xref: /linux/fs/ext2/super.c (revision b7019ac550eb3916f34d79db583e9b7ea2524afa)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  *  linux/fs/ext2/super.c
4  *
5  * Copyright (C) 1992, 1993, 1994, 1995
6  * Remy Card (card@masi.ibp.fr)
7  * Laboratoire MASI - Institut Blaise Pascal
8  * Universite Pierre et Marie Curie (Paris VI)
9  *
10  *  from
11  *
12  *  linux/fs/minix/inode.c
13  *
14  *  Copyright (C) 1991, 1992  Linus Torvalds
15  *
16  *  Big-endian to little-endian byte-swapping/bitmaps by
17  *        David S. Miller (davem@caip.rutgers.edu), 1995
18  */
19 
20 #include <linux/module.h>
21 #include <linux/string.h>
22 #include <linux/fs.h>
23 #include <linux/slab.h>
24 #include <linux/init.h>
25 #include <linux/blkdev.h>
26 #include <linux/parser.h>
27 #include <linux/random.h>
28 #include <linux/buffer_head.h>
29 #include <linux/exportfs.h>
30 #include <linux/vfs.h>
31 #include <linux/seq_file.h>
32 #include <linux/mount.h>
33 #include <linux/log2.h>
34 #include <linux/quotaops.h>
35 #include <linux/uaccess.h>
36 #include <linux/dax.h>
37 #include <linux/iversion.h>
38 #include "ext2.h"
39 #include "xattr.h"
40 #include "acl.h"
41 
42 static void ext2_write_super(struct super_block *sb);
43 static int ext2_remount (struct super_block * sb, int * flags, char * data);
44 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
45 static int ext2_sync_fs(struct super_block *sb, int wait);
46 static int ext2_freeze(struct super_block *sb);
47 static int ext2_unfreeze(struct super_block *sb);
48 
49 void ext2_error(struct super_block *sb, const char *function,
50 		const char *fmt, ...)
51 {
52 	struct va_format vaf;
53 	va_list args;
54 	struct ext2_sb_info *sbi = EXT2_SB(sb);
55 	struct ext2_super_block *es = sbi->s_es;
56 
57 	if (!sb_rdonly(sb)) {
58 		spin_lock(&sbi->s_lock);
59 		sbi->s_mount_state |= EXT2_ERROR_FS;
60 		es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
61 		spin_unlock(&sbi->s_lock);
62 		ext2_sync_super(sb, es, 1);
63 	}
64 
65 	va_start(args, fmt);
66 
67 	vaf.fmt = fmt;
68 	vaf.va = &args;
69 
70 	printk(KERN_CRIT "EXT2-fs (%s): error: %s: %pV\n",
71 	       sb->s_id, function, &vaf);
72 
73 	va_end(args);
74 
75 	if (test_opt(sb, ERRORS_PANIC))
76 		panic("EXT2-fs: panic from previous error\n");
77 	if (!sb_rdonly(sb) && test_opt(sb, ERRORS_RO)) {
78 		ext2_msg(sb, KERN_CRIT,
79 			     "error: remounting filesystem read-only");
80 		sb->s_flags |= SB_RDONLY;
81 	}
82 }
83 
84 void ext2_msg(struct super_block *sb, const char *prefix,
85 		const char *fmt, ...)
86 {
87 	struct va_format vaf;
88 	va_list args;
89 
90 	va_start(args, fmt);
91 
92 	vaf.fmt = fmt;
93 	vaf.va = &args;
94 
95 	printk("%sEXT2-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
96 
97 	va_end(args);
98 }
99 
100 /*
101  * This must be called with sbi->s_lock held.
102  */
103 void ext2_update_dynamic_rev(struct super_block *sb)
104 {
105 	struct ext2_super_block *es = EXT2_SB(sb)->s_es;
106 
107 	if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
108 		return;
109 
110 	ext2_msg(sb, KERN_WARNING,
111 		     "warning: updating to rev %d because of "
112 		     "new feature flag, running e2fsck is recommended",
113 		     EXT2_DYNAMIC_REV);
114 
115 	es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
116 	es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
117 	es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
118 	/* leave es->s_feature_*compat flags alone */
119 	/* es->s_uuid will be set by e2fsck if empty */
120 
121 	/*
122 	 * The rest of the superblock fields should be zero, and if not it
123 	 * means they are likely already in use, so leave them alone.  We
124 	 * can leave it up to e2fsck to clean up any inconsistencies there.
125 	 */
126 }
127 
128 #ifdef CONFIG_QUOTA
129 static int ext2_quota_off(struct super_block *sb, int type);
130 
131 static void ext2_quota_off_umount(struct super_block *sb)
132 {
133 	int type;
134 
135 	for (type = 0; type < MAXQUOTAS; type++)
136 		ext2_quota_off(sb, type);
137 }
138 #else
139 static inline void ext2_quota_off_umount(struct super_block *sb)
140 {
141 }
142 #endif
143 
144 static void ext2_put_super (struct super_block * sb)
145 {
146 	int db_count;
147 	int i;
148 	struct ext2_sb_info *sbi = EXT2_SB(sb);
149 
150 	ext2_quota_off_umount(sb);
151 
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_free_in_core_inode(struct inode *inode)
197 {
198 	kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
199 }
200 
201 static void init_once(void *foo)
202 {
203 	struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
204 
205 	rwlock_init(&ei->i_meta_lock);
206 #ifdef CONFIG_EXT2_FS_XATTR
207 	init_rwsem(&ei->xattr_sem);
208 #endif
209 	mutex_init(&ei->truncate_mutex);
210 #ifdef CONFIG_FS_DAX
211 	init_rwsem(&ei->dax_sem);
212 #endif
213 	inode_init_once(&ei->vfs_inode);
214 }
215 
216 static int __init init_inodecache(void)
217 {
218 	ext2_inode_cachep = kmem_cache_create_usercopy("ext2_inode_cache",
219 				sizeof(struct ext2_inode_info), 0,
220 				(SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
221 					SLAB_ACCOUNT),
222 				offsetof(struct ext2_inode_info, i_data),
223 				sizeof_field(struct ext2_inode_info, i_data),
224 				init_once);
225 	if (ext2_inode_cachep == NULL)
226 		return -ENOMEM;
227 	return 0;
228 }
229 
230 static void destroy_inodecache(void)
231 {
232 	/*
233 	 * Make sure all delayed rcu free inodes are flushed before we
234 	 * destroy cache.
235 	 */
236 	rcu_barrier();
237 	kmem_cache_destroy(ext2_inode_cachep);
238 }
239 
240 static int ext2_show_options(struct seq_file *seq, struct dentry *root)
241 {
242 	struct super_block *sb = root->d_sb;
243 	struct ext2_sb_info *sbi = EXT2_SB(sb);
244 	struct ext2_super_block *es = sbi->s_es;
245 	unsigned long def_mount_opts;
246 
247 	spin_lock(&sbi->s_lock);
248 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
249 
250 	if (sbi->s_sb_block != 1)
251 		seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
252 	if (test_opt(sb, MINIX_DF))
253 		seq_puts(seq, ",minixdf");
254 	if (test_opt(sb, GRPID))
255 		seq_puts(seq, ",grpid");
256 	if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
257 		seq_puts(seq, ",nogrpid");
258 	if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT2_DEF_RESUID)) ||
259 	    le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
260 		seq_printf(seq, ",resuid=%u",
261 				from_kuid_munged(&init_user_ns, sbi->s_resuid));
262 	}
263 	if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT2_DEF_RESGID)) ||
264 	    le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
265 		seq_printf(seq, ",resgid=%u",
266 				from_kgid_munged(&init_user_ns, sbi->s_resgid));
267 	}
268 	if (test_opt(sb, ERRORS_RO)) {
269 		int def_errors = le16_to_cpu(es->s_errors);
270 
271 		if (def_errors == EXT2_ERRORS_PANIC ||
272 		    def_errors == EXT2_ERRORS_CONTINUE) {
273 			seq_puts(seq, ",errors=remount-ro");
274 		}
275 	}
276 	if (test_opt(sb, ERRORS_CONT))
277 		seq_puts(seq, ",errors=continue");
278 	if (test_opt(sb, ERRORS_PANIC))
279 		seq_puts(seq, ",errors=panic");
280 	if (test_opt(sb, NO_UID32))
281 		seq_puts(seq, ",nouid32");
282 	if (test_opt(sb, DEBUG))
283 		seq_puts(seq, ",debug");
284 	if (test_opt(sb, OLDALLOC))
285 		seq_puts(seq, ",oldalloc");
286 
287 #ifdef CONFIG_EXT2_FS_XATTR
288 	if (test_opt(sb, XATTR_USER))
289 		seq_puts(seq, ",user_xattr");
290 	if (!test_opt(sb, XATTR_USER) &&
291 	    (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
292 		seq_puts(seq, ",nouser_xattr");
293 	}
294 #endif
295 
296 #ifdef CONFIG_EXT2_FS_POSIX_ACL
297 	if (test_opt(sb, POSIX_ACL))
298 		seq_puts(seq, ",acl");
299 	if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
300 		seq_puts(seq, ",noacl");
301 #endif
302 
303 	if (test_opt(sb, NOBH))
304 		seq_puts(seq, ",nobh");
305 
306 	if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
307 		seq_puts(seq, ",usrquota");
308 
309 	if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
310 		seq_puts(seq, ",grpquota");
311 
312 	if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
313 		seq_puts(seq, ",xip");
314 
315 	if (sbi->s_mount_opt & EXT2_MOUNT_DAX)
316 		seq_puts(seq, ",dax");
317 
318 	if (!test_opt(sb, RESERVATION))
319 		seq_puts(seq, ",noreservation");
320 
321 	spin_unlock(&sbi->s_lock);
322 	return 0;
323 }
324 
325 #ifdef CONFIG_QUOTA
326 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
327 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
328 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
329 			 const struct path *path);
330 static struct dquot **ext2_get_dquots(struct inode *inode)
331 {
332 	return EXT2_I(inode)->i_dquot;
333 }
334 
335 static const struct quotactl_ops ext2_quotactl_ops = {
336 	.quota_on	= ext2_quota_on,
337 	.quota_off	= ext2_quota_off,
338 	.quota_sync	= dquot_quota_sync,
339 	.get_state	= dquot_get_state,
340 	.set_info	= dquot_set_dqinfo,
341 	.get_dqblk	= dquot_get_dqblk,
342 	.set_dqblk	= dquot_set_dqblk,
343 	.get_nextdqblk	= dquot_get_next_dqblk,
344 };
345 #endif
346 
347 static const struct super_operations ext2_sops = {
348 	.alloc_inode	= ext2_alloc_inode,
349 	.free_inode	= ext2_free_in_core_inode,
350 	.write_inode	= ext2_write_inode,
351 	.evict_inode	= ext2_evict_inode,
352 	.put_super	= ext2_put_super,
353 	.sync_fs	= ext2_sync_fs,
354 	.freeze_fs	= ext2_freeze,
355 	.unfreeze_fs	= ext2_unfreeze,
356 	.statfs		= ext2_statfs,
357 	.remount_fs	= ext2_remount,
358 	.show_options	= ext2_show_options,
359 #ifdef CONFIG_QUOTA
360 	.quota_read	= ext2_quota_read,
361 	.quota_write	= ext2_quota_write,
362 	.get_dquots	= ext2_get_dquots,
363 #endif
364 };
365 
366 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
367 		u64 ino, u32 generation)
368 {
369 	struct inode *inode;
370 
371 	if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
372 		return ERR_PTR(-ESTALE);
373 	if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
374 		return ERR_PTR(-ESTALE);
375 
376 	/*
377 	 * ext2_iget isn't quite right if the inode is currently unallocated!
378 	 * However ext2_iget currently does appropriate checks to handle stale
379 	 * inodes so everything is OK.
380 	 */
381 	inode = ext2_iget(sb, ino);
382 	if (IS_ERR(inode))
383 		return ERR_CAST(inode);
384 	if (generation && inode->i_generation != generation) {
385 		/* we didn't find the right inode.. */
386 		iput(inode);
387 		return ERR_PTR(-ESTALE);
388 	}
389 	return inode;
390 }
391 
392 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
393 		int fh_len, int fh_type)
394 {
395 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
396 				    ext2_nfs_get_inode);
397 }
398 
399 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
400 		int fh_len, int fh_type)
401 {
402 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
403 				    ext2_nfs_get_inode);
404 }
405 
406 static const struct export_operations ext2_export_ops = {
407 	.fh_to_dentry = ext2_fh_to_dentry,
408 	.fh_to_parent = ext2_fh_to_parent,
409 	.get_parent = ext2_get_parent,
410 };
411 
412 static unsigned long get_sb_block(void **data)
413 {
414 	unsigned long 	sb_block;
415 	char 		*options = (char *) *data;
416 
417 	if (!options || strncmp(options, "sb=", 3) != 0)
418 		return 1;	/* Default location */
419 	options += 3;
420 	sb_block = simple_strtoul(options, &options, 0);
421 	if (*options && *options != ',') {
422 		printk("EXT2-fs: Invalid sb specification: %s\n",
423 		       (char *) *data);
424 		return 1;
425 	}
426 	if (*options == ',')
427 		options++;
428 	*data = (void *) options;
429 	return sb_block;
430 }
431 
432 enum {
433 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
434 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
435 	Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
436 	Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
437 	Opt_acl, Opt_noacl, Opt_xip, Opt_dax, Opt_ignore, Opt_err, Opt_quota,
438 	Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
439 };
440 
441 static const match_table_t tokens = {
442 	{Opt_bsd_df, "bsddf"},
443 	{Opt_minix_df, "minixdf"},
444 	{Opt_grpid, "grpid"},
445 	{Opt_grpid, "bsdgroups"},
446 	{Opt_nogrpid, "nogrpid"},
447 	{Opt_nogrpid, "sysvgroups"},
448 	{Opt_resgid, "resgid=%u"},
449 	{Opt_resuid, "resuid=%u"},
450 	{Opt_sb, "sb=%u"},
451 	{Opt_err_cont, "errors=continue"},
452 	{Opt_err_panic, "errors=panic"},
453 	{Opt_err_ro, "errors=remount-ro"},
454 	{Opt_nouid32, "nouid32"},
455 	{Opt_nocheck, "check=none"},
456 	{Opt_nocheck, "nocheck"},
457 	{Opt_debug, "debug"},
458 	{Opt_oldalloc, "oldalloc"},
459 	{Opt_orlov, "orlov"},
460 	{Opt_nobh, "nobh"},
461 	{Opt_user_xattr, "user_xattr"},
462 	{Opt_nouser_xattr, "nouser_xattr"},
463 	{Opt_acl, "acl"},
464 	{Opt_noacl, "noacl"},
465 	{Opt_xip, "xip"},
466 	{Opt_dax, "dax"},
467 	{Opt_grpquota, "grpquota"},
468 	{Opt_ignore, "noquota"},
469 	{Opt_quota, "quota"},
470 	{Opt_usrquota, "usrquota"},
471 	{Opt_reservation, "reservation"},
472 	{Opt_noreservation, "noreservation"},
473 	{Opt_err, NULL}
474 };
475 
476 static int parse_options(char *options, struct super_block *sb,
477 			 struct ext2_mount_options *opts)
478 {
479 	char *p;
480 	substring_t args[MAX_OPT_ARGS];
481 	int option;
482 	kuid_t uid;
483 	kgid_t gid;
484 
485 	if (!options)
486 		return 1;
487 
488 	while ((p = strsep (&options, ",")) != NULL) {
489 		int token;
490 		if (!*p)
491 			continue;
492 
493 		token = match_token(p, tokens, args);
494 		switch (token) {
495 		case Opt_bsd_df:
496 			clear_opt (opts->s_mount_opt, MINIX_DF);
497 			break;
498 		case Opt_minix_df:
499 			set_opt (opts->s_mount_opt, MINIX_DF);
500 			break;
501 		case Opt_grpid:
502 			set_opt (opts->s_mount_opt, GRPID);
503 			break;
504 		case Opt_nogrpid:
505 			clear_opt (opts->s_mount_opt, GRPID);
506 			break;
507 		case Opt_resuid:
508 			if (match_int(&args[0], &option))
509 				return 0;
510 			uid = make_kuid(current_user_ns(), option);
511 			if (!uid_valid(uid)) {
512 				ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
513 				return 0;
514 
515 			}
516 			opts->s_resuid = uid;
517 			break;
518 		case Opt_resgid:
519 			if (match_int(&args[0], &option))
520 				return 0;
521 			gid = make_kgid(current_user_ns(), option);
522 			if (!gid_valid(gid)) {
523 				ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
524 				return 0;
525 			}
526 			opts->s_resgid = gid;
527 			break;
528 		case Opt_sb:
529 			/* handled by get_sb_block() instead of here */
530 			/* *sb_block = match_int(&args[0]); */
531 			break;
532 		case Opt_err_panic:
533 			clear_opt (opts->s_mount_opt, ERRORS_CONT);
534 			clear_opt (opts->s_mount_opt, ERRORS_RO);
535 			set_opt (opts->s_mount_opt, ERRORS_PANIC);
536 			break;
537 		case Opt_err_ro:
538 			clear_opt (opts->s_mount_opt, ERRORS_CONT);
539 			clear_opt (opts->s_mount_opt, ERRORS_PANIC);
540 			set_opt (opts->s_mount_opt, ERRORS_RO);
541 			break;
542 		case Opt_err_cont:
543 			clear_opt (opts->s_mount_opt, ERRORS_RO);
544 			clear_opt (opts->s_mount_opt, ERRORS_PANIC);
545 			set_opt (opts->s_mount_opt, ERRORS_CONT);
546 			break;
547 		case Opt_nouid32:
548 			set_opt (opts->s_mount_opt, NO_UID32);
549 			break;
550 		case Opt_nocheck:
551 			ext2_msg(sb, KERN_WARNING,
552 				"Option nocheck/check=none is deprecated and"
553 				" will be removed in June 2020.");
554 			clear_opt (opts->s_mount_opt, CHECK);
555 			break;
556 		case Opt_debug:
557 			set_opt (opts->s_mount_opt, DEBUG);
558 			break;
559 		case Opt_oldalloc:
560 			set_opt (opts->s_mount_opt, OLDALLOC);
561 			break;
562 		case Opt_orlov:
563 			clear_opt (opts->s_mount_opt, OLDALLOC);
564 			break;
565 		case Opt_nobh:
566 			set_opt (opts->s_mount_opt, NOBH);
567 			break;
568 #ifdef CONFIG_EXT2_FS_XATTR
569 		case Opt_user_xattr:
570 			set_opt (opts->s_mount_opt, XATTR_USER);
571 			break;
572 		case Opt_nouser_xattr:
573 			clear_opt (opts->s_mount_opt, XATTR_USER);
574 			break;
575 #else
576 		case Opt_user_xattr:
577 		case Opt_nouser_xattr:
578 			ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
579 				"not supported");
580 			break;
581 #endif
582 #ifdef CONFIG_EXT2_FS_POSIX_ACL
583 		case Opt_acl:
584 			set_opt(opts->s_mount_opt, POSIX_ACL);
585 			break;
586 		case Opt_noacl:
587 			clear_opt(opts->s_mount_opt, POSIX_ACL);
588 			break;
589 #else
590 		case Opt_acl:
591 		case Opt_noacl:
592 			ext2_msg(sb, KERN_INFO,
593 				"(no)acl options not supported");
594 			break;
595 #endif
596 		case Opt_xip:
597 			ext2_msg(sb, KERN_INFO, "use dax instead of xip");
598 			set_opt(opts->s_mount_opt, XIP);
599 			/* Fall through */
600 		case Opt_dax:
601 #ifdef CONFIG_FS_DAX
602 			ext2_msg(sb, KERN_WARNING,
603 		"DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
604 			set_opt(opts->s_mount_opt, DAX);
605 #else
606 			ext2_msg(sb, KERN_INFO, "dax option not supported");
607 #endif
608 			break;
609 
610 #if defined(CONFIG_QUOTA)
611 		case Opt_quota:
612 		case Opt_usrquota:
613 			set_opt(opts->s_mount_opt, USRQUOTA);
614 			break;
615 
616 		case Opt_grpquota:
617 			set_opt(opts->s_mount_opt, GRPQUOTA);
618 			break;
619 #else
620 		case Opt_quota:
621 		case Opt_usrquota:
622 		case Opt_grpquota:
623 			ext2_msg(sb, KERN_INFO,
624 				"quota operations not supported");
625 			break;
626 #endif
627 
628 		case Opt_reservation:
629 			set_opt(opts->s_mount_opt, RESERVATION);
630 			ext2_msg(sb, KERN_INFO, "reservations ON");
631 			break;
632 		case Opt_noreservation:
633 			clear_opt(opts->s_mount_opt, RESERVATION);
634 			ext2_msg(sb, KERN_INFO, "reservations OFF");
635 			break;
636 		case Opt_ignore:
637 			break;
638 		default:
639 			return 0;
640 		}
641 	}
642 	return 1;
643 }
644 
645 static int ext2_setup_super (struct super_block * sb,
646 			      struct ext2_super_block * es,
647 			      int read_only)
648 {
649 	int res = 0;
650 	struct ext2_sb_info *sbi = EXT2_SB(sb);
651 
652 	if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
653 		ext2_msg(sb, KERN_ERR,
654 			"error: revision level too high, "
655 			"forcing read-only mode");
656 		res = SB_RDONLY;
657 	}
658 	if (read_only)
659 		return res;
660 	if (!(sbi->s_mount_state & EXT2_VALID_FS))
661 		ext2_msg(sb, KERN_WARNING,
662 			"warning: mounting unchecked fs, "
663 			"running e2fsck is recommended");
664 	else if ((sbi->s_mount_state & EXT2_ERROR_FS))
665 		ext2_msg(sb, KERN_WARNING,
666 			"warning: mounting fs with errors, "
667 			"running e2fsck is recommended");
668 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
669 		 le16_to_cpu(es->s_mnt_count) >=
670 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
671 		ext2_msg(sb, KERN_WARNING,
672 			"warning: maximal mount count reached, "
673 			"running e2fsck is recommended");
674 	else if (le32_to_cpu(es->s_checkinterval) &&
675 		(le32_to_cpu(es->s_lastcheck) +
676 			le32_to_cpu(es->s_checkinterval) <=
677 			ktime_get_real_seconds()))
678 		ext2_msg(sb, KERN_WARNING,
679 			"warning: checktime reached, "
680 			"running e2fsck is recommended");
681 	if (!le16_to_cpu(es->s_max_mnt_count))
682 		es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
683 	le16_add_cpu(&es->s_mnt_count, 1);
684 	if (test_opt (sb, DEBUG))
685 		ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
686 			"bpg=%lu, ipg=%lu, mo=%04lx]",
687 			EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
688 			sbi->s_frag_size,
689 			sbi->s_groups_count,
690 			EXT2_BLOCKS_PER_GROUP(sb),
691 			EXT2_INODES_PER_GROUP(sb),
692 			sbi->s_mount_opt);
693 	return res;
694 }
695 
696 static int ext2_check_descriptors(struct super_block *sb)
697 {
698 	int i;
699 	struct ext2_sb_info *sbi = EXT2_SB(sb);
700 
701 	ext2_debug ("Checking group descriptors");
702 
703 	for (i = 0; i < sbi->s_groups_count; i++) {
704 		struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
705 		ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
706 		ext2_fsblk_t last_block;
707 
708 		if (i == sbi->s_groups_count - 1)
709 			last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
710 		else
711 			last_block = first_block +
712 				(EXT2_BLOCKS_PER_GROUP(sb) - 1);
713 
714 		if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
715 		    le32_to_cpu(gdp->bg_block_bitmap) > last_block)
716 		{
717 			ext2_error (sb, "ext2_check_descriptors",
718 				    "Block bitmap for group %d"
719 				    " not in group (block %lu)!",
720 				    i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
721 			return 0;
722 		}
723 		if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
724 		    le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
725 		{
726 			ext2_error (sb, "ext2_check_descriptors",
727 				    "Inode bitmap for group %d"
728 				    " not in group (block %lu)!",
729 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
730 			return 0;
731 		}
732 		if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
733 		    le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
734 		    last_block)
735 		{
736 			ext2_error (sb, "ext2_check_descriptors",
737 				    "Inode table for group %d"
738 				    " not in group (block %lu)!",
739 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
740 			return 0;
741 		}
742 	}
743 	return 1;
744 }
745 
746 /*
747  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
748  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
749  * We need to be 1 filesystem block less than the 2^32 sector limit.
750  */
751 static loff_t ext2_max_size(int bits)
752 {
753 	loff_t res = EXT2_NDIR_BLOCKS;
754 	int meta_blocks;
755 	unsigned int upper_limit;
756 	unsigned int ppb = 1 << (bits-2);
757 
758 	/* This is calculated to be the largest file size for a
759 	 * dense, file such that the total number of
760 	 * sectors in the file, including data and all indirect blocks,
761 	 * does not exceed 2^32 -1
762 	 * __u32 i_blocks representing the total number of
763 	 * 512 bytes blocks of the file
764 	 */
765 	upper_limit = (1LL << 32) - 1;
766 
767 	/* total blocks in file system block size */
768 	upper_limit >>= (bits - 9);
769 
770 	/* Compute how many blocks we can address by block tree */
771 	res += 1LL << (bits-2);
772 	res += 1LL << (2*(bits-2));
773 	res += 1LL << (3*(bits-2));
774 	/* Does block tree limit file size? */
775 	if (res < upper_limit)
776 		goto check_lfs;
777 
778 	res = upper_limit;
779 	/* How many metadata blocks are needed for addressing upper_limit? */
780 	upper_limit -= EXT2_NDIR_BLOCKS;
781 	/* indirect blocks */
782 	meta_blocks = 1;
783 	upper_limit -= ppb;
784 	/* double indirect blocks */
785 	if (upper_limit < ppb * ppb) {
786 		meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb);
787 		res -= meta_blocks;
788 		goto check_lfs;
789 	}
790 	meta_blocks += 1 + ppb;
791 	upper_limit -= ppb * ppb;
792 	/* tripple indirect blocks for the rest */
793 	meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb) +
794 		DIV_ROUND_UP(upper_limit, ppb*ppb);
795 	res -= meta_blocks;
796 check_lfs:
797 	res <<= bits;
798 	if (res > MAX_LFS_FILESIZE)
799 		res = MAX_LFS_FILESIZE;
800 
801 	return res;
802 }
803 
804 static unsigned long descriptor_loc(struct super_block *sb,
805 				    unsigned long logic_sb_block,
806 				    int nr)
807 {
808 	struct ext2_sb_info *sbi = EXT2_SB(sb);
809 	unsigned long bg, first_meta_bg;
810 	int has_super = 0;
811 
812 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
813 
814 	if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
815 	    nr < first_meta_bg)
816 		return (logic_sb_block + nr + 1);
817 	bg = sbi->s_desc_per_block * nr;
818 	if (ext2_bg_has_super(sb, bg))
819 		has_super = 1;
820 
821 	return ext2_group_first_block_no(sb, bg) + has_super;
822 }
823 
824 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
825 {
826 	struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
827 	struct buffer_head * bh;
828 	struct ext2_sb_info * sbi;
829 	struct ext2_super_block * es;
830 	struct inode *root;
831 	unsigned long block;
832 	unsigned long sb_block = get_sb_block(&data);
833 	unsigned long logic_sb_block;
834 	unsigned long offset = 0;
835 	unsigned long def_mount_opts;
836 	long ret = -ENOMEM;
837 	int blocksize = BLOCK_SIZE;
838 	int db_count;
839 	int i, j;
840 	__le32 features;
841 	int err;
842 	struct ext2_mount_options opts;
843 
844 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
845 	if (!sbi)
846 		goto failed;
847 
848 	sbi->s_blockgroup_lock =
849 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
850 	if (!sbi->s_blockgroup_lock) {
851 		kfree(sbi);
852 		goto failed;
853 	}
854 	sb->s_fs_info = sbi;
855 	sbi->s_sb_block = sb_block;
856 	sbi->s_daxdev = dax_dev;
857 
858 	spin_lock_init(&sbi->s_lock);
859 	ret = -EINVAL;
860 
861 	/*
862 	 * See what the current blocksize for the device is, and
863 	 * use that as the blocksize.  Otherwise (or if the blocksize
864 	 * is smaller than the default) use the default.
865 	 * This is important for devices that have a hardware
866 	 * sectorsize that is larger than the default.
867 	 */
868 	blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
869 	if (!blocksize) {
870 		ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
871 		goto failed_sbi;
872 	}
873 
874 	/*
875 	 * If the superblock doesn't start on a hardware sector boundary,
876 	 * calculate the offset.
877 	 */
878 	if (blocksize != BLOCK_SIZE) {
879 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
880 		offset = (sb_block*BLOCK_SIZE) % blocksize;
881 	} else {
882 		logic_sb_block = sb_block;
883 	}
884 
885 	if (!(bh = sb_bread(sb, logic_sb_block))) {
886 		ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
887 		goto failed_sbi;
888 	}
889 	/*
890 	 * Note: s_es must be initialized as soon as possible because
891 	 *       some ext2 macro-instructions depend on its value
892 	 */
893 	es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
894 	sbi->s_es = es;
895 	sb->s_magic = le16_to_cpu(es->s_magic);
896 
897 	if (sb->s_magic != EXT2_SUPER_MAGIC)
898 		goto cantfind_ext2;
899 
900 	opts.s_mount_opt = 0;
901 	/* Set defaults before we parse the mount options */
902 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
903 	if (def_mount_opts & EXT2_DEFM_DEBUG)
904 		set_opt(opts.s_mount_opt, DEBUG);
905 	if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
906 		set_opt(opts.s_mount_opt, GRPID);
907 	if (def_mount_opts & EXT2_DEFM_UID16)
908 		set_opt(opts.s_mount_opt, NO_UID32);
909 #ifdef CONFIG_EXT2_FS_XATTR
910 	if (def_mount_opts & EXT2_DEFM_XATTR_USER)
911 		set_opt(opts.s_mount_opt, XATTR_USER);
912 #endif
913 #ifdef CONFIG_EXT2_FS_POSIX_ACL
914 	if (def_mount_opts & EXT2_DEFM_ACL)
915 		set_opt(opts.s_mount_opt, POSIX_ACL);
916 #endif
917 
918 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
919 		set_opt(opts.s_mount_opt, ERRORS_PANIC);
920 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
921 		set_opt(opts.s_mount_opt, ERRORS_CONT);
922 	else
923 		set_opt(opts.s_mount_opt, ERRORS_RO);
924 
925 	opts.s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
926 	opts.s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
927 
928 	set_opt(opts.s_mount_opt, RESERVATION);
929 
930 	if (!parse_options((char *) data, sb, &opts))
931 		goto failed_mount;
932 
933 	sbi->s_mount_opt = opts.s_mount_opt;
934 	sbi->s_resuid = opts.s_resuid;
935 	sbi->s_resgid = opts.s_resgid;
936 
937 	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
938 		((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
939 		 SB_POSIXACL : 0);
940 	sb->s_iflags |= SB_I_CGROUPWB;
941 
942 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
943 	    (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
944 	     EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
945 	     EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
946 		ext2_msg(sb, KERN_WARNING,
947 			"warning: feature flags set on rev 0 fs, "
948 			"running e2fsck is recommended");
949 	/*
950 	 * Check feature flags regardless of the revision level, since we
951 	 * previously didn't change the revision level when setting the flags,
952 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
953 	 */
954 	features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
955 	if (features) {
956 		ext2_msg(sb, KERN_ERR,	"error: couldn't mount because of "
957 		       "unsupported optional features (%x)",
958 			le32_to_cpu(features));
959 		goto failed_mount;
960 	}
961 	if (!sb_rdonly(sb) && (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
962 		ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
963 		       "unsupported optional features (%x)",
964 		       le32_to_cpu(features));
965 		goto failed_mount;
966 	}
967 
968 	blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
969 
970 	if (sbi->s_mount_opt & EXT2_MOUNT_DAX) {
971 		if (!bdev_dax_supported(sb->s_bdev, blocksize)) {
972 			ext2_msg(sb, KERN_ERR,
973 				"DAX unsupported by block device. Turning off DAX.");
974 			sbi->s_mount_opt &= ~EXT2_MOUNT_DAX;
975 		}
976 	}
977 
978 	/* If the blocksize doesn't match, re-read the thing.. */
979 	if (sb->s_blocksize != blocksize) {
980 		brelse(bh);
981 
982 		if (!sb_set_blocksize(sb, blocksize)) {
983 			ext2_msg(sb, KERN_ERR,
984 				"error: bad blocksize %d", blocksize);
985 			goto failed_sbi;
986 		}
987 
988 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
989 		offset = (sb_block*BLOCK_SIZE) % blocksize;
990 		bh = sb_bread(sb, logic_sb_block);
991 		if(!bh) {
992 			ext2_msg(sb, KERN_ERR, "error: couldn't read"
993 				"superblock on 2nd try");
994 			goto failed_sbi;
995 		}
996 		es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
997 		sbi->s_es = es;
998 		if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
999 			ext2_msg(sb, KERN_ERR, "error: magic mismatch");
1000 			goto failed_mount;
1001 		}
1002 	}
1003 
1004 	sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
1005 	sb->s_max_links = EXT2_LINK_MAX;
1006 
1007 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
1008 		sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
1009 		sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
1010 	} else {
1011 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1012 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1013 		if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
1014 		    !is_power_of_2(sbi->s_inode_size) ||
1015 		    (sbi->s_inode_size > blocksize)) {
1016 			ext2_msg(sb, KERN_ERR,
1017 				"error: unsupported inode size: %d",
1018 				sbi->s_inode_size);
1019 			goto failed_mount;
1020 		}
1021 	}
1022 
1023 	sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
1024 				   le32_to_cpu(es->s_log_frag_size);
1025 	if (sbi->s_frag_size == 0)
1026 		goto cantfind_ext2;
1027 	sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
1028 
1029 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1030 	sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1031 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1032 
1033 	sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
1034 	if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
1035 		goto cantfind_ext2;
1036 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
1037 					sbi->s_inodes_per_block;
1038 	sbi->s_desc_per_block = sb->s_blocksize /
1039 					sizeof (struct ext2_group_desc);
1040 	sbi->s_sbh = bh;
1041 	sbi->s_mount_state = le16_to_cpu(es->s_state);
1042 	sbi->s_addr_per_block_bits =
1043 		ilog2 (EXT2_ADDR_PER_BLOCK(sb));
1044 	sbi->s_desc_per_block_bits =
1045 		ilog2 (EXT2_DESC_PER_BLOCK(sb));
1046 
1047 	if (sb->s_magic != EXT2_SUPER_MAGIC)
1048 		goto cantfind_ext2;
1049 
1050 	if (sb->s_blocksize != bh->b_size) {
1051 		if (!silent)
1052 			ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
1053 		goto failed_mount;
1054 	}
1055 
1056 	if (sb->s_blocksize != sbi->s_frag_size) {
1057 		ext2_msg(sb, KERN_ERR,
1058 			"error: fragsize %lu != blocksize %lu"
1059 			"(not supported yet)",
1060 			sbi->s_frag_size, sb->s_blocksize);
1061 		goto failed_mount;
1062 	}
1063 
1064 	if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
1065 		ext2_msg(sb, KERN_ERR,
1066 			"error: #blocks per group too big: %lu",
1067 			sbi->s_blocks_per_group);
1068 		goto failed_mount;
1069 	}
1070 	if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
1071 		ext2_msg(sb, KERN_ERR,
1072 			"error: #fragments per group too big: %lu",
1073 			sbi->s_frags_per_group);
1074 		goto failed_mount;
1075 	}
1076 	if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1077 		ext2_msg(sb, KERN_ERR,
1078 			"error: #inodes per group too big: %lu",
1079 			sbi->s_inodes_per_group);
1080 		goto failed_mount;
1081 	}
1082 
1083 	if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1084 		goto cantfind_ext2;
1085  	sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1086  				le32_to_cpu(es->s_first_data_block) - 1)
1087  					/ EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1088 	db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1089 		   EXT2_DESC_PER_BLOCK(sb);
1090 	sbi->s_group_desc = kmalloc_array (db_count,
1091 					   sizeof(struct buffer_head *),
1092 					   GFP_KERNEL);
1093 	if (sbi->s_group_desc == NULL) {
1094 		ret = -ENOMEM;
1095 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1096 		goto failed_mount;
1097 	}
1098 	bgl_lock_init(sbi->s_blockgroup_lock);
1099 	sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1100 	if (!sbi->s_debts) {
1101 		ret = -ENOMEM;
1102 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1103 		goto failed_mount_group_desc;
1104 	}
1105 	for (i = 0; i < db_count; i++) {
1106 		block = descriptor_loc(sb, logic_sb_block, i);
1107 		sbi->s_group_desc[i] = sb_bread(sb, block);
1108 		if (!sbi->s_group_desc[i]) {
1109 			for (j = 0; j < i; j++)
1110 				brelse (sbi->s_group_desc[j]);
1111 			ext2_msg(sb, KERN_ERR,
1112 				"error: unable to read group descriptors");
1113 			goto failed_mount_group_desc;
1114 		}
1115 	}
1116 	if (!ext2_check_descriptors (sb)) {
1117 		ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1118 		goto failed_mount2;
1119 	}
1120 	sbi->s_gdb_count = db_count;
1121 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1122 	spin_lock_init(&sbi->s_next_gen_lock);
1123 
1124 	/* per fileystem reservation list head & lock */
1125 	spin_lock_init(&sbi->s_rsv_window_lock);
1126 	sbi->s_rsv_window_root = RB_ROOT;
1127 	/*
1128 	 * Add a single, static dummy reservation to the start of the
1129 	 * reservation window list --- it gives us a placeholder for
1130 	 * append-at-start-of-list which makes the allocation logic
1131 	 * _much_ simpler.
1132 	 */
1133 	sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1134 	sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1135 	sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1136 	sbi->s_rsv_window_head.rsv_goal_size = 0;
1137 	ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1138 
1139 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
1140 				ext2_count_free_blocks(sb), GFP_KERNEL);
1141 	if (!err) {
1142 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
1143 				ext2_count_free_inodes(sb), GFP_KERNEL);
1144 	}
1145 	if (!err) {
1146 		err = percpu_counter_init(&sbi->s_dirs_counter,
1147 				ext2_count_dirs(sb), GFP_KERNEL);
1148 	}
1149 	if (err) {
1150 		ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1151 		goto failed_mount3;
1152 	}
1153 
1154 #ifdef CONFIG_EXT2_FS_XATTR
1155 	sbi->s_ea_block_cache = ext2_xattr_create_cache();
1156 	if (!sbi->s_ea_block_cache) {
1157 		ret = -ENOMEM;
1158 		ext2_msg(sb, KERN_ERR, "Failed to create ea_block_cache");
1159 		goto failed_mount3;
1160 	}
1161 #endif
1162 	/*
1163 	 * set up enough so that it can read an inode
1164 	 */
1165 	sb->s_op = &ext2_sops;
1166 	sb->s_export_op = &ext2_export_ops;
1167 	sb->s_xattr = ext2_xattr_handlers;
1168 
1169 #ifdef CONFIG_QUOTA
1170 	sb->dq_op = &dquot_operations;
1171 	sb->s_qcop = &ext2_quotactl_ops;
1172 	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1173 #endif
1174 
1175 	root = ext2_iget(sb, EXT2_ROOT_INO);
1176 	if (IS_ERR(root)) {
1177 		ret = PTR_ERR(root);
1178 		goto failed_mount3;
1179 	}
1180 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1181 		iput(root);
1182 		ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1183 		goto failed_mount3;
1184 	}
1185 
1186 	sb->s_root = d_make_root(root);
1187 	if (!sb->s_root) {
1188 		ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1189 		ret = -ENOMEM;
1190 		goto failed_mount3;
1191 	}
1192 	if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1193 		ext2_msg(sb, KERN_WARNING,
1194 			"warning: mounting ext3 filesystem as ext2");
1195 	if (ext2_setup_super (sb, es, sb_rdonly(sb)))
1196 		sb->s_flags |= SB_RDONLY;
1197 	ext2_write_super(sb);
1198 	return 0;
1199 
1200 cantfind_ext2:
1201 	if (!silent)
1202 		ext2_msg(sb, KERN_ERR,
1203 			"error: can't find an ext2 filesystem on dev %s.",
1204 			sb->s_id);
1205 	goto failed_mount;
1206 failed_mount3:
1207 	ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
1208 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
1209 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
1210 	percpu_counter_destroy(&sbi->s_dirs_counter);
1211 failed_mount2:
1212 	for (i = 0; i < db_count; i++)
1213 		brelse(sbi->s_group_desc[i]);
1214 failed_mount_group_desc:
1215 	kfree(sbi->s_group_desc);
1216 	kfree(sbi->s_debts);
1217 failed_mount:
1218 	brelse(bh);
1219 failed_sbi:
1220 	sb->s_fs_info = NULL;
1221 	kfree(sbi->s_blockgroup_lock);
1222 	kfree(sbi);
1223 failed:
1224 	fs_put_dax(dax_dev);
1225 	return ret;
1226 }
1227 
1228 static void ext2_clear_super_error(struct super_block *sb)
1229 {
1230 	struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1231 
1232 	if (buffer_write_io_error(sbh)) {
1233 		/*
1234 		 * Oh, dear.  A previous attempt to write the
1235 		 * superblock failed.  This could happen because the
1236 		 * USB device was yanked out.  Or it could happen to
1237 		 * be a transient write error and maybe the block will
1238 		 * be remapped.  Nothing we can do but to retry the
1239 		 * write and hope for the best.
1240 		 */
1241 		ext2_msg(sb, KERN_ERR,
1242 		       "previous I/O error to superblock detected");
1243 		clear_buffer_write_io_error(sbh);
1244 		set_buffer_uptodate(sbh);
1245 	}
1246 }
1247 
1248 void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1249 		     int wait)
1250 {
1251 	ext2_clear_super_error(sb);
1252 	spin_lock(&EXT2_SB(sb)->s_lock);
1253 	es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1254 	es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1255 	es->s_wtime = cpu_to_le32(ktime_get_real_seconds());
1256 	/* unlock before we do IO */
1257 	spin_unlock(&EXT2_SB(sb)->s_lock);
1258 	mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1259 	if (wait)
1260 		sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1261 }
1262 
1263 /*
1264  * In the second extended file system, it is not necessary to
1265  * write the super block since we use a mapping of the
1266  * disk super block in a buffer.
1267  *
1268  * However, this function is still used to set the fs valid
1269  * flags to 0.  We need to set this flag to 0 since the fs
1270  * may have been checked while mounted and e2fsck may have
1271  * set s_state to EXT2_VALID_FS after some corrections.
1272  */
1273 static int ext2_sync_fs(struct super_block *sb, int wait)
1274 {
1275 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1276 	struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1277 
1278 	/*
1279 	 * Write quota structures to quota file, sync_blockdev() will write
1280 	 * them to disk later
1281 	 */
1282 	dquot_writeback_dquots(sb, -1);
1283 
1284 	spin_lock(&sbi->s_lock);
1285 	if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1286 		ext2_debug("setting valid to 0\n");
1287 		es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1288 	}
1289 	spin_unlock(&sbi->s_lock);
1290 	ext2_sync_super(sb, es, wait);
1291 	return 0;
1292 }
1293 
1294 static int ext2_freeze(struct super_block *sb)
1295 {
1296 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1297 
1298 	/*
1299 	 * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1300 	 * because we have unattached inodes and thus filesystem is not fully
1301 	 * consistent.
1302 	 */
1303 	if (atomic_long_read(&sb->s_remove_count)) {
1304 		ext2_sync_fs(sb, 1);
1305 		return 0;
1306 	}
1307 	/* Set EXT2_FS_VALID flag */
1308 	spin_lock(&sbi->s_lock);
1309 	sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1310 	spin_unlock(&sbi->s_lock);
1311 	ext2_sync_super(sb, sbi->s_es, 1);
1312 
1313 	return 0;
1314 }
1315 
1316 static int ext2_unfreeze(struct super_block *sb)
1317 {
1318 	/* Just write sb to clear EXT2_VALID_FS flag */
1319 	ext2_write_super(sb);
1320 
1321 	return 0;
1322 }
1323 
1324 static void ext2_write_super(struct super_block *sb)
1325 {
1326 	if (!sb_rdonly(sb))
1327 		ext2_sync_fs(sb, 1);
1328 }
1329 
1330 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1331 {
1332 	struct ext2_sb_info * sbi = EXT2_SB(sb);
1333 	struct ext2_super_block * es;
1334 	struct ext2_mount_options new_opts;
1335 	int err;
1336 
1337 	sync_filesystem(sb);
1338 
1339 	spin_lock(&sbi->s_lock);
1340 	new_opts.s_mount_opt = sbi->s_mount_opt;
1341 	new_opts.s_resuid = sbi->s_resuid;
1342 	new_opts.s_resgid = sbi->s_resgid;
1343 	spin_unlock(&sbi->s_lock);
1344 
1345 	if (!parse_options(data, sb, &new_opts))
1346 		return -EINVAL;
1347 
1348 	spin_lock(&sbi->s_lock);
1349 	es = sbi->s_es;
1350 	if ((sbi->s_mount_opt ^ new_opts.s_mount_opt) & EXT2_MOUNT_DAX) {
1351 		ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1352 			 "dax flag with busy inodes while remounting");
1353 		new_opts.s_mount_opt ^= EXT2_MOUNT_DAX;
1354 	}
1355 	if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1356 		goto out_set;
1357 	if (*flags & SB_RDONLY) {
1358 		if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1359 		    !(sbi->s_mount_state & EXT2_VALID_FS))
1360 			goto out_set;
1361 
1362 		/*
1363 		 * OK, we are remounting a valid rw partition rdonly, so set
1364 		 * the rdonly flag and then mark the partition as valid again.
1365 		 */
1366 		es->s_state = cpu_to_le16(sbi->s_mount_state);
1367 		es->s_mtime = cpu_to_le32(ktime_get_real_seconds());
1368 		spin_unlock(&sbi->s_lock);
1369 
1370 		err = dquot_suspend(sb, -1);
1371 		if (err < 0)
1372 			return err;
1373 
1374 		ext2_sync_super(sb, es, 1);
1375 	} else {
1376 		__le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1377 					       ~EXT2_FEATURE_RO_COMPAT_SUPP);
1378 		if (ret) {
1379 			spin_unlock(&sbi->s_lock);
1380 			ext2_msg(sb, KERN_WARNING,
1381 				"warning: couldn't remount RDWR because of "
1382 				"unsupported optional features (%x).",
1383 				le32_to_cpu(ret));
1384 			return -EROFS;
1385 		}
1386 		/*
1387 		 * Mounting a RDONLY partition read-write, so reread and
1388 		 * store the current valid flag.  (It may have been changed
1389 		 * by e2fsck since we originally mounted the partition.)
1390 		 */
1391 		sbi->s_mount_state = le16_to_cpu(es->s_state);
1392 		if (!ext2_setup_super (sb, es, 0))
1393 			sb->s_flags &= ~SB_RDONLY;
1394 		spin_unlock(&sbi->s_lock);
1395 
1396 		ext2_write_super(sb);
1397 
1398 		dquot_resume(sb, -1);
1399 	}
1400 
1401 	spin_lock(&sbi->s_lock);
1402 out_set:
1403 	sbi->s_mount_opt = new_opts.s_mount_opt;
1404 	sbi->s_resuid = new_opts.s_resuid;
1405 	sbi->s_resgid = new_opts.s_resgid;
1406 	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1407 		((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? SB_POSIXACL : 0);
1408 	spin_unlock(&sbi->s_lock);
1409 
1410 	return 0;
1411 }
1412 
1413 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1414 {
1415 	struct super_block *sb = dentry->d_sb;
1416 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1417 	struct ext2_super_block *es = sbi->s_es;
1418 	u64 fsid;
1419 
1420 	spin_lock(&sbi->s_lock);
1421 
1422 	if (test_opt (sb, MINIX_DF))
1423 		sbi->s_overhead_last = 0;
1424 	else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1425 		unsigned long i, overhead = 0;
1426 		smp_rmb();
1427 
1428 		/*
1429 		 * Compute the overhead (FS structures). This is constant
1430 		 * for a given filesystem unless the number of block groups
1431 		 * changes so we cache the previous value until it does.
1432 		 */
1433 
1434 		/*
1435 		 * All of the blocks before first_data_block are
1436 		 * overhead
1437 		 */
1438 		overhead = le32_to_cpu(es->s_first_data_block);
1439 
1440 		/*
1441 		 * Add the overhead attributed to the superblock and
1442 		 * block group descriptors.  If the sparse superblocks
1443 		 * feature is turned on, then not all groups have this.
1444 		 */
1445 		for (i = 0; i < sbi->s_groups_count; i++)
1446 			overhead += ext2_bg_has_super(sb, i) +
1447 				ext2_bg_num_gdb(sb, i);
1448 
1449 		/*
1450 		 * Every block group has an inode bitmap, a block
1451 		 * bitmap, and an inode table.
1452 		 */
1453 		overhead += (sbi->s_groups_count *
1454 			     (2 + sbi->s_itb_per_group));
1455 		sbi->s_overhead_last = overhead;
1456 		smp_wmb();
1457 		sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1458 	}
1459 
1460 	buf->f_type = EXT2_SUPER_MAGIC;
1461 	buf->f_bsize = sb->s_blocksize;
1462 	buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1463 	buf->f_bfree = ext2_count_free_blocks(sb);
1464 	es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1465 	buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1466 	if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1467 		buf->f_bavail = 0;
1468 	buf->f_files = le32_to_cpu(es->s_inodes_count);
1469 	buf->f_ffree = ext2_count_free_inodes(sb);
1470 	es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1471 	buf->f_namelen = EXT2_NAME_LEN;
1472 	fsid = le64_to_cpup((void *)es->s_uuid) ^
1473 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1474 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1475 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1476 	spin_unlock(&sbi->s_lock);
1477 	return 0;
1478 }
1479 
1480 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1481 	int flags, const char *dev_name, void *data)
1482 {
1483 	return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1484 }
1485 
1486 #ifdef CONFIG_QUOTA
1487 
1488 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1489  * acquiring the locks... As quota files are never truncated and quota code
1490  * itself serializes the operations (and no one else should touch the files)
1491  * we don't have to be afraid of races */
1492 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1493 			       size_t len, loff_t off)
1494 {
1495 	struct inode *inode = sb_dqopt(sb)->files[type];
1496 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1497 	int err = 0;
1498 	int offset = off & (sb->s_blocksize - 1);
1499 	int tocopy;
1500 	size_t toread;
1501 	struct buffer_head tmp_bh;
1502 	struct buffer_head *bh;
1503 	loff_t i_size = i_size_read(inode);
1504 
1505 	if (off > i_size)
1506 		return 0;
1507 	if (off+len > i_size)
1508 		len = i_size-off;
1509 	toread = len;
1510 	while (toread > 0) {
1511 		tocopy = sb->s_blocksize - offset < toread ?
1512 				sb->s_blocksize - offset : toread;
1513 
1514 		tmp_bh.b_state = 0;
1515 		tmp_bh.b_size = sb->s_blocksize;
1516 		err = ext2_get_block(inode, blk, &tmp_bh, 0);
1517 		if (err < 0)
1518 			return err;
1519 		if (!buffer_mapped(&tmp_bh))	/* A hole? */
1520 			memset(data, 0, tocopy);
1521 		else {
1522 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1523 			if (!bh)
1524 				return -EIO;
1525 			memcpy(data, bh->b_data+offset, tocopy);
1526 			brelse(bh);
1527 		}
1528 		offset = 0;
1529 		toread -= tocopy;
1530 		data += tocopy;
1531 		blk++;
1532 	}
1533 	return len;
1534 }
1535 
1536 /* Write to quotafile */
1537 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1538 				const char *data, size_t len, loff_t off)
1539 {
1540 	struct inode *inode = sb_dqopt(sb)->files[type];
1541 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1542 	int err = 0;
1543 	int offset = off & (sb->s_blocksize - 1);
1544 	int tocopy;
1545 	size_t towrite = len;
1546 	struct buffer_head tmp_bh;
1547 	struct buffer_head *bh;
1548 
1549 	while (towrite > 0) {
1550 		tocopy = sb->s_blocksize - offset < towrite ?
1551 				sb->s_blocksize - offset : towrite;
1552 
1553 		tmp_bh.b_state = 0;
1554 		tmp_bh.b_size = sb->s_blocksize;
1555 		err = ext2_get_block(inode, blk, &tmp_bh, 1);
1556 		if (err < 0)
1557 			goto out;
1558 		if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1559 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1560 		else
1561 			bh = sb_getblk(sb, tmp_bh.b_blocknr);
1562 		if (unlikely(!bh)) {
1563 			err = -EIO;
1564 			goto out;
1565 		}
1566 		lock_buffer(bh);
1567 		memcpy(bh->b_data+offset, data, tocopy);
1568 		flush_dcache_page(bh->b_page);
1569 		set_buffer_uptodate(bh);
1570 		mark_buffer_dirty(bh);
1571 		unlock_buffer(bh);
1572 		brelse(bh);
1573 		offset = 0;
1574 		towrite -= tocopy;
1575 		data += tocopy;
1576 		blk++;
1577 	}
1578 out:
1579 	if (len == towrite)
1580 		return err;
1581 	if (inode->i_size < off+len-towrite)
1582 		i_size_write(inode, off+len-towrite);
1583 	inode_inc_iversion(inode);
1584 	inode->i_mtime = inode->i_ctime = current_time(inode);
1585 	mark_inode_dirty(inode);
1586 	return len - towrite;
1587 }
1588 
1589 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
1590 			 const struct path *path)
1591 {
1592 	int err;
1593 	struct inode *inode;
1594 
1595 	err = dquot_quota_on(sb, type, format_id, path);
1596 	if (err)
1597 		return err;
1598 
1599 	inode = d_inode(path->dentry);
1600 	inode_lock(inode);
1601 	EXT2_I(inode)->i_flags |= EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL;
1602 	inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1603 			S_NOATIME | S_IMMUTABLE);
1604 	inode_unlock(inode);
1605 	mark_inode_dirty(inode);
1606 
1607 	return 0;
1608 }
1609 
1610 static int ext2_quota_off(struct super_block *sb, int type)
1611 {
1612 	struct inode *inode = sb_dqopt(sb)->files[type];
1613 	int err;
1614 
1615 	if (!inode || !igrab(inode))
1616 		goto out;
1617 
1618 	err = dquot_quota_off(sb, type);
1619 	if (err)
1620 		goto out_put;
1621 
1622 	inode_lock(inode);
1623 	EXT2_I(inode)->i_flags &= ~(EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL);
1624 	inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1625 	inode_unlock(inode);
1626 	mark_inode_dirty(inode);
1627 out_put:
1628 	iput(inode);
1629 	return err;
1630 out:
1631 	return dquot_quota_off(sb, type);
1632 }
1633 
1634 #endif
1635 
1636 static struct file_system_type ext2_fs_type = {
1637 	.owner		= THIS_MODULE,
1638 	.name		= "ext2",
1639 	.mount		= ext2_mount,
1640 	.kill_sb	= kill_block_super,
1641 	.fs_flags	= FS_REQUIRES_DEV,
1642 };
1643 MODULE_ALIAS_FS("ext2");
1644 
1645 static int __init init_ext2_fs(void)
1646 {
1647 	int err;
1648 
1649 	err = init_inodecache();
1650 	if (err)
1651 		return err;
1652         err = register_filesystem(&ext2_fs_type);
1653 	if (err)
1654 		goto out;
1655 	return 0;
1656 out:
1657 	destroy_inodecache();
1658 	return err;
1659 }
1660 
1661 static void __exit exit_ext2_fs(void)
1662 {
1663 	unregister_filesystem(&ext2_fs_type);
1664 	destroy_inodecache();
1665 }
1666 
1667 MODULE_AUTHOR("Remy Card and others");
1668 MODULE_DESCRIPTION("Second Extended Filesystem");
1669 MODULE_LICENSE("GPL");
1670 module_init(init_ext2_fs)
1671 module_exit(exit_ext2_fs)
1672