xref: /linux/fs/ext2/super.c (revision 4f77c3462308c62ffe7129cc18b9ac937f44b5a5)
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_ACCOUNT,
217 				offsetof(struct ext2_inode_info, i_data),
218 				sizeof_field(struct ext2_inode_info, i_data),
219 				init_once);
220 	if (ext2_inode_cachep == NULL)
221 		return -ENOMEM;
222 	return 0;
223 }
224 
225 static void destroy_inodecache(void)
226 {
227 	/*
228 	 * Make sure all delayed rcu free inodes are flushed before we
229 	 * destroy cache.
230 	 */
231 	rcu_barrier();
232 	kmem_cache_destroy(ext2_inode_cachep);
233 }
234 
235 static int ext2_show_options(struct seq_file *seq, struct dentry *root)
236 {
237 	struct super_block *sb = root->d_sb;
238 	struct ext2_sb_info *sbi = EXT2_SB(sb);
239 	struct ext2_super_block *es = sbi->s_es;
240 	unsigned long def_mount_opts;
241 
242 	spin_lock(&sbi->s_lock);
243 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
244 
245 	if (sbi->s_sb_block != 1)
246 		seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
247 	if (test_opt(sb, MINIX_DF))
248 		seq_puts(seq, ",minixdf");
249 	if (test_opt(sb, GRPID))
250 		seq_puts(seq, ",grpid");
251 	if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
252 		seq_puts(seq, ",nogrpid");
253 	if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT2_DEF_RESUID)) ||
254 	    le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
255 		seq_printf(seq, ",resuid=%u",
256 				from_kuid_munged(&init_user_ns, sbi->s_resuid));
257 	}
258 	if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT2_DEF_RESGID)) ||
259 	    le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
260 		seq_printf(seq, ",resgid=%u",
261 				from_kgid_munged(&init_user_ns, sbi->s_resgid));
262 	}
263 	if (test_opt(sb, ERRORS_RO)) {
264 		int def_errors = le16_to_cpu(es->s_errors);
265 
266 		if (def_errors == EXT2_ERRORS_PANIC ||
267 		    def_errors == EXT2_ERRORS_CONTINUE) {
268 			seq_puts(seq, ",errors=remount-ro");
269 		}
270 	}
271 	if (test_opt(sb, ERRORS_CONT))
272 		seq_puts(seq, ",errors=continue");
273 	if (test_opt(sb, ERRORS_PANIC))
274 		seq_puts(seq, ",errors=panic");
275 	if (test_opt(sb, NO_UID32))
276 		seq_puts(seq, ",nouid32");
277 	if (test_opt(sb, DEBUG))
278 		seq_puts(seq, ",debug");
279 	if (test_opt(sb, OLDALLOC))
280 		seq_puts(seq, ",oldalloc");
281 
282 #ifdef CONFIG_EXT2_FS_XATTR
283 	if (test_opt(sb, XATTR_USER))
284 		seq_puts(seq, ",user_xattr");
285 	if (!test_opt(sb, XATTR_USER) &&
286 	    (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
287 		seq_puts(seq, ",nouser_xattr");
288 	}
289 #endif
290 
291 #ifdef CONFIG_EXT2_FS_POSIX_ACL
292 	if (test_opt(sb, POSIX_ACL))
293 		seq_puts(seq, ",acl");
294 	if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
295 		seq_puts(seq, ",noacl");
296 #endif
297 
298 	if (test_opt(sb, USRQUOTA))
299 		seq_puts(seq, ",usrquota");
300 
301 	if (test_opt(sb, GRPQUOTA))
302 		seq_puts(seq, ",grpquota");
303 
304 	if (test_opt(sb, XIP))
305 		seq_puts(seq, ",xip");
306 
307 	if (test_opt(sb, DAX))
308 		seq_puts(seq, ",dax");
309 
310 	if (!test_opt(sb, RESERVATION))
311 		seq_puts(seq, ",noreservation");
312 
313 	spin_unlock(&sbi->s_lock);
314 	return 0;
315 }
316 
317 #ifdef CONFIG_QUOTA
318 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
319 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
320 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
321 			 const struct path *path);
322 static struct dquot __rcu **ext2_get_dquots(struct inode *inode)
323 {
324 	return EXT2_I(inode)->i_dquot;
325 }
326 
327 static const struct quotactl_ops ext2_quotactl_ops = {
328 	.quota_on	= ext2_quota_on,
329 	.quota_off	= ext2_quota_off,
330 	.quota_sync	= dquot_quota_sync,
331 	.get_state	= dquot_get_state,
332 	.set_info	= dquot_set_dqinfo,
333 	.get_dqblk	= dquot_get_dqblk,
334 	.set_dqblk	= dquot_set_dqblk,
335 	.get_nextdqblk	= dquot_get_next_dqblk,
336 };
337 #endif
338 
339 static const struct super_operations ext2_sops = {
340 	.alloc_inode	= ext2_alloc_inode,
341 	.free_inode	= ext2_free_in_core_inode,
342 	.write_inode	= ext2_write_inode,
343 	.evict_inode	= ext2_evict_inode,
344 	.put_super	= ext2_put_super,
345 	.sync_fs	= ext2_sync_fs,
346 	.freeze_fs	= ext2_freeze,
347 	.unfreeze_fs	= ext2_unfreeze,
348 	.statfs		= ext2_statfs,
349 	.remount_fs	= ext2_remount,
350 	.show_options	= ext2_show_options,
351 #ifdef CONFIG_QUOTA
352 	.quota_read	= ext2_quota_read,
353 	.quota_write	= ext2_quota_write,
354 	.get_dquots	= ext2_get_dquots,
355 #endif
356 };
357 
358 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
359 		u64 ino, u32 generation)
360 {
361 	struct inode *inode;
362 
363 	if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
364 		return ERR_PTR(-ESTALE);
365 	if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
366 		return ERR_PTR(-ESTALE);
367 
368 	/*
369 	 * ext2_iget isn't quite right if the inode is currently unallocated!
370 	 * However ext2_iget currently does appropriate checks to handle stale
371 	 * inodes so everything is OK.
372 	 */
373 	inode = ext2_iget(sb, ino);
374 	if (IS_ERR(inode))
375 		return ERR_CAST(inode);
376 	if (generation && inode->i_generation != generation) {
377 		/* we didn't find the right inode.. */
378 		iput(inode);
379 		return ERR_PTR(-ESTALE);
380 	}
381 	return inode;
382 }
383 
384 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
385 		int fh_len, int fh_type)
386 {
387 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
388 				    ext2_nfs_get_inode);
389 }
390 
391 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
392 		int fh_len, int fh_type)
393 {
394 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
395 				    ext2_nfs_get_inode);
396 }
397 
398 static const struct export_operations ext2_export_ops = {
399 	.encode_fh = generic_encode_ino32_fh,
400 	.fh_to_dentry = ext2_fh_to_dentry,
401 	.fh_to_parent = ext2_fh_to_parent,
402 	.get_parent = ext2_get_parent,
403 };
404 
405 static unsigned long get_sb_block(void **data)
406 {
407 	unsigned long 	sb_block;
408 	char 		*options = (char *) *data;
409 
410 	if (!options || strncmp(options, "sb=", 3) != 0)
411 		return 1;	/* Default location */
412 	options += 3;
413 	sb_block = simple_strtoul(options, &options, 0);
414 	if (*options && *options != ',') {
415 		printk("EXT2-fs: Invalid sb specification: %s\n",
416 		       (char *) *data);
417 		return 1;
418 	}
419 	if (*options == ',')
420 		options++;
421 	*data = (void *) options;
422 	return sb_block;
423 }
424 
425 enum {
426 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
427 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
428 	Opt_err_ro, Opt_nouid32, Opt_debug,
429 	Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
430 	Opt_acl, Opt_noacl, Opt_xip, Opt_dax, Opt_ignore, Opt_err, Opt_quota,
431 	Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
432 };
433 
434 static const match_table_t tokens = {
435 	{Opt_bsd_df, "bsddf"},
436 	{Opt_minix_df, "minixdf"},
437 	{Opt_grpid, "grpid"},
438 	{Opt_grpid, "bsdgroups"},
439 	{Opt_nogrpid, "nogrpid"},
440 	{Opt_nogrpid, "sysvgroups"},
441 	{Opt_resgid, "resgid=%u"},
442 	{Opt_resuid, "resuid=%u"},
443 	{Opt_sb, "sb=%u"},
444 	{Opt_err_cont, "errors=continue"},
445 	{Opt_err_panic, "errors=panic"},
446 	{Opt_err_ro, "errors=remount-ro"},
447 	{Opt_nouid32, "nouid32"},
448 	{Opt_debug, "debug"},
449 	{Opt_oldalloc, "oldalloc"},
450 	{Opt_orlov, "orlov"},
451 	{Opt_nobh, "nobh"},
452 	{Opt_user_xattr, "user_xattr"},
453 	{Opt_nouser_xattr, "nouser_xattr"},
454 	{Opt_acl, "acl"},
455 	{Opt_noacl, "noacl"},
456 	{Opt_xip, "xip"},
457 	{Opt_dax, "dax"},
458 	{Opt_grpquota, "grpquota"},
459 	{Opt_ignore, "noquota"},
460 	{Opt_quota, "quota"},
461 	{Opt_usrquota, "usrquota"},
462 	{Opt_reservation, "reservation"},
463 	{Opt_noreservation, "noreservation"},
464 	{Opt_err, NULL}
465 };
466 
467 static int parse_options(char *options, struct super_block *sb,
468 			 struct ext2_mount_options *opts)
469 {
470 	char *p;
471 	substring_t args[MAX_OPT_ARGS];
472 	int option;
473 	kuid_t uid;
474 	kgid_t gid;
475 
476 	if (!options)
477 		return 1;
478 
479 	while ((p = strsep (&options, ",")) != NULL) {
480 		int token;
481 		if (!*p)
482 			continue;
483 
484 		token = match_token(p, tokens, args);
485 		switch (token) {
486 		case Opt_bsd_df:
487 			clear_opt (opts->s_mount_opt, MINIX_DF);
488 			break;
489 		case Opt_minix_df:
490 			set_opt (opts->s_mount_opt, MINIX_DF);
491 			break;
492 		case Opt_grpid:
493 			set_opt (opts->s_mount_opt, GRPID);
494 			break;
495 		case Opt_nogrpid:
496 			clear_opt (opts->s_mount_opt, GRPID);
497 			break;
498 		case Opt_resuid:
499 			if (match_int(&args[0], &option))
500 				return 0;
501 			uid = make_kuid(current_user_ns(), option);
502 			if (!uid_valid(uid)) {
503 				ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
504 				return 0;
505 
506 			}
507 			opts->s_resuid = uid;
508 			break;
509 		case Opt_resgid:
510 			if (match_int(&args[0], &option))
511 				return 0;
512 			gid = make_kgid(current_user_ns(), option);
513 			if (!gid_valid(gid)) {
514 				ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
515 				return 0;
516 			}
517 			opts->s_resgid = gid;
518 			break;
519 		case Opt_sb:
520 			/* handled by get_sb_block() instead of here */
521 			/* *sb_block = match_int(&args[0]); */
522 			break;
523 		case Opt_err_panic:
524 			clear_opt (opts->s_mount_opt, ERRORS_CONT);
525 			clear_opt (opts->s_mount_opt, ERRORS_RO);
526 			set_opt (opts->s_mount_opt, ERRORS_PANIC);
527 			break;
528 		case Opt_err_ro:
529 			clear_opt (opts->s_mount_opt, ERRORS_CONT);
530 			clear_opt (opts->s_mount_opt, ERRORS_PANIC);
531 			set_opt (opts->s_mount_opt, ERRORS_RO);
532 			break;
533 		case Opt_err_cont:
534 			clear_opt (opts->s_mount_opt, ERRORS_RO);
535 			clear_opt (opts->s_mount_opt, ERRORS_PANIC);
536 			set_opt (opts->s_mount_opt, ERRORS_CONT);
537 			break;
538 		case Opt_nouid32:
539 			set_opt (opts->s_mount_opt, NO_UID32);
540 			break;
541 		case Opt_debug:
542 			set_opt (opts->s_mount_opt, DEBUG);
543 			break;
544 		case Opt_oldalloc:
545 			set_opt (opts->s_mount_opt, OLDALLOC);
546 			break;
547 		case Opt_orlov:
548 			clear_opt (opts->s_mount_opt, OLDALLOC);
549 			break;
550 		case Opt_nobh:
551 			ext2_msg(sb, KERN_INFO,
552 				"nobh option not supported");
553 			break;
554 #ifdef CONFIG_EXT2_FS_XATTR
555 		case Opt_user_xattr:
556 			set_opt (opts->s_mount_opt, XATTR_USER);
557 			break;
558 		case Opt_nouser_xattr:
559 			clear_opt (opts->s_mount_opt, XATTR_USER);
560 			break;
561 #else
562 		case Opt_user_xattr:
563 		case Opt_nouser_xattr:
564 			ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
565 				"not supported");
566 			break;
567 #endif
568 #ifdef CONFIG_EXT2_FS_POSIX_ACL
569 		case Opt_acl:
570 			set_opt(opts->s_mount_opt, POSIX_ACL);
571 			break;
572 		case Opt_noacl:
573 			clear_opt(opts->s_mount_opt, POSIX_ACL);
574 			break;
575 #else
576 		case Opt_acl:
577 		case Opt_noacl:
578 			ext2_msg(sb, KERN_INFO,
579 				"(no)acl options not supported");
580 			break;
581 #endif
582 		case Opt_xip:
583 			ext2_msg(sb, KERN_INFO, "use dax instead of xip");
584 			set_opt(opts->s_mount_opt, XIP);
585 			fallthrough;
586 		case Opt_dax:
587 #ifdef CONFIG_FS_DAX
588 			ext2_msg(sb, KERN_WARNING,
589 		"DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
590 			set_opt(opts->s_mount_opt, DAX);
591 #else
592 			ext2_msg(sb, KERN_INFO, "dax option not supported");
593 #endif
594 			break;
595 
596 #if defined(CONFIG_QUOTA)
597 		case Opt_quota:
598 		case Opt_usrquota:
599 			set_opt(opts->s_mount_opt, USRQUOTA);
600 			break;
601 
602 		case Opt_grpquota:
603 			set_opt(opts->s_mount_opt, GRPQUOTA);
604 			break;
605 #else
606 		case Opt_quota:
607 		case Opt_usrquota:
608 		case Opt_grpquota:
609 			ext2_msg(sb, KERN_INFO,
610 				"quota operations not supported");
611 			break;
612 #endif
613 
614 		case Opt_reservation:
615 			set_opt(opts->s_mount_opt, RESERVATION);
616 			ext2_msg(sb, KERN_INFO, "reservations ON");
617 			break;
618 		case Opt_noreservation:
619 			clear_opt(opts->s_mount_opt, RESERVATION);
620 			ext2_msg(sb, KERN_INFO, "reservations OFF");
621 			break;
622 		case Opt_ignore:
623 			break;
624 		default:
625 			return 0;
626 		}
627 	}
628 	return 1;
629 }
630 
631 static int ext2_setup_super (struct super_block * sb,
632 			      struct ext2_super_block * es,
633 			      int read_only)
634 {
635 	int res = 0;
636 	struct ext2_sb_info *sbi = EXT2_SB(sb);
637 
638 	if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
639 		ext2_msg(sb, KERN_ERR,
640 			"error: revision level too high, "
641 			"forcing read-only mode");
642 		res = SB_RDONLY;
643 	}
644 	if (read_only)
645 		return res;
646 	if (!(sbi->s_mount_state & EXT2_VALID_FS))
647 		ext2_msg(sb, KERN_WARNING,
648 			"warning: mounting unchecked fs, "
649 			"running e2fsck is recommended");
650 	else if ((sbi->s_mount_state & EXT2_ERROR_FS))
651 		ext2_msg(sb, KERN_WARNING,
652 			"warning: mounting fs with errors, "
653 			"running e2fsck is recommended");
654 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
655 		 le16_to_cpu(es->s_mnt_count) >=
656 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
657 		ext2_msg(sb, KERN_WARNING,
658 			"warning: maximal mount count reached, "
659 			"running e2fsck is recommended");
660 	else if (le32_to_cpu(es->s_checkinterval) &&
661 		(le32_to_cpu(es->s_lastcheck) +
662 			le32_to_cpu(es->s_checkinterval) <=
663 			ktime_get_real_seconds()))
664 		ext2_msg(sb, KERN_WARNING,
665 			"warning: checktime reached, "
666 			"running e2fsck is recommended");
667 	if (!le16_to_cpu(es->s_max_mnt_count))
668 		es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
669 	le16_add_cpu(&es->s_mnt_count, 1);
670 	if (test_opt (sb, DEBUG))
671 		ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, gc=%lu, "
672 			"bpg=%lu, ipg=%lu, mo=%04lx]",
673 			EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
674 			sbi->s_groups_count,
675 			EXT2_BLOCKS_PER_GROUP(sb),
676 			EXT2_INODES_PER_GROUP(sb),
677 			sbi->s_mount_opt);
678 	return res;
679 }
680 
681 static int ext2_check_descriptors(struct super_block *sb)
682 {
683 	int i;
684 	struct ext2_sb_info *sbi = EXT2_SB(sb);
685 
686 	ext2_debug ("Checking group descriptors");
687 
688 	for (i = 0; i < sbi->s_groups_count; i++) {
689 		struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
690 		ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
691 		ext2_fsblk_t last_block = ext2_group_last_block_no(sb, i);
692 
693 		if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
694 		    le32_to_cpu(gdp->bg_block_bitmap) > last_block)
695 		{
696 			ext2_error (sb, "ext2_check_descriptors",
697 				    "Block bitmap for group %d"
698 				    " not in group (block %lu)!",
699 				    i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
700 			return 0;
701 		}
702 		if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
703 		    le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
704 		{
705 			ext2_error (sb, "ext2_check_descriptors",
706 				    "Inode bitmap for group %d"
707 				    " not in group (block %lu)!",
708 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
709 			return 0;
710 		}
711 		if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
712 		    le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
713 		    last_block)
714 		{
715 			ext2_error (sb, "ext2_check_descriptors",
716 				    "Inode table for group %d"
717 				    " not in group (block %lu)!",
718 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
719 			return 0;
720 		}
721 	}
722 	return 1;
723 }
724 
725 /*
726  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
727  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
728  * We need to be 1 filesystem block less than the 2^32 sector limit.
729  */
730 static loff_t ext2_max_size(int bits)
731 {
732 	loff_t res = EXT2_NDIR_BLOCKS;
733 	int meta_blocks;
734 	unsigned int upper_limit;
735 	unsigned int ppb = 1 << (bits-2);
736 
737 	/* This is calculated to be the largest file size for a
738 	 * dense, file such that the total number of
739 	 * sectors in the file, including data and all indirect blocks,
740 	 * does not exceed 2^32 -1
741 	 * __u32 i_blocks representing the total number of
742 	 * 512 bytes blocks of the file
743 	 */
744 	upper_limit = (1LL << 32) - 1;
745 
746 	/* total blocks in file system block size */
747 	upper_limit >>= (bits - 9);
748 
749 	/* Compute how many blocks we can address by block tree */
750 	res += 1LL << (bits-2);
751 	res += 1LL << (2*(bits-2));
752 	res += 1LL << (3*(bits-2));
753 	/* Compute how many metadata blocks are needed */
754 	meta_blocks = 1;
755 	meta_blocks += 1 + ppb;
756 	meta_blocks += 1 + ppb + ppb * ppb;
757 	/* Does block tree limit file size? */
758 	if (res + meta_blocks <= upper_limit)
759 		goto check_lfs;
760 
761 	res = upper_limit;
762 	/* How many metadata blocks are needed for addressing upper_limit? */
763 	upper_limit -= EXT2_NDIR_BLOCKS;
764 	/* indirect blocks */
765 	meta_blocks = 1;
766 	upper_limit -= ppb;
767 	/* double indirect blocks */
768 	if (upper_limit < ppb * ppb) {
769 		meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb);
770 		res -= meta_blocks;
771 		goto check_lfs;
772 	}
773 	meta_blocks += 1 + ppb;
774 	upper_limit -= ppb * ppb;
775 	/* tripple indirect blocks for the rest */
776 	meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb) +
777 		DIV_ROUND_UP(upper_limit, ppb*ppb);
778 	res -= meta_blocks;
779 check_lfs:
780 	res <<= bits;
781 	if (res > MAX_LFS_FILESIZE)
782 		res = MAX_LFS_FILESIZE;
783 
784 	return res;
785 }
786 
787 static unsigned long descriptor_loc(struct super_block *sb,
788 				    unsigned long logic_sb_block,
789 				    int nr)
790 {
791 	struct ext2_sb_info *sbi = EXT2_SB(sb);
792 	unsigned long bg, first_meta_bg;
793 
794 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
795 
796 	if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
797 	    nr < first_meta_bg)
798 		return (logic_sb_block + nr + 1);
799 	bg = sbi->s_desc_per_block * nr;
800 
801 	return ext2_group_first_block_no(sb, bg) + ext2_bg_has_super(sb, bg);
802 }
803 
804 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
805 {
806 	struct buffer_head * bh;
807 	struct ext2_sb_info * sbi;
808 	struct ext2_super_block * es;
809 	struct inode *root;
810 	unsigned long block;
811 	unsigned long sb_block = get_sb_block(&data);
812 	unsigned long logic_sb_block;
813 	unsigned long offset = 0;
814 	unsigned long def_mount_opts;
815 	long ret = -ENOMEM;
816 	int blocksize = BLOCK_SIZE;
817 	int db_count;
818 	int i, j;
819 	__le32 features;
820 	int err;
821 	struct ext2_mount_options opts;
822 
823 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
824 	if (!sbi)
825 		return -ENOMEM;
826 
827 	sbi->s_blockgroup_lock =
828 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
829 	if (!sbi->s_blockgroup_lock) {
830 		kfree(sbi);
831 		return -ENOMEM;
832 	}
833 	sb->s_fs_info = sbi;
834 	sbi->s_sb_block = sb_block;
835 	sbi->s_daxdev = fs_dax_get_by_bdev(sb->s_bdev, &sbi->s_dax_part_off,
836 					   NULL, NULL);
837 
838 	spin_lock_init(&sbi->s_lock);
839 	ret = -EINVAL;
840 
841 	/*
842 	 * See what the current blocksize for the device is, and
843 	 * use that as the blocksize.  Otherwise (or if the blocksize
844 	 * is smaller than the default) use the default.
845 	 * This is important for devices that have a hardware
846 	 * sectorsize that is larger than the default.
847 	 */
848 	blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
849 	if (!blocksize) {
850 		ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
851 		goto failed_sbi;
852 	}
853 
854 	/*
855 	 * If the superblock doesn't start on a hardware sector boundary,
856 	 * calculate the offset.
857 	 */
858 	if (blocksize != BLOCK_SIZE) {
859 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
860 		offset = (sb_block*BLOCK_SIZE) % blocksize;
861 	} else {
862 		logic_sb_block = sb_block;
863 	}
864 
865 	if (!(bh = sb_bread(sb, logic_sb_block))) {
866 		ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
867 		goto failed_sbi;
868 	}
869 	/*
870 	 * Note: s_es must be initialized as soon as possible because
871 	 *       some ext2 macro-instructions depend on its value
872 	 */
873 	es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
874 	sbi->s_es = es;
875 	sb->s_magic = le16_to_cpu(es->s_magic);
876 
877 	if (sb->s_magic != EXT2_SUPER_MAGIC)
878 		goto cantfind_ext2;
879 
880 	opts.s_mount_opt = 0;
881 	/* Set defaults before we parse the mount options */
882 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
883 	if (def_mount_opts & EXT2_DEFM_DEBUG)
884 		set_opt(opts.s_mount_opt, DEBUG);
885 	if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
886 		set_opt(opts.s_mount_opt, GRPID);
887 	if (def_mount_opts & EXT2_DEFM_UID16)
888 		set_opt(opts.s_mount_opt, NO_UID32);
889 #ifdef CONFIG_EXT2_FS_XATTR
890 	if (def_mount_opts & EXT2_DEFM_XATTR_USER)
891 		set_opt(opts.s_mount_opt, XATTR_USER);
892 #endif
893 #ifdef CONFIG_EXT2_FS_POSIX_ACL
894 	if (def_mount_opts & EXT2_DEFM_ACL)
895 		set_opt(opts.s_mount_opt, POSIX_ACL);
896 #endif
897 
898 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
899 		set_opt(opts.s_mount_opt, ERRORS_PANIC);
900 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
901 		set_opt(opts.s_mount_opt, ERRORS_CONT);
902 	else
903 		set_opt(opts.s_mount_opt, ERRORS_RO);
904 
905 	opts.s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
906 	opts.s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
907 
908 	set_opt(opts.s_mount_opt, RESERVATION);
909 
910 	if (!parse_options((char *) data, sb, &opts))
911 		goto failed_mount;
912 
913 	sbi->s_mount_opt = opts.s_mount_opt;
914 	sbi->s_resuid = opts.s_resuid;
915 	sbi->s_resgid = opts.s_resgid;
916 
917 	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
918 		(test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
919 	sb->s_iflags |= SB_I_CGROUPWB;
920 
921 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
922 	    (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
923 	     EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
924 	     EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
925 		ext2_msg(sb, KERN_WARNING,
926 			"warning: feature flags set on rev 0 fs, "
927 			"running e2fsck is recommended");
928 	/*
929 	 * Check feature flags regardless of the revision level, since we
930 	 * previously didn't change the revision level when setting the flags,
931 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
932 	 */
933 	features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
934 	if (features) {
935 		ext2_msg(sb, KERN_ERR,	"error: couldn't mount because of "
936 		       "unsupported optional features (%x)",
937 			le32_to_cpu(features));
938 		goto failed_mount;
939 	}
940 	if (!sb_rdonly(sb) && (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
941 		ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
942 		       "unsupported optional features (%x)",
943 		       le32_to_cpu(features));
944 		goto failed_mount;
945 	}
946 
947 	if (le32_to_cpu(es->s_log_block_size) >
948 	    (EXT2_MAX_BLOCK_LOG_SIZE - BLOCK_SIZE_BITS)) {
949 		ext2_msg(sb, KERN_ERR,
950 			 "Invalid log block size: %u",
951 			 le32_to_cpu(es->s_log_block_size));
952 		goto failed_mount;
953 	}
954 	blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
955 
956 	if (test_opt(sb, DAX)) {
957 		if (!sbi->s_daxdev) {
958 			ext2_msg(sb, KERN_ERR,
959 				"DAX unsupported by block device. Turning off DAX.");
960 			clear_opt(sbi->s_mount_opt, DAX);
961 		} else if (blocksize != PAGE_SIZE) {
962 			ext2_msg(sb, KERN_ERR, "unsupported blocksize for DAX\n");
963 			clear_opt(sbi->s_mount_opt, DAX);
964 		}
965 	}
966 
967 	/* If the blocksize doesn't match, re-read the thing.. */
968 	if (sb->s_blocksize != blocksize) {
969 		brelse(bh);
970 
971 		if (!sb_set_blocksize(sb, blocksize)) {
972 			ext2_msg(sb, KERN_ERR,
973 				"error: bad blocksize %d", blocksize);
974 			goto failed_sbi;
975 		}
976 
977 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
978 		offset = (sb_block*BLOCK_SIZE) % blocksize;
979 		bh = sb_bread(sb, logic_sb_block);
980 		if(!bh) {
981 			ext2_msg(sb, KERN_ERR, "error: couldn't read"
982 				"superblock on 2nd try");
983 			goto failed_sbi;
984 		}
985 		es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
986 		sbi->s_es = es;
987 		if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
988 			ext2_msg(sb, KERN_ERR, "error: magic mismatch");
989 			goto failed_mount;
990 		}
991 	}
992 
993 	sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
994 	sb->s_max_links = EXT2_LINK_MAX;
995 	sb->s_time_min = S32_MIN;
996 	sb->s_time_max = S32_MAX;
997 
998 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
999 		sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
1000 		sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
1001 	} else {
1002 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1003 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1004 		if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
1005 		    !is_power_of_2(sbi->s_inode_size) ||
1006 		    (sbi->s_inode_size > blocksize)) {
1007 			ext2_msg(sb, KERN_ERR,
1008 				"error: unsupported inode size: %d",
1009 				sbi->s_inode_size);
1010 			goto failed_mount;
1011 		}
1012 	}
1013 
1014 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1015 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1016 
1017 	sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
1018 	if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
1019 		goto cantfind_ext2;
1020 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
1021 					sbi->s_inodes_per_block;
1022 	sbi->s_desc_per_block = sb->s_blocksize /
1023 					sizeof (struct ext2_group_desc);
1024 	sbi->s_sbh = bh;
1025 	sbi->s_mount_state = le16_to_cpu(es->s_state);
1026 	sbi->s_addr_per_block_bits =
1027 		ilog2 (EXT2_ADDR_PER_BLOCK(sb));
1028 	sbi->s_desc_per_block_bits =
1029 		ilog2 (EXT2_DESC_PER_BLOCK(sb));
1030 
1031 	if (sb->s_magic != EXT2_SUPER_MAGIC)
1032 		goto cantfind_ext2;
1033 
1034 	if (sb->s_blocksize != bh->b_size) {
1035 		if (!silent)
1036 			ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
1037 		goto failed_mount;
1038 	}
1039 
1040 	if (es->s_log_frag_size != es->s_log_block_size) {
1041 		ext2_msg(sb, KERN_ERR,
1042 			"error: fragsize log %u != blocksize log %u",
1043 			le32_to_cpu(es->s_log_frag_size), sb->s_blocksize_bits);
1044 		goto failed_mount;
1045 	}
1046 
1047 	if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
1048 		ext2_msg(sb, KERN_ERR,
1049 			"error: #blocks per group too big: %lu",
1050 			sbi->s_blocks_per_group);
1051 		goto failed_mount;
1052 	}
1053 	/* At least inode table, bitmaps, and sb have to fit in one group */
1054 	if (sbi->s_blocks_per_group <= sbi->s_itb_per_group + 3) {
1055 		ext2_msg(sb, KERN_ERR,
1056 			"error: #blocks per group smaller than metadata size: %lu <= %lu",
1057 			sbi->s_blocks_per_group, sbi->s_inodes_per_group + 3);
1058 		goto failed_mount;
1059 	}
1060 	if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
1061 	    sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1062 		ext2_msg(sb, KERN_ERR,
1063 			"error: invalid #inodes per group: %lu",
1064 			sbi->s_inodes_per_group);
1065 		goto failed_mount;
1066 	}
1067 	if (sb_bdev_nr_blocks(sb) < le32_to_cpu(es->s_blocks_count)) {
1068 		ext2_msg(sb, KERN_ERR,
1069 			 "bad geometry: block count %u exceeds size of device (%u blocks)",
1070 			 le32_to_cpu(es->s_blocks_count),
1071 			 (unsigned)sb_bdev_nr_blocks(sb));
1072 		goto failed_mount;
1073 	}
1074 
1075 	sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1076 				le32_to_cpu(es->s_first_data_block) - 1)
1077 					/ EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1078 	if ((u64)sbi->s_groups_count * sbi->s_inodes_per_group !=
1079 	    le32_to_cpu(es->s_inodes_count)) {
1080 		ext2_msg(sb, KERN_ERR, "error: invalid #inodes: %u vs computed %llu",
1081 			 le32_to_cpu(es->s_inodes_count),
1082 			 (u64)sbi->s_groups_count * sbi->s_inodes_per_group);
1083 		goto failed_mount;
1084 	}
1085 	db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1086 		   EXT2_DESC_PER_BLOCK(sb);
1087 	sbi->s_group_desc = kvmalloc_array(db_count,
1088 					   sizeof(struct buffer_head *),
1089 					   GFP_KERNEL);
1090 	if (sbi->s_group_desc == NULL) {
1091 		ret = -ENOMEM;
1092 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1093 		goto failed_mount;
1094 	}
1095 	bgl_lock_init(sbi->s_blockgroup_lock);
1096 	sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1097 	if (!sbi->s_debts) {
1098 		ret = -ENOMEM;
1099 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1100 		goto failed_mount_group_desc;
1101 	}
1102 	for (i = 0; i < db_count; i++) {
1103 		block = descriptor_loc(sb, logic_sb_block, i);
1104 		sbi->s_group_desc[i] = sb_bread(sb, block);
1105 		if (!sbi->s_group_desc[i]) {
1106 			for (j = 0; j < i; j++)
1107 				brelse (sbi->s_group_desc[j]);
1108 			ext2_msg(sb, KERN_ERR,
1109 				"error: unable to read group descriptors");
1110 			goto failed_mount_group_desc;
1111 		}
1112 	}
1113 	if (!ext2_check_descriptors (sb)) {
1114 		ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1115 		goto failed_mount2;
1116 	}
1117 	sbi->s_gdb_count = db_count;
1118 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1119 	spin_lock_init(&sbi->s_next_gen_lock);
1120 
1121 	/* per filesystem reservation list head & lock */
1122 	spin_lock_init(&sbi->s_rsv_window_lock);
1123 	sbi->s_rsv_window_root = RB_ROOT;
1124 	/*
1125 	 * Add a single, static dummy reservation to the start of the
1126 	 * reservation window list --- it gives us a placeholder for
1127 	 * append-at-start-of-list which makes the allocation logic
1128 	 * _much_ simpler.
1129 	 */
1130 	sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1131 	sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1132 	sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1133 	sbi->s_rsv_window_head.rsv_goal_size = 0;
1134 	ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1135 
1136 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
1137 				ext2_count_free_blocks(sb), GFP_KERNEL);
1138 	if (!err) {
1139 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
1140 				ext2_count_free_inodes(sb), GFP_KERNEL);
1141 	}
1142 	if (!err) {
1143 		err = percpu_counter_init(&sbi->s_dirs_counter,
1144 				ext2_count_dirs(sb), GFP_KERNEL);
1145 	}
1146 	if (err) {
1147 		ret = err;
1148 		ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1149 		goto failed_mount3;
1150 	}
1151 
1152 #ifdef CONFIG_EXT2_FS_XATTR
1153 	sbi->s_ea_block_cache = ext2_xattr_create_cache();
1154 	if (!sbi->s_ea_block_cache) {
1155 		ret = -ENOMEM;
1156 		ext2_msg(sb, KERN_ERR, "Failed to create ea_block_cache");
1157 		goto failed_mount3;
1158 	}
1159 #endif
1160 	/*
1161 	 * set up enough so that it can read an inode
1162 	 */
1163 	sb->s_op = &ext2_sops;
1164 	sb->s_export_op = &ext2_export_ops;
1165 	sb->s_xattr = ext2_xattr_handlers;
1166 
1167 #ifdef CONFIG_QUOTA
1168 	sb->dq_op = &dquot_operations;
1169 	sb->s_qcop = &ext2_quotactl_ops;
1170 	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1171 #endif
1172 
1173 	root = ext2_iget(sb, EXT2_ROOT_INO);
1174 	if (IS_ERR(root)) {
1175 		ret = PTR_ERR(root);
1176 		goto failed_mount3;
1177 	}
1178 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1179 		iput(root);
1180 		ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1181 		goto failed_mount3;
1182 	}
1183 
1184 	sb->s_root = d_make_root(root);
1185 	if (!sb->s_root) {
1186 		ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1187 		ret = -ENOMEM;
1188 		goto failed_mount3;
1189 	}
1190 	if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1191 		ext2_msg(sb, KERN_WARNING,
1192 			"warning: mounting ext3 filesystem as ext2");
1193 	if (ext2_setup_super (sb, es, sb_rdonly(sb)))
1194 		sb->s_flags |= SB_RDONLY;
1195 	ext2_write_super(sb);
1196 	return 0;
1197 
1198 cantfind_ext2:
1199 	if (!silent)
1200 		ext2_msg(sb, KERN_ERR,
1201 			"error: can't find an ext2 filesystem on dev %s.",
1202 			sb->s_id);
1203 	goto failed_mount;
1204 failed_mount3:
1205 	ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
1206 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
1207 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
1208 	percpu_counter_destroy(&sbi->s_dirs_counter);
1209 failed_mount2:
1210 	for (i = 0; i < db_count; i++)
1211 		brelse(sbi->s_group_desc[i]);
1212 failed_mount_group_desc:
1213 	kvfree(sbi->s_group_desc);
1214 	kfree(sbi->s_debts);
1215 failed_mount:
1216 	brelse(bh);
1217 failed_sbi:
1218 	fs_put_dax(sbi->s_daxdev, NULL);
1219 	sb->s_fs_info = NULL;
1220 	kfree(sbi->s_blockgroup_lock);
1221 	kfree(sbi);
1222 	return ret;
1223 }
1224 
1225 static void ext2_clear_super_error(struct super_block *sb)
1226 {
1227 	struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1228 
1229 	if (buffer_write_io_error(sbh)) {
1230 		/*
1231 		 * Oh, dear.  A previous attempt to write the
1232 		 * superblock failed.  This could happen because the
1233 		 * USB device was yanked out.  Or it could happen to
1234 		 * be a transient write error and maybe the block will
1235 		 * be remapped.  Nothing we can do but to retry the
1236 		 * write and hope for the best.
1237 		 */
1238 		ext2_msg(sb, KERN_ERR,
1239 		       "previous I/O error to superblock detected");
1240 		clear_buffer_write_io_error(sbh);
1241 		set_buffer_uptodate(sbh);
1242 	}
1243 }
1244 
1245 void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1246 		     int wait)
1247 {
1248 	ext2_clear_super_error(sb);
1249 	spin_lock(&EXT2_SB(sb)->s_lock);
1250 	es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1251 	es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1252 	es->s_wtime = cpu_to_le32(ktime_get_real_seconds());
1253 	/* unlock before we do IO */
1254 	spin_unlock(&EXT2_SB(sb)->s_lock);
1255 	mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1256 	if (wait)
1257 		sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1258 }
1259 
1260 /*
1261  * In the second extended file system, it is not necessary to
1262  * write the super block since we use a mapping of the
1263  * disk super block in a buffer.
1264  *
1265  * However, this function is still used to set the fs valid
1266  * flags to 0.  We need to set this flag to 0 since the fs
1267  * may have been checked while mounted and e2fsck may have
1268  * set s_state to EXT2_VALID_FS after some corrections.
1269  */
1270 static int ext2_sync_fs(struct super_block *sb, int wait)
1271 {
1272 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1273 	struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1274 
1275 	/*
1276 	 * Write quota structures to quota file, sync_blockdev() will write
1277 	 * them to disk later
1278 	 */
1279 	dquot_writeback_dquots(sb, -1);
1280 
1281 	spin_lock(&sbi->s_lock);
1282 	if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1283 		ext2_debug("setting valid to 0\n");
1284 		es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1285 	}
1286 	spin_unlock(&sbi->s_lock);
1287 	ext2_sync_super(sb, es, wait);
1288 	return 0;
1289 }
1290 
1291 static int ext2_freeze(struct super_block *sb)
1292 {
1293 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1294 
1295 	/*
1296 	 * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1297 	 * because we have unattached inodes and thus filesystem is not fully
1298 	 * consistent.
1299 	 */
1300 	if (atomic_long_read(&sb->s_remove_count)) {
1301 		ext2_sync_fs(sb, 1);
1302 		return 0;
1303 	}
1304 	/* Set EXT2_FS_VALID flag */
1305 	spin_lock(&sbi->s_lock);
1306 	sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1307 	spin_unlock(&sbi->s_lock);
1308 	ext2_sync_super(sb, sbi->s_es, 1);
1309 
1310 	return 0;
1311 }
1312 
1313 static int ext2_unfreeze(struct super_block *sb)
1314 {
1315 	/* Just write sb to clear EXT2_VALID_FS flag */
1316 	ext2_write_super(sb);
1317 
1318 	return 0;
1319 }
1320 
1321 static void ext2_write_super(struct super_block *sb)
1322 {
1323 	if (!sb_rdonly(sb))
1324 		ext2_sync_fs(sb, 1);
1325 }
1326 
1327 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1328 {
1329 	struct ext2_sb_info * sbi = EXT2_SB(sb);
1330 	struct ext2_super_block * es;
1331 	struct ext2_mount_options new_opts;
1332 	int err;
1333 
1334 	sync_filesystem(sb);
1335 
1336 	spin_lock(&sbi->s_lock);
1337 	new_opts.s_mount_opt = sbi->s_mount_opt;
1338 	new_opts.s_resuid = sbi->s_resuid;
1339 	new_opts.s_resgid = sbi->s_resgid;
1340 	spin_unlock(&sbi->s_lock);
1341 
1342 	if (!parse_options(data, sb, &new_opts))
1343 		return -EINVAL;
1344 
1345 	spin_lock(&sbi->s_lock);
1346 	es = sbi->s_es;
1347 	if ((sbi->s_mount_opt ^ new_opts.s_mount_opt) & EXT2_MOUNT_DAX) {
1348 		ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1349 			 "dax flag with busy inodes while remounting");
1350 		new_opts.s_mount_opt ^= EXT2_MOUNT_DAX;
1351 	}
1352 	if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1353 		goto out_set;
1354 	if (*flags & SB_RDONLY) {
1355 		if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1356 		    !(sbi->s_mount_state & EXT2_VALID_FS))
1357 			goto out_set;
1358 
1359 		/*
1360 		 * OK, we are remounting a valid rw partition rdonly, so set
1361 		 * the rdonly flag and then mark the partition as valid again.
1362 		 */
1363 		es->s_state = cpu_to_le16(sbi->s_mount_state);
1364 		es->s_mtime = cpu_to_le32(ktime_get_real_seconds());
1365 		spin_unlock(&sbi->s_lock);
1366 
1367 		err = dquot_suspend(sb, -1);
1368 		if (err < 0)
1369 			return err;
1370 
1371 		ext2_sync_super(sb, es, 1);
1372 	} else {
1373 		__le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1374 					       ~EXT2_FEATURE_RO_COMPAT_SUPP);
1375 		if (ret) {
1376 			spin_unlock(&sbi->s_lock);
1377 			ext2_msg(sb, KERN_WARNING,
1378 				"warning: couldn't remount RDWR because of "
1379 				"unsupported optional features (%x).",
1380 				le32_to_cpu(ret));
1381 			return -EROFS;
1382 		}
1383 		/*
1384 		 * Mounting a RDONLY partition read-write, so reread and
1385 		 * store the current valid flag.  (It may have been changed
1386 		 * by e2fsck since we originally mounted the partition.)
1387 		 */
1388 		sbi->s_mount_state = le16_to_cpu(es->s_state);
1389 		if (!ext2_setup_super (sb, es, 0))
1390 			sb->s_flags &= ~SB_RDONLY;
1391 		spin_unlock(&sbi->s_lock);
1392 
1393 		ext2_write_super(sb);
1394 
1395 		dquot_resume(sb, -1);
1396 	}
1397 
1398 	spin_lock(&sbi->s_lock);
1399 out_set:
1400 	sbi->s_mount_opt = new_opts.s_mount_opt;
1401 	sbi->s_resuid = new_opts.s_resuid;
1402 	sbi->s_resgid = new_opts.s_resgid;
1403 	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1404 		(test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
1405 	spin_unlock(&sbi->s_lock);
1406 
1407 	return 0;
1408 }
1409 
1410 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1411 {
1412 	struct super_block *sb = dentry->d_sb;
1413 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1414 	struct ext2_super_block *es = sbi->s_es;
1415 
1416 	spin_lock(&sbi->s_lock);
1417 
1418 	if (test_opt (sb, MINIX_DF))
1419 		sbi->s_overhead_last = 0;
1420 	else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1421 		unsigned long i, overhead = 0;
1422 		smp_rmb();
1423 
1424 		/*
1425 		 * Compute the overhead (FS structures). This is constant
1426 		 * for a given filesystem unless the number of block groups
1427 		 * changes so we cache the previous value until it does.
1428 		 */
1429 
1430 		/*
1431 		 * All of the blocks before first_data_block are
1432 		 * overhead
1433 		 */
1434 		overhead = le32_to_cpu(es->s_first_data_block);
1435 
1436 		/*
1437 		 * Add the overhead attributed to the superblock and
1438 		 * block group descriptors.  If the sparse superblocks
1439 		 * feature is turned on, then not all groups have this.
1440 		 */
1441 		for (i = 0; i < sbi->s_groups_count; i++)
1442 			overhead += ext2_bg_has_super(sb, i) +
1443 				ext2_bg_num_gdb(sb, i);
1444 
1445 		/*
1446 		 * Every block group has an inode bitmap, a block
1447 		 * bitmap, and an inode table.
1448 		 */
1449 		overhead += (sbi->s_groups_count *
1450 			     (2 + sbi->s_itb_per_group));
1451 		sbi->s_overhead_last = overhead;
1452 		smp_wmb();
1453 		sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1454 	}
1455 
1456 	buf->f_type = EXT2_SUPER_MAGIC;
1457 	buf->f_bsize = sb->s_blocksize;
1458 	buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1459 	buf->f_bfree = ext2_count_free_blocks(sb);
1460 	es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1461 	buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1462 	if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1463 		buf->f_bavail = 0;
1464 	buf->f_files = le32_to_cpu(es->s_inodes_count);
1465 	buf->f_ffree = ext2_count_free_inodes(sb);
1466 	es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1467 	buf->f_namelen = EXT2_NAME_LEN;
1468 	buf->f_fsid = uuid_to_fsid(es->s_uuid);
1469 	spin_unlock(&sbi->s_lock);
1470 	return 0;
1471 }
1472 
1473 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1474 	int flags, const char *dev_name, void *data)
1475 {
1476 	return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1477 }
1478 
1479 #ifdef CONFIG_QUOTA
1480 
1481 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1482  * acquiring the locks... As quota files are never truncated and quota code
1483  * itself serializes the operations (and no one else should touch the files)
1484  * we don't have to be afraid of races */
1485 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1486 			       size_t len, loff_t off)
1487 {
1488 	struct inode *inode = sb_dqopt(sb)->files[type];
1489 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1490 	int err = 0;
1491 	int offset = off & (sb->s_blocksize - 1);
1492 	int tocopy;
1493 	size_t toread;
1494 	struct buffer_head tmp_bh;
1495 	struct buffer_head *bh;
1496 	loff_t i_size = i_size_read(inode);
1497 
1498 	if (off > i_size)
1499 		return 0;
1500 	if (off+len > i_size)
1501 		len = i_size-off;
1502 	toread = len;
1503 	while (toread > 0) {
1504 		tocopy = min_t(size_t, sb->s_blocksize - offset, toread);
1505 
1506 		tmp_bh.b_state = 0;
1507 		tmp_bh.b_size = sb->s_blocksize;
1508 		err = ext2_get_block(inode, blk, &tmp_bh, 0);
1509 		if (err < 0)
1510 			return err;
1511 		if (!buffer_mapped(&tmp_bh))	/* A hole? */
1512 			memset(data, 0, tocopy);
1513 		else {
1514 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1515 			if (!bh)
1516 				return -EIO;
1517 			memcpy(data, bh->b_data+offset, tocopy);
1518 			brelse(bh);
1519 		}
1520 		offset = 0;
1521 		toread -= tocopy;
1522 		data += tocopy;
1523 		blk++;
1524 	}
1525 	return len;
1526 }
1527 
1528 /* Write to quotafile */
1529 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1530 				const char *data, size_t len, loff_t off)
1531 {
1532 	struct inode *inode = sb_dqopt(sb)->files[type];
1533 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1534 	int err = 0;
1535 	int offset = off & (sb->s_blocksize - 1);
1536 	int tocopy;
1537 	size_t towrite = len;
1538 	struct buffer_head tmp_bh;
1539 	struct buffer_head *bh;
1540 
1541 	while (towrite > 0) {
1542 		tocopy = min_t(size_t, sb->s_blocksize - offset, towrite);
1543 
1544 		tmp_bh.b_state = 0;
1545 		tmp_bh.b_size = sb->s_blocksize;
1546 		err = ext2_get_block(inode, blk, &tmp_bh, 1);
1547 		if (err < 0)
1548 			goto out;
1549 		if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1550 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1551 		else
1552 			bh = sb_getblk(sb, tmp_bh.b_blocknr);
1553 		if (unlikely(!bh)) {
1554 			err = -EIO;
1555 			goto out;
1556 		}
1557 		lock_buffer(bh);
1558 		memcpy(bh->b_data+offset, data, tocopy);
1559 		flush_dcache_page(bh->b_page);
1560 		set_buffer_uptodate(bh);
1561 		mark_buffer_dirty(bh);
1562 		unlock_buffer(bh);
1563 		brelse(bh);
1564 		offset = 0;
1565 		towrite -= tocopy;
1566 		data += tocopy;
1567 		blk++;
1568 	}
1569 out:
1570 	if (len == towrite)
1571 		return err;
1572 	if (inode->i_size < off+len-towrite)
1573 		i_size_write(inode, off+len-towrite);
1574 	inode_inc_iversion(inode);
1575 	inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
1576 	mark_inode_dirty(inode);
1577 	return len - towrite;
1578 }
1579 
1580 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
1581 			 const struct path *path)
1582 {
1583 	int err;
1584 	struct inode *inode;
1585 
1586 	err = dquot_quota_on(sb, type, format_id, path);
1587 	if (err)
1588 		return err;
1589 
1590 	inode = d_inode(path->dentry);
1591 	inode_lock(inode);
1592 	EXT2_I(inode)->i_flags |= EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL;
1593 	inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1594 			S_NOATIME | S_IMMUTABLE);
1595 	inode_unlock(inode);
1596 	mark_inode_dirty(inode);
1597 
1598 	return 0;
1599 }
1600 
1601 static int ext2_quota_off(struct super_block *sb, int type)
1602 {
1603 	struct inode *inode = sb_dqopt(sb)->files[type];
1604 	int err;
1605 
1606 	if (!inode || !igrab(inode))
1607 		goto out;
1608 
1609 	err = dquot_quota_off(sb, type);
1610 	if (err)
1611 		goto out_put;
1612 
1613 	inode_lock(inode);
1614 	EXT2_I(inode)->i_flags &= ~(EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL);
1615 	inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1616 	inode_unlock(inode);
1617 	mark_inode_dirty(inode);
1618 out_put:
1619 	iput(inode);
1620 	return err;
1621 out:
1622 	return dquot_quota_off(sb, type);
1623 }
1624 
1625 #endif
1626 
1627 static struct file_system_type ext2_fs_type = {
1628 	.owner		= THIS_MODULE,
1629 	.name		= "ext2",
1630 	.mount		= ext2_mount,
1631 	.kill_sb	= kill_block_super,
1632 	.fs_flags	= FS_REQUIRES_DEV,
1633 };
1634 MODULE_ALIAS_FS("ext2");
1635 
1636 static int __init init_ext2_fs(void)
1637 {
1638 	int err;
1639 
1640 	err = init_inodecache();
1641 	if (err)
1642 		return err;
1643 	err = register_filesystem(&ext2_fs_type);
1644 	if (err)
1645 		goto out;
1646 	return 0;
1647 out:
1648 	destroy_inodecache();
1649 	return err;
1650 }
1651 
1652 static void __exit exit_ext2_fs(void)
1653 {
1654 	unregister_filesystem(&ext2_fs_type);
1655 	destroy_inodecache();
1656 }
1657 
1658 MODULE_AUTHOR("Remy Card and others");
1659 MODULE_DESCRIPTION("Second Extended Filesystem");
1660 MODULE_LICENSE("GPL");
1661 module_init(init_ext2_fs)
1662 module_exit(exit_ext2_fs)
1663