xref: /linux/fs/ext2/super.c (revision b1a54551dd9ed5ef1763b97b35a0999ca002b95c)
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 		brelse(sbi->s_group_desc[i]);
166 	kvfree(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, NULL);
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 = alloc_inode_sb(sb, 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_free_in_core_inode(struct inode *inode)
196 {
197 	kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
198 }
199 
200 static void init_once(void *foo)
201 {
202 	struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
203 
204 	rwlock_init(&ei->i_meta_lock);
205 #ifdef CONFIG_EXT2_FS_XATTR
206 	init_rwsem(&ei->xattr_sem);
207 #endif
208 	mutex_init(&ei->truncate_mutex);
209 	inode_init_once(&ei->vfs_inode);
210 }
211 
212 static int __init init_inodecache(void)
213 {
214 	ext2_inode_cachep = kmem_cache_create_usercopy("ext2_inode_cache",
215 				sizeof(struct ext2_inode_info), 0,
216 				(SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
217 					SLAB_ACCOUNT),
218 				offsetof(struct ext2_inode_info, i_data),
219 				sizeof_field(struct ext2_inode_info, i_data),
220 				init_once);
221 	if (ext2_inode_cachep == NULL)
222 		return -ENOMEM;
223 	return 0;
224 }
225 
226 static void destroy_inodecache(void)
227 {
228 	/*
229 	 * Make sure all delayed rcu free inodes are flushed before we
230 	 * destroy cache.
231 	 */
232 	rcu_barrier();
233 	kmem_cache_destroy(ext2_inode_cachep);
234 }
235 
236 static int ext2_show_options(struct seq_file *seq, struct dentry *root)
237 {
238 	struct super_block *sb = root->d_sb;
239 	struct ext2_sb_info *sbi = EXT2_SB(sb);
240 	struct ext2_super_block *es = sbi->s_es;
241 	unsigned long def_mount_opts;
242 
243 	spin_lock(&sbi->s_lock);
244 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
245 
246 	if (sbi->s_sb_block != 1)
247 		seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
248 	if (test_opt(sb, MINIX_DF))
249 		seq_puts(seq, ",minixdf");
250 	if (test_opt(sb, GRPID))
251 		seq_puts(seq, ",grpid");
252 	if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
253 		seq_puts(seq, ",nogrpid");
254 	if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT2_DEF_RESUID)) ||
255 	    le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
256 		seq_printf(seq, ",resuid=%u",
257 				from_kuid_munged(&init_user_ns, sbi->s_resuid));
258 	}
259 	if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT2_DEF_RESGID)) ||
260 	    le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
261 		seq_printf(seq, ",resgid=%u",
262 				from_kgid_munged(&init_user_ns, sbi->s_resgid));
263 	}
264 	if (test_opt(sb, ERRORS_RO)) {
265 		int def_errors = le16_to_cpu(es->s_errors);
266 
267 		if (def_errors == EXT2_ERRORS_PANIC ||
268 		    def_errors == EXT2_ERRORS_CONTINUE) {
269 			seq_puts(seq, ",errors=remount-ro");
270 		}
271 	}
272 	if (test_opt(sb, ERRORS_CONT))
273 		seq_puts(seq, ",errors=continue");
274 	if (test_opt(sb, ERRORS_PANIC))
275 		seq_puts(seq, ",errors=panic");
276 	if (test_opt(sb, NO_UID32))
277 		seq_puts(seq, ",nouid32");
278 	if (test_opt(sb, DEBUG))
279 		seq_puts(seq, ",debug");
280 	if (test_opt(sb, OLDALLOC))
281 		seq_puts(seq, ",oldalloc");
282 
283 #ifdef CONFIG_EXT2_FS_XATTR
284 	if (test_opt(sb, XATTR_USER))
285 		seq_puts(seq, ",user_xattr");
286 	if (!test_opt(sb, XATTR_USER) &&
287 	    (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
288 		seq_puts(seq, ",nouser_xattr");
289 	}
290 #endif
291 
292 #ifdef CONFIG_EXT2_FS_POSIX_ACL
293 	if (test_opt(sb, POSIX_ACL))
294 		seq_puts(seq, ",acl");
295 	if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
296 		seq_puts(seq, ",noacl");
297 #endif
298 
299 	if (test_opt(sb, USRQUOTA))
300 		seq_puts(seq, ",usrquota");
301 
302 	if (test_opt(sb, GRPQUOTA))
303 		seq_puts(seq, ",grpquota");
304 
305 	if (test_opt(sb, XIP))
306 		seq_puts(seq, ",xip");
307 
308 	if (test_opt(sb, DAX))
309 		seq_puts(seq, ",dax");
310 
311 	if (!test_opt(sb, RESERVATION))
312 		seq_puts(seq, ",noreservation");
313 
314 	spin_unlock(&sbi->s_lock);
315 	return 0;
316 }
317 
318 #ifdef CONFIG_QUOTA
319 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
320 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
321 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
322 			 const struct path *path);
323 static struct dquot **ext2_get_dquots(struct inode *inode)
324 {
325 	return EXT2_I(inode)->i_dquot;
326 }
327 
328 static const struct quotactl_ops ext2_quotactl_ops = {
329 	.quota_on	= ext2_quota_on,
330 	.quota_off	= ext2_quota_off,
331 	.quota_sync	= dquot_quota_sync,
332 	.get_state	= dquot_get_state,
333 	.set_info	= dquot_set_dqinfo,
334 	.get_dqblk	= dquot_get_dqblk,
335 	.set_dqblk	= dquot_set_dqblk,
336 	.get_nextdqblk	= dquot_get_next_dqblk,
337 };
338 #endif
339 
340 static const struct super_operations ext2_sops = {
341 	.alloc_inode	= ext2_alloc_inode,
342 	.free_inode	= ext2_free_in_core_inode,
343 	.write_inode	= ext2_write_inode,
344 	.evict_inode	= ext2_evict_inode,
345 	.put_super	= ext2_put_super,
346 	.sync_fs	= ext2_sync_fs,
347 	.freeze_fs	= ext2_freeze,
348 	.unfreeze_fs	= ext2_unfreeze,
349 	.statfs		= ext2_statfs,
350 	.remount_fs	= ext2_remount,
351 	.show_options	= ext2_show_options,
352 #ifdef CONFIG_QUOTA
353 	.quota_read	= ext2_quota_read,
354 	.quota_write	= ext2_quota_write,
355 	.get_dquots	= ext2_get_dquots,
356 #endif
357 };
358 
359 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
360 		u64 ino, u32 generation)
361 {
362 	struct inode *inode;
363 
364 	if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
365 		return ERR_PTR(-ESTALE);
366 	if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
367 		return ERR_PTR(-ESTALE);
368 
369 	/*
370 	 * ext2_iget isn't quite right if the inode is currently unallocated!
371 	 * However ext2_iget currently does appropriate checks to handle stale
372 	 * inodes so everything is OK.
373 	 */
374 	inode = ext2_iget(sb, ino);
375 	if (IS_ERR(inode))
376 		return ERR_CAST(inode);
377 	if (generation && inode->i_generation != generation) {
378 		/* we didn't find the right inode.. */
379 		iput(inode);
380 		return ERR_PTR(-ESTALE);
381 	}
382 	return inode;
383 }
384 
385 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
386 		int fh_len, int fh_type)
387 {
388 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
389 				    ext2_nfs_get_inode);
390 }
391 
392 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
393 		int fh_len, int fh_type)
394 {
395 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
396 				    ext2_nfs_get_inode);
397 }
398 
399 static const struct export_operations ext2_export_ops = {
400 	.encode_fh = generic_encode_ino32_fh,
401 	.fh_to_dentry = ext2_fh_to_dentry,
402 	.fh_to_parent = ext2_fh_to_parent,
403 	.get_parent = ext2_get_parent,
404 };
405 
406 static unsigned long get_sb_block(void **data)
407 {
408 	unsigned long 	sb_block;
409 	char 		*options = (char *) *data;
410 
411 	if (!options || strncmp(options, "sb=", 3) != 0)
412 		return 1;	/* Default location */
413 	options += 3;
414 	sb_block = simple_strtoul(options, &options, 0);
415 	if (*options && *options != ',') {
416 		printk("EXT2-fs: Invalid sb specification: %s\n",
417 		       (char *) *data);
418 		return 1;
419 	}
420 	if (*options == ',')
421 		options++;
422 	*data = (void *) options;
423 	return sb_block;
424 }
425 
426 enum {
427 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
428 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
429 	Opt_err_ro, Opt_nouid32, Opt_debug,
430 	Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
431 	Opt_acl, Opt_noacl, Opt_xip, Opt_dax, Opt_ignore, Opt_err, Opt_quota,
432 	Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
433 };
434 
435 static const match_table_t tokens = {
436 	{Opt_bsd_df, "bsddf"},
437 	{Opt_minix_df, "minixdf"},
438 	{Opt_grpid, "grpid"},
439 	{Opt_grpid, "bsdgroups"},
440 	{Opt_nogrpid, "nogrpid"},
441 	{Opt_nogrpid, "sysvgroups"},
442 	{Opt_resgid, "resgid=%u"},
443 	{Opt_resuid, "resuid=%u"},
444 	{Opt_sb, "sb=%u"},
445 	{Opt_err_cont, "errors=continue"},
446 	{Opt_err_panic, "errors=panic"},
447 	{Opt_err_ro, "errors=remount-ro"},
448 	{Opt_nouid32, "nouid32"},
449 	{Opt_debug, "debug"},
450 	{Opt_oldalloc, "oldalloc"},
451 	{Opt_orlov, "orlov"},
452 	{Opt_nobh, "nobh"},
453 	{Opt_user_xattr, "user_xattr"},
454 	{Opt_nouser_xattr, "nouser_xattr"},
455 	{Opt_acl, "acl"},
456 	{Opt_noacl, "noacl"},
457 	{Opt_xip, "xip"},
458 	{Opt_dax, "dax"},
459 	{Opt_grpquota, "grpquota"},
460 	{Opt_ignore, "noquota"},
461 	{Opt_quota, "quota"},
462 	{Opt_usrquota, "usrquota"},
463 	{Opt_reservation, "reservation"},
464 	{Opt_noreservation, "noreservation"},
465 	{Opt_err, NULL}
466 };
467 
468 static int parse_options(char *options, struct super_block *sb,
469 			 struct ext2_mount_options *opts)
470 {
471 	char *p;
472 	substring_t args[MAX_OPT_ARGS];
473 	int option;
474 	kuid_t uid;
475 	kgid_t gid;
476 
477 	if (!options)
478 		return 1;
479 
480 	while ((p = strsep (&options, ",")) != NULL) {
481 		int token;
482 		if (!*p)
483 			continue;
484 
485 		token = match_token(p, tokens, args);
486 		switch (token) {
487 		case Opt_bsd_df:
488 			clear_opt (opts->s_mount_opt, MINIX_DF);
489 			break;
490 		case Opt_minix_df:
491 			set_opt (opts->s_mount_opt, MINIX_DF);
492 			break;
493 		case Opt_grpid:
494 			set_opt (opts->s_mount_opt, GRPID);
495 			break;
496 		case Opt_nogrpid:
497 			clear_opt (opts->s_mount_opt, GRPID);
498 			break;
499 		case Opt_resuid:
500 			if (match_int(&args[0], &option))
501 				return 0;
502 			uid = make_kuid(current_user_ns(), option);
503 			if (!uid_valid(uid)) {
504 				ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
505 				return 0;
506 
507 			}
508 			opts->s_resuid = uid;
509 			break;
510 		case Opt_resgid:
511 			if (match_int(&args[0], &option))
512 				return 0;
513 			gid = make_kgid(current_user_ns(), option);
514 			if (!gid_valid(gid)) {
515 				ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
516 				return 0;
517 			}
518 			opts->s_resgid = gid;
519 			break;
520 		case Opt_sb:
521 			/* handled by get_sb_block() instead of here */
522 			/* *sb_block = match_int(&args[0]); */
523 			break;
524 		case Opt_err_panic:
525 			clear_opt (opts->s_mount_opt, ERRORS_CONT);
526 			clear_opt (opts->s_mount_opt, ERRORS_RO);
527 			set_opt (opts->s_mount_opt, ERRORS_PANIC);
528 			break;
529 		case Opt_err_ro:
530 			clear_opt (opts->s_mount_opt, ERRORS_CONT);
531 			clear_opt (opts->s_mount_opt, ERRORS_PANIC);
532 			set_opt (opts->s_mount_opt, ERRORS_RO);
533 			break;
534 		case Opt_err_cont:
535 			clear_opt (opts->s_mount_opt, ERRORS_RO);
536 			clear_opt (opts->s_mount_opt, ERRORS_PANIC);
537 			set_opt (opts->s_mount_opt, ERRORS_CONT);
538 			break;
539 		case Opt_nouid32:
540 			set_opt (opts->s_mount_opt, NO_UID32);
541 			break;
542 		case Opt_debug:
543 			set_opt (opts->s_mount_opt, DEBUG);
544 			break;
545 		case Opt_oldalloc:
546 			set_opt (opts->s_mount_opt, OLDALLOC);
547 			break;
548 		case Opt_orlov:
549 			clear_opt (opts->s_mount_opt, OLDALLOC);
550 			break;
551 		case Opt_nobh:
552 			ext2_msg(sb, KERN_INFO,
553 				"nobh option not supported");
554 			break;
555 #ifdef CONFIG_EXT2_FS_XATTR
556 		case Opt_user_xattr:
557 			set_opt (opts->s_mount_opt, XATTR_USER);
558 			break;
559 		case Opt_nouser_xattr:
560 			clear_opt (opts->s_mount_opt, XATTR_USER);
561 			break;
562 #else
563 		case Opt_user_xattr:
564 		case Opt_nouser_xattr:
565 			ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
566 				"not supported");
567 			break;
568 #endif
569 #ifdef CONFIG_EXT2_FS_POSIX_ACL
570 		case Opt_acl:
571 			set_opt(opts->s_mount_opt, POSIX_ACL);
572 			break;
573 		case Opt_noacl:
574 			clear_opt(opts->s_mount_opt, POSIX_ACL);
575 			break;
576 #else
577 		case Opt_acl:
578 		case Opt_noacl:
579 			ext2_msg(sb, KERN_INFO,
580 				"(no)acl options not supported");
581 			break;
582 #endif
583 		case Opt_xip:
584 			ext2_msg(sb, KERN_INFO, "use dax instead of xip");
585 			set_opt(opts->s_mount_opt, XIP);
586 			fallthrough;
587 		case Opt_dax:
588 #ifdef CONFIG_FS_DAX
589 			ext2_msg(sb, KERN_WARNING,
590 		"DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
591 			set_opt(opts->s_mount_opt, DAX);
592 #else
593 			ext2_msg(sb, KERN_INFO, "dax option not supported");
594 #endif
595 			break;
596 
597 #if defined(CONFIG_QUOTA)
598 		case Opt_quota:
599 		case Opt_usrquota:
600 			set_opt(opts->s_mount_opt, USRQUOTA);
601 			break;
602 
603 		case Opt_grpquota:
604 			set_opt(opts->s_mount_opt, GRPQUOTA);
605 			break;
606 #else
607 		case Opt_quota:
608 		case Opt_usrquota:
609 		case Opt_grpquota:
610 			ext2_msg(sb, KERN_INFO,
611 				"quota operations not supported");
612 			break;
613 #endif
614 
615 		case Opt_reservation:
616 			set_opt(opts->s_mount_opt, RESERVATION);
617 			ext2_msg(sb, KERN_INFO, "reservations ON");
618 			break;
619 		case Opt_noreservation:
620 			clear_opt(opts->s_mount_opt, RESERVATION);
621 			ext2_msg(sb, KERN_INFO, "reservations OFF");
622 			break;
623 		case Opt_ignore:
624 			break;
625 		default:
626 			return 0;
627 		}
628 	}
629 	return 1;
630 }
631 
632 static int ext2_setup_super (struct super_block * sb,
633 			      struct ext2_super_block * es,
634 			      int read_only)
635 {
636 	int res = 0;
637 	struct ext2_sb_info *sbi = EXT2_SB(sb);
638 
639 	if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
640 		ext2_msg(sb, KERN_ERR,
641 			"error: revision level too high, "
642 			"forcing read-only mode");
643 		res = SB_RDONLY;
644 	}
645 	if (read_only)
646 		return res;
647 	if (!(sbi->s_mount_state & EXT2_VALID_FS))
648 		ext2_msg(sb, KERN_WARNING,
649 			"warning: mounting unchecked fs, "
650 			"running e2fsck is recommended");
651 	else if ((sbi->s_mount_state & EXT2_ERROR_FS))
652 		ext2_msg(sb, KERN_WARNING,
653 			"warning: mounting fs with errors, "
654 			"running e2fsck is recommended");
655 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
656 		 le16_to_cpu(es->s_mnt_count) >=
657 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
658 		ext2_msg(sb, KERN_WARNING,
659 			"warning: maximal mount count reached, "
660 			"running e2fsck is recommended");
661 	else if (le32_to_cpu(es->s_checkinterval) &&
662 		(le32_to_cpu(es->s_lastcheck) +
663 			le32_to_cpu(es->s_checkinterval) <=
664 			ktime_get_real_seconds()))
665 		ext2_msg(sb, KERN_WARNING,
666 			"warning: checktime reached, "
667 			"running e2fsck is recommended");
668 	if (!le16_to_cpu(es->s_max_mnt_count))
669 		es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
670 	le16_add_cpu(&es->s_mnt_count, 1);
671 	if (test_opt (sb, DEBUG))
672 		ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, gc=%lu, "
673 			"bpg=%lu, ipg=%lu, mo=%04lx]",
674 			EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
675 			sbi->s_groups_count,
676 			EXT2_BLOCKS_PER_GROUP(sb),
677 			EXT2_INODES_PER_GROUP(sb),
678 			sbi->s_mount_opt);
679 	return res;
680 }
681 
682 static int ext2_check_descriptors(struct super_block *sb)
683 {
684 	int i;
685 	struct ext2_sb_info *sbi = EXT2_SB(sb);
686 
687 	ext2_debug ("Checking group descriptors");
688 
689 	for (i = 0; i < sbi->s_groups_count; i++) {
690 		struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
691 		ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
692 		ext2_fsblk_t last_block = ext2_group_last_block_no(sb, i);
693 
694 		if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
695 		    le32_to_cpu(gdp->bg_block_bitmap) > last_block)
696 		{
697 			ext2_error (sb, "ext2_check_descriptors",
698 				    "Block bitmap for group %d"
699 				    " not in group (block %lu)!",
700 				    i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
701 			return 0;
702 		}
703 		if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
704 		    le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
705 		{
706 			ext2_error (sb, "ext2_check_descriptors",
707 				    "Inode bitmap for group %d"
708 				    " not in group (block %lu)!",
709 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
710 			return 0;
711 		}
712 		if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
713 		    le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
714 		    last_block)
715 		{
716 			ext2_error (sb, "ext2_check_descriptors",
717 				    "Inode table for group %d"
718 				    " not in group (block %lu)!",
719 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
720 			return 0;
721 		}
722 	}
723 	return 1;
724 }
725 
726 /*
727  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
728  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
729  * We need to be 1 filesystem block less than the 2^32 sector limit.
730  */
731 static loff_t ext2_max_size(int bits)
732 {
733 	loff_t res = EXT2_NDIR_BLOCKS;
734 	int meta_blocks;
735 	unsigned int upper_limit;
736 	unsigned int ppb = 1 << (bits-2);
737 
738 	/* This is calculated to be the largest file size for a
739 	 * dense, file such that the total number of
740 	 * sectors in the file, including data and all indirect blocks,
741 	 * does not exceed 2^32 -1
742 	 * __u32 i_blocks representing the total number of
743 	 * 512 bytes blocks of the file
744 	 */
745 	upper_limit = (1LL << 32) - 1;
746 
747 	/* total blocks in file system block size */
748 	upper_limit >>= (bits - 9);
749 
750 	/* Compute how many blocks we can address by block tree */
751 	res += 1LL << (bits-2);
752 	res += 1LL << (2*(bits-2));
753 	res += 1LL << (3*(bits-2));
754 	/* Compute how many metadata blocks are needed */
755 	meta_blocks = 1;
756 	meta_blocks += 1 + ppb;
757 	meta_blocks += 1 + ppb + ppb * ppb;
758 	/* Does block tree limit file size? */
759 	if (res + meta_blocks <= upper_limit)
760 		goto check_lfs;
761 
762 	res = upper_limit;
763 	/* How many metadata blocks are needed for addressing upper_limit? */
764 	upper_limit -= EXT2_NDIR_BLOCKS;
765 	/* indirect blocks */
766 	meta_blocks = 1;
767 	upper_limit -= ppb;
768 	/* double indirect blocks */
769 	if (upper_limit < ppb * ppb) {
770 		meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb);
771 		res -= meta_blocks;
772 		goto check_lfs;
773 	}
774 	meta_blocks += 1 + ppb;
775 	upper_limit -= ppb * ppb;
776 	/* tripple indirect blocks for the rest */
777 	meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb) +
778 		DIV_ROUND_UP(upper_limit, ppb*ppb);
779 	res -= meta_blocks;
780 check_lfs:
781 	res <<= bits;
782 	if (res > MAX_LFS_FILESIZE)
783 		res = MAX_LFS_FILESIZE;
784 
785 	return res;
786 }
787 
788 static unsigned long descriptor_loc(struct super_block *sb,
789 				    unsigned long logic_sb_block,
790 				    int nr)
791 {
792 	struct ext2_sb_info *sbi = EXT2_SB(sb);
793 	unsigned long bg, first_meta_bg;
794 
795 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
796 
797 	if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
798 	    nr < first_meta_bg)
799 		return (logic_sb_block + nr + 1);
800 	bg = sbi->s_desc_per_block * nr;
801 
802 	return ext2_group_first_block_no(sb, bg) + ext2_bg_has_super(sb, bg);
803 }
804 
805 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
806 {
807 	struct buffer_head * bh;
808 	struct ext2_sb_info * sbi;
809 	struct ext2_super_block * es;
810 	struct inode *root;
811 	unsigned long block;
812 	unsigned long sb_block = get_sb_block(&data);
813 	unsigned long logic_sb_block;
814 	unsigned long offset = 0;
815 	unsigned long def_mount_opts;
816 	long ret = -ENOMEM;
817 	int blocksize = BLOCK_SIZE;
818 	int db_count;
819 	int i, j;
820 	__le32 features;
821 	int err;
822 	struct ext2_mount_options opts;
823 
824 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
825 	if (!sbi)
826 		return -ENOMEM;
827 
828 	sbi->s_blockgroup_lock =
829 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
830 	if (!sbi->s_blockgroup_lock) {
831 		kfree(sbi);
832 		return -ENOMEM;
833 	}
834 	sb->s_fs_info = sbi;
835 	sbi->s_sb_block = sb_block;
836 	sbi->s_daxdev = fs_dax_get_by_bdev(sb->s_bdev, &sbi->s_dax_part_off,
837 					   NULL, NULL);
838 
839 	spin_lock_init(&sbi->s_lock);
840 	ret = -EINVAL;
841 
842 	/*
843 	 * See what the current blocksize for the device is, and
844 	 * use that as the blocksize.  Otherwise (or if the blocksize
845 	 * is smaller than the default) use the default.
846 	 * This is important for devices that have a hardware
847 	 * sectorsize that is larger than the default.
848 	 */
849 	blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
850 	if (!blocksize) {
851 		ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
852 		goto failed_sbi;
853 	}
854 
855 	/*
856 	 * If the superblock doesn't start on a hardware sector boundary,
857 	 * calculate the offset.
858 	 */
859 	if (blocksize != BLOCK_SIZE) {
860 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
861 		offset = (sb_block*BLOCK_SIZE) % blocksize;
862 	} else {
863 		logic_sb_block = sb_block;
864 	}
865 
866 	if (!(bh = sb_bread(sb, logic_sb_block))) {
867 		ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
868 		goto failed_sbi;
869 	}
870 	/*
871 	 * Note: s_es must be initialized as soon as possible because
872 	 *       some ext2 macro-instructions depend on its value
873 	 */
874 	es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
875 	sbi->s_es = es;
876 	sb->s_magic = le16_to_cpu(es->s_magic);
877 
878 	if (sb->s_magic != EXT2_SUPER_MAGIC)
879 		goto cantfind_ext2;
880 
881 	opts.s_mount_opt = 0;
882 	/* Set defaults before we parse the mount options */
883 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
884 	if (def_mount_opts & EXT2_DEFM_DEBUG)
885 		set_opt(opts.s_mount_opt, DEBUG);
886 	if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
887 		set_opt(opts.s_mount_opt, GRPID);
888 	if (def_mount_opts & EXT2_DEFM_UID16)
889 		set_opt(opts.s_mount_opt, NO_UID32);
890 #ifdef CONFIG_EXT2_FS_XATTR
891 	if (def_mount_opts & EXT2_DEFM_XATTR_USER)
892 		set_opt(opts.s_mount_opt, XATTR_USER);
893 #endif
894 #ifdef CONFIG_EXT2_FS_POSIX_ACL
895 	if (def_mount_opts & EXT2_DEFM_ACL)
896 		set_opt(opts.s_mount_opt, POSIX_ACL);
897 #endif
898 
899 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
900 		set_opt(opts.s_mount_opt, ERRORS_PANIC);
901 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
902 		set_opt(opts.s_mount_opt, ERRORS_CONT);
903 	else
904 		set_opt(opts.s_mount_opt, ERRORS_RO);
905 
906 	opts.s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
907 	opts.s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
908 
909 	set_opt(opts.s_mount_opt, RESERVATION);
910 
911 	if (!parse_options((char *) data, sb, &opts))
912 		goto failed_mount;
913 
914 	sbi->s_mount_opt = opts.s_mount_opt;
915 	sbi->s_resuid = opts.s_resuid;
916 	sbi->s_resgid = opts.s_resgid;
917 
918 	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
919 		(test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
920 	sb->s_iflags |= SB_I_CGROUPWB;
921 
922 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
923 	    (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
924 	     EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
925 	     EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
926 		ext2_msg(sb, KERN_WARNING,
927 			"warning: feature flags set on rev 0 fs, "
928 			"running e2fsck is recommended");
929 	/*
930 	 * Check feature flags regardless of the revision level, since we
931 	 * previously didn't change the revision level when setting the flags,
932 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
933 	 */
934 	features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
935 	if (features) {
936 		ext2_msg(sb, KERN_ERR,	"error: couldn't mount because of "
937 		       "unsupported optional features (%x)",
938 			le32_to_cpu(features));
939 		goto failed_mount;
940 	}
941 	if (!sb_rdonly(sb) && (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
942 		ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
943 		       "unsupported optional features (%x)",
944 		       le32_to_cpu(features));
945 		goto failed_mount;
946 	}
947 
948 	if (le32_to_cpu(es->s_log_block_size) >
949 	    (EXT2_MAX_BLOCK_LOG_SIZE - BLOCK_SIZE_BITS)) {
950 		ext2_msg(sb, KERN_ERR,
951 			 "Invalid log block size: %u",
952 			 le32_to_cpu(es->s_log_block_size));
953 		goto failed_mount;
954 	}
955 	blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
956 
957 	if (test_opt(sb, DAX)) {
958 		if (!sbi->s_daxdev) {
959 			ext2_msg(sb, KERN_ERR,
960 				"DAX unsupported by block device. Turning off DAX.");
961 			clear_opt(sbi->s_mount_opt, DAX);
962 		} else if (blocksize != PAGE_SIZE) {
963 			ext2_msg(sb, KERN_ERR, "unsupported blocksize for DAX\n");
964 			clear_opt(sbi->s_mount_opt, DAX);
965 		}
966 	}
967 
968 	/* If the blocksize doesn't match, re-read the thing.. */
969 	if (sb->s_blocksize != blocksize) {
970 		brelse(bh);
971 
972 		if (!sb_set_blocksize(sb, blocksize)) {
973 			ext2_msg(sb, KERN_ERR,
974 				"error: bad blocksize %d", blocksize);
975 			goto failed_sbi;
976 		}
977 
978 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
979 		offset = (sb_block*BLOCK_SIZE) % blocksize;
980 		bh = sb_bread(sb, logic_sb_block);
981 		if(!bh) {
982 			ext2_msg(sb, KERN_ERR, "error: couldn't read"
983 				"superblock on 2nd try");
984 			goto failed_sbi;
985 		}
986 		es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
987 		sbi->s_es = es;
988 		if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
989 			ext2_msg(sb, KERN_ERR, "error: magic mismatch");
990 			goto failed_mount;
991 		}
992 	}
993 
994 	sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
995 	sb->s_max_links = EXT2_LINK_MAX;
996 	sb->s_time_min = S32_MIN;
997 	sb->s_time_max = S32_MAX;
998 
999 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
1000 		sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
1001 		sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
1002 	} else {
1003 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1004 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1005 		if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
1006 		    !is_power_of_2(sbi->s_inode_size) ||
1007 		    (sbi->s_inode_size > blocksize)) {
1008 			ext2_msg(sb, KERN_ERR,
1009 				"error: unsupported inode size: %d",
1010 				sbi->s_inode_size);
1011 			goto failed_mount;
1012 		}
1013 	}
1014 
1015 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1016 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1017 
1018 	sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
1019 	if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
1020 		goto cantfind_ext2;
1021 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
1022 					sbi->s_inodes_per_block;
1023 	sbi->s_desc_per_block = sb->s_blocksize /
1024 					sizeof (struct ext2_group_desc);
1025 	sbi->s_sbh = bh;
1026 	sbi->s_mount_state = le16_to_cpu(es->s_state);
1027 	sbi->s_addr_per_block_bits =
1028 		ilog2 (EXT2_ADDR_PER_BLOCK(sb));
1029 	sbi->s_desc_per_block_bits =
1030 		ilog2 (EXT2_DESC_PER_BLOCK(sb));
1031 
1032 	if (sb->s_magic != EXT2_SUPER_MAGIC)
1033 		goto cantfind_ext2;
1034 
1035 	if (sb->s_blocksize != bh->b_size) {
1036 		if (!silent)
1037 			ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
1038 		goto failed_mount;
1039 	}
1040 
1041 	if (es->s_log_frag_size != es->s_log_block_size) {
1042 		ext2_msg(sb, KERN_ERR,
1043 			"error: fragsize log %u != blocksize log %u",
1044 			le32_to_cpu(es->s_log_frag_size), sb->s_blocksize_bits);
1045 		goto failed_mount;
1046 	}
1047 
1048 	if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
1049 		ext2_msg(sb, KERN_ERR,
1050 			"error: #blocks per group too big: %lu",
1051 			sbi->s_blocks_per_group);
1052 		goto failed_mount;
1053 	}
1054 	/* At least inode table, bitmaps, and sb have to fit in one group */
1055 	if (sbi->s_blocks_per_group <= sbi->s_itb_per_group + 3) {
1056 		ext2_msg(sb, KERN_ERR,
1057 			"error: #blocks per group smaller than metadata size: %lu <= %lu",
1058 			sbi->s_blocks_per_group, sbi->s_inodes_per_group + 3);
1059 		goto failed_mount;
1060 	}
1061 	if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
1062 	    sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1063 		ext2_msg(sb, KERN_ERR,
1064 			"error: invalid #inodes per group: %lu",
1065 			sbi->s_inodes_per_group);
1066 		goto failed_mount;
1067 	}
1068 	if (sb_bdev_nr_blocks(sb) < le32_to_cpu(es->s_blocks_count)) {
1069 		ext2_msg(sb, KERN_ERR,
1070 			 "bad geometry: block count %u exceeds size of device (%u blocks)",
1071 			 le32_to_cpu(es->s_blocks_count),
1072 			 (unsigned)sb_bdev_nr_blocks(sb));
1073 		goto failed_mount;
1074 	}
1075 
1076 	sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1077 				le32_to_cpu(es->s_first_data_block) - 1)
1078 					/ EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1079 	if ((u64)sbi->s_groups_count * sbi->s_inodes_per_group !=
1080 	    le32_to_cpu(es->s_inodes_count)) {
1081 		ext2_msg(sb, KERN_ERR, "error: invalid #inodes: %u vs computed %llu",
1082 			 le32_to_cpu(es->s_inodes_count),
1083 			 (u64)sbi->s_groups_count * sbi->s_inodes_per_group);
1084 		goto failed_mount;
1085 	}
1086 	db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1087 		   EXT2_DESC_PER_BLOCK(sb);
1088 	sbi->s_group_desc = kvmalloc_array(db_count,
1089 					   sizeof(struct buffer_head *),
1090 					   GFP_KERNEL);
1091 	if (sbi->s_group_desc == NULL) {
1092 		ret = -ENOMEM;
1093 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1094 		goto failed_mount;
1095 	}
1096 	bgl_lock_init(sbi->s_blockgroup_lock);
1097 	sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1098 	if (!sbi->s_debts) {
1099 		ret = -ENOMEM;
1100 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1101 		goto failed_mount_group_desc;
1102 	}
1103 	for (i = 0; i < db_count; i++) {
1104 		block = descriptor_loc(sb, logic_sb_block, i);
1105 		sbi->s_group_desc[i] = sb_bread(sb, block);
1106 		if (!sbi->s_group_desc[i]) {
1107 			for (j = 0; j < i; j++)
1108 				brelse (sbi->s_group_desc[j]);
1109 			ext2_msg(sb, KERN_ERR,
1110 				"error: unable to read group descriptors");
1111 			goto failed_mount_group_desc;
1112 		}
1113 	}
1114 	if (!ext2_check_descriptors (sb)) {
1115 		ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1116 		goto failed_mount2;
1117 	}
1118 	sbi->s_gdb_count = db_count;
1119 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1120 	spin_lock_init(&sbi->s_next_gen_lock);
1121 
1122 	/* per filesystem reservation list head & lock */
1123 	spin_lock_init(&sbi->s_rsv_window_lock);
1124 	sbi->s_rsv_window_root = RB_ROOT;
1125 	/*
1126 	 * Add a single, static dummy reservation to the start of the
1127 	 * reservation window list --- it gives us a placeholder for
1128 	 * append-at-start-of-list which makes the allocation logic
1129 	 * _much_ simpler.
1130 	 */
1131 	sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1132 	sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1133 	sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1134 	sbi->s_rsv_window_head.rsv_goal_size = 0;
1135 	ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1136 
1137 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
1138 				ext2_count_free_blocks(sb), GFP_KERNEL);
1139 	if (!err) {
1140 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
1141 				ext2_count_free_inodes(sb), GFP_KERNEL);
1142 	}
1143 	if (!err) {
1144 		err = percpu_counter_init(&sbi->s_dirs_counter,
1145 				ext2_count_dirs(sb), GFP_KERNEL);
1146 	}
1147 	if (err) {
1148 		ret = err;
1149 		ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1150 		goto failed_mount3;
1151 	}
1152 
1153 #ifdef CONFIG_EXT2_FS_XATTR
1154 	sbi->s_ea_block_cache = ext2_xattr_create_cache();
1155 	if (!sbi->s_ea_block_cache) {
1156 		ret = -ENOMEM;
1157 		ext2_msg(sb, KERN_ERR, "Failed to create ea_block_cache");
1158 		goto failed_mount3;
1159 	}
1160 #endif
1161 	/*
1162 	 * set up enough so that it can read an inode
1163 	 */
1164 	sb->s_op = &ext2_sops;
1165 	sb->s_export_op = &ext2_export_ops;
1166 	sb->s_xattr = ext2_xattr_handlers;
1167 
1168 #ifdef CONFIG_QUOTA
1169 	sb->dq_op = &dquot_operations;
1170 	sb->s_qcop = &ext2_quotactl_ops;
1171 	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1172 #endif
1173 
1174 	root = ext2_iget(sb, EXT2_ROOT_INO);
1175 	if (IS_ERR(root)) {
1176 		ret = PTR_ERR(root);
1177 		goto failed_mount3;
1178 	}
1179 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1180 		iput(root);
1181 		ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1182 		goto failed_mount3;
1183 	}
1184 
1185 	sb->s_root = d_make_root(root);
1186 	if (!sb->s_root) {
1187 		ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1188 		ret = -ENOMEM;
1189 		goto failed_mount3;
1190 	}
1191 	if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1192 		ext2_msg(sb, KERN_WARNING,
1193 			"warning: mounting ext3 filesystem as ext2");
1194 	if (ext2_setup_super (sb, es, sb_rdonly(sb)))
1195 		sb->s_flags |= SB_RDONLY;
1196 	ext2_write_super(sb);
1197 	return 0;
1198 
1199 cantfind_ext2:
1200 	if (!silent)
1201 		ext2_msg(sb, KERN_ERR,
1202 			"error: can't find an ext2 filesystem on dev %s.",
1203 			sb->s_id);
1204 	goto failed_mount;
1205 failed_mount3:
1206 	ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
1207 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
1208 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
1209 	percpu_counter_destroy(&sbi->s_dirs_counter);
1210 failed_mount2:
1211 	for (i = 0; i < db_count; i++)
1212 		brelse(sbi->s_group_desc[i]);
1213 failed_mount_group_desc:
1214 	kvfree(sbi->s_group_desc);
1215 	kfree(sbi->s_debts);
1216 failed_mount:
1217 	brelse(bh);
1218 failed_sbi:
1219 	fs_put_dax(sbi->s_daxdev, NULL);
1220 	sb->s_fs_info = NULL;
1221 	kfree(sbi->s_blockgroup_lock);
1222 	kfree(sbi);
1223 	return ret;
1224 }
1225 
1226 static void ext2_clear_super_error(struct super_block *sb)
1227 {
1228 	struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1229 
1230 	if (buffer_write_io_error(sbh)) {
1231 		/*
1232 		 * Oh, dear.  A previous attempt to write the
1233 		 * superblock failed.  This could happen because the
1234 		 * USB device was yanked out.  Or it could happen to
1235 		 * be a transient write error and maybe the block will
1236 		 * be remapped.  Nothing we can do but to retry the
1237 		 * write and hope for the best.
1238 		 */
1239 		ext2_msg(sb, KERN_ERR,
1240 		       "previous I/O error to superblock detected");
1241 		clear_buffer_write_io_error(sbh);
1242 		set_buffer_uptodate(sbh);
1243 	}
1244 }
1245 
1246 void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1247 		     int wait)
1248 {
1249 	ext2_clear_super_error(sb);
1250 	spin_lock(&EXT2_SB(sb)->s_lock);
1251 	es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1252 	es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1253 	es->s_wtime = cpu_to_le32(ktime_get_real_seconds());
1254 	/* unlock before we do IO */
1255 	spin_unlock(&EXT2_SB(sb)->s_lock);
1256 	mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1257 	if (wait)
1258 		sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1259 }
1260 
1261 /*
1262  * In the second extended file system, it is not necessary to
1263  * write the super block since we use a mapping of the
1264  * disk super block in a buffer.
1265  *
1266  * However, this function is still used to set the fs valid
1267  * flags to 0.  We need to set this flag to 0 since the fs
1268  * may have been checked while mounted and e2fsck may have
1269  * set s_state to EXT2_VALID_FS after some corrections.
1270  */
1271 static int ext2_sync_fs(struct super_block *sb, int wait)
1272 {
1273 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1274 	struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1275 
1276 	/*
1277 	 * Write quota structures to quota file, sync_blockdev() will write
1278 	 * them to disk later
1279 	 */
1280 	dquot_writeback_dquots(sb, -1);
1281 
1282 	spin_lock(&sbi->s_lock);
1283 	if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1284 		ext2_debug("setting valid to 0\n");
1285 		es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1286 	}
1287 	spin_unlock(&sbi->s_lock);
1288 	ext2_sync_super(sb, es, wait);
1289 	return 0;
1290 }
1291 
1292 static int ext2_freeze(struct super_block *sb)
1293 {
1294 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1295 
1296 	/*
1297 	 * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1298 	 * because we have unattached inodes and thus filesystem is not fully
1299 	 * consistent.
1300 	 */
1301 	if (atomic_long_read(&sb->s_remove_count)) {
1302 		ext2_sync_fs(sb, 1);
1303 		return 0;
1304 	}
1305 	/* Set EXT2_FS_VALID flag */
1306 	spin_lock(&sbi->s_lock);
1307 	sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1308 	spin_unlock(&sbi->s_lock);
1309 	ext2_sync_super(sb, sbi->s_es, 1);
1310 
1311 	return 0;
1312 }
1313 
1314 static int ext2_unfreeze(struct super_block *sb)
1315 {
1316 	/* Just write sb to clear EXT2_VALID_FS flag */
1317 	ext2_write_super(sb);
1318 
1319 	return 0;
1320 }
1321 
1322 static void ext2_write_super(struct super_block *sb)
1323 {
1324 	if (!sb_rdonly(sb))
1325 		ext2_sync_fs(sb, 1);
1326 }
1327 
1328 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1329 {
1330 	struct ext2_sb_info * sbi = EXT2_SB(sb);
1331 	struct ext2_super_block * es;
1332 	struct ext2_mount_options new_opts;
1333 	int err;
1334 
1335 	sync_filesystem(sb);
1336 
1337 	spin_lock(&sbi->s_lock);
1338 	new_opts.s_mount_opt = sbi->s_mount_opt;
1339 	new_opts.s_resuid = sbi->s_resuid;
1340 	new_opts.s_resgid = sbi->s_resgid;
1341 	spin_unlock(&sbi->s_lock);
1342 
1343 	if (!parse_options(data, sb, &new_opts))
1344 		return -EINVAL;
1345 
1346 	spin_lock(&sbi->s_lock);
1347 	es = sbi->s_es;
1348 	if ((sbi->s_mount_opt ^ new_opts.s_mount_opt) & EXT2_MOUNT_DAX) {
1349 		ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1350 			 "dax flag with busy inodes while remounting");
1351 		new_opts.s_mount_opt ^= EXT2_MOUNT_DAX;
1352 	}
1353 	if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1354 		goto out_set;
1355 	if (*flags & SB_RDONLY) {
1356 		if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1357 		    !(sbi->s_mount_state & EXT2_VALID_FS))
1358 			goto out_set;
1359 
1360 		/*
1361 		 * OK, we are remounting a valid rw partition rdonly, so set
1362 		 * the rdonly flag and then mark the partition as valid again.
1363 		 */
1364 		es->s_state = cpu_to_le16(sbi->s_mount_state);
1365 		es->s_mtime = cpu_to_le32(ktime_get_real_seconds());
1366 		spin_unlock(&sbi->s_lock);
1367 
1368 		err = dquot_suspend(sb, -1);
1369 		if (err < 0)
1370 			return err;
1371 
1372 		ext2_sync_super(sb, es, 1);
1373 	} else {
1374 		__le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1375 					       ~EXT2_FEATURE_RO_COMPAT_SUPP);
1376 		if (ret) {
1377 			spin_unlock(&sbi->s_lock);
1378 			ext2_msg(sb, KERN_WARNING,
1379 				"warning: couldn't remount RDWR because of "
1380 				"unsupported optional features (%x).",
1381 				le32_to_cpu(ret));
1382 			return -EROFS;
1383 		}
1384 		/*
1385 		 * Mounting a RDONLY partition read-write, so reread and
1386 		 * store the current valid flag.  (It may have been changed
1387 		 * by e2fsck since we originally mounted the partition.)
1388 		 */
1389 		sbi->s_mount_state = le16_to_cpu(es->s_state);
1390 		if (!ext2_setup_super (sb, es, 0))
1391 			sb->s_flags &= ~SB_RDONLY;
1392 		spin_unlock(&sbi->s_lock);
1393 
1394 		ext2_write_super(sb);
1395 
1396 		dquot_resume(sb, -1);
1397 	}
1398 
1399 	spin_lock(&sbi->s_lock);
1400 out_set:
1401 	sbi->s_mount_opt = new_opts.s_mount_opt;
1402 	sbi->s_resuid = new_opts.s_resuid;
1403 	sbi->s_resgid = new_opts.s_resgid;
1404 	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1405 		(test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
1406 	spin_unlock(&sbi->s_lock);
1407 
1408 	return 0;
1409 }
1410 
1411 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1412 {
1413 	struct super_block *sb = dentry->d_sb;
1414 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1415 	struct ext2_super_block *es = sbi->s_es;
1416 
1417 	spin_lock(&sbi->s_lock);
1418 
1419 	if (test_opt (sb, MINIX_DF))
1420 		sbi->s_overhead_last = 0;
1421 	else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1422 		unsigned long i, overhead = 0;
1423 		smp_rmb();
1424 
1425 		/*
1426 		 * Compute the overhead (FS structures). This is constant
1427 		 * for a given filesystem unless the number of block groups
1428 		 * changes so we cache the previous value until it does.
1429 		 */
1430 
1431 		/*
1432 		 * All of the blocks before first_data_block are
1433 		 * overhead
1434 		 */
1435 		overhead = le32_to_cpu(es->s_first_data_block);
1436 
1437 		/*
1438 		 * Add the overhead attributed to the superblock and
1439 		 * block group descriptors.  If the sparse superblocks
1440 		 * feature is turned on, then not all groups have this.
1441 		 */
1442 		for (i = 0; i < sbi->s_groups_count; i++)
1443 			overhead += ext2_bg_has_super(sb, i) +
1444 				ext2_bg_num_gdb(sb, i);
1445 
1446 		/*
1447 		 * Every block group has an inode bitmap, a block
1448 		 * bitmap, and an inode table.
1449 		 */
1450 		overhead += (sbi->s_groups_count *
1451 			     (2 + sbi->s_itb_per_group));
1452 		sbi->s_overhead_last = overhead;
1453 		smp_wmb();
1454 		sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1455 	}
1456 
1457 	buf->f_type = EXT2_SUPER_MAGIC;
1458 	buf->f_bsize = sb->s_blocksize;
1459 	buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1460 	buf->f_bfree = ext2_count_free_blocks(sb);
1461 	es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1462 	buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1463 	if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1464 		buf->f_bavail = 0;
1465 	buf->f_files = le32_to_cpu(es->s_inodes_count);
1466 	buf->f_ffree = ext2_count_free_inodes(sb);
1467 	es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1468 	buf->f_namelen = EXT2_NAME_LEN;
1469 	buf->f_fsid = uuid_to_fsid(es->s_uuid);
1470 	spin_unlock(&sbi->s_lock);
1471 	return 0;
1472 }
1473 
1474 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1475 	int flags, const char *dev_name, void *data)
1476 {
1477 	return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1478 }
1479 
1480 #ifdef CONFIG_QUOTA
1481 
1482 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1483  * acquiring the locks... As quota files are never truncated and quota code
1484  * itself serializes the operations (and no one else should touch the files)
1485  * we don't have to be afraid of races */
1486 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1487 			       size_t len, loff_t off)
1488 {
1489 	struct inode *inode = sb_dqopt(sb)->files[type];
1490 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1491 	int err = 0;
1492 	int offset = off & (sb->s_blocksize - 1);
1493 	int tocopy;
1494 	size_t toread;
1495 	struct buffer_head tmp_bh;
1496 	struct buffer_head *bh;
1497 	loff_t i_size = i_size_read(inode);
1498 
1499 	if (off > i_size)
1500 		return 0;
1501 	if (off+len > i_size)
1502 		len = i_size-off;
1503 	toread = len;
1504 	while (toread > 0) {
1505 		tocopy = min_t(size_t, sb->s_blocksize - offset, toread);
1506 
1507 		tmp_bh.b_state = 0;
1508 		tmp_bh.b_size = sb->s_blocksize;
1509 		err = ext2_get_block(inode, blk, &tmp_bh, 0);
1510 		if (err < 0)
1511 			return err;
1512 		if (!buffer_mapped(&tmp_bh))	/* A hole? */
1513 			memset(data, 0, tocopy);
1514 		else {
1515 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1516 			if (!bh)
1517 				return -EIO;
1518 			memcpy(data, bh->b_data+offset, tocopy);
1519 			brelse(bh);
1520 		}
1521 		offset = 0;
1522 		toread -= tocopy;
1523 		data += tocopy;
1524 		blk++;
1525 	}
1526 	return len;
1527 }
1528 
1529 /* Write to quotafile */
1530 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1531 				const char *data, size_t len, loff_t off)
1532 {
1533 	struct inode *inode = sb_dqopt(sb)->files[type];
1534 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1535 	int err = 0;
1536 	int offset = off & (sb->s_blocksize - 1);
1537 	int tocopy;
1538 	size_t towrite = len;
1539 	struct buffer_head tmp_bh;
1540 	struct buffer_head *bh;
1541 
1542 	while (towrite > 0) {
1543 		tocopy = min_t(size_t, sb->s_blocksize - offset, towrite);
1544 
1545 		tmp_bh.b_state = 0;
1546 		tmp_bh.b_size = sb->s_blocksize;
1547 		err = ext2_get_block(inode, blk, &tmp_bh, 1);
1548 		if (err < 0)
1549 			goto out;
1550 		if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1551 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1552 		else
1553 			bh = sb_getblk(sb, tmp_bh.b_blocknr);
1554 		if (unlikely(!bh)) {
1555 			err = -EIO;
1556 			goto out;
1557 		}
1558 		lock_buffer(bh);
1559 		memcpy(bh->b_data+offset, data, tocopy);
1560 		flush_dcache_page(bh->b_page);
1561 		set_buffer_uptodate(bh);
1562 		mark_buffer_dirty(bh);
1563 		unlock_buffer(bh);
1564 		brelse(bh);
1565 		offset = 0;
1566 		towrite -= tocopy;
1567 		data += tocopy;
1568 		blk++;
1569 	}
1570 out:
1571 	if (len == towrite)
1572 		return err;
1573 	if (inode->i_size < off+len-towrite)
1574 		i_size_write(inode, off+len-towrite);
1575 	inode_inc_iversion(inode);
1576 	inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
1577 	mark_inode_dirty(inode);
1578 	return len - towrite;
1579 }
1580 
1581 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
1582 			 const struct path *path)
1583 {
1584 	int err;
1585 	struct inode *inode;
1586 
1587 	err = dquot_quota_on(sb, type, format_id, path);
1588 	if (err)
1589 		return err;
1590 
1591 	inode = d_inode(path->dentry);
1592 	inode_lock(inode);
1593 	EXT2_I(inode)->i_flags |= EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL;
1594 	inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1595 			S_NOATIME | S_IMMUTABLE);
1596 	inode_unlock(inode);
1597 	mark_inode_dirty(inode);
1598 
1599 	return 0;
1600 }
1601 
1602 static int ext2_quota_off(struct super_block *sb, int type)
1603 {
1604 	struct inode *inode = sb_dqopt(sb)->files[type];
1605 	int err;
1606 
1607 	if (!inode || !igrab(inode))
1608 		goto out;
1609 
1610 	err = dquot_quota_off(sb, type);
1611 	if (err)
1612 		goto out_put;
1613 
1614 	inode_lock(inode);
1615 	EXT2_I(inode)->i_flags &= ~(EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL);
1616 	inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1617 	inode_unlock(inode);
1618 	mark_inode_dirty(inode);
1619 out_put:
1620 	iput(inode);
1621 	return err;
1622 out:
1623 	return dquot_quota_off(sb, type);
1624 }
1625 
1626 #endif
1627 
1628 static struct file_system_type ext2_fs_type = {
1629 	.owner		= THIS_MODULE,
1630 	.name		= "ext2",
1631 	.mount		= ext2_mount,
1632 	.kill_sb	= kill_block_super,
1633 	.fs_flags	= FS_REQUIRES_DEV,
1634 };
1635 MODULE_ALIAS_FS("ext2");
1636 
1637 static int __init init_ext2_fs(void)
1638 {
1639 	int err;
1640 
1641 	err = init_inodecache();
1642 	if (err)
1643 		return err;
1644 	err = register_filesystem(&ext2_fs_type);
1645 	if (err)
1646 		goto out;
1647 	return 0;
1648 out:
1649 	destroy_inodecache();
1650 	return err;
1651 }
1652 
1653 static void __exit exit_ext2_fs(void)
1654 {
1655 	unregister_filesystem(&ext2_fs_type);
1656 	destroy_inodecache();
1657 }
1658 
1659 MODULE_AUTHOR("Remy Card and others");
1660 MODULE_DESCRIPTION("Second Extended Filesystem");
1661 MODULE_LICENSE("GPL");
1662 module_init(init_ext2_fs)
1663 module_exit(exit_ext2_fs)
1664