xref: /linux/fs/ocfs2/super.c (revision 092e0e7e520a1fca03e13c9f2d157432a8657ff2)
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 	if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
634 	    !ocfs2_check_set_options(sb, &parsed_options)) {
635 		ret = -EINVAL;
636 		goto out;
637 	}
638 
639 	tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
640 		OCFS2_MOUNT_HB_NONE;
641 	if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
642 		ret = -EINVAL;
643 		mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
644 		goto out;
645 	}
646 
647 	if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
648 	    (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
649 		ret = -EINVAL;
650 		mlog(ML_ERROR, "Cannot change data mode on remount\n");
651 		goto out;
652 	}
653 
654 	/* Probably don't want this on remount; it might
655 	 * mess with other nodes */
656 	if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
657 	    (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
658 		ret = -EINVAL;
659 		mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
660 		goto out;
661 	}
662 
663 	/* We're going to/from readonly mode. */
664 	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
665 		/* Disable quota accounting before remounting RO */
666 		if (*flags & MS_RDONLY) {
667 			ret = ocfs2_susp_quotas(osb, 0);
668 			if (ret < 0)
669 				goto out;
670 		}
671 		/* Lock here so the check of HARD_RO and the potential
672 		 * setting of SOFT_RO is atomic. */
673 		spin_lock(&osb->osb_lock);
674 		if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
675 			mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
676 			ret = -EROFS;
677 			goto unlock_osb;
678 		}
679 
680 		if (*flags & MS_RDONLY) {
681 			mlog(0, "Going to ro mode.\n");
682 			sb->s_flags |= MS_RDONLY;
683 			osb->osb_flags |= OCFS2_OSB_SOFT_RO;
684 		} else {
685 			mlog(0, "Making ro filesystem writeable.\n");
686 
687 			if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
688 				mlog(ML_ERROR, "Cannot remount RDWR "
689 				     "filesystem due to previous errors.\n");
690 				ret = -EROFS;
691 				goto unlock_osb;
692 			}
693 			incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
694 			if (incompat_features) {
695 				mlog(ML_ERROR, "Cannot remount RDWR because "
696 				     "of unsupported optional features "
697 				     "(%x).\n", incompat_features);
698 				ret = -EINVAL;
699 				goto unlock_osb;
700 			}
701 			sb->s_flags &= ~MS_RDONLY;
702 			osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
703 		}
704 unlock_osb:
705 		spin_unlock(&osb->osb_lock);
706 		/* Enable quota accounting after remounting RW */
707 		if (!ret && !(*flags & MS_RDONLY)) {
708 			if (sb_any_quota_suspended(sb))
709 				ret = ocfs2_susp_quotas(osb, 1);
710 			else
711 				ret = ocfs2_enable_quotas(osb);
712 			if (ret < 0) {
713 				/* Return back changes... */
714 				spin_lock(&osb->osb_lock);
715 				sb->s_flags |= MS_RDONLY;
716 				osb->osb_flags |= OCFS2_OSB_SOFT_RO;
717 				spin_unlock(&osb->osb_lock);
718 				goto out;
719 			}
720 		}
721 	}
722 
723 	if (!ret) {
724 		/* Only save off the new mount options in case of a successful
725 		 * remount. */
726 		osb->s_mount_opt = parsed_options.mount_opt;
727 		osb->s_atime_quantum = parsed_options.atime_quantum;
728 		osb->preferred_slot = parsed_options.slot;
729 		if (parsed_options.commit_interval)
730 			osb->osb_commit_interval = parsed_options.commit_interval;
731 
732 		if (!ocfs2_is_hard_readonly(osb))
733 			ocfs2_set_journal_params(osb);
734 
735 		sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
736 			((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
737 							MS_POSIXACL : 0);
738 	}
739 out:
740 	return ret;
741 }
742 
743 static int ocfs2_sb_probe(struct super_block *sb,
744 			  struct buffer_head **bh,
745 			  int *sector_size,
746 			  struct ocfs2_blockcheck_stats *stats)
747 {
748 	int status, tmpstat;
749 	struct ocfs1_vol_disk_hdr *hdr;
750 	struct ocfs2_dinode *di;
751 	int blksize;
752 
753 	*bh = NULL;
754 
755 	/* may be > 512 */
756 	*sector_size = bdev_logical_block_size(sb->s_bdev);
757 	if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
758 		mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
759 		     *sector_size, OCFS2_MAX_BLOCKSIZE);
760 		status = -EINVAL;
761 		goto bail;
762 	}
763 
764 	/* Can this really happen? */
765 	if (*sector_size < OCFS2_MIN_BLOCKSIZE)
766 		*sector_size = OCFS2_MIN_BLOCKSIZE;
767 
768 	/* check block zero for old format */
769 	status = ocfs2_get_sector(sb, bh, 0, *sector_size);
770 	if (status < 0) {
771 		mlog_errno(status);
772 		goto bail;
773 	}
774 	hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
775 	if (hdr->major_version == OCFS1_MAJOR_VERSION) {
776 		mlog(ML_ERROR, "incompatible version: %u.%u\n",
777 		     hdr->major_version, hdr->minor_version);
778 		status = -EINVAL;
779 	}
780 	if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
781 		   strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
782 		mlog(ML_ERROR, "incompatible volume signature: %8s\n",
783 		     hdr->signature);
784 		status = -EINVAL;
785 	}
786 	brelse(*bh);
787 	*bh = NULL;
788 	if (status < 0) {
789 		mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
790 		     "upgraded before mounting with ocfs v2\n");
791 		goto bail;
792 	}
793 
794 	/*
795 	 * Now check at magic offset for 512, 1024, 2048, 4096
796 	 * blocksizes.  4096 is the maximum blocksize because it is
797 	 * the minimum clustersize.
798 	 */
799 	status = -EINVAL;
800 	for (blksize = *sector_size;
801 	     blksize <= OCFS2_MAX_BLOCKSIZE;
802 	     blksize <<= 1) {
803 		tmpstat = ocfs2_get_sector(sb, bh,
804 					   OCFS2_SUPER_BLOCK_BLKNO,
805 					   blksize);
806 		if (tmpstat < 0) {
807 			status = tmpstat;
808 			mlog_errno(status);
809 			break;
810 		}
811 		di = (struct ocfs2_dinode *) (*bh)->b_data;
812 		memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
813 		spin_lock_init(&stats->b_lock);
814 		tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
815 		if (tmpstat < 0) {
816 			brelse(*bh);
817 			*bh = NULL;
818 		}
819 		if (tmpstat != -EAGAIN) {
820 			status = tmpstat;
821 			break;
822 		}
823 	}
824 
825 bail:
826 	return status;
827 }
828 
829 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
830 {
831 	u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
832 
833 	if (osb->s_mount_opt & hb_enabled) {
834 		if (ocfs2_mount_local(osb)) {
835 			mlog(ML_ERROR, "Cannot heartbeat on a locally "
836 			     "mounted device.\n");
837 			return -EINVAL;
838 		}
839 		if (ocfs2_userspace_stack(osb)) {
840 			mlog(ML_ERROR, "Userspace stack expected, but "
841 			     "o2cb heartbeat arguments passed to mount\n");
842 			return -EINVAL;
843 		}
844 		if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
845 		     !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
846 		    ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
847 		     ocfs2_cluster_o2cb_global_heartbeat(osb))) {
848 			mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
849 			return -EINVAL;
850 		}
851 	}
852 
853 	if (!(osb->s_mount_opt & hb_enabled)) {
854 		if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
855 		    !ocfs2_userspace_stack(osb)) {
856 			mlog(ML_ERROR, "Heartbeat has to be started to mount "
857 			     "a read-write clustered device.\n");
858 			return -EINVAL;
859 		}
860 	}
861 
862 	return 0;
863 }
864 
865 /*
866  * If we're using a userspace stack, mount should have passed
867  * a name that matches the disk.  If not, mount should not
868  * have passed a stack.
869  */
870 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
871 					struct mount_options *mopt)
872 {
873 	if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
874 		mlog(ML_ERROR,
875 		     "cluster stack passed to mount, but this filesystem "
876 		     "does not support it\n");
877 		return -EINVAL;
878 	}
879 
880 	if (ocfs2_userspace_stack(osb) &&
881 	    strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
882 		    OCFS2_STACK_LABEL_LEN)) {
883 		mlog(ML_ERROR,
884 		     "cluster stack passed to mount (\"%s\") does not "
885 		     "match the filesystem (\"%s\")\n",
886 		     mopt->cluster_stack,
887 		     osb->osb_cluster_stack);
888 		return -EINVAL;
889 	}
890 
891 	return 0;
892 }
893 
894 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
895 {
896 	int type;
897 	struct super_block *sb = osb->sb;
898 	unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
899 					     OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
900 	int status = 0;
901 
902 	for (type = 0; type < MAXQUOTAS; type++) {
903 		if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
904 			continue;
905 		if (unsuspend)
906 			status = dquot_resume(sb, type);
907 		else {
908 			struct ocfs2_mem_dqinfo *oinfo;
909 
910 			/* Cancel periodic syncing before suspending */
911 			oinfo = sb_dqinfo(sb, type)->dqi_priv;
912 			cancel_delayed_work_sync(&oinfo->dqi_sync_work);
913 			status = dquot_suspend(sb, type);
914 		}
915 		if (status < 0)
916 			break;
917 	}
918 	if (status < 0)
919 		mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
920 		     "remount (error = %d).\n", status);
921 	return status;
922 }
923 
924 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
925 {
926 	struct inode *inode[MAXQUOTAS] = { NULL, NULL };
927 	struct super_block *sb = osb->sb;
928 	unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
929 					     OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
930 	unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
931 					LOCAL_GROUP_QUOTA_SYSTEM_INODE };
932 	int status;
933 	int type;
934 
935 	sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
936 	for (type = 0; type < MAXQUOTAS; type++) {
937 		if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
938 			continue;
939 		inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
940 							osb->slot_num);
941 		if (!inode[type]) {
942 			status = -ENOENT;
943 			goto out_quota_off;
944 		}
945 		status = dquot_enable(inode[type], type, QFMT_OCFS2,
946 				      DQUOT_USAGE_ENABLED);
947 		if (status < 0)
948 			goto out_quota_off;
949 	}
950 
951 	for (type = 0; type < MAXQUOTAS; type++)
952 		iput(inode[type]);
953 	return 0;
954 out_quota_off:
955 	ocfs2_disable_quotas(osb);
956 	for (type = 0; type < MAXQUOTAS; type++)
957 		iput(inode[type]);
958 	mlog_errno(status);
959 	return status;
960 }
961 
962 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
963 {
964 	int type;
965 	struct inode *inode;
966 	struct super_block *sb = osb->sb;
967 	struct ocfs2_mem_dqinfo *oinfo;
968 
969 	/* We mostly ignore errors in this function because there's not much
970 	 * we can do when we see them */
971 	for (type = 0; type < MAXQUOTAS; type++) {
972 		if (!sb_has_quota_loaded(sb, type))
973 			continue;
974 		/* Cancel periodic syncing before we grab dqonoff_mutex */
975 		oinfo = sb_dqinfo(sb, type)->dqi_priv;
976 		cancel_delayed_work_sync(&oinfo->dqi_sync_work);
977 		inode = igrab(sb->s_dquot.files[type]);
978 		/* Turn off quotas. This will remove all dquot structures from
979 		 * memory and so they will be automatically synced to global
980 		 * quota files */
981 		dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
982 					DQUOT_LIMITS_ENABLED);
983 		if (!inode)
984 			continue;
985 		iput(inode);
986 	}
987 }
988 
989 /* Handle quota on quotactl */
990 static int ocfs2_quota_on(struct super_block *sb, int type, int format_id,
991 			  char *path)
992 {
993 	unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
994 					     OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
995 
996 	if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
997 		return -EINVAL;
998 
999 	return dquot_enable(sb_dqopt(sb)->files[type], type,
1000 			    format_id, DQUOT_LIMITS_ENABLED);
1001 }
1002 
1003 /* Handle quota off quotactl */
1004 static int ocfs2_quota_off(struct super_block *sb, int type)
1005 {
1006 	return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
1007 }
1008 
1009 static const struct quotactl_ops ocfs2_quotactl_ops = {
1010 	.quota_on	= ocfs2_quota_on,
1011 	.quota_off	= ocfs2_quota_off,
1012 	.quota_sync	= dquot_quota_sync,
1013 	.get_info	= dquot_get_dqinfo,
1014 	.set_info	= dquot_set_dqinfo,
1015 	.get_dqblk	= dquot_get_dqblk,
1016 	.set_dqblk	= dquot_set_dqblk,
1017 };
1018 
1019 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
1020 {
1021 	struct dentry *root;
1022 	int status, sector_size;
1023 	struct mount_options parsed_options;
1024 	struct inode *inode = NULL;
1025 	struct ocfs2_super *osb = NULL;
1026 	struct buffer_head *bh = NULL;
1027 	char nodestr[8];
1028 	struct ocfs2_blockcheck_stats stats;
1029 
1030 	mlog_entry("%p, %p, %i", sb, data, silent);
1031 
1032 	if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
1033 		status = -EINVAL;
1034 		goto read_super_error;
1035 	}
1036 
1037 	/* probe for superblock */
1038 	status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
1039 	if (status < 0) {
1040 		mlog(ML_ERROR, "superblock probe failed!\n");
1041 		goto read_super_error;
1042 	}
1043 
1044 	status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1045 	osb = OCFS2_SB(sb);
1046 	if (status < 0) {
1047 		mlog_errno(status);
1048 		goto read_super_error;
1049 	}
1050 	brelse(bh);
1051 	bh = NULL;
1052 
1053 	if (!ocfs2_check_set_options(sb, &parsed_options)) {
1054 		status = -EINVAL;
1055 		goto read_super_error;
1056 	}
1057 	osb->s_mount_opt = parsed_options.mount_opt;
1058 	osb->s_atime_quantum = parsed_options.atime_quantum;
1059 	osb->preferred_slot = parsed_options.slot;
1060 	osb->osb_commit_interval = parsed_options.commit_interval;
1061 
1062 	ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1063 	osb->osb_resv_level = parsed_options.resv_level;
1064 	osb->osb_dir_resv_level = parsed_options.resv_level;
1065 	if (parsed_options.dir_resv_level == -1)
1066 		osb->osb_dir_resv_level = parsed_options.resv_level;
1067 	else
1068 		osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1069 
1070 	status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1071 	if (status)
1072 		goto read_super_error;
1073 
1074 	sb->s_magic = OCFS2_SUPER_MAGIC;
1075 
1076 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1077 		((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1078 
1079 	/* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1080 	 * heartbeat=none */
1081 	if (bdev_read_only(sb->s_bdev)) {
1082 		if (!(sb->s_flags & MS_RDONLY)) {
1083 			status = -EACCES;
1084 			mlog(ML_ERROR, "Readonly device detected but readonly "
1085 			     "mount was not specified.\n");
1086 			goto read_super_error;
1087 		}
1088 
1089 		/* You should not be able to start a local heartbeat
1090 		 * on a readonly device. */
1091 		if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1092 			status = -EROFS;
1093 			mlog(ML_ERROR, "Local heartbeat specified on readonly "
1094 			     "device.\n");
1095 			goto read_super_error;
1096 		}
1097 
1098 		status = ocfs2_check_journals_nolocks(osb);
1099 		if (status < 0) {
1100 			if (status == -EROFS)
1101 				mlog(ML_ERROR, "Recovery required on readonly "
1102 				     "file system, but write access is "
1103 				     "unavailable.\n");
1104 			else
1105 				mlog_errno(status);
1106 			goto read_super_error;
1107 		}
1108 
1109 		ocfs2_set_ro_flag(osb, 1);
1110 
1111 		printk(KERN_NOTICE "Readonly device detected. No cluster "
1112 		       "services will be utilized for this mount. Recovery "
1113 		       "will be skipped.\n");
1114 	}
1115 
1116 	if (!ocfs2_is_hard_readonly(osb)) {
1117 		if (sb->s_flags & MS_RDONLY)
1118 			ocfs2_set_ro_flag(osb, 0);
1119 	}
1120 
1121 	status = ocfs2_verify_heartbeat(osb);
1122 	if (status < 0) {
1123 		mlog_errno(status);
1124 		goto read_super_error;
1125 	}
1126 
1127 	osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1128 						 ocfs2_debugfs_root);
1129 	if (!osb->osb_debug_root) {
1130 		status = -EINVAL;
1131 		mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1132 		goto read_super_error;
1133 	}
1134 
1135 	osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1136 					    osb->osb_debug_root,
1137 					    osb,
1138 					    &ocfs2_osb_debug_fops);
1139 	if (!osb->osb_ctxt) {
1140 		status = -EINVAL;
1141 		mlog_errno(status);
1142 		goto read_super_error;
1143 	}
1144 
1145 	if (ocfs2_meta_ecc(osb)) {
1146 		status = ocfs2_blockcheck_stats_debugfs_install(
1147 						&osb->osb_ecc_stats,
1148 						osb->osb_debug_root);
1149 		if (status) {
1150 			mlog(ML_ERROR,
1151 			     "Unable to create blockcheck statistics "
1152 			     "files\n");
1153 			goto read_super_error;
1154 		}
1155 	}
1156 
1157 	status = ocfs2_mount_volume(sb);
1158 	if (osb->root_inode)
1159 		inode = igrab(osb->root_inode);
1160 
1161 	if (status < 0)
1162 		goto read_super_error;
1163 
1164 	if (!inode) {
1165 		status = -EIO;
1166 		mlog_errno(status);
1167 		goto read_super_error;
1168 	}
1169 
1170 	root = d_alloc_root(inode);
1171 	if (!root) {
1172 		status = -ENOMEM;
1173 		mlog_errno(status);
1174 		goto read_super_error;
1175 	}
1176 
1177 	sb->s_root = root;
1178 
1179 	ocfs2_complete_mount_recovery(osb);
1180 
1181 	if (ocfs2_mount_local(osb))
1182 		snprintf(nodestr, sizeof(nodestr), "local");
1183 	else
1184 		snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1185 
1186 	printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1187 	       "with %s data mode.\n",
1188 	       osb->dev_str, nodestr, osb->slot_num,
1189 	       osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1190 	       "ordered");
1191 
1192 	atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1193 	wake_up(&osb->osb_mount_event);
1194 
1195 	/* Now we can initialize quotas because we can afford to wait
1196 	 * for cluster locks recovery now. That also means that truncation
1197 	 * log recovery can happen but that waits for proper quota setup */
1198 	if (!(sb->s_flags & MS_RDONLY)) {
1199 		status = ocfs2_enable_quotas(osb);
1200 		if (status < 0) {
1201 			/* We have to err-out specially here because
1202 			 * s_root is already set */
1203 			mlog_errno(status);
1204 			atomic_set(&osb->vol_state, VOLUME_DISABLED);
1205 			wake_up(&osb->osb_mount_event);
1206 			mlog_exit(status);
1207 			return status;
1208 		}
1209 	}
1210 
1211 	ocfs2_complete_quota_recovery(osb);
1212 
1213 	/* Now we wake up again for processes waiting for quotas */
1214 	atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1215 	wake_up(&osb->osb_mount_event);
1216 
1217 	/* Start this when the mount is almost sure of being successful */
1218 	ocfs2_orphan_scan_start(osb);
1219 
1220 	mlog_exit(status);
1221 	return status;
1222 
1223 read_super_error:
1224 	brelse(bh);
1225 
1226 	if (inode)
1227 		iput(inode);
1228 
1229 	if (osb) {
1230 		atomic_set(&osb->vol_state, VOLUME_DISABLED);
1231 		wake_up(&osb->osb_mount_event);
1232 		ocfs2_dismount_volume(sb, 1);
1233 	}
1234 
1235 	mlog_exit(status);
1236 	return status;
1237 }
1238 
1239 static int ocfs2_get_sb(struct file_system_type *fs_type,
1240 			int flags,
1241 			const char *dev_name,
1242 			void *data,
1243 			struct vfsmount *mnt)
1244 {
1245 	return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super,
1246 			   mnt);
1247 }
1248 
1249 static void ocfs2_kill_sb(struct super_block *sb)
1250 {
1251 	struct ocfs2_super *osb = OCFS2_SB(sb);
1252 
1253 	/* Failed mount? */
1254 	if (!osb || atomic_read(&osb->vol_state) == VOLUME_DISABLED)
1255 		goto out;
1256 
1257 	/* Prevent further queueing of inode drop events */
1258 	spin_lock(&dentry_list_lock);
1259 	ocfs2_set_osb_flag(osb, OCFS2_OSB_DROP_DENTRY_LOCK_IMMED);
1260 	spin_unlock(&dentry_list_lock);
1261 	/* Wait for work to finish and/or remove it */
1262 	cancel_work_sync(&osb->dentry_lock_work);
1263 out:
1264 	kill_block_super(sb);
1265 }
1266 
1267 static struct file_system_type ocfs2_fs_type = {
1268 	.owner          = THIS_MODULE,
1269 	.name           = "ocfs2",
1270 	.get_sb         = ocfs2_get_sb, /* is this called when we mount
1271 					* the fs? */
1272 	.kill_sb        = ocfs2_kill_sb,
1273 
1274 	.fs_flags       = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1275 	.next           = NULL
1276 };
1277 
1278 static int ocfs2_check_set_options(struct super_block *sb,
1279 				   struct mount_options *options)
1280 {
1281 	if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1282 	    !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1283 					 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1284 		mlog(ML_ERROR, "User quotas were requested, but this "
1285 		     "filesystem does not have the feature enabled.\n");
1286 		return 0;
1287 	}
1288 	if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1289 	    !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1290 					 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1291 		mlog(ML_ERROR, "Group quotas were requested, but this "
1292 		     "filesystem does not have the feature enabled.\n");
1293 		return 0;
1294 	}
1295 	if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1296 	    !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1297 		mlog(ML_ERROR, "ACL support requested but extended attributes "
1298 		     "feature is not enabled\n");
1299 		return 0;
1300 	}
1301 	/* No ACL setting specified? Use XATTR feature... */
1302 	if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1303 				    OCFS2_MOUNT_NO_POSIX_ACL))) {
1304 		if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1305 			options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1306 		else
1307 			options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1308 	}
1309 	return 1;
1310 }
1311 
1312 static int ocfs2_parse_options(struct super_block *sb,
1313 			       char *options,
1314 			       struct mount_options *mopt,
1315 			       int is_remount)
1316 {
1317 	int status;
1318 	char *p;
1319 	u32 tmp;
1320 
1321 	mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
1322 		   options ? options : "(none)");
1323 
1324 	mopt->commit_interval = 0;
1325 	mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1326 	mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1327 	mopt->slot = OCFS2_INVALID_SLOT;
1328 	mopt->localalloc_opt = -1;
1329 	mopt->cluster_stack[0] = '\0';
1330 	mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1331 	mopt->dir_resv_level = -1;
1332 
1333 	if (!options) {
1334 		status = 1;
1335 		goto bail;
1336 	}
1337 
1338 	while ((p = strsep(&options, ",")) != NULL) {
1339 		int token, option;
1340 		substring_t args[MAX_OPT_ARGS];
1341 
1342 		if (!*p)
1343 			continue;
1344 
1345 		token = match_token(p, tokens, args);
1346 		switch (token) {
1347 		case Opt_hb_local:
1348 			mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1349 			break;
1350 		case Opt_hb_none:
1351 			mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1352 			break;
1353 		case Opt_hb_global:
1354 			mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1355 			break;
1356 		case Opt_barrier:
1357 			if (match_int(&args[0], &option)) {
1358 				status = 0;
1359 				goto bail;
1360 			}
1361 			if (option)
1362 				mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1363 			else
1364 				mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1365 			break;
1366 		case Opt_intr:
1367 			mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1368 			break;
1369 		case Opt_nointr:
1370 			mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1371 			break;
1372 		case Opt_err_panic:
1373 			mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1374 			break;
1375 		case Opt_err_ro:
1376 			mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1377 			break;
1378 		case Opt_data_ordered:
1379 			mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1380 			break;
1381 		case Opt_data_writeback:
1382 			mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1383 			break;
1384 		case Opt_user_xattr:
1385 			mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1386 			break;
1387 		case Opt_nouser_xattr:
1388 			mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1389 			break;
1390 		case Opt_atime_quantum:
1391 			if (match_int(&args[0], &option)) {
1392 				status = 0;
1393 				goto bail;
1394 			}
1395 			if (option >= 0)
1396 				mopt->atime_quantum = option;
1397 			break;
1398 		case Opt_slot:
1399 			option = 0;
1400 			if (match_int(&args[0], &option)) {
1401 				status = 0;
1402 				goto bail;
1403 			}
1404 			if (option)
1405 				mopt->slot = (s16)option;
1406 			break;
1407 		case Opt_commit:
1408 			option = 0;
1409 			if (match_int(&args[0], &option)) {
1410 				status = 0;
1411 				goto bail;
1412 			}
1413 			if (option < 0)
1414 				return 0;
1415 			if (option == 0)
1416 				option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1417 			mopt->commit_interval = HZ * option;
1418 			break;
1419 		case Opt_localalloc:
1420 			option = 0;
1421 			if (match_int(&args[0], &option)) {
1422 				status = 0;
1423 				goto bail;
1424 			}
1425 			if (option >= 0)
1426 				mopt->localalloc_opt = option;
1427 			break;
1428 		case Opt_localflocks:
1429 			/*
1430 			 * Changing this during remount could race
1431 			 * flock() requests, or "unbalance" existing
1432 			 * ones (e.g., a lock is taken in one mode but
1433 			 * dropped in the other). If users care enough
1434 			 * to flip locking modes during remount, we
1435 			 * could add a "local" flag to individual
1436 			 * flock structures for proper tracking of
1437 			 * state.
1438 			 */
1439 			if (!is_remount)
1440 				mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1441 			break;
1442 		case Opt_stack:
1443 			/* Check both that the option we were passed
1444 			 * is of the right length and that it is a proper
1445 			 * string of the right length.
1446 			 */
1447 			if (((args[0].to - args[0].from) !=
1448 			     OCFS2_STACK_LABEL_LEN) ||
1449 			    (strnlen(args[0].from,
1450 				     OCFS2_STACK_LABEL_LEN) !=
1451 			     OCFS2_STACK_LABEL_LEN)) {
1452 				mlog(ML_ERROR,
1453 				     "Invalid cluster_stack option\n");
1454 				status = 0;
1455 				goto bail;
1456 			}
1457 			memcpy(mopt->cluster_stack, args[0].from,
1458 			       OCFS2_STACK_LABEL_LEN);
1459 			mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1460 			break;
1461 		case Opt_inode64:
1462 			mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1463 			break;
1464 		case Opt_usrquota:
1465 			mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1466 			break;
1467 		case Opt_grpquota:
1468 			mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1469 			break;
1470 		case Opt_coherency_buffered:
1471 			mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1472 			break;
1473 		case Opt_coherency_full:
1474 			mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1475 			break;
1476 		case Opt_acl:
1477 			mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1478 			mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1479 			break;
1480 		case Opt_noacl:
1481 			mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1482 			mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1483 			break;
1484 		case Opt_resv_level:
1485 			if (is_remount)
1486 				break;
1487 			if (match_int(&args[0], &option)) {
1488 				status = 0;
1489 				goto bail;
1490 			}
1491 			if (option >= OCFS2_MIN_RESV_LEVEL &&
1492 			    option < OCFS2_MAX_RESV_LEVEL)
1493 				mopt->resv_level = option;
1494 			break;
1495 		case Opt_dir_resv_level:
1496 			if (is_remount)
1497 				break;
1498 			if (match_int(&args[0], &option)) {
1499 				status = 0;
1500 				goto bail;
1501 			}
1502 			if (option >= OCFS2_MIN_RESV_LEVEL &&
1503 			    option < OCFS2_MAX_RESV_LEVEL)
1504 				mopt->dir_resv_level = option;
1505 			break;
1506 		default:
1507 			mlog(ML_ERROR,
1508 			     "Unrecognized mount option \"%s\" "
1509 			     "or missing value\n", p);
1510 			status = 0;
1511 			goto bail;
1512 		}
1513 	}
1514 
1515 	/* Ensure only one heartbeat mode */
1516 	tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
1517 				 OCFS2_MOUNT_HB_NONE);
1518 	if (hweight32(tmp) != 1) {
1519 		mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1520 		status = 0;
1521 		goto bail;
1522 	}
1523 
1524 	status = 1;
1525 
1526 bail:
1527 	mlog_exit(status);
1528 	return status;
1529 }
1530 
1531 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1532 {
1533 	struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb);
1534 	unsigned long opts = osb->s_mount_opt;
1535 	unsigned int local_alloc_megs;
1536 
1537 	if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1538 		seq_printf(s, ",_netdev");
1539 		if (opts & OCFS2_MOUNT_HB_LOCAL)
1540 			seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1541 		else
1542 			seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1543 	} else
1544 		seq_printf(s, ",%s", OCFS2_HB_NONE);
1545 
1546 	if (opts & OCFS2_MOUNT_NOINTR)
1547 		seq_printf(s, ",nointr");
1548 
1549 	if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1550 		seq_printf(s, ",data=writeback");
1551 	else
1552 		seq_printf(s, ",data=ordered");
1553 
1554 	if (opts & OCFS2_MOUNT_BARRIER)
1555 		seq_printf(s, ",barrier=1");
1556 
1557 	if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1558 		seq_printf(s, ",errors=panic");
1559 	else
1560 		seq_printf(s, ",errors=remount-ro");
1561 
1562 	if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1563 		seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1564 
1565 	if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM)
1566 		seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1567 
1568 	if (osb->osb_commit_interval)
1569 		seq_printf(s, ",commit=%u",
1570 			   (unsigned) (osb->osb_commit_interval / HZ));
1571 
1572 	local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1573 	if (local_alloc_megs != ocfs2_la_default_mb(osb))
1574 		seq_printf(s, ",localalloc=%d", local_alloc_megs);
1575 
1576 	if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1577 		seq_printf(s, ",localflocks,");
1578 
1579 	if (osb->osb_cluster_stack[0])
1580 		seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1581 			   osb->osb_cluster_stack);
1582 	if (opts & OCFS2_MOUNT_USRQUOTA)
1583 		seq_printf(s, ",usrquota");
1584 	if (opts & OCFS2_MOUNT_GRPQUOTA)
1585 		seq_printf(s, ",grpquota");
1586 
1587 	if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1588 		seq_printf(s, ",coherency=buffered");
1589 	else
1590 		seq_printf(s, ",coherency=full");
1591 
1592 	if (opts & OCFS2_MOUNT_NOUSERXATTR)
1593 		seq_printf(s, ",nouser_xattr");
1594 	else
1595 		seq_printf(s, ",user_xattr");
1596 
1597 	if (opts & OCFS2_MOUNT_INODE64)
1598 		seq_printf(s, ",inode64");
1599 
1600 	if (opts & OCFS2_MOUNT_POSIX_ACL)
1601 		seq_printf(s, ",acl");
1602 	else
1603 		seq_printf(s, ",noacl");
1604 
1605 	if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1606 		seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1607 
1608 	if (osb->osb_dir_resv_level != osb->osb_resv_level)
1609 		seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1610 
1611 	return 0;
1612 }
1613 
1614 static int __init ocfs2_init(void)
1615 {
1616 	int status;
1617 
1618 	mlog_entry_void();
1619 
1620 	ocfs2_print_version();
1621 
1622 	status = init_ocfs2_uptodate_cache();
1623 	if (status < 0) {
1624 		mlog_errno(status);
1625 		goto leave;
1626 	}
1627 
1628 	status = ocfs2_initialize_mem_caches();
1629 	if (status < 0) {
1630 		mlog_errno(status);
1631 		goto leave;
1632 	}
1633 
1634 	ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1635 	if (!ocfs2_wq) {
1636 		status = -ENOMEM;
1637 		goto leave;
1638 	}
1639 
1640 	ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1641 	if (!ocfs2_debugfs_root) {
1642 		status = -EFAULT;
1643 		mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1644 	}
1645 
1646 	status = ocfs2_quota_setup();
1647 	if (status)
1648 		goto leave;
1649 
1650 	ocfs2_set_locking_protocol();
1651 
1652 	status = register_quota_format(&ocfs2_quota_format);
1653 leave:
1654 	if (status < 0) {
1655 		ocfs2_quota_shutdown();
1656 		ocfs2_free_mem_caches();
1657 		exit_ocfs2_uptodate_cache();
1658 	}
1659 
1660 	mlog_exit(status);
1661 
1662 	if (status >= 0) {
1663 		return register_filesystem(&ocfs2_fs_type);
1664 	} else
1665 		return -1;
1666 }
1667 
1668 static void __exit ocfs2_exit(void)
1669 {
1670 	mlog_entry_void();
1671 
1672 	ocfs2_quota_shutdown();
1673 
1674 	if (ocfs2_wq) {
1675 		flush_workqueue(ocfs2_wq);
1676 		destroy_workqueue(ocfs2_wq);
1677 	}
1678 
1679 	unregister_quota_format(&ocfs2_quota_format);
1680 
1681 	debugfs_remove(ocfs2_debugfs_root);
1682 
1683 	ocfs2_free_mem_caches();
1684 
1685 	unregister_filesystem(&ocfs2_fs_type);
1686 
1687 	exit_ocfs2_uptodate_cache();
1688 
1689 	mlog_exit_void();
1690 }
1691 
1692 static void ocfs2_put_super(struct super_block *sb)
1693 {
1694 	mlog_entry("(0x%p)\n", sb);
1695 
1696 	ocfs2_sync_blockdev(sb);
1697 	ocfs2_dismount_volume(sb, 0);
1698 
1699 	mlog_exit_void();
1700 }
1701 
1702 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1703 {
1704 	struct ocfs2_super *osb;
1705 	u32 numbits, freebits;
1706 	int status;
1707 	struct ocfs2_dinode *bm_lock;
1708 	struct buffer_head *bh = NULL;
1709 	struct inode *inode = NULL;
1710 
1711 	mlog_entry("(%p, %p)\n", dentry->d_sb, buf);
1712 
1713 	osb = OCFS2_SB(dentry->d_sb);
1714 
1715 	inode = ocfs2_get_system_file_inode(osb,
1716 					    GLOBAL_BITMAP_SYSTEM_INODE,
1717 					    OCFS2_INVALID_SLOT);
1718 	if (!inode) {
1719 		mlog(ML_ERROR, "failed to get bitmap inode\n");
1720 		status = -EIO;
1721 		goto bail;
1722 	}
1723 
1724 	status = ocfs2_inode_lock(inode, &bh, 0);
1725 	if (status < 0) {
1726 		mlog_errno(status);
1727 		goto bail;
1728 	}
1729 
1730 	bm_lock = (struct ocfs2_dinode *) bh->b_data;
1731 
1732 	numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1733 	freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1734 
1735 	buf->f_type = OCFS2_SUPER_MAGIC;
1736 	buf->f_bsize = dentry->d_sb->s_blocksize;
1737 	buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1738 	buf->f_blocks = ((sector_t) numbits) *
1739 			(osb->s_clustersize >> osb->sb->s_blocksize_bits);
1740 	buf->f_bfree = ((sector_t) freebits) *
1741 		       (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1742 	buf->f_bavail = buf->f_bfree;
1743 	buf->f_files = numbits;
1744 	buf->f_ffree = freebits;
1745 	buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1746 				& 0xFFFFFFFFUL;
1747 	buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1748 				OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1749 
1750 	brelse(bh);
1751 
1752 	ocfs2_inode_unlock(inode, 0);
1753 	status = 0;
1754 bail:
1755 	if (inode)
1756 		iput(inode);
1757 
1758 	mlog_exit(status);
1759 
1760 	return status;
1761 }
1762 
1763 static void ocfs2_inode_init_once(void *data)
1764 {
1765 	struct ocfs2_inode_info *oi = data;
1766 
1767 	oi->ip_flags = 0;
1768 	oi->ip_open_count = 0;
1769 	spin_lock_init(&oi->ip_lock);
1770 	ocfs2_extent_map_init(&oi->vfs_inode);
1771 	INIT_LIST_HEAD(&oi->ip_io_markers);
1772 	oi->ip_dir_start_lookup = 0;
1773 
1774 	init_rwsem(&oi->ip_alloc_sem);
1775 	init_rwsem(&oi->ip_xattr_sem);
1776 	mutex_init(&oi->ip_io_mutex);
1777 
1778 	oi->ip_blkno = 0ULL;
1779 	oi->ip_clusters = 0;
1780 
1781 	ocfs2_resv_init_once(&oi->ip_la_data_resv);
1782 
1783 	ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1784 	ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1785 	ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1786 
1787 	ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1788 				  &ocfs2_inode_caching_ops);
1789 
1790 	inode_init_once(&oi->vfs_inode);
1791 }
1792 
1793 static int ocfs2_initialize_mem_caches(void)
1794 {
1795 	ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1796 				       sizeof(struct ocfs2_inode_info),
1797 				       0,
1798 				       (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1799 						SLAB_MEM_SPREAD),
1800 				       ocfs2_inode_init_once);
1801 	ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1802 					sizeof(struct ocfs2_dquot),
1803 					0,
1804 					(SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1805 						SLAB_MEM_SPREAD),
1806 					NULL);
1807 	ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1808 					sizeof(struct ocfs2_quota_chunk),
1809 					0,
1810 					(SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1811 					NULL);
1812 	if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1813 	    !ocfs2_qf_chunk_cachep) {
1814 		if (ocfs2_inode_cachep)
1815 			kmem_cache_destroy(ocfs2_inode_cachep);
1816 		if (ocfs2_dquot_cachep)
1817 			kmem_cache_destroy(ocfs2_dquot_cachep);
1818 		if (ocfs2_qf_chunk_cachep)
1819 			kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1820 		return -ENOMEM;
1821 	}
1822 
1823 	return 0;
1824 }
1825 
1826 static void ocfs2_free_mem_caches(void)
1827 {
1828 	if (ocfs2_inode_cachep)
1829 		kmem_cache_destroy(ocfs2_inode_cachep);
1830 	ocfs2_inode_cachep = NULL;
1831 
1832 	if (ocfs2_dquot_cachep)
1833 		kmem_cache_destroy(ocfs2_dquot_cachep);
1834 	ocfs2_dquot_cachep = NULL;
1835 
1836 	if (ocfs2_qf_chunk_cachep)
1837 		kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1838 	ocfs2_qf_chunk_cachep = NULL;
1839 }
1840 
1841 static int ocfs2_get_sector(struct super_block *sb,
1842 			    struct buffer_head **bh,
1843 			    int block,
1844 			    int sect_size)
1845 {
1846 	if (!sb_set_blocksize(sb, sect_size)) {
1847 		mlog(ML_ERROR, "unable to set blocksize\n");
1848 		return -EIO;
1849 	}
1850 
1851 	*bh = sb_getblk(sb, block);
1852 	if (!*bh) {
1853 		mlog_errno(-EIO);
1854 		return -EIO;
1855 	}
1856 	lock_buffer(*bh);
1857 	if (!buffer_dirty(*bh))
1858 		clear_buffer_uptodate(*bh);
1859 	unlock_buffer(*bh);
1860 	ll_rw_block(READ, 1, bh);
1861 	wait_on_buffer(*bh);
1862 	if (!buffer_uptodate(*bh)) {
1863 		mlog_errno(-EIO);
1864 		brelse(*bh);
1865 		*bh = NULL;
1866 		return -EIO;
1867 	}
1868 
1869 	return 0;
1870 }
1871 
1872 static int ocfs2_mount_volume(struct super_block *sb)
1873 {
1874 	int status = 0;
1875 	int unlock_super = 0;
1876 	struct ocfs2_super *osb = OCFS2_SB(sb);
1877 
1878 	mlog_entry_void();
1879 
1880 	if (ocfs2_is_hard_readonly(osb))
1881 		goto leave;
1882 
1883 	status = ocfs2_dlm_init(osb);
1884 	if (status < 0) {
1885 		mlog_errno(status);
1886 		goto leave;
1887 	}
1888 
1889 	status = ocfs2_super_lock(osb, 1);
1890 	if (status < 0) {
1891 		mlog_errno(status);
1892 		goto leave;
1893 	}
1894 	unlock_super = 1;
1895 
1896 	/* This will load up the node map and add ourselves to it. */
1897 	status = ocfs2_find_slot(osb);
1898 	if (status < 0) {
1899 		mlog_errno(status);
1900 		goto leave;
1901 	}
1902 
1903 	/* load all node-local system inodes */
1904 	status = ocfs2_init_local_system_inodes(osb);
1905 	if (status < 0) {
1906 		mlog_errno(status);
1907 		goto leave;
1908 	}
1909 
1910 	status = ocfs2_check_volume(osb);
1911 	if (status < 0) {
1912 		mlog_errno(status);
1913 		goto leave;
1914 	}
1915 
1916 	status = ocfs2_truncate_log_init(osb);
1917 	if (status < 0)
1918 		mlog_errno(status);
1919 
1920 leave:
1921 	if (unlock_super)
1922 		ocfs2_super_unlock(osb, 1);
1923 
1924 	mlog_exit(status);
1925 	return status;
1926 }
1927 
1928 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1929 {
1930 	int tmp, hangup_needed = 0;
1931 	struct ocfs2_super *osb = NULL;
1932 	char nodestr[8];
1933 
1934 	mlog_entry("(0x%p)\n", sb);
1935 
1936 	BUG_ON(!sb);
1937 	osb = OCFS2_SB(sb);
1938 	BUG_ON(!osb);
1939 
1940 	debugfs_remove(osb->osb_ctxt);
1941 
1942 	/*
1943 	 * Flush inode dropping work queue so that deletes are
1944 	 * performed while the filesystem is still working
1945 	 */
1946 	ocfs2_drop_all_dl_inodes(osb);
1947 
1948 	/* Orphan scan should be stopped as early as possible */
1949 	ocfs2_orphan_scan_stop(osb);
1950 
1951 	ocfs2_disable_quotas(osb);
1952 
1953 	ocfs2_shutdown_local_alloc(osb);
1954 
1955 	ocfs2_truncate_log_shutdown(osb);
1956 
1957 	/* This will disable recovery and flush any recovery work. */
1958 	ocfs2_recovery_exit(osb);
1959 
1960 	ocfs2_journal_shutdown(osb);
1961 
1962 	ocfs2_sync_blockdev(sb);
1963 
1964 	ocfs2_purge_refcount_trees(osb);
1965 
1966 	/* No cluster connection means we've failed during mount, so skip
1967 	 * all the steps which depended on that to complete. */
1968 	if (osb->cconn) {
1969 		tmp = ocfs2_super_lock(osb, 1);
1970 		if (tmp < 0) {
1971 			mlog_errno(tmp);
1972 			return;
1973 		}
1974 	}
1975 
1976 	if (osb->slot_num != OCFS2_INVALID_SLOT)
1977 		ocfs2_put_slot(osb);
1978 
1979 	if (osb->cconn)
1980 		ocfs2_super_unlock(osb, 1);
1981 
1982 	ocfs2_release_system_inodes(osb);
1983 
1984 	/*
1985 	 * If we're dismounting due to mount error, mount.ocfs2 will clean
1986 	 * up heartbeat.  If we're a local mount, there is no heartbeat.
1987 	 * If we failed before we got a uuid_str yet, we can't stop
1988 	 * heartbeat.  Otherwise, do it.
1989 	 */
1990 	if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str)
1991 		hangup_needed = 1;
1992 
1993 	if (osb->cconn)
1994 		ocfs2_dlm_shutdown(osb, hangup_needed);
1995 
1996 	ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1997 	debugfs_remove(osb->osb_debug_root);
1998 
1999 	if (hangup_needed)
2000 		ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
2001 
2002 	atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
2003 
2004 	if (ocfs2_mount_local(osb))
2005 		snprintf(nodestr, sizeof(nodestr), "local");
2006 	else
2007 		snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
2008 
2009 	printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
2010 	       osb->dev_str, nodestr);
2011 
2012 	ocfs2_delete_osb(osb);
2013 	kfree(osb);
2014 	sb->s_dev = 0;
2015 	sb->s_fs_info = NULL;
2016 }
2017 
2018 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
2019 				unsigned uuid_bytes)
2020 {
2021 	int i, ret;
2022 	char *ptr;
2023 
2024 	BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
2025 
2026 	osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
2027 	if (osb->uuid_str == NULL)
2028 		return -ENOMEM;
2029 
2030 	for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
2031 		/* print with null */
2032 		ret = snprintf(ptr, 3, "%02X", uuid[i]);
2033 		if (ret != 2) /* drop super cleans up */
2034 			return -EINVAL;
2035 		/* then only advance past the last char */
2036 		ptr += 2;
2037 	}
2038 
2039 	return 0;
2040 }
2041 
2042 /* Make sure entire volume is addressable by our journal.  Requires
2043    osb_clusters_at_boot to be valid and for the journal to have been
2044    initialized by ocfs2_journal_init(). */
2045 static int ocfs2_journal_addressable(struct ocfs2_super *osb)
2046 {
2047 	int status = 0;
2048 	u64 max_block =
2049 		ocfs2_clusters_to_blocks(osb->sb,
2050 					 osb->osb_clusters_at_boot) - 1;
2051 
2052 	/* 32-bit block number is always OK. */
2053 	if (max_block <= (u32)~0ULL)
2054 		goto out;
2055 
2056 	/* Volume is "huge", so see if our journal is new enough to
2057 	   support it. */
2058 	if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2059 				       OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2060 	      jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2061 					       JBD2_FEATURE_INCOMPAT_64BIT))) {
2062 		mlog(ML_ERROR, "The journal cannot address the entire volume. "
2063 		     "Enable the 'block64' journal option with tunefs.ocfs2");
2064 		status = -EFBIG;
2065 		goto out;
2066 	}
2067 
2068  out:
2069 	return status;
2070 }
2071 
2072 static int ocfs2_initialize_super(struct super_block *sb,
2073 				  struct buffer_head *bh,
2074 				  int sector_size,
2075 				  struct ocfs2_blockcheck_stats *stats)
2076 {
2077 	int status;
2078 	int i, cbits, bbits;
2079 	struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2080 	struct inode *inode = NULL;
2081 	struct ocfs2_journal *journal;
2082 	__le32 uuid_net_key;
2083 	struct ocfs2_super *osb;
2084 	u64 total_blocks;
2085 
2086 	mlog_entry_void();
2087 
2088 	osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2089 	if (!osb) {
2090 		status = -ENOMEM;
2091 		mlog_errno(status);
2092 		goto bail;
2093 	}
2094 
2095 	sb->s_fs_info = osb;
2096 	sb->s_op = &ocfs2_sops;
2097 	sb->s_export_op = &ocfs2_export_ops;
2098 	sb->s_qcop = &ocfs2_quotactl_ops;
2099 	sb->dq_op = &ocfs2_quota_operations;
2100 	sb->s_xattr = ocfs2_xattr_handlers;
2101 	sb->s_time_gran = 1;
2102 	sb->s_flags |= MS_NOATIME;
2103 	/* this is needed to support O_LARGEFILE */
2104 	cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2105 	bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2106 	sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2107 
2108 	osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2109 
2110 	for (i = 0; i < 3; i++)
2111 		osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2112 	osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2113 
2114 	osb->sb = sb;
2115 	/* Save off for ocfs2_rw_direct */
2116 	osb->s_sectsize_bits = blksize_bits(sector_size);
2117 	BUG_ON(!osb->s_sectsize_bits);
2118 
2119 	spin_lock_init(&osb->dc_task_lock);
2120 	init_waitqueue_head(&osb->dc_event);
2121 	osb->dc_work_sequence = 0;
2122 	osb->dc_wake_sequence = 0;
2123 	INIT_LIST_HEAD(&osb->blocked_lock_list);
2124 	osb->blocked_lock_count = 0;
2125 	spin_lock_init(&osb->osb_lock);
2126 	spin_lock_init(&osb->osb_xattr_lock);
2127 	ocfs2_init_steal_slots(osb);
2128 
2129 	atomic_set(&osb->alloc_stats.moves, 0);
2130 	atomic_set(&osb->alloc_stats.local_data, 0);
2131 	atomic_set(&osb->alloc_stats.bitmap_data, 0);
2132 	atomic_set(&osb->alloc_stats.bg_allocs, 0);
2133 	atomic_set(&osb->alloc_stats.bg_extends, 0);
2134 
2135 	/* Copy the blockcheck stats from the superblock probe */
2136 	osb->osb_ecc_stats = *stats;
2137 
2138 	ocfs2_init_node_maps(osb);
2139 
2140 	snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2141 		 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2142 
2143 	osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2144 	if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2145 		mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2146 		     osb->max_slots);
2147 		status = -EINVAL;
2148 		goto bail;
2149 	}
2150 	mlog(0, "max_slots for this device: %u\n", osb->max_slots);
2151 
2152 	ocfs2_orphan_scan_init(osb);
2153 
2154 	status = ocfs2_recovery_init(osb);
2155 	if (status) {
2156 		mlog(ML_ERROR, "Unable to initialize recovery state\n");
2157 		mlog_errno(status);
2158 		goto bail;
2159 	}
2160 
2161 	init_waitqueue_head(&osb->checkpoint_event);
2162 	atomic_set(&osb->needs_checkpoint, 0);
2163 
2164 	osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2165 
2166 	osb->slot_num = OCFS2_INVALID_SLOT;
2167 
2168 	osb->s_xattr_inline_size = le16_to_cpu(
2169 					di->id2.i_super.s_xattr_inline_size);
2170 
2171 	osb->local_alloc_state = OCFS2_LA_UNUSED;
2172 	osb->local_alloc_bh = NULL;
2173 	INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2174 
2175 	init_waitqueue_head(&osb->osb_mount_event);
2176 
2177 	status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2178 	if (status) {
2179 		mlog_errno(status);
2180 		goto bail;
2181 	}
2182 
2183 	osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2184 	if (!osb->vol_label) {
2185 		mlog(ML_ERROR, "unable to alloc vol label\n");
2186 		status = -ENOMEM;
2187 		goto bail;
2188 	}
2189 
2190 	osb->slot_recovery_generations =
2191 		kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2192 			GFP_KERNEL);
2193 	if (!osb->slot_recovery_generations) {
2194 		status = -ENOMEM;
2195 		mlog_errno(status);
2196 		goto bail;
2197 	}
2198 
2199 	init_waitqueue_head(&osb->osb_wipe_event);
2200 	osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2201 					sizeof(*osb->osb_orphan_wipes),
2202 					GFP_KERNEL);
2203 	if (!osb->osb_orphan_wipes) {
2204 		status = -ENOMEM;
2205 		mlog_errno(status);
2206 		goto bail;
2207 	}
2208 
2209 	osb->osb_rf_lock_tree = RB_ROOT;
2210 
2211 	osb->s_feature_compat =
2212 		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2213 	osb->s_feature_ro_compat =
2214 		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2215 	osb->s_feature_incompat =
2216 		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2217 
2218 	if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2219 		mlog(ML_ERROR, "couldn't mount because of unsupported "
2220 		     "optional features (%x).\n", i);
2221 		status = -EINVAL;
2222 		goto bail;
2223 	}
2224 	if (!(osb->sb->s_flags & MS_RDONLY) &&
2225 	    (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2226 		mlog(ML_ERROR, "couldn't mount RDWR because of "
2227 		     "unsupported optional features (%x).\n", i);
2228 		status = -EINVAL;
2229 		goto bail;
2230 	}
2231 
2232 	if (ocfs2_clusterinfo_valid(osb)) {
2233 		osb->osb_stackflags =
2234 			OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2235 		memcpy(osb->osb_cluster_stack,
2236 		       OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2237 		       OCFS2_STACK_LABEL_LEN);
2238 		osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
2239 		if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2240 			mlog(ML_ERROR,
2241 			     "couldn't mount because of an invalid "
2242 			     "cluster stack label (%s) \n",
2243 			     osb->osb_cluster_stack);
2244 			status = -EINVAL;
2245 			goto bail;
2246 		}
2247 	} else {
2248 		/* The empty string is identical with classic tools that
2249 		 * don't know about s_cluster_info. */
2250 		osb->osb_cluster_stack[0] = '\0';
2251 	}
2252 
2253 	get_random_bytes(&osb->s_next_generation, sizeof(u32));
2254 
2255 	/* FIXME
2256 	 * This should be done in ocfs2_journal_init(), but unknown
2257 	 * ordering issues will cause the filesystem to crash.
2258 	 * If anyone wants to figure out what part of the code
2259 	 * refers to osb->journal before ocfs2_journal_init() is run,
2260 	 * be my guest.
2261 	 */
2262 	/* initialize our journal structure */
2263 
2264 	journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2265 	if (!journal) {
2266 		mlog(ML_ERROR, "unable to alloc journal\n");
2267 		status = -ENOMEM;
2268 		goto bail;
2269 	}
2270 	osb->journal = journal;
2271 	journal->j_osb = osb;
2272 
2273 	atomic_set(&journal->j_num_trans, 0);
2274 	init_rwsem(&journal->j_trans_barrier);
2275 	init_waitqueue_head(&journal->j_checkpointed);
2276 	spin_lock_init(&journal->j_lock);
2277 	journal->j_trans_id = (unsigned long) 1;
2278 	INIT_LIST_HEAD(&journal->j_la_cleanups);
2279 	INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2280 	journal->j_state = OCFS2_JOURNAL_FREE;
2281 
2282 	INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes);
2283 	osb->dentry_lock_list = NULL;
2284 
2285 	/* get some pseudo constants for clustersize bits */
2286 	osb->s_clustersize_bits =
2287 		le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2288 	osb->s_clustersize = 1 << osb->s_clustersize_bits;
2289 	mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
2290 
2291 	if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2292 	    osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2293 		mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2294 		     osb->s_clustersize);
2295 		status = -EINVAL;
2296 		goto bail;
2297 	}
2298 
2299 	total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2300 						le32_to_cpu(di->i_clusters));
2301 
2302 	status = generic_check_addressable(osb->sb->s_blocksize_bits,
2303 					   total_blocks);
2304 	if (status) {
2305 		mlog(ML_ERROR, "Volume too large "
2306 		     "to mount safely on this system");
2307 		status = -EFBIG;
2308 		goto bail;
2309 	}
2310 
2311 	if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2312 				 sizeof(di->id2.i_super.s_uuid))) {
2313 		mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2314 		status = -ENOMEM;
2315 		goto bail;
2316 	}
2317 
2318 	memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
2319 
2320 	strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2321 	osb->vol_label[63] = '\0';
2322 	osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2323 	osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2324 	osb->first_cluster_group_blkno =
2325 		le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2326 	osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2327 	osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2328 	mlog(0, "vol_label: %s\n", osb->vol_label);
2329 	mlog(0, "uuid: %s\n", osb->uuid_str);
2330 	mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
2331 	     (unsigned long long)osb->root_blkno,
2332 	     (unsigned long long)osb->system_dir_blkno);
2333 
2334 	osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2335 	if (!osb->osb_dlm_debug) {
2336 		status = -ENOMEM;
2337 		mlog_errno(status);
2338 		goto bail;
2339 	}
2340 
2341 	atomic_set(&osb->vol_state, VOLUME_INIT);
2342 
2343 	/* load root, system_dir, and all global system inodes */
2344 	status = ocfs2_init_global_system_inodes(osb);
2345 	if (status < 0) {
2346 		mlog_errno(status);
2347 		goto bail;
2348 	}
2349 
2350 	/*
2351 	 * global bitmap
2352 	 */
2353 	inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2354 					    OCFS2_INVALID_SLOT);
2355 	if (!inode) {
2356 		status = -EINVAL;
2357 		mlog_errno(status);
2358 		goto bail;
2359 	}
2360 
2361 	osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2362 	osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2363 	iput(inode);
2364 
2365 	osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2366 				 osb->s_feature_incompat) * 8;
2367 
2368 	status = ocfs2_init_slot_info(osb);
2369 	if (status < 0) {
2370 		mlog_errno(status);
2371 		goto bail;
2372 	}
2373 
2374 bail:
2375 	mlog_exit(status);
2376 	return status;
2377 }
2378 
2379 /*
2380  * will return: -EAGAIN if it is ok to keep searching for superblocks
2381  *              -EINVAL if there is a bad superblock
2382  *              0 on success
2383  */
2384 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2385 			       struct buffer_head *bh,
2386 			       u32 blksz,
2387 			       struct ocfs2_blockcheck_stats *stats)
2388 {
2389 	int status = -EAGAIN;
2390 
2391 	mlog_entry_void();
2392 
2393 	if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2394 		   strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2395 		/* We have to do a raw check of the feature here */
2396 		if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2397 		    OCFS2_FEATURE_INCOMPAT_META_ECC) {
2398 			status = ocfs2_block_check_validate(bh->b_data,
2399 							    bh->b_size,
2400 							    &di->i_check,
2401 							    stats);
2402 			if (status)
2403 				goto out;
2404 		}
2405 		status = -EINVAL;
2406 		if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2407 			mlog(ML_ERROR, "found superblock with incorrect block "
2408 			     "size: found %u, should be %u\n",
2409 			     1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2410 			       blksz);
2411 		} else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2412 			   OCFS2_MAJOR_REV_LEVEL ||
2413 			   le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2414 			   OCFS2_MINOR_REV_LEVEL) {
2415 			mlog(ML_ERROR, "found superblock with bad version: "
2416 			     "found %u.%u, should be %u.%u\n",
2417 			     le16_to_cpu(di->id2.i_super.s_major_rev_level),
2418 			     le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2419 			     OCFS2_MAJOR_REV_LEVEL,
2420 			     OCFS2_MINOR_REV_LEVEL);
2421 		} else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2422 			mlog(ML_ERROR, "bad block number on superblock: "
2423 			     "found %llu, should be %llu\n",
2424 			     (unsigned long long)le64_to_cpu(di->i_blkno),
2425 			     (unsigned long long)bh->b_blocknr);
2426 		} else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2427 			    le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2428 			mlog(ML_ERROR, "bad cluster size found: %u\n",
2429 			     1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2430 		} else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2431 			mlog(ML_ERROR, "bad root_blkno: 0\n");
2432 		} else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2433 			mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2434 		} else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2435 			mlog(ML_ERROR,
2436 			     "Superblock slots found greater than file system "
2437 			     "maximum: found %u, max %u\n",
2438 			     le16_to_cpu(di->id2.i_super.s_max_slots),
2439 			     OCFS2_MAX_SLOTS);
2440 		} else {
2441 			/* found it! */
2442 			status = 0;
2443 		}
2444 	}
2445 
2446 out:
2447 	mlog_exit(status);
2448 	return status;
2449 }
2450 
2451 static int ocfs2_check_volume(struct ocfs2_super *osb)
2452 {
2453 	int status;
2454 	int dirty;
2455 	int local;
2456 	struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2457 						  * recover
2458 						  * ourselves. */
2459 
2460 	mlog_entry_void();
2461 
2462 	/* Init our journal object. */
2463 	status = ocfs2_journal_init(osb->journal, &dirty);
2464 	if (status < 0) {
2465 		mlog(ML_ERROR, "Could not initialize journal!\n");
2466 		goto finally;
2467 	}
2468 
2469 	/* Now that journal has been initialized, check to make sure
2470 	   entire volume is addressable. */
2471 	status = ocfs2_journal_addressable(osb);
2472 	if (status)
2473 		goto finally;
2474 
2475 	/* If the journal was unmounted cleanly then we don't want to
2476 	 * recover anything. Otherwise, journal_load will do that
2477 	 * dirty work for us :) */
2478 	if (!dirty) {
2479 		status = ocfs2_journal_wipe(osb->journal, 0);
2480 		if (status < 0) {
2481 			mlog_errno(status);
2482 			goto finally;
2483 		}
2484 	} else {
2485 		mlog(ML_NOTICE, "File system was not unmounted cleanly, "
2486 		     "recovering volume.\n");
2487 	}
2488 
2489 	local = ocfs2_mount_local(osb);
2490 
2491 	/* will play back anything left in the journal. */
2492 	status = ocfs2_journal_load(osb->journal, local, dirty);
2493 	if (status < 0) {
2494 		mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2495 		goto finally;
2496 	}
2497 
2498 	if (dirty) {
2499 		/* recover my local alloc if we didn't unmount cleanly. */
2500 		status = ocfs2_begin_local_alloc_recovery(osb,
2501 							  osb->slot_num,
2502 							  &local_alloc);
2503 		if (status < 0) {
2504 			mlog_errno(status);
2505 			goto finally;
2506 		}
2507 		/* we complete the recovery process after we've marked
2508 		 * ourselves as mounted. */
2509 	}
2510 
2511 	mlog(0, "Journal loaded.\n");
2512 
2513 	status = ocfs2_load_local_alloc(osb);
2514 	if (status < 0) {
2515 		mlog_errno(status);
2516 		goto finally;
2517 	}
2518 
2519 	if (dirty) {
2520 		/* Recovery will be completed after we've mounted the
2521 		 * rest of the volume. */
2522 		osb->dirty = 1;
2523 		osb->local_alloc_copy = local_alloc;
2524 		local_alloc = NULL;
2525 	}
2526 
2527 	/* go through each journal, trylock it and if you get the
2528 	 * lock, and it's marked as dirty, set the bit in the recover
2529 	 * map and launch a recovery thread for it. */
2530 	status = ocfs2_mark_dead_nodes(osb);
2531 	if (status < 0) {
2532 		mlog_errno(status);
2533 		goto finally;
2534 	}
2535 
2536 	status = ocfs2_compute_replay_slots(osb);
2537 	if (status < 0)
2538 		mlog_errno(status);
2539 
2540 finally:
2541 	if (local_alloc)
2542 		kfree(local_alloc);
2543 
2544 	mlog_exit(status);
2545 	return status;
2546 }
2547 
2548 /*
2549  * The routine gets called from dismount or close whenever a dismount on
2550  * volume is requested and the osb open count becomes 1.
2551  * It will remove the osb from the global list and also free up all the
2552  * initialized resources and fileobject.
2553  */
2554 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2555 {
2556 	mlog_entry_void();
2557 
2558 	/* This function assumes that the caller has the main osb resource */
2559 
2560 	ocfs2_free_slot_info(osb);
2561 
2562 	kfree(osb->osb_orphan_wipes);
2563 	kfree(osb->slot_recovery_generations);
2564 	/* FIXME
2565 	 * This belongs in journal shutdown, but because we have to
2566 	 * allocate osb->journal at the start of ocfs2_initialize_osb(),
2567 	 * we free it here.
2568 	 */
2569 	kfree(osb->journal);
2570 	if (osb->local_alloc_copy)
2571 		kfree(osb->local_alloc_copy);
2572 	kfree(osb->uuid_str);
2573 	ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2574 	memset(osb, 0, sizeof(struct ocfs2_super));
2575 
2576 	mlog_exit_void();
2577 }
2578 
2579 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2580  * panic(). We do not support continue-on-error operation. */
2581 static void ocfs2_handle_error(struct super_block *sb)
2582 {
2583 	struct ocfs2_super *osb = OCFS2_SB(sb);
2584 
2585 	if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2586 		panic("OCFS2: (device %s): panic forced after error\n",
2587 		      sb->s_id);
2588 
2589 	ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2590 
2591 	if (sb->s_flags & MS_RDONLY &&
2592 	    (ocfs2_is_soft_readonly(osb) ||
2593 	     ocfs2_is_hard_readonly(osb)))
2594 		return;
2595 
2596 	printk(KERN_CRIT "File system is now read-only due to the potential "
2597 	       "of on-disk corruption. Please run fsck.ocfs2 once the file "
2598 	       "system is unmounted.\n");
2599 	sb->s_flags |= MS_RDONLY;
2600 	ocfs2_set_ro_flag(osb, 0);
2601 }
2602 
2603 static char error_buf[1024];
2604 
2605 void __ocfs2_error(struct super_block *sb,
2606 		   const char *function,
2607 		   const char *fmt, ...)
2608 {
2609 	va_list args;
2610 
2611 	va_start(args, fmt);
2612 	vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2613 	va_end(args);
2614 
2615 	/* Not using mlog here because we want to show the actual
2616 	 * function the error came from. */
2617 	printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2618 	       sb->s_id, function, error_buf);
2619 
2620 	ocfs2_handle_error(sb);
2621 }
2622 
2623 /* Handle critical errors. This is intentionally more drastic than
2624  * ocfs2_handle_error, so we only use for things like journal errors,
2625  * etc. */
2626 void __ocfs2_abort(struct super_block* sb,
2627 		   const char *function,
2628 		   const char *fmt, ...)
2629 {
2630 	va_list args;
2631 
2632 	va_start(args, fmt);
2633 	vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2634 	va_end(args);
2635 
2636 	printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2637 	       sb->s_id, function, error_buf);
2638 
2639 	/* We don't have the cluster support yet to go straight to
2640 	 * hard readonly in here. Until then, we want to keep
2641 	 * ocfs2_abort() so that we can at least mark critical
2642 	 * errors.
2643 	 *
2644 	 * TODO: This should abort the journal and alert other nodes
2645 	 * that our slot needs recovery. */
2646 
2647 	/* Force a panic(). This stinks, but it's better than letting
2648 	 * things continue without having a proper hard readonly
2649 	 * here. */
2650 	if (!ocfs2_mount_local(OCFS2_SB(sb)))
2651 		OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2652 	ocfs2_handle_error(sb);
2653 }
2654 
2655 /*
2656  * Void signal blockers, because in-kernel sigprocmask() only fails
2657  * when SIG_* is wrong.
2658  */
2659 void ocfs2_block_signals(sigset_t *oldset)
2660 {
2661 	int rc;
2662 	sigset_t blocked;
2663 
2664 	sigfillset(&blocked);
2665 	rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2666 	BUG_ON(rc);
2667 }
2668 
2669 void ocfs2_unblock_signals(sigset_t *oldset)
2670 {
2671 	int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2672 	BUG_ON(rc);
2673 }
2674 
2675 module_init(ocfs2_init);
2676 module_exit(ocfs2_exit);
2677