xref: /linux/fs/ext2/super.c (revision f2ee442115c9b6219083c019939a9cc0c9abb2f8)
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 
46 void ext2_error(struct super_block *sb, const char *function,
47 		const char *fmt, ...)
48 {
49 	struct va_format vaf;
50 	va_list args;
51 	struct ext2_sb_info *sbi = EXT2_SB(sb);
52 	struct ext2_super_block *es = sbi->s_es;
53 
54 	if (!(sb->s_flags & MS_RDONLY)) {
55 		spin_lock(&sbi->s_lock);
56 		sbi->s_mount_state |= EXT2_ERROR_FS;
57 		es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
58 		spin_unlock(&sbi->s_lock);
59 		ext2_sync_super(sb, es, 1);
60 	}
61 
62 	va_start(args, fmt);
63 
64 	vaf.fmt = fmt;
65 	vaf.va = &args;
66 
67 	printk(KERN_CRIT "EXT2-fs (%s): error: %s: %pV\n",
68 	       sb->s_id, function, &vaf);
69 
70 	va_end(args);
71 
72 	if (test_opt(sb, ERRORS_PANIC))
73 		panic("EXT2-fs: panic from previous error\n");
74 	if (test_opt(sb, ERRORS_RO)) {
75 		ext2_msg(sb, KERN_CRIT,
76 			     "error: remounting filesystem read-only");
77 		sb->s_flags |= MS_RDONLY;
78 	}
79 }
80 
81 void ext2_msg(struct super_block *sb, const char *prefix,
82 		const char *fmt, ...)
83 {
84 	struct va_format vaf;
85 	va_list args;
86 
87 	va_start(args, fmt);
88 
89 	vaf.fmt = fmt;
90 	vaf.va = &args;
91 
92 	printk("%sEXT2-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
93 
94 	va_end(args);
95 }
96 
97 /*
98  * This must be called with sbi->s_lock held.
99  */
100 void ext2_update_dynamic_rev(struct super_block *sb)
101 {
102 	struct ext2_super_block *es = EXT2_SB(sb)->s_es;
103 
104 	if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
105 		return;
106 
107 	ext2_msg(sb, KERN_WARNING,
108 		     "warning: updating to rev %d because of "
109 		     "new feature flag, running e2fsck is recommended",
110 		     EXT2_DYNAMIC_REV);
111 
112 	es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
113 	es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
114 	es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
115 	/* leave es->s_feature_*compat flags alone */
116 	/* es->s_uuid will be set by e2fsck if empty */
117 
118 	/*
119 	 * The rest of the superblock fields should be zero, and if not it
120 	 * means they are likely already in use, so leave them alone.  We
121 	 * can leave it up to e2fsck to clean up any inconsistencies there.
122 	 */
123 }
124 
125 static void ext2_put_super (struct super_block * sb)
126 {
127 	int db_count;
128 	int i;
129 	struct ext2_sb_info *sbi = EXT2_SB(sb);
130 
131 	dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
132 
133 	if (sb->s_dirt)
134 		ext2_write_super(sb);
135 
136 	ext2_xattr_put_super(sb);
137 	if (!(sb->s_flags & MS_RDONLY)) {
138 		struct ext2_super_block *es = sbi->s_es;
139 
140 		spin_lock(&sbi->s_lock);
141 		es->s_state = cpu_to_le16(sbi->s_mount_state);
142 		spin_unlock(&sbi->s_lock);
143 		ext2_sync_super(sb, es, 1);
144 	}
145 	db_count = sbi->s_gdb_count;
146 	for (i = 0; i < db_count; i++)
147 		if (sbi->s_group_desc[i])
148 			brelse (sbi->s_group_desc[i]);
149 	kfree(sbi->s_group_desc);
150 	kfree(sbi->s_debts);
151 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
152 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
153 	percpu_counter_destroy(&sbi->s_dirs_counter);
154 	brelse (sbi->s_sbh);
155 	sb->s_fs_info = NULL;
156 	kfree(sbi->s_blockgroup_lock);
157 	kfree(sbi);
158 }
159 
160 static struct kmem_cache * ext2_inode_cachep;
161 
162 static struct inode *ext2_alloc_inode(struct super_block *sb)
163 {
164 	struct ext2_inode_info *ei;
165 	ei = (struct ext2_inode_info *)kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
166 	if (!ei)
167 		return NULL;
168 	ei->i_block_alloc_info = NULL;
169 	ei->vfs_inode.i_version = 1;
170 	return &ei->vfs_inode;
171 }
172 
173 static void ext2_i_callback(struct rcu_head *head)
174 {
175 	struct inode *inode = container_of(head, struct inode, i_rcu);
176 	INIT_LIST_HEAD(&inode->i_dentry);
177 	kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
178 }
179 
180 static void ext2_destroy_inode(struct inode *inode)
181 {
182 	call_rcu(&inode->i_rcu, ext2_i_callback);
183 }
184 
185 static void init_once(void *foo)
186 {
187 	struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
188 
189 	rwlock_init(&ei->i_meta_lock);
190 #ifdef CONFIG_EXT2_FS_XATTR
191 	init_rwsem(&ei->xattr_sem);
192 #endif
193 	mutex_init(&ei->truncate_mutex);
194 	inode_init_once(&ei->vfs_inode);
195 }
196 
197 static int init_inodecache(void)
198 {
199 	ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
200 					     sizeof(struct ext2_inode_info),
201 					     0, (SLAB_RECLAIM_ACCOUNT|
202 						SLAB_MEM_SPREAD),
203 					     init_once);
204 	if (ext2_inode_cachep == NULL)
205 		return -ENOMEM;
206 	return 0;
207 }
208 
209 static void destroy_inodecache(void)
210 {
211 	kmem_cache_destroy(ext2_inode_cachep);
212 }
213 
214 static int ext2_show_options(struct seq_file *seq, struct vfsmount *vfs)
215 {
216 	struct super_block *sb = vfs->mnt_sb;
217 	struct ext2_sb_info *sbi = EXT2_SB(sb);
218 	struct ext2_super_block *es = sbi->s_es;
219 	unsigned long def_mount_opts;
220 
221 	spin_lock(&sbi->s_lock);
222 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
223 
224 	if (sbi->s_sb_block != 1)
225 		seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
226 	if (test_opt(sb, MINIX_DF))
227 		seq_puts(seq, ",minixdf");
228 	if (test_opt(sb, GRPID))
229 		seq_puts(seq, ",grpid");
230 	if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
231 		seq_puts(seq, ",nogrpid");
232 	if (sbi->s_resuid != EXT2_DEF_RESUID ||
233 	    le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
234 		seq_printf(seq, ",resuid=%u", sbi->s_resuid);
235 	}
236 	if (sbi->s_resgid != EXT2_DEF_RESGID ||
237 	    le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
238 		seq_printf(seq, ",resgid=%u", sbi->s_resgid);
239 	}
240 	if (test_opt(sb, ERRORS_RO)) {
241 		int def_errors = le16_to_cpu(es->s_errors);
242 
243 		if (def_errors == EXT2_ERRORS_PANIC ||
244 		    def_errors == EXT2_ERRORS_CONTINUE) {
245 			seq_puts(seq, ",errors=remount-ro");
246 		}
247 	}
248 	if (test_opt(sb, ERRORS_CONT))
249 		seq_puts(seq, ",errors=continue");
250 	if (test_opt(sb, ERRORS_PANIC))
251 		seq_puts(seq, ",errors=panic");
252 	if (test_opt(sb, NO_UID32))
253 		seq_puts(seq, ",nouid32");
254 	if (test_opt(sb, DEBUG))
255 		seq_puts(seq, ",debug");
256 	if (test_opt(sb, OLDALLOC))
257 		seq_puts(seq, ",oldalloc");
258 
259 #ifdef CONFIG_EXT2_FS_XATTR
260 	if (test_opt(sb, XATTR_USER))
261 		seq_puts(seq, ",user_xattr");
262 	if (!test_opt(sb, XATTR_USER) &&
263 	    (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
264 		seq_puts(seq, ",nouser_xattr");
265 	}
266 #endif
267 
268 #ifdef CONFIG_EXT2_FS_POSIX_ACL
269 	if (test_opt(sb, POSIX_ACL))
270 		seq_puts(seq, ",acl");
271 	if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
272 		seq_puts(seq, ",noacl");
273 #endif
274 
275 	if (test_opt(sb, NOBH))
276 		seq_puts(seq, ",nobh");
277 
278 #if defined(CONFIG_QUOTA)
279 	if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
280 		seq_puts(seq, ",usrquota");
281 
282 	if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
283 		seq_puts(seq, ",grpquota");
284 #endif
285 
286 #if defined(CONFIG_EXT2_FS_XIP)
287 	if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
288 		seq_puts(seq, ",xip");
289 #endif
290 
291 	if (!test_opt(sb, RESERVATION))
292 		seq_puts(seq, ",noreservation");
293 
294 	spin_unlock(&sbi->s_lock);
295 	return 0;
296 }
297 
298 #ifdef CONFIG_QUOTA
299 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
300 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
301 #endif
302 
303 static const struct super_operations ext2_sops = {
304 	.alloc_inode	= ext2_alloc_inode,
305 	.destroy_inode	= ext2_destroy_inode,
306 	.write_inode	= ext2_write_inode,
307 	.evict_inode	= ext2_evict_inode,
308 	.put_super	= ext2_put_super,
309 	.write_super	= ext2_write_super,
310 	.sync_fs	= ext2_sync_fs,
311 	.statfs		= ext2_statfs,
312 	.remount_fs	= ext2_remount,
313 	.show_options	= ext2_show_options,
314 #ifdef CONFIG_QUOTA
315 	.quota_read	= ext2_quota_read,
316 	.quota_write	= ext2_quota_write,
317 #endif
318 };
319 
320 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
321 		u64 ino, u32 generation)
322 {
323 	struct inode *inode;
324 
325 	if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
326 		return ERR_PTR(-ESTALE);
327 	if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
328 		return ERR_PTR(-ESTALE);
329 
330 	/*
331 	 * ext2_iget isn't quite right if the inode is currently unallocated!
332 	 * However ext2_iget currently does appropriate checks to handle stale
333 	 * inodes so everything is OK.
334 	 */
335 	inode = ext2_iget(sb, ino);
336 	if (IS_ERR(inode))
337 		return ERR_CAST(inode);
338 	if (generation && inode->i_generation != generation) {
339 		/* we didn't find the right inode.. */
340 		iput(inode);
341 		return ERR_PTR(-ESTALE);
342 	}
343 	return inode;
344 }
345 
346 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
347 		int fh_len, int fh_type)
348 {
349 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
350 				    ext2_nfs_get_inode);
351 }
352 
353 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
354 		int fh_len, int fh_type)
355 {
356 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
357 				    ext2_nfs_get_inode);
358 }
359 
360 /* Yes, most of these are left as NULL!!
361  * A NULL value implies the default, which works with ext2-like file
362  * systems, but can be improved upon.
363  * Currently only get_parent is required.
364  */
365 static const struct export_operations ext2_export_ops = {
366 	.fh_to_dentry = ext2_fh_to_dentry,
367 	.fh_to_parent = ext2_fh_to_parent,
368 	.get_parent = ext2_get_parent,
369 };
370 
371 static unsigned long get_sb_block(void **data)
372 {
373 	unsigned long 	sb_block;
374 	char 		*options = (char *) *data;
375 
376 	if (!options || strncmp(options, "sb=", 3) != 0)
377 		return 1;	/* Default location */
378 	options += 3;
379 	sb_block = simple_strtoul(options, &options, 0);
380 	if (*options && *options != ',') {
381 		printk("EXT2-fs: Invalid sb specification: %s\n",
382 		       (char *) *data);
383 		return 1;
384 	}
385 	if (*options == ',')
386 		options++;
387 	*data = (void *) options;
388 	return sb_block;
389 }
390 
391 enum {
392 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
393 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
394 	Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
395 	Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
396 	Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
397 	Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
398 };
399 
400 static const match_table_t tokens = {
401 	{Opt_bsd_df, "bsddf"},
402 	{Opt_minix_df, "minixdf"},
403 	{Opt_grpid, "grpid"},
404 	{Opt_grpid, "bsdgroups"},
405 	{Opt_nogrpid, "nogrpid"},
406 	{Opt_nogrpid, "sysvgroups"},
407 	{Opt_resgid, "resgid=%u"},
408 	{Opt_resuid, "resuid=%u"},
409 	{Opt_sb, "sb=%u"},
410 	{Opt_err_cont, "errors=continue"},
411 	{Opt_err_panic, "errors=panic"},
412 	{Opt_err_ro, "errors=remount-ro"},
413 	{Opt_nouid32, "nouid32"},
414 	{Opt_nocheck, "check=none"},
415 	{Opt_nocheck, "nocheck"},
416 	{Opt_debug, "debug"},
417 	{Opt_oldalloc, "oldalloc"},
418 	{Opt_orlov, "orlov"},
419 	{Opt_nobh, "nobh"},
420 	{Opt_user_xattr, "user_xattr"},
421 	{Opt_nouser_xattr, "nouser_xattr"},
422 	{Opt_acl, "acl"},
423 	{Opt_noacl, "noacl"},
424 	{Opt_xip, "xip"},
425 	{Opt_grpquota, "grpquota"},
426 	{Opt_ignore, "noquota"},
427 	{Opt_quota, "quota"},
428 	{Opt_usrquota, "usrquota"},
429 	{Opt_reservation, "reservation"},
430 	{Opt_noreservation, "noreservation"},
431 	{Opt_err, NULL}
432 };
433 
434 static int parse_options(char *options, struct super_block *sb)
435 {
436 	char *p;
437 	struct ext2_sb_info *sbi = EXT2_SB(sb);
438 	substring_t args[MAX_OPT_ARGS];
439 	int option;
440 
441 	if (!options)
442 		return 1;
443 
444 	while ((p = strsep (&options, ",")) != NULL) {
445 		int token;
446 		if (!*p)
447 			continue;
448 
449 		token = match_token(p, tokens, args);
450 		switch (token) {
451 		case Opt_bsd_df:
452 			clear_opt (sbi->s_mount_opt, MINIX_DF);
453 			break;
454 		case Opt_minix_df:
455 			set_opt (sbi->s_mount_opt, MINIX_DF);
456 			break;
457 		case Opt_grpid:
458 			set_opt (sbi->s_mount_opt, GRPID);
459 			break;
460 		case Opt_nogrpid:
461 			clear_opt (sbi->s_mount_opt, GRPID);
462 			break;
463 		case Opt_resuid:
464 			if (match_int(&args[0], &option))
465 				return 0;
466 			sbi->s_resuid = option;
467 			break;
468 		case Opt_resgid:
469 			if (match_int(&args[0], &option))
470 				return 0;
471 			sbi->s_resgid = option;
472 			break;
473 		case Opt_sb:
474 			/* handled by get_sb_block() instead of here */
475 			/* *sb_block = match_int(&args[0]); */
476 			break;
477 		case Opt_err_panic:
478 			clear_opt (sbi->s_mount_opt, ERRORS_CONT);
479 			clear_opt (sbi->s_mount_opt, ERRORS_RO);
480 			set_opt (sbi->s_mount_opt, ERRORS_PANIC);
481 			break;
482 		case Opt_err_ro:
483 			clear_opt (sbi->s_mount_opt, ERRORS_CONT);
484 			clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
485 			set_opt (sbi->s_mount_opt, ERRORS_RO);
486 			break;
487 		case Opt_err_cont:
488 			clear_opt (sbi->s_mount_opt, ERRORS_RO);
489 			clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
490 			set_opt (sbi->s_mount_opt, ERRORS_CONT);
491 			break;
492 		case Opt_nouid32:
493 			set_opt (sbi->s_mount_opt, NO_UID32);
494 			break;
495 		case Opt_nocheck:
496 			clear_opt (sbi->s_mount_opt, CHECK);
497 			break;
498 		case Opt_debug:
499 			set_opt (sbi->s_mount_opt, DEBUG);
500 			break;
501 		case Opt_oldalloc:
502 			set_opt (sbi->s_mount_opt, OLDALLOC);
503 			break;
504 		case Opt_orlov:
505 			clear_opt (sbi->s_mount_opt, OLDALLOC);
506 			break;
507 		case Opt_nobh:
508 			set_opt (sbi->s_mount_opt, NOBH);
509 			break;
510 #ifdef CONFIG_EXT2_FS_XATTR
511 		case Opt_user_xattr:
512 			set_opt (sbi->s_mount_opt, XATTR_USER);
513 			break;
514 		case Opt_nouser_xattr:
515 			clear_opt (sbi->s_mount_opt, XATTR_USER);
516 			break;
517 #else
518 		case Opt_user_xattr:
519 		case Opt_nouser_xattr:
520 			ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
521 				"not supported");
522 			break;
523 #endif
524 #ifdef CONFIG_EXT2_FS_POSIX_ACL
525 		case Opt_acl:
526 			set_opt(sbi->s_mount_opt, POSIX_ACL);
527 			break;
528 		case Opt_noacl:
529 			clear_opt(sbi->s_mount_opt, POSIX_ACL);
530 			break;
531 #else
532 		case Opt_acl:
533 		case Opt_noacl:
534 			ext2_msg(sb, KERN_INFO,
535 				"(no)acl options not supported");
536 			break;
537 #endif
538 		case Opt_xip:
539 #ifdef CONFIG_EXT2_FS_XIP
540 			set_opt (sbi->s_mount_opt, XIP);
541 #else
542 			ext2_msg(sb, KERN_INFO, "xip option not supported");
543 #endif
544 			break;
545 
546 #if defined(CONFIG_QUOTA)
547 		case Opt_quota:
548 		case Opt_usrquota:
549 			set_opt(sbi->s_mount_opt, USRQUOTA);
550 			break;
551 
552 		case Opt_grpquota:
553 			set_opt(sbi->s_mount_opt, GRPQUOTA);
554 			break;
555 #else
556 		case Opt_quota:
557 		case Opt_usrquota:
558 		case Opt_grpquota:
559 			ext2_msg(sb, KERN_INFO,
560 				"quota operations not supported");
561 			break;
562 #endif
563 
564 		case Opt_reservation:
565 			set_opt(sbi->s_mount_opt, RESERVATION);
566 			ext2_msg(sb, KERN_INFO, "reservations ON");
567 			break;
568 		case Opt_noreservation:
569 			clear_opt(sbi->s_mount_opt, RESERVATION);
570 			ext2_msg(sb, KERN_INFO, "reservations OFF");
571 			break;
572 		case Opt_ignore:
573 			break;
574 		default:
575 			return 0;
576 		}
577 	}
578 	return 1;
579 }
580 
581 static int ext2_setup_super (struct super_block * sb,
582 			      struct ext2_super_block * es,
583 			      int read_only)
584 {
585 	int res = 0;
586 	struct ext2_sb_info *sbi = EXT2_SB(sb);
587 
588 	if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
589 		ext2_msg(sb, KERN_ERR,
590 			"error: revision level too high, "
591 			"forcing read-only mode");
592 		res = MS_RDONLY;
593 	}
594 	if (read_only)
595 		return res;
596 	if (!(sbi->s_mount_state & EXT2_VALID_FS))
597 		ext2_msg(sb, KERN_WARNING,
598 			"warning: mounting unchecked fs, "
599 			"running e2fsck is recommended");
600 	else if ((sbi->s_mount_state & EXT2_ERROR_FS))
601 		ext2_msg(sb, KERN_WARNING,
602 			"warning: mounting fs with errors, "
603 			"running e2fsck is recommended");
604 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
605 		 le16_to_cpu(es->s_mnt_count) >=
606 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
607 		ext2_msg(sb, KERN_WARNING,
608 			"warning: maximal mount count reached, "
609 			"running e2fsck is recommended");
610 	else if (le32_to_cpu(es->s_checkinterval) &&
611 		(le32_to_cpu(es->s_lastcheck) +
612 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
613 		ext2_msg(sb, KERN_WARNING,
614 			"warning: checktime reached, "
615 			"running e2fsck is recommended");
616 	if (!le16_to_cpu(es->s_max_mnt_count))
617 		es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
618 	le16_add_cpu(&es->s_mnt_count, 1);
619 	if (test_opt (sb, DEBUG))
620 		ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
621 			"bpg=%lu, ipg=%lu, mo=%04lx]",
622 			EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
623 			sbi->s_frag_size,
624 			sbi->s_groups_count,
625 			EXT2_BLOCKS_PER_GROUP(sb),
626 			EXT2_INODES_PER_GROUP(sb),
627 			sbi->s_mount_opt);
628 	return res;
629 }
630 
631 static int ext2_check_descriptors(struct super_block *sb)
632 {
633 	int i;
634 	struct ext2_sb_info *sbi = EXT2_SB(sb);
635 
636 	ext2_debug ("Checking group descriptors");
637 
638 	for (i = 0; i < sbi->s_groups_count; i++) {
639 		struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
640 		ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
641 		ext2_fsblk_t last_block;
642 
643 		if (i == sbi->s_groups_count - 1)
644 			last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
645 		else
646 			last_block = first_block +
647 				(EXT2_BLOCKS_PER_GROUP(sb) - 1);
648 
649 		if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
650 		    le32_to_cpu(gdp->bg_block_bitmap) > last_block)
651 		{
652 			ext2_error (sb, "ext2_check_descriptors",
653 				    "Block bitmap for group %d"
654 				    " not in group (block %lu)!",
655 				    i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
656 			return 0;
657 		}
658 		if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
659 		    le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
660 		{
661 			ext2_error (sb, "ext2_check_descriptors",
662 				    "Inode bitmap for group %d"
663 				    " not in group (block %lu)!",
664 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
665 			return 0;
666 		}
667 		if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
668 		    le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
669 		    last_block)
670 		{
671 			ext2_error (sb, "ext2_check_descriptors",
672 				    "Inode table for group %d"
673 				    " not in group (block %lu)!",
674 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
675 			return 0;
676 		}
677 	}
678 	return 1;
679 }
680 
681 /*
682  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
683  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
684  * We need to be 1 filesystem block less than the 2^32 sector limit.
685  */
686 static loff_t ext2_max_size(int bits)
687 {
688 	loff_t res = EXT2_NDIR_BLOCKS;
689 	int meta_blocks;
690 	loff_t upper_limit;
691 
692 	/* This is calculated to be the largest file size for a
693 	 * dense, file such that the total number of
694 	 * sectors in the file, including data and all indirect blocks,
695 	 * does not exceed 2^32 -1
696 	 * __u32 i_blocks representing the total number of
697 	 * 512 bytes blocks of the file
698 	 */
699 	upper_limit = (1LL << 32) - 1;
700 
701 	/* total blocks in file system block size */
702 	upper_limit >>= (bits - 9);
703 
704 
705 	/* indirect blocks */
706 	meta_blocks = 1;
707 	/* double indirect blocks */
708 	meta_blocks += 1 + (1LL << (bits-2));
709 	/* tripple indirect blocks */
710 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
711 
712 	upper_limit -= meta_blocks;
713 	upper_limit <<= bits;
714 
715 	res += 1LL << (bits-2);
716 	res += 1LL << (2*(bits-2));
717 	res += 1LL << (3*(bits-2));
718 	res <<= bits;
719 	if (res > upper_limit)
720 		res = upper_limit;
721 
722 	if (res > MAX_LFS_FILESIZE)
723 		res = MAX_LFS_FILESIZE;
724 
725 	return res;
726 }
727 
728 static unsigned long descriptor_loc(struct super_block *sb,
729 				    unsigned long logic_sb_block,
730 				    int nr)
731 {
732 	struct ext2_sb_info *sbi = EXT2_SB(sb);
733 	unsigned long bg, first_meta_bg;
734 	int has_super = 0;
735 
736 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
737 
738 	if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
739 	    nr < first_meta_bg)
740 		return (logic_sb_block + nr + 1);
741 	bg = sbi->s_desc_per_block * nr;
742 	if (ext2_bg_has_super(sb, bg))
743 		has_super = 1;
744 
745 	return ext2_group_first_block_no(sb, bg) + has_super;
746 }
747 
748 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
749 {
750 	struct buffer_head * bh;
751 	struct ext2_sb_info * sbi;
752 	struct ext2_super_block * es;
753 	struct inode *root;
754 	unsigned long block;
755 	unsigned long sb_block = get_sb_block(&data);
756 	unsigned long logic_sb_block;
757 	unsigned long offset = 0;
758 	unsigned long def_mount_opts;
759 	long ret = -EINVAL;
760 	int blocksize = BLOCK_SIZE;
761 	int db_count;
762 	int i, j;
763 	__le32 features;
764 	int err;
765 
766 	err = -ENOMEM;
767 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
768 	if (!sbi)
769 		goto failed_unlock;
770 
771 	sbi->s_blockgroup_lock =
772 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
773 	if (!sbi->s_blockgroup_lock) {
774 		kfree(sbi);
775 		goto failed_unlock;
776 	}
777 	sb->s_fs_info = sbi;
778 	sbi->s_sb_block = sb_block;
779 
780 	spin_lock_init(&sbi->s_lock);
781 
782 	/*
783 	 * See what the current blocksize for the device is, and
784 	 * use that as the blocksize.  Otherwise (or if the blocksize
785 	 * is smaller than the default) use the default.
786 	 * This is important for devices that have a hardware
787 	 * sectorsize that is larger than the default.
788 	 */
789 	blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
790 	if (!blocksize) {
791 		ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
792 		goto failed_sbi;
793 	}
794 
795 	/*
796 	 * If the superblock doesn't start on a hardware sector boundary,
797 	 * calculate the offset.
798 	 */
799 	if (blocksize != BLOCK_SIZE) {
800 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
801 		offset = (sb_block*BLOCK_SIZE) % blocksize;
802 	} else {
803 		logic_sb_block = sb_block;
804 	}
805 
806 	if (!(bh = sb_bread(sb, logic_sb_block))) {
807 		ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
808 		goto failed_sbi;
809 	}
810 	/*
811 	 * Note: s_es must be initialized as soon as possible because
812 	 *       some ext2 macro-instructions depend on its value
813 	 */
814 	es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
815 	sbi->s_es = es;
816 	sb->s_magic = le16_to_cpu(es->s_magic);
817 
818 	if (sb->s_magic != EXT2_SUPER_MAGIC)
819 		goto cantfind_ext2;
820 
821 	/* Set defaults before we parse the mount options */
822 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
823 	if (def_mount_opts & EXT2_DEFM_DEBUG)
824 		set_opt(sbi->s_mount_opt, DEBUG);
825 	if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
826 		set_opt(sbi->s_mount_opt, GRPID);
827 	if (def_mount_opts & EXT2_DEFM_UID16)
828 		set_opt(sbi->s_mount_opt, NO_UID32);
829 #ifdef CONFIG_EXT2_FS_XATTR
830 	if (def_mount_opts & EXT2_DEFM_XATTR_USER)
831 		set_opt(sbi->s_mount_opt, XATTR_USER);
832 #endif
833 #ifdef CONFIG_EXT2_FS_POSIX_ACL
834 	if (def_mount_opts & EXT2_DEFM_ACL)
835 		set_opt(sbi->s_mount_opt, POSIX_ACL);
836 #endif
837 
838 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
839 		set_opt(sbi->s_mount_opt, ERRORS_PANIC);
840 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
841 		set_opt(sbi->s_mount_opt, ERRORS_CONT);
842 	else
843 		set_opt(sbi->s_mount_opt, ERRORS_RO);
844 
845 	sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
846 	sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
847 
848 	set_opt(sbi->s_mount_opt, RESERVATION);
849 
850 	if (!parse_options((char *) data, sb))
851 		goto failed_mount;
852 
853 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
854 		((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
855 		 MS_POSIXACL : 0);
856 
857 	ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
858 				    EXT2_MOUNT_XIP if not */
859 
860 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
861 	    (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
862 	     EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
863 	     EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
864 		ext2_msg(sb, KERN_WARNING,
865 			"warning: feature flags set on rev 0 fs, "
866 			"running e2fsck is recommended");
867 	/*
868 	 * Check feature flags regardless of the revision level, since we
869 	 * previously didn't change the revision level when setting the flags,
870 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
871 	 */
872 	features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
873 	if (features) {
874 		ext2_msg(sb, KERN_ERR,	"error: couldn't mount because of "
875 		       "unsupported optional features (%x)",
876 			le32_to_cpu(features));
877 		goto failed_mount;
878 	}
879 	if (!(sb->s_flags & MS_RDONLY) &&
880 	    (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
881 		ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
882 		       "unsupported optional features (%x)",
883 		       le32_to_cpu(features));
884 		goto failed_mount;
885 	}
886 
887 	blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
888 
889 	if (ext2_use_xip(sb) && blocksize != PAGE_SIZE) {
890 		if (!silent)
891 			ext2_msg(sb, KERN_ERR,
892 				"error: unsupported blocksize for xip");
893 		goto failed_mount;
894 	}
895 
896 	/* If the blocksize doesn't match, re-read the thing.. */
897 	if (sb->s_blocksize != blocksize) {
898 		brelse(bh);
899 
900 		if (!sb_set_blocksize(sb, blocksize)) {
901 			ext2_msg(sb, KERN_ERR,
902 				"error: bad blocksize %d", blocksize);
903 			goto failed_sbi;
904 		}
905 
906 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
907 		offset = (sb_block*BLOCK_SIZE) % blocksize;
908 		bh = sb_bread(sb, logic_sb_block);
909 		if(!bh) {
910 			ext2_msg(sb, KERN_ERR, "error: couldn't read"
911 				"superblock on 2nd try");
912 			goto failed_sbi;
913 		}
914 		es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
915 		sbi->s_es = es;
916 		if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
917 			ext2_msg(sb, KERN_ERR, "error: magic mismatch");
918 			goto failed_mount;
919 		}
920 	}
921 
922 	sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
923 
924 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
925 		sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
926 		sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
927 	} else {
928 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
929 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
930 		if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
931 		    !is_power_of_2(sbi->s_inode_size) ||
932 		    (sbi->s_inode_size > blocksize)) {
933 			ext2_msg(sb, KERN_ERR,
934 				"error: unsupported inode size: %d",
935 				sbi->s_inode_size);
936 			goto failed_mount;
937 		}
938 	}
939 
940 	sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
941 				   le32_to_cpu(es->s_log_frag_size);
942 	if (sbi->s_frag_size == 0)
943 		goto cantfind_ext2;
944 	sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
945 
946 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
947 	sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
948 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
949 
950 	if (EXT2_INODE_SIZE(sb) == 0)
951 		goto cantfind_ext2;
952 	sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
953 	if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
954 		goto cantfind_ext2;
955 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
956 					sbi->s_inodes_per_block;
957 	sbi->s_desc_per_block = sb->s_blocksize /
958 					sizeof (struct ext2_group_desc);
959 	sbi->s_sbh = bh;
960 	sbi->s_mount_state = le16_to_cpu(es->s_state);
961 	sbi->s_addr_per_block_bits =
962 		ilog2 (EXT2_ADDR_PER_BLOCK(sb));
963 	sbi->s_desc_per_block_bits =
964 		ilog2 (EXT2_DESC_PER_BLOCK(sb));
965 
966 	if (sb->s_magic != EXT2_SUPER_MAGIC)
967 		goto cantfind_ext2;
968 
969 	if (sb->s_blocksize != bh->b_size) {
970 		if (!silent)
971 			ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
972 		goto failed_mount;
973 	}
974 
975 	if (sb->s_blocksize != sbi->s_frag_size) {
976 		ext2_msg(sb, KERN_ERR,
977 			"error: fragsize %lu != blocksize %lu"
978 			"(not supported yet)",
979 			sbi->s_frag_size, sb->s_blocksize);
980 		goto failed_mount;
981 	}
982 
983 	if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
984 		ext2_msg(sb, KERN_ERR,
985 			"error: #blocks per group too big: %lu",
986 			sbi->s_blocks_per_group);
987 		goto failed_mount;
988 	}
989 	if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
990 		ext2_msg(sb, KERN_ERR,
991 			"error: #fragments per group too big: %lu",
992 			sbi->s_frags_per_group);
993 		goto failed_mount;
994 	}
995 	if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
996 		ext2_msg(sb, KERN_ERR,
997 			"error: #inodes per group too big: %lu",
998 			sbi->s_inodes_per_group);
999 		goto failed_mount;
1000 	}
1001 
1002 	if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1003 		goto cantfind_ext2;
1004  	sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1005  				le32_to_cpu(es->s_first_data_block) - 1)
1006  					/ EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1007 	db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1008 		   EXT2_DESC_PER_BLOCK(sb);
1009 	sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
1010 	if (sbi->s_group_desc == NULL) {
1011 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1012 		goto failed_mount;
1013 	}
1014 	bgl_lock_init(sbi->s_blockgroup_lock);
1015 	sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1016 	if (!sbi->s_debts) {
1017 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1018 		goto failed_mount_group_desc;
1019 	}
1020 	for (i = 0; i < db_count; i++) {
1021 		block = descriptor_loc(sb, logic_sb_block, i);
1022 		sbi->s_group_desc[i] = sb_bread(sb, block);
1023 		if (!sbi->s_group_desc[i]) {
1024 			for (j = 0; j < i; j++)
1025 				brelse (sbi->s_group_desc[j]);
1026 			ext2_msg(sb, KERN_ERR,
1027 				"error: unable to read group descriptors");
1028 			goto failed_mount_group_desc;
1029 		}
1030 	}
1031 	if (!ext2_check_descriptors (sb)) {
1032 		ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1033 		goto failed_mount2;
1034 	}
1035 	sbi->s_gdb_count = db_count;
1036 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1037 	spin_lock_init(&sbi->s_next_gen_lock);
1038 
1039 	/* per fileystem reservation list head & lock */
1040 	spin_lock_init(&sbi->s_rsv_window_lock);
1041 	sbi->s_rsv_window_root = RB_ROOT;
1042 	/*
1043 	 * Add a single, static dummy reservation to the start of the
1044 	 * reservation window list --- it gives us a placeholder for
1045 	 * append-at-start-of-list which makes the allocation logic
1046 	 * _much_ simpler.
1047 	 */
1048 	sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1049 	sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1050 	sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1051 	sbi->s_rsv_window_head.rsv_goal_size = 0;
1052 	ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1053 
1054 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
1055 				ext2_count_free_blocks(sb));
1056 	if (!err) {
1057 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
1058 				ext2_count_free_inodes(sb));
1059 	}
1060 	if (!err) {
1061 		err = percpu_counter_init(&sbi->s_dirs_counter,
1062 				ext2_count_dirs(sb));
1063 	}
1064 	if (err) {
1065 		ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1066 		goto failed_mount3;
1067 	}
1068 	/*
1069 	 * set up enough so that it can read an inode
1070 	 */
1071 	sb->s_op = &ext2_sops;
1072 	sb->s_export_op = &ext2_export_ops;
1073 	sb->s_xattr = ext2_xattr_handlers;
1074 
1075 #ifdef CONFIG_QUOTA
1076 	sb->dq_op = &dquot_operations;
1077 	sb->s_qcop = &dquot_quotactl_ops;
1078 #endif
1079 
1080 	root = ext2_iget(sb, EXT2_ROOT_INO);
1081 	if (IS_ERR(root)) {
1082 		ret = PTR_ERR(root);
1083 		goto failed_mount3;
1084 	}
1085 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1086 		iput(root);
1087 		ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1088 		goto failed_mount3;
1089 	}
1090 
1091 	sb->s_root = d_alloc_root(root);
1092 	if (!sb->s_root) {
1093 		iput(root);
1094 		ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1095 		ret = -ENOMEM;
1096 		goto failed_mount3;
1097 	}
1098 	if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1099 		ext2_msg(sb, KERN_WARNING,
1100 			"warning: mounting ext3 filesystem as ext2");
1101 	if (ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY))
1102 		sb->s_flags |= MS_RDONLY;
1103 	ext2_write_super(sb);
1104 	return 0;
1105 
1106 cantfind_ext2:
1107 	if (!silent)
1108 		ext2_msg(sb, KERN_ERR,
1109 			"error: can't find an ext2 filesystem on dev %s.",
1110 			sb->s_id);
1111 	goto failed_mount;
1112 failed_mount3:
1113 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
1114 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
1115 	percpu_counter_destroy(&sbi->s_dirs_counter);
1116 failed_mount2:
1117 	for (i = 0; i < db_count; i++)
1118 		brelse(sbi->s_group_desc[i]);
1119 failed_mount_group_desc:
1120 	kfree(sbi->s_group_desc);
1121 	kfree(sbi->s_debts);
1122 failed_mount:
1123 	brelse(bh);
1124 failed_sbi:
1125 	sb->s_fs_info = NULL;
1126 	kfree(sbi->s_blockgroup_lock);
1127 	kfree(sbi);
1128 failed_unlock:
1129 	return ret;
1130 }
1131 
1132 static void ext2_clear_super_error(struct super_block *sb)
1133 {
1134 	struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1135 
1136 	if (buffer_write_io_error(sbh)) {
1137 		/*
1138 		 * Oh, dear.  A previous attempt to write the
1139 		 * superblock failed.  This could happen because the
1140 		 * USB device was yanked out.  Or it could happen to
1141 		 * be a transient write error and maybe the block will
1142 		 * be remapped.  Nothing we can do but to retry the
1143 		 * write and hope for the best.
1144 		 */
1145 		ext2_msg(sb, KERN_ERR,
1146 		       "previous I/O error to superblock detected\n");
1147 		clear_buffer_write_io_error(sbh);
1148 		set_buffer_uptodate(sbh);
1149 	}
1150 }
1151 
1152 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1153 			    int wait)
1154 {
1155 	ext2_clear_super_error(sb);
1156 	spin_lock(&EXT2_SB(sb)->s_lock);
1157 	es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1158 	es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1159 	es->s_wtime = cpu_to_le32(get_seconds());
1160 	/* unlock before we do IO */
1161 	spin_unlock(&EXT2_SB(sb)->s_lock);
1162 	mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1163 	if (wait)
1164 		sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1165 	sb->s_dirt = 0;
1166 }
1167 
1168 /*
1169  * In the second extended file system, it is not necessary to
1170  * write the super block since we use a mapping of the
1171  * disk super block in a buffer.
1172  *
1173  * However, this function is still used to set the fs valid
1174  * flags to 0.  We need to set this flag to 0 since the fs
1175  * may have been checked while mounted and e2fsck may have
1176  * set s_state to EXT2_VALID_FS after some corrections.
1177  */
1178 static int ext2_sync_fs(struct super_block *sb, int wait)
1179 {
1180 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1181 	struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1182 
1183 	spin_lock(&sbi->s_lock);
1184 	if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1185 		ext2_debug("setting valid to 0\n");
1186 		es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1187 	}
1188 	spin_unlock(&sbi->s_lock);
1189 	ext2_sync_super(sb, es, wait);
1190 	return 0;
1191 }
1192 
1193 
1194 void ext2_write_super(struct super_block *sb)
1195 {
1196 	if (!(sb->s_flags & MS_RDONLY))
1197 		ext2_sync_fs(sb, 1);
1198 	else
1199 		sb->s_dirt = 0;
1200 }
1201 
1202 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1203 {
1204 	struct ext2_sb_info * sbi = EXT2_SB(sb);
1205 	struct ext2_super_block * es;
1206 	unsigned long old_mount_opt = sbi->s_mount_opt;
1207 	struct ext2_mount_options old_opts;
1208 	unsigned long old_sb_flags;
1209 	int err;
1210 
1211 	spin_lock(&sbi->s_lock);
1212 
1213 	/* Store the old options */
1214 	old_sb_flags = sb->s_flags;
1215 	old_opts.s_mount_opt = sbi->s_mount_opt;
1216 	old_opts.s_resuid = sbi->s_resuid;
1217 	old_opts.s_resgid = sbi->s_resgid;
1218 
1219 	/*
1220 	 * Allow the "check" option to be passed as a remount option.
1221 	 */
1222 	if (!parse_options(data, sb)) {
1223 		err = -EINVAL;
1224 		goto restore_opts;
1225 	}
1226 
1227 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1228 		((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1229 
1230 	ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1231 				    EXT2_MOUNT_XIP if not */
1232 
1233 	if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
1234 		ext2_msg(sb, KERN_WARNING,
1235 			"warning: unsupported blocksize for xip");
1236 		err = -EINVAL;
1237 		goto restore_opts;
1238 	}
1239 
1240 	es = sbi->s_es;
1241 	if ((sbi->s_mount_opt ^ old_mount_opt) & EXT2_MOUNT_XIP) {
1242 		ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1243 			 "xip flag with busy inodes while remounting");
1244 		sbi->s_mount_opt &= ~EXT2_MOUNT_XIP;
1245 		sbi->s_mount_opt |= old_mount_opt & EXT2_MOUNT_XIP;
1246 	}
1247 	if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) {
1248 		spin_unlock(&sbi->s_lock);
1249 		return 0;
1250 	}
1251 	if (*flags & MS_RDONLY) {
1252 		if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1253 		    !(sbi->s_mount_state & EXT2_VALID_FS)) {
1254 			spin_unlock(&sbi->s_lock);
1255 			return 0;
1256 		}
1257 
1258 		/*
1259 		 * OK, we are remounting a valid rw partition rdonly, so set
1260 		 * the rdonly flag and then mark the partition as valid again.
1261 		 */
1262 		es->s_state = cpu_to_le16(sbi->s_mount_state);
1263 		es->s_mtime = cpu_to_le32(get_seconds());
1264 		spin_unlock(&sbi->s_lock);
1265 
1266 		err = dquot_suspend(sb, -1);
1267 		if (err < 0) {
1268 			spin_lock(&sbi->s_lock);
1269 			goto restore_opts;
1270 		}
1271 
1272 		ext2_sync_super(sb, es, 1);
1273 	} else {
1274 		__le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1275 					       ~EXT2_FEATURE_RO_COMPAT_SUPP);
1276 		if (ret) {
1277 			ext2_msg(sb, KERN_WARNING,
1278 				"warning: couldn't remount RDWR because of "
1279 				"unsupported optional features (%x).",
1280 				le32_to_cpu(ret));
1281 			err = -EROFS;
1282 			goto restore_opts;
1283 		}
1284 		/*
1285 		 * Mounting a RDONLY partition read-write, so reread and
1286 		 * store the current valid flag.  (It may have been changed
1287 		 * by e2fsck since we originally mounted the partition.)
1288 		 */
1289 		sbi->s_mount_state = le16_to_cpu(es->s_state);
1290 		if (!ext2_setup_super (sb, es, 0))
1291 			sb->s_flags &= ~MS_RDONLY;
1292 		spin_unlock(&sbi->s_lock);
1293 
1294 		ext2_write_super(sb);
1295 
1296 		dquot_resume(sb, -1);
1297 	}
1298 
1299 	return 0;
1300 restore_opts:
1301 	sbi->s_mount_opt = old_opts.s_mount_opt;
1302 	sbi->s_resuid = old_opts.s_resuid;
1303 	sbi->s_resgid = old_opts.s_resgid;
1304 	sb->s_flags = old_sb_flags;
1305 	spin_unlock(&sbi->s_lock);
1306 	return err;
1307 }
1308 
1309 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1310 {
1311 	struct super_block *sb = dentry->d_sb;
1312 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1313 	struct ext2_super_block *es = sbi->s_es;
1314 	u64 fsid;
1315 
1316 	spin_lock(&sbi->s_lock);
1317 
1318 	if (test_opt (sb, MINIX_DF))
1319 		sbi->s_overhead_last = 0;
1320 	else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1321 		unsigned long i, overhead = 0;
1322 		smp_rmb();
1323 
1324 		/*
1325 		 * Compute the overhead (FS structures). This is constant
1326 		 * for a given filesystem unless the number of block groups
1327 		 * changes so we cache the previous value until it does.
1328 		 */
1329 
1330 		/*
1331 		 * All of the blocks before first_data_block are
1332 		 * overhead
1333 		 */
1334 		overhead = le32_to_cpu(es->s_first_data_block);
1335 
1336 		/*
1337 		 * Add the overhead attributed to the superblock and
1338 		 * block group descriptors.  If the sparse superblocks
1339 		 * feature is turned on, then not all groups have this.
1340 		 */
1341 		for (i = 0; i < sbi->s_groups_count; i++)
1342 			overhead += ext2_bg_has_super(sb, i) +
1343 				ext2_bg_num_gdb(sb, i);
1344 
1345 		/*
1346 		 * Every block group has an inode bitmap, a block
1347 		 * bitmap, and an inode table.
1348 		 */
1349 		overhead += (sbi->s_groups_count *
1350 			     (2 + sbi->s_itb_per_group));
1351 		sbi->s_overhead_last = overhead;
1352 		smp_wmb();
1353 		sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1354 	}
1355 
1356 	buf->f_type = EXT2_SUPER_MAGIC;
1357 	buf->f_bsize = sb->s_blocksize;
1358 	buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1359 	buf->f_bfree = ext2_count_free_blocks(sb);
1360 	es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1361 	buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1362 	if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1363 		buf->f_bavail = 0;
1364 	buf->f_files = le32_to_cpu(es->s_inodes_count);
1365 	buf->f_ffree = ext2_count_free_inodes(sb);
1366 	es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1367 	buf->f_namelen = EXT2_NAME_LEN;
1368 	fsid = le64_to_cpup((void *)es->s_uuid) ^
1369 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1370 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1371 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1372 	spin_unlock(&sbi->s_lock);
1373 	return 0;
1374 }
1375 
1376 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1377 	int flags, const char *dev_name, void *data)
1378 {
1379 	return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1380 }
1381 
1382 #ifdef CONFIG_QUOTA
1383 
1384 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1385  * acquiring the locks... As quota files are never truncated and quota code
1386  * itself serializes the operations (and no one else should touch the files)
1387  * we don't have to be afraid of races */
1388 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1389 			       size_t len, loff_t off)
1390 {
1391 	struct inode *inode = sb_dqopt(sb)->files[type];
1392 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1393 	int err = 0;
1394 	int offset = off & (sb->s_blocksize - 1);
1395 	int tocopy;
1396 	size_t toread;
1397 	struct buffer_head tmp_bh;
1398 	struct buffer_head *bh;
1399 	loff_t i_size = i_size_read(inode);
1400 
1401 	if (off > i_size)
1402 		return 0;
1403 	if (off+len > i_size)
1404 		len = i_size-off;
1405 	toread = len;
1406 	while (toread > 0) {
1407 		tocopy = sb->s_blocksize - offset < toread ?
1408 				sb->s_blocksize - offset : toread;
1409 
1410 		tmp_bh.b_state = 0;
1411 		tmp_bh.b_size = sb->s_blocksize;
1412 		err = ext2_get_block(inode, blk, &tmp_bh, 0);
1413 		if (err < 0)
1414 			return err;
1415 		if (!buffer_mapped(&tmp_bh))	/* A hole? */
1416 			memset(data, 0, tocopy);
1417 		else {
1418 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1419 			if (!bh)
1420 				return -EIO;
1421 			memcpy(data, bh->b_data+offset, tocopy);
1422 			brelse(bh);
1423 		}
1424 		offset = 0;
1425 		toread -= tocopy;
1426 		data += tocopy;
1427 		blk++;
1428 	}
1429 	return len;
1430 }
1431 
1432 /* Write to quotafile */
1433 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1434 				const char *data, size_t len, loff_t off)
1435 {
1436 	struct inode *inode = sb_dqopt(sb)->files[type];
1437 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1438 	int err = 0;
1439 	int offset = off & (sb->s_blocksize - 1);
1440 	int tocopy;
1441 	size_t towrite = len;
1442 	struct buffer_head tmp_bh;
1443 	struct buffer_head *bh;
1444 
1445 	mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
1446 	while (towrite > 0) {
1447 		tocopy = sb->s_blocksize - offset < towrite ?
1448 				sb->s_blocksize - offset : towrite;
1449 
1450 		tmp_bh.b_state = 0;
1451 		err = ext2_get_block(inode, blk, &tmp_bh, 1);
1452 		if (err < 0)
1453 			goto out;
1454 		if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1455 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1456 		else
1457 			bh = sb_getblk(sb, tmp_bh.b_blocknr);
1458 		if (!bh) {
1459 			err = -EIO;
1460 			goto out;
1461 		}
1462 		lock_buffer(bh);
1463 		memcpy(bh->b_data+offset, data, tocopy);
1464 		flush_dcache_page(bh->b_page);
1465 		set_buffer_uptodate(bh);
1466 		mark_buffer_dirty(bh);
1467 		unlock_buffer(bh);
1468 		brelse(bh);
1469 		offset = 0;
1470 		towrite -= tocopy;
1471 		data += tocopy;
1472 		blk++;
1473 	}
1474 out:
1475 	if (len == towrite) {
1476 		mutex_unlock(&inode->i_mutex);
1477 		return err;
1478 	}
1479 	if (inode->i_size < off+len-towrite)
1480 		i_size_write(inode, off+len-towrite);
1481 	inode->i_version++;
1482 	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1483 	mark_inode_dirty(inode);
1484 	mutex_unlock(&inode->i_mutex);
1485 	return len - towrite;
1486 }
1487 
1488 #endif
1489 
1490 static struct file_system_type ext2_fs_type = {
1491 	.owner		= THIS_MODULE,
1492 	.name		= "ext2",
1493 	.mount		= ext2_mount,
1494 	.kill_sb	= kill_block_super,
1495 	.fs_flags	= FS_REQUIRES_DEV,
1496 };
1497 
1498 static int __init init_ext2_fs(void)
1499 {
1500 	int err = init_ext2_xattr();
1501 	if (err)
1502 		return err;
1503 	err = init_inodecache();
1504 	if (err)
1505 		goto out1;
1506         err = register_filesystem(&ext2_fs_type);
1507 	if (err)
1508 		goto out;
1509 	return 0;
1510 out:
1511 	destroy_inodecache();
1512 out1:
1513 	exit_ext2_xattr();
1514 	return err;
1515 }
1516 
1517 static void __exit exit_ext2_fs(void)
1518 {
1519 	unregister_filesystem(&ext2_fs_type);
1520 	destroy_inodecache();
1521 	exit_ext2_xattr();
1522 }
1523 
1524 module_init(init_ext2_fs)
1525 module_exit(exit_ext2_fs)
1526