xref: /openbmc/linux/fs/ext2/super.c (revision a1177825)
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 	struct va_format vaf;
50 	va_list args;
51 	struct ext2_sb_info *sbi = EXT2_SB(sb);
52 	struct ext2_super_block *es = sbi->s_es;
53 
54 	if (!(sb->s_flags & MS_RDONLY)) {
55 		spin_lock(&sbi->s_lock);
56 		sbi->s_mount_state |= EXT2_ERROR_FS;
57 		es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
58 		spin_unlock(&sbi->s_lock);
59 		ext2_sync_super(sb, es, 1);
60 	}
61 
62 	va_start(args, fmt);
63 
64 	vaf.fmt = fmt;
65 	vaf.va = &args;
66 
67 	printk(KERN_CRIT "EXT2-fs (%s): error: %s: %pV\n",
68 	       sb->s_id, function, &vaf);
69 
70 	va_end(args);
71 
72 	if (test_opt(sb, ERRORS_PANIC))
73 		panic("EXT2-fs: panic from previous error\n");
74 	if (test_opt(sb, ERRORS_RO)) {
75 		ext2_msg(sb, KERN_CRIT,
76 			     "error: remounting filesystem read-only");
77 		sb->s_flags |= MS_RDONLY;
78 	}
79 }
80 
81 void ext2_msg(struct super_block *sb, const char *prefix,
82 		const char *fmt, ...)
83 {
84 	struct va_format vaf;
85 	va_list args;
86 
87 	va_start(args, fmt);
88 
89 	vaf.fmt = fmt;
90 	vaf.va = &args;
91 
92 	printk("%sEXT2-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
93 
94 	va_end(args);
95 }
96 
97 /*
98  * This must be called with sbi->s_lock held.
99  */
100 void ext2_update_dynamic_rev(struct super_block *sb)
101 {
102 	struct ext2_super_block *es = EXT2_SB(sb)->s_es;
103 
104 	if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
105 		return;
106 
107 	ext2_msg(sb, KERN_WARNING,
108 		     "warning: updating to rev %d because of "
109 		     "new feature flag, running e2fsck is recommended",
110 		     EXT2_DYNAMIC_REV);
111 
112 	es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
113 	es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
114 	es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
115 	/* leave es->s_feature_*compat flags alone */
116 	/* es->s_uuid will be set by e2fsck if empty */
117 
118 	/*
119 	 * The rest of the superblock fields should be zero, and if not it
120 	 * means they are likely already in use, so leave them alone.  We
121 	 * can leave it up to e2fsck to clean up any inconsistencies there.
122 	 */
123 }
124 
125 static void ext2_put_super (struct super_block * sb)
126 {
127 	int db_count;
128 	int i;
129 	struct ext2_sb_info *sbi = EXT2_SB(sb);
130 
131 	dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
132 
133 	ext2_xattr_put_super(sb);
134 	if (!(sb->s_flags & MS_RDONLY)) {
135 		struct ext2_super_block *es = sbi->s_es;
136 
137 		spin_lock(&sbi->s_lock);
138 		es->s_state = cpu_to_le16(sbi->s_mount_state);
139 		spin_unlock(&sbi->s_lock);
140 		ext2_sync_super(sb, es, 1);
141 	}
142 	db_count = sbi->s_gdb_count;
143 	for (i = 0; i < db_count; i++)
144 		if (sbi->s_group_desc[i])
145 			brelse (sbi->s_group_desc[i]);
146 	kfree(sbi->s_group_desc);
147 	kfree(sbi->s_debts);
148 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
149 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
150 	percpu_counter_destroy(&sbi->s_dirs_counter);
151 	brelse (sbi->s_sbh);
152 	sb->s_fs_info = NULL;
153 	kfree(sbi->s_blockgroup_lock);
154 	kfree(sbi);
155 }
156 
157 static struct kmem_cache * ext2_inode_cachep;
158 
159 static struct inode *ext2_alloc_inode(struct super_block *sb)
160 {
161 	struct ext2_inode_info *ei;
162 	ei = (struct ext2_inode_info *)kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
163 	if (!ei)
164 		return NULL;
165 	ei->i_block_alloc_info = NULL;
166 	ei->vfs_inode.i_version = 1;
167 	return &ei->vfs_inode;
168 }
169 
170 static void ext2_i_callback(struct rcu_head *head)
171 {
172 	struct inode *inode = container_of(head, struct inode, i_rcu);
173 	kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
174 }
175 
176 static void ext2_destroy_inode(struct inode *inode)
177 {
178 	call_rcu(&inode->i_rcu, ext2_i_callback);
179 }
180 
181 static void init_once(void *foo)
182 {
183 	struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
184 
185 	rwlock_init(&ei->i_meta_lock);
186 #ifdef CONFIG_EXT2_FS_XATTR
187 	init_rwsem(&ei->xattr_sem);
188 #endif
189 	mutex_init(&ei->truncate_mutex);
190 	inode_init_once(&ei->vfs_inode);
191 }
192 
193 static int init_inodecache(void)
194 {
195 	ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
196 					     sizeof(struct ext2_inode_info),
197 					     0, (SLAB_RECLAIM_ACCOUNT|
198 						SLAB_MEM_SPREAD),
199 					     init_once);
200 	if (ext2_inode_cachep == NULL)
201 		return -ENOMEM;
202 	return 0;
203 }
204 
205 static void destroy_inodecache(void)
206 {
207 	kmem_cache_destroy(ext2_inode_cachep);
208 }
209 
210 static int ext2_show_options(struct seq_file *seq, struct dentry *root)
211 {
212 	struct super_block *sb = root->d_sb;
213 	struct ext2_sb_info *sbi = EXT2_SB(sb);
214 	struct ext2_super_block *es = sbi->s_es;
215 	unsigned long def_mount_opts;
216 
217 	spin_lock(&sbi->s_lock);
218 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
219 
220 	if (sbi->s_sb_block != 1)
221 		seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
222 	if (test_opt(sb, MINIX_DF))
223 		seq_puts(seq, ",minixdf");
224 	if (test_opt(sb, GRPID))
225 		seq_puts(seq, ",grpid");
226 	if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
227 		seq_puts(seq, ",nogrpid");
228 	if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT2_DEF_RESUID)) ||
229 	    le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
230 		seq_printf(seq, ",resuid=%u",
231 				from_kuid_munged(&init_user_ns, sbi->s_resuid));
232 	}
233 	if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT2_DEF_RESGID)) ||
234 	    le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
235 		seq_printf(seq, ",resgid=%u",
236 				from_kgid_munged(&init_user_ns, sbi->s_resgid));
237 	}
238 	if (test_opt(sb, ERRORS_RO)) {
239 		int def_errors = le16_to_cpu(es->s_errors);
240 
241 		if (def_errors == EXT2_ERRORS_PANIC ||
242 		    def_errors == EXT2_ERRORS_CONTINUE) {
243 			seq_puts(seq, ",errors=remount-ro");
244 		}
245 	}
246 	if (test_opt(sb, ERRORS_CONT))
247 		seq_puts(seq, ",errors=continue");
248 	if (test_opt(sb, ERRORS_PANIC))
249 		seq_puts(seq, ",errors=panic");
250 	if (test_opt(sb, NO_UID32))
251 		seq_puts(seq, ",nouid32");
252 	if (test_opt(sb, DEBUG))
253 		seq_puts(seq, ",debug");
254 	if (test_opt(sb, OLDALLOC))
255 		seq_puts(seq, ",oldalloc");
256 
257 #ifdef CONFIG_EXT2_FS_XATTR
258 	if (test_opt(sb, XATTR_USER))
259 		seq_puts(seq, ",user_xattr");
260 	if (!test_opt(sb, XATTR_USER) &&
261 	    (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
262 		seq_puts(seq, ",nouser_xattr");
263 	}
264 #endif
265 
266 #ifdef CONFIG_EXT2_FS_POSIX_ACL
267 	if (test_opt(sb, POSIX_ACL))
268 		seq_puts(seq, ",acl");
269 	if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
270 		seq_puts(seq, ",noacl");
271 #endif
272 
273 	if (test_opt(sb, NOBH))
274 		seq_puts(seq, ",nobh");
275 
276 #if defined(CONFIG_QUOTA)
277 	if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
278 		seq_puts(seq, ",usrquota");
279 
280 	if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
281 		seq_puts(seq, ",grpquota");
282 #endif
283 
284 #if defined(CONFIG_EXT2_FS_XIP)
285 	if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
286 		seq_puts(seq, ",xip");
287 #endif
288 
289 	if (!test_opt(sb, RESERVATION))
290 		seq_puts(seq, ",noreservation");
291 
292 	spin_unlock(&sbi->s_lock);
293 	return 0;
294 }
295 
296 #ifdef CONFIG_QUOTA
297 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
298 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
299 #endif
300 
301 static const struct super_operations ext2_sops = {
302 	.alloc_inode	= ext2_alloc_inode,
303 	.destroy_inode	= ext2_destroy_inode,
304 	.write_inode	= ext2_write_inode,
305 	.evict_inode	= ext2_evict_inode,
306 	.put_super	= ext2_put_super,
307 	.sync_fs	= ext2_sync_fs,
308 	.statfs		= ext2_statfs,
309 	.remount_fs	= ext2_remount,
310 	.show_options	= ext2_show_options,
311 #ifdef CONFIG_QUOTA
312 	.quota_read	= ext2_quota_read,
313 	.quota_write	= ext2_quota_write,
314 #endif
315 };
316 
317 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
318 		u64 ino, u32 generation)
319 {
320 	struct inode *inode;
321 
322 	if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
323 		return ERR_PTR(-ESTALE);
324 	if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
325 		return ERR_PTR(-ESTALE);
326 
327 	/*
328 	 * ext2_iget isn't quite right if the inode is currently unallocated!
329 	 * However ext2_iget currently does appropriate checks to handle stale
330 	 * inodes so everything is OK.
331 	 */
332 	inode = ext2_iget(sb, ino);
333 	if (IS_ERR(inode))
334 		return ERR_CAST(inode);
335 	if (generation && inode->i_generation != generation) {
336 		/* we didn't find the right inode.. */
337 		iput(inode);
338 		return ERR_PTR(-ESTALE);
339 	}
340 	return inode;
341 }
342 
343 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
344 		int fh_len, int fh_type)
345 {
346 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
347 				    ext2_nfs_get_inode);
348 }
349 
350 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
351 		int fh_len, int fh_type)
352 {
353 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
354 				    ext2_nfs_get_inode);
355 }
356 
357 static const struct export_operations ext2_export_ops = {
358 	.fh_to_dentry = ext2_fh_to_dentry,
359 	.fh_to_parent = ext2_fh_to_parent,
360 	.get_parent = ext2_get_parent,
361 };
362 
363 static unsigned long get_sb_block(void **data)
364 {
365 	unsigned long 	sb_block;
366 	char 		*options = (char *) *data;
367 
368 	if (!options || strncmp(options, "sb=", 3) != 0)
369 		return 1;	/* Default location */
370 	options += 3;
371 	sb_block = simple_strtoul(options, &options, 0);
372 	if (*options && *options != ',') {
373 		printk("EXT2-fs: Invalid sb specification: %s\n",
374 		       (char *) *data);
375 		return 1;
376 	}
377 	if (*options == ',')
378 		options++;
379 	*data = (void *) options;
380 	return sb_block;
381 }
382 
383 enum {
384 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
385 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
386 	Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
387 	Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
388 	Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
389 	Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
390 };
391 
392 static const match_table_t tokens = {
393 	{Opt_bsd_df, "bsddf"},
394 	{Opt_minix_df, "minixdf"},
395 	{Opt_grpid, "grpid"},
396 	{Opt_grpid, "bsdgroups"},
397 	{Opt_nogrpid, "nogrpid"},
398 	{Opt_nogrpid, "sysvgroups"},
399 	{Opt_resgid, "resgid=%u"},
400 	{Opt_resuid, "resuid=%u"},
401 	{Opt_sb, "sb=%u"},
402 	{Opt_err_cont, "errors=continue"},
403 	{Opt_err_panic, "errors=panic"},
404 	{Opt_err_ro, "errors=remount-ro"},
405 	{Opt_nouid32, "nouid32"},
406 	{Opt_nocheck, "check=none"},
407 	{Opt_nocheck, "nocheck"},
408 	{Opt_debug, "debug"},
409 	{Opt_oldalloc, "oldalloc"},
410 	{Opt_orlov, "orlov"},
411 	{Opt_nobh, "nobh"},
412 	{Opt_user_xattr, "user_xattr"},
413 	{Opt_nouser_xattr, "nouser_xattr"},
414 	{Opt_acl, "acl"},
415 	{Opt_noacl, "noacl"},
416 	{Opt_xip, "xip"},
417 	{Opt_grpquota, "grpquota"},
418 	{Opt_ignore, "noquota"},
419 	{Opt_quota, "quota"},
420 	{Opt_usrquota, "usrquota"},
421 	{Opt_reservation, "reservation"},
422 	{Opt_noreservation, "noreservation"},
423 	{Opt_err, NULL}
424 };
425 
426 static int parse_options(char *options, struct super_block *sb)
427 {
428 	char *p;
429 	struct ext2_sb_info *sbi = EXT2_SB(sb);
430 	substring_t args[MAX_OPT_ARGS];
431 	int option;
432 	kuid_t uid;
433 	kgid_t gid;
434 
435 	if (!options)
436 		return 1;
437 
438 	while ((p = strsep (&options, ",")) != NULL) {
439 		int token;
440 		if (!*p)
441 			continue;
442 
443 		token = match_token(p, tokens, args);
444 		switch (token) {
445 		case Opt_bsd_df:
446 			clear_opt (sbi->s_mount_opt, MINIX_DF);
447 			break;
448 		case Opt_minix_df:
449 			set_opt (sbi->s_mount_opt, MINIX_DF);
450 			break;
451 		case Opt_grpid:
452 			set_opt (sbi->s_mount_opt, GRPID);
453 			break;
454 		case Opt_nogrpid:
455 			clear_opt (sbi->s_mount_opt, GRPID);
456 			break;
457 		case Opt_resuid:
458 			if (match_int(&args[0], &option))
459 				return 0;
460 			uid = make_kuid(current_user_ns(), option);
461 			if (!uid_valid(uid)) {
462 				ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
463 				return -1;
464 
465 			}
466 			sbi->s_resuid = uid;
467 			break;
468 		case Opt_resgid:
469 			if (match_int(&args[0], &option))
470 				return 0;
471 			gid = make_kgid(current_user_ns(), option);
472 			if (!gid_valid(gid)) {
473 				ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
474 				return -1;
475 			}
476 			sbi->s_resgid = gid;
477 			break;
478 		case Opt_sb:
479 			/* handled by get_sb_block() instead of here */
480 			/* *sb_block = match_int(&args[0]); */
481 			break;
482 		case Opt_err_panic:
483 			clear_opt (sbi->s_mount_opt, ERRORS_CONT);
484 			clear_opt (sbi->s_mount_opt, ERRORS_RO);
485 			set_opt (sbi->s_mount_opt, ERRORS_PANIC);
486 			break;
487 		case Opt_err_ro:
488 			clear_opt (sbi->s_mount_opt, ERRORS_CONT);
489 			clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
490 			set_opt (sbi->s_mount_opt, ERRORS_RO);
491 			break;
492 		case Opt_err_cont:
493 			clear_opt (sbi->s_mount_opt, ERRORS_RO);
494 			clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
495 			set_opt (sbi->s_mount_opt, ERRORS_CONT);
496 			break;
497 		case Opt_nouid32:
498 			set_opt (sbi->s_mount_opt, NO_UID32);
499 			break;
500 		case Opt_nocheck:
501 			clear_opt (sbi->s_mount_opt, CHECK);
502 			break;
503 		case Opt_debug:
504 			set_opt (sbi->s_mount_opt, DEBUG);
505 			break;
506 		case Opt_oldalloc:
507 			set_opt (sbi->s_mount_opt, OLDALLOC);
508 			break;
509 		case Opt_orlov:
510 			clear_opt (sbi->s_mount_opt, OLDALLOC);
511 			break;
512 		case Opt_nobh:
513 			set_opt (sbi->s_mount_opt, NOBH);
514 			break;
515 #ifdef CONFIG_EXT2_FS_XATTR
516 		case Opt_user_xattr:
517 			set_opt (sbi->s_mount_opt, XATTR_USER);
518 			break;
519 		case Opt_nouser_xattr:
520 			clear_opt (sbi->s_mount_opt, XATTR_USER);
521 			break;
522 #else
523 		case Opt_user_xattr:
524 		case Opt_nouser_xattr:
525 			ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
526 				"not supported");
527 			break;
528 #endif
529 #ifdef CONFIG_EXT2_FS_POSIX_ACL
530 		case Opt_acl:
531 			set_opt(sbi->s_mount_opt, POSIX_ACL);
532 			break;
533 		case Opt_noacl:
534 			clear_opt(sbi->s_mount_opt, POSIX_ACL);
535 			break;
536 #else
537 		case Opt_acl:
538 		case Opt_noacl:
539 			ext2_msg(sb, KERN_INFO,
540 				"(no)acl options not supported");
541 			break;
542 #endif
543 		case Opt_xip:
544 #ifdef CONFIG_EXT2_FS_XIP
545 			set_opt (sbi->s_mount_opt, XIP);
546 #else
547 			ext2_msg(sb, KERN_INFO, "xip option not supported");
548 #endif
549 			break;
550 
551 #if defined(CONFIG_QUOTA)
552 		case Opt_quota:
553 		case Opt_usrquota:
554 			set_opt(sbi->s_mount_opt, USRQUOTA);
555 			break;
556 
557 		case Opt_grpquota:
558 			set_opt(sbi->s_mount_opt, GRPQUOTA);
559 			break;
560 #else
561 		case Opt_quota:
562 		case Opt_usrquota:
563 		case Opt_grpquota:
564 			ext2_msg(sb, KERN_INFO,
565 				"quota operations not supported");
566 			break;
567 #endif
568 
569 		case Opt_reservation:
570 			set_opt(sbi->s_mount_opt, RESERVATION);
571 			ext2_msg(sb, KERN_INFO, "reservations ON");
572 			break;
573 		case Opt_noreservation:
574 			clear_opt(sbi->s_mount_opt, RESERVATION);
575 			ext2_msg(sb, KERN_INFO, "reservations OFF");
576 			break;
577 		case Opt_ignore:
578 			break;
579 		default:
580 			return 0;
581 		}
582 	}
583 	return 1;
584 }
585 
586 static int ext2_setup_super (struct super_block * sb,
587 			      struct ext2_super_block * es,
588 			      int read_only)
589 {
590 	int res = 0;
591 	struct ext2_sb_info *sbi = EXT2_SB(sb);
592 
593 	if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
594 		ext2_msg(sb, KERN_ERR,
595 			"error: revision level too high, "
596 			"forcing read-only mode");
597 		res = MS_RDONLY;
598 	}
599 	if (read_only)
600 		return res;
601 	if (!(sbi->s_mount_state & EXT2_VALID_FS))
602 		ext2_msg(sb, KERN_WARNING,
603 			"warning: mounting unchecked fs, "
604 			"running e2fsck is recommended");
605 	else if ((sbi->s_mount_state & EXT2_ERROR_FS))
606 		ext2_msg(sb, KERN_WARNING,
607 			"warning: mounting fs with errors, "
608 			"running e2fsck is recommended");
609 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
610 		 le16_to_cpu(es->s_mnt_count) >=
611 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
612 		ext2_msg(sb, KERN_WARNING,
613 			"warning: maximal mount count reached, "
614 			"running e2fsck is recommended");
615 	else if (le32_to_cpu(es->s_checkinterval) &&
616 		(le32_to_cpu(es->s_lastcheck) +
617 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
618 		ext2_msg(sb, KERN_WARNING,
619 			"warning: checktime reached, "
620 			"running e2fsck is recommended");
621 	if (!le16_to_cpu(es->s_max_mnt_count))
622 		es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
623 	le16_add_cpu(&es->s_mnt_count, 1);
624 	if (test_opt (sb, DEBUG))
625 		ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
626 			"bpg=%lu, ipg=%lu, mo=%04lx]",
627 			EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
628 			sbi->s_frag_size,
629 			sbi->s_groups_count,
630 			EXT2_BLOCKS_PER_GROUP(sb),
631 			EXT2_INODES_PER_GROUP(sb),
632 			sbi->s_mount_opt);
633 	return res;
634 }
635 
636 static int ext2_check_descriptors(struct super_block *sb)
637 {
638 	int i;
639 	struct ext2_sb_info *sbi = EXT2_SB(sb);
640 
641 	ext2_debug ("Checking group descriptors");
642 
643 	for (i = 0; i < sbi->s_groups_count; i++) {
644 		struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
645 		ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
646 		ext2_fsblk_t last_block;
647 
648 		if (i == sbi->s_groups_count - 1)
649 			last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
650 		else
651 			last_block = first_block +
652 				(EXT2_BLOCKS_PER_GROUP(sb) - 1);
653 
654 		if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
655 		    le32_to_cpu(gdp->bg_block_bitmap) > last_block)
656 		{
657 			ext2_error (sb, "ext2_check_descriptors",
658 				    "Block bitmap for group %d"
659 				    " not in group (block %lu)!",
660 				    i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
661 			return 0;
662 		}
663 		if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
664 		    le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
665 		{
666 			ext2_error (sb, "ext2_check_descriptors",
667 				    "Inode bitmap for group %d"
668 				    " not in group (block %lu)!",
669 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
670 			return 0;
671 		}
672 		if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
673 		    le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
674 		    last_block)
675 		{
676 			ext2_error (sb, "ext2_check_descriptors",
677 				    "Inode table for group %d"
678 				    " not in group (block %lu)!",
679 				    i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
680 			return 0;
681 		}
682 	}
683 	return 1;
684 }
685 
686 /*
687  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
688  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
689  * We need to be 1 filesystem block less than the 2^32 sector limit.
690  */
691 static loff_t ext2_max_size(int bits)
692 {
693 	loff_t res = EXT2_NDIR_BLOCKS;
694 	int meta_blocks;
695 	loff_t upper_limit;
696 
697 	/* This is calculated to be the largest file size for a
698 	 * dense, file such that the total number of
699 	 * sectors in the file, including data and all indirect blocks,
700 	 * does not exceed 2^32 -1
701 	 * __u32 i_blocks representing the total number of
702 	 * 512 bytes blocks of the file
703 	 */
704 	upper_limit = (1LL << 32) - 1;
705 
706 	/* total blocks in file system block size */
707 	upper_limit >>= (bits - 9);
708 
709 
710 	/* indirect blocks */
711 	meta_blocks = 1;
712 	/* double indirect blocks */
713 	meta_blocks += 1 + (1LL << (bits-2));
714 	/* tripple indirect blocks */
715 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
716 
717 	upper_limit -= meta_blocks;
718 	upper_limit <<= bits;
719 
720 	res += 1LL << (bits-2);
721 	res += 1LL << (2*(bits-2));
722 	res += 1LL << (3*(bits-2));
723 	res <<= bits;
724 	if (res > upper_limit)
725 		res = upper_limit;
726 
727 	if (res > MAX_LFS_FILESIZE)
728 		res = MAX_LFS_FILESIZE;
729 
730 	return res;
731 }
732 
733 static unsigned long descriptor_loc(struct super_block *sb,
734 				    unsigned long logic_sb_block,
735 				    int nr)
736 {
737 	struct ext2_sb_info *sbi = EXT2_SB(sb);
738 	unsigned long bg, first_meta_bg;
739 	int has_super = 0;
740 
741 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
742 
743 	if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
744 	    nr < first_meta_bg)
745 		return (logic_sb_block + nr + 1);
746 	bg = sbi->s_desc_per_block * nr;
747 	if (ext2_bg_has_super(sb, bg))
748 		has_super = 1;
749 
750 	return ext2_group_first_block_no(sb, bg) + has_super;
751 }
752 
753 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
754 {
755 	struct buffer_head * bh;
756 	struct ext2_sb_info * sbi;
757 	struct ext2_super_block * es;
758 	struct inode *root;
759 	unsigned long block;
760 	unsigned long sb_block = get_sb_block(&data);
761 	unsigned long logic_sb_block;
762 	unsigned long offset = 0;
763 	unsigned long def_mount_opts;
764 	long ret = -EINVAL;
765 	int blocksize = BLOCK_SIZE;
766 	int db_count;
767 	int i, j;
768 	__le32 features;
769 	int err;
770 
771 	err = -ENOMEM;
772 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
773 	if (!sbi)
774 		goto failed_unlock;
775 
776 	sbi->s_blockgroup_lock =
777 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
778 	if (!sbi->s_blockgroup_lock) {
779 		kfree(sbi);
780 		goto failed_unlock;
781 	}
782 	sb->s_fs_info = sbi;
783 	sbi->s_sb_block = sb_block;
784 
785 	spin_lock_init(&sbi->s_lock);
786 
787 	/*
788 	 * See what the current blocksize for the device is, and
789 	 * use that as the blocksize.  Otherwise (or if the blocksize
790 	 * is smaller than the default) use the default.
791 	 * This is important for devices that have a hardware
792 	 * sectorsize that is larger than the default.
793 	 */
794 	blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
795 	if (!blocksize) {
796 		ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
797 		goto failed_sbi;
798 	}
799 
800 	/*
801 	 * If the superblock doesn't start on a hardware sector boundary,
802 	 * calculate the offset.
803 	 */
804 	if (blocksize != BLOCK_SIZE) {
805 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
806 		offset = (sb_block*BLOCK_SIZE) % blocksize;
807 	} else {
808 		logic_sb_block = sb_block;
809 	}
810 
811 	if (!(bh = sb_bread(sb, logic_sb_block))) {
812 		ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
813 		goto failed_sbi;
814 	}
815 	/*
816 	 * Note: s_es must be initialized as soon as possible because
817 	 *       some ext2 macro-instructions depend on its value
818 	 */
819 	es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
820 	sbi->s_es = es;
821 	sb->s_magic = le16_to_cpu(es->s_magic);
822 
823 	if (sb->s_magic != EXT2_SUPER_MAGIC)
824 		goto cantfind_ext2;
825 
826 	/* Set defaults before we parse the mount options */
827 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
828 	if (def_mount_opts & EXT2_DEFM_DEBUG)
829 		set_opt(sbi->s_mount_opt, DEBUG);
830 	if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
831 		set_opt(sbi->s_mount_opt, GRPID);
832 	if (def_mount_opts & EXT2_DEFM_UID16)
833 		set_opt(sbi->s_mount_opt, NO_UID32);
834 #ifdef CONFIG_EXT2_FS_XATTR
835 	if (def_mount_opts & EXT2_DEFM_XATTR_USER)
836 		set_opt(sbi->s_mount_opt, XATTR_USER);
837 #endif
838 #ifdef CONFIG_EXT2_FS_POSIX_ACL
839 	if (def_mount_opts & EXT2_DEFM_ACL)
840 		set_opt(sbi->s_mount_opt, POSIX_ACL);
841 #endif
842 
843 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
844 		set_opt(sbi->s_mount_opt, ERRORS_PANIC);
845 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
846 		set_opt(sbi->s_mount_opt, ERRORS_CONT);
847 	else
848 		set_opt(sbi->s_mount_opt, ERRORS_RO);
849 
850 	sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
851 	sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
852 
853 	set_opt(sbi->s_mount_opt, RESERVATION);
854 
855 	if (!parse_options((char *) data, sb))
856 		goto failed_mount;
857 
858 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
859 		((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
860 		 MS_POSIXACL : 0);
861 
862 	ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
863 				    EXT2_MOUNT_XIP if not */
864 
865 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
866 	    (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
867 	     EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
868 	     EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
869 		ext2_msg(sb, KERN_WARNING,
870 			"warning: feature flags set on rev 0 fs, "
871 			"running e2fsck is recommended");
872 	/*
873 	 * Check feature flags regardless of the revision level, since we
874 	 * previously didn't change the revision level when setting the flags,
875 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
876 	 */
877 	features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
878 	if (features) {
879 		ext2_msg(sb, KERN_ERR,	"error: couldn't mount because of "
880 		       "unsupported optional features (%x)",
881 			le32_to_cpu(features));
882 		goto failed_mount;
883 	}
884 	if (!(sb->s_flags & MS_RDONLY) &&
885 	    (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
886 		ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
887 		       "unsupported optional features (%x)",
888 		       le32_to_cpu(features));
889 		goto failed_mount;
890 	}
891 
892 	blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
893 
894 	if (ext2_use_xip(sb) && blocksize != PAGE_SIZE) {
895 		if (!silent)
896 			ext2_msg(sb, KERN_ERR,
897 				"error: unsupported blocksize for xip");
898 		goto failed_mount;
899 	}
900 
901 	/* If the blocksize doesn't match, re-read the thing.. */
902 	if (sb->s_blocksize != blocksize) {
903 		brelse(bh);
904 
905 		if (!sb_set_blocksize(sb, blocksize)) {
906 			ext2_msg(sb, KERN_ERR,
907 				"error: bad blocksize %d", blocksize);
908 			goto failed_sbi;
909 		}
910 
911 		logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
912 		offset = (sb_block*BLOCK_SIZE) % blocksize;
913 		bh = sb_bread(sb, logic_sb_block);
914 		if(!bh) {
915 			ext2_msg(sb, KERN_ERR, "error: couldn't read"
916 				"superblock on 2nd try");
917 			goto failed_sbi;
918 		}
919 		es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
920 		sbi->s_es = es;
921 		if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
922 			ext2_msg(sb, KERN_ERR, "error: magic mismatch");
923 			goto failed_mount;
924 		}
925 	}
926 
927 	sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
928 	sb->s_max_links = EXT2_LINK_MAX;
929 
930 	if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
931 		sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
932 		sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
933 	} else {
934 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
935 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
936 		if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
937 		    !is_power_of_2(sbi->s_inode_size) ||
938 		    (sbi->s_inode_size > blocksize)) {
939 			ext2_msg(sb, KERN_ERR,
940 				"error: unsupported inode size: %d",
941 				sbi->s_inode_size);
942 			goto failed_mount;
943 		}
944 	}
945 
946 	sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
947 				   le32_to_cpu(es->s_log_frag_size);
948 	if (sbi->s_frag_size == 0)
949 		goto cantfind_ext2;
950 	sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
951 
952 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
953 	sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
954 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
955 
956 	if (EXT2_INODE_SIZE(sb) == 0)
957 		goto cantfind_ext2;
958 	sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
959 	if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
960 		goto cantfind_ext2;
961 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
962 					sbi->s_inodes_per_block;
963 	sbi->s_desc_per_block = sb->s_blocksize /
964 					sizeof (struct ext2_group_desc);
965 	sbi->s_sbh = bh;
966 	sbi->s_mount_state = le16_to_cpu(es->s_state);
967 	sbi->s_addr_per_block_bits =
968 		ilog2 (EXT2_ADDR_PER_BLOCK(sb));
969 	sbi->s_desc_per_block_bits =
970 		ilog2 (EXT2_DESC_PER_BLOCK(sb));
971 
972 	if (sb->s_magic != EXT2_SUPER_MAGIC)
973 		goto cantfind_ext2;
974 
975 	if (sb->s_blocksize != bh->b_size) {
976 		if (!silent)
977 			ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
978 		goto failed_mount;
979 	}
980 
981 	if (sb->s_blocksize != sbi->s_frag_size) {
982 		ext2_msg(sb, KERN_ERR,
983 			"error: fragsize %lu != blocksize %lu"
984 			"(not supported yet)",
985 			sbi->s_frag_size, sb->s_blocksize);
986 		goto failed_mount;
987 	}
988 
989 	if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
990 		ext2_msg(sb, KERN_ERR,
991 			"error: #blocks per group too big: %lu",
992 			sbi->s_blocks_per_group);
993 		goto failed_mount;
994 	}
995 	if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
996 		ext2_msg(sb, KERN_ERR,
997 			"error: #fragments per group too big: %lu",
998 			sbi->s_frags_per_group);
999 		goto failed_mount;
1000 	}
1001 	if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1002 		ext2_msg(sb, KERN_ERR,
1003 			"error: #inodes per group too big: %lu",
1004 			sbi->s_inodes_per_group);
1005 		goto failed_mount;
1006 	}
1007 
1008 	if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1009 		goto cantfind_ext2;
1010  	sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1011  				le32_to_cpu(es->s_first_data_block) - 1)
1012  					/ EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1013 	db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1014 		   EXT2_DESC_PER_BLOCK(sb);
1015 	sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
1016 	if (sbi->s_group_desc == NULL) {
1017 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1018 		goto failed_mount;
1019 	}
1020 	bgl_lock_init(sbi->s_blockgroup_lock);
1021 	sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1022 	if (!sbi->s_debts) {
1023 		ext2_msg(sb, KERN_ERR, "error: not enough memory");
1024 		goto failed_mount_group_desc;
1025 	}
1026 	for (i = 0; i < db_count; i++) {
1027 		block = descriptor_loc(sb, logic_sb_block, i);
1028 		sbi->s_group_desc[i] = sb_bread(sb, block);
1029 		if (!sbi->s_group_desc[i]) {
1030 			for (j = 0; j < i; j++)
1031 				brelse (sbi->s_group_desc[j]);
1032 			ext2_msg(sb, KERN_ERR,
1033 				"error: unable to read group descriptors");
1034 			goto failed_mount_group_desc;
1035 		}
1036 	}
1037 	if (!ext2_check_descriptors (sb)) {
1038 		ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1039 		goto failed_mount2;
1040 	}
1041 	sbi->s_gdb_count = db_count;
1042 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1043 	spin_lock_init(&sbi->s_next_gen_lock);
1044 
1045 	/* per fileystem reservation list head & lock */
1046 	spin_lock_init(&sbi->s_rsv_window_lock);
1047 	sbi->s_rsv_window_root = RB_ROOT;
1048 	/*
1049 	 * Add a single, static dummy reservation to the start of the
1050 	 * reservation window list --- it gives us a placeholder for
1051 	 * append-at-start-of-list which makes the allocation logic
1052 	 * _much_ simpler.
1053 	 */
1054 	sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1055 	sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1056 	sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1057 	sbi->s_rsv_window_head.rsv_goal_size = 0;
1058 	ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1059 
1060 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
1061 				ext2_count_free_blocks(sb));
1062 	if (!err) {
1063 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
1064 				ext2_count_free_inodes(sb));
1065 	}
1066 	if (!err) {
1067 		err = percpu_counter_init(&sbi->s_dirs_counter,
1068 				ext2_count_dirs(sb));
1069 	}
1070 	if (err) {
1071 		ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1072 		goto failed_mount3;
1073 	}
1074 	/*
1075 	 * set up enough so that it can read an inode
1076 	 */
1077 	sb->s_op = &ext2_sops;
1078 	sb->s_export_op = &ext2_export_ops;
1079 	sb->s_xattr = ext2_xattr_handlers;
1080 
1081 #ifdef CONFIG_QUOTA
1082 	sb->dq_op = &dquot_operations;
1083 	sb->s_qcop = &dquot_quotactl_ops;
1084 #endif
1085 
1086 	root = ext2_iget(sb, EXT2_ROOT_INO);
1087 	if (IS_ERR(root)) {
1088 		ret = PTR_ERR(root);
1089 		goto failed_mount3;
1090 	}
1091 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1092 		iput(root);
1093 		ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1094 		goto failed_mount3;
1095 	}
1096 
1097 	sb->s_root = d_make_root(root);
1098 	if (!sb->s_root) {
1099 		ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1100 		ret = -ENOMEM;
1101 		goto failed_mount3;
1102 	}
1103 	if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1104 		ext2_msg(sb, KERN_WARNING,
1105 			"warning: mounting ext3 filesystem as ext2");
1106 	if (ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY))
1107 		sb->s_flags |= MS_RDONLY;
1108 	ext2_write_super(sb);
1109 	return 0;
1110 
1111 cantfind_ext2:
1112 	if (!silent)
1113 		ext2_msg(sb, KERN_ERR,
1114 			"error: can't find an ext2 filesystem on dev %s.",
1115 			sb->s_id);
1116 	goto failed_mount;
1117 failed_mount3:
1118 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
1119 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
1120 	percpu_counter_destroy(&sbi->s_dirs_counter);
1121 failed_mount2:
1122 	for (i = 0; i < db_count; i++)
1123 		brelse(sbi->s_group_desc[i]);
1124 failed_mount_group_desc:
1125 	kfree(sbi->s_group_desc);
1126 	kfree(sbi->s_debts);
1127 failed_mount:
1128 	brelse(bh);
1129 failed_sbi:
1130 	sb->s_fs_info = NULL;
1131 	kfree(sbi->s_blockgroup_lock);
1132 	kfree(sbi);
1133 failed_unlock:
1134 	return ret;
1135 }
1136 
1137 static void ext2_clear_super_error(struct super_block *sb)
1138 {
1139 	struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1140 
1141 	if (buffer_write_io_error(sbh)) {
1142 		/*
1143 		 * Oh, dear.  A previous attempt to write the
1144 		 * superblock failed.  This could happen because the
1145 		 * USB device was yanked out.  Or it could happen to
1146 		 * be a transient write error and maybe the block will
1147 		 * be remapped.  Nothing we can do but to retry the
1148 		 * write and hope for the best.
1149 		 */
1150 		ext2_msg(sb, KERN_ERR,
1151 		       "previous I/O error to superblock detected\n");
1152 		clear_buffer_write_io_error(sbh);
1153 		set_buffer_uptodate(sbh);
1154 	}
1155 }
1156 
1157 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1158 			    int wait)
1159 {
1160 	ext2_clear_super_error(sb);
1161 	spin_lock(&EXT2_SB(sb)->s_lock);
1162 	es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1163 	es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1164 	es->s_wtime = cpu_to_le32(get_seconds());
1165 	/* unlock before we do IO */
1166 	spin_unlock(&EXT2_SB(sb)->s_lock);
1167 	mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1168 	if (wait)
1169 		sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1170 }
1171 
1172 /*
1173  * In the second extended file system, it is not necessary to
1174  * write the super block since we use a mapping of the
1175  * disk super block in a buffer.
1176  *
1177  * However, this function is still used to set the fs valid
1178  * flags to 0.  We need to set this flag to 0 since the fs
1179  * may have been checked while mounted and e2fsck may have
1180  * set s_state to EXT2_VALID_FS after some corrections.
1181  */
1182 static int ext2_sync_fs(struct super_block *sb, int wait)
1183 {
1184 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1185 	struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1186 
1187 	/*
1188 	 * Write quota structures to quota file, sync_blockdev() will write
1189 	 * them to disk later
1190 	 */
1191 	dquot_writeback_dquots(sb, -1);
1192 
1193 	spin_lock(&sbi->s_lock);
1194 	if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1195 		ext2_debug("setting valid to 0\n");
1196 		es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1197 	}
1198 	spin_unlock(&sbi->s_lock);
1199 	ext2_sync_super(sb, es, wait);
1200 	return 0;
1201 }
1202 
1203 
1204 void ext2_write_super(struct super_block *sb)
1205 {
1206 	if (!(sb->s_flags & MS_RDONLY))
1207 		ext2_sync_fs(sb, 1);
1208 }
1209 
1210 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1211 {
1212 	struct ext2_sb_info * sbi = EXT2_SB(sb);
1213 	struct ext2_super_block * es;
1214 	unsigned long old_mount_opt = sbi->s_mount_opt;
1215 	struct ext2_mount_options old_opts;
1216 	unsigned long old_sb_flags;
1217 	int err;
1218 
1219 	spin_lock(&sbi->s_lock);
1220 
1221 	/* Store the old options */
1222 	old_sb_flags = sb->s_flags;
1223 	old_opts.s_mount_opt = sbi->s_mount_opt;
1224 	old_opts.s_resuid = sbi->s_resuid;
1225 	old_opts.s_resgid = sbi->s_resgid;
1226 
1227 	/*
1228 	 * Allow the "check" option to be passed as a remount option.
1229 	 */
1230 	if (!parse_options(data, sb)) {
1231 		err = -EINVAL;
1232 		goto restore_opts;
1233 	}
1234 
1235 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1236 		((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1237 
1238 	ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1239 				    EXT2_MOUNT_XIP if not */
1240 
1241 	if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
1242 		ext2_msg(sb, KERN_WARNING,
1243 			"warning: unsupported blocksize for xip");
1244 		err = -EINVAL;
1245 		goto restore_opts;
1246 	}
1247 
1248 	es = sbi->s_es;
1249 	if ((sbi->s_mount_opt ^ old_mount_opt) & EXT2_MOUNT_XIP) {
1250 		ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1251 			 "xip flag with busy inodes while remounting");
1252 		sbi->s_mount_opt &= ~EXT2_MOUNT_XIP;
1253 		sbi->s_mount_opt |= old_mount_opt & EXT2_MOUNT_XIP;
1254 	}
1255 	if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) {
1256 		spin_unlock(&sbi->s_lock);
1257 		return 0;
1258 	}
1259 	if (*flags & MS_RDONLY) {
1260 		if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1261 		    !(sbi->s_mount_state & EXT2_VALID_FS)) {
1262 			spin_unlock(&sbi->s_lock);
1263 			return 0;
1264 		}
1265 
1266 		/*
1267 		 * OK, we are remounting a valid rw partition rdonly, so set
1268 		 * the rdonly flag and then mark the partition as valid again.
1269 		 */
1270 		es->s_state = cpu_to_le16(sbi->s_mount_state);
1271 		es->s_mtime = cpu_to_le32(get_seconds());
1272 		spin_unlock(&sbi->s_lock);
1273 
1274 		err = dquot_suspend(sb, -1);
1275 		if (err < 0) {
1276 			spin_lock(&sbi->s_lock);
1277 			goto restore_opts;
1278 		}
1279 
1280 		ext2_sync_super(sb, es, 1);
1281 	} else {
1282 		__le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1283 					       ~EXT2_FEATURE_RO_COMPAT_SUPP);
1284 		if (ret) {
1285 			ext2_msg(sb, KERN_WARNING,
1286 				"warning: couldn't remount RDWR because of "
1287 				"unsupported optional features (%x).",
1288 				le32_to_cpu(ret));
1289 			err = -EROFS;
1290 			goto restore_opts;
1291 		}
1292 		/*
1293 		 * Mounting a RDONLY partition read-write, so reread and
1294 		 * store the current valid flag.  (It may have been changed
1295 		 * by e2fsck since we originally mounted the partition.)
1296 		 */
1297 		sbi->s_mount_state = le16_to_cpu(es->s_state);
1298 		if (!ext2_setup_super (sb, es, 0))
1299 			sb->s_flags &= ~MS_RDONLY;
1300 		spin_unlock(&sbi->s_lock);
1301 
1302 		ext2_write_super(sb);
1303 
1304 		dquot_resume(sb, -1);
1305 	}
1306 
1307 	return 0;
1308 restore_opts:
1309 	sbi->s_mount_opt = old_opts.s_mount_opt;
1310 	sbi->s_resuid = old_opts.s_resuid;
1311 	sbi->s_resgid = old_opts.s_resgid;
1312 	sb->s_flags = old_sb_flags;
1313 	spin_unlock(&sbi->s_lock);
1314 	return err;
1315 }
1316 
1317 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1318 {
1319 	struct super_block *sb = dentry->d_sb;
1320 	struct ext2_sb_info *sbi = EXT2_SB(sb);
1321 	struct ext2_super_block *es = sbi->s_es;
1322 	u64 fsid;
1323 
1324 	spin_lock(&sbi->s_lock);
1325 
1326 	if (test_opt (sb, MINIX_DF))
1327 		sbi->s_overhead_last = 0;
1328 	else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1329 		unsigned long i, overhead = 0;
1330 		smp_rmb();
1331 
1332 		/*
1333 		 * Compute the overhead (FS structures). This is constant
1334 		 * for a given filesystem unless the number of block groups
1335 		 * changes so we cache the previous value until it does.
1336 		 */
1337 
1338 		/*
1339 		 * All of the blocks before first_data_block are
1340 		 * overhead
1341 		 */
1342 		overhead = le32_to_cpu(es->s_first_data_block);
1343 
1344 		/*
1345 		 * Add the overhead attributed to the superblock and
1346 		 * block group descriptors.  If the sparse superblocks
1347 		 * feature is turned on, then not all groups have this.
1348 		 */
1349 		for (i = 0; i < sbi->s_groups_count; i++)
1350 			overhead += ext2_bg_has_super(sb, i) +
1351 				ext2_bg_num_gdb(sb, i);
1352 
1353 		/*
1354 		 * Every block group has an inode bitmap, a block
1355 		 * bitmap, and an inode table.
1356 		 */
1357 		overhead += (sbi->s_groups_count *
1358 			     (2 + sbi->s_itb_per_group));
1359 		sbi->s_overhead_last = overhead;
1360 		smp_wmb();
1361 		sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1362 	}
1363 
1364 	buf->f_type = EXT2_SUPER_MAGIC;
1365 	buf->f_bsize = sb->s_blocksize;
1366 	buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1367 	buf->f_bfree = ext2_count_free_blocks(sb);
1368 	es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1369 	buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1370 	if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1371 		buf->f_bavail = 0;
1372 	buf->f_files = le32_to_cpu(es->s_inodes_count);
1373 	buf->f_ffree = ext2_count_free_inodes(sb);
1374 	es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1375 	buf->f_namelen = EXT2_NAME_LEN;
1376 	fsid = le64_to_cpup((void *)es->s_uuid) ^
1377 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1378 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1379 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1380 	spin_unlock(&sbi->s_lock);
1381 	return 0;
1382 }
1383 
1384 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1385 	int flags, const char *dev_name, void *data)
1386 {
1387 	return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1388 }
1389 
1390 #ifdef CONFIG_QUOTA
1391 
1392 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1393  * acquiring the locks... As quota files are never truncated and quota code
1394  * itself serializes the operations (and no one else should touch the files)
1395  * we don't have to be afraid of races */
1396 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1397 			       size_t len, loff_t off)
1398 {
1399 	struct inode *inode = sb_dqopt(sb)->files[type];
1400 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1401 	int err = 0;
1402 	int offset = off & (sb->s_blocksize - 1);
1403 	int tocopy;
1404 	size_t toread;
1405 	struct buffer_head tmp_bh;
1406 	struct buffer_head *bh;
1407 	loff_t i_size = i_size_read(inode);
1408 
1409 	if (off > i_size)
1410 		return 0;
1411 	if (off+len > i_size)
1412 		len = i_size-off;
1413 	toread = len;
1414 	while (toread > 0) {
1415 		tocopy = sb->s_blocksize - offset < toread ?
1416 				sb->s_blocksize - offset : toread;
1417 
1418 		tmp_bh.b_state = 0;
1419 		tmp_bh.b_size = sb->s_blocksize;
1420 		err = ext2_get_block(inode, blk, &tmp_bh, 0);
1421 		if (err < 0)
1422 			return err;
1423 		if (!buffer_mapped(&tmp_bh))	/* A hole? */
1424 			memset(data, 0, tocopy);
1425 		else {
1426 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1427 			if (!bh)
1428 				return -EIO;
1429 			memcpy(data, bh->b_data+offset, tocopy);
1430 			brelse(bh);
1431 		}
1432 		offset = 0;
1433 		toread -= tocopy;
1434 		data += tocopy;
1435 		blk++;
1436 	}
1437 	return len;
1438 }
1439 
1440 /* Write to quotafile */
1441 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1442 				const char *data, size_t len, loff_t off)
1443 {
1444 	struct inode *inode = sb_dqopt(sb)->files[type];
1445 	sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1446 	int err = 0;
1447 	int offset = off & (sb->s_blocksize - 1);
1448 	int tocopy;
1449 	size_t towrite = len;
1450 	struct buffer_head tmp_bh;
1451 	struct buffer_head *bh;
1452 
1453 	while (towrite > 0) {
1454 		tocopy = sb->s_blocksize - offset < towrite ?
1455 				sb->s_blocksize - offset : towrite;
1456 
1457 		tmp_bh.b_state = 0;
1458 		err = ext2_get_block(inode, blk, &tmp_bh, 1);
1459 		if (err < 0)
1460 			goto out;
1461 		if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1462 			bh = sb_bread(sb, tmp_bh.b_blocknr);
1463 		else
1464 			bh = sb_getblk(sb, tmp_bh.b_blocknr);
1465 		if (!bh) {
1466 			err = -EIO;
1467 			goto out;
1468 		}
1469 		lock_buffer(bh);
1470 		memcpy(bh->b_data+offset, data, tocopy);
1471 		flush_dcache_page(bh->b_page);
1472 		set_buffer_uptodate(bh);
1473 		mark_buffer_dirty(bh);
1474 		unlock_buffer(bh);
1475 		brelse(bh);
1476 		offset = 0;
1477 		towrite -= tocopy;
1478 		data += tocopy;
1479 		blk++;
1480 	}
1481 out:
1482 	if (len == towrite)
1483 		return err;
1484 	if (inode->i_size < off+len-towrite)
1485 		i_size_write(inode, off+len-towrite);
1486 	inode->i_version++;
1487 	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1488 	mark_inode_dirty(inode);
1489 	return len - towrite;
1490 }
1491 
1492 #endif
1493 
1494 static struct file_system_type ext2_fs_type = {
1495 	.owner		= THIS_MODULE,
1496 	.name		= "ext2",
1497 	.mount		= ext2_mount,
1498 	.kill_sb	= kill_block_super,
1499 	.fs_flags	= FS_REQUIRES_DEV,
1500 };
1501 
1502 static int __init init_ext2_fs(void)
1503 {
1504 	int err = init_ext2_xattr();
1505 	if (err)
1506 		return err;
1507 	err = init_inodecache();
1508 	if (err)
1509 		goto out1;
1510         err = register_filesystem(&ext2_fs_type);
1511 	if (err)
1512 		goto out;
1513 	return 0;
1514 out:
1515 	destroy_inodecache();
1516 out1:
1517 	exit_ext2_xattr();
1518 	return err;
1519 }
1520 
1521 static void __exit exit_ext2_fs(void)
1522 {
1523 	unregister_filesystem(&ext2_fs_type);
1524 	destroy_inodecache();
1525 	exit_ext2_xattr();
1526 }
1527 
1528 MODULE_AUTHOR("Remy Card and others");
1529 MODULE_DESCRIPTION("Second Extended Filesystem");
1530 MODULE_LICENSE("GPL");
1531 module_init(init_ext2_fs)
1532 module_exit(exit_ext2_fs)
1533