1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* -*- mode: c; c-basic-offset: 8; -*-
3 * vim: noexpandtab sw=8 ts=8 sts=0:
7 * load/unload driver, mount/dismount volumes
9 * Copyright (C) 2002, 2004 Oracle. All rights reserved.
12 #include <linux/module.h>
14 #include <linux/types.h>
15 #include <linux/slab.h>
16 #include <linux/highmem.h>
17 #include <linux/init.h>
18 #include <linux/random.h>
19 #include <linux/statfs.h>
20 #include <linux/moduleparam.h>
21 #include <linux/blkdev.h>
22 #include <linux/socket.h>
23 #include <linux/inet.h>
24 #include <linux/parser.h>
25 #include <linux/crc32.h>
26 #include <linux/debugfs.h>
27 #include <linux/mount.h>
28 #include <linux/seq_file.h>
29 #include <linux/quotaops.h>
30 #include <linux/cleancache.h>
31 #include <linux/signal.h>
33 #define CREATE_TRACE_POINTS
34 #include "ocfs2_trace.h"
36 #include <cluster/masklog.h>
40 /* this should be the only file to include a version 1 header */
41 #include "ocfs1_fs_compat.h"
45 #include "blockcheck.h"
48 #include "extent_map.h"
49 #include "heartbeat.h"
52 #include "localalloc.h"
60 #include "refcounttree.h"
63 #include "buffer_head_io.h"
64 #include "filecheck.h"
66 static struct kmem_cache *ocfs2_inode_cachep;
67 struct kmem_cache *ocfs2_dquot_cachep;
68 struct kmem_cache *ocfs2_qf_chunk_cachep;
70 static struct dentry *ocfs2_debugfs_root;
72 MODULE_AUTHOR("Oracle");
73 MODULE_LICENSE("GPL");
74 MODULE_DESCRIPTION("OCFS2 cluster file system");
78 unsigned long commit_interval;
79 unsigned long mount_opt;
80 unsigned int atime_quantum;
83 unsigned int resv_level;
85 char cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
88 static int ocfs2_parse_options(struct super_block *sb, char *options,
89 struct mount_options *mopt,
91 static int ocfs2_check_set_options(struct super_block *sb,
92 struct mount_options *options);
93 static int ocfs2_show_options(struct seq_file *s, struct dentry *root);
94 static void ocfs2_put_super(struct super_block *sb);
95 static int ocfs2_mount_volume(struct super_block *sb);
96 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
97 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
98 static int ocfs2_initialize_mem_caches(void);
99 static void ocfs2_free_mem_caches(void);
100 static void ocfs2_delete_osb(struct ocfs2_super *osb);
102 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
104 static int ocfs2_sync_fs(struct super_block *sb, int wait);
106 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
107 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
108 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
109 static int ocfs2_check_volume(struct ocfs2_super *osb);
110 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
111 struct buffer_head *bh,
113 struct ocfs2_blockcheck_stats *stats);
114 static int ocfs2_initialize_super(struct super_block *sb,
115 struct buffer_head *bh,
117 struct ocfs2_blockcheck_stats *stats);
118 static int ocfs2_get_sector(struct super_block *sb,
119 struct buffer_head **bh,
122 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
123 static void ocfs2_free_inode(struct inode *inode);
124 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
125 static int ocfs2_enable_quotas(struct ocfs2_super *osb);
126 static void ocfs2_disable_quotas(struct ocfs2_super *osb);
128 static struct dquot **ocfs2_get_dquots(struct inode *inode)
130 return OCFS2_I(inode)->i_dquot;
133 static const struct super_operations ocfs2_sops = {
134 .statfs = ocfs2_statfs,
135 .alloc_inode = ocfs2_alloc_inode,
136 .free_inode = ocfs2_free_inode,
137 .drop_inode = ocfs2_drop_inode,
138 .evict_inode = ocfs2_evict_inode,
139 .sync_fs = ocfs2_sync_fs,
140 .put_super = ocfs2_put_super,
141 .remount_fs = ocfs2_remount,
142 .show_options = ocfs2_show_options,
143 .quota_read = ocfs2_quota_read,
144 .quota_write = ocfs2_quota_write,
145 .get_dquots = ocfs2_get_dquots,
172 Opt_coherency_buffered,
176 Opt_journal_async_commit,
182 static const match_table_t tokens = {
183 {Opt_barrier, "barrier=%u"},
184 {Opt_err_panic, "errors=panic"},
185 {Opt_err_ro, "errors=remount-ro"},
187 {Opt_nointr, "nointr"},
188 {Opt_hb_none, OCFS2_HB_NONE},
189 {Opt_hb_local, OCFS2_HB_LOCAL},
190 {Opt_hb_global, OCFS2_HB_GLOBAL},
191 {Opt_data_ordered, "data=ordered"},
192 {Opt_data_writeback, "data=writeback"},
193 {Opt_atime_quantum, "atime_quantum=%u"},
194 {Opt_slot, "preferred_slot=%u"},
195 {Opt_commit, "commit=%u"},
196 {Opt_localalloc, "localalloc=%d"},
197 {Opt_localflocks, "localflocks"},
198 {Opt_stack, "cluster_stack=%s"},
199 {Opt_user_xattr, "user_xattr"},
200 {Opt_nouser_xattr, "nouser_xattr"},
201 {Opt_inode64, "inode64"},
203 {Opt_noacl, "noacl"},
204 {Opt_usrquota, "usrquota"},
205 {Opt_grpquota, "grpquota"},
206 {Opt_coherency_buffered, "coherency=buffered"},
207 {Opt_coherency_full, "coherency=full"},
208 {Opt_resv_level, "resv_level=%u"},
209 {Opt_dir_resv_level, "dir_resv_level=%u"},
210 {Opt_journal_async_commit, "journal_async_commit"},
211 {Opt_err_cont, "errors=continue"},
212 {Opt_nocluster, "nocluster"},
216 #ifdef CONFIG_DEBUG_FS
217 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
219 struct ocfs2_cluster_connection *cconn = osb->cconn;
220 struct ocfs2_recovery_map *rm = osb->recovery_map;
221 struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
225 out += scnprintf(buf + out, len - out,
226 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n",
227 "Device", osb->dev_str, osb->uuid_str,
228 osb->fs_generation, osb->vol_label);
230 out += scnprintf(buf + out, len - out,
231 "%10s => State: %d Flags: 0x%lX\n", "Volume",
232 atomic_read(&osb->vol_state), osb->osb_flags);
234 out += scnprintf(buf + out, len - out,
235 "%10s => Block: %lu Cluster: %d\n", "Sizes",
236 osb->sb->s_blocksize, osb->s_clustersize);
238 out += scnprintf(buf + out, len - out,
239 "%10s => Compat: 0x%X Incompat: 0x%X "
241 "Features", osb->s_feature_compat,
242 osb->s_feature_incompat, osb->s_feature_ro_compat);
244 out += scnprintf(buf + out, len - out,
245 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount",
246 osb->s_mount_opt, osb->s_atime_quantum);
249 out += scnprintf(buf + out, len - out,
250 "%10s => Stack: %s Name: %*s "
251 "Version: %d.%d\n", "Cluster",
252 (*osb->osb_cluster_stack == '\0' ?
253 "o2cb" : osb->osb_cluster_stack),
254 cconn->cc_namelen, cconn->cc_name,
255 cconn->cc_version.pv_major,
256 cconn->cc_version.pv_minor);
259 spin_lock_irqsave(&osb->dc_task_lock, flags);
260 out += scnprintf(buf + out, len - out,
261 "%10s => Pid: %d Count: %lu WakeSeq: %lu "
262 "WorkSeq: %lu\n", "DownCnvt",
263 (osb->dc_task ? task_pid_nr(osb->dc_task) : -1),
264 osb->blocked_lock_count, osb->dc_wake_sequence,
265 osb->dc_work_sequence);
266 spin_unlock_irqrestore(&osb->dc_task_lock, flags);
268 spin_lock(&osb->osb_lock);
269 out += scnprintf(buf + out, len - out, "%10s => Pid: %d Nodes:",
271 (osb->recovery_thread_task ?
272 task_pid_nr(osb->recovery_thread_task) : -1));
273 if (rm->rm_used == 0)
274 out += scnprintf(buf + out, len - out, " None\n");
276 for (i = 0; i < rm->rm_used; i++)
277 out += scnprintf(buf + out, len - out, " %d",
279 out += scnprintf(buf + out, len - out, "\n");
281 spin_unlock(&osb->osb_lock);
283 out += scnprintf(buf + out, len - out,
284 "%10s => Pid: %d Interval: %lu\n", "Commit",
285 (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
286 osb->osb_commit_interval);
288 out += scnprintf(buf + out, len - out,
289 "%10s => State: %d TxnId: %lu NumTxns: %d\n",
290 "Journal", osb->journal->j_state,
291 osb->journal->j_trans_id,
292 atomic_read(&osb->journal->j_num_trans));
294 out += scnprintf(buf + out, len - out,
295 "%10s => GlobalAllocs: %d LocalAllocs: %d "
296 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
298 atomic_read(&osb->alloc_stats.bitmap_data),
299 atomic_read(&osb->alloc_stats.local_data),
300 atomic_read(&osb->alloc_stats.bg_allocs),
301 atomic_read(&osb->alloc_stats.moves),
302 atomic_read(&osb->alloc_stats.bg_extends));
304 out += scnprintf(buf + out, len - out,
305 "%10s => State: %u Descriptor: %llu Size: %u bits "
306 "Default: %u bits\n",
307 "LocalAlloc", osb->local_alloc_state,
308 (unsigned long long)osb->la_last_gd,
309 osb->local_alloc_bits, osb->local_alloc_default_bits);
311 spin_lock(&osb->osb_lock);
312 out += scnprintf(buf + out, len - out,
313 "%10s => InodeSlot: %d StolenInodes: %d, "
314 "MetaSlot: %d StolenMeta: %d\n", "Steal",
315 osb->s_inode_steal_slot,
316 atomic_read(&osb->s_num_inodes_stolen),
317 osb->s_meta_steal_slot,
318 atomic_read(&osb->s_num_meta_stolen));
319 spin_unlock(&osb->osb_lock);
321 out += scnprintf(buf + out, len - out, "OrphanScan => ");
322 out += scnprintf(buf + out, len - out, "Local: %u Global: %u ",
323 os->os_count, os->os_seqno);
324 out += scnprintf(buf + out, len - out, " Last Scan: ");
325 if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
326 out += scnprintf(buf + out, len - out, "Disabled\n");
328 out += scnprintf(buf + out, len - out, "%lu seconds ago\n",
329 (unsigned long)(ktime_get_seconds() - os->os_scantime));
331 out += scnprintf(buf + out, len - out, "%10s => %3s %10s\n",
332 "Slots", "Num", "RecoGen");
333 for (i = 0; i < osb->max_slots; ++i) {
334 out += scnprintf(buf + out, len - out,
335 "%10s %c %3d %10d\n",
337 (i == osb->slot_num ? '*' : ' '),
338 i, osb->slot_recovery_generations[i]);
344 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
346 struct ocfs2_super *osb = inode->i_private;
349 buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
353 i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
355 file->private_data = buf;
362 static int ocfs2_debug_release(struct inode *inode, struct file *file)
364 kfree(file->private_data);
368 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
369 size_t nbytes, loff_t *ppos)
371 return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
372 i_size_read(file->f_mapping->host));
375 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
379 static int ocfs2_debug_release(struct inode *inode, struct file *file)
383 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
384 size_t nbytes, loff_t *ppos)
388 #endif /* CONFIG_DEBUG_FS */
390 static const struct file_operations ocfs2_osb_debug_fops = {
391 .open = ocfs2_osb_debug_open,
392 .release = ocfs2_debug_release,
393 .read = ocfs2_debug_read,
394 .llseek = generic_file_llseek,
397 static int ocfs2_sync_fs(struct super_block *sb, int wait)
401 struct ocfs2_super *osb = OCFS2_SB(sb);
403 if (ocfs2_is_hard_readonly(osb))
407 status = ocfs2_flush_truncate_log(osb);
411 ocfs2_schedule_truncate_log_flush(osb, 0);
414 if (jbd2_journal_start_commit(osb->journal->j_journal,
417 jbd2_log_wait_commit(osb->journal->j_journal,
423 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
425 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
426 && (ino == USER_QUOTA_SYSTEM_INODE
427 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
429 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
430 && (ino == GROUP_QUOTA_SYSTEM_INODE
431 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
436 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
438 struct inode *new = NULL;
442 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
444 status = PTR_ERR(new);
448 osb->root_inode = new;
450 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
452 status = PTR_ERR(new);
456 osb->sys_root_inode = new;
458 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
459 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
460 if (!ocfs2_need_system_inode(osb, i))
462 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
464 ocfs2_release_system_inodes(osb);
465 status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
467 mlog(ML_ERROR, "Unable to load system inode %d, "
468 "possibly corrupt fs?", i);
471 // the array now has one ref, so drop this one
481 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
483 struct inode *new = NULL;
487 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
488 i < NUM_SYSTEM_INODES;
490 if (!ocfs2_need_system_inode(osb, i))
492 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
494 ocfs2_release_system_inodes(osb);
495 status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
496 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
497 status, i, osb->slot_num);
500 /* the array now has one ref, so drop this one */
510 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
515 for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
516 inode = osb->global_system_inodes[i];
519 osb->global_system_inodes[i] = NULL;
523 inode = osb->sys_root_inode;
526 osb->sys_root_inode = NULL;
529 inode = osb->root_inode;
532 osb->root_inode = NULL;
535 if (!osb->local_system_inodes)
538 for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
539 if (osb->local_system_inodes[i]) {
540 iput(osb->local_system_inodes[i]);
541 osb->local_system_inodes[i] = NULL;
545 kfree(osb->local_system_inodes);
546 osb->local_system_inodes = NULL;
549 /* We're allocating fs objects, use GFP_NOFS */
550 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
552 struct ocfs2_inode_info *oi;
554 oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
559 oi->i_datasync_tid = 0;
560 memset(&oi->i_dquot, 0, sizeof(oi->i_dquot));
562 jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
563 return &oi->vfs_inode;
566 static void ocfs2_free_inode(struct inode *inode)
568 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
571 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
574 unsigned int bytes = 1 << cbits;
575 unsigned int trim = bytes;
576 unsigned int bitshift = 32;
579 * i_size and all block offsets in ocfs2 are always 64 bits
580 * wide. i_clusters is 32 bits, in cluster-sized units. So on
581 * 64 bit platforms, cluster size will be the limiting factor.
584 #if BITS_PER_LONG == 32
585 BUILD_BUG_ON(sizeof(sector_t) != 8);
587 * We might be limited by page cache size.
589 if (bytes > PAGE_SIZE) {
593 * Shift by 31 here so that we don't get larger than
601 * Trim by a whole cluster when we can actually approach the
602 * on-disk limits. Otherwise we can overflow i_clusters when
603 * an extent start is at the max offset.
605 return (((unsigned long long)bytes) << bitshift) - trim;
608 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
610 int incompat_features;
612 struct mount_options parsed_options;
613 struct ocfs2_super *osb = OCFS2_SB(sb);
618 if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
619 !ocfs2_check_set_options(sb, &parsed_options)) {
624 tmp = OCFS2_MOUNT_NOCLUSTER;
625 if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
627 mlog(ML_ERROR, "Cannot change nocluster option on remount\n");
631 tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
633 if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
635 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
639 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
640 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
642 mlog(ML_ERROR, "Cannot change data mode on remount\n");
646 /* Probably don't want this on remount; it might
647 * mess with other nodes */
648 if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
649 (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
651 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
655 /* We're going to/from readonly mode. */
656 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
657 /* Disable quota accounting before remounting RO */
658 if (*flags & SB_RDONLY) {
659 ret = ocfs2_susp_quotas(osb, 0);
663 /* Lock here so the check of HARD_RO and the potential
664 * setting of SOFT_RO is atomic. */
665 spin_lock(&osb->osb_lock);
666 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
667 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
672 if (*flags & SB_RDONLY) {
673 sb->s_flags |= SB_RDONLY;
674 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
676 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
677 mlog(ML_ERROR, "Cannot remount RDWR "
678 "filesystem due to previous errors.\n");
682 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
683 if (incompat_features) {
684 mlog(ML_ERROR, "Cannot remount RDWR because "
685 "of unsupported optional features "
686 "(%x).\n", incompat_features);
690 sb->s_flags &= ~SB_RDONLY;
691 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
693 trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags);
695 spin_unlock(&osb->osb_lock);
696 /* Enable quota accounting after remounting RW */
697 if (!ret && !(*flags & SB_RDONLY)) {
698 if (sb_any_quota_suspended(sb))
699 ret = ocfs2_susp_quotas(osb, 1);
701 ret = ocfs2_enable_quotas(osb);
703 /* Return back changes... */
704 spin_lock(&osb->osb_lock);
705 sb->s_flags |= SB_RDONLY;
706 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
707 spin_unlock(&osb->osb_lock);
714 /* Only save off the new mount options in case of a successful
716 osb->s_mount_opt = parsed_options.mount_opt;
717 osb->s_atime_quantum = parsed_options.atime_quantum;
718 osb->preferred_slot = parsed_options.slot;
719 if (parsed_options.commit_interval)
720 osb->osb_commit_interval = parsed_options.commit_interval;
722 if (!ocfs2_is_hard_readonly(osb))
723 ocfs2_set_journal_params(osb);
725 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
726 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
733 static int ocfs2_sb_probe(struct super_block *sb,
734 struct buffer_head **bh,
736 struct ocfs2_blockcheck_stats *stats)
739 struct ocfs1_vol_disk_hdr *hdr;
740 struct ocfs2_dinode *di;
746 *sector_size = bdev_logical_block_size(sb->s_bdev);
747 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
748 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
749 *sector_size, OCFS2_MAX_BLOCKSIZE);
754 /* Can this really happen? */
755 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
756 *sector_size = OCFS2_MIN_BLOCKSIZE;
758 /* check block zero for old format */
759 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
764 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
765 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
766 mlog(ML_ERROR, "incompatible version: %u.%u\n",
767 hdr->major_version, hdr->minor_version);
770 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
771 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
772 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
779 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
780 "upgraded before mounting with ocfs v2\n");
785 * Now check at magic offset for 512, 1024, 2048, 4096
786 * blocksizes. 4096 is the maximum blocksize because it is
787 * the minimum clustersize.
790 for (blksize = *sector_size;
791 blksize <= OCFS2_MAX_BLOCKSIZE;
793 tmpstat = ocfs2_get_sector(sb, bh,
794 OCFS2_SUPER_BLOCK_BLKNO,
801 di = (struct ocfs2_dinode *) (*bh)->b_data;
802 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
803 spin_lock_init(&stats->b_lock);
804 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
809 if (tmpstat != -EAGAIN) {
819 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
821 u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
823 if (osb->s_mount_opt & hb_enabled) {
824 if (ocfs2_mount_local(osb)) {
825 mlog(ML_ERROR, "Cannot heartbeat on a locally "
826 "mounted device.\n");
829 if (ocfs2_userspace_stack(osb)) {
830 mlog(ML_ERROR, "Userspace stack expected, but "
831 "o2cb heartbeat arguments passed to mount\n");
834 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
835 !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
836 ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
837 ocfs2_cluster_o2cb_global_heartbeat(osb))) {
838 mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
843 if (!(osb->s_mount_opt & hb_enabled)) {
844 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
845 !ocfs2_userspace_stack(osb)) {
846 mlog(ML_ERROR, "Heartbeat has to be started to mount "
847 "a read-write clustered device.\n");
856 * If we're using a userspace stack, mount should have passed
857 * a name that matches the disk. If not, mount should not
858 * have passed a stack.
860 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
861 struct mount_options *mopt)
863 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
865 "cluster stack passed to mount, but this filesystem "
866 "does not support it\n");
870 if (ocfs2_userspace_stack(osb) &&
871 !(osb->s_mount_opt & OCFS2_MOUNT_NOCLUSTER) &&
872 strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
873 OCFS2_STACK_LABEL_LEN)) {
875 "cluster stack passed to mount (\"%s\") does not "
876 "match the filesystem (\"%s\")\n",
878 osb->osb_cluster_stack);
885 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
888 struct super_block *sb = osb->sb;
889 unsigned int feature[OCFS2_MAXQUOTAS] = {
890 OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
891 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
894 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
895 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
898 status = dquot_resume(sb, type);
900 struct ocfs2_mem_dqinfo *oinfo;
902 /* Cancel periodic syncing before suspending */
903 oinfo = sb_dqinfo(sb, type)->dqi_priv;
904 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
905 status = dquot_suspend(sb, type);
911 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
912 "remount (error = %d).\n", status);
916 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
918 struct inode *inode[OCFS2_MAXQUOTAS] = { NULL, NULL };
919 struct super_block *sb = osb->sb;
920 unsigned int feature[OCFS2_MAXQUOTAS] = {
921 OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
922 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
923 unsigned int ino[OCFS2_MAXQUOTAS] = {
924 LOCAL_USER_QUOTA_SYSTEM_INODE,
925 LOCAL_GROUP_QUOTA_SYSTEM_INODE };
929 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
930 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
931 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
933 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
939 status = dquot_load_quota_inode(inode[type], type, QFMT_OCFS2,
940 DQUOT_USAGE_ENABLED);
945 for (type = 0; type < OCFS2_MAXQUOTAS; type++)
949 ocfs2_disable_quotas(osb);
950 for (type = 0; type < OCFS2_MAXQUOTAS; type++)
956 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
960 struct super_block *sb = osb->sb;
961 struct ocfs2_mem_dqinfo *oinfo;
963 /* We mostly ignore errors in this function because there's not much
964 * we can do when we see them */
965 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
966 if (!sb_has_quota_loaded(sb, type))
968 oinfo = sb_dqinfo(sb, type)->dqi_priv;
969 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
970 inode = igrab(sb->s_dquot.files[type]);
971 /* Turn off quotas. This will remove all dquot structures from
972 * memory and so they will be automatically synced to global
974 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
975 DQUOT_LIMITS_ENABLED);
982 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
985 int status, sector_size;
986 struct mount_options parsed_options;
987 struct inode *inode = NULL;
988 struct ocfs2_super *osb = NULL;
989 struct buffer_head *bh = NULL;
991 struct ocfs2_blockcheck_stats stats;
993 trace_ocfs2_fill_super(sb, data, silent);
995 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
997 goto read_super_error;
1000 /* probe for superblock */
1001 status = ocfs2_sb_probe(sb, &bh, §or_size, &stats);
1003 mlog(ML_ERROR, "superblock probe failed!\n");
1004 goto read_super_error;
1007 status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1011 goto read_super_error;
1016 if (!ocfs2_check_set_options(sb, &parsed_options)) {
1018 goto read_super_error;
1020 osb->s_mount_opt = parsed_options.mount_opt;
1021 osb->s_atime_quantum = parsed_options.atime_quantum;
1022 osb->preferred_slot = parsed_options.slot;
1023 osb->osb_commit_interval = parsed_options.commit_interval;
1025 ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1026 osb->osb_resv_level = parsed_options.resv_level;
1027 osb->osb_dir_resv_level = parsed_options.resv_level;
1028 if (parsed_options.dir_resv_level == -1)
1029 osb->osb_dir_resv_level = parsed_options.resv_level;
1031 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1033 status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1035 goto read_super_error;
1037 sb->s_magic = OCFS2_SUPER_MAGIC;
1039 sb->s_flags = (sb->s_flags & ~(SB_POSIXACL | SB_NOSEC)) |
1040 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? SB_POSIXACL : 0);
1042 /* Hard readonly mode only if: bdev_read_only, SB_RDONLY,
1044 if (bdev_read_only(sb->s_bdev)) {
1045 if (!sb_rdonly(sb)) {
1047 mlog(ML_ERROR, "Readonly device detected but readonly "
1048 "mount was not specified.\n");
1049 goto read_super_error;
1052 /* You should not be able to start a local heartbeat
1053 * on a readonly device. */
1054 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1056 mlog(ML_ERROR, "Local heartbeat specified on readonly "
1058 goto read_super_error;
1061 status = ocfs2_check_journals_nolocks(osb);
1063 if (status == -EROFS)
1064 mlog(ML_ERROR, "Recovery required on readonly "
1065 "file system, but write access is "
1069 goto read_super_error;
1072 ocfs2_set_ro_flag(osb, 1);
1074 printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. "
1075 "Cluster services will not be used for this mount. "
1076 "Recovery will be skipped.\n", osb->dev_str);
1079 if (!ocfs2_is_hard_readonly(osb)) {
1081 ocfs2_set_ro_flag(osb, 0);
1084 status = ocfs2_verify_heartbeat(osb);
1087 goto read_super_error;
1090 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1091 ocfs2_debugfs_root);
1093 debugfs_create_file("fs_state", S_IFREG|S_IRUSR, osb->osb_debug_root,
1094 osb, &ocfs2_osb_debug_fops);
1096 if (ocfs2_meta_ecc(osb))
1097 ocfs2_blockcheck_stats_debugfs_install( &osb->osb_ecc_stats,
1098 osb->osb_debug_root);
1100 status = ocfs2_mount_volume(sb);
1102 goto read_super_error;
1104 if (osb->root_inode)
1105 inode = igrab(osb->root_inode);
1110 goto read_super_error;
1113 root = d_make_root(inode);
1117 goto read_super_error;
1122 ocfs2_complete_mount_recovery(osb);
1124 osb->osb_dev_kset = kset_create_and_add(sb->s_id, NULL,
1126 if (!osb->osb_dev_kset) {
1128 mlog(ML_ERROR, "Unable to create device kset %s.\n", sb->s_id);
1129 goto read_super_error;
1132 /* Create filecheck sysfs related directories/files at
1133 * /sys/fs/ocfs2/<devname>/filecheck */
1134 if (ocfs2_filecheck_create_sysfs(osb)) {
1136 mlog(ML_ERROR, "Unable to create filecheck sysfs directory at "
1137 "/sys/fs/ocfs2/%s/filecheck.\n", sb->s_id);
1138 goto read_super_error;
1141 if (ocfs2_mount_local(osb))
1142 snprintf(nodestr, sizeof(nodestr), "local");
1144 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1146 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1147 "with %s data mode.\n",
1148 osb->dev_str, nodestr, osb->slot_num,
1149 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1152 if ((osb->s_mount_opt & OCFS2_MOUNT_NOCLUSTER) &&
1153 !(osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT))
1154 printk(KERN_NOTICE "ocfs2: The shared device (%s) is mounted "
1155 "without cluster aware mode.\n", osb->dev_str);
1157 atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1158 wake_up(&osb->osb_mount_event);
1160 /* Now we can initialize quotas because we can afford to wait
1161 * for cluster locks recovery now. That also means that truncation
1162 * log recovery can happen but that waits for proper quota setup */
1163 if (!sb_rdonly(sb)) {
1164 status = ocfs2_enable_quotas(osb);
1166 /* We have to err-out specially here because
1167 * s_root is already set */
1169 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1170 wake_up(&osb->osb_mount_event);
1175 ocfs2_complete_quota_recovery(osb);
1177 /* Now we wake up again for processes waiting for quotas */
1178 atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1179 wake_up(&osb->osb_mount_event);
1181 /* Start this when the mount is almost sure of being successful */
1182 ocfs2_orphan_scan_start(osb);
1193 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1194 wake_up(&osb->osb_mount_event);
1195 ocfs2_dismount_volume(sb, 1);
1201 static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
1203 const char *dev_name,
1206 return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
1209 static struct file_system_type ocfs2_fs_type = {
1210 .owner = THIS_MODULE,
1212 .mount = ocfs2_mount,
1213 .kill_sb = kill_block_super,
1214 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1217 MODULE_ALIAS_FS("ocfs2");
1219 static int ocfs2_check_set_options(struct super_block *sb,
1220 struct mount_options *options)
1222 if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1223 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1224 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1225 mlog(ML_ERROR, "User quotas were requested, but this "
1226 "filesystem does not have the feature enabled.\n");
1229 if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1230 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1231 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1232 mlog(ML_ERROR, "Group quotas were requested, but this "
1233 "filesystem does not have the feature enabled.\n");
1236 if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1237 !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1238 mlog(ML_ERROR, "ACL support requested but extended attributes "
1239 "feature is not enabled\n");
1242 /* No ACL setting specified? Use XATTR feature... */
1243 if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1244 OCFS2_MOUNT_NO_POSIX_ACL))) {
1245 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1246 options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1248 options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1253 static int ocfs2_parse_options(struct super_block *sb,
1255 struct mount_options *mopt,
1258 int status, user_stack = 0;
1262 substring_t args[MAX_OPT_ARGS];
1264 trace_ocfs2_parse_options(is_remount, options ? options : "(none)");
1266 mopt->commit_interval = 0;
1267 mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1268 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1269 mopt->slot = OCFS2_INVALID_SLOT;
1270 mopt->localalloc_opt = -1;
1271 mopt->cluster_stack[0] = '\0';
1272 mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1273 mopt->dir_resv_level = -1;
1280 while ((p = strsep(&options, ",")) != NULL) {
1284 token = match_token(p, tokens, args);
1287 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1290 mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1293 mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1296 if (match_int(&args[0], &option)) {
1301 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1303 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1306 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1309 mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1312 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1313 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1314 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1317 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1318 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1319 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_ROFS;
1322 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1323 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1324 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_CONT;
1326 case Opt_data_ordered:
1327 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1329 case Opt_data_writeback:
1330 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1332 case Opt_user_xattr:
1333 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1335 case Opt_nouser_xattr:
1336 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1338 case Opt_atime_quantum:
1339 if (match_int(&args[0], &option)) {
1344 mopt->atime_quantum = option;
1347 if (match_int(&args[0], &option)) {
1352 mopt->slot = (s16)option;
1355 if (match_int(&args[0], &option)) {
1362 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1363 mopt->commit_interval = HZ * option;
1365 case Opt_localalloc:
1366 if (match_int(&args[0], &option)) {
1371 mopt->localalloc_opt = option;
1373 case Opt_localflocks:
1375 * Changing this during remount could race
1376 * flock() requests, or "unbalance" existing
1377 * ones (e.g., a lock is taken in one mode but
1378 * dropped in the other). If users care enough
1379 * to flip locking modes during remount, we
1380 * could add a "local" flag to individual
1381 * flock structures for proper tracking of
1385 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1388 /* Check both that the option we were passed
1389 * is of the right length and that it is a proper
1390 * string of the right length.
1392 if (((args[0].to - args[0].from) !=
1393 OCFS2_STACK_LABEL_LEN) ||
1394 (strnlen(args[0].from,
1395 OCFS2_STACK_LABEL_LEN) !=
1396 OCFS2_STACK_LABEL_LEN)) {
1398 "Invalid cluster_stack option\n");
1402 memcpy(mopt->cluster_stack, args[0].from,
1403 OCFS2_STACK_LABEL_LEN);
1404 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1406 * Open code the memcmp here as we don't have
1408 * ocfs2_userspace_stack().
1410 if (memcmp(mopt->cluster_stack,
1411 OCFS2_CLASSIC_CLUSTER_STACK,
1412 OCFS2_STACK_LABEL_LEN))
1416 mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1419 mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1422 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1424 case Opt_coherency_buffered:
1425 mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1427 case Opt_coherency_full:
1428 mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1431 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1432 mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1435 mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1436 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1438 case Opt_resv_level:
1441 if (match_int(&args[0], &option)) {
1445 if (option >= OCFS2_MIN_RESV_LEVEL &&
1446 option < OCFS2_MAX_RESV_LEVEL)
1447 mopt->resv_level = option;
1449 case Opt_dir_resv_level:
1452 if (match_int(&args[0], &option)) {
1456 if (option >= OCFS2_MIN_RESV_LEVEL &&
1457 option < OCFS2_MAX_RESV_LEVEL)
1458 mopt->dir_resv_level = option;
1460 case Opt_journal_async_commit:
1461 mopt->mount_opt |= OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT;
1464 mopt->mount_opt |= OCFS2_MOUNT_NOCLUSTER;
1468 "Unrecognized mount option \"%s\" "
1469 "or missing value\n", p);
1475 if (user_stack == 0) {
1476 /* Ensure only one heartbeat mode */
1477 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1478 OCFS2_MOUNT_HB_GLOBAL |
1479 OCFS2_MOUNT_HB_NONE);
1480 if (hweight32(tmp) != 1) {
1481 mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1493 static int ocfs2_show_options(struct seq_file *s, struct dentry *root)
1495 struct ocfs2_super *osb = OCFS2_SB(root->d_sb);
1496 unsigned long opts = osb->s_mount_opt;
1497 unsigned int local_alloc_megs;
1499 if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1500 seq_printf(s, ",_netdev");
1501 if (opts & OCFS2_MOUNT_HB_LOCAL)
1502 seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1504 seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1506 seq_printf(s, ",%s", OCFS2_HB_NONE);
1508 if (opts & OCFS2_MOUNT_NOINTR)
1509 seq_printf(s, ",nointr");
1511 if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1512 seq_printf(s, ",data=writeback");
1514 seq_printf(s, ",data=ordered");
1516 if (opts & OCFS2_MOUNT_BARRIER)
1517 seq_printf(s, ",barrier=1");
1519 if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1520 seq_printf(s, ",errors=panic");
1521 else if (opts & OCFS2_MOUNT_ERRORS_CONT)
1522 seq_printf(s, ",errors=continue");
1524 seq_printf(s, ",errors=remount-ro");
1526 if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1527 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1529 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1531 if (osb->osb_commit_interval)
1532 seq_printf(s, ",commit=%u",
1533 (unsigned) (osb->osb_commit_interval / HZ));
1535 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1536 if (local_alloc_megs != ocfs2_la_default_mb(osb))
1537 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1539 if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1540 seq_printf(s, ",localflocks,");
1542 if (osb->osb_cluster_stack[0])
1543 seq_show_option_n(s, "cluster_stack", osb->osb_cluster_stack,
1544 OCFS2_STACK_LABEL_LEN);
1545 if (opts & OCFS2_MOUNT_USRQUOTA)
1546 seq_printf(s, ",usrquota");
1547 if (opts & OCFS2_MOUNT_GRPQUOTA)
1548 seq_printf(s, ",grpquota");
1550 if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1551 seq_printf(s, ",coherency=buffered");
1553 seq_printf(s, ",coherency=full");
1555 if (opts & OCFS2_MOUNT_NOUSERXATTR)
1556 seq_printf(s, ",nouser_xattr");
1558 seq_printf(s, ",user_xattr");
1560 if (opts & OCFS2_MOUNT_INODE64)
1561 seq_printf(s, ",inode64");
1563 if (opts & OCFS2_MOUNT_POSIX_ACL)
1564 seq_printf(s, ",acl");
1566 seq_printf(s, ",noacl");
1568 if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1569 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1571 if (osb->osb_dir_resv_level != osb->osb_resv_level)
1572 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1574 if (opts & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
1575 seq_printf(s, ",journal_async_commit");
1577 if (opts & OCFS2_MOUNT_NOCLUSTER)
1578 seq_printf(s, ",nocluster");
1583 static int __init ocfs2_init(void)
1587 status = init_ocfs2_uptodate_cache();
1591 status = ocfs2_initialize_mem_caches();
1595 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1597 ocfs2_set_locking_protocol();
1599 status = register_quota_format(&ocfs2_quota_format);
1602 status = register_filesystem(&ocfs2_fs_type);
1606 unregister_quota_format(&ocfs2_quota_format);
1608 debugfs_remove(ocfs2_debugfs_root);
1609 ocfs2_free_mem_caches();
1611 exit_ocfs2_uptodate_cache();
1617 static void __exit ocfs2_exit(void)
1619 unregister_quota_format(&ocfs2_quota_format);
1621 debugfs_remove(ocfs2_debugfs_root);
1623 ocfs2_free_mem_caches();
1625 unregister_filesystem(&ocfs2_fs_type);
1627 exit_ocfs2_uptodate_cache();
1630 static void ocfs2_put_super(struct super_block *sb)
1632 trace_ocfs2_put_super(sb);
1634 ocfs2_sync_blockdev(sb);
1635 ocfs2_dismount_volume(sb, 0);
1638 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1640 struct ocfs2_super *osb;
1641 u32 numbits, freebits;
1643 struct ocfs2_dinode *bm_lock;
1644 struct buffer_head *bh = NULL;
1645 struct inode *inode = NULL;
1647 trace_ocfs2_statfs(dentry->d_sb, buf);
1649 osb = OCFS2_SB(dentry->d_sb);
1651 inode = ocfs2_get_system_file_inode(osb,
1652 GLOBAL_BITMAP_SYSTEM_INODE,
1653 OCFS2_INVALID_SLOT);
1655 mlog(ML_ERROR, "failed to get bitmap inode\n");
1660 status = ocfs2_inode_lock(inode, &bh, 0);
1666 bm_lock = (struct ocfs2_dinode *) bh->b_data;
1668 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1669 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1671 buf->f_type = OCFS2_SUPER_MAGIC;
1672 buf->f_bsize = dentry->d_sb->s_blocksize;
1673 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1674 buf->f_blocks = ((sector_t) numbits) *
1675 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1676 buf->f_bfree = ((sector_t) freebits) *
1677 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1678 buf->f_bavail = buf->f_bfree;
1679 buf->f_files = numbits;
1680 buf->f_ffree = freebits;
1681 buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1683 buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1684 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1688 ocfs2_inode_unlock(inode, 0);
1699 static void ocfs2_inode_init_once(void *data)
1701 struct ocfs2_inode_info *oi = data;
1704 oi->ip_open_count = 0;
1705 spin_lock_init(&oi->ip_lock);
1706 ocfs2_extent_map_init(&oi->vfs_inode);
1707 INIT_LIST_HEAD(&oi->ip_io_markers);
1708 INIT_LIST_HEAD(&oi->ip_unwritten_list);
1709 oi->ip_dir_start_lookup = 0;
1710 init_rwsem(&oi->ip_alloc_sem);
1711 init_rwsem(&oi->ip_xattr_sem);
1712 mutex_init(&oi->ip_io_mutex);
1714 oi->ip_blkno = 0ULL;
1715 oi->ip_clusters = 0;
1717 ocfs2_resv_init_once(&oi->ip_la_data_resv);
1719 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1720 ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1721 ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1723 ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1724 &ocfs2_inode_caching_ops);
1726 inode_init_once(&oi->vfs_inode);
1729 static int ocfs2_initialize_mem_caches(void)
1731 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1732 sizeof(struct ocfs2_inode_info),
1734 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1735 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1736 ocfs2_inode_init_once);
1737 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1738 sizeof(struct ocfs2_dquot),
1740 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1743 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1744 sizeof(struct ocfs2_quota_chunk),
1746 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1748 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1749 !ocfs2_qf_chunk_cachep) {
1750 kmem_cache_destroy(ocfs2_inode_cachep);
1751 kmem_cache_destroy(ocfs2_dquot_cachep);
1752 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1759 static void ocfs2_free_mem_caches(void)
1762 * Make sure all delayed rcu free inodes are flushed before we
1766 kmem_cache_destroy(ocfs2_inode_cachep);
1767 ocfs2_inode_cachep = NULL;
1769 kmem_cache_destroy(ocfs2_dquot_cachep);
1770 ocfs2_dquot_cachep = NULL;
1772 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1773 ocfs2_qf_chunk_cachep = NULL;
1776 static int ocfs2_get_sector(struct super_block *sb,
1777 struct buffer_head **bh,
1781 if (!sb_set_blocksize(sb, sect_size)) {
1782 mlog(ML_ERROR, "unable to set blocksize\n");
1786 *bh = sb_getblk(sb, block);
1788 mlog_errno(-ENOMEM);
1792 if (!buffer_dirty(*bh))
1793 clear_buffer_uptodate(*bh);
1795 ll_rw_block(REQ_OP_READ, 0, 1, bh);
1796 wait_on_buffer(*bh);
1797 if (!buffer_uptodate(*bh)) {
1807 static int ocfs2_mount_volume(struct super_block *sb)
1810 int unlock_super = 0;
1811 struct ocfs2_super *osb = OCFS2_SB(sb);
1813 if (ocfs2_is_hard_readonly(osb))
1816 mutex_init(&osb->obs_trim_fs_mutex);
1818 status = ocfs2_dlm_init(osb);
1821 if (status == -EBADR && ocfs2_userspace_stack(osb))
1822 mlog(ML_ERROR, "couldn't mount because cluster name on"
1823 " disk does not match the running cluster name.\n");
1827 status = ocfs2_super_lock(osb, 1);
1834 /* This will load up the node map and add ourselves to it. */
1835 status = ocfs2_find_slot(osb);
1841 /* load all node-local system inodes */
1842 status = ocfs2_init_local_system_inodes(osb);
1848 status = ocfs2_check_volume(osb);
1854 status = ocfs2_truncate_log_init(osb);
1860 ocfs2_super_unlock(osb, 1);
1865 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1867 int tmp, hangup_needed = 0;
1868 struct ocfs2_super *osb = NULL;
1871 trace_ocfs2_dismount_volume(sb);
1877 /* Remove file check sysfs related directores/files,
1878 * and wait for the pending file check operations */
1879 ocfs2_filecheck_remove_sysfs(osb);
1881 kset_unregister(osb->osb_dev_kset);
1883 /* Orphan scan should be stopped as early as possible */
1884 ocfs2_orphan_scan_stop(osb);
1886 ocfs2_disable_quotas(osb);
1888 /* All dquots should be freed by now */
1889 WARN_ON(!llist_empty(&osb->dquot_drop_list));
1890 /* Wait for worker to be done with the work structure in osb */
1891 cancel_work_sync(&osb->dquot_drop_work);
1893 ocfs2_shutdown_local_alloc(osb);
1895 ocfs2_truncate_log_shutdown(osb);
1897 /* This will disable recovery and flush any recovery work. */
1898 ocfs2_recovery_exit(osb);
1900 ocfs2_journal_shutdown(osb);
1902 ocfs2_sync_blockdev(sb);
1904 ocfs2_purge_refcount_trees(osb);
1906 /* No cluster connection means we've failed during mount, so skip
1907 * all the steps which depended on that to complete. */
1909 tmp = ocfs2_super_lock(osb, 1);
1916 if (osb->slot_num != OCFS2_INVALID_SLOT)
1917 ocfs2_put_slot(osb);
1920 ocfs2_super_unlock(osb, 1);
1922 ocfs2_release_system_inodes(osb);
1925 * If we're dismounting due to mount error, mount.ocfs2 will clean
1926 * up heartbeat. If we're a local mount, there is no heartbeat.
1927 * If we failed before we got a uuid_str yet, we can't stop
1928 * heartbeat. Otherwise, do it.
1930 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str &&
1931 !ocfs2_is_hard_readonly(osb))
1935 ocfs2_dlm_shutdown(osb, hangup_needed);
1937 ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1938 debugfs_remove_recursive(osb->osb_debug_root);
1941 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1943 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1945 if (ocfs2_mount_local(osb))
1946 snprintf(nodestr, sizeof(nodestr), "local");
1948 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1950 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1951 osb->dev_str, nodestr);
1953 ocfs2_delete_osb(osb);
1956 sb->s_fs_info = NULL;
1959 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1960 unsigned uuid_bytes)
1965 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1967 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1968 if (osb->uuid_str == NULL)
1971 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1972 /* print with null */
1973 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1974 if (ret != 2) /* drop super cleans up */
1976 /* then only advance past the last char */
1983 /* Make sure entire volume is addressable by our journal. Requires
1984 osb_clusters_at_boot to be valid and for the journal to have been
1985 initialized by ocfs2_journal_init(). */
1986 static int ocfs2_journal_addressable(struct ocfs2_super *osb)
1990 ocfs2_clusters_to_blocks(osb->sb,
1991 osb->osb_clusters_at_boot) - 1;
1993 /* 32-bit block number is always OK. */
1994 if (max_block <= (u32)~0ULL)
1997 /* Volume is "huge", so see if our journal is new enough to
1999 if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2000 OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2001 jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2002 JBD2_FEATURE_INCOMPAT_64BIT))) {
2003 mlog(ML_ERROR, "The journal cannot address the entire volume. "
2004 "Enable the 'block64' journal option with tunefs.ocfs2");
2013 static int ocfs2_initialize_super(struct super_block *sb,
2014 struct buffer_head *bh,
2016 struct ocfs2_blockcheck_stats *stats)
2019 int i, cbits, bbits;
2020 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2021 struct inode *inode = NULL;
2022 struct ocfs2_journal *journal;
2023 struct ocfs2_super *osb;
2026 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2033 sb->s_fs_info = osb;
2034 sb->s_op = &ocfs2_sops;
2035 sb->s_d_op = &ocfs2_dentry_ops;
2036 sb->s_export_op = &ocfs2_export_ops;
2037 sb->s_qcop = &dquot_quotactl_sysfile_ops;
2038 sb->dq_op = &ocfs2_quota_operations;
2039 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
2040 sb->s_xattr = ocfs2_xattr_handlers;
2041 sb->s_time_gran = 1;
2042 sb->s_flags |= SB_NOATIME;
2043 /* this is needed to support O_LARGEFILE */
2044 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2045 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2046 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2047 memcpy(&sb->s_uuid, di->id2.i_super.s_uuid,
2048 sizeof(di->id2.i_super.s_uuid));
2050 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2052 for (i = 0; i < 3; i++)
2053 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2054 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2057 osb->s_sectsize_bits = blksize_bits(sector_size);
2058 BUG_ON(!osb->s_sectsize_bits);
2060 spin_lock_init(&osb->dc_task_lock);
2061 init_waitqueue_head(&osb->dc_event);
2062 osb->dc_work_sequence = 0;
2063 osb->dc_wake_sequence = 0;
2064 INIT_LIST_HEAD(&osb->blocked_lock_list);
2065 osb->blocked_lock_count = 0;
2066 spin_lock_init(&osb->osb_lock);
2067 spin_lock_init(&osb->osb_xattr_lock);
2068 ocfs2_init_steal_slots(osb);
2070 mutex_init(&osb->system_file_mutex);
2072 atomic_set(&osb->alloc_stats.moves, 0);
2073 atomic_set(&osb->alloc_stats.local_data, 0);
2074 atomic_set(&osb->alloc_stats.bitmap_data, 0);
2075 atomic_set(&osb->alloc_stats.bg_allocs, 0);
2076 atomic_set(&osb->alloc_stats.bg_extends, 0);
2078 /* Copy the blockcheck stats from the superblock probe */
2079 osb->osb_ecc_stats = *stats;
2081 ocfs2_init_node_maps(osb);
2083 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2084 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2086 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2087 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2088 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2094 ocfs2_orphan_scan_init(osb);
2096 status = ocfs2_recovery_init(osb);
2098 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2103 init_waitqueue_head(&osb->checkpoint_event);
2105 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2107 osb->slot_num = OCFS2_INVALID_SLOT;
2109 osb->s_xattr_inline_size = le16_to_cpu(
2110 di->id2.i_super.s_xattr_inline_size);
2112 osb->local_alloc_state = OCFS2_LA_UNUSED;
2113 osb->local_alloc_bh = NULL;
2114 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2116 init_waitqueue_head(&osb->osb_mount_event);
2118 status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2124 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2125 if (!osb->vol_label) {
2126 mlog(ML_ERROR, "unable to alloc vol label\n");
2131 osb->slot_recovery_generations =
2132 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2134 if (!osb->slot_recovery_generations) {
2140 init_waitqueue_head(&osb->osb_wipe_event);
2141 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2142 sizeof(*osb->osb_orphan_wipes),
2144 if (!osb->osb_orphan_wipes) {
2150 osb->osb_rf_lock_tree = RB_ROOT;
2152 osb->s_feature_compat =
2153 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2154 osb->s_feature_ro_compat =
2155 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2156 osb->s_feature_incompat =
2157 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2159 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2160 mlog(ML_ERROR, "couldn't mount because of unsupported "
2161 "optional features (%x).\n", i);
2165 if (!sb_rdonly(osb->sb) && (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2166 mlog(ML_ERROR, "couldn't mount RDWR because of "
2167 "unsupported optional features (%x).\n", i);
2172 if (ocfs2_clusterinfo_valid(osb)) {
2173 osb->osb_stackflags =
2174 OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2175 strlcpy(osb->osb_cluster_stack,
2176 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2177 OCFS2_STACK_LABEL_LEN + 1);
2178 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2180 "couldn't mount because of an invalid "
2181 "cluster stack label (%s) \n",
2182 osb->osb_cluster_stack);
2186 strlcpy(osb->osb_cluster_name,
2187 OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster,
2188 OCFS2_CLUSTER_NAME_LEN + 1);
2190 /* The empty string is identical with classic tools that
2191 * don't know about s_cluster_info. */
2192 osb->osb_cluster_stack[0] = '\0';
2195 get_random_bytes(&osb->s_next_generation, sizeof(u32));
2198 * This should be done in ocfs2_journal_init(), but unknown
2199 * ordering issues will cause the filesystem to crash.
2200 * If anyone wants to figure out what part of the code
2201 * refers to osb->journal before ocfs2_journal_init() is run,
2204 /* initialize our journal structure */
2206 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2208 mlog(ML_ERROR, "unable to alloc journal\n");
2212 osb->journal = journal;
2213 journal->j_osb = osb;
2215 atomic_set(&journal->j_num_trans, 0);
2216 init_rwsem(&journal->j_trans_barrier);
2217 init_waitqueue_head(&journal->j_checkpointed);
2218 spin_lock_init(&journal->j_lock);
2219 journal->j_trans_id = (unsigned long) 1;
2220 INIT_LIST_HEAD(&journal->j_la_cleanups);
2221 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2222 journal->j_state = OCFS2_JOURNAL_FREE;
2224 INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs);
2225 init_llist_head(&osb->dquot_drop_list);
2227 /* get some pseudo constants for clustersize bits */
2228 osb->s_clustersize_bits =
2229 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2230 osb->s_clustersize = 1 << osb->s_clustersize_bits;
2232 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2233 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2234 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2235 osb->s_clustersize);
2240 total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2241 le32_to_cpu(di->i_clusters));
2243 status = generic_check_addressable(osb->sb->s_blocksize_bits,
2246 mlog(ML_ERROR, "Volume too large "
2247 "to mount safely on this system");
2252 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2253 sizeof(di->id2.i_super.s_uuid))) {
2254 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2259 strlcpy(osb->vol_label, di->id2.i_super.s_label,
2260 OCFS2_MAX_VOL_LABEL_LEN);
2261 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2262 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2263 osb->first_cluster_group_blkno =
2264 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2265 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2266 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2267 trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str,
2268 (unsigned long long)osb->root_blkno,
2269 (unsigned long long)osb->system_dir_blkno,
2270 osb->s_clustersize_bits);
2272 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2273 if (!osb->osb_dlm_debug) {
2279 atomic_set(&osb->vol_state, VOLUME_INIT);
2281 /* load root, system_dir, and all global system inodes */
2282 status = ocfs2_init_global_system_inodes(osb);
2291 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2292 OCFS2_INVALID_SLOT);
2299 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2300 osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2303 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2304 osb->s_feature_incompat) * 8;
2306 status = ocfs2_init_slot_info(osb);
2311 cleancache_init_shared_fs(sb);
2313 osb->ocfs2_wq = alloc_ordered_workqueue("ocfs2_wq", WQ_MEM_RECLAIM);
2314 if (!osb->ocfs2_wq) {
2324 * will return: -EAGAIN if it is ok to keep searching for superblocks
2325 * -EINVAL if there is a bad superblock
2328 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2329 struct buffer_head *bh,
2331 struct ocfs2_blockcheck_stats *stats)
2333 int status = -EAGAIN;
2335 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2336 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2337 /* We have to do a raw check of the feature here */
2338 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2339 OCFS2_FEATURE_INCOMPAT_META_ECC) {
2340 status = ocfs2_block_check_validate(bh->b_data,
2348 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2349 mlog(ML_ERROR, "found superblock with incorrect block "
2350 "size: found %u, should be %u\n",
2351 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2353 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2354 OCFS2_MAJOR_REV_LEVEL ||
2355 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2356 OCFS2_MINOR_REV_LEVEL) {
2357 mlog(ML_ERROR, "found superblock with bad version: "
2358 "found %u.%u, should be %u.%u\n",
2359 le16_to_cpu(di->id2.i_super.s_major_rev_level),
2360 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2361 OCFS2_MAJOR_REV_LEVEL,
2362 OCFS2_MINOR_REV_LEVEL);
2363 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2364 mlog(ML_ERROR, "bad block number on superblock: "
2365 "found %llu, should be %llu\n",
2366 (unsigned long long)le64_to_cpu(di->i_blkno),
2367 (unsigned long long)bh->b_blocknr);
2368 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2369 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2370 mlog(ML_ERROR, "bad cluster size found: %u\n",
2371 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2372 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2373 mlog(ML_ERROR, "bad root_blkno: 0\n");
2374 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2375 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2376 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2378 "Superblock slots found greater than file system "
2379 "maximum: found %u, max %u\n",
2380 le16_to_cpu(di->id2.i_super.s_max_slots),
2389 if (status && status != -EAGAIN)
2394 static int ocfs2_check_volume(struct ocfs2_super *osb)
2399 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2403 /* Init our journal object. */
2404 status = ocfs2_journal_init(osb->journal, &dirty);
2406 mlog(ML_ERROR, "Could not initialize journal!\n");
2410 /* Now that journal has been initialized, check to make sure
2411 entire volume is addressable. */
2412 status = ocfs2_journal_addressable(osb);
2416 /* If the journal was unmounted cleanly then we don't want to
2417 * recover anything. Otherwise, journal_load will do that
2418 * dirty work for us :) */
2420 status = ocfs2_journal_wipe(osb->journal, 0);
2426 printk(KERN_NOTICE "ocfs2: File system on device (%s) was not "
2427 "unmounted cleanly, recovering it.\n", osb->dev_str);
2430 local = ocfs2_mount_local(osb);
2432 /* will play back anything left in the journal. */
2433 status = ocfs2_journal_load(osb->journal, local, dirty);
2435 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2439 if (osb->s_mount_opt & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
2440 jbd2_journal_set_features(osb->journal->j_journal,
2441 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2442 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2444 jbd2_journal_clear_features(osb->journal->j_journal,
2445 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2446 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2449 /* recover my local alloc if we didn't unmount cleanly. */
2450 status = ocfs2_begin_local_alloc_recovery(osb,
2457 /* we complete the recovery process after we've marked
2458 * ourselves as mounted. */
2461 status = ocfs2_load_local_alloc(osb);
2468 /* Recovery will be completed after we've mounted the
2469 * rest of the volume. */
2470 osb->local_alloc_copy = local_alloc;
2474 /* go through each journal, trylock it and if you get the
2475 * lock, and it's marked as dirty, set the bit in the recover
2476 * map and launch a recovery thread for it. */
2477 status = ocfs2_mark_dead_nodes(osb);
2483 status = ocfs2_compute_replay_slots(osb);
2496 * The routine gets called from dismount or close whenever a dismount on
2497 * volume is requested and the osb open count becomes 1.
2498 * It will remove the osb from the global list and also free up all the
2499 * initialized resources and fileobject.
2501 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2503 /* This function assumes that the caller has the main osb resource */
2505 /* ocfs2_initializer_super have already created this workqueue */
2507 destroy_workqueue(osb->ocfs2_wq);
2509 ocfs2_free_slot_info(osb);
2511 kfree(osb->osb_orphan_wipes);
2512 kfree(osb->slot_recovery_generations);
2514 * This belongs in journal shutdown, but because we have to
2515 * allocate osb->journal at the start of ocfs2_initialize_osb(),
2518 kfree(osb->journal);
2519 kfree(osb->local_alloc_copy);
2520 kfree(osb->uuid_str);
2521 kfree(osb->vol_label);
2522 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2523 memset(osb, 0, sizeof(struct ocfs2_super));
2526 /* Depending on the mount option passed, perform one of the following:
2527 * Put OCFS2 into a readonly state (default)
2528 * Return EIO so that only the process errs
2529 * Fix the error as if fsck.ocfs2 -y
2532 static int ocfs2_handle_error(struct super_block *sb)
2534 struct ocfs2_super *osb = OCFS2_SB(sb);
2537 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2538 pr_crit("On-disk corruption discovered. "
2539 "Please run fsck.ocfs2 once the filesystem is unmounted.\n");
2541 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC) {
2542 panic("OCFS2: (device %s): panic forced after error\n",
2544 } else if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_CONT) {
2545 pr_crit("OCFS2: Returning error to the calling process.\n");
2547 } else { /* default option */
2549 if (sb_rdonly(sb) && (ocfs2_is_soft_readonly(osb) || ocfs2_is_hard_readonly(osb)))
2552 pr_crit("OCFS2: File system is now read-only.\n");
2553 sb->s_flags |= SB_RDONLY;
2554 ocfs2_set_ro_flag(osb, 0);
2560 int __ocfs2_error(struct super_block *sb, const char *function,
2561 const char *fmt, ...)
2563 struct va_format vaf;
2566 va_start(args, fmt);
2570 /* Not using mlog here because we want to show the actual
2571 * function the error came from. */
2572 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %pV",
2573 sb->s_id, function, &vaf);
2577 return ocfs2_handle_error(sb);
2580 /* Handle critical errors. This is intentionally more drastic than
2581 * ocfs2_handle_error, so we only use for things like journal errors,
2583 void __ocfs2_abort(struct super_block *sb, const char *function,
2584 const char *fmt, ...)
2586 struct va_format vaf;
2589 va_start(args, fmt);
2594 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %pV",
2595 sb->s_id, function, &vaf);
2599 /* We don't have the cluster support yet to go straight to
2600 * hard readonly in here. Until then, we want to keep
2601 * ocfs2_abort() so that we can at least mark critical
2604 * TODO: This should abort the journal and alert other nodes
2605 * that our slot needs recovery. */
2607 /* Force a panic(). This stinks, but it's better than letting
2608 * things continue without having a proper hard readonly
2610 if (!ocfs2_mount_local(OCFS2_SB(sb)))
2611 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2612 ocfs2_handle_error(sb);
2616 * Void signal blockers, because in-kernel sigprocmask() only fails
2617 * when SIG_* is wrong.
2619 void ocfs2_block_signals(sigset_t *oldset)
2624 sigfillset(&blocked);
2625 rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2629 void ocfs2_unblock_signals(sigset_t *oldset)
2631 int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2635 module_init(ocfs2_init);
2636 module_exit(ocfs2_exit);