xref: /openbmc/linux/fs/ext4/super.c (revision b43fa8284d7790d9cca32c9c55e24f29be2fa33b)
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/log2.h>
41 #include <linux/crc16.h>
42 #include <asm/uaccess.h>
43 
44 #include "ext4.h"
45 #include "ext4_jbd2.h"
46 #include "xattr.h"
47 #include "acl.h"
48 #include "mballoc.h"
49 
50 #define CREATE_TRACE_POINTS
51 #include <trace/events/ext4.h>
52 
53 struct proc_dir_entry *ext4_proc_root;
54 static struct kset *ext4_kset;
55 
56 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
57 			     unsigned long journal_devnum);
58 static int ext4_commit_super(struct super_block *sb, int sync);
59 static void ext4_mark_recovery_complete(struct super_block *sb,
60 					struct ext4_super_block *es);
61 static void ext4_clear_journal_err(struct super_block *sb,
62 				   struct ext4_super_block *es);
63 static int ext4_sync_fs(struct super_block *sb, int wait);
64 static const char *ext4_decode_error(struct super_block *sb, int errno,
65 				     char nbuf[16]);
66 static int ext4_remount(struct super_block *sb, int *flags, char *data);
67 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
68 static int ext4_unfreeze(struct super_block *sb);
69 static void ext4_write_super(struct super_block *sb);
70 static int ext4_freeze(struct super_block *sb);
71 
72 
73 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
74 			       struct ext4_group_desc *bg)
75 {
76 	return le32_to_cpu(bg->bg_block_bitmap_lo) |
77 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
78 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
79 }
80 
81 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
82 			       struct ext4_group_desc *bg)
83 {
84 	return le32_to_cpu(bg->bg_inode_bitmap_lo) |
85 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
86 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
87 }
88 
89 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
90 			      struct ext4_group_desc *bg)
91 {
92 	return le32_to_cpu(bg->bg_inode_table_lo) |
93 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
94 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
95 }
96 
97 __u32 ext4_free_blks_count(struct super_block *sb,
98 			      struct ext4_group_desc *bg)
99 {
100 	return le16_to_cpu(bg->bg_free_blocks_count_lo) |
101 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
102 		 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
103 }
104 
105 __u32 ext4_free_inodes_count(struct super_block *sb,
106 			      struct ext4_group_desc *bg)
107 {
108 	return le16_to_cpu(bg->bg_free_inodes_count_lo) |
109 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
110 		 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
111 }
112 
113 __u32 ext4_used_dirs_count(struct super_block *sb,
114 			      struct ext4_group_desc *bg)
115 {
116 	return le16_to_cpu(bg->bg_used_dirs_count_lo) |
117 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
118 		 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
119 }
120 
121 __u32 ext4_itable_unused_count(struct super_block *sb,
122 			      struct ext4_group_desc *bg)
123 {
124 	return le16_to_cpu(bg->bg_itable_unused_lo) |
125 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
126 		 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
127 }
128 
129 void ext4_block_bitmap_set(struct super_block *sb,
130 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
131 {
132 	bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
133 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
134 		bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
135 }
136 
137 void ext4_inode_bitmap_set(struct super_block *sb,
138 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
139 {
140 	bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
141 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
142 		bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
143 }
144 
145 void ext4_inode_table_set(struct super_block *sb,
146 			  struct ext4_group_desc *bg, ext4_fsblk_t blk)
147 {
148 	bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
149 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
150 		bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
151 }
152 
153 void ext4_free_blks_set(struct super_block *sb,
154 			  struct ext4_group_desc *bg, __u32 count)
155 {
156 	bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
157 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
158 		bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
159 }
160 
161 void ext4_free_inodes_set(struct super_block *sb,
162 			  struct ext4_group_desc *bg, __u32 count)
163 {
164 	bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
165 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
166 		bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
167 }
168 
169 void ext4_used_dirs_set(struct super_block *sb,
170 			  struct ext4_group_desc *bg, __u32 count)
171 {
172 	bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
173 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
174 		bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
175 }
176 
177 void ext4_itable_unused_set(struct super_block *sb,
178 			  struct ext4_group_desc *bg, __u32 count)
179 {
180 	bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
181 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
182 		bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
183 }
184 
185 
186 /* Just increment the non-pointer handle value */
187 static handle_t *ext4_get_nojournal(void)
188 {
189 	handle_t *handle = current->journal_info;
190 	unsigned long ref_cnt = (unsigned long)handle;
191 
192 	BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
193 
194 	ref_cnt++;
195 	handle = (handle_t *)ref_cnt;
196 
197 	current->journal_info = handle;
198 	return handle;
199 }
200 
201 
202 /* Decrement the non-pointer handle value */
203 static void ext4_put_nojournal(handle_t *handle)
204 {
205 	unsigned long ref_cnt = (unsigned long)handle;
206 
207 	BUG_ON(ref_cnt == 0);
208 
209 	ref_cnt--;
210 	handle = (handle_t *)ref_cnt;
211 
212 	current->journal_info = handle;
213 }
214 
215 /*
216  * Wrappers for jbd2_journal_start/end.
217  *
218  * The only special thing we need to do here is to make sure that all
219  * journal_end calls result in the superblock being marked dirty, so
220  * that sync() will call the filesystem's write_super callback if
221  * appropriate.
222  */
223 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
224 {
225 	journal_t *journal;
226 
227 	if (sb->s_flags & MS_RDONLY)
228 		return ERR_PTR(-EROFS);
229 
230 	/* Special case here: if the journal has aborted behind our
231 	 * backs (eg. EIO in the commit thread), then we still need to
232 	 * take the FS itself readonly cleanly. */
233 	journal = EXT4_SB(sb)->s_journal;
234 	if (journal) {
235 		if (is_journal_aborted(journal)) {
236 			ext4_abort(sb, __func__, "Detected aborted journal");
237 			return ERR_PTR(-EROFS);
238 		}
239 		return jbd2_journal_start(journal, nblocks);
240 	}
241 	return ext4_get_nojournal();
242 }
243 
244 /*
245  * The only special thing we need to do here is to make sure that all
246  * jbd2_journal_stop calls result in the superblock being marked dirty, so
247  * that sync() will call the filesystem's write_super callback if
248  * appropriate.
249  */
250 int __ext4_journal_stop(const char *where, handle_t *handle)
251 {
252 	struct super_block *sb;
253 	int err;
254 	int rc;
255 
256 	if (!ext4_handle_valid(handle)) {
257 		ext4_put_nojournal(handle);
258 		return 0;
259 	}
260 	sb = handle->h_transaction->t_journal->j_private;
261 	err = handle->h_err;
262 	rc = jbd2_journal_stop(handle);
263 
264 	if (!err)
265 		err = rc;
266 	if (err)
267 		__ext4_std_error(sb, where, err);
268 	return err;
269 }
270 
271 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
272 		struct buffer_head *bh, handle_t *handle, int err)
273 {
274 	char nbuf[16];
275 	const char *errstr = ext4_decode_error(NULL, err, nbuf);
276 
277 	BUG_ON(!ext4_handle_valid(handle));
278 
279 	if (bh)
280 		BUFFER_TRACE(bh, "abort");
281 
282 	if (!handle->h_err)
283 		handle->h_err = err;
284 
285 	if (is_handle_aborted(handle))
286 		return;
287 
288 	printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
289 	       caller, errstr, err_fn);
290 
291 	jbd2_journal_abort_handle(handle);
292 }
293 
294 /* Deal with the reporting of failure conditions on a filesystem such as
295  * inconsistencies detected or read IO failures.
296  *
297  * On ext2, we can store the error state of the filesystem in the
298  * superblock.  That is not possible on ext4, because we may have other
299  * write ordering constraints on the superblock which prevent us from
300  * writing it out straight away; and given that the journal is about to
301  * be aborted, we can't rely on the current, or future, transactions to
302  * write out the superblock safely.
303  *
304  * We'll just use the jbd2_journal_abort() error code to record an error in
305  * the journal instead.  On recovery, the journal will compain about
306  * that error until we've noted it down and cleared it.
307  */
308 
309 static void ext4_handle_error(struct super_block *sb)
310 {
311 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
312 
313 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
314 	es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
315 
316 	if (sb->s_flags & MS_RDONLY)
317 		return;
318 
319 	if (!test_opt(sb, ERRORS_CONT)) {
320 		journal_t *journal = EXT4_SB(sb)->s_journal;
321 
322 		EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
323 		if (journal)
324 			jbd2_journal_abort(journal, -EIO);
325 	}
326 	if (test_opt(sb, ERRORS_RO)) {
327 		ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
328 		sb->s_flags |= MS_RDONLY;
329 	}
330 	ext4_commit_super(sb, 1);
331 	if (test_opt(sb, ERRORS_PANIC))
332 		panic("EXT4-fs (device %s): panic forced after error\n",
333 			sb->s_id);
334 }
335 
336 void ext4_error(struct super_block *sb, const char *function,
337 		const char *fmt, ...)
338 {
339 	va_list args;
340 
341 	va_start(args, fmt);
342 	printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
343 	vprintk(fmt, args);
344 	printk("\n");
345 	va_end(args);
346 
347 	ext4_handle_error(sb);
348 }
349 
350 static const char *ext4_decode_error(struct super_block *sb, int errno,
351 				     char nbuf[16])
352 {
353 	char *errstr = NULL;
354 
355 	switch (errno) {
356 	case -EIO:
357 		errstr = "IO failure";
358 		break;
359 	case -ENOMEM:
360 		errstr = "Out of memory";
361 		break;
362 	case -EROFS:
363 		if (!sb || (EXT4_SB(sb)->s_journal &&
364 			    EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
365 			errstr = "Journal has aborted";
366 		else
367 			errstr = "Readonly filesystem";
368 		break;
369 	default:
370 		/* If the caller passed in an extra buffer for unknown
371 		 * errors, textualise them now.  Else we just return
372 		 * NULL. */
373 		if (nbuf) {
374 			/* Check for truncated error codes... */
375 			if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
376 				errstr = nbuf;
377 		}
378 		break;
379 	}
380 
381 	return errstr;
382 }
383 
384 /* __ext4_std_error decodes expected errors from journaling functions
385  * automatically and invokes the appropriate error response.  */
386 
387 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
388 {
389 	char nbuf[16];
390 	const char *errstr;
391 
392 	/* Special case: if the error is EROFS, and we're not already
393 	 * inside a transaction, then there's really no point in logging
394 	 * an error. */
395 	if (errno == -EROFS && journal_current_handle() == NULL &&
396 	    (sb->s_flags & MS_RDONLY))
397 		return;
398 
399 	errstr = ext4_decode_error(sb, errno, nbuf);
400 	printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
401 	       sb->s_id, function, errstr);
402 
403 	ext4_handle_error(sb);
404 }
405 
406 /*
407  * ext4_abort is a much stronger failure handler than ext4_error.  The
408  * abort function may be used to deal with unrecoverable failures such
409  * as journal IO errors or ENOMEM at a critical moment in log management.
410  *
411  * We unconditionally force the filesystem into an ABORT|READONLY state,
412  * unless the error response on the fs has been set to panic in which
413  * case we take the easy way out and panic immediately.
414  */
415 
416 void ext4_abort(struct super_block *sb, const char *function,
417 		const char *fmt, ...)
418 {
419 	va_list args;
420 
421 	va_start(args, fmt);
422 	printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
423 	vprintk(fmt, args);
424 	printk("\n");
425 	va_end(args);
426 
427 	if (test_opt(sb, ERRORS_PANIC))
428 		panic("EXT4-fs panic from previous error\n");
429 
430 	if (sb->s_flags & MS_RDONLY)
431 		return;
432 
433 	ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
434 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
435 	sb->s_flags |= MS_RDONLY;
436 	EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
437 	if (EXT4_SB(sb)->s_journal)
438 		jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
439 }
440 
441 void ext4_msg (struct super_block * sb, const char *prefix,
442 		   const char *fmt, ...)
443 {
444 	va_list args;
445 
446 	va_start(args, fmt);
447 	printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
448 	vprintk(fmt, args);
449 	printk("\n");
450 	va_end(args);
451 }
452 
453 void ext4_warning(struct super_block *sb, const char *function,
454 		  const char *fmt, ...)
455 {
456 	va_list args;
457 
458 	va_start(args, fmt);
459 	printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
460 	       sb->s_id, function);
461 	vprintk(fmt, args);
462 	printk("\n");
463 	va_end(args);
464 }
465 
466 void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
467 			   const char *function, const char *fmt, ...)
468 __releases(bitlock)
469 __acquires(bitlock)
470 {
471 	va_list args;
472 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
473 
474 	va_start(args, fmt);
475 	printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
476 	vprintk(fmt, args);
477 	printk("\n");
478 	va_end(args);
479 
480 	if (test_opt(sb, ERRORS_CONT)) {
481 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
482 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
483 		ext4_commit_super(sb, 0);
484 		return;
485 	}
486 	ext4_unlock_group(sb, grp);
487 	ext4_handle_error(sb);
488 	/*
489 	 * We only get here in the ERRORS_RO case; relocking the group
490 	 * may be dangerous, but nothing bad will happen since the
491 	 * filesystem will have already been marked read/only and the
492 	 * journal has been aborted.  We return 1 as a hint to callers
493 	 * who might what to use the return value from
494 	 * ext4_grp_locked_error() to distinguish beween the
495 	 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
496 	 * aggressively from the ext4 function in question, with a
497 	 * more appropriate error code.
498 	 */
499 	ext4_lock_group(sb, grp);
500 	return;
501 }
502 
503 void ext4_update_dynamic_rev(struct super_block *sb)
504 {
505 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
506 
507 	if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
508 		return;
509 
510 	ext4_warning(sb, __func__,
511 		     "updating to rev %d because of new feature flag, "
512 		     "running e2fsck is recommended",
513 		     EXT4_DYNAMIC_REV);
514 
515 	es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
516 	es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
517 	es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
518 	/* leave es->s_feature_*compat flags alone */
519 	/* es->s_uuid will be set by e2fsck if empty */
520 
521 	/*
522 	 * The rest of the superblock fields should be zero, and if not it
523 	 * means they are likely already in use, so leave them alone.  We
524 	 * can leave it up to e2fsck to clean up any inconsistencies there.
525 	 */
526 }
527 
528 /*
529  * Open the external journal device
530  */
531 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
532 {
533 	struct block_device *bdev;
534 	char b[BDEVNAME_SIZE];
535 
536 	bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
537 	if (IS_ERR(bdev))
538 		goto fail;
539 	return bdev;
540 
541 fail:
542 	ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
543 			__bdevname(dev, b), PTR_ERR(bdev));
544 	return NULL;
545 }
546 
547 /*
548  * Release the journal device
549  */
550 static int ext4_blkdev_put(struct block_device *bdev)
551 {
552 	bd_release(bdev);
553 	return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
554 }
555 
556 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
557 {
558 	struct block_device *bdev;
559 	int ret = -ENODEV;
560 
561 	bdev = sbi->journal_bdev;
562 	if (bdev) {
563 		ret = ext4_blkdev_put(bdev);
564 		sbi->journal_bdev = NULL;
565 	}
566 	return ret;
567 }
568 
569 static inline struct inode *orphan_list_entry(struct list_head *l)
570 {
571 	return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
572 }
573 
574 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
575 {
576 	struct list_head *l;
577 
578 	ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
579 		 le32_to_cpu(sbi->s_es->s_last_orphan));
580 
581 	printk(KERN_ERR "sb_info orphan list:\n");
582 	list_for_each(l, &sbi->s_orphan) {
583 		struct inode *inode = orphan_list_entry(l);
584 		printk(KERN_ERR "  "
585 		       "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
586 		       inode->i_sb->s_id, inode->i_ino, inode,
587 		       inode->i_mode, inode->i_nlink,
588 		       NEXT_ORPHAN(inode));
589 	}
590 }
591 
592 static void ext4_put_super(struct super_block *sb)
593 {
594 	struct ext4_sb_info *sbi = EXT4_SB(sb);
595 	struct ext4_super_block *es = sbi->s_es;
596 	int i, err;
597 
598 	flush_workqueue(sbi->dio_unwritten_wq);
599 	destroy_workqueue(sbi->dio_unwritten_wq);
600 
601 	lock_super(sb);
602 	lock_kernel();
603 	if (sb->s_dirt)
604 		ext4_commit_super(sb, 1);
605 
606 	if (sbi->s_journal) {
607 		err = jbd2_journal_destroy(sbi->s_journal);
608 		sbi->s_journal = NULL;
609 		if (err < 0)
610 			ext4_abort(sb, __func__,
611 				   "Couldn't clean up the journal");
612 	}
613 
614 	ext4_release_system_zone(sb);
615 	ext4_mb_release(sb);
616 	ext4_ext_release(sb);
617 	ext4_xattr_put_super(sb);
618 
619 	if (!(sb->s_flags & MS_RDONLY)) {
620 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
621 		es->s_state = cpu_to_le16(sbi->s_mount_state);
622 		ext4_commit_super(sb, 1);
623 	}
624 	if (sbi->s_proc) {
625 		remove_proc_entry(sb->s_id, ext4_proc_root);
626 	}
627 	kobject_del(&sbi->s_kobj);
628 
629 	for (i = 0; i < sbi->s_gdb_count; i++)
630 		brelse(sbi->s_group_desc[i]);
631 	kfree(sbi->s_group_desc);
632 	if (is_vmalloc_addr(sbi->s_flex_groups))
633 		vfree(sbi->s_flex_groups);
634 	else
635 		kfree(sbi->s_flex_groups);
636 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
637 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
638 	percpu_counter_destroy(&sbi->s_dirs_counter);
639 	percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
640 	brelse(sbi->s_sbh);
641 #ifdef CONFIG_QUOTA
642 	for (i = 0; i < MAXQUOTAS; i++)
643 		kfree(sbi->s_qf_names[i]);
644 #endif
645 
646 	/* Debugging code just in case the in-memory inode orphan list
647 	 * isn't empty.  The on-disk one can be non-empty if we've
648 	 * detected an error and taken the fs readonly, but the
649 	 * in-memory list had better be clean by this point. */
650 	if (!list_empty(&sbi->s_orphan))
651 		dump_orphan_list(sb, sbi);
652 	J_ASSERT(list_empty(&sbi->s_orphan));
653 
654 	invalidate_bdev(sb->s_bdev);
655 	if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
656 		/*
657 		 * Invalidate the journal device's buffers.  We don't want them
658 		 * floating about in memory - the physical journal device may
659 		 * hotswapped, and it breaks the `ro-after' testing code.
660 		 */
661 		sync_blockdev(sbi->journal_bdev);
662 		invalidate_bdev(sbi->journal_bdev);
663 		ext4_blkdev_remove(sbi);
664 	}
665 	sb->s_fs_info = NULL;
666 	/*
667 	 * Now that we are completely done shutting down the
668 	 * superblock, we need to actually destroy the kobject.
669 	 */
670 	unlock_kernel();
671 	unlock_super(sb);
672 	kobject_put(&sbi->s_kobj);
673 	wait_for_completion(&sbi->s_kobj_unregister);
674 	kfree(sbi->s_blockgroup_lock);
675 	kfree(sbi);
676 }
677 
678 static struct kmem_cache *ext4_inode_cachep;
679 
680 /*
681  * Called inside transaction, so use GFP_NOFS
682  */
683 static struct inode *ext4_alloc_inode(struct super_block *sb)
684 {
685 	struct ext4_inode_info *ei;
686 
687 	ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
688 	if (!ei)
689 		return NULL;
690 
691 	ei->vfs_inode.i_version = 1;
692 	ei->vfs_inode.i_data.writeback_index = 0;
693 	memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
694 	INIT_LIST_HEAD(&ei->i_prealloc_list);
695 	spin_lock_init(&ei->i_prealloc_lock);
696 	/*
697 	 * Note:  We can be called before EXT4_SB(sb)->s_journal is set,
698 	 * therefore it can be null here.  Don't check it, just initialize
699 	 * jinode.
700 	 */
701 	jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
702 	ei->i_reserved_data_blocks = 0;
703 	ei->i_reserved_meta_blocks = 0;
704 	ei->i_allocated_meta_blocks = 0;
705 	ei->i_da_metadata_calc_len = 0;
706 	ei->i_delalloc_reserved_flag = 0;
707 	spin_lock_init(&(ei->i_block_reservation_lock));
708 #ifdef CONFIG_QUOTA
709 	ei->i_reserved_quota = 0;
710 #endif
711 	INIT_LIST_HEAD(&ei->i_aio_dio_complete_list);
712 	ei->cur_aio_dio = NULL;
713 	ei->i_sync_tid = 0;
714 	ei->i_datasync_tid = 0;
715 
716 	return &ei->vfs_inode;
717 }
718 
719 static void ext4_destroy_inode(struct inode *inode)
720 {
721 	if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
722 		ext4_msg(inode->i_sb, KERN_ERR,
723 			 "Inode %lu (%p): orphan list check failed!",
724 			 inode->i_ino, EXT4_I(inode));
725 		print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
726 				EXT4_I(inode), sizeof(struct ext4_inode_info),
727 				true);
728 		dump_stack();
729 	}
730 	kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
731 }
732 
733 static void init_once(void *foo)
734 {
735 	struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
736 
737 	INIT_LIST_HEAD(&ei->i_orphan);
738 #ifdef CONFIG_EXT4_FS_XATTR
739 	init_rwsem(&ei->xattr_sem);
740 #endif
741 	init_rwsem(&ei->i_data_sem);
742 	inode_init_once(&ei->vfs_inode);
743 }
744 
745 static int init_inodecache(void)
746 {
747 	ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
748 					     sizeof(struct ext4_inode_info),
749 					     0, (SLAB_RECLAIM_ACCOUNT|
750 						SLAB_MEM_SPREAD),
751 					     init_once);
752 	if (ext4_inode_cachep == NULL)
753 		return -ENOMEM;
754 	return 0;
755 }
756 
757 static void destroy_inodecache(void)
758 {
759 	kmem_cache_destroy(ext4_inode_cachep);
760 }
761 
762 static void ext4_clear_inode(struct inode *inode)
763 {
764 	ext4_discard_preallocations(inode);
765 	if (EXT4_JOURNAL(inode))
766 		jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
767 				       &EXT4_I(inode)->jinode);
768 }
769 
770 static inline void ext4_show_quota_options(struct seq_file *seq,
771 					   struct super_block *sb)
772 {
773 #if defined(CONFIG_QUOTA)
774 	struct ext4_sb_info *sbi = EXT4_SB(sb);
775 
776 	if (sbi->s_jquota_fmt) {
777 		char *fmtname = "";
778 
779 		switch (sbi->s_jquota_fmt) {
780 		case QFMT_VFS_OLD:
781 			fmtname = "vfsold";
782 			break;
783 		case QFMT_VFS_V0:
784 			fmtname = "vfsv0";
785 			break;
786 		case QFMT_VFS_V1:
787 			fmtname = "vfsv1";
788 			break;
789 		}
790 		seq_printf(seq, ",jqfmt=%s", fmtname);
791 	}
792 
793 	if (sbi->s_qf_names[USRQUOTA])
794 		seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
795 
796 	if (sbi->s_qf_names[GRPQUOTA])
797 		seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
798 
799 	if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
800 		seq_puts(seq, ",usrquota");
801 
802 	if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
803 		seq_puts(seq, ",grpquota");
804 #endif
805 }
806 
807 /*
808  * Show an option if
809  *  - it's set to a non-default value OR
810  *  - if the per-sb default is different from the global default
811  */
812 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
813 {
814 	int def_errors;
815 	unsigned long def_mount_opts;
816 	struct super_block *sb = vfs->mnt_sb;
817 	struct ext4_sb_info *sbi = EXT4_SB(sb);
818 	struct ext4_super_block *es = sbi->s_es;
819 
820 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
821 	def_errors     = le16_to_cpu(es->s_errors);
822 
823 	if (sbi->s_sb_block != 1)
824 		seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
825 	if (test_opt(sb, MINIX_DF))
826 		seq_puts(seq, ",minixdf");
827 	if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
828 		seq_puts(seq, ",grpid");
829 	if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
830 		seq_puts(seq, ",nogrpid");
831 	if (sbi->s_resuid != EXT4_DEF_RESUID ||
832 	    le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
833 		seq_printf(seq, ",resuid=%u", sbi->s_resuid);
834 	}
835 	if (sbi->s_resgid != EXT4_DEF_RESGID ||
836 	    le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
837 		seq_printf(seq, ",resgid=%u", sbi->s_resgid);
838 	}
839 	if (test_opt(sb, ERRORS_RO)) {
840 		if (def_errors == EXT4_ERRORS_PANIC ||
841 		    def_errors == EXT4_ERRORS_CONTINUE) {
842 			seq_puts(seq, ",errors=remount-ro");
843 		}
844 	}
845 	if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
846 		seq_puts(seq, ",errors=continue");
847 	if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
848 		seq_puts(seq, ",errors=panic");
849 	if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
850 		seq_puts(seq, ",nouid32");
851 	if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
852 		seq_puts(seq, ",debug");
853 	if (test_opt(sb, OLDALLOC))
854 		seq_puts(seq, ",oldalloc");
855 #ifdef CONFIG_EXT4_FS_XATTR
856 	if (test_opt(sb, XATTR_USER) &&
857 		!(def_mount_opts & EXT4_DEFM_XATTR_USER))
858 		seq_puts(seq, ",user_xattr");
859 	if (!test_opt(sb, XATTR_USER) &&
860 	    (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
861 		seq_puts(seq, ",nouser_xattr");
862 	}
863 #endif
864 #ifdef CONFIG_EXT4_FS_POSIX_ACL
865 	if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
866 		seq_puts(seq, ",acl");
867 	if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
868 		seq_puts(seq, ",noacl");
869 #endif
870 	if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
871 		seq_printf(seq, ",commit=%u",
872 			   (unsigned) (sbi->s_commit_interval / HZ));
873 	}
874 	if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
875 		seq_printf(seq, ",min_batch_time=%u",
876 			   (unsigned) sbi->s_min_batch_time);
877 	}
878 	if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
879 		seq_printf(seq, ",max_batch_time=%u",
880 			   (unsigned) sbi->s_min_batch_time);
881 	}
882 
883 	/*
884 	 * We're changing the default of barrier mount option, so
885 	 * let's always display its mount state so it's clear what its
886 	 * status is.
887 	 */
888 	seq_puts(seq, ",barrier=");
889 	seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
890 	if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
891 		seq_puts(seq, ",journal_async_commit");
892 	if (test_opt(sb, NOBH))
893 		seq_puts(seq, ",nobh");
894 	if (test_opt(sb, I_VERSION))
895 		seq_puts(seq, ",i_version");
896 	if (!test_opt(sb, DELALLOC))
897 		seq_puts(seq, ",nodelalloc");
898 
899 
900 	if (sbi->s_stripe)
901 		seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
902 	/*
903 	 * journal mode get enabled in different ways
904 	 * So just print the value even if we didn't specify it
905 	 */
906 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
907 		seq_puts(seq, ",data=journal");
908 	else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
909 		seq_puts(seq, ",data=ordered");
910 	else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
911 		seq_puts(seq, ",data=writeback");
912 
913 	if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
914 		seq_printf(seq, ",inode_readahead_blks=%u",
915 			   sbi->s_inode_readahead_blks);
916 
917 	if (test_opt(sb, DATA_ERR_ABORT))
918 		seq_puts(seq, ",data_err=abort");
919 
920 	if (test_opt(sb, NO_AUTO_DA_ALLOC))
921 		seq_puts(seq, ",noauto_da_alloc");
922 
923 	if (test_opt(sb, DISCARD))
924 		seq_puts(seq, ",discard");
925 
926 	if (test_opt(sb, NOLOAD))
927 		seq_puts(seq, ",norecovery");
928 
929 	ext4_show_quota_options(seq, sb);
930 
931 	return 0;
932 }
933 
934 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
935 					u64 ino, u32 generation)
936 {
937 	struct inode *inode;
938 
939 	if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
940 		return ERR_PTR(-ESTALE);
941 	if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
942 		return ERR_PTR(-ESTALE);
943 
944 	/* iget isn't really right if the inode is currently unallocated!!
945 	 *
946 	 * ext4_read_inode will return a bad_inode if the inode had been
947 	 * deleted, so we should be safe.
948 	 *
949 	 * Currently we don't know the generation for parent directory, so
950 	 * a generation of 0 means "accept any"
951 	 */
952 	inode = ext4_iget(sb, ino);
953 	if (IS_ERR(inode))
954 		return ERR_CAST(inode);
955 	if (generation && inode->i_generation != generation) {
956 		iput(inode);
957 		return ERR_PTR(-ESTALE);
958 	}
959 
960 	return inode;
961 }
962 
963 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
964 					int fh_len, int fh_type)
965 {
966 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
967 				    ext4_nfs_get_inode);
968 }
969 
970 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
971 					int fh_len, int fh_type)
972 {
973 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
974 				    ext4_nfs_get_inode);
975 }
976 
977 /*
978  * Try to release metadata pages (indirect blocks, directories) which are
979  * mapped via the block device.  Since these pages could have journal heads
980  * which would prevent try_to_free_buffers() from freeing them, we must use
981  * jbd2 layer's try_to_free_buffers() function to release them.
982  */
983 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
984 				 gfp_t wait)
985 {
986 	journal_t *journal = EXT4_SB(sb)->s_journal;
987 
988 	WARN_ON(PageChecked(page));
989 	if (!page_has_buffers(page))
990 		return 0;
991 	if (journal)
992 		return jbd2_journal_try_to_free_buffers(journal, page,
993 							wait & ~__GFP_WAIT);
994 	return try_to_free_buffers(page);
995 }
996 
997 #ifdef CONFIG_QUOTA
998 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
999 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1000 
1001 static int ext4_write_dquot(struct dquot *dquot);
1002 static int ext4_acquire_dquot(struct dquot *dquot);
1003 static int ext4_release_dquot(struct dquot *dquot);
1004 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1005 static int ext4_write_info(struct super_block *sb, int type);
1006 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1007 				char *path, int remount);
1008 static int ext4_quota_on_mount(struct super_block *sb, int type);
1009 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1010 			       size_t len, loff_t off);
1011 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1012 				const char *data, size_t len, loff_t off);
1013 
1014 static const struct dquot_operations ext4_quota_operations = {
1015 	.initialize	= dquot_initialize,
1016 	.drop		= dquot_drop,
1017 #ifdef CONFIG_QUOTA
1018 	.get_reserved_space = ext4_get_reserved_space,
1019 #endif
1020 	.write_dquot	= ext4_write_dquot,
1021 	.acquire_dquot	= ext4_acquire_dquot,
1022 	.release_dquot	= ext4_release_dquot,
1023 	.mark_dirty	= ext4_mark_dquot_dirty,
1024 	.write_info	= ext4_write_info,
1025 	.alloc_dquot	= dquot_alloc,
1026 	.destroy_dquot	= dquot_destroy,
1027 };
1028 
1029 static const struct quotactl_ops ext4_qctl_operations = {
1030 	.quota_on	= ext4_quota_on,
1031 	.quota_off	= vfs_quota_off,
1032 	.quota_sync	= vfs_quota_sync,
1033 	.get_info	= vfs_get_dqinfo,
1034 	.set_info	= vfs_set_dqinfo,
1035 	.get_dqblk	= vfs_get_dqblk,
1036 	.set_dqblk	= vfs_set_dqblk
1037 };
1038 #endif
1039 
1040 static const struct super_operations ext4_sops = {
1041 	.alloc_inode	= ext4_alloc_inode,
1042 	.destroy_inode	= ext4_destroy_inode,
1043 	.write_inode	= ext4_write_inode,
1044 	.dirty_inode	= ext4_dirty_inode,
1045 	.delete_inode	= ext4_delete_inode,
1046 	.put_super	= ext4_put_super,
1047 	.sync_fs	= ext4_sync_fs,
1048 	.freeze_fs	= ext4_freeze,
1049 	.unfreeze_fs	= ext4_unfreeze,
1050 	.statfs		= ext4_statfs,
1051 	.remount_fs	= ext4_remount,
1052 	.clear_inode	= ext4_clear_inode,
1053 	.show_options	= ext4_show_options,
1054 #ifdef CONFIG_QUOTA
1055 	.quota_read	= ext4_quota_read,
1056 	.quota_write	= ext4_quota_write,
1057 #endif
1058 	.bdev_try_to_free_page = bdev_try_to_free_page,
1059 };
1060 
1061 static const struct super_operations ext4_nojournal_sops = {
1062 	.alloc_inode	= ext4_alloc_inode,
1063 	.destroy_inode	= ext4_destroy_inode,
1064 	.write_inode	= ext4_write_inode,
1065 	.dirty_inode	= ext4_dirty_inode,
1066 	.delete_inode	= ext4_delete_inode,
1067 	.write_super	= ext4_write_super,
1068 	.put_super	= ext4_put_super,
1069 	.statfs		= ext4_statfs,
1070 	.remount_fs	= ext4_remount,
1071 	.clear_inode	= ext4_clear_inode,
1072 	.show_options	= ext4_show_options,
1073 #ifdef CONFIG_QUOTA
1074 	.quota_read	= ext4_quota_read,
1075 	.quota_write	= ext4_quota_write,
1076 #endif
1077 	.bdev_try_to_free_page = bdev_try_to_free_page,
1078 };
1079 
1080 static const struct export_operations ext4_export_ops = {
1081 	.fh_to_dentry = ext4_fh_to_dentry,
1082 	.fh_to_parent = ext4_fh_to_parent,
1083 	.get_parent = ext4_get_parent,
1084 };
1085 
1086 enum {
1087 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1088 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1089 	Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1090 	Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1091 	Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1092 	Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1093 	Opt_journal_update, Opt_journal_dev,
1094 	Opt_journal_checksum, Opt_journal_async_commit,
1095 	Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1096 	Opt_data_err_abort, Opt_data_err_ignore,
1097 	Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1098 	Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1099 	Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1100 	Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1101 	Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1102 	Opt_block_validity, Opt_noblock_validity,
1103 	Opt_inode_readahead_blks, Opt_journal_ioprio,
1104 	Opt_discard, Opt_nodiscard,
1105 };
1106 
1107 static const match_table_t tokens = {
1108 	{Opt_bsd_df, "bsddf"},
1109 	{Opt_minix_df, "minixdf"},
1110 	{Opt_grpid, "grpid"},
1111 	{Opt_grpid, "bsdgroups"},
1112 	{Opt_nogrpid, "nogrpid"},
1113 	{Opt_nogrpid, "sysvgroups"},
1114 	{Opt_resgid, "resgid=%u"},
1115 	{Opt_resuid, "resuid=%u"},
1116 	{Opt_sb, "sb=%u"},
1117 	{Opt_err_cont, "errors=continue"},
1118 	{Opt_err_panic, "errors=panic"},
1119 	{Opt_err_ro, "errors=remount-ro"},
1120 	{Opt_nouid32, "nouid32"},
1121 	{Opt_debug, "debug"},
1122 	{Opt_oldalloc, "oldalloc"},
1123 	{Opt_orlov, "orlov"},
1124 	{Opt_user_xattr, "user_xattr"},
1125 	{Opt_nouser_xattr, "nouser_xattr"},
1126 	{Opt_acl, "acl"},
1127 	{Opt_noacl, "noacl"},
1128 	{Opt_noload, "noload"},
1129 	{Opt_noload, "norecovery"},
1130 	{Opt_nobh, "nobh"},
1131 	{Opt_bh, "bh"},
1132 	{Opt_commit, "commit=%u"},
1133 	{Opt_min_batch_time, "min_batch_time=%u"},
1134 	{Opt_max_batch_time, "max_batch_time=%u"},
1135 	{Opt_journal_update, "journal=update"},
1136 	{Opt_journal_dev, "journal_dev=%u"},
1137 	{Opt_journal_checksum, "journal_checksum"},
1138 	{Opt_journal_async_commit, "journal_async_commit"},
1139 	{Opt_abort, "abort"},
1140 	{Opt_data_journal, "data=journal"},
1141 	{Opt_data_ordered, "data=ordered"},
1142 	{Opt_data_writeback, "data=writeback"},
1143 	{Opt_data_err_abort, "data_err=abort"},
1144 	{Opt_data_err_ignore, "data_err=ignore"},
1145 	{Opt_offusrjquota, "usrjquota="},
1146 	{Opt_usrjquota, "usrjquota=%s"},
1147 	{Opt_offgrpjquota, "grpjquota="},
1148 	{Opt_grpjquota, "grpjquota=%s"},
1149 	{Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1150 	{Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1151 	{Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1152 	{Opt_grpquota, "grpquota"},
1153 	{Opt_noquota, "noquota"},
1154 	{Opt_quota, "quota"},
1155 	{Opt_usrquota, "usrquota"},
1156 	{Opt_barrier, "barrier=%u"},
1157 	{Opt_barrier, "barrier"},
1158 	{Opt_nobarrier, "nobarrier"},
1159 	{Opt_i_version, "i_version"},
1160 	{Opt_stripe, "stripe=%u"},
1161 	{Opt_resize, "resize"},
1162 	{Opt_delalloc, "delalloc"},
1163 	{Opt_nodelalloc, "nodelalloc"},
1164 	{Opt_block_validity, "block_validity"},
1165 	{Opt_noblock_validity, "noblock_validity"},
1166 	{Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1167 	{Opt_journal_ioprio, "journal_ioprio=%u"},
1168 	{Opt_auto_da_alloc, "auto_da_alloc=%u"},
1169 	{Opt_auto_da_alloc, "auto_da_alloc"},
1170 	{Opt_noauto_da_alloc, "noauto_da_alloc"},
1171 	{Opt_discard, "discard"},
1172 	{Opt_nodiscard, "nodiscard"},
1173 	{Opt_err, NULL},
1174 };
1175 
1176 static ext4_fsblk_t get_sb_block(void **data)
1177 {
1178 	ext4_fsblk_t	sb_block;
1179 	char		*options = (char *) *data;
1180 
1181 	if (!options || strncmp(options, "sb=", 3) != 0)
1182 		return 1;	/* Default location */
1183 
1184 	options += 3;
1185 	/* TODO: use simple_strtoll with >32bit ext4 */
1186 	sb_block = simple_strtoul(options, &options, 0);
1187 	if (*options && *options != ',') {
1188 		printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1189 		       (char *) *data);
1190 		return 1;
1191 	}
1192 	if (*options == ',')
1193 		options++;
1194 	*data = (void *) options;
1195 
1196 	return sb_block;
1197 }
1198 
1199 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1200 
1201 static int parse_options(char *options, struct super_block *sb,
1202 			 unsigned long *journal_devnum,
1203 			 unsigned int *journal_ioprio,
1204 			 ext4_fsblk_t *n_blocks_count, int is_remount)
1205 {
1206 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1207 	char *p;
1208 	substring_t args[MAX_OPT_ARGS];
1209 	int data_opt = 0;
1210 	int option;
1211 #ifdef CONFIG_QUOTA
1212 	int qtype, qfmt;
1213 	char *qname;
1214 #endif
1215 
1216 	if (!options)
1217 		return 1;
1218 
1219 	while ((p = strsep(&options, ",")) != NULL) {
1220 		int token;
1221 		if (!*p)
1222 			continue;
1223 
1224 		token = match_token(p, tokens, args);
1225 		switch (token) {
1226 		case Opt_bsd_df:
1227 			clear_opt(sbi->s_mount_opt, MINIX_DF);
1228 			break;
1229 		case Opt_minix_df:
1230 			set_opt(sbi->s_mount_opt, MINIX_DF);
1231 			break;
1232 		case Opt_grpid:
1233 			set_opt(sbi->s_mount_opt, GRPID);
1234 			break;
1235 		case Opt_nogrpid:
1236 			clear_opt(sbi->s_mount_opt, GRPID);
1237 			break;
1238 		case Opt_resuid:
1239 			if (match_int(&args[0], &option))
1240 				return 0;
1241 			sbi->s_resuid = option;
1242 			break;
1243 		case Opt_resgid:
1244 			if (match_int(&args[0], &option))
1245 				return 0;
1246 			sbi->s_resgid = option;
1247 			break;
1248 		case Opt_sb:
1249 			/* handled by get_sb_block() instead of here */
1250 			/* *sb_block = match_int(&args[0]); */
1251 			break;
1252 		case Opt_err_panic:
1253 			clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1254 			clear_opt(sbi->s_mount_opt, ERRORS_RO);
1255 			set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1256 			break;
1257 		case Opt_err_ro:
1258 			clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1259 			clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1260 			set_opt(sbi->s_mount_opt, ERRORS_RO);
1261 			break;
1262 		case Opt_err_cont:
1263 			clear_opt(sbi->s_mount_opt, ERRORS_RO);
1264 			clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1265 			set_opt(sbi->s_mount_opt, ERRORS_CONT);
1266 			break;
1267 		case Opt_nouid32:
1268 			set_opt(sbi->s_mount_opt, NO_UID32);
1269 			break;
1270 		case Opt_debug:
1271 			set_opt(sbi->s_mount_opt, DEBUG);
1272 			break;
1273 		case Opt_oldalloc:
1274 			set_opt(sbi->s_mount_opt, OLDALLOC);
1275 			break;
1276 		case Opt_orlov:
1277 			clear_opt(sbi->s_mount_opt, OLDALLOC);
1278 			break;
1279 #ifdef CONFIG_EXT4_FS_XATTR
1280 		case Opt_user_xattr:
1281 			set_opt(sbi->s_mount_opt, XATTR_USER);
1282 			break;
1283 		case Opt_nouser_xattr:
1284 			clear_opt(sbi->s_mount_opt, XATTR_USER);
1285 			break;
1286 #else
1287 		case Opt_user_xattr:
1288 		case Opt_nouser_xattr:
1289 			ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1290 			break;
1291 #endif
1292 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1293 		case Opt_acl:
1294 			set_opt(sbi->s_mount_opt, POSIX_ACL);
1295 			break;
1296 		case Opt_noacl:
1297 			clear_opt(sbi->s_mount_opt, POSIX_ACL);
1298 			break;
1299 #else
1300 		case Opt_acl:
1301 		case Opt_noacl:
1302 			ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1303 			break;
1304 #endif
1305 		case Opt_journal_update:
1306 			/* @@@ FIXME */
1307 			/* Eventually we will want to be able to create
1308 			   a journal file here.  For now, only allow the
1309 			   user to specify an existing inode to be the
1310 			   journal file. */
1311 			if (is_remount) {
1312 				ext4_msg(sb, KERN_ERR,
1313 					 "Cannot specify journal on remount");
1314 				return 0;
1315 			}
1316 			set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1317 			break;
1318 		case Opt_journal_dev:
1319 			if (is_remount) {
1320 				ext4_msg(sb, KERN_ERR,
1321 					"Cannot specify journal on remount");
1322 				return 0;
1323 			}
1324 			if (match_int(&args[0], &option))
1325 				return 0;
1326 			*journal_devnum = option;
1327 			break;
1328 		case Opt_journal_checksum:
1329 			set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1330 			break;
1331 		case Opt_journal_async_commit:
1332 			set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1333 			set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1334 			break;
1335 		case Opt_noload:
1336 			set_opt(sbi->s_mount_opt, NOLOAD);
1337 			break;
1338 		case Opt_commit:
1339 			if (match_int(&args[0], &option))
1340 				return 0;
1341 			if (option < 0)
1342 				return 0;
1343 			if (option == 0)
1344 				option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1345 			sbi->s_commit_interval = HZ * option;
1346 			break;
1347 		case Opt_max_batch_time:
1348 			if (match_int(&args[0], &option))
1349 				return 0;
1350 			if (option < 0)
1351 				return 0;
1352 			if (option == 0)
1353 				option = EXT4_DEF_MAX_BATCH_TIME;
1354 			sbi->s_max_batch_time = option;
1355 			break;
1356 		case Opt_min_batch_time:
1357 			if (match_int(&args[0], &option))
1358 				return 0;
1359 			if (option < 0)
1360 				return 0;
1361 			sbi->s_min_batch_time = option;
1362 			break;
1363 		case Opt_data_journal:
1364 			data_opt = EXT4_MOUNT_JOURNAL_DATA;
1365 			goto datacheck;
1366 		case Opt_data_ordered:
1367 			data_opt = EXT4_MOUNT_ORDERED_DATA;
1368 			goto datacheck;
1369 		case Opt_data_writeback:
1370 			data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1371 		datacheck:
1372 			if (is_remount) {
1373 				if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1374 						!= data_opt) {
1375 					ext4_msg(sb, KERN_ERR,
1376 						"Cannot change data mode on remount");
1377 					return 0;
1378 				}
1379 			} else {
1380 				sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1381 				sbi->s_mount_opt |= data_opt;
1382 			}
1383 			break;
1384 		case Opt_data_err_abort:
1385 			set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1386 			break;
1387 		case Opt_data_err_ignore:
1388 			clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1389 			break;
1390 #ifdef CONFIG_QUOTA
1391 		case Opt_usrjquota:
1392 			qtype = USRQUOTA;
1393 			goto set_qf_name;
1394 		case Opt_grpjquota:
1395 			qtype = GRPQUOTA;
1396 set_qf_name:
1397 			if (sb_any_quota_loaded(sb) &&
1398 			    !sbi->s_qf_names[qtype]) {
1399 				ext4_msg(sb, KERN_ERR,
1400 				       "Cannot change journaled "
1401 				       "quota options when quota turned on");
1402 				return 0;
1403 			}
1404 			qname = match_strdup(&args[0]);
1405 			if (!qname) {
1406 				ext4_msg(sb, KERN_ERR,
1407 					"Not enough memory for "
1408 					"storing quotafile name");
1409 				return 0;
1410 			}
1411 			if (sbi->s_qf_names[qtype] &&
1412 			    strcmp(sbi->s_qf_names[qtype], qname)) {
1413 				ext4_msg(sb, KERN_ERR,
1414 					"%s quota file already "
1415 					"specified", QTYPE2NAME(qtype));
1416 				kfree(qname);
1417 				return 0;
1418 			}
1419 			sbi->s_qf_names[qtype] = qname;
1420 			if (strchr(sbi->s_qf_names[qtype], '/')) {
1421 				ext4_msg(sb, KERN_ERR,
1422 					"quotafile must be on "
1423 					"filesystem root");
1424 				kfree(sbi->s_qf_names[qtype]);
1425 				sbi->s_qf_names[qtype] = NULL;
1426 				return 0;
1427 			}
1428 			set_opt(sbi->s_mount_opt, QUOTA);
1429 			break;
1430 		case Opt_offusrjquota:
1431 			qtype = USRQUOTA;
1432 			goto clear_qf_name;
1433 		case Opt_offgrpjquota:
1434 			qtype = GRPQUOTA;
1435 clear_qf_name:
1436 			if (sb_any_quota_loaded(sb) &&
1437 			    sbi->s_qf_names[qtype]) {
1438 				ext4_msg(sb, KERN_ERR, "Cannot change "
1439 					"journaled quota options when "
1440 					"quota turned on");
1441 				return 0;
1442 			}
1443 			/*
1444 			 * The space will be released later when all options
1445 			 * are confirmed to be correct
1446 			 */
1447 			sbi->s_qf_names[qtype] = NULL;
1448 			break;
1449 		case Opt_jqfmt_vfsold:
1450 			qfmt = QFMT_VFS_OLD;
1451 			goto set_qf_format;
1452 		case Opt_jqfmt_vfsv0:
1453 			qfmt = QFMT_VFS_V0;
1454 			goto set_qf_format;
1455 		case Opt_jqfmt_vfsv1:
1456 			qfmt = QFMT_VFS_V1;
1457 set_qf_format:
1458 			if (sb_any_quota_loaded(sb) &&
1459 			    sbi->s_jquota_fmt != qfmt) {
1460 				ext4_msg(sb, KERN_ERR, "Cannot change "
1461 					"journaled quota options when "
1462 					"quota turned on");
1463 				return 0;
1464 			}
1465 			sbi->s_jquota_fmt = qfmt;
1466 			break;
1467 		case Opt_quota:
1468 		case Opt_usrquota:
1469 			set_opt(sbi->s_mount_opt, QUOTA);
1470 			set_opt(sbi->s_mount_opt, USRQUOTA);
1471 			break;
1472 		case Opt_grpquota:
1473 			set_opt(sbi->s_mount_opt, QUOTA);
1474 			set_opt(sbi->s_mount_opt, GRPQUOTA);
1475 			break;
1476 		case Opt_noquota:
1477 			if (sb_any_quota_loaded(sb)) {
1478 				ext4_msg(sb, KERN_ERR, "Cannot change quota "
1479 					"options when quota turned on");
1480 				return 0;
1481 			}
1482 			clear_opt(sbi->s_mount_opt, QUOTA);
1483 			clear_opt(sbi->s_mount_opt, USRQUOTA);
1484 			clear_opt(sbi->s_mount_opt, GRPQUOTA);
1485 			break;
1486 #else
1487 		case Opt_quota:
1488 		case Opt_usrquota:
1489 		case Opt_grpquota:
1490 			ext4_msg(sb, KERN_ERR,
1491 				"quota options not supported");
1492 			break;
1493 		case Opt_usrjquota:
1494 		case Opt_grpjquota:
1495 		case Opt_offusrjquota:
1496 		case Opt_offgrpjquota:
1497 		case Opt_jqfmt_vfsold:
1498 		case Opt_jqfmt_vfsv0:
1499 		case Opt_jqfmt_vfsv1:
1500 			ext4_msg(sb, KERN_ERR,
1501 				"journaled quota options not supported");
1502 			break;
1503 		case Opt_noquota:
1504 			break;
1505 #endif
1506 		case Opt_abort:
1507 			sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1508 			break;
1509 		case Opt_nobarrier:
1510 			clear_opt(sbi->s_mount_opt, BARRIER);
1511 			break;
1512 		case Opt_barrier:
1513 			if (match_int(&args[0], &option)) {
1514 				set_opt(sbi->s_mount_opt, BARRIER);
1515 				break;
1516 			}
1517 			if (option)
1518 				set_opt(sbi->s_mount_opt, BARRIER);
1519 			else
1520 				clear_opt(sbi->s_mount_opt, BARRIER);
1521 			break;
1522 		case Opt_ignore:
1523 			break;
1524 		case Opt_resize:
1525 			if (!is_remount) {
1526 				ext4_msg(sb, KERN_ERR,
1527 					"resize option only available "
1528 					"for remount");
1529 				return 0;
1530 			}
1531 			if (match_int(&args[0], &option) != 0)
1532 				return 0;
1533 			*n_blocks_count = option;
1534 			break;
1535 		case Opt_nobh:
1536 			set_opt(sbi->s_mount_opt, NOBH);
1537 			break;
1538 		case Opt_bh:
1539 			clear_opt(sbi->s_mount_opt, NOBH);
1540 			break;
1541 		case Opt_i_version:
1542 			set_opt(sbi->s_mount_opt, I_VERSION);
1543 			sb->s_flags |= MS_I_VERSION;
1544 			break;
1545 		case Opt_nodelalloc:
1546 			clear_opt(sbi->s_mount_opt, DELALLOC);
1547 			break;
1548 		case Opt_stripe:
1549 			if (match_int(&args[0], &option))
1550 				return 0;
1551 			if (option < 0)
1552 				return 0;
1553 			sbi->s_stripe = option;
1554 			break;
1555 		case Opt_delalloc:
1556 			set_opt(sbi->s_mount_opt, DELALLOC);
1557 			break;
1558 		case Opt_block_validity:
1559 			set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1560 			break;
1561 		case Opt_noblock_validity:
1562 			clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1563 			break;
1564 		case Opt_inode_readahead_blks:
1565 			if (match_int(&args[0], &option))
1566 				return 0;
1567 			if (option < 0 || option > (1 << 30))
1568 				return 0;
1569 			if (!is_power_of_2(option)) {
1570 				ext4_msg(sb, KERN_ERR,
1571 					 "EXT4-fs: inode_readahead_blks"
1572 					 " must be a power of 2");
1573 				return 0;
1574 			}
1575 			sbi->s_inode_readahead_blks = option;
1576 			break;
1577 		case Opt_journal_ioprio:
1578 			if (match_int(&args[0], &option))
1579 				return 0;
1580 			if (option < 0 || option > 7)
1581 				break;
1582 			*journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1583 							    option);
1584 			break;
1585 		case Opt_noauto_da_alloc:
1586 			set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1587 			break;
1588 		case Opt_auto_da_alloc:
1589 			if (match_int(&args[0], &option)) {
1590 				clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1591 				break;
1592 			}
1593 			if (option)
1594 				clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1595 			else
1596 				set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1597 			break;
1598 		case Opt_discard:
1599 			set_opt(sbi->s_mount_opt, DISCARD);
1600 			break;
1601 		case Opt_nodiscard:
1602 			clear_opt(sbi->s_mount_opt, DISCARD);
1603 			break;
1604 		default:
1605 			ext4_msg(sb, KERN_ERR,
1606 			       "Unrecognized mount option \"%s\" "
1607 			       "or missing value", p);
1608 			return 0;
1609 		}
1610 	}
1611 #ifdef CONFIG_QUOTA
1612 	if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1613 		if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1614 		     sbi->s_qf_names[USRQUOTA])
1615 			clear_opt(sbi->s_mount_opt, USRQUOTA);
1616 
1617 		if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1618 		     sbi->s_qf_names[GRPQUOTA])
1619 			clear_opt(sbi->s_mount_opt, GRPQUOTA);
1620 
1621 		if ((sbi->s_qf_names[USRQUOTA] &&
1622 				(sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1623 		    (sbi->s_qf_names[GRPQUOTA] &&
1624 				(sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1625 			ext4_msg(sb, KERN_ERR, "old and new quota "
1626 					"format mixing");
1627 			return 0;
1628 		}
1629 
1630 		if (!sbi->s_jquota_fmt) {
1631 			ext4_msg(sb, KERN_ERR, "journaled quota format "
1632 					"not specified");
1633 			return 0;
1634 		}
1635 	} else {
1636 		if (sbi->s_jquota_fmt) {
1637 			ext4_msg(sb, KERN_ERR, "journaled quota format "
1638 					"specified with no journaling "
1639 					"enabled");
1640 			return 0;
1641 		}
1642 	}
1643 #endif
1644 	return 1;
1645 }
1646 
1647 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1648 			    int read_only)
1649 {
1650 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1651 	int res = 0;
1652 
1653 	if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1654 		ext4_msg(sb, KERN_ERR, "revision level too high, "
1655 			 "forcing read-only mode");
1656 		res = MS_RDONLY;
1657 	}
1658 	if (read_only)
1659 		return res;
1660 	if (!(sbi->s_mount_state & EXT4_VALID_FS))
1661 		ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1662 			 "running e2fsck is recommended");
1663 	else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1664 		ext4_msg(sb, KERN_WARNING,
1665 			 "warning: mounting fs with errors, "
1666 			 "running e2fsck is recommended");
1667 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1668 		 le16_to_cpu(es->s_mnt_count) >=
1669 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1670 		ext4_msg(sb, KERN_WARNING,
1671 			 "warning: maximal mount count reached, "
1672 			 "running e2fsck is recommended");
1673 	else if (le32_to_cpu(es->s_checkinterval) &&
1674 		(le32_to_cpu(es->s_lastcheck) +
1675 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1676 		ext4_msg(sb, KERN_WARNING,
1677 			 "warning: checktime reached, "
1678 			 "running e2fsck is recommended");
1679 	if (!sbi->s_journal)
1680 		es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1681 	if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1682 		es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1683 	le16_add_cpu(&es->s_mnt_count, 1);
1684 	es->s_mtime = cpu_to_le32(get_seconds());
1685 	ext4_update_dynamic_rev(sb);
1686 	if (sbi->s_journal)
1687 		EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1688 
1689 	ext4_commit_super(sb, 1);
1690 	if (test_opt(sb, DEBUG))
1691 		printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1692 				"bpg=%lu, ipg=%lu, mo=%04x]\n",
1693 			sb->s_blocksize,
1694 			sbi->s_groups_count,
1695 			EXT4_BLOCKS_PER_GROUP(sb),
1696 			EXT4_INODES_PER_GROUP(sb),
1697 			sbi->s_mount_opt);
1698 
1699 	return res;
1700 }
1701 
1702 static int ext4_fill_flex_info(struct super_block *sb)
1703 {
1704 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1705 	struct ext4_group_desc *gdp = NULL;
1706 	ext4_group_t flex_group_count;
1707 	ext4_group_t flex_group;
1708 	int groups_per_flex = 0;
1709 	size_t size;
1710 	int i;
1711 
1712 	sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1713 	groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1714 
1715 	if (groups_per_flex < 2) {
1716 		sbi->s_log_groups_per_flex = 0;
1717 		return 1;
1718 	}
1719 
1720 	/* We allocate both existing and potentially added groups */
1721 	flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1722 			((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1723 			      EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1724 	size = flex_group_count * sizeof(struct flex_groups);
1725 	sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1726 	if (sbi->s_flex_groups == NULL) {
1727 		sbi->s_flex_groups = vmalloc(size);
1728 		if (sbi->s_flex_groups)
1729 			memset(sbi->s_flex_groups, 0, size);
1730 	}
1731 	if (sbi->s_flex_groups == NULL) {
1732 		ext4_msg(sb, KERN_ERR, "not enough memory for "
1733 				"%u flex groups", flex_group_count);
1734 		goto failed;
1735 	}
1736 
1737 	for (i = 0; i < sbi->s_groups_count; i++) {
1738 		gdp = ext4_get_group_desc(sb, i, NULL);
1739 
1740 		flex_group = ext4_flex_group(sbi, i);
1741 		atomic_add(ext4_free_inodes_count(sb, gdp),
1742 			   &sbi->s_flex_groups[flex_group].free_inodes);
1743 		atomic_add(ext4_free_blks_count(sb, gdp),
1744 			   &sbi->s_flex_groups[flex_group].free_blocks);
1745 		atomic_add(ext4_used_dirs_count(sb, gdp),
1746 			   &sbi->s_flex_groups[flex_group].used_dirs);
1747 	}
1748 
1749 	return 1;
1750 failed:
1751 	return 0;
1752 }
1753 
1754 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1755 			    struct ext4_group_desc *gdp)
1756 {
1757 	__u16 crc = 0;
1758 
1759 	if (sbi->s_es->s_feature_ro_compat &
1760 	    cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1761 		int offset = offsetof(struct ext4_group_desc, bg_checksum);
1762 		__le32 le_group = cpu_to_le32(block_group);
1763 
1764 		crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1765 		crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1766 		crc = crc16(crc, (__u8 *)gdp, offset);
1767 		offset += sizeof(gdp->bg_checksum); /* skip checksum */
1768 		/* for checksum of struct ext4_group_desc do the rest...*/
1769 		if ((sbi->s_es->s_feature_incompat &
1770 		     cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1771 		    offset < le16_to_cpu(sbi->s_es->s_desc_size))
1772 			crc = crc16(crc, (__u8 *)gdp + offset,
1773 				    le16_to_cpu(sbi->s_es->s_desc_size) -
1774 					offset);
1775 	}
1776 
1777 	return cpu_to_le16(crc);
1778 }
1779 
1780 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1781 				struct ext4_group_desc *gdp)
1782 {
1783 	if ((sbi->s_es->s_feature_ro_compat &
1784 	     cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1785 	    (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1786 		return 0;
1787 
1788 	return 1;
1789 }
1790 
1791 /* Called at mount-time, super-block is locked */
1792 static int ext4_check_descriptors(struct super_block *sb)
1793 {
1794 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1795 	ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1796 	ext4_fsblk_t last_block;
1797 	ext4_fsblk_t block_bitmap;
1798 	ext4_fsblk_t inode_bitmap;
1799 	ext4_fsblk_t inode_table;
1800 	int flexbg_flag = 0;
1801 	ext4_group_t i;
1802 
1803 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1804 		flexbg_flag = 1;
1805 
1806 	ext4_debug("Checking group descriptors");
1807 
1808 	for (i = 0; i < sbi->s_groups_count; i++) {
1809 		struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1810 
1811 		if (i == sbi->s_groups_count - 1 || flexbg_flag)
1812 			last_block = ext4_blocks_count(sbi->s_es) - 1;
1813 		else
1814 			last_block = first_block +
1815 				(EXT4_BLOCKS_PER_GROUP(sb) - 1);
1816 
1817 		block_bitmap = ext4_block_bitmap(sb, gdp);
1818 		if (block_bitmap < first_block || block_bitmap > last_block) {
1819 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1820 			       "Block bitmap for group %u not in group "
1821 			       "(block %llu)!", i, block_bitmap);
1822 			return 0;
1823 		}
1824 		inode_bitmap = ext4_inode_bitmap(sb, gdp);
1825 		if (inode_bitmap < first_block || inode_bitmap > last_block) {
1826 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1827 			       "Inode bitmap for group %u not in group "
1828 			       "(block %llu)!", i, inode_bitmap);
1829 			return 0;
1830 		}
1831 		inode_table = ext4_inode_table(sb, gdp);
1832 		if (inode_table < first_block ||
1833 		    inode_table + sbi->s_itb_per_group - 1 > last_block) {
1834 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1835 			       "Inode table for group %u not in group "
1836 			       "(block %llu)!", i, inode_table);
1837 			return 0;
1838 		}
1839 		ext4_lock_group(sb, i);
1840 		if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1841 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1842 				 "Checksum for group %u failed (%u!=%u)",
1843 				 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1844 				     gdp)), le16_to_cpu(gdp->bg_checksum));
1845 			if (!(sb->s_flags & MS_RDONLY)) {
1846 				ext4_unlock_group(sb, i);
1847 				return 0;
1848 			}
1849 		}
1850 		ext4_unlock_group(sb, i);
1851 		if (!flexbg_flag)
1852 			first_block += EXT4_BLOCKS_PER_GROUP(sb);
1853 	}
1854 
1855 	ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1856 	sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
1857 	return 1;
1858 }
1859 
1860 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1861  * the superblock) which were deleted from all directories, but held open by
1862  * a process at the time of a crash.  We walk the list and try to delete these
1863  * inodes at recovery time (only with a read-write filesystem).
1864  *
1865  * In order to keep the orphan inode chain consistent during traversal (in
1866  * case of crash during recovery), we link each inode into the superblock
1867  * orphan list_head and handle it the same way as an inode deletion during
1868  * normal operation (which journals the operations for us).
1869  *
1870  * We only do an iget() and an iput() on each inode, which is very safe if we
1871  * accidentally point at an in-use or already deleted inode.  The worst that
1872  * can happen in this case is that we get a "bit already cleared" message from
1873  * ext4_free_inode().  The only reason we would point at a wrong inode is if
1874  * e2fsck was run on this filesystem, and it must have already done the orphan
1875  * inode cleanup for us, so we can safely abort without any further action.
1876  */
1877 static void ext4_orphan_cleanup(struct super_block *sb,
1878 				struct ext4_super_block *es)
1879 {
1880 	unsigned int s_flags = sb->s_flags;
1881 	int nr_orphans = 0, nr_truncates = 0;
1882 #ifdef CONFIG_QUOTA
1883 	int i;
1884 #endif
1885 	if (!es->s_last_orphan) {
1886 		jbd_debug(4, "no orphan inodes to clean up\n");
1887 		return;
1888 	}
1889 
1890 	if (bdev_read_only(sb->s_bdev)) {
1891 		ext4_msg(sb, KERN_ERR, "write access "
1892 			"unavailable, skipping orphan cleanup");
1893 		return;
1894 	}
1895 
1896 	if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1897 		if (es->s_last_orphan)
1898 			jbd_debug(1, "Errors on filesystem, "
1899 				  "clearing orphan list.\n");
1900 		es->s_last_orphan = 0;
1901 		jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1902 		return;
1903 	}
1904 
1905 	if (s_flags & MS_RDONLY) {
1906 		ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1907 		sb->s_flags &= ~MS_RDONLY;
1908 	}
1909 #ifdef CONFIG_QUOTA
1910 	/* Needed for iput() to work correctly and not trash data */
1911 	sb->s_flags |= MS_ACTIVE;
1912 	/* Turn on quotas so that they are updated correctly */
1913 	for (i = 0; i < MAXQUOTAS; i++) {
1914 		if (EXT4_SB(sb)->s_qf_names[i]) {
1915 			int ret = ext4_quota_on_mount(sb, i);
1916 			if (ret < 0)
1917 				ext4_msg(sb, KERN_ERR,
1918 					"Cannot turn on journaled "
1919 					"quota: error %d", ret);
1920 		}
1921 	}
1922 #endif
1923 
1924 	while (es->s_last_orphan) {
1925 		struct inode *inode;
1926 
1927 		inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1928 		if (IS_ERR(inode)) {
1929 			es->s_last_orphan = 0;
1930 			break;
1931 		}
1932 
1933 		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1934 		vfs_dq_init(inode);
1935 		if (inode->i_nlink) {
1936 			ext4_msg(sb, KERN_DEBUG,
1937 				"%s: truncating inode %lu to %lld bytes",
1938 				__func__, inode->i_ino, inode->i_size);
1939 			jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1940 				  inode->i_ino, inode->i_size);
1941 			ext4_truncate(inode);
1942 			nr_truncates++;
1943 		} else {
1944 			ext4_msg(sb, KERN_DEBUG,
1945 				"%s: deleting unreferenced inode %lu",
1946 				__func__, inode->i_ino);
1947 			jbd_debug(2, "deleting unreferenced inode %lu\n",
1948 				  inode->i_ino);
1949 			nr_orphans++;
1950 		}
1951 		iput(inode);  /* The delete magic happens here! */
1952 	}
1953 
1954 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1955 
1956 	if (nr_orphans)
1957 		ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1958 		       PLURAL(nr_orphans));
1959 	if (nr_truncates)
1960 		ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1961 		       PLURAL(nr_truncates));
1962 #ifdef CONFIG_QUOTA
1963 	/* Turn quotas off */
1964 	for (i = 0; i < MAXQUOTAS; i++) {
1965 		if (sb_dqopt(sb)->files[i])
1966 			vfs_quota_off(sb, i, 0);
1967 	}
1968 #endif
1969 	sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1970 }
1971 
1972 /*
1973  * Maximal extent format file size.
1974  * Resulting logical blkno at s_maxbytes must fit in our on-disk
1975  * extent format containers, within a sector_t, and within i_blocks
1976  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
1977  * so that won't be a limiting factor.
1978  *
1979  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1980  */
1981 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1982 {
1983 	loff_t res;
1984 	loff_t upper_limit = MAX_LFS_FILESIZE;
1985 
1986 	/* small i_blocks in vfs inode? */
1987 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1988 		/*
1989 		 * CONFIG_LBDAF is not enabled implies the inode
1990 		 * i_block represent total blocks in 512 bytes
1991 		 * 32 == size of vfs inode i_blocks * 8
1992 		 */
1993 		upper_limit = (1LL << 32) - 1;
1994 
1995 		/* total blocks in file system block size */
1996 		upper_limit >>= (blkbits - 9);
1997 		upper_limit <<= blkbits;
1998 	}
1999 
2000 	/* 32-bit extent-start container, ee_block */
2001 	res = 1LL << 32;
2002 	res <<= blkbits;
2003 	res -= 1;
2004 
2005 	/* Sanity check against vm- & vfs- imposed limits */
2006 	if (res > upper_limit)
2007 		res = upper_limit;
2008 
2009 	return res;
2010 }
2011 
2012 /*
2013  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
2014  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2015  * We need to be 1 filesystem block less than the 2^48 sector limit.
2016  */
2017 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2018 {
2019 	loff_t res = EXT4_NDIR_BLOCKS;
2020 	int meta_blocks;
2021 	loff_t upper_limit;
2022 	/* This is calculated to be the largest file size for a dense, block
2023 	 * mapped file such that the file's total number of 512-byte sectors,
2024 	 * including data and all indirect blocks, does not exceed (2^48 - 1).
2025 	 *
2026 	 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2027 	 * number of 512-byte sectors of the file.
2028 	 */
2029 
2030 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2031 		/*
2032 		 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2033 		 * the inode i_block field represents total file blocks in
2034 		 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2035 		 */
2036 		upper_limit = (1LL << 32) - 1;
2037 
2038 		/* total blocks in file system block size */
2039 		upper_limit >>= (bits - 9);
2040 
2041 	} else {
2042 		/*
2043 		 * We use 48 bit ext4_inode i_blocks
2044 		 * With EXT4_HUGE_FILE_FL set the i_blocks
2045 		 * represent total number of blocks in
2046 		 * file system block size
2047 		 */
2048 		upper_limit = (1LL << 48) - 1;
2049 
2050 	}
2051 
2052 	/* indirect blocks */
2053 	meta_blocks = 1;
2054 	/* double indirect blocks */
2055 	meta_blocks += 1 + (1LL << (bits-2));
2056 	/* tripple indirect blocks */
2057 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2058 
2059 	upper_limit -= meta_blocks;
2060 	upper_limit <<= bits;
2061 
2062 	res += 1LL << (bits-2);
2063 	res += 1LL << (2*(bits-2));
2064 	res += 1LL << (3*(bits-2));
2065 	res <<= bits;
2066 	if (res > upper_limit)
2067 		res = upper_limit;
2068 
2069 	if (res > MAX_LFS_FILESIZE)
2070 		res = MAX_LFS_FILESIZE;
2071 
2072 	return res;
2073 }
2074 
2075 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2076 				   ext4_fsblk_t logical_sb_block, int nr)
2077 {
2078 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2079 	ext4_group_t bg, first_meta_bg;
2080 	int has_super = 0;
2081 
2082 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2083 
2084 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2085 	    nr < first_meta_bg)
2086 		return logical_sb_block + nr + 1;
2087 	bg = sbi->s_desc_per_block * nr;
2088 	if (ext4_bg_has_super(sb, bg))
2089 		has_super = 1;
2090 
2091 	return (has_super + ext4_group_first_block_no(sb, bg));
2092 }
2093 
2094 /**
2095  * ext4_get_stripe_size: Get the stripe size.
2096  * @sbi: In memory super block info
2097  *
2098  * If we have specified it via mount option, then
2099  * use the mount option value. If the value specified at mount time is
2100  * greater than the blocks per group use the super block value.
2101  * If the super block value is greater than blocks per group return 0.
2102  * Allocator needs it be less than blocks per group.
2103  *
2104  */
2105 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2106 {
2107 	unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2108 	unsigned long stripe_width =
2109 			le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2110 
2111 	if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2112 		return sbi->s_stripe;
2113 
2114 	if (stripe_width <= sbi->s_blocks_per_group)
2115 		return stripe_width;
2116 
2117 	if (stride <= sbi->s_blocks_per_group)
2118 		return stride;
2119 
2120 	return 0;
2121 }
2122 
2123 /* sysfs supprt */
2124 
2125 struct ext4_attr {
2126 	struct attribute attr;
2127 	ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2128 	ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2129 			 const char *, size_t);
2130 	int offset;
2131 };
2132 
2133 static int parse_strtoul(const char *buf,
2134 		unsigned long max, unsigned long *value)
2135 {
2136 	char *endp;
2137 
2138 	*value = simple_strtoul(skip_spaces(buf), &endp, 0);
2139 	endp = skip_spaces(endp);
2140 	if (*endp || *value > max)
2141 		return -EINVAL;
2142 
2143 	return 0;
2144 }
2145 
2146 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2147 					      struct ext4_sb_info *sbi,
2148 					      char *buf)
2149 {
2150 	return snprintf(buf, PAGE_SIZE, "%llu\n",
2151 			(s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2152 }
2153 
2154 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2155 					 struct ext4_sb_info *sbi, char *buf)
2156 {
2157 	struct super_block *sb = sbi->s_buddy_cache->i_sb;
2158 
2159 	return snprintf(buf, PAGE_SIZE, "%lu\n",
2160 			(part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2161 			 sbi->s_sectors_written_start) >> 1);
2162 }
2163 
2164 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2165 					  struct ext4_sb_info *sbi, char *buf)
2166 {
2167 	struct super_block *sb = sbi->s_buddy_cache->i_sb;
2168 
2169 	return snprintf(buf, PAGE_SIZE, "%llu\n",
2170 			(unsigned long long)(sbi->s_kbytes_written +
2171 			((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2172 			  EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2173 }
2174 
2175 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2176 					  struct ext4_sb_info *sbi,
2177 					  const char *buf, size_t count)
2178 {
2179 	unsigned long t;
2180 
2181 	if (parse_strtoul(buf, 0x40000000, &t))
2182 		return -EINVAL;
2183 
2184 	if (!is_power_of_2(t))
2185 		return -EINVAL;
2186 
2187 	sbi->s_inode_readahead_blks = t;
2188 	return count;
2189 }
2190 
2191 static ssize_t sbi_ui_show(struct ext4_attr *a,
2192 			   struct ext4_sb_info *sbi, char *buf)
2193 {
2194 	unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2195 
2196 	return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2197 }
2198 
2199 static ssize_t sbi_ui_store(struct ext4_attr *a,
2200 			    struct ext4_sb_info *sbi,
2201 			    const char *buf, size_t count)
2202 {
2203 	unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2204 	unsigned long t;
2205 
2206 	if (parse_strtoul(buf, 0xffffffff, &t))
2207 		return -EINVAL;
2208 	*ui = t;
2209 	return count;
2210 }
2211 
2212 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2213 static struct ext4_attr ext4_attr_##_name = {			\
2214 	.attr = {.name = __stringify(_name), .mode = _mode },	\
2215 	.show	= _show,					\
2216 	.store	= _store,					\
2217 	.offset = offsetof(struct ext4_sb_info, _elname),	\
2218 }
2219 #define EXT4_ATTR(name, mode, show, store) \
2220 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2221 
2222 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2223 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2224 #define EXT4_RW_ATTR_SBI_UI(name, elname)	\
2225 	EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2226 #define ATTR_LIST(name) &ext4_attr_##name.attr
2227 
2228 EXT4_RO_ATTR(delayed_allocation_blocks);
2229 EXT4_RO_ATTR(session_write_kbytes);
2230 EXT4_RO_ATTR(lifetime_write_kbytes);
2231 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2232 		 inode_readahead_blks_store, s_inode_readahead_blks);
2233 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2234 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2235 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2236 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2237 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2238 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2239 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2240 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2241 
2242 static struct attribute *ext4_attrs[] = {
2243 	ATTR_LIST(delayed_allocation_blocks),
2244 	ATTR_LIST(session_write_kbytes),
2245 	ATTR_LIST(lifetime_write_kbytes),
2246 	ATTR_LIST(inode_readahead_blks),
2247 	ATTR_LIST(inode_goal),
2248 	ATTR_LIST(mb_stats),
2249 	ATTR_LIST(mb_max_to_scan),
2250 	ATTR_LIST(mb_min_to_scan),
2251 	ATTR_LIST(mb_order2_req),
2252 	ATTR_LIST(mb_stream_req),
2253 	ATTR_LIST(mb_group_prealloc),
2254 	ATTR_LIST(max_writeback_mb_bump),
2255 	NULL,
2256 };
2257 
2258 static ssize_t ext4_attr_show(struct kobject *kobj,
2259 			      struct attribute *attr, char *buf)
2260 {
2261 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2262 						s_kobj);
2263 	struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2264 
2265 	return a->show ? a->show(a, sbi, buf) : 0;
2266 }
2267 
2268 static ssize_t ext4_attr_store(struct kobject *kobj,
2269 			       struct attribute *attr,
2270 			       const char *buf, size_t len)
2271 {
2272 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2273 						s_kobj);
2274 	struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2275 
2276 	return a->store ? a->store(a, sbi, buf, len) : 0;
2277 }
2278 
2279 static void ext4_sb_release(struct kobject *kobj)
2280 {
2281 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2282 						s_kobj);
2283 	complete(&sbi->s_kobj_unregister);
2284 }
2285 
2286 
2287 static struct sysfs_ops ext4_attr_ops = {
2288 	.show	= ext4_attr_show,
2289 	.store	= ext4_attr_store,
2290 };
2291 
2292 static struct kobj_type ext4_ktype = {
2293 	.default_attrs	= ext4_attrs,
2294 	.sysfs_ops	= &ext4_attr_ops,
2295 	.release	= ext4_sb_release,
2296 };
2297 
2298 /*
2299  * Check whether this filesystem can be mounted based on
2300  * the features present and the RDONLY/RDWR mount requested.
2301  * Returns 1 if this filesystem can be mounted as requested,
2302  * 0 if it cannot be.
2303  */
2304 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2305 {
2306 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2307 		ext4_msg(sb, KERN_ERR,
2308 			"Couldn't mount because of "
2309 			"unsupported optional features (%x)",
2310 			(le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2311 			~EXT4_FEATURE_INCOMPAT_SUPP));
2312 		return 0;
2313 	}
2314 
2315 	if (readonly)
2316 		return 1;
2317 
2318 	/* Check that feature set is OK for a read-write mount */
2319 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2320 		ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2321 			 "unsupported optional features (%x)",
2322 			 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2323 				~EXT4_FEATURE_RO_COMPAT_SUPP));
2324 		return 0;
2325 	}
2326 	/*
2327 	 * Large file size enabled file system can only be mounted
2328 	 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2329 	 */
2330 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2331 		if (sizeof(blkcnt_t) < sizeof(u64)) {
2332 			ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2333 				 "cannot be mounted RDWR without "
2334 				 "CONFIG_LBDAF");
2335 			return 0;
2336 		}
2337 	}
2338 	return 1;
2339 }
2340 
2341 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2342 				__releases(kernel_lock)
2343 				__acquires(kernel_lock)
2344 {
2345 	struct buffer_head *bh;
2346 	struct ext4_super_block *es = NULL;
2347 	struct ext4_sb_info *sbi;
2348 	ext4_fsblk_t block;
2349 	ext4_fsblk_t sb_block = get_sb_block(&data);
2350 	ext4_fsblk_t logical_sb_block;
2351 	unsigned long offset = 0;
2352 	unsigned long journal_devnum = 0;
2353 	unsigned long def_mount_opts;
2354 	struct inode *root;
2355 	char *cp;
2356 	const char *descr;
2357 	int ret = -EINVAL;
2358 	int blocksize;
2359 	unsigned int db_count;
2360 	unsigned int i;
2361 	int needs_recovery, has_huge_files;
2362 	__u64 blocks_count;
2363 	int err;
2364 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2365 
2366 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2367 	if (!sbi)
2368 		return -ENOMEM;
2369 
2370 	sbi->s_blockgroup_lock =
2371 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2372 	if (!sbi->s_blockgroup_lock) {
2373 		kfree(sbi);
2374 		return -ENOMEM;
2375 	}
2376 	sb->s_fs_info = sbi;
2377 	sbi->s_mount_opt = 0;
2378 	sbi->s_resuid = EXT4_DEF_RESUID;
2379 	sbi->s_resgid = EXT4_DEF_RESGID;
2380 	sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2381 	sbi->s_sb_block = sb_block;
2382 	sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2383 						      sectors[1]);
2384 
2385 	unlock_kernel();
2386 
2387 	/* Cleanup superblock name */
2388 	for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2389 		*cp = '!';
2390 
2391 	blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2392 	if (!blocksize) {
2393 		ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2394 		goto out_fail;
2395 	}
2396 
2397 	/*
2398 	 * The ext4 superblock will not be buffer aligned for other than 1kB
2399 	 * block sizes.  We need to calculate the offset from buffer start.
2400 	 */
2401 	if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2402 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2403 		offset = do_div(logical_sb_block, blocksize);
2404 	} else {
2405 		logical_sb_block = sb_block;
2406 	}
2407 
2408 	if (!(bh = sb_bread(sb, logical_sb_block))) {
2409 		ext4_msg(sb, KERN_ERR, "unable to read superblock");
2410 		goto out_fail;
2411 	}
2412 	/*
2413 	 * Note: s_es must be initialized as soon as possible because
2414 	 *       some ext4 macro-instructions depend on its value
2415 	 */
2416 	es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2417 	sbi->s_es = es;
2418 	sb->s_magic = le16_to_cpu(es->s_magic);
2419 	if (sb->s_magic != EXT4_SUPER_MAGIC)
2420 		goto cantfind_ext4;
2421 	sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2422 
2423 	/* Set defaults before we parse the mount options */
2424 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2425 	if (def_mount_opts & EXT4_DEFM_DEBUG)
2426 		set_opt(sbi->s_mount_opt, DEBUG);
2427 	if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2428 		set_opt(sbi->s_mount_opt, GRPID);
2429 	if (def_mount_opts & EXT4_DEFM_UID16)
2430 		set_opt(sbi->s_mount_opt, NO_UID32);
2431 #ifdef CONFIG_EXT4_FS_XATTR
2432 	if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2433 		set_opt(sbi->s_mount_opt, XATTR_USER);
2434 #endif
2435 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2436 	if (def_mount_opts & EXT4_DEFM_ACL)
2437 		set_opt(sbi->s_mount_opt, POSIX_ACL);
2438 #endif
2439 	if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2440 		sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2441 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2442 		sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2443 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2444 		sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2445 
2446 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2447 		set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2448 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2449 		set_opt(sbi->s_mount_opt, ERRORS_CONT);
2450 	else
2451 		set_opt(sbi->s_mount_opt, ERRORS_RO);
2452 
2453 	sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2454 	sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2455 	sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2456 	sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2457 	sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2458 
2459 	set_opt(sbi->s_mount_opt, BARRIER);
2460 
2461 	/*
2462 	 * enable delayed allocation by default
2463 	 * Use -o nodelalloc to turn it off
2464 	 */
2465 	set_opt(sbi->s_mount_opt, DELALLOC);
2466 
2467 	if (!parse_options((char *) data, sb, &journal_devnum,
2468 			   &journal_ioprio, NULL, 0))
2469 		goto failed_mount;
2470 
2471 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2472 		((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2473 
2474 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2475 	    (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2476 	     EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2477 	     EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2478 		ext4_msg(sb, KERN_WARNING,
2479 		       "feature flags set on rev 0 fs, "
2480 		       "running e2fsck is recommended");
2481 
2482 	/*
2483 	 * Check feature flags regardless of the revision level, since we
2484 	 * previously didn't change the revision level when setting the flags,
2485 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
2486 	 */
2487 	if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2488 		goto failed_mount;
2489 
2490 	blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2491 
2492 	if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2493 	    blocksize > EXT4_MAX_BLOCK_SIZE) {
2494 		ext4_msg(sb, KERN_ERR,
2495 		       "Unsupported filesystem blocksize %d", blocksize);
2496 		goto failed_mount;
2497 	}
2498 
2499 	if (sb->s_blocksize != blocksize) {
2500 		/* Validate the filesystem blocksize */
2501 		if (!sb_set_blocksize(sb, blocksize)) {
2502 			ext4_msg(sb, KERN_ERR, "bad block size %d",
2503 					blocksize);
2504 			goto failed_mount;
2505 		}
2506 
2507 		brelse(bh);
2508 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2509 		offset = do_div(logical_sb_block, blocksize);
2510 		bh = sb_bread(sb, logical_sb_block);
2511 		if (!bh) {
2512 			ext4_msg(sb, KERN_ERR,
2513 			       "Can't read superblock on 2nd try");
2514 			goto failed_mount;
2515 		}
2516 		es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2517 		sbi->s_es = es;
2518 		if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2519 			ext4_msg(sb, KERN_ERR,
2520 			       "Magic mismatch, very weird!");
2521 			goto failed_mount;
2522 		}
2523 	}
2524 
2525 	has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2526 				EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2527 	sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2528 						      has_huge_files);
2529 	sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2530 
2531 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2532 		sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2533 		sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2534 	} else {
2535 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2536 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2537 		if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2538 		    (!is_power_of_2(sbi->s_inode_size)) ||
2539 		    (sbi->s_inode_size > blocksize)) {
2540 			ext4_msg(sb, KERN_ERR,
2541 			       "unsupported inode size: %d",
2542 			       sbi->s_inode_size);
2543 			goto failed_mount;
2544 		}
2545 		if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2546 			sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2547 	}
2548 
2549 	sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2550 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2551 		if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2552 		    sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2553 		    !is_power_of_2(sbi->s_desc_size)) {
2554 			ext4_msg(sb, KERN_ERR,
2555 			       "unsupported descriptor size %lu",
2556 			       sbi->s_desc_size);
2557 			goto failed_mount;
2558 		}
2559 	} else
2560 		sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2561 
2562 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2563 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2564 	if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2565 		goto cantfind_ext4;
2566 
2567 	sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2568 	if (sbi->s_inodes_per_block == 0)
2569 		goto cantfind_ext4;
2570 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
2571 					sbi->s_inodes_per_block;
2572 	sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2573 	sbi->s_sbh = bh;
2574 	sbi->s_mount_state = le16_to_cpu(es->s_state);
2575 	sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2576 	sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2577 
2578 	for (i = 0; i < 4; i++)
2579 		sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2580 	sbi->s_def_hash_version = es->s_def_hash_version;
2581 	i = le32_to_cpu(es->s_flags);
2582 	if (i & EXT2_FLAGS_UNSIGNED_HASH)
2583 		sbi->s_hash_unsigned = 3;
2584 	else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2585 #ifdef __CHAR_UNSIGNED__
2586 		es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2587 		sbi->s_hash_unsigned = 3;
2588 #else
2589 		es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2590 #endif
2591 		sb->s_dirt = 1;
2592 	}
2593 
2594 	if (sbi->s_blocks_per_group > blocksize * 8) {
2595 		ext4_msg(sb, KERN_ERR,
2596 		       "#blocks per group too big: %lu",
2597 		       sbi->s_blocks_per_group);
2598 		goto failed_mount;
2599 	}
2600 	if (sbi->s_inodes_per_group > blocksize * 8) {
2601 		ext4_msg(sb, KERN_ERR,
2602 		       "#inodes per group too big: %lu",
2603 		       sbi->s_inodes_per_group);
2604 		goto failed_mount;
2605 	}
2606 
2607 	/*
2608 	 * Test whether we have more sectors than will fit in sector_t,
2609 	 * and whether the max offset is addressable by the page cache.
2610 	 */
2611 	if ((ext4_blocks_count(es) >
2612 	     (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2613 	    (ext4_blocks_count(es) >
2614 	     (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2615 		ext4_msg(sb, KERN_ERR, "filesystem"
2616 			 " too large to mount safely on this system");
2617 		if (sizeof(sector_t) < 8)
2618 			ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2619 		ret = -EFBIG;
2620 		goto failed_mount;
2621 	}
2622 
2623 	if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2624 		goto cantfind_ext4;
2625 
2626 	/* check blocks count against device size */
2627 	blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2628 	if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2629 		ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2630 		       "exceeds size of device (%llu blocks)",
2631 		       ext4_blocks_count(es), blocks_count);
2632 		goto failed_mount;
2633 	}
2634 
2635 	/*
2636 	 * It makes no sense for the first data block to be beyond the end
2637 	 * of the filesystem.
2638 	 */
2639 	if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2640                 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2641 			 "block %u is beyond end of filesystem (%llu)",
2642 			 le32_to_cpu(es->s_first_data_block),
2643 			 ext4_blocks_count(es));
2644 		goto failed_mount;
2645 	}
2646 	blocks_count = (ext4_blocks_count(es) -
2647 			le32_to_cpu(es->s_first_data_block) +
2648 			EXT4_BLOCKS_PER_GROUP(sb) - 1);
2649 	do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2650 	if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2651 		ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2652 		       "(block count %llu, first data block %u, "
2653 		       "blocks per group %lu)", sbi->s_groups_count,
2654 		       ext4_blocks_count(es),
2655 		       le32_to_cpu(es->s_first_data_block),
2656 		       EXT4_BLOCKS_PER_GROUP(sb));
2657 		goto failed_mount;
2658 	}
2659 	sbi->s_groups_count = blocks_count;
2660 	sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2661 			(EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2662 	db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2663 		   EXT4_DESC_PER_BLOCK(sb);
2664 	sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2665 				    GFP_KERNEL);
2666 	if (sbi->s_group_desc == NULL) {
2667 		ext4_msg(sb, KERN_ERR, "not enough memory");
2668 		goto failed_mount;
2669 	}
2670 
2671 #ifdef CONFIG_PROC_FS
2672 	if (ext4_proc_root)
2673 		sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2674 #endif
2675 
2676 	bgl_lock_init(sbi->s_blockgroup_lock);
2677 
2678 	for (i = 0; i < db_count; i++) {
2679 		block = descriptor_loc(sb, logical_sb_block, i);
2680 		sbi->s_group_desc[i] = sb_bread(sb, block);
2681 		if (!sbi->s_group_desc[i]) {
2682 			ext4_msg(sb, KERN_ERR,
2683 			       "can't read group descriptor %d", i);
2684 			db_count = i;
2685 			goto failed_mount2;
2686 		}
2687 	}
2688 	if (!ext4_check_descriptors(sb)) {
2689 		ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2690 		goto failed_mount2;
2691 	}
2692 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2693 		if (!ext4_fill_flex_info(sb)) {
2694 			ext4_msg(sb, KERN_ERR,
2695 			       "unable to initialize "
2696 			       "flex_bg meta info!");
2697 			goto failed_mount2;
2698 		}
2699 
2700 	sbi->s_gdb_count = db_count;
2701 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2702 	spin_lock_init(&sbi->s_next_gen_lock);
2703 
2704 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
2705 			ext4_count_free_blocks(sb));
2706 	if (!err) {
2707 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
2708 				ext4_count_free_inodes(sb));
2709 	}
2710 	if (!err) {
2711 		err = percpu_counter_init(&sbi->s_dirs_counter,
2712 				ext4_count_dirs(sb));
2713 	}
2714 	if (!err) {
2715 		err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2716 	}
2717 	if (err) {
2718 		ext4_msg(sb, KERN_ERR, "insufficient memory");
2719 		goto failed_mount3;
2720 	}
2721 
2722 	sbi->s_stripe = ext4_get_stripe_size(sbi);
2723 	sbi->s_max_writeback_mb_bump = 128;
2724 
2725 	/*
2726 	 * set up enough so that it can read an inode
2727 	 */
2728 	if (!test_opt(sb, NOLOAD) &&
2729 	    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2730 		sb->s_op = &ext4_sops;
2731 	else
2732 		sb->s_op = &ext4_nojournal_sops;
2733 	sb->s_export_op = &ext4_export_ops;
2734 	sb->s_xattr = ext4_xattr_handlers;
2735 #ifdef CONFIG_QUOTA
2736 	sb->s_qcop = &ext4_qctl_operations;
2737 	sb->dq_op = &ext4_quota_operations;
2738 #endif
2739 	INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2740 	mutex_init(&sbi->s_orphan_lock);
2741 	mutex_init(&sbi->s_resize_lock);
2742 
2743 	sb->s_root = NULL;
2744 
2745 	needs_recovery = (es->s_last_orphan != 0 ||
2746 			  EXT4_HAS_INCOMPAT_FEATURE(sb,
2747 				    EXT4_FEATURE_INCOMPAT_RECOVER));
2748 
2749 	/*
2750 	 * The first inode we look at is the journal inode.  Don't try
2751 	 * root first: it may be modified in the journal!
2752 	 */
2753 	if (!test_opt(sb, NOLOAD) &&
2754 	    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2755 		if (ext4_load_journal(sb, es, journal_devnum))
2756 			goto failed_mount3;
2757 	} else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2758 	      EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2759 		ext4_msg(sb, KERN_ERR, "required journal recovery "
2760 		       "suppressed and not mounted read-only");
2761 		goto failed_mount4;
2762 	} else {
2763 		clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2764 		set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2765 		sbi->s_journal = NULL;
2766 		needs_recovery = 0;
2767 		goto no_journal;
2768 	}
2769 
2770 	if (ext4_blocks_count(es) > 0xffffffffULL &&
2771 	    !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2772 				       JBD2_FEATURE_INCOMPAT_64BIT)) {
2773 		ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2774 		goto failed_mount4;
2775 	}
2776 
2777 	if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2778 		jbd2_journal_set_features(sbi->s_journal,
2779 				JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2780 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2781 	} else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2782 		jbd2_journal_set_features(sbi->s_journal,
2783 				JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2784 		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2785 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2786 	} else {
2787 		jbd2_journal_clear_features(sbi->s_journal,
2788 				JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2789 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2790 	}
2791 
2792 	/* We have now updated the journal if required, so we can
2793 	 * validate the data journaling mode. */
2794 	switch (test_opt(sb, DATA_FLAGS)) {
2795 	case 0:
2796 		/* No mode set, assume a default based on the journal
2797 		 * capabilities: ORDERED_DATA if the journal can
2798 		 * cope, else JOURNAL_DATA
2799 		 */
2800 		if (jbd2_journal_check_available_features
2801 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2802 			set_opt(sbi->s_mount_opt, ORDERED_DATA);
2803 		else
2804 			set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2805 		break;
2806 
2807 	case EXT4_MOUNT_ORDERED_DATA:
2808 	case EXT4_MOUNT_WRITEBACK_DATA:
2809 		if (!jbd2_journal_check_available_features
2810 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2811 			ext4_msg(sb, KERN_ERR, "Journal does not support "
2812 			       "requested data journaling mode");
2813 			goto failed_mount4;
2814 		}
2815 	default:
2816 		break;
2817 	}
2818 	set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2819 
2820 no_journal:
2821 
2822 	if (test_opt(sb, NOBH)) {
2823 		if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2824 			ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2825 				"its supported only with writeback mode");
2826 			clear_opt(sbi->s_mount_opt, NOBH);
2827 		}
2828 	}
2829 	EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2830 	if (!EXT4_SB(sb)->dio_unwritten_wq) {
2831 		printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2832 		goto failed_mount_wq;
2833 	}
2834 
2835 	/*
2836 	 * The jbd2_journal_load will have done any necessary log recovery,
2837 	 * so we can safely mount the rest of the filesystem now.
2838 	 */
2839 
2840 	root = ext4_iget(sb, EXT4_ROOT_INO);
2841 	if (IS_ERR(root)) {
2842 		ext4_msg(sb, KERN_ERR, "get root inode failed");
2843 		ret = PTR_ERR(root);
2844 		goto failed_mount4;
2845 	}
2846 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2847 		iput(root);
2848 		ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
2849 		goto failed_mount4;
2850 	}
2851 	sb->s_root = d_alloc_root(root);
2852 	if (!sb->s_root) {
2853 		ext4_msg(sb, KERN_ERR, "get root dentry failed");
2854 		iput(root);
2855 		ret = -ENOMEM;
2856 		goto failed_mount4;
2857 	}
2858 
2859 	ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2860 
2861 	/* determine the minimum size of new large inodes, if present */
2862 	if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2863 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2864 						     EXT4_GOOD_OLD_INODE_SIZE;
2865 		if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2866 				       EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2867 			if (sbi->s_want_extra_isize <
2868 			    le16_to_cpu(es->s_want_extra_isize))
2869 				sbi->s_want_extra_isize =
2870 					le16_to_cpu(es->s_want_extra_isize);
2871 			if (sbi->s_want_extra_isize <
2872 			    le16_to_cpu(es->s_min_extra_isize))
2873 				sbi->s_want_extra_isize =
2874 					le16_to_cpu(es->s_min_extra_isize);
2875 		}
2876 	}
2877 	/* Check if enough inode space is available */
2878 	if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2879 							sbi->s_inode_size) {
2880 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2881 						       EXT4_GOOD_OLD_INODE_SIZE;
2882 		ext4_msg(sb, KERN_INFO, "required extra inode space not"
2883 			 "available");
2884 	}
2885 
2886 	if (test_opt(sb, DELALLOC) &&
2887 	    (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2888 		ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2889 			 "requested data journaling mode");
2890 		clear_opt(sbi->s_mount_opt, DELALLOC);
2891 	}
2892 
2893 	err = ext4_setup_system_zone(sb);
2894 	if (err) {
2895 		ext4_msg(sb, KERN_ERR, "failed to initialize system "
2896 			 "zone (%d)\n", err);
2897 		goto failed_mount4;
2898 	}
2899 
2900 	ext4_ext_init(sb);
2901 	err = ext4_mb_init(sb, needs_recovery);
2902 	if (err) {
2903 		ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2904 			 err);
2905 		goto failed_mount4;
2906 	}
2907 
2908 	sbi->s_kobj.kset = ext4_kset;
2909 	init_completion(&sbi->s_kobj_unregister);
2910 	err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2911 				   "%s", sb->s_id);
2912 	if (err) {
2913 		ext4_mb_release(sb);
2914 		ext4_ext_release(sb);
2915 		goto failed_mount4;
2916 	};
2917 
2918 	EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2919 	ext4_orphan_cleanup(sb, es);
2920 	EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2921 	if (needs_recovery) {
2922 		ext4_msg(sb, KERN_INFO, "recovery complete");
2923 		ext4_mark_recovery_complete(sb, es);
2924 	}
2925 	if (EXT4_SB(sb)->s_journal) {
2926 		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2927 			descr = " journalled data mode";
2928 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2929 			descr = " ordered data mode";
2930 		else
2931 			descr = " writeback data mode";
2932 	} else
2933 		descr = "out journal";
2934 
2935 	ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
2936 
2937 	lock_kernel();
2938 	return 0;
2939 
2940 cantfind_ext4:
2941 	if (!silent)
2942 		ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
2943 	goto failed_mount;
2944 
2945 failed_mount4:
2946 	ext4_msg(sb, KERN_ERR, "mount failed");
2947 	destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
2948 failed_mount_wq:
2949 	ext4_release_system_zone(sb);
2950 	if (sbi->s_journal) {
2951 		jbd2_journal_destroy(sbi->s_journal);
2952 		sbi->s_journal = NULL;
2953 	}
2954 failed_mount3:
2955 	if (sbi->s_flex_groups) {
2956 		if (is_vmalloc_addr(sbi->s_flex_groups))
2957 			vfree(sbi->s_flex_groups);
2958 		else
2959 			kfree(sbi->s_flex_groups);
2960 	}
2961 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
2962 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
2963 	percpu_counter_destroy(&sbi->s_dirs_counter);
2964 	percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2965 failed_mount2:
2966 	for (i = 0; i < db_count; i++)
2967 		brelse(sbi->s_group_desc[i]);
2968 	kfree(sbi->s_group_desc);
2969 failed_mount:
2970 	if (sbi->s_proc) {
2971 		remove_proc_entry(sb->s_id, ext4_proc_root);
2972 	}
2973 #ifdef CONFIG_QUOTA
2974 	for (i = 0; i < MAXQUOTAS; i++)
2975 		kfree(sbi->s_qf_names[i]);
2976 #endif
2977 	ext4_blkdev_remove(sbi);
2978 	brelse(bh);
2979 out_fail:
2980 	sb->s_fs_info = NULL;
2981 	kfree(sbi->s_blockgroup_lock);
2982 	kfree(sbi);
2983 	lock_kernel();
2984 	return ret;
2985 }
2986 
2987 /*
2988  * Setup any per-fs journal parameters now.  We'll do this both on
2989  * initial mount, once the journal has been initialised but before we've
2990  * done any recovery; and again on any subsequent remount.
2991  */
2992 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2993 {
2994 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2995 
2996 	journal->j_commit_interval = sbi->s_commit_interval;
2997 	journal->j_min_batch_time = sbi->s_min_batch_time;
2998 	journal->j_max_batch_time = sbi->s_max_batch_time;
2999 
3000 	spin_lock(&journal->j_state_lock);
3001 	if (test_opt(sb, BARRIER))
3002 		journal->j_flags |= JBD2_BARRIER;
3003 	else
3004 		journal->j_flags &= ~JBD2_BARRIER;
3005 	if (test_opt(sb, DATA_ERR_ABORT))
3006 		journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3007 	else
3008 		journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3009 	spin_unlock(&journal->j_state_lock);
3010 }
3011 
3012 static journal_t *ext4_get_journal(struct super_block *sb,
3013 				   unsigned int journal_inum)
3014 {
3015 	struct inode *journal_inode;
3016 	journal_t *journal;
3017 
3018 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3019 
3020 	/* First, test for the existence of a valid inode on disk.  Bad
3021 	 * things happen if we iget() an unused inode, as the subsequent
3022 	 * iput() will try to delete it. */
3023 
3024 	journal_inode = ext4_iget(sb, journal_inum);
3025 	if (IS_ERR(journal_inode)) {
3026 		ext4_msg(sb, KERN_ERR, "no journal found");
3027 		return NULL;
3028 	}
3029 	if (!journal_inode->i_nlink) {
3030 		make_bad_inode(journal_inode);
3031 		iput(journal_inode);
3032 		ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3033 		return NULL;
3034 	}
3035 
3036 	jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3037 		  journal_inode, journal_inode->i_size);
3038 	if (!S_ISREG(journal_inode->i_mode)) {
3039 		ext4_msg(sb, KERN_ERR, "invalid journal inode");
3040 		iput(journal_inode);
3041 		return NULL;
3042 	}
3043 
3044 	journal = jbd2_journal_init_inode(journal_inode);
3045 	if (!journal) {
3046 		ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3047 		iput(journal_inode);
3048 		return NULL;
3049 	}
3050 	journal->j_private = sb;
3051 	ext4_init_journal_params(sb, journal);
3052 	return journal;
3053 }
3054 
3055 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3056 				       dev_t j_dev)
3057 {
3058 	struct buffer_head *bh;
3059 	journal_t *journal;
3060 	ext4_fsblk_t start;
3061 	ext4_fsblk_t len;
3062 	int hblock, blocksize;
3063 	ext4_fsblk_t sb_block;
3064 	unsigned long offset;
3065 	struct ext4_super_block *es;
3066 	struct block_device *bdev;
3067 
3068 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3069 
3070 	bdev = ext4_blkdev_get(j_dev, sb);
3071 	if (bdev == NULL)
3072 		return NULL;
3073 
3074 	if (bd_claim(bdev, sb)) {
3075 		ext4_msg(sb, KERN_ERR,
3076 			"failed to claim external journal device");
3077 		blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3078 		return NULL;
3079 	}
3080 
3081 	blocksize = sb->s_blocksize;
3082 	hblock = bdev_logical_block_size(bdev);
3083 	if (blocksize < hblock) {
3084 		ext4_msg(sb, KERN_ERR,
3085 			"blocksize too small for journal device");
3086 		goto out_bdev;
3087 	}
3088 
3089 	sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3090 	offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3091 	set_blocksize(bdev, blocksize);
3092 	if (!(bh = __bread(bdev, sb_block, blocksize))) {
3093 		ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3094 		       "external journal");
3095 		goto out_bdev;
3096 	}
3097 
3098 	es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3099 	if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3100 	    !(le32_to_cpu(es->s_feature_incompat) &
3101 	      EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3102 		ext4_msg(sb, KERN_ERR, "external journal has "
3103 					"bad superblock");
3104 		brelse(bh);
3105 		goto out_bdev;
3106 	}
3107 
3108 	if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3109 		ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3110 		brelse(bh);
3111 		goto out_bdev;
3112 	}
3113 
3114 	len = ext4_blocks_count(es);
3115 	start = sb_block + 1;
3116 	brelse(bh);	/* we're done with the superblock */
3117 
3118 	journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3119 					start, len, blocksize);
3120 	if (!journal) {
3121 		ext4_msg(sb, KERN_ERR, "failed to create device journal");
3122 		goto out_bdev;
3123 	}
3124 	journal->j_private = sb;
3125 	ll_rw_block(READ, 1, &journal->j_sb_buffer);
3126 	wait_on_buffer(journal->j_sb_buffer);
3127 	if (!buffer_uptodate(journal->j_sb_buffer)) {
3128 		ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3129 		goto out_journal;
3130 	}
3131 	if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3132 		ext4_msg(sb, KERN_ERR, "External journal has more than one "
3133 					"user (unsupported) - %d",
3134 			be32_to_cpu(journal->j_superblock->s_nr_users));
3135 		goto out_journal;
3136 	}
3137 	EXT4_SB(sb)->journal_bdev = bdev;
3138 	ext4_init_journal_params(sb, journal);
3139 	return journal;
3140 
3141 out_journal:
3142 	jbd2_journal_destroy(journal);
3143 out_bdev:
3144 	ext4_blkdev_put(bdev);
3145 	return NULL;
3146 }
3147 
3148 static int ext4_load_journal(struct super_block *sb,
3149 			     struct ext4_super_block *es,
3150 			     unsigned long journal_devnum)
3151 {
3152 	journal_t *journal;
3153 	unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3154 	dev_t journal_dev;
3155 	int err = 0;
3156 	int really_read_only;
3157 
3158 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3159 
3160 	if (journal_devnum &&
3161 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3162 		ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3163 			"numbers have changed");
3164 		journal_dev = new_decode_dev(journal_devnum);
3165 	} else
3166 		journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3167 
3168 	really_read_only = bdev_read_only(sb->s_bdev);
3169 
3170 	/*
3171 	 * Are we loading a blank journal or performing recovery after a
3172 	 * crash?  For recovery, we need to check in advance whether we
3173 	 * can get read-write access to the device.
3174 	 */
3175 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3176 		if (sb->s_flags & MS_RDONLY) {
3177 			ext4_msg(sb, KERN_INFO, "INFO: recovery "
3178 					"required on readonly filesystem");
3179 			if (really_read_only) {
3180 				ext4_msg(sb, KERN_ERR, "write access "
3181 					"unavailable, cannot proceed");
3182 				return -EROFS;
3183 			}
3184 			ext4_msg(sb, KERN_INFO, "write access will "
3185 			       "be enabled during recovery");
3186 		}
3187 	}
3188 
3189 	if (journal_inum && journal_dev) {
3190 		ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3191 		       "and inode journals!");
3192 		return -EINVAL;
3193 	}
3194 
3195 	if (journal_inum) {
3196 		if (!(journal = ext4_get_journal(sb, journal_inum)))
3197 			return -EINVAL;
3198 	} else {
3199 		if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3200 			return -EINVAL;
3201 	}
3202 
3203 	if (!(journal->j_flags & JBD2_BARRIER))
3204 		ext4_msg(sb, KERN_INFO, "barriers disabled");
3205 
3206 	if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3207 		err = jbd2_journal_update_format(journal);
3208 		if (err)  {
3209 			ext4_msg(sb, KERN_ERR, "error updating journal");
3210 			jbd2_journal_destroy(journal);
3211 			return err;
3212 		}
3213 	}
3214 
3215 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3216 		err = jbd2_journal_wipe(journal, !really_read_only);
3217 	if (!err)
3218 		err = jbd2_journal_load(journal);
3219 
3220 	if (err) {
3221 		ext4_msg(sb, KERN_ERR, "error loading journal");
3222 		jbd2_journal_destroy(journal);
3223 		return err;
3224 	}
3225 
3226 	EXT4_SB(sb)->s_journal = journal;
3227 	ext4_clear_journal_err(sb, es);
3228 
3229 	if (journal_devnum &&
3230 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3231 		es->s_journal_dev = cpu_to_le32(journal_devnum);
3232 
3233 		/* Make sure we flush the recovery flag to disk. */
3234 		ext4_commit_super(sb, 1);
3235 	}
3236 
3237 	return 0;
3238 }
3239 
3240 static int ext4_commit_super(struct super_block *sb, int sync)
3241 {
3242 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3243 	struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3244 	int error = 0;
3245 
3246 	if (!sbh)
3247 		return error;
3248 	if (buffer_write_io_error(sbh)) {
3249 		/*
3250 		 * Oh, dear.  A previous attempt to write the
3251 		 * superblock failed.  This could happen because the
3252 		 * USB device was yanked out.  Or it could happen to
3253 		 * be a transient write error and maybe the block will
3254 		 * be remapped.  Nothing we can do but to retry the
3255 		 * write and hope for the best.
3256 		 */
3257 		ext4_msg(sb, KERN_ERR, "previous I/O error to "
3258 		       "superblock detected");
3259 		clear_buffer_write_io_error(sbh);
3260 		set_buffer_uptodate(sbh);
3261 	}
3262 	/*
3263 	 * If the file system is mounted read-only, don't update the
3264 	 * superblock write time.  This avoids updating the superblock
3265 	 * write time when we are mounting the root file system
3266 	 * read/only but we need to replay the journal; at that point,
3267 	 * for people who are east of GMT and who make their clock
3268 	 * tick in localtime for Windows bug-for-bug compatibility,
3269 	 * the clock is set in the future, and this will cause e2fsck
3270 	 * to complain and force a full file system check.
3271 	 */
3272 	if (!(sb->s_flags & MS_RDONLY))
3273 		es->s_wtime = cpu_to_le32(get_seconds());
3274 	es->s_kbytes_written =
3275 		cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3276 			    ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3277 			      EXT4_SB(sb)->s_sectors_written_start) >> 1));
3278 	ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3279 					&EXT4_SB(sb)->s_freeblocks_counter));
3280 	es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3281 					&EXT4_SB(sb)->s_freeinodes_counter));
3282 	sb->s_dirt = 0;
3283 	BUFFER_TRACE(sbh, "marking dirty");
3284 	mark_buffer_dirty(sbh);
3285 	if (sync) {
3286 		error = sync_dirty_buffer(sbh);
3287 		if (error)
3288 			return error;
3289 
3290 		error = buffer_write_io_error(sbh);
3291 		if (error) {
3292 			ext4_msg(sb, KERN_ERR, "I/O error while writing "
3293 			       "superblock");
3294 			clear_buffer_write_io_error(sbh);
3295 			set_buffer_uptodate(sbh);
3296 		}
3297 	}
3298 	return error;
3299 }
3300 
3301 /*
3302  * Have we just finished recovery?  If so, and if we are mounting (or
3303  * remounting) the filesystem readonly, then we will end up with a
3304  * consistent fs on disk.  Record that fact.
3305  */
3306 static void ext4_mark_recovery_complete(struct super_block *sb,
3307 					struct ext4_super_block *es)
3308 {
3309 	journal_t *journal = EXT4_SB(sb)->s_journal;
3310 
3311 	if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3312 		BUG_ON(journal != NULL);
3313 		return;
3314 	}
3315 	jbd2_journal_lock_updates(journal);
3316 	if (jbd2_journal_flush(journal) < 0)
3317 		goto out;
3318 
3319 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3320 	    sb->s_flags & MS_RDONLY) {
3321 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3322 		ext4_commit_super(sb, 1);
3323 	}
3324 
3325 out:
3326 	jbd2_journal_unlock_updates(journal);
3327 }
3328 
3329 /*
3330  * If we are mounting (or read-write remounting) a filesystem whose journal
3331  * has recorded an error from a previous lifetime, move that error to the
3332  * main filesystem now.
3333  */
3334 static void ext4_clear_journal_err(struct super_block *sb,
3335 				   struct ext4_super_block *es)
3336 {
3337 	journal_t *journal;
3338 	int j_errno;
3339 	const char *errstr;
3340 
3341 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3342 
3343 	journal = EXT4_SB(sb)->s_journal;
3344 
3345 	/*
3346 	 * Now check for any error status which may have been recorded in the
3347 	 * journal by a prior ext4_error() or ext4_abort()
3348 	 */
3349 
3350 	j_errno = jbd2_journal_errno(journal);
3351 	if (j_errno) {
3352 		char nbuf[16];
3353 
3354 		errstr = ext4_decode_error(sb, j_errno, nbuf);
3355 		ext4_warning(sb, __func__, "Filesystem error recorded "
3356 			     "from previous mount: %s", errstr);
3357 		ext4_warning(sb, __func__, "Marking fs in need of "
3358 			     "filesystem check.");
3359 
3360 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3361 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3362 		ext4_commit_super(sb, 1);
3363 
3364 		jbd2_journal_clear_err(journal);
3365 	}
3366 }
3367 
3368 /*
3369  * Force the running and committing transactions to commit,
3370  * and wait on the commit.
3371  */
3372 int ext4_force_commit(struct super_block *sb)
3373 {
3374 	journal_t *journal;
3375 	int ret = 0;
3376 
3377 	if (sb->s_flags & MS_RDONLY)
3378 		return 0;
3379 
3380 	journal = EXT4_SB(sb)->s_journal;
3381 	if (journal)
3382 		ret = ext4_journal_force_commit(journal);
3383 
3384 	return ret;
3385 }
3386 
3387 static void ext4_write_super(struct super_block *sb)
3388 {
3389 	lock_super(sb);
3390 	ext4_commit_super(sb, 1);
3391 	unlock_super(sb);
3392 }
3393 
3394 static int ext4_sync_fs(struct super_block *sb, int wait)
3395 {
3396 	int ret = 0;
3397 	tid_t target;
3398 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3399 
3400 	trace_ext4_sync_fs(sb, wait);
3401 	flush_workqueue(sbi->dio_unwritten_wq);
3402 	if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
3403 		if (wait)
3404 			jbd2_log_wait_commit(sbi->s_journal, target);
3405 	}
3406 	return ret;
3407 }
3408 
3409 /*
3410  * LVM calls this function before a (read-only) snapshot is created.  This
3411  * gives us a chance to flush the journal completely and mark the fs clean.
3412  */
3413 static int ext4_freeze(struct super_block *sb)
3414 {
3415 	int error = 0;
3416 	journal_t *journal;
3417 
3418 	if (sb->s_flags & MS_RDONLY)
3419 		return 0;
3420 
3421 	journal = EXT4_SB(sb)->s_journal;
3422 
3423 	/* Now we set up the journal barrier. */
3424 	jbd2_journal_lock_updates(journal);
3425 
3426 	/*
3427 	 * Don't clear the needs_recovery flag if we failed to flush
3428 	 * the journal.
3429 	 */
3430 	error = jbd2_journal_flush(journal);
3431 	if (error < 0) {
3432 	out:
3433 		jbd2_journal_unlock_updates(journal);
3434 		return error;
3435 	}
3436 
3437 	/* Journal blocked and flushed, clear needs_recovery flag. */
3438 	EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3439 	error = ext4_commit_super(sb, 1);
3440 	if (error)
3441 		goto out;
3442 	return 0;
3443 }
3444 
3445 /*
3446  * Called by LVM after the snapshot is done.  We need to reset the RECOVER
3447  * flag here, even though the filesystem is not technically dirty yet.
3448  */
3449 static int ext4_unfreeze(struct super_block *sb)
3450 {
3451 	if (sb->s_flags & MS_RDONLY)
3452 		return 0;
3453 
3454 	lock_super(sb);
3455 	/* Reset the needs_recovery flag before the fs is unlocked. */
3456 	EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3457 	ext4_commit_super(sb, 1);
3458 	unlock_super(sb);
3459 	jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3460 	return 0;
3461 }
3462 
3463 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3464 {
3465 	struct ext4_super_block *es;
3466 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3467 	ext4_fsblk_t n_blocks_count = 0;
3468 	unsigned long old_sb_flags;
3469 	struct ext4_mount_options old_opts;
3470 	ext4_group_t g;
3471 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3472 	int err;
3473 #ifdef CONFIG_QUOTA
3474 	int i;
3475 #endif
3476 
3477 	lock_kernel();
3478 
3479 	/* Store the original options */
3480 	lock_super(sb);
3481 	old_sb_flags = sb->s_flags;
3482 	old_opts.s_mount_opt = sbi->s_mount_opt;
3483 	old_opts.s_resuid = sbi->s_resuid;
3484 	old_opts.s_resgid = sbi->s_resgid;
3485 	old_opts.s_commit_interval = sbi->s_commit_interval;
3486 	old_opts.s_min_batch_time = sbi->s_min_batch_time;
3487 	old_opts.s_max_batch_time = sbi->s_max_batch_time;
3488 #ifdef CONFIG_QUOTA
3489 	old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3490 	for (i = 0; i < MAXQUOTAS; i++)
3491 		old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3492 #endif
3493 	if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3494 		journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3495 
3496 	/*
3497 	 * Allow the "check" option to be passed as a remount option.
3498 	 */
3499 	if (!parse_options(data, sb, NULL, &journal_ioprio,
3500 			   &n_blocks_count, 1)) {
3501 		err = -EINVAL;
3502 		goto restore_opts;
3503 	}
3504 
3505 	if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3506 		ext4_abort(sb, __func__, "Abort forced by user");
3507 
3508 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3509 		((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3510 
3511 	es = sbi->s_es;
3512 
3513 	if (sbi->s_journal) {
3514 		ext4_init_journal_params(sb, sbi->s_journal);
3515 		set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3516 	}
3517 
3518 	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3519 		n_blocks_count > ext4_blocks_count(es)) {
3520 		if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3521 			err = -EROFS;
3522 			goto restore_opts;
3523 		}
3524 
3525 		if (*flags & MS_RDONLY) {
3526 			/*
3527 			 * First of all, the unconditional stuff we have to do
3528 			 * to disable replay of the journal when we next remount
3529 			 */
3530 			sb->s_flags |= MS_RDONLY;
3531 
3532 			/*
3533 			 * OK, test if we are remounting a valid rw partition
3534 			 * readonly, and if so set the rdonly flag and then
3535 			 * mark the partition as valid again.
3536 			 */
3537 			if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3538 			    (sbi->s_mount_state & EXT4_VALID_FS))
3539 				es->s_state = cpu_to_le16(sbi->s_mount_state);
3540 
3541 			if (sbi->s_journal)
3542 				ext4_mark_recovery_complete(sb, es);
3543 		} else {
3544 			/* Make sure we can mount this feature set readwrite */
3545 			if (!ext4_feature_set_ok(sb, 0)) {
3546 				err = -EROFS;
3547 				goto restore_opts;
3548 			}
3549 			/*
3550 			 * Make sure the group descriptor checksums
3551 			 * are sane.  If they aren't, refuse to remount r/w.
3552 			 */
3553 			for (g = 0; g < sbi->s_groups_count; g++) {
3554 				struct ext4_group_desc *gdp =
3555 					ext4_get_group_desc(sb, g, NULL);
3556 
3557 				if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3558 					ext4_msg(sb, KERN_ERR,
3559 	       "ext4_remount: Checksum for group %u failed (%u!=%u)",
3560 		g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3561 					       le16_to_cpu(gdp->bg_checksum));
3562 					err = -EINVAL;
3563 					goto restore_opts;
3564 				}
3565 			}
3566 
3567 			/*
3568 			 * If we have an unprocessed orphan list hanging
3569 			 * around from a previously readonly bdev mount,
3570 			 * require a full umount/remount for now.
3571 			 */
3572 			if (es->s_last_orphan) {
3573 				ext4_msg(sb, KERN_WARNING, "Couldn't "
3574 				       "remount RDWR because of unprocessed "
3575 				       "orphan inode list.  Please "
3576 				       "umount/remount instead");
3577 				err = -EINVAL;
3578 				goto restore_opts;
3579 			}
3580 
3581 			/*
3582 			 * Mounting a RDONLY partition read-write, so reread
3583 			 * and store the current valid flag.  (It may have
3584 			 * been changed by e2fsck since we originally mounted
3585 			 * the partition.)
3586 			 */
3587 			if (sbi->s_journal)
3588 				ext4_clear_journal_err(sb, es);
3589 			sbi->s_mount_state = le16_to_cpu(es->s_state);
3590 			if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3591 				goto restore_opts;
3592 			if (!ext4_setup_super(sb, es, 0))
3593 				sb->s_flags &= ~MS_RDONLY;
3594 		}
3595 	}
3596 	ext4_setup_system_zone(sb);
3597 	if (sbi->s_journal == NULL)
3598 		ext4_commit_super(sb, 1);
3599 
3600 #ifdef CONFIG_QUOTA
3601 	/* Release old quota file names */
3602 	for (i = 0; i < MAXQUOTAS; i++)
3603 		if (old_opts.s_qf_names[i] &&
3604 		    old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3605 			kfree(old_opts.s_qf_names[i]);
3606 #endif
3607 	unlock_super(sb);
3608 	unlock_kernel();
3609 	return 0;
3610 
3611 restore_opts:
3612 	sb->s_flags = old_sb_flags;
3613 	sbi->s_mount_opt = old_opts.s_mount_opt;
3614 	sbi->s_resuid = old_opts.s_resuid;
3615 	sbi->s_resgid = old_opts.s_resgid;
3616 	sbi->s_commit_interval = old_opts.s_commit_interval;
3617 	sbi->s_min_batch_time = old_opts.s_min_batch_time;
3618 	sbi->s_max_batch_time = old_opts.s_max_batch_time;
3619 #ifdef CONFIG_QUOTA
3620 	sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3621 	for (i = 0; i < MAXQUOTAS; i++) {
3622 		if (sbi->s_qf_names[i] &&
3623 		    old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3624 			kfree(sbi->s_qf_names[i]);
3625 		sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3626 	}
3627 #endif
3628 	unlock_super(sb);
3629 	unlock_kernel();
3630 	return err;
3631 }
3632 
3633 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3634 {
3635 	struct super_block *sb = dentry->d_sb;
3636 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3637 	struct ext4_super_block *es = sbi->s_es;
3638 	u64 fsid;
3639 
3640 	if (test_opt(sb, MINIX_DF)) {
3641 		sbi->s_overhead_last = 0;
3642 	} else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3643 		ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3644 		ext4_fsblk_t overhead = 0;
3645 
3646 		/*
3647 		 * Compute the overhead (FS structures).  This is constant
3648 		 * for a given filesystem unless the number of block groups
3649 		 * changes so we cache the previous value until it does.
3650 		 */
3651 
3652 		/*
3653 		 * All of the blocks before first_data_block are
3654 		 * overhead
3655 		 */
3656 		overhead = le32_to_cpu(es->s_first_data_block);
3657 
3658 		/*
3659 		 * Add the overhead attributed to the superblock and
3660 		 * block group descriptors.  If the sparse superblocks
3661 		 * feature is turned on, then not all groups have this.
3662 		 */
3663 		for (i = 0; i < ngroups; i++) {
3664 			overhead += ext4_bg_has_super(sb, i) +
3665 				ext4_bg_num_gdb(sb, i);
3666 			cond_resched();
3667 		}
3668 
3669 		/*
3670 		 * Every block group has an inode bitmap, a block
3671 		 * bitmap, and an inode table.
3672 		 */
3673 		overhead += ngroups * (2 + sbi->s_itb_per_group);
3674 		sbi->s_overhead_last = overhead;
3675 		smp_wmb();
3676 		sbi->s_blocks_last = ext4_blocks_count(es);
3677 	}
3678 
3679 	buf->f_type = EXT4_SUPER_MAGIC;
3680 	buf->f_bsize = sb->s_blocksize;
3681 	buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3682 	buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3683 		       percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3684 	buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3685 	if (buf->f_bfree < ext4_r_blocks_count(es))
3686 		buf->f_bavail = 0;
3687 	buf->f_files = le32_to_cpu(es->s_inodes_count);
3688 	buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3689 	buf->f_namelen = EXT4_NAME_LEN;
3690 	fsid = le64_to_cpup((void *)es->s_uuid) ^
3691 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3692 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3693 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3694 
3695 	return 0;
3696 }
3697 
3698 /* Helper function for writing quotas on sync - we need to start transaction
3699  * before quota file is locked for write. Otherwise the are possible deadlocks:
3700  * Process 1                         Process 2
3701  * ext4_create()                     quota_sync()
3702  *   jbd2_journal_start()                  write_dquot()
3703  *   vfs_dq_init()                         down(dqio_mutex)
3704  *     down(dqio_mutex)                    jbd2_journal_start()
3705  *
3706  */
3707 
3708 #ifdef CONFIG_QUOTA
3709 
3710 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3711 {
3712 	return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3713 }
3714 
3715 static int ext4_write_dquot(struct dquot *dquot)
3716 {
3717 	int ret, err;
3718 	handle_t *handle;
3719 	struct inode *inode;
3720 
3721 	inode = dquot_to_inode(dquot);
3722 	handle = ext4_journal_start(inode,
3723 				    EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3724 	if (IS_ERR(handle))
3725 		return PTR_ERR(handle);
3726 	ret = dquot_commit(dquot);
3727 	err = ext4_journal_stop(handle);
3728 	if (!ret)
3729 		ret = err;
3730 	return ret;
3731 }
3732 
3733 static int ext4_acquire_dquot(struct dquot *dquot)
3734 {
3735 	int ret, err;
3736 	handle_t *handle;
3737 
3738 	handle = ext4_journal_start(dquot_to_inode(dquot),
3739 				    EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3740 	if (IS_ERR(handle))
3741 		return PTR_ERR(handle);
3742 	ret = dquot_acquire(dquot);
3743 	err = ext4_journal_stop(handle);
3744 	if (!ret)
3745 		ret = err;
3746 	return ret;
3747 }
3748 
3749 static int ext4_release_dquot(struct dquot *dquot)
3750 {
3751 	int ret, err;
3752 	handle_t *handle;
3753 
3754 	handle = ext4_journal_start(dquot_to_inode(dquot),
3755 				    EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3756 	if (IS_ERR(handle)) {
3757 		/* Release dquot anyway to avoid endless cycle in dqput() */
3758 		dquot_release(dquot);
3759 		return PTR_ERR(handle);
3760 	}
3761 	ret = dquot_release(dquot);
3762 	err = ext4_journal_stop(handle);
3763 	if (!ret)
3764 		ret = err;
3765 	return ret;
3766 }
3767 
3768 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3769 {
3770 	/* Are we journaling quotas? */
3771 	if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3772 	    EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3773 		dquot_mark_dquot_dirty(dquot);
3774 		return ext4_write_dquot(dquot);
3775 	} else {
3776 		return dquot_mark_dquot_dirty(dquot);
3777 	}
3778 }
3779 
3780 static int ext4_write_info(struct super_block *sb, int type)
3781 {
3782 	int ret, err;
3783 	handle_t *handle;
3784 
3785 	/* Data block + inode block */
3786 	handle = ext4_journal_start(sb->s_root->d_inode, 2);
3787 	if (IS_ERR(handle))
3788 		return PTR_ERR(handle);
3789 	ret = dquot_commit_info(sb, type);
3790 	err = ext4_journal_stop(handle);
3791 	if (!ret)
3792 		ret = err;
3793 	return ret;
3794 }
3795 
3796 /*
3797  * Turn on quotas during mount time - we need to find
3798  * the quota file and such...
3799  */
3800 static int ext4_quota_on_mount(struct super_block *sb, int type)
3801 {
3802 	return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3803 				  EXT4_SB(sb)->s_jquota_fmt, type);
3804 }
3805 
3806 /*
3807  * Standard function to be called on quota_on
3808  */
3809 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3810 			 char *name, int remount)
3811 {
3812 	int err;
3813 	struct path path;
3814 
3815 	if (!test_opt(sb, QUOTA))
3816 		return -EINVAL;
3817 	/* When remounting, no checks are needed and in fact, name is NULL */
3818 	if (remount)
3819 		return vfs_quota_on(sb, type, format_id, name, remount);
3820 
3821 	err = kern_path(name, LOOKUP_FOLLOW, &path);
3822 	if (err)
3823 		return err;
3824 
3825 	/* Quotafile not on the same filesystem? */
3826 	if (path.mnt->mnt_sb != sb) {
3827 		path_put(&path);
3828 		return -EXDEV;
3829 	}
3830 	/* Journaling quota? */
3831 	if (EXT4_SB(sb)->s_qf_names[type]) {
3832 		/* Quotafile not in fs root? */
3833 		if (path.dentry->d_parent != sb->s_root)
3834 			ext4_msg(sb, KERN_WARNING,
3835 				"Quota file not on filesystem root. "
3836 				"Journaled quota will not work");
3837 	}
3838 
3839 	/*
3840 	 * When we journal data on quota file, we have to flush journal to see
3841 	 * all updates to the file when we bypass pagecache...
3842 	 */
3843 	if (EXT4_SB(sb)->s_journal &&
3844 	    ext4_should_journal_data(path.dentry->d_inode)) {
3845 		/*
3846 		 * We don't need to lock updates but journal_flush() could
3847 		 * otherwise be livelocked...
3848 		 */
3849 		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3850 		err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3851 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3852 		if (err) {
3853 			path_put(&path);
3854 			return err;
3855 		}
3856 	}
3857 
3858 	err = vfs_quota_on_path(sb, type, format_id, &path);
3859 	path_put(&path);
3860 	return err;
3861 }
3862 
3863 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3864  * acquiring the locks... As quota files are never truncated and quota code
3865  * itself serializes the operations (and noone else should touch the files)
3866  * we don't have to be afraid of races */
3867 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3868 			       size_t len, loff_t off)
3869 {
3870 	struct inode *inode = sb_dqopt(sb)->files[type];
3871 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3872 	int err = 0;
3873 	int offset = off & (sb->s_blocksize - 1);
3874 	int tocopy;
3875 	size_t toread;
3876 	struct buffer_head *bh;
3877 	loff_t i_size = i_size_read(inode);
3878 
3879 	if (off > i_size)
3880 		return 0;
3881 	if (off+len > i_size)
3882 		len = i_size-off;
3883 	toread = len;
3884 	while (toread > 0) {
3885 		tocopy = sb->s_blocksize - offset < toread ?
3886 				sb->s_blocksize - offset : toread;
3887 		bh = ext4_bread(NULL, inode, blk, 0, &err);
3888 		if (err)
3889 			return err;
3890 		if (!bh)	/* A hole? */
3891 			memset(data, 0, tocopy);
3892 		else
3893 			memcpy(data, bh->b_data+offset, tocopy);
3894 		brelse(bh);
3895 		offset = 0;
3896 		toread -= tocopy;
3897 		data += tocopy;
3898 		blk++;
3899 	}
3900 	return len;
3901 }
3902 
3903 /* Write to quotafile (we know the transaction is already started and has
3904  * enough credits) */
3905 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3906 				const char *data, size_t len, loff_t off)
3907 {
3908 	struct inode *inode = sb_dqopt(sb)->files[type];
3909 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3910 	int err = 0;
3911 	int offset = off & (sb->s_blocksize - 1);
3912 	int tocopy;
3913 	int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3914 	size_t towrite = len;
3915 	struct buffer_head *bh;
3916 	handle_t *handle = journal_current_handle();
3917 
3918 	if (EXT4_SB(sb)->s_journal && !handle) {
3919 		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3920 			" cancelled because transaction is not started",
3921 			(unsigned long long)off, (unsigned long long)len);
3922 		return -EIO;
3923 	}
3924 	mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3925 	while (towrite > 0) {
3926 		tocopy = sb->s_blocksize - offset < towrite ?
3927 				sb->s_blocksize - offset : towrite;
3928 		bh = ext4_bread(handle, inode, blk, 1, &err);
3929 		if (!bh)
3930 			goto out;
3931 		if (journal_quota) {
3932 			err = ext4_journal_get_write_access(handle, bh);
3933 			if (err) {
3934 				brelse(bh);
3935 				goto out;
3936 			}
3937 		}
3938 		lock_buffer(bh);
3939 		memcpy(bh->b_data+offset, data, tocopy);
3940 		flush_dcache_page(bh->b_page);
3941 		unlock_buffer(bh);
3942 		if (journal_quota)
3943 			err = ext4_handle_dirty_metadata(handle, NULL, bh);
3944 		else {
3945 			/* Always do at least ordered writes for quotas */
3946 			err = ext4_jbd2_file_inode(handle, inode);
3947 			mark_buffer_dirty(bh);
3948 		}
3949 		brelse(bh);
3950 		if (err)
3951 			goto out;
3952 		offset = 0;
3953 		towrite -= tocopy;
3954 		data += tocopy;
3955 		blk++;
3956 	}
3957 out:
3958 	if (len == towrite) {
3959 		mutex_unlock(&inode->i_mutex);
3960 		return err;
3961 	}
3962 	if (inode->i_size < off+len-towrite) {
3963 		i_size_write(inode, off+len-towrite);
3964 		EXT4_I(inode)->i_disksize = inode->i_size;
3965 	}
3966 	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3967 	ext4_mark_inode_dirty(handle, inode);
3968 	mutex_unlock(&inode->i_mutex);
3969 	return len - towrite;
3970 }
3971 
3972 #endif
3973 
3974 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3975 		       const char *dev_name, void *data, struct vfsmount *mnt)
3976 {
3977 	return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3978 }
3979 
3980 #if !defined(CONTIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
3981 static struct file_system_type ext2_fs_type = {
3982 	.owner		= THIS_MODULE,
3983 	.name		= "ext2",
3984 	.get_sb		= ext4_get_sb,
3985 	.kill_sb	= kill_block_super,
3986 	.fs_flags	= FS_REQUIRES_DEV,
3987 };
3988 
3989 static inline void register_as_ext2(void)
3990 {
3991 	int err = register_filesystem(&ext2_fs_type);
3992 	if (err)
3993 		printk(KERN_WARNING
3994 		       "EXT4-fs: Unable to register as ext2 (%d)\n", err);
3995 }
3996 
3997 static inline void unregister_as_ext2(void)
3998 {
3999 	unregister_filesystem(&ext2_fs_type);
4000 }
4001 MODULE_ALIAS("ext2");
4002 #else
4003 static inline void register_as_ext2(void) { }
4004 static inline void unregister_as_ext2(void) { }
4005 #endif
4006 
4007 #if !defined(CONTIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4008 static struct file_system_type ext3_fs_type = {
4009 	.owner		= THIS_MODULE,
4010 	.name		= "ext3",
4011 	.get_sb		= ext4_get_sb,
4012 	.kill_sb	= kill_block_super,
4013 	.fs_flags	= FS_REQUIRES_DEV,
4014 };
4015 
4016 static inline void register_as_ext3(void)
4017 {
4018 	int err = register_filesystem(&ext3_fs_type);
4019 	if (err)
4020 		printk(KERN_WARNING
4021 		       "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4022 }
4023 
4024 static inline void unregister_as_ext3(void)
4025 {
4026 	unregister_filesystem(&ext3_fs_type);
4027 }
4028 MODULE_ALIAS("ext3");
4029 #else
4030 static inline void register_as_ext3(void) { }
4031 static inline void unregister_as_ext3(void) { }
4032 #endif
4033 
4034 static struct file_system_type ext4_fs_type = {
4035 	.owner		= THIS_MODULE,
4036 	.name		= "ext4",
4037 	.get_sb		= ext4_get_sb,
4038 	.kill_sb	= kill_block_super,
4039 	.fs_flags	= FS_REQUIRES_DEV,
4040 };
4041 
4042 static int __init init_ext4_fs(void)
4043 {
4044 	int err;
4045 
4046 	err = init_ext4_system_zone();
4047 	if (err)
4048 		return err;
4049 	ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4050 	if (!ext4_kset)
4051 		goto out4;
4052 	ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4053 	err = init_ext4_mballoc();
4054 	if (err)
4055 		goto out3;
4056 
4057 	err = init_ext4_xattr();
4058 	if (err)
4059 		goto out2;
4060 	err = init_inodecache();
4061 	if (err)
4062 		goto out1;
4063 	register_as_ext2();
4064 	register_as_ext3();
4065 	err = register_filesystem(&ext4_fs_type);
4066 	if (err)
4067 		goto out;
4068 	return 0;
4069 out:
4070 	unregister_as_ext2();
4071 	unregister_as_ext3();
4072 	destroy_inodecache();
4073 out1:
4074 	exit_ext4_xattr();
4075 out2:
4076 	exit_ext4_mballoc();
4077 out3:
4078 	remove_proc_entry("fs/ext4", NULL);
4079 	kset_unregister(ext4_kset);
4080 out4:
4081 	exit_ext4_system_zone();
4082 	return err;
4083 }
4084 
4085 static void __exit exit_ext4_fs(void)
4086 {
4087 	unregister_as_ext2();
4088 	unregister_as_ext3();
4089 	unregister_filesystem(&ext4_fs_type);
4090 	destroy_inodecache();
4091 	exit_ext4_xattr();
4092 	exit_ext4_mballoc();
4093 	remove_proc_entry("fs/ext4", NULL);
4094 	kset_unregister(ext4_kset);
4095 	exit_ext4_system_zone();
4096 }
4097 
4098 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4099 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4100 MODULE_LICENSE("GPL");
4101 module_init(init_ext4_fs)
4102 module_exit(exit_ext4_fs)
4103