xref: /openbmc/linux/fs/ext4/super.c (revision 78c99ba1)
1 /*
2  *  linux/fs/ext4/super.c
3  *
4  * Copyright (C) 1992, 1993, 1994, 1995
5  * Remy Card (card@masi.ibp.fr)
6  * Laboratoire MASI - Institut Blaise Pascal
7  * Universite Pierre et Marie Curie (Paris VI)
8  *
9  *  from
10  *
11  *  linux/fs/minix/inode.c
12  *
13  *  Copyright (C) 1991, 1992  Linus Torvalds
14  *
15  *  Big-endian to little-endian byte-swapping/bitmaps by
16  *        David S. Miller (davem@caip.rutgers.edu), 1995
17  */
18 
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/smp_lock.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/proc_fs.h>
39 #include <linux/ctype.h>
40 #include <linux/marker.h>
41 #include <linux/log2.h>
42 #include <linux/crc16.h>
43 #include <asm/uaccess.h>
44 
45 #include "ext4.h"
46 #include "ext4_jbd2.h"
47 #include "xattr.h"
48 #include "acl.h"
49 
50 static int default_mb_history_length = 1000;
51 
52 module_param_named(default_mb_history_length, default_mb_history_length,
53 		   int, 0644);
54 MODULE_PARM_DESC(default_mb_history_length,
55 		 "Default number of entries saved for mb_history");
56 
57 struct proc_dir_entry *ext4_proc_root;
58 static struct kset *ext4_kset;
59 
60 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
61 			     unsigned long journal_devnum);
62 static int ext4_commit_super(struct super_block *sb, int sync);
63 static void ext4_mark_recovery_complete(struct super_block *sb,
64 					struct ext4_super_block *es);
65 static void ext4_clear_journal_err(struct super_block *sb,
66 				   struct ext4_super_block *es);
67 static int ext4_sync_fs(struct super_block *sb, int wait);
68 static const char *ext4_decode_error(struct super_block *sb, int errno,
69 				     char nbuf[16]);
70 static int ext4_remount(struct super_block *sb, int *flags, char *data);
71 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
72 static int ext4_unfreeze(struct super_block *sb);
73 static void ext4_write_super(struct super_block *sb);
74 static int ext4_freeze(struct super_block *sb);
75 
76 
77 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
78 			       struct ext4_group_desc *bg)
79 {
80 	return le32_to_cpu(bg->bg_block_bitmap_lo) |
81 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
82 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
83 }
84 
85 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
86 			       struct ext4_group_desc *bg)
87 {
88 	return le32_to_cpu(bg->bg_inode_bitmap_lo) |
89 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
90 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
91 }
92 
93 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
94 			      struct ext4_group_desc *bg)
95 {
96 	return le32_to_cpu(bg->bg_inode_table_lo) |
97 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
98 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
99 }
100 
101 __u32 ext4_free_blks_count(struct super_block *sb,
102 			      struct ext4_group_desc *bg)
103 {
104 	return le16_to_cpu(bg->bg_free_blocks_count_lo) |
105 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
106 		 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
107 }
108 
109 __u32 ext4_free_inodes_count(struct super_block *sb,
110 			      struct ext4_group_desc *bg)
111 {
112 	return le16_to_cpu(bg->bg_free_inodes_count_lo) |
113 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
114 		 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
115 }
116 
117 __u32 ext4_used_dirs_count(struct super_block *sb,
118 			      struct ext4_group_desc *bg)
119 {
120 	return le16_to_cpu(bg->bg_used_dirs_count_lo) |
121 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
122 		 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
123 }
124 
125 __u32 ext4_itable_unused_count(struct super_block *sb,
126 			      struct ext4_group_desc *bg)
127 {
128 	return le16_to_cpu(bg->bg_itable_unused_lo) |
129 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
130 		 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
131 }
132 
133 void ext4_block_bitmap_set(struct super_block *sb,
134 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
135 {
136 	bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
137 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
138 		bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
139 }
140 
141 void ext4_inode_bitmap_set(struct super_block *sb,
142 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
143 {
144 	bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
145 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
146 		bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
147 }
148 
149 void ext4_inode_table_set(struct super_block *sb,
150 			  struct ext4_group_desc *bg, ext4_fsblk_t blk)
151 {
152 	bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
153 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
154 		bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
155 }
156 
157 void ext4_free_blks_set(struct super_block *sb,
158 			  struct ext4_group_desc *bg, __u32 count)
159 {
160 	bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
161 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
162 		bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
163 }
164 
165 void ext4_free_inodes_set(struct super_block *sb,
166 			  struct ext4_group_desc *bg, __u32 count)
167 {
168 	bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
169 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
170 		bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
171 }
172 
173 void ext4_used_dirs_set(struct super_block *sb,
174 			  struct ext4_group_desc *bg, __u32 count)
175 {
176 	bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
177 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
178 		bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
179 }
180 
181 void ext4_itable_unused_set(struct super_block *sb,
182 			  struct ext4_group_desc *bg, __u32 count)
183 {
184 	bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
185 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
186 		bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
187 }
188 
189 /*
190  * Wrappers for jbd2_journal_start/end.
191  *
192  * The only special thing we need to do here is to make sure that all
193  * journal_end calls result in the superblock being marked dirty, so
194  * that sync() will call the filesystem's write_super callback if
195  * appropriate.
196  */
197 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
198 {
199 	journal_t *journal;
200 
201 	if (sb->s_flags & MS_RDONLY)
202 		return ERR_PTR(-EROFS);
203 
204 	/* Special case here: if the journal has aborted behind our
205 	 * backs (eg. EIO in the commit thread), then we still need to
206 	 * take the FS itself readonly cleanly. */
207 	journal = EXT4_SB(sb)->s_journal;
208 	if (journal) {
209 		if (is_journal_aborted(journal)) {
210 			ext4_abort(sb, __func__, "Detected aborted journal");
211 			return ERR_PTR(-EROFS);
212 		}
213 		return jbd2_journal_start(journal, nblocks);
214 	}
215 	/*
216 	 * We're not journaling, return the appropriate indication.
217 	 */
218 	current->journal_info = EXT4_NOJOURNAL_HANDLE;
219 	return current->journal_info;
220 }
221 
222 /*
223  * The only special thing we need to do here is to make sure that all
224  * jbd2_journal_stop calls result in the superblock being marked dirty, so
225  * that sync() will call the filesystem's write_super callback if
226  * appropriate.
227  */
228 int __ext4_journal_stop(const char *where, handle_t *handle)
229 {
230 	struct super_block *sb;
231 	int err;
232 	int rc;
233 
234 	if (!ext4_handle_valid(handle)) {
235 		/*
236 		 * Do this here since we don't call jbd2_journal_stop() in
237 		 * no-journal mode.
238 		 */
239 		current->journal_info = NULL;
240 		return 0;
241 	}
242 	sb = handle->h_transaction->t_journal->j_private;
243 	err = handle->h_err;
244 	rc = jbd2_journal_stop(handle);
245 
246 	if (!err)
247 		err = rc;
248 	if (err)
249 		__ext4_std_error(sb, where, err);
250 	return err;
251 }
252 
253 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
254 		struct buffer_head *bh, handle_t *handle, int err)
255 {
256 	char nbuf[16];
257 	const char *errstr = ext4_decode_error(NULL, err, nbuf);
258 
259 	BUG_ON(!ext4_handle_valid(handle));
260 
261 	if (bh)
262 		BUFFER_TRACE(bh, "abort");
263 
264 	if (!handle->h_err)
265 		handle->h_err = err;
266 
267 	if (is_handle_aborted(handle))
268 		return;
269 
270 	printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
271 	       caller, errstr, err_fn);
272 
273 	jbd2_journal_abort_handle(handle);
274 }
275 
276 /* Deal with the reporting of failure conditions on a filesystem such as
277  * inconsistencies detected or read IO failures.
278  *
279  * On ext2, we can store the error state of the filesystem in the
280  * superblock.  That is not possible on ext4, because we may have other
281  * write ordering constraints on the superblock which prevent us from
282  * writing it out straight away; and given that the journal is about to
283  * be aborted, we can't rely on the current, or future, transactions to
284  * write out the superblock safely.
285  *
286  * We'll just use the jbd2_journal_abort() error code to record an error in
287  * the journal instead.  On recovery, the journal will compain about
288  * that error until we've noted it down and cleared it.
289  */
290 
291 static void ext4_handle_error(struct super_block *sb)
292 {
293 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
294 
295 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
296 	es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
297 
298 	if (sb->s_flags & MS_RDONLY)
299 		return;
300 
301 	if (!test_opt(sb, ERRORS_CONT)) {
302 		journal_t *journal = EXT4_SB(sb)->s_journal;
303 
304 		EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
305 		if (journal)
306 			jbd2_journal_abort(journal, -EIO);
307 	}
308 	if (test_opt(sb, ERRORS_RO)) {
309 		ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
310 		sb->s_flags |= MS_RDONLY;
311 	}
312 	ext4_commit_super(sb, 1);
313 	if (test_opt(sb, ERRORS_PANIC))
314 		panic("EXT4-fs (device %s): panic forced after error\n",
315 			sb->s_id);
316 }
317 
318 void ext4_error(struct super_block *sb, const char *function,
319 		const char *fmt, ...)
320 {
321 	va_list args;
322 
323 	va_start(args, fmt);
324 	printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
325 	vprintk(fmt, args);
326 	printk("\n");
327 	va_end(args);
328 
329 	ext4_handle_error(sb);
330 }
331 
332 static const char *ext4_decode_error(struct super_block *sb, int errno,
333 				     char nbuf[16])
334 {
335 	char *errstr = NULL;
336 
337 	switch (errno) {
338 	case -EIO:
339 		errstr = "IO failure";
340 		break;
341 	case -ENOMEM:
342 		errstr = "Out of memory";
343 		break;
344 	case -EROFS:
345 		if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
346 			errstr = "Journal has aborted";
347 		else
348 			errstr = "Readonly filesystem";
349 		break;
350 	default:
351 		/* If the caller passed in an extra buffer for unknown
352 		 * errors, textualise them now.  Else we just return
353 		 * NULL. */
354 		if (nbuf) {
355 			/* Check for truncated error codes... */
356 			if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
357 				errstr = nbuf;
358 		}
359 		break;
360 	}
361 
362 	return errstr;
363 }
364 
365 /* __ext4_std_error decodes expected errors from journaling functions
366  * automatically and invokes the appropriate error response.  */
367 
368 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
369 {
370 	char nbuf[16];
371 	const char *errstr;
372 
373 	/* Special case: if the error is EROFS, and we're not already
374 	 * inside a transaction, then there's really no point in logging
375 	 * an error. */
376 	if (errno == -EROFS && journal_current_handle() == NULL &&
377 	    (sb->s_flags & MS_RDONLY))
378 		return;
379 
380 	errstr = ext4_decode_error(sb, errno, nbuf);
381 	printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
382 	       sb->s_id, function, errstr);
383 
384 	ext4_handle_error(sb);
385 }
386 
387 /*
388  * ext4_abort is a much stronger failure handler than ext4_error.  The
389  * abort function may be used to deal with unrecoverable failures such
390  * as journal IO errors or ENOMEM at a critical moment in log management.
391  *
392  * We unconditionally force the filesystem into an ABORT|READONLY state,
393  * unless the error response on the fs has been set to panic in which
394  * case we take the easy way out and panic immediately.
395  */
396 
397 void ext4_abort(struct super_block *sb, const char *function,
398 		const char *fmt, ...)
399 {
400 	va_list args;
401 
402 	va_start(args, fmt);
403 	printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
404 	vprintk(fmt, args);
405 	printk("\n");
406 	va_end(args);
407 
408 	if (test_opt(sb, ERRORS_PANIC))
409 		panic("EXT4-fs panic from previous error\n");
410 
411 	if (sb->s_flags & MS_RDONLY)
412 		return;
413 
414 	ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
415 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
416 	sb->s_flags |= MS_RDONLY;
417 	EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
418 	if (EXT4_SB(sb)->s_journal)
419 		jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
420 }
421 
422 void ext4_msg (struct super_block * sb, const char *prefix,
423 		   const char *fmt, ...)
424 {
425 	va_list args;
426 
427 	va_start(args, fmt);
428 	printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
429 	vprintk(fmt, args);
430 	printk("\n");
431 	va_end(args);
432 }
433 
434 void ext4_warning(struct super_block *sb, const char *function,
435 		  const char *fmt, ...)
436 {
437 	va_list args;
438 
439 	va_start(args, fmt);
440 	printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
441 	       sb->s_id, function);
442 	vprintk(fmt, args);
443 	printk("\n");
444 	va_end(args);
445 }
446 
447 void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
448 			   const char *function, const char *fmt, ...)
449 __releases(bitlock)
450 __acquires(bitlock)
451 {
452 	va_list args;
453 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
454 
455 	va_start(args, fmt);
456 	printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
457 	vprintk(fmt, args);
458 	printk("\n");
459 	va_end(args);
460 
461 	if (test_opt(sb, ERRORS_CONT)) {
462 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
463 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
464 		ext4_commit_super(sb, 0);
465 		return;
466 	}
467 	ext4_unlock_group(sb, grp);
468 	ext4_handle_error(sb);
469 	/*
470 	 * We only get here in the ERRORS_RO case; relocking the group
471 	 * may be dangerous, but nothing bad will happen since the
472 	 * filesystem will have already been marked read/only and the
473 	 * journal has been aborted.  We return 1 as a hint to callers
474 	 * who might what to use the return value from
475 	 * ext4_grp_locked_error() to distinguish beween the
476 	 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
477 	 * aggressively from the ext4 function in question, with a
478 	 * more appropriate error code.
479 	 */
480 	ext4_lock_group(sb, grp);
481 	return;
482 }
483 
484 void ext4_update_dynamic_rev(struct super_block *sb)
485 {
486 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
487 
488 	if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
489 		return;
490 
491 	ext4_warning(sb, __func__,
492 		     "updating to rev %d because of new feature flag, "
493 		     "running e2fsck is recommended",
494 		     EXT4_DYNAMIC_REV);
495 
496 	es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
497 	es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
498 	es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
499 	/* leave es->s_feature_*compat flags alone */
500 	/* es->s_uuid will be set by e2fsck if empty */
501 
502 	/*
503 	 * The rest of the superblock fields should be zero, and if not it
504 	 * means they are likely already in use, so leave them alone.  We
505 	 * can leave it up to e2fsck to clean up any inconsistencies there.
506 	 */
507 }
508 
509 /*
510  * Open the external journal device
511  */
512 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
513 {
514 	struct block_device *bdev;
515 	char b[BDEVNAME_SIZE];
516 
517 	bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
518 	if (IS_ERR(bdev))
519 		goto fail;
520 	return bdev;
521 
522 fail:
523 	ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
524 			__bdevname(dev, b), PTR_ERR(bdev));
525 	return NULL;
526 }
527 
528 /*
529  * Release the journal device
530  */
531 static int ext4_blkdev_put(struct block_device *bdev)
532 {
533 	bd_release(bdev);
534 	return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
535 }
536 
537 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
538 {
539 	struct block_device *bdev;
540 	int ret = -ENODEV;
541 
542 	bdev = sbi->journal_bdev;
543 	if (bdev) {
544 		ret = ext4_blkdev_put(bdev);
545 		sbi->journal_bdev = NULL;
546 	}
547 	return ret;
548 }
549 
550 static inline struct inode *orphan_list_entry(struct list_head *l)
551 {
552 	return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
553 }
554 
555 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
556 {
557 	struct list_head *l;
558 
559 	ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
560 		 le32_to_cpu(sbi->s_es->s_last_orphan));
561 
562 	printk(KERN_ERR "sb_info orphan list:\n");
563 	list_for_each(l, &sbi->s_orphan) {
564 		struct inode *inode = orphan_list_entry(l);
565 		printk(KERN_ERR "  "
566 		       "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
567 		       inode->i_sb->s_id, inode->i_ino, inode,
568 		       inode->i_mode, inode->i_nlink,
569 		       NEXT_ORPHAN(inode));
570 	}
571 }
572 
573 static void ext4_put_super(struct super_block *sb)
574 {
575 	struct ext4_sb_info *sbi = EXT4_SB(sb);
576 	struct ext4_super_block *es = sbi->s_es;
577 	int i, err;
578 
579 	ext4_release_system_zone(sb);
580 	ext4_mb_release(sb);
581 	ext4_ext_release(sb);
582 	ext4_xattr_put_super(sb);
583 	if (sbi->s_journal) {
584 		err = jbd2_journal_destroy(sbi->s_journal);
585 		sbi->s_journal = NULL;
586 		if (err < 0)
587 			ext4_abort(sb, __func__,
588 				   "Couldn't clean up the journal");
589 	}
590 	if (!(sb->s_flags & MS_RDONLY)) {
591 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
592 		es->s_state = cpu_to_le16(sbi->s_mount_state);
593 		ext4_commit_super(sb, 1);
594 	}
595 	if (sbi->s_proc) {
596 		remove_proc_entry(sb->s_id, ext4_proc_root);
597 	}
598 	kobject_del(&sbi->s_kobj);
599 
600 	for (i = 0; i < sbi->s_gdb_count; i++)
601 		brelse(sbi->s_group_desc[i]);
602 	kfree(sbi->s_group_desc);
603 	if (is_vmalloc_addr(sbi->s_flex_groups))
604 		vfree(sbi->s_flex_groups);
605 	else
606 		kfree(sbi->s_flex_groups);
607 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
608 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
609 	percpu_counter_destroy(&sbi->s_dirs_counter);
610 	percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
611 	brelse(sbi->s_sbh);
612 #ifdef CONFIG_QUOTA
613 	for (i = 0; i < MAXQUOTAS; i++)
614 		kfree(sbi->s_qf_names[i]);
615 #endif
616 
617 	/* Debugging code just in case the in-memory inode orphan list
618 	 * isn't empty.  The on-disk one can be non-empty if we've
619 	 * detected an error and taken the fs readonly, but the
620 	 * in-memory list had better be clean by this point. */
621 	if (!list_empty(&sbi->s_orphan))
622 		dump_orphan_list(sb, sbi);
623 	J_ASSERT(list_empty(&sbi->s_orphan));
624 
625 	invalidate_bdev(sb->s_bdev);
626 	if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
627 		/*
628 		 * Invalidate the journal device's buffers.  We don't want them
629 		 * floating about in memory - the physical journal device may
630 		 * hotswapped, and it breaks the `ro-after' testing code.
631 		 */
632 		sync_blockdev(sbi->journal_bdev);
633 		invalidate_bdev(sbi->journal_bdev);
634 		ext4_blkdev_remove(sbi);
635 	}
636 	sb->s_fs_info = NULL;
637 	/*
638 	 * Now that we are completely done shutting down the
639 	 * superblock, we need to actually destroy the kobject.
640 	 */
641 	unlock_kernel();
642 	unlock_super(sb);
643 	kobject_put(&sbi->s_kobj);
644 	wait_for_completion(&sbi->s_kobj_unregister);
645 	lock_super(sb);
646 	lock_kernel();
647 	kfree(sbi->s_blockgroup_lock);
648 	kfree(sbi);
649 }
650 
651 static struct kmem_cache *ext4_inode_cachep;
652 
653 /*
654  * Called inside transaction, so use GFP_NOFS
655  */
656 static struct inode *ext4_alloc_inode(struct super_block *sb)
657 {
658 	struct ext4_inode_info *ei;
659 
660 	ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
661 	if (!ei)
662 		return NULL;
663 
664 #ifdef CONFIG_EXT4_FS_POSIX_ACL
665 	ei->i_acl = EXT4_ACL_NOT_CACHED;
666 	ei->i_default_acl = EXT4_ACL_NOT_CACHED;
667 #endif
668 	ei->vfs_inode.i_version = 1;
669 	ei->vfs_inode.i_data.writeback_index = 0;
670 	memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
671 	INIT_LIST_HEAD(&ei->i_prealloc_list);
672 	spin_lock_init(&ei->i_prealloc_lock);
673 	/*
674 	 * Note:  We can be called before EXT4_SB(sb)->s_journal is set,
675 	 * therefore it can be null here.  Don't check it, just initialize
676 	 * jinode.
677 	 */
678 	jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
679 	ei->i_reserved_data_blocks = 0;
680 	ei->i_reserved_meta_blocks = 0;
681 	ei->i_allocated_meta_blocks = 0;
682 	ei->i_delalloc_reserved_flag = 0;
683 	spin_lock_init(&(ei->i_block_reservation_lock));
684 
685 	return &ei->vfs_inode;
686 }
687 
688 static void ext4_destroy_inode(struct inode *inode)
689 {
690 	if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
691 		ext4_msg(inode->i_sb, KERN_ERR,
692 			 "Inode %lu (%p): orphan list check failed!",
693 			 inode->i_ino, EXT4_I(inode));
694 		print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
695 				EXT4_I(inode), sizeof(struct ext4_inode_info),
696 				true);
697 		dump_stack();
698 	}
699 	kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
700 }
701 
702 static void init_once(void *foo)
703 {
704 	struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
705 
706 	INIT_LIST_HEAD(&ei->i_orphan);
707 #ifdef CONFIG_EXT4_FS_XATTR
708 	init_rwsem(&ei->xattr_sem);
709 #endif
710 	init_rwsem(&ei->i_data_sem);
711 	inode_init_once(&ei->vfs_inode);
712 }
713 
714 static int init_inodecache(void)
715 {
716 	ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
717 					     sizeof(struct ext4_inode_info),
718 					     0, (SLAB_RECLAIM_ACCOUNT|
719 						SLAB_MEM_SPREAD),
720 					     init_once);
721 	if (ext4_inode_cachep == NULL)
722 		return -ENOMEM;
723 	return 0;
724 }
725 
726 static void destroy_inodecache(void)
727 {
728 	kmem_cache_destroy(ext4_inode_cachep);
729 }
730 
731 static void ext4_clear_inode(struct inode *inode)
732 {
733 #ifdef CONFIG_EXT4_FS_POSIX_ACL
734 	if (EXT4_I(inode)->i_acl &&
735 			EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
736 		posix_acl_release(EXT4_I(inode)->i_acl);
737 		EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
738 	}
739 	if (EXT4_I(inode)->i_default_acl &&
740 			EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
741 		posix_acl_release(EXT4_I(inode)->i_default_acl);
742 		EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
743 	}
744 #endif
745 	ext4_discard_preallocations(inode);
746 	if (EXT4_JOURNAL(inode))
747 		jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
748 				       &EXT4_I(inode)->jinode);
749 }
750 
751 static inline void ext4_show_quota_options(struct seq_file *seq,
752 					   struct super_block *sb)
753 {
754 #if defined(CONFIG_QUOTA)
755 	struct ext4_sb_info *sbi = EXT4_SB(sb);
756 
757 	if (sbi->s_jquota_fmt)
758 		seq_printf(seq, ",jqfmt=%s",
759 		(sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
760 
761 	if (sbi->s_qf_names[USRQUOTA])
762 		seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
763 
764 	if (sbi->s_qf_names[GRPQUOTA])
765 		seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
766 
767 	if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
768 		seq_puts(seq, ",usrquota");
769 
770 	if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
771 		seq_puts(seq, ",grpquota");
772 #endif
773 }
774 
775 /*
776  * Show an option if
777  *  - it's set to a non-default value OR
778  *  - if the per-sb default is different from the global default
779  */
780 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
781 {
782 	int def_errors;
783 	unsigned long def_mount_opts;
784 	struct super_block *sb = vfs->mnt_sb;
785 	struct ext4_sb_info *sbi = EXT4_SB(sb);
786 	struct ext4_super_block *es = sbi->s_es;
787 
788 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
789 	def_errors     = le16_to_cpu(es->s_errors);
790 
791 	if (sbi->s_sb_block != 1)
792 		seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
793 	if (test_opt(sb, MINIX_DF))
794 		seq_puts(seq, ",minixdf");
795 	if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
796 		seq_puts(seq, ",grpid");
797 	if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
798 		seq_puts(seq, ",nogrpid");
799 	if (sbi->s_resuid != EXT4_DEF_RESUID ||
800 	    le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
801 		seq_printf(seq, ",resuid=%u", sbi->s_resuid);
802 	}
803 	if (sbi->s_resgid != EXT4_DEF_RESGID ||
804 	    le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
805 		seq_printf(seq, ",resgid=%u", sbi->s_resgid);
806 	}
807 	if (test_opt(sb, ERRORS_RO)) {
808 		if (def_errors == EXT4_ERRORS_PANIC ||
809 		    def_errors == EXT4_ERRORS_CONTINUE) {
810 			seq_puts(seq, ",errors=remount-ro");
811 		}
812 	}
813 	if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
814 		seq_puts(seq, ",errors=continue");
815 	if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
816 		seq_puts(seq, ",errors=panic");
817 	if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
818 		seq_puts(seq, ",nouid32");
819 	if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
820 		seq_puts(seq, ",debug");
821 	if (test_opt(sb, OLDALLOC))
822 		seq_puts(seq, ",oldalloc");
823 #ifdef CONFIG_EXT4_FS_XATTR
824 	if (test_opt(sb, XATTR_USER) &&
825 		!(def_mount_opts & EXT4_DEFM_XATTR_USER))
826 		seq_puts(seq, ",user_xattr");
827 	if (!test_opt(sb, XATTR_USER) &&
828 	    (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
829 		seq_puts(seq, ",nouser_xattr");
830 	}
831 #endif
832 #ifdef CONFIG_EXT4_FS_POSIX_ACL
833 	if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
834 		seq_puts(seq, ",acl");
835 	if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
836 		seq_puts(seq, ",noacl");
837 #endif
838 	if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
839 		seq_printf(seq, ",commit=%u",
840 			   (unsigned) (sbi->s_commit_interval / HZ));
841 	}
842 	if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
843 		seq_printf(seq, ",min_batch_time=%u",
844 			   (unsigned) sbi->s_min_batch_time);
845 	}
846 	if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
847 		seq_printf(seq, ",max_batch_time=%u",
848 			   (unsigned) sbi->s_min_batch_time);
849 	}
850 
851 	/*
852 	 * We're changing the default of barrier mount option, so
853 	 * let's always display its mount state so it's clear what its
854 	 * status is.
855 	 */
856 	seq_puts(seq, ",barrier=");
857 	seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
858 	if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
859 		seq_puts(seq, ",journal_async_commit");
860 	if (test_opt(sb, NOBH))
861 		seq_puts(seq, ",nobh");
862 	if (test_opt(sb, I_VERSION))
863 		seq_puts(seq, ",i_version");
864 	if (!test_opt(sb, DELALLOC))
865 		seq_puts(seq, ",nodelalloc");
866 
867 
868 	if (sbi->s_stripe)
869 		seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
870 	/*
871 	 * journal mode get enabled in different ways
872 	 * So just print the value even if we didn't specify it
873 	 */
874 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
875 		seq_puts(seq, ",data=journal");
876 	else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
877 		seq_puts(seq, ",data=ordered");
878 	else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
879 		seq_puts(seq, ",data=writeback");
880 
881 	if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
882 		seq_printf(seq, ",inode_readahead_blks=%u",
883 			   sbi->s_inode_readahead_blks);
884 
885 	if (test_opt(sb, DATA_ERR_ABORT))
886 		seq_puts(seq, ",data_err=abort");
887 
888 	if (test_opt(sb, NO_AUTO_DA_ALLOC))
889 		seq_puts(seq, ",noauto_da_alloc");
890 
891 	ext4_show_quota_options(seq, sb);
892 
893 	return 0;
894 }
895 
896 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
897 					u64 ino, u32 generation)
898 {
899 	struct inode *inode;
900 
901 	if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
902 		return ERR_PTR(-ESTALE);
903 	if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
904 		return ERR_PTR(-ESTALE);
905 
906 	/* iget isn't really right if the inode is currently unallocated!!
907 	 *
908 	 * ext4_read_inode will return a bad_inode if the inode had been
909 	 * deleted, so we should be safe.
910 	 *
911 	 * Currently we don't know the generation for parent directory, so
912 	 * a generation of 0 means "accept any"
913 	 */
914 	inode = ext4_iget(sb, ino);
915 	if (IS_ERR(inode))
916 		return ERR_CAST(inode);
917 	if (generation && inode->i_generation != generation) {
918 		iput(inode);
919 		return ERR_PTR(-ESTALE);
920 	}
921 
922 	return inode;
923 }
924 
925 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
926 					int fh_len, int fh_type)
927 {
928 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
929 				    ext4_nfs_get_inode);
930 }
931 
932 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
933 					int fh_len, int fh_type)
934 {
935 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
936 				    ext4_nfs_get_inode);
937 }
938 
939 /*
940  * Try to release metadata pages (indirect blocks, directories) which are
941  * mapped via the block device.  Since these pages could have journal heads
942  * which would prevent try_to_free_buffers() from freeing them, we must use
943  * jbd2 layer's try_to_free_buffers() function to release them.
944  */
945 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
946 				 gfp_t wait)
947 {
948 	journal_t *journal = EXT4_SB(sb)->s_journal;
949 
950 	WARN_ON(PageChecked(page));
951 	if (!page_has_buffers(page))
952 		return 0;
953 	if (journal)
954 		return jbd2_journal_try_to_free_buffers(journal, page,
955 							wait & ~__GFP_WAIT);
956 	return try_to_free_buffers(page);
957 }
958 
959 #ifdef CONFIG_QUOTA
960 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
961 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
962 
963 static int ext4_write_dquot(struct dquot *dquot);
964 static int ext4_acquire_dquot(struct dquot *dquot);
965 static int ext4_release_dquot(struct dquot *dquot);
966 static int ext4_mark_dquot_dirty(struct dquot *dquot);
967 static int ext4_write_info(struct super_block *sb, int type);
968 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
969 				char *path, int remount);
970 static int ext4_quota_on_mount(struct super_block *sb, int type);
971 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
972 			       size_t len, loff_t off);
973 static ssize_t ext4_quota_write(struct super_block *sb, int type,
974 				const char *data, size_t len, loff_t off);
975 
976 static struct dquot_operations ext4_quota_operations = {
977 	.initialize	= dquot_initialize,
978 	.drop		= dquot_drop,
979 	.alloc_space	= dquot_alloc_space,
980 	.reserve_space	= dquot_reserve_space,
981 	.claim_space	= dquot_claim_space,
982 	.release_rsv	= dquot_release_reserved_space,
983 	.get_reserved_space = ext4_get_reserved_space,
984 	.alloc_inode	= dquot_alloc_inode,
985 	.free_space	= dquot_free_space,
986 	.free_inode	= dquot_free_inode,
987 	.transfer	= dquot_transfer,
988 	.write_dquot	= ext4_write_dquot,
989 	.acquire_dquot	= ext4_acquire_dquot,
990 	.release_dquot	= ext4_release_dquot,
991 	.mark_dirty	= ext4_mark_dquot_dirty,
992 	.write_info	= ext4_write_info,
993 	.alloc_dquot	= dquot_alloc,
994 	.destroy_dquot	= dquot_destroy,
995 };
996 
997 static struct quotactl_ops ext4_qctl_operations = {
998 	.quota_on	= ext4_quota_on,
999 	.quota_off	= vfs_quota_off,
1000 	.quota_sync	= vfs_quota_sync,
1001 	.get_info	= vfs_get_dqinfo,
1002 	.set_info	= vfs_set_dqinfo,
1003 	.get_dqblk	= vfs_get_dqblk,
1004 	.set_dqblk	= vfs_set_dqblk
1005 };
1006 #endif
1007 
1008 static const struct super_operations ext4_sops = {
1009 	.alloc_inode	= ext4_alloc_inode,
1010 	.destroy_inode	= ext4_destroy_inode,
1011 	.write_inode	= ext4_write_inode,
1012 	.dirty_inode	= ext4_dirty_inode,
1013 	.delete_inode	= ext4_delete_inode,
1014 	.put_super	= ext4_put_super,
1015 	.sync_fs	= ext4_sync_fs,
1016 	.freeze_fs	= ext4_freeze,
1017 	.unfreeze_fs	= ext4_unfreeze,
1018 	.statfs		= ext4_statfs,
1019 	.remount_fs	= ext4_remount,
1020 	.clear_inode	= ext4_clear_inode,
1021 	.show_options	= ext4_show_options,
1022 #ifdef CONFIG_QUOTA
1023 	.quota_read	= ext4_quota_read,
1024 	.quota_write	= ext4_quota_write,
1025 #endif
1026 	.bdev_try_to_free_page = bdev_try_to_free_page,
1027 };
1028 
1029 static const struct super_operations ext4_nojournal_sops = {
1030 	.alloc_inode	= ext4_alloc_inode,
1031 	.destroy_inode	= ext4_destroy_inode,
1032 	.write_inode	= ext4_write_inode,
1033 	.dirty_inode	= ext4_dirty_inode,
1034 	.delete_inode	= ext4_delete_inode,
1035 	.write_super	= ext4_write_super,
1036 	.put_super	= ext4_put_super,
1037 	.statfs		= ext4_statfs,
1038 	.remount_fs	= ext4_remount,
1039 	.clear_inode	= ext4_clear_inode,
1040 	.show_options	= ext4_show_options,
1041 #ifdef CONFIG_QUOTA
1042 	.quota_read	= ext4_quota_read,
1043 	.quota_write	= ext4_quota_write,
1044 #endif
1045 	.bdev_try_to_free_page = bdev_try_to_free_page,
1046 };
1047 
1048 static const struct export_operations ext4_export_ops = {
1049 	.fh_to_dentry = ext4_fh_to_dentry,
1050 	.fh_to_parent = ext4_fh_to_parent,
1051 	.get_parent = ext4_get_parent,
1052 };
1053 
1054 enum {
1055 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1056 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1057 	Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1058 	Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1059 	Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1060 	Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1061 	Opt_journal_update, Opt_journal_dev,
1062 	Opt_journal_checksum, Opt_journal_async_commit,
1063 	Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1064 	Opt_data_err_abort, Opt_data_err_ignore, Opt_mb_history_length,
1065 	Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1066 	Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1067 	Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1068 	Opt_usrquota, Opt_grpquota, Opt_i_version,
1069 	Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1070 	Opt_block_validity, Opt_noblock_validity,
1071 	Opt_inode_readahead_blks, Opt_journal_ioprio
1072 };
1073 
1074 static const match_table_t tokens = {
1075 	{Opt_bsd_df, "bsddf"},
1076 	{Opt_minix_df, "minixdf"},
1077 	{Opt_grpid, "grpid"},
1078 	{Opt_grpid, "bsdgroups"},
1079 	{Opt_nogrpid, "nogrpid"},
1080 	{Opt_nogrpid, "sysvgroups"},
1081 	{Opt_resgid, "resgid=%u"},
1082 	{Opt_resuid, "resuid=%u"},
1083 	{Opt_sb, "sb=%u"},
1084 	{Opt_err_cont, "errors=continue"},
1085 	{Opt_err_panic, "errors=panic"},
1086 	{Opt_err_ro, "errors=remount-ro"},
1087 	{Opt_nouid32, "nouid32"},
1088 	{Opt_debug, "debug"},
1089 	{Opt_oldalloc, "oldalloc"},
1090 	{Opt_orlov, "orlov"},
1091 	{Opt_user_xattr, "user_xattr"},
1092 	{Opt_nouser_xattr, "nouser_xattr"},
1093 	{Opt_acl, "acl"},
1094 	{Opt_noacl, "noacl"},
1095 	{Opt_noload, "noload"},
1096 	{Opt_nobh, "nobh"},
1097 	{Opt_bh, "bh"},
1098 	{Opt_commit, "commit=%u"},
1099 	{Opt_min_batch_time, "min_batch_time=%u"},
1100 	{Opt_max_batch_time, "max_batch_time=%u"},
1101 	{Opt_journal_update, "journal=update"},
1102 	{Opt_journal_dev, "journal_dev=%u"},
1103 	{Opt_journal_checksum, "journal_checksum"},
1104 	{Opt_journal_async_commit, "journal_async_commit"},
1105 	{Opt_abort, "abort"},
1106 	{Opt_data_journal, "data=journal"},
1107 	{Opt_data_ordered, "data=ordered"},
1108 	{Opt_data_writeback, "data=writeback"},
1109 	{Opt_data_err_abort, "data_err=abort"},
1110 	{Opt_data_err_ignore, "data_err=ignore"},
1111 	{Opt_mb_history_length, "mb_history_length=%u"},
1112 	{Opt_offusrjquota, "usrjquota="},
1113 	{Opt_usrjquota, "usrjquota=%s"},
1114 	{Opt_offgrpjquota, "grpjquota="},
1115 	{Opt_grpjquota, "grpjquota=%s"},
1116 	{Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1117 	{Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1118 	{Opt_grpquota, "grpquota"},
1119 	{Opt_noquota, "noquota"},
1120 	{Opt_quota, "quota"},
1121 	{Opt_usrquota, "usrquota"},
1122 	{Opt_barrier, "barrier=%u"},
1123 	{Opt_barrier, "barrier"},
1124 	{Opt_nobarrier, "nobarrier"},
1125 	{Opt_i_version, "i_version"},
1126 	{Opt_stripe, "stripe=%u"},
1127 	{Opt_resize, "resize"},
1128 	{Opt_delalloc, "delalloc"},
1129 	{Opt_nodelalloc, "nodelalloc"},
1130 	{Opt_block_validity, "block_validity"},
1131 	{Opt_noblock_validity, "noblock_validity"},
1132 	{Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1133 	{Opt_journal_ioprio, "journal_ioprio=%u"},
1134 	{Opt_auto_da_alloc, "auto_da_alloc=%u"},
1135 	{Opt_auto_da_alloc, "auto_da_alloc"},
1136 	{Opt_noauto_da_alloc, "noauto_da_alloc"},
1137 	{Opt_err, NULL},
1138 };
1139 
1140 static ext4_fsblk_t get_sb_block(void **data)
1141 {
1142 	ext4_fsblk_t	sb_block;
1143 	char		*options = (char *) *data;
1144 
1145 	if (!options || strncmp(options, "sb=", 3) != 0)
1146 		return 1;	/* Default location */
1147 
1148 	options += 3;
1149 	/* TODO: use simple_strtoll with >32bit ext4 */
1150 	sb_block = simple_strtoul(options, &options, 0);
1151 	if (*options && *options != ',') {
1152 		printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1153 		       (char *) *data);
1154 		return 1;
1155 	}
1156 	if (*options == ',')
1157 		options++;
1158 	*data = (void *) options;
1159 
1160 	return sb_block;
1161 }
1162 
1163 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1164 
1165 static int parse_options(char *options, struct super_block *sb,
1166 			 unsigned long *journal_devnum,
1167 			 unsigned int *journal_ioprio,
1168 			 ext4_fsblk_t *n_blocks_count, int is_remount)
1169 {
1170 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1171 	char *p;
1172 	substring_t args[MAX_OPT_ARGS];
1173 	int data_opt = 0;
1174 	int option;
1175 #ifdef CONFIG_QUOTA
1176 	int qtype, qfmt;
1177 	char *qname;
1178 #endif
1179 
1180 	if (!options)
1181 		return 1;
1182 
1183 	while ((p = strsep(&options, ",")) != NULL) {
1184 		int token;
1185 		if (!*p)
1186 			continue;
1187 
1188 		token = match_token(p, tokens, args);
1189 		switch (token) {
1190 		case Opt_bsd_df:
1191 			clear_opt(sbi->s_mount_opt, MINIX_DF);
1192 			break;
1193 		case Opt_minix_df:
1194 			set_opt(sbi->s_mount_opt, MINIX_DF);
1195 			break;
1196 		case Opt_grpid:
1197 			set_opt(sbi->s_mount_opt, GRPID);
1198 			break;
1199 		case Opt_nogrpid:
1200 			clear_opt(sbi->s_mount_opt, GRPID);
1201 			break;
1202 		case Opt_resuid:
1203 			if (match_int(&args[0], &option))
1204 				return 0;
1205 			sbi->s_resuid = option;
1206 			break;
1207 		case Opt_resgid:
1208 			if (match_int(&args[0], &option))
1209 				return 0;
1210 			sbi->s_resgid = option;
1211 			break;
1212 		case Opt_sb:
1213 			/* handled by get_sb_block() instead of here */
1214 			/* *sb_block = match_int(&args[0]); */
1215 			break;
1216 		case Opt_err_panic:
1217 			clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1218 			clear_opt(sbi->s_mount_opt, ERRORS_RO);
1219 			set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1220 			break;
1221 		case Opt_err_ro:
1222 			clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1223 			clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1224 			set_opt(sbi->s_mount_opt, ERRORS_RO);
1225 			break;
1226 		case Opt_err_cont:
1227 			clear_opt(sbi->s_mount_opt, ERRORS_RO);
1228 			clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1229 			set_opt(sbi->s_mount_opt, ERRORS_CONT);
1230 			break;
1231 		case Opt_nouid32:
1232 			set_opt(sbi->s_mount_opt, NO_UID32);
1233 			break;
1234 		case Opt_debug:
1235 			set_opt(sbi->s_mount_opt, DEBUG);
1236 			break;
1237 		case Opt_oldalloc:
1238 			set_opt(sbi->s_mount_opt, OLDALLOC);
1239 			break;
1240 		case Opt_orlov:
1241 			clear_opt(sbi->s_mount_opt, OLDALLOC);
1242 			break;
1243 #ifdef CONFIG_EXT4_FS_XATTR
1244 		case Opt_user_xattr:
1245 			set_opt(sbi->s_mount_opt, XATTR_USER);
1246 			break;
1247 		case Opt_nouser_xattr:
1248 			clear_opt(sbi->s_mount_opt, XATTR_USER);
1249 			break;
1250 #else
1251 		case Opt_user_xattr:
1252 		case Opt_nouser_xattr:
1253 			ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1254 			break;
1255 #endif
1256 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1257 		case Opt_acl:
1258 			set_opt(sbi->s_mount_opt, POSIX_ACL);
1259 			break;
1260 		case Opt_noacl:
1261 			clear_opt(sbi->s_mount_opt, POSIX_ACL);
1262 			break;
1263 #else
1264 		case Opt_acl:
1265 		case Opt_noacl:
1266 			ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1267 			break;
1268 #endif
1269 		case Opt_journal_update:
1270 			/* @@@ FIXME */
1271 			/* Eventually we will want to be able to create
1272 			   a journal file here.  For now, only allow the
1273 			   user to specify an existing inode to be the
1274 			   journal file. */
1275 			if (is_remount) {
1276 				ext4_msg(sb, KERN_ERR,
1277 					 "Cannot specify journal on remount");
1278 				return 0;
1279 			}
1280 			set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1281 			break;
1282 		case Opt_journal_dev:
1283 			if (is_remount) {
1284 				ext4_msg(sb, KERN_ERR,
1285 					"Cannot specify journal on remount");
1286 				return 0;
1287 			}
1288 			if (match_int(&args[0], &option))
1289 				return 0;
1290 			*journal_devnum = option;
1291 			break;
1292 		case Opt_journal_checksum:
1293 			set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1294 			break;
1295 		case Opt_journal_async_commit:
1296 			set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1297 			set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1298 			break;
1299 		case Opt_noload:
1300 			set_opt(sbi->s_mount_opt, NOLOAD);
1301 			break;
1302 		case Opt_commit:
1303 			if (match_int(&args[0], &option))
1304 				return 0;
1305 			if (option < 0)
1306 				return 0;
1307 			if (option == 0)
1308 				option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1309 			sbi->s_commit_interval = HZ * option;
1310 			break;
1311 		case Opt_max_batch_time:
1312 			if (match_int(&args[0], &option))
1313 				return 0;
1314 			if (option < 0)
1315 				return 0;
1316 			if (option == 0)
1317 				option = EXT4_DEF_MAX_BATCH_TIME;
1318 			sbi->s_max_batch_time = option;
1319 			break;
1320 		case Opt_min_batch_time:
1321 			if (match_int(&args[0], &option))
1322 				return 0;
1323 			if (option < 0)
1324 				return 0;
1325 			sbi->s_min_batch_time = option;
1326 			break;
1327 		case Opt_data_journal:
1328 			data_opt = EXT4_MOUNT_JOURNAL_DATA;
1329 			goto datacheck;
1330 		case Opt_data_ordered:
1331 			data_opt = EXT4_MOUNT_ORDERED_DATA;
1332 			goto datacheck;
1333 		case Opt_data_writeback:
1334 			data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1335 		datacheck:
1336 			if (is_remount) {
1337 				if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1338 						!= data_opt) {
1339 					ext4_msg(sb, KERN_ERR,
1340 						"Cannot change data mode on remount");
1341 					return 0;
1342 				}
1343 			} else {
1344 				sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1345 				sbi->s_mount_opt |= data_opt;
1346 			}
1347 			break;
1348 		case Opt_data_err_abort:
1349 			set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1350 			break;
1351 		case Opt_data_err_ignore:
1352 			clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1353 			break;
1354 		case Opt_mb_history_length:
1355 			if (match_int(&args[0], &option))
1356 				return 0;
1357 			if (option < 0)
1358 				return 0;
1359 			sbi->s_mb_history_max = option;
1360 			break;
1361 #ifdef CONFIG_QUOTA
1362 		case Opt_usrjquota:
1363 			qtype = USRQUOTA;
1364 			goto set_qf_name;
1365 		case Opt_grpjquota:
1366 			qtype = GRPQUOTA;
1367 set_qf_name:
1368 			if (sb_any_quota_loaded(sb) &&
1369 			    !sbi->s_qf_names[qtype]) {
1370 				ext4_msg(sb, KERN_ERR,
1371 				       "Cannot change journaled "
1372 				       "quota options when quota turned on");
1373 				return 0;
1374 			}
1375 			qname = match_strdup(&args[0]);
1376 			if (!qname) {
1377 				ext4_msg(sb, KERN_ERR,
1378 					"Not enough memory for "
1379 					"storing quotafile name");
1380 				return 0;
1381 			}
1382 			if (sbi->s_qf_names[qtype] &&
1383 			    strcmp(sbi->s_qf_names[qtype], qname)) {
1384 				ext4_msg(sb, KERN_ERR,
1385 					"%s quota file already "
1386 					"specified", QTYPE2NAME(qtype));
1387 				kfree(qname);
1388 				return 0;
1389 			}
1390 			sbi->s_qf_names[qtype] = qname;
1391 			if (strchr(sbi->s_qf_names[qtype], '/')) {
1392 				ext4_msg(sb, KERN_ERR,
1393 					"quotafile must be on "
1394 					"filesystem root");
1395 				kfree(sbi->s_qf_names[qtype]);
1396 				sbi->s_qf_names[qtype] = NULL;
1397 				return 0;
1398 			}
1399 			set_opt(sbi->s_mount_opt, QUOTA);
1400 			break;
1401 		case Opt_offusrjquota:
1402 			qtype = USRQUOTA;
1403 			goto clear_qf_name;
1404 		case Opt_offgrpjquota:
1405 			qtype = GRPQUOTA;
1406 clear_qf_name:
1407 			if (sb_any_quota_loaded(sb) &&
1408 			    sbi->s_qf_names[qtype]) {
1409 				ext4_msg(sb, KERN_ERR, "Cannot change "
1410 					"journaled quota options when "
1411 					"quota turned on");
1412 				return 0;
1413 			}
1414 			/*
1415 			 * The space will be released later when all options
1416 			 * are confirmed to be correct
1417 			 */
1418 			sbi->s_qf_names[qtype] = NULL;
1419 			break;
1420 		case Opt_jqfmt_vfsold:
1421 			qfmt = QFMT_VFS_OLD;
1422 			goto set_qf_format;
1423 		case Opt_jqfmt_vfsv0:
1424 			qfmt = QFMT_VFS_V0;
1425 set_qf_format:
1426 			if (sb_any_quota_loaded(sb) &&
1427 			    sbi->s_jquota_fmt != qfmt) {
1428 				ext4_msg(sb, KERN_ERR, "Cannot change "
1429 					"journaled quota options when "
1430 					"quota turned on");
1431 				return 0;
1432 			}
1433 			sbi->s_jquota_fmt = qfmt;
1434 			break;
1435 		case Opt_quota:
1436 		case Opt_usrquota:
1437 			set_opt(sbi->s_mount_opt, QUOTA);
1438 			set_opt(sbi->s_mount_opt, USRQUOTA);
1439 			break;
1440 		case Opt_grpquota:
1441 			set_opt(sbi->s_mount_opt, QUOTA);
1442 			set_opt(sbi->s_mount_opt, GRPQUOTA);
1443 			break;
1444 		case Opt_noquota:
1445 			if (sb_any_quota_loaded(sb)) {
1446 				ext4_msg(sb, KERN_ERR, "Cannot change quota "
1447 					"options when quota turned on");
1448 				return 0;
1449 			}
1450 			clear_opt(sbi->s_mount_opt, QUOTA);
1451 			clear_opt(sbi->s_mount_opt, USRQUOTA);
1452 			clear_opt(sbi->s_mount_opt, GRPQUOTA);
1453 			break;
1454 #else
1455 		case Opt_quota:
1456 		case Opt_usrquota:
1457 		case Opt_grpquota:
1458 			ext4_msg(sb, KERN_ERR,
1459 				"quota options not supported");
1460 			break;
1461 		case Opt_usrjquota:
1462 		case Opt_grpjquota:
1463 		case Opt_offusrjquota:
1464 		case Opt_offgrpjquota:
1465 		case Opt_jqfmt_vfsold:
1466 		case Opt_jqfmt_vfsv0:
1467 			ext4_msg(sb, KERN_ERR,
1468 				"journaled quota options not supported");
1469 			break;
1470 		case Opt_noquota:
1471 			break;
1472 #endif
1473 		case Opt_abort:
1474 			set_opt(sbi->s_mount_opt, ABORT);
1475 			break;
1476 		case Opt_nobarrier:
1477 			clear_opt(sbi->s_mount_opt, BARRIER);
1478 			break;
1479 		case Opt_barrier:
1480 			if (match_int(&args[0], &option)) {
1481 				set_opt(sbi->s_mount_opt, BARRIER);
1482 				break;
1483 			}
1484 			if (option)
1485 				set_opt(sbi->s_mount_opt, BARRIER);
1486 			else
1487 				clear_opt(sbi->s_mount_opt, BARRIER);
1488 			break;
1489 		case Opt_ignore:
1490 			break;
1491 		case Opt_resize:
1492 			if (!is_remount) {
1493 				ext4_msg(sb, KERN_ERR,
1494 					"resize option only available "
1495 					"for remount");
1496 				return 0;
1497 			}
1498 			if (match_int(&args[0], &option) != 0)
1499 				return 0;
1500 			*n_blocks_count = option;
1501 			break;
1502 		case Opt_nobh:
1503 			set_opt(sbi->s_mount_opt, NOBH);
1504 			break;
1505 		case Opt_bh:
1506 			clear_opt(sbi->s_mount_opt, NOBH);
1507 			break;
1508 		case Opt_i_version:
1509 			set_opt(sbi->s_mount_opt, I_VERSION);
1510 			sb->s_flags |= MS_I_VERSION;
1511 			break;
1512 		case Opt_nodelalloc:
1513 			clear_opt(sbi->s_mount_opt, DELALLOC);
1514 			break;
1515 		case Opt_stripe:
1516 			if (match_int(&args[0], &option))
1517 				return 0;
1518 			if (option < 0)
1519 				return 0;
1520 			sbi->s_stripe = option;
1521 			break;
1522 		case Opt_delalloc:
1523 			set_opt(sbi->s_mount_opt, DELALLOC);
1524 			break;
1525 		case Opt_block_validity:
1526 			set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1527 			break;
1528 		case Opt_noblock_validity:
1529 			clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1530 			break;
1531 		case Opt_inode_readahead_blks:
1532 			if (match_int(&args[0], &option))
1533 				return 0;
1534 			if (option < 0 || option > (1 << 30))
1535 				return 0;
1536 			if (!is_power_of_2(option)) {
1537 				ext4_msg(sb, KERN_ERR,
1538 					 "EXT4-fs: inode_readahead_blks"
1539 					 " must be a power of 2");
1540 				return 0;
1541 			}
1542 			sbi->s_inode_readahead_blks = option;
1543 			break;
1544 		case Opt_journal_ioprio:
1545 			if (match_int(&args[0], &option))
1546 				return 0;
1547 			if (option < 0 || option > 7)
1548 				break;
1549 			*journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1550 							    option);
1551 			break;
1552 		case Opt_noauto_da_alloc:
1553 			set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1554 			break;
1555 		case Opt_auto_da_alloc:
1556 			if (match_int(&args[0], &option)) {
1557 				clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1558 				break;
1559 			}
1560 			if (option)
1561 				clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1562 			else
1563 				set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1564 			break;
1565 		default:
1566 			ext4_msg(sb, KERN_ERR,
1567 			       "Unrecognized mount option \"%s\" "
1568 			       "or missing value", p);
1569 			return 0;
1570 		}
1571 	}
1572 #ifdef CONFIG_QUOTA
1573 	if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1574 		if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1575 		     sbi->s_qf_names[USRQUOTA])
1576 			clear_opt(sbi->s_mount_opt, USRQUOTA);
1577 
1578 		if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1579 		     sbi->s_qf_names[GRPQUOTA])
1580 			clear_opt(sbi->s_mount_opt, GRPQUOTA);
1581 
1582 		if ((sbi->s_qf_names[USRQUOTA] &&
1583 				(sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1584 		    (sbi->s_qf_names[GRPQUOTA] &&
1585 				(sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1586 			ext4_msg(sb, KERN_ERR, "old and new quota "
1587 					"format mixing");
1588 			return 0;
1589 		}
1590 
1591 		if (!sbi->s_jquota_fmt) {
1592 			ext4_msg(sb, KERN_ERR, "journaled quota format "
1593 					"not specified");
1594 			return 0;
1595 		}
1596 	} else {
1597 		if (sbi->s_jquota_fmt) {
1598 			ext4_msg(sb, KERN_ERR, "journaled quota format "
1599 					"specified with no journaling "
1600 					"enabled");
1601 			return 0;
1602 		}
1603 	}
1604 #endif
1605 	return 1;
1606 }
1607 
1608 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1609 			    int read_only)
1610 {
1611 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1612 	int res = 0;
1613 
1614 	if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1615 		ext4_msg(sb, KERN_ERR, "revision level too high, "
1616 			 "forcing read-only mode");
1617 		res = MS_RDONLY;
1618 	}
1619 	if (read_only)
1620 		return res;
1621 	if (!(sbi->s_mount_state & EXT4_VALID_FS))
1622 		ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1623 			 "running e2fsck is recommended");
1624 	else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1625 		ext4_msg(sb, KERN_WARNING,
1626 			 "warning: mounting fs with errors, "
1627 			 "running e2fsck is recommended");
1628 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1629 		 le16_to_cpu(es->s_mnt_count) >=
1630 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1631 		ext4_msg(sb, KERN_WARNING,
1632 			 "warning: maximal mount count reached, "
1633 			 "running e2fsck is recommended");
1634 	else if (le32_to_cpu(es->s_checkinterval) &&
1635 		(le32_to_cpu(es->s_lastcheck) +
1636 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1637 		ext4_msg(sb, KERN_WARNING,
1638 			 "warning: checktime reached, "
1639 			 "running e2fsck is recommended");
1640 	if (!sbi->s_journal)
1641 		es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1642 	if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1643 		es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1644 	le16_add_cpu(&es->s_mnt_count, 1);
1645 	es->s_mtime = cpu_to_le32(get_seconds());
1646 	ext4_update_dynamic_rev(sb);
1647 	if (sbi->s_journal)
1648 		EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1649 
1650 	ext4_commit_super(sb, 1);
1651 	if (test_opt(sb, DEBUG))
1652 		printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1653 				"bpg=%lu, ipg=%lu, mo=%04lx]\n",
1654 			sb->s_blocksize,
1655 			sbi->s_groups_count,
1656 			EXT4_BLOCKS_PER_GROUP(sb),
1657 			EXT4_INODES_PER_GROUP(sb),
1658 			sbi->s_mount_opt);
1659 
1660 	if (EXT4_SB(sb)->s_journal) {
1661 		ext4_msg(sb, KERN_INFO, "%s journal on %s",
1662 		       EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1663 		       "external", EXT4_SB(sb)->s_journal->j_devname);
1664 	} else {
1665 		ext4_msg(sb, KERN_INFO, "no journal");
1666 	}
1667 	return res;
1668 }
1669 
1670 static int ext4_fill_flex_info(struct super_block *sb)
1671 {
1672 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1673 	struct ext4_group_desc *gdp = NULL;
1674 	ext4_group_t flex_group_count;
1675 	ext4_group_t flex_group;
1676 	int groups_per_flex = 0;
1677 	size_t size;
1678 	int i;
1679 
1680 	if (!sbi->s_es->s_log_groups_per_flex) {
1681 		sbi->s_log_groups_per_flex = 0;
1682 		return 1;
1683 	}
1684 
1685 	sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1686 	groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1687 
1688 	/* We allocate both existing and potentially added groups */
1689 	flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1690 			((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1691 			      EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1692 	size = flex_group_count * sizeof(struct flex_groups);
1693 	sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1694 	if (sbi->s_flex_groups == NULL) {
1695 		sbi->s_flex_groups = vmalloc(size);
1696 		if (sbi->s_flex_groups)
1697 			memset(sbi->s_flex_groups, 0, size);
1698 	}
1699 	if (sbi->s_flex_groups == NULL) {
1700 		ext4_msg(sb, KERN_ERR, "not enough memory for "
1701 				"%u flex groups", flex_group_count);
1702 		goto failed;
1703 	}
1704 
1705 	for (i = 0; i < sbi->s_groups_count; i++) {
1706 		gdp = ext4_get_group_desc(sb, i, NULL);
1707 
1708 		flex_group = ext4_flex_group(sbi, i);
1709 		atomic_set(&sbi->s_flex_groups[flex_group].free_inodes,
1710 			   ext4_free_inodes_count(sb, gdp));
1711 		atomic_set(&sbi->s_flex_groups[flex_group].free_blocks,
1712 			   ext4_free_blks_count(sb, gdp));
1713 		atomic_set(&sbi->s_flex_groups[flex_group].used_dirs,
1714 			   ext4_used_dirs_count(sb, gdp));
1715 	}
1716 
1717 	return 1;
1718 failed:
1719 	return 0;
1720 }
1721 
1722 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1723 			    struct ext4_group_desc *gdp)
1724 {
1725 	__u16 crc = 0;
1726 
1727 	if (sbi->s_es->s_feature_ro_compat &
1728 	    cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1729 		int offset = offsetof(struct ext4_group_desc, bg_checksum);
1730 		__le32 le_group = cpu_to_le32(block_group);
1731 
1732 		crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1733 		crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1734 		crc = crc16(crc, (__u8 *)gdp, offset);
1735 		offset += sizeof(gdp->bg_checksum); /* skip checksum */
1736 		/* for checksum of struct ext4_group_desc do the rest...*/
1737 		if ((sbi->s_es->s_feature_incompat &
1738 		     cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1739 		    offset < le16_to_cpu(sbi->s_es->s_desc_size))
1740 			crc = crc16(crc, (__u8 *)gdp + offset,
1741 				    le16_to_cpu(sbi->s_es->s_desc_size) -
1742 					offset);
1743 	}
1744 
1745 	return cpu_to_le16(crc);
1746 }
1747 
1748 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1749 				struct ext4_group_desc *gdp)
1750 {
1751 	if ((sbi->s_es->s_feature_ro_compat &
1752 	     cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1753 	    (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1754 		return 0;
1755 
1756 	return 1;
1757 }
1758 
1759 /* Called at mount-time, super-block is locked */
1760 static int ext4_check_descriptors(struct super_block *sb)
1761 {
1762 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1763 	ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1764 	ext4_fsblk_t last_block;
1765 	ext4_fsblk_t block_bitmap;
1766 	ext4_fsblk_t inode_bitmap;
1767 	ext4_fsblk_t inode_table;
1768 	int flexbg_flag = 0;
1769 	ext4_group_t i;
1770 
1771 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1772 		flexbg_flag = 1;
1773 
1774 	ext4_debug("Checking group descriptors");
1775 
1776 	for (i = 0; i < sbi->s_groups_count; i++) {
1777 		struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1778 
1779 		if (i == sbi->s_groups_count - 1 || flexbg_flag)
1780 			last_block = ext4_blocks_count(sbi->s_es) - 1;
1781 		else
1782 			last_block = first_block +
1783 				(EXT4_BLOCKS_PER_GROUP(sb) - 1);
1784 
1785 		block_bitmap = ext4_block_bitmap(sb, gdp);
1786 		if (block_bitmap < first_block || block_bitmap > last_block) {
1787 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1788 			       "Block bitmap for group %u not in group "
1789 			       "(block %llu)!", i, block_bitmap);
1790 			return 0;
1791 		}
1792 		inode_bitmap = ext4_inode_bitmap(sb, gdp);
1793 		if (inode_bitmap < first_block || inode_bitmap > last_block) {
1794 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1795 			       "Inode bitmap for group %u not in group "
1796 			       "(block %llu)!", i, inode_bitmap);
1797 			return 0;
1798 		}
1799 		inode_table = ext4_inode_table(sb, gdp);
1800 		if (inode_table < first_block ||
1801 		    inode_table + sbi->s_itb_per_group - 1 > last_block) {
1802 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1803 			       "Inode table for group %u not in group "
1804 			       "(block %llu)!", i, inode_table);
1805 			return 0;
1806 		}
1807 		ext4_lock_group(sb, i);
1808 		if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1809 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1810 				 "Checksum for group %u failed (%u!=%u)",
1811 				 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1812 				     gdp)), le16_to_cpu(gdp->bg_checksum));
1813 			if (!(sb->s_flags & MS_RDONLY)) {
1814 				ext4_unlock_group(sb, i);
1815 				return 0;
1816 			}
1817 		}
1818 		ext4_unlock_group(sb, i);
1819 		if (!flexbg_flag)
1820 			first_block += EXT4_BLOCKS_PER_GROUP(sb);
1821 	}
1822 
1823 	ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1824 	sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
1825 	return 1;
1826 }
1827 
1828 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1829  * the superblock) which were deleted from all directories, but held open by
1830  * a process at the time of a crash.  We walk the list and try to delete these
1831  * inodes at recovery time (only with a read-write filesystem).
1832  *
1833  * In order to keep the orphan inode chain consistent during traversal (in
1834  * case of crash during recovery), we link each inode into the superblock
1835  * orphan list_head and handle it the same way as an inode deletion during
1836  * normal operation (which journals the operations for us).
1837  *
1838  * We only do an iget() and an iput() on each inode, which is very safe if we
1839  * accidentally point at an in-use or already deleted inode.  The worst that
1840  * can happen in this case is that we get a "bit already cleared" message from
1841  * ext4_free_inode().  The only reason we would point at a wrong inode is if
1842  * e2fsck was run on this filesystem, and it must have already done the orphan
1843  * inode cleanup for us, so we can safely abort without any further action.
1844  */
1845 static void ext4_orphan_cleanup(struct super_block *sb,
1846 				struct ext4_super_block *es)
1847 {
1848 	unsigned int s_flags = sb->s_flags;
1849 	int nr_orphans = 0, nr_truncates = 0;
1850 #ifdef CONFIG_QUOTA
1851 	int i;
1852 #endif
1853 	if (!es->s_last_orphan) {
1854 		jbd_debug(4, "no orphan inodes to clean up\n");
1855 		return;
1856 	}
1857 
1858 	if (bdev_read_only(sb->s_bdev)) {
1859 		ext4_msg(sb, KERN_ERR, "write access "
1860 			"unavailable, skipping orphan cleanup");
1861 		return;
1862 	}
1863 
1864 	if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1865 		if (es->s_last_orphan)
1866 			jbd_debug(1, "Errors on filesystem, "
1867 				  "clearing orphan list.\n");
1868 		es->s_last_orphan = 0;
1869 		jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1870 		return;
1871 	}
1872 
1873 	if (s_flags & MS_RDONLY) {
1874 		ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1875 		sb->s_flags &= ~MS_RDONLY;
1876 	}
1877 #ifdef CONFIG_QUOTA
1878 	/* Needed for iput() to work correctly and not trash data */
1879 	sb->s_flags |= MS_ACTIVE;
1880 	/* Turn on quotas so that they are updated correctly */
1881 	for (i = 0; i < MAXQUOTAS; i++) {
1882 		if (EXT4_SB(sb)->s_qf_names[i]) {
1883 			int ret = ext4_quota_on_mount(sb, i);
1884 			if (ret < 0)
1885 				ext4_msg(sb, KERN_ERR,
1886 					"Cannot turn on journaled "
1887 					"quota: error %d", ret);
1888 		}
1889 	}
1890 #endif
1891 
1892 	while (es->s_last_orphan) {
1893 		struct inode *inode;
1894 
1895 		inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1896 		if (IS_ERR(inode)) {
1897 			es->s_last_orphan = 0;
1898 			break;
1899 		}
1900 
1901 		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1902 		vfs_dq_init(inode);
1903 		if (inode->i_nlink) {
1904 			ext4_msg(sb, KERN_DEBUG,
1905 				"%s: truncating inode %lu to %lld bytes",
1906 				__func__, inode->i_ino, inode->i_size);
1907 			jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1908 				  inode->i_ino, inode->i_size);
1909 			ext4_truncate(inode);
1910 			nr_truncates++;
1911 		} else {
1912 			ext4_msg(sb, KERN_DEBUG,
1913 				"%s: deleting unreferenced inode %lu",
1914 				__func__, inode->i_ino);
1915 			jbd_debug(2, "deleting unreferenced inode %lu\n",
1916 				  inode->i_ino);
1917 			nr_orphans++;
1918 		}
1919 		iput(inode);  /* The delete magic happens here! */
1920 	}
1921 
1922 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1923 
1924 	if (nr_orphans)
1925 		ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1926 		       PLURAL(nr_orphans));
1927 	if (nr_truncates)
1928 		ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1929 		       PLURAL(nr_truncates));
1930 #ifdef CONFIG_QUOTA
1931 	/* Turn quotas off */
1932 	for (i = 0; i < MAXQUOTAS; i++) {
1933 		if (sb_dqopt(sb)->files[i])
1934 			vfs_quota_off(sb, i, 0);
1935 	}
1936 #endif
1937 	sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1938 }
1939 
1940 /*
1941  * Maximal extent format file size.
1942  * Resulting logical blkno at s_maxbytes must fit in our on-disk
1943  * extent format containers, within a sector_t, and within i_blocks
1944  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
1945  * so that won't be a limiting factor.
1946  *
1947  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1948  */
1949 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1950 {
1951 	loff_t res;
1952 	loff_t upper_limit = MAX_LFS_FILESIZE;
1953 
1954 	/* small i_blocks in vfs inode? */
1955 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1956 		/*
1957 		 * CONFIG_LBD is not enabled implies the inode
1958 		 * i_block represent total blocks in 512 bytes
1959 		 * 32 == size of vfs inode i_blocks * 8
1960 		 */
1961 		upper_limit = (1LL << 32) - 1;
1962 
1963 		/* total blocks in file system block size */
1964 		upper_limit >>= (blkbits - 9);
1965 		upper_limit <<= blkbits;
1966 	}
1967 
1968 	/* 32-bit extent-start container, ee_block */
1969 	res = 1LL << 32;
1970 	res <<= blkbits;
1971 	res -= 1;
1972 
1973 	/* Sanity check against vm- & vfs- imposed limits */
1974 	if (res > upper_limit)
1975 		res = upper_limit;
1976 
1977 	return res;
1978 }
1979 
1980 /*
1981  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
1982  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1983  * We need to be 1 filesystem block less than the 2^48 sector limit.
1984  */
1985 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1986 {
1987 	loff_t res = EXT4_NDIR_BLOCKS;
1988 	int meta_blocks;
1989 	loff_t upper_limit;
1990 	/* This is calculated to be the largest file size for a dense, block
1991 	 * mapped file such that the file's total number of 512-byte sectors,
1992 	 * including data and all indirect blocks, does not exceed (2^48 - 1).
1993 	 *
1994 	 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
1995 	 * number of 512-byte sectors of the file.
1996 	 */
1997 
1998 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1999 		/*
2000 		 * !has_huge_files or CONFIG_LBD not enabled implies that
2001 		 * the inode i_block field represents total file blocks in
2002 		 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2003 		 */
2004 		upper_limit = (1LL << 32) - 1;
2005 
2006 		/* total blocks in file system block size */
2007 		upper_limit >>= (bits - 9);
2008 
2009 	} else {
2010 		/*
2011 		 * We use 48 bit ext4_inode i_blocks
2012 		 * With EXT4_HUGE_FILE_FL set the i_blocks
2013 		 * represent total number of blocks in
2014 		 * file system block size
2015 		 */
2016 		upper_limit = (1LL << 48) - 1;
2017 
2018 	}
2019 
2020 	/* indirect blocks */
2021 	meta_blocks = 1;
2022 	/* double indirect blocks */
2023 	meta_blocks += 1 + (1LL << (bits-2));
2024 	/* tripple indirect blocks */
2025 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2026 
2027 	upper_limit -= meta_blocks;
2028 	upper_limit <<= bits;
2029 
2030 	res += 1LL << (bits-2);
2031 	res += 1LL << (2*(bits-2));
2032 	res += 1LL << (3*(bits-2));
2033 	res <<= bits;
2034 	if (res > upper_limit)
2035 		res = upper_limit;
2036 
2037 	if (res > MAX_LFS_FILESIZE)
2038 		res = MAX_LFS_FILESIZE;
2039 
2040 	return res;
2041 }
2042 
2043 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2044 				   ext4_fsblk_t logical_sb_block, int nr)
2045 {
2046 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2047 	ext4_group_t bg, first_meta_bg;
2048 	int has_super = 0;
2049 
2050 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2051 
2052 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2053 	    nr < first_meta_bg)
2054 		return logical_sb_block + nr + 1;
2055 	bg = sbi->s_desc_per_block * nr;
2056 	if (ext4_bg_has_super(sb, bg))
2057 		has_super = 1;
2058 
2059 	return (has_super + ext4_group_first_block_no(sb, bg));
2060 }
2061 
2062 /**
2063  * ext4_get_stripe_size: Get the stripe size.
2064  * @sbi: In memory super block info
2065  *
2066  * If we have specified it via mount option, then
2067  * use the mount option value. If the value specified at mount time is
2068  * greater than the blocks per group use the super block value.
2069  * If the super block value is greater than blocks per group return 0.
2070  * Allocator needs it be less than blocks per group.
2071  *
2072  */
2073 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2074 {
2075 	unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2076 	unsigned long stripe_width =
2077 			le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2078 
2079 	if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2080 		return sbi->s_stripe;
2081 
2082 	if (stripe_width <= sbi->s_blocks_per_group)
2083 		return stripe_width;
2084 
2085 	if (stride <= sbi->s_blocks_per_group)
2086 		return stride;
2087 
2088 	return 0;
2089 }
2090 
2091 /* sysfs supprt */
2092 
2093 struct ext4_attr {
2094 	struct attribute attr;
2095 	ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2096 	ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2097 			 const char *, size_t);
2098 	int offset;
2099 };
2100 
2101 static int parse_strtoul(const char *buf,
2102 		unsigned long max, unsigned long *value)
2103 {
2104 	char *endp;
2105 
2106 	while (*buf && isspace(*buf))
2107 		buf++;
2108 	*value = simple_strtoul(buf, &endp, 0);
2109 	while (*endp && isspace(*endp))
2110 		endp++;
2111 	if (*endp || *value > max)
2112 		return -EINVAL;
2113 
2114 	return 0;
2115 }
2116 
2117 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2118 					      struct ext4_sb_info *sbi,
2119 					      char *buf)
2120 {
2121 	return snprintf(buf, PAGE_SIZE, "%llu\n",
2122 			(s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2123 }
2124 
2125 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2126 					 struct ext4_sb_info *sbi, char *buf)
2127 {
2128 	struct super_block *sb = sbi->s_buddy_cache->i_sb;
2129 
2130 	return snprintf(buf, PAGE_SIZE, "%lu\n",
2131 			(part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2132 			 sbi->s_sectors_written_start) >> 1);
2133 }
2134 
2135 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2136 					  struct ext4_sb_info *sbi, char *buf)
2137 {
2138 	struct super_block *sb = sbi->s_buddy_cache->i_sb;
2139 
2140 	return snprintf(buf, PAGE_SIZE, "%llu\n",
2141 			sbi->s_kbytes_written +
2142 			((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2143 			  EXT4_SB(sb)->s_sectors_written_start) >> 1));
2144 }
2145 
2146 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2147 					  struct ext4_sb_info *sbi,
2148 					  const char *buf, size_t count)
2149 {
2150 	unsigned long t;
2151 
2152 	if (parse_strtoul(buf, 0x40000000, &t))
2153 		return -EINVAL;
2154 
2155 	if (!is_power_of_2(t))
2156 		return -EINVAL;
2157 
2158 	sbi->s_inode_readahead_blks = t;
2159 	return count;
2160 }
2161 
2162 static ssize_t sbi_ui_show(struct ext4_attr *a,
2163 			   struct ext4_sb_info *sbi, char *buf)
2164 {
2165 	unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2166 
2167 	return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2168 }
2169 
2170 static ssize_t sbi_ui_store(struct ext4_attr *a,
2171 			    struct ext4_sb_info *sbi,
2172 			    const char *buf, size_t count)
2173 {
2174 	unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2175 	unsigned long t;
2176 
2177 	if (parse_strtoul(buf, 0xffffffff, &t))
2178 		return -EINVAL;
2179 	*ui = t;
2180 	return count;
2181 }
2182 
2183 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2184 static struct ext4_attr ext4_attr_##_name = {			\
2185 	.attr = {.name = __stringify(_name), .mode = _mode },	\
2186 	.show	= _show,					\
2187 	.store	= _store,					\
2188 	.offset = offsetof(struct ext4_sb_info, _elname),	\
2189 }
2190 #define EXT4_ATTR(name, mode, show, store) \
2191 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2192 
2193 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2194 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2195 #define EXT4_RW_ATTR_SBI_UI(name, elname)	\
2196 	EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2197 #define ATTR_LIST(name) &ext4_attr_##name.attr
2198 
2199 EXT4_RO_ATTR(delayed_allocation_blocks);
2200 EXT4_RO_ATTR(session_write_kbytes);
2201 EXT4_RO_ATTR(lifetime_write_kbytes);
2202 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2203 		 inode_readahead_blks_store, s_inode_readahead_blks);
2204 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2205 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2206 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2207 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2208 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2209 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2210 
2211 static struct attribute *ext4_attrs[] = {
2212 	ATTR_LIST(delayed_allocation_blocks),
2213 	ATTR_LIST(session_write_kbytes),
2214 	ATTR_LIST(lifetime_write_kbytes),
2215 	ATTR_LIST(inode_readahead_blks),
2216 	ATTR_LIST(mb_stats),
2217 	ATTR_LIST(mb_max_to_scan),
2218 	ATTR_LIST(mb_min_to_scan),
2219 	ATTR_LIST(mb_order2_req),
2220 	ATTR_LIST(mb_stream_req),
2221 	ATTR_LIST(mb_group_prealloc),
2222 	NULL,
2223 };
2224 
2225 static ssize_t ext4_attr_show(struct kobject *kobj,
2226 			      struct attribute *attr, char *buf)
2227 {
2228 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2229 						s_kobj);
2230 	struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2231 
2232 	return a->show ? a->show(a, sbi, buf) : 0;
2233 }
2234 
2235 static ssize_t ext4_attr_store(struct kobject *kobj,
2236 			       struct attribute *attr,
2237 			       const char *buf, size_t len)
2238 {
2239 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2240 						s_kobj);
2241 	struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2242 
2243 	return a->store ? a->store(a, sbi, buf, len) : 0;
2244 }
2245 
2246 static void ext4_sb_release(struct kobject *kobj)
2247 {
2248 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2249 						s_kobj);
2250 	complete(&sbi->s_kobj_unregister);
2251 }
2252 
2253 
2254 static struct sysfs_ops ext4_attr_ops = {
2255 	.show	= ext4_attr_show,
2256 	.store	= ext4_attr_store,
2257 };
2258 
2259 static struct kobj_type ext4_ktype = {
2260 	.default_attrs	= ext4_attrs,
2261 	.sysfs_ops	= &ext4_attr_ops,
2262 	.release	= ext4_sb_release,
2263 };
2264 
2265 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2266 				__releases(kernel_lock)
2267 				__acquires(kernel_lock)
2268 {
2269 	struct buffer_head *bh;
2270 	struct ext4_super_block *es = NULL;
2271 	struct ext4_sb_info *sbi;
2272 	ext4_fsblk_t block;
2273 	ext4_fsblk_t sb_block = get_sb_block(&data);
2274 	ext4_fsblk_t logical_sb_block;
2275 	unsigned long offset = 0;
2276 	unsigned long journal_devnum = 0;
2277 	unsigned long def_mount_opts;
2278 	struct inode *root;
2279 	char *cp;
2280 	const char *descr;
2281 	int ret = -EINVAL;
2282 	int blocksize;
2283 	unsigned int db_count;
2284 	unsigned int i;
2285 	int needs_recovery, has_huge_files;
2286 	int features;
2287 	__u64 blocks_count;
2288 	int err;
2289 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2290 
2291 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2292 	if (!sbi)
2293 		return -ENOMEM;
2294 
2295 	sbi->s_blockgroup_lock =
2296 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2297 	if (!sbi->s_blockgroup_lock) {
2298 		kfree(sbi);
2299 		return -ENOMEM;
2300 	}
2301 	sb->s_fs_info = sbi;
2302 	sbi->s_mount_opt = 0;
2303 	sbi->s_resuid = EXT4_DEF_RESUID;
2304 	sbi->s_resgid = EXT4_DEF_RESGID;
2305 	sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2306 	sbi->s_sb_block = sb_block;
2307 	sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2308 						      sectors[1]);
2309 
2310 	unlock_kernel();
2311 
2312 	/* Cleanup superblock name */
2313 	for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2314 		*cp = '!';
2315 
2316 	blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2317 	if (!blocksize) {
2318 		ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2319 		goto out_fail;
2320 	}
2321 
2322 	/*
2323 	 * The ext4 superblock will not be buffer aligned for other than 1kB
2324 	 * block sizes.  We need to calculate the offset from buffer start.
2325 	 */
2326 	if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2327 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2328 		offset = do_div(logical_sb_block, blocksize);
2329 	} else {
2330 		logical_sb_block = sb_block;
2331 	}
2332 
2333 	if (!(bh = sb_bread(sb, logical_sb_block))) {
2334 		ext4_msg(sb, KERN_ERR, "unable to read superblock");
2335 		goto out_fail;
2336 	}
2337 	/*
2338 	 * Note: s_es must be initialized as soon as possible because
2339 	 *       some ext4 macro-instructions depend on its value
2340 	 */
2341 	es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2342 	sbi->s_es = es;
2343 	sb->s_magic = le16_to_cpu(es->s_magic);
2344 	if (sb->s_magic != EXT4_SUPER_MAGIC)
2345 		goto cantfind_ext4;
2346 	sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2347 
2348 	/* Set defaults before we parse the mount options */
2349 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2350 	if (def_mount_opts & EXT4_DEFM_DEBUG)
2351 		set_opt(sbi->s_mount_opt, DEBUG);
2352 	if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2353 		set_opt(sbi->s_mount_opt, GRPID);
2354 	if (def_mount_opts & EXT4_DEFM_UID16)
2355 		set_opt(sbi->s_mount_opt, NO_UID32);
2356 #ifdef CONFIG_EXT4_FS_XATTR
2357 	if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2358 		set_opt(sbi->s_mount_opt, XATTR_USER);
2359 #endif
2360 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2361 	if (def_mount_opts & EXT4_DEFM_ACL)
2362 		set_opt(sbi->s_mount_opt, POSIX_ACL);
2363 #endif
2364 	if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2365 		sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2366 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2367 		sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2368 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2369 		sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2370 
2371 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2372 		set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2373 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2374 		set_opt(sbi->s_mount_opt, ERRORS_CONT);
2375 	else
2376 		set_opt(sbi->s_mount_opt, ERRORS_RO);
2377 
2378 	sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2379 	sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2380 	sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2381 	sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2382 	sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2383 	sbi->s_mb_history_max = default_mb_history_length;
2384 
2385 	set_opt(sbi->s_mount_opt, BARRIER);
2386 
2387 	/*
2388 	 * enable delayed allocation by default
2389 	 * Use -o nodelalloc to turn it off
2390 	 */
2391 	set_opt(sbi->s_mount_opt, DELALLOC);
2392 
2393 	if (!parse_options((char *) data, sb, &journal_devnum,
2394 			   &journal_ioprio, NULL, 0))
2395 		goto failed_mount;
2396 
2397 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2398 		((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2399 
2400 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2401 	    (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2402 	     EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2403 	     EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2404 		ext4_msg(sb, KERN_WARNING,
2405 		       "feature flags set on rev 0 fs, "
2406 		       "running e2fsck is recommended");
2407 
2408 	/*
2409 	 * Check feature flags regardless of the revision level, since we
2410 	 * previously didn't change the revision level when setting the flags,
2411 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
2412 	 */
2413 	features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2414 	if (features) {
2415 		ext4_msg(sb, KERN_ERR,
2416 			"Couldn't mount because of "
2417 			"unsupported optional features (%x)",
2418 			(le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2419 			~EXT4_FEATURE_INCOMPAT_SUPP));
2420 		goto failed_mount;
2421 	}
2422 	features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2423 	if (!(sb->s_flags & MS_RDONLY) && features) {
2424 		ext4_msg(sb, KERN_ERR,
2425 			"Couldn't mount RDWR because of "
2426 			"unsupported optional features (%x)",
2427 			(le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2428 			~EXT4_FEATURE_RO_COMPAT_SUPP));
2429 		goto failed_mount;
2430 	}
2431 	has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2432 				    EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2433 	if (has_huge_files) {
2434 		/*
2435 		 * Large file size enabled file system can only be
2436 		 * mount if kernel is build with CONFIG_LBD
2437 		 */
2438 		if (sizeof(root->i_blocks) < sizeof(u64) &&
2439 				!(sb->s_flags & MS_RDONLY)) {
2440 			ext4_msg(sb, KERN_ERR, "Filesystem with huge "
2441 					"files cannot be mounted read-write "
2442 					"without CONFIG_LBD");
2443 			goto failed_mount;
2444 		}
2445 	}
2446 	blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2447 
2448 	if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2449 	    blocksize > EXT4_MAX_BLOCK_SIZE) {
2450 		ext4_msg(sb, KERN_ERR,
2451 		       "Unsupported filesystem blocksize %d", blocksize);
2452 		goto failed_mount;
2453 	}
2454 
2455 	if (sb->s_blocksize != blocksize) {
2456 		/* Validate the filesystem blocksize */
2457 		if (!sb_set_blocksize(sb, blocksize)) {
2458 			ext4_msg(sb, KERN_ERR, "bad block size %d",
2459 					blocksize);
2460 			goto failed_mount;
2461 		}
2462 
2463 		brelse(bh);
2464 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2465 		offset = do_div(logical_sb_block, blocksize);
2466 		bh = sb_bread(sb, logical_sb_block);
2467 		if (!bh) {
2468 			ext4_msg(sb, KERN_ERR,
2469 			       "Can't read superblock on 2nd try");
2470 			goto failed_mount;
2471 		}
2472 		es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2473 		sbi->s_es = es;
2474 		if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2475 			ext4_msg(sb, KERN_ERR,
2476 			       "Magic mismatch, very weird!");
2477 			goto failed_mount;
2478 		}
2479 	}
2480 
2481 	sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2482 						      has_huge_files);
2483 	sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2484 
2485 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2486 		sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2487 		sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2488 	} else {
2489 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2490 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2491 		if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2492 		    (!is_power_of_2(sbi->s_inode_size)) ||
2493 		    (sbi->s_inode_size > blocksize)) {
2494 			ext4_msg(sb, KERN_ERR,
2495 			       "unsupported inode size: %d",
2496 			       sbi->s_inode_size);
2497 			goto failed_mount;
2498 		}
2499 		if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2500 			sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2501 	}
2502 
2503 	sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2504 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2505 		if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2506 		    sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2507 		    !is_power_of_2(sbi->s_desc_size)) {
2508 			ext4_msg(sb, KERN_ERR,
2509 			       "unsupported descriptor size %lu",
2510 			       sbi->s_desc_size);
2511 			goto failed_mount;
2512 		}
2513 	} else
2514 		sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2515 
2516 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2517 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2518 	if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2519 		goto cantfind_ext4;
2520 
2521 	sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2522 	if (sbi->s_inodes_per_block == 0)
2523 		goto cantfind_ext4;
2524 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
2525 					sbi->s_inodes_per_block;
2526 	sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2527 	sbi->s_sbh = bh;
2528 	sbi->s_mount_state = le16_to_cpu(es->s_state);
2529 	sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2530 	sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2531 
2532 	for (i = 0; i < 4; i++)
2533 		sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2534 	sbi->s_def_hash_version = es->s_def_hash_version;
2535 	i = le32_to_cpu(es->s_flags);
2536 	if (i & EXT2_FLAGS_UNSIGNED_HASH)
2537 		sbi->s_hash_unsigned = 3;
2538 	else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2539 #ifdef __CHAR_UNSIGNED__
2540 		es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2541 		sbi->s_hash_unsigned = 3;
2542 #else
2543 		es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2544 #endif
2545 		sb->s_dirt = 1;
2546 	}
2547 
2548 	if (sbi->s_blocks_per_group > blocksize * 8) {
2549 		ext4_msg(sb, KERN_ERR,
2550 		       "#blocks per group too big: %lu",
2551 		       sbi->s_blocks_per_group);
2552 		goto failed_mount;
2553 	}
2554 	if (sbi->s_inodes_per_group > blocksize * 8) {
2555 		ext4_msg(sb, KERN_ERR,
2556 		       "#inodes per group too big: %lu",
2557 		       sbi->s_inodes_per_group);
2558 		goto failed_mount;
2559 	}
2560 
2561 	if (ext4_blocks_count(es) >
2562 		    (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2563 		ext4_msg(sb, KERN_ERR, "filesystem"
2564 			" too large to mount safely");
2565 		if (sizeof(sector_t) < 8)
2566 			ext4_msg(sb, KERN_WARNING, "CONFIG_LBD not enabled");
2567 		goto failed_mount;
2568 	}
2569 
2570 	if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2571 		goto cantfind_ext4;
2572 
2573 	/* check blocks count against device size */
2574 	blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2575 	if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2576 		ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2577 		       "exceeds size of device (%llu blocks)",
2578 		       ext4_blocks_count(es), blocks_count);
2579 		goto failed_mount;
2580 	}
2581 
2582 	/*
2583 	 * It makes no sense for the first data block to be beyond the end
2584 	 * of the filesystem.
2585 	 */
2586 	if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2587                 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2588 			 "block %u is beyond end of filesystem (%llu)",
2589 			 le32_to_cpu(es->s_first_data_block),
2590 			 ext4_blocks_count(es));
2591 		goto failed_mount;
2592 	}
2593 	blocks_count = (ext4_blocks_count(es) -
2594 			le32_to_cpu(es->s_first_data_block) +
2595 			EXT4_BLOCKS_PER_GROUP(sb) - 1);
2596 	do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2597 	if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2598 		ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2599 		       "(block count %llu, first data block %u, "
2600 		       "blocks per group %lu)", sbi->s_groups_count,
2601 		       ext4_blocks_count(es),
2602 		       le32_to_cpu(es->s_first_data_block),
2603 		       EXT4_BLOCKS_PER_GROUP(sb));
2604 		goto failed_mount;
2605 	}
2606 	sbi->s_groups_count = blocks_count;
2607 	db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2608 		   EXT4_DESC_PER_BLOCK(sb);
2609 	sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2610 				    GFP_KERNEL);
2611 	if (sbi->s_group_desc == NULL) {
2612 		ext4_msg(sb, KERN_ERR, "not enough memory");
2613 		goto failed_mount;
2614 	}
2615 
2616 #ifdef CONFIG_PROC_FS
2617 	if (ext4_proc_root)
2618 		sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2619 #endif
2620 
2621 	bgl_lock_init(sbi->s_blockgroup_lock);
2622 
2623 	for (i = 0; i < db_count; i++) {
2624 		block = descriptor_loc(sb, logical_sb_block, i);
2625 		sbi->s_group_desc[i] = sb_bread(sb, block);
2626 		if (!sbi->s_group_desc[i]) {
2627 			ext4_msg(sb, KERN_ERR,
2628 			       "can't read group descriptor %d", i);
2629 			db_count = i;
2630 			goto failed_mount2;
2631 		}
2632 	}
2633 	if (!ext4_check_descriptors(sb)) {
2634 		ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2635 		goto failed_mount2;
2636 	}
2637 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2638 		if (!ext4_fill_flex_info(sb)) {
2639 			ext4_msg(sb, KERN_ERR,
2640 			       "unable to initialize "
2641 			       "flex_bg meta info!");
2642 			goto failed_mount2;
2643 		}
2644 
2645 	sbi->s_gdb_count = db_count;
2646 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2647 	spin_lock_init(&sbi->s_next_gen_lock);
2648 
2649 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
2650 			ext4_count_free_blocks(sb));
2651 	if (!err) {
2652 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
2653 				ext4_count_free_inodes(sb));
2654 	}
2655 	if (!err) {
2656 		err = percpu_counter_init(&sbi->s_dirs_counter,
2657 				ext4_count_dirs(sb));
2658 	}
2659 	if (!err) {
2660 		err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2661 	}
2662 	if (err) {
2663 		ext4_msg(sb, KERN_ERR, "insufficient memory");
2664 		goto failed_mount3;
2665 	}
2666 
2667 	sbi->s_stripe = ext4_get_stripe_size(sbi);
2668 
2669 	/*
2670 	 * set up enough so that it can read an inode
2671 	 */
2672 	if (!test_opt(sb, NOLOAD) &&
2673 	    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2674 		sb->s_op = &ext4_sops;
2675 	else
2676 		sb->s_op = &ext4_nojournal_sops;
2677 	sb->s_export_op = &ext4_export_ops;
2678 	sb->s_xattr = ext4_xattr_handlers;
2679 #ifdef CONFIG_QUOTA
2680 	sb->s_qcop = &ext4_qctl_operations;
2681 	sb->dq_op = &ext4_quota_operations;
2682 #endif
2683 	INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2684 	mutex_init(&sbi->s_orphan_lock);
2685 	mutex_init(&sbi->s_resize_lock);
2686 
2687 	sb->s_root = NULL;
2688 
2689 	needs_recovery = (es->s_last_orphan != 0 ||
2690 			  EXT4_HAS_INCOMPAT_FEATURE(sb,
2691 				    EXT4_FEATURE_INCOMPAT_RECOVER));
2692 
2693 	/*
2694 	 * The first inode we look at is the journal inode.  Don't try
2695 	 * root first: it may be modified in the journal!
2696 	 */
2697 	if (!test_opt(sb, NOLOAD) &&
2698 	    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2699 		if (ext4_load_journal(sb, es, journal_devnum))
2700 			goto failed_mount3;
2701 		if (!(sb->s_flags & MS_RDONLY) &&
2702 		    EXT4_SB(sb)->s_journal->j_failed_commit) {
2703 			ext4_msg(sb, KERN_CRIT, "error: "
2704 			       "ext4_fill_super: Journal transaction "
2705 			       "%u is corrupt",
2706 			       EXT4_SB(sb)->s_journal->j_failed_commit);
2707 			if (test_opt(sb, ERRORS_RO)) {
2708 				ext4_msg(sb, KERN_CRIT,
2709 				       "Mounting filesystem read-only");
2710 				sb->s_flags |= MS_RDONLY;
2711 				EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2712 				es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2713 			}
2714 			if (test_opt(sb, ERRORS_PANIC)) {
2715 				EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2716 				es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2717 				ext4_commit_super(sb, 1);
2718 				goto failed_mount4;
2719 			}
2720 		}
2721 	} else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2722 	      EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2723 		ext4_msg(sb, KERN_ERR, "required journal recovery "
2724 		       "suppressed and not mounted read-only");
2725 		goto failed_mount4;
2726 	} else {
2727 		clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2728 		set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2729 		sbi->s_journal = NULL;
2730 		needs_recovery = 0;
2731 		goto no_journal;
2732 	}
2733 
2734 	if (ext4_blocks_count(es) > 0xffffffffULL &&
2735 	    !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2736 				       JBD2_FEATURE_INCOMPAT_64BIT)) {
2737 		ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2738 		goto failed_mount4;
2739 	}
2740 
2741 	if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2742 		jbd2_journal_set_features(sbi->s_journal,
2743 				JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2744 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2745 	} else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2746 		jbd2_journal_set_features(sbi->s_journal,
2747 				JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2748 		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2749 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2750 	} else {
2751 		jbd2_journal_clear_features(sbi->s_journal,
2752 				JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2753 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2754 	}
2755 
2756 	/* We have now updated the journal if required, so we can
2757 	 * validate the data journaling mode. */
2758 	switch (test_opt(sb, DATA_FLAGS)) {
2759 	case 0:
2760 		/* No mode set, assume a default based on the journal
2761 		 * capabilities: ORDERED_DATA if the journal can
2762 		 * cope, else JOURNAL_DATA
2763 		 */
2764 		if (jbd2_journal_check_available_features
2765 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2766 			set_opt(sbi->s_mount_opt, ORDERED_DATA);
2767 		else
2768 			set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2769 		break;
2770 
2771 	case EXT4_MOUNT_ORDERED_DATA:
2772 	case EXT4_MOUNT_WRITEBACK_DATA:
2773 		if (!jbd2_journal_check_available_features
2774 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2775 			ext4_msg(sb, KERN_ERR, "Journal does not support "
2776 			       "requested data journaling mode");
2777 			goto failed_mount4;
2778 		}
2779 	default:
2780 		break;
2781 	}
2782 	set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2783 
2784 no_journal:
2785 
2786 	if (test_opt(sb, NOBH)) {
2787 		if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2788 			ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2789 				"its supported only with writeback mode");
2790 			clear_opt(sbi->s_mount_opt, NOBH);
2791 		}
2792 	}
2793 	/*
2794 	 * The jbd2_journal_load will have done any necessary log recovery,
2795 	 * so we can safely mount the rest of the filesystem now.
2796 	 */
2797 
2798 	root = ext4_iget(sb, EXT4_ROOT_INO);
2799 	if (IS_ERR(root)) {
2800 		ext4_msg(sb, KERN_ERR, "get root inode failed");
2801 		ret = PTR_ERR(root);
2802 		goto failed_mount4;
2803 	}
2804 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2805 		iput(root);
2806 		ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
2807 		goto failed_mount4;
2808 	}
2809 	sb->s_root = d_alloc_root(root);
2810 	if (!sb->s_root) {
2811 		ext4_msg(sb, KERN_ERR, "get root dentry failed");
2812 		iput(root);
2813 		ret = -ENOMEM;
2814 		goto failed_mount4;
2815 	}
2816 
2817 	ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2818 
2819 	/* determine the minimum size of new large inodes, if present */
2820 	if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2821 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2822 						     EXT4_GOOD_OLD_INODE_SIZE;
2823 		if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2824 				       EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2825 			if (sbi->s_want_extra_isize <
2826 			    le16_to_cpu(es->s_want_extra_isize))
2827 				sbi->s_want_extra_isize =
2828 					le16_to_cpu(es->s_want_extra_isize);
2829 			if (sbi->s_want_extra_isize <
2830 			    le16_to_cpu(es->s_min_extra_isize))
2831 				sbi->s_want_extra_isize =
2832 					le16_to_cpu(es->s_min_extra_isize);
2833 		}
2834 	}
2835 	/* Check if enough inode space is available */
2836 	if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2837 							sbi->s_inode_size) {
2838 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2839 						       EXT4_GOOD_OLD_INODE_SIZE;
2840 		ext4_msg(sb, KERN_INFO, "required extra inode space not"
2841 			 "available");
2842 	}
2843 
2844 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2845 		ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2846 			 "requested data journaling mode");
2847 		clear_opt(sbi->s_mount_opt, DELALLOC);
2848 	} else if (test_opt(sb, DELALLOC))
2849 		ext4_msg(sb, KERN_INFO, "delayed allocation enabled");
2850 
2851 	err = ext4_setup_system_zone(sb);
2852 	if (err) {
2853 		ext4_msg(sb, KERN_ERR, "failed to initialize system "
2854 			 "zone (%d)\n", err);
2855 		goto failed_mount4;
2856 	}
2857 
2858 	ext4_ext_init(sb);
2859 	err = ext4_mb_init(sb, needs_recovery);
2860 	if (err) {
2861 		ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2862 			 err);
2863 		goto failed_mount4;
2864 	}
2865 
2866 	sbi->s_kobj.kset = ext4_kset;
2867 	init_completion(&sbi->s_kobj_unregister);
2868 	err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2869 				   "%s", sb->s_id);
2870 	if (err) {
2871 		ext4_mb_release(sb);
2872 		ext4_ext_release(sb);
2873 		goto failed_mount4;
2874 	};
2875 
2876 	EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2877 	ext4_orphan_cleanup(sb, es);
2878 	EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2879 	if (needs_recovery) {
2880 		ext4_msg(sb, KERN_INFO, "recovery complete");
2881 		ext4_mark_recovery_complete(sb, es);
2882 	}
2883 	if (EXT4_SB(sb)->s_journal) {
2884 		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2885 			descr = " journalled data mode";
2886 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2887 			descr = " ordered data mode";
2888 		else
2889 			descr = " writeback data mode";
2890 	} else
2891 		descr = "out journal";
2892 
2893 	ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
2894 
2895 	lock_kernel();
2896 	return 0;
2897 
2898 cantfind_ext4:
2899 	if (!silent)
2900 		ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
2901 	goto failed_mount;
2902 
2903 failed_mount4:
2904 	ext4_msg(sb, KERN_ERR, "mount failed");
2905 	ext4_release_system_zone(sb);
2906 	if (sbi->s_journal) {
2907 		jbd2_journal_destroy(sbi->s_journal);
2908 		sbi->s_journal = NULL;
2909 	}
2910 failed_mount3:
2911 	if (sbi->s_flex_groups) {
2912 		if (is_vmalloc_addr(sbi->s_flex_groups))
2913 			vfree(sbi->s_flex_groups);
2914 		else
2915 			kfree(sbi->s_flex_groups);
2916 	}
2917 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
2918 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
2919 	percpu_counter_destroy(&sbi->s_dirs_counter);
2920 	percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2921 failed_mount2:
2922 	for (i = 0; i < db_count; i++)
2923 		brelse(sbi->s_group_desc[i]);
2924 	kfree(sbi->s_group_desc);
2925 failed_mount:
2926 	if (sbi->s_proc) {
2927 		remove_proc_entry(sb->s_id, ext4_proc_root);
2928 	}
2929 #ifdef CONFIG_QUOTA
2930 	for (i = 0; i < MAXQUOTAS; i++)
2931 		kfree(sbi->s_qf_names[i]);
2932 #endif
2933 	ext4_blkdev_remove(sbi);
2934 	brelse(bh);
2935 out_fail:
2936 	sb->s_fs_info = NULL;
2937 	kfree(sbi->s_blockgroup_lock);
2938 	kfree(sbi);
2939 	lock_kernel();
2940 	return ret;
2941 }
2942 
2943 /*
2944  * Setup any per-fs journal parameters now.  We'll do this both on
2945  * initial mount, once the journal has been initialised but before we've
2946  * done any recovery; and again on any subsequent remount.
2947  */
2948 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2949 {
2950 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2951 
2952 	journal->j_commit_interval = sbi->s_commit_interval;
2953 	journal->j_min_batch_time = sbi->s_min_batch_time;
2954 	journal->j_max_batch_time = sbi->s_max_batch_time;
2955 
2956 	spin_lock(&journal->j_state_lock);
2957 	if (test_opt(sb, BARRIER))
2958 		journal->j_flags |= JBD2_BARRIER;
2959 	else
2960 		journal->j_flags &= ~JBD2_BARRIER;
2961 	if (test_opt(sb, DATA_ERR_ABORT))
2962 		journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2963 	else
2964 		journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2965 	spin_unlock(&journal->j_state_lock);
2966 }
2967 
2968 static journal_t *ext4_get_journal(struct super_block *sb,
2969 				   unsigned int journal_inum)
2970 {
2971 	struct inode *journal_inode;
2972 	journal_t *journal;
2973 
2974 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2975 
2976 	/* First, test for the existence of a valid inode on disk.  Bad
2977 	 * things happen if we iget() an unused inode, as the subsequent
2978 	 * iput() will try to delete it. */
2979 
2980 	journal_inode = ext4_iget(sb, journal_inum);
2981 	if (IS_ERR(journal_inode)) {
2982 		ext4_msg(sb, KERN_ERR, "no journal found");
2983 		return NULL;
2984 	}
2985 	if (!journal_inode->i_nlink) {
2986 		make_bad_inode(journal_inode);
2987 		iput(journal_inode);
2988 		ext4_msg(sb, KERN_ERR, "journal inode is deleted");
2989 		return NULL;
2990 	}
2991 
2992 	jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2993 		  journal_inode, journal_inode->i_size);
2994 	if (!S_ISREG(journal_inode->i_mode)) {
2995 		ext4_msg(sb, KERN_ERR, "invalid journal inode");
2996 		iput(journal_inode);
2997 		return NULL;
2998 	}
2999 
3000 	journal = jbd2_journal_init_inode(journal_inode);
3001 	if (!journal) {
3002 		ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3003 		iput(journal_inode);
3004 		return NULL;
3005 	}
3006 	journal->j_private = sb;
3007 	ext4_init_journal_params(sb, journal);
3008 	return journal;
3009 }
3010 
3011 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3012 				       dev_t j_dev)
3013 {
3014 	struct buffer_head *bh;
3015 	journal_t *journal;
3016 	ext4_fsblk_t start;
3017 	ext4_fsblk_t len;
3018 	int hblock, blocksize;
3019 	ext4_fsblk_t sb_block;
3020 	unsigned long offset;
3021 	struct ext4_super_block *es;
3022 	struct block_device *bdev;
3023 
3024 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3025 
3026 	bdev = ext4_blkdev_get(j_dev, sb);
3027 	if (bdev == NULL)
3028 		return NULL;
3029 
3030 	if (bd_claim(bdev, sb)) {
3031 		ext4_msg(sb, KERN_ERR,
3032 			"failed to claim external journal device");
3033 		blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3034 		return NULL;
3035 	}
3036 
3037 	blocksize = sb->s_blocksize;
3038 	hblock = bdev_logical_block_size(bdev);
3039 	if (blocksize < hblock) {
3040 		ext4_msg(sb, KERN_ERR,
3041 			"blocksize too small for journal device");
3042 		goto out_bdev;
3043 	}
3044 
3045 	sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3046 	offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3047 	set_blocksize(bdev, blocksize);
3048 	if (!(bh = __bread(bdev, sb_block, blocksize))) {
3049 		ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3050 		       "external journal");
3051 		goto out_bdev;
3052 	}
3053 
3054 	es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3055 	if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3056 	    !(le32_to_cpu(es->s_feature_incompat) &
3057 	      EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3058 		ext4_msg(sb, KERN_ERR, "external journal has "
3059 					"bad superblock");
3060 		brelse(bh);
3061 		goto out_bdev;
3062 	}
3063 
3064 	if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3065 		ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3066 		brelse(bh);
3067 		goto out_bdev;
3068 	}
3069 
3070 	len = ext4_blocks_count(es);
3071 	start = sb_block + 1;
3072 	brelse(bh);	/* we're done with the superblock */
3073 
3074 	journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3075 					start, len, blocksize);
3076 	if (!journal) {
3077 		ext4_msg(sb, KERN_ERR, "failed to create device journal");
3078 		goto out_bdev;
3079 	}
3080 	journal->j_private = sb;
3081 	ll_rw_block(READ, 1, &journal->j_sb_buffer);
3082 	wait_on_buffer(journal->j_sb_buffer);
3083 	if (!buffer_uptodate(journal->j_sb_buffer)) {
3084 		ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3085 		goto out_journal;
3086 	}
3087 	if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3088 		ext4_msg(sb, KERN_ERR, "External journal has more than one "
3089 					"user (unsupported) - %d",
3090 			be32_to_cpu(journal->j_superblock->s_nr_users));
3091 		goto out_journal;
3092 	}
3093 	EXT4_SB(sb)->journal_bdev = bdev;
3094 	ext4_init_journal_params(sb, journal);
3095 	return journal;
3096 
3097 out_journal:
3098 	jbd2_journal_destroy(journal);
3099 out_bdev:
3100 	ext4_blkdev_put(bdev);
3101 	return NULL;
3102 }
3103 
3104 static int ext4_load_journal(struct super_block *sb,
3105 			     struct ext4_super_block *es,
3106 			     unsigned long journal_devnum)
3107 {
3108 	journal_t *journal;
3109 	unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3110 	dev_t journal_dev;
3111 	int err = 0;
3112 	int really_read_only;
3113 
3114 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3115 
3116 	if (journal_devnum &&
3117 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3118 		ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3119 			"numbers have changed");
3120 		journal_dev = new_decode_dev(journal_devnum);
3121 	} else
3122 		journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3123 
3124 	really_read_only = bdev_read_only(sb->s_bdev);
3125 
3126 	/*
3127 	 * Are we loading a blank journal or performing recovery after a
3128 	 * crash?  For recovery, we need to check in advance whether we
3129 	 * can get read-write access to the device.
3130 	 */
3131 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3132 		if (sb->s_flags & MS_RDONLY) {
3133 			ext4_msg(sb, KERN_INFO, "INFO: recovery "
3134 					"required on readonly filesystem");
3135 			if (really_read_only) {
3136 				ext4_msg(sb, KERN_ERR, "write access "
3137 					"unavailable, cannot proceed");
3138 				return -EROFS;
3139 			}
3140 			ext4_msg(sb, KERN_INFO, "write access will "
3141 			       "be enabled during recovery");
3142 		}
3143 	}
3144 
3145 	if (journal_inum && journal_dev) {
3146 		ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3147 		       "and inode journals!");
3148 		return -EINVAL;
3149 	}
3150 
3151 	if (journal_inum) {
3152 		if (!(journal = ext4_get_journal(sb, journal_inum)))
3153 			return -EINVAL;
3154 	} else {
3155 		if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3156 			return -EINVAL;
3157 	}
3158 
3159 	if (journal->j_flags & JBD2_BARRIER)
3160 		ext4_msg(sb, KERN_INFO, "barriers enabled");
3161 	else
3162 		ext4_msg(sb, KERN_INFO, "barriers disabled");
3163 
3164 	if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3165 		err = jbd2_journal_update_format(journal);
3166 		if (err)  {
3167 			ext4_msg(sb, KERN_ERR, "error updating journal");
3168 			jbd2_journal_destroy(journal);
3169 			return err;
3170 		}
3171 	}
3172 
3173 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3174 		err = jbd2_journal_wipe(journal, !really_read_only);
3175 	if (!err)
3176 		err = jbd2_journal_load(journal);
3177 
3178 	if (err) {
3179 		ext4_msg(sb, KERN_ERR, "error loading journal");
3180 		jbd2_journal_destroy(journal);
3181 		return err;
3182 	}
3183 
3184 	EXT4_SB(sb)->s_journal = journal;
3185 	ext4_clear_journal_err(sb, es);
3186 
3187 	if (journal_devnum &&
3188 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3189 		es->s_journal_dev = cpu_to_le32(journal_devnum);
3190 
3191 		/* Make sure we flush the recovery flag to disk. */
3192 		ext4_commit_super(sb, 1);
3193 	}
3194 
3195 	return 0;
3196 }
3197 
3198 static int ext4_commit_super(struct super_block *sb, int sync)
3199 {
3200 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3201 	struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3202 	int error = 0;
3203 
3204 	if (!sbh)
3205 		return error;
3206 	if (buffer_write_io_error(sbh)) {
3207 		/*
3208 		 * Oh, dear.  A previous attempt to write the
3209 		 * superblock failed.  This could happen because the
3210 		 * USB device was yanked out.  Or it could happen to
3211 		 * be a transient write error and maybe the block will
3212 		 * be remapped.  Nothing we can do but to retry the
3213 		 * write and hope for the best.
3214 		 */
3215 		ext4_msg(sb, KERN_ERR, "previous I/O error to "
3216 		       "superblock detected");
3217 		clear_buffer_write_io_error(sbh);
3218 		set_buffer_uptodate(sbh);
3219 	}
3220 	es->s_wtime = cpu_to_le32(get_seconds());
3221 	es->s_kbytes_written =
3222 		cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3223 			    ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3224 			      EXT4_SB(sb)->s_sectors_written_start) >> 1));
3225 	ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3226 					&EXT4_SB(sb)->s_freeblocks_counter));
3227 	es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3228 					&EXT4_SB(sb)->s_freeinodes_counter));
3229 	sb->s_dirt = 0;
3230 	BUFFER_TRACE(sbh, "marking dirty");
3231 	mark_buffer_dirty(sbh);
3232 	if (sync) {
3233 		error = sync_dirty_buffer(sbh);
3234 		if (error)
3235 			return error;
3236 
3237 		error = buffer_write_io_error(sbh);
3238 		if (error) {
3239 			ext4_msg(sb, KERN_ERR, "I/O error while writing "
3240 			       "superblock");
3241 			clear_buffer_write_io_error(sbh);
3242 			set_buffer_uptodate(sbh);
3243 		}
3244 	}
3245 	return error;
3246 }
3247 
3248 /*
3249  * Have we just finished recovery?  If so, and if we are mounting (or
3250  * remounting) the filesystem readonly, then we will end up with a
3251  * consistent fs on disk.  Record that fact.
3252  */
3253 static void ext4_mark_recovery_complete(struct super_block *sb,
3254 					struct ext4_super_block *es)
3255 {
3256 	journal_t *journal = EXT4_SB(sb)->s_journal;
3257 
3258 	if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3259 		BUG_ON(journal != NULL);
3260 		return;
3261 	}
3262 	jbd2_journal_lock_updates(journal);
3263 	if (jbd2_journal_flush(journal) < 0)
3264 		goto out;
3265 
3266 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3267 	    sb->s_flags & MS_RDONLY) {
3268 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3269 		ext4_commit_super(sb, 1);
3270 	}
3271 
3272 out:
3273 	jbd2_journal_unlock_updates(journal);
3274 }
3275 
3276 /*
3277  * If we are mounting (or read-write remounting) a filesystem whose journal
3278  * has recorded an error from a previous lifetime, move that error to the
3279  * main filesystem now.
3280  */
3281 static void ext4_clear_journal_err(struct super_block *sb,
3282 				   struct ext4_super_block *es)
3283 {
3284 	journal_t *journal;
3285 	int j_errno;
3286 	const char *errstr;
3287 
3288 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3289 
3290 	journal = EXT4_SB(sb)->s_journal;
3291 
3292 	/*
3293 	 * Now check for any error status which may have been recorded in the
3294 	 * journal by a prior ext4_error() or ext4_abort()
3295 	 */
3296 
3297 	j_errno = jbd2_journal_errno(journal);
3298 	if (j_errno) {
3299 		char nbuf[16];
3300 
3301 		errstr = ext4_decode_error(sb, j_errno, nbuf);
3302 		ext4_warning(sb, __func__, "Filesystem error recorded "
3303 			     "from previous mount: %s", errstr);
3304 		ext4_warning(sb, __func__, "Marking fs in need of "
3305 			     "filesystem check.");
3306 
3307 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3308 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3309 		ext4_commit_super(sb, 1);
3310 
3311 		jbd2_journal_clear_err(journal);
3312 	}
3313 }
3314 
3315 /*
3316  * Force the running and committing transactions to commit,
3317  * and wait on the commit.
3318  */
3319 int ext4_force_commit(struct super_block *sb)
3320 {
3321 	journal_t *journal;
3322 	int ret = 0;
3323 
3324 	if (sb->s_flags & MS_RDONLY)
3325 		return 0;
3326 
3327 	journal = EXT4_SB(sb)->s_journal;
3328 	if (journal)
3329 		ret = ext4_journal_force_commit(journal);
3330 
3331 	return ret;
3332 }
3333 
3334 static void ext4_write_super(struct super_block *sb)
3335 {
3336 	ext4_commit_super(sb, 1);
3337 }
3338 
3339 static int ext4_sync_fs(struct super_block *sb, int wait)
3340 {
3341 	int ret = 0;
3342 	tid_t target;
3343 
3344 	trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
3345 	if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
3346 		if (wait)
3347 			jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
3348 	}
3349 	return ret;
3350 }
3351 
3352 /*
3353  * LVM calls this function before a (read-only) snapshot is created.  This
3354  * gives us a chance to flush the journal completely and mark the fs clean.
3355  */
3356 static int ext4_freeze(struct super_block *sb)
3357 {
3358 	int error = 0;
3359 	journal_t *journal;
3360 
3361 	if (sb->s_flags & MS_RDONLY)
3362 		return 0;
3363 
3364 	journal = EXT4_SB(sb)->s_journal;
3365 
3366 	/* Now we set up the journal barrier. */
3367 	jbd2_journal_lock_updates(journal);
3368 
3369 	/*
3370 	 * Don't clear the needs_recovery flag if we failed to flush
3371 	 * the journal.
3372 	 */
3373 	error = jbd2_journal_flush(journal);
3374 	if (error < 0) {
3375 	out:
3376 		jbd2_journal_unlock_updates(journal);
3377 		return error;
3378 	}
3379 
3380 	/* Journal blocked and flushed, clear needs_recovery flag. */
3381 	EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3382 	error = ext4_commit_super(sb, 1);
3383 	if (error)
3384 		goto out;
3385 	return 0;
3386 }
3387 
3388 /*
3389  * Called by LVM after the snapshot is done.  We need to reset the RECOVER
3390  * flag here, even though the filesystem is not technically dirty yet.
3391  */
3392 static int ext4_unfreeze(struct super_block *sb)
3393 {
3394 	if (sb->s_flags & MS_RDONLY)
3395 		return 0;
3396 
3397 	lock_super(sb);
3398 	/* Reset the needs_recovery flag before the fs is unlocked. */
3399 	EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3400 	ext4_commit_super(sb, 1);
3401 	unlock_super(sb);
3402 	jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3403 	return 0;
3404 }
3405 
3406 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3407 {
3408 	struct ext4_super_block *es;
3409 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3410 	ext4_fsblk_t n_blocks_count = 0;
3411 	unsigned long old_sb_flags;
3412 	struct ext4_mount_options old_opts;
3413 	ext4_group_t g;
3414 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3415 	int err;
3416 #ifdef CONFIG_QUOTA
3417 	int i;
3418 #endif
3419 
3420 	/* Store the original options */
3421 	old_sb_flags = sb->s_flags;
3422 	old_opts.s_mount_opt = sbi->s_mount_opt;
3423 	old_opts.s_resuid = sbi->s_resuid;
3424 	old_opts.s_resgid = sbi->s_resgid;
3425 	old_opts.s_commit_interval = sbi->s_commit_interval;
3426 	old_opts.s_min_batch_time = sbi->s_min_batch_time;
3427 	old_opts.s_max_batch_time = sbi->s_max_batch_time;
3428 #ifdef CONFIG_QUOTA
3429 	old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3430 	for (i = 0; i < MAXQUOTAS; i++)
3431 		old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3432 #endif
3433 	if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3434 		journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3435 
3436 	/*
3437 	 * Allow the "check" option to be passed as a remount option.
3438 	 */
3439 	if (!parse_options(data, sb, NULL, &journal_ioprio,
3440 			   &n_blocks_count, 1)) {
3441 		err = -EINVAL;
3442 		goto restore_opts;
3443 	}
3444 
3445 	if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
3446 		ext4_abort(sb, __func__, "Abort forced by user");
3447 
3448 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3449 		((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3450 
3451 	es = sbi->s_es;
3452 
3453 	if (sbi->s_journal) {
3454 		ext4_init_journal_params(sb, sbi->s_journal);
3455 		set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3456 	}
3457 
3458 	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3459 		n_blocks_count > ext4_blocks_count(es)) {
3460 		if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
3461 			err = -EROFS;
3462 			goto restore_opts;
3463 		}
3464 
3465 		if (*flags & MS_RDONLY) {
3466 			/*
3467 			 * First of all, the unconditional stuff we have to do
3468 			 * to disable replay of the journal when we next remount
3469 			 */
3470 			sb->s_flags |= MS_RDONLY;
3471 
3472 			/*
3473 			 * OK, test if we are remounting a valid rw partition
3474 			 * readonly, and if so set the rdonly flag and then
3475 			 * mark the partition as valid again.
3476 			 */
3477 			if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3478 			    (sbi->s_mount_state & EXT4_VALID_FS))
3479 				es->s_state = cpu_to_le16(sbi->s_mount_state);
3480 
3481 			if (sbi->s_journal)
3482 				ext4_mark_recovery_complete(sb, es);
3483 		} else {
3484 			int ret;
3485 			if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3486 					~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3487 				ext4_msg(sb, KERN_WARNING, "couldn't "
3488 				       "remount RDWR because of unsupported "
3489 				       "optional features (%x)",
3490 				(le32_to_cpu(sbi->s_es->s_feature_ro_compat) &
3491 					~EXT4_FEATURE_RO_COMPAT_SUPP));
3492 				err = -EROFS;
3493 				goto restore_opts;
3494 			}
3495 
3496 			/*
3497 			 * Make sure the group descriptor checksums
3498 			 * are sane.  If they aren't, refuse to remount r/w.
3499 			 */
3500 			for (g = 0; g < sbi->s_groups_count; g++) {
3501 				struct ext4_group_desc *gdp =
3502 					ext4_get_group_desc(sb, g, NULL);
3503 
3504 				if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3505 					ext4_msg(sb, KERN_ERR,
3506 	       "ext4_remount: Checksum for group %u failed (%u!=%u)",
3507 		g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3508 					       le16_to_cpu(gdp->bg_checksum));
3509 					err = -EINVAL;
3510 					goto restore_opts;
3511 				}
3512 			}
3513 
3514 			/*
3515 			 * If we have an unprocessed orphan list hanging
3516 			 * around from a previously readonly bdev mount,
3517 			 * require a full umount/remount for now.
3518 			 */
3519 			if (es->s_last_orphan) {
3520 				ext4_msg(sb, KERN_WARNING, "Couldn't "
3521 				       "remount RDWR because of unprocessed "
3522 				       "orphan inode list.  Please "
3523 				       "umount/remount instead");
3524 				err = -EINVAL;
3525 				goto restore_opts;
3526 			}
3527 
3528 			/*
3529 			 * Mounting a RDONLY partition read-write, so reread
3530 			 * and store the current valid flag.  (It may have
3531 			 * been changed by e2fsck since we originally mounted
3532 			 * the partition.)
3533 			 */
3534 			if (sbi->s_journal)
3535 				ext4_clear_journal_err(sb, es);
3536 			sbi->s_mount_state = le16_to_cpu(es->s_state);
3537 			if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3538 				goto restore_opts;
3539 			if (!ext4_setup_super(sb, es, 0))
3540 				sb->s_flags &= ~MS_RDONLY;
3541 		}
3542 	}
3543 	ext4_setup_system_zone(sb);
3544 	if (sbi->s_journal == NULL)
3545 		ext4_commit_super(sb, 1);
3546 
3547 #ifdef CONFIG_QUOTA
3548 	/* Release old quota file names */
3549 	for (i = 0; i < MAXQUOTAS; i++)
3550 		if (old_opts.s_qf_names[i] &&
3551 		    old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3552 			kfree(old_opts.s_qf_names[i]);
3553 #endif
3554 	return 0;
3555 
3556 restore_opts:
3557 	sb->s_flags = old_sb_flags;
3558 	sbi->s_mount_opt = old_opts.s_mount_opt;
3559 	sbi->s_resuid = old_opts.s_resuid;
3560 	sbi->s_resgid = old_opts.s_resgid;
3561 	sbi->s_commit_interval = old_opts.s_commit_interval;
3562 	sbi->s_min_batch_time = old_opts.s_min_batch_time;
3563 	sbi->s_max_batch_time = old_opts.s_max_batch_time;
3564 #ifdef CONFIG_QUOTA
3565 	sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3566 	for (i = 0; i < MAXQUOTAS; i++) {
3567 		if (sbi->s_qf_names[i] &&
3568 		    old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3569 			kfree(sbi->s_qf_names[i]);
3570 		sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3571 	}
3572 #endif
3573 	return err;
3574 }
3575 
3576 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3577 {
3578 	struct super_block *sb = dentry->d_sb;
3579 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3580 	struct ext4_super_block *es = sbi->s_es;
3581 	u64 fsid;
3582 
3583 	if (test_opt(sb, MINIX_DF)) {
3584 		sbi->s_overhead_last = 0;
3585 	} else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3586 		ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3587 		ext4_fsblk_t overhead = 0;
3588 
3589 		/*
3590 		 * Compute the overhead (FS structures).  This is constant
3591 		 * for a given filesystem unless the number of block groups
3592 		 * changes so we cache the previous value until it does.
3593 		 */
3594 
3595 		/*
3596 		 * All of the blocks before first_data_block are
3597 		 * overhead
3598 		 */
3599 		overhead = le32_to_cpu(es->s_first_data_block);
3600 
3601 		/*
3602 		 * Add the overhead attributed to the superblock and
3603 		 * block group descriptors.  If the sparse superblocks
3604 		 * feature is turned on, then not all groups have this.
3605 		 */
3606 		for (i = 0; i < ngroups; i++) {
3607 			overhead += ext4_bg_has_super(sb, i) +
3608 				ext4_bg_num_gdb(sb, i);
3609 			cond_resched();
3610 		}
3611 
3612 		/*
3613 		 * Every block group has an inode bitmap, a block
3614 		 * bitmap, and an inode table.
3615 		 */
3616 		overhead += ngroups * (2 + sbi->s_itb_per_group);
3617 		sbi->s_overhead_last = overhead;
3618 		smp_wmb();
3619 		sbi->s_blocks_last = ext4_blocks_count(es);
3620 	}
3621 
3622 	buf->f_type = EXT4_SUPER_MAGIC;
3623 	buf->f_bsize = sb->s_blocksize;
3624 	buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3625 	buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3626 		       percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3627 	ext4_free_blocks_count_set(es, buf->f_bfree);
3628 	buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3629 	if (buf->f_bfree < ext4_r_blocks_count(es))
3630 		buf->f_bavail = 0;
3631 	buf->f_files = le32_to_cpu(es->s_inodes_count);
3632 	buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3633 	es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3634 	buf->f_namelen = EXT4_NAME_LEN;
3635 	fsid = le64_to_cpup((void *)es->s_uuid) ^
3636 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3637 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3638 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3639 
3640 	return 0;
3641 }
3642 
3643 /* Helper function for writing quotas on sync - we need to start transaction
3644  * before quota file is locked for write. Otherwise the are possible deadlocks:
3645  * Process 1                         Process 2
3646  * ext4_create()                     quota_sync()
3647  *   jbd2_journal_start()                  write_dquot()
3648  *   vfs_dq_init()                         down(dqio_mutex)
3649  *     down(dqio_mutex)                    jbd2_journal_start()
3650  *
3651  */
3652 
3653 #ifdef CONFIG_QUOTA
3654 
3655 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3656 {
3657 	return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3658 }
3659 
3660 static int ext4_write_dquot(struct dquot *dquot)
3661 {
3662 	int ret, err;
3663 	handle_t *handle;
3664 	struct inode *inode;
3665 
3666 	inode = dquot_to_inode(dquot);
3667 	handle = ext4_journal_start(inode,
3668 				    EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3669 	if (IS_ERR(handle))
3670 		return PTR_ERR(handle);
3671 	ret = dquot_commit(dquot);
3672 	err = ext4_journal_stop(handle);
3673 	if (!ret)
3674 		ret = err;
3675 	return ret;
3676 }
3677 
3678 static int ext4_acquire_dquot(struct dquot *dquot)
3679 {
3680 	int ret, err;
3681 	handle_t *handle;
3682 
3683 	handle = ext4_journal_start(dquot_to_inode(dquot),
3684 				    EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3685 	if (IS_ERR(handle))
3686 		return PTR_ERR(handle);
3687 	ret = dquot_acquire(dquot);
3688 	err = ext4_journal_stop(handle);
3689 	if (!ret)
3690 		ret = err;
3691 	return ret;
3692 }
3693 
3694 static int ext4_release_dquot(struct dquot *dquot)
3695 {
3696 	int ret, err;
3697 	handle_t *handle;
3698 
3699 	handle = ext4_journal_start(dquot_to_inode(dquot),
3700 				    EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3701 	if (IS_ERR(handle)) {
3702 		/* Release dquot anyway to avoid endless cycle in dqput() */
3703 		dquot_release(dquot);
3704 		return PTR_ERR(handle);
3705 	}
3706 	ret = dquot_release(dquot);
3707 	err = ext4_journal_stop(handle);
3708 	if (!ret)
3709 		ret = err;
3710 	return ret;
3711 }
3712 
3713 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3714 {
3715 	/* Are we journaling quotas? */
3716 	if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3717 	    EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3718 		dquot_mark_dquot_dirty(dquot);
3719 		return ext4_write_dquot(dquot);
3720 	} else {
3721 		return dquot_mark_dquot_dirty(dquot);
3722 	}
3723 }
3724 
3725 static int ext4_write_info(struct super_block *sb, int type)
3726 {
3727 	int ret, err;
3728 	handle_t *handle;
3729 
3730 	/* Data block + inode block */
3731 	handle = ext4_journal_start(sb->s_root->d_inode, 2);
3732 	if (IS_ERR(handle))
3733 		return PTR_ERR(handle);
3734 	ret = dquot_commit_info(sb, type);
3735 	err = ext4_journal_stop(handle);
3736 	if (!ret)
3737 		ret = err;
3738 	return ret;
3739 }
3740 
3741 /*
3742  * Turn on quotas during mount time - we need to find
3743  * the quota file and such...
3744  */
3745 static int ext4_quota_on_mount(struct super_block *sb, int type)
3746 {
3747 	return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3748 				  EXT4_SB(sb)->s_jquota_fmt, type);
3749 }
3750 
3751 /*
3752  * Standard function to be called on quota_on
3753  */
3754 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3755 			 char *name, int remount)
3756 {
3757 	int err;
3758 	struct path path;
3759 
3760 	if (!test_opt(sb, QUOTA))
3761 		return -EINVAL;
3762 	/* When remounting, no checks are needed and in fact, name is NULL */
3763 	if (remount)
3764 		return vfs_quota_on(sb, type, format_id, name, remount);
3765 
3766 	err = kern_path(name, LOOKUP_FOLLOW, &path);
3767 	if (err)
3768 		return err;
3769 
3770 	/* Quotafile not on the same filesystem? */
3771 	if (path.mnt->mnt_sb != sb) {
3772 		path_put(&path);
3773 		return -EXDEV;
3774 	}
3775 	/* Journaling quota? */
3776 	if (EXT4_SB(sb)->s_qf_names[type]) {
3777 		/* Quotafile not in fs root? */
3778 		if (path.dentry->d_parent != sb->s_root)
3779 			ext4_msg(sb, KERN_WARNING,
3780 				"Quota file not on filesystem root. "
3781 				"Journaled quota will not work");
3782 	}
3783 
3784 	/*
3785 	 * When we journal data on quota file, we have to flush journal to see
3786 	 * all updates to the file when we bypass pagecache...
3787 	 */
3788 	if (EXT4_SB(sb)->s_journal &&
3789 	    ext4_should_journal_data(path.dentry->d_inode)) {
3790 		/*
3791 		 * We don't need to lock updates but journal_flush() could
3792 		 * otherwise be livelocked...
3793 		 */
3794 		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3795 		err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3796 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3797 		if (err) {
3798 			path_put(&path);
3799 			return err;
3800 		}
3801 	}
3802 
3803 	err = vfs_quota_on_path(sb, type, format_id, &path);
3804 	path_put(&path);
3805 	return err;
3806 }
3807 
3808 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3809  * acquiring the locks... As quota files are never truncated and quota code
3810  * itself serializes the operations (and noone else should touch the files)
3811  * we don't have to be afraid of races */
3812 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3813 			       size_t len, loff_t off)
3814 {
3815 	struct inode *inode = sb_dqopt(sb)->files[type];
3816 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3817 	int err = 0;
3818 	int offset = off & (sb->s_blocksize - 1);
3819 	int tocopy;
3820 	size_t toread;
3821 	struct buffer_head *bh;
3822 	loff_t i_size = i_size_read(inode);
3823 
3824 	if (off > i_size)
3825 		return 0;
3826 	if (off+len > i_size)
3827 		len = i_size-off;
3828 	toread = len;
3829 	while (toread > 0) {
3830 		tocopy = sb->s_blocksize - offset < toread ?
3831 				sb->s_blocksize - offset : toread;
3832 		bh = ext4_bread(NULL, inode, blk, 0, &err);
3833 		if (err)
3834 			return err;
3835 		if (!bh)	/* A hole? */
3836 			memset(data, 0, tocopy);
3837 		else
3838 			memcpy(data, bh->b_data+offset, tocopy);
3839 		brelse(bh);
3840 		offset = 0;
3841 		toread -= tocopy;
3842 		data += tocopy;
3843 		blk++;
3844 	}
3845 	return len;
3846 }
3847 
3848 /* Write to quotafile (we know the transaction is already started and has
3849  * enough credits) */
3850 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3851 				const char *data, size_t len, loff_t off)
3852 {
3853 	struct inode *inode = sb_dqopt(sb)->files[type];
3854 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3855 	int err = 0;
3856 	int offset = off & (sb->s_blocksize - 1);
3857 	int tocopy;
3858 	int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3859 	size_t towrite = len;
3860 	struct buffer_head *bh;
3861 	handle_t *handle = journal_current_handle();
3862 
3863 	if (EXT4_SB(sb)->s_journal && !handle) {
3864 		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3865 			" cancelled because transaction is not started",
3866 			(unsigned long long)off, (unsigned long long)len);
3867 		return -EIO;
3868 	}
3869 	mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3870 	while (towrite > 0) {
3871 		tocopy = sb->s_blocksize - offset < towrite ?
3872 				sb->s_blocksize - offset : towrite;
3873 		bh = ext4_bread(handle, inode, blk, 1, &err);
3874 		if (!bh)
3875 			goto out;
3876 		if (journal_quota) {
3877 			err = ext4_journal_get_write_access(handle, bh);
3878 			if (err) {
3879 				brelse(bh);
3880 				goto out;
3881 			}
3882 		}
3883 		lock_buffer(bh);
3884 		memcpy(bh->b_data+offset, data, tocopy);
3885 		flush_dcache_page(bh->b_page);
3886 		unlock_buffer(bh);
3887 		if (journal_quota)
3888 			err = ext4_handle_dirty_metadata(handle, NULL, bh);
3889 		else {
3890 			/* Always do at least ordered writes for quotas */
3891 			err = ext4_jbd2_file_inode(handle, inode);
3892 			mark_buffer_dirty(bh);
3893 		}
3894 		brelse(bh);
3895 		if (err)
3896 			goto out;
3897 		offset = 0;
3898 		towrite -= tocopy;
3899 		data += tocopy;
3900 		blk++;
3901 	}
3902 out:
3903 	if (len == towrite) {
3904 		mutex_unlock(&inode->i_mutex);
3905 		return err;
3906 	}
3907 	if (inode->i_size < off+len-towrite) {
3908 		i_size_write(inode, off+len-towrite);
3909 		EXT4_I(inode)->i_disksize = inode->i_size;
3910 	}
3911 	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3912 	ext4_mark_inode_dirty(handle, inode);
3913 	mutex_unlock(&inode->i_mutex);
3914 	return len - towrite;
3915 }
3916 
3917 #endif
3918 
3919 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3920 		       const char *dev_name, void *data, struct vfsmount *mnt)
3921 {
3922 	return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3923 }
3924 
3925 static struct file_system_type ext4_fs_type = {
3926 	.owner		= THIS_MODULE,
3927 	.name		= "ext4",
3928 	.get_sb		= ext4_get_sb,
3929 	.kill_sb	= kill_block_super,
3930 	.fs_flags	= FS_REQUIRES_DEV,
3931 };
3932 
3933 #ifdef CONFIG_EXT4DEV_COMPAT
3934 static int ext4dev_get_sb(struct file_system_type *fs_type, int flags,
3935 			  const char *dev_name, void *data,struct vfsmount *mnt)
3936 {
3937 	printk(KERN_WARNING "EXT4-fs (%s): Update your userspace programs "
3938 	       "to mount using ext4\n", dev_name);
3939 	printk(KERN_WARNING "EXT4-fs (%s): ext4dev backwards compatibility "
3940 	       "will go away by 2.6.31\n", dev_name);
3941 	return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3942 }
3943 
3944 static struct file_system_type ext4dev_fs_type = {
3945 	.owner		= THIS_MODULE,
3946 	.name		= "ext4dev",
3947 	.get_sb		= ext4dev_get_sb,
3948 	.kill_sb	= kill_block_super,
3949 	.fs_flags	= FS_REQUIRES_DEV,
3950 };
3951 MODULE_ALIAS("ext4dev");
3952 #endif
3953 
3954 static int __init init_ext4_fs(void)
3955 {
3956 	int err;
3957 
3958 	err = init_ext4_system_zone();
3959 	if (err)
3960 		return err;
3961 	ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3962 	if (!ext4_kset)
3963 		goto out4;
3964 	ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3965 	err = init_ext4_mballoc();
3966 	if (err)
3967 		goto out3;
3968 
3969 	err = init_ext4_xattr();
3970 	if (err)
3971 		goto out2;
3972 	err = init_inodecache();
3973 	if (err)
3974 		goto out1;
3975 	err = register_filesystem(&ext4_fs_type);
3976 	if (err)
3977 		goto out;
3978 #ifdef CONFIG_EXT4DEV_COMPAT
3979 	err = register_filesystem(&ext4dev_fs_type);
3980 	if (err) {
3981 		unregister_filesystem(&ext4_fs_type);
3982 		goto out;
3983 	}
3984 #endif
3985 	return 0;
3986 out:
3987 	destroy_inodecache();
3988 out1:
3989 	exit_ext4_xattr();
3990 out2:
3991 	exit_ext4_mballoc();
3992 out3:
3993 	remove_proc_entry("fs/ext4", NULL);
3994 	kset_unregister(ext4_kset);
3995 out4:
3996 	exit_ext4_system_zone();
3997 	return err;
3998 }
3999 
4000 static void __exit exit_ext4_fs(void)
4001 {
4002 	unregister_filesystem(&ext4_fs_type);
4003 #ifdef CONFIG_EXT4DEV_COMPAT
4004 	unregister_filesystem(&ext4dev_fs_type);
4005 #endif
4006 	destroy_inodecache();
4007 	exit_ext4_xattr();
4008 	exit_ext4_mballoc();
4009 	remove_proc_entry("fs/ext4", NULL);
4010 	kset_unregister(ext4_kset);
4011 	exit_ext4_system_zone();
4012 }
4013 
4014 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4015 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4016 MODULE_LICENSE("GPL");
4017 module_init(init_ext4_fs)
4018 module_exit(exit_ext4_fs)
4019