xref: /linux/fs/ext2/super.c (revision 0dd9ac63ce26ec87b080ca9c3e6efed33c23ace6)
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 #include "xip.h"
39 
40 static void ext2_sync_super(struct super_block *sb,
41 			    struct ext2_super_block *es, int wait);
42 static int ext2_remount (struct super_block * sb, int * flags, char * data);
43 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
44 static int ext2_sync_fs(struct super_block *sb, int wait);
45 
46 void ext2_error (struct super_block * sb, const char * function,
47 		 const char * fmt, ...)
48 {
49 	va_list args;
50 	struct ext2_sb_info *sbi = EXT2_SB(sb);
51 	struct ext2_super_block *es = sbi->s_es;
52 
53 	if (!(sb->s_flags & MS_RDONLY)) {
54 		spin_lock(&sbi->s_lock);
55 		sbi->s_mount_state |= EXT2_ERROR_FS;
56 		es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
57 		spin_unlock(&sbi->s_lock);
58 		ext2_sync_super(sb, es, 1);
59 	}
60 
61 	va_start(args, fmt);
62 	printk(KERN_CRIT "EXT2-fs (%s): error: %s: ", sb->s_id, function);
63 	vprintk(fmt, args);
64 	printk("\n");
65 	va_end(args);
66 
67 	if (test_opt(sb, ERRORS_PANIC))
68 		panic("EXT2-fs: panic from previous error\n");
69 	if (test_opt(sb, ERRORS_RO)) {
70 		ext2_msg(sb, KERN_CRIT,
71 			     "error: remounting filesystem read-only");
72 		sb->s_flags |= MS_RDONLY;
73 	}
74 }
75 
76 void ext2_msg(struct super_block *sb, const char *prefix,
77 		const char *fmt, ...)
78 {
79 	va_list args;
80 
81 	va_start(args, fmt);
82 	printk("%sEXT2-fs (%s): ", prefix, sb->s_id);
83 	vprintk(fmt, args);
84 	printk("\n");
85 	va_end(args);
86 }
87 
88 /*
89  * This must be called with sbi->s_lock held.
90  */
91 void ext2_update_dynamic_rev(struct super_block *sb)
92 {
93 	struct ext2_super_block *es = EXT2_SB(sb)->s_es;
94 
95 	if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
96 		return;
97 
98 	ext2_msg(sb, KERN_WARNING,
99 		     "warning: updating to rev %d because of "
100 		     "new feature flag, running e2fsck is recommended",
101 		     EXT2_DYNAMIC_REV);
102 
103 	es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
104 	es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
105 	es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
106 	/* leave es->s_feature_*compat flags alone */
107 	/* es->s_uuid will be set by e2fsck if empty */
108 
109 	/*
110 	 * The rest of the superblock fields should be zero, and if not it
111 	 * means they are likely already in use, so leave them alone.  We
112 	 * can leave it up to e2fsck to clean up any inconsistencies there.
113 	 */
114 }
115 
116 static void ext2_put_super (struct super_block * sb)
117 {
118 	int db_count;
119 	int i;
120 	struct ext2_sb_info *sbi = EXT2_SB(sb);
121 
122 	dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
123 
124 	if (sb->s_dirt)
125 		ext2_write_super(sb);
126 
127 	ext2_xattr_put_super(sb);
128 	if (!(sb->s_flags & MS_RDONLY)) {
129 		struct ext2_super_block *es = sbi->s_es;
130 
131 		spin_lock(&sbi->s_lock);
132 		es->s_state = cpu_to_le16(sbi->s_mount_state);
133 		spin_unlock(&sbi->s_lock);
134 		ext2_sync_super(sb, es, 1);
135 	}
136 	db_count = sbi->s_gdb_count;
137 	for (i = 0; i < db_count; i++)
138 		if (sbi->s_group_desc[i])
139 			brelse (sbi->s_group_desc[i]);
140 	kfree(sbi->s_group_desc);
141 	kfree(sbi->s_debts);
142 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
143 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
144 	percpu_counter_destroy(&sbi->s_dirs_counter);
145 	brelse (sbi->s_sbh);
146 	sb->s_fs_info = NULL;
147 	kfree(sbi->s_blockgroup_lock);
148 	kfree(sbi);
149 }
150 
151 static struct kmem_cache * ext2_inode_cachep;
152 
153 static struct inode *ext2_alloc_inode(struct super_block *sb)
154 {
155 	struct ext2_inode_info *ei;
156 	ei = (struct ext2_inode_info *)kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
157 	if (!ei)
158 		return NULL;
159 	ei->i_block_alloc_info = NULL;
160 	ei->vfs_inode.i_version = 1;
161 	return &ei->vfs_inode;
162 }
163 
164 static void ext2_destroy_inode(struct inode *inode)
165 {
166 	kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
167 }
168 
169 static void init_once(void *foo)
170 {
171 	struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
172 
173 	rwlock_init(&ei->i_meta_lock);
174 #ifdef CONFIG_EXT2_FS_XATTR
175 	init_rwsem(&ei->xattr_sem);
176 #endif
177 	mutex_init(&ei->truncate_mutex);
178 	inode_init_once(&ei->vfs_inode);
179 }
180 
181 static int init_inodecache(void)
182 {
183 	ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
184 					     sizeof(struct ext2_inode_info),
185 					     0, (SLAB_RECLAIM_ACCOUNT|
186 						SLAB_MEM_SPREAD),
187 					     init_once);
188 	if (ext2_inode_cachep == NULL)
189 		return -ENOMEM;
190 	return 0;
191 }
192 
193 static void destroy_inodecache(void)
194 {
195 	kmem_cache_destroy(ext2_inode_cachep);
196 }
197 
198 static void ext2_clear_inode(struct inode *inode)
199 {
200 	struct ext2_block_alloc_info *rsv = EXT2_I(inode)->i_block_alloc_info;
201 
202 	dquot_drop(inode);
203 	ext2_discard_reservation(inode);
204 	EXT2_I(inode)->i_block_alloc_info = NULL;
205 	if (unlikely(rsv))
206 		kfree(rsv);
207 }
208 
209 static int ext2_show_options(struct seq_file *seq, struct vfsmount *vfs)
210 {
211 	struct super_block *sb = vfs->mnt_sb;
212 	struct ext2_sb_info *sbi = EXT2_SB(sb);
213 	struct ext2_super_block *es = sbi->s_es;
214 	unsigned long def_mount_opts;
215 
216 	spin_lock(&sbi->s_lock);
217 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
218 
219 	if (sbi->s_sb_block != 1)
220 		seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
221 	if (test_opt(sb, MINIX_DF))
222 		seq_puts(seq, ",minixdf");
223 	if (test_opt(sb, GRPID))
224 		seq_puts(seq, ",grpid");
225 	if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
226 		seq_puts(seq, ",nogrpid");
227 	if (sbi->s_resuid != EXT2_DEF_RESUID ||
228 	    le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
229 		seq_printf(seq, ",resuid=%u", sbi->s_resuid);
230 	}
231 	if (sbi->s_resgid != EXT2_DEF_RESGID ||
232 	    le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
233 		seq_printf(seq, ",resgid=%u", sbi->s_resgid);
234 	}
235 	if (test_opt(sb, ERRORS_RO)) {
236 		int def_errors = le16_to_cpu(es->s_errors);
237 
238 		if (def_errors == EXT2_ERRORS_PANIC ||
239 		    def_errors == EXT2_ERRORS_CONTINUE) {
240 			seq_puts(seq, ",errors=remount-ro");
241 		}
242 	}
243 	if (test_opt(sb, ERRORS_CONT))
244 		seq_puts(seq, ",errors=continue");
245 	if (test_opt(sb, ERRORS_PANIC))
246 		seq_puts(seq, ",errors=panic");
247 	if (test_opt(sb, NO_UID32))
248 		seq_puts(seq, ",nouid32");
249 	if (test_opt(sb, DEBUG))
250 		seq_puts(seq, ",debug");
251 	if (test_opt(sb, OLDALLOC))
252 		seq_puts(seq, ",oldalloc");
253 
254 #ifdef CONFIG_EXT2_FS_XATTR
255 	if (test_opt(sb, XATTR_USER))
256 		seq_puts(seq, ",user_xattr");
257 	if (!test_opt(sb, XATTR_USER) &&
258 	    (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
259 		seq_puts(seq, ",nouser_xattr");
260 	}
261 #endif
262 
263 #ifdef CONFIG_EXT2_FS_POSIX_ACL
264 	if (test_opt(sb, POSIX_ACL))
265 		seq_puts(seq, ",acl");
266 	if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
267 		seq_puts(seq, ",noacl");
268 #endif
269 
270 	if (test_opt(sb, NOBH))
271 		seq_puts(seq, ",nobh");
272 
273 #if defined(CONFIG_QUOTA)
274 	if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
275 		seq_puts(seq, ",usrquota");
276 
277 	if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
278 		seq_puts(seq, ",grpquota");
279 #endif
280 
281 #if defined(CONFIG_EXT2_FS_XIP)
282 	if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
283 		seq_puts(seq, ",xip");
284 #endif
285 
286 	if (!test_opt(sb, RESERVATION))
287 		seq_puts(seq, ",noreservation");
288 
289 	spin_unlock(&sbi->s_lock);
290 	return 0;
291 }
292 
293 #ifdef CONFIG_QUOTA
294 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
295 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
296 #endif
297 
298 static const struct super_operations ext2_sops = {
299 	.alloc_inode	= ext2_alloc_inode,
300 	.destroy_inode	= ext2_destroy_inode,
301 	.write_inode	= ext2_write_inode,
302 	.delete_inode	= ext2_delete_inode,
303 	.put_super	= ext2_put_super,
304 	.write_super	= ext2_write_super,
305 	.sync_fs	= ext2_sync_fs,
306 	.statfs		= ext2_statfs,
307 	.remount_fs	= ext2_remount,
308 	.clear_inode	= ext2_clear_inode,
309 	.show_options	= ext2_show_options,
310 #ifdef CONFIG_QUOTA
311 	.quota_read	= ext2_quota_read,
312 	.quota_write	= ext2_quota_write,
313 #endif
314 };
315 
316 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
317 		u64 ino, u32 generation)
318 {
319 	struct inode *inode;
320 
321 	if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
322 		return ERR_PTR(-ESTALE);
323 	if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
324 		return ERR_PTR(-ESTALE);
325 
326 	/* iget isn't really right if the inode is currently unallocated!!
327 	 * ext2_read_inode currently does appropriate checks, but
328 	 * it might be "neater" to call ext2_get_inode first and check
329 	 * if the inode is valid.....
330 	 */
331 	inode = ext2_iget(sb, ino);
332 	if (IS_ERR(inode))
333 		return ERR_CAST(inode);
334 	if (generation && inode->i_generation != generation) {
335 		/* we didn't find the right inode.. */
336 		iput(inode);
337 		return ERR_PTR(-ESTALE);
338 	}
339 	return inode;
340 }
341 
342 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
343 		int fh_len, int fh_type)
344 {
345 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
346 				    ext2_nfs_get_inode);
347 }
348 
349 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
350 		int fh_len, int fh_type)
351 {
352 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
353 				    ext2_nfs_get_inode);
354 }
355 
356 /* Yes, most of these are left as NULL!!
357  * A NULL value implies the default, which works with ext2-like file
358  * systems, but can be improved upon.
359  * Currently only get_parent is required.
360  */
361 static const struct export_operations ext2_export_ops = {
362 	.fh_to_dentry = ext2_fh_to_dentry,
363 	.fh_to_parent = ext2_fh_to_parent,
364 	.get_parent = ext2_get_parent,
365 };
366 
367 static unsigned long get_sb_block(void **data)
368 {
369 	unsigned long 	sb_block;
370 	char 		*options = (char *) *data;
371 
372 	if (!options || strncmp(options, "sb=", 3) != 0)
373 		return 1;	/* Default location */
374 	options += 3;
375 	sb_block = simple_strtoul(options, &options, 0);
376 	if (*options && *options != ',') {
377 		printk("EXT2-fs: Invalid sb specification: %s\n",
378 		       (char *) *data);
379 		return 1;
380 	}
381 	if (*options == ',')
382 		options++;
383 	*data = (void *) options;
384 	return sb_block;
385 }
386 
387 enum {
388 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
389 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
390 	Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
391 	Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
392 	Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
393 	Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
394 };
395 
396 static const match_table_t tokens = {
397 	{Opt_bsd_df, "bsddf"},
398 	{Opt_minix_df, "minixdf"},
399 	{Opt_grpid, "grpid"},
400 	{Opt_grpid, "bsdgroups"},
401 	{Opt_nogrpid, "nogrpid"},
402 	{Opt_nogrpid, "sysvgroups"},
403 	{Opt_resgid, "resgid=%u"},
404 	{Opt_resuid, "resuid=%u"},
405 	{Opt_sb, "sb=%u"},
406 	{Opt_err_cont, "errors=continue"},
407 	{Opt_err_panic, "errors=panic"},
408 	{Opt_err_ro, "errors=remount-ro"},
409 	{Opt_nouid32, "nouid32"},
410 	{Opt_nocheck, "check=none"},
411 	{Opt_nocheck, "nocheck"},
412 	{Opt_debug, "debug"},
413 	{Opt_oldalloc, "oldalloc"},
414 	{Opt_orlov, "orlov"},
415 	{Opt_nobh, "nobh"},
416 	{Opt_user_xattr, "user_xattr"},
417 	{Opt_nouser_xattr, "nouser_xattr"},
418 	{Opt_acl, "acl"},
419 	{Opt_noacl, "noacl"},
420 	{Opt_xip, "xip"},
421 	{Opt_grpquota, "grpquota"},
422 	{Opt_ignore, "noquota"},
423 	{Opt_quota, "quota"},
424 	{Opt_usrquota, "usrquota"},
425 	{Opt_reservation, "reservation"},
426 	{Opt_noreservation, "noreservation"},
427 	{Opt_err, NULL}
428 };
429 
430 static int parse_options(char *options, struct super_block *sb)
431 {
432 	char *p;
433 	struct ext2_sb_info *sbi = EXT2_SB(sb);
434 	substring_t args[MAX_OPT_ARGS];
435 	int option;
436 
437 	if (!options)
438 		return 1;
439 
440 	while ((p = strsep (&options, ",")) != NULL) {
441 		int token;
442 		if (!*p)
443 			continue;
444 
445 		token = match_token(p, tokens, args);
446 		switch (token) {
447 		case Opt_bsd_df:
448 			clear_opt (sbi->s_mount_opt, MINIX_DF);
449 			break;
450 		case Opt_minix_df:
451 			set_opt (sbi->s_mount_opt, MINIX_DF);
452 			break;
453 		case Opt_grpid:
454 			set_opt (sbi->s_mount_opt, GRPID);
455 			break;
456 		case Opt_nogrpid:
457 			clear_opt (sbi->s_mount_opt, GRPID);
458 			break;
459 		case Opt_resuid:
460 			if (match_int(&args[0], &option))
461 				return 0;
462 			sbi->s_resuid = option;
463 			break;
464 		case Opt_resgid:
465 			if (match_int(&args[0], &option))
466 				return 0;
467 			sbi->s_resgid = option;
468 			break;
469 		case Opt_sb:
470 			/* handled by get_sb_block() instead of here */
471 			/* *sb_block = match_int(&args[0]); */
472 			break;
473 		case Opt_err_panic:
474 			clear_opt (sbi->s_mount_opt, ERRORS_CONT);
475 			clear_opt (sbi->s_mount_opt, ERRORS_RO);
476 			set_opt (sbi->s_mount_opt, ERRORS_PANIC);
477 			break;
478 		case Opt_err_ro:
479 			clear_opt (sbi->s_mount_opt, ERRORS_CONT);
480 			clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
481 			set_opt (sbi->s_mount_opt, ERRORS_RO);
482 			break;
483 		case Opt_err_cont:
484 			clear_opt (sbi->s_mount_opt, ERRORS_RO);
485 			clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
486 			set_opt (sbi->s_mount_opt, ERRORS_CONT);
487 			break;
488 		case Opt_nouid32:
489 			set_opt (sbi->s_mount_opt, NO_UID32);
490 			break;
491 		case Opt_nocheck:
492 			clear_opt (sbi->s_mount_opt, CHECK);
493 			break;
494 		case Opt_debug:
495 			set_opt (sbi->s_mount_opt, DEBUG);
496 			break;
497 		case Opt_oldalloc:
498 			set_opt (sbi->s_mount_opt, OLDALLOC);
499 			break;
500 		case Opt_orlov:
501 			clear_opt (sbi->s_mount_opt, OLDALLOC);
502 			break;
503 		case Opt_nobh:
504 			set_opt (sbi->s_mount_opt, NOBH);
505 			break;
506 #ifdef CONFIG_EXT2_FS_XATTR
507 		case Opt_user_xattr:
508 			set_opt (sbi->s_mount_opt, XATTR_USER);
509 			break;
510 		case Opt_nouser_xattr:
511 			clear_opt (sbi->s_mount_opt, XATTR_USER);
512 			break;
513 #else
514 		case Opt_user_xattr:
515 		case Opt_nouser_xattr:
516 			ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
517 				"not supported");
518 			break;
519 #endif
520 #ifdef CONFIG_EXT2_FS_POSIX_ACL
521 		case Opt_acl:
522 			set_opt(sbi->s_mount_opt, POSIX_ACL);
523 			break;
524 		case Opt_noacl:
525 			clear_opt(sbi->s_mount_opt, POSIX_ACL);
526 			break;
527 #else
528 		case Opt_acl:
529 		case Opt_noacl:
530 			ext2_msg(sb, KERN_INFO,
531 				"(no)acl options not supported");
532 			break;
533 #endif
534 		case Opt_xip:
535 #ifdef CONFIG_EXT2_FS_XIP
536 			set_opt (sbi->s_mount_opt, XIP);
537 #else
538 			ext2_msg(sb, KERN_INFO, "xip option not supported");
539 #endif
540 			break;
541 
542 #if defined(CONFIG_QUOTA)
543 		case Opt_quota:
544 		case Opt_usrquota:
545 			set_opt(sbi->s_mount_opt, USRQUOTA);
546 			break;
547 
548 		case Opt_grpquota:
549 			set_opt(sbi->s_mount_opt, GRPQUOTA);
550 			break;
551 #else
552 		case Opt_quota:
553 		case Opt_usrquota:
554 		case Opt_grpquota:
555 			ext2_msg(sb, KERN_INFO,
556 				"quota operations not supported");
557 			break;
558 #endif
559 
560 		case Opt_reservation:
561 			set_opt(sbi->s_mount_opt, RESERVATION);
562 			ext2_msg(sb, KERN_INFO, "reservations ON");
563 			break;
564 		case Opt_noreservation:
565 			clear_opt(sbi->s_mount_opt, RESERVATION);
566 			ext2_msg(sb, KERN_INFO, "reservations OFF");
567 			break;
568 		case Opt_ignore:
569 			break;
570 		default:
571 			return 0;
572 		}
573 	}
574 	return 1;
575 }
576 
577 static int ext2_setup_super (struct super_block * sb,
578 			      struct ext2_super_block * es,
579 			      int read_only)
580 {
581 	int res = 0;
582 	struct ext2_sb_info *sbi = EXT2_SB(sb);
583 
584 	if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
585 		ext2_msg(sb, KERN_ERR,
586 			"error: revision level too high, "
587 			"forcing read-only mode");
588 		res = MS_RDONLY;
589 	}
590 	if (read_only)
591 		return res;
592 	if (!(sbi->s_mount_state & EXT2_VALID_FS))
593 		ext2_msg(sb, KERN_WARNING,
594 			"warning: mounting unchecked fs, "
595 			"running e2fsck is recommended");
596 	else if ((sbi->s_mount_state & EXT2_ERROR_FS))
597 		ext2_msg(sb, KERN_WARNING,
598 			"warning: mounting fs with errors, "
599 			"running e2fsck is recommended");
600 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
601 		 le16_to_cpu(es->s_mnt_count) >=
602 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
603 		ext2_msg(sb, KERN_WARNING,
604 			"warning: maximal mount count reached, "
605 			"running e2fsck is recommended");
606 	else if (le32_to_cpu(es->s_checkinterval) &&
607 		(le32_to_cpu(es->s_lastcheck) +
608 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
609 		ext2_msg(sb, KERN_WARNING,
610 			"warning: checktime reached, "
611 			"running e2fsck is recommended");
612 	if (!le16_to_cpu(es->s_max_mnt_count))
613 		es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
614 	le16_add_cpu(&es->s_mnt_count, 1);
615 	if (test_opt (sb, DEBUG))
616 		ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
617 			"bpg=%lu, ipg=%lu, mo=%04lx]",
618 			EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
619 			sbi->s_frag_size,
620 			sbi->s_groups_count,
621 			EXT2_BLOCKS_PER_GROUP(sb),
622 			EXT2_INODES_PER_GROUP(sb),
623 			sbi->s_mount_opt);
624 	return res;
625 }
626 
627 static int ext2_check_descriptors(struct super_block *sb)
628 {
629 	int i;
630 	struct ext2_sb_info *sbi = EXT2_SB(sb);
631 
632 	ext2_debug ("Checking group descriptors");
633 
634 	for (i = 0; i < sbi->s_groups_count; i++) {
635 		struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
636 		ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
637 		ext2_fsblk_t last_block;
638 
639 		if (i == sbi->s_groups_count - 1)
640 			last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
641 		else
642 			last_block = first_block +
643 				(EXT2_BLOCKS_PER_GROUP(sb) - 1);
644 
645 		if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
646 		    le32_to_cpu(gdp->bg_block_bitmap) > last_block)
647 		{
648 			ext2_error (sb, "ext2_check_descriptors",
649 				    "Block bitmap for group %d"
650 				    " not in group (block %lu)!",
651 				    i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
652 			return 0;
653 		}
654 		if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
655 		    le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
656 		{
657 			ext2_error (sb, "ext2_check_descriptors",
658 				    "Inode bitmap for group %d"
659 				    " not in group (block %lu)!",
660 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
661 			return 0;
662 		}
663 		if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
664 		    le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
665 		    last_block)
666 		{
667 			ext2_error (sb, "ext2_check_descriptors",
668 				    "Inode table for group %d"
669 				    " not in group (block %lu)!",
670 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
671 			return 0;
672 		}
673 	}
674 	return 1;
675 }
676 
677 /*
678  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
679  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
680  * We need to be 1 filesystem block less than the 2^32 sector limit.
681  */
682 static loff_t ext2_max_size(int bits)
683 {
684 	loff_t res = EXT2_NDIR_BLOCKS;
685 	int meta_blocks;
686 	loff_t upper_limit;
687 
688 	/* This is calculated to be the largest file size for a
689 	 * dense, file such that the total number of
690 	 * sectors in the file, including data and all indirect blocks,
691 	 * does not exceed 2^32 -1
692 	 * __u32 i_blocks representing the total number of
693 	 * 512 bytes blocks of the file
694 	 */
695 	upper_limit = (1LL << 32) - 1;
696 
697 	/* total blocks in file system block size */
698 	upper_limit >>= (bits - 9);
699 
700 
701 	/* indirect blocks */
702 	meta_blocks = 1;
703 	/* double indirect blocks */
704 	meta_blocks += 1 + (1LL << (bits-2));
705 	/* tripple indirect blocks */
706 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
707 
708 	upper_limit -= meta_blocks;
709 	upper_limit <<= bits;
710 
711 	res += 1LL << (bits-2);
712 	res += 1LL << (2*(bits-2));
713 	res += 1LL << (3*(bits-2));
714 	res <<= bits;
715 	if (res > upper_limit)
716 		res = upper_limit;
717 
718 	if (res > MAX_LFS_FILESIZE)
719 		res = MAX_LFS_FILESIZE;
720 
721 	return res;
722 }
723 
724 static unsigned long descriptor_loc(struct super_block *sb,
725 				    unsigned long logic_sb_block,
726 				    int nr)
727 {
728 	struct ext2_sb_info *sbi = EXT2_SB(sb);
729 	unsigned long bg, first_meta_bg;
730 	int has_super = 0;
731 
732 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
733 
734 	if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
735 	    nr < first_meta_bg)
736 		return (logic_sb_block + nr + 1);
737 	bg = sbi->s_desc_per_block * nr;
738 	if (ext2_bg_has_super(sb, bg))
739 		has_super = 1;
740 
741 	return ext2_group_first_block_no(sb, bg) + has_super;
742 }
743 
744 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
745 {
746 	struct buffer_head * bh;
747 	struct ext2_sb_info * sbi;
748 	struct ext2_super_block * es;
749 	struct inode *root;
750 	unsigned long block;
751 	unsigned long sb_block = get_sb_block(&data);
752 	unsigned long logic_sb_block;
753 	unsigned long offset = 0;
754 	unsigned long def_mount_opts;
755 	long ret = -EINVAL;
756 	int blocksize = BLOCK_SIZE;
757 	int db_count;
758 	int i, j;
759 	__le32 features;
760 	int err;
761 
762 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
763 	if (!sbi)
764 		return -ENOMEM;
765 
766 	sbi->s_blockgroup_lock =
767 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
768 	if (!sbi->s_blockgroup_lock) {
769 		kfree(sbi);
770 		return -ENOMEM;
771 	}
772 	sb->s_fs_info = sbi;
773 	sbi->s_sb_block = sb_block;
774 
775 	spin_lock_init(&sbi->s_lock);
776 
777 	/*
778 	 * See what the current blocksize for the device is, and
779 	 * use that as the blocksize.  Otherwise (or if the blocksize
780 	 * is smaller than the default) use the default.
781 	 * This is important for devices that have a hardware
782 	 * sectorsize that is larger than the default.
783 	 */
784 	blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
785 	if (!blocksize) {
786 		ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
787 		goto failed_sbi;
788 	}
789 
790 	/*
791 	 * If the superblock doesn't start on a hardware sector boundary,
792 	 * calculate the offset.
793 	 */
794 	if (blocksize != BLOCK_SIZE) {
795 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
796 		offset = (sb_block*BLOCK_SIZE) % blocksize;
797 	} else {
798 		logic_sb_block = sb_block;
799 	}
800 
801 	if (!(bh = sb_bread(sb, logic_sb_block))) {
802 		ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
803 		goto failed_sbi;
804 	}
805 	/*
806 	 * Note: s_es must be initialized as soon as possible because
807 	 *       some ext2 macro-instructions depend on its value
808 	 */
809 	es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
810 	sbi->s_es = es;
811 	sb->s_magic = le16_to_cpu(es->s_magic);
812 
813 	if (sb->s_magic != EXT2_SUPER_MAGIC)
814 		goto cantfind_ext2;
815 
816 	/* Set defaults before we parse the mount options */
817 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
818 	if (def_mount_opts & EXT2_DEFM_DEBUG)
819 		set_opt(sbi->s_mount_opt, DEBUG);
820 	if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
821 		set_opt(sbi->s_mount_opt, GRPID);
822 	if (def_mount_opts & EXT2_DEFM_UID16)
823 		set_opt(sbi->s_mount_opt, NO_UID32);
824 #ifdef CONFIG_EXT2_FS_XATTR
825 	if (def_mount_opts & EXT2_DEFM_XATTR_USER)
826 		set_opt(sbi->s_mount_opt, XATTR_USER);
827 #endif
828 #ifdef CONFIG_EXT2_FS_POSIX_ACL
829 	if (def_mount_opts & EXT2_DEFM_ACL)
830 		set_opt(sbi->s_mount_opt, POSIX_ACL);
831 #endif
832 
833 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
834 		set_opt(sbi->s_mount_opt, ERRORS_PANIC);
835 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
836 		set_opt(sbi->s_mount_opt, ERRORS_CONT);
837 	else
838 		set_opt(sbi->s_mount_opt, ERRORS_RO);
839 
840 	sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
841 	sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
842 
843 	set_opt(sbi->s_mount_opt, RESERVATION);
844 
845 	if (!parse_options((char *) data, sb))
846 		goto failed_mount;
847 
848 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
849 		((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
850 		 MS_POSIXACL : 0);
851 
852 	ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
853 				    EXT2_MOUNT_XIP if not */
854 
855 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
856 	    (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
857 	     EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
858 	     EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
859 		ext2_msg(sb, KERN_WARNING,
860 			"warning: feature flags set on rev 0 fs, "
861 			"running e2fsck is recommended");
862 	/*
863 	 * Check feature flags regardless of the revision level, since we
864 	 * previously didn't change the revision level when setting the flags,
865 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
866 	 */
867 	features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
868 	if (features) {
869 		ext2_msg(sb, KERN_ERR,	"error: couldn't mount because of "
870 		       "unsupported optional features (%x)",
871 			le32_to_cpu(features));
872 		goto failed_mount;
873 	}
874 	if (!(sb->s_flags & MS_RDONLY) &&
875 	    (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
876 		ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
877 		       "unsupported optional features (%x)",
878 		       le32_to_cpu(features));
879 		goto failed_mount;
880 	}
881 
882 	blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
883 
884 	if (ext2_use_xip(sb) && blocksize != PAGE_SIZE) {
885 		if (!silent)
886 			ext2_msg(sb, KERN_ERR,
887 				"error: unsupported blocksize for xip");
888 		goto failed_mount;
889 	}
890 
891 	/* If the blocksize doesn't match, re-read the thing.. */
892 	if (sb->s_blocksize != blocksize) {
893 		brelse(bh);
894 
895 		if (!sb_set_blocksize(sb, blocksize)) {
896 			ext2_msg(sb, KERN_ERR, "error: blocksize is too small");
897 			goto failed_sbi;
898 		}
899 
900 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
901 		offset = (sb_block*BLOCK_SIZE) % blocksize;
902 		bh = sb_bread(sb, logic_sb_block);
903 		if(!bh) {
904 			ext2_msg(sb, KERN_ERR, "error: couldn't read"
905 				"superblock on 2nd try");
906 			goto failed_sbi;
907 		}
908 		es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
909 		sbi->s_es = es;
910 		if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
911 			ext2_msg(sb, KERN_ERR, "error: magic mismatch");
912 			goto failed_mount;
913 		}
914 	}
915 
916 	sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
917 
918 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
919 		sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
920 		sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
921 	} else {
922 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
923 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
924 		if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
925 		    !is_power_of_2(sbi->s_inode_size) ||
926 		    (sbi->s_inode_size > blocksize)) {
927 			ext2_msg(sb, KERN_ERR,
928 				"error: unsupported inode size: %d",
929 				sbi->s_inode_size);
930 			goto failed_mount;
931 		}
932 	}
933 
934 	sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
935 				   le32_to_cpu(es->s_log_frag_size);
936 	if (sbi->s_frag_size == 0)
937 		goto cantfind_ext2;
938 	sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
939 
940 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
941 	sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
942 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
943 
944 	if (EXT2_INODE_SIZE(sb) == 0)
945 		goto cantfind_ext2;
946 	sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
947 	if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
948 		goto cantfind_ext2;
949 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
950 					sbi->s_inodes_per_block;
951 	sbi->s_desc_per_block = sb->s_blocksize /
952 					sizeof (struct ext2_group_desc);
953 	sbi->s_sbh = bh;
954 	sbi->s_mount_state = le16_to_cpu(es->s_state);
955 	sbi->s_addr_per_block_bits =
956 		ilog2 (EXT2_ADDR_PER_BLOCK(sb));
957 	sbi->s_desc_per_block_bits =
958 		ilog2 (EXT2_DESC_PER_BLOCK(sb));
959 
960 	if (sb->s_magic != EXT2_SUPER_MAGIC)
961 		goto cantfind_ext2;
962 
963 	if (sb->s_blocksize != bh->b_size) {
964 		if (!silent)
965 			ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
966 		goto failed_mount;
967 	}
968 
969 	if (sb->s_blocksize != sbi->s_frag_size) {
970 		ext2_msg(sb, KERN_ERR,
971 			"error: fragsize %lu != blocksize %lu"
972 			"(not supported yet)",
973 			sbi->s_frag_size, sb->s_blocksize);
974 		goto failed_mount;
975 	}
976 
977 	if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
978 		ext2_msg(sb, KERN_ERR,
979 			"error: #blocks per group too big: %lu",
980 			sbi->s_blocks_per_group);
981 		goto failed_mount;
982 	}
983 	if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
984 		ext2_msg(sb, KERN_ERR,
985 			"error: #fragments per group too big: %lu",
986 			sbi->s_frags_per_group);
987 		goto failed_mount;
988 	}
989 	if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
990 		ext2_msg(sb, KERN_ERR,
991 			"error: #inodes per group too big: %lu",
992 			sbi->s_inodes_per_group);
993 		goto failed_mount;
994 	}
995 
996 	if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
997 		goto cantfind_ext2;
998  	sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
999  				le32_to_cpu(es->s_first_data_block) - 1)
1000  					/ EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1001 	db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1002 		   EXT2_DESC_PER_BLOCK(sb);
1003 	sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
1004 	if (sbi->s_group_desc == NULL) {
1005 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1006 		goto failed_mount;
1007 	}
1008 	bgl_lock_init(sbi->s_blockgroup_lock);
1009 	sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1010 	if (!sbi->s_debts) {
1011 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1012 		goto failed_mount_group_desc;
1013 	}
1014 	for (i = 0; i < db_count; i++) {
1015 		block = descriptor_loc(sb, logic_sb_block, i);
1016 		sbi->s_group_desc[i] = sb_bread(sb, block);
1017 		if (!sbi->s_group_desc[i]) {
1018 			for (j = 0; j < i; j++)
1019 				brelse (sbi->s_group_desc[j]);
1020 			ext2_msg(sb, KERN_ERR,
1021 				"error: unable to read group descriptors");
1022 			goto failed_mount_group_desc;
1023 		}
1024 	}
1025 	if (!ext2_check_descriptors (sb)) {
1026 		ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1027 		goto failed_mount2;
1028 	}
1029 	sbi->s_gdb_count = db_count;
1030 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1031 	spin_lock_init(&sbi->s_next_gen_lock);
1032 
1033 	/* per fileystem reservation list head & lock */
1034 	spin_lock_init(&sbi->s_rsv_window_lock);
1035 	sbi->s_rsv_window_root = RB_ROOT;
1036 	/*
1037 	 * Add a single, static dummy reservation to the start of the
1038 	 * reservation window list --- it gives us a placeholder for
1039 	 * append-at-start-of-list which makes the allocation logic
1040 	 * _much_ simpler.
1041 	 */
1042 	sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1043 	sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1044 	sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1045 	sbi->s_rsv_window_head.rsv_goal_size = 0;
1046 	ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1047 
1048 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
1049 				ext2_count_free_blocks(sb));
1050 	if (!err) {
1051 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
1052 				ext2_count_free_inodes(sb));
1053 	}
1054 	if (!err) {
1055 		err = percpu_counter_init(&sbi->s_dirs_counter,
1056 				ext2_count_dirs(sb));
1057 	}
1058 	if (err) {
1059 		ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1060 		goto failed_mount3;
1061 	}
1062 	/*
1063 	 * set up enough so that it can read an inode
1064 	 */
1065 	sb->s_op = &ext2_sops;
1066 	sb->s_export_op = &ext2_export_ops;
1067 	sb->s_xattr = ext2_xattr_handlers;
1068 
1069 #ifdef CONFIG_QUOTA
1070 	sb->dq_op = &dquot_operations;
1071 	sb->s_qcop = &dquot_quotactl_ops;
1072 #endif
1073 
1074 	root = ext2_iget(sb, EXT2_ROOT_INO);
1075 	if (IS_ERR(root)) {
1076 		ret = PTR_ERR(root);
1077 		goto failed_mount3;
1078 	}
1079 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1080 		iput(root);
1081 		ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1082 		goto failed_mount3;
1083 	}
1084 
1085 	sb->s_root = d_alloc_root(root);
1086 	if (!sb->s_root) {
1087 		iput(root);
1088 		ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1089 		ret = -ENOMEM;
1090 		goto failed_mount3;
1091 	}
1092 	if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1093 		ext2_msg(sb, KERN_WARNING,
1094 			"warning: mounting ext3 filesystem as ext2");
1095 	if (ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY))
1096 		sb->s_flags |= MS_RDONLY;
1097 	ext2_write_super(sb);
1098 	return 0;
1099 
1100 cantfind_ext2:
1101 	if (!silent)
1102 		ext2_msg(sb, KERN_ERR,
1103 			"error: can't find an ext2 filesystem on dev %s.",
1104 			sb->s_id);
1105 	goto failed_mount;
1106 failed_mount3:
1107 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
1108 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
1109 	percpu_counter_destroy(&sbi->s_dirs_counter);
1110 failed_mount2:
1111 	for (i = 0; i < db_count; i++)
1112 		brelse(sbi->s_group_desc[i]);
1113 failed_mount_group_desc:
1114 	kfree(sbi->s_group_desc);
1115 	kfree(sbi->s_debts);
1116 failed_mount:
1117 	brelse(bh);
1118 failed_sbi:
1119 	sb->s_fs_info = NULL;
1120 	kfree(sbi->s_blockgroup_lock);
1121 	kfree(sbi);
1122 	return ret;
1123 }
1124 
1125 static void ext2_clear_super_error(struct super_block *sb)
1126 {
1127 	struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1128 
1129 	if (buffer_write_io_error(sbh)) {
1130 		/*
1131 		 * Oh, dear.  A previous attempt to write the
1132 		 * superblock failed.  This could happen because the
1133 		 * USB device was yanked out.  Or it could happen to
1134 		 * be a transient write error and maybe the block will
1135 		 * be remapped.  Nothing we can do but to retry the
1136 		 * write and hope for the best.
1137 		 */
1138 		ext2_msg(sb, KERN_ERR,
1139 		       "previous I/O error to superblock detected\n");
1140 		clear_buffer_write_io_error(sbh);
1141 		set_buffer_uptodate(sbh);
1142 	}
1143 }
1144 
1145 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1146 			    int wait)
1147 {
1148 	ext2_clear_super_error(sb);
1149 	spin_lock(&EXT2_SB(sb)->s_lock);
1150 	es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1151 	es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1152 	es->s_wtime = cpu_to_le32(get_seconds());
1153 	/* unlock before we do IO */
1154 	spin_unlock(&EXT2_SB(sb)->s_lock);
1155 	mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1156 	if (wait)
1157 		sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1158 	sb->s_dirt = 0;
1159 }
1160 
1161 /*
1162  * In the second extended file system, it is not necessary to
1163  * write the super block since we use a mapping of the
1164  * disk super block in a buffer.
1165  *
1166  * However, this function is still used to set the fs valid
1167  * flags to 0.  We need to set this flag to 0 since the fs
1168  * may have been checked while mounted and e2fsck may have
1169  * set s_state to EXT2_VALID_FS after some corrections.
1170  */
1171 static int ext2_sync_fs(struct super_block *sb, int wait)
1172 {
1173 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1174 	struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1175 
1176 	spin_lock(&sbi->s_lock);
1177 	if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1178 		ext2_debug("setting valid to 0\n");
1179 		es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1180 	}
1181 	spin_unlock(&sbi->s_lock);
1182 	ext2_sync_super(sb, es, wait);
1183 	return 0;
1184 }
1185 
1186 
1187 void ext2_write_super(struct super_block *sb)
1188 {
1189 	if (!(sb->s_flags & MS_RDONLY))
1190 		ext2_sync_fs(sb, 1);
1191 	else
1192 		sb->s_dirt = 0;
1193 }
1194 
1195 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1196 {
1197 	struct ext2_sb_info * sbi = EXT2_SB(sb);
1198 	struct ext2_super_block * es;
1199 	unsigned long old_mount_opt = sbi->s_mount_opt;
1200 	struct ext2_mount_options old_opts;
1201 	unsigned long old_sb_flags;
1202 	int err;
1203 
1204 	spin_lock(&sbi->s_lock);
1205 
1206 	/* Store the old options */
1207 	old_sb_flags = sb->s_flags;
1208 	old_opts.s_mount_opt = sbi->s_mount_opt;
1209 	old_opts.s_resuid = sbi->s_resuid;
1210 	old_opts.s_resgid = sbi->s_resgid;
1211 
1212 	/*
1213 	 * Allow the "check" option to be passed as a remount option.
1214 	 */
1215 	if (!parse_options(data, sb)) {
1216 		err = -EINVAL;
1217 		goto restore_opts;
1218 	}
1219 
1220 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1221 		((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1222 
1223 	ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1224 				    EXT2_MOUNT_XIP if not */
1225 
1226 	if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
1227 		ext2_msg(sb, KERN_WARNING,
1228 			"warning: unsupported blocksize for xip");
1229 		err = -EINVAL;
1230 		goto restore_opts;
1231 	}
1232 
1233 	es = sbi->s_es;
1234 	if (((sbi->s_mount_opt & EXT2_MOUNT_XIP) !=
1235 	    (old_mount_opt & EXT2_MOUNT_XIP)) &&
1236 	    invalidate_inodes(sb)) {
1237 		ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1238 			 "xip flag with busy inodes while remounting");
1239 		sbi->s_mount_opt &= ~EXT2_MOUNT_XIP;
1240 		sbi->s_mount_opt |= old_mount_opt & EXT2_MOUNT_XIP;
1241 	}
1242 	if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) {
1243 		spin_unlock(&sbi->s_lock);
1244 		return 0;
1245 	}
1246 	if (*flags & MS_RDONLY) {
1247 		if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1248 		    !(sbi->s_mount_state & EXT2_VALID_FS)) {
1249 			spin_unlock(&sbi->s_lock);
1250 			return 0;
1251 		}
1252 
1253 		/*
1254 		 * OK, we are remounting a valid rw partition rdonly, so set
1255 		 * the rdonly flag and then mark the partition as valid again.
1256 		 */
1257 		es->s_state = cpu_to_le16(sbi->s_mount_state);
1258 		es->s_mtime = cpu_to_le32(get_seconds());
1259 		spin_unlock(&sbi->s_lock);
1260 
1261 		err = dquot_suspend(sb, -1);
1262 		if (err < 0) {
1263 			spin_lock(&sbi->s_lock);
1264 			goto restore_opts;
1265 		}
1266 
1267 		ext2_sync_super(sb, es, 1);
1268 	} else {
1269 		__le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1270 					       ~EXT2_FEATURE_RO_COMPAT_SUPP);
1271 		if (ret) {
1272 			ext2_msg(sb, KERN_WARNING,
1273 				"warning: couldn't remount RDWR because of "
1274 				"unsupported optional features (%x).",
1275 				le32_to_cpu(ret));
1276 			err = -EROFS;
1277 			goto restore_opts;
1278 		}
1279 		/*
1280 		 * Mounting a RDONLY partition read-write, so reread and
1281 		 * store the current valid flag.  (It may have been changed
1282 		 * by e2fsck since we originally mounted the partition.)
1283 		 */
1284 		sbi->s_mount_state = le16_to_cpu(es->s_state);
1285 		if (!ext2_setup_super (sb, es, 0))
1286 			sb->s_flags &= ~MS_RDONLY;
1287 		spin_unlock(&sbi->s_lock);
1288 
1289 		ext2_write_super(sb);
1290 
1291 		dquot_resume(sb, -1);
1292 	}
1293 
1294 	return 0;
1295 restore_opts:
1296 	sbi->s_mount_opt = old_opts.s_mount_opt;
1297 	sbi->s_resuid = old_opts.s_resuid;
1298 	sbi->s_resgid = old_opts.s_resgid;
1299 	sb->s_flags = old_sb_flags;
1300 	spin_unlock(&sbi->s_lock);
1301 	return err;
1302 }
1303 
1304 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1305 {
1306 	struct super_block *sb = dentry->d_sb;
1307 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1308 	struct ext2_super_block *es = sbi->s_es;
1309 	u64 fsid;
1310 
1311 	spin_lock(&sbi->s_lock);
1312 
1313 	if (test_opt (sb, MINIX_DF))
1314 		sbi->s_overhead_last = 0;
1315 	else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1316 		unsigned long i, overhead = 0;
1317 		smp_rmb();
1318 
1319 		/*
1320 		 * Compute the overhead (FS structures). This is constant
1321 		 * for a given filesystem unless the number of block groups
1322 		 * changes so we cache the previous value until it does.
1323 		 */
1324 
1325 		/*
1326 		 * All of the blocks before first_data_block are
1327 		 * overhead
1328 		 */
1329 		overhead = le32_to_cpu(es->s_first_data_block);
1330 
1331 		/*
1332 		 * Add the overhead attributed to the superblock and
1333 		 * block group descriptors.  If the sparse superblocks
1334 		 * feature is turned on, then not all groups have this.
1335 		 */
1336 		for (i = 0; i < sbi->s_groups_count; i++)
1337 			overhead += ext2_bg_has_super(sb, i) +
1338 				ext2_bg_num_gdb(sb, i);
1339 
1340 		/*
1341 		 * Every block group has an inode bitmap, a block
1342 		 * bitmap, and an inode table.
1343 		 */
1344 		overhead += (sbi->s_groups_count *
1345 			     (2 + sbi->s_itb_per_group));
1346 		sbi->s_overhead_last = overhead;
1347 		smp_wmb();
1348 		sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1349 	}
1350 
1351 	buf->f_type = EXT2_SUPER_MAGIC;
1352 	buf->f_bsize = sb->s_blocksize;
1353 	buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1354 	buf->f_bfree = ext2_count_free_blocks(sb);
1355 	es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1356 	buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1357 	if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1358 		buf->f_bavail = 0;
1359 	buf->f_files = le32_to_cpu(es->s_inodes_count);
1360 	buf->f_ffree = ext2_count_free_inodes(sb);
1361 	es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1362 	buf->f_namelen = EXT2_NAME_LEN;
1363 	fsid = le64_to_cpup((void *)es->s_uuid) ^
1364 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1365 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1366 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1367 	spin_unlock(&sbi->s_lock);
1368 	return 0;
1369 }
1370 
1371 static int ext2_get_sb(struct file_system_type *fs_type,
1372 	int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1373 {
1374 	return get_sb_bdev(fs_type, flags, dev_name, data, ext2_fill_super, mnt);
1375 }
1376 
1377 #ifdef CONFIG_QUOTA
1378 
1379 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1380  * acquiring the locks... As quota files are never truncated and quota code
1381  * itself serializes the operations (and noone else should touch the files)
1382  * we don't have to be afraid of races */
1383 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1384 			       size_t len, loff_t off)
1385 {
1386 	struct inode *inode = sb_dqopt(sb)->files[type];
1387 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1388 	int err = 0;
1389 	int offset = off & (sb->s_blocksize - 1);
1390 	int tocopy;
1391 	size_t toread;
1392 	struct buffer_head tmp_bh;
1393 	struct buffer_head *bh;
1394 	loff_t i_size = i_size_read(inode);
1395 
1396 	if (off > i_size)
1397 		return 0;
1398 	if (off+len > i_size)
1399 		len = i_size-off;
1400 	toread = len;
1401 	while (toread > 0) {
1402 		tocopy = sb->s_blocksize - offset < toread ?
1403 				sb->s_blocksize - offset : toread;
1404 
1405 		tmp_bh.b_state = 0;
1406 		tmp_bh.b_size = sb->s_blocksize;
1407 		err = ext2_get_block(inode, blk, &tmp_bh, 0);
1408 		if (err < 0)
1409 			return err;
1410 		if (!buffer_mapped(&tmp_bh))	/* A hole? */
1411 			memset(data, 0, tocopy);
1412 		else {
1413 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1414 			if (!bh)
1415 				return -EIO;
1416 			memcpy(data, bh->b_data+offset, tocopy);
1417 			brelse(bh);
1418 		}
1419 		offset = 0;
1420 		toread -= tocopy;
1421 		data += tocopy;
1422 		blk++;
1423 	}
1424 	return len;
1425 }
1426 
1427 /* Write to quotafile */
1428 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1429 				const char *data, size_t len, loff_t off)
1430 {
1431 	struct inode *inode = sb_dqopt(sb)->files[type];
1432 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1433 	int err = 0;
1434 	int offset = off & (sb->s_blocksize - 1);
1435 	int tocopy;
1436 	size_t towrite = len;
1437 	struct buffer_head tmp_bh;
1438 	struct buffer_head *bh;
1439 
1440 	mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
1441 	while (towrite > 0) {
1442 		tocopy = sb->s_blocksize - offset < towrite ?
1443 				sb->s_blocksize - offset : towrite;
1444 
1445 		tmp_bh.b_state = 0;
1446 		err = ext2_get_block(inode, blk, &tmp_bh, 1);
1447 		if (err < 0)
1448 			goto out;
1449 		if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1450 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1451 		else
1452 			bh = sb_getblk(sb, tmp_bh.b_blocknr);
1453 		if (!bh) {
1454 			err = -EIO;
1455 			goto out;
1456 		}
1457 		lock_buffer(bh);
1458 		memcpy(bh->b_data+offset, data, tocopy);
1459 		flush_dcache_page(bh->b_page);
1460 		set_buffer_uptodate(bh);
1461 		mark_buffer_dirty(bh);
1462 		unlock_buffer(bh);
1463 		brelse(bh);
1464 		offset = 0;
1465 		towrite -= tocopy;
1466 		data += tocopy;
1467 		blk++;
1468 	}
1469 out:
1470 	if (len == towrite) {
1471 		mutex_unlock(&inode->i_mutex);
1472 		return err;
1473 	}
1474 	if (inode->i_size < off+len-towrite)
1475 		i_size_write(inode, off+len-towrite);
1476 	inode->i_version++;
1477 	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1478 	mark_inode_dirty(inode);
1479 	mutex_unlock(&inode->i_mutex);
1480 	return len - towrite;
1481 }
1482 
1483 #endif
1484 
1485 static struct file_system_type ext2_fs_type = {
1486 	.owner		= THIS_MODULE,
1487 	.name		= "ext2",
1488 	.get_sb		= ext2_get_sb,
1489 	.kill_sb	= kill_block_super,
1490 	.fs_flags	= FS_REQUIRES_DEV,
1491 };
1492 
1493 static int __init init_ext2_fs(void)
1494 {
1495 	int err = init_ext2_xattr();
1496 	if (err)
1497 		return err;
1498 	err = init_inodecache();
1499 	if (err)
1500 		goto out1;
1501         err = register_filesystem(&ext2_fs_type);
1502 	if (err)
1503 		goto out;
1504 	return 0;
1505 out:
1506 	destroy_inodecache();
1507 out1:
1508 	exit_ext2_xattr();
1509 	return err;
1510 }
1511 
1512 static void __exit exit_ext2_fs(void)
1513 {
1514 	unregister_filesystem(&ext2_fs_type);
1515 	destroy_inodecache();
1516 	exit_ext2_xattr();
1517 }
1518 
1519 module_init(init_ext2_fs)
1520 module_exit(exit_ext2_fs)
1521