1 // SPDX-License-Identifier: GPL-2.0-or-later
5 * load/unload driver, mount/dismount volumes
7 * Copyright (C) 2002, 2004 Oracle. All rights reserved.
10 #include <linux/module.h>
12 #include <linux/types.h>
13 #include <linux/slab.h>
14 #include <linux/highmem.h>
15 #include <linux/init.h>
16 #include <linux/random.h>
17 #include <linux/statfs.h>
18 #include <linux/moduleparam.h>
19 #include <linux/blkdev.h>
20 #include <linux/socket.h>
21 #include <linux/inet.h>
22 #include <linux/parser.h>
23 #include <linux/crc32.h>
24 #include <linux/debugfs.h>
25 #include <linux/mount.h>
26 #include <linux/seq_file.h>
27 #include <linux/quotaops.h>
28 #include <linux/cleancache.h>
29 #include <linux/signal.h>
31 #define CREATE_TRACE_POINTS
32 #include "ocfs2_trace.h"
34 #include <cluster/masklog.h>
38 /* this should be the only file to include a version 1 header */
39 #include "ocfs1_fs_compat.h"
43 #include "blockcheck.h"
46 #include "extent_map.h"
47 #include "heartbeat.h"
50 #include "localalloc.h"
58 #include "refcounttree.h"
61 #include "buffer_head_io.h"
62 #include "filecheck.h"
64 static struct kmem_cache *ocfs2_inode_cachep;
65 struct kmem_cache *ocfs2_dquot_cachep;
66 struct kmem_cache *ocfs2_qf_chunk_cachep;
68 static struct dentry *ocfs2_debugfs_root;
70 MODULE_AUTHOR("Oracle");
71 MODULE_LICENSE("GPL");
72 MODULE_DESCRIPTION("OCFS2 cluster file system");
76 unsigned long commit_interval;
77 unsigned long mount_opt;
78 unsigned int atime_quantum;
81 unsigned int resv_level;
83 char cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
86 static int ocfs2_parse_options(struct super_block *sb, char *options,
87 struct mount_options *mopt,
89 static int ocfs2_check_set_options(struct super_block *sb,
90 struct mount_options *options);
91 static int ocfs2_show_options(struct seq_file *s, struct dentry *root);
92 static void ocfs2_put_super(struct super_block *sb);
93 static int ocfs2_mount_volume(struct super_block *sb);
94 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
95 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
96 static int ocfs2_initialize_mem_caches(void);
97 static void ocfs2_free_mem_caches(void);
98 static void ocfs2_delete_osb(struct ocfs2_super *osb);
100 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
102 static int ocfs2_sync_fs(struct super_block *sb, int wait);
104 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
105 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
106 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
107 static int ocfs2_check_volume(struct ocfs2_super *osb);
108 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
109 struct buffer_head *bh,
111 struct ocfs2_blockcheck_stats *stats);
112 static int ocfs2_initialize_super(struct super_block *sb,
113 struct buffer_head *bh,
115 struct ocfs2_blockcheck_stats *stats);
116 static int ocfs2_get_sector(struct super_block *sb,
117 struct buffer_head **bh,
120 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
121 static void ocfs2_free_inode(struct inode *inode);
122 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
123 static int ocfs2_enable_quotas(struct ocfs2_super *osb);
124 static void ocfs2_disable_quotas(struct ocfs2_super *osb);
126 static struct dquot **ocfs2_get_dquots(struct inode *inode)
128 return OCFS2_I(inode)->i_dquot;
131 static const struct super_operations ocfs2_sops = {
132 .statfs = ocfs2_statfs,
133 .alloc_inode = ocfs2_alloc_inode,
134 .free_inode = ocfs2_free_inode,
135 .drop_inode = ocfs2_drop_inode,
136 .evict_inode = ocfs2_evict_inode,
137 .sync_fs = ocfs2_sync_fs,
138 .put_super = ocfs2_put_super,
139 .remount_fs = ocfs2_remount,
140 .show_options = ocfs2_show_options,
141 .quota_read = ocfs2_quota_read,
142 .quota_write = ocfs2_quota_write,
143 .get_dquots = ocfs2_get_dquots,
170 Opt_coherency_buffered,
174 Opt_journal_async_commit,
179 static const match_table_t tokens = {
180 {Opt_barrier, "barrier=%u"},
181 {Opt_err_panic, "errors=panic"},
182 {Opt_err_ro, "errors=remount-ro"},
184 {Opt_nointr, "nointr"},
185 {Opt_hb_none, OCFS2_HB_NONE},
186 {Opt_hb_local, OCFS2_HB_LOCAL},
187 {Opt_hb_global, OCFS2_HB_GLOBAL},
188 {Opt_data_ordered, "data=ordered"},
189 {Opt_data_writeback, "data=writeback"},
190 {Opt_atime_quantum, "atime_quantum=%u"},
191 {Opt_slot, "preferred_slot=%u"},
192 {Opt_commit, "commit=%u"},
193 {Opt_localalloc, "localalloc=%d"},
194 {Opt_localflocks, "localflocks"},
195 {Opt_stack, "cluster_stack=%s"},
196 {Opt_user_xattr, "user_xattr"},
197 {Opt_nouser_xattr, "nouser_xattr"},
198 {Opt_inode64, "inode64"},
200 {Opt_noacl, "noacl"},
201 {Opt_usrquota, "usrquota"},
202 {Opt_grpquota, "grpquota"},
203 {Opt_coherency_buffered, "coherency=buffered"},
204 {Opt_coherency_full, "coherency=full"},
205 {Opt_resv_level, "resv_level=%u"},
206 {Opt_dir_resv_level, "dir_resv_level=%u"},
207 {Opt_journal_async_commit, "journal_async_commit"},
208 {Opt_err_cont, "errors=continue"},
212 #ifdef CONFIG_DEBUG_FS
213 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
215 struct ocfs2_cluster_connection *cconn = osb->cconn;
216 struct ocfs2_recovery_map *rm = osb->recovery_map;
217 struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
221 out += scnprintf(buf + out, len - out,
222 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n",
223 "Device", osb->dev_str, osb->uuid_str,
224 osb->fs_generation, osb->vol_label);
226 out += scnprintf(buf + out, len - out,
227 "%10s => State: %d Flags: 0x%lX\n", "Volume",
228 atomic_read(&osb->vol_state), osb->osb_flags);
230 out += scnprintf(buf + out, len - out,
231 "%10s => Block: %lu Cluster: %d\n", "Sizes",
232 osb->sb->s_blocksize, osb->s_clustersize);
234 out += scnprintf(buf + out, len - out,
235 "%10s => Compat: 0x%X Incompat: 0x%X "
237 "Features", osb->s_feature_compat,
238 osb->s_feature_incompat, osb->s_feature_ro_compat);
240 out += scnprintf(buf + out, len - out,
241 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount",
242 osb->s_mount_opt, osb->s_atime_quantum);
245 out += scnprintf(buf + out, len - out,
246 "%10s => Stack: %s Name: %*s "
247 "Version: %d.%d\n", "Cluster",
248 (*osb->osb_cluster_stack == '\0' ?
249 "o2cb" : osb->osb_cluster_stack),
250 cconn->cc_namelen, cconn->cc_name,
251 cconn->cc_version.pv_major,
252 cconn->cc_version.pv_minor);
255 spin_lock_irqsave(&osb->dc_task_lock, flags);
256 out += scnprintf(buf + out, len - out,
257 "%10s => Pid: %d Count: %lu WakeSeq: %lu "
258 "WorkSeq: %lu\n", "DownCnvt",
259 (osb->dc_task ? task_pid_nr(osb->dc_task) : -1),
260 osb->blocked_lock_count, osb->dc_wake_sequence,
261 osb->dc_work_sequence);
262 spin_unlock_irqrestore(&osb->dc_task_lock, flags);
264 spin_lock(&osb->osb_lock);
265 out += scnprintf(buf + out, len - out, "%10s => Pid: %d Nodes:",
267 (osb->recovery_thread_task ?
268 task_pid_nr(osb->recovery_thread_task) : -1));
269 if (rm->rm_used == 0)
270 out += scnprintf(buf + out, len - out, " None\n");
272 for (i = 0; i < rm->rm_used; i++)
273 out += scnprintf(buf + out, len - out, " %d",
275 out += scnprintf(buf + out, len - out, "\n");
277 spin_unlock(&osb->osb_lock);
279 out += scnprintf(buf + out, len - out,
280 "%10s => Pid: %d Interval: %lu\n", "Commit",
281 (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
282 osb->osb_commit_interval);
284 out += scnprintf(buf + out, len - out,
285 "%10s => State: %d TxnId: %lu NumTxns: %d\n",
286 "Journal", osb->journal->j_state,
287 osb->journal->j_trans_id,
288 atomic_read(&osb->journal->j_num_trans));
290 out += scnprintf(buf + out, len - out,
291 "%10s => GlobalAllocs: %d LocalAllocs: %d "
292 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
294 atomic_read(&osb->alloc_stats.bitmap_data),
295 atomic_read(&osb->alloc_stats.local_data),
296 atomic_read(&osb->alloc_stats.bg_allocs),
297 atomic_read(&osb->alloc_stats.moves),
298 atomic_read(&osb->alloc_stats.bg_extends));
300 out += scnprintf(buf + out, len - out,
301 "%10s => State: %u Descriptor: %llu Size: %u bits "
302 "Default: %u bits\n",
303 "LocalAlloc", osb->local_alloc_state,
304 (unsigned long long)osb->la_last_gd,
305 osb->local_alloc_bits, osb->local_alloc_default_bits);
307 spin_lock(&osb->osb_lock);
308 out += scnprintf(buf + out, len - out,
309 "%10s => InodeSlot: %d StolenInodes: %d, "
310 "MetaSlot: %d StolenMeta: %d\n", "Steal",
311 osb->s_inode_steal_slot,
312 atomic_read(&osb->s_num_inodes_stolen),
313 osb->s_meta_steal_slot,
314 atomic_read(&osb->s_num_meta_stolen));
315 spin_unlock(&osb->osb_lock);
317 out += scnprintf(buf + out, len - out, "OrphanScan => ");
318 out += scnprintf(buf + out, len - out, "Local: %u Global: %u ",
319 os->os_count, os->os_seqno);
320 out += scnprintf(buf + out, len - out, " Last Scan: ");
321 if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
322 out += scnprintf(buf + out, len - out, "Disabled\n");
324 out += scnprintf(buf + out, len - out, "%lu seconds ago\n",
325 (unsigned long)(ktime_get_seconds() - os->os_scantime));
327 out += scnprintf(buf + out, len - out, "%10s => %3s %10s\n",
328 "Slots", "Num", "RecoGen");
329 for (i = 0; i < osb->max_slots; ++i) {
330 out += scnprintf(buf + out, len - out,
331 "%10s %c %3d %10d\n",
333 (i == osb->slot_num ? '*' : ' '),
334 i, osb->slot_recovery_generations[i]);
340 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
342 struct ocfs2_super *osb = inode->i_private;
345 buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
349 i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
351 file->private_data = buf;
358 static int ocfs2_debug_release(struct inode *inode, struct file *file)
360 kfree(file->private_data);
364 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
365 size_t nbytes, loff_t *ppos)
367 return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
368 i_size_read(file->f_mapping->host));
371 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
375 static int ocfs2_debug_release(struct inode *inode, struct file *file)
379 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
380 size_t nbytes, loff_t *ppos)
384 #endif /* CONFIG_DEBUG_FS */
386 static const struct file_operations ocfs2_osb_debug_fops = {
387 .open = ocfs2_osb_debug_open,
388 .release = ocfs2_debug_release,
389 .read = ocfs2_debug_read,
390 .llseek = generic_file_llseek,
393 static int ocfs2_sync_fs(struct super_block *sb, int wait)
397 struct ocfs2_super *osb = OCFS2_SB(sb);
399 if (ocfs2_is_hard_readonly(osb))
403 status = ocfs2_flush_truncate_log(osb);
407 ocfs2_schedule_truncate_log_flush(osb, 0);
410 if (jbd2_journal_start_commit(osb->journal->j_journal,
413 jbd2_log_wait_commit(osb->journal->j_journal,
419 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
421 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
422 && (ino == USER_QUOTA_SYSTEM_INODE
423 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
425 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
426 && (ino == GROUP_QUOTA_SYSTEM_INODE
427 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
432 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
434 struct inode *new = NULL;
438 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
440 status = PTR_ERR(new);
444 osb->root_inode = new;
446 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
448 status = PTR_ERR(new);
452 osb->sys_root_inode = new;
454 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
455 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
456 if (!ocfs2_need_system_inode(osb, i))
458 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
460 ocfs2_release_system_inodes(osb);
461 status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
463 mlog(ML_ERROR, "Unable to load system inode %d, "
464 "possibly corrupt fs?", i);
467 // the array now has one ref, so drop this one
477 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
479 struct inode *new = NULL;
483 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
484 i < NUM_SYSTEM_INODES;
486 if (!ocfs2_need_system_inode(osb, i))
488 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
490 ocfs2_release_system_inodes(osb);
491 status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
492 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
493 status, i, osb->slot_num);
496 /* the array now has one ref, so drop this one */
506 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
511 for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
512 inode = osb->global_system_inodes[i];
515 osb->global_system_inodes[i] = NULL;
519 inode = osb->sys_root_inode;
522 osb->sys_root_inode = NULL;
525 inode = osb->root_inode;
528 osb->root_inode = NULL;
531 if (!osb->local_system_inodes)
534 for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
535 if (osb->local_system_inodes[i]) {
536 iput(osb->local_system_inodes[i]);
537 osb->local_system_inodes[i] = NULL;
541 kfree(osb->local_system_inodes);
542 osb->local_system_inodes = NULL;
545 /* We're allocating fs objects, use GFP_NOFS */
546 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
548 struct ocfs2_inode_info *oi;
550 oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
555 oi->i_datasync_tid = 0;
556 memset(&oi->i_dquot, 0, sizeof(oi->i_dquot));
558 jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
559 return &oi->vfs_inode;
562 static void ocfs2_free_inode(struct inode *inode)
564 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
567 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
570 unsigned int bytes = 1 << cbits;
571 unsigned int trim = bytes;
572 unsigned int bitshift = 32;
575 * i_size and all block offsets in ocfs2 are always 64 bits
576 * wide. i_clusters is 32 bits, in cluster-sized units. So on
577 * 64 bit platforms, cluster size will be the limiting factor.
580 #if BITS_PER_LONG == 32
581 BUILD_BUG_ON(sizeof(sector_t) != 8);
583 * We might be limited by page cache size.
585 if (bytes > PAGE_SIZE) {
589 * Shift by 31 here so that we don't get larger than
597 * Trim by a whole cluster when we can actually approach the
598 * on-disk limits. Otherwise we can overflow i_clusters when
599 * an extent start is at the max offset.
601 return (((unsigned long long)bytes) << bitshift) - trim;
604 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
606 int incompat_features;
608 struct mount_options parsed_options;
609 struct ocfs2_super *osb = OCFS2_SB(sb);
614 if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
615 !ocfs2_check_set_options(sb, &parsed_options)) {
620 tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
622 if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
624 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
628 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
629 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
631 mlog(ML_ERROR, "Cannot change data mode on remount\n");
635 /* Probably don't want this on remount; it might
636 * mess with other nodes */
637 if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
638 (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
640 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
644 /* We're going to/from readonly mode. */
645 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
646 /* Disable quota accounting before remounting RO */
647 if (*flags & SB_RDONLY) {
648 ret = ocfs2_susp_quotas(osb, 0);
652 /* Lock here so the check of HARD_RO and the potential
653 * setting of SOFT_RO is atomic. */
654 spin_lock(&osb->osb_lock);
655 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
656 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
661 if (*flags & SB_RDONLY) {
662 sb->s_flags |= SB_RDONLY;
663 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
665 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
666 mlog(ML_ERROR, "Cannot remount RDWR "
667 "filesystem due to previous errors.\n");
671 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
672 if (incompat_features) {
673 mlog(ML_ERROR, "Cannot remount RDWR because "
674 "of unsupported optional features "
675 "(%x).\n", incompat_features);
679 sb->s_flags &= ~SB_RDONLY;
680 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
682 trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags);
684 spin_unlock(&osb->osb_lock);
685 /* Enable quota accounting after remounting RW */
686 if (!ret && !(*flags & SB_RDONLY)) {
687 if (sb_any_quota_suspended(sb))
688 ret = ocfs2_susp_quotas(osb, 1);
690 ret = ocfs2_enable_quotas(osb);
692 /* Return back changes... */
693 spin_lock(&osb->osb_lock);
694 sb->s_flags |= SB_RDONLY;
695 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
696 spin_unlock(&osb->osb_lock);
703 /* Only save off the new mount options in case of a successful
705 osb->s_mount_opt = parsed_options.mount_opt;
706 osb->s_atime_quantum = parsed_options.atime_quantum;
707 osb->preferred_slot = parsed_options.slot;
708 if (parsed_options.commit_interval)
709 osb->osb_commit_interval = parsed_options.commit_interval;
711 if (!ocfs2_is_hard_readonly(osb))
712 ocfs2_set_journal_params(osb);
714 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
715 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
722 static int ocfs2_sb_probe(struct super_block *sb,
723 struct buffer_head **bh,
725 struct ocfs2_blockcheck_stats *stats)
728 struct ocfs1_vol_disk_hdr *hdr;
729 struct ocfs2_dinode *di;
735 *sector_size = bdev_logical_block_size(sb->s_bdev);
736 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
737 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
738 *sector_size, OCFS2_MAX_BLOCKSIZE);
743 /* Can this really happen? */
744 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
745 *sector_size = OCFS2_MIN_BLOCKSIZE;
747 /* check block zero for old format */
748 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
753 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
754 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
755 mlog(ML_ERROR, "incompatible version: %u.%u\n",
756 hdr->major_version, hdr->minor_version);
759 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
760 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
761 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
768 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
769 "upgraded before mounting with ocfs v2\n");
774 * Now check at magic offset for 512, 1024, 2048, 4096
775 * blocksizes. 4096 is the maximum blocksize because it is
776 * the minimum clustersize.
779 for (blksize = *sector_size;
780 blksize <= OCFS2_MAX_BLOCKSIZE;
782 tmpstat = ocfs2_get_sector(sb, bh,
783 OCFS2_SUPER_BLOCK_BLKNO,
790 di = (struct ocfs2_dinode *) (*bh)->b_data;
791 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
792 spin_lock_init(&stats->b_lock);
793 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
798 if (tmpstat != -EAGAIN) {
808 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
810 u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
812 if (osb->s_mount_opt & hb_enabled) {
813 if (ocfs2_mount_local(osb)) {
814 mlog(ML_ERROR, "Cannot heartbeat on a locally "
815 "mounted device.\n");
818 if (ocfs2_userspace_stack(osb)) {
819 mlog(ML_ERROR, "Userspace stack expected, but "
820 "o2cb heartbeat arguments passed to mount\n");
823 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
824 !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
825 ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
826 ocfs2_cluster_o2cb_global_heartbeat(osb))) {
827 mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
832 if (!(osb->s_mount_opt & hb_enabled)) {
833 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
834 !ocfs2_userspace_stack(osb)) {
835 mlog(ML_ERROR, "Heartbeat has to be started to mount "
836 "a read-write clustered device.\n");
845 * If we're using a userspace stack, mount should have passed
846 * a name that matches the disk. If not, mount should not
847 * have passed a stack.
849 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
850 struct mount_options *mopt)
852 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
854 "cluster stack passed to mount, but this filesystem "
855 "does not support it\n");
859 if (ocfs2_userspace_stack(osb) &&
860 strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
861 OCFS2_STACK_LABEL_LEN)) {
863 "cluster stack passed to mount (\"%s\") does not "
864 "match the filesystem (\"%s\")\n",
866 osb->osb_cluster_stack);
873 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
876 struct super_block *sb = osb->sb;
877 unsigned int feature[OCFS2_MAXQUOTAS] = {
878 OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
879 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
882 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
883 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
886 status = dquot_resume(sb, type);
888 struct ocfs2_mem_dqinfo *oinfo;
890 /* Cancel periodic syncing before suspending */
891 oinfo = sb_dqinfo(sb, type)->dqi_priv;
892 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
893 status = dquot_suspend(sb, type);
899 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
900 "remount (error = %d).\n", status);
904 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
906 struct inode *inode[OCFS2_MAXQUOTAS] = { NULL, NULL };
907 struct super_block *sb = osb->sb;
908 unsigned int feature[OCFS2_MAXQUOTAS] = {
909 OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
910 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
911 unsigned int ino[OCFS2_MAXQUOTAS] = {
912 LOCAL_USER_QUOTA_SYSTEM_INODE,
913 LOCAL_GROUP_QUOTA_SYSTEM_INODE };
917 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
918 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
919 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
921 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
927 status = dquot_load_quota_inode(inode[type], type, QFMT_OCFS2,
928 DQUOT_USAGE_ENABLED);
933 for (type = 0; type < OCFS2_MAXQUOTAS; type++)
937 ocfs2_disable_quotas(osb);
938 for (type = 0; type < OCFS2_MAXQUOTAS; type++)
944 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
948 struct super_block *sb = osb->sb;
949 struct ocfs2_mem_dqinfo *oinfo;
951 /* We mostly ignore errors in this function because there's not much
952 * we can do when we see them */
953 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
954 if (!sb_has_quota_loaded(sb, type))
956 oinfo = sb_dqinfo(sb, type)->dqi_priv;
957 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
958 inode = igrab(sb->s_dquot.files[type]);
959 /* Turn off quotas. This will remove all dquot structures from
960 * memory and so they will be automatically synced to global
962 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
963 DQUOT_LIMITS_ENABLED);
968 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
971 int status, sector_size;
972 struct mount_options parsed_options;
973 struct inode *inode = NULL;
974 struct ocfs2_super *osb = NULL;
975 struct buffer_head *bh = NULL;
977 struct ocfs2_blockcheck_stats stats;
979 trace_ocfs2_fill_super(sb, data, silent);
981 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
983 goto read_super_error;
986 /* probe for superblock */
987 status = ocfs2_sb_probe(sb, &bh, §or_size, &stats);
989 mlog(ML_ERROR, "superblock probe failed!\n");
990 goto read_super_error;
993 status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
997 goto read_super_error;
1002 if (!ocfs2_check_set_options(sb, &parsed_options)) {
1004 goto read_super_error;
1006 osb->s_mount_opt = parsed_options.mount_opt;
1007 osb->s_atime_quantum = parsed_options.atime_quantum;
1008 osb->preferred_slot = parsed_options.slot;
1009 osb->osb_commit_interval = parsed_options.commit_interval;
1011 ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1012 osb->osb_resv_level = parsed_options.resv_level;
1013 osb->osb_dir_resv_level = parsed_options.resv_level;
1014 if (parsed_options.dir_resv_level == -1)
1015 osb->osb_dir_resv_level = parsed_options.resv_level;
1017 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1019 status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1021 goto read_super_error;
1023 sb->s_magic = OCFS2_SUPER_MAGIC;
1025 sb->s_flags = (sb->s_flags & ~(SB_POSIXACL | SB_NOSEC)) |
1026 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? SB_POSIXACL : 0);
1028 /* Hard readonly mode only if: bdev_read_only, SB_RDONLY,
1030 if (bdev_read_only(sb->s_bdev)) {
1031 if (!sb_rdonly(sb)) {
1033 mlog(ML_ERROR, "Readonly device detected but readonly "
1034 "mount was not specified.\n");
1035 goto read_super_error;
1038 /* You should not be able to start a local heartbeat
1039 * on a readonly device. */
1040 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1042 mlog(ML_ERROR, "Local heartbeat specified on readonly "
1044 goto read_super_error;
1047 status = ocfs2_check_journals_nolocks(osb);
1049 if (status == -EROFS)
1050 mlog(ML_ERROR, "Recovery required on readonly "
1051 "file system, but write access is "
1055 goto read_super_error;
1058 ocfs2_set_ro_flag(osb, 1);
1060 printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. "
1061 "Cluster services will not be used for this mount. "
1062 "Recovery will be skipped.\n", osb->dev_str);
1065 if (!ocfs2_is_hard_readonly(osb)) {
1067 ocfs2_set_ro_flag(osb, 0);
1070 status = ocfs2_verify_heartbeat(osb);
1073 goto read_super_error;
1076 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1077 ocfs2_debugfs_root);
1079 debugfs_create_file("fs_state", S_IFREG|S_IRUSR, osb->osb_debug_root,
1080 osb, &ocfs2_osb_debug_fops);
1082 if (ocfs2_meta_ecc(osb))
1083 ocfs2_blockcheck_stats_debugfs_install( &osb->osb_ecc_stats,
1084 osb->osb_debug_root);
1086 status = ocfs2_mount_volume(sb);
1088 goto read_super_error;
1090 if (osb->root_inode)
1091 inode = igrab(osb->root_inode);
1096 goto read_super_error;
1099 osb->osb_dev_kset = kset_create_and_add(sb->s_id, NULL,
1101 if (!osb->osb_dev_kset) {
1103 mlog(ML_ERROR, "Unable to create device kset %s.\n", sb->s_id);
1104 goto read_super_error;
1107 /* Create filecheck sysfs related directories/files at
1108 * /sys/fs/ocfs2/<devname>/filecheck */
1109 if (ocfs2_filecheck_create_sysfs(osb)) {
1111 mlog(ML_ERROR, "Unable to create filecheck sysfs directory at "
1112 "/sys/fs/ocfs2/%s/filecheck.\n", sb->s_id);
1113 goto read_super_error;
1116 root = d_make_root(inode);
1120 goto read_super_error;
1125 ocfs2_complete_mount_recovery(osb);
1127 if (ocfs2_mount_local(osb))
1128 snprintf(nodestr, sizeof(nodestr), "local");
1130 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1132 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1133 "with %s data mode.\n",
1134 osb->dev_str, nodestr, osb->slot_num,
1135 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1138 atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1139 wake_up(&osb->osb_mount_event);
1141 /* Now we can initialize quotas because we can afford to wait
1142 * for cluster locks recovery now. That also means that truncation
1143 * log recovery can happen but that waits for proper quota setup */
1144 if (!sb_rdonly(sb)) {
1145 status = ocfs2_enable_quotas(osb);
1147 /* We have to err-out specially here because
1148 * s_root is already set */
1150 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1151 wake_up(&osb->osb_mount_event);
1156 ocfs2_complete_quota_recovery(osb);
1158 /* Now we wake up again for processes waiting for quotas */
1159 atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1160 wake_up(&osb->osb_mount_event);
1162 /* Start this when the mount is almost sure of being successful */
1163 ocfs2_orphan_scan_start(osb);
1174 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1175 wake_up(&osb->osb_mount_event);
1176 ocfs2_dismount_volume(sb, 1);
1182 static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
1184 const char *dev_name,
1187 return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
1190 static struct file_system_type ocfs2_fs_type = {
1191 .owner = THIS_MODULE,
1193 .mount = ocfs2_mount,
1194 .kill_sb = kill_block_super,
1195 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1198 MODULE_ALIAS_FS("ocfs2");
1200 static int ocfs2_check_set_options(struct super_block *sb,
1201 struct mount_options *options)
1203 if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1204 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1205 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1206 mlog(ML_ERROR, "User quotas were requested, but this "
1207 "filesystem does not have the feature enabled.\n");
1210 if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1211 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1212 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1213 mlog(ML_ERROR, "Group quotas were requested, but this "
1214 "filesystem does not have the feature enabled.\n");
1217 if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1218 !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1219 mlog(ML_ERROR, "ACL support requested but extended attributes "
1220 "feature is not enabled\n");
1223 /* No ACL setting specified? Use XATTR feature... */
1224 if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1225 OCFS2_MOUNT_NO_POSIX_ACL))) {
1226 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1227 options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1229 options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1234 static int ocfs2_parse_options(struct super_block *sb,
1236 struct mount_options *mopt,
1239 int status, user_stack = 0;
1243 substring_t args[MAX_OPT_ARGS];
1245 trace_ocfs2_parse_options(is_remount, options ? options : "(none)");
1247 mopt->commit_interval = 0;
1248 mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1249 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1250 mopt->slot = OCFS2_INVALID_SLOT;
1251 mopt->localalloc_opt = -1;
1252 mopt->cluster_stack[0] = '\0';
1253 mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1254 mopt->dir_resv_level = -1;
1261 while ((p = strsep(&options, ",")) != NULL) {
1265 token = match_token(p, tokens, args);
1268 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1271 mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1274 mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1277 if (match_int(&args[0], &option)) {
1282 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1284 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1287 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1290 mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1293 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1294 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1295 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1298 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1299 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1300 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_ROFS;
1303 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1304 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1305 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_CONT;
1307 case Opt_data_ordered:
1308 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1310 case Opt_data_writeback:
1311 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1313 case Opt_user_xattr:
1314 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1316 case Opt_nouser_xattr:
1317 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1319 case Opt_atime_quantum:
1320 if (match_int(&args[0], &option)) {
1325 mopt->atime_quantum = option;
1328 if (match_int(&args[0], &option)) {
1333 mopt->slot = (u16)option;
1336 if (match_int(&args[0], &option)) {
1343 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1344 mopt->commit_interval = HZ * option;
1346 case Opt_localalloc:
1347 if (match_int(&args[0], &option)) {
1352 mopt->localalloc_opt = option;
1354 case Opt_localflocks:
1356 * Changing this during remount could race
1357 * flock() requests, or "unbalance" existing
1358 * ones (e.g., a lock is taken in one mode but
1359 * dropped in the other). If users care enough
1360 * to flip locking modes during remount, we
1361 * could add a "local" flag to individual
1362 * flock structures for proper tracking of
1366 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1369 /* Check both that the option we were passed
1370 * is of the right length and that it is a proper
1371 * string of the right length.
1373 if (((args[0].to - args[0].from) !=
1374 OCFS2_STACK_LABEL_LEN) ||
1375 (strnlen(args[0].from,
1376 OCFS2_STACK_LABEL_LEN) !=
1377 OCFS2_STACK_LABEL_LEN)) {
1379 "Invalid cluster_stack option\n");
1383 memcpy(mopt->cluster_stack, args[0].from,
1384 OCFS2_STACK_LABEL_LEN);
1385 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1387 * Open code the memcmp here as we don't have
1389 * ocfs2_userspace_stack().
1391 if (memcmp(mopt->cluster_stack,
1392 OCFS2_CLASSIC_CLUSTER_STACK,
1393 OCFS2_STACK_LABEL_LEN))
1397 mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1400 mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1403 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1405 case Opt_coherency_buffered:
1406 mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1408 case Opt_coherency_full:
1409 mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1412 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1413 mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1416 mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1417 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1419 case Opt_resv_level:
1422 if (match_int(&args[0], &option)) {
1426 if (option >= OCFS2_MIN_RESV_LEVEL &&
1427 option < OCFS2_MAX_RESV_LEVEL)
1428 mopt->resv_level = option;
1430 case Opt_dir_resv_level:
1433 if (match_int(&args[0], &option)) {
1437 if (option >= OCFS2_MIN_RESV_LEVEL &&
1438 option < OCFS2_MAX_RESV_LEVEL)
1439 mopt->dir_resv_level = option;
1441 case Opt_journal_async_commit:
1442 mopt->mount_opt |= OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT;
1446 "Unrecognized mount option \"%s\" "
1447 "or missing value\n", p);
1453 if (user_stack == 0) {
1454 /* Ensure only one heartbeat mode */
1455 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1456 OCFS2_MOUNT_HB_GLOBAL |
1457 OCFS2_MOUNT_HB_NONE);
1458 if (hweight32(tmp) != 1) {
1459 mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1471 static int ocfs2_show_options(struct seq_file *s, struct dentry *root)
1473 struct ocfs2_super *osb = OCFS2_SB(root->d_sb);
1474 unsigned long opts = osb->s_mount_opt;
1475 unsigned int local_alloc_megs;
1477 if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1478 seq_printf(s, ",_netdev");
1479 if (opts & OCFS2_MOUNT_HB_LOCAL)
1480 seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1482 seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1484 seq_printf(s, ",%s", OCFS2_HB_NONE);
1486 if (opts & OCFS2_MOUNT_NOINTR)
1487 seq_printf(s, ",nointr");
1489 if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1490 seq_printf(s, ",data=writeback");
1492 seq_printf(s, ",data=ordered");
1494 if (opts & OCFS2_MOUNT_BARRIER)
1495 seq_printf(s, ",barrier=1");
1497 if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1498 seq_printf(s, ",errors=panic");
1499 else if (opts & OCFS2_MOUNT_ERRORS_CONT)
1500 seq_printf(s, ",errors=continue");
1502 seq_printf(s, ",errors=remount-ro");
1504 if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1505 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1507 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1509 if (osb->osb_commit_interval)
1510 seq_printf(s, ",commit=%u",
1511 (unsigned) (osb->osb_commit_interval / HZ));
1513 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1514 if (local_alloc_megs != ocfs2_la_default_mb(osb))
1515 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1517 if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1518 seq_printf(s, ",localflocks,");
1520 if (osb->osb_cluster_stack[0])
1521 seq_show_option_n(s, "cluster_stack", osb->osb_cluster_stack,
1522 OCFS2_STACK_LABEL_LEN);
1523 if (opts & OCFS2_MOUNT_USRQUOTA)
1524 seq_printf(s, ",usrquota");
1525 if (opts & OCFS2_MOUNT_GRPQUOTA)
1526 seq_printf(s, ",grpquota");
1528 if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1529 seq_printf(s, ",coherency=buffered");
1531 seq_printf(s, ",coherency=full");
1533 if (opts & OCFS2_MOUNT_NOUSERXATTR)
1534 seq_printf(s, ",nouser_xattr");
1536 seq_printf(s, ",user_xattr");
1538 if (opts & OCFS2_MOUNT_INODE64)
1539 seq_printf(s, ",inode64");
1541 if (opts & OCFS2_MOUNT_POSIX_ACL)
1542 seq_printf(s, ",acl");
1544 seq_printf(s, ",noacl");
1546 if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1547 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1549 if (osb->osb_dir_resv_level != osb->osb_resv_level)
1550 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1552 if (opts & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
1553 seq_printf(s, ",journal_async_commit");
1558 static int __init ocfs2_init(void)
1562 status = init_ocfs2_uptodate_cache();
1566 status = ocfs2_initialize_mem_caches();
1570 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1572 ocfs2_set_locking_protocol();
1574 status = register_quota_format(&ocfs2_quota_format);
1577 status = register_filesystem(&ocfs2_fs_type);
1581 unregister_quota_format(&ocfs2_quota_format);
1583 debugfs_remove(ocfs2_debugfs_root);
1584 ocfs2_free_mem_caches();
1586 exit_ocfs2_uptodate_cache();
1592 static void __exit ocfs2_exit(void)
1594 unregister_quota_format(&ocfs2_quota_format);
1596 debugfs_remove(ocfs2_debugfs_root);
1598 ocfs2_free_mem_caches();
1600 unregister_filesystem(&ocfs2_fs_type);
1602 exit_ocfs2_uptodate_cache();
1605 static void ocfs2_put_super(struct super_block *sb)
1607 trace_ocfs2_put_super(sb);
1609 ocfs2_sync_blockdev(sb);
1610 ocfs2_dismount_volume(sb, 0);
1613 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1615 struct ocfs2_super *osb;
1616 u32 numbits, freebits;
1618 struct ocfs2_dinode *bm_lock;
1619 struct buffer_head *bh = NULL;
1620 struct inode *inode = NULL;
1622 trace_ocfs2_statfs(dentry->d_sb, buf);
1624 osb = OCFS2_SB(dentry->d_sb);
1626 inode = ocfs2_get_system_file_inode(osb,
1627 GLOBAL_BITMAP_SYSTEM_INODE,
1628 OCFS2_INVALID_SLOT);
1630 mlog(ML_ERROR, "failed to get bitmap inode\n");
1635 status = ocfs2_inode_lock(inode, &bh, 0);
1641 bm_lock = (struct ocfs2_dinode *) bh->b_data;
1643 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1644 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1646 buf->f_type = OCFS2_SUPER_MAGIC;
1647 buf->f_bsize = dentry->d_sb->s_blocksize;
1648 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1649 buf->f_blocks = ((sector_t) numbits) *
1650 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1651 buf->f_bfree = ((sector_t) freebits) *
1652 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1653 buf->f_bavail = buf->f_bfree;
1654 buf->f_files = numbits;
1655 buf->f_ffree = freebits;
1656 buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1658 buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1659 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1663 ocfs2_inode_unlock(inode, 0);
1674 static void ocfs2_inode_init_once(void *data)
1676 struct ocfs2_inode_info *oi = data;
1679 oi->ip_open_count = 0;
1680 spin_lock_init(&oi->ip_lock);
1681 ocfs2_extent_map_init(&oi->vfs_inode);
1682 INIT_LIST_HEAD(&oi->ip_io_markers);
1683 INIT_LIST_HEAD(&oi->ip_unwritten_list);
1684 oi->ip_dir_start_lookup = 0;
1685 init_rwsem(&oi->ip_alloc_sem);
1686 init_rwsem(&oi->ip_xattr_sem);
1687 mutex_init(&oi->ip_io_mutex);
1689 oi->ip_blkno = 0ULL;
1690 oi->ip_clusters = 0;
1691 oi->ip_next_orphan = NULL;
1693 ocfs2_resv_init_once(&oi->ip_la_data_resv);
1695 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1696 ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1697 ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1699 ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1700 &ocfs2_inode_caching_ops);
1702 inode_init_once(&oi->vfs_inode);
1705 static int ocfs2_initialize_mem_caches(void)
1707 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1708 sizeof(struct ocfs2_inode_info),
1710 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1711 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1712 ocfs2_inode_init_once);
1713 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1714 sizeof(struct ocfs2_dquot),
1716 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1719 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1720 sizeof(struct ocfs2_quota_chunk),
1722 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1724 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1725 !ocfs2_qf_chunk_cachep) {
1726 kmem_cache_destroy(ocfs2_inode_cachep);
1727 kmem_cache_destroy(ocfs2_dquot_cachep);
1728 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1735 static void ocfs2_free_mem_caches(void)
1738 * Make sure all delayed rcu free inodes are flushed before we
1742 kmem_cache_destroy(ocfs2_inode_cachep);
1743 ocfs2_inode_cachep = NULL;
1745 kmem_cache_destroy(ocfs2_dquot_cachep);
1746 ocfs2_dquot_cachep = NULL;
1748 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1749 ocfs2_qf_chunk_cachep = NULL;
1752 static int ocfs2_get_sector(struct super_block *sb,
1753 struct buffer_head **bh,
1757 if (!sb_set_blocksize(sb, sect_size)) {
1758 mlog(ML_ERROR, "unable to set blocksize\n");
1762 *bh = sb_getblk(sb, block);
1764 mlog_errno(-ENOMEM);
1768 if (!buffer_dirty(*bh))
1769 clear_buffer_uptodate(*bh);
1771 ll_rw_block(REQ_OP_READ, 0, 1, bh);
1772 wait_on_buffer(*bh);
1773 if (!buffer_uptodate(*bh)) {
1783 static int ocfs2_mount_volume(struct super_block *sb)
1786 int unlock_super = 0;
1787 struct ocfs2_super *osb = OCFS2_SB(sb);
1789 if (ocfs2_is_hard_readonly(osb))
1792 mutex_init(&osb->obs_trim_fs_mutex);
1794 status = ocfs2_dlm_init(osb);
1797 if (status == -EBADR && ocfs2_userspace_stack(osb))
1798 mlog(ML_ERROR, "couldn't mount because cluster name on"
1799 " disk does not match the running cluster name.\n");
1803 status = ocfs2_super_lock(osb, 1);
1810 /* This will load up the node map and add ourselves to it. */
1811 status = ocfs2_find_slot(osb);
1817 /* load all node-local system inodes */
1818 status = ocfs2_init_local_system_inodes(osb);
1824 status = ocfs2_check_volume(osb);
1830 status = ocfs2_truncate_log_init(osb);
1836 ocfs2_super_unlock(osb, 1);
1841 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1843 int tmp, hangup_needed = 0;
1844 struct ocfs2_super *osb = NULL;
1847 trace_ocfs2_dismount_volume(sb);
1853 /* Remove file check sysfs related directores/files,
1854 * and wait for the pending file check operations */
1855 ocfs2_filecheck_remove_sysfs(osb);
1857 kset_unregister(osb->osb_dev_kset);
1859 /* Orphan scan should be stopped as early as possible */
1860 ocfs2_orphan_scan_stop(osb);
1862 ocfs2_disable_quotas(osb);
1864 /* All dquots should be freed by now */
1865 WARN_ON(!llist_empty(&osb->dquot_drop_list));
1866 /* Wait for worker to be done with the work structure in osb */
1867 cancel_work_sync(&osb->dquot_drop_work);
1869 ocfs2_shutdown_local_alloc(osb);
1871 ocfs2_truncate_log_shutdown(osb);
1873 /* This will disable recovery and flush any recovery work. */
1874 ocfs2_recovery_exit(osb);
1876 ocfs2_journal_shutdown(osb);
1878 ocfs2_sync_blockdev(sb);
1880 ocfs2_purge_refcount_trees(osb);
1882 /* No cluster connection means we've failed during mount, so skip
1883 * all the steps which depended on that to complete. */
1885 tmp = ocfs2_super_lock(osb, 1);
1892 if (osb->slot_num != OCFS2_INVALID_SLOT)
1893 ocfs2_put_slot(osb);
1896 ocfs2_super_unlock(osb, 1);
1898 ocfs2_release_system_inodes(osb);
1901 * If we're dismounting due to mount error, mount.ocfs2 will clean
1902 * up heartbeat. If we're a local mount, there is no heartbeat.
1903 * If we failed before we got a uuid_str yet, we can't stop
1904 * heartbeat. Otherwise, do it.
1906 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str &&
1907 !ocfs2_is_hard_readonly(osb))
1911 ocfs2_dlm_shutdown(osb, hangup_needed);
1913 ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1914 debugfs_remove_recursive(osb->osb_debug_root);
1917 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1919 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1921 if (ocfs2_mount_local(osb))
1922 snprintf(nodestr, sizeof(nodestr), "local");
1924 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1926 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1927 osb->dev_str, nodestr);
1929 ocfs2_delete_osb(osb);
1932 sb->s_fs_info = NULL;
1935 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1936 unsigned uuid_bytes)
1941 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1943 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1944 if (osb->uuid_str == NULL)
1947 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1948 /* print with null */
1949 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1950 if (ret != 2) /* drop super cleans up */
1952 /* then only advance past the last char */
1959 /* Make sure entire volume is addressable by our journal. Requires
1960 osb_clusters_at_boot to be valid and for the journal to have been
1961 initialized by ocfs2_journal_init(). */
1962 static int ocfs2_journal_addressable(struct ocfs2_super *osb)
1966 ocfs2_clusters_to_blocks(osb->sb,
1967 osb->osb_clusters_at_boot) - 1;
1969 /* 32-bit block number is always OK. */
1970 if (max_block <= (u32)~0ULL)
1973 /* Volume is "huge", so see if our journal is new enough to
1975 if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
1976 OCFS2_FEATURE_COMPAT_JBD2_SB) &&
1977 jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
1978 JBD2_FEATURE_INCOMPAT_64BIT))) {
1979 mlog(ML_ERROR, "The journal cannot address the entire volume. "
1980 "Enable the 'block64' journal option with tunefs.ocfs2");
1989 static int ocfs2_initialize_super(struct super_block *sb,
1990 struct buffer_head *bh,
1992 struct ocfs2_blockcheck_stats *stats)
1995 int i, cbits, bbits;
1996 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
1997 struct inode *inode = NULL;
1998 struct ocfs2_journal *journal;
1999 struct ocfs2_super *osb;
2002 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2009 sb->s_fs_info = osb;
2010 sb->s_op = &ocfs2_sops;
2011 sb->s_d_op = &ocfs2_dentry_ops;
2012 sb->s_export_op = &ocfs2_export_ops;
2013 sb->s_qcop = &dquot_quotactl_sysfile_ops;
2014 sb->dq_op = &ocfs2_quota_operations;
2015 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
2016 sb->s_xattr = ocfs2_xattr_handlers;
2017 sb->s_time_gran = 1;
2018 sb->s_flags |= SB_NOATIME;
2019 /* this is needed to support O_LARGEFILE */
2020 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2021 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2022 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2023 memcpy(&sb->s_uuid, di->id2.i_super.s_uuid,
2024 sizeof(di->id2.i_super.s_uuid));
2026 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2028 for (i = 0; i < 3; i++)
2029 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2030 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2033 osb->s_sectsize_bits = blksize_bits(sector_size);
2034 BUG_ON(!osb->s_sectsize_bits);
2036 spin_lock_init(&osb->dc_task_lock);
2037 init_waitqueue_head(&osb->dc_event);
2038 osb->dc_work_sequence = 0;
2039 osb->dc_wake_sequence = 0;
2040 INIT_LIST_HEAD(&osb->blocked_lock_list);
2041 osb->blocked_lock_count = 0;
2042 spin_lock_init(&osb->osb_lock);
2043 spin_lock_init(&osb->osb_xattr_lock);
2044 ocfs2_init_steal_slots(osb);
2046 mutex_init(&osb->system_file_mutex);
2048 atomic_set(&osb->alloc_stats.moves, 0);
2049 atomic_set(&osb->alloc_stats.local_data, 0);
2050 atomic_set(&osb->alloc_stats.bitmap_data, 0);
2051 atomic_set(&osb->alloc_stats.bg_allocs, 0);
2052 atomic_set(&osb->alloc_stats.bg_extends, 0);
2054 /* Copy the blockcheck stats from the superblock probe */
2055 osb->osb_ecc_stats = *stats;
2057 ocfs2_init_node_maps(osb);
2059 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2060 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2062 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2063 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2064 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2070 ocfs2_orphan_scan_init(osb);
2072 status = ocfs2_recovery_init(osb);
2074 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2079 init_waitqueue_head(&osb->checkpoint_event);
2081 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2083 osb->slot_num = OCFS2_INVALID_SLOT;
2085 osb->s_xattr_inline_size = le16_to_cpu(
2086 di->id2.i_super.s_xattr_inline_size);
2088 osb->local_alloc_state = OCFS2_LA_UNUSED;
2089 osb->local_alloc_bh = NULL;
2090 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2092 init_waitqueue_head(&osb->osb_mount_event);
2094 status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2100 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2101 if (!osb->vol_label) {
2102 mlog(ML_ERROR, "unable to alloc vol label\n");
2107 osb->slot_recovery_generations =
2108 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2110 if (!osb->slot_recovery_generations) {
2116 init_waitqueue_head(&osb->osb_wipe_event);
2117 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2118 sizeof(*osb->osb_orphan_wipes),
2120 if (!osb->osb_orphan_wipes) {
2126 osb->osb_rf_lock_tree = RB_ROOT;
2128 osb->s_feature_compat =
2129 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2130 osb->s_feature_ro_compat =
2131 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2132 osb->s_feature_incompat =
2133 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2135 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2136 mlog(ML_ERROR, "couldn't mount because of unsupported "
2137 "optional features (%x).\n", i);
2141 if (!sb_rdonly(osb->sb) && (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2142 mlog(ML_ERROR, "couldn't mount RDWR because of "
2143 "unsupported optional features (%x).\n", i);
2148 if (ocfs2_clusterinfo_valid(osb)) {
2150 * ci_stack and ci_cluster in ocfs2_cluster_info may not be null
2151 * terminated, so make sure no overflow happens here by using
2152 * memcpy. Destination strings will always be null terminated
2153 * because osb is allocated using kzalloc.
2155 osb->osb_stackflags =
2156 OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2157 memcpy(osb->osb_cluster_stack,
2158 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2159 OCFS2_STACK_LABEL_LEN);
2160 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2162 "couldn't mount because of an invalid "
2163 "cluster stack label (%s) \n",
2164 osb->osb_cluster_stack);
2168 memcpy(osb->osb_cluster_name,
2169 OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster,
2170 OCFS2_CLUSTER_NAME_LEN);
2172 /* The empty string is identical with classic tools that
2173 * don't know about s_cluster_info. */
2174 osb->osb_cluster_stack[0] = '\0';
2177 get_random_bytes(&osb->s_next_generation, sizeof(u32));
2180 * This should be done in ocfs2_journal_init(), but unknown
2181 * ordering issues will cause the filesystem to crash.
2182 * If anyone wants to figure out what part of the code
2183 * refers to osb->journal before ocfs2_journal_init() is run,
2186 /* initialize our journal structure */
2188 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2190 mlog(ML_ERROR, "unable to alloc journal\n");
2194 osb->journal = journal;
2195 journal->j_osb = osb;
2197 atomic_set(&journal->j_num_trans, 0);
2198 init_rwsem(&journal->j_trans_barrier);
2199 init_waitqueue_head(&journal->j_checkpointed);
2200 spin_lock_init(&journal->j_lock);
2201 journal->j_trans_id = (unsigned long) 1;
2202 INIT_LIST_HEAD(&journal->j_la_cleanups);
2203 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2204 journal->j_state = OCFS2_JOURNAL_FREE;
2206 INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs);
2207 init_llist_head(&osb->dquot_drop_list);
2209 /* get some pseudo constants for clustersize bits */
2210 osb->s_clustersize_bits =
2211 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2212 osb->s_clustersize = 1 << osb->s_clustersize_bits;
2214 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2215 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2216 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2217 osb->s_clustersize);
2222 total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2223 le32_to_cpu(di->i_clusters));
2225 status = generic_check_addressable(osb->sb->s_blocksize_bits,
2228 mlog(ML_ERROR, "Volume too large "
2229 "to mount safely on this system");
2234 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2235 sizeof(di->id2.i_super.s_uuid))) {
2236 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2241 strlcpy(osb->vol_label, di->id2.i_super.s_label,
2242 OCFS2_MAX_VOL_LABEL_LEN);
2243 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2244 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2245 osb->first_cluster_group_blkno =
2246 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2247 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2248 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2249 trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str,
2250 (unsigned long long)osb->root_blkno,
2251 (unsigned long long)osb->system_dir_blkno,
2252 osb->s_clustersize_bits);
2254 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2255 if (!osb->osb_dlm_debug) {
2261 atomic_set(&osb->vol_state, VOLUME_INIT);
2263 /* load root, system_dir, and all global system inodes */
2264 status = ocfs2_init_global_system_inodes(osb);
2273 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2274 OCFS2_INVALID_SLOT);
2281 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2282 osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2285 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2286 osb->s_feature_incompat) * 8;
2288 status = ocfs2_init_slot_info(osb);
2293 cleancache_init_shared_fs(sb);
2295 osb->ocfs2_wq = alloc_ordered_workqueue("ocfs2_wq", WQ_MEM_RECLAIM);
2296 if (!osb->ocfs2_wq) {
2306 * will return: -EAGAIN if it is ok to keep searching for superblocks
2307 * -EINVAL if there is a bad superblock
2310 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2311 struct buffer_head *bh,
2313 struct ocfs2_blockcheck_stats *stats)
2315 int status = -EAGAIN;
2317 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2318 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2319 /* We have to do a raw check of the feature here */
2320 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2321 OCFS2_FEATURE_INCOMPAT_META_ECC) {
2322 status = ocfs2_block_check_validate(bh->b_data,
2330 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2331 mlog(ML_ERROR, "found superblock with incorrect block "
2332 "size: found %u, should be %u\n",
2333 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2335 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2336 OCFS2_MAJOR_REV_LEVEL ||
2337 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2338 OCFS2_MINOR_REV_LEVEL) {
2339 mlog(ML_ERROR, "found superblock with bad version: "
2340 "found %u.%u, should be %u.%u\n",
2341 le16_to_cpu(di->id2.i_super.s_major_rev_level),
2342 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2343 OCFS2_MAJOR_REV_LEVEL,
2344 OCFS2_MINOR_REV_LEVEL);
2345 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2346 mlog(ML_ERROR, "bad block number on superblock: "
2347 "found %llu, should be %llu\n",
2348 (unsigned long long)le64_to_cpu(di->i_blkno),
2349 (unsigned long long)bh->b_blocknr);
2350 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2351 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2352 mlog(ML_ERROR, "bad cluster size found: %u\n",
2353 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2354 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2355 mlog(ML_ERROR, "bad root_blkno: 0\n");
2356 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2357 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2358 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2360 "Superblock slots found greater than file system "
2361 "maximum: found %u, max %u\n",
2362 le16_to_cpu(di->id2.i_super.s_max_slots),
2371 if (status && status != -EAGAIN)
2376 static int ocfs2_check_volume(struct ocfs2_super *osb)
2381 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2385 /* Init our journal object. */
2386 status = ocfs2_journal_init(osb->journal, &dirty);
2388 mlog(ML_ERROR, "Could not initialize journal!\n");
2392 /* Now that journal has been initialized, check to make sure
2393 entire volume is addressable. */
2394 status = ocfs2_journal_addressable(osb);
2398 /* If the journal was unmounted cleanly then we don't want to
2399 * recover anything. Otherwise, journal_load will do that
2400 * dirty work for us :) */
2402 status = ocfs2_journal_wipe(osb->journal, 0);
2408 printk(KERN_NOTICE "ocfs2: File system on device (%s) was not "
2409 "unmounted cleanly, recovering it.\n", osb->dev_str);
2412 local = ocfs2_mount_local(osb);
2414 /* will play back anything left in the journal. */
2415 status = ocfs2_journal_load(osb->journal, local, dirty);
2417 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2421 if (osb->s_mount_opt & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
2422 jbd2_journal_set_features(osb->journal->j_journal,
2423 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2424 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2426 jbd2_journal_clear_features(osb->journal->j_journal,
2427 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2428 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2431 /* recover my local alloc if we didn't unmount cleanly. */
2432 status = ocfs2_begin_local_alloc_recovery(osb,
2439 /* we complete the recovery process after we've marked
2440 * ourselves as mounted. */
2443 status = ocfs2_load_local_alloc(osb);
2450 /* Recovery will be completed after we've mounted the
2451 * rest of the volume. */
2452 osb->local_alloc_copy = local_alloc;
2456 /* go through each journal, trylock it and if you get the
2457 * lock, and it's marked as dirty, set the bit in the recover
2458 * map and launch a recovery thread for it. */
2459 status = ocfs2_mark_dead_nodes(osb);
2465 status = ocfs2_compute_replay_slots(osb);
2478 * The routine gets called from dismount or close whenever a dismount on
2479 * volume is requested and the osb open count becomes 1.
2480 * It will remove the osb from the global list and also free up all the
2481 * initialized resources and fileobject.
2483 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2485 /* This function assumes that the caller has the main osb resource */
2487 /* ocfs2_initializer_super have already created this workqueue */
2489 destroy_workqueue(osb->ocfs2_wq);
2491 ocfs2_free_slot_info(osb);
2493 kfree(osb->osb_orphan_wipes);
2494 kfree(osb->slot_recovery_generations);
2496 * This belongs in journal shutdown, but because we have to
2497 * allocate osb->journal at the start of ocfs2_initialize_osb(),
2500 kfree(osb->journal);
2501 kfree(osb->local_alloc_copy);
2502 kfree(osb->uuid_str);
2503 kfree(osb->vol_label);
2504 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2505 memset(osb, 0, sizeof(struct ocfs2_super));
2508 /* Depending on the mount option passed, perform one of the following:
2509 * Put OCFS2 into a readonly state (default)
2510 * Return EIO so that only the process errs
2511 * Fix the error as if fsck.ocfs2 -y
2514 static int ocfs2_handle_error(struct super_block *sb)
2516 struct ocfs2_super *osb = OCFS2_SB(sb);
2519 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2520 pr_crit("On-disk corruption discovered. "
2521 "Please run fsck.ocfs2 once the filesystem is unmounted.\n");
2523 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC) {
2524 panic("OCFS2: (device %s): panic forced after error\n",
2526 } else if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_CONT) {
2527 pr_crit("OCFS2: Returning error to the calling process.\n");
2529 } else { /* default option */
2531 if (sb_rdonly(sb) && (ocfs2_is_soft_readonly(osb) || ocfs2_is_hard_readonly(osb)))
2534 pr_crit("OCFS2: File system is now read-only.\n");
2535 sb->s_flags |= SB_RDONLY;
2536 ocfs2_set_ro_flag(osb, 0);
2542 int __ocfs2_error(struct super_block *sb, const char *function,
2543 const char *fmt, ...)
2545 struct va_format vaf;
2548 va_start(args, fmt);
2552 /* Not using mlog here because we want to show the actual
2553 * function the error came from. */
2554 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %pV",
2555 sb->s_id, function, &vaf);
2559 return ocfs2_handle_error(sb);
2562 /* Handle critical errors. This is intentionally more drastic than
2563 * ocfs2_handle_error, so we only use for things like journal errors,
2565 void __ocfs2_abort(struct super_block *sb, const char *function,
2566 const char *fmt, ...)
2568 struct va_format vaf;
2571 va_start(args, fmt);
2576 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %pV",
2577 sb->s_id, function, &vaf);
2581 /* We don't have the cluster support yet to go straight to
2582 * hard readonly in here. Until then, we want to keep
2583 * ocfs2_abort() so that we can at least mark critical
2586 * TODO: This should abort the journal and alert other nodes
2587 * that our slot needs recovery. */
2589 /* Force a panic(). This stinks, but it's better than letting
2590 * things continue without having a proper hard readonly
2592 if (!ocfs2_mount_local(OCFS2_SB(sb)))
2593 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2594 ocfs2_handle_error(sb);
2598 * Void signal blockers, because in-kernel sigprocmask() only fails
2599 * when SIG_* is wrong.
2601 void ocfs2_block_signals(sigset_t *oldset)
2606 sigfillset(&blocked);
2607 rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2611 void ocfs2_unblock_signals(sigset_t *oldset)
2613 int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2617 module_init(ocfs2_init);
2618 module_exit(ocfs2_exit);