xref: /linux/fs/ext4/super.c (revision 93d546399c2b7d66a54d5fbd5eee17de19246bf6)
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/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/parser.h>
28 #include <linux/smp_lock.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/marker.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <asm/uaccess.h>
42 
43 #include "ext4.h"
44 #include "ext4_jbd2.h"
45 #include "xattr.h"
46 #include "acl.h"
47 #include "namei.h"
48 #include "group.h"
49 
50 struct proc_dir_entry *ext4_proc_root;
51 
52 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
53 			     unsigned long journal_devnum);
54 static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
55 			       unsigned int);
56 static void ext4_commit_super(struct super_block *sb,
57 			      struct ext4_super_block *es, int sync);
58 static void ext4_mark_recovery_complete(struct super_block *sb,
59 					struct ext4_super_block *es);
60 static void ext4_clear_journal_err(struct super_block *sb,
61 				   struct ext4_super_block *es);
62 static int ext4_sync_fs(struct super_block *sb, int wait);
63 static const char *ext4_decode_error(struct super_block *sb, int errno,
64 				     char nbuf[16]);
65 static int ext4_remount(struct super_block *sb, int *flags, char *data);
66 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
67 static void ext4_unlockfs(struct super_block *sb);
68 static void ext4_write_super(struct super_block *sb);
69 static void ext4_write_super_lockfs(struct super_block *sb);
70 
71 
72 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
73 			       struct ext4_group_desc *bg)
74 {
75 	return le32_to_cpu(bg->bg_block_bitmap_lo) |
76 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
77 		(ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
78 }
79 
80 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
81 			       struct ext4_group_desc *bg)
82 {
83 	return le32_to_cpu(bg->bg_inode_bitmap_lo) |
84 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
85 		(ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
86 }
87 
88 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
89 			      struct ext4_group_desc *bg)
90 {
91 	return le32_to_cpu(bg->bg_inode_table_lo) |
92 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
93 		(ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
94 }
95 
96 void ext4_block_bitmap_set(struct super_block *sb,
97 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
98 {
99 	bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
100 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
101 		bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
102 }
103 
104 void ext4_inode_bitmap_set(struct super_block *sb,
105 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
106 {
107 	bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
108 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
109 		bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
110 }
111 
112 void ext4_inode_table_set(struct super_block *sb,
113 			  struct ext4_group_desc *bg, ext4_fsblk_t blk)
114 {
115 	bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
116 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
117 		bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
118 }
119 
120 /*
121  * Wrappers for jbd2_journal_start/end.
122  *
123  * The only special thing we need to do here is to make sure that all
124  * journal_end calls result in the superblock being marked dirty, so
125  * that sync() will call the filesystem's write_super callback if
126  * appropriate.
127  */
128 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
129 {
130 	journal_t *journal;
131 
132 	if (sb->s_flags & MS_RDONLY)
133 		return ERR_PTR(-EROFS);
134 
135 	/* Special case here: if the journal has aborted behind our
136 	 * backs (eg. EIO in the commit thread), then we still need to
137 	 * take the FS itself readonly cleanly. */
138 	journal = EXT4_SB(sb)->s_journal;
139 	if (is_journal_aborted(journal)) {
140 		ext4_abort(sb, __func__,
141 			   "Detected aborted journal");
142 		return ERR_PTR(-EROFS);
143 	}
144 
145 	return jbd2_journal_start(journal, nblocks);
146 }
147 
148 /*
149  * The only special thing we need to do here is to make sure that all
150  * jbd2_journal_stop calls result in the superblock being marked dirty, so
151  * that sync() will call the filesystem's write_super callback if
152  * appropriate.
153  */
154 int __ext4_journal_stop(const char *where, handle_t *handle)
155 {
156 	struct super_block *sb;
157 	int err;
158 	int rc;
159 
160 	sb = handle->h_transaction->t_journal->j_private;
161 	err = handle->h_err;
162 	rc = jbd2_journal_stop(handle);
163 
164 	if (!err)
165 		err = rc;
166 	if (err)
167 		__ext4_std_error(sb, where, err);
168 	return err;
169 }
170 
171 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
172 		struct buffer_head *bh, handle_t *handle, int err)
173 {
174 	char nbuf[16];
175 	const char *errstr = ext4_decode_error(NULL, err, nbuf);
176 
177 	if (bh)
178 		BUFFER_TRACE(bh, "abort");
179 
180 	if (!handle->h_err)
181 		handle->h_err = err;
182 
183 	if (is_handle_aborted(handle))
184 		return;
185 
186 	printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
187 	       caller, errstr, err_fn);
188 
189 	jbd2_journal_abort_handle(handle);
190 }
191 
192 /* Deal with the reporting of failure conditions on a filesystem such as
193  * inconsistencies detected or read IO failures.
194  *
195  * On ext2, we can store the error state of the filesystem in the
196  * superblock.  That is not possible on ext4, because we may have other
197  * write ordering constraints on the superblock which prevent us from
198  * writing it out straight away; and given that the journal is about to
199  * be aborted, we can't rely on the current, or future, transactions to
200  * write out the superblock safely.
201  *
202  * We'll just use the jbd2_journal_abort() error code to record an error in
203  * the journal instead.  On recovery, the journal will compain about
204  * that error until we've noted it down and cleared it.
205  */
206 
207 static void ext4_handle_error(struct super_block *sb)
208 {
209 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
210 
211 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
212 	es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
213 
214 	if (sb->s_flags & MS_RDONLY)
215 		return;
216 
217 	if (!test_opt(sb, ERRORS_CONT)) {
218 		journal_t *journal = EXT4_SB(sb)->s_journal;
219 
220 		EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
221 		if (journal)
222 			jbd2_journal_abort(journal, -EIO);
223 	}
224 	if (test_opt(sb, ERRORS_RO)) {
225 		printk(KERN_CRIT "Remounting filesystem read-only\n");
226 		sb->s_flags |= MS_RDONLY;
227 	}
228 	ext4_commit_super(sb, es, 1);
229 	if (test_opt(sb, ERRORS_PANIC))
230 		panic("EXT4-fs (device %s): panic forced after error\n",
231 			sb->s_id);
232 }
233 
234 void ext4_error(struct super_block *sb, const char *function,
235 		const char *fmt, ...)
236 {
237 	va_list args;
238 
239 	va_start(args, fmt);
240 	printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
241 	vprintk(fmt, args);
242 	printk("\n");
243 	va_end(args);
244 
245 	ext4_handle_error(sb);
246 }
247 
248 static const char *ext4_decode_error(struct super_block *sb, int errno,
249 				     char nbuf[16])
250 {
251 	char *errstr = NULL;
252 
253 	switch (errno) {
254 	case -EIO:
255 		errstr = "IO failure";
256 		break;
257 	case -ENOMEM:
258 		errstr = "Out of memory";
259 		break;
260 	case -EROFS:
261 		if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
262 			errstr = "Journal has aborted";
263 		else
264 			errstr = "Readonly filesystem";
265 		break;
266 	default:
267 		/* If the caller passed in an extra buffer for unknown
268 		 * errors, textualise them now.  Else we just return
269 		 * NULL. */
270 		if (nbuf) {
271 			/* Check for truncated error codes... */
272 			if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
273 				errstr = nbuf;
274 		}
275 		break;
276 	}
277 
278 	return errstr;
279 }
280 
281 /* __ext4_std_error decodes expected errors from journaling functions
282  * automatically and invokes the appropriate error response.  */
283 
284 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
285 {
286 	char nbuf[16];
287 	const char *errstr;
288 
289 	/* Special case: if the error is EROFS, and we're not already
290 	 * inside a transaction, then there's really no point in logging
291 	 * an error. */
292 	if (errno == -EROFS && journal_current_handle() == NULL &&
293 	    (sb->s_flags & MS_RDONLY))
294 		return;
295 
296 	errstr = ext4_decode_error(sb, errno, nbuf);
297 	printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
298 	       sb->s_id, function, errstr);
299 
300 	ext4_handle_error(sb);
301 }
302 
303 /*
304  * ext4_abort is a much stronger failure handler than ext4_error.  The
305  * abort function may be used to deal with unrecoverable failures such
306  * as journal IO errors or ENOMEM at a critical moment in log management.
307  *
308  * We unconditionally force the filesystem into an ABORT|READONLY state,
309  * unless the error response on the fs has been set to panic in which
310  * case we take the easy way out and panic immediately.
311  */
312 
313 void ext4_abort(struct super_block *sb, const char *function,
314 		const char *fmt, ...)
315 {
316 	va_list args;
317 
318 	printk(KERN_CRIT "ext4_abort called.\n");
319 
320 	va_start(args, fmt);
321 	printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
322 	vprintk(fmt, args);
323 	printk("\n");
324 	va_end(args);
325 
326 	if (test_opt(sb, ERRORS_PANIC))
327 		panic("EXT4-fs panic from previous error\n");
328 
329 	if (sb->s_flags & MS_RDONLY)
330 		return;
331 
332 	printk(KERN_CRIT "Remounting filesystem read-only\n");
333 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
334 	sb->s_flags |= MS_RDONLY;
335 	EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
336 	if (EXT4_SB(sb)->s_journal)
337 		jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
338 }
339 
340 void ext4_warning(struct super_block *sb, const char *function,
341 		  const char *fmt, ...)
342 {
343 	va_list args;
344 
345 	va_start(args, fmt);
346 	printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
347 	       sb->s_id, function);
348 	vprintk(fmt, args);
349 	printk("\n");
350 	va_end(args);
351 }
352 
353 void ext4_update_dynamic_rev(struct super_block *sb)
354 {
355 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
356 
357 	if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
358 		return;
359 
360 	ext4_warning(sb, __func__,
361 		     "updating to rev %d because of new feature flag, "
362 		     "running e2fsck is recommended",
363 		     EXT4_DYNAMIC_REV);
364 
365 	es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
366 	es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
367 	es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
368 	/* leave es->s_feature_*compat flags alone */
369 	/* es->s_uuid will be set by e2fsck if empty */
370 
371 	/*
372 	 * The rest of the superblock fields should be zero, and if not it
373 	 * means they are likely already in use, so leave them alone.  We
374 	 * can leave it up to e2fsck to clean up any inconsistencies there.
375 	 */
376 }
377 
378 /*
379  * Open the external journal device
380  */
381 static struct block_device *ext4_blkdev_get(dev_t dev)
382 {
383 	struct block_device *bdev;
384 	char b[BDEVNAME_SIZE];
385 
386 	bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
387 	if (IS_ERR(bdev))
388 		goto fail;
389 	return bdev;
390 
391 fail:
392 	printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
393 			__bdevname(dev, b), PTR_ERR(bdev));
394 	return NULL;
395 }
396 
397 /*
398  * Release the journal device
399  */
400 static int ext4_blkdev_put(struct block_device *bdev)
401 {
402 	bd_release(bdev);
403 	return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
404 }
405 
406 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
407 {
408 	struct block_device *bdev;
409 	int ret = -ENODEV;
410 
411 	bdev = sbi->journal_bdev;
412 	if (bdev) {
413 		ret = ext4_blkdev_put(bdev);
414 		sbi->journal_bdev = NULL;
415 	}
416 	return ret;
417 }
418 
419 static inline struct inode *orphan_list_entry(struct list_head *l)
420 {
421 	return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
422 }
423 
424 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
425 {
426 	struct list_head *l;
427 
428 	printk(KERN_ERR "sb orphan head is %d\n",
429 	       le32_to_cpu(sbi->s_es->s_last_orphan));
430 
431 	printk(KERN_ERR "sb_info orphan list:\n");
432 	list_for_each(l, &sbi->s_orphan) {
433 		struct inode *inode = orphan_list_entry(l);
434 		printk(KERN_ERR "  "
435 		       "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
436 		       inode->i_sb->s_id, inode->i_ino, inode,
437 		       inode->i_mode, inode->i_nlink,
438 		       NEXT_ORPHAN(inode));
439 	}
440 }
441 
442 static void ext4_put_super(struct super_block *sb)
443 {
444 	struct ext4_sb_info *sbi = EXT4_SB(sb);
445 	struct ext4_super_block *es = sbi->s_es;
446 	int i, err;
447 
448 	ext4_mb_release(sb);
449 	ext4_ext_release(sb);
450 	ext4_xattr_put_super(sb);
451 	err = jbd2_journal_destroy(sbi->s_journal);
452 	sbi->s_journal = NULL;
453 	if (err < 0)
454 		ext4_abort(sb, __func__, "Couldn't clean up the journal");
455 
456 	if (!(sb->s_flags & MS_RDONLY)) {
457 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
458 		es->s_state = cpu_to_le16(sbi->s_mount_state);
459 		ext4_commit_super(sb, es, 1);
460 	}
461 	if (sbi->s_proc) {
462 		remove_proc_entry("inode_readahead_blks", sbi->s_proc);
463 		remove_proc_entry(sb->s_id, ext4_proc_root);
464 	}
465 
466 	for (i = 0; i < sbi->s_gdb_count; i++)
467 		brelse(sbi->s_group_desc[i]);
468 	kfree(sbi->s_group_desc);
469 	kfree(sbi->s_flex_groups);
470 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
471 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
472 	percpu_counter_destroy(&sbi->s_dirs_counter);
473 	percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
474 	brelse(sbi->s_sbh);
475 #ifdef CONFIG_QUOTA
476 	for (i = 0; i < MAXQUOTAS; i++)
477 		kfree(sbi->s_qf_names[i]);
478 #endif
479 
480 	/* Debugging code just in case the in-memory inode orphan list
481 	 * isn't empty.  The on-disk one can be non-empty if we've
482 	 * detected an error and taken the fs readonly, but the
483 	 * in-memory list had better be clean by this point. */
484 	if (!list_empty(&sbi->s_orphan))
485 		dump_orphan_list(sb, sbi);
486 	J_ASSERT(list_empty(&sbi->s_orphan));
487 
488 	invalidate_bdev(sb->s_bdev);
489 	if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
490 		/*
491 		 * Invalidate the journal device's buffers.  We don't want them
492 		 * floating about in memory - the physical journal device may
493 		 * hotswapped, and it breaks the `ro-after' testing code.
494 		 */
495 		sync_blockdev(sbi->journal_bdev);
496 		invalidate_bdev(sbi->journal_bdev);
497 		ext4_blkdev_remove(sbi);
498 	}
499 	sb->s_fs_info = NULL;
500 	kfree(sbi);
501 	return;
502 }
503 
504 static struct kmem_cache *ext4_inode_cachep;
505 
506 /*
507  * Called inside transaction, so use GFP_NOFS
508  */
509 static struct inode *ext4_alloc_inode(struct super_block *sb)
510 {
511 	struct ext4_inode_info *ei;
512 
513 	ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
514 	if (!ei)
515 		return NULL;
516 #ifdef CONFIG_EXT4_FS_POSIX_ACL
517 	ei->i_acl = EXT4_ACL_NOT_CACHED;
518 	ei->i_default_acl = EXT4_ACL_NOT_CACHED;
519 #endif
520 	ei->vfs_inode.i_version = 1;
521 	ei->vfs_inode.i_data.writeback_index = 0;
522 	memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
523 	INIT_LIST_HEAD(&ei->i_prealloc_list);
524 	spin_lock_init(&ei->i_prealloc_lock);
525 	jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
526 	ei->i_reserved_data_blocks = 0;
527 	ei->i_reserved_meta_blocks = 0;
528 	ei->i_allocated_meta_blocks = 0;
529 	ei->i_delalloc_reserved_flag = 0;
530 	spin_lock_init(&(ei->i_block_reservation_lock));
531 	return &ei->vfs_inode;
532 }
533 
534 static void ext4_destroy_inode(struct inode *inode)
535 {
536 	if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
537 		printk("EXT4 Inode %p: orphan list check failed!\n",
538 			EXT4_I(inode));
539 		print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
540 				EXT4_I(inode), sizeof(struct ext4_inode_info),
541 				true);
542 		dump_stack();
543 	}
544 	kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
545 }
546 
547 static void init_once(void *foo)
548 {
549 	struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
550 
551 	INIT_LIST_HEAD(&ei->i_orphan);
552 #ifdef CONFIG_EXT4_FS_XATTR
553 	init_rwsem(&ei->xattr_sem);
554 #endif
555 	init_rwsem(&ei->i_data_sem);
556 	inode_init_once(&ei->vfs_inode);
557 }
558 
559 static int init_inodecache(void)
560 {
561 	ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
562 					     sizeof(struct ext4_inode_info),
563 					     0, (SLAB_RECLAIM_ACCOUNT|
564 						SLAB_MEM_SPREAD),
565 					     init_once);
566 	if (ext4_inode_cachep == NULL)
567 		return -ENOMEM;
568 	return 0;
569 }
570 
571 static void destroy_inodecache(void)
572 {
573 	kmem_cache_destroy(ext4_inode_cachep);
574 }
575 
576 static void ext4_clear_inode(struct inode *inode)
577 {
578 #ifdef CONFIG_EXT4_FS_POSIX_ACL
579 	if (EXT4_I(inode)->i_acl &&
580 			EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
581 		posix_acl_release(EXT4_I(inode)->i_acl);
582 		EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
583 	}
584 	if (EXT4_I(inode)->i_default_acl &&
585 			EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
586 		posix_acl_release(EXT4_I(inode)->i_default_acl);
587 		EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
588 	}
589 #endif
590 	ext4_discard_preallocations(inode);
591 	jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
592 				       &EXT4_I(inode)->jinode);
593 }
594 
595 static inline void ext4_show_quota_options(struct seq_file *seq,
596 					   struct super_block *sb)
597 {
598 #if defined(CONFIG_QUOTA)
599 	struct ext4_sb_info *sbi = EXT4_SB(sb);
600 
601 	if (sbi->s_jquota_fmt)
602 		seq_printf(seq, ",jqfmt=%s",
603 		(sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
604 
605 	if (sbi->s_qf_names[USRQUOTA])
606 		seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
607 
608 	if (sbi->s_qf_names[GRPQUOTA])
609 		seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
610 
611 	if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
612 		seq_puts(seq, ",usrquota");
613 
614 	if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
615 		seq_puts(seq, ",grpquota");
616 #endif
617 }
618 
619 /*
620  * Show an option if
621  *  - it's set to a non-default value OR
622  *  - if the per-sb default is different from the global default
623  */
624 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
625 {
626 	int def_errors;
627 	unsigned long def_mount_opts;
628 	struct super_block *sb = vfs->mnt_sb;
629 	struct ext4_sb_info *sbi = EXT4_SB(sb);
630 	struct ext4_super_block *es = sbi->s_es;
631 
632 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
633 	def_errors     = le16_to_cpu(es->s_errors);
634 
635 	if (sbi->s_sb_block != 1)
636 		seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
637 	if (test_opt(sb, MINIX_DF))
638 		seq_puts(seq, ",minixdf");
639 	if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
640 		seq_puts(seq, ",grpid");
641 	if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
642 		seq_puts(seq, ",nogrpid");
643 	if (sbi->s_resuid != EXT4_DEF_RESUID ||
644 	    le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
645 		seq_printf(seq, ",resuid=%u", sbi->s_resuid);
646 	}
647 	if (sbi->s_resgid != EXT4_DEF_RESGID ||
648 	    le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
649 		seq_printf(seq, ",resgid=%u", sbi->s_resgid);
650 	}
651 	if (test_opt(sb, ERRORS_RO)) {
652 		if (def_errors == EXT4_ERRORS_PANIC ||
653 		    def_errors == EXT4_ERRORS_CONTINUE) {
654 			seq_puts(seq, ",errors=remount-ro");
655 		}
656 	}
657 	if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
658 		seq_puts(seq, ",errors=continue");
659 	if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
660 		seq_puts(seq, ",errors=panic");
661 	if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
662 		seq_puts(seq, ",nouid32");
663 	if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
664 		seq_puts(seq, ",debug");
665 	if (test_opt(sb, OLDALLOC))
666 		seq_puts(seq, ",oldalloc");
667 #ifdef CONFIG_EXT4_FS_XATTR
668 	if (test_opt(sb, XATTR_USER) &&
669 		!(def_mount_opts & EXT4_DEFM_XATTR_USER))
670 		seq_puts(seq, ",user_xattr");
671 	if (!test_opt(sb, XATTR_USER) &&
672 	    (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
673 		seq_puts(seq, ",nouser_xattr");
674 	}
675 #endif
676 #ifdef CONFIG_EXT4_FS_POSIX_ACL
677 	if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
678 		seq_puts(seq, ",acl");
679 	if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
680 		seq_puts(seq, ",noacl");
681 #endif
682 	if (!test_opt(sb, RESERVATION))
683 		seq_puts(seq, ",noreservation");
684 	if (sbi->s_commit_interval) {
685 		seq_printf(seq, ",commit=%u",
686 			   (unsigned) (sbi->s_commit_interval / HZ));
687 	}
688 	/*
689 	 * We're changing the default of barrier mount option, so
690 	 * let's always display its mount state so it's clear what its
691 	 * status is.
692 	 */
693 	seq_puts(seq, ",barrier=");
694 	seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
695 	if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
696 		seq_puts(seq, ",journal_async_commit");
697 	if (test_opt(sb, NOBH))
698 		seq_puts(seq, ",nobh");
699 	if (!test_opt(sb, EXTENTS))
700 		seq_puts(seq, ",noextents");
701 	if (test_opt(sb, I_VERSION))
702 		seq_puts(seq, ",i_version");
703 	if (!test_opt(sb, DELALLOC))
704 		seq_puts(seq, ",nodelalloc");
705 
706 
707 	if (sbi->s_stripe)
708 		seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
709 	/*
710 	 * journal mode get enabled in different ways
711 	 * So just print the value even if we didn't specify it
712 	 */
713 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
714 		seq_puts(seq, ",data=journal");
715 	else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
716 		seq_puts(seq, ",data=ordered");
717 	else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
718 		seq_puts(seq, ",data=writeback");
719 
720 	if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
721 		seq_printf(seq, ",inode_readahead_blks=%u",
722 			   sbi->s_inode_readahead_blks);
723 
724 	if (test_opt(sb, DATA_ERR_ABORT))
725 		seq_puts(seq, ",data_err=abort");
726 
727 	ext4_show_quota_options(seq, sb);
728 	return 0;
729 }
730 
731 
732 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
733 		u64 ino, u32 generation)
734 {
735 	struct inode *inode;
736 
737 	if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
738 		return ERR_PTR(-ESTALE);
739 	if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
740 		return ERR_PTR(-ESTALE);
741 
742 	/* iget isn't really right if the inode is currently unallocated!!
743 	 *
744 	 * ext4_read_inode will return a bad_inode if the inode had been
745 	 * deleted, so we should be safe.
746 	 *
747 	 * Currently we don't know the generation for parent directory, so
748 	 * a generation of 0 means "accept any"
749 	 */
750 	inode = ext4_iget(sb, ino);
751 	if (IS_ERR(inode))
752 		return ERR_CAST(inode);
753 	if (generation && inode->i_generation != generation) {
754 		iput(inode);
755 		return ERR_PTR(-ESTALE);
756 	}
757 
758 	return inode;
759 }
760 
761 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
762 		int fh_len, int fh_type)
763 {
764 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
765 				    ext4_nfs_get_inode);
766 }
767 
768 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
769 		int fh_len, int fh_type)
770 {
771 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
772 				    ext4_nfs_get_inode);
773 }
774 
775 #ifdef CONFIG_QUOTA
776 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
777 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
778 
779 static int ext4_dquot_initialize(struct inode *inode, int type);
780 static int ext4_dquot_drop(struct inode *inode);
781 static int ext4_write_dquot(struct dquot *dquot);
782 static int ext4_acquire_dquot(struct dquot *dquot);
783 static int ext4_release_dquot(struct dquot *dquot);
784 static int ext4_mark_dquot_dirty(struct dquot *dquot);
785 static int ext4_write_info(struct super_block *sb, int type);
786 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
787 				char *path, int remount);
788 static int ext4_quota_on_mount(struct super_block *sb, int type);
789 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
790 			       size_t len, loff_t off);
791 static ssize_t ext4_quota_write(struct super_block *sb, int type,
792 				const char *data, size_t len, loff_t off);
793 
794 static struct dquot_operations ext4_quota_operations = {
795 	.initialize	= ext4_dquot_initialize,
796 	.drop		= ext4_dquot_drop,
797 	.alloc_space	= dquot_alloc_space,
798 	.alloc_inode	= dquot_alloc_inode,
799 	.free_space	= dquot_free_space,
800 	.free_inode	= dquot_free_inode,
801 	.transfer	= dquot_transfer,
802 	.write_dquot	= ext4_write_dquot,
803 	.acquire_dquot	= ext4_acquire_dquot,
804 	.release_dquot	= ext4_release_dquot,
805 	.mark_dirty	= ext4_mark_dquot_dirty,
806 	.write_info	= ext4_write_info
807 };
808 
809 static struct quotactl_ops ext4_qctl_operations = {
810 	.quota_on	= ext4_quota_on,
811 	.quota_off	= vfs_quota_off,
812 	.quota_sync	= vfs_quota_sync,
813 	.get_info	= vfs_get_dqinfo,
814 	.set_info	= vfs_set_dqinfo,
815 	.get_dqblk	= vfs_get_dqblk,
816 	.set_dqblk	= vfs_set_dqblk
817 };
818 #endif
819 
820 static const struct super_operations ext4_sops = {
821 	.alloc_inode	= ext4_alloc_inode,
822 	.destroy_inode	= ext4_destroy_inode,
823 	.write_inode	= ext4_write_inode,
824 	.dirty_inode	= ext4_dirty_inode,
825 	.delete_inode	= ext4_delete_inode,
826 	.put_super	= ext4_put_super,
827 	.write_super	= ext4_write_super,
828 	.sync_fs	= ext4_sync_fs,
829 	.write_super_lockfs = ext4_write_super_lockfs,
830 	.unlockfs	= ext4_unlockfs,
831 	.statfs		= ext4_statfs,
832 	.remount_fs	= ext4_remount,
833 	.clear_inode	= ext4_clear_inode,
834 	.show_options	= ext4_show_options,
835 #ifdef CONFIG_QUOTA
836 	.quota_read	= ext4_quota_read,
837 	.quota_write	= ext4_quota_write,
838 #endif
839 };
840 
841 static const struct export_operations ext4_export_ops = {
842 	.fh_to_dentry = ext4_fh_to_dentry,
843 	.fh_to_parent = ext4_fh_to_parent,
844 	.get_parent = ext4_get_parent,
845 };
846 
847 enum {
848 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
849 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
850 	Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
851 	Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
852 	Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
853 	Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
854 	Opt_journal_checksum, Opt_journal_async_commit,
855 	Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
856 	Opt_data_err_abort, Opt_data_err_ignore,
857 	Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
858 	Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
859 	Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
860 	Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
861 	Opt_stripe, Opt_delalloc, Opt_nodelalloc,
862 	Opt_inode_readahead_blks
863 };
864 
865 static const match_table_t tokens = {
866 	{Opt_bsd_df, "bsddf"},
867 	{Opt_minix_df, "minixdf"},
868 	{Opt_grpid, "grpid"},
869 	{Opt_grpid, "bsdgroups"},
870 	{Opt_nogrpid, "nogrpid"},
871 	{Opt_nogrpid, "sysvgroups"},
872 	{Opt_resgid, "resgid=%u"},
873 	{Opt_resuid, "resuid=%u"},
874 	{Opt_sb, "sb=%u"},
875 	{Opt_err_cont, "errors=continue"},
876 	{Opt_err_panic, "errors=panic"},
877 	{Opt_err_ro, "errors=remount-ro"},
878 	{Opt_nouid32, "nouid32"},
879 	{Opt_debug, "debug"},
880 	{Opt_oldalloc, "oldalloc"},
881 	{Opt_orlov, "orlov"},
882 	{Opt_user_xattr, "user_xattr"},
883 	{Opt_nouser_xattr, "nouser_xattr"},
884 	{Opt_acl, "acl"},
885 	{Opt_noacl, "noacl"},
886 	{Opt_reservation, "reservation"},
887 	{Opt_noreservation, "noreservation"},
888 	{Opt_noload, "noload"},
889 	{Opt_nobh, "nobh"},
890 	{Opt_bh, "bh"},
891 	{Opt_commit, "commit=%u"},
892 	{Opt_journal_update, "journal=update"},
893 	{Opt_journal_inum, "journal=%u"},
894 	{Opt_journal_dev, "journal_dev=%u"},
895 	{Opt_journal_checksum, "journal_checksum"},
896 	{Opt_journal_async_commit, "journal_async_commit"},
897 	{Opt_abort, "abort"},
898 	{Opt_data_journal, "data=journal"},
899 	{Opt_data_ordered, "data=ordered"},
900 	{Opt_data_writeback, "data=writeback"},
901 	{Opt_data_err_abort, "data_err=abort"},
902 	{Opt_data_err_ignore, "data_err=ignore"},
903 	{Opt_offusrjquota, "usrjquota="},
904 	{Opt_usrjquota, "usrjquota=%s"},
905 	{Opt_offgrpjquota, "grpjquota="},
906 	{Opt_grpjquota, "grpjquota=%s"},
907 	{Opt_jqfmt_vfsold, "jqfmt=vfsold"},
908 	{Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
909 	{Opt_grpquota, "grpquota"},
910 	{Opt_noquota, "noquota"},
911 	{Opt_quota, "quota"},
912 	{Opt_usrquota, "usrquota"},
913 	{Opt_barrier, "barrier=%u"},
914 	{Opt_extents, "extents"},
915 	{Opt_noextents, "noextents"},
916 	{Opt_i_version, "i_version"},
917 	{Opt_stripe, "stripe=%u"},
918 	{Opt_resize, "resize"},
919 	{Opt_delalloc, "delalloc"},
920 	{Opt_nodelalloc, "nodelalloc"},
921 	{Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
922 	{Opt_err, NULL},
923 };
924 
925 static ext4_fsblk_t get_sb_block(void **data)
926 {
927 	ext4_fsblk_t	sb_block;
928 	char		*options = (char *) *data;
929 
930 	if (!options || strncmp(options, "sb=", 3) != 0)
931 		return 1;	/* Default location */
932 	options += 3;
933 	/*todo: use simple_strtoll with >32bit ext4 */
934 	sb_block = simple_strtoul(options, &options, 0);
935 	if (*options && *options != ',') {
936 		printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
937 		       (char *) *data);
938 		return 1;
939 	}
940 	if (*options == ',')
941 		options++;
942 	*data = (void *) options;
943 	return sb_block;
944 }
945 
946 static int parse_options(char *options, struct super_block *sb,
947 			 unsigned int *inum, unsigned long *journal_devnum,
948 			 ext4_fsblk_t *n_blocks_count, int is_remount)
949 {
950 	struct ext4_sb_info *sbi = EXT4_SB(sb);
951 	char *p;
952 	substring_t args[MAX_OPT_ARGS];
953 	int data_opt = 0;
954 	int option;
955 #ifdef CONFIG_QUOTA
956 	int qtype, qfmt;
957 	char *qname;
958 #endif
959 	ext4_fsblk_t last_block;
960 
961 	if (!options)
962 		return 1;
963 
964 	while ((p = strsep(&options, ",")) != NULL) {
965 		int token;
966 		if (!*p)
967 			continue;
968 
969 		token = match_token(p, tokens, args);
970 		switch (token) {
971 		case Opt_bsd_df:
972 			clear_opt(sbi->s_mount_opt, MINIX_DF);
973 			break;
974 		case Opt_minix_df:
975 			set_opt(sbi->s_mount_opt, MINIX_DF);
976 			break;
977 		case Opt_grpid:
978 			set_opt(sbi->s_mount_opt, GRPID);
979 			break;
980 		case Opt_nogrpid:
981 			clear_opt(sbi->s_mount_opt, GRPID);
982 			break;
983 		case Opt_resuid:
984 			if (match_int(&args[0], &option))
985 				return 0;
986 			sbi->s_resuid = option;
987 			break;
988 		case Opt_resgid:
989 			if (match_int(&args[0], &option))
990 				return 0;
991 			sbi->s_resgid = option;
992 			break;
993 		case Opt_sb:
994 			/* handled by get_sb_block() instead of here */
995 			/* *sb_block = match_int(&args[0]); */
996 			break;
997 		case Opt_err_panic:
998 			clear_opt(sbi->s_mount_opt, ERRORS_CONT);
999 			clear_opt(sbi->s_mount_opt, ERRORS_RO);
1000 			set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1001 			break;
1002 		case Opt_err_ro:
1003 			clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1004 			clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1005 			set_opt(sbi->s_mount_opt, ERRORS_RO);
1006 			break;
1007 		case Opt_err_cont:
1008 			clear_opt(sbi->s_mount_opt, ERRORS_RO);
1009 			clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1010 			set_opt(sbi->s_mount_opt, ERRORS_CONT);
1011 			break;
1012 		case Opt_nouid32:
1013 			set_opt(sbi->s_mount_opt, NO_UID32);
1014 			break;
1015 		case Opt_debug:
1016 			set_opt(sbi->s_mount_opt, DEBUG);
1017 			break;
1018 		case Opt_oldalloc:
1019 			set_opt(sbi->s_mount_opt, OLDALLOC);
1020 			break;
1021 		case Opt_orlov:
1022 			clear_opt(sbi->s_mount_opt, OLDALLOC);
1023 			break;
1024 #ifdef CONFIG_EXT4_FS_XATTR
1025 		case Opt_user_xattr:
1026 			set_opt(sbi->s_mount_opt, XATTR_USER);
1027 			break;
1028 		case Opt_nouser_xattr:
1029 			clear_opt(sbi->s_mount_opt, XATTR_USER);
1030 			break;
1031 #else
1032 		case Opt_user_xattr:
1033 		case Opt_nouser_xattr:
1034 			printk(KERN_ERR "EXT4 (no)user_xattr options "
1035 			       "not supported\n");
1036 			break;
1037 #endif
1038 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1039 		case Opt_acl:
1040 			set_opt(sbi->s_mount_opt, POSIX_ACL);
1041 			break;
1042 		case Opt_noacl:
1043 			clear_opt(sbi->s_mount_opt, POSIX_ACL);
1044 			break;
1045 #else
1046 		case Opt_acl:
1047 		case Opt_noacl:
1048 			printk(KERN_ERR "EXT4 (no)acl options "
1049 			       "not supported\n");
1050 			break;
1051 #endif
1052 		case Opt_reservation:
1053 			set_opt(sbi->s_mount_opt, RESERVATION);
1054 			break;
1055 		case Opt_noreservation:
1056 			clear_opt(sbi->s_mount_opt, RESERVATION);
1057 			break;
1058 		case Opt_journal_update:
1059 			/* @@@ FIXME */
1060 			/* Eventually we will want to be able to create
1061 			   a journal file here.  For now, only allow the
1062 			   user to specify an existing inode to be the
1063 			   journal file. */
1064 			if (is_remount) {
1065 				printk(KERN_ERR "EXT4-fs: cannot specify "
1066 				       "journal on remount\n");
1067 				return 0;
1068 			}
1069 			set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1070 			break;
1071 		case Opt_journal_inum:
1072 			if (is_remount) {
1073 				printk(KERN_ERR "EXT4-fs: cannot specify "
1074 				       "journal on remount\n");
1075 				return 0;
1076 			}
1077 			if (match_int(&args[0], &option))
1078 				return 0;
1079 			*inum = option;
1080 			break;
1081 		case Opt_journal_dev:
1082 			if (is_remount) {
1083 				printk(KERN_ERR "EXT4-fs: cannot specify "
1084 				       "journal on remount\n");
1085 				return 0;
1086 			}
1087 			if (match_int(&args[0], &option))
1088 				return 0;
1089 			*journal_devnum = option;
1090 			break;
1091 		case Opt_journal_checksum:
1092 			set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1093 			break;
1094 		case Opt_journal_async_commit:
1095 			set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1096 			set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1097 			break;
1098 		case Opt_noload:
1099 			set_opt(sbi->s_mount_opt, NOLOAD);
1100 			break;
1101 		case Opt_commit:
1102 			if (match_int(&args[0], &option))
1103 				return 0;
1104 			if (option < 0)
1105 				return 0;
1106 			if (option == 0)
1107 				option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1108 			sbi->s_commit_interval = HZ * option;
1109 			break;
1110 		case Opt_data_journal:
1111 			data_opt = EXT4_MOUNT_JOURNAL_DATA;
1112 			goto datacheck;
1113 		case Opt_data_ordered:
1114 			data_opt = EXT4_MOUNT_ORDERED_DATA;
1115 			goto datacheck;
1116 		case Opt_data_writeback:
1117 			data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1118 		datacheck:
1119 			if (is_remount) {
1120 				if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1121 						!= data_opt) {
1122 					printk(KERN_ERR
1123 						"EXT4-fs: cannot change data "
1124 						"mode on remount\n");
1125 					return 0;
1126 				}
1127 			} else {
1128 				sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1129 				sbi->s_mount_opt |= data_opt;
1130 			}
1131 			break;
1132 		case Opt_data_err_abort:
1133 			set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1134 			break;
1135 		case Opt_data_err_ignore:
1136 			clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1137 			break;
1138 #ifdef CONFIG_QUOTA
1139 		case Opt_usrjquota:
1140 			qtype = USRQUOTA;
1141 			goto set_qf_name;
1142 		case Opt_grpjquota:
1143 			qtype = GRPQUOTA;
1144 set_qf_name:
1145 			if ((sb_any_quota_enabled(sb) ||
1146 			     sb_any_quota_suspended(sb)) &&
1147 			    !sbi->s_qf_names[qtype]) {
1148 				printk(KERN_ERR
1149 				       "EXT4-fs: Cannot change journaled "
1150 				       "quota options when quota turned on.\n");
1151 				return 0;
1152 			}
1153 			qname = match_strdup(&args[0]);
1154 			if (!qname) {
1155 				printk(KERN_ERR
1156 					"EXT4-fs: not enough memory for "
1157 					"storing quotafile name.\n");
1158 				return 0;
1159 			}
1160 			if (sbi->s_qf_names[qtype] &&
1161 			    strcmp(sbi->s_qf_names[qtype], qname)) {
1162 				printk(KERN_ERR
1163 					"EXT4-fs: %s quota file already "
1164 					"specified.\n", QTYPE2NAME(qtype));
1165 				kfree(qname);
1166 				return 0;
1167 			}
1168 			sbi->s_qf_names[qtype] = qname;
1169 			if (strchr(sbi->s_qf_names[qtype], '/')) {
1170 				printk(KERN_ERR
1171 					"EXT4-fs: quotafile must be on "
1172 					"filesystem root.\n");
1173 				kfree(sbi->s_qf_names[qtype]);
1174 				sbi->s_qf_names[qtype] = NULL;
1175 				return 0;
1176 			}
1177 			set_opt(sbi->s_mount_opt, QUOTA);
1178 			break;
1179 		case Opt_offusrjquota:
1180 			qtype = USRQUOTA;
1181 			goto clear_qf_name;
1182 		case Opt_offgrpjquota:
1183 			qtype = GRPQUOTA;
1184 clear_qf_name:
1185 			if ((sb_any_quota_enabled(sb) ||
1186 			     sb_any_quota_suspended(sb)) &&
1187 			    sbi->s_qf_names[qtype]) {
1188 				printk(KERN_ERR "EXT4-fs: Cannot change "
1189 					"journaled quota options when "
1190 					"quota turned on.\n");
1191 				return 0;
1192 			}
1193 			/*
1194 			 * The space will be released later when all options
1195 			 * are confirmed to be correct
1196 			 */
1197 			sbi->s_qf_names[qtype] = NULL;
1198 			break;
1199 		case Opt_jqfmt_vfsold:
1200 			qfmt = QFMT_VFS_OLD;
1201 			goto set_qf_format;
1202 		case Opt_jqfmt_vfsv0:
1203 			qfmt = QFMT_VFS_V0;
1204 set_qf_format:
1205 			if ((sb_any_quota_enabled(sb) ||
1206 			     sb_any_quota_suspended(sb)) &&
1207 			    sbi->s_jquota_fmt != qfmt) {
1208 				printk(KERN_ERR "EXT4-fs: Cannot change "
1209 					"journaled quota options when "
1210 					"quota turned on.\n");
1211 				return 0;
1212 			}
1213 			sbi->s_jquota_fmt = qfmt;
1214 			break;
1215 		case Opt_quota:
1216 		case Opt_usrquota:
1217 			set_opt(sbi->s_mount_opt, QUOTA);
1218 			set_opt(sbi->s_mount_opt, USRQUOTA);
1219 			break;
1220 		case Opt_grpquota:
1221 			set_opt(sbi->s_mount_opt, QUOTA);
1222 			set_opt(sbi->s_mount_opt, GRPQUOTA);
1223 			break;
1224 		case Opt_noquota:
1225 			if (sb_any_quota_enabled(sb)) {
1226 				printk(KERN_ERR "EXT4-fs: Cannot change quota "
1227 					"options when quota turned on.\n");
1228 				return 0;
1229 			}
1230 			clear_opt(sbi->s_mount_opt, QUOTA);
1231 			clear_opt(sbi->s_mount_opt, USRQUOTA);
1232 			clear_opt(sbi->s_mount_opt, GRPQUOTA);
1233 			break;
1234 #else
1235 		case Opt_quota:
1236 		case Opt_usrquota:
1237 		case Opt_grpquota:
1238 			printk(KERN_ERR
1239 				"EXT4-fs: quota options not supported.\n");
1240 			break;
1241 		case Opt_usrjquota:
1242 		case Opt_grpjquota:
1243 		case Opt_offusrjquota:
1244 		case Opt_offgrpjquota:
1245 		case Opt_jqfmt_vfsold:
1246 		case Opt_jqfmt_vfsv0:
1247 			printk(KERN_ERR
1248 				"EXT4-fs: journaled quota options not "
1249 				"supported.\n");
1250 			break;
1251 		case Opt_noquota:
1252 			break;
1253 #endif
1254 		case Opt_abort:
1255 			set_opt(sbi->s_mount_opt, ABORT);
1256 			break;
1257 		case Opt_barrier:
1258 			if (match_int(&args[0], &option))
1259 				return 0;
1260 			if (option)
1261 				set_opt(sbi->s_mount_opt, BARRIER);
1262 			else
1263 				clear_opt(sbi->s_mount_opt, BARRIER);
1264 			break;
1265 		case Opt_ignore:
1266 			break;
1267 		case Opt_resize:
1268 			if (!is_remount) {
1269 				printk("EXT4-fs: resize option only available "
1270 					"for remount\n");
1271 				return 0;
1272 			}
1273 			if (match_int(&args[0], &option) != 0)
1274 				return 0;
1275 			*n_blocks_count = option;
1276 			break;
1277 		case Opt_nobh:
1278 			set_opt(sbi->s_mount_opt, NOBH);
1279 			break;
1280 		case Opt_bh:
1281 			clear_opt(sbi->s_mount_opt, NOBH);
1282 			break;
1283 		case Opt_extents:
1284 			if (!EXT4_HAS_INCOMPAT_FEATURE(sb,
1285 					EXT4_FEATURE_INCOMPAT_EXTENTS)) {
1286 				ext4_warning(sb, __func__,
1287 					"extents feature not enabled "
1288 					"on this filesystem, use tune2fs\n");
1289 				return 0;
1290 			}
1291 			set_opt(sbi->s_mount_opt, EXTENTS);
1292 			break;
1293 		case Opt_noextents:
1294 			/*
1295 			 * When e2fsprogs support resizing an already existing
1296 			 * ext3 file system to greater than 2**32 we need to
1297 			 * add support to block allocator to handle growing
1298 			 * already existing block  mapped inode so that blocks
1299 			 * allocated for them fall within 2**32
1300 			 */
1301 			last_block = ext4_blocks_count(sbi->s_es) - 1;
1302 			if (last_block  > 0xffffffffULL) {
1303 				printk(KERN_ERR "EXT4-fs: Filesystem too "
1304 						"large to mount with "
1305 						"-o noextents options\n");
1306 				return 0;
1307 			}
1308 			clear_opt(sbi->s_mount_opt, EXTENTS);
1309 			break;
1310 		case Opt_i_version:
1311 			set_opt(sbi->s_mount_opt, I_VERSION);
1312 			sb->s_flags |= MS_I_VERSION;
1313 			break;
1314 		case Opt_nodelalloc:
1315 			clear_opt(sbi->s_mount_opt, DELALLOC);
1316 			break;
1317 		case Opt_stripe:
1318 			if (match_int(&args[0], &option))
1319 				return 0;
1320 			if (option < 0)
1321 				return 0;
1322 			sbi->s_stripe = option;
1323 			break;
1324 		case Opt_delalloc:
1325 			set_opt(sbi->s_mount_opt, DELALLOC);
1326 			break;
1327 		case Opt_inode_readahead_blks:
1328 			if (match_int(&args[0], &option))
1329 				return 0;
1330 			if (option < 0 || option > (1 << 30))
1331 				return 0;
1332 			sbi->s_inode_readahead_blks = option;
1333 			break;
1334 		default:
1335 			printk(KERN_ERR
1336 			       "EXT4-fs: Unrecognized mount option \"%s\" "
1337 			       "or missing value\n", p);
1338 			return 0;
1339 		}
1340 	}
1341 #ifdef CONFIG_QUOTA
1342 	if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1343 		if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1344 		     sbi->s_qf_names[USRQUOTA])
1345 			clear_opt(sbi->s_mount_opt, USRQUOTA);
1346 
1347 		if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1348 		     sbi->s_qf_names[GRPQUOTA])
1349 			clear_opt(sbi->s_mount_opt, GRPQUOTA);
1350 
1351 		if ((sbi->s_qf_names[USRQUOTA] &&
1352 				(sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1353 		    (sbi->s_qf_names[GRPQUOTA] &&
1354 				(sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1355 			printk(KERN_ERR "EXT4-fs: old and new quota "
1356 					"format mixing.\n");
1357 			return 0;
1358 		}
1359 
1360 		if (!sbi->s_jquota_fmt) {
1361 			printk(KERN_ERR "EXT4-fs: journaled quota format "
1362 					"not specified.\n");
1363 			return 0;
1364 		}
1365 	} else {
1366 		if (sbi->s_jquota_fmt) {
1367 			printk(KERN_ERR "EXT4-fs: journaled quota format "
1368 					"specified with no journaling "
1369 					"enabled.\n");
1370 			return 0;
1371 		}
1372 	}
1373 #endif
1374 	return 1;
1375 }
1376 
1377 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1378 			    int read_only)
1379 {
1380 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1381 	int res = 0;
1382 
1383 	if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1384 		printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1385 		       "forcing read-only mode\n");
1386 		res = MS_RDONLY;
1387 	}
1388 	if (read_only)
1389 		return res;
1390 	if (!(sbi->s_mount_state & EXT4_VALID_FS))
1391 		printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1392 		       "running e2fsck is recommended\n");
1393 	else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1394 		printk(KERN_WARNING
1395 		       "EXT4-fs warning: mounting fs with errors, "
1396 		       "running e2fsck is recommended\n");
1397 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1398 		 le16_to_cpu(es->s_mnt_count) >=
1399 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1400 		printk(KERN_WARNING
1401 		       "EXT4-fs warning: maximal mount count reached, "
1402 		       "running e2fsck is recommended\n");
1403 	else if (le32_to_cpu(es->s_checkinterval) &&
1404 		(le32_to_cpu(es->s_lastcheck) +
1405 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1406 		printk(KERN_WARNING
1407 		       "EXT4-fs warning: checktime reached, "
1408 		       "running e2fsck is recommended\n");
1409 #if 0
1410 		/* @@@ We _will_ want to clear the valid bit if we find
1411 		 * inconsistencies, to force a fsck at reboot.  But for
1412 		 * a plain journaled filesystem we can keep it set as
1413 		 * valid forever! :)
1414 		 */
1415 	es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1416 #endif
1417 	if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1418 		es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1419 	le16_add_cpu(&es->s_mnt_count, 1);
1420 	es->s_mtime = cpu_to_le32(get_seconds());
1421 	ext4_update_dynamic_rev(sb);
1422 	EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1423 
1424 	ext4_commit_super(sb, es, 1);
1425 	if (test_opt(sb, DEBUG))
1426 		printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
1427 				"bpg=%lu, ipg=%lu, mo=%04lx]\n",
1428 			sb->s_blocksize,
1429 			sbi->s_groups_count,
1430 			EXT4_BLOCKS_PER_GROUP(sb),
1431 			EXT4_INODES_PER_GROUP(sb),
1432 			sbi->s_mount_opt);
1433 
1434 	printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1435 	       sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1436 	       "external", EXT4_SB(sb)->s_journal->j_devname);
1437 	return res;
1438 }
1439 
1440 static int ext4_fill_flex_info(struct super_block *sb)
1441 {
1442 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1443 	struct ext4_group_desc *gdp = NULL;
1444 	struct buffer_head *bh;
1445 	ext4_group_t flex_group_count;
1446 	ext4_group_t flex_group;
1447 	int groups_per_flex = 0;
1448 	__u64 block_bitmap = 0;
1449 	int i;
1450 
1451 	if (!sbi->s_es->s_log_groups_per_flex) {
1452 		sbi->s_log_groups_per_flex = 0;
1453 		return 1;
1454 	}
1455 
1456 	sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1457 	groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1458 
1459 	/* We allocate both existing and potentially added groups */
1460 	flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1461 			((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1462 			      EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1463 	sbi->s_flex_groups = kzalloc(flex_group_count *
1464 				     sizeof(struct flex_groups), GFP_KERNEL);
1465 	if (sbi->s_flex_groups == NULL) {
1466 		printk(KERN_ERR "EXT4-fs: not enough memory for "
1467 				"%lu flex groups\n", flex_group_count);
1468 		goto failed;
1469 	}
1470 
1471 	gdp = ext4_get_group_desc(sb, 1, &bh);
1472 	block_bitmap = ext4_block_bitmap(sb, gdp) - 1;
1473 
1474 	for (i = 0; i < sbi->s_groups_count; i++) {
1475 		gdp = ext4_get_group_desc(sb, i, &bh);
1476 
1477 		flex_group = ext4_flex_group(sbi, i);
1478 		sbi->s_flex_groups[flex_group].free_inodes +=
1479 			le16_to_cpu(gdp->bg_free_inodes_count);
1480 		sbi->s_flex_groups[flex_group].free_blocks +=
1481 			le16_to_cpu(gdp->bg_free_blocks_count);
1482 	}
1483 
1484 	return 1;
1485 failed:
1486 	return 0;
1487 }
1488 
1489 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1490 			    struct ext4_group_desc *gdp)
1491 {
1492 	__u16 crc = 0;
1493 
1494 	if (sbi->s_es->s_feature_ro_compat &
1495 	    cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1496 		int offset = offsetof(struct ext4_group_desc, bg_checksum);
1497 		__le32 le_group = cpu_to_le32(block_group);
1498 
1499 		crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1500 		crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1501 		crc = crc16(crc, (__u8 *)gdp, offset);
1502 		offset += sizeof(gdp->bg_checksum); /* skip checksum */
1503 		/* for checksum of struct ext4_group_desc do the rest...*/
1504 		if ((sbi->s_es->s_feature_incompat &
1505 		     cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1506 		    offset < le16_to_cpu(sbi->s_es->s_desc_size))
1507 			crc = crc16(crc, (__u8 *)gdp + offset,
1508 				    le16_to_cpu(sbi->s_es->s_desc_size) -
1509 					offset);
1510 	}
1511 
1512 	return cpu_to_le16(crc);
1513 }
1514 
1515 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1516 				struct ext4_group_desc *gdp)
1517 {
1518 	if ((sbi->s_es->s_feature_ro_compat &
1519 	     cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1520 	    (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1521 		return 0;
1522 
1523 	return 1;
1524 }
1525 
1526 /* Called at mount-time, super-block is locked */
1527 static int ext4_check_descriptors(struct super_block *sb)
1528 {
1529 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1530 	ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1531 	ext4_fsblk_t last_block;
1532 	ext4_fsblk_t block_bitmap;
1533 	ext4_fsblk_t inode_bitmap;
1534 	ext4_fsblk_t inode_table;
1535 	int flexbg_flag = 0;
1536 	ext4_group_t i;
1537 
1538 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1539 		flexbg_flag = 1;
1540 
1541 	ext4_debug("Checking group descriptors");
1542 
1543 	for (i = 0; i < sbi->s_groups_count; i++) {
1544 		struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1545 
1546 		if (i == sbi->s_groups_count - 1 || flexbg_flag)
1547 			last_block = ext4_blocks_count(sbi->s_es) - 1;
1548 		else
1549 			last_block = first_block +
1550 				(EXT4_BLOCKS_PER_GROUP(sb) - 1);
1551 
1552 		block_bitmap = ext4_block_bitmap(sb, gdp);
1553 		if (block_bitmap < first_block || block_bitmap > last_block) {
1554 			printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1555 			       "Block bitmap for group %lu not in group "
1556 			       "(block %llu)!\n", i, block_bitmap);
1557 			return 0;
1558 		}
1559 		inode_bitmap = ext4_inode_bitmap(sb, gdp);
1560 		if (inode_bitmap < first_block || inode_bitmap > last_block) {
1561 			printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1562 			       "Inode bitmap for group %lu not in group "
1563 			       "(block %llu)!\n", i, inode_bitmap);
1564 			return 0;
1565 		}
1566 		inode_table = ext4_inode_table(sb, gdp);
1567 		if (inode_table < first_block ||
1568 		    inode_table + sbi->s_itb_per_group - 1 > last_block) {
1569 			printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1570 			       "Inode table for group %lu not in group "
1571 			       "(block %llu)!\n", i, inode_table);
1572 			return 0;
1573 		}
1574 		spin_lock(sb_bgl_lock(sbi, i));
1575 		if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1576 			printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1577 			       "Checksum for group %lu failed (%u!=%u)\n",
1578 			       i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1579 			       gdp)), le16_to_cpu(gdp->bg_checksum));
1580 			if (!(sb->s_flags & MS_RDONLY)) {
1581 				spin_unlock(sb_bgl_lock(sbi, i));
1582 				return 0;
1583 			}
1584 		}
1585 		spin_unlock(sb_bgl_lock(sbi, i));
1586 		if (!flexbg_flag)
1587 			first_block += EXT4_BLOCKS_PER_GROUP(sb);
1588 	}
1589 
1590 	ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1591 	sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
1592 	return 1;
1593 }
1594 
1595 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1596  * the superblock) which were deleted from all directories, but held open by
1597  * a process at the time of a crash.  We walk the list and try to delete these
1598  * inodes at recovery time (only with a read-write filesystem).
1599  *
1600  * In order to keep the orphan inode chain consistent during traversal (in
1601  * case of crash during recovery), we link each inode into the superblock
1602  * orphan list_head and handle it the same way as an inode deletion during
1603  * normal operation (which journals the operations for us).
1604  *
1605  * We only do an iget() and an iput() on each inode, which is very safe if we
1606  * accidentally point at an in-use or already deleted inode.  The worst that
1607  * can happen in this case is that we get a "bit already cleared" message from
1608  * ext4_free_inode().  The only reason we would point at a wrong inode is if
1609  * e2fsck was run on this filesystem, and it must have already done the orphan
1610  * inode cleanup for us, so we can safely abort without any further action.
1611  */
1612 static void ext4_orphan_cleanup(struct super_block *sb,
1613 				struct ext4_super_block *es)
1614 {
1615 	unsigned int s_flags = sb->s_flags;
1616 	int nr_orphans = 0, nr_truncates = 0;
1617 #ifdef CONFIG_QUOTA
1618 	int i;
1619 #endif
1620 	if (!es->s_last_orphan) {
1621 		jbd_debug(4, "no orphan inodes to clean up\n");
1622 		return;
1623 	}
1624 
1625 	if (bdev_read_only(sb->s_bdev)) {
1626 		printk(KERN_ERR "EXT4-fs: write access "
1627 			"unavailable, skipping orphan cleanup.\n");
1628 		return;
1629 	}
1630 
1631 	if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1632 		if (es->s_last_orphan)
1633 			jbd_debug(1, "Errors on filesystem, "
1634 				  "clearing orphan list.\n");
1635 		es->s_last_orphan = 0;
1636 		jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1637 		return;
1638 	}
1639 
1640 	if (s_flags & MS_RDONLY) {
1641 		printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1642 		       sb->s_id);
1643 		sb->s_flags &= ~MS_RDONLY;
1644 	}
1645 #ifdef CONFIG_QUOTA
1646 	/* Needed for iput() to work correctly and not trash data */
1647 	sb->s_flags |= MS_ACTIVE;
1648 	/* Turn on quotas so that they are updated correctly */
1649 	for (i = 0; i < MAXQUOTAS; i++) {
1650 		if (EXT4_SB(sb)->s_qf_names[i]) {
1651 			int ret = ext4_quota_on_mount(sb, i);
1652 			if (ret < 0)
1653 				printk(KERN_ERR
1654 					"EXT4-fs: Cannot turn on journaled "
1655 					"quota: error %d\n", ret);
1656 		}
1657 	}
1658 #endif
1659 
1660 	while (es->s_last_orphan) {
1661 		struct inode *inode;
1662 
1663 		inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1664 		if (IS_ERR(inode)) {
1665 			es->s_last_orphan = 0;
1666 			break;
1667 		}
1668 
1669 		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1670 		DQUOT_INIT(inode);
1671 		if (inode->i_nlink) {
1672 			printk(KERN_DEBUG
1673 				"%s: truncating inode %lu to %lld bytes\n",
1674 				__func__, inode->i_ino, inode->i_size);
1675 			jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1676 				  inode->i_ino, inode->i_size);
1677 			ext4_truncate(inode);
1678 			nr_truncates++;
1679 		} else {
1680 			printk(KERN_DEBUG
1681 				"%s: deleting unreferenced inode %lu\n",
1682 				__func__, inode->i_ino);
1683 			jbd_debug(2, "deleting unreferenced inode %lu\n",
1684 				  inode->i_ino);
1685 			nr_orphans++;
1686 		}
1687 		iput(inode);  /* The delete magic happens here! */
1688 	}
1689 
1690 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1691 
1692 	if (nr_orphans)
1693 		printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1694 		       sb->s_id, PLURAL(nr_orphans));
1695 	if (nr_truncates)
1696 		printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1697 		       sb->s_id, PLURAL(nr_truncates));
1698 #ifdef CONFIG_QUOTA
1699 	/* Turn quotas off */
1700 	for (i = 0; i < MAXQUOTAS; i++) {
1701 		if (sb_dqopt(sb)->files[i])
1702 			vfs_quota_off(sb, i, 0);
1703 	}
1704 #endif
1705 	sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1706 }
1707 /*
1708  * Maximal extent format file size.
1709  * Resulting logical blkno at s_maxbytes must fit in our on-disk
1710  * extent format containers, within a sector_t, and within i_blocks
1711  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
1712  * so that won't be a limiting factor.
1713  *
1714  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1715  */
1716 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1717 {
1718 	loff_t res;
1719 	loff_t upper_limit = MAX_LFS_FILESIZE;
1720 
1721 	/* small i_blocks in vfs inode? */
1722 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1723 		/*
1724 		 * CONFIG_LSF is not enabled implies the inode
1725 		 * i_block represent total blocks in 512 bytes
1726 		 * 32 == size of vfs inode i_blocks * 8
1727 		 */
1728 		upper_limit = (1LL << 32) - 1;
1729 
1730 		/* total blocks in file system block size */
1731 		upper_limit >>= (blkbits - 9);
1732 		upper_limit <<= blkbits;
1733 	}
1734 
1735 	/* 32-bit extent-start container, ee_block */
1736 	res = 1LL << 32;
1737 	res <<= blkbits;
1738 	res -= 1;
1739 
1740 	/* Sanity check against vm- & vfs- imposed limits */
1741 	if (res > upper_limit)
1742 		res = upper_limit;
1743 
1744 	return res;
1745 }
1746 
1747 /*
1748  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
1749  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1750  * We need to be 1 filesystem block less than the 2^48 sector limit.
1751  */
1752 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1753 {
1754 	loff_t res = EXT4_NDIR_BLOCKS;
1755 	int meta_blocks;
1756 	loff_t upper_limit;
1757 	/* This is calculated to be the largest file size for a
1758 	 * dense, bitmapped file such that the total number of
1759 	 * sectors in the file, including data and all indirect blocks,
1760 	 * does not exceed 2^48 -1
1761 	 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1762 	 * total number of  512 bytes blocks of the file
1763 	 */
1764 
1765 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1766 		/*
1767 		 * !has_huge_files or CONFIG_LSF is not enabled
1768 		 * implies the inode i_block represent total blocks in
1769 		 * 512 bytes 32 == size of vfs inode i_blocks * 8
1770 		 */
1771 		upper_limit = (1LL << 32) - 1;
1772 
1773 		/* total blocks in file system block size */
1774 		upper_limit >>= (bits - 9);
1775 
1776 	} else {
1777 		/*
1778 		 * We use 48 bit ext4_inode i_blocks
1779 		 * With EXT4_HUGE_FILE_FL set the i_blocks
1780 		 * represent total number of blocks in
1781 		 * file system block size
1782 		 */
1783 		upper_limit = (1LL << 48) - 1;
1784 
1785 	}
1786 
1787 	/* indirect blocks */
1788 	meta_blocks = 1;
1789 	/* double indirect blocks */
1790 	meta_blocks += 1 + (1LL << (bits-2));
1791 	/* tripple indirect blocks */
1792 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1793 
1794 	upper_limit -= meta_blocks;
1795 	upper_limit <<= bits;
1796 
1797 	res += 1LL << (bits-2);
1798 	res += 1LL << (2*(bits-2));
1799 	res += 1LL << (3*(bits-2));
1800 	res <<= bits;
1801 	if (res > upper_limit)
1802 		res = upper_limit;
1803 
1804 	if (res > MAX_LFS_FILESIZE)
1805 		res = MAX_LFS_FILESIZE;
1806 
1807 	return res;
1808 }
1809 
1810 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1811 				ext4_fsblk_t logical_sb_block, int nr)
1812 {
1813 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1814 	ext4_group_t bg, first_meta_bg;
1815 	int has_super = 0;
1816 
1817 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1818 
1819 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1820 	    nr < first_meta_bg)
1821 		return logical_sb_block + nr + 1;
1822 	bg = sbi->s_desc_per_block * nr;
1823 	if (ext4_bg_has_super(sb, bg))
1824 		has_super = 1;
1825 	return (has_super + ext4_group_first_block_no(sb, bg));
1826 }
1827 
1828 /**
1829  * ext4_get_stripe_size: Get the stripe size.
1830  * @sbi: In memory super block info
1831  *
1832  * If we have specified it via mount option, then
1833  * use the mount option value. If the value specified at mount time is
1834  * greater than the blocks per group use the super block value.
1835  * If the super block value is greater than blocks per group return 0.
1836  * Allocator needs it be less than blocks per group.
1837  *
1838  */
1839 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1840 {
1841 	unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1842 	unsigned long stripe_width =
1843 			le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1844 
1845 	if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1846 		return sbi->s_stripe;
1847 
1848 	if (stripe_width <= sbi->s_blocks_per_group)
1849 		return stripe_width;
1850 
1851 	if (stride <= sbi->s_blocks_per_group)
1852 		return stride;
1853 
1854 	return 0;
1855 }
1856 
1857 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
1858 				__releases(kernel_lock)
1859 				__acquires(kernel_lock)
1860 
1861 {
1862 	struct buffer_head *bh;
1863 	struct ext4_super_block *es = NULL;
1864 	struct ext4_sb_info *sbi;
1865 	ext4_fsblk_t block;
1866 	ext4_fsblk_t sb_block = get_sb_block(&data);
1867 	ext4_fsblk_t logical_sb_block;
1868 	unsigned long offset = 0;
1869 	unsigned int journal_inum = 0;
1870 	unsigned long journal_devnum = 0;
1871 	unsigned long def_mount_opts;
1872 	struct inode *root;
1873 	char *cp;
1874 	int ret = -EINVAL;
1875 	int blocksize;
1876 	int db_count;
1877 	int i;
1878 	int needs_recovery, has_huge_files;
1879 	__le32 features;
1880 	__u64 blocks_count;
1881 	int err;
1882 
1883 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1884 	if (!sbi)
1885 		return -ENOMEM;
1886 	sb->s_fs_info = sbi;
1887 	sbi->s_mount_opt = 0;
1888 	sbi->s_resuid = EXT4_DEF_RESUID;
1889 	sbi->s_resgid = EXT4_DEF_RESGID;
1890 	sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
1891 	sbi->s_sb_block = sb_block;
1892 
1893 	unlock_kernel();
1894 
1895 	/* Cleanup superblock name */
1896 	for (cp = sb->s_id; (cp = strchr(cp, '/'));)
1897 		*cp = '!';
1898 
1899 	blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
1900 	if (!blocksize) {
1901 		printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
1902 		goto out_fail;
1903 	}
1904 
1905 	/*
1906 	 * The ext4 superblock will not be buffer aligned for other than 1kB
1907 	 * block sizes.  We need to calculate the offset from buffer start.
1908 	 */
1909 	if (blocksize != EXT4_MIN_BLOCK_SIZE) {
1910 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1911 		offset = do_div(logical_sb_block, blocksize);
1912 	} else {
1913 		logical_sb_block = sb_block;
1914 	}
1915 
1916 	if (!(bh = sb_bread(sb, logical_sb_block))) {
1917 		printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
1918 		goto out_fail;
1919 	}
1920 	/*
1921 	 * Note: s_es must be initialized as soon as possible because
1922 	 *       some ext4 macro-instructions depend on its value
1923 	 */
1924 	es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
1925 	sbi->s_es = es;
1926 	sb->s_magic = le16_to_cpu(es->s_magic);
1927 	if (sb->s_magic != EXT4_SUPER_MAGIC)
1928 		goto cantfind_ext4;
1929 
1930 	/* Set defaults before we parse the mount options */
1931 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1932 	if (def_mount_opts & EXT4_DEFM_DEBUG)
1933 		set_opt(sbi->s_mount_opt, DEBUG);
1934 	if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
1935 		set_opt(sbi->s_mount_opt, GRPID);
1936 	if (def_mount_opts & EXT4_DEFM_UID16)
1937 		set_opt(sbi->s_mount_opt, NO_UID32);
1938 #ifdef CONFIG_EXT4_FS_XATTR
1939 	if (def_mount_opts & EXT4_DEFM_XATTR_USER)
1940 		set_opt(sbi->s_mount_opt, XATTR_USER);
1941 #endif
1942 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1943 	if (def_mount_opts & EXT4_DEFM_ACL)
1944 		set_opt(sbi->s_mount_opt, POSIX_ACL);
1945 #endif
1946 	if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1947 		sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1948 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1949 		sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1950 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1951 		sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1952 
1953 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
1954 		set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1955 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
1956 		set_opt(sbi->s_mount_opt, ERRORS_CONT);
1957 	else
1958 		set_opt(sbi->s_mount_opt, ERRORS_RO);
1959 
1960 	sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1961 	sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1962 
1963 	set_opt(sbi->s_mount_opt, RESERVATION);
1964 	set_opt(sbi->s_mount_opt, BARRIER);
1965 
1966 	/*
1967 	 * turn on extents feature by default in ext4 filesystem
1968 	 * only if feature flag already set by mkfs or tune2fs.
1969 	 * Use -o noextents to turn it off
1970 	 */
1971 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
1972 		set_opt(sbi->s_mount_opt, EXTENTS);
1973 	else
1974 		ext4_warning(sb, __func__,
1975 			"extents feature not enabled on this filesystem, "
1976 			"use tune2fs.\n");
1977 
1978 	/*
1979 	 * enable delayed allocation by default
1980 	 * Use -o nodelalloc to turn it off
1981 	 */
1982 	set_opt(sbi->s_mount_opt, DELALLOC);
1983 
1984 
1985 	if (!parse_options((char *) data, sb, &journal_inum, &journal_devnum,
1986 			   NULL, 0))
1987 		goto failed_mount;
1988 
1989 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1990 		((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1991 
1992 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1993 	    (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1994 	     EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1995 	     EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1996 		printk(KERN_WARNING
1997 		       "EXT4-fs warning: feature flags set on rev 0 fs, "
1998 		       "running e2fsck is recommended\n");
1999 
2000 	/*
2001 	 * Check feature flags regardless of the revision level, since we
2002 	 * previously didn't change the revision level when setting the flags,
2003 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
2004 	 */
2005 	features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2006 	if (features) {
2007 		printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
2008 		       "unsupported optional features (%x).\n",
2009 		       sb->s_id, le32_to_cpu(features));
2010 		goto failed_mount;
2011 	}
2012 	features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2013 	if (!(sb->s_flags & MS_RDONLY) && features) {
2014 		printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
2015 		       "unsupported optional features (%x).\n",
2016 		       sb->s_id, le32_to_cpu(features));
2017 		goto failed_mount;
2018 	}
2019 	has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2020 				    EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2021 	if (has_huge_files) {
2022 		/*
2023 		 * Large file size enabled file system can only be
2024 		 * mount if kernel is build with CONFIG_LSF
2025 		 */
2026 		if (sizeof(root->i_blocks) < sizeof(u64) &&
2027 				!(sb->s_flags & MS_RDONLY)) {
2028 			printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2029 					"files cannot be mounted read-write "
2030 					"without CONFIG_LSF.\n", sb->s_id);
2031 			goto failed_mount;
2032 		}
2033 	}
2034 	blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2035 
2036 	if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2037 	    blocksize > EXT4_MAX_BLOCK_SIZE) {
2038 		printk(KERN_ERR
2039 		       "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2040 		       blocksize, sb->s_id);
2041 		goto failed_mount;
2042 	}
2043 
2044 	if (sb->s_blocksize != blocksize) {
2045 
2046 		/* Validate the filesystem blocksize */
2047 		if (!sb_set_blocksize(sb, blocksize)) {
2048 			printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2049 					blocksize);
2050 			goto failed_mount;
2051 		}
2052 
2053 		brelse(bh);
2054 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2055 		offset = do_div(logical_sb_block, blocksize);
2056 		bh = sb_bread(sb, logical_sb_block);
2057 		if (!bh) {
2058 			printk(KERN_ERR
2059 			       "EXT4-fs: Can't read superblock on 2nd try.\n");
2060 			goto failed_mount;
2061 		}
2062 		es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2063 		sbi->s_es = es;
2064 		if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2065 			printk(KERN_ERR
2066 			       "EXT4-fs: Magic mismatch, very weird !\n");
2067 			goto failed_mount;
2068 		}
2069 	}
2070 
2071 	sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2072 						      has_huge_files);
2073 	sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2074 
2075 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2076 		sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2077 		sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2078 	} else {
2079 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2080 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2081 		if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2082 		    (!is_power_of_2(sbi->s_inode_size)) ||
2083 		    (sbi->s_inode_size > blocksize)) {
2084 			printk(KERN_ERR
2085 			       "EXT4-fs: unsupported inode size: %d\n",
2086 			       sbi->s_inode_size);
2087 			goto failed_mount;
2088 		}
2089 		if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2090 			sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2091 	}
2092 	sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2093 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2094 		if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2095 		    sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2096 		    !is_power_of_2(sbi->s_desc_size)) {
2097 			printk(KERN_ERR
2098 			       "EXT4-fs: unsupported descriptor size %lu\n",
2099 			       sbi->s_desc_size);
2100 			goto failed_mount;
2101 		}
2102 	} else
2103 		sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2104 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2105 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2106 	if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2107 		goto cantfind_ext4;
2108 	sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2109 	if (sbi->s_inodes_per_block == 0)
2110 		goto cantfind_ext4;
2111 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
2112 					sbi->s_inodes_per_block;
2113 	sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2114 	sbi->s_sbh = bh;
2115 	sbi->s_mount_state = le16_to_cpu(es->s_state);
2116 	sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2117 	sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2118 	for (i = 0; i < 4; i++)
2119 		sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2120 	sbi->s_def_hash_version = es->s_def_hash_version;
2121 
2122 	if (sbi->s_blocks_per_group > blocksize * 8) {
2123 		printk(KERN_ERR
2124 		       "EXT4-fs: #blocks per group too big: %lu\n",
2125 		       sbi->s_blocks_per_group);
2126 		goto failed_mount;
2127 	}
2128 	if (sbi->s_inodes_per_group > blocksize * 8) {
2129 		printk(KERN_ERR
2130 		       "EXT4-fs: #inodes per group too big: %lu\n",
2131 		       sbi->s_inodes_per_group);
2132 		goto failed_mount;
2133 	}
2134 
2135 	if (ext4_blocks_count(es) >
2136 		    (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2137 		printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2138 			" too large to mount safely\n", sb->s_id);
2139 		if (sizeof(sector_t) < 8)
2140 			printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2141 					"enabled\n");
2142 		goto failed_mount;
2143 	}
2144 
2145 	if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2146 		goto cantfind_ext4;
2147 
2148 	/* ensure blocks_count calculation below doesn't sign-extend */
2149 	if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2150 	    le32_to_cpu(es->s_first_data_block) + 1) {
2151 		printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2152 		       "first data block %u, blocks per group %lu\n",
2153 			ext4_blocks_count(es),
2154 			le32_to_cpu(es->s_first_data_block),
2155 			EXT4_BLOCKS_PER_GROUP(sb));
2156 		goto failed_mount;
2157 	}
2158 	blocks_count = (ext4_blocks_count(es) -
2159 			le32_to_cpu(es->s_first_data_block) +
2160 			EXT4_BLOCKS_PER_GROUP(sb) - 1);
2161 	do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2162 	sbi->s_groups_count = blocks_count;
2163 	db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2164 		   EXT4_DESC_PER_BLOCK(sb);
2165 	sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2166 				    GFP_KERNEL);
2167 	if (sbi->s_group_desc == NULL) {
2168 		printk(KERN_ERR "EXT4-fs: not enough memory\n");
2169 		goto failed_mount;
2170 	}
2171 
2172 #ifdef CONFIG_PROC_FS
2173 	if (ext4_proc_root)
2174 		sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2175 
2176 	if (sbi->s_proc)
2177 		proc_create_data("inode_readahead_blks", 0644, sbi->s_proc,
2178 				 &ext4_ui_proc_fops,
2179 				 &sbi->s_inode_readahead_blks);
2180 #endif
2181 
2182 	bgl_lock_init(&sbi->s_blockgroup_lock);
2183 
2184 	for (i = 0; i < db_count; i++) {
2185 		block = descriptor_loc(sb, logical_sb_block, i);
2186 		sbi->s_group_desc[i] = sb_bread(sb, block);
2187 		if (!sbi->s_group_desc[i]) {
2188 			printk(KERN_ERR "EXT4-fs: "
2189 			       "can't read group descriptor %d\n", i);
2190 			db_count = i;
2191 			goto failed_mount2;
2192 		}
2193 	}
2194 	if (!ext4_check_descriptors(sb)) {
2195 		printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2196 		goto failed_mount2;
2197 	}
2198 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2199 		if (!ext4_fill_flex_info(sb)) {
2200 			printk(KERN_ERR
2201 			       "EXT4-fs: unable to initialize "
2202 			       "flex_bg meta info!\n");
2203 			goto failed_mount2;
2204 		}
2205 
2206 	sbi->s_gdb_count = db_count;
2207 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2208 	spin_lock_init(&sbi->s_next_gen_lock);
2209 
2210 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
2211 			ext4_count_free_blocks(sb));
2212 	if (!err) {
2213 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
2214 				ext4_count_free_inodes(sb));
2215 	}
2216 	if (!err) {
2217 		err = percpu_counter_init(&sbi->s_dirs_counter,
2218 				ext4_count_dirs(sb));
2219 	}
2220 	if (!err) {
2221 		err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2222 	}
2223 	if (err) {
2224 		printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2225 		goto failed_mount3;
2226 	}
2227 
2228 	sbi->s_stripe = ext4_get_stripe_size(sbi);
2229 
2230 	/*
2231 	 * set up enough so that it can read an inode
2232 	 */
2233 	sb->s_op = &ext4_sops;
2234 	sb->s_export_op = &ext4_export_ops;
2235 	sb->s_xattr = ext4_xattr_handlers;
2236 #ifdef CONFIG_QUOTA
2237 	sb->s_qcop = &ext4_qctl_operations;
2238 	sb->dq_op = &ext4_quota_operations;
2239 #endif
2240 	INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2241 
2242 	sb->s_root = NULL;
2243 
2244 	needs_recovery = (es->s_last_orphan != 0 ||
2245 			  EXT4_HAS_INCOMPAT_FEATURE(sb,
2246 				    EXT4_FEATURE_INCOMPAT_RECOVER));
2247 
2248 	/*
2249 	 * The first inode we look at is the journal inode.  Don't try
2250 	 * root first: it may be modified in the journal!
2251 	 */
2252 	if (!test_opt(sb, NOLOAD) &&
2253 	    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2254 		if (ext4_load_journal(sb, es, journal_devnum))
2255 			goto failed_mount3;
2256 		if (!(sb->s_flags & MS_RDONLY) &&
2257 		    EXT4_SB(sb)->s_journal->j_failed_commit) {
2258 			printk(KERN_CRIT "EXT4-fs error (device %s): "
2259 			       "ext4_fill_super: Journal transaction "
2260 			       "%u is corrupt\n", sb->s_id,
2261 			       EXT4_SB(sb)->s_journal->j_failed_commit);
2262 			if (test_opt(sb, ERRORS_RO)) {
2263 				printk(KERN_CRIT
2264 				       "Mounting filesystem read-only\n");
2265 				sb->s_flags |= MS_RDONLY;
2266 				EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2267 				es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2268 			}
2269 			if (test_opt(sb, ERRORS_PANIC)) {
2270 				EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2271 				es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2272 				ext4_commit_super(sb, es, 1);
2273 				printk(KERN_CRIT
2274 				       "EXT4-fs (device %s): mount failed\n",
2275 				      sb->s_id);
2276 				goto failed_mount4;
2277 			}
2278 		}
2279 	} else if (journal_inum) {
2280 		if (ext4_create_journal(sb, es, journal_inum))
2281 			goto failed_mount3;
2282 	} else {
2283 		if (!silent)
2284 			printk(KERN_ERR
2285 			       "ext4: No journal on filesystem on %s\n",
2286 			       sb->s_id);
2287 		goto failed_mount3;
2288 	}
2289 
2290 	if (ext4_blocks_count(es) > 0xffffffffULL &&
2291 	    !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2292 				       JBD2_FEATURE_INCOMPAT_64BIT)) {
2293 		printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2294 		goto failed_mount4;
2295 	}
2296 
2297 	if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2298 		jbd2_journal_set_features(sbi->s_journal,
2299 				JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2300 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2301 	} else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2302 		jbd2_journal_set_features(sbi->s_journal,
2303 				JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2304 		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2305 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2306 	} else {
2307 		jbd2_journal_clear_features(sbi->s_journal,
2308 				JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2309 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2310 	}
2311 
2312 	/* We have now updated the journal if required, so we can
2313 	 * validate the data journaling mode. */
2314 	switch (test_opt(sb, DATA_FLAGS)) {
2315 	case 0:
2316 		/* No mode set, assume a default based on the journal
2317 		 * capabilities: ORDERED_DATA if the journal can
2318 		 * cope, else JOURNAL_DATA
2319 		 */
2320 		if (jbd2_journal_check_available_features
2321 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2322 			set_opt(sbi->s_mount_opt, ORDERED_DATA);
2323 		else
2324 			set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2325 		break;
2326 
2327 	case EXT4_MOUNT_ORDERED_DATA:
2328 	case EXT4_MOUNT_WRITEBACK_DATA:
2329 		if (!jbd2_journal_check_available_features
2330 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2331 			printk(KERN_ERR "EXT4-fs: Journal does not support "
2332 			       "requested data journaling mode\n");
2333 			goto failed_mount4;
2334 		}
2335 	default:
2336 		break;
2337 	}
2338 
2339 	if (test_opt(sb, NOBH)) {
2340 		if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2341 			printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2342 				"its supported only with writeback mode\n");
2343 			clear_opt(sbi->s_mount_opt, NOBH);
2344 		}
2345 	}
2346 	/*
2347 	 * The jbd2_journal_load will have done any necessary log recovery,
2348 	 * so we can safely mount the rest of the filesystem now.
2349 	 */
2350 
2351 	root = ext4_iget(sb, EXT4_ROOT_INO);
2352 	if (IS_ERR(root)) {
2353 		printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2354 		ret = PTR_ERR(root);
2355 		goto failed_mount4;
2356 	}
2357 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2358 		iput(root);
2359 		printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2360 		goto failed_mount4;
2361 	}
2362 	sb->s_root = d_alloc_root(root);
2363 	if (!sb->s_root) {
2364 		printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2365 		iput(root);
2366 		ret = -ENOMEM;
2367 		goto failed_mount4;
2368 	}
2369 
2370 	ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2371 
2372 	/* determine the minimum size of new large inodes, if present */
2373 	if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2374 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2375 						     EXT4_GOOD_OLD_INODE_SIZE;
2376 		if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2377 				       EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2378 			if (sbi->s_want_extra_isize <
2379 			    le16_to_cpu(es->s_want_extra_isize))
2380 				sbi->s_want_extra_isize =
2381 					le16_to_cpu(es->s_want_extra_isize);
2382 			if (sbi->s_want_extra_isize <
2383 			    le16_to_cpu(es->s_min_extra_isize))
2384 				sbi->s_want_extra_isize =
2385 					le16_to_cpu(es->s_min_extra_isize);
2386 		}
2387 	}
2388 	/* Check if enough inode space is available */
2389 	if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2390 							sbi->s_inode_size) {
2391 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2392 						       EXT4_GOOD_OLD_INODE_SIZE;
2393 		printk(KERN_INFO "EXT4-fs: required extra inode space not"
2394 			"available.\n");
2395 	}
2396 
2397 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2398 		printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2399 				"requested data journaling mode\n");
2400 		clear_opt(sbi->s_mount_opt, DELALLOC);
2401 	} else if (test_opt(sb, DELALLOC))
2402 		printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2403 
2404 	ext4_ext_init(sb);
2405 	err = ext4_mb_init(sb, needs_recovery);
2406 	if (err) {
2407 		printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2408 		       err);
2409 		goto failed_mount4;
2410 	}
2411 
2412 	/*
2413 	 * akpm: core read_super() calls in here with the superblock locked.
2414 	 * That deadlocks, because orphan cleanup needs to lock the superblock
2415 	 * in numerous places.  Here we just pop the lock - it's relatively
2416 	 * harmless, because we are now ready to accept write_super() requests,
2417 	 * and aviro says that's the only reason for hanging onto the
2418 	 * superblock lock.
2419 	 */
2420 	EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2421 	ext4_orphan_cleanup(sb, es);
2422 	EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2423 	if (needs_recovery)
2424 		printk(KERN_INFO "EXT4-fs: recovery complete.\n");
2425 	ext4_mark_recovery_complete(sb, es);
2426 	printk(KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2427 	       test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2428 	       test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
2429 	       "writeback");
2430 
2431 	lock_kernel();
2432 	return 0;
2433 
2434 cantfind_ext4:
2435 	if (!silent)
2436 		printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2437 		       sb->s_id);
2438 	goto failed_mount;
2439 
2440 failed_mount4:
2441 	jbd2_journal_destroy(sbi->s_journal);
2442 	sbi->s_journal = NULL;
2443 failed_mount3:
2444 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
2445 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
2446 	percpu_counter_destroy(&sbi->s_dirs_counter);
2447 	percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2448 failed_mount2:
2449 	for (i = 0; i < db_count; i++)
2450 		brelse(sbi->s_group_desc[i]);
2451 	kfree(sbi->s_group_desc);
2452 failed_mount:
2453 	if (sbi->s_proc) {
2454 		remove_proc_entry("inode_readahead_blks", sbi->s_proc);
2455 		remove_proc_entry(sb->s_id, ext4_proc_root);
2456 	}
2457 #ifdef CONFIG_QUOTA
2458 	for (i = 0; i < MAXQUOTAS; i++)
2459 		kfree(sbi->s_qf_names[i]);
2460 #endif
2461 	ext4_blkdev_remove(sbi);
2462 	brelse(bh);
2463 out_fail:
2464 	sb->s_fs_info = NULL;
2465 	kfree(sbi);
2466 	lock_kernel();
2467 	return ret;
2468 }
2469 
2470 /*
2471  * Setup any per-fs journal parameters now.  We'll do this both on
2472  * initial mount, once the journal has been initialised but before we've
2473  * done any recovery; and again on any subsequent remount.
2474  */
2475 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2476 {
2477 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2478 
2479 	if (sbi->s_commit_interval)
2480 		journal->j_commit_interval = sbi->s_commit_interval;
2481 	/* We could also set up an ext4-specific default for the commit
2482 	 * interval here, but for now we'll just fall back to the jbd
2483 	 * default. */
2484 
2485 	spin_lock(&journal->j_state_lock);
2486 	if (test_opt(sb, BARRIER))
2487 		journal->j_flags |= JBD2_BARRIER;
2488 	else
2489 		journal->j_flags &= ~JBD2_BARRIER;
2490 	if (test_opt(sb, DATA_ERR_ABORT))
2491 		journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2492 	else
2493 		journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2494 	spin_unlock(&journal->j_state_lock);
2495 }
2496 
2497 static journal_t *ext4_get_journal(struct super_block *sb,
2498 				   unsigned int journal_inum)
2499 {
2500 	struct inode *journal_inode;
2501 	journal_t *journal;
2502 
2503 	/* First, test for the existence of a valid inode on disk.  Bad
2504 	 * things happen if we iget() an unused inode, as the subsequent
2505 	 * iput() will try to delete it. */
2506 
2507 	journal_inode = ext4_iget(sb, journal_inum);
2508 	if (IS_ERR(journal_inode)) {
2509 		printk(KERN_ERR "EXT4-fs: no journal found.\n");
2510 		return NULL;
2511 	}
2512 	if (!journal_inode->i_nlink) {
2513 		make_bad_inode(journal_inode);
2514 		iput(journal_inode);
2515 		printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2516 		return NULL;
2517 	}
2518 
2519 	jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2520 		  journal_inode, journal_inode->i_size);
2521 	if (!S_ISREG(journal_inode->i_mode)) {
2522 		printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2523 		iput(journal_inode);
2524 		return NULL;
2525 	}
2526 
2527 	journal = jbd2_journal_init_inode(journal_inode);
2528 	if (!journal) {
2529 		printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2530 		iput(journal_inode);
2531 		return NULL;
2532 	}
2533 	journal->j_private = sb;
2534 	ext4_init_journal_params(sb, journal);
2535 	return journal;
2536 }
2537 
2538 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2539 				       dev_t j_dev)
2540 {
2541 	struct buffer_head *bh;
2542 	journal_t *journal;
2543 	ext4_fsblk_t start;
2544 	ext4_fsblk_t len;
2545 	int hblock, blocksize;
2546 	ext4_fsblk_t sb_block;
2547 	unsigned long offset;
2548 	struct ext4_super_block *es;
2549 	struct block_device *bdev;
2550 
2551 	bdev = ext4_blkdev_get(j_dev);
2552 	if (bdev == NULL)
2553 		return NULL;
2554 
2555 	if (bd_claim(bdev, sb)) {
2556 		printk(KERN_ERR
2557 			"EXT4: failed to claim external journal device.\n");
2558 		blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2559 		return NULL;
2560 	}
2561 
2562 	blocksize = sb->s_blocksize;
2563 	hblock = bdev_hardsect_size(bdev);
2564 	if (blocksize < hblock) {
2565 		printk(KERN_ERR
2566 			"EXT4-fs: blocksize too small for journal device.\n");
2567 		goto out_bdev;
2568 	}
2569 
2570 	sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2571 	offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2572 	set_blocksize(bdev, blocksize);
2573 	if (!(bh = __bread(bdev, sb_block, blocksize))) {
2574 		printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2575 		       "external journal\n");
2576 		goto out_bdev;
2577 	}
2578 
2579 	es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2580 	if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2581 	    !(le32_to_cpu(es->s_feature_incompat) &
2582 	      EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2583 		printk(KERN_ERR "EXT4-fs: external journal has "
2584 					"bad superblock\n");
2585 		brelse(bh);
2586 		goto out_bdev;
2587 	}
2588 
2589 	if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2590 		printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2591 		brelse(bh);
2592 		goto out_bdev;
2593 	}
2594 
2595 	len = ext4_blocks_count(es);
2596 	start = sb_block + 1;
2597 	brelse(bh);	/* we're done with the superblock */
2598 
2599 	journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2600 					start, len, blocksize);
2601 	if (!journal) {
2602 		printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2603 		goto out_bdev;
2604 	}
2605 	journal->j_private = sb;
2606 	ll_rw_block(READ, 1, &journal->j_sb_buffer);
2607 	wait_on_buffer(journal->j_sb_buffer);
2608 	if (!buffer_uptodate(journal->j_sb_buffer)) {
2609 		printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2610 		goto out_journal;
2611 	}
2612 	if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2613 		printk(KERN_ERR "EXT4-fs: External journal has more than one "
2614 					"user (unsupported) - %d\n",
2615 			be32_to_cpu(journal->j_superblock->s_nr_users));
2616 		goto out_journal;
2617 	}
2618 	EXT4_SB(sb)->journal_bdev = bdev;
2619 	ext4_init_journal_params(sb, journal);
2620 	return journal;
2621 out_journal:
2622 	jbd2_journal_destroy(journal);
2623 out_bdev:
2624 	ext4_blkdev_put(bdev);
2625 	return NULL;
2626 }
2627 
2628 static int ext4_load_journal(struct super_block *sb,
2629 			     struct ext4_super_block *es,
2630 			     unsigned long journal_devnum)
2631 {
2632 	journal_t *journal;
2633 	unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2634 	dev_t journal_dev;
2635 	int err = 0;
2636 	int really_read_only;
2637 
2638 	if (journal_devnum &&
2639 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2640 		printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2641 			"numbers have changed\n");
2642 		journal_dev = new_decode_dev(journal_devnum);
2643 	} else
2644 		journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2645 
2646 	really_read_only = bdev_read_only(sb->s_bdev);
2647 
2648 	/*
2649 	 * Are we loading a blank journal or performing recovery after a
2650 	 * crash?  For recovery, we need to check in advance whether we
2651 	 * can get read-write access to the device.
2652 	 */
2653 
2654 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2655 		if (sb->s_flags & MS_RDONLY) {
2656 			printk(KERN_INFO "EXT4-fs: INFO: recovery "
2657 					"required on readonly filesystem.\n");
2658 			if (really_read_only) {
2659 				printk(KERN_ERR "EXT4-fs: write access "
2660 					"unavailable, cannot proceed.\n");
2661 				return -EROFS;
2662 			}
2663 			printk(KERN_INFO "EXT4-fs: write access will "
2664 			       "be enabled during recovery.\n");
2665 		}
2666 	}
2667 
2668 	if (journal_inum && journal_dev) {
2669 		printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2670 		       "and inode journals!\n");
2671 		return -EINVAL;
2672 	}
2673 
2674 	if (journal_inum) {
2675 		if (!(journal = ext4_get_journal(sb, journal_inum)))
2676 			return -EINVAL;
2677 	} else {
2678 		if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2679 			return -EINVAL;
2680 	}
2681 
2682 	if (journal->j_flags & JBD2_BARRIER)
2683 		printk(KERN_INFO "EXT4-fs: barriers enabled\n");
2684 	else
2685 		printk(KERN_INFO "EXT4-fs: barriers disabled\n");
2686 
2687 	if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2688 		err = jbd2_journal_update_format(journal);
2689 		if (err)  {
2690 			printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2691 			jbd2_journal_destroy(journal);
2692 			return err;
2693 		}
2694 	}
2695 
2696 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2697 		err = jbd2_journal_wipe(journal, !really_read_only);
2698 	if (!err)
2699 		err = jbd2_journal_load(journal);
2700 
2701 	if (err) {
2702 		printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2703 		jbd2_journal_destroy(journal);
2704 		return err;
2705 	}
2706 
2707 	EXT4_SB(sb)->s_journal = journal;
2708 	ext4_clear_journal_err(sb, es);
2709 
2710 	if (journal_devnum &&
2711 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2712 		es->s_journal_dev = cpu_to_le32(journal_devnum);
2713 		sb->s_dirt = 1;
2714 
2715 		/* Make sure we flush the recovery flag to disk. */
2716 		ext4_commit_super(sb, es, 1);
2717 	}
2718 
2719 	return 0;
2720 }
2721 
2722 static int ext4_create_journal(struct super_block *sb,
2723 			       struct ext4_super_block *es,
2724 			       unsigned int journal_inum)
2725 {
2726 	journal_t *journal;
2727 	int err;
2728 
2729 	if (sb->s_flags & MS_RDONLY) {
2730 		printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
2731 				"create journal.\n");
2732 		return -EROFS;
2733 	}
2734 
2735 	journal = ext4_get_journal(sb, journal_inum);
2736 	if (!journal)
2737 		return -EINVAL;
2738 
2739 	printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
2740 	       journal_inum);
2741 
2742 	err = jbd2_journal_create(journal);
2743 	if (err) {
2744 		printk(KERN_ERR "EXT4-fs: error creating journal.\n");
2745 		jbd2_journal_destroy(journal);
2746 		return -EIO;
2747 	}
2748 
2749 	EXT4_SB(sb)->s_journal = journal;
2750 
2751 	ext4_update_dynamic_rev(sb);
2752 	EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2753 	EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
2754 
2755 	es->s_journal_inum = cpu_to_le32(journal_inum);
2756 	sb->s_dirt = 1;
2757 
2758 	/* Make sure we flush the recovery flag to disk. */
2759 	ext4_commit_super(sb, es, 1);
2760 
2761 	return 0;
2762 }
2763 
2764 static void ext4_commit_super(struct super_block *sb,
2765 			      struct ext4_super_block *es, int sync)
2766 {
2767 	struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2768 
2769 	if (!sbh)
2770 		return;
2771 	if (buffer_write_io_error(sbh)) {
2772 		/*
2773 		 * Oh, dear.  A previous attempt to write the
2774 		 * superblock failed.  This could happen because the
2775 		 * USB device was yanked out.  Or it could happen to
2776 		 * be a transient write error and maybe the block will
2777 		 * be remapped.  Nothing we can do but to retry the
2778 		 * write and hope for the best.
2779 		 */
2780 		printk(KERN_ERR "ext4: previous I/O error to "
2781 		       "superblock detected for %s.\n", sb->s_id);
2782 		clear_buffer_write_io_error(sbh);
2783 		set_buffer_uptodate(sbh);
2784 	}
2785 	es->s_wtime = cpu_to_le32(get_seconds());
2786 	ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
2787 	es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
2788 	BUFFER_TRACE(sbh, "marking dirty");
2789 	mark_buffer_dirty(sbh);
2790 	if (sync) {
2791 		sync_dirty_buffer(sbh);
2792 		if (buffer_write_io_error(sbh)) {
2793 			printk(KERN_ERR "ext4: I/O error while writing "
2794 			       "superblock for %s.\n", sb->s_id);
2795 			clear_buffer_write_io_error(sbh);
2796 			set_buffer_uptodate(sbh);
2797 		}
2798 	}
2799 }
2800 
2801 
2802 /*
2803  * Have we just finished recovery?  If so, and if we are mounting (or
2804  * remounting) the filesystem readonly, then we will end up with a
2805  * consistent fs on disk.  Record that fact.
2806  */
2807 static void ext4_mark_recovery_complete(struct super_block *sb,
2808 					struct ext4_super_block *es)
2809 {
2810 	journal_t *journal = EXT4_SB(sb)->s_journal;
2811 
2812 	jbd2_journal_lock_updates(journal);
2813 	if (jbd2_journal_flush(journal) < 0)
2814 		goto out;
2815 
2816 	lock_super(sb);
2817 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2818 	    sb->s_flags & MS_RDONLY) {
2819 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2820 		sb->s_dirt = 0;
2821 		ext4_commit_super(sb, es, 1);
2822 	}
2823 	unlock_super(sb);
2824 
2825 out:
2826 	jbd2_journal_unlock_updates(journal);
2827 }
2828 
2829 /*
2830  * If we are mounting (or read-write remounting) a filesystem whose journal
2831  * has recorded an error from a previous lifetime, move that error to the
2832  * main filesystem now.
2833  */
2834 static void ext4_clear_journal_err(struct super_block *sb,
2835 				   struct ext4_super_block *es)
2836 {
2837 	journal_t *journal;
2838 	int j_errno;
2839 	const char *errstr;
2840 
2841 	journal = EXT4_SB(sb)->s_journal;
2842 
2843 	/*
2844 	 * Now check for any error status which may have been recorded in the
2845 	 * journal by a prior ext4_error() or ext4_abort()
2846 	 */
2847 
2848 	j_errno = jbd2_journal_errno(journal);
2849 	if (j_errno) {
2850 		char nbuf[16];
2851 
2852 		errstr = ext4_decode_error(sb, j_errno, nbuf);
2853 		ext4_warning(sb, __func__, "Filesystem error recorded "
2854 			     "from previous mount: %s", errstr);
2855 		ext4_warning(sb, __func__, "Marking fs in need of "
2856 			     "filesystem check.");
2857 
2858 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2859 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2860 		ext4_commit_super(sb, es, 1);
2861 
2862 		jbd2_journal_clear_err(journal);
2863 	}
2864 }
2865 
2866 /*
2867  * Force the running and committing transactions to commit,
2868  * and wait on the commit.
2869  */
2870 int ext4_force_commit(struct super_block *sb)
2871 {
2872 	journal_t *journal;
2873 	int ret;
2874 
2875 	if (sb->s_flags & MS_RDONLY)
2876 		return 0;
2877 
2878 	journal = EXT4_SB(sb)->s_journal;
2879 	sb->s_dirt = 0;
2880 	ret = ext4_journal_force_commit(journal);
2881 	return ret;
2882 }
2883 
2884 /*
2885  * Ext4 always journals updates to the superblock itself, so we don't
2886  * have to propagate any other updates to the superblock on disk at this
2887  * point.  (We can probably nuke this function altogether, and remove
2888  * any mention to sb->s_dirt in all of fs/ext4; eventual cleanup...)
2889  */
2890 static void ext4_write_super(struct super_block *sb)
2891 {
2892 	if (mutex_trylock(&sb->s_lock) != 0)
2893 		BUG();
2894 	sb->s_dirt = 0;
2895 }
2896 
2897 static int ext4_sync_fs(struct super_block *sb, int wait)
2898 {
2899 	int ret = 0;
2900 
2901 	trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
2902 	sb->s_dirt = 0;
2903 	if (wait)
2904 		ret = ext4_force_commit(sb);
2905 	else
2906 		jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, NULL);
2907 	return ret;
2908 }
2909 
2910 /*
2911  * LVM calls this function before a (read-only) snapshot is created.  This
2912  * gives us a chance to flush the journal completely and mark the fs clean.
2913  */
2914 static void ext4_write_super_lockfs(struct super_block *sb)
2915 {
2916 	sb->s_dirt = 0;
2917 
2918 	if (!(sb->s_flags & MS_RDONLY)) {
2919 		journal_t *journal = EXT4_SB(sb)->s_journal;
2920 
2921 		/* Now we set up the journal barrier. */
2922 		jbd2_journal_lock_updates(journal);
2923 
2924 		/*
2925 		 * We don't want to clear needs_recovery flag when we failed
2926 		 * to flush the journal.
2927 		 */
2928 		if (jbd2_journal_flush(journal) < 0)
2929 			return;
2930 
2931 		/* Journal blocked and flushed, clear needs_recovery flag. */
2932 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2933 		ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2934 	}
2935 }
2936 
2937 /*
2938  * Called by LVM after the snapshot is done.  We need to reset the RECOVER
2939  * flag here, even though the filesystem is not technically dirty yet.
2940  */
2941 static void ext4_unlockfs(struct super_block *sb)
2942 {
2943 	if (!(sb->s_flags & MS_RDONLY)) {
2944 		lock_super(sb);
2945 		/* Reser the needs_recovery flag before the fs is unlocked. */
2946 		EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2947 		ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2948 		unlock_super(sb);
2949 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
2950 	}
2951 }
2952 
2953 static int ext4_remount(struct super_block *sb, int *flags, char *data)
2954 {
2955 	struct ext4_super_block *es;
2956 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2957 	ext4_fsblk_t n_blocks_count = 0;
2958 	unsigned long old_sb_flags;
2959 	struct ext4_mount_options old_opts;
2960 	ext4_group_t g;
2961 	int err;
2962 #ifdef CONFIG_QUOTA
2963 	int i;
2964 #endif
2965 
2966 	/* Store the original options */
2967 	old_sb_flags = sb->s_flags;
2968 	old_opts.s_mount_opt = sbi->s_mount_opt;
2969 	old_opts.s_resuid = sbi->s_resuid;
2970 	old_opts.s_resgid = sbi->s_resgid;
2971 	old_opts.s_commit_interval = sbi->s_commit_interval;
2972 #ifdef CONFIG_QUOTA
2973 	old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2974 	for (i = 0; i < MAXQUOTAS; i++)
2975 		old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2976 #endif
2977 
2978 	/*
2979 	 * Allow the "check" option to be passed as a remount option.
2980 	 */
2981 	if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2982 		err = -EINVAL;
2983 		goto restore_opts;
2984 	}
2985 
2986 	if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
2987 		ext4_abort(sb, __func__, "Abort forced by user");
2988 
2989 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2990 		((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2991 
2992 	es = sbi->s_es;
2993 
2994 	ext4_init_journal_params(sb, sbi->s_journal);
2995 
2996 	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2997 		n_blocks_count > ext4_blocks_count(es)) {
2998 		if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
2999 			err = -EROFS;
3000 			goto restore_opts;
3001 		}
3002 
3003 		if (*flags & MS_RDONLY) {
3004 			/*
3005 			 * First of all, the unconditional stuff we have to do
3006 			 * to disable replay of the journal when we next remount
3007 			 */
3008 			sb->s_flags |= MS_RDONLY;
3009 
3010 			/*
3011 			 * OK, test if we are remounting a valid rw partition
3012 			 * readonly, and if so set the rdonly flag and then
3013 			 * mark the partition as valid again.
3014 			 */
3015 			if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3016 			    (sbi->s_mount_state & EXT4_VALID_FS))
3017 				es->s_state = cpu_to_le16(sbi->s_mount_state);
3018 
3019 			/*
3020 			 * We have to unlock super so that we can wait for
3021 			 * transactions.
3022 			 */
3023 			unlock_super(sb);
3024 			ext4_mark_recovery_complete(sb, es);
3025 			lock_super(sb);
3026 		} else {
3027 			__le32 ret;
3028 			if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3029 					~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3030 				printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3031 				       "remount RDWR because of unsupported "
3032 				       "optional features (%x).\n",
3033 				       sb->s_id, le32_to_cpu(ret));
3034 				err = -EROFS;
3035 				goto restore_opts;
3036 			}
3037 
3038 			/*
3039 			 * Make sure the group descriptor checksums
3040 			 * are sane.  If they aren't, refuse to
3041 			 * remount r/w.
3042 			 */
3043 			for (g = 0; g < sbi->s_groups_count; g++) {
3044 				struct ext4_group_desc *gdp =
3045 					ext4_get_group_desc(sb, g, NULL);
3046 
3047 				if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3048 					printk(KERN_ERR
3049 	       "EXT4-fs: ext4_remount: "
3050 		"Checksum for group %lu failed (%u!=%u)\n",
3051 		g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3052 					       le16_to_cpu(gdp->bg_checksum));
3053 					err = -EINVAL;
3054 					goto restore_opts;
3055 				}
3056 			}
3057 
3058 			/*
3059 			 * If we have an unprocessed orphan list hanging
3060 			 * around from a previously readonly bdev mount,
3061 			 * require a full umount/remount for now.
3062 			 */
3063 			if (es->s_last_orphan) {
3064 				printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3065 				       "remount RDWR because of unprocessed "
3066 				       "orphan inode list.  Please "
3067 				       "umount/remount instead.\n",
3068 				       sb->s_id);
3069 				err = -EINVAL;
3070 				goto restore_opts;
3071 			}
3072 
3073 			/*
3074 			 * Mounting a RDONLY partition read-write, so reread
3075 			 * and store the current valid flag.  (It may have
3076 			 * been changed by e2fsck since we originally mounted
3077 			 * the partition.)
3078 			 */
3079 			ext4_clear_journal_err(sb, es);
3080 			sbi->s_mount_state = le16_to_cpu(es->s_state);
3081 			if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3082 				goto restore_opts;
3083 			if (!ext4_setup_super(sb, es, 0))
3084 				sb->s_flags &= ~MS_RDONLY;
3085 		}
3086 	}
3087 #ifdef CONFIG_QUOTA
3088 	/* Release old quota file names */
3089 	for (i = 0; i < MAXQUOTAS; i++)
3090 		if (old_opts.s_qf_names[i] &&
3091 		    old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3092 			kfree(old_opts.s_qf_names[i]);
3093 #endif
3094 	return 0;
3095 restore_opts:
3096 	sb->s_flags = old_sb_flags;
3097 	sbi->s_mount_opt = old_opts.s_mount_opt;
3098 	sbi->s_resuid = old_opts.s_resuid;
3099 	sbi->s_resgid = old_opts.s_resgid;
3100 	sbi->s_commit_interval = old_opts.s_commit_interval;
3101 #ifdef CONFIG_QUOTA
3102 	sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3103 	for (i = 0; i < MAXQUOTAS; i++) {
3104 		if (sbi->s_qf_names[i] &&
3105 		    old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3106 			kfree(sbi->s_qf_names[i]);
3107 		sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3108 	}
3109 #endif
3110 	return err;
3111 }
3112 
3113 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3114 {
3115 	struct super_block *sb = dentry->d_sb;
3116 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3117 	struct ext4_super_block *es = sbi->s_es;
3118 	u64 fsid;
3119 
3120 	if (test_opt(sb, MINIX_DF)) {
3121 		sbi->s_overhead_last = 0;
3122 	} else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3123 		ext4_group_t ngroups = sbi->s_groups_count, i;
3124 		ext4_fsblk_t overhead = 0;
3125 		smp_rmb();
3126 
3127 		/*
3128 		 * Compute the overhead (FS structures).  This is constant
3129 		 * for a given filesystem unless the number of block groups
3130 		 * changes so we cache the previous value until it does.
3131 		 */
3132 
3133 		/*
3134 		 * All of the blocks before first_data_block are
3135 		 * overhead
3136 		 */
3137 		overhead = le32_to_cpu(es->s_first_data_block);
3138 
3139 		/*
3140 		 * Add the overhead attributed to the superblock and
3141 		 * block group descriptors.  If the sparse superblocks
3142 		 * feature is turned on, then not all groups have this.
3143 		 */
3144 		for (i = 0; i < ngroups; i++) {
3145 			overhead += ext4_bg_has_super(sb, i) +
3146 				ext4_bg_num_gdb(sb, i);
3147 			cond_resched();
3148 		}
3149 
3150 		/*
3151 		 * Every block group has an inode bitmap, a block
3152 		 * bitmap, and an inode table.
3153 		 */
3154 		overhead += ngroups * (2 + sbi->s_itb_per_group);
3155 		sbi->s_overhead_last = overhead;
3156 		smp_wmb();
3157 		sbi->s_blocks_last = ext4_blocks_count(es);
3158 	}
3159 
3160 	buf->f_type = EXT4_SUPER_MAGIC;
3161 	buf->f_bsize = sb->s_blocksize;
3162 	buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3163 	buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3164 		       percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3165 	ext4_free_blocks_count_set(es, buf->f_bfree);
3166 	buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3167 	if (buf->f_bfree < ext4_r_blocks_count(es))
3168 		buf->f_bavail = 0;
3169 	buf->f_files = le32_to_cpu(es->s_inodes_count);
3170 	buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3171 	es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3172 	buf->f_namelen = EXT4_NAME_LEN;
3173 	fsid = le64_to_cpup((void *)es->s_uuid) ^
3174 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3175 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3176 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3177 	return 0;
3178 }
3179 
3180 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3181  * is locked for write. Otherwise the are possible deadlocks:
3182  * Process 1                         Process 2
3183  * ext4_create()                     quota_sync()
3184  *   jbd2_journal_start()                   write_dquot()
3185  *   DQUOT_INIT()                        down(dqio_mutex)
3186  *     down(dqio_mutex)                    jbd2_journal_start()
3187  *
3188  */
3189 
3190 #ifdef CONFIG_QUOTA
3191 
3192 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3193 {
3194 	return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3195 }
3196 
3197 static int ext4_dquot_initialize(struct inode *inode, int type)
3198 {
3199 	handle_t *handle;
3200 	int ret, err;
3201 
3202 	/* We may create quota structure so we need to reserve enough blocks */
3203 	handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
3204 	if (IS_ERR(handle))
3205 		return PTR_ERR(handle);
3206 	ret = dquot_initialize(inode, type);
3207 	err = ext4_journal_stop(handle);
3208 	if (!ret)
3209 		ret = err;
3210 	return ret;
3211 }
3212 
3213 static int ext4_dquot_drop(struct inode *inode)
3214 {
3215 	handle_t *handle;
3216 	int ret, err;
3217 
3218 	/* We may delete quota structure so we need to reserve enough blocks */
3219 	handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
3220 	if (IS_ERR(handle)) {
3221 		/*
3222 		 * We call dquot_drop() anyway to at least release references
3223 		 * to quota structures so that umount does not hang.
3224 		 */
3225 		dquot_drop(inode);
3226 		return PTR_ERR(handle);
3227 	}
3228 	ret = dquot_drop(inode);
3229 	err = ext4_journal_stop(handle);
3230 	if (!ret)
3231 		ret = err;
3232 	return ret;
3233 }
3234 
3235 static int ext4_write_dquot(struct dquot *dquot)
3236 {
3237 	int ret, err;
3238 	handle_t *handle;
3239 	struct inode *inode;
3240 
3241 	inode = dquot_to_inode(dquot);
3242 	handle = ext4_journal_start(inode,
3243 					EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3244 	if (IS_ERR(handle))
3245 		return PTR_ERR(handle);
3246 	ret = dquot_commit(dquot);
3247 	err = ext4_journal_stop(handle);
3248 	if (!ret)
3249 		ret = err;
3250 	return ret;
3251 }
3252 
3253 static int ext4_acquire_dquot(struct dquot *dquot)
3254 {
3255 	int ret, err;
3256 	handle_t *handle;
3257 
3258 	handle = ext4_journal_start(dquot_to_inode(dquot),
3259 					EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3260 	if (IS_ERR(handle))
3261 		return PTR_ERR(handle);
3262 	ret = dquot_acquire(dquot);
3263 	err = ext4_journal_stop(handle);
3264 	if (!ret)
3265 		ret = err;
3266 	return ret;
3267 }
3268 
3269 static int ext4_release_dquot(struct dquot *dquot)
3270 {
3271 	int ret, err;
3272 	handle_t *handle;
3273 
3274 	handle = ext4_journal_start(dquot_to_inode(dquot),
3275 					EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3276 	if (IS_ERR(handle)) {
3277 		/* Release dquot anyway to avoid endless cycle in dqput() */
3278 		dquot_release(dquot);
3279 		return PTR_ERR(handle);
3280 	}
3281 	ret = dquot_release(dquot);
3282 	err = ext4_journal_stop(handle);
3283 	if (!ret)
3284 		ret = err;
3285 	return ret;
3286 }
3287 
3288 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3289 {
3290 	/* Are we journaling quotas? */
3291 	if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3292 	    EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3293 		dquot_mark_dquot_dirty(dquot);
3294 		return ext4_write_dquot(dquot);
3295 	} else {
3296 		return dquot_mark_dquot_dirty(dquot);
3297 	}
3298 }
3299 
3300 static int ext4_write_info(struct super_block *sb, int type)
3301 {
3302 	int ret, err;
3303 	handle_t *handle;
3304 
3305 	/* Data block + inode block */
3306 	handle = ext4_journal_start(sb->s_root->d_inode, 2);
3307 	if (IS_ERR(handle))
3308 		return PTR_ERR(handle);
3309 	ret = dquot_commit_info(sb, type);
3310 	err = ext4_journal_stop(handle);
3311 	if (!ret)
3312 		ret = err;
3313 	return ret;
3314 }
3315 
3316 /*
3317  * Turn on quotas during mount time - we need to find
3318  * the quota file and such...
3319  */
3320 static int ext4_quota_on_mount(struct super_block *sb, int type)
3321 {
3322 	return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3323 			EXT4_SB(sb)->s_jquota_fmt, type);
3324 }
3325 
3326 /*
3327  * Standard function to be called on quota_on
3328  */
3329 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3330 			 char *name, int remount)
3331 {
3332 	int err;
3333 	struct path path;
3334 
3335 	if (!test_opt(sb, QUOTA))
3336 		return -EINVAL;
3337 	/* When remounting, no checks are needed and in fact, name is NULL */
3338 	if (remount)
3339 		return vfs_quota_on(sb, type, format_id, name, remount);
3340 
3341 	err = kern_path(name, LOOKUP_FOLLOW, &path);
3342 	if (err)
3343 		return err;
3344 
3345 	/* Quotafile not on the same filesystem? */
3346 	if (path.mnt->mnt_sb != sb) {
3347 		path_put(&path);
3348 		return -EXDEV;
3349 	}
3350 	/* Journaling quota? */
3351 	if (EXT4_SB(sb)->s_qf_names[type]) {
3352 		/* Quotafile not in fs root? */
3353 		if (path.dentry->d_parent != sb->s_root)
3354 			printk(KERN_WARNING
3355 				"EXT4-fs: Quota file not on filesystem root. "
3356 				"Journaled quota will not work.\n");
3357 	}
3358 
3359 	/*
3360 	 * When we journal data on quota file, we have to flush journal to see
3361 	 * all updates to the file when we bypass pagecache...
3362 	 */
3363 	if (ext4_should_journal_data(path.dentry->d_inode)) {
3364 		/*
3365 		 * We don't need to lock updates but journal_flush() could
3366 		 * otherwise be livelocked...
3367 		 */
3368 		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3369 		err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3370 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3371 		if (err) {
3372 			path_put(&path);
3373 			return err;
3374 		}
3375 	}
3376 
3377 	err = vfs_quota_on_path(sb, type, format_id, &path);
3378 	path_put(&path);
3379 	return err;
3380 }
3381 
3382 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3383  * acquiring the locks... As quota files are never truncated and quota code
3384  * itself serializes the operations (and noone else should touch the files)
3385  * we don't have to be afraid of races */
3386 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3387 			       size_t len, loff_t off)
3388 {
3389 	struct inode *inode = sb_dqopt(sb)->files[type];
3390 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3391 	int err = 0;
3392 	int offset = off & (sb->s_blocksize - 1);
3393 	int tocopy;
3394 	size_t toread;
3395 	struct buffer_head *bh;
3396 	loff_t i_size = i_size_read(inode);
3397 
3398 	if (off > i_size)
3399 		return 0;
3400 	if (off+len > i_size)
3401 		len = i_size-off;
3402 	toread = len;
3403 	while (toread > 0) {
3404 		tocopy = sb->s_blocksize - offset < toread ?
3405 				sb->s_blocksize - offset : toread;
3406 		bh = ext4_bread(NULL, inode, blk, 0, &err);
3407 		if (err)
3408 			return err;
3409 		if (!bh)	/* A hole? */
3410 			memset(data, 0, tocopy);
3411 		else
3412 			memcpy(data, bh->b_data+offset, tocopy);
3413 		brelse(bh);
3414 		offset = 0;
3415 		toread -= tocopy;
3416 		data += tocopy;
3417 		blk++;
3418 	}
3419 	return len;
3420 }
3421 
3422 /* Write to quotafile (we know the transaction is already started and has
3423  * enough credits) */
3424 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3425 				const char *data, size_t len, loff_t off)
3426 {
3427 	struct inode *inode = sb_dqopt(sb)->files[type];
3428 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3429 	int err = 0;
3430 	int offset = off & (sb->s_blocksize - 1);
3431 	int tocopy;
3432 	int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3433 	size_t towrite = len;
3434 	struct buffer_head *bh;
3435 	handle_t *handle = journal_current_handle();
3436 
3437 	if (!handle) {
3438 		printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
3439 			" cancelled because transaction is not started.\n",
3440 			(unsigned long long)off, (unsigned long long)len);
3441 		return -EIO;
3442 	}
3443 	mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3444 	while (towrite > 0) {
3445 		tocopy = sb->s_blocksize - offset < towrite ?
3446 				sb->s_blocksize - offset : towrite;
3447 		bh = ext4_bread(handle, inode, blk, 1, &err);
3448 		if (!bh)
3449 			goto out;
3450 		if (journal_quota) {
3451 			err = ext4_journal_get_write_access(handle, bh);
3452 			if (err) {
3453 				brelse(bh);
3454 				goto out;
3455 			}
3456 		}
3457 		lock_buffer(bh);
3458 		memcpy(bh->b_data+offset, data, tocopy);
3459 		flush_dcache_page(bh->b_page);
3460 		unlock_buffer(bh);
3461 		if (journal_quota)
3462 			err = ext4_journal_dirty_metadata(handle, bh);
3463 		else {
3464 			/* Always do at least ordered writes for quotas */
3465 			err = ext4_jbd2_file_inode(handle, inode);
3466 			mark_buffer_dirty(bh);
3467 		}
3468 		brelse(bh);
3469 		if (err)
3470 			goto out;
3471 		offset = 0;
3472 		towrite -= tocopy;
3473 		data += tocopy;
3474 		blk++;
3475 	}
3476 out:
3477 	if (len == towrite) {
3478 		mutex_unlock(&inode->i_mutex);
3479 		return err;
3480 	}
3481 	if (inode->i_size < off+len-towrite) {
3482 		i_size_write(inode, off+len-towrite);
3483 		EXT4_I(inode)->i_disksize = inode->i_size;
3484 	}
3485 	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3486 	ext4_mark_inode_dirty(handle, inode);
3487 	mutex_unlock(&inode->i_mutex);
3488 	return len - towrite;
3489 }
3490 
3491 #endif
3492 
3493 static int ext4_get_sb(struct file_system_type *fs_type,
3494 	int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3495 {
3496 	return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3497 }
3498 
3499 #ifdef CONFIG_PROC_FS
3500 static int ext4_ui_proc_show(struct seq_file *m, void *v)
3501 {
3502 	unsigned int *p = m->private;
3503 
3504 	seq_printf(m, "%u\n", *p);
3505 	return 0;
3506 }
3507 
3508 static int ext4_ui_proc_open(struct inode *inode, struct file *file)
3509 {
3510 	return single_open(file, ext4_ui_proc_show, PDE(inode)->data);
3511 }
3512 
3513 static ssize_t ext4_ui_proc_write(struct file *file, const char __user *buf,
3514 			       size_t cnt, loff_t *ppos)
3515 {
3516 	unsigned int *p = PDE(file->f_path.dentry->d_inode)->data;
3517 	char str[32];
3518 	unsigned long value;
3519 
3520 	if (cnt >= sizeof(str))
3521 		return -EINVAL;
3522 	if (copy_from_user(str, buf, cnt))
3523 		return -EFAULT;
3524 	value = simple_strtol(str, NULL, 0);
3525 	if (value < 0)
3526 		return -ERANGE;
3527 	*p = value;
3528 	return cnt;
3529 }
3530 
3531 const struct file_operations ext4_ui_proc_fops = {
3532 	.owner		= THIS_MODULE,
3533 	.open		= ext4_ui_proc_open,
3534 	.read		= seq_read,
3535 	.llseek		= seq_lseek,
3536 	.release	= single_release,
3537 	.write		= ext4_ui_proc_write,
3538 };
3539 #endif
3540 
3541 static struct file_system_type ext4_fs_type = {
3542 	.owner		= THIS_MODULE,
3543 	.name		= "ext4",
3544 	.get_sb		= ext4_get_sb,
3545 	.kill_sb	= kill_block_super,
3546 	.fs_flags	= FS_REQUIRES_DEV,
3547 };
3548 
3549 #ifdef CONFIG_EXT4DEV_COMPAT
3550 static int ext4dev_get_sb(struct file_system_type *fs_type,
3551 	int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3552 {
3553 	printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3554 	       "to mount using ext4\n");
3555 	printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3556 	       "will go away by 2.6.31\n");
3557 	return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3558 }
3559 
3560 static struct file_system_type ext4dev_fs_type = {
3561 	.owner		= THIS_MODULE,
3562 	.name		= "ext4dev",
3563 	.get_sb		= ext4dev_get_sb,
3564 	.kill_sb	= kill_block_super,
3565 	.fs_flags	= FS_REQUIRES_DEV,
3566 };
3567 MODULE_ALIAS("ext4dev");
3568 #endif
3569 
3570 static int __init init_ext4_fs(void)
3571 {
3572 	int err;
3573 
3574 	ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3575 	err = init_ext4_mballoc();
3576 	if (err)
3577 		return err;
3578 
3579 	err = init_ext4_xattr();
3580 	if (err)
3581 		goto out2;
3582 	err = init_inodecache();
3583 	if (err)
3584 		goto out1;
3585 	err = register_filesystem(&ext4_fs_type);
3586 	if (err)
3587 		goto out;
3588 #ifdef CONFIG_EXT4DEV_COMPAT
3589 	err = register_filesystem(&ext4dev_fs_type);
3590 	if (err) {
3591 		unregister_filesystem(&ext4_fs_type);
3592 		goto out;
3593 	}
3594 #endif
3595 	return 0;
3596 out:
3597 	destroy_inodecache();
3598 out1:
3599 	exit_ext4_xattr();
3600 out2:
3601 	exit_ext4_mballoc();
3602 	return err;
3603 }
3604 
3605 static void __exit exit_ext4_fs(void)
3606 {
3607 	unregister_filesystem(&ext4_fs_type);
3608 #ifdef CONFIG_EXT4DEV_COMPAT
3609 	unregister_filesystem(&ext4dev_fs_type);
3610 #endif
3611 	destroy_inodecache();
3612 	exit_ext4_xattr();
3613 	exit_ext4_mballoc();
3614 	remove_proc_entry("fs/ext4", NULL);
3615 }
3616 
3617 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3618 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
3619 MODULE_LICENSE("GPL");
3620 module_init(init_ext4_fs)
3621 module_exit(exit_ext4_fs)
3622