_alloc_recovery(osb,
2352 /* we complete the recovery process after we've marked
2353 * ourselves as mounted. */
2356 mlog(0, "Journal loaded.\n");
2358 status = ocfs2_load_local_alloc(osb);
2365 /* Recovery will be completed after we've mounted the
2366 * rest of the volume. */
2368 osb->local_alloc_copy = local_alloc;
2372 /* go through each journal, trylock it and if you get the
2373 * lock, and it's marked as dirty, set the bit in the recover
2374 * map and launch a recovery thread for it. */
2375 status = ocfs2_mark_dead_nodes(osb);
2381 status = ocfs2_compute_replay_slots(osb);
2394 * The routine gets called from dismount or close whenever a dismount on
2395 * volume is requested and the osb open count becomes 1.
2396 * It will remove the osb from the global list and also free up all the
2397 * initialized resources and fileobject.
2399 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2403 /* This function assumes that the caller has the main osb resource */
2405 ocfs2_free_slot_info(osb);
2407 kfree(osb->osb_orphan_wipes);
2408 kfree(osb->slot_recovery_generations);
2410 * This belongs in journal shutdown, but because we have to
2411 * allocate osb->journal at the start of ocfs2_initalize_osb(),
2414 kfree(osb->journal);
2415 if (osb->local_alloc_copy)
2416 kfree(osb->local_alloc_copy);
2417 kfree(osb->uuid_str);
2418 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2419 memset(osb, 0, sizeof(struct ocfs2_super));
2424 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2425 * panic(). We do not support continue-on-error operation. */
2426 static void ocfs2_handle_error(struct super_block *sb)
2428 struct ocfs2_super *osb = OCFS2_SB(sb);
2430 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2431 panic("OCFS2: (device %s): panic forced after error\n",
2434 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2436 if (sb->s_flags & MS_RDONLY &&
2437 (ocfs2_is_soft_readonly(osb) ||
2438 ocfs2_is_hard_readonly(osb)))
2441 printk(KERN_CRIT "File system is now read-only due to the potential "
2442 "of on-disk corruption. Please run fsck.ocfs2 once the file "
2443 "system is unmounted.\n");
2444 sb->s_flags |= MS_RDONLY;
2445 ocfs2_set_ro_flag(osb, 0);
2448 static char error_buf[1024];
2450 void __ocfs2_error(struct super_block *sb,
2451 const char *function,
2452 const char *fmt, ...)
2456 va_start(args, fmt);
2457 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2460 /* Not using mlog here because we want to show the actual
2461 * function the error came from. */
2462 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2463 sb->s_id, function, error_buf);
2465 ocfs2_handle_error(sb);
2468 /* Handle critical errors. This is intentionally more drastic than
2469 * ocfs2_handle_error, so we only use for things like journal errors,
2471 void __ocfs2_abort(struct super_block* sb,
2472 const char *function,
2473 const char *fmt, ...)
2477 va_start(args, fmt);
2478 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2481 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2482 sb->s_id, function, error_buf);
2484 /* We don't have the cluster support yet to go straight to
2485 * hard readonly in here. Until then, we want to keep
2486 * ocfs2_abort() so that we can at least mark critical
2489 * TODO: This should abort the journal and alert other nodes
2490 * that our slot needs recovery. */
2492 /* Force a panic(). This stinks, but it's better than letting
2493 * things continue without having a proper hard readonly
2495 if (!ocfs2_mount_local(OCFS2_SB(sb)))
2496 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2497 ocfs2_handle_error(sb);
2500 module_init(ocfs2_init);
2501 module_exit(ocfs2_exit);