xref: /openbmc/linux/fs/ext4/super.c (revision 1d1fe1ee02b9ac2660995b10e35dd41448fef011)
1 /*
2  *  linux/fs/ext4/super.c
3  *
4  * Copyright (C) 1992, 1993, 1994, 1995
5  * Remy Card (card@masi.ibp.fr)
6  * Laboratoire MASI - Institut Blaise Pascal
7  * Universite Pierre et Marie Curie (Paris VI)
8  *
9  *  from
10  *
11  *  linux/fs/minix/inode.c
12  *
13  *  Copyright (C) 1991, 1992  Linus Torvalds
14  *
15  *  Big-endian to little-endian byte-swapping/bitmaps by
16  *        David S. Miller (davem@caip.rutgers.edu), 1995
17  */
18 
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/time.h>
23 #include <linux/jbd2.h>
24 #include <linux/ext4_fs.h>
25 #include <linux/ext4_jbd2.h>
26 #include <linux/slab.h>
27 #include <linux/init.h>
28 #include <linux/blkdev.h>
29 #include <linux/parser.h>
30 #include <linux/smp_lock.h>
31 #include <linux/buffer_head.h>
32 #include <linux/exportfs.h>
33 #include <linux/vfs.h>
34 #include <linux/random.h>
35 #include <linux/mount.h>
36 #include <linux/namei.h>
37 #include <linux/quotaops.h>
38 #include <linux/seq_file.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 
42 #include <asm/uaccess.h>
43 
44 #include "xattr.h"
45 #include "acl.h"
46 #include "namei.h"
47 #include "group.h"
48 
49 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
50 			     unsigned long journal_devnum);
51 static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
52 			       unsigned int);
53 static void ext4_commit_super (struct super_block * sb,
54 			       struct ext4_super_block * es,
55 			       int sync);
56 static void ext4_mark_recovery_complete(struct super_block * sb,
57 					struct ext4_super_block * es);
58 static void ext4_clear_journal_err(struct super_block * sb,
59 				   struct ext4_super_block * es);
60 static int ext4_sync_fs(struct super_block *sb, int wait);
61 static const char *ext4_decode_error(struct super_block * sb, int errno,
62 				     char nbuf[16]);
63 static int ext4_remount (struct super_block * sb, int * flags, char * data);
64 static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf);
65 static void ext4_unlockfs(struct super_block *sb);
66 static void ext4_write_super (struct super_block * sb);
67 static void ext4_write_super_lockfs(struct super_block *sb);
68 
69 
70 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
71 			       struct ext4_group_desc *bg)
72 {
73 	return le32_to_cpu(bg->bg_block_bitmap_lo) |
74 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
75 		(ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
76 }
77 
78 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
79 			       struct ext4_group_desc *bg)
80 {
81 	return le32_to_cpu(bg->bg_inode_bitmap_lo) |
82 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
83 		(ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
84 }
85 
86 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
87 			      struct ext4_group_desc *bg)
88 {
89 	return le32_to_cpu(bg->bg_inode_table_lo) |
90 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
91 		(ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
92 }
93 
94 void ext4_block_bitmap_set(struct super_block *sb,
95 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
96 {
97 	bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
98 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
99 		bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
100 }
101 
102 void ext4_inode_bitmap_set(struct super_block *sb,
103 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
104 {
105 	bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
106 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
107 		bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
108 }
109 
110 void ext4_inode_table_set(struct super_block *sb,
111 			  struct ext4_group_desc *bg, ext4_fsblk_t blk)
112 {
113 	bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
114 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
115 		bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
116 }
117 
118 /*
119  * Wrappers for jbd2_journal_start/end.
120  *
121  * The only special thing we need to do here is to make sure that all
122  * journal_end calls result in the superblock being marked dirty, so
123  * that sync() will call the filesystem's write_super callback if
124  * appropriate.
125  */
126 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
127 {
128 	journal_t *journal;
129 
130 	if (sb->s_flags & MS_RDONLY)
131 		return ERR_PTR(-EROFS);
132 
133 	/* Special case here: if the journal has aborted behind our
134 	 * backs (eg. EIO in the commit thread), then we still need to
135 	 * take the FS itself readonly cleanly. */
136 	journal = EXT4_SB(sb)->s_journal;
137 	if (is_journal_aborted(journal)) {
138 		ext4_abort(sb, __FUNCTION__,
139 			   "Detected aborted journal");
140 		return ERR_PTR(-EROFS);
141 	}
142 
143 	return jbd2_journal_start(journal, nblocks);
144 }
145 
146 /*
147  * The only special thing we need to do here is to make sure that all
148  * jbd2_journal_stop calls result in the superblock being marked dirty, so
149  * that sync() will call the filesystem's write_super callback if
150  * appropriate.
151  */
152 int __ext4_journal_stop(const char *where, handle_t *handle)
153 {
154 	struct super_block *sb;
155 	int err;
156 	int rc;
157 
158 	sb = handle->h_transaction->t_journal->j_private;
159 	err = handle->h_err;
160 	rc = jbd2_journal_stop(handle);
161 
162 	if (!err)
163 		err = rc;
164 	if (err)
165 		__ext4_std_error(sb, where, err);
166 	return err;
167 }
168 
169 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
170 		struct buffer_head *bh, handle_t *handle, int err)
171 {
172 	char nbuf[16];
173 	const char *errstr = ext4_decode_error(NULL, err, nbuf);
174 
175 	if (bh)
176 		BUFFER_TRACE(bh, "abort");
177 
178 	if (!handle->h_err)
179 		handle->h_err = err;
180 
181 	if (is_handle_aborted(handle))
182 		return;
183 
184 	printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
185 	       caller, errstr, err_fn);
186 
187 	jbd2_journal_abort_handle(handle);
188 }
189 
190 /* Deal with the reporting of failure conditions on a filesystem such as
191  * inconsistencies detected or read IO failures.
192  *
193  * On ext2, we can store the error state of the filesystem in the
194  * superblock.  That is not possible on ext4, because we may have other
195  * write ordering constraints on the superblock which prevent us from
196  * writing it out straight away; and given that the journal is about to
197  * be aborted, we can't rely on the current, or future, transactions to
198  * write out the superblock safely.
199  *
200  * We'll just use the jbd2_journal_abort() error code to record an error in
201  * the journal instead.  On recovery, the journal will compain about
202  * that error until we've noted it down and cleared it.
203  */
204 
205 static void ext4_handle_error(struct super_block *sb)
206 {
207 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
208 
209 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
210 	es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
211 
212 	if (sb->s_flags & MS_RDONLY)
213 		return;
214 
215 	if (!test_opt (sb, ERRORS_CONT)) {
216 		journal_t *journal = EXT4_SB(sb)->s_journal;
217 
218 		EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
219 		if (journal)
220 			jbd2_journal_abort(journal, -EIO);
221 	}
222 	if (test_opt (sb, ERRORS_RO)) {
223 		printk (KERN_CRIT "Remounting filesystem read-only\n");
224 		sb->s_flags |= MS_RDONLY;
225 	}
226 	ext4_commit_super(sb, es, 1);
227 	if (test_opt(sb, ERRORS_PANIC))
228 		panic("EXT4-fs (device %s): panic forced after error\n",
229 			sb->s_id);
230 }
231 
232 void ext4_error (struct super_block * sb, const char * function,
233 		 const char * fmt, ...)
234 {
235 	va_list args;
236 
237 	va_start(args, fmt);
238 	printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
239 	vprintk(fmt, args);
240 	printk("\n");
241 	va_end(args);
242 
243 	ext4_handle_error(sb);
244 }
245 
246 static const char *ext4_decode_error(struct super_block * sb, int errno,
247 				     char nbuf[16])
248 {
249 	char *errstr = NULL;
250 
251 	switch (errno) {
252 	case -EIO:
253 		errstr = "IO failure";
254 		break;
255 	case -ENOMEM:
256 		errstr = "Out of memory";
257 		break;
258 	case -EROFS:
259 		if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
260 			errstr = "Journal has aborted";
261 		else
262 			errstr = "Readonly filesystem";
263 		break;
264 	default:
265 		/* If the caller passed in an extra buffer for unknown
266 		 * errors, textualise them now.  Else we just return
267 		 * NULL. */
268 		if (nbuf) {
269 			/* Check for truncated error codes... */
270 			if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
271 				errstr = nbuf;
272 		}
273 		break;
274 	}
275 
276 	return errstr;
277 }
278 
279 /* __ext4_std_error decodes expected errors from journaling functions
280  * automatically and invokes the appropriate error response.  */
281 
282 void __ext4_std_error (struct super_block * sb, const char * function,
283 		       int errno)
284 {
285 	char nbuf[16];
286 	const char *errstr;
287 
288 	/* Special case: if the error is EROFS, and we're not already
289 	 * inside a transaction, then there's really no point in logging
290 	 * an error. */
291 	if (errno == -EROFS && journal_current_handle() == NULL &&
292 	    (sb->s_flags & MS_RDONLY))
293 		return;
294 
295 	errstr = ext4_decode_error(sb, errno, nbuf);
296 	printk (KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
297 		sb->s_id, function, errstr);
298 
299 	ext4_handle_error(sb);
300 }
301 
302 /*
303  * ext4_abort is a much stronger failure handler than ext4_error.  The
304  * abort function may be used to deal with unrecoverable failures such
305  * as journal IO errors or ENOMEM at a critical moment in log management.
306  *
307  * We unconditionally force the filesystem into an ABORT|READONLY state,
308  * unless the error response on the fs has been set to panic in which
309  * case we take the easy way out and panic immediately.
310  */
311 
312 void ext4_abort (struct super_block * sb, const char * function,
313 		 const char * fmt, ...)
314 {
315 	va_list args;
316 
317 	printk (KERN_CRIT "ext4_abort called.\n");
318 
319 	va_start(args, fmt);
320 	printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
321 	vprintk(fmt, args);
322 	printk("\n");
323 	va_end(args);
324 
325 	if (test_opt(sb, ERRORS_PANIC))
326 		panic("EXT4-fs panic from previous error\n");
327 
328 	if (sb->s_flags & MS_RDONLY)
329 		return;
330 
331 	printk(KERN_CRIT "Remounting filesystem read-only\n");
332 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
333 	sb->s_flags |= MS_RDONLY;
334 	EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
335 	jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
336 }
337 
338 void ext4_warning (struct super_block * sb, const char * function,
339 		   const char * fmt, ...)
340 {
341 	va_list args;
342 
343 	va_start(args, fmt);
344 	printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
345 	       sb->s_id, function);
346 	vprintk(fmt, args);
347 	printk("\n");
348 	va_end(args);
349 }
350 
351 void ext4_update_dynamic_rev(struct super_block *sb)
352 {
353 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
354 
355 	if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
356 		return;
357 
358 	ext4_warning(sb, __FUNCTION__,
359 		     "updating to rev %d because of new feature flag, "
360 		     "running e2fsck is recommended",
361 		     EXT4_DYNAMIC_REV);
362 
363 	es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
364 	es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
365 	es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
366 	/* leave es->s_feature_*compat flags alone */
367 	/* es->s_uuid will be set by e2fsck if empty */
368 
369 	/*
370 	 * The rest of the superblock fields should be zero, and if not it
371 	 * means they are likely already in use, so leave them alone.  We
372 	 * can leave it up to e2fsck to clean up any inconsistencies there.
373 	 */
374 }
375 
376 int ext4_update_compat_feature(handle_t *handle,
377 					struct super_block *sb, __u32 compat)
378 {
379 	int err = 0;
380 	if (!EXT4_HAS_COMPAT_FEATURE(sb, compat)) {
381 		err = ext4_journal_get_write_access(handle,
382 				EXT4_SB(sb)->s_sbh);
383 		if (err)
384 			return err;
385 		EXT4_SET_COMPAT_FEATURE(sb, compat);
386 		sb->s_dirt = 1;
387 		handle->h_sync = 1;
388 		BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
389 					"call ext4_journal_dirty_met adata");
390 		err = ext4_journal_dirty_metadata(handle,
391 				EXT4_SB(sb)->s_sbh);
392 	}
393 	return err;
394 }
395 
396 int ext4_update_rocompat_feature(handle_t *handle,
397 					struct super_block *sb, __u32 rocompat)
398 {
399 	int err = 0;
400 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, rocompat)) {
401 		err = ext4_journal_get_write_access(handle,
402 				EXT4_SB(sb)->s_sbh);
403 		if (err)
404 			return err;
405 		EXT4_SET_RO_COMPAT_FEATURE(sb, rocompat);
406 		sb->s_dirt = 1;
407 		handle->h_sync = 1;
408 		BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
409 					"call ext4_journal_dirty_met adata");
410 		err = ext4_journal_dirty_metadata(handle,
411 				EXT4_SB(sb)->s_sbh);
412 	}
413 	return err;
414 }
415 
416 int ext4_update_incompat_feature(handle_t *handle,
417 					struct super_block *sb, __u32 incompat)
418 {
419 	int err = 0;
420 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, incompat)) {
421 		err = ext4_journal_get_write_access(handle,
422 				EXT4_SB(sb)->s_sbh);
423 		if (err)
424 			return err;
425 		EXT4_SET_INCOMPAT_FEATURE(sb, incompat);
426 		sb->s_dirt = 1;
427 		handle->h_sync = 1;
428 		BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
429 					"call ext4_journal_dirty_met adata");
430 		err = ext4_journal_dirty_metadata(handle,
431 				EXT4_SB(sb)->s_sbh);
432 	}
433 	return err;
434 }
435 
436 /*
437  * Open the external journal device
438  */
439 static struct block_device *ext4_blkdev_get(dev_t dev)
440 {
441 	struct block_device *bdev;
442 	char b[BDEVNAME_SIZE];
443 
444 	bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
445 	if (IS_ERR(bdev))
446 		goto fail;
447 	return bdev;
448 
449 fail:
450 	printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
451 			__bdevname(dev, b), PTR_ERR(bdev));
452 	return NULL;
453 }
454 
455 /*
456  * Release the journal device
457  */
458 static int ext4_blkdev_put(struct block_device *bdev)
459 {
460 	bd_release(bdev);
461 	return blkdev_put(bdev);
462 }
463 
464 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
465 {
466 	struct block_device *bdev;
467 	int ret = -ENODEV;
468 
469 	bdev = sbi->journal_bdev;
470 	if (bdev) {
471 		ret = ext4_blkdev_put(bdev);
472 		sbi->journal_bdev = NULL;
473 	}
474 	return ret;
475 }
476 
477 static inline struct inode *orphan_list_entry(struct list_head *l)
478 {
479 	return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
480 }
481 
482 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
483 {
484 	struct list_head *l;
485 
486 	printk(KERN_ERR "sb orphan head is %d\n",
487 	       le32_to_cpu(sbi->s_es->s_last_orphan));
488 
489 	printk(KERN_ERR "sb_info orphan list:\n");
490 	list_for_each(l, &sbi->s_orphan) {
491 		struct inode *inode = orphan_list_entry(l);
492 		printk(KERN_ERR "  "
493 		       "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
494 		       inode->i_sb->s_id, inode->i_ino, inode,
495 		       inode->i_mode, inode->i_nlink,
496 		       NEXT_ORPHAN(inode));
497 	}
498 }
499 
500 static void ext4_put_super (struct super_block * sb)
501 {
502 	struct ext4_sb_info *sbi = EXT4_SB(sb);
503 	struct ext4_super_block *es = sbi->s_es;
504 	int i;
505 
506 	ext4_mb_release(sb);
507 	ext4_ext_release(sb);
508 	ext4_xattr_put_super(sb);
509 	jbd2_journal_destroy(sbi->s_journal);
510 	if (!(sb->s_flags & MS_RDONLY)) {
511 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
512 		es->s_state = cpu_to_le16(sbi->s_mount_state);
513 		BUFFER_TRACE(sbi->s_sbh, "marking dirty");
514 		mark_buffer_dirty(sbi->s_sbh);
515 		ext4_commit_super(sb, es, 1);
516 	}
517 
518 	for (i = 0; i < sbi->s_gdb_count; i++)
519 		brelse(sbi->s_group_desc[i]);
520 	kfree(sbi->s_group_desc);
521 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
522 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
523 	percpu_counter_destroy(&sbi->s_dirs_counter);
524 	brelse(sbi->s_sbh);
525 #ifdef CONFIG_QUOTA
526 	for (i = 0; i < MAXQUOTAS; i++)
527 		kfree(sbi->s_qf_names[i]);
528 #endif
529 
530 	/* Debugging code just in case the in-memory inode orphan list
531 	 * isn't empty.  The on-disk one can be non-empty if we've
532 	 * detected an error and taken the fs readonly, but the
533 	 * in-memory list had better be clean by this point. */
534 	if (!list_empty(&sbi->s_orphan))
535 		dump_orphan_list(sb, sbi);
536 	J_ASSERT(list_empty(&sbi->s_orphan));
537 
538 	invalidate_bdev(sb->s_bdev);
539 	if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
540 		/*
541 		 * Invalidate the journal device's buffers.  We don't want them
542 		 * floating about in memory - the physical journal device may
543 		 * hotswapped, and it breaks the `ro-after' testing code.
544 		 */
545 		sync_blockdev(sbi->journal_bdev);
546 		invalidate_bdev(sbi->journal_bdev);
547 		ext4_blkdev_remove(sbi);
548 	}
549 	sb->s_fs_info = NULL;
550 	kfree(sbi);
551 	return;
552 }
553 
554 static struct kmem_cache *ext4_inode_cachep;
555 
556 /*
557  * Called inside transaction, so use GFP_NOFS
558  */
559 static struct inode *ext4_alloc_inode(struct super_block *sb)
560 {
561 	struct ext4_inode_info *ei;
562 
563 	ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
564 	if (!ei)
565 		return NULL;
566 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
567 	ei->i_acl = EXT4_ACL_NOT_CACHED;
568 	ei->i_default_acl = EXT4_ACL_NOT_CACHED;
569 #endif
570 	ei->i_block_alloc_info = NULL;
571 	ei->vfs_inode.i_version = 1;
572 	memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
573 	INIT_LIST_HEAD(&ei->i_prealloc_list);
574 	spin_lock_init(&ei->i_prealloc_lock);
575 	return &ei->vfs_inode;
576 }
577 
578 static void ext4_destroy_inode(struct inode *inode)
579 {
580 	if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
581 		printk("EXT4 Inode %p: orphan list check failed!\n",
582 			EXT4_I(inode));
583 		print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
584 				EXT4_I(inode), sizeof(struct ext4_inode_info),
585 				true);
586 		dump_stack();
587 	}
588 	kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
589 }
590 
591 static void init_once(struct kmem_cache *cachep, void *foo)
592 {
593 	struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
594 
595 	INIT_LIST_HEAD(&ei->i_orphan);
596 #ifdef CONFIG_EXT4DEV_FS_XATTR
597 	init_rwsem(&ei->xattr_sem);
598 #endif
599 	init_rwsem(&ei->i_data_sem);
600 	inode_init_once(&ei->vfs_inode);
601 }
602 
603 static int init_inodecache(void)
604 {
605 	ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
606 					     sizeof(struct ext4_inode_info),
607 					     0, (SLAB_RECLAIM_ACCOUNT|
608 						SLAB_MEM_SPREAD),
609 					     init_once);
610 	if (ext4_inode_cachep == NULL)
611 		return -ENOMEM;
612 	return 0;
613 }
614 
615 static void destroy_inodecache(void)
616 {
617 	kmem_cache_destroy(ext4_inode_cachep);
618 }
619 
620 static void ext4_clear_inode(struct inode *inode)
621 {
622 	struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info;
623 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
624 	if (EXT4_I(inode)->i_acl &&
625 			EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
626 		posix_acl_release(EXT4_I(inode)->i_acl);
627 		EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
628 	}
629 	if (EXT4_I(inode)->i_default_acl &&
630 			EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
631 		posix_acl_release(EXT4_I(inode)->i_default_acl);
632 		EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
633 	}
634 #endif
635 	ext4_discard_reservation(inode);
636 	EXT4_I(inode)->i_block_alloc_info = NULL;
637 	if (unlikely(rsv))
638 		kfree(rsv);
639 }
640 
641 static inline void ext4_show_quota_options(struct seq_file *seq, struct super_block *sb)
642 {
643 #if defined(CONFIG_QUOTA)
644 	struct ext4_sb_info *sbi = EXT4_SB(sb);
645 
646 	if (sbi->s_jquota_fmt)
647 		seq_printf(seq, ",jqfmt=%s",
648 		(sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
649 
650 	if (sbi->s_qf_names[USRQUOTA])
651 		seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
652 
653 	if (sbi->s_qf_names[GRPQUOTA])
654 		seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
655 
656 	if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
657 		seq_puts(seq, ",usrquota");
658 
659 	if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
660 		seq_puts(seq, ",grpquota");
661 #endif
662 }
663 
664 /*
665  * Show an option if
666  *  - it's set to a non-default value OR
667  *  - if the per-sb default is different from the global default
668  */
669 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
670 {
671 	int def_errors;
672 	unsigned long def_mount_opts;
673 	struct super_block *sb = vfs->mnt_sb;
674 	struct ext4_sb_info *sbi = EXT4_SB(sb);
675 	struct ext4_super_block *es = sbi->s_es;
676 
677 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
678 	def_errors     = le16_to_cpu(es->s_errors);
679 
680 	if (sbi->s_sb_block != 1)
681 		seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
682 	if (test_opt(sb, MINIX_DF))
683 		seq_puts(seq, ",minixdf");
684 	if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
685 		seq_puts(seq, ",grpid");
686 	if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
687 		seq_puts(seq, ",nogrpid");
688 	if (sbi->s_resuid != EXT4_DEF_RESUID ||
689 	    le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
690 		seq_printf(seq, ",resuid=%u", sbi->s_resuid);
691 	}
692 	if (sbi->s_resgid != EXT4_DEF_RESGID ||
693 	    le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
694 		seq_printf(seq, ",resgid=%u", sbi->s_resgid);
695 	}
696 	if (test_opt(sb, ERRORS_RO)) {
697 		if (def_errors == EXT4_ERRORS_PANIC ||
698 		    def_errors == EXT4_ERRORS_CONTINUE) {
699 			seq_puts(seq, ",errors=remount-ro");
700 		}
701 	}
702 	if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
703 		seq_puts(seq, ",errors=continue");
704 	if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
705 		seq_puts(seq, ",errors=panic");
706 	if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
707 		seq_puts(seq, ",nouid32");
708 	if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
709 		seq_puts(seq, ",debug");
710 	if (test_opt(sb, OLDALLOC))
711 		seq_puts(seq, ",oldalloc");
712 #ifdef CONFIG_EXT4DEV_FS_XATTR
713 	if (test_opt(sb, XATTR_USER) &&
714 		!(def_mount_opts & EXT4_DEFM_XATTR_USER))
715 		seq_puts(seq, ",user_xattr");
716 	if (!test_opt(sb, XATTR_USER) &&
717 	    (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
718 		seq_puts(seq, ",nouser_xattr");
719 	}
720 #endif
721 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
722 	if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
723 		seq_puts(seq, ",acl");
724 	if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
725 		seq_puts(seq, ",noacl");
726 #endif
727 	if (!test_opt(sb, RESERVATION))
728 		seq_puts(seq, ",noreservation");
729 	if (sbi->s_commit_interval) {
730 		seq_printf(seq, ",commit=%u",
731 			   (unsigned) (sbi->s_commit_interval / HZ));
732 	}
733 	if (test_opt(sb, BARRIER))
734 		seq_puts(seq, ",barrier=1");
735 	if (test_opt(sb, NOBH))
736 		seq_puts(seq, ",nobh");
737 	if (!test_opt(sb, EXTENTS))
738 		seq_puts(seq, ",noextents");
739 	if (!test_opt(sb, MBALLOC))
740 		seq_puts(seq, ",nomballoc");
741 	if (test_opt(sb, I_VERSION))
742 		seq_puts(seq, ",i_version");
743 
744 	if (sbi->s_stripe)
745 		seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
746 	/*
747 	 * journal mode get enabled in different ways
748 	 * So just print the value even if we didn't specify it
749 	 */
750 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
751 		seq_puts(seq, ",data=journal");
752 	else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
753 		seq_puts(seq, ",data=ordered");
754 	else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
755 		seq_puts(seq, ",data=writeback");
756 
757 	ext4_show_quota_options(seq, sb);
758 	return 0;
759 }
760 
761 
762 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
763 		u64 ino, u32 generation)
764 {
765 	struct inode *inode;
766 
767 	if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
768 		return ERR_PTR(-ESTALE);
769 	if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
770 		return ERR_PTR(-ESTALE);
771 
772 	/* iget isn't really right if the inode is currently unallocated!!
773 	 *
774 	 * ext4_read_inode will return a bad_inode if the inode had been
775 	 * deleted, so we should be safe.
776 	 *
777 	 * Currently we don't know the generation for parent directory, so
778 	 * a generation of 0 means "accept any"
779 	 */
780 	inode = ext4_iget(sb, ino);
781 	if (IS_ERR(inode))
782 		return ERR_CAST(inode);
783 	if (generation && inode->i_generation != generation) {
784 		iput(inode);
785 		return ERR_PTR(-ESTALE);
786 	}
787 
788 	return inode;
789 }
790 
791 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
792 		int fh_len, int fh_type)
793 {
794 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
795 				    ext4_nfs_get_inode);
796 }
797 
798 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
799 		int fh_len, int fh_type)
800 {
801 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
802 				    ext4_nfs_get_inode);
803 }
804 
805 #ifdef CONFIG_QUOTA
806 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
807 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
808 
809 static int ext4_dquot_initialize(struct inode *inode, int type);
810 static int ext4_dquot_drop(struct inode *inode);
811 static int ext4_write_dquot(struct dquot *dquot);
812 static int ext4_acquire_dquot(struct dquot *dquot);
813 static int ext4_release_dquot(struct dquot *dquot);
814 static int ext4_mark_dquot_dirty(struct dquot *dquot);
815 static int ext4_write_info(struct super_block *sb, int type);
816 static int ext4_quota_on(struct super_block *sb, int type, int format_id, char *path);
817 static int ext4_quota_on_mount(struct super_block *sb, int type);
818 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
819 			       size_t len, loff_t off);
820 static ssize_t ext4_quota_write(struct super_block *sb, int type,
821 				const char *data, size_t len, loff_t off);
822 
823 static struct dquot_operations ext4_quota_operations = {
824 	.initialize	= ext4_dquot_initialize,
825 	.drop		= ext4_dquot_drop,
826 	.alloc_space	= dquot_alloc_space,
827 	.alloc_inode	= dquot_alloc_inode,
828 	.free_space	= dquot_free_space,
829 	.free_inode	= dquot_free_inode,
830 	.transfer	= dquot_transfer,
831 	.write_dquot	= ext4_write_dquot,
832 	.acquire_dquot	= ext4_acquire_dquot,
833 	.release_dquot	= ext4_release_dquot,
834 	.mark_dirty	= ext4_mark_dquot_dirty,
835 	.write_info	= ext4_write_info
836 };
837 
838 static struct quotactl_ops ext4_qctl_operations = {
839 	.quota_on	= ext4_quota_on,
840 	.quota_off	= vfs_quota_off,
841 	.quota_sync	= vfs_quota_sync,
842 	.get_info	= vfs_get_dqinfo,
843 	.set_info	= vfs_set_dqinfo,
844 	.get_dqblk	= vfs_get_dqblk,
845 	.set_dqblk	= vfs_set_dqblk
846 };
847 #endif
848 
849 static const struct super_operations ext4_sops = {
850 	.alloc_inode	= ext4_alloc_inode,
851 	.destroy_inode	= ext4_destroy_inode,
852 	.write_inode	= ext4_write_inode,
853 	.dirty_inode	= ext4_dirty_inode,
854 	.delete_inode	= ext4_delete_inode,
855 	.put_super	= ext4_put_super,
856 	.write_super	= ext4_write_super,
857 	.sync_fs	= ext4_sync_fs,
858 	.write_super_lockfs = ext4_write_super_lockfs,
859 	.unlockfs	= ext4_unlockfs,
860 	.statfs		= ext4_statfs,
861 	.remount_fs	= ext4_remount,
862 	.clear_inode	= ext4_clear_inode,
863 	.show_options	= ext4_show_options,
864 #ifdef CONFIG_QUOTA
865 	.quota_read	= ext4_quota_read,
866 	.quota_write	= ext4_quota_write,
867 #endif
868 };
869 
870 static const struct export_operations ext4_export_ops = {
871 	.fh_to_dentry = ext4_fh_to_dentry,
872 	.fh_to_parent = ext4_fh_to_parent,
873 	.get_parent = ext4_get_parent,
874 };
875 
876 enum {
877 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
878 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
879 	Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
880 	Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
881 	Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
882 	Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
883 	Opt_journal_checksum, Opt_journal_async_commit,
884 	Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
885 	Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
886 	Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
887 	Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
888 	Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
889 	Opt_mballoc, Opt_nomballoc, Opt_stripe,
890 };
891 
892 static match_table_t tokens = {
893 	{Opt_bsd_df, "bsddf"},
894 	{Opt_minix_df, "minixdf"},
895 	{Opt_grpid, "grpid"},
896 	{Opt_grpid, "bsdgroups"},
897 	{Opt_nogrpid, "nogrpid"},
898 	{Opt_nogrpid, "sysvgroups"},
899 	{Opt_resgid, "resgid=%u"},
900 	{Opt_resuid, "resuid=%u"},
901 	{Opt_sb, "sb=%u"},
902 	{Opt_err_cont, "errors=continue"},
903 	{Opt_err_panic, "errors=panic"},
904 	{Opt_err_ro, "errors=remount-ro"},
905 	{Opt_nouid32, "nouid32"},
906 	{Opt_nocheck, "nocheck"},
907 	{Opt_nocheck, "check=none"},
908 	{Opt_debug, "debug"},
909 	{Opt_oldalloc, "oldalloc"},
910 	{Opt_orlov, "orlov"},
911 	{Opt_user_xattr, "user_xattr"},
912 	{Opt_nouser_xattr, "nouser_xattr"},
913 	{Opt_acl, "acl"},
914 	{Opt_noacl, "noacl"},
915 	{Opt_reservation, "reservation"},
916 	{Opt_noreservation, "noreservation"},
917 	{Opt_noload, "noload"},
918 	{Opt_nobh, "nobh"},
919 	{Opt_bh, "bh"},
920 	{Opt_commit, "commit=%u"},
921 	{Opt_journal_update, "journal=update"},
922 	{Opt_journal_inum, "journal=%u"},
923 	{Opt_journal_dev, "journal_dev=%u"},
924 	{Opt_journal_checksum, "journal_checksum"},
925 	{Opt_journal_async_commit, "journal_async_commit"},
926 	{Opt_abort, "abort"},
927 	{Opt_data_journal, "data=journal"},
928 	{Opt_data_ordered, "data=ordered"},
929 	{Opt_data_writeback, "data=writeback"},
930 	{Opt_offusrjquota, "usrjquota="},
931 	{Opt_usrjquota, "usrjquota=%s"},
932 	{Opt_offgrpjquota, "grpjquota="},
933 	{Opt_grpjquota, "grpjquota=%s"},
934 	{Opt_jqfmt_vfsold, "jqfmt=vfsold"},
935 	{Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
936 	{Opt_grpquota, "grpquota"},
937 	{Opt_noquota, "noquota"},
938 	{Opt_quota, "quota"},
939 	{Opt_usrquota, "usrquota"},
940 	{Opt_barrier, "barrier=%u"},
941 	{Opt_extents, "extents"},
942 	{Opt_noextents, "noextents"},
943 	{Opt_i_version, "i_version"},
944 	{Opt_mballoc, "mballoc"},
945 	{Opt_nomballoc, "nomballoc"},
946 	{Opt_stripe, "stripe=%u"},
947 	{Opt_err, NULL},
948 	{Opt_resize, "resize"},
949 };
950 
951 static ext4_fsblk_t get_sb_block(void **data)
952 {
953 	ext4_fsblk_t	sb_block;
954 	char		*options = (char *) *data;
955 
956 	if (!options || strncmp(options, "sb=", 3) != 0)
957 		return 1;	/* Default location */
958 	options += 3;
959 	/*todo: use simple_strtoll with >32bit ext4 */
960 	sb_block = simple_strtoul(options, &options, 0);
961 	if (*options && *options != ',') {
962 		printk("EXT4-fs: Invalid sb specification: %s\n",
963 		       (char *) *data);
964 		return 1;
965 	}
966 	if (*options == ',')
967 		options++;
968 	*data = (void *) options;
969 	return sb_block;
970 }
971 
972 static int parse_options (char *options, struct super_block *sb,
973 			  unsigned int *inum, unsigned long *journal_devnum,
974 			  ext4_fsblk_t *n_blocks_count, int is_remount)
975 {
976 	struct ext4_sb_info *sbi = EXT4_SB(sb);
977 	char * p;
978 	substring_t args[MAX_OPT_ARGS];
979 	int data_opt = 0;
980 	int option;
981 #ifdef CONFIG_QUOTA
982 	int qtype;
983 	char *qname;
984 #endif
985 
986 	if (!options)
987 		return 1;
988 
989 	while ((p = strsep (&options, ",")) != NULL) {
990 		int token;
991 		if (!*p)
992 			continue;
993 
994 		token = match_token(p, tokens, args);
995 		switch (token) {
996 		case Opt_bsd_df:
997 			clear_opt (sbi->s_mount_opt, MINIX_DF);
998 			break;
999 		case Opt_minix_df:
1000 			set_opt (sbi->s_mount_opt, MINIX_DF);
1001 			break;
1002 		case Opt_grpid:
1003 			set_opt (sbi->s_mount_opt, GRPID);
1004 			break;
1005 		case Opt_nogrpid:
1006 			clear_opt (sbi->s_mount_opt, GRPID);
1007 			break;
1008 		case Opt_resuid:
1009 			if (match_int(&args[0], &option))
1010 				return 0;
1011 			sbi->s_resuid = option;
1012 			break;
1013 		case Opt_resgid:
1014 			if (match_int(&args[0], &option))
1015 				return 0;
1016 			sbi->s_resgid = option;
1017 			break;
1018 		case Opt_sb:
1019 			/* handled by get_sb_block() instead of here */
1020 			/* *sb_block = match_int(&args[0]); */
1021 			break;
1022 		case Opt_err_panic:
1023 			clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1024 			clear_opt (sbi->s_mount_opt, ERRORS_RO);
1025 			set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1026 			break;
1027 		case Opt_err_ro:
1028 			clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1029 			clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1030 			set_opt (sbi->s_mount_opt, ERRORS_RO);
1031 			break;
1032 		case Opt_err_cont:
1033 			clear_opt (sbi->s_mount_opt, ERRORS_RO);
1034 			clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1035 			set_opt (sbi->s_mount_opt, ERRORS_CONT);
1036 			break;
1037 		case Opt_nouid32:
1038 			set_opt (sbi->s_mount_opt, NO_UID32);
1039 			break;
1040 		case Opt_nocheck:
1041 			clear_opt (sbi->s_mount_opt, CHECK);
1042 			break;
1043 		case Opt_debug:
1044 			set_opt (sbi->s_mount_opt, DEBUG);
1045 			break;
1046 		case Opt_oldalloc:
1047 			set_opt (sbi->s_mount_opt, OLDALLOC);
1048 			break;
1049 		case Opt_orlov:
1050 			clear_opt (sbi->s_mount_opt, OLDALLOC);
1051 			break;
1052 #ifdef CONFIG_EXT4DEV_FS_XATTR
1053 		case Opt_user_xattr:
1054 			set_opt (sbi->s_mount_opt, XATTR_USER);
1055 			break;
1056 		case Opt_nouser_xattr:
1057 			clear_opt (sbi->s_mount_opt, XATTR_USER);
1058 			break;
1059 #else
1060 		case Opt_user_xattr:
1061 		case Opt_nouser_xattr:
1062 			printk("EXT4 (no)user_xattr options not supported\n");
1063 			break;
1064 #endif
1065 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1066 		case Opt_acl:
1067 			set_opt(sbi->s_mount_opt, POSIX_ACL);
1068 			break;
1069 		case Opt_noacl:
1070 			clear_opt(sbi->s_mount_opt, POSIX_ACL);
1071 			break;
1072 #else
1073 		case Opt_acl:
1074 		case Opt_noacl:
1075 			printk("EXT4 (no)acl options not supported\n");
1076 			break;
1077 #endif
1078 		case Opt_reservation:
1079 			set_opt(sbi->s_mount_opt, RESERVATION);
1080 			break;
1081 		case Opt_noreservation:
1082 			clear_opt(sbi->s_mount_opt, RESERVATION);
1083 			break;
1084 		case Opt_journal_update:
1085 			/* @@@ FIXME */
1086 			/* Eventually we will want to be able to create
1087 			   a journal file here.  For now, only allow the
1088 			   user to specify an existing inode to be the
1089 			   journal file. */
1090 			if (is_remount) {
1091 				printk(KERN_ERR "EXT4-fs: cannot specify "
1092 				       "journal on remount\n");
1093 				return 0;
1094 			}
1095 			set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1096 			break;
1097 		case Opt_journal_inum:
1098 			if (is_remount) {
1099 				printk(KERN_ERR "EXT4-fs: cannot specify "
1100 				       "journal on remount\n");
1101 				return 0;
1102 			}
1103 			if (match_int(&args[0], &option))
1104 				return 0;
1105 			*inum = option;
1106 			break;
1107 		case Opt_journal_dev:
1108 			if (is_remount) {
1109 				printk(KERN_ERR "EXT4-fs: cannot specify "
1110 				       "journal on remount\n");
1111 				return 0;
1112 			}
1113 			if (match_int(&args[0], &option))
1114 				return 0;
1115 			*journal_devnum = option;
1116 			break;
1117 		case Opt_journal_checksum:
1118 			set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1119 			break;
1120 		case Opt_journal_async_commit:
1121 			set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1122 			set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1123 			break;
1124 		case Opt_noload:
1125 			set_opt (sbi->s_mount_opt, NOLOAD);
1126 			break;
1127 		case Opt_commit:
1128 			if (match_int(&args[0], &option))
1129 				return 0;
1130 			if (option < 0)
1131 				return 0;
1132 			if (option == 0)
1133 				option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1134 			sbi->s_commit_interval = HZ * option;
1135 			break;
1136 		case Opt_data_journal:
1137 			data_opt = EXT4_MOUNT_JOURNAL_DATA;
1138 			goto datacheck;
1139 		case Opt_data_ordered:
1140 			data_opt = EXT4_MOUNT_ORDERED_DATA;
1141 			goto datacheck;
1142 		case Opt_data_writeback:
1143 			data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1144 		datacheck:
1145 			if (is_remount) {
1146 				if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1147 						!= data_opt) {
1148 					printk(KERN_ERR
1149 						"EXT4-fs: cannot change data "
1150 						"mode on remount\n");
1151 					return 0;
1152 				}
1153 			} else {
1154 				sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1155 				sbi->s_mount_opt |= data_opt;
1156 			}
1157 			break;
1158 #ifdef CONFIG_QUOTA
1159 		case Opt_usrjquota:
1160 			qtype = USRQUOTA;
1161 			goto set_qf_name;
1162 		case Opt_grpjquota:
1163 			qtype = GRPQUOTA;
1164 set_qf_name:
1165 			if (sb_any_quota_enabled(sb)) {
1166 				printk(KERN_ERR
1167 					"EXT4-fs: Cannot change journalled "
1168 					"quota options when quota turned on.\n");
1169 				return 0;
1170 			}
1171 			qname = match_strdup(&args[0]);
1172 			if (!qname) {
1173 				printk(KERN_ERR
1174 					"EXT4-fs: not enough memory for "
1175 					"storing quotafile name.\n");
1176 				return 0;
1177 			}
1178 			if (sbi->s_qf_names[qtype] &&
1179 			    strcmp(sbi->s_qf_names[qtype], qname)) {
1180 				printk(KERN_ERR
1181 					"EXT4-fs: %s quota file already "
1182 					"specified.\n", QTYPE2NAME(qtype));
1183 				kfree(qname);
1184 				return 0;
1185 			}
1186 			sbi->s_qf_names[qtype] = qname;
1187 			if (strchr(sbi->s_qf_names[qtype], '/')) {
1188 				printk(KERN_ERR
1189 					"EXT4-fs: quotafile must be on "
1190 					"filesystem root.\n");
1191 				kfree(sbi->s_qf_names[qtype]);
1192 				sbi->s_qf_names[qtype] = NULL;
1193 				return 0;
1194 			}
1195 			set_opt(sbi->s_mount_opt, QUOTA);
1196 			break;
1197 		case Opt_offusrjquota:
1198 			qtype = USRQUOTA;
1199 			goto clear_qf_name;
1200 		case Opt_offgrpjquota:
1201 			qtype = GRPQUOTA;
1202 clear_qf_name:
1203 			if (sb_any_quota_enabled(sb)) {
1204 				printk(KERN_ERR "EXT4-fs: Cannot change "
1205 					"journalled quota options when "
1206 					"quota turned on.\n");
1207 				return 0;
1208 			}
1209 			/*
1210 			 * The space will be released later when all options
1211 			 * are confirmed to be correct
1212 			 */
1213 			sbi->s_qf_names[qtype] = NULL;
1214 			break;
1215 		case Opt_jqfmt_vfsold:
1216 			sbi->s_jquota_fmt = QFMT_VFS_OLD;
1217 			break;
1218 		case Opt_jqfmt_vfsv0:
1219 			sbi->s_jquota_fmt = QFMT_VFS_V0;
1220 			break;
1221 		case Opt_quota:
1222 		case Opt_usrquota:
1223 			set_opt(sbi->s_mount_opt, QUOTA);
1224 			set_opt(sbi->s_mount_opt, USRQUOTA);
1225 			break;
1226 		case Opt_grpquota:
1227 			set_opt(sbi->s_mount_opt, QUOTA);
1228 			set_opt(sbi->s_mount_opt, GRPQUOTA);
1229 			break;
1230 		case Opt_noquota:
1231 			if (sb_any_quota_enabled(sb)) {
1232 				printk(KERN_ERR "EXT4-fs: Cannot change quota "
1233 					"options when quota turned on.\n");
1234 				return 0;
1235 			}
1236 			clear_opt(sbi->s_mount_opt, QUOTA);
1237 			clear_opt(sbi->s_mount_opt, USRQUOTA);
1238 			clear_opt(sbi->s_mount_opt, GRPQUOTA);
1239 			break;
1240 #else
1241 		case Opt_quota:
1242 		case Opt_usrquota:
1243 		case Opt_grpquota:
1244 		case Opt_usrjquota:
1245 		case Opt_grpjquota:
1246 		case Opt_offusrjquota:
1247 		case Opt_offgrpjquota:
1248 		case Opt_jqfmt_vfsold:
1249 		case Opt_jqfmt_vfsv0:
1250 			printk(KERN_ERR
1251 				"EXT4-fs: journalled quota options not "
1252 				"supported.\n");
1253 			break;
1254 		case Opt_noquota:
1255 			break;
1256 #endif
1257 		case Opt_abort:
1258 			set_opt(sbi->s_mount_opt, ABORT);
1259 			break;
1260 		case Opt_barrier:
1261 			if (match_int(&args[0], &option))
1262 				return 0;
1263 			if (option)
1264 				set_opt(sbi->s_mount_opt, BARRIER);
1265 			else
1266 				clear_opt(sbi->s_mount_opt, BARRIER);
1267 			break;
1268 		case Opt_ignore:
1269 			break;
1270 		case Opt_resize:
1271 			if (!is_remount) {
1272 				printk("EXT4-fs: resize option only available "
1273 					"for remount\n");
1274 				return 0;
1275 			}
1276 			if (match_int(&args[0], &option) != 0)
1277 				return 0;
1278 			*n_blocks_count = option;
1279 			break;
1280 		case Opt_nobh:
1281 			set_opt(sbi->s_mount_opt, NOBH);
1282 			break;
1283 		case Opt_bh:
1284 			clear_opt(sbi->s_mount_opt, NOBH);
1285 			break;
1286 		case Opt_extents:
1287 			set_opt (sbi->s_mount_opt, EXTENTS);
1288 			break;
1289 		case Opt_noextents:
1290 			clear_opt (sbi->s_mount_opt, EXTENTS);
1291 			break;
1292 		case Opt_i_version:
1293 			set_opt(sbi->s_mount_opt, I_VERSION);
1294 			sb->s_flags |= MS_I_VERSION;
1295 			break;
1296 		case Opt_mballoc:
1297 			set_opt(sbi->s_mount_opt, MBALLOC);
1298 			break;
1299 		case Opt_nomballoc:
1300 			clear_opt(sbi->s_mount_opt, MBALLOC);
1301 			break;
1302 		case Opt_stripe:
1303 			if (match_int(&args[0], &option))
1304 				return 0;
1305 			if (option < 0)
1306 				return 0;
1307 			sbi->s_stripe = option;
1308 			break;
1309 		default:
1310 			printk (KERN_ERR
1311 				"EXT4-fs: Unrecognized mount option \"%s\" "
1312 				"or missing value\n", p);
1313 			return 0;
1314 		}
1315 	}
1316 #ifdef CONFIG_QUOTA
1317 	if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1318 		if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1319 		     sbi->s_qf_names[USRQUOTA])
1320 			clear_opt(sbi->s_mount_opt, USRQUOTA);
1321 
1322 		if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1323 		     sbi->s_qf_names[GRPQUOTA])
1324 			clear_opt(sbi->s_mount_opt, GRPQUOTA);
1325 
1326 		if ((sbi->s_qf_names[USRQUOTA] &&
1327 				(sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1328 		    (sbi->s_qf_names[GRPQUOTA] &&
1329 				(sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1330 			printk(KERN_ERR "EXT4-fs: old and new quota "
1331 					"format mixing.\n");
1332 			return 0;
1333 		}
1334 
1335 		if (!sbi->s_jquota_fmt) {
1336 			printk(KERN_ERR "EXT4-fs: journalled quota format "
1337 					"not specified.\n");
1338 			return 0;
1339 		}
1340 	} else {
1341 		if (sbi->s_jquota_fmt) {
1342 			printk(KERN_ERR "EXT4-fs: journalled quota format "
1343 					"specified with no journalling "
1344 					"enabled.\n");
1345 			return 0;
1346 		}
1347 	}
1348 #endif
1349 	return 1;
1350 }
1351 
1352 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1353 			    int read_only)
1354 {
1355 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1356 	int res = 0;
1357 
1358 	if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1359 		printk (KERN_ERR "EXT4-fs warning: revision level too high, "
1360 			"forcing read-only mode\n");
1361 		res = MS_RDONLY;
1362 	}
1363 	if (read_only)
1364 		return res;
1365 	if (!(sbi->s_mount_state & EXT4_VALID_FS))
1366 		printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1367 			"running e2fsck is recommended\n");
1368 	else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1369 		printk (KERN_WARNING
1370 			"EXT4-fs warning: mounting fs with errors, "
1371 			"running e2fsck is recommended\n");
1372 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1373 		 le16_to_cpu(es->s_mnt_count) >=
1374 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1375 		printk (KERN_WARNING
1376 			"EXT4-fs warning: maximal mount count reached, "
1377 			"running e2fsck is recommended\n");
1378 	else if (le32_to_cpu(es->s_checkinterval) &&
1379 		(le32_to_cpu(es->s_lastcheck) +
1380 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1381 		printk (KERN_WARNING
1382 			"EXT4-fs warning: checktime reached, "
1383 			"running e2fsck is recommended\n");
1384 #if 0
1385 		/* @@@ We _will_ want to clear the valid bit if we find
1386 		 * inconsistencies, to force a fsck at reboot.  But for
1387 		 * a plain journaled filesystem we can keep it set as
1388 		 * valid forever! :)
1389 		 */
1390 	es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT4_VALID_FS);
1391 #endif
1392 	if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1393 		es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1394 	es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
1395 	es->s_mtime = cpu_to_le32(get_seconds());
1396 	ext4_update_dynamic_rev(sb);
1397 	EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1398 
1399 	ext4_commit_super(sb, es, 1);
1400 	if (test_opt(sb, DEBUG))
1401 		printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
1402 				"bpg=%lu, ipg=%lu, mo=%04lx]\n",
1403 			sb->s_blocksize,
1404 			sbi->s_groups_count,
1405 			EXT4_BLOCKS_PER_GROUP(sb),
1406 			EXT4_INODES_PER_GROUP(sb),
1407 			sbi->s_mount_opt);
1408 
1409 	printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1410 	if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
1411 		char b[BDEVNAME_SIZE];
1412 
1413 		printk("external journal on %s\n",
1414 			bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
1415 	} else {
1416 		printk("internal journal\n");
1417 	}
1418 	return res;
1419 }
1420 
1421 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1422 			    struct ext4_group_desc *gdp)
1423 {
1424 	__u16 crc = 0;
1425 
1426 	if (sbi->s_es->s_feature_ro_compat &
1427 	    cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1428 		int offset = offsetof(struct ext4_group_desc, bg_checksum);
1429 		__le32 le_group = cpu_to_le32(block_group);
1430 
1431 		crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1432 		crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1433 		crc = crc16(crc, (__u8 *)gdp, offset);
1434 		offset += sizeof(gdp->bg_checksum); /* skip checksum */
1435 		/* for checksum of struct ext4_group_desc do the rest...*/
1436 		if ((sbi->s_es->s_feature_incompat &
1437 		     cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1438 		    offset < le16_to_cpu(sbi->s_es->s_desc_size))
1439 			crc = crc16(crc, (__u8 *)gdp + offset,
1440 				    le16_to_cpu(sbi->s_es->s_desc_size) -
1441 					offset);
1442 	}
1443 
1444 	return cpu_to_le16(crc);
1445 }
1446 
1447 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1448 				struct ext4_group_desc *gdp)
1449 {
1450 	if ((sbi->s_es->s_feature_ro_compat &
1451 	     cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1452 	    (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1453 		return 0;
1454 
1455 	return 1;
1456 }
1457 
1458 /* Called at mount-time, super-block is locked */
1459 static int ext4_check_descriptors(struct super_block *sb)
1460 {
1461 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1462 	ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1463 	ext4_fsblk_t last_block;
1464 	ext4_fsblk_t block_bitmap;
1465 	ext4_fsblk_t inode_bitmap;
1466 	ext4_fsblk_t inode_table;
1467 	int flexbg_flag = 0;
1468 	ext4_group_t i;
1469 
1470 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1471 		flexbg_flag = 1;
1472 
1473 	ext4_debug ("Checking group descriptors");
1474 
1475 	for (i = 0; i < sbi->s_groups_count; i++) {
1476 		struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1477 
1478 		if (i == sbi->s_groups_count - 1 || flexbg_flag)
1479 			last_block = ext4_blocks_count(sbi->s_es) - 1;
1480 		else
1481 			last_block = first_block +
1482 				(EXT4_BLOCKS_PER_GROUP(sb) - 1);
1483 
1484 		block_bitmap = ext4_block_bitmap(sb, gdp);
1485 		if (block_bitmap < first_block || block_bitmap > last_block)
1486 		{
1487 			ext4_error (sb, "ext4_check_descriptors",
1488 				    "Block bitmap for group %lu"
1489 				    " not in group (block %llu)!",
1490 				    i, block_bitmap);
1491 			return 0;
1492 		}
1493 		inode_bitmap = ext4_inode_bitmap(sb, gdp);
1494 		if (inode_bitmap < first_block || inode_bitmap > last_block)
1495 		{
1496 			ext4_error (sb, "ext4_check_descriptors",
1497 				    "Inode bitmap for group %lu"
1498 				    " not in group (block %llu)!",
1499 				    i, inode_bitmap);
1500 			return 0;
1501 		}
1502 		inode_table = ext4_inode_table(sb, gdp);
1503 		if (inode_table < first_block ||
1504 		    inode_table + sbi->s_itb_per_group - 1 > last_block)
1505 		{
1506 			ext4_error (sb, "ext4_check_descriptors",
1507 				    "Inode table for group %lu"
1508 				    " not in group (block %llu)!",
1509 				    i, inode_table);
1510 			return 0;
1511 		}
1512 		if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1513 			ext4_error(sb, __FUNCTION__,
1514 				   "Checksum for group %lu failed (%u!=%u)\n",
1515 				    i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1516 				    gdp)), le16_to_cpu(gdp->bg_checksum));
1517 			return 0;
1518 		}
1519 		if (!flexbg_flag)
1520 			first_block += EXT4_BLOCKS_PER_GROUP(sb);
1521 	}
1522 
1523 	ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1524 	sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb));
1525 	return 1;
1526 }
1527 
1528 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1529  * the superblock) which were deleted from all directories, but held open by
1530  * a process at the time of a crash.  We walk the list and try to delete these
1531  * inodes at recovery time (only with a read-write filesystem).
1532  *
1533  * In order to keep the orphan inode chain consistent during traversal (in
1534  * case of crash during recovery), we link each inode into the superblock
1535  * orphan list_head and handle it the same way as an inode deletion during
1536  * normal operation (which journals the operations for us).
1537  *
1538  * We only do an iget() and an iput() on each inode, which is very safe if we
1539  * accidentally point at an in-use or already deleted inode.  The worst that
1540  * can happen in this case is that we get a "bit already cleared" message from
1541  * ext4_free_inode().  The only reason we would point at a wrong inode is if
1542  * e2fsck was run on this filesystem, and it must have already done the orphan
1543  * inode cleanup for us, so we can safely abort without any further action.
1544  */
1545 static void ext4_orphan_cleanup (struct super_block * sb,
1546 				 struct ext4_super_block * es)
1547 {
1548 	unsigned int s_flags = sb->s_flags;
1549 	int nr_orphans = 0, nr_truncates = 0;
1550 #ifdef CONFIG_QUOTA
1551 	int i;
1552 #endif
1553 	if (!es->s_last_orphan) {
1554 		jbd_debug(4, "no orphan inodes to clean up\n");
1555 		return;
1556 	}
1557 
1558 	if (bdev_read_only(sb->s_bdev)) {
1559 		printk(KERN_ERR "EXT4-fs: write access "
1560 			"unavailable, skipping orphan cleanup.\n");
1561 		return;
1562 	}
1563 
1564 	if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1565 		if (es->s_last_orphan)
1566 			jbd_debug(1, "Errors on filesystem, "
1567 				  "clearing orphan list.\n");
1568 		es->s_last_orphan = 0;
1569 		jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1570 		return;
1571 	}
1572 
1573 	if (s_flags & MS_RDONLY) {
1574 		printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1575 		       sb->s_id);
1576 		sb->s_flags &= ~MS_RDONLY;
1577 	}
1578 #ifdef CONFIG_QUOTA
1579 	/* Needed for iput() to work correctly and not trash data */
1580 	sb->s_flags |= MS_ACTIVE;
1581 	/* Turn on quotas so that they are updated correctly */
1582 	for (i = 0; i < MAXQUOTAS; i++) {
1583 		if (EXT4_SB(sb)->s_qf_names[i]) {
1584 			int ret = ext4_quota_on_mount(sb, i);
1585 			if (ret < 0)
1586 				printk(KERN_ERR
1587 					"EXT4-fs: Cannot turn on journalled "
1588 					"quota: error %d\n", ret);
1589 		}
1590 	}
1591 #endif
1592 
1593 	while (es->s_last_orphan) {
1594 		struct inode *inode;
1595 
1596 		if (!(inode =
1597 		      ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
1598 			es->s_last_orphan = 0;
1599 			break;
1600 		}
1601 
1602 		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1603 		DQUOT_INIT(inode);
1604 		if (inode->i_nlink) {
1605 			printk(KERN_DEBUG
1606 				"%s: truncating inode %lu to %Ld bytes\n",
1607 				__FUNCTION__, inode->i_ino, inode->i_size);
1608 			jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1609 				  inode->i_ino, inode->i_size);
1610 			ext4_truncate(inode);
1611 			nr_truncates++;
1612 		} else {
1613 			printk(KERN_DEBUG
1614 				"%s: deleting unreferenced inode %lu\n",
1615 				__FUNCTION__, inode->i_ino);
1616 			jbd_debug(2, "deleting unreferenced inode %lu\n",
1617 				  inode->i_ino);
1618 			nr_orphans++;
1619 		}
1620 		iput(inode);  /* The delete magic happens here! */
1621 	}
1622 
1623 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1624 
1625 	if (nr_orphans)
1626 		printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1627 		       sb->s_id, PLURAL(nr_orphans));
1628 	if (nr_truncates)
1629 		printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1630 		       sb->s_id, PLURAL(nr_truncates));
1631 #ifdef CONFIG_QUOTA
1632 	/* Turn quotas off */
1633 	for (i = 0; i < MAXQUOTAS; i++) {
1634 		if (sb_dqopt(sb)->files[i])
1635 			vfs_quota_off(sb, i);
1636 	}
1637 #endif
1638 	sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1639 }
1640 /*
1641  * Maximal extent format file size.
1642  * Resulting logical blkno at s_maxbytes must fit in our on-disk
1643  * extent format containers, within a sector_t, and within i_blocks
1644  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
1645  * so that won't be a limiting factor.
1646  *
1647  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1648  */
1649 static loff_t ext4_max_size(int blkbits)
1650 {
1651 	loff_t res;
1652 	loff_t upper_limit = MAX_LFS_FILESIZE;
1653 
1654 	/* small i_blocks in vfs inode? */
1655 	if (sizeof(blkcnt_t) < sizeof(u64)) {
1656 		/*
1657 		 * CONFIG_LSF is not enabled implies the inode
1658 		 * i_block represent total blocks in 512 bytes
1659 		 * 32 == size of vfs inode i_blocks * 8
1660 		 */
1661 		upper_limit = (1LL << 32) - 1;
1662 
1663 		/* total blocks in file system block size */
1664 		upper_limit >>= (blkbits - 9);
1665 		upper_limit <<= blkbits;
1666 	}
1667 
1668 	/* 32-bit extent-start container, ee_block */
1669 	res = 1LL << 32;
1670 	res <<= blkbits;
1671 	res -= 1;
1672 
1673 	/* Sanity check against vm- & vfs- imposed limits */
1674 	if (res > upper_limit)
1675 		res = upper_limit;
1676 
1677 	return res;
1678 }
1679 
1680 /*
1681  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
1682  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1683  * We need to be 1 filesystem block less than the 2^48 sector limit.
1684  */
1685 static loff_t ext4_max_bitmap_size(int bits)
1686 {
1687 	loff_t res = EXT4_NDIR_BLOCKS;
1688 	int meta_blocks;
1689 	loff_t upper_limit;
1690 	/* This is calculated to be the largest file size for a
1691 	 * dense, bitmapped file such that the total number of
1692 	 * sectors in the file, including data and all indirect blocks,
1693 	 * does not exceed 2^48 -1
1694 	 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1695 	 * total number of  512 bytes blocks of the file
1696 	 */
1697 
1698 	if (sizeof(blkcnt_t) < sizeof(u64)) {
1699 		/*
1700 		 * CONFIG_LSF is not enabled implies the inode
1701 		 * i_block represent total blocks in 512 bytes
1702 		 * 32 == size of vfs inode i_blocks * 8
1703 		 */
1704 		upper_limit = (1LL << 32) - 1;
1705 
1706 		/* total blocks in file system block size */
1707 		upper_limit >>= (bits - 9);
1708 
1709 	} else {
1710 		/*
1711 		 * We use 48 bit ext4_inode i_blocks
1712 		 * With EXT4_HUGE_FILE_FL set the i_blocks
1713 		 * represent total number of blocks in
1714 		 * file system block size
1715 		 */
1716 		upper_limit = (1LL << 48) - 1;
1717 
1718 	}
1719 
1720 	/* indirect blocks */
1721 	meta_blocks = 1;
1722 	/* double indirect blocks */
1723 	meta_blocks += 1 + (1LL << (bits-2));
1724 	/* tripple indirect blocks */
1725 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1726 
1727 	upper_limit -= meta_blocks;
1728 	upper_limit <<= bits;
1729 
1730 	res += 1LL << (bits-2);
1731 	res += 1LL << (2*(bits-2));
1732 	res += 1LL << (3*(bits-2));
1733 	res <<= bits;
1734 	if (res > upper_limit)
1735 		res = upper_limit;
1736 
1737 	if (res > MAX_LFS_FILESIZE)
1738 		res = MAX_LFS_FILESIZE;
1739 
1740 	return res;
1741 }
1742 
1743 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1744 				ext4_fsblk_t logical_sb_block, int nr)
1745 {
1746 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1747 	ext4_group_t bg, first_meta_bg;
1748 	int has_super = 0;
1749 
1750 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1751 
1752 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1753 	    nr < first_meta_bg)
1754 		return logical_sb_block + nr + 1;
1755 	bg = sbi->s_desc_per_block * nr;
1756 	if (ext4_bg_has_super(sb, bg))
1757 		has_super = 1;
1758 	return (has_super + ext4_group_first_block_no(sb, bg));
1759 }
1760 
1761 /**
1762  * ext4_get_stripe_size: Get the stripe size.
1763  * @sbi: In memory super block info
1764  *
1765  * If we have specified it via mount option, then
1766  * use the mount option value. If the value specified at mount time is
1767  * greater than the blocks per group use the super block value.
1768  * If the super block value is greater than blocks per group return 0.
1769  * Allocator needs it be less than blocks per group.
1770  *
1771  */
1772 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1773 {
1774 	unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1775 	unsigned long stripe_width =
1776 			le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1777 
1778 	if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1779 		return sbi->s_stripe;
1780 
1781 	if (stripe_width <= sbi->s_blocks_per_group)
1782 		return stripe_width;
1783 
1784 	if (stride <= sbi->s_blocks_per_group)
1785 		return stride;
1786 
1787 	return 0;
1788 }
1789 
1790 static int ext4_fill_super (struct super_block *sb, void *data, int silent)
1791 				__releases(kernel_sem)
1792 				__acquires(kernel_sem)
1793 
1794 {
1795 	struct buffer_head * bh;
1796 	struct ext4_super_block *es = NULL;
1797 	struct ext4_sb_info *sbi;
1798 	ext4_fsblk_t block;
1799 	ext4_fsblk_t sb_block = get_sb_block(&data);
1800 	ext4_fsblk_t logical_sb_block;
1801 	unsigned long offset = 0;
1802 	unsigned int journal_inum = 0;
1803 	unsigned long journal_devnum = 0;
1804 	unsigned long def_mount_opts;
1805 	struct inode *root;
1806 	int ret = -EINVAL;
1807 	int blocksize;
1808 	int db_count;
1809 	int i;
1810 	int needs_recovery;
1811 	__le32 features;
1812 	__u64 blocks_count;
1813 	int err;
1814 
1815 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1816 	if (!sbi)
1817 		return -ENOMEM;
1818 	sb->s_fs_info = sbi;
1819 	sbi->s_mount_opt = 0;
1820 	sbi->s_resuid = EXT4_DEF_RESUID;
1821 	sbi->s_resgid = EXT4_DEF_RESGID;
1822 	sbi->s_sb_block = sb_block;
1823 
1824 	unlock_kernel();
1825 
1826 	blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
1827 	if (!blocksize) {
1828 		printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
1829 		goto out_fail;
1830 	}
1831 
1832 	if (!sb_set_blocksize(sb, blocksize)) {
1833 		printk(KERN_ERR "EXT4-fs: bad blocksize %d.\n", blocksize);
1834 		goto out_fail;
1835 	}
1836 
1837 	/*
1838 	 * The ext4 superblock will not be buffer aligned for other than 1kB
1839 	 * block sizes.  We need to calculate the offset from buffer start.
1840 	 */
1841 	if (blocksize != EXT4_MIN_BLOCK_SIZE) {
1842 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1843 		offset = do_div(logical_sb_block, blocksize);
1844 	} else {
1845 		logical_sb_block = sb_block;
1846 	}
1847 
1848 	if (!(bh = sb_bread(sb, logical_sb_block))) {
1849 		printk (KERN_ERR "EXT4-fs: unable to read superblock\n");
1850 		goto out_fail;
1851 	}
1852 	/*
1853 	 * Note: s_es must be initialized as soon as possible because
1854 	 *       some ext4 macro-instructions depend on its value
1855 	 */
1856 	es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
1857 	sbi->s_es = es;
1858 	sb->s_magic = le16_to_cpu(es->s_magic);
1859 	if (sb->s_magic != EXT4_SUPER_MAGIC)
1860 		goto cantfind_ext4;
1861 
1862 	/* Set defaults before we parse the mount options */
1863 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1864 	if (def_mount_opts & EXT4_DEFM_DEBUG)
1865 		set_opt(sbi->s_mount_opt, DEBUG);
1866 	if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
1867 		set_opt(sbi->s_mount_opt, GRPID);
1868 	if (def_mount_opts & EXT4_DEFM_UID16)
1869 		set_opt(sbi->s_mount_opt, NO_UID32);
1870 #ifdef CONFIG_EXT4DEV_FS_XATTR
1871 	if (def_mount_opts & EXT4_DEFM_XATTR_USER)
1872 		set_opt(sbi->s_mount_opt, XATTR_USER);
1873 #endif
1874 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1875 	if (def_mount_opts & EXT4_DEFM_ACL)
1876 		set_opt(sbi->s_mount_opt, POSIX_ACL);
1877 #endif
1878 	if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1879 		sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1880 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1881 		sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1882 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1883 		sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1884 
1885 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
1886 		set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1887 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
1888 		set_opt(sbi->s_mount_opt, ERRORS_CONT);
1889 	else
1890 		set_opt(sbi->s_mount_opt, ERRORS_RO);
1891 
1892 	sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1893 	sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1894 
1895 	set_opt(sbi->s_mount_opt, RESERVATION);
1896 
1897 	/*
1898 	 * turn on extents feature by default in ext4 filesystem
1899 	 * User -o noextents to turn it off
1900 	 */
1901 	set_opt(sbi->s_mount_opt, EXTENTS);
1902 	/*
1903 	 * turn on mballoc feature by default in ext4 filesystem
1904 	 * User -o nomballoc to turn it off
1905 	 */
1906 	set_opt(sbi->s_mount_opt, MBALLOC);
1907 
1908 	if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1909 			    NULL, 0))
1910 		goto failed_mount;
1911 
1912 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1913 		((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1914 
1915 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1916 	    (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1917 	     EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1918 	     EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1919 		printk(KERN_WARNING
1920 		       "EXT4-fs warning: feature flags set on rev 0 fs, "
1921 		       "running e2fsck is recommended\n");
1922 	/*
1923 	 * Check feature flags regardless of the revision level, since we
1924 	 * previously didn't change the revision level when setting the flags,
1925 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
1926 	 */
1927 	features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
1928 	if (features) {
1929 		printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
1930 		       "unsupported optional features (%x).\n",
1931 		       sb->s_id, le32_to_cpu(features));
1932 		goto failed_mount;
1933 	}
1934 	features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
1935 	if (!(sb->s_flags & MS_RDONLY) && features) {
1936 		printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
1937 		       "unsupported optional features (%x).\n",
1938 		       sb->s_id, le32_to_cpu(features));
1939 		goto failed_mount;
1940 	}
1941 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
1942 		/*
1943 		 * Large file size enabled file system can only be
1944 		 * mount if kernel is build with CONFIG_LSF
1945 		 */
1946 		if (sizeof(root->i_blocks) < sizeof(u64) &&
1947 				!(sb->s_flags & MS_RDONLY)) {
1948 			printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
1949 					"files cannot be mounted read-write "
1950 					"without CONFIG_LSF.\n", sb->s_id);
1951 			goto failed_mount;
1952 		}
1953 	}
1954 	blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1955 
1956 	if (blocksize < EXT4_MIN_BLOCK_SIZE ||
1957 	    blocksize > EXT4_MAX_BLOCK_SIZE) {
1958 		printk(KERN_ERR
1959 		       "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
1960 		       blocksize, sb->s_id);
1961 		goto failed_mount;
1962 	}
1963 
1964 	if (sb->s_blocksize != blocksize) {
1965 
1966 		/* Validate the filesystem blocksize */
1967 		if (!sb_set_blocksize(sb, blocksize)) {
1968 			printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
1969 					blocksize);
1970 			goto failed_mount;
1971 		}
1972 
1973 		brelse (bh);
1974 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1975 		offset = do_div(logical_sb_block, blocksize);
1976 		bh = sb_bread(sb, logical_sb_block);
1977 		if (!bh) {
1978 			printk(KERN_ERR
1979 			       "EXT4-fs: Can't read superblock on 2nd try.\n");
1980 			goto failed_mount;
1981 		}
1982 		es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
1983 		sbi->s_es = es;
1984 		if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
1985 			printk (KERN_ERR
1986 				"EXT4-fs: Magic mismatch, very weird !\n");
1987 			goto failed_mount;
1988 		}
1989 	}
1990 
1991 	sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits);
1992 	sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
1993 
1994 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
1995 		sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
1996 		sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
1997 	} else {
1998 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1999 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2000 		if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2001 		    (!is_power_of_2(sbi->s_inode_size)) ||
2002 		    (sbi->s_inode_size > blocksize)) {
2003 			printk (KERN_ERR
2004 				"EXT4-fs: unsupported inode size: %d\n",
2005 				sbi->s_inode_size);
2006 			goto failed_mount;
2007 		}
2008 		if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2009 			sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2010 	}
2011 	sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2012 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2013 		if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2014 		    sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2015 		    !is_power_of_2(sbi->s_desc_size)) {
2016 			printk(KERN_ERR
2017 			       "EXT4-fs: unsupported descriptor size %lu\n",
2018 			       sbi->s_desc_size);
2019 			goto failed_mount;
2020 		}
2021 	} else
2022 		sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2023 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2024 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2025 	if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2026 		goto cantfind_ext4;
2027 	sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2028 	if (sbi->s_inodes_per_block == 0)
2029 		goto cantfind_ext4;
2030 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
2031 					sbi->s_inodes_per_block;
2032 	sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2033 	sbi->s_sbh = bh;
2034 	sbi->s_mount_state = le16_to_cpu(es->s_state);
2035 	sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2036 	sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2037 	for (i=0; i < 4; i++)
2038 		sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2039 	sbi->s_def_hash_version = es->s_def_hash_version;
2040 
2041 	if (sbi->s_blocks_per_group > blocksize * 8) {
2042 		printk (KERN_ERR
2043 			"EXT4-fs: #blocks per group too big: %lu\n",
2044 			sbi->s_blocks_per_group);
2045 		goto failed_mount;
2046 	}
2047 	if (sbi->s_inodes_per_group > blocksize * 8) {
2048 		printk (KERN_ERR
2049 			"EXT4-fs: #inodes per group too big: %lu\n",
2050 			sbi->s_inodes_per_group);
2051 		goto failed_mount;
2052 	}
2053 
2054 	if (ext4_blocks_count(es) >
2055 		    (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2056 		printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2057 			" too large to mount safely\n", sb->s_id);
2058 		if (sizeof(sector_t) < 8)
2059 			printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2060 					"enabled\n");
2061 		goto failed_mount;
2062 	}
2063 
2064 	if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2065 		goto cantfind_ext4;
2066 
2067 	/* ensure blocks_count calculation below doesn't sign-extend */
2068 	if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2069 	    le32_to_cpu(es->s_first_data_block) + 1) {
2070 		printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2071 		       "first data block %u, blocks per group %lu\n",
2072 			ext4_blocks_count(es),
2073 			le32_to_cpu(es->s_first_data_block),
2074 			EXT4_BLOCKS_PER_GROUP(sb));
2075 		goto failed_mount;
2076 	}
2077 	blocks_count = (ext4_blocks_count(es) -
2078 			le32_to_cpu(es->s_first_data_block) +
2079 			EXT4_BLOCKS_PER_GROUP(sb) - 1);
2080 	do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2081 	sbi->s_groups_count = blocks_count;
2082 	db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2083 		   EXT4_DESC_PER_BLOCK(sb);
2084 	sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
2085 				    GFP_KERNEL);
2086 	if (sbi->s_group_desc == NULL) {
2087 		printk (KERN_ERR "EXT4-fs: not enough memory\n");
2088 		goto failed_mount;
2089 	}
2090 
2091 	bgl_lock_init(&sbi->s_blockgroup_lock);
2092 
2093 	for (i = 0; i < db_count; i++) {
2094 		block = descriptor_loc(sb, logical_sb_block, i);
2095 		sbi->s_group_desc[i] = sb_bread(sb, block);
2096 		if (!sbi->s_group_desc[i]) {
2097 			printk (KERN_ERR "EXT4-fs: "
2098 				"can't read group descriptor %d\n", i);
2099 			db_count = i;
2100 			goto failed_mount2;
2101 		}
2102 	}
2103 	if (!ext4_check_descriptors (sb)) {
2104 		printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2105 		goto failed_mount2;
2106 	}
2107 	sbi->s_gdb_count = db_count;
2108 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2109 	spin_lock_init(&sbi->s_next_gen_lock);
2110 
2111 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
2112 			ext4_count_free_blocks(sb));
2113 	if (!err) {
2114 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
2115 				ext4_count_free_inodes(sb));
2116 	}
2117 	if (!err) {
2118 		err = percpu_counter_init(&sbi->s_dirs_counter,
2119 				ext4_count_dirs(sb));
2120 	}
2121 	if (err) {
2122 		printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2123 		goto failed_mount3;
2124 	}
2125 
2126 	/* per fileystem reservation list head & lock */
2127 	spin_lock_init(&sbi->s_rsv_window_lock);
2128 	sbi->s_rsv_window_root = RB_ROOT;
2129 	/* Add a single, static dummy reservation to the start of the
2130 	 * reservation window list --- it gives us a placeholder for
2131 	 * append-at-start-of-list which makes the allocation logic
2132 	 * _much_ simpler. */
2133 	sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2134 	sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2135 	sbi->s_rsv_window_head.rsv_alloc_hit = 0;
2136 	sbi->s_rsv_window_head.rsv_goal_size = 0;
2137 	ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
2138 
2139 	sbi->s_stripe = ext4_get_stripe_size(sbi);
2140 
2141 	/*
2142 	 * set up enough so that it can read an inode
2143 	 */
2144 	sb->s_op = &ext4_sops;
2145 	sb->s_export_op = &ext4_export_ops;
2146 	sb->s_xattr = ext4_xattr_handlers;
2147 #ifdef CONFIG_QUOTA
2148 	sb->s_qcop = &ext4_qctl_operations;
2149 	sb->dq_op = &ext4_quota_operations;
2150 #endif
2151 	INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2152 
2153 	sb->s_root = NULL;
2154 
2155 	needs_recovery = (es->s_last_orphan != 0 ||
2156 			  EXT4_HAS_INCOMPAT_FEATURE(sb,
2157 				    EXT4_FEATURE_INCOMPAT_RECOVER));
2158 
2159 	/*
2160 	 * The first inode we look at is the journal inode.  Don't try
2161 	 * root first: it may be modified in the journal!
2162 	 */
2163 	if (!test_opt(sb, NOLOAD) &&
2164 	    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2165 		if (ext4_load_journal(sb, es, journal_devnum))
2166 			goto failed_mount3;
2167 	} else if (journal_inum) {
2168 		if (ext4_create_journal(sb, es, journal_inum))
2169 			goto failed_mount3;
2170 	} else {
2171 		if (!silent)
2172 			printk (KERN_ERR
2173 				"ext4: No journal on filesystem on %s\n",
2174 				sb->s_id);
2175 		goto failed_mount3;
2176 	}
2177 
2178 	if (ext4_blocks_count(es) > 0xffffffffULL &&
2179 	    !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2180 				       JBD2_FEATURE_INCOMPAT_64BIT)) {
2181 		printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2182 		goto failed_mount4;
2183 	}
2184 
2185 	if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2186 		jbd2_journal_set_features(sbi->s_journal,
2187 				JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2188 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2189 	} else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2190 		jbd2_journal_set_features(sbi->s_journal,
2191 				JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2192 		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2193 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2194 	} else {
2195 		jbd2_journal_clear_features(sbi->s_journal,
2196 				JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2197 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2198 	}
2199 
2200 	/* We have now updated the journal if required, so we can
2201 	 * validate the data journaling mode. */
2202 	switch (test_opt(sb, DATA_FLAGS)) {
2203 	case 0:
2204 		/* No mode set, assume a default based on the journal
2205 		 * capabilities: ORDERED_DATA if the journal can
2206 		 * cope, else JOURNAL_DATA
2207 		 */
2208 		if (jbd2_journal_check_available_features
2209 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2210 			set_opt(sbi->s_mount_opt, ORDERED_DATA);
2211 		else
2212 			set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2213 		break;
2214 
2215 	case EXT4_MOUNT_ORDERED_DATA:
2216 	case EXT4_MOUNT_WRITEBACK_DATA:
2217 		if (!jbd2_journal_check_available_features
2218 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2219 			printk(KERN_ERR "EXT4-fs: Journal does not support "
2220 			       "requested data journaling mode\n");
2221 			goto failed_mount4;
2222 		}
2223 	default:
2224 		break;
2225 	}
2226 
2227 	if (test_opt(sb, NOBH)) {
2228 		if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2229 			printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2230 				"its supported only with writeback mode\n");
2231 			clear_opt(sbi->s_mount_opt, NOBH);
2232 		}
2233 	}
2234 	/*
2235 	 * The jbd2_journal_load will have done any necessary log recovery,
2236 	 * so we can safely mount the rest of the filesystem now.
2237 	 */
2238 
2239 	root = ext4_iget(sb, EXT4_ROOT_INO);
2240 	if (IS_ERR(root)) {
2241 		printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2242 		ret = PTR_ERR(root);
2243 		goto failed_mount4;
2244 	}
2245 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2246 		iput(root);
2247 		printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2248 		goto failed_mount4;
2249 	}
2250 	sb->s_root = d_alloc_root(root);
2251 	if (!sb->s_root) {
2252 		printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2253 		iput(root);
2254 		ret = -ENOMEM;
2255 		goto failed_mount4;
2256 	}
2257 
2258 	ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
2259 
2260 	/* determine the minimum size of new large inodes, if present */
2261 	if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2262 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2263 						     EXT4_GOOD_OLD_INODE_SIZE;
2264 		if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2265 				       EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2266 			if (sbi->s_want_extra_isize <
2267 			    le16_to_cpu(es->s_want_extra_isize))
2268 				sbi->s_want_extra_isize =
2269 					le16_to_cpu(es->s_want_extra_isize);
2270 			if (sbi->s_want_extra_isize <
2271 			    le16_to_cpu(es->s_min_extra_isize))
2272 				sbi->s_want_extra_isize =
2273 					le16_to_cpu(es->s_min_extra_isize);
2274 		}
2275 	}
2276 	/* Check if enough inode space is available */
2277 	if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2278 							sbi->s_inode_size) {
2279 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2280 						       EXT4_GOOD_OLD_INODE_SIZE;
2281 		printk(KERN_INFO "EXT4-fs: required extra inode space not"
2282 			"available.\n");
2283 	}
2284 
2285 	/*
2286 	 * akpm: core read_super() calls in here with the superblock locked.
2287 	 * That deadlocks, because orphan cleanup needs to lock the superblock
2288 	 * in numerous places.  Here we just pop the lock - it's relatively
2289 	 * harmless, because we are now ready to accept write_super() requests,
2290 	 * and aviro says that's the only reason for hanging onto the
2291 	 * superblock lock.
2292 	 */
2293 	EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2294 	ext4_orphan_cleanup(sb, es);
2295 	EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2296 	if (needs_recovery)
2297 		printk (KERN_INFO "EXT4-fs: recovery complete.\n");
2298 	ext4_mark_recovery_complete(sb, es);
2299 	printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2300 		test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2301 		test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
2302 		"writeback");
2303 
2304 	ext4_ext_init(sb);
2305 	ext4_mb_init(sb, needs_recovery);
2306 
2307 	lock_kernel();
2308 	return 0;
2309 
2310 cantfind_ext4:
2311 	if (!silent)
2312 		printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2313 		       sb->s_id);
2314 	goto failed_mount;
2315 
2316 failed_mount4:
2317 	jbd2_journal_destroy(sbi->s_journal);
2318 failed_mount3:
2319 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
2320 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
2321 	percpu_counter_destroy(&sbi->s_dirs_counter);
2322 failed_mount2:
2323 	for (i = 0; i < db_count; i++)
2324 		brelse(sbi->s_group_desc[i]);
2325 	kfree(sbi->s_group_desc);
2326 failed_mount:
2327 #ifdef CONFIG_QUOTA
2328 	for (i = 0; i < MAXQUOTAS; i++)
2329 		kfree(sbi->s_qf_names[i]);
2330 #endif
2331 	ext4_blkdev_remove(sbi);
2332 	brelse(bh);
2333 out_fail:
2334 	sb->s_fs_info = NULL;
2335 	kfree(sbi);
2336 	lock_kernel();
2337 	return ret;
2338 }
2339 
2340 /*
2341  * Setup any per-fs journal parameters now.  We'll do this both on
2342  * initial mount, once the journal has been initialised but before we've
2343  * done any recovery; and again on any subsequent remount.
2344  */
2345 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2346 {
2347 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2348 
2349 	if (sbi->s_commit_interval)
2350 		journal->j_commit_interval = sbi->s_commit_interval;
2351 	/* We could also set up an ext4-specific default for the commit
2352 	 * interval here, but for now we'll just fall back to the jbd
2353 	 * default. */
2354 
2355 	spin_lock(&journal->j_state_lock);
2356 	if (test_opt(sb, BARRIER))
2357 		journal->j_flags |= JBD2_BARRIER;
2358 	else
2359 		journal->j_flags &= ~JBD2_BARRIER;
2360 	spin_unlock(&journal->j_state_lock);
2361 }
2362 
2363 static journal_t *ext4_get_journal(struct super_block *sb,
2364 				   unsigned int journal_inum)
2365 {
2366 	struct inode *journal_inode;
2367 	journal_t *journal;
2368 
2369 	/* First, test for the existence of a valid inode on disk.  Bad
2370 	 * things happen if we iget() an unused inode, as the subsequent
2371 	 * iput() will try to delete it. */
2372 
2373 	journal_inode = ext4_iget(sb, journal_inum);
2374 	if (IS_ERR(journal_inode)) {
2375 		printk(KERN_ERR "EXT4-fs: no journal found.\n");
2376 		return NULL;
2377 	}
2378 	if (!journal_inode->i_nlink) {
2379 		make_bad_inode(journal_inode);
2380 		iput(journal_inode);
2381 		printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2382 		return NULL;
2383 	}
2384 
2385 	jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2386 		  journal_inode, journal_inode->i_size);
2387 	if (!S_ISREG(journal_inode->i_mode)) {
2388 		printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2389 		iput(journal_inode);
2390 		return NULL;
2391 	}
2392 
2393 	journal = jbd2_journal_init_inode(journal_inode);
2394 	if (!journal) {
2395 		printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2396 		iput(journal_inode);
2397 		return NULL;
2398 	}
2399 	journal->j_private = sb;
2400 	ext4_init_journal_params(sb, journal);
2401 	return journal;
2402 }
2403 
2404 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2405 				       dev_t j_dev)
2406 {
2407 	struct buffer_head * bh;
2408 	journal_t *journal;
2409 	ext4_fsblk_t start;
2410 	ext4_fsblk_t len;
2411 	int hblock, blocksize;
2412 	ext4_fsblk_t sb_block;
2413 	unsigned long offset;
2414 	struct ext4_super_block * es;
2415 	struct block_device *bdev;
2416 
2417 	bdev = ext4_blkdev_get(j_dev);
2418 	if (bdev == NULL)
2419 		return NULL;
2420 
2421 	if (bd_claim(bdev, sb)) {
2422 		printk(KERN_ERR
2423 			"EXT4: failed to claim external journal device.\n");
2424 		blkdev_put(bdev);
2425 		return NULL;
2426 	}
2427 
2428 	blocksize = sb->s_blocksize;
2429 	hblock = bdev_hardsect_size(bdev);
2430 	if (blocksize < hblock) {
2431 		printk(KERN_ERR
2432 			"EXT4-fs: blocksize too small for journal device.\n");
2433 		goto out_bdev;
2434 	}
2435 
2436 	sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2437 	offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2438 	set_blocksize(bdev, blocksize);
2439 	if (!(bh = __bread(bdev, sb_block, blocksize))) {
2440 		printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2441 		       "external journal\n");
2442 		goto out_bdev;
2443 	}
2444 
2445 	es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2446 	if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2447 	    !(le32_to_cpu(es->s_feature_incompat) &
2448 	      EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2449 		printk(KERN_ERR "EXT4-fs: external journal has "
2450 					"bad superblock\n");
2451 		brelse(bh);
2452 		goto out_bdev;
2453 	}
2454 
2455 	if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2456 		printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2457 		brelse(bh);
2458 		goto out_bdev;
2459 	}
2460 
2461 	len = ext4_blocks_count(es);
2462 	start = sb_block + 1;
2463 	brelse(bh);	/* we're done with the superblock */
2464 
2465 	journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2466 					start, len, blocksize);
2467 	if (!journal) {
2468 		printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2469 		goto out_bdev;
2470 	}
2471 	journal->j_private = sb;
2472 	ll_rw_block(READ, 1, &journal->j_sb_buffer);
2473 	wait_on_buffer(journal->j_sb_buffer);
2474 	if (!buffer_uptodate(journal->j_sb_buffer)) {
2475 		printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2476 		goto out_journal;
2477 	}
2478 	if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2479 		printk(KERN_ERR "EXT4-fs: External journal has more than one "
2480 					"user (unsupported) - %d\n",
2481 			be32_to_cpu(journal->j_superblock->s_nr_users));
2482 		goto out_journal;
2483 	}
2484 	EXT4_SB(sb)->journal_bdev = bdev;
2485 	ext4_init_journal_params(sb, journal);
2486 	return journal;
2487 out_journal:
2488 	jbd2_journal_destroy(journal);
2489 out_bdev:
2490 	ext4_blkdev_put(bdev);
2491 	return NULL;
2492 }
2493 
2494 static int ext4_load_journal(struct super_block *sb,
2495 			     struct ext4_super_block *es,
2496 			     unsigned long journal_devnum)
2497 {
2498 	journal_t *journal;
2499 	unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2500 	dev_t journal_dev;
2501 	int err = 0;
2502 	int really_read_only;
2503 
2504 	if (journal_devnum &&
2505 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2506 		printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2507 			"numbers have changed\n");
2508 		journal_dev = new_decode_dev(journal_devnum);
2509 	} else
2510 		journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2511 
2512 	really_read_only = bdev_read_only(sb->s_bdev);
2513 
2514 	/*
2515 	 * Are we loading a blank journal or performing recovery after a
2516 	 * crash?  For recovery, we need to check in advance whether we
2517 	 * can get read-write access to the device.
2518 	 */
2519 
2520 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2521 		if (sb->s_flags & MS_RDONLY) {
2522 			printk(KERN_INFO "EXT4-fs: INFO: recovery "
2523 					"required on readonly filesystem.\n");
2524 			if (really_read_only) {
2525 				printk(KERN_ERR "EXT4-fs: write access "
2526 					"unavailable, cannot proceed.\n");
2527 				return -EROFS;
2528 			}
2529 			printk (KERN_INFO "EXT4-fs: write access will "
2530 					"be enabled during recovery.\n");
2531 		}
2532 	}
2533 
2534 	if (journal_inum && journal_dev) {
2535 		printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2536 		       "and inode journals!\n");
2537 		return -EINVAL;
2538 	}
2539 
2540 	if (journal_inum) {
2541 		if (!(journal = ext4_get_journal(sb, journal_inum)))
2542 			return -EINVAL;
2543 	} else {
2544 		if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2545 			return -EINVAL;
2546 	}
2547 
2548 	if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2549 		err = jbd2_journal_update_format(journal);
2550 		if (err)  {
2551 			printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2552 			jbd2_journal_destroy(journal);
2553 			return err;
2554 		}
2555 	}
2556 
2557 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2558 		err = jbd2_journal_wipe(journal, !really_read_only);
2559 	if (!err)
2560 		err = jbd2_journal_load(journal);
2561 
2562 	if (err) {
2563 		printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2564 		jbd2_journal_destroy(journal);
2565 		return err;
2566 	}
2567 
2568 	EXT4_SB(sb)->s_journal = journal;
2569 	ext4_clear_journal_err(sb, es);
2570 
2571 	if (journal_devnum &&
2572 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2573 		es->s_journal_dev = cpu_to_le32(journal_devnum);
2574 		sb->s_dirt = 1;
2575 
2576 		/* Make sure we flush the recovery flag to disk. */
2577 		ext4_commit_super(sb, es, 1);
2578 	}
2579 
2580 	return 0;
2581 }
2582 
2583 static int ext4_create_journal(struct super_block * sb,
2584 			       struct ext4_super_block * es,
2585 			       unsigned int journal_inum)
2586 {
2587 	journal_t *journal;
2588 	int err;
2589 
2590 	if (sb->s_flags & MS_RDONLY) {
2591 		printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
2592 				"create journal.\n");
2593 		return -EROFS;
2594 	}
2595 
2596 	journal = ext4_get_journal(sb, journal_inum);
2597 	if (!journal)
2598 		return -EINVAL;
2599 
2600 	printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
2601 	       journal_inum);
2602 
2603 	err = jbd2_journal_create(journal);
2604 	if (err) {
2605 		printk(KERN_ERR "EXT4-fs: error creating journal.\n");
2606 		jbd2_journal_destroy(journal);
2607 		return -EIO;
2608 	}
2609 
2610 	EXT4_SB(sb)->s_journal = journal;
2611 
2612 	ext4_update_dynamic_rev(sb);
2613 	EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2614 	EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
2615 
2616 	es->s_journal_inum = cpu_to_le32(journal_inum);
2617 	sb->s_dirt = 1;
2618 
2619 	/* Make sure we flush the recovery flag to disk. */
2620 	ext4_commit_super(sb, es, 1);
2621 
2622 	return 0;
2623 }
2624 
2625 static void ext4_commit_super (struct super_block * sb,
2626 			       struct ext4_super_block * es,
2627 			       int sync)
2628 {
2629 	struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2630 
2631 	if (!sbh)
2632 		return;
2633 	es->s_wtime = cpu_to_le32(get_seconds());
2634 	ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
2635 	es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
2636 	BUFFER_TRACE(sbh, "marking dirty");
2637 	mark_buffer_dirty(sbh);
2638 	if (sync)
2639 		sync_dirty_buffer(sbh);
2640 }
2641 
2642 
2643 /*
2644  * Have we just finished recovery?  If so, and if we are mounting (or
2645  * remounting) the filesystem readonly, then we will end up with a
2646  * consistent fs on disk.  Record that fact.
2647  */
2648 static void ext4_mark_recovery_complete(struct super_block * sb,
2649 					struct ext4_super_block * es)
2650 {
2651 	journal_t *journal = EXT4_SB(sb)->s_journal;
2652 
2653 	jbd2_journal_lock_updates(journal);
2654 	jbd2_journal_flush(journal);
2655 	lock_super(sb);
2656 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2657 	    sb->s_flags & MS_RDONLY) {
2658 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2659 		sb->s_dirt = 0;
2660 		ext4_commit_super(sb, es, 1);
2661 	}
2662 	unlock_super(sb);
2663 	jbd2_journal_unlock_updates(journal);
2664 }
2665 
2666 /*
2667  * If we are mounting (or read-write remounting) a filesystem whose journal
2668  * has recorded an error from a previous lifetime, move that error to the
2669  * main filesystem now.
2670  */
2671 static void ext4_clear_journal_err(struct super_block * sb,
2672 				   struct ext4_super_block * es)
2673 {
2674 	journal_t *journal;
2675 	int j_errno;
2676 	const char *errstr;
2677 
2678 	journal = EXT4_SB(sb)->s_journal;
2679 
2680 	/*
2681 	 * Now check for any error status which may have been recorded in the
2682 	 * journal by a prior ext4_error() or ext4_abort()
2683 	 */
2684 
2685 	j_errno = jbd2_journal_errno(journal);
2686 	if (j_errno) {
2687 		char nbuf[16];
2688 
2689 		errstr = ext4_decode_error(sb, j_errno, nbuf);
2690 		ext4_warning(sb, __FUNCTION__, "Filesystem error recorded "
2691 			     "from previous mount: %s", errstr);
2692 		ext4_warning(sb, __FUNCTION__, "Marking fs in need of "
2693 			     "filesystem check.");
2694 
2695 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2696 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2697 		ext4_commit_super (sb, es, 1);
2698 
2699 		jbd2_journal_clear_err(journal);
2700 	}
2701 }
2702 
2703 /*
2704  * Force the running and committing transactions to commit,
2705  * and wait on the commit.
2706  */
2707 int ext4_force_commit(struct super_block *sb)
2708 {
2709 	journal_t *journal;
2710 	int ret;
2711 
2712 	if (sb->s_flags & MS_RDONLY)
2713 		return 0;
2714 
2715 	journal = EXT4_SB(sb)->s_journal;
2716 	sb->s_dirt = 0;
2717 	ret = ext4_journal_force_commit(journal);
2718 	return ret;
2719 }
2720 
2721 /*
2722  * Ext4 always journals updates to the superblock itself, so we don't
2723  * have to propagate any other updates to the superblock on disk at this
2724  * point.  Just start an async writeback to get the buffers on their way
2725  * to the disk.
2726  *
2727  * This implicitly triggers the writebehind on sync().
2728  */
2729 
2730 static void ext4_write_super (struct super_block * sb)
2731 {
2732 	if (mutex_trylock(&sb->s_lock) != 0)
2733 		BUG();
2734 	sb->s_dirt = 0;
2735 }
2736 
2737 static int ext4_sync_fs(struct super_block *sb, int wait)
2738 {
2739 	tid_t target;
2740 
2741 	sb->s_dirt = 0;
2742 	if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
2743 		if (wait)
2744 			jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
2745 	}
2746 	return 0;
2747 }
2748 
2749 /*
2750  * LVM calls this function before a (read-only) snapshot is created.  This
2751  * gives us a chance to flush the journal completely and mark the fs clean.
2752  */
2753 static void ext4_write_super_lockfs(struct super_block *sb)
2754 {
2755 	sb->s_dirt = 0;
2756 
2757 	if (!(sb->s_flags & MS_RDONLY)) {
2758 		journal_t *journal = EXT4_SB(sb)->s_journal;
2759 
2760 		/* Now we set up the journal barrier. */
2761 		jbd2_journal_lock_updates(journal);
2762 		jbd2_journal_flush(journal);
2763 
2764 		/* Journal blocked and flushed, clear needs_recovery flag. */
2765 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2766 		ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2767 	}
2768 }
2769 
2770 /*
2771  * Called by LVM after the snapshot is done.  We need to reset the RECOVER
2772  * flag here, even though the filesystem is not technically dirty yet.
2773  */
2774 static void ext4_unlockfs(struct super_block *sb)
2775 {
2776 	if (!(sb->s_flags & MS_RDONLY)) {
2777 		lock_super(sb);
2778 		/* Reser the needs_recovery flag before the fs is unlocked. */
2779 		EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2780 		ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2781 		unlock_super(sb);
2782 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
2783 	}
2784 }
2785 
2786 static int ext4_remount (struct super_block * sb, int * flags, char * data)
2787 {
2788 	struct ext4_super_block * es;
2789 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2790 	ext4_fsblk_t n_blocks_count = 0;
2791 	unsigned long old_sb_flags;
2792 	struct ext4_mount_options old_opts;
2793 	int err;
2794 #ifdef CONFIG_QUOTA
2795 	int i;
2796 #endif
2797 
2798 	/* Store the original options */
2799 	old_sb_flags = sb->s_flags;
2800 	old_opts.s_mount_opt = sbi->s_mount_opt;
2801 	old_opts.s_resuid = sbi->s_resuid;
2802 	old_opts.s_resgid = sbi->s_resgid;
2803 	old_opts.s_commit_interval = sbi->s_commit_interval;
2804 #ifdef CONFIG_QUOTA
2805 	old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2806 	for (i = 0; i < MAXQUOTAS; i++)
2807 		old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2808 #endif
2809 
2810 	/*
2811 	 * Allow the "check" option to be passed as a remount option.
2812 	 */
2813 	if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2814 		err = -EINVAL;
2815 		goto restore_opts;
2816 	}
2817 
2818 	if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
2819 		ext4_abort(sb, __FUNCTION__, "Abort forced by user");
2820 
2821 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2822 		((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2823 
2824 	es = sbi->s_es;
2825 
2826 	ext4_init_journal_params(sb, sbi->s_journal);
2827 
2828 	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2829 		n_blocks_count > ext4_blocks_count(es)) {
2830 		if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
2831 			err = -EROFS;
2832 			goto restore_opts;
2833 		}
2834 
2835 		if (*flags & MS_RDONLY) {
2836 			/*
2837 			 * First of all, the unconditional stuff we have to do
2838 			 * to disable replay of the journal when we next remount
2839 			 */
2840 			sb->s_flags |= MS_RDONLY;
2841 
2842 			/*
2843 			 * OK, test if we are remounting a valid rw partition
2844 			 * readonly, and if so set the rdonly flag and then
2845 			 * mark the partition as valid again.
2846 			 */
2847 			if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
2848 			    (sbi->s_mount_state & EXT4_VALID_FS))
2849 				es->s_state = cpu_to_le16(sbi->s_mount_state);
2850 
2851 			/*
2852 			 * We have to unlock super so that we can wait for
2853 			 * transactions.
2854 			 */
2855 			unlock_super(sb);
2856 			ext4_mark_recovery_complete(sb, es);
2857 			lock_super(sb);
2858 		} else {
2859 			__le32 ret;
2860 			if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2861 					~EXT4_FEATURE_RO_COMPAT_SUPP))) {
2862 				printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2863 				       "remount RDWR because of unsupported "
2864 				       "optional features (%x).\n",
2865 				       sb->s_id, le32_to_cpu(ret));
2866 				err = -EROFS;
2867 				goto restore_opts;
2868 			}
2869 
2870 			/*
2871 			 * If we have an unprocessed orphan list hanging
2872 			 * around from a previously readonly bdev mount,
2873 			 * require a full umount/remount for now.
2874 			 */
2875 			if (es->s_last_orphan) {
2876 				printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2877 				       "remount RDWR because of unprocessed "
2878 				       "orphan inode list.  Please "
2879 				       "umount/remount instead.\n",
2880 				       sb->s_id);
2881 				err = -EINVAL;
2882 				goto restore_opts;
2883 			}
2884 
2885 			/*
2886 			 * Mounting a RDONLY partition read-write, so reread
2887 			 * and store the current valid flag.  (It may have
2888 			 * been changed by e2fsck since we originally mounted
2889 			 * the partition.)
2890 			 */
2891 			ext4_clear_journal_err(sb, es);
2892 			sbi->s_mount_state = le16_to_cpu(es->s_state);
2893 			if ((err = ext4_group_extend(sb, es, n_blocks_count)))
2894 				goto restore_opts;
2895 			if (!ext4_setup_super (sb, es, 0))
2896 				sb->s_flags &= ~MS_RDONLY;
2897 		}
2898 	}
2899 #ifdef CONFIG_QUOTA
2900 	/* Release old quota file names */
2901 	for (i = 0; i < MAXQUOTAS; i++)
2902 		if (old_opts.s_qf_names[i] &&
2903 		    old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2904 			kfree(old_opts.s_qf_names[i]);
2905 #endif
2906 	return 0;
2907 restore_opts:
2908 	sb->s_flags = old_sb_flags;
2909 	sbi->s_mount_opt = old_opts.s_mount_opt;
2910 	sbi->s_resuid = old_opts.s_resuid;
2911 	sbi->s_resgid = old_opts.s_resgid;
2912 	sbi->s_commit_interval = old_opts.s_commit_interval;
2913 #ifdef CONFIG_QUOTA
2914 	sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2915 	for (i = 0; i < MAXQUOTAS; i++) {
2916 		if (sbi->s_qf_names[i] &&
2917 		    old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2918 			kfree(sbi->s_qf_names[i]);
2919 		sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2920 	}
2921 #endif
2922 	return err;
2923 }
2924 
2925 static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf)
2926 {
2927 	struct super_block *sb = dentry->d_sb;
2928 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2929 	struct ext4_super_block *es = sbi->s_es;
2930 	u64 fsid;
2931 
2932 	if (test_opt(sb, MINIX_DF)) {
2933 		sbi->s_overhead_last = 0;
2934 	} else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
2935 		ext4_group_t ngroups = sbi->s_groups_count, i;
2936 		ext4_fsblk_t overhead = 0;
2937 		smp_rmb();
2938 
2939 		/*
2940 		 * Compute the overhead (FS structures).  This is constant
2941 		 * for a given filesystem unless the number of block groups
2942 		 * changes so we cache the previous value until it does.
2943 		 */
2944 
2945 		/*
2946 		 * All of the blocks before first_data_block are
2947 		 * overhead
2948 		 */
2949 		overhead = le32_to_cpu(es->s_first_data_block);
2950 
2951 		/*
2952 		 * Add the overhead attributed to the superblock and
2953 		 * block group descriptors.  If the sparse superblocks
2954 		 * feature is turned on, then not all groups have this.
2955 		 */
2956 		for (i = 0; i < ngroups; i++) {
2957 			overhead += ext4_bg_has_super(sb, i) +
2958 				ext4_bg_num_gdb(sb, i);
2959 			cond_resched();
2960 		}
2961 
2962 		/*
2963 		 * Every block group has an inode bitmap, a block
2964 		 * bitmap, and an inode table.
2965 		 */
2966 		overhead += ngroups * (2 + sbi->s_itb_per_group);
2967 		sbi->s_overhead_last = overhead;
2968 		smp_wmb();
2969 		sbi->s_blocks_last = ext4_blocks_count(es);
2970 	}
2971 
2972 	buf->f_type = EXT4_SUPER_MAGIC;
2973 	buf->f_bsize = sb->s_blocksize;
2974 	buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
2975 	buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2976 	ext4_free_blocks_count_set(es, buf->f_bfree);
2977 	buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
2978 	if (buf->f_bfree < ext4_r_blocks_count(es))
2979 		buf->f_bavail = 0;
2980 	buf->f_files = le32_to_cpu(es->s_inodes_count);
2981 	buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2982 	es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
2983 	buf->f_namelen = EXT4_NAME_LEN;
2984 	fsid = le64_to_cpup((void *)es->s_uuid) ^
2985 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2986 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2987 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2988 	return 0;
2989 }
2990 
2991 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2992  * is locked for write. Otherwise the are possible deadlocks:
2993  * Process 1                         Process 2
2994  * ext4_create()                     quota_sync()
2995  *   jbd2_journal_start()                   write_dquot()
2996  *   DQUOT_INIT()                        down(dqio_mutex)
2997  *     down(dqio_mutex)                    jbd2_journal_start()
2998  *
2999  */
3000 
3001 #ifdef CONFIG_QUOTA
3002 
3003 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3004 {
3005 	return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3006 }
3007 
3008 static int ext4_dquot_initialize(struct inode *inode, int type)
3009 {
3010 	handle_t *handle;
3011 	int ret, err;
3012 
3013 	/* We may create quota structure so we need to reserve enough blocks */
3014 	handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
3015 	if (IS_ERR(handle))
3016 		return PTR_ERR(handle);
3017 	ret = dquot_initialize(inode, type);
3018 	err = ext4_journal_stop(handle);
3019 	if (!ret)
3020 		ret = err;
3021 	return ret;
3022 }
3023 
3024 static int ext4_dquot_drop(struct inode *inode)
3025 {
3026 	handle_t *handle;
3027 	int ret, err;
3028 
3029 	/* We may delete quota structure so we need to reserve enough blocks */
3030 	handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
3031 	if (IS_ERR(handle))
3032 		return PTR_ERR(handle);
3033 	ret = dquot_drop(inode);
3034 	err = ext4_journal_stop(handle);
3035 	if (!ret)
3036 		ret = err;
3037 	return ret;
3038 }
3039 
3040 static int ext4_write_dquot(struct dquot *dquot)
3041 {
3042 	int ret, err;
3043 	handle_t *handle;
3044 	struct inode *inode;
3045 
3046 	inode = dquot_to_inode(dquot);
3047 	handle = ext4_journal_start(inode,
3048 					EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3049 	if (IS_ERR(handle))
3050 		return PTR_ERR(handle);
3051 	ret = dquot_commit(dquot);
3052 	err = ext4_journal_stop(handle);
3053 	if (!ret)
3054 		ret = err;
3055 	return ret;
3056 }
3057 
3058 static int ext4_acquire_dquot(struct dquot *dquot)
3059 {
3060 	int ret, err;
3061 	handle_t *handle;
3062 
3063 	handle = ext4_journal_start(dquot_to_inode(dquot),
3064 					EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3065 	if (IS_ERR(handle))
3066 		return PTR_ERR(handle);
3067 	ret = dquot_acquire(dquot);
3068 	err = ext4_journal_stop(handle);
3069 	if (!ret)
3070 		ret = err;
3071 	return ret;
3072 }
3073 
3074 static int ext4_release_dquot(struct dquot *dquot)
3075 {
3076 	int ret, err;
3077 	handle_t *handle;
3078 
3079 	handle = ext4_journal_start(dquot_to_inode(dquot),
3080 					EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3081 	if (IS_ERR(handle)) {
3082 		/* Release dquot anyway to avoid endless cycle in dqput() */
3083 		dquot_release(dquot);
3084 		return PTR_ERR(handle);
3085 	}
3086 	ret = dquot_release(dquot);
3087 	err = ext4_journal_stop(handle);
3088 	if (!ret)
3089 		ret = err;
3090 	return ret;
3091 }
3092 
3093 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3094 {
3095 	/* Are we journalling quotas? */
3096 	if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3097 	    EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3098 		dquot_mark_dquot_dirty(dquot);
3099 		return ext4_write_dquot(dquot);
3100 	} else {
3101 		return dquot_mark_dquot_dirty(dquot);
3102 	}
3103 }
3104 
3105 static int ext4_write_info(struct super_block *sb, int type)
3106 {
3107 	int ret, err;
3108 	handle_t *handle;
3109 
3110 	/* Data block + inode block */
3111 	handle = ext4_journal_start(sb->s_root->d_inode, 2);
3112 	if (IS_ERR(handle))
3113 		return PTR_ERR(handle);
3114 	ret = dquot_commit_info(sb, type);
3115 	err = ext4_journal_stop(handle);
3116 	if (!ret)
3117 		ret = err;
3118 	return ret;
3119 }
3120 
3121 /*
3122  * Turn on quotas during mount time - we need to find
3123  * the quota file and such...
3124  */
3125 static int ext4_quota_on_mount(struct super_block *sb, int type)
3126 {
3127 	return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3128 			EXT4_SB(sb)->s_jquota_fmt, type);
3129 }
3130 
3131 /*
3132  * Standard function to be called on quota_on
3133  */
3134 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3135 			 char *path)
3136 {
3137 	int err;
3138 	struct nameidata nd;
3139 
3140 	if (!test_opt(sb, QUOTA))
3141 		return -EINVAL;
3142 	/* Not journalling quota? */
3143 	if (!EXT4_SB(sb)->s_qf_names[USRQUOTA] &&
3144 	    !EXT4_SB(sb)->s_qf_names[GRPQUOTA])
3145 		return vfs_quota_on(sb, type, format_id, path);
3146 	err = path_lookup(path, LOOKUP_FOLLOW, &nd);
3147 	if (err)
3148 		return err;
3149 	/* Quotafile not on the same filesystem? */
3150 	if (nd.mnt->mnt_sb != sb) {
3151 		path_release(&nd);
3152 		return -EXDEV;
3153 	}
3154 	/* Quotafile not of fs root? */
3155 	if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
3156 		printk(KERN_WARNING
3157 			"EXT4-fs: Quota file not on filesystem root. "
3158 			"Journalled quota will not work.\n");
3159 	path_release(&nd);
3160 	return vfs_quota_on(sb, type, format_id, path);
3161 }
3162 
3163 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3164  * acquiring the locks... As quota files are never truncated and quota code
3165  * itself serializes the operations (and noone else should touch the files)
3166  * we don't have to be afraid of races */
3167 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3168 			       size_t len, loff_t off)
3169 {
3170 	struct inode *inode = sb_dqopt(sb)->files[type];
3171 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3172 	int err = 0;
3173 	int offset = off & (sb->s_blocksize - 1);
3174 	int tocopy;
3175 	size_t toread;
3176 	struct buffer_head *bh;
3177 	loff_t i_size = i_size_read(inode);
3178 
3179 	if (off > i_size)
3180 		return 0;
3181 	if (off+len > i_size)
3182 		len = i_size-off;
3183 	toread = len;
3184 	while (toread > 0) {
3185 		tocopy = sb->s_blocksize - offset < toread ?
3186 				sb->s_blocksize - offset : toread;
3187 		bh = ext4_bread(NULL, inode, blk, 0, &err);
3188 		if (err)
3189 			return err;
3190 		if (!bh)	/* A hole? */
3191 			memset(data, 0, tocopy);
3192 		else
3193 			memcpy(data, bh->b_data+offset, tocopy);
3194 		brelse(bh);
3195 		offset = 0;
3196 		toread -= tocopy;
3197 		data += tocopy;
3198 		blk++;
3199 	}
3200 	return len;
3201 }
3202 
3203 /* Write to quotafile (we know the transaction is already started and has
3204  * enough credits) */
3205 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3206 				const char *data, size_t len, loff_t off)
3207 {
3208 	struct inode *inode = sb_dqopt(sb)->files[type];
3209 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3210 	int err = 0;
3211 	int offset = off & (sb->s_blocksize - 1);
3212 	int tocopy;
3213 	int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3214 	size_t towrite = len;
3215 	struct buffer_head *bh;
3216 	handle_t *handle = journal_current_handle();
3217 
3218 	if (!handle) {
3219 		printk(KERN_WARNING "EXT4-fs: Quota write (off=%Lu, len=%Lu)"
3220 			" cancelled because transaction is not started.\n",
3221 			(unsigned long long)off, (unsigned long long)len);
3222 		return -EIO;
3223 	}
3224 	mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3225 	while (towrite > 0) {
3226 		tocopy = sb->s_blocksize - offset < towrite ?
3227 				sb->s_blocksize - offset : towrite;
3228 		bh = ext4_bread(handle, inode, blk, 1, &err);
3229 		if (!bh)
3230 			goto out;
3231 		if (journal_quota) {
3232 			err = ext4_journal_get_write_access(handle, bh);
3233 			if (err) {
3234 				brelse(bh);
3235 				goto out;
3236 			}
3237 		}
3238 		lock_buffer(bh);
3239 		memcpy(bh->b_data+offset, data, tocopy);
3240 		flush_dcache_page(bh->b_page);
3241 		unlock_buffer(bh);
3242 		if (journal_quota)
3243 			err = ext4_journal_dirty_metadata(handle, bh);
3244 		else {
3245 			/* Always do at least ordered writes for quotas */
3246 			err = ext4_journal_dirty_data(handle, bh);
3247 			mark_buffer_dirty(bh);
3248 		}
3249 		brelse(bh);
3250 		if (err)
3251 			goto out;
3252 		offset = 0;
3253 		towrite -= tocopy;
3254 		data += tocopy;
3255 		blk++;
3256 	}
3257 out:
3258 	if (len == towrite)
3259 		return err;
3260 	if (inode->i_size < off+len-towrite) {
3261 		i_size_write(inode, off+len-towrite);
3262 		EXT4_I(inode)->i_disksize = inode->i_size;
3263 	}
3264 	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3265 	ext4_mark_inode_dirty(handle, inode);
3266 	mutex_unlock(&inode->i_mutex);
3267 	return len - towrite;
3268 }
3269 
3270 #endif
3271 
3272 static int ext4_get_sb(struct file_system_type *fs_type,
3273 	int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3274 {
3275 	return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3276 }
3277 
3278 static struct file_system_type ext4dev_fs_type = {
3279 	.owner		= THIS_MODULE,
3280 	.name		= "ext4dev",
3281 	.get_sb		= ext4_get_sb,
3282 	.kill_sb	= kill_block_super,
3283 	.fs_flags	= FS_REQUIRES_DEV,
3284 };
3285 
3286 static int __init init_ext4_fs(void)
3287 {
3288 	int err;
3289 
3290 	err = init_ext4_mballoc();
3291 	if (err)
3292 		return err;
3293 
3294 	err = init_ext4_xattr();
3295 	if (err)
3296 		goto out2;
3297 	err = init_inodecache();
3298 	if (err)
3299 		goto out1;
3300 	err = register_filesystem(&ext4dev_fs_type);
3301 	if (err)
3302 		goto out;
3303 	return 0;
3304 out:
3305 	destroy_inodecache();
3306 out1:
3307 	exit_ext4_xattr();
3308 out2:
3309 	exit_ext4_mballoc();
3310 	return err;
3311 }
3312 
3313 static void __exit exit_ext4_fs(void)
3314 {
3315 	unregister_filesystem(&ext4dev_fs_type);
3316 	destroy_inodecache();
3317 	exit_ext4_xattr();
3318 	exit_ext4_mballoc();
3319 }
3320 
3321 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3322 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
3323 MODULE_LICENSE("GPL");
3324 module_init(init_ext4_fs)
3325 module_exit(exit_ext4_fs)
3326