xref: /openbmc/linux/fs/ocfs2/super.c (revision df2634f43f5106947f3735a0b61a6527a4b278cd)
1 /* -*- mode: c; c-basic-offset: 8; -*-
2  * vim: noexpandtab sw=8 ts=8 sts=0:
3  *
4  * super.c
5  *
6  * load/unload driver, mount/dismount volumes
7  *
8  * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
9  *
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.
14  *
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.
19  *
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.
24  */
25 
26 #include <linux/module.h>
27 #include <linux/fs.h>
28 #include <linux/types.h>
29 #include <linux/slab.h>
30 #include <linux/highmem.h>
31 #include <linux/init.h>
32 #include <linux/random.h>
33 #include <linux/statfs.h>
34 #include <linux/moduleparam.h>
35 #include <linux/blkdev.h>
36 #include <linux/socket.h>
37 #include <linux/inet.h>
38 #include <linux/parser.h>
39 #include <linux/crc32.h>
40 #include <linux/debugfs.h>
41 #include <linux/mount.h>
42 #include <linux/seq_file.h>
43 #include <linux/quotaops.h>
44 
45 #define MLOG_MASK_PREFIX ML_SUPER
46 #include <cluster/masklog.h>
47 
48 #include "ocfs2.h"
49 
50 /* this should be the only file to include a version 1 header */
51 #include "ocfs1_fs_compat.h"
52 
53 #include "alloc.h"
54 #include "blockcheck.h"
55 #include "dlmglue.h"
56 #include "export.h"
57 #include "extent_map.h"
58 #include "heartbeat.h"
59 #include "inode.h"
60 #include "journal.h"
61 #include "localalloc.h"
62 #include "namei.h"
63 #include "slot_map.h"
64 #include "super.h"
65 #include "sysfile.h"
66 #include "uptodate.h"
67 #include "ver.h"
68 #include "xattr.h"
69 #include "quota.h"
70 #include "refcounttree.h"
71 #include "suballoc.h"
72 
73 #include "buffer_head_io.h"
74 
75 static struct kmem_cache *ocfs2_inode_cachep = NULL;
76 struct kmem_cache *ocfs2_dquot_cachep;
77 struct kmem_cache *ocfs2_qf_chunk_cachep;
78 
79 /* OCFS2 needs to schedule several differnt types of work which
80  * require cluster locking, disk I/O, recovery waits, etc. Since these
81  * types of work tend to be heavy we avoid using the kernel events
82  * workqueue and schedule on our own. */
83 struct workqueue_struct *ocfs2_wq = NULL;
84 
85 static struct dentry *ocfs2_debugfs_root = NULL;
86 
87 MODULE_AUTHOR("Oracle");
88 MODULE_LICENSE("GPL");
89 
90 struct mount_options
91 {
92 	unsigned long	commit_interval;
93 	unsigned long	mount_opt;
94 	unsigned int	atime_quantum;
95 	signed short	slot;
96 	int		localalloc_opt;
97 	unsigned int	resv_level;
98 	int		dir_resv_level;
99 	char		cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
100 };
101 
102 static int ocfs2_parse_options(struct super_block *sb, char *options,
103 			       struct mount_options *mopt,
104 			       int is_remount);
105 static int ocfs2_check_set_options(struct super_block *sb,
106 				   struct mount_options *options);
107 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt);
108 static void ocfs2_put_super(struct super_block *sb);
109 static int ocfs2_mount_volume(struct super_block *sb);
110 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
111 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
112 static int ocfs2_initialize_mem_caches(void);
113 static void ocfs2_free_mem_caches(void);
114 static void ocfs2_delete_osb(struct ocfs2_super *osb);
115 
116 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
117 
118 static int ocfs2_sync_fs(struct super_block *sb, int wait);
119 
120 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
121 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
122 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
123 static int ocfs2_check_volume(struct ocfs2_super *osb);
124 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
125 			       struct buffer_head *bh,
126 			       u32 sectsize,
127 			       struct ocfs2_blockcheck_stats *stats);
128 static int ocfs2_initialize_super(struct super_block *sb,
129 				  struct buffer_head *bh,
130 				  int sector_size,
131 				  struct ocfs2_blockcheck_stats *stats);
132 static int ocfs2_get_sector(struct super_block *sb,
133 			    struct buffer_head **bh,
134 			    int block,
135 			    int sect_size);
136 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
137 static void ocfs2_destroy_inode(struct inode *inode);
138 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
139 static int ocfs2_enable_quotas(struct ocfs2_super *osb);
140 static void ocfs2_disable_quotas(struct ocfs2_super *osb);
141 
142 static const struct super_operations ocfs2_sops = {
143 	.statfs		= ocfs2_statfs,
144 	.alloc_inode	= ocfs2_alloc_inode,
145 	.destroy_inode	= ocfs2_destroy_inode,
146 	.drop_inode	= ocfs2_drop_inode,
147 	.evict_inode	= ocfs2_evict_inode,
148 	.sync_fs	= ocfs2_sync_fs,
149 	.put_super	= ocfs2_put_super,
150 	.remount_fs	= ocfs2_remount,
151 	.show_options   = ocfs2_show_options,
152 	.quota_read	= ocfs2_quota_read,
153 	.quota_write	= ocfs2_quota_write,
154 };
155 
156 enum {
157 	Opt_barrier,
158 	Opt_err_panic,
159 	Opt_err_ro,
160 	Opt_intr,
161 	Opt_nointr,
162 	Opt_hb_none,
163 	Opt_hb_local,
164 	Opt_hb_global,
165 	Opt_data_ordered,
166 	Opt_data_writeback,
167 	Opt_atime_quantum,
168 	Opt_slot,
169 	Opt_commit,
170 	Opt_localalloc,
171 	Opt_localflocks,
172 	Opt_stack,
173 	Opt_user_xattr,
174 	Opt_nouser_xattr,
175 	Opt_inode64,
176 	Opt_acl,
177 	Opt_noacl,
178 	Opt_usrquota,
179 	Opt_grpquota,
180 	Opt_coherency_buffered,
181 	Opt_coherency_full,
182 	Opt_resv_level,
183 	Opt_dir_resv_level,
184 	Opt_err,
185 };
186 
187 static const match_table_t tokens = {
188 	{Opt_barrier, "barrier=%u"},
189 	{Opt_err_panic, "errors=panic"},
190 	{Opt_err_ro, "errors=remount-ro"},
191 	{Opt_intr, "intr"},
192 	{Opt_nointr, "nointr"},
193 	{Opt_hb_none, OCFS2_HB_NONE},
194 	{Opt_hb_local, OCFS2_HB_LOCAL},
195 	{Opt_hb_global, OCFS2_HB_GLOBAL},
196 	{Opt_data_ordered, "data=ordered"},
197 	{Opt_data_writeback, "data=writeback"},
198 	{Opt_atime_quantum, "atime_quantum=%u"},
199 	{Opt_slot, "preferred_slot=%u"},
200 	{Opt_commit, "commit=%u"},
201 	{Opt_localalloc, "localalloc=%d"},
202 	{Opt_localflocks, "localflocks"},
203 	{Opt_stack, "cluster_stack=%s"},
204 	{Opt_user_xattr, "user_xattr"},
205 	{Opt_nouser_xattr, "nouser_xattr"},
206 	{Opt_inode64, "inode64"},
207 	{Opt_acl, "acl"},
208 	{Opt_noacl, "noacl"},
209 	{Opt_usrquota, "usrquota"},
210 	{Opt_grpquota, "grpquota"},
211 	{Opt_coherency_buffered, "coherency=buffered"},
212 	{Opt_coherency_full, "coherency=full"},
213 	{Opt_resv_level, "resv_level=%u"},
214 	{Opt_dir_resv_level, "dir_resv_level=%u"},
215 	{Opt_err, NULL}
216 };
217 
218 #ifdef CONFIG_DEBUG_FS
219 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
220 {
221 	struct ocfs2_cluster_connection *cconn = osb->cconn;
222 	struct ocfs2_recovery_map *rm = osb->recovery_map;
223 	struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
224 	int i, out = 0;
225 
226 	out += snprintf(buf + out, len - out,
227 			"%10s => Id: %-s  Uuid: %-s  Gen: 0x%X  Label: %-s\n",
228 			"Device", osb->dev_str, osb->uuid_str,
229 			osb->fs_generation, osb->vol_label);
230 
231 	out += snprintf(buf + out, len - out,
232 			"%10s => State: %d  Flags: 0x%lX\n", "Volume",
233 			atomic_read(&osb->vol_state), osb->osb_flags);
234 
235 	out += snprintf(buf + out, len - out,
236 			"%10s => Block: %lu  Cluster: %d\n", "Sizes",
237 			osb->sb->s_blocksize, osb->s_clustersize);
238 
239 	out += snprintf(buf + out, len - out,
240 			"%10s => Compat: 0x%X  Incompat: 0x%X  "
241 			"ROcompat: 0x%X\n",
242 			"Features", osb->s_feature_compat,
243 			osb->s_feature_incompat, osb->s_feature_ro_compat);
244 
245 	out += snprintf(buf + out, len - out,
246 			"%10s => Opts: 0x%lX  AtimeQuanta: %u\n", "Mount",
247 			osb->s_mount_opt, osb->s_atime_quantum);
248 
249 	if (cconn) {
250 		out += snprintf(buf + out, len - out,
251 				"%10s => Stack: %s  Name: %*s  "
252 				"Version: %d.%d\n", "Cluster",
253 				(*osb->osb_cluster_stack == '\0' ?
254 				 "o2cb" : osb->osb_cluster_stack),
255 				cconn->cc_namelen, cconn->cc_name,
256 				cconn->cc_version.pv_major,
257 				cconn->cc_version.pv_minor);
258 	}
259 
260 	spin_lock(&osb->dc_task_lock);
261 	out += snprintf(buf + out, len - out,
262 			"%10s => Pid: %d  Count: %lu  WakeSeq: %lu  "
263 			"WorkSeq: %lu\n", "DownCnvt",
264 			(osb->dc_task ?  task_pid_nr(osb->dc_task) : -1),
265 			osb->blocked_lock_count, osb->dc_wake_sequence,
266 			osb->dc_work_sequence);
267 	spin_unlock(&osb->dc_task_lock);
268 
269 	spin_lock(&osb->osb_lock);
270 	out += snprintf(buf + out, len - out, "%10s => Pid: %d  Nodes:",
271 			"Recovery",
272 			(osb->recovery_thread_task ?
273 			 task_pid_nr(osb->recovery_thread_task) : -1));
274 	if (rm->rm_used == 0)
275 		out += snprintf(buf + out, len - out, " None\n");
276 	else {
277 		for (i = 0; i < rm->rm_used; i++)
278 			out += snprintf(buf + out, len - out, " %d",
279 					rm->rm_entries[i]);
280 		out += snprintf(buf + out, len - out, "\n");
281 	}
282 	spin_unlock(&osb->osb_lock);
283 
284 	out += snprintf(buf + out, len - out,
285 			"%10s => Pid: %d  Interval: %lu  Needs: %d\n", "Commit",
286 			(osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
287 			osb->osb_commit_interval,
288 			atomic_read(&osb->needs_checkpoint));
289 
290 	out += snprintf(buf + out, len - out,
291 			"%10s => State: %d  TxnId: %lu  NumTxns: %d\n",
292 			"Journal", osb->journal->j_state,
293 			osb->journal->j_trans_id,
294 			atomic_read(&osb->journal->j_num_trans));
295 
296 	out += snprintf(buf + out, len - out,
297 			"%10s => GlobalAllocs: %d  LocalAllocs: %d  "
298 			"SubAllocs: %d  LAWinMoves: %d  SAExtends: %d\n",
299 			"Stats",
300 			atomic_read(&osb->alloc_stats.bitmap_data),
301 			atomic_read(&osb->alloc_stats.local_data),
302 			atomic_read(&osb->alloc_stats.bg_allocs),
303 			atomic_read(&osb->alloc_stats.moves),
304 			atomic_read(&osb->alloc_stats.bg_extends));
305 
306 	out += snprintf(buf + out, len - out,
307 			"%10s => State: %u  Descriptor: %llu  Size: %u bits  "
308 			"Default: %u bits\n",
309 			"LocalAlloc", osb->local_alloc_state,
310 			(unsigned long long)osb->la_last_gd,
311 			osb->local_alloc_bits, osb->local_alloc_default_bits);
312 
313 	spin_lock(&osb->osb_lock);
314 	out += snprintf(buf + out, len - out,
315 			"%10s => InodeSlot: %d  StolenInodes: %d, "
316 			"MetaSlot: %d  StolenMeta: %d\n", "Steal",
317 			osb->s_inode_steal_slot,
318 			atomic_read(&osb->s_num_inodes_stolen),
319 			osb->s_meta_steal_slot,
320 			atomic_read(&osb->s_num_meta_stolen));
321 	spin_unlock(&osb->osb_lock);
322 
323 	out += snprintf(buf + out, len - out, "OrphanScan => ");
324 	out += snprintf(buf + out, len - out, "Local: %u  Global: %u ",
325 			os->os_count, os->os_seqno);
326 	out += snprintf(buf + out, len - out, " Last Scan: ");
327 	if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
328 		out += snprintf(buf + out, len - out, "Disabled\n");
329 	else
330 		out += snprintf(buf + out, len - out, "%lu seconds ago\n",
331 				(get_seconds() - os->os_scantime.tv_sec));
332 
333 	out += snprintf(buf + out, len - out, "%10s => %3s  %10s\n",
334 			"Slots", "Num", "RecoGen");
335 	for (i = 0; i < osb->max_slots; ++i) {
336 		out += snprintf(buf + out, len - out,
337 				"%10s  %c %3d  %10d\n",
338 				" ",
339 				(i == osb->slot_num ? '*' : ' '),
340 				i, osb->slot_recovery_generations[i]);
341 	}
342 
343 	return out;
344 }
345 
346 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
347 {
348 	struct ocfs2_super *osb = inode->i_private;
349 	char *buf = NULL;
350 
351 	buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
352 	if (!buf)
353 		goto bail;
354 
355 	i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
356 
357 	file->private_data = buf;
358 
359 	return 0;
360 bail:
361 	return -ENOMEM;
362 }
363 
364 static int ocfs2_debug_release(struct inode *inode, struct file *file)
365 {
366 	kfree(file->private_data);
367 	return 0;
368 }
369 
370 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
371 				size_t nbytes, loff_t *ppos)
372 {
373 	return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
374 				       i_size_read(file->f_mapping->host));
375 }
376 #else
377 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
378 {
379 	return 0;
380 }
381 static int ocfs2_debug_release(struct inode *inode, struct file *file)
382 {
383 	return 0;
384 }
385 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
386 				size_t nbytes, loff_t *ppos)
387 {
388 	return 0;
389 }
390 #endif	/* CONFIG_DEBUG_FS */
391 
392 static const struct file_operations ocfs2_osb_debug_fops = {
393 	.open =		ocfs2_osb_debug_open,
394 	.release =	ocfs2_debug_release,
395 	.read =		ocfs2_debug_read,
396 	.llseek =	generic_file_llseek,
397 };
398 
399 static int ocfs2_sync_fs(struct super_block *sb, int wait)
400 {
401 	int status;
402 	tid_t target;
403 	struct ocfs2_super *osb = OCFS2_SB(sb);
404 
405 	if (ocfs2_is_hard_readonly(osb))
406 		return -EROFS;
407 
408 	if (wait) {
409 		status = ocfs2_flush_truncate_log(osb);
410 		if (status < 0)
411 			mlog_errno(status);
412 	} else {
413 		ocfs2_schedule_truncate_log_flush(osb, 0);
414 	}
415 
416 	if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
417 				      &target)) {
418 		if (wait)
419 			jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
420 					     target);
421 	}
422 	return 0;
423 }
424 
425 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
426 {
427 	if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
428 	    && (ino == USER_QUOTA_SYSTEM_INODE
429 		|| ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
430 		return 0;
431 	if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
432 	    && (ino == GROUP_QUOTA_SYSTEM_INODE
433 		|| ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
434 		return 0;
435 	return 1;
436 }
437 
438 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
439 {
440 	struct inode *new = NULL;
441 	int status = 0;
442 	int i;
443 
444 	mlog_entry_void();
445 
446 	new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
447 	if (IS_ERR(new)) {
448 		status = PTR_ERR(new);
449 		mlog_errno(status);
450 		goto bail;
451 	}
452 	osb->root_inode = new;
453 
454 	new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
455 	if (IS_ERR(new)) {
456 		status = PTR_ERR(new);
457 		mlog_errno(status);
458 		goto bail;
459 	}
460 	osb->sys_root_inode = new;
461 
462 	for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
463 	     i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
464 		if (!ocfs2_need_system_inode(osb, i))
465 			continue;
466 		new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
467 		if (!new) {
468 			ocfs2_release_system_inodes(osb);
469 			status = -EINVAL;
470 			mlog_errno(status);
471 			/* FIXME: Should ERROR_RO_FS */
472 			mlog(ML_ERROR, "Unable to load system inode %d, "
473 			     "possibly corrupt fs?", i);
474 			goto bail;
475 		}
476 		// the array now has one ref, so drop this one
477 		iput(new);
478 	}
479 
480 bail:
481 	mlog_exit(status);
482 	return status;
483 }
484 
485 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
486 {
487 	struct inode *new = NULL;
488 	int status = 0;
489 	int i;
490 
491 	mlog_entry_void();
492 
493 	for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
494 	     i < NUM_SYSTEM_INODES;
495 	     i++) {
496 		if (!ocfs2_need_system_inode(osb, i))
497 			continue;
498 		new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
499 		if (!new) {
500 			ocfs2_release_system_inodes(osb);
501 			status = -EINVAL;
502 			mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
503 			     status, i, osb->slot_num);
504 			goto bail;
505 		}
506 		/* the array now has one ref, so drop this one */
507 		iput(new);
508 	}
509 
510 bail:
511 	mlog_exit(status);
512 	return status;
513 }
514 
515 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
516 {
517 	int i;
518 	struct inode *inode;
519 
520 	mlog_entry_void();
521 
522 	for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
523 		inode = osb->global_system_inodes[i];
524 		if (inode) {
525 			iput(inode);
526 			osb->global_system_inodes[i] = NULL;
527 		}
528 	}
529 
530 	inode = osb->sys_root_inode;
531 	if (inode) {
532 		iput(inode);
533 		osb->sys_root_inode = NULL;
534 	}
535 
536 	inode = osb->root_inode;
537 	if (inode) {
538 		iput(inode);
539 		osb->root_inode = NULL;
540 	}
541 
542 	if (!osb->local_system_inodes)
543 		goto out;
544 
545 	for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
546 		if (osb->local_system_inodes[i]) {
547 			iput(osb->local_system_inodes[i]);
548 			osb->local_system_inodes[i] = NULL;
549 		}
550 	}
551 
552 	kfree(osb->local_system_inodes);
553 	osb->local_system_inodes = NULL;
554 
555 out:
556 	mlog_exit(0);
557 }
558 
559 /* We're allocating fs objects, use GFP_NOFS */
560 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
561 {
562 	struct ocfs2_inode_info *oi;
563 
564 	oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
565 	if (!oi)
566 		return NULL;
567 
568 	jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
569 	return &oi->vfs_inode;
570 }
571 
572 static void ocfs2_i_callback(struct rcu_head *head)
573 {
574 	struct inode *inode = container_of(head, struct inode, i_rcu);
575 	INIT_LIST_HEAD(&inode->i_dentry);
576 	kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
577 }
578 
579 static void ocfs2_destroy_inode(struct inode *inode)
580 {
581 	call_rcu(&inode->i_rcu, ocfs2_i_callback);
582 }
583 
584 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
585 						unsigned int cbits)
586 {
587 	unsigned int bytes = 1 << cbits;
588 	unsigned int trim = bytes;
589 	unsigned int bitshift = 32;
590 
591 	/*
592 	 * i_size and all block offsets in ocfs2 are always 64 bits
593 	 * wide. i_clusters is 32 bits, in cluster-sized units. So on
594 	 * 64 bit platforms, cluster size will be the limiting factor.
595 	 */
596 
597 #if BITS_PER_LONG == 32
598 # if defined(CONFIG_LBDAF)
599 	BUILD_BUG_ON(sizeof(sector_t) != 8);
600 	/*
601 	 * We might be limited by page cache size.
602 	 */
603 	if (bytes > PAGE_CACHE_SIZE) {
604 		bytes = PAGE_CACHE_SIZE;
605 		trim = 1;
606 		/*
607 		 * Shift by 31 here so that we don't get larger than
608 		 * MAX_LFS_FILESIZE
609 		 */
610 		bitshift = 31;
611 	}
612 # else
613 	/*
614 	 * We are limited by the size of sector_t. Use block size, as
615 	 * that's what we expose to the VFS.
616 	 */
617 	bytes = 1 << bbits;
618 	trim = 1;
619 	bitshift = 31;
620 # endif
621 #endif
622 
623 	/*
624 	 * Trim by a whole cluster when we can actually approach the
625 	 * on-disk limits. Otherwise we can overflow i_clusters when
626 	 * an extent start is at the max offset.
627 	 */
628 	return (((unsigned long long)bytes) << bitshift) - trim;
629 }
630 
631 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
632 {
633 	int incompat_features;
634 	int ret = 0;
635 	struct mount_options parsed_options;
636 	struct ocfs2_super *osb = OCFS2_SB(sb);
637 	u32 tmp;
638 
639 	if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
640 	    !ocfs2_check_set_options(sb, &parsed_options)) {
641 		ret = -EINVAL;
642 		goto out;
643 	}
644 
645 	tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
646 		OCFS2_MOUNT_HB_NONE;
647 	if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
648 		ret = -EINVAL;
649 		mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
650 		goto out;
651 	}
652 
653 	if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
654 	    (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
655 		ret = -EINVAL;
656 		mlog(ML_ERROR, "Cannot change data mode on remount\n");
657 		goto out;
658 	}
659 
660 	/* Probably don't want this on remount; it might
661 	 * mess with other nodes */
662 	if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
663 	    (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
664 		ret = -EINVAL;
665 		mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
666 		goto out;
667 	}
668 
669 	/* We're going to/from readonly mode. */
670 	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
671 		/* Disable quota accounting before remounting RO */
672 		if (*flags & MS_RDONLY) {
673 			ret = ocfs2_susp_quotas(osb, 0);
674 			if (ret < 0)
675 				goto out;
676 		}
677 		/* Lock here so the check of HARD_RO and the potential
678 		 * setting of SOFT_RO is atomic. */
679 		spin_lock(&osb->osb_lock);
680 		if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
681 			mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
682 			ret = -EROFS;
683 			goto unlock_osb;
684 		}
685 
686 		if (*flags & MS_RDONLY) {
687 			mlog(0, "Going to ro mode.\n");
688 			sb->s_flags |= MS_RDONLY;
689 			osb->osb_flags |= OCFS2_OSB_SOFT_RO;
690 		} else {
691 			mlog(0, "Making ro filesystem writeable.\n");
692 
693 			if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
694 				mlog(ML_ERROR, "Cannot remount RDWR "
695 				     "filesystem due to previous errors.\n");
696 				ret = -EROFS;
697 				goto unlock_osb;
698 			}
699 			incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
700 			if (incompat_features) {
701 				mlog(ML_ERROR, "Cannot remount RDWR because "
702 				     "of unsupported optional features "
703 				     "(%x).\n", incompat_features);
704 				ret = -EINVAL;
705 				goto unlock_osb;
706 			}
707 			sb->s_flags &= ~MS_RDONLY;
708 			osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
709 		}
710 unlock_osb:
711 		spin_unlock(&osb->osb_lock);
712 		/* Enable quota accounting after remounting RW */
713 		if (!ret && !(*flags & MS_RDONLY)) {
714 			if (sb_any_quota_suspended(sb))
715 				ret = ocfs2_susp_quotas(osb, 1);
716 			else
717 				ret = ocfs2_enable_quotas(osb);
718 			if (ret < 0) {
719 				/* Return back changes... */
720 				spin_lock(&osb->osb_lock);
721 				sb->s_flags |= MS_RDONLY;
722 				osb->osb_flags |= OCFS2_OSB_SOFT_RO;
723 				spin_unlock(&osb->osb_lock);
724 				goto out;
725 			}
726 		}
727 	}
728 
729 	if (!ret) {
730 		/* Only save off the new mount options in case of a successful
731 		 * remount. */
732 		osb->s_mount_opt = parsed_options.mount_opt;
733 		osb->s_atime_quantum = parsed_options.atime_quantum;
734 		osb->preferred_slot = parsed_options.slot;
735 		if (parsed_options.commit_interval)
736 			osb->osb_commit_interval = parsed_options.commit_interval;
737 
738 		if (!ocfs2_is_hard_readonly(osb))
739 			ocfs2_set_journal_params(osb);
740 
741 		sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
742 			((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
743 							MS_POSIXACL : 0);
744 	}
745 out:
746 	return ret;
747 }
748 
749 static int ocfs2_sb_probe(struct super_block *sb,
750 			  struct buffer_head **bh,
751 			  int *sector_size,
752 			  struct ocfs2_blockcheck_stats *stats)
753 {
754 	int status, tmpstat;
755 	struct ocfs1_vol_disk_hdr *hdr;
756 	struct ocfs2_dinode *di;
757 	int blksize;
758 
759 	*bh = NULL;
760 
761 	/* may be > 512 */
762 	*sector_size = bdev_logical_block_size(sb->s_bdev);
763 	if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
764 		mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
765 		     *sector_size, OCFS2_MAX_BLOCKSIZE);
766 		status = -EINVAL;
767 		goto bail;
768 	}
769 
770 	/* Can this really happen? */
771 	if (*sector_size < OCFS2_MIN_BLOCKSIZE)
772 		*sector_size = OCFS2_MIN_BLOCKSIZE;
773 
774 	/* check block zero for old format */
775 	status = ocfs2_get_sector(sb, bh, 0, *sector_size);
776 	if (status < 0) {
777 		mlog_errno(status);
778 		goto bail;
779 	}
780 	hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
781 	if (hdr->major_version == OCFS1_MAJOR_VERSION) {
782 		mlog(ML_ERROR, "incompatible version: %u.%u\n",
783 		     hdr->major_version, hdr->minor_version);
784 		status = -EINVAL;
785 	}
786 	if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
787 		   strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
788 		mlog(ML_ERROR, "incompatible volume signature: %8s\n",
789 		     hdr->signature);
790 		status = -EINVAL;
791 	}
792 	brelse(*bh);
793 	*bh = NULL;
794 	if (status < 0) {
795 		mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
796 		     "upgraded before mounting with ocfs v2\n");
797 		goto bail;
798 	}
799 
800 	/*
801 	 * Now check at magic offset for 512, 1024, 2048, 4096
802 	 * blocksizes.  4096 is the maximum blocksize because it is
803 	 * the minimum clustersize.
804 	 */
805 	status = -EINVAL;
806 	for (blksize = *sector_size;
807 	     blksize <= OCFS2_MAX_BLOCKSIZE;
808 	     blksize <<= 1) {
809 		tmpstat = ocfs2_get_sector(sb, bh,
810 					   OCFS2_SUPER_BLOCK_BLKNO,
811 					   blksize);
812 		if (tmpstat < 0) {
813 			status = tmpstat;
814 			mlog_errno(status);
815 			break;
816 		}
817 		di = (struct ocfs2_dinode *) (*bh)->b_data;
818 		memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
819 		spin_lock_init(&stats->b_lock);
820 		tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
821 		if (tmpstat < 0) {
822 			brelse(*bh);
823 			*bh = NULL;
824 		}
825 		if (tmpstat != -EAGAIN) {
826 			status = tmpstat;
827 			break;
828 		}
829 	}
830 
831 bail:
832 	return status;
833 }
834 
835 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
836 {
837 	u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
838 
839 	if (osb->s_mount_opt & hb_enabled) {
840 		if (ocfs2_mount_local(osb)) {
841 			mlog(ML_ERROR, "Cannot heartbeat on a locally "
842 			     "mounted device.\n");
843 			return -EINVAL;
844 		}
845 		if (ocfs2_userspace_stack(osb)) {
846 			mlog(ML_ERROR, "Userspace stack expected, but "
847 			     "o2cb heartbeat arguments passed to mount\n");
848 			return -EINVAL;
849 		}
850 		if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
851 		     !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
852 		    ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
853 		     ocfs2_cluster_o2cb_global_heartbeat(osb))) {
854 			mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
855 			return -EINVAL;
856 		}
857 	}
858 
859 	if (!(osb->s_mount_opt & hb_enabled)) {
860 		if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
861 		    !ocfs2_userspace_stack(osb)) {
862 			mlog(ML_ERROR, "Heartbeat has to be started to mount "
863 			     "a read-write clustered device.\n");
864 			return -EINVAL;
865 		}
866 	}
867 
868 	return 0;
869 }
870 
871 /*
872  * If we're using a userspace stack, mount should have passed
873  * a name that matches the disk.  If not, mount should not
874  * have passed a stack.
875  */
876 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
877 					struct mount_options *mopt)
878 {
879 	if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
880 		mlog(ML_ERROR,
881 		     "cluster stack passed to mount, but this filesystem "
882 		     "does not support it\n");
883 		return -EINVAL;
884 	}
885 
886 	if (ocfs2_userspace_stack(osb) &&
887 	    strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
888 		    OCFS2_STACK_LABEL_LEN)) {
889 		mlog(ML_ERROR,
890 		     "cluster stack passed to mount (\"%s\") does not "
891 		     "match the filesystem (\"%s\")\n",
892 		     mopt->cluster_stack,
893 		     osb->osb_cluster_stack);
894 		return -EINVAL;
895 	}
896 
897 	return 0;
898 }
899 
900 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
901 {
902 	int type;
903 	struct super_block *sb = osb->sb;
904 	unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
905 					     OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
906 	int status = 0;
907 
908 	for (type = 0; type < MAXQUOTAS; type++) {
909 		if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
910 			continue;
911 		if (unsuspend)
912 			status = dquot_resume(sb, type);
913 		else {
914 			struct ocfs2_mem_dqinfo *oinfo;
915 
916 			/* Cancel periodic syncing before suspending */
917 			oinfo = sb_dqinfo(sb, type)->dqi_priv;
918 			cancel_delayed_work_sync(&oinfo->dqi_sync_work);
919 			status = dquot_suspend(sb, type);
920 		}
921 		if (status < 0)
922 			break;
923 	}
924 	if (status < 0)
925 		mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
926 		     "remount (error = %d).\n", status);
927 	return status;
928 }
929 
930 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
931 {
932 	struct inode *inode[MAXQUOTAS] = { NULL, NULL };
933 	struct super_block *sb = osb->sb;
934 	unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
935 					     OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
936 	unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
937 					LOCAL_GROUP_QUOTA_SYSTEM_INODE };
938 	int status;
939 	int type;
940 
941 	sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
942 	for (type = 0; type < MAXQUOTAS; type++) {
943 		if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
944 			continue;
945 		inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
946 							osb->slot_num);
947 		if (!inode[type]) {
948 			status = -ENOENT;
949 			goto out_quota_off;
950 		}
951 		status = dquot_enable(inode[type], type, QFMT_OCFS2,
952 				      DQUOT_USAGE_ENABLED);
953 		if (status < 0)
954 			goto out_quota_off;
955 	}
956 
957 	for (type = 0; type < MAXQUOTAS; type++)
958 		iput(inode[type]);
959 	return 0;
960 out_quota_off:
961 	ocfs2_disable_quotas(osb);
962 	for (type = 0; type < MAXQUOTAS; type++)
963 		iput(inode[type]);
964 	mlog_errno(status);
965 	return status;
966 }
967 
968 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
969 {
970 	int type;
971 	struct inode *inode;
972 	struct super_block *sb = osb->sb;
973 	struct ocfs2_mem_dqinfo *oinfo;
974 
975 	/* We mostly ignore errors in this function because there's not much
976 	 * we can do when we see them */
977 	for (type = 0; type < MAXQUOTAS; type++) {
978 		if (!sb_has_quota_loaded(sb, type))
979 			continue;
980 		/* Cancel periodic syncing before we grab dqonoff_mutex */
981 		oinfo = sb_dqinfo(sb, type)->dqi_priv;
982 		cancel_delayed_work_sync(&oinfo->dqi_sync_work);
983 		inode = igrab(sb->s_dquot.files[type]);
984 		/* Turn off quotas. This will remove all dquot structures from
985 		 * memory and so they will be automatically synced to global
986 		 * quota files */
987 		dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
988 					DQUOT_LIMITS_ENABLED);
989 		if (!inode)
990 			continue;
991 		iput(inode);
992 	}
993 }
994 
995 /* Handle quota on quotactl */
996 static int ocfs2_quota_on(struct super_block *sb, int type, int format_id)
997 {
998 	unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
999 					     OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
1000 
1001 	if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
1002 		return -EINVAL;
1003 
1004 	return dquot_enable(sb_dqopt(sb)->files[type], type,
1005 			    format_id, DQUOT_LIMITS_ENABLED);
1006 }
1007 
1008 /* Handle quota off quotactl */
1009 static int ocfs2_quota_off(struct super_block *sb, int type)
1010 {
1011 	return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
1012 }
1013 
1014 static const struct quotactl_ops ocfs2_quotactl_ops = {
1015 	.quota_on_meta	= ocfs2_quota_on,
1016 	.quota_off	= ocfs2_quota_off,
1017 	.quota_sync	= dquot_quota_sync,
1018 	.get_info	= dquot_get_dqinfo,
1019 	.set_info	= dquot_set_dqinfo,
1020 	.get_dqblk	= dquot_get_dqblk,
1021 	.set_dqblk	= dquot_set_dqblk,
1022 };
1023 
1024 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
1025 {
1026 	struct dentry *root;
1027 	int status, sector_size;
1028 	struct mount_options parsed_options;
1029 	struct inode *inode = NULL;
1030 	struct ocfs2_super *osb = NULL;
1031 	struct buffer_head *bh = NULL;
1032 	char nodestr[8];
1033 	struct ocfs2_blockcheck_stats stats;
1034 
1035 	mlog_entry("%p, %p, %i", sb, data, silent);
1036 
1037 	if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
1038 		status = -EINVAL;
1039 		goto read_super_error;
1040 	}
1041 
1042 	/* probe for superblock */
1043 	status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
1044 	if (status < 0) {
1045 		mlog(ML_ERROR, "superblock probe failed!\n");
1046 		goto read_super_error;
1047 	}
1048 
1049 	status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1050 	osb = OCFS2_SB(sb);
1051 	if (status < 0) {
1052 		mlog_errno(status);
1053 		goto read_super_error;
1054 	}
1055 	brelse(bh);
1056 	bh = NULL;
1057 
1058 	if (!ocfs2_check_set_options(sb, &parsed_options)) {
1059 		status = -EINVAL;
1060 		goto read_super_error;
1061 	}
1062 	osb->s_mount_opt = parsed_options.mount_opt;
1063 	osb->s_atime_quantum = parsed_options.atime_quantum;
1064 	osb->preferred_slot = parsed_options.slot;
1065 	osb->osb_commit_interval = parsed_options.commit_interval;
1066 
1067 	ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1068 	osb->osb_resv_level = parsed_options.resv_level;
1069 	osb->osb_dir_resv_level = parsed_options.resv_level;
1070 	if (parsed_options.dir_resv_level == -1)
1071 		osb->osb_dir_resv_level = parsed_options.resv_level;
1072 	else
1073 		osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1074 
1075 	status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1076 	if (status)
1077 		goto read_super_error;
1078 
1079 	sb->s_magic = OCFS2_SUPER_MAGIC;
1080 
1081 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1082 		((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1083 
1084 	/* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1085 	 * heartbeat=none */
1086 	if (bdev_read_only(sb->s_bdev)) {
1087 		if (!(sb->s_flags & MS_RDONLY)) {
1088 			status = -EACCES;
1089 			mlog(ML_ERROR, "Readonly device detected but readonly "
1090 			     "mount was not specified.\n");
1091 			goto read_super_error;
1092 		}
1093 
1094 		/* You should not be able to start a local heartbeat
1095 		 * on a readonly device. */
1096 		if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1097 			status = -EROFS;
1098 			mlog(ML_ERROR, "Local heartbeat specified on readonly "
1099 			     "device.\n");
1100 			goto read_super_error;
1101 		}
1102 
1103 		status = ocfs2_check_journals_nolocks(osb);
1104 		if (status < 0) {
1105 			if (status == -EROFS)
1106 				mlog(ML_ERROR, "Recovery required on readonly "
1107 				     "file system, but write access is "
1108 				     "unavailable.\n");
1109 			else
1110 				mlog_errno(status);
1111 			goto read_super_error;
1112 		}
1113 
1114 		ocfs2_set_ro_flag(osb, 1);
1115 
1116 		printk(KERN_NOTICE "Readonly device detected. No cluster "
1117 		       "services will be utilized for this mount. Recovery "
1118 		       "will be skipped.\n");
1119 	}
1120 
1121 	if (!ocfs2_is_hard_readonly(osb)) {
1122 		if (sb->s_flags & MS_RDONLY)
1123 			ocfs2_set_ro_flag(osb, 0);
1124 	}
1125 
1126 	status = ocfs2_verify_heartbeat(osb);
1127 	if (status < 0) {
1128 		mlog_errno(status);
1129 		goto read_super_error;
1130 	}
1131 
1132 	osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1133 						 ocfs2_debugfs_root);
1134 	if (!osb->osb_debug_root) {
1135 		status = -EINVAL;
1136 		mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1137 		goto read_super_error;
1138 	}
1139 
1140 	osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1141 					    osb->osb_debug_root,
1142 					    osb,
1143 					    &ocfs2_osb_debug_fops);
1144 	if (!osb->osb_ctxt) {
1145 		status = -EINVAL;
1146 		mlog_errno(status);
1147 		goto read_super_error;
1148 	}
1149 
1150 	if (ocfs2_meta_ecc(osb)) {
1151 		status = ocfs2_blockcheck_stats_debugfs_install(
1152 						&osb->osb_ecc_stats,
1153 						osb->osb_debug_root);
1154 		if (status) {
1155 			mlog(ML_ERROR,
1156 			     "Unable to create blockcheck statistics "
1157 			     "files\n");
1158 			goto read_super_error;
1159 		}
1160 	}
1161 
1162 	status = ocfs2_mount_volume(sb);
1163 	if (osb->root_inode)
1164 		inode = igrab(osb->root_inode);
1165 
1166 	if (status < 0)
1167 		goto read_super_error;
1168 
1169 	if (!inode) {
1170 		status = -EIO;
1171 		mlog_errno(status);
1172 		goto read_super_error;
1173 	}
1174 
1175 	root = d_alloc_root(inode);
1176 	if (!root) {
1177 		status = -ENOMEM;
1178 		mlog_errno(status);
1179 		goto read_super_error;
1180 	}
1181 
1182 	sb->s_root = root;
1183 
1184 	ocfs2_complete_mount_recovery(osb);
1185 
1186 	if (ocfs2_mount_local(osb))
1187 		snprintf(nodestr, sizeof(nodestr), "local");
1188 	else
1189 		snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1190 
1191 	printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1192 	       "with %s data mode.\n",
1193 	       osb->dev_str, nodestr, osb->slot_num,
1194 	       osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1195 	       "ordered");
1196 
1197 	atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1198 	wake_up(&osb->osb_mount_event);
1199 
1200 	/* Now we can initialize quotas because we can afford to wait
1201 	 * for cluster locks recovery now. That also means that truncation
1202 	 * log recovery can happen but that waits for proper quota setup */
1203 	if (!(sb->s_flags & MS_RDONLY)) {
1204 		status = ocfs2_enable_quotas(osb);
1205 		if (status < 0) {
1206 			/* We have to err-out specially here because
1207 			 * s_root is already set */
1208 			mlog_errno(status);
1209 			atomic_set(&osb->vol_state, VOLUME_DISABLED);
1210 			wake_up(&osb->osb_mount_event);
1211 			mlog_exit(status);
1212 			return status;
1213 		}
1214 	}
1215 
1216 	ocfs2_complete_quota_recovery(osb);
1217 
1218 	/* Now we wake up again for processes waiting for quotas */
1219 	atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1220 	wake_up(&osb->osb_mount_event);
1221 
1222 	/* Start this when the mount is almost sure of being successful */
1223 	ocfs2_orphan_scan_start(osb);
1224 
1225 	mlog_exit(status);
1226 	return status;
1227 
1228 read_super_error:
1229 	brelse(bh);
1230 
1231 	if (inode)
1232 		iput(inode);
1233 
1234 	if (osb) {
1235 		atomic_set(&osb->vol_state, VOLUME_DISABLED);
1236 		wake_up(&osb->osb_mount_event);
1237 		ocfs2_dismount_volume(sb, 1);
1238 	}
1239 
1240 	mlog_exit(status);
1241 	return status;
1242 }
1243 
1244 static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
1245 			int flags,
1246 			const char *dev_name,
1247 			void *data)
1248 {
1249 	return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
1250 }
1251 
1252 static void ocfs2_kill_sb(struct super_block *sb)
1253 {
1254 	struct ocfs2_super *osb = OCFS2_SB(sb);
1255 
1256 	/* Failed mount? */
1257 	if (!osb || atomic_read(&osb->vol_state) == VOLUME_DISABLED)
1258 		goto out;
1259 
1260 	/* Prevent further queueing of inode drop events */
1261 	spin_lock(&dentry_list_lock);
1262 	ocfs2_set_osb_flag(osb, OCFS2_OSB_DROP_DENTRY_LOCK_IMMED);
1263 	spin_unlock(&dentry_list_lock);
1264 	/* Wait for work to finish and/or remove it */
1265 	cancel_work_sync(&osb->dentry_lock_work);
1266 out:
1267 	kill_block_super(sb);
1268 }
1269 
1270 static struct file_system_type ocfs2_fs_type = {
1271 	.owner          = THIS_MODULE,
1272 	.name           = "ocfs2",
1273 	.mount          = ocfs2_mount,
1274 	.kill_sb        = ocfs2_kill_sb,
1275 
1276 	.fs_flags       = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1277 	.next           = NULL
1278 };
1279 
1280 static int ocfs2_check_set_options(struct super_block *sb,
1281 				   struct mount_options *options)
1282 {
1283 	if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1284 	    !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1285 					 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1286 		mlog(ML_ERROR, "User quotas were requested, but this "
1287 		     "filesystem does not have the feature enabled.\n");
1288 		return 0;
1289 	}
1290 	if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1291 	    !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1292 					 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1293 		mlog(ML_ERROR, "Group quotas were requested, but this "
1294 		     "filesystem does not have the feature enabled.\n");
1295 		return 0;
1296 	}
1297 	if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1298 	    !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1299 		mlog(ML_ERROR, "ACL support requested but extended attributes "
1300 		     "feature is not enabled\n");
1301 		return 0;
1302 	}
1303 	/* No ACL setting specified? Use XATTR feature... */
1304 	if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1305 				    OCFS2_MOUNT_NO_POSIX_ACL))) {
1306 		if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1307 			options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1308 		else
1309 			options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1310 	}
1311 	return 1;
1312 }
1313 
1314 static int ocfs2_parse_options(struct super_block *sb,
1315 			       char *options,
1316 			       struct mount_options *mopt,
1317 			       int is_remount)
1318 {
1319 	int status;
1320 	char *p;
1321 	u32 tmp;
1322 
1323 	mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
1324 		   options ? options : "(none)");
1325 
1326 	mopt->commit_interval = 0;
1327 	mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1328 	mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1329 	mopt->slot = OCFS2_INVALID_SLOT;
1330 	mopt->localalloc_opt = -1;
1331 	mopt->cluster_stack[0] = '\0';
1332 	mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1333 	mopt->dir_resv_level = -1;
1334 
1335 	if (!options) {
1336 		status = 1;
1337 		goto bail;
1338 	}
1339 
1340 	while ((p = strsep(&options, ",")) != NULL) {
1341 		int token, option;
1342 		substring_t args[MAX_OPT_ARGS];
1343 
1344 		if (!*p)
1345 			continue;
1346 
1347 		token = match_token(p, tokens, args);
1348 		switch (token) {
1349 		case Opt_hb_local:
1350 			mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1351 			break;
1352 		case Opt_hb_none:
1353 			mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1354 			break;
1355 		case Opt_hb_global:
1356 			mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1357 			break;
1358 		case Opt_barrier:
1359 			if (match_int(&args[0], &option)) {
1360 				status = 0;
1361 				goto bail;
1362 			}
1363 			if (option)
1364 				mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1365 			else
1366 				mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1367 			break;
1368 		case Opt_intr:
1369 			mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1370 			break;
1371 		case Opt_nointr:
1372 			mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1373 			break;
1374 		case Opt_err_panic:
1375 			mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1376 			break;
1377 		case Opt_err_ro:
1378 			mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1379 			break;
1380 		case Opt_data_ordered:
1381 			mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1382 			break;
1383 		case Opt_data_writeback:
1384 			mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1385 			break;
1386 		case Opt_user_xattr:
1387 			mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1388 			break;
1389 		case Opt_nouser_xattr:
1390 			mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1391 			break;
1392 		case Opt_atime_quantum:
1393 			if (match_int(&args[0], &option)) {
1394 				status = 0;
1395 				goto bail;
1396 			}
1397 			if (option >= 0)
1398 				mopt->atime_quantum = option;
1399 			break;
1400 		case Opt_slot:
1401 			option = 0;
1402 			if (match_int(&args[0], &option)) {
1403 				status = 0;
1404 				goto bail;
1405 			}
1406 			if (option)
1407 				mopt->slot = (s16)option;
1408 			break;
1409 		case Opt_commit:
1410 			option = 0;
1411 			if (match_int(&args[0], &option)) {
1412 				status = 0;
1413 				goto bail;
1414 			}
1415 			if (option < 0)
1416 				return 0;
1417 			if (option == 0)
1418 				option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1419 			mopt->commit_interval = HZ * option;
1420 			break;
1421 		case Opt_localalloc:
1422 			option = 0;
1423 			if (match_int(&args[0], &option)) {
1424 				status = 0;
1425 				goto bail;
1426 			}
1427 			if (option >= 0)
1428 				mopt->localalloc_opt = option;
1429 			break;
1430 		case Opt_localflocks:
1431 			/*
1432 			 * Changing this during remount could race
1433 			 * flock() requests, or "unbalance" existing
1434 			 * ones (e.g., a lock is taken in one mode but
1435 			 * dropped in the other). If users care enough
1436 			 * to flip locking modes during remount, we
1437 			 * could add a "local" flag to individual
1438 			 * flock structures for proper tracking of
1439 			 * state.
1440 			 */
1441 			if (!is_remount)
1442 				mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1443 			break;
1444 		case Opt_stack:
1445 			/* Check both that the option we were passed
1446 			 * is of the right length and that it is a proper
1447 			 * string of the right length.
1448 			 */
1449 			if (((args[0].to - args[0].from) !=
1450 			     OCFS2_STACK_LABEL_LEN) ||
1451 			    (strnlen(args[0].from,
1452 				     OCFS2_STACK_LABEL_LEN) !=
1453 			     OCFS2_STACK_LABEL_LEN)) {
1454 				mlog(ML_ERROR,
1455 				     "Invalid cluster_stack option\n");
1456 				status = 0;
1457 				goto bail;
1458 			}
1459 			memcpy(mopt->cluster_stack, args[0].from,
1460 			       OCFS2_STACK_LABEL_LEN);
1461 			mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1462 			break;
1463 		case Opt_inode64:
1464 			mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1465 			break;
1466 		case Opt_usrquota:
1467 			mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1468 			break;
1469 		case Opt_grpquota:
1470 			mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1471 			break;
1472 		case Opt_coherency_buffered:
1473 			mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1474 			break;
1475 		case Opt_coherency_full:
1476 			mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1477 			break;
1478 		case Opt_acl:
1479 			mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1480 			mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1481 			break;
1482 		case Opt_noacl:
1483 			mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1484 			mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1485 			break;
1486 		case Opt_resv_level:
1487 			if (is_remount)
1488 				break;
1489 			if (match_int(&args[0], &option)) {
1490 				status = 0;
1491 				goto bail;
1492 			}
1493 			if (option >= OCFS2_MIN_RESV_LEVEL &&
1494 			    option < OCFS2_MAX_RESV_LEVEL)
1495 				mopt->resv_level = option;
1496 			break;
1497 		case Opt_dir_resv_level:
1498 			if (is_remount)
1499 				break;
1500 			if (match_int(&args[0], &option)) {
1501 				status = 0;
1502 				goto bail;
1503 			}
1504 			if (option >= OCFS2_MIN_RESV_LEVEL &&
1505 			    option < OCFS2_MAX_RESV_LEVEL)
1506 				mopt->dir_resv_level = option;
1507 			break;
1508 		default:
1509 			mlog(ML_ERROR,
1510 			     "Unrecognized mount option \"%s\" "
1511 			     "or missing value\n", p);
1512 			status = 0;
1513 			goto bail;
1514 		}
1515 	}
1516 
1517 	/* Ensure only one heartbeat mode */
1518 	tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
1519 				 OCFS2_MOUNT_HB_NONE);
1520 	if (hweight32(tmp) != 1) {
1521 		mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1522 		status = 0;
1523 		goto bail;
1524 	}
1525 
1526 	status = 1;
1527 
1528 bail:
1529 	mlog_exit(status);
1530 	return status;
1531 }
1532 
1533 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1534 {
1535 	struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb);
1536 	unsigned long opts = osb->s_mount_opt;
1537 	unsigned int local_alloc_megs;
1538 
1539 	if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1540 		seq_printf(s, ",_netdev");
1541 		if (opts & OCFS2_MOUNT_HB_LOCAL)
1542 			seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1543 		else
1544 			seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1545 	} else
1546 		seq_printf(s, ",%s", OCFS2_HB_NONE);
1547 
1548 	if (opts & OCFS2_MOUNT_NOINTR)
1549 		seq_printf(s, ",nointr");
1550 
1551 	if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1552 		seq_printf(s, ",data=writeback");
1553 	else
1554 		seq_printf(s, ",data=ordered");
1555 
1556 	if (opts & OCFS2_MOUNT_BARRIER)
1557 		seq_printf(s, ",barrier=1");
1558 
1559 	if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1560 		seq_printf(s, ",errors=panic");
1561 	else
1562 		seq_printf(s, ",errors=remount-ro");
1563 
1564 	if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1565 		seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1566 
1567 	if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM)
1568 		seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1569 
1570 	if (osb->osb_commit_interval)
1571 		seq_printf(s, ",commit=%u",
1572 			   (unsigned) (osb->osb_commit_interval / HZ));
1573 
1574 	local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1575 	if (local_alloc_megs != ocfs2_la_default_mb(osb))
1576 		seq_printf(s, ",localalloc=%d", local_alloc_megs);
1577 
1578 	if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1579 		seq_printf(s, ",localflocks,");
1580 
1581 	if (osb->osb_cluster_stack[0])
1582 		seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1583 			   osb->osb_cluster_stack);
1584 	if (opts & OCFS2_MOUNT_USRQUOTA)
1585 		seq_printf(s, ",usrquota");
1586 	if (opts & OCFS2_MOUNT_GRPQUOTA)
1587 		seq_printf(s, ",grpquota");
1588 
1589 	if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1590 		seq_printf(s, ",coherency=buffered");
1591 	else
1592 		seq_printf(s, ",coherency=full");
1593 
1594 	if (opts & OCFS2_MOUNT_NOUSERXATTR)
1595 		seq_printf(s, ",nouser_xattr");
1596 	else
1597 		seq_printf(s, ",user_xattr");
1598 
1599 	if (opts & OCFS2_MOUNT_INODE64)
1600 		seq_printf(s, ",inode64");
1601 
1602 	if (opts & OCFS2_MOUNT_POSIX_ACL)
1603 		seq_printf(s, ",acl");
1604 	else
1605 		seq_printf(s, ",noacl");
1606 
1607 	if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1608 		seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1609 
1610 	if (osb->osb_dir_resv_level != osb->osb_resv_level)
1611 		seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1612 
1613 	return 0;
1614 }
1615 
1616 static int __init ocfs2_init(void)
1617 {
1618 	int status;
1619 
1620 	mlog_entry_void();
1621 
1622 	ocfs2_print_version();
1623 
1624 	status = init_ocfs2_uptodate_cache();
1625 	if (status < 0) {
1626 		mlog_errno(status);
1627 		goto leave;
1628 	}
1629 
1630 	status = ocfs2_initialize_mem_caches();
1631 	if (status < 0) {
1632 		mlog_errno(status);
1633 		goto leave;
1634 	}
1635 
1636 	ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1637 	if (!ocfs2_wq) {
1638 		status = -ENOMEM;
1639 		goto leave;
1640 	}
1641 
1642 	ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1643 	if (!ocfs2_debugfs_root) {
1644 		status = -EFAULT;
1645 		mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1646 	}
1647 
1648 	status = ocfs2_quota_setup();
1649 	if (status)
1650 		goto leave;
1651 
1652 	ocfs2_set_locking_protocol();
1653 
1654 	status = register_quota_format(&ocfs2_quota_format);
1655 leave:
1656 	if (status < 0) {
1657 		ocfs2_quota_shutdown();
1658 		ocfs2_free_mem_caches();
1659 		exit_ocfs2_uptodate_cache();
1660 	}
1661 
1662 	mlog_exit(status);
1663 
1664 	if (status >= 0) {
1665 		return register_filesystem(&ocfs2_fs_type);
1666 	} else
1667 		return -1;
1668 }
1669 
1670 static void __exit ocfs2_exit(void)
1671 {
1672 	mlog_entry_void();
1673 
1674 	ocfs2_quota_shutdown();
1675 
1676 	if (ocfs2_wq) {
1677 		flush_workqueue(ocfs2_wq);
1678 		destroy_workqueue(ocfs2_wq);
1679 	}
1680 
1681 	unregister_quota_format(&ocfs2_quota_format);
1682 
1683 	debugfs_remove(ocfs2_debugfs_root);
1684 
1685 	ocfs2_free_mem_caches();
1686 
1687 	unregister_filesystem(&ocfs2_fs_type);
1688 
1689 	exit_ocfs2_uptodate_cache();
1690 
1691 	mlog_exit_void();
1692 }
1693 
1694 static void ocfs2_put_super(struct super_block *sb)
1695 {
1696 	mlog_entry("(0x%p)\n", sb);
1697 
1698 	ocfs2_sync_blockdev(sb);
1699 	ocfs2_dismount_volume(sb, 0);
1700 
1701 	mlog_exit_void();
1702 }
1703 
1704 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1705 {
1706 	struct ocfs2_super *osb;
1707 	u32 numbits, freebits;
1708 	int status;
1709 	struct ocfs2_dinode *bm_lock;
1710 	struct buffer_head *bh = NULL;
1711 	struct inode *inode = NULL;
1712 
1713 	mlog_entry("(%p, %p)\n", dentry->d_sb, buf);
1714 
1715 	osb = OCFS2_SB(dentry->d_sb);
1716 
1717 	inode = ocfs2_get_system_file_inode(osb,
1718 					    GLOBAL_BITMAP_SYSTEM_INODE,
1719 					    OCFS2_INVALID_SLOT);
1720 	if (!inode) {
1721 		mlog(ML_ERROR, "failed to get bitmap inode\n");
1722 		status = -EIO;
1723 		goto bail;
1724 	}
1725 
1726 	status = ocfs2_inode_lock(inode, &bh, 0);
1727 	if (status < 0) {
1728 		mlog_errno(status);
1729 		goto bail;
1730 	}
1731 
1732 	bm_lock = (struct ocfs2_dinode *) bh->b_data;
1733 
1734 	numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1735 	freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1736 
1737 	buf->f_type = OCFS2_SUPER_MAGIC;
1738 	buf->f_bsize = dentry->d_sb->s_blocksize;
1739 	buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1740 	buf->f_blocks = ((sector_t) numbits) *
1741 			(osb->s_clustersize >> osb->sb->s_blocksize_bits);
1742 	buf->f_bfree = ((sector_t) freebits) *
1743 		       (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1744 	buf->f_bavail = buf->f_bfree;
1745 	buf->f_files = numbits;
1746 	buf->f_ffree = freebits;
1747 	buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1748 				& 0xFFFFFFFFUL;
1749 	buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1750 				OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1751 
1752 	brelse(bh);
1753 
1754 	ocfs2_inode_unlock(inode, 0);
1755 	status = 0;
1756 bail:
1757 	if (inode)
1758 		iput(inode);
1759 
1760 	mlog_exit(status);
1761 
1762 	return status;
1763 }
1764 
1765 static void ocfs2_inode_init_once(void *data)
1766 {
1767 	struct ocfs2_inode_info *oi = data;
1768 
1769 	oi->ip_flags = 0;
1770 	oi->ip_open_count = 0;
1771 	spin_lock_init(&oi->ip_lock);
1772 	ocfs2_extent_map_init(&oi->vfs_inode);
1773 	INIT_LIST_HEAD(&oi->ip_io_markers);
1774 	oi->ip_dir_start_lookup = 0;
1775 
1776 	init_rwsem(&oi->ip_alloc_sem);
1777 	init_rwsem(&oi->ip_xattr_sem);
1778 	mutex_init(&oi->ip_io_mutex);
1779 
1780 	oi->ip_blkno = 0ULL;
1781 	oi->ip_clusters = 0;
1782 
1783 	ocfs2_resv_init_once(&oi->ip_la_data_resv);
1784 
1785 	ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1786 	ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1787 	ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1788 
1789 	ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1790 				  &ocfs2_inode_caching_ops);
1791 
1792 	inode_init_once(&oi->vfs_inode);
1793 }
1794 
1795 static int ocfs2_initialize_mem_caches(void)
1796 {
1797 	ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1798 				       sizeof(struct ocfs2_inode_info),
1799 				       0,
1800 				       (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1801 						SLAB_MEM_SPREAD),
1802 				       ocfs2_inode_init_once);
1803 	ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1804 					sizeof(struct ocfs2_dquot),
1805 					0,
1806 					(SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1807 						SLAB_MEM_SPREAD),
1808 					NULL);
1809 	ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1810 					sizeof(struct ocfs2_quota_chunk),
1811 					0,
1812 					(SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1813 					NULL);
1814 	if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1815 	    !ocfs2_qf_chunk_cachep) {
1816 		if (ocfs2_inode_cachep)
1817 			kmem_cache_destroy(ocfs2_inode_cachep);
1818 		if (ocfs2_dquot_cachep)
1819 			kmem_cache_destroy(ocfs2_dquot_cachep);
1820 		if (ocfs2_qf_chunk_cachep)
1821 			kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1822 		return -ENOMEM;
1823 	}
1824 
1825 	return 0;
1826 }
1827 
1828 static void ocfs2_free_mem_caches(void)
1829 {
1830 	if (ocfs2_inode_cachep)
1831 		kmem_cache_destroy(ocfs2_inode_cachep);
1832 	ocfs2_inode_cachep = NULL;
1833 
1834 	if (ocfs2_dquot_cachep)
1835 		kmem_cache_destroy(ocfs2_dquot_cachep);
1836 	ocfs2_dquot_cachep = NULL;
1837 
1838 	if (ocfs2_qf_chunk_cachep)
1839 		kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1840 	ocfs2_qf_chunk_cachep = NULL;
1841 }
1842 
1843 static int ocfs2_get_sector(struct super_block *sb,
1844 			    struct buffer_head **bh,
1845 			    int block,
1846 			    int sect_size)
1847 {
1848 	if (!sb_set_blocksize(sb, sect_size)) {
1849 		mlog(ML_ERROR, "unable to set blocksize\n");
1850 		return -EIO;
1851 	}
1852 
1853 	*bh = sb_getblk(sb, block);
1854 	if (!*bh) {
1855 		mlog_errno(-EIO);
1856 		return -EIO;
1857 	}
1858 	lock_buffer(*bh);
1859 	if (!buffer_dirty(*bh))
1860 		clear_buffer_uptodate(*bh);
1861 	unlock_buffer(*bh);
1862 	ll_rw_block(READ, 1, bh);
1863 	wait_on_buffer(*bh);
1864 	if (!buffer_uptodate(*bh)) {
1865 		mlog_errno(-EIO);
1866 		brelse(*bh);
1867 		*bh = NULL;
1868 		return -EIO;
1869 	}
1870 
1871 	return 0;
1872 }
1873 
1874 static int ocfs2_mount_volume(struct super_block *sb)
1875 {
1876 	int status = 0;
1877 	int unlock_super = 0;
1878 	struct ocfs2_super *osb = OCFS2_SB(sb);
1879 
1880 	mlog_entry_void();
1881 
1882 	if (ocfs2_is_hard_readonly(osb))
1883 		goto leave;
1884 
1885 	status = ocfs2_dlm_init(osb);
1886 	if (status < 0) {
1887 		mlog_errno(status);
1888 		goto leave;
1889 	}
1890 
1891 	status = ocfs2_super_lock(osb, 1);
1892 	if (status < 0) {
1893 		mlog_errno(status);
1894 		goto leave;
1895 	}
1896 	unlock_super = 1;
1897 
1898 	/* This will load up the node map and add ourselves to it. */
1899 	status = ocfs2_find_slot(osb);
1900 	if (status < 0) {
1901 		mlog_errno(status);
1902 		goto leave;
1903 	}
1904 
1905 	/* load all node-local system inodes */
1906 	status = ocfs2_init_local_system_inodes(osb);
1907 	if (status < 0) {
1908 		mlog_errno(status);
1909 		goto leave;
1910 	}
1911 
1912 	status = ocfs2_check_volume(osb);
1913 	if (status < 0) {
1914 		mlog_errno(status);
1915 		goto leave;
1916 	}
1917 
1918 	status = ocfs2_truncate_log_init(osb);
1919 	if (status < 0)
1920 		mlog_errno(status);
1921 
1922 leave:
1923 	if (unlock_super)
1924 		ocfs2_super_unlock(osb, 1);
1925 
1926 	mlog_exit(status);
1927 	return status;
1928 }
1929 
1930 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1931 {
1932 	int tmp, hangup_needed = 0;
1933 	struct ocfs2_super *osb = NULL;
1934 	char nodestr[8];
1935 
1936 	mlog_entry("(0x%p)\n", sb);
1937 
1938 	BUG_ON(!sb);
1939 	osb = OCFS2_SB(sb);
1940 	BUG_ON(!osb);
1941 
1942 	debugfs_remove(osb->osb_ctxt);
1943 
1944 	/*
1945 	 * Flush inode dropping work queue so that deletes are
1946 	 * performed while the filesystem is still working
1947 	 */
1948 	ocfs2_drop_all_dl_inodes(osb);
1949 
1950 	/* Orphan scan should be stopped as early as possible */
1951 	ocfs2_orphan_scan_stop(osb);
1952 
1953 	ocfs2_disable_quotas(osb);
1954 
1955 	ocfs2_shutdown_local_alloc(osb);
1956 
1957 	ocfs2_truncate_log_shutdown(osb);
1958 
1959 	/* This will disable recovery and flush any recovery work. */
1960 	ocfs2_recovery_exit(osb);
1961 
1962 	ocfs2_journal_shutdown(osb);
1963 
1964 	ocfs2_sync_blockdev(sb);
1965 
1966 	ocfs2_purge_refcount_trees(osb);
1967 
1968 	/* No cluster connection means we've failed during mount, so skip
1969 	 * all the steps which depended on that to complete. */
1970 	if (osb->cconn) {
1971 		tmp = ocfs2_super_lock(osb, 1);
1972 		if (tmp < 0) {
1973 			mlog_errno(tmp);
1974 			return;
1975 		}
1976 	}
1977 
1978 	if (osb->slot_num != OCFS2_INVALID_SLOT)
1979 		ocfs2_put_slot(osb);
1980 
1981 	if (osb->cconn)
1982 		ocfs2_super_unlock(osb, 1);
1983 
1984 	ocfs2_release_system_inodes(osb);
1985 
1986 	/*
1987 	 * If we're dismounting due to mount error, mount.ocfs2 will clean
1988 	 * up heartbeat.  If we're a local mount, there is no heartbeat.
1989 	 * If we failed before we got a uuid_str yet, we can't stop
1990 	 * heartbeat.  Otherwise, do it.
1991 	 */
1992 	if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str)
1993 		hangup_needed = 1;
1994 
1995 	if (osb->cconn)
1996 		ocfs2_dlm_shutdown(osb, hangup_needed);
1997 
1998 	ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1999 	debugfs_remove(osb->osb_debug_root);
2000 
2001 	if (hangup_needed)
2002 		ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
2003 
2004 	atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
2005 
2006 	if (ocfs2_mount_local(osb))
2007 		snprintf(nodestr, sizeof(nodestr), "local");
2008 	else
2009 		snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
2010 
2011 	printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
2012 	       osb->dev_str, nodestr);
2013 
2014 	ocfs2_delete_osb(osb);
2015 	kfree(osb);
2016 	sb->s_dev = 0;
2017 	sb->s_fs_info = NULL;
2018 }
2019 
2020 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
2021 				unsigned uuid_bytes)
2022 {
2023 	int i, ret;
2024 	char *ptr;
2025 
2026 	BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
2027 
2028 	osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
2029 	if (osb->uuid_str == NULL)
2030 		return -ENOMEM;
2031 
2032 	for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
2033 		/* print with null */
2034 		ret = snprintf(ptr, 3, "%02X", uuid[i]);
2035 		if (ret != 2) /* drop super cleans up */
2036 			return -EINVAL;
2037 		/* then only advance past the last char */
2038 		ptr += 2;
2039 	}
2040 
2041 	return 0;
2042 }
2043 
2044 /* Make sure entire volume is addressable by our journal.  Requires
2045    osb_clusters_at_boot to be valid and for the journal to have been
2046    initialized by ocfs2_journal_init(). */
2047 static int ocfs2_journal_addressable(struct ocfs2_super *osb)
2048 {
2049 	int status = 0;
2050 	u64 max_block =
2051 		ocfs2_clusters_to_blocks(osb->sb,
2052 					 osb->osb_clusters_at_boot) - 1;
2053 
2054 	/* 32-bit block number is always OK. */
2055 	if (max_block <= (u32)~0ULL)
2056 		goto out;
2057 
2058 	/* Volume is "huge", so see if our journal is new enough to
2059 	   support it. */
2060 	if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2061 				       OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2062 	      jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2063 					       JBD2_FEATURE_INCOMPAT_64BIT))) {
2064 		mlog(ML_ERROR, "The journal cannot address the entire volume. "
2065 		     "Enable the 'block64' journal option with tunefs.ocfs2");
2066 		status = -EFBIG;
2067 		goto out;
2068 	}
2069 
2070  out:
2071 	return status;
2072 }
2073 
2074 static int ocfs2_initialize_super(struct super_block *sb,
2075 				  struct buffer_head *bh,
2076 				  int sector_size,
2077 				  struct ocfs2_blockcheck_stats *stats)
2078 {
2079 	int status;
2080 	int i, cbits, bbits;
2081 	struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2082 	struct inode *inode = NULL;
2083 	struct ocfs2_journal *journal;
2084 	__le32 uuid_net_key;
2085 	struct ocfs2_super *osb;
2086 	u64 total_blocks;
2087 
2088 	mlog_entry_void();
2089 
2090 	osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2091 	if (!osb) {
2092 		status = -ENOMEM;
2093 		mlog_errno(status);
2094 		goto bail;
2095 	}
2096 
2097 	sb->s_fs_info = osb;
2098 	sb->s_op = &ocfs2_sops;
2099 	sb->s_d_op = &ocfs2_dentry_ops;
2100 	sb->s_export_op = &ocfs2_export_ops;
2101 	sb->s_qcop = &ocfs2_quotactl_ops;
2102 	sb->dq_op = &ocfs2_quota_operations;
2103 	sb->s_xattr = ocfs2_xattr_handlers;
2104 	sb->s_time_gran = 1;
2105 	sb->s_flags |= MS_NOATIME;
2106 	/* this is needed to support O_LARGEFILE */
2107 	cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2108 	bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2109 	sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2110 
2111 	osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2112 
2113 	for (i = 0; i < 3; i++)
2114 		osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2115 	osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2116 
2117 	osb->sb = sb;
2118 	/* Save off for ocfs2_rw_direct */
2119 	osb->s_sectsize_bits = blksize_bits(sector_size);
2120 	BUG_ON(!osb->s_sectsize_bits);
2121 
2122 	spin_lock_init(&osb->dc_task_lock);
2123 	init_waitqueue_head(&osb->dc_event);
2124 	osb->dc_work_sequence = 0;
2125 	osb->dc_wake_sequence = 0;
2126 	INIT_LIST_HEAD(&osb->blocked_lock_list);
2127 	osb->blocked_lock_count = 0;
2128 	spin_lock_init(&osb->osb_lock);
2129 	spin_lock_init(&osb->osb_xattr_lock);
2130 	ocfs2_init_steal_slots(osb);
2131 
2132 	atomic_set(&osb->alloc_stats.moves, 0);
2133 	atomic_set(&osb->alloc_stats.local_data, 0);
2134 	atomic_set(&osb->alloc_stats.bitmap_data, 0);
2135 	atomic_set(&osb->alloc_stats.bg_allocs, 0);
2136 	atomic_set(&osb->alloc_stats.bg_extends, 0);
2137 
2138 	/* Copy the blockcheck stats from the superblock probe */
2139 	osb->osb_ecc_stats = *stats;
2140 
2141 	ocfs2_init_node_maps(osb);
2142 
2143 	snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2144 		 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2145 
2146 	osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2147 	if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2148 		mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2149 		     osb->max_slots);
2150 		status = -EINVAL;
2151 		goto bail;
2152 	}
2153 	mlog(0, "max_slots for this device: %u\n", osb->max_slots);
2154 
2155 	ocfs2_orphan_scan_init(osb);
2156 
2157 	status = ocfs2_recovery_init(osb);
2158 	if (status) {
2159 		mlog(ML_ERROR, "Unable to initialize recovery state\n");
2160 		mlog_errno(status);
2161 		goto bail;
2162 	}
2163 
2164 	init_waitqueue_head(&osb->checkpoint_event);
2165 	atomic_set(&osb->needs_checkpoint, 0);
2166 
2167 	osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2168 
2169 	osb->slot_num = OCFS2_INVALID_SLOT;
2170 
2171 	osb->s_xattr_inline_size = le16_to_cpu(
2172 					di->id2.i_super.s_xattr_inline_size);
2173 
2174 	osb->local_alloc_state = OCFS2_LA_UNUSED;
2175 	osb->local_alloc_bh = NULL;
2176 	INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2177 
2178 	init_waitqueue_head(&osb->osb_mount_event);
2179 
2180 	status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2181 	if (status) {
2182 		mlog_errno(status);
2183 		goto bail;
2184 	}
2185 
2186 	osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2187 	if (!osb->vol_label) {
2188 		mlog(ML_ERROR, "unable to alloc vol label\n");
2189 		status = -ENOMEM;
2190 		goto bail;
2191 	}
2192 
2193 	osb->slot_recovery_generations =
2194 		kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2195 			GFP_KERNEL);
2196 	if (!osb->slot_recovery_generations) {
2197 		status = -ENOMEM;
2198 		mlog_errno(status);
2199 		goto bail;
2200 	}
2201 
2202 	init_waitqueue_head(&osb->osb_wipe_event);
2203 	osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2204 					sizeof(*osb->osb_orphan_wipes),
2205 					GFP_KERNEL);
2206 	if (!osb->osb_orphan_wipes) {
2207 		status = -ENOMEM;
2208 		mlog_errno(status);
2209 		goto bail;
2210 	}
2211 
2212 	osb->osb_rf_lock_tree = RB_ROOT;
2213 
2214 	osb->s_feature_compat =
2215 		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2216 	osb->s_feature_ro_compat =
2217 		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2218 	osb->s_feature_incompat =
2219 		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2220 
2221 	if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2222 		mlog(ML_ERROR, "couldn't mount because of unsupported "
2223 		     "optional features (%x).\n", i);
2224 		status = -EINVAL;
2225 		goto bail;
2226 	}
2227 	if (!(osb->sb->s_flags & MS_RDONLY) &&
2228 	    (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2229 		mlog(ML_ERROR, "couldn't mount RDWR because of "
2230 		     "unsupported optional features (%x).\n", i);
2231 		status = -EINVAL;
2232 		goto bail;
2233 	}
2234 
2235 	if (ocfs2_clusterinfo_valid(osb)) {
2236 		osb->osb_stackflags =
2237 			OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2238 		memcpy(osb->osb_cluster_stack,
2239 		       OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2240 		       OCFS2_STACK_LABEL_LEN);
2241 		osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
2242 		if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2243 			mlog(ML_ERROR,
2244 			     "couldn't mount because of an invalid "
2245 			     "cluster stack label (%s) \n",
2246 			     osb->osb_cluster_stack);
2247 			status = -EINVAL;
2248 			goto bail;
2249 		}
2250 	} else {
2251 		/* The empty string is identical with classic tools that
2252 		 * don't know about s_cluster_info. */
2253 		osb->osb_cluster_stack[0] = '\0';
2254 	}
2255 
2256 	get_random_bytes(&osb->s_next_generation, sizeof(u32));
2257 
2258 	/* FIXME
2259 	 * This should be done in ocfs2_journal_init(), but unknown
2260 	 * ordering issues will cause the filesystem to crash.
2261 	 * If anyone wants to figure out what part of the code
2262 	 * refers to osb->journal before ocfs2_journal_init() is run,
2263 	 * be my guest.
2264 	 */
2265 	/* initialize our journal structure */
2266 
2267 	journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2268 	if (!journal) {
2269 		mlog(ML_ERROR, "unable to alloc journal\n");
2270 		status = -ENOMEM;
2271 		goto bail;
2272 	}
2273 	osb->journal = journal;
2274 	journal->j_osb = osb;
2275 
2276 	atomic_set(&journal->j_num_trans, 0);
2277 	init_rwsem(&journal->j_trans_barrier);
2278 	init_waitqueue_head(&journal->j_checkpointed);
2279 	spin_lock_init(&journal->j_lock);
2280 	journal->j_trans_id = (unsigned long) 1;
2281 	INIT_LIST_HEAD(&journal->j_la_cleanups);
2282 	INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2283 	journal->j_state = OCFS2_JOURNAL_FREE;
2284 
2285 	INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes);
2286 	osb->dentry_lock_list = NULL;
2287 
2288 	/* get some pseudo constants for clustersize bits */
2289 	osb->s_clustersize_bits =
2290 		le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2291 	osb->s_clustersize = 1 << osb->s_clustersize_bits;
2292 	mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
2293 
2294 	if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2295 	    osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2296 		mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2297 		     osb->s_clustersize);
2298 		status = -EINVAL;
2299 		goto bail;
2300 	}
2301 
2302 	total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2303 						le32_to_cpu(di->i_clusters));
2304 
2305 	status = generic_check_addressable(osb->sb->s_blocksize_bits,
2306 					   total_blocks);
2307 	if (status) {
2308 		mlog(ML_ERROR, "Volume too large "
2309 		     "to mount safely on this system");
2310 		status = -EFBIG;
2311 		goto bail;
2312 	}
2313 
2314 	if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2315 				 sizeof(di->id2.i_super.s_uuid))) {
2316 		mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2317 		status = -ENOMEM;
2318 		goto bail;
2319 	}
2320 
2321 	memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
2322 
2323 	strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2324 	osb->vol_label[63] = '\0';
2325 	osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2326 	osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2327 	osb->first_cluster_group_blkno =
2328 		le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2329 	osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2330 	osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2331 	mlog(0, "vol_label: %s\n", osb->vol_label);
2332 	mlog(0, "uuid: %s\n", osb->uuid_str);
2333 	mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
2334 	     (unsigned long long)osb->root_blkno,
2335 	     (unsigned long long)osb->system_dir_blkno);
2336 
2337 	osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2338 	if (!osb->osb_dlm_debug) {
2339 		status = -ENOMEM;
2340 		mlog_errno(status);
2341 		goto bail;
2342 	}
2343 
2344 	atomic_set(&osb->vol_state, VOLUME_INIT);
2345 
2346 	/* load root, system_dir, and all global system inodes */
2347 	status = ocfs2_init_global_system_inodes(osb);
2348 	if (status < 0) {
2349 		mlog_errno(status);
2350 		goto bail;
2351 	}
2352 
2353 	/*
2354 	 * global bitmap
2355 	 */
2356 	inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2357 					    OCFS2_INVALID_SLOT);
2358 	if (!inode) {
2359 		status = -EINVAL;
2360 		mlog_errno(status);
2361 		goto bail;
2362 	}
2363 
2364 	osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2365 	osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2366 	iput(inode);
2367 
2368 	osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2369 				 osb->s_feature_incompat) * 8;
2370 
2371 	status = ocfs2_init_slot_info(osb);
2372 	if (status < 0) {
2373 		mlog_errno(status);
2374 		goto bail;
2375 	}
2376 
2377 bail:
2378 	mlog_exit(status);
2379 	return status;
2380 }
2381 
2382 /*
2383  * will return: -EAGAIN if it is ok to keep searching for superblocks
2384  *              -EINVAL if there is a bad superblock
2385  *              0 on success
2386  */
2387 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2388 			       struct buffer_head *bh,
2389 			       u32 blksz,
2390 			       struct ocfs2_blockcheck_stats *stats)
2391 {
2392 	int status = -EAGAIN;
2393 
2394 	mlog_entry_void();
2395 
2396 	if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2397 		   strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2398 		/* We have to do a raw check of the feature here */
2399 		if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2400 		    OCFS2_FEATURE_INCOMPAT_META_ECC) {
2401 			status = ocfs2_block_check_validate(bh->b_data,
2402 							    bh->b_size,
2403 							    &di->i_check,
2404 							    stats);
2405 			if (status)
2406 				goto out;
2407 		}
2408 		status = -EINVAL;
2409 		if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2410 			mlog(ML_ERROR, "found superblock with incorrect block "
2411 			     "size: found %u, should be %u\n",
2412 			     1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2413 			       blksz);
2414 		} else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2415 			   OCFS2_MAJOR_REV_LEVEL ||
2416 			   le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2417 			   OCFS2_MINOR_REV_LEVEL) {
2418 			mlog(ML_ERROR, "found superblock with bad version: "
2419 			     "found %u.%u, should be %u.%u\n",
2420 			     le16_to_cpu(di->id2.i_super.s_major_rev_level),
2421 			     le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2422 			     OCFS2_MAJOR_REV_LEVEL,
2423 			     OCFS2_MINOR_REV_LEVEL);
2424 		} else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2425 			mlog(ML_ERROR, "bad block number on superblock: "
2426 			     "found %llu, should be %llu\n",
2427 			     (unsigned long long)le64_to_cpu(di->i_blkno),
2428 			     (unsigned long long)bh->b_blocknr);
2429 		} else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2430 			    le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2431 			mlog(ML_ERROR, "bad cluster size found: %u\n",
2432 			     1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2433 		} else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2434 			mlog(ML_ERROR, "bad root_blkno: 0\n");
2435 		} else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2436 			mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2437 		} else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2438 			mlog(ML_ERROR,
2439 			     "Superblock slots found greater than file system "
2440 			     "maximum: found %u, max %u\n",
2441 			     le16_to_cpu(di->id2.i_super.s_max_slots),
2442 			     OCFS2_MAX_SLOTS);
2443 		} else {
2444 			/* found it! */
2445 			status = 0;
2446 		}
2447 	}
2448 
2449 out:
2450 	mlog_exit(status);
2451 	return status;
2452 }
2453 
2454 static int ocfs2_check_volume(struct ocfs2_super *osb)
2455 {
2456 	int status;
2457 	int dirty;
2458 	int local;
2459 	struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2460 						  * recover
2461 						  * ourselves. */
2462 
2463 	mlog_entry_void();
2464 
2465 	/* Init our journal object. */
2466 	status = ocfs2_journal_init(osb->journal, &dirty);
2467 	if (status < 0) {
2468 		mlog(ML_ERROR, "Could not initialize journal!\n");
2469 		goto finally;
2470 	}
2471 
2472 	/* Now that journal has been initialized, check to make sure
2473 	   entire volume is addressable. */
2474 	status = ocfs2_journal_addressable(osb);
2475 	if (status)
2476 		goto finally;
2477 
2478 	/* If the journal was unmounted cleanly then we don't want to
2479 	 * recover anything. Otherwise, journal_load will do that
2480 	 * dirty work for us :) */
2481 	if (!dirty) {
2482 		status = ocfs2_journal_wipe(osb->journal, 0);
2483 		if (status < 0) {
2484 			mlog_errno(status);
2485 			goto finally;
2486 		}
2487 	} else {
2488 		mlog(ML_NOTICE, "File system was not unmounted cleanly, "
2489 		     "recovering volume.\n");
2490 	}
2491 
2492 	local = ocfs2_mount_local(osb);
2493 
2494 	/* will play back anything left in the journal. */
2495 	status = ocfs2_journal_load(osb->journal, local, dirty);
2496 	if (status < 0) {
2497 		mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2498 		goto finally;
2499 	}
2500 
2501 	if (dirty) {
2502 		/* recover my local alloc if we didn't unmount cleanly. */
2503 		status = ocfs2_begin_local_alloc_recovery(osb,
2504 							  osb->slot_num,
2505 							  &local_alloc);
2506 		if (status < 0) {
2507 			mlog_errno(status);
2508 			goto finally;
2509 		}
2510 		/* we complete the recovery process after we've marked
2511 		 * ourselves as mounted. */
2512 	}
2513 
2514 	mlog(0, "Journal loaded.\n");
2515 
2516 	status = ocfs2_load_local_alloc(osb);
2517 	if (status < 0) {
2518 		mlog_errno(status);
2519 		goto finally;
2520 	}
2521 
2522 	if (dirty) {
2523 		/* Recovery will be completed after we've mounted the
2524 		 * rest of the volume. */
2525 		osb->dirty = 1;
2526 		osb->local_alloc_copy = local_alloc;
2527 		local_alloc = NULL;
2528 	}
2529 
2530 	/* go through each journal, trylock it and if you get the
2531 	 * lock, and it's marked as dirty, set the bit in the recover
2532 	 * map and launch a recovery thread for it. */
2533 	status = ocfs2_mark_dead_nodes(osb);
2534 	if (status < 0) {
2535 		mlog_errno(status);
2536 		goto finally;
2537 	}
2538 
2539 	status = ocfs2_compute_replay_slots(osb);
2540 	if (status < 0)
2541 		mlog_errno(status);
2542 
2543 finally:
2544 	if (local_alloc)
2545 		kfree(local_alloc);
2546 
2547 	mlog_exit(status);
2548 	return status;
2549 }
2550 
2551 /*
2552  * The routine gets called from dismount or close whenever a dismount on
2553  * volume is requested and the osb open count becomes 1.
2554  * It will remove the osb from the global list and also free up all the
2555  * initialized resources and fileobject.
2556  */
2557 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2558 {
2559 	mlog_entry_void();
2560 
2561 	/* This function assumes that the caller has the main osb resource */
2562 
2563 	ocfs2_free_slot_info(osb);
2564 
2565 	kfree(osb->osb_orphan_wipes);
2566 	kfree(osb->slot_recovery_generations);
2567 	/* FIXME
2568 	 * This belongs in journal shutdown, but because we have to
2569 	 * allocate osb->journal at the start of ocfs2_initialize_osb(),
2570 	 * we free it here.
2571 	 */
2572 	kfree(osb->journal);
2573 	if (osb->local_alloc_copy)
2574 		kfree(osb->local_alloc_copy);
2575 	kfree(osb->uuid_str);
2576 	ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2577 	memset(osb, 0, sizeof(struct ocfs2_super));
2578 
2579 	mlog_exit_void();
2580 }
2581 
2582 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2583  * panic(). We do not support continue-on-error operation. */
2584 static void ocfs2_handle_error(struct super_block *sb)
2585 {
2586 	struct ocfs2_super *osb = OCFS2_SB(sb);
2587 
2588 	if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2589 		panic("OCFS2: (device %s): panic forced after error\n",
2590 		      sb->s_id);
2591 
2592 	ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2593 
2594 	if (sb->s_flags & MS_RDONLY &&
2595 	    (ocfs2_is_soft_readonly(osb) ||
2596 	     ocfs2_is_hard_readonly(osb)))
2597 		return;
2598 
2599 	printk(KERN_CRIT "File system is now read-only due to the potential "
2600 	       "of on-disk corruption. Please run fsck.ocfs2 once the file "
2601 	       "system is unmounted.\n");
2602 	sb->s_flags |= MS_RDONLY;
2603 	ocfs2_set_ro_flag(osb, 0);
2604 }
2605 
2606 static char error_buf[1024];
2607 
2608 void __ocfs2_error(struct super_block *sb,
2609 		   const char *function,
2610 		   const char *fmt, ...)
2611 {
2612 	va_list args;
2613 
2614 	va_start(args, fmt);
2615 	vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2616 	va_end(args);
2617 
2618 	/* Not using mlog here because we want to show the actual
2619 	 * function the error came from. */
2620 	printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2621 	       sb->s_id, function, error_buf);
2622 
2623 	ocfs2_handle_error(sb);
2624 }
2625 
2626 /* Handle critical errors. This is intentionally more drastic than
2627  * ocfs2_handle_error, so we only use for things like journal errors,
2628  * etc. */
2629 void __ocfs2_abort(struct super_block* sb,
2630 		   const char *function,
2631 		   const char *fmt, ...)
2632 {
2633 	va_list args;
2634 
2635 	va_start(args, fmt);
2636 	vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2637 	va_end(args);
2638 
2639 	printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2640 	       sb->s_id, function, error_buf);
2641 
2642 	/* We don't have the cluster support yet to go straight to
2643 	 * hard readonly in here. Until then, we want to keep
2644 	 * ocfs2_abort() so that we can at least mark critical
2645 	 * errors.
2646 	 *
2647 	 * TODO: This should abort the journal and alert other nodes
2648 	 * that our slot needs recovery. */
2649 
2650 	/* Force a panic(). This stinks, but it's better than letting
2651 	 * things continue without having a proper hard readonly
2652 	 * here. */
2653 	if (!ocfs2_mount_local(OCFS2_SB(sb)))
2654 		OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2655 	ocfs2_handle_error(sb);
2656 }
2657 
2658 /*
2659  * Void signal blockers, because in-kernel sigprocmask() only fails
2660  * when SIG_* is wrong.
2661  */
2662 void ocfs2_block_signals(sigset_t *oldset)
2663 {
2664 	int rc;
2665 	sigset_t blocked;
2666 
2667 	sigfillset(&blocked);
2668 	rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2669 	BUG_ON(rc);
2670 }
2671 
2672 void ocfs2_unblock_signals(sigset_t *oldset)
2673 {
2674 	int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2675 	BUG_ON(rc);
2676 }
2677 
2678 module_init(ocfs2_init);
2679 module_exit(ocfs2_exit);
2680