• 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 "xattr.h"
72 #include "quota.h"
73 #include "refcounttree.h"
74 #include "suballoc.h"
75 
76 #include "buffer_head_io.h"
77 
78 static struct kmem_cache *ocfs2_inode_cachep;
79 struct kmem_cache *ocfs2_dquot_cachep;
80 struct kmem_cache *ocfs2_qf_chunk_cachep;
81 
82 /* OCFS2 needs to schedule several different types of work which
83  * require cluster locking, disk I/O, recovery waits, etc. Since these
84  * types of work tend to be heavy we avoid using the kernel events
85  * workqueue and schedule on our own. */
86 struct workqueue_struct *ocfs2_wq = NULL;
87 
88 static struct dentry *ocfs2_debugfs_root;
89 
90 MODULE_AUTHOR("Oracle");
91 MODULE_LICENSE("GPL");
92 MODULE_DESCRIPTION("OCFS2 cluster file system");
93 
94 struct mount_options
95 {
96 	unsigned long	commit_interval;
97 	unsigned long	mount_opt;
98 	unsigned int	atime_quantum;
99 	unsigned 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 
ocfs2_get_dquots(struct inode * inode)146 static struct dquot **ocfs2_get_dquots(struct inode *inode)
147 {
148 	return OCFS2_I(inode)->i_dquot;
149 }
150 
151 static const struct super_operations ocfs2_sops = {
152 	.statfs		= ocfs2_statfs,
153 	.alloc_inode	= ocfs2_alloc_inode,
154 	.destroy_inode	= ocfs2_destroy_inode,
155 	.drop_inode	= ocfs2_drop_inode,
156 	.evict_inode	= ocfs2_evict_inode,
157 	.sync_fs	= ocfs2_sync_fs,
158 	.put_super	= ocfs2_put_super,
159 	.remount_fs	= ocfs2_remount,
160 	.show_options   = ocfs2_show_options,
161 	.quota_read	= ocfs2_quota_read,
162 	.quota_write	= ocfs2_quota_write,
163 	.get_dquots	= ocfs2_get_dquots,
164 };
165 
166 enum {
167 	Opt_barrier,
168 	Opt_err_panic,
169 	Opt_err_ro,
170 	Opt_intr,
171 	Opt_nointr,
172 	Opt_hb_none,
173 	Opt_hb_local,
174 	Opt_hb_global,
175 	Opt_data_ordered,
176 	Opt_data_writeback,
177 	Opt_atime_quantum,
178 	Opt_slot,
179 	Opt_commit,
180 	Opt_localalloc,
181 	Opt_localflocks,
182 	Opt_stack,
183 	Opt_user_xattr,
184 	Opt_nouser_xattr,
185 	Opt_inode64,
186 	Opt_acl,
187 	Opt_noacl,
188 	Opt_usrquota,
189 	Opt_grpquota,
190 	Opt_coherency_buffered,
191 	Opt_coherency_full,
192 	Opt_resv_level,
193 	Opt_dir_resv_level,
194 	Opt_journal_async_commit,
195 	Opt_err_cont,
196 	Opt_err,
197 };
198 
199 static const match_table_t tokens = {
200 	{Opt_barrier, "barrier=%u"},
201 	{Opt_err_panic, "errors=panic"},
202 	{Opt_err_ro, "errors=remount-ro"},
203 	{Opt_intr, "intr"},
204 	{Opt_nointr, "nointr"},
205 	{Opt_hb_none, OCFS2_HB_NONE},
206 	{Opt_hb_local, OCFS2_HB_LOCAL},
207 	{Opt_hb_global, OCFS2_HB_GLOBAL},
208 	{Opt_data_ordered, "data=ordered"},
209 	{Opt_data_writeback, "data=writeback"},
210 	{Opt_atime_quantum, "atime_quantum=%u"},
211 	{Opt_slot, "preferred_slot=%u"},
212 	{Opt_commit, "commit=%u"},
213 	{Opt_localalloc, "localalloc=%d"},
214 	{Opt_localflocks, "localflocks"},
215 	{Opt_stack, "cluster_stack=%s"},
216 	{Opt_user_xattr, "user_xattr"},
217 	{Opt_nouser_xattr, "nouser_xattr"},
218 	{Opt_inode64, "inode64"},
219 	{Opt_acl, "acl"},
220 	{Opt_noacl, "noacl"},
221 	{Opt_usrquota, "usrquota"},
222 	{Opt_grpquota, "grpquota"},
223 	{Opt_coherency_buffered, "coherency=buffered"},
224 	{Opt_coherency_full, "coherency=full"},
225 	{Opt_resv_level, "resv_level=%u"},
226 	{Opt_dir_resv_level, "dir_resv_level=%u"},
227 	{Opt_journal_async_commit, "journal_async_commit"},
228 	{Opt_err_cont, "errors=continue"},
229 	{Opt_err, NULL}
230 };
231 
232 #ifdef CONFIG_DEBUG_FS
ocfs2_osb_dump(struct ocfs2_super * osb,char * buf,int len)233 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
234 {
235 	struct ocfs2_cluster_connection *cconn = osb->cconn;
236 	struct ocfs2_recovery_map *rm = osb->recovery_map;
237 	struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
238 	int i, out = 0;
239 
240 	out += snprintf(buf + out, len - out,
241 			"%10s => Id: %-s  Uuid: %-s  Gen: 0x%X  Label: %-s\n",
242 			"Device", osb->dev_str, osb->uuid_str,
243 			osb->fs_generation, osb->vol_label);
244 
245 	out += snprintf(buf + out, len - out,
246 			"%10s => State: %d  Flags: 0x%lX\n", "Volume",
247 			atomic_read(&osb->vol_state), osb->osb_flags);
248 
249 	out += snprintf(buf + out, len - out,
250 			"%10s => Block: %lu  Cluster: %d\n", "Sizes",
251 			osb->sb->s_blocksize, osb->s_clustersize);
252 
253 	out += snprintf(buf + out, len - out,
254 			"%10s => Compat: 0x%X  Incompat: 0x%X  "
255 			"ROcompat: 0x%X\n",
256 			"Features", osb->s_feature_compat,
257 			osb->s_feature_incompat, osb->s_feature_ro_compat);
258 
259 	out += snprintf(buf + out, len - out,
260 			"%10s => Opts: 0x%lX  AtimeQuanta: %u\n", "Mount",
261 			osb->s_mount_opt, osb->s_atime_quantum);
262 
263 	if (cconn) {
264 		out += snprintf(buf + out, len - out,
265 				"%10s => Stack: %s  Name: %*s  "
266 				"Version: %d.%d\n", "Cluster",
267 				(*osb->osb_cluster_stack == '\0' ?
268 				 "o2cb" : osb->osb_cluster_stack),
269 				cconn->cc_namelen, cconn->cc_name,
270 				cconn->cc_version.pv_major,
271 				cconn->cc_version.pv_minor);
272 	}
273 
274 	spin_lock(&osb->dc_task_lock);
275 	out += snprintf(buf + out, len - out,
276 			"%10s => Pid: %d  Count: %lu  WakeSeq: %lu  "
277 			"WorkSeq: %lu\n", "DownCnvt",
278 			(osb->dc_task ?  task_pid_nr(osb->dc_task) : -1),
279 			osb->blocked_lock_count, osb->dc_wake_sequence,
280 			osb->dc_work_sequence);
281 	spin_unlock(&osb->dc_task_lock);
282 
283 	spin_lock(&osb->osb_lock);
284 	out += snprintf(buf + out, len - out, "%10s => Pid: %d  Nodes:",
285 			"Recovery",
286 			(osb->recovery_thread_task ?
287 			 task_pid_nr(osb->recovery_thread_task) : -1));
288 	if (rm->rm_used == 0)
289 		out += snprintf(buf + out, len - out, " None\n");
290 	else {
291 		for (i = 0; i < rm->rm_used; i++)
292 			out += snprintf(buf + out, len - out, " %d",
293 					rm->rm_entries[i]);
294 		out += snprintf(buf + out, len - out, "\n");
295 	}
296 	spin_unlock(&osb->osb_lock);
297 
298 	out += snprintf(buf + out, len - out,
299 			"%10s => Pid: %d  Interval: %lu\n", "Commit",
300 			(osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
301 			osb->osb_commit_interval);
302 
303 	out += snprintf(buf + out, len - out,
304 			"%10s => State: %d  TxnId: %lu  NumTxns: %d\n",
305 			"Journal", osb->journal->j_state,
306 			osb->journal->j_trans_id,
307 			atomic_read(&osb->journal->j_num_trans));
308 
309 	out += snprintf(buf + out, len - out,
310 			"%10s => GlobalAllocs: %d  LocalAllocs: %d  "
311 			"SubAllocs: %d  LAWinMoves: %d  SAExtends: %d\n",
312 			"Stats",
313 			atomic_read(&osb->alloc_stats.bitmap_data),
314 			atomic_read(&osb->alloc_stats.local_data),
315 			atomic_read(&osb->alloc_stats.bg_allocs),
316 			atomic_read(&osb->alloc_stats.moves),
317 			atomic_read(&osb->alloc_stats.bg_extends));
318 
319 	out += snprintf(buf + out, len - out,
320 			"%10s => State: %u  Descriptor: %llu  Size: %u bits  "
321 			"Default: %u bits\n",
322 			"LocalAlloc", osb->local_alloc_state,
323 			(unsigned long long)osb->la_last_gd,
324 			osb->local_alloc_bits, osb->local_alloc_default_bits);
325 
326 	spin_lock(&osb->osb_lock);
327 	out += snprintf(buf + out, len - out,
328 			"%10s => InodeSlot: %d  StolenInodes: %d, "
329 			"MetaSlot: %d  StolenMeta: %d\n", "Steal",
330 			osb->s_inode_steal_slot,
331 			atomic_read(&osb->s_num_inodes_stolen),
332 			osb->s_meta_steal_slot,
333 			atomic_read(&osb->s_num_meta_stolen));
334 	spin_unlock(&osb->osb_lock);
335 
336 	out += snprintf(buf + out, len - out, "OrphanScan => ");
337 	out += snprintf(buf + out, len - out, "Local: %u  Global: %u ",
338 			os->os_count, os->os_seqno);
339 	out += snprintf(buf + out, len - out, " Last Scan: ");
340 	if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
341 		out += snprintf(buf + out, len - out, "Disabled\n");
342 	else
343 		out += snprintf(buf + out, len - out, "%lu seconds ago\n",
344 				(get_seconds() - os->os_scantime.tv_sec));
345 
346 	out += snprintf(buf + out, len - out, "%10s => %3s  %10s\n",
347 			"Slots", "Num", "RecoGen");
348 	for (i = 0; i < osb->max_slots; ++i) {
349 		out += snprintf(buf + out, len - out,
350 				"%10s  %c %3d  %10d\n",
351 				" ",
352 				(i == osb->slot_num ? '*' : ' '),
353 				i, osb->slot_recovery_generations[i]);
354 	}
355 
356 	return out;
357 }
358 
ocfs2_osb_debug_open(struct inode * inode,struct file * file)359 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
360 {
361 	struct ocfs2_super *osb = inode->i_private;
362 	char *buf = NULL;
363 
364 	buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
365 	if (!buf)
366 		goto bail;
367 
368 	i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
369 
370 	file->private_data = buf;
371 
372 	return 0;
373 bail:
374 	return -ENOMEM;
375 }
376 
ocfs2_debug_release(struct inode * inode,struct file * file)377 static int ocfs2_debug_release(struct inode *inode, struct file *file)
378 {
379 	kfree(file->private_data);
380 	return 0;
381 }
382 
ocfs2_debug_read(struct file * file,char __user * buf,size_t nbytes,loff_t * ppos)383 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
384 				size_t nbytes, loff_t *ppos)
385 {
386 	return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
387 				       i_size_read(file->f_mapping->host));
388 }
389 #else
ocfs2_osb_debug_open(struct inode * inode,struct file * file)390 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
391 {
392 	return 0;
393 }
ocfs2_debug_release(struct inode * inode,struct file * file)394 static int ocfs2_debug_release(struct inode *inode, struct file *file)
395 {
396 	return 0;
397 }
ocfs2_debug_read(struct file * file,char __user * buf,size_t nbytes,loff_t * ppos)398 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
399 				size_t nbytes, loff_t *ppos)
400 {
401 	return 0;
402 }
403 #endif	/* CONFIG_DEBUG_FS */
404 
405 static const struct file_operations ocfs2_osb_debug_fops = {
406 	.open =		ocfs2_osb_debug_open,
407 	.release =	ocfs2_debug_release,
408 	.read =		ocfs2_debug_read,
409 	.llseek =	generic_file_llseek,
410 };
411 
ocfs2_sync_fs(struct super_block * sb,int wait)412 static int ocfs2_sync_fs(struct super_block *sb, int wait)
413 {
414 	int status;
415 	tid_t target;
416 	struct ocfs2_super *osb = OCFS2_SB(sb);
417 
418 	if (ocfs2_is_hard_readonly(osb))
419 		return -EROFS;
420 
421 	if (wait) {
422 		status = ocfs2_flush_truncate_log(osb);
423 		if (status < 0)
424 			mlog_errno(status);
425 	} else {
426 		ocfs2_schedule_truncate_log_flush(osb, 0);
427 	}
428 
429 	if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
430 				      &target)) {
431 		if (wait)
432 			jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
433 					     target);
434 	}
435 	return 0;
436 }
437 
ocfs2_need_system_inode(struct ocfs2_super * osb,int ino)438 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
439 {
440 	if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
441 	    && (ino == USER_QUOTA_SYSTEM_INODE
442 		|| ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
443 		return 0;
444 	if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
445 	    && (ino == GROUP_QUOTA_SYSTEM_INODE
446 		|| ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
447 		return 0;
448 	return 1;
449 }
450 
ocfs2_init_global_system_inodes(struct ocfs2_super * osb)451 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
452 {
453 	struct inode *new = NULL;
454 	int status = 0;
455 	int i;
456 
457 	new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
458 	if (IS_ERR(new)) {
459 		status = PTR_ERR(new);
460 		mlog_errno(status);
461 		goto bail;
462 	}
463 	osb->root_inode = new;
464 
465 	new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
466 	if (IS_ERR(new)) {
467 		status = PTR_ERR(new);
468 		mlog_errno(status);
469 		goto bail;
470 	}
471 	osb->sys_root_inode = new;
472 
473 	for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
474 	     i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
475 		if (!ocfs2_need_system_inode(osb, i))
476 			continue;
477 		new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
478 		if (!new) {
479 			ocfs2_release_system_inodes(osb);
480 			status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
481 			mlog_errno(status);
482 			mlog(ML_ERROR, "Unable to load system inode %d, "
483 			     "possibly corrupt fs?", i);
484 			goto bail;
485 		}
486 		// the array now has one ref, so drop this one
487 		iput(new);
488 	}
489 
490 bail:
491 	if (status)
492 		mlog_errno(status);
493 	return status;
494 }
495 
ocfs2_init_local_system_inodes(struct ocfs2_super * osb)496 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
497 {
498 	struct inode *new = NULL;
499 	int status = 0;
500 	int i;
501 
502 	for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
503 	     i < NUM_SYSTEM_INODES;
504 	     i++) {
505 		if (!ocfs2_need_system_inode(osb, i))
506 			continue;
507 		new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
508 		if (!new) {
509 			ocfs2_release_system_inodes(osb);
510 			status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
511 			mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
512 			     status, i, osb->slot_num);
513 			goto bail;
514 		}
515 		/* the array now has one ref, so drop this one */
516 		iput(new);
517 	}
518 
519 bail:
520 	if (status)
521 		mlog_errno(status);
522 	return status;
523 }
524 
ocfs2_release_system_inodes(struct ocfs2_super * osb)525 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
526 {
527 	int i;
528 	struct inode *inode;
529 
530 	for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
531 		inode = osb->global_system_inodes[i];
532 		if (inode) {
533 			iput(inode);
534 			osb->global_system_inodes[i] = NULL;
535 		}
536 	}
537 
538 	inode = osb->sys_root_inode;
539 	if (inode) {
540 		iput(inode);
541 		osb->sys_root_inode = NULL;
542 	}
543 
544 	inode = osb->root_inode;
545 	if (inode) {
546 		iput(inode);
547 		osb->root_inode = NULL;
548 	}
549 
550 	if (!osb->local_system_inodes)
551 		return;
552 
553 	for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
554 		if (osb->local_system_inodes[i]) {
555 			iput(osb->local_system_inodes[i]);
556 			osb->local_system_inodes[i] = NULL;
557 		}
558 	}
559 
560 	kfree(osb->local_system_inodes);
561 	osb->local_system_inodes = NULL;
562 }
563 
564 /* We're allocating fs objects, use GFP_NOFS */
ocfs2_alloc_inode(struct super_block * sb)565 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
566 {
567 	struct ocfs2_inode_info *oi;
568 
569 	oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
570 	if (!oi)
571 		return NULL;
572 
573 	oi->i_sync_tid = 0;
574 	oi->i_datasync_tid = 0;
575 	memset(&oi->i_dquot, 0, sizeof(oi->i_dquot));
576 
577 	jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
578 	return &oi->vfs_inode;
579 }
580 
ocfs2_i_callback(struct rcu_head * head)581 static void ocfs2_i_callback(struct rcu_head *head)
582 {
583 	struct inode *inode = container_of(head, struct inode, i_rcu);
584 	kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
585 }
586 
ocfs2_destroy_inode(struct inode * inode)587 static void ocfs2_destroy_inode(struct inode *inode)
588 {
589 	call_rcu(&inode->i_rcu, ocfs2_i_callback);
590 }
591 
ocfs2_max_file_offset(unsigned int bbits,unsigned int cbits)592 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
593 						unsigned int cbits)
594 {
595 	unsigned int bytes = 1 << cbits;
596 	unsigned int trim = bytes;
597 	unsigned int bitshift = 32;
598 
599 	/*
600 	 * i_size and all block offsets in ocfs2 are always 64 bits
601 	 * wide. i_clusters is 32 bits, in cluster-sized units. So on
602 	 * 64 bit platforms, cluster size will be the limiting factor.
603 	 */
604 
605 #if BITS_PER_LONG == 32
606 # if defined(CONFIG_LBDAF)
607 	BUILD_BUG_ON(sizeof(sector_t) != 8);
608 	/*
609 	 * We might be limited by page cache size.
610 	 */
611 	if (bytes > PAGE_CACHE_SIZE) {
612 		bytes = PAGE_CACHE_SIZE;
613 		trim = 1;
614 		/*
615 		 * Shift by 31 here so that we don't get larger than
616 		 * MAX_LFS_FILESIZE
617 		 */
618 		bitshift = 31;
619 	}
620 # else
621 	/*
622 	 * We are limited by the size of sector_t. Use block size, as
623 	 * that's what we expose to the VFS.
624 	 */
625 	bytes = 1 << bbits;
626 	trim = 1;
627 	bitshift = 31;
628 # endif
629 #endif
630 
631 	/*
632 	 * Trim by a whole cluster when we can actually approach the
633 	 * on-disk limits. Otherwise we can overflow i_clusters when
634 	 * an extent start is at the max offset.
635 	 */
636 	return (((unsigned long long)bytes) << bitshift) - trim;
637 }
638 
ocfs2_remount(struct super_block * sb,int * flags,char * data)639 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
640 {
641 	int incompat_features;
642 	int ret = 0;
643 	struct mount_options parsed_options;
644 	struct ocfs2_super *osb = OCFS2_SB(sb);
645 	u32 tmp;
646 
647 	sync_filesystem(sb);
648 
649 	if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
650 	    !ocfs2_check_set_options(sb, &parsed_options)) {
651 		ret = -EINVAL;
652 		goto out;
653 	}
654 
655 	tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
656 		OCFS2_MOUNT_HB_NONE;
657 	if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
658 		ret = -EINVAL;
659 		mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
660 		goto out;
661 	}
662 
663 	if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
664 	    (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
665 		ret = -EINVAL;
666 		mlog(ML_ERROR, "Cannot change data mode on remount\n");
667 		goto out;
668 	}
669 
670 	/* Probably don't want this on remount; it might
671 	 * mess with other nodes */
672 	if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
673 	    (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
674 		ret = -EINVAL;
675 		mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
676 		goto out;
677 	}
678 
679 	/* We're going to/from readonly mode. */
680 	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
681 		/* Disable quota accounting before remounting RO */
682 		if (*flags & MS_RDONLY) {
683 			ret = ocfs2_susp_quotas(osb, 0);
684 			if (ret < 0)
685 				goto out;
686 		}
687 		/* Lock here so the check of HARD_RO and the potential
688 		 * setting of SOFT_RO is atomic. */
689 		spin_lock(&osb->osb_lock);
690 		if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
691 			mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
692 			ret = -EROFS;
693 			goto unlock_osb;
694 		}
695 
696 		if (*flags & MS_RDONLY) {
697 			sb->s_flags |= MS_RDONLY;
698 			osb->osb_flags |= OCFS2_OSB_SOFT_RO;
699 		} else {
700 			if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
701 				mlog(ML_ERROR, "Cannot remount RDWR "
702 				     "filesystem due to previous errors.\n");
703 				ret = -EROFS;
704 				goto unlock_osb;
705 			}
706 			incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
707 			if (incompat_features) {
708 				mlog(ML_ERROR, "Cannot remount RDWR because "
709 				     "of unsupported optional features "
710 				     "(%x).\n", incompat_features);
711 				ret = -EINVAL;
712 				goto unlock_osb;
713 			}
714 			sb->s_flags &= ~MS_RDONLY;
715 			osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
716 		}
717 		trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags);
718 unlock_osb:
719 		spin_unlock(&osb->osb_lock);
720 		/* Enable quota accounting after remounting RW */
721 		if (!ret && !(*flags & MS_RDONLY)) {
722 			if (sb_any_quota_suspended(sb))
723 				ret = ocfs2_susp_quotas(osb, 1);
724 			else
725 				ret = ocfs2_enable_quotas(osb);
726 			if (ret < 0) {
727 				/* Return back changes... */
728 				spin_lock(&osb->osb_lock);
729 				sb->s_flags |= MS_RDONLY;
730 				osb->osb_flags |= OCFS2_OSB_SOFT_RO;
731 				spin_unlock(&osb->osb_lock);
732 				goto out;
733 			}
734 		}
735 	}
736 
737 	if (!ret) {
738 		/* Only save off the new mount options in case of a successful
739 		 * remount. */
740 		osb->s_mount_opt = parsed_options.mount_opt;
741 		osb->s_atime_quantum = parsed_options.atime_quantum;
742 		osb->preferred_slot = parsed_options.slot;
743 		if (parsed_options.commit_interval)
744 			osb->osb_commit_interval = parsed_options.commit_interval;
745 
746 		if (!ocfs2_is_hard_readonly(osb))
747 			ocfs2_set_journal_params(osb);
748 
749 		sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
750 			((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
751 							MS_POSIXACL : 0);
752 	}
753 out:
754 	return ret;
755 }
756 
ocfs2_sb_probe(struct super_block * sb,struct buffer_head ** bh,int * sector_size,struct ocfs2_blockcheck_stats * stats)757 static int ocfs2_sb_probe(struct super_block *sb,
758 			  struct buffer_head **bh,
759 			  int *sector_size,
760 			  struct ocfs2_blockcheck_stats *stats)
761 {
762 	int status, tmpstat;
763 	struct ocfs1_vol_disk_hdr *hdr;
764 	struct ocfs2_dinode *di;
765 	int blksize;
766 
767 	*bh = NULL;
768 
769 	/* may be > 512 */
770 	*sector_size = bdev_logical_block_size(sb->s_bdev);
771 	if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
772 		mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
773 		     *sector_size, OCFS2_MAX_BLOCKSIZE);
774 		status = -EINVAL;
775 		goto bail;
776 	}
777 
778 	/* Can this really happen? */
779 	if (*sector_size < OCFS2_MIN_BLOCKSIZE)
780 		*sector_size = OCFS2_MIN_BLOCKSIZE;
781 
782 	/* check block zero for old format */
783 	status = ocfs2_get_sector(sb, bh, 0, *sector_size);
784 	if (status < 0) {
785 		mlog_errno(status);
786 		goto bail;
787 	}
788 	hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
789 	if (hdr->major_version == OCFS1_MAJOR_VERSION) {
790 		mlog(ML_ERROR, "incompatible version: %u.%u\n",
791 		     hdr->major_version, hdr->minor_version);
792 		status = -EINVAL;
793 	}
794 	if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
795 		   strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
796 		mlog(ML_ERROR, "incompatible volume signature: %8s\n",
797 		     hdr->signature);
798 		status = -EINVAL;
799 	}
800 	brelse(*bh);
801 	*bh = NULL;
802 	if (status < 0) {
803 		mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
804 		     "upgraded before mounting with ocfs v2\n");
805 		goto bail;
806 	}
807 
808 	/*
809 	 * Now check at magic offset for 512, 1024, 2048, 4096
810 	 * blocksizes.  4096 is the maximum blocksize because it is
811 	 * the minimum clustersize.
812 	 */
813 	status = -EINVAL;
814 	for (blksize = *sector_size;
815 	     blksize <= OCFS2_MAX_BLOCKSIZE;
816 	     blksize <<= 1) {
817 		tmpstat = ocfs2_get_sector(sb, bh,
818 					   OCFS2_SUPER_BLOCK_BLKNO,
819 					   blksize);
820 		if (tmpstat < 0) {
821 			status = tmpstat;
822 			mlog_errno(status);
823 			break;
824 		}
825 		di = (struct ocfs2_dinode *) (*bh)->b_data;
826 		memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
827 		spin_lock_init(&stats->b_lock);
828 		tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
829 		if (tmpstat < 0) {
830 			brelse(*bh);
831 			*bh = NULL;
832 		}
833 		if (tmpstat != -EAGAIN) {
834 			status = tmpstat;
835 			break;
836 		}
837 	}
838 
839 bail:
840 	return status;
841 }
842 
ocfs2_verify_heartbeat(struct ocfs2_super * osb)843 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
844 {
845 	u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
846 
847 	if (osb->s_mount_opt & hb_enabled) {
848 		if (ocfs2_mount_local(osb)) {
849 			mlog(ML_ERROR, "Cannot heartbeat on a locally "
850 			     "mounted device.\n");
851 			return -EINVAL;
852 		}
853 		if (ocfs2_userspace_stack(osb)) {
854 			mlog(ML_ERROR, "Userspace stack expected, but "
855 			     "o2cb heartbeat arguments passed to mount\n");
856 			return -EINVAL;
857 		}
858 		if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
859 		     !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
860 		    ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
861 		     ocfs2_cluster_o2cb_global_heartbeat(osb))) {
862 			mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
863 			return -EINVAL;
864 		}
865 	}
866 
867 	if (!(osb->s_mount_opt & hb_enabled)) {
868 		if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
869 		    !ocfs2_userspace_stack(osb)) {
870 			mlog(ML_ERROR, "Heartbeat has to be started to mount "
871 			     "a read-write clustered device.\n");
872 			return -EINVAL;
873 		}
874 	}
875 
876 	return 0;
877 }
878 
879 /*
880  * If we're using a userspace stack, mount should have passed
881  * a name that matches the disk.  If not, mount should not
882  * have passed a stack.
883  */
ocfs2_verify_userspace_stack(struct ocfs2_super * osb,struct mount_options * mopt)884 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
885 					struct mount_options *mopt)
886 {
887 	if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
888 		mlog(ML_ERROR,
889 		     "cluster stack passed to mount, but this filesystem "
890 		     "does not support it\n");
891 		return -EINVAL;
892 	}
893 
894 	if (ocfs2_userspace_stack(osb) &&
895 	    strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
896 		    OCFS2_STACK_LABEL_LEN)) {
897 		mlog(ML_ERROR,
898 		     "cluster stack passed to mount (\"%s\") does not "
899 		     "match the filesystem (\"%s\")\n",
900 		     mopt->cluster_stack,
901 		     osb->osb_cluster_stack);
902 		return -EINVAL;
903 	}
904 
905 	return 0;
906 }
907 
ocfs2_susp_quotas(struct ocfs2_super * osb,int unsuspend)908 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
909 {
910 	int type;
911 	struct super_block *sb = osb->sb;
912 	unsigned int feature[OCFS2_MAXQUOTAS] = {
913 					OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
914 					OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
915 	int status = 0;
916 
917 	for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
918 		if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
919 			continue;
920 		if (unsuspend)
921 			status = dquot_resume(sb, type);
922 		else {
923 			struct ocfs2_mem_dqinfo *oinfo;
924 
925 			/* Cancel periodic syncing before suspending */
926 			oinfo = sb_dqinfo(sb, type)->dqi_priv;
927 			cancel_delayed_work_sync(&oinfo->dqi_sync_work);
928 			status = dquot_suspend(sb, type);
929 		}
930 		if (status < 0)
931 			break;
932 	}
933 	if (status < 0)
934 		mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
935 		     "remount (error = %d).\n", status);
936 	return status;
937 }
938 
ocfs2_enable_quotas(struct ocfs2_super * osb)939 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
940 {
941 	struct inode *inode[OCFS2_MAXQUOTAS] = { NULL, NULL };
942 	struct super_block *sb = osb->sb;
943 	unsigned int feature[OCFS2_MAXQUOTAS] = {
944 					OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
945 					OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
946 	unsigned int ino[OCFS2_MAXQUOTAS] = {
947 					LOCAL_USER_QUOTA_SYSTEM_INODE,
948 					LOCAL_GROUP_QUOTA_SYSTEM_INODE };
949 	int status;
950 	int type;
951 
952 	sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
953 	for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
954 		if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
955 			continue;
956 		inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
957 							osb->slot_num);
958 		if (!inode[type]) {
959 			status = -ENOENT;
960 			goto out_quota_off;
961 		}
962 		status = dquot_enable(inode[type], type, QFMT_OCFS2,
963 				      DQUOT_USAGE_ENABLED);
964 		if (status < 0)
965 			goto out_quota_off;
966 	}
967 
968 	for (type = 0; type < OCFS2_MAXQUOTAS; type++)
969 		iput(inode[type]);
970 	return 0;
971 out_quota_off:
972 	ocfs2_disable_quotas(osb);
973 	for (type = 0; type < OCFS2_MAXQUOTAS; type++)
974 		iput(inode[type]);
975 	mlog_errno(status);
976 	return status;
977 }
978 
ocfs2_disable_quotas(struct ocfs2_super * osb)979 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
980 {
981 	int type;
982 	struct inode *inode;
983 	struct super_block *sb = osb->sb;
984 	struct ocfs2_mem_dqinfo *oinfo;
985 
986 	/* We mostly ignore errors in this function because there's not much
987 	 * we can do when we see them */
988 	for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
989 		if (!sb_has_quota_loaded(sb, type))
990 			continue;
991 		/* Cancel periodic syncing before we grab dqonoff_mutex */
992 		oinfo = sb_dqinfo(sb, type)->dqi_priv;
993 		cancel_delayed_work_sync(&oinfo->dqi_sync_work);
994 		inode = igrab(sb->s_dquot.files[type]);
995 		/* Turn off quotas. This will remove all dquot structures from
996 		 * memory and so they will be automatically synced to global
997 		 * quota files */
998 		dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
999 					DQUOT_LIMITS_ENABLED);
1000 		if (!inode)
1001 			continue;
1002 		iput(inode);
1003 	}
1004 }
1005 
ocfs2_fill_super(struct super_block * sb,void * data,int silent)1006 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
1007 {
1008 	struct dentry *root;
1009 	int status, sector_size;
1010 	struct mount_options parsed_options;
1011 	struct inode *inode = NULL;
1012 	struct ocfs2_super *osb = NULL;
1013 	struct buffer_head *bh = NULL;
1014 	char nodestr[12];
1015 	struct ocfs2_blockcheck_stats stats;
1016 
1017 	trace_ocfs2_fill_super(sb, data, silent);
1018 
1019 	if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
1020 		status = -EINVAL;
1021 		goto read_super_error;
1022 	}
1023 
1024 	/* probe for superblock */
1025 	status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
1026 	if (status < 0) {
1027 		mlog(ML_ERROR, "superblock probe failed!\n");
1028 		goto read_super_error;
1029 	}
1030 
1031 	status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1032 	osb = OCFS2_SB(sb);
1033 	if (status < 0) {
1034 		mlog_errno(status);
1035 		goto read_super_error;
1036 	}
1037 	brelse(bh);
1038 	bh = NULL;
1039 
1040 	if (!ocfs2_check_set_options(sb, &parsed_options)) {
1041 		status = -EINVAL;
1042 		goto read_super_error;
1043 	}
1044 	osb->s_mount_opt = parsed_options.mount_opt;
1045 	osb->s_atime_quantum = parsed_options.atime_quantum;
1046 	osb->preferred_slot = parsed_options.slot;
1047 	osb->osb_commit_interval = parsed_options.commit_interval;
1048 
1049 	ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1050 	osb->osb_resv_level = parsed_options.resv_level;
1051 	osb->osb_dir_resv_level = parsed_options.resv_level;
1052 	if (parsed_options.dir_resv_level == -1)
1053 		osb->osb_dir_resv_level = parsed_options.resv_level;
1054 	else
1055 		osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1056 
1057 	status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1058 	if (status)
1059 		goto read_super_error;
1060 
1061 	sb->s_magic = OCFS2_SUPER_MAGIC;
1062 
1063 	sb->s_flags = (sb->s_flags & ~(MS_POSIXACL | MS_NOSEC)) |
1064 		((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1065 
1066 	/* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1067 	 * heartbeat=none */
1068 	if (bdev_read_only(sb->s_bdev)) {
1069 		if (!(sb->s_flags & MS_RDONLY)) {
1070 			status = -EACCES;
1071 			mlog(ML_ERROR, "Readonly device detected but readonly "
1072 			     "mount was not specified.\n");
1073 			goto read_super_error;
1074 		}
1075 
1076 		/* You should not be able to start a local heartbeat
1077 		 * on a readonly device. */
1078 		if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1079 			status = -EROFS;
1080 			mlog(ML_ERROR, "Local heartbeat specified on readonly "
1081 			     "device.\n");
1082 			goto read_super_error;
1083 		}
1084 
1085 		status = ocfs2_check_journals_nolocks(osb);
1086 		if (status < 0) {
1087 			if (status == -EROFS)
1088 				mlog(ML_ERROR, "Recovery required on readonly "
1089 				     "file system, but write access is "
1090 				     "unavailable.\n");
1091 			else
1092 				mlog_errno(status);
1093 			goto read_super_error;
1094 		}
1095 
1096 		ocfs2_set_ro_flag(osb, 1);
1097 
1098 		printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. "
1099 		       "Cluster services will not be used for this mount. "
1100 		       "Recovery will be skipped.\n", osb->dev_str);
1101 	}
1102 
1103 	if (!ocfs2_is_hard_readonly(osb)) {
1104 		if (sb->s_flags & MS_RDONLY)
1105 			ocfs2_set_ro_flag(osb, 0);
1106 	}
1107 
1108 	status = ocfs2_verify_heartbeat(osb);
1109 	if (status < 0) {
1110 		mlog_errno(status);
1111 		goto read_super_error;
1112 	}
1113 
1114 	osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1115 						 ocfs2_debugfs_root);
1116 	if (!osb->osb_debug_root) {
1117 		status = -EINVAL;
1118 		mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1119 		goto read_super_error;
1120 	}
1121 
1122 	osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1123 					    osb->osb_debug_root,
1124 					    osb,
1125 					    &ocfs2_osb_debug_fops);
1126 	if (!osb->osb_ctxt) {
1127 		status = -EINVAL;
1128 		mlog_errno(status);
1129 		goto read_super_error;
1130 	}
1131 
1132 	if (ocfs2_meta_ecc(osb)) {
1133 		status = ocfs2_blockcheck_stats_debugfs_install(
1134 						&osb->osb_ecc_stats,
1135 						osb->osb_debug_root);
1136 		if (status) {
1137 			mlog(ML_ERROR,
1138 			     "Unable to create blockcheck statistics "
1139 			     "files\n");
1140 			goto read_super_error;
1141 		}
1142 	}
1143 
1144 	status = ocfs2_mount_volume(sb);
1145 	if (status < 0)
1146 		goto read_super_error;
1147 
1148 	if (osb->root_inode)
1149 		inode = igrab(osb->root_inode);
1150 
1151 	if (!inode) {
1152 		status = -EIO;
1153 		mlog_errno(status);
1154 		goto read_super_error;
1155 	}
1156 
1157 	root = d_make_root(inode);
1158 	if (!root) {
1159 		status = -ENOMEM;
1160 		mlog_errno(status);
1161 		goto read_super_error;
1162 	}
1163 
1164 	sb->s_root = root;
1165 
1166 	ocfs2_complete_mount_recovery(osb);
1167 
1168 	if (ocfs2_mount_local(osb))
1169 		snprintf(nodestr, sizeof(nodestr), "local");
1170 	else
1171 		snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1172 
1173 	printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1174 	       "with %s data mode.\n",
1175 	       osb->dev_str, nodestr, osb->slot_num,
1176 	       osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1177 	       "ordered");
1178 
1179 	atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1180 	wake_up(&osb->osb_mount_event);
1181 
1182 	/* Now we can initialize quotas because we can afford to wait
1183 	 * for cluster locks recovery now. That also means that truncation
1184 	 * log recovery can happen but that waits for proper quota setup */
1185 	if (!(sb->s_flags & MS_RDONLY)) {
1186 		status = ocfs2_enable_quotas(osb);
1187 		if (status < 0) {
1188 			/* We have to err-out specially here because
1189 			 * s_root is already set */
1190 			mlog_errno(status);
1191 			atomic_set(&osb->vol_state, VOLUME_DISABLED);
1192 			wake_up(&osb->osb_mount_event);
1193 			return status;
1194 		}
1195 	}
1196 
1197 	ocfs2_complete_quota_recovery(osb);
1198 
1199 	/* Now we wake up again for processes waiting for quotas */
1200 	atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1201 	wake_up(&osb->osb_mount_event);
1202 
1203 	/* Start this when the mount is almost sure of being successful */
1204 	ocfs2_orphan_scan_start(osb);
1205 
1206 	return status;
1207 
1208 read_super_error:
1209 	brelse(bh);
1210 
1211 	if (osb) {
1212 		atomic_set(&osb->vol_state, VOLUME_DISABLED);
1213 		wake_up(&osb->osb_mount_event);
1214 		ocfs2_dismount_volume(sb, 1);
1215 	}
1216 
1217 	if (status)
1218 		mlog_errno(status);
1219 	return status;
1220 }
1221 
ocfs2_mount(struct file_system_type * fs_type,int flags,const char * dev_name,void * data)1222 static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
1223 			int flags,
1224 			const char *dev_name,
1225 			void *data)
1226 {
1227 	return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
1228 }
1229 
1230 static struct file_system_type ocfs2_fs_type = {
1231 	.owner          = THIS_MODULE,
1232 	.name           = "ocfs2",
1233 	.mount          = ocfs2_mount,
1234 	.kill_sb        = kill_block_super,
1235 	.fs_flags       = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1236 	.next           = NULL
1237 };
1238 MODULE_ALIAS_FS("ocfs2");
1239 
ocfs2_check_set_options(struct super_block * sb,struct mount_options * options)1240 static int ocfs2_check_set_options(struct super_block *sb,
1241 				   struct mount_options *options)
1242 {
1243 	if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1244 	    !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1245 					 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1246 		mlog(ML_ERROR, "User quotas were requested, but this "
1247 		     "filesystem does not have the feature enabled.\n");
1248 		return 0;
1249 	}
1250 	if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1251 	    !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1252 					 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1253 		mlog(ML_ERROR, "Group quotas were requested, but this "
1254 		     "filesystem does not have the feature enabled.\n");
1255 		return 0;
1256 	}
1257 	if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1258 	    !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1259 		mlog(ML_ERROR, "ACL support requested but extended attributes "
1260 		     "feature is not enabled\n");
1261 		return 0;
1262 	}
1263 	/* No ACL setting specified? Use XATTR feature... */
1264 	if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1265 				    OCFS2_MOUNT_NO_POSIX_ACL))) {
1266 		if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1267 			options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1268 		else
1269 			options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1270 	}
1271 	return 1;
1272 }
1273 
ocfs2_parse_options(struct super_block * sb,char * options,struct mount_options * mopt,int is_remount)1274 static int ocfs2_parse_options(struct super_block *sb,
1275 			       char *options,
1276 			       struct mount_options *mopt,
1277 			       int is_remount)
1278 {
1279 	int status, user_stack = 0;
1280 	char *p;
1281 	u32 tmp;
1282 
1283 	trace_ocfs2_parse_options(is_remount, options ? options : "(none)");
1284 
1285 	mopt->commit_interval = 0;
1286 	mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1287 	mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1288 	mopt->slot = OCFS2_INVALID_SLOT;
1289 	mopt->localalloc_opt = -1;
1290 	mopt->cluster_stack[0] = '\0';
1291 	mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1292 	mopt->dir_resv_level = -1;
1293 
1294 	if (!options) {
1295 		status = 1;
1296 		goto bail;
1297 	}
1298 
1299 	while ((p = strsep(&options, ",")) != NULL) {
1300 		int token, option;
1301 		substring_t args[MAX_OPT_ARGS];
1302 
1303 		if (!*p)
1304 			continue;
1305 
1306 		token = match_token(p, tokens, args);
1307 		switch (token) {
1308 		case Opt_hb_local:
1309 			mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1310 			break;
1311 		case Opt_hb_none:
1312 			mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1313 			break;
1314 		case Opt_hb_global:
1315 			mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1316 			break;
1317 		case Opt_barrier:
1318 			if (match_int(&args[0], &option)) {
1319 				status = 0;
1320 				goto bail;
1321 			}
1322 			if (option)
1323 				mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1324 			else
1325 				mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1326 			break;
1327 		case Opt_intr:
1328 			mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1329 			break;
1330 		case Opt_nointr:
1331 			mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1332 			break;
1333 		case Opt_err_panic:
1334 			mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1335 			mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1336 			mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1337 			break;
1338 		case Opt_err_ro:
1339 			mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1340 			mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1341 			mopt->mount_opt |= OCFS2_MOUNT_ERRORS_ROFS;
1342 			break;
1343 		case Opt_err_cont:
1344 			mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1345 			mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1346 			mopt->mount_opt |= OCFS2_MOUNT_ERRORS_CONT;
1347 			break;
1348 		case Opt_data_ordered:
1349 			mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1350 			break;
1351 		case Opt_data_writeback:
1352 			mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1353 			break;
1354 		case Opt_user_xattr:
1355 			mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1356 			break;
1357 		case Opt_nouser_xattr:
1358 			mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1359 			break;
1360 		case Opt_atime_quantum:
1361 			if (match_int(&args[0], &option)) {
1362 				status = 0;
1363 				goto bail;
1364 			}
1365 			if (option >= 0)
1366 				mopt->atime_quantum = option;
1367 			break;
1368 		case Opt_slot:
1369 			option = 0;
1370 			if (match_int(&args[0], &option)) {
1371 				status = 0;
1372 				goto bail;
1373 			}
1374 			if (option)
1375 				mopt->slot = (u16)option;
1376 			break;
1377 		case Opt_commit:
1378 			option = 0;
1379 			if (match_int(&args[0], &option)) {
1380 				status = 0;
1381 				goto bail;
1382 			}
1383 			if (option < 0)
1384 				return 0;
1385 			if (option == 0)
1386 				option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1387 			mopt->commit_interval = HZ * option;
1388 			break;
1389 		case Opt_localalloc:
1390 			option = 0;
1391 			if (match_int(&args[0], &option)) {
1392 				status = 0;
1393 				goto bail;
1394 			}
1395 			if (option >= 0)
1396 				mopt->localalloc_opt = option;
1397 			break;
1398 		case Opt_localflocks:
1399 			/*
1400 			 * Changing this during remount could race
1401 			 * flock() requests, or "unbalance" existing
1402 			 * ones (e.g., a lock is taken in one mode but
1403 			 * dropped in the other). If users care enough
1404 			 * to flip locking modes during remount, we
1405 			 * could add a "local" flag to individual
1406 			 * flock structures for proper tracking of
1407 			 * state.
1408 			 */
1409 			if (!is_remount)
1410 				mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1411 			break;
1412 		case Opt_stack:
1413 			/* Check both that the option we were passed
1414 			 * is of the right length and that it is a proper
1415 			 * string of the right length.
1416 			 */
1417 			if (((args[0].to - args[0].from) !=
1418 			     OCFS2_STACK_LABEL_LEN) ||
1419 			    (strnlen(args[0].from,
1420 				     OCFS2_STACK_LABEL_LEN) !=
1421 			     OCFS2_STACK_LABEL_LEN)) {
1422 				mlog(ML_ERROR,
1423 				     "Invalid cluster_stack option\n");
1424 				status = 0;
1425 				goto bail;
1426 			}
1427 			memcpy(mopt->cluster_stack, args[0].from,
1428 			       OCFS2_STACK_LABEL_LEN);
1429 			mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1430 			/*
1431 			 * Open code the memcmp here as we don't have
1432 			 * an osb to pass to
1433 			 * ocfs2_userspace_stack().
1434 			 */
1435 			if (memcmp(mopt->cluster_stack,
1436 				   OCFS2_CLASSIC_CLUSTER_STACK,
1437 				   OCFS2_STACK_LABEL_LEN))
1438 				user_stack = 1;
1439 			break;
1440 		case Opt_inode64:
1441 			mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1442 			break;
1443 		case Opt_usrquota:
1444 			mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1445 			break;
1446 		case Opt_grpquota:
1447 			mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1448 			break;
1449 		case Opt_coherency_buffered:
1450 			mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1451 			break;
1452 		case Opt_coherency_full:
1453 			mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1454 			break;
1455 		case Opt_acl:
1456 			mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1457 			mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1458 			break;
1459 		case Opt_noacl:
1460 			mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1461 			mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1462 			break;
1463 		case Opt_resv_level:
1464 			if (is_remount)
1465 				break;
1466 			if (match_int(&args[0], &option)) {
1467 				status = 0;
1468 				goto bail;
1469 			}
1470 			if (option >= OCFS2_MIN_RESV_LEVEL &&
1471 			    option < OCFS2_MAX_RESV_LEVEL)
1472 				mopt->resv_level = option;
1473 			break;
1474 		case Opt_dir_resv_level:
1475 			if (is_remount)
1476 				break;
1477 			if (match_int(&args[0], &option)) {
1478 				status = 0;
1479 				goto bail;
1480 			}
1481 			if (option >= OCFS2_MIN_RESV_LEVEL &&
1482 			    option < OCFS2_MAX_RESV_LEVEL)
1483 				mopt->dir_resv_level = option;
1484 			break;
1485 		case Opt_journal_async_commit:
1486 			mopt->mount_opt |= OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT;
1487 			break;
1488 		default:
1489 			mlog(ML_ERROR,
1490 			     "Unrecognized mount option \"%s\" "
1491 			     "or missing value\n", p);
1492 			status = 0;
1493 			goto bail;
1494 		}
1495 	}
1496 
1497 	if (user_stack == 0) {
1498 		/* Ensure only one heartbeat mode */
1499 		tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1500 					 OCFS2_MOUNT_HB_GLOBAL |
1501 					 OCFS2_MOUNT_HB_NONE);
1502 		if (hweight32(tmp) != 1) {
1503 			mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1504 			status = 0;
1505 			goto bail;
1506 		}
1507 	}
1508 
1509 	status = 1;
1510 
1511 bail:
1512 	return status;
1513 }
1514 
ocfs2_show_options(struct seq_file * s,struct dentry * root)1515 static int ocfs2_show_options(struct seq_file *s, struct dentry *root)
1516 {
1517 	struct ocfs2_super *osb = OCFS2_SB(root->d_sb);
1518 	unsigned long opts = osb->s_mount_opt;
1519 	unsigned int local_alloc_megs;
1520 
1521 	if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1522 		seq_printf(s, ",_netdev");
1523 		if (opts & OCFS2_MOUNT_HB_LOCAL)
1524 			seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1525 		else
1526 			seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1527 	} else
1528 		seq_printf(s, ",%s", OCFS2_HB_NONE);
1529 
1530 	if (opts & OCFS2_MOUNT_NOINTR)
1531 		seq_printf(s, ",nointr");
1532 
1533 	if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1534 		seq_printf(s, ",data=writeback");
1535 	else
1536 		seq_printf(s, ",data=ordered");
1537 
1538 	if (opts & OCFS2_MOUNT_BARRIER)
1539 		seq_printf(s, ",barrier=1");
1540 
1541 	if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1542 		seq_printf(s, ",errors=panic");
1543 	else if (opts & OCFS2_MOUNT_ERRORS_CONT)
1544 		seq_printf(s, ",errors=continue");
1545 	else
1546 		seq_printf(s, ",errors=remount-ro");
1547 
1548 	if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1549 		seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1550 
1551 	seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1552 
1553 	if (osb->osb_commit_interval)
1554 		seq_printf(s, ",commit=%u",
1555 			   (unsigned) (osb->osb_commit_interval / HZ));
1556 
1557 	local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1558 	if (local_alloc_megs != ocfs2_la_default_mb(osb))
1559 		seq_printf(s, ",localalloc=%d", local_alloc_megs);
1560 
1561 	if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1562 		seq_printf(s, ",localflocks,");
1563 
1564 	if (osb->osb_cluster_stack[0])
1565 		seq_show_option_n(s, "cluster_stack", osb->osb_cluster_stack,
1566 				  OCFS2_STACK_LABEL_LEN);
1567 	if (opts & OCFS2_MOUNT_USRQUOTA)
1568 		seq_printf(s, ",usrquota");
1569 	if (opts & OCFS2_MOUNT_GRPQUOTA)
1570 		seq_printf(s, ",grpquota");
1571 
1572 	if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1573 		seq_printf(s, ",coherency=buffered");
1574 	else
1575 		seq_printf(s, ",coherency=full");
1576 
1577 	if (opts & OCFS2_MOUNT_NOUSERXATTR)
1578 		seq_printf(s, ",nouser_xattr");
1579 	else
1580 		seq_printf(s, ",user_xattr");
1581 
1582 	if (opts & OCFS2_MOUNT_INODE64)
1583 		seq_printf(s, ",inode64");
1584 
1585 	if (opts & OCFS2_MOUNT_POSIX_ACL)
1586 		seq_printf(s, ",acl");
1587 	else
1588 		seq_printf(s, ",noacl");
1589 
1590 	if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1591 		seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1592 
1593 	if (osb->osb_dir_resv_level != osb->osb_resv_level)
1594 		seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1595 
1596 	if (opts & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
1597 		seq_printf(s, ",journal_async_commit");
1598 
1599 	return 0;
1600 }
1601 
ocfs2_init(void)1602 static int __init ocfs2_init(void)
1603 {
1604 	int status;
1605 
1606 	status = init_ocfs2_uptodate_cache();
1607 	if (status < 0)
1608 		goto out1;
1609 
1610 	status = ocfs2_initialize_mem_caches();
1611 	if (status < 0)
1612 		goto out2;
1613 
1614 	ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1615 	if (!ocfs2_wq) {
1616 		status = -ENOMEM;
1617 		goto out3;
1618 	}
1619 
1620 	ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1621 	if (!ocfs2_debugfs_root) {
1622 		status = -ENOMEM;
1623 		mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1624 		goto out4;
1625 	}
1626 
1627 	ocfs2_set_locking_protocol();
1628 
1629 	status = register_quota_format(&ocfs2_quota_format);
1630 	if (status < 0)
1631 		goto out4;
1632 	status = register_filesystem(&ocfs2_fs_type);
1633 	if (!status)
1634 		return 0;
1635 
1636 	unregister_quota_format(&ocfs2_quota_format);
1637 out4:
1638 	destroy_workqueue(ocfs2_wq);
1639 	debugfs_remove(ocfs2_debugfs_root);
1640 out3:
1641 	ocfs2_free_mem_caches();
1642 out2:
1643 	exit_ocfs2_uptodate_cache();
1644 out1:
1645 	mlog_errno(status);
1646 	return status;
1647 }
1648 
ocfs2_exit(void)1649 static void __exit ocfs2_exit(void)
1650 {
1651 	if (ocfs2_wq) {
1652 		flush_workqueue(ocfs2_wq);
1653 		destroy_workqueue(ocfs2_wq);
1654 	}
1655 
1656 	unregister_quota_format(&ocfs2_quota_format);
1657 
1658 	debugfs_remove(ocfs2_debugfs_root);
1659 
1660 	ocfs2_free_mem_caches();
1661 
1662 	unregister_filesystem(&ocfs2_fs_type);
1663 
1664 	exit_ocfs2_uptodate_cache();
1665 }
1666 
ocfs2_put_super(struct super_block * sb)1667 static void ocfs2_put_super(struct super_block *sb)
1668 {
1669 	trace_ocfs2_put_super(sb);
1670 
1671 	ocfs2_sync_blockdev(sb);
1672 	ocfs2_dismount_volume(sb, 0);
1673 }
1674 
ocfs2_statfs(struct dentry * dentry,struct kstatfs * buf)1675 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1676 {
1677 	struct ocfs2_super *osb;
1678 	u32 numbits, freebits;
1679 	int status;
1680 	struct ocfs2_dinode *bm_lock;
1681 	struct buffer_head *bh = NULL;
1682 	struct inode *inode = NULL;
1683 
1684 	trace_ocfs2_statfs(dentry->d_sb, buf);
1685 
1686 	osb = OCFS2_SB(dentry->d_sb);
1687 
1688 	inode = ocfs2_get_system_file_inode(osb,
1689 					    GLOBAL_BITMAP_SYSTEM_INODE,
1690 					    OCFS2_INVALID_SLOT);
1691 	if (!inode) {
1692 		mlog(ML_ERROR, "failed to get bitmap inode\n");
1693 		status = -EIO;
1694 		goto bail;
1695 	}
1696 
1697 	status = ocfs2_inode_lock(inode, &bh, 0);
1698 	if (status < 0) {
1699 		mlog_errno(status);
1700 		goto bail;
1701 	}
1702 
1703 	bm_lock = (struct ocfs2_dinode *) bh->b_data;
1704 
1705 	numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1706 	freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1707 
1708 	buf->f_type = OCFS2_SUPER_MAGIC;
1709 	buf->f_bsize = dentry->d_sb->s_blocksize;
1710 	buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1711 	buf->f_blocks = ((sector_t) numbits) *
1712 			(osb->s_clustersize >> osb->sb->s_blocksize_bits);
1713 	buf->f_bfree = ((sector_t) freebits) *
1714 		       (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1715 	buf->f_bavail = buf->f_bfree;
1716 	buf->f_files = numbits;
1717 	buf->f_ffree = freebits;
1718 	buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1719 				& 0xFFFFFFFFUL;
1720 	buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1721 				OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1722 
1723 	brelse(bh);
1724 
1725 	ocfs2_inode_unlock(inode, 0);
1726 	status = 0;
1727 bail:
1728 	if (inode)
1729 		iput(inode);
1730 
1731 	if (status)
1732 		mlog_errno(status);
1733 
1734 	return status;
1735 }
1736 
ocfs2_inode_init_once(void * data)1737 static void ocfs2_inode_init_once(void *data)
1738 {
1739 	struct ocfs2_inode_info *oi = data;
1740 
1741 	oi->ip_flags = 0;
1742 	oi->ip_open_count = 0;
1743 	spin_lock_init(&oi->ip_lock);
1744 	ocfs2_extent_map_init(&oi->vfs_inode);
1745 	INIT_LIST_HEAD(&oi->ip_io_markers);
1746 	oi->ip_dir_start_lookup = 0;
1747 	mutex_init(&oi->ip_unaligned_aio);
1748 	init_rwsem(&oi->ip_alloc_sem);
1749 	init_rwsem(&oi->ip_xattr_sem);
1750 	mutex_init(&oi->ip_io_mutex);
1751 
1752 	oi->ip_blkno = 0ULL;
1753 	oi->ip_clusters = 0;
1754 	oi->ip_next_orphan = NULL;
1755 
1756 	ocfs2_resv_init_once(&oi->ip_la_data_resv);
1757 
1758 	ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1759 	ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1760 	ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1761 
1762 	ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1763 				  &ocfs2_inode_caching_ops);
1764 
1765 	inode_init_once(&oi->vfs_inode);
1766 }
1767 
ocfs2_initialize_mem_caches(void)1768 static int ocfs2_initialize_mem_caches(void)
1769 {
1770 	ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1771 				       sizeof(struct ocfs2_inode_info),
1772 				       0,
1773 				       (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1774 						SLAB_MEM_SPREAD),
1775 				       ocfs2_inode_init_once);
1776 	ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1777 					sizeof(struct ocfs2_dquot),
1778 					0,
1779 					(SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1780 						SLAB_MEM_SPREAD),
1781 					NULL);
1782 	ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1783 					sizeof(struct ocfs2_quota_chunk),
1784 					0,
1785 					(SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1786 					NULL);
1787 	if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1788 	    !ocfs2_qf_chunk_cachep) {
1789 		if (ocfs2_inode_cachep)
1790 			kmem_cache_destroy(ocfs2_inode_cachep);
1791 		if (ocfs2_dquot_cachep)
1792 			kmem_cache_destroy(ocfs2_dquot_cachep);
1793 		if (ocfs2_qf_chunk_cachep)
1794 			kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1795 		return -ENOMEM;
1796 	}
1797 
1798 	return 0;
1799 }
1800 
ocfs2_free_mem_caches(void)1801 static void ocfs2_free_mem_caches(void)
1802 {
1803 	/*
1804 	 * Make sure all delayed rcu free inodes are flushed before we
1805 	 * destroy cache.
1806 	 */
1807 	rcu_barrier();
1808 	if (ocfs2_inode_cachep)
1809 		kmem_cache_destroy(ocfs2_inode_cachep);
1810 	ocfs2_inode_cachep = NULL;
1811 
1812 	if (ocfs2_dquot_cachep)
1813 		kmem_cache_destroy(ocfs2_dquot_cachep);
1814 	ocfs2_dquot_cachep = NULL;
1815 
1816 	if (ocfs2_qf_chunk_cachep)
1817 		kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1818 	ocfs2_qf_chunk_cachep = NULL;
1819 }
1820 
ocfs2_get_sector(struct super_block * sb,struct buffer_head ** bh,int block,int sect_size)1821 static int ocfs2_get_sector(struct super_block *sb,
1822 			    struct buffer_head **bh,
1823 			    int block,
1824 			    int sect_size)
1825 {
1826 	if (!sb_set_blocksize(sb, sect_size)) {
1827 		mlog(ML_ERROR, "unable to set blocksize\n");
1828 		return -EIO;
1829 	}
1830 
1831 	*bh = sb_getblk(sb, block);
1832 	if (!*bh) {
1833 		mlog_errno(-ENOMEM);
1834 		return -ENOMEM;
1835 	}
1836 	lock_buffer(*bh);
1837 	if (!buffer_dirty(*bh))
1838 		clear_buffer_uptodate(*bh);
1839 	unlock_buffer(*bh);
1840 	ll_rw_block(READ, 1, bh);
1841 	wait_on_buffer(*bh);
1842 	if (!buffer_uptodate(*bh)) {
1843 		mlog_errno(-EIO);
1844 		brelse(*bh);
1845 		*bh = NULL;
1846 		return -EIO;
1847 	}
1848 
1849 	return 0;
1850 }
1851 
ocfs2_mount_volume(struct super_block * sb)1852 static int ocfs2_mount_volume(struct super_block *sb)
1853 {
1854 	int status = 0;
1855 	int unlock_super = 0;
1856 	struct ocfs2_super *osb = OCFS2_SB(sb);
1857 
1858 	if (ocfs2_is_hard_readonly(osb))
1859 		goto leave;
1860 
1861 	status = ocfs2_dlm_init(osb);
1862 	if (status < 0) {
1863 		mlog_errno(status);
1864 		goto leave;
1865 	}
1866 
1867 	status = ocfs2_super_lock(osb, 1);
1868 	if (status < 0) {
1869 		mlog_errno(status);
1870 		goto leave;
1871 	}
1872 	unlock_super = 1;
1873 
1874 	/* This will load up the node map and add ourselves to it. */
1875 	status = ocfs2_find_slot(osb);
1876 	if (status < 0) {
1877 		mlog_errno(status);
1878 		goto leave;
1879 	}
1880 
1881 	/* load all node-local system inodes */
1882 	status = ocfs2_init_local_system_inodes(osb);
1883 	if (status < 0) {
1884 		mlog_errno(status);
1885 		goto leave;
1886 	}
1887 
1888 	status = ocfs2_check_volume(osb);
1889 	if (status < 0) {
1890 		mlog_errno(status);
1891 		goto leave;
1892 	}
1893 
1894 	status = ocfs2_truncate_log_init(osb);
1895 	if (status < 0)
1896 		mlog_errno(status);
1897 
1898 leave:
1899 	if (unlock_super)
1900 		ocfs2_super_unlock(osb, 1);
1901 
1902 	return status;
1903 }
1904 
ocfs2_dismount_volume(struct super_block * sb,int mnt_err)1905 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1906 {
1907 	int tmp, hangup_needed = 0;
1908 	struct ocfs2_super *osb = NULL;
1909 	char nodestr[12];
1910 
1911 	trace_ocfs2_dismount_volume(sb);
1912 
1913 	BUG_ON(!sb);
1914 	osb = OCFS2_SB(sb);
1915 	BUG_ON(!osb);
1916 
1917 	debugfs_remove(osb->osb_ctxt);
1918 
1919 	/* Orphan scan should be stopped as early as possible */
1920 	ocfs2_orphan_scan_stop(osb);
1921 
1922 	ocfs2_disable_quotas(osb);
1923 
1924 	/* All dquots should be freed by now */
1925 	WARN_ON(!llist_empty(&osb->dquot_drop_list));
1926 	/* Wait for worker to be done with the work structure in osb */
1927 	cancel_work_sync(&osb->dquot_drop_work);
1928 
1929 	ocfs2_shutdown_local_alloc(osb);
1930 
1931 	ocfs2_truncate_log_shutdown(osb);
1932 
1933 	/* This will disable recovery and flush any recovery work. */
1934 	ocfs2_recovery_exit(osb);
1935 
1936 	ocfs2_journal_shutdown(osb);
1937 
1938 	ocfs2_sync_blockdev(sb);
1939 
1940 	ocfs2_purge_refcount_trees(osb);
1941 
1942 	/* No cluster connection means we've failed during mount, so skip
1943 	 * all the steps which depended on that to complete. */
1944 	if (osb->cconn) {
1945 		tmp = ocfs2_super_lock(osb, 1);
1946 		if (tmp < 0) {
1947 			mlog_errno(tmp);
1948 			return;
1949 		}
1950 	}
1951 
1952 	if (osb->slot_num != OCFS2_INVALID_SLOT)
1953 		ocfs2_put_slot(osb);
1954 
1955 	if (osb->cconn)
1956 		ocfs2_super_unlock(osb, 1);
1957 
1958 	ocfs2_release_system_inodes(osb);
1959 
1960 	/*
1961 	 * If we're dismounting due to mount error, mount.ocfs2 will clean
1962 	 * up heartbeat.  If we're a local mount, there is no heartbeat.
1963 	 * If we failed before we got a uuid_str yet, we can't stop
1964 	 * heartbeat.  Otherwise, do it.
1965 	 */
1966 	if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str &&
1967 	    !ocfs2_is_hard_readonly(osb))
1968 		hangup_needed = 1;
1969 
1970 	if (osb->cconn)
1971 		ocfs2_dlm_shutdown(osb, hangup_needed);
1972 
1973 	ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1974 	debugfs_remove(osb->osb_debug_root);
1975 
1976 	if (hangup_needed)
1977 		ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1978 
1979 	atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1980 
1981 	if (ocfs2_mount_local(osb))
1982 		snprintf(nodestr, sizeof(nodestr), "local");
1983 	else
1984 		snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1985 
1986 	printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1987 	       osb->dev_str, nodestr);
1988 
1989 	ocfs2_delete_osb(osb);
1990 	kfree(osb);
1991 	sb->s_dev = 0;
1992 	sb->s_fs_info = NULL;
1993 }
1994 
ocfs2_setup_osb_uuid(struct ocfs2_super * osb,const unsigned char * uuid,unsigned uuid_bytes)1995 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1996 				unsigned uuid_bytes)
1997 {
1998 	int i, ret;
1999 	char *ptr;
2000 
2001 	BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
2002 
2003 	osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
2004 	if (osb->uuid_str == NULL)
2005 		return -ENOMEM;
2006 
2007 	for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
2008 		/* print with null */
2009 		ret = snprintf(ptr, 3, "%02X", uuid[i]);
2010 		if (ret != 2) /* drop super cleans up */
2011 			return -EINVAL;
2012 		/* then only advance past the last char */
2013 		ptr += 2;
2014 	}
2015 
2016 	return 0;
2017 }
2018 
2019 /* Make sure entire volume is addressable by our journal.  Requires
2020    osb_clusters_at_boot to be valid and for the journal to have been
2021    initialized by ocfs2_journal_init(). */
ocfs2_journal_addressable(struct ocfs2_super * osb)2022 static int ocfs2_journal_addressable(struct ocfs2_super *osb)
2023 {
2024 	int status = 0;
2025 	u64 max_block =
2026 		ocfs2_clusters_to_blocks(osb->sb,
2027 					 osb->osb_clusters_at_boot) - 1;
2028 
2029 	/* 32-bit block number is always OK. */
2030 	if (max_block <= (u32)~0ULL)
2031 		goto out;
2032 
2033 	/* Volume is "huge", so see if our journal is new enough to
2034 	   support it. */
2035 	if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2036 				       OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2037 	      jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2038 					       JBD2_FEATURE_INCOMPAT_64BIT))) {
2039 		mlog(ML_ERROR, "The journal cannot address the entire volume. "
2040 		     "Enable the 'block64' journal option with tunefs.ocfs2");
2041 		status = -EFBIG;
2042 		goto out;
2043 	}
2044 
2045  out:
2046 	return status;
2047 }
2048 
ocfs2_initialize_super(struct super_block * sb,struct buffer_head * bh,int sector_size,struct ocfs2_blockcheck_stats * stats)2049 static int ocfs2_initialize_super(struct super_block *sb,
2050 				  struct buffer_head *bh,
2051 				  int sector_size,
2052 				  struct ocfs2_blockcheck_stats *stats)
2053 {
2054 	int status;
2055 	int i, cbits, bbits;
2056 	struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2057 	struct inode *inode = NULL;
2058 	struct ocfs2_journal *journal;
2059 	struct ocfs2_super *osb;
2060 	u64 total_blocks;
2061 
2062 	osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2063 	if (!osb) {
2064 		status = -ENOMEM;
2065 		mlog_errno(status);
2066 		goto bail;
2067 	}
2068 
2069 	sb->s_fs_info = osb;
2070 	sb->s_op = &ocfs2_sops;
2071 	sb->s_d_op = &ocfs2_dentry_ops;
2072 	sb->s_export_op = &ocfs2_export_ops;
2073 	sb->s_qcop = &dquot_quotactl_sysfile_ops;
2074 	sb->dq_op = &ocfs2_quota_operations;
2075 	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
2076 	sb->s_xattr = ocfs2_xattr_handlers;
2077 	sb->s_time_gran = 1;
2078 	sb->s_flags |= MS_NOATIME;
2079 	/* this is needed to support O_LARGEFILE */
2080 	cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2081 	bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2082 	sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2083 	memcpy(sb->s_uuid, di->id2.i_super.s_uuid,
2084 	       sizeof(di->id2.i_super.s_uuid));
2085 
2086 	osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2087 
2088 	for (i = 0; i < 3; i++)
2089 		osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2090 	osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2091 
2092 	osb->sb = sb;
2093 	/* Save off for ocfs2_rw_direct */
2094 	osb->s_sectsize_bits = blksize_bits(sector_size);
2095 	BUG_ON(!osb->s_sectsize_bits);
2096 
2097 	spin_lock_init(&osb->dc_task_lock);
2098 	init_waitqueue_head(&osb->dc_event);
2099 	osb->dc_work_sequence = 0;
2100 	osb->dc_wake_sequence = 0;
2101 	INIT_LIST_HEAD(&osb->blocked_lock_list);
2102 	osb->blocked_lock_count = 0;
2103 	spin_lock_init(&osb->osb_lock);
2104 	spin_lock_init(&osb->osb_xattr_lock);
2105 	ocfs2_init_steal_slots(osb);
2106 
2107 	mutex_init(&osb->system_file_mutex);
2108 
2109 	atomic_set(&osb->alloc_stats.moves, 0);
2110 	atomic_set(&osb->alloc_stats.local_data, 0);
2111 	atomic_set(&osb->alloc_stats.bitmap_data, 0);
2112 	atomic_set(&osb->alloc_stats.bg_allocs, 0);
2113 	atomic_set(&osb->alloc_stats.bg_extends, 0);
2114 
2115 	/* Copy the blockcheck stats from the superblock probe */
2116 	osb->osb_ecc_stats = *stats;
2117 
2118 	ocfs2_init_node_maps(osb);
2119 
2120 	snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2121 		 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2122 
2123 	osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2124 	if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2125 		mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2126 		     osb->max_slots);
2127 		status = -EINVAL;
2128 		goto bail;
2129 	}
2130 
2131 	ocfs2_orphan_scan_init(osb);
2132 
2133 	status = ocfs2_recovery_init(osb);
2134 	if (status) {
2135 		mlog(ML_ERROR, "Unable to initialize recovery state\n");
2136 		mlog_errno(status);
2137 		goto bail;
2138 	}
2139 
2140 	init_waitqueue_head(&osb->checkpoint_event);
2141 
2142 	osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2143 
2144 	osb->slot_num = OCFS2_INVALID_SLOT;
2145 
2146 	osb->s_xattr_inline_size = le16_to_cpu(
2147 					di->id2.i_super.s_xattr_inline_size);
2148 
2149 	osb->local_alloc_state = OCFS2_LA_UNUSED;
2150 	osb->local_alloc_bh = NULL;
2151 	INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2152 
2153 	init_waitqueue_head(&osb->osb_mount_event);
2154 
2155 	status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2156 	if (status) {
2157 		mlog_errno(status);
2158 		goto bail;
2159 	}
2160 
2161 	osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2162 	if (!osb->vol_label) {
2163 		mlog(ML_ERROR, "unable to alloc vol label\n");
2164 		status = -ENOMEM;
2165 		goto bail;
2166 	}
2167 
2168 	osb->slot_recovery_generations =
2169 		kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2170 			GFP_KERNEL);
2171 	if (!osb->slot_recovery_generations) {
2172 		status = -ENOMEM;
2173 		mlog_errno(status);
2174 		goto bail;
2175 	}
2176 
2177 	init_waitqueue_head(&osb->osb_wipe_event);
2178 	osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2179 					sizeof(*osb->osb_orphan_wipes),
2180 					GFP_KERNEL);
2181 	if (!osb->osb_orphan_wipes) {
2182 		status = -ENOMEM;
2183 		mlog_errno(status);
2184 		goto bail;
2185 	}
2186 
2187 	osb->osb_rf_lock_tree = RB_ROOT;
2188 
2189 	osb->s_feature_compat =
2190 		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2191 	osb->s_feature_ro_compat =
2192 		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2193 	osb->s_feature_incompat =
2194 		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2195 
2196 	if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2197 		mlog(ML_ERROR, "couldn't mount because of unsupported "
2198 		     "optional features (%x).\n", i);
2199 		status = -EINVAL;
2200 		goto bail;
2201 	}
2202 	if (!(osb->sb->s_flags & MS_RDONLY) &&
2203 	    (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2204 		mlog(ML_ERROR, "couldn't mount RDWR because of "
2205 		     "unsupported optional features (%x).\n", i);
2206 		status = -EINVAL;
2207 		goto bail;
2208 	}
2209 
2210 	if (ocfs2_clusterinfo_valid(osb)) {
2211 		/*
2212 		 * ci_stack and ci_cluster in ocfs2_cluster_info may not be null
2213 		 * terminated, so make sure no overflow happens here by using
2214 		 * memcpy. Destination strings will always be null terminated
2215 		 * because osb is allocated using kzalloc.
2216 		 */
2217 		osb->osb_stackflags =
2218 			OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2219 		memcpy(osb->osb_cluster_stack,
2220 		       OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2221 		       OCFS2_STACK_LABEL_LEN);
2222 		if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2223 			mlog(ML_ERROR,
2224 			     "couldn't mount because of an invalid "
2225 			     "cluster stack label (%s) \n",
2226 			     osb->osb_cluster_stack);
2227 			status = -EINVAL;
2228 			goto bail;
2229 		}
2230 		memcpy(osb->osb_cluster_name,
2231 			OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster,
2232 			OCFS2_CLUSTER_NAME_LEN);
2233 	} else {
2234 		/* The empty string is identical with classic tools that
2235 		 * don't know about s_cluster_info. */
2236 		osb->osb_cluster_stack[0] = '\0';
2237 	}
2238 
2239 	get_random_bytes(&osb->s_next_generation, sizeof(u32));
2240 
2241 	/* FIXME
2242 	 * This should be done in ocfs2_journal_init(), but unknown
2243 	 * ordering issues will cause the filesystem to crash.
2244 	 * If anyone wants to figure out what part of the code
2245 	 * refers to osb->journal before ocfs2_journal_init() is run,
2246 	 * be my guest.
2247 	 */
2248 	/* initialize our journal structure */
2249 
2250 	journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2251 	if (!journal) {
2252 		mlog(ML_ERROR, "unable to alloc journal\n");
2253 		status = -ENOMEM;
2254 		goto bail;
2255 	}
2256 	osb->journal = journal;
2257 	journal->j_osb = osb;
2258 
2259 	atomic_set(&journal->j_num_trans, 0);
2260 	init_rwsem(&journal->j_trans_barrier);
2261 	init_waitqueue_head(&journal->j_checkpointed);
2262 	spin_lock_init(&journal->j_lock);
2263 	journal->j_trans_id = (unsigned long) 1;
2264 	INIT_LIST_HEAD(&journal->j_la_cleanups);
2265 	INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2266 	journal->j_state = OCFS2_JOURNAL_FREE;
2267 
2268 	INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs);
2269 	init_llist_head(&osb->dquot_drop_list);
2270 
2271 	/* get some pseudo constants for clustersize bits */
2272 	osb->s_clustersize_bits =
2273 		le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2274 	osb->s_clustersize = 1 << osb->s_clustersize_bits;
2275 
2276 	if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2277 	    osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2278 		mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2279 		     osb->s_clustersize);
2280 		status = -EINVAL;
2281 		goto bail;
2282 	}
2283 
2284 	total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2285 						le32_to_cpu(di->i_clusters));
2286 
2287 	status = generic_check_addressable(osb->sb->s_blocksize_bits,
2288 					   total_blocks);
2289 	if (status) {
2290 		mlog(ML_ERROR, "Volume too large "
2291 		     "to mount safely on this system");
2292 		status = -EFBIG;
2293 		goto bail;
2294 	}
2295 
2296 	if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2297 				 sizeof(di->id2.i_super.s_uuid))) {
2298 		mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2299 		status = -ENOMEM;
2300 		goto bail;
2301 	}
2302 
2303 	strlcpy(osb->vol_label, di->id2.i_super.s_label,
2304 		OCFS2_MAX_VOL_LABEL_LEN);
2305 	osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2306 	osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2307 	osb->first_cluster_group_blkno =
2308 		le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2309 	osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2310 	osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2311 	trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str,
2312 				     (unsigned long long)osb->root_blkno,
2313 				     (unsigned long long)osb->system_dir_blkno,
2314 				     osb->s_clustersize_bits);
2315 
2316 	osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2317 	if (!osb->osb_dlm_debug) {
2318 		status = -ENOMEM;
2319 		mlog_errno(status);
2320 		goto bail;
2321 	}
2322 
2323 	atomic_set(&osb->vol_state, VOLUME_INIT);
2324 
2325 	/* load root, system_dir, and all global system inodes */
2326 	status = ocfs2_init_global_system_inodes(osb);
2327 	if (status < 0) {
2328 		mlog_errno(status);
2329 		goto bail;
2330 	}
2331 
2332 	/*
2333 	 * global bitmap
2334 	 */
2335 	inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2336 					    OCFS2_INVALID_SLOT);
2337 	if (!inode) {
2338 		status = -EINVAL;
2339 		mlog_errno(status);
2340 		goto bail;
2341 	}
2342 
2343 	osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2344 	osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2345 	iput(inode);
2346 
2347 	osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2348 				 osb->s_feature_incompat) * 8;
2349 
2350 	status = ocfs2_init_slot_info(osb);
2351 	if (status < 0) {
2352 		mlog_errno(status);
2353 		goto bail;
2354 	}
2355 	cleancache_init_shared_fs(sb);
2356 
2357 bail:
2358 	return status;
2359 }
2360 
2361 /*
2362  * will return: -EAGAIN if it is ok to keep searching for superblocks
2363  *              -EINVAL if there is a bad superblock
2364  *              0 on success
2365  */
ocfs2_verify_volume(struct ocfs2_dinode * di,struct buffer_head * bh,u32 blksz,struct ocfs2_blockcheck_stats * stats)2366 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2367 			       struct buffer_head *bh,
2368 			       u32 blksz,
2369 			       struct ocfs2_blockcheck_stats *stats)
2370 {
2371 	int status = -EAGAIN;
2372 
2373 	if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2374 		   strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2375 		/* We have to do a raw check of the feature here */
2376 		if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2377 		    OCFS2_FEATURE_INCOMPAT_META_ECC) {
2378 			status = ocfs2_block_check_validate(bh->b_data,
2379 							    bh->b_size,
2380 							    &di->i_check,
2381 							    stats);
2382 			if (status)
2383 				goto out;
2384 		}
2385 		status = -EINVAL;
2386 		if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2387 			mlog(ML_ERROR, "found superblock with incorrect block "
2388 			     "size: found %u, should be %u\n",
2389 			     1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2390 			       blksz);
2391 		} else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2392 			   OCFS2_MAJOR_REV_LEVEL ||
2393 			   le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2394 			   OCFS2_MINOR_REV_LEVEL) {
2395 			mlog(ML_ERROR, "found superblock with bad version: "
2396 			     "found %u.%u, should be %u.%u\n",
2397 			     le16_to_cpu(di->id2.i_super.s_major_rev_level),
2398 			     le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2399 			     OCFS2_MAJOR_REV_LEVEL,
2400 			     OCFS2_MINOR_REV_LEVEL);
2401 		} else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2402 			mlog(ML_ERROR, "bad block number on superblock: "
2403 			     "found %llu, should be %llu\n",
2404 			     (unsigned long long)le64_to_cpu(di->i_blkno),
2405 			     (unsigned long long)bh->b_blocknr);
2406 		} else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2407 			    le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2408 			mlog(ML_ERROR, "bad cluster size found: %u\n",
2409 			     1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2410 		} else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2411 			mlog(ML_ERROR, "bad root_blkno: 0\n");
2412 		} else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2413 			mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2414 		} else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2415 			mlog(ML_ERROR,
2416 			     "Superblock slots found greater than file system "
2417 			     "maximum: found %u, max %u\n",
2418 			     le16_to_cpu(di->id2.i_super.s_max_slots),
2419 			     OCFS2_MAX_SLOTS);
2420 		} else {
2421 			/* found it! */
2422 			status = 0;
2423 		}
2424 	}
2425 
2426 out:
2427 	if (status && status != -EAGAIN)
2428 		mlog_errno(status);
2429 	return status;
2430 }
2431 
ocfs2_check_volume(struct ocfs2_super * osb)2432 static int ocfs2_check_volume(struct ocfs2_super *osb)
2433 {
2434 	int status;
2435 	int dirty;
2436 	int local;
2437 	struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2438 						  * recover
2439 						  * ourselves. */
2440 
2441 	/* Init our journal object. */
2442 	status = ocfs2_journal_init(osb->journal, &dirty);
2443 	if (status < 0) {
2444 		mlog(ML_ERROR, "Could not initialize journal!\n");
2445 		goto finally;
2446 	}
2447 
2448 	/* Now that journal has been initialized, check to make sure
2449 	   entire volume is addressable. */
2450 	status = ocfs2_journal_addressable(osb);
2451 	if (status)
2452 		goto finally;
2453 
2454 	/* If the journal was unmounted cleanly then we don't want to
2455 	 * recover anything. Otherwise, journal_load will do that
2456 	 * dirty work for us :) */
2457 	if (!dirty) {
2458 		status = ocfs2_journal_wipe(osb->journal, 0);
2459 		if (status < 0) {
2460 			mlog_errno(status);
2461 			goto finally;
2462 		}
2463 	} else {
2464 		printk(KERN_NOTICE "ocfs2: File system on device (%s) was not "
2465 		       "unmounted cleanly, recovering it.\n", osb->dev_str);
2466 	}
2467 
2468 	local = ocfs2_mount_local(osb);
2469 
2470 	/* will play back anything left in the journal. */
2471 	status = ocfs2_journal_load(osb->journal, local, dirty);
2472 	if (status < 0) {
2473 		mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2474 		goto finally;
2475 	}
2476 
2477 	if (osb->s_mount_opt & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
2478 		jbd2_journal_set_features(osb->journal->j_journal,
2479 				JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2480 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2481 	else
2482 		jbd2_journal_clear_features(osb->journal->j_journal,
2483 				JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2484 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2485 
2486 	if (dirty) {
2487 		/* recover my local alloc if we didn't unmount cleanly. */
2488 		status = ocfs2_begin_local_alloc_recovery(osb,
2489 							  osb->slot_num,
2490 							  &local_alloc);
2491 		if (status < 0) {
2492 			mlog_errno(status);
2493 			goto finally;
2494 		}
2495 		/* we complete the recovery process after we've marked
2496 		 * ourselves as mounted. */
2497 	}
2498 
2499 	status = ocfs2_load_local_alloc(osb);
2500 	if (status < 0) {
2501 		mlog_errno(status);
2502 		goto finally;
2503 	}
2504 
2505 	if (dirty) {
2506 		/* Recovery will be completed after we've mounted the
2507 		 * rest of the volume. */
2508 		osb->dirty = 1;
2509 		osb->local_alloc_copy = local_alloc;
2510 		local_alloc = NULL;
2511 	}
2512 
2513 	/* go through each journal, trylock it and if you get the
2514 	 * lock, and it's marked as dirty, set the bit in the recover
2515 	 * map and launch a recovery thread for it. */
2516 	status = ocfs2_mark_dead_nodes(osb);
2517 	if (status < 0) {
2518 		mlog_errno(status);
2519 		goto finally;
2520 	}
2521 
2522 	status = ocfs2_compute_replay_slots(osb);
2523 	if (status < 0)
2524 		mlog_errno(status);
2525 
2526 finally:
2527 	kfree(local_alloc);
2528 
2529 	if (status)
2530 		mlog_errno(status);
2531 	return status;
2532 }
2533 
2534 /*
2535  * The routine gets called from dismount or close whenever a dismount on
2536  * volume is requested and the osb open count becomes 1.
2537  * It will remove the osb from the global list and also free up all the
2538  * initialized resources and fileobject.
2539  */
ocfs2_delete_osb(struct ocfs2_super * osb)2540 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2541 {
2542 	/* This function assumes that the caller has the main osb resource */
2543 
2544 	ocfs2_free_slot_info(osb);
2545 
2546 	kfree(osb->osb_orphan_wipes);
2547 	kfree(osb->slot_recovery_generations);
2548 	/* FIXME
2549 	 * This belongs in journal shutdown, but because we have to
2550 	 * allocate osb->journal at the start of ocfs2_initialize_osb(),
2551 	 * we free it here.
2552 	 */
2553 	kfree(osb->journal);
2554 	kfree(osb->local_alloc_copy);
2555 	kfree(osb->uuid_str);
2556 	kfree(osb->vol_label);
2557 	ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2558 	memset(osb, 0, sizeof(struct ocfs2_super));
2559 }
2560 
2561 /* Depending on the mount option passed, perform one of the following:
2562  * Put OCFS2 into a readonly state (default)
2563  * Return EIO so that only the process errs
2564  * Fix the error as if fsck.ocfs2 -y
2565  * panic
2566  */
ocfs2_handle_error(struct super_block * sb)2567 static int ocfs2_handle_error(struct super_block *sb)
2568 {
2569 	struct ocfs2_super *osb = OCFS2_SB(sb);
2570 	int rv = 0;
2571 
2572 	ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2573 	pr_crit("On-disk corruption discovered. "
2574 		"Please run fsck.ocfs2 once the filesystem is unmounted.\n");
2575 
2576 	if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC) {
2577 		panic("OCFS2: (device %s): panic forced after error\n",
2578 		      sb->s_id);
2579 	} else if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_CONT) {
2580 		pr_crit("OCFS2: Returning error to the calling process.\n");
2581 		rv = -EIO;
2582 	} else { /* default option */
2583 		rv = -EROFS;
2584 		if (sb->s_flags & MS_RDONLY &&
2585 				(ocfs2_is_soft_readonly(osb) ||
2586 				 ocfs2_is_hard_readonly(osb)))
2587 			return rv;
2588 
2589 		pr_crit("OCFS2: File system is now read-only.\n");
2590 		sb->s_flags |= MS_RDONLY;
2591 		ocfs2_set_ro_flag(osb, 0);
2592 	}
2593 
2594 	return rv;
2595 }
2596 
__ocfs2_error(struct super_block * sb,const char * function,const char * fmt,...)2597 int __ocfs2_error(struct super_block *sb, const char *function,
2598 		  const char *fmt, ...)
2599 {
2600 	struct va_format vaf;
2601 	va_list args;
2602 
2603 	va_start(args, fmt);
2604 	vaf.fmt = fmt;
2605 	vaf.va = &args;
2606 
2607 	/* Not using mlog here because we want to show the actual
2608 	 * function the error came from. */
2609 	printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %pV",
2610 	       sb->s_id, function, &vaf);
2611 
2612 	va_end(args);
2613 
2614 	return ocfs2_handle_error(sb);
2615 }
2616 
2617 /* Handle critical errors. This is intentionally more drastic than
2618  * ocfs2_handle_error, so we only use for things like journal errors,
2619  * etc. */
__ocfs2_abort(struct super_block * sb,const char * function,const char * fmt,...)2620 void __ocfs2_abort(struct super_block *sb, const char *function,
2621 		   const char *fmt, ...)
2622 {
2623 	struct va_format vaf;
2624 	va_list args;
2625 
2626 	va_start(args, fmt);
2627 
2628 	vaf.fmt = fmt;
2629 	vaf.va = &args;
2630 
2631 	printk(KERN_CRIT "OCFS2: abort (device %s): %s: %pV",
2632 	       sb->s_id, function, &vaf);
2633 
2634 	va_end(args);
2635 
2636 	/* We don't have the cluster support yet to go straight to
2637 	 * hard readonly in here. Until then, we want to keep
2638 	 * ocfs2_abort() so that we can at least mark critical
2639 	 * errors.
2640 	 *
2641 	 * TODO: This should abort the journal and alert other nodes
2642 	 * that our slot needs recovery. */
2643 
2644 	/* Force a panic(). This stinks, but it's better than letting
2645 	 * things continue without having a proper hard readonly
2646 	 * here. */
2647 	if (!ocfs2_mount_local(OCFS2_SB(sb)))
2648 		OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2649 	ocfs2_handle_error(sb);
2650 }
2651 
2652 /*
2653  * Void signal blockers, because in-kernel sigprocmask() only fails
2654  * when SIG_* is wrong.
2655  */
ocfs2_block_signals(sigset_t * oldset)2656 void ocfs2_block_signals(sigset_t *oldset)
2657 {
2658 	int rc;
2659 	sigset_t blocked;
2660 
2661 	sigfillset(&blocked);
2662 	rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2663 	BUG_ON(rc);
2664 }
2665 
ocfs2_unblock_signals(sigset_t * oldset)2666 void ocfs2_unblock_signals(sigset_t *oldset)
2667 {
2668 	int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2669 	BUG_ON(rc);
2670 }
2671 
2672 module_init(ocfs2_init);
2673 module_exit(ocfs2_exit);
2674