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/cleancache.h>
46 #define CREATE_TRACE_POINTS
47 #include "ocfs2_trace.h"
49 #include <cluster/masklog.h>
53 /* this should be the only file to include a version 1 header */
54 #include "ocfs1_fs_compat.h"
58 #include "blockcheck.h"
61 #include "extent_map.h"
62 #include "heartbeat.h"
65 #include "localalloc.h"
73 #include "refcounttree.h"
76 #include "buffer_head_io.h"
78 static struct kmem_cache *ocfs2_inode_cachep = NULL;
79 struct kmem_cache *ocfs2_dquot_cachep;
80 struct kmem_cache *ocfs2_qf_chunk_cachep;
82 /* OCFS2 needs to schedule several different types of work which
83 * require cluster locking, disk I/O, recovery waits, etc. Since these
84 * types of work tend to be heavy we avoid using the kernel events
85 * workqueue and schedule on our own. */
86 struct workqueue_struct *ocfs2_wq = NULL;
88 static struct dentry *ocfs2_debugfs_root = NULL;
90 MODULE_AUTHOR("Oracle");
91 MODULE_LICENSE("GPL");
92 MODULE_DESCRIPTION("OCFS2 cluster file system");
96 unsigned long commit_interval;
97 unsigned long mount_opt;
98 unsigned int atime_quantum;
101 unsigned int resv_level;
103 char cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
106 static int ocfs2_parse_options(struct super_block *sb, char *options,
107 struct mount_options *mopt,
109 static int ocfs2_check_set_options(struct super_block *sb,
110 struct mount_options *options);
111 static int ocfs2_show_options(struct seq_file *s, struct dentry *root);
112 static void ocfs2_put_super(struct super_block *sb);
113 static int ocfs2_mount_volume(struct super_block *sb);
114 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
115 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
116 static int ocfs2_initialize_mem_caches(void);
117 static void ocfs2_free_mem_caches(void);
118 static void ocfs2_delete_osb(struct ocfs2_super *osb);
120 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
122 static int ocfs2_sync_fs(struct super_block *sb, int wait);
124 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
125 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
126 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
127 static int ocfs2_check_volume(struct ocfs2_super *osb);
128 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
129 struct buffer_head *bh,
131 struct ocfs2_blockcheck_stats *stats);
132 static int ocfs2_initialize_super(struct super_block *sb,
133 struct buffer_head *bh,
135 struct ocfs2_blockcheck_stats *stats);
136 static int ocfs2_get_sector(struct super_block *sb,
137 struct buffer_head **bh,
140 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
141 static void ocfs2_destroy_inode(struct inode *inode);
142 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
143 static int ocfs2_enable_quotas(struct ocfs2_super *osb);
144 static void ocfs2_disable_quotas(struct ocfs2_super *osb);
146 static const struct super_operations ocfs2_sops = {
147 .statfs = ocfs2_statfs,
148 .alloc_inode = ocfs2_alloc_inode,
149 .destroy_inode = ocfs2_destroy_inode,
150 .drop_inode = ocfs2_drop_inode,
151 .evict_inode = ocfs2_evict_inode,
152 .sync_fs = ocfs2_sync_fs,
153 .put_super = ocfs2_put_super,
154 .remount_fs = ocfs2_remount,
155 .show_options = ocfs2_show_options,
156 .quota_read = ocfs2_quota_read,
157 .quota_write = ocfs2_quota_write,
184 Opt_coherency_buffered,
191 static const match_table_t tokens = {
192 {Opt_barrier, "barrier=%u"},
193 {Opt_err_panic, "errors=panic"},
194 {Opt_err_ro, "errors=remount-ro"},
196 {Opt_nointr, "nointr"},
197 {Opt_hb_none, OCFS2_HB_NONE},
198 {Opt_hb_local, OCFS2_HB_LOCAL},
199 {Opt_hb_global, OCFS2_HB_GLOBAL},
200 {Opt_data_ordered, "data=ordered"},
201 {Opt_data_writeback, "data=writeback"},
202 {Opt_atime_quantum, "atime_quantum=%u"},
203 {Opt_slot, "preferred_slot=%u"},
204 {Opt_commit, "commit=%u"},
205 {Opt_localalloc, "localalloc=%d"},
206 {Opt_localflocks, "localflocks"},
207 {Opt_stack, "cluster_stack=%s"},
208 {Opt_user_xattr, "user_xattr"},
209 {Opt_nouser_xattr, "nouser_xattr"},
210 {Opt_inode64, "inode64"},
212 {Opt_noacl, "noacl"},
213 {Opt_usrquota, "usrquota"},
214 {Opt_grpquota, "grpquota"},
215 {Opt_coherency_buffered, "coherency=buffered"},
216 {Opt_coherency_full, "coherency=full"},
217 {Opt_resv_level, "resv_level=%u"},
218 {Opt_dir_resv_level, "dir_resv_level=%u"},
222 #ifdef CONFIG_DEBUG_FS
223 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
225 struct ocfs2_cluster_connection *cconn = osb->cconn;
226 struct ocfs2_recovery_map *rm = osb->recovery_map;
227 struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
230 out += snprintf(buf + out, len - out,
231 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n",
232 "Device", osb->dev_str, osb->uuid_str,
233 osb->fs_generation, osb->vol_label);
235 out += snprintf(buf + out, len - out,
236 "%10s => State: %d Flags: 0x%lX\n", "Volume",
237 atomic_read(&osb->vol_state), osb->osb_flags);
239 out += snprintf(buf + out, len - out,
240 "%10s => Block: %lu Cluster: %d\n", "Sizes",
241 osb->sb->s_blocksize, osb->s_clustersize);
243 out += snprintf(buf + out, len - out,
244 "%10s => Compat: 0x%X Incompat: 0x%X "
246 "Features", osb->s_feature_compat,
247 osb->s_feature_incompat, osb->s_feature_ro_compat);
249 out += snprintf(buf + out, len - out,
250 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount",
251 osb->s_mount_opt, osb->s_atime_quantum);
254 out += snprintf(buf + out, len - out,
255 "%10s => Stack: %s Name: %*s "
256 "Version: %d.%d\n", "Cluster",
257 (*osb->osb_cluster_stack == '\0' ?
258 "o2cb" : osb->osb_cluster_stack),
259 cconn->cc_namelen, cconn->cc_name,
260 cconn->cc_version.pv_major,
261 cconn->cc_version.pv_minor);
264 spin_lock(&osb->dc_task_lock);
265 out += snprintf(buf + out, len - out,
266 "%10s => Pid: %d Count: %lu WakeSeq: %lu "
267 "WorkSeq: %lu\n", "DownCnvt",
268 (osb->dc_task ? task_pid_nr(osb->dc_task) : -1),
269 osb->blocked_lock_count, osb->dc_wake_sequence,
270 osb->dc_work_sequence);
271 spin_unlock(&osb->dc_task_lock);
273 spin_lock(&osb->osb_lock);
274 out += snprintf(buf + out, len - out, "%10s => Pid: %d Nodes:",
276 (osb->recovery_thread_task ?
277 task_pid_nr(osb->recovery_thread_task) : -1));
278 if (rm->rm_used == 0)
279 out += snprintf(buf + out, len - out, " None\n");
281 for (i = 0; i < rm->rm_used; i++)
282 out += snprintf(buf + out, len - out, " %d",
284 out += snprintf(buf + out, len - out, "\n");
286 spin_unlock(&osb->osb_lock);
288 out += snprintf(buf + out, len - out,
289 "%10s => Pid: %d Interval: %lu\n", "Commit",
290 (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
291 osb->osb_commit_interval);
293 out += snprintf(buf + out, len - out,
294 "%10s => State: %d TxnId: %lu NumTxns: %d\n",
295 "Journal", osb->journal->j_state,
296 osb->journal->j_trans_id,
297 atomic_read(&osb->journal->j_num_trans));
299 out += snprintf(buf + out, len - out,
300 "%10s => GlobalAllocs: %d LocalAllocs: %d "
301 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
303 atomic_read(&osb->alloc_stats.bitmap_data),
304 atomic_read(&osb->alloc_stats.local_data),
305 atomic_read(&osb->alloc_stats.bg_allocs),
306 atomic_read(&osb->alloc_stats.moves),
307 atomic_read(&osb->alloc_stats.bg_extends));
309 out += snprintf(buf + out, len - out,
310 "%10s => State: %u Descriptor: %llu Size: %u bits "
311 "Default: %u bits\n",
312 "LocalAlloc", osb->local_alloc_state,
313 (unsigned long long)osb->la_last_gd,
314 osb->local_alloc_bits, osb->local_alloc_default_bits);
316 spin_lock(&osb->osb_lock);
317 out += snprintf(buf + out, len - out,
318 "%10s => InodeSlot: %d StolenInodes: %d, "
319 "MetaSlot: %d StolenMeta: %d\n", "Steal",
320 osb->s_inode_steal_slot,
321 atomic_read(&osb->s_num_inodes_stolen),
322 osb->s_meta_steal_slot,
323 atomic_read(&osb->s_num_meta_stolen));
324 spin_unlock(&osb->osb_lock);
326 out += snprintf(buf + out, len - out, "OrphanScan => ");
327 out += snprintf(buf + out, len - out, "Local: %u Global: %u ",
328 os->os_count, os->os_seqno);
329 out += snprintf(buf + out, len - out, " Last Scan: ");
330 if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
331 out += snprintf(buf + out, len - out, "Disabled\n");
333 out += snprintf(buf + out, len - out, "%lu seconds ago\n",
334 (get_seconds() - os->os_scantime.tv_sec));
336 out += snprintf(buf + out, len - out, "%10s => %3s %10s\n",
337 "Slots", "Num", "RecoGen");
338 for (i = 0; i < osb->max_slots; ++i) {
339 out += snprintf(buf + out, len - out,
340 "%10s %c %3d %10d\n",
342 (i == osb->slot_num ? '*' : ' '),
343 i, osb->slot_recovery_generations[i]);
349 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
351 struct ocfs2_super *osb = inode->i_private;
354 buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
358 i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
360 file->private_data = buf;
367 static int ocfs2_debug_release(struct inode *inode, struct file *file)
369 kfree(file->private_data);
373 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
374 size_t nbytes, loff_t *ppos)
376 return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
377 i_size_read(file->f_mapping->host));
380 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
384 static int ocfs2_debug_release(struct inode *inode, struct file *file)
388 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
389 size_t nbytes, loff_t *ppos)
393 #endif /* CONFIG_DEBUG_FS */
395 static const struct file_operations ocfs2_osb_debug_fops = {
396 .open = ocfs2_osb_debug_open,
397 .release = ocfs2_debug_release,
398 .read = ocfs2_debug_read,
399 .llseek = generic_file_llseek,
402 static int ocfs2_sync_fs(struct super_block *sb, int wait)
406 struct ocfs2_super *osb = OCFS2_SB(sb);
408 if (ocfs2_is_hard_readonly(osb))
412 status = ocfs2_flush_truncate_log(osb);
416 ocfs2_schedule_truncate_log_flush(osb, 0);
419 if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
422 jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
428 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
430 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
431 && (ino == USER_QUOTA_SYSTEM_INODE
432 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
434 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
435 && (ino == GROUP_QUOTA_SYSTEM_INODE
436 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
441 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
443 struct inode *new = NULL;
447 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
449 status = PTR_ERR(new);
453 osb->root_inode = new;
455 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
457 status = PTR_ERR(new);
461 osb->sys_root_inode = new;
463 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
464 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
465 if (!ocfs2_need_system_inode(osb, i))
467 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
469 ocfs2_release_system_inodes(osb);
472 /* FIXME: Should ERROR_RO_FS */
473 mlog(ML_ERROR, "Unable to load system inode %d, "
474 "possibly corrupt fs?", i);
477 // the array now has one ref, so drop this one
487 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
489 struct inode *new = NULL;
493 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
494 i < NUM_SYSTEM_INODES;
496 if (!ocfs2_need_system_inode(osb, i))
498 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
500 ocfs2_release_system_inodes(osb);
502 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
503 status, i, osb->slot_num);
506 /* the array now has one ref, so drop this one */
516 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
521 for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
522 inode = osb->global_system_inodes[i];
525 osb->global_system_inodes[i] = NULL;
529 inode = osb->sys_root_inode;
532 osb->sys_root_inode = NULL;
535 inode = osb->root_inode;
538 osb->root_inode = NULL;
541 if (!osb->local_system_inodes)
544 for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
545 if (osb->local_system_inodes[i]) {
546 iput(osb->local_system_inodes[i]);
547 osb->local_system_inodes[i] = NULL;
551 kfree(osb->local_system_inodes);
552 osb->local_system_inodes = NULL;
555 /* We're allocating fs objects, use GFP_NOFS */
556 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
558 struct ocfs2_inode_info *oi;
560 oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
564 jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
565 return &oi->vfs_inode;
568 static void ocfs2_i_callback(struct rcu_head *head)
570 struct inode *inode = container_of(head, struct inode, i_rcu);
571 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
574 static void ocfs2_destroy_inode(struct inode *inode)
576 call_rcu(&inode->i_rcu, ocfs2_i_callback);
579 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
582 unsigned int bytes = 1 << cbits;
583 unsigned int trim = bytes;
584 unsigned int bitshift = 32;
587 * i_size and all block offsets in ocfs2 are always 64 bits
588 * wide. i_clusters is 32 bits, in cluster-sized units. So on
589 * 64 bit platforms, cluster size will be the limiting factor.
592 #if BITS_PER_LONG == 32
593 # if defined(CONFIG_LBDAF)
594 BUILD_BUG_ON(sizeof(sector_t) != 8);
596 * We might be limited by page cache size.
598 if (bytes > PAGE_CACHE_SIZE) {
599 bytes = PAGE_CACHE_SIZE;
602 * Shift by 31 here so that we don't get larger than
609 * We are limited by the size of sector_t. Use block size, as
610 * that's what we expose to the VFS.
619 * Trim by a whole cluster when we can actually approach the
620 * on-disk limits. Otherwise we can overflow i_clusters when
621 * an extent start is at the max offset.
623 return (((unsigned long long)bytes) << bitshift) - trim;
626 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
628 int incompat_features;
630 struct mount_options parsed_options;
631 struct ocfs2_super *osb = OCFS2_SB(sb);
636 if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
637 !ocfs2_check_set_options(sb, &parsed_options)) {
642 tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
644 if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
646 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
650 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
651 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
653 mlog(ML_ERROR, "Cannot change data mode on remount\n");
657 /* Probably don't want this on remount; it might
658 * mess with other nodes */
659 if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
660 (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
662 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
666 /* We're going to/from readonly mode. */
667 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
668 /* Disable quota accounting before remounting RO */
669 if (*flags & MS_RDONLY) {
670 ret = ocfs2_susp_quotas(osb, 0);
674 /* Lock here so the check of HARD_RO and the potential
675 * setting of SOFT_RO is atomic. */
676 spin_lock(&osb->osb_lock);
677 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
678 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
683 if (*flags & MS_RDONLY) {
684 sb->s_flags |= MS_RDONLY;
685 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
687 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
688 mlog(ML_ERROR, "Cannot remount RDWR "
689 "filesystem due to previous errors.\n");
693 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
694 if (incompat_features) {
695 mlog(ML_ERROR, "Cannot remount RDWR because "
696 "of unsupported optional features "
697 "(%x).\n", incompat_features);
701 sb->s_flags &= ~MS_RDONLY;
702 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
704 trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags);
706 spin_unlock(&osb->osb_lock);
707 /* Enable quota accounting after remounting RW */
708 if (!ret && !(*flags & MS_RDONLY)) {
709 if (sb_any_quota_suspended(sb))
710 ret = ocfs2_susp_quotas(osb, 1);
712 ret = ocfs2_enable_quotas(osb);
714 /* Return back changes... */
715 spin_lock(&osb->osb_lock);
716 sb->s_flags |= MS_RDONLY;
717 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
718 spin_unlock(&osb->osb_lock);
725 /* Only save off the new mount options in case of a successful
727 osb->s_mount_opt = parsed_options.mount_opt;
728 osb->s_atime_quantum = parsed_options.atime_quantum;
729 osb->preferred_slot = parsed_options.slot;
730 if (parsed_options.commit_interval)
731 osb->osb_commit_interval = parsed_options.commit_interval;
733 if (!ocfs2_is_hard_readonly(osb))
734 ocfs2_set_journal_params(osb);
736 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
737 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
744 static int ocfs2_sb_probe(struct super_block *sb,
745 struct buffer_head **bh,
747 struct ocfs2_blockcheck_stats *stats)
750 struct ocfs1_vol_disk_hdr *hdr;
751 struct ocfs2_dinode *di;
757 *sector_size = bdev_logical_block_size(sb->s_bdev);
758 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
759 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
760 *sector_size, OCFS2_MAX_BLOCKSIZE);
765 /* Can this really happen? */
766 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
767 *sector_size = OCFS2_MIN_BLOCKSIZE;
769 /* check block zero for old format */
770 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
775 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
776 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
777 mlog(ML_ERROR, "incompatible version: %u.%u\n",
778 hdr->major_version, hdr->minor_version);
781 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
782 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
783 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
790 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
791 "upgraded before mounting with ocfs v2\n");
796 * Now check at magic offset for 512, 1024, 2048, 4096
797 * blocksizes. 4096 is the maximum blocksize because it is
798 * the minimum clustersize.
801 for (blksize = *sector_size;
802 blksize <= OCFS2_MAX_BLOCKSIZE;
804 tmpstat = ocfs2_get_sector(sb, bh,
805 OCFS2_SUPER_BLOCK_BLKNO,
812 di = (struct ocfs2_dinode *) (*bh)->b_data;
813 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
814 spin_lock_init(&stats->b_lock);
815 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
820 if (tmpstat != -EAGAIN) {
830 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
832 u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
834 if (osb->s_mount_opt & hb_enabled) {
835 if (ocfs2_mount_local(osb)) {
836 mlog(ML_ERROR, "Cannot heartbeat on a locally "
837 "mounted device.\n");
840 if (ocfs2_userspace_stack(osb)) {
841 mlog(ML_ERROR, "Userspace stack expected, but "
842 "o2cb heartbeat arguments passed to mount\n");
845 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
846 !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
847 ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
848 ocfs2_cluster_o2cb_global_heartbeat(osb))) {
849 mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
854 if (!(osb->s_mount_opt & hb_enabled)) {
855 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
856 !ocfs2_userspace_stack(osb)) {
857 mlog(ML_ERROR, "Heartbeat has to be started to mount "
858 "a read-write clustered device.\n");
867 * If we're using a userspace stack, mount should have passed
868 * a name that matches the disk. If not, mount should not
869 * have passed a stack.
871 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
872 struct mount_options *mopt)
874 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
876 "cluster stack passed to mount, but this filesystem "
877 "does not support it\n");
881 if (ocfs2_userspace_stack(osb) &&
882 strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
883 OCFS2_STACK_LABEL_LEN)) {
885 "cluster stack passed to mount (\"%s\") does not "
886 "match the filesystem (\"%s\")\n",
888 osb->osb_cluster_stack);
895 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
898 struct super_block *sb = osb->sb;
899 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
900 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
903 for (type = 0; type < MAXQUOTAS; type++) {
904 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
907 status = dquot_resume(sb, type);
909 struct ocfs2_mem_dqinfo *oinfo;
911 /* Cancel periodic syncing before suspending */
912 oinfo = sb_dqinfo(sb, type)->dqi_priv;
913 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
914 status = dquot_suspend(sb, type);
920 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
921 "remount (error = %d).\n", status);
925 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
927 struct inode *inode[MAXQUOTAS] = { NULL, NULL };
928 struct super_block *sb = osb->sb;
929 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
930 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
931 unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
932 LOCAL_GROUP_QUOTA_SYSTEM_INODE };
936 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
937 for (type = 0; type < MAXQUOTAS; type++) {
938 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
940 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
946 status = dquot_enable(inode[type], type, QFMT_OCFS2,
947 DQUOT_USAGE_ENABLED);
952 for (type = 0; type < MAXQUOTAS; type++)
956 ocfs2_disable_quotas(osb);
957 for (type = 0; type < MAXQUOTAS; type++)
963 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
967 struct super_block *sb = osb->sb;
968 struct ocfs2_mem_dqinfo *oinfo;
970 /* We mostly ignore errors in this function because there's not much
971 * we can do when we see them */
972 for (type = 0; type < MAXQUOTAS; type++) {
973 if (!sb_has_quota_loaded(sb, type))
975 /* Cancel periodic syncing before we grab dqonoff_mutex */
976 oinfo = sb_dqinfo(sb, type)->dqi_priv;
977 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
978 inode = igrab(sb->s_dquot.files[type]);
979 /* Turn off quotas. This will remove all dquot structures from
980 * memory and so they will be automatically synced to global
982 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
983 DQUOT_LIMITS_ENABLED);
990 /* Handle quota on quotactl */
991 static int ocfs2_quota_on(struct super_block *sb, int type, int format_id)
993 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
994 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
996 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
999 return dquot_enable(sb_dqopt(sb)->files[type], type,
1000 format_id, DQUOT_LIMITS_ENABLED);
1003 /* Handle quota off quotactl */
1004 static int ocfs2_quota_off(struct super_block *sb, int type)
1006 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
1009 static const struct quotactl_ops ocfs2_quotactl_ops = {
1010 .quota_on_meta = ocfs2_quota_on,
1011 .quota_off = ocfs2_quota_off,
1012 .quota_sync = dquot_quota_sync,
1013 .get_info = dquot_get_dqinfo,
1014 .set_info = dquot_set_dqinfo,
1015 .get_dqblk = dquot_get_dqblk,
1016 .set_dqblk = dquot_set_dqblk,
1019 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
1021 struct dentry *root;
1022 int status, sector_size;
1023 struct mount_options parsed_options;
1024 struct inode *inode = NULL;
1025 struct ocfs2_super *osb = NULL;
1026 struct buffer_head *bh = NULL;
1028 struct ocfs2_blockcheck_stats stats;
1030 trace_ocfs2_fill_super(sb, data, silent);
1032 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
1034 goto read_super_error;
1037 /* probe for superblock */
1038 status = ocfs2_sb_probe(sb, &bh, §or_size, &stats);
1040 mlog(ML_ERROR, "superblock probe failed!\n");
1041 goto read_super_error;
1044 status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1048 goto read_super_error;
1053 if (!ocfs2_check_set_options(sb, &parsed_options)) {
1055 goto read_super_error;
1057 osb->s_mount_opt = parsed_options.mount_opt;
1058 osb->s_atime_quantum = parsed_options.atime_quantum;
1059 osb->preferred_slot = parsed_options.slot;
1060 osb->osb_commit_interval = parsed_options.commit_interval;
1062 ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1063 osb->osb_resv_level = parsed_options.resv_level;
1064 osb->osb_dir_resv_level = parsed_options.resv_level;
1065 if (parsed_options.dir_resv_level == -1)
1066 osb->osb_dir_resv_level = parsed_options.resv_level;
1068 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1070 status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1072 goto read_super_error;
1074 sb->s_magic = OCFS2_SUPER_MAGIC;
1076 sb->s_flags = (sb->s_flags & ~(MS_POSIXACL | MS_NOSEC)) |
1077 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1079 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1081 if (bdev_read_only(sb->s_bdev)) {
1082 if (!(sb->s_flags & MS_RDONLY)) {
1084 mlog(ML_ERROR, "Readonly device detected but readonly "
1085 "mount was not specified.\n");
1086 goto read_super_error;
1089 /* You should not be able to start a local heartbeat
1090 * on a readonly device. */
1091 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1093 mlog(ML_ERROR, "Local heartbeat specified on readonly "
1095 goto read_super_error;
1098 status = ocfs2_check_journals_nolocks(osb);
1100 if (status == -EROFS)
1101 mlog(ML_ERROR, "Recovery required on readonly "
1102 "file system, but write access is "
1106 goto read_super_error;
1109 ocfs2_set_ro_flag(osb, 1);
1111 printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. "
1112 "Cluster services will not be used for this mount. "
1113 "Recovery will be skipped.\n", osb->dev_str);
1116 if (!ocfs2_is_hard_readonly(osb)) {
1117 if (sb->s_flags & MS_RDONLY)
1118 ocfs2_set_ro_flag(osb, 0);
1121 status = ocfs2_verify_heartbeat(osb);
1124 goto read_super_error;
1127 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1128 ocfs2_debugfs_root);
1129 if (!osb->osb_debug_root) {
1131 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1132 goto read_super_error;
1135 osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1136 osb->osb_debug_root,
1138 &ocfs2_osb_debug_fops);
1139 if (!osb->osb_ctxt) {
1142 goto read_super_error;
1145 if (ocfs2_meta_ecc(osb)) {
1146 status = ocfs2_blockcheck_stats_debugfs_install(
1147 &osb->osb_ecc_stats,
1148 osb->osb_debug_root);
1151 "Unable to create blockcheck statistics "
1153 goto read_super_error;
1157 status = ocfs2_mount_volume(sb);
1159 goto read_super_error;
1161 if (osb->root_inode)
1162 inode = igrab(osb->root_inode);
1167 goto read_super_error;
1170 root = d_make_root(inode);
1174 goto read_super_error;
1179 ocfs2_complete_mount_recovery(osb);
1181 if (ocfs2_mount_local(osb))
1182 snprintf(nodestr, sizeof(nodestr), "local");
1184 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1186 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1187 "with %s data mode.\n",
1188 osb->dev_str, nodestr, osb->slot_num,
1189 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1192 atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1193 wake_up(&osb->osb_mount_event);
1195 /* Now we can initialize quotas because we can afford to wait
1196 * for cluster locks recovery now. That also means that truncation
1197 * log recovery can happen but that waits for proper quota setup */
1198 if (!(sb->s_flags & MS_RDONLY)) {
1199 status = ocfs2_enable_quotas(osb);
1201 /* We have to err-out specially here because
1202 * s_root is already set */
1204 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1205 wake_up(&osb->osb_mount_event);
1210 ocfs2_complete_quota_recovery(osb);
1212 /* Now we wake up again for processes waiting for quotas */
1213 atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1214 wake_up(&osb->osb_mount_event);
1216 /* Start this when the mount is almost sure of being successful */
1217 ocfs2_orphan_scan_start(osb);
1225 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1226 wake_up(&osb->osb_mount_event);
1227 ocfs2_dismount_volume(sb, 1);
1235 static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
1237 const char *dev_name,
1240 return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
1243 static void ocfs2_kill_sb(struct super_block *sb)
1245 struct ocfs2_super *osb = OCFS2_SB(sb);
1248 if (!osb || atomic_read(&osb->vol_state) == VOLUME_DISABLED)
1251 /* Prevent further queueing of inode drop events */
1252 spin_lock(&dentry_list_lock);
1253 ocfs2_set_osb_flag(osb, OCFS2_OSB_DROP_DENTRY_LOCK_IMMED);
1254 spin_unlock(&dentry_list_lock);
1255 /* Wait for work to finish and/or remove it */
1256 cancel_work_sync(&osb->dentry_lock_work);
1258 kill_block_super(sb);
1261 static struct file_system_type ocfs2_fs_type = {
1262 .owner = THIS_MODULE,
1264 .mount = ocfs2_mount,
1265 .kill_sb = ocfs2_kill_sb,
1267 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1270 MODULE_ALIAS_FS("ocfs2");
1272 static int ocfs2_check_set_options(struct super_block *sb,
1273 struct mount_options *options)
1275 if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1276 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1277 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1278 mlog(ML_ERROR, "User quotas were requested, but this "
1279 "filesystem does not have the feature enabled.\n");
1282 if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1283 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1284 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1285 mlog(ML_ERROR, "Group quotas were requested, but this "
1286 "filesystem does not have the feature enabled.\n");
1289 if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1290 !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1291 mlog(ML_ERROR, "ACL support requested but extended attributes "
1292 "feature is not enabled\n");
1295 /* No ACL setting specified? Use XATTR feature... */
1296 if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1297 OCFS2_MOUNT_NO_POSIX_ACL))) {
1298 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1299 options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1301 options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1306 static int ocfs2_parse_options(struct super_block *sb,
1308 struct mount_options *mopt,
1311 int status, user_stack = 0;
1315 trace_ocfs2_parse_options(is_remount, options ? options : "(none)");
1317 mopt->commit_interval = 0;
1318 mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1319 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1320 mopt->slot = OCFS2_INVALID_SLOT;
1321 mopt->localalloc_opt = -1;
1322 mopt->cluster_stack[0] = '\0';
1323 mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1324 mopt->dir_resv_level = -1;
1331 while ((p = strsep(&options, ",")) != NULL) {
1333 substring_t args[MAX_OPT_ARGS];
1338 token = match_token(p, tokens, args);
1341 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1344 mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1347 mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1350 if (match_int(&args[0], &option)) {
1355 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1357 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1360 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1363 mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1366 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1369 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1371 case Opt_data_ordered:
1372 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1374 case Opt_data_writeback:
1375 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1377 case Opt_user_xattr:
1378 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1380 case Opt_nouser_xattr:
1381 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1383 case Opt_atime_quantum:
1384 if (match_int(&args[0], &option)) {
1389 mopt->atime_quantum = option;
1393 if (match_int(&args[0], &option)) {
1398 mopt->slot = (s16)option;
1402 if (match_int(&args[0], &option)) {
1409 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1410 mopt->commit_interval = HZ * option;
1412 case Opt_localalloc:
1414 if (match_int(&args[0], &option)) {
1419 mopt->localalloc_opt = option;
1421 case Opt_localflocks:
1423 * Changing this during remount could race
1424 * flock() requests, or "unbalance" existing
1425 * ones (e.g., a lock is taken in one mode but
1426 * dropped in the other). If users care enough
1427 * to flip locking modes during remount, we
1428 * could add a "local" flag to individual
1429 * flock structures for proper tracking of
1433 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1436 /* Check both that the option we were passed
1437 * is of the right length and that it is a proper
1438 * string of the right length.
1440 if (((args[0].to - args[0].from) !=
1441 OCFS2_STACK_LABEL_LEN) ||
1442 (strnlen(args[0].from,
1443 OCFS2_STACK_LABEL_LEN) !=
1444 OCFS2_STACK_LABEL_LEN)) {
1446 "Invalid cluster_stack option\n");
1450 memcpy(mopt->cluster_stack, args[0].from,
1451 OCFS2_STACK_LABEL_LEN);
1452 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1454 * Open code the memcmp here as we don't have
1456 * ocfs2_userspace_stack().
1458 if (memcmp(mopt->cluster_stack,
1459 OCFS2_CLASSIC_CLUSTER_STACK,
1460 OCFS2_STACK_LABEL_LEN))
1464 mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1467 mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1470 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1472 case Opt_coherency_buffered:
1473 mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1475 case Opt_coherency_full:
1476 mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1479 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1480 mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1483 mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1484 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1486 case Opt_resv_level:
1489 if (match_int(&args[0], &option)) {
1493 if (option >= OCFS2_MIN_RESV_LEVEL &&
1494 option < OCFS2_MAX_RESV_LEVEL)
1495 mopt->resv_level = option;
1497 case Opt_dir_resv_level:
1500 if (match_int(&args[0], &option)) {
1504 if (option >= OCFS2_MIN_RESV_LEVEL &&
1505 option < OCFS2_MAX_RESV_LEVEL)
1506 mopt->dir_resv_level = option;
1510 "Unrecognized mount option \"%s\" "
1511 "or missing value\n", p);
1517 if (user_stack == 0) {
1518 /* Ensure only one heartbeat mode */
1519 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1520 OCFS2_MOUNT_HB_GLOBAL |
1521 OCFS2_MOUNT_HB_NONE);
1522 if (hweight32(tmp) != 1) {
1523 mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1535 static int ocfs2_show_options(struct seq_file *s, struct dentry *root)
1537 struct ocfs2_super *osb = OCFS2_SB(root->d_sb);
1538 unsigned long opts = osb->s_mount_opt;
1539 unsigned int local_alloc_megs;
1541 if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1542 seq_printf(s, ",_netdev");
1543 if (opts & OCFS2_MOUNT_HB_LOCAL)
1544 seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1546 seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1548 seq_printf(s, ",%s", OCFS2_HB_NONE);
1550 if (opts & OCFS2_MOUNT_NOINTR)
1551 seq_printf(s, ",nointr");
1553 if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1554 seq_printf(s, ",data=writeback");
1556 seq_printf(s, ",data=ordered");
1558 if (opts & OCFS2_MOUNT_BARRIER)
1559 seq_printf(s, ",barrier=1");
1561 if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1562 seq_printf(s, ",errors=panic");
1564 seq_printf(s, ",errors=remount-ro");
1566 if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1567 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1569 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1571 if (osb->osb_commit_interval)
1572 seq_printf(s, ",commit=%u",
1573 (unsigned) (osb->osb_commit_interval / HZ));
1575 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1576 if (local_alloc_megs != ocfs2_la_default_mb(osb))
1577 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1579 if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1580 seq_printf(s, ",localflocks,");
1582 if (osb->osb_cluster_stack[0])
1583 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1584 osb->osb_cluster_stack);
1585 if (opts & OCFS2_MOUNT_USRQUOTA)
1586 seq_printf(s, ",usrquota");
1587 if (opts & OCFS2_MOUNT_GRPQUOTA)
1588 seq_printf(s, ",grpquota");
1590 if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1591 seq_printf(s, ",coherency=buffered");
1593 seq_printf(s, ",coherency=full");
1595 if (opts & OCFS2_MOUNT_NOUSERXATTR)
1596 seq_printf(s, ",nouser_xattr");
1598 seq_printf(s, ",user_xattr");
1600 if (opts & OCFS2_MOUNT_INODE64)
1601 seq_printf(s, ",inode64");
1603 if (opts & OCFS2_MOUNT_POSIX_ACL)
1604 seq_printf(s, ",acl");
1606 seq_printf(s, ",noacl");
1608 if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1609 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1611 if (osb->osb_dir_resv_level != osb->osb_resv_level)
1612 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1617 wait_queue_head_t ocfs2__ioend_wq[OCFS2_IOEND_WQ_HASH_SZ];
1619 static int __init ocfs2_init(void)
1623 for (i = 0; i < OCFS2_IOEND_WQ_HASH_SZ; i++)
1624 init_waitqueue_head(&ocfs2__ioend_wq[i]);
1626 status = init_ocfs2_uptodate_cache();
1630 status = ocfs2_initialize_mem_caches();
1634 ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1640 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1641 if (!ocfs2_debugfs_root) {
1643 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1646 ocfs2_set_locking_protocol();
1648 status = register_quota_format(&ocfs2_quota_format);
1651 status = register_filesystem(&ocfs2_fs_type);
1655 unregister_quota_format(&ocfs2_quota_format);
1657 destroy_workqueue(ocfs2_wq);
1658 debugfs_remove(ocfs2_debugfs_root);
1660 ocfs2_free_mem_caches();
1662 exit_ocfs2_uptodate_cache();
1668 static void __exit ocfs2_exit(void)
1671 flush_workqueue(ocfs2_wq);
1672 destroy_workqueue(ocfs2_wq);
1675 unregister_quota_format(&ocfs2_quota_format);
1677 debugfs_remove(ocfs2_debugfs_root);
1679 ocfs2_free_mem_caches();
1681 unregister_filesystem(&ocfs2_fs_type);
1683 exit_ocfs2_uptodate_cache();
1686 static void ocfs2_put_super(struct super_block *sb)
1688 trace_ocfs2_put_super(sb);
1690 ocfs2_sync_blockdev(sb);
1691 ocfs2_dismount_volume(sb, 0);
1694 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1696 struct ocfs2_super *osb;
1697 u32 numbits, freebits;
1699 struct ocfs2_dinode *bm_lock;
1700 struct buffer_head *bh = NULL;
1701 struct inode *inode = NULL;
1703 trace_ocfs2_statfs(dentry->d_sb, buf);
1705 osb = OCFS2_SB(dentry->d_sb);
1707 inode = ocfs2_get_system_file_inode(osb,
1708 GLOBAL_BITMAP_SYSTEM_INODE,
1709 OCFS2_INVALID_SLOT);
1711 mlog(ML_ERROR, "failed to get bitmap inode\n");
1716 status = ocfs2_inode_lock(inode, &bh, 0);
1722 bm_lock = (struct ocfs2_dinode *) bh->b_data;
1724 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1725 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1727 buf->f_type = OCFS2_SUPER_MAGIC;
1728 buf->f_bsize = dentry->d_sb->s_blocksize;
1729 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1730 buf->f_blocks = ((sector_t) numbits) *
1731 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1732 buf->f_bfree = ((sector_t) freebits) *
1733 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1734 buf->f_bavail = buf->f_bfree;
1735 buf->f_files = numbits;
1736 buf->f_ffree = freebits;
1737 buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1739 buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1740 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1744 ocfs2_inode_unlock(inode, 0);
1756 static void ocfs2_inode_init_once(void *data)
1758 struct ocfs2_inode_info *oi = data;
1761 oi->ip_open_count = 0;
1762 spin_lock_init(&oi->ip_lock);
1763 ocfs2_extent_map_init(&oi->vfs_inode);
1764 INIT_LIST_HEAD(&oi->ip_io_markers);
1765 oi->ip_dir_start_lookup = 0;
1766 atomic_set(&oi->ip_unaligned_aio, 0);
1767 init_rwsem(&oi->ip_alloc_sem);
1768 init_rwsem(&oi->ip_xattr_sem);
1769 mutex_init(&oi->ip_io_mutex);
1771 oi->ip_blkno = 0ULL;
1772 oi->ip_clusters = 0;
1774 ocfs2_resv_init_once(&oi->ip_la_data_resv);
1776 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1777 ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1778 ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1780 ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1781 &ocfs2_inode_caching_ops);
1783 inode_init_once(&oi->vfs_inode);
1786 static int ocfs2_initialize_mem_caches(void)
1788 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1789 sizeof(struct ocfs2_inode_info),
1791 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1793 ocfs2_inode_init_once);
1794 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1795 sizeof(struct ocfs2_dquot),
1797 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1800 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1801 sizeof(struct ocfs2_quota_chunk),
1803 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1805 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1806 !ocfs2_qf_chunk_cachep) {
1807 if (ocfs2_inode_cachep)
1808 kmem_cache_destroy(ocfs2_inode_cachep);
1809 if (ocfs2_dquot_cachep)
1810 kmem_cache_destroy(ocfs2_dquot_cachep);
1811 if (ocfs2_qf_chunk_cachep)
1812 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1819 static void ocfs2_free_mem_caches(void)
1822 * Make sure all delayed rcu free inodes are flushed before we
1826 if (ocfs2_inode_cachep)
1827 kmem_cache_destroy(ocfs2_inode_cachep);
1828 ocfs2_inode_cachep = NULL;
1830 if (ocfs2_dquot_cachep)
1831 kmem_cache_destroy(ocfs2_dquot_cachep);
1832 ocfs2_dquot_cachep = NULL;
1834 if (ocfs2_qf_chunk_cachep)
1835 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1836 ocfs2_qf_chunk_cachep = NULL;
1839 static int ocfs2_get_sector(struct super_block *sb,
1840 struct buffer_head **bh,
1844 if (!sb_set_blocksize(sb, sect_size)) {
1845 mlog(ML_ERROR, "unable to set blocksize\n");
1849 *bh = sb_getblk(sb, block);
1851 mlog_errno(-ENOMEM);
1855 if (!buffer_dirty(*bh))
1856 clear_buffer_uptodate(*bh);
1858 ll_rw_block(READ, 1, bh);
1859 wait_on_buffer(*bh);
1860 if (!buffer_uptodate(*bh)) {
1870 static int ocfs2_mount_volume(struct super_block *sb)
1873 int unlock_super = 0;
1874 struct ocfs2_super *osb = OCFS2_SB(sb);
1876 if (ocfs2_is_hard_readonly(osb))
1879 status = ocfs2_dlm_init(osb);
1885 status = ocfs2_super_lock(osb, 1);
1892 /* This will load up the node map and add ourselves to it. */
1893 status = ocfs2_find_slot(osb);
1899 /* load all node-local system inodes */
1900 status = ocfs2_init_local_system_inodes(osb);
1906 status = ocfs2_check_volume(osb);
1912 status = ocfs2_truncate_log_init(osb);
1918 ocfs2_super_unlock(osb, 1);
1923 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1925 int tmp, hangup_needed = 0;
1926 struct ocfs2_super *osb = NULL;
1929 trace_ocfs2_dismount_volume(sb);
1935 debugfs_remove(osb->osb_ctxt);
1938 * Flush inode dropping work queue so that deletes are
1939 * performed while the filesystem is still working
1941 ocfs2_drop_all_dl_inodes(osb);
1943 /* Orphan scan should be stopped as early as possible */
1944 ocfs2_orphan_scan_stop(osb);
1946 ocfs2_disable_quotas(osb);
1948 ocfs2_shutdown_local_alloc(osb);
1950 /* This will disable recovery and flush any recovery work. */
1951 ocfs2_recovery_exit(osb);
1954 * During dismount, when it recovers another node it will call
1955 * ocfs2_recover_orphans and queue delayed work osb_truncate_log_wq.
1957 ocfs2_truncate_log_shutdown(osb);
1959 ocfs2_journal_shutdown(osb);
1961 ocfs2_sync_blockdev(sb);
1963 ocfs2_purge_refcount_trees(osb);
1965 /* No cluster connection means we've failed during mount, so skip
1966 * all the steps which depended on that to complete. */
1968 tmp = ocfs2_super_lock(osb, 1);
1975 if (osb->slot_num != OCFS2_INVALID_SLOT)
1976 ocfs2_put_slot(osb);
1979 ocfs2_super_unlock(osb, 1);
1981 ocfs2_release_system_inodes(osb);
1984 * If we're dismounting due to mount error, mount.ocfs2 will clean
1985 * up heartbeat. If we're a local mount, there is no heartbeat.
1986 * If we failed before we got a uuid_str yet, we can't stop
1987 * heartbeat. Otherwise, do it.
1989 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str &&
1990 !ocfs2_is_hard_readonly(osb))
1994 ocfs2_dlm_shutdown(osb, hangup_needed);
1996 ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1997 debugfs_remove(osb->osb_debug_root);
2000 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
2002 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
2004 if (ocfs2_mount_local(osb))
2005 snprintf(nodestr, sizeof(nodestr), "local");
2007 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
2009 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
2010 osb->dev_str, nodestr);
2012 ocfs2_delete_osb(osb);
2015 sb->s_fs_info = NULL;
2018 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
2019 unsigned uuid_bytes)
2024 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
2026 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
2027 if (osb->uuid_str == NULL)
2030 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
2031 /* print with null */
2032 ret = snprintf(ptr, 3, "%02X", uuid[i]);
2033 if (ret != 2) /* drop super cleans up */
2035 /* then only advance past the last char */
2042 /* Make sure entire volume is addressable by our journal. Requires
2043 osb_clusters_at_boot to be valid and for the journal to have been
2044 initialized by ocfs2_journal_init(). */
2045 static int ocfs2_journal_addressable(struct ocfs2_super *osb)
2049 ocfs2_clusters_to_blocks(osb->sb,
2050 osb->osb_clusters_at_boot) - 1;
2052 /* 32-bit block number is always OK. */
2053 if (max_block <= (u32)~0ULL)
2056 /* Volume is "huge", so see if our journal is new enough to
2058 if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2059 OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2060 jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2061 JBD2_FEATURE_INCOMPAT_64BIT))) {
2062 mlog(ML_ERROR, "The journal cannot address the entire volume. "
2063 "Enable the 'block64' journal option with tunefs.ocfs2");
2072 static int ocfs2_initialize_super(struct super_block *sb,
2073 struct buffer_head *bh,
2075 struct ocfs2_blockcheck_stats *stats)
2078 int i, cbits, bbits;
2079 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2080 struct inode *inode = NULL;
2081 struct ocfs2_journal *journal;
2082 __le32 uuid_net_key;
2083 struct ocfs2_super *osb;
2086 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2093 sb->s_fs_info = osb;
2094 sb->s_op = &ocfs2_sops;
2095 sb->s_d_op = &ocfs2_dentry_ops;
2096 sb->s_export_op = &ocfs2_export_ops;
2097 sb->s_qcop = &ocfs2_quotactl_ops;
2098 sb->dq_op = &ocfs2_quota_operations;
2099 sb->s_xattr = ocfs2_xattr_handlers;
2100 sb->s_time_gran = 1;
2101 sb->s_flags |= MS_NOATIME;
2102 /* this is needed to support O_LARGEFILE */
2103 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2104 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2105 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2107 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2109 for (i = 0; i < 3; i++)
2110 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2111 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2114 /* Save off for ocfs2_rw_direct */
2115 osb->s_sectsize_bits = blksize_bits(sector_size);
2116 BUG_ON(!osb->s_sectsize_bits);
2118 spin_lock_init(&osb->dc_task_lock);
2119 init_waitqueue_head(&osb->dc_event);
2120 osb->dc_work_sequence = 0;
2121 osb->dc_wake_sequence = 0;
2122 INIT_LIST_HEAD(&osb->blocked_lock_list);
2123 osb->blocked_lock_count = 0;
2124 spin_lock_init(&osb->osb_lock);
2125 spin_lock_init(&osb->osb_xattr_lock);
2126 ocfs2_init_steal_slots(osb);
2128 atomic_set(&osb->alloc_stats.moves, 0);
2129 atomic_set(&osb->alloc_stats.local_data, 0);
2130 atomic_set(&osb->alloc_stats.bitmap_data, 0);
2131 atomic_set(&osb->alloc_stats.bg_allocs, 0);
2132 atomic_set(&osb->alloc_stats.bg_extends, 0);
2134 /* Copy the blockcheck stats from the superblock probe */
2135 osb->osb_ecc_stats = *stats;
2137 ocfs2_init_node_maps(osb);
2139 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2140 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2142 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2143 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2144 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2150 ocfs2_orphan_scan_init(osb);
2152 status = ocfs2_recovery_init(osb);
2154 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2159 init_waitqueue_head(&osb->checkpoint_event);
2161 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2163 osb->slot_num = OCFS2_INVALID_SLOT;
2165 osb->s_xattr_inline_size = le16_to_cpu(
2166 di->id2.i_super.s_xattr_inline_size);
2168 osb->local_alloc_state = OCFS2_LA_UNUSED;
2169 osb->local_alloc_bh = NULL;
2170 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2172 init_waitqueue_head(&osb->osb_mount_event);
2174 status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2180 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2181 if (!osb->vol_label) {
2182 mlog(ML_ERROR, "unable to alloc vol label\n");
2187 osb->slot_recovery_generations =
2188 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2190 if (!osb->slot_recovery_generations) {
2196 init_waitqueue_head(&osb->osb_wipe_event);
2197 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2198 sizeof(*osb->osb_orphan_wipes),
2200 if (!osb->osb_orphan_wipes) {
2206 osb->osb_rf_lock_tree = RB_ROOT;
2208 osb->s_feature_compat =
2209 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2210 osb->s_feature_ro_compat =
2211 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2212 osb->s_feature_incompat =
2213 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2215 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2216 mlog(ML_ERROR, "couldn't mount because of unsupported "
2217 "optional features (%x).\n", i);
2221 if (!(osb->sb->s_flags & MS_RDONLY) &&
2222 (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2223 mlog(ML_ERROR, "couldn't mount RDWR because of "
2224 "unsupported optional features (%x).\n", i);
2229 if (ocfs2_clusterinfo_valid(osb)) {
2230 osb->osb_stackflags =
2231 OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2232 strlcpy(osb->osb_cluster_stack,
2233 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2234 OCFS2_STACK_LABEL_LEN + 1);
2235 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2237 "couldn't mount because of an invalid "
2238 "cluster stack label (%s) \n",
2239 osb->osb_cluster_stack);
2243 strlcpy(osb->osb_cluster_name,
2244 OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster,
2245 OCFS2_CLUSTER_NAME_LEN + 1);
2247 /* The empty string is identical with classic tools that
2248 * don't know about s_cluster_info. */
2249 osb->osb_cluster_stack[0] = '\0';
2252 get_random_bytes(&osb->s_next_generation, sizeof(u32));
2255 * This should be done in ocfs2_journal_init(), but unknown
2256 * ordering issues will cause the filesystem to crash.
2257 * If anyone wants to figure out what part of the code
2258 * refers to osb->journal before ocfs2_journal_init() is run,
2261 /* initialize our journal structure */
2263 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2265 mlog(ML_ERROR, "unable to alloc journal\n");
2269 osb->journal = journal;
2270 journal->j_osb = osb;
2272 atomic_set(&journal->j_num_trans, 0);
2273 init_rwsem(&journal->j_trans_barrier);
2274 init_waitqueue_head(&journal->j_checkpointed);
2275 spin_lock_init(&journal->j_lock);
2276 journal->j_trans_id = (unsigned long) 1;
2277 INIT_LIST_HEAD(&journal->j_la_cleanups);
2278 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2279 journal->j_state = OCFS2_JOURNAL_FREE;
2281 INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes);
2282 osb->dentry_lock_list = NULL;
2284 /* get some pseudo constants for clustersize bits */
2285 osb->s_clustersize_bits =
2286 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2287 osb->s_clustersize = 1 << osb->s_clustersize_bits;
2289 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2290 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2291 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2292 osb->s_clustersize);
2297 total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2298 le32_to_cpu(di->i_clusters));
2300 status = generic_check_addressable(osb->sb->s_blocksize_bits,
2303 mlog(ML_ERROR, "Volume too large "
2304 "to mount safely on this system");
2309 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2310 sizeof(di->id2.i_super.s_uuid))) {
2311 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2316 memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
2318 strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2319 osb->vol_label[63] = '\0';
2320 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2321 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2322 osb->first_cluster_group_blkno =
2323 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2324 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2325 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2326 trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str,
2327 (unsigned long long)osb->root_blkno,
2328 (unsigned long long)osb->system_dir_blkno,
2329 osb->s_clustersize_bits);
2331 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2332 if (!osb->osb_dlm_debug) {
2338 atomic_set(&osb->vol_state, VOLUME_INIT);
2340 /* load root, system_dir, and all global system inodes */
2341 status = ocfs2_init_global_system_inodes(osb);
2350 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2351 OCFS2_INVALID_SLOT);
2358 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2359 osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2362 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2363 osb->s_feature_incompat) * 8;
2365 status = ocfs2_init_slot_info(osb);
2370 cleancache_init_shared_fs((char *)&di->id2.i_super.s_uuid, sb);
2377 * will return: -EAGAIN if it is ok to keep searching for superblocks
2378 * -EINVAL if there is a bad superblock
2381 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2382 struct buffer_head *bh,
2384 struct ocfs2_blockcheck_stats *stats)
2386 int status = -EAGAIN;
2388 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2389 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2390 /* We have to do a raw check of the feature here */
2391 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2392 OCFS2_FEATURE_INCOMPAT_META_ECC) {
2393 status = ocfs2_block_check_validate(bh->b_data,
2401 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2402 mlog(ML_ERROR, "found superblock with incorrect block "
2403 "size: found %u, should be %u\n",
2404 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2406 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2407 OCFS2_MAJOR_REV_LEVEL ||
2408 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2409 OCFS2_MINOR_REV_LEVEL) {
2410 mlog(ML_ERROR, "found superblock with bad version: "
2411 "found %u.%u, should be %u.%u\n",
2412 le16_to_cpu(di->id2.i_super.s_major_rev_level),
2413 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2414 OCFS2_MAJOR_REV_LEVEL,
2415 OCFS2_MINOR_REV_LEVEL);
2416 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2417 mlog(ML_ERROR, "bad block number on superblock: "
2418 "found %llu, should be %llu\n",
2419 (unsigned long long)le64_to_cpu(di->i_blkno),
2420 (unsigned long long)bh->b_blocknr);
2421 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2422 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2423 mlog(ML_ERROR, "bad cluster size found: %u\n",
2424 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2425 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2426 mlog(ML_ERROR, "bad root_blkno: 0\n");
2427 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2428 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2429 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2431 "Superblock slots found greater than file system "
2432 "maximum: found %u, max %u\n",
2433 le16_to_cpu(di->id2.i_super.s_max_slots),
2442 if (status && status != -EAGAIN)
2447 static int ocfs2_check_volume(struct ocfs2_super *osb)
2452 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2456 /* Init our journal object. */
2457 status = ocfs2_journal_init(osb->journal, &dirty);
2459 mlog(ML_ERROR, "Could not initialize journal!\n");
2463 /* Now that journal has been initialized, check to make sure
2464 entire volume is addressable. */
2465 status = ocfs2_journal_addressable(osb);
2469 /* If the journal was unmounted cleanly then we don't want to
2470 * recover anything. Otherwise, journal_load will do that
2471 * dirty work for us :) */
2473 status = ocfs2_journal_wipe(osb->journal, 0);
2479 printk(KERN_NOTICE "ocfs2: File system on device (%s) was not "
2480 "unmounted cleanly, recovering it.\n", osb->dev_str);
2483 local = ocfs2_mount_local(osb);
2485 /* will play back anything left in the journal. */
2486 status = ocfs2_journal_load(osb->journal, local, dirty);
2488 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2493 /* recover my local alloc if we didn't unmount cleanly. */
2494 status = ocfs2_begin_local_alloc_recovery(osb,
2501 /* we complete the recovery process after we've marked
2502 * ourselves as mounted. */
2505 status = ocfs2_load_local_alloc(osb);
2512 /* Recovery will be completed after we've mounted the
2513 * rest of the volume. */
2515 osb->local_alloc_copy = local_alloc;
2519 /* go through each journal, trylock it and if you get the
2520 * lock, and it's marked as dirty, set the bit in the recover
2521 * map and launch a recovery thread for it. */
2522 status = ocfs2_mark_dead_nodes(osb);
2528 status = ocfs2_compute_replay_slots(osb);
2541 * The routine gets called from dismount or close whenever a dismount on
2542 * volume is requested and the osb open count becomes 1.
2543 * It will remove the osb from the global list and also free up all the
2544 * initialized resources and fileobject.
2546 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2548 /* This function assumes that the caller has the main osb resource */
2550 ocfs2_free_slot_info(osb);
2552 kfree(osb->osb_orphan_wipes);
2553 kfree(osb->slot_recovery_generations);
2555 * This belongs in journal shutdown, but because we have to
2556 * allocate osb->journal at the start of ocfs2_initialize_osb(),
2559 kfree(osb->journal);
2560 kfree(osb->local_alloc_copy);
2561 kfree(osb->uuid_str);
2562 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2563 memset(osb, 0, sizeof(struct ocfs2_super));
2566 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2567 * panic(). We do not support continue-on-error operation. */
2568 static void ocfs2_handle_error(struct super_block *sb)
2570 struct ocfs2_super *osb = OCFS2_SB(sb);
2572 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2573 panic("OCFS2: (device %s): panic forced after error\n",
2576 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2578 if (sb->s_flags & MS_RDONLY &&
2579 (ocfs2_is_soft_readonly(osb) ||
2580 ocfs2_is_hard_readonly(osb)))
2583 printk(KERN_CRIT "File system is now read-only due to the potential "
2584 "of on-disk corruption. Please run fsck.ocfs2 once the file "
2585 "system is unmounted.\n");
2586 sb->s_flags |= MS_RDONLY;
2587 ocfs2_set_ro_flag(osb, 0);
2590 static char error_buf[1024];
2592 void __ocfs2_error(struct super_block *sb,
2593 const char *function,
2594 const char *fmt, ...)
2598 va_start(args, fmt);
2599 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2602 /* Not using mlog here because we want to show the actual
2603 * function the error came from. */
2604 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2605 sb->s_id, function, error_buf);
2607 ocfs2_handle_error(sb);
2610 /* Handle critical errors. This is intentionally more drastic than
2611 * ocfs2_handle_error, so we only use for things like journal errors,
2613 void __ocfs2_abort(struct super_block* sb,
2614 const char *function,
2615 const char *fmt, ...)
2619 va_start(args, fmt);
2620 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2623 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2624 sb->s_id, function, error_buf);
2626 /* We don't have the cluster support yet to go straight to
2627 * hard readonly in here. Until then, we want to keep
2628 * ocfs2_abort() so that we can at least mark critical
2631 * TODO: This should abort the journal and alert other nodes
2632 * that our slot needs recovery. */
2634 /* Force a panic(). This stinks, but it's better than letting
2635 * things continue without having a proper hard readonly
2637 if (!ocfs2_mount_local(OCFS2_SB(sb)))
2638 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2639 ocfs2_handle_error(sb);
2643 * Void signal blockers, because in-kernel sigprocmask() only fails
2644 * when SIG_* is wrong.
2646 void ocfs2_block_signals(sigset_t *oldset)
2651 sigfillset(&blocked);
2652 rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2656 void ocfs2_unblock_signals(sigset_t *oldset)
2658 int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2662 module_init(ocfs2_init);
2663 module_exit(ocfs2_exit);