xref: /linux/fs/ext4/super.c (revision e27ecdd94d81e5bc3d1f68591701db5adb342f0d)
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/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/smp_lock.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/proc_fs.h>
39 #include <linux/ctype.h>
40 #include <linux/marker.h>
41 #include <linux/log2.h>
42 #include <linux/crc16.h>
43 #include <asm/uaccess.h>
44 
45 #include "ext4.h"
46 #include "ext4_jbd2.h"
47 #include "xattr.h"
48 #include "acl.h"
49 
50 static int default_mb_history_length = 1000;
51 
52 module_param_named(default_mb_history_length, default_mb_history_length,
53 		   int, 0644);
54 MODULE_PARM_DESC(default_mb_history_length,
55 		 "Default number of entries saved for mb_history");
56 
57 struct proc_dir_entry *ext4_proc_root;
58 static struct kset *ext4_kset;
59 
60 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
61 			     unsigned long journal_devnum);
62 static int ext4_commit_super(struct super_block *sb, int sync);
63 static void ext4_mark_recovery_complete(struct super_block *sb,
64 					struct ext4_super_block *es);
65 static void ext4_clear_journal_err(struct super_block *sb,
66 				   struct ext4_super_block *es);
67 static int ext4_sync_fs(struct super_block *sb, int wait);
68 static const char *ext4_decode_error(struct super_block *sb, int errno,
69 				     char nbuf[16]);
70 static int ext4_remount(struct super_block *sb, int *flags, char *data);
71 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
72 static int ext4_unfreeze(struct super_block *sb);
73 static void ext4_write_super(struct super_block *sb);
74 static int ext4_freeze(struct super_block *sb);
75 
76 
77 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
78 			       struct ext4_group_desc *bg)
79 {
80 	return le32_to_cpu(bg->bg_block_bitmap_lo) |
81 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
82 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
83 }
84 
85 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
86 			       struct ext4_group_desc *bg)
87 {
88 	return le32_to_cpu(bg->bg_inode_bitmap_lo) |
89 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
90 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
91 }
92 
93 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
94 			      struct ext4_group_desc *bg)
95 {
96 	return le32_to_cpu(bg->bg_inode_table_lo) |
97 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
98 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
99 }
100 
101 __u32 ext4_free_blks_count(struct super_block *sb,
102 			      struct ext4_group_desc *bg)
103 {
104 	return le16_to_cpu(bg->bg_free_blocks_count_lo) |
105 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
106 		 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
107 }
108 
109 __u32 ext4_free_inodes_count(struct super_block *sb,
110 			      struct ext4_group_desc *bg)
111 {
112 	return le16_to_cpu(bg->bg_free_inodes_count_lo) |
113 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
114 		 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
115 }
116 
117 __u32 ext4_used_dirs_count(struct super_block *sb,
118 			      struct ext4_group_desc *bg)
119 {
120 	return le16_to_cpu(bg->bg_used_dirs_count_lo) |
121 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
122 		 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
123 }
124 
125 __u32 ext4_itable_unused_count(struct super_block *sb,
126 			      struct ext4_group_desc *bg)
127 {
128 	return le16_to_cpu(bg->bg_itable_unused_lo) |
129 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
130 		 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
131 }
132 
133 void ext4_block_bitmap_set(struct super_block *sb,
134 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
135 {
136 	bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
137 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
138 		bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
139 }
140 
141 void ext4_inode_bitmap_set(struct super_block *sb,
142 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
143 {
144 	bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
145 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
146 		bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
147 }
148 
149 void ext4_inode_table_set(struct super_block *sb,
150 			  struct ext4_group_desc *bg, ext4_fsblk_t blk)
151 {
152 	bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
153 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
154 		bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
155 }
156 
157 void ext4_free_blks_set(struct super_block *sb,
158 			  struct ext4_group_desc *bg, __u32 count)
159 {
160 	bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
161 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
162 		bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
163 }
164 
165 void ext4_free_inodes_set(struct super_block *sb,
166 			  struct ext4_group_desc *bg, __u32 count)
167 {
168 	bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
169 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
170 		bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
171 }
172 
173 void ext4_used_dirs_set(struct super_block *sb,
174 			  struct ext4_group_desc *bg, __u32 count)
175 {
176 	bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
177 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
178 		bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
179 }
180 
181 void ext4_itable_unused_set(struct super_block *sb,
182 			  struct ext4_group_desc *bg, __u32 count)
183 {
184 	bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
185 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
186 		bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
187 }
188 
189 /*
190  * Wrappers for jbd2_journal_start/end.
191  *
192  * The only special thing we need to do here is to make sure that all
193  * journal_end calls result in the superblock being marked dirty, so
194  * that sync() will call the filesystem's write_super callback if
195  * appropriate.
196  */
197 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
198 {
199 	journal_t *journal;
200 
201 	if (sb->s_flags & MS_RDONLY)
202 		return ERR_PTR(-EROFS);
203 
204 	/* Special case here: if the journal has aborted behind our
205 	 * backs (eg. EIO in the commit thread), then we still need to
206 	 * take the FS itself readonly cleanly. */
207 	journal = EXT4_SB(sb)->s_journal;
208 	if (journal) {
209 		if (is_journal_aborted(journal)) {
210 			ext4_abort(sb, __func__, "Detected aborted journal");
211 			return ERR_PTR(-EROFS);
212 		}
213 		return jbd2_journal_start(journal, nblocks);
214 	}
215 	/*
216 	 * We're not journaling, return the appropriate indication.
217 	 */
218 	current->journal_info = EXT4_NOJOURNAL_HANDLE;
219 	return current->journal_info;
220 }
221 
222 /*
223  * The only special thing we need to do here is to make sure that all
224  * jbd2_journal_stop calls result in the superblock being marked dirty, so
225  * that sync() will call the filesystem's write_super callback if
226  * appropriate.
227  */
228 int __ext4_journal_stop(const char *where, handle_t *handle)
229 {
230 	struct super_block *sb;
231 	int err;
232 	int rc;
233 
234 	if (!ext4_handle_valid(handle)) {
235 		/*
236 		 * Do this here since we don't call jbd2_journal_stop() in
237 		 * no-journal mode.
238 		 */
239 		current->journal_info = NULL;
240 		return 0;
241 	}
242 	sb = handle->h_transaction->t_journal->j_private;
243 	err = handle->h_err;
244 	rc = jbd2_journal_stop(handle);
245 
246 	if (!err)
247 		err = rc;
248 	if (err)
249 		__ext4_std_error(sb, where, err);
250 	return err;
251 }
252 
253 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
254 		struct buffer_head *bh, handle_t *handle, int err)
255 {
256 	char nbuf[16];
257 	const char *errstr = ext4_decode_error(NULL, err, nbuf);
258 
259 	BUG_ON(!ext4_handle_valid(handle));
260 
261 	if (bh)
262 		BUFFER_TRACE(bh, "abort");
263 
264 	if (!handle->h_err)
265 		handle->h_err = err;
266 
267 	if (is_handle_aborted(handle))
268 		return;
269 
270 	printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
271 	       caller, errstr, err_fn);
272 
273 	jbd2_journal_abort_handle(handle);
274 }
275 
276 /* Deal with the reporting of failure conditions on a filesystem such as
277  * inconsistencies detected or read IO failures.
278  *
279  * On ext2, we can store the error state of the filesystem in the
280  * superblock.  That is not possible on ext4, because we may have other
281  * write ordering constraints on the superblock which prevent us from
282  * writing it out straight away; and given that the journal is about to
283  * be aborted, we can't rely on the current, or future, transactions to
284  * write out the superblock safely.
285  *
286  * We'll just use the jbd2_journal_abort() error code to record an error in
287  * the journal instead.  On recovery, the journal will compain about
288  * that error until we've noted it down and cleared it.
289  */
290 
291 static void ext4_handle_error(struct super_block *sb)
292 {
293 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
294 
295 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
296 	es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
297 
298 	if (sb->s_flags & MS_RDONLY)
299 		return;
300 
301 	if (!test_opt(sb, ERRORS_CONT)) {
302 		journal_t *journal = EXT4_SB(sb)->s_journal;
303 
304 		EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
305 		if (journal)
306 			jbd2_journal_abort(journal, -EIO);
307 	}
308 	if (test_opt(sb, ERRORS_RO)) {
309 		ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
310 		sb->s_flags |= MS_RDONLY;
311 	}
312 	ext4_commit_super(sb, 1);
313 	if (test_opt(sb, ERRORS_PANIC))
314 		panic("EXT4-fs (device %s): panic forced after error\n",
315 			sb->s_id);
316 }
317 
318 void ext4_error(struct super_block *sb, const char *function,
319 		const char *fmt, ...)
320 {
321 	va_list args;
322 
323 	va_start(args, fmt);
324 	printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
325 	vprintk(fmt, args);
326 	printk("\n");
327 	va_end(args);
328 
329 	ext4_handle_error(sb);
330 }
331 
332 static const char *ext4_decode_error(struct super_block *sb, int errno,
333 				     char nbuf[16])
334 {
335 	char *errstr = NULL;
336 
337 	switch (errno) {
338 	case -EIO:
339 		errstr = "IO failure";
340 		break;
341 	case -ENOMEM:
342 		errstr = "Out of memory";
343 		break;
344 	case -EROFS:
345 		if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
346 			errstr = "Journal has aborted";
347 		else
348 			errstr = "Readonly filesystem";
349 		break;
350 	default:
351 		/* If the caller passed in an extra buffer for unknown
352 		 * errors, textualise them now.  Else we just return
353 		 * NULL. */
354 		if (nbuf) {
355 			/* Check for truncated error codes... */
356 			if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
357 				errstr = nbuf;
358 		}
359 		break;
360 	}
361 
362 	return errstr;
363 }
364 
365 /* __ext4_std_error decodes expected errors from journaling functions
366  * automatically and invokes the appropriate error response.  */
367 
368 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
369 {
370 	char nbuf[16];
371 	const char *errstr;
372 
373 	/* Special case: if the error is EROFS, and we're not already
374 	 * inside a transaction, then there's really no point in logging
375 	 * an error. */
376 	if (errno == -EROFS && journal_current_handle() == NULL &&
377 	    (sb->s_flags & MS_RDONLY))
378 		return;
379 
380 	errstr = ext4_decode_error(sb, errno, nbuf);
381 	printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
382 	       sb->s_id, function, errstr);
383 
384 	ext4_handle_error(sb);
385 }
386 
387 /*
388  * ext4_abort is a much stronger failure handler than ext4_error.  The
389  * abort function may be used to deal with unrecoverable failures such
390  * as journal IO errors or ENOMEM at a critical moment in log management.
391  *
392  * We unconditionally force the filesystem into an ABORT|READONLY state,
393  * unless the error response on the fs has been set to panic in which
394  * case we take the easy way out and panic immediately.
395  */
396 
397 void ext4_abort(struct super_block *sb, const char *function,
398 		const char *fmt, ...)
399 {
400 	va_list args;
401 
402 	va_start(args, fmt);
403 	printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
404 	vprintk(fmt, args);
405 	printk("\n");
406 	va_end(args);
407 
408 	if (test_opt(sb, ERRORS_PANIC))
409 		panic("EXT4-fs panic from previous error\n");
410 
411 	if (sb->s_flags & MS_RDONLY)
412 		return;
413 
414 	ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
415 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
416 	sb->s_flags |= MS_RDONLY;
417 	EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
418 	if (EXT4_SB(sb)->s_journal)
419 		jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
420 }
421 
422 void ext4_msg (struct super_block * sb, const char *prefix,
423 		   const char *fmt, ...)
424 {
425 	va_list args;
426 
427 	va_start(args, fmt);
428 	printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
429 	vprintk(fmt, args);
430 	printk("\n");
431 	va_end(args);
432 }
433 
434 void ext4_warning(struct super_block *sb, const char *function,
435 		  const char *fmt, ...)
436 {
437 	va_list args;
438 
439 	va_start(args, fmt);
440 	printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
441 	       sb->s_id, function);
442 	vprintk(fmt, args);
443 	printk("\n");
444 	va_end(args);
445 }
446 
447 void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
448 			   const char *function, const char *fmt, ...)
449 __releases(bitlock)
450 __acquires(bitlock)
451 {
452 	va_list args;
453 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
454 
455 	va_start(args, fmt);
456 	printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
457 	vprintk(fmt, args);
458 	printk("\n");
459 	va_end(args);
460 
461 	if (test_opt(sb, ERRORS_CONT)) {
462 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
463 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
464 		ext4_commit_super(sb, 0);
465 		return;
466 	}
467 	ext4_unlock_group(sb, grp);
468 	ext4_handle_error(sb);
469 	/*
470 	 * We only get here in the ERRORS_RO case; relocking the group
471 	 * may be dangerous, but nothing bad will happen since the
472 	 * filesystem will have already been marked read/only and the
473 	 * journal has been aborted.  We return 1 as a hint to callers
474 	 * who might what to use the return value from
475 	 * ext4_grp_locked_error() to distinguish beween the
476 	 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
477 	 * aggressively from the ext4 function in question, with a
478 	 * more appropriate error code.
479 	 */
480 	ext4_lock_group(sb, grp);
481 	return;
482 }
483 
484 void ext4_update_dynamic_rev(struct super_block *sb)
485 {
486 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
487 
488 	if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
489 		return;
490 
491 	ext4_warning(sb, __func__,
492 		     "updating to rev %d because of new feature flag, "
493 		     "running e2fsck is recommended",
494 		     EXT4_DYNAMIC_REV);
495 
496 	es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
497 	es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
498 	es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
499 	/* leave es->s_feature_*compat flags alone */
500 	/* es->s_uuid will be set by e2fsck if empty */
501 
502 	/*
503 	 * The rest of the superblock fields should be zero, and if not it
504 	 * means they are likely already in use, so leave them alone.  We
505 	 * can leave it up to e2fsck to clean up any inconsistencies there.
506 	 */
507 }
508 
509 /*
510  * Open the external journal device
511  */
512 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
513 {
514 	struct block_device *bdev;
515 	char b[BDEVNAME_SIZE];
516 
517 	bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
518 	if (IS_ERR(bdev))
519 		goto fail;
520 	return bdev;
521 
522 fail:
523 	ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
524 			__bdevname(dev, b), PTR_ERR(bdev));
525 	return NULL;
526 }
527 
528 /*
529  * Release the journal device
530  */
531 static int ext4_blkdev_put(struct block_device *bdev)
532 {
533 	bd_release(bdev);
534 	return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
535 }
536 
537 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
538 {
539 	struct block_device *bdev;
540 	int ret = -ENODEV;
541 
542 	bdev = sbi->journal_bdev;
543 	if (bdev) {
544 		ret = ext4_blkdev_put(bdev);
545 		sbi->journal_bdev = NULL;
546 	}
547 	return ret;
548 }
549 
550 static inline struct inode *orphan_list_entry(struct list_head *l)
551 {
552 	return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
553 }
554 
555 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
556 {
557 	struct list_head *l;
558 
559 	ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
560 		 le32_to_cpu(sbi->s_es->s_last_orphan));
561 
562 	printk(KERN_ERR "sb_info orphan list:\n");
563 	list_for_each(l, &sbi->s_orphan) {
564 		struct inode *inode = orphan_list_entry(l);
565 		printk(KERN_ERR "  "
566 		       "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
567 		       inode->i_sb->s_id, inode->i_ino, inode,
568 		       inode->i_mode, inode->i_nlink,
569 		       NEXT_ORPHAN(inode));
570 	}
571 }
572 
573 static void ext4_put_super(struct super_block *sb)
574 {
575 	struct ext4_sb_info *sbi = EXT4_SB(sb);
576 	struct ext4_super_block *es = sbi->s_es;
577 	int i, err;
578 
579 	lock_super(sb);
580 	lock_kernel();
581 	if (sb->s_dirt)
582 		ext4_commit_super(sb, 1);
583 
584 	ext4_release_system_zone(sb);
585 	ext4_mb_release(sb);
586 	ext4_ext_release(sb);
587 	ext4_xattr_put_super(sb);
588 	if (sbi->s_journal) {
589 		err = jbd2_journal_destroy(sbi->s_journal);
590 		sbi->s_journal = NULL;
591 		if (err < 0)
592 			ext4_abort(sb, __func__,
593 				   "Couldn't clean up the journal");
594 	}
595 	if (!(sb->s_flags & MS_RDONLY)) {
596 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
597 		es->s_state = cpu_to_le16(sbi->s_mount_state);
598 		ext4_commit_super(sb, 1);
599 	}
600 	if (sbi->s_proc) {
601 		remove_proc_entry(sb->s_id, ext4_proc_root);
602 	}
603 	kobject_del(&sbi->s_kobj);
604 
605 	for (i = 0; i < sbi->s_gdb_count; i++)
606 		brelse(sbi->s_group_desc[i]);
607 	kfree(sbi->s_group_desc);
608 	if (is_vmalloc_addr(sbi->s_flex_groups))
609 		vfree(sbi->s_flex_groups);
610 	else
611 		kfree(sbi->s_flex_groups);
612 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
613 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
614 	percpu_counter_destroy(&sbi->s_dirs_counter);
615 	percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
616 	brelse(sbi->s_sbh);
617 #ifdef CONFIG_QUOTA
618 	for (i = 0; i < MAXQUOTAS; i++)
619 		kfree(sbi->s_qf_names[i]);
620 #endif
621 
622 	/* Debugging code just in case the in-memory inode orphan list
623 	 * isn't empty.  The on-disk one can be non-empty if we've
624 	 * detected an error and taken the fs readonly, but the
625 	 * in-memory list had better be clean by this point. */
626 	if (!list_empty(&sbi->s_orphan))
627 		dump_orphan_list(sb, sbi);
628 	J_ASSERT(list_empty(&sbi->s_orphan));
629 
630 	invalidate_bdev(sb->s_bdev);
631 	if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
632 		/*
633 		 * Invalidate the journal device's buffers.  We don't want them
634 		 * floating about in memory - the physical journal device may
635 		 * hotswapped, and it breaks the `ro-after' testing code.
636 		 */
637 		sync_blockdev(sbi->journal_bdev);
638 		invalidate_bdev(sbi->journal_bdev);
639 		ext4_blkdev_remove(sbi);
640 	}
641 	sb->s_fs_info = NULL;
642 	/*
643 	 * Now that we are completely done shutting down the
644 	 * superblock, we need to actually destroy the kobject.
645 	 */
646 	unlock_kernel();
647 	unlock_super(sb);
648 	kobject_put(&sbi->s_kobj);
649 	wait_for_completion(&sbi->s_kobj_unregister);
650 	kfree(sbi->s_blockgroup_lock);
651 	kfree(sbi);
652 }
653 
654 static struct kmem_cache *ext4_inode_cachep;
655 
656 /*
657  * Called inside transaction, so use GFP_NOFS
658  */
659 static struct inode *ext4_alloc_inode(struct super_block *sb)
660 {
661 	struct ext4_inode_info *ei;
662 
663 	ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
664 	if (!ei)
665 		return NULL;
666 
667 #ifdef CONFIG_EXT4_FS_POSIX_ACL
668 	ei->i_acl = EXT4_ACL_NOT_CACHED;
669 	ei->i_default_acl = EXT4_ACL_NOT_CACHED;
670 #endif
671 	ei->vfs_inode.i_version = 1;
672 	ei->vfs_inode.i_data.writeback_index = 0;
673 	memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
674 	INIT_LIST_HEAD(&ei->i_prealloc_list);
675 	spin_lock_init(&ei->i_prealloc_lock);
676 	/*
677 	 * Note:  We can be called before EXT4_SB(sb)->s_journal is set,
678 	 * therefore it can be null here.  Don't check it, just initialize
679 	 * jinode.
680 	 */
681 	jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
682 	ei->i_reserved_data_blocks = 0;
683 	ei->i_reserved_meta_blocks = 0;
684 	ei->i_allocated_meta_blocks = 0;
685 	ei->i_delalloc_reserved_flag = 0;
686 	spin_lock_init(&(ei->i_block_reservation_lock));
687 
688 	return &ei->vfs_inode;
689 }
690 
691 static void ext4_destroy_inode(struct inode *inode)
692 {
693 	if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
694 		ext4_msg(inode->i_sb, KERN_ERR,
695 			 "Inode %lu (%p): orphan list check failed!",
696 			 inode->i_ino, EXT4_I(inode));
697 		print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
698 				EXT4_I(inode), sizeof(struct ext4_inode_info),
699 				true);
700 		dump_stack();
701 	}
702 	kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
703 }
704 
705 static void init_once(void *foo)
706 {
707 	struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
708 
709 	INIT_LIST_HEAD(&ei->i_orphan);
710 #ifdef CONFIG_EXT4_FS_XATTR
711 	init_rwsem(&ei->xattr_sem);
712 #endif
713 	init_rwsem(&ei->i_data_sem);
714 	inode_init_once(&ei->vfs_inode);
715 }
716 
717 static int init_inodecache(void)
718 {
719 	ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
720 					     sizeof(struct ext4_inode_info),
721 					     0, (SLAB_RECLAIM_ACCOUNT|
722 						SLAB_MEM_SPREAD),
723 					     init_once);
724 	if (ext4_inode_cachep == NULL)
725 		return -ENOMEM;
726 	return 0;
727 }
728 
729 static void destroy_inodecache(void)
730 {
731 	kmem_cache_destroy(ext4_inode_cachep);
732 }
733 
734 static void ext4_clear_inode(struct inode *inode)
735 {
736 #ifdef CONFIG_EXT4_FS_POSIX_ACL
737 	if (EXT4_I(inode)->i_acl &&
738 			EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
739 		posix_acl_release(EXT4_I(inode)->i_acl);
740 		EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
741 	}
742 	if (EXT4_I(inode)->i_default_acl &&
743 			EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
744 		posix_acl_release(EXT4_I(inode)->i_default_acl);
745 		EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
746 	}
747 #endif
748 	ext4_discard_preallocations(inode);
749 	if (EXT4_JOURNAL(inode))
750 		jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
751 				       &EXT4_I(inode)->jinode);
752 }
753 
754 static inline void ext4_show_quota_options(struct seq_file *seq,
755 					   struct super_block *sb)
756 {
757 #if defined(CONFIG_QUOTA)
758 	struct ext4_sb_info *sbi = EXT4_SB(sb);
759 
760 	if (sbi->s_jquota_fmt)
761 		seq_printf(seq, ",jqfmt=%s",
762 		(sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
763 
764 	if (sbi->s_qf_names[USRQUOTA])
765 		seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
766 
767 	if (sbi->s_qf_names[GRPQUOTA])
768 		seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
769 
770 	if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
771 		seq_puts(seq, ",usrquota");
772 
773 	if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
774 		seq_puts(seq, ",grpquota");
775 #endif
776 }
777 
778 /*
779  * Show an option if
780  *  - it's set to a non-default value OR
781  *  - if the per-sb default is different from the global default
782  */
783 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
784 {
785 	int def_errors;
786 	unsigned long def_mount_opts;
787 	struct super_block *sb = vfs->mnt_sb;
788 	struct ext4_sb_info *sbi = EXT4_SB(sb);
789 	struct ext4_super_block *es = sbi->s_es;
790 
791 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
792 	def_errors     = le16_to_cpu(es->s_errors);
793 
794 	if (sbi->s_sb_block != 1)
795 		seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
796 	if (test_opt(sb, MINIX_DF))
797 		seq_puts(seq, ",minixdf");
798 	if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
799 		seq_puts(seq, ",grpid");
800 	if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
801 		seq_puts(seq, ",nogrpid");
802 	if (sbi->s_resuid != EXT4_DEF_RESUID ||
803 	    le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
804 		seq_printf(seq, ",resuid=%u", sbi->s_resuid);
805 	}
806 	if (sbi->s_resgid != EXT4_DEF_RESGID ||
807 	    le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
808 		seq_printf(seq, ",resgid=%u", sbi->s_resgid);
809 	}
810 	if (test_opt(sb, ERRORS_RO)) {
811 		if (def_errors == EXT4_ERRORS_PANIC ||
812 		    def_errors == EXT4_ERRORS_CONTINUE) {
813 			seq_puts(seq, ",errors=remount-ro");
814 		}
815 	}
816 	if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
817 		seq_puts(seq, ",errors=continue");
818 	if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
819 		seq_puts(seq, ",errors=panic");
820 	if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
821 		seq_puts(seq, ",nouid32");
822 	if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
823 		seq_puts(seq, ",debug");
824 	if (test_opt(sb, OLDALLOC))
825 		seq_puts(seq, ",oldalloc");
826 #ifdef CONFIG_EXT4_FS_XATTR
827 	if (test_opt(sb, XATTR_USER) &&
828 		!(def_mount_opts & EXT4_DEFM_XATTR_USER))
829 		seq_puts(seq, ",user_xattr");
830 	if (!test_opt(sb, XATTR_USER) &&
831 	    (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
832 		seq_puts(seq, ",nouser_xattr");
833 	}
834 #endif
835 #ifdef CONFIG_EXT4_FS_POSIX_ACL
836 	if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
837 		seq_puts(seq, ",acl");
838 	if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
839 		seq_puts(seq, ",noacl");
840 #endif
841 	if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
842 		seq_printf(seq, ",commit=%u",
843 			   (unsigned) (sbi->s_commit_interval / HZ));
844 	}
845 	if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
846 		seq_printf(seq, ",min_batch_time=%u",
847 			   (unsigned) sbi->s_min_batch_time);
848 	}
849 	if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
850 		seq_printf(seq, ",max_batch_time=%u",
851 			   (unsigned) sbi->s_min_batch_time);
852 	}
853 
854 	/*
855 	 * We're changing the default of barrier mount option, so
856 	 * let's always display its mount state so it's clear what its
857 	 * status is.
858 	 */
859 	seq_puts(seq, ",barrier=");
860 	seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
861 	if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
862 		seq_puts(seq, ",journal_async_commit");
863 	if (test_opt(sb, NOBH))
864 		seq_puts(seq, ",nobh");
865 	if (test_opt(sb, I_VERSION))
866 		seq_puts(seq, ",i_version");
867 	if (!test_opt(sb, DELALLOC))
868 		seq_puts(seq, ",nodelalloc");
869 
870 
871 	if (sbi->s_stripe)
872 		seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
873 	/*
874 	 * journal mode get enabled in different ways
875 	 * So just print the value even if we didn't specify it
876 	 */
877 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
878 		seq_puts(seq, ",data=journal");
879 	else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
880 		seq_puts(seq, ",data=ordered");
881 	else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
882 		seq_puts(seq, ",data=writeback");
883 
884 	if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
885 		seq_printf(seq, ",inode_readahead_blks=%u",
886 			   sbi->s_inode_readahead_blks);
887 
888 	if (test_opt(sb, DATA_ERR_ABORT))
889 		seq_puts(seq, ",data_err=abort");
890 
891 	if (test_opt(sb, NO_AUTO_DA_ALLOC))
892 		seq_puts(seq, ",noauto_da_alloc");
893 
894 	ext4_show_quota_options(seq, sb);
895 
896 	return 0;
897 }
898 
899 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
900 					u64 ino, u32 generation)
901 {
902 	struct inode *inode;
903 
904 	if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
905 		return ERR_PTR(-ESTALE);
906 	if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
907 		return ERR_PTR(-ESTALE);
908 
909 	/* iget isn't really right if the inode is currently unallocated!!
910 	 *
911 	 * ext4_read_inode will return a bad_inode if the inode had been
912 	 * deleted, so we should be safe.
913 	 *
914 	 * Currently we don't know the generation for parent directory, so
915 	 * a generation of 0 means "accept any"
916 	 */
917 	inode = ext4_iget(sb, ino);
918 	if (IS_ERR(inode))
919 		return ERR_CAST(inode);
920 	if (generation && inode->i_generation != generation) {
921 		iput(inode);
922 		return ERR_PTR(-ESTALE);
923 	}
924 
925 	return inode;
926 }
927 
928 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
929 					int fh_len, int fh_type)
930 {
931 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
932 				    ext4_nfs_get_inode);
933 }
934 
935 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
936 					int fh_len, int fh_type)
937 {
938 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
939 				    ext4_nfs_get_inode);
940 }
941 
942 /*
943  * Try to release metadata pages (indirect blocks, directories) which are
944  * mapped via the block device.  Since these pages could have journal heads
945  * which would prevent try_to_free_buffers() from freeing them, we must use
946  * jbd2 layer's try_to_free_buffers() function to release them.
947  */
948 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
949 				 gfp_t wait)
950 {
951 	journal_t *journal = EXT4_SB(sb)->s_journal;
952 
953 	WARN_ON(PageChecked(page));
954 	if (!page_has_buffers(page))
955 		return 0;
956 	if (journal)
957 		return jbd2_journal_try_to_free_buffers(journal, page,
958 							wait & ~__GFP_WAIT);
959 	return try_to_free_buffers(page);
960 }
961 
962 #ifdef CONFIG_QUOTA
963 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
964 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
965 
966 static int ext4_write_dquot(struct dquot *dquot);
967 static int ext4_acquire_dquot(struct dquot *dquot);
968 static int ext4_release_dquot(struct dquot *dquot);
969 static int ext4_mark_dquot_dirty(struct dquot *dquot);
970 static int ext4_write_info(struct super_block *sb, int type);
971 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
972 				char *path, int remount);
973 static int ext4_quota_on_mount(struct super_block *sb, int type);
974 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
975 			       size_t len, loff_t off);
976 static ssize_t ext4_quota_write(struct super_block *sb, int type,
977 				const char *data, size_t len, loff_t off);
978 
979 static struct dquot_operations ext4_quota_operations = {
980 	.initialize	= dquot_initialize,
981 	.drop		= dquot_drop,
982 	.alloc_space	= dquot_alloc_space,
983 	.reserve_space	= dquot_reserve_space,
984 	.claim_space	= dquot_claim_space,
985 	.release_rsv	= dquot_release_reserved_space,
986 	.get_reserved_space = ext4_get_reserved_space,
987 	.alloc_inode	= dquot_alloc_inode,
988 	.free_space	= dquot_free_space,
989 	.free_inode	= dquot_free_inode,
990 	.transfer	= dquot_transfer,
991 	.write_dquot	= ext4_write_dquot,
992 	.acquire_dquot	= ext4_acquire_dquot,
993 	.release_dquot	= ext4_release_dquot,
994 	.mark_dirty	= ext4_mark_dquot_dirty,
995 	.write_info	= ext4_write_info,
996 	.alloc_dquot	= dquot_alloc,
997 	.destroy_dquot	= dquot_destroy,
998 };
999 
1000 static struct quotactl_ops ext4_qctl_operations = {
1001 	.quota_on	= ext4_quota_on,
1002 	.quota_off	= vfs_quota_off,
1003 	.quota_sync	= vfs_quota_sync,
1004 	.get_info	= vfs_get_dqinfo,
1005 	.set_info	= vfs_set_dqinfo,
1006 	.get_dqblk	= vfs_get_dqblk,
1007 	.set_dqblk	= vfs_set_dqblk
1008 };
1009 #endif
1010 
1011 static const struct super_operations ext4_sops = {
1012 	.alloc_inode	= ext4_alloc_inode,
1013 	.destroy_inode	= ext4_destroy_inode,
1014 	.write_inode	= ext4_write_inode,
1015 	.dirty_inode	= ext4_dirty_inode,
1016 	.delete_inode	= ext4_delete_inode,
1017 	.put_super	= ext4_put_super,
1018 	.sync_fs	= ext4_sync_fs,
1019 	.freeze_fs	= ext4_freeze,
1020 	.unfreeze_fs	= ext4_unfreeze,
1021 	.statfs		= ext4_statfs,
1022 	.remount_fs	= ext4_remount,
1023 	.clear_inode	= ext4_clear_inode,
1024 	.show_options	= ext4_show_options,
1025 #ifdef CONFIG_QUOTA
1026 	.quota_read	= ext4_quota_read,
1027 	.quota_write	= ext4_quota_write,
1028 #endif
1029 	.bdev_try_to_free_page = bdev_try_to_free_page,
1030 };
1031 
1032 static const struct super_operations ext4_nojournal_sops = {
1033 	.alloc_inode	= ext4_alloc_inode,
1034 	.destroy_inode	= ext4_destroy_inode,
1035 	.write_inode	= ext4_write_inode,
1036 	.dirty_inode	= ext4_dirty_inode,
1037 	.delete_inode	= ext4_delete_inode,
1038 	.write_super	= ext4_write_super,
1039 	.put_super	= ext4_put_super,
1040 	.statfs		= ext4_statfs,
1041 	.remount_fs	= ext4_remount,
1042 	.clear_inode	= ext4_clear_inode,
1043 	.show_options	= ext4_show_options,
1044 #ifdef CONFIG_QUOTA
1045 	.quota_read	= ext4_quota_read,
1046 	.quota_write	= ext4_quota_write,
1047 #endif
1048 	.bdev_try_to_free_page = bdev_try_to_free_page,
1049 };
1050 
1051 static const struct export_operations ext4_export_ops = {
1052 	.fh_to_dentry = ext4_fh_to_dentry,
1053 	.fh_to_parent = ext4_fh_to_parent,
1054 	.get_parent = ext4_get_parent,
1055 };
1056 
1057 enum {
1058 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1059 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1060 	Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1061 	Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1062 	Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1063 	Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1064 	Opt_journal_update, Opt_journal_dev,
1065 	Opt_journal_checksum, Opt_journal_async_commit,
1066 	Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1067 	Opt_data_err_abort, Opt_data_err_ignore, Opt_mb_history_length,
1068 	Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1069 	Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1070 	Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1071 	Opt_usrquota, Opt_grpquota, Opt_i_version,
1072 	Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1073 	Opt_block_validity, Opt_noblock_validity,
1074 	Opt_inode_readahead_blks, Opt_journal_ioprio
1075 };
1076 
1077 static const match_table_t tokens = {
1078 	{Opt_bsd_df, "bsddf"},
1079 	{Opt_minix_df, "minixdf"},
1080 	{Opt_grpid, "grpid"},
1081 	{Opt_grpid, "bsdgroups"},
1082 	{Opt_nogrpid, "nogrpid"},
1083 	{Opt_nogrpid, "sysvgroups"},
1084 	{Opt_resgid, "resgid=%u"},
1085 	{Opt_resuid, "resuid=%u"},
1086 	{Opt_sb, "sb=%u"},
1087 	{Opt_err_cont, "errors=continue"},
1088 	{Opt_err_panic, "errors=panic"},
1089 	{Opt_err_ro, "errors=remount-ro"},
1090 	{Opt_nouid32, "nouid32"},
1091 	{Opt_debug, "debug"},
1092 	{Opt_oldalloc, "oldalloc"},
1093 	{Opt_orlov, "orlov"},
1094 	{Opt_user_xattr, "user_xattr"},
1095 	{Opt_nouser_xattr, "nouser_xattr"},
1096 	{Opt_acl, "acl"},
1097 	{Opt_noacl, "noacl"},
1098 	{Opt_noload, "noload"},
1099 	{Opt_nobh, "nobh"},
1100 	{Opt_bh, "bh"},
1101 	{Opt_commit, "commit=%u"},
1102 	{Opt_min_batch_time, "min_batch_time=%u"},
1103 	{Opt_max_batch_time, "max_batch_time=%u"},
1104 	{Opt_journal_update, "journal=update"},
1105 	{Opt_journal_dev, "journal_dev=%u"},
1106 	{Opt_journal_checksum, "journal_checksum"},
1107 	{Opt_journal_async_commit, "journal_async_commit"},
1108 	{Opt_abort, "abort"},
1109 	{Opt_data_journal, "data=journal"},
1110 	{Opt_data_ordered, "data=ordered"},
1111 	{Opt_data_writeback, "data=writeback"},
1112 	{Opt_data_err_abort, "data_err=abort"},
1113 	{Opt_data_err_ignore, "data_err=ignore"},
1114 	{Opt_mb_history_length, "mb_history_length=%u"},
1115 	{Opt_offusrjquota, "usrjquota="},
1116 	{Opt_usrjquota, "usrjquota=%s"},
1117 	{Opt_offgrpjquota, "grpjquota="},
1118 	{Opt_grpjquota, "grpjquota=%s"},
1119 	{Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1120 	{Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1121 	{Opt_grpquota, "grpquota"},
1122 	{Opt_noquota, "noquota"},
1123 	{Opt_quota, "quota"},
1124 	{Opt_usrquota, "usrquota"},
1125 	{Opt_barrier, "barrier=%u"},
1126 	{Opt_barrier, "barrier"},
1127 	{Opt_nobarrier, "nobarrier"},
1128 	{Opt_i_version, "i_version"},
1129 	{Opt_stripe, "stripe=%u"},
1130 	{Opt_resize, "resize"},
1131 	{Opt_delalloc, "delalloc"},
1132 	{Opt_nodelalloc, "nodelalloc"},
1133 	{Opt_block_validity, "block_validity"},
1134 	{Opt_noblock_validity, "noblock_validity"},
1135 	{Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1136 	{Opt_journal_ioprio, "journal_ioprio=%u"},
1137 	{Opt_auto_da_alloc, "auto_da_alloc=%u"},
1138 	{Opt_auto_da_alloc, "auto_da_alloc"},
1139 	{Opt_noauto_da_alloc, "noauto_da_alloc"},
1140 	{Opt_err, NULL},
1141 };
1142 
1143 static ext4_fsblk_t get_sb_block(void **data)
1144 {
1145 	ext4_fsblk_t	sb_block;
1146 	char		*options = (char *) *data;
1147 
1148 	if (!options || strncmp(options, "sb=", 3) != 0)
1149 		return 1;	/* Default location */
1150 
1151 	options += 3;
1152 	/* TODO: use simple_strtoll with >32bit ext4 */
1153 	sb_block = simple_strtoul(options, &options, 0);
1154 	if (*options && *options != ',') {
1155 		printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1156 		       (char *) *data);
1157 		return 1;
1158 	}
1159 	if (*options == ',')
1160 		options++;
1161 	*data = (void *) options;
1162 
1163 	return sb_block;
1164 }
1165 
1166 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1167 
1168 static int parse_options(char *options, struct super_block *sb,
1169 			 unsigned long *journal_devnum,
1170 			 unsigned int *journal_ioprio,
1171 			 ext4_fsblk_t *n_blocks_count, int is_remount)
1172 {
1173 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1174 	char *p;
1175 	substring_t args[MAX_OPT_ARGS];
1176 	int data_opt = 0;
1177 	int option;
1178 #ifdef CONFIG_QUOTA
1179 	int qtype, qfmt;
1180 	char *qname;
1181 #endif
1182 
1183 	if (!options)
1184 		return 1;
1185 
1186 	while ((p = strsep(&options, ",")) != NULL) {
1187 		int token;
1188 		if (!*p)
1189 			continue;
1190 
1191 		token = match_token(p, tokens, args);
1192 		switch (token) {
1193 		case Opt_bsd_df:
1194 			clear_opt(sbi->s_mount_opt, MINIX_DF);
1195 			break;
1196 		case Opt_minix_df:
1197 			set_opt(sbi->s_mount_opt, MINIX_DF);
1198 			break;
1199 		case Opt_grpid:
1200 			set_opt(sbi->s_mount_opt, GRPID);
1201 			break;
1202 		case Opt_nogrpid:
1203 			clear_opt(sbi->s_mount_opt, GRPID);
1204 			break;
1205 		case Opt_resuid:
1206 			if (match_int(&args[0], &option))
1207 				return 0;
1208 			sbi->s_resuid = option;
1209 			break;
1210 		case Opt_resgid:
1211 			if (match_int(&args[0], &option))
1212 				return 0;
1213 			sbi->s_resgid = option;
1214 			break;
1215 		case Opt_sb:
1216 			/* handled by get_sb_block() instead of here */
1217 			/* *sb_block = match_int(&args[0]); */
1218 			break;
1219 		case Opt_err_panic:
1220 			clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1221 			clear_opt(sbi->s_mount_opt, ERRORS_RO);
1222 			set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1223 			break;
1224 		case Opt_err_ro:
1225 			clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1226 			clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1227 			set_opt(sbi->s_mount_opt, ERRORS_RO);
1228 			break;
1229 		case Opt_err_cont:
1230 			clear_opt(sbi->s_mount_opt, ERRORS_RO);
1231 			clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1232 			set_opt(sbi->s_mount_opt, ERRORS_CONT);
1233 			break;
1234 		case Opt_nouid32:
1235 			set_opt(sbi->s_mount_opt, NO_UID32);
1236 			break;
1237 		case Opt_debug:
1238 			set_opt(sbi->s_mount_opt, DEBUG);
1239 			break;
1240 		case Opt_oldalloc:
1241 			set_opt(sbi->s_mount_opt, OLDALLOC);
1242 			break;
1243 		case Opt_orlov:
1244 			clear_opt(sbi->s_mount_opt, OLDALLOC);
1245 			break;
1246 #ifdef CONFIG_EXT4_FS_XATTR
1247 		case Opt_user_xattr:
1248 			set_opt(sbi->s_mount_opt, XATTR_USER);
1249 			break;
1250 		case Opt_nouser_xattr:
1251 			clear_opt(sbi->s_mount_opt, XATTR_USER);
1252 			break;
1253 #else
1254 		case Opt_user_xattr:
1255 		case Opt_nouser_xattr:
1256 			ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1257 			break;
1258 #endif
1259 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1260 		case Opt_acl:
1261 			set_opt(sbi->s_mount_opt, POSIX_ACL);
1262 			break;
1263 		case Opt_noacl:
1264 			clear_opt(sbi->s_mount_opt, POSIX_ACL);
1265 			break;
1266 #else
1267 		case Opt_acl:
1268 		case Opt_noacl:
1269 			ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1270 			break;
1271 #endif
1272 		case Opt_journal_update:
1273 			/* @@@ FIXME */
1274 			/* Eventually we will want to be able to create
1275 			   a journal file here.  For now, only allow the
1276 			   user to specify an existing inode to be the
1277 			   journal file. */
1278 			if (is_remount) {
1279 				ext4_msg(sb, KERN_ERR,
1280 					 "Cannot specify journal on remount");
1281 				return 0;
1282 			}
1283 			set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1284 			break;
1285 		case Opt_journal_dev:
1286 			if (is_remount) {
1287 				ext4_msg(sb, KERN_ERR,
1288 					"Cannot specify journal on remount");
1289 				return 0;
1290 			}
1291 			if (match_int(&args[0], &option))
1292 				return 0;
1293 			*journal_devnum = option;
1294 			break;
1295 		case Opt_journal_checksum:
1296 			set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1297 			break;
1298 		case Opt_journal_async_commit:
1299 			set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1300 			set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1301 			break;
1302 		case Opt_noload:
1303 			set_opt(sbi->s_mount_opt, NOLOAD);
1304 			break;
1305 		case Opt_commit:
1306 			if (match_int(&args[0], &option))
1307 				return 0;
1308 			if (option < 0)
1309 				return 0;
1310 			if (option == 0)
1311 				option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1312 			sbi->s_commit_interval = HZ * option;
1313 			break;
1314 		case Opt_max_batch_time:
1315 			if (match_int(&args[0], &option))
1316 				return 0;
1317 			if (option < 0)
1318 				return 0;
1319 			if (option == 0)
1320 				option = EXT4_DEF_MAX_BATCH_TIME;
1321 			sbi->s_max_batch_time = option;
1322 			break;
1323 		case Opt_min_batch_time:
1324 			if (match_int(&args[0], &option))
1325 				return 0;
1326 			if (option < 0)
1327 				return 0;
1328 			sbi->s_min_batch_time = option;
1329 			break;
1330 		case Opt_data_journal:
1331 			data_opt = EXT4_MOUNT_JOURNAL_DATA;
1332 			goto datacheck;
1333 		case Opt_data_ordered:
1334 			data_opt = EXT4_MOUNT_ORDERED_DATA;
1335 			goto datacheck;
1336 		case Opt_data_writeback:
1337 			data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1338 		datacheck:
1339 			if (is_remount) {
1340 				if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1341 						!= data_opt) {
1342 					ext4_msg(sb, KERN_ERR,
1343 						"Cannot change data mode on remount");
1344 					return 0;
1345 				}
1346 			} else {
1347 				sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1348 				sbi->s_mount_opt |= data_opt;
1349 			}
1350 			break;
1351 		case Opt_data_err_abort:
1352 			set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1353 			break;
1354 		case Opt_data_err_ignore:
1355 			clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1356 			break;
1357 		case Opt_mb_history_length:
1358 			if (match_int(&args[0], &option))
1359 				return 0;
1360 			if (option < 0)
1361 				return 0;
1362 			sbi->s_mb_history_max = option;
1363 			break;
1364 #ifdef CONFIG_QUOTA
1365 		case Opt_usrjquota:
1366 			qtype = USRQUOTA;
1367 			goto set_qf_name;
1368 		case Opt_grpjquota:
1369 			qtype = GRPQUOTA;
1370 set_qf_name:
1371 			if (sb_any_quota_loaded(sb) &&
1372 			    !sbi->s_qf_names[qtype]) {
1373 				ext4_msg(sb, KERN_ERR,
1374 				       "Cannot change journaled "
1375 				       "quota options when quota turned on");
1376 				return 0;
1377 			}
1378 			qname = match_strdup(&args[0]);
1379 			if (!qname) {
1380 				ext4_msg(sb, KERN_ERR,
1381 					"Not enough memory for "
1382 					"storing quotafile name");
1383 				return 0;
1384 			}
1385 			if (sbi->s_qf_names[qtype] &&
1386 			    strcmp(sbi->s_qf_names[qtype], qname)) {
1387 				ext4_msg(sb, KERN_ERR,
1388 					"%s quota file already "
1389 					"specified", QTYPE2NAME(qtype));
1390 				kfree(qname);
1391 				return 0;
1392 			}
1393 			sbi->s_qf_names[qtype] = qname;
1394 			if (strchr(sbi->s_qf_names[qtype], '/')) {
1395 				ext4_msg(sb, KERN_ERR,
1396 					"quotafile must be on "
1397 					"filesystem root");
1398 				kfree(sbi->s_qf_names[qtype]);
1399 				sbi->s_qf_names[qtype] = NULL;
1400 				return 0;
1401 			}
1402 			set_opt(sbi->s_mount_opt, QUOTA);
1403 			break;
1404 		case Opt_offusrjquota:
1405 			qtype = USRQUOTA;
1406 			goto clear_qf_name;
1407 		case Opt_offgrpjquota:
1408 			qtype = GRPQUOTA;
1409 clear_qf_name:
1410 			if (sb_any_quota_loaded(sb) &&
1411 			    sbi->s_qf_names[qtype]) {
1412 				ext4_msg(sb, KERN_ERR, "Cannot change "
1413 					"journaled quota options when "
1414 					"quota turned on");
1415 				return 0;
1416 			}
1417 			/*
1418 			 * The space will be released later when all options
1419 			 * are confirmed to be correct
1420 			 */
1421 			sbi->s_qf_names[qtype] = NULL;
1422 			break;
1423 		case Opt_jqfmt_vfsold:
1424 			qfmt = QFMT_VFS_OLD;
1425 			goto set_qf_format;
1426 		case Opt_jqfmt_vfsv0:
1427 			qfmt = QFMT_VFS_V0;
1428 set_qf_format:
1429 			if (sb_any_quota_loaded(sb) &&
1430 			    sbi->s_jquota_fmt != qfmt) {
1431 				ext4_msg(sb, KERN_ERR, "Cannot change "
1432 					"journaled quota options when "
1433 					"quota turned on");
1434 				return 0;
1435 			}
1436 			sbi->s_jquota_fmt = qfmt;
1437 			break;
1438 		case Opt_quota:
1439 		case Opt_usrquota:
1440 			set_opt(sbi->s_mount_opt, QUOTA);
1441 			set_opt(sbi->s_mount_opt, USRQUOTA);
1442 			break;
1443 		case Opt_grpquota:
1444 			set_opt(sbi->s_mount_opt, QUOTA);
1445 			set_opt(sbi->s_mount_opt, GRPQUOTA);
1446 			break;
1447 		case Opt_noquota:
1448 			if (sb_any_quota_loaded(sb)) {
1449 				ext4_msg(sb, KERN_ERR, "Cannot change quota "
1450 					"options when quota turned on");
1451 				return 0;
1452 			}
1453 			clear_opt(sbi->s_mount_opt, QUOTA);
1454 			clear_opt(sbi->s_mount_opt, USRQUOTA);
1455 			clear_opt(sbi->s_mount_opt, GRPQUOTA);
1456 			break;
1457 #else
1458 		case Opt_quota:
1459 		case Opt_usrquota:
1460 		case Opt_grpquota:
1461 			ext4_msg(sb, KERN_ERR,
1462 				"quota options not supported");
1463 			break;
1464 		case Opt_usrjquota:
1465 		case Opt_grpjquota:
1466 		case Opt_offusrjquota:
1467 		case Opt_offgrpjquota:
1468 		case Opt_jqfmt_vfsold:
1469 		case Opt_jqfmt_vfsv0:
1470 			ext4_msg(sb, KERN_ERR,
1471 				"journaled quota options not supported");
1472 			break;
1473 		case Opt_noquota:
1474 			break;
1475 #endif
1476 		case Opt_abort:
1477 			set_opt(sbi->s_mount_opt, ABORT);
1478 			break;
1479 		case Opt_nobarrier:
1480 			clear_opt(sbi->s_mount_opt, BARRIER);
1481 			break;
1482 		case Opt_barrier:
1483 			if (match_int(&args[0], &option)) {
1484 				set_opt(sbi->s_mount_opt, BARRIER);
1485 				break;
1486 			}
1487 			if (option)
1488 				set_opt(sbi->s_mount_opt, BARRIER);
1489 			else
1490 				clear_opt(sbi->s_mount_opt, BARRIER);
1491 			break;
1492 		case Opt_ignore:
1493 			break;
1494 		case Opt_resize:
1495 			if (!is_remount) {
1496 				ext4_msg(sb, KERN_ERR,
1497 					"resize option only available "
1498 					"for remount");
1499 				return 0;
1500 			}
1501 			if (match_int(&args[0], &option) != 0)
1502 				return 0;
1503 			*n_blocks_count = option;
1504 			break;
1505 		case Opt_nobh:
1506 			set_opt(sbi->s_mount_opt, NOBH);
1507 			break;
1508 		case Opt_bh:
1509 			clear_opt(sbi->s_mount_opt, NOBH);
1510 			break;
1511 		case Opt_i_version:
1512 			set_opt(sbi->s_mount_opt, I_VERSION);
1513 			sb->s_flags |= MS_I_VERSION;
1514 			break;
1515 		case Opt_nodelalloc:
1516 			clear_opt(sbi->s_mount_opt, DELALLOC);
1517 			break;
1518 		case Opt_stripe:
1519 			if (match_int(&args[0], &option))
1520 				return 0;
1521 			if (option < 0)
1522 				return 0;
1523 			sbi->s_stripe = option;
1524 			break;
1525 		case Opt_delalloc:
1526 			set_opt(sbi->s_mount_opt, DELALLOC);
1527 			break;
1528 		case Opt_block_validity:
1529 			set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1530 			break;
1531 		case Opt_noblock_validity:
1532 			clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1533 			break;
1534 		case Opt_inode_readahead_blks:
1535 			if (match_int(&args[0], &option))
1536 				return 0;
1537 			if (option < 0 || option > (1 << 30))
1538 				return 0;
1539 			if (!is_power_of_2(option)) {
1540 				ext4_msg(sb, KERN_ERR,
1541 					 "EXT4-fs: inode_readahead_blks"
1542 					 " must be a power of 2");
1543 				return 0;
1544 			}
1545 			sbi->s_inode_readahead_blks = option;
1546 			break;
1547 		case Opt_journal_ioprio:
1548 			if (match_int(&args[0], &option))
1549 				return 0;
1550 			if (option < 0 || option > 7)
1551 				break;
1552 			*journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1553 							    option);
1554 			break;
1555 		case Opt_noauto_da_alloc:
1556 			set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1557 			break;
1558 		case Opt_auto_da_alloc:
1559 			if (match_int(&args[0], &option)) {
1560 				clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1561 				break;
1562 			}
1563 			if (option)
1564 				clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1565 			else
1566 				set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1567 			break;
1568 		default:
1569 			ext4_msg(sb, KERN_ERR,
1570 			       "Unrecognized mount option \"%s\" "
1571 			       "or missing value", p);
1572 			return 0;
1573 		}
1574 	}
1575 #ifdef CONFIG_QUOTA
1576 	if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1577 		if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1578 		     sbi->s_qf_names[USRQUOTA])
1579 			clear_opt(sbi->s_mount_opt, USRQUOTA);
1580 
1581 		if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1582 		     sbi->s_qf_names[GRPQUOTA])
1583 			clear_opt(sbi->s_mount_opt, GRPQUOTA);
1584 
1585 		if ((sbi->s_qf_names[USRQUOTA] &&
1586 				(sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1587 		    (sbi->s_qf_names[GRPQUOTA] &&
1588 				(sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1589 			ext4_msg(sb, KERN_ERR, "old and new quota "
1590 					"format mixing");
1591 			return 0;
1592 		}
1593 
1594 		if (!sbi->s_jquota_fmt) {
1595 			ext4_msg(sb, KERN_ERR, "journaled quota format "
1596 					"not specified");
1597 			return 0;
1598 		}
1599 	} else {
1600 		if (sbi->s_jquota_fmt) {
1601 			ext4_msg(sb, KERN_ERR, "journaled quota format "
1602 					"specified with no journaling "
1603 					"enabled");
1604 			return 0;
1605 		}
1606 	}
1607 #endif
1608 	return 1;
1609 }
1610 
1611 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1612 			    int read_only)
1613 {
1614 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1615 	int res = 0;
1616 
1617 	if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1618 		ext4_msg(sb, KERN_ERR, "revision level too high, "
1619 			 "forcing read-only mode");
1620 		res = MS_RDONLY;
1621 	}
1622 	if (read_only)
1623 		return res;
1624 	if (!(sbi->s_mount_state & EXT4_VALID_FS))
1625 		ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1626 			 "running e2fsck is recommended");
1627 	else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1628 		ext4_msg(sb, KERN_WARNING,
1629 			 "warning: mounting fs with errors, "
1630 			 "running e2fsck is recommended");
1631 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1632 		 le16_to_cpu(es->s_mnt_count) >=
1633 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1634 		ext4_msg(sb, KERN_WARNING,
1635 			 "warning: maximal mount count reached, "
1636 			 "running e2fsck is recommended");
1637 	else if (le32_to_cpu(es->s_checkinterval) &&
1638 		(le32_to_cpu(es->s_lastcheck) +
1639 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1640 		ext4_msg(sb, KERN_WARNING,
1641 			 "warning: checktime reached, "
1642 			 "running e2fsck is recommended");
1643 	if (!sbi->s_journal)
1644 		es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1645 	if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1646 		es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1647 	le16_add_cpu(&es->s_mnt_count, 1);
1648 	es->s_mtime = cpu_to_le32(get_seconds());
1649 	ext4_update_dynamic_rev(sb);
1650 	if (sbi->s_journal)
1651 		EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1652 
1653 	ext4_commit_super(sb, 1);
1654 	if (test_opt(sb, DEBUG))
1655 		printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1656 				"bpg=%lu, ipg=%lu, mo=%04lx]\n",
1657 			sb->s_blocksize,
1658 			sbi->s_groups_count,
1659 			EXT4_BLOCKS_PER_GROUP(sb),
1660 			EXT4_INODES_PER_GROUP(sb),
1661 			sbi->s_mount_opt);
1662 
1663 	if (EXT4_SB(sb)->s_journal) {
1664 		ext4_msg(sb, KERN_INFO, "%s journal on %s",
1665 		       EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1666 		       "external", EXT4_SB(sb)->s_journal->j_devname);
1667 	} else {
1668 		ext4_msg(sb, KERN_INFO, "no journal");
1669 	}
1670 	return res;
1671 }
1672 
1673 static int ext4_fill_flex_info(struct super_block *sb)
1674 {
1675 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1676 	struct ext4_group_desc *gdp = NULL;
1677 	ext4_group_t flex_group_count;
1678 	ext4_group_t flex_group;
1679 	int groups_per_flex = 0;
1680 	size_t size;
1681 	int i;
1682 
1683 	if (!sbi->s_es->s_log_groups_per_flex) {
1684 		sbi->s_log_groups_per_flex = 0;
1685 		return 1;
1686 	}
1687 
1688 	sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1689 	groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1690 
1691 	/* We allocate both existing and potentially added groups */
1692 	flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1693 			((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1694 			      EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1695 	size = flex_group_count * sizeof(struct flex_groups);
1696 	sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1697 	if (sbi->s_flex_groups == NULL) {
1698 		sbi->s_flex_groups = vmalloc(size);
1699 		if (sbi->s_flex_groups)
1700 			memset(sbi->s_flex_groups, 0, size);
1701 	}
1702 	if (sbi->s_flex_groups == NULL) {
1703 		ext4_msg(sb, KERN_ERR, "not enough memory for "
1704 				"%u flex groups", flex_group_count);
1705 		goto failed;
1706 	}
1707 
1708 	for (i = 0; i < sbi->s_groups_count; i++) {
1709 		gdp = ext4_get_group_desc(sb, i, NULL);
1710 
1711 		flex_group = ext4_flex_group(sbi, i);
1712 		atomic_set(&sbi->s_flex_groups[flex_group].free_inodes,
1713 			   ext4_free_inodes_count(sb, gdp));
1714 		atomic_set(&sbi->s_flex_groups[flex_group].free_blocks,
1715 			   ext4_free_blks_count(sb, gdp));
1716 		atomic_set(&sbi->s_flex_groups[flex_group].used_dirs,
1717 			   ext4_used_dirs_count(sb, gdp));
1718 	}
1719 
1720 	return 1;
1721 failed:
1722 	return 0;
1723 }
1724 
1725 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1726 			    struct ext4_group_desc *gdp)
1727 {
1728 	__u16 crc = 0;
1729 
1730 	if (sbi->s_es->s_feature_ro_compat &
1731 	    cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1732 		int offset = offsetof(struct ext4_group_desc, bg_checksum);
1733 		__le32 le_group = cpu_to_le32(block_group);
1734 
1735 		crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1736 		crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1737 		crc = crc16(crc, (__u8 *)gdp, offset);
1738 		offset += sizeof(gdp->bg_checksum); /* skip checksum */
1739 		/* for checksum of struct ext4_group_desc do the rest...*/
1740 		if ((sbi->s_es->s_feature_incompat &
1741 		     cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1742 		    offset < le16_to_cpu(sbi->s_es->s_desc_size))
1743 			crc = crc16(crc, (__u8 *)gdp + offset,
1744 				    le16_to_cpu(sbi->s_es->s_desc_size) -
1745 					offset);
1746 	}
1747 
1748 	return cpu_to_le16(crc);
1749 }
1750 
1751 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1752 				struct ext4_group_desc *gdp)
1753 {
1754 	if ((sbi->s_es->s_feature_ro_compat &
1755 	     cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1756 	    (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1757 		return 0;
1758 
1759 	return 1;
1760 }
1761 
1762 /* Called at mount-time, super-block is locked */
1763 static int ext4_check_descriptors(struct super_block *sb)
1764 {
1765 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1766 	ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1767 	ext4_fsblk_t last_block;
1768 	ext4_fsblk_t block_bitmap;
1769 	ext4_fsblk_t inode_bitmap;
1770 	ext4_fsblk_t inode_table;
1771 	int flexbg_flag = 0;
1772 	ext4_group_t i;
1773 
1774 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1775 		flexbg_flag = 1;
1776 
1777 	ext4_debug("Checking group descriptors");
1778 
1779 	for (i = 0; i < sbi->s_groups_count; i++) {
1780 		struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1781 
1782 		if (i == sbi->s_groups_count - 1 || flexbg_flag)
1783 			last_block = ext4_blocks_count(sbi->s_es) - 1;
1784 		else
1785 			last_block = first_block +
1786 				(EXT4_BLOCKS_PER_GROUP(sb) - 1);
1787 
1788 		block_bitmap = ext4_block_bitmap(sb, gdp);
1789 		if (block_bitmap < first_block || block_bitmap > last_block) {
1790 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1791 			       "Block bitmap for group %u not in group "
1792 			       "(block %llu)!", i, block_bitmap);
1793 			return 0;
1794 		}
1795 		inode_bitmap = ext4_inode_bitmap(sb, gdp);
1796 		if (inode_bitmap < first_block || inode_bitmap > last_block) {
1797 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1798 			       "Inode bitmap for group %u not in group "
1799 			       "(block %llu)!", i, inode_bitmap);
1800 			return 0;
1801 		}
1802 		inode_table = ext4_inode_table(sb, gdp);
1803 		if (inode_table < first_block ||
1804 		    inode_table + sbi->s_itb_per_group - 1 > last_block) {
1805 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1806 			       "Inode table for group %u not in group "
1807 			       "(block %llu)!", i, inode_table);
1808 			return 0;
1809 		}
1810 		ext4_lock_group(sb, i);
1811 		if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1812 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1813 				 "Checksum for group %u failed (%u!=%u)",
1814 				 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1815 				     gdp)), le16_to_cpu(gdp->bg_checksum));
1816 			if (!(sb->s_flags & MS_RDONLY)) {
1817 				ext4_unlock_group(sb, i);
1818 				return 0;
1819 			}
1820 		}
1821 		ext4_unlock_group(sb, i);
1822 		if (!flexbg_flag)
1823 			first_block += EXT4_BLOCKS_PER_GROUP(sb);
1824 	}
1825 
1826 	ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1827 	sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
1828 	return 1;
1829 }
1830 
1831 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1832  * the superblock) which were deleted from all directories, but held open by
1833  * a process at the time of a crash.  We walk the list and try to delete these
1834  * inodes at recovery time (only with a read-write filesystem).
1835  *
1836  * In order to keep the orphan inode chain consistent during traversal (in
1837  * case of crash during recovery), we link each inode into the superblock
1838  * orphan list_head and handle it the same way as an inode deletion during
1839  * normal operation (which journals the operations for us).
1840  *
1841  * We only do an iget() and an iput() on each inode, which is very safe if we
1842  * accidentally point at an in-use or already deleted inode.  The worst that
1843  * can happen in this case is that we get a "bit already cleared" message from
1844  * ext4_free_inode().  The only reason we would point at a wrong inode is if
1845  * e2fsck was run on this filesystem, and it must have already done the orphan
1846  * inode cleanup for us, so we can safely abort without any further action.
1847  */
1848 static void ext4_orphan_cleanup(struct super_block *sb,
1849 				struct ext4_super_block *es)
1850 {
1851 	unsigned int s_flags = sb->s_flags;
1852 	int nr_orphans = 0, nr_truncates = 0;
1853 #ifdef CONFIG_QUOTA
1854 	int i;
1855 #endif
1856 	if (!es->s_last_orphan) {
1857 		jbd_debug(4, "no orphan inodes to clean up\n");
1858 		return;
1859 	}
1860 
1861 	if (bdev_read_only(sb->s_bdev)) {
1862 		ext4_msg(sb, KERN_ERR, "write access "
1863 			"unavailable, skipping orphan cleanup");
1864 		return;
1865 	}
1866 
1867 	if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1868 		if (es->s_last_orphan)
1869 			jbd_debug(1, "Errors on filesystem, "
1870 				  "clearing orphan list.\n");
1871 		es->s_last_orphan = 0;
1872 		jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1873 		return;
1874 	}
1875 
1876 	if (s_flags & MS_RDONLY) {
1877 		ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1878 		sb->s_flags &= ~MS_RDONLY;
1879 	}
1880 #ifdef CONFIG_QUOTA
1881 	/* Needed for iput() to work correctly and not trash data */
1882 	sb->s_flags |= MS_ACTIVE;
1883 	/* Turn on quotas so that they are updated correctly */
1884 	for (i = 0; i < MAXQUOTAS; i++) {
1885 		if (EXT4_SB(sb)->s_qf_names[i]) {
1886 			int ret = ext4_quota_on_mount(sb, i);
1887 			if (ret < 0)
1888 				ext4_msg(sb, KERN_ERR,
1889 					"Cannot turn on journaled "
1890 					"quota: error %d", ret);
1891 		}
1892 	}
1893 #endif
1894 
1895 	while (es->s_last_orphan) {
1896 		struct inode *inode;
1897 
1898 		inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1899 		if (IS_ERR(inode)) {
1900 			es->s_last_orphan = 0;
1901 			break;
1902 		}
1903 
1904 		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1905 		vfs_dq_init(inode);
1906 		if (inode->i_nlink) {
1907 			ext4_msg(sb, KERN_DEBUG,
1908 				"%s: truncating inode %lu to %lld bytes",
1909 				__func__, inode->i_ino, inode->i_size);
1910 			jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1911 				  inode->i_ino, inode->i_size);
1912 			ext4_truncate(inode);
1913 			nr_truncates++;
1914 		} else {
1915 			ext4_msg(sb, KERN_DEBUG,
1916 				"%s: deleting unreferenced inode %lu",
1917 				__func__, inode->i_ino);
1918 			jbd_debug(2, "deleting unreferenced inode %lu\n",
1919 				  inode->i_ino);
1920 			nr_orphans++;
1921 		}
1922 		iput(inode);  /* The delete magic happens here! */
1923 	}
1924 
1925 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1926 
1927 	if (nr_orphans)
1928 		ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1929 		       PLURAL(nr_orphans));
1930 	if (nr_truncates)
1931 		ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1932 		       PLURAL(nr_truncates));
1933 #ifdef CONFIG_QUOTA
1934 	/* Turn quotas off */
1935 	for (i = 0; i < MAXQUOTAS; i++) {
1936 		if (sb_dqopt(sb)->files[i])
1937 			vfs_quota_off(sb, i, 0);
1938 	}
1939 #endif
1940 	sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1941 }
1942 
1943 /*
1944  * Maximal extent format file size.
1945  * Resulting logical blkno at s_maxbytes must fit in our on-disk
1946  * extent format containers, within a sector_t, and within i_blocks
1947  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
1948  * so that won't be a limiting factor.
1949  *
1950  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1951  */
1952 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1953 {
1954 	loff_t res;
1955 	loff_t upper_limit = MAX_LFS_FILESIZE;
1956 
1957 	/* small i_blocks in vfs inode? */
1958 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1959 		/*
1960 		 * CONFIG_LBD is not enabled implies the inode
1961 		 * i_block represent total blocks in 512 bytes
1962 		 * 32 == size of vfs inode i_blocks * 8
1963 		 */
1964 		upper_limit = (1LL << 32) - 1;
1965 
1966 		/* total blocks in file system block size */
1967 		upper_limit >>= (blkbits - 9);
1968 		upper_limit <<= blkbits;
1969 	}
1970 
1971 	/* 32-bit extent-start container, ee_block */
1972 	res = 1LL << 32;
1973 	res <<= blkbits;
1974 	res -= 1;
1975 
1976 	/* Sanity check against vm- & vfs- imposed limits */
1977 	if (res > upper_limit)
1978 		res = upper_limit;
1979 
1980 	return res;
1981 }
1982 
1983 /*
1984  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
1985  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1986  * We need to be 1 filesystem block less than the 2^48 sector limit.
1987  */
1988 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1989 {
1990 	loff_t res = EXT4_NDIR_BLOCKS;
1991 	int meta_blocks;
1992 	loff_t upper_limit;
1993 	/* This is calculated to be the largest file size for a dense, block
1994 	 * mapped file such that the file's total number of 512-byte sectors,
1995 	 * including data and all indirect blocks, does not exceed (2^48 - 1).
1996 	 *
1997 	 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
1998 	 * number of 512-byte sectors of the file.
1999 	 */
2000 
2001 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2002 		/*
2003 		 * !has_huge_files or CONFIG_LBD not enabled implies that
2004 		 * the inode i_block field represents total file blocks in
2005 		 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2006 		 */
2007 		upper_limit = (1LL << 32) - 1;
2008 
2009 		/* total blocks in file system block size */
2010 		upper_limit >>= (bits - 9);
2011 
2012 	} else {
2013 		/*
2014 		 * We use 48 bit ext4_inode i_blocks
2015 		 * With EXT4_HUGE_FILE_FL set the i_blocks
2016 		 * represent total number of blocks in
2017 		 * file system block size
2018 		 */
2019 		upper_limit = (1LL << 48) - 1;
2020 
2021 	}
2022 
2023 	/* indirect blocks */
2024 	meta_blocks = 1;
2025 	/* double indirect blocks */
2026 	meta_blocks += 1 + (1LL << (bits-2));
2027 	/* tripple indirect blocks */
2028 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2029 
2030 	upper_limit -= meta_blocks;
2031 	upper_limit <<= bits;
2032 
2033 	res += 1LL << (bits-2);
2034 	res += 1LL << (2*(bits-2));
2035 	res += 1LL << (3*(bits-2));
2036 	res <<= bits;
2037 	if (res > upper_limit)
2038 		res = upper_limit;
2039 
2040 	if (res > MAX_LFS_FILESIZE)
2041 		res = MAX_LFS_FILESIZE;
2042 
2043 	return res;
2044 }
2045 
2046 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2047 				   ext4_fsblk_t logical_sb_block, int nr)
2048 {
2049 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2050 	ext4_group_t bg, first_meta_bg;
2051 	int has_super = 0;
2052 
2053 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2054 
2055 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2056 	    nr < first_meta_bg)
2057 		return logical_sb_block + nr + 1;
2058 	bg = sbi->s_desc_per_block * nr;
2059 	if (ext4_bg_has_super(sb, bg))
2060 		has_super = 1;
2061 
2062 	return (has_super + ext4_group_first_block_no(sb, bg));
2063 }
2064 
2065 /**
2066  * ext4_get_stripe_size: Get the stripe size.
2067  * @sbi: In memory super block info
2068  *
2069  * If we have specified it via mount option, then
2070  * use the mount option value. If the value specified at mount time is
2071  * greater than the blocks per group use the super block value.
2072  * If the super block value is greater than blocks per group return 0.
2073  * Allocator needs it be less than blocks per group.
2074  *
2075  */
2076 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2077 {
2078 	unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2079 	unsigned long stripe_width =
2080 			le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2081 
2082 	if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2083 		return sbi->s_stripe;
2084 
2085 	if (stripe_width <= sbi->s_blocks_per_group)
2086 		return stripe_width;
2087 
2088 	if (stride <= sbi->s_blocks_per_group)
2089 		return stride;
2090 
2091 	return 0;
2092 }
2093 
2094 /* sysfs supprt */
2095 
2096 struct ext4_attr {
2097 	struct attribute attr;
2098 	ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2099 	ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2100 			 const char *, size_t);
2101 	int offset;
2102 };
2103 
2104 static int parse_strtoul(const char *buf,
2105 		unsigned long max, unsigned long *value)
2106 {
2107 	char *endp;
2108 
2109 	while (*buf && isspace(*buf))
2110 		buf++;
2111 	*value = simple_strtoul(buf, &endp, 0);
2112 	while (*endp && isspace(*endp))
2113 		endp++;
2114 	if (*endp || *value > max)
2115 		return -EINVAL;
2116 
2117 	return 0;
2118 }
2119 
2120 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2121 					      struct ext4_sb_info *sbi,
2122 					      char *buf)
2123 {
2124 	return snprintf(buf, PAGE_SIZE, "%llu\n",
2125 			(s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2126 }
2127 
2128 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2129 					 struct ext4_sb_info *sbi, char *buf)
2130 {
2131 	struct super_block *sb = sbi->s_buddy_cache->i_sb;
2132 
2133 	return snprintf(buf, PAGE_SIZE, "%lu\n",
2134 			(part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2135 			 sbi->s_sectors_written_start) >> 1);
2136 }
2137 
2138 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2139 					  struct ext4_sb_info *sbi, char *buf)
2140 {
2141 	struct super_block *sb = sbi->s_buddy_cache->i_sb;
2142 
2143 	return snprintf(buf, PAGE_SIZE, "%llu\n",
2144 			sbi->s_kbytes_written +
2145 			((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2146 			  EXT4_SB(sb)->s_sectors_written_start) >> 1));
2147 }
2148 
2149 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2150 					  struct ext4_sb_info *sbi,
2151 					  const char *buf, size_t count)
2152 {
2153 	unsigned long t;
2154 
2155 	if (parse_strtoul(buf, 0x40000000, &t))
2156 		return -EINVAL;
2157 
2158 	if (!is_power_of_2(t))
2159 		return -EINVAL;
2160 
2161 	sbi->s_inode_readahead_blks = t;
2162 	return count;
2163 }
2164 
2165 static ssize_t sbi_ui_show(struct ext4_attr *a,
2166 			   struct ext4_sb_info *sbi, char *buf)
2167 {
2168 	unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2169 
2170 	return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2171 }
2172 
2173 static ssize_t sbi_ui_store(struct ext4_attr *a,
2174 			    struct ext4_sb_info *sbi,
2175 			    const char *buf, size_t count)
2176 {
2177 	unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2178 	unsigned long t;
2179 
2180 	if (parse_strtoul(buf, 0xffffffff, &t))
2181 		return -EINVAL;
2182 	*ui = t;
2183 	return count;
2184 }
2185 
2186 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2187 static struct ext4_attr ext4_attr_##_name = {			\
2188 	.attr = {.name = __stringify(_name), .mode = _mode },	\
2189 	.show	= _show,					\
2190 	.store	= _store,					\
2191 	.offset = offsetof(struct ext4_sb_info, _elname),	\
2192 }
2193 #define EXT4_ATTR(name, mode, show, store) \
2194 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2195 
2196 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2197 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2198 #define EXT4_RW_ATTR_SBI_UI(name, elname)	\
2199 	EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2200 #define ATTR_LIST(name) &ext4_attr_##name.attr
2201 
2202 EXT4_RO_ATTR(delayed_allocation_blocks);
2203 EXT4_RO_ATTR(session_write_kbytes);
2204 EXT4_RO_ATTR(lifetime_write_kbytes);
2205 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2206 		 inode_readahead_blks_store, s_inode_readahead_blks);
2207 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2208 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2209 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2210 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2211 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2212 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2213 
2214 static struct attribute *ext4_attrs[] = {
2215 	ATTR_LIST(delayed_allocation_blocks),
2216 	ATTR_LIST(session_write_kbytes),
2217 	ATTR_LIST(lifetime_write_kbytes),
2218 	ATTR_LIST(inode_readahead_blks),
2219 	ATTR_LIST(mb_stats),
2220 	ATTR_LIST(mb_max_to_scan),
2221 	ATTR_LIST(mb_min_to_scan),
2222 	ATTR_LIST(mb_order2_req),
2223 	ATTR_LIST(mb_stream_req),
2224 	ATTR_LIST(mb_group_prealloc),
2225 	NULL,
2226 };
2227 
2228 static ssize_t ext4_attr_show(struct kobject *kobj,
2229 			      struct attribute *attr, char *buf)
2230 {
2231 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2232 						s_kobj);
2233 	struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2234 
2235 	return a->show ? a->show(a, sbi, buf) : 0;
2236 }
2237 
2238 static ssize_t ext4_attr_store(struct kobject *kobj,
2239 			       struct attribute *attr,
2240 			       const char *buf, size_t len)
2241 {
2242 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2243 						s_kobj);
2244 	struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2245 
2246 	return a->store ? a->store(a, sbi, buf, len) : 0;
2247 }
2248 
2249 static void ext4_sb_release(struct kobject *kobj)
2250 {
2251 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2252 						s_kobj);
2253 	complete(&sbi->s_kobj_unregister);
2254 }
2255 
2256 
2257 static struct sysfs_ops ext4_attr_ops = {
2258 	.show	= ext4_attr_show,
2259 	.store	= ext4_attr_store,
2260 };
2261 
2262 static struct kobj_type ext4_ktype = {
2263 	.default_attrs	= ext4_attrs,
2264 	.sysfs_ops	= &ext4_attr_ops,
2265 	.release	= ext4_sb_release,
2266 };
2267 
2268 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2269 				__releases(kernel_lock)
2270 				__acquires(kernel_lock)
2271 {
2272 	struct buffer_head *bh;
2273 	struct ext4_super_block *es = NULL;
2274 	struct ext4_sb_info *sbi;
2275 	ext4_fsblk_t block;
2276 	ext4_fsblk_t sb_block = get_sb_block(&data);
2277 	ext4_fsblk_t logical_sb_block;
2278 	unsigned long offset = 0;
2279 	unsigned long journal_devnum = 0;
2280 	unsigned long def_mount_opts;
2281 	struct inode *root;
2282 	char *cp;
2283 	const char *descr;
2284 	int ret = -EINVAL;
2285 	int blocksize;
2286 	unsigned int db_count;
2287 	unsigned int i;
2288 	int needs_recovery, has_huge_files;
2289 	int features;
2290 	__u64 blocks_count;
2291 	int err;
2292 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2293 
2294 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2295 	if (!sbi)
2296 		return -ENOMEM;
2297 
2298 	sbi->s_blockgroup_lock =
2299 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2300 	if (!sbi->s_blockgroup_lock) {
2301 		kfree(sbi);
2302 		return -ENOMEM;
2303 	}
2304 	sb->s_fs_info = sbi;
2305 	sbi->s_mount_opt = 0;
2306 	sbi->s_resuid = EXT4_DEF_RESUID;
2307 	sbi->s_resgid = EXT4_DEF_RESGID;
2308 	sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2309 	sbi->s_sb_block = sb_block;
2310 	sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2311 						      sectors[1]);
2312 
2313 	unlock_kernel();
2314 
2315 	/* Cleanup superblock name */
2316 	for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2317 		*cp = '!';
2318 
2319 	blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2320 	if (!blocksize) {
2321 		ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2322 		goto out_fail;
2323 	}
2324 
2325 	/*
2326 	 * The ext4 superblock will not be buffer aligned for other than 1kB
2327 	 * block sizes.  We need to calculate the offset from buffer start.
2328 	 */
2329 	if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2330 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2331 		offset = do_div(logical_sb_block, blocksize);
2332 	} else {
2333 		logical_sb_block = sb_block;
2334 	}
2335 
2336 	if (!(bh = sb_bread(sb, logical_sb_block))) {
2337 		ext4_msg(sb, KERN_ERR, "unable to read superblock");
2338 		goto out_fail;
2339 	}
2340 	/*
2341 	 * Note: s_es must be initialized as soon as possible because
2342 	 *       some ext4 macro-instructions depend on its value
2343 	 */
2344 	es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2345 	sbi->s_es = es;
2346 	sb->s_magic = le16_to_cpu(es->s_magic);
2347 	if (sb->s_magic != EXT4_SUPER_MAGIC)
2348 		goto cantfind_ext4;
2349 	sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2350 
2351 	/* Set defaults before we parse the mount options */
2352 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2353 	if (def_mount_opts & EXT4_DEFM_DEBUG)
2354 		set_opt(sbi->s_mount_opt, DEBUG);
2355 	if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2356 		set_opt(sbi->s_mount_opt, GRPID);
2357 	if (def_mount_opts & EXT4_DEFM_UID16)
2358 		set_opt(sbi->s_mount_opt, NO_UID32);
2359 #ifdef CONFIG_EXT4_FS_XATTR
2360 	if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2361 		set_opt(sbi->s_mount_opt, XATTR_USER);
2362 #endif
2363 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2364 	if (def_mount_opts & EXT4_DEFM_ACL)
2365 		set_opt(sbi->s_mount_opt, POSIX_ACL);
2366 #endif
2367 	if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2368 		sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2369 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2370 		sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2371 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2372 		sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2373 
2374 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2375 		set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2376 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2377 		set_opt(sbi->s_mount_opt, ERRORS_CONT);
2378 	else
2379 		set_opt(sbi->s_mount_opt, ERRORS_RO);
2380 
2381 	sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2382 	sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2383 	sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2384 	sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2385 	sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2386 	sbi->s_mb_history_max = default_mb_history_length;
2387 
2388 	set_opt(sbi->s_mount_opt, BARRIER);
2389 
2390 	/*
2391 	 * enable delayed allocation by default
2392 	 * Use -o nodelalloc to turn it off
2393 	 */
2394 	set_opt(sbi->s_mount_opt, DELALLOC);
2395 
2396 	if (!parse_options((char *) data, sb, &journal_devnum,
2397 			   &journal_ioprio, NULL, 0))
2398 		goto failed_mount;
2399 
2400 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2401 		((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2402 
2403 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2404 	    (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2405 	     EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2406 	     EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2407 		ext4_msg(sb, KERN_WARNING,
2408 		       "feature flags set on rev 0 fs, "
2409 		       "running e2fsck is recommended");
2410 
2411 	/*
2412 	 * Check feature flags regardless of the revision level, since we
2413 	 * previously didn't change the revision level when setting the flags,
2414 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
2415 	 */
2416 	features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2417 	if (features) {
2418 		ext4_msg(sb, KERN_ERR,
2419 			"Couldn't mount because of "
2420 			"unsupported optional features (%x)",
2421 			(le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2422 			~EXT4_FEATURE_INCOMPAT_SUPP));
2423 		goto failed_mount;
2424 	}
2425 	features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2426 	if (!(sb->s_flags & MS_RDONLY) && features) {
2427 		ext4_msg(sb, KERN_ERR,
2428 			"Couldn't mount RDWR because of "
2429 			"unsupported optional features (%x)",
2430 			(le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2431 			~EXT4_FEATURE_RO_COMPAT_SUPP));
2432 		goto failed_mount;
2433 	}
2434 	has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2435 				    EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2436 	if (has_huge_files) {
2437 		/*
2438 		 * Large file size enabled file system can only be
2439 		 * mount if kernel is build with CONFIG_LBD
2440 		 */
2441 		if (sizeof(root->i_blocks) < sizeof(u64) &&
2442 				!(sb->s_flags & MS_RDONLY)) {
2443 			ext4_msg(sb, KERN_ERR, "Filesystem with huge "
2444 					"files cannot be mounted read-write "
2445 					"without CONFIG_LBD");
2446 			goto failed_mount;
2447 		}
2448 	}
2449 	blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2450 
2451 	if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2452 	    blocksize > EXT4_MAX_BLOCK_SIZE) {
2453 		ext4_msg(sb, KERN_ERR,
2454 		       "Unsupported filesystem blocksize %d", blocksize);
2455 		goto failed_mount;
2456 	}
2457 
2458 	if (sb->s_blocksize != blocksize) {
2459 		/* Validate the filesystem blocksize */
2460 		if (!sb_set_blocksize(sb, blocksize)) {
2461 			ext4_msg(sb, KERN_ERR, "bad block size %d",
2462 					blocksize);
2463 			goto failed_mount;
2464 		}
2465 
2466 		brelse(bh);
2467 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2468 		offset = do_div(logical_sb_block, blocksize);
2469 		bh = sb_bread(sb, logical_sb_block);
2470 		if (!bh) {
2471 			ext4_msg(sb, KERN_ERR,
2472 			       "Can't read superblock on 2nd try");
2473 			goto failed_mount;
2474 		}
2475 		es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2476 		sbi->s_es = es;
2477 		if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2478 			ext4_msg(sb, KERN_ERR,
2479 			       "Magic mismatch, very weird!");
2480 			goto failed_mount;
2481 		}
2482 	}
2483 
2484 	sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2485 						      has_huge_files);
2486 	sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2487 
2488 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2489 		sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2490 		sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2491 	} else {
2492 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2493 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2494 		if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2495 		    (!is_power_of_2(sbi->s_inode_size)) ||
2496 		    (sbi->s_inode_size > blocksize)) {
2497 			ext4_msg(sb, KERN_ERR,
2498 			       "unsupported inode size: %d",
2499 			       sbi->s_inode_size);
2500 			goto failed_mount;
2501 		}
2502 		if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2503 			sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2504 	}
2505 
2506 	sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2507 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2508 		if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2509 		    sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2510 		    !is_power_of_2(sbi->s_desc_size)) {
2511 			ext4_msg(sb, KERN_ERR,
2512 			       "unsupported descriptor size %lu",
2513 			       sbi->s_desc_size);
2514 			goto failed_mount;
2515 		}
2516 	} else
2517 		sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2518 
2519 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2520 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2521 	if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2522 		goto cantfind_ext4;
2523 
2524 	sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2525 	if (sbi->s_inodes_per_block == 0)
2526 		goto cantfind_ext4;
2527 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
2528 					sbi->s_inodes_per_block;
2529 	sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2530 	sbi->s_sbh = bh;
2531 	sbi->s_mount_state = le16_to_cpu(es->s_state);
2532 	sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2533 	sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2534 
2535 	for (i = 0; i < 4; i++)
2536 		sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2537 	sbi->s_def_hash_version = es->s_def_hash_version;
2538 	i = le32_to_cpu(es->s_flags);
2539 	if (i & EXT2_FLAGS_UNSIGNED_HASH)
2540 		sbi->s_hash_unsigned = 3;
2541 	else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2542 #ifdef __CHAR_UNSIGNED__
2543 		es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2544 		sbi->s_hash_unsigned = 3;
2545 #else
2546 		es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2547 #endif
2548 		sb->s_dirt = 1;
2549 	}
2550 
2551 	if (sbi->s_blocks_per_group > blocksize * 8) {
2552 		ext4_msg(sb, KERN_ERR,
2553 		       "#blocks per group too big: %lu",
2554 		       sbi->s_blocks_per_group);
2555 		goto failed_mount;
2556 	}
2557 	if (sbi->s_inodes_per_group > blocksize * 8) {
2558 		ext4_msg(sb, KERN_ERR,
2559 		       "#inodes per group too big: %lu",
2560 		       sbi->s_inodes_per_group);
2561 		goto failed_mount;
2562 	}
2563 
2564 	if (ext4_blocks_count(es) >
2565 		    (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2566 		ext4_msg(sb, KERN_ERR, "filesystem"
2567 			" too large to mount safely");
2568 		if (sizeof(sector_t) < 8)
2569 			ext4_msg(sb, KERN_WARNING, "CONFIG_LBD not enabled");
2570 		goto failed_mount;
2571 	}
2572 
2573 	if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2574 		goto cantfind_ext4;
2575 
2576 	/* check blocks count against device size */
2577 	blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2578 	if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2579 		ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2580 		       "exceeds size of device (%llu blocks)",
2581 		       ext4_blocks_count(es), blocks_count);
2582 		goto failed_mount;
2583 	}
2584 
2585 	/*
2586 	 * It makes no sense for the first data block to be beyond the end
2587 	 * of the filesystem.
2588 	 */
2589 	if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2590                 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2591 			 "block %u is beyond end of filesystem (%llu)",
2592 			 le32_to_cpu(es->s_first_data_block),
2593 			 ext4_blocks_count(es));
2594 		goto failed_mount;
2595 	}
2596 	blocks_count = (ext4_blocks_count(es) -
2597 			le32_to_cpu(es->s_first_data_block) +
2598 			EXT4_BLOCKS_PER_GROUP(sb) - 1);
2599 	do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2600 	if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2601 		ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2602 		       "(block count %llu, first data block %u, "
2603 		       "blocks per group %lu)", sbi->s_groups_count,
2604 		       ext4_blocks_count(es),
2605 		       le32_to_cpu(es->s_first_data_block),
2606 		       EXT4_BLOCKS_PER_GROUP(sb));
2607 		goto failed_mount;
2608 	}
2609 	sbi->s_groups_count = blocks_count;
2610 	db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2611 		   EXT4_DESC_PER_BLOCK(sb);
2612 	sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2613 				    GFP_KERNEL);
2614 	if (sbi->s_group_desc == NULL) {
2615 		ext4_msg(sb, KERN_ERR, "not enough memory");
2616 		goto failed_mount;
2617 	}
2618 
2619 #ifdef CONFIG_PROC_FS
2620 	if (ext4_proc_root)
2621 		sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2622 #endif
2623 
2624 	bgl_lock_init(sbi->s_blockgroup_lock);
2625 
2626 	for (i = 0; i < db_count; i++) {
2627 		block = descriptor_loc(sb, logical_sb_block, i);
2628 		sbi->s_group_desc[i] = sb_bread(sb, block);
2629 		if (!sbi->s_group_desc[i]) {
2630 			ext4_msg(sb, KERN_ERR,
2631 			       "can't read group descriptor %d", i);
2632 			db_count = i;
2633 			goto failed_mount2;
2634 		}
2635 	}
2636 	if (!ext4_check_descriptors(sb)) {
2637 		ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2638 		goto failed_mount2;
2639 	}
2640 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2641 		if (!ext4_fill_flex_info(sb)) {
2642 			ext4_msg(sb, KERN_ERR,
2643 			       "unable to initialize "
2644 			       "flex_bg meta info!");
2645 			goto failed_mount2;
2646 		}
2647 
2648 	sbi->s_gdb_count = db_count;
2649 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2650 	spin_lock_init(&sbi->s_next_gen_lock);
2651 
2652 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
2653 			ext4_count_free_blocks(sb));
2654 	if (!err) {
2655 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
2656 				ext4_count_free_inodes(sb));
2657 	}
2658 	if (!err) {
2659 		err = percpu_counter_init(&sbi->s_dirs_counter,
2660 				ext4_count_dirs(sb));
2661 	}
2662 	if (!err) {
2663 		err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2664 	}
2665 	if (err) {
2666 		ext4_msg(sb, KERN_ERR, "insufficient memory");
2667 		goto failed_mount3;
2668 	}
2669 
2670 	sbi->s_stripe = ext4_get_stripe_size(sbi);
2671 
2672 	/*
2673 	 * set up enough so that it can read an inode
2674 	 */
2675 	if (!test_opt(sb, NOLOAD) &&
2676 	    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2677 		sb->s_op = &ext4_sops;
2678 	else
2679 		sb->s_op = &ext4_nojournal_sops;
2680 	sb->s_export_op = &ext4_export_ops;
2681 	sb->s_xattr = ext4_xattr_handlers;
2682 #ifdef CONFIG_QUOTA
2683 	sb->s_qcop = &ext4_qctl_operations;
2684 	sb->dq_op = &ext4_quota_operations;
2685 #endif
2686 	INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2687 	mutex_init(&sbi->s_orphan_lock);
2688 	mutex_init(&sbi->s_resize_lock);
2689 
2690 	sb->s_root = NULL;
2691 
2692 	needs_recovery = (es->s_last_orphan != 0 ||
2693 			  EXT4_HAS_INCOMPAT_FEATURE(sb,
2694 				    EXT4_FEATURE_INCOMPAT_RECOVER));
2695 
2696 	/*
2697 	 * The first inode we look at is the journal inode.  Don't try
2698 	 * root first: it may be modified in the journal!
2699 	 */
2700 	if (!test_opt(sb, NOLOAD) &&
2701 	    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2702 		if (ext4_load_journal(sb, es, journal_devnum))
2703 			goto failed_mount3;
2704 		if (!(sb->s_flags & MS_RDONLY) &&
2705 		    EXT4_SB(sb)->s_journal->j_failed_commit) {
2706 			ext4_msg(sb, KERN_CRIT, "error: "
2707 			       "ext4_fill_super: Journal transaction "
2708 			       "%u is corrupt",
2709 			       EXT4_SB(sb)->s_journal->j_failed_commit);
2710 			if (test_opt(sb, ERRORS_RO)) {
2711 				ext4_msg(sb, KERN_CRIT,
2712 				       "Mounting filesystem read-only");
2713 				sb->s_flags |= MS_RDONLY;
2714 				EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2715 				es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2716 			}
2717 			if (test_opt(sb, ERRORS_PANIC)) {
2718 				EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2719 				es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2720 				ext4_commit_super(sb, 1);
2721 				goto failed_mount4;
2722 			}
2723 		}
2724 	} else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2725 	      EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2726 		ext4_msg(sb, KERN_ERR, "required journal recovery "
2727 		       "suppressed and not mounted read-only");
2728 		goto failed_mount4;
2729 	} else {
2730 		clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2731 		set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2732 		sbi->s_journal = NULL;
2733 		needs_recovery = 0;
2734 		goto no_journal;
2735 	}
2736 
2737 	if (ext4_blocks_count(es) > 0xffffffffULL &&
2738 	    !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2739 				       JBD2_FEATURE_INCOMPAT_64BIT)) {
2740 		ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2741 		goto failed_mount4;
2742 	}
2743 
2744 	if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2745 		jbd2_journal_set_features(sbi->s_journal,
2746 				JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2747 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2748 	} else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2749 		jbd2_journal_set_features(sbi->s_journal,
2750 				JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2751 		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2752 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2753 	} else {
2754 		jbd2_journal_clear_features(sbi->s_journal,
2755 				JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2756 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2757 	}
2758 
2759 	/* We have now updated the journal if required, so we can
2760 	 * validate the data journaling mode. */
2761 	switch (test_opt(sb, DATA_FLAGS)) {
2762 	case 0:
2763 		/* No mode set, assume a default based on the journal
2764 		 * capabilities: ORDERED_DATA if the journal can
2765 		 * cope, else JOURNAL_DATA
2766 		 */
2767 		if (jbd2_journal_check_available_features
2768 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2769 			set_opt(sbi->s_mount_opt, ORDERED_DATA);
2770 		else
2771 			set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2772 		break;
2773 
2774 	case EXT4_MOUNT_ORDERED_DATA:
2775 	case EXT4_MOUNT_WRITEBACK_DATA:
2776 		if (!jbd2_journal_check_available_features
2777 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2778 			ext4_msg(sb, KERN_ERR, "Journal does not support "
2779 			       "requested data journaling mode");
2780 			goto failed_mount4;
2781 		}
2782 	default:
2783 		break;
2784 	}
2785 	set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2786 
2787 no_journal:
2788 
2789 	if (test_opt(sb, NOBH)) {
2790 		if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2791 			ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2792 				"its supported only with writeback mode");
2793 			clear_opt(sbi->s_mount_opt, NOBH);
2794 		}
2795 	}
2796 	/*
2797 	 * The jbd2_journal_load will have done any necessary log recovery,
2798 	 * so we can safely mount the rest of the filesystem now.
2799 	 */
2800 
2801 	root = ext4_iget(sb, EXT4_ROOT_INO);
2802 	if (IS_ERR(root)) {
2803 		ext4_msg(sb, KERN_ERR, "get root inode failed");
2804 		ret = PTR_ERR(root);
2805 		goto failed_mount4;
2806 	}
2807 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2808 		iput(root);
2809 		ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
2810 		goto failed_mount4;
2811 	}
2812 	sb->s_root = d_alloc_root(root);
2813 	if (!sb->s_root) {
2814 		ext4_msg(sb, KERN_ERR, "get root dentry failed");
2815 		iput(root);
2816 		ret = -ENOMEM;
2817 		goto failed_mount4;
2818 	}
2819 
2820 	ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2821 
2822 	/* determine the minimum size of new large inodes, if present */
2823 	if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2824 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2825 						     EXT4_GOOD_OLD_INODE_SIZE;
2826 		if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2827 				       EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2828 			if (sbi->s_want_extra_isize <
2829 			    le16_to_cpu(es->s_want_extra_isize))
2830 				sbi->s_want_extra_isize =
2831 					le16_to_cpu(es->s_want_extra_isize);
2832 			if (sbi->s_want_extra_isize <
2833 			    le16_to_cpu(es->s_min_extra_isize))
2834 				sbi->s_want_extra_isize =
2835 					le16_to_cpu(es->s_min_extra_isize);
2836 		}
2837 	}
2838 	/* Check if enough inode space is available */
2839 	if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2840 							sbi->s_inode_size) {
2841 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2842 						       EXT4_GOOD_OLD_INODE_SIZE;
2843 		ext4_msg(sb, KERN_INFO, "required extra inode space not"
2844 			 "available");
2845 	}
2846 
2847 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2848 		ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2849 			 "requested data journaling mode");
2850 		clear_opt(sbi->s_mount_opt, DELALLOC);
2851 	} else if (test_opt(sb, DELALLOC))
2852 		ext4_msg(sb, KERN_INFO, "delayed allocation enabled");
2853 
2854 	err = ext4_setup_system_zone(sb);
2855 	if (err) {
2856 		ext4_msg(sb, KERN_ERR, "failed to initialize system "
2857 			 "zone (%d)\n", err);
2858 		goto failed_mount4;
2859 	}
2860 
2861 	ext4_ext_init(sb);
2862 	err = ext4_mb_init(sb, needs_recovery);
2863 	if (err) {
2864 		ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2865 			 err);
2866 		goto failed_mount4;
2867 	}
2868 
2869 	sbi->s_kobj.kset = ext4_kset;
2870 	init_completion(&sbi->s_kobj_unregister);
2871 	err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2872 				   "%s", sb->s_id);
2873 	if (err) {
2874 		ext4_mb_release(sb);
2875 		ext4_ext_release(sb);
2876 		goto failed_mount4;
2877 	};
2878 
2879 	EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2880 	ext4_orphan_cleanup(sb, es);
2881 	EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2882 	if (needs_recovery) {
2883 		ext4_msg(sb, KERN_INFO, "recovery complete");
2884 		ext4_mark_recovery_complete(sb, es);
2885 	}
2886 	if (EXT4_SB(sb)->s_journal) {
2887 		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2888 			descr = " journalled data mode";
2889 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2890 			descr = " ordered data mode";
2891 		else
2892 			descr = " writeback data mode";
2893 	} else
2894 		descr = "out journal";
2895 
2896 	ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
2897 
2898 	lock_kernel();
2899 	return 0;
2900 
2901 cantfind_ext4:
2902 	if (!silent)
2903 		ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
2904 	goto failed_mount;
2905 
2906 failed_mount4:
2907 	ext4_msg(sb, KERN_ERR, "mount failed");
2908 	ext4_release_system_zone(sb);
2909 	if (sbi->s_journal) {
2910 		jbd2_journal_destroy(sbi->s_journal);
2911 		sbi->s_journal = NULL;
2912 	}
2913 failed_mount3:
2914 	if (sbi->s_flex_groups) {
2915 		if (is_vmalloc_addr(sbi->s_flex_groups))
2916 			vfree(sbi->s_flex_groups);
2917 		else
2918 			kfree(sbi->s_flex_groups);
2919 	}
2920 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
2921 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
2922 	percpu_counter_destroy(&sbi->s_dirs_counter);
2923 	percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2924 failed_mount2:
2925 	for (i = 0; i < db_count; i++)
2926 		brelse(sbi->s_group_desc[i]);
2927 	kfree(sbi->s_group_desc);
2928 failed_mount:
2929 	if (sbi->s_proc) {
2930 		remove_proc_entry(sb->s_id, ext4_proc_root);
2931 	}
2932 #ifdef CONFIG_QUOTA
2933 	for (i = 0; i < MAXQUOTAS; i++)
2934 		kfree(sbi->s_qf_names[i]);
2935 #endif
2936 	ext4_blkdev_remove(sbi);
2937 	brelse(bh);
2938 out_fail:
2939 	sb->s_fs_info = NULL;
2940 	kfree(sbi->s_blockgroup_lock);
2941 	kfree(sbi);
2942 	lock_kernel();
2943 	return ret;
2944 }
2945 
2946 /*
2947  * Setup any per-fs journal parameters now.  We'll do this both on
2948  * initial mount, once the journal has been initialised but before we've
2949  * done any recovery; and again on any subsequent remount.
2950  */
2951 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2952 {
2953 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2954 
2955 	journal->j_commit_interval = sbi->s_commit_interval;
2956 	journal->j_min_batch_time = sbi->s_min_batch_time;
2957 	journal->j_max_batch_time = sbi->s_max_batch_time;
2958 
2959 	spin_lock(&journal->j_state_lock);
2960 	if (test_opt(sb, BARRIER))
2961 		journal->j_flags |= JBD2_BARRIER;
2962 	else
2963 		journal->j_flags &= ~JBD2_BARRIER;
2964 	if (test_opt(sb, DATA_ERR_ABORT))
2965 		journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2966 	else
2967 		journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2968 	spin_unlock(&journal->j_state_lock);
2969 }
2970 
2971 static journal_t *ext4_get_journal(struct super_block *sb,
2972 				   unsigned int journal_inum)
2973 {
2974 	struct inode *journal_inode;
2975 	journal_t *journal;
2976 
2977 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2978 
2979 	/* First, test for the existence of a valid inode on disk.  Bad
2980 	 * things happen if we iget() an unused inode, as the subsequent
2981 	 * iput() will try to delete it. */
2982 
2983 	journal_inode = ext4_iget(sb, journal_inum);
2984 	if (IS_ERR(journal_inode)) {
2985 		ext4_msg(sb, KERN_ERR, "no journal found");
2986 		return NULL;
2987 	}
2988 	if (!journal_inode->i_nlink) {
2989 		make_bad_inode(journal_inode);
2990 		iput(journal_inode);
2991 		ext4_msg(sb, KERN_ERR, "journal inode is deleted");
2992 		return NULL;
2993 	}
2994 
2995 	jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2996 		  journal_inode, journal_inode->i_size);
2997 	if (!S_ISREG(journal_inode->i_mode)) {
2998 		ext4_msg(sb, KERN_ERR, "invalid journal inode");
2999 		iput(journal_inode);
3000 		return NULL;
3001 	}
3002 
3003 	journal = jbd2_journal_init_inode(journal_inode);
3004 	if (!journal) {
3005 		ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3006 		iput(journal_inode);
3007 		return NULL;
3008 	}
3009 	journal->j_private = sb;
3010 	ext4_init_journal_params(sb, journal);
3011 	return journal;
3012 }
3013 
3014 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3015 				       dev_t j_dev)
3016 {
3017 	struct buffer_head *bh;
3018 	journal_t *journal;
3019 	ext4_fsblk_t start;
3020 	ext4_fsblk_t len;
3021 	int hblock, blocksize;
3022 	ext4_fsblk_t sb_block;
3023 	unsigned long offset;
3024 	struct ext4_super_block *es;
3025 	struct block_device *bdev;
3026 
3027 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3028 
3029 	bdev = ext4_blkdev_get(j_dev, sb);
3030 	if (bdev == NULL)
3031 		return NULL;
3032 
3033 	if (bd_claim(bdev, sb)) {
3034 		ext4_msg(sb, KERN_ERR,
3035 			"failed to claim external journal device");
3036 		blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3037 		return NULL;
3038 	}
3039 
3040 	blocksize = sb->s_blocksize;
3041 	hblock = bdev_logical_block_size(bdev);
3042 	if (blocksize < hblock) {
3043 		ext4_msg(sb, KERN_ERR,
3044 			"blocksize too small for journal device");
3045 		goto out_bdev;
3046 	}
3047 
3048 	sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3049 	offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3050 	set_blocksize(bdev, blocksize);
3051 	if (!(bh = __bread(bdev, sb_block, blocksize))) {
3052 		ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3053 		       "external journal");
3054 		goto out_bdev;
3055 	}
3056 
3057 	es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3058 	if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3059 	    !(le32_to_cpu(es->s_feature_incompat) &
3060 	      EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3061 		ext4_msg(sb, KERN_ERR, "external journal has "
3062 					"bad superblock");
3063 		brelse(bh);
3064 		goto out_bdev;
3065 	}
3066 
3067 	if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3068 		ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3069 		brelse(bh);
3070 		goto out_bdev;
3071 	}
3072 
3073 	len = ext4_blocks_count(es);
3074 	start = sb_block + 1;
3075 	brelse(bh);	/* we're done with the superblock */
3076 
3077 	journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3078 					start, len, blocksize);
3079 	if (!journal) {
3080 		ext4_msg(sb, KERN_ERR, "failed to create device journal");
3081 		goto out_bdev;
3082 	}
3083 	journal->j_private = sb;
3084 	ll_rw_block(READ, 1, &journal->j_sb_buffer);
3085 	wait_on_buffer(journal->j_sb_buffer);
3086 	if (!buffer_uptodate(journal->j_sb_buffer)) {
3087 		ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3088 		goto out_journal;
3089 	}
3090 	if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3091 		ext4_msg(sb, KERN_ERR, "External journal has more than one "
3092 					"user (unsupported) - %d",
3093 			be32_to_cpu(journal->j_superblock->s_nr_users));
3094 		goto out_journal;
3095 	}
3096 	EXT4_SB(sb)->journal_bdev = bdev;
3097 	ext4_init_journal_params(sb, journal);
3098 	return journal;
3099 
3100 out_journal:
3101 	jbd2_journal_destroy(journal);
3102 out_bdev:
3103 	ext4_blkdev_put(bdev);
3104 	return NULL;
3105 }
3106 
3107 static int ext4_load_journal(struct super_block *sb,
3108 			     struct ext4_super_block *es,
3109 			     unsigned long journal_devnum)
3110 {
3111 	journal_t *journal;
3112 	unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3113 	dev_t journal_dev;
3114 	int err = 0;
3115 	int really_read_only;
3116 
3117 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3118 
3119 	if (journal_devnum &&
3120 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3121 		ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3122 			"numbers have changed");
3123 		journal_dev = new_decode_dev(journal_devnum);
3124 	} else
3125 		journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3126 
3127 	really_read_only = bdev_read_only(sb->s_bdev);
3128 
3129 	/*
3130 	 * Are we loading a blank journal or performing recovery after a
3131 	 * crash?  For recovery, we need to check in advance whether we
3132 	 * can get read-write access to the device.
3133 	 */
3134 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3135 		if (sb->s_flags & MS_RDONLY) {
3136 			ext4_msg(sb, KERN_INFO, "INFO: recovery "
3137 					"required on readonly filesystem");
3138 			if (really_read_only) {
3139 				ext4_msg(sb, KERN_ERR, "write access "
3140 					"unavailable, cannot proceed");
3141 				return -EROFS;
3142 			}
3143 			ext4_msg(sb, KERN_INFO, "write access will "
3144 			       "be enabled during recovery");
3145 		}
3146 	}
3147 
3148 	if (journal_inum && journal_dev) {
3149 		ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3150 		       "and inode journals!");
3151 		return -EINVAL;
3152 	}
3153 
3154 	if (journal_inum) {
3155 		if (!(journal = ext4_get_journal(sb, journal_inum)))
3156 			return -EINVAL;
3157 	} else {
3158 		if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3159 			return -EINVAL;
3160 	}
3161 
3162 	if (journal->j_flags & JBD2_BARRIER)
3163 		ext4_msg(sb, KERN_INFO, "barriers enabled");
3164 	else
3165 		ext4_msg(sb, KERN_INFO, "barriers disabled");
3166 
3167 	if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3168 		err = jbd2_journal_update_format(journal);
3169 		if (err)  {
3170 			ext4_msg(sb, KERN_ERR, "error updating journal");
3171 			jbd2_journal_destroy(journal);
3172 			return err;
3173 		}
3174 	}
3175 
3176 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3177 		err = jbd2_journal_wipe(journal, !really_read_only);
3178 	if (!err)
3179 		err = jbd2_journal_load(journal);
3180 
3181 	if (err) {
3182 		ext4_msg(sb, KERN_ERR, "error loading journal");
3183 		jbd2_journal_destroy(journal);
3184 		return err;
3185 	}
3186 
3187 	EXT4_SB(sb)->s_journal = journal;
3188 	ext4_clear_journal_err(sb, es);
3189 
3190 	if (journal_devnum &&
3191 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3192 		es->s_journal_dev = cpu_to_le32(journal_devnum);
3193 
3194 		/* Make sure we flush the recovery flag to disk. */
3195 		ext4_commit_super(sb, 1);
3196 	}
3197 
3198 	return 0;
3199 }
3200 
3201 static int ext4_commit_super(struct super_block *sb, int sync)
3202 {
3203 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3204 	struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3205 	int error = 0;
3206 
3207 	if (!sbh)
3208 		return error;
3209 	if (buffer_write_io_error(sbh)) {
3210 		/*
3211 		 * Oh, dear.  A previous attempt to write the
3212 		 * superblock failed.  This could happen because the
3213 		 * USB device was yanked out.  Or it could happen to
3214 		 * be a transient write error and maybe the block will
3215 		 * be remapped.  Nothing we can do but to retry the
3216 		 * write and hope for the best.
3217 		 */
3218 		ext4_msg(sb, KERN_ERR, "previous I/O error to "
3219 		       "superblock detected");
3220 		clear_buffer_write_io_error(sbh);
3221 		set_buffer_uptodate(sbh);
3222 	}
3223 	es->s_wtime = cpu_to_le32(get_seconds());
3224 	es->s_kbytes_written =
3225 		cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3226 			    ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3227 			      EXT4_SB(sb)->s_sectors_written_start) >> 1));
3228 	ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3229 					&EXT4_SB(sb)->s_freeblocks_counter));
3230 	es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3231 					&EXT4_SB(sb)->s_freeinodes_counter));
3232 	sb->s_dirt = 0;
3233 	BUFFER_TRACE(sbh, "marking dirty");
3234 	mark_buffer_dirty(sbh);
3235 	if (sync) {
3236 		error = sync_dirty_buffer(sbh);
3237 		if (error)
3238 			return error;
3239 
3240 		error = buffer_write_io_error(sbh);
3241 		if (error) {
3242 			ext4_msg(sb, KERN_ERR, "I/O error while writing "
3243 			       "superblock");
3244 			clear_buffer_write_io_error(sbh);
3245 			set_buffer_uptodate(sbh);
3246 		}
3247 	}
3248 	return error;
3249 }
3250 
3251 /*
3252  * Have we just finished recovery?  If so, and if we are mounting (or
3253  * remounting) the filesystem readonly, then we will end up with a
3254  * consistent fs on disk.  Record that fact.
3255  */
3256 static void ext4_mark_recovery_complete(struct super_block *sb,
3257 					struct ext4_super_block *es)
3258 {
3259 	journal_t *journal = EXT4_SB(sb)->s_journal;
3260 
3261 	if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3262 		BUG_ON(journal != NULL);
3263 		return;
3264 	}
3265 	jbd2_journal_lock_updates(journal);
3266 	if (jbd2_journal_flush(journal) < 0)
3267 		goto out;
3268 
3269 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3270 	    sb->s_flags & MS_RDONLY) {
3271 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3272 		ext4_commit_super(sb, 1);
3273 	}
3274 
3275 out:
3276 	jbd2_journal_unlock_updates(journal);
3277 }
3278 
3279 /*
3280  * If we are mounting (or read-write remounting) a filesystem whose journal
3281  * has recorded an error from a previous lifetime, move that error to the
3282  * main filesystem now.
3283  */
3284 static void ext4_clear_journal_err(struct super_block *sb,
3285 				   struct ext4_super_block *es)
3286 {
3287 	journal_t *journal;
3288 	int j_errno;
3289 	const char *errstr;
3290 
3291 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3292 
3293 	journal = EXT4_SB(sb)->s_journal;
3294 
3295 	/*
3296 	 * Now check for any error status which may have been recorded in the
3297 	 * journal by a prior ext4_error() or ext4_abort()
3298 	 */
3299 
3300 	j_errno = jbd2_journal_errno(journal);
3301 	if (j_errno) {
3302 		char nbuf[16];
3303 
3304 		errstr = ext4_decode_error(sb, j_errno, nbuf);
3305 		ext4_warning(sb, __func__, "Filesystem error recorded "
3306 			     "from previous mount: %s", errstr);
3307 		ext4_warning(sb, __func__, "Marking fs in need of "
3308 			     "filesystem check.");
3309 
3310 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3311 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3312 		ext4_commit_super(sb, 1);
3313 
3314 		jbd2_journal_clear_err(journal);
3315 	}
3316 }
3317 
3318 /*
3319  * Force the running and committing transactions to commit,
3320  * and wait on the commit.
3321  */
3322 int ext4_force_commit(struct super_block *sb)
3323 {
3324 	journal_t *journal;
3325 	int ret = 0;
3326 
3327 	if (sb->s_flags & MS_RDONLY)
3328 		return 0;
3329 
3330 	journal = EXT4_SB(sb)->s_journal;
3331 	if (journal)
3332 		ret = ext4_journal_force_commit(journal);
3333 
3334 	return ret;
3335 }
3336 
3337 static void ext4_write_super(struct super_block *sb)
3338 {
3339 	lock_super(sb);
3340 	ext4_commit_super(sb, 1);
3341 	unlock_super(sb);
3342 }
3343 
3344 static int ext4_sync_fs(struct super_block *sb, int wait)
3345 {
3346 	int ret = 0;
3347 	tid_t target;
3348 
3349 	trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
3350 	if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
3351 		if (wait)
3352 			jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
3353 	}
3354 	return ret;
3355 }
3356 
3357 /*
3358  * LVM calls this function before a (read-only) snapshot is created.  This
3359  * gives us a chance to flush the journal completely and mark the fs clean.
3360  */
3361 static int ext4_freeze(struct super_block *sb)
3362 {
3363 	int error = 0;
3364 	journal_t *journal;
3365 
3366 	if (sb->s_flags & MS_RDONLY)
3367 		return 0;
3368 
3369 	journal = EXT4_SB(sb)->s_journal;
3370 
3371 	/* Now we set up the journal barrier. */
3372 	jbd2_journal_lock_updates(journal);
3373 
3374 	/*
3375 	 * Don't clear the needs_recovery flag if we failed to flush
3376 	 * the journal.
3377 	 */
3378 	error = jbd2_journal_flush(journal);
3379 	if (error < 0) {
3380 	out:
3381 		jbd2_journal_unlock_updates(journal);
3382 		return error;
3383 	}
3384 
3385 	/* Journal blocked and flushed, clear needs_recovery flag. */
3386 	EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3387 	error = ext4_commit_super(sb, 1);
3388 	if (error)
3389 		goto out;
3390 	return 0;
3391 }
3392 
3393 /*
3394  * Called by LVM after the snapshot is done.  We need to reset the RECOVER
3395  * flag here, even though the filesystem is not technically dirty yet.
3396  */
3397 static int ext4_unfreeze(struct super_block *sb)
3398 {
3399 	if (sb->s_flags & MS_RDONLY)
3400 		return 0;
3401 
3402 	lock_super(sb);
3403 	/* Reset the needs_recovery flag before the fs is unlocked. */
3404 	EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3405 	ext4_commit_super(sb, 1);
3406 	unlock_super(sb);
3407 	jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3408 	return 0;
3409 }
3410 
3411 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3412 {
3413 	struct ext4_super_block *es;
3414 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3415 	ext4_fsblk_t n_blocks_count = 0;
3416 	unsigned long old_sb_flags;
3417 	struct ext4_mount_options old_opts;
3418 	ext4_group_t g;
3419 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3420 	int err;
3421 #ifdef CONFIG_QUOTA
3422 	int i;
3423 #endif
3424 
3425 	lock_kernel();
3426 
3427 	/* Store the original options */
3428 	lock_super(sb);
3429 	old_sb_flags = sb->s_flags;
3430 	old_opts.s_mount_opt = sbi->s_mount_opt;
3431 	old_opts.s_resuid = sbi->s_resuid;
3432 	old_opts.s_resgid = sbi->s_resgid;
3433 	old_opts.s_commit_interval = sbi->s_commit_interval;
3434 	old_opts.s_min_batch_time = sbi->s_min_batch_time;
3435 	old_opts.s_max_batch_time = sbi->s_max_batch_time;
3436 #ifdef CONFIG_QUOTA
3437 	old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3438 	for (i = 0; i < MAXQUOTAS; i++)
3439 		old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3440 #endif
3441 	if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3442 		journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3443 
3444 	/*
3445 	 * Allow the "check" option to be passed as a remount option.
3446 	 */
3447 	if (!parse_options(data, sb, NULL, &journal_ioprio,
3448 			   &n_blocks_count, 1)) {
3449 		err = -EINVAL;
3450 		goto restore_opts;
3451 	}
3452 
3453 	if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
3454 		ext4_abort(sb, __func__, "Abort forced by user");
3455 
3456 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3457 		((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3458 
3459 	es = sbi->s_es;
3460 
3461 	if (sbi->s_journal) {
3462 		ext4_init_journal_params(sb, sbi->s_journal);
3463 		set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3464 	}
3465 
3466 	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3467 		n_blocks_count > ext4_blocks_count(es)) {
3468 		if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
3469 			err = -EROFS;
3470 			goto restore_opts;
3471 		}
3472 
3473 		if (*flags & MS_RDONLY) {
3474 			/*
3475 			 * First of all, the unconditional stuff we have to do
3476 			 * to disable replay of the journal when we next remount
3477 			 */
3478 			sb->s_flags |= MS_RDONLY;
3479 
3480 			/*
3481 			 * OK, test if we are remounting a valid rw partition
3482 			 * readonly, and if so set the rdonly flag and then
3483 			 * mark the partition as valid again.
3484 			 */
3485 			if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3486 			    (sbi->s_mount_state & EXT4_VALID_FS))
3487 				es->s_state = cpu_to_le16(sbi->s_mount_state);
3488 
3489 			if (sbi->s_journal)
3490 				ext4_mark_recovery_complete(sb, es);
3491 		} else {
3492 			int ret;
3493 			if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3494 					~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3495 				ext4_msg(sb, KERN_WARNING, "couldn't "
3496 				       "remount RDWR because of unsupported "
3497 				       "optional features (%x)",
3498 				(le32_to_cpu(sbi->s_es->s_feature_ro_compat) &
3499 					~EXT4_FEATURE_RO_COMPAT_SUPP));
3500 				err = -EROFS;
3501 				goto restore_opts;
3502 			}
3503 
3504 			/*
3505 			 * Make sure the group descriptor checksums
3506 			 * are sane.  If they aren't, refuse to remount r/w.
3507 			 */
3508 			for (g = 0; g < sbi->s_groups_count; g++) {
3509 				struct ext4_group_desc *gdp =
3510 					ext4_get_group_desc(sb, g, NULL);
3511 
3512 				if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3513 					ext4_msg(sb, KERN_ERR,
3514 	       "ext4_remount: Checksum for group %u failed (%u!=%u)",
3515 		g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3516 					       le16_to_cpu(gdp->bg_checksum));
3517 					err = -EINVAL;
3518 					goto restore_opts;
3519 				}
3520 			}
3521 
3522 			/*
3523 			 * If we have an unprocessed orphan list hanging
3524 			 * around from a previously readonly bdev mount,
3525 			 * require a full umount/remount for now.
3526 			 */
3527 			if (es->s_last_orphan) {
3528 				ext4_msg(sb, KERN_WARNING, "Couldn't "
3529 				       "remount RDWR because of unprocessed "
3530 				       "orphan inode list.  Please "
3531 				       "umount/remount instead");
3532 				err = -EINVAL;
3533 				goto restore_opts;
3534 			}
3535 
3536 			/*
3537 			 * Mounting a RDONLY partition read-write, so reread
3538 			 * and store the current valid flag.  (It may have
3539 			 * been changed by e2fsck since we originally mounted
3540 			 * the partition.)
3541 			 */
3542 			if (sbi->s_journal)
3543 				ext4_clear_journal_err(sb, es);
3544 			sbi->s_mount_state = le16_to_cpu(es->s_state);
3545 			if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3546 				goto restore_opts;
3547 			if (!ext4_setup_super(sb, es, 0))
3548 				sb->s_flags &= ~MS_RDONLY;
3549 		}
3550 	}
3551 	ext4_setup_system_zone(sb);
3552 	if (sbi->s_journal == NULL)
3553 		ext4_commit_super(sb, 1);
3554 
3555 #ifdef CONFIG_QUOTA
3556 	/* Release old quota file names */
3557 	for (i = 0; i < MAXQUOTAS; i++)
3558 		if (old_opts.s_qf_names[i] &&
3559 		    old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3560 			kfree(old_opts.s_qf_names[i]);
3561 #endif
3562 	unlock_super(sb);
3563 	unlock_kernel();
3564 	return 0;
3565 
3566 restore_opts:
3567 	sb->s_flags = old_sb_flags;
3568 	sbi->s_mount_opt = old_opts.s_mount_opt;
3569 	sbi->s_resuid = old_opts.s_resuid;
3570 	sbi->s_resgid = old_opts.s_resgid;
3571 	sbi->s_commit_interval = old_opts.s_commit_interval;
3572 	sbi->s_min_batch_time = old_opts.s_min_batch_time;
3573 	sbi->s_max_batch_time = old_opts.s_max_batch_time;
3574 #ifdef CONFIG_QUOTA
3575 	sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3576 	for (i = 0; i < MAXQUOTAS; i++) {
3577 		if (sbi->s_qf_names[i] &&
3578 		    old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3579 			kfree(sbi->s_qf_names[i]);
3580 		sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3581 	}
3582 #endif
3583 	unlock_super(sb);
3584 	unlock_kernel();
3585 	return err;
3586 }
3587 
3588 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3589 {
3590 	struct super_block *sb = dentry->d_sb;
3591 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3592 	struct ext4_super_block *es = sbi->s_es;
3593 	u64 fsid;
3594 
3595 	if (test_opt(sb, MINIX_DF)) {
3596 		sbi->s_overhead_last = 0;
3597 	} else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3598 		ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3599 		ext4_fsblk_t overhead = 0;
3600 
3601 		/*
3602 		 * Compute the overhead (FS structures).  This is constant
3603 		 * for a given filesystem unless the number of block groups
3604 		 * changes so we cache the previous value until it does.
3605 		 */
3606 
3607 		/*
3608 		 * All of the blocks before first_data_block are
3609 		 * overhead
3610 		 */
3611 		overhead = le32_to_cpu(es->s_first_data_block);
3612 
3613 		/*
3614 		 * Add the overhead attributed to the superblock and
3615 		 * block group descriptors.  If the sparse superblocks
3616 		 * feature is turned on, then not all groups have this.
3617 		 */
3618 		for (i = 0; i < ngroups; i++) {
3619 			overhead += ext4_bg_has_super(sb, i) +
3620 				ext4_bg_num_gdb(sb, i);
3621 			cond_resched();
3622 		}
3623 
3624 		/*
3625 		 * Every block group has an inode bitmap, a block
3626 		 * bitmap, and an inode table.
3627 		 */
3628 		overhead += ngroups * (2 + sbi->s_itb_per_group);
3629 		sbi->s_overhead_last = overhead;
3630 		smp_wmb();
3631 		sbi->s_blocks_last = ext4_blocks_count(es);
3632 	}
3633 
3634 	buf->f_type = EXT4_SUPER_MAGIC;
3635 	buf->f_bsize = sb->s_blocksize;
3636 	buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3637 	buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3638 		       percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3639 	ext4_free_blocks_count_set(es, buf->f_bfree);
3640 	buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3641 	if (buf->f_bfree < ext4_r_blocks_count(es))
3642 		buf->f_bavail = 0;
3643 	buf->f_files = le32_to_cpu(es->s_inodes_count);
3644 	buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3645 	es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3646 	buf->f_namelen = EXT4_NAME_LEN;
3647 	fsid = le64_to_cpup((void *)es->s_uuid) ^
3648 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3649 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3650 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3651 
3652 	return 0;
3653 }
3654 
3655 /* Helper function for writing quotas on sync - we need to start transaction
3656  * before quota file is locked for write. Otherwise the are possible deadlocks:
3657  * Process 1                         Process 2
3658  * ext4_create()                     quota_sync()
3659  *   jbd2_journal_start()                  write_dquot()
3660  *   vfs_dq_init()                         down(dqio_mutex)
3661  *     down(dqio_mutex)                    jbd2_journal_start()
3662  *
3663  */
3664 
3665 #ifdef CONFIG_QUOTA
3666 
3667 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3668 {
3669 	return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3670 }
3671 
3672 static int ext4_write_dquot(struct dquot *dquot)
3673 {
3674 	int ret, err;
3675 	handle_t *handle;
3676 	struct inode *inode;
3677 
3678 	inode = dquot_to_inode(dquot);
3679 	handle = ext4_journal_start(inode,
3680 				    EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3681 	if (IS_ERR(handle))
3682 		return PTR_ERR(handle);
3683 	ret = dquot_commit(dquot);
3684 	err = ext4_journal_stop(handle);
3685 	if (!ret)
3686 		ret = err;
3687 	return ret;
3688 }
3689 
3690 static int ext4_acquire_dquot(struct dquot *dquot)
3691 {
3692 	int ret, err;
3693 	handle_t *handle;
3694 
3695 	handle = ext4_journal_start(dquot_to_inode(dquot),
3696 				    EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3697 	if (IS_ERR(handle))
3698 		return PTR_ERR(handle);
3699 	ret = dquot_acquire(dquot);
3700 	err = ext4_journal_stop(handle);
3701 	if (!ret)
3702 		ret = err;
3703 	return ret;
3704 }
3705 
3706 static int ext4_release_dquot(struct dquot *dquot)
3707 {
3708 	int ret, err;
3709 	handle_t *handle;
3710 
3711 	handle = ext4_journal_start(dquot_to_inode(dquot),
3712 				    EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3713 	if (IS_ERR(handle)) {
3714 		/* Release dquot anyway to avoid endless cycle in dqput() */
3715 		dquot_release(dquot);
3716 		return PTR_ERR(handle);
3717 	}
3718 	ret = dquot_release(dquot);
3719 	err = ext4_journal_stop(handle);
3720 	if (!ret)
3721 		ret = err;
3722 	return ret;
3723 }
3724 
3725 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3726 {
3727 	/* Are we journaling quotas? */
3728 	if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3729 	    EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3730 		dquot_mark_dquot_dirty(dquot);
3731 		return ext4_write_dquot(dquot);
3732 	} else {
3733 		return dquot_mark_dquot_dirty(dquot);
3734 	}
3735 }
3736 
3737 static int ext4_write_info(struct super_block *sb, int type)
3738 {
3739 	int ret, err;
3740 	handle_t *handle;
3741 
3742 	/* Data block + inode block */
3743 	handle = ext4_journal_start(sb->s_root->d_inode, 2);
3744 	if (IS_ERR(handle))
3745 		return PTR_ERR(handle);
3746 	ret = dquot_commit_info(sb, type);
3747 	err = ext4_journal_stop(handle);
3748 	if (!ret)
3749 		ret = err;
3750 	return ret;
3751 }
3752 
3753 /*
3754  * Turn on quotas during mount time - we need to find
3755  * the quota file and such...
3756  */
3757 static int ext4_quota_on_mount(struct super_block *sb, int type)
3758 {
3759 	return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3760 				  EXT4_SB(sb)->s_jquota_fmt, type);
3761 }
3762 
3763 /*
3764  * Standard function to be called on quota_on
3765  */
3766 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3767 			 char *name, int remount)
3768 {
3769 	int err;
3770 	struct path path;
3771 
3772 	if (!test_opt(sb, QUOTA))
3773 		return -EINVAL;
3774 	/* When remounting, no checks are needed and in fact, name is NULL */
3775 	if (remount)
3776 		return vfs_quota_on(sb, type, format_id, name, remount);
3777 
3778 	err = kern_path(name, LOOKUP_FOLLOW, &path);
3779 	if (err)
3780 		return err;
3781 
3782 	/* Quotafile not on the same filesystem? */
3783 	if (path.mnt->mnt_sb != sb) {
3784 		path_put(&path);
3785 		return -EXDEV;
3786 	}
3787 	/* Journaling quota? */
3788 	if (EXT4_SB(sb)->s_qf_names[type]) {
3789 		/* Quotafile not in fs root? */
3790 		if (path.dentry->d_parent != sb->s_root)
3791 			ext4_msg(sb, KERN_WARNING,
3792 				"Quota file not on filesystem root. "
3793 				"Journaled quota will not work");
3794 	}
3795 
3796 	/*
3797 	 * When we journal data on quota file, we have to flush journal to see
3798 	 * all updates to the file when we bypass pagecache...
3799 	 */
3800 	if (EXT4_SB(sb)->s_journal &&
3801 	    ext4_should_journal_data(path.dentry->d_inode)) {
3802 		/*
3803 		 * We don't need to lock updates but journal_flush() could
3804 		 * otherwise be livelocked...
3805 		 */
3806 		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3807 		err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3808 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3809 		if (err) {
3810 			path_put(&path);
3811 			return err;
3812 		}
3813 	}
3814 
3815 	err = vfs_quota_on_path(sb, type, format_id, &path);
3816 	path_put(&path);
3817 	return err;
3818 }
3819 
3820 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3821  * acquiring the locks... As quota files are never truncated and quota code
3822  * itself serializes the operations (and noone else should touch the files)
3823  * we don't have to be afraid of races */
3824 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3825 			       size_t len, loff_t off)
3826 {
3827 	struct inode *inode = sb_dqopt(sb)->files[type];
3828 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3829 	int err = 0;
3830 	int offset = off & (sb->s_blocksize - 1);
3831 	int tocopy;
3832 	size_t toread;
3833 	struct buffer_head *bh;
3834 	loff_t i_size = i_size_read(inode);
3835 
3836 	if (off > i_size)
3837 		return 0;
3838 	if (off+len > i_size)
3839 		len = i_size-off;
3840 	toread = len;
3841 	while (toread > 0) {
3842 		tocopy = sb->s_blocksize - offset < toread ?
3843 				sb->s_blocksize - offset : toread;
3844 		bh = ext4_bread(NULL, inode, blk, 0, &err);
3845 		if (err)
3846 			return err;
3847 		if (!bh)	/* A hole? */
3848 			memset(data, 0, tocopy);
3849 		else
3850 			memcpy(data, bh->b_data+offset, tocopy);
3851 		brelse(bh);
3852 		offset = 0;
3853 		toread -= tocopy;
3854 		data += tocopy;
3855 		blk++;
3856 	}
3857 	return len;
3858 }
3859 
3860 /* Write to quotafile (we know the transaction is already started and has
3861  * enough credits) */
3862 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3863 				const char *data, size_t len, loff_t off)
3864 {
3865 	struct inode *inode = sb_dqopt(sb)->files[type];
3866 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3867 	int err = 0;
3868 	int offset = off & (sb->s_blocksize - 1);
3869 	int tocopy;
3870 	int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3871 	size_t towrite = len;
3872 	struct buffer_head *bh;
3873 	handle_t *handle = journal_current_handle();
3874 
3875 	if (EXT4_SB(sb)->s_journal && !handle) {
3876 		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3877 			" cancelled because transaction is not started",
3878 			(unsigned long long)off, (unsigned long long)len);
3879 		return -EIO;
3880 	}
3881 	mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3882 	while (towrite > 0) {
3883 		tocopy = sb->s_blocksize - offset < towrite ?
3884 				sb->s_blocksize - offset : towrite;
3885 		bh = ext4_bread(handle, inode, blk, 1, &err);
3886 		if (!bh)
3887 			goto out;
3888 		if (journal_quota) {
3889 			err = ext4_journal_get_write_access(handle, bh);
3890 			if (err) {
3891 				brelse(bh);
3892 				goto out;
3893 			}
3894 		}
3895 		lock_buffer(bh);
3896 		memcpy(bh->b_data+offset, data, tocopy);
3897 		flush_dcache_page(bh->b_page);
3898 		unlock_buffer(bh);
3899 		if (journal_quota)
3900 			err = ext4_handle_dirty_metadata(handle, NULL, bh);
3901 		else {
3902 			/* Always do at least ordered writes for quotas */
3903 			err = ext4_jbd2_file_inode(handle, inode);
3904 			mark_buffer_dirty(bh);
3905 		}
3906 		brelse(bh);
3907 		if (err)
3908 			goto out;
3909 		offset = 0;
3910 		towrite -= tocopy;
3911 		data += tocopy;
3912 		blk++;
3913 	}
3914 out:
3915 	if (len == towrite) {
3916 		mutex_unlock(&inode->i_mutex);
3917 		return err;
3918 	}
3919 	if (inode->i_size < off+len-towrite) {
3920 		i_size_write(inode, off+len-towrite);
3921 		EXT4_I(inode)->i_disksize = inode->i_size;
3922 	}
3923 	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3924 	ext4_mark_inode_dirty(handle, inode);
3925 	mutex_unlock(&inode->i_mutex);
3926 	return len - towrite;
3927 }
3928 
3929 #endif
3930 
3931 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3932 		       const char *dev_name, void *data, struct vfsmount *mnt)
3933 {
3934 	return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3935 }
3936 
3937 static struct file_system_type ext4_fs_type = {
3938 	.owner		= THIS_MODULE,
3939 	.name		= "ext4",
3940 	.get_sb		= ext4_get_sb,
3941 	.kill_sb	= kill_block_super,
3942 	.fs_flags	= FS_REQUIRES_DEV,
3943 };
3944 
3945 #ifdef CONFIG_EXT4DEV_COMPAT
3946 static int ext4dev_get_sb(struct file_system_type *fs_type, int flags,
3947 			  const char *dev_name, void *data,struct vfsmount *mnt)
3948 {
3949 	printk(KERN_WARNING "EXT4-fs (%s): Update your userspace programs "
3950 	       "to mount using ext4\n", dev_name);
3951 	printk(KERN_WARNING "EXT4-fs (%s): ext4dev backwards compatibility "
3952 	       "will go away by 2.6.31\n", dev_name);
3953 	return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3954 }
3955 
3956 static struct file_system_type ext4dev_fs_type = {
3957 	.owner		= THIS_MODULE,
3958 	.name		= "ext4dev",
3959 	.get_sb		= ext4dev_get_sb,
3960 	.kill_sb	= kill_block_super,
3961 	.fs_flags	= FS_REQUIRES_DEV,
3962 };
3963 MODULE_ALIAS("ext4dev");
3964 #endif
3965 
3966 static int __init init_ext4_fs(void)
3967 {
3968 	int err;
3969 
3970 	err = init_ext4_system_zone();
3971 	if (err)
3972 		return err;
3973 	ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3974 	if (!ext4_kset)
3975 		goto out4;
3976 	ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3977 	err = init_ext4_mballoc();
3978 	if (err)
3979 		goto out3;
3980 
3981 	err = init_ext4_xattr();
3982 	if (err)
3983 		goto out2;
3984 	err = init_inodecache();
3985 	if (err)
3986 		goto out1;
3987 	err = register_filesystem(&ext4_fs_type);
3988 	if (err)
3989 		goto out;
3990 #ifdef CONFIG_EXT4DEV_COMPAT
3991 	err = register_filesystem(&ext4dev_fs_type);
3992 	if (err) {
3993 		unregister_filesystem(&ext4_fs_type);
3994 		goto out;
3995 	}
3996 #endif
3997 	return 0;
3998 out:
3999 	destroy_inodecache();
4000 out1:
4001 	exit_ext4_xattr();
4002 out2:
4003 	exit_ext4_mballoc();
4004 out3:
4005 	remove_proc_entry("fs/ext4", NULL);
4006 	kset_unregister(ext4_kset);
4007 out4:
4008 	exit_ext4_system_zone();
4009 	return err;
4010 }
4011 
4012 static void __exit exit_ext4_fs(void)
4013 {
4014 	unregister_filesystem(&ext4_fs_type);
4015 #ifdef CONFIG_EXT4DEV_COMPAT
4016 	unregister_filesystem(&ext4dev_fs_type);
4017 #endif
4018 	destroy_inodecache();
4019 	exit_ext4_xattr();
4020 	exit_ext4_mballoc();
4021 	remove_proc_entry("fs/ext4", NULL);
4022 	kset_unregister(ext4_kset);
4023 	exit_ext4_system_zone();
4024 }
4025 
4026 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4027 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4028 MODULE_LICENSE("GPL");
4029 module_init(init_ext4_fs)
4030 module_exit(exit_ext4_fs)
4031