1 /* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
6 * load/unload driver, mount/dismount volumes
8 * Copyright (C) 2002, 2004 Oracle. All rights reserved.
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public
12 * License as published by the Free Software Foundation; either
13 * version 2 of the License, or (at your option) any later version.
15 * This program is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * General Public License for more details.
20 * You should have received a copy of the GNU General Public
21 * License along with this program; if not, write to the
22 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
23 * Boston, MA 021110-1307, USA.
26 #include <linux/module.h>
28 #include <linux/types.h>
29 #include <linux/slab.h>
30 #include <linux/highmem.h>
31 #include <linux/utsname.h>
32 #include <linux/init.h>
33 #include <linux/random.h>
34 #include <linux/statfs.h>
35 #include <linux/moduleparam.h>
36 #include <linux/blkdev.h>
37 #include <linux/socket.h>
38 #include <linux/inet.h>
39 #include <linux/parser.h>
40 #include <linux/crc32.h>
41 #include <linux/debugfs.h>
42 #include <linux/mount.h>
43 #include <linux/seq_file.h>
44 #include <linux/quotaops.h>
46 #define MLOG_MASK_PREFIX ML_SUPER
47 #include <cluster/masklog.h>
51 /* this should be the only file to include a version 1 header */
52 #include "ocfs1_fs_compat.h"
55 #include "blockcheck.h"
58 #include "extent_map.h"
59 #include "heartbeat.h"
62 #include "localalloc.h"
72 #include "buffer_head_io.h"
74 static struct kmem_cache *ocfs2_inode_cachep = NULL;
75 struct kmem_cache *ocfs2_dquot_cachep;
76 struct kmem_cache *ocfs2_qf_chunk_cachep;
78 /* OCFS2 needs to schedule several differnt types of work which
79 * require cluster locking, disk I/O, recovery waits, etc. Since these
80 * types of work tend to be heavy we avoid using the kernel events
81 * workqueue and schedule on our own. */
82 struct workqueue_struct *ocfs2_wq = NULL;
84 static struct dentry *ocfs2_debugfs_root = NULL;
86 MODULE_AUTHOR("Oracle");
87 MODULE_LICENSE("GPL");
91 unsigned long commit_interval;
92 unsigned long mount_opt;
93 unsigned int atime_quantum;
95 unsigned int localalloc_opt;
96 char cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
99 static int ocfs2_parse_options(struct super_block *sb, char *options,
100 struct mount_options *mopt,
102 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt);
103 static void ocfs2_put_super(struct super_block *sb);
104 static int ocfs2_mount_volume(struct super_block *sb);
105 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
106 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
107 static int ocfs2_initialize_mem_caches(void);
108 static void ocfs2_free_mem_caches(void);
109 static void ocfs2_delete_osb(struct ocfs2_super *osb);
111 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
113 static int ocfs2_sync_fs(struct super_block *sb, int wait);
115 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
116 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
117 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
118 static int ocfs2_check_volume(struct ocfs2_super *osb);
119 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
120 struct buffer_head *bh,
122 static int ocfs2_initialize_super(struct super_block *sb,
123 struct buffer_head *bh,
125 static int ocfs2_get_sector(struct super_block *sb,
126 struct buffer_head **bh,
129 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
130 static void ocfs2_destroy_inode(struct inode *inode);
131 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
132 static int ocfs2_enable_quotas(struct ocfs2_super *osb);
133 static void ocfs2_disable_quotas(struct ocfs2_super *osb);
135 static const struct super_operations ocfs2_sops = {
136 .statfs = ocfs2_statfs,
137 .alloc_inode = ocfs2_alloc_inode,
138 .destroy_inode = ocfs2_destroy_inode,
139 .drop_inode = ocfs2_drop_inode,
140 .clear_inode = ocfs2_clear_inode,
141 .delete_inode = ocfs2_delete_inode,
142 .sync_fs = ocfs2_sync_fs,
143 .put_super = ocfs2_put_super,
144 .remount_fs = ocfs2_remount,
145 .show_options = ocfs2_show_options,
146 .quota_read = ocfs2_quota_read,
147 .quota_write = ocfs2_quota_write,
176 static const match_table_t tokens = {
177 {Opt_barrier, "barrier=%u"},
178 {Opt_err_panic, "errors=panic"},
179 {Opt_err_ro, "errors=remount-ro"},
181 {Opt_nointr, "nointr"},
182 {Opt_hb_none, OCFS2_HB_NONE},
183 {Opt_hb_local, OCFS2_HB_LOCAL},
184 {Opt_data_ordered, "data=ordered"},
185 {Opt_data_writeback, "data=writeback"},
186 {Opt_atime_quantum, "atime_quantum=%u"},
187 {Opt_slot, "preferred_slot=%u"},
188 {Opt_commit, "commit=%u"},
189 {Opt_localalloc, "localalloc=%d"},
190 {Opt_localflocks, "localflocks"},
191 {Opt_stack, "cluster_stack=%s"},
192 {Opt_user_xattr, "user_xattr"},
193 {Opt_nouser_xattr, "nouser_xattr"},
194 {Opt_inode64, "inode64"},
196 {Opt_noacl, "noacl"},
197 {Opt_usrquota, "usrquota"},
198 {Opt_grpquota, "grpquota"},
202 #ifdef CONFIG_DEBUG_FS
203 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
207 struct ocfs2_cluster_connection *cconn = osb->cconn;
208 struct ocfs2_recovery_map *rm = osb->recovery_map;
210 out += snprintf(buf + out, len - out,
211 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n",
212 "Device", osb->dev_str, osb->uuid_str,
213 osb->fs_generation, osb->vol_label);
215 out += snprintf(buf + out, len - out,
216 "%10s => State: %d Flags: 0x%lX\n", "Volume",
217 atomic_read(&osb->vol_state), osb->osb_flags);
219 out += snprintf(buf + out, len - out,
220 "%10s => Block: %lu Cluster: %d\n", "Sizes",
221 osb->sb->s_blocksize, osb->s_clustersize);
223 out += snprintf(buf + out, len - out,
224 "%10s => Compat: 0x%X Incompat: 0x%X "
226 "Features", osb->s_feature_compat,
227 osb->s_feature_incompat, osb->s_feature_ro_compat);
229 out += snprintf(buf + out, len - out,
230 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount",
231 osb->s_mount_opt, osb->s_atime_quantum);
233 out += snprintf(buf + out, len - out,
234 "%10s => Stack: %s Name: %*s Version: %d.%d\n",
236 (*osb->osb_cluster_stack == '\0' ?
237 "o2cb" : osb->osb_cluster_stack),
238 cconn->cc_namelen, cconn->cc_name,
239 cconn->cc_version.pv_major, cconn->cc_version.pv_minor);
241 spin_lock(&osb->dc_task_lock);
242 out += snprintf(buf + out, len - out,
243 "%10s => Pid: %d Count: %lu WakeSeq: %lu "
244 "WorkSeq: %lu\n", "DownCnvt",
245 task_pid_nr(osb->dc_task), osb->blocked_lock_count,
246 osb->dc_wake_sequence, osb->dc_work_sequence);
247 spin_unlock(&osb->dc_task_lock);
249 spin_lock(&osb->osb_lock);
250 out += snprintf(buf + out, len - out, "%10s => Pid: %d Nodes:",
252 (osb->recovery_thread_task ?
253 task_pid_nr(osb->recovery_thread_task) : -1));
254 if (rm->rm_used == 0)
255 out += snprintf(buf + out, len - out, " None\n");
257 for (i = 0; i < rm->rm_used; i++)
258 out += snprintf(buf + out, len - out, " %d",
260 out += snprintf(buf + out, len - out, "\n");
262 spin_unlock(&osb->osb_lock);
264 out += snprintf(buf + out, len - out,
265 "%10s => Pid: %d Interval: %lu Needs: %d\n", "Commit",
266 task_pid_nr(osb->commit_task), osb->osb_commit_interval,
267 atomic_read(&osb->needs_checkpoint));
269 out += snprintf(buf + out, len - out,
270 "%10s => State: %d NumTxns: %d TxnId: %lu\n",
271 "Journal", osb->journal->j_state,
272 atomic_read(&osb->journal->j_num_trans),
273 osb->journal->j_trans_id);
275 out += snprintf(buf + out, len - out,
276 "%10s => GlobalAllocs: %d LocalAllocs: %d "
277 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
279 atomic_read(&osb->alloc_stats.bitmap_data),
280 atomic_read(&osb->alloc_stats.local_data),
281 atomic_read(&osb->alloc_stats.bg_allocs),
282 atomic_read(&osb->alloc_stats.moves),
283 atomic_read(&osb->alloc_stats.bg_extends));
285 out += snprintf(buf + out, len - out,
286 "%10s => State: %u Descriptor: %llu Size: %u bits "
287 "Default: %u bits\n",
288 "LocalAlloc", osb->local_alloc_state,
289 (unsigned long long)osb->la_last_gd,
290 osb->local_alloc_bits, osb->local_alloc_default_bits);
292 spin_lock(&osb->osb_lock);
293 out += snprintf(buf + out, len - out,
294 "%10s => Slot: %d NumStolen: %d\n", "Steal",
295 osb->s_inode_steal_slot,
296 atomic_read(&osb->s_num_inodes_stolen));
297 spin_unlock(&osb->osb_lock);
299 out += snprintf(buf + out, len - out, "%10s => %3s %10s\n",
300 "Slots", "Num", "RecoGen");
302 for (i = 0; i < osb->max_slots; ++i) {
303 out += snprintf(buf + out, len - out,
304 "%10s %c %3d %10d\n",
306 (i == osb->slot_num ? '*' : ' '),
307 i, osb->slot_recovery_generations[i]);
313 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
315 struct ocfs2_super *osb = inode->i_private;
318 buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
322 i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
324 file->private_data = buf;
331 static int ocfs2_debug_release(struct inode *inode, struct file *file)
333 kfree(file->private_data);
337 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
338 size_t nbytes, loff_t *ppos)
340 return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
341 i_size_read(file->f_mapping->host));
344 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
348 static int ocfs2_debug_release(struct inode *inode, struct file *file)
352 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
353 size_t nbytes, loff_t *ppos)
357 #endif /* CONFIG_DEBUG_FS */
359 static struct file_operations ocfs2_osb_debug_fops = {
360 .open = ocfs2_osb_debug_open,
361 .release = ocfs2_debug_release,
362 .read = ocfs2_debug_read,
363 .llseek = generic_file_llseek,
366 static int ocfs2_sync_fs(struct super_block *sb, int wait)
370 struct ocfs2_super *osb = OCFS2_SB(sb);
372 if (ocfs2_is_hard_readonly(osb))
376 status = ocfs2_flush_truncate_log(osb);
380 ocfs2_schedule_truncate_log_flush(osb, 0);
383 if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
386 jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
392 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
394 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
395 && (ino == USER_QUOTA_SYSTEM_INODE
396 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
398 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
399 && (ino == GROUP_QUOTA_SYSTEM_INODE
400 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
405 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
407 struct inode *new = NULL;
413 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
415 status = PTR_ERR(new);
419 osb->root_inode = new;
421 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
423 status = PTR_ERR(new);
427 osb->sys_root_inode = new;
429 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
430 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
431 if (!ocfs2_need_system_inode(osb, i))
433 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
435 ocfs2_release_system_inodes(osb);
438 /* FIXME: Should ERROR_RO_FS */
439 mlog(ML_ERROR, "Unable to load system inode %d, "
440 "possibly corrupt fs?", i);
443 // the array now has one ref, so drop this one
452 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
454 struct inode *new = NULL;
460 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
461 i < NUM_SYSTEM_INODES;
463 if (!ocfs2_need_system_inode(osb, i))
465 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
467 ocfs2_release_system_inodes(osb);
469 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
470 status, i, osb->slot_num);
473 /* the array now has one ref, so drop this one */
482 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
489 for (i = 0; i < NUM_SYSTEM_INODES; i++) {
490 inode = osb->system_inodes[i];
493 osb->system_inodes[i] = NULL;
497 inode = osb->sys_root_inode;
500 osb->sys_root_inode = NULL;
503 inode = osb->root_inode;
506 osb->root_inode = NULL;
512 /* We're allocating fs objects, use GFP_NOFS */
513 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
515 struct ocfs2_inode_info *oi;
517 oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
521 jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
522 return &oi->vfs_inode;
525 static void ocfs2_destroy_inode(struct inode *inode)
527 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
530 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
533 unsigned int bytes = 1 << cbits;
534 unsigned int trim = bytes;
535 unsigned int bitshift = 32;
538 * i_size and all block offsets in ocfs2 are always 64 bits
539 * wide. i_clusters is 32 bits, in cluster-sized units. So on
540 * 64 bit platforms, cluster size will be the limiting factor.
543 #if BITS_PER_LONG == 32
544 # if defined(CONFIG_LBD)
545 BUILD_BUG_ON(sizeof(sector_t) != 8);
547 * We might be limited by page cache size.
549 if (bytes > PAGE_CACHE_SIZE) {
550 bytes = PAGE_CACHE_SIZE;
553 * Shift by 31 here so that we don't get larger than
560 * We are limited by the size of sector_t. Use block size, as
561 * that's what we expose to the VFS.
570 * Trim by a whole cluster when we can actually approach the
571 * on-disk limits. Otherwise we can overflow i_clusters when
572 * an extent start is at the max offset.
574 return (((unsigned long long)bytes) << bitshift) - trim;
577 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
579 int incompat_features;
581 struct mount_options parsed_options;
582 struct ocfs2_super *osb = OCFS2_SB(sb);
584 if (!ocfs2_parse_options(sb, data, &parsed_options, 1)) {
589 if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) !=
590 (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
592 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
596 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
597 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
599 mlog(ML_ERROR, "Cannot change data mode on remount\n");
603 /* Probably don't want this on remount; it might
604 * mess with other nodes */
605 if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
606 (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
608 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
612 /* We're going to/from readonly mode. */
613 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
614 /* Disable quota accounting before remounting RO */
615 if (*flags & MS_RDONLY) {
616 ret = ocfs2_susp_quotas(osb, 0);
620 /* Lock here so the check of HARD_RO and the potential
621 * setting of SOFT_RO is atomic. */
622 spin_lock(&osb->osb_lock);
623 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
624 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
629 if (*flags & MS_RDONLY) {
630 mlog(0, "Going to ro mode.\n");
631 sb->s_flags |= MS_RDONLY;
632 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
634 mlog(0, "Making ro filesystem writeable.\n");
636 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
637 mlog(ML_ERROR, "Cannot remount RDWR "
638 "filesystem due to previous errors.\n");
642 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
643 if (incompat_features) {
644 mlog(ML_ERROR, "Cannot remount RDWR because "
645 "of unsupported optional features "
646 "(%x).\n", incompat_features);
650 sb->s_flags &= ~MS_RDONLY;
651 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
654 spin_unlock(&osb->osb_lock);
655 /* Enable quota accounting after remounting RW */
656 if (!ret && !(*flags & MS_RDONLY)) {
657 if (sb_any_quota_suspended(sb))
658 ret = ocfs2_susp_quotas(osb, 1);
660 ret = ocfs2_enable_quotas(osb);
662 /* Return back changes... */
663 spin_lock(&osb->osb_lock);
664 sb->s_flags |= MS_RDONLY;
665 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
666 spin_unlock(&osb->osb_lock);
673 /* Only save off the new mount options in case of a successful
675 if (!(osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_XATTR))
676 parsed_options.mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
677 osb->s_mount_opt = parsed_options.mount_opt;
678 osb->s_atime_quantum = parsed_options.atime_quantum;
679 osb->preferred_slot = parsed_options.slot;
680 if (parsed_options.commit_interval)
681 osb->osb_commit_interval = parsed_options.commit_interval;
683 if (!ocfs2_is_hard_readonly(osb))
684 ocfs2_set_journal_params(osb);
690 static int ocfs2_sb_probe(struct super_block *sb,
691 struct buffer_head **bh,
695 struct ocfs1_vol_disk_hdr *hdr;
696 struct ocfs2_dinode *di;
702 *sector_size = bdev_logical_block_size(sb->s_bdev);
703 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
704 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
705 *sector_size, OCFS2_MAX_BLOCKSIZE);
710 /* Can this really happen? */
711 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
712 *sector_size = OCFS2_MIN_BLOCKSIZE;
714 /* check block zero for old format */
715 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
720 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
721 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
722 mlog(ML_ERROR, "incompatible version: %u.%u\n",
723 hdr->major_version, hdr->minor_version);
726 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
727 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
728 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
735 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
736 "upgraded before mounting with ocfs v2\n");
741 * Now check at magic offset for 512, 1024, 2048, 4096
742 * blocksizes. 4096 is the maximum blocksize because it is
743 * the minimum clustersize.
746 for (blksize = *sector_size;
747 blksize <= OCFS2_MAX_BLOCKSIZE;
749 tmpstat = ocfs2_get_sector(sb, bh,
750 OCFS2_SUPER_BLOCK_BLKNO,
757 di = (struct ocfs2_dinode *) (*bh)->b_data;
758 status = ocfs2_verify_volume(di, *bh, blksize);
763 if (status != -EAGAIN)
771 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
773 if (ocfs2_mount_local(osb)) {
774 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
775 mlog(ML_ERROR, "Cannot heartbeat on a locally "
776 "mounted device.\n");
781 if (ocfs2_userspace_stack(osb)) {
782 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
783 mlog(ML_ERROR, "Userspace stack expected, but "
784 "o2cb heartbeat arguments passed to mount\n");
789 if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
790 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
791 !ocfs2_userspace_stack(osb)) {
792 mlog(ML_ERROR, "Heartbeat has to be started to mount "
793 "a read-write clustered device.\n");
802 * If we're using a userspace stack, mount should have passed
803 * a name that matches the disk. If not, mount should not
804 * have passed a stack.
806 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
807 struct mount_options *mopt)
809 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
811 "cluster stack passed to mount, but this filesystem "
812 "does not support it\n");
816 if (ocfs2_userspace_stack(osb) &&
817 strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
818 OCFS2_STACK_LABEL_LEN)) {
820 "cluster stack passed to mount (\"%s\") does not "
821 "match the filesystem (\"%s\")\n",
823 osb->osb_cluster_stack);
830 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
833 struct super_block *sb = osb->sb;
834 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
835 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
838 for (type = 0; type < MAXQUOTAS; type++) {
839 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
842 status = vfs_quota_enable(
843 sb_dqopt(sb)->files[type],
847 status = vfs_quota_disable(sb, type,
853 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
854 "remount (error = %d).\n", status);
858 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
860 struct inode *inode[MAXQUOTAS] = { NULL, NULL };
861 struct super_block *sb = osb->sb;
862 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
863 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
864 unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
865 LOCAL_GROUP_QUOTA_SYSTEM_INODE };
869 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
870 for (type = 0; type < MAXQUOTAS; type++) {
871 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
873 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
879 status = vfs_quota_enable(inode[type], type, QFMT_OCFS2,
880 DQUOT_USAGE_ENABLED);
885 for (type = 0; type < MAXQUOTAS; type++)
889 ocfs2_disable_quotas(osb);
890 for (type = 0; type < MAXQUOTAS; type++)
896 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
900 struct super_block *sb = osb->sb;
902 /* We mostly ignore errors in this function because there's not much
903 * we can do when we see them */
904 for (type = 0; type < MAXQUOTAS; type++) {
905 if (!sb_has_quota_loaded(sb, type))
907 inode = igrab(sb->s_dquot.files[type]);
908 /* Turn off quotas. This will remove all dquot structures from
909 * memory and so they will be automatically synced to global
911 vfs_quota_disable(sb, type, DQUOT_USAGE_ENABLED |
912 DQUOT_LIMITS_ENABLED);
919 /* Handle quota on quotactl */
920 static int ocfs2_quota_on(struct super_block *sb, int type, int format_id,
921 char *path, int remount)
923 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
924 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
926 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
930 return 0; /* Just ignore it has been handled in
932 return vfs_quota_enable(sb_dqopt(sb)->files[type], type,
933 format_id, DQUOT_LIMITS_ENABLED);
936 /* Handle quota off quotactl */
937 static int ocfs2_quota_off(struct super_block *sb, int type, int remount)
940 return 0; /* Ignore now and handle later in
942 return vfs_quota_disable(sb, type, DQUOT_LIMITS_ENABLED);
945 static struct quotactl_ops ocfs2_quotactl_ops = {
946 .quota_on = ocfs2_quota_on,
947 .quota_off = ocfs2_quota_off,
948 .quota_sync = vfs_quota_sync,
949 .get_info = vfs_get_dqinfo,
950 .set_info = vfs_set_dqinfo,
951 .get_dqblk = vfs_get_dqblk,
952 .set_dqblk = vfs_set_dqblk,
955 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
958 int status, sector_size;
959 struct mount_options parsed_options;
960 struct inode *inode = NULL;
961 struct ocfs2_super *osb = NULL;
962 struct buffer_head *bh = NULL;
965 mlog_entry("%p, %p, %i", sb, data, silent);
967 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
969 goto read_super_error;
972 /* probe for superblock */
973 status = ocfs2_sb_probe(sb, &bh, §or_size);
975 mlog(ML_ERROR, "superblock probe failed!\n");
976 goto read_super_error;
979 status = ocfs2_initialize_super(sb, bh, sector_size);
983 goto read_super_error;
988 if (!(osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_XATTR))
989 parsed_options.mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
991 osb->s_mount_opt = parsed_options.mount_opt;
992 osb->s_atime_quantum = parsed_options.atime_quantum;
993 osb->preferred_slot = parsed_options.slot;
994 osb->osb_commit_interval = parsed_options.commit_interval;
995 osb->local_alloc_default_bits = ocfs2_megabytes_to_clusters(sb, parsed_options.localalloc_opt);
996 osb->local_alloc_bits = osb->local_alloc_default_bits;
997 if (osb->s_mount_opt & OCFS2_MOUNT_USRQUOTA &&
998 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
999 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1001 mlog(ML_ERROR, "User quotas were requested, but this "
1002 "filesystem does not have the feature enabled.\n");
1003 goto read_super_error;
1005 if (osb->s_mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1006 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1007 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1009 mlog(ML_ERROR, "Group quotas were requested, but this "
1010 "filesystem does not have the feature enabled.\n");
1011 goto read_super_error;
1014 status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1016 goto read_super_error;
1018 sb->s_magic = OCFS2_SUPER_MAGIC;
1020 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1021 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1023 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1025 if (bdev_read_only(sb->s_bdev)) {
1026 if (!(sb->s_flags & MS_RDONLY)) {
1028 mlog(ML_ERROR, "Readonly device detected but readonly "
1029 "mount was not specified.\n");
1030 goto read_super_error;
1033 /* You should not be able to start a local heartbeat
1034 * on a readonly device. */
1035 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1037 mlog(ML_ERROR, "Local heartbeat specified on readonly "
1039 goto read_super_error;
1042 status = ocfs2_check_journals_nolocks(osb);
1044 if (status == -EROFS)
1045 mlog(ML_ERROR, "Recovery required on readonly "
1046 "file system, but write access is "
1050 goto read_super_error;
1053 ocfs2_set_ro_flag(osb, 1);
1055 printk(KERN_NOTICE "Readonly device detected. No cluster "
1056 "services will be utilized for this mount. Recovery "
1057 "will be skipped.\n");
1060 if (!ocfs2_is_hard_readonly(osb)) {
1061 if (sb->s_flags & MS_RDONLY)
1062 ocfs2_set_ro_flag(osb, 0);
1065 status = ocfs2_verify_heartbeat(osb);
1068 goto read_super_error;
1071 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1072 ocfs2_debugfs_root);
1073 if (!osb->osb_debug_root) {
1075 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1076 goto read_super_error;
1079 osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1080 osb->osb_debug_root,
1082 &ocfs2_osb_debug_fops);
1083 if (!osb->osb_ctxt) {
1086 goto read_super_error;
1089 status = ocfs2_mount_volume(sb);
1090 if (osb->root_inode)
1091 inode = igrab(osb->root_inode);
1094 goto read_super_error;
1099 goto read_super_error;
1102 root = d_alloc_root(inode);
1106 goto read_super_error;
1111 ocfs2_complete_mount_recovery(osb);
1113 if (ocfs2_mount_local(osb))
1114 snprintf(nodestr, sizeof(nodestr), "local");
1116 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1118 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1119 "with %s data mode.\n",
1120 osb->dev_str, nodestr, osb->slot_num,
1121 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1124 atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1125 wake_up(&osb->osb_mount_event);
1127 /* Now we can initialize quotas because we can afford to wait
1128 * for cluster locks recovery now. That also means that truncation
1129 * log recovery can happen but that waits for proper quota setup */
1130 if (!(sb->s_flags & MS_RDONLY)) {
1131 status = ocfs2_enable_quotas(osb);
1133 /* We have to err-out specially here because
1134 * s_root is already set */
1136 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1137 wake_up(&osb->osb_mount_event);
1143 ocfs2_complete_quota_recovery(osb);
1145 /* Now we wake up again for processes waiting for quotas */
1146 atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1147 wake_up(&osb->osb_mount_event);
1159 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1160 wake_up(&osb->osb_mount_event);
1161 ocfs2_dismount_volume(sb, 1);
1168 static int ocfs2_get_sb(struct file_system_type *fs_type,
1170 const char *dev_name,
1172 struct vfsmount *mnt)
1174 return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super,
1178 static struct file_system_type ocfs2_fs_type = {
1179 .owner = THIS_MODULE,
1181 .get_sb = ocfs2_get_sb, /* is this called when we mount
1183 .kill_sb = kill_block_super, /* set to the generic one
1184 * right now, but do we
1185 * need to change that? */
1186 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1190 static int ocfs2_parse_options(struct super_block *sb,
1192 struct mount_options *mopt,
1198 mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
1199 options ? options : "(none)");
1201 mopt->commit_interval = 0;
1202 mopt->mount_opt = 0;
1203 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1204 mopt->slot = OCFS2_INVALID_SLOT;
1205 mopt->localalloc_opt = OCFS2_DEFAULT_LOCAL_ALLOC_SIZE;
1206 mopt->cluster_stack[0] = '\0';
1213 while ((p = strsep(&options, ",")) != NULL) {
1215 substring_t args[MAX_OPT_ARGS];
1220 token = match_token(p, tokens, args);
1223 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1226 mopt->mount_opt &= ~OCFS2_MOUNT_HB_LOCAL;
1229 if (match_int(&args[0], &option)) {
1234 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1236 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1239 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1242 mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1245 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1248 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1250 case Opt_data_ordered:
1251 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1253 case Opt_data_writeback:
1254 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1256 case Opt_user_xattr:
1257 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1259 case Opt_nouser_xattr:
1260 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1262 case Opt_atime_quantum:
1263 if (match_int(&args[0], &option)) {
1268 mopt->atime_quantum = option;
1272 if (match_int(&args[0], &option)) {
1277 mopt->slot = (s16)option;
1281 if (match_int(&args[0], &option)) {
1288 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1289 mopt->commit_interval = HZ * option;
1291 case Opt_localalloc:
1293 if (match_int(&args[0], &option)) {
1297 if (option >= 0 && (option <= ocfs2_local_alloc_size(sb) * 8))
1298 mopt->localalloc_opt = option;
1300 case Opt_localflocks:
1302 * Changing this during remount could race
1303 * flock() requests, or "unbalance" existing
1304 * ones (e.g., a lock is taken in one mode but
1305 * dropped in the other). If users care enough
1306 * to flip locking modes during remount, we
1307 * could add a "local" flag to individual
1308 * flock structures for proper tracking of
1312 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1315 /* Check both that the option we were passed
1316 * is of the right length and that it is a proper
1317 * string of the right length.
1319 if (((args[0].to - args[0].from) !=
1320 OCFS2_STACK_LABEL_LEN) ||
1321 (strnlen(args[0].from,
1322 OCFS2_STACK_LABEL_LEN) !=
1323 OCFS2_STACK_LABEL_LEN)) {
1325 "Invalid cluster_stack option\n");
1329 memcpy(mopt->cluster_stack, args[0].from,
1330 OCFS2_STACK_LABEL_LEN);
1331 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1334 mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1337 /* We check only on remount, otherwise features
1338 * aren't yet initialized. */
1339 if (is_remount && !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1340 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1341 mlog(ML_ERROR, "User quota requested but "
1342 "filesystem feature is not set\n");
1346 mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1349 if (is_remount && !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1350 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1351 mlog(ML_ERROR, "Group quota requested but "
1352 "filesystem feature is not set\n");
1356 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1358 #ifdef CONFIG_OCFS2_FS_POSIX_ACL
1360 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1363 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1368 printk(KERN_INFO "ocfs2 (no)acl options not supported\n");
1373 "Unrecognized mount option \"%s\" "
1374 "or missing value\n", p);
1387 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1389 struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb);
1390 unsigned long opts = osb->s_mount_opt;
1391 unsigned int local_alloc_megs;
1393 if (opts & OCFS2_MOUNT_HB_LOCAL)
1394 seq_printf(s, ",_netdev,heartbeat=local");
1396 seq_printf(s, ",heartbeat=none");
1398 if (opts & OCFS2_MOUNT_NOINTR)
1399 seq_printf(s, ",nointr");
1401 if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1402 seq_printf(s, ",data=writeback");
1404 seq_printf(s, ",data=ordered");
1406 if (opts & OCFS2_MOUNT_BARRIER)
1407 seq_printf(s, ",barrier=1");
1409 if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1410 seq_printf(s, ",errors=panic");
1412 seq_printf(s, ",errors=remount-ro");
1414 if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1415 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1417 if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM)
1418 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1420 if (osb->osb_commit_interval)
1421 seq_printf(s, ",commit=%u",
1422 (unsigned) (osb->osb_commit_interval / HZ));
1424 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1425 if (local_alloc_megs != OCFS2_DEFAULT_LOCAL_ALLOC_SIZE)
1426 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1428 if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1429 seq_printf(s, ",localflocks,");
1431 if (osb->osb_cluster_stack[0])
1432 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1433 osb->osb_cluster_stack);
1434 if (opts & OCFS2_MOUNT_USRQUOTA)
1435 seq_printf(s, ",usrquota");
1436 if (opts & OCFS2_MOUNT_GRPQUOTA)
1437 seq_printf(s, ",grpquota");
1439 if (opts & OCFS2_MOUNT_NOUSERXATTR)
1440 seq_printf(s, ",nouser_xattr");
1442 seq_printf(s, ",user_xattr");
1444 if (opts & OCFS2_MOUNT_INODE64)
1445 seq_printf(s, ",inode64");
1447 #ifdef CONFIG_OCFS2_FS_POSIX_ACL
1448 if (opts & OCFS2_MOUNT_POSIX_ACL)
1449 seq_printf(s, ",acl");
1451 seq_printf(s, ",noacl");
1457 static int __init ocfs2_init(void)
1463 ocfs2_print_version();
1465 status = init_ocfs2_uptodate_cache();
1471 status = ocfs2_initialize_mem_caches();
1477 ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1483 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1484 if (!ocfs2_debugfs_root) {
1486 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1489 status = ocfs2_quota_setup();
1493 ocfs2_set_locking_protocol();
1495 status = register_quota_format(&ocfs2_quota_format);
1498 ocfs2_quota_shutdown();
1499 ocfs2_free_mem_caches();
1500 exit_ocfs2_uptodate_cache();
1506 return register_filesystem(&ocfs2_fs_type);
1511 static void __exit ocfs2_exit(void)
1515 ocfs2_quota_shutdown();
1518 flush_workqueue(ocfs2_wq);
1519 destroy_workqueue(ocfs2_wq);
1522 unregister_quota_format(&ocfs2_quota_format);
1524 debugfs_remove(ocfs2_debugfs_root);
1526 ocfs2_free_mem_caches();
1528 unregister_filesystem(&ocfs2_fs_type);
1530 exit_ocfs2_uptodate_cache();
1535 static void ocfs2_put_super(struct super_block *sb)
1537 mlog_entry("(0x%p)\n", sb);
1539 ocfs2_sync_blockdev(sb);
1540 ocfs2_dismount_volume(sb, 0);
1545 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1547 struct ocfs2_super *osb;
1548 u32 numbits, freebits;
1550 struct ocfs2_dinode *bm_lock;
1551 struct buffer_head *bh = NULL;
1552 struct inode *inode = NULL;
1554 mlog_entry("(%p, %p)\n", dentry->d_sb, buf);
1556 osb = OCFS2_SB(dentry->d_sb);
1558 inode = ocfs2_get_system_file_inode(osb,
1559 GLOBAL_BITMAP_SYSTEM_INODE,
1560 OCFS2_INVALID_SLOT);
1562 mlog(ML_ERROR, "failed to get bitmap inode\n");
1567 status = ocfs2_inode_lock(inode, &bh, 0);
1573 bm_lock = (struct ocfs2_dinode *) bh->b_data;
1575 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1576 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1578 buf->f_type = OCFS2_SUPER_MAGIC;
1579 buf->f_bsize = dentry->d_sb->s_blocksize;
1580 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1581 buf->f_blocks = ((sector_t) numbits) *
1582 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1583 buf->f_bfree = ((sector_t) freebits) *
1584 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1585 buf->f_bavail = buf->f_bfree;
1586 buf->f_files = numbits;
1587 buf->f_ffree = freebits;
1591 ocfs2_inode_unlock(inode, 0);
1602 static void ocfs2_inode_init_once(void *data)
1604 struct ocfs2_inode_info *oi = data;
1607 oi->ip_open_count = 0;
1608 spin_lock_init(&oi->ip_lock);
1609 ocfs2_extent_map_init(&oi->vfs_inode);
1610 INIT_LIST_HEAD(&oi->ip_io_markers);
1611 oi->ip_created_trans = 0;
1612 oi->ip_last_trans = 0;
1613 oi->ip_dir_start_lookup = 0;
1615 init_rwsem(&oi->ip_alloc_sem);
1616 init_rwsem(&oi->ip_xattr_sem);
1617 mutex_init(&oi->ip_io_mutex);
1619 oi->ip_blkno = 0ULL;
1620 oi->ip_clusters = 0;
1622 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1623 ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1624 ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1626 ocfs2_metadata_cache_init(&oi->vfs_inode);
1628 inode_init_once(&oi->vfs_inode);
1631 static int ocfs2_initialize_mem_caches(void)
1633 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1634 sizeof(struct ocfs2_inode_info),
1636 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1638 ocfs2_inode_init_once);
1639 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1640 sizeof(struct ocfs2_dquot),
1642 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1645 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1646 sizeof(struct ocfs2_quota_chunk),
1648 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1650 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1651 !ocfs2_qf_chunk_cachep) {
1652 if (ocfs2_inode_cachep)
1653 kmem_cache_destroy(ocfs2_inode_cachep);
1654 if (ocfs2_dquot_cachep)
1655 kmem_cache_destroy(ocfs2_dquot_cachep);
1656 if (ocfs2_qf_chunk_cachep)
1657 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1664 static void ocfs2_free_mem_caches(void)
1666 if (ocfs2_inode_cachep)
1667 kmem_cache_destroy(ocfs2_inode_cachep);
1668 ocfs2_inode_cachep = NULL;
1670 if (ocfs2_dquot_cachep)
1671 kmem_cache_destroy(ocfs2_dquot_cachep);
1672 ocfs2_dquot_cachep = NULL;
1674 if (ocfs2_qf_chunk_cachep)
1675 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1676 ocfs2_qf_chunk_cachep = NULL;
1679 static int ocfs2_get_sector(struct super_block *sb,
1680 struct buffer_head **bh,
1684 if (!sb_set_blocksize(sb, sect_size)) {
1685 mlog(ML_ERROR, "unable to set blocksize\n");
1689 *bh = sb_getblk(sb, block);
1695 if (!buffer_dirty(*bh))
1696 clear_buffer_uptodate(*bh);
1698 ll_rw_block(READ, 1, bh);
1699 wait_on_buffer(*bh);
1700 if (!buffer_uptodate(*bh)) {
1710 static int ocfs2_mount_volume(struct super_block *sb)
1713 int unlock_super = 0;
1714 struct ocfs2_super *osb = OCFS2_SB(sb);
1718 if (ocfs2_is_hard_readonly(osb))
1721 status = ocfs2_dlm_init(osb);
1727 status = ocfs2_super_lock(osb, 1);
1734 /* This will load up the node map and add ourselves to it. */
1735 status = ocfs2_find_slot(osb);
1741 /* load all node-local system inodes */
1742 status = ocfs2_init_local_system_inodes(osb);
1748 status = ocfs2_check_volume(osb);
1754 status = ocfs2_truncate_log_init(osb);
1760 if (ocfs2_mount_local(osb))
1765 ocfs2_super_unlock(osb, 1);
1771 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1773 int tmp, hangup_needed = 0;
1774 struct ocfs2_super *osb = NULL;
1777 mlog_entry("(0x%p)\n", sb);
1783 debugfs_remove(osb->osb_ctxt);
1785 ocfs2_disable_quotas(osb);
1787 ocfs2_shutdown_local_alloc(osb);
1789 ocfs2_truncate_log_shutdown(osb);
1791 /* This will disable recovery and flush any recovery work. */
1792 ocfs2_recovery_exit(osb);
1794 ocfs2_journal_shutdown(osb);
1796 ocfs2_sync_blockdev(sb);
1798 /* No cluster connection means we've failed during mount, so skip
1799 * all the steps which depended on that to complete. */
1801 tmp = ocfs2_super_lock(osb, 1);
1808 if (osb->slot_num != OCFS2_INVALID_SLOT)
1809 ocfs2_put_slot(osb);
1812 ocfs2_super_unlock(osb, 1);
1814 ocfs2_release_system_inodes(osb);
1817 * If we're dismounting due to mount error, mount.ocfs2 will clean
1818 * up heartbeat. If we're a local mount, there is no heartbeat.
1819 * If we failed before we got a uuid_str yet, we can't stop
1820 * heartbeat. Otherwise, do it.
1822 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str)
1826 ocfs2_dlm_shutdown(osb, hangup_needed);
1828 debugfs_remove(osb->osb_debug_root);
1831 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1833 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1835 if (ocfs2_mount_local(osb))
1836 snprintf(nodestr, sizeof(nodestr), "local");
1838 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1840 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1841 osb->dev_str, nodestr);
1843 ocfs2_delete_osb(osb);
1846 sb->s_fs_info = NULL;
1849 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1850 unsigned uuid_bytes)
1855 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1857 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1858 if (osb->uuid_str == NULL)
1861 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1862 /* print with null */
1863 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1864 if (ret != 2) /* drop super cleans up */
1866 /* then only advance past the last char */
1873 static int ocfs2_initialize_super(struct super_block *sb,
1874 struct buffer_head *bh,
1878 int i, cbits, bbits;
1879 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
1880 struct inode *inode = NULL;
1881 struct ocfs2_journal *journal;
1882 __le32 uuid_net_key;
1883 struct ocfs2_super *osb;
1887 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
1894 sb->s_fs_info = osb;
1895 sb->s_op = &ocfs2_sops;
1896 sb->s_export_op = &ocfs2_export_ops;
1897 sb->s_qcop = &ocfs2_quotactl_ops;
1898 sb->dq_op = &ocfs2_quota_operations;
1899 sb->s_xattr = ocfs2_xattr_handlers;
1900 sb->s_time_gran = 1;
1901 sb->s_flags |= MS_NOATIME;
1902 /* this is needed to support O_LARGEFILE */
1903 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
1904 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
1905 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
1907 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
1909 for (i = 0; i < 3; i++)
1910 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
1911 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
1914 /* Save off for ocfs2_rw_direct */
1915 osb->s_sectsize_bits = blksize_bits(sector_size);
1916 BUG_ON(!osb->s_sectsize_bits);
1918 spin_lock_init(&osb->dc_task_lock);
1919 init_waitqueue_head(&osb->dc_event);
1920 osb->dc_work_sequence = 0;
1921 osb->dc_wake_sequence = 0;
1922 INIT_LIST_HEAD(&osb->blocked_lock_list);
1923 osb->blocked_lock_count = 0;
1924 spin_lock_init(&osb->osb_lock);
1925 spin_lock_init(&osb->osb_xattr_lock);
1926 ocfs2_init_inode_steal_slot(osb);
1928 atomic_set(&osb->alloc_stats.moves, 0);
1929 atomic_set(&osb->alloc_stats.local_data, 0);
1930 atomic_set(&osb->alloc_stats.bitmap_data, 0);
1931 atomic_set(&osb->alloc_stats.bg_allocs, 0);
1932 atomic_set(&osb->alloc_stats.bg_extends, 0);
1934 ocfs2_init_node_maps(osb);
1936 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
1937 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
1939 status = ocfs2_recovery_init(osb);
1941 mlog(ML_ERROR, "Unable to initialize recovery state\n");
1946 init_waitqueue_head(&osb->checkpoint_event);
1947 atomic_set(&osb->needs_checkpoint, 0);
1949 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1951 osb->slot_num = OCFS2_INVALID_SLOT;
1953 osb->s_xattr_inline_size = le16_to_cpu(
1954 di->id2.i_super.s_xattr_inline_size);
1956 osb->local_alloc_state = OCFS2_LA_UNUSED;
1957 osb->local_alloc_bh = NULL;
1958 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
1960 init_waitqueue_head(&osb->osb_mount_event);
1962 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
1963 if (!osb->vol_label) {
1964 mlog(ML_ERROR, "unable to alloc vol label\n");
1969 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
1970 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
1971 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
1976 mlog(0, "max_slots for this device: %u\n", osb->max_slots);
1978 osb->slot_recovery_generations =
1979 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
1981 if (!osb->slot_recovery_generations) {
1987 init_waitqueue_head(&osb->osb_wipe_event);
1988 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
1989 sizeof(*osb->osb_orphan_wipes),
1991 if (!osb->osb_orphan_wipes) {
1997 osb->s_feature_compat =
1998 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
1999 osb->s_feature_ro_compat =
2000 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2001 osb->s_feature_incompat =
2002 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2004 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2005 mlog(ML_ERROR, "couldn't mount because of unsupported "
2006 "optional features (%x).\n", i);
2010 if (!(osb->sb->s_flags & MS_RDONLY) &&
2011 (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2012 mlog(ML_ERROR, "couldn't mount RDWR because of "
2013 "unsupported optional features (%x).\n", i);
2018 if (ocfs2_userspace_stack(osb)) {
2019 memcpy(osb->osb_cluster_stack,
2020 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2021 OCFS2_STACK_LABEL_LEN);
2022 osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
2023 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2025 "couldn't mount because of an invalid "
2026 "cluster stack label (%s) \n",
2027 osb->osb_cluster_stack);
2032 /* The empty string is identical with classic tools that
2033 * don't know about s_cluster_info. */
2034 osb->osb_cluster_stack[0] = '\0';
2037 get_random_bytes(&osb->s_next_generation, sizeof(u32));
2040 * This should be done in ocfs2_journal_init(), but unknown
2041 * ordering issues will cause the filesystem to crash.
2042 * If anyone wants to figure out what part of the code
2043 * refers to osb->journal before ocfs2_journal_init() is run,
2046 /* initialize our journal structure */
2048 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2050 mlog(ML_ERROR, "unable to alloc journal\n");
2054 osb->journal = journal;
2055 journal->j_osb = osb;
2057 atomic_set(&journal->j_num_trans, 0);
2058 init_rwsem(&journal->j_trans_barrier);
2059 init_waitqueue_head(&journal->j_checkpointed);
2060 spin_lock_init(&journal->j_lock);
2061 journal->j_trans_id = (unsigned long) 1;
2062 INIT_LIST_HEAD(&journal->j_la_cleanups);
2063 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2064 journal->j_state = OCFS2_JOURNAL_FREE;
2066 INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes);
2067 osb->dentry_lock_list = NULL;
2069 /* get some pseudo constants for clustersize bits */
2070 osb->s_clustersize_bits =
2071 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2072 osb->s_clustersize = 1 << osb->s_clustersize_bits;
2073 mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
2075 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2076 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2077 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2078 osb->s_clustersize);
2083 if (ocfs2_clusters_to_blocks(osb->sb, le32_to_cpu(di->i_clusters) - 1)
2085 mlog(ML_ERROR, "Volume might try to write to blocks beyond "
2086 "what jbd can address in 32 bits.\n");
2091 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2092 sizeof(di->id2.i_super.s_uuid))) {
2093 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2098 memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
2100 strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2101 osb->vol_label[63] = '\0';
2102 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2103 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2104 osb->first_cluster_group_blkno =
2105 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2106 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2107 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2108 mlog(0, "vol_label: %s\n", osb->vol_label);
2109 mlog(0, "uuid: %s\n", osb->uuid_str);
2110 mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
2111 (unsigned long long)osb->root_blkno,
2112 (unsigned long long)osb->system_dir_blkno);
2114 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2115 if (!osb->osb_dlm_debug) {
2121 atomic_set(&osb->vol_state, VOLUME_INIT);
2123 /* load root, system_dir, and all global system inodes */
2124 status = ocfs2_init_global_system_inodes(osb);
2133 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2134 OCFS2_INVALID_SLOT);
2141 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2144 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb) * 8;
2146 status = ocfs2_init_slot_info(osb);
2158 * will return: -EAGAIN if it is ok to keep searching for superblocks
2159 * -EINVAL if there is a bad superblock
2162 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2163 struct buffer_head *bh,
2166 int status = -EAGAIN;
2170 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2171 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2172 /* We have to do a raw check of the feature here */
2173 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2174 OCFS2_FEATURE_INCOMPAT_META_ECC) {
2175 status = ocfs2_block_check_validate(bh->b_data,
2182 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2183 mlog(ML_ERROR, "found superblock with incorrect block "
2184 "size: found %u, should be %u\n",
2185 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2187 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2188 OCFS2_MAJOR_REV_LEVEL ||
2189 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2190 OCFS2_MINOR_REV_LEVEL) {
2191 mlog(ML_ERROR, "found superblock with bad version: "
2192 "found %u.%u, should be %u.%u\n",
2193 le16_to_cpu(di->id2.i_super.s_major_rev_level),
2194 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2195 OCFS2_MAJOR_REV_LEVEL,
2196 OCFS2_MINOR_REV_LEVEL);
2197 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2198 mlog(ML_ERROR, "bad block number on superblock: "
2199 "found %llu, should be %llu\n",
2200 (unsigned long long)le64_to_cpu(di->i_blkno),
2201 (unsigned long long)bh->b_blocknr);
2202 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2203 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2204 mlog(ML_ERROR, "bad cluster size found: %u\n",
2205 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2206 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2207 mlog(ML_ERROR, "bad root_blkno: 0\n");
2208 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2209 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2210 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2212 "Superblock slots found greater than file system "
2213 "maximum: found %u, max %u\n",
2214 le16_to_cpu(di->id2.i_super.s_max_slots),
2227 static int ocfs2_check_volume(struct ocfs2_super *osb)
2232 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2238 /* Init our journal object. */
2239 status = ocfs2_journal_init(osb->journal, &dirty);
2241 mlog(ML_ERROR, "Could not initialize journal!\n");
2245 /* If the journal was unmounted cleanly then we don't want to
2246 * recover anything. Otherwise, journal_load will do that
2247 * dirty work for us :) */
2249 status = ocfs2_journal_wipe(osb->journal, 0);
2255 mlog(ML_NOTICE, "File system was not unmounted cleanly, "
2256 "recovering volume.\n");
2259 local = ocfs2_mount_local(osb);
2261 /* will play back anything left in the journal. */
2262 status = ocfs2_journal_load(osb->journal, local, dirty);
2264 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2269 /* recover my local alloc if we didn't unmount cleanly. */
2270 status = ocfs2_begin_local_alloc_recovery(osb,
2277 /* we complete the recovery process after we've marked
2278 * ourselves as mounted. */
2281 mlog(0, "Journal loaded.\n");
2283 status = ocfs2_load_local_alloc(osb);
2290 /* Recovery will be completed after we've mounted the
2291 * rest of the volume. */
2293 osb->local_alloc_copy = local_alloc;
2297 /* go through each journal, trylock it and if you get the
2298 * lock, and it's marked as dirty, set the bit in the recover
2299 * map and launch a recovery thread for it. */
2300 status = ocfs2_mark_dead_nodes(osb);
2306 status = ocfs2_compute_replay_slots(osb);
2319 * The routine gets called from dismount or close whenever a dismount on
2320 * volume is requested and the osb open count becomes 1.
2321 * It will remove the osb from the global list and also free up all the
2322 * initialized resources and fileobject.
2324 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2328 /* This function assumes that the caller has the main osb resource */
2330 ocfs2_free_slot_info(osb);
2332 kfree(osb->osb_orphan_wipes);
2333 kfree(osb->slot_recovery_generations);
2335 * This belongs in journal shutdown, but because we have to
2336 * allocate osb->journal at the start of ocfs2_initalize_osb(),
2339 kfree(osb->journal);
2340 if (osb->local_alloc_copy)
2341 kfree(osb->local_alloc_copy);
2342 kfree(osb->uuid_str);
2343 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2344 memset(osb, 0, sizeof(struct ocfs2_super));
2349 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2350 * panic(). We do not support continue-on-error operation. */
2351 static void ocfs2_handle_error(struct super_block *sb)
2353 struct ocfs2_super *osb = OCFS2_SB(sb);
2355 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2356 panic("OCFS2: (device %s): panic forced after error\n",
2359 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2361 if (sb->s_flags & MS_RDONLY &&
2362 (ocfs2_is_soft_readonly(osb) ||
2363 ocfs2_is_hard_readonly(osb)))
2366 printk(KERN_CRIT "File system is now read-only due to the potential "
2367 "of on-disk corruption. Please run fsck.ocfs2 once the file "
2368 "system is unmounted.\n");
2369 sb->s_flags |= MS_RDONLY;
2370 ocfs2_set_ro_flag(osb, 0);
2373 static char error_buf[1024];
2375 void __ocfs2_error(struct super_block *sb,
2376 const char *function,
2377 const char *fmt, ...)
2381 va_start(args, fmt);
2382 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2385 /* Not using mlog here because we want to show the actual
2386 * function the error came from. */
2387 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2388 sb->s_id, function, error_buf);
2390 ocfs2_handle_error(sb);
2393 /* Handle critical errors. This is intentionally more drastic than
2394 * ocfs2_handle_error, so we only use for things like journal errors,
2396 void __ocfs2_abort(struct super_block* sb,
2397 const char *function,
2398 const char *fmt, ...)
2402 va_start(args, fmt);
2403 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2406 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2407 sb->s_id, function, error_buf);
2409 /* We don't have the cluster support yet to go straight to
2410 * hard readonly in here. Until then, we want to keep
2411 * ocfs2_abort() so that we can at least mark critical
2414 * TODO: This should abort the journal and alert other nodes
2415 * that our slot needs recovery. */
2417 /* Force a panic(). This stinks, but it's better than letting
2418 * things continue without having a proper hard readonly
2420 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2421 ocfs2_handle_error(sb);
2424 module_init(ocfs2_init);
2425 module_exit(ocfs2_exit);