xref: /openbmc/linux/fs/ext2/super.c (revision 3b23dc52)
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 <linux/iversion.h>
37 #include "ext2.h"
38 #include "xattr.h"
39 #include "acl.h"
40 
41 static void ext2_write_super(struct super_block *sb);
42 static int ext2_remount (struct super_block * sb, int * flags, char * data);
43 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
44 static int ext2_sync_fs(struct super_block *sb, int wait);
45 static int ext2_freeze(struct super_block *sb);
46 static int ext2_unfreeze(struct super_block *sb);
47 
48 void ext2_error(struct super_block *sb, const char *function,
49 		const char *fmt, ...)
50 {
51 	struct va_format vaf;
52 	va_list args;
53 	struct ext2_sb_info *sbi = EXT2_SB(sb);
54 	struct ext2_super_block *es = sbi->s_es;
55 
56 	if (!sb_rdonly(sb)) {
57 		spin_lock(&sbi->s_lock);
58 		sbi->s_mount_state |= EXT2_ERROR_FS;
59 		es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
60 		spin_unlock(&sbi->s_lock);
61 		ext2_sync_super(sb, es, 1);
62 	}
63 
64 	va_start(args, fmt);
65 
66 	vaf.fmt = fmt;
67 	vaf.va = &args;
68 
69 	printk(KERN_CRIT "EXT2-fs (%s): error: %s: %pV\n",
70 	       sb->s_id, function, &vaf);
71 
72 	va_end(args);
73 
74 	if (test_opt(sb, ERRORS_PANIC))
75 		panic("EXT2-fs: panic from previous error\n");
76 	if (test_opt(sb, ERRORS_RO)) {
77 		ext2_msg(sb, KERN_CRIT,
78 			     "error: remounting filesystem read-only");
79 		sb->s_flags |= SB_RDONLY;
80 	}
81 }
82 
83 void ext2_msg(struct super_block *sb, const char *prefix,
84 		const char *fmt, ...)
85 {
86 	struct va_format vaf;
87 	va_list args;
88 
89 	va_start(args, fmt);
90 
91 	vaf.fmt = fmt;
92 	vaf.va = &args;
93 
94 	printk("%sEXT2-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
95 
96 	va_end(args);
97 }
98 
99 /*
100  * This must be called with sbi->s_lock held.
101  */
102 void ext2_update_dynamic_rev(struct super_block *sb)
103 {
104 	struct ext2_super_block *es = EXT2_SB(sb)->s_es;
105 
106 	if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
107 		return;
108 
109 	ext2_msg(sb, KERN_WARNING,
110 		     "warning: updating to rev %d because of "
111 		     "new feature flag, running e2fsck is recommended",
112 		     EXT2_DYNAMIC_REV);
113 
114 	es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
115 	es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
116 	es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
117 	/* leave es->s_feature_*compat flags alone */
118 	/* es->s_uuid will be set by e2fsck if empty */
119 
120 	/*
121 	 * The rest of the superblock fields should be zero, and if not it
122 	 * means they are likely already in use, so leave them alone.  We
123 	 * can leave it up to e2fsck to clean up any inconsistencies there.
124 	 */
125 }
126 
127 #ifdef CONFIG_QUOTA
128 static int ext2_quota_off(struct super_block *sb, int type);
129 
130 static void ext2_quota_off_umount(struct super_block *sb)
131 {
132 	int type;
133 
134 	for (type = 0; type < MAXQUOTAS; type++)
135 		ext2_quota_off(sb, type);
136 }
137 #else
138 static inline void ext2_quota_off_umount(struct super_block *sb)
139 {
140 }
141 #endif
142 
143 static void ext2_put_super (struct super_block * sb)
144 {
145 	int db_count;
146 	int i;
147 	struct ext2_sb_info *sbi = EXT2_SB(sb);
148 
149 	ext2_quota_off_umount(sb);
150 
151 	if (sbi->s_ea_block_cache) {
152 		ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
153 		sbi->s_ea_block_cache = NULL;
154 	}
155 	if (!sb_rdonly(sb)) {
156 		struct ext2_super_block *es = sbi->s_es;
157 
158 		spin_lock(&sbi->s_lock);
159 		es->s_state = cpu_to_le16(sbi->s_mount_state);
160 		spin_unlock(&sbi->s_lock);
161 		ext2_sync_super(sb, es, 1);
162 	}
163 	db_count = sbi->s_gdb_count;
164 	for (i = 0; i < db_count; i++)
165 		if (sbi->s_group_desc[i])
166 			brelse (sbi->s_group_desc[i]);
167 	kfree(sbi->s_group_desc);
168 	kfree(sbi->s_debts);
169 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
170 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
171 	percpu_counter_destroy(&sbi->s_dirs_counter);
172 	brelse (sbi->s_sbh);
173 	sb->s_fs_info = NULL;
174 	kfree(sbi->s_blockgroup_lock);
175 	fs_put_dax(sbi->s_daxdev);
176 	kfree(sbi);
177 }
178 
179 static struct kmem_cache * ext2_inode_cachep;
180 
181 static struct inode *ext2_alloc_inode(struct super_block *sb)
182 {
183 	struct ext2_inode_info *ei;
184 	ei = kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
185 	if (!ei)
186 		return NULL;
187 	ei->i_block_alloc_info = NULL;
188 	inode_set_iversion(&ei->vfs_inode, 1);
189 #ifdef CONFIG_QUOTA
190 	memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
191 #endif
192 
193 	return &ei->vfs_inode;
194 }
195 
196 static void ext2_i_callback(struct rcu_head *head)
197 {
198 	struct inode *inode = container_of(head, struct inode, i_rcu);
199 	kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
200 }
201 
202 static void ext2_destroy_inode(struct inode *inode)
203 {
204 	call_rcu(&inode->i_rcu, ext2_i_callback);
205 }
206 
207 static void init_once(void *foo)
208 {
209 	struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
210 
211 	rwlock_init(&ei->i_meta_lock);
212 #ifdef CONFIG_EXT2_FS_XATTR
213 	init_rwsem(&ei->xattr_sem);
214 #endif
215 	mutex_init(&ei->truncate_mutex);
216 #ifdef CONFIG_FS_DAX
217 	init_rwsem(&ei->dax_sem);
218 #endif
219 	inode_init_once(&ei->vfs_inode);
220 }
221 
222 static int __init init_inodecache(void)
223 {
224 	ext2_inode_cachep = kmem_cache_create_usercopy("ext2_inode_cache",
225 				sizeof(struct ext2_inode_info), 0,
226 				(SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
227 					SLAB_ACCOUNT),
228 				offsetof(struct ext2_inode_info, i_data),
229 				sizeof_field(struct ext2_inode_info, i_data),
230 				init_once);
231 	if (ext2_inode_cachep == NULL)
232 		return -ENOMEM;
233 	return 0;
234 }
235 
236 static void destroy_inodecache(void)
237 {
238 	/*
239 	 * Make sure all delayed rcu free inodes are flushed before we
240 	 * destroy cache.
241 	 */
242 	rcu_barrier();
243 	kmem_cache_destroy(ext2_inode_cachep);
244 }
245 
246 static int ext2_show_options(struct seq_file *seq, struct dentry *root)
247 {
248 	struct super_block *sb = root->d_sb;
249 	struct ext2_sb_info *sbi = EXT2_SB(sb);
250 	struct ext2_super_block *es = sbi->s_es;
251 	unsigned long def_mount_opts;
252 
253 	spin_lock(&sbi->s_lock);
254 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
255 
256 	if (sbi->s_sb_block != 1)
257 		seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
258 	if (test_opt(sb, MINIX_DF))
259 		seq_puts(seq, ",minixdf");
260 	if (test_opt(sb, GRPID))
261 		seq_puts(seq, ",grpid");
262 	if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
263 		seq_puts(seq, ",nogrpid");
264 	if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT2_DEF_RESUID)) ||
265 	    le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
266 		seq_printf(seq, ",resuid=%u",
267 				from_kuid_munged(&init_user_ns, sbi->s_resuid));
268 	}
269 	if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT2_DEF_RESGID)) ||
270 	    le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
271 		seq_printf(seq, ",resgid=%u",
272 				from_kgid_munged(&init_user_ns, sbi->s_resgid));
273 	}
274 	if (test_opt(sb, ERRORS_RO)) {
275 		int def_errors = le16_to_cpu(es->s_errors);
276 
277 		if (def_errors == EXT2_ERRORS_PANIC ||
278 		    def_errors == EXT2_ERRORS_CONTINUE) {
279 			seq_puts(seq, ",errors=remount-ro");
280 		}
281 	}
282 	if (test_opt(sb, ERRORS_CONT))
283 		seq_puts(seq, ",errors=continue");
284 	if (test_opt(sb, ERRORS_PANIC))
285 		seq_puts(seq, ",errors=panic");
286 	if (test_opt(sb, NO_UID32))
287 		seq_puts(seq, ",nouid32");
288 	if (test_opt(sb, DEBUG))
289 		seq_puts(seq, ",debug");
290 	if (test_opt(sb, OLDALLOC))
291 		seq_puts(seq, ",oldalloc");
292 
293 #ifdef CONFIG_EXT2_FS_XATTR
294 	if (test_opt(sb, XATTR_USER))
295 		seq_puts(seq, ",user_xattr");
296 	if (!test_opt(sb, XATTR_USER) &&
297 	    (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
298 		seq_puts(seq, ",nouser_xattr");
299 	}
300 #endif
301 
302 #ifdef CONFIG_EXT2_FS_POSIX_ACL
303 	if (test_opt(sb, POSIX_ACL))
304 		seq_puts(seq, ",acl");
305 	if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
306 		seq_puts(seq, ",noacl");
307 #endif
308 
309 	if (test_opt(sb, NOBH))
310 		seq_puts(seq, ",nobh");
311 
312 #if defined(CONFIG_QUOTA)
313 	if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
314 		seq_puts(seq, ",usrquota");
315 
316 	if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
317 		seq_puts(seq, ",grpquota");
318 #endif
319 
320 #ifdef CONFIG_FS_DAX
321 	if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
322 		seq_puts(seq, ",xip");
323 	if (sbi->s_mount_opt & EXT2_MOUNT_DAX)
324 		seq_puts(seq, ",dax");
325 #endif
326 
327 	if (!test_opt(sb, RESERVATION))
328 		seq_puts(seq, ",noreservation");
329 
330 	spin_unlock(&sbi->s_lock);
331 	return 0;
332 }
333 
334 #ifdef CONFIG_QUOTA
335 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
336 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
337 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
338 			 const struct path *path);
339 static struct dquot **ext2_get_dquots(struct inode *inode)
340 {
341 	return EXT2_I(inode)->i_dquot;
342 }
343 
344 static const struct quotactl_ops ext2_quotactl_ops = {
345 	.quota_on	= ext2_quota_on,
346 	.quota_off	= ext2_quota_off,
347 	.quota_sync	= dquot_quota_sync,
348 	.get_state	= dquot_get_state,
349 	.set_info	= dquot_set_dqinfo,
350 	.get_dqblk	= dquot_get_dqblk,
351 	.set_dqblk	= dquot_set_dqblk,
352 	.get_nextdqblk	= dquot_get_next_dqblk,
353 };
354 #endif
355 
356 static const struct super_operations ext2_sops = {
357 	.alloc_inode	= ext2_alloc_inode,
358 	.destroy_inode	= ext2_destroy_inode,
359 	.write_inode	= ext2_write_inode,
360 	.evict_inode	= ext2_evict_inode,
361 	.put_super	= ext2_put_super,
362 	.sync_fs	= ext2_sync_fs,
363 	.freeze_fs	= ext2_freeze,
364 	.unfreeze_fs	= ext2_unfreeze,
365 	.statfs		= ext2_statfs,
366 	.remount_fs	= ext2_remount,
367 	.show_options	= ext2_show_options,
368 #ifdef CONFIG_QUOTA
369 	.quota_read	= ext2_quota_read,
370 	.quota_write	= ext2_quota_write,
371 	.get_dquots	= ext2_get_dquots,
372 #endif
373 };
374 
375 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
376 		u64 ino, u32 generation)
377 {
378 	struct inode *inode;
379 
380 	if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
381 		return ERR_PTR(-ESTALE);
382 	if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
383 		return ERR_PTR(-ESTALE);
384 
385 	/*
386 	 * ext2_iget isn't quite right if the inode is currently unallocated!
387 	 * However ext2_iget currently does appropriate checks to handle stale
388 	 * inodes so everything is OK.
389 	 */
390 	inode = ext2_iget(sb, ino);
391 	if (IS_ERR(inode))
392 		return ERR_CAST(inode);
393 	if (generation && inode->i_generation != generation) {
394 		/* we didn't find the right inode.. */
395 		iput(inode);
396 		return ERR_PTR(-ESTALE);
397 	}
398 	return inode;
399 }
400 
401 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
402 		int fh_len, int fh_type)
403 {
404 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
405 				    ext2_nfs_get_inode);
406 }
407 
408 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
409 		int fh_len, int fh_type)
410 {
411 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
412 				    ext2_nfs_get_inode);
413 }
414 
415 static const struct export_operations ext2_export_ops = {
416 	.fh_to_dentry = ext2_fh_to_dentry,
417 	.fh_to_parent = ext2_fh_to_parent,
418 	.get_parent = ext2_get_parent,
419 };
420 
421 static unsigned long get_sb_block(void **data)
422 {
423 	unsigned long 	sb_block;
424 	char 		*options = (char *) *data;
425 
426 	if (!options || strncmp(options, "sb=", 3) != 0)
427 		return 1;	/* Default location */
428 	options += 3;
429 	sb_block = simple_strtoul(options, &options, 0);
430 	if (*options && *options != ',') {
431 		printk("EXT2-fs: Invalid sb specification: %s\n",
432 		       (char *) *data);
433 		return 1;
434 	}
435 	if (*options == ',')
436 		options++;
437 	*data = (void *) options;
438 	return sb_block;
439 }
440 
441 enum {
442 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
443 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
444 	Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
445 	Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
446 	Opt_acl, Opt_noacl, Opt_xip, Opt_dax, Opt_ignore, Opt_err, Opt_quota,
447 	Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
448 };
449 
450 static const match_table_t tokens = {
451 	{Opt_bsd_df, "bsddf"},
452 	{Opt_minix_df, "minixdf"},
453 	{Opt_grpid, "grpid"},
454 	{Opt_grpid, "bsdgroups"},
455 	{Opt_nogrpid, "nogrpid"},
456 	{Opt_nogrpid, "sysvgroups"},
457 	{Opt_resgid, "resgid=%u"},
458 	{Opt_resuid, "resuid=%u"},
459 	{Opt_sb, "sb=%u"},
460 	{Opt_err_cont, "errors=continue"},
461 	{Opt_err_panic, "errors=panic"},
462 	{Opt_err_ro, "errors=remount-ro"},
463 	{Opt_nouid32, "nouid32"},
464 	{Opt_nocheck, "check=none"},
465 	{Opt_nocheck, "nocheck"},
466 	{Opt_debug, "debug"},
467 	{Opt_oldalloc, "oldalloc"},
468 	{Opt_orlov, "orlov"},
469 	{Opt_nobh, "nobh"},
470 	{Opt_user_xattr, "user_xattr"},
471 	{Opt_nouser_xattr, "nouser_xattr"},
472 	{Opt_acl, "acl"},
473 	{Opt_noacl, "noacl"},
474 	{Opt_xip, "xip"},
475 	{Opt_dax, "dax"},
476 	{Opt_grpquota, "grpquota"},
477 	{Opt_ignore, "noquota"},
478 	{Opt_quota, "quota"},
479 	{Opt_usrquota, "usrquota"},
480 	{Opt_reservation, "reservation"},
481 	{Opt_noreservation, "noreservation"},
482 	{Opt_err, NULL}
483 };
484 
485 static int parse_options(char *options, struct super_block *sb,
486 			 struct ext2_mount_options *opts)
487 {
488 	char *p;
489 	substring_t args[MAX_OPT_ARGS];
490 	int option;
491 	kuid_t uid;
492 	kgid_t gid;
493 
494 	if (!options)
495 		return 1;
496 
497 	while ((p = strsep (&options, ",")) != NULL) {
498 		int token;
499 		if (!*p)
500 			continue;
501 
502 		token = match_token(p, tokens, args);
503 		switch (token) {
504 		case Opt_bsd_df:
505 			clear_opt (opts->s_mount_opt, MINIX_DF);
506 			break;
507 		case Opt_minix_df:
508 			set_opt (opts->s_mount_opt, MINIX_DF);
509 			break;
510 		case Opt_grpid:
511 			set_opt (opts->s_mount_opt, GRPID);
512 			break;
513 		case Opt_nogrpid:
514 			clear_opt (opts->s_mount_opt, GRPID);
515 			break;
516 		case Opt_resuid:
517 			if (match_int(&args[0], &option))
518 				return 0;
519 			uid = make_kuid(current_user_ns(), option);
520 			if (!uid_valid(uid)) {
521 				ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
522 				return 0;
523 
524 			}
525 			opts->s_resuid = uid;
526 			break;
527 		case Opt_resgid:
528 			if (match_int(&args[0], &option))
529 				return 0;
530 			gid = make_kgid(current_user_ns(), option);
531 			if (!gid_valid(gid)) {
532 				ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
533 				return 0;
534 			}
535 			opts->s_resgid = gid;
536 			break;
537 		case Opt_sb:
538 			/* handled by get_sb_block() instead of here */
539 			/* *sb_block = match_int(&args[0]); */
540 			break;
541 		case Opt_err_panic:
542 			clear_opt (opts->s_mount_opt, ERRORS_CONT);
543 			clear_opt (opts->s_mount_opt, ERRORS_RO);
544 			set_opt (opts->s_mount_opt, ERRORS_PANIC);
545 			break;
546 		case Opt_err_ro:
547 			clear_opt (opts->s_mount_opt, ERRORS_CONT);
548 			clear_opt (opts->s_mount_opt, ERRORS_PANIC);
549 			set_opt (opts->s_mount_opt, ERRORS_RO);
550 			break;
551 		case Opt_err_cont:
552 			clear_opt (opts->s_mount_opt, ERRORS_RO);
553 			clear_opt (opts->s_mount_opt, ERRORS_PANIC);
554 			set_opt (opts->s_mount_opt, ERRORS_CONT);
555 			break;
556 		case Opt_nouid32:
557 			set_opt (opts->s_mount_opt, NO_UID32);
558 			break;
559 		case Opt_nocheck:
560 			ext2_msg(sb, KERN_WARNING,
561 				"Option nocheck/check=none is deprecated and"
562 				" will be removed in June 2020.");
563 			clear_opt (opts->s_mount_opt, CHECK);
564 			break;
565 		case Opt_debug:
566 			set_opt (opts->s_mount_opt, DEBUG);
567 			break;
568 		case Opt_oldalloc:
569 			set_opt (opts->s_mount_opt, OLDALLOC);
570 			break;
571 		case Opt_orlov:
572 			clear_opt (opts->s_mount_opt, OLDALLOC);
573 			break;
574 		case Opt_nobh:
575 			set_opt (opts->s_mount_opt, NOBH);
576 			break;
577 #ifdef CONFIG_EXT2_FS_XATTR
578 		case Opt_user_xattr:
579 			set_opt (opts->s_mount_opt, XATTR_USER);
580 			break;
581 		case Opt_nouser_xattr:
582 			clear_opt (opts->s_mount_opt, XATTR_USER);
583 			break;
584 #else
585 		case Opt_user_xattr:
586 		case Opt_nouser_xattr:
587 			ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
588 				"not supported");
589 			break;
590 #endif
591 #ifdef CONFIG_EXT2_FS_POSIX_ACL
592 		case Opt_acl:
593 			set_opt(opts->s_mount_opt, POSIX_ACL);
594 			break;
595 		case Opt_noacl:
596 			clear_opt(opts->s_mount_opt, POSIX_ACL);
597 			break;
598 #else
599 		case Opt_acl:
600 		case Opt_noacl:
601 			ext2_msg(sb, KERN_INFO,
602 				"(no)acl options not supported");
603 			break;
604 #endif
605 		case Opt_xip:
606 			ext2_msg(sb, KERN_INFO, "use dax instead of xip");
607 			set_opt(opts->s_mount_opt, XIP);
608 			/* Fall through */
609 		case Opt_dax:
610 #ifdef CONFIG_FS_DAX
611 			ext2_msg(sb, KERN_WARNING,
612 		"DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
613 			set_opt(opts->s_mount_opt, DAX);
614 #else
615 			ext2_msg(sb, KERN_INFO, "dax option not supported");
616 #endif
617 			break;
618 
619 #if defined(CONFIG_QUOTA)
620 		case Opt_quota:
621 		case Opt_usrquota:
622 			set_opt(opts->s_mount_opt, USRQUOTA);
623 			break;
624 
625 		case Opt_grpquota:
626 			set_opt(opts->s_mount_opt, GRPQUOTA);
627 			break;
628 #else
629 		case Opt_quota:
630 		case Opt_usrquota:
631 		case Opt_grpquota:
632 			ext2_msg(sb, KERN_INFO,
633 				"quota operations not supported");
634 			break;
635 #endif
636 
637 		case Opt_reservation:
638 			set_opt(opts->s_mount_opt, RESERVATION);
639 			ext2_msg(sb, KERN_INFO, "reservations ON");
640 			break;
641 		case Opt_noreservation:
642 			clear_opt(opts->s_mount_opt, RESERVATION);
643 			ext2_msg(sb, KERN_INFO, "reservations OFF");
644 			break;
645 		case Opt_ignore:
646 			break;
647 		default:
648 			return 0;
649 		}
650 	}
651 	return 1;
652 }
653 
654 static int ext2_setup_super (struct super_block * sb,
655 			      struct ext2_super_block * es,
656 			      int read_only)
657 {
658 	int res = 0;
659 	struct ext2_sb_info *sbi = EXT2_SB(sb);
660 
661 	if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
662 		ext2_msg(sb, KERN_ERR,
663 			"error: revision level too high, "
664 			"forcing read-only mode");
665 		res = SB_RDONLY;
666 	}
667 	if (read_only)
668 		return res;
669 	if (!(sbi->s_mount_state & EXT2_VALID_FS))
670 		ext2_msg(sb, KERN_WARNING,
671 			"warning: mounting unchecked fs, "
672 			"running e2fsck is recommended");
673 	else if ((sbi->s_mount_state & EXT2_ERROR_FS))
674 		ext2_msg(sb, KERN_WARNING,
675 			"warning: mounting fs with errors, "
676 			"running e2fsck is recommended");
677 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
678 		 le16_to_cpu(es->s_mnt_count) >=
679 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
680 		ext2_msg(sb, KERN_WARNING,
681 			"warning: maximal mount count reached, "
682 			"running e2fsck is recommended");
683 	else if (le32_to_cpu(es->s_checkinterval) &&
684 		(le32_to_cpu(es->s_lastcheck) +
685 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
686 		ext2_msg(sb, KERN_WARNING,
687 			"warning: checktime reached, "
688 			"running e2fsck is recommended");
689 	if (!le16_to_cpu(es->s_max_mnt_count))
690 		es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
691 	le16_add_cpu(&es->s_mnt_count, 1);
692 	if (test_opt (sb, DEBUG))
693 		ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
694 			"bpg=%lu, ipg=%lu, mo=%04lx]",
695 			EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
696 			sbi->s_frag_size,
697 			sbi->s_groups_count,
698 			EXT2_BLOCKS_PER_GROUP(sb),
699 			EXT2_INODES_PER_GROUP(sb),
700 			sbi->s_mount_opt);
701 	return res;
702 }
703 
704 static int ext2_check_descriptors(struct super_block *sb)
705 {
706 	int i;
707 	struct ext2_sb_info *sbi = EXT2_SB(sb);
708 
709 	ext2_debug ("Checking group descriptors");
710 
711 	for (i = 0; i < sbi->s_groups_count; i++) {
712 		struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
713 		ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
714 		ext2_fsblk_t last_block;
715 
716 		if (i == sbi->s_groups_count - 1)
717 			last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
718 		else
719 			last_block = first_block +
720 				(EXT2_BLOCKS_PER_GROUP(sb) - 1);
721 
722 		if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
723 		    le32_to_cpu(gdp->bg_block_bitmap) > last_block)
724 		{
725 			ext2_error (sb, "ext2_check_descriptors",
726 				    "Block bitmap for group %d"
727 				    " not in group (block %lu)!",
728 				    i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
729 			return 0;
730 		}
731 		if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
732 		    le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
733 		{
734 			ext2_error (sb, "ext2_check_descriptors",
735 				    "Inode bitmap for group %d"
736 				    " not in group (block %lu)!",
737 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
738 			return 0;
739 		}
740 		if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
741 		    le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
742 		    last_block)
743 		{
744 			ext2_error (sb, "ext2_check_descriptors",
745 				    "Inode table for group %d"
746 				    " not in group (block %lu)!",
747 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
748 			return 0;
749 		}
750 	}
751 	return 1;
752 }
753 
754 /*
755  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
756  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
757  * We need to be 1 filesystem block less than the 2^32 sector limit.
758  */
759 static loff_t ext2_max_size(int bits)
760 {
761 	loff_t res = EXT2_NDIR_BLOCKS;
762 	int meta_blocks;
763 	loff_t upper_limit;
764 
765 	/* This is calculated to be the largest file size for a
766 	 * dense, file such that the total number of
767 	 * sectors in the file, including data and all indirect blocks,
768 	 * does not exceed 2^32 -1
769 	 * __u32 i_blocks representing the total number of
770 	 * 512 bytes blocks of the file
771 	 */
772 	upper_limit = (1LL << 32) - 1;
773 
774 	/* total blocks in file system block size */
775 	upper_limit >>= (bits - 9);
776 
777 
778 	/* indirect blocks */
779 	meta_blocks = 1;
780 	/* double indirect blocks */
781 	meta_blocks += 1 + (1LL << (bits-2));
782 	/* tripple indirect blocks */
783 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
784 
785 	upper_limit -= meta_blocks;
786 	upper_limit <<= bits;
787 
788 	res += 1LL << (bits-2);
789 	res += 1LL << (2*(bits-2));
790 	res += 1LL << (3*(bits-2));
791 	res <<= bits;
792 	if (res > upper_limit)
793 		res = upper_limit;
794 
795 	if (res > MAX_LFS_FILESIZE)
796 		res = MAX_LFS_FILESIZE;
797 
798 	return res;
799 }
800 
801 static unsigned long descriptor_loc(struct super_block *sb,
802 				    unsigned long logic_sb_block,
803 				    int nr)
804 {
805 	struct ext2_sb_info *sbi = EXT2_SB(sb);
806 	unsigned long bg, first_meta_bg;
807 	int has_super = 0;
808 
809 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
810 
811 	if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
812 	    nr < first_meta_bg)
813 		return (logic_sb_block + nr + 1);
814 	bg = sbi->s_desc_per_block * nr;
815 	if (ext2_bg_has_super(sb, bg))
816 		has_super = 1;
817 
818 	return ext2_group_first_block_no(sb, bg) + has_super;
819 }
820 
821 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
822 {
823 	struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
824 	struct buffer_head * bh;
825 	struct ext2_sb_info * sbi;
826 	struct ext2_super_block * es;
827 	struct inode *root;
828 	unsigned long block;
829 	unsigned long sb_block = get_sb_block(&data);
830 	unsigned long logic_sb_block;
831 	unsigned long offset = 0;
832 	unsigned long def_mount_opts;
833 	long ret = -ENOMEM;
834 	int blocksize = BLOCK_SIZE;
835 	int db_count;
836 	int i, j;
837 	__le32 features;
838 	int err;
839 	struct ext2_mount_options opts;
840 
841 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
842 	if (!sbi)
843 		goto failed;
844 
845 	sbi->s_blockgroup_lock =
846 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
847 	if (!sbi->s_blockgroup_lock) {
848 		kfree(sbi);
849 		goto failed;
850 	}
851 	sb->s_fs_info = sbi;
852 	sbi->s_sb_block = sb_block;
853 	sbi->s_daxdev = dax_dev;
854 
855 	spin_lock_init(&sbi->s_lock);
856 	ret = -EINVAL;
857 
858 	/*
859 	 * See what the current blocksize for the device is, and
860 	 * use that as the blocksize.  Otherwise (or if the blocksize
861 	 * is smaller than the default) use the default.
862 	 * This is important for devices that have a hardware
863 	 * sectorsize that is larger than the default.
864 	 */
865 	blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
866 	if (!blocksize) {
867 		ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
868 		goto failed_sbi;
869 	}
870 
871 	/*
872 	 * If the superblock doesn't start on a hardware sector boundary,
873 	 * calculate the offset.
874 	 */
875 	if (blocksize != BLOCK_SIZE) {
876 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
877 		offset = (sb_block*BLOCK_SIZE) % blocksize;
878 	} else {
879 		logic_sb_block = sb_block;
880 	}
881 
882 	if (!(bh = sb_bread(sb, logic_sb_block))) {
883 		ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
884 		goto failed_sbi;
885 	}
886 	/*
887 	 * Note: s_es must be initialized as soon as possible because
888 	 *       some ext2 macro-instructions depend on its value
889 	 */
890 	es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
891 	sbi->s_es = es;
892 	sb->s_magic = le16_to_cpu(es->s_magic);
893 
894 	if (sb->s_magic != EXT2_SUPER_MAGIC)
895 		goto cantfind_ext2;
896 
897 	/* Set defaults before we parse the mount options */
898 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
899 	if (def_mount_opts & EXT2_DEFM_DEBUG)
900 		set_opt(opts.s_mount_opt, DEBUG);
901 	if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
902 		set_opt(opts.s_mount_opt, GRPID);
903 	if (def_mount_opts & EXT2_DEFM_UID16)
904 		set_opt(opts.s_mount_opt, NO_UID32);
905 #ifdef CONFIG_EXT2_FS_XATTR
906 	if (def_mount_opts & EXT2_DEFM_XATTR_USER)
907 		set_opt(opts.s_mount_opt, XATTR_USER);
908 #endif
909 #ifdef CONFIG_EXT2_FS_POSIX_ACL
910 	if (def_mount_opts & EXT2_DEFM_ACL)
911 		set_opt(opts.s_mount_opt, POSIX_ACL);
912 #endif
913 
914 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
915 		set_opt(opts.s_mount_opt, ERRORS_PANIC);
916 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
917 		set_opt(opts.s_mount_opt, ERRORS_CONT);
918 	else
919 		set_opt(opts.s_mount_opt, ERRORS_RO);
920 
921 	opts.s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
922 	opts.s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
923 
924 	set_opt(opts.s_mount_opt, RESERVATION);
925 
926 	if (!parse_options((char *) data, sb, &opts))
927 		goto failed_mount;
928 
929 	sbi->s_mount_opt = opts.s_mount_opt;
930 	sbi->s_resuid = opts.s_resuid;
931 	sbi->s_resgid = opts.s_resgid;
932 
933 	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
934 		((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
935 		 SB_POSIXACL : 0);
936 	sb->s_iflags |= SB_I_CGROUPWB;
937 
938 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
939 	    (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
940 	     EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
941 	     EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
942 		ext2_msg(sb, KERN_WARNING,
943 			"warning: feature flags set on rev 0 fs, "
944 			"running e2fsck is recommended");
945 	/*
946 	 * Check feature flags regardless of the revision level, since we
947 	 * previously didn't change the revision level when setting the flags,
948 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
949 	 */
950 	features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
951 	if (features) {
952 		ext2_msg(sb, KERN_ERR,	"error: couldn't mount because of "
953 		       "unsupported optional features (%x)",
954 			le32_to_cpu(features));
955 		goto failed_mount;
956 	}
957 	if (!sb_rdonly(sb) && (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
958 		ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
959 		       "unsupported optional features (%x)",
960 		       le32_to_cpu(features));
961 		goto failed_mount;
962 	}
963 
964 	blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
965 
966 	if (sbi->s_mount_opt & EXT2_MOUNT_DAX) {
967 		if (!bdev_dax_supported(sb->s_bdev, blocksize)) {
968 			ext2_msg(sb, KERN_ERR,
969 				"DAX unsupported by block device. Turning off DAX.");
970 			sbi->s_mount_opt &= ~EXT2_MOUNT_DAX;
971 		}
972 	}
973 
974 	/* If the blocksize doesn't match, re-read the thing.. */
975 	if (sb->s_blocksize != blocksize) {
976 		brelse(bh);
977 
978 		if (!sb_set_blocksize(sb, blocksize)) {
979 			ext2_msg(sb, KERN_ERR,
980 				"error: bad blocksize %d", blocksize);
981 			goto failed_sbi;
982 		}
983 
984 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
985 		offset = (sb_block*BLOCK_SIZE) % blocksize;
986 		bh = sb_bread(sb, logic_sb_block);
987 		if(!bh) {
988 			ext2_msg(sb, KERN_ERR, "error: couldn't read"
989 				"superblock on 2nd try");
990 			goto failed_sbi;
991 		}
992 		es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
993 		sbi->s_es = es;
994 		if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
995 			ext2_msg(sb, KERN_ERR, "error: magic mismatch");
996 			goto failed_mount;
997 		}
998 	}
999 
1000 	sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
1001 	sb->s_max_links = EXT2_LINK_MAX;
1002 
1003 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
1004 		sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
1005 		sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
1006 	} else {
1007 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1008 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1009 		if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
1010 		    !is_power_of_2(sbi->s_inode_size) ||
1011 		    (sbi->s_inode_size > blocksize)) {
1012 			ext2_msg(sb, KERN_ERR,
1013 				"error: unsupported inode size: %d",
1014 				sbi->s_inode_size);
1015 			goto failed_mount;
1016 		}
1017 	}
1018 
1019 	sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
1020 				   le32_to_cpu(es->s_log_frag_size);
1021 	if (sbi->s_frag_size == 0)
1022 		goto cantfind_ext2;
1023 	sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
1024 
1025 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1026 	sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1027 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1028 
1029 	if (EXT2_INODE_SIZE(sb) == 0)
1030 		goto cantfind_ext2;
1031 	sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
1032 	if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
1033 		goto cantfind_ext2;
1034 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
1035 					sbi->s_inodes_per_block;
1036 	sbi->s_desc_per_block = sb->s_blocksize /
1037 					sizeof (struct ext2_group_desc);
1038 	sbi->s_sbh = bh;
1039 	sbi->s_mount_state = le16_to_cpu(es->s_state);
1040 	sbi->s_addr_per_block_bits =
1041 		ilog2 (EXT2_ADDR_PER_BLOCK(sb));
1042 	sbi->s_desc_per_block_bits =
1043 		ilog2 (EXT2_DESC_PER_BLOCK(sb));
1044 
1045 	if (sb->s_magic != EXT2_SUPER_MAGIC)
1046 		goto cantfind_ext2;
1047 
1048 	if (sb->s_blocksize != bh->b_size) {
1049 		if (!silent)
1050 			ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
1051 		goto failed_mount;
1052 	}
1053 
1054 	if (sb->s_blocksize != sbi->s_frag_size) {
1055 		ext2_msg(sb, KERN_ERR,
1056 			"error: fragsize %lu != blocksize %lu"
1057 			"(not supported yet)",
1058 			sbi->s_frag_size, sb->s_blocksize);
1059 		goto failed_mount;
1060 	}
1061 
1062 	if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
1063 		ext2_msg(sb, KERN_ERR,
1064 			"error: #blocks per group too big: %lu",
1065 			sbi->s_blocks_per_group);
1066 		goto failed_mount;
1067 	}
1068 	if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
1069 		ext2_msg(sb, KERN_ERR,
1070 			"error: #fragments per group too big: %lu",
1071 			sbi->s_frags_per_group);
1072 		goto failed_mount;
1073 	}
1074 	if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1075 		ext2_msg(sb, KERN_ERR,
1076 			"error: #inodes per group too big: %lu",
1077 			sbi->s_inodes_per_group);
1078 		goto failed_mount;
1079 	}
1080 
1081 	if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1082 		goto cantfind_ext2;
1083  	sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1084  				le32_to_cpu(es->s_first_data_block) - 1)
1085  					/ EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1086 	db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1087 		   EXT2_DESC_PER_BLOCK(sb);
1088 	sbi->s_group_desc = kmalloc_array (db_count,
1089 					   sizeof(struct buffer_head *),
1090 					   GFP_KERNEL);
1091 	if (sbi->s_group_desc == NULL) {
1092 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1093 		goto failed_mount;
1094 	}
1095 	bgl_lock_init(sbi->s_blockgroup_lock);
1096 	sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1097 	if (!sbi->s_debts) {
1098 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1099 		goto failed_mount_group_desc;
1100 	}
1101 	for (i = 0; i < db_count; i++) {
1102 		block = descriptor_loc(sb, logic_sb_block, i);
1103 		sbi->s_group_desc[i] = sb_bread(sb, block);
1104 		if (!sbi->s_group_desc[i]) {
1105 			for (j = 0; j < i; j++)
1106 				brelse (sbi->s_group_desc[j]);
1107 			ext2_msg(sb, KERN_ERR,
1108 				"error: unable to read group descriptors");
1109 			goto failed_mount_group_desc;
1110 		}
1111 	}
1112 	if (!ext2_check_descriptors (sb)) {
1113 		ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1114 		goto failed_mount2;
1115 	}
1116 	sbi->s_gdb_count = db_count;
1117 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1118 	spin_lock_init(&sbi->s_next_gen_lock);
1119 
1120 	/* per fileystem reservation list head & lock */
1121 	spin_lock_init(&sbi->s_rsv_window_lock);
1122 	sbi->s_rsv_window_root = RB_ROOT;
1123 	/*
1124 	 * Add a single, static dummy reservation to the start of the
1125 	 * reservation window list --- it gives us a placeholder for
1126 	 * append-at-start-of-list which makes the allocation logic
1127 	 * _much_ simpler.
1128 	 */
1129 	sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1130 	sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1131 	sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1132 	sbi->s_rsv_window_head.rsv_goal_size = 0;
1133 	ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1134 
1135 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
1136 				ext2_count_free_blocks(sb), GFP_KERNEL);
1137 	if (!err) {
1138 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
1139 				ext2_count_free_inodes(sb), GFP_KERNEL);
1140 	}
1141 	if (!err) {
1142 		err = percpu_counter_init(&sbi->s_dirs_counter,
1143 				ext2_count_dirs(sb), GFP_KERNEL);
1144 	}
1145 	if (err) {
1146 		ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1147 		goto failed_mount3;
1148 	}
1149 
1150 #ifdef CONFIG_EXT2_FS_XATTR
1151 	sbi->s_ea_block_cache = ext2_xattr_create_cache();
1152 	if (!sbi->s_ea_block_cache) {
1153 		ext2_msg(sb, KERN_ERR, "Failed to create ea_block_cache");
1154 		goto failed_mount3;
1155 	}
1156 #endif
1157 	/*
1158 	 * set up enough so that it can read an inode
1159 	 */
1160 	sb->s_op = &ext2_sops;
1161 	sb->s_export_op = &ext2_export_ops;
1162 	sb->s_xattr = ext2_xattr_handlers;
1163 
1164 #ifdef CONFIG_QUOTA
1165 	sb->dq_op = &dquot_operations;
1166 	sb->s_qcop = &ext2_quotactl_ops;
1167 	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1168 #endif
1169 
1170 	root = ext2_iget(sb, EXT2_ROOT_INO);
1171 	if (IS_ERR(root)) {
1172 		ret = PTR_ERR(root);
1173 		goto failed_mount3;
1174 	}
1175 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1176 		iput(root);
1177 		ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1178 		goto failed_mount3;
1179 	}
1180 
1181 	sb->s_root = d_make_root(root);
1182 	if (!sb->s_root) {
1183 		ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1184 		ret = -ENOMEM;
1185 		goto failed_mount3;
1186 	}
1187 	if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1188 		ext2_msg(sb, KERN_WARNING,
1189 			"warning: mounting ext3 filesystem as ext2");
1190 	if (ext2_setup_super (sb, es, sb_rdonly(sb)))
1191 		sb->s_flags |= SB_RDONLY;
1192 	ext2_write_super(sb);
1193 	return 0;
1194 
1195 cantfind_ext2:
1196 	if (!silent)
1197 		ext2_msg(sb, KERN_ERR,
1198 			"error: can't find an ext2 filesystem on dev %s.",
1199 			sb->s_id);
1200 	goto failed_mount;
1201 failed_mount3:
1202 	if (sbi->s_ea_block_cache)
1203 		ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
1204 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
1205 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
1206 	percpu_counter_destroy(&sbi->s_dirs_counter);
1207 failed_mount2:
1208 	for (i = 0; i < db_count; i++)
1209 		brelse(sbi->s_group_desc[i]);
1210 failed_mount_group_desc:
1211 	kfree(sbi->s_group_desc);
1212 	kfree(sbi->s_debts);
1213 failed_mount:
1214 	brelse(bh);
1215 failed_sbi:
1216 	sb->s_fs_info = NULL;
1217 	kfree(sbi->s_blockgroup_lock);
1218 	kfree(sbi);
1219 failed:
1220 	fs_put_dax(dax_dev);
1221 	return ret;
1222 }
1223 
1224 static void ext2_clear_super_error(struct super_block *sb)
1225 {
1226 	struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1227 
1228 	if (buffer_write_io_error(sbh)) {
1229 		/*
1230 		 * Oh, dear.  A previous attempt to write the
1231 		 * superblock failed.  This could happen because the
1232 		 * USB device was yanked out.  Or it could happen to
1233 		 * be a transient write error and maybe the block will
1234 		 * be remapped.  Nothing we can do but to retry the
1235 		 * write and hope for the best.
1236 		 */
1237 		ext2_msg(sb, KERN_ERR,
1238 		       "previous I/O error to superblock detected");
1239 		clear_buffer_write_io_error(sbh);
1240 		set_buffer_uptodate(sbh);
1241 	}
1242 }
1243 
1244 void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1245 		     int wait)
1246 {
1247 	ext2_clear_super_error(sb);
1248 	spin_lock(&EXT2_SB(sb)->s_lock);
1249 	es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1250 	es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1251 	es->s_wtime = cpu_to_le32(get_seconds());
1252 	/* unlock before we do IO */
1253 	spin_unlock(&EXT2_SB(sb)->s_lock);
1254 	mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1255 	if (wait)
1256 		sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1257 }
1258 
1259 /*
1260  * In the second extended file system, it is not necessary to
1261  * write the super block since we use a mapping of the
1262  * disk super block in a buffer.
1263  *
1264  * However, this function is still used to set the fs valid
1265  * flags to 0.  We need to set this flag to 0 since the fs
1266  * may have been checked while mounted and e2fsck may have
1267  * set s_state to EXT2_VALID_FS after some corrections.
1268  */
1269 static int ext2_sync_fs(struct super_block *sb, int wait)
1270 {
1271 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1272 	struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1273 
1274 	/*
1275 	 * Write quota structures to quota file, sync_blockdev() will write
1276 	 * them to disk later
1277 	 */
1278 	dquot_writeback_dquots(sb, -1);
1279 
1280 	spin_lock(&sbi->s_lock);
1281 	if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1282 		ext2_debug("setting valid to 0\n");
1283 		es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1284 	}
1285 	spin_unlock(&sbi->s_lock);
1286 	ext2_sync_super(sb, es, wait);
1287 	return 0;
1288 }
1289 
1290 static int ext2_freeze(struct super_block *sb)
1291 {
1292 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1293 
1294 	/*
1295 	 * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1296 	 * because we have unattached inodes and thus filesystem is not fully
1297 	 * consistent.
1298 	 */
1299 	if (atomic_long_read(&sb->s_remove_count)) {
1300 		ext2_sync_fs(sb, 1);
1301 		return 0;
1302 	}
1303 	/* Set EXT2_FS_VALID flag */
1304 	spin_lock(&sbi->s_lock);
1305 	sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1306 	spin_unlock(&sbi->s_lock);
1307 	ext2_sync_super(sb, sbi->s_es, 1);
1308 
1309 	return 0;
1310 }
1311 
1312 static int ext2_unfreeze(struct super_block *sb)
1313 {
1314 	/* Just write sb to clear EXT2_VALID_FS flag */
1315 	ext2_write_super(sb);
1316 
1317 	return 0;
1318 }
1319 
1320 static void ext2_write_super(struct super_block *sb)
1321 {
1322 	if (!sb_rdonly(sb))
1323 		ext2_sync_fs(sb, 1);
1324 }
1325 
1326 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1327 {
1328 	struct ext2_sb_info * sbi = EXT2_SB(sb);
1329 	struct ext2_super_block * es;
1330 	struct ext2_mount_options new_opts;
1331 	int err;
1332 
1333 	sync_filesystem(sb);
1334 
1335 	spin_lock(&sbi->s_lock);
1336 	new_opts.s_mount_opt = sbi->s_mount_opt;
1337 	new_opts.s_resuid = sbi->s_resuid;
1338 	new_opts.s_resgid = sbi->s_resgid;
1339 	spin_unlock(&sbi->s_lock);
1340 
1341 	if (!parse_options(data, sb, &new_opts))
1342 		return -EINVAL;
1343 
1344 	spin_lock(&sbi->s_lock);
1345 	es = sbi->s_es;
1346 	if ((sbi->s_mount_opt ^ new_opts.s_mount_opt) & EXT2_MOUNT_DAX) {
1347 		ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1348 			 "dax flag with busy inodes while remounting");
1349 		new_opts.s_mount_opt ^= EXT2_MOUNT_DAX;
1350 	}
1351 	if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1352 		goto out_set;
1353 	if (*flags & SB_RDONLY) {
1354 		if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1355 		    !(sbi->s_mount_state & EXT2_VALID_FS))
1356 			goto out_set;
1357 
1358 		/*
1359 		 * OK, we are remounting a valid rw partition rdonly, so set
1360 		 * the rdonly flag and then mark the partition as valid again.
1361 		 */
1362 		es->s_state = cpu_to_le16(sbi->s_mount_state);
1363 		es->s_mtime = cpu_to_le32(get_seconds());
1364 		spin_unlock(&sbi->s_lock);
1365 
1366 		err = dquot_suspend(sb, -1);
1367 		if (err < 0)
1368 			return err;
1369 
1370 		ext2_sync_super(sb, es, 1);
1371 	} else {
1372 		__le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1373 					       ~EXT2_FEATURE_RO_COMPAT_SUPP);
1374 		if (ret) {
1375 			spin_unlock(&sbi->s_lock);
1376 			ext2_msg(sb, KERN_WARNING,
1377 				"warning: couldn't remount RDWR because of "
1378 				"unsupported optional features (%x).",
1379 				le32_to_cpu(ret));
1380 			return -EROFS;
1381 		}
1382 		/*
1383 		 * Mounting a RDONLY partition read-write, so reread and
1384 		 * store the current valid flag.  (It may have been changed
1385 		 * by e2fsck since we originally mounted the partition.)
1386 		 */
1387 		sbi->s_mount_state = le16_to_cpu(es->s_state);
1388 		if (!ext2_setup_super (sb, es, 0))
1389 			sb->s_flags &= ~SB_RDONLY;
1390 		spin_unlock(&sbi->s_lock);
1391 
1392 		ext2_write_super(sb);
1393 
1394 		dquot_resume(sb, -1);
1395 	}
1396 
1397 	spin_lock(&sbi->s_lock);
1398 out_set:
1399 	sbi->s_mount_opt = new_opts.s_mount_opt;
1400 	sbi->s_resuid = new_opts.s_resuid;
1401 	sbi->s_resgid = new_opts.s_resgid;
1402 	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1403 		((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? SB_POSIXACL : 0);
1404 	spin_unlock(&sbi->s_lock);
1405 
1406 	return 0;
1407 }
1408 
1409 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1410 {
1411 	struct super_block *sb = dentry->d_sb;
1412 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1413 	struct ext2_super_block *es = sbi->s_es;
1414 	u64 fsid;
1415 
1416 	spin_lock(&sbi->s_lock);
1417 
1418 	if (test_opt (sb, MINIX_DF))
1419 		sbi->s_overhead_last = 0;
1420 	else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1421 		unsigned long i, overhead = 0;
1422 		smp_rmb();
1423 
1424 		/*
1425 		 * Compute the overhead (FS structures). This is constant
1426 		 * for a given filesystem unless the number of block groups
1427 		 * changes so we cache the previous value until it does.
1428 		 */
1429 
1430 		/*
1431 		 * All of the blocks before first_data_block are
1432 		 * overhead
1433 		 */
1434 		overhead = le32_to_cpu(es->s_first_data_block);
1435 
1436 		/*
1437 		 * Add the overhead attributed to the superblock and
1438 		 * block group descriptors.  If the sparse superblocks
1439 		 * feature is turned on, then not all groups have this.
1440 		 */
1441 		for (i = 0; i < sbi->s_groups_count; i++)
1442 			overhead += ext2_bg_has_super(sb, i) +
1443 				ext2_bg_num_gdb(sb, i);
1444 
1445 		/*
1446 		 * Every block group has an inode bitmap, a block
1447 		 * bitmap, and an inode table.
1448 		 */
1449 		overhead += (sbi->s_groups_count *
1450 			     (2 + sbi->s_itb_per_group));
1451 		sbi->s_overhead_last = overhead;
1452 		smp_wmb();
1453 		sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1454 	}
1455 
1456 	buf->f_type = EXT2_SUPER_MAGIC;
1457 	buf->f_bsize = sb->s_blocksize;
1458 	buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1459 	buf->f_bfree = ext2_count_free_blocks(sb);
1460 	es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1461 	buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1462 	if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1463 		buf->f_bavail = 0;
1464 	buf->f_files = le32_to_cpu(es->s_inodes_count);
1465 	buf->f_ffree = ext2_count_free_inodes(sb);
1466 	es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1467 	buf->f_namelen = EXT2_NAME_LEN;
1468 	fsid = le64_to_cpup((void *)es->s_uuid) ^
1469 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1470 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1471 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1472 	spin_unlock(&sbi->s_lock);
1473 	return 0;
1474 }
1475 
1476 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1477 	int flags, const char *dev_name, void *data)
1478 {
1479 	return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1480 }
1481 
1482 #ifdef CONFIG_QUOTA
1483 
1484 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1485  * acquiring the locks... As quota files are never truncated and quota code
1486  * itself serializes the operations (and no one else should touch the files)
1487  * we don't have to be afraid of races */
1488 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1489 			       size_t len, loff_t off)
1490 {
1491 	struct inode *inode = sb_dqopt(sb)->files[type];
1492 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1493 	int err = 0;
1494 	int offset = off & (sb->s_blocksize - 1);
1495 	int tocopy;
1496 	size_t toread;
1497 	struct buffer_head tmp_bh;
1498 	struct buffer_head *bh;
1499 	loff_t i_size = i_size_read(inode);
1500 
1501 	if (off > i_size)
1502 		return 0;
1503 	if (off+len > i_size)
1504 		len = i_size-off;
1505 	toread = len;
1506 	while (toread > 0) {
1507 		tocopy = sb->s_blocksize - offset < toread ?
1508 				sb->s_blocksize - offset : toread;
1509 
1510 		tmp_bh.b_state = 0;
1511 		tmp_bh.b_size = sb->s_blocksize;
1512 		err = ext2_get_block(inode, blk, &tmp_bh, 0);
1513 		if (err < 0)
1514 			return err;
1515 		if (!buffer_mapped(&tmp_bh))	/* A hole? */
1516 			memset(data, 0, tocopy);
1517 		else {
1518 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1519 			if (!bh)
1520 				return -EIO;
1521 			memcpy(data, bh->b_data+offset, tocopy);
1522 			brelse(bh);
1523 		}
1524 		offset = 0;
1525 		toread -= tocopy;
1526 		data += tocopy;
1527 		blk++;
1528 	}
1529 	return len;
1530 }
1531 
1532 /* Write to quotafile */
1533 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1534 				const char *data, size_t len, loff_t off)
1535 {
1536 	struct inode *inode = sb_dqopt(sb)->files[type];
1537 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1538 	int err = 0;
1539 	int offset = off & (sb->s_blocksize - 1);
1540 	int tocopy;
1541 	size_t towrite = len;
1542 	struct buffer_head tmp_bh;
1543 	struct buffer_head *bh;
1544 
1545 	while (towrite > 0) {
1546 		tocopy = sb->s_blocksize - offset < towrite ?
1547 				sb->s_blocksize - offset : towrite;
1548 
1549 		tmp_bh.b_state = 0;
1550 		tmp_bh.b_size = sb->s_blocksize;
1551 		err = ext2_get_block(inode, blk, &tmp_bh, 1);
1552 		if (err < 0)
1553 			goto out;
1554 		if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1555 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1556 		else
1557 			bh = sb_getblk(sb, tmp_bh.b_blocknr);
1558 		if (unlikely(!bh)) {
1559 			err = -EIO;
1560 			goto out;
1561 		}
1562 		lock_buffer(bh);
1563 		memcpy(bh->b_data+offset, data, tocopy);
1564 		flush_dcache_page(bh->b_page);
1565 		set_buffer_uptodate(bh);
1566 		mark_buffer_dirty(bh);
1567 		unlock_buffer(bh);
1568 		brelse(bh);
1569 		offset = 0;
1570 		towrite -= tocopy;
1571 		data += tocopy;
1572 		blk++;
1573 	}
1574 out:
1575 	if (len == towrite)
1576 		return err;
1577 	if (inode->i_size < off+len-towrite)
1578 		i_size_write(inode, off+len-towrite);
1579 	inode_inc_iversion(inode);
1580 	inode->i_mtime = inode->i_ctime = current_time(inode);
1581 	mark_inode_dirty(inode);
1582 	return len - towrite;
1583 }
1584 
1585 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
1586 			 const struct path *path)
1587 {
1588 	int err;
1589 	struct inode *inode;
1590 
1591 	err = dquot_quota_on(sb, type, format_id, path);
1592 	if (err)
1593 		return err;
1594 
1595 	inode = d_inode(path->dentry);
1596 	inode_lock(inode);
1597 	EXT2_I(inode)->i_flags |= EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL;
1598 	inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1599 			S_NOATIME | S_IMMUTABLE);
1600 	inode_unlock(inode);
1601 	mark_inode_dirty(inode);
1602 
1603 	return 0;
1604 }
1605 
1606 static int ext2_quota_off(struct super_block *sb, int type)
1607 {
1608 	struct inode *inode = sb_dqopt(sb)->files[type];
1609 	int err;
1610 
1611 	if (!inode || !igrab(inode))
1612 		goto out;
1613 
1614 	err = dquot_quota_off(sb, type);
1615 	if (err)
1616 		goto out_put;
1617 
1618 	inode_lock(inode);
1619 	EXT2_I(inode)->i_flags &= ~(EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL);
1620 	inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1621 	inode_unlock(inode);
1622 	mark_inode_dirty(inode);
1623 out_put:
1624 	iput(inode);
1625 	return err;
1626 out:
1627 	return dquot_quota_off(sb, type);
1628 }
1629 
1630 #endif
1631 
1632 static struct file_system_type ext2_fs_type = {
1633 	.owner		= THIS_MODULE,
1634 	.name		= "ext2",
1635 	.mount		= ext2_mount,
1636 	.kill_sb	= kill_block_super,
1637 	.fs_flags	= FS_REQUIRES_DEV,
1638 };
1639 MODULE_ALIAS_FS("ext2");
1640 
1641 static int __init init_ext2_fs(void)
1642 {
1643 	int err;
1644 
1645 	err = init_inodecache();
1646 	if (err)
1647 		return err;
1648         err = register_filesystem(&ext2_fs_type);
1649 	if (err)
1650 		goto out;
1651 	return 0;
1652 out:
1653 	destroy_inodecache();
1654 	return err;
1655 }
1656 
1657 static void __exit exit_ext2_fs(void)
1658 {
1659 	unregister_filesystem(&ext2_fs_type);
1660 	destroy_inodecache();
1661 }
1662 
1663 MODULE_AUTHOR("Remy Card and others");
1664 MODULE_DESCRIPTION("Second Extended Filesystem");
1665 MODULE_LICENSE("GPL");
1666 module_init(init_ext2_fs)
1667 module_exit(exit_ext2_fs)
1668