xref: /openbmc/linux/fs/ext2/super.c (revision 029f7f3b8701cc7aca8bdb31f0c7edd6a479e357)
1 /*
2  *  linux/fs/ext2/super.c
3  *
4  * Copyright (C) 1992, 1993, 1994, 1995
5  * Remy Card (card@masi.ibp.fr)
6  * Laboratoire MASI - Institut Blaise Pascal
7  * Universite Pierre et Marie Curie (Paris VI)
8  *
9  *  from
10  *
11  *  linux/fs/minix/inode.c
12  *
13  *  Copyright (C) 1991, 1992  Linus Torvalds
14  *
15  *  Big-endian to little-endian byte-swapping/bitmaps by
16  *        David S. Miller (davem@caip.rutgers.edu), 1995
17  */
18 
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/slab.h>
23 #include <linux/init.h>
24 #include <linux/blkdev.h>
25 #include <linux/parser.h>
26 #include <linux/random.h>
27 #include <linux/buffer_head.h>
28 #include <linux/exportfs.h>
29 #include <linux/vfs.h>
30 #include <linux/seq_file.h>
31 #include <linux/mount.h>
32 #include <linux/log2.h>
33 #include <linux/quotaops.h>
34 #include <asm/uaccess.h>
35 #include "ext2.h"
36 #include "xattr.h"
37 #include "acl.h"
38 
39 static void ext2_sync_super(struct super_block *sb,
40 			    struct ext2_super_block *es, int wait);
41 static int ext2_remount (struct super_block * sb, int * flags, char * data);
42 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
43 static int ext2_sync_fs(struct super_block *sb, int wait);
44 static int ext2_freeze(struct super_block *sb);
45 static int ext2_unfreeze(struct super_block *sb);
46 
47 void ext2_error(struct super_block *sb, const char *function,
48 		const char *fmt, ...)
49 {
50 	struct va_format vaf;
51 	va_list args;
52 	struct ext2_sb_info *sbi = EXT2_SB(sb);
53 	struct ext2_super_block *es = sbi->s_es;
54 
55 	if (!(sb->s_flags & MS_RDONLY)) {
56 		spin_lock(&sbi->s_lock);
57 		sbi->s_mount_state |= EXT2_ERROR_FS;
58 		es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
59 		spin_unlock(&sbi->s_lock);
60 		ext2_sync_super(sb, es, 1);
61 	}
62 
63 	va_start(args, fmt);
64 
65 	vaf.fmt = fmt;
66 	vaf.va = &args;
67 
68 	printk(KERN_CRIT "EXT2-fs (%s): error: %s: %pV\n",
69 	       sb->s_id, function, &vaf);
70 
71 	va_end(args);
72 
73 	if (test_opt(sb, ERRORS_PANIC))
74 		panic("EXT2-fs: panic from previous error\n");
75 	if (test_opt(sb, ERRORS_RO)) {
76 		ext2_msg(sb, KERN_CRIT,
77 			     "error: remounting filesystem read-only");
78 		sb->s_flags |= MS_RDONLY;
79 	}
80 }
81 
82 void ext2_msg(struct super_block *sb, const char *prefix,
83 		const char *fmt, ...)
84 {
85 	struct va_format vaf;
86 	va_list args;
87 
88 	va_start(args, fmt);
89 
90 	vaf.fmt = fmt;
91 	vaf.va = &args;
92 
93 	printk("%sEXT2-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
94 
95 	va_end(args);
96 }
97 
98 /*
99  * This must be called with sbi->s_lock held.
100  */
101 void ext2_update_dynamic_rev(struct super_block *sb)
102 {
103 	struct ext2_super_block *es = EXT2_SB(sb)->s_es;
104 
105 	if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
106 		return;
107 
108 	ext2_msg(sb, KERN_WARNING,
109 		     "warning: updating to rev %d because of "
110 		     "new feature flag, running e2fsck is recommended",
111 		     EXT2_DYNAMIC_REV);
112 
113 	es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
114 	es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
115 	es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
116 	/* leave es->s_feature_*compat flags alone */
117 	/* es->s_uuid will be set by e2fsck if empty */
118 
119 	/*
120 	 * The rest of the superblock fields should be zero, and if not it
121 	 * means they are likely already in use, so leave them alone.  We
122 	 * can leave it up to e2fsck to clean up any inconsistencies there.
123 	 */
124 }
125 
126 static void ext2_put_super (struct super_block * sb)
127 {
128 	int db_count;
129 	int i;
130 	struct ext2_sb_info *sbi = EXT2_SB(sb);
131 
132 	dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
133 
134 	ext2_xattr_put_super(sb);
135 	if (!(sb->s_flags & MS_RDONLY)) {
136 		struct ext2_super_block *es = sbi->s_es;
137 
138 		spin_lock(&sbi->s_lock);
139 		es->s_state = cpu_to_le16(sbi->s_mount_state);
140 		spin_unlock(&sbi->s_lock);
141 		ext2_sync_super(sb, es, 1);
142 	}
143 	db_count = sbi->s_gdb_count;
144 	for (i = 0; i < db_count; i++)
145 		if (sbi->s_group_desc[i])
146 			brelse (sbi->s_group_desc[i]);
147 	kfree(sbi->s_group_desc);
148 	kfree(sbi->s_debts);
149 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
150 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
151 	percpu_counter_destroy(&sbi->s_dirs_counter);
152 	brelse (sbi->s_sbh);
153 	sb->s_fs_info = NULL;
154 	kfree(sbi->s_blockgroup_lock);
155 	kfree(sbi);
156 }
157 
158 static struct kmem_cache * ext2_inode_cachep;
159 
160 static struct inode *ext2_alloc_inode(struct super_block *sb)
161 {
162 	struct ext2_inode_info *ei;
163 	ei = kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
164 	if (!ei)
165 		return NULL;
166 	ei->i_block_alloc_info = NULL;
167 	ei->vfs_inode.i_version = 1;
168 #ifdef CONFIG_QUOTA
169 	memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
170 #endif
171 
172 	return &ei->vfs_inode;
173 }
174 
175 static void ext2_i_callback(struct rcu_head *head)
176 {
177 	struct inode *inode = container_of(head, struct inode, i_rcu);
178 	kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
179 }
180 
181 static void ext2_destroy_inode(struct inode *inode)
182 {
183 	call_rcu(&inode->i_rcu, ext2_i_callback);
184 }
185 
186 static void init_once(void *foo)
187 {
188 	struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
189 
190 	rwlock_init(&ei->i_meta_lock);
191 #ifdef CONFIG_EXT2_FS_XATTR
192 	init_rwsem(&ei->xattr_sem);
193 #endif
194 	mutex_init(&ei->truncate_mutex);
195 #ifdef CONFIG_FS_DAX
196 	init_rwsem(&ei->dax_sem);
197 #endif
198 	inode_init_once(&ei->vfs_inode);
199 }
200 
201 static int __init init_inodecache(void)
202 {
203 	ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
204 					     sizeof(struct ext2_inode_info),
205 					     0, (SLAB_RECLAIM_ACCOUNT|
206 						SLAB_MEM_SPREAD),
207 					     init_once);
208 	if (ext2_inode_cachep == NULL)
209 		return -ENOMEM;
210 	return 0;
211 }
212 
213 static void destroy_inodecache(void)
214 {
215 	/*
216 	 * Make sure all delayed rcu free inodes are flushed before we
217 	 * destroy cache.
218 	 */
219 	rcu_barrier();
220 	kmem_cache_destroy(ext2_inode_cachep);
221 }
222 
223 static int ext2_show_options(struct seq_file *seq, struct dentry *root)
224 {
225 	struct super_block *sb = root->d_sb;
226 	struct ext2_sb_info *sbi = EXT2_SB(sb);
227 	struct ext2_super_block *es = sbi->s_es;
228 	unsigned long def_mount_opts;
229 
230 	spin_lock(&sbi->s_lock);
231 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
232 
233 	if (sbi->s_sb_block != 1)
234 		seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
235 	if (test_opt(sb, MINIX_DF))
236 		seq_puts(seq, ",minixdf");
237 	if (test_opt(sb, GRPID))
238 		seq_puts(seq, ",grpid");
239 	if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
240 		seq_puts(seq, ",nogrpid");
241 	if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT2_DEF_RESUID)) ||
242 	    le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
243 		seq_printf(seq, ",resuid=%u",
244 				from_kuid_munged(&init_user_ns, sbi->s_resuid));
245 	}
246 	if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT2_DEF_RESGID)) ||
247 	    le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
248 		seq_printf(seq, ",resgid=%u",
249 				from_kgid_munged(&init_user_ns, sbi->s_resgid));
250 	}
251 	if (test_opt(sb, ERRORS_RO)) {
252 		int def_errors = le16_to_cpu(es->s_errors);
253 
254 		if (def_errors == EXT2_ERRORS_PANIC ||
255 		    def_errors == EXT2_ERRORS_CONTINUE) {
256 			seq_puts(seq, ",errors=remount-ro");
257 		}
258 	}
259 	if (test_opt(sb, ERRORS_CONT))
260 		seq_puts(seq, ",errors=continue");
261 	if (test_opt(sb, ERRORS_PANIC))
262 		seq_puts(seq, ",errors=panic");
263 	if (test_opt(sb, NO_UID32))
264 		seq_puts(seq, ",nouid32");
265 	if (test_opt(sb, DEBUG))
266 		seq_puts(seq, ",debug");
267 	if (test_opt(sb, OLDALLOC))
268 		seq_puts(seq, ",oldalloc");
269 
270 #ifdef CONFIG_EXT2_FS_XATTR
271 	if (test_opt(sb, XATTR_USER))
272 		seq_puts(seq, ",user_xattr");
273 	if (!test_opt(sb, XATTR_USER) &&
274 	    (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
275 		seq_puts(seq, ",nouser_xattr");
276 	}
277 #endif
278 
279 #ifdef CONFIG_EXT2_FS_POSIX_ACL
280 	if (test_opt(sb, POSIX_ACL))
281 		seq_puts(seq, ",acl");
282 	if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
283 		seq_puts(seq, ",noacl");
284 #endif
285 
286 	if (test_opt(sb, NOBH))
287 		seq_puts(seq, ",nobh");
288 
289 #if defined(CONFIG_QUOTA)
290 	if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
291 		seq_puts(seq, ",usrquota");
292 
293 	if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
294 		seq_puts(seq, ",grpquota");
295 #endif
296 
297 #ifdef CONFIG_FS_DAX
298 	if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
299 		seq_puts(seq, ",xip");
300 	if (sbi->s_mount_opt & EXT2_MOUNT_DAX)
301 		seq_puts(seq, ",dax");
302 #endif
303 
304 	if (!test_opt(sb, RESERVATION))
305 		seq_puts(seq, ",noreservation");
306 
307 	spin_unlock(&sbi->s_lock);
308 	return 0;
309 }
310 
311 #ifdef CONFIG_QUOTA
312 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
313 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
314 static struct dquot **ext2_get_dquots(struct inode *inode)
315 {
316 	return EXT2_I(inode)->i_dquot;
317 }
318 #endif
319 
320 static const struct super_operations ext2_sops = {
321 	.alloc_inode	= ext2_alloc_inode,
322 	.destroy_inode	= ext2_destroy_inode,
323 	.write_inode	= ext2_write_inode,
324 	.evict_inode	= ext2_evict_inode,
325 	.put_super	= ext2_put_super,
326 	.sync_fs	= ext2_sync_fs,
327 	.freeze_fs	= ext2_freeze,
328 	.unfreeze_fs	= ext2_unfreeze,
329 	.statfs		= ext2_statfs,
330 	.remount_fs	= ext2_remount,
331 	.show_options	= ext2_show_options,
332 #ifdef CONFIG_QUOTA
333 	.quota_read	= ext2_quota_read,
334 	.quota_write	= ext2_quota_write,
335 	.get_dquots	= ext2_get_dquots,
336 #endif
337 };
338 
339 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
340 		u64 ino, u32 generation)
341 {
342 	struct inode *inode;
343 
344 	if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
345 		return ERR_PTR(-ESTALE);
346 	if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
347 		return ERR_PTR(-ESTALE);
348 
349 	/*
350 	 * ext2_iget isn't quite right if the inode is currently unallocated!
351 	 * However ext2_iget currently does appropriate checks to handle stale
352 	 * inodes so everything is OK.
353 	 */
354 	inode = ext2_iget(sb, ino);
355 	if (IS_ERR(inode))
356 		return ERR_CAST(inode);
357 	if (generation && inode->i_generation != generation) {
358 		/* we didn't find the right inode.. */
359 		iput(inode);
360 		return ERR_PTR(-ESTALE);
361 	}
362 	return inode;
363 }
364 
365 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
366 		int fh_len, int fh_type)
367 {
368 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
369 				    ext2_nfs_get_inode);
370 }
371 
372 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
373 		int fh_len, int fh_type)
374 {
375 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
376 				    ext2_nfs_get_inode);
377 }
378 
379 static const struct export_operations ext2_export_ops = {
380 	.fh_to_dentry = ext2_fh_to_dentry,
381 	.fh_to_parent = ext2_fh_to_parent,
382 	.get_parent = ext2_get_parent,
383 };
384 
385 static unsigned long get_sb_block(void **data)
386 {
387 	unsigned long 	sb_block;
388 	char 		*options = (char *) *data;
389 
390 	if (!options || strncmp(options, "sb=", 3) != 0)
391 		return 1;	/* Default location */
392 	options += 3;
393 	sb_block = simple_strtoul(options, &options, 0);
394 	if (*options && *options != ',') {
395 		printk("EXT2-fs: Invalid sb specification: %s\n",
396 		       (char *) *data);
397 		return 1;
398 	}
399 	if (*options == ',')
400 		options++;
401 	*data = (void *) options;
402 	return sb_block;
403 }
404 
405 enum {
406 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
407 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
408 	Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
409 	Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
410 	Opt_acl, Opt_noacl, Opt_xip, Opt_dax, Opt_ignore, Opt_err, Opt_quota,
411 	Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
412 };
413 
414 static const match_table_t tokens = {
415 	{Opt_bsd_df, "bsddf"},
416 	{Opt_minix_df, "minixdf"},
417 	{Opt_grpid, "grpid"},
418 	{Opt_grpid, "bsdgroups"},
419 	{Opt_nogrpid, "nogrpid"},
420 	{Opt_nogrpid, "sysvgroups"},
421 	{Opt_resgid, "resgid=%u"},
422 	{Opt_resuid, "resuid=%u"},
423 	{Opt_sb, "sb=%u"},
424 	{Opt_err_cont, "errors=continue"},
425 	{Opt_err_panic, "errors=panic"},
426 	{Opt_err_ro, "errors=remount-ro"},
427 	{Opt_nouid32, "nouid32"},
428 	{Opt_nocheck, "check=none"},
429 	{Opt_nocheck, "nocheck"},
430 	{Opt_debug, "debug"},
431 	{Opt_oldalloc, "oldalloc"},
432 	{Opt_orlov, "orlov"},
433 	{Opt_nobh, "nobh"},
434 	{Opt_user_xattr, "user_xattr"},
435 	{Opt_nouser_xattr, "nouser_xattr"},
436 	{Opt_acl, "acl"},
437 	{Opt_noacl, "noacl"},
438 	{Opt_xip, "xip"},
439 	{Opt_dax, "dax"},
440 	{Opt_grpquota, "grpquota"},
441 	{Opt_ignore, "noquota"},
442 	{Opt_quota, "quota"},
443 	{Opt_usrquota, "usrquota"},
444 	{Opt_reservation, "reservation"},
445 	{Opt_noreservation, "noreservation"},
446 	{Opt_err, NULL}
447 };
448 
449 static int parse_options(char *options, struct super_block *sb)
450 {
451 	char *p;
452 	struct ext2_sb_info *sbi = EXT2_SB(sb);
453 	substring_t args[MAX_OPT_ARGS];
454 	int option;
455 	kuid_t uid;
456 	kgid_t gid;
457 
458 	if (!options)
459 		return 1;
460 
461 	while ((p = strsep (&options, ",")) != NULL) {
462 		int token;
463 		if (!*p)
464 			continue;
465 
466 		token = match_token(p, tokens, args);
467 		switch (token) {
468 		case Opt_bsd_df:
469 			clear_opt (sbi->s_mount_opt, MINIX_DF);
470 			break;
471 		case Opt_minix_df:
472 			set_opt (sbi->s_mount_opt, MINIX_DF);
473 			break;
474 		case Opt_grpid:
475 			set_opt (sbi->s_mount_opt, GRPID);
476 			break;
477 		case Opt_nogrpid:
478 			clear_opt (sbi->s_mount_opt, GRPID);
479 			break;
480 		case Opt_resuid:
481 			if (match_int(&args[0], &option))
482 				return 0;
483 			uid = make_kuid(current_user_ns(), option);
484 			if (!uid_valid(uid)) {
485 				ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
486 				return 0;
487 
488 			}
489 			sbi->s_resuid = uid;
490 			break;
491 		case Opt_resgid:
492 			if (match_int(&args[0], &option))
493 				return 0;
494 			gid = make_kgid(current_user_ns(), option);
495 			if (!gid_valid(gid)) {
496 				ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
497 				return 0;
498 			}
499 			sbi->s_resgid = gid;
500 			break;
501 		case Opt_sb:
502 			/* handled by get_sb_block() instead of here */
503 			/* *sb_block = match_int(&args[0]); */
504 			break;
505 		case Opt_err_panic:
506 			clear_opt (sbi->s_mount_opt, ERRORS_CONT);
507 			clear_opt (sbi->s_mount_opt, ERRORS_RO);
508 			set_opt (sbi->s_mount_opt, ERRORS_PANIC);
509 			break;
510 		case Opt_err_ro:
511 			clear_opt (sbi->s_mount_opt, ERRORS_CONT);
512 			clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
513 			set_opt (sbi->s_mount_opt, ERRORS_RO);
514 			break;
515 		case Opt_err_cont:
516 			clear_opt (sbi->s_mount_opt, ERRORS_RO);
517 			clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
518 			set_opt (sbi->s_mount_opt, ERRORS_CONT);
519 			break;
520 		case Opt_nouid32:
521 			set_opt (sbi->s_mount_opt, NO_UID32);
522 			break;
523 		case Opt_nocheck:
524 			clear_opt (sbi->s_mount_opt, CHECK);
525 			break;
526 		case Opt_debug:
527 			set_opt (sbi->s_mount_opt, DEBUG);
528 			break;
529 		case Opt_oldalloc:
530 			set_opt (sbi->s_mount_opt, OLDALLOC);
531 			break;
532 		case Opt_orlov:
533 			clear_opt (sbi->s_mount_opt, OLDALLOC);
534 			break;
535 		case Opt_nobh:
536 			set_opt (sbi->s_mount_opt, NOBH);
537 			break;
538 #ifdef CONFIG_EXT2_FS_XATTR
539 		case Opt_user_xattr:
540 			set_opt (sbi->s_mount_opt, XATTR_USER);
541 			break;
542 		case Opt_nouser_xattr:
543 			clear_opt (sbi->s_mount_opt, XATTR_USER);
544 			break;
545 #else
546 		case Opt_user_xattr:
547 		case Opt_nouser_xattr:
548 			ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
549 				"not supported");
550 			break;
551 #endif
552 #ifdef CONFIG_EXT2_FS_POSIX_ACL
553 		case Opt_acl:
554 			set_opt(sbi->s_mount_opt, POSIX_ACL);
555 			break;
556 		case Opt_noacl:
557 			clear_opt(sbi->s_mount_opt, POSIX_ACL);
558 			break;
559 #else
560 		case Opt_acl:
561 		case Opt_noacl:
562 			ext2_msg(sb, KERN_INFO,
563 				"(no)acl options not supported");
564 			break;
565 #endif
566 		case Opt_xip:
567 			ext2_msg(sb, KERN_INFO, "use dax instead of xip");
568 			set_opt(sbi->s_mount_opt, XIP);
569 			/* Fall through */
570 		case Opt_dax:
571 #ifdef CONFIG_FS_DAX
572 			set_opt(sbi->s_mount_opt, DAX);
573 #else
574 			ext2_msg(sb, KERN_INFO, "dax option not supported");
575 #endif
576 			break;
577 
578 #if defined(CONFIG_QUOTA)
579 		case Opt_quota:
580 		case Opt_usrquota:
581 			set_opt(sbi->s_mount_opt, USRQUOTA);
582 			break;
583 
584 		case Opt_grpquota:
585 			set_opt(sbi->s_mount_opt, GRPQUOTA);
586 			break;
587 #else
588 		case Opt_quota:
589 		case Opt_usrquota:
590 		case Opt_grpquota:
591 			ext2_msg(sb, KERN_INFO,
592 				"quota operations not supported");
593 			break;
594 #endif
595 
596 		case Opt_reservation:
597 			set_opt(sbi->s_mount_opt, RESERVATION);
598 			ext2_msg(sb, KERN_INFO, "reservations ON");
599 			break;
600 		case Opt_noreservation:
601 			clear_opt(sbi->s_mount_opt, RESERVATION);
602 			ext2_msg(sb, KERN_INFO, "reservations OFF");
603 			break;
604 		case Opt_ignore:
605 			break;
606 		default:
607 			return 0;
608 		}
609 	}
610 	return 1;
611 }
612 
613 static int ext2_setup_super (struct super_block * sb,
614 			      struct ext2_super_block * es,
615 			      int read_only)
616 {
617 	int res = 0;
618 	struct ext2_sb_info *sbi = EXT2_SB(sb);
619 
620 	if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
621 		ext2_msg(sb, KERN_ERR,
622 			"error: revision level too high, "
623 			"forcing read-only mode");
624 		res = MS_RDONLY;
625 	}
626 	if (read_only)
627 		return res;
628 	if (!(sbi->s_mount_state & EXT2_VALID_FS))
629 		ext2_msg(sb, KERN_WARNING,
630 			"warning: mounting unchecked fs, "
631 			"running e2fsck is recommended");
632 	else if ((sbi->s_mount_state & EXT2_ERROR_FS))
633 		ext2_msg(sb, KERN_WARNING,
634 			"warning: mounting fs with errors, "
635 			"running e2fsck is recommended");
636 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
637 		 le16_to_cpu(es->s_mnt_count) >=
638 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
639 		ext2_msg(sb, KERN_WARNING,
640 			"warning: maximal mount count reached, "
641 			"running e2fsck is recommended");
642 	else if (le32_to_cpu(es->s_checkinterval) &&
643 		(le32_to_cpu(es->s_lastcheck) +
644 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
645 		ext2_msg(sb, KERN_WARNING,
646 			"warning: checktime reached, "
647 			"running e2fsck is recommended");
648 	if (!le16_to_cpu(es->s_max_mnt_count))
649 		es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
650 	le16_add_cpu(&es->s_mnt_count, 1);
651 	if (test_opt (sb, DEBUG))
652 		ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
653 			"bpg=%lu, ipg=%lu, mo=%04lx]",
654 			EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
655 			sbi->s_frag_size,
656 			sbi->s_groups_count,
657 			EXT2_BLOCKS_PER_GROUP(sb),
658 			EXT2_INODES_PER_GROUP(sb),
659 			sbi->s_mount_opt);
660 	return res;
661 }
662 
663 static int ext2_check_descriptors(struct super_block *sb)
664 {
665 	int i;
666 	struct ext2_sb_info *sbi = EXT2_SB(sb);
667 
668 	ext2_debug ("Checking group descriptors");
669 
670 	for (i = 0; i < sbi->s_groups_count; i++) {
671 		struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
672 		ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
673 		ext2_fsblk_t last_block;
674 
675 		if (i == sbi->s_groups_count - 1)
676 			last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
677 		else
678 			last_block = first_block +
679 				(EXT2_BLOCKS_PER_GROUP(sb) - 1);
680 
681 		if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
682 		    le32_to_cpu(gdp->bg_block_bitmap) > last_block)
683 		{
684 			ext2_error (sb, "ext2_check_descriptors",
685 				    "Block bitmap for group %d"
686 				    " not in group (block %lu)!",
687 				    i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
688 			return 0;
689 		}
690 		if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
691 		    le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
692 		{
693 			ext2_error (sb, "ext2_check_descriptors",
694 				    "Inode bitmap for group %d"
695 				    " not in group (block %lu)!",
696 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
697 			return 0;
698 		}
699 		if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
700 		    le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
701 		    last_block)
702 		{
703 			ext2_error (sb, "ext2_check_descriptors",
704 				    "Inode table for group %d"
705 				    " not in group (block %lu)!",
706 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
707 			return 0;
708 		}
709 	}
710 	return 1;
711 }
712 
713 /*
714  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
715  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
716  * We need to be 1 filesystem block less than the 2^32 sector limit.
717  */
718 static loff_t ext2_max_size(int bits)
719 {
720 	loff_t res = EXT2_NDIR_BLOCKS;
721 	int meta_blocks;
722 	loff_t upper_limit;
723 
724 	/* This is calculated to be the largest file size for a
725 	 * dense, file such that the total number of
726 	 * sectors in the file, including data and all indirect blocks,
727 	 * does not exceed 2^32 -1
728 	 * __u32 i_blocks representing the total number of
729 	 * 512 bytes blocks of the file
730 	 */
731 	upper_limit = (1LL << 32) - 1;
732 
733 	/* total blocks in file system block size */
734 	upper_limit >>= (bits - 9);
735 
736 
737 	/* indirect blocks */
738 	meta_blocks = 1;
739 	/* double indirect blocks */
740 	meta_blocks += 1 + (1LL << (bits-2));
741 	/* tripple indirect blocks */
742 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
743 
744 	upper_limit -= meta_blocks;
745 	upper_limit <<= bits;
746 
747 	res += 1LL << (bits-2);
748 	res += 1LL << (2*(bits-2));
749 	res += 1LL << (3*(bits-2));
750 	res <<= bits;
751 	if (res > upper_limit)
752 		res = upper_limit;
753 
754 	if (res > MAX_LFS_FILESIZE)
755 		res = MAX_LFS_FILESIZE;
756 
757 	return res;
758 }
759 
760 static unsigned long descriptor_loc(struct super_block *sb,
761 				    unsigned long logic_sb_block,
762 				    int nr)
763 {
764 	struct ext2_sb_info *sbi = EXT2_SB(sb);
765 	unsigned long bg, first_meta_bg;
766 	int has_super = 0;
767 
768 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
769 
770 	if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
771 	    nr < first_meta_bg)
772 		return (logic_sb_block + nr + 1);
773 	bg = sbi->s_desc_per_block * nr;
774 	if (ext2_bg_has_super(sb, bg))
775 		has_super = 1;
776 
777 	return ext2_group_first_block_no(sb, bg) + has_super;
778 }
779 
780 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
781 {
782 	struct buffer_head * bh;
783 	struct ext2_sb_info * sbi;
784 	struct ext2_super_block * es;
785 	struct inode *root;
786 	unsigned long block;
787 	unsigned long sb_block = get_sb_block(&data);
788 	unsigned long logic_sb_block;
789 	unsigned long offset = 0;
790 	unsigned long def_mount_opts;
791 	long ret = -EINVAL;
792 	int blocksize = BLOCK_SIZE;
793 	int db_count;
794 	int i, j;
795 	__le32 features;
796 	int err;
797 
798 	err = -ENOMEM;
799 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
800 	if (!sbi)
801 		goto failed;
802 
803 	sbi->s_blockgroup_lock =
804 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
805 	if (!sbi->s_blockgroup_lock) {
806 		kfree(sbi);
807 		goto failed;
808 	}
809 	sb->s_fs_info = sbi;
810 	sbi->s_sb_block = sb_block;
811 
812 	spin_lock_init(&sbi->s_lock);
813 
814 	/*
815 	 * See what the current blocksize for the device is, and
816 	 * use that as the blocksize.  Otherwise (or if the blocksize
817 	 * is smaller than the default) use the default.
818 	 * This is important for devices that have a hardware
819 	 * sectorsize that is larger than the default.
820 	 */
821 	blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
822 	if (!blocksize) {
823 		ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
824 		goto failed_sbi;
825 	}
826 
827 	/*
828 	 * If the superblock doesn't start on a hardware sector boundary,
829 	 * calculate the offset.
830 	 */
831 	if (blocksize != BLOCK_SIZE) {
832 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
833 		offset = (sb_block*BLOCK_SIZE) % blocksize;
834 	} else {
835 		logic_sb_block = sb_block;
836 	}
837 
838 	if (!(bh = sb_bread(sb, logic_sb_block))) {
839 		ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
840 		goto failed_sbi;
841 	}
842 	/*
843 	 * Note: s_es must be initialized as soon as possible because
844 	 *       some ext2 macro-instructions depend on its value
845 	 */
846 	es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
847 	sbi->s_es = es;
848 	sb->s_magic = le16_to_cpu(es->s_magic);
849 
850 	if (sb->s_magic != EXT2_SUPER_MAGIC)
851 		goto cantfind_ext2;
852 
853 	/* Set defaults before we parse the mount options */
854 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
855 	if (def_mount_opts & EXT2_DEFM_DEBUG)
856 		set_opt(sbi->s_mount_opt, DEBUG);
857 	if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
858 		set_opt(sbi->s_mount_opt, GRPID);
859 	if (def_mount_opts & EXT2_DEFM_UID16)
860 		set_opt(sbi->s_mount_opt, NO_UID32);
861 #ifdef CONFIG_EXT2_FS_XATTR
862 	if (def_mount_opts & EXT2_DEFM_XATTR_USER)
863 		set_opt(sbi->s_mount_opt, XATTR_USER);
864 #endif
865 #ifdef CONFIG_EXT2_FS_POSIX_ACL
866 	if (def_mount_opts & EXT2_DEFM_ACL)
867 		set_opt(sbi->s_mount_opt, POSIX_ACL);
868 #endif
869 
870 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
871 		set_opt(sbi->s_mount_opt, ERRORS_PANIC);
872 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
873 		set_opt(sbi->s_mount_opt, ERRORS_CONT);
874 	else
875 		set_opt(sbi->s_mount_opt, ERRORS_RO);
876 
877 	sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
878 	sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
879 
880 	set_opt(sbi->s_mount_opt, RESERVATION);
881 
882 	if (!parse_options((char *) data, sb))
883 		goto failed_mount;
884 
885 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
886 		((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
887 		 MS_POSIXACL : 0);
888 	sb->s_iflags |= SB_I_CGROUPWB;
889 
890 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
891 	    (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
892 	     EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
893 	     EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
894 		ext2_msg(sb, KERN_WARNING,
895 			"warning: feature flags set on rev 0 fs, "
896 			"running e2fsck is recommended");
897 	/*
898 	 * Check feature flags regardless of the revision level, since we
899 	 * previously didn't change the revision level when setting the flags,
900 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
901 	 */
902 	features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
903 	if (features) {
904 		ext2_msg(sb, KERN_ERR,	"error: couldn't mount because of "
905 		       "unsupported optional features (%x)",
906 			le32_to_cpu(features));
907 		goto failed_mount;
908 	}
909 	if (!(sb->s_flags & MS_RDONLY) &&
910 	    (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
911 		ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
912 		       "unsupported optional features (%x)",
913 		       le32_to_cpu(features));
914 		goto failed_mount;
915 	}
916 
917 	blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
918 
919 	if (sbi->s_mount_opt & EXT2_MOUNT_DAX) {
920 		if (blocksize != PAGE_SIZE) {
921 			ext2_msg(sb, KERN_ERR,
922 					"error: unsupported blocksize for dax");
923 			goto failed_mount;
924 		}
925 		if (!sb->s_bdev->bd_disk->fops->direct_access) {
926 			ext2_msg(sb, KERN_ERR,
927 					"error: device does not support dax");
928 			goto failed_mount;
929 		}
930 	}
931 
932 	/* If the blocksize doesn't match, re-read the thing.. */
933 	if (sb->s_blocksize != blocksize) {
934 		brelse(bh);
935 
936 		if (!sb_set_blocksize(sb, blocksize)) {
937 			ext2_msg(sb, KERN_ERR,
938 				"error: bad blocksize %d", blocksize);
939 			goto failed_sbi;
940 		}
941 
942 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
943 		offset = (sb_block*BLOCK_SIZE) % blocksize;
944 		bh = sb_bread(sb, logic_sb_block);
945 		if(!bh) {
946 			ext2_msg(sb, KERN_ERR, "error: couldn't read"
947 				"superblock on 2nd try");
948 			goto failed_sbi;
949 		}
950 		es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
951 		sbi->s_es = es;
952 		if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
953 			ext2_msg(sb, KERN_ERR, "error: magic mismatch");
954 			goto failed_mount;
955 		}
956 	}
957 
958 	sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
959 	sb->s_max_links = EXT2_LINK_MAX;
960 
961 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
962 		sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
963 		sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
964 	} else {
965 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
966 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
967 		if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
968 		    !is_power_of_2(sbi->s_inode_size) ||
969 		    (sbi->s_inode_size > blocksize)) {
970 			ext2_msg(sb, KERN_ERR,
971 				"error: unsupported inode size: %d",
972 				sbi->s_inode_size);
973 			goto failed_mount;
974 		}
975 	}
976 
977 	sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
978 				   le32_to_cpu(es->s_log_frag_size);
979 	if (sbi->s_frag_size == 0)
980 		goto cantfind_ext2;
981 	sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
982 
983 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
984 	sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
985 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
986 
987 	if (EXT2_INODE_SIZE(sb) == 0)
988 		goto cantfind_ext2;
989 	sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
990 	if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
991 		goto cantfind_ext2;
992 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
993 					sbi->s_inodes_per_block;
994 	sbi->s_desc_per_block = sb->s_blocksize /
995 					sizeof (struct ext2_group_desc);
996 	sbi->s_sbh = bh;
997 	sbi->s_mount_state = le16_to_cpu(es->s_state);
998 	sbi->s_addr_per_block_bits =
999 		ilog2 (EXT2_ADDR_PER_BLOCK(sb));
1000 	sbi->s_desc_per_block_bits =
1001 		ilog2 (EXT2_DESC_PER_BLOCK(sb));
1002 
1003 	if (sb->s_magic != EXT2_SUPER_MAGIC)
1004 		goto cantfind_ext2;
1005 
1006 	if (sb->s_blocksize != bh->b_size) {
1007 		if (!silent)
1008 			ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
1009 		goto failed_mount;
1010 	}
1011 
1012 	if (sb->s_blocksize != sbi->s_frag_size) {
1013 		ext2_msg(sb, KERN_ERR,
1014 			"error: fragsize %lu != blocksize %lu"
1015 			"(not supported yet)",
1016 			sbi->s_frag_size, sb->s_blocksize);
1017 		goto failed_mount;
1018 	}
1019 
1020 	if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
1021 		ext2_msg(sb, KERN_ERR,
1022 			"error: #blocks per group too big: %lu",
1023 			sbi->s_blocks_per_group);
1024 		goto failed_mount;
1025 	}
1026 	if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
1027 		ext2_msg(sb, KERN_ERR,
1028 			"error: #fragments per group too big: %lu",
1029 			sbi->s_frags_per_group);
1030 		goto failed_mount;
1031 	}
1032 	if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1033 		ext2_msg(sb, KERN_ERR,
1034 			"error: #inodes per group too big: %lu",
1035 			sbi->s_inodes_per_group);
1036 		goto failed_mount;
1037 	}
1038 
1039 	if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1040 		goto cantfind_ext2;
1041  	sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1042  				le32_to_cpu(es->s_first_data_block) - 1)
1043  					/ EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1044 	db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1045 		   EXT2_DESC_PER_BLOCK(sb);
1046 	sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
1047 	if (sbi->s_group_desc == NULL) {
1048 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1049 		goto failed_mount;
1050 	}
1051 	bgl_lock_init(sbi->s_blockgroup_lock);
1052 	sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1053 	if (!sbi->s_debts) {
1054 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1055 		goto failed_mount_group_desc;
1056 	}
1057 	for (i = 0; i < db_count; i++) {
1058 		block = descriptor_loc(sb, logic_sb_block, i);
1059 		sbi->s_group_desc[i] = sb_bread(sb, block);
1060 		if (!sbi->s_group_desc[i]) {
1061 			for (j = 0; j < i; j++)
1062 				brelse (sbi->s_group_desc[j]);
1063 			ext2_msg(sb, KERN_ERR,
1064 				"error: unable to read group descriptors");
1065 			goto failed_mount_group_desc;
1066 		}
1067 	}
1068 	if (!ext2_check_descriptors (sb)) {
1069 		ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1070 		goto failed_mount2;
1071 	}
1072 	sbi->s_gdb_count = db_count;
1073 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1074 	spin_lock_init(&sbi->s_next_gen_lock);
1075 
1076 	/* per fileystem reservation list head & lock */
1077 	spin_lock_init(&sbi->s_rsv_window_lock);
1078 	sbi->s_rsv_window_root = RB_ROOT;
1079 	/*
1080 	 * Add a single, static dummy reservation to the start of the
1081 	 * reservation window list --- it gives us a placeholder for
1082 	 * append-at-start-of-list which makes the allocation logic
1083 	 * _much_ simpler.
1084 	 */
1085 	sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1086 	sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1087 	sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1088 	sbi->s_rsv_window_head.rsv_goal_size = 0;
1089 	ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1090 
1091 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
1092 				ext2_count_free_blocks(sb), GFP_KERNEL);
1093 	if (!err) {
1094 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
1095 				ext2_count_free_inodes(sb), GFP_KERNEL);
1096 	}
1097 	if (!err) {
1098 		err = percpu_counter_init(&sbi->s_dirs_counter,
1099 				ext2_count_dirs(sb), GFP_KERNEL);
1100 	}
1101 	if (err) {
1102 		ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1103 		goto failed_mount3;
1104 	}
1105 	/*
1106 	 * set up enough so that it can read an inode
1107 	 */
1108 	sb->s_op = &ext2_sops;
1109 	sb->s_export_op = &ext2_export_ops;
1110 	sb->s_xattr = ext2_xattr_handlers;
1111 
1112 #ifdef CONFIG_QUOTA
1113 	sb->dq_op = &dquot_operations;
1114 	sb->s_qcop = &dquot_quotactl_ops;
1115 	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1116 #endif
1117 
1118 	root = ext2_iget(sb, EXT2_ROOT_INO);
1119 	if (IS_ERR(root)) {
1120 		ret = PTR_ERR(root);
1121 		goto failed_mount3;
1122 	}
1123 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1124 		iput(root);
1125 		ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1126 		goto failed_mount3;
1127 	}
1128 
1129 	sb->s_root = d_make_root(root);
1130 	if (!sb->s_root) {
1131 		ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1132 		ret = -ENOMEM;
1133 		goto failed_mount3;
1134 	}
1135 	if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1136 		ext2_msg(sb, KERN_WARNING,
1137 			"warning: mounting ext3 filesystem as ext2");
1138 	if (ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY))
1139 		sb->s_flags |= MS_RDONLY;
1140 	ext2_write_super(sb);
1141 	return 0;
1142 
1143 cantfind_ext2:
1144 	if (!silent)
1145 		ext2_msg(sb, KERN_ERR,
1146 			"error: can't find an ext2 filesystem on dev %s.",
1147 			sb->s_id);
1148 	goto failed_mount;
1149 failed_mount3:
1150 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
1151 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
1152 	percpu_counter_destroy(&sbi->s_dirs_counter);
1153 failed_mount2:
1154 	for (i = 0; i < db_count; i++)
1155 		brelse(sbi->s_group_desc[i]);
1156 failed_mount_group_desc:
1157 	kfree(sbi->s_group_desc);
1158 	kfree(sbi->s_debts);
1159 failed_mount:
1160 	brelse(bh);
1161 failed_sbi:
1162 	sb->s_fs_info = NULL;
1163 	kfree(sbi->s_blockgroup_lock);
1164 	kfree(sbi);
1165 failed:
1166 	return ret;
1167 }
1168 
1169 static void ext2_clear_super_error(struct super_block *sb)
1170 {
1171 	struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1172 
1173 	if (buffer_write_io_error(sbh)) {
1174 		/*
1175 		 * Oh, dear.  A previous attempt to write the
1176 		 * superblock failed.  This could happen because the
1177 		 * USB device was yanked out.  Or it could happen to
1178 		 * be a transient write error and maybe the block will
1179 		 * be remapped.  Nothing we can do but to retry the
1180 		 * write and hope for the best.
1181 		 */
1182 		ext2_msg(sb, KERN_ERR,
1183 		       "previous I/O error to superblock detected\n");
1184 		clear_buffer_write_io_error(sbh);
1185 		set_buffer_uptodate(sbh);
1186 	}
1187 }
1188 
1189 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1190 			    int wait)
1191 {
1192 	ext2_clear_super_error(sb);
1193 	spin_lock(&EXT2_SB(sb)->s_lock);
1194 	es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1195 	es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1196 	es->s_wtime = cpu_to_le32(get_seconds());
1197 	/* unlock before we do IO */
1198 	spin_unlock(&EXT2_SB(sb)->s_lock);
1199 	mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1200 	if (wait)
1201 		sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1202 }
1203 
1204 /*
1205  * In the second extended file system, it is not necessary to
1206  * write the super block since we use a mapping of the
1207  * disk super block in a buffer.
1208  *
1209  * However, this function is still used to set the fs valid
1210  * flags to 0.  We need to set this flag to 0 since the fs
1211  * may have been checked while mounted and e2fsck may have
1212  * set s_state to EXT2_VALID_FS after some corrections.
1213  */
1214 static int ext2_sync_fs(struct super_block *sb, int wait)
1215 {
1216 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1217 	struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1218 
1219 	/*
1220 	 * Write quota structures to quota file, sync_blockdev() will write
1221 	 * them to disk later
1222 	 */
1223 	dquot_writeback_dquots(sb, -1);
1224 
1225 	spin_lock(&sbi->s_lock);
1226 	if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1227 		ext2_debug("setting valid to 0\n");
1228 		es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1229 	}
1230 	spin_unlock(&sbi->s_lock);
1231 	ext2_sync_super(sb, es, wait);
1232 	return 0;
1233 }
1234 
1235 static int ext2_freeze(struct super_block *sb)
1236 {
1237 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1238 
1239 	/*
1240 	 * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1241 	 * because we have unattached inodes and thus filesystem is not fully
1242 	 * consistent.
1243 	 */
1244 	if (atomic_long_read(&sb->s_remove_count)) {
1245 		ext2_sync_fs(sb, 1);
1246 		return 0;
1247 	}
1248 	/* Set EXT2_FS_VALID flag */
1249 	spin_lock(&sbi->s_lock);
1250 	sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1251 	spin_unlock(&sbi->s_lock);
1252 	ext2_sync_super(sb, sbi->s_es, 1);
1253 
1254 	return 0;
1255 }
1256 
1257 static int ext2_unfreeze(struct super_block *sb)
1258 {
1259 	/* Just write sb to clear EXT2_VALID_FS flag */
1260 	ext2_write_super(sb);
1261 
1262 	return 0;
1263 }
1264 
1265 void ext2_write_super(struct super_block *sb)
1266 {
1267 	if (!(sb->s_flags & MS_RDONLY))
1268 		ext2_sync_fs(sb, 1);
1269 }
1270 
1271 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1272 {
1273 	struct ext2_sb_info * sbi = EXT2_SB(sb);
1274 	struct ext2_super_block * es;
1275 	struct ext2_mount_options old_opts;
1276 	unsigned long old_sb_flags;
1277 	int err;
1278 
1279 	sync_filesystem(sb);
1280 	spin_lock(&sbi->s_lock);
1281 
1282 	/* Store the old options */
1283 	old_sb_flags = sb->s_flags;
1284 	old_opts.s_mount_opt = sbi->s_mount_opt;
1285 	old_opts.s_resuid = sbi->s_resuid;
1286 	old_opts.s_resgid = sbi->s_resgid;
1287 
1288 	/*
1289 	 * Allow the "check" option to be passed as a remount option.
1290 	 */
1291 	if (!parse_options(data, sb)) {
1292 		err = -EINVAL;
1293 		goto restore_opts;
1294 	}
1295 
1296 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1297 		((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1298 
1299 	es = sbi->s_es;
1300 	if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT2_MOUNT_DAX) {
1301 		ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1302 			 "dax flag with busy inodes while remounting");
1303 		sbi->s_mount_opt ^= EXT2_MOUNT_DAX;
1304 	}
1305 	if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) {
1306 		spin_unlock(&sbi->s_lock);
1307 		return 0;
1308 	}
1309 	if (*flags & MS_RDONLY) {
1310 		if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1311 		    !(sbi->s_mount_state & EXT2_VALID_FS)) {
1312 			spin_unlock(&sbi->s_lock);
1313 			return 0;
1314 		}
1315 
1316 		/*
1317 		 * OK, we are remounting a valid rw partition rdonly, so set
1318 		 * the rdonly flag and then mark the partition as valid again.
1319 		 */
1320 		es->s_state = cpu_to_le16(sbi->s_mount_state);
1321 		es->s_mtime = cpu_to_le32(get_seconds());
1322 		spin_unlock(&sbi->s_lock);
1323 
1324 		err = dquot_suspend(sb, -1);
1325 		if (err < 0) {
1326 			spin_lock(&sbi->s_lock);
1327 			goto restore_opts;
1328 		}
1329 
1330 		ext2_sync_super(sb, es, 1);
1331 	} else {
1332 		__le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1333 					       ~EXT2_FEATURE_RO_COMPAT_SUPP);
1334 		if (ret) {
1335 			ext2_msg(sb, KERN_WARNING,
1336 				"warning: couldn't remount RDWR because of "
1337 				"unsupported optional features (%x).",
1338 				le32_to_cpu(ret));
1339 			err = -EROFS;
1340 			goto restore_opts;
1341 		}
1342 		/*
1343 		 * Mounting a RDONLY partition read-write, so reread and
1344 		 * store the current valid flag.  (It may have been changed
1345 		 * by e2fsck since we originally mounted the partition.)
1346 		 */
1347 		sbi->s_mount_state = le16_to_cpu(es->s_state);
1348 		if (!ext2_setup_super (sb, es, 0))
1349 			sb->s_flags &= ~MS_RDONLY;
1350 		spin_unlock(&sbi->s_lock);
1351 
1352 		ext2_write_super(sb);
1353 
1354 		dquot_resume(sb, -1);
1355 	}
1356 
1357 	return 0;
1358 restore_opts:
1359 	sbi->s_mount_opt = old_opts.s_mount_opt;
1360 	sbi->s_resuid = old_opts.s_resuid;
1361 	sbi->s_resgid = old_opts.s_resgid;
1362 	sb->s_flags = old_sb_flags;
1363 	spin_unlock(&sbi->s_lock);
1364 	return err;
1365 }
1366 
1367 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1368 {
1369 	struct super_block *sb = dentry->d_sb;
1370 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1371 	struct ext2_super_block *es = sbi->s_es;
1372 	u64 fsid;
1373 
1374 	spin_lock(&sbi->s_lock);
1375 
1376 	if (test_opt (sb, MINIX_DF))
1377 		sbi->s_overhead_last = 0;
1378 	else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1379 		unsigned long i, overhead = 0;
1380 		smp_rmb();
1381 
1382 		/*
1383 		 * Compute the overhead (FS structures). This is constant
1384 		 * for a given filesystem unless the number of block groups
1385 		 * changes so we cache the previous value until it does.
1386 		 */
1387 
1388 		/*
1389 		 * All of the blocks before first_data_block are
1390 		 * overhead
1391 		 */
1392 		overhead = le32_to_cpu(es->s_first_data_block);
1393 
1394 		/*
1395 		 * Add the overhead attributed to the superblock and
1396 		 * block group descriptors.  If the sparse superblocks
1397 		 * feature is turned on, then not all groups have this.
1398 		 */
1399 		for (i = 0; i < sbi->s_groups_count; i++)
1400 			overhead += ext2_bg_has_super(sb, i) +
1401 				ext2_bg_num_gdb(sb, i);
1402 
1403 		/*
1404 		 * Every block group has an inode bitmap, a block
1405 		 * bitmap, and an inode table.
1406 		 */
1407 		overhead += (sbi->s_groups_count *
1408 			     (2 + sbi->s_itb_per_group));
1409 		sbi->s_overhead_last = overhead;
1410 		smp_wmb();
1411 		sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1412 	}
1413 
1414 	buf->f_type = EXT2_SUPER_MAGIC;
1415 	buf->f_bsize = sb->s_blocksize;
1416 	buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1417 	buf->f_bfree = ext2_count_free_blocks(sb);
1418 	es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1419 	buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1420 	if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1421 		buf->f_bavail = 0;
1422 	buf->f_files = le32_to_cpu(es->s_inodes_count);
1423 	buf->f_ffree = ext2_count_free_inodes(sb);
1424 	es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1425 	buf->f_namelen = EXT2_NAME_LEN;
1426 	fsid = le64_to_cpup((void *)es->s_uuid) ^
1427 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1428 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1429 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1430 	spin_unlock(&sbi->s_lock);
1431 	return 0;
1432 }
1433 
1434 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1435 	int flags, const char *dev_name, void *data)
1436 {
1437 	return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1438 }
1439 
1440 #ifdef CONFIG_QUOTA
1441 
1442 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1443  * acquiring the locks... As quota files are never truncated and quota code
1444  * itself serializes the operations (and no one else should touch the files)
1445  * we don't have to be afraid of races */
1446 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1447 			       size_t len, loff_t off)
1448 {
1449 	struct inode *inode = sb_dqopt(sb)->files[type];
1450 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1451 	int err = 0;
1452 	int offset = off & (sb->s_blocksize - 1);
1453 	int tocopy;
1454 	size_t toread;
1455 	struct buffer_head tmp_bh;
1456 	struct buffer_head *bh;
1457 	loff_t i_size = i_size_read(inode);
1458 
1459 	if (off > i_size)
1460 		return 0;
1461 	if (off+len > i_size)
1462 		len = i_size-off;
1463 	toread = len;
1464 	while (toread > 0) {
1465 		tocopy = sb->s_blocksize - offset < toread ?
1466 				sb->s_blocksize - offset : toread;
1467 
1468 		tmp_bh.b_state = 0;
1469 		tmp_bh.b_size = sb->s_blocksize;
1470 		err = ext2_get_block(inode, blk, &tmp_bh, 0);
1471 		if (err < 0)
1472 			return err;
1473 		if (!buffer_mapped(&tmp_bh))	/* A hole? */
1474 			memset(data, 0, tocopy);
1475 		else {
1476 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1477 			if (!bh)
1478 				return -EIO;
1479 			memcpy(data, bh->b_data+offset, tocopy);
1480 			brelse(bh);
1481 		}
1482 		offset = 0;
1483 		toread -= tocopy;
1484 		data += tocopy;
1485 		blk++;
1486 	}
1487 	return len;
1488 }
1489 
1490 /* Write to quotafile */
1491 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1492 				const char *data, size_t len, loff_t off)
1493 {
1494 	struct inode *inode = sb_dqopt(sb)->files[type];
1495 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1496 	int err = 0;
1497 	int offset = off & (sb->s_blocksize - 1);
1498 	int tocopy;
1499 	size_t towrite = len;
1500 	struct buffer_head tmp_bh;
1501 	struct buffer_head *bh;
1502 
1503 	while (towrite > 0) {
1504 		tocopy = sb->s_blocksize - offset < towrite ?
1505 				sb->s_blocksize - offset : towrite;
1506 
1507 		tmp_bh.b_state = 0;
1508 		tmp_bh.b_size = sb->s_blocksize;
1509 		err = ext2_get_block(inode, blk, &tmp_bh, 1);
1510 		if (err < 0)
1511 			goto out;
1512 		if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1513 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1514 		else
1515 			bh = sb_getblk(sb, tmp_bh.b_blocknr);
1516 		if (unlikely(!bh)) {
1517 			err = -EIO;
1518 			goto out;
1519 		}
1520 		lock_buffer(bh);
1521 		memcpy(bh->b_data+offset, data, tocopy);
1522 		flush_dcache_page(bh->b_page);
1523 		set_buffer_uptodate(bh);
1524 		mark_buffer_dirty(bh);
1525 		unlock_buffer(bh);
1526 		brelse(bh);
1527 		offset = 0;
1528 		towrite -= tocopy;
1529 		data += tocopy;
1530 		blk++;
1531 	}
1532 out:
1533 	if (len == towrite)
1534 		return err;
1535 	if (inode->i_size < off+len-towrite)
1536 		i_size_write(inode, off+len-towrite);
1537 	inode->i_version++;
1538 	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1539 	mark_inode_dirty(inode);
1540 	return len - towrite;
1541 }
1542 
1543 #endif
1544 
1545 static struct file_system_type ext2_fs_type = {
1546 	.owner		= THIS_MODULE,
1547 	.name		= "ext2",
1548 	.mount		= ext2_mount,
1549 	.kill_sb	= kill_block_super,
1550 	.fs_flags	= FS_REQUIRES_DEV,
1551 };
1552 MODULE_ALIAS_FS("ext2");
1553 
1554 static int __init init_ext2_fs(void)
1555 {
1556 	int err = init_ext2_xattr();
1557 	if (err)
1558 		return err;
1559 	err = init_inodecache();
1560 	if (err)
1561 		goto out1;
1562         err = register_filesystem(&ext2_fs_type);
1563 	if (err)
1564 		goto out;
1565 	return 0;
1566 out:
1567 	destroy_inodecache();
1568 out1:
1569 	exit_ext2_xattr();
1570 	return err;
1571 }
1572 
1573 static void __exit exit_ext2_fs(void)
1574 {
1575 	unregister_filesystem(&ext2_fs_type);
1576 	destroy_inodecache();
1577 	exit_ext2_xattr();
1578 }
1579 
1580 MODULE_AUTHOR("Remy Card and others");
1581 MODULE_DESCRIPTION("Second Extended Filesystem");
1582 MODULE_LICENSE("GPL");
1583 module_init(init_ext2_fs)
1584 module_exit(exit_ext2_fs)
1585