xref: /linux/fs/ext2/super.c (revision 4c62e9764ab403d42f9b8871b1241fe7812f19d4)
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 = (struct ext2_inode_info *)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 	return &ei->vfs_inode;
170 }
171 
172 static void ext2_i_callback(struct rcu_head *head)
173 {
174 	struct inode *inode = container_of(head, struct inode, i_rcu);
175 	kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
176 }
177 
178 static void ext2_destroy_inode(struct inode *inode)
179 {
180 	call_rcu(&inode->i_rcu, ext2_i_callback);
181 }
182 
183 static void init_once(void *foo)
184 {
185 	struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
186 
187 	rwlock_init(&ei->i_meta_lock);
188 #ifdef CONFIG_EXT2_FS_XATTR
189 	init_rwsem(&ei->xattr_sem);
190 #endif
191 	mutex_init(&ei->truncate_mutex);
192 	inode_init_once(&ei->vfs_inode);
193 }
194 
195 static int init_inodecache(void)
196 {
197 	ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
198 					     sizeof(struct ext2_inode_info),
199 					     0, (SLAB_RECLAIM_ACCOUNT|
200 						SLAB_MEM_SPREAD),
201 					     init_once);
202 	if (ext2_inode_cachep == NULL)
203 		return -ENOMEM;
204 	return 0;
205 }
206 
207 static void destroy_inodecache(void)
208 {
209 	/*
210 	 * Make sure all delayed rcu free inodes are flushed before we
211 	 * destroy cache.
212 	 */
213 	rcu_barrier();
214 	kmem_cache_destroy(ext2_inode_cachep);
215 }
216 
217 static int ext2_show_options(struct seq_file *seq, struct dentry *root)
218 {
219 	struct super_block *sb = root->d_sb;
220 	struct ext2_sb_info *sbi = EXT2_SB(sb);
221 	struct ext2_super_block *es = sbi->s_es;
222 	unsigned long def_mount_opts;
223 
224 	spin_lock(&sbi->s_lock);
225 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
226 
227 	if (sbi->s_sb_block != 1)
228 		seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
229 	if (test_opt(sb, MINIX_DF))
230 		seq_puts(seq, ",minixdf");
231 	if (test_opt(sb, GRPID))
232 		seq_puts(seq, ",grpid");
233 	if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
234 		seq_puts(seq, ",nogrpid");
235 	if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT2_DEF_RESUID)) ||
236 	    le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
237 		seq_printf(seq, ",resuid=%u",
238 				from_kuid_munged(&init_user_ns, sbi->s_resuid));
239 	}
240 	if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT2_DEF_RESGID)) ||
241 	    le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
242 		seq_printf(seq, ",resgid=%u",
243 				from_kgid_munged(&init_user_ns, sbi->s_resgid));
244 	}
245 	if (test_opt(sb, ERRORS_RO)) {
246 		int def_errors = le16_to_cpu(es->s_errors);
247 
248 		if (def_errors == EXT2_ERRORS_PANIC ||
249 		    def_errors == EXT2_ERRORS_CONTINUE) {
250 			seq_puts(seq, ",errors=remount-ro");
251 		}
252 	}
253 	if (test_opt(sb, ERRORS_CONT))
254 		seq_puts(seq, ",errors=continue");
255 	if (test_opt(sb, ERRORS_PANIC))
256 		seq_puts(seq, ",errors=panic");
257 	if (test_opt(sb, NO_UID32))
258 		seq_puts(seq, ",nouid32");
259 	if (test_opt(sb, DEBUG))
260 		seq_puts(seq, ",debug");
261 	if (test_opt(sb, OLDALLOC))
262 		seq_puts(seq, ",oldalloc");
263 
264 #ifdef CONFIG_EXT2_FS_XATTR
265 	if (test_opt(sb, XATTR_USER))
266 		seq_puts(seq, ",user_xattr");
267 	if (!test_opt(sb, XATTR_USER) &&
268 	    (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
269 		seq_puts(seq, ",nouser_xattr");
270 	}
271 #endif
272 
273 #ifdef CONFIG_EXT2_FS_POSIX_ACL
274 	if (test_opt(sb, POSIX_ACL))
275 		seq_puts(seq, ",acl");
276 	if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
277 		seq_puts(seq, ",noacl");
278 #endif
279 
280 	if (test_opt(sb, NOBH))
281 		seq_puts(seq, ",nobh");
282 
283 #if defined(CONFIG_QUOTA)
284 	if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
285 		seq_puts(seq, ",usrquota");
286 
287 	if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
288 		seq_puts(seq, ",grpquota");
289 #endif
290 
291 #if defined(CONFIG_EXT2_FS_XIP)
292 	if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
293 		seq_puts(seq, ",xip");
294 #endif
295 
296 	if (!test_opt(sb, RESERVATION))
297 		seq_puts(seq, ",noreservation");
298 
299 	spin_unlock(&sbi->s_lock);
300 	return 0;
301 }
302 
303 #ifdef CONFIG_QUOTA
304 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
305 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
306 #endif
307 
308 static const struct super_operations ext2_sops = {
309 	.alloc_inode	= ext2_alloc_inode,
310 	.destroy_inode	= ext2_destroy_inode,
311 	.write_inode	= ext2_write_inode,
312 	.evict_inode	= ext2_evict_inode,
313 	.put_super	= ext2_put_super,
314 	.sync_fs	= ext2_sync_fs,
315 	.freeze_fs	= ext2_freeze,
316 	.unfreeze_fs	= ext2_unfreeze,
317 	.statfs		= ext2_statfs,
318 	.remount_fs	= ext2_remount,
319 	.show_options	= ext2_show_options,
320 #ifdef CONFIG_QUOTA
321 	.quota_read	= ext2_quota_read,
322 	.quota_write	= ext2_quota_write,
323 #endif
324 };
325 
326 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
327 		u64 ino, u32 generation)
328 {
329 	struct inode *inode;
330 
331 	if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
332 		return ERR_PTR(-ESTALE);
333 	if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
334 		return ERR_PTR(-ESTALE);
335 
336 	/*
337 	 * ext2_iget isn't quite right if the inode is currently unallocated!
338 	 * However ext2_iget currently does appropriate checks to handle stale
339 	 * inodes so everything is OK.
340 	 */
341 	inode = ext2_iget(sb, ino);
342 	if (IS_ERR(inode))
343 		return ERR_CAST(inode);
344 	if (generation && inode->i_generation != generation) {
345 		/* we didn't find the right inode.. */
346 		iput(inode);
347 		return ERR_PTR(-ESTALE);
348 	}
349 	return inode;
350 }
351 
352 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
353 		int fh_len, int fh_type)
354 {
355 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
356 				    ext2_nfs_get_inode);
357 }
358 
359 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
360 		int fh_len, int fh_type)
361 {
362 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
363 				    ext2_nfs_get_inode);
364 }
365 
366 static const struct export_operations ext2_export_ops = {
367 	.fh_to_dentry = ext2_fh_to_dentry,
368 	.fh_to_parent = ext2_fh_to_parent,
369 	.get_parent = ext2_get_parent,
370 };
371 
372 static unsigned long get_sb_block(void **data)
373 {
374 	unsigned long 	sb_block;
375 	char 		*options = (char *) *data;
376 
377 	if (!options || strncmp(options, "sb=", 3) != 0)
378 		return 1;	/* Default location */
379 	options += 3;
380 	sb_block = simple_strtoul(options, &options, 0);
381 	if (*options && *options != ',') {
382 		printk("EXT2-fs: Invalid sb specification: %s\n",
383 		       (char *) *data);
384 		return 1;
385 	}
386 	if (*options == ',')
387 		options++;
388 	*data = (void *) options;
389 	return sb_block;
390 }
391 
392 enum {
393 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
394 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
395 	Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
396 	Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
397 	Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
398 	Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
399 };
400 
401 static const match_table_t tokens = {
402 	{Opt_bsd_df, "bsddf"},
403 	{Opt_minix_df, "minixdf"},
404 	{Opt_grpid, "grpid"},
405 	{Opt_grpid, "bsdgroups"},
406 	{Opt_nogrpid, "nogrpid"},
407 	{Opt_nogrpid, "sysvgroups"},
408 	{Opt_resgid, "resgid=%u"},
409 	{Opt_resuid, "resuid=%u"},
410 	{Opt_sb, "sb=%u"},
411 	{Opt_err_cont, "errors=continue"},
412 	{Opt_err_panic, "errors=panic"},
413 	{Opt_err_ro, "errors=remount-ro"},
414 	{Opt_nouid32, "nouid32"},
415 	{Opt_nocheck, "check=none"},
416 	{Opt_nocheck, "nocheck"},
417 	{Opt_debug, "debug"},
418 	{Opt_oldalloc, "oldalloc"},
419 	{Opt_orlov, "orlov"},
420 	{Opt_nobh, "nobh"},
421 	{Opt_user_xattr, "user_xattr"},
422 	{Opt_nouser_xattr, "nouser_xattr"},
423 	{Opt_acl, "acl"},
424 	{Opt_noacl, "noacl"},
425 	{Opt_xip, "xip"},
426 	{Opt_grpquota, "grpquota"},
427 	{Opt_ignore, "noquota"},
428 	{Opt_quota, "quota"},
429 	{Opt_usrquota, "usrquota"},
430 	{Opt_reservation, "reservation"},
431 	{Opt_noreservation, "noreservation"},
432 	{Opt_err, NULL}
433 };
434 
435 static int parse_options(char *options, struct super_block *sb)
436 {
437 	char *p;
438 	struct ext2_sb_info *sbi = EXT2_SB(sb);
439 	substring_t args[MAX_OPT_ARGS];
440 	int option;
441 	kuid_t uid;
442 	kgid_t gid;
443 
444 	if (!options)
445 		return 1;
446 
447 	while ((p = strsep (&options, ",")) != NULL) {
448 		int token;
449 		if (!*p)
450 			continue;
451 
452 		token = match_token(p, tokens, args);
453 		switch (token) {
454 		case Opt_bsd_df:
455 			clear_opt (sbi->s_mount_opt, MINIX_DF);
456 			break;
457 		case Opt_minix_df:
458 			set_opt (sbi->s_mount_opt, MINIX_DF);
459 			break;
460 		case Opt_grpid:
461 			set_opt (sbi->s_mount_opt, GRPID);
462 			break;
463 		case Opt_nogrpid:
464 			clear_opt (sbi->s_mount_opt, GRPID);
465 			break;
466 		case Opt_resuid:
467 			if (match_int(&args[0], &option))
468 				return 0;
469 			uid = make_kuid(current_user_ns(), option);
470 			if (!uid_valid(uid)) {
471 				ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
472 				return 0;
473 
474 			}
475 			sbi->s_resuid = uid;
476 			break;
477 		case Opt_resgid:
478 			if (match_int(&args[0], &option))
479 				return 0;
480 			gid = make_kgid(current_user_ns(), option);
481 			if (!gid_valid(gid)) {
482 				ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
483 				return 0;
484 			}
485 			sbi->s_resgid = gid;
486 			break;
487 		case Opt_sb:
488 			/* handled by get_sb_block() instead of here */
489 			/* *sb_block = match_int(&args[0]); */
490 			break;
491 		case Opt_err_panic:
492 			clear_opt (sbi->s_mount_opt, ERRORS_CONT);
493 			clear_opt (sbi->s_mount_opt, ERRORS_RO);
494 			set_opt (sbi->s_mount_opt, ERRORS_PANIC);
495 			break;
496 		case Opt_err_ro:
497 			clear_opt (sbi->s_mount_opt, ERRORS_CONT);
498 			clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
499 			set_opt (sbi->s_mount_opt, ERRORS_RO);
500 			break;
501 		case Opt_err_cont:
502 			clear_opt (sbi->s_mount_opt, ERRORS_RO);
503 			clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
504 			set_opt (sbi->s_mount_opt, ERRORS_CONT);
505 			break;
506 		case Opt_nouid32:
507 			set_opt (sbi->s_mount_opt, NO_UID32);
508 			break;
509 		case Opt_nocheck:
510 			clear_opt (sbi->s_mount_opt, CHECK);
511 			break;
512 		case Opt_debug:
513 			set_opt (sbi->s_mount_opt, DEBUG);
514 			break;
515 		case Opt_oldalloc:
516 			set_opt (sbi->s_mount_opt, OLDALLOC);
517 			break;
518 		case Opt_orlov:
519 			clear_opt (sbi->s_mount_opt, OLDALLOC);
520 			break;
521 		case Opt_nobh:
522 			set_opt (sbi->s_mount_opt, NOBH);
523 			break;
524 #ifdef CONFIG_EXT2_FS_XATTR
525 		case Opt_user_xattr:
526 			set_opt (sbi->s_mount_opt, XATTR_USER);
527 			break;
528 		case Opt_nouser_xattr:
529 			clear_opt (sbi->s_mount_opt, XATTR_USER);
530 			break;
531 #else
532 		case Opt_user_xattr:
533 		case Opt_nouser_xattr:
534 			ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
535 				"not supported");
536 			break;
537 #endif
538 #ifdef CONFIG_EXT2_FS_POSIX_ACL
539 		case Opt_acl:
540 			set_opt(sbi->s_mount_opt, POSIX_ACL);
541 			break;
542 		case Opt_noacl:
543 			clear_opt(sbi->s_mount_opt, POSIX_ACL);
544 			break;
545 #else
546 		case Opt_acl:
547 		case Opt_noacl:
548 			ext2_msg(sb, KERN_INFO,
549 				"(no)acl options not supported");
550 			break;
551 #endif
552 		case Opt_xip:
553 #ifdef CONFIG_EXT2_FS_XIP
554 			set_opt (sbi->s_mount_opt, XIP);
555 #else
556 			ext2_msg(sb, KERN_INFO, "xip option not supported");
557 #endif
558 			break;
559 
560 #if defined(CONFIG_QUOTA)
561 		case Opt_quota:
562 		case Opt_usrquota:
563 			set_opt(sbi->s_mount_opt, USRQUOTA);
564 			break;
565 
566 		case Opt_grpquota:
567 			set_opt(sbi->s_mount_opt, GRPQUOTA);
568 			break;
569 #else
570 		case Opt_quota:
571 		case Opt_usrquota:
572 		case Opt_grpquota:
573 			ext2_msg(sb, KERN_INFO,
574 				"quota operations not supported");
575 			break;
576 #endif
577 
578 		case Opt_reservation:
579 			set_opt(sbi->s_mount_opt, RESERVATION);
580 			ext2_msg(sb, KERN_INFO, "reservations ON");
581 			break;
582 		case Opt_noreservation:
583 			clear_opt(sbi->s_mount_opt, RESERVATION);
584 			ext2_msg(sb, KERN_INFO, "reservations OFF");
585 			break;
586 		case Opt_ignore:
587 			break;
588 		default:
589 			return 0;
590 		}
591 	}
592 	return 1;
593 }
594 
595 static int ext2_setup_super (struct super_block * sb,
596 			      struct ext2_super_block * es,
597 			      int read_only)
598 {
599 	int res = 0;
600 	struct ext2_sb_info *sbi = EXT2_SB(sb);
601 
602 	if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
603 		ext2_msg(sb, KERN_ERR,
604 			"error: revision level too high, "
605 			"forcing read-only mode");
606 		res = MS_RDONLY;
607 	}
608 	if (read_only)
609 		return res;
610 	if (!(sbi->s_mount_state & EXT2_VALID_FS))
611 		ext2_msg(sb, KERN_WARNING,
612 			"warning: mounting unchecked fs, "
613 			"running e2fsck is recommended");
614 	else if ((sbi->s_mount_state & EXT2_ERROR_FS))
615 		ext2_msg(sb, KERN_WARNING,
616 			"warning: mounting fs with errors, "
617 			"running e2fsck is recommended");
618 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
619 		 le16_to_cpu(es->s_mnt_count) >=
620 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
621 		ext2_msg(sb, KERN_WARNING,
622 			"warning: maximal mount count reached, "
623 			"running e2fsck is recommended");
624 	else if (le32_to_cpu(es->s_checkinterval) &&
625 		(le32_to_cpu(es->s_lastcheck) +
626 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
627 		ext2_msg(sb, KERN_WARNING,
628 			"warning: checktime reached, "
629 			"running e2fsck is recommended");
630 	if (!le16_to_cpu(es->s_max_mnt_count))
631 		es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
632 	le16_add_cpu(&es->s_mnt_count, 1);
633 	if (test_opt (sb, DEBUG))
634 		ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
635 			"bpg=%lu, ipg=%lu, mo=%04lx]",
636 			EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
637 			sbi->s_frag_size,
638 			sbi->s_groups_count,
639 			EXT2_BLOCKS_PER_GROUP(sb),
640 			EXT2_INODES_PER_GROUP(sb),
641 			sbi->s_mount_opt);
642 	return res;
643 }
644 
645 static int ext2_check_descriptors(struct super_block *sb)
646 {
647 	int i;
648 	struct ext2_sb_info *sbi = EXT2_SB(sb);
649 
650 	ext2_debug ("Checking group descriptors");
651 
652 	for (i = 0; i < sbi->s_groups_count; i++) {
653 		struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
654 		ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
655 		ext2_fsblk_t last_block;
656 
657 		if (i == sbi->s_groups_count - 1)
658 			last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
659 		else
660 			last_block = first_block +
661 				(EXT2_BLOCKS_PER_GROUP(sb) - 1);
662 
663 		if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
664 		    le32_to_cpu(gdp->bg_block_bitmap) > last_block)
665 		{
666 			ext2_error (sb, "ext2_check_descriptors",
667 				    "Block bitmap for group %d"
668 				    " not in group (block %lu)!",
669 				    i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
670 			return 0;
671 		}
672 		if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
673 		    le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
674 		{
675 			ext2_error (sb, "ext2_check_descriptors",
676 				    "Inode bitmap for group %d"
677 				    " not in group (block %lu)!",
678 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
679 			return 0;
680 		}
681 		if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
682 		    le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
683 		    last_block)
684 		{
685 			ext2_error (sb, "ext2_check_descriptors",
686 				    "Inode table for group %d"
687 				    " not in group (block %lu)!",
688 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
689 			return 0;
690 		}
691 	}
692 	return 1;
693 }
694 
695 /*
696  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
697  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
698  * We need to be 1 filesystem block less than the 2^32 sector limit.
699  */
700 static loff_t ext2_max_size(int bits)
701 {
702 	loff_t res = EXT2_NDIR_BLOCKS;
703 	int meta_blocks;
704 	loff_t upper_limit;
705 
706 	/* This is calculated to be the largest file size for a
707 	 * dense, file such that the total number of
708 	 * sectors in the file, including data and all indirect blocks,
709 	 * does not exceed 2^32 -1
710 	 * __u32 i_blocks representing the total number of
711 	 * 512 bytes blocks of the file
712 	 */
713 	upper_limit = (1LL << 32) - 1;
714 
715 	/* total blocks in file system block size */
716 	upper_limit >>= (bits - 9);
717 
718 
719 	/* indirect blocks */
720 	meta_blocks = 1;
721 	/* double indirect blocks */
722 	meta_blocks += 1 + (1LL << (bits-2));
723 	/* tripple indirect blocks */
724 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
725 
726 	upper_limit -= meta_blocks;
727 	upper_limit <<= bits;
728 
729 	res += 1LL << (bits-2);
730 	res += 1LL << (2*(bits-2));
731 	res += 1LL << (3*(bits-2));
732 	res <<= bits;
733 	if (res > upper_limit)
734 		res = upper_limit;
735 
736 	if (res > MAX_LFS_FILESIZE)
737 		res = MAX_LFS_FILESIZE;
738 
739 	return res;
740 }
741 
742 static unsigned long descriptor_loc(struct super_block *sb,
743 				    unsigned long logic_sb_block,
744 				    int nr)
745 {
746 	struct ext2_sb_info *sbi = EXT2_SB(sb);
747 	unsigned long bg, first_meta_bg;
748 	int has_super = 0;
749 
750 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
751 
752 	if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
753 	    nr < first_meta_bg)
754 		return (logic_sb_block + nr + 1);
755 	bg = sbi->s_desc_per_block * nr;
756 	if (ext2_bg_has_super(sb, bg))
757 		has_super = 1;
758 
759 	return ext2_group_first_block_no(sb, bg) + has_super;
760 }
761 
762 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
763 {
764 	struct buffer_head * bh;
765 	struct ext2_sb_info * sbi;
766 	struct ext2_super_block * es;
767 	struct inode *root;
768 	unsigned long block;
769 	unsigned long sb_block = get_sb_block(&data);
770 	unsigned long logic_sb_block;
771 	unsigned long offset = 0;
772 	unsigned long def_mount_opts;
773 	long ret = -EINVAL;
774 	int blocksize = BLOCK_SIZE;
775 	int db_count;
776 	int i, j;
777 	__le32 features;
778 	int err;
779 
780 	err = -ENOMEM;
781 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
782 	if (!sbi)
783 		goto failed;
784 
785 	sbi->s_blockgroup_lock =
786 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
787 	if (!sbi->s_blockgroup_lock) {
788 		kfree(sbi);
789 		goto failed;
790 	}
791 	sb->s_fs_info = sbi;
792 	sbi->s_sb_block = sb_block;
793 
794 	spin_lock_init(&sbi->s_lock);
795 
796 	/*
797 	 * See what the current blocksize for the device is, and
798 	 * use that as the blocksize.  Otherwise (or if the blocksize
799 	 * is smaller than the default) use the default.
800 	 * This is important for devices that have a hardware
801 	 * sectorsize that is larger than the default.
802 	 */
803 	blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
804 	if (!blocksize) {
805 		ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
806 		goto failed_sbi;
807 	}
808 
809 	/*
810 	 * If the superblock doesn't start on a hardware sector boundary,
811 	 * calculate the offset.
812 	 */
813 	if (blocksize != BLOCK_SIZE) {
814 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
815 		offset = (sb_block*BLOCK_SIZE) % blocksize;
816 	} else {
817 		logic_sb_block = sb_block;
818 	}
819 
820 	if (!(bh = sb_bread(sb, logic_sb_block))) {
821 		ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
822 		goto failed_sbi;
823 	}
824 	/*
825 	 * Note: s_es must be initialized as soon as possible because
826 	 *       some ext2 macro-instructions depend on its value
827 	 */
828 	es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
829 	sbi->s_es = es;
830 	sb->s_magic = le16_to_cpu(es->s_magic);
831 
832 	if (sb->s_magic != EXT2_SUPER_MAGIC)
833 		goto cantfind_ext2;
834 
835 	/* Set defaults before we parse the mount options */
836 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
837 	if (def_mount_opts & EXT2_DEFM_DEBUG)
838 		set_opt(sbi->s_mount_opt, DEBUG);
839 	if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
840 		set_opt(sbi->s_mount_opt, GRPID);
841 	if (def_mount_opts & EXT2_DEFM_UID16)
842 		set_opt(sbi->s_mount_opt, NO_UID32);
843 #ifdef CONFIG_EXT2_FS_XATTR
844 	if (def_mount_opts & EXT2_DEFM_XATTR_USER)
845 		set_opt(sbi->s_mount_opt, XATTR_USER);
846 #endif
847 #ifdef CONFIG_EXT2_FS_POSIX_ACL
848 	if (def_mount_opts & EXT2_DEFM_ACL)
849 		set_opt(sbi->s_mount_opt, POSIX_ACL);
850 #endif
851 
852 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
853 		set_opt(sbi->s_mount_opt, ERRORS_PANIC);
854 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
855 		set_opt(sbi->s_mount_opt, ERRORS_CONT);
856 	else
857 		set_opt(sbi->s_mount_opt, ERRORS_RO);
858 
859 	sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
860 	sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
861 
862 	set_opt(sbi->s_mount_opt, RESERVATION);
863 
864 	if (!parse_options((char *) data, sb))
865 		goto failed_mount;
866 
867 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
868 		((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
869 		 MS_POSIXACL : 0);
870 
871 	ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
872 				    EXT2_MOUNT_XIP if not */
873 
874 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
875 	    (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
876 	     EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
877 	     EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
878 		ext2_msg(sb, KERN_WARNING,
879 			"warning: feature flags set on rev 0 fs, "
880 			"running e2fsck is recommended");
881 	/*
882 	 * Check feature flags regardless of the revision level, since we
883 	 * previously didn't change the revision level when setting the flags,
884 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
885 	 */
886 	features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
887 	if (features) {
888 		ext2_msg(sb, KERN_ERR,	"error: couldn't mount because of "
889 		       "unsupported optional features (%x)",
890 			le32_to_cpu(features));
891 		goto failed_mount;
892 	}
893 	if (!(sb->s_flags & MS_RDONLY) &&
894 	    (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
895 		ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
896 		       "unsupported optional features (%x)",
897 		       le32_to_cpu(features));
898 		goto failed_mount;
899 	}
900 
901 	blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
902 
903 	if (ext2_use_xip(sb) && blocksize != PAGE_SIZE) {
904 		if (!silent)
905 			ext2_msg(sb, KERN_ERR,
906 				"error: unsupported blocksize for xip");
907 		goto failed_mount;
908 	}
909 
910 	/* If the blocksize doesn't match, re-read the thing.. */
911 	if (sb->s_blocksize != blocksize) {
912 		brelse(bh);
913 
914 		if (!sb_set_blocksize(sb, blocksize)) {
915 			ext2_msg(sb, KERN_ERR,
916 				"error: bad blocksize %d", blocksize);
917 			goto failed_sbi;
918 		}
919 
920 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
921 		offset = (sb_block*BLOCK_SIZE) % blocksize;
922 		bh = sb_bread(sb, logic_sb_block);
923 		if(!bh) {
924 			ext2_msg(sb, KERN_ERR, "error: couldn't read"
925 				"superblock on 2nd try");
926 			goto failed_sbi;
927 		}
928 		es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
929 		sbi->s_es = es;
930 		if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
931 			ext2_msg(sb, KERN_ERR, "error: magic mismatch");
932 			goto failed_mount;
933 		}
934 	}
935 
936 	sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
937 	sb->s_max_links = EXT2_LINK_MAX;
938 
939 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
940 		sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
941 		sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
942 	} else {
943 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
944 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
945 		if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
946 		    !is_power_of_2(sbi->s_inode_size) ||
947 		    (sbi->s_inode_size > blocksize)) {
948 			ext2_msg(sb, KERN_ERR,
949 				"error: unsupported inode size: %d",
950 				sbi->s_inode_size);
951 			goto failed_mount;
952 		}
953 	}
954 
955 	sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
956 				   le32_to_cpu(es->s_log_frag_size);
957 	if (sbi->s_frag_size == 0)
958 		goto cantfind_ext2;
959 	sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
960 
961 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
962 	sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
963 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
964 
965 	if (EXT2_INODE_SIZE(sb) == 0)
966 		goto cantfind_ext2;
967 	sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
968 	if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
969 		goto cantfind_ext2;
970 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
971 					sbi->s_inodes_per_block;
972 	sbi->s_desc_per_block = sb->s_blocksize /
973 					sizeof (struct ext2_group_desc);
974 	sbi->s_sbh = bh;
975 	sbi->s_mount_state = le16_to_cpu(es->s_state);
976 	sbi->s_addr_per_block_bits =
977 		ilog2 (EXT2_ADDR_PER_BLOCK(sb));
978 	sbi->s_desc_per_block_bits =
979 		ilog2 (EXT2_DESC_PER_BLOCK(sb));
980 
981 	if (sb->s_magic != EXT2_SUPER_MAGIC)
982 		goto cantfind_ext2;
983 
984 	if (sb->s_blocksize != bh->b_size) {
985 		if (!silent)
986 			ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
987 		goto failed_mount;
988 	}
989 
990 	if (sb->s_blocksize != sbi->s_frag_size) {
991 		ext2_msg(sb, KERN_ERR,
992 			"error: fragsize %lu != blocksize %lu"
993 			"(not supported yet)",
994 			sbi->s_frag_size, sb->s_blocksize);
995 		goto failed_mount;
996 	}
997 
998 	if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
999 		ext2_msg(sb, KERN_ERR,
1000 			"error: #blocks per group too big: %lu",
1001 			sbi->s_blocks_per_group);
1002 		goto failed_mount;
1003 	}
1004 	if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
1005 		ext2_msg(sb, KERN_ERR,
1006 			"error: #fragments per group too big: %lu",
1007 			sbi->s_frags_per_group);
1008 		goto failed_mount;
1009 	}
1010 	if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1011 		ext2_msg(sb, KERN_ERR,
1012 			"error: #inodes per group too big: %lu",
1013 			sbi->s_inodes_per_group);
1014 		goto failed_mount;
1015 	}
1016 
1017 	if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1018 		goto cantfind_ext2;
1019  	sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1020  				le32_to_cpu(es->s_first_data_block) - 1)
1021  					/ EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1022 	db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1023 		   EXT2_DESC_PER_BLOCK(sb);
1024 	sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
1025 	if (sbi->s_group_desc == NULL) {
1026 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1027 		goto failed_mount;
1028 	}
1029 	bgl_lock_init(sbi->s_blockgroup_lock);
1030 	sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1031 	if (!sbi->s_debts) {
1032 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1033 		goto failed_mount_group_desc;
1034 	}
1035 	for (i = 0; i < db_count; i++) {
1036 		block = descriptor_loc(sb, logic_sb_block, i);
1037 		sbi->s_group_desc[i] = sb_bread(sb, block);
1038 		if (!sbi->s_group_desc[i]) {
1039 			for (j = 0; j < i; j++)
1040 				brelse (sbi->s_group_desc[j]);
1041 			ext2_msg(sb, KERN_ERR,
1042 				"error: unable to read group descriptors");
1043 			goto failed_mount_group_desc;
1044 		}
1045 	}
1046 	if (!ext2_check_descriptors (sb)) {
1047 		ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1048 		goto failed_mount2;
1049 	}
1050 	sbi->s_gdb_count = db_count;
1051 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1052 	spin_lock_init(&sbi->s_next_gen_lock);
1053 
1054 	/* per fileystem reservation list head & lock */
1055 	spin_lock_init(&sbi->s_rsv_window_lock);
1056 	sbi->s_rsv_window_root = RB_ROOT;
1057 	/*
1058 	 * Add a single, static dummy reservation to the start of the
1059 	 * reservation window list --- it gives us a placeholder for
1060 	 * append-at-start-of-list which makes the allocation logic
1061 	 * _much_ simpler.
1062 	 */
1063 	sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1064 	sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1065 	sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1066 	sbi->s_rsv_window_head.rsv_goal_size = 0;
1067 	ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1068 
1069 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
1070 				ext2_count_free_blocks(sb));
1071 	if (!err) {
1072 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
1073 				ext2_count_free_inodes(sb));
1074 	}
1075 	if (!err) {
1076 		err = percpu_counter_init(&sbi->s_dirs_counter,
1077 				ext2_count_dirs(sb));
1078 	}
1079 	if (err) {
1080 		ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1081 		goto failed_mount3;
1082 	}
1083 	/*
1084 	 * set up enough so that it can read an inode
1085 	 */
1086 	sb->s_op = &ext2_sops;
1087 	sb->s_export_op = &ext2_export_ops;
1088 	sb->s_xattr = ext2_xattr_handlers;
1089 
1090 #ifdef CONFIG_QUOTA
1091 	sb->dq_op = &dquot_operations;
1092 	sb->s_qcop = &dquot_quotactl_ops;
1093 #endif
1094 
1095 	root = ext2_iget(sb, EXT2_ROOT_INO);
1096 	if (IS_ERR(root)) {
1097 		ret = PTR_ERR(root);
1098 		goto failed_mount3;
1099 	}
1100 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1101 		iput(root);
1102 		ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1103 		goto failed_mount3;
1104 	}
1105 
1106 	sb->s_root = d_make_root(root);
1107 	if (!sb->s_root) {
1108 		ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1109 		ret = -ENOMEM;
1110 		goto failed_mount3;
1111 	}
1112 	if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1113 		ext2_msg(sb, KERN_WARNING,
1114 			"warning: mounting ext3 filesystem as ext2");
1115 	if (ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY))
1116 		sb->s_flags |= MS_RDONLY;
1117 	ext2_write_super(sb);
1118 	return 0;
1119 
1120 cantfind_ext2:
1121 	if (!silent)
1122 		ext2_msg(sb, KERN_ERR,
1123 			"error: can't find an ext2 filesystem on dev %s.",
1124 			sb->s_id);
1125 	goto failed_mount;
1126 failed_mount3:
1127 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
1128 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
1129 	percpu_counter_destroy(&sbi->s_dirs_counter);
1130 failed_mount2:
1131 	for (i = 0; i < db_count; i++)
1132 		brelse(sbi->s_group_desc[i]);
1133 failed_mount_group_desc:
1134 	kfree(sbi->s_group_desc);
1135 	kfree(sbi->s_debts);
1136 failed_mount:
1137 	brelse(bh);
1138 failed_sbi:
1139 	sb->s_fs_info = NULL;
1140 	kfree(sbi->s_blockgroup_lock);
1141 	kfree(sbi);
1142 failed:
1143 	return ret;
1144 }
1145 
1146 static void ext2_clear_super_error(struct super_block *sb)
1147 {
1148 	struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1149 
1150 	if (buffer_write_io_error(sbh)) {
1151 		/*
1152 		 * Oh, dear.  A previous attempt to write the
1153 		 * superblock failed.  This could happen because the
1154 		 * USB device was yanked out.  Or it could happen to
1155 		 * be a transient write error and maybe the block will
1156 		 * be remapped.  Nothing we can do but to retry the
1157 		 * write and hope for the best.
1158 		 */
1159 		ext2_msg(sb, KERN_ERR,
1160 		       "previous I/O error to superblock detected\n");
1161 		clear_buffer_write_io_error(sbh);
1162 		set_buffer_uptodate(sbh);
1163 	}
1164 }
1165 
1166 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1167 			    int wait)
1168 {
1169 	ext2_clear_super_error(sb);
1170 	spin_lock(&EXT2_SB(sb)->s_lock);
1171 	es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1172 	es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1173 	es->s_wtime = cpu_to_le32(get_seconds());
1174 	/* unlock before we do IO */
1175 	spin_unlock(&EXT2_SB(sb)->s_lock);
1176 	mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1177 	if (wait)
1178 		sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1179 }
1180 
1181 /*
1182  * In the second extended file system, it is not necessary to
1183  * write the super block since we use a mapping of the
1184  * disk super block in a buffer.
1185  *
1186  * However, this function is still used to set the fs valid
1187  * flags to 0.  We need to set this flag to 0 since the fs
1188  * may have been checked while mounted and e2fsck may have
1189  * set s_state to EXT2_VALID_FS after some corrections.
1190  */
1191 static int ext2_sync_fs(struct super_block *sb, int wait)
1192 {
1193 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1194 	struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1195 
1196 	/*
1197 	 * Write quota structures to quota file, sync_blockdev() will write
1198 	 * them to disk later
1199 	 */
1200 	dquot_writeback_dquots(sb, -1);
1201 
1202 	spin_lock(&sbi->s_lock);
1203 	if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1204 		ext2_debug("setting valid to 0\n");
1205 		es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1206 	}
1207 	spin_unlock(&sbi->s_lock);
1208 	ext2_sync_super(sb, es, wait);
1209 	return 0;
1210 }
1211 
1212 static int ext2_freeze(struct super_block *sb)
1213 {
1214 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1215 
1216 	/*
1217 	 * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1218 	 * because we have unattached inodes and thus filesystem is not fully
1219 	 * consistent.
1220 	 */
1221 	if (atomic_long_read(&sb->s_remove_count)) {
1222 		ext2_sync_fs(sb, 1);
1223 		return 0;
1224 	}
1225 	/* Set EXT2_FS_VALID flag */
1226 	spin_lock(&sbi->s_lock);
1227 	sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1228 	spin_unlock(&sbi->s_lock);
1229 	ext2_sync_super(sb, sbi->s_es, 1);
1230 
1231 	return 0;
1232 }
1233 
1234 static int ext2_unfreeze(struct super_block *sb)
1235 {
1236 	/* Just write sb to clear EXT2_VALID_FS flag */
1237 	ext2_write_super(sb);
1238 
1239 	return 0;
1240 }
1241 
1242 void ext2_write_super(struct super_block *sb)
1243 {
1244 	if (!(sb->s_flags & MS_RDONLY))
1245 		ext2_sync_fs(sb, 1);
1246 }
1247 
1248 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1249 {
1250 	struct ext2_sb_info * sbi = EXT2_SB(sb);
1251 	struct ext2_super_block * es;
1252 	unsigned long old_mount_opt = sbi->s_mount_opt;
1253 	struct ext2_mount_options old_opts;
1254 	unsigned long old_sb_flags;
1255 	int err;
1256 
1257 	spin_lock(&sbi->s_lock);
1258 
1259 	/* Store the old options */
1260 	old_sb_flags = sb->s_flags;
1261 	old_opts.s_mount_opt = sbi->s_mount_opt;
1262 	old_opts.s_resuid = sbi->s_resuid;
1263 	old_opts.s_resgid = sbi->s_resgid;
1264 
1265 	/*
1266 	 * Allow the "check" option to be passed as a remount option.
1267 	 */
1268 	if (!parse_options(data, sb)) {
1269 		err = -EINVAL;
1270 		goto restore_opts;
1271 	}
1272 
1273 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1274 		((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1275 
1276 	ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1277 				    EXT2_MOUNT_XIP if not */
1278 
1279 	if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
1280 		ext2_msg(sb, KERN_WARNING,
1281 			"warning: unsupported blocksize for xip");
1282 		err = -EINVAL;
1283 		goto restore_opts;
1284 	}
1285 
1286 	es = sbi->s_es;
1287 	if ((sbi->s_mount_opt ^ old_mount_opt) & EXT2_MOUNT_XIP) {
1288 		ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1289 			 "xip flag with busy inodes while remounting");
1290 		sbi->s_mount_opt &= ~EXT2_MOUNT_XIP;
1291 		sbi->s_mount_opt |= old_mount_opt & EXT2_MOUNT_XIP;
1292 	}
1293 	if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) {
1294 		spin_unlock(&sbi->s_lock);
1295 		return 0;
1296 	}
1297 	if (*flags & MS_RDONLY) {
1298 		if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1299 		    !(sbi->s_mount_state & EXT2_VALID_FS)) {
1300 			spin_unlock(&sbi->s_lock);
1301 			return 0;
1302 		}
1303 
1304 		/*
1305 		 * OK, we are remounting a valid rw partition rdonly, so set
1306 		 * the rdonly flag and then mark the partition as valid again.
1307 		 */
1308 		es->s_state = cpu_to_le16(sbi->s_mount_state);
1309 		es->s_mtime = cpu_to_le32(get_seconds());
1310 		spin_unlock(&sbi->s_lock);
1311 
1312 		err = dquot_suspend(sb, -1);
1313 		if (err < 0) {
1314 			spin_lock(&sbi->s_lock);
1315 			goto restore_opts;
1316 		}
1317 
1318 		ext2_sync_super(sb, es, 1);
1319 	} else {
1320 		__le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1321 					       ~EXT2_FEATURE_RO_COMPAT_SUPP);
1322 		if (ret) {
1323 			ext2_msg(sb, KERN_WARNING,
1324 				"warning: couldn't remount RDWR because of "
1325 				"unsupported optional features (%x).",
1326 				le32_to_cpu(ret));
1327 			err = -EROFS;
1328 			goto restore_opts;
1329 		}
1330 		/*
1331 		 * Mounting a RDONLY partition read-write, so reread and
1332 		 * store the current valid flag.  (It may have been changed
1333 		 * by e2fsck since we originally mounted the partition.)
1334 		 */
1335 		sbi->s_mount_state = le16_to_cpu(es->s_state);
1336 		if (!ext2_setup_super (sb, es, 0))
1337 			sb->s_flags &= ~MS_RDONLY;
1338 		spin_unlock(&sbi->s_lock);
1339 
1340 		ext2_write_super(sb);
1341 
1342 		dquot_resume(sb, -1);
1343 	}
1344 
1345 	return 0;
1346 restore_opts:
1347 	sbi->s_mount_opt = old_opts.s_mount_opt;
1348 	sbi->s_resuid = old_opts.s_resuid;
1349 	sbi->s_resgid = old_opts.s_resgid;
1350 	sb->s_flags = old_sb_flags;
1351 	spin_unlock(&sbi->s_lock);
1352 	return err;
1353 }
1354 
1355 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1356 {
1357 	struct super_block *sb = dentry->d_sb;
1358 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1359 	struct ext2_super_block *es = sbi->s_es;
1360 	u64 fsid;
1361 
1362 	spin_lock(&sbi->s_lock);
1363 
1364 	if (test_opt (sb, MINIX_DF))
1365 		sbi->s_overhead_last = 0;
1366 	else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1367 		unsigned long i, overhead = 0;
1368 		smp_rmb();
1369 
1370 		/*
1371 		 * Compute the overhead (FS structures). This is constant
1372 		 * for a given filesystem unless the number of block groups
1373 		 * changes so we cache the previous value until it does.
1374 		 */
1375 
1376 		/*
1377 		 * All of the blocks before first_data_block are
1378 		 * overhead
1379 		 */
1380 		overhead = le32_to_cpu(es->s_first_data_block);
1381 
1382 		/*
1383 		 * Add the overhead attributed to the superblock and
1384 		 * block group descriptors.  If the sparse superblocks
1385 		 * feature is turned on, then not all groups have this.
1386 		 */
1387 		for (i = 0; i < sbi->s_groups_count; i++)
1388 			overhead += ext2_bg_has_super(sb, i) +
1389 				ext2_bg_num_gdb(sb, i);
1390 
1391 		/*
1392 		 * Every block group has an inode bitmap, a block
1393 		 * bitmap, and an inode table.
1394 		 */
1395 		overhead += (sbi->s_groups_count *
1396 			     (2 + sbi->s_itb_per_group));
1397 		sbi->s_overhead_last = overhead;
1398 		smp_wmb();
1399 		sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1400 	}
1401 
1402 	buf->f_type = EXT2_SUPER_MAGIC;
1403 	buf->f_bsize = sb->s_blocksize;
1404 	buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1405 	buf->f_bfree = ext2_count_free_blocks(sb);
1406 	es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1407 	buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1408 	if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1409 		buf->f_bavail = 0;
1410 	buf->f_files = le32_to_cpu(es->s_inodes_count);
1411 	buf->f_ffree = ext2_count_free_inodes(sb);
1412 	es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1413 	buf->f_namelen = EXT2_NAME_LEN;
1414 	fsid = le64_to_cpup((void *)es->s_uuid) ^
1415 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1416 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1417 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1418 	spin_unlock(&sbi->s_lock);
1419 	return 0;
1420 }
1421 
1422 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1423 	int flags, const char *dev_name, void *data)
1424 {
1425 	return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1426 }
1427 
1428 #ifdef CONFIG_QUOTA
1429 
1430 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1431  * acquiring the locks... As quota files are never truncated and quota code
1432  * itself serializes the operations (and no one else should touch the files)
1433  * we don't have to be afraid of races */
1434 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1435 			       size_t len, loff_t off)
1436 {
1437 	struct inode *inode = sb_dqopt(sb)->files[type];
1438 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1439 	int err = 0;
1440 	int offset = off & (sb->s_blocksize - 1);
1441 	int tocopy;
1442 	size_t toread;
1443 	struct buffer_head tmp_bh;
1444 	struct buffer_head *bh;
1445 	loff_t i_size = i_size_read(inode);
1446 
1447 	if (off > i_size)
1448 		return 0;
1449 	if (off+len > i_size)
1450 		len = i_size-off;
1451 	toread = len;
1452 	while (toread > 0) {
1453 		tocopy = sb->s_blocksize - offset < toread ?
1454 				sb->s_blocksize - offset : toread;
1455 
1456 		tmp_bh.b_state = 0;
1457 		tmp_bh.b_size = sb->s_blocksize;
1458 		err = ext2_get_block(inode, blk, &tmp_bh, 0);
1459 		if (err < 0)
1460 			return err;
1461 		if (!buffer_mapped(&tmp_bh))	/* A hole? */
1462 			memset(data, 0, tocopy);
1463 		else {
1464 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1465 			if (!bh)
1466 				return -EIO;
1467 			memcpy(data, bh->b_data+offset, tocopy);
1468 			brelse(bh);
1469 		}
1470 		offset = 0;
1471 		toread -= tocopy;
1472 		data += tocopy;
1473 		blk++;
1474 	}
1475 	return len;
1476 }
1477 
1478 /* Write to quotafile */
1479 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1480 				const char *data, size_t len, loff_t off)
1481 {
1482 	struct inode *inode = sb_dqopt(sb)->files[type];
1483 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1484 	int err = 0;
1485 	int offset = off & (sb->s_blocksize - 1);
1486 	int tocopy;
1487 	size_t towrite = len;
1488 	struct buffer_head tmp_bh;
1489 	struct buffer_head *bh;
1490 
1491 	while (towrite > 0) {
1492 		tocopy = sb->s_blocksize - offset < towrite ?
1493 				sb->s_blocksize - offset : towrite;
1494 
1495 		tmp_bh.b_state = 0;
1496 		err = ext2_get_block(inode, blk, &tmp_bh, 1);
1497 		if (err < 0)
1498 			goto out;
1499 		if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1500 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1501 		else
1502 			bh = sb_getblk(sb, tmp_bh.b_blocknr);
1503 		if (!bh) {
1504 			err = -EIO;
1505 			goto out;
1506 		}
1507 		lock_buffer(bh);
1508 		memcpy(bh->b_data+offset, data, tocopy);
1509 		flush_dcache_page(bh->b_page);
1510 		set_buffer_uptodate(bh);
1511 		mark_buffer_dirty(bh);
1512 		unlock_buffer(bh);
1513 		brelse(bh);
1514 		offset = 0;
1515 		towrite -= tocopy;
1516 		data += tocopy;
1517 		blk++;
1518 	}
1519 out:
1520 	if (len == towrite)
1521 		return err;
1522 	if (inode->i_size < off+len-towrite)
1523 		i_size_write(inode, off+len-towrite);
1524 	inode->i_version++;
1525 	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1526 	mark_inode_dirty(inode);
1527 	return len - towrite;
1528 }
1529 
1530 #endif
1531 
1532 static struct file_system_type ext2_fs_type = {
1533 	.owner		= THIS_MODULE,
1534 	.name		= "ext2",
1535 	.mount		= ext2_mount,
1536 	.kill_sb	= kill_block_super,
1537 	.fs_flags	= FS_REQUIRES_DEV,
1538 };
1539 
1540 static int __init init_ext2_fs(void)
1541 {
1542 	int err = init_ext2_xattr();
1543 	if (err)
1544 		return err;
1545 	err = init_inodecache();
1546 	if (err)
1547 		goto out1;
1548         err = register_filesystem(&ext2_fs_type);
1549 	if (err)
1550 		goto out;
1551 	return 0;
1552 out:
1553 	destroy_inodecache();
1554 out1:
1555 	exit_ext2_xattr();
1556 	return err;
1557 }
1558 
1559 static void __exit exit_ext2_fs(void)
1560 {
1561 	unregister_filesystem(&ext2_fs_type);
1562 	destroy_inodecache();
1563 	exit_ext2_xattr();
1564 }
1565 
1566 MODULE_AUTHOR("Remy Card and others");
1567 MODULE_DESCRIPTION("Second Extended Filesystem");
1568 MODULE_LICENSE("GPL");
1569 module_init(init_ext2_fs)
1570 module_exit(exit_ext2_fs)
1571