xref: /linux/fs/ocfs2/super.c (revision ff10fca5ceacf7bc59636f5ab808e775d1717167)
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 #include <linux/smp_lock.h>
45 
46 #define MLOG_MASK_PREFIX ML_SUPER
47 #include <cluster/masklog.h>
48 
49 #include "ocfs2.h"
50 
51 /* this should be the only file to include a version 1 header */
52 #include "ocfs1_fs_compat.h"
53 
54 #include "alloc.h"
55 #include "blockcheck.h"
56 #include "dlmglue.h"
57 #include "export.h"
58 #include "extent_map.h"
59 #include "heartbeat.h"
60 #include "inode.h"
61 #include "journal.h"
62 #include "localalloc.h"
63 #include "namei.h"
64 #include "slot_map.h"
65 #include "super.h"
66 #include "sysfile.h"
67 #include "uptodate.h"
68 #include "ver.h"
69 #include "xattr.h"
70 #include "quota.h"
71 #include "refcounttree.h"
72 #include "suballoc.h"
73 
74 #include "buffer_head_io.h"
75 
76 static struct kmem_cache *ocfs2_inode_cachep = NULL;
77 struct kmem_cache *ocfs2_dquot_cachep;
78 struct kmem_cache *ocfs2_qf_chunk_cachep;
79 
80 /* OCFS2 needs to schedule several differnt types of work which
81  * require cluster locking, disk I/O, recovery waits, etc. Since these
82  * types of work tend to be heavy we avoid using the kernel events
83  * workqueue and schedule on our own. */
84 struct workqueue_struct *ocfs2_wq = NULL;
85 
86 static struct dentry *ocfs2_debugfs_root = NULL;
87 
88 MODULE_AUTHOR("Oracle");
89 MODULE_LICENSE("GPL");
90 
91 struct mount_options
92 {
93 	unsigned long	commit_interval;
94 	unsigned long	mount_opt;
95 	unsigned int	atime_quantum;
96 	signed short	slot;
97 	int		localalloc_opt;
98 	unsigned int	resv_level;
99 	int		dir_resv_level;
100 	char		cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
101 };
102 
103 static int ocfs2_parse_options(struct super_block *sb, char *options,
104 			       struct mount_options *mopt,
105 			       int is_remount);
106 static int ocfs2_check_set_options(struct super_block *sb,
107 				   struct mount_options *options);
108 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt);
109 static void ocfs2_put_super(struct super_block *sb);
110 static int ocfs2_mount_volume(struct super_block *sb);
111 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
112 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
113 static int ocfs2_initialize_mem_caches(void);
114 static void ocfs2_free_mem_caches(void);
115 static void ocfs2_delete_osb(struct ocfs2_super *osb);
116 
117 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
118 
119 static int ocfs2_sync_fs(struct super_block *sb, int wait);
120 
121 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
122 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
123 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
124 static int ocfs2_check_volume(struct ocfs2_super *osb);
125 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
126 			       struct buffer_head *bh,
127 			       u32 sectsize,
128 			       struct ocfs2_blockcheck_stats *stats);
129 static int ocfs2_initialize_super(struct super_block *sb,
130 				  struct buffer_head *bh,
131 				  int sector_size,
132 				  struct ocfs2_blockcheck_stats *stats);
133 static int ocfs2_get_sector(struct super_block *sb,
134 			    struct buffer_head **bh,
135 			    int block,
136 			    int sect_size);
137 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
138 static void ocfs2_destroy_inode(struct inode *inode);
139 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
140 static int ocfs2_enable_quotas(struct ocfs2_super *osb);
141 static void ocfs2_disable_quotas(struct ocfs2_super *osb);
142 
143 static const struct super_operations ocfs2_sops = {
144 	.statfs		= ocfs2_statfs,
145 	.alloc_inode	= ocfs2_alloc_inode,
146 	.destroy_inode	= ocfs2_destroy_inode,
147 	.drop_inode	= ocfs2_drop_inode,
148 	.evict_inode	= ocfs2_evict_inode,
149 	.sync_fs	= ocfs2_sync_fs,
150 	.put_super	= ocfs2_put_super,
151 	.remount_fs	= ocfs2_remount,
152 	.show_options   = ocfs2_show_options,
153 	.quota_read	= ocfs2_quota_read,
154 	.quota_write	= ocfs2_quota_write,
155 };
156 
157 enum {
158 	Opt_barrier,
159 	Opt_err_panic,
160 	Opt_err_ro,
161 	Opt_intr,
162 	Opt_nointr,
163 	Opt_hb_none,
164 	Opt_hb_local,
165 	Opt_hb_global,
166 	Opt_data_ordered,
167 	Opt_data_writeback,
168 	Opt_atime_quantum,
169 	Opt_slot,
170 	Opt_commit,
171 	Opt_localalloc,
172 	Opt_localflocks,
173 	Opt_stack,
174 	Opt_user_xattr,
175 	Opt_nouser_xattr,
176 	Opt_inode64,
177 	Opt_acl,
178 	Opt_noacl,
179 	Opt_usrquota,
180 	Opt_grpquota,
181 	Opt_coherency_buffered,
182 	Opt_coherency_full,
183 	Opt_resv_level,
184 	Opt_dir_resv_level,
185 	Opt_err,
186 };
187 
188 static const match_table_t tokens = {
189 	{Opt_barrier, "barrier=%u"},
190 	{Opt_err_panic, "errors=panic"},
191 	{Opt_err_ro, "errors=remount-ro"},
192 	{Opt_intr, "intr"},
193 	{Opt_nointr, "nointr"},
194 	{Opt_hb_none, OCFS2_HB_NONE},
195 	{Opt_hb_local, OCFS2_HB_LOCAL},
196 	{Opt_hb_global, OCFS2_HB_GLOBAL},
197 	{Opt_data_ordered, "data=ordered"},
198 	{Opt_data_writeback, "data=writeback"},
199 	{Opt_atime_quantum, "atime_quantum=%u"},
200 	{Opt_slot, "preferred_slot=%u"},
201 	{Opt_commit, "commit=%u"},
202 	{Opt_localalloc, "localalloc=%d"},
203 	{Opt_localflocks, "localflocks"},
204 	{Opt_stack, "cluster_stack=%s"},
205 	{Opt_user_xattr, "user_xattr"},
206 	{Opt_nouser_xattr, "nouser_xattr"},
207 	{Opt_inode64, "inode64"},
208 	{Opt_acl, "acl"},
209 	{Opt_noacl, "noacl"},
210 	{Opt_usrquota, "usrquota"},
211 	{Opt_grpquota, "grpquota"},
212 	{Opt_coherency_buffered, "coherency=buffered"},
213 	{Opt_coherency_full, "coherency=full"},
214 	{Opt_resv_level, "resv_level=%u"},
215 	{Opt_dir_resv_level, "dir_resv_level=%u"},
216 	{Opt_err, NULL}
217 };
218 
219 #ifdef CONFIG_DEBUG_FS
220 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
221 {
222 	struct ocfs2_cluster_connection *cconn = osb->cconn;
223 	struct ocfs2_recovery_map *rm = osb->recovery_map;
224 	struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
225 	int i, out = 0;
226 
227 	out += snprintf(buf + out, len - out,
228 			"%10s => Id: %-s  Uuid: %-s  Gen: 0x%X  Label: %-s\n",
229 			"Device", osb->dev_str, osb->uuid_str,
230 			osb->fs_generation, osb->vol_label);
231 
232 	out += snprintf(buf + out, len - out,
233 			"%10s => State: %d  Flags: 0x%lX\n", "Volume",
234 			atomic_read(&osb->vol_state), osb->osb_flags);
235 
236 	out += snprintf(buf + out, len - out,
237 			"%10s => Block: %lu  Cluster: %d\n", "Sizes",
238 			osb->sb->s_blocksize, osb->s_clustersize);
239 
240 	out += snprintf(buf + out, len - out,
241 			"%10s => Compat: 0x%X  Incompat: 0x%X  "
242 			"ROcompat: 0x%X\n",
243 			"Features", osb->s_feature_compat,
244 			osb->s_feature_incompat, osb->s_feature_ro_compat);
245 
246 	out += snprintf(buf + out, len - out,
247 			"%10s => Opts: 0x%lX  AtimeQuanta: %u\n", "Mount",
248 			osb->s_mount_opt, osb->s_atime_quantum);
249 
250 	if (cconn) {
251 		out += snprintf(buf + out, len - out,
252 				"%10s => Stack: %s  Name: %*s  "
253 				"Version: %d.%d\n", "Cluster",
254 				(*osb->osb_cluster_stack == '\0' ?
255 				 "o2cb" : osb->osb_cluster_stack),
256 				cconn->cc_namelen, cconn->cc_name,
257 				cconn->cc_version.pv_major,
258 				cconn->cc_version.pv_minor);
259 	}
260 
261 	spin_lock(&osb->dc_task_lock);
262 	out += snprintf(buf + out, len - out,
263 			"%10s => Pid: %d  Count: %lu  WakeSeq: %lu  "
264 			"WorkSeq: %lu\n", "DownCnvt",
265 			(osb->dc_task ?  task_pid_nr(osb->dc_task) : -1),
266 			osb->blocked_lock_count, osb->dc_wake_sequence,
267 			osb->dc_work_sequence);
268 	spin_unlock(&osb->dc_task_lock);
269 
270 	spin_lock(&osb->osb_lock);
271 	out += snprintf(buf + out, len - out, "%10s => Pid: %d  Nodes:",
272 			"Recovery",
273 			(osb->recovery_thread_task ?
274 			 task_pid_nr(osb->recovery_thread_task) : -1));
275 	if (rm->rm_used == 0)
276 		out += snprintf(buf + out, len - out, " None\n");
277 	else {
278 		for (i = 0; i < rm->rm_used; i++)
279 			out += snprintf(buf + out, len - out, " %d",
280 					rm->rm_entries[i]);
281 		out += snprintf(buf + out, len - out, "\n");
282 	}
283 	spin_unlock(&osb->osb_lock);
284 
285 	out += snprintf(buf + out, len - out,
286 			"%10s => Pid: %d  Interval: %lu  Needs: %d\n", "Commit",
287 			(osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
288 			osb->osb_commit_interval,
289 			atomic_read(&osb->needs_checkpoint));
290 
291 	out += snprintf(buf + out, len - out,
292 			"%10s => State: %d  TxnId: %lu  NumTxns: %d\n",
293 			"Journal", osb->journal->j_state,
294 			osb->journal->j_trans_id,
295 			atomic_read(&osb->journal->j_num_trans));
296 
297 	out += snprintf(buf + out, len - out,
298 			"%10s => GlobalAllocs: %d  LocalAllocs: %d  "
299 			"SubAllocs: %d  LAWinMoves: %d  SAExtends: %d\n",
300 			"Stats",
301 			atomic_read(&osb->alloc_stats.bitmap_data),
302 			atomic_read(&osb->alloc_stats.local_data),
303 			atomic_read(&osb->alloc_stats.bg_allocs),
304 			atomic_read(&osb->alloc_stats.moves),
305 			atomic_read(&osb->alloc_stats.bg_extends));
306 
307 	out += snprintf(buf + out, len - out,
308 			"%10s => State: %u  Descriptor: %llu  Size: %u bits  "
309 			"Default: %u bits\n",
310 			"LocalAlloc", osb->local_alloc_state,
311 			(unsigned long long)osb->la_last_gd,
312 			osb->local_alloc_bits, osb->local_alloc_default_bits);
313 
314 	spin_lock(&osb->osb_lock);
315 	out += snprintf(buf + out, len - out,
316 			"%10s => InodeSlot: %d  StolenInodes: %d, "
317 			"MetaSlot: %d  StolenMeta: %d\n", "Steal",
318 			osb->s_inode_steal_slot,
319 			atomic_read(&osb->s_num_inodes_stolen),
320 			osb->s_meta_steal_slot,
321 			atomic_read(&osb->s_num_meta_stolen));
322 	spin_unlock(&osb->osb_lock);
323 
324 	out += snprintf(buf + out, len - out, "OrphanScan => ");
325 	out += snprintf(buf + out, len - out, "Local: %u  Global: %u ",
326 			os->os_count, os->os_seqno);
327 	out += snprintf(buf + out, len - out, " Last Scan: ");
328 	if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
329 		out += snprintf(buf + out, len - out, "Disabled\n");
330 	else
331 		out += snprintf(buf + out, len - out, "%lu seconds ago\n",
332 				(get_seconds() - os->os_scantime.tv_sec));
333 
334 	out += snprintf(buf + out, len - out, "%10s => %3s  %10s\n",
335 			"Slots", "Num", "RecoGen");
336 	for (i = 0; i < osb->max_slots; ++i) {
337 		out += snprintf(buf + out, len - out,
338 				"%10s  %c %3d  %10d\n",
339 				" ",
340 				(i == osb->slot_num ? '*' : ' '),
341 				i, osb->slot_recovery_generations[i]);
342 	}
343 
344 	return out;
345 }
346 
347 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
348 {
349 	struct ocfs2_super *osb = inode->i_private;
350 	char *buf = NULL;
351 
352 	buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
353 	if (!buf)
354 		goto bail;
355 
356 	i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
357 
358 	file->private_data = buf;
359 
360 	return 0;
361 bail:
362 	return -ENOMEM;
363 }
364 
365 static int ocfs2_debug_release(struct inode *inode, struct file *file)
366 {
367 	kfree(file->private_data);
368 	return 0;
369 }
370 
371 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
372 				size_t nbytes, loff_t *ppos)
373 {
374 	return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
375 				       i_size_read(file->f_mapping->host));
376 }
377 #else
378 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
379 {
380 	return 0;
381 }
382 static int ocfs2_debug_release(struct inode *inode, struct file *file)
383 {
384 	return 0;
385 }
386 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
387 				size_t nbytes, loff_t *ppos)
388 {
389 	return 0;
390 }
391 #endif	/* CONFIG_DEBUG_FS */
392 
393 static const struct file_operations ocfs2_osb_debug_fops = {
394 	.open =		ocfs2_osb_debug_open,
395 	.release =	ocfs2_debug_release,
396 	.read =		ocfs2_debug_read,
397 	.llseek =	generic_file_llseek,
398 };
399 
400 static int ocfs2_sync_fs(struct super_block *sb, int wait)
401 {
402 	int status;
403 	tid_t target;
404 	struct ocfs2_super *osb = OCFS2_SB(sb);
405 
406 	if (ocfs2_is_hard_readonly(osb))
407 		return -EROFS;
408 
409 	if (wait) {
410 		status = ocfs2_flush_truncate_log(osb);
411 		if (status < 0)
412 			mlog_errno(status);
413 	} else {
414 		ocfs2_schedule_truncate_log_flush(osb, 0);
415 	}
416 
417 	if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
418 				      &target)) {
419 		if (wait)
420 			jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
421 					     target);
422 	}
423 	return 0;
424 }
425 
426 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
427 {
428 	if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
429 	    && (ino == USER_QUOTA_SYSTEM_INODE
430 		|| ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
431 		return 0;
432 	if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
433 	    && (ino == GROUP_QUOTA_SYSTEM_INODE
434 		|| ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
435 		return 0;
436 	return 1;
437 }
438 
439 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
440 {
441 	struct inode *new = NULL;
442 	int status = 0;
443 	int i;
444 
445 	mlog_entry_void();
446 
447 	new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
448 	if (IS_ERR(new)) {
449 		status = PTR_ERR(new);
450 		mlog_errno(status);
451 		goto bail;
452 	}
453 	osb->root_inode = new;
454 
455 	new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
456 	if (IS_ERR(new)) {
457 		status = PTR_ERR(new);
458 		mlog_errno(status);
459 		goto bail;
460 	}
461 	osb->sys_root_inode = new;
462 
463 	for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
464 	     i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
465 		if (!ocfs2_need_system_inode(osb, i))
466 			continue;
467 		new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
468 		if (!new) {
469 			ocfs2_release_system_inodes(osb);
470 			status = -EINVAL;
471 			mlog_errno(status);
472 			/* FIXME: Should ERROR_RO_FS */
473 			mlog(ML_ERROR, "Unable to load system inode %d, "
474 			     "possibly corrupt fs?", i);
475 			goto bail;
476 		}
477 		// the array now has one ref, so drop this one
478 		iput(new);
479 	}
480 
481 bail:
482 	mlog_exit(status);
483 	return status;
484 }
485 
486 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
487 {
488 	struct inode *new = NULL;
489 	int status = 0;
490 	int i;
491 
492 	mlog_entry_void();
493 
494 	for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
495 	     i < NUM_SYSTEM_INODES;
496 	     i++) {
497 		if (!ocfs2_need_system_inode(osb, i))
498 			continue;
499 		new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
500 		if (!new) {
501 			ocfs2_release_system_inodes(osb);
502 			status = -EINVAL;
503 			mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
504 			     status, i, osb->slot_num);
505 			goto bail;
506 		}
507 		/* the array now has one ref, so drop this one */
508 		iput(new);
509 	}
510 
511 bail:
512 	mlog_exit(status);
513 	return status;
514 }
515 
516 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
517 {
518 	int i;
519 	struct inode *inode;
520 
521 	mlog_entry_void();
522 
523 	for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
524 		inode = osb->global_system_inodes[i];
525 		if (inode) {
526 			iput(inode);
527 			osb->global_system_inodes[i] = NULL;
528 		}
529 	}
530 
531 	inode = osb->sys_root_inode;
532 	if (inode) {
533 		iput(inode);
534 		osb->sys_root_inode = NULL;
535 	}
536 
537 	inode = osb->root_inode;
538 	if (inode) {
539 		iput(inode);
540 		osb->root_inode = NULL;
541 	}
542 
543 	if (!osb->local_system_inodes)
544 		goto out;
545 
546 	for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
547 		if (osb->local_system_inodes[i]) {
548 			iput(osb->local_system_inodes[i]);
549 			osb->local_system_inodes[i] = NULL;
550 		}
551 	}
552 
553 	kfree(osb->local_system_inodes);
554 	osb->local_system_inodes = NULL;
555 
556 out:
557 	mlog_exit(0);
558 }
559 
560 /* We're allocating fs objects, use GFP_NOFS */
561 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
562 {
563 	struct ocfs2_inode_info *oi;
564 
565 	oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
566 	if (!oi)
567 		return NULL;
568 
569 	jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
570 	return &oi->vfs_inode;
571 }
572 
573 static void ocfs2_destroy_inode(struct inode *inode)
574 {
575 	kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
576 }
577 
578 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
579 						unsigned int cbits)
580 {
581 	unsigned int bytes = 1 << cbits;
582 	unsigned int trim = bytes;
583 	unsigned int bitshift = 32;
584 
585 	/*
586 	 * i_size and all block offsets in ocfs2 are always 64 bits
587 	 * wide. i_clusters is 32 bits, in cluster-sized units. So on
588 	 * 64 bit platforms, cluster size will be the limiting factor.
589 	 */
590 
591 #if BITS_PER_LONG == 32
592 # if defined(CONFIG_LBDAF)
593 	BUILD_BUG_ON(sizeof(sector_t) != 8);
594 	/*
595 	 * We might be limited by page cache size.
596 	 */
597 	if (bytes > PAGE_CACHE_SIZE) {
598 		bytes = PAGE_CACHE_SIZE;
599 		trim = 1;
600 		/*
601 		 * Shift by 31 here so that we don't get larger than
602 		 * MAX_LFS_FILESIZE
603 		 */
604 		bitshift = 31;
605 	}
606 # else
607 	/*
608 	 * We are limited by the size of sector_t. Use block size, as
609 	 * that's what we expose to the VFS.
610 	 */
611 	bytes = 1 << bbits;
612 	trim = 1;
613 	bitshift = 31;
614 # endif
615 #endif
616 
617 	/*
618 	 * Trim by a whole cluster when we can actually approach the
619 	 * on-disk limits. Otherwise we can overflow i_clusters when
620 	 * an extent start is at the max offset.
621 	 */
622 	return (((unsigned long long)bytes) << bitshift) - trim;
623 }
624 
625 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
626 {
627 	int incompat_features;
628 	int ret = 0;
629 	struct mount_options parsed_options;
630 	struct ocfs2_super *osb = OCFS2_SB(sb);
631 	u32 tmp;
632 
633 	lock_kernel();
634 
635 	if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
636 	    !ocfs2_check_set_options(sb, &parsed_options)) {
637 		ret = -EINVAL;
638 		goto out;
639 	}
640 
641 	tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
642 		OCFS2_MOUNT_HB_NONE;
643 	if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
644 		ret = -EINVAL;
645 		mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
646 		goto out;
647 	}
648 
649 	if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
650 	    (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
651 		ret = -EINVAL;
652 		mlog(ML_ERROR, "Cannot change data mode on remount\n");
653 		goto out;
654 	}
655 
656 	/* Probably don't want this on remount; it might
657 	 * mess with other nodes */
658 	if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
659 	    (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
660 		ret = -EINVAL;
661 		mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
662 		goto out;
663 	}
664 
665 	/* We're going to/from readonly mode. */
666 	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
667 		/* Disable quota accounting before remounting RO */
668 		if (*flags & MS_RDONLY) {
669 			ret = ocfs2_susp_quotas(osb, 0);
670 			if (ret < 0)
671 				goto out;
672 		}
673 		/* Lock here so the check of HARD_RO and the potential
674 		 * setting of SOFT_RO is atomic. */
675 		spin_lock(&osb->osb_lock);
676 		if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
677 			mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
678 			ret = -EROFS;
679 			goto unlock_osb;
680 		}
681 
682 		if (*flags & MS_RDONLY) {
683 			mlog(0, "Going to ro mode.\n");
684 			sb->s_flags |= MS_RDONLY;
685 			osb->osb_flags |= OCFS2_OSB_SOFT_RO;
686 		} else {
687 			mlog(0, "Making ro filesystem writeable.\n");
688 
689 			if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
690 				mlog(ML_ERROR, "Cannot remount RDWR "
691 				     "filesystem due to previous errors.\n");
692 				ret = -EROFS;
693 				goto unlock_osb;
694 			}
695 			incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
696 			if (incompat_features) {
697 				mlog(ML_ERROR, "Cannot remount RDWR because "
698 				     "of unsupported optional features "
699 				     "(%x).\n", incompat_features);
700 				ret = -EINVAL;
701 				goto unlock_osb;
702 			}
703 			sb->s_flags &= ~MS_RDONLY;
704 			osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
705 		}
706 unlock_osb:
707 		spin_unlock(&osb->osb_lock);
708 		/* Enable quota accounting after remounting RW */
709 		if (!ret && !(*flags & MS_RDONLY)) {
710 			if (sb_any_quota_suspended(sb))
711 				ret = ocfs2_susp_quotas(osb, 1);
712 			else
713 				ret = ocfs2_enable_quotas(osb);
714 			if (ret < 0) {
715 				/* Return back changes... */
716 				spin_lock(&osb->osb_lock);
717 				sb->s_flags |= MS_RDONLY;
718 				osb->osb_flags |= OCFS2_OSB_SOFT_RO;
719 				spin_unlock(&osb->osb_lock);
720 				goto out;
721 			}
722 		}
723 	}
724 
725 	if (!ret) {
726 		/* Only save off the new mount options in case of a successful
727 		 * remount. */
728 		osb->s_mount_opt = parsed_options.mount_opt;
729 		osb->s_atime_quantum = parsed_options.atime_quantum;
730 		osb->preferred_slot = parsed_options.slot;
731 		if (parsed_options.commit_interval)
732 			osb->osb_commit_interval = parsed_options.commit_interval;
733 
734 		if (!ocfs2_is_hard_readonly(osb))
735 			ocfs2_set_journal_params(osb);
736 
737 		sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
738 			((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
739 							MS_POSIXACL : 0);
740 	}
741 out:
742 	unlock_kernel();
743 	return ret;
744 }
745 
746 static int ocfs2_sb_probe(struct super_block *sb,
747 			  struct buffer_head **bh,
748 			  int *sector_size,
749 			  struct ocfs2_blockcheck_stats *stats)
750 {
751 	int status, tmpstat;
752 	struct ocfs1_vol_disk_hdr *hdr;
753 	struct ocfs2_dinode *di;
754 	int blksize;
755 
756 	*bh = NULL;
757 
758 	/* may be > 512 */
759 	*sector_size = bdev_logical_block_size(sb->s_bdev);
760 	if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
761 		mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
762 		     *sector_size, OCFS2_MAX_BLOCKSIZE);
763 		status = -EINVAL;
764 		goto bail;
765 	}
766 
767 	/* Can this really happen? */
768 	if (*sector_size < OCFS2_MIN_BLOCKSIZE)
769 		*sector_size = OCFS2_MIN_BLOCKSIZE;
770 
771 	/* check block zero for old format */
772 	status = ocfs2_get_sector(sb, bh, 0, *sector_size);
773 	if (status < 0) {
774 		mlog_errno(status);
775 		goto bail;
776 	}
777 	hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
778 	if (hdr->major_version == OCFS1_MAJOR_VERSION) {
779 		mlog(ML_ERROR, "incompatible version: %u.%u\n",
780 		     hdr->major_version, hdr->minor_version);
781 		status = -EINVAL;
782 	}
783 	if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
784 		   strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
785 		mlog(ML_ERROR, "incompatible volume signature: %8s\n",
786 		     hdr->signature);
787 		status = -EINVAL;
788 	}
789 	brelse(*bh);
790 	*bh = NULL;
791 	if (status < 0) {
792 		mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
793 		     "upgraded before mounting with ocfs v2\n");
794 		goto bail;
795 	}
796 
797 	/*
798 	 * Now check at magic offset for 512, 1024, 2048, 4096
799 	 * blocksizes.  4096 is the maximum blocksize because it is
800 	 * the minimum clustersize.
801 	 */
802 	status = -EINVAL;
803 	for (blksize = *sector_size;
804 	     blksize <= OCFS2_MAX_BLOCKSIZE;
805 	     blksize <<= 1) {
806 		tmpstat = ocfs2_get_sector(sb, bh,
807 					   OCFS2_SUPER_BLOCK_BLKNO,
808 					   blksize);
809 		if (tmpstat < 0) {
810 			status = tmpstat;
811 			mlog_errno(status);
812 			break;
813 		}
814 		di = (struct ocfs2_dinode *) (*bh)->b_data;
815 		memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
816 		spin_lock_init(&stats->b_lock);
817 		tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
818 		if (tmpstat < 0) {
819 			brelse(*bh);
820 			*bh = NULL;
821 		}
822 		if (tmpstat != -EAGAIN) {
823 			status = tmpstat;
824 			break;
825 		}
826 	}
827 
828 bail:
829 	return status;
830 }
831 
832 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
833 {
834 	u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
835 
836 	if (osb->s_mount_opt & hb_enabled) {
837 		if (ocfs2_mount_local(osb)) {
838 			mlog(ML_ERROR, "Cannot heartbeat on a locally "
839 			     "mounted device.\n");
840 			return -EINVAL;
841 		}
842 		if (ocfs2_userspace_stack(osb)) {
843 			mlog(ML_ERROR, "Userspace stack expected, but "
844 			     "o2cb heartbeat arguments passed to mount\n");
845 			return -EINVAL;
846 		}
847 		if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
848 		     !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
849 		    ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
850 		     ocfs2_cluster_o2cb_global_heartbeat(osb))) {
851 			mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
852 			return -EINVAL;
853 		}
854 	}
855 
856 	if (!(osb->s_mount_opt & hb_enabled)) {
857 		if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
858 		    !ocfs2_userspace_stack(osb)) {
859 			mlog(ML_ERROR, "Heartbeat has to be started to mount "
860 			     "a read-write clustered device.\n");
861 			return -EINVAL;
862 		}
863 	}
864 
865 	return 0;
866 }
867 
868 /*
869  * If we're using a userspace stack, mount should have passed
870  * a name that matches the disk.  If not, mount should not
871  * have passed a stack.
872  */
873 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
874 					struct mount_options *mopt)
875 {
876 	if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
877 		mlog(ML_ERROR,
878 		     "cluster stack passed to mount, but this filesystem "
879 		     "does not support it\n");
880 		return -EINVAL;
881 	}
882 
883 	if (ocfs2_userspace_stack(osb) &&
884 	    strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
885 		    OCFS2_STACK_LABEL_LEN)) {
886 		mlog(ML_ERROR,
887 		     "cluster stack passed to mount (\"%s\") does not "
888 		     "match the filesystem (\"%s\")\n",
889 		     mopt->cluster_stack,
890 		     osb->osb_cluster_stack);
891 		return -EINVAL;
892 	}
893 
894 	return 0;
895 }
896 
897 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
898 {
899 	int type;
900 	struct super_block *sb = osb->sb;
901 	unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
902 					     OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
903 	int status = 0;
904 
905 	for (type = 0; type < MAXQUOTAS; type++) {
906 		if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
907 			continue;
908 		if (unsuspend)
909 			status = dquot_resume(sb, type);
910 		else {
911 			struct ocfs2_mem_dqinfo *oinfo;
912 
913 			/* Cancel periodic syncing before suspending */
914 			oinfo = sb_dqinfo(sb, type)->dqi_priv;
915 			cancel_delayed_work_sync(&oinfo->dqi_sync_work);
916 			status = dquot_suspend(sb, type);
917 		}
918 		if (status < 0)
919 			break;
920 	}
921 	if (status < 0)
922 		mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
923 		     "remount (error = %d).\n", status);
924 	return status;
925 }
926 
927 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
928 {
929 	struct inode *inode[MAXQUOTAS] = { NULL, NULL };
930 	struct super_block *sb = osb->sb;
931 	unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
932 					     OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
933 	unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
934 					LOCAL_GROUP_QUOTA_SYSTEM_INODE };
935 	int status;
936 	int type;
937 
938 	sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
939 	for (type = 0; type < MAXQUOTAS; type++) {
940 		if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
941 			continue;
942 		inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
943 							osb->slot_num);
944 		if (!inode[type]) {
945 			status = -ENOENT;
946 			goto out_quota_off;
947 		}
948 		status = dquot_enable(inode[type], type, QFMT_OCFS2,
949 				      DQUOT_USAGE_ENABLED);
950 		if (status < 0)
951 			goto out_quota_off;
952 	}
953 
954 	for (type = 0; type < MAXQUOTAS; type++)
955 		iput(inode[type]);
956 	return 0;
957 out_quota_off:
958 	ocfs2_disable_quotas(osb);
959 	for (type = 0; type < MAXQUOTAS; type++)
960 		iput(inode[type]);
961 	mlog_errno(status);
962 	return status;
963 }
964 
965 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
966 {
967 	int type;
968 	struct inode *inode;
969 	struct super_block *sb = osb->sb;
970 	struct ocfs2_mem_dqinfo *oinfo;
971 
972 	/* We mostly ignore errors in this function because there's not much
973 	 * we can do when we see them */
974 	for (type = 0; type < MAXQUOTAS; type++) {
975 		if (!sb_has_quota_loaded(sb, type))
976 			continue;
977 		/* Cancel periodic syncing before we grab dqonoff_mutex */
978 		oinfo = sb_dqinfo(sb, type)->dqi_priv;
979 		cancel_delayed_work_sync(&oinfo->dqi_sync_work);
980 		inode = igrab(sb->s_dquot.files[type]);
981 		/* Turn off quotas. This will remove all dquot structures from
982 		 * memory and so they will be automatically synced to global
983 		 * quota files */
984 		dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
985 					DQUOT_LIMITS_ENABLED);
986 		if (!inode)
987 			continue;
988 		iput(inode);
989 	}
990 }
991 
992 /* Handle quota on quotactl */
993 static int ocfs2_quota_on(struct super_block *sb, int type, int format_id,
994 			  char *path)
995 {
996 	unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
997 					     OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
998 
999 	if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
1000 		return -EINVAL;
1001 
1002 	return dquot_enable(sb_dqopt(sb)->files[type], type,
1003 			    format_id, DQUOT_LIMITS_ENABLED);
1004 }
1005 
1006 /* Handle quota off quotactl */
1007 static int ocfs2_quota_off(struct super_block *sb, int type)
1008 {
1009 	return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
1010 }
1011 
1012 static const struct quotactl_ops ocfs2_quotactl_ops = {
1013 	.quota_on	= ocfs2_quota_on,
1014 	.quota_off	= ocfs2_quota_off,
1015 	.quota_sync	= dquot_quota_sync,
1016 	.get_info	= dquot_get_dqinfo,
1017 	.set_info	= dquot_set_dqinfo,
1018 	.get_dqblk	= dquot_get_dqblk,
1019 	.set_dqblk	= dquot_set_dqblk,
1020 };
1021 
1022 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
1023 {
1024 	struct dentry *root;
1025 	int status, sector_size;
1026 	struct mount_options parsed_options;
1027 	struct inode *inode = NULL;
1028 	struct ocfs2_super *osb = NULL;
1029 	struct buffer_head *bh = NULL;
1030 	char nodestr[8];
1031 	struct ocfs2_blockcheck_stats stats;
1032 
1033 	mlog_entry("%p, %p, %i", sb, data, silent);
1034 
1035 	if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
1036 		status = -EINVAL;
1037 		goto read_super_error;
1038 	}
1039 
1040 	/* probe for superblock */
1041 	status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
1042 	if (status < 0) {
1043 		mlog(ML_ERROR, "superblock probe failed!\n");
1044 		goto read_super_error;
1045 	}
1046 
1047 	status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1048 	osb = OCFS2_SB(sb);
1049 	if (status < 0) {
1050 		mlog_errno(status);
1051 		goto read_super_error;
1052 	}
1053 	brelse(bh);
1054 	bh = NULL;
1055 
1056 	if (!ocfs2_check_set_options(sb, &parsed_options)) {
1057 		status = -EINVAL;
1058 		goto read_super_error;
1059 	}
1060 	osb->s_mount_opt = parsed_options.mount_opt;
1061 	osb->s_atime_quantum = parsed_options.atime_quantum;
1062 	osb->preferred_slot = parsed_options.slot;
1063 	osb->osb_commit_interval = parsed_options.commit_interval;
1064 
1065 	ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1066 	osb->osb_resv_level = parsed_options.resv_level;
1067 	osb->osb_dir_resv_level = parsed_options.resv_level;
1068 	if (parsed_options.dir_resv_level == -1)
1069 		osb->osb_dir_resv_level = parsed_options.resv_level;
1070 	else
1071 		osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1072 
1073 	status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1074 	if (status)
1075 		goto read_super_error;
1076 
1077 	sb->s_magic = OCFS2_SUPER_MAGIC;
1078 
1079 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1080 		((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1081 
1082 	/* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1083 	 * heartbeat=none */
1084 	if (bdev_read_only(sb->s_bdev)) {
1085 		if (!(sb->s_flags & MS_RDONLY)) {
1086 			status = -EACCES;
1087 			mlog(ML_ERROR, "Readonly device detected but readonly "
1088 			     "mount was not specified.\n");
1089 			goto read_super_error;
1090 		}
1091 
1092 		/* You should not be able to start a local heartbeat
1093 		 * on a readonly device. */
1094 		if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1095 			status = -EROFS;
1096 			mlog(ML_ERROR, "Local heartbeat specified on readonly "
1097 			     "device.\n");
1098 			goto read_super_error;
1099 		}
1100 
1101 		status = ocfs2_check_journals_nolocks(osb);
1102 		if (status < 0) {
1103 			if (status == -EROFS)
1104 				mlog(ML_ERROR, "Recovery required on readonly "
1105 				     "file system, but write access is "
1106 				     "unavailable.\n");
1107 			else
1108 				mlog_errno(status);
1109 			goto read_super_error;
1110 		}
1111 
1112 		ocfs2_set_ro_flag(osb, 1);
1113 
1114 		printk(KERN_NOTICE "Readonly device detected. No cluster "
1115 		       "services will be utilized for this mount. Recovery "
1116 		       "will be skipped.\n");
1117 	}
1118 
1119 	if (!ocfs2_is_hard_readonly(osb)) {
1120 		if (sb->s_flags & MS_RDONLY)
1121 			ocfs2_set_ro_flag(osb, 0);
1122 	}
1123 
1124 	status = ocfs2_verify_heartbeat(osb);
1125 	if (status < 0) {
1126 		mlog_errno(status);
1127 		goto read_super_error;
1128 	}
1129 
1130 	osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1131 						 ocfs2_debugfs_root);
1132 	if (!osb->osb_debug_root) {
1133 		status = -EINVAL;
1134 		mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1135 		goto read_super_error;
1136 	}
1137 
1138 	osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1139 					    osb->osb_debug_root,
1140 					    osb,
1141 					    &ocfs2_osb_debug_fops);
1142 	if (!osb->osb_ctxt) {
1143 		status = -EINVAL;
1144 		mlog_errno(status);
1145 		goto read_super_error;
1146 	}
1147 
1148 	if (ocfs2_meta_ecc(osb)) {
1149 		status = ocfs2_blockcheck_stats_debugfs_install(
1150 						&osb->osb_ecc_stats,
1151 						osb->osb_debug_root);
1152 		if (status) {
1153 			mlog(ML_ERROR,
1154 			     "Unable to create blockcheck statistics "
1155 			     "files\n");
1156 			goto read_super_error;
1157 		}
1158 	}
1159 
1160 	status = ocfs2_mount_volume(sb);
1161 	if (osb->root_inode)
1162 		inode = igrab(osb->root_inode);
1163 
1164 	if (status < 0)
1165 		goto read_super_error;
1166 
1167 	if (!inode) {
1168 		status = -EIO;
1169 		mlog_errno(status);
1170 		goto read_super_error;
1171 	}
1172 
1173 	root = d_alloc_root(inode);
1174 	if (!root) {
1175 		status = -ENOMEM;
1176 		mlog_errno(status);
1177 		goto read_super_error;
1178 	}
1179 
1180 	sb->s_root = root;
1181 
1182 	ocfs2_complete_mount_recovery(osb);
1183 
1184 	if (ocfs2_mount_local(osb))
1185 		snprintf(nodestr, sizeof(nodestr), "local");
1186 	else
1187 		snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1188 
1189 	printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1190 	       "with %s data mode.\n",
1191 	       osb->dev_str, nodestr, osb->slot_num,
1192 	       osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1193 	       "ordered");
1194 
1195 	atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1196 	wake_up(&osb->osb_mount_event);
1197 
1198 	/* Now we can initialize quotas because we can afford to wait
1199 	 * for cluster locks recovery now. That also means that truncation
1200 	 * log recovery can happen but that waits for proper quota setup */
1201 	if (!(sb->s_flags & MS_RDONLY)) {
1202 		status = ocfs2_enable_quotas(osb);
1203 		if (status < 0) {
1204 			/* We have to err-out specially here because
1205 			 * s_root is already set */
1206 			mlog_errno(status);
1207 			atomic_set(&osb->vol_state, VOLUME_DISABLED);
1208 			wake_up(&osb->osb_mount_event);
1209 			mlog_exit(status);
1210 			return status;
1211 		}
1212 	}
1213 
1214 	ocfs2_complete_quota_recovery(osb);
1215 
1216 	/* Now we wake up again for processes waiting for quotas */
1217 	atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1218 	wake_up(&osb->osb_mount_event);
1219 
1220 	/* Start this when the mount is almost sure of being successful */
1221 	ocfs2_orphan_scan_start(osb);
1222 
1223 	mlog_exit(status);
1224 	return status;
1225 
1226 read_super_error:
1227 	brelse(bh);
1228 
1229 	if (inode)
1230 		iput(inode);
1231 
1232 	if (osb) {
1233 		atomic_set(&osb->vol_state, VOLUME_DISABLED);
1234 		wake_up(&osb->osb_mount_event);
1235 		ocfs2_dismount_volume(sb, 1);
1236 	}
1237 
1238 	mlog_exit(status);
1239 	return status;
1240 }
1241 
1242 static int ocfs2_get_sb(struct file_system_type *fs_type,
1243 			int flags,
1244 			const char *dev_name,
1245 			void *data,
1246 			struct vfsmount *mnt)
1247 {
1248 	return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super,
1249 			   mnt);
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 	.get_sb         = ocfs2_get_sb, /* is this called when we mount
1274 					* the fs? */
1275 	.kill_sb        = ocfs2_kill_sb,
1276 
1277 	.fs_flags       = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1278 	.next           = NULL
1279 };
1280 
1281 static int ocfs2_check_set_options(struct super_block *sb,
1282 				   struct mount_options *options)
1283 {
1284 	if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1285 	    !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1286 					 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1287 		mlog(ML_ERROR, "User quotas were requested, but this "
1288 		     "filesystem does not have the feature enabled.\n");
1289 		return 0;
1290 	}
1291 	if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1292 	    !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1293 					 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1294 		mlog(ML_ERROR, "Group quotas were requested, but this "
1295 		     "filesystem does not have the feature enabled.\n");
1296 		return 0;
1297 	}
1298 	if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1299 	    !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1300 		mlog(ML_ERROR, "ACL support requested but extended attributes "
1301 		     "feature is not enabled\n");
1302 		return 0;
1303 	}
1304 	/* No ACL setting specified? Use XATTR feature... */
1305 	if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1306 				    OCFS2_MOUNT_NO_POSIX_ACL))) {
1307 		if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1308 			options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1309 		else
1310 			options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1311 	}
1312 	return 1;
1313 }
1314 
1315 static int ocfs2_parse_options(struct super_block *sb,
1316 			       char *options,
1317 			       struct mount_options *mopt,
1318 			       int is_remount)
1319 {
1320 	int status;
1321 	char *p;
1322 	u32 tmp;
1323 
1324 	mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
1325 		   options ? options : "(none)");
1326 
1327 	mopt->commit_interval = 0;
1328 	mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1329 	mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1330 	mopt->slot = OCFS2_INVALID_SLOT;
1331 	mopt->localalloc_opt = -1;
1332 	mopt->cluster_stack[0] = '\0';
1333 	mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1334 	mopt->dir_resv_level = -1;
1335 
1336 	if (!options) {
1337 		status = 1;
1338 		goto bail;
1339 	}
1340 
1341 	while ((p = strsep(&options, ",")) != NULL) {
1342 		int token, option;
1343 		substring_t args[MAX_OPT_ARGS];
1344 
1345 		if (!*p)
1346 			continue;
1347 
1348 		token = match_token(p, tokens, args);
1349 		switch (token) {
1350 		case Opt_hb_local:
1351 			mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1352 			break;
1353 		case Opt_hb_none:
1354 			mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1355 			break;
1356 		case Opt_hb_global:
1357 			mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1358 			break;
1359 		case Opt_barrier:
1360 			if (match_int(&args[0], &option)) {
1361 				status = 0;
1362 				goto bail;
1363 			}
1364 			if (option)
1365 				mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1366 			else
1367 				mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1368 			break;
1369 		case Opt_intr:
1370 			mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1371 			break;
1372 		case Opt_nointr:
1373 			mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1374 			break;
1375 		case Opt_err_panic:
1376 			mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1377 			break;
1378 		case Opt_err_ro:
1379 			mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1380 			break;
1381 		case Opt_data_ordered:
1382 			mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1383 			break;
1384 		case Opt_data_writeback:
1385 			mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1386 			break;
1387 		case Opt_user_xattr:
1388 			mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1389 			break;
1390 		case Opt_nouser_xattr:
1391 			mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1392 			break;
1393 		case Opt_atime_quantum:
1394 			if (match_int(&args[0], &option)) {
1395 				status = 0;
1396 				goto bail;
1397 			}
1398 			if (option >= 0)
1399 				mopt->atime_quantum = option;
1400 			break;
1401 		case Opt_slot:
1402 			option = 0;
1403 			if (match_int(&args[0], &option)) {
1404 				status = 0;
1405 				goto bail;
1406 			}
1407 			if (option)
1408 				mopt->slot = (s16)option;
1409 			break;
1410 		case Opt_commit:
1411 			option = 0;
1412 			if (match_int(&args[0], &option)) {
1413 				status = 0;
1414 				goto bail;
1415 			}
1416 			if (option < 0)
1417 				return 0;
1418 			if (option == 0)
1419 				option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1420 			mopt->commit_interval = HZ * option;
1421 			break;
1422 		case Opt_localalloc:
1423 			option = 0;
1424 			if (match_int(&args[0], &option)) {
1425 				status = 0;
1426 				goto bail;
1427 			}
1428 			if (option >= 0)
1429 				mopt->localalloc_opt = option;
1430 			break;
1431 		case Opt_localflocks:
1432 			/*
1433 			 * Changing this during remount could race
1434 			 * flock() requests, or "unbalance" existing
1435 			 * ones (e.g., a lock is taken in one mode but
1436 			 * dropped in the other). If users care enough
1437 			 * to flip locking modes during remount, we
1438 			 * could add a "local" flag to individual
1439 			 * flock structures for proper tracking of
1440 			 * state.
1441 			 */
1442 			if (!is_remount)
1443 				mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1444 			break;
1445 		case Opt_stack:
1446 			/* Check both that the option we were passed
1447 			 * is of the right length and that it is a proper
1448 			 * string of the right length.
1449 			 */
1450 			if (((args[0].to - args[0].from) !=
1451 			     OCFS2_STACK_LABEL_LEN) ||
1452 			    (strnlen(args[0].from,
1453 				     OCFS2_STACK_LABEL_LEN) !=
1454 			     OCFS2_STACK_LABEL_LEN)) {
1455 				mlog(ML_ERROR,
1456 				     "Invalid cluster_stack option\n");
1457 				status = 0;
1458 				goto bail;
1459 			}
1460 			memcpy(mopt->cluster_stack, args[0].from,
1461 			       OCFS2_STACK_LABEL_LEN);
1462 			mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1463 			break;
1464 		case Opt_inode64:
1465 			mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1466 			break;
1467 		case Opt_usrquota:
1468 			mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1469 			break;
1470 		case Opt_grpquota:
1471 			mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1472 			break;
1473 		case Opt_coherency_buffered:
1474 			mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1475 			break;
1476 		case Opt_coherency_full:
1477 			mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1478 			break;
1479 		case Opt_acl:
1480 			mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1481 			mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1482 			break;
1483 		case Opt_noacl:
1484 			mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1485 			mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1486 			break;
1487 		case Opt_resv_level:
1488 			if (is_remount)
1489 				break;
1490 			if (match_int(&args[0], &option)) {
1491 				status = 0;
1492 				goto bail;
1493 			}
1494 			if (option >= OCFS2_MIN_RESV_LEVEL &&
1495 			    option < OCFS2_MAX_RESV_LEVEL)
1496 				mopt->resv_level = option;
1497 			break;
1498 		case Opt_dir_resv_level:
1499 			if (is_remount)
1500 				break;
1501 			if (match_int(&args[0], &option)) {
1502 				status = 0;
1503 				goto bail;
1504 			}
1505 			if (option >= OCFS2_MIN_RESV_LEVEL &&
1506 			    option < OCFS2_MAX_RESV_LEVEL)
1507 				mopt->dir_resv_level = option;
1508 			break;
1509 		default:
1510 			mlog(ML_ERROR,
1511 			     "Unrecognized mount option \"%s\" "
1512 			     "or missing value\n", p);
1513 			status = 0;
1514 			goto bail;
1515 		}
1516 	}
1517 
1518 	/* Ensure only one heartbeat mode */
1519 	tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
1520 				 OCFS2_MOUNT_HB_NONE);
1521 	if (hweight32(tmp) != 1) {
1522 		mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1523 		status = 0;
1524 		goto bail;
1525 	}
1526 
1527 	status = 1;
1528 
1529 bail:
1530 	mlog_exit(status);
1531 	return status;
1532 }
1533 
1534 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1535 {
1536 	struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb);
1537 	unsigned long opts = osb->s_mount_opt;
1538 	unsigned int local_alloc_megs;
1539 
1540 	if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1541 		seq_printf(s, ",_netdev");
1542 		if (opts & OCFS2_MOUNT_HB_LOCAL)
1543 			seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1544 		else
1545 			seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1546 	} else
1547 		seq_printf(s, ",%s", OCFS2_HB_NONE);
1548 
1549 	if (opts & OCFS2_MOUNT_NOINTR)
1550 		seq_printf(s, ",nointr");
1551 
1552 	if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1553 		seq_printf(s, ",data=writeback");
1554 	else
1555 		seq_printf(s, ",data=ordered");
1556 
1557 	if (opts & OCFS2_MOUNT_BARRIER)
1558 		seq_printf(s, ",barrier=1");
1559 
1560 	if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1561 		seq_printf(s, ",errors=panic");
1562 	else
1563 		seq_printf(s, ",errors=remount-ro");
1564 
1565 	if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1566 		seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1567 
1568 	if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM)
1569 		seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1570 
1571 	if (osb->osb_commit_interval)
1572 		seq_printf(s, ",commit=%u",
1573 			   (unsigned) (osb->osb_commit_interval / HZ));
1574 
1575 	local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1576 	if (local_alloc_megs != ocfs2_la_default_mb(osb))
1577 		seq_printf(s, ",localalloc=%d", local_alloc_megs);
1578 
1579 	if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1580 		seq_printf(s, ",localflocks,");
1581 
1582 	if (osb->osb_cluster_stack[0])
1583 		seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1584 			   osb->osb_cluster_stack);
1585 	if (opts & OCFS2_MOUNT_USRQUOTA)
1586 		seq_printf(s, ",usrquota");
1587 	if (opts & OCFS2_MOUNT_GRPQUOTA)
1588 		seq_printf(s, ",grpquota");
1589 
1590 	if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1591 		seq_printf(s, ",coherency=buffered");
1592 	else
1593 		seq_printf(s, ",coherency=full");
1594 
1595 	if (opts & OCFS2_MOUNT_NOUSERXATTR)
1596 		seq_printf(s, ",nouser_xattr");
1597 	else
1598 		seq_printf(s, ",user_xattr");
1599 
1600 	if (opts & OCFS2_MOUNT_INODE64)
1601 		seq_printf(s, ",inode64");
1602 
1603 	if (opts & OCFS2_MOUNT_POSIX_ACL)
1604 		seq_printf(s, ",acl");
1605 	else
1606 		seq_printf(s, ",noacl");
1607 
1608 	if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1609 		seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1610 
1611 	if (osb->osb_dir_resv_level != osb->osb_resv_level)
1612 		seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1613 
1614 	return 0;
1615 }
1616 
1617 static int __init ocfs2_init(void)
1618 {
1619 	int status;
1620 
1621 	mlog_entry_void();
1622 
1623 	ocfs2_print_version();
1624 
1625 	status = init_ocfs2_uptodate_cache();
1626 	if (status < 0) {
1627 		mlog_errno(status);
1628 		goto leave;
1629 	}
1630 
1631 	status = ocfs2_initialize_mem_caches();
1632 	if (status < 0) {
1633 		mlog_errno(status);
1634 		goto leave;
1635 	}
1636 
1637 	ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1638 	if (!ocfs2_wq) {
1639 		status = -ENOMEM;
1640 		goto leave;
1641 	}
1642 
1643 	ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1644 	if (!ocfs2_debugfs_root) {
1645 		status = -EFAULT;
1646 		mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1647 	}
1648 
1649 	status = ocfs2_quota_setup();
1650 	if (status)
1651 		goto leave;
1652 
1653 	ocfs2_set_locking_protocol();
1654 
1655 	status = register_quota_format(&ocfs2_quota_format);
1656 leave:
1657 	if (status < 0) {
1658 		ocfs2_quota_shutdown();
1659 		ocfs2_free_mem_caches();
1660 		exit_ocfs2_uptodate_cache();
1661 	}
1662 
1663 	mlog_exit(status);
1664 
1665 	if (status >= 0) {
1666 		return register_filesystem(&ocfs2_fs_type);
1667 	} else
1668 		return -1;
1669 }
1670 
1671 static void __exit ocfs2_exit(void)
1672 {
1673 	mlog_entry_void();
1674 
1675 	ocfs2_quota_shutdown();
1676 
1677 	if (ocfs2_wq) {
1678 		flush_workqueue(ocfs2_wq);
1679 		destroy_workqueue(ocfs2_wq);
1680 	}
1681 
1682 	unregister_quota_format(&ocfs2_quota_format);
1683 
1684 	debugfs_remove(ocfs2_debugfs_root);
1685 
1686 	ocfs2_free_mem_caches();
1687 
1688 	unregister_filesystem(&ocfs2_fs_type);
1689 
1690 	exit_ocfs2_uptodate_cache();
1691 
1692 	mlog_exit_void();
1693 }
1694 
1695 static void ocfs2_put_super(struct super_block *sb)
1696 {
1697 	mlog_entry("(0x%p)\n", sb);
1698 
1699 	lock_kernel();
1700 
1701 	ocfs2_sync_blockdev(sb);
1702 	ocfs2_dismount_volume(sb, 0);
1703 
1704 	unlock_kernel();
1705 
1706 	mlog_exit_void();
1707 }
1708 
1709 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1710 {
1711 	struct ocfs2_super *osb;
1712 	u32 numbits, freebits;
1713 	int status;
1714 	struct ocfs2_dinode *bm_lock;
1715 	struct buffer_head *bh = NULL;
1716 	struct inode *inode = NULL;
1717 
1718 	mlog_entry("(%p, %p)\n", dentry->d_sb, buf);
1719 
1720 	osb = OCFS2_SB(dentry->d_sb);
1721 
1722 	inode = ocfs2_get_system_file_inode(osb,
1723 					    GLOBAL_BITMAP_SYSTEM_INODE,
1724 					    OCFS2_INVALID_SLOT);
1725 	if (!inode) {
1726 		mlog(ML_ERROR, "failed to get bitmap inode\n");
1727 		status = -EIO;
1728 		goto bail;
1729 	}
1730 
1731 	status = ocfs2_inode_lock(inode, &bh, 0);
1732 	if (status < 0) {
1733 		mlog_errno(status);
1734 		goto bail;
1735 	}
1736 
1737 	bm_lock = (struct ocfs2_dinode *) bh->b_data;
1738 
1739 	numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1740 	freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1741 
1742 	buf->f_type = OCFS2_SUPER_MAGIC;
1743 	buf->f_bsize = dentry->d_sb->s_blocksize;
1744 	buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1745 	buf->f_blocks = ((sector_t) numbits) *
1746 			(osb->s_clustersize >> osb->sb->s_blocksize_bits);
1747 	buf->f_bfree = ((sector_t) freebits) *
1748 		       (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1749 	buf->f_bavail = buf->f_bfree;
1750 	buf->f_files = numbits;
1751 	buf->f_ffree = freebits;
1752 	buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1753 				& 0xFFFFFFFFUL;
1754 	buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1755 				OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1756 
1757 	brelse(bh);
1758 
1759 	ocfs2_inode_unlock(inode, 0);
1760 	status = 0;
1761 bail:
1762 	if (inode)
1763 		iput(inode);
1764 
1765 	mlog_exit(status);
1766 
1767 	return status;
1768 }
1769 
1770 static void ocfs2_inode_init_once(void *data)
1771 {
1772 	struct ocfs2_inode_info *oi = data;
1773 
1774 	oi->ip_flags = 0;
1775 	oi->ip_open_count = 0;
1776 	spin_lock_init(&oi->ip_lock);
1777 	ocfs2_extent_map_init(&oi->vfs_inode);
1778 	INIT_LIST_HEAD(&oi->ip_io_markers);
1779 	oi->ip_dir_start_lookup = 0;
1780 
1781 	init_rwsem(&oi->ip_alloc_sem);
1782 	init_rwsem(&oi->ip_xattr_sem);
1783 	mutex_init(&oi->ip_io_mutex);
1784 
1785 	oi->ip_blkno = 0ULL;
1786 	oi->ip_clusters = 0;
1787 
1788 	ocfs2_resv_init_once(&oi->ip_la_data_resv);
1789 
1790 	ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1791 	ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1792 	ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1793 
1794 	ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1795 				  &ocfs2_inode_caching_ops);
1796 
1797 	inode_init_once(&oi->vfs_inode);
1798 }
1799 
1800 static int ocfs2_initialize_mem_caches(void)
1801 {
1802 	ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1803 				       sizeof(struct ocfs2_inode_info),
1804 				       0,
1805 				       (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1806 						SLAB_MEM_SPREAD),
1807 				       ocfs2_inode_init_once);
1808 	ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1809 					sizeof(struct ocfs2_dquot),
1810 					0,
1811 					(SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1812 						SLAB_MEM_SPREAD),
1813 					NULL);
1814 	ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1815 					sizeof(struct ocfs2_quota_chunk),
1816 					0,
1817 					(SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1818 					NULL);
1819 	if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1820 	    !ocfs2_qf_chunk_cachep) {
1821 		if (ocfs2_inode_cachep)
1822 			kmem_cache_destroy(ocfs2_inode_cachep);
1823 		if (ocfs2_dquot_cachep)
1824 			kmem_cache_destroy(ocfs2_dquot_cachep);
1825 		if (ocfs2_qf_chunk_cachep)
1826 			kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1827 		return -ENOMEM;
1828 	}
1829 
1830 	return 0;
1831 }
1832 
1833 static void ocfs2_free_mem_caches(void)
1834 {
1835 	if (ocfs2_inode_cachep)
1836 		kmem_cache_destroy(ocfs2_inode_cachep);
1837 	ocfs2_inode_cachep = NULL;
1838 
1839 	if (ocfs2_dquot_cachep)
1840 		kmem_cache_destroy(ocfs2_dquot_cachep);
1841 	ocfs2_dquot_cachep = NULL;
1842 
1843 	if (ocfs2_qf_chunk_cachep)
1844 		kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1845 	ocfs2_qf_chunk_cachep = NULL;
1846 }
1847 
1848 static int ocfs2_get_sector(struct super_block *sb,
1849 			    struct buffer_head **bh,
1850 			    int block,
1851 			    int sect_size)
1852 {
1853 	if (!sb_set_blocksize(sb, sect_size)) {
1854 		mlog(ML_ERROR, "unable to set blocksize\n");
1855 		return -EIO;
1856 	}
1857 
1858 	*bh = sb_getblk(sb, block);
1859 	if (!*bh) {
1860 		mlog_errno(-EIO);
1861 		return -EIO;
1862 	}
1863 	lock_buffer(*bh);
1864 	if (!buffer_dirty(*bh))
1865 		clear_buffer_uptodate(*bh);
1866 	unlock_buffer(*bh);
1867 	ll_rw_block(READ, 1, bh);
1868 	wait_on_buffer(*bh);
1869 	if (!buffer_uptodate(*bh)) {
1870 		mlog_errno(-EIO);
1871 		brelse(*bh);
1872 		*bh = NULL;
1873 		return -EIO;
1874 	}
1875 
1876 	return 0;
1877 }
1878 
1879 static int ocfs2_mount_volume(struct super_block *sb)
1880 {
1881 	int status = 0;
1882 	int unlock_super = 0;
1883 	struct ocfs2_super *osb = OCFS2_SB(sb);
1884 
1885 	mlog_entry_void();
1886 
1887 	if (ocfs2_is_hard_readonly(osb))
1888 		goto leave;
1889 
1890 	status = ocfs2_dlm_init(osb);
1891 	if (status < 0) {
1892 		mlog_errno(status);
1893 		goto leave;
1894 	}
1895 
1896 	status = ocfs2_super_lock(osb, 1);
1897 	if (status < 0) {
1898 		mlog_errno(status);
1899 		goto leave;
1900 	}
1901 	unlock_super = 1;
1902 
1903 	/* This will load up the node map and add ourselves to it. */
1904 	status = ocfs2_find_slot(osb);
1905 	if (status < 0) {
1906 		mlog_errno(status);
1907 		goto leave;
1908 	}
1909 
1910 	/* load all node-local system inodes */
1911 	status = ocfs2_init_local_system_inodes(osb);
1912 	if (status < 0) {
1913 		mlog_errno(status);
1914 		goto leave;
1915 	}
1916 
1917 	status = ocfs2_check_volume(osb);
1918 	if (status < 0) {
1919 		mlog_errno(status);
1920 		goto leave;
1921 	}
1922 
1923 	status = ocfs2_truncate_log_init(osb);
1924 	if (status < 0)
1925 		mlog_errno(status);
1926 
1927 leave:
1928 	if (unlock_super)
1929 		ocfs2_super_unlock(osb, 1);
1930 
1931 	mlog_exit(status);
1932 	return status;
1933 }
1934 
1935 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1936 {
1937 	int tmp, hangup_needed = 0;
1938 	struct ocfs2_super *osb = NULL;
1939 	char nodestr[8];
1940 
1941 	mlog_entry("(0x%p)\n", sb);
1942 
1943 	BUG_ON(!sb);
1944 	osb = OCFS2_SB(sb);
1945 	BUG_ON(!osb);
1946 
1947 	debugfs_remove(osb->osb_ctxt);
1948 
1949 	/*
1950 	 * Flush inode dropping work queue so that deletes are
1951 	 * performed while the filesystem is still working
1952 	 */
1953 	ocfs2_drop_all_dl_inodes(osb);
1954 
1955 	/* Orphan scan should be stopped as early as possible */
1956 	ocfs2_orphan_scan_stop(osb);
1957 
1958 	ocfs2_disable_quotas(osb);
1959 
1960 	ocfs2_shutdown_local_alloc(osb);
1961 
1962 	ocfs2_truncate_log_shutdown(osb);
1963 
1964 	/* This will disable recovery and flush any recovery work. */
1965 	ocfs2_recovery_exit(osb);
1966 
1967 	ocfs2_journal_shutdown(osb);
1968 
1969 	ocfs2_sync_blockdev(sb);
1970 
1971 	ocfs2_purge_refcount_trees(osb);
1972 
1973 	/* No cluster connection means we've failed during mount, so skip
1974 	 * all the steps which depended on that to complete. */
1975 	if (osb->cconn) {
1976 		tmp = ocfs2_super_lock(osb, 1);
1977 		if (tmp < 0) {
1978 			mlog_errno(tmp);
1979 			return;
1980 		}
1981 	}
1982 
1983 	if (osb->slot_num != OCFS2_INVALID_SLOT)
1984 		ocfs2_put_slot(osb);
1985 
1986 	if (osb->cconn)
1987 		ocfs2_super_unlock(osb, 1);
1988 
1989 	ocfs2_release_system_inodes(osb);
1990 
1991 	/*
1992 	 * If we're dismounting due to mount error, mount.ocfs2 will clean
1993 	 * up heartbeat.  If we're a local mount, there is no heartbeat.
1994 	 * If we failed before we got a uuid_str yet, we can't stop
1995 	 * heartbeat.  Otherwise, do it.
1996 	 */
1997 	if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str)
1998 		hangup_needed = 1;
1999 
2000 	if (osb->cconn)
2001 		ocfs2_dlm_shutdown(osb, hangup_needed);
2002 
2003 	ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
2004 	debugfs_remove(osb->osb_debug_root);
2005 
2006 	if (hangup_needed)
2007 		ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
2008 
2009 	atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
2010 
2011 	if (ocfs2_mount_local(osb))
2012 		snprintf(nodestr, sizeof(nodestr), "local");
2013 	else
2014 		snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
2015 
2016 	printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
2017 	       osb->dev_str, nodestr);
2018 
2019 	ocfs2_delete_osb(osb);
2020 	kfree(osb);
2021 	sb->s_dev = 0;
2022 	sb->s_fs_info = NULL;
2023 }
2024 
2025 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
2026 				unsigned uuid_bytes)
2027 {
2028 	int i, ret;
2029 	char *ptr;
2030 
2031 	BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
2032 
2033 	osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
2034 	if (osb->uuid_str == NULL)
2035 		return -ENOMEM;
2036 
2037 	for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
2038 		/* print with null */
2039 		ret = snprintf(ptr, 3, "%02X", uuid[i]);
2040 		if (ret != 2) /* drop super cleans up */
2041 			return -EINVAL;
2042 		/* then only advance past the last char */
2043 		ptr += 2;
2044 	}
2045 
2046 	return 0;
2047 }
2048 
2049 /* Make sure entire volume is addressable by our journal.  Requires
2050    osb_clusters_at_boot to be valid and for the journal to have been
2051    initialized by ocfs2_journal_init(). */
2052 static int ocfs2_journal_addressable(struct ocfs2_super *osb)
2053 {
2054 	int status = 0;
2055 	u64 max_block =
2056 		ocfs2_clusters_to_blocks(osb->sb,
2057 					 osb->osb_clusters_at_boot) - 1;
2058 
2059 	/* 32-bit block number is always OK. */
2060 	if (max_block <= (u32)~0ULL)
2061 		goto out;
2062 
2063 	/* Volume is "huge", so see if our journal is new enough to
2064 	   support it. */
2065 	if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2066 				       OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2067 	      jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2068 					       JBD2_FEATURE_INCOMPAT_64BIT))) {
2069 		mlog(ML_ERROR, "The journal cannot address the entire volume. "
2070 		     "Enable the 'block64' journal option with tunefs.ocfs2");
2071 		status = -EFBIG;
2072 		goto out;
2073 	}
2074 
2075  out:
2076 	return status;
2077 }
2078 
2079 static int ocfs2_initialize_super(struct super_block *sb,
2080 				  struct buffer_head *bh,
2081 				  int sector_size,
2082 				  struct ocfs2_blockcheck_stats *stats)
2083 {
2084 	int status;
2085 	int i, cbits, bbits;
2086 	struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2087 	struct inode *inode = NULL;
2088 	struct ocfs2_journal *journal;
2089 	__le32 uuid_net_key;
2090 	struct ocfs2_super *osb;
2091 	u64 total_blocks;
2092 
2093 	mlog_entry_void();
2094 
2095 	osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2096 	if (!osb) {
2097 		status = -ENOMEM;
2098 		mlog_errno(status);
2099 		goto bail;
2100 	}
2101 
2102 	sb->s_fs_info = osb;
2103 	sb->s_op = &ocfs2_sops;
2104 	sb->s_export_op = &ocfs2_export_ops;
2105 	sb->s_qcop = &ocfs2_quotactl_ops;
2106 	sb->dq_op = &ocfs2_quota_operations;
2107 	sb->s_xattr = ocfs2_xattr_handlers;
2108 	sb->s_time_gran = 1;
2109 	sb->s_flags |= MS_NOATIME;
2110 	/* this is needed to support O_LARGEFILE */
2111 	cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2112 	bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2113 	sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2114 
2115 	osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2116 
2117 	for (i = 0; i < 3; i++)
2118 		osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2119 	osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2120 
2121 	osb->sb = sb;
2122 	/* Save off for ocfs2_rw_direct */
2123 	osb->s_sectsize_bits = blksize_bits(sector_size);
2124 	BUG_ON(!osb->s_sectsize_bits);
2125 
2126 	spin_lock_init(&osb->dc_task_lock);
2127 	init_waitqueue_head(&osb->dc_event);
2128 	osb->dc_work_sequence = 0;
2129 	osb->dc_wake_sequence = 0;
2130 	INIT_LIST_HEAD(&osb->blocked_lock_list);
2131 	osb->blocked_lock_count = 0;
2132 	spin_lock_init(&osb->osb_lock);
2133 	spin_lock_init(&osb->osb_xattr_lock);
2134 	ocfs2_init_steal_slots(osb);
2135 
2136 	atomic_set(&osb->alloc_stats.moves, 0);
2137 	atomic_set(&osb->alloc_stats.local_data, 0);
2138 	atomic_set(&osb->alloc_stats.bitmap_data, 0);
2139 	atomic_set(&osb->alloc_stats.bg_allocs, 0);
2140 	atomic_set(&osb->alloc_stats.bg_extends, 0);
2141 
2142 	/* Copy the blockcheck stats from the superblock probe */
2143 	osb->osb_ecc_stats = *stats;
2144 
2145 	ocfs2_init_node_maps(osb);
2146 
2147 	snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2148 		 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2149 
2150 	osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2151 	if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2152 		mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2153 		     osb->max_slots);
2154 		status = -EINVAL;
2155 		goto bail;
2156 	}
2157 	mlog(0, "max_slots for this device: %u\n", osb->max_slots);
2158 
2159 	ocfs2_orphan_scan_init(osb);
2160 
2161 	status = ocfs2_recovery_init(osb);
2162 	if (status) {
2163 		mlog(ML_ERROR, "Unable to initialize recovery state\n");
2164 		mlog_errno(status);
2165 		goto bail;
2166 	}
2167 
2168 	init_waitqueue_head(&osb->checkpoint_event);
2169 	atomic_set(&osb->needs_checkpoint, 0);
2170 
2171 	osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2172 
2173 	osb->slot_num = OCFS2_INVALID_SLOT;
2174 
2175 	osb->s_xattr_inline_size = le16_to_cpu(
2176 					di->id2.i_super.s_xattr_inline_size);
2177 
2178 	osb->local_alloc_state = OCFS2_LA_UNUSED;
2179 	osb->local_alloc_bh = NULL;
2180 	INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2181 
2182 	init_waitqueue_head(&osb->osb_mount_event);
2183 
2184 	status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2185 	if (status) {
2186 		mlog_errno(status);
2187 		goto bail;
2188 	}
2189 
2190 	osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2191 	if (!osb->vol_label) {
2192 		mlog(ML_ERROR, "unable to alloc vol label\n");
2193 		status = -ENOMEM;
2194 		goto bail;
2195 	}
2196 
2197 	osb->slot_recovery_generations =
2198 		kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2199 			GFP_KERNEL);
2200 	if (!osb->slot_recovery_generations) {
2201 		status = -ENOMEM;
2202 		mlog_errno(status);
2203 		goto bail;
2204 	}
2205 
2206 	init_waitqueue_head(&osb->osb_wipe_event);
2207 	osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2208 					sizeof(*osb->osb_orphan_wipes),
2209 					GFP_KERNEL);
2210 	if (!osb->osb_orphan_wipes) {
2211 		status = -ENOMEM;
2212 		mlog_errno(status);
2213 		goto bail;
2214 	}
2215 
2216 	osb->osb_rf_lock_tree = RB_ROOT;
2217 
2218 	osb->s_feature_compat =
2219 		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2220 	osb->s_feature_ro_compat =
2221 		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2222 	osb->s_feature_incompat =
2223 		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2224 
2225 	if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2226 		mlog(ML_ERROR, "couldn't mount because of unsupported "
2227 		     "optional features (%x).\n", i);
2228 		status = -EINVAL;
2229 		goto bail;
2230 	}
2231 	if (!(osb->sb->s_flags & MS_RDONLY) &&
2232 	    (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2233 		mlog(ML_ERROR, "couldn't mount RDWR because of "
2234 		     "unsupported optional features (%x).\n", i);
2235 		status = -EINVAL;
2236 		goto bail;
2237 	}
2238 
2239 	if (ocfs2_clusterinfo_valid(osb)) {
2240 		osb->osb_stackflags =
2241 			OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2242 		memcpy(osb->osb_cluster_stack,
2243 		       OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2244 		       OCFS2_STACK_LABEL_LEN);
2245 		osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
2246 		if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2247 			mlog(ML_ERROR,
2248 			     "couldn't mount because of an invalid "
2249 			     "cluster stack label (%s) \n",
2250 			     osb->osb_cluster_stack);
2251 			status = -EINVAL;
2252 			goto bail;
2253 		}
2254 	} else {
2255 		/* The empty string is identical with classic tools that
2256 		 * don't know about s_cluster_info. */
2257 		osb->osb_cluster_stack[0] = '\0';
2258 	}
2259 
2260 	get_random_bytes(&osb->s_next_generation, sizeof(u32));
2261 
2262 	/* FIXME
2263 	 * This should be done in ocfs2_journal_init(), but unknown
2264 	 * ordering issues will cause the filesystem to crash.
2265 	 * If anyone wants to figure out what part of the code
2266 	 * refers to osb->journal before ocfs2_journal_init() is run,
2267 	 * be my guest.
2268 	 */
2269 	/* initialize our journal structure */
2270 
2271 	journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2272 	if (!journal) {
2273 		mlog(ML_ERROR, "unable to alloc journal\n");
2274 		status = -ENOMEM;
2275 		goto bail;
2276 	}
2277 	osb->journal = journal;
2278 	journal->j_osb = osb;
2279 
2280 	atomic_set(&journal->j_num_trans, 0);
2281 	init_rwsem(&journal->j_trans_barrier);
2282 	init_waitqueue_head(&journal->j_checkpointed);
2283 	spin_lock_init(&journal->j_lock);
2284 	journal->j_trans_id = (unsigned long) 1;
2285 	INIT_LIST_HEAD(&journal->j_la_cleanups);
2286 	INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2287 	journal->j_state = OCFS2_JOURNAL_FREE;
2288 
2289 	INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes);
2290 	osb->dentry_lock_list = NULL;
2291 
2292 	/* get some pseudo constants for clustersize bits */
2293 	osb->s_clustersize_bits =
2294 		le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2295 	osb->s_clustersize = 1 << osb->s_clustersize_bits;
2296 	mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
2297 
2298 	if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2299 	    osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2300 		mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2301 		     osb->s_clustersize);
2302 		status = -EINVAL;
2303 		goto bail;
2304 	}
2305 
2306 	total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2307 						le32_to_cpu(di->i_clusters));
2308 
2309 	status = generic_check_addressable(osb->sb->s_blocksize_bits,
2310 					   total_blocks);
2311 	if (status) {
2312 		mlog(ML_ERROR, "Volume too large "
2313 		     "to mount safely on this system");
2314 		status = -EFBIG;
2315 		goto bail;
2316 	}
2317 
2318 	if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2319 				 sizeof(di->id2.i_super.s_uuid))) {
2320 		mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2321 		status = -ENOMEM;
2322 		goto bail;
2323 	}
2324 
2325 	memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
2326 
2327 	strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2328 	osb->vol_label[63] = '\0';
2329 	osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2330 	osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2331 	osb->first_cluster_group_blkno =
2332 		le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2333 	osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2334 	osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2335 	mlog(0, "vol_label: %s\n", osb->vol_label);
2336 	mlog(0, "uuid: %s\n", osb->uuid_str);
2337 	mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
2338 	     (unsigned long long)osb->root_blkno,
2339 	     (unsigned long long)osb->system_dir_blkno);
2340 
2341 	osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2342 	if (!osb->osb_dlm_debug) {
2343 		status = -ENOMEM;
2344 		mlog_errno(status);
2345 		goto bail;
2346 	}
2347 
2348 	atomic_set(&osb->vol_state, VOLUME_INIT);
2349 
2350 	/* load root, system_dir, and all global system inodes */
2351 	status = ocfs2_init_global_system_inodes(osb);
2352 	if (status < 0) {
2353 		mlog_errno(status);
2354 		goto bail;
2355 	}
2356 
2357 	/*
2358 	 * global bitmap
2359 	 */
2360 	inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2361 					    OCFS2_INVALID_SLOT);
2362 	if (!inode) {
2363 		status = -EINVAL;
2364 		mlog_errno(status);
2365 		goto bail;
2366 	}
2367 
2368 	osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2369 	osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2370 	iput(inode);
2371 
2372 	osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2373 				 osb->s_feature_incompat) * 8;
2374 
2375 	status = ocfs2_init_slot_info(osb);
2376 	if (status < 0) {
2377 		mlog_errno(status);
2378 		goto bail;
2379 	}
2380 
2381 bail:
2382 	mlog_exit(status);
2383 	return status;
2384 }
2385 
2386 /*
2387  * will return: -EAGAIN if it is ok to keep searching for superblocks
2388  *              -EINVAL if there is a bad superblock
2389  *              0 on success
2390  */
2391 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2392 			       struct buffer_head *bh,
2393 			       u32 blksz,
2394 			       struct ocfs2_blockcheck_stats *stats)
2395 {
2396 	int status = -EAGAIN;
2397 
2398 	mlog_entry_void();
2399 
2400 	if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2401 		   strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2402 		/* We have to do a raw check of the feature here */
2403 		if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2404 		    OCFS2_FEATURE_INCOMPAT_META_ECC) {
2405 			status = ocfs2_block_check_validate(bh->b_data,
2406 							    bh->b_size,
2407 							    &di->i_check,
2408 							    stats);
2409 			if (status)
2410 				goto out;
2411 		}
2412 		status = -EINVAL;
2413 		if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2414 			mlog(ML_ERROR, "found superblock with incorrect block "
2415 			     "size: found %u, should be %u\n",
2416 			     1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2417 			       blksz);
2418 		} else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2419 			   OCFS2_MAJOR_REV_LEVEL ||
2420 			   le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2421 			   OCFS2_MINOR_REV_LEVEL) {
2422 			mlog(ML_ERROR, "found superblock with bad version: "
2423 			     "found %u.%u, should be %u.%u\n",
2424 			     le16_to_cpu(di->id2.i_super.s_major_rev_level),
2425 			     le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2426 			     OCFS2_MAJOR_REV_LEVEL,
2427 			     OCFS2_MINOR_REV_LEVEL);
2428 		} else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2429 			mlog(ML_ERROR, "bad block number on superblock: "
2430 			     "found %llu, should be %llu\n",
2431 			     (unsigned long long)le64_to_cpu(di->i_blkno),
2432 			     (unsigned long long)bh->b_blocknr);
2433 		} else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2434 			    le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2435 			mlog(ML_ERROR, "bad cluster size found: %u\n",
2436 			     1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2437 		} else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2438 			mlog(ML_ERROR, "bad root_blkno: 0\n");
2439 		} else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2440 			mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2441 		} else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2442 			mlog(ML_ERROR,
2443 			     "Superblock slots found greater than file system "
2444 			     "maximum: found %u, max %u\n",
2445 			     le16_to_cpu(di->id2.i_super.s_max_slots),
2446 			     OCFS2_MAX_SLOTS);
2447 		} else {
2448 			/* found it! */
2449 			status = 0;
2450 		}
2451 	}
2452 
2453 out:
2454 	mlog_exit(status);
2455 	return status;
2456 }
2457 
2458 static int ocfs2_check_volume(struct ocfs2_super *osb)
2459 {
2460 	int status;
2461 	int dirty;
2462 	int local;
2463 	struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2464 						  * recover
2465 						  * ourselves. */
2466 
2467 	mlog_entry_void();
2468 
2469 	/* Init our journal object. */
2470 	status = ocfs2_journal_init(osb->journal, &dirty);
2471 	if (status < 0) {
2472 		mlog(ML_ERROR, "Could not initialize journal!\n");
2473 		goto finally;
2474 	}
2475 
2476 	/* Now that journal has been initialized, check to make sure
2477 	   entire volume is addressable. */
2478 	status = ocfs2_journal_addressable(osb);
2479 	if (status)
2480 		goto finally;
2481 
2482 	/* If the journal was unmounted cleanly then we don't want to
2483 	 * recover anything. Otherwise, journal_load will do that
2484 	 * dirty work for us :) */
2485 	if (!dirty) {
2486 		status = ocfs2_journal_wipe(osb->journal, 0);
2487 		if (status < 0) {
2488 			mlog_errno(status);
2489 			goto finally;
2490 		}
2491 	} else {
2492 		mlog(ML_NOTICE, "File system was not unmounted cleanly, "
2493 		     "recovering volume.\n");
2494 	}
2495 
2496 	local = ocfs2_mount_local(osb);
2497 
2498 	/* will play back anything left in the journal. */
2499 	status = ocfs2_journal_load(osb->journal, local, dirty);
2500 	if (status < 0) {
2501 		mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2502 		goto finally;
2503 	}
2504 
2505 	if (dirty) {
2506 		/* recover my local alloc if we didn't unmount cleanly. */
2507 		status = ocfs2_begin_local_alloc_recovery(osb,
2508 							  osb->slot_num,
2509 							  &local_alloc);
2510 		if (status < 0) {
2511 			mlog_errno(status);
2512 			goto finally;
2513 		}
2514 		/* we complete the recovery process after we've marked
2515 		 * ourselves as mounted. */
2516 	}
2517 
2518 	mlog(0, "Journal loaded.\n");
2519 
2520 	status = ocfs2_load_local_alloc(osb);
2521 	if (status < 0) {
2522 		mlog_errno(status);
2523 		goto finally;
2524 	}
2525 
2526 	if (dirty) {
2527 		/* Recovery will be completed after we've mounted the
2528 		 * rest of the volume. */
2529 		osb->dirty = 1;
2530 		osb->local_alloc_copy = local_alloc;
2531 		local_alloc = NULL;
2532 	}
2533 
2534 	/* go through each journal, trylock it and if you get the
2535 	 * lock, and it's marked as dirty, set the bit in the recover
2536 	 * map and launch a recovery thread for it. */
2537 	status = ocfs2_mark_dead_nodes(osb);
2538 	if (status < 0) {
2539 		mlog_errno(status);
2540 		goto finally;
2541 	}
2542 
2543 	status = ocfs2_compute_replay_slots(osb);
2544 	if (status < 0)
2545 		mlog_errno(status);
2546 
2547 finally:
2548 	if (local_alloc)
2549 		kfree(local_alloc);
2550 
2551 	mlog_exit(status);
2552 	return status;
2553 }
2554 
2555 /*
2556  * The routine gets called from dismount or close whenever a dismount on
2557  * volume is requested and the osb open count becomes 1.
2558  * It will remove the osb from the global list and also free up all the
2559  * initialized resources and fileobject.
2560  */
2561 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2562 {
2563 	mlog_entry_void();
2564 
2565 	/* This function assumes that the caller has the main osb resource */
2566 
2567 	ocfs2_free_slot_info(osb);
2568 
2569 	kfree(osb->osb_orphan_wipes);
2570 	kfree(osb->slot_recovery_generations);
2571 	/* FIXME
2572 	 * This belongs in journal shutdown, but because we have to
2573 	 * allocate osb->journal at the start of ocfs2_initialize_osb(),
2574 	 * we free it here.
2575 	 */
2576 	kfree(osb->journal);
2577 	if (osb->local_alloc_copy)
2578 		kfree(osb->local_alloc_copy);
2579 	kfree(osb->uuid_str);
2580 	ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2581 	memset(osb, 0, sizeof(struct ocfs2_super));
2582 
2583 	mlog_exit_void();
2584 }
2585 
2586 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2587  * panic(). We do not support continue-on-error operation. */
2588 static void ocfs2_handle_error(struct super_block *sb)
2589 {
2590 	struct ocfs2_super *osb = OCFS2_SB(sb);
2591 
2592 	if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2593 		panic("OCFS2: (device %s): panic forced after error\n",
2594 		      sb->s_id);
2595 
2596 	ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2597 
2598 	if (sb->s_flags & MS_RDONLY &&
2599 	    (ocfs2_is_soft_readonly(osb) ||
2600 	     ocfs2_is_hard_readonly(osb)))
2601 		return;
2602 
2603 	printk(KERN_CRIT "File system is now read-only due to the potential "
2604 	       "of on-disk corruption. Please run fsck.ocfs2 once the file "
2605 	       "system is unmounted.\n");
2606 	sb->s_flags |= MS_RDONLY;
2607 	ocfs2_set_ro_flag(osb, 0);
2608 }
2609 
2610 static char error_buf[1024];
2611 
2612 void __ocfs2_error(struct super_block *sb,
2613 		   const char *function,
2614 		   const char *fmt, ...)
2615 {
2616 	va_list args;
2617 
2618 	va_start(args, fmt);
2619 	vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2620 	va_end(args);
2621 
2622 	/* Not using mlog here because we want to show the actual
2623 	 * function the error came from. */
2624 	printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2625 	       sb->s_id, function, error_buf);
2626 
2627 	ocfs2_handle_error(sb);
2628 }
2629 
2630 /* Handle critical errors. This is intentionally more drastic than
2631  * ocfs2_handle_error, so we only use for things like journal errors,
2632  * etc. */
2633 void __ocfs2_abort(struct super_block* sb,
2634 		   const char *function,
2635 		   const char *fmt, ...)
2636 {
2637 	va_list args;
2638 
2639 	va_start(args, fmt);
2640 	vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2641 	va_end(args);
2642 
2643 	printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2644 	       sb->s_id, function, error_buf);
2645 
2646 	/* We don't have the cluster support yet to go straight to
2647 	 * hard readonly in here. Until then, we want to keep
2648 	 * ocfs2_abort() so that we can at least mark critical
2649 	 * errors.
2650 	 *
2651 	 * TODO: This should abort the journal and alert other nodes
2652 	 * that our slot needs recovery. */
2653 
2654 	/* Force a panic(). This stinks, but it's better than letting
2655 	 * things continue without having a proper hard readonly
2656 	 * here. */
2657 	if (!ocfs2_mount_local(OCFS2_SB(sb)))
2658 		OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2659 	ocfs2_handle_error(sb);
2660 }
2661 
2662 /*
2663  * Void signal blockers, because in-kernel sigprocmask() only fails
2664  * when SIG_* is wrong.
2665  */
2666 void ocfs2_block_signals(sigset_t *oldset)
2667 {
2668 	int rc;
2669 	sigset_t blocked;
2670 
2671 	sigfillset(&blocked);
2672 	rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2673 	BUG_ON(rc);
2674 }
2675 
2676 void ocfs2_unblock_signals(sigset_t *oldset)
2677 {
2678 	int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2679 	BUG_ON(rc);
2680 }
2681 
2682 module_init(ocfs2_init);
2683 module_exit(ocfs2_exit);
2684