xref: /openbmc/linux/fs/ext4/super.c (revision b8bb76713ec50df2f11efee386e16f93d51e1076)
1 /*
2  *  linux/fs/ext4/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/time.h>
23 #include <linux/jbd2.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/parser.h>
28 #include <linux/smp_lock.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/marker.h>
40 #include <linux/log2.h>
41 #include <linux/crc16.h>
42 #include <asm/uaccess.h>
43 
44 #include "ext4.h"
45 #include "ext4_jbd2.h"
46 #include "xattr.h"
47 #include "acl.h"
48 #include "namei.h"
49 #include "group.h"
50 
51 struct proc_dir_entry *ext4_proc_root;
52 static struct kset *ext4_kset;
53 
54 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
55 			     unsigned long journal_devnum);
56 static int ext4_commit_super(struct super_block *sb,
57 			      struct ext4_super_block *es, int sync);
58 static void ext4_mark_recovery_complete(struct super_block *sb,
59 					struct ext4_super_block *es);
60 static void ext4_clear_journal_err(struct super_block *sb,
61 				   struct ext4_super_block *es);
62 static int ext4_sync_fs(struct super_block *sb, int wait);
63 static const char *ext4_decode_error(struct super_block *sb, int errno,
64 				     char nbuf[16]);
65 static int ext4_remount(struct super_block *sb, int *flags, char *data);
66 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
67 static int ext4_unfreeze(struct super_block *sb);
68 static void ext4_write_super(struct super_block *sb);
69 static int ext4_freeze(struct super_block *sb);
70 
71 
72 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
73 			       struct ext4_group_desc *bg)
74 {
75 	return le32_to_cpu(bg->bg_block_bitmap_lo) |
76 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
77 		(ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
78 }
79 
80 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
81 			       struct ext4_group_desc *bg)
82 {
83 	return le32_to_cpu(bg->bg_inode_bitmap_lo) |
84 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
85 		(ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
86 }
87 
88 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
89 			      struct ext4_group_desc *bg)
90 {
91 	return le32_to_cpu(bg->bg_inode_table_lo) |
92 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
93 		(ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
94 }
95 
96 __u32 ext4_free_blks_count(struct super_block *sb,
97 			      struct ext4_group_desc *bg)
98 {
99 	return le16_to_cpu(bg->bg_free_blocks_count_lo) |
100 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
101 		(__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
102 }
103 
104 __u32 ext4_free_inodes_count(struct super_block *sb,
105 			      struct ext4_group_desc *bg)
106 {
107 	return le16_to_cpu(bg->bg_free_inodes_count_lo) |
108 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
109 		(__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
110 }
111 
112 __u32 ext4_used_dirs_count(struct super_block *sb,
113 			      struct ext4_group_desc *bg)
114 {
115 	return le16_to_cpu(bg->bg_used_dirs_count_lo) |
116 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
117 		(__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
118 }
119 
120 __u32 ext4_itable_unused_count(struct super_block *sb,
121 			      struct ext4_group_desc *bg)
122 {
123 	return le16_to_cpu(bg->bg_itable_unused_lo) |
124 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
125 		(__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
126 }
127 
128 void ext4_block_bitmap_set(struct super_block *sb,
129 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
130 {
131 	bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
132 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
133 		bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
134 }
135 
136 void ext4_inode_bitmap_set(struct super_block *sb,
137 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
138 {
139 	bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
140 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
141 		bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
142 }
143 
144 void ext4_inode_table_set(struct super_block *sb,
145 			  struct ext4_group_desc *bg, ext4_fsblk_t blk)
146 {
147 	bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
148 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
149 		bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
150 }
151 
152 void ext4_free_blks_set(struct super_block *sb,
153 			  struct ext4_group_desc *bg, __u32 count)
154 {
155 	bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
156 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
157 		bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
158 }
159 
160 void ext4_free_inodes_set(struct super_block *sb,
161 			  struct ext4_group_desc *bg, __u32 count)
162 {
163 	bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
164 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
165 		bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
166 }
167 
168 void ext4_used_dirs_set(struct super_block *sb,
169 			  struct ext4_group_desc *bg, __u32 count)
170 {
171 	bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
172 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
173 		bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
174 }
175 
176 void ext4_itable_unused_set(struct super_block *sb,
177 			  struct ext4_group_desc *bg, __u32 count)
178 {
179 	bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
180 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
181 		bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
182 }
183 
184 /*
185  * Wrappers for jbd2_journal_start/end.
186  *
187  * The only special thing we need to do here is to make sure that all
188  * journal_end calls result in the superblock being marked dirty, so
189  * that sync() will call the filesystem's write_super callback if
190  * appropriate.
191  */
192 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
193 {
194 	journal_t *journal;
195 
196 	if (sb->s_flags & MS_RDONLY)
197 		return ERR_PTR(-EROFS);
198 
199 	/* Special case here: if the journal has aborted behind our
200 	 * backs (eg. EIO in the commit thread), then we still need to
201 	 * take the FS itself readonly cleanly. */
202 	journal = EXT4_SB(sb)->s_journal;
203 	if (journal) {
204 		if (is_journal_aborted(journal)) {
205 			ext4_abort(sb, __func__,
206 				   "Detected aborted journal");
207 			return ERR_PTR(-EROFS);
208 		}
209 		return jbd2_journal_start(journal, nblocks);
210 	}
211 	/*
212 	 * We're not journaling, return the appropriate indication.
213 	 */
214 	current->journal_info = EXT4_NOJOURNAL_HANDLE;
215 	return current->journal_info;
216 }
217 
218 /*
219  * The only special thing we need to do here is to make sure that all
220  * jbd2_journal_stop calls result in the superblock being marked dirty, so
221  * that sync() will call the filesystem's write_super callback if
222  * appropriate.
223  */
224 int __ext4_journal_stop(const char *where, handle_t *handle)
225 {
226 	struct super_block *sb;
227 	int err;
228 	int rc;
229 
230 	if (!ext4_handle_valid(handle)) {
231 		/*
232 		 * Do this here since we don't call jbd2_journal_stop() in
233 		 * no-journal mode.
234 		 */
235 		current->journal_info = NULL;
236 		return 0;
237 	}
238 	sb = handle->h_transaction->t_journal->j_private;
239 	err = handle->h_err;
240 	rc = jbd2_journal_stop(handle);
241 
242 	if (!err)
243 		err = rc;
244 	if (err)
245 		__ext4_std_error(sb, where, err);
246 	return err;
247 }
248 
249 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
250 		struct buffer_head *bh, handle_t *handle, int err)
251 {
252 	char nbuf[16];
253 	const char *errstr = ext4_decode_error(NULL, err, nbuf);
254 
255 	BUG_ON(!ext4_handle_valid(handle));
256 
257 	if (bh)
258 		BUFFER_TRACE(bh, "abort");
259 
260 	if (!handle->h_err)
261 		handle->h_err = err;
262 
263 	if (is_handle_aborted(handle))
264 		return;
265 
266 	printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
267 	       caller, errstr, err_fn);
268 
269 	jbd2_journal_abort_handle(handle);
270 }
271 
272 /* Deal with the reporting of failure conditions on a filesystem such as
273  * inconsistencies detected or read IO failures.
274  *
275  * On ext2, we can store the error state of the filesystem in the
276  * superblock.  That is not possible on ext4, because we may have other
277  * write ordering constraints on the superblock which prevent us from
278  * writing it out straight away; and given that the journal is about to
279  * be aborted, we can't rely on the current, or future, transactions to
280  * write out the superblock safely.
281  *
282  * We'll just use the jbd2_journal_abort() error code to record an error in
283  * the journal instead.  On recovery, the journal will compain about
284  * that error until we've noted it down and cleared it.
285  */
286 
287 static void ext4_handle_error(struct super_block *sb)
288 {
289 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
290 
291 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
292 	es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
293 
294 	if (sb->s_flags & MS_RDONLY)
295 		return;
296 
297 	if (!test_opt(sb, ERRORS_CONT)) {
298 		journal_t *journal = EXT4_SB(sb)->s_journal;
299 
300 		EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
301 		if (journal)
302 			jbd2_journal_abort(journal, -EIO);
303 	}
304 	if (test_opt(sb, ERRORS_RO)) {
305 		printk(KERN_CRIT "Remounting filesystem read-only\n");
306 		sb->s_flags |= MS_RDONLY;
307 	}
308 	ext4_commit_super(sb, es, 1);
309 	if (test_opt(sb, ERRORS_PANIC))
310 		panic("EXT4-fs (device %s): panic forced after error\n",
311 			sb->s_id);
312 }
313 
314 void ext4_error(struct super_block *sb, const char *function,
315 		const char *fmt, ...)
316 {
317 	va_list args;
318 
319 	va_start(args, fmt);
320 	printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
321 	vprintk(fmt, args);
322 	printk("\n");
323 	va_end(args);
324 
325 	ext4_handle_error(sb);
326 }
327 
328 static const char *ext4_decode_error(struct super_block *sb, int errno,
329 				     char nbuf[16])
330 {
331 	char *errstr = NULL;
332 
333 	switch (errno) {
334 	case -EIO:
335 		errstr = "IO failure";
336 		break;
337 	case -ENOMEM:
338 		errstr = "Out of memory";
339 		break;
340 	case -EROFS:
341 		if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
342 			errstr = "Journal has aborted";
343 		else
344 			errstr = "Readonly filesystem";
345 		break;
346 	default:
347 		/* If the caller passed in an extra buffer for unknown
348 		 * errors, textualise them now.  Else we just return
349 		 * NULL. */
350 		if (nbuf) {
351 			/* Check for truncated error codes... */
352 			if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
353 				errstr = nbuf;
354 		}
355 		break;
356 	}
357 
358 	return errstr;
359 }
360 
361 /* __ext4_std_error decodes expected errors from journaling functions
362  * automatically and invokes the appropriate error response.  */
363 
364 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
365 {
366 	char nbuf[16];
367 	const char *errstr;
368 
369 	/* Special case: if the error is EROFS, and we're not already
370 	 * inside a transaction, then there's really no point in logging
371 	 * an error. */
372 	if (errno == -EROFS && journal_current_handle() == NULL &&
373 	    (sb->s_flags & MS_RDONLY))
374 		return;
375 
376 	errstr = ext4_decode_error(sb, errno, nbuf);
377 	printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
378 	       sb->s_id, function, errstr);
379 
380 	ext4_handle_error(sb);
381 }
382 
383 /*
384  * ext4_abort is a much stronger failure handler than ext4_error.  The
385  * abort function may be used to deal with unrecoverable failures such
386  * as journal IO errors or ENOMEM at a critical moment in log management.
387  *
388  * We unconditionally force the filesystem into an ABORT|READONLY state,
389  * unless the error response on the fs has been set to panic in which
390  * case we take the easy way out and panic immediately.
391  */
392 
393 void ext4_abort(struct super_block *sb, const char *function,
394 		const char *fmt, ...)
395 {
396 	va_list args;
397 
398 	printk(KERN_CRIT "ext4_abort called.\n");
399 
400 	va_start(args, fmt);
401 	printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
402 	vprintk(fmt, args);
403 	printk("\n");
404 	va_end(args);
405 
406 	if (test_opt(sb, ERRORS_PANIC))
407 		panic("EXT4-fs panic from previous error\n");
408 
409 	if (sb->s_flags & MS_RDONLY)
410 		return;
411 
412 	printk(KERN_CRIT "Remounting filesystem read-only\n");
413 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
414 	sb->s_flags |= MS_RDONLY;
415 	EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
416 	if (EXT4_SB(sb)->s_journal)
417 		jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
418 }
419 
420 void ext4_warning(struct super_block *sb, const char *function,
421 		  const char *fmt, ...)
422 {
423 	va_list args;
424 
425 	va_start(args, fmt);
426 	printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
427 	       sb->s_id, function);
428 	vprintk(fmt, args);
429 	printk("\n");
430 	va_end(args);
431 }
432 
433 void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
434 				const char *function, const char *fmt, ...)
435 __releases(bitlock)
436 __acquires(bitlock)
437 {
438 	va_list args;
439 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
440 
441 	va_start(args, fmt);
442 	printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
443 	vprintk(fmt, args);
444 	printk("\n");
445 	va_end(args);
446 
447 	if (test_opt(sb, ERRORS_CONT)) {
448 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
449 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
450 		ext4_commit_super(sb, es, 0);
451 		return;
452 	}
453 	ext4_unlock_group(sb, grp);
454 	ext4_handle_error(sb);
455 	/*
456 	 * We only get here in the ERRORS_RO case; relocking the group
457 	 * may be dangerous, but nothing bad will happen since the
458 	 * filesystem will have already been marked read/only and the
459 	 * journal has been aborted.  We return 1 as a hint to callers
460 	 * who might what to use the return value from
461 	 * ext4_grp_locked_error() to distinguish beween the
462 	 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
463 	 * aggressively from the ext4 function in question, with a
464 	 * more appropriate error code.
465 	 */
466 	ext4_lock_group(sb, grp);
467 	return;
468 }
469 
470 
471 void ext4_update_dynamic_rev(struct super_block *sb)
472 {
473 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
474 
475 	if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
476 		return;
477 
478 	ext4_warning(sb, __func__,
479 		     "updating to rev %d because of new feature flag, "
480 		     "running e2fsck is recommended",
481 		     EXT4_DYNAMIC_REV);
482 
483 	es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
484 	es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
485 	es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
486 	/* leave es->s_feature_*compat flags alone */
487 	/* es->s_uuid will be set by e2fsck if empty */
488 
489 	/*
490 	 * The rest of the superblock fields should be zero, and if not it
491 	 * means they are likely already in use, so leave them alone.  We
492 	 * can leave it up to e2fsck to clean up any inconsistencies there.
493 	 */
494 }
495 
496 /*
497  * Open the external journal device
498  */
499 static struct block_device *ext4_blkdev_get(dev_t dev)
500 {
501 	struct block_device *bdev;
502 	char b[BDEVNAME_SIZE];
503 
504 	bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
505 	if (IS_ERR(bdev))
506 		goto fail;
507 	return bdev;
508 
509 fail:
510 	printk(KERN_ERR "EXT4-fs: failed to open journal device %s: %ld\n",
511 			__bdevname(dev, b), PTR_ERR(bdev));
512 	return NULL;
513 }
514 
515 /*
516  * Release the journal device
517  */
518 static int ext4_blkdev_put(struct block_device *bdev)
519 {
520 	bd_release(bdev);
521 	return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
522 }
523 
524 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
525 {
526 	struct block_device *bdev;
527 	int ret = -ENODEV;
528 
529 	bdev = sbi->journal_bdev;
530 	if (bdev) {
531 		ret = ext4_blkdev_put(bdev);
532 		sbi->journal_bdev = NULL;
533 	}
534 	return ret;
535 }
536 
537 static inline struct inode *orphan_list_entry(struct list_head *l)
538 {
539 	return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
540 }
541 
542 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
543 {
544 	struct list_head *l;
545 
546 	printk(KERN_ERR "sb orphan head is %d\n",
547 	       le32_to_cpu(sbi->s_es->s_last_orphan));
548 
549 	printk(KERN_ERR "sb_info orphan list:\n");
550 	list_for_each(l, &sbi->s_orphan) {
551 		struct inode *inode = orphan_list_entry(l);
552 		printk(KERN_ERR "  "
553 		       "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
554 		       inode->i_sb->s_id, inode->i_ino, inode,
555 		       inode->i_mode, inode->i_nlink,
556 		       NEXT_ORPHAN(inode));
557 	}
558 }
559 
560 static void ext4_put_super(struct super_block *sb)
561 {
562 	struct ext4_sb_info *sbi = EXT4_SB(sb);
563 	struct ext4_super_block *es = sbi->s_es;
564 	int i, err;
565 
566 	ext4_mb_release(sb);
567 	ext4_ext_release(sb);
568 	ext4_xattr_put_super(sb);
569 	if (sbi->s_journal) {
570 		err = jbd2_journal_destroy(sbi->s_journal);
571 		sbi->s_journal = NULL;
572 		if (err < 0)
573 			ext4_abort(sb, __func__,
574 				   "Couldn't clean up the journal");
575 	}
576 	if (!(sb->s_flags & MS_RDONLY)) {
577 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
578 		es->s_state = cpu_to_le16(sbi->s_mount_state);
579 		ext4_commit_super(sb, es, 1);
580 	}
581 	if (sbi->s_proc) {
582 		remove_proc_entry(sb->s_id, ext4_proc_root);
583 	}
584 	kobject_del(&sbi->s_kobj);
585 
586 	for (i = 0; i < sbi->s_gdb_count; i++)
587 		brelse(sbi->s_group_desc[i]);
588 	kfree(sbi->s_group_desc);
589 	kfree(sbi->s_flex_groups);
590 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
591 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
592 	percpu_counter_destroy(&sbi->s_dirs_counter);
593 	percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
594 	brelse(sbi->s_sbh);
595 #ifdef CONFIG_QUOTA
596 	for (i = 0; i < MAXQUOTAS; i++)
597 		kfree(sbi->s_qf_names[i]);
598 #endif
599 
600 	/* Debugging code just in case the in-memory inode orphan list
601 	 * isn't empty.  The on-disk one can be non-empty if we've
602 	 * detected an error and taken the fs readonly, but the
603 	 * in-memory list had better be clean by this point. */
604 	if (!list_empty(&sbi->s_orphan))
605 		dump_orphan_list(sb, sbi);
606 	J_ASSERT(list_empty(&sbi->s_orphan));
607 
608 	invalidate_bdev(sb->s_bdev);
609 	if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
610 		/*
611 		 * Invalidate the journal device's buffers.  We don't want them
612 		 * floating about in memory - the physical journal device may
613 		 * hotswapped, and it breaks the `ro-after' testing code.
614 		 */
615 		sync_blockdev(sbi->journal_bdev);
616 		invalidate_bdev(sbi->journal_bdev);
617 		ext4_blkdev_remove(sbi);
618 	}
619 	sb->s_fs_info = NULL;
620 	/*
621 	 * Now that we are completely done shutting down the
622 	 * superblock, we need to actually destroy the kobject.
623 	 */
624 	unlock_kernel();
625 	unlock_super(sb);
626 	kobject_put(&sbi->s_kobj);
627 	wait_for_completion(&sbi->s_kobj_unregister);
628 	lock_super(sb);
629 	lock_kernel();
630 	kfree(sbi->s_blockgroup_lock);
631 	kfree(sbi);
632 	return;
633 }
634 
635 static struct kmem_cache *ext4_inode_cachep;
636 
637 /*
638  * Called inside transaction, so use GFP_NOFS
639  */
640 static struct inode *ext4_alloc_inode(struct super_block *sb)
641 {
642 	struct ext4_inode_info *ei;
643 
644 	ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
645 	if (!ei)
646 		return NULL;
647 #ifdef CONFIG_EXT4_FS_POSIX_ACL
648 	ei->i_acl = EXT4_ACL_NOT_CACHED;
649 	ei->i_default_acl = EXT4_ACL_NOT_CACHED;
650 #endif
651 	ei->vfs_inode.i_version = 1;
652 	ei->vfs_inode.i_data.writeback_index = 0;
653 	memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
654 	INIT_LIST_HEAD(&ei->i_prealloc_list);
655 	spin_lock_init(&ei->i_prealloc_lock);
656 	/*
657 	 * Note:  We can be called before EXT4_SB(sb)->s_journal is set,
658 	 * therefore it can be null here.  Don't check it, just initialize
659 	 * jinode.
660 	 */
661 	jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
662 	ei->i_reserved_data_blocks = 0;
663 	ei->i_reserved_meta_blocks = 0;
664 	ei->i_allocated_meta_blocks = 0;
665 	ei->i_delalloc_reserved_flag = 0;
666 	spin_lock_init(&(ei->i_block_reservation_lock));
667 	return &ei->vfs_inode;
668 }
669 
670 static void ext4_destroy_inode(struct inode *inode)
671 {
672 	if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
673 		printk("EXT4 Inode %p: orphan list check failed!\n",
674 			EXT4_I(inode));
675 		print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
676 				EXT4_I(inode), sizeof(struct ext4_inode_info),
677 				true);
678 		dump_stack();
679 	}
680 	kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
681 }
682 
683 static void init_once(void *foo)
684 {
685 	struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
686 
687 	INIT_LIST_HEAD(&ei->i_orphan);
688 #ifdef CONFIG_EXT4_FS_XATTR
689 	init_rwsem(&ei->xattr_sem);
690 #endif
691 	init_rwsem(&ei->i_data_sem);
692 	inode_init_once(&ei->vfs_inode);
693 }
694 
695 static int init_inodecache(void)
696 {
697 	ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
698 					     sizeof(struct ext4_inode_info),
699 					     0, (SLAB_RECLAIM_ACCOUNT|
700 						SLAB_MEM_SPREAD),
701 					     init_once);
702 	if (ext4_inode_cachep == NULL)
703 		return -ENOMEM;
704 	return 0;
705 }
706 
707 static void destroy_inodecache(void)
708 {
709 	kmem_cache_destroy(ext4_inode_cachep);
710 }
711 
712 static void ext4_clear_inode(struct inode *inode)
713 {
714 #ifdef CONFIG_EXT4_FS_POSIX_ACL
715 	if (EXT4_I(inode)->i_acl &&
716 			EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
717 		posix_acl_release(EXT4_I(inode)->i_acl);
718 		EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
719 	}
720 	if (EXT4_I(inode)->i_default_acl &&
721 			EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
722 		posix_acl_release(EXT4_I(inode)->i_default_acl);
723 		EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
724 	}
725 #endif
726 	ext4_discard_preallocations(inode);
727 	if (EXT4_JOURNAL(inode))
728 		jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
729 				       &EXT4_I(inode)->jinode);
730 }
731 
732 static inline void ext4_show_quota_options(struct seq_file *seq,
733 					   struct super_block *sb)
734 {
735 #if defined(CONFIG_QUOTA)
736 	struct ext4_sb_info *sbi = EXT4_SB(sb);
737 
738 	if (sbi->s_jquota_fmt)
739 		seq_printf(seq, ",jqfmt=%s",
740 		(sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
741 
742 	if (sbi->s_qf_names[USRQUOTA])
743 		seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
744 
745 	if (sbi->s_qf_names[GRPQUOTA])
746 		seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
747 
748 	if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
749 		seq_puts(seq, ",usrquota");
750 
751 	if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
752 		seq_puts(seq, ",grpquota");
753 #endif
754 }
755 
756 /*
757  * Show an option if
758  *  - it's set to a non-default value OR
759  *  - if the per-sb default is different from the global default
760  */
761 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
762 {
763 	int def_errors;
764 	unsigned long def_mount_opts;
765 	struct super_block *sb = vfs->mnt_sb;
766 	struct ext4_sb_info *sbi = EXT4_SB(sb);
767 	struct ext4_super_block *es = sbi->s_es;
768 
769 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
770 	def_errors     = le16_to_cpu(es->s_errors);
771 
772 	if (sbi->s_sb_block != 1)
773 		seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
774 	if (test_opt(sb, MINIX_DF))
775 		seq_puts(seq, ",minixdf");
776 	if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
777 		seq_puts(seq, ",grpid");
778 	if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
779 		seq_puts(seq, ",nogrpid");
780 	if (sbi->s_resuid != EXT4_DEF_RESUID ||
781 	    le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
782 		seq_printf(seq, ",resuid=%u", sbi->s_resuid);
783 	}
784 	if (sbi->s_resgid != EXT4_DEF_RESGID ||
785 	    le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
786 		seq_printf(seq, ",resgid=%u", sbi->s_resgid);
787 	}
788 	if (test_opt(sb, ERRORS_RO)) {
789 		if (def_errors == EXT4_ERRORS_PANIC ||
790 		    def_errors == EXT4_ERRORS_CONTINUE) {
791 			seq_puts(seq, ",errors=remount-ro");
792 		}
793 	}
794 	if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
795 		seq_puts(seq, ",errors=continue");
796 	if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
797 		seq_puts(seq, ",errors=panic");
798 	if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
799 		seq_puts(seq, ",nouid32");
800 	if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
801 		seq_puts(seq, ",debug");
802 	if (test_opt(sb, OLDALLOC))
803 		seq_puts(seq, ",oldalloc");
804 #ifdef CONFIG_EXT4_FS_XATTR
805 	if (test_opt(sb, XATTR_USER) &&
806 		!(def_mount_opts & EXT4_DEFM_XATTR_USER))
807 		seq_puts(seq, ",user_xattr");
808 	if (!test_opt(sb, XATTR_USER) &&
809 	    (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
810 		seq_puts(seq, ",nouser_xattr");
811 	}
812 #endif
813 #ifdef CONFIG_EXT4_FS_POSIX_ACL
814 	if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
815 		seq_puts(seq, ",acl");
816 	if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
817 		seq_puts(seq, ",noacl");
818 #endif
819 	if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
820 		seq_printf(seq, ",commit=%u",
821 			   (unsigned) (sbi->s_commit_interval / HZ));
822 	}
823 	if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
824 		seq_printf(seq, ",min_batch_time=%u",
825 			   (unsigned) sbi->s_min_batch_time);
826 	}
827 	if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
828 		seq_printf(seq, ",max_batch_time=%u",
829 			   (unsigned) sbi->s_min_batch_time);
830 	}
831 
832 	/*
833 	 * We're changing the default of barrier mount option, so
834 	 * let's always display its mount state so it's clear what its
835 	 * status is.
836 	 */
837 	seq_puts(seq, ",barrier=");
838 	seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
839 	if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
840 		seq_puts(seq, ",journal_async_commit");
841 	if (test_opt(sb, NOBH))
842 		seq_puts(seq, ",nobh");
843 	if (test_opt(sb, I_VERSION))
844 		seq_puts(seq, ",i_version");
845 	if (!test_opt(sb, DELALLOC))
846 		seq_puts(seq, ",nodelalloc");
847 
848 
849 	if (sbi->s_stripe)
850 		seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
851 	/*
852 	 * journal mode get enabled in different ways
853 	 * So just print the value even if we didn't specify it
854 	 */
855 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
856 		seq_puts(seq, ",data=journal");
857 	else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
858 		seq_puts(seq, ",data=ordered");
859 	else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
860 		seq_puts(seq, ",data=writeback");
861 
862 	if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
863 		seq_printf(seq, ",inode_readahead_blks=%u",
864 			   sbi->s_inode_readahead_blks);
865 
866 	if (test_opt(sb, DATA_ERR_ABORT))
867 		seq_puts(seq, ",data_err=abort");
868 
869 	if (test_opt(sb, NO_AUTO_DA_ALLOC))
870 		seq_puts(seq, ",noauto_da_alloc");
871 
872 	ext4_show_quota_options(seq, sb);
873 	return 0;
874 }
875 
876 
877 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
878 		u64 ino, u32 generation)
879 {
880 	struct inode *inode;
881 
882 	if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
883 		return ERR_PTR(-ESTALE);
884 	if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
885 		return ERR_PTR(-ESTALE);
886 
887 	/* iget isn't really right if the inode is currently unallocated!!
888 	 *
889 	 * ext4_read_inode will return a bad_inode if the inode had been
890 	 * deleted, so we should be safe.
891 	 *
892 	 * Currently we don't know the generation for parent directory, so
893 	 * a generation of 0 means "accept any"
894 	 */
895 	inode = ext4_iget(sb, ino);
896 	if (IS_ERR(inode))
897 		return ERR_CAST(inode);
898 	if (generation && inode->i_generation != generation) {
899 		iput(inode);
900 		return ERR_PTR(-ESTALE);
901 	}
902 
903 	return inode;
904 }
905 
906 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
907 		int fh_len, int fh_type)
908 {
909 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
910 				    ext4_nfs_get_inode);
911 }
912 
913 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
914 		int fh_len, int fh_type)
915 {
916 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
917 				    ext4_nfs_get_inode);
918 }
919 
920 /*
921  * Try to release metadata pages (indirect blocks, directories) which are
922  * mapped via the block device.  Since these pages could have journal heads
923  * which would prevent try_to_free_buffers() from freeing them, we must use
924  * jbd2 layer's try_to_free_buffers() function to release them.
925  */
926 static int bdev_try_to_free_page(struct super_block *sb, struct page *page, gfp_t wait)
927 {
928 	journal_t *journal = EXT4_SB(sb)->s_journal;
929 
930 	WARN_ON(PageChecked(page));
931 	if (!page_has_buffers(page))
932 		return 0;
933 	if (journal)
934 		return jbd2_journal_try_to_free_buffers(journal, page,
935 							wait & ~__GFP_WAIT);
936 	return try_to_free_buffers(page);
937 }
938 
939 #ifdef CONFIG_QUOTA
940 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
941 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
942 
943 static int ext4_write_dquot(struct dquot *dquot);
944 static int ext4_acquire_dquot(struct dquot *dquot);
945 static int ext4_release_dquot(struct dquot *dquot);
946 static int ext4_mark_dquot_dirty(struct dquot *dquot);
947 static int ext4_write_info(struct super_block *sb, int type);
948 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
949 				char *path, int remount);
950 static int ext4_quota_on_mount(struct super_block *sb, int type);
951 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
952 			       size_t len, loff_t off);
953 static ssize_t ext4_quota_write(struct super_block *sb, int type,
954 				const char *data, size_t len, loff_t off);
955 
956 static struct dquot_operations ext4_quota_operations = {
957 	.initialize	= dquot_initialize,
958 	.drop		= dquot_drop,
959 	.alloc_space	= dquot_alloc_space,
960 	.reserve_space	= dquot_reserve_space,
961 	.claim_space	= dquot_claim_space,
962 	.release_rsv	= dquot_release_reserved_space,
963 	.get_reserved_space = ext4_get_reserved_space,
964 	.alloc_inode	= dquot_alloc_inode,
965 	.free_space	= dquot_free_space,
966 	.free_inode	= dquot_free_inode,
967 	.transfer	= dquot_transfer,
968 	.write_dquot	= ext4_write_dquot,
969 	.acquire_dquot	= ext4_acquire_dquot,
970 	.release_dquot	= ext4_release_dquot,
971 	.mark_dirty	= ext4_mark_dquot_dirty,
972 	.write_info	= ext4_write_info,
973 	.alloc_dquot	= dquot_alloc,
974 	.destroy_dquot	= dquot_destroy,
975 };
976 
977 static struct quotactl_ops ext4_qctl_operations = {
978 	.quota_on	= ext4_quota_on,
979 	.quota_off	= vfs_quota_off,
980 	.quota_sync	= vfs_quota_sync,
981 	.get_info	= vfs_get_dqinfo,
982 	.set_info	= vfs_set_dqinfo,
983 	.get_dqblk	= vfs_get_dqblk,
984 	.set_dqblk	= vfs_set_dqblk
985 };
986 #endif
987 
988 static const struct super_operations ext4_sops = {
989 	.alloc_inode	= ext4_alloc_inode,
990 	.destroy_inode	= ext4_destroy_inode,
991 	.write_inode	= ext4_write_inode,
992 	.dirty_inode	= ext4_dirty_inode,
993 	.delete_inode	= ext4_delete_inode,
994 	.put_super	= ext4_put_super,
995 	.write_super	= ext4_write_super,
996 	.sync_fs	= ext4_sync_fs,
997 	.freeze_fs	= ext4_freeze,
998 	.unfreeze_fs	= ext4_unfreeze,
999 	.statfs		= ext4_statfs,
1000 	.remount_fs	= ext4_remount,
1001 	.clear_inode	= ext4_clear_inode,
1002 	.show_options	= ext4_show_options,
1003 #ifdef CONFIG_QUOTA
1004 	.quota_read	= ext4_quota_read,
1005 	.quota_write	= ext4_quota_write,
1006 #endif
1007 	.bdev_try_to_free_page = bdev_try_to_free_page,
1008 };
1009 
1010 static const struct export_operations ext4_export_ops = {
1011 	.fh_to_dentry = ext4_fh_to_dentry,
1012 	.fh_to_parent = ext4_fh_to_parent,
1013 	.get_parent = ext4_get_parent,
1014 };
1015 
1016 enum {
1017 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1018 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1019 	Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1020 	Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1021 	Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1022 	Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1023 	Opt_journal_update, Opt_journal_dev,
1024 	Opt_journal_checksum, Opt_journal_async_commit,
1025 	Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1026 	Opt_data_err_abort, Opt_data_err_ignore,
1027 	Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1028 	Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1029 	Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1030 	Opt_usrquota, Opt_grpquota, Opt_i_version,
1031 	Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1032 	Opt_inode_readahead_blks, Opt_journal_ioprio
1033 };
1034 
1035 static const match_table_t tokens = {
1036 	{Opt_bsd_df, "bsddf"},
1037 	{Opt_minix_df, "minixdf"},
1038 	{Opt_grpid, "grpid"},
1039 	{Opt_grpid, "bsdgroups"},
1040 	{Opt_nogrpid, "nogrpid"},
1041 	{Opt_nogrpid, "sysvgroups"},
1042 	{Opt_resgid, "resgid=%u"},
1043 	{Opt_resuid, "resuid=%u"},
1044 	{Opt_sb, "sb=%u"},
1045 	{Opt_err_cont, "errors=continue"},
1046 	{Opt_err_panic, "errors=panic"},
1047 	{Opt_err_ro, "errors=remount-ro"},
1048 	{Opt_nouid32, "nouid32"},
1049 	{Opt_debug, "debug"},
1050 	{Opt_oldalloc, "oldalloc"},
1051 	{Opt_orlov, "orlov"},
1052 	{Opt_user_xattr, "user_xattr"},
1053 	{Opt_nouser_xattr, "nouser_xattr"},
1054 	{Opt_acl, "acl"},
1055 	{Opt_noacl, "noacl"},
1056 	{Opt_noload, "noload"},
1057 	{Opt_nobh, "nobh"},
1058 	{Opt_bh, "bh"},
1059 	{Opt_commit, "commit=%u"},
1060 	{Opt_min_batch_time, "min_batch_time=%u"},
1061 	{Opt_max_batch_time, "max_batch_time=%u"},
1062 	{Opt_journal_update, "journal=update"},
1063 	{Opt_journal_dev, "journal_dev=%u"},
1064 	{Opt_journal_checksum, "journal_checksum"},
1065 	{Opt_journal_async_commit, "journal_async_commit"},
1066 	{Opt_abort, "abort"},
1067 	{Opt_data_journal, "data=journal"},
1068 	{Opt_data_ordered, "data=ordered"},
1069 	{Opt_data_writeback, "data=writeback"},
1070 	{Opt_data_err_abort, "data_err=abort"},
1071 	{Opt_data_err_ignore, "data_err=ignore"},
1072 	{Opt_offusrjquota, "usrjquota="},
1073 	{Opt_usrjquota, "usrjquota=%s"},
1074 	{Opt_offgrpjquota, "grpjquota="},
1075 	{Opt_grpjquota, "grpjquota=%s"},
1076 	{Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1077 	{Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1078 	{Opt_grpquota, "grpquota"},
1079 	{Opt_noquota, "noquota"},
1080 	{Opt_quota, "quota"},
1081 	{Opt_usrquota, "usrquota"},
1082 	{Opt_barrier, "barrier=%u"},
1083 	{Opt_barrier, "barrier"},
1084 	{Opt_nobarrier, "nobarrier"},
1085 	{Opt_i_version, "i_version"},
1086 	{Opt_stripe, "stripe=%u"},
1087 	{Opt_resize, "resize"},
1088 	{Opt_delalloc, "delalloc"},
1089 	{Opt_nodelalloc, "nodelalloc"},
1090 	{Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1091 	{Opt_journal_ioprio, "journal_ioprio=%u"},
1092 	{Opt_auto_da_alloc, "auto_da_alloc=%u"},
1093 	{Opt_auto_da_alloc, "auto_da_alloc"},
1094 	{Opt_noauto_da_alloc, "noauto_da_alloc"},
1095 	{Opt_err, NULL},
1096 };
1097 
1098 static ext4_fsblk_t get_sb_block(void **data)
1099 {
1100 	ext4_fsblk_t	sb_block;
1101 	char		*options = (char *) *data;
1102 
1103 	if (!options || strncmp(options, "sb=", 3) != 0)
1104 		return 1;	/* Default location */
1105 	options += 3;
1106 	/*todo: use simple_strtoll with >32bit ext4 */
1107 	sb_block = simple_strtoul(options, &options, 0);
1108 	if (*options && *options != ',') {
1109 		printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1110 		       (char *) *data);
1111 		return 1;
1112 	}
1113 	if (*options == ',')
1114 		options++;
1115 	*data = (void *) options;
1116 	return sb_block;
1117 }
1118 
1119 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1120 
1121 static int parse_options(char *options, struct super_block *sb,
1122 			 unsigned long *journal_devnum,
1123 			 unsigned int *journal_ioprio,
1124 			 ext4_fsblk_t *n_blocks_count, int is_remount)
1125 {
1126 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1127 	char *p;
1128 	substring_t args[MAX_OPT_ARGS];
1129 	int data_opt = 0;
1130 	int option;
1131 #ifdef CONFIG_QUOTA
1132 	int qtype, qfmt;
1133 	char *qname;
1134 #endif
1135 
1136 	if (!options)
1137 		return 1;
1138 
1139 	while ((p = strsep(&options, ",")) != NULL) {
1140 		int token;
1141 		if (!*p)
1142 			continue;
1143 
1144 		token = match_token(p, tokens, args);
1145 		switch (token) {
1146 		case Opt_bsd_df:
1147 			clear_opt(sbi->s_mount_opt, MINIX_DF);
1148 			break;
1149 		case Opt_minix_df:
1150 			set_opt(sbi->s_mount_opt, MINIX_DF);
1151 			break;
1152 		case Opt_grpid:
1153 			set_opt(sbi->s_mount_opt, GRPID);
1154 			break;
1155 		case Opt_nogrpid:
1156 			clear_opt(sbi->s_mount_opt, GRPID);
1157 			break;
1158 		case Opt_resuid:
1159 			if (match_int(&args[0], &option))
1160 				return 0;
1161 			sbi->s_resuid = option;
1162 			break;
1163 		case Opt_resgid:
1164 			if (match_int(&args[0], &option))
1165 				return 0;
1166 			sbi->s_resgid = option;
1167 			break;
1168 		case Opt_sb:
1169 			/* handled by get_sb_block() instead of here */
1170 			/* *sb_block = match_int(&args[0]); */
1171 			break;
1172 		case Opt_err_panic:
1173 			clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1174 			clear_opt(sbi->s_mount_opt, ERRORS_RO);
1175 			set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1176 			break;
1177 		case Opt_err_ro:
1178 			clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1179 			clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1180 			set_opt(sbi->s_mount_opt, ERRORS_RO);
1181 			break;
1182 		case Opt_err_cont:
1183 			clear_opt(sbi->s_mount_opt, ERRORS_RO);
1184 			clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1185 			set_opt(sbi->s_mount_opt, ERRORS_CONT);
1186 			break;
1187 		case Opt_nouid32:
1188 			set_opt(sbi->s_mount_opt, NO_UID32);
1189 			break;
1190 		case Opt_debug:
1191 			set_opt(sbi->s_mount_opt, DEBUG);
1192 			break;
1193 		case Opt_oldalloc:
1194 			set_opt(sbi->s_mount_opt, OLDALLOC);
1195 			break;
1196 		case Opt_orlov:
1197 			clear_opt(sbi->s_mount_opt, OLDALLOC);
1198 			break;
1199 #ifdef CONFIG_EXT4_FS_XATTR
1200 		case Opt_user_xattr:
1201 			set_opt(sbi->s_mount_opt, XATTR_USER);
1202 			break;
1203 		case Opt_nouser_xattr:
1204 			clear_opt(sbi->s_mount_opt, XATTR_USER);
1205 			break;
1206 #else
1207 		case Opt_user_xattr:
1208 		case Opt_nouser_xattr:
1209 			printk(KERN_ERR "EXT4 (no)user_xattr options "
1210 			       "not supported\n");
1211 			break;
1212 #endif
1213 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1214 		case Opt_acl:
1215 			set_opt(sbi->s_mount_opt, POSIX_ACL);
1216 			break;
1217 		case Opt_noacl:
1218 			clear_opt(sbi->s_mount_opt, POSIX_ACL);
1219 			break;
1220 #else
1221 		case Opt_acl:
1222 		case Opt_noacl:
1223 			printk(KERN_ERR "EXT4 (no)acl options "
1224 			       "not supported\n");
1225 			break;
1226 #endif
1227 		case Opt_journal_update:
1228 			/* @@@ FIXME */
1229 			/* Eventually we will want to be able to create
1230 			   a journal file here.  For now, only allow the
1231 			   user to specify an existing inode to be the
1232 			   journal file. */
1233 			if (is_remount) {
1234 				printk(KERN_ERR "EXT4-fs: cannot specify "
1235 				       "journal on remount\n");
1236 				return 0;
1237 			}
1238 			set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1239 			break;
1240 		case Opt_journal_dev:
1241 			if (is_remount) {
1242 				printk(KERN_ERR "EXT4-fs: cannot specify "
1243 				       "journal on remount\n");
1244 				return 0;
1245 			}
1246 			if (match_int(&args[0], &option))
1247 				return 0;
1248 			*journal_devnum = option;
1249 			break;
1250 		case Opt_journal_checksum:
1251 			set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1252 			break;
1253 		case Opt_journal_async_commit:
1254 			set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1255 			set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1256 			break;
1257 		case Opt_noload:
1258 			set_opt(sbi->s_mount_opt, NOLOAD);
1259 			break;
1260 		case Opt_commit:
1261 			if (match_int(&args[0], &option))
1262 				return 0;
1263 			if (option < 0)
1264 				return 0;
1265 			if (option == 0)
1266 				option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1267 			sbi->s_commit_interval = HZ * option;
1268 			break;
1269 		case Opt_max_batch_time:
1270 			if (match_int(&args[0], &option))
1271 				return 0;
1272 			if (option < 0)
1273 				return 0;
1274 			if (option == 0)
1275 				option = EXT4_DEF_MAX_BATCH_TIME;
1276 			sbi->s_max_batch_time = option;
1277 			break;
1278 		case Opt_min_batch_time:
1279 			if (match_int(&args[0], &option))
1280 				return 0;
1281 			if (option < 0)
1282 				return 0;
1283 			sbi->s_min_batch_time = option;
1284 			break;
1285 		case Opt_data_journal:
1286 			data_opt = EXT4_MOUNT_JOURNAL_DATA;
1287 			goto datacheck;
1288 		case Opt_data_ordered:
1289 			data_opt = EXT4_MOUNT_ORDERED_DATA;
1290 			goto datacheck;
1291 		case Opt_data_writeback:
1292 			data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1293 		datacheck:
1294 			if (is_remount) {
1295 				if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1296 						!= data_opt) {
1297 					printk(KERN_ERR
1298 						"EXT4-fs: cannot change data "
1299 						"mode on remount\n");
1300 					return 0;
1301 				}
1302 			} else {
1303 				sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1304 				sbi->s_mount_opt |= data_opt;
1305 			}
1306 			break;
1307 		case Opt_data_err_abort:
1308 			set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1309 			break;
1310 		case Opt_data_err_ignore:
1311 			clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1312 			break;
1313 #ifdef CONFIG_QUOTA
1314 		case Opt_usrjquota:
1315 			qtype = USRQUOTA;
1316 			goto set_qf_name;
1317 		case Opt_grpjquota:
1318 			qtype = GRPQUOTA;
1319 set_qf_name:
1320 			if (sb_any_quota_loaded(sb) &&
1321 			    !sbi->s_qf_names[qtype]) {
1322 				printk(KERN_ERR
1323 				       "EXT4-fs: Cannot change journaled "
1324 				       "quota options when quota turned on.\n");
1325 				return 0;
1326 			}
1327 			qname = match_strdup(&args[0]);
1328 			if (!qname) {
1329 				printk(KERN_ERR
1330 					"EXT4-fs: not enough memory for "
1331 					"storing quotafile name.\n");
1332 				return 0;
1333 			}
1334 			if (sbi->s_qf_names[qtype] &&
1335 			    strcmp(sbi->s_qf_names[qtype], qname)) {
1336 				printk(KERN_ERR
1337 					"EXT4-fs: %s quota file already "
1338 					"specified.\n", QTYPE2NAME(qtype));
1339 				kfree(qname);
1340 				return 0;
1341 			}
1342 			sbi->s_qf_names[qtype] = qname;
1343 			if (strchr(sbi->s_qf_names[qtype], '/')) {
1344 				printk(KERN_ERR
1345 					"EXT4-fs: quotafile must be on "
1346 					"filesystem root.\n");
1347 				kfree(sbi->s_qf_names[qtype]);
1348 				sbi->s_qf_names[qtype] = NULL;
1349 				return 0;
1350 			}
1351 			set_opt(sbi->s_mount_opt, QUOTA);
1352 			break;
1353 		case Opt_offusrjquota:
1354 			qtype = USRQUOTA;
1355 			goto clear_qf_name;
1356 		case Opt_offgrpjquota:
1357 			qtype = GRPQUOTA;
1358 clear_qf_name:
1359 			if (sb_any_quota_loaded(sb) &&
1360 			    sbi->s_qf_names[qtype]) {
1361 				printk(KERN_ERR "EXT4-fs: Cannot change "
1362 					"journaled quota options when "
1363 					"quota turned on.\n");
1364 				return 0;
1365 			}
1366 			/*
1367 			 * The space will be released later when all options
1368 			 * are confirmed to be correct
1369 			 */
1370 			sbi->s_qf_names[qtype] = NULL;
1371 			break;
1372 		case Opt_jqfmt_vfsold:
1373 			qfmt = QFMT_VFS_OLD;
1374 			goto set_qf_format;
1375 		case Opt_jqfmt_vfsv0:
1376 			qfmt = QFMT_VFS_V0;
1377 set_qf_format:
1378 			if (sb_any_quota_loaded(sb) &&
1379 			    sbi->s_jquota_fmt != qfmt) {
1380 				printk(KERN_ERR "EXT4-fs: Cannot change "
1381 					"journaled quota options when "
1382 					"quota turned on.\n");
1383 				return 0;
1384 			}
1385 			sbi->s_jquota_fmt = qfmt;
1386 			break;
1387 		case Opt_quota:
1388 		case Opt_usrquota:
1389 			set_opt(sbi->s_mount_opt, QUOTA);
1390 			set_opt(sbi->s_mount_opt, USRQUOTA);
1391 			break;
1392 		case Opt_grpquota:
1393 			set_opt(sbi->s_mount_opt, QUOTA);
1394 			set_opt(sbi->s_mount_opt, GRPQUOTA);
1395 			break;
1396 		case Opt_noquota:
1397 			if (sb_any_quota_loaded(sb)) {
1398 				printk(KERN_ERR "EXT4-fs: Cannot change quota "
1399 					"options when quota turned on.\n");
1400 				return 0;
1401 			}
1402 			clear_opt(sbi->s_mount_opt, QUOTA);
1403 			clear_opt(sbi->s_mount_opt, USRQUOTA);
1404 			clear_opt(sbi->s_mount_opt, GRPQUOTA);
1405 			break;
1406 #else
1407 		case Opt_quota:
1408 		case Opt_usrquota:
1409 		case Opt_grpquota:
1410 			printk(KERN_ERR
1411 				"EXT4-fs: quota options not supported.\n");
1412 			break;
1413 		case Opt_usrjquota:
1414 		case Opt_grpjquota:
1415 		case Opt_offusrjquota:
1416 		case Opt_offgrpjquota:
1417 		case Opt_jqfmt_vfsold:
1418 		case Opt_jqfmt_vfsv0:
1419 			printk(KERN_ERR
1420 				"EXT4-fs: journaled quota options not "
1421 				"supported.\n");
1422 			break;
1423 		case Opt_noquota:
1424 			break;
1425 #endif
1426 		case Opt_abort:
1427 			set_opt(sbi->s_mount_opt, ABORT);
1428 			break;
1429 		case Opt_nobarrier:
1430 			clear_opt(sbi->s_mount_opt, BARRIER);
1431 			break;
1432 		case Opt_barrier:
1433 			if (match_int(&args[0], &option)) {
1434 				set_opt(sbi->s_mount_opt, BARRIER);
1435 				break;
1436 			}
1437 			if (option)
1438 				set_opt(sbi->s_mount_opt, BARRIER);
1439 			else
1440 				clear_opt(sbi->s_mount_opt, BARRIER);
1441 			break;
1442 		case Opt_ignore:
1443 			break;
1444 		case Opt_resize:
1445 			if (!is_remount) {
1446 				printk("EXT4-fs: resize option only available "
1447 					"for remount\n");
1448 				return 0;
1449 			}
1450 			if (match_int(&args[0], &option) != 0)
1451 				return 0;
1452 			*n_blocks_count = option;
1453 			break;
1454 		case Opt_nobh:
1455 			set_opt(sbi->s_mount_opt, NOBH);
1456 			break;
1457 		case Opt_bh:
1458 			clear_opt(sbi->s_mount_opt, NOBH);
1459 			break;
1460 		case Opt_i_version:
1461 			set_opt(sbi->s_mount_opt, I_VERSION);
1462 			sb->s_flags |= MS_I_VERSION;
1463 			break;
1464 		case Opt_nodelalloc:
1465 			clear_opt(sbi->s_mount_opt, DELALLOC);
1466 			break;
1467 		case Opt_stripe:
1468 			if (match_int(&args[0], &option))
1469 				return 0;
1470 			if (option < 0)
1471 				return 0;
1472 			sbi->s_stripe = option;
1473 			break;
1474 		case Opt_delalloc:
1475 			set_opt(sbi->s_mount_opt, DELALLOC);
1476 			break;
1477 		case Opt_inode_readahead_blks:
1478 			if (match_int(&args[0], &option))
1479 				return 0;
1480 			if (option < 0 || option > (1 << 30))
1481 				return 0;
1482 			if (option & (option - 1)) {
1483 				printk(KERN_ERR "EXT4-fs: inode_readahead_blks"
1484 				       " must be a power of 2\n");
1485 				return 0;
1486 			}
1487 			sbi->s_inode_readahead_blks = option;
1488 			break;
1489 		case Opt_journal_ioprio:
1490 			if (match_int(&args[0], &option))
1491 				return 0;
1492 			if (option < 0 || option > 7)
1493 				break;
1494 			*journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1495 							    option);
1496 			break;
1497 		case Opt_noauto_da_alloc:
1498 			set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1499 			break;
1500 		case Opt_auto_da_alloc:
1501 			if (match_int(&args[0], &option)) {
1502 				clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1503 				break;
1504 			}
1505 			if (option)
1506 				clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1507 			else
1508 				set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1509 			break;
1510 		default:
1511 			printk(KERN_ERR
1512 			       "EXT4-fs: Unrecognized mount option \"%s\" "
1513 			       "or missing value\n", p);
1514 			return 0;
1515 		}
1516 	}
1517 #ifdef CONFIG_QUOTA
1518 	if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1519 		if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1520 		     sbi->s_qf_names[USRQUOTA])
1521 			clear_opt(sbi->s_mount_opt, USRQUOTA);
1522 
1523 		if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1524 		     sbi->s_qf_names[GRPQUOTA])
1525 			clear_opt(sbi->s_mount_opt, GRPQUOTA);
1526 
1527 		if ((sbi->s_qf_names[USRQUOTA] &&
1528 				(sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1529 		    (sbi->s_qf_names[GRPQUOTA] &&
1530 				(sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1531 			printk(KERN_ERR "EXT4-fs: old and new quota "
1532 					"format mixing.\n");
1533 			return 0;
1534 		}
1535 
1536 		if (!sbi->s_jquota_fmt) {
1537 			printk(KERN_ERR "EXT4-fs: journaled quota format "
1538 					"not specified.\n");
1539 			return 0;
1540 		}
1541 	} else {
1542 		if (sbi->s_jquota_fmt) {
1543 			printk(KERN_ERR "EXT4-fs: journaled quota format "
1544 					"specified with no journaling "
1545 					"enabled.\n");
1546 			return 0;
1547 		}
1548 	}
1549 #endif
1550 	return 1;
1551 }
1552 
1553 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1554 			    int read_only)
1555 {
1556 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1557 	int res = 0;
1558 
1559 	if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1560 		printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1561 		       "forcing read-only mode\n");
1562 		res = MS_RDONLY;
1563 	}
1564 	if (read_only)
1565 		return res;
1566 	if (!(sbi->s_mount_state & EXT4_VALID_FS))
1567 		printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1568 		       "running e2fsck is recommended\n");
1569 	else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1570 		printk(KERN_WARNING
1571 		       "EXT4-fs warning: mounting fs with errors, "
1572 		       "running e2fsck is recommended\n");
1573 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1574 		 le16_to_cpu(es->s_mnt_count) >=
1575 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1576 		printk(KERN_WARNING
1577 		       "EXT4-fs warning: maximal mount count reached, "
1578 		       "running e2fsck is recommended\n");
1579 	else if (le32_to_cpu(es->s_checkinterval) &&
1580 		(le32_to_cpu(es->s_lastcheck) +
1581 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1582 		printk(KERN_WARNING
1583 		       "EXT4-fs warning: checktime reached, "
1584 		       "running e2fsck is recommended\n");
1585 	if (!sbi->s_journal)
1586 		es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1587 	if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1588 		es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1589 	le16_add_cpu(&es->s_mnt_count, 1);
1590 	es->s_mtime = cpu_to_le32(get_seconds());
1591 	ext4_update_dynamic_rev(sb);
1592 	if (sbi->s_journal)
1593 		EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1594 
1595 	ext4_commit_super(sb, es, 1);
1596 	if (test_opt(sb, DEBUG))
1597 		printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1598 				"bpg=%lu, ipg=%lu, mo=%04lx]\n",
1599 			sb->s_blocksize,
1600 			sbi->s_groups_count,
1601 			EXT4_BLOCKS_PER_GROUP(sb),
1602 			EXT4_INODES_PER_GROUP(sb),
1603 			sbi->s_mount_opt);
1604 
1605 	if (EXT4_SB(sb)->s_journal) {
1606 		printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1607 		       sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1608 		       "external", EXT4_SB(sb)->s_journal->j_devname);
1609 	} else {
1610 		printk(KERN_INFO "EXT4 FS on %s, no journal\n", sb->s_id);
1611 	}
1612 	return res;
1613 }
1614 
1615 static int ext4_fill_flex_info(struct super_block *sb)
1616 {
1617 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1618 	struct ext4_group_desc *gdp = NULL;
1619 	struct buffer_head *bh;
1620 	ext4_group_t flex_group_count;
1621 	ext4_group_t flex_group;
1622 	int groups_per_flex = 0;
1623 	int i;
1624 
1625 	if (!sbi->s_es->s_log_groups_per_flex) {
1626 		sbi->s_log_groups_per_flex = 0;
1627 		return 1;
1628 	}
1629 
1630 	sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1631 	groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1632 
1633 	/* We allocate both existing and potentially added groups */
1634 	flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1635 			((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1636 			      EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1637 	sbi->s_flex_groups = kzalloc(flex_group_count *
1638 				     sizeof(struct flex_groups), GFP_KERNEL);
1639 	if (sbi->s_flex_groups == NULL) {
1640 		printk(KERN_ERR "EXT4-fs: not enough memory for "
1641 				"%u flex groups\n", flex_group_count);
1642 		goto failed;
1643 	}
1644 
1645 	for (i = 0; i < sbi->s_groups_count; i++) {
1646 		gdp = ext4_get_group_desc(sb, i, &bh);
1647 
1648 		flex_group = ext4_flex_group(sbi, i);
1649 		atomic_set(&sbi->s_flex_groups[flex_group].free_inodes,
1650 			   ext4_free_inodes_count(sb, gdp));
1651 		atomic_set(&sbi->s_flex_groups[flex_group].free_blocks,
1652 			   ext4_free_blks_count(sb, gdp));
1653 		atomic_set(&sbi->s_flex_groups[flex_group].used_dirs,
1654 			   ext4_used_dirs_count(sb, gdp));
1655 	}
1656 
1657 	return 1;
1658 failed:
1659 	return 0;
1660 }
1661 
1662 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1663 			    struct ext4_group_desc *gdp)
1664 {
1665 	__u16 crc = 0;
1666 
1667 	if (sbi->s_es->s_feature_ro_compat &
1668 	    cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1669 		int offset = offsetof(struct ext4_group_desc, bg_checksum);
1670 		__le32 le_group = cpu_to_le32(block_group);
1671 
1672 		crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1673 		crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1674 		crc = crc16(crc, (__u8 *)gdp, offset);
1675 		offset += sizeof(gdp->bg_checksum); /* skip checksum */
1676 		/* for checksum of struct ext4_group_desc do the rest...*/
1677 		if ((sbi->s_es->s_feature_incompat &
1678 		     cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1679 		    offset < le16_to_cpu(sbi->s_es->s_desc_size))
1680 			crc = crc16(crc, (__u8 *)gdp + offset,
1681 				    le16_to_cpu(sbi->s_es->s_desc_size) -
1682 					offset);
1683 	}
1684 
1685 	return cpu_to_le16(crc);
1686 }
1687 
1688 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1689 				struct ext4_group_desc *gdp)
1690 {
1691 	if ((sbi->s_es->s_feature_ro_compat &
1692 	     cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1693 	    (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1694 		return 0;
1695 
1696 	return 1;
1697 }
1698 
1699 /* Called at mount-time, super-block is locked */
1700 static int ext4_check_descriptors(struct super_block *sb)
1701 {
1702 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1703 	ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1704 	ext4_fsblk_t last_block;
1705 	ext4_fsblk_t block_bitmap;
1706 	ext4_fsblk_t inode_bitmap;
1707 	ext4_fsblk_t inode_table;
1708 	int flexbg_flag = 0;
1709 	ext4_group_t i;
1710 
1711 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1712 		flexbg_flag = 1;
1713 
1714 	ext4_debug("Checking group descriptors");
1715 
1716 	for (i = 0; i < sbi->s_groups_count; i++) {
1717 		struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1718 
1719 		if (i == sbi->s_groups_count - 1 || flexbg_flag)
1720 			last_block = ext4_blocks_count(sbi->s_es) - 1;
1721 		else
1722 			last_block = first_block +
1723 				(EXT4_BLOCKS_PER_GROUP(sb) - 1);
1724 
1725 		block_bitmap = ext4_block_bitmap(sb, gdp);
1726 		if (block_bitmap < first_block || block_bitmap > last_block) {
1727 			printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1728 			       "Block bitmap for group %u not in group "
1729 			       "(block %llu)!\n", i, block_bitmap);
1730 			return 0;
1731 		}
1732 		inode_bitmap = ext4_inode_bitmap(sb, gdp);
1733 		if (inode_bitmap < first_block || inode_bitmap > last_block) {
1734 			printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1735 			       "Inode bitmap for group %u not in group "
1736 			       "(block %llu)!\n", i, inode_bitmap);
1737 			return 0;
1738 		}
1739 		inode_table = ext4_inode_table(sb, gdp);
1740 		if (inode_table < first_block ||
1741 		    inode_table + sbi->s_itb_per_group - 1 > last_block) {
1742 			printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1743 			       "Inode table for group %u not in group "
1744 			       "(block %llu)!\n", i, inode_table);
1745 			return 0;
1746 		}
1747 		spin_lock(sb_bgl_lock(sbi, i));
1748 		if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1749 			printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1750 			       "Checksum for group %u failed (%u!=%u)\n",
1751 			       i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1752 			       gdp)), le16_to_cpu(gdp->bg_checksum));
1753 			if (!(sb->s_flags & MS_RDONLY)) {
1754 				spin_unlock(sb_bgl_lock(sbi, i));
1755 				return 0;
1756 			}
1757 		}
1758 		spin_unlock(sb_bgl_lock(sbi, i));
1759 		if (!flexbg_flag)
1760 			first_block += EXT4_BLOCKS_PER_GROUP(sb);
1761 	}
1762 
1763 	ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1764 	sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
1765 	return 1;
1766 }
1767 
1768 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1769  * the superblock) which were deleted from all directories, but held open by
1770  * a process at the time of a crash.  We walk the list and try to delete these
1771  * inodes at recovery time (only with a read-write filesystem).
1772  *
1773  * In order to keep the orphan inode chain consistent during traversal (in
1774  * case of crash during recovery), we link each inode into the superblock
1775  * orphan list_head and handle it the same way as an inode deletion during
1776  * normal operation (which journals the operations for us).
1777  *
1778  * We only do an iget() and an iput() on each inode, which is very safe if we
1779  * accidentally point at an in-use or already deleted inode.  The worst that
1780  * can happen in this case is that we get a "bit already cleared" message from
1781  * ext4_free_inode().  The only reason we would point at a wrong inode is if
1782  * e2fsck was run on this filesystem, and it must have already done the orphan
1783  * inode cleanup for us, so we can safely abort without any further action.
1784  */
1785 static void ext4_orphan_cleanup(struct super_block *sb,
1786 				struct ext4_super_block *es)
1787 {
1788 	unsigned int s_flags = sb->s_flags;
1789 	int nr_orphans = 0, nr_truncates = 0;
1790 #ifdef CONFIG_QUOTA
1791 	int i;
1792 #endif
1793 	if (!es->s_last_orphan) {
1794 		jbd_debug(4, "no orphan inodes to clean up\n");
1795 		return;
1796 	}
1797 
1798 	if (bdev_read_only(sb->s_bdev)) {
1799 		printk(KERN_ERR "EXT4-fs: write access "
1800 			"unavailable, skipping orphan cleanup.\n");
1801 		return;
1802 	}
1803 
1804 	if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1805 		if (es->s_last_orphan)
1806 			jbd_debug(1, "Errors on filesystem, "
1807 				  "clearing orphan list.\n");
1808 		es->s_last_orphan = 0;
1809 		jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1810 		return;
1811 	}
1812 
1813 	if (s_flags & MS_RDONLY) {
1814 		printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1815 		       sb->s_id);
1816 		sb->s_flags &= ~MS_RDONLY;
1817 	}
1818 #ifdef CONFIG_QUOTA
1819 	/* Needed for iput() to work correctly and not trash data */
1820 	sb->s_flags |= MS_ACTIVE;
1821 	/* Turn on quotas so that they are updated correctly */
1822 	for (i = 0; i < MAXQUOTAS; i++) {
1823 		if (EXT4_SB(sb)->s_qf_names[i]) {
1824 			int ret = ext4_quota_on_mount(sb, i);
1825 			if (ret < 0)
1826 				printk(KERN_ERR
1827 					"EXT4-fs: Cannot turn on journaled "
1828 					"quota: error %d\n", ret);
1829 		}
1830 	}
1831 #endif
1832 
1833 	while (es->s_last_orphan) {
1834 		struct inode *inode;
1835 
1836 		inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1837 		if (IS_ERR(inode)) {
1838 			es->s_last_orphan = 0;
1839 			break;
1840 		}
1841 
1842 		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1843 		vfs_dq_init(inode);
1844 		if (inode->i_nlink) {
1845 			printk(KERN_DEBUG
1846 				"%s: truncating inode %lu to %lld bytes\n",
1847 				__func__, inode->i_ino, inode->i_size);
1848 			jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1849 				  inode->i_ino, inode->i_size);
1850 			ext4_truncate(inode);
1851 			nr_truncates++;
1852 		} else {
1853 			printk(KERN_DEBUG
1854 				"%s: deleting unreferenced inode %lu\n",
1855 				__func__, inode->i_ino);
1856 			jbd_debug(2, "deleting unreferenced inode %lu\n",
1857 				  inode->i_ino);
1858 			nr_orphans++;
1859 		}
1860 		iput(inode);  /* The delete magic happens here! */
1861 	}
1862 
1863 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1864 
1865 	if (nr_orphans)
1866 		printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1867 		       sb->s_id, PLURAL(nr_orphans));
1868 	if (nr_truncates)
1869 		printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1870 		       sb->s_id, PLURAL(nr_truncates));
1871 #ifdef CONFIG_QUOTA
1872 	/* Turn quotas off */
1873 	for (i = 0; i < MAXQUOTAS; i++) {
1874 		if (sb_dqopt(sb)->files[i])
1875 			vfs_quota_off(sb, i, 0);
1876 	}
1877 #endif
1878 	sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1879 }
1880 /*
1881  * Maximal extent format file size.
1882  * Resulting logical blkno at s_maxbytes must fit in our on-disk
1883  * extent format containers, within a sector_t, and within i_blocks
1884  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
1885  * so that won't be a limiting factor.
1886  *
1887  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1888  */
1889 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1890 {
1891 	loff_t res;
1892 	loff_t upper_limit = MAX_LFS_FILESIZE;
1893 
1894 	/* small i_blocks in vfs inode? */
1895 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1896 		/*
1897 		 * CONFIG_LBD is not enabled implies the inode
1898 		 * i_block represent total blocks in 512 bytes
1899 		 * 32 == size of vfs inode i_blocks * 8
1900 		 */
1901 		upper_limit = (1LL << 32) - 1;
1902 
1903 		/* total blocks in file system block size */
1904 		upper_limit >>= (blkbits - 9);
1905 		upper_limit <<= blkbits;
1906 	}
1907 
1908 	/* 32-bit extent-start container, ee_block */
1909 	res = 1LL << 32;
1910 	res <<= blkbits;
1911 	res -= 1;
1912 
1913 	/* Sanity check against vm- & vfs- imposed limits */
1914 	if (res > upper_limit)
1915 		res = upper_limit;
1916 
1917 	return res;
1918 }
1919 
1920 /*
1921  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
1922  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1923  * We need to be 1 filesystem block less than the 2^48 sector limit.
1924  */
1925 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1926 {
1927 	loff_t res = EXT4_NDIR_BLOCKS;
1928 	int meta_blocks;
1929 	loff_t upper_limit;
1930 	/* This is calculated to be the largest file size for a
1931 	 * dense, bitmapped file such that the total number of
1932 	 * sectors in the file, including data and all indirect blocks,
1933 	 * does not exceed 2^48 -1
1934 	 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1935 	 * total number of  512 bytes blocks of the file
1936 	 */
1937 
1938 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1939 		/*
1940 		 * !has_huge_files or CONFIG_LBD is not enabled
1941 		 * implies the inode i_block represent total blocks in
1942 		 * 512 bytes 32 == size of vfs inode i_blocks * 8
1943 		 */
1944 		upper_limit = (1LL << 32) - 1;
1945 
1946 		/* total blocks in file system block size */
1947 		upper_limit >>= (bits - 9);
1948 
1949 	} else {
1950 		/*
1951 		 * We use 48 bit ext4_inode i_blocks
1952 		 * With EXT4_HUGE_FILE_FL set the i_blocks
1953 		 * represent total number of blocks in
1954 		 * file system block size
1955 		 */
1956 		upper_limit = (1LL << 48) - 1;
1957 
1958 	}
1959 
1960 	/* indirect blocks */
1961 	meta_blocks = 1;
1962 	/* double indirect blocks */
1963 	meta_blocks += 1 + (1LL << (bits-2));
1964 	/* tripple indirect blocks */
1965 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1966 
1967 	upper_limit -= meta_blocks;
1968 	upper_limit <<= bits;
1969 
1970 	res += 1LL << (bits-2);
1971 	res += 1LL << (2*(bits-2));
1972 	res += 1LL << (3*(bits-2));
1973 	res <<= bits;
1974 	if (res > upper_limit)
1975 		res = upper_limit;
1976 
1977 	if (res > MAX_LFS_FILESIZE)
1978 		res = MAX_LFS_FILESIZE;
1979 
1980 	return res;
1981 }
1982 
1983 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1984 				ext4_fsblk_t logical_sb_block, int nr)
1985 {
1986 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1987 	ext4_group_t bg, first_meta_bg;
1988 	int has_super = 0;
1989 
1990 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1991 
1992 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1993 	    nr < first_meta_bg)
1994 		return logical_sb_block + nr + 1;
1995 	bg = sbi->s_desc_per_block * nr;
1996 	if (ext4_bg_has_super(sb, bg))
1997 		has_super = 1;
1998 	return (has_super + ext4_group_first_block_no(sb, bg));
1999 }
2000 
2001 /**
2002  * ext4_get_stripe_size: Get the stripe size.
2003  * @sbi: In memory super block info
2004  *
2005  * If we have specified it via mount option, then
2006  * use the mount option value. If the value specified at mount time is
2007  * greater than the blocks per group use the super block value.
2008  * If the super block value is greater than blocks per group return 0.
2009  * Allocator needs it be less than blocks per group.
2010  *
2011  */
2012 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2013 {
2014 	unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2015 	unsigned long stripe_width =
2016 			le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2017 
2018 	if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2019 		return sbi->s_stripe;
2020 
2021 	if (stripe_width <= sbi->s_blocks_per_group)
2022 		return stripe_width;
2023 
2024 	if (stride <= sbi->s_blocks_per_group)
2025 		return stride;
2026 
2027 	return 0;
2028 }
2029 
2030 /* sysfs supprt */
2031 
2032 struct ext4_attr {
2033 	struct attribute attr;
2034 	ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2035 	ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2036 			 const char *, size_t);
2037 	int offset;
2038 };
2039 
2040 static int parse_strtoul(const char *buf,
2041 		unsigned long max, unsigned long *value)
2042 {
2043 	char *endp;
2044 
2045 	while (*buf && isspace(*buf))
2046 		buf++;
2047 	*value = simple_strtoul(buf, &endp, 0);
2048 	while (*endp && isspace(*endp))
2049 		endp++;
2050 	if (*endp || *value > max)
2051 		return -EINVAL;
2052 
2053 	return 0;
2054 }
2055 
2056 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2057 					      struct ext4_sb_info *sbi,
2058 					      char *buf)
2059 {
2060 	return snprintf(buf, PAGE_SIZE, "%llu\n",
2061 			(s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2062 }
2063 
2064 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2065 					 struct ext4_sb_info *sbi, char *buf)
2066 {
2067 	struct super_block *sb = sbi->s_buddy_cache->i_sb;
2068 
2069 	return snprintf(buf, PAGE_SIZE, "%lu\n",
2070 			(part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2071 			 sbi->s_sectors_written_start) >> 1);
2072 }
2073 
2074 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2075 					  struct ext4_sb_info *sbi, char *buf)
2076 {
2077 	struct super_block *sb = sbi->s_buddy_cache->i_sb;
2078 
2079 	return snprintf(buf, PAGE_SIZE, "%llu\n",
2080 			sbi->s_kbytes_written +
2081 			((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2082 			  EXT4_SB(sb)->s_sectors_written_start) >> 1));
2083 }
2084 
2085 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2086 					  struct ext4_sb_info *sbi,
2087 					  const char *buf, size_t count)
2088 {
2089 	unsigned long t;
2090 
2091 	if (parse_strtoul(buf, 0x40000000, &t))
2092 		return -EINVAL;
2093 
2094 	/* inode_readahead_blks must be a power of 2 */
2095 	if (t & (t-1))
2096 		return -EINVAL;
2097 
2098 	sbi->s_inode_readahead_blks = t;
2099 	return count;
2100 }
2101 
2102 static ssize_t sbi_ui_show(struct ext4_attr *a,
2103 				struct ext4_sb_info *sbi, char *buf)
2104 {
2105 	unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2106 
2107 	return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2108 }
2109 
2110 static ssize_t sbi_ui_store(struct ext4_attr *a,
2111 			    struct ext4_sb_info *sbi,
2112 			    const char *buf, size_t count)
2113 {
2114 	unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2115 	unsigned long t;
2116 
2117 	if (parse_strtoul(buf, 0xffffffff, &t))
2118 		return -EINVAL;
2119 	*ui = t;
2120 	return count;
2121 }
2122 
2123 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2124 static struct ext4_attr ext4_attr_##_name = {			\
2125 	.attr = {.name = __stringify(_name), .mode = _mode },	\
2126 	.show	= _show,					\
2127 	.store	= _store,					\
2128 	.offset = offsetof(struct ext4_sb_info, _elname),	\
2129 }
2130 #define EXT4_ATTR(name, mode, show, store) \
2131 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2132 
2133 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2134 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2135 #define EXT4_RW_ATTR_SBI_UI(name, elname)	\
2136 	EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2137 #define ATTR_LIST(name) &ext4_attr_##name.attr
2138 
2139 EXT4_RO_ATTR(delayed_allocation_blocks);
2140 EXT4_RO_ATTR(session_write_kbytes);
2141 EXT4_RO_ATTR(lifetime_write_kbytes);
2142 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2143 		 inode_readahead_blks_store, s_inode_readahead_blks);
2144 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2145 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2146 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2147 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2148 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2149 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2150 
2151 static struct attribute *ext4_attrs[] = {
2152 	ATTR_LIST(delayed_allocation_blocks),
2153 	ATTR_LIST(session_write_kbytes),
2154 	ATTR_LIST(lifetime_write_kbytes),
2155 	ATTR_LIST(inode_readahead_blks),
2156 	ATTR_LIST(mb_stats),
2157 	ATTR_LIST(mb_max_to_scan),
2158 	ATTR_LIST(mb_min_to_scan),
2159 	ATTR_LIST(mb_order2_req),
2160 	ATTR_LIST(mb_stream_req),
2161 	ATTR_LIST(mb_group_prealloc),
2162 	NULL,
2163 };
2164 
2165 static ssize_t ext4_attr_show(struct kobject *kobj,
2166 			      struct attribute *attr, char *buf)
2167 {
2168 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2169 						s_kobj);
2170 	struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2171 
2172 	return a->show ? a->show(a, sbi, buf) : 0;
2173 }
2174 
2175 static ssize_t ext4_attr_store(struct kobject *kobj,
2176 			       struct attribute *attr,
2177 			       const char *buf, size_t len)
2178 {
2179 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2180 						s_kobj);
2181 	struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2182 
2183 	return a->store ? a->store(a, sbi, buf, len) : 0;
2184 }
2185 
2186 static void ext4_sb_release(struct kobject *kobj)
2187 {
2188 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2189 						s_kobj);
2190 	complete(&sbi->s_kobj_unregister);
2191 }
2192 
2193 
2194 static struct sysfs_ops ext4_attr_ops = {
2195 	.show	= ext4_attr_show,
2196 	.store	= ext4_attr_store,
2197 };
2198 
2199 static struct kobj_type ext4_ktype = {
2200 	.default_attrs	= ext4_attrs,
2201 	.sysfs_ops	= &ext4_attr_ops,
2202 	.release	= ext4_sb_release,
2203 };
2204 
2205 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2206 				__releases(kernel_lock)
2207 				__acquires(kernel_lock)
2208 
2209 {
2210 	struct buffer_head *bh;
2211 	struct ext4_super_block *es = NULL;
2212 	struct ext4_sb_info *sbi;
2213 	ext4_fsblk_t block;
2214 	ext4_fsblk_t sb_block = get_sb_block(&data);
2215 	ext4_fsblk_t logical_sb_block;
2216 	unsigned long offset = 0;
2217 	unsigned long journal_devnum = 0;
2218 	unsigned long def_mount_opts;
2219 	struct inode *root;
2220 	char *cp;
2221 	const char *descr;
2222 	int ret = -EINVAL;
2223 	int blocksize;
2224 	unsigned int db_count;
2225 	unsigned int i;
2226 	int needs_recovery, has_huge_files;
2227 	int features;
2228 	__u64 blocks_count;
2229 	int err;
2230 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2231 
2232 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2233 	if (!sbi)
2234 		return -ENOMEM;
2235 
2236 	sbi->s_blockgroup_lock =
2237 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2238 	if (!sbi->s_blockgroup_lock) {
2239 		kfree(sbi);
2240 		return -ENOMEM;
2241 	}
2242 	sb->s_fs_info = sbi;
2243 	sbi->s_mount_opt = 0;
2244 	sbi->s_resuid = EXT4_DEF_RESUID;
2245 	sbi->s_resgid = EXT4_DEF_RESGID;
2246 	sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2247 	sbi->s_sb_block = sb_block;
2248 	sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2249 						      sectors[1]);
2250 
2251 	unlock_kernel();
2252 
2253 	/* Cleanup superblock name */
2254 	for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2255 		*cp = '!';
2256 
2257 	blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2258 	if (!blocksize) {
2259 		printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
2260 		goto out_fail;
2261 	}
2262 
2263 	/*
2264 	 * The ext4 superblock will not be buffer aligned for other than 1kB
2265 	 * block sizes.  We need to calculate the offset from buffer start.
2266 	 */
2267 	if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2268 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2269 		offset = do_div(logical_sb_block, blocksize);
2270 	} else {
2271 		logical_sb_block = sb_block;
2272 	}
2273 
2274 	if (!(bh = sb_bread(sb, logical_sb_block))) {
2275 		printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
2276 		goto out_fail;
2277 	}
2278 	/*
2279 	 * Note: s_es must be initialized as soon as possible because
2280 	 *       some ext4 macro-instructions depend on its value
2281 	 */
2282 	es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2283 	sbi->s_es = es;
2284 	sb->s_magic = le16_to_cpu(es->s_magic);
2285 	if (sb->s_magic != EXT4_SUPER_MAGIC)
2286 		goto cantfind_ext4;
2287 	sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2288 
2289 	/* Set defaults before we parse the mount options */
2290 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2291 	if (def_mount_opts & EXT4_DEFM_DEBUG)
2292 		set_opt(sbi->s_mount_opt, DEBUG);
2293 	if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2294 		set_opt(sbi->s_mount_opt, GRPID);
2295 	if (def_mount_opts & EXT4_DEFM_UID16)
2296 		set_opt(sbi->s_mount_opt, NO_UID32);
2297 #ifdef CONFIG_EXT4_FS_XATTR
2298 	if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2299 		set_opt(sbi->s_mount_opt, XATTR_USER);
2300 #endif
2301 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2302 	if (def_mount_opts & EXT4_DEFM_ACL)
2303 		set_opt(sbi->s_mount_opt, POSIX_ACL);
2304 #endif
2305 	if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2306 		sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2307 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2308 		sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2309 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2310 		sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2311 
2312 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2313 		set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2314 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2315 		set_opt(sbi->s_mount_opt, ERRORS_CONT);
2316 	else
2317 		set_opt(sbi->s_mount_opt, ERRORS_RO);
2318 
2319 	sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2320 	sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2321 	sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2322 	sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2323 	sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2324 
2325 	set_opt(sbi->s_mount_opt, BARRIER);
2326 
2327 	/*
2328 	 * enable delayed allocation by default
2329 	 * Use -o nodelalloc to turn it off
2330 	 */
2331 	set_opt(sbi->s_mount_opt, DELALLOC);
2332 
2333 
2334 	if (!parse_options((char *) data, sb, &journal_devnum,
2335 			   &journal_ioprio, NULL, 0))
2336 		goto failed_mount;
2337 
2338 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2339 		((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2340 
2341 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2342 	    (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2343 	     EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2344 	     EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2345 		printk(KERN_WARNING
2346 		       "EXT4-fs warning: feature flags set on rev 0 fs, "
2347 		       "running e2fsck is recommended\n");
2348 
2349 	/*
2350 	 * Check feature flags regardless of the revision level, since we
2351 	 * previously didn't change the revision level when setting the flags,
2352 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
2353 	 */
2354 	features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2355 	if (features) {
2356 		printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
2357 		       "unsupported optional features (%x).\n", sb->s_id,
2358 			(le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2359 			~EXT4_FEATURE_INCOMPAT_SUPP));
2360 		goto failed_mount;
2361 	}
2362 	features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2363 	if (!(sb->s_flags & MS_RDONLY) && features) {
2364 		printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
2365 		       "unsupported optional features (%x).\n", sb->s_id,
2366 			(le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2367 			~EXT4_FEATURE_RO_COMPAT_SUPP));
2368 		goto failed_mount;
2369 	}
2370 	has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2371 				    EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2372 	if (has_huge_files) {
2373 		/*
2374 		 * Large file size enabled file system can only be
2375 		 * mount if kernel is build with CONFIG_LBD
2376 		 */
2377 		if (sizeof(root->i_blocks) < sizeof(u64) &&
2378 				!(sb->s_flags & MS_RDONLY)) {
2379 			printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2380 					"files cannot be mounted read-write "
2381 					"without CONFIG_LBD.\n", sb->s_id);
2382 			goto failed_mount;
2383 		}
2384 	}
2385 	blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2386 
2387 	if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2388 	    blocksize > EXT4_MAX_BLOCK_SIZE) {
2389 		printk(KERN_ERR
2390 		       "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2391 		       blocksize, sb->s_id);
2392 		goto failed_mount;
2393 	}
2394 
2395 	if (sb->s_blocksize != blocksize) {
2396 
2397 		/* Validate the filesystem blocksize */
2398 		if (!sb_set_blocksize(sb, blocksize)) {
2399 			printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2400 					blocksize);
2401 			goto failed_mount;
2402 		}
2403 
2404 		brelse(bh);
2405 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2406 		offset = do_div(logical_sb_block, blocksize);
2407 		bh = sb_bread(sb, logical_sb_block);
2408 		if (!bh) {
2409 			printk(KERN_ERR
2410 			       "EXT4-fs: Can't read superblock on 2nd try.\n");
2411 			goto failed_mount;
2412 		}
2413 		es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2414 		sbi->s_es = es;
2415 		if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2416 			printk(KERN_ERR
2417 			       "EXT4-fs: Magic mismatch, very weird !\n");
2418 			goto failed_mount;
2419 		}
2420 	}
2421 
2422 	sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2423 						      has_huge_files);
2424 	sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2425 
2426 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2427 		sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2428 		sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2429 	} else {
2430 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2431 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2432 		if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2433 		    (!is_power_of_2(sbi->s_inode_size)) ||
2434 		    (sbi->s_inode_size > blocksize)) {
2435 			printk(KERN_ERR
2436 			       "EXT4-fs: unsupported inode size: %d\n",
2437 			       sbi->s_inode_size);
2438 			goto failed_mount;
2439 		}
2440 		if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2441 			sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2442 	}
2443 	sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2444 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2445 		if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2446 		    sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2447 		    !is_power_of_2(sbi->s_desc_size)) {
2448 			printk(KERN_ERR
2449 			       "EXT4-fs: unsupported descriptor size %lu\n",
2450 			       sbi->s_desc_size);
2451 			goto failed_mount;
2452 		}
2453 	} else
2454 		sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2455 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2456 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2457 	if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2458 		goto cantfind_ext4;
2459 	sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2460 	if (sbi->s_inodes_per_block == 0)
2461 		goto cantfind_ext4;
2462 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
2463 					sbi->s_inodes_per_block;
2464 	sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2465 	sbi->s_sbh = bh;
2466 	sbi->s_mount_state = le16_to_cpu(es->s_state);
2467 	sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2468 	sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2469 	for (i = 0; i < 4; i++)
2470 		sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2471 	sbi->s_def_hash_version = es->s_def_hash_version;
2472 	i = le32_to_cpu(es->s_flags);
2473 	if (i & EXT2_FLAGS_UNSIGNED_HASH)
2474 		sbi->s_hash_unsigned = 3;
2475 	else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2476 #ifdef __CHAR_UNSIGNED__
2477 		es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2478 		sbi->s_hash_unsigned = 3;
2479 #else
2480 		es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2481 #endif
2482 		sb->s_dirt = 1;
2483 	}
2484 
2485 	if (sbi->s_blocks_per_group > blocksize * 8) {
2486 		printk(KERN_ERR
2487 		       "EXT4-fs: #blocks per group too big: %lu\n",
2488 		       sbi->s_blocks_per_group);
2489 		goto failed_mount;
2490 	}
2491 	if (sbi->s_inodes_per_group > blocksize * 8) {
2492 		printk(KERN_ERR
2493 		       "EXT4-fs: #inodes per group too big: %lu\n",
2494 		       sbi->s_inodes_per_group);
2495 		goto failed_mount;
2496 	}
2497 
2498 	if (ext4_blocks_count(es) >
2499 		    (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2500 		printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2501 			" too large to mount safely\n", sb->s_id);
2502 		if (sizeof(sector_t) < 8)
2503 			printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2504 					"enabled\n");
2505 		goto failed_mount;
2506 	}
2507 
2508 	if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2509 		goto cantfind_ext4;
2510 
2511         /*
2512          * It makes no sense for the first data block to be beyond the end
2513          * of the filesystem.
2514          */
2515         if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2516                 printk(KERN_WARNING "EXT4-fs: bad geometry: first data"
2517 		       "block %u is beyond end of filesystem (%llu)\n",
2518 		       le32_to_cpu(es->s_first_data_block),
2519 		       ext4_blocks_count(es));
2520 		goto failed_mount;
2521 	}
2522 	blocks_count = (ext4_blocks_count(es) -
2523 			le32_to_cpu(es->s_first_data_block) +
2524 			EXT4_BLOCKS_PER_GROUP(sb) - 1);
2525 	do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2526 	if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2527 		printk(KERN_WARNING "EXT4-fs: groups count too large: %u "
2528 		       "(block count %llu, first data block %u, "
2529 		       "blocks per group %lu)\n", sbi->s_groups_count,
2530 		       ext4_blocks_count(es),
2531 		       le32_to_cpu(es->s_first_data_block),
2532 		       EXT4_BLOCKS_PER_GROUP(sb));
2533 		goto failed_mount;
2534 	}
2535 	sbi->s_groups_count = blocks_count;
2536 	db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2537 		   EXT4_DESC_PER_BLOCK(sb);
2538 	sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2539 				    GFP_KERNEL);
2540 	if (sbi->s_group_desc == NULL) {
2541 		printk(KERN_ERR "EXT4-fs: not enough memory\n");
2542 		goto failed_mount;
2543 	}
2544 
2545 #ifdef CONFIG_PROC_FS
2546 	if (ext4_proc_root)
2547 		sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2548 #endif
2549 
2550 	bgl_lock_init(sbi->s_blockgroup_lock);
2551 
2552 	for (i = 0; i < db_count; i++) {
2553 		block = descriptor_loc(sb, logical_sb_block, i);
2554 		sbi->s_group_desc[i] = sb_bread(sb, block);
2555 		if (!sbi->s_group_desc[i]) {
2556 			printk(KERN_ERR "EXT4-fs: "
2557 			       "can't read group descriptor %d\n", i);
2558 			db_count = i;
2559 			goto failed_mount2;
2560 		}
2561 	}
2562 	if (!ext4_check_descriptors(sb)) {
2563 		printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2564 		goto failed_mount2;
2565 	}
2566 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2567 		if (!ext4_fill_flex_info(sb)) {
2568 			printk(KERN_ERR
2569 			       "EXT4-fs: unable to initialize "
2570 			       "flex_bg meta info!\n");
2571 			goto failed_mount2;
2572 		}
2573 
2574 	sbi->s_gdb_count = db_count;
2575 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2576 	spin_lock_init(&sbi->s_next_gen_lock);
2577 
2578 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
2579 			ext4_count_free_blocks(sb));
2580 	if (!err) {
2581 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
2582 				ext4_count_free_inodes(sb));
2583 	}
2584 	if (!err) {
2585 		err = percpu_counter_init(&sbi->s_dirs_counter,
2586 				ext4_count_dirs(sb));
2587 	}
2588 	if (!err) {
2589 		err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2590 	}
2591 	if (err) {
2592 		printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2593 		goto failed_mount3;
2594 	}
2595 
2596 	sbi->s_stripe = ext4_get_stripe_size(sbi);
2597 
2598 	/*
2599 	 * set up enough so that it can read an inode
2600 	 */
2601 	sb->s_op = &ext4_sops;
2602 	sb->s_export_op = &ext4_export_ops;
2603 	sb->s_xattr = ext4_xattr_handlers;
2604 #ifdef CONFIG_QUOTA
2605 	sb->s_qcop = &ext4_qctl_operations;
2606 	sb->dq_op = &ext4_quota_operations;
2607 #endif
2608 	INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2609 
2610 	sb->s_root = NULL;
2611 
2612 	needs_recovery = (es->s_last_orphan != 0 ||
2613 			  EXT4_HAS_INCOMPAT_FEATURE(sb,
2614 				    EXT4_FEATURE_INCOMPAT_RECOVER));
2615 
2616 	/*
2617 	 * The first inode we look at is the journal inode.  Don't try
2618 	 * root first: it may be modified in the journal!
2619 	 */
2620 	if (!test_opt(sb, NOLOAD) &&
2621 	    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2622 		if (ext4_load_journal(sb, es, journal_devnum))
2623 			goto failed_mount3;
2624 		if (!(sb->s_flags & MS_RDONLY) &&
2625 		    EXT4_SB(sb)->s_journal->j_failed_commit) {
2626 			printk(KERN_CRIT "EXT4-fs error (device %s): "
2627 			       "ext4_fill_super: Journal transaction "
2628 			       "%u is corrupt\n", sb->s_id,
2629 			       EXT4_SB(sb)->s_journal->j_failed_commit);
2630 			if (test_opt(sb, ERRORS_RO)) {
2631 				printk(KERN_CRIT
2632 				       "Mounting filesystem read-only\n");
2633 				sb->s_flags |= MS_RDONLY;
2634 				EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2635 				es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2636 			}
2637 			if (test_opt(sb, ERRORS_PANIC)) {
2638 				EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2639 				es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2640 				ext4_commit_super(sb, es, 1);
2641 				goto failed_mount4;
2642 			}
2643 		}
2644 	} else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2645 	      EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2646 		printk(KERN_ERR "EXT4-fs: required journal recovery "
2647 		       "suppressed and not mounted read-only\n");
2648 		goto failed_mount4;
2649 	} else {
2650 		clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2651 		set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2652 		sbi->s_journal = NULL;
2653 		needs_recovery = 0;
2654 		goto no_journal;
2655 	}
2656 
2657 	if (ext4_blocks_count(es) > 0xffffffffULL &&
2658 	    !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2659 				       JBD2_FEATURE_INCOMPAT_64BIT)) {
2660 		printk(KERN_ERR "EXT4-fs: Failed to set 64-bit journal feature\n");
2661 		goto failed_mount4;
2662 	}
2663 
2664 	if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2665 		jbd2_journal_set_features(sbi->s_journal,
2666 				JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2667 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2668 	} else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2669 		jbd2_journal_set_features(sbi->s_journal,
2670 				JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2671 		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2672 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2673 	} else {
2674 		jbd2_journal_clear_features(sbi->s_journal,
2675 				JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2676 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2677 	}
2678 
2679 	/* We have now updated the journal if required, so we can
2680 	 * validate the data journaling mode. */
2681 	switch (test_opt(sb, DATA_FLAGS)) {
2682 	case 0:
2683 		/* No mode set, assume a default based on the journal
2684 		 * capabilities: ORDERED_DATA if the journal can
2685 		 * cope, else JOURNAL_DATA
2686 		 */
2687 		if (jbd2_journal_check_available_features
2688 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2689 			set_opt(sbi->s_mount_opt, ORDERED_DATA);
2690 		else
2691 			set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2692 		break;
2693 
2694 	case EXT4_MOUNT_ORDERED_DATA:
2695 	case EXT4_MOUNT_WRITEBACK_DATA:
2696 		if (!jbd2_journal_check_available_features
2697 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2698 			printk(KERN_ERR "EXT4-fs: Journal does not support "
2699 			       "requested data journaling mode\n");
2700 			goto failed_mount4;
2701 		}
2702 	default:
2703 		break;
2704 	}
2705 	set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2706 
2707 no_journal:
2708 
2709 	if (test_opt(sb, NOBH)) {
2710 		if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2711 			printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2712 				"its supported only with writeback mode\n");
2713 			clear_opt(sbi->s_mount_opt, NOBH);
2714 		}
2715 	}
2716 	/*
2717 	 * The jbd2_journal_load will have done any necessary log recovery,
2718 	 * so we can safely mount the rest of the filesystem now.
2719 	 */
2720 
2721 	root = ext4_iget(sb, EXT4_ROOT_INO);
2722 	if (IS_ERR(root)) {
2723 		printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2724 		ret = PTR_ERR(root);
2725 		goto failed_mount4;
2726 	}
2727 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2728 		iput(root);
2729 		printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2730 		goto failed_mount4;
2731 	}
2732 	sb->s_root = d_alloc_root(root);
2733 	if (!sb->s_root) {
2734 		printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2735 		iput(root);
2736 		ret = -ENOMEM;
2737 		goto failed_mount4;
2738 	}
2739 
2740 	ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2741 
2742 	/* determine the minimum size of new large inodes, if present */
2743 	if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2744 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2745 						     EXT4_GOOD_OLD_INODE_SIZE;
2746 		if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2747 				       EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2748 			if (sbi->s_want_extra_isize <
2749 			    le16_to_cpu(es->s_want_extra_isize))
2750 				sbi->s_want_extra_isize =
2751 					le16_to_cpu(es->s_want_extra_isize);
2752 			if (sbi->s_want_extra_isize <
2753 			    le16_to_cpu(es->s_min_extra_isize))
2754 				sbi->s_want_extra_isize =
2755 					le16_to_cpu(es->s_min_extra_isize);
2756 		}
2757 	}
2758 	/* Check if enough inode space is available */
2759 	if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2760 							sbi->s_inode_size) {
2761 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2762 						       EXT4_GOOD_OLD_INODE_SIZE;
2763 		printk(KERN_INFO "EXT4-fs: required extra inode space not"
2764 			"available.\n");
2765 	}
2766 
2767 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2768 		printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2769 				"requested data journaling mode\n");
2770 		clear_opt(sbi->s_mount_opt, DELALLOC);
2771 	} else if (test_opt(sb, DELALLOC))
2772 		printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2773 
2774 	ext4_ext_init(sb);
2775 	err = ext4_mb_init(sb, needs_recovery);
2776 	if (err) {
2777 		printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2778 		       err);
2779 		goto failed_mount4;
2780 	}
2781 
2782 	sbi->s_kobj.kset = ext4_kset;
2783 	init_completion(&sbi->s_kobj_unregister);
2784 	err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2785 				   "%s", sb->s_id);
2786 	if (err) {
2787 		ext4_mb_release(sb);
2788 		ext4_ext_release(sb);
2789 		goto failed_mount4;
2790 	};
2791 
2792 	/*
2793 	 * akpm: core read_super() calls in here with the superblock locked.
2794 	 * That deadlocks, because orphan cleanup needs to lock the superblock
2795 	 * in numerous places.  Here we just pop the lock - it's relatively
2796 	 * harmless, because we are now ready to accept write_super() requests,
2797 	 * and aviro says that's the only reason for hanging onto the
2798 	 * superblock lock.
2799 	 */
2800 	EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2801 	ext4_orphan_cleanup(sb, es);
2802 	EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2803 	if (needs_recovery) {
2804 		printk(KERN_INFO "EXT4-fs: recovery complete.\n");
2805 		ext4_mark_recovery_complete(sb, es);
2806 	}
2807 	if (EXT4_SB(sb)->s_journal) {
2808 		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2809 			descr = " journalled data mode";
2810 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2811 			descr = " ordered data mode";
2812 		else
2813 			descr = " writeback data mode";
2814 	} else
2815 		descr = "out journal";
2816 
2817 	printk(KERN_INFO "EXT4-fs: mounted filesystem %s with%s\n",
2818 	       sb->s_id, descr);
2819 
2820 	lock_kernel();
2821 	return 0;
2822 
2823 cantfind_ext4:
2824 	if (!silent)
2825 		printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2826 		       sb->s_id);
2827 	goto failed_mount;
2828 
2829 failed_mount4:
2830 	printk(KERN_ERR "EXT4-fs (device %s): mount failed\n", sb->s_id);
2831 	if (sbi->s_journal) {
2832 		jbd2_journal_destroy(sbi->s_journal);
2833 		sbi->s_journal = NULL;
2834 	}
2835 failed_mount3:
2836 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
2837 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
2838 	percpu_counter_destroy(&sbi->s_dirs_counter);
2839 	percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2840 failed_mount2:
2841 	for (i = 0; i < db_count; i++)
2842 		brelse(sbi->s_group_desc[i]);
2843 	kfree(sbi->s_group_desc);
2844 failed_mount:
2845 	if (sbi->s_proc) {
2846 		remove_proc_entry(sb->s_id, ext4_proc_root);
2847 	}
2848 #ifdef CONFIG_QUOTA
2849 	for (i = 0; i < MAXQUOTAS; i++)
2850 		kfree(sbi->s_qf_names[i]);
2851 #endif
2852 	ext4_blkdev_remove(sbi);
2853 	brelse(bh);
2854 out_fail:
2855 	sb->s_fs_info = NULL;
2856 	kfree(sbi);
2857 	lock_kernel();
2858 	return ret;
2859 }
2860 
2861 /*
2862  * Setup any per-fs journal parameters now.  We'll do this both on
2863  * initial mount, once the journal has been initialised but before we've
2864  * done any recovery; and again on any subsequent remount.
2865  */
2866 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2867 {
2868 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2869 
2870 	journal->j_commit_interval = sbi->s_commit_interval;
2871 	journal->j_min_batch_time = sbi->s_min_batch_time;
2872 	journal->j_max_batch_time = sbi->s_max_batch_time;
2873 
2874 	spin_lock(&journal->j_state_lock);
2875 	if (test_opt(sb, BARRIER))
2876 		journal->j_flags |= JBD2_BARRIER;
2877 	else
2878 		journal->j_flags &= ~JBD2_BARRIER;
2879 	if (test_opt(sb, DATA_ERR_ABORT))
2880 		journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2881 	else
2882 		journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2883 	spin_unlock(&journal->j_state_lock);
2884 }
2885 
2886 static journal_t *ext4_get_journal(struct super_block *sb,
2887 				   unsigned int journal_inum)
2888 {
2889 	struct inode *journal_inode;
2890 	journal_t *journal;
2891 
2892 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2893 
2894 	/* First, test for the existence of a valid inode on disk.  Bad
2895 	 * things happen if we iget() an unused inode, as the subsequent
2896 	 * iput() will try to delete it. */
2897 
2898 	journal_inode = ext4_iget(sb, journal_inum);
2899 	if (IS_ERR(journal_inode)) {
2900 		printk(KERN_ERR "EXT4-fs: no journal found.\n");
2901 		return NULL;
2902 	}
2903 	if (!journal_inode->i_nlink) {
2904 		make_bad_inode(journal_inode);
2905 		iput(journal_inode);
2906 		printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2907 		return NULL;
2908 	}
2909 
2910 	jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2911 		  journal_inode, journal_inode->i_size);
2912 	if (!S_ISREG(journal_inode->i_mode)) {
2913 		printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2914 		iput(journal_inode);
2915 		return NULL;
2916 	}
2917 
2918 	journal = jbd2_journal_init_inode(journal_inode);
2919 	if (!journal) {
2920 		printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2921 		iput(journal_inode);
2922 		return NULL;
2923 	}
2924 	journal->j_private = sb;
2925 	ext4_init_journal_params(sb, journal);
2926 	return journal;
2927 }
2928 
2929 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2930 				       dev_t j_dev)
2931 {
2932 	struct buffer_head *bh;
2933 	journal_t *journal;
2934 	ext4_fsblk_t start;
2935 	ext4_fsblk_t len;
2936 	int hblock, blocksize;
2937 	ext4_fsblk_t sb_block;
2938 	unsigned long offset;
2939 	struct ext4_super_block *es;
2940 	struct block_device *bdev;
2941 
2942 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2943 
2944 	bdev = ext4_blkdev_get(j_dev);
2945 	if (bdev == NULL)
2946 		return NULL;
2947 
2948 	if (bd_claim(bdev, sb)) {
2949 		printk(KERN_ERR
2950 			"EXT4-fs: failed to claim external journal device.\n");
2951 		blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2952 		return NULL;
2953 	}
2954 
2955 	blocksize = sb->s_blocksize;
2956 	hblock = bdev_hardsect_size(bdev);
2957 	if (blocksize < hblock) {
2958 		printk(KERN_ERR
2959 			"EXT4-fs: blocksize too small for journal device.\n");
2960 		goto out_bdev;
2961 	}
2962 
2963 	sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2964 	offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2965 	set_blocksize(bdev, blocksize);
2966 	if (!(bh = __bread(bdev, sb_block, blocksize))) {
2967 		printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2968 		       "external journal\n");
2969 		goto out_bdev;
2970 	}
2971 
2972 	es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2973 	if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2974 	    !(le32_to_cpu(es->s_feature_incompat) &
2975 	      EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2976 		printk(KERN_ERR "EXT4-fs: external journal has "
2977 					"bad superblock\n");
2978 		brelse(bh);
2979 		goto out_bdev;
2980 	}
2981 
2982 	if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2983 		printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2984 		brelse(bh);
2985 		goto out_bdev;
2986 	}
2987 
2988 	len = ext4_blocks_count(es);
2989 	start = sb_block + 1;
2990 	brelse(bh);	/* we're done with the superblock */
2991 
2992 	journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2993 					start, len, blocksize);
2994 	if (!journal) {
2995 		printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2996 		goto out_bdev;
2997 	}
2998 	journal->j_private = sb;
2999 	ll_rw_block(READ, 1, &journal->j_sb_buffer);
3000 	wait_on_buffer(journal->j_sb_buffer);
3001 	if (!buffer_uptodate(journal->j_sb_buffer)) {
3002 		printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
3003 		goto out_journal;
3004 	}
3005 	if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3006 		printk(KERN_ERR "EXT4-fs: External journal has more than one "
3007 					"user (unsupported) - %d\n",
3008 			be32_to_cpu(journal->j_superblock->s_nr_users));
3009 		goto out_journal;
3010 	}
3011 	EXT4_SB(sb)->journal_bdev = bdev;
3012 	ext4_init_journal_params(sb, journal);
3013 	return journal;
3014 out_journal:
3015 	jbd2_journal_destroy(journal);
3016 out_bdev:
3017 	ext4_blkdev_put(bdev);
3018 	return NULL;
3019 }
3020 
3021 static int ext4_load_journal(struct super_block *sb,
3022 			     struct ext4_super_block *es,
3023 			     unsigned long journal_devnum)
3024 {
3025 	journal_t *journal;
3026 	unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3027 	dev_t journal_dev;
3028 	int err = 0;
3029 	int really_read_only;
3030 
3031 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3032 
3033 	if (journal_devnum &&
3034 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3035 		printk(KERN_INFO "EXT4-fs: external journal device major/minor "
3036 			"numbers have changed\n");
3037 		journal_dev = new_decode_dev(journal_devnum);
3038 	} else
3039 		journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3040 
3041 	really_read_only = bdev_read_only(sb->s_bdev);
3042 
3043 	/*
3044 	 * Are we loading a blank journal or performing recovery after a
3045 	 * crash?  For recovery, we need to check in advance whether we
3046 	 * can get read-write access to the device.
3047 	 */
3048 
3049 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3050 		if (sb->s_flags & MS_RDONLY) {
3051 			printk(KERN_INFO "EXT4-fs: INFO: recovery "
3052 					"required on readonly filesystem.\n");
3053 			if (really_read_only) {
3054 				printk(KERN_ERR "EXT4-fs: write access "
3055 					"unavailable, cannot proceed.\n");
3056 				return -EROFS;
3057 			}
3058 			printk(KERN_INFO "EXT4-fs: write access will "
3059 			       "be enabled during recovery.\n");
3060 		}
3061 	}
3062 
3063 	if (journal_inum && journal_dev) {
3064 		printk(KERN_ERR "EXT4-fs: filesystem has both journal "
3065 		       "and inode journals!\n");
3066 		return -EINVAL;
3067 	}
3068 
3069 	if (journal_inum) {
3070 		if (!(journal = ext4_get_journal(sb, journal_inum)))
3071 			return -EINVAL;
3072 	} else {
3073 		if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3074 			return -EINVAL;
3075 	}
3076 
3077 	if (journal->j_flags & JBD2_BARRIER)
3078 		printk(KERN_INFO "EXT4-fs: barriers enabled\n");
3079 	else
3080 		printk(KERN_INFO "EXT4-fs: barriers disabled\n");
3081 
3082 	if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3083 		err = jbd2_journal_update_format(journal);
3084 		if (err)  {
3085 			printk(KERN_ERR "EXT4-fs: error updating journal.\n");
3086 			jbd2_journal_destroy(journal);
3087 			return err;
3088 		}
3089 	}
3090 
3091 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3092 		err = jbd2_journal_wipe(journal, !really_read_only);
3093 	if (!err)
3094 		err = jbd2_journal_load(journal);
3095 
3096 	if (err) {
3097 		printk(KERN_ERR "EXT4-fs: error loading journal.\n");
3098 		jbd2_journal_destroy(journal);
3099 		return err;
3100 	}
3101 
3102 	EXT4_SB(sb)->s_journal = journal;
3103 	ext4_clear_journal_err(sb, es);
3104 
3105 	if (journal_devnum &&
3106 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3107 		es->s_journal_dev = cpu_to_le32(journal_devnum);
3108 		sb->s_dirt = 1;
3109 
3110 		/* Make sure we flush the recovery flag to disk. */
3111 		ext4_commit_super(sb, es, 1);
3112 	}
3113 
3114 	return 0;
3115 }
3116 
3117 static int ext4_commit_super(struct super_block *sb,
3118 			      struct ext4_super_block *es, int sync)
3119 {
3120 	struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3121 	int error = 0;
3122 
3123 	if (!sbh)
3124 		return error;
3125 	if (buffer_write_io_error(sbh)) {
3126 		/*
3127 		 * Oh, dear.  A previous attempt to write the
3128 		 * superblock failed.  This could happen because the
3129 		 * USB device was yanked out.  Or it could happen to
3130 		 * be a transient write error and maybe the block will
3131 		 * be remapped.  Nothing we can do but to retry the
3132 		 * write and hope for the best.
3133 		 */
3134 		printk(KERN_ERR "EXT4-fs: previous I/O error to "
3135 		       "superblock detected for %s.\n", sb->s_id);
3136 		clear_buffer_write_io_error(sbh);
3137 		set_buffer_uptodate(sbh);
3138 	}
3139 	es->s_wtime = cpu_to_le32(get_seconds());
3140 	es->s_kbytes_written =
3141 		cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3142 			    ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3143 			      EXT4_SB(sb)->s_sectors_written_start) >> 1));
3144 	ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3145 					&EXT4_SB(sb)->s_freeblocks_counter));
3146 	es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3147 					&EXT4_SB(sb)->s_freeinodes_counter));
3148 
3149 	BUFFER_TRACE(sbh, "marking dirty");
3150 	mark_buffer_dirty(sbh);
3151 	if (sync) {
3152 		error = sync_dirty_buffer(sbh);
3153 		if (error)
3154 			return error;
3155 
3156 		error = buffer_write_io_error(sbh);
3157 		if (error) {
3158 			printk(KERN_ERR "EXT4-fs: I/O error while writing "
3159 			       "superblock for %s.\n", sb->s_id);
3160 			clear_buffer_write_io_error(sbh);
3161 			set_buffer_uptodate(sbh);
3162 		}
3163 	}
3164 	return error;
3165 }
3166 
3167 
3168 /*
3169  * Have we just finished recovery?  If so, and if we are mounting (or
3170  * remounting) the filesystem readonly, then we will end up with a
3171  * consistent fs on disk.  Record that fact.
3172  */
3173 static void ext4_mark_recovery_complete(struct super_block *sb,
3174 					struct ext4_super_block *es)
3175 {
3176 	journal_t *journal = EXT4_SB(sb)->s_journal;
3177 
3178 	if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3179 		BUG_ON(journal != NULL);
3180 		return;
3181 	}
3182 	jbd2_journal_lock_updates(journal);
3183 	if (jbd2_journal_flush(journal) < 0)
3184 		goto out;
3185 
3186 	lock_super(sb);
3187 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3188 	    sb->s_flags & MS_RDONLY) {
3189 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3190 		sb->s_dirt = 0;
3191 		ext4_commit_super(sb, es, 1);
3192 	}
3193 	unlock_super(sb);
3194 
3195 out:
3196 	jbd2_journal_unlock_updates(journal);
3197 }
3198 
3199 /*
3200  * If we are mounting (or read-write remounting) a filesystem whose journal
3201  * has recorded an error from a previous lifetime, move that error to the
3202  * main filesystem now.
3203  */
3204 static void ext4_clear_journal_err(struct super_block *sb,
3205 				   struct ext4_super_block *es)
3206 {
3207 	journal_t *journal;
3208 	int j_errno;
3209 	const char *errstr;
3210 
3211 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3212 
3213 	journal = EXT4_SB(sb)->s_journal;
3214 
3215 	/*
3216 	 * Now check for any error status which may have been recorded in the
3217 	 * journal by a prior ext4_error() or ext4_abort()
3218 	 */
3219 
3220 	j_errno = jbd2_journal_errno(journal);
3221 	if (j_errno) {
3222 		char nbuf[16];
3223 
3224 		errstr = ext4_decode_error(sb, j_errno, nbuf);
3225 		ext4_warning(sb, __func__, "Filesystem error recorded "
3226 			     "from previous mount: %s", errstr);
3227 		ext4_warning(sb, __func__, "Marking fs in need of "
3228 			     "filesystem check.");
3229 
3230 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3231 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3232 		ext4_commit_super(sb, es, 1);
3233 
3234 		jbd2_journal_clear_err(journal);
3235 	}
3236 }
3237 
3238 /*
3239  * Force the running and committing transactions to commit,
3240  * and wait on the commit.
3241  */
3242 int ext4_force_commit(struct super_block *sb)
3243 {
3244 	journal_t *journal;
3245 	int ret = 0;
3246 
3247 	if (sb->s_flags & MS_RDONLY)
3248 		return 0;
3249 
3250 	journal = EXT4_SB(sb)->s_journal;
3251 	if (journal) {
3252 		sb->s_dirt = 0;
3253 		ret = ext4_journal_force_commit(journal);
3254 	}
3255 
3256 	return ret;
3257 }
3258 
3259 /*
3260  * Ext4 always journals updates to the superblock itself, so we don't
3261  * have to propagate any other updates to the superblock on disk at this
3262  * point.  (We can probably nuke this function altogether, and remove
3263  * any mention to sb->s_dirt in all of fs/ext4; eventual cleanup...)
3264  */
3265 static void ext4_write_super(struct super_block *sb)
3266 {
3267 	if (EXT4_SB(sb)->s_journal) {
3268 		if (mutex_trylock(&sb->s_lock) != 0)
3269 			BUG();
3270 		sb->s_dirt = 0;
3271 	} else {
3272 		ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3273 	}
3274 }
3275 
3276 static int ext4_sync_fs(struct super_block *sb, int wait)
3277 {
3278 	int ret = 0;
3279 	tid_t target;
3280 
3281 	trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
3282 	sb->s_dirt = 0;
3283 	if (EXT4_SB(sb)->s_journal) {
3284 		if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal,
3285 					      &target)) {
3286 			if (wait)
3287 				jbd2_log_wait_commit(EXT4_SB(sb)->s_journal,
3288 						     target);
3289 		}
3290 	} else {
3291 		ext4_commit_super(sb, EXT4_SB(sb)->s_es, wait);
3292 	}
3293 	return ret;
3294 }
3295 
3296 /*
3297  * LVM calls this function before a (read-only) snapshot is created.  This
3298  * gives us a chance to flush the journal completely and mark the fs clean.
3299  */
3300 static int ext4_freeze(struct super_block *sb)
3301 {
3302 	int error = 0;
3303 	journal_t *journal;
3304 	sb->s_dirt = 0;
3305 
3306 	if (!(sb->s_flags & MS_RDONLY)) {
3307 		journal = EXT4_SB(sb)->s_journal;
3308 
3309 		if (journal) {
3310 			/* Now we set up the journal barrier. */
3311 			jbd2_journal_lock_updates(journal);
3312 
3313 			/*
3314 			 * We don't want to clear needs_recovery flag when we
3315 			 * failed to flush the journal.
3316 			 */
3317 			error = jbd2_journal_flush(journal);
3318 			if (error < 0)
3319 				goto out;
3320 		}
3321 
3322 		/* Journal blocked and flushed, clear needs_recovery flag. */
3323 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3324 		error = ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3325 		if (error)
3326 			goto out;
3327 	}
3328 	return 0;
3329 out:
3330 	jbd2_journal_unlock_updates(journal);
3331 	return error;
3332 }
3333 
3334 /*
3335  * Called by LVM after the snapshot is done.  We need to reset the RECOVER
3336  * flag here, even though the filesystem is not technically dirty yet.
3337  */
3338 static int ext4_unfreeze(struct super_block *sb)
3339 {
3340 	if (EXT4_SB(sb)->s_journal && !(sb->s_flags & MS_RDONLY)) {
3341 		lock_super(sb);
3342 		/* Reser the needs_recovery flag before the fs is unlocked. */
3343 		EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3344 		ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3345 		unlock_super(sb);
3346 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3347 	}
3348 	return 0;
3349 }
3350 
3351 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3352 {
3353 	struct ext4_super_block *es;
3354 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3355 	ext4_fsblk_t n_blocks_count = 0;
3356 	unsigned long old_sb_flags;
3357 	struct ext4_mount_options old_opts;
3358 	ext4_group_t g;
3359 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3360 	int err;
3361 #ifdef CONFIG_QUOTA
3362 	int i;
3363 #endif
3364 
3365 	/* Store the original options */
3366 	old_sb_flags = sb->s_flags;
3367 	old_opts.s_mount_opt = sbi->s_mount_opt;
3368 	old_opts.s_resuid = sbi->s_resuid;
3369 	old_opts.s_resgid = sbi->s_resgid;
3370 	old_opts.s_commit_interval = sbi->s_commit_interval;
3371 	old_opts.s_min_batch_time = sbi->s_min_batch_time;
3372 	old_opts.s_max_batch_time = sbi->s_max_batch_time;
3373 #ifdef CONFIG_QUOTA
3374 	old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3375 	for (i = 0; i < MAXQUOTAS; i++)
3376 		old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3377 #endif
3378 	if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3379 		journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3380 
3381 	/*
3382 	 * Allow the "check" option to be passed as a remount option.
3383 	 */
3384 	if (!parse_options(data, sb, NULL, &journal_ioprio,
3385 			   &n_blocks_count, 1)) {
3386 		err = -EINVAL;
3387 		goto restore_opts;
3388 	}
3389 
3390 	if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
3391 		ext4_abort(sb, __func__, "Abort forced by user");
3392 
3393 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3394 		((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3395 
3396 	es = sbi->s_es;
3397 
3398 	if (sbi->s_journal) {
3399 		ext4_init_journal_params(sb, sbi->s_journal);
3400 		set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3401 	}
3402 
3403 	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3404 		n_blocks_count > ext4_blocks_count(es)) {
3405 		if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
3406 			err = -EROFS;
3407 			goto restore_opts;
3408 		}
3409 
3410 		if (*flags & MS_RDONLY) {
3411 			/*
3412 			 * First of all, the unconditional stuff we have to do
3413 			 * to disable replay of the journal when we next remount
3414 			 */
3415 			sb->s_flags |= MS_RDONLY;
3416 
3417 			/*
3418 			 * OK, test if we are remounting a valid rw partition
3419 			 * readonly, and if so set the rdonly flag and then
3420 			 * mark the partition as valid again.
3421 			 */
3422 			if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3423 			    (sbi->s_mount_state & EXT4_VALID_FS))
3424 				es->s_state = cpu_to_le16(sbi->s_mount_state);
3425 
3426 			/*
3427 			 * We have to unlock super so that we can wait for
3428 			 * transactions.
3429 			 */
3430 			if (sbi->s_journal) {
3431 				unlock_super(sb);
3432 				ext4_mark_recovery_complete(sb, es);
3433 				lock_super(sb);
3434 			}
3435 		} else {
3436 			int ret;
3437 			if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3438 					~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3439 				printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3440 				       "remount RDWR because of unsupported "
3441 				       "optional features (%x).\n", sb->s_id,
3442 				(le32_to_cpu(sbi->s_es->s_feature_ro_compat) &
3443 					~EXT4_FEATURE_RO_COMPAT_SUPP));
3444 				err = -EROFS;
3445 				goto restore_opts;
3446 			}
3447 
3448 			/*
3449 			 * Make sure the group descriptor checksums
3450 			 * are sane.  If they aren't, refuse to
3451 			 * remount r/w.
3452 			 */
3453 			for (g = 0; g < sbi->s_groups_count; g++) {
3454 				struct ext4_group_desc *gdp =
3455 					ext4_get_group_desc(sb, g, NULL);
3456 
3457 				if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3458 					printk(KERN_ERR
3459 	       "EXT4-fs: ext4_remount: "
3460 		"Checksum for group %u failed (%u!=%u)\n",
3461 		g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3462 					       le16_to_cpu(gdp->bg_checksum));
3463 					err = -EINVAL;
3464 					goto restore_opts;
3465 				}
3466 			}
3467 
3468 			/*
3469 			 * If we have an unprocessed orphan list hanging
3470 			 * around from a previously readonly bdev mount,
3471 			 * require a full umount/remount for now.
3472 			 */
3473 			if (es->s_last_orphan) {
3474 				printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3475 				       "remount RDWR because of unprocessed "
3476 				       "orphan inode list.  Please "
3477 				       "umount/remount instead.\n",
3478 				       sb->s_id);
3479 				err = -EINVAL;
3480 				goto restore_opts;
3481 			}
3482 
3483 			/*
3484 			 * Mounting a RDONLY partition read-write, so reread
3485 			 * and store the current valid flag.  (It may have
3486 			 * been changed by e2fsck since we originally mounted
3487 			 * the partition.)
3488 			 */
3489 			if (sbi->s_journal)
3490 				ext4_clear_journal_err(sb, es);
3491 			sbi->s_mount_state = le16_to_cpu(es->s_state);
3492 			if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3493 				goto restore_opts;
3494 			if (!ext4_setup_super(sb, es, 0))
3495 				sb->s_flags &= ~MS_RDONLY;
3496 		}
3497 	}
3498 	if (sbi->s_journal == NULL)
3499 		ext4_commit_super(sb, es, 1);
3500 
3501 #ifdef CONFIG_QUOTA
3502 	/* Release old quota file names */
3503 	for (i = 0; i < MAXQUOTAS; i++)
3504 		if (old_opts.s_qf_names[i] &&
3505 		    old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3506 			kfree(old_opts.s_qf_names[i]);
3507 #endif
3508 	return 0;
3509 restore_opts:
3510 	sb->s_flags = old_sb_flags;
3511 	sbi->s_mount_opt = old_opts.s_mount_opt;
3512 	sbi->s_resuid = old_opts.s_resuid;
3513 	sbi->s_resgid = old_opts.s_resgid;
3514 	sbi->s_commit_interval = old_opts.s_commit_interval;
3515 	sbi->s_min_batch_time = old_opts.s_min_batch_time;
3516 	sbi->s_max_batch_time = old_opts.s_max_batch_time;
3517 #ifdef CONFIG_QUOTA
3518 	sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3519 	for (i = 0; i < MAXQUOTAS; i++) {
3520 		if (sbi->s_qf_names[i] &&
3521 		    old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3522 			kfree(sbi->s_qf_names[i]);
3523 		sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3524 	}
3525 #endif
3526 	return err;
3527 }
3528 
3529 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3530 {
3531 	struct super_block *sb = dentry->d_sb;
3532 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3533 	struct ext4_super_block *es = sbi->s_es;
3534 	u64 fsid;
3535 
3536 	if (test_opt(sb, MINIX_DF)) {
3537 		sbi->s_overhead_last = 0;
3538 	} else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3539 		ext4_group_t ngroups = sbi->s_groups_count, i;
3540 		ext4_fsblk_t overhead = 0;
3541 		smp_rmb();
3542 
3543 		/*
3544 		 * Compute the overhead (FS structures).  This is constant
3545 		 * for a given filesystem unless the number of block groups
3546 		 * changes so we cache the previous value until it does.
3547 		 */
3548 
3549 		/*
3550 		 * All of the blocks before first_data_block are
3551 		 * overhead
3552 		 */
3553 		overhead = le32_to_cpu(es->s_first_data_block);
3554 
3555 		/*
3556 		 * Add the overhead attributed to the superblock and
3557 		 * block group descriptors.  If the sparse superblocks
3558 		 * feature is turned on, then not all groups have this.
3559 		 */
3560 		for (i = 0; i < ngroups; i++) {
3561 			overhead += ext4_bg_has_super(sb, i) +
3562 				ext4_bg_num_gdb(sb, i);
3563 			cond_resched();
3564 		}
3565 
3566 		/*
3567 		 * Every block group has an inode bitmap, a block
3568 		 * bitmap, and an inode table.
3569 		 */
3570 		overhead += ngroups * (2 + sbi->s_itb_per_group);
3571 		sbi->s_overhead_last = overhead;
3572 		smp_wmb();
3573 		sbi->s_blocks_last = ext4_blocks_count(es);
3574 	}
3575 
3576 	buf->f_type = EXT4_SUPER_MAGIC;
3577 	buf->f_bsize = sb->s_blocksize;
3578 	buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3579 	buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3580 		       percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3581 	ext4_free_blocks_count_set(es, buf->f_bfree);
3582 	buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3583 	if (buf->f_bfree < ext4_r_blocks_count(es))
3584 		buf->f_bavail = 0;
3585 	buf->f_files = le32_to_cpu(es->s_inodes_count);
3586 	buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3587 	es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3588 	buf->f_namelen = EXT4_NAME_LEN;
3589 	fsid = le64_to_cpup((void *)es->s_uuid) ^
3590 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3591 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3592 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3593 	return 0;
3594 }
3595 
3596 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3597  * is locked for write. Otherwise the are possible deadlocks:
3598  * Process 1                         Process 2
3599  * ext4_create()                     quota_sync()
3600  *   jbd2_journal_start()                  write_dquot()
3601  *   vfs_dq_init()                         down(dqio_mutex)
3602  *     down(dqio_mutex)                    jbd2_journal_start()
3603  *
3604  */
3605 
3606 #ifdef CONFIG_QUOTA
3607 
3608 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3609 {
3610 	return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3611 }
3612 
3613 static int ext4_write_dquot(struct dquot *dquot)
3614 {
3615 	int ret, err;
3616 	handle_t *handle;
3617 	struct inode *inode;
3618 
3619 	inode = dquot_to_inode(dquot);
3620 	handle = ext4_journal_start(inode,
3621 					EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3622 	if (IS_ERR(handle))
3623 		return PTR_ERR(handle);
3624 	ret = dquot_commit(dquot);
3625 	err = ext4_journal_stop(handle);
3626 	if (!ret)
3627 		ret = err;
3628 	return ret;
3629 }
3630 
3631 static int ext4_acquire_dquot(struct dquot *dquot)
3632 {
3633 	int ret, err;
3634 	handle_t *handle;
3635 
3636 	handle = ext4_journal_start(dquot_to_inode(dquot),
3637 					EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3638 	if (IS_ERR(handle))
3639 		return PTR_ERR(handle);
3640 	ret = dquot_acquire(dquot);
3641 	err = ext4_journal_stop(handle);
3642 	if (!ret)
3643 		ret = err;
3644 	return ret;
3645 }
3646 
3647 static int ext4_release_dquot(struct dquot *dquot)
3648 {
3649 	int ret, err;
3650 	handle_t *handle;
3651 
3652 	handle = ext4_journal_start(dquot_to_inode(dquot),
3653 					EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3654 	if (IS_ERR(handle)) {
3655 		/* Release dquot anyway to avoid endless cycle in dqput() */
3656 		dquot_release(dquot);
3657 		return PTR_ERR(handle);
3658 	}
3659 	ret = dquot_release(dquot);
3660 	err = ext4_journal_stop(handle);
3661 	if (!ret)
3662 		ret = err;
3663 	return ret;
3664 }
3665 
3666 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3667 {
3668 	/* Are we journaling quotas? */
3669 	if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3670 	    EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3671 		dquot_mark_dquot_dirty(dquot);
3672 		return ext4_write_dquot(dquot);
3673 	} else {
3674 		return dquot_mark_dquot_dirty(dquot);
3675 	}
3676 }
3677 
3678 static int ext4_write_info(struct super_block *sb, int type)
3679 {
3680 	int ret, err;
3681 	handle_t *handle;
3682 
3683 	/* Data block + inode block */
3684 	handle = ext4_journal_start(sb->s_root->d_inode, 2);
3685 	if (IS_ERR(handle))
3686 		return PTR_ERR(handle);
3687 	ret = dquot_commit_info(sb, type);
3688 	err = ext4_journal_stop(handle);
3689 	if (!ret)
3690 		ret = err;
3691 	return ret;
3692 }
3693 
3694 /*
3695  * Turn on quotas during mount time - we need to find
3696  * the quota file and such...
3697  */
3698 static int ext4_quota_on_mount(struct super_block *sb, int type)
3699 {
3700 	return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3701 			EXT4_SB(sb)->s_jquota_fmt, type);
3702 }
3703 
3704 /*
3705  * Standard function to be called on quota_on
3706  */
3707 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3708 			 char *name, int remount)
3709 {
3710 	int err;
3711 	struct path path;
3712 
3713 	if (!test_opt(sb, QUOTA))
3714 		return -EINVAL;
3715 	/* When remounting, no checks are needed and in fact, name is NULL */
3716 	if (remount)
3717 		return vfs_quota_on(sb, type, format_id, name, remount);
3718 
3719 	err = kern_path(name, LOOKUP_FOLLOW, &path);
3720 	if (err)
3721 		return err;
3722 
3723 	/* Quotafile not on the same filesystem? */
3724 	if (path.mnt->mnt_sb != sb) {
3725 		path_put(&path);
3726 		return -EXDEV;
3727 	}
3728 	/* Journaling quota? */
3729 	if (EXT4_SB(sb)->s_qf_names[type]) {
3730 		/* Quotafile not in fs root? */
3731 		if (path.dentry->d_parent != sb->s_root)
3732 			printk(KERN_WARNING
3733 				"EXT4-fs: Quota file not on filesystem root. "
3734 				"Journaled quota will not work.\n");
3735 	}
3736 
3737 	/*
3738 	 * When we journal data on quota file, we have to flush journal to see
3739 	 * all updates to the file when we bypass pagecache...
3740 	 */
3741 	if (EXT4_SB(sb)->s_journal &&
3742 	    ext4_should_journal_data(path.dentry->d_inode)) {
3743 		/*
3744 		 * We don't need to lock updates but journal_flush() could
3745 		 * otherwise be livelocked...
3746 		 */
3747 		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3748 		err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3749 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3750 		if (err) {
3751 			path_put(&path);
3752 			return err;
3753 		}
3754 	}
3755 
3756 	err = vfs_quota_on_path(sb, type, format_id, &path);
3757 	path_put(&path);
3758 	return err;
3759 }
3760 
3761 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3762  * acquiring the locks... As quota files are never truncated and quota code
3763  * itself serializes the operations (and noone else should touch the files)
3764  * we don't have to be afraid of races */
3765 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3766 			       size_t len, loff_t off)
3767 {
3768 	struct inode *inode = sb_dqopt(sb)->files[type];
3769 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3770 	int err = 0;
3771 	int offset = off & (sb->s_blocksize - 1);
3772 	int tocopy;
3773 	size_t toread;
3774 	struct buffer_head *bh;
3775 	loff_t i_size = i_size_read(inode);
3776 
3777 	if (off > i_size)
3778 		return 0;
3779 	if (off+len > i_size)
3780 		len = i_size-off;
3781 	toread = len;
3782 	while (toread > 0) {
3783 		tocopy = sb->s_blocksize - offset < toread ?
3784 				sb->s_blocksize - offset : toread;
3785 		bh = ext4_bread(NULL, inode, blk, 0, &err);
3786 		if (err)
3787 			return err;
3788 		if (!bh)	/* A hole? */
3789 			memset(data, 0, tocopy);
3790 		else
3791 			memcpy(data, bh->b_data+offset, tocopy);
3792 		brelse(bh);
3793 		offset = 0;
3794 		toread -= tocopy;
3795 		data += tocopy;
3796 		blk++;
3797 	}
3798 	return len;
3799 }
3800 
3801 /* Write to quotafile (we know the transaction is already started and has
3802  * enough credits) */
3803 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3804 				const char *data, size_t len, loff_t off)
3805 {
3806 	struct inode *inode = sb_dqopt(sb)->files[type];
3807 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3808 	int err = 0;
3809 	int offset = off & (sb->s_blocksize - 1);
3810 	int tocopy;
3811 	int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3812 	size_t towrite = len;
3813 	struct buffer_head *bh;
3814 	handle_t *handle = journal_current_handle();
3815 
3816 	if (EXT4_SB(sb)->s_journal && !handle) {
3817 		printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
3818 			" cancelled because transaction is not started.\n",
3819 			(unsigned long long)off, (unsigned long long)len);
3820 		return -EIO;
3821 	}
3822 	mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3823 	while (towrite > 0) {
3824 		tocopy = sb->s_blocksize - offset < towrite ?
3825 				sb->s_blocksize - offset : towrite;
3826 		bh = ext4_bread(handle, inode, blk, 1, &err);
3827 		if (!bh)
3828 			goto out;
3829 		if (journal_quota) {
3830 			err = ext4_journal_get_write_access(handle, bh);
3831 			if (err) {
3832 				brelse(bh);
3833 				goto out;
3834 			}
3835 		}
3836 		lock_buffer(bh);
3837 		memcpy(bh->b_data+offset, data, tocopy);
3838 		flush_dcache_page(bh->b_page);
3839 		unlock_buffer(bh);
3840 		if (journal_quota)
3841 			err = ext4_handle_dirty_metadata(handle, NULL, bh);
3842 		else {
3843 			/* Always do at least ordered writes for quotas */
3844 			err = ext4_jbd2_file_inode(handle, inode);
3845 			mark_buffer_dirty(bh);
3846 		}
3847 		brelse(bh);
3848 		if (err)
3849 			goto out;
3850 		offset = 0;
3851 		towrite -= tocopy;
3852 		data += tocopy;
3853 		blk++;
3854 	}
3855 out:
3856 	if (len == towrite) {
3857 		mutex_unlock(&inode->i_mutex);
3858 		return err;
3859 	}
3860 	if (inode->i_size < off+len-towrite) {
3861 		i_size_write(inode, off+len-towrite);
3862 		EXT4_I(inode)->i_disksize = inode->i_size;
3863 	}
3864 	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3865 	ext4_mark_inode_dirty(handle, inode);
3866 	mutex_unlock(&inode->i_mutex);
3867 	return len - towrite;
3868 }
3869 
3870 #endif
3871 
3872 static int ext4_get_sb(struct file_system_type *fs_type,
3873 	int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3874 {
3875 	return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3876 }
3877 
3878 static struct file_system_type ext4_fs_type = {
3879 	.owner		= THIS_MODULE,
3880 	.name		= "ext4",
3881 	.get_sb		= ext4_get_sb,
3882 	.kill_sb	= kill_block_super,
3883 	.fs_flags	= FS_REQUIRES_DEV,
3884 };
3885 
3886 #ifdef CONFIG_EXT4DEV_COMPAT
3887 static int ext4dev_get_sb(struct file_system_type *fs_type,
3888 	int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3889 {
3890 	printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3891 	       "to mount using ext4\n");
3892 	printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3893 	       "will go away by 2.6.31\n");
3894 	return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3895 }
3896 
3897 static struct file_system_type ext4dev_fs_type = {
3898 	.owner		= THIS_MODULE,
3899 	.name		= "ext4dev",
3900 	.get_sb		= ext4dev_get_sb,
3901 	.kill_sb	= kill_block_super,
3902 	.fs_flags	= FS_REQUIRES_DEV,
3903 };
3904 MODULE_ALIAS("ext4dev");
3905 #endif
3906 
3907 static int __init init_ext4_fs(void)
3908 {
3909 	int err;
3910 
3911 	ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3912 	if (!ext4_kset)
3913 		return -ENOMEM;
3914 	ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3915 	err = init_ext4_mballoc();
3916 	if (err)
3917 		return err;
3918 
3919 	err = init_ext4_xattr();
3920 	if (err)
3921 		goto out2;
3922 	err = init_inodecache();
3923 	if (err)
3924 		goto out1;
3925 	err = register_filesystem(&ext4_fs_type);
3926 	if (err)
3927 		goto out;
3928 #ifdef CONFIG_EXT4DEV_COMPAT
3929 	err = register_filesystem(&ext4dev_fs_type);
3930 	if (err) {
3931 		unregister_filesystem(&ext4_fs_type);
3932 		goto out;
3933 	}
3934 #endif
3935 	return 0;
3936 out:
3937 	destroy_inodecache();
3938 out1:
3939 	exit_ext4_xattr();
3940 out2:
3941 	exit_ext4_mballoc();
3942 	return err;
3943 }
3944 
3945 static void __exit exit_ext4_fs(void)
3946 {
3947 	unregister_filesystem(&ext4_fs_type);
3948 #ifdef CONFIG_EXT4DEV_COMPAT
3949 	unregister_filesystem(&ext4dev_fs_type);
3950 #endif
3951 	destroy_inodecache();
3952 	exit_ext4_xattr();
3953 	exit_ext4_mballoc();
3954 	remove_proc_entry("fs/ext4", NULL);
3955 	kset_unregister(ext4_kset);
3956 }
3957 
3958 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3959 MODULE_DESCRIPTION("Fourth Extended Filesystem");
3960 MODULE_LICENSE("GPL");
3961 module_init(init_ext4_fs)
3962 module_exit(exit_ext4_fs)
3963