xref: /openbmc/linux/fs/ext2/super.c (revision 161f3b74)
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_write_super(struct super_block *sb);
40 static int ext2_remount (struct super_block * sb, int * flags, char * data);
41 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
42 static int ext2_sync_fs(struct super_block *sb, int wait);
43 static int ext2_freeze(struct super_block *sb);
44 static int ext2_unfreeze(struct super_block *sb);
45 
46 void ext2_error(struct super_block *sb, const char *function,
47 		const char *fmt, ...)
48 {
49 	struct va_format vaf;
50 	va_list args;
51 	struct ext2_sb_info *sbi = EXT2_SB(sb);
52 	struct ext2_super_block *es = sbi->s_es;
53 
54 	if (!(sb->s_flags & MS_RDONLY)) {
55 		spin_lock(&sbi->s_lock);
56 		sbi->s_mount_state |= EXT2_ERROR_FS;
57 		es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
58 		spin_unlock(&sbi->s_lock);
59 		ext2_sync_super(sb, es, 1);
60 	}
61 
62 	va_start(args, fmt);
63 
64 	vaf.fmt = fmt;
65 	vaf.va = &args;
66 
67 	printk(KERN_CRIT "EXT2-fs (%s): error: %s: %pV\n",
68 	       sb->s_id, function, &vaf);
69 
70 	va_end(args);
71 
72 	if (test_opt(sb, ERRORS_PANIC))
73 		panic("EXT2-fs: panic from previous error\n");
74 	if (test_opt(sb, ERRORS_RO)) {
75 		ext2_msg(sb, KERN_CRIT,
76 			     "error: remounting filesystem read-only");
77 		sb->s_flags |= MS_RDONLY;
78 	}
79 }
80 
81 void ext2_msg(struct super_block *sb, const char *prefix,
82 		const char *fmt, ...)
83 {
84 	struct va_format vaf;
85 	va_list args;
86 
87 	va_start(args, fmt);
88 
89 	vaf.fmt = fmt;
90 	vaf.va = &args;
91 
92 	printk("%sEXT2-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
93 
94 	va_end(args);
95 }
96 
97 /*
98  * This must be called with sbi->s_lock held.
99  */
100 void ext2_update_dynamic_rev(struct super_block *sb)
101 {
102 	struct ext2_super_block *es = EXT2_SB(sb)->s_es;
103 
104 	if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
105 		return;
106 
107 	ext2_msg(sb, KERN_WARNING,
108 		     "warning: updating to rev %d because of "
109 		     "new feature flag, running e2fsck is recommended",
110 		     EXT2_DYNAMIC_REV);
111 
112 	es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
113 	es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
114 	es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
115 	/* leave es->s_feature_*compat flags alone */
116 	/* es->s_uuid will be set by e2fsck if empty */
117 
118 	/*
119 	 * The rest of the superblock fields should be zero, and if not it
120 	 * means they are likely already in use, so leave them alone.  We
121 	 * can leave it up to e2fsck to clean up any inconsistencies there.
122 	 */
123 }
124 
125 #ifdef CONFIG_QUOTA
126 static int ext2_quota_off(struct super_block *sb, int type);
127 
128 static void ext2_quota_off_umount(struct super_block *sb)
129 {
130 	int type;
131 
132 	for (type = 0; type < MAXQUOTAS; type++)
133 		ext2_quota_off(sb, type);
134 }
135 #else
136 static inline void ext2_quota_off_umount(struct super_block *sb)
137 {
138 }
139 #endif
140 
141 static void ext2_put_super (struct super_block * sb)
142 {
143 	int db_count;
144 	int i;
145 	struct ext2_sb_info *sbi = EXT2_SB(sb);
146 
147 	ext2_quota_off_umount(sb);
148 
149 	if (sbi->s_mb_cache) {
150 		ext2_xattr_destroy_cache(sbi->s_mb_cache);
151 		sbi->s_mb_cache = NULL;
152 	}
153 	if (!(sb->s_flags & MS_RDONLY)) {
154 		struct ext2_super_block *es = sbi->s_es;
155 
156 		spin_lock(&sbi->s_lock);
157 		es->s_state = cpu_to_le16(sbi->s_mount_state);
158 		spin_unlock(&sbi->s_lock);
159 		ext2_sync_super(sb, es, 1);
160 	}
161 	db_count = sbi->s_gdb_count;
162 	for (i = 0; i < db_count; i++)
163 		if (sbi->s_group_desc[i])
164 			brelse (sbi->s_group_desc[i]);
165 	kfree(sbi->s_group_desc);
166 	kfree(sbi->s_debts);
167 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
168 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
169 	percpu_counter_destroy(&sbi->s_dirs_counter);
170 	brelse (sbi->s_sbh);
171 	sb->s_fs_info = NULL;
172 	kfree(sbi->s_blockgroup_lock);
173 	kfree(sbi);
174 }
175 
176 static struct kmem_cache * ext2_inode_cachep;
177 
178 static struct inode *ext2_alloc_inode(struct super_block *sb)
179 {
180 	struct ext2_inode_info *ei;
181 	ei = kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
182 	if (!ei)
183 		return NULL;
184 	ei->i_block_alloc_info = NULL;
185 	ei->vfs_inode.i_version = 1;
186 #ifdef CONFIG_QUOTA
187 	memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
188 #endif
189 
190 	return &ei->vfs_inode;
191 }
192 
193 static void ext2_i_callback(struct rcu_head *head)
194 {
195 	struct inode *inode = container_of(head, struct inode, i_rcu);
196 	kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
197 }
198 
199 static void ext2_destroy_inode(struct inode *inode)
200 {
201 	call_rcu(&inode->i_rcu, ext2_i_callback);
202 }
203 
204 static void init_once(void *foo)
205 {
206 	struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
207 
208 	rwlock_init(&ei->i_meta_lock);
209 #ifdef CONFIG_EXT2_FS_XATTR
210 	init_rwsem(&ei->xattr_sem);
211 #endif
212 	mutex_init(&ei->truncate_mutex);
213 #ifdef CONFIG_FS_DAX
214 	init_rwsem(&ei->dax_sem);
215 #endif
216 	inode_init_once(&ei->vfs_inode);
217 }
218 
219 static int __init init_inodecache(void)
220 {
221 	ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
222 					     sizeof(struct ext2_inode_info),
223 					     0, (SLAB_RECLAIM_ACCOUNT|
224 						SLAB_MEM_SPREAD|SLAB_ACCOUNT),
225 					     init_once);
226 	if (ext2_inode_cachep == NULL)
227 		return -ENOMEM;
228 	return 0;
229 }
230 
231 static void destroy_inodecache(void)
232 {
233 	/*
234 	 * Make sure all delayed rcu free inodes are flushed before we
235 	 * destroy cache.
236 	 */
237 	rcu_barrier();
238 	kmem_cache_destroy(ext2_inode_cachep);
239 }
240 
241 static int ext2_show_options(struct seq_file *seq, struct dentry *root)
242 {
243 	struct super_block *sb = root->d_sb;
244 	struct ext2_sb_info *sbi = EXT2_SB(sb);
245 	struct ext2_super_block *es = sbi->s_es;
246 	unsigned long def_mount_opts;
247 
248 	spin_lock(&sbi->s_lock);
249 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
250 
251 	if (sbi->s_sb_block != 1)
252 		seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
253 	if (test_opt(sb, MINIX_DF))
254 		seq_puts(seq, ",minixdf");
255 	if (test_opt(sb, GRPID))
256 		seq_puts(seq, ",grpid");
257 	if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
258 		seq_puts(seq, ",nogrpid");
259 	if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT2_DEF_RESUID)) ||
260 	    le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
261 		seq_printf(seq, ",resuid=%u",
262 				from_kuid_munged(&init_user_ns, sbi->s_resuid));
263 	}
264 	if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT2_DEF_RESGID)) ||
265 	    le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
266 		seq_printf(seq, ",resgid=%u",
267 				from_kgid_munged(&init_user_ns, sbi->s_resgid));
268 	}
269 	if (test_opt(sb, ERRORS_RO)) {
270 		int def_errors = le16_to_cpu(es->s_errors);
271 
272 		if (def_errors == EXT2_ERRORS_PANIC ||
273 		    def_errors == EXT2_ERRORS_CONTINUE) {
274 			seq_puts(seq, ",errors=remount-ro");
275 		}
276 	}
277 	if (test_opt(sb, ERRORS_CONT))
278 		seq_puts(seq, ",errors=continue");
279 	if (test_opt(sb, ERRORS_PANIC))
280 		seq_puts(seq, ",errors=panic");
281 	if (test_opt(sb, NO_UID32))
282 		seq_puts(seq, ",nouid32");
283 	if (test_opt(sb, DEBUG))
284 		seq_puts(seq, ",debug");
285 	if (test_opt(sb, OLDALLOC))
286 		seq_puts(seq, ",oldalloc");
287 
288 #ifdef CONFIG_EXT2_FS_XATTR
289 	if (test_opt(sb, XATTR_USER))
290 		seq_puts(seq, ",user_xattr");
291 	if (!test_opt(sb, XATTR_USER) &&
292 	    (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
293 		seq_puts(seq, ",nouser_xattr");
294 	}
295 #endif
296 
297 #ifdef CONFIG_EXT2_FS_POSIX_ACL
298 	if (test_opt(sb, POSIX_ACL))
299 		seq_puts(seq, ",acl");
300 	if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
301 		seq_puts(seq, ",noacl");
302 #endif
303 
304 	if (test_opt(sb, NOBH))
305 		seq_puts(seq, ",nobh");
306 
307 #if defined(CONFIG_QUOTA)
308 	if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
309 		seq_puts(seq, ",usrquota");
310 
311 	if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
312 		seq_puts(seq, ",grpquota");
313 #endif
314 
315 #ifdef CONFIG_FS_DAX
316 	if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
317 		seq_puts(seq, ",xip");
318 	if (sbi->s_mount_opt & EXT2_MOUNT_DAX)
319 		seq_puts(seq, ",dax");
320 #endif
321 
322 	if (!test_opt(sb, RESERVATION))
323 		seq_puts(seq, ",noreservation");
324 
325 	spin_unlock(&sbi->s_lock);
326 	return 0;
327 }
328 
329 #ifdef CONFIG_QUOTA
330 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
331 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
332 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
333 			 const struct path *path);
334 static struct dquot **ext2_get_dquots(struct inode *inode)
335 {
336 	return EXT2_I(inode)->i_dquot;
337 }
338 
339 static const struct quotactl_ops ext2_quotactl_ops = {
340 	.quota_on	= ext2_quota_on,
341 	.quota_off	= ext2_quota_off,
342 	.quota_sync	= dquot_quota_sync,
343 	.get_state	= dquot_get_state,
344 	.set_info	= dquot_set_dqinfo,
345 	.get_dqblk	= dquot_get_dqblk,
346 	.set_dqblk	= dquot_set_dqblk,
347 	.get_nextdqblk	= dquot_get_next_dqblk,
348 };
349 #endif
350 
351 static const struct super_operations ext2_sops = {
352 	.alloc_inode	= ext2_alloc_inode,
353 	.destroy_inode	= ext2_destroy_inode,
354 	.write_inode	= ext2_write_inode,
355 	.evict_inode	= ext2_evict_inode,
356 	.put_super	= ext2_put_super,
357 	.sync_fs	= ext2_sync_fs,
358 	.freeze_fs	= ext2_freeze,
359 	.unfreeze_fs	= ext2_unfreeze,
360 	.statfs		= ext2_statfs,
361 	.remount_fs	= ext2_remount,
362 	.show_options	= ext2_show_options,
363 #ifdef CONFIG_QUOTA
364 	.quota_read	= ext2_quota_read,
365 	.quota_write	= ext2_quota_write,
366 	.get_dquots	= ext2_get_dquots,
367 #endif
368 };
369 
370 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
371 		u64 ino, u32 generation)
372 {
373 	struct inode *inode;
374 
375 	if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
376 		return ERR_PTR(-ESTALE);
377 	if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
378 		return ERR_PTR(-ESTALE);
379 
380 	/*
381 	 * ext2_iget isn't quite right if the inode is currently unallocated!
382 	 * However ext2_iget currently does appropriate checks to handle stale
383 	 * inodes so everything is OK.
384 	 */
385 	inode = ext2_iget(sb, ino);
386 	if (IS_ERR(inode))
387 		return ERR_CAST(inode);
388 	if (generation && inode->i_generation != generation) {
389 		/* we didn't find the right inode.. */
390 		iput(inode);
391 		return ERR_PTR(-ESTALE);
392 	}
393 	return inode;
394 }
395 
396 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
397 		int fh_len, int fh_type)
398 {
399 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
400 				    ext2_nfs_get_inode);
401 }
402 
403 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
404 		int fh_len, int fh_type)
405 {
406 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
407 				    ext2_nfs_get_inode);
408 }
409 
410 static const struct export_operations ext2_export_ops = {
411 	.fh_to_dentry = ext2_fh_to_dentry,
412 	.fh_to_parent = ext2_fh_to_parent,
413 	.get_parent = ext2_get_parent,
414 };
415 
416 static unsigned long get_sb_block(void **data)
417 {
418 	unsigned long 	sb_block;
419 	char 		*options = (char *) *data;
420 
421 	if (!options || strncmp(options, "sb=", 3) != 0)
422 		return 1;	/* Default location */
423 	options += 3;
424 	sb_block = simple_strtoul(options, &options, 0);
425 	if (*options && *options != ',') {
426 		printk("EXT2-fs: Invalid sb specification: %s\n",
427 		       (char *) *data);
428 		return 1;
429 	}
430 	if (*options == ',')
431 		options++;
432 	*data = (void *) options;
433 	return sb_block;
434 }
435 
436 enum {
437 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
438 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
439 	Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
440 	Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
441 	Opt_acl, Opt_noacl, Opt_xip, Opt_dax, Opt_ignore, Opt_err, Opt_quota,
442 	Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
443 };
444 
445 static const match_table_t tokens = {
446 	{Opt_bsd_df, "bsddf"},
447 	{Opt_minix_df, "minixdf"},
448 	{Opt_grpid, "grpid"},
449 	{Opt_grpid, "bsdgroups"},
450 	{Opt_nogrpid, "nogrpid"},
451 	{Opt_nogrpid, "sysvgroups"},
452 	{Opt_resgid, "resgid=%u"},
453 	{Opt_resuid, "resuid=%u"},
454 	{Opt_sb, "sb=%u"},
455 	{Opt_err_cont, "errors=continue"},
456 	{Opt_err_panic, "errors=panic"},
457 	{Opt_err_ro, "errors=remount-ro"},
458 	{Opt_nouid32, "nouid32"},
459 	{Opt_nocheck, "check=none"},
460 	{Opt_nocheck, "nocheck"},
461 	{Opt_debug, "debug"},
462 	{Opt_oldalloc, "oldalloc"},
463 	{Opt_orlov, "orlov"},
464 	{Opt_nobh, "nobh"},
465 	{Opt_user_xattr, "user_xattr"},
466 	{Opt_nouser_xattr, "nouser_xattr"},
467 	{Opt_acl, "acl"},
468 	{Opt_noacl, "noacl"},
469 	{Opt_xip, "xip"},
470 	{Opt_dax, "dax"},
471 	{Opt_grpquota, "grpquota"},
472 	{Opt_ignore, "noquota"},
473 	{Opt_quota, "quota"},
474 	{Opt_usrquota, "usrquota"},
475 	{Opt_reservation, "reservation"},
476 	{Opt_noreservation, "noreservation"},
477 	{Opt_err, NULL}
478 };
479 
480 static int parse_options(char *options, struct super_block *sb)
481 {
482 	char *p;
483 	struct ext2_sb_info *sbi = EXT2_SB(sb);
484 	substring_t args[MAX_OPT_ARGS];
485 	int option;
486 	kuid_t uid;
487 	kgid_t gid;
488 
489 	if (!options)
490 		return 1;
491 
492 	while ((p = strsep (&options, ",")) != NULL) {
493 		int token;
494 		if (!*p)
495 			continue;
496 
497 		token = match_token(p, tokens, args);
498 		switch (token) {
499 		case Opt_bsd_df:
500 			clear_opt (sbi->s_mount_opt, MINIX_DF);
501 			break;
502 		case Opt_minix_df:
503 			set_opt (sbi->s_mount_opt, MINIX_DF);
504 			break;
505 		case Opt_grpid:
506 			set_opt (sbi->s_mount_opt, GRPID);
507 			break;
508 		case Opt_nogrpid:
509 			clear_opt (sbi->s_mount_opt, GRPID);
510 			break;
511 		case Opt_resuid:
512 			if (match_int(&args[0], &option))
513 				return 0;
514 			uid = make_kuid(current_user_ns(), option);
515 			if (!uid_valid(uid)) {
516 				ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
517 				return 0;
518 
519 			}
520 			sbi->s_resuid = uid;
521 			break;
522 		case Opt_resgid:
523 			if (match_int(&args[0], &option))
524 				return 0;
525 			gid = make_kgid(current_user_ns(), option);
526 			if (!gid_valid(gid)) {
527 				ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
528 				return 0;
529 			}
530 			sbi->s_resgid = gid;
531 			break;
532 		case Opt_sb:
533 			/* handled by get_sb_block() instead of here */
534 			/* *sb_block = match_int(&args[0]); */
535 			break;
536 		case Opt_err_panic:
537 			clear_opt (sbi->s_mount_opt, ERRORS_CONT);
538 			clear_opt (sbi->s_mount_opt, ERRORS_RO);
539 			set_opt (sbi->s_mount_opt, ERRORS_PANIC);
540 			break;
541 		case Opt_err_ro:
542 			clear_opt (sbi->s_mount_opt, ERRORS_CONT);
543 			clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
544 			set_opt (sbi->s_mount_opt, ERRORS_RO);
545 			break;
546 		case Opt_err_cont:
547 			clear_opt (sbi->s_mount_opt, ERRORS_RO);
548 			clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
549 			set_opt (sbi->s_mount_opt, ERRORS_CONT);
550 			break;
551 		case Opt_nouid32:
552 			set_opt (sbi->s_mount_opt, NO_UID32);
553 			break;
554 		case Opt_nocheck:
555 			clear_opt (sbi->s_mount_opt, CHECK);
556 			break;
557 		case Opt_debug:
558 			set_opt (sbi->s_mount_opt, DEBUG);
559 			break;
560 		case Opt_oldalloc:
561 			set_opt (sbi->s_mount_opt, OLDALLOC);
562 			break;
563 		case Opt_orlov:
564 			clear_opt (sbi->s_mount_opt, OLDALLOC);
565 			break;
566 		case Opt_nobh:
567 			set_opt (sbi->s_mount_opt, NOBH);
568 			break;
569 #ifdef CONFIG_EXT2_FS_XATTR
570 		case Opt_user_xattr:
571 			set_opt (sbi->s_mount_opt, XATTR_USER);
572 			break;
573 		case Opt_nouser_xattr:
574 			clear_opt (sbi->s_mount_opt, XATTR_USER);
575 			break;
576 #else
577 		case Opt_user_xattr:
578 		case Opt_nouser_xattr:
579 			ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
580 				"not supported");
581 			break;
582 #endif
583 #ifdef CONFIG_EXT2_FS_POSIX_ACL
584 		case Opt_acl:
585 			set_opt(sbi->s_mount_opt, POSIX_ACL);
586 			break;
587 		case Opt_noacl:
588 			clear_opt(sbi->s_mount_opt, POSIX_ACL);
589 			break;
590 #else
591 		case Opt_acl:
592 		case Opt_noacl:
593 			ext2_msg(sb, KERN_INFO,
594 				"(no)acl options not supported");
595 			break;
596 #endif
597 		case Opt_xip:
598 			ext2_msg(sb, KERN_INFO, "use dax instead of xip");
599 			set_opt(sbi->s_mount_opt, XIP);
600 			/* Fall through */
601 		case Opt_dax:
602 #ifdef CONFIG_FS_DAX
603 			ext2_msg(sb, KERN_WARNING,
604 		"DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
605 			set_opt(sbi->s_mount_opt, DAX);
606 #else
607 			ext2_msg(sb, KERN_INFO, "dax option not supported");
608 #endif
609 			break;
610 
611 #if defined(CONFIG_QUOTA)
612 		case Opt_quota:
613 		case Opt_usrquota:
614 			set_opt(sbi->s_mount_opt, USRQUOTA);
615 			break;
616 
617 		case Opt_grpquota:
618 			set_opt(sbi->s_mount_opt, GRPQUOTA);
619 			break;
620 #else
621 		case Opt_quota:
622 		case Opt_usrquota:
623 		case Opt_grpquota:
624 			ext2_msg(sb, KERN_INFO,
625 				"quota operations not supported");
626 			break;
627 #endif
628 
629 		case Opt_reservation:
630 			set_opt(sbi->s_mount_opt, RESERVATION);
631 			ext2_msg(sb, KERN_INFO, "reservations ON");
632 			break;
633 		case Opt_noreservation:
634 			clear_opt(sbi->s_mount_opt, RESERVATION);
635 			ext2_msg(sb, KERN_INFO, "reservations OFF");
636 			break;
637 		case Opt_ignore:
638 			break;
639 		default:
640 			return 0;
641 		}
642 	}
643 	return 1;
644 }
645 
646 static int ext2_setup_super (struct super_block * sb,
647 			      struct ext2_super_block * es,
648 			      int read_only)
649 {
650 	int res = 0;
651 	struct ext2_sb_info *sbi = EXT2_SB(sb);
652 
653 	if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
654 		ext2_msg(sb, KERN_ERR,
655 			"error: revision level too high, "
656 			"forcing read-only mode");
657 		res = MS_RDONLY;
658 	}
659 	if (read_only)
660 		return res;
661 	if (!(sbi->s_mount_state & EXT2_VALID_FS))
662 		ext2_msg(sb, KERN_WARNING,
663 			"warning: mounting unchecked fs, "
664 			"running e2fsck is recommended");
665 	else if ((sbi->s_mount_state & EXT2_ERROR_FS))
666 		ext2_msg(sb, KERN_WARNING,
667 			"warning: mounting fs with errors, "
668 			"running e2fsck is recommended");
669 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
670 		 le16_to_cpu(es->s_mnt_count) >=
671 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
672 		ext2_msg(sb, KERN_WARNING,
673 			"warning: maximal mount count reached, "
674 			"running e2fsck is recommended");
675 	else if (le32_to_cpu(es->s_checkinterval) &&
676 		(le32_to_cpu(es->s_lastcheck) +
677 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
678 		ext2_msg(sb, KERN_WARNING,
679 			"warning: checktime reached, "
680 			"running e2fsck is recommended");
681 	if (!le16_to_cpu(es->s_max_mnt_count))
682 		es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
683 	le16_add_cpu(&es->s_mnt_count, 1);
684 	if (test_opt (sb, DEBUG))
685 		ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
686 			"bpg=%lu, ipg=%lu, mo=%04lx]",
687 			EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
688 			sbi->s_frag_size,
689 			sbi->s_groups_count,
690 			EXT2_BLOCKS_PER_GROUP(sb),
691 			EXT2_INODES_PER_GROUP(sb),
692 			sbi->s_mount_opt);
693 	return res;
694 }
695 
696 static int ext2_check_descriptors(struct super_block *sb)
697 {
698 	int i;
699 	struct ext2_sb_info *sbi = EXT2_SB(sb);
700 
701 	ext2_debug ("Checking group descriptors");
702 
703 	for (i = 0; i < sbi->s_groups_count; i++) {
704 		struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
705 		ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
706 		ext2_fsblk_t last_block;
707 
708 		if (i == sbi->s_groups_count - 1)
709 			last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
710 		else
711 			last_block = first_block +
712 				(EXT2_BLOCKS_PER_GROUP(sb) - 1);
713 
714 		if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
715 		    le32_to_cpu(gdp->bg_block_bitmap) > last_block)
716 		{
717 			ext2_error (sb, "ext2_check_descriptors",
718 				    "Block bitmap for group %d"
719 				    " not in group (block %lu)!",
720 				    i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
721 			return 0;
722 		}
723 		if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
724 		    le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
725 		{
726 			ext2_error (sb, "ext2_check_descriptors",
727 				    "Inode bitmap for group %d"
728 				    " not in group (block %lu)!",
729 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
730 			return 0;
731 		}
732 		if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
733 		    le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
734 		    last_block)
735 		{
736 			ext2_error (sb, "ext2_check_descriptors",
737 				    "Inode table for group %d"
738 				    " not in group (block %lu)!",
739 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
740 			return 0;
741 		}
742 	}
743 	return 1;
744 }
745 
746 /*
747  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
748  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
749  * We need to be 1 filesystem block less than the 2^32 sector limit.
750  */
751 static loff_t ext2_max_size(int bits)
752 {
753 	loff_t res = EXT2_NDIR_BLOCKS;
754 	int meta_blocks;
755 	loff_t upper_limit;
756 
757 	/* This is calculated to be the largest file size for a
758 	 * dense, file such that the total number of
759 	 * sectors in the file, including data and all indirect blocks,
760 	 * does not exceed 2^32 -1
761 	 * __u32 i_blocks representing the total number of
762 	 * 512 bytes blocks of the file
763 	 */
764 	upper_limit = (1LL << 32) - 1;
765 
766 	/* total blocks in file system block size */
767 	upper_limit >>= (bits - 9);
768 
769 
770 	/* indirect blocks */
771 	meta_blocks = 1;
772 	/* double indirect blocks */
773 	meta_blocks += 1 + (1LL << (bits-2));
774 	/* tripple indirect blocks */
775 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
776 
777 	upper_limit -= meta_blocks;
778 	upper_limit <<= bits;
779 
780 	res += 1LL << (bits-2);
781 	res += 1LL << (2*(bits-2));
782 	res += 1LL << (3*(bits-2));
783 	res <<= bits;
784 	if (res > upper_limit)
785 		res = upper_limit;
786 
787 	if (res > MAX_LFS_FILESIZE)
788 		res = MAX_LFS_FILESIZE;
789 
790 	return res;
791 }
792 
793 static unsigned long descriptor_loc(struct super_block *sb,
794 				    unsigned long logic_sb_block,
795 				    int nr)
796 {
797 	struct ext2_sb_info *sbi = EXT2_SB(sb);
798 	unsigned long bg, first_meta_bg;
799 	int has_super = 0;
800 
801 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
802 
803 	if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
804 	    nr < first_meta_bg)
805 		return (logic_sb_block + nr + 1);
806 	bg = sbi->s_desc_per_block * nr;
807 	if (ext2_bg_has_super(sb, bg))
808 		has_super = 1;
809 
810 	return ext2_group_first_block_no(sb, bg) + has_super;
811 }
812 
813 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
814 {
815 	struct buffer_head * bh;
816 	struct ext2_sb_info * sbi;
817 	struct ext2_super_block * es;
818 	struct inode *root;
819 	unsigned long block;
820 	unsigned long sb_block = get_sb_block(&data);
821 	unsigned long logic_sb_block;
822 	unsigned long offset = 0;
823 	unsigned long def_mount_opts;
824 	long ret = -EINVAL;
825 	int blocksize = BLOCK_SIZE;
826 	int db_count;
827 	int i, j;
828 	__le32 features;
829 	int err;
830 
831 	err = -ENOMEM;
832 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
833 	if (!sbi)
834 		goto failed;
835 
836 	sbi->s_blockgroup_lock =
837 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
838 	if (!sbi->s_blockgroup_lock) {
839 		kfree(sbi);
840 		goto failed;
841 	}
842 	sb->s_fs_info = sbi;
843 	sbi->s_sb_block = sb_block;
844 
845 	spin_lock_init(&sbi->s_lock);
846 
847 	/*
848 	 * See what the current blocksize for the device is, and
849 	 * use that as the blocksize.  Otherwise (or if the blocksize
850 	 * is smaller than the default) use the default.
851 	 * This is important for devices that have a hardware
852 	 * sectorsize that is larger than the default.
853 	 */
854 	blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
855 	if (!blocksize) {
856 		ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
857 		goto failed_sbi;
858 	}
859 
860 	/*
861 	 * If the superblock doesn't start on a hardware sector boundary,
862 	 * calculate the offset.
863 	 */
864 	if (blocksize != BLOCK_SIZE) {
865 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
866 		offset = (sb_block*BLOCK_SIZE) % blocksize;
867 	} else {
868 		logic_sb_block = sb_block;
869 	}
870 
871 	if (!(bh = sb_bread(sb, logic_sb_block))) {
872 		ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
873 		goto failed_sbi;
874 	}
875 	/*
876 	 * Note: s_es must be initialized as soon as possible because
877 	 *       some ext2 macro-instructions depend on its value
878 	 */
879 	es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
880 	sbi->s_es = es;
881 	sb->s_magic = le16_to_cpu(es->s_magic);
882 
883 	if (sb->s_magic != EXT2_SUPER_MAGIC)
884 		goto cantfind_ext2;
885 
886 	/* Set defaults before we parse the mount options */
887 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
888 	if (def_mount_opts & EXT2_DEFM_DEBUG)
889 		set_opt(sbi->s_mount_opt, DEBUG);
890 	if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
891 		set_opt(sbi->s_mount_opt, GRPID);
892 	if (def_mount_opts & EXT2_DEFM_UID16)
893 		set_opt(sbi->s_mount_opt, NO_UID32);
894 #ifdef CONFIG_EXT2_FS_XATTR
895 	if (def_mount_opts & EXT2_DEFM_XATTR_USER)
896 		set_opt(sbi->s_mount_opt, XATTR_USER);
897 #endif
898 #ifdef CONFIG_EXT2_FS_POSIX_ACL
899 	if (def_mount_opts & EXT2_DEFM_ACL)
900 		set_opt(sbi->s_mount_opt, POSIX_ACL);
901 #endif
902 
903 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
904 		set_opt(sbi->s_mount_opt, ERRORS_PANIC);
905 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
906 		set_opt(sbi->s_mount_opt, ERRORS_CONT);
907 	else
908 		set_opt(sbi->s_mount_opt, ERRORS_RO);
909 
910 	sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
911 	sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
912 
913 	set_opt(sbi->s_mount_opt, RESERVATION);
914 
915 	if (!parse_options((char *) data, sb))
916 		goto failed_mount;
917 
918 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
919 		((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
920 		 MS_POSIXACL : 0);
921 	sb->s_iflags |= SB_I_CGROUPWB;
922 
923 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
924 	    (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
925 	     EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
926 	     EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
927 		ext2_msg(sb, KERN_WARNING,
928 			"warning: feature flags set on rev 0 fs, "
929 			"running e2fsck is recommended");
930 	/*
931 	 * Check feature flags regardless of the revision level, since we
932 	 * previously didn't change the revision level when setting the flags,
933 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
934 	 */
935 	features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
936 	if (features) {
937 		ext2_msg(sb, KERN_ERR,	"error: couldn't mount because of "
938 		       "unsupported optional features (%x)",
939 			le32_to_cpu(features));
940 		goto failed_mount;
941 	}
942 	if (!(sb->s_flags & MS_RDONLY) &&
943 	    (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
944 		ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
945 		       "unsupported optional features (%x)",
946 		       le32_to_cpu(features));
947 		goto failed_mount;
948 	}
949 
950 	blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
951 
952 	if (sbi->s_mount_opt & EXT2_MOUNT_DAX) {
953 		err = bdev_dax_supported(sb, blocksize);
954 		if (err)
955 			goto failed_mount;
956 	}
957 
958 	/* If the blocksize doesn't match, re-read the thing.. */
959 	if (sb->s_blocksize != blocksize) {
960 		brelse(bh);
961 
962 		if (!sb_set_blocksize(sb, blocksize)) {
963 			ext2_msg(sb, KERN_ERR,
964 				"error: bad blocksize %d", blocksize);
965 			goto failed_sbi;
966 		}
967 
968 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
969 		offset = (sb_block*BLOCK_SIZE) % blocksize;
970 		bh = sb_bread(sb, logic_sb_block);
971 		if(!bh) {
972 			ext2_msg(sb, KERN_ERR, "error: couldn't read"
973 				"superblock on 2nd try");
974 			goto failed_sbi;
975 		}
976 		es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
977 		sbi->s_es = es;
978 		if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
979 			ext2_msg(sb, KERN_ERR, "error: magic mismatch");
980 			goto failed_mount;
981 		}
982 	}
983 
984 	sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
985 	sb->s_max_links = EXT2_LINK_MAX;
986 
987 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
988 		sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
989 		sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
990 	} else {
991 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
992 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
993 		if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
994 		    !is_power_of_2(sbi->s_inode_size) ||
995 		    (sbi->s_inode_size > blocksize)) {
996 			ext2_msg(sb, KERN_ERR,
997 				"error: unsupported inode size: %d",
998 				sbi->s_inode_size);
999 			goto failed_mount;
1000 		}
1001 	}
1002 
1003 	sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
1004 				   le32_to_cpu(es->s_log_frag_size);
1005 	if (sbi->s_frag_size == 0)
1006 		goto cantfind_ext2;
1007 	sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
1008 
1009 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1010 	sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1011 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1012 
1013 	if (EXT2_INODE_SIZE(sb) == 0)
1014 		goto cantfind_ext2;
1015 	sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
1016 	if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
1017 		goto cantfind_ext2;
1018 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
1019 					sbi->s_inodes_per_block;
1020 	sbi->s_desc_per_block = sb->s_blocksize /
1021 					sizeof (struct ext2_group_desc);
1022 	sbi->s_sbh = bh;
1023 	sbi->s_mount_state = le16_to_cpu(es->s_state);
1024 	sbi->s_addr_per_block_bits =
1025 		ilog2 (EXT2_ADDR_PER_BLOCK(sb));
1026 	sbi->s_desc_per_block_bits =
1027 		ilog2 (EXT2_DESC_PER_BLOCK(sb));
1028 
1029 	if (sb->s_magic != EXT2_SUPER_MAGIC)
1030 		goto cantfind_ext2;
1031 
1032 	if (sb->s_blocksize != bh->b_size) {
1033 		if (!silent)
1034 			ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
1035 		goto failed_mount;
1036 	}
1037 
1038 	if (sb->s_blocksize != sbi->s_frag_size) {
1039 		ext2_msg(sb, KERN_ERR,
1040 			"error: fragsize %lu != blocksize %lu"
1041 			"(not supported yet)",
1042 			sbi->s_frag_size, sb->s_blocksize);
1043 		goto failed_mount;
1044 	}
1045 
1046 	if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
1047 		ext2_msg(sb, KERN_ERR,
1048 			"error: #blocks per group too big: %lu",
1049 			sbi->s_blocks_per_group);
1050 		goto failed_mount;
1051 	}
1052 	if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
1053 		ext2_msg(sb, KERN_ERR,
1054 			"error: #fragments per group too big: %lu",
1055 			sbi->s_frags_per_group);
1056 		goto failed_mount;
1057 	}
1058 	if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1059 		ext2_msg(sb, KERN_ERR,
1060 			"error: #inodes per group too big: %lu",
1061 			sbi->s_inodes_per_group);
1062 		goto failed_mount;
1063 	}
1064 
1065 	if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1066 		goto cantfind_ext2;
1067  	sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1068  				le32_to_cpu(es->s_first_data_block) - 1)
1069  					/ EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1070 	db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1071 		   EXT2_DESC_PER_BLOCK(sb);
1072 	sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
1073 	if (sbi->s_group_desc == NULL) {
1074 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1075 		goto failed_mount;
1076 	}
1077 	bgl_lock_init(sbi->s_blockgroup_lock);
1078 	sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1079 	if (!sbi->s_debts) {
1080 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1081 		goto failed_mount_group_desc;
1082 	}
1083 	for (i = 0; i < db_count; i++) {
1084 		block = descriptor_loc(sb, logic_sb_block, i);
1085 		sbi->s_group_desc[i] = sb_bread(sb, block);
1086 		if (!sbi->s_group_desc[i]) {
1087 			for (j = 0; j < i; j++)
1088 				brelse (sbi->s_group_desc[j]);
1089 			ext2_msg(sb, KERN_ERR,
1090 				"error: unable to read group descriptors");
1091 			goto failed_mount_group_desc;
1092 		}
1093 	}
1094 	if (!ext2_check_descriptors (sb)) {
1095 		ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1096 		goto failed_mount2;
1097 	}
1098 	sbi->s_gdb_count = db_count;
1099 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1100 	spin_lock_init(&sbi->s_next_gen_lock);
1101 
1102 	/* per fileystem reservation list head & lock */
1103 	spin_lock_init(&sbi->s_rsv_window_lock);
1104 	sbi->s_rsv_window_root = RB_ROOT;
1105 	/*
1106 	 * Add a single, static dummy reservation to the start of the
1107 	 * reservation window list --- it gives us a placeholder for
1108 	 * append-at-start-of-list which makes the allocation logic
1109 	 * _much_ simpler.
1110 	 */
1111 	sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1112 	sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1113 	sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1114 	sbi->s_rsv_window_head.rsv_goal_size = 0;
1115 	ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1116 
1117 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
1118 				ext2_count_free_blocks(sb), GFP_KERNEL);
1119 	if (!err) {
1120 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
1121 				ext2_count_free_inodes(sb), GFP_KERNEL);
1122 	}
1123 	if (!err) {
1124 		err = percpu_counter_init(&sbi->s_dirs_counter,
1125 				ext2_count_dirs(sb), GFP_KERNEL);
1126 	}
1127 	if (err) {
1128 		ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1129 		goto failed_mount3;
1130 	}
1131 
1132 #ifdef CONFIG_EXT2_FS_XATTR
1133 	sbi->s_mb_cache = ext2_xattr_create_cache();
1134 	if (!sbi->s_mb_cache) {
1135 		ext2_msg(sb, KERN_ERR, "Failed to create an mb_cache");
1136 		goto failed_mount3;
1137 	}
1138 #endif
1139 	/*
1140 	 * set up enough so that it can read an inode
1141 	 */
1142 	sb->s_op = &ext2_sops;
1143 	sb->s_export_op = &ext2_export_ops;
1144 	sb->s_xattr = ext2_xattr_handlers;
1145 
1146 #ifdef CONFIG_QUOTA
1147 	sb->dq_op = &dquot_operations;
1148 	sb->s_qcop = &ext2_quotactl_ops;
1149 	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1150 #endif
1151 
1152 	root = ext2_iget(sb, EXT2_ROOT_INO);
1153 	if (IS_ERR(root)) {
1154 		ret = PTR_ERR(root);
1155 		goto failed_mount3;
1156 	}
1157 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1158 		iput(root);
1159 		ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1160 		goto failed_mount3;
1161 	}
1162 
1163 	sb->s_root = d_make_root(root);
1164 	if (!sb->s_root) {
1165 		ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1166 		ret = -ENOMEM;
1167 		goto failed_mount3;
1168 	}
1169 	if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1170 		ext2_msg(sb, KERN_WARNING,
1171 			"warning: mounting ext3 filesystem as ext2");
1172 	if (ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY))
1173 		sb->s_flags |= MS_RDONLY;
1174 	ext2_write_super(sb);
1175 	return 0;
1176 
1177 cantfind_ext2:
1178 	if (!silent)
1179 		ext2_msg(sb, KERN_ERR,
1180 			"error: can't find an ext2 filesystem on dev %s.",
1181 			sb->s_id);
1182 	goto failed_mount;
1183 failed_mount3:
1184 	if (sbi->s_mb_cache)
1185 		ext2_xattr_destroy_cache(sbi->s_mb_cache);
1186 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
1187 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
1188 	percpu_counter_destroy(&sbi->s_dirs_counter);
1189 failed_mount2:
1190 	for (i = 0; i < db_count; i++)
1191 		brelse(sbi->s_group_desc[i]);
1192 failed_mount_group_desc:
1193 	kfree(sbi->s_group_desc);
1194 	kfree(sbi->s_debts);
1195 failed_mount:
1196 	brelse(bh);
1197 failed_sbi:
1198 	sb->s_fs_info = NULL;
1199 	kfree(sbi->s_blockgroup_lock);
1200 	kfree(sbi);
1201 failed:
1202 	return ret;
1203 }
1204 
1205 static void ext2_clear_super_error(struct super_block *sb)
1206 {
1207 	struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1208 
1209 	if (buffer_write_io_error(sbh)) {
1210 		/*
1211 		 * Oh, dear.  A previous attempt to write the
1212 		 * superblock failed.  This could happen because the
1213 		 * USB device was yanked out.  Or it could happen to
1214 		 * be a transient write error and maybe the block will
1215 		 * be remapped.  Nothing we can do but to retry the
1216 		 * write and hope for the best.
1217 		 */
1218 		ext2_msg(sb, KERN_ERR,
1219 		       "previous I/O error to superblock detected\n");
1220 		clear_buffer_write_io_error(sbh);
1221 		set_buffer_uptodate(sbh);
1222 	}
1223 }
1224 
1225 void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1226 		     int wait)
1227 {
1228 	ext2_clear_super_error(sb);
1229 	spin_lock(&EXT2_SB(sb)->s_lock);
1230 	es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1231 	es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1232 	es->s_wtime = cpu_to_le32(get_seconds());
1233 	/* unlock before we do IO */
1234 	spin_unlock(&EXT2_SB(sb)->s_lock);
1235 	mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1236 	if (wait)
1237 		sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1238 }
1239 
1240 /*
1241  * In the second extended file system, it is not necessary to
1242  * write the super block since we use a mapping of the
1243  * disk super block in a buffer.
1244  *
1245  * However, this function is still used to set the fs valid
1246  * flags to 0.  We need to set this flag to 0 since the fs
1247  * may have been checked while mounted and e2fsck may have
1248  * set s_state to EXT2_VALID_FS after some corrections.
1249  */
1250 static int ext2_sync_fs(struct super_block *sb, int wait)
1251 {
1252 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1253 	struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1254 
1255 	/*
1256 	 * Write quota structures to quota file, sync_blockdev() will write
1257 	 * them to disk later
1258 	 */
1259 	dquot_writeback_dquots(sb, -1);
1260 
1261 	spin_lock(&sbi->s_lock);
1262 	if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1263 		ext2_debug("setting valid to 0\n");
1264 		es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1265 	}
1266 	spin_unlock(&sbi->s_lock);
1267 	ext2_sync_super(sb, es, wait);
1268 	return 0;
1269 }
1270 
1271 static int ext2_freeze(struct super_block *sb)
1272 {
1273 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1274 
1275 	/*
1276 	 * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1277 	 * because we have unattached inodes and thus filesystem is not fully
1278 	 * consistent.
1279 	 */
1280 	if (atomic_long_read(&sb->s_remove_count)) {
1281 		ext2_sync_fs(sb, 1);
1282 		return 0;
1283 	}
1284 	/* Set EXT2_FS_VALID flag */
1285 	spin_lock(&sbi->s_lock);
1286 	sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1287 	spin_unlock(&sbi->s_lock);
1288 	ext2_sync_super(sb, sbi->s_es, 1);
1289 
1290 	return 0;
1291 }
1292 
1293 static int ext2_unfreeze(struct super_block *sb)
1294 {
1295 	/* Just write sb to clear EXT2_VALID_FS flag */
1296 	ext2_write_super(sb);
1297 
1298 	return 0;
1299 }
1300 
1301 static void ext2_write_super(struct super_block *sb)
1302 {
1303 	if (!(sb->s_flags & MS_RDONLY))
1304 		ext2_sync_fs(sb, 1);
1305 }
1306 
1307 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1308 {
1309 	struct ext2_sb_info * sbi = EXT2_SB(sb);
1310 	struct ext2_super_block * es;
1311 	struct ext2_mount_options old_opts;
1312 	unsigned long old_sb_flags;
1313 	int err;
1314 
1315 	sync_filesystem(sb);
1316 	spin_lock(&sbi->s_lock);
1317 
1318 	/* Store the old options */
1319 	old_sb_flags = sb->s_flags;
1320 	old_opts.s_mount_opt = sbi->s_mount_opt;
1321 	old_opts.s_resuid = sbi->s_resuid;
1322 	old_opts.s_resgid = sbi->s_resgid;
1323 
1324 	/*
1325 	 * Allow the "check" option to be passed as a remount option.
1326 	 */
1327 	if (!parse_options(data, sb)) {
1328 		err = -EINVAL;
1329 		goto restore_opts;
1330 	}
1331 
1332 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1333 		((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1334 
1335 	es = sbi->s_es;
1336 	if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT2_MOUNT_DAX) {
1337 		ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1338 			 "dax flag with busy inodes while remounting");
1339 		sbi->s_mount_opt ^= EXT2_MOUNT_DAX;
1340 	}
1341 	if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) {
1342 		spin_unlock(&sbi->s_lock);
1343 		return 0;
1344 	}
1345 	if (*flags & MS_RDONLY) {
1346 		if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1347 		    !(sbi->s_mount_state & EXT2_VALID_FS)) {
1348 			spin_unlock(&sbi->s_lock);
1349 			return 0;
1350 		}
1351 
1352 		/*
1353 		 * OK, we are remounting a valid rw partition rdonly, so set
1354 		 * the rdonly flag and then mark the partition as valid again.
1355 		 */
1356 		es->s_state = cpu_to_le16(sbi->s_mount_state);
1357 		es->s_mtime = cpu_to_le32(get_seconds());
1358 		spin_unlock(&sbi->s_lock);
1359 
1360 		err = dquot_suspend(sb, -1);
1361 		if (err < 0) {
1362 			spin_lock(&sbi->s_lock);
1363 			goto restore_opts;
1364 		}
1365 
1366 		ext2_sync_super(sb, es, 1);
1367 	} else {
1368 		__le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1369 					       ~EXT2_FEATURE_RO_COMPAT_SUPP);
1370 		if (ret) {
1371 			ext2_msg(sb, KERN_WARNING,
1372 				"warning: couldn't remount RDWR because of "
1373 				"unsupported optional features (%x).",
1374 				le32_to_cpu(ret));
1375 			err = -EROFS;
1376 			goto restore_opts;
1377 		}
1378 		/*
1379 		 * Mounting a RDONLY partition read-write, so reread and
1380 		 * store the current valid flag.  (It may have been changed
1381 		 * by e2fsck since we originally mounted the partition.)
1382 		 */
1383 		sbi->s_mount_state = le16_to_cpu(es->s_state);
1384 		if (!ext2_setup_super (sb, es, 0))
1385 			sb->s_flags &= ~MS_RDONLY;
1386 		spin_unlock(&sbi->s_lock);
1387 
1388 		ext2_write_super(sb);
1389 
1390 		dquot_resume(sb, -1);
1391 	}
1392 
1393 	return 0;
1394 restore_opts:
1395 	sbi->s_mount_opt = old_opts.s_mount_opt;
1396 	sbi->s_resuid = old_opts.s_resuid;
1397 	sbi->s_resgid = old_opts.s_resgid;
1398 	sb->s_flags = old_sb_flags;
1399 	spin_unlock(&sbi->s_lock);
1400 	return err;
1401 }
1402 
1403 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1404 {
1405 	struct super_block *sb = dentry->d_sb;
1406 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1407 	struct ext2_super_block *es = sbi->s_es;
1408 	u64 fsid;
1409 
1410 	spin_lock(&sbi->s_lock);
1411 
1412 	if (test_opt (sb, MINIX_DF))
1413 		sbi->s_overhead_last = 0;
1414 	else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1415 		unsigned long i, overhead = 0;
1416 		smp_rmb();
1417 
1418 		/*
1419 		 * Compute the overhead (FS structures). This is constant
1420 		 * for a given filesystem unless the number of block groups
1421 		 * changes so we cache the previous value until it does.
1422 		 */
1423 
1424 		/*
1425 		 * All of the blocks before first_data_block are
1426 		 * overhead
1427 		 */
1428 		overhead = le32_to_cpu(es->s_first_data_block);
1429 
1430 		/*
1431 		 * Add the overhead attributed to the superblock and
1432 		 * block group descriptors.  If the sparse superblocks
1433 		 * feature is turned on, then not all groups have this.
1434 		 */
1435 		for (i = 0; i < sbi->s_groups_count; i++)
1436 			overhead += ext2_bg_has_super(sb, i) +
1437 				ext2_bg_num_gdb(sb, i);
1438 
1439 		/*
1440 		 * Every block group has an inode bitmap, a block
1441 		 * bitmap, and an inode table.
1442 		 */
1443 		overhead += (sbi->s_groups_count *
1444 			     (2 + sbi->s_itb_per_group));
1445 		sbi->s_overhead_last = overhead;
1446 		smp_wmb();
1447 		sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1448 	}
1449 
1450 	buf->f_type = EXT2_SUPER_MAGIC;
1451 	buf->f_bsize = sb->s_blocksize;
1452 	buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1453 	buf->f_bfree = ext2_count_free_blocks(sb);
1454 	es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1455 	buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1456 	if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1457 		buf->f_bavail = 0;
1458 	buf->f_files = le32_to_cpu(es->s_inodes_count);
1459 	buf->f_ffree = ext2_count_free_inodes(sb);
1460 	es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1461 	buf->f_namelen = EXT2_NAME_LEN;
1462 	fsid = le64_to_cpup((void *)es->s_uuid) ^
1463 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1464 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1465 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1466 	spin_unlock(&sbi->s_lock);
1467 	return 0;
1468 }
1469 
1470 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1471 	int flags, const char *dev_name, void *data)
1472 {
1473 	return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1474 }
1475 
1476 #ifdef CONFIG_QUOTA
1477 
1478 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1479  * acquiring the locks... As quota files are never truncated and quota code
1480  * itself serializes the operations (and no one else should touch the files)
1481  * we don't have to be afraid of races */
1482 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1483 			       size_t len, loff_t off)
1484 {
1485 	struct inode *inode = sb_dqopt(sb)->files[type];
1486 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1487 	int err = 0;
1488 	int offset = off & (sb->s_blocksize - 1);
1489 	int tocopy;
1490 	size_t toread;
1491 	struct buffer_head tmp_bh;
1492 	struct buffer_head *bh;
1493 	loff_t i_size = i_size_read(inode);
1494 
1495 	if (off > i_size)
1496 		return 0;
1497 	if (off+len > i_size)
1498 		len = i_size-off;
1499 	toread = len;
1500 	while (toread > 0) {
1501 		tocopy = sb->s_blocksize - offset < toread ?
1502 				sb->s_blocksize - offset : toread;
1503 
1504 		tmp_bh.b_state = 0;
1505 		tmp_bh.b_size = sb->s_blocksize;
1506 		err = ext2_get_block(inode, blk, &tmp_bh, 0);
1507 		if (err < 0)
1508 			return err;
1509 		if (!buffer_mapped(&tmp_bh))	/* A hole? */
1510 			memset(data, 0, tocopy);
1511 		else {
1512 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1513 			if (!bh)
1514 				return -EIO;
1515 			memcpy(data, bh->b_data+offset, tocopy);
1516 			brelse(bh);
1517 		}
1518 		offset = 0;
1519 		toread -= tocopy;
1520 		data += tocopy;
1521 		blk++;
1522 	}
1523 	return len;
1524 }
1525 
1526 /* Write to quotafile */
1527 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1528 				const char *data, size_t len, loff_t off)
1529 {
1530 	struct inode *inode = sb_dqopt(sb)->files[type];
1531 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1532 	int err = 0;
1533 	int offset = off & (sb->s_blocksize - 1);
1534 	int tocopy;
1535 	size_t towrite = len;
1536 	struct buffer_head tmp_bh;
1537 	struct buffer_head *bh;
1538 
1539 	while (towrite > 0) {
1540 		tocopy = sb->s_blocksize - offset < towrite ?
1541 				sb->s_blocksize - offset : towrite;
1542 
1543 		tmp_bh.b_state = 0;
1544 		tmp_bh.b_size = sb->s_blocksize;
1545 		err = ext2_get_block(inode, blk, &tmp_bh, 1);
1546 		if (err < 0)
1547 			goto out;
1548 		if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1549 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1550 		else
1551 			bh = sb_getblk(sb, tmp_bh.b_blocknr);
1552 		if (unlikely(!bh)) {
1553 			err = -EIO;
1554 			goto out;
1555 		}
1556 		lock_buffer(bh);
1557 		memcpy(bh->b_data+offset, data, tocopy);
1558 		flush_dcache_page(bh->b_page);
1559 		set_buffer_uptodate(bh);
1560 		mark_buffer_dirty(bh);
1561 		unlock_buffer(bh);
1562 		brelse(bh);
1563 		offset = 0;
1564 		towrite -= tocopy;
1565 		data += tocopy;
1566 		blk++;
1567 	}
1568 out:
1569 	if (len == towrite)
1570 		return err;
1571 	if (inode->i_size < off+len-towrite)
1572 		i_size_write(inode, off+len-towrite);
1573 	inode->i_version++;
1574 	inode->i_mtime = inode->i_ctime = current_time(inode);
1575 	mark_inode_dirty(inode);
1576 	return len - towrite;
1577 }
1578 
1579 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
1580 			 const struct path *path)
1581 {
1582 	int err;
1583 	struct inode *inode;
1584 
1585 	err = dquot_quota_on(sb, type, format_id, path);
1586 	if (err)
1587 		return err;
1588 
1589 	inode = d_inode(path->dentry);
1590 	inode_lock(inode);
1591 	EXT2_I(inode)->i_flags |= EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL;
1592 	inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1593 			S_NOATIME | S_IMMUTABLE);
1594 	inode_unlock(inode);
1595 	mark_inode_dirty(inode);
1596 
1597 	return 0;
1598 }
1599 
1600 static int ext2_quota_off(struct super_block *sb, int type)
1601 {
1602 	struct inode *inode = sb_dqopt(sb)->files[type];
1603 	int err;
1604 
1605 	if (!inode || !igrab(inode))
1606 		goto out;
1607 
1608 	err = dquot_quota_off(sb, type);
1609 	if (err)
1610 		goto out_put;
1611 
1612 	inode_lock(inode);
1613 	EXT2_I(inode)->i_flags &= ~(EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL);
1614 	inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1615 	inode_unlock(inode);
1616 	mark_inode_dirty(inode);
1617 out_put:
1618 	iput(inode);
1619 	return err;
1620 out:
1621 	return dquot_quota_off(sb, type);
1622 }
1623 
1624 #endif
1625 
1626 static struct file_system_type ext2_fs_type = {
1627 	.owner		= THIS_MODULE,
1628 	.name		= "ext2",
1629 	.mount		= ext2_mount,
1630 	.kill_sb	= kill_block_super,
1631 	.fs_flags	= FS_REQUIRES_DEV,
1632 };
1633 MODULE_ALIAS_FS("ext2");
1634 
1635 static int __init init_ext2_fs(void)
1636 {
1637 	int err;
1638 
1639 	err = init_inodecache();
1640 	if (err)
1641 		return err;
1642         err = register_filesystem(&ext2_fs_type);
1643 	if (err)
1644 		goto out;
1645 	return 0;
1646 out:
1647 	destroy_inodecache();
1648 	return err;
1649 }
1650 
1651 static void __exit exit_ext2_fs(void)
1652 {
1653 	unregister_filesystem(&ext2_fs_type);
1654 	destroy_inodecache();
1655 }
1656 
1657 MODULE_AUTHOR("Remy Card and others");
1658 MODULE_DESCRIPTION("Second Extended Filesystem");
1659 MODULE_LICENSE("GPL");
1660 module_init(init_ext2_fs)
1661 module_exit(exit_ext2_fs)
1662