xref: /openbmc/linux/fs/ext2/super.c (revision f7d84fa7)
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 <linux/dax.h>
36 #include "ext2.h"
37 #include "xattr.h"
38 #include "acl.h"
39 
40 static void ext2_write_super(struct super_block *sb);
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 #ifdef CONFIG_QUOTA
127 static int ext2_quota_off(struct super_block *sb, int type);
128 
129 static void ext2_quota_off_umount(struct super_block *sb)
130 {
131 	int type;
132 
133 	for (type = 0; type < MAXQUOTAS; type++)
134 		ext2_quota_off(sb, type);
135 }
136 #else
137 static inline void ext2_quota_off_umount(struct super_block *sb)
138 {
139 }
140 #endif
141 
142 static void ext2_put_super (struct super_block * sb)
143 {
144 	int db_count;
145 	int i;
146 	struct ext2_sb_info *sbi = EXT2_SB(sb);
147 
148 	ext2_quota_off_umount(sb);
149 
150 	if (sbi->s_mb_cache) {
151 		ext2_xattr_destroy_cache(sbi->s_mb_cache);
152 		sbi->s_mb_cache = NULL;
153 	}
154 	if (!(sb->s_flags & MS_RDONLY)) {
155 		struct ext2_super_block *es = sbi->s_es;
156 
157 		spin_lock(&sbi->s_lock);
158 		es->s_state = cpu_to_le16(sbi->s_mount_state);
159 		spin_unlock(&sbi->s_lock);
160 		ext2_sync_super(sb, es, 1);
161 	}
162 	db_count = sbi->s_gdb_count;
163 	for (i = 0; i < db_count; i++)
164 		if (sbi->s_group_desc[i])
165 			brelse (sbi->s_group_desc[i]);
166 	kfree(sbi->s_group_desc);
167 	kfree(sbi->s_debts);
168 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
169 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
170 	percpu_counter_destroy(&sbi->s_dirs_counter);
171 	brelse (sbi->s_sbh);
172 	sb->s_fs_info = NULL;
173 	kfree(sbi->s_blockgroup_lock);
174 	kfree(sbi);
175 }
176 
177 static struct kmem_cache * ext2_inode_cachep;
178 
179 static struct inode *ext2_alloc_inode(struct super_block *sb)
180 {
181 	struct ext2_inode_info *ei;
182 	ei = kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
183 	if (!ei)
184 		return NULL;
185 	ei->i_block_alloc_info = NULL;
186 	ei->vfs_inode.i_version = 1;
187 #ifdef CONFIG_QUOTA
188 	memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
189 #endif
190 
191 	return &ei->vfs_inode;
192 }
193 
194 static void ext2_i_callback(struct rcu_head *head)
195 {
196 	struct inode *inode = container_of(head, struct inode, i_rcu);
197 	kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
198 }
199 
200 static void ext2_destroy_inode(struct inode *inode)
201 {
202 	call_rcu(&inode->i_rcu, ext2_i_callback);
203 }
204 
205 static void init_once(void *foo)
206 {
207 	struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
208 
209 	rwlock_init(&ei->i_meta_lock);
210 #ifdef CONFIG_EXT2_FS_XATTR
211 	init_rwsem(&ei->xattr_sem);
212 #endif
213 	mutex_init(&ei->truncate_mutex);
214 #ifdef CONFIG_FS_DAX
215 	init_rwsem(&ei->dax_sem);
216 #endif
217 	inode_init_once(&ei->vfs_inode);
218 }
219 
220 static int __init init_inodecache(void)
221 {
222 	ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
223 					     sizeof(struct ext2_inode_info),
224 					     0, (SLAB_RECLAIM_ACCOUNT|
225 						SLAB_MEM_SPREAD|SLAB_ACCOUNT),
226 					     init_once);
227 	if (ext2_inode_cachep == NULL)
228 		return -ENOMEM;
229 	return 0;
230 }
231 
232 static void destroy_inodecache(void)
233 {
234 	/*
235 	 * Make sure all delayed rcu free inodes are flushed before we
236 	 * destroy cache.
237 	 */
238 	rcu_barrier();
239 	kmem_cache_destroy(ext2_inode_cachep);
240 }
241 
242 static int ext2_show_options(struct seq_file *seq, struct dentry *root)
243 {
244 	struct super_block *sb = root->d_sb;
245 	struct ext2_sb_info *sbi = EXT2_SB(sb);
246 	struct ext2_super_block *es = sbi->s_es;
247 	unsigned long def_mount_opts;
248 
249 	spin_lock(&sbi->s_lock);
250 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
251 
252 	if (sbi->s_sb_block != 1)
253 		seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
254 	if (test_opt(sb, MINIX_DF))
255 		seq_puts(seq, ",minixdf");
256 	if (test_opt(sb, GRPID))
257 		seq_puts(seq, ",grpid");
258 	if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
259 		seq_puts(seq, ",nogrpid");
260 	if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT2_DEF_RESUID)) ||
261 	    le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
262 		seq_printf(seq, ",resuid=%u",
263 				from_kuid_munged(&init_user_ns, sbi->s_resuid));
264 	}
265 	if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT2_DEF_RESGID)) ||
266 	    le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
267 		seq_printf(seq, ",resgid=%u",
268 				from_kgid_munged(&init_user_ns, sbi->s_resgid));
269 	}
270 	if (test_opt(sb, ERRORS_RO)) {
271 		int def_errors = le16_to_cpu(es->s_errors);
272 
273 		if (def_errors == EXT2_ERRORS_PANIC ||
274 		    def_errors == EXT2_ERRORS_CONTINUE) {
275 			seq_puts(seq, ",errors=remount-ro");
276 		}
277 	}
278 	if (test_opt(sb, ERRORS_CONT))
279 		seq_puts(seq, ",errors=continue");
280 	if (test_opt(sb, ERRORS_PANIC))
281 		seq_puts(seq, ",errors=panic");
282 	if (test_opt(sb, NO_UID32))
283 		seq_puts(seq, ",nouid32");
284 	if (test_opt(sb, DEBUG))
285 		seq_puts(seq, ",debug");
286 	if (test_opt(sb, OLDALLOC))
287 		seq_puts(seq, ",oldalloc");
288 
289 #ifdef CONFIG_EXT2_FS_XATTR
290 	if (test_opt(sb, XATTR_USER))
291 		seq_puts(seq, ",user_xattr");
292 	if (!test_opt(sb, XATTR_USER) &&
293 	    (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
294 		seq_puts(seq, ",nouser_xattr");
295 	}
296 #endif
297 
298 #ifdef CONFIG_EXT2_FS_POSIX_ACL
299 	if (test_opt(sb, POSIX_ACL))
300 		seq_puts(seq, ",acl");
301 	if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
302 		seq_puts(seq, ",noacl");
303 #endif
304 
305 	if (test_opt(sb, NOBH))
306 		seq_puts(seq, ",nobh");
307 
308 #if defined(CONFIG_QUOTA)
309 	if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
310 		seq_puts(seq, ",usrquota");
311 
312 	if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
313 		seq_puts(seq, ",grpquota");
314 #endif
315 
316 #ifdef CONFIG_FS_DAX
317 	if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
318 		seq_puts(seq, ",xip");
319 	if (sbi->s_mount_opt & EXT2_MOUNT_DAX)
320 		seq_puts(seq, ",dax");
321 #endif
322 
323 	if (!test_opt(sb, RESERVATION))
324 		seq_puts(seq, ",noreservation");
325 
326 	spin_unlock(&sbi->s_lock);
327 	return 0;
328 }
329 
330 #ifdef CONFIG_QUOTA
331 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
332 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
333 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
334 			 const struct path *path);
335 static struct dquot **ext2_get_dquots(struct inode *inode)
336 {
337 	return EXT2_I(inode)->i_dquot;
338 }
339 
340 static const struct quotactl_ops ext2_quotactl_ops = {
341 	.quota_on	= ext2_quota_on,
342 	.quota_off	= ext2_quota_off,
343 	.quota_sync	= dquot_quota_sync,
344 	.get_state	= dquot_get_state,
345 	.set_info	= dquot_set_dqinfo,
346 	.get_dqblk	= dquot_get_dqblk,
347 	.set_dqblk	= dquot_set_dqblk,
348 	.get_nextdqblk	= dquot_get_next_dqblk,
349 };
350 #endif
351 
352 static const struct super_operations ext2_sops = {
353 	.alloc_inode	= ext2_alloc_inode,
354 	.destroy_inode	= ext2_destroy_inode,
355 	.write_inode	= ext2_write_inode,
356 	.evict_inode	= ext2_evict_inode,
357 	.put_super	= ext2_put_super,
358 	.sync_fs	= ext2_sync_fs,
359 	.freeze_fs	= ext2_freeze,
360 	.unfreeze_fs	= ext2_unfreeze,
361 	.statfs		= ext2_statfs,
362 	.remount_fs	= ext2_remount,
363 	.show_options	= ext2_show_options,
364 #ifdef CONFIG_QUOTA
365 	.quota_read	= ext2_quota_read,
366 	.quota_write	= ext2_quota_write,
367 	.get_dquots	= ext2_get_dquots,
368 #endif
369 };
370 
371 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
372 		u64 ino, u32 generation)
373 {
374 	struct inode *inode;
375 
376 	if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
377 		return ERR_PTR(-ESTALE);
378 	if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
379 		return ERR_PTR(-ESTALE);
380 
381 	/*
382 	 * ext2_iget isn't quite right if the inode is currently unallocated!
383 	 * However ext2_iget currently does appropriate checks to handle stale
384 	 * inodes so everything is OK.
385 	 */
386 	inode = ext2_iget(sb, ino);
387 	if (IS_ERR(inode))
388 		return ERR_CAST(inode);
389 	if (generation && inode->i_generation != generation) {
390 		/* we didn't find the right inode.. */
391 		iput(inode);
392 		return ERR_PTR(-ESTALE);
393 	}
394 	return inode;
395 }
396 
397 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
398 		int fh_len, int fh_type)
399 {
400 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
401 				    ext2_nfs_get_inode);
402 }
403 
404 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
405 		int fh_len, int fh_type)
406 {
407 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
408 				    ext2_nfs_get_inode);
409 }
410 
411 static const struct export_operations ext2_export_ops = {
412 	.fh_to_dentry = ext2_fh_to_dentry,
413 	.fh_to_parent = ext2_fh_to_parent,
414 	.get_parent = ext2_get_parent,
415 };
416 
417 static unsigned long get_sb_block(void **data)
418 {
419 	unsigned long 	sb_block;
420 	char 		*options = (char *) *data;
421 
422 	if (!options || strncmp(options, "sb=", 3) != 0)
423 		return 1;	/* Default location */
424 	options += 3;
425 	sb_block = simple_strtoul(options, &options, 0);
426 	if (*options && *options != ',') {
427 		printk("EXT2-fs: Invalid sb specification: %s\n",
428 		       (char *) *data);
429 		return 1;
430 	}
431 	if (*options == ',')
432 		options++;
433 	*data = (void *) options;
434 	return sb_block;
435 }
436 
437 enum {
438 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
439 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
440 	Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
441 	Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
442 	Opt_acl, Opt_noacl, Opt_xip, Opt_dax, Opt_ignore, Opt_err, Opt_quota,
443 	Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
444 };
445 
446 static const match_table_t tokens = {
447 	{Opt_bsd_df, "bsddf"},
448 	{Opt_minix_df, "minixdf"},
449 	{Opt_grpid, "grpid"},
450 	{Opt_grpid, "bsdgroups"},
451 	{Opt_nogrpid, "nogrpid"},
452 	{Opt_nogrpid, "sysvgroups"},
453 	{Opt_resgid, "resgid=%u"},
454 	{Opt_resuid, "resuid=%u"},
455 	{Opt_sb, "sb=%u"},
456 	{Opt_err_cont, "errors=continue"},
457 	{Opt_err_panic, "errors=panic"},
458 	{Opt_err_ro, "errors=remount-ro"},
459 	{Opt_nouid32, "nouid32"},
460 	{Opt_nocheck, "check=none"},
461 	{Opt_nocheck, "nocheck"},
462 	{Opt_debug, "debug"},
463 	{Opt_oldalloc, "oldalloc"},
464 	{Opt_orlov, "orlov"},
465 	{Opt_nobh, "nobh"},
466 	{Opt_user_xattr, "user_xattr"},
467 	{Opt_nouser_xattr, "nouser_xattr"},
468 	{Opt_acl, "acl"},
469 	{Opt_noacl, "noacl"},
470 	{Opt_xip, "xip"},
471 	{Opt_dax, "dax"},
472 	{Opt_grpquota, "grpquota"},
473 	{Opt_ignore, "noquota"},
474 	{Opt_quota, "quota"},
475 	{Opt_usrquota, "usrquota"},
476 	{Opt_reservation, "reservation"},
477 	{Opt_noreservation, "noreservation"},
478 	{Opt_err, NULL}
479 };
480 
481 static int parse_options(char *options, struct super_block *sb)
482 {
483 	char *p;
484 	struct ext2_sb_info *sbi = EXT2_SB(sb);
485 	substring_t args[MAX_OPT_ARGS];
486 	int option;
487 	kuid_t uid;
488 	kgid_t gid;
489 
490 	if (!options)
491 		return 1;
492 
493 	while ((p = strsep (&options, ",")) != NULL) {
494 		int token;
495 		if (!*p)
496 			continue;
497 
498 		token = match_token(p, tokens, args);
499 		switch (token) {
500 		case Opt_bsd_df:
501 			clear_opt (sbi->s_mount_opt, MINIX_DF);
502 			break;
503 		case Opt_minix_df:
504 			set_opt (sbi->s_mount_opt, MINIX_DF);
505 			break;
506 		case Opt_grpid:
507 			set_opt (sbi->s_mount_opt, GRPID);
508 			break;
509 		case Opt_nogrpid:
510 			clear_opt (sbi->s_mount_opt, GRPID);
511 			break;
512 		case Opt_resuid:
513 			if (match_int(&args[0], &option))
514 				return 0;
515 			uid = make_kuid(current_user_ns(), option);
516 			if (!uid_valid(uid)) {
517 				ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
518 				return 0;
519 
520 			}
521 			sbi->s_resuid = uid;
522 			break;
523 		case Opt_resgid:
524 			if (match_int(&args[0], &option))
525 				return 0;
526 			gid = make_kgid(current_user_ns(), option);
527 			if (!gid_valid(gid)) {
528 				ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
529 				return 0;
530 			}
531 			sbi->s_resgid = gid;
532 			break;
533 		case Opt_sb:
534 			/* handled by get_sb_block() instead of here */
535 			/* *sb_block = match_int(&args[0]); */
536 			break;
537 		case Opt_err_panic:
538 			clear_opt (sbi->s_mount_opt, ERRORS_CONT);
539 			clear_opt (sbi->s_mount_opt, ERRORS_RO);
540 			set_opt (sbi->s_mount_opt, ERRORS_PANIC);
541 			break;
542 		case Opt_err_ro:
543 			clear_opt (sbi->s_mount_opt, ERRORS_CONT);
544 			clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
545 			set_opt (sbi->s_mount_opt, ERRORS_RO);
546 			break;
547 		case Opt_err_cont:
548 			clear_opt (sbi->s_mount_opt, ERRORS_RO);
549 			clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
550 			set_opt (sbi->s_mount_opt, ERRORS_CONT);
551 			break;
552 		case Opt_nouid32:
553 			set_opt (sbi->s_mount_opt, NO_UID32);
554 			break;
555 		case Opt_nocheck:
556 			clear_opt (sbi->s_mount_opt, CHECK);
557 			break;
558 		case Opt_debug:
559 			set_opt (sbi->s_mount_opt, DEBUG);
560 			break;
561 		case Opt_oldalloc:
562 			set_opt (sbi->s_mount_opt, OLDALLOC);
563 			break;
564 		case Opt_orlov:
565 			clear_opt (sbi->s_mount_opt, OLDALLOC);
566 			break;
567 		case Opt_nobh:
568 			set_opt (sbi->s_mount_opt, NOBH);
569 			break;
570 #ifdef CONFIG_EXT2_FS_XATTR
571 		case Opt_user_xattr:
572 			set_opt (sbi->s_mount_opt, XATTR_USER);
573 			break;
574 		case Opt_nouser_xattr:
575 			clear_opt (sbi->s_mount_opt, XATTR_USER);
576 			break;
577 #else
578 		case Opt_user_xattr:
579 		case Opt_nouser_xattr:
580 			ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
581 				"not supported");
582 			break;
583 #endif
584 #ifdef CONFIG_EXT2_FS_POSIX_ACL
585 		case Opt_acl:
586 			set_opt(sbi->s_mount_opt, POSIX_ACL);
587 			break;
588 		case Opt_noacl:
589 			clear_opt(sbi->s_mount_opt, POSIX_ACL);
590 			break;
591 #else
592 		case Opt_acl:
593 		case Opt_noacl:
594 			ext2_msg(sb, KERN_INFO,
595 				"(no)acl options not supported");
596 			break;
597 #endif
598 		case Opt_xip:
599 			ext2_msg(sb, KERN_INFO, "use dax instead of xip");
600 			set_opt(sbi->s_mount_opt, XIP);
601 			/* Fall through */
602 		case Opt_dax:
603 #ifdef CONFIG_FS_DAX
604 			ext2_msg(sb, KERN_WARNING,
605 		"DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
606 			set_opt(sbi->s_mount_opt, DAX);
607 #else
608 			ext2_msg(sb, KERN_INFO, "dax option not supported");
609 #endif
610 			break;
611 
612 #if defined(CONFIG_QUOTA)
613 		case Opt_quota:
614 		case Opt_usrquota:
615 			set_opt(sbi->s_mount_opt, USRQUOTA);
616 			break;
617 
618 		case Opt_grpquota:
619 			set_opt(sbi->s_mount_opt, GRPQUOTA);
620 			break;
621 #else
622 		case Opt_quota:
623 		case Opt_usrquota:
624 		case Opt_grpquota:
625 			ext2_msg(sb, KERN_INFO,
626 				"quota operations not supported");
627 			break;
628 #endif
629 
630 		case Opt_reservation:
631 			set_opt(sbi->s_mount_opt, RESERVATION);
632 			ext2_msg(sb, KERN_INFO, "reservations ON");
633 			break;
634 		case Opt_noreservation:
635 			clear_opt(sbi->s_mount_opt, RESERVATION);
636 			ext2_msg(sb, KERN_INFO, "reservations OFF");
637 			break;
638 		case Opt_ignore:
639 			break;
640 		default:
641 			return 0;
642 		}
643 	}
644 	return 1;
645 }
646 
647 static int ext2_setup_super (struct super_block * sb,
648 			      struct ext2_super_block * es,
649 			      int read_only)
650 {
651 	int res = 0;
652 	struct ext2_sb_info *sbi = EXT2_SB(sb);
653 
654 	if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
655 		ext2_msg(sb, KERN_ERR,
656 			"error: revision level too high, "
657 			"forcing read-only mode");
658 		res = MS_RDONLY;
659 	}
660 	if (read_only)
661 		return res;
662 	if (!(sbi->s_mount_state & EXT2_VALID_FS))
663 		ext2_msg(sb, KERN_WARNING,
664 			"warning: mounting unchecked fs, "
665 			"running e2fsck is recommended");
666 	else if ((sbi->s_mount_state & EXT2_ERROR_FS))
667 		ext2_msg(sb, KERN_WARNING,
668 			"warning: mounting fs with errors, "
669 			"running e2fsck is recommended");
670 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
671 		 le16_to_cpu(es->s_mnt_count) >=
672 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
673 		ext2_msg(sb, KERN_WARNING,
674 			"warning: maximal mount count reached, "
675 			"running e2fsck is recommended");
676 	else if (le32_to_cpu(es->s_checkinterval) &&
677 		(le32_to_cpu(es->s_lastcheck) +
678 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
679 		ext2_msg(sb, KERN_WARNING,
680 			"warning: checktime reached, "
681 			"running e2fsck is recommended");
682 	if (!le16_to_cpu(es->s_max_mnt_count))
683 		es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
684 	le16_add_cpu(&es->s_mnt_count, 1);
685 	if (test_opt (sb, DEBUG))
686 		ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
687 			"bpg=%lu, ipg=%lu, mo=%04lx]",
688 			EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
689 			sbi->s_frag_size,
690 			sbi->s_groups_count,
691 			EXT2_BLOCKS_PER_GROUP(sb),
692 			EXT2_INODES_PER_GROUP(sb),
693 			sbi->s_mount_opt);
694 	return res;
695 }
696 
697 static int ext2_check_descriptors(struct super_block *sb)
698 {
699 	int i;
700 	struct ext2_sb_info *sbi = EXT2_SB(sb);
701 
702 	ext2_debug ("Checking group descriptors");
703 
704 	for (i = 0; i < sbi->s_groups_count; i++) {
705 		struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
706 		ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
707 		ext2_fsblk_t last_block;
708 
709 		if (i == sbi->s_groups_count - 1)
710 			last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
711 		else
712 			last_block = first_block +
713 				(EXT2_BLOCKS_PER_GROUP(sb) - 1);
714 
715 		if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
716 		    le32_to_cpu(gdp->bg_block_bitmap) > last_block)
717 		{
718 			ext2_error (sb, "ext2_check_descriptors",
719 				    "Block bitmap for group %d"
720 				    " not in group (block %lu)!",
721 				    i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
722 			return 0;
723 		}
724 		if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
725 		    le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
726 		{
727 			ext2_error (sb, "ext2_check_descriptors",
728 				    "Inode bitmap for group %d"
729 				    " not in group (block %lu)!",
730 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
731 			return 0;
732 		}
733 		if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
734 		    le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
735 		    last_block)
736 		{
737 			ext2_error (sb, "ext2_check_descriptors",
738 				    "Inode table for group %d"
739 				    " not in group (block %lu)!",
740 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
741 			return 0;
742 		}
743 	}
744 	return 1;
745 }
746 
747 /*
748  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
749  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
750  * We need to be 1 filesystem block less than the 2^32 sector limit.
751  */
752 static loff_t ext2_max_size(int bits)
753 {
754 	loff_t res = EXT2_NDIR_BLOCKS;
755 	int meta_blocks;
756 	loff_t upper_limit;
757 
758 	/* This is calculated to be the largest file size for a
759 	 * dense, file such that the total number of
760 	 * sectors in the file, including data and all indirect blocks,
761 	 * does not exceed 2^32 -1
762 	 * __u32 i_blocks representing the total number of
763 	 * 512 bytes blocks of the file
764 	 */
765 	upper_limit = (1LL << 32) - 1;
766 
767 	/* total blocks in file system block size */
768 	upper_limit >>= (bits - 9);
769 
770 
771 	/* indirect blocks */
772 	meta_blocks = 1;
773 	/* double indirect blocks */
774 	meta_blocks += 1 + (1LL << (bits-2));
775 	/* tripple indirect blocks */
776 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
777 
778 	upper_limit -= meta_blocks;
779 	upper_limit <<= bits;
780 
781 	res += 1LL << (bits-2);
782 	res += 1LL << (2*(bits-2));
783 	res += 1LL << (3*(bits-2));
784 	res <<= bits;
785 	if (res > upper_limit)
786 		res = upper_limit;
787 
788 	if (res > MAX_LFS_FILESIZE)
789 		res = MAX_LFS_FILESIZE;
790 
791 	return res;
792 }
793 
794 static unsigned long descriptor_loc(struct super_block *sb,
795 				    unsigned long logic_sb_block,
796 				    int nr)
797 {
798 	struct ext2_sb_info *sbi = EXT2_SB(sb);
799 	unsigned long bg, first_meta_bg;
800 	int has_super = 0;
801 
802 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
803 
804 	if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
805 	    nr < first_meta_bg)
806 		return (logic_sb_block + nr + 1);
807 	bg = sbi->s_desc_per_block * nr;
808 	if (ext2_bg_has_super(sb, bg))
809 		has_super = 1;
810 
811 	return ext2_group_first_block_no(sb, bg) + has_super;
812 }
813 
814 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
815 {
816 	struct buffer_head * bh;
817 	struct ext2_sb_info * sbi;
818 	struct ext2_super_block * es;
819 	struct inode *root;
820 	unsigned long block;
821 	unsigned long sb_block = get_sb_block(&data);
822 	unsigned long logic_sb_block;
823 	unsigned long offset = 0;
824 	unsigned long def_mount_opts;
825 	long ret = -EINVAL;
826 	int blocksize = BLOCK_SIZE;
827 	int db_count;
828 	int i, j;
829 	__le32 features;
830 	int err;
831 
832 	err = -ENOMEM;
833 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
834 	if (!sbi)
835 		goto failed;
836 
837 	sbi->s_blockgroup_lock =
838 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
839 	if (!sbi->s_blockgroup_lock) {
840 		kfree(sbi);
841 		goto failed;
842 	}
843 	sb->s_fs_info = sbi;
844 	sbi->s_sb_block = sb_block;
845 
846 	spin_lock_init(&sbi->s_lock);
847 
848 	/*
849 	 * See what the current blocksize for the device is, and
850 	 * use that as the blocksize.  Otherwise (or if the blocksize
851 	 * is smaller than the default) use the default.
852 	 * This is important for devices that have a hardware
853 	 * sectorsize that is larger than the default.
854 	 */
855 	blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
856 	if (!blocksize) {
857 		ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
858 		goto failed_sbi;
859 	}
860 
861 	/*
862 	 * If the superblock doesn't start on a hardware sector boundary,
863 	 * calculate the offset.
864 	 */
865 	if (blocksize != BLOCK_SIZE) {
866 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
867 		offset = (sb_block*BLOCK_SIZE) % blocksize;
868 	} else {
869 		logic_sb_block = sb_block;
870 	}
871 
872 	if (!(bh = sb_bread(sb, logic_sb_block))) {
873 		ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
874 		goto failed_sbi;
875 	}
876 	/*
877 	 * Note: s_es must be initialized as soon as possible because
878 	 *       some ext2 macro-instructions depend on its value
879 	 */
880 	es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
881 	sbi->s_es = es;
882 	sb->s_magic = le16_to_cpu(es->s_magic);
883 
884 	if (sb->s_magic != EXT2_SUPER_MAGIC)
885 		goto cantfind_ext2;
886 
887 	/* Set defaults before we parse the mount options */
888 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
889 	if (def_mount_opts & EXT2_DEFM_DEBUG)
890 		set_opt(sbi->s_mount_opt, DEBUG);
891 	if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
892 		set_opt(sbi->s_mount_opt, GRPID);
893 	if (def_mount_opts & EXT2_DEFM_UID16)
894 		set_opt(sbi->s_mount_opt, NO_UID32);
895 #ifdef CONFIG_EXT2_FS_XATTR
896 	if (def_mount_opts & EXT2_DEFM_XATTR_USER)
897 		set_opt(sbi->s_mount_opt, XATTR_USER);
898 #endif
899 #ifdef CONFIG_EXT2_FS_POSIX_ACL
900 	if (def_mount_opts & EXT2_DEFM_ACL)
901 		set_opt(sbi->s_mount_opt, POSIX_ACL);
902 #endif
903 
904 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
905 		set_opt(sbi->s_mount_opt, ERRORS_PANIC);
906 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
907 		set_opt(sbi->s_mount_opt, ERRORS_CONT);
908 	else
909 		set_opt(sbi->s_mount_opt, ERRORS_RO);
910 
911 	sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
912 	sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
913 
914 	set_opt(sbi->s_mount_opt, RESERVATION);
915 
916 	if (!parse_options((char *) data, sb))
917 		goto failed_mount;
918 
919 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
920 		((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
921 		 MS_POSIXACL : 0);
922 	sb->s_iflags |= SB_I_CGROUPWB;
923 
924 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
925 	    (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
926 	     EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
927 	     EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
928 		ext2_msg(sb, KERN_WARNING,
929 			"warning: feature flags set on rev 0 fs, "
930 			"running e2fsck is recommended");
931 	/*
932 	 * Check feature flags regardless of the revision level, since we
933 	 * previously didn't change the revision level when setting the flags,
934 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
935 	 */
936 	features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
937 	if (features) {
938 		ext2_msg(sb, KERN_ERR,	"error: couldn't mount because of "
939 		       "unsupported optional features (%x)",
940 			le32_to_cpu(features));
941 		goto failed_mount;
942 	}
943 	if (!(sb->s_flags & MS_RDONLY) &&
944 	    (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
945 		ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
946 		       "unsupported optional features (%x)",
947 		       le32_to_cpu(features));
948 		goto failed_mount;
949 	}
950 
951 	blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
952 
953 	if (sbi->s_mount_opt & EXT2_MOUNT_DAX) {
954 		err = bdev_dax_supported(sb, blocksize);
955 		if (err)
956 			goto failed_mount;
957 	}
958 
959 	/* If the blocksize doesn't match, re-read the thing.. */
960 	if (sb->s_blocksize != blocksize) {
961 		brelse(bh);
962 
963 		if (!sb_set_blocksize(sb, blocksize)) {
964 			ext2_msg(sb, KERN_ERR,
965 				"error: bad blocksize %d", blocksize);
966 			goto failed_sbi;
967 		}
968 
969 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
970 		offset = (sb_block*BLOCK_SIZE) % blocksize;
971 		bh = sb_bread(sb, logic_sb_block);
972 		if(!bh) {
973 			ext2_msg(sb, KERN_ERR, "error: couldn't read"
974 				"superblock on 2nd try");
975 			goto failed_sbi;
976 		}
977 		es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
978 		sbi->s_es = es;
979 		if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
980 			ext2_msg(sb, KERN_ERR, "error: magic mismatch");
981 			goto failed_mount;
982 		}
983 	}
984 
985 	sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
986 	sb->s_max_links = EXT2_LINK_MAX;
987 
988 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
989 		sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
990 		sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
991 	} else {
992 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
993 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
994 		if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
995 		    !is_power_of_2(sbi->s_inode_size) ||
996 		    (sbi->s_inode_size > blocksize)) {
997 			ext2_msg(sb, KERN_ERR,
998 				"error: unsupported inode size: %d",
999 				sbi->s_inode_size);
1000 			goto failed_mount;
1001 		}
1002 	}
1003 
1004 	sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
1005 				   le32_to_cpu(es->s_log_frag_size);
1006 	if (sbi->s_frag_size == 0)
1007 		goto cantfind_ext2;
1008 	sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
1009 
1010 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1011 	sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1012 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1013 
1014 	if (EXT2_INODE_SIZE(sb) == 0)
1015 		goto cantfind_ext2;
1016 	sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
1017 	if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
1018 		goto cantfind_ext2;
1019 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
1020 					sbi->s_inodes_per_block;
1021 	sbi->s_desc_per_block = sb->s_blocksize /
1022 					sizeof (struct ext2_group_desc);
1023 	sbi->s_sbh = bh;
1024 	sbi->s_mount_state = le16_to_cpu(es->s_state);
1025 	sbi->s_addr_per_block_bits =
1026 		ilog2 (EXT2_ADDR_PER_BLOCK(sb));
1027 	sbi->s_desc_per_block_bits =
1028 		ilog2 (EXT2_DESC_PER_BLOCK(sb));
1029 
1030 	if (sb->s_magic != EXT2_SUPER_MAGIC)
1031 		goto cantfind_ext2;
1032 
1033 	if (sb->s_blocksize != bh->b_size) {
1034 		if (!silent)
1035 			ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
1036 		goto failed_mount;
1037 	}
1038 
1039 	if (sb->s_blocksize != sbi->s_frag_size) {
1040 		ext2_msg(sb, KERN_ERR,
1041 			"error: fragsize %lu != blocksize %lu"
1042 			"(not supported yet)",
1043 			sbi->s_frag_size, sb->s_blocksize);
1044 		goto failed_mount;
1045 	}
1046 
1047 	if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
1048 		ext2_msg(sb, KERN_ERR,
1049 			"error: #blocks per group too big: %lu",
1050 			sbi->s_blocks_per_group);
1051 		goto failed_mount;
1052 	}
1053 	if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
1054 		ext2_msg(sb, KERN_ERR,
1055 			"error: #fragments per group too big: %lu",
1056 			sbi->s_frags_per_group);
1057 		goto failed_mount;
1058 	}
1059 	if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1060 		ext2_msg(sb, KERN_ERR,
1061 			"error: #inodes per group too big: %lu",
1062 			sbi->s_inodes_per_group);
1063 		goto failed_mount;
1064 	}
1065 
1066 	if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1067 		goto cantfind_ext2;
1068  	sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1069  				le32_to_cpu(es->s_first_data_block) - 1)
1070  					/ EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1071 	db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1072 		   EXT2_DESC_PER_BLOCK(sb);
1073 	sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
1074 	if (sbi->s_group_desc == NULL) {
1075 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1076 		goto failed_mount;
1077 	}
1078 	bgl_lock_init(sbi->s_blockgroup_lock);
1079 	sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1080 	if (!sbi->s_debts) {
1081 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1082 		goto failed_mount_group_desc;
1083 	}
1084 	for (i = 0; i < db_count; i++) {
1085 		block = descriptor_loc(sb, logic_sb_block, i);
1086 		sbi->s_group_desc[i] = sb_bread(sb, block);
1087 		if (!sbi->s_group_desc[i]) {
1088 			for (j = 0; j < i; j++)
1089 				brelse (sbi->s_group_desc[j]);
1090 			ext2_msg(sb, KERN_ERR,
1091 				"error: unable to read group descriptors");
1092 			goto failed_mount_group_desc;
1093 		}
1094 	}
1095 	if (!ext2_check_descriptors (sb)) {
1096 		ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1097 		goto failed_mount2;
1098 	}
1099 	sbi->s_gdb_count = db_count;
1100 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1101 	spin_lock_init(&sbi->s_next_gen_lock);
1102 
1103 	/* per fileystem reservation list head & lock */
1104 	spin_lock_init(&sbi->s_rsv_window_lock);
1105 	sbi->s_rsv_window_root = RB_ROOT;
1106 	/*
1107 	 * Add a single, static dummy reservation to the start of the
1108 	 * reservation window list --- it gives us a placeholder for
1109 	 * append-at-start-of-list which makes the allocation logic
1110 	 * _much_ simpler.
1111 	 */
1112 	sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1113 	sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1114 	sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1115 	sbi->s_rsv_window_head.rsv_goal_size = 0;
1116 	ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1117 
1118 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
1119 				ext2_count_free_blocks(sb), GFP_KERNEL);
1120 	if (!err) {
1121 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
1122 				ext2_count_free_inodes(sb), GFP_KERNEL);
1123 	}
1124 	if (!err) {
1125 		err = percpu_counter_init(&sbi->s_dirs_counter,
1126 				ext2_count_dirs(sb), GFP_KERNEL);
1127 	}
1128 	if (err) {
1129 		ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1130 		goto failed_mount3;
1131 	}
1132 
1133 #ifdef CONFIG_EXT2_FS_XATTR
1134 	sbi->s_mb_cache = ext2_xattr_create_cache();
1135 	if (!sbi->s_mb_cache) {
1136 		ext2_msg(sb, KERN_ERR, "Failed to create an mb_cache");
1137 		goto failed_mount3;
1138 	}
1139 #endif
1140 	/*
1141 	 * set up enough so that it can read an inode
1142 	 */
1143 	sb->s_op = &ext2_sops;
1144 	sb->s_export_op = &ext2_export_ops;
1145 	sb->s_xattr = ext2_xattr_handlers;
1146 
1147 #ifdef CONFIG_QUOTA
1148 	sb->dq_op = &dquot_operations;
1149 	sb->s_qcop = &ext2_quotactl_ops;
1150 	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1151 #endif
1152 
1153 	root = ext2_iget(sb, EXT2_ROOT_INO);
1154 	if (IS_ERR(root)) {
1155 		ret = PTR_ERR(root);
1156 		goto failed_mount3;
1157 	}
1158 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1159 		iput(root);
1160 		ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1161 		goto failed_mount3;
1162 	}
1163 
1164 	sb->s_root = d_make_root(root);
1165 	if (!sb->s_root) {
1166 		ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1167 		ret = -ENOMEM;
1168 		goto failed_mount3;
1169 	}
1170 	if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1171 		ext2_msg(sb, KERN_WARNING,
1172 			"warning: mounting ext3 filesystem as ext2");
1173 	if (ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY))
1174 		sb->s_flags |= MS_RDONLY;
1175 	ext2_write_super(sb);
1176 	return 0;
1177 
1178 cantfind_ext2:
1179 	if (!silent)
1180 		ext2_msg(sb, KERN_ERR,
1181 			"error: can't find an ext2 filesystem on dev %s.",
1182 			sb->s_id);
1183 	goto failed_mount;
1184 failed_mount3:
1185 	if (sbi->s_mb_cache)
1186 		ext2_xattr_destroy_cache(sbi->s_mb_cache);
1187 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
1188 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
1189 	percpu_counter_destroy(&sbi->s_dirs_counter);
1190 failed_mount2:
1191 	for (i = 0; i < db_count; i++)
1192 		brelse(sbi->s_group_desc[i]);
1193 failed_mount_group_desc:
1194 	kfree(sbi->s_group_desc);
1195 	kfree(sbi->s_debts);
1196 failed_mount:
1197 	brelse(bh);
1198 failed_sbi:
1199 	sb->s_fs_info = NULL;
1200 	kfree(sbi->s_blockgroup_lock);
1201 	kfree(sbi);
1202 failed:
1203 	return ret;
1204 }
1205 
1206 static void ext2_clear_super_error(struct super_block *sb)
1207 {
1208 	struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1209 
1210 	if (buffer_write_io_error(sbh)) {
1211 		/*
1212 		 * Oh, dear.  A previous attempt to write the
1213 		 * superblock failed.  This could happen because the
1214 		 * USB device was yanked out.  Or it could happen to
1215 		 * be a transient write error and maybe the block will
1216 		 * be remapped.  Nothing we can do but to retry the
1217 		 * write and hope for the best.
1218 		 */
1219 		ext2_msg(sb, KERN_ERR,
1220 		       "previous I/O error to superblock detected\n");
1221 		clear_buffer_write_io_error(sbh);
1222 		set_buffer_uptodate(sbh);
1223 	}
1224 }
1225 
1226 void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1227 		     int wait)
1228 {
1229 	ext2_clear_super_error(sb);
1230 	spin_lock(&EXT2_SB(sb)->s_lock);
1231 	es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1232 	es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1233 	es->s_wtime = cpu_to_le32(get_seconds());
1234 	/* unlock before we do IO */
1235 	spin_unlock(&EXT2_SB(sb)->s_lock);
1236 	mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1237 	if (wait)
1238 		sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1239 }
1240 
1241 /*
1242  * In the second extended file system, it is not necessary to
1243  * write the super block since we use a mapping of the
1244  * disk super block in a buffer.
1245  *
1246  * However, this function is still used to set the fs valid
1247  * flags to 0.  We need to set this flag to 0 since the fs
1248  * may have been checked while mounted and e2fsck may have
1249  * set s_state to EXT2_VALID_FS after some corrections.
1250  */
1251 static int ext2_sync_fs(struct super_block *sb, int wait)
1252 {
1253 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1254 	struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1255 
1256 	/*
1257 	 * Write quota structures to quota file, sync_blockdev() will write
1258 	 * them to disk later
1259 	 */
1260 	dquot_writeback_dquots(sb, -1);
1261 
1262 	spin_lock(&sbi->s_lock);
1263 	if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1264 		ext2_debug("setting valid to 0\n");
1265 		es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1266 	}
1267 	spin_unlock(&sbi->s_lock);
1268 	ext2_sync_super(sb, es, wait);
1269 	return 0;
1270 }
1271 
1272 static int ext2_freeze(struct super_block *sb)
1273 {
1274 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1275 
1276 	/*
1277 	 * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1278 	 * because we have unattached inodes and thus filesystem is not fully
1279 	 * consistent.
1280 	 */
1281 	if (atomic_long_read(&sb->s_remove_count)) {
1282 		ext2_sync_fs(sb, 1);
1283 		return 0;
1284 	}
1285 	/* Set EXT2_FS_VALID flag */
1286 	spin_lock(&sbi->s_lock);
1287 	sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1288 	spin_unlock(&sbi->s_lock);
1289 	ext2_sync_super(sb, sbi->s_es, 1);
1290 
1291 	return 0;
1292 }
1293 
1294 static int ext2_unfreeze(struct super_block *sb)
1295 {
1296 	/* Just write sb to clear EXT2_VALID_FS flag */
1297 	ext2_write_super(sb);
1298 
1299 	return 0;
1300 }
1301 
1302 static void ext2_write_super(struct super_block *sb)
1303 {
1304 	if (!(sb->s_flags & MS_RDONLY))
1305 		ext2_sync_fs(sb, 1);
1306 }
1307 
1308 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1309 {
1310 	struct ext2_sb_info * sbi = EXT2_SB(sb);
1311 	struct ext2_super_block * es;
1312 	struct ext2_mount_options old_opts;
1313 	unsigned long old_sb_flags;
1314 	int err;
1315 
1316 	sync_filesystem(sb);
1317 	spin_lock(&sbi->s_lock);
1318 
1319 	/* Store the old options */
1320 	old_sb_flags = sb->s_flags;
1321 	old_opts.s_mount_opt = sbi->s_mount_opt;
1322 	old_opts.s_resuid = sbi->s_resuid;
1323 	old_opts.s_resgid = sbi->s_resgid;
1324 
1325 	/*
1326 	 * Allow the "check" option to be passed as a remount option.
1327 	 */
1328 	if (!parse_options(data, sb)) {
1329 		err = -EINVAL;
1330 		goto restore_opts;
1331 	}
1332 
1333 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1334 		((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1335 
1336 	es = sbi->s_es;
1337 	if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT2_MOUNT_DAX) {
1338 		ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1339 			 "dax flag with busy inodes while remounting");
1340 		sbi->s_mount_opt ^= EXT2_MOUNT_DAX;
1341 	}
1342 	if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) {
1343 		spin_unlock(&sbi->s_lock);
1344 		return 0;
1345 	}
1346 	if (*flags & MS_RDONLY) {
1347 		if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1348 		    !(sbi->s_mount_state & EXT2_VALID_FS)) {
1349 			spin_unlock(&sbi->s_lock);
1350 			return 0;
1351 		}
1352 
1353 		/*
1354 		 * OK, we are remounting a valid rw partition rdonly, so set
1355 		 * the rdonly flag and then mark the partition as valid again.
1356 		 */
1357 		es->s_state = cpu_to_le16(sbi->s_mount_state);
1358 		es->s_mtime = cpu_to_le32(get_seconds());
1359 		spin_unlock(&sbi->s_lock);
1360 
1361 		err = dquot_suspend(sb, -1);
1362 		if (err < 0) {
1363 			spin_lock(&sbi->s_lock);
1364 			goto restore_opts;
1365 		}
1366 
1367 		ext2_sync_super(sb, es, 1);
1368 	} else {
1369 		__le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1370 					       ~EXT2_FEATURE_RO_COMPAT_SUPP);
1371 		if (ret) {
1372 			ext2_msg(sb, KERN_WARNING,
1373 				"warning: couldn't remount RDWR because of "
1374 				"unsupported optional features (%x).",
1375 				le32_to_cpu(ret));
1376 			err = -EROFS;
1377 			goto restore_opts;
1378 		}
1379 		/*
1380 		 * Mounting a RDONLY partition read-write, so reread and
1381 		 * store the current valid flag.  (It may have been changed
1382 		 * by e2fsck since we originally mounted the partition.)
1383 		 */
1384 		sbi->s_mount_state = le16_to_cpu(es->s_state);
1385 		if (!ext2_setup_super (sb, es, 0))
1386 			sb->s_flags &= ~MS_RDONLY;
1387 		spin_unlock(&sbi->s_lock);
1388 
1389 		ext2_write_super(sb);
1390 
1391 		dquot_resume(sb, -1);
1392 	}
1393 
1394 	return 0;
1395 restore_opts:
1396 	sbi->s_mount_opt = old_opts.s_mount_opt;
1397 	sbi->s_resuid = old_opts.s_resuid;
1398 	sbi->s_resgid = old_opts.s_resgid;
1399 	sb->s_flags = old_sb_flags;
1400 	spin_unlock(&sbi->s_lock);
1401 	return err;
1402 }
1403 
1404 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1405 {
1406 	struct super_block *sb = dentry->d_sb;
1407 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1408 	struct ext2_super_block *es = sbi->s_es;
1409 	u64 fsid;
1410 
1411 	spin_lock(&sbi->s_lock);
1412 
1413 	if (test_opt (sb, MINIX_DF))
1414 		sbi->s_overhead_last = 0;
1415 	else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1416 		unsigned long i, overhead = 0;
1417 		smp_rmb();
1418 
1419 		/*
1420 		 * Compute the overhead (FS structures). This is constant
1421 		 * for a given filesystem unless the number of block groups
1422 		 * changes so we cache the previous value until it does.
1423 		 */
1424 
1425 		/*
1426 		 * All of the blocks before first_data_block are
1427 		 * overhead
1428 		 */
1429 		overhead = le32_to_cpu(es->s_first_data_block);
1430 
1431 		/*
1432 		 * Add the overhead attributed to the superblock and
1433 		 * block group descriptors.  If the sparse superblocks
1434 		 * feature is turned on, then not all groups have this.
1435 		 */
1436 		for (i = 0; i < sbi->s_groups_count; i++)
1437 			overhead += ext2_bg_has_super(sb, i) +
1438 				ext2_bg_num_gdb(sb, i);
1439 
1440 		/*
1441 		 * Every block group has an inode bitmap, a block
1442 		 * bitmap, and an inode table.
1443 		 */
1444 		overhead += (sbi->s_groups_count *
1445 			     (2 + sbi->s_itb_per_group));
1446 		sbi->s_overhead_last = overhead;
1447 		smp_wmb();
1448 		sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1449 	}
1450 
1451 	buf->f_type = EXT2_SUPER_MAGIC;
1452 	buf->f_bsize = sb->s_blocksize;
1453 	buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1454 	buf->f_bfree = ext2_count_free_blocks(sb);
1455 	es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1456 	buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1457 	if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1458 		buf->f_bavail = 0;
1459 	buf->f_files = le32_to_cpu(es->s_inodes_count);
1460 	buf->f_ffree = ext2_count_free_inodes(sb);
1461 	es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1462 	buf->f_namelen = EXT2_NAME_LEN;
1463 	fsid = le64_to_cpup((void *)es->s_uuid) ^
1464 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1465 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1466 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1467 	spin_unlock(&sbi->s_lock);
1468 	return 0;
1469 }
1470 
1471 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1472 	int flags, const char *dev_name, void *data)
1473 {
1474 	return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1475 }
1476 
1477 #ifdef CONFIG_QUOTA
1478 
1479 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1480  * acquiring the locks... As quota files are never truncated and quota code
1481  * itself serializes the operations (and no one else should touch the files)
1482  * we don't have to be afraid of races */
1483 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1484 			       size_t len, loff_t off)
1485 {
1486 	struct inode *inode = sb_dqopt(sb)->files[type];
1487 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1488 	int err = 0;
1489 	int offset = off & (sb->s_blocksize - 1);
1490 	int tocopy;
1491 	size_t toread;
1492 	struct buffer_head tmp_bh;
1493 	struct buffer_head *bh;
1494 	loff_t i_size = i_size_read(inode);
1495 
1496 	if (off > i_size)
1497 		return 0;
1498 	if (off+len > i_size)
1499 		len = i_size-off;
1500 	toread = len;
1501 	while (toread > 0) {
1502 		tocopy = sb->s_blocksize - offset < toread ?
1503 				sb->s_blocksize - offset : toread;
1504 
1505 		tmp_bh.b_state = 0;
1506 		tmp_bh.b_size = sb->s_blocksize;
1507 		err = ext2_get_block(inode, blk, &tmp_bh, 0);
1508 		if (err < 0)
1509 			return err;
1510 		if (!buffer_mapped(&tmp_bh))	/* A hole? */
1511 			memset(data, 0, tocopy);
1512 		else {
1513 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1514 			if (!bh)
1515 				return -EIO;
1516 			memcpy(data, bh->b_data+offset, tocopy);
1517 			brelse(bh);
1518 		}
1519 		offset = 0;
1520 		toread -= tocopy;
1521 		data += tocopy;
1522 		blk++;
1523 	}
1524 	return len;
1525 }
1526 
1527 /* Write to quotafile */
1528 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1529 				const char *data, size_t len, loff_t off)
1530 {
1531 	struct inode *inode = sb_dqopt(sb)->files[type];
1532 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1533 	int err = 0;
1534 	int offset = off & (sb->s_blocksize - 1);
1535 	int tocopy;
1536 	size_t towrite = len;
1537 	struct buffer_head tmp_bh;
1538 	struct buffer_head *bh;
1539 
1540 	while (towrite > 0) {
1541 		tocopy = sb->s_blocksize - offset < towrite ?
1542 				sb->s_blocksize - offset : towrite;
1543 
1544 		tmp_bh.b_state = 0;
1545 		tmp_bh.b_size = sb->s_blocksize;
1546 		err = ext2_get_block(inode, blk, &tmp_bh, 1);
1547 		if (err < 0)
1548 			goto out;
1549 		if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1550 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1551 		else
1552 			bh = sb_getblk(sb, tmp_bh.b_blocknr);
1553 		if (unlikely(!bh)) {
1554 			err = -EIO;
1555 			goto out;
1556 		}
1557 		lock_buffer(bh);
1558 		memcpy(bh->b_data+offset, data, tocopy);
1559 		flush_dcache_page(bh->b_page);
1560 		set_buffer_uptodate(bh);
1561 		mark_buffer_dirty(bh);
1562 		unlock_buffer(bh);
1563 		brelse(bh);
1564 		offset = 0;
1565 		towrite -= tocopy;
1566 		data += tocopy;
1567 		blk++;
1568 	}
1569 out:
1570 	if (len == towrite)
1571 		return err;
1572 	if (inode->i_size < off+len-towrite)
1573 		i_size_write(inode, off+len-towrite);
1574 	inode->i_version++;
1575 	inode->i_mtime = inode->i_ctime = current_time(inode);
1576 	mark_inode_dirty(inode);
1577 	return len - towrite;
1578 }
1579 
1580 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
1581 			 const struct path *path)
1582 {
1583 	int err;
1584 	struct inode *inode;
1585 
1586 	err = dquot_quota_on(sb, type, format_id, path);
1587 	if (err)
1588 		return err;
1589 
1590 	inode = d_inode(path->dentry);
1591 	inode_lock(inode);
1592 	EXT2_I(inode)->i_flags |= EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL;
1593 	inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1594 			S_NOATIME | S_IMMUTABLE);
1595 	inode_unlock(inode);
1596 	mark_inode_dirty(inode);
1597 
1598 	return 0;
1599 }
1600 
1601 static int ext2_quota_off(struct super_block *sb, int type)
1602 {
1603 	struct inode *inode = sb_dqopt(sb)->files[type];
1604 	int err;
1605 
1606 	if (!inode || !igrab(inode))
1607 		goto out;
1608 
1609 	err = dquot_quota_off(sb, type);
1610 	if (err)
1611 		goto out_put;
1612 
1613 	inode_lock(inode);
1614 	EXT2_I(inode)->i_flags &= ~(EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL);
1615 	inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1616 	inode_unlock(inode);
1617 	mark_inode_dirty(inode);
1618 out_put:
1619 	iput(inode);
1620 	return err;
1621 out:
1622 	return dquot_quota_off(sb, type);
1623 }
1624 
1625 #endif
1626 
1627 static struct file_system_type ext2_fs_type = {
1628 	.owner		= THIS_MODULE,
1629 	.name		= "ext2",
1630 	.mount		= ext2_mount,
1631 	.kill_sb	= kill_block_super,
1632 	.fs_flags	= FS_REQUIRES_DEV,
1633 };
1634 MODULE_ALIAS_FS("ext2");
1635 
1636 static int __init init_ext2_fs(void)
1637 {
1638 	int err;
1639 
1640 	err = init_inodecache();
1641 	if (err)
1642 		return err;
1643         err = register_filesystem(&ext2_fs_type);
1644 	if (err)
1645 		goto out;
1646 	return 0;
1647 out:
1648 	destroy_inodecache();
1649 	return err;
1650 }
1651 
1652 static void __exit exit_ext2_fs(void)
1653 {
1654 	unregister_filesystem(&ext2_fs_type);
1655 	destroy_inodecache();
1656 }
1657 
1658 MODULE_AUTHOR("Remy Card and others");
1659 MODULE_DESCRIPTION("Second Extended Filesystem");
1660 MODULE_LICENSE("GPL");
1661 module_init(init_ext2_fs)
1662 module_exit(exit_ext2_fs)
1663