xref: /linux/fs/ext4/super.c (revision 44eeab67416711db9b84610ef18c99a60415dff8)
1 /*
2  *  linux/fs/ext4/super.c
3  *
4  * Copyright (C) 1992, 1993, 1994, 1995
5  * Remy Card (card@masi.ibp.fr)
6  * Laboratoire MASI - Institut Blaise Pascal
7  * Universite Pierre et Marie Curie (Paris VI)
8  *
9  *  from
10  *
11  *  linux/fs/minix/inode.c
12  *
13  *  Copyright (C) 1991, 1992  Linus Torvalds
14  *
15  *  Big-endian to little-endian byte-swapping/bitmaps by
16  *        David S. Miller (davem@caip.rutgers.edu), 1995
17  */
18 
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/smp_lock.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/proc_fs.h>
39 #include <linux/ctype.h>
40 #include <linux/log2.h>
41 #include <linux/crc16.h>
42 #include <asm/uaccess.h>
43 
44 #include "ext4.h"
45 #include "ext4_jbd2.h"
46 #include "xattr.h"
47 #include "acl.h"
48 
49 #define CREATE_TRACE_POINTS
50 #include <trace/events/ext4.h>
51 
52 static int default_mb_history_length = 1000;
53 
54 module_param_named(default_mb_history_length, default_mb_history_length,
55 		   int, 0644);
56 MODULE_PARM_DESC(default_mb_history_length,
57 		 "Default number of entries saved for mb_history");
58 
59 struct proc_dir_entry *ext4_proc_root;
60 static struct kset *ext4_kset;
61 
62 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
63 			     unsigned long journal_devnum);
64 static int ext4_commit_super(struct super_block *sb, int sync);
65 static void ext4_mark_recovery_complete(struct super_block *sb,
66 					struct ext4_super_block *es);
67 static void ext4_clear_journal_err(struct super_block *sb,
68 				   struct ext4_super_block *es);
69 static int ext4_sync_fs(struct super_block *sb, int wait);
70 static const char *ext4_decode_error(struct super_block *sb, int errno,
71 				     char nbuf[16]);
72 static int ext4_remount(struct super_block *sb, int *flags, char *data);
73 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
74 static int ext4_unfreeze(struct super_block *sb);
75 static void ext4_write_super(struct super_block *sb);
76 static int ext4_freeze(struct super_block *sb);
77 
78 
79 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
80 			       struct ext4_group_desc *bg)
81 {
82 	return le32_to_cpu(bg->bg_block_bitmap_lo) |
83 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
84 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
85 }
86 
87 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
88 			       struct ext4_group_desc *bg)
89 {
90 	return le32_to_cpu(bg->bg_inode_bitmap_lo) |
91 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
92 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
93 }
94 
95 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
96 			      struct ext4_group_desc *bg)
97 {
98 	return le32_to_cpu(bg->bg_inode_table_lo) |
99 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
100 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
101 }
102 
103 __u32 ext4_free_blks_count(struct super_block *sb,
104 			      struct ext4_group_desc *bg)
105 {
106 	return le16_to_cpu(bg->bg_free_blocks_count_lo) |
107 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
108 		 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
109 }
110 
111 __u32 ext4_free_inodes_count(struct super_block *sb,
112 			      struct ext4_group_desc *bg)
113 {
114 	return le16_to_cpu(bg->bg_free_inodes_count_lo) |
115 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
116 		 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
117 }
118 
119 __u32 ext4_used_dirs_count(struct super_block *sb,
120 			      struct ext4_group_desc *bg)
121 {
122 	return le16_to_cpu(bg->bg_used_dirs_count_lo) |
123 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
124 		 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
125 }
126 
127 __u32 ext4_itable_unused_count(struct super_block *sb,
128 			      struct ext4_group_desc *bg)
129 {
130 	return le16_to_cpu(bg->bg_itable_unused_lo) |
131 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
132 		 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
133 }
134 
135 void ext4_block_bitmap_set(struct super_block *sb,
136 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
137 {
138 	bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
139 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
140 		bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
141 }
142 
143 void ext4_inode_bitmap_set(struct super_block *sb,
144 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
145 {
146 	bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
147 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
148 		bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
149 }
150 
151 void ext4_inode_table_set(struct super_block *sb,
152 			  struct ext4_group_desc *bg, ext4_fsblk_t blk)
153 {
154 	bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
155 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
156 		bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
157 }
158 
159 void ext4_free_blks_set(struct super_block *sb,
160 			  struct ext4_group_desc *bg, __u32 count)
161 {
162 	bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
163 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
164 		bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
165 }
166 
167 void ext4_free_inodes_set(struct super_block *sb,
168 			  struct ext4_group_desc *bg, __u32 count)
169 {
170 	bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
171 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
172 		bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
173 }
174 
175 void ext4_used_dirs_set(struct super_block *sb,
176 			  struct ext4_group_desc *bg, __u32 count)
177 {
178 	bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
179 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
180 		bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
181 }
182 
183 void ext4_itable_unused_set(struct super_block *sb,
184 			  struct ext4_group_desc *bg, __u32 count)
185 {
186 	bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
187 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
188 		bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
189 }
190 
191 /*
192  * Wrappers for jbd2_journal_start/end.
193  *
194  * The only special thing we need to do here is to make sure that all
195  * journal_end calls result in the superblock being marked dirty, so
196  * that sync() will call the filesystem's write_super callback if
197  * appropriate.
198  */
199 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
200 {
201 	journal_t *journal;
202 
203 	if (sb->s_flags & MS_RDONLY)
204 		return ERR_PTR(-EROFS);
205 
206 	/* Special case here: if the journal has aborted behind our
207 	 * backs (eg. EIO in the commit thread), then we still need to
208 	 * take the FS itself readonly cleanly. */
209 	journal = EXT4_SB(sb)->s_journal;
210 	if (journal) {
211 		if (is_journal_aborted(journal)) {
212 			ext4_abort(sb, __func__, "Detected aborted journal");
213 			return ERR_PTR(-EROFS);
214 		}
215 		return jbd2_journal_start(journal, nblocks);
216 	}
217 	/*
218 	 * We're not journaling, return the appropriate indication.
219 	 */
220 	current->journal_info = EXT4_NOJOURNAL_HANDLE;
221 	return current->journal_info;
222 }
223 
224 /*
225  * The only special thing we need to do here is to make sure that all
226  * jbd2_journal_stop calls result in the superblock being marked dirty, so
227  * that sync() will call the filesystem's write_super callback if
228  * appropriate.
229  */
230 int __ext4_journal_stop(const char *where, handle_t *handle)
231 {
232 	struct super_block *sb;
233 	int err;
234 	int rc;
235 
236 	if (!ext4_handle_valid(handle)) {
237 		/*
238 		 * Do this here since we don't call jbd2_journal_stop() in
239 		 * no-journal mode.
240 		 */
241 		current->journal_info = NULL;
242 		return 0;
243 	}
244 	sb = handle->h_transaction->t_journal->j_private;
245 	err = handle->h_err;
246 	rc = jbd2_journal_stop(handle);
247 
248 	if (!err)
249 		err = rc;
250 	if (err)
251 		__ext4_std_error(sb, where, err);
252 	return err;
253 }
254 
255 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
256 		struct buffer_head *bh, handle_t *handle, int err)
257 {
258 	char nbuf[16];
259 	const char *errstr = ext4_decode_error(NULL, err, nbuf);
260 
261 	BUG_ON(!ext4_handle_valid(handle));
262 
263 	if (bh)
264 		BUFFER_TRACE(bh, "abort");
265 
266 	if (!handle->h_err)
267 		handle->h_err = err;
268 
269 	if (is_handle_aborted(handle))
270 		return;
271 
272 	printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
273 	       caller, errstr, err_fn);
274 
275 	jbd2_journal_abort_handle(handle);
276 }
277 
278 /* Deal with the reporting of failure conditions on a filesystem such as
279  * inconsistencies detected or read IO failures.
280  *
281  * On ext2, we can store the error state of the filesystem in the
282  * superblock.  That is not possible on ext4, because we may have other
283  * write ordering constraints on the superblock which prevent us from
284  * writing it out straight away; and given that the journal is about to
285  * be aborted, we can't rely on the current, or future, transactions to
286  * write out the superblock safely.
287  *
288  * We'll just use the jbd2_journal_abort() error code to record an error in
289  * the journal instead.  On recovery, the journal will compain about
290  * that error until we've noted it down and cleared it.
291  */
292 
293 static void ext4_handle_error(struct super_block *sb)
294 {
295 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
296 
297 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
298 	es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
299 
300 	if (sb->s_flags & MS_RDONLY)
301 		return;
302 
303 	if (!test_opt(sb, ERRORS_CONT)) {
304 		journal_t *journal = EXT4_SB(sb)->s_journal;
305 
306 		EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
307 		if (journal)
308 			jbd2_journal_abort(journal, -EIO);
309 	}
310 	if (test_opt(sb, ERRORS_RO)) {
311 		ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
312 		sb->s_flags |= MS_RDONLY;
313 	}
314 	ext4_commit_super(sb, 1);
315 	if (test_opt(sb, ERRORS_PANIC))
316 		panic("EXT4-fs (device %s): panic forced after error\n",
317 			sb->s_id);
318 }
319 
320 void ext4_error(struct super_block *sb, const char *function,
321 		const char *fmt, ...)
322 {
323 	va_list args;
324 
325 	va_start(args, fmt);
326 	printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
327 	vprintk(fmt, args);
328 	printk("\n");
329 	va_end(args);
330 
331 	ext4_handle_error(sb);
332 }
333 
334 static const char *ext4_decode_error(struct super_block *sb, int errno,
335 				     char nbuf[16])
336 {
337 	char *errstr = NULL;
338 
339 	switch (errno) {
340 	case -EIO:
341 		errstr = "IO failure";
342 		break;
343 	case -ENOMEM:
344 		errstr = "Out of memory";
345 		break;
346 	case -EROFS:
347 		if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
348 			errstr = "Journal has aborted";
349 		else
350 			errstr = "Readonly filesystem";
351 		break;
352 	default:
353 		/* If the caller passed in an extra buffer for unknown
354 		 * errors, textualise them now.  Else we just return
355 		 * NULL. */
356 		if (nbuf) {
357 			/* Check for truncated error codes... */
358 			if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
359 				errstr = nbuf;
360 		}
361 		break;
362 	}
363 
364 	return errstr;
365 }
366 
367 /* __ext4_std_error decodes expected errors from journaling functions
368  * automatically and invokes the appropriate error response.  */
369 
370 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
371 {
372 	char nbuf[16];
373 	const char *errstr;
374 
375 	/* Special case: if the error is EROFS, and we're not already
376 	 * inside a transaction, then there's really no point in logging
377 	 * an error. */
378 	if (errno == -EROFS && journal_current_handle() == NULL &&
379 	    (sb->s_flags & MS_RDONLY))
380 		return;
381 
382 	errstr = ext4_decode_error(sb, errno, nbuf);
383 	printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
384 	       sb->s_id, function, errstr);
385 
386 	ext4_handle_error(sb);
387 }
388 
389 /*
390  * ext4_abort is a much stronger failure handler than ext4_error.  The
391  * abort function may be used to deal with unrecoverable failures such
392  * as journal IO errors or ENOMEM at a critical moment in log management.
393  *
394  * We unconditionally force the filesystem into an ABORT|READONLY state,
395  * unless the error response on the fs has been set to panic in which
396  * case we take the easy way out and panic immediately.
397  */
398 
399 void ext4_abort(struct super_block *sb, const char *function,
400 		const char *fmt, ...)
401 {
402 	va_list args;
403 
404 	va_start(args, fmt);
405 	printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
406 	vprintk(fmt, args);
407 	printk("\n");
408 	va_end(args);
409 
410 	if (test_opt(sb, ERRORS_PANIC))
411 		panic("EXT4-fs panic from previous error\n");
412 
413 	if (sb->s_flags & MS_RDONLY)
414 		return;
415 
416 	ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
417 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
418 	sb->s_flags |= MS_RDONLY;
419 	EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
420 	if (EXT4_SB(sb)->s_journal)
421 		jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
422 }
423 
424 void ext4_msg (struct super_block * sb, const char *prefix,
425 		   const char *fmt, ...)
426 {
427 	va_list args;
428 
429 	va_start(args, fmt);
430 	printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
431 	vprintk(fmt, args);
432 	printk("\n");
433 	va_end(args);
434 }
435 
436 void ext4_warning(struct super_block *sb, const char *function,
437 		  const char *fmt, ...)
438 {
439 	va_list args;
440 
441 	va_start(args, fmt);
442 	printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
443 	       sb->s_id, function);
444 	vprintk(fmt, args);
445 	printk("\n");
446 	va_end(args);
447 }
448 
449 void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
450 			   const char *function, const char *fmt, ...)
451 __releases(bitlock)
452 __acquires(bitlock)
453 {
454 	va_list args;
455 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
456 
457 	va_start(args, fmt);
458 	printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
459 	vprintk(fmt, args);
460 	printk("\n");
461 	va_end(args);
462 
463 	if (test_opt(sb, ERRORS_CONT)) {
464 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
465 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
466 		ext4_commit_super(sb, 0);
467 		return;
468 	}
469 	ext4_unlock_group(sb, grp);
470 	ext4_handle_error(sb);
471 	/*
472 	 * We only get here in the ERRORS_RO case; relocking the group
473 	 * may be dangerous, but nothing bad will happen since the
474 	 * filesystem will have already been marked read/only and the
475 	 * journal has been aborted.  We return 1 as a hint to callers
476 	 * who might what to use the return value from
477 	 * ext4_grp_locked_error() to distinguish beween the
478 	 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
479 	 * aggressively from the ext4 function in question, with a
480 	 * more appropriate error code.
481 	 */
482 	ext4_lock_group(sb, grp);
483 	return;
484 }
485 
486 void ext4_update_dynamic_rev(struct super_block *sb)
487 {
488 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
489 
490 	if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
491 		return;
492 
493 	ext4_warning(sb, __func__,
494 		     "updating to rev %d because of new feature flag, "
495 		     "running e2fsck is recommended",
496 		     EXT4_DYNAMIC_REV);
497 
498 	es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
499 	es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
500 	es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
501 	/* leave es->s_feature_*compat flags alone */
502 	/* es->s_uuid will be set by e2fsck if empty */
503 
504 	/*
505 	 * The rest of the superblock fields should be zero, and if not it
506 	 * means they are likely already in use, so leave them alone.  We
507 	 * can leave it up to e2fsck to clean up any inconsistencies there.
508 	 */
509 }
510 
511 /*
512  * Open the external journal device
513  */
514 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
515 {
516 	struct block_device *bdev;
517 	char b[BDEVNAME_SIZE];
518 
519 	bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
520 	if (IS_ERR(bdev))
521 		goto fail;
522 	return bdev;
523 
524 fail:
525 	ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
526 			__bdevname(dev, b), PTR_ERR(bdev));
527 	return NULL;
528 }
529 
530 /*
531  * Release the journal device
532  */
533 static int ext4_blkdev_put(struct block_device *bdev)
534 {
535 	bd_release(bdev);
536 	return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
537 }
538 
539 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
540 {
541 	struct block_device *bdev;
542 	int ret = -ENODEV;
543 
544 	bdev = sbi->journal_bdev;
545 	if (bdev) {
546 		ret = ext4_blkdev_put(bdev);
547 		sbi->journal_bdev = NULL;
548 	}
549 	return ret;
550 }
551 
552 static inline struct inode *orphan_list_entry(struct list_head *l)
553 {
554 	return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
555 }
556 
557 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
558 {
559 	struct list_head *l;
560 
561 	ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
562 		 le32_to_cpu(sbi->s_es->s_last_orphan));
563 
564 	printk(KERN_ERR "sb_info orphan list:\n");
565 	list_for_each(l, &sbi->s_orphan) {
566 		struct inode *inode = orphan_list_entry(l);
567 		printk(KERN_ERR "  "
568 		       "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
569 		       inode->i_sb->s_id, inode->i_ino, inode,
570 		       inode->i_mode, inode->i_nlink,
571 		       NEXT_ORPHAN(inode));
572 	}
573 }
574 
575 static void ext4_put_super(struct super_block *sb)
576 {
577 	struct ext4_sb_info *sbi = EXT4_SB(sb);
578 	struct ext4_super_block *es = sbi->s_es;
579 	int i, err;
580 
581 	lock_super(sb);
582 	lock_kernel();
583 	if (sb->s_dirt)
584 		ext4_commit_super(sb, 1);
585 
586 	ext4_release_system_zone(sb);
587 	ext4_mb_release(sb);
588 	ext4_ext_release(sb);
589 	ext4_xattr_put_super(sb);
590 	if (sbi->s_journal) {
591 		err = jbd2_journal_destroy(sbi->s_journal);
592 		sbi->s_journal = NULL;
593 		if (err < 0)
594 			ext4_abort(sb, __func__,
595 				   "Couldn't clean up the journal");
596 	}
597 	if (!(sb->s_flags & MS_RDONLY)) {
598 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
599 		es->s_state = cpu_to_le16(sbi->s_mount_state);
600 		ext4_commit_super(sb, 1);
601 	}
602 	if (sbi->s_proc) {
603 		remove_proc_entry(sb->s_id, ext4_proc_root);
604 	}
605 	kobject_del(&sbi->s_kobj);
606 
607 	for (i = 0; i < sbi->s_gdb_count; i++)
608 		brelse(sbi->s_group_desc[i]);
609 	kfree(sbi->s_group_desc);
610 	if (is_vmalloc_addr(sbi->s_flex_groups))
611 		vfree(sbi->s_flex_groups);
612 	else
613 		kfree(sbi->s_flex_groups);
614 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
615 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
616 	percpu_counter_destroy(&sbi->s_dirs_counter);
617 	percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
618 	brelse(sbi->s_sbh);
619 #ifdef CONFIG_QUOTA
620 	for (i = 0; i < MAXQUOTAS; i++)
621 		kfree(sbi->s_qf_names[i]);
622 #endif
623 
624 	/* Debugging code just in case the in-memory inode orphan list
625 	 * isn't empty.  The on-disk one can be non-empty if we've
626 	 * detected an error and taken the fs readonly, but the
627 	 * in-memory list had better be clean by this point. */
628 	if (!list_empty(&sbi->s_orphan))
629 		dump_orphan_list(sb, sbi);
630 	J_ASSERT(list_empty(&sbi->s_orphan));
631 
632 	invalidate_bdev(sb->s_bdev);
633 	if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
634 		/*
635 		 * Invalidate the journal device's buffers.  We don't want them
636 		 * floating about in memory - the physical journal device may
637 		 * hotswapped, and it breaks the `ro-after' testing code.
638 		 */
639 		sync_blockdev(sbi->journal_bdev);
640 		invalidate_bdev(sbi->journal_bdev);
641 		ext4_blkdev_remove(sbi);
642 	}
643 	sb->s_fs_info = NULL;
644 	/*
645 	 * Now that we are completely done shutting down the
646 	 * superblock, we need to actually destroy the kobject.
647 	 */
648 	unlock_kernel();
649 	unlock_super(sb);
650 	kobject_put(&sbi->s_kobj);
651 	wait_for_completion(&sbi->s_kobj_unregister);
652 	kfree(sbi->s_blockgroup_lock);
653 	kfree(sbi);
654 }
655 
656 static struct kmem_cache *ext4_inode_cachep;
657 
658 /*
659  * Called inside transaction, so use GFP_NOFS
660  */
661 static struct inode *ext4_alloc_inode(struct super_block *sb)
662 {
663 	struct ext4_inode_info *ei;
664 
665 	ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
666 	if (!ei)
667 		return NULL;
668 
669 #ifdef CONFIG_EXT4_FS_POSIX_ACL
670 	ei->i_acl = EXT4_ACL_NOT_CACHED;
671 	ei->i_default_acl = EXT4_ACL_NOT_CACHED;
672 #endif
673 	ei->vfs_inode.i_version = 1;
674 	ei->vfs_inode.i_data.writeback_index = 0;
675 	memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
676 	INIT_LIST_HEAD(&ei->i_prealloc_list);
677 	spin_lock_init(&ei->i_prealloc_lock);
678 	/*
679 	 * Note:  We can be called before EXT4_SB(sb)->s_journal is set,
680 	 * therefore it can be null here.  Don't check it, just initialize
681 	 * jinode.
682 	 */
683 	jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
684 	ei->i_reserved_data_blocks = 0;
685 	ei->i_reserved_meta_blocks = 0;
686 	ei->i_allocated_meta_blocks = 0;
687 	ei->i_delalloc_reserved_flag = 0;
688 	spin_lock_init(&(ei->i_block_reservation_lock));
689 
690 	return &ei->vfs_inode;
691 }
692 
693 static void ext4_destroy_inode(struct inode *inode)
694 {
695 	if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
696 		ext4_msg(inode->i_sb, KERN_ERR,
697 			 "Inode %lu (%p): orphan list check failed!",
698 			 inode->i_ino, EXT4_I(inode));
699 		print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
700 				EXT4_I(inode), sizeof(struct ext4_inode_info),
701 				true);
702 		dump_stack();
703 	}
704 	kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
705 }
706 
707 static void init_once(void *foo)
708 {
709 	struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
710 
711 	INIT_LIST_HEAD(&ei->i_orphan);
712 #ifdef CONFIG_EXT4_FS_XATTR
713 	init_rwsem(&ei->xattr_sem);
714 #endif
715 	init_rwsem(&ei->i_data_sem);
716 	inode_init_once(&ei->vfs_inode);
717 }
718 
719 static int init_inodecache(void)
720 {
721 	ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
722 					     sizeof(struct ext4_inode_info),
723 					     0, (SLAB_RECLAIM_ACCOUNT|
724 						SLAB_MEM_SPREAD),
725 					     init_once);
726 	if (ext4_inode_cachep == NULL)
727 		return -ENOMEM;
728 	return 0;
729 }
730 
731 static void destroy_inodecache(void)
732 {
733 	kmem_cache_destroy(ext4_inode_cachep);
734 }
735 
736 static void ext4_clear_inode(struct inode *inode)
737 {
738 #ifdef CONFIG_EXT4_FS_POSIX_ACL
739 	if (EXT4_I(inode)->i_acl &&
740 			EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
741 		posix_acl_release(EXT4_I(inode)->i_acl);
742 		EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
743 	}
744 	if (EXT4_I(inode)->i_default_acl &&
745 			EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
746 		posix_acl_release(EXT4_I(inode)->i_default_acl);
747 		EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
748 	}
749 #endif
750 	ext4_discard_preallocations(inode);
751 	if (EXT4_JOURNAL(inode))
752 		jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
753 				       &EXT4_I(inode)->jinode);
754 }
755 
756 static inline void ext4_show_quota_options(struct seq_file *seq,
757 					   struct super_block *sb)
758 {
759 #if defined(CONFIG_QUOTA)
760 	struct ext4_sb_info *sbi = EXT4_SB(sb);
761 
762 	if (sbi->s_jquota_fmt)
763 		seq_printf(seq, ",jqfmt=%s",
764 		(sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
765 
766 	if (sbi->s_qf_names[USRQUOTA])
767 		seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
768 
769 	if (sbi->s_qf_names[GRPQUOTA])
770 		seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
771 
772 	if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
773 		seq_puts(seq, ",usrquota");
774 
775 	if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
776 		seq_puts(seq, ",grpquota");
777 #endif
778 }
779 
780 /*
781  * Show an option if
782  *  - it's set to a non-default value OR
783  *  - if the per-sb default is different from the global default
784  */
785 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
786 {
787 	int def_errors;
788 	unsigned long def_mount_opts;
789 	struct super_block *sb = vfs->mnt_sb;
790 	struct ext4_sb_info *sbi = EXT4_SB(sb);
791 	struct ext4_super_block *es = sbi->s_es;
792 
793 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
794 	def_errors     = le16_to_cpu(es->s_errors);
795 
796 	if (sbi->s_sb_block != 1)
797 		seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
798 	if (test_opt(sb, MINIX_DF))
799 		seq_puts(seq, ",minixdf");
800 	if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
801 		seq_puts(seq, ",grpid");
802 	if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
803 		seq_puts(seq, ",nogrpid");
804 	if (sbi->s_resuid != EXT4_DEF_RESUID ||
805 	    le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
806 		seq_printf(seq, ",resuid=%u", sbi->s_resuid);
807 	}
808 	if (sbi->s_resgid != EXT4_DEF_RESGID ||
809 	    le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
810 		seq_printf(seq, ",resgid=%u", sbi->s_resgid);
811 	}
812 	if (test_opt(sb, ERRORS_RO)) {
813 		if (def_errors == EXT4_ERRORS_PANIC ||
814 		    def_errors == EXT4_ERRORS_CONTINUE) {
815 			seq_puts(seq, ",errors=remount-ro");
816 		}
817 	}
818 	if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
819 		seq_puts(seq, ",errors=continue");
820 	if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
821 		seq_puts(seq, ",errors=panic");
822 	if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
823 		seq_puts(seq, ",nouid32");
824 	if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
825 		seq_puts(seq, ",debug");
826 	if (test_opt(sb, OLDALLOC))
827 		seq_puts(seq, ",oldalloc");
828 #ifdef CONFIG_EXT4_FS_XATTR
829 	if (test_opt(sb, XATTR_USER) &&
830 		!(def_mount_opts & EXT4_DEFM_XATTR_USER))
831 		seq_puts(seq, ",user_xattr");
832 	if (!test_opt(sb, XATTR_USER) &&
833 	    (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
834 		seq_puts(seq, ",nouser_xattr");
835 	}
836 #endif
837 #ifdef CONFIG_EXT4_FS_POSIX_ACL
838 	if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
839 		seq_puts(seq, ",acl");
840 	if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
841 		seq_puts(seq, ",noacl");
842 #endif
843 	if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
844 		seq_printf(seq, ",commit=%u",
845 			   (unsigned) (sbi->s_commit_interval / HZ));
846 	}
847 	if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
848 		seq_printf(seq, ",min_batch_time=%u",
849 			   (unsigned) sbi->s_min_batch_time);
850 	}
851 	if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
852 		seq_printf(seq, ",max_batch_time=%u",
853 			   (unsigned) sbi->s_min_batch_time);
854 	}
855 
856 	/*
857 	 * We're changing the default of barrier mount option, so
858 	 * let's always display its mount state so it's clear what its
859 	 * status is.
860 	 */
861 	seq_puts(seq, ",barrier=");
862 	seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
863 	if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
864 		seq_puts(seq, ",journal_async_commit");
865 	if (test_opt(sb, NOBH))
866 		seq_puts(seq, ",nobh");
867 	if (test_opt(sb, I_VERSION))
868 		seq_puts(seq, ",i_version");
869 	if (!test_opt(sb, DELALLOC))
870 		seq_puts(seq, ",nodelalloc");
871 
872 
873 	if (sbi->s_stripe)
874 		seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
875 	/*
876 	 * journal mode get enabled in different ways
877 	 * So just print the value even if we didn't specify it
878 	 */
879 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
880 		seq_puts(seq, ",data=journal");
881 	else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
882 		seq_puts(seq, ",data=ordered");
883 	else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
884 		seq_puts(seq, ",data=writeback");
885 
886 	if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
887 		seq_printf(seq, ",inode_readahead_blks=%u",
888 			   sbi->s_inode_readahead_blks);
889 
890 	if (test_opt(sb, DATA_ERR_ABORT))
891 		seq_puts(seq, ",data_err=abort");
892 
893 	if (test_opt(sb, NO_AUTO_DA_ALLOC))
894 		seq_puts(seq, ",noauto_da_alloc");
895 
896 	ext4_show_quota_options(seq, sb);
897 
898 	return 0;
899 }
900 
901 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
902 					u64 ino, u32 generation)
903 {
904 	struct inode *inode;
905 
906 	if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
907 		return ERR_PTR(-ESTALE);
908 	if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
909 		return ERR_PTR(-ESTALE);
910 
911 	/* iget isn't really right if the inode is currently unallocated!!
912 	 *
913 	 * ext4_read_inode will return a bad_inode if the inode had been
914 	 * deleted, so we should be safe.
915 	 *
916 	 * Currently we don't know the generation for parent directory, so
917 	 * a generation of 0 means "accept any"
918 	 */
919 	inode = ext4_iget(sb, ino);
920 	if (IS_ERR(inode))
921 		return ERR_CAST(inode);
922 	if (generation && inode->i_generation != generation) {
923 		iput(inode);
924 		return ERR_PTR(-ESTALE);
925 	}
926 
927 	return inode;
928 }
929 
930 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
931 					int fh_len, int fh_type)
932 {
933 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
934 				    ext4_nfs_get_inode);
935 }
936 
937 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
938 					int fh_len, int fh_type)
939 {
940 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
941 				    ext4_nfs_get_inode);
942 }
943 
944 /*
945  * Try to release metadata pages (indirect blocks, directories) which are
946  * mapped via the block device.  Since these pages could have journal heads
947  * which would prevent try_to_free_buffers() from freeing them, we must use
948  * jbd2 layer's try_to_free_buffers() function to release them.
949  */
950 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
951 				 gfp_t wait)
952 {
953 	journal_t *journal = EXT4_SB(sb)->s_journal;
954 
955 	WARN_ON(PageChecked(page));
956 	if (!page_has_buffers(page))
957 		return 0;
958 	if (journal)
959 		return jbd2_journal_try_to_free_buffers(journal, page,
960 							wait & ~__GFP_WAIT);
961 	return try_to_free_buffers(page);
962 }
963 
964 #ifdef CONFIG_QUOTA
965 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
966 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
967 
968 static int ext4_write_dquot(struct dquot *dquot);
969 static int ext4_acquire_dquot(struct dquot *dquot);
970 static int ext4_release_dquot(struct dquot *dquot);
971 static int ext4_mark_dquot_dirty(struct dquot *dquot);
972 static int ext4_write_info(struct super_block *sb, int type);
973 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
974 				char *path, int remount);
975 static int ext4_quota_on_mount(struct super_block *sb, int type);
976 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
977 			       size_t len, loff_t off);
978 static ssize_t ext4_quota_write(struct super_block *sb, int type,
979 				const char *data, size_t len, loff_t off);
980 
981 static struct dquot_operations ext4_quota_operations = {
982 	.initialize	= dquot_initialize,
983 	.drop		= dquot_drop,
984 	.alloc_space	= dquot_alloc_space,
985 	.reserve_space	= dquot_reserve_space,
986 	.claim_space	= dquot_claim_space,
987 	.release_rsv	= dquot_release_reserved_space,
988 	.get_reserved_space = ext4_get_reserved_space,
989 	.alloc_inode	= dquot_alloc_inode,
990 	.free_space	= dquot_free_space,
991 	.free_inode	= dquot_free_inode,
992 	.transfer	= dquot_transfer,
993 	.write_dquot	= ext4_write_dquot,
994 	.acquire_dquot	= ext4_acquire_dquot,
995 	.release_dquot	= ext4_release_dquot,
996 	.mark_dirty	= ext4_mark_dquot_dirty,
997 	.write_info	= ext4_write_info,
998 	.alloc_dquot	= dquot_alloc,
999 	.destroy_dquot	= dquot_destroy,
1000 };
1001 
1002 static struct quotactl_ops ext4_qctl_operations = {
1003 	.quota_on	= ext4_quota_on,
1004 	.quota_off	= vfs_quota_off,
1005 	.quota_sync	= vfs_quota_sync,
1006 	.get_info	= vfs_get_dqinfo,
1007 	.set_info	= vfs_set_dqinfo,
1008 	.get_dqblk	= vfs_get_dqblk,
1009 	.set_dqblk	= vfs_set_dqblk
1010 };
1011 #endif
1012 
1013 static const struct super_operations ext4_sops = {
1014 	.alloc_inode	= ext4_alloc_inode,
1015 	.destroy_inode	= ext4_destroy_inode,
1016 	.write_inode	= ext4_write_inode,
1017 	.dirty_inode	= ext4_dirty_inode,
1018 	.delete_inode	= ext4_delete_inode,
1019 	.put_super	= ext4_put_super,
1020 	.sync_fs	= ext4_sync_fs,
1021 	.freeze_fs	= ext4_freeze,
1022 	.unfreeze_fs	= ext4_unfreeze,
1023 	.statfs		= ext4_statfs,
1024 	.remount_fs	= ext4_remount,
1025 	.clear_inode	= ext4_clear_inode,
1026 	.show_options	= ext4_show_options,
1027 #ifdef CONFIG_QUOTA
1028 	.quota_read	= ext4_quota_read,
1029 	.quota_write	= ext4_quota_write,
1030 #endif
1031 	.bdev_try_to_free_page = bdev_try_to_free_page,
1032 };
1033 
1034 static const struct super_operations ext4_nojournal_sops = {
1035 	.alloc_inode	= ext4_alloc_inode,
1036 	.destroy_inode	= ext4_destroy_inode,
1037 	.write_inode	= ext4_write_inode,
1038 	.dirty_inode	= ext4_dirty_inode,
1039 	.delete_inode	= ext4_delete_inode,
1040 	.write_super	= ext4_write_super,
1041 	.put_super	= ext4_put_super,
1042 	.statfs		= ext4_statfs,
1043 	.remount_fs	= ext4_remount,
1044 	.clear_inode	= ext4_clear_inode,
1045 	.show_options	= ext4_show_options,
1046 #ifdef CONFIG_QUOTA
1047 	.quota_read	= ext4_quota_read,
1048 	.quota_write	= ext4_quota_write,
1049 #endif
1050 	.bdev_try_to_free_page = bdev_try_to_free_page,
1051 };
1052 
1053 static const struct export_operations ext4_export_ops = {
1054 	.fh_to_dentry = ext4_fh_to_dentry,
1055 	.fh_to_parent = ext4_fh_to_parent,
1056 	.get_parent = ext4_get_parent,
1057 };
1058 
1059 enum {
1060 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1061 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1062 	Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1063 	Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1064 	Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1065 	Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1066 	Opt_journal_update, Opt_journal_dev,
1067 	Opt_journal_checksum, Opt_journal_async_commit,
1068 	Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1069 	Opt_data_err_abort, Opt_data_err_ignore, Opt_mb_history_length,
1070 	Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1071 	Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1072 	Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1073 	Opt_usrquota, Opt_grpquota, Opt_i_version,
1074 	Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1075 	Opt_block_validity, Opt_noblock_validity,
1076 	Opt_inode_readahead_blks, Opt_journal_ioprio
1077 };
1078 
1079 static const match_table_t tokens = {
1080 	{Opt_bsd_df, "bsddf"},
1081 	{Opt_minix_df, "minixdf"},
1082 	{Opt_grpid, "grpid"},
1083 	{Opt_grpid, "bsdgroups"},
1084 	{Opt_nogrpid, "nogrpid"},
1085 	{Opt_nogrpid, "sysvgroups"},
1086 	{Opt_resgid, "resgid=%u"},
1087 	{Opt_resuid, "resuid=%u"},
1088 	{Opt_sb, "sb=%u"},
1089 	{Opt_err_cont, "errors=continue"},
1090 	{Opt_err_panic, "errors=panic"},
1091 	{Opt_err_ro, "errors=remount-ro"},
1092 	{Opt_nouid32, "nouid32"},
1093 	{Opt_debug, "debug"},
1094 	{Opt_oldalloc, "oldalloc"},
1095 	{Opt_orlov, "orlov"},
1096 	{Opt_user_xattr, "user_xattr"},
1097 	{Opt_nouser_xattr, "nouser_xattr"},
1098 	{Opt_acl, "acl"},
1099 	{Opt_noacl, "noacl"},
1100 	{Opt_noload, "noload"},
1101 	{Opt_nobh, "nobh"},
1102 	{Opt_bh, "bh"},
1103 	{Opt_commit, "commit=%u"},
1104 	{Opt_min_batch_time, "min_batch_time=%u"},
1105 	{Opt_max_batch_time, "max_batch_time=%u"},
1106 	{Opt_journal_update, "journal=update"},
1107 	{Opt_journal_dev, "journal_dev=%u"},
1108 	{Opt_journal_checksum, "journal_checksum"},
1109 	{Opt_journal_async_commit, "journal_async_commit"},
1110 	{Opt_abort, "abort"},
1111 	{Opt_data_journal, "data=journal"},
1112 	{Opt_data_ordered, "data=ordered"},
1113 	{Opt_data_writeback, "data=writeback"},
1114 	{Opt_data_err_abort, "data_err=abort"},
1115 	{Opt_data_err_ignore, "data_err=ignore"},
1116 	{Opt_mb_history_length, "mb_history_length=%u"},
1117 	{Opt_offusrjquota, "usrjquota="},
1118 	{Opt_usrjquota, "usrjquota=%s"},
1119 	{Opt_offgrpjquota, "grpjquota="},
1120 	{Opt_grpjquota, "grpjquota=%s"},
1121 	{Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1122 	{Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1123 	{Opt_grpquota, "grpquota"},
1124 	{Opt_noquota, "noquota"},
1125 	{Opt_quota, "quota"},
1126 	{Opt_usrquota, "usrquota"},
1127 	{Opt_barrier, "barrier=%u"},
1128 	{Opt_barrier, "barrier"},
1129 	{Opt_nobarrier, "nobarrier"},
1130 	{Opt_i_version, "i_version"},
1131 	{Opt_stripe, "stripe=%u"},
1132 	{Opt_resize, "resize"},
1133 	{Opt_delalloc, "delalloc"},
1134 	{Opt_nodelalloc, "nodelalloc"},
1135 	{Opt_block_validity, "block_validity"},
1136 	{Opt_noblock_validity, "noblock_validity"},
1137 	{Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1138 	{Opt_journal_ioprio, "journal_ioprio=%u"},
1139 	{Opt_auto_da_alloc, "auto_da_alloc=%u"},
1140 	{Opt_auto_da_alloc, "auto_da_alloc"},
1141 	{Opt_noauto_da_alloc, "noauto_da_alloc"},
1142 	{Opt_err, NULL},
1143 };
1144 
1145 static ext4_fsblk_t get_sb_block(void **data)
1146 {
1147 	ext4_fsblk_t	sb_block;
1148 	char		*options = (char *) *data;
1149 
1150 	if (!options || strncmp(options, "sb=", 3) != 0)
1151 		return 1;	/* Default location */
1152 
1153 	options += 3;
1154 	/* TODO: use simple_strtoll with >32bit ext4 */
1155 	sb_block = simple_strtoul(options, &options, 0);
1156 	if (*options && *options != ',') {
1157 		printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1158 		       (char *) *data);
1159 		return 1;
1160 	}
1161 	if (*options == ',')
1162 		options++;
1163 	*data = (void *) options;
1164 
1165 	return sb_block;
1166 }
1167 
1168 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1169 
1170 static int parse_options(char *options, struct super_block *sb,
1171 			 unsigned long *journal_devnum,
1172 			 unsigned int *journal_ioprio,
1173 			 ext4_fsblk_t *n_blocks_count, int is_remount)
1174 {
1175 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1176 	char *p;
1177 	substring_t args[MAX_OPT_ARGS];
1178 	int data_opt = 0;
1179 	int option;
1180 #ifdef CONFIG_QUOTA
1181 	int qtype, qfmt;
1182 	char *qname;
1183 #endif
1184 
1185 	if (!options)
1186 		return 1;
1187 
1188 	while ((p = strsep(&options, ",")) != NULL) {
1189 		int token;
1190 		if (!*p)
1191 			continue;
1192 
1193 		token = match_token(p, tokens, args);
1194 		switch (token) {
1195 		case Opt_bsd_df:
1196 			clear_opt(sbi->s_mount_opt, MINIX_DF);
1197 			break;
1198 		case Opt_minix_df:
1199 			set_opt(sbi->s_mount_opt, MINIX_DF);
1200 			break;
1201 		case Opt_grpid:
1202 			set_opt(sbi->s_mount_opt, GRPID);
1203 			break;
1204 		case Opt_nogrpid:
1205 			clear_opt(sbi->s_mount_opt, GRPID);
1206 			break;
1207 		case Opt_resuid:
1208 			if (match_int(&args[0], &option))
1209 				return 0;
1210 			sbi->s_resuid = option;
1211 			break;
1212 		case Opt_resgid:
1213 			if (match_int(&args[0], &option))
1214 				return 0;
1215 			sbi->s_resgid = option;
1216 			break;
1217 		case Opt_sb:
1218 			/* handled by get_sb_block() instead of here */
1219 			/* *sb_block = match_int(&args[0]); */
1220 			break;
1221 		case Opt_err_panic:
1222 			clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1223 			clear_opt(sbi->s_mount_opt, ERRORS_RO);
1224 			set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1225 			break;
1226 		case Opt_err_ro:
1227 			clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1228 			clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1229 			set_opt(sbi->s_mount_opt, ERRORS_RO);
1230 			break;
1231 		case Opt_err_cont:
1232 			clear_opt(sbi->s_mount_opt, ERRORS_RO);
1233 			clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1234 			set_opt(sbi->s_mount_opt, ERRORS_CONT);
1235 			break;
1236 		case Opt_nouid32:
1237 			set_opt(sbi->s_mount_opt, NO_UID32);
1238 			break;
1239 		case Opt_debug:
1240 			set_opt(sbi->s_mount_opt, DEBUG);
1241 			break;
1242 		case Opt_oldalloc:
1243 			set_opt(sbi->s_mount_opt, OLDALLOC);
1244 			break;
1245 		case Opt_orlov:
1246 			clear_opt(sbi->s_mount_opt, OLDALLOC);
1247 			break;
1248 #ifdef CONFIG_EXT4_FS_XATTR
1249 		case Opt_user_xattr:
1250 			set_opt(sbi->s_mount_opt, XATTR_USER);
1251 			break;
1252 		case Opt_nouser_xattr:
1253 			clear_opt(sbi->s_mount_opt, XATTR_USER);
1254 			break;
1255 #else
1256 		case Opt_user_xattr:
1257 		case Opt_nouser_xattr:
1258 			ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1259 			break;
1260 #endif
1261 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1262 		case Opt_acl:
1263 			set_opt(sbi->s_mount_opt, POSIX_ACL);
1264 			break;
1265 		case Opt_noacl:
1266 			clear_opt(sbi->s_mount_opt, POSIX_ACL);
1267 			break;
1268 #else
1269 		case Opt_acl:
1270 		case Opt_noacl:
1271 			ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1272 			break;
1273 #endif
1274 		case Opt_journal_update:
1275 			/* @@@ FIXME */
1276 			/* Eventually we will want to be able to create
1277 			   a journal file here.  For now, only allow the
1278 			   user to specify an existing inode to be the
1279 			   journal file. */
1280 			if (is_remount) {
1281 				ext4_msg(sb, KERN_ERR,
1282 					 "Cannot specify journal on remount");
1283 				return 0;
1284 			}
1285 			set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1286 			break;
1287 		case Opt_journal_dev:
1288 			if (is_remount) {
1289 				ext4_msg(sb, KERN_ERR,
1290 					"Cannot specify journal on remount");
1291 				return 0;
1292 			}
1293 			if (match_int(&args[0], &option))
1294 				return 0;
1295 			*journal_devnum = option;
1296 			break;
1297 		case Opt_journal_checksum:
1298 			set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1299 			break;
1300 		case Opt_journal_async_commit:
1301 			set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1302 			set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1303 			break;
1304 		case Opt_noload:
1305 			set_opt(sbi->s_mount_opt, NOLOAD);
1306 			break;
1307 		case Opt_commit:
1308 			if (match_int(&args[0], &option))
1309 				return 0;
1310 			if (option < 0)
1311 				return 0;
1312 			if (option == 0)
1313 				option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1314 			sbi->s_commit_interval = HZ * option;
1315 			break;
1316 		case Opt_max_batch_time:
1317 			if (match_int(&args[0], &option))
1318 				return 0;
1319 			if (option < 0)
1320 				return 0;
1321 			if (option == 0)
1322 				option = EXT4_DEF_MAX_BATCH_TIME;
1323 			sbi->s_max_batch_time = option;
1324 			break;
1325 		case Opt_min_batch_time:
1326 			if (match_int(&args[0], &option))
1327 				return 0;
1328 			if (option < 0)
1329 				return 0;
1330 			sbi->s_min_batch_time = option;
1331 			break;
1332 		case Opt_data_journal:
1333 			data_opt = EXT4_MOUNT_JOURNAL_DATA;
1334 			goto datacheck;
1335 		case Opt_data_ordered:
1336 			data_opt = EXT4_MOUNT_ORDERED_DATA;
1337 			goto datacheck;
1338 		case Opt_data_writeback:
1339 			data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1340 		datacheck:
1341 			if (is_remount) {
1342 				if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1343 						!= data_opt) {
1344 					ext4_msg(sb, KERN_ERR,
1345 						"Cannot change data mode on remount");
1346 					return 0;
1347 				}
1348 			} else {
1349 				sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1350 				sbi->s_mount_opt |= data_opt;
1351 			}
1352 			break;
1353 		case Opt_data_err_abort:
1354 			set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1355 			break;
1356 		case Opt_data_err_ignore:
1357 			clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1358 			break;
1359 		case Opt_mb_history_length:
1360 			if (match_int(&args[0], &option))
1361 				return 0;
1362 			if (option < 0)
1363 				return 0;
1364 			sbi->s_mb_history_max = option;
1365 			break;
1366 #ifdef CONFIG_QUOTA
1367 		case Opt_usrjquota:
1368 			qtype = USRQUOTA;
1369 			goto set_qf_name;
1370 		case Opt_grpjquota:
1371 			qtype = GRPQUOTA;
1372 set_qf_name:
1373 			if (sb_any_quota_loaded(sb) &&
1374 			    !sbi->s_qf_names[qtype]) {
1375 				ext4_msg(sb, KERN_ERR,
1376 				       "Cannot change journaled "
1377 				       "quota options when quota turned on");
1378 				return 0;
1379 			}
1380 			qname = match_strdup(&args[0]);
1381 			if (!qname) {
1382 				ext4_msg(sb, KERN_ERR,
1383 					"Not enough memory for "
1384 					"storing quotafile name");
1385 				return 0;
1386 			}
1387 			if (sbi->s_qf_names[qtype] &&
1388 			    strcmp(sbi->s_qf_names[qtype], qname)) {
1389 				ext4_msg(sb, KERN_ERR,
1390 					"%s quota file already "
1391 					"specified", QTYPE2NAME(qtype));
1392 				kfree(qname);
1393 				return 0;
1394 			}
1395 			sbi->s_qf_names[qtype] = qname;
1396 			if (strchr(sbi->s_qf_names[qtype], '/')) {
1397 				ext4_msg(sb, KERN_ERR,
1398 					"quotafile must be on "
1399 					"filesystem root");
1400 				kfree(sbi->s_qf_names[qtype]);
1401 				sbi->s_qf_names[qtype] = NULL;
1402 				return 0;
1403 			}
1404 			set_opt(sbi->s_mount_opt, QUOTA);
1405 			break;
1406 		case Opt_offusrjquota:
1407 			qtype = USRQUOTA;
1408 			goto clear_qf_name;
1409 		case Opt_offgrpjquota:
1410 			qtype = GRPQUOTA;
1411 clear_qf_name:
1412 			if (sb_any_quota_loaded(sb) &&
1413 			    sbi->s_qf_names[qtype]) {
1414 				ext4_msg(sb, KERN_ERR, "Cannot change "
1415 					"journaled quota options when "
1416 					"quota turned on");
1417 				return 0;
1418 			}
1419 			/*
1420 			 * The space will be released later when all options
1421 			 * are confirmed to be correct
1422 			 */
1423 			sbi->s_qf_names[qtype] = NULL;
1424 			break;
1425 		case Opt_jqfmt_vfsold:
1426 			qfmt = QFMT_VFS_OLD;
1427 			goto set_qf_format;
1428 		case Opt_jqfmt_vfsv0:
1429 			qfmt = QFMT_VFS_V0;
1430 set_qf_format:
1431 			if (sb_any_quota_loaded(sb) &&
1432 			    sbi->s_jquota_fmt != qfmt) {
1433 				ext4_msg(sb, KERN_ERR, "Cannot change "
1434 					"journaled quota options when "
1435 					"quota turned on");
1436 				return 0;
1437 			}
1438 			sbi->s_jquota_fmt = qfmt;
1439 			break;
1440 		case Opt_quota:
1441 		case Opt_usrquota:
1442 			set_opt(sbi->s_mount_opt, QUOTA);
1443 			set_opt(sbi->s_mount_opt, USRQUOTA);
1444 			break;
1445 		case Opt_grpquota:
1446 			set_opt(sbi->s_mount_opt, QUOTA);
1447 			set_opt(sbi->s_mount_opt, GRPQUOTA);
1448 			break;
1449 		case Opt_noquota:
1450 			if (sb_any_quota_loaded(sb)) {
1451 				ext4_msg(sb, KERN_ERR, "Cannot change quota "
1452 					"options when quota turned on");
1453 				return 0;
1454 			}
1455 			clear_opt(sbi->s_mount_opt, QUOTA);
1456 			clear_opt(sbi->s_mount_opt, USRQUOTA);
1457 			clear_opt(sbi->s_mount_opt, GRPQUOTA);
1458 			break;
1459 #else
1460 		case Opt_quota:
1461 		case Opt_usrquota:
1462 		case Opt_grpquota:
1463 			ext4_msg(sb, KERN_ERR,
1464 				"quota options not supported");
1465 			break;
1466 		case Opt_usrjquota:
1467 		case Opt_grpjquota:
1468 		case Opt_offusrjquota:
1469 		case Opt_offgrpjquota:
1470 		case Opt_jqfmt_vfsold:
1471 		case Opt_jqfmt_vfsv0:
1472 			ext4_msg(sb, KERN_ERR,
1473 				"journaled quota options not supported");
1474 			break;
1475 		case Opt_noquota:
1476 			break;
1477 #endif
1478 		case Opt_abort:
1479 			sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1480 			break;
1481 		case Opt_nobarrier:
1482 			clear_opt(sbi->s_mount_opt, BARRIER);
1483 			break;
1484 		case Opt_barrier:
1485 			if (match_int(&args[0], &option)) {
1486 				set_opt(sbi->s_mount_opt, BARRIER);
1487 				break;
1488 			}
1489 			if (option)
1490 				set_opt(sbi->s_mount_opt, BARRIER);
1491 			else
1492 				clear_opt(sbi->s_mount_opt, BARRIER);
1493 			break;
1494 		case Opt_ignore:
1495 			break;
1496 		case Opt_resize:
1497 			if (!is_remount) {
1498 				ext4_msg(sb, KERN_ERR,
1499 					"resize option only available "
1500 					"for remount");
1501 				return 0;
1502 			}
1503 			if (match_int(&args[0], &option) != 0)
1504 				return 0;
1505 			*n_blocks_count = option;
1506 			break;
1507 		case Opt_nobh:
1508 			set_opt(sbi->s_mount_opt, NOBH);
1509 			break;
1510 		case Opt_bh:
1511 			clear_opt(sbi->s_mount_opt, NOBH);
1512 			break;
1513 		case Opt_i_version:
1514 			set_opt(sbi->s_mount_opt, I_VERSION);
1515 			sb->s_flags |= MS_I_VERSION;
1516 			break;
1517 		case Opt_nodelalloc:
1518 			clear_opt(sbi->s_mount_opt, DELALLOC);
1519 			break;
1520 		case Opt_stripe:
1521 			if (match_int(&args[0], &option))
1522 				return 0;
1523 			if (option < 0)
1524 				return 0;
1525 			sbi->s_stripe = option;
1526 			break;
1527 		case Opt_delalloc:
1528 			set_opt(sbi->s_mount_opt, DELALLOC);
1529 			break;
1530 		case Opt_block_validity:
1531 			set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1532 			break;
1533 		case Opt_noblock_validity:
1534 			clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1535 			break;
1536 		case Opt_inode_readahead_blks:
1537 			if (match_int(&args[0], &option))
1538 				return 0;
1539 			if (option < 0 || option > (1 << 30))
1540 				return 0;
1541 			if (!is_power_of_2(option)) {
1542 				ext4_msg(sb, KERN_ERR,
1543 					 "EXT4-fs: inode_readahead_blks"
1544 					 " must be a power of 2");
1545 				return 0;
1546 			}
1547 			sbi->s_inode_readahead_blks = option;
1548 			break;
1549 		case Opt_journal_ioprio:
1550 			if (match_int(&args[0], &option))
1551 				return 0;
1552 			if (option < 0 || option > 7)
1553 				break;
1554 			*journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1555 							    option);
1556 			break;
1557 		case Opt_noauto_da_alloc:
1558 			set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1559 			break;
1560 		case Opt_auto_da_alloc:
1561 			if (match_int(&args[0], &option)) {
1562 				clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1563 				break;
1564 			}
1565 			if (option)
1566 				clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1567 			else
1568 				set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1569 			break;
1570 		default:
1571 			ext4_msg(sb, KERN_ERR,
1572 			       "Unrecognized mount option \"%s\" "
1573 			       "or missing value", p);
1574 			return 0;
1575 		}
1576 	}
1577 #ifdef CONFIG_QUOTA
1578 	if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1579 		if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1580 		     sbi->s_qf_names[USRQUOTA])
1581 			clear_opt(sbi->s_mount_opt, USRQUOTA);
1582 
1583 		if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1584 		     sbi->s_qf_names[GRPQUOTA])
1585 			clear_opt(sbi->s_mount_opt, GRPQUOTA);
1586 
1587 		if ((sbi->s_qf_names[USRQUOTA] &&
1588 				(sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1589 		    (sbi->s_qf_names[GRPQUOTA] &&
1590 				(sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1591 			ext4_msg(sb, KERN_ERR, "old and new quota "
1592 					"format mixing");
1593 			return 0;
1594 		}
1595 
1596 		if (!sbi->s_jquota_fmt) {
1597 			ext4_msg(sb, KERN_ERR, "journaled quota format "
1598 					"not specified");
1599 			return 0;
1600 		}
1601 	} else {
1602 		if (sbi->s_jquota_fmt) {
1603 			ext4_msg(sb, KERN_ERR, "journaled quota format "
1604 					"specified with no journaling "
1605 					"enabled");
1606 			return 0;
1607 		}
1608 	}
1609 #endif
1610 	return 1;
1611 }
1612 
1613 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1614 			    int read_only)
1615 {
1616 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1617 	int res = 0;
1618 
1619 	if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1620 		ext4_msg(sb, KERN_ERR, "revision level too high, "
1621 			 "forcing read-only mode");
1622 		res = MS_RDONLY;
1623 	}
1624 	if (read_only)
1625 		return res;
1626 	if (!(sbi->s_mount_state & EXT4_VALID_FS))
1627 		ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1628 			 "running e2fsck is recommended");
1629 	else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1630 		ext4_msg(sb, KERN_WARNING,
1631 			 "warning: mounting fs with errors, "
1632 			 "running e2fsck is recommended");
1633 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1634 		 le16_to_cpu(es->s_mnt_count) >=
1635 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1636 		ext4_msg(sb, KERN_WARNING,
1637 			 "warning: maximal mount count reached, "
1638 			 "running e2fsck is recommended");
1639 	else if (le32_to_cpu(es->s_checkinterval) &&
1640 		(le32_to_cpu(es->s_lastcheck) +
1641 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1642 		ext4_msg(sb, KERN_WARNING,
1643 			 "warning: checktime reached, "
1644 			 "running e2fsck is recommended");
1645 	if (!sbi->s_journal)
1646 		es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1647 	if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1648 		es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1649 	le16_add_cpu(&es->s_mnt_count, 1);
1650 	es->s_mtime = cpu_to_le32(get_seconds());
1651 	ext4_update_dynamic_rev(sb);
1652 	if (sbi->s_journal)
1653 		EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1654 
1655 	ext4_commit_super(sb, 1);
1656 	if (test_opt(sb, DEBUG))
1657 		printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1658 				"bpg=%lu, ipg=%lu, mo=%04x]\n",
1659 			sb->s_blocksize,
1660 			sbi->s_groups_count,
1661 			EXT4_BLOCKS_PER_GROUP(sb),
1662 			EXT4_INODES_PER_GROUP(sb),
1663 			sbi->s_mount_opt);
1664 
1665 	if (EXT4_SB(sb)->s_journal) {
1666 		ext4_msg(sb, KERN_INFO, "%s journal on %s",
1667 		       EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1668 		       "external", EXT4_SB(sb)->s_journal->j_devname);
1669 	} else {
1670 		ext4_msg(sb, KERN_INFO, "no journal");
1671 	}
1672 	return res;
1673 }
1674 
1675 static int ext4_fill_flex_info(struct super_block *sb)
1676 {
1677 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1678 	struct ext4_group_desc *gdp = NULL;
1679 	ext4_group_t flex_group_count;
1680 	ext4_group_t flex_group;
1681 	int groups_per_flex = 0;
1682 	size_t size;
1683 	int i;
1684 
1685 	if (!sbi->s_es->s_log_groups_per_flex) {
1686 		sbi->s_log_groups_per_flex = 0;
1687 		return 1;
1688 	}
1689 
1690 	sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1691 	groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1692 
1693 	/* We allocate both existing and potentially added groups */
1694 	flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1695 			((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1696 			      EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1697 	size = flex_group_count * sizeof(struct flex_groups);
1698 	sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1699 	if (sbi->s_flex_groups == NULL) {
1700 		sbi->s_flex_groups = vmalloc(size);
1701 		if (sbi->s_flex_groups)
1702 			memset(sbi->s_flex_groups, 0, size);
1703 	}
1704 	if (sbi->s_flex_groups == NULL) {
1705 		ext4_msg(sb, KERN_ERR, "not enough memory for "
1706 				"%u flex groups", flex_group_count);
1707 		goto failed;
1708 	}
1709 
1710 	for (i = 0; i < sbi->s_groups_count; i++) {
1711 		gdp = ext4_get_group_desc(sb, i, NULL);
1712 
1713 		flex_group = ext4_flex_group(sbi, i);
1714 		atomic_set(&sbi->s_flex_groups[flex_group].free_inodes,
1715 			   ext4_free_inodes_count(sb, gdp));
1716 		atomic_set(&sbi->s_flex_groups[flex_group].free_blocks,
1717 			   ext4_free_blks_count(sb, gdp));
1718 		atomic_set(&sbi->s_flex_groups[flex_group].used_dirs,
1719 			   ext4_used_dirs_count(sb, gdp));
1720 	}
1721 
1722 	return 1;
1723 failed:
1724 	return 0;
1725 }
1726 
1727 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1728 			    struct ext4_group_desc *gdp)
1729 {
1730 	__u16 crc = 0;
1731 
1732 	if (sbi->s_es->s_feature_ro_compat &
1733 	    cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1734 		int offset = offsetof(struct ext4_group_desc, bg_checksum);
1735 		__le32 le_group = cpu_to_le32(block_group);
1736 
1737 		crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1738 		crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1739 		crc = crc16(crc, (__u8 *)gdp, offset);
1740 		offset += sizeof(gdp->bg_checksum); /* skip checksum */
1741 		/* for checksum of struct ext4_group_desc do the rest...*/
1742 		if ((sbi->s_es->s_feature_incompat &
1743 		     cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1744 		    offset < le16_to_cpu(sbi->s_es->s_desc_size))
1745 			crc = crc16(crc, (__u8 *)gdp + offset,
1746 				    le16_to_cpu(sbi->s_es->s_desc_size) -
1747 					offset);
1748 	}
1749 
1750 	return cpu_to_le16(crc);
1751 }
1752 
1753 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1754 				struct ext4_group_desc *gdp)
1755 {
1756 	if ((sbi->s_es->s_feature_ro_compat &
1757 	     cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1758 	    (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1759 		return 0;
1760 
1761 	return 1;
1762 }
1763 
1764 /* Called at mount-time, super-block is locked */
1765 static int ext4_check_descriptors(struct super_block *sb)
1766 {
1767 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1768 	ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1769 	ext4_fsblk_t last_block;
1770 	ext4_fsblk_t block_bitmap;
1771 	ext4_fsblk_t inode_bitmap;
1772 	ext4_fsblk_t inode_table;
1773 	int flexbg_flag = 0;
1774 	ext4_group_t i;
1775 
1776 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1777 		flexbg_flag = 1;
1778 
1779 	ext4_debug("Checking group descriptors");
1780 
1781 	for (i = 0; i < sbi->s_groups_count; i++) {
1782 		struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1783 
1784 		if (i == sbi->s_groups_count - 1 || flexbg_flag)
1785 			last_block = ext4_blocks_count(sbi->s_es) - 1;
1786 		else
1787 			last_block = first_block +
1788 				(EXT4_BLOCKS_PER_GROUP(sb) - 1);
1789 
1790 		block_bitmap = ext4_block_bitmap(sb, gdp);
1791 		if (block_bitmap < first_block || block_bitmap > last_block) {
1792 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1793 			       "Block bitmap for group %u not in group "
1794 			       "(block %llu)!", i, block_bitmap);
1795 			return 0;
1796 		}
1797 		inode_bitmap = ext4_inode_bitmap(sb, gdp);
1798 		if (inode_bitmap < first_block || inode_bitmap > last_block) {
1799 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1800 			       "Inode bitmap for group %u not in group "
1801 			       "(block %llu)!", i, inode_bitmap);
1802 			return 0;
1803 		}
1804 		inode_table = ext4_inode_table(sb, gdp);
1805 		if (inode_table < first_block ||
1806 		    inode_table + sbi->s_itb_per_group - 1 > last_block) {
1807 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1808 			       "Inode table for group %u not in group "
1809 			       "(block %llu)!", i, inode_table);
1810 			return 0;
1811 		}
1812 		ext4_lock_group(sb, i);
1813 		if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1814 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1815 				 "Checksum for group %u failed (%u!=%u)",
1816 				 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1817 				     gdp)), le16_to_cpu(gdp->bg_checksum));
1818 			if (!(sb->s_flags & MS_RDONLY)) {
1819 				ext4_unlock_group(sb, i);
1820 				return 0;
1821 			}
1822 		}
1823 		ext4_unlock_group(sb, i);
1824 		if (!flexbg_flag)
1825 			first_block += EXT4_BLOCKS_PER_GROUP(sb);
1826 	}
1827 
1828 	ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1829 	sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
1830 	return 1;
1831 }
1832 
1833 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1834  * the superblock) which were deleted from all directories, but held open by
1835  * a process at the time of a crash.  We walk the list and try to delete these
1836  * inodes at recovery time (only with a read-write filesystem).
1837  *
1838  * In order to keep the orphan inode chain consistent during traversal (in
1839  * case of crash during recovery), we link each inode into the superblock
1840  * orphan list_head and handle it the same way as an inode deletion during
1841  * normal operation (which journals the operations for us).
1842  *
1843  * We only do an iget() and an iput() on each inode, which is very safe if we
1844  * accidentally point at an in-use or already deleted inode.  The worst that
1845  * can happen in this case is that we get a "bit already cleared" message from
1846  * ext4_free_inode().  The only reason we would point at a wrong inode is if
1847  * e2fsck was run on this filesystem, and it must have already done the orphan
1848  * inode cleanup for us, so we can safely abort without any further action.
1849  */
1850 static void ext4_orphan_cleanup(struct super_block *sb,
1851 				struct ext4_super_block *es)
1852 {
1853 	unsigned int s_flags = sb->s_flags;
1854 	int nr_orphans = 0, nr_truncates = 0;
1855 #ifdef CONFIG_QUOTA
1856 	int i;
1857 #endif
1858 	if (!es->s_last_orphan) {
1859 		jbd_debug(4, "no orphan inodes to clean up\n");
1860 		return;
1861 	}
1862 
1863 	if (bdev_read_only(sb->s_bdev)) {
1864 		ext4_msg(sb, KERN_ERR, "write access "
1865 			"unavailable, skipping orphan cleanup");
1866 		return;
1867 	}
1868 
1869 	if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1870 		if (es->s_last_orphan)
1871 			jbd_debug(1, "Errors on filesystem, "
1872 				  "clearing orphan list.\n");
1873 		es->s_last_orphan = 0;
1874 		jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1875 		return;
1876 	}
1877 
1878 	if (s_flags & MS_RDONLY) {
1879 		ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1880 		sb->s_flags &= ~MS_RDONLY;
1881 	}
1882 #ifdef CONFIG_QUOTA
1883 	/* Needed for iput() to work correctly and not trash data */
1884 	sb->s_flags |= MS_ACTIVE;
1885 	/* Turn on quotas so that they are updated correctly */
1886 	for (i = 0; i < MAXQUOTAS; i++) {
1887 		if (EXT4_SB(sb)->s_qf_names[i]) {
1888 			int ret = ext4_quota_on_mount(sb, i);
1889 			if (ret < 0)
1890 				ext4_msg(sb, KERN_ERR,
1891 					"Cannot turn on journaled "
1892 					"quota: error %d", ret);
1893 		}
1894 	}
1895 #endif
1896 
1897 	while (es->s_last_orphan) {
1898 		struct inode *inode;
1899 
1900 		inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1901 		if (IS_ERR(inode)) {
1902 			es->s_last_orphan = 0;
1903 			break;
1904 		}
1905 
1906 		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1907 		vfs_dq_init(inode);
1908 		if (inode->i_nlink) {
1909 			ext4_msg(sb, KERN_DEBUG,
1910 				"%s: truncating inode %lu to %lld bytes",
1911 				__func__, inode->i_ino, inode->i_size);
1912 			jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1913 				  inode->i_ino, inode->i_size);
1914 			ext4_truncate(inode);
1915 			nr_truncates++;
1916 		} else {
1917 			ext4_msg(sb, KERN_DEBUG,
1918 				"%s: deleting unreferenced inode %lu",
1919 				__func__, inode->i_ino);
1920 			jbd_debug(2, "deleting unreferenced inode %lu\n",
1921 				  inode->i_ino);
1922 			nr_orphans++;
1923 		}
1924 		iput(inode);  /* The delete magic happens here! */
1925 	}
1926 
1927 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1928 
1929 	if (nr_orphans)
1930 		ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1931 		       PLURAL(nr_orphans));
1932 	if (nr_truncates)
1933 		ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1934 		       PLURAL(nr_truncates));
1935 #ifdef CONFIG_QUOTA
1936 	/* Turn quotas off */
1937 	for (i = 0; i < MAXQUOTAS; i++) {
1938 		if (sb_dqopt(sb)->files[i])
1939 			vfs_quota_off(sb, i, 0);
1940 	}
1941 #endif
1942 	sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1943 }
1944 
1945 /*
1946  * Maximal extent format file size.
1947  * Resulting logical blkno at s_maxbytes must fit in our on-disk
1948  * extent format containers, within a sector_t, and within i_blocks
1949  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
1950  * so that won't be a limiting factor.
1951  *
1952  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1953  */
1954 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1955 {
1956 	loff_t res;
1957 	loff_t upper_limit = MAX_LFS_FILESIZE;
1958 
1959 	/* small i_blocks in vfs inode? */
1960 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1961 		/*
1962 		 * CONFIG_LBDAF is not enabled implies the inode
1963 		 * i_block represent total blocks in 512 bytes
1964 		 * 32 == size of vfs inode i_blocks * 8
1965 		 */
1966 		upper_limit = (1LL << 32) - 1;
1967 
1968 		/* total blocks in file system block size */
1969 		upper_limit >>= (blkbits - 9);
1970 		upper_limit <<= blkbits;
1971 	}
1972 
1973 	/* 32-bit extent-start container, ee_block */
1974 	res = 1LL << 32;
1975 	res <<= blkbits;
1976 	res -= 1;
1977 
1978 	/* Sanity check against vm- & vfs- imposed limits */
1979 	if (res > upper_limit)
1980 		res = upper_limit;
1981 
1982 	return res;
1983 }
1984 
1985 /*
1986  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
1987  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1988  * We need to be 1 filesystem block less than the 2^48 sector limit.
1989  */
1990 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1991 {
1992 	loff_t res = EXT4_NDIR_BLOCKS;
1993 	int meta_blocks;
1994 	loff_t upper_limit;
1995 	/* This is calculated to be the largest file size for a dense, block
1996 	 * mapped file such that the file's total number of 512-byte sectors,
1997 	 * including data and all indirect blocks, does not exceed (2^48 - 1).
1998 	 *
1999 	 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2000 	 * number of 512-byte sectors of the file.
2001 	 */
2002 
2003 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2004 		/*
2005 		 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2006 		 * the inode i_block field represents total file blocks in
2007 		 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2008 		 */
2009 		upper_limit = (1LL << 32) - 1;
2010 
2011 		/* total blocks in file system block size */
2012 		upper_limit >>= (bits - 9);
2013 
2014 	} else {
2015 		/*
2016 		 * We use 48 bit ext4_inode i_blocks
2017 		 * With EXT4_HUGE_FILE_FL set the i_blocks
2018 		 * represent total number of blocks in
2019 		 * file system block size
2020 		 */
2021 		upper_limit = (1LL << 48) - 1;
2022 
2023 	}
2024 
2025 	/* indirect blocks */
2026 	meta_blocks = 1;
2027 	/* double indirect blocks */
2028 	meta_blocks += 1 + (1LL << (bits-2));
2029 	/* tripple indirect blocks */
2030 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2031 
2032 	upper_limit -= meta_blocks;
2033 	upper_limit <<= bits;
2034 
2035 	res += 1LL << (bits-2);
2036 	res += 1LL << (2*(bits-2));
2037 	res += 1LL << (3*(bits-2));
2038 	res <<= bits;
2039 	if (res > upper_limit)
2040 		res = upper_limit;
2041 
2042 	if (res > MAX_LFS_FILESIZE)
2043 		res = MAX_LFS_FILESIZE;
2044 
2045 	return res;
2046 }
2047 
2048 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2049 				   ext4_fsblk_t logical_sb_block, int nr)
2050 {
2051 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2052 	ext4_group_t bg, first_meta_bg;
2053 	int has_super = 0;
2054 
2055 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2056 
2057 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2058 	    nr < first_meta_bg)
2059 		return logical_sb_block + nr + 1;
2060 	bg = sbi->s_desc_per_block * nr;
2061 	if (ext4_bg_has_super(sb, bg))
2062 		has_super = 1;
2063 
2064 	return (has_super + ext4_group_first_block_no(sb, bg));
2065 }
2066 
2067 /**
2068  * ext4_get_stripe_size: Get the stripe size.
2069  * @sbi: In memory super block info
2070  *
2071  * If we have specified it via mount option, then
2072  * use the mount option value. If the value specified at mount time is
2073  * greater than the blocks per group use the super block value.
2074  * If the super block value is greater than blocks per group return 0.
2075  * Allocator needs it be less than blocks per group.
2076  *
2077  */
2078 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2079 {
2080 	unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2081 	unsigned long stripe_width =
2082 			le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2083 
2084 	if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2085 		return sbi->s_stripe;
2086 
2087 	if (stripe_width <= sbi->s_blocks_per_group)
2088 		return stripe_width;
2089 
2090 	if (stride <= sbi->s_blocks_per_group)
2091 		return stride;
2092 
2093 	return 0;
2094 }
2095 
2096 /* sysfs supprt */
2097 
2098 struct ext4_attr {
2099 	struct attribute attr;
2100 	ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2101 	ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2102 			 const char *, size_t);
2103 	int offset;
2104 };
2105 
2106 static int parse_strtoul(const char *buf,
2107 		unsigned long max, unsigned long *value)
2108 {
2109 	char *endp;
2110 
2111 	while (*buf && isspace(*buf))
2112 		buf++;
2113 	*value = simple_strtoul(buf, &endp, 0);
2114 	while (*endp && isspace(*endp))
2115 		endp++;
2116 	if (*endp || *value > max)
2117 		return -EINVAL;
2118 
2119 	return 0;
2120 }
2121 
2122 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2123 					      struct ext4_sb_info *sbi,
2124 					      char *buf)
2125 {
2126 	return snprintf(buf, PAGE_SIZE, "%llu\n",
2127 			(s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2128 }
2129 
2130 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2131 					 struct ext4_sb_info *sbi, char *buf)
2132 {
2133 	struct super_block *sb = sbi->s_buddy_cache->i_sb;
2134 
2135 	return snprintf(buf, PAGE_SIZE, "%lu\n",
2136 			(part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2137 			 sbi->s_sectors_written_start) >> 1);
2138 }
2139 
2140 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2141 					  struct ext4_sb_info *sbi, char *buf)
2142 {
2143 	struct super_block *sb = sbi->s_buddy_cache->i_sb;
2144 
2145 	return snprintf(buf, PAGE_SIZE, "%llu\n",
2146 			sbi->s_kbytes_written +
2147 			((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2148 			  EXT4_SB(sb)->s_sectors_written_start) >> 1));
2149 }
2150 
2151 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2152 					  struct ext4_sb_info *sbi,
2153 					  const char *buf, size_t count)
2154 {
2155 	unsigned long t;
2156 
2157 	if (parse_strtoul(buf, 0x40000000, &t))
2158 		return -EINVAL;
2159 
2160 	if (!is_power_of_2(t))
2161 		return -EINVAL;
2162 
2163 	sbi->s_inode_readahead_blks = t;
2164 	return count;
2165 }
2166 
2167 static ssize_t sbi_ui_show(struct ext4_attr *a,
2168 			   struct ext4_sb_info *sbi, char *buf)
2169 {
2170 	unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2171 
2172 	return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2173 }
2174 
2175 static ssize_t sbi_ui_store(struct ext4_attr *a,
2176 			    struct ext4_sb_info *sbi,
2177 			    const char *buf, size_t count)
2178 {
2179 	unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2180 	unsigned long t;
2181 
2182 	if (parse_strtoul(buf, 0xffffffff, &t))
2183 		return -EINVAL;
2184 	*ui = t;
2185 	return count;
2186 }
2187 
2188 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2189 static struct ext4_attr ext4_attr_##_name = {			\
2190 	.attr = {.name = __stringify(_name), .mode = _mode },	\
2191 	.show	= _show,					\
2192 	.store	= _store,					\
2193 	.offset = offsetof(struct ext4_sb_info, _elname),	\
2194 }
2195 #define EXT4_ATTR(name, mode, show, store) \
2196 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2197 
2198 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2199 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2200 #define EXT4_RW_ATTR_SBI_UI(name, elname)	\
2201 	EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2202 #define ATTR_LIST(name) &ext4_attr_##name.attr
2203 
2204 EXT4_RO_ATTR(delayed_allocation_blocks);
2205 EXT4_RO_ATTR(session_write_kbytes);
2206 EXT4_RO_ATTR(lifetime_write_kbytes);
2207 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2208 		 inode_readahead_blks_store, s_inode_readahead_blks);
2209 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2210 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2211 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2212 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2213 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2214 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2215 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2216 
2217 static struct attribute *ext4_attrs[] = {
2218 	ATTR_LIST(delayed_allocation_blocks),
2219 	ATTR_LIST(session_write_kbytes),
2220 	ATTR_LIST(lifetime_write_kbytes),
2221 	ATTR_LIST(inode_readahead_blks),
2222 	ATTR_LIST(inode_goal),
2223 	ATTR_LIST(mb_stats),
2224 	ATTR_LIST(mb_max_to_scan),
2225 	ATTR_LIST(mb_min_to_scan),
2226 	ATTR_LIST(mb_order2_req),
2227 	ATTR_LIST(mb_stream_req),
2228 	ATTR_LIST(mb_group_prealloc),
2229 	NULL,
2230 };
2231 
2232 static ssize_t ext4_attr_show(struct kobject *kobj,
2233 			      struct attribute *attr, char *buf)
2234 {
2235 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2236 						s_kobj);
2237 	struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2238 
2239 	return a->show ? a->show(a, sbi, buf) : 0;
2240 }
2241 
2242 static ssize_t ext4_attr_store(struct kobject *kobj,
2243 			       struct attribute *attr,
2244 			       const char *buf, size_t len)
2245 {
2246 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2247 						s_kobj);
2248 	struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2249 
2250 	return a->store ? a->store(a, sbi, buf, len) : 0;
2251 }
2252 
2253 static void ext4_sb_release(struct kobject *kobj)
2254 {
2255 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2256 						s_kobj);
2257 	complete(&sbi->s_kobj_unregister);
2258 }
2259 
2260 
2261 static struct sysfs_ops ext4_attr_ops = {
2262 	.show	= ext4_attr_show,
2263 	.store	= ext4_attr_store,
2264 };
2265 
2266 static struct kobj_type ext4_ktype = {
2267 	.default_attrs	= ext4_attrs,
2268 	.sysfs_ops	= &ext4_attr_ops,
2269 	.release	= ext4_sb_release,
2270 };
2271 
2272 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2273 				__releases(kernel_lock)
2274 				__acquires(kernel_lock)
2275 {
2276 	struct buffer_head *bh;
2277 	struct ext4_super_block *es = NULL;
2278 	struct ext4_sb_info *sbi;
2279 	ext4_fsblk_t block;
2280 	ext4_fsblk_t sb_block = get_sb_block(&data);
2281 	ext4_fsblk_t logical_sb_block;
2282 	unsigned long offset = 0;
2283 	unsigned long journal_devnum = 0;
2284 	unsigned long def_mount_opts;
2285 	struct inode *root;
2286 	char *cp;
2287 	const char *descr;
2288 	int ret = -EINVAL;
2289 	int blocksize;
2290 	unsigned int db_count;
2291 	unsigned int i;
2292 	int needs_recovery, has_huge_files;
2293 	int features;
2294 	__u64 blocks_count;
2295 	int err;
2296 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2297 
2298 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2299 	if (!sbi)
2300 		return -ENOMEM;
2301 
2302 	sbi->s_blockgroup_lock =
2303 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2304 	if (!sbi->s_blockgroup_lock) {
2305 		kfree(sbi);
2306 		return -ENOMEM;
2307 	}
2308 	sb->s_fs_info = sbi;
2309 	sbi->s_mount_opt = 0;
2310 	sbi->s_resuid = EXT4_DEF_RESUID;
2311 	sbi->s_resgid = EXT4_DEF_RESGID;
2312 	sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2313 	sbi->s_sb_block = sb_block;
2314 	sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2315 						      sectors[1]);
2316 
2317 	unlock_kernel();
2318 
2319 	/* Cleanup superblock name */
2320 	for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2321 		*cp = '!';
2322 
2323 	blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2324 	if (!blocksize) {
2325 		ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2326 		goto out_fail;
2327 	}
2328 
2329 	/*
2330 	 * The ext4 superblock will not be buffer aligned for other than 1kB
2331 	 * block sizes.  We need to calculate the offset from buffer start.
2332 	 */
2333 	if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2334 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2335 		offset = do_div(logical_sb_block, blocksize);
2336 	} else {
2337 		logical_sb_block = sb_block;
2338 	}
2339 
2340 	if (!(bh = sb_bread(sb, logical_sb_block))) {
2341 		ext4_msg(sb, KERN_ERR, "unable to read superblock");
2342 		goto out_fail;
2343 	}
2344 	/*
2345 	 * Note: s_es must be initialized as soon as possible because
2346 	 *       some ext4 macro-instructions depend on its value
2347 	 */
2348 	es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2349 	sbi->s_es = es;
2350 	sb->s_magic = le16_to_cpu(es->s_magic);
2351 	if (sb->s_magic != EXT4_SUPER_MAGIC)
2352 		goto cantfind_ext4;
2353 	sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2354 
2355 	/* Set defaults before we parse the mount options */
2356 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2357 	if (def_mount_opts & EXT4_DEFM_DEBUG)
2358 		set_opt(sbi->s_mount_opt, DEBUG);
2359 	if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2360 		set_opt(sbi->s_mount_opt, GRPID);
2361 	if (def_mount_opts & EXT4_DEFM_UID16)
2362 		set_opt(sbi->s_mount_opt, NO_UID32);
2363 #ifdef CONFIG_EXT4_FS_XATTR
2364 	if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2365 		set_opt(sbi->s_mount_opt, XATTR_USER);
2366 #endif
2367 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2368 	if (def_mount_opts & EXT4_DEFM_ACL)
2369 		set_opt(sbi->s_mount_opt, POSIX_ACL);
2370 #endif
2371 	if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2372 		sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2373 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2374 		sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2375 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2376 		sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2377 
2378 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2379 		set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2380 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2381 		set_opt(sbi->s_mount_opt, ERRORS_CONT);
2382 	else
2383 		set_opt(sbi->s_mount_opt, ERRORS_RO);
2384 
2385 	sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2386 	sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2387 	sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2388 	sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2389 	sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2390 	sbi->s_mb_history_max = default_mb_history_length;
2391 
2392 	set_opt(sbi->s_mount_opt, BARRIER);
2393 
2394 	/*
2395 	 * enable delayed allocation by default
2396 	 * Use -o nodelalloc to turn it off
2397 	 */
2398 	set_opt(sbi->s_mount_opt, DELALLOC);
2399 
2400 	if (!parse_options((char *) data, sb, &journal_devnum,
2401 			   &journal_ioprio, NULL, 0))
2402 		goto failed_mount;
2403 
2404 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2405 		((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2406 
2407 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2408 	    (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2409 	     EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2410 	     EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2411 		ext4_msg(sb, KERN_WARNING,
2412 		       "feature flags set on rev 0 fs, "
2413 		       "running e2fsck is recommended");
2414 
2415 	/*
2416 	 * Check feature flags regardless of the revision level, since we
2417 	 * previously didn't change the revision level when setting the flags,
2418 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
2419 	 */
2420 	features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2421 	if (features) {
2422 		ext4_msg(sb, KERN_ERR,
2423 			"Couldn't mount because of "
2424 			"unsupported optional features (%x)",
2425 			(le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2426 			~EXT4_FEATURE_INCOMPAT_SUPP));
2427 		goto failed_mount;
2428 	}
2429 	features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2430 	if (!(sb->s_flags & MS_RDONLY) && features) {
2431 		ext4_msg(sb, KERN_ERR,
2432 			"Couldn't mount RDWR because of "
2433 			"unsupported optional features (%x)",
2434 			(le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2435 			~EXT4_FEATURE_RO_COMPAT_SUPP));
2436 		goto failed_mount;
2437 	}
2438 	has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2439 				    EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2440 	if (has_huge_files) {
2441 		/*
2442 		 * Large file size enabled file system can only be
2443 		 * mount if kernel is build with CONFIG_LBDAF
2444 		 */
2445 		if (sizeof(root->i_blocks) < sizeof(u64) &&
2446 				!(sb->s_flags & MS_RDONLY)) {
2447 			ext4_msg(sb, KERN_ERR, "Filesystem with huge "
2448 					"files cannot be mounted read-write "
2449 					"without CONFIG_LBDAF");
2450 			goto failed_mount;
2451 		}
2452 	}
2453 	blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2454 
2455 	if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2456 	    blocksize > EXT4_MAX_BLOCK_SIZE) {
2457 		ext4_msg(sb, KERN_ERR,
2458 		       "Unsupported filesystem blocksize %d", blocksize);
2459 		goto failed_mount;
2460 	}
2461 
2462 	if (sb->s_blocksize != blocksize) {
2463 		/* Validate the filesystem blocksize */
2464 		if (!sb_set_blocksize(sb, blocksize)) {
2465 			ext4_msg(sb, KERN_ERR, "bad block size %d",
2466 					blocksize);
2467 			goto failed_mount;
2468 		}
2469 
2470 		brelse(bh);
2471 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2472 		offset = do_div(logical_sb_block, blocksize);
2473 		bh = sb_bread(sb, logical_sb_block);
2474 		if (!bh) {
2475 			ext4_msg(sb, KERN_ERR,
2476 			       "Can't read superblock on 2nd try");
2477 			goto failed_mount;
2478 		}
2479 		es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2480 		sbi->s_es = es;
2481 		if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2482 			ext4_msg(sb, KERN_ERR,
2483 			       "Magic mismatch, very weird!");
2484 			goto failed_mount;
2485 		}
2486 	}
2487 
2488 	sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2489 						      has_huge_files);
2490 	sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2491 
2492 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2493 		sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2494 		sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2495 	} else {
2496 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2497 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2498 		if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2499 		    (!is_power_of_2(sbi->s_inode_size)) ||
2500 		    (sbi->s_inode_size > blocksize)) {
2501 			ext4_msg(sb, KERN_ERR,
2502 			       "unsupported inode size: %d",
2503 			       sbi->s_inode_size);
2504 			goto failed_mount;
2505 		}
2506 		if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2507 			sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2508 	}
2509 
2510 	sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2511 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2512 		if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2513 		    sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2514 		    !is_power_of_2(sbi->s_desc_size)) {
2515 			ext4_msg(sb, KERN_ERR,
2516 			       "unsupported descriptor size %lu",
2517 			       sbi->s_desc_size);
2518 			goto failed_mount;
2519 		}
2520 	} else
2521 		sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2522 
2523 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2524 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2525 	if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2526 		goto cantfind_ext4;
2527 
2528 	sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2529 	if (sbi->s_inodes_per_block == 0)
2530 		goto cantfind_ext4;
2531 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
2532 					sbi->s_inodes_per_block;
2533 	sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2534 	sbi->s_sbh = bh;
2535 	sbi->s_mount_state = le16_to_cpu(es->s_state);
2536 	sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2537 	sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2538 
2539 	for (i = 0; i < 4; i++)
2540 		sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2541 	sbi->s_def_hash_version = es->s_def_hash_version;
2542 	i = le32_to_cpu(es->s_flags);
2543 	if (i & EXT2_FLAGS_UNSIGNED_HASH)
2544 		sbi->s_hash_unsigned = 3;
2545 	else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2546 #ifdef __CHAR_UNSIGNED__
2547 		es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2548 		sbi->s_hash_unsigned = 3;
2549 #else
2550 		es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2551 #endif
2552 		sb->s_dirt = 1;
2553 	}
2554 
2555 	if (sbi->s_blocks_per_group > blocksize * 8) {
2556 		ext4_msg(sb, KERN_ERR,
2557 		       "#blocks per group too big: %lu",
2558 		       sbi->s_blocks_per_group);
2559 		goto failed_mount;
2560 	}
2561 	if (sbi->s_inodes_per_group > blocksize * 8) {
2562 		ext4_msg(sb, KERN_ERR,
2563 		       "#inodes per group too big: %lu",
2564 		       sbi->s_inodes_per_group);
2565 		goto failed_mount;
2566 	}
2567 
2568 	if (ext4_blocks_count(es) >
2569 		    (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2570 		ext4_msg(sb, KERN_ERR, "filesystem"
2571 			" too large to mount safely");
2572 		if (sizeof(sector_t) < 8)
2573 			ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2574 		goto failed_mount;
2575 	}
2576 
2577 	if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2578 		goto cantfind_ext4;
2579 
2580 	/* check blocks count against device size */
2581 	blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2582 	if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2583 		ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2584 		       "exceeds size of device (%llu blocks)",
2585 		       ext4_blocks_count(es), blocks_count);
2586 		goto failed_mount;
2587 	}
2588 
2589 	/*
2590 	 * It makes no sense for the first data block to be beyond the end
2591 	 * of the filesystem.
2592 	 */
2593 	if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2594                 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2595 			 "block %u is beyond end of filesystem (%llu)",
2596 			 le32_to_cpu(es->s_first_data_block),
2597 			 ext4_blocks_count(es));
2598 		goto failed_mount;
2599 	}
2600 	blocks_count = (ext4_blocks_count(es) -
2601 			le32_to_cpu(es->s_first_data_block) +
2602 			EXT4_BLOCKS_PER_GROUP(sb) - 1);
2603 	do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2604 	if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2605 		ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2606 		       "(block count %llu, first data block %u, "
2607 		       "blocks per group %lu)", sbi->s_groups_count,
2608 		       ext4_blocks_count(es),
2609 		       le32_to_cpu(es->s_first_data_block),
2610 		       EXT4_BLOCKS_PER_GROUP(sb));
2611 		goto failed_mount;
2612 	}
2613 	sbi->s_groups_count = blocks_count;
2614 	db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2615 		   EXT4_DESC_PER_BLOCK(sb);
2616 	sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2617 				    GFP_KERNEL);
2618 	if (sbi->s_group_desc == NULL) {
2619 		ext4_msg(sb, KERN_ERR, "not enough memory");
2620 		goto failed_mount;
2621 	}
2622 
2623 #ifdef CONFIG_PROC_FS
2624 	if (ext4_proc_root)
2625 		sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2626 #endif
2627 
2628 	bgl_lock_init(sbi->s_blockgroup_lock);
2629 
2630 	for (i = 0; i < db_count; i++) {
2631 		block = descriptor_loc(sb, logical_sb_block, i);
2632 		sbi->s_group_desc[i] = sb_bread(sb, block);
2633 		if (!sbi->s_group_desc[i]) {
2634 			ext4_msg(sb, KERN_ERR,
2635 			       "can't read group descriptor %d", i);
2636 			db_count = i;
2637 			goto failed_mount2;
2638 		}
2639 	}
2640 	if (!ext4_check_descriptors(sb)) {
2641 		ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2642 		goto failed_mount2;
2643 	}
2644 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2645 		if (!ext4_fill_flex_info(sb)) {
2646 			ext4_msg(sb, KERN_ERR,
2647 			       "unable to initialize "
2648 			       "flex_bg meta info!");
2649 			goto failed_mount2;
2650 		}
2651 
2652 	sbi->s_gdb_count = db_count;
2653 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2654 	spin_lock_init(&sbi->s_next_gen_lock);
2655 
2656 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
2657 			ext4_count_free_blocks(sb));
2658 	if (!err) {
2659 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
2660 				ext4_count_free_inodes(sb));
2661 	}
2662 	if (!err) {
2663 		err = percpu_counter_init(&sbi->s_dirs_counter,
2664 				ext4_count_dirs(sb));
2665 	}
2666 	if (!err) {
2667 		err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2668 	}
2669 	if (err) {
2670 		ext4_msg(sb, KERN_ERR, "insufficient memory");
2671 		goto failed_mount3;
2672 	}
2673 
2674 	sbi->s_stripe = ext4_get_stripe_size(sbi);
2675 
2676 	/*
2677 	 * set up enough so that it can read an inode
2678 	 */
2679 	if (!test_opt(sb, NOLOAD) &&
2680 	    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2681 		sb->s_op = &ext4_sops;
2682 	else
2683 		sb->s_op = &ext4_nojournal_sops;
2684 	sb->s_export_op = &ext4_export_ops;
2685 	sb->s_xattr = ext4_xattr_handlers;
2686 #ifdef CONFIG_QUOTA
2687 	sb->s_qcop = &ext4_qctl_operations;
2688 	sb->dq_op = &ext4_quota_operations;
2689 #endif
2690 	INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2691 	mutex_init(&sbi->s_orphan_lock);
2692 	mutex_init(&sbi->s_resize_lock);
2693 
2694 	sb->s_root = NULL;
2695 
2696 	needs_recovery = (es->s_last_orphan != 0 ||
2697 			  EXT4_HAS_INCOMPAT_FEATURE(sb,
2698 				    EXT4_FEATURE_INCOMPAT_RECOVER));
2699 
2700 	/*
2701 	 * The first inode we look at is the journal inode.  Don't try
2702 	 * root first: it may be modified in the journal!
2703 	 */
2704 	if (!test_opt(sb, NOLOAD) &&
2705 	    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2706 		if (ext4_load_journal(sb, es, journal_devnum))
2707 			goto failed_mount3;
2708 		if (!(sb->s_flags & MS_RDONLY) &&
2709 		    EXT4_SB(sb)->s_journal->j_failed_commit) {
2710 			ext4_msg(sb, KERN_CRIT, "error: "
2711 			       "ext4_fill_super: Journal transaction "
2712 			       "%u is corrupt",
2713 			       EXT4_SB(sb)->s_journal->j_failed_commit);
2714 			if (test_opt(sb, ERRORS_RO)) {
2715 				ext4_msg(sb, KERN_CRIT,
2716 				       "Mounting filesystem read-only");
2717 				sb->s_flags |= MS_RDONLY;
2718 				EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2719 				es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2720 			}
2721 			if (test_opt(sb, ERRORS_PANIC)) {
2722 				EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2723 				es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2724 				ext4_commit_super(sb, 1);
2725 				goto failed_mount4;
2726 			}
2727 		}
2728 	} else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2729 	      EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2730 		ext4_msg(sb, KERN_ERR, "required journal recovery "
2731 		       "suppressed and not mounted read-only");
2732 		goto failed_mount4;
2733 	} else {
2734 		clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2735 		set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2736 		sbi->s_journal = NULL;
2737 		needs_recovery = 0;
2738 		goto no_journal;
2739 	}
2740 
2741 	if (ext4_blocks_count(es) > 0xffffffffULL &&
2742 	    !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2743 				       JBD2_FEATURE_INCOMPAT_64BIT)) {
2744 		ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2745 		goto failed_mount4;
2746 	}
2747 
2748 	if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2749 		jbd2_journal_set_features(sbi->s_journal,
2750 				JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2751 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2752 	} else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2753 		jbd2_journal_set_features(sbi->s_journal,
2754 				JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2755 		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2756 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2757 	} else {
2758 		jbd2_journal_clear_features(sbi->s_journal,
2759 				JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2760 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2761 	}
2762 
2763 	/* We have now updated the journal if required, so we can
2764 	 * validate the data journaling mode. */
2765 	switch (test_opt(sb, DATA_FLAGS)) {
2766 	case 0:
2767 		/* No mode set, assume a default based on the journal
2768 		 * capabilities: ORDERED_DATA if the journal can
2769 		 * cope, else JOURNAL_DATA
2770 		 */
2771 		if (jbd2_journal_check_available_features
2772 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2773 			set_opt(sbi->s_mount_opt, ORDERED_DATA);
2774 		else
2775 			set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2776 		break;
2777 
2778 	case EXT4_MOUNT_ORDERED_DATA:
2779 	case EXT4_MOUNT_WRITEBACK_DATA:
2780 		if (!jbd2_journal_check_available_features
2781 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2782 			ext4_msg(sb, KERN_ERR, "Journal does not support "
2783 			       "requested data journaling mode");
2784 			goto failed_mount4;
2785 		}
2786 	default:
2787 		break;
2788 	}
2789 	set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2790 
2791 no_journal:
2792 
2793 	if (test_opt(sb, NOBH)) {
2794 		if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2795 			ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2796 				"its supported only with writeback mode");
2797 			clear_opt(sbi->s_mount_opt, NOBH);
2798 		}
2799 	}
2800 	/*
2801 	 * The jbd2_journal_load will have done any necessary log recovery,
2802 	 * so we can safely mount the rest of the filesystem now.
2803 	 */
2804 
2805 	root = ext4_iget(sb, EXT4_ROOT_INO);
2806 	if (IS_ERR(root)) {
2807 		ext4_msg(sb, KERN_ERR, "get root inode failed");
2808 		ret = PTR_ERR(root);
2809 		goto failed_mount4;
2810 	}
2811 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2812 		iput(root);
2813 		ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
2814 		goto failed_mount4;
2815 	}
2816 	sb->s_root = d_alloc_root(root);
2817 	if (!sb->s_root) {
2818 		ext4_msg(sb, KERN_ERR, "get root dentry failed");
2819 		iput(root);
2820 		ret = -ENOMEM;
2821 		goto failed_mount4;
2822 	}
2823 
2824 	ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2825 
2826 	/* determine the minimum size of new large inodes, if present */
2827 	if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2828 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2829 						     EXT4_GOOD_OLD_INODE_SIZE;
2830 		if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2831 				       EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2832 			if (sbi->s_want_extra_isize <
2833 			    le16_to_cpu(es->s_want_extra_isize))
2834 				sbi->s_want_extra_isize =
2835 					le16_to_cpu(es->s_want_extra_isize);
2836 			if (sbi->s_want_extra_isize <
2837 			    le16_to_cpu(es->s_min_extra_isize))
2838 				sbi->s_want_extra_isize =
2839 					le16_to_cpu(es->s_min_extra_isize);
2840 		}
2841 	}
2842 	/* Check if enough inode space is available */
2843 	if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2844 							sbi->s_inode_size) {
2845 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2846 						       EXT4_GOOD_OLD_INODE_SIZE;
2847 		ext4_msg(sb, KERN_INFO, "required extra inode space not"
2848 			 "available");
2849 	}
2850 
2851 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2852 		ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2853 			 "requested data journaling mode");
2854 		clear_opt(sbi->s_mount_opt, DELALLOC);
2855 	} else if (test_opt(sb, DELALLOC))
2856 		ext4_msg(sb, KERN_INFO, "delayed allocation enabled");
2857 
2858 	err = ext4_setup_system_zone(sb);
2859 	if (err) {
2860 		ext4_msg(sb, KERN_ERR, "failed to initialize system "
2861 			 "zone (%d)\n", err);
2862 		goto failed_mount4;
2863 	}
2864 
2865 	ext4_ext_init(sb);
2866 	err = ext4_mb_init(sb, needs_recovery);
2867 	if (err) {
2868 		ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2869 			 err);
2870 		goto failed_mount4;
2871 	}
2872 
2873 	sbi->s_kobj.kset = ext4_kset;
2874 	init_completion(&sbi->s_kobj_unregister);
2875 	err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2876 				   "%s", sb->s_id);
2877 	if (err) {
2878 		ext4_mb_release(sb);
2879 		ext4_ext_release(sb);
2880 		goto failed_mount4;
2881 	};
2882 
2883 	EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2884 	ext4_orphan_cleanup(sb, es);
2885 	EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2886 	if (needs_recovery) {
2887 		ext4_msg(sb, KERN_INFO, "recovery complete");
2888 		ext4_mark_recovery_complete(sb, es);
2889 	}
2890 	if (EXT4_SB(sb)->s_journal) {
2891 		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2892 			descr = " journalled data mode";
2893 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2894 			descr = " ordered data mode";
2895 		else
2896 			descr = " writeback data mode";
2897 	} else
2898 		descr = "out journal";
2899 
2900 	ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
2901 
2902 	lock_kernel();
2903 	return 0;
2904 
2905 cantfind_ext4:
2906 	if (!silent)
2907 		ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
2908 	goto failed_mount;
2909 
2910 failed_mount4:
2911 	ext4_msg(sb, KERN_ERR, "mount failed");
2912 	ext4_release_system_zone(sb);
2913 	if (sbi->s_journal) {
2914 		jbd2_journal_destroy(sbi->s_journal);
2915 		sbi->s_journal = NULL;
2916 	}
2917 failed_mount3:
2918 	if (sbi->s_flex_groups) {
2919 		if (is_vmalloc_addr(sbi->s_flex_groups))
2920 			vfree(sbi->s_flex_groups);
2921 		else
2922 			kfree(sbi->s_flex_groups);
2923 	}
2924 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
2925 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
2926 	percpu_counter_destroy(&sbi->s_dirs_counter);
2927 	percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2928 failed_mount2:
2929 	for (i = 0; i < db_count; i++)
2930 		brelse(sbi->s_group_desc[i]);
2931 	kfree(sbi->s_group_desc);
2932 failed_mount:
2933 	if (sbi->s_proc) {
2934 		remove_proc_entry(sb->s_id, ext4_proc_root);
2935 	}
2936 #ifdef CONFIG_QUOTA
2937 	for (i = 0; i < MAXQUOTAS; i++)
2938 		kfree(sbi->s_qf_names[i]);
2939 #endif
2940 	ext4_blkdev_remove(sbi);
2941 	brelse(bh);
2942 out_fail:
2943 	sb->s_fs_info = NULL;
2944 	kfree(sbi->s_blockgroup_lock);
2945 	kfree(sbi);
2946 	lock_kernel();
2947 	return ret;
2948 }
2949 
2950 /*
2951  * Setup any per-fs journal parameters now.  We'll do this both on
2952  * initial mount, once the journal has been initialised but before we've
2953  * done any recovery; and again on any subsequent remount.
2954  */
2955 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2956 {
2957 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2958 
2959 	journal->j_commit_interval = sbi->s_commit_interval;
2960 	journal->j_min_batch_time = sbi->s_min_batch_time;
2961 	journal->j_max_batch_time = sbi->s_max_batch_time;
2962 
2963 	spin_lock(&journal->j_state_lock);
2964 	if (test_opt(sb, BARRIER))
2965 		journal->j_flags |= JBD2_BARRIER;
2966 	else
2967 		journal->j_flags &= ~JBD2_BARRIER;
2968 	if (test_opt(sb, DATA_ERR_ABORT))
2969 		journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2970 	else
2971 		journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2972 	spin_unlock(&journal->j_state_lock);
2973 }
2974 
2975 static journal_t *ext4_get_journal(struct super_block *sb,
2976 				   unsigned int journal_inum)
2977 {
2978 	struct inode *journal_inode;
2979 	journal_t *journal;
2980 
2981 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2982 
2983 	/* First, test for the existence of a valid inode on disk.  Bad
2984 	 * things happen if we iget() an unused inode, as the subsequent
2985 	 * iput() will try to delete it. */
2986 
2987 	journal_inode = ext4_iget(sb, journal_inum);
2988 	if (IS_ERR(journal_inode)) {
2989 		ext4_msg(sb, KERN_ERR, "no journal found");
2990 		return NULL;
2991 	}
2992 	if (!journal_inode->i_nlink) {
2993 		make_bad_inode(journal_inode);
2994 		iput(journal_inode);
2995 		ext4_msg(sb, KERN_ERR, "journal inode is deleted");
2996 		return NULL;
2997 	}
2998 
2999 	jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3000 		  journal_inode, journal_inode->i_size);
3001 	if (!S_ISREG(journal_inode->i_mode)) {
3002 		ext4_msg(sb, KERN_ERR, "invalid journal inode");
3003 		iput(journal_inode);
3004 		return NULL;
3005 	}
3006 
3007 	journal = jbd2_journal_init_inode(journal_inode);
3008 	if (!journal) {
3009 		ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3010 		iput(journal_inode);
3011 		return NULL;
3012 	}
3013 	journal->j_private = sb;
3014 	ext4_init_journal_params(sb, journal);
3015 	return journal;
3016 }
3017 
3018 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3019 				       dev_t j_dev)
3020 {
3021 	struct buffer_head *bh;
3022 	journal_t *journal;
3023 	ext4_fsblk_t start;
3024 	ext4_fsblk_t len;
3025 	int hblock, blocksize;
3026 	ext4_fsblk_t sb_block;
3027 	unsigned long offset;
3028 	struct ext4_super_block *es;
3029 	struct block_device *bdev;
3030 
3031 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3032 
3033 	bdev = ext4_blkdev_get(j_dev, sb);
3034 	if (bdev == NULL)
3035 		return NULL;
3036 
3037 	if (bd_claim(bdev, sb)) {
3038 		ext4_msg(sb, KERN_ERR,
3039 			"failed to claim external journal device");
3040 		blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3041 		return NULL;
3042 	}
3043 
3044 	blocksize = sb->s_blocksize;
3045 	hblock = bdev_logical_block_size(bdev);
3046 	if (blocksize < hblock) {
3047 		ext4_msg(sb, KERN_ERR,
3048 			"blocksize too small for journal device");
3049 		goto out_bdev;
3050 	}
3051 
3052 	sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3053 	offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3054 	set_blocksize(bdev, blocksize);
3055 	if (!(bh = __bread(bdev, sb_block, blocksize))) {
3056 		ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3057 		       "external journal");
3058 		goto out_bdev;
3059 	}
3060 
3061 	es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3062 	if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3063 	    !(le32_to_cpu(es->s_feature_incompat) &
3064 	      EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3065 		ext4_msg(sb, KERN_ERR, "external journal has "
3066 					"bad superblock");
3067 		brelse(bh);
3068 		goto out_bdev;
3069 	}
3070 
3071 	if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3072 		ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3073 		brelse(bh);
3074 		goto out_bdev;
3075 	}
3076 
3077 	len = ext4_blocks_count(es);
3078 	start = sb_block + 1;
3079 	brelse(bh);	/* we're done with the superblock */
3080 
3081 	journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3082 					start, len, blocksize);
3083 	if (!journal) {
3084 		ext4_msg(sb, KERN_ERR, "failed to create device journal");
3085 		goto out_bdev;
3086 	}
3087 	journal->j_private = sb;
3088 	ll_rw_block(READ, 1, &journal->j_sb_buffer);
3089 	wait_on_buffer(journal->j_sb_buffer);
3090 	if (!buffer_uptodate(journal->j_sb_buffer)) {
3091 		ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3092 		goto out_journal;
3093 	}
3094 	if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3095 		ext4_msg(sb, KERN_ERR, "External journal has more than one "
3096 					"user (unsupported) - %d",
3097 			be32_to_cpu(journal->j_superblock->s_nr_users));
3098 		goto out_journal;
3099 	}
3100 	EXT4_SB(sb)->journal_bdev = bdev;
3101 	ext4_init_journal_params(sb, journal);
3102 	return journal;
3103 
3104 out_journal:
3105 	jbd2_journal_destroy(journal);
3106 out_bdev:
3107 	ext4_blkdev_put(bdev);
3108 	return NULL;
3109 }
3110 
3111 static int ext4_load_journal(struct super_block *sb,
3112 			     struct ext4_super_block *es,
3113 			     unsigned long journal_devnum)
3114 {
3115 	journal_t *journal;
3116 	unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3117 	dev_t journal_dev;
3118 	int err = 0;
3119 	int really_read_only;
3120 
3121 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3122 
3123 	if (journal_devnum &&
3124 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3125 		ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3126 			"numbers have changed");
3127 		journal_dev = new_decode_dev(journal_devnum);
3128 	} else
3129 		journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3130 
3131 	really_read_only = bdev_read_only(sb->s_bdev);
3132 
3133 	/*
3134 	 * Are we loading a blank journal or performing recovery after a
3135 	 * crash?  For recovery, we need to check in advance whether we
3136 	 * can get read-write access to the device.
3137 	 */
3138 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3139 		if (sb->s_flags & MS_RDONLY) {
3140 			ext4_msg(sb, KERN_INFO, "INFO: recovery "
3141 					"required on readonly filesystem");
3142 			if (really_read_only) {
3143 				ext4_msg(sb, KERN_ERR, "write access "
3144 					"unavailable, cannot proceed");
3145 				return -EROFS;
3146 			}
3147 			ext4_msg(sb, KERN_INFO, "write access will "
3148 			       "be enabled during recovery");
3149 		}
3150 	}
3151 
3152 	if (journal_inum && journal_dev) {
3153 		ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3154 		       "and inode journals!");
3155 		return -EINVAL;
3156 	}
3157 
3158 	if (journal_inum) {
3159 		if (!(journal = ext4_get_journal(sb, journal_inum)))
3160 			return -EINVAL;
3161 	} else {
3162 		if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3163 			return -EINVAL;
3164 	}
3165 
3166 	if (journal->j_flags & JBD2_BARRIER)
3167 		ext4_msg(sb, KERN_INFO, "barriers enabled");
3168 	else
3169 		ext4_msg(sb, KERN_INFO, "barriers disabled");
3170 
3171 	if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3172 		err = jbd2_journal_update_format(journal);
3173 		if (err)  {
3174 			ext4_msg(sb, KERN_ERR, "error updating journal");
3175 			jbd2_journal_destroy(journal);
3176 			return err;
3177 		}
3178 	}
3179 
3180 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3181 		err = jbd2_journal_wipe(journal, !really_read_only);
3182 	if (!err)
3183 		err = jbd2_journal_load(journal);
3184 
3185 	if (err) {
3186 		ext4_msg(sb, KERN_ERR, "error loading journal");
3187 		jbd2_journal_destroy(journal);
3188 		return err;
3189 	}
3190 
3191 	EXT4_SB(sb)->s_journal = journal;
3192 	ext4_clear_journal_err(sb, es);
3193 
3194 	if (journal_devnum &&
3195 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3196 		es->s_journal_dev = cpu_to_le32(journal_devnum);
3197 
3198 		/* Make sure we flush the recovery flag to disk. */
3199 		ext4_commit_super(sb, 1);
3200 	}
3201 
3202 	return 0;
3203 }
3204 
3205 static int ext4_commit_super(struct super_block *sb, int sync)
3206 {
3207 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3208 	struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3209 	int error = 0;
3210 
3211 	if (!sbh)
3212 		return error;
3213 	if (buffer_write_io_error(sbh)) {
3214 		/*
3215 		 * Oh, dear.  A previous attempt to write the
3216 		 * superblock failed.  This could happen because the
3217 		 * USB device was yanked out.  Or it could happen to
3218 		 * be a transient write error and maybe the block will
3219 		 * be remapped.  Nothing we can do but to retry the
3220 		 * write and hope for the best.
3221 		 */
3222 		ext4_msg(sb, KERN_ERR, "previous I/O error to "
3223 		       "superblock detected");
3224 		clear_buffer_write_io_error(sbh);
3225 		set_buffer_uptodate(sbh);
3226 	}
3227 	es->s_wtime = cpu_to_le32(get_seconds());
3228 	es->s_kbytes_written =
3229 		cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3230 			    ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3231 			      EXT4_SB(sb)->s_sectors_written_start) >> 1));
3232 	ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3233 					&EXT4_SB(sb)->s_freeblocks_counter));
3234 	es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3235 					&EXT4_SB(sb)->s_freeinodes_counter));
3236 	sb->s_dirt = 0;
3237 	BUFFER_TRACE(sbh, "marking dirty");
3238 	mark_buffer_dirty(sbh);
3239 	if (sync) {
3240 		error = sync_dirty_buffer(sbh);
3241 		if (error)
3242 			return error;
3243 
3244 		error = buffer_write_io_error(sbh);
3245 		if (error) {
3246 			ext4_msg(sb, KERN_ERR, "I/O error while writing "
3247 			       "superblock");
3248 			clear_buffer_write_io_error(sbh);
3249 			set_buffer_uptodate(sbh);
3250 		}
3251 	}
3252 	return error;
3253 }
3254 
3255 /*
3256  * Have we just finished recovery?  If so, and if we are mounting (or
3257  * remounting) the filesystem readonly, then we will end up with a
3258  * consistent fs on disk.  Record that fact.
3259  */
3260 static void ext4_mark_recovery_complete(struct super_block *sb,
3261 					struct ext4_super_block *es)
3262 {
3263 	journal_t *journal = EXT4_SB(sb)->s_journal;
3264 
3265 	if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3266 		BUG_ON(journal != NULL);
3267 		return;
3268 	}
3269 	jbd2_journal_lock_updates(journal);
3270 	if (jbd2_journal_flush(journal) < 0)
3271 		goto out;
3272 
3273 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3274 	    sb->s_flags & MS_RDONLY) {
3275 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3276 		ext4_commit_super(sb, 1);
3277 	}
3278 
3279 out:
3280 	jbd2_journal_unlock_updates(journal);
3281 }
3282 
3283 /*
3284  * If we are mounting (or read-write remounting) a filesystem whose journal
3285  * has recorded an error from a previous lifetime, move that error to the
3286  * main filesystem now.
3287  */
3288 static void ext4_clear_journal_err(struct super_block *sb,
3289 				   struct ext4_super_block *es)
3290 {
3291 	journal_t *journal;
3292 	int j_errno;
3293 	const char *errstr;
3294 
3295 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3296 
3297 	journal = EXT4_SB(sb)->s_journal;
3298 
3299 	/*
3300 	 * Now check for any error status which may have been recorded in the
3301 	 * journal by a prior ext4_error() or ext4_abort()
3302 	 */
3303 
3304 	j_errno = jbd2_journal_errno(journal);
3305 	if (j_errno) {
3306 		char nbuf[16];
3307 
3308 		errstr = ext4_decode_error(sb, j_errno, nbuf);
3309 		ext4_warning(sb, __func__, "Filesystem error recorded "
3310 			     "from previous mount: %s", errstr);
3311 		ext4_warning(sb, __func__, "Marking fs in need of "
3312 			     "filesystem check.");
3313 
3314 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3315 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3316 		ext4_commit_super(sb, 1);
3317 
3318 		jbd2_journal_clear_err(journal);
3319 	}
3320 }
3321 
3322 /*
3323  * Force the running and committing transactions to commit,
3324  * and wait on the commit.
3325  */
3326 int ext4_force_commit(struct super_block *sb)
3327 {
3328 	journal_t *journal;
3329 	int ret = 0;
3330 
3331 	if (sb->s_flags & MS_RDONLY)
3332 		return 0;
3333 
3334 	journal = EXT4_SB(sb)->s_journal;
3335 	if (journal)
3336 		ret = ext4_journal_force_commit(journal);
3337 
3338 	return ret;
3339 }
3340 
3341 static void ext4_write_super(struct super_block *sb)
3342 {
3343 	lock_super(sb);
3344 	ext4_commit_super(sb, 1);
3345 	unlock_super(sb);
3346 }
3347 
3348 static int ext4_sync_fs(struct super_block *sb, int wait)
3349 {
3350 	int ret = 0;
3351 	tid_t target;
3352 
3353 	trace_ext4_sync_fs(sb, wait);
3354 	if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
3355 		if (wait)
3356 			jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
3357 	}
3358 	return ret;
3359 }
3360 
3361 /*
3362  * LVM calls this function before a (read-only) snapshot is created.  This
3363  * gives us a chance to flush the journal completely and mark the fs clean.
3364  */
3365 static int ext4_freeze(struct super_block *sb)
3366 {
3367 	int error = 0;
3368 	journal_t *journal;
3369 
3370 	if (sb->s_flags & MS_RDONLY)
3371 		return 0;
3372 
3373 	journal = EXT4_SB(sb)->s_journal;
3374 
3375 	/* Now we set up the journal barrier. */
3376 	jbd2_journal_lock_updates(journal);
3377 
3378 	/*
3379 	 * Don't clear the needs_recovery flag if we failed to flush
3380 	 * the journal.
3381 	 */
3382 	error = jbd2_journal_flush(journal);
3383 	if (error < 0) {
3384 	out:
3385 		jbd2_journal_unlock_updates(journal);
3386 		return error;
3387 	}
3388 
3389 	/* Journal blocked and flushed, clear needs_recovery flag. */
3390 	EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3391 	error = ext4_commit_super(sb, 1);
3392 	if (error)
3393 		goto out;
3394 	return 0;
3395 }
3396 
3397 /*
3398  * Called by LVM after the snapshot is done.  We need to reset the RECOVER
3399  * flag here, even though the filesystem is not technically dirty yet.
3400  */
3401 static int ext4_unfreeze(struct super_block *sb)
3402 {
3403 	if (sb->s_flags & MS_RDONLY)
3404 		return 0;
3405 
3406 	lock_super(sb);
3407 	/* Reset the needs_recovery flag before the fs is unlocked. */
3408 	EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3409 	ext4_commit_super(sb, 1);
3410 	unlock_super(sb);
3411 	jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3412 	return 0;
3413 }
3414 
3415 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3416 {
3417 	struct ext4_super_block *es;
3418 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3419 	ext4_fsblk_t n_blocks_count = 0;
3420 	unsigned long old_sb_flags;
3421 	struct ext4_mount_options old_opts;
3422 	ext4_group_t g;
3423 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3424 	int err;
3425 #ifdef CONFIG_QUOTA
3426 	int i;
3427 #endif
3428 
3429 	lock_kernel();
3430 
3431 	/* Store the original options */
3432 	lock_super(sb);
3433 	old_sb_flags = sb->s_flags;
3434 	old_opts.s_mount_opt = sbi->s_mount_opt;
3435 	old_opts.s_resuid = sbi->s_resuid;
3436 	old_opts.s_resgid = sbi->s_resgid;
3437 	old_opts.s_commit_interval = sbi->s_commit_interval;
3438 	old_opts.s_min_batch_time = sbi->s_min_batch_time;
3439 	old_opts.s_max_batch_time = sbi->s_max_batch_time;
3440 #ifdef CONFIG_QUOTA
3441 	old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3442 	for (i = 0; i < MAXQUOTAS; i++)
3443 		old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3444 #endif
3445 	if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3446 		journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3447 
3448 	/*
3449 	 * Allow the "check" option to be passed as a remount option.
3450 	 */
3451 	if (!parse_options(data, sb, NULL, &journal_ioprio,
3452 			   &n_blocks_count, 1)) {
3453 		err = -EINVAL;
3454 		goto restore_opts;
3455 	}
3456 
3457 	if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3458 		ext4_abort(sb, __func__, "Abort forced by user");
3459 
3460 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3461 		((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3462 
3463 	es = sbi->s_es;
3464 
3465 	if (sbi->s_journal) {
3466 		ext4_init_journal_params(sb, sbi->s_journal);
3467 		set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3468 	}
3469 
3470 	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3471 		n_blocks_count > ext4_blocks_count(es)) {
3472 		if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3473 			err = -EROFS;
3474 			goto restore_opts;
3475 		}
3476 
3477 		if (*flags & MS_RDONLY) {
3478 			/*
3479 			 * First of all, the unconditional stuff we have to do
3480 			 * to disable replay of the journal when we next remount
3481 			 */
3482 			sb->s_flags |= MS_RDONLY;
3483 
3484 			/*
3485 			 * OK, test if we are remounting a valid rw partition
3486 			 * readonly, and if so set the rdonly flag and then
3487 			 * mark the partition as valid again.
3488 			 */
3489 			if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3490 			    (sbi->s_mount_state & EXT4_VALID_FS))
3491 				es->s_state = cpu_to_le16(sbi->s_mount_state);
3492 
3493 			if (sbi->s_journal)
3494 				ext4_mark_recovery_complete(sb, es);
3495 		} else {
3496 			int ret;
3497 			if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3498 					~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3499 				ext4_msg(sb, KERN_WARNING, "couldn't "
3500 				       "remount RDWR because of unsupported "
3501 				       "optional features (%x)",
3502 				(le32_to_cpu(sbi->s_es->s_feature_ro_compat) &
3503 					~EXT4_FEATURE_RO_COMPAT_SUPP));
3504 				err = -EROFS;
3505 				goto restore_opts;
3506 			}
3507 
3508 			/*
3509 			 * Make sure the group descriptor checksums
3510 			 * are sane.  If they aren't, refuse to remount r/w.
3511 			 */
3512 			for (g = 0; g < sbi->s_groups_count; g++) {
3513 				struct ext4_group_desc *gdp =
3514 					ext4_get_group_desc(sb, g, NULL);
3515 
3516 				if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3517 					ext4_msg(sb, KERN_ERR,
3518 	       "ext4_remount: Checksum for group %u failed (%u!=%u)",
3519 		g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3520 					       le16_to_cpu(gdp->bg_checksum));
3521 					err = -EINVAL;
3522 					goto restore_opts;
3523 				}
3524 			}
3525 
3526 			/*
3527 			 * If we have an unprocessed orphan list hanging
3528 			 * around from a previously readonly bdev mount,
3529 			 * require a full umount/remount for now.
3530 			 */
3531 			if (es->s_last_orphan) {
3532 				ext4_msg(sb, KERN_WARNING, "Couldn't "
3533 				       "remount RDWR because of unprocessed "
3534 				       "orphan inode list.  Please "
3535 				       "umount/remount instead");
3536 				err = -EINVAL;
3537 				goto restore_opts;
3538 			}
3539 
3540 			/*
3541 			 * Mounting a RDONLY partition read-write, so reread
3542 			 * and store the current valid flag.  (It may have
3543 			 * been changed by e2fsck since we originally mounted
3544 			 * the partition.)
3545 			 */
3546 			if (sbi->s_journal)
3547 				ext4_clear_journal_err(sb, es);
3548 			sbi->s_mount_state = le16_to_cpu(es->s_state);
3549 			if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3550 				goto restore_opts;
3551 			if (!ext4_setup_super(sb, es, 0))
3552 				sb->s_flags &= ~MS_RDONLY;
3553 		}
3554 	}
3555 	ext4_setup_system_zone(sb);
3556 	if (sbi->s_journal == NULL)
3557 		ext4_commit_super(sb, 1);
3558 
3559 #ifdef CONFIG_QUOTA
3560 	/* Release old quota file names */
3561 	for (i = 0; i < MAXQUOTAS; i++)
3562 		if (old_opts.s_qf_names[i] &&
3563 		    old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3564 			kfree(old_opts.s_qf_names[i]);
3565 #endif
3566 	unlock_super(sb);
3567 	unlock_kernel();
3568 	return 0;
3569 
3570 restore_opts:
3571 	sb->s_flags = old_sb_flags;
3572 	sbi->s_mount_opt = old_opts.s_mount_opt;
3573 	sbi->s_resuid = old_opts.s_resuid;
3574 	sbi->s_resgid = old_opts.s_resgid;
3575 	sbi->s_commit_interval = old_opts.s_commit_interval;
3576 	sbi->s_min_batch_time = old_opts.s_min_batch_time;
3577 	sbi->s_max_batch_time = old_opts.s_max_batch_time;
3578 #ifdef CONFIG_QUOTA
3579 	sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3580 	for (i = 0; i < MAXQUOTAS; i++) {
3581 		if (sbi->s_qf_names[i] &&
3582 		    old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3583 			kfree(sbi->s_qf_names[i]);
3584 		sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3585 	}
3586 #endif
3587 	unlock_super(sb);
3588 	unlock_kernel();
3589 	return err;
3590 }
3591 
3592 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3593 {
3594 	struct super_block *sb = dentry->d_sb;
3595 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3596 	struct ext4_super_block *es = sbi->s_es;
3597 	u64 fsid;
3598 
3599 	if (test_opt(sb, MINIX_DF)) {
3600 		sbi->s_overhead_last = 0;
3601 	} else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3602 		ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3603 		ext4_fsblk_t overhead = 0;
3604 
3605 		/*
3606 		 * Compute the overhead (FS structures).  This is constant
3607 		 * for a given filesystem unless the number of block groups
3608 		 * changes so we cache the previous value until it does.
3609 		 */
3610 
3611 		/*
3612 		 * All of the blocks before first_data_block are
3613 		 * overhead
3614 		 */
3615 		overhead = le32_to_cpu(es->s_first_data_block);
3616 
3617 		/*
3618 		 * Add the overhead attributed to the superblock and
3619 		 * block group descriptors.  If the sparse superblocks
3620 		 * feature is turned on, then not all groups have this.
3621 		 */
3622 		for (i = 0; i < ngroups; i++) {
3623 			overhead += ext4_bg_has_super(sb, i) +
3624 				ext4_bg_num_gdb(sb, i);
3625 			cond_resched();
3626 		}
3627 
3628 		/*
3629 		 * Every block group has an inode bitmap, a block
3630 		 * bitmap, and an inode table.
3631 		 */
3632 		overhead += ngroups * (2 + sbi->s_itb_per_group);
3633 		sbi->s_overhead_last = overhead;
3634 		smp_wmb();
3635 		sbi->s_blocks_last = ext4_blocks_count(es);
3636 	}
3637 
3638 	buf->f_type = EXT4_SUPER_MAGIC;
3639 	buf->f_bsize = sb->s_blocksize;
3640 	buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3641 	buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3642 		       percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3643 	ext4_free_blocks_count_set(es, buf->f_bfree);
3644 	buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3645 	if (buf->f_bfree < ext4_r_blocks_count(es))
3646 		buf->f_bavail = 0;
3647 	buf->f_files = le32_to_cpu(es->s_inodes_count);
3648 	buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3649 	es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3650 	buf->f_namelen = EXT4_NAME_LEN;
3651 	fsid = le64_to_cpup((void *)es->s_uuid) ^
3652 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3653 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3654 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3655 
3656 	return 0;
3657 }
3658 
3659 /* Helper function for writing quotas on sync - we need to start transaction
3660  * before quota file is locked for write. Otherwise the are possible deadlocks:
3661  * Process 1                         Process 2
3662  * ext4_create()                     quota_sync()
3663  *   jbd2_journal_start()                  write_dquot()
3664  *   vfs_dq_init()                         down(dqio_mutex)
3665  *     down(dqio_mutex)                    jbd2_journal_start()
3666  *
3667  */
3668 
3669 #ifdef CONFIG_QUOTA
3670 
3671 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3672 {
3673 	return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3674 }
3675 
3676 static int ext4_write_dquot(struct dquot *dquot)
3677 {
3678 	int ret, err;
3679 	handle_t *handle;
3680 	struct inode *inode;
3681 
3682 	inode = dquot_to_inode(dquot);
3683 	handle = ext4_journal_start(inode,
3684 				    EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3685 	if (IS_ERR(handle))
3686 		return PTR_ERR(handle);
3687 	ret = dquot_commit(dquot);
3688 	err = ext4_journal_stop(handle);
3689 	if (!ret)
3690 		ret = err;
3691 	return ret;
3692 }
3693 
3694 static int ext4_acquire_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_INIT_BLOCKS(dquot->dq_sb));
3701 	if (IS_ERR(handle))
3702 		return PTR_ERR(handle);
3703 	ret = dquot_acquire(dquot);
3704 	err = ext4_journal_stop(handle);
3705 	if (!ret)
3706 		ret = err;
3707 	return ret;
3708 }
3709 
3710 static int ext4_release_dquot(struct dquot *dquot)
3711 {
3712 	int ret, err;
3713 	handle_t *handle;
3714 
3715 	handle = ext4_journal_start(dquot_to_inode(dquot),
3716 				    EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3717 	if (IS_ERR(handle)) {
3718 		/* Release dquot anyway to avoid endless cycle in dqput() */
3719 		dquot_release(dquot);
3720 		return PTR_ERR(handle);
3721 	}
3722 	ret = dquot_release(dquot);
3723 	err = ext4_journal_stop(handle);
3724 	if (!ret)
3725 		ret = err;
3726 	return ret;
3727 }
3728 
3729 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3730 {
3731 	/* Are we journaling quotas? */
3732 	if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3733 	    EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3734 		dquot_mark_dquot_dirty(dquot);
3735 		return ext4_write_dquot(dquot);
3736 	} else {
3737 		return dquot_mark_dquot_dirty(dquot);
3738 	}
3739 }
3740 
3741 static int ext4_write_info(struct super_block *sb, int type)
3742 {
3743 	int ret, err;
3744 	handle_t *handle;
3745 
3746 	/* Data block + inode block */
3747 	handle = ext4_journal_start(sb->s_root->d_inode, 2);
3748 	if (IS_ERR(handle))
3749 		return PTR_ERR(handle);
3750 	ret = dquot_commit_info(sb, type);
3751 	err = ext4_journal_stop(handle);
3752 	if (!ret)
3753 		ret = err;
3754 	return ret;
3755 }
3756 
3757 /*
3758  * Turn on quotas during mount time - we need to find
3759  * the quota file and such...
3760  */
3761 static int ext4_quota_on_mount(struct super_block *sb, int type)
3762 {
3763 	return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3764 				  EXT4_SB(sb)->s_jquota_fmt, type);
3765 }
3766 
3767 /*
3768  * Standard function to be called on quota_on
3769  */
3770 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3771 			 char *name, int remount)
3772 {
3773 	int err;
3774 	struct path path;
3775 
3776 	if (!test_opt(sb, QUOTA))
3777 		return -EINVAL;
3778 	/* When remounting, no checks are needed and in fact, name is NULL */
3779 	if (remount)
3780 		return vfs_quota_on(sb, type, format_id, name, remount);
3781 
3782 	err = kern_path(name, LOOKUP_FOLLOW, &path);
3783 	if (err)
3784 		return err;
3785 
3786 	/* Quotafile not on the same filesystem? */
3787 	if (path.mnt->mnt_sb != sb) {
3788 		path_put(&path);
3789 		return -EXDEV;
3790 	}
3791 	/* Journaling quota? */
3792 	if (EXT4_SB(sb)->s_qf_names[type]) {
3793 		/* Quotafile not in fs root? */
3794 		if (path.dentry->d_parent != sb->s_root)
3795 			ext4_msg(sb, KERN_WARNING,
3796 				"Quota file not on filesystem root. "
3797 				"Journaled quota will not work");
3798 	}
3799 
3800 	/*
3801 	 * When we journal data on quota file, we have to flush journal to see
3802 	 * all updates to the file when we bypass pagecache...
3803 	 */
3804 	if (EXT4_SB(sb)->s_journal &&
3805 	    ext4_should_journal_data(path.dentry->d_inode)) {
3806 		/*
3807 		 * We don't need to lock updates but journal_flush() could
3808 		 * otherwise be livelocked...
3809 		 */
3810 		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3811 		err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3812 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3813 		if (err) {
3814 			path_put(&path);
3815 			return err;
3816 		}
3817 	}
3818 
3819 	err = vfs_quota_on_path(sb, type, format_id, &path);
3820 	path_put(&path);
3821 	return err;
3822 }
3823 
3824 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3825  * acquiring the locks... As quota files are never truncated and quota code
3826  * itself serializes the operations (and noone else should touch the files)
3827  * we don't have to be afraid of races */
3828 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3829 			       size_t len, loff_t off)
3830 {
3831 	struct inode *inode = sb_dqopt(sb)->files[type];
3832 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3833 	int err = 0;
3834 	int offset = off & (sb->s_blocksize - 1);
3835 	int tocopy;
3836 	size_t toread;
3837 	struct buffer_head *bh;
3838 	loff_t i_size = i_size_read(inode);
3839 
3840 	if (off > i_size)
3841 		return 0;
3842 	if (off+len > i_size)
3843 		len = i_size-off;
3844 	toread = len;
3845 	while (toread > 0) {
3846 		tocopy = sb->s_blocksize - offset < toread ?
3847 				sb->s_blocksize - offset : toread;
3848 		bh = ext4_bread(NULL, inode, blk, 0, &err);
3849 		if (err)
3850 			return err;
3851 		if (!bh)	/* A hole? */
3852 			memset(data, 0, tocopy);
3853 		else
3854 			memcpy(data, bh->b_data+offset, tocopy);
3855 		brelse(bh);
3856 		offset = 0;
3857 		toread -= tocopy;
3858 		data += tocopy;
3859 		blk++;
3860 	}
3861 	return len;
3862 }
3863 
3864 /* Write to quotafile (we know the transaction is already started and has
3865  * enough credits) */
3866 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3867 				const char *data, size_t len, loff_t off)
3868 {
3869 	struct inode *inode = sb_dqopt(sb)->files[type];
3870 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3871 	int err = 0;
3872 	int offset = off & (sb->s_blocksize - 1);
3873 	int tocopy;
3874 	int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3875 	size_t towrite = len;
3876 	struct buffer_head *bh;
3877 	handle_t *handle = journal_current_handle();
3878 
3879 	if (EXT4_SB(sb)->s_journal && !handle) {
3880 		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3881 			" cancelled because transaction is not started",
3882 			(unsigned long long)off, (unsigned long long)len);
3883 		return -EIO;
3884 	}
3885 	mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3886 	while (towrite > 0) {
3887 		tocopy = sb->s_blocksize - offset < towrite ?
3888 				sb->s_blocksize - offset : towrite;
3889 		bh = ext4_bread(handle, inode, blk, 1, &err);
3890 		if (!bh)
3891 			goto out;
3892 		if (journal_quota) {
3893 			err = ext4_journal_get_write_access(handle, bh);
3894 			if (err) {
3895 				brelse(bh);
3896 				goto out;
3897 			}
3898 		}
3899 		lock_buffer(bh);
3900 		memcpy(bh->b_data+offset, data, tocopy);
3901 		flush_dcache_page(bh->b_page);
3902 		unlock_buffer(bh);
3903 		if (journal_quota)
3904 			err = ext4_handle_dirty_metadata(handle, NULL, bh);
3905 		else {
3906 			/* Always do at least ordered writes for quotas */
3907 			err = ext4_jbd2_file_inode(handle, inode);
3908 			mark_buffer_dirty(bh);
3909 		}
3910 		brelse(bh);
3911 		if (err)
3912 			goto out;
3913 		offset = 0;
3914 		towrite -= tocopy;
3915 		data += tocopy;
3916 		blk++;
3917 	}
3918 out:
3919 	if (len == towrite) {
3920 		mutex_unlock(&inode->i_mutex);
3921 		return err;
3922 	}
3923 	if (inode->i_size < off+len-towrite) {
3924 		i_size_write(inode, off+len-towrite);
3925 		EXT4_I(inode)->i_disksize = inode->i_size;
3926 	}
3927 	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3928 	ext4_mark_inode_dirty(handle, inode);
3929 	mutex_unlock(&inode->i_mutex);
3930 	return len - towrite;
3931 }
3932 
3933 #endif
3934 
3935 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3936 		       const char *dev_name, void *data, struct vfsmount *mnt)
3937 {
3938 	return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3939 }
3940 
3941 static struct file_system_type ext4_fs_type = {
3942 	.owner		= THIS_MODULE,
3943 	.name		= "ext4",
3944 	.get_sb		= ext4_get_sb,
3945 	.kill_sb	= kill_block_super,
3946 	.fs_flags	= FS_REQUIRES_DEV,
3947 };
3948 
3949 #ifdef CONFIG_EXT4DEV_COMPAT
3950 static int ext4dev_get_sb(struct file_system_type *fs_type, int flags,
3951 			  const char *dev_name, void *data,struct vfsmount *mnt)
3952 {
3953 	printk(KERN_WARNING "EXT4-fs (%s): Update your userspace programs "
3954 	       "to mount using ext4\n", dev_name);
3955 	printk(KERN_WARNING "EXT4-fs (%s): ext4dev backwards compatibility "
3956 	       "will go away by 2.6.31\n", dev_name);
3957 	return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3958 }
3959 
3960 static struct file_system_type ext4dev_fs_type = {
3961 	.owner		= THIS_MODULE,
3962 	.name		= "ext4dev",
3963 	.get_sb		= ext4dev_get_sb,
3964 	.kill_sb	= kill_block_super,
3965 	.fs_flags	= FS_REQUIRES_DEV,
3966 };
3967 MODULE_ALIAS("ext4dev");
3968 #endif
3969 
3970 static int __init init_ext4_fs(void)
3971 {
3972 	int err;
3973 
3974 	err = init_ext4_system_zone();
3975 	if (err)
3976 		return err;
3977 	ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3978 	if (!ext4_kset)
3979 		goto out4;
3980 	ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3981 	err = init_ext4_mballoc();
3982 	if (err)
3983 		goto out3;
3984 
3985 	err = init_ext4_xattr();
3986 	if (err)
3987 		goto out2;
3988 	err = init_inodecache();
3989 	if (err)
3990 		goto out1;
3991 	err = register_filesystem(&ext4_fs_type);
3992 	if (err)
3993 		goto out;
3994 #ifdef CONFIG_EXT4DEV_COMPAT
3995 	err = register_filesystem(&ext4dev_fs_type);
3996 	if (err) {
3997 		unregister_filesystem(&ext4_fs_type);
3998 		goto out;
3999 	}
4000 #endif
4001 	return 0;
4002 out:
4003 	destroy_inodecache();
4004 out1:
4005 	exit_ext4_xattr();
4006 out2:
4007 	exit_ext4_mballoc();
4008 out3:
4009 	remove_proc_entry("fs/ext4", NULL);
4010 	kset_unregister(ext4_kset);
4011 out4:
4012 	exit_ext4_system_zone();
4013 	return err;
4014 }
4015 
4016 static void __exit exit_ext4_fs(void)
4017 {
4018 	unregister_filesystem(&ext4_fs_type);
4019 #ifdef CONFIG_EXT4DEV_COMPAT
4020 	unregister_filesystem(&ext4dev_fs_type);
4021 #endif
4022 	destroy_inodecache();
4023 	exit_ext4_xattr();
4024 	exit_ext4_mballoc();
4025 	remove_proc_entry("fs/ext4", NULL);
4026 	kset_unregister(ext4_kset);
4027 	exit_ext4_system_zone();
4028 }
4029 
4030 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4031 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4032 MODULE_LICENSE("GPL");
4033 module_init(init_ext4_fs)
4034 module_exit(exit_ext4_fs)
4035