xref: /openbmc/linux/fs/ext2/super.c (revision 3557b3fd)
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 (!sb_rdonly(sb) && 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 	ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
152 	sbi->s_ea_block_cache = NULL;
153 
154 	if (!sb_rdonly(sb)) {
155 		struct ext2_super_block *es = sbi->s_es;
156 
157 		spin_lock(&sbi->s_lock);
158 		es->s_state = cpu_to_le16(sbi->s_mount_state);
159 		spin_unlock(&sbi->s_lock);
160 		ext2_sync_super(sb, es, 1);
161 	}
162 	db_count = sbi->s_gdb_count;
163 	for (i = 0; i < db_count; i++)
164 		if (sbi->s_group_desc[i])
165 			brelse (sbi->s_group_desc[i]);
166 	kfree(sbi->s_group_desc);
167 	kfree(sbi->s_debts);
168 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
169 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
170 	percpu_counter_destroy(&sbi->s_dirs_counter);
171 	brelse (sbi->s_sbh);
172 	sb->s_fs_info = NULL;
173 	kfree(sbi->s_blockgroup_lock);
174 	fs_put_dax(sbi->s_daxdev);
175 	kfree(sbi);
176 }
177 
178 static struct kmem_cache * ext2_inode_cachep;
179 
180 static struct inode *ext2_alloc_inode(struct super_block *sb)
181 {
182 	struct ext2_inode_info *ei;
183 	ei = kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
184 	if (!ei)
185 		return NULL;
186 	ei->i_block_alloc_info = NULL;
187 	inode_set_iversion(&ei->vfs_inode, 1);
188 #ifdef CONFIG_QUOTA
189 	memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
190 #endif
191 
192 	return &ei->vfs_inode;
193 }
194 
195 static void ext2_i_callback(struct rcu_head *head)
196 {
197 	struct inode *inode = container_of(head, struct inode, i_rcu);
198 	kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
199 }
200 
201 static void ext2_destroy_inode(struct inode *inode)
202 {
203 	call_rcu(&inode->i_rcu, ext2_i_callback);
204 }
205 
206 static void init_once(void *foo)
207 {
208 	struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
209 
210 	rwlock_init(&ei->i_meta_lock);
211 #ifdef CONFIG_EXT2_FS_XATTR
212 	init_rwsem(&ei->xattr_sem);
213 #endif
214 	mutex_init(&ei->truncate_mutex);
215 #ifdef CONFIG_FS_DAX
216 	init_rwsem(&ei->dax_sem);
217 #endif
218 	inode_init_once(&ei->vfs_inode);
219 }
220 
221 static int __init init_inodecache(void)
222 {
223 	ext2_inode_cachep = kmem_cache_create_usercopy("ext2_inode_cache",
224 				sizeof(struct ext2_inode_info), 0,
225 				(SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
226 					SLAB_ACCOUNT),
227 				offsetof(struct ext2_inode_info, i_data),
228 				sizeof_field(struct ext2_inode_info, i_data),
229 				init_once);
230 	if (ext2_inode_cachep == NULL)
231 		return -ENOMEM;
232 	return 0;
233 }
234 
235 static void destroy_inodecache(void)
236 {
237 	/*
238 	 * Make sure all delayed rcu free inodes are flushed before we
239 	 * destroy cache.
240 	 */
241 	rcu_barrier();
242 	kmem_cache_destroy(ext2_inode_cachep);
243 }
244 
245 static int ext2_show_options(struct seq_file *seq, struct dentry *root)
246 {
247 	struct super_block *sb = root->d_sb;
248 	struct ext2_sb_info *sbi = EXT2_SB(sb);
249 	struct ext2_super_block *es = sbi->s_es;
250 	unsigned long def_mount_opts;
251 
252 	spin_lock(&sbi->s_lock);
253 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
254 
255 	if (sbi->s_sb_block != 1)
256 		seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
257 	if (test_opt(sb, MINIX_DF))
258 		seq_puts(seq, ",minixdf");
259 	if (test_opt(sb, GRPID))
260 		seq_puts(seq, ",grpid");
261 	if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
262 		seq_puts(seq, ",nogrpid");
263 	if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT2_DEF_RESUID)) ||
264 	    le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
265 		seq_printf(seq, ",resuid=%u",
266 				from_kuid_munged(&init_user_ns, sbi->s_resuid));
267 	}
268 	if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT2_DEF_RESGID)) ||
269 	    le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
270 		seq_printf(seq, ",resgid=%u",
271 				from_kgid_munged(&init_user_ns, sbi->s_resgid));
272 	}
273 	if (test_opt(sb, ERRORS_RO)) {
274 		int def_errors = le16_to_cpu(es->s_errors);
275 
276 		if (def_errors == EXT2_ERRORS_PANIC ||
277 		    def_errors == EXT2_ERRORS_CONTINUE) {
278 			seq_puts(seq, ",errors=remount-ro");
279 		}
280 	}
281 	if (test_opt(sb, ERRORS_CONT))
282 		seq_puts(seq, ",errors=continue");
283 	if (test_opt(sb, ERRORS_PANIC))
284 		seq_puts(seq, ",errors=panic");
285 	if (test_opt(sb, NO_UID32))
286 		seq_puts(seq, ",nouid32");
287 	if (test_opt(sb, DEBUG))
288 		seq_puts(seq, ",debug");
289 	if (test_opt(sb, OLDALLOC))
290 		seq_puts(seq, ",oldalloc");
291 
292 #ifdef CONFIG_EXT2_FS_XATTR
293 	if (test_opt(sb, XATTR_USER))
294 		seq_puts(seq, ",user_xattr");
295 	if (!test_opt(sb, XATTR_USER) &&
296 	    (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
297 		seq_puts(seq, ",nouser_xattr");
298 	}
299 #endif
300 
301 #ifdef CONFIG_EXT2_FS_POSIX_ACL
302 	if (test_opt(sb, POSIX_ACL))
303 		seq_puts(seq, ",acl");
304 	if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
305 		seq_puts(seq, ",noacl");
306 #endif
307 
308 	if (test_opt(sb, NOBH))
309 		seq_puts(seq, ",nobh");
310 
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 
317 	if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
318 		seq_puts(seq, ",xip");
319 
320 	if (sbi->s_mount_opt & EXT2_MOUNT_DAX)
321 		seq_puts(seq, ",dax");
322 
323 	if (!test_opt(sb, RESERVATION))
324 		seq_puts(seq, ",noreservation");
325 
326 	spin_unlock(&sbi->s_lock);
327 	return 0;
328 }
329 
330 #ifdef CONFIG_QUOTA
331 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
332 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
333 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
334 			 const struct path *path);
335 static struct dquot **ext2_get_dquots(struct inode *inode)
336 {
337 	return EXT2_I(inode)->i_dquot;
338 }
339 
340 static const struct quotactl_ops ext2_quotactl_ops = {
341 	.quota_on	= ext2_quota_on,
342 	.quota_off	= ext2_quota_off,
343 	.quota_sync	= dquot_quota_sync,
344 	.get_state	= dquot_get_state,
345 	.set_info	= dquot_set_dqinfo,
346 	.get_dqblk	= dquot_get_dqblk,
347 	.set_dqblk	= dquot_set_dqblk,
348 	.get_nextdqblk	= dquot_get_next_dqblk,
349 };
350 #endif
351 
352 static const struct super_operations ext2_sops = {
353 	.alloc_inode	= ext2_alloc_inode,
354 	.destroy_inode	= ext2_destroy_inode,
355 	.write_inode	= ext2_write_inode,
356 	.evict_inode	= ext2_evict_inode,
357 	.put_super	= ext2_put_super,
358 	.sync_fs	= ext2_sync_fs,
359 	.freeze_fs	= ext2_freeze,
360 	.unfreeze_fs	= ext2_unfreeze,
361 	.statfs		= ext2_statfs,
362 	.remount_fs	= ext2_remount,
363 	.show_options	= ext2_show_options,
364 #ifdef CONFIG_QUOTA
365 	.quota_read	= ext2_quota_read,
366 	.quota_write	= ext2_quota_write,
367 	.get_dquots	= ext2_get_dquots,
368 #endif
369 };
370 
371 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
372 		u64 ino, u32 generation)
373 {
374 	struct inode *inode;
375 
376 	if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
377 		return ERR_PTR(-ESTALE);
378 	if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
379 		return ERR_PTR(-ESTALE);
380 
381 	/*
382 	 * ext2_iget isn't quite right if the inode is currently unallocated!
383 	 * However ext2_iget currently does appropriate checks to handle stale
384 	 * inodes so everything is OK.
385 	 */
386 	inode = ext2_iget(sb, ino);
387 	if (IS_ERR(inode))
388 		return ERR_CAST(inode);
389 	if (generation && inode->i_generation != generation) {
390 		/* we didn't find the right inode.. */
391 		iput(inode);
392 		return ERR_PTR(-ESTALE);
393 	}
394 	return inode;
395 }
396 
397 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
398 		int fh_len, int fh_type)
399 {
400 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
401 				    ext2_nfs_get_inode);
402 }
403 
404 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
405 		int fh_len, int fh_type)
406 {
407 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
408 				    ext2_nfs_get_inode);
409 }
410 
411 static const struct export_operations ext2_export_ops = {
412 	.fh_to_dentry = ext2_fh_to_dentry,
413 	.fh_to_parent = ext2_fh_to_parent,
414 	.get_parent = ext2_get_parent,
415 };
416 
417 static unsigned long get_sb_block(void **data)
418 {
419 	unsigned long 	sb_block;
420 	char 		*options = (char *) *data;
421 
422 	if (!options || strncmp(options, "sb=", 3) != 0)
423 		return 1;	/* Default location */
424 	options += 3;
425 	sb_block = simple_strtoul(options, &options, 0);
426 	if (*options && *options != ',') {
427 		printk("EXT2-fs: Invalid sb specification: %s\n",
428 		       (char *) *data);
429 		return 1;
430 	}
431 	if (*options == ',')
432 		options++;
433 	*data = (void *) options;
434 	return sb_block;
435 }
436 
437 enum {
438 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
439 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
440 	Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
441 	Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
442 	Opt_acl, Opt_noacl, Opt_xip, Opt_dax, Opt_ignore, Opt_err, Opt_quota,
443 	Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
444 };
445 
446 static const match_table_t tokens = {
447 	{Opt_bsd_df, "bsddf"},
448 	{Opt_minix_df, "minixdf"},
449 	{Opt_grpid, "grpid"},
450 	{Opt_grpid, "bsdgroups"},
451 	{Opt_nogrpid, "nogrpid"},
452 	{Opt_nogrpid, "sysvgroups"},
453 	{Opt_resgid, "resgid=%u"},
454 	{Opt_resuid, "resuid=%u"},
455 	{Opt_sb, "sb=%u"},
456 	{Opt_err_cont, "errors=continue"},
457 	{Opt_err_panic, "errors=panic"},
458 	{Opt_err_ro, "errors=remount-ro"},
459 	{Opt_nouid32, "nouid32"},
460 	{Opt_nocheck, "check=none"},
461 	{Opt_nocheck, "nocheck"},
462 	{Opt_debug, "debug"},
463 	{Opt_oldalloc, "oldalloc"},
464 	{Opt_orlov, "orlov"},
465 	{Opt_nobh, "nobh"},
466 	{Opt_user_xattr, "user_xattr"},
467 	{Opt_nouser_xattr, "nouser_xattr"},
468 	{Opt_acl, "acl"},
469 	{Opt_noacl, "noacl"},
470 	{Opt_xip, "xip"},
471 	{Opt_dax, "dax"},
472 	{Opt_grpquota, "grpquota"},
473 	{Opt_ignore, "noquota"},
474 	{Opt_quota, "quota"},
475 	{Opt_usrquota, "usrquota"},
476 	{Opt_reservation, "reservation"},
477 	{Opt_noreservation, "noreservation"},
478 	{Opt_err, NULL}
479 };
480 
481 static int parse_options(char *options, struct super_block *sb,
482 			 struct ext2_mount_options *opts)
483 {
484 	char *p;
485 	substring_t args[MAX_OPT_ARGS];
486 	int option;
487 	kuid_t uid;
488 	kgid_t gid;
489 
490 	if (!options)
491 		return 1;
492 
493 	while ((p = strsep (&options, ",")) != NULL) {
494 		int token;
495 		if (!*p)
496 			continue;
497 
498 		token = match_token(p, tokens, args);
499 		switch (token) {
500 		case Opt_bsd_df:
501 			clear_opt (opts->s_mount_opt, MINIX_DF);
502 			break;
503 		case Opt_minix_df:
504 			set_opt (opts->s_mount_opt, MINIX_DF);
505 			break;
506 		case Opt_grpid:
507 			set_opt (opts->s_mount_opt, GRPID);
508 			break;
509 		case Opt_nogrpid:
510 			clear_opt (opts->s_mount_opt, GRPID);
511 			break;
512 		case Opt_resuid:
513 			if (match_int(&args[0], &option))
514 				return 0;
515 			uid = make_kuid(current_user_ns(), option);
516 			if (!uid_valid(uid)) {
517 				ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
518 				return 0;
519 
520 			}
521 			opts->s_resuid = uid;
522 			break;
523 		case Opt_resgid:
524 			if (match_int(&args[0], &option))
525 				return 0;
526 			gid = make_kgid(current_user_ns(), option);
527 			if (!gid_valid(gid)) {
528 				ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
529 				return 0;
530 			}
531 			opts->s_resgid = gid;
532 			break;
533 		case Opt_sb:
534 			/* handled by get_sb_block() instead of here */
535 			/* *sb_block = match_int(&args[0]); */
536 			break;
537 		case Opt_err_panic:
538 			clear_opt (opts->s_mount_opt, ERRORS_CONT);
539 			clear_opt (opts->s_mount_opt, ERRORS_RO);
540 			set_opt (opts->s_mount_opt, ERRORS_PANIC);
541 			break;
542 		case Opt_err_ro:
543 			clear_opt (opts->s_mount_opt, ERRORS_CONT);
544 			clear_opt (opts->s_mount_opt, ERRORS_PANIC);
545 			set_opt (opts->s_mount_opt, ERRORS_RO);
546 			break;
547 		case Opt_err_cont:
548 			clear_opt (opts->s_mount_opt, ERRORS_RO);
549 			clear_opt (opts->s_mount_opt, ERRORS_PANIC);
550 			set_opt (opts->s_mount_opt, ERRORS_CONT);
551 			break;
552 		case Opt_nouid32:
553 			set_opt (opts->s_mount_opt, NO_UID32);
554 			break;
555 		case Opt_nocheck:
556 			ext2_msg(sb, KERN_WARNING,
557 				"Option nocheck/check=none is deprecated and"
558 				" will be removed in June 2020.");
559 			clear_opt (opts->s_mount_opt, CHECK);
560 			break;
561 		case Opt_debug:
562 			set_opt (opts->s_mount_opt, DEBUG);
563 			break;
564 		case Opt_oldalloc:
565 			set_opt (opts->s_mount_opt, OLDALLOC);
566 			break;
567 		case Opt_orlov:
568 			clear_opt (opts->s_mount_opt, OLDALLOC);
569 			break;
570 		case Opt_nobh:
571 			set_opt (opts->s_mount_opt, NOBH);
572 			break;
573 #ifdef CONFIG_EXT2_FS_XATTR
574 		case Opt_user_xattr:
575 			set_opt (opts->s_mount_opt, XATTR_USER);
576 			break;
577 		case Opt_nouser_xattr:
578 			clear_opt (opts->s_mount_opt, XATTR_USER);
579 			break;
580 #else
581 		case Opt_user_xattr:
582 		case Opt_nouser_xattr:
583 			ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
584 				"not supported");
585 			break;
586 #endif
587 #ifdef CONFIG_EXT2_FS_POSIX_ACL
588 		case Opt_acl:
589 			set_opt(opts->s_mount_opt, POSIX_ACL);
590 			break;
591 		case Opt_noacl:
592 			clear_opt(opts->s_mount_opt, POSIX_ACL);
593 			break;
594 #else
595 		case Opt_acl:
596 		case Opt_noacl:
597 			ext2_msg(sb, KERN_INFO,
598 				"(no)acl options not supported");
599 			break;
600 #endif
601 		case Opt_xip:
602 			ext2_msg(sb, KERN_INFO, "use dax instead of xip");
603 			set_opt(opts->s_mount_opt, XIP);
604 			/* Fall through */
605 		case Opt_dax:
606 #ifdef CONFIG_FS_DAX
607 			ext2_msg(sb, KERN_WARNING,
608 		"DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
609 			set_opt(opts->s_mount_opt, DAX);
610 #else
611 			ext2_msg(sb, KERN_INFO, "dax option not supported");
612 #endif
613 			break;
614 
615 #if defined(CONFIG_QUOTA)
616 		case Opt_quota:
617 		case Opt_usrquota:
618 			set_opt(opts->s_mount_opt, USRQUOTA);
619 			break;
620 
621 		case Opt_grpquota:
622 			set_opt(opts->s_mount_opt, GRPQUOTA);
623 			break;
624 #else
625 		case Opt_quota:
626 		case Opt_usrquota:
627 		case Opt_grpquota:
628 			ext2_msg(sb, KERN_INFO,
629 				"quota operations not supported");
630 			break;
631 #endif
632 
633 		case Opt_reservation:
634 			set_opt(opts->s_mount_opt, RESERVATION);
635 			ext2_msg(sb, KERN_INFO, "reservations ON");
636 			break;
637 		case Opt_noreservation:
638 			clear_opt(opts->s_mount_opt, RESERVATION);
639 			ext2_msg(sb, KERN_INFO, "reservations OFF");
640 			break;
641 		case Opt_ignore:
642 			break;
643 		default:
644 			return 0;
645 		}
646 	}
647 	return 1;
648 }
649 
650 static int ext2_setup_super (struct super_block * sb,
651 			      struct ext2_super_block * es,
652 			      int read_only)
653 {
654 	int res = 0;
655 	struct ext2_sb_info *sbi = EXT2_SB(sb);
656 
657 	if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
658 		ext2_msg(sb, KERN_ERR,
659 			"error: revision level too high, "
660 			"forcing read-only mode");
661 		res = SB_RDONLY;
662 	}
663 	if (read_only)
664 		return res;
665 	if (!(sbi->s_mount_state & EXT2_VALID_FS))
666 		ext2_msg(sb, KERN_WARNING,
667 			"warning: mounting unchecked fs, "
668 			"running e2fsck is recommended");
669 	else if ((sbi->s_mount_state & EXT2_ERROR_FS))
670 		ext2_msg(sb, KERN_WARNING,
671 			"warning: mounting fs with errors, "
672 			"running e2fsck is recommended");
673 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
674 		 le16_to_cpu(es->s_mnt_count) >=
675 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
676 		ext2_msg(sb, KERN_WARNING,
677 			"warning: maximal mount count reached, "
678 			"running e2fsck is recommended");
679 	else if (le32_to_cpu(es->s_checkinterval) &&
680 		(le32_to_cpu(es->s_lastcheck) +
681 			le32_to_cpu(es->s_checkinterval) <=
682 			ktime_get_real_seconds()))
683 		ext2_msg(sb, KERN_WARNING,
684 			"warning: checktime reached, "
685 			"running e2fsck is recommended");
686 	if (!le16_to_cpu(es->s_max_mnt_count))
687 		es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
688 	le16_add_cpu(&es->s_mnt_count, 1);
689 	if (test_opt (sb, DEBUG))
690 		ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
691 			"bpg=%lu, ipg=%lu, mo=%04lx]",
692 			EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
693 			sbi->s_frag_size,
694 			sbi->s_groups_count,
695 			EXT2_BLOCKS_PER_GROUP(sb),
696 			EXT2_INODES_PER_GROUP(sb),
697 			sbi->s_mount_opt);
698 	return res;
699 }
700 
701 static int ext2_check_descriptors(struct super_block *sb)
702 {
703 	int i;
704 	struct ext2_sb_info *sbi = EXT2_SB(sb);
705 
706 	ext2_debug ("Checking group descriptors");
707 
708 	for (i = 0; i < sbi->s_groups_count; i++) {
709 		struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
710 		ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
711 		ext2_fsblk_t last_block;
712 
713 		if (i == sbi->s_groups_count - 1)
714 			last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
715 		else
716 			last_block = first_block +
717 				(EXT2_BLOCKS_PER_GROUP(sb) - 1);
718 
719 		if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
720 		    le32_to_cpu(gdp->bg_block_bitmap) > last_block)
721 		{
722 			ext2_error (sb, "ext2_check_descriptors",
723 				    "Block bitmap for group %d"
724 				    " not in group (block %lu)!",
725 				    i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
726 			return 0;
727 		}
728 		if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
729 		    le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
730 		{
731 			ext2_error (sb, "ext2_check_descriptors",
732 				    "Inode bitmap for group %d"
733 				    " not in group (block %lu)!",
734 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
735 			return 0;
736 		}
737 		if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
738 		    le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
739 		    last_block)
740 		{
741 			ext2_error (sb, "ext2_check_descriptors",
742 				    "Inode table for group %d"
743 				    " not in group (block %lu)!",
744 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
745 			return 0;
746 		}
747 	}
748 	return 1;
749 }
750 
751 /*
752  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
753  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
754  * We need to be 1 filesystem block less than the 2^32 sector limit.
755  */
756 static loff_t ext2_max_size(int bits)
757 {
758 	loff_t res = EXT2_NDIR_BLOCKS;
759 	int meta_blocks;
760 	unsigned int upper_limit;
761 	unsigned int ppb = 1 << (bits-2);
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 	/* Compute how many blocks we can address by block tree */
776 	res += 1LL << (bits-2);
777 	res += 1LL << (2*(bits-2));
778 	res += 1LL << (3*(bits-2));
779 	/* Does block tree limit file size? */
780 	if (res < upper_limit)
781 		goto check_lfs;
782 
783 	res = upper_limit;
784 	/* How many metadata blocks are needed for addressing upper_limit? */
785 	upper_limit -= EXT2_NDIR_BLOCKS;
786 	/* indirect blocks */
787 	meta_blocks = 1;
788 	upper_limit -= ppb;
789 	/* double indirect blocks */
790 	if (upper_limit < ppb * ppb) {
791 		meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb);
792 		res -= meta_blocks;
793 		goto check_lfs;
794 	}
795 	meta_blocks += 1 + ppb;
796 	upper_limit -= ppb * ppb;
797 	/* tripple indirect blocks for the rest */
798 	meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb) +
799 		DIV_ROUND_UP(upper_limit, ppb*ppb);
800 	res -= meta_blocks;
801 check_lfs:
802 	res <<= bits;
803 	if (res > MAX_LFS_FILESIZE)
804 		res = MAX_LFS_FILESIZE;
805 
806 	return res;
807 }
808 
809 static unsigned long descriptor_loc(struct super_block *sb,
810 				    unsigned long logic_sb_block,
811 				    int nr)
812 {
813 	struct ext2_sb_info *sbi = EXT2_SB(sb);
814 	unsigned long bg, first_meta_bg;
815 	int has_super = 0;
816 
817 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
818 
819 	if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
820 	    nr < first_meta_bg)
821 		return (logic_sb_block + nr + 1);
822 	bg = sbi->s_desc_per_block * nr;
823 	if (ext2_bg_has_super(sb, bg))
824 		has_super = 1;
825 
826 	return ext2_group_first_block_no(sb, bg) + has_super;
827 }
828 
829 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
830 {
831 	struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
832 	struct buffer_head * bh;
833 	struct ext2_sb_info * sbi;
834 	struct ext2_super_block * es;
835 	struct inode *root;
836 	unsigned long block;
837 	unsigned long sb_block = get_sb_block(&data);
838 	unsigned long logic_sb_block;
839 	unsigned long offset = 0;
840 	unsigned long def_mount_opts;
841 	long ret = -ENOMEM;
842 	int blocksize = BLOCK_SIZE;
843 	int db_count;
844 	int i, j;
845 	__le32 features;
846 	int err;
847 	struct ext2_mount_options opts;
848 
849 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
850 	if (!sbi)
851 		goto failed;
852 
853 	sbi->s_blockgroup_lock =
854 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
855 	if (!sbi->s_blockgroup_lock) {
856 		kfree(sbi);
857 		goto failed;
858 	}
859 	sb->s_fs_info = sbi;
860 	sbi->s_sb_block = sb_block;
861 	sbi->s_daxdev = dax_dev;
862 
863 	spin_lock_init(&sbi->s_lock);
864 	ret = -EINVAL;
865 
866 	/*
867 	 * See what the current blocksize for the device is, and
868 	 * use that as the blocksize.  Otherwise (or if the blocksize
869 	 * is smaller than the default) use the default.
870 	 * This is important for devices that have a hardware
871 	 * sectorsize that is larger than the default.
872 	 */
873 	blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
874 	if (!blocksize) {
875 		ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
876 		goto failed_sbi;
877 	}
878 
879 	/*
880 	 * If the superblock doesn't start on a hardware sector boundary,
881 	 * calculate the offset.
882 	 */
883 	if (blocksize != BLOCK_SIZE) {
884 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
885 		offset = (sb_block*BLOCK_SIZE) % blocksize;
886 	} else {
887 		logic_sb_block = sb_block;
888 	}
889 
890 	if (!(bh = sb_bread(sb, logic_sb_block))) {
891 		ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
892 		goto failed_sbi;
893 	}
894 	/*
895 	 * Note: s_es must be initialized as soon as possible because
896 	 *       some ext2 macro-instructions depend on its value
897 	 */
898 	es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
899 	sbi->s_es = es;
900 	sb->s_magic = le16_to_cpu(es->s_magic);
901 
902 	if (sb->s_magic != EXT2_SUPER_MAGIC)
903 		goto cantfind_ext2;
904 
905 	opts.s_mount_opt = 0;
906 	/* Set defaults before we parse the mount options */
907 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
908 	if (def_mount_opts & EXT2_DEFM_DEBUG)
909 		set_opt(opts.s_mount_opt, DEBUG);
910 	if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
911 		set_opt(opts.s_mount_opt, GRPID);
912 	if (def_mount_opts & EXT2_DEFM_UID16)
913 		set_opt(opts.s_mount_opt, NO_UID32);
914 #ifdef CONFIG_EXT2_FS_XATTR
915 	if (def_mount_opts & EXT2_DEFM_XATTR_USER)
916 		set_opt(opts.s_mount_opt, XATTR_USER);
917 #endif
918 #ifdef CONFIG_EXT2_FS_POSIX_ACL
919 	if (def_mount_opts & EXT2_DEFM_ACL)
920 		set_opt(opts.s_mount_opt, POSIX_ACL);
921 #endif
922 
923 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
924 		set_opt(opts.s_mount_opt, ERRORS_PANIC);
925 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
926 		set_opt(opts.s_mount_opt, ERRORS_CONT);
927 	else
928 		set_opt(opts.s_mount_opt, ERRORS_RO);
929 
930 	opts.s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
931 	opts.s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
932 
933 	set_opt(opts.s_mount_opt, RESERVATION);
934 
935 	if (!parse_options((char *) data, sb, &opts))
936 		goto failed_mount;
937 
938 	sbi->s_mount_opt = opts.s_mount_opt;
939 	sbi->s_resuid = opts.s_resuid;
940 	sbi->s_resgid = opts.s_resgid;
941 
942 	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
943 		((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
944 		 SB_POSIXACL : 0);
945 	sb->s_iflags |= SB_I_CGROUPWB;
946 
947 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
948 	    (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
949 	     EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
950 	     EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
951 		ext2_msg(sb, KERN_WARNING,
952 			"warning: feature flags set on rev 0 fs, "
953 			"running e2fsck is recommended");
954 	/*
955 	 * Check feature flags regardless of the revision level, since we
956 	 * previously didn't change the revision level when setting the flags,
957 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
958 	 */
959 	features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
960 	if (features) {
961 		ext2_msg(sb, KERN_ERR,	"error: couldn't mount because of "
962 		       "unsupported optional features (%x)",
963 			le32_to_cpu(features));
964 		goto failed_mount;
965 	}
966 	if (!sb_rdonly(sb) && (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
967 		ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
968 		       "unsupported optional features (%x)",
969 		       le32_to_cpu(features));
970 		goto failed_mount;
971 	}
972 
973 	blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
974 
975 	if (sbi->s_mount_opt & EXT2_MOUNT_DAX) {
976 		if (!bdev_dax_supported(sb->s_bdev, blocksize)) {
977 			ext2_msg(sb, KERN_ERR,
978 				"DAX unsupported by block device. Turning off DAX.");
979 			sbi->s_mount_opt &= ~EXT2_MOUNT_DAX;
980 		}
981 	}
982 
983 	/* If the blocksize doesn't match, re-read the thing.. */
984 	if (sb->s_blocksize != blocksize) {
985 		brelse(bh);
986 
987 		if (!sb_set_blocksize(sb, blocksize)) {
988 			ext2_msg(sb, KERN_ERR,
989 				"error: bad blocksize %d", blocksize);
990 			goto failed_sbi;
991 		}
992 
993 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
994 		offset = (sb_block*BLOCK_SIZE) % blocksize;
995 		bh = sb_bread(sb, logic_sb_block);
996 		if(!bh) {
997 			ext2_msg(sb, KERN_ERR, "error: couldn't read"
998 				"superblock on 2nd try");
999 			goto failed_sbi;
1000 		}
1001 		es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
1002 		sbi->s_es = es;
1003 		if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
1004 			ext2_msg(sb, KERN_ERR, "error: magic mismatch");
1005 			goto failed_mount;
1006 		}
1007 	}
1008 
1009 	sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
1010 	sb->s_max_links = EXT2_LINK_MAX;
1011 
1012 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
1013 		sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
1014 		sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
1015 	} else {
1016 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1017 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1018 		if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
1019 		    !is_power_of_2(sbi->s_inode_size) ||
1020 		    (sbi->s_inode_size > blocksize)) {
1021 			ext2_msg(sb, KERN_ERR,
1022 				"error: unsupported inode size: %d",
1023 				sbi->s_inode_size);
1024 			goto failed_mount;
1025 		}
1026 	}
1027 
1028 	sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
1029 				   le32_to_cpu(es->s_log_frag_size);
1030 	if (sbi->s_frag_size == 0)
1031 		goto cantfind_ext2;
1032 	sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
1033 
1034 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1035 	sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1036 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1037 
1038 	sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
1039 	if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
1040 		goto cantfind_ext2;
1041 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
1042 					sbi->s_inodes_per_block;
1043 	sbi->s_desc_per_block = sb->s_blocksize /
1044 					sizeof (struct ext2_group_desc);
1045 	sbi->s_sbh = bh;
1046 	sbi->s_mount_state = le16_to_cpu(es->s_state);
1047 	sbi->s_addr_per_block_bits =
1048 		ilog2 (EXT2_ADDR_PER_BLOCK(sb));
1049 	sbi->s_desc_per_block_bits =
1050 		ilog2 (EXT2_DESC_PER_BLOCK(sb));
1051 
1052 	if (sb->s_magic != EXT2_SUPER_MAGIC)
1053 		goto cantfind_ext2;
1054 
1055 	if (sb->s_blocksize != bh->b_size) {
1056 		if (!silent)
1057 			ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
1058 		goto failed_mount;
1059 	}
1060 
1061 	if (sb->s_blocksize != sbi->s_frag_size) {
1062 		ext2_msg(sb, KERN_ERR,
1063 			"error: fragsize %lu != blocksize %lu"
1064 			"(not supported yet)",
1065 			sbi->s_frag_size, sb->s_blocksize);
1066 		goto failed_mount;
1067 	}
1068 
1069 	if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
1070 		ext2_msg(sb, KERN_ERR,
1071 			"error: #blocks per group too big: %lu",
1072 			sbi->s_blocks_per_group);
1073 		goto failed_mount;
1074 	}
1075 	if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
1076 		ext2_msg(sb, KERN_ERR,
1077 			"error: #fragments per group too big: %lu",
1078 			sbi->s_frags_per_group);
1079 		goto failed_mount;
1080 	}
1081 	if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1082 		ext2_msg(sb, KERN_ERR,
1083 			"error: #inodes per group too big: %lu",
1084 			sbi->s_inodes_per_group);
1085 		goto failed_mount;
1086 	}
1087 
1088 	if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1089 		goto cantfind_ext2;
1090  	sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1091  				le32_to_cpu(es->s_first_data_block) - 1)
1092  					/ EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1093 	db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1094 		   EXT2_DESC_PER_BLOCK(sb);
1095 	sbi->s_group_desc = kmalloc_array (db_count,
1096 					   sizeof(struct buffer_head *),
1097 					   GFP_KERNEL);
1098 	if (sbi->s_group_desc == NULL) {
1099 		ret = -ENOMEM;
1100 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1101 		goto failed_mount;
1102 	}
1103 	bgl_lock_init(sbi->s_blockgroup_lock);
1104 	sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1105 	if (!sbi->s_debts) {
1106 		ret = -ENOMEM;
1107 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1108 		goto failed_mount_group_desc;
1109 	}
1110 	for (i = 0; i < db_count; i++) {
1111 		block = descriptor_loc(sb, logic_sb_block, i);
1112 		sbi->s_group_desc[i] = sb_bread(sb, block);
1113 		if (!sbi->s_group_desc[i]) {
1114 			for (j = 0; j < i; j++)
1115 				brelse (sbi->s_group_desc[j]);
1116 			ext2_msg(sb, KERN_ERR,
1117 				"error: unable to read group descriptors");
1118 			goto failed_mount_group_desc;
1119 		}
1120 	}
1121 	if (!ext2_check_descriptors (sb)) {
1122 		ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1123 		goto failed_mount2;
1124 	}
1125 	sbi->s_gdb_count = db_count;
1126 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1127 	spin_lock_init(&sbi->s_next_gen_lock);
1128 
1129 	/* per fileystem reservation list head & lock */
1130 	spin_lock_init(&sbi->s_rsv_window_lock);
1131 	sbi->s_rsv_window_root = RB_ROOT;
1132 	/*
1133 	 * Add a single, static dummy reservation to the start of the
1134 	 * reservation window list --- it gives us a placeholder for
1135 	 * append-at-start-of-list which makes the allocation logic
1136 	 * _much_ simpler.
1137 	 */
1138 	sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1139 	sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1140 	sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1141 	sbi->s_rsv_window_head.rsv_goal_size = 0;
1142 	ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1143 
1144 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
1145 				ext2_count_free_blocks(sb), GFP_KERNEL);
1146 	if (!err) {
1147 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
1148 				ext2_count_free_inodes(sb), GFP_KERNEL);
1149 	}
1150 	if (!err) {
1151 		err = percpu_counter_init(&sbi->s_dirs_counter,
1152 				ext2_count_dirs(sb), GFP_KERNEL);
1153 	}
1154 	if (err) {
1155 		ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1156 		goto failed_mount3;
1157 	}
1158 
1159 #ifdef CONFIG_EXT2_FS_XATTR
1160 	sbi->s_ea_block_cache = ext2_xattr_create_cache();
1161 	if (!sbi->s_ea_block_cache) {
1162 		ret = -ENOMEM;
1163 		ext2_msg(sb, KERN_ERR, "Failed to create ea_block_cache");
1164 		goto failed_mount3;
1165 	}
1166 #endif
1167 	/*
1168 	 * set up enough so that it can read an inode
1169 	 */
1170 	sb->s_op = &ext2_sops;
1171 	sb->s_export_op = &ext2_export_ops;
1172 	sb->s_xattr = ext2_xattr_handlers;
1173 
1174 #ifdef CONFIG_QUOTA
1175 	sb->dq_op = &dquot_operations;
1176 	sb->s_qcop = &ext2_quotactl_ops;
1177 	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1178 #endif
1179 
1180 	root = ext2_iget(sb, EXT2_ROOT_INO);
1181 	if (IS_ERR(root)) {
1182 		ret = PTR_ERR(root);
1183 		goto failed_mount3;
1184 	}
1185 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1186 		iput(root);
1187 		ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1188 		goto failed_mount3;
1189 	}
1190 
1191 	sb->s_root = d_make_root(root);
1192 	if (!sb->s_root) {
1193 		ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1194 		ret = -ENOMEM;
1195 		goto failed_mount3;
1196 	}
1197 	if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1198 		ext2_msg(sb, KERN_WARNING,
1199 			"warning: mounting ext3 filesystem as ext2");
1200 	if (ext2_setup_super (sb, es, sb_rdonly(sb)))
1201 		sb->s_flags |= SB_RDONLY;
1202 	ext2_write_super(sb);
1203 	return 0;
1204 
1205 cantfind_ext2:
1206 	if (!silent)
1207 		ext2_msg(sb, KERN_ERR,
1208 			"error: can't find an ext2 filesystem on dev %s.",
1209 			sb->s_id);
1210 	goto failed_mount;
1211 failed_mount3:
1212 	ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
1213 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
1214 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
1215 	percpu_counter_destroy(&sbi->s_dirs_counter);
1216 failed_mount2:
1217 	for (i = 0; i < db_count; i++)
1218 		brelse(sbi->s_group_desc[i]);
1219 failed_mount_group_desc:
1220 	kfree(sbi->s_group_desc);
1221 	kfree(sbi->s_debts);
1222 failed_mount:
1223 	brelse(bh);
1224 failed_sbi:
1225 	sb->s_fs_info = NULL;
1226 	kfree(sbi->s_blockgroup_lock);
1227 	kfree(sbi);
1228 failed:
1229 	fs_put_dax(dax_dev);
1230 	return ret;
1231 }
1232 
1233 static void ext2_clear_super_error(struct super_block *sb)
1234 {
1235 	struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1236 
1237 	if (buffer_write_io_error(sbh)) {
1238 		/*
1239 		 * Oh, dear.  A previous attempt to write the
1240 		 * superblock failed.  This could happen because the
1241 		 * USB device was yanked out.  Or it could happen to
1242 		 * be a transient write error and maybe the block will
1243 		 * be remapped.  Nothing we can do but to retry the
1244 		 * write and hope for the best.
1245 		 */
1246 		ext2_msg(sb, KERN_ERR,
1247 		       "previous I/O error to superblock detected");
1248 		clear_buffer_write_io_error(sbh);
1249 		set_buffer_uptodate(sbh);
1250 	}
1251 }
1252 
1253 void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1254 		     int wait)
1255 {
1256 	ext2_clear_super_error(sb);
1257 	spin_lock(&EXT2_SB(sb)->s_lock);
1258 	es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1259 	es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1260 	es->s_wtime = cpu_to_le32(ktime_get_real_seconds());
1261 	/* unlock before we do IO */
1262 	spin_unlock(&EXT2_SB(sb)->s_lock);
1263 	mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1264 	if (wait)
1265 		sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1266 }
1267 
1268 /*
1269  * In the second extended file system, it is not necessary to
1270  * write the super block since we use a mapping of the
1271  * disk super block in a buffer.
1272  *
1273  * However, this function is still used to set the fs valid
1274  * flags to 0.  We need to set this flag to 0 since the fs
1275  * may have been checked while mounted and e2fsck may have
1276  * set s_state to EXT2_VALID_FS after some corrections.
1277  */
1278 static int ext2_sync_fs(struct super_block *sb, int wait)
1279 {
1280 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1281 	struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1282 
1283 	/*
1284 	 * Write quota structures to quota file, sync_blockdev() will write
1285 	 * them to disk later
1286 	 */
1287 	dquot_writeback_dquots(sb, -1);
1288 
1289 	spin_lock(&sbi->s_lock);
1290 	if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1291 		ext2_debug("setting valid to 0\n");
1292 		es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1293 	}
1294 	spin_unlock(&sbi->s_lock);
1295 	ext2_sync_super(sb, es, wait);
1296 	return 0;
1297 }
1298 
1299 static int ext2_freeze(struct super_block *sb)
1300 {
1301 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1302 
1303 	/*
1304 	 * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1305 	 * because we have unattached inodes and thus filesystem is not fully
1306 	 * consistent.
1307 	 */
1308 	if (atomic_long_read(&sb->s_remove_count)) {
1309 		ext2_sync_fs(sb, 1);
1310 		return 0;
1311 	}
1312 	/* Set EXT2_FS_VALID flag */
1313 	spin_lock(&sbi->s_lock);
1314 	sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1315 	spin_unlock(&sbi->s_lock);
1316 	ext2_sync_super(sb, sbi->s_es, 1);
1317 
1318 	return 0;
1319 }
1320 
1321 static int ext2_unfreeze(struct super_block *sb)
1322 {
1323 	/* Just write sb to clear EXT2_VALID_FS flag */
1324 	ext2_write_super(sb);
1325 
1326 	return 0;
1327 }
1328 
1329 static void ext2_write_super(struct super_block *sb)
1330 {
1331 	if (!sb_rdonly(sb))
1332 		ext2_sync_fs(sb, 1);
1333 }
1334 
1335 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1336 {
1337 	struct ext2_sb_info * sbi = EXT2_SB(sb);
1338 	struct ext2_super_block * es;
1339 	struct ext2_mount_options new_opts;
1340 	int err;
1341 
1342 	sync_filesystem(sb);
1343 
1344 	spin_lock(&sbi->s_lock);
1345 	new_opts.s_mount_opt = sbi->s_mount_opt;
1346 	new_opts.s_resuid = sbi->s_resuid;
1347 	new_opts.s_resgid = sbi->s_resgid;
1348 	spin_unlock(&sbi->s_lock);
1349 
1350 	if (!parse_options(data, sb, &new_opts))
1351 		return -EINVAL;
1352 
1353 	spin_lock(&sbi->s_lock);
1354 	es = sbi->s_es;
1355 	if ((sbi->s_mount_opt ^ new_opts.s_mount_opt) & EXT2_MOUNT_DAX) {
1356 		ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1357 			 "dax flag with busy inodes while remounting");
1358 		new_opts.s_mount_opt ^= EXT2_MOUNT_DAX;
1359 	}
1360 	if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1361 		goto out_set;
1362 	if (*flags & SB_RDONLY) {
1363 		if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1364 		    !(sbi->s_mount_state & EXT2_VALID_FS))
1365 			goto out_set;
1366 
1367 		/*
1368 		 * OK, we are remounting a valid rw partition rdonly, so set
1369 		 * the rdonly flag and then mark the partition as valid again.
1370 		 */
1371 		es->s_state = cpu_to_le16(sbi->s_mount_state);
1372 		es->s_mtime = cpu_to_le32(ktime_get_real_seconds());
1373 		spin_unlock(&sbi->s_lock);
1374 
1375 		err = dquot_suspend(sb, -1);
1376 		if (err < 0)
1377 			return err;
1378 
1379 		ext2_sync_super(sb, es, 1);
1380 	} else {
1381 		__le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1382 					       ~EXT2_FEATURE_RO_COMPAT_SUPP);
1383 		if (ret) {
1384 			spin_unlock(&sbi->s_lock);
1385 			ext2_msg(sb, KERN_WARNING,
1386 				"warning: couldn't remount RDWR because of "
1387 				"unsupported optional features (%x).",
1388 				le32_to_cpu(ret));
1389 			return -EROFS;
1390 		}
1391 		/*
1392 		 * Mounting a RDONLY partition read-write, so reread and
1393 		 * store the current valid flag.  (It may have been changed
1394 		 * by e2fsck since we originally mounted the partition.)
1395 		 */
1396 		sbi->s_mount_state = le16_to_cpu(es->s_state);
1397 		if (!ext2_setup_super (sb, es, 0))
1398 			sb->s_flags &= ~SB_RDONLY;
1399 		spin_unlock(&sbi->s_lock);
1400 
1401 		ext2_write_super(sb);
1402 
1403 		dquot_resume(sb, -1);
1404 	}
1405 
1406 	spin_lock(&sbi->s_lock);
1407 out_set:
1408 	sbi->s_mount_opt = new_opts.s_mount_opt;
1409 	sbi->s_resuid = new_opts.s_resuid;
1410 	sbi->s_resgid = new_opts.s_resgid;
1411 	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1412 		((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? SB_POSIXACL : 0);
1413 	spin_unlock(&sbi->s_lock);
1414 
1415 	return 0;
1416 }
1417 
1418 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1419 {
1420 	struct super_block *sb = dentry->d_sb;
1421 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1422 	struct ext2_super_block *es = sbi->s_es;
1423 	u64 fsid;
1424 
1425 	spin_lock(&sbi->s_lock);
1426 
1427 	if (test_opt (sb, MINIX_DF))
1428 		sbi->s_overhead_last = 0;
1429 	else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1430 		unsigned long i, overhead = 0;
1431 		smp_rmb();
1432 
1433 		/*
1434 		 * Compute the overhead (FS structures). This is constant
1435 		 * for a given filesystem unless the number of block groups
1436 		 * changes so we cache the previous value until it does.
1437 		 */
1438 
1439 		/*
1440 		 * All of the blocks before first_data_block are
1441 		 * overhead
1442 		 */
1443 		overhead = le32_to_cpu(es->s_first_data_block);
1444 
1445 		/*
1446 		 * Add the overhead attributed to the superblock and
1447 		 * block group descriptors.  If the sparse superblocks
1448 		 * feature is turned on, then not all groups have this.
1449 		 */
1450 		for (i = 0; i < sbi->s_groups_count; i++)
1451 			overhead += ext2_bg_has_super(sb, i) +
1452 				ext2_bg_num_gdb(sb, i);
1453 
1454 		/*
1455 		 * Every block group has an inode bitmap, a block
1456 		 * bitmap, and an inode table.
1457 		 */
1458 		overhead += (sbi->s_groups_count *
1459 			     (2 + sbi->s_itb_per_group));
1460 		sbi->s_overhead_last = overhead;
1461 		smp_wmb();
1462 		sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1463 	}
1464 
1465 	buf->f_type = EXT2_SUPER_MAGIC;
1466 	buf->f_bsize = sb->s_blocksize;
1467 	buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1468 	buf->f_bfree = ext2_count_free_blocks(sb);
1469 	es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1470 	buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1471 	if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1472 		buf->f_bavail = 0;
1473 	buf->f_files = le32_to_cpu(es->s_inodes_count);
1474 	buf->f_ffree = ext2_count_free_inodes(sb);
1475 	es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1476 	buf->f_namelen = EXT2_NAME_LEN;
1477 	fsid = le64_to_cpup((void *)es->s_uuid) ^
1478 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1479 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1480 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1481 	spin_unlock(&sbi->s_lock);
1482 	return 0;
1483 }
1484 
1485 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1486 	int flags, const char *dev_name, void *data)
1487 {
1488 	return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1489 }
1490 
1491 #ifdef CONFIG_QUOTA
1492 
1493 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1494  * acquiring the locks... As quota files are never truncated and quota code
1495  * itself serializes the operations (and no one else should touch the files)
1496  * we don't have to be afraid of races */
1497 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1498 			       size_t len, loff_t off)
1499 {
1500 	struct inode *inode = sb_dqopt(sb)->files[type];
1501 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1502 	int err = 0;
1503 	int offset = off & (sb->s_blocksize - 1);
1504 	int tocopy;
1505 	size_t toread;
1506 	struct buffer_head tmp_bh;
1507 	struct buffer_head *bh;
1508 	loff_t i_size = i_size_read(inode);
1509 
1510 	if (off > i_size)
1511 		return 0;
1512 	if (off+len > i_size)
1513 		len = i_size-off;
1514 	toread = len;
1515 	while (toread > 0) {
1516 		tocopy = sb->s_blocksize - offset < toread ?
1517 				sb->s_blocksize - offset : toread;
1518 
1519 		tmp_bh.b_state = 0;
1520 		tmp_bh.b_size = sb->s_blocksize;
1521 		err = ext2_get_block(inode, blk, &tmp_bh, 0);
1522 		if (err < 0)
1523 			return err;
1524 		if (!buffer_mapped(&tmp_bh))	/* A hole? */
1525 			memset(data, 0, tocopy);
1526 		else {
1527 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1528 			if (!bh)
1529 				return -EIO;
1530 			memcpy(data, bh->b_data+offset, tocopy);
1531 			brelse(bh);
1532 		}
1533 		offset = 0;
1534 		toread -= tocopy;
1535 		data += tocopy;
1536 		blk++;
1537 	}
1538 	return len;
1539 }
1540 
1541 /* Write to quotafile */
1542 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1543 				const char *data, size_t len, loff_t off)
1544 {
1545 	struct inode *inode = sb_dqopt(sb)->files[type];
1546 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1547 	int err = 0;
1548 	int offset = off & (sb->s_blocksize - 1);
1549 	int tocopy;
1550 	size_t towrite = len;
1551 	struct buffer_head tmp_bh;
1552 	struct buffer_head *bh;
1553 
1554 	while (towrite > 0) {
1555 		tocopy = sb->s_blocksize - offset < towrite ?
1556 				sb->s_blocksize - offset : towrite;
1557 
1558 		tmp_bh.b_state = 0;
1559 		tmp_bh.b_size = sb->s_blocksize;
1560 		err = ext2_get_block(inode, blk, &tmp_bh, 1);
1561 		if (err < 0)
1562 			goto out;
1563 		if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1564 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1565 		else
1566 			bh = sb_getblk(sb, tmp_bh.b_blocknr);
1567 		if (unlikely(!bh)) {
1568 			err = -EIO;
1569 			goto out;
1570 		}
1571 		lock_buffer(bh);
1572 		memcpy(bh->b_data+offset, data, tocopy);
1573 		flush_dcache_page(bh->b_page);
1574 		set_buffer_uptodate(bh);
1575 		mark_buffer_dirty(bh);
1576 		unlock_buffer(bh);
1577 		brelse(bh);
1578 		offset = 0;
1579 		towrite -= tocopy;
1580 		data += tocopy;
1581 		blk++;
1582 	}
1583 out:
1584 	if (len == towrite)
1585 		return err;
1586 	if (inode->i_size < off+len-towrite)
1587 		i_size_write(inode, off+len-towrite);
1588 	inode_inc_iversion(inode);
1589 	inode->i_mtime = inode->i_ctime = current_time(inode);
1590 	mark_inode_dirty(inode);
1591 	return len - towrite;
1592 }
1593 
1594 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
1595 			 const struct path *path)
1596 {
1597 	int err;
1598 	struct inode *inode;
1599 
1600 	err = dquot_quota_on(sb, type, format_id, path);
1601 	if (err)
1602 		return err;
1603 
1604 	inode = d_inode(path->dentry);
1605 	inode_lock(inode);
1606 	EXT2_I(inode)->i_flags |= EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL;
1607 	inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1608 			S_NOATIME | S_IMMUTABLE);
1609 	inode_unlock(inode);
1610 	mark_inode_dirty(inode);
1611 
1612 	return 0;
1613 }
1614 
1615 static int ext2_quota_off(struct super_block *sb, int type)
1616 {
1617 	struct inode *inode = sb_dqopt(sb)->files[type];
1618 	int err;
1619 
1620 	if (!inode || !igrab(inode))
1621 		goto out;
1622 
1623 	err = dquot_quota_off(sb, type);
1624 	if (err)
1625 		goto out_put;
1626 
1627 	inode_lock(inode);
1628 	EXT2_I(inode)->i_flags &= ~(EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL);
1629 	inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1630 	inode_unlock(inode);
1631 	mark_inode_dirty(inode);
1632 out_put:
1633 	iput(inode);
1634 	return err;
1635 out:
1636 	return dquot_quota_off(sb, type);
1637 }
1638 
1639 #endif
1640 
1641 static struct file_system_type ext2_fs_type = {
1642 	.owner		= THIS_MODULE,
1643 	.name		= "ext2",
1644 	.mount		= ext2_mount,
1645 	.kill_sb	= kill_block_super,
1646 	.fs_flags	= FS_REQUIRES_DEV,
1647 };
1648 MODULE_ALIAS_FS("ext2");
1649 
1650 static int __init init_ext2_fs(void)
1651 {
1652 	int err;
1653 
1654 	err = init_inodecache();
1655 	if (err)
1656 		return err;
1657         err = register_filesystem(&ext2_fs_type);
1658 	if (err)
1659 		goto out;
1660 	return 0;
1661 out:
1662 	destroy_inodecache();
1663 	return err;
1664 }
1665 
1666 static void __exit exit_ext2_fs(void)
1667 {
1668 	unregister_filesystem(&ext2_fs_type);
1669 	destroy_inodecache();
1670 }
1671 
1672 MODULE_AUTHOR("Remy Card and others");
1673 MODULE_DESCRIPTION("Second Extended Filesystem");
1674 MODULE_LICENSE("GPL");
1675 module_init(init_ext2_fs)
1676 module_exit(exit_ext2_fs)
1677