xref: /openbmc/linux/fs/ext2/super.c (revision ac8b6f14)
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 (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
313 		seq_puts(seq, ",usrquota");
314 
315 	if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
316 		seq_puts(seq, ",grpquota");
317 
318 	if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
319 		seq_puts(seq, ",xip");
320 
321 	if (sbi->s_mount_opt & EXT2_MOUNT_DAX)
322 		seq_puts(seq, ",dax");
323 
324 	if (!test_opt(sb, RESERVATION))
325 		seq_puts(seq, ",noreservation");
326 
327 	spin_unlock(&sbi->s_lock);
328 	return 0;
329 }
330 
331 #ifdef CONFIG_QUOTA
332 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
333 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
334 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
335 			 const struct path *path);
336 static struct dquot **ext2_get_dquots(struct inode *inode)
337 {
338 	return EXT2_I(inode)->i_dquot;
339 }
340 
341 static const struct quotactl_ops ext2_quotactl_ops = {
342 	.quota_on	= ext2_quota_on,
343 	.quota_off	= ext2_quota_off,
344 	.quota_sync	= dquot_quota_sync,
345 	.get_state	= dquot_get_state,
346 	.set_info	= dquot_set_dqinfo,
347 	.get_dqblk	= dquot_get_dqblk,
348 	.set_dqblk	= dquot_set_dqblk,
349 	.get_nextdqblk	= dquot_get_next_dqblk,
350 };
351 #endif
352 
353 static const struct super_operations ext2_sops = {
354 	.alloc_inode	= ext2_alloc_inode,
355 	.destroy_inode	= ext2_destroy_inode,
356 	.write_inode	= ext2_write_inode,
357 	.evict_inode	= ext2_evict_inode,
358 	.put_super	= ext2_put_super,
359 	.sync_fs	= ext2_sync_fs,
360 	.freeze_fs	= ext2_freeze,
361 	.unfreeze_fs	= ext2_unfreeze,
362 	.statfs		= ext2_statfs,
363 	.remount_fs	= ext2_remount,
364 	.show_options	= ext2_show_options,
365 #ifdef CONFIG_QUOTA
366 	.quota_read	= ext2_quota_read,
367 	.quota_write	= ext2_quota_write,
368 	.get_dquots	= ext2_get_dquots,
369 #endif
370 };
371 
372 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
373 		u64 ino, u32 generation)
374 {
375 	struct inode *inode;
376 
377 	if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
378 		return ERR_PTR(-ESTALE);
379 	if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
380 		return ERR_PTR(-ESTALE);
381 
382 	/*
383 	 * ext2_iget isn't quite right if the inode is currently unallocated!
384 	 * However ext2_iget currently does appropriate checks to handle stale
385 	 * inodes so everything is OK.
386 	 */
387 	inode = ext2_iget(sb, ino);
388 	if (IS_ERR(inode))
389 		return ERR_CAST(inode);
390 	if (generation && inode->i_generation != generation) {
391 		/* we didn't find the right inode.. */
392 		iput(inode);
393 		return ERR_PTR(-ESTALE);
394 	}
395 	return inode;
396 }
397 
398 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
399 		int fh_len, int fh_type)
400 {
401 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
402 				    ext2_nfs_get_inode);
403 }
404 
405 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
406 		int fh_len, int fh_type)
407 {
408 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
409 				    ext2_nfs_get_inode);
410 }
411 
412 static const struct export_operations ext2_export_ops = {
413 	.fh_to_dentry = ext2_fh_to_dentry,
414 	.fh_to_parent = ext2_fh_to_parent,
415 	.get_parent = ext2_get_parent,
416 };
417 
418 static unsigned long get_sb_block(void **data)
419 {
420 	unsigned long 	sb_block;
421 	char 		*options = (char *) *data;
422 
423 	if (!options || strncmp(options, "sb=", 3) != 0)
424 		return 1;	/* Default location */
425 	options += 3;
426 	sb_block = simple_strtoul(options, &options, 0);
427 	if (*options && *options != ',') {
428 		printk("EXT2-fs: Invalid sb specification: %s\n",
429 		       (char *) *data);
430 		return 1;
431 	}
432 	if (*options == ',')
433 		options++;
434 	*data = (void *) options;
435 	return sb_block;
436 }
437 
438 enum {
439 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
440 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
441 	Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
442 	Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
443 	Opt_acl, Opt_noacl, Opt_xip, Opt_dax, Opt_ignore, Opt_err, Opt_quota,
444 	Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
445 };
446 
447 static const match_table_t tokens = {
448 	{Opt_bsd_df, "bsddf"},
449 	{Opt_minix_df, "minixdf"},
450 	{Opt_grpid, "grpid"},
451 	{Opt_grpid, "bsdgroups"},
452 	{Opt_nogrpid, "nogrpid"},
453 	{Opt_nogrpid, "sysvgroups"},
454 	{Opt_resgid, "resgid=%u"},
455 	{Opt_resuid, "resuid=%u"},
456 	{Opt_sb, "sb=%u"},
457 	{Opt_err_cont, "errors=continue"},
458 	{Opt_err_panic, "errors=panic"},
459 	{Opt_err_ro, "errors=remount-ro"},
460 	{Opt_nouid32, "nouid32"},
461 	{Opt_nocheck, "check=none"},
462 	{Opt_nocheck, "nocheck"},
463 	{Opt_debug, "debug"},
464 	{Opt_oldalloc, "oldalloc"},
465 	{Opt_orlov, "orlov"},
466 	{Opt_nobh, "nobh"},
467 	{Opt_user_xattr, "user_xattr"},
468 	{Opt_nouser_xattr, "nouser_xattr"},
469 	{Opt_acl, "acl"},
470 	{Opt_noacl, "noacl"},
471 	{Opt_xip, "xip"},
472 	{Opt_dax, "dax"},
473 	{Opt_grpquota, "grpquota"},
474 	{Opt_ignore, "noquota"},
475 	{Opt_quota, "quota"},
476 	{Opt_usrquota, "usrquota"},
477 	{Opt_reservation, "reservation"},
478 	{Opt_noreservation, "noreservation"},
479 	{Opt_err, NULL}
480 };
481 
482 static int parse_options(char *options, struct super_block *sb,
483 			 struct ext2_mount_options *opts)
484 {
485 	char *p;
486 	substring_t args[MAX_OPT_ARGS];
487 	int option;
488 	kuid_t uid;
489 	kgid_t gid;
490 
491 	if (!options)
492 		return 1;
493 
494 	while ((p = strsep (&options, ",")) != NULL) {
495 		int token;
496 		if (!*p)
497 			continue;
498 
499 		token = match_token(p, tokens, args);
500 		switch (token) {
501 		case Opt_bsd_df:
502 			clear_opt (opts->s_mount_opt, MINIX_DF);
503 			break;
504 		case Opt_minix_df:
505 			set_opt (opts->s_mount_opt, MINIX_DF);
506 			break;
507 		case Opt_grpid:
508 			set_opt (opts->s_mount_opt, GRPID);
509 			break;
510 		case Opt_nogrpid:
511 			clear_opt (opts->s_mount_opt, GRPID);
512 			break;
513 		case Opt_resuid:
514 			if (match_int(&args[0], &option))
515 				return 0;
516 			uid = make_kuid(current_user_ns(), option);
517 			if (!uid_valid(uid)) {
518 				ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
519 				return 0;
520 
521 			}
522 			opts->s_resuid = uid;
523 			break;
524 		case Opt_resgid:
525 			if (match_int(&args[0], &option))
526 				return 0;
527 			gid = make_kgid(current_user_ns(), option);
528 			if (!gid_valid(gid)) {
529 				ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
530 				return 0;
531 			}
532 			opts->s_resgid = gid;
533 			break;
534 		case Opt_sb:
535 			/* handled by get_sb_block() instead of here */
536 			/* *sb_block = match_int(&args[0]); */
537 			break;
538 		case Opt_err_panic:
539 			clear_opt (opts->s_mount_opt, ERRORS_CONT);
540 			clear_opt (opts->s_mount_opt, ERRORS_RO);
541 			set_opt (opts->s_mount_opt, ERRORS_PANIC);
542 			break;
543 		case Opt_err_ro:
544 			clear_opt (opts->s_mount_opt, ERRORS_CONT);
545 			clear_opt (opts->s_mount_opt, ERRORS_PANIC);
546 			set_opt (opts->s_mount_opt, ERRORS_RO);
547 			break;
548 		case Opt_err_cont:
549 			clear_opt (opts->s_mount_opt, ERRORS_RO);
550 			clear_opt (opts->s_mount_opt, ERRORS_PANIC);
551 			set_opt (opts->s_mount_opt, ERRORS_CONT);
552 			break;
553 		case Opt_nouid32:
554 			set_opt (opts->s_mount_opt, NO_UID32);
555 			break;
556 		case Opt_nocheck:
557 			ext2_msg(sb, KERN_WARNING,
558 				"Option nocheck/check=none is deprecated and"
559 				" will be removed in June 2020.");
560 			clear_opt (opts->s_mount_opt, CHECK);
561 			break;
562 		case Opt_debug:
563 			set_opt (opts->s_mount_opt, DEBUG);
564 			break;
565 		case Opt_oldalloc:
566 			set_opt (opts->s_mount_opt, OLDALLOC);
567 			break;
568 		case Opt_orlov:
569 			clear_opt (opts->s_mount_opt, OLDALLOC);
570 			break;
571 		case Opt_nobh:
572 			set_opt (opts->s_mount_opt, NOBH);
573 			break;
574 #ifdef CONFIG_EXT2_FS_XATTR
575 		case Opt_user_xattr:
576 			set_opt (opts->s_mount_opt, XATTR_USER);
577 			break;
578 		case Opt_nouser_xattr:
579 			clear_opt (opts->s_mount_opt, XATTR_USER);
580 			break;
581 #else
582 		case Opt_user_xattr:
583 		case Opt_nouser_xattr:
584 			ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
585 				"not supported");
586 			break;
587 #endif
588 #ifdef CONFIG_EXT2_FS_POSIX_ACL
589 		case Opt_acl:
590 			set_opt(opts->s_mount_opt, POSIX_ACL);
591 			break;
592 		case Opt_noacl:
593 			clear_opt(opts->s_mount_opt, POSIX_ACL);
594 			break;
595 #else
596 		case Opt_acl:
597 		case Opt_noacl:
598 			ext2_msg(sb, KERN_INFO,
599 				"(no)acl options not supported");
600 			break;
601 #endif
602 		case Opt_xip:
603 			ext2_msg(sb, KERN_INFO, "use dax instead of xip");
604 			set_opt(opts->s_mount_opt, XIP);
605 			/* Fall through */
606 		case Opt_dax:
607 #ifdef CONFIG_FS_DAX
608 			ext2_msg(sb, KERN_WARNING,
609 		"DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
610 			set_opt(opts->s_mount_opt, DAX);
611 #else
612 			ext2_msg(sb, KERN_INFO, "dax option not supported");
613 #endif
614 			break;
615 
616 #if defined(CONFIG_QUOTA)
617 		case Opt_quota:
618 		case Opt_usrquota:
619 			set_opt(opts->s_mount_opt, USRQUOTA);
620 			break;
621 
622 		case Opt_grpquota:
623 			set_opt(opts->s_mount_opt, GRPQUOTA);
624 			break;
625 #else
626 		case Opt_quota:
627 		case Opt_usrquota:
628 		case Opt_grpquota:
629 			ext2_msg(sb, KERN_INFO,
630 				"quota operations not supported");
631 			break;
632 #endif
633 
634 		case Opt_reservation:
635 			set_opt(opts->s_mount_opt, RESERVATION);
636 			ext2_msg(sb, KERN_INFO, "reservations ON");
637 			break;
638 		case Opt_noreservation:
639 			clear_opt(opts->s_mount_opt, RESERVATION);
640 			ext2_msg(sb, KERN_INFO, "reservations OFF");
641 			break;
642 		case Opt_ignore:
643 			break;
644 		default:
645 			return 0;
646 		}
647 	}
648 	return 1;
649 }
650 
651 static int ext2_setup_super (struct super_block * sb,
652 			      struct ext2_super_block * es,
653 			      int read_only)
654 {
655 	int res = 0;
656 	struct ext2_sb_info *sbi = EXT2_SB(sb);
657 
658 	if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
659 		ext2_msg(sb, KERN_ERR,
660 			"error: revision level too high, "
661 			"forcing read-only mode");
662 		res = SB_RDONLY;
663 	}
664 	if (read_only)
665 		return res;
666 	if (!(sbi->s_mount_state & EXT2_VALID_FS))
667 		ext2_msg(sb, KERN_WARNING,
668 			"warning: mounting unchecked fs, "
669 			"running e2fsck is recommended");
670 	else if ((sbi->s_mount_state & EXT2_ERROR_FS))
671 		ext2_msg(sb, KERN_WARNING,
672 			"warning: mounting fs with errors, "
673 			"running e2fsck is recommended");
674 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
675 		 le16_to_cpu(es->s_mnt_count) >=
676 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
677 		ext2_msg(sb, KERN_WARNING,
678 			"warning: maximal mount count reached, "
679 			"running e2fsck is recommended");
680 	else if (le32_to_cpu(es->s_checkinterval) &&
681 		(le32_to_cpu(es->s_lastcheck) +
682 			le32_to_cpu(es->s_checkinterval) <=
683 			ktime_get_real_seconds()))
684 		ext2_msg(sb, KERN_WARNING,
685 			"warning: checktime reached, "
686 			"running e2fsck is recommended");
687 	if (!le16_to_cpu(es->s_max_mnt_count))
688 		es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
689 	le16_add_cpu(&es->s_mnt_count, 1);
690 	if (test_opt (sb, DEBUG))
691 		ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
692 			"bpg=%lu, ipg=%lu, mo=%04lx]",
693 			EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
694 			sbi->s_frag_size,
695 			sbi->s_groups_count,
696 			EXT2_BLOCKS_PER_GROUP(sb),
697 			EXT2_INODES_PER_GROUP(sb),
698 			sbi->s_mount_opt);
699 	return res;
700 }
701 
702 static int ext2_check_descriptors(struct super_block *sb)
703 {
704 	int i;
705 	struct ext2_sb_info *sbi = EXT2_SB(sb);
706 
707 	ext2_debug ("Checking group descriptors");
708 
709 	for (i = 0; i < sbi->s_groups_count; i++) {
710 		struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
711 		ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
712 		ext2_fsblk_t last_block;
713 
714 		if (i == sbi->s_groups_count - 1)
715 			last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
716 		else
717 			last_block = first_block +
718 				(EXT2_BLOCKS_PER_GROUP(sb) - 1);
719 
720 		if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
721 		    le32_to_cpu(gdp->bg_block_bitmap) > last_block)
722 		{
723 			ext2_error (sb, "ext2_check_descriptors",
724 				    "Block bitmap for group %d"
725 				    " not in group (block %lu)!",
726 				    i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
727 			return 0;
728 		}
729 		if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
730 		    le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
731 		{
732 			ext2_error (sb, "ext2_check_descriptors",
733 				    "Inode bitmap for group %d"
734 				    " not in group (block %lu)!",
735 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
736 			return 0;
737 		}
738 		if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
739 		    le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
740 		    last_block)
741 		{
742 			ext2_error (sb, "ext2_check_descriptors",
743 				    "Inode table for group %d"
744 				    " not in group (block %lu)!",
745 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
746 			return 0;
747 		}
748 	}
749 	return 1;
750 }
751 
752 /*
753  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
754  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
755  * We need to be 1 filesystem block less than the 2^32 sector limit.
756  */
757 static loff_t ext2_max_size(int bits)
758 {
759 	loff_t res = EXT2_NDIR_BLOCKS;
760 	int meta_blocks;
761 	loff_t upper_limit;
762 
763 	/* This is calculated to be the largest file size for a
764 	 * dense, file such that the total number of
765 	 * sectors in the file, including data and all indirect blocks,
766 	 * does not exceed 2^32 -1
767 	 * __u32 i_blocks representing the total number of
768 	 * 512 bytes blocks of the file
769 	 */
770 	upper_limit = (1LL << 32) - 1;
771 
772 	/* total blocks in file system block size */
773 	upper_limit >>= (bits - 9);
774 
775 
776 	/* indirect blocks */
777 	meta_blocks = 1;
778 	/* double indirect blocks */
779 	meta_blocks += 1 + (1LL << (bits-2));
780 	/* tripple indirect blocks */
781 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
782 
783 	upper_limit -= meta_blocks;
784 	upper_limit <<= bits;
785 
786 	res += 1LL << (bits-2);
787 	res += 1LL << (2*(bits-2));
788 	res += 1LL << (3*(bits-2));
789 	res <<= bits;
790 	if (res > upper_limit)
791 		res = upper_limit;
792 
793 	if (res > MAX_LFS_FILESIZE)
794 		res = MAX_LFS_FILESIZE;
795 
796 	return res;
797 }
798 
799 static unsigned long descriptor_loc(struct super_block *sb,
800 				    unsigned long logic_sb_block,
801 				    int nr)
802 {
803 	struct ext2_sb_info *sbi = EXT2_SB(sb);
804 	unsigned long bg, first_meta_bg;
805 	int has_super = 0;
806 
807 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
808 
809 	if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
810 	    nr < first_meta_bg)
811 		return (logic_sb_block + nr + 1);
812 	bg = sbi->s_desc_per_block * nr;
813 	if (ext2_bg_has_super(sb, bg))
814 		has_super = 1;
815 
816 	return ext2_group_first_block_no(sb, bg) + has_super;
817 }
818 
819 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
820 {
821 	struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
822 	struct buffer_head * bh;
823 	struct ext2_sb_info * sbi;
824 	struct ext2_super_block * es;
825 	struct inode *root;
826 	unsigned long block;
827 	unsigned long sb_block = get_sb_block(&data);
828 	unsigned long logic_sb_block;
829 	unsigned long offset = 0;
830 	unsigned long def_mount_opts;
831 	long ret = -ENOMEM;
832 	int blocksize = BLOCK_SIZE;
833 	int db_count;
834 	int i, j;
835 	__le32 features;
836 	int err;
837 	struct ext2_mount_options opts;
838 
839 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
840 	if (!sbi)
841 		goto failed;
842 
843 	sbi->s_blockgroup_lock =
844 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
845 	if (!sbi->s_blockgroup_lock) {
846 		kfree(sbi);
847 		goto failed;
848 	}
849 	sb->s_fs_info = sbi;
850 	sbi->s_sb_block = sb_block;
851 	sbi->s_daxdev = dax_dev;
852 
853 	spin_lock_init(&sbi->s_lock);
854 	ret = -EINVAL;
855 
856 	/*
857 	 * See what the current blocksize for the device is, and
858 	 * use that as the blocksize.  Otherwise (or if the blocksize
859 	 * is smaller than the default) use the default.
860 	 * This is important for devices that have a hardware
861 	 * sectorsize that is larger than the default.
862 	 */
863 	blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
864 	if (!blocksize) {
865 		ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
866 		goto failed_sbi;
867 	}
868 
869 	/*
870 	 * If the superblock doesn't start on a hardware sector boundary,
871 	 * calculate the offset.
872 	 */
873 	if (blocksize != BLOCK_SIZE) {
874 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
875 		offset = (sb_block*BLOCK_SIZE) % blocksize;
876 	} else {
877 		logic_sb_block = sb_block;
878 	}
879 
880 	if (!(bh = sb_bread(sb, logic_sb_block))) {
881 		ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
882 		goto failed_sbi;
883 	}
884 	/*
885 	 * Note: s_es must be initialized as soon as possible because
886 	 *       some ext2 macro-instructions depend on its value
887 	 */
888 	es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
889 	sbi->s_es = es;
890 	sb->s_magic = le16_to_cpu(es->s_magic);
891 
892 	if (sb->s_magic != EXT2_SUPER_MAGIC)
893 		goto cantfind_ext2;
894 
895 	opts.s_mount_opt = 0;
896 	/* Set defaults before we parse the mount options */
897 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
898 	if (def_mount_opts & EXT2_DEFM_DEBUG)
899 		set_opt(opts.s_mount_opt, DEBUG);
900 	if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
901 		set_opt(opts.s_mount_opt, GRPID);
902 	if (def_mount_opts & EXT2_DEFM_UID16)
903 		set_opt(opts.s_mount_opt, NO_UID32);
904 #ifdef CONFIG_EXT2_FS_XATTR
905 	if (def_mount_opts & EXT2_DEFM_XATTR_USER)
906 		set_opt(opts.s_mount_opt, XATTR_USER);
907 #endif
908 #ifdef CONFIG_EXT2_FS_POSIX_ACL
909 	if (def_mount_opts & EXT2_DEFM_ACL)
910 		set_opt(opts.s_mount_opt, POSIX_ACL);
911 #endif
912 
913 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
914 		set_opt(opts.s_mount_opt, ERRORS_PANIC);
915 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
916 		set_opt(opts.s_mount_opt, ERRORS_CONT);
917 	else
918 		set_opt(opts.s_mount_opt, ERRORS_RO);
919 
920 	opts.s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
921 	opts.s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
922 
923 	set_opt(opts.s_mount_opt, RESERVATION);
924 
925 	if (!parse_options((char *) data, sb, &opts))
926 		goto failed_mount;
927 
928 	sbi->s_mount_opt = opts.s_mount_opt;
929 	sbi->s_resuid = opts.s_resuid;
930 	sbi->s_resgid = opts.s_resgid;
931 
932 	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
933 		((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
934 		 SB_POSIXACL : 0);
935 	sb->s_iflags |= SB_I_CGROUPWB;
936 
937 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
938 	    (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
939 	     EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
940 	     EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
941 		ext2_msg(sb, KERN_WARNING,
942 			"warning: feature flags set on rev 0 fs, "
943 			"running e2fsck is recommended");
944 	/*
945 	 * Check feature flags regardless of the revision level, since we
946 	 * previously didn't change the revision level when setting the flags,
947 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
948 	 */
949 	features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
950 	if (features) {
951 		ext2_msg(sb, KERN_ERR,	"error: couldn't mount because of "
952 		       "unsupported optional features (%x)",
953 			le32_to_cpu(features));
954 		goto failed_mount;
955 	}
956 	if (!sb_rdonly(sb) && (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
957 		ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
958 		       "unsupported optional features (%x)",
959 		       le32_to_cpu(features));
960 		goto failed_mount;
961 	}
962 
963 	blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
964 
965 	if (sbi->s_mount_opt & EXT2_MOUNT_DAX) {
966 		if (!bdev_dax_supported(sb->s_bdev, blocksize)) {
967 			ext2_msg(sb, KERN_ERR,
968 				"DAX unsupported by block device. Turning off DAX.");
969 			sbi->s_mount_opt &= ~EXT2_MOUNT_DAX;
970 		}
971 	}
972 
973 	/* If the blocksize doesn't match, re-read the thing.. */
974 	if (sb->s_blocksize != blocksize) {
975 		brelse(bh);
976 
977 		if (!sb_set_blocksize(sb, blocksize)) {
978 			ext2_msg(sb, KERN_ERR,
979 				"error: bad blocksize %d", blocksize);
980 			goto failed_sbi;
981 		}
982 
983 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
984 		offset = (sb_block*BLOCK_SIZE) % blocksize;
985 		bh = sb_bread(sb, logic_sb_block);
986 		if(!bh) {
987 			ext2_msg(sb, KERN_ERR, "error: couldn't read"
988 				"superblock on 2nd try");
989 			goto failed_sbi;
990 		}
991 		es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
992 		sbi->s_es = es;
993 		if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
994 			ext2_msg(sb, KERN_ERR, "error: magic mismatch");
995 			goto failed_mount;
996 		}
997 	}
998 
999 	sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
1000 	sb->s_max_links = EXT2_LINK_MAX;
1001 
1002 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
1003 		sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
1004 		sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
1005 	} else {
1006 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1007 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1008 		if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
1009 		    !is_power_of_2(sbi->s_inode_size) ||
1010 		    (sbi->s_inode_size > blocksize)) {
1011 			ext2_msg(sb, KERN_ERR,
1012 				"error: unsupported inode size: %d",
1013 				sbi->s_inode_size);
1014 			goto failed_mount;
1015 		}
1016 	}
1017 
1018 	sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
1019 				   le32_to_cpu(es->s_log_frag_size);
1020 	if (sbi->s_frag_size == 0)
1021 		goto cantfind_ext2;
1022 	sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
1023 
1024 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1025 	sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1026 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1027 
1028 	if (EXT2_INODE_SIZE(sb) == 0)
1029 		goto cantfind_ext2;
1030 	sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
1031 	if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
1032 		goto cantfind_ext2;
1033 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
1034 					sbi->s_inodes_per_block;
1035 	sbi->s_desc_per_block = sb->s_blocksize /
1036 					sizeof (struct ext2_group_desc);
1037 	sbi->s_sbh = bh;
1038 	sbi->s_mount_state = le16_to_cpu(es->s_state);
1039 	sbi->s_addr_per_block_bits =
1040 		ilog2 (EXT2_ADDR_PER_BLOCK(sb));
1041 	sbi->s_desc_per_block_bits =
1042 		ilog2 (EXT2_DESC_PER_BLOCK(sb));
1043 
1044 	if (sb->s_magic != EXT2_SUPER_MAGIC)
1045 		goto cantfind_ext2;
1046 
1047 	if (sb->s_blocksize != bh->b_size) {
1048 		if (!silent)
1049 			ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
1050 		goto failed_mount;
1051 	}
1052 
1053 	if (sb->s_blocksize != sbi->s_frag_size) {
1054 		ext2_msg(sb, KERN_ERR,
1055 			"error: fragsize %lu != blocksize %lu"
1056 			"(not supported yet)",
1057 			sbi->s_frag_size, sb->s_blocksize);
1058 		goto failed_mount;
1059 	}
1060 
1061 	if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
1062 		ext2_msg(sb, KERN_ERR,
1063 			"error: #blocks per group too big: %lu",
1064 			sbi->s_blocks_per_group);
1065 		goto failed_mount;
1066 	}
1067 	if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
1068 		ext2_msg(sb, KERN_ERR,
1069 			"error: #fragments per group too big: %lu",
1070 			sbi->s_frags_per_group);
1071 		goto failed_mount;
1072 	}
1073 	if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1074 		ext2_msg(sb, KERN_ERR,
1075 			"error: #inodes per group too big: %lu",
1076 			sbi->s_inodes_per_group);
1077 		goto failed_mount;
1078 	}
1079 
1080 	if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1081 		goto cantfind_ext2;
1082  	sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1083  				le32_to_cpu(es->s_first_data_block) - 1)
1084  					/ EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1085 	db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1086 		   EXT2_DESC_PER_BLOCK(sb);
1087 	sbi->s_group_desc = kmalloc_array (db_count,
1088 					   sizeof(struct buffer_head *),
1089 					   GFP_KERNEL);
1090 	if (sbi->s_group_desc == NULL) {
1091 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1092 		goto failed_mount;
1093 	}
1094 	bgl_lock_init(sbi->s_blockgroup_lock);
1095 	sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1096 	if (!sbi->s_debts) {
1097 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1098 		goto failed_mount_group_desc;
1099 	}
1100 	for (i = 0; i < db_count; i++) {
1101 		block = descriptor_loc(sb, logic_sb_block, i);
1102 		sbi->s_group_desc[i] = sb_bread(sb, block);
1103 		if (!sbi->s_group_desc[i]) {
1104 			for (j = 0; j < i; j++)
1105 				brelse (sbi->s_group_desc[j]);
1106 			ext2_msg(sb, KERN_ERR,
1107 				"error: unable to read group descriptors");
1108 			goto failed_mount_group_desc;
1109 		}
1110 	}
1111 	if (!ext2_check_descriptors (sb)) {
1112 		ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1113 		goto failed_mount2;
1114 	}
1115 	sbi->s_gdb_count = db_count;
1116 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1117 	spin_lock_init(&sbi->s_next_gen_lock);
1118 
1119 	/* per fileystem reservation list head & lock */
1120 	spin_lock_init(&sbi->s_rsv_window_lock);
1121 	sbi->s_rsv_window_root = RB_ROOT;
1122 	/*
1123 	 * Add a single, static dummy reservation to the start of the
1124 	 * reservation window list --- it gives us a placeholder for
1125 	 * append-at-start-of-list which makes the allocation logic
1126 	 * _much_ simpler.
1127 	 */
1128 	sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1129 	sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1130 	sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1131 	sbi->s_rsv_window_head.rsv_goal_size = 0;
1132 	ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1133 
1134 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
1135 				ext2_count_free_blocks(sb), GFP_KERNEL);
1136 	if (!err) {
1137 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
1138 				ext2_count_free_inodes(sb), GFP_KERNEL);
1139 	}
1140 	if (!err) {
1141 		err = percpu_counter_init(&sbi->s_dirs_counter,
1142 				ext2_count_dirs(sb), GFP_KERNEL);
1143 	}
1144 	if (err) {
1145 		ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1146 		goto failed_mount3;
1147 	}
1148 
1149 #ifdef CONFIG_EXT2_FS_XATTR
1150 	sbi->s_ea_block_cache = ext2_xattr_create_cache();
1151 	if (!sbi->s_ea_block_cache) {
1152 		ext2_msg(sb, KERN_ERR, "Failed to create ea_block_cache");
1153 		goto failed_mount3;
1154 	}
1155 #endif
1156 	/*
1157 	 * set up enough so that it can read an inode
1158 	 */
1159 	sb->s_op = &ext2_sops;
1160 	sb->s_export_op = &ext2_export_ops;
1161 	sb->s_xattr = ext2_xattr_handlers;
1162 
1163 #ifdef CONFIG_QUOTA
1164 	sb->dq_op = &dquot_operations;
1165 	sb->s_qcop = &ext2_quotactl_ops;
1166 	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1167 #endif
1168 
1169 	root = ext2_iget(sb, EXT2_ROOT_INO);
1170 	if (IS_ERR(root)) {
1171 		ret = PTR_ERR(root);
1172 		goto failed_mount3;
1173 	}
1174 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1175 		iput(root);
1176 		ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1177 		goto failed_mount3;
1178 	}
1179 
1180 	sb->s_root = d_make_root(root);
1181 	if (!sb->s_root) {
1182 		ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1183 		ret = -ENOMEM;
1184 		goto failed_mount3;
1185 	}
1186 	if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1187 		ext2_msg(sb, KERN_WARNING,
1188 			"warning: mounting ext3 filesystem as ext2");
1189 	if (ext2_setup_super (sb, es, sb_rdonly(sb)))
1190 		sb->s_flags |= SB_RDONLY;
1191 	ext2_write_super(sb);
1192 	return 0;
1193 
1194 cantfind_ext2:
1195 	if (!silent)
1196 		ext2_msg(sb, KERN_ERR,
1197 			"error: can't find an ext2 filesystem on dev %s.",
1198 			sb->s_id);
1199 	goto failed_mount;
1200 failed_mount3:
1201 	if (sbi->s_ea_block_cache)
1202 		ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
1203 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
1204 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
1205 	percpu_counter_destroy(&sbi->s_dirs_counter);
1206 failed_mount2:
1207 	for (i = 0; i < db_count; i++)
1208 		brelse(sbi->s_group_desc[i]);
1209 failed_mount_group_desc:
1210 	kfree(sbi->s_group_desc);
1211 	kfree(sbi->s_debts);
1212 failed_mount:
1213 	brelse(bh);
1214 failed_sbi:
1215 	sb->s_fs_info = NULL;
1216 	kfree(sbi->s_blockgroup_lock);
1217 	kfree(sbi);
1218 failed:
1219 	fs_put_dax(dax_dev);
1220 	return ret;
1221 }
1222 
1223 static void ext2_clear_super_error(struct super_block *sb)
1224 {
1225 	struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1226 
1227 	if (buffer_write_io_error(sbh)) {
1228 		/*
1229 		 * Oh, dear.  A previous attempt to write the
1230 		 * superblock failed.  This could happen because the
1231 		 * USB device was yanked out.  Or it could happen to
1232 		 * be a transient write error and maybe the block will
1233 		 * be remapped.  Nothing we can do but to retry the
1234 		 * write and hope for the best.
1235 		 */
1236 		ext2_msg(sb, KERN_ERR,
1237 		       "previous I/O error to superblock detected");
1238 		clear_buffer_write_io_error(sbh);
1239 		set_buffer_uptodate(sbh);
1240 	}
1241 }
1242 
1243 void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1244 		     int wait)
1245 {
1246 	ext2_clear_super_error(sb);
1247 	spin_lock(&EXT2_SB(sb)->s_lock);
1248 	es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1249 	es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1250 	es->s_wtime = cpu_to_le32(ktime_get_real_seconds());
1251 	/* unlock before we do IO */
1252 	spin_unlock(&EXT2_SB(sb)->s_lock);
1253 	mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1254 	if (wait)
1255 		sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1256 }
1257 
1258 /*
1259  * In the second extended file system, it is not necessary to
1260  * write the super block since we use a mapping of the
1261  * disk super block in a buffer.
1262  *
1263  * However, this function is still used to set the fs valid
1264  * flags to 0.  We need to set this flag to 0 since the fs
1265  * may have been checked while mounted and e2fsck may have
1266  * set s_state to EXT2_VALID_FS after some corrections.
1267  */
1268 static int ext2_sync_fs(struct super_block *sb, int wait)
1269 {
1270 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1271 	struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1272 
1273 	/*
1274 	 * Write quota structures to quota file, sync_blockdev() will write
1275 	 * them to disk later
1276 	 */
1277 	dquot_writeback_dquots(sb, -1);
1278 
1279 	spin_lock(&sbi->s_lock);
1280 	if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1281 		ext2_debug("setting valid to 0\n");
1282 		es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1283 	}
1284 	spin_unlock(&sbi->s_lock);
1285 	ext2_sync_super(sb, es, wait);
1286 	return 0;
1287 }
1288 
1289 static int ext2_freeze(struct super_block *sb)
1290 {
1291 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1292 
1293 	/*
1294 	 * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1295 	 * because we have unattached inodes and thus filesystem is not fully
1296 	 * consistent.
1297 	 */
1298 	if (atomic_long_read(&sb->s_remove_count)) {
1299 		ext2_sync_fs(sb, 1);
1300 		return 0;
1301 	}
1302 	/* Set EXT2_FS_VALID flag */
1303 	spin_lock(&sbi->s_lock);
1304 	sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1305 	spin_unlock(&sbi->s_lock);
1306 	ext2_sync_super(sb, sbi->s_es, 1);
1307 
1308 	return 0;
1309 }
1310 
1311 static int ext2_unfreeze(struct super_block *sb)
1312 {
1313 	/* Just write sb to clear EXT2_VALID_FS flag */
1314 	ext2_write_super(sb);
1315 
1316 	return 0;
1317 }
1318 
1319 static void ext2_write_super(struct super_block *sb)
1320 {
1321 	if (!sb_rdonly(sb))
1322 		ext2_sync_fs(sb, 1);
1323 }
1324 
1325 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1326 {
1327 	struct ext2_sb_info * sbi = EXT2_SB(sb);
1328 	struct ext2_super_block * es;
1329 	struct ext2_mount_options new_opts;
1330 	int err;
1331 
1332 	sync_filesystem(sb);
1333 
1334 	spin_lock(&sbi->s_lock);
1335 	new_opts.s_mount_opt = sbi->s_mount_opt;
1336 	new_opts.s_resuid = sbi->s_resuid;
1337 	new_opts.s_resgid = sbi->s_resgid;
1338 	spin_unlock(&sbi->s_lock);
1339 
1340 	if (!parse_options(data, sb, &new_opts))
1341 		return -EINVAL;
1342 
1343 	spin_lock(&sbi->s_lock);
1344 	es = sbi->s_es;
1345 	if ((sbi->s_mount_opt ^ new_opts.s_mount_opt) & EXT2_MOUNT_DAX) {
1346 		ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1347 			 "dax flag with busy inodes while remounting");
1348 		new_opts.s_mount_opt ^= EXT2_MOUNT_DAX;
1349 	}
1350 	if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1351 		goto out_set;
1352 	if (*flags & SB_RDONLY) {
1353 		if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1354 		    !(sbi->s_mount_state & EXT2_VALID_FS))
1355 			goto out_set;
1356 
1357 		/*
1358 		 * OK, we are remounting a valid rw partition rdonly, so set
1359 		 * the rdonly flag and then mark the partition as valid again.
1360 		 */
1361 		es->s_state = cpu_to_le16(sbi->s_mount_state);
1362 		es->s_mtime = cpu_to_le32(ktime_get_real_seconds());
1363 		spin_unlock(&sbi->s_lock);
1364 
1365 		err = dquot_suspend(sb, -1);
1366 		if (err < 0)
1367 			return err;
1368 
1369 		ext2_sync_super(sb, es, 1);
1370 	} else {
1371 		__le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1372 					       ~EXT2_FEATURE_RO_COMPAT_SUPP);
1373 		if (ret) {
1374 			spin_unlock(&sbi->s_lock);
1375 			ext2_msg(sb, KERN_WARNING,
1376 				"warning: couldn't remount RDWR because of "
1377 				"unsupported optional features (%x).",
1378 				le32_to_cpu(ret));
1379 			return -EROFS;
1380 		}
1381 		/*
1382 		 * Mounting a RDONLY partition read-write, so reread and
1383 		 * store the current valid flag.  (It may have been changed
1384 		 * by e2fsck since we originally mounted the partition.)
1385 		 */
1386 		sbi->s_mount_state = le16_to_cpu(es->s_state);
1387 		if (!ext2_setup_super (sb, es, 0))
1388 			sb->s_flags &= ~SB_RDONLY;
1389 		spin_unlock(&sbi->s_lock);
1390 
1391 		ext2_write_super(sb);
1392 
1393 		dquot_resume(sb, -1);
1394 	}
1395 
1396 	spin_lock(&sbi->s_lock);
1397 out_set:
1398 	sbi->s_mount_opt = new_opts.s_mount_opt;
1399 	sbi->s_resuid = new_opts.s_resuid;
1400 	sbi->s_resgid = new_opts.s_resgid;
1401 	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1402 		((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? SB_POSIXACL : 0);
1403 	spin_unlock(&sbi->s_lock);
1404 
1405 	return 0;
1406 }
1407 
1408 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1409 {
1410 	struct super_block *sb = dentry->d_sb;
1411 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1412 	struct ext2_super_block *es = sbi->s_es;
1413 	u64 fsid;
1414 
1415 	spin_lock(&sbi->s_lock);
1416 
1417 	if (test_opt (sb, MINIX_DF))
1418 		sbi->s_overhead_last = 0;
1419 	else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1420 		unsigned long i, overhead = 0;
1421 		smp_rmb();
1422 
1423 		/*
1424 		 * Compute the overhead (FS structures). This is constant
1425 		 * for a given filesystem unless the number of block groups
1426 		 * changes so we cache the previous value until it does.
1427 		 */
1428 
1429 		/*
1430 		 * All of the blocks before first_data_block are
1431 		 * overhead
1432 		 */
1433 		overhead = le32_to_cpu(es->s_first_data_block);
1434 
1435 		/*
1436 		 * Add the overhead attributed to the superblock and
1437 		 * block group descriptors.  If the sparse superblocks
1438 		 * feature is turned on, then not all groups have this.
1439 		 */
1440 		for (i = 0; i < sbi->s_groups_count; i++)
1441 			overhead += ext2_bg_has_super(sb, i) +
1442 				ext2_bg_num_gdb(sb, i);
1443 
1444 		/*
1445 		 * Every block group has an inode bitmap, a block
1446 		 * bitmap, and an inode table.
1447 		 */
1448 		overhead += (sbi->s_groups_count *
1449 			     (2 + sbi->s_itb_per_group));
1450 		sbi->s_overhead_last = overhead;
1451 		smp_wmb();
1452 		sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1453 	}
1454 
1455 	buf->f_type = EXT2_SUPER_MAGIC;
1456 	buf->f_bsize = sb->s_blocksize;
1457 	buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1458 	buf->f_bfree = ext2_count_free_blocks(sb);
1459 	es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1460 	buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1461 	if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1462 		buf->f_bavail = 0;
1463 	buf->f_files = le32_to_cpu(es->s_inodes_count);
1464 	buf->f_ffree = ext2_count_free_inodes(sb);
1465 	es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1466 	buf->f_namelen = EXT2_NAME_LEN;
1467 	fsid = le64_to_cpup((void *)es->s_uuid) ^
1468 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1469 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1470 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1471 	spin_unlock(&sbi->s_lock);
1472 	return 0;
1473 }
1474 
1475 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1476 	int flags, const char *dev_name, void *data)
1477 {
1478 	return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1479 }
1480 
1481 #ifdef CONFIG_QUOTA
1482 
1483 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1484  * acquiring the locks... As quota files are never truncated and quota code
1485  * itself serializes the operations (and no one else should touch the files)
1486  * we don't have to be afraid of races */
1487 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1488 			       size_t len, loff_t off)
1489 {
1490 	struct inode *inode = sb_dqopt(sb)->files[type];
1491 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1492 	int err = 0;
1493 	int offset = off & (sb->s_blocksize - 1);
1494 	int tocopy;
1495 	size_t toread;
1496 	struct buffer_head tmp_bh;
1497 	struct buffer_head *bh;
1498 	loff_t i_size = i_size_read(inode);
1499 
1500 	if (off > i_size)
1501 		return 0;
1502 	if (off+len > i_size)
1503 		len = i_size-off;
1504 	toread = len;
1505 	while (toread > 0) {
1506 		tocopy = sb->s_blocksize - offset < toread ?
1507 				sb->s_blocksize - offset : toread;
1508 
1509 		tmp_bh.b_state = 0;
1510 		tmp_bh.b_size = sb->s_blocksize;
1511 		err = ext2_get_block(inode, blk, &tmp_bh, 0);
1512 		if (err < 0)
1513 			return err;
1514 		if (!buffer_mapped(&tmp_bh))	/* A hole? */
1515 			memset(data, 0, tocopy);
1516 		else {
1517 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1518 			if (!bh)
1519 				return -EIO;
1520 			memcpy(data, bh->b_data+offset, tocopy);
1521 			brelse(bh);
1522 		}
1523 		offset = 0;
1524 		toread -= tocopy;
1525 		data += tocopy;
1526 		blk++;
1527 	}
1528 	return len;
1529 }
1530 
1531 /* Write to quotafile */
1532 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1533 				const char *data, size_t len, loff_t off)
1534 {
1535 	struct inode *inode = sb_dqopt(sb)->files[type];
1536 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1537 	int err = 0;
1538 	int offset = off & (sb->s_blocksize - 1);
1539 	int tocopy;
1540 	size_t towrite = len;
1541 	struct buffer_head tmp_bh;
1542 	struct buffer_head *bh;
1543 
1544 	while (towrite > 0) {
1545 		tocopy = sb->s_blocksize - offset < towrite ?
1546 				sb->s_blocksize - offset : towrite;
1547 
1548 		tmp_bh.b_state = 0;
1549 		tmp_bh.b_size = sb->s_blocksize;
1550 		err = ext2_get_block(inode, blk, &tmp_bh, 1);
1551 		if (err < 0)
1552 			goto out;
1553 		if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1554 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1555 		else
1556 			bh = sb_getblk(sb, tmp_bh.b_blocknr);
1557 		if (unlikely(!bh)) {
1558 			err = -EIO;
1559 			goto out;
1560 		}
1561 		lock_buffer(bh);
1562 		memcpy(bh->b_data+offset, data, tocopy);
1563 		flush_dcache_page(bh->b_page);
1564 		set_buffer_uptodate(bh);
1565 		mark_buffer_dirty(bh);
1566 		unlock_buffer(bh);
1567 		brelse(bh);
1568 		offset = 0;
1569 		towrite -= tocopy;
1570 		data += tocopy;
1571 		blk++;
1572 	}
1573 out:
1574 	if (len == towrite)
1575 		return err;
1576 	if (inode->i_size < off+len-towrite)
1577 		i_size_write(inode, off+len-towrite);
1578 	inode_inc_iversion(inode);
1579 	inode->i_mtime = inode->i_ctime = current_time(inode);
1580 	mark_inode_dirty(inode);
1581 	return len - towrite;
1582 }
1583 
1584 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
1585 			 const struct path *path)
1586 {
1587 	int err;
1588 	struct inode *inode;
1589 
1590 	err = dquot_quota_on(sb, type, format_id, path);
1591 	if (err)
1592 		return err;
1593 
1594 	inode = d_inode(path->dentry);
1595 	inode_lock(inode);
1596 	EXT2_I(inode)->i_flags |= EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL;
1597 	inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1598 			S_NOATIME | S_IMMUTABLE);
1599 	inode_unlock(inode);
1600 	mark_inode_dirty(inode);
1601 
1602 	return 0;
1603 }
1604 
1605 static int ext2_quota_off(struct super_block *sb, int type)
1606 {
1607 	struct inode *inode = sb_dqopt(sb)->files[type];
1608 	int err;
1609 
1610 	if (!inode || !igrab(inode))
1611 		goto out;
1612 
1613 	err = dquot_quota_off(sb, type);
1614 	if (err)
1615 		goto out_put;
1616 
1617 	inode_lock(inode);
1618 	EXT2_I(inode)->i_flags &= ~(EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL);
1619 	inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1620 	inode_unlock(inode);
1621 	mark_inode_dirty(inode);
1622 out_put:
1623 	iput(inode);
1624 	return err;
1625 out:
1626 	return dquot_quota_off(sb, type);
1627 }
1628 
1629 #endif
1630 
1631 static struct file_system_type ext2_fs_type = {
1632 	.owner		= THIS_MODULE,
1633 	.name		= "ext2",
1634 	.mount		= ext2_mount,
1635 	.kill_sb	= kill_block_super,
1636 	.fs_flags	= FS_REQUIRES_DEV,
1637 };
1638 MODULE_ALIAS_FS("ext2");
1639 
1640 static int __init init_ext2_fs(void)
1641 {
1642 	int err;
1643 
1644 	err = init_inodecache();
1645 	if (err)
1646 		return err;
1647         err = register_filesystem(&ext2_fs_type);
1648 	if (err)
1649 		goto out;
1650 	return 0;
1651 out:
1652 	destroy_inodecache();
1653 	return err;
1654 }
1655 
1656 static void __exit exit_ext2_fs(void)
1657 {
1658 	unregister_filesystem(&ext2_fs_type);
1659 	destroy_inodecache();
1660 }
1661 
1662 MODULE_AUTHOR("Remy Card and others");
1663 MODULE_DESCRIPTION("Second Extended Filesystem");
1664 MODULE_LICENSE("GPL");
1665 module_init(init_ext2_fs)
1666 module_exit(exit_ext2_fs)
1667