xref: /linux/fs/ext2/super.c (revision 80d443e8876602be2c130f79c4de81e12e2a700d)
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 <linux/uaccess.h>
35 #include "ext2.h"
36 #include "xattr.h"
37 #include "acl.h"
38 
39 static void ext2_sync_super(struct super_block *sb,
40 			    struct ext2_super_block *es, int wait);
41 static int ext2_remount (struct super_block * sb, int * flags, char * data);
42 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
43 static int ext2_sync_fs(struct super_block *sb, int wait);
44 static int ext2_freeze(struct super_block *sb);
45 static int ext2_unfreeze(struct super_block *sb);
46 
47 void ext2_error(struct super_block *sb, const char *function,
48 		const char *fmt, ...)
49 {
50 	struct va_format vaf;
51 	va_list args;
52 	struct ext2_sb_info *sbi = EXT2_SB(sb);
53 	struct ext2_super_block *es = sbi->s_es;
54 
55 	if (!(sb->s_flags & MS_RDONLY)) {
56 		spin_lock(&sbi->s_lock);
57 		sbi->s_mount_state |= EXT2_ERROR_FS;
58 		es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
59 		spin_unlock(&sbi->s_lock);
60 		ext2_sync_super(sb, es, 1);
61 	}
62 
63 	va_start(args, fmt);
64 
65 	vaf.fmt = fmt;
66 	vaf.va = &args;
67 
68 	printk(KERN_CRIT "EXT2-fs (%s): error: %s: %pV\n",
69 	       sb->s_id, function, &vaf);
70 
71 	va_end(args);
72 
73 	if (test_opt(sb, ERRORS_PANIC))
74 		panic("EXT2-fs: panic from previous error\n");
75 	if (test_opt(sb, ERRORS_RO)) {
76 		ext2_msg(sb, KERN_CRIT,
77 			     "error: remounting filesystem read-only");
78 		sb->s_flags |= MS_RDONLY;
79 	}
80 }
81 
82 void ext2_msg(struct super_block *sb, const char *prefix,
83 		const char *fmt, ...)
84 {
85 	struct va_format vaf;
86 	va_list args;
87 
88 	va_start(args, fmt);
89 
90 	vaf.fmt = fmt;
91 	vaf.va = &args;
92 
93 	printk("%sEXT2-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
94 
95 	va_end(args);
96 }
97 
98 /*
99  * This must be called with sbi->s_lock held.
100  */
101 void ext2_update_dynamic_rev(struct super_block *sb)
102 {
103 	struct ext2_super_block *es = EXT2_SB(sb)->s_es;
104 
105 	if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
106 		return;
107 
108 	ext2_msg(sb, KERN_WARNING,
109 		     "warning: updating to rev %d because of "
110 		     "new feature flag, running e2fsck is recommended",
111 		     EXT2_DYNAMIC_REV);
112 
113 	es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
114 	es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
115 	es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
116 	/* leave es->s_feature_*compat flags alone */
117 	/* es->s_uuid will be set by e2fsck if empty */
118 
119 	/*
120 	 * The rest of the superblock fields should be zero, and if not it
121 	 * means they are likely already in use, so leave them alone.  We
122 	 * can leave it up to e2fsck to clean up any inconsistencies there.
123 	 */
124 }
125 
126 static void ext2_put_super (struct super_block * sb)
127 {
128 	int db_count;
129 	int i;
130 	struct ext2_sb_info *sbi = EXT2_SB(sb);
131 
132 	dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
133 
134 	if (sbi->s_mb_cache) {
135 		ext2_xattr_destroy_cache(sbi->s_mb_cache);
136 		sbi->s_mb_cache = NULL;
137 	}
138 	if (!(sb->s_flags & MS_RDONLY)) {
139 		struct ext2_super_block *es = sbi->s_es;
140 
141 		spin_lock(&sbi->s_lock);
142 		es->s_state = cpu_to_le16(sbi->s_mount_state);
143 		spin_unlock(&sbi->s_lock);
144 		ext2_sync_super(sb, es, 1);
145 	}
146 	db_count = sbi->s_gdb_count;
147 	for (i = 0; i < db_count; i++)
148 		if (sbi->s_group_desc[i])
149 			brelse (sbi->s_group_desc[i]);
150 	kfree(sbi->s_group_desc);
151 	kfree(sbi->s_debts);
152 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
153 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
154 	percpu_counter_destroy(&sbi->s_dirs_counter);
155 	brelse (sbi->s_sbh);
156 	sb->s_fs_info = NULL;
157 	kfree(sbi->s_blockgroup_lock);
158 	kfree(sbi);
159 }
160 
161 static struct kmem_cache * ext2_inode_cachep;
162 
163 static struct inode *ext2_alloc_inode(struct super_block *sb)
164 {
165 	struct ext2_inode_info *ei;
166 	ei = kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
167 	if (!ei)
168 		return NULL;
169 	ei->i_block_alloc_info = NULL;
170 	ei->vfs_inode.i_version = 1;
171 #ifdef CONFIG_QUOTA
172 	memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
173 #endif
174 
175 	return &ei->vfs_inode;
176 }
177 
178 static void ext2_i_callback(struct rcu_head *head)
179 {
180 	struct inode *inode = container_of(head, struct inode, i_rcu);
181 	kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
182 }
183 
184 static void ext2_destroy_inode(struct inode *inode)
185 {
186 	call_rcu(&inode->i_rcu, ext2_i_callback);
187 }
188 
189 static void init_once(void *foo)
190 {
191 	struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
192 
193 	rwlock_init(&ei->i_meta_lock);
194 #ifdef CONFIG_EXT2_FS_XATTR
195 	init_rwsem(&ei->xattr_sem);
196 #endif
197 	mutex_init(&ei->truncate_mutex);
198 #ifdef CONFIG_FS_DAX
199 	init_rwsem(&ei->dax_sem);
200 #endif
201 	inode_init_once(&ei->vfs_inode);
202 }
203 
204 static int __init init_inodecache(void)
205 {
206 	ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
207 					     sizeof(struct ext2_inode_info),
208 					     0, (SLAB_RECLAIM_ACCOUNT|
209 						SLAB_MEM_SPREAD|SLAB_ACCOUNT),
210 					     init_once);
211 	if (ext2_inode_cachep == NULL)
212 		return -ENOMEM;
213 	return 0;
214 }
215 
216 static void destroy_inodecache(void)
217 {
218 	/*
219 	 * Make sure all delayed rcu free inodes are flushed before we
220 	 * destroy cache.
221 	 */
222 	rcu_barrier();
223 	kmem_cache_destroy(ext2_inode_cachep);
224 }
225 
226 static int ext2_show_options(struct seq_file *seq, struct dentry *root)
227 {
228 	struct super_block *sb = root->d_sb;
229 	struct ext2_sb_info *sbi = EXT2_SB(sb);
230 	struct ext2_super_block *es = sbi->s_es;
231 	unsigned long def_mount_opts;
232 
233 	spin_lock(&sbi->s_lock);
234 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
235 
236 	if (sbi->s_sb_block != 1)
237 		seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
238 	if (test_opt(sb, MINIX_DF))
239 		seq_puts(seq, ",minixdf");
240 	if (test_opt(sb, GRPID))
241 		seq_puts(seq, ",grpid");
242 	if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
243 		seq_puts(seq, ",nogrpid");
244 	if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT2_DEF_RESUID)) ||
245 	    le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
246 		seq_printf(seq, ",resuid=%u",
247 				from_kuid_munged(&init_user_ns, sbi->s_resuid));
248 	}
249 	if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT2_DEF_RESGID)) ||
250 	    le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
251 		seq_printf(seq, ",resgid=%u",
252 				from_kgid_munged(&init_user_ns, sbi->s_resgid));
253 	}
254 	if (test_opt(sb, ERRORS_RO)) {
255 		int def_errors = le16_to_cpu(es->s_errors);
256 
257 		if (def_errors == EXT2_ERRORS_PANIC ||
258 		    def_errors == EXT2_ERRORS_CONTINUE) {
259 			seq_puts(seq, ",errors=remount-ro");
260 		}
261 	}
262 	if (test_opt(sb, ERRORS_CONT))
263 		seq_puts(seq, ",errors=continue");
264 	if (test_opt(sb, ERRORS_PANIC))
265 		seq_puts(seq, ",errors=panic");
266 	if (test_opt(sb, NO_UID32))
267 		seq_puts(seq, ",nouid32");
268 	if (test_opt(sb, DEBUG))
269 		seq_puts(seq, ",debug");
270 	if (test_opt(sb, OLDALLOC))
271 		seq_puts(seq, ",oldalloc");
272 
273 #ifdef CONFIG_EXT2_FS_XATTR
274 	if (test_opt(sb, XATTR_USER))
275 		seq_puts(seq, ",user_xattr");
276 	if (!test_opt(sb, XATTR_USER) &&
277 	    (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
278 		seq_puts(seq, ",nouser_xattr");
279 	}
280 #endif
281 
282 #ifdef CONFIG_EXT2_FS_POSIX_ACL
283 	if (test_opt(sb, POSIX_ACL))
284 		seq_puts(seq, ",acl");
285 	if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
286 		seq_puts(seq, ",noacl");
287 #endif
288 
289 	if (test_opt(sb, NOBH))
290 		seq_puts(seq, ",nobh");
291 
292 #if defined(CONFIG_QUOTA)
293 	if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
294 		seq_puts(seq, ",usrquota");
295 
296 	if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
297 		seq_puts(seq, ",grpquota");
298 #endif
299 
300 #ifdef CONFIG_FS_DAX
301 	if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
302 		seq_puts(seq, ",xip");
303 	if (sbi->s_mount_opt & EXT2_MOUNT_DAX)
304 		seq_puts(seq, ",dax");
305 #endif
306 
307 	if (!test_opt(sb, RESERVATION))
308 		seq_puts(seq, ",noreservation");
309 
310 	spin_unlock(&sbi->s_lock);
311 	return 0;
312 }
313 
314 #ifdef CONFIG_QUOTA
315 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
316 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
317 static struct dquot **ext2_get_dquots(struct inode *inode)
318 {
319 	return EXT2_I(inode)->i_dquot;
320 }
321 #endif
322 
323 static const struct super_operations ext2_sops = {
324 	.alloc_inode	= ext2_alloc_inode,
325 	.destroy_inode	= ext2_destroy_inode,
326 	.write_inode	= ext2_write_inode,
327 	.evict_inode	= ext2_evict_inode,
328 	.put_super	= ext2_put_super,
329 	.sync_fs	= ext2_sync_fs,
330 	.freeze_fs	= ext2_freeze,
331 	.unfreeze_fs	= ext2_unfreeze,
332 	.statfs		= ext2_statfs,
333 	.remount_fs	= ext2_remount,
334 	.show_options	= ext2_show_options,
335 #ifdef CONFIG_QUOTA
336 	.quota_read	= ext2_quota_read,
337 	.quota_write	= ext2_quota_write,
338 	.get_dquots	= ext2_get_dquots,
339 #endif
340 };
341 
342 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
343 		u64 ino, u32 generation)
344 {
345 	struct inode *inode;
346 
347 	if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
348 		return ERR_PTR(-ESTALE);
349 	if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
350 		return ERR_PTR(-ESTALE);
351 
352 	/*
353 	 * ext2_iget isn't quite right if the inode is currently unallocated!
354 	 * However ext2_iget currently does appropriate checks to handle stale
355 	 * inodes so everything is OK.
356 	 */
357 	inode = ext2_iget(sb, ino);
358 	if (IS_ERR(inode))
359 		return ERR_CAST(inode);
360 	if (generation && inode->i_generation != generation) {
361 		/* we didn't find the right inode.. */
362 		iput(inode);
363 		return ERR_PTR(-ESTALE);
364 	}
365 	return inode;
366 }
367 
368 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
369 		int fh_len, int fh_type)
370 {
371 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
372 				    ext2_nfs_get_inode);
373 }
374 
375 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
376 		int fh_len, int fh_type)
377 {
378 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
379 				    ext2_nfs_get_inode);
380 }
381 
382 static const struct export_operations ext2_export_ops = {
383 	.fh_to_dentry = ext2_fh_to_dentry,
384 	.fh_to_parent = ext2_fh_to_parent,
385 	.get_parent = ext2_get_parent,
386 };
387 
388 static unsigned long get_sb_block(void **data)
389 {
390 	unsigned long 	sb_block;
391 	char 		*options = (char *) *data;
392 
393 	if (!options || strncmp(options, "sb=", 3) != 0)
394 		return 1;	/* Default location */
395 	options += 3;
396 	sb_block = simple_strtoul(options, &options, 0);
397 	if (*options && *options != ',') {
398 		printk("EXT2-fs: Invalid sb specification: %s\n",
399 		       (char *) *data);
400 		return 1;
401 	}
402 	if (*options == ',')
403 		options++;
404 	*data = (void *) options;
405 	return sb_block;
406 }
407 
408 enum {
409 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
410 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
411 	Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
412 	Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
413 	Opt_acl, Opt_noacl, Opt_xip, Opt_dax, Opt_ignore, Opt_err, Opt_quota,
414 	Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
415 };
416 
417 static const match_table_t tokens = {
418 	{Opt_bsd_df, "bsddf"},
419 	{Opt_minix_df, "minixdf"},
420 	{Opt_grpid, "grpid"},
421 	{Opt_grpid, "bsdgroups"},
422 	{Opt_nogrpid, "nogrpid"},
423 	{Opt_nogrpid, "sysvgroups"},
424 	{Opt_resgid, "resgid=%u"},
425 	{Opt_resuid, "resuid=%u"},
426 	{Opt_sb, "sb=%u"},
427 	{Opt_err_cont, "errors=continue"},
428 	{Opt_err_panic, "errors=panic"},
429 	{Opt_err_ro, "errors=remount-ro"},
430 	{Opt_nouid32, "nouid32"},
431 	{Opt_nocheck, "check=none"},
432 	{Opt_nocheck, "nocheck"},
433 	{Opt_debug, "debug"},
434 	{Opt_oldalloc, "oldalloc"},
435 	{Opt_orlov, "orlov"},
436 	{Opt_nobh, "nobh"},
437 	{Opt_user_xattr, "user_xattr"},
438 	{Opt_nouser_xattr, "nouser_xattr"},
439 	{Opt_acl, "acl"},
440 	{Opt_noacl, "noacl"},
441 	{Opt_xip, "xip"},
442 	{Opt_dax, "dax"},
443 	{Opt_grpquota, "grpquota"},
444 	{Opt_ignore, "noquota"},
445 	{Opt_quota, "quota"},
446 	{Opt_usrquota, "usrquota"},
447 	{Opt_reservation, "reservation"},
448 	{Opt_noreservation, "noreservation"},
449 	{Opt_err, NULL}
450 };
451 
452 static int parse_options(char *options, struct super_block *sb)
453 {
454 	char *p;
455 	struct ext2_sb_info *sbi = EXT2_SB(sb);
456 	substring_t args[MAX_OPT_ARGS];
457 	int option;
458 	kuid_t uid;
459 	kgid_t gid;
460 
461 	if (!options)
462 		return 1;
463 
464 	while ((p = strsep (&options, ",")) != NULL) {
465 		int token;
466 		if (!*p)
467 			continue;
468 
469 		token = match_token(p, tokens, args);
470 		switch (token) {
471 		case Opt_bsd_df:
472 			clear_opt (sbi->s_mount_opt, MINIX_DF);
473 			break;
474 		case Opt_minix_df:
475 			set_opt (sbi->s_mount_opt, MINIX_DF);
476 			break;
477 		case Opt_grpid:
478 			set_opt (sbi->s_mount_opt, GRPID);
479 			break;
480 		case Opt_nogrpid:
481 			clear_opt (sbi->s_mount_opt, GRPID);
482 			break;
483 		case Opt_resuid:
484 			if (match_int(&args[0], &option))
485 				return 0;
486 			uid = make_kuid(current_user_ns(), option);
487 			if (!uid_valid(uid)) {
488 				ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
489 				return 0;
490 
491 			}
492 			sbi->s_resuid = uid;
493 			break;
494 		case Opt_resgid:
495 			if (match_int(&args[0], &option))
496 				return 0;
497 			gid = make_kgid(current_user_ns(), option);
498 			if (!gid_valid(gid)) {
499 				ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
500 				return 0;
501 			}
502 			sbi->s_resgid = gid;
503 			break;
504 		case Opt_sb:
505 			/* handled by get_sb_block() instead of here */
506 			/* *sb_block = match_int(&args[0]); */
507 			break;
508 		case Opt_err_panic:
509 			clear_opt (sbi->s_mount_opt, ERRORS_CONT);
510 			clear_opt (sbi->s_mount_opt, ERRORS_RO);
511 			set_opt (sbi->s_mount_opt, ERRORS_PANIC);
512 			break;
513 		case Opt_err_ro:
514 			clear_opt (sbi->s_mount_opt, ERRORS_CONT);
515 			clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
516 			set_opt (sbi->s_mount_opt, ERRORS_RO);
517 			break;
518 		case Opt_err_cont:
519 			clear_opt (sbi->s_mount_opt, ERRORS_RO);
520 			clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
521 			set_opt (sbi->s_mount_opt, ERRORS_CONT);
522 			break;
523 		case Opt_nouid32:
524 			set_opt (sbi->s_mount_opt, NO_UID32);
525 			break;
526 		case Opt_nocheck:
527 			clear_opt (sbi->s_mount_opt, CHECK);
528 			break;
529 		case Opt_debug:
530 			set_opt (sbi->s_mount_opt, DEBUG);
531 			break;
532 		case Opt_oldalloc:
533 			set_opt (sbi->s_mount_opt, OLDALLOC);
534 			break;
535 		case Opt_orlov:
536 			clear_opt (sbi->s_mount_opt, OLDALLOC);
537 			break;
538 		case Opt_nobh:
539 			set_opt (sbi->s_mount_opt, NOBH);
540 			break;
541 #ifdef CONFIG_EXT2_FS_XATTR
542 		case Opt_user_xattr:
543 			set_opt (sbi->s_mount_opt, XATTR_USER);
544 			break;
545 		case Opt_nouser_xattr:
546 			clear_opt (sbi->s_mount_opt, XATTR_USER);
547 			break;
548 #else
549 		case Opt_user_xattr:
550 		case Opt_nouser_xattr:
551 			ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
552 				"not supported");
553 			break;
554 #endif
555 #ifdef CONFIG_EXT2_FS_POSIX_ACL
556 		case Opt_acl:
557 			set_opt(sbi->s_mount_opt, POSIX_ACL);
558 			break;
559 		case Opt_noacl:
560 			clear_opt(sbi->s_mount_opt, POSIX_ACL);
561 			break;
562 #else
563 		case Opt_acl:
564 		case Opt_noacl:
565 			ext2_msg(sb, KERN_INFO,
566 				"(no)acl options not supported");
567 			break;
568 #endif
569 		case Opt_xip:
570 			ext2_msg(sb, KERN_INFO, "use dax instead of xip");
571 			set_opt(sbi->s_mount_opt, XIP);
572 			/* Fall through */
573 		case Opt_dax:
574 #ifdef CONFIG_FS_DAX
575 			ext2_msg(sb, KERN_WARNING,
576 		"DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
577 			set_opt(sbi->s_mount_opt, DAX);
578 #else
579 			ext2_msg(sb, KERN_INFO, "dax option not supported");
580 #endif
581 			break;
582 
583 #if defined(CONFIG_QUOTA)
584 		case Opt_quota:
585 		case Opt_usrquota:
586 			set_opt(sbi->s_mount_opt, USRQUOTA);
587 			break;
588 
589 		case Opt_grpquota:
590 			set_opt(sbi->s_mount_opt, GRPQUOTA);
591 			break;
592 #else
593 		case Opt_quota:
594 		case Opt_usrquota:
595 		case Opt_grpquota:
596 			ext2_msg(sb, KERN_INFO,
597 				"quota operations not supported");
598 			break;
599 #endif
600 
601 		case Opt_reservation:
602 			set_opt(sbi->s_mount_opt, RESERVATION);
603 			ext2_msg(sb, KERN_INFO, "reservations ON");
604 			break;
605 		case Opt_noreservation:
606 			clear_opt(sbi->s_mount_opt, RESERVATION);
607 			ext2_msg(sb, KERN_INFO, "reservations OFF");
608 			break;
609 		case Opt_ignore:
610 			break;
611 		default:
612 			return 0;
613 		}
614 	}
615 	return 1;
616 }
617 
618 static int ext2_setup_super (struct super_block * sb,
619 			      struct ext2_super_block * es,
620 			      int read_only)
621 {
622 	int res = 0;
623 	struct ext2_sb_info *sbi = EXT2_SB(sb);
624 
625 	if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
626 		ext2_msg(sb, KERN_ERR,
627 			"error: revision level too high, "
628 			"forcing read-only mode");
629 		res = MS_RDONLY;
630 	}
631 	if (read_only)
632 		return res;
633 	if (!(sbi->s_mount_state & EXT2_VALID_FS))
634 		ext2_msg(sb, KERN_WARNING,
635 			"warning: mounting unchecked fs, "
636 			"running e2fsck is recommended");
637 	else if ((sbi->s_mount_state & EXT2_ERROR_FS))
638 		ext2_msg(sb, KERN_WARNING,
639 			"warning: mounting fs with errors, "
640 			"running e2fsck is recommended");
641 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
642 		 le16_to_cpu(es->s_mnt_count) >=
643 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
644 		ext2_msg(sb, KERN_WARNING,
645 			"warning: maximal mount count reached, "
646 			"running e2fsck is recommended");
647 	else if (le32_to_cpu(es->s_checkinterval) &&
648 		(le32_to_cpu(es->s_lastcheck) +
649 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
650 		ext2_msg(sb, KERN_WARNING,
651 			"warning: checktime reached, "
652 			"running e2fsck is recommended");
653 	if (!le16_to_cpu(es->s_max_mnt_count))
654 		es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
655 	le16_add_cpu(&es->s_mnt_count, 1);
656 	if (test_opt (sb, DEBUG))
657 		ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
658 			"bpg=%lu, ipg=%lu, mo=%04lx]",
659 			EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
660 			sbi->s_frag_size,
661 			sbi->s_groups_count,
662 			EXT2_BLOCKS_PER_GROUP(sb),
663 			EXT2_INODES_PER_GROUP(sb),
664 			sbi->s_mount_opt);
665 	return res;
666 }
667 
668 static int ext2_check_descriptors(struct super_block *sb)
669 {
670 	int i;
671 	struct ext2_sb_info *sbi = EXT2_SB(sb);
672 
673 	ext2_debug ("Checking group descriptors");
674 
675 	for (i = 0; i < sbi->s_groups_count; i++) {
676 		struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
677 		ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
678 		ext2_fsblk_t last_block;
679 
680 		if (i == sbi->s_groups_count - 1)
681 			last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
682 		else
683 			last_block = first_block +
684 				(EXT2_BLOCKS_PER_GROUP(sb) - 1);
685 
686 		if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
687 		    le32_to_cpu(gdp->bg_block_bitmap) > last_block)
688 		{
689 			ext2_error (sb, "ext2_check_descriptors",
690 				    "Block bitmap for group %d"
691 				    " not in group (block %lu)!",
692 				    i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
693 			return 0;
694 		}
695 		if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
696 		    le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
697 		{
698 			ext2_error (sb, "ext2_check_descriptors",
699 				    "Inode bitmap for group %d"
700 				    " not in group (block %lu)!",
701 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
702 			return 0;
703 		}
704 		if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
705 		    le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
706 		    last_block)
707 		{
708 			ext2_error (sb, "ext2_check_descriptors",
709 				    "Inode table for group %d"
710 				    " not in group (block %lu)!",
711 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
712 			return 0;
713 		}
714 	}
715 	return 1;
716 }
717 
718 /*
719  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
720  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
721  * We need to be 1 filesystem block less than the 2^32 sector limit.
722  */
723 static loff_t ext2_max_size(int bits)
724 {
725 	loff_t res = EXT2_NDIR_BLOCKS;
726 	int meta_blocks;
727 	loff_t upper_limit;
728 
729 	/* This is calculated to be the largest file size for a
730 	 * dense, file such that the total number of
731 	 * sectors in the file, including data and all indirect blocks,
732 	 * does not exceed 2^32 -1
733 	 * __u32 i_blocks representing the total number of
734 	 * 512 bytes blocks of the file
735 	 */
736 	upper_limit = (1LL << 32) - 1;
737 
738 	/* total blocks in file system block size */
739 	upper_limit >>= (bits - 9);
740 
741 
742 	/* indirect blocks */
743 	meta_blocks = 1;
744 	/* double indirect blocks */
745 	meta_blocks += 1 + (1LL << (bits-2));
746 	/* tripple indirect blocks */
747 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
748 
749 	upper_limit -= meta_blocks;
750 	upper_limit <<= bits;
751 
752 	res += 1LL << (bits-2);
753 	res += 1LL << (2*(bits-2));
754 	res += 1LL << (3*(bits-2));
755 	res <<= bits;
756 	if (res > upper_limit)
757 		res = upper_limit;
758 
759 	if (res > MAX_LFS_FILESIZE)
760 		res = MAX_LFS_FILESIZE;
761 
762 	return res;
763 }
764 
765 static unsigned long descriptor_loc(struct super_block *sb,
766 				    unsigned long logic_sb_block,
767 				    int nr)
768 {
769 	struct ext2_sb_info *sbi = EXT2_SB(sb);
770 	unsigned long bg, first_meta_bg;
771 	int has_super = 0;
772 
773 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
774 
775 	if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
776 	    nr < first_meta_bg)
777 		return (logic_sb_block + nr + 1);
778 	bg = sbi->s_desc_per_block * nr;
779 	if (ext2_bg_has_super(sb, bg))
780 		has_super = 1;
781 
782 	return ext2_group_first_block_no(sb, bg) + has_super;
783 }
784 
785 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
786 {
787 	struct buffer_head * bh;
788 	struct ext2_sb_info * sbi;
789 	struct ext2_super_block * es;
790 	struct inode *root;
791 	unsigned long block;
792 	unsigned long sb_block = get_sb_block(&data);
793 	unsigned long logic_sb_block;
794 	unsigned long offset = 0;
795 	unsigned long def_mount_opts;
796 	long ret = -EINVAL;
797 	int blocksize = BLOCK_SIZE;
798 	int db_count;
799 	int i, j;
800 	__le32 features;
801 	int err;
802 
803 	err = -ENOMEM;
804 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
805 	if (!sbi)
806 		goto failed;
807 
808 	sbi->s_blockgroup_lock =
809 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
810 	if (!sbi->s_blockgroup_lock) {
811 		kfree(sbi);
812 		goto failed;
813 	}
814 	sb->s_fs_info = sbi;
815 	sbi->s_sb_block = sb_block;
816 
817 	spin_lock_init(&sbi->s_lock);
818 
819 	/*
820 	 * See what the current blocksize for the device is, and
821 	 * use that as the blocksize.  Otherwise (or if the blocksize
822 	 * is smaller than the default) use the default.
823 	 * This is important for devices that have a hardware
824 	 * sectorsize that is larger than the default.
825 	 */
826 	blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
827 	if (!blocksize) {
828 		ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
829 		goto failed_sbi;
830 	}
831 
832 	/*
833 	 * If the superblock doesn't start on a hardware sector boundary,
834 	 * calculate the offset.
835 	 */
836 	if (blocksize != BLOCK_SIZE) {
837 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
838 		offset = (sb_block*BLOCK_SIZE) % blocksize;
839 	} else {
840 		logic_sb_block = sb_block;
841 	}
842 
843 	if (!(bh = sb_bread(sb, logic_sb_block))) {
844 		ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
845 		goto failed_sbi;
846 	}
847 	/*
848 	 * Note: s_es must be initialized as soon as possible because
849 	 *       some ext2 macro-instructions depend on its value
850 	 */
851 	es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
852 	sbi->s_es = es;
853 	sb->s_magic = le16_to_cpu(es->s_magic);
854 
855 	if (sb->s_magic != EXT2_SUPER_MAGIC)
856 		goto cantfind_ext2;
857 
858 	/* Set defaults before we parse the mount options */
859 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
860 	if (def_mount_opts & EXT2_DEFM_DEBUG)
861 		set_opt(sbi->s_mount_opt, DEBUG);
862 	if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
863 		set_opt(sbi->s_mount_opt, GRPID);
864 	if (def_mount_opts & EXT2_DEFM_UID16)
865 		set_opt(sbi->s_mount_opt, NO_UID32);
866 #ifdef CONFIG_EXT2_FS_XATTR
867 	if (def_mount_opts & EXT2_DEFM_XATTR_USER)
868 		set_opt(sbi->s_mount_opt, XATTR_USER);
869 #endif
870 #ifdef CONFIG_EXT2_FS_POSIX_ACL
871 	if (def_mount_opts & EXT2_DEFM_ACL)
872 		set_opt(sbi->s_mount_opt, POSIX_ACL);
873 #endif
874 
875 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
876 		set_opt(sbi->s_mount_opt, ERRORS_PANIC);
877 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
878 		set_opt(sbi->s_mount_opt, ERRORS_CONT);
879 	else
880 		set_opt(sbi->s_mount_opt, ERRORS_RO);
881 
882 	sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
883 	sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
884 
885 	set_opt(sbi->s_mount_opt, RESERVATION);
886 
887 	if (!parse_options((char *) data, sb))
888 		goto failed_mount;
889 
890 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
891 		((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
892 		 MS_POSIXACL : 0);
893 	sb->s_iflags |= SB_I_CGROUPWB;
894 
895 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
896 	    (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
897 	     EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
898 	     EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
899 		ext2_msg(sb, KERN_WARNING,
900 			"warning: feature flags set on rev 0 fs, "
901 			"running e2fsck is recommended");
902 	/*
903 	 * Check feature flags regardless of the revision level, since we
904 	 * previously didn't change the revision level when setting the flags,
905 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
906 	 */
907 	features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
908 	if (features) {
909 		ext2_msg(sb, KERN_ERR,	"error: couldn't mount because of "
910 		       "unsupported optional features (%x)",
911 			le32_to_cpu(features));
912 		goto failed_mount;
913 	}
914 	if (!(sb->s_flags & MS_RDONLY) &&
915 	    (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
916 		ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
917 		       "unsupported optional features (%x)",
918 		       le32_to_cpu(features));
919 		goto failed_mount;
920 	}
921 
922 	blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
923 
924 	if (sbi->s_mount_opt & EXT2_MOUNT_DAX) {
925 		err = bdev_dax_supported(sb, blocksize);
926 		if (err)
927 			goto failed_mount;
928 	}
929 
930 	/* If the blocksize doesn't match, re-read the thing.. */
931 	if (sb->s_blocksize != blocksize) {
932 		brelse(bh);
933 
934 		if (!sb_set_blocksize(sb, blocksize)) {
935 			ext2_msg(sb, KERN_ERR,
936 				"error: bad blocksize %d", blocksize);
937 			goto failed_sbi;
938 		}
939 
940 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
941 		offset = (sb_block*BLOCK_SIZE) % blocksize;
942 		bh = sb_bread(sb, logic_sb_block);
943 		if(!bh) {
944 			ext2_msg(sb, KERN_ERR, "error: couldn't read"
945 				"superblock on 2nd try");
946 			goto failed_sbi;
947 		}
948 		es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
949 		sbi->s_es = es;
950 		if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
951 			ext2_msg(sb, KERN_ERR, "error: magic mismatch");
952 			goto failed_mount;
953 		}
954 	}
955 
956 	sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
957 	sb->s_max_links = EXT2_LINK_MAX;
958 
959 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
960 		sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
961 		sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
962 	} else {
963 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
964 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
965 		if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
966 		    !is_power_of_2(sbi->s_inode_size) ||
967 		    (sbi->s_inode_size > blocksize)) {
968 			ext2_msg(sb, KERN_ERR,
969 				"error: unsupported inode size: %d",
970 				sbi->s_inode_size);
971 			goto failed_mount;
972 		}
973 	}
974 
975 	sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
976 				   le32_to_cpu(es->s_log_frag_size);
977 	if (sbi->s_frag_size == 0)
978 		goto cantfind_ext2;
979 	sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
980 
981 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
982 	sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
983 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
984 
985 	if (EXT2_INODE_SIZE(sb) == 0)
986 		goto cantfind_ext2;
987 	sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
988 	if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
989 		goto cantfind_ext2;
990 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
991 					sbi->s_inodes_per_block;
992 	sbi->s_desc_per_block = sb->s_blocksize /
993 					sizeof (struct ext2_group_desc);
994 	sbi->s_sbh = bh;
995 	sbi->s_mount_state = le16_to_cpu(es->s_state);
996 	sbi->s_addr_per_block_bits =
997 		ilog2 (EXT2_ADDR_PER_BLOCK(sb));
998 	sbi->s_desc_per_block_bits =
999 		ilog2 (EXT2_DESC_PER_BLOCK(sb));
1000 
1001 	if (sb->s_magic != EXT2_SUPER_MAGIC)
1002 		goto cantfind_ext2;
1003 
1004 	if (sb->s_blocksize != bh->b_size) {
1005 		if (!silent)
1006 			ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
1007 		goto failed_mount;
1008 	}
1009 
1010 	if (sb->s_blocksize != sbi->s_frag_size) {
1011 		ext2_msg(sb, KERN_ERR,
1012 			"error: fragsize %lu != blocksize %lu"
1013 			"(not supported yet)",
1014 			sbi->s_frag_size, sb->s_blocksize);
1015 		goto failed_mount;
1016 	}
1017 
1018 	if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
1019 		ext2_msg(sb, KERN_ERR,
1020 			"error: #blocks per group too big: %lu",
1021 			sbi->s_blocks_per_group);
1022 		goto failed_mount;
1023 	}
1024 	if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
1025 		ext2_msg(sb, KERN_ERR,
1026 			"error: #fragments per group too big: %lu",
1027 			sbi->s_frags_per_group);
1028 		goto failed_mount;
1029 	}
1030 	if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1031 		ext2_msg(sb, KERN_ERR,
1032 			"error: #inodes per group too big: %lu",
1033 			sbi->s_inodes_per_group);
1034 		goto failed_mount;
1035 	}
1036 
1037 	if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1038 		goto cantfind_ext2;
1039  	sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1040  				le32_to_cpu(es->s_first_data_block) - 1)
1041  					/ EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1042 	db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1043 		   EXT2_DESC_PER_BLOCK(sb);
1044 	sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
1045 	if (sbi->s_group_desc == NULL) {
1046 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1047 		goto failed_mount;
1048 	}
1049 	bgl_lock_init(sbi->s_blockgroup_lock);
1050 	sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1051 	if (!sbi->s_debts) {
1052 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1053 		goto failed_mount_group_desc;
1054 	}
1055 	for (i = 0; i < db_count; i++) {
1056 		block = descriptor_loc(sb, logic_sb_block, i);
1057 		sbi->s_group_desc[i] = sb_bread(sb, block);
1058 		if (!sbi->s_group_desc[i]) {
1059 			for (j = 0; j < i; j++)
1060 				brelse (sbi->s_group_desc[j]);
1061 			ext2_msg(sb, KERN_ERR,
1062 				"error: unable to read group descriptors");
1063 			goto failed_mount_group_desc;
1064 		}
1065 	}
1066 	if (!ext2_check_descriptors (sb)) {
1067 		ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1068 		goto failed_mount2;
1069 	}
1070 	sbi->s_gdb_count = db_count;
1071 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1072 	spin_lock_init(&sbi->s_next_gen_lock);
1073 
1074 	/* per fileystem reservation list head & lock */
1075 	spin_lock_init(&sbi->s_rsv_window_lock);
1076 	sbi->s_rsv_window_root = RB_ROOT;
1077 	/*
1078 	 * Add a single, static dummy reservation to the start of the
1079 	 * reservation window list --- it gives us a placeholder for
1080 	 * append-at-start-of-list which makes the allocation logic
1081 	 * _much_ simpler.
1082 	 */
1083 	sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1084 	sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1085 	sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1086 	sbi->s_rsv_window_head.rsv_goal_size = 0;
1087 	ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1088 
1089 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
1090 				ext2_count_free_blocks(sb), GFP_KERNEL);
1091 	if (!err) {
1092 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
1093 				ext2_count_free_inodes(sb), GFP_KERNEL);
1094 	}
1095 	if (!err) {
1096 		err = percpu_counter_init(&sbi->s_dirs_counter,
1097 				ext2_count_dirs(sb), GFP_KERNEL);
1098 	}
1099 	if (err) {
1100 		ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1101 		goto failed_mount3;
1102 	}
1103 
1104 #ifdef CONFIG_EXT2_FS_XATTR
1105 	sbi->s_mb_cache = ext2_xattr_create_cache();
1106 	if (!sbi->s_mb_cache) {
1107 		ext2_msg(sb, KERN_ERR, "Failed to create an mb_cache");
1108 		goto failed_mount3;
1109 	}
1110 #endif
1111 	/*
1112 	 * set up enough so that it can read an inode
1113 	 */
1114 	sb->s_op = &ext2_sops;
1115 	sb->s_export_op = &ext2_export_ops;
1116 	sb->s_xattr = ext2_xattr_handlers;
1117 
1118 #ifdef CONFIG_QUOTA
1119 	sb->dq_op = &dquot_operations;
1120 	sb->s_qcop = &dquot_quotactl_ops;
1121 	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1122 #endif
1123 
1124 	root = ext2_iget(sb, EXT2_ROOT_INO);
1125 	if (IS_ERR(root)) {
1126 		ret = PTR_ERR(root);
1127 		goto failed_mount3;
1128 	}
1129 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1130 		iput(root);
1131 		ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1132 		goto failed_mount3;
1133 	}
1134 
1135 	sb->s_root = d_make_root(root);
1136 	if (!sb->s_root) {
1137 		ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1138 		ret = -ENOMEM;
1139 		goto failed_mount3;
1140 	}
1141 	if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1142 		ext2_msg(sb, KERN_WARNING,
1143 			"warning: mounting ext3 filesystem as ext2");
1144 	if (ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY))
1145 		sb->s_flags |= MS_RDONLY;
1146 	ext2_write_super(sb);
1147 	return 0;
1148 
1149 cantfind_ext2:
1150 	if (!silent)
1151 		ext2_msg(sb, KERN_ERR,
1152 			"error: can't find an ext2 filesystem on dev %s.",
1153 			sb->s_id);
1154 	goto failed_mount;
1155 failed_mount3:
1156 	if (sbi->s_mb_cache)
1157 		ext2_xattr_destroy_cache(sbi->s_mb_cache);
1158 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
1159 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
1160 	percpu_counter_destroy(&sbi->s_dirs_counter);
1161 failed_mount2:
1162 	for (i = 0; i < db_count; i++)
1163 		brelse(sbi->s_group_desc[i]);
1164 failed_mount_group_desc:
1165 	kfree(sbi->s_group_desc);
1166 	kfree(sbi->s_debts);
1167 failed_mount:
1168 	brelse(bh);
1169 failed_sbi:
1170 	sb->s_fs_info = NULL;
1171 	kfree(sbi->s_blockgroup_lock);
1172 	kfree(sbi);
1173 failed:
1174 	return ret;
1175 }
1176 
1177 static void ext2_clear_super_error(struct super_block *sb)
1178 {
1179 	struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1180 
1181 	if (buffer_write_io_error(sbh)) {
1182 		/*
1183 		 * Oh, dear.  A previous attempt to write the
1184 		 * superblock failed.  This could happen because the
1185 		 * USB device was yanked out.  Or it could happen to
1186 		 * be a transient write error and maybe the block will
1187 		 * be remapped.  Nothing we can do but to retry the
1188 		 * write and hope for the best.
1189 		 */
1190 		ext2_msg(sb, KERN_ERR,
1191 		       "previous I/O error to superblock detected\n");
1192 		clear_buffer_write_io_error(sbh);
1193 		set_buffer_uptodate(sbh);
1194 	}
1195 }
1196 
1197 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1198 			    int wait)
1199 {
1200 	ext2_clear_super_error(sb);
1201 	spin_lock(&EXT2_SB(sb)->s_lock);
1202 	es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1203 	es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1204 	es->s_wtime = cpu_to_le32(get_seconds());
1205 	/* unlock before we do IO */
1206 	spin_unlock(&EXT2_SB(sb)->s_lock);
1207 	mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1208 	if (wait)
1209 		sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1210 }
1211 
1212 /*
1213  * In the second extended file system, it is not necessary to
1214  * write the super block since we use a mapping of the
1215  * disk super block in a buffer.
1216  *
1217  * However, this function is still used to set the fs valid
1218  * flags to 0.  We need to set this flag to 0 since the fs
1219  * may have been checked while mounted and e2fsck may have
1220  * set s_state to EXT2_VALID_FS after some corrections.
1221  */
1222 static int ext2_sync_fs(struct super_block *sb, int wait)
1223 {
1224 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1225 	struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1226 
1227 	/*
1228 	 * Write quota structures to quota file, sync_blockdev() will write
1229 	 * them to disk later
1230 	 */
1231 	dquot_writeback_dquots(sb, -1);
1232 
1233 	spin_lock(&sbi->s_lock);
1234 	if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1235 		ext2_debug("setting valid to 0\n");
1236 		es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1237 	}
1238 	spin_unlock(&sbi->s_lock);
1239 	ext2_sync_super(sb, es, wait);
1240 	return 0;
1241 }
1242 
1243 static int ext2_freeze(struct super_block *sb)
1244 {
1245 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1246 
1247 	/*
1248 	 * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1249 	 * because we have unattached inodes and thus filesystem is not fully
1250 	 * consistent.
1251 	 */
1252 	if (atomic_long_read(&sb->s_remove_count)) {
1253 		ext2_sync_fs(sb, 1);
1254 		return 0;
1255 	}
1256 	/* Set EXT2_FS_VALID flag */
1257 	spin_lock(&sbi->s_lock);
1258 	sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1259 	spin_unlock(&sbi->s_lock);
1260 	ext2_sync_super(sb, sbi->s_es, 1);
1261 
1262 	return 0;
1263 }
1264 
1265 static int ext2_unfreeze(struct super_block *sb)
1266 {
1267 	/* Just write sb to clear EXT2_VALID_FS flag */
1268 	ext2_write_super(sb);
1269 
1270 	return 0;
1271 }
1272 
1273 void ext2_write_super(struct super_block *sb)
1274 {
1275 	if (!(sb->s_flags & MS_RDONLY))
1276 		ext2_sync_fs(sb, 1);
1277 }
1278 
1279 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1280 {
1281 	struct ext2_sb_info * sbi = EXT2_SB(sb);
1282 	struct ext2_super_block * es;
1283 	struct ext2_mount_options old_opts;
1284 	unsigned long old_sb_flags;
1285 	int err;
1286 
1287 	sync_filesystem(sb);
1288 	spin_lock(&sbi->s_lock);
1289 
1290 	/* Store the old options */
1291 	old_sb_flags = sb->s_flags;
1292 	old_opts.s_mount_opt = sbi->s_mount_opt;
1293 	old_opts.s_resuid = sbi->s_resuid;
1294 	old_opts.s_resgid = sbi->s_resgid;
1295 
1296 	/*
1297 	 * Allow the "check" option to be passed as a remount option.
1298 	 */
1299 	if (!parse_options(data, sb)) {
1300 		err = -EINVAL;
1301 		goto restore_opts;
1302 	}
1303 
1304 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1305 		((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1306 
1307 	es = sbi->s_es;
1308 	if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT2_MOUNT_DAX) {
1309 		ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1310 			 "dax flag with busy inodes while remounting");
1311 		sbi->s_mount_opt ^= EXT2_MOUNT_DAX;
1312 	}
1313 	if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) {
1314 		spin_unlock(&sbi->s_lock);
1315 		return 0;
1316 	}
1317 	if (*flags & MS_RDONLY) {
1318 		if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1319 		    !(sbi->s_mount_state & EXT2_VALID_FS)) {
1320 			spin_unlock(&sbi->s_lock);
1321 			return 0;
1322 		}
1323 
1324 		/*
1325 		 * OK, we are remounting a valid rw partition rdonly, so set
1326 		 * the rdonly flag and then mark the partition as valid again.
1327 		 */
1328 		es->s_state = cpu_to_le16(sbi->s_mount_state);
1329 		es->s_mtime = cpu_to_le32(get_seconds());
1330 		spin_unlock(&sbi->s_lock);
1331 
1332 		err = dquot_suspend(sb, -1);
1333 		if (err < 0) {
1334 			spin_lock(&sbi->s_lock);
1335 			goto restore_opts;
1336 		}
1337 
1338 		ext2_sync_super(sb, es, 1);
1339 	} else {
1340 		__le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1341 					       ~EXT2_FEATURE_RO_COMPAT_SUPP);
1342 		if (ret) {
1343 			ext2_msg(sb, KERN_WARNING,
1344 				"warning: couldn't remount RDWR because of "
1345 				"unsupported optional features (%x).",
1346 				le32_to_cpu(ret));
1347 			err = -EROFS;
1348 			goto restore_opts;
1349 		}
1350 		/*
1351 		 * Mounting a RDONLY partition read-write, so reread and
1352 		 * store the current valid flag.  (It may have been changed
1353 		 * by e2fsck since we originally mounted the partition.)
1354 		 */
1355 		sbi->s_mount_state = le16_to_cpu(es->s_state);
1356 		if (!ext2_setup_super (sb, es, 0))
1357 			sb->s_flags &= ~MS_RDONLY;
1358 		spin_unlock(&sbi->s_lock);
1359 
1360 		ext2_write_super(sb);
1361 
1362 		dquot_resume(sb, -1);
1363 	}
1364 
1365 	return 0;
1366 restore_opts:
1367 	sbi->s_mount_opt = old_opts.s_mount_opt;
1368 	sbi->s_resuid = old_opts.s_resuid;
1369 	sbi->s_resgid = old_opts.s_resgid;
1370 	sb->s_flags = old_sb_flags;
1371 	spin_unlock(&sbi->s_lock);
1372 	return err;
1373 }
1374 
1375 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1376 {
1377 	struct super_block *sb = dentry->d_sb;
1378 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1379 	struct ext2_super_block *es = sbi->s_es;
1380 	u64 fsid;
1381 
1382 	spin_lock(&sbi->s_lock);
1383 
1384 	if (test_opt (sb, MINIX_DF))
1385 		sbi->s_overhead_last = 0;
1386 	else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1387 		unsigned long i, overhead = 0;
1388 		smp_rmb();
1389 
1390 		/*
1391 		 * Compute the overhead (FS structures). This is constant
1392 		 * for a given filesystem unless the number of block groups
1393 		 * changes so we cache the previous value until it does.
1394 		 */
1395 
1396 		/*
1397 		 * All of the blocks before first_data_block are
1398 		 * overhead
1399 		 */
1400 		overhead = le32_to_cpu(es->s_first_data_block);
1401 
1402 		/*
1403 		 * Add the overhead attributed to the superblock and
1404 		 * block group descriptors.  If the sparse superblocks
1405 		 * feature is turned on, then not all groups have this.
1406 		 */
1407 		for (i = 0; i < sbi->s_groups_count; i++)
1408 			overhead += ext2_bg_has_super(sb, i) +
1409 				ext2_bg_num_gdb(sb, i);
1410 
1411 		/*
1412 		 * Every block group has an inode bitmap, a block
1413 		 * bitmap, and an inode table.
1414 		 */
1415 		overhead += (sbi->s_groups_count *
1416 			     (2 + sbi->s_itb_per_group));
1417 		sbi->s_overhead_last = overhead;
1418 		smp_wmb();
1419 		sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1420 	}
1421 
1422 	buf->f_type = EXT2_SUPER_MAGIC;
1423 	buf->f_bsize = sb->s_blocksize;
1424 	buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1425 	buf->f_bfree = ext2_count_free_blocks(sb);
1426 	es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1427 	buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1428 	if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1429 		buf->f_bavail = 0;
1430 	buf->f_files = le32_to_cpu(es->s_inodes_count);
1431 	buf->f_ffree = ext2_count_free_inodes(sb);
1432 	es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1433 	buf->f_namelen = EXT2_NAME_LEN;
1434 	fsid = le64_to_cpup((void *)es->s_uuid) ^
1435 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1436 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1437 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1438 	spin_unlock(&sbi->s_lock);
1439 	return 0;
1440 }
1441 
1442 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1443 	int flags, const char *dev_name, void *data)
1444 {
1445 	return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1446 }
1447 
1448 #ifdef CONFIG_QUOTA
1449 
1450 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1451  * acquiring the locks... As quota files are never truncated and quota code
1452  * itself serializes the operations (and no one else should touch the files)
1453  * we don't have to be afraid of races */
1454 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1455 			       size_t len, loff_t off)
1456 {
1457 	struct inode *inode = sb_dqopt(sb)->files[type];
1458 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1459 	int err = 0;
1460 	int offset = off & (sb->s_blocksize - 1);
1461 	int tocopy;
1462 	size_t toread;
1463 	struct buffer_head tmp_bh;
1464 	struct buffer_head *bh;
1465 	loff_t i_size = i_size_read(inode);
1466 
1467 	if (off > i_size)
1468 		return 0;
1469 	if (off+len > i_size)
1470 		len = i_size-off;
1471 	toread = len;
1472 	while (toread > 0) {
1473 		tocopy = sb->s_blocksize - offset < toread ?
1474 				sb->s_blocksize - offset : toread;
1475 
1476 		tmp_bh.b_state = 0;
1477 		tmp_bh.b_size = sb->s_blocksize;
1478 		err = ext2_get_block(inode, blk, &tmp_bh, 0);
1479 		if (err < 0)
1480 			return err;
1481 		if (!buffer_mapped(&tmp_bh))	/* A hole? */
1482 			memset(data, 0, tocopy);
1483 		else {
1484 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1485 			if (!bh)
1486 				return -EIO;
1487 			memcpy(data, bh->b_data+offset, tocopy);
1488 			brelse(bh);
1489 		}
1490 		offset = 0;
1491 		toread -= tocopy;
1492 		data += tocopy;
1493 		blk++;
1494 	}
1495 	return len;
1496 }
1497 
1498 /* Write to quotafile */
1499 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1500 				const char *data, size_t len, loff_t off)
1501 {
1502 	struct inode *inode = sb_dqopt(sb)->files[type];
1503 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1504 	int err = 0;
1505 	int offset = off & (sb->s_blocksize - 1);
1506 	int tocopy;
1507 	size_t towrite = len;
1508 	struct buffer_head tmp_bh;
1509 	struct buffer_head *bh;
1510 
1511 	while (towrite > 0) {
1512 		tocopy = sb->s_blocksize - offset < towrite ?
1513 				sb->s_blocksize - offset : towrite;
1514 
1515 		tmp_bh.b_state = 0;
1516 		tmp_bh.b_size = sb->s_blocksize;
1517 		err = ext2_get_block(inode, blk, &tmp_bh, 1);
1518 		if (err < 0)
1519 			goto out;
1520 		if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1521 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1522 		else
1523 			bh = sb_getblk(sb, tmp_bh.b_blocknr);
1524 		if (unlikely(!bh)) {
1525 			err = -EIO;
1526 			goto out;
1527 		}
1528 		lock_buffer(bh);
1529 		memcpy(bh->b_data+offset, data, tocopy);
1530 		flush_dcache_page(bh->b_page);
1531 		set_buffer_uptodate(bh);
1532 		mark_buffer_dirty(bh);
1533 		unlock_buffer(bh);
1534 		brelse(bh);
1535 		offset = 0;
1536 		towrite -= tocopy;
1537 		data += tocopy;
1538 		blk++;
1539 	}
1540 out:
1541 	if (len == towrite)
1542 		return err;
1543 	if (inode->i_size < off+len-towrite)
1544 		i_size_write(inode, off+len-towrite);
1545 	inode->i_version++;
1546 	inode->i_mtime = inode->i_ctime = current_time(inode);
1547 	mark_inode_dirty(inode);
1548 	return len - towrite;
1549 }
1550 
1551 #endif
1552 
1553 static struct file_system_type ext2_fs_type = {
1554 	.owner		= THIS_MODULE,
1555 	.name		= "ext2",
1556 	.mount		= ext2_mount,
1557 	.kill_sb	= kill_block_super,
1558 	.fs_flags	= FS_REQUIRES_DEV,
1559 };
1560 MODULE_ALIAS_FS("ext2");
1561 
1562 static int __init init_ext2_fs(void)
1563 {
1564 	int err;
1565 
1566 	err = init_inodecache();
1567 	if (err)
1568 		return err;
1569         err = register_filesystem(&ext2_fs_type);
1570 	if (err)
1571 		goto out;
1572 	return 0;
1573 out:
1574 	destroy_inodecache();
1575 	return err;
1576 }
1577 
1578 static void __exit exit_ext2_fs(void)
1579 {
1580 	unregister_filesystem(&ext2_fs_type);
1581 	destroy_inodecache();
1582 }
1583 
1584 MODULE_AUTHOR("Remy Card and others");
1585 MODULE_DESCRIPTION("Second Extended Filesystem");
1586 MODULE_LICENSE("GPL");
1587 module_init(init_ext2_fs)
1588 module_exit(exit_ext2_fs)
1589