xref: /linux/fs/ext2/super.c (revision 93d90ad708b8da6efc0e487b66111aa9db7f70c7)
1 /*
2  *  linux/fs/ext2/super.c
3  *
4  * Copyright (C) 1992, 1993, 1994, 1995
5  * Remy Card (card@masi.ibp.fr)
6  * Laboratoire MASI - Institut Blaise Pascal
7  * Universite Pierre et Marie Curie (Paris VI)
8  *
9  *  from
10  *
11  *  linux/fs/minix/inode.c
12  *
13  *  Copyright (C) 1991, 1992  Linus Torvalds
14  *
15  *  Big-endian to little-endian byte-swapping/bitmaps by
16  *        David S. Miller (davem@caip.rutgers.edu), 1995
17  */
18 
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/slab.h>
23 #include <linux/init.h>
24 #include <linux/blkdev.h>
25 #include <linux/parser.h>
26 #include <linux/random.h>
27 #include <linux/buffer_head.h>
28 #include <linux/exportfs.h>
29 #include <linux/vfs.h>
30 #include <linux/seq_file.h>
31 #include <linux/mount.h>
32 #include <linux/log2.h>
33 #include <linux/quotaops.h>
34 #include <asm/uaccess.h>
35 #include "ext2.h"
36 #include "xattr.h"
37 #include "acl.h"
38 #include "xip.h"
39 
40 static void ext2_sync_super(struct super_block *sb,
41 			    struct ext2_super_block *es, int wait);
42 static int ext2_remount (struct super_block * sb, int * flags, char * data);
43 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
44 static int ext2_sync_fs(struct super_block *sb, int wait);
45 static int ext2_freeze(struct super_block *sb);
46 static int ext2_unfreeze(struct super_block *sb);
47 
48 void ext2_error(struct super_block *sb, const char *function,
49 		const char *fmt, ...)
50 {
51 	struct va_format vaf;
52 	va_list args;
53 	struct ext2_sb_info *sbi = EXT2_SB(sb);
54 	struct ext2_super_block *es = sbi->s_es;
55 
56 	if (!(sb->s_flags & MS_RDONLY)) {
57 		spin_lock(&sbi->s_lock);
58 		sbi->s_mount_state |= EXT2_ERROR_FS;
59 		es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
60 		spin_unlock(&sbi->s_lock);
61 		ext2_sync_super(sb, es, 1);
62 	}
63 
64 	va_start(args, fmt);
65 
66 	vaf.fmt = fmt;
67 	vaf.va = &args;
68 
69 	printk(KERN_CRIT "EXT2-fs (%s): error: %s: %pV\n",
70 	       sb->s_id, function, &vaf);
71 
72 	va_end(args);
73 
74 	if (test_opt(sb, ERRORS_PANIC))
75 		panic("EXT2-fs: panic from previous error\n");
76 	if (test_opt(sb, ERRORS_RO)) {
77 		ext2_msg(sb, KERN_CRIT,
78 			     "error: remounting filesystem read-only");
79 		sb->s_flags |= MS_RDONLY;
80 	}
81 }
82 
83 void ext2_msg(struct super_block *sb, const char *prefix,
84 		const char *fmt, ...)
85 {
86 	struct va_format vaf;
87 	va_list args;
88 
89 	va_start(args, fmt);
90 
91 	vaf.fmt = fmt;
92 	vaf.va = &args;
93 
94 	printk("%sEXT2-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
95 
96 	va_end(args);
97 }
98 
99 /*
100  * This must be called with sbi->s_lock held.
101  */
102 void ext2_update_dynamic_rev(struct super_block *sb)
103 {
104 	struct ext2_super_block *es = EXT2_SB(sb)->s_es;
105 
106 	if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
107 		return;
108 
109 	ext2_msg(sb, KERN_WARNING,
110 		     "warning: updating to rev %d because of "
111 		     "new feature flag, running e2fsck is recommended",
112 		     EXT2_DYNAMIC_REV);
113 
114 	es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
115 	es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
116 	es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
117 	/* leave es->s_feature_*compat flags alone */
118 	/* es->s_uuid will be set by e2fsck if empty */
119 
120 	/*
121 	 * The rest of the superblock fields should be zero, and if not it
122 	 * means they are likely already in use, so leave them alone.  We
123 	 * can leave it up to e2fsck to clean up any inconsistencies there.
124 	 */
125 }
126 
127 static void ext2_put_super (struct super_block * sb)
128 {
129 	int db_count;
130 	int i;
131 	struct ext2_sb_info *sbi = EXT2_SB(sb);
132 
133 	dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
134 
135 	ext2_xattr_put_super(sb);
136 	if (!(sb->s_flags & MS_RDONLY)) {
137 		struct ext2_super_block *es = sbi->s_es;
138 
139 		spin_lock(&sbi->s_lock);
140 		es->s_state = cpu_to_le16(sbi->s_mount_state);
141 		spin_unlock(&sbi->s_lock);
142 		ext2_sync_super(sb, es, 1);
143 	}
144 	db_count = sbi->s_gdb_count;
145 	for (i = 0; i < db_count; i++)
146 		if (sbi->s_group_desc[i])
147 			brelse (sbi->s_group_desc[i]);
148 	kfree(sbi->s_group_desc);
149 	kfree(sbi->s_debts);
150 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
151 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
152 	percpu_counter_destroy(&sbi->s_dirs_counter);
153 	brelse (sbi->s_sbh);
154 	sb->s_fs_info = NULL;
155 	kfree(sbi->s_blockgroup_lock);
156 	kfree(sbi);
157 }
158 
159 static struct kmem_cache * ext2_inode_cachep;
160 
161 static struct inode *ext2_alloc_inode(struct super_block *sb)
162 {
163 	struct ext2_inode_info *ei;
164 	ei = kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
165 	if (!ei)
166 		return NULL;
167 	ei->i_block_alloc_info = NULL;
168 	ei->vfs_inode.i_version = 1;
169 #ifdef CONFIG_QUOTA
170 	memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
171 #endif
172 
173 	return &ei->vfs_inode;
174 }
175 
176 static void ext2_i_callback(struct rcu_head *head)
177 {
178 	struct inode *inode = container_of(head, struct inode, i_rcu);
179 	kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
180 }
181 
182 static void ext2_destroy_inode(struct inode *inode)
183 {
184 	call_rcu(&inode->i_rcu, ext2_i_callback);
185 }
186 
187 static void init_once(void *foo)
188 {
189 	struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
190 
191 	rwlock_init(&ei->i_meta_lock);
192 #ifdef CONFIG_EXT2_FS_XATTR
193 	init_rwsem(&ei->xattr_sem);
194 #endif
195 	mutex_init(&ei->truncate_mutex);
196 	inode_init_once(&ei->vfs_inode);
197 }
198 
199 static int __init init_inodecache(void)
200 {
201 	ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
202 					     sizeof(struct ext2_inode_info),
203 					     0, (SLAB_RECLAIM_ACCOUNT|
204 						SLAB_MEM_SPREAD),
205 					     init_once);
206 	if (ext2_inode_cachep == NULL)
207 		return -ENOMEM;
208 	return 0;
209 }
210 
211 static void destroy_inodecache(void)
212 {
213 	/*
214 	 * Make sure all delayed rcu free inodes are flushed before we
215 	 * destroy cache.
216 	 */
217 	rcu_barrier();
218 	kmem_cache_destroy(ext2_inode_cachep);
219 }
220 
221 static int ext2_show_options(struct seq_file *seq, struct dentry *root)
222 {
223 	struct super_block *sb = root->d_sb;
224 	struct ext2_sb_info *sbi = EXT2_SB(sb);
225 	struct ext2_super_block *es = sbi->s_es;
226 	unsigned long def_mount_opts;
227 
228 	spin_lock(&sbi->s_lock);
229 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
230 
231 	if (sbi->s_sb_block != 1)
232 		seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
233 	if (test_opt(sb, MINIX_DF))
234 		seq_puts(seq, ",minixdf");
235 	if (test_opt(sb, GRPID))
236 		seq_puts(seq, ",grpid");
237 	if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
238 		seq_puts(seq, ",nogrpid");
239 	if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT2_DEF_RESUID)) ||
240 	    le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
241 		seq_printf(seq, ",resuid=%u",
242 				from_kuid_munged(&init_user_ns, sbi->s_resuid));
243 	}
244 	if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT2_DEF_RESGID)) ||
245 	    le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
246 		seq_printf(seq, ",resgid=%u",
247 				from_kgid_munged(&init_user_ns, sbi->s_resgid));
248 	}
249 	if (test_opt(sb, ERRORS_RO)) {
250 		int def_errors = le16_to_cpu(es->s_errors);
251 
252 		if (def_errors == EXT2_ERRORS_PANIC ||
253 		    def_errors == EXT2_ERRORS_CONTINUE) {
254 			seq_puts(seq, ",errors=remount-ro");
255 		}
256 	}
257 	if (test_opt(sb, ERRORS_CONT))
258 		seq_puts(seq, ",errors=continue");
259 	if (test_opt(sb, ERRORS_PANIC))
260 		seq_puts(seq, ",errors=panic");
261 	if (test_opt(sb, NO_UID32))
262 		seq_puts(seq, ",nouid32");
263 	if (test_opt(sb, DEBUG))
264 		seq_puts(seq, ",debug");
265 	if (test_opt(sb, OLDALLOC))
266 		seq_puts(seq, ",oldalloc");
267 
268 #ifdef CONFIG_EXT2_FS_XATTR
269 	if (test_opt(sb, XATTR_USER))
270 		seq_puts(seq, ",user_xattr");
271 	if (!test_opt(sb, XATTR_USER) &&
272 	    (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
273 		seq_puts(seq, ",nouser_xattr");
274 	}
275 #endif
276 
277 #ifdef CONFIG_EXT2_FS_POSIX_ACL
278 	if (test_opt(sb, POSIX_ACL))
279 		seq_puts(seq, ",acl");
280 	if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
281 		seq_puts(seq, ",noacl");
282 #endif
283 
284 	if (test_opt(sb, NOBH))
285 		seq_puts(seq, ",nobh");
286 
287 #if defined(CONFIG_QUOTA)
288 	if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
289 		seq_puts(seq, ",usrquota");
290 
291 	if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
292 		seq_puts(seq, ",grpquota");
293 #endif
294 
295 #if defined(CONFIG_EXT2_FS_XIP)
296 	if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
297 		seq_puts(seq, ",xip");
298 #endif
299 
300 	if (!test_opt(sb, RESERVATION))
301 		seq_puts(seq, ",noreservation");
302 
303 	spin_unlock(&sbi->s_lock);
304 	return 0;
305 }
306 
307 #ifdef CONFIG_QUOTA
308 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
309 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
310 static struct dquot **ext2_get_dquots(struct inode *inode)
311 {
312 	return EXT2_I(inode)->i_dquot;
313 }
314 #endif
315 
316 static const struct super_operations ext2_sops = {
317 	.alloc_inode	= ext2_alloc_inode,
318 	.destroy_inode	= ext2_destroy_inode,
319 	.write_inode	= ext2_write_inode,
320 	.evict_inode	= ext2_evict_inode,
321 	.put_super	= ext2_put_super,
322 	.sync_fs	= ext2_sync_fs,
323 	.freeze_fs	= ext2_freeze,
324 	.unfreeze_fs	= ext2_unfreeze,
325 	.statfs		= ext2_statfs,
326 	.remount_fs	= ext2_remount,
327 	.show_options	= ext2_show_options,
328 #ifdef CONFIG_QUOTA
329 	.quota_read	= ext2_quota_read,
330 	.quota_write	= ext2_quota_write,
331 	.get_dquots	= ext2_get_dquots,
332 #endif
333 };
334 
335 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
336 		u64 ino, u32 generation)
337 {
338 	struct inode *inode;
339 
340 	if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
341 		return ERR_PTR(-ESTALE);
342 	if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
343 		return ERR_PTR(-ESTALE);
344 
345 	/*
346 	 * ext2_iget isn't quite right if the inode is currently unallocated!
347 	 * However ext2_iget currently does appropriate checks to handle stale
348 	 * inodes so everything is OK.
349 	 */
350 	inode = ext2_iget(sb, ino);
351 	if (IS_ERR(inode))
352 		return ERR_CAST(inode);
353 	if (generation && inode->i_generation != generation) {
354 		/* we didn't find the right inode.. */
355 		iput(inode);
356 		return ERR_PTR(-ESTALE);
357 	}
358 	return inode;
359 }
360 
361 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
362 		int fh_len, int fh_type)
363 {
364 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
365 				    ext2_nfs_get_inode);
366 }
367 
368 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
369 		int fh_len, int fh_type)
370 {
371 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
372 				    ext2_nfs_get_inode);
373 }
374 
375 static const struct export_operations ext2_export_ops = {
376 	.fh_to_dentry = ext2_fh_to_dentry,
377 	.fh_to_parent = ext2_fh_to_parent,
378 	.get_parent = ext2_get_parent,
379 };
380 
381 static unsigned long get_sb_block(void **data)
382 {
383 	unsigned long 	sb_block;
384 	char 		*options = (char *) *data;
385 
386 	if (!options || strncmp(options, "sb=", 3) != 0)
387 		return 1;	/* Default location */
388 	options += 3;
389 	sb_block = simple_strtoul(options, &options, 0);
390 	if (*options && *options != ',') {
391 		printk("EXT2-fs: Invalid sb specification: %s\n",
392 		       (char *) *data);
393 		return 1;
394 	}
395 	if (*options == ',')
396 		options++;
397 	*data = (void *) options;
398 	return sb_block;
399 }
400 
401 enum {
402 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
403 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
404 	Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
405 	Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
406 	Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
407 	Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
408 };
409 
410 static const match_table_t tokens = {
411 	{Opt_bsd_df, "bsddf"},
412 	{Opt_minix_df, "minixdf"},
413 	{Opt_grpid, "grpid"},
414 	{Opt_grpid, "bsdgroups"},
415 	{Opt_nogrpid, "nogrpid"},
416 	{Opt_nogrpid, "sysvgroups"},
417 	{Opt_resgid, "resgid=%u"},
418 	{Opt_resuid, "resuid=%u"},
419 	{Opt_sb, "sb=%u"},
420 	{Opt_err_cont, "errors=continue"},
421 	{Opt_err_panic, "errors=panic"},
422 	{Opt_err_ro, "errors=remount-ro"},
423 	{Opt_nouid32, "nouid32"},
424 	{Opt_nocheck, "check=none"},
425 	{Opt_nocheck, "nocheck"},
426 	{Opt_debug, "debug"},
427 	{Opt_oldalloc, "oldalloc"},
428 	{Opt_orlov, "orlov"},
429 	{Opt_nobh, "nobh"},
430 	{Opt_user_xattr, "user_xattr"},
431 	{Opt_nouser_xattr, "nouser_xattr"},
432 	{Opt_acl, "acl"},
433 	{Opt_noacl, "noacl"},
434 	{Opt_xip, "xip"},
435 	{Opt_grpquota, "grpquota"},
436 	{Opt_ignore, "noquota"},
437 	{Opt_quota, "quota"},
438 	{Opt_usrquota, "usrquota"},
439 	{Opt_reservation, "reservation"},
440 	{Opt_noreservation, "noreservation"},
441 	{Opt_err, NULL}
442 };
443 
444 static int parse_options(char *options, struct super_block *sb)
445 {
446 	char *p;
447 	struct ext2_sb_info *sbi = EXT2_SB(sb);
448 	substring_t args[MAX_OPT_ARGS];
449 	int option;
450 	kuid_t uid;
451 	kgid_t gid;
452 
453 	if (!options)
454 		return 1;
455 
456 	while ((p = strsep (&options, ",")) != NULL) {
457 		int token;
458 		if (!*p)
459 			continue;
460 
461 		token = match_token(p, tokens, args);
462 		switch (token) {
463 		case Opt_bsd_df:
464 			clear_opt (sbi->s_mount_opt, MINIX_DF);
465 			break;
466 		case Opt_minix_df:
467 			set_opt (sbi->s_mount_opt, MINIX_DF);
468 			break;
469 		case Opt_grpid:
470 			set_opt (sbi->s_mount_opt, GRPID);
471 			break;
472 		case Opt_nogrpid:
473 			clear_opt (sbi->s_mount_opt, GRPID);
474 			break;
475 		case Opt_resuid:
476 			if (match_int(&args[0], &option))
477 				return 0;
478 			uid = make_kuid(current_user_ns(), option);
479 			if (!uid_valid(uid)) {
480 				ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
481 				return 0;
482 
483 			}
484 			sbi->s_resuid = uid;
485 			break;
486 		case Opt_resgid:
487 			if (match_int(&args[0], &option))
488 				return 0;
489 			gid = make_kgid(current_user_ns(), option);
490 			if (!gid_valid(gid)) {
491 				ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
492 				return 0;
493 			}
494 			sbi->s_resgid = gid;
495 			break;
496 		case Opt_sb:
497 			/* handled by get_sb_block() instead of here */
498 			/* *sb_block = match_int(&args[0]); */
499 			break;
500 		case Opt_err_panic:
501 			clear_opt (sbi->s_mount_opt, ERRORS_CONT);
502 			clear_opt (sbi->s_mount_opt, ERRORS_RO);
503 			set_opt (sbi->s_mount_opt, ERRORS_PANIC);
504 			break;
505 		case Opt_err_ro:
506 			clear_opt (sbi->s_mount_opt, ERRORS_CONT);
507 			clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
508 			set_opt (sbi->s_mount_opt, ERRORS_RO);
509 			break;
510 		case Opt_err_cont:
511 			clear_opt (sbi->s_mount_opt, ERRORS_RO);
512 			clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
513 			set_opt (sbi->s_mount_opt, ERRORS_CONT);
514 			break;
515 		case Opt_nouid32:
516 			set_opt (sbi->s_mount_opt, NO_UID32);
517 			break;
518 		case Opt_nocheck:
519 			clear_opt (sbi->s_mount_opt, CHECK);
520 			break;
521 		case Opt_debug:
522 			set_opt (sbi->s_mount_opt, DEBUG);
523 			break;
524 		case Opt_oldalloc:
525 			set_opt (sbi->s_mount_opt, OLDALLOC);
526 			break;
527 		case Opt_orlov:
528 			clear_opt (sbi->s_mount_opt, OLDALLOC);
529 			break;
530 		case Opt_nobh:
531 			set_opt (sbi->s_mount_opt, NOBH);
532 			break;
533 #ifdef CONFIG_EXT2_FS_XATTR
534 		case Opt_user_xattr:
535 			set_opt (sbi->s_mount_opt, XATTR_USER);
536 			break;
537 		case Opt_nouser_xattr:
538 			clear_opt (sbi->s_mount_opt, XATTR_USER);
539 			break;
540 #else
541 		case Opt_user_xattr:
542 		case Opt_nouser_xattr:
543 			ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
544 				"not supported");
545 			break;
546 #endif
547 #ifdef CONFIG_EXT2_FS_POSIX_ACL
548 		case Opt_acl:
549 			set_opt(sbi->s_mount_opt, POSIX_ACL);
550 			break;
551 		case Opt_noacl:
552 			clear_opt(sbi->s_mount_opt, POSIX_ACL);
553 			break;
554 #else
555 		case Opt_acl:
556 		case Opt_noacl:
557 			ext2_msg(sb, KERN_INFO,
558 				"(no)acl options not supported");
559 			break;
560 #endif
561 		case Opt_xip:
562 #ifdef CONFIG_EXT2_FS_XIP
563 			set_opt (sbi->s_mount_opt, XIP);
564 #else
565 			ext2_msg(sb, KERN_INFO, "xip option not supported");
566 #endif
567 			break;
568 
569 #if defined(CONFIG_QUOTA)
570 		case Opt_quota:
571 		case Opt_usrquota:
572 			set_opt(sbi->s_mount_opt, USRQUOTA);
573 			break;
574 
575 		case Opt_grpquota:
576 			set_opt(sbi->s_mount_opt, GRPQUOTA);
577 			break;
578 #else
579 		case Opt_quota:
580 		case Opt_usrquota:
581 		case Opt_grpquota:
582 			ext2_msg(sb, KERN_INFO,
583 				"quota operations not supported");
584 			break;
585 #endif
586 
587 		case Opt_reservation:
588 			set_opt(sbi->s_mount_opt, RESERVATION);
589 			ext2_msg(sb, KERN_INFO, "reservations ON");
590 			break;
591 		case Opt_noreservation:
592 			clear_opt(sbi->s_mount_opt, RESERVATION);
593 			ext2_msg(sb, KERN_INFO, "reservations OFF");
594 			break;
595 		case Opt_ignore:
596 			break;
597 		default:
598 			return 0;
599 		}
600 	}
601 	return 1;
602 }
603 
604 static int ext2_setup_super (struct super_block * sb,
605 			      struct ext2_super_block * es,
606 			      int read_only)
607 {
608 	int res = 0;
609 	struct ext2_sb_info *sbi = EXT2_SB(sb);
610 
611 	if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
612 		ext2_msg(sb, KERN_ERR,
613 			"error: revision level too high, "
614 			"forcing read-only mode");
615 		res = MS_RDONLY;
616 	}
617 	if (read_only)
618 		return res;
619 	if (!(sbi->s_mount_state & EXT2_VALID_FS))
620 		ext2_msg(sb, KERN_WARNING,
621 			"warning: mounting unchecked fs, "
622 			"running e2fsck is recommended");
623 	else if ((sbi->s_mount_state & EXT2_ERROR_FS))
624 		ext2_msg(sb, KERN_WARNING,
625 			"warning: mounting fs with errors, "
626 			"running e2fsck is recommended");
627 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
628 		 le16_to_cpu(es->s_mnt_count) >=
629 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
630 		ext2_msg(sb, KERN_WARNING,
631 			"warning: maximal mount count reached, "
632 			"running e2fsck is recommended");
633 	else if (le32_to_cpu(es->s_checkinterval) &&
634 		(le32_to_cpu(es->s_lastcheck) +
635 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
636 		ext2_msg(sb, KERN_WARNING,
637 			"warning: checktime reached, "
638 			"running e2fsck is recommended");
639 	if (!le16_to_cpu(es->s_max_mnt_count))
640 		es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
641 	le16_add_cpu(&es->s_mnt_count, 1);
642 	if (test_opt (sb, DEBUG))
643 		ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
644 			"bpg=%lu, ipg=%lu, mo=%04lx]",
645 			EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
646 			sbi->s_frag_size,
647 			sbi->s_groups_count,
648 			EXT2_BLOCKS_PER_GROUP(sb),
649 			EXT2_INODES_PER_GROUP(sb),
650 			sbi->s_mount_opt);
651 	return res;
652 }
653 
654 static int ext2_check_descriptors(struct super_block *sb)
655 {
656 	int i;
657 	struct ext2_sb_info *sbi = EXT2_SB(sb);
658 
659 	ext2_debug ("Checking group descriptors");
660 
661 	for (i = 0; i < sbi->s_groups_count; i++) {
662 		struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
663 		ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
664 		ext2_fsblk_t last_block;
665 
666 		if (i == sbi->s_groups_count - 1)
667 			last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
668 		else
669 			last_block = first_block +
670 				(EXT2_BLOCKS_PER_GROUP(sb) - 1);
671 
672 		if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
673 		    le32_to_cpu(gdp->bg_block_bitmap) > last_block)
674 		{
675 			ext2_error (sb, "ext2_check_descriptors",
676 				    "Block bitmap for group %d"
677 				    " not in group (block %lu)!",
678 				    i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
679 			return 0;
680 		}
681 		if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
682 		    le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
683 		{
684 			ext2_error (sb, "ext2_check_descriptors",
685 				    "Inode bitmap for group %d"
686 				    " not in group (block %lu)!",
687 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
688 			return 0;
689 		}
690 		if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
691 		    le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
692 		    last_block)
693 		{
694 			ext2_error (sb, "ext2_check_descriptors",
695 				    "Inode table for group %d"
696 				    " not in group (block %lu)!",
697 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
698 			return 0;
699 		}
700 	}
701 	return 1;
702 }
703 
704 /*
705  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
706  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
707  * We need to be 1 filesystem block less than the 2^32 sector limit.
708  */
709 static loff_t ext2_max_size(int bits)
710 {
711 	loff_t res = EXT2_NDIR_BLOCKS;
712 	int meta_blocks;
713 	loff_t upper_limit;
714 
715 	/* This is calculated to be the largest file size for a
716 	 * dense, file such that the total number of
717 	 * sectors in the file, including data and all indirect blocks,
718 	 * does not exceed 2^32 -1
719 	 * __u32 i_blocks representing the total number of
720 	 * 512 bytes blocks of the file
721 	 */
722 	upper_limit = (1LL << 32) - 1;
723 
724 	/* total blocks in file system block size */
725 	upper_limit >>= (bits - 9);
726 
727 
728 	/* indirect blocks */
729 	meta_blocks = 1;
730 	/* double indirect blocks */
731 	meta_blocks += 1 + (1LL << (bits-2));
732 	/* tripple indirect blocks */
733 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
734 
735 	upper_limit -= meta_blocks;
736 	upper_limit <<= bits;
737 
738 	res += 1LL << (bits-2);
739 	res += 1LL << (2*(bits-2));
740 	res += 1LL << (3*(bits-2));
741 	res <<= bits;
742 	if (res > upper_limit)
743 		res = upper_limit;
744 
745 	if (res > MAX_LFS_FILESIZE)
746 		res = MAX_LFS_FILESIZE;
747 
748 	return res;
749 }
750 
751 static unsigned long descriptor_loc(struct super_block *sb,
752 				    unsigned long logic_sb_block,
753 				    int nr)
754 {
755 	struct ext2_sb_info *sbi = EXT2_SB(sb);
756 	unsigned long bg, first_meta_bg;
757 	int has_super = 0;
758 
759 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
760 
761 	if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
762 	    nr < first_meta_bg)
763 		return (logic_sb_block + nr + 1);
764 	bg = sbi->s_desc_per_block * nr;
765 	if (ext2_bg_has_super(sb, bg))
766 		has_super = 1;
767 
768 	return ext2_group_first_block_no(sb, bg) + has_super;
769 }
770 
771 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
772 {
773 	struct buffer_head * bh;
774 	struct ext2_sb_info * sbi;
775 	struct ext2_super_block * es;
776 	struct inode *root;
777 	unsigned long block;
778 	unsigned long sb_block = get_sb_block(&data);
779 	unsigned long logic_sb_block;
780 	unsigned long offset = 0;
781 	unsigned long def_mount_opts;
782 	long ret = -EINVAL;
783 	int blocksize = BLOCK_SIZE;
784 	int db_count;
785 	int i, j;
786 	__le32 features;
787 	int err;
788 
789 	err = -ENOMEM;
790 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
791 	if (!sbi)
792 		goto failed;
793 
794 	sbi->s_blockgroup_lock =
795 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
796 	if (!sbi->s_blockgroup_lock) {
797 		kfree(sbi);
798 		goto failed;
799 	}
800 	sb->s_fs_info = sbi;
801 	sbi->s_sb_block = sb_block;
802 
803 	spin_lock_init(&sbi->s_lock);
804 
805 	/*
806 	 * See what the current blocksize for the device is, and
807 	 * use that as the blocksize.  Otherwise (or if the blocksize
808 	 * is smaller than the default) use the default.
809 	 * This is important for devices that have a hardware
810 	 * sectorsize that is larger than the default.
811 	 */
812 	blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
813 	if (!blocksize) {
814 		ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
815 		goto failed_sbi;
816 	}
817 
818 	/*
819 	 * If the superblock doesn't start on a hardware sector boundary,
820 	 * calculate the offset.
821 	 */
822 	if (blocksize != BLOCK_SIZE) {
823 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
824 		offset = (sb_block*BLOCK_SIZE) % blocksize;
825 	} else {
826 		logic_sb_block = sb_block;
827 	}
828 
829 	if (!(bh = sb_bread(sb, logic_sb_block))) {
830 		ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
831 		goto failed_sbi;
832 	}
833 	/*
834 	 * Note: s_es must be initialized as soon as possible because
835 	 *       some ext2 macro-instructions depend on its value
836 	 */
837 	es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
838 	sbi->s_es = es;
839 	sb->s_magic = le16_to_cpu(es->s_magic);
840 
841 	if (sb->s_magic != EXT2_SUPER_MAGIC)
842 		goto cantfind_ext2;
843 
844 	/* Set defaults before we parse the mount options */
845 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
846 	if (def_mount_opts & EXT2_DEFM_DEBUG)
847 		set_opt(sbi->s_mount_opt, DEBUG);
848 	if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
849 		set_opt(sbi->s_mount_opt, GRPID);
850 	if (def_mount_opts & EXT2_DEFM_UID16)
851 		set_opt(sbi->s_mount_opt, NO_UID32);
852 #ifdef CONFIG_EXT2_FS_XATTR
853 	if (def_mount_opts & EXT2_DEFM_XATTR_USER)
854 		set_opt(sbi->s_mount_opt, XATTR_USER);
855 #endif
856 #ifdef CONFIG_EXT2_FS_POSIX_ACL
857 	if (def_mount_opts & EXT2_DEFM_ACL)
858 		set_opt(sbi->s_mount_opt, POSIX_ACL);
859 #endif
860 
861 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
862 		set_opt(sbi->s_mount_opt, ERRORS_PANIC);
863 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
864 		set_opt(sbi->s_mount_opt, ERRORS_CONT);
865 	else
866 		set_opt(sbi->s_mount_opt, ERRORS_RO);
867 
868 	sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
869 	sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
870 
871 	set_opt(sbi->s_mount_opt, RESERVATION);
872 
873 	if (!parse_options((char *) data, sb))
874 		goto failed_mount;
875 
876 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
877 		((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
878 		 MS_POSIXACL : 0);
879 
880 	ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
881 				    EXT2_MOUNT_XIP if not */
882 
883 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
884 	    (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
885 	     EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
886 	     EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
887 		ext2_msg(sb, KERN_WARNING,
888 			"warning: feature flags set on rev 0 fs, "
889 			"running e2fsck is recommended");
890 	/*
891 	 * Check feature flags regardless of the revision level, since we
892 	 * previously didn't change the revision level when setting the flags,
893 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
894 	 */
895 	features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
896 	if (features) {
897 		ext2_msg(sb, KERN_ERR,	"error: couldn't mount because of "
898 		       "unsupported optional features (%x)",
899 			le32_to_cpu(features));
900 		goto failed_mount;
901 	}
902 	if (!(sb->s_flags & MS_RDONLY) &&
903 	    (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
904 		ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
905 		       "unsupported optional features (%x)",
906 		       le32_to_cpu(features));
907 		goto failed_mount;
908 	}
909 
910 	blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
911 
912 	if (ext2_use_xip(sb) && blocksize != PAGE_SIZE) {
913 		if (!silent)
914 			ext2_msg(sb, KERN_ERR,
915 				"error: unsupported blocksize for xip");
916 		goto failed_mount;
917 	}
918 
919 	/* If the blocksize doesn't match, re-read the thing.. */
920 	if (sb->s_blocksize != blocksize) {
921 		brelse(bh);
922 
923 		if (!sb_set_blocksize(sb, blocksize)) {
924 			ext2_msg(sb, KERN_ERR,
925 				"error: bad blocksize %d", blocksize);
926 			goto failed_sbi;
927 		}
928 
929 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
930 		offset = (sb_block*BLOCK_SIZE) % blocksize;
931 		bh = sb_bread(sb, logic_sb_block);
932 		if(!bh) {
933 			ext2_msg(sb, KERN_ERR, "error: couldn't read"
934 				"superblock on 2nd try");
935 			goto failed_sbi;
936 		}
937 		es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
938 		sbi->s_es = es;
939 		if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
940 			ext2_msg(sb, KERN_ERR, "error: magic mismatch");
941 			goto failed_mount;
942 		}
943 	}
944 
945 	sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
946 	sb->s_max_links = EXT2_LINK_MAX;
947 
948 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
949 		sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
950 		sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
951 	} else {
952 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
953 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
954 		if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
955 		    !is_power_of_2(sbi->s_inode_size) ||
956 		    (sbi->s_inode_size > blocksize)) {
957 			ext2_msg(sb, KERN_ERR,
958 				"error: unsupported inode size: %d",
959 				sbi->s_inode_size);
960 			goto failed_mount;
961 		}
962 	}
963 
964 	sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
965 				   le32_to_cpu(es->s_log_frag_size);
966 	if (sbi->s_frag_size == 0)
967 		goto cantfind_ext2;
968 	sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
969 
970 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
971 	sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
972 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
973 
974 	if (EXT2_INODE_SIZE(sb) == 0)
975 		goto cantfind_ext2;
976 	sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
977 	if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
978 		goto cantfind_ext2;
979 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
980 					sbi->s_inodes_per_block;
981 	sbi->s_desc_per_block = sb->s_blocksize /
982 					sizeof (struct ext2_group_desc);
983 	sbi->s_sbh = bh;
984 	sbi->s_mount_state = le16_to_cpu(es->s_state);
985 	sbi->s_addr_per_block_bits =
986 		ilog2 (EXT2_ADDR_PER_BLOCK(sb));
987 	sbi->s_desc_per_block_bits =
988 		ilog2 (EXT2_DESC_PER_BLOCK(sb));
989 
990 	if (sb->s_magic != EXT2_SUPER_MAGIC)
991 		goto cantfind_ext2;
992 
993 	if (sb->s_blocksize != bh->b_size) {
994 		if (!silent)
995 			ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
996 		goto failed_mount;
997 	}
998 
999 	if (sb->s_blocksize != sbi->s_frag_size) {
1000 		ext2_msg(sb, KERN_ERR,
1001 			"error: fragsize %lu != blocksize %lu"
1002 			"(not supported yet)",
1003 			sbi->s_frag_size, sb->s_blocksize);
1004 		goto failed_mount;
1005 	}
1006 
1007 	if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
1008 		ext2_msg(sb, KERN_ERR,
1009 			"error: #blocks per group too big: %lu",
1010 			sbi->s_blocks_per_group);
1011 		goto failed_mount;
1012 	}
1013 	if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
1014 		ext2_msg(sb, KERN_ERR,
1015 			"error: #fragments per group too big: %lu",
1016 			sbi->s_frags_per_group);
1017 		goto failed_mount;
1018 	}
1019 	if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1020 		ext2_msg(sb, KERN_ERR,
1021 			"error: #inodes per group too big: %lu",
1022 			sbi->s_inodes_per_group);
1023 		goto failed_mount;
1024 	}
1025 
1026 	if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1027 		goto cantfind_ext2;
1028  	sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1029  				le32_to_cpu(es->s_first_data_block) - 1)
1030  					/ EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1031 	db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1032 		   EXT2_DESC_PER_BLOCK(sb);
1033 	sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
1034 	if (sbi->s_group_desc == NULL) {
1035 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1036 		goto failed_mount;
1037 	}
1038 	bgl_lock_init(sbi->s_blockgroup_lock);
1039 	sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1040 	if (!sbi->s_debts) {
1041 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1042 		goto failed_mount_group_desc;
1043 	}
1044 	for (i = 0; i < db_count; i++) {
1045 		block = descriptor_loc(sb, logic_sb_block, i);
1046 		sbi->s_group_desc[i] = sb_bread(sb, block);
1047 		if (!sbi->s_group_desc[i]) {
1048 			for (j = 0; j < i; j++)
1049 				brelse (sbi->s_group_desc[j]);
1050 			ext2_msg(sb, KERN_ERR,
1051 				"error: unable to read group descriptors");
1052 			goto failed_mount_group_desc;
1053 		}
1054 	}
1055 	if (!ext2_check_descriptors (sb)) {
1056 		ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1057 		goto failed_mount2;
1058 	}
1059 	sbi->s_gdb_count = db_count;
1060 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1061 	spin_lock_init(&sbi->s_next_gen_lock);
1062 
1063 	/* per fileystem reservation list head & lock */
1064 	spin_lock_init(&sbi->s_rsv_window_lock);
1065 	sbi->s_rsv_window_root = RB_ROOT;
1066 	/*
1067 	 * Add a single, static dummy reservation to the start of the
1068 	 * reservation window list --- it gives us a placeholder for
1069 	 * append-at-start-of-list which makes the allocation logic
1070 	 * _much_ simpler.
1071 	 */
1072 	sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1073 	sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1074 	sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1075 	sbi->s_rsv_window_head.rsv_goal_size = 0;
1076 	ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1077 
1078 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
1079 				ext2_count_free_blocks(sb), GFP_KERNEL);
1080 	if (!err) {
1081 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
1082 				ext2_count_free_inodes(sb), GFP_KERNEL);
1083 	}
1084 	if (!err) {
1085 		err = percpu_counter_init(&sbi->s_dirs_counter,
1086 				ext2_count_dirs(sb), GFP_KERNEL);
1087 	}
1088 	if (err) {
1089 		ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1090 		goto failed_mount3;
1091 	}
1092 	/*
1093 	 * set up enough so that it can read an inode
1094 	 */
1095 	sb->s_op = &ext2_sops;
1096 	sb->s_export_op = &ext2_export_ops;
1097 	sb->s_xattr = ext2_xattr_handlers;
1098 
1099 #ifdef CONFIG_QUOTA
1100 	sb->dq_op = &dquot_operations;
1101 	sb->s_qcop = &dquot_quotactl_ops;
1102 	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1103 #endif
1104 
1105 	root = ext2_iget(sb, EXT2_ROOT_INO);
1106 	if (IS_ERR(root)) {
1107 		ret = PTR_ERR(root);
1108 		goto failed_mount3;
1109 	}
1110 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1111 		iput(root);
1112 		ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1113 		goto failed_mount3;
1114 	}
1115 
1116 	sb->s_root = d_make_root(root);
1117 	if (!sb->s_root) {
1118 		ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1119 		ret = -ENOMEM;
1120 		goto failed_mount3;
1121 	}
1122 	if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1123 		ext2_msg(sb, KERN_WARNING,
1124 			"warning: mounting ext3 filesystem as ext2");
1125 	if (ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY))
1126 		sb->s_flags |= MS_RDONLY;
1127 	ext2_write_super(sb);
1128 	return 0;
1129 
1130 cantfind_ext2:
1131 	if (!silent)
1132 		ext2_msg(sb, KERN_ERR,
1133 			"error: can't find an ext2 filesystem on dev %s.",
1134 			sb->s_id);
1135 	goto failed_mount;
1136 failed_mount3:
1137 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
1138 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
1139 	percpu_counter_destroy(&sbi->s_dirs_counter);
1140 failed_mount2:
1141 	for (i = 0; i < db_count; i++)
1142 		brelse(sbi->s_group_desc[i]);
1143 failed_mount_group_desc:
1144 	kfree(sbi->s_group_desc);
1145 	kfree(sbi->s_debts);
1146 failed_mount:
1147 	brelse(bh);
1148 failed_sbi:
1149 	sb->s_fs_info = NULL;
1150 	kfree(sbi->s_blockgroup_lock);
1151 	kfree(sbi);
1152 failed:
1153 	return ret;
1154 }
1155 
1156 static void ext2_clear_super_error(struct super_block *sb)
1157 {
1158 	struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1159 
1160 	if (buffer_write_io_error(sbh)) {
1161 		/*
1162 		 * Oh, dear.  A previous attempt to write the
1163 		 * superblock failed.  This could happen because the
1164 		 * USB device was yanked out.  Or it could happen to
1165 		 * be a transient write error and maybe the block will
1166 		 * be remapped.  Nothing we can do but to retry the
1167 		 * write and hope for the best.
1168 		 */
1169 		ext2_msg(sb, KERN_ERR,
1170 		       "previous I/O error to superblock detected\n");
1171 		clear_buffer_write_io_error(sbh);
1172 		set_buffer_uptodate(sbh);
1173 	}
1174 }
1175 
1176 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1177 			    int wait)
1178 {
1179 	ext2_clear_super_error(sb);
1180 	spin_lock(&EXT2_SB(sb)->s_lock);
1181 	es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1182 	es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1183 	es->s_wtime = cpu_to_le32(get_seconds());
1184 	/* unlock before we do IO */
1185 	spin_unlock(&EXT2_SB(sb)->s_lock);
1186 	mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1187 	if (wait)
1188 		sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1189 }
1190 
1191 /*
1192  * In the second extended file system, it is not necessary to
1193  * write the super block since we use a mapping of the
1194  * disk super block in a buffer.
1195  *
1196  * However, this function is still used to set the fs valid
1197  * flags to 0.  We need to set this flag to 0 since the fs
1198  * may have been checked while mounted and e2fsck may have
1199  * set s_state to EXT2_VALID_FS after some corrections.
1200  */
1201 static int ext2_sync_fs(struct super_block *sb, int wait)
1202 {
1203 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1204 	struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1205 
1206 	/*
1207 	 * Write quota structures to quota file, sync_blockdev() will write
1208 	 * them to disk later
1209 	 */
1210 	dquot_writeback_dquots(sb, -1);
1211 
1212 	spin_lock(&sbi->s_lock);
1213 	if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1214 		ext2_debug("setting valid to 0\n");
1215 		es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1216 	}
1217 	spin_unlock(&sbi->s_lock);
1218 	ext2_sync_super(sb, es, wait);
1219 	return 0;
1220 }
1221 
1222 static int ext2_freeze(struct super_block *sb)
1223 {
1224 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1225 
1226 	/*
1227 	 * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1228 	 * because we have unattached inodes and thus filesystem is not fully
1229 	 * consistent.
1230 	 */
1231 	if (atomic_long_read(&sb->s_remove_count)) {
1232 		ext2_sync_fs(sb, 1);
1233 		return 0;
1234 	}
1235 	/* Set EXT2_FS_VALID flag */
1236 	spin_lock(&sbi->s_lock);
1237 	sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1238 	spin_unlock(&sbi->s_lock);
1239 	ext2_sync_super(sb, sbi->s_es, 1);
1240 
1241 	return 0;
1242 }
1243 
1244 static int ext2_unfreeze(struct super_block *sb)
1245 {
1246 	/* Just write sb to clear EXT2_VALID_FS flag */
1247 	ext2_write_super(sb);
1248 
1249 	return 0;
1250 }
1251 
1252 void ext2_write_super(struct super_block *sb)
1253 {
1254 	if (!(sb->s_flags & MS_RDONLY))
1255 		ext2_sync_fs(sb, 1);
1256 }
1257 
1258 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1259 {
1260 	struct ext2_sb_info * sbi = EXT2_SB(sb);
1261 	struct ext2_super_block * es;
1262 	unsigned long old_mount_opt = sbi->s_mount_opt;
1263 	struct ext2_mount_options old_opts;
1264 	unsigned long old_sb_flags;
1265 	int err;
1266 
1267 	sync_filesystem(sb);
1268 	spin_lock(&sbi->s_lock);
1269 
1270 	/* Store the old options */
1271 	old_sb_flags = sb->s_flags;
1272 	old_opts.s_mount_opt = sbi->s_mount_opt;
1273 	old_opts.s_resuid = sbi->s_resuid;
1274 	old_opts.s_resgid = sbi->s_resgid;
1275 
1276 	/*
1277 	 * Allow the "check" option to be passed as a remount option.
1278 	 */
1279 	if (!parse_options(data, sb)) {
1280 		err = -EINVAL;
1281 		goto restore_opts;
1282 	}
1283 
1284 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1285 		((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1286 
1287 	ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1288 				    EXT2_MOUNT_XIP if not */
1289 
1290 	if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
1291 		ext2_msg(sb, KERN_WARNING,
1292 			"warning: unsupported blocksize for xip");
1293 		err = -EINVAL;
1294 		goto restore_opts;
1295 	}
1296 
1297 	es = sbi->s_es;
1298 	if ((sbi->s_mount_opt ^ old_mount_opt) & EXT2_MOUNT_XIP) {
1299 		ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1300 			 "xip flag with busy inodes while remounting");
1301 		sbi->s_mount_opt &= ~EXT2_MOUNT_XIP;
1302 		sbi->s_mount_opt |= old_mount_opt & EXT2_MOUNT_XIP;
1303 	}
1304 	if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) {
1305 		spin_unlock(&sbi->s_lock);
1306 		return 0;
1307 	}
1308 	if (*flags & MS_RDONLY) {
1309 		if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1310 		    !(sbi->s_mount_state & EXT2_VALID_FS)) {
1311 			spin_unlock(&sbi->s_lock);
1312 			return 0;
1313 		}
1314 
1315 		/*
1316 		 * OK, we are remounting a valid rw partition rdonly, so set
1317 		 * the rdonly flag and then mark the partition as valid again.
1318 		 */
1319 		es->s_state = cpu_to_le16(sbi->s_mount_state);
1320 		es->s_mtime = cpu_to_le32(get_seconds());
1321 		spin_unlock(&sbi->s_lock);
1322 
1323 		err = dquot_suspend(sb, -1);
1324 		if (err < 0) {
1325 			spin_lock(&sbi->s_lock);
1326 			goto restore_opts;
1327 		}
1328 
1329 		ext2_sync_super(sb, es, 1);
1330 	} else {
1331 		__le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1332 					       ~EXT2_FEATURE_RO_COMPAT_SUPP);
1333 		if (ret) {
1334 			ext2_msg(sb, KERN_WARNING,
1335 				"warning: couldn't remount RDWR because of "
1336 				"unsupported optional features (%x).",
1337 				le32_to_cpu(ret));
1338 			err = -EROFS;
1339 			goto restore_opts;
1340 		}
1341 		/*
1342 		 * Mounting a RDONLY partition read-write, so reread and
1343 		 * store the current valid flag.  (It may have been changed
1344 		 * by e2fsck since we originally mounted the partition.)
1345 		 */
1346 		sbi->s_mount_state = le16_to_cpu(es->s_state);
1347 		if (!ext2_setup_super (sb, es, 0))
1348 			sb->s_flags &= ~MS_RDONLY;
1349 		spin_unlock(&sbi->s_lock);
1350 
1351 		ext2_write_super(sb);
1352 
1353 		dquot_resume(sb, -1);
1354 	}
1355 
1356 	return 0;
1357 restore_opts:
1358 	sbi->s_mount_opt = old_opts.s_mount_opt;
1359 	sbi->s_resuid = old_opts.s_resuid;
1360 	sbi->s_resgid = old_opts.s_resgid;
1361 	sb->s_flags = old_sb_flags;
1362 	spin_unlock(&sbi->s_lock);
1363 	return err;
1364 }
1365 
1366 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1367 {
1368 	struct super_block *sb = dentry->d_sb;
1369 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1370 	struct ext2_super_block *es = sbi->s_es;
1371 	u64 fsid;
1372 
1373 	spin_lock(&sbi->s_lock);
1374 
1375 	if (test_opt (sb, MINIX_DF))
1376 		sbi->s_overhead_last = 0;
1377 	else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1378 		unsigned long i, overhead = 0;
1379 		smp_rmb();
1380 
1381 		/*
1382 		 * Compute the overhead (FS structures). This is constant
1383 		 * for a given filesystem unless the number of block groups
1384 		 * changes so we cache the previous value until it does.
1385 		 */
1386 
1387 		/*
1388 		 * All of the blocks before first_data_block are
1389 		 * overhead
1390 		 */
1391 		overhead = le32_to_cpu(es->s_first_data_block);
1392 
1393 		/*
1394 		 * Add the overhead attributed to the superblock and
1395 		 * block group descriptors.  If the sparse superblocks
1396 		 * feature is turned on, then not all groups have this.
1397 		 */
1398 		for (i = 0; i < sbi->s_groups_count; i++)
1399 			overhead += ext2_bg_has_super(sb, i) +
1400 				ext2_bg_num_gdb(sb, i);
1401 
1402 		/*
1403 		 * Every block group has an inode bitmap, a block
1404 		 * bitmap, and an inode table.
1405 		 */
1406 		overhead += (sbi->s_groups_count *
1407 			     (2 + sbi->s_itb_per_group));
1408 		sbi->s_overhead_last = overhead;
1409 		smp_wmb();
1410 		sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1411 	}
1412 
1413 	buf->f_type = EXT2_SUPER_MAGIC;
1414 	buf->f_bsize = sb->s_blocksize;
1415 	buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1416 	buf->f_bfree = ext2_count_free_blocks(sb);
1417 	es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1418 	buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1419 	if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1420 		buf->f_bavail = 0;
1421 	buf->f_files = le32_to_cpu(es->s_inodes_count);
1422 	buf->f_ffree = ext2_count_free_inodes(sb);
1423 	es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1424 	buf->f_namelen = EXT2_NAME_LEN;
1425 	fsid = le64_to_cpup((void *)es->s_uuid) ^
1426 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1427 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1428 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1429 	spin_unlock(&sbi->s_lock);
1430 	return 0;
1431 }
1432 
1433 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1434 	int flags, const char *dev_name, void *data)
1435 {
1436 	return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1437 }
1438 
1439 #ifdef CONFIG_QUOTA
1440 
1441 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1442  * acquiring the locks... As quota files are never truncated and quota code
1443  * itself serializes the operations (and no one else should touch the files)
1444  * we don't have to be afraid of races */
1445 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1446 			       size_t len, loff_t off)
1447 {
1448 	struct inode *inode = sb_dqopt(sb)->files[type];
1449 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1450 	int err = 0;
1451 	int offset = off & (sb->s_blocksize - 1);
1452 	int tocopy;
1453 	size_t toread;
1454 	struct buffer_head tmp_bh;
1455 	struct buffer_head *bh;
1456 	loff_t i_size = i_size_read(inode);
1457 
1458 	if (off > i_size)
1459 		return 0;
1460 	if (off+len > i_size)
1461 		len = i_size-off;
1462 	toread = len;
1463 	while (toread > 0) {
1464 		tocopy = sb->s_blocksize - offset < toread ?
1465 				sb->s_blocksize - offset : toread;
1466 
1467 		tmp_bh.b_state = 0;
1468 		tmp_bh.b_size = sb->s_blocksize;
1469 		err = ext2_get_block(inode, blk, &tmp_bh, 0);
1470 		if (err < 0)
1471 			return err;
1472 		if (!buffer_mapped(&tmp_bh))	/* A hole? */
1473 			memset(data, 0, tocopy);
1474 		else {
1475 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1476 			if (!bh)
1477 				return -EIO;
1478 			memcpy(data, bh->b_data+offset, tocopy);
1479 			brelse(bh);
1480 		}
1481 		offset = 0;
1482 		toread -= tocopy;
1483 		data += tocopy;
1484 		blk++;
1485 	}
1486 	return len;
1487 }
1488 
1489 /* Write to quotafile */
1490 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1491 				const char *data, size_t len, loff_t off)
1492 {
1493 	struct inode *inode = sb_dqopt(sb)->files[type];
1494 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1495 	int err = 0;
1496 	int offset = off & (sb->s_blocksize - 1);
1497 	int tocopy;
1498 	size_t towrite = len;
1499 	struct buffer_head tmp_bh;
1500 	struct buffer_head *bh;
1501 
1502 	while (towrite > 0) {
1503 		tocopy = sb->s_blocksize - offset < towrite ?
1504 				sb->s_blocksize - offset : towrite;
1505 
1506 		tmp_bh.b_state = 0;
1507 		tmp_bh.b_size = sb->s_blocksize;
1508 		err = ext2_get_block(inode, blk, &tmp_bh, 1);
1509 		if (err < 0)
1510 			goto out;
1511 		if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1512 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1513 		else
1514 			bh = sb_getblk(sb, tmp_bh.b_blocknr);
1515 		if (unlikely(!bh)) {
1516 			err = -EIO;
1517 			goto out;
1518 		}
1519 		lock_buffer(bh);
1520 		memcpy(bh->b_data+offset, data, tocopy);
1521 		flush_dcache_page(bh->b_page);
1522 		set_buffer_uptodate(bh);
1523 		mark_buffer_dirty(bh);
1524 		unlock_buffer(bh);
1525 		brelse(bh);
1526 		offset = 0;
1527 		towrite -= tocopy;
1528 		data += tocopy;
1529 		blk++;
1530 	}
1531 out:
1532 	if (len == towrite)
1533 		return err;
1534 	if (inode->i_size < off+len-towrite)
1535 		i_size_write(inode, off+len-towrite);
1536 	inode->i_version++;
1537 	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1538 	mark_inode_dirty(inode);
1539 	return len - towrite;
1540 }
1541 
1542 #endif
1543 
1544 static struct file_system_type ext2_fs_type = {
1545 	.owner		= THIS_MODULE,
1546 	.name		= "ext2",
1547 	.mount		= ext2_mount,
1548 	.kill_sb	= kill_block_super,
1549 	.fs_flags	= FS_REQUIRES_DEV,
1550 };
1551 MODULE_ALIAS_FS("ext2");
1552 
1553 static int __init init_ext2_fs(void)
1554 {
1555 	int err = init_ext2_xattr();
1556 	if (err)
1557 		return err;
1558 	err = init_inodecache();
1559 	if (err)
1560 		goto out1;
1561         err = register_filesystem(&ext2_fs_type);
1562 	if (err)
1563 		goto out;
1564 	return 0;
1565 out:
1566 	destroy_inodecache();
1567 out1:
1568 	exit_ext2_xattr();
1569 	return err;
1570 }
1571 
1572 static void __exit exit_ext2_fs(void)
1573 {
1574 	unregister_filesystem(&ext2_fs_type);
1575 	destroy_inodecache();
1576 	exit_ext2_xattr();
1577 }
1578 
1579 MODULE_AUTHOR("Remy Card and others");
1580 MODULE_DESCRIPTION("Second Extended Filesystem");
1581 MODULE_LICENSE("GPL");
1582 module_init(init_ext2_fs)
1583 module_exit(exit_ext2_fs)
1584