xref: /linux/fs/ext2/super.c (revision 667ca1ec7c9eb7ac3b80590b6597151b4c2a750b)
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("ext2_inode_cache",
225 					     sizeof(struct ext2_inode_info),
226 					     0, (SLAB_RECLAIM_ACCOUNT|
227 						SLAB_MEM_SPREAD|SLAB_ACCOUNT),
228 					     init_once);
229 	if (ext2_inode_cachep == NULL)
230 		return -ENOMEM;
231 	return 0;
232 }
233 
234 static void destroy_inodecache(void)
235 {
236 	/*
237 	 * Make sure all delayed rcu free inodes are flushed before we
238 	 * destroy cache.
239 	 */
240 	rcu_barrier();
241 	kmem_cache_destroy(ext2_inode_cachep);
242 }
243 
244 static int ext2_show_options(struct seq_file *seq, struct dentry *root)
245 {
246 	struct super_block *sb = root->d_sb;
247 	struct ext2_sb_info *sbi = EXT2_SB(sb);
248 	struct ext2_super_block *es = sbi->s_es;
249 	unsigned long def_mount_opts;
250 
251 	spin_lock(&sbi->s_lock);
252 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
253 
254 	if (sbi->s_sb_block != 1)
255 		seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
256 	if (test_opt(sb, MINIX_DF))
257 		seq_puts(seq, ",minixdf");
258 	if (test_opt(sb, GRPID))
259 		seq_puts(seq, ",grpid");
260 	if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
261 		seq_puts(seq, ",nogrpid");
262 	if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT2_DEF_RESUID)) ||
263 	    le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
264 		seq_printf(seq, ",resuid=%u",
265 				from_kuid_munged(&init_user_ns, sbi->s_resuid));
266 	}
267 	if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT2_DEF_RESGID)) ||
268 	    le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
269 		seq_printf(seq, ",resgid=%u",
270 				from_kgid_munged(&init_user_ns, sbi->s_resgid));
271 	}
272 	if (test_opt(sb, ERRORS_RO)) {
273 		int def_errors = le16_to_cpu(es->s_errors);
274 
275 		if (def_errors == EXT2_ERRORS_PANIC ||
276 		    def_errors == EXT2_ERRORS_CONTINUE) {
277 			seq_puts(seq, ",errors=remount-ro");
278 		}
279 	}
280 	if (test_opt(sb, ERRORS_CONT))
281 		seq_puts(seq, ",errors=continue");
282 	if (test_opt(sb, ERRORS_PANIC))
283 		seq_puts(seq, ",errors=panic");
284 	if (test_opt(sb, NO_UID32))
285 		seq_puts(seq, ",nouid32");
286 	if (test_opt(sb, DEBUG))
287 		seq_puts(seq, ",debug");
288 	if (test_opt(sb, OLDALLOC))
289 		seq_puts(seq, ",oldalloc");
290 
291 #ifdef CONFIG_EXT2_FS_XATTR
292 	if (test_opt(sb, XATTR_USER))
293 		seq_puts(seq, ",user_xattr");
294 	if (!test_opt(sb, XATTR_USER) &&
295 	    (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
296 		seq_puts(seq, ",nouser_xattr");
297 	}
298 #endif
299 
300 #ifdef CONFIG_EXT2_FS_POSIX_ACL
301 	if (test_opt(sb, POSIX_ACL))
302 		seq_puts(seq, ",acl");
303 	if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
304 		seq_puts(seq, ",noacl");
305 #endif
306 
307 	if (test_opt(sb, NOBH))
308 		seq_puts(seq, ",nobh");
309 
310 #if defined(CONFIG_QUOTA)
311 	if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
312 		seq_puts(seq, ",usrquota");
313 
314 	if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
315 		seq_puts(seq, ",grpquota");
316 #endif
317 
318 #ifdef CONFIG_FS_DAX
319 	if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
320 		seq_puts(seq, ",xip");
321 	if (sbi->s_mount_opt & EXT2_MOUNT_DAX)
322 		seq_puts(seq, ",dax");
323 #endif
324 
325 	if (!test_opt(sb, RESERVATION))
326 		seq_puts(seq, ",noreservation");
327 
328 	spin_unlock(&sbi->s_lock);
329 	return 0;
330 }
331 
332 #ifdef CONFIG_QUOTA
333 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
334 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
335 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
336 			 const struct path *path);
337 static struct dquot **ext2_get_dquots(struct inode *inode)
338 {
339 	return EXT2_I(inode)->i_dquot;
340 }
341 
342 static const struct quotactl_ops ext2_quotactl_ops = {
343 	.quota_on	= ext2_quota_on,
344 	.quota_off	= ext2_quota_off,
345 	.quota_sync	= dquot_quota_sync,
346 	.get_state	= dquot_get_state,
347 	.set_info	= dquot_set_dqinfo,
348 	.get_dqblk	= dquot_get_dqblk,
349 	.set_dqblk	= dquot_set_dqblk,
350 	.get_nextdqblk	= dquot_get_next_dqblk,
351 };
352 #endif
353 
354 static const struct super_operations ext2_sops = {
355 	.alloc_inode	= ext2_alloc_inode,
356 	.destroy_inode	= ext2_destroy_inode,
357 	.write_inode	= ext2_write_inode,
358 	.evict_inode	= ext2_evict_inode,
359 	.put_super	= ext2_put_super,
360 	.sync_fs	= ext2_sync_fs,
361 	.freeze_fs	= ext2_freeze,
362 	.unfreeze_fs	= ext2_unfreeze,
363 	.statfs		= ext2_statfs,
364 	.remount_fs	= ext2_remount,
365 	.show_options	= ext2_show_options,
366 #ifdef CONFIG_QUOTA
367 	.quota_read	= ext2_quota_read,
368 	.quota_write	= ext2_quota_write,
369 	.get_dquots	= ext2_get_dquots,
370 #endif
371 };
372 
373 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
374 		u64 ino, u32 generation)
375 {
376 	struct inode *inode;
377 
378 	if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
379 		return ERR_PTR(-ESTALE);
380 	if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
381 		return ERR_PTR(-ESTALE);
382 
383 	/*
384 	 * ext2_iget isn't quite right if the inode is currently unallocated!
385 	 * However ext2_iget currently does appropriate checks to handle stale
386 	 * inodes so everything is OK.
387 	 */
388 	inode = ext2_iget(sb, ino);
389 	if (IS_ERR(inode))
390 		return ERR_CAST(inode);
391 	if (generation && inode->i_generation != generation) {
392 		/* we didn't find the right inode.. */
393 		iput(inode);
394 		return ERR_PTR(-ESTALE);
395 	}
396 	return inode;
397 }
398 
399 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
400 		int fh_len, int fh_type)
401 {
402 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
403 				    ext2_nfs_get_inode);
404 }
405 
406 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
407 		int fh_len, int fh_type)
408 {
409 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
410 				    ext2_nfs_get_inode);
411 }
412 
413 static const struct export_operations ext2_export_ops = {
414 	.fh_to_dentry = ext2_fh_to_dentry,
415 	.fh_to_parent = ext2_fh_to_parent,
416 	.get_parent = ext2_get_parent,
417 };
418 
419 static unsigned long get_sb_block(void **data)
420 {
421 	unsigned long 	sb_block;
422 	char 		*options = (char *) *data;
423 
424 	if (!options || strncmp(options, "sb=", 3) != 0)
425 		return 1;	/* Default location */
426 	options += 3;
427 	sb_block = simple_strtoul(options, &options, 0);
428 	if (*options && *options != ',') {
429 		printk("EXT2-fs: Invalid sb specification: %s\n",
430 		       (char *) *data);
431 		return 1;
432 	}
433 	if (*options == ',')
434 		options++;
435 	*data = (void *) options;
436 	return sb_block;
437 }
438 
439 enum {
440 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
441 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
442 	Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
443 	Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
444 	Opt_acl, Opt_noacl, Opt_xip, Opt_dax, Opt_ignore, Opt_err, Opt_quota,
445 	Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
446 };
447 
448 static const match_table_t tokens = {
449 	{Opt_bsd_df, "bsddf"},
450 	{Opt_minix_df, "minixdf"},
451 	{Opt_grpid, "grpid"},
452 	{Opt_grpid, "bsdgroups"},
453 	{Opt_nogrpid, "nogrpid"},
454 	{Opt_nogrpid, "sysvgroups"},
455 	{Opt_resgid, "resgid=%u"},
456 	{Opt_resuid, "resuid=%u"},
457 	{Opt_sb, "sb=%u"},
458 	{Opt_err_cont, "errors=continue"},
459 	{Opt_err_panic, "errors=panic"},
460 	{Opt_err_ro, "errors=remount-ro"},
461 	{Opt_nouid32, "nouid32"},
462 	{Opt_nocheck, "check=none"},
463 	{Opt_nocheck, "nocheck"},
464 	{Opt_debug, "debug"},
465 	{Opt_oldalloc, "oldalloc"},
466 	{Opt_orlov, "orlov"},
467 	{Opt_nobh, "nobh"},
468 	{Opt_user_xattr, "user_xattr"},
469 	{Opt_nouser_xattr, "nouser_xattr"},
470 	{Opt_acl, "acl"},
471 	{Opt_noacl, "noacl"},
472 	{Opt_xip, "xip"},
473 	{Opt_dax, "dax"},
474 	{Opt_grpquota, "grpquota"},
475 	{Opt_ignore, "noquota"},
476 	{Opt_quota, "quota"},
477 	{Opt_usrquota, "usrquota"},
478 	{Opt_reservation, "reservation"},
479 	{Opt_noreservation, "noreservation"},
480 	{Opt_err, NULL}
481 };
482 
483 static int parse_options(char *options, struct super_block *sb,
484 			 struct ext2_mount_options *opts)
485 {
486 	char *p;
487 	substring_t args[MAX_OPT_ARGS];
488 	int option;
489 	kuid_t uid;
490 	kgid_t gid;
491 
492 	if (!options)
493 		return 1;
494 
495 	while ((p = strsep (&options, ",")) != NULL) {
496 		int token;
497 		if (!*p)
498 			continue;
499 
500 		token = match_token(p, tokens, args);
501 		switch (token) {
502 		case Opt_bsd_df:
503 			clear_opt (opts->s_mount_opt, MINIX_DF);
504 			break;
505 		case Opt_minix_df:
506 			set_opt (opts->s_mount_opt, MINIX_DF);
507 			break;
508 		case Opt_grpid:
509 			set_opt (opts->s_mount_opt, GRPID);
510 			break;
511 		case Opt_nogrpid:
512 			clear_opt (opts->s_mount_opt, GRPID);
513 			break;
514 		case Opt_resuid:
515 			if (match_int(&args[0], &option))
516 				return 0;
517 			uid = make_kuid(current_user_ns(), option);
518 			if (!uid_valid(uid)) {
519 				ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
520 				return 0;
521 
522 			}
523 			opts->s_resuid = uid;
524 			break;
525 		case Opt_resgid:
526 			if (match_int(&args[0], &option))
527 				return 0;
528 			gid = make_kgid(current_user_ns(), option);
529 			if (!gid_valid(gid)) {
530 				ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
531 				return 0;
532 			}
533 			opts->s_resgid = gid;
534 			break;
535 		case Opt_sb:
536 			/* handled by get_sb_block() instead of here */
537 			/* *sb_block = match_int(&args[0]); */
538 			break;
539 		case Opt_err_panic:
540 			clear_opt (opts->s_mount_opt, ERRORS_CONT);
541 			clear_opt (opts->s_mount_opt, ERRORS_RO);
542 			set_opt (opts->s_mount_opt, ERRORS_PANIC);
543 			break;
544 		case Opt_err_ro:
545 			clear_opt (opts->s_mount_opt, ERRORS_CONT);
546 			clear_opt (opts->s_mount_opt, ERRORS_PANIC);
547 			set_opt (opts->s_mount_opt, ERRORS_RO);
548 			break;
549 		case Opt_err_cont:
550 			clear_opt (opts->s_mount_opt, ERRORS_RO);
551 			clear_opt (opts->s_mount_opt, ERRORS_PANIC);
552 			set_opt (opts->s_mount_opt, ERRORS_CONT);
553 			break;
554 		case Opt_nouid32:
555 			set_opt (opts->s_mount_opt, NO_UID32);
556 			break;
557 		case Opt_nocheck:
558 			clear_opt (opts->s_mount_opt, CHECK);
559 			break;
560 		case Opt_debug:
561 			set_opt (opts->s_mount_opt, DEBUG);
562 			break;
563 		case Opt_oldalloc:
564 			set_opt (opts->s_mount_opt, OLDALLOC);
565 			break;
566 		case Opt_orlov:
567 			clear_opt (opts->s_mount_opt, OLDALLOC);
568 			break;
569 		case Opt_nobh:
570 			set_opt (opts->s_mount_opt, NOBH);
571 			break;
572 #ifdef CONFIG_EXT2_FS_XATTR
573 		case Opt_user_xattr:
574 			set_opt (opts->s_mount_opt, XATTR_USER);
575 			break;
576 		case Opt_nouser_xattr:
577 			clear_opt (opts->s_mount_opt, XATTR_USER);
578 			break;
579 #else
580 		case Opt_user_xattr:
581 		case Opt_nouser_xattr:
582 			ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
583 				"not supported");
584 			break;
585 #endif
586 #ifdef CONFIG_EXT2_FS_POSIX_ACL
587 		case Opt_acl:
588 			set_opt(opts->s_mount_opt, POSIX_ACL);
589 			break;
590 		case Opt_noacl:
591 			clear_opt(opts->s_mount_opt, POSIX_ACL);
592 			break;
593 #else
594 		case Opt_acl:
595 		case Opt_noacl:
596 			ext2_msg(sb, KERN_INFO,
597 				"(no)acl options not supported");
598 			break;
599 #endif
600 		case Opt_xip:
601 			ext2_msg(sb, KERN_INFO, "use dax instead of xip");
602 			set_opt(opts->s_mount_opt, XIP);
603 			/* Fall through */
604 		case Opt_dax:
605 #ifdef CONFIG_FS_DAX
606 			ext2_msg(sb, KERN_WARNING,
607 		"DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
608 			set_opt(opts->s_mount_opt, DAX);
609 #else
610 			ext2_msg(sb, KERN_INFO, "dax option not supported");
611 #endif
612 			break;
613 
614 #if defined(CONFIG_QUOTA)
615 		case Opt_quota:
616 		case Opt_usrquota:
617 			set_opt(opts->s_mount_opt, USRQUOTA);
618 			break;
619 
620 		case Opt_grpquota:
621 			set_opt(opts->s_mount_opt, GRPQUOTA);
622 			break;
623 #else
624 		case Opt_quota:
625 		case Opt_usrquota:
626 		case Opt_grpquota:
627 			ext2_msg(sb, KERN_INFO,
628 				"quota operations not supported");
629 			break;
630 #endif
631 
632 		case Opt_reservation:
633 			set_opt(opts->s_mount_opt, RESERVATION);
634 			ext2_msg(sb, KERN_INFO, "reservations ON");
635 			break;
636 		case Opt_noreservation:
637 			clear_opt(opts->s_mount_opt, RESERVATION);
638 			ext2_msg(sb, KERN_INFO, "reservations OFF");
639 			break;
640 		case Opt_ignore:
641 			break;
642 		default:
643 			return 0;
644 		}
645 	}
646 	return 1;
647 }
648 
649 static int ext2_setup_super (struct super_block * sb,
650 			      struct ext2_super_block * es,
651 			      int read_only)
652 {
653 	int res = 0;
654 	struct ext2_sb_info *sbi = EXT2_SB(sb);
655 
656 	if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
657 		ext2_msg(sb, KERN_ERR,
658 			"error: revision level too high, "
659 			"forcing read-only mode");
660 		res = SB_RDONLY;
661 	}
662 	if (read_only)
663 		return res;
664 	if (!(sbi->s_mount_state & EXT2_VALID_FS))
665 		ext2_msg(sb, KERN_WARNING,
666 			"warning: mounting unchecked fs, "
667 			"running e2fsck is recommended");
668 	else if ((sbi->s_mount_state & EXT2_ERROR_FS))
669 		ext2_msg(sb, KERN_WARNING,
670 			"warning: mounting fs with errors, "
671 			"running e2fsck is recommended");
672 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
673 		 le16_to_cpu(es->s_mnt_count) >=
674 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
675 		ext2_msg(sb, KERN_WARNING,
676 			"warning: maximal mount count reached, "
677 			"running e2fsck is recommended");
678 	else if (le32_to_cpu(es->s_checkinterval) &&
679 		(le32_to_cpu(es->s_lastcheck) +
680 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
681 		ext2_msg(sb, KERN_WARNING,
682 			"warning: checktime reached, "
683 			"running e2fsck is recommended");
684 	if (!le16_to_cpu(es->s_max_mnt_count))
685 		es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
686 	le16_add_cpu(&es->s_mnt_count, 1);
687 	if (test_opt (sb, DEBUG))
688 		ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
689 			"bpg=%lu, ipg=%lu, mo=%04lx]",
690 			EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
691 			sbi->s_frag_size,
692 			sbi->s_groups_count,
693 			EXT2_BLOCKS_PER_GROUP(sb),
694 			EXT2_INODES_PER_GROUP(sb),
695 			sbi->s_mount_opt);
696 	return res;
697 }
698 
699 static int ext2_check_descriptors(struct super_block *sb)
700 {
701 	int i;
702 	struct ext2_sb_info *sbi = EXT2_SB(sb);
703 
704 	ext2_debug ("Checking group descriptors");
705 
706 	for (i = 0; i < sbi->s_groups_count; i++) {
707 		struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
708 		ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
709 		ext2_fsblk_t last_block;
710 
711 		if (i == sbi->s_groups_count - 1)
712 			last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
713 		else
714 			last_block = first_block +
715 				(EXT2_BLOCKS_PER_GROUP(sb) - 1);
716 
717 		if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
718 		    le32_to_cpu(gdp->bg_block_bitmap) > last_block)
719 		{
720 			ext2_error (sb, "ext2_check_descriptors",
721 				    "Block bitmap for group %d"
722 				    " not in group (block %lu)!",
723 				    i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
724 			return 0;
725 		}
726 		if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
727 		    le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
728 		{
729 			ext2_error (sb, "ext2_check_descriptors",
730 				    "Inode bitmap for group %d"
731 				    " not in group (block %lu)!",
732 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
733 			return 0;
734 		}
735 		if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
736 		    le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
737 		    last_block)
738 		{
739 			ext2_error (sb, "ext2_check_descriptors",
740 				    "Inode table for group %d"
741 				    " not in group (block %lu)!",
742 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
743 			return 0;
744 		}
745 	}
746 	return 1;
747 }
748 
749 /*
750  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
751  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
752  * We need to be 1 filesystem block less than the 2^32 sector limit.
753  */
754 static loff_t ext2_max_size(int bits)
755 {
756 	loff_t res = EXT2_NDIR_BLOCKS;
757 	int meta_blocks;
758 	loff_t upper_limit;
759 
760 	/* This is calculated to be the largest file size for a
761 	 * dense, file such that the total number of
762 	 * sectors in the file, including data and all indirect blocks,
763 	 * does not exceed 2^32 -1
764 	 * __u32 i_blocks representing the total number of
765 	 * 512 bytes blocks of the file
766 	 */
767 	upper_limit = (1LL << 32) - 1;
768 
769 	/* total blocks in file system block size */
770 	upper_limit >>= (bits - 9);
771 
772 
773 	/* indirect blocks */
774 	meta_blocks = 1;
775 	/* double indirect blocks */
776 	meta_blocks += 1 + (1LL << (bits-2));
777 	/* tripple indirect blocks */
778 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
779 
780 	upper_limit -= meta_blocks;
781 	upper_limit <<= bits;
782 
783 	res += 1LL << (bits-2);
784 	res += 1LL << (2*(bits-2));
785 	res += 1LL << (3*(bits-2));
786 	res <<= bits;
787 	if (res > upper_limit)
788 		res = upper_limit;
789 
790 	if (res > MAX_LFS_FILESIZE)
791 		res = MAX_LFS_FILESIZE;
792 
793 	return res;
794 }
795 
796 static unsigned long descriptor_loc(struct super_block *sb,
797 				    unsigned long logic_sb_block,
798 				    int nr)
799 {
800 	struct ext2_sb_info *sbi = EXT2_SB(sb);
801 	unsigned long bg, first_meta_bg;
802 	int has_super = 0;
803 
804 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
805 
806 	if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
807 	    nr < first_meta_bg)
808 		return (logic_sb_block + nr + 1);
809 	bg = sbi->s_desc_per_block * nr;
810 	if (ext2_bg_has_super(sb, bg))
811 		has_super = 1;
812 
813 	return ext2_group_first_block_no(sb, bg) + has_super;
814 }
815 
816 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
817 {
818 	struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
819 	struct buffer_head * bh;
820 	struct ext2_sb_info * sbi;
821 	struct ext2_super_block * es;
822 	struct inode *root;
823 	unsigned long block;
824 	unsigned long sb_block = get_sb_block(&data);
825 	unsigned long logic_sb_block;
826 	unsigned long offset = 0;
827 	unsigned long def_mount_opts;
828 	long ret = -EINVAL;
829 	int blocksize = BLOCK_SIZE;
830 	int db_count;
831 	int i, j;
832 	__le32 features;
833 	int err;
834 	struct ext2_mount_options opts;
835 
836 	err = -ENOMEM;
837 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
838 	if (!sbi)
839 		goto failed;
840 
841 	sbi->s_blockgroup_lock =
842 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
843 	if (!sbi->s_blockgroup_lock) {
844 		kfree(sbi);
845 		goto failed;
846 	}
847 	sb->s_fs_info = sbi;
848 	sbi->s_sb_block = sb_block;
849 	sbi->s_daxdev = dax_dev;
850 
851 	spin_lock_init(&sbi->s_lock);
852 
853 	/*
854 	 * See what the current blocksize for the device is, and
855 	 * use that as the blocksize.  Otherwise (or if the blocksize
856 	 * is smaller than the default) use the default.
857 	 * This is important for devices that have a hardware
858 	 * sectorsize that is larger than the default.
859 	 */
860 	blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
861 	if (!blocksize) {
862 		ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
863 		goto failed_sbi;
864 	}
865 
866 	/*
867 	 * If the superblock doesn't start on a hardware sector boundary,
868 	 * calculate the offset.
869 	 */
870 	if (blocksize != BLOCK_SIZE) {
871 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
872 		offset = (sb_block*BLOCK_SIZE) % blocksize;
873 	} else {
874 		logic_sb_block = sb_block;
875 	}
876 
877 	if (!(bh = sb_bread(sb, logic_sb_block))) {
878 		ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
879 		goto failed_sbi;
880 	}
881 	/*
882 	 * Note: s_es must be initialized as soon as possible because
883 	 *       some ext2 macro-instructions depend on its value
884 	 */
885 	es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
886 	sbi->s_es = es;
887 	sb->s_magic = le16_to_cpu(es->s_magic);
888 
889 	if (sb->s_magic != EXT2_SUPER_MAGIC)
890 		goto cantfind_ext2;
891 
892 	/* Set defaults before we parse the mount options */
893 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
894 	if (def_mount_opts & EXT2_DEFM_DEBUG)
895 		set_opt(opts.s_mount_opt, DEBUG);
896 	if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
897 		set_opt(opts.s_mount_opt, GRPID);
898 	if (def_mount_opts & EXT2_DEFM_UID16)
899 		set_opt(opts.s_mount_opt, NO_UID32);
900 #ifdef CONFIG_EXT2_FS_XATTR
901 	if (def_mount_opts & EXT2_DEFM_XATTR_USER)
902 		set_opt(opts.s_mount_opt, XATTR_USER);
903 #endif
904 #ifdef CONFIG_EXT2_FS_POSIX_ACL
905 	if (def_mount_opts & EXT2_DEFM_ACL)
906 		set_opt(opts.s_mount_opt, POSIX_ACL);
907 #endif
908 
909 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
910 		set_opt(opts.s_mount_opt, ERRORS_PANIC);
911 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
912 		set_opt(opts.s_mount_opt, ERRORS_CONT);
913 	else
914 		set_opt(opts.s_mount_opt, ERRORS_RO);
915 
916 	opts.s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
917 	opts.s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
918 
919 	set_opt(opts.s_mount_opt, RESERVATION);
920 
921 	if (!parse_options((char *) data, sb, &opts))
922 		goto failed_mount;
923 
924 	sbi->s_mount_opt = opts.s_mount_opt;
925 	sbi->s_resuid = opts.s_resuid;
926 	sbi->s_resgid = opts.s_resgid;
927 
928 	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
929 		((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
930 		 SB_POSIXACL : 0);
931 	sb->s_iflags |= SB_I_CGROUPWB;
932 
933 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
934 	    (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
935 	     EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
936 	     EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
937 		ext2_msg(sb, KERN_WARNING,
938 			"warning: feature flags set on rev 0 fs, "
939 			"running e2fsck is recommended");
940 	/*
941 	 * Check feature flags regardless of the revision level, since we
942 	 * previously didn't change the revision level when setting the flags,
943 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
944 	 */
945 	features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
946 	if (features) {
947 		ext2_msg(sb, KERN_ERR,	"error: couldn't mount because of "
948 		       "unsupported optional features (%x)",
949 			le32_to_cpu(features));
950 		goto failed_mount;
951 	}
952 	if (!sb_rdonly(sb) && (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
953 		ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
954 		       "unsupported optional features (%x)",
955 		       le32_to_cpu(features));
956 		goto failed_mount;
957 	}
958 
959 	blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
960 
961 	if (sbi->s_mount_opt & EXT2_MOUNT_DAX) {
962 		err = bdev_dax_supported(sb, blocksize);
963 		if (err)
964 			goto failed_mount;
965 	}
966 
967 	/* If the blocksize doesn't match, re-read the thing.. */
968 	if (sb->s_blocksize != blocksize) {
969 		brelse(bh);
970 
971 		if (!sb_set_blocksize(sb, blocksize)) {
972 			ext2_msg(sb, KERN_ERR,
973 				"error: bad blocksize %d", blocksize);
974 			goto failed_sbi;
975 		}
976 
977 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
978 		offset = (sb_block*BLOCK_SIZE) % blocksize;
979 		bh = sb_bread(sb, logic_sb_block);
980 		if(!bh) {
981 			ext2_msg(sb, KERN_ERR, "error: couldn't read"
982 				"superblock on 2nd try");
983 			goto failed_sbi;
984 		}
985 		es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
986 		sbi->s_es = es;
987 		if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
988 			ext2_msg(sb, KERN_ERR, "error: magic mismatch");
989 			goto failed_mount;
990 		}
991 	}
992 
993 	sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
994 	sb->s_max_links = EXT2_LINK_MAX;
995 
996 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
997 		sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
998 		sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
999 	} else {
1000 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1001 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1002 		if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
1003 		    !is_power_of_2(sbi->s_inode_size) ||
1004 		    (sbi->s_inode_size > blocksize)) {
1005 			ext2_msg(sb, KERN_ERR,
1006 				"error: unsupported inode size: %d",
1007 				sbi->s_inode_size);
1008 			goto failed_mount;
1009 		}
1010 	}
1011 
1012 	sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
1013 				   le32_to_cpu(es->s_log_frag_size);
1014 	if (sbi->s_frag_size == 0)
1015 		goto cantfind_ext2;
1016 	sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
1017 
1018 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1019 	sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1020 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1021 
1022 	if (EXT2_INODE_SIZE(sb) == 0)
1023 		goto cantfind_ext2;
1024 	sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
1025 	if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
1026 		goto cantfind_ext2;
1027 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
1028 					sbi->s_inodes_per_block;
1029 	sbi->s_desc_per_block = sb->s_blocksize /
1030 					sizeof (struct ext2_group_desc);
1031 	sbi->s_sbh = bh;
1032 	sbi->s_mount_state = le16_to_cpu(es->s_state);
1033 	sbi->s_addr_per_block_bits =
1034 		ilog2 (EXT2_ADDR_PER_BLOCK(sb));
1035 	sbi->s_desc_per_block_bits =
1036 		ilog2 (EXT2_DESC_PER_BLOCK(sb));
1037 
1038 	if (sb->s_magic != EXT2_SUPER_MAGIC)
1039 		goto cantfind_ext2;
1040 
1041 	if (sb->s_blocksize != bh->b_size) {
1042 		if (!silent)
1043 			ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
1044 		goto failed_mount;
1045 	}
1046 
1047 	if (sb->s_blocksize != sbi->s_frag_size) {
1048 		ext2_msg(sb, KERN_ERR,
1049 			"error: fragsize %lu != blocksize %lu"
1050 			"(not supported yet)",
1051 			sbi->s_frag_size, sb->s_blocksize);
1052 		goto failed_mount;
1053 	}
1054 
1055 	if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
1056 		ext2_msg(sb, KERN_ERR,
1057 			"error: #blocks per group too big: %lu",
1058 			sbi->s_blocks_per_group);
1059 		goto failed_mount;
1060 	}
1061 	if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
1062 		ext2_msg(sb, KERN_ERR,
1063 			"error: #fragments per group too big: %lu",
1064 			sbi->s_frags_per_group);
1065 		goto failed_mount;
1066 	}
1067 	if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1068 		ext2_msg(sb, KERN_ERR,
1069 			"error: #inodes per group too big: %lu",
1070 			sbi->s_inodes_per_group);
1071 		goto failed_mount;
1072 	}
1073 
1074 	if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1075 		goto cantfind_ext2;
1076  	sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1077  				le32_to_cpu(es->s_first_data_block) - 1)
1078  					/ EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1079 	db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1080 		   EXT2_DESC_PER_BLOCK(sb);
1081 	sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
1082 	if (sbi->s_group_desc == NULL) {
1083 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1084 		goto failed_mount;
1085 	}
1086 	bgl_lock_init(sbi->s_blockgroup_lock);
1087 	sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1088 	if (!sbi->s_debts) {
1089 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1090 		goto failed_mount_group_desc;
1091 	}
1092 	for (i = 0; i < db_count; i++) {
1093 		block = descriptor_loc(sb, logic_sb_block, i);
1094 		sbi->s_group_desc[i] = sb_bread(sb, block);
1095 		if (!sbi->s_group_desc[i]) {
1096 			for (j = 0; j < i; j++)
1097 				brelse (sbi->s_group_desc[j]);
1098 			ext2_msg(sb, KERN_ERR,
1099 				"error: unable to read group descriptors");
1100 			goto failed_mount_group_desc;
1101 		}
1102 	}
1103 	if (!ext2_check_descriptors (sb)) {
1104 		ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1105 		goto failed_mount2;
1106 	}
1107 	sbi->s_gdb_count = db_count;
1108 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1109 	spin_lock_init(&sbi->s_next_gen_lock);
1110 
1111 	/* per fileystem reservation list head & lock */
1112 	spin_lock_init(&sbi->s_rsv_window_lock);
1113 	sbi->s_rsv_window_root = RB_ROOT;
1114 	/*
1115 	 * Add a single, static dummy reservation to the start of the
1116 	 * reservation window list --- it gives us a placeholder for
1117 	 * append-at-start-of-list which makes the allocation logic
1118 	 * _much_ simpler.
1119 	 */
1120 	sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1121 	sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1122 	sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1123 	sbi->s_rsv_window_head.rsv_goal_size = 0;
1124 	ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1125 
1126 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
1127 				ext2_count_free_blocks(sb), GFP_KERNEL);
1128 	if (!err) {
1129 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
1130 				ext2_count_free_inodes(sb), GFP_KERNEL);
1131 	}
1132 	if (!err) {
1133 		err = percpu_counter_init(&sbi->s_dirs_counter,
1134 				ext2_count_dirs(sb), GFP_KERNEL);
1135 	}
1136 	if (err) {
1137 		ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1138 		goto failed_mount3;
1139 	}
1140 
1141 #ifdef CONFIG_EXT2_FS_XATTR
1142 	sbi->s_ea_block_cache = ext2_xattr_create_cache();
1143 	if (!sbi->s_ea_block_cache) {
1144 		ext2_msg(sb, KERN_ERR, "Failed to create ea_block_cache");
1145 		goto failed_mount3;
1146 	}
1147 #endif
1148 	/*
1149 	 * set up enough so that it can read an inode
1150 	 */
1151 	sb->s_op = &ext2_sops;
1152 	sb->s_export_op = &ext2_export_ops;
1153 	sb->s_xattr = ext2_xattr_handlers;
1154 
1155 #ifdef CONFIG_QUOTA
1156 	sb->dq_op = &dquot_operations;
1157 	sb->s_qcop = &ext2_quotactl_ops;
1158 	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1159 #endif
1160 
1161 	root = ext2_iget(sb, EXT2_ROOT_INO);
1162 	if (IS_ERR(root)) {
1163 		ret = PTR_ERR(root);
1164 		goto failed_mount3;
1165 	}
1166 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1167 		iput(root);
1168 		ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1169 		goto failed_mount3;
1170 	}
1171 
1172 	sb->s_root = d_make_root(root);
1173 	if (!sb->s_root) {
1174 		ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1175 		ret = -ENOMEM;
1176 		goto failed_mount3;
1177 	}
1178 	if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1179 		ext2_msg(sb, KERN_WARNING,
1180 			"warning: mounting ext3 filesystem as ext2");
1181 	if (ext2_setup_super (sb, es, sb_rdonly(sb)))
1182 		sb->s_flags |= SB_RDONLY;
1183 	ext2_write_super(sb);
1184 	return 0;
1185 
1186 cantfind_ext2:
1187 	if (!silent)
1188 		ext2_msg(sb, KERN_ERR,
1189 			"error: can't find an ext2 filesystem on dev %s.",
1190 			sb->s_id);
1191 	goto failed_mount;
1192 failed_mount3:
1193 	if (sbi->s_ea_block_cache)
1194 		ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
1195 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
1196 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
1197 	percpu_counter_destroy(&sbi->s_dirs_counter);
1198 failed_mount2:
1199 	for (i = 0; i < db_count; i++)
1200 		brelse(sbi->s_group_desc[i]);
1201 failed_mount_group_desc:
1202 	kfree(sbi->s_group_desc);
1203 	kfree(sbi->s_debts);
1204 failed_mount:
1205 	brelse(bh);
1206 failed_sbi:
1207 	sb->s_fs_info = NULL;
1208 	kfree(sbi->s_blockgroup_lock);
1209 	kfree(sbi);
1210 failed:
1211 	fs_put_dax(dax_dev);
1212 	return ret;
1213 }
1214 
1215 static void ext2_clear_super_error(struct super_block *sb)
1216 {
1217 	struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1218 
1219 	if (buffer_write_io_error(sbh)) {
1220 		/*
1221 		 * Oh, dear.  A previous attempt to write the
1222 		 * superblock failed.  This could happen because the
1223 		 * USB device was yanked out.  Or it could happen to
1224 		 * be a transient write error and maybe the block will
1225 		 * be remapped.  Nothing we can do but to retry the
1226 		 * write and hope for the best.
1227 		 */
1228 		ext2_msg(sb, KERN_ERR,
1229 		       "previous I/O error to superblock detected\n");
1230 		clear_buffer_write_io_error(sbh);
1231 		set_buffer_uptodate(sbh);
1232 	}
1233 }
1234 
1235 void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1236 		     int wait)
1237 {
1238 	ext2_clear_super_error(sb);
1239 	spin_lock(&EXT2_SB(sb)->s_lock);
1240 	es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1241 	es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1242 	es->s_wtime = cpu_to_le32(get_seconds());
1243 	/* unlock before we do IO */
1244 	spin_unlock(&EXT2_SB(sb)->s_lock);
1245 	mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1246 	if (wait)
1247 		sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1248 }
1249 
1250 /*
1251  * In the second extended file system, it is not necessary to
1252  * write the super block since we use a mapping of the
1253  * disk super block in a buffer.
1254  *
1255  * However, this function is still used to set the fs valid
1256  * flags to 0.  We need to set this flag to 0 since the fs
1257  * may have been checked while mounted and e2fsck may have
1258  * set s_state to EXT2_VALID_FS after some corrections.
1259  */
1260 static int ext2_sync_fs(struct super_block *sb, int wait)
1261 {
1262 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1263 	struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1264 
1265 	/*
1266 	 * Write quota structures to quota file, sync_blockdev() will write
1267 	 * them to disk later
1268 	 */
1269 	dquot_writeback_dquots(sb, -1);
1270 
1271 	spin_lock(&sbi->s_lock);
1272 	if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1273 		ext2_debug("setting valid to 0\n");
1274 		es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1275 	}
1276 	spin_unlock(&sbi->s_lock);
1277 	ext2_sync_super(sb, es, wait);
1278 	return 0;
1279 }
1280 
1281 static int ext2_freeze(struct super_block *sb)
1282 {
1283 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1284 
1285 	/*
1286 	 * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1287 	 * because we have unattached inodes and thus filesystem is not fully
1288 	 * consistent.
1289 	 */
1290 	if (atomic_long_read(&sb->s_remove_count)) {
1291 		ext2_sync_fs(sb, 1);
1292 		return 0;
1293 	}
1294 	/* Set EXT2_FS_VALID flag */
1295 	spin_lock(&sbi->s_lock);
1296 	sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1297 	spin_unlock(&sbi->s_lock);
1298 	ext2_sync_super(sb, sbi->s_es, 1);
1299 
1300 	return 0;
1301 }
1302 
1303 static int ext2_unfreeze(struct super_block *sb)
1304 {
1305 	/* Just write sb to clear EXT2_VALID_FS flag */
1306 	ext2_write_super(sb);
1307 
1308 	return 0;
1309 }
1310 
1311 static void ext2_write_super(struct super_block *sb)
1312 {
1313 	if (!sb_rdonly(sb))
1314 		ext2_sync_fs(sb, 1);
1315 }
1316 
1317 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1318 {
1319 	struct ext2_sb_info * sbi = EXT2_SB(sb);
1320 	struct ext2_super_block * es;
1321 	struct ext2_mount_options new_opts;
1322 	int err;
1323 
1324 	sync_filesystem(sb);
1325 
1326 	spin_lock(&sbi->s_lock);
1327 	new_opts.s_mount_opt = sbi->s_mount_opt;
1328 	new_opts.s_resuid = sbi->s_resuid;
1329 	new_opts.s_resgid = sbi->s_resgid;
1330 	spin_unlock(&sbi->s_lock);
1331 
1332 	/*
1333 	 * Allow the "check" option to be passed as a remount option.
1334 	 */
1335 	if (!parse_options(data, sb, &new_opts))
1336 		return -EINVAL;
1337 
1338 	spin_lock(&sbi->s_lock);
1339 	es = sbi->s_es;
1340 	if ((sbi->s_mount_opt ^ new_opts.s_mount_opt) & EXT2_MOUNT_DAX) {
1341 		ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1342 			 "dax flag with busy inodes while remounting");
1343 		new_opts.s_mount_opt ^= EXT2_MOUNT_DAX;
1344 	}
1345 	if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1346 		goto out_set;
1347 	if (*flags & SB_RDONLY) {
1348 		if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1349 		    !(sbi->s_mount_state & EXT2_VALID_FS))
1350 			goto out_set;
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 			return err;
1363 
1364 		ext2_sync_super(sb, es, 1);
1365 	} else {
1366 		__le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1367 					       ~EXT2_FEATURE_RO_COMPAT_SUPP);
1368 		if (ret) {
1369 			spin_unlock(&sbi->s_lock);
1370 			ext2_msg(sb, KERN_WARNING,
1371 				"warning: couldn't remount RDWR because of "
1372 				"unsupported optional features (%x).",
1373 				le32_to_cpu(ret));
1374 			return -EROFS;
1375 		}
1376 		/*
1377 		 * Mounting a RDONLY partition read-write, so reread and
1378 		 * store the current valid flag.  (It may have been changed
1379 		 * by e2fsck since we originally mounted the partition.)
1380 		 */
1381 		sbi->s_mount_state = le16_to_cpu(es->s_state);
1382 		if (!ext2_setup_super (sb, es, 0))
1383 			sb->s_flags &= ~SB_RDONLY;
1384 		spin_unlock(&sbi->s_lock);
1385 
1386 		ext2_write_super(sb);
1387 
1388 		dquot_resume(sb, -1);
1389 	}
1390 
1391 	spin_lock(&sbi->s_lock);
1392 out_set:
1393 	sbi->s_mount_opt = new_opts.s_mount_opt;
1394 	sbi->s_resuid = new_opts.s_resuid;
1395 	sbi->s_resgid = new_opts.s_resgid;
1396 	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1397 		((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? SB_POSIXACL : 0);
1398 	spin_unlock(&sbi->s_lock);
1399 
1400 	return 0;
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_inc_iversion(inode);
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