• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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/cleancache.h>
45 
46 #define CREATE_TRACE_POINTS
47 #include "ocfs2_trace.h"
48 
49 #include <cluster/masklog.h>
50 
51 #include "ocfs2.h"
52 
53 /* this should be the only file to include a version 1 header */
54 #include "ocfs1_fs_compat.h"
55 
56 #include "alloc.h"
57 #include "aops.h"
58 #include "blockcheck.h"
59 #include "dlmglue.h"
60 #include "export.h"
61 #include "extent_map.h"
62 #include "heartbeat.h"
63 #include "inode.h"
64 #include "journal.h"
65 #include "localalloc.h"
66 #include "namei.h"
67 #include "slot_map.h"
68 #include "super.h"
69 #include "sysfile.h"
70 #include "uptodate.h"
71 #include "ver.h"
72 #include "xattr.h"
73 #include "quota.h"
74 #include "refcounttree.h"
75 #include "suballoc.h"
76 
77 #include "buffer_head_io.h"
78 
79 static struct kmem_cache *ocfs2_inode_cachep = NULL;
80 struct kmem_cache *ocfs2_dquot_cachep;
81 struct kmem_cache *ocfs2_qf_chunk_cachep;
82 
83 /* OCFS2 needs to schedule several different types of work which
84  * require cluster locking, disk I/O, recovery waits, etc. Since these
85  * types of work tend to be heavy we avoid using the kernel events
86  * workqueue and schedule on our own. */
87 struct workqueue_struct *ocfs2_wq = NULL;
88 
89 static struct dentry *ocfs2_debugfs_root = NULL;
90 
91 MODULE_AUTHOR("Oracle");
92 MODULE_LICENSE("GPL");
93 
94 struct mount_options
95 {
96 	unsigned long	commit_interval;
97 	unsigned long	mount_opt;
98 	unsigned int	atime_quantum;
99 	signed short	slot;
100 	int		localalloc_opt;
101 	unsigned int	resv_level;
102 	int		dir_resv_level;
103 	char		cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
104 };
105 
106 static int ocfs2_parse_options(struct super_block *sb, char *options,
107 			       struct mount_options *mopt,
108 			       int is_remount);
109 static int ocfs2_check_set_options(struct super_block *sb,
110 				   struct mount_options *options);
111 static int ocfs2_show_options(struct seq_file *s, struct dentry *root);
112 static void ocfs2_put_super(struct super_block *sb);
113 static int ocfs2_mount_volume(struct super_block *sb);
114 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
115 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
116 static int ocfs2_initialize_mem_caches(void);
117 static void ocfs2_free_mem_caches(void);
118 static void ocfs2_delete_osb(struct ocfs2_super *osb);
119 
120 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
121 
122 static int ocfs2_sync_fs(struct super_block *sb, int wait);
123 
124 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
125 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
126 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
127 static int ocfs2_check_volume(struct ocfs2_super *osb);
128 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
129 			       struct buffer_head *bh,
130 			       u32 sectsize,
131 			       struct ocfs2_blockcheck_stats *stats);
132 static int ocfs2_initialize_super(struct super_block *sb,
133 				  struct buffer_head *bh,
134 				  int sector_size,
135 				  struct ocfs2_blockcheck_stats *stats);
136 static int ocfs2_get_sector(struct super_block *sb,
137 			    struct buffer_head **bh,
138 			    int block,
139 			    int sect_size);
140 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
141 static void ocfs2_destroy_inode(struct inode *inode);
142 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
143 static int ocfs2_enable_quotas(struct ocfs2_super *osb);
144 static void ocfs2_disable_quotas(struct ocfs2_super *osb);
145 
146 static const struct super_operations ocfs2_sops = {
147 	.statfs		= ocfs2_statfs,
148 	.alloc_inode	= ocfs2_alloc_inode,
149 	.destroy_inode	= ocfs2_destroy_inode,
150 	.drop_inode	= ocfs2_drop_inode,
151 	.evict_inode	= ocfs2_evict_inode,
152 	.sync_fs	= ocfs2_sync_fs,
153 	.put_super	= ocfs2_put_super,
154 	.remount_fs	= ocfs2_remount,
155 	.show_options   = ocfs2_show_options,
156 	.quota_read	= ocfs2_quota_read,
157 	.quota_write	= ocfs2_quota_write,
158 };
159 
160 enum {
161 	Opt_barrier,
162 	Opt_err_panic,
163 	Opt_err_ro,
164 	Opt_intr,
165 	Opt_nointr,
166 	Opt_hb_none,
167 	Opt_hb_local,
168 	Opt_hb_global,
169 	Opt_data_ordered,
170 	Opt_data_writeback,
171 	Opt_atime_quantum,
172 	Opt_slot,
173 	Opt_commit,
174 	Opt_localalloc,
175 	Opt_localflocks,
176 	Opt_stack,
177 	Opt_user_xattr,
178 	Opt_nouser_xattr,
179 	Opt_inode64,
180 	Opt_acl,
181 	Opt_noacl,
182 	Opt_usrquota,
183 	Opt_grpquota,
184 	Opt_coherency_buffered,
185 	Opt_coherency_full,
186 	Opt_resv_level,
187 	Opt_dir_resv_level,
188 	Opt_err,
189 };
190 
191 static const match_table_t tokens = {
192 	{Opt_barrier, "barrier=%u"},
193 	{Opt_err_panic, "errors=panic"},
194 	{Opt_err_ro, "errors=remount-ro"},
195 	{Opt_intr, "intr"},
196 	{Opt_nointr, "nointr"},
197 	{Opt_hb_none, OCFS2_HB_NONE},
198 	{Opt_hb_local, OCFS2_HB_LOCAL},
199 	{Opt_hb_global, OCFS2_HB_GLOBAL},
200 	{Opt_data_ordered, "data=ordered"},
201 	{Opt_data_writeback, "data=writeback"},
202 	{Opt_atime_quantum, "atime_quantum=%u"},
203 	{Opt_slot, "preferred_slot=%u"},
204 	{Opt_commit, "commit=%u"},
205 	{Opt_localalloc, "localalloc=%d"},
206 	{Opt_localflocks, "localflocks"},
207 	{Opt_stack, "cluster_stack=%s"},
208 	{Opt_user_xattr, "user_xattr"},
209 	{Opt_nouser_xattr, "nouser_xattr"},
210 	{Opt_inode64, "inode64"},
211 	{Opt_acl, "acl"},
212 	{Opt_noacl, "noacl"},
213 	{Opt_usrquota, "usrquota"},
214 	{Opt_grpquota, "grpquota"},
215 	{Opt_coherency_buffered, "coherency=buffered"},
216 	{Opt_coherency_full, "coherency=full"},
217 	{Opt_resv_level, "resv_level=%u"},
218 	{Opt_dir_resv_level, "dir_resv_level=%u"},
219 	{Opt_err, NULL}
220 };
221 
222 #ifdef CONFIG_DEBUG_FS
ocfs2_osb_dump(struct ocfs2_super * osb,char * buf,int len)223 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
224 {
225 	struct ocfs2_cluster_connection *cconn = osb->cconn;
226 	struct ocfs2_recovery_map *rm = osb->recovery_map;
227 	struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
228 	int i, out = 0;
229 
230 	out += snprintf(buf + out, len - out,
231 			"%10s => Id: %-s  Uuid: %-s  Gen: 0x%X  Label: %-s\n",
232 			"Device", osb->dev_str, osb->uuid_str,
233 			osb->fs_generation, osb->vol_label);
234 
235 	out += snprintf(buf + out, len - out,
236 			"%10s => State: %d  Flags: 0x%lX\n", "Volume",
237 			atomic_read(&osb->vol_state), osb->osb_flags);
238 
239 	out += snprintf(buf + out, len - out,
240 			"%10s => Block: %lu  Cluster: %d\n", "Sizes",
241 			osb->sb->s_blocksize, osb->s_clustersize);
242 
243 	out += snprintf(buf + out, len - out,
244 			"%10s => Compat: 0x%X  Incompat: 0x%X  "
245 			"ROcompat: 0x%X\n",
246 			"Features", osb->s_feature_compat,
247 			osb->s_feature_incompat, osb->s_feature_ro_compat);
248 
249 	out += snprintf(buf + out, len - out,
250 			"%10s => Opts: 0x%lX  AtimeQuanta: %u\n", "Mount",
251 			osb->s_mount_opt, osb->s_atime_quantum);
252 
253 	if (cconn) {
254 		out += snprintf(buf + out, len - out,
255 				"%10s => Stack: %s  Name: %*s  "
256 				"Version: %d.%d\n", "Cluster",
257 				(*osb->osb_cluster_stack == '\0' ?
258 				 "o2cb" : osb->osb_cluster_stack),
259 				cconn->cc_namelen, cconn->cc_name,
260 				cconn->cc_version.pv_major,
261 				cconn->cc_version.pv_minor);
262 	}
263 
264 	spin_lock(&osb->dc_task_lock);
265 	out += snprintf(buf + out, len - out,
266 			"%10s => Pid: %d  Count: %lu  WakeSeq: %lu  "
267 			"WorkSeq: %lu\n", "DownCnvt",
268 			(osb->dc_task ?  task_pid_nr(osb->dc_task) : -1),
269 			osb->blocked_lock_count, osb->dc_wake_sequence,
270 			osb->dc_work_sequence);
271 	spin_unlock(&osb->dc_task_lock);
272 
273 	spin_lock(&osb->osb_lock);
274 	out += snprintf(buf + out, len - out, "%10s => Pid: %d  Nodes:",
275 			"Recovery",
276 			(osb->recovery_thread_task ?
277 			 task_pid_nr(osb->recovery_thread_task) : -1));
278 	if (rm->rm_used == 0)
279 		out += snprintf(buf + out, len - out, " None\n");
280 	else {
281 		for (i = 0; i < rm->rm_used; i++)
282 			out += snprintf(buf + out, len - out, " %d",
283 					rm->rm_entries[i]);
284 		out += snprintf(buf + out, len - out, "\n");
285 	}
286 	spin_unlock(&osb->osb_lock);
287 
288 	out += snprintf(buf + out, len - out,
289 			"%10s => Pid: %d  Interval: %lu  Needs: %d\n", "Commit",
290 			(osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
291 			osb->osb_commit_interval,
292 			atomic_read(&osb->needs_checkpoint));
293 
294 	out += snprintf(buf + out, len - out,
295 			"%10s => State: %d  TxnId: %lu  NumTxns: %d\n",
296 			"Journal", osb->journal->j_state,
297 			osb->journal->j_trans_id,
298 			atomic_read(&osb->journal->j_num_trans));
299 
300 	out += snprintf(buf + out, len - out,
301 			"%10s => GlobalAllocs: %d  LocalAllocs: %d  "
302 			"SubAllocs: %d  LAWinMoves: %d  SAExtends: %d\n",
303 			"Stats",
304 			atomic_read(&osb->alloc_stats.bitmap_data),
305 			atomic_read(&osb->alloc_stats.local_data),
306 			atomic_read(&osb->alloc_stats.bg_allocs),
307 			atomic_read(&osb->alloc_stats.moves),
308 			atomic_read(&osb->alloc_stats.bg_extends));
309 
310 	out += snprintf(buf + out, len - out,
311 			"%10s => State: %u  Descriptor: %llu  Size: %u bits  "
312 			"Default: %u bits\n",
313 			"LocalAlloc", osb->local_alloc_state,
314 			(unsigned long long)osb->la_last_gd,
315 			osb->local_alloc_bits, osb->local_alloc_default_bits);
316 
317 	spin_lock(&osb->osb_lock);
318 	out += snprintf(buf + out, len - out,
319 			"%10s => InodeSlot: %d  StolenInodes: %d, "
320 			"MetaSlot: %d  StolenMeta: %d\n", "Steal",
321 			osb->s_inode_steal_slot,
322 			atomic_read(&osb->s_num_inodes_stolen),
323 			osb->s_meta_steal_slot,
324 			atomic_read(&osb->s_num_meta_stolen));
325 	spin_unlock(&osb->osb_lock);
326 
327 	out += snprintf(buf + out, len - out, "OrphanScan => ");
328 	out += snprintf(buf + out, len - out, "Local: %u  Global: %u ",
329 			os->os_count, os->os_seqno);
330 	out += snprintf(buf + out, len - out, " Last Scan: ");
331 	if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
332 		out += snprintf(buf + out, len - out, "Disabled\n");
333 	else
334 		out += snprintf(buf + out, len - out, "%lu seconds ago\n",
335 				(get_seconds() - os->os_scantime.tv_sec));
336 
337 	out += snprintf(buf + out, len - out, "%10s => %3s  %10s\n",
338 			"Slots", "Num", "RecoGen");
339 	for (i = 0; i < osb->max_slots; ++i) {
340 		out += snprintf(buf + out, len - out,
341 				"%10s  %c %3d  %10d\n",
342 				" ",
343 				(i == osb->slot_num ? '*' : ' '),
344 				i, osb->slot_recovery_generations[i]);
345 	}
346 
347 	return out;
348 }
349 
ocfs2_osb_debug_open(struct inode * inode,struct file * file)350 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
351 {
352 	struct ocfs2_super *osb = inode->i_private;
353 	char *buf = NULL;
354 
355 	buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
356 	if (!buf)
357 		goto bail;
358 
359 	i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
360 
361 	file->private_data = buf;
362 
363 	return 0;
364 bail:
365 	return -ENOMEM;
366 }
367 
ocfs2_debug_release(struct inode * inode,struct file * file)368 static int ocfs2_debug_release(struct inode *inode, struct file *file)
369 {
370 	kfree(file->private_data);
371 	return 0;
372 }
373 
ocfs2_debug_read(struct file * file,char __user * buf,size_t nbytes,loff_t * ppos)374 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
375 				size_t nbytes, loff_t *ppos)
376 {
377 	return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
378 				       i_size_read(file->f_mapping->host));
379 }
380 #else
ocfs2_osb_debug_open(struct inode * inode,struct file * file)381 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
382 {
383 	return 0;
384 }
ocfs2_debug_release(struct inode * inode,struct file * file)385 static int ocfs2_debug_release(struct inode *inode, struct file *file)
386 {
387 	return 0;
388 }
ocfs2_debug_read(struct file * file,char __user * buf,size_t nbytes,loff_t * ppos)389 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
390 				size_t nbytes, loff_t *ppos)
391 {
392 	return 0;
393 }
394 #endif	/* CONFIG_DEBUG_FS */
395 
396 static const struct file_operations ocfs2_osb_debug_fops = {
397 	.open =		ocfs2_osb_debug_open,
398 	.release =	ocfs2_debug_release,
399 	.read =		ocfs2_debug_read,
400 	.llseek =	generic_file_llseek,
401 };
402 
ocfs2_sync_fs(struct super_block * sb,int wait)403 static int ocfs2_sync_fs(struct super_block *sb, int wait)
404 {
405 	int status;
406 	tid_t target;
407 	struct ocfs2_super *osb = OCFS2_SB(sb);
408 
409 	if (ocfs2_is_hard_readonly(osb))
410 		return -EROFS;
411 
412 	if (wait) {
413 		status = ocfs2_flush_truncate_log(osb);
414 		if (status < 0)
415 			mlog_errno(status);
416 	} else {
417 		ocfs2_schedule_truncate_log_flush(osb, 0);
418 	}
419 
420 	if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
421 				      &target)) {
422 		if (wait)
423 			jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
424 					     target);
425 	}
426 	return 0;
427 }
428 
ocfs2_need_system_inode(struct ocfs2_super * osb,int ino)429 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
430 {
431 	if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
432 	    && (ino == USER_QUOTA_SYSTEM_INODE
433 		|| ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
434 		return 0;
435 	if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
436 	    && (ino == GROUP_QUOTA_SYSTEM_INODE
437 		|| ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
438 		return 0;
439 	return 1;
440 }
441 
ocfs2_init_global_system_inodes(struct ocfs2_super * osb)442 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
443 {
444 	struct inode *new = NULL;
445 	int status = 0;
446 	int i;
447 
448 	new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
449 	if (IS_ERR(new)) {
450 		status = PTR_ERR(new);
451 		mlog_errno(status);
452 		goto bail;
453 	}
454 	osb->root_inode = new;
455 
456 	new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
457 	if (IS_ERR(new)) {
458 		status = PTR_ERR(new);
459 		mlog_errno(status);
460 		goto bail;
461 	}
462 	osb->sys_root_inode = new;
463 
464 	for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
465 	     i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
466 		if (!ocfs2_need_system_inode(osb, i))
467 			continue;
468 		new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
469 		if (!new) {
470 			ocfs2_release_system_inodes(osb);
471 			status = -EINVAL;
472 			mlog_errno(status);
473 			/* FIXME: Should ERROR_RO_FS */
474 			mlog(ML_ERROR, "Unable to load system inode %d, "
475 			     "possibly corrupt fs?", i);
476 			goto bail;
477 		}
478 		// the array now has one ref, so drop this one
479 		iput(new);
480 	}
481 
482 bail:
483 	if (status)
484 		mlog_errno(status);
485 	return status;
486 }
487 
ocfs2_init_local_system_inodes(struct ocfs2_super * osb)488 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
489 {
490 	struct inode *new = NULL;
491 	int status = 0;
492 	int i;
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 	if (status)
513 		mlog_errno(status);
514 	return status;
515 }
516 
ocfs2_release_system_inodes(struct ocfs2_super * osb)517 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
518 {
519 	int i;
520 	struct inode *inode;
521 
522 	for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
523 		inode = osb->global_system_inodes[i];
524 		if (inode) {
525 			iput(inode);
526 			osb->global_system_inodes[i] = NULL;
527 		}
528 	}
529 
530 	inode = osb->sys_root_inode;
531 	if (inode) {
532 		iput(inode);
533 		osb->sys_root_inode = NULL;
534 	}
535 
536 	inode = osb->root_inode;
537 	if (inode) {
538 		iput(inode);
539 		osb->root_inode = NULL;
540 	}
541 
542 	if (!osb->local_system_inodes)
543 		return;
544 
545 	for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
546 		if (osb->local_system_inodes[i]) {
547 			iput(osb->local_system_inodes[i]);
548 			osb->local_system_inodes[i] = NULL;
549 		}
550 	}
551 
552 	kfree(osb->local_system_inodes);
553 	osb->local_system_inodes = NULL;
554 }
555 
556 /* We're allocating fs objects, use GFP_NOFS */
ocfs2_alloc_inode(struct super_block * sb)557 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
558 {
559 	struct ocfs2_inode_info *oi;
560 
561 	oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
562 	if (!oi)
563 		return NULL;
564 
565 	jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
566 	return &oi->vfs_inode;
567 }
568 
ocfs2_i_callback(struct rcu_head * head)569 static void ocfs2_i_callback(struct rcu_head *head)
570 {
571 	struct inode *inode = container_of(head, struct inode, i_rcu);
572 	kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
573 }
574 
ocfs2_destroy_inode(struct inode * inode)575 static void ocfs2_destroy_inode(struct inode *inode)
576 {
577 	call_rcu(&inode->i_rcu, ocfs2_i_callback);
578 }
579 
ocfs2_max_file_offset(unsigned int bbits,unsigned int cbits)580 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
581 						unsigned int cbits)
582 {
583 	unsigned int bytes = 1 << cbits;
584 	unsigned int trim = bytes;
585 	unsigned int bitshift = 32;
586 
587 	/*
588 	 * i_size and all block offsets in ocfs2 are always 64 bits
589 	 * wide. i_clusters is 32 bits, in cluster-sized units. So on
590 	 * 64 bit platforms, cluster size will be the limiting factor.
591 	 */
592 
593 #if BITS_PER_LONG == 32
594 # if defined(CONFIG_LBDAF)
595 	BUILD_BUG_ON(sizeof(sector_t) != 8);
596 	/*
597 	 * We might be limited by page cache size.
598 	 */
599 	if (bytes > PAGE_CACHE_SIZE) {
600 		bytes = PAGE_CACHE_SIZE;
601 		trim = 1;
602 		/*
603 		 * Shift by 31 here so that we don't get larger than
604 		 * MAX_LFS_FILESIZE
605 		 */
606 		bitshift = 31;
607 	}
608 # else
609 	/*
610 	 * We are limited by the size of sector_t. Use block size, as
611 	 * that's what we expose to the VFS.
612 	 */
613 	bytes = 1 << bbits;
614 	trim = 1;
615 	bitshift = 31;
616 # endif
617 #endif
618 
619 	/*
620 	 * Trim by a whole cluster when we can actually approach the
621 	 * on-disk limits. Otherwise we can overflow i_clusters when
622 	 * an extent start is at the max offset.
623 	 */
624 	return (((unsigned long long)bytes) << bitshift) - trim;
625 }
626 
ocfs2_remount(struct super_block * sb,int * flags,char * data)627 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
628 {
629 	int incompat_features;
630 	int ret = 0;
631 	struct mount_options parsed_options;
632 	struct ocfs2_super *osb = OCFS2_SB(sb);
633 	u32 tmp;
634 
635 	sync_filesystem(sb);
636 
637 	if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
638 	    !ocfs2_check_set_options(sb, &parsed_options)) {
639 		ret = -EINVAL;
640 		goto out;
641 	}
642 
643 	tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
644 		OCFS2_MOUNT_HB_NONE;
645 	if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
646 		ret = -EINVAL;
647 		mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
648 		goto out;
649 	}
650 
651 	if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
652 	    (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
653 		ret = -EINVAL;
654 		mlog(ML_ERROR, "Cannot change data mode on remount\n");
655 		goto out;
656 	}
657 
658 	/* Probably don't want this on remount; it might
659 	 * mess with other nodes */
660 	if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
661 	    (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
662 		ret = -EINVAL;
663 		mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
664 		goto out;
665 	}
666 
667 	/* We're going to/from readonly mode. */
668 	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
669 		/* Disable quota accounting before remounting RO */
670 		if (*flags & MS_RDONLY) {
671 			ret = ocfs2_susp_quotas(osb, 0);
672 			if (ret < 0)
673 				goto out;
674 		}
675 		/* Lock here so the check of HARD_RO and the potential
676 		 * setting of SOFT_RO is atomic. */
677 		spin_lock(&osb->osb_lock);
678 		if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
679 			mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
680 			ret = -EROFS;
681 			goto unlock_osb;
682 		}
683 
684 		if (*flags & MS_RDONLY) {
685 			sb->s_flags |= MS_RDONLY;
686 			osb->osb_flags |= OCFS2_OSB_SOFT_RO;
687 		} else {
688 			if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
689 				mlog(ML_ERROR, "Cannot remount RDWR "
690 				     "filesystem due to previous errors.\n");
691 				ret = -EROFS;
692 				goto unlock_osb;
693 			}
694 			incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
695 			if (incompat_features) {
696 				mlog(ML_ERROR, "Cannot remount RDWR because "
697 				     "of unsupported optional features "
698 				     "(%x).\n", incompat_features);
699 				ret = -EINVAL;
700 				goto unlock_osb;
701 			}
702 			sb->s_flags &= ~MS_RDONLY;
703 			osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
704 		}
705 		trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags);
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 	return ret;
743 }
744 
ocfs2_sb_probe(struct super_block * sb,struct buffer_head ** bh,int * sector_size,struct ocfs2_blockcheck_stats * stats)745 static int ocfs2_sb_probe(struct super_block *sb,
746 			  struct buffer_head **bh,
747 			  int *sector_size,
748 			  struct ocfs2_blockcheck_stats *stats)
749 {
750 	int status, tmpstat;
751 	struct ocfs1_vol_disk_hdr *hdr;
752 	struct ocfs2_dinode *di;
753 	int blksize;
754 
755 	*bh = NULL;
756 
757 	/* may be > 512 */
758 	*sector_size = bdev_logical_block_size(sb->s_bdev);
759 	if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
760 		mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
761 		     *sector_size, OCFS2_MAX_BLOCKSIZE);
762 		status = -EINVAL;
763 		goto bail;
764 	}
765 
766 	/* Can this really happen? */
767 	if (*sector_size < OCFS2_MIN_BLOCKSIZE)
768 		*sector_size = OCFS2_MIN_BLOCKSIZE;
769 
770 	/* check block zero for old format */
771 	status = ocfs2_get_sector(sb, bh, 0, *sector_size);
772 	if (status < 0) {
773 		mlog_errno(status);
774 		goto bail;
775 	}
776 	hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
777 	if (hdr->major_version == OCFS1_MAJOR_VERSION) {
778 		mlog(ML_ERROR, "incompatible version: %u.%u\n",
779 		     hdr->major_version, hdr->minor_version);
780 		status = -EINVAL;
781 	}
782 	if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
783 		   strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
784 		mlog(ML_ERROR, "incompatible volume signature: %8s\n",
785 		     hdr->signature);
786 		status = -EINVAL;
787 	}
788 	brelse(*bh);
789 	*bh = NULL;
790 	if (status < 0) {
791 		mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
792 		     "upgraded before mounting with ocfs v2\n");
793 		goto bail;
794 	}
795 
796 	/*
797 	 * Now check at magic offset for 512, 1024, 2048, 4096
798 	 * blocksizes.  4096 is the maximum blocksize because it is
799 	 * the minimum clustersize.
800 	 */
801 	status = -EINVAL;
802 	for (blksize = *sector_size;
803 	     blksize <= OCFS2_MAX_BLOCKSIZE;
804 	     blksize <<= 1) {
805 		tmpstat = ocfs2_get_sector(sb, bh,
806 					   OCFS2_SUPER_BLOCK_BLKNO,
807 					   blksize);
808 		if (tmpstat < 0) {
809 			status = tmpstat;
810 			mlog_errno(status);
811 			break;
812 		}
813 		di = (struct ocfs2_dinode *) (*bh)->b_data;
814 		memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
815 		spin_lock_init(&stats->b_lock);
816 		tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
817 		if (tmpstat < 0) {
818 			brelse(*bh);
819 			*bh = NULL;
820 		}
821 		if (tmpstat != -EAGAIN) {
822 			status = tmpstat;
823 			break;
824 		}
825 	}
826 
827 bail:
828 	return status;
829 }
830 
ocfs2_verify_heartbeat(struct ocfs2_super * osb)831 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
832 {
833 	u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
834 
835 	if (osb->s_mount_opt & hb_enabled) {
836 		if (ocfs2_mount_local(osb)) {
837 			mlog(ML_ERROR, "Cannot heartbeat on a locally "
838 			     "mounted device.\n");
839 			return -EINVAL;
840 		}
841 		if (ocfs2_userspace_stack(osb)) {
842 			mlog(ML_ERROR, "Userspace stack expected, but "
843 			     "o2cb heartbeat arguments passed to mount\n");
844 			return -EINVAL;
845 		}
846 		if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
847 		     !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
848 		    ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
849 		     ocfs2_cluster_o2cb_global_heartbeat(osb))) {
850 			mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
851 			return -EINVAL;
852 		}
853 	}
854 
855 	if (!(osb->s_mount_opt & hb_enabled)) {
856 		if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
857 		    !ocfs2_userspace_stack(osb)) {
858 			mlog(ML_ERROR, "Heartbeat has to be started to mount "
859 			     "a read-write clustered device.\n");
860 			return -EINVAL;
861 		}
862 	}
863 
864 	return 0;
865 }
866 
867 /*
868  * If we're using a userspace stack, mount should have passed
869  * a name that matches the disk.  If not, mount should not
870  * have passed a stack.
871  */
ocfs2_verify_userspace_stack(struct ocfs2_super * osb,struct mount_options * mopt)872 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
873 					struct mount_options *mopt)
874 {
875 	if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
876 		mlog(ML_ERROR,
877 		     "cluster stack passed to mount, but this filesystem "
878 		     "does not support it\n");
879 		return -EINVAL;
880 	}
881 
882 	if (ocfs2_userspace_stack(osb) &&
883 	    strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
884 		    OCFS2_STACK_LABEL_LEN)) {
885 		mlog(ML_ERROR,
886 		     "cluster stack passed to mount (\"%s\") does not "
887 		     "match the filesystem (\"%s\")\n",
888 		     mopt->cluster_stack,
889 		     osb->osb_cluster_stack);
890 		return -EINVAL;
891 	}
892 
893 	return 0;
894 }
895 
ocfs2_susp_quotas(struct ocfs2_super * osb,int unsuspend)896 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
897 {
898 	int type;
899 	struct super_block *sb = osb->sb;
900 	unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
901 					     OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
902 	int status = 0;
903 
904 	for (type = 0; type < MAXQUOTAS; type++) {
905 		if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
906 			continue;
907 		if (unsuspend)
908 			status = dquot_resume(sb, type);
909 		else {
910 			struct ocfs2_mem_dqinfo *oinfo;
911 
912 			/* Cancel periodic syncing before suspending */
913 			oinfo = sb_dqinfo(sb, type)->dqi_priv;
914 			cancel_delayed_work_sync(&oinfo->dqi_sync_work);
915 			status = dquot_suspend(sb, type);
916 		}
917 		if (status < 0)
918 			break;
919 	}
920 	if (status < 0)
921 		mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
922 		     "remount (error = %d).\n", status);
923 	return status;
924 }
925 
ocfs2_enable_quotas(struct ocfs2_super * osb)926 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
927 {
928 	struct inode *inode[MAXQUOTAS] = { NULL, NULL };
929 	struct super_block *sb = osb->sb;
930 	unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
931 					     OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
932 	unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
933 					LOCAL_GROUP_QUOTA_SYSTEM_INODE };
934 	int status;
935 	int type;
936 
937 	sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
938 	for (type = 0; type < MAXQUOTAS; type++) {
939 		if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
940 			continue;
941 		inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
942 							osb->slot_num);
943 		if (!inode[type]) {
944 			status = -ENOENT;
945 			goto out_quota_off;
946 		}
947 		status = dquot_enable(inode[type], type, QFMT_OCFS2,
948 				      DQUOT_USAGE_ENABLED);
949 		if (status < 0)
950 			goto out_quota_off;
951 	}
952 
953 	for (type = 0; type < MAXQUOTAS; type++)
954 		iput(inode[type]);
955 	return 0;
956 out_quota_off:
957 	ocfs2_disable_quotas(osb);
958 	for (type = 0; type < MAXQUOTAS; type++)
959 		iput(inode[type]);
960 	mlog_errno(status);
961 	return status;
962 }
963 
ocfs2_disable_quotas(struct ocfs2_super * osb)964 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
965 {
966 	int type;
967 	struct inode *inode;
968 	struct super_block *sb = osb->sb;
969 	struct ocfs2_mem_dqinfo *oinfo;
970 
971 	/* We mostly ignore errors in this function because there's not much
972 	 * we can do when we see them */
973 	for (type = 0; type < MAXQUOTAS; type++) {
974 		if (!sb_has_quota_loaded(sb, type))
975 			continue;
976 		/* Cancel periodic syncing before we grab dqonoff_mutex */
977 		oinfo = sb_dqinfo(sb, type)->dqi_priv;
978 		cancel_delayed_work_sync(&oinfo->dqi_sync_work);
979 		inode = igrab(sb->s_dquot.files[type]);
980 		/* Turn off quotas. This will remove all dquot structures from
981 		 * memory and so they will be automatically synced to global
982 		 * quota files */
983 		dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
984 					DQUOT_LIMITS_ENABLED);
985 		if (!inode)
986 			continue;
987 		iput(inode);
988 	}
989 }
990 
991 /* Handle quota on quotactl */
ocfs2_quota_on(struct super_block * sb,int type,int format_id)992 static int ocfs2_quota_on(struct super_block *sb, int type, int format_id)
993 {
994 	unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
995 					     OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
996 
997 	if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
998 		return -EINVAL;
999 
1000 	return dquot_enable(sb_dqopt(sb)->files[type], type,
1001 			    format_id, DQUOT_LIMITS_ENABLED);
1002 }
1003 
1004 /* Handle quota off quotactl */
ocfs2_quota_off(struct super_block * sb,int type)1005 static int ocfs2_quota_off(struct super_block *sb, int type)
1006 {
1007 	return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
1008 }
1009 
1010 static const struct quotactl_ops ocfs2_quotactl_ops = {
1011 	.quota_on_meta	= ocfs2_quota_on,
1012 	.quota_off	= ocfs2_quota_off,
1013 	.quota_sync	= dquot_quota_sync,
1014 	.get_info	= dquot_get_dqinfo,
1015 	.set_info	= dquot_set_dqinfo,
1016 	.get_dqblk	= dquot_get_dqblk,
1017 	.set_dqblk	= dquot_set_dqblk,
1018 };
1019 
ocfs2_fill_super(struct super_block * sb,void * data,int silent)1020 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
1021 {
1022 	struct dentry *root;
1023 	int status, sector_size;
1024 	struct mount_options parsed_options;
1025 	struct inode *inode = NULL;
1026 	struct ocfs2_super *osb = NULL;
1027 	struct buffer_head *bh = NULL;
1028 	char nodestr[8];
1029 	struct ocfs2_blockcheck_stats stats;
1030 
1031 	trace_ocfs2_fill_super(sb, data, silent);
1032 
1033 	if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
1034 		status = -EINVAL;
1035 		goto read_super_error;
1036 	}
1037 
1038 	/* probe for superblock */
1039 	status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
1040 	if (status < 0) {
1041 		mlog(ML_ERROR, "superblock probe failed!\n");
1042 		goto read_super_error;
1043 	}
1044 
1045 	status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1046 	osb = OCFS2_SB(sb);
1047 	if (status < 0) {
1048 		mlog_errno(status);
1049 		goto read_super_error;
1050 	}
1051 	brelse(bh);
1052 	bh = NULL;
1053 
1054 	if (!ocfs2_check_set_options(sb, &parsed_options)) {
1055 		status = -EINVAL;
1056 		goto read_super_error;
1057 	}
1058 	osb->s_mount_opt = parsed_options.mount_opt;
1059 	osb->s_atime_quantum = parsed_options.atime_quantum;
1060 	osb->preferred_slot = parsed_options.slot;
1061 	osb->osb_commit_interval = parsed_options.commit_interval;
1062 
1063 	ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1064 	osb->osb_resv_level = parsed_options.resv_level;
1065 	osb->osb_dir_resv_level = parsed_options.resv_level;
1066 	if (parsed_options.dir_resv_level == -1)
1067 		osb->osb_dir_resv_level = parsed_options.resv_level;
1068 	else
1069 		osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1070 
1071 	status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1072 	if (status)
1073 		goto read_super_error;
1074 
1075 	sb->s_magic = OCFS2_SUPER_MAGIC;
1076 
1077 	sb->s_flags = (sb->s_flags & ~(MS_POSIXACL | MS_NOSEC)) |
1078 		((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1079 
1080 	/* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1081 	 * heartbeat=none */
1082 	if (bdev_read_only(sb->s_bdev)) {
1083 		if (!(sb->s_flags & MS_RDONLY)) {
1084 			status = -EACCES;
1085 			mlog(ML_ERROR, "Readonly device detected but readonly "
1086 			     "mount was not specified.\n");
1087 			goto read_super_error;
1088 		}
1089 
1090 		/* You should not be able to start a local heartbeat
1091 		 * on a readonly device. */
1092 		if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1093 			status = -EROFS;
1094 			mlog(ML_ERROR, "Local heartbeat specified on readonly "
1095 			     "device.\n");
1096 			goto read_super_error;
1097 		}
1098 
1099 		status = ocfs2_check_journals_nolocks(osb);
1100 		if (status < 0) {
1101 			if (status == -EROFS)
1102 				mlog(ML_ERROR, "Recovery required on readonly "
1103 				     "file system, but write access is "
1104 				     "unavailable.\n");
1105 			else
1106 				mlog_errno(status);
1107 			goto read_super_error;
1108 		}
1109 
1110 		ocfs2_set_ro_flag(osb, 1);
1111 
1112 		printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. "
1113 		       "Cluster services will not be used for this mount. "
1114 		       "Recovery will be skipped.\n", osb->dev_str);
1115 	}
1116 
1117 	if (!ocfs2_is_hard_readonly(osb)) {
1118 		if (sb->s_flags & MS_RDONLY)
1119 			ocfs2_set_ro_flag(osb, 0);
1120 	}
1121 
1122 	status = ocfs2_verify_heartbeat(osb);
1123 	if (status < 0) {
1124 		mlog_errno(status);
1125 		goto read_super_error;
1126 	}
1127 
1128 	osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1129 						 ocfs2_debugfs_root);
1130 	if (!osb->osb_debug_root) {
1131 		status = -EINVAL;
1132 		mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1133 		goto read_super_error;
1134 	}
1135 
1136 	osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1137 					    osb->osb_debug_root,
1138 					    osb,
1139 					    &ocfs2_osb_debug_fops);
1140 	if (!osb->osb_ctxt) {
1141 		status = -EINVAL;
1142 		mlog_errno(status);
1143 		goto read_super_error;
1144 	}
1145 
1146 	if (ocfs2_meta_ecc(osb)) {
1147 		status = ocfs2_blockcheck_stats_debugfs_install(
1148 						&osb->osb_ecc_stats,
1149 						osb->osb_debug_root);
1150 		if (status) {
1151 			mlog(ML_ERROR,
1152 			     "Unable to create blockcheck statistics "
1153 			     "files\n");
1154 			goto read_super_error;
1155 		}
1156 	}
1157 
1158 	status = ocfs2_mount_volume(sb);
1159 	if (status < 0)
1160 		goto read_super_error;
1161 
1162 	if (osb->root_inode)
1163 		inode = igrab(osb->root_inode);
1164 
1165 	if (!inode) {
1166 		status = -EIO;
1167 		mlog_errno(status);
1168 		goto read_super_error;
1169 	}
1170 
1171 	root = d_make_root(inode);
1172 	if (!root) {
1173 		status = -ENOMEM;
1174 		mlog_errno(status);
1175 		goto read_super_error;
1176 	}
1177 
1178 	sb->s_root = root;
1179 
1180 	ocfs2_complete_mount_recovery(osb);
1181 
1182 	if (ocfs2_mount_local(osb))
1183 		snprintf(nodestr, sizeof(nodestr), "local");
1184 	else
1185 		snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1186 
1187 	printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1188 	       "with %s data mode.\n",
1189 	       osb->dev_str, nodestr, osb->slot_num,
1190 	       osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1191 	       "ordered");
1192 
1193 	atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1194 	wake_up(&osb->osb_mount_event);
1195 
1196 	/* Now we can initialize quotas because we can afford to wait
1197 	 * for cluster locks recovery now. That also means that truncation
1198 	 * log recovery can happen but that waits for proper quota setup */
1199 	if (!(sb->s_flags & MS_RDONLY)) {
1200 		status = ocfs2_enable_quotas(osb);
1201 		if (status < 0) {
1202 			/* We have to err-out specially here because
1203 			 * s_root is already set */
1204 			mlog_errno(status);
1205 			atomic_set(&osb->vol_state, VOLUME_DISABLED);
1206 			wake_up(&osb->osb_mount_event);
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 	return status;
1221 
1222 read_super_error:
1223 	brelse(bh);
1224 
1225 	if (osb) {
1226 		atomic_set(&osb->vol_state, VOLUME_DISABLED);
1227 		wake_up(&osb->osb_mount_event);
1228 		ocfs2_dismount_volume(sb, 1);
1229 	}
1230 
1231 	if (status)
1232 		mlog_errno(status);
1233 	return status;
1234 }
1235 
ocfs2_mount(struct file_system_type * fs_type,int flags,const char * dev_name,void * data)1236 static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
1237 			int flags,
1238 			const char *dev_name,
1239 			void *data)
1240 {
1241 	return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
1242 }
1243 
ocfs2_kill_sb(struct super_block * sb)1244 static void ocfs2_kill_sb(struct super_block *sb)
1245 {
1246 	struct ocfs2_super *osb = OCFS2_SB(sb);
1247 
1248 	/* Failed mount? */
1249 	if (!osb || atomic_read(&osb->vol_state) == VOLUME_DISABLED)
1250 		goto out;
1251 
1252 	/* Prevent further queueing of inode drop events */
1253 	spin_lock(&dentry_list_lock);
1254 	ocfs2_set_osb_flag(osb, OCFS2_OSB_DROP_DENTRY_LOCK_IMMED);
1255 	spin_unlock(&dentry_list_lock);
1256 	/* Wait for work to finish and/or remove it */
1257 	cancel_work_sync(&osb->dentry_lock_work);
1258 out:
1259 	kill_block_super(sb);
1260 }
1261 
1262 static struct file_system_type ocfs2_fs_type = {
1263 	.owner          = THIS_MODULE,
1264 	.name           = "ocfs2",
1265 	.mount          = ocfs2_mount,
1266 	.kill_sb        = ocfs2_kill_sb,
1267 
1268 	.fs_flags       = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1269 	.next           = NULL
1270 };
1271 MODULE_ALIAS_FS("ocfs2");
1272 
ocfs2_check_set_options(struct super_block * sb,struct mount_options * options)1273 static int ocfs2_check_set_options(struct super_block *sb,
1274 				   struct mount_options *options)
1275 {
1276 	if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1277 	    !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1278 					 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1279 		mlog(ML_ERROR, "User quotas were requested, but this "
1280 		     "filesystem does not have the feature enabled.\n");
1281 		return 0;
1282 	}
1283 	if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1284 	    !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1285 					 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1286 		mlog(ML_ERROR, "Group quotas were requested, but this "
1287 		     "filesystem does not have the feature enabled.\n");
1288 		return 0;
1289 	}
1290 	if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1291 	    !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1292 		mlog(ML_ERROR, "ACL support requested but extended attributes "
1293 		     "feature is not enabled\n");
1294 		return 0;
1295 	}
1296 	/* No ACL setting specified? Use XATTR feature... */
1297 	if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1298 				    OCFS2_MOUNT_NO_POSIX_ACL))) {
1299 		if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1300 			options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1301 		else
1302 			options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1303 	}
1304 	return 1;
1305 }
1306 
ocfs2_parse_options(struct super_block * sb,char * options,struct mount_options * mopt,int is_remount)1307 static int ocfs2_parse_options(struct super_block *sb,
1308 			       char *options,
1309 			       struct mount_options *mopt,
1310 			       int is_remount)
1311 {
1312 	int status, user_stack = 0;
1313 	char *p;
1314 	u32 tmp;
1315 
1316 	trace_ocfs2_parse_options(is_remount, options ? options : "(none)");
1317 
1318 	mopt->commit_interval = 0;
1319 	mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1320 	mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1321 	mopt->slot = OCFS2_INVALID_SLOT;
1322 	mopt->localalloc_opt = -1;
1323 	mopt->cluster_stack[0] = '\0';
1324 	mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1325 	mopt->dir_resv_level = -1;
1326 
1327 	if (!options) {
1328 		status = 1;
1329 		goto bail;
1330 	}
1331 
1332 	while ((p = strsep(&options, ",")) != NULL) {
1333 		int token, option;
1334 		substring_t args[MAX_OPT_ARGS];
1335 
1336 		if (!*p)
1337 			continue;
1338 
1339 		token = match_token(p, tokens, args);
1340 		switch (token) {
1341 		case Opt_hb_local:
1342 			mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1343 			break;
1344 		case Opt_hb_none:
1345 			mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1346 			break;
1347 		case Opt_hb_global:
1348 			mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1349 			break;
1350 		case Opt_barrier:
1351 			if (match_int(&args[0], &option)) {
1352 				status = 0;
1353 				goto bail;
1354 			}
1355 			if (option)
1356 				mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1357 			else
1358 				mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1359 			break;
1360 		case Opt_intr:
1361 			mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1362 			break;
1363 		case Opt_nointr:
1364 			mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1365 			break;
1366 		case Opt_err_panic:
1367 			mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1368 			break;
1369 		case Opt_err_ro:
1370 			mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1371 			break;
1372 		case Opt_data_ordered:
1373 			mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1374 			break;
1375 		case Opt_data_writeback:
1376 			mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1377 			break;
1378 		case Opt_user_xattr:
1379 			mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1380 			break;
1381 		case Opt_nouser_xattr:
1382 			mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1383 			break;
1384 		case Opt_atime_quantum:
1385 			if (match_int(&args[0], &option)) {
1386 				status = 0;
1387 				goto bail;
1388 			}
1389 			if (option >= 0)
1390 				mopt->atime_quantum = option;
1391 			break;
1392 		case Opt_slot:
1393 			option = 0;
1394 			if (match_int(&args[0], &option)) {
1395 				status = 0;
1396 				goto bail;
1397 			}
1398 			if (option)
1399 				mopt->slot = (s16)option;
1400 			break;
1401 		case Opt_commit:
1402 			option = 0;
1403 			if (match_int(&args[0], &option)) {
1404 				status = 0;
1405 				goto bail;
1406 			}
1407 			if (option < 0)
1408 				return 0;
1409 			if (option == 0)
1410 				option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1411 			mopt->commit_interval = HZ * option;
1412 			break;
1413 		case Opt_localalloc:
1414 			option = 0;
1415 			if (match_int(&args[0], &option)) {
1416 				status = 0;
1417 				goto bail;
1418 			}
1419 			if (option >= 0)
1420 				mopt->localalloc_opt = option;
1421 			break;
1422 		case Opt_localflocks:
1423 			/*
1424 			 * Changing this during remount could race
1425 			 * flock() requests, or "unbalance" existing
1426 			 * ones (e.g., a lock is taken in one mode but
1427 			 * dropped in the other). If users care enough
1428 			 * to flip locking modes during remount, we
1429 			 * could add a "local" flag to individual
1430 			 * flock structures for proper tracking of
1431 			 * state.
1432 			 */
1433 			if (!is_remount)
1434 				mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1435 			break;
1436 		case Opt_stack:
1437 			/* Check both that the option we were passed
1438 			 * is of the right length and that it is a proper
1439 			 * string of the right length.
1440 			 */
1441 			if (((args[0].to - args[0].from) !=
1442 			     OCFS2_STACK_LABEL_LEN) ||
1443 			    (strnlen(args[0].from,
1444 				     OCFS2_STACK_LABEL_LEN) !=
1445 			     OCFS2_STACK_LABEL_LEN)) {
1446 				mlog(ML_ERROR,
1447 				     "Invalid cluster_stack option\n");
1448 				status = 0;
1449 				goto bail;
1450 			}
1451 			memcpy(mopt->cluster_stack, args[0].from,
1452 			       OCFS2_STACK_LABEL_LEN);
1453 			mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1454 			/*
1455 			 * Open code the memcmp here as we don't have
1456 			 * an osb to pass to
1457 			 * ocfs2_userspace_stack().
1458 			 */
1459 			if (memcmp(mopt->cluster_stack,
1460 				   OCFS2_CLASSIC_CLUSTER_STACK,
1461 				   OCFS2_STACK_LABEL_LEN))
1462 				user_stack = 1;
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 	if (user_stack == 0) {
1519 		/* Ensure only one heartbeat mode */
1520 		tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1521 					 OCFS2_MOUNT_HB_GLOBAL |
1522 					 OCFS2_MOUNT_HB_NONE);
1523 		if (hweight32(tmp) != 1) {
1524 			mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1525 			status = 0;
1526 			goto bail;
1527 		}
1528 	}
1529 
1530 	status = 1;
1531 
1532 bail:
1533 	return status;
1534 }
1535 
ocfs2_show_options(struct seq_file * s,struct dentry * root)1536 static int ocfs2_show_options(struct seq_file *s, struct dentry *root)
1537 {
1538 	struct ocfs2_super *osb = OCFS2_SB(root->d_sb);
1539 	unsigned long opts = osb->s_mount_opt;
1540 	unsigned int local_alloc_megs;
1541 
1542 	if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1543 		seq_printf(s, ",_netdev");
1544 		if (opts & OCFS2_MOUNT_HB_LOCAL)
1545 			seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1546 		else
1547 			seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1548 	} else
1549 		seq_printf(s, ",%s", OCFS2_HB_NONE);
1550 
1551 	if (opts & OCFS2_MOUNT_NOINTR)
1552 		seq_printf(s, ",nointr");
1553 
1554 	if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1555 		seq_printf(s, ",data=writeback");
1556 	else
1557 		seq_printf(s, ",data=ordered");
1558 
1559 	if (opts & OCFS2_MOUNT_BARRIER)
1560 		seq_printf(s, ",barrier=1");
1561 
1562 	if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1563 		seq_printf(s, ",errors=panic");
1564 	else
1565 		seq_printf(s, ",errors=remount-ro");
1566 
1567 	if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1568 		seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1569 
1570 	seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1571 
1572 	if (osb->osb_commit_interval)
1573 		seq_printf(s, ",commit=%u",
1574 			   (unsigned) (osb->osb_commit_interval / HZ));
1575 
1576 	local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1577 	if (local_alloc_megs != ocfs2_la_default_mb(osb))
1578 		seq_printf(s, ",localalloc=%d", local_alloc_megs);
1579 
1580 	if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1581 		seq_printf(s, ",localflocks,");
1582 
1583 	if (osb->osb_cluster_stack[0])
1584 		seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1585 			   osb->osb_cluster_stack);
1586 	if (opts & OCFS2_MOUNT_USRQUOTA)
1587 		seq_printf(s, ",usrquota");
1588 	if (opts & OCFS2_MOUNT_GRPQUOTA)
1589 		seq_printf(s, ",grpquota");
1590 
1591 	if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1592 		seq_printf(s, ",coherency=buffered");
1593 	else
1594 		seq_printf(s, ",coherency=full");
1595 
1596 	if (opts & OCFS2_MOUNT_NOUSERXATTR)
1597 		seq_printf(s, ",nouser_xattr");
1598 	else
1599 		seq_printf(s, ",user_xattr");
1600 
1601 	if (opts & OCFS2_MOUNT_INODE64)
1602 		seq_printf(s, ",inode64");
1603 
1604 	if (opts & OCFS2_MOUNT_POSIX_ACL)
1605 		seq_printf(s, ",acl");
1606 	else
1607 		seq_printf(s, ",noacl");
1608 
1609 	if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1610 		seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1611 
1612 	if (osb->osb_dir_resv_level != osb->osb_resv_level)
1613 		seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1614 
1615 	return 0;
1616 }
1617 
1618 wait_queue_head_t ocfs2__ioend_wq[OCFS2_IOEND_WQ_HASH_SZ];
1619 
ocfs2_init(void)1620 static int __init ocfs2_init(void)
1621 {
1622 	int status, i;
1623 
1624 	ocfs2_print_version();
1625 
1626 	for (i = 0; i < OCFS2_IOEND_WQ_HASH_SZ; i++)
1627 		init_waitqueue_head(&ocfs2__ioend_wq[i]);
1628 
1629 	status = init_ocfs2_uptodate_cache();
1630 	if (status < 0)
1631 		goto out1;
1632 
1633 	status = ocfs2_initialize_mem_caches();
1634 	if (status < 0)
1635 		goto out2;
1636 
1637 	ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1638 	if (!ocfs2_wq) {
1639 		status = -ENOMEM;
1640 		goto out3;
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 	ocfs2_set_locking_protocol();
1650 
1651 	status = register_quota_format(&ocfs2_quota_format);
1652 	if (status < 0)
1653 		goto out4;
1654 	status = register_filesystem(&ocfs2_fs_type);
1655 	if (!status)
1656 		return 0;
1657 
1658 	unregister_quota_format(&ocfs2_quota_format);
1659 out4:
1660 	destroy_workqueue(ocfs2_wq);
1661 	debugfs_remove(ocfs2_debugfs_root);
1662 out3:
1663 	ocfs2_free_mem_caches();
1664 out2:
1665 	exit_ocfs2_uptodate_cache();
1666 out1:
1667 	mlog_errno(status);
1668 	return status;
1669 }
1670 
ocfs2_exit(void)1671 static void __exit ocfs2_exit(void)
1672 {
1673 	if (ocfs2_wq) {
1674 		flush_workqueue(ocfs2_wq);
1675 		destroy_workqueue(ocfs2_wq);
1676 	}
1677 
1678 	unregister_quota_format(&ocfs2_quota_format);
1679 
1680 	debugfs_remove(ocfs2_debugfs_root);
1681 
1682 	ocfs2_free_mem_caches();
1683 
1684 	unregister_filesystem(&ocfs2_fs_type);
1685 
1686 	exit_ocfs2_uptodate_cache();
1687 }
1688 
ocfs2_put_super(struct super_block * sb)1689 static void ocfs2_put_super(struct super_block *sb)
1690 {
1691 	trace_ocfs2_put_super(sb);
1692 
1693 	ocfs2_sync_blockdev(sb);
1694 	ocfs2_dismount_volume(sb, 0);
1695 }
1696 
ocfs2_statfs(struct dentry * dentry,struct kstatfs * buf)1697 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1698 {
1699 	struct ocfs2_super *osb;
1700 	u32 numbits, freebits;
1701 	int status;
1702 	struct ocfs2_dinode *bm_lock;
1703 	struct buffer_head *bh = NULL;
1704 	struct inode *inode = NULL;
1705 
1706 	trace_ocfs2_statfs(dentry->d_sb, buf);
1707 
1708 	osb = OCFS2_SB(dentry->d_sb);
1709 
1710 	inode = ocfs2_get_system_file_inode(osb,
1711 					    GLOBAL_BITMAP_SYSTEM_INODE,
1712 					    OCFS2_INVALID_SLOT);
1713 	if (!inode) {
1714 		mlog(ML_ERROR, "failed to get bitmap inode\n");
1715 		status = -EIO;
1716 		goto bail;
1717 	}
1718 
1719 	status = ocfs2_inode_lock(inode, &bh, 0);
1720 	if (status < 0) {
1721 		mlog_errno(status);
1722 		goto bail;
1723 	}
1724 
1725 	bm_lock = (struct ocfs2_dinode *) bh->b_data;
1726 
1727 	numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1728 	freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1729 
1730 	buf->f_type = OCFS2_SUPER_MAGIC;
1731 	buf->f_bsize = dentry->d_sb->s_blocksize;
1732 	buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1733 	buf->f_blocks = ((sector_t) numbits) *
1734 			(osb->s_clustersize >> osb->sb->s_blocksize_bits);
1735 	buf->f_bfree = ((sector_t) freebits) *
1736 		       (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1737 	buf->f_bavail = buf->f_bfree;
1738 	buf->f_files = numbits;
1739 	buf->f_ffree = freebits;
1740 	buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1741 				& 0xFFFFFFFFUL;
1742 	buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1743 				OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1744 
1745 	brelse(bh);
1746 
1747 	ocfs2_inode_unlock(inode, 0);
1748 	status = 0;
1749 bail:
1750 	if (inode)
1751 		iput(inode);
1752 
1753 	if (status)
1754 		mlog_errno(status);
1755 
1756 	return status;
1757 }
1758 
ocfs2_inode_init_once(void * data)1759 static void ocfs2_inode_init_once(void *data)
1760 {
1761 	struct ocfs2_inode_info *oi = data;
1762 
1763 	oi->ip_flags = 0;
1764 	oi->ip_open_count = 0;
1765 	spin_lock_init(&oi->ip_lock);
1766 	ocfs2_extent_map_init(&oi->vfs_inode);
1767 	INIT_LIST_HEAD(&oi->ip_io_markers);
1768 	oi->ip_dir_start_lookup = 0;
1769 	atomic_set(&oi->ip_unaligned_aio, 0);
1770 	init_rwsem(&oi->ip_alloc_sem);
1771 	init_rwsem(&oi->ip_xattr_sem);
1772 	mutex_init(&oi->ip_io_mutex);
1773 
1774 	oi->ip_blkno = 0ULL;
1775 	oi->ip_clusters = 0;
1776 
1777 	ocfs2_resv_init_once(&oi->ip_la_data_resv);
1778 
1779 	ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1780 	ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1781 	ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1782 
1783 	ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1784 				  &ocfs2_inode_caching_ops);
1785 
1786 	inode_init_once(&oi->vfs_inode);
1787 }
1788 
ocfs2_initialize_mem_caches(void)1789 static int ocfs2_initialize_mem_caches(void)
1790 {
1791 	ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1792 				       sizeof(struct ocfs2_inode_info),
1793 				       0,
1794 				       (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1795 						SLAB_MEM_SPREAD),
1796 				       ocfs2_inode_init_once);
1797 	ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1798 					sizeof(struct ocfs2_dquot),
1799 					0,
1800 					(SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1801 						SLAB_MEM_SPREAD),
1802 					NULL);
1803 	ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1804 					sizeof(struct ocfs2_quota_chunk),
1805 					0,
1806 					(SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1807 					NULL);
1808 	if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1809 	    !ocfs2_qf_chunk_cachep) {
1810 		if (ocfs2_inode_cachep)
1811 			kmem_cache_destroy(ocfs2_inode_cachep);
1812 		if (ocfs2_dquot_cachep)
1813 			kmem_cache_destroy(ocfs2_dquot_cachep);
1814 		if (ocfs2_qf_chunk_cachep)
1815 			kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1816 		return -ENOMEM;
1817 	}
1818 
1819 	return 0;
1820 }
1821 
ocfs2_free_mem_caches(void)1822 static void ocfs2_free_mem_caches(void)
1823 {
1824 	/*
1825 	 * Make sure all delayed rcu free inodes are flushed before we
1826 	 * destroy cache.
1827 	 */
1828 	rcu_barrier();
1829 	if (ocfs2_inode_cachep)
1830 		kmem_cache_destroy(ocfs2_inode_cachep);
1831 	ocfs2_inode_cachep = NULL;
1832 
1833 	if (ocfs2_dquot_cachep)
1834 		kmem_cache_destroy(ocfs2_dquot_cachep);
1835 	ocfs2_dquot_cachep = NULL;
1836 
1837 	if (ocfs2_qf_chunk_cachep)
1838 		kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1839 	ocfs2_qf_chunk_cachep = NULL;
1840 }
1841 
ocfs2_get_sector(struct super_block * sb,struct buffer_head ** bh,int block,int sect_size)1842 static int ocfs2_get_sector(struct super_block *sb,
1843 			    struct buffer_head **bh,
1844 			    int block,
1845 			    int sect_size)
1846 {
1847 	if (!sb_set_blocksize(sb, sect_size)) {
1848 		mlog(ML_ERROR, "unable to set blocksize\n");
1849 		return -EIO;
1850 	}
1851 
1852 	*bh = sb_getblk(sb, block);
1853 	if (!*bh) {
1854 		mlog_errno(-EIO);
1855 		return -EIO;
1856 	}
1857 	lock_buffer(*bh);
1858 	if (!buffer_dirty(*bh))
1859 		clear_buffer_uptodate(*bh);
1860 	unlock_buffer(*bh);
1861 	ll_rw_block(READ, 1, bh);
1862 	wait_on_buffer(*bh);
1863 	if (!buffer_uptodate(*bh)) {
1864 		mlog_errno(-EIO);
1865 		brelse(*bh);
1866 		*bh = NULL;
1867 		return -EIO;
1868 	}
1869 
1870 	return 0;
1871 }
1872 
ocfs2_mount_volume(struct super_block * sb)1873 static int ocfs2_mount_volume(struct super_block *sb)
1874 {
1875 	int status = 0;
1876 	int unlock_super = 0;
1877 	struct ocfs2_super *osb = OCFS2_SB(sb);
1878 
1879 	if (ocfs2_is_hard_readonly(osb))
1880 		goto leave;
1881 
1882 	status = ocfs2_dlm_init(osb);
1883 	if (status < 0) {
1884 		mlog_errno(status);
1885 		goto leave;
1886 	}
1887 
1888 	status = ocfs2_super_lock(osb, 1);
1889 	if (status < 0) {
1890 		mlog_errno(status);
1891 		goto leave;
1892 	}
1893 	unlock_super = 1;
1894 
1895 	/* This will load up the node map and add ourselves to it. */
1896 	status = ocfs2_find_slot(osb);
1897 	if (status < 0) {
1898 		mlog_errno(status);
1899 		goto leave;
1900 	}
1901 
1902 	/* load all node-local system inodes */
1903 	status = ocfs2_init_local_system_inodes(osb);
1904 	if (status < 0) {
1905 		mlog_errno(status);
1906 		goto leave;
1907 	}
1908 
1909 	status = ocfs2_check_volume(osb);
1910 	if (status < 0) {
1911 		mlog_errno(status);
1912 		goto leave;
1913 	}
1914 
1915 	status = ocfs2_truncate_log_init(osb);
1916 	if (status < 0)
1917 		mlog_errno(status);
1918 
1919 leave:
1920 	if (unlock_super)
1921 		ocfs2_super_unlock(osb, 1);
1922 
1923 	return status;
1924 }
1925 
ocfs2_dismount_volume(struct super_block * sb,int mnt_err)1926 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1927 {
1928 	int tmp, hangup_needed = 0;
1929 	struct ocfs2_super *osb = NULL;
1930 	char nodestr[8];
1931 
1932 	trace_ocfs2_dismount_volume(sb);
1933 
1934 	BUG_ON(!sb);
1935 	osb = OCFS2_SB(sb);
1936 	BUG_ON(!osb);
1937 
1938 	debugfs_remove(osb->osb_ctxt);
1939 
1940 	/*
1941 	 * Flush inode dropping work queue so that deletes are
1942 	 * performed while the filesystem is still working
1943 	 */
1944 	ocfs2_drop_all_dl_inodes(osb);
1945 
1946 	/* Orphan scan should be stopped as early as possible */
1947 	ocfs2_orphan_scan_stop(osb);
1948 
1949 	ocfs2_disable_quotas(osb);
1950 
1951 	ocfs2_shutdown_local_alloc(osb);
1952 
1953 	ocfs2_truncate_log_shutdown(osb);
1954 
1955 	/* This will disable recovery and flush any recovery work. */
1956 	ocfs2_recovery_exit(osb);
1957 
1958 	ocfs2_journal_shutdown(osb);
1959 
1960 	ocfs2_sync_blockdev(sb);
1961 
1962 	ocfs2_purge_refcount_trees(osb);
1963 
1964 	/* No cluster connection means we've failed during mount, so skip
1965 	 * all the steps which depended on that to complete. */
1966 	if (osb->cconn) {
1967 		tmp = ocfs2_super_lock(osb, 1);
1968 		if (tmp < 0) {
1969 			mlog_errno(tmp);
1970 			return;
1971 		}
1972 	}
1973 
1974 	if (osb->slot_num != OCFS2_INVALID_SLOT)
1975 		ocfs2_put_slot(osb);
1976 
1977 	if (osb->cconn)
1978 		ocfs2_super_unlock(osb, 1);
1979 
1980 	ocfs2_release_system_inodes(osb);
1981 
1982 	/*
1983 	 * If we're dismounting due to mount error, mount.ocfs2 will clean
1984 	 * up heartbeat.  If we're a local mount, there is no heartbeat.
1985 	 * If we failed before we got a uuid_str yet, we can't stop
1986 	 * heartbeat.  Otherwise, do it.
1987 	 */
1988 	if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str &&
1989 	    !ocfs2_is_hard_readonly(osb))
1990 		hangup_needed = 1;
1991 
1992 	if (osb->cconn)
1993 		ocfs2_dlm_shutdown(osb, hangup_needed);
1994 
1995 	ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1996 	debugfs_remove(osb->osb_debug_root);
1997 
1998 	if (hangup_needed)
1999 		ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
2000 
2001 	atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
2002 
2003 	if (ocfs2_mount_local(osb))
2004 		snprintf(nodestr, sizeof(nodestr), "local");
2005 	else
2006 		snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
2007 
2008 	printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
2009 	       osb->dev_str, nodestr);
2010 
2011 	ocfs2_delete_osb(osb);
2012 	kfree(osb);
2013 	sb->s_dev = 0;
2014 	sb->s_fs_info = NULL;
2015 }
2016 
ocfs2_setup_osb_uuid(struct ocfs2_super * osb,const unsigned char * uuid,unsigned uuid_bytes)2017 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
2018 				unsigned uuid_bytes)
2019 {
2020 	int i, ret;
2021 	char *ptr;
2022 
2023 	BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
2024 
2025 	osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
2026 	if (osb->uuid_str == NULL)
2027 		return -ENOMEM;
2028 
2029 	for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
2030 		/* print with null */
2031 		ret = snprintf(ptr, 3, "%02X", uuid[i]);
2032 		if (ret != 2) /* drop super cleans up */
2033 			return -EINVAL;
2034 		/* then only advance past the last char */
2035 		ptr += 2;
2036 	}
2037 
2038 	return 0;
2039 }
2040 
2041 /* Make sure entire volume is addressable by our journal.  Requires
2042    osb_clusters_at_boot to be valid and for the journal to have been
2043    initialized by ocfs2_journal_init(). */
ocfs2_journal_addressable(struct ocfs2_super * osb)2044 static int ocfs2_journal_addressable(struct ocfs2_super *osb)
2045 {
2046 	int status = 0;
2047 	u64 max_block =
2048 		ocfs2_clusters_to_blocks(osb->sb,
2049 					 osb->osb_clusters_at_boot) - 1;
2050 
2051 	/* 32-bit block number is always OK. */
2052 	if (max_block <= (u32)~0ULL)
2053 		goto out;
2054 
2055 	/* Volume is "huge", so see if our journal is new enough to
2056 	   support it. */
2057 	if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2058 				       OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2059 	      jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2060 					       JBD2_FEATURE_INCOMPAT_64BIT))) {
2061 		mlog(ML_ERROR, "The journal cannot address the entire volume. "
2062 		     "Enable the 'block64' journal option with tunefs.ocfs2");
2063 		status = -EFBIG;
2064 		goto out;
2065 	}
2066 
2067  out:
2068 	return status;
2069 }
2070 
ocfs2_initialize_super(struct super_block * sb,struct buffer_head * bh,int sector_size,struct ocfs2_blockcheck_stats * stats)2071 static int ocfs2_initialize_super(struct super_block *sb,
2072 				  struct buffer_head *bh,
2073 				  int sector_size,
2074 				  struct ocfs2_blockcheck_stats *stats)
2075 {
2076 	int status;
2077 	int i, cbits, bbits;
2078 	struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2079 	struct inode *inode = NULL;
2080 	struct ocfs2_journal *journal;
2081 	__le32 uuid_net_key;
2082 	struct ocfs2_super *osb;
2083 	u64 total_blocks;
2084 
2085 	osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2086 	if (!osb) {
2087 		status = -ENOMEM;
2088 		mlog_errno(status);
2089 		goto bail;
2090 	}
2091 
2092 	sb->s_fs_info = osb;
2093 	sb->s_op = &ocfs2_sops;
2094 	sb->s_d_op = &ocfs2_dentry_ops;
2095 	sb->s_export_op = &ocfs2_export_ops;
2096 	sb->s_qcop = &ocfs2_quotactl_ops;
2097 	sb->dq_op = &ocfs2_quota_operations;
2098 	sb->s_xattr = ocfs2_xattr_handlers;
2099 	sb->s_time_gran = 1;
2100 	sb->s_flags |= MS_NOATIME;
2101 	/* this is needed to support O_LARGEFILE */
2102 	cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2103 	bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2104 	sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2105 
2106 	osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2107 
2108 	for (i = 0; i < 3; i++)
2109 		osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2110 	osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2111 
2112 	osb->sb = sb;
2113 	/* Save off for ocfs2_rw_direct */
2114 	osb->s_sectsize_bits = blksize_bits(sector_size);
2115 	BUG_ON(!osb->s_sectsize_bits);
2116 
2117 	spin_lock_init(&osb->dc_task_lock);
2118 	init_waitqueue_head(&osb->dc_event);
2119 	osb->dc_work_sequence = 0;
2120 	osb->dc_wake_sequence = 0;
2121 	INIT_LIST_HEAD(&osb->blocked_lock_list);
2122 	osb->blocked_lock_count = 0;
2123 	spin_lock_init(&osb->osb_lock);
2124 	spin_lock_init(&osb->osb_xattr_lock);
2125 	ocfs2_init_steal_slots(osb);
2126 
2127 	atomic_set(&osb->alloc_stats.moves, 0);
2128 	atomic_set(&osb->alloc_stats.local_data, 0);
2129 	atomic_set(&osb->alloc_stats.bitmap_data, 0);
2130 	atomic_set(&osb->alloc_stats.bg_allocs, 0);
2131 	atomic_set(&osb->alloc_stats.bg_extends, 0);
2132 
2133 	/* Copy the blockcheck stats from the superblock probe */
2134 	osb->osb_ecc_stats = *stats;
2135 
2136 	ocfs2_init_node_maps(osb);
2137 
2138 	snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2139 		 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2140 
2141 	osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2142 	if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2143 		mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2144 		     osb->max_slots);
2145 		status = -EINVAL;
2146 		goto bail;
2147 	}
2148 
2149 	ocfs2_orphan_scan_init(osb);
2150 
2151 	status = ocfs2_recovery_init(osb);
2152 	if (status) {
2153 		mlog(ML_ERROR, "Unable to initialize recovery state\n");
2154 		mlog_errno(status);
2155 		goto bail;
2156 	}
2157 
2158 	init_waitqueue_head(&osb->checkpoint_event);
2159 	atomic_set(&osb->needs_checkpoint, 0);
2160 
2161 	osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2162 
2163 	osb->slot_num = OCFS2_INVALID_SLOT;
2164 
2165 	osb->s_xattr_inline_size = le16_to_cpu(
2166 					di->id2.i_super.s_xattr_inline_size);
2167 
2168 	osb->local_alloc_state = OCFS2_LA_UNUSED;
2169 	osb->local_alloc_bh = NULL;
2170 	INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2171 
2172 	init_waitqueue_head(&osb->osb_mount_event);
2173 
2174 	status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2175 	if (status) {
2176 		mlog_errno(status);
2177 		goto bail;
2178 	}
2179 
2180 	osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2181 	if (!osb->vol_label) {
2182 		mlog(ML_ERROR, "unable to alloc vol label\n");
2183 		status = -ENOMEM;
2184 		goto bail;
2185 	}
2186 
2187 	osb->slot_recovery_generations =
2188 		kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2189 			GFP_KERNEL);
2190 	if (!osb->slot_recovery_generations) {
2191 		status = -ENOMEM;
2192 		mlog_errno(status);
2193 		goto bail;
2194 	}
2195 
2196 	init_waitqueue_head(&osb->osb_wipe_event);
2197 	osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2198 					sizeof(*osb->osb_orphan_wipes),
2199 					GFP_KERNEL);
2200 	if (!osb->osb_orphan_wipes) {
2201 		status = -ENOMEM;
2202 		mlog_errno(status);
2203 		goto bail;
2204 	}
2205 
2206 	osb->osb_rf_lock_tree = RB_ROOT;
2207 
2208 	osb->s_feature_compat =
2209 		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2210 	osb->s_feature_ro_compat =
2211 		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2212 	osb->s_feature_incompat =
2213 		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2214 
2215 	if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2216 		mlog(ML_ERROR, "couldn't mount because of unsupported "
2217 		     "optional features (%x).\n", i);
2218 		status = -EINVAL;
2219 		goto bail;
2220 	}
2221 	if (!(osb->sb->s_flags & MS_RDONLY) &&
2222 	    (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2223 		mlog(ML_ERROR, "couldn't mount RDWR because of "
2224 		     "unsupported optional features (%x).\n", i);
2225 		status = -EINVAL;
2226 		goto bail;
2227 	}
2228 
2229 	if (ocfs2_clusterinfo_valid(osb)) {
2230 		osb->osb_stackflags =
2231 			OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2232 		memcpy(osb->osb_cluster_stack,
2233 		       OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2234 		       OCFS2_STACK_LABEL_LEN);
2235 		osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
2236 		if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2237 			mlog(ML_ERROR,
2238 			     "couldn't mount because of an invalid "
2239 			     "cluster stack label (%s) \n",
2240 			     osb->osb_cluster_stack);
2241 			status = -EINVAL;
2242 			goto bail;
2243 		}
2244 	} else {
2245 		/* The empty string is identical with classic tools that
2246 		 * don't know about s_cluster_info. */
2247 		osb->osb_cluster_stack[0] = '\0';
2248 	}
2249 
2250 	get_random_bytes(&osb->s_next_generation, sizeof(u32));
2251 
2252 	/* FIXME
2253 	 * This should be done in ocfs2_journal_init(), but unknown
2254 	 * ordering issues will cause the filesystem to crash.
2255 	 * If anyone wants to figure out what part of the code
2256 	 * refers to osb->journal before ocfs2_journal_init() is run,
2257 	 * be my guest.
2258 	 */
2259 	/* initialize our journal structure */
2260 
2261 	journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2262 	if (!journal) {
2263 		mlog(ML_ERROR, "unable to alloc journal\n");
2264 		status = -ENOMEM;
2265 		goto bail;
2266 	}
2267 	osb->journal = journal;
2268 	journal->j_osb = osb;
2269 
2270 	atomic_set(&journal->j_num_trans, 0);
2271 	init_rwsem(&journal->j_trans_barrier);
2272 	init_waitqueue_head(&journal->j_checkpointed);
2273 	spin_lock_init(&journal->j_lock);
2274 	journal->j_trans_id = (unsigned long) 1;
2275 	INIT_LIST_HEAD(&journal->j_la_cleanups);
2276 	INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2277 	journal->j_state = OCFS2_JOURNAL_FREE;
2278 
2279 	INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes);
2280 	osb->dentry_lock_list = NULL;
2281 
2282 	/* get some pseudo constants for clustersize bits */
2283 	osb->s_clustersize_bits =
2284 		le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2285 	osb->s_clustersize = 1 << osb->s_clustersize_bits;
2286 
2287 	if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2288 	    osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2289 		mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2290 		     osb->s_clustersize);
2291 		status = -EINVAL;
2292 		goto bail;
2293 	}
2294 
2295 	total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2296 						le32_to_cpu(di->i_clusters));
2297 
2298 	status = generic_check_addressable(osb->sb->s_blocksize_bits,
2299 					   total_blocks);
2300 	if (status) {
2301 		mlog(ML_ERROR, "Volume too large "
2302 		     "to mount safely on this system");
2303 		status = -EFBIG;
2304 		goto bail;
2305 	}
2306 
2307 	if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2308 				 sizeof(di->id2.i_super.s_uuid))) {
2309 		mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2310 		status = -ENOMEM;
2311 		goto bail;
2312 	}
2313 
2314 	memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
2315 
2316 	strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2317 	osb->vol_label[63] = '\0';
2318 	osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2319 	osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2320 	osb->first_cluster_group_blkno =
2321 		le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2322 	osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2323 	osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2324 	trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str,
2325 				     (unsigned long long)osb->root_blkno,
2326 				     (unsigned long long)osb->system_dir_blkno,
2327 				     osb->s_clustersize_bits);
2328 
2329 	osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2330 	if (!osb->osb_dlm_debug) {
2331 		status = -ENOMEM;
2332 		mlog_errno(status);
2333 		goto bail;
2334 	}
2335 
2336 	atomic_set(&osb->vol_state, VOLUME_INIT);
2337 
2338 	/* load root, system_dir, and all global system inodes */
2339 	status = ocfs2_init_global_system_inodes(osb);
2340 	if (status < 0) {
2341 		mlog_errno(status);
2342 		goto bail;
2343 	}
2344 
2345 	/*
2346 	 * global bitmap
2347 	 */
2348 	inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2349 					    OCFS2_INVALID_SLOT);
2350 	if (!inode) {
2351 		status = -EINVAL;
2352 		mlog_errno(status);
2353 		goto bail;
2354 	}
2355 
2356 	osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2357 	osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2358 	iput(inode);
2359 
2360 	osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2361 				 osb->s_feature_incompat) * 8;
2362 
2363 	status = ocfs2_init_slot_info(osb);
2364 	if (status < 0) {
2365 		mlog_errno(status);
2366 		goto bail;
2367 	}
2368 	cleancache_init_shared_fs((char *)&di->id2.i_super.s_uuid, sb);
2369 
2370 bail:
2371 	return status;
2372 }
2373 
2374 /*
2375  * will return: -EAGAIN if it is ok to keep searching for superblocks
2376  *              -EINVAL if there is a bad superblock
2377  *              0 on success
2378  */
ocfs2_verify_volume(struct ocfs2_dinode * di,struct buffer_head * bh,u32 blksz,struct ocfs2_blockcheck_stats * stats)2379 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2380 			       struct buffer_head *bh,
2381 			       u32 blksz,
2382 			       struct ocfs2_blockcheck_stats *stats)
2383 {
2384 	int status = -EAGAIN;
2385 
2386 	if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2387 		   strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2388 		/* We have to do a raw check of the feature here */
2389 		if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2390 		    OCFS2_FEATURE_INCOMPAT_META_ECC) {
2391 			status = ocfs2_block_check_validate(bh->b_data,
2392 							    bh->b_size,
2393 							    &di->i_check,
2394 							    stats);
2395 			if (status)
2396 				goto out;
2397 		}
2398 		status = -EINVAL;
2399 		if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2400 			mlog(ML_ERROR, "found superblock with incorrect block "
2401 			     "size: found %u, should be %u\n",
2402 			     1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2403 			       blksz);
2404 		} else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2405 			   OCFS2_MAJOR_REV_LEVEL ||
2406 			   le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2407 			   OCFS2_MINOR_REV_LEVEL) {
2408 			mlog(ML_ERROR, "found superblock with bad version: "
2409 			     "found %u.%u, should be %u.%u\n",
2410 			     le16_to_cpu(di->id2.i_super.s_major_rev_level),
2411 			     le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2412 			     OCFS2_MAJOR_REV_LEVEL,
2413 			     OCFS2_MINOR_REV_LEVEL);
2414 		} else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2415 			mlog(ML_ERROR, "bad block number on superblock: "
2416 			     "found %llu, should be %llu\n",
2417 			     (unsigned long long)le64_to_cpu(di->i_blkno),
2418 			     (unsigned long long)bh->b_blocknr);
2419 		} else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2420 			    le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2421 			mlog(ML_ERROR, "bad cluster size found: %u\n",
2422 			     1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2423 		} else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2424 			mlog(ML_ERROR, "bad root_blkno: 0\n");
2425 		} else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2426 			mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2427 		} else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2428 			mlog(ML_ERROR,
2429 			     "Superblock slots found greater than file system "
2430 			     "maximum: found %u, max %u\n",
2431 			     le16_to_cpu(di->id2.i_super.s_max_slots),
2432 			     OCFS2_MAX_SLOTS);
2433 		} else {
2434 			/* found it! */
2435 			status = 0;
2436 		}
2437 	}
2438 
2439 out:
2440 	if (status && status != -EAGAIN)
2441 		mlog_errno(status);
2442 	return status;
2443 }
2444 
ocfs2_check_volume(struct ocfs2_super * osb)2445 static int ocfs2_check_volume(struct ocfs2_super *osb)
2446 {
2447 	int status;
2448 	int dirty;
2449 	int local;
2450 	struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2451 						  * recover
2452 						  * ourselves. */
2453 
2454 	/* Init our journal object. */
2455 	status = ocfs2_journal_init(osb->journal, &dirty);
2456 	if (status < 0) {
2457 		mlog(ML_ERROR, "Could not initialize journal!\n");
2458 		goto finally;
2459 	}
2460 
2461 	/* Now that journal has been initialized, check to make sure
2462 	   entire volume is addressable. */
2463 	status = ocfs2_journal_addressable(osb);
2464 	if (status)
2465 		goto finally;
2466 
2467 	/* If the journal was unmounted cleanly then we don't want to
2468 	 * recover anything. Otherwise, journal_load will do that
2469 	 * dirty work for us :) */
2470 	if (!dirty) {
2471 		status = ocfs2_journal_wipe(osb->journal, 0);
2472 		if (status < 0) {
2473 			mlog_errno(status);
2474 			goto finally;
2475 		}
2476 	} else {
2477 		printk(KERN_NOTICE "ocfs2: File system on device (%s) was not "
2478 		       "unmounted cleanly, recovering it.\n", osb->dev_str);
2479 	}
2480 
2481 	local = ocfs2_mount_local(osb);
2482 
2483 	/* will play back anything left in the journal. */
2484 	status = ocfs2_journal_load(osb->journal, local, dirty);
2485 	if (status < 0) {
2486 		mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2487 		goto finally;
2488 	}
2489 
2490 	if (dirty) {
2491 		/* recover my local alloc if we didn't unmount cleanly. */
2492 		status = ocfs2_begin_local_alloc_recovery(osb,
2493 							  osb->slot_num,
2494 							  &local_alloc);
2495 		if (status < 0) {
2496 			mlog_errno(status);
2497 			goto finally;
2498 		}
2499 		/* we complete the recovery process after we've marked
2500 		 * ourselves as mounted. */
2501 	}
2502 
2503 	status = ocfs2_load_local_alloc(osb);
2504 	if (status < 0) {
2505 		mlog_errno(status);
2506 		goto finally;
2507 	}
2508 
2509 	if (dirty) {
2510 		/* Recovery will be completed after we've mounted the
2511 		 * rest of the volume. */
2512 		osb->dirty = 1;
2513 		osb->local_alloc_copy = local_alloc;
2514 		local_alloc = NULL;
2515 	}
2516 
2517 	/* go through each journal, trylock it and if you get the
2518 	 * lock, and it's marked as dirty, set the bit in the recover
2519 	 * map and launch a recovery thread for it. */
2520 	status = ocfs2_mark_dead_nodes(osb);
2521 	if (status < 0) {
2522 		mlog_errno(status);
2523 		goto finally;
2524 	}
2525 
2526 	status = ocfs2_compute_replay_slots(osb);
2527 	if (status < 0)
2528 		mlog_errno(status);
2529 
2530 finally:
2531 	kfree(local_alloc);
2532 
2533 	if (status)
2534 		mlog_errno(status);
2535 	return status;
2536 }
2537 
2538 /*
2539  * The routine gets called from dismount or close whenever a dismount on
2540  * volume is requested and the osb open count becomes 1.
2541  * It will remove the osb from the global list and also free up all the
2542  * initialized resources and fileobject.
2543  */
ocfs2_delete_osb(struct ocfs2_super * osb)2544 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2545 {
2546 	/* This function assumes that the caller has the main osb resource */
2547 
2548 	ocfs2_free_slot_info(osb);
2549 
2550 	kfree(osb->osb_orphan_wipes);
2551 	kfree(osb->slot_recovery_generations);
2552 	/* FIXME
2553 	 * This belongs in journal shutdown, but because we have to
2554 	 * allocate osb->journal at the start of ocfs2_initialize_osb(),
2555 	 * we free it here.
2556 	 */
2557 	kfree(osb->journal);
2558 	kfree(osb->local_alloc_copy);
2559 	kfree(osb->uuid_str);
2560 	ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2561 	memset(osb, 0, sizeof(struct ocfs2_super));
2562 }
2563 
2564 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2565  * panic(). We do not support continue-on-error operation. */
ocfs2_handle_error(struct super_block * sb)2566 static void ocfs2_handle_error(struct super_block *sb)
2567 {
2568 	struct ocfs2_super *osb = OCFS2_SB(sb);
2569 
2570 	if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2571 		panic("OCFS2: (device %s): panic forced after error\n",
2572 		      sb->s_id);
2573 
2574 	ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2575 
2576 	if (sb->s_flags & MS_RDONLY &&
2577 	    (ocfs2_is_soft_readonly(osb) ||
2578 	     ocfs2_is_hard_readonly(osb)))
2579 		return;
2580 
2581 	printk(KERN_CRIT "File system is now read-only due to the potential "
2582 	       "of on-disk corruption. Please run fsck.ocfs2 once the file "
2583 	       "system is unmounted.\n");
2584 	sb->s_flags |= MS_RDONLY;
2585 	ocfs2_set_ro_flag(osb, 0);
2586 }
2587 
2588 static char error_buf[1024];
2589 
__ocfs2_error(struct super_block * sb,const char * function,const char * fmt,...)2590 void __ocfs2_error(struct super_block *sb,
2591 		   const char *function,
2592 		   const char *fmt, ...)
2593 {
2594 	va_list args;
2595 
2596 	va_start(args, fmt);
2597 	vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2598 	va_end(args);
2599 
2600 	/* Not using mlog here because we want to show the actual
2601 	 * function the error came from. */
2602 	printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2603 	       sb->s_id, function, error_buf);
2604 
2605 	ocfs2_handle_error(sb);
2606 }
2607 
2608 /* Handle critical errors. This is intentionally more drastic than
2609  * ocfs2_handle_error, so we only use for things like journal errors,
2610  * etc. */
__ocfs2_abort(struct super_block * sb,const char * function,const char * fmt,...)2611 void __ocfs2_abort(struct super_block* sb,
2612 		   const char *function,
2613 		   const char *fmt, ...)
2614 {
2615 	va_list args;
2616 
2617 	va_start(args, fmt);
2618 	vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2619 	va_end(args);
2620 
2621 	printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2622 	       sb->s_id, function, error_buf);
2623 
2624 	/* We don't have the cluster support yet to go straight to
2625 	 * hard readonly in here. Until then, we want to keep
2626 	 * ocfs2_abort() so that we can at least mark critical
2627 	 * errors.
2628 	 *
2629 	 * TODO: This should abort the journal and alert other nodes
2630 	 * that our slot needs recovery. */
2631 
2632 	/* Force a panic(). This stinks, but it's better than letting
2633 	 * things continue without having a proper hard readonly
2634 	 * here. */
2635 	if (!ocfs2_mount_local(OCFS2_SB(sb)))
2636 		OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2637 	ocfs2_handle_error(sb);
2638 }
2639 
2640 /*
2641  * Void signal blockers, because in-kernel sigprocmask() only fails
2642  * when SIG_* is wrong.
2643  */
ocfs2_block_signals(sigset_t * oldset)2644 void ocfs2_block_signals(sigset_t *oldset)
2645 {
2646 	int rc;
2647 	sigset_t blocked;
2648 
2649 	sigfillset(&blocked);
2650 	rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2651 	BUG_ON(rc);
2652 }
2653 
ocfs2_unblock_signals(sigset_t * oldset)2654 void ocfs2_unblock_signals(sigset_t *oldset)
2655 {
2656 	int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2657 	BUG_ON(rc);
2658 }
2659 
2660 module_init(ocfs2_init);
2661 module_exit(ocfs2_exit);
2662