• 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 	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\n", "Commit",
290 			(osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
291 			osb->osb_commit_interval);
292 
293 	out += snprintf(buf + out, len - out,
294 			"%10s => State: %d  TxnId: %lu  NumTxns: %d\n",
295 			"Journal", osb->journal->j_state,
296 			osb->journal->j_trans_id,
297 			atomic_read(&osb->journal->j_num_trans));
298 
299 	out += snprintf(buf + out, len - out,
300 			"%10s => GlobalAllocs: %d  LocalAllocs: %d  "
301 			"SubAllocs: %d  LAWinMoves: %d  SAExtends: %d\n",
302 			"Stats",
303 			atomic_read(&osb->alloc_stats.bitmap_data),
304 			atomic_read(&osb->alloc_stats.local_data),
305 			atomic_read(&osb->alloc_stats.bg_allocs),
306 			atomic_read(&osb->alloc_stats.moves),
307 			atomic_read(&osb->alloc_stats.bg_extends));
308 
309 	out += snprintf(buf + out, len - out,
310 			"%10s => State: %u  Descriptor: %llu  Size: %u bits  "
311 			"Default: %u bits\n",
312 			"LocalAlloc", osb->local_alloc_state,
313 			(unsigned long long)osb->la_last_gd,
314 			osb->local_alloc_bits, osb->local_alloc_default_bits);
315 
316 	spin_lock(&osb->osb_lock);
317 	out += snprintf(buf + out, len - out,
318 			"%10s => InodeSlot: %d  StolenInodes: %d, "
319 			"MetaSlot: %d  StolenMeta: %d\n", "Steal",
320 			osb->s_inode_steal_slot,
321 			atomic_read(&osb->s_num_inodes_stolen),
322 			osb->s_meta_steal_slot,
323 			atomic_read(&osb->s_num_meta_stolen));
324 	spin_unlock(&osb->osb_lock);
325 
326 	out += snprintf(buf + out, len - out, "OrphanScan => ");
327 	out += snprintf(buf + out, len - out, "Local: %u  Global: %u ",
328 			os->os_count, os->os_seqno);
329 	out += snprintf(buf + out, len - out, " Last Scan: ");
330 	if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
331 		out += snprintf(buf + out, len - out, "Disabled\n");
332 	else
333 		out += snprintf(buf + out, len - out, "%lu seconds ago\n",
334 				(get_seconds() - os->os_scantime.tv_sec));
335 
336 	out += snprintf(buf + out, len - out, "%10s => %3s  %10s\n",
337 			"Slots", "Num", "RecoGen");
338 	for (i = 0; i < osb->max_slots; ++i) {
339 		out += snprintf(buf + out, len - out,
340 				"%10s  %c %3d  %10d\n",
341 				" ",
342 				(i == osb->slot_num ? '*' : ' '),
343 				i, osb->slot_recovery_generations[i]);
344 	}
345 
346 	return out;
347 }
348 
ocfs2_osb_debug_open(struct inode * inode,struct file * file)349 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
350 {
351 	struct ocfs2_super *osb = inode->i_private;
352 	char *buf = NULL;
353 
354 	buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
355 	if (!buf)
356 		goto bail;
357 
358 	i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
359 
360 	file->private_data = buf;
361 
362 	return 0;
363 bail:
364 	return -ENOMEM;
365 }
366 
ocfs2_debug_release(struct inode * inode,struct file * file)367 static int ocfs2_debug_release(struct inode *inode, struct file *file)
368 {
369 	kfree(file->private_data);
370 	return 0;
371 }
372 
ocfs2_debug_read(struct file * file,char __user * buf,size_t nbytes,loff_t * ppos)373 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
374 				size_t nbytes, loff_t *ppos)
375 {
376 	return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
377 				       i_size_read(file->f_mapping->host));
378 }
379 #else
ocfs2_osb_debug_open(struct inode * inode,struct file * file)380 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
381 {
382 	return 0;
383 }
ocfs2_debug_release(struct inode * inode,struct file * file)384 static int ocfs2_debug_release(struct inode *inode, struct file *file)
385 {
386 	return 0;
387 }
ocfs2_debug_read(struct file * file,char __user * buf,size_t nbytes,loff_t * ppos)388 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
389 				size_t nbytes, loff_t *ppos)
390 {
391 	return 0;
392 }
393 #endif	/* CONFIG_DEBUG_FS */
394 
395 static const struct file_operations ocfs2_osb_debug_fops = {
396 	.open =		ocfs2_osb_debug_open,
397 	.release =	ocfs2_debug_release,
398 	.read =		ocfs2_debug_read,
399 	.llseek =	generic_file_llseek,
400 };
401 
ocfs2_sync_fs(struct super_block * sb,int wait)402 static int ocfs2_sync_fs(struct super_block *sb, int wait)
403 {
404 	int status;
405 	tid_t target;
406 	struct ocfs2_super *osb = OCFS2_SB(sb);
407 
408 	if (ocfs2_is_hard_readonly(osb))
409 		return -EROFS;
410 
411 	if (wait) {
412 		status = ocfs2_flush_truncate_log(osb);
413 		if (status < 0)
414 			mlog_errno(status);
415 	} else {
416 		ocfs2_schedule_truncate_log_flush(osb, 0);
417 	}
418 
419 	if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
420 				      &target)) {
421 		if (wait)
422 			jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
423 					     target);
424 	}
425 	return 0;
426 }
427 
ocfs2_need_system_inode(struct ocfs2_super * osb,int ino)428 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
429 {
430 	if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
431 	    && (ino == USER_QUOTA_SYSTEM_INODE
432 		|| ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
433 		return 0;
434 	if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
435 	    && (ino == GROUP_QUOTA_SYSTEM_INODE
436 		|| ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
437 		return 0;
438 	return 1;
439 }
440 
ocfs2_init_global_system_inodes(struct ocfs2_super * osb)441 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
442 {
443 	struct inode *new = NULL;
444 	int status = 0;
445 	int i;
446 
447 	new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
448 	if (IS_ERR(new)) {
449 		status = PTR_ERR(new);
450 		mlog_errno(status);
451 		goto bail;
452 	}
453 	osb->root_inode = new;
454 
455 	new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
456 	if (IS_ERR(new)) {
457 		status = PTR_ERR(new);
458 		mlog_errno(status);
459 		goto bail;
460 	}
461 	osb->sys_root_inode = new;
462 
463 	for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
464 	     i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
465 		if (!ocfs2_need_system_inode(osb, i))
466 			continue;
467 		new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
468 		if (!new) {
469 			ocfs2_release_system_inodes(osb);
470 			status = -EINVAL;
471 			mlog_errno(status);
472 			/* FIXME: Should ERROR_RO_FS */
473 			mlog(ML_ERROR, "Unable to load system inode %d, "
474 			     "possibly corrupt fs?", i);
475 			goto bail;
476 		}
477 		// the array now has one ref, so drop this one
478 		iput(new);
479 	}
480 
481 bail:
482 	if (status)
483 		mlog_errno(status);
484 	return status;
485 }
486 
ocfs2_init_local_system_inodes(struct ocfs2_super * osb)487 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
488 {
489 	struct inode *new = NULL;
490 	int status = 0;
491 	int i;
492 
493 	for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
494 	     i < NUM_SYSTEM_INODES;
495 	     i++) {
496 		if (!ocfs2_need_system_inode(osb, i))
497 			continue;
498 		new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
499 		if (!new) {
500 			ocfs2_release_system_inodes(osb);
501 			status = -EINVAL;
502 			mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
503 			     status, i, osb->slot_num);
504 			goto bail;
505 		}
506 		/* the array now has one ref, so drop this one */
507 		iput(new);
508 	}
509 
510 bail:
511 	if (status)
512 		mlog_errno(status);
513 	return status;
514 }
515 
ocfs2_release_system_inodes(struct ocfs2_super * osb)516 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
517 {
518 	int i;
519 	struct inode *inode;
520 
521 	for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
522 		inode = osb->global_system_inodes[i];
523 		if (inode) {
524 			iput(inode);
525 			osb->global_system_inodes[i] = NULL;
526 		}
527 	}
528 
529 	inode = osb->sys_root_inode;
530 	if (inode) {
531 		iput(inode);
532 		osb->sys_root_inode = NULL;
533 	}
534 
535 	inode = osb->root_inode;
536 	if (inode) {
537 		iput(inode);
538 		osb->root_inode = NULL;
539 	}
540 
541 	if (!osb->local_system_inodes)
542 		return;
543 
544 	for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
545 		if (osb->local_system_inodes[i]) {
546 			iput(osb->local_system_inodes[i]);
547 			osb->local_system_inodes[i] = NULL;
548 		}
549 	}
550 
551 	kfree(osb->local_system_inodes);
552 	osb->local_system_inodes = NULL;
553 }
554 
555 /* We're allocating fs objects, use GFP_NOFS */
ocfs2_alloc_inode(struct super_block * sb)556 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
557 {
558 	struct ocfs2_inode_info *oi;
559 
560 	oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
561 	if (!oi)
562 		return NULL;
563 
564 	oi->i_sync_tid = 0;
565 	oi->i_datasync_tid = 0;
566 
567 	jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
568 	return &oi->vfs_inode;
569 }
570 
ocfs2_i_callback(struct rcu_head * head)571 static void ocfs2_i_callback(struct rcu_head *head)
572 {
573 	struct inode *inode = container_of(head, struct inode, i_rcu);
574 	kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
575 }
576 
ocfs2_destroy_inode(struct inode * inode)577 static void ocfs2_destroy_inode(struct inode *inode)
578 {
579 	call_rcu(&inode->i_rcu, ocfs2_i_callback);
580 }
581 
ocfs2_max_file_offset(unsigned int bbits,unsigned int cbits)582 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
583 						unsigned int cbits)
584 {
585 	unsigned int bytes = 1 << cbits;
586 	unsigned int trim = bytes;
587 	unsigned int bitshift = 32;
588 
589 	/*
590 	 * i_size and all block offsets in ocfs2 are always 64 bits
591 	 * wide. i_clusters is 32 bits, in cluster-sized units. So on
592 	 * 64 bit platforms, cluster size will be the limiting factor.
593 	 */
594 
595 #if BITS_PER_LONG == 32
596 # if defined(CONFIG_LBDAF)
597 	BUILD_BUG_ON(sizeof(sector_t) != 8);
598 	/*
599 	 * We might be limited by page cache size.
600 	 */
601 	if (bytes > PAGE_CACHE_SIZE) {
602 		bytes = PAGE_CACHE_SIZE;
603 		trim = 1;
604 		/*
605 		 * Shift by 31 here so that we don't get larger than
606 		 * MAX_LFS_FILESIZE
607 		 */
608 		bitshift = 31;
609 	}
610 # else
611 	/*
612 	 * We are limited by the size of sector_t. Use block size, as
613 	 * that's what we expose to the VFS.
614 	 */
615 	bytes = 1 << bbits;
616 	trim = 1;
617 	bitshift = 31;
618 # endif
619 #endif
620 
621 	/*
622 	 * Trim by a whole cluster when we can actually approach the
623 	 * on-disk limits. Otherwise we can overflow i_clusters when
624 	 * an extent start is at the max offset.
625 	 */
626 	return (((unsigned long long)bytes) << bitshift) - trim;
627 }
628 
ocfs2_remount(struct super_block * sb,int * flags,char * data)629 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
630 {
631 	int incompat_features;
632 	int ret = 0;
633 	struct mount_options parsed_options;
634 	struct ocfs2_super *osb = OCFS2_SB(sb);
635 	u32 tmp;
636 
637 	sync_filesystem(sb);
638 
639 	if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
640 	    !ocfs2_check_set_options(sb, &parsed_options)) {
641 		ret = -EINVAL;
642 		goto out;
643 	}
644 
645 	tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
646 		OCFS2_MOUNT_HB_NONE;
647 	if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
648 		ret = -EINVAL;
649 		mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
650 		goto out;
651 	}
652 
653 	if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
654 	    (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
655 		ret = -EINVAL;
656 		mlog(ML_ERROR, "Cannot change data mode on remount\n");
657 		goto out;
658 	}
659 
660 	/* Probably don't want this on remount; it might
661 	 * mess with other nodes */
662 	if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
663 	    (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
664 		ret = -EINVAL;
665 		mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
666 		goto out;
667 	}
668 
669 	/* We're going to/from readonly mode. */
670 	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
671 		/* Disable quota accounting before remounting RO */
672 		if (*flags & MS_RDONLY) {
673 			ret = ocfs2_susp_quotas(osb, 0);
674 			if (ret < 0)
675 				goto out;
676 		}
677 		/* Lock here so the check of HARD_RO and the potential
678 		 * setting of SOFT_RO is atomic. */
679 		spin_lock(&osb->osb_lock);
680 		if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
681 			mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
682 			ret = -EROFS;
683 			goto unlock_osb;
684 		}
685 
686 		if (*flags & MS_RDONLY) {
687 			sb->s_flags |= MS_RDONLY;
688 			osb->osb_flags |= OCFS2_OSB_SOFT_RO;
689 		} else {
690 			if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
691 				mlog(ML_ERROR, "Cannot remount RDWR "
692 				     "filesystem due to previous errors.\n");
693 				ret = -EROFS;
694 				goto unlock_osb;
695 			}
696 			incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
697 			if (incompat_features) {
698 				mlog(ML_ERROR, "Cannot remount RDWR because "
699 				     "of unsupported optional features "
700 				     "(%x).\n", incompat_features);
701 				ret = -EINVAL;
702 				goto unlock_osb;
703 			}
704 			sb->s_flags &= ~MS_RDONLY;
705 			osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
706 		}
707 		trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags);
708 unlock_osb:
709 		spin_unlock(&osb->osb_lock);
710 		/* Enable quota accounting after remounting RW */
711 		if (!ret && !(*flags & MS_RDONLY)) {
712 			if (sb_any_quota_suspended(sb))
713 				ret = ocfs2_susp_quotas(osb, 1);
714 			else
715 				ret = ocfs2_enable_quotas(osb);
716 			if (ret < 0) {
717 				/* Return back changes... */
718 				spin_lock(&osb->osb_lock);
719 				sb->s_flags |= MS_RDONLY;
720 				osb->osb_flags |= OCFS2_OSB_SOFT_RO;
721 				spin_unlock(&osb->osb_lock);
722 				goto out;
723 			}
724 		}
725 	}
726 
727 	if (!ret) {
728 		/* Only save off the new mount options in case of a successful
729 		 * remount. */
730 		osb->s_mount_opt = parsed_options.mount_opt;
731 		osb->s_atime_quantum = parsed_options.atime_quantum;
732 		osb->preferred_slot = parsed_options.slot;
733 		if (parsed_options.commit_interval)
734 			osb->osb_commit_interval = parsed_options.commit_interval;
735 
736 		if (!ocfs2_is_hard_readonly(osb))
737 			ocfs2_set_journal_params(osb);
738 
739 		sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
740 			((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
741 							MS_POSIXACL : 0);
742 	}
743 out:
744 	return ret;
745 }
746 
ocfs2_sb_probe(struct super_block * sb,struct buffer_head ** bh,int * sector_size,struct ocfs2_blockcheck_stats * stats)747 static int ocfs2_sb_probe(struct super_block *sb,
748 			  struct buffer_head **bh,
749 			  int *sector_size,
750 			  struct ocfs2_blockcheck_stats *stats)
751 {
752 	int status, tmpstat;
753 	struct ocfs1_vol_disk_hdr *hdr;
754 	struct ocfs2_dinode *di;
755 	int blksize;
756 
757 	*bh = NULL;
758 
759 	/* may be > 512 */
760 	*sector_size = bdev_logical_block_size(sb->s_bdev);
761 	if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
762 		mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
763 		     *sector_size, OCFS2_MAX_BLOCKSIZE);
764 		status = -EINVAL;
765 		goto bail;
766 	}
767 
768 	/* Can this really happen? */
769 	if (*sector_size < OCFS2_MIN_BLOCKSIZE)
770 		*sector_size = OCFS2_MIN_BLOCKSIZE;
771 
772 	/* check block zero for old format */
773 	status = ocfs2_get_sector(sb, bh, 0, *sector_size);
774 	if (status < 0) {
775 		mlog_errno(status);
776 		goto bail;
777 	}
778 	hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
779 	if (hdr->major_version == OCFS1_MAJOR_VERSION) {
780 		mlog(ML_ERROR, "incompatible version: %u.%u\n",
781 		     hdr->major_version, hdr->minor_version);
782 		status = -EINVAL;
783 	}
784 	if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
785 		   strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
786 		mlog(ML_ERROR, "incompatible volume signature: %8s\n",
787 		     hdr->signature);
788 		status = -EINVAL;
789 	}
790 	brelse(*bh);
791 	*bh = NULL;
792 	if (status < 0) {
793 		mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
794 		     "upgraded before mounting with ocfs v2\n");
795 		goto bail;
796 	}
797 
798 	/*
799 	 * Now check at magic offset for 512, 1024, 2048, 4096
800 	 * blocksizes.  4096 is the maximum blocksize because it is
801 	 * the minimum clustersize.
802 	 */
803 	status = -EINVAL;
804 	for (blksize = *sector_size;
805 	     blksize <= OCFS2_MAX_BLOCKSIZE;
806 	     blksize <<= 1) {
807 		tmpstat = ocfs2_get_sector(sb, bh,
808 					   OCFS2_SUPER_BLOCK_BLKNO,
809 					   blksize);
810 		if (tmpstat < 0) {
811 			status = tmpstat;
812 			mlog_errno(status);
813 			break;
814 		}
815 		di = (struct ocfs2_dinode *) (*bh)->b_data;
816 		memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
817 		spin_lock_init(&stats->b_lock);
818 		tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
819 		if (tmpstat < 0) {
820 			brelse(*bh);
821 			*bh = NULL;
822 		}
823 		if (tmpstat != -EAGAIN) {
824 			status = tmpstat;
825 			break;
826 		}
827 	}
828 
829 bail:
830 	return status;
831 }
832 
ocfs2_verify_heartbeat(struct ocfs2_super * osb)833 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
834 {
835 	u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
836 
837 	if (osb->s_mount_opt & hb_enabled) {
838 		if (ocfs2_mount_local(osb)) {
839 			mlog(ML_ERROR, "Cannot heartbeat on a locally "
840 			     "mounted device.\n");
841 			return -EINVAL;
842 		}
843 		if (ocfs2_userspace_stack(osb)) {
844 			mlog(ML_ERROR, "Userspace stack expected, but "
845 			     "o2cb heartbeat arguments passed to mount\n");
846 			return -EINVAL;
847 		}
848 		if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
849 		     !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
850 		    ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
851 		     ocfs2_cluster_o2cb_global_heartbeat(osb))) {
852 			mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
853 			return -EINVAL;
854 		}
855 	}
856 
857 	if (!(osb->s_mount_opt & hb_enabled)) {
858 		if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
859 		    !ocfs2_userspace_stack(osb)) {
860 			mlog(ML_ERROR, "Heartbeat has to be started to mount "
861 			     "a read-write clustered device.\n");
862 			return -EINVAL;
863 		}
864 	}
865 
866 	return 0;
867 }
868 
869 /*
870  * If we're using a userspace stack, mount should have passed
871  * a name that matches the disk.  If not, mount should not
872  * have passed a stack.
873  */
ocfs2_verify_userspace_stack(struct ocfs2_super * osb,struct mount_options * mopt)874 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
875 					struct mount_options *mopt)
876 {
877 	if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
878 		mlog(ML_ERROR,
879 		     "cluster stack passed to mount, but this filesystem "
880 		     "does not support it\n");
881 		return -EINVAL;
882 	}
883 
884 	if (ocfs2_userspace_stack(osb) &&
885 	    strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
886 		    OCFS2_STACK_LABEL_LEN)) {
887 		mlog(ML_ERROR,
888 		     "cluster stack passed to mount (\"%s\") does not "
889 		     "match the filesystem (\"%s\")\n",
890 		     mopt->cluster_stack,
891 		     osb->osb_cluster_stack);
892 		return -EINVAL;
893 	}
894 
895 	return 0;
896 }
897 
ocfs2_susp_quotas(struct ocfs2_super * osb,int unsuspend)898 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
899 {
900 	int type;
901 	struct super_block *sb = osb->sb;
902 	unsigned int feature[OCFS2_MAXQUOTAS] = {
903 					OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
904 					OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
905 	int status = 0;
906 
907 	for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
908 		if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
909 			continue;
910 		if (unsuspend)
911 			status = dquot_resume(sb, type);
912 		else {
913 			struct ocfs2_mem_dqinfo *oinfo;
914 
915 			/* Cancel periodic syncing before suspending */
916 			oinfo = sb_dqinfo(sb, type)->dqi_priv;
917 			cancel_delayed_work_sync(&oinfo->dqi_sync_work);
918 			status = dquot_suspend(sb, type);
919 		}
920 		if (status < 0)
921 			break;
922 	}
923 	if (status < 0)
924 		mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
925 		     "remount (error = %d).\n", status);
926 	return status;
927 }
928 
ocfs2_enable_quotas(struct ocfs2_super * osb)929 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
930 {
931 	struct inode *inode[OCFS2_MAXQUOTAS] = { NULL, NULL };
932 	struct super_block *sb = osb->sb;
933 	unsigned int feature[OCFS2_MAXQUOTAS] = {
934 					OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
935 					OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
936 	unsigned int ino[OCFS2_MAXQUOTAS] = {
937 					LOCAL_USER_QUOTA_SYSTEM_INODE,
938 					LOCAL_GROUP_QUOTA_SYSTEM_INODE };
939 	int status;
940 	int type;
941 
942 	sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
943 	for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
944 		if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
945 			continue;
946 		inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
947 							osb->slot_num);
948 		if (!inode[type]) {
949 			status = -ENOENT;
950 			goto out_quota_off;
951 		}
952 		status = dquot_enable(inode[type], type, QFMT_OCFS2,
953 				      DQUOT_USAGE_ENABLED);
954 		if (status < 0)
955 			goto out_quota_off;
956 	}
957 
958 	for (type = 0; type < OCFS2_MAXQUOTAS; type++)
959 		iput(inode[type]);
960 	return 0;
961 out_quota_off:
962 	ocfs2_disable_quotas(osb);
963 	for (type = 0; type < OCFS2_MAXQUOTAS; type++)
964 		iput(inode[type]);
965 	mlog_errno(status);
966 	return status;
967 }
968 
ocfs2_disable_quotas(struct ocfs2_super * osb)969 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
970 {
971 	int type;
972 	struct inode *inode;
973 	struct super_block *sb = osb->sb;
974 	struct ocfs2_mem_dqinfo *oinfo;
975 
976 	/* We mostly ignore errors in this function because there's not much
977 	 * we can do when we see them */
978 	for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
979 		if (!sb_has_quota_loaded(sb, type))
980 			continue;
981 		/* Cancel periodic syncing before we grab dqonoff_mutex */
982 		oinfo = sb_dqinfo(sb, type)->dqi_priv;
983 		cancel_delayed_work_sync(&oinfo->dqi_sync_work);
984 		inode = igrab(sb->s_dquot.files[type]);
985 		/* Turn off quotas. This will remove all dquot structures from
986 		 * memory and so they will be automatically synced to global
987 		 * quota files */
988 		dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
989 					DQUOT_LIMITS_ENABLED);
990 		if (!inode)
991 			continue;
992 		iput(inode);
993 	}
994 }
995 
996 /* Handle quota on quotactl */
ocfs2_quota_on(struct super_block * sb,int type,int format_id)997 static int ocfs2_quota_on(struct super_block *sb, int type, int format_id)
998 {
999 	unsigned int feature[OCFS2_MAXQUOTAS] = {
1000 					OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
1001 					OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
1002 
1003 	if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
1004 		return -EINVAL;
1005 
1006 	return dquot_enable(sb_dqopt(sb)->files[type], type,
1007 			    format_id, DQUOT_LIMITS_ENABLED);
1008 }
1009 
1010 /* Handle quota off quotactl */
ocfs2_quota_off(struct super_block * sb,int type)1011 static int ocfs2_quota_off(struct super_block *sb, int type)
1012 {
1013 	return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
1014 }
1015 
1016 static const struct quotactl_ops ocfs2_quotactl_ops = {
1017 	.quota_on_meta	= ocfs2_quota_on,
1018 	.quota_off	= ocfs2_quota_off,
1019 	.quota_sync	= dquot_quota_sync,
1020 	.get_info	= dquot_get_dqinfo,
1021 	.set_info	= dquot_set_dqinfo,
1022 	.get_dqblk	= dquot_get_dqblk,
1023 	.set_dqblk	= dquot_set_dqblk,
1024 };
1025 
ocfs2_fill_super(struct super_block * sb,void * data,int silent)1026 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
1027 {
1028 	struct dentry *root;
1029 	int status, sector_size;
1030 	struct mount_options parsed_options;
1031 	struct inode *inode = NULL;
1032 	struct ocfs2_super *osb = NULL;
1033 	struct buffer_head *bh = NULL;
1034 	char nodestr[12];
1035 	struct ocfs2_blockcheck_stats stats;
1036 
1037 	trace_ocfs2_fill_super(sb, data, silent);
1038 
1039 	if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
1040 		status = -EINVAL;
1041 		goto read_super_error;
1042 	}
1043 
1044 	/* probe for superblock */
1045 	status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
1046 	if (status < 0) {
1047 		mlog(ML_ERROR, "superblock probe failed!\n");
1048 		goto read_super_error;
1049 	}
1050 
1051 	status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1052 	osb = OCFS2_SB(sb);
1053 	if (status < 0) {
1054 		mlog_errno(status);
1055 		goto read_super_error;
1056 	}
1057 	brelse(bh);
1058 	bh = NULL;
1059 
1060 	if (!ocfs2_check_set_options(sb, &parsed_options)) {
1061 		status = -EINVAL;
1062 		goto read_super_error;
1063 	}
1064 	osb->s_mount_opt = parsed_options.mount_opt;
1065 	osb->s_atime_quantum = parsed_options.atime_quantum;
1066 	osb->preferred_slot = parsed_options.slot;
1067 	osb->osb_commit_interval = parsed_options.commit_interval;
1068 
1069 	ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1070 	osb->osb_resv_level = parsed_options.resv_level;
1071 	osb->osb_dir_resv_level = parsed_options.resv_level;
1072 	if (parsed_options.dir_resv_level == -1)
1073 		osb->osb_dir_resv_level = parsed_options.resv_level;
1074 	else
1075 		osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1076 
1077 	status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1078 	if (status)
1079 		goto read_super_error;
1080 
1081 	sb->s_magic = OCFS2_SUPER_MAGIC;
1082 
1083 	sb->s_flags = (sb->s_flags & ~(MS_POSIXACL | MS_NOSEC)) |
1084 		((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1085 
1086 	/* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1087 	 * heartbeat=none */
1088 	if (bdev_read_only(sb->s_bdev)) {
1089 		if (!(sb->s_flags & MS_RDONLY)) {
1090 			status = -EACCES;
1091 			mlog(ML_ERROR, "Readonly device detected but readonly "
1092 			     "mount was not specified.\n");
1093 			goto read_super_error;
1094 		}
1095 
1096 		/* You should not be able to start a local heartbeat
1097 		 * on a readonly device. */
1098 		if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1099 			status = -EROFS;
1100 			mlog(ML_ERROR, "Local heartbeat specified on readonly "
1101 			     "device.\n");
1102 			goto read_super_error;
1103 		}
1104 
1105 		status = ocfs2_check_journals_nolocks(osb);
1106 		if (status < 0) {
1107 			if (status == -EROFS)
1108 				mlog(ML_ERROR, "Recovery required on readonly "
1109 				     "file system, but write access is "
1110 				     "unavailable.\n");
1111 			else
1112 				mlog_errno(status);
1113 			goto read_super_error;
1114 		}
1115 
1116 		ocfs2_set_ro_flag(osb, 1);
1117 
1118 		printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. "
1119 		       "Cluster services will not be used for this mount. "
1120 		       "Recovery will be skipped.\n", osb->dev_str);
1121 	}
1122 
1123 	if (!ocfs2_is_hard_readonly(osb)) {
1124 		if (sb->s_flags & MS_RDONLY)
1125 			ocfs2_set_ro_flag(osb, 0);
1126 	}
1127 
1128 	status = ocfs2_verify_heartbeat(osb);
1129 	if (status < 0) {
1130 		mlog_errno(status);
1131 		goto read_super_error;
1132 	}
1133 
1134 	osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1135 						 ocfs2_debugfs_root);
1136 	if (!osb->osb_debug_root) {
1137 		status = -EINVAL;
1138 		mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1139 		goto read_super_error;
1140 	}
1141 
1142 	osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1143 					    osb->osb_debug_root,
1144 					    osb,
1145 					    &ocfs2_osb_debug_fops);
1146 	if (!osb->osb_ctxt) {
1147 		status = -EINVAL;
1148 		mlog_errno(status);
1149 		goto read_super_error;
1150 	}
1151 
1152 	if (ocfs2_meta_ecc(osb)) {
1153 		status = ocfs2_blockcheck_stats_debugfs_install(
1154 						&osb->osb_ecc_stats,
1155 						osb->osb_debug_root);
1156 		if (status) {
1157 			mlog(ML_ERROR,
1158 			     "Unable to create blockcheck statistics "
1159 			     "files\n");
1160 			goto read_super_error;
1161 		}
1162 	}
1163 
1164 	status = ocfs2_mount_volume(sb);
1165 	if (status < 0)
1166 		goto read_super_error;
1167 
1168 	if (osb->root_inode)
1169 		inode = igrab(osb->root_inode);
1170 
1171 	if (!inode) {
1172 		status = -EIO;
1173 		mlog_errno(status);
1174 		goto read_super_error;
1175 	}
1176 
1177 	root = d_make_root(inode);
1178 	if (!root) {
1179 		status = -ENOMEM;
1180 		mlog_errno(status);
1181 		goto read_super_error;
1182 	}
1183 
1184 	sb->s_root = root;
1185 
1186 	ocfs2_complete_mount_recovery(osb);
1187 
1188 	if (ocfs2_mount_local(osb))
1189 		snprintf(nodestr, sizeof(nodestr), "local");
1190 	else
1191 		snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1192 
1193 	printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1194 	       "with %s data mode.\n",
1195 	       osb->dev_str, nodestr, osb->slot_num,
1196 	       osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1197 	       "ordered");
1198 
1199 	atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1200 	wake_up(&osb->osb_mount_event);
1201 
1202 	/* Now we can initialize quotas because we can afford to wait
1203 	 * for cluster locks recovery now. That also means that truncation
1204 	 * log recovery can happen but that waits for proper quota setup */
1205 	if (!(sb->s_flags & MS_RDONLY)) {
1206 		status = ocfs2_enable_quotas(osb);
1207 		if (status < 0) {
1208 			/* We have to err-out specially here because
1209 			 * s_root is already set */
1210 			mlog_errno(status);
1211 			atomic_set(&osb->vol_state, VOLUME_DISABLED);
1212 			wake_up(&osb->osb_mount_event);
1213 			return status;
1214 		}
1215 	}
1216 
1217 	ocfs2_complete_quota_recovery(osb);
1218 
1219 	/* Now we wake up again for processes waiting for quotas */
1220 	atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1221 	wake_up(&osb->osb_mount_event);
1222 
1223 	/* Start this when the mount is almost sure of being successful */
1224 	ocfs2_orphan_scan_start(osb);
1225 
1226 	return status;
1227 
1228 read_super_error:
1229 	brelse(bh);
1230 
1231 	if (osb) {
1232 		atomic_set(&osb->vol_state, VOLUME_DISABLED);
1233 		wake_up(&osb->osb_mount_event);
1234 		ocfs2_dismount_volume(sb, 1);
1235 	}
1236 
1237 	if (status)
1238 		mlog_errno(status);
1239 	return status;
1240 }
1241 
ocfs2_mount(struct file_system_type * fs_type,int flags,const char * dev_name,void * data)1242 static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
1243 			int flags,
1244 			const char *dev_name,
1245 			void *data)
1246 {
1247 	return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
1248 }
1249 
1250 static struct file_system_type ocfs2_fs_type = {
1251 	.owner          = THIS_MODULE,
1252 	.name           = "ocfs2",
1253 	.mount          = ocfs2_mount,
1254 	.kill_sb        = kill_block_super,
1255 	.fs_flags       = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1256 	.next           = NULL
1257 };
1258 MODULE_ALIAS_FS("ocfs2");
1259 
ocfs2_check_set_options(struct super_block * sb,struct mount_options * options)1260 static int ocfs2_check_set_options(struct super_block *sb,
1261 				   struct mount_options *options)
1262 {
1263 	if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1264 	    !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1265 					 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1266 		mlog(ML_ERROR, "User quotas were requested, but this "
1267 		     "filesystem does not have the feature enabled.\n");
1268 		return 0;
1269 	}
1270 	if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1271 	    !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1272 					 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1273 		mlog(ML_ERROR, "Group quotas were requested, but this "
1274 		     "filesystem does not have the feature enabled.\n");
1275 		return 0;
1276 	}
1277 	if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1278 	    !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1279 		mlog(ML_ERROR, "ACL support requested but extended attributes "
1280 		     "feature is not enabled\n");
1281 		return 0;
1282 	}
1283 	/* No ACL setting specified? Use XATTR feature... */
1284 	if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1285 				    OCFS2_MOUNT_NO_POSIX_ACL))) {
1286 		if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1287 			options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1288 		else
1289 			options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1290 	}
1291 	return 1;
1292 }
1293 
ocfs2_parse_options(struct super_block * sb,char * options,struct mount_options * mopt,int is_remount)1294 static int ocfs2_parse_options(struct super_block *sb,
1295 			       char *options,
1296 			       struct mount_options *mopt,
1297 			       int is_remount)
1298 {
1299 	int status, user_stack = 0;
1300 	char *p;
1301 	u32 tmp;
1302 
1303 	trace_ocfs2_parse_options(is_remount, options ? options : "(none)");
1304 
1305 	mopt->commit_interval = 0;
1306 	mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1307 	mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1308 	mopt->slot = OCFS2_INVALID_SLOT;
1309 	mopt->localalloc_opt = -1;
1310 	mopt->cluster_stack[0] = '\0';
1311 	mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1312 	mopt->dir_resv_level = -1;
1313 
1314 	if (!options) {
1315 		status = 1;
1316 		goto bail;
1317 	}
1318 
1319 	while ((p = strsep(&options, ",")) != NULL) {
1320 		int token, option;
1321 		substring_t args[MAX_OPT_ARGS];
1322 
1323 		if (!*p)
1324 			continue;
1325 
1326 		token = match_token(p, tokens, args);
1327 		switch (token) {
1328 		case Opt_hb_local:
1329 			mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1330 			break;
1331 		case Opt_hb_none:
1332 			mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1333 			break;
1334 		case Opt_hb_global:
1335 			mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1336 			break;
1337 		case Opt_barrier:
1338 			if (match_int(&args[0], &option)) {
1339 				status = 0;
1340 				goto bail;
1341 			}
1342 			if (option)
1343 				mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1344 			else
1345 				mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1346 			break;
1347 		case Opt_intr:
1348 			mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1349 			break;
1350 		case Opt_nointr:
1351 			mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1352 			break;
1353 		case Opt_err_panic:
1354 			mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1355 			break;
1356 		case Opt_err_ro:
1357 			mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1358 			break;
1359 		case Opt_data_ordered:
1360 			mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1361 			break;
1362 		case Opt_data_writeback:
1363 			mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1364 			break;
1365 		case Opt_user_xattr:
1366 			mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1367 			break;
1368 		case Opt_nouser_xattr:
1369 			mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1370 			break;
1371 		case Opt_atime_quantum:
1372 			if (match_int(&args[0], &option)) {
1373 				status = 0;
1374 				goto bail;
1375 			}
1376 			if (option >= 0)
1377 				mopt->atime_quantum = option;
1378 			break;
1379 		case Opt_slot:
1380 			option = 0;
1381 			if (match_int(&args[0], &option)) {
1382 				status = 0;
1383 				goto bail;
1384 			}
1385 			if (option)
1386 				mopt->slot = (s16)option;
1387 			break;
1388 		case Opt_commit:
1389 			option = 0;
1390 			if (match_int(&args[0], &option)) {
1391 				status = 0;
1392 				goto bail;
1393 			}
1394 			if (option < 0)
1395 				return 0;
1396 			if (option == 0)
1397 				option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1398 			mopt->commit_interval = HZ * option;
1399 			break;
1400 		case Opt_localalloc:
1401 			option = 0;
1402 			if (match_int(&args[0], &option)) {
1403 				status = 0;
1404 				goto bail;
1405 			}
1406 			if (option >= 0)
1407 				mopt->localalloc_opt = option;
1408 			break;
1409 		case Opt_localflocks:
1410 			/*
1411 			 * Changing this during remount could race
1412 			 * flock() requests, or "unbalance" existing
1413 			 * ones (e.g., a lock is taken in one mode but
1414 			 * dropped in the other). If users care enough
1415 			 * to flip locking modes during remount, we
1416 			 * could add a "local" flag to individual
1417 			 * flock structures for proper tracking of
1418 			 * state.
1419 			 */
1420 			if (!is_remount)
1421 				mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1422 			break;
1423 		case Opt_stack:
1424 			/* Check both that the option we were passed
1425 			 * is of the right length and that it is a proper
1426 			 * string of the right length.
1427 			 */
1428 			if (((args[0].to - args[0].from) !=
1429 			     OCFS2_STACK_LABEL_LEN) ||
1430 			    (strnlen(args[0].from,
1431 				     OCFS2_STACK_LABEL_LEN) !=
1432 			     OCFS2_STACK_LABEL_LEN)) {
1433 				mlog(ML_ERROR,
1434 				     "Invalid cluster_stack option\n");
1435 				status = 0;
1436 				goto bail;
1437 			}
1438 			memcpy(mopt->cluster_stack, args[0].from,
1439 			       OCFS2_STACK_LABEL_LEN);
1440 			mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1441 			/*
1442 			 * Open code the memcmp here as we don't have
1443 			 * an osb to pass to
1444 			 * ocfs2_userspace_stack().
1445 			 */
1446 			if (memcmp(mopt->cluster_stack,
1447 				   OCFS2_CLASSIC_CLUSTER_STACK,
1448 				   OCFS2_STACK_LABEL_LEN))
1449 				user_stack = 1;
1450 			break;
1451 		case Opt_inode64:
1452 			mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1453 			break;
1454 		case Opt_usrquota:
1455 			mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1456 			break;
1457 		case Opt_grpquota:
1458 			mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1459 			break;
1460 		case Opt_coherency_buffered:
1461 			mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1462 			break;
1463 		case Opt_coherency_full:
1464 			mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1465 			break;
1466 		case Opt_acl:
1467 			mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1468 			mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1469 			break;
1470 		case Opt_noacl:
1471 			mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1472 			mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1473 			break;
1474 		case Opt_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->resv_level = option;
1484 			break;
1485 		case Opt_dir_resv_level:
1486 			if (is_remount)
1487 				break;
1488 			if (match_int(&args[0], &option)) {
1489 				status = 0;
1490 				goto bail;
1491 			}
1492 			if (option >= OCFS2_MIN_RESV_LEVEL &&
1493 			    option < OCFS2_MAX_RESV_LEVEL)
1494 				mopt->dir_resv_level = option;
1495 			break;
1496 		default:
1497 			mlog(ML_ERROR,
1498 			     "Unrecognized mount option \"%s\" "
1499 			     "or missing value\n", p);
1500 			status = 0;
1501 			goto bail;
1502 		}
1503 	}
1504 
1505 	if (user_stack == 0) {
1506 		/* Ensure only one heartbeat mode */
1507 		tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1508 					 OCFS2_MOUNT_HB_GLOBAL |
1509 					 OCFS2_MOUNT_HB_NONE);
1510 		if (hweight32(tmp) != 1) {
1511 			mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1512 			status = 0;
1513 			goto bail;
1514 		}
1515 	}
1516 
1517 	status = 1;
1518 
1519 bail:
1520 	return status;
1521 }
1522 
ocfs2_show_options(struct seq_file * s,struct dentry * root)1523 static int ocfs2_show_options(struct seq_file *s, struct dentry *root)
1524 {
1525 	struct ocfs2_super *osb = OCFS2_SB(root->d_sb);
1526 	unsigned long opts = osb->s_mount_opt;
1527 	unsigned int local_alloc_megs;
1528 
1529 	if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1530 		seq_printf(s, ",_netdev");
1531 		if (opts & OCFS2_MOUNT_HB_LOCAL)
1532 			seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1533 		else
1534 			seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1535 	} else
1536 		seq_printf(s, ",%s", OCFS2_HB_NONE);
1537 
1538 	if (opts & OCFS2_MOUNT_NOINTR)
1539 		seq_printf(s, ",nointr");
1540 
1541 	if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1542 		seq_printf(s, ",data=writeback");
1543 	else
1544 		seq_printf(s, ",data=ordered");
1545 
1546 	if (opts & OCFS2_MOUNT_BARRIER)
1547 		seq_printf(s, ",barrier=1");
1548 
1549 	if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1550 		seq_printf(s, ",errors=panic");
1551 	else
1552 		seq_printf(s, ",errors=remount-ro");
1553 
1554 	if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1555 		seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1556 
1557 	seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1558 
1559 	if (osb->osb_commit_interval)
1560 		seq_printf(s, ",commit=%u",
1561 			   (unsigned) (osb->osb_commit_interval / HZ));
1562 
1563 	local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1564 	if (local_alloc_megs != ocfs2_la_default_mb(osb))
1565 		seq_printf(s, ",localalloc=%d", local_alloc_megs);
1566 
1567 	if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1568 		seq_printf(s, ",localflocks,");
1569 
1570 	if (osb->osb_cluster_stack[0])
1571 		seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1572 			   osb->osb_cluster_stack);
1573 	if (opts & OCFS2_MOUNT_USRQUOTA)
1574 		seq_printf(s, ",usrquota");
1575 	if (opts & OCFS2_MOUNT_GRPQUOTA)
1576 		seq_printf(s, ",grpquota");
1577 
1578 	if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1579 		seq_printf(s, ",coherency=buffered");
1580 	else
1581 		seq_printf(s, ",coherency=full");
1582 
1583 	if (opts & OCFS2_MOUNT_NOUSERXATTR)
1584 		seq_printf(s, ",nouser_xattr");
1585 	else
1586 		seq_printf(s, ",user_xattr");
1587 
1588 	if (opts & OCFS2_MOUNT_INODE64)
1589 		seq_printf(s, ",inode64");
1590 
1591 	if (opts & OCFS2_MOUNT_POSIX_ACL)
1592 		seq_printf(s, ",acl");
1593 	else
1594 		seq_printf(s, ",noacl");
1595 
1596 	if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1597 		seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1598 
1599 	if (osb->osb_dir_resv_level != osb->osb_resv_level)
1600 		seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1601 
1602 	return 0;
1603 }
1604 
ocfs2_init(void)1605 static int __init ocfs2_init(void)
1606 {
1607 	int status;
1608 
1609 	status = init_ocfs2_uptodate_cache();
1610 	if (status < 0)
1611 		goto out1;
1612 
1613 	status = ocfs2_initialize_mem_caches();
1614 	if (status < 0)
1615 		goto out2;
1616 
1617 	ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1618 	if (!ocfs2_wq) {
1619 		status = -ENOMEM;
1620 		goto out3;
1621 	}
1622 
1623 	ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1624 	if (!ocfs2_debugfs_root) {
1625 		status = -EFAULT;
1626 		mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1627 	}
1628 
1629 	ocfs2_set_locking_protocol();
1630 
1631 	status = register_quota_format(&ocfs2_quota_format);
1632 	if (status < 0)
1633 		goto out4;
1634 	status = register_filesystem(&ocfs2_fs_type);
1635 	if (!status)
1636 		return 0;
1637 
1638 	unregister_quota_format(&ocfs2_quota_format);
1639 out4:
1640 	destroy_workqueue(ocfs2_wq);
1641 	debugfs_remove(ocfs2_debugfs_root);
1642 out3:
1643 	ocfs2_free_mem_caches();
1644 out2:
1645 	exit_ocfs2_uptodate_cache();
1646 out1:
1647 	mlog_errno(status);
1648 	return status;
1649 }
1650 
ocfs2_exit(void)1651 static void __exit ocfs2_exit(void)
1652 {
1653 	if (ocfs2_wq) {
1654 		flush_workqueue(ocfs2_wq);
1655 		destroy_workqueue(ocfs2_wq);
1656 	}
1657 
1658 	unregister_quota_format(&ocfs2_quota_format);
1659 
1660 	debugfs_remove(ocfs2_debugfs_root);
1661 
1662 	ocfs2_free_mem_caches();
1663 
1664 	unregister_filesystem(&ocfs2_fs_type);
1665 
1666 	exit_ocfs2_uptodate_cache();
1667 }
1668 
ocfs2_put_super(struct super_block * sb)1669 static void ocfs2_put_super(struct super_block *sb)
1670 {
1671 	trace_ocfs2_put_super(sb);
1672 
1673 	ocfs2_sync_blockdev(sb);
1674 	ocfs2_dismount_volume(sb, 0);
1675 }
1676 
ocfs2_statfs(struct dentry * dentry,struct kstatfs * buf)1677 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1678 {
1679 	struct ocfs2_super *osb;
1680 	u32 numbits, freebits;
1681 	int status;
1682 	struct ocfs2_dinode *bm_lock;
1683 	struct buffer_head *bh = NULL;
1684 	struct inode *inode = NULL;
1685 
1686 	trace_ocfs2_statfs(dentry->d_sb, buf);
1687 
1688 	osb = OCFS2_SB(dentry->d_sb);
1689 
1690 	inode = ocfs2_get_system_file_inode(osb,
1691 					    GLOBAL_BITMAP_SYSTEM_INODE,
1692 					    OCFS2_INVALID_SLOT);
1693 	if (!inode) {
1694 		mlog(ML_ERROR, "failed to get bitmap inode\n");
1695 		status = -EIO;
1696 		goto bail;
1697 	}
1698 
1699 	status = ocfs2_inode_lock(inode, &bh, 0);
1700 	if (status < 0) {
1701 		mlog_errno(status);
1702 		goto bail;
1703 	}
1704 
1705 	bm_lock = (struct ocfs2_dinode *) bh->b_data;
1706 
1707 	numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1708 	freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1709 
1710 	buf->f_type = OCFS2_SUPER_MAGIC;
1711 	buf->f_bsize = dentry->d_sb->s_blocksize;
1712 	buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1713 	buf->f_blocks = ((sector_t) numbits) *
1714 			(osb->s_clustersize >> osb->sb->s_blocksize_bits);
1715 	buf->f_bfree = ((sector_t) freebits) *
1716 		       (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1717 	buf->f_bavail = buf->f_bfree;
1718 	buf->f_files = numbits;
1719 	buf->f_ffree = freebits;
1720 	buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1721 				& 0xFFFFFFFFUL;
1722 	buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1723 				OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1724 
1725 	brelse(bh);
1726 
1727 	ocfs2_inode_unlock(inode, 0);
1728 	status = 0;
1729 bail:
1730 	if (inode)
1731 		iput(inode);
1732 
1733 	if (status)
1734 		mlog_errno(status);
1735 
1736 	return status;
1737 }
1738 
ocfs2_inode_init_once(void * data)1739 static void ocfs2_inode_init_once(void *data)
1740 {
1741 	struct ocfs2_inode_info *oi = data;
1742 
1743 	oi->ip_flags = 0;
1744 	oi->ip_open_count = 0;
1745 	spin_lock_init(&oi->ip_lock);
1746 	ocfs2_extent_map_init(&oi->vfs_inode);
1747 	INIT_LIST_HEAD(&oi->ip_io_markers);
1748 	oi->ip_dir_start_lookup = 0;
1749 	mutex_init(&oi->ip_unaligned_aio);
1750 	init_rwsem(&oi->ip_alloc_sem);
1751 	init_rwsem(&oi->ip_xattr_sem);
1752 	mutex_init(&oi->ip_io_mutex);
1753 
1754 	oi->ip_blkno = 0ULL;
1755 	oi->ip_clusters = 0;
1756 
1757 	ocfs2_resv_init_once(&oi->ip_la_data_resv);
1758 
1759 	ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1760 	ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1761 	ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1762 
1763 	ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1764 				  &ocfs2_inode_caching_ops);
1765 
1766 	inode_init_once(&oi->vfs_inode);
1767 }
1768 
ocfs2_initialize_mem_caches(void)1769 static int ocfs2_initialize_mem_caches(void)
1770 {
1771 	ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1772 				       sizeof(struct ocfs2_inode_info),
1773 				       0,
1774 				       (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1775 						SLAB_MEM_SPREAD),
1776 				       ocfs2_inode_init_once);
1777 	ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1778 					sizeof(struct ocfs2_dquot),
1779 					0,
1780 					(SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1781 						SLAB_MEM_SPREAD),
1782 					NULL);
1783 	ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1784 					sizeof(struct ocfs2_quota_chunk),
1785 					0,
1786 					(SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1787 					NULL);
1788 	if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1789 	    !ocfs2_qf_chunk_cachep) {
1790 		if (ocfs2_inode_cachep)
1791 			kmem_cache_destroy(ocfs2_inode_cachep);
1792 		if (ocfs2_dquot_cachep)
1793 			kmem_cache_destroy(ocfs2_dquot_cachep);
1794 		if (ocfs2_qf_chunk_cachep)
1795 			kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1796 		return -ENOMEM;
1797 	}
1798 
1799 	return 0;
1800 }
1801 
ocfs2_free_mem_caches(void)1802 static void ocfs2_free_mem_caches(void)
1803 {
1804 	/*
1805 	 * Make sure all delayed rcu free inodes are flushed before we
1806 	 * destroy cache.
1807 	 */
1808 	rcu_barrier();
1809 	if (ocfs2_inode_cachep)
1810 		kmem_cache_destroy(ocfs2_inode_cachep);
1811 	ocfs2_inode_cachep = NULL;
1812 
1813 	if (ocfs2_dquot_cachep)
1814 		kmem_cache_destroy(ocfs2_dquot_cachep);
1815 	ocfs2_dquot_cachep = NULL;
1816 
1817 	if (ocfs2_qf_chunk_cachep)
1818 		kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1819 	ocfs2_qf_chunk_cachep = NULL;
1820 }
1821 
ocfs2_get_sector(struct super_block * sb,struct buffer_head ** bh,int block,int sect_size)1822 static int ocfs2_get_sector(struct super_block *sb,
1823 			    struct buffer_head **bh,
1824 			    int block,
1825 			    int sect_size)
1826 {
1827 	if (!sb_set_blocksize(sb, sect_size)) {
1828 		mlog(ML_ERROR, "unable to set blocksize\n");
1829 		return -EIO;
1830 	}
1831 
1832 	*bh = sb_getblk(sb, block);
1833 	if (!*bh) {
1834 		mlog_errno(-ENOMEM);
1835 		return -ENOMEM;
1836 	}
1837 	lock_buffer(*bh);
1838 	if (!buffer_dirty(*bh))
1839 		clear_buffer_uptodate(*bh);
1840 	unlock_buffer(*bh);
1841 	ll_rw_block(READ, 1, bh);
1842 	wait_on_buffer(*bh);
1843 	if (!buffer_uptodate(*bh)) {
1844 		mlog_errno(-EIO);
1845 		brelse(*bh);
1846 		*bh = NULL;
1847 		return -EIO;
1848 	}
1849 
1850 	return 0;
1851 }
1852 
ocfs2_mount_volume(struct super_block * sb)1853 static int ocfs2_mount_volume(struct super_block *sb)
1854 {
1855 	int status = 0;
1856 	int unlock_super = 0;
1857 	struct ocfs2_super *osb = OCFS2_SB(sb);
1858 
1859 	if (ocfs2_is_hard_readonly(osb))
1860 		goto leave;
1861 
1862 	status = ocfs2_dlm_init(osb);
1863 	if (status < 0) {
1864 		mlog_errno(status);
1865 		goto leave;
1866 	}
1867 
1868 	status = ocfs2_super_lock(osb, 1);
1869 	if (status < 0) {
1870 		mlog_errno(status);
1871 		goto leave;
1872 	}
1873 	unlock_super = 1;
1874 
1875 	/* This will load up the node map and add ourselves to it. */
1876 	status = ocfs2_find_slot(osb);
1877 	if (status < 0) {
1878 		mlog_errno(status);
1879 		goto leave;
1880 	}
1881 
1882 	/* load all node-local system inodes */
1883 	status = ocfs2_init_local_system_inodes(osb);
1884 	if (status < 0) {
1885 		mlog_errno(status);
1886 		goto leave;
1887 	}
1888 
1889 	status = ocfs2_check_volume(osb);
1890 	if (status < 0) {
1891 		mlog_errno(status);
1892 		goto leave;
1893 	}
1894 
1895 	status = ocfs2_truncate_log_init(osb);
1896 	if (status < 0)
1897 		mlog_errno(status);
1898 
1899 leave:
1900 	if (unlock_super)
1901 		ocfs2_super_unlock(osb, 1);
1902 
1903 	return status;
1904 }
1905 
ocfs2_dismount_volume(struct super_block * sb,int mnt_err)1906 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1907 {
1908 	int tmp, hangup_needed = 0;
1909 	struct ocfs2_super *osb = NULL;
1910 	char nodestr[12];
1911 
1912 	trace_ocfs2_dismount_volume(sb);
1913 
1914 	BUG_ON(!sb);
1915 	osb = OCFS2_SB(sb);
1916 	BUG_ON(!osb);
1917 
1918 	debugfs_remove(osb->osb_ctxt);
1919 
1920 	/* Orphan scan should be stopped as early as possible */
1921 	ocfs2_orphan_scan_stop(osb);
1922 
1923 	ocfs2_disable_quotas(osb);
1924 
1925 	/* All dquots should be freed by now */
1926 	WARN_ON(!llist_empty(&osb->dquot_drop_list));
1927 	/* Wait for worker to be done with the work structure in osb */
1928 	cancel_work_sync(&osb->dquot_drop_work);
1929 
1930 	ocfs2_shutdown_local_alloc(osb);
1931 
1932 	ocfs2_truncate_log_shutdown(osb);
1933 
1934 	/* This will disable recovery and flush any recovery work. */
1935 	ocfs2_recovery_exit(osb);
1936 
1937 	ocfs2_journal_shutdown(osb);
1938 
1939 	ocfs2_sync_blockdev(sb);
1940 
1941 	ocfs2_purge_refcount_trees(osb);
1942 
1943 	/* No cluster connection means we've failed during mount, so skip
1944 	 * all the steps which depended on that to complete. */
1945 	if (osb->cconn) {
1946 		tmp = ocfs2_super_lock(osb, 1);
1947 		if (tmp < 0) {
1948 			mlog_errno(tmp);
1949 			return;
1950 		}
1951 	}
1952 
1953 	if (osb->slot_num != OCFS2_INVALID_SLOT)
1954 		ocfs2_put_slot(osb);
1955 
1956 	if (osb->cconn)
1957 		ocfs2_super_unlock(osb, 1);
1958 
1959 	ocfs2_release_system_inodes(osb);
1960 
1961 	/*
1962 	 * If we're dismounting due to mount error, mount.ocfs2 will clean
1963 	 * up heartbeat.  If we're a local mount, there is no heartbeat.
1964 	 * If we failed before we got a uuid_str yet, we can't stop
1965 	 * heartbeat.  Otherwise, do it.
1966 	 */
1967 	if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str &&
1968 	    !ocfs2_is_hard_readonly(osb))
1969 		hangup_needed = 1;
1970 
1971 	if (osb->cconn)
1972 		ocfs2_dlm_shutdown(osb, hangup_needed);
1973 
1974 	ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1975 	debugfs_remove(osb->osb_debug_root);
1976 
1977 	if (hangup_needed)
1978 		ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1979 
1980 	atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1981 
1982 	if (ocfs2_mount_local(osb))
1983 		snprintf(nodestr, sizeof(nodestr), "local");
1984 	else
1985 		snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1986 
1987 	printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1988 	       osb->dev_str, nodestr);
1989 
1990 	ocfs2_delete_osb(osb);
1991 	kfree(osb);
1992 	sb->s_dev = 0;
1993 	sb->s_fs_info = NULL;
1994 }
1995 
ocfs2_setup_osb_uuid(struct ocfs2_super * osb,const unsigned char * uuid,unsigned uuid_bytes)1996 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1997 				unsigned uuid_bytes)
1998 {
1999 	int i, ret;
2000 	char *ptr;
2001 
2002 	BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
2003 
2004 	osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
2005 	if (osb->uuid_str == NULL)
2006 		return -ENOMEM;
2007 
2008 	for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
2009 		/* print with null */
2010 		ret = snprintf(ptr, 3, "%02X", uuid[i]);
2011 		if (ret != 2) /* drop super cleans up */
2012 			return -EINVAL;
2013 		/* then only advance past the last char */
2014 		ptr += 2;
2015 	}
2016 
2017 	return 0;
2018 }
2019 
2020 /* Make sure entire volume is addressable by our journal.  Requires
2021    osb_clusters_at_boot to be valid and for the journal to have been
2022    initialized by ocfs2_journal_init(). */
ocfs2_journal_addressable(struct ocfs2_super * osb)2023 static int ocfs2_journal_addressable(struct ocfs2_super *osb)
2024 {
2025 	int status = 0;
2026 	u64 max_block =
2027 		ocfs2_clusters_to_blocks(osb->sb,
2028 					 osb->osb_clusters_at_boot) - 1;
2029 
2030 	/* 32-bit block number is always OK. */
2031 	if (max_block <= (u32)~0ULL)
2032 		goto out;
2033 
2034 	/* Volume is "huge", so see if our journal is new enough to
2035 	   support it. */
2036 	if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2037 				       OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2038 	      jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2039 					       JBD2_FEATURE_INCOMPAT_64BIT))) {
2040 		mlog(ML_ERROR, "The journal cannot address the entire volume. "
2041 		     "Enable the 'block64' journal option with tunefs.ocfs2");
2042 		status = -EFBIG;
2043 		goto out;
2044 	}
2045 
2046  out:
2047 	return status;
2048 }
2049 
ocfs2_initialize_super(struct super_block * sb,struct buffer_head * bh,int sector_size,struct ocfs2_blockcheck_stats * stats)2050 static int ocfs2_initialize_super(struct super_block *sb,
2051 				  struct buffer_head *bh,
2052 				  int sector_size,
2053 				  struct ocfs2_blockcheck_stats *stats)
2054 {
2055 	int status;
2056 	int i, cbits, bbits;
2057 	struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2058 	struct inode *inode = NULL;
2059 	struct ocfs2_journal *journal;
2060 	struct ocfs2_super *osb;
2061 	u64 total_blocks;
2062 
2063 	osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2064 	if (!osb) {
2065 		status = -ENOMEM;
2066 		mlog_errno(status);
2067 		goto bail;
2068 	}
2069 
2070 	sb->s_fs_info = osb;
2071 	sb->s_op = &ocfs2_sops;
2072 	sb->s_d_op = &ocfs2_dentry_ops;
2073 	sb->s_export_op = &ocfs2_export_ops;
2074 	sb->s_qcop = &ocfs2_quotactl_ops;
2075 	sb->dq_op = &ocfs2_quota_operations;
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 
2084 	osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2085 
2086 	for (i = 0; i < 3; i++)
2087 		osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2088 	osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2089 
2090 	osb->sb = sb;
2091 	/* Save off for ocfs2_rw_direct */
2092 	osb->s_sectsize_bits = blksize_bits(sector_size);
2093 	BUG_ON(!osb->s_sectsize_bits);
2094 
2095 	spin_lock_init(&osb->dc_task_lock);
2096 	init_waitqueue_head(&osb->dc_event);
2097 	osb->dc_work_sequence = 0;
2098 	osb->dc_wake_sequence = 0;
2099 	INIT_LIST_HEAD(&osb->blocked_lock_list);
2100 	osb->blocked_lock_count = 0;
2101 	spin_lock_init(&osb->osb_lock);
2102 	spin_lock_init(&osb->osb_xattr_lock);
2103 	ocfs2_init_steal_slots(osb);
2104 
2105 	mutex_init(&osb->system_file_mutex);
2106 
2107 	atomic_set(&osb->alloc_stats.moves, 0);
2108 	atomic_set(&osb->alloc_stats.local_data, 0);
2109 	atomic_set(&osb->alloc_stats.bitmap_data, 0);
2110 	atomic_set(&osb->alloc_stats.bg_allocs, 0);
2111 	atomic_set(&osb->alloc_stats.bg_extends, 0);
2112 
2113 	/* Copy the blockcheck stats from the superblock probe */
2114 	osb->osb_ecc_stats = *stats;
2115 
2116 	ocfs2_init_node_maps(osb);
2117 
2118 	snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2119 		 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2120 
2121 	osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2122 	if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2123 		mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2124 		     osb->max_slots);
2125 		status = -EINVAL;
2126 		goto bail;
2127 	}
2128 
2129 	ocfs2_orphan_scan_init(osb);
2130 
2131 	status = ocfs2_recovery_init(osb);
2132 	if (status) {
2133 		mlog(ML_ERROR, "Unable to initialize recovery state\n");
2134 		mlog_errno(status);
2135 		goto bail;
2136 	}
2137 
2138 	init_waitqueue_head(&osb->checkpoint_event);
2139 
2140 	osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2141 
2142 	osb->slot_num = OCFS2_INVALID_SLOT;
2143 
2144 	osb->s_xattr_inline_size = le16_to_cpu(
2145 					di->id2.i_super.s_xattr_inline_size);
2146 
2147 	osb->local_alloc_state = OCFS2_LA_UNUSED;
2148 	osb->local_alloc_bh = NULL;
2149 	INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2150 
2151 	init_waitqueue_head(&osb->osb_mount_event);
2152 
2153 	status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2154 	if (status) {
2155 		mlog_errno(status);
2156 		goto bail;
2157 	}
2158 
2159 	osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2160 	if (!osb->vol_label) {
2161 		mlog(ML_ERROR, "unable to alloc vol label\n");
2162 		status = -ENOMEM;
2163 		goto bail;
2164 	}
2165 
2166 	osb->slot_recovery_generations =
2167 		kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2168 			GFP_KERNEL);
2169 	if (!osb->slot_recovery_generations) {
2170 		status = -ENOMEM;
2171 		mlog_errno(status);
2172 		goto bail;
2173 	}
2174 
2175 	init_waitqueue_head(&osb->osb_wipe_event);
2176 	osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2177 					sizeof(*osb->osb_orphan_wipes),
2178 					GFP_KERNEL);
2179 	if (!osb->osb_orphan_wipes) {
2180 		status = -ENOMEM;
2181 		mlog_errno(status);
2182 		goto bail;
2183 	}
2184 
2185 	osb->osb_rf_lock_tree = RB_ROOT;
2186 
2187 	osb->s_feature_compat =
2188 		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2189 	osb->s_feature_ro_compat =
2190 		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2191 	osb->s_feature_incompat =
2192 		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2193 
2194 	if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2195 		mlog(ML_ERROR, "couldn't mount because of unsupported "
2196 		     "optional features (%x).\n", i);
2197 		status = -EINVAL;
2198 		goto bail;
2199 	}
2200 	if (!(osb->sb->s_flags & MS_RDONLY) &&
2201 	    (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2202 		mlog(ML_ERROR, "couldn't mount RDWR because of "
2203 		     "unsupported optional features (%x).\n", i);
2204 		status = -EINVAL;
2205 		goto bail;
2206 	}
2207 
2208 	if (ocfs2_clusterinfo_valid(osb)) {
2209 		osb->osb_stackflags =
2210 			OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2211 		strlcpy(osb->osb_cluster_stack,
2212 		       OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2213 		       OCFS2_STACK_LABEL_LEN + 1);
2214 		if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2215 			mlog(ML_ERROR,
2216 			     "couldn't mount because of an invalid "
2217 			     "cluster stack label (%s) \n",
2218 			     osb->osb_cluster_stack);
2219 			status = -EINVAL;
2220 			goto bail;
2221 		}
2222 		strlcpy(osb->osb_cluster_name,
2223 			OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster,
2224 			OCFS2_CLUSTER_NAME_LEN + 1);
2225 	} else {
2226 		/* The empty string is identical with classic tools that
2227 		 * don't know about s_cluster_info. */
2228 		osb->osb_cluster_stack[0] = '\0';
2229 	}
2230 
2231 	get_random_bytes(&osb->s_next_generation, sizeof(u32));
2232 
2233 	/* FIXME
2234 	 * This should be done in ocfs2_journal_init(), but unknown
2235 	 * ordering issues will cause the filesystem to crash.
2236 	 * If anyone wants to figure out what part of the code
2237 	 * refers to osb->journal before ocfs2_journal_init() is run,
2238 	 * be my guest.
2239 	 */
2240 	/* initialize our journal structure */
2241 
2242 	journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2243 	if (!journal) {
2244 		mlog(ML_ERROR, "unable to alloc journal\n");
2245 		status = -ENOMEM;
2246 		goto bail;
2247 	}
2248 	osb->journal = journal;
2249 	journal->j_osb = osb;
2250 
2251 	atomic_set(&journal->j_num_trans, 0);
2252 	init_rwsem(&journal->j_trans_barrier);
2253 	init_waitqueue_head(&journal->j_checkpointed);
2254 	spin_lock_init(&journal->j_lock);
2255 	journal->j_trans_id = (unsigned long) 1;
2256 	INIT_LIST_HEAD(&journal->j_la_cleanups);
2257 	INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2258 	journal->j_state = OCFS2_JOURNAL_FREE;
2259 
2260 	INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs);
2261 	init_llist_head(&osb->dquot_drop_list);
2262 
2263 	/* get some pseudo constants for clustersize bits */
2264 	osb->s_clustersize_bits =
2265 		le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2266 	osb->s_clustersize = 1 << osb->s_clustersize_bits;
2267 
2268 	if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2269 	    osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2270 		mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2271 		     osb->s_clustersize);
2272 		status = -EINVAL;
2273 		goto bail;
2274 	}
2275 
2276 	total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2277 						le32_to_cpu(di->i_clusters));
2278 
2279 	status = generic_check_addressable(osb->sb->s_blocksize_bits,
2280 					   total_blocks);
2281 	if (status) {
2282 		mlog(ML_ERROR, "Volume too large "
2283 		     "to mount safely on this system");
2284 		status = -EFBIG;
2285 		goto bail;
2286 	}
2287 
2288 	if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2289 				 sizeof(di->id2.i_super.s_uuid))) {
2290 		mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2291 		status = -ENOMEM;
2292 		goto bail;
2293 	}
2294 
2295 	strlcpy(osb->vol_label, di->id2.i_super.s_label,
2296 		OCFS2_MAX_VOL_LABEL_LEN);
2297 	osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2298 	osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2299 	osb->first_cluster_group_blkno =
2300 		le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2301 	osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2302 	osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2303 	trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str,
2304 				     (unsigned long long)osb->root_blkno,
2305 				     (unsigned long long)osb->system_dir_blkno,
2306 				     osb->s_clustersize_bits);
2307 
2308 	osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2309 	if (!osb->osb_dlm_debug) {
2310 		status = -ENOMEM;
2311 		mlog_errno(status);
2312 		goto bail;
2313 	}
2314 
2315 	atomic_set(&osb->vol_state, VOLUME_INIT);
2316 
2317 	/* load root, system_dir, and all global system inodes */
2318 	status = ocfs2_init_global_system_inodes(osb);
2319 	if (status < 0) {
2320 		mlog_errno(status);
2321 		goto bail;
2322 	}
2323 
2324 	/*
2325 	 * global bitmap
2326 	 */
2327 	inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2328 					    OCFS2_INVALID_SLOT);
2329 	if (!inode) {
2330 		status = -EINVAL;
2331 		mlog_errno(status);
2332 		goto bail;
2333 	}
2334 
2335 	osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2336 	osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2337 	iput(inode);
2338 
2339 	osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2340 				 osb->s_feature_incompat) * 8;
2341 
2342 	status = ocfs2_init_slot_info(osb);
2343 	if (status < 0) {
2344 		mlog_errno(status);
2345 		goto bail;
2346 	}
2347 	cleancache_init_shared_fs((char *)&di->id2.i_super.s_uuid, sb);
2348 
2349 bail:
2350 	return status;
2351 }
2352 
2353 /*
2354  * will return: -EAGAIN if it is ok to keep searching for superblocks
2355  *              -EINVAL if there is a bad superblock
2356  *              0 on success
2357  */
ocfs2_verify_volume(struct ocfs2_dinode * di,struct buffer_head * bh,u32 blksz,struct ocfs2_blockcheck_stats * stats)2358 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2359 			       struct buffer_head *bh,
2360 			       u32 blksz,
2361 			       struct ocfs2_blockcheck_stats *stats)
2362 {
2363 	int status = -EAGAIN;
2364 
2365 	if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2366 		   strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2367 		/* We have to do a raw check of the feature here */
2368 		if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2369 		    OCFS2_FEATURE_INCOMPAT_META_ECC) {
2370 			status = ocfs2_block_check_validate(bh->b_data,
2371 							    bh->b_size,
2372 							    &di->i_check,
2373 							    stats);
2374 			if (status)
2375 				goto out;
2376 		}
2377 		status = -EINVAL;
2378 		if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2379 			mlog(ML_ERROR, "found superblock with incorrect block "
2380 			     "size: found %u, should be %u\n",
2381 			     1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2382 			       blksz);
2383 		} else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2384 			   OCFS2_MAJOR_REV_LEVEL ||
2385 			   le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2386 			   OCFS2_MINOR_REV_LEVEL) {
2387 			mlog(ML_ERROR, "found superblock with bad version: "
2388 			     "found %u.%u, should be %u.%u\n",
2389 			     le16_to_cpu(di->id2.i_super.s_major_rev_level),
2390 			     le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2391 			     OCFS2_MAJOR_REV_LEVEL,
2392 			     OCFS2_MINOR_REV_LEVEL);
2393 		} else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2394 			mlog(ML_ERROR, "bad block number on superblock: "
2395 			     "found %llu, should be %llu\n",
2396 			     (unsigned long long)le64_to_cpu(di->i_blkno),
2397 			     (unsigned long long)bh->b_blocknr);
2398 		} else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2399 			    le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2400 			mlog(ML_ERROR, "bad cluster size found: %u\n",
2401 			     1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2402 		} else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2403 			mlog(ML_ERROR, "bad root_blkno: 0\n");
2404 		} else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2405 			mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2406 		} else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2407 			mlog(ML_ERROR,
2408 			     "Superblock slots found greater than file system "
2409 			     "maximum: found %u, max %u\n",
2410 			     le16_to_cpu(di->id2.i_super.s_max_slots),
2411 			     OCFS2_MAX_SLOTS);
2412 		} else {
2413 			/* found it! */
2414 			status = 0;
2415 		}
2416 	}
2417 
2418 out:
2419 	if (status && status != -EAGAIN)
2420 		mlog_errno(status);
2421 	return status;
2422 }
2423 
ocfs2_check_volume(struct ocfs2_super * osb)2424 static int ocfs2_check_volume(struct ocfs2_super *osb)
2425 {
2426 	int status;
2427 	int dirty;
2428 	int local;
2429 	struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2430 						  * recover
2431 						  * ourselves. */
2432 
2433 	/* Init our journal object. */
2434 	status = ocfs2_journal_init(osb->journal, &dirty);
2435 	if (status < 0) {
2436 		mlog(ML_ERROR, "Could not initialize journal!\n");
2437 		goto finally;
2438 	}
2439 
2440 	/* Now that journal has been initialized, check to make sure
2441 	   entire volume is addressable. */
2442 	status = ocfs2_journal_addressable(osb);
2443 	if (status)
2444 		goto finally;
2445 
2446 	/* If the journal was unmounted cleanly then we don't want to
2447 	 * recover anything. Otherwise, journal_load will do that
2448 	 * dirty work for us :) */
2449 	if (!dirty) {
2450 		status = ocfs2_journal_wipe(osb->journal, 0);
2451 		if (status < 0) {
2452 			mlog_errno(status);
2453 			goto finally;
2454 		}
2455 	} else {
2456 		printk(KERN_NOTICE "ocfs2: File system on device (%s) was not "
2457 		       "unmounted cleanly, recovering it.\n", osb->dev_str);
2458 	}
2459 
2460 	local = ocfs2_mount_local(osb);
2461 
2462 	/* will play back anything left in the journal. */
2463 	status = ocfs2_journal_load(osb->journal, local, dirty);
2464 	if (status < 0) {
2465 		mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2466 		goto finally;
2467 	}
2468 
2469 	if (dirty) {
2470 		/* recover my local alloc if we didn't unmount cleanly. */
2471 		status = ocfs2_begin_local_alloc_recovery(osb,
2472 							  osb->slot_num,
2473 							  &local_alloc);
2474 		if (status < 0) {
2475 			mlog_errno(status);
2476 			goto finally;
2477 		}
2478 		/* we complete the recovery process after we've marked
2479 		 * ourselves as mounted. */
2480 	}
2481 
2482 	status = ocfs2_load_local_alloc(osb);
2483 	if (status < 0) {
2484 		mlog_errno(status);
2485 		goto finally;
2486 	}
2487 
2488 	if (dirty) {
2489 		/* Recovery will be completed after we've mounted the
2490 		 * rest of the volume. */
2491 		osb->dirty = 1;
2492 		osb->local_alloc_copy = local_alloc;
2493 		local_alloc = NULL;
2494 	}
2495 
2496 	/* go through each journal, trylock it and if you get the
2497 	 * lock, and it's marked as dirty, set the bit in the recover
2498 	 * map and launch a recovery thread for it. */
2499 	status = ocfs2_mark_dead_nodes(osb);
2500 	if (status < 0) {
2501 		mlog_errno(status);
2502 		goto finally;
2503 	}
2504 
2505 	status = ocfs2_compute_replay_slots(osb);
2506 	if (status < 0)
2507 		mlog_errno(status);
2508 
2509 finally:
2510 	kfree(local_alloc);
2511 
2512 	if (status)
2513 		mlog_errno(status);
2514 	return status;
2515 }
2516 
2517 /*
2518  * The routine gets called from dismount or close whenever a dismount on
2519  * volume is requested and the osb open count becomes 1.
2520  * It will remove the osb from the global list and also free up all the
2521  * initialized resources and fileobject.
2522  */
ocfs2_delete_osb(struct ocfs2_super * osb)2523 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2524 {
2525 	/* This function assumes that the caller has the main osb resource */
2526 
2527 	ocfs2_free_slot_info(osb);
2528 
2529 	kfree(osb->osb_orphan_wipes);
2530 	kfree(osb->slot_recovery_generations);
2531 	/* FIXME
2532 	 * This belongs in journal shutdown, but because we have to
2533 	 * allocate osb->journal at the start of ocfs2_initialize_osb(),
2534 	 * we free it here.
2535 	 */
2536 	kfree(osb->journal);
2537 	kfree(osb->local_alloc_copy);
2538 	kfree(osb->uuid_str);
2539 	kfree(osb->vol_label);
2540 	ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2541 	memset(osb, 0, sizeof(struct ocfs2_super));
2542 }
2543 
2544 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2545  * panic(). We do not support continue-on-error operation. */
ocfs2_handle_error(struct super_block * sb)2546 static void ocfs2_handle_error(struct super_block *sb)
2547 {
2548 	struct ocfs2_super *osb = OCFS2_SB(sb);
2549 
2550 	if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2551 		panic("OCFS2: (device %s): panic forced after error\n",
2552 		      sb->s_id);
2553 
2554 	ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2555 
2556 	if (sb->s_flags & MS_RDONLY &&
2557 	    (ocfs2_is_soft_readonly(osb) ||
2558 	     ocfs2_is_hard_readonly(osb)))
2559 		return;
2560 
2561 	printk(KERN_CRIT "File system is now read-only due to the potential "
2562 	       "of on-disk corruption. Please run fsck.ocfs2 once the file "
2563 	       "system is unmounted.\n");
2564 	sb->s_flags |= MS_RDONLY;
2565 	ocfs2_set_ro_flag(osb, 0);
2566 }
2567 
2568 static char error_buf[1024];
2569 
__ocfs2_error(struct super_block * sb,const char * function,const char * fmt,...)2570 void __ocfs2_error(struct super_block *sb,
2571 		   const char *function,
2572 		   const char *fmt, ...)
2573 {
2574 	va_list args;
2575 
2576 	va_start(args, fmt);
2577 	vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2578 	va_end(args);
2579 
2580 	/* Not using mlog here because we want to show the actual
2581 	 * function the error came from. */
2582 	printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2583 	       sb->s_id, function, error_buf);
2584 
2585 	ocfs2_handle_error(sb);
2586 }
2587 
2588 /* Handle critical errors. This is intentionally more drastic than
2589  * ocfs2_handle_error, so we only use for things like journal errors,
2590  * etc. */
__ocfs2_abort(struct super_block * sb,const char * function,const char * fmt,...)2591 void __ocfs2_abort(struct super_block* sb,
2592 		   const char *function,
2593 		   const char *fmt, ...)
2594 {
2595 	va_list args;
2596 
2597 	va_start(args, fmt);
2598 	vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2599 	va_end(args);
2600 
2601 	printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2602 	       sb->s_id, function, error_buf);
2603 
2604 	/* We don't have the cluster support yet to go straight to
2605 	 * hard readonly in here. Until then, we want to keep
2606 	 * ocfs2_abort() so that we can at least mark critical
2607 	 * errors.
2608 	 *
2609 	 * TODO: This should abort the journal and alert other nodes
2610 	 * that our slot needs recovery. */
2611 
2612 	/* Force a panic(). This stinks, but it's better than letting
2613 	 * things continue without having a proper hard readonly
2614 	 * here. */
2615 	if (!ocfs2_mount_local(OCFS2_SB(sb)))
2616 		OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2617 	ocfs2_handle_error(sb);
2618 }
2619 
2620 /*
2621  * Void signal blockers, because in-kernel sigprocmask() only fails
2622  * when SIG_* is wrong.
2623  */
ocfs2_block_signals(sigset_t * oldset)2624 void ocfs2_block_signals(sigset_t *oldset)
2625 {
2626 	int rc;
2627 	sigset_t blocked;
2628 
2629 	sigfillset(&blocked);
2630 	rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2631 	BUG_ON(rc);
2632 }
2633 
ocfs2_unblock_signals(sigset_t * oldset)2634 void ocfs2_unblock_signals(sigset_t *oldset)
2635 {
2636 	int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2637 	BUG_ON(rc);
2638 }
2639 
2640 module_init(ocfs2_init);
2641 module_exit(ocfs2_exit);
2642