xref: /linux/fs/ext4/super.c (revision 3eeebf17f31c583f83e081b17b3076477cb96886)
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 	jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
337 }
338 
339 void ext4_warning(struct super_block *sb, const char *function,
340 		  const char *fmt, ...)
341 {
342 	va_list args;
343 
344 	va_start(args, fmt);
345 	printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
346 	       sb->s_id, function);
347 	vprintk(fmt, args);
348 	printk("\n");
349 	va_end(args);
350 }
351 
352 void ext4_update_dynamic_rev(struct super_block *sb)
353 {
354 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
355 
356 	if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
357 		return;
358 
359 	ext4_warning(sb, __func__,
360 		     "updating to rev %d because of new feature flag, "
361 		     "running e2fsck is recommended",
362 		     EXT4_DYNAMIC_REV);
363 
364 	es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
365 	es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
366 	es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
367 	/* leave es->s_feature_*compat flags alone */
368 	/* es->s_uuid will be set by e2fsck if empty */
369 
370 	/*
371 	 * The rest of the superblock fields should be zero, and if not it
372 	 * means they are likely already in use, so leave them alone.  We
373 	 * can leave it up to e2fsck to clean up any inconsistencies there.
374 	 */
375 }
376 
377 /*
378  * Open the external journal device
379  */
380 static struct block_device *ext4_blkdev_get(dev_t dev)
381 {
382 	struct block_device *bdev;
383 	char b[BDEVNAME_SIZE];
384 
385 	bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
386 	if (IS_ERR(bdev))
387 		goto fail;
388 	return bdev;
389 
390 fail:
391 	printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
392 			__bdevname(dev, b), PTR_ERR(bdev));
393 	return NULL;
394 }
395 
396 /*
397  * Release the journal device
398  */
399 static int ext4_blkdev_put(struct block_device *bdev)
400 {
401 	bd_release(bdev);
402 	return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
403 }
404 
405 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
406 {
407 	struct block_device *bdev;
408 	int ret = -ENODEV;
409 
410 	bdev = sbi->journal_bdev;
411 	if (bdev) {
412 		ret = ext4_blkdev_put(bdev);
413 		sbi->journal_bdev = NULL;
414 	}
415 	return ret;
416 }
417 
418 static inline struct inode *orphan_list_entry(struct list_head *l)
419 {
420 	return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
421 }
422 
423 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
424 {
425 	struct list_head *l;
426 
427 	printk(KERN_ERR "sb orphan head is %d\n",
428 	       le32_to_cpu(sbi->s_es->s_last_orphan));
429 
430 	printk(KERN_ERR "sb_info orphan list:\n");
431 	list_for_each(l, &sbi->s_orphan) {
432 		struct inode *inode = orphan_list_entry(l);
433 		printk(KERN_ERR "  "
434 		       "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
435 		       inode->i_sb->s_id, inode->i_ino, inode,
436 		       inode->i_mode, inode->i_nlink,
437 		       NEXT_ORPHAN(inode));
438 	}
439 }
440 
441 static void ext4_put_super(struct super_block *sb)
442 {
443 	struct ext4_sb_info *sbi = EXT4_SB(sb);
444 	struct ext4_super_block *es = sbi->s_es;
445 	int i;
446 
447 	ext4_mb_release(sb);
448 	ext4_ext_release(sb);
449 	ext4_xattr_put_super(sb);
450 	if (jbd2_journal_destroy(sbi->s_journal) < 0)
451 		ext4_abort(sb, __func__, "Couldn't clean up the journal");
452 	sbi->s_journal = NULL;
453 	if (!(sb->s_flags & MS_RDONLY)) {
454 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
455 		es->s_state = cpu_to_le16(sbi->s_mount_state);
456 		ext4_commit_super(sb, es, 1);
457 	}
458 	if (sbi->s_proc) {
459 		remove_proc_entry("inode_readahead_blks", sbi->s_proc);
460 		remove_proc_entry(sb->s_id, ext4_proc_root);
461 	}
462 
463 	for (i = 0; i < sbi->s_gdb_count; i++)
464 		brelse(sbi->s_group_desc[i]);
465 	kfree(sbi->s_group_desc);
466 	kfree(sbi->s_flex_groups);
467 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
468 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
469 	percpu_counter_destroy(&sbi->s_dirs_counter);
470 	percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
471 	brelse(sbi->s_sbh);
472 #ifdef CONFIG_QUOTA
473 	for (i = 0; i < MAXQUOTAS; i++)
474 		kfree(sbi->s_qf_names[i]);
475 #endif
476 
477 	/* Debugging code just in case the in-memory inode orphan list
478 	 * isn't empty.  The on-disk one can be non-empty if we've
479 	 * detected an error and taken the fs readonly, but the
480 	 * in-memory list had better be clean by this point. */
481 	if (!list_empty(&sbi->s_orphan))
482 		dump_orphan_list(sb, sbi);
483 	J_ASSERT(list_empty(&sbi->s_orphan));
484 
485 	invalidate_bdev(sb->s_bdev);
486 	if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
487 		/*
488 		 * Invalidate the journal device's buffers.  We don't want them
489 		 * floating about in memory - the physical journal device may
490 		 * hotswapped, and it breaks the `ro-after' testing code.
491 		 */
492 		sync_blockdev(sbi->journal_bdev);
493 		invalidate_bdev(sbi->journal_bdev);
494 		ext4_blkdev_remove(sbi);
495 	}
496 	sb->s_fs_info = NULL;
497 	kfree(sbi);
498 	return;
499 }
500 
501 static struct kmem_cache *ext4_inode_cachep;
502 
503 /*
504  * Called inside transaction, so use GFP_NOFS
505  */
506 static struct inode *ext4_alloc_inode(struct super_block *sb)
507 {
508 	struct ext4_inode_info *ei;
509 
510 	ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
511 	if (!ei)
512 		return NULL;
513 #ifdef CONFIG_EXT4_FS_POSIX_ACL
514 	ei->i_acl = EXT4_ACL_NOT_CACHED;
515 	ei->i_default_acl = EXT4_ACL_NOT_CACHED;
516 #endif
517 	ei->vfs_inode.i_version = 1;
518 	ei->vfs_inode.i_data.writeback_index = 0;
519 	memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
520 	INIT_LIST_HEAD(&ei->i_prealloc_list);
521 	spin_lock_init(&ei->i_prealloc_lock);
522 	jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
523 	ei->i_reserved_data_blocks = 0;
524 	ei->i_reserved_meta_blocks = 0;
525 	ei->i_allocated_meta_blocks = 0;
526 	ei->i_delalloc_reserved_flag = 0;
527 	spin_lock_init(&(ei->i_block_reservation_lock));
528 	return &ei->vfs_inode;
529 }
530 
531 static void ext4_destroy_inode(struct inode *inode)
532 {
533 	if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
534 		printk("EXT4 Inode %p: orphan list check failed!\n",
535 			EXT4_I(inode));
536 		print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
537 				EXT4_I(inode), sizeof(struct ext4_inode_info),
538 				true);
539 		dump_stack();
540 	}
541 	kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
542 }
543 
544 static void init_once(void *foo)
545 {
546 	struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
547 
548 	INIT_LIST_HEAD(&ei->i_orphan);
549 #ifdef CONFIG_EXT4_FS_XATTR
550 	init_rwsem(&ei->xattr_sem);
551 #endif
552 	init_rwsem(&ei->i_data_sem);
553 	inode_init_once(&ei->vfs_inode);
554 }
555 
556 static int init_inodecache(void)
557 {
558 	ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
559 					     sizeof(struct ext4_inode_info),
560 					     0, (SLAB_RECLAIM_ACCOUNT|
561 						SLAB_MEM_SPREAD),
562 					     init_once);
563 	if (ext4_inode_cachep == NULL)
564 		return -ENOMEM;
565 	return 0;
566 }
567 
568 static void destroy_inodecache(void)
569 {
570 	kmem_cache_destroy(ext4_inode_cachep);
571 }
572 
573 static void ext4_clear_inode(struct inode *inode)
574 {
575 #ifdef CONFIG_EXT4_FS_POSIX_ACL
576 	if (EXT4_I(inode)->i_acl &&
577 			EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
578 		posix_acl_release(EXT4_I(inode)->i_acl);
579 		EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
580 	}
581 	if (EXT4_I(inode)->i_default_acl &&
582 			EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
583 		posix_acl_release(EXT4_I(inode)->i_default_acl);
584 		EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
585 	}
586 #endif
587 	ext4_discard_preallocations(inode);
588 	jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
589 				       &EXT4_I(inode)->jinode);
590 }
591 
592 static inline void ext4_show_quota_options(struct seq_file *seq,
593 					   struct super_block *sb)
594 {
595 #if defined(CONFIG_QUOTA)
596 	struct ext4_sb_info *sbi = EXT4_SB(sb);
597 
598 	if (sbi->s_jquota_fmt)
599 		seq_printf(seq, ",jqfmt=%s",
600 		(sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
601 
602 	if (sbi->s_qf_names[USRQUOTA])
603 		seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
604 
605 	if (sbi->s_qf_names[GRPQUOTA])
606 		seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
607 
608 	if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
609 		seq_puts(seq, ",usrquota");
610 
611 	if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
612 		seq_puts(seq, ",grpquota");
613 #endif
614 }
615 
616 /*
617  * Show an option if
618  *  - it's set to a non-default value OR
619  *  - if the per-sb default is different from the global default
620  */
621 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
622 {
623 	int def_errors;
624 	unsigned long def_mount_opts;
625 	struct super_block *sb = vfs->mnt_sb;
626 	struct ext4_sb_info *sbi = EXT4_SB(sb);
627 	struct ext4_super_block *es = sbi->s_es;
628 
629 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
630 	def_errors     = le16_to_cpu(es->s_errors);
631 
632 	if (sbi->s_sb_block != 1)
633 		seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
634 	if (test_opt(sb, MINIX_DF))
635 		seq_puts(seq, ",minixdf");
636 	if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
637 		seq_puts(seq, ",grpid");
638 	if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
639 		seq_puts(seq, ",nogrpid");
640 	if (sbi->s_resuid != EXT4_DEF_RESUID ||
641 	    le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
642 		seq_printf(seq, ",resuid=%u", sbi->s_resuid);
643 	}
644 	if (sbi->s_resgid != EXT4_DEF_RESGID ||
645 	    le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
646 		seq_printf(seq, ",resgid=%u", sbi->s_resgid);
647 	}
648 	if (test_opt(sb, ERRORS_RO)) {
649 		if (def_errors == EXT4_ERRORS_PANIC ||
650 		    def_errors == EXT4_ERRORS_CONTINUE) {
651 			seq_puts(seq, ",errors=remount-ro");
652 		}
653 	}
654 	if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
655 		seq_puts(seq, ",errors=continue");
656 	if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
657 		seq_puts(seq, ",errors=panic");
658 	if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
659 		seq_puts(seq, ",nouid32");
660 	if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
661 		seq_puts(seq, ",debug");
662 	if (test_opt(sb, OLDALLOC))
663 		seq_puts(seq, ",oldalloc");
664 #ifdef CONFIG_EXT4_FS_XATTR
665 	if (test_opt(sb, XATTR_USER) &&
666 		!(def_mount_opts & EXT4_DEFM_XATTR_USER))
667 		seq_puts(seq, ",user_xattr");
668 	if (!test_opt(sb, XATTR_USER) &&
669 	    (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
670 		seq_puts(seq, ",nouser_xattr");
671 	}
672 #endif
673 #ifdef CONFIG_EXT4_FS_POSIX_ACL
674 	if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
675 		seq_puts(seq, ",acl");
676 	if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
677 		seq_puts(seq, ",noacl");
678 #endif
679 	if (!test_opt(sb, RESERVATION))
680 		seq_puts(seq, ",noreservation");
681 	if (sbi->s_commit_interval) {
682 		seq_printf(seq, ",commit=%u",
683 			   (unsigned) (sbi->s_commit_interval / HZ));
684 	}
685 	/*
686 	 * We're changing the default of barrier mount option, so
687 	 * let's always display its mount state so it's clear what its
688 	 * status is.
689 	 */
690 	seq_puts(seq, ",barrier=");
691 	seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
692 	if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
693 		seq_puts(seq, ",journal_async_commit");
694 	if (test_opt(sb, NOBH))
695 		seq_puts(seq, ",nobh");
696 	if (!test_opt(sb, EXTENTS))
697 		seq_puts(seq, ",noextents");
698 	if (test_opt(sb, I_VERSION))
699 		seq_puts(seq, ",i_version");
700 	if (!test_opt(sb, DELALLOC))
701 		seq_puts(seq, ",nodelalloc");
702 
703 
704 	if (sbi->s_stripe)
705 		seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
706 	/*
707 	 * journal mode get enabled in different ways
708 	 * So just print the value even if we didn't specify it
709 	 */
710 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
711 		seq_puts(seq, ",data=journal");
712 	else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
713 		seq_puts(seq, ",data=ordered");
714 	else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
715 		seq_puts(seq, ",data=writeback");
716 
717 	if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
718 		seq_printf(seq, ",inode_readahead_blks=%u",
719 			   sbi->s_inode_readahead_blks);
720 
721 	if (test_opt(sb, DATA_ERR_ABORT))
722 		seq_puts(seq, ",data_err=abort");
723 
724 	ext4_show_quota_options(seq, sb);
725 	return 0;
726 }
727 
728 
729 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
730 		u64 ino, u32 generation)
731 {
732 	struct inode *inode;
733 
734 	if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
735 		return ERR_PTR(-ESTALE);
736 	if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
737 		return ERR_PTR(-ESTALE);
738 
739 	/* iget isn't really right if the inode is currently unallocated!!
740 	 *
741 	 * ext4_read_inode will return a bad_inode if the inode had been
742 	 * deleted, so we should be safe.
743 	 *
744 	 * Currently we don't know the generation for parent directory, so
745 	 * a generation of 0 means "accept any"
746 	 */
747 	inode = ext4_iget(sb, ino);
748 	if (IS_ERR(inode))
749 		return ERR_CAST(inode);
750 	if (generation && inode->i_generation != generation) {
751 		iput(inode);
752 		return ERR_PTR(-ESTALE);
753 	}
754 
755 	return inode;
756 }
757 
758 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
759 		int fh_len, int fh_type)
760 {
761 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
762 				    ext4_nfs_get_inode);
763 }
764 
765 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
766 		int fh_len, int fh_type)
767 {
768 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
769 				    ext4_nfs_get_inode);
770 }
771 
772 #ifdef CONFIG_QUOTA
773 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
774 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
775 
776 static int ext4_dquot_initialize(struct inode *inode, int type);
777 static int ext4_dquot_drop(struct inode *inode);
778 static int ext4_write_dquot(struct dquot *dquot);
779 static int ext4_acquire_dquot(struct dquot *dquot);
780 static int ext4_release_dquot(struct dquot *dquot);
781 static int ext4_mark_dquot_dirty(struct dquot *dquot);
782 static int ext4_write_info(struct super_block *sb, int type);
783 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
784 				char *path, int remount);
785 static int ext4_quota_on_mount(struct super_block *sb, int type);
786 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
787 			       size_t len, loff_t off);
788 static ssize_t ext4_quota_write(struct super_block *sb, int type,
789 				const char *data, size_t len, loff_t off);
790 
791 static struct dquot_operations ext4_quota_operations = {
792 	.initialize	= ext4_dquot_initialize,
793 	.drop		= ext4_dquot_drop,
794 	.alloc_space	= dquot_alloc_space,
795 	.alloc_inode	= dquot_alloc_inode,
796 	.free_space	= dquot_free_space,
797 	.free_inode	= dquot_free_inode,
798 	.transfer	= dquot_transfer,
799 	.write_dquot	= ext4_write_dquot,
800 	.acquire_dquot	= ext4_acquire_dquot,
801 	.release_dquot	= ext4_release_dquot,
802 	.mark_dirty	= ext4_mark_dquot_dirty,
803 	.write_info	= ext4_write_info
804 };
805 
806 static struct quotactl_ops ext4_qctl_operations = {
807 	.quota_on	= ext4_quota_on,
808 	.quota_off	= vfs_quota_off,
809 	.quota_sync	= vfs_quota_sync,
810 	.get_info	= vfs_get_dqinfo,
811 	.set_info	= vfs_set_dqinfo,
812 	.get_dqblk	= vfs_get_dqblk,
813 	.set_dqblk	= vfs_set_dqblk
814 };
815 #endif
816 
817 static const struct super_operations ext4_sops = {
818 	.alloc_inode	= ext4_alloc_inode,
819 	.destroy_inode	= ext4_destroy_inode,
820 	.write_inode	= ext4_write_inode,
821 	.dirty_inode	= ext4_dirty_inode,
822 	.delete_inode	= ext4_delete_inode,
823 	.put_super	= ext4_put_super,
824 	.write_super	= ext4_write_super,
825 	.sync_fs	= ext4_sync_fs,
826 	.write_super_lockfs = ext4_write_super_lockfs,
827 	.unlockfs	= ext4_unlockfs,
828 	.statfs		= ext4_statfs,
829 	.remount_fs	= ext4_remount,
830 	.clear_inode	= ext4_clear_inode,
831 	.show_options	= ext4_show_options,
832 #ifdef CONFIG_QUOTA
833 	.quota_read	= ext4_quota_read,
834 	.quota_write	= ext4_quota_write,
835 #endif
836 };
837 
838 static const struct export_operations ext4_export_ops = {
839 	.fh_to_dentry = ext4_fh_to_dentry,
840 	.fh_to_parent = ext4_fh_to_parent,
841 	.get_parent = ext4_get_parent,
842 };
843 
844 enum {
845 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
846 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
847 	Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
848 	Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
849 	Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
850 	Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
851 	Opt_journal_checksum, Opt_journal_async_commit,
852 	Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
853 	Opt_data_err_abort, Opt_data_err_ignore,
854 	Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
855 	Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
856 	Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
857 	Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
858 	Opt_stripe, Opt_delalloc, Opt_nodelalloc,
859 	Opt_inode_readahead_blks
860 };
861 
862 static const match_table_t tokens = {
863 	{Opt_bsd_df, "bsddf"},
864 	{Opt_minix_df, "minixdf"},
865 	{Opt_grpid, "grpid"},
866 	{Opt_grpid, "bsdgroups"},
867 	{Opt_nogrpid, "nogrpid"},
868 	{Opt_nogrpid, "sysvgroups"},
869 	{Opt_resgid, "resgid=%u"},
870 	{Opt_resuid, "resuid=%u"},
871 	{Opt_sb, "sb=%u"},
872 	{Opt_err_cont, "errors=continue"},
873 	{Opt_err_panic, "errors=panic"},
874 	{Opt_err_ro, "errors=remount-ro"},
875 	{Opt_nouid32, "nouid32"},
876 	{Opt_debug, "debug"},
877 	{Opt_oldalloc, "oldalloc"},
878 	{Opt_orlov, "orlov"},
879 	{Opt_user_xattr, "user_xattr"},
880 	{Opt_nouser_xattr, "nouser_xattr"},
881 	{Opt_acl, "acl"},
882 	{Opt_noacl, "noacl"},
883 	{Opt_reservation, "reservation"},
884 	{Opt_noreservation, "noreservation"},
885 	{Opt_noload, "noload"},
886 	{Opt_nobh, "nobh"},
887 	{Opt_bh, "bh"},
888 	{Opt_commit, "commit=%u"},
889 	{Opt_journal_update, "journal=update"},
890 	{Opt_journal_inum, "journal=%u"},
891 	{Opt_journal_dev, "journal_dev=%u"},
892 	{Opt_journal_checksum, "journal_checksum"},
893 	{Opt_journal_async_commit, "journal_async_commit"},
894 	{Opt_abort, "abort"},
895 	{Opt_data_journal, "data=journal"},
896 	{Opt_data_ordered, "data=ordered"},
897 	{Opt_data_writeback, "data=writeback"},
898 	{Opt_data_err_abort, "data_err=abort"},
899 	{Opt_data_err_ignore, "data_err=ignore"},
900 	{Opt_offusrjquota, "usrjquota="},
901 	{Opt_usrjquota, "usrjquota=%s"},
902 	{Opt_offgrpjquota, "grpjquota="},
903 	{Opt_grpjquota, "grpjquota=%s"},
904 	{Opt_jqfmt_vfsold, "jqfmt=vfsold"},
905 	{Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
906 	{Opt_grpquota, "grpquota"},
907 	{Opt_noquota, "noquota"},
908 	{Opt_quota, "quota"},
909 	{Opt_usrquota, "usrquota"},
910 	{Opt_barrier, "barrier=%u"},
911 	{Opt_extents, "extents"},
912 	{Opt_noextents, "noextents"},
913 	{Opt_i_version, "i_version"},
914 	{Opt_stripe, "stripe=%u"},
915 	{Opt_resize, "resize"},
916 	{Opt_delalloc, "delalloc"},
917 	{Opt_nodelalloc, "nodelalloc"},
918 	{Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
919 	{Opt_err, NULL},
920 };
921 
922 static ext4_fsblk_t get_sb_block(void **data)
923 {
924 	ext4_fsblk_t	sb_block;
925 	char		*options = (char *) *data;
926 
927 	if (!options || strncmp(options, "sb=", 3) != 0)
928 		return 1;	/* Default location */
929 	options += 3;
930 	/*todo: use simple_strtoll with >32bit ext4 */
931 	sb_block = simple_strtoul(options, &options, 0);
932 	if (*options && *options != ',') {
933 		printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
934 		       (char *) *data);
935 		return 1;
936 	}
937 	if (*options == ',')
938 		options++;
939 	*data = (void *) options;
940 	return sb_block;
941 }
942 
943 static int parse_options(char *options, struct super_block *sb,
944 			 unsigned int *inum, unsigned long *journal_devnum,
945 			 ext4_fsblk_t *n_blocks_count, int is_remount)
946 {
947 	struct ext4_sb_info *sbi = EXT4_SB(sb);
948 	char *p;
949 	substring_t args[MAX_OPT_ARGS];
950 	int data_opt = 0;
951 	int option;
952 #ifdef CONFIG_QUOTA
953 	int qtype, qfmt;
954 	char *qname;
955 #endif
956 	ext4_fsblk_t last_block;
957 
958 	if (!options)
959 		return 1;
960 
961 	while ((p = strsep(&options, ",")) != NULL) {
962 		int token;
963 		if (!*p)
964 			continue;
965 
966 		token = match_token(p, tokens, args);
967 		switch (token) {
968 		case Opt_bsd_df:
969 			clear_opt(sbi->s_mount_opt, MINIX_DF);
970 			break;
971 		case Opt_minix_df:
972 			set_opt(sbi->s_mount_opt, MINIX_DF);
973 			break;
974 		case Opt_grpid:
975 			set_opt(sbi->s_mount_opt, GRPID);
976 			break;
977 		case Opt_nogrpid:
978 			clear_opt(sbi->s_mount_opt, GRPID);
979 			break;
980 		case Opt_resuid:
981 			if (match_int(&args[0], &option))
982 				return 0;
983 			sbi->s_resuid = option;
984 			break;
985 		case Opt_resgid:
986 			if (match_int(&args[0], &option))
987 				return 0;
988 			sbi->s_resgid = option;
989 			break;
990 		case Opt_sb:
991 			/* handled by get_sb_block() instead of here */
992 			/* *sb_block = match_int(&args[0]); */
993 			break;
994 		case Opt_err_panic:
995 			clear_opt(sbi->s_mount_opt, ERRORS_CONT);
996 			clear_opt(sbi->s_mount_opt, ERRORS_RO);
997 			set_opt(sbi->s_mount_opt, ERRORS_PANIC);
998 			break;
999 		case Opt_err_ro:
1000 			clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1001 			clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1002 			set_opt(sbi->s_mount_opt, ERRORS_RO);
1003 			break;
1004 		case Opt_err_cont:
1005 			clear_opt(sbi->s_mount_opt, ERRORS_RO);
1006 			clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1007 			set_opt(sbi->s_mount_opt, ERRORS_CONT);
1008 			break;
1009 		case Opt_nouid32:
1010 			set_opt(sbi->s_mount_opt, NO_UID32);
1011 			break;
1012 		case Opt_debug:
1013 			set_opt(sbi->s_mount_opt, DEBUG);
1014 			break;
1015 		case Opt_oldalloc:
1016 			set_opt(sbi->s_mount_opt, OLDALLOC);
1017 			break;
1018 		case Opt_orlov:
1019 			clear_opt(sbi->s_mount_opt, OLDALLOC);
1020 			break;
1021 #ifdef CONFIG_EXT4_FS_XATTR
1022 		case Opt_user_xattr:
1023 			set_opt(sbi->s_mount_opt, XATTR_USER);
1024 			break;
1025 		case Opt_nouser_xattr:
1026 			clear_opt(sbi->s_mount_opt, XATTR_USER);
1027 			break;
1028 #else
1029 		case Opt_user_xattr:
1030 		case Opt_nouser_xattr:
1031 			printk(KERN_ERR "EXT4 (no)user_xattr options "
1032 			       "not supported\n");
1033 			break;
1034 #endif
1035 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1036 		case Opt_acl:
1037 			set_opt(sbi->s_mount_opt, POSIX_ACL);
1038 			break;
1039 		case Opt_noacl:
1040 			clear_opt(sbi->s_mount_opt, POSIX_ACL);
1041 			break;
1042 #else
1043 		case Opt_acl:
1044 		case Opt_noacl:
1045 			printk(KERN_ERR "EXT4 (no)acl options "
1046 			       "not supported\n");
1047 			break;
1048 #endif
1049 		case Opt_reservation:
1050 			set_opt(sbi->s_mount_opt, RESERVATION);
1051 			break;
1052 		case Opt_noreservation:
1053 			clear_opt(sbi->s_mount_opt, RESERVATION);
1054 			break;
1055 		case Opt_journal_update:
1056 			/* @@@ FIXME */
1057 			/* Eventually we will want to be able to create
1058 			   a journal file here.  For now, only allow the
1059 			   user to specify an existing inode to be the
1060 			   journal file. */
1061 			if (is_remount) {
1062 				printk(KERN_ERR "EXT4-fs: cannot specify "
1063 				       "journal on remount\n");
1064 				return 0;
1065 			}
1066 			set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1067 			break;
1068 		case Opt_journal_inum:
1069 			if (is_remount) {
1070 				printk(KERN_ERR "EXT4-fs: cannot specify "
1071 				       "journal on remount\n");
1072 				return 0;
1073 			}
1074 			if (match_int(&args[0], &option))
1075 				return 0;
1076 			*inum = option;
1077 			break;
1078 		case Opt_journal_dev:
1079 			if (is_remount) {
1080 				printk(KERN_ERR "EXT4-fs: cannot specify "
1081 				       "journal on remount\n");
1082 				return 0;
1083 			}
1084 			if (match_int(&args[0], &option))
1085 				return 0;
1086 			*journal_devnum = option;
1087 			break;
1088 		case Opt_journal_checksum:
1089 			set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1090 			break;
1091 		case Opt_journal_async_commit:
1092 			set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1093 			set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1094 			break;
1095 		case Opt_noload:
1096 			set_opt(sbi->s_mount_opt, NOLOAD);
1097 			break;
1098 		case Opt_commit:
1099 			if (match_int(&args[0], &option))
1100 				return 0;
1101 			if (option < 0)
1102 				return 0;
1103 			if (option == 0)
1104 				option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1105 			sbi->s_commit_interval = HZ * option;
1106 			break;
1107 		case Opt_data_journal:
1108 			data_opt = EXT4_MOUNT_JOURNAL_DATA;
1109 			goto datacheck;
1110 		case Opt_data_ordered:
1111 			data_opt = EXT4_MOUNT_ORDERED_DATA;
1112 			goto datacheck;
1113 		case Opt_data_writeback:
1114 			data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1115 		datacheck:
1116 			if (is_remount) {
1117 				if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1118 						!= data_opt) {
1119 					printk(KERN_ERR
1120 						"EXT4-fs: cannot change data "
1121 						"mode on remount\n");
1122 					return 0;
1123 				}
1124 			} else {
1125 				sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1126 				sbi->s_mount_opt |= data_opt;
1127 			}
1128 			break;
1129 		case Opt_data_err_abort:
1130 			set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1131 			break;
1132 		case Opt_data_err_ignore:
1133 			clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1134 			break;
1135 #ifdef CONFIG_QUOTA
1136 		case Opt_usrjquota:
1137 			qtype = USRQUOTA;
1138 			goto set_qf_name;
1139 		case Opt_grpjquota:
1140 			qtype = GRPQUOTA;
1141 set_qf_name:
1142 			if ((sb_any_quota_enabled(sb) ||
1143 			     sb_any_quota_suspended(sb)) &&
1144 			    !sbi->s_qf_names[qtype]) {
1145 				printk(KERN_ERR
1146 				       "EXT4-fs: Cannot change journaled "
1147 				       "quota options when quota turned on.\n");
1148 				return 0;
1149 			}
1150 			qname = match_strdup(&args[0]);
1151 			if (!qname) {
1152 				printk(KERN_ERR
1153 					"EXT4-fs: not enough memory for "
1154 					"storing quotafile name.\n");
1155 				return 0;
1156 			}
1157 			if (sbi->s_qf_names[qtype] &&
1158 			    strcmp(sbi->s_qf_names[qtype], qname)) {
1159 				printk(KERN_ERR
1160 					"EXT4-fs: %s quota file already "
1161 					"specified.\n", QTYPE2NAME(qtype));
1162 				kfree(qname);
1163 				return 0;
1164 			}
1165 			sbi->s_qf_names[qtype] = qname;
1166 			if (strchr(sbi->s_qf_names[qtype], '/')) {
1167 				printk(KERN_ERR
1168 					"EXT4-fs: quotafile must be on "
1169 					"filesystem root.\n");
1170 				kfree(sbi->s_qf_names[qtype]);
1171 				sbi->s_qf_names[qtype] = NULL;
1172 				return 0;
1173 			}
1174 			set_opt(sbi->s_mount_opt, QUOTA);
1175 			break;
1176 		case Opt_offusrjquota:
1177 			qtype = USRQUOTA;
1178 			goto clear_qf_name;
1179 		case Opt_offgrpjquota:
1180 			qtype = GRPQUOTA;
1181 clear_qf_name:
1182 			if ((sb_any_quota_enabled(sb) ||
1183 			     sb_any_quota_suspended(sb)) &&
1184 			    sbi->s_qf_names[qtype]) {
1185 				printk(KERN_ERR "EXT4-fs: Cannot change "
1186 					"journaled quota options when "
1187 					"quota turned on.\n");
1188 				return 0;
1189 			}
1190 			/*
1191 			 * The space will be released later when all options
1192 			 * are confirmed to be correct
1193 			 */
1194 			sbi->s_qf_names[qtype] = NULL;
1195 			break;
1196 		case Opt_jqfmt_vfsold:
1197 			qfmt = QFMT_VFS_OLD;
1198 			goto set_qf_format;
1199 		case Opt_jqfmt_vfsv0:
1200 			qfmt = QFMT_VFS_V0;
1201 set_qf_format:
1202 			if ((sb_any_quota_enabled(sb) ||
1203 			     sb_any_quota_suspended(sb)) &&
1204 			    sbi->s_jquota_fmt != qfmt) {
1205 				printk(KERN_ERR "EXT4-fs: Cannot change "
1206 					"journaled quota options when "
1207 					"quota turned on.\n");
1208 				return 0;
1209 			}
1210 			sbi->s_jquota_fmt = qfmt;
1211 			break;
1212 		case Opt_quota:
1213 		case Opt_usrquota:
1214 			set_opt(sbi->s_mount_opt, QUOTA);
1215 			set_opt(sbi->s_mount_opt, USRQUOTA);
1216 			break;
1217 		case Opt_grpquota:
1218 			set_opt(sbi->s_mount_opt, QUOTA);
1219 			set_opt(sbi->s_mount_opt, GRPQUOTA);
1220 			break;
1221 		case Opt_noquota:
1222 			if (sb_any_quota_enabled(sb)) {
1223 				printk(KERN_ERR "EXT4-fs: Cannot change quota "
1224 					"options when quota turned on.\n");
1225 				return 0;
1226 			}
1227 			clear_opt(sbi->s_mount_opt, QUOTA);
1228 			clear_opt(sbi->s_mount_opt, USRQUOTA);
1229 			clear_opt(sbi->s_mount_opt, GRPQUOTA);
1230 			break;
1231 #else
1232 		case Opt_quota:
1233 		case Opt_usrquota:
1234 		case Opt_grpquota:
1235 			printk(KERN_ERR
1236 				"EXT4-fs: quota options not supported.\n");
1237 			break;
1238 		case Opt_usrjquota:
1239 		case Opt_grpjquota:
1240 		case Opt_offusrjquota:
1241 		case Opt_offgrpjquota:
1242 		case Opt_jqfmt_vfsold:
1243 		case Opt_jqfmt_vfsv0:
1244 			printk(KERN_ERR
1245 				"EXT4-fs: journaled quota options not "
1246 				"supported.\n");
1247 			break;
1248 		case Opt_noquota:
1249 			break;
1250 #endif
1251 		case Opt_abort:
1252 			set_opt(sbi->s_mount_opt, ABORT);
1253 			break;
1254 		case Opt_barrier:
1255 			if (match_int(&args[0], &option))
1256 				return 0;
1257 			if (option)
1258 				set_opt(sbi->s_mount_opt, BARRIER);
1259 			else
1260 				clear_opt(sbi->s_mount_opt, BARRIER);
1261 			break;
1262 		case Opt_ignore:
1263 			break;
1264 		case Opt_resize:
1265 			if (!is_remount) {
1266 				printk("EXT4-fs: resize option only available "
1267 					"for remount\n");
1268 				return 0;
1269 			}
1270 			if (match_int(&args[0], &option) != 0)
1271 				return 0;
1272 			*n_blocks_count = option;
1273 			break;
1274 		case Opt_nobh:
1275 			set_opt(sbi->s_mount_opt, NOBH);
1276 			break;
1277 		case Opt_bh:
1278 			clear_opt(sbi->s_mount_opt, NOBH);
1279 			break;
1280 		case Opt_extents:
1281 			if (!EXT4_HAS_INCOMPAT_FEATURE(sb,
1282 					EXT4_FEATURE_INCOMPAT_EXTENTS)) {
1283 				ext4_warning(sb, __func__,
1284 					"extents feature not enabled "
1285 					"on this filesystem, use tune2fs\n");
1286 				return 0;
1287 			}
1288 			set_opt(sbi->s_mount_opt, EXTENTS);
1289 			break;
1290 		case Opt_noextents:
1291 			/*
1292 			 * When e2fsprogs support resizing an already existing
1293 			 * ext3 file system to greater than 2**32 we need to
1294 			 * add support to block allocator to handle growing
1295 			 * already existing block  mapped inode so that blocks
1296 			 * allocated for them fall within 2**32
1297 			 */
1298 			last_block = ext4_blocks_count(sbi->s_es) - 1;
1299 			if (last_block  > 0xffffffffULL) {
1300 				printk(KERN_ERR "EXT4-fs: Filesystem too "
1301 						"large to mount with "
1302 						"-o noextents options\n");
1303 				return 0;
1304 			}
1305 			clear_opt(sbi->s_mount_opt, EXTENTS);
1306 			break;
1307 		case Opt_i_version:
1308 			set_opt(sbi->s_mount_opt, I_VERSION);
1309 			sb->s_flags |= MS_I_VERSION;
1310 			break;
1311 		case Opt_nodelalloc:
1312 			clear_opt(sbi->s_mount_opt, DELALLOC);
1313 			break;
1314 		case Opt_stripe:
1315 			if (match_int(&args[0], &option))
1316 				return 0;
1317 			if (option < 0)
1318 				return 0;
1319 			sbi->s_stripe = option;
1320 			break;
1321 		case Opt_delalloc:
1322 			set_opt(sbi->s_mount_opt, DELALLOC);
1323 			break;
1324 		case Opt_inode_readahead_blks:
1325 			if (match_int(&args[0], &option))
1326 				return 0;
1327 			if (option < 0 || option > (1 << 30))
1328 				return 0;
1329 			sbi->s_inode_readahead_blks = option;
1330 			break;
1331 		default:
1332 			printk(KERN_ERR
1333 			       "EXT4-fs: Unrecognized mount option \"%s\" "
1334 			       "or missing value\n", p);
1335 			return 0;
1336 		}
1337 	}
1338 #ifdef CONFIG_QUOTA
1339 	if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1340 		if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1341 		     sbi->s_qf_names[USRQUOTA])
1342 			clear_opt(sbi->s_mount_opt, USRQUOTA);
1343 
1344 		if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1345 		     sbi->s_qf_names[GRPQUOTA])
1346 			clear_opt(sbi->s_mount_opt, GRPQUOTA);
1347 
1348 		if ((sbi->s_qf_names[USRQUOTA] &&
1349 				(sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1350 		    (sbi->s_qf_names[GRPQUOTA] &&
1351 				(sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1352 			printk(KERN_ERR "EXT4-fs: old and new quota "
1353 					"format mixing.\n");
1354 			return 0;
1355 		}
1356 
1357 		if (!sbi->s_jquota_fmt) {
1358 			printk(KERN_ERR "EXT4-fs: journaled quota format "
1359 					"not specified.\n");
1360 			return 0;
1361 		}
1362 	} else {
1363 		if (sbi->s_jquota_fmt) {
1364 			printk(KERN_ERR "EXT4-fs: journaled quota format "
1365 					"specified with no journaling "
1366 					"enabled.\n");
1367 			return 0;
1368 		}
1369 	}
1370 #endif
1371 	return 1;
1372 }
1373 
1374 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1375 			    int read_only)
1376 {
1377 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1378 	int res = 0;
1379 
1380 	if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1381 		printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1382 		       "forcing read-only mode\n");
1383 		res = MS_RDONLY;
1384 	}
1385 	if (read_only)
1386 		return res;
1387 	if (!(sbi->s_mount_state & EXT4_VALID_FS))
1388 		printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1389 		       "running e2fsck is recommended\n");
1390 	else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1391 		printk(KERN_WARNING
1392 		       "EXT4-fs warning: mounting fs with errors, "
1393 		       "running e2fsck is recommended\n");
1394 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1395 		 le16_to_cpu(es->s_mnt_count) >=
1396 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1397 		printk(KERN_WARNING
1398 		       "EXT4-fs warning: maximal mount count reached, "
1399 		       "running e2fsck is recommended\n");
1400 	else if (le32_to_cpu(es->s_checkinterval) &&
1401 		(le32_to_cpu(es->s_lastcheck) +
1402 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1403 		printk(KERN_WARNING
1404 		       "EXT4-fs warning: checktime reached, "
1405 		       "running e2fsck is recommended\n");
1406 #if 0
1407 		/* @@@ We _will_ want to clear the valid bit if we find
1408 		 * inconsistencies, to force a fsck at reboot.  But for
1409 		 * a plain journaled filesystem we can keep it set as
1410 		 * valid forever! :)
1411 		 */
1412 	es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1413 #endif
1414 	if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1415 		es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1416 	le16_add_cpu(&es->s_mnt_count, 1);
1417 	es->s_mtime = cpu_to_le32(get_seconds());
1418 	ext4_update_dynamic_rev(sb);
1419 	EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1420 
1421 	ext4_commit_super(sb, es, 1);
1422 	if (test_opt(sb, DEBUG))
1423 		printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
1424 				"bpg=%lu, ipg=%lu, mo=%04lx]\n",
1425 			sb->s_blocksize,
1426 			sbi->s_groups_count,
1427 			EXT4_BLOCKS_PER_GROUP(sb),
1428 			EXT4_INODES_PER_GROUP(sb),
1429 			sbi->s_mount_opt);
1430 
1431 	printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1432 	       sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1433 	       "external", EXT4_SB(sb)->s_journal->j_devname);
1434 	return res;
1435 }
1436 
1437 static int ext4_fill_flex_info(struct super_block *sb)
1438 {
1439 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1440 	struct ext4_group_desc *gdp = NULL;
1441 	struct buffer_head *bh;
1442 	ext4_group_t flex_group_count;
1443 	ext4_group_t flex_group;
1444 	int groups_per_flex = 0;
1445 	__u64 block_bitmap = 0;
1446 	int i;
1447 
1448 	if (!sbi->s_es->s_log_groups_per_flex) {
1449 		sbi->s_log_groups_per_flex = 0;
1450 		return 1;
1451 	}
1452 
1453 	sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1454 	groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1455 
1456 	/* We allocate both existing and potentially added groups */
1457 	flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1458 			    ((sbi->s_es->s_reserved_gdt_blocks +1 ) <<
1459 			      EXT4_DESC_PER_BLOCK_BITS(sb))) /
1460 			   groups_per_flex;
1461 	sbi->s_flex_groups = kzalloc(flex_group_count *
1462 				     sizeof(struct flex_groups), GFP_KERNEL);
1463 	if (sbi->s_flex_groups == NULL) {
1464 		printk(KERN_ERR "EXT4-fs: not enough memory for "
1465 				"%lu flex groups\n", flex_group_count);
1466 		goto failed;
1467 	}
1468 
1469 	gdp = ext4_get_group_desc(sb, 1, &bh);
1470 	block_bitmap = ext4_block_bitmap(sb, gdp) - 1;
1471 
1472 	for (i = 0; i < sbi->s_groups_count; i++) {
1473 		gdp = ext4_get_group_desc(sb, i, &bh);
1474 
1475 		flex_group = ext4_flex_group(sbi, i);
1476 		sbi->s_flex_groups[flex_group].free_inodes +=
1477 			le16_to_cpu(gdp->bg_free_inodes_count);
1478 		sbi->s_flex_groups[flex_group].free_blocks +=
1479 			le16_to_cpu(gdp->bg_free_blocks_count);
1480 	}
1481 
1482 	return 1;
1483 failed:
1484 	return 0;
1485 }
1486 
1487 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1488 			    struct ext4_group_desc *gdp)
1489 {
1490 	__u16 crc = 0;
1491 
1492 	if (sbi->s_es->s_feature_ro_compat &
1493 	    cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1494 		int offset = offsetof(struct ext4_group_desc, bg_checksum);
1495 		__le32 le_group = cpu_to_le32(block_group);
1496 
1497 		crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1498 		crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1499 		crc = crc16(crc, (__u8 *)gdp, offset);
1500 		offset += sizeof(gdp->bg_checksum); /* skip checksum */
1501 		/* for checksum of struct ext4_group_desc do the rest...*/
1502 		if ((sbi->s_es->s_feature_incompat &
1503 		     cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1504 		    offset < le16_to_cpu(sbi->s_es->s_desc_size))
1505 			crc = crc16(crc, (__u8 *)gdp + offset,
1506 				    le16_to_cpu(sbi->s_es->s_desc_size) -
1507 					offset);
1508 	}
1509 
1510 	return cpu_to_le16(crc);
1511 }
1512 
1513 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1514 				struct ext4_group_desc *gdp)
1515 {
1516 	if ((sbi->s_es->s_feature_ro_compat &
1517 	     cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1518 	    (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1519 		return 0;
1520 
1521 	return 1;
1522 }
1523 
1524 /* Called at mount-time, super-block is locked */
1525 static int ext4_check_descriptors(struct super_block *sb)
1526 {
1527 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1528 	ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1529 	ext4_fsblk_t last_block;
1530 	ext4_fsblk_t block_bitmap;
1531 	ext4_fsblk_t inode_bitmap;
1532 	ext4_fsblk_t inode_table;
1533 	int flexbg_flag = 0;
1534 	ext4_group_t i;
1535 
1536 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1537 		flexbg_flag = 1;
1538 
1539 	ext4_debug("Checking group descriptors");
1540 
1541 	for (i = 0; i < sbi->s_groups_count; i++) {
1542 		struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1543 
1544 		if (i == sbi->s_groups_count - 1 || flexbg_flag)
1545 			last_block = ext4_blocks_count(sbi->s_es) - 1;
1546 		else
1547 			last_block = first_block +
1548 				(EXT4_BLOCKS_PER_GROUP(sb) - 1);
1549 
1550 		block_bitmap = ext4_block_bitmap(sb, gdp);
1551 		if (block_bitmap < first_block || block_bitmap > last_block) {
1552 			printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1553 			       "Block bitmap for group %lu not in group "
1554 			       "(block %llu)!\n", i, block_bitmap);
1555 			return 0;
1556 		}
1557 		inode_bitmap = ext4_inode_bitmap(sb, gdp);
1558 		if (inode_bitmap < first_block || inode_bitmap > last_block) {
1559 			printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1560 			       "Inode bitmap for group %lu not in group "
1561 			       "(block %llu)!\n", i, inode_bitmap);
1562 			return 0;
1563 		}
1564 		inode_table = ext4_inode_table(sb, gdp);
1565 		if (inode_table < first_block ||
1566 		    inode_table + sbi->s_itb_per_group - 1 > last_block) {
1567 			printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1568 			       "Inode table for group %lu not in group "
1569 			       "(block %llu)!\n", i, inode_table);
1570 			return 0;
1571 		}
1572 		spin_lock(sb_bgl_lock(sbi, i));
1573 		if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1574 			printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1575 			       "Checksum for group %lu failed (%u!=%u)\n",
1576 			       i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1577 			       gdp)), le16_to_cpu(gdp->bg_checksum));
1578 			if (!(sb->s_flags & MS_RDONLY)) {
1579 				spin_unlock(sb_bgl_lock(sbi, i));
1580 				return 0;
1581 			}
1582 		}
1583 		spin_unlock(sb_bgl_lock(sbi, i));
1584 		if (!flexbg_flag)
1585 			first_block += EXT4_BLOCKS_PER_GROUP(sb);
1586 	}
1587 
1588 	ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1589 	sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
1590 	return 1;
1591 }
1592 
1593 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1594  * the superblock) which were deleted from all directories, but held open by
1595  * a process at the time of a crash.  We walk the list and try to delete these
1596  * inodes at recovery time (only with a read-write filesystem).
1597  *
1598  * In order to keep the orphan inode chain consistent during traversal (in
1599  * case of crash during recovery), we link each inode into the superblock
1600  * orphan list_head and handle it the same way as an inode deletion during
1601  * normal operation (which journals the operations for us).
1602  *
1603  * We only do an iget() and an iput() on each inode, which is very safe if we
1604  * accidentally point at an in-use or already deleted inode.  The worst that
1605  * can happen in this case is that we get a "bit already cleared" message from
1606  * ext4_free_inode().  The only reason we would point at a wrong inode is if
1607  * e2fsck was run on this filesystem, and it must have already done the orphan
1608  * inode cleanup for us, so we can safely abort without any further action.
1609  */
1610 static void ext4_orphan_cleanup(struct super_block *sb,
1611 				struct ext4_super_block *es)
1612 {
1613 	unsigned int s_flags = sb->s_flags;
1614 	int nr_orphans = 0, nr_truncates = 0;
1615 #ifdef CONFIG_QUOTA
1616 	int i;
1617 #endif
1618 	if (!es->s_last_orphan) {
1619 		jbd_debug(4, "no orphan inodes to clean up\n");
1620 		return;
1621 	}
1622 
1623 	if (bdev_read_only(sb->s_bdev)) {
1624 		printk(KERN_ERR "EXT4-fs: write access "
1625 			"unavailable, skipping orphan cleanup.\n");
1626 		return;
1627 	}
1628 
1629 	if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1630 		if (es->s_last_orphan)
1631 			jbd_debug(1, "Errors on filesystem, "
1632 				  "clearing orphan list.\n");
1633 		es->s_last_orphan = 0;
1634 		jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1635 		return;
1636 	}
1637 
1638 	if (s_flags & MS_RDONLY) {
1639 		printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1640 		       sb->s_id);
1641 		sb->s_flags &= ~MS_RDONLY;
1642 	}
1643 #ifdef CONFIG_QUOTA
1644 	/* Needed for iput() to work correctly and not trash data */
1645 	sb->s_flags |= MS_ACTIVE;
1646 	/* Turn on quotas so that they are updated correctly */
1647 	for (i = 0; i < MAXQUOTAS; i++) {
1648 		if (EXT4_SB(sb)->s_qf_names[i]) {
1649 			int ret = ext4_quota_on_mount(sb, i);
1650 			if (ret < 0)
1651 				printk(KERN_ERR
1652 					"EXT4-fs: Cannot turn on journaled "
1653 					"quota: error %d\n", ret);
1654 		}
1655 	}
1656 #endif
1657 
1658 	while (es->s_last_orphan) {
1659 		struct inode *inode;
1660 
1661 		inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1662 		if (IS_ERR(inode)) {
1663 			es->s_last_orphan = 0;
1664 			break;
1665 		}
1666 
1667 		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1668 		DQUOT_INIT(inode);
1669 		if (inode->i_nlink) {
1670 			printk(KERN_DEBUG
1671 				"%s: truncating inode %lu to %lld bytes\n",
1672 				__func__, inode->i_ino, inode->i_size);
1673 			jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1674 				  inode->i_ino, inode->i_size);
1675 			ext4_truncate(inode);
1676 			nr_truncates++;
1677 		} else {
1678 			printk(KERN_DEBUG
1679 				"%s: deleting unreferenced inode %lu\n",
1680 				__func__, inode->i_ino);
1681 			jbd_debug(2, "deleting unreferenced inode %lu\n",
1682 				  inode->i_ino);
1683 			nr_orphans++;
1684 		}
1685 		iput(inode);  /* The delete magic happens here! */
1686 	}
1687 
1688 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1689 
1690 	if (nr_orphans)
1691 		printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1692 		       sb->s_id, PLURAL(nr_orphans));
1693 	if (nr_truncates)
1694 		printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1695 		       sb->s_id, PLURAL(nr_truncates));
1696 #ifdef CONFIG_QUOTA
1697 	/* Turn quotas off */
1698 	for (i = 0; i < MAXQUOTAS; i++) {
1699 		if (sb_dqopt(sb)->files[i])
1700 			vfs_quota_off(sb, i, 0);
1701 	}
1702 #endif
1703 	sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1704 }
1705 /*
1706  * Maximal extent format file size.
1707  * Resulting logical blkno at s_maxbytes must fit in our on-disk
1708  * extent format containers, within a sector_t, and within i_blocks
1709  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
1710  * so that won't be a limiting factor.
1711  *
1712  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1713  */
1714 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1715 {
1716 	loff_t res;
1717 	loff_t upper_limit = MAX_LFS_FILESIZE;
1718 
1719 	/* small i_blocks in vfs inode? */
1720 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1721 		/*
1722 		 * CONFIG_LSF is not enabled implies the inode
1723 		 * i_block represent total blocks in 512 bytes
1724 		 * 32 == size of vfs inode i_blocks * 8
1725 		 */
1726 		upper_limit = (1LL << 32) - 1;
1727 
1728 		/* total blocks in file system block size */
1729 		upper_limit >>= (blkbits - 9);
1730 		upper_limit <<= blkbits;
1731 	}
1732 
1733 	/* 32-bit extent-start container, ee_block */
1734 	res = 1LL << 32;
1735 	res <<= blkbits;
1736 	res -= 1;
1737 
1738 	/* Sanity check against vm- & vfs- imposed limits */
1739 	if (res > upper_limit)
1740 		res = upper_limit;
1741 
1742 	return res;
1743 }
1744 
1745 /*
1746  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
1747  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1748  * We need to be 1 filesystem block less than the 2^48 sector limit.
1749  */
1750 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1751 {
1752 	loff_t res = EXT4_NDIR_BLOCKS;
1753 	int meta_blocks;
1754 	loff_t upper_limit;
1755 	/* This is calculated to be the largest file size for a
1756 	 * dense, bitmapped file such that the total number of
1757 	 * sectors in the file, including data and all indirect blocks,
1758 	 * does not exceed 2^48 -1
1759 	 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1760 	 * total number of  512 bytes blocks of the file
1761 	 */
1762 
1763 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1764 		/*
1765 		 * !has_huge_files or CONFIG_LSF is not enabled
1766 		 * implies the inode i_block represent total blocks in
1767 		 * 512 bytes 32 == size of vfs inode i_blocks * 8
1768 		 */
1769 		upper_limit = (1LL << 32) - 1;
1770 
1771 		/* total blocks in file system block size */
1772 		upper_limit >>= (bits - 9);
1773 
1774 	} else {
1775 		/*
1776 		 * We use 48 bit ext4_inode i_blocks
1777 		 * With EXT4_HUGE_FILE_FL set the i_blocks
1778 		 * represent total number of blocks in
1779 		 * file system block size
1780 		 */
1781 		upper_limit = (1LL << 48) - 1;
1782 
1783 	}
1784 
1785 	/* indirect blocks */
1786 	meta_blocks = 1;
1787 	/* double indirect blocks */
1788 	meta_blocks += 1 + (1LL << (bits-2));
1789 	/* tripple indirect blocks */
1790 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1791 
1792 	upper_limit -= meta_blocks;
1793 	upper_limit <<= bits;
1794 
1795 	res += 1LL << (bits-2);
1796 	res += 1LL << (2*(bits-2));
1797 	res += 1LL << (3*(bits-2));
1798 	res <<= bits;
1799 	if (res > upper_limit)
1800 		res = upper_limit;
1801 
1802 	if (res > MAX_LFS_FILESIZE)
1803 		res = MAX_LFS_FILESIZE;
1804 
1805 	return res;
1806 }
1807 
1808 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1809 				ext4_fsblk_t logical_sb_block, int nr)
1810 {
1811 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1812 	ext4_group_t bg, first_meta_bg;
1813 	int has_super = 0;
1814 
1815 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1816 
1817 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1818 	    nr < first_meta_bg)
1819 		return logical_sb_block + nr + 1;
1820 	bg = sbi->s_desc_per_block * nr;
1821 	if (ext4_bg_has_super(sb, bg))
1822 		has_super = 1;
1823 	return (has_super + ext4_group_first_block_no(sb, bg));
1824 }
1825 
1826 /**
1827  * ext4_get_stripe_size: Get the stripe size.
1828  * @sbi: In memory super block info
1829  *
1830  * If we have specified it via mount option, then
1831  * use the mount option value. If the value specified at mount time is
1832  * greater than the blocks per group use the super block value.
1833  * If the super block value is greater than blocks per group return 0.
1834  * Allocator needs it be less than blocks per group.
1835  *
1836  */
1837 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1838 {
1839 	unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1840 	unsigned long stripe_width =
1841 			le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1842 
1843 	if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1844 		return sbi->s_stripe;
1845 
1846 	if (stripe_width <= sbi->s_blocks_per_group)
1847 		return stripe_width;
1848 
1849 	if (stride <= sbi->s_blocks_per_group)
1850 		return stride;
1851 
1852 	return 0;
1853 }
1854 
1855 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
1856 				__releases(kernel_lock)
1857 				__acquires(kernel_lock)
1858 
1859 {
1860 	struct buffer_head *bh;
1861 	struct ext4_super_block *es = NULL;
1862 	struct ext4_sb_info *sbi;
1863 	ext4_fsblk_t block;
1864 	ext4_fsblk_t sb_block = get_sb_block(&data);
1865 	ext4_fsblk_t logical_sb_block;
1866 	unsigned long offset = 0;
1867 	unsigned int journal_inum = 0;
1868 	unsigned long journal_devnum = 0;
1869 	unsigned long def_mount_opts;
1870 	struct inode *root;
1871 	char *cp;
1872 	int ret = -EINVAL;
1873 	int blocksize;
1874 	int db_count;
1875 	int i;
1876 	int needs_recovery, has_huge_files;
1877 	__le32 features;
1878 	__u64 blocks_count;
1879 	int err;
1880 
1881 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1882 	if (!sbi)
1883 		return -ENOMEM;
1884 	sb->s_fs_info = sbi;
1885 	sbi->s_mount_opt = 0;
1886 	sbi->s_resuid = EXT4_DEF_RESUID;
1887 	sbi->s_resgid = EXT4_DEF_RESGID;
1888 	sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
1889 	sbi->s_sb_block = sb_block;
1890 
1891 	unlock_kernel();
1892 
1893 	/* Cleanup superblock name */
1894 	for (cp = sb->s_id; (cp = strchr(cp, '/'));)
1895 		*cp = '!';
1896 
1897 	blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
1898 	if (!blocksize) {
1899 		printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
1900 		goto out_fail;
1901 	}
1902 
1903 	/*
1904 	 * The ext4 superblock will not be buffer aligned for other than 1kB
1905 	 * block sizes.  We need to calculate the offset from buffer start.
1906 	 */
1907 	if (blocksize != EXT4_MIN_BLOCK_SIZE) {
1908 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1909 		offset = do_div(logical_sb_block, blocksize);
1910 	} else {
1911 		logical_sb_block = sb_block;
1912 	}
1913 
1914 	if (!(bh = sb_bread(sb, logical_sb_block))) {
1915 		printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
1916 		goto out_fail;
1917 	}
1918 	/*
1919 	 * Note: s_es must be initialized as soon as possible because
1920 	 *       some ext4 macro-instructions depend on its value
1921 	 */
1922 	es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
1923 	sbi->s_es = es;
1924 	sb->s_magic = le16_to_cpu(es->s_magic);
1925 	if (sb->s_magic != EXT4_SUPER_MAGIC)
1926 		goto cantfind_ext4;
1927 
1928 	/* Set defaults before we parse the mount options */
1929 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1930 	if (def_mount_opts & EXT4_DEFM_DEBUG)
1931 		set_opt(sbi->s_mount_opt, DEBUG);
1932 	if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
1933 		set_opt(sbi->s_mount_opt, GRPID);
1934 	if (def_mount_opts & EXT4_DEFM_UID16)
1935 		set_opt(sbi->s_mount_opt, NO_UID32);
1936 #ifdef CONFIG_EXT4_FS_XATTR
1937 	if (def_mount_opts & EXT4_DEFM_XATTR_USER)
1938 		set_opt(sbi->s_mount_opt, XATTR_USER);
1939 #endif
1940 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1941 	if (def_mount_opts & EXT4_DEFM_ACL)
1942 		set_opt(sbi->s_mount_opt, POSIX_ACL);
1943 #endif
1944 	if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1945 		sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1946 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1947 		sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1948 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1949 		sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1950 
1951 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
1952 		set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1953 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
1954 		set_opt(sbi->s_mount_opt, ERRORS_CONT);
1955 	else
1956 		set_opt(sbi->s_mount_opt, ERRORS_RO);
1957 
1958 	sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1959 	sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1960 
1961 	set_opt(sbi->s_mount_opt, RESERVATION);
1962 	set_opt(sbi->s_mount_opt, BARRIER);
1963 
1964 	/*
1965 	 * turn on extents feature by default in ext4 filesystem
1966 	 * only if feature flag already set by mkfs or tune2fs.
1967 	 * Use -o noextents to turn it off
1968 	 */
1969 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
1970 		set_opt(sbi->s_mount_opt, EXTENTS);
1971 	else
1972 		ext4_warning(sb, __func__,
1973 			"extents feature not enabled on this filesystem, "
1974 			"use tune2fs.\n");
1975 
1976 	/*
1977 	 * enable delayed allocation by default
1978 	 * Use -o nodelalloc to turn it off
1979 	 */
1980 	set_opt(sbi->s_mount_opt, DELALLOC);
1981 
1982 
1983 	if (!parse_options((char *) data, sb, &journal_inum, &journal_devnum,
1984 			   NULL, 0))
1985 		goto failed_mount;
1986 
1987 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1988 		((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1989 
1990 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1991 	    (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1992 	     EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1993 	     EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1994 		printk(KERN_WARNING
1995 		       "EXT4-fs warning: feature flags set on rev 0 fs, "
1996 		       "running e2fsck is recommended\n");
1997 
1998 	/*
1999 	 * Check feature flags regardless of the revision level, since we
2000 	 * previously didn't change the revision level when setting the flags,
2001 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
2002 	 */
2003 	features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2004 	if (features) {
2005 		printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
2006 		       "unsupported optional features (%x).\n",
2007 		       sb->s_id, le32_to_cpu(features));
2008 		goto failed_mount;
2009 	}
2010 	features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2011 	if (!(sb->s_flags & MS_RDONLY) && features) {
2012 		printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
2013 		       "unsupported optional features (%x).\n",
2014 		       sb->s_id, le32_to_cpu(features));
2015 		goto failed_mount;
2016 	}
2017 	has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2018 				    EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2019 	if (has_huge_files) {
2020 		/*
2021 		 * Large file size enabled file system can only be
2022 		 * mount if kernel is build with CONFIG_LSF
2023 		 */
2024 		if (sizeof(root->i_blocks) < sizeof(u64) &&
2025 				!(sb->s_flags & MS_RDONLY)) {
2026 			printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2027 					"files cannot be mounted read-write "
2028 					"without CONFIG_LSF.\n", sb->s_id);
2029 			goto failed_mount;
2030 		}
2031 	}
2032 	blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2033 
2034 	if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2035 	    blocksize > EXT4_MAX_BLOCK_SIZE) {
2036 		printk(KERN_ERR
2037 		       "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2038 		       blocksize, sb->s_id);
2039 		goto failed_mount;
2040 	}
2041 
2042 	if (sb->s_blocksize != blocksize) {
2043 
2044 		/* Validate the filesystem blocksize */
2045 		if (!sb_set_blocksize(sb, blocksize)) {
2046 			printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2047 					blocksize);
2048 			goto failed_mount;
2049 		}
2050 
2051 		brelse(bh);
2052 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2053 		offset = do_div(logical_sb_block, blocksize);
2054 		bh = sb_bread(sb, logical_sb_block);
2055 		if (!bh) {
2056 			printk(KERN_ERR
2057 			       "EXT4-fs: Can't read superblock on 2nd try.\n");
2058 			goto failed_mount;
2059 		}
2060 		es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2061 		sbi->s_es = es;
2062 		if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2063 			printk(KERN_ERR
2064 			       "EXT4-fs: Magic mismatch, very weird !\n");
2065 			goto failed_mount;
2066 		}
2067 	}
2068 
2069 	sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2070 						      has_huge_files);
2071 	sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2072 
2073 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2074 		sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2075 		sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2076 	} else {
2077 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2078 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2079 		if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2080 		    (!is_power_of_2(sbi->s_inode_size)) ||
2081 		    (sbi->s_inode_size > blocksize)) {
2082 			printk(KERN_ERR
2083 			       "EXT4-fs: unsupported inode size: %d\n",
2084 			       sbi->s_inode_size);
2085 			goto failed_mount;
2086 		}
2087 		if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2088 			sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2089 	}
2090 	sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2091 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2092 		if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2093 		    sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2094 		    !is_power_of_2(sbi->s_desc_size)) {
2095 			printk(KERN_ERR
2096 			       "EXT4-fs: unsupported descriptor size %lu\n",
2097 			       sbi->s_desc_size);
2098 			goto failed_mount;
2099 		}
2100 	} else
2101 		sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2102 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2103 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2104 	if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2105 		goto cantfind_ext4;
2106 	sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2107 	if (sbi->s_inodes_per_block == 0)
2108 		goto cantfind_ext4;
2109 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
2110 					sbi->s_inodes_per_block;
2111 	sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2112 	sbi->s_sbh = bh;
2113 	sbi->s_mount_state = le16_to_cpu(es->s_state);
2114 	sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2115 	sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2116 	for (i = 0; i < 4; i++)
2117 		sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2118 	sbi->s_def_hash_version = es->s_def_hash_version;
2119 
2120 	if (sbi->s_blocks_per_group > blocksize * 8) {
2121 		printk(KERN_ERR
2122 		       "EXT4-fs: #blocks per group too big: %lu\n",
2123 		       sbi->s_blocks_per_group);
2124 		goto failed_mount;
2125 	}
2126 	if (sbi->s_inodes_per_group > blocksize * 8) {
2127 		printk(KERN_ERR
2128 		       "EXT4-fs: #inodes per group too big: %lu\n",
2129 		       sbi->s_inodes_per_group);
2130 		goto failed_mount;
2131 	}
2132 
2133 	if (ext4_blocks_count(es) >
2134 		    (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2135 		printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2136 			" too large to mount safely\n", sb->s_id);
2137 		if (sizeof(sector_t) < 8)
2138 			printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2139 					"enabled\n");
2140 		goto failed_mount;
2141 	}
2142 
2143 	if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2144 		goto cantfind_ext4;
2145 
2146 	/* ensure blocks_count calculation below doesn't sign-extend */
2147 	if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2148 	    le32_to_cpu(es->s_first_data_block) + 1) {
2149 		printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2150 		       "first data block %u, blocks per group %lu\n",
2151 			ext4_blocks_count(es),
2152 			le32_to_cpu(es->s_first_data_block),
2153 			EXT4_BLOCKS_PER_GROUP(sb));
2154 		goto failed_mount;
2155 	}
2156 	blocks_count = (ext4_blocks_count(es) -
2157 			le32_to_cpu(es->s_first_data_block) +
2158 			EXT4_BLOCKS_PER_GROUP(sb) - 1);
2159 	do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2160 	sbi->s_groups_count = blocks_count;
2161 	db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2162 		   EXT4_DESC_PER_BLOCK(sb);
2163 	sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2164 				    GFP_KERNEL);
2165 	if (sbi->s_group_desc == NULL) {
2166 		printk(KERN_ERR "EXT4-fs: not enough memory\n");
2167 		goto failed_mount;
2168 	}
2169 
2170 #ifdef CONFIG_PROC_FS
2171 	if (ext4_proc_root)
2172 		sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2173 
2174 	if (sbi->s_proc)
2175 		proc_create_data("inode_readahead_blks", 0644, sbi->s_proc,
2176 				 &ext4_ui_proc_fops,
2177 				 &sbi->s_inode_readahead_blks);
2178 #endif
2179 
2180 	bgl_lock_init(&sbi->s_blockgroup_lock);
2181 
2182 	for (i = 0; i < db_count; i++) {
2183 		block = descriptor_loc(sb, logical_sb_block, i);
2184 		sbi->s_group_desc[i] = sb_bread(sb, block);
2185 		if (!sbi->s_group_desc[i]) {
2186 			printk(KERN_ERR "EXT4-fs: "
2187 			       "can't read group descriptor %d\n", i);
2188 			db_count = i;
2189 			goto failed_mount2;
2190 		}
2191 	}
2192 	if (!ext4_check_descriptors(sb)) {
2193 		printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2194 		goto failed_mount2;
2195 	}
2196 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2197 		if (!ext4_fill_flex_info(sb)) {
2198 			printk(KERN_ERR
2199 			       "EXT4-fs: unable to initialize "
2200 			       "flex_bg meta info!\n");
2201 			goto failed_mount2;
2202 		}
2203 
2204 	sbi->s_gdb_count = db_count;
2205 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2206 	spin_lock_init(&sbi->s_next_gen_lock);
2207 
2208 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
2209 			ext4_count_free_blocks(sb));
2210 	if (!err) {
2211 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
2212 				ext4_count_free_inodes(sb));
2213 	}
2214 	if (!err) {
2215 		err = percpu_counter_init(&sbi->s_dirs_counter,
2216 				ext4_count_dirs(sb));
2217 	}
2218 	if (!err) {
2219 		err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2220 	}
2221 	if (err) {
2222 		printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2223 		goto failed_mount3;
2224 	}
2225 
2226 	sbi->s_stripe = ext4_get_stripe_size(sbi);
2227 
2228 	/*
2229 	 * set up enough so that it can read an inode
2230 	 */
2231 	sb->s_op = &ext4_sops;
2232 	sb->s_export_op = &ext4_export_ops;
2233 	sb->s_xattr = ext4_xattr_handlers;
2234 #ifdef CONFIG_QUOTA
2235 	sb->s_qcop = &ext4_qctl_operations;
2236 	sb->dq_op = &ext4_quota_operations;
2237 #endif
2238 	INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2239 
2240 	sb->s_root = NULL;
2241 
2242 	needs_recovery = (es->s_last_orphan != 0 ||
2243 			  EXT4_HAS_INCOMPAT_FEATURE(sb,
2244 				    EXT4_FEATURE_INCOMPAT_RECOVER));
2245 
2246 	/*
2247 	 * The first inode we look at is the journal inode.  Don't try
2248 	 * root first: it may be modified in the journal!
2249 	 */
2250 	if (!test_opt(sb, NOLOAD) &&
2251 	    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2252 		if (ext4_load_journal(sb, es, journal_devnum))
2253 			goto failed_mount3;
2254 		if (!(sb->s_flags & MS_RDONLY) &&
2255 		    EXT4_SB(sb)->s_journal->j_failed_commit) {
2256 			printk(KERN_CRIT "EXT4-fs error (device %s): "
2257 			       "ext4_fill_super: Journal transaction "
2258 			       "%u is corrupt\n", sb->s_id,
2259 			       EXT4_SB(sb)->s_journal->j_failed_commit);
2260 			if (test_opt(sb, ERRORS_RO)) {
2261 				printk(KERN_CRIT
2262 				       "Mounting filesystem read-only\n");
2263 				sb->s_flags |= MS_RDONLY;
2264 				EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2265 				es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2266 			}
2267 			if (test_opt(sb, ERRORS_PANIC)) {
2268 				EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2269 				es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2270 				ext4_commit_super(sb, es, 1);
2271 				printk(KERN_CRIT
2272 				       "EXT4-fs (device %s): mount failed\n",
2273 				      sb->s_id);
2274 				goto failed_mount4;
2275 			}
2276 		}
2277 	} else if (journal_inum) {
2278 		if (ext4_create_journal(sb, es, journal_inum))
2279 			goto failed_mount3;
2280 	} else {
2281 		if (!silent)
2282 			printk(KERN_ERR
2283 			       "ext4: No journal on filesystem on %s\n",
2284 			       sb->s_id);
2285 		goto failed_mount3;
2286 	}
2287 
2288 	if (ext4_blocks_count(es) > 0xffffffffULL &&
2289 	    !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2290 				       JBD2_FEATURE_INCOMPAT_64BIT)) {
2291 		printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2292 		goto failed_mount4;
2293 	}
2294 
2295 	if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2296 		jbd2_journal_set_features(sbi->s_journal,
2297 				JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2298 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2299 	} else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2300 		jbd2_journal_set_features(sbi->s_journal,
2301 				JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2302 		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2303 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2304 	} else {
2305 		jbd2_journal_clear_features(sbi->s_journal,
2306 				JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2307 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2308 	}
2309 
2310 	/* We have now updated the journal if required, so we can
2311 	 * validate the data journaling mode. */
2312 	switch (test_opt(sb, DATA_FLAGS)) {
2313 	case 0:
2314 		/* No mode set, assume a default based on the journal
2315 		 * capabilities: ORDERED_DATA if the journal can
2316 		 * cope, else JOURNAL_DATA
2317 		 */
2318 		if (jbd2_journal_check_available_features
2319 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2320 			set_opt(sbi->s_mount_opt, ORDERED_DATA);
2321 		else
2322 			set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2323 		break;
2324 
2325 	case EXT4_MOUNT_ORDERED_DATA:
2326 	case EXT4_MOUNT_WRITEBACK_DATA:
2327 		if (!jbd2_journal_check_available_features
2328 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2329 			printk(KERN_ERR "EXT4-fs: Journal does not support "
2330 			       "requested data journaling mode\n");
2331 			goto failed_mount4;
2332 		}
2333 	default:
2334 		break;
2335 	}
2336 
2337 	if (test_opt(sb, NOBH)) {
2338 		if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2339 			printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2340 				"its supported only with writeback mode\n");
2341 			clear_opt(sbi->s_mount_opt, NOBH);
2342 		}
2343 	}
2344 	/*
2345 	 * The jbd2_journal_load will have done any necessary log recovery,
2346 	 * so we can safely mount the rest of the filesystem now.
2347 	 */
2348 
2349 	root = ext4_iget(sb, EXT4_ROOT_INO);
2350 	if (IS_ERR(root)) {
2351 		printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2352 		ret = PTR_ERR(root);
2353 		goto failed_mount4;
2354 	}
2355 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2356 		iput(root);
2357 		printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2358 		goto failed_mount4;
2359 	}
2360 	sb->s_root = d_alloc_root(root);
2361 	if (!sb->s_root) {
2362 		printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2363 		iput(root);
2364 		ret = -ENOMEM;
2365 		goto failed_mount4;
2366 	}
2367 
2368 	ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2369 
2370 	/* determine the minimum size of new large inodes, if present */
2371 	if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2372 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2373 						     EXT4_GOOD_OLD_INODE_SIZE;
2374 		if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2375 				       EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2376 			if (sbi->s_want_extra_isize <
2377 			    le16_to_cpu(es->s_want_extra_isize))
2378 				sbi->s_want_extra_isize =
2379 					le16_to_cpu(es->s_want_extra_isize);
2380 			if (sbi->s_want_extra_isize <
2381 			    le16_to_cpu(es->s_min_extra_isize))
2382 				sbi->s_want_extra_isize =
2383 					le16_to_cpu(es->s_min_extra_isize);
2384 		}
2385 	}
2386 	/* Check if enough inode space is available */
2387 	if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2388 							sbi->s_inode_size) {
2389 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2390 						       EXT4_GOOD_OLD_INODE_SIZE;
2391 		printk(KERN_INFO "EXT4-fs: required extra inode space not"
2392 			"available.\n");
2393 	}
2394 
2395 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2396 		printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2397 				"requested data journaling mode\n");
2398 		clear_opt(sbi->s_mount_opt, DELALLOC);
2399 	} else if (test_opt(sb, DELALLOC))
2400 		printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2401 
2402 	ext4_ext_init(sb);
2403 	err = ext4_mb_init(sb, needs_recovery);
2404 	if (err) {
2405 		printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2406 		       err);
2407 		goto failed_mount4;
2408 	}
2409 
2410 	/*
2411 	 * akpm: core read_super() calls in here with the superblock locked.
2412 	 * That deadlocks, because orphan cleanup needs to lock the superblock
2413 	 * in numerous places.  Here we just pop the lock - it's relatively
2414 	 * harmless, because we are now ready to accept write_super() requests,
2415 	 * and aviro says that's the only reason for hanging onto the
2416 	 * superblock lock.
2417 	 */
2418 	EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2419 	ext4_orphan_cleanup(sb, es);
2420 	EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2421 	if (needs_recovery)
2422 		printk(KERN_INFO "EXT4-fs: recovery complete.\n");
2423 	ext4_mark_recovery_complete(sb, es);
2424 	printk(KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2425 	       test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2426 	       test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
2427 	       "writeback");
2428 
2429 	lock_kernel();
2430 	return 0;
2431 
2432 cantfind_ext4:
2433 	if (!silent)
2434 		printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2435 		       sb->s_id);
2436 	goto failed_mount;
2437 
2438 failed_mount4:
2439 	jbd2_journal_destroy(sbi->s_journal);
2440 	sbi->s_journal = NULL;
2441 failed_mount3:
2442 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
2443 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
2444 	percpu_counter_destroy(&sbi->s_dirs_counter);
2445 	percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2446 failed_mount2:
2447 	for (i = 0; i < db_count; i++)
2448 		brelse(sbi->s_group_desc[i]);
2449 	kfree(sbi->s_group_desc);
2450 failed_mount:
2451 	if (sbi->s_proc) {
2452 		remove_proc_entry("inode_readahead_blks", sbi->s_proc);
2453 		remove_proc_entry(sb->s_id, ext4_proc_root);
2454 	}
2455 #ifdef CONFIG_QUOTA
2456 	for (i = 0; i < MAXQUOTAS; i++)
2457 		kfree(sbi->s_qf_names[i]);
2458 #endif
2459 	ext4_blkdev_remove(sbi);
2460 	brelse(bh);
2461 out_fail:
2462 	sb->s_fs_info = NULL;
2463 	kfree(sbi);
2464 	lock_kernel();
2465 	return ret;
2466 }
2467 
2468 /*
2469  * Setup any per-fs journal parameters now.  We'll do this both on
2470  * initial mount, once the journal has been initialised but before we've
2471  * done any recovery; and again on any subsequent remount.
2472  */
2473 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2474 {
2475 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2476 
2477 	if (sbi->s_commit_interval)
2478 		journal->j_commit_interval = sbi->s_commit_interval;
2479 	/* We could also set up an ext4-specific default for the commit
2480 	 * interval here, but for now we'll just fall back to the jbd
2481 	 * default. */
2482 
2483 	spin_lock(&journal->j_state_lock);
2484 	if (test_opt(sb, BARRIER))
2485 		journal->j_flags |= JBD2_BARRIER;
2486 	else
2487 		journal->j_flags &= ~JBD2_BARRIER;
2488 	if (test_opt(sb, DATA_ERR_ABORT))
2489 		journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2490 	else
2491 		journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2492 	spin_unlock(&journal->j_state_lock);
2493 }
2494 
2495 static journal_t *ext4_get_journal(struct super_block *sb,
2496 				   unsigned int journal_inum)
2497 {
2498 	struct inode *journal_inode;
2499 	journal_t *journal;
2500 
2501 	/* First, test for the existence of a valid inode on disk.  Bad
2502 	 * things happen if we iget() an unused inode, as the subsequent
2503 	 * iput() will try to delete it. */
2504 
2505 	journal_inode = ext4_iget(sb, journal_inum);
2506 	if (IS_ERR(journal_inode)) {
2507 		printk(KERN_ERR "EXT4-fs: no journal found.\n");
2508 		return NULL;
2509 	}
2510 	if (!journal_inode->i_nlink) {
2511 		make_bad_inode(journal_inode);
2512 		iput(journal_inode);
2513 		printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2514 		return NULL;
2515 	}
2516 
2517 	jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2518 		  journal_inode, journal_inode->i_size);
2519 	if (!S_ISREG(journal_inode->i_mode)) {
2520 		printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2521 		iput(journal_inode);
2522 		return NULL;
2523 	}
2524 
2525 	journal = jbd2_journal_init_inode(journal_inode);
2526 	if (!journal) {
2527 		printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2528 		iput(journal_inode);
2529 		return NULL;
2530 	}
2531 	journal->j_private = sb;
2532 	ext4_init_journal_params(sb, journal);
2533 	return journal;
2534 }
2535 
2536 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2537 				       dev_t j_dev)
2538 {
2539 	struct buffer_head *bh;
2540 	journal_t *journal;
2541 	ext4_fsblk_t start;
2542 	ext4_fsblk_t len;
2543 	int hblock, blocksize;
2544 	ext4_fsblk_t sb_block;
2545 	unsigned long offset;
2546 	struct ext4_super_block *es;
2547 	struct block_device *bdev;
2548 
2549 	bdev = ext4_blkdev_get(j_dev);
2550 	if (bdev == NULL)
2551 		return NULL;
2552 
2553 	if (bd_claim(bdev, sb)) {
2554 		printk(KERN_ERR
2555 			"EXT4: failed to claim external journal device.\n");
2556 		blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2557 		return NULL;
2558 	}
2559 
2560 	blocksize = sb->s_blocksize;
2561 	hblock = bdev_hardsect_size(bdev);
2562 	if (blocksize < hblock) {
2563 		printk(KERN_ERR
2564 			"EXT4-fs: blocksize too small for journal device.\n");
2565 		goto out_bdev;
2566 	}
2567 
2568 	sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2569 	offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2570 	set_blocksize(bdev, blocksize);
2571 	if (!(bh = __bread(bdev, sb_block, blocksize))) {
2572 		printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2573 		       "external journal\n");
2574 		goto out_bdev;
2575 	}
2576 
2577 	es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2578 	if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2579 	    !(le32_to_cpu(es->s_feature_incompat) &
2580 	      EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2581 		printk(KERN_ERR "EXT4-fs: external journal has "
2582 					"bad superblock\n");
2583 		brelse(bh);
2584 		goto out_bdev;
2585 	}
2586 
2587 	if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2588 		printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2589 		brelse(bh);
2590 		goto out_bdev;
2591 	}
2592 
2593 	len = ext4_blocks_count(es);
2594 	start = sb_block + 1;
2595 	brelse(bh);	/* we're done with the superblock */
2596 
2597 	journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2598 					start, len, blocksize);
2599 	if (!journal) {
2600 		printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2601 		goto out_bdev;
2602 	}
2603 	journal->j_private = sb;
2604 	ll_rw_block(READ, 1, &journal->j_sb_buffer);
2605 	wait_on_buffer(journal->j_sb_buffer);
2606 	if (!buffer_uptodate(journal->j_sb_buffer)) {
2607 		printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2608 		goto out_journal;
2609 	}
2610 	if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2611 		printk(KERN_ERR "EXT4-fs: External journal has more than one "
2612 					"user (unsupported) - %d\n",
2613 			be32_to_cpu(journal->j_superblock->s_nr_users));
2614 		goto out_journal;
2615 	}
2616 	EXT4_SB(sb)->journal_bdev = bdev;
2617 	ext4_init_journal_params(sb, journal);
2618 	return journal;
2619 out_journal:
2620 	jbd2_journal_destroy(journal);
2621 out_bdev:
2622 	ext4_blkdev_put(bdev);
2623 	return NULL;
2624 }
2625 
2626 static int ext4_load_journal(struct super_block *sb,
2627 			     struct ext4_super_block *es,
2628 			     unsigned long journal_devnum)
2629 {
2630 	journal_t *journal;
2631 	unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2632 	dev_t journal_dev;
2633 	int err = 0;
2634 	int really_read_only;
2635 
2636 	if (journal_devnum &&
2637 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2638 		printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2639 			"numbers have changed\n");
2640 		journal_dev = new_decode_dev(journal_devnum);
2641 	} else
2642 		journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2643 
2644 	really_read_only = bdev_read_only(sb->s_bdev);
2645 
2646 	/*
2647 	 * Are we loading a blank journal or performing recovery after a
2648 	 * crash?  For recovery, we need to check in advance whether we
2649 	 * can get read-write access to the device.
2650 	 */
2651 
2652 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2653 		if (sb->s_flags & MS_RDONLY) {
2654 			printk(KERN_INFO "EXT4-fs: INFO: recovery "
2655 					"required on readonly filesystem.\n");
2656 			if (really_read_only) {
2657 				printk(KERN_ERR "EXT4-fs: write access "
2658 					"unavailable, cannot proceed.\n");
2659 				return -EROFS;
2660 			}
2661 			printk(KERN_INFO "EXT4-fs: write access will "
2662 			       "be enabled during recovery.\n");
2663 		}
2664 	}
2665 
2666 	if (journal_inum && journal_dev) {
2667 		printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2668 		       "and inode journals!\n");
2669 		return -EINVAL;
2670 	}
2671 
2672 	if (journal_inum) {
2673 		if (!(journal = ext4_get_journal(sb, journal_inum)))
2674 			return -EINVAL;
2675 	} else {
2676 		if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2677 			return -EINVAL;
2678 	}
2679 
2680 	if (journal->j_flags & JBD2_BARRIER)
2681 		printk(KERN_INFO "EXT4-fs: barriers enabled\n");
2682 	else
2683 		printk(KERN_INFO "EXT4-fs: barriers disabled\n");
2684 
2685 	if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2686 		err = jbd2_journal_update_format(journal);
2687 		if (err)  {
2688 			printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2689 			jbd2_journal_destroy(journal);
2690 			return err;
2691 		}
2692 	}
2693 
2694 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2695 		err = jbd2_journal_wipe(journal, !really_read_only);
2696 	if (!err)
2697 		err = jbd2_journal_load(journal);
2698 
2699 	if (err) {
2700 		printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2701 		jbd2_journal_destroy(journal);
2702 		return err;
2703 	}
2704 
2705 	EXT4_SB(sb)->s_journal = journal;
2706 	ext4_clear_journal_err(sb, es);
2707 
2708 	if (journal_devnum &&
2709 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2710 		es->s_journal_dev = cpu_to_le32(journal_devnum);
2711 		sb->s_dirt = 1;
2712 
2713 		/* Make sure we flush the recovery flag to disk. */
2714 		ext4_commit_super(sb, es, 1);
2715 	}
2716 
2717 	return 0;
2718 }
2719 
2720 static int ext4_create_journal(struct super_block *sb,
2721 			       struct ext4_super_block *es,
2722 			       unsigned int journal_inum)
2723 {
2724 	journal_t *journal;
2725 	int err;
2726 
2727 	if (sb->s_flags & MS_RDONLY) {
2728 		printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
2729 				"create journal.\n");
2730 		return -EROFS;
2731 	}
2732 
2733 	journal = ext4_get_journal(sb, journal_inum);
2734 	if (!journal)
2735 		return -EINVAL;
2736 
2737 	printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
2738 	       journal_inum);
2739 
2740 	err = jbd2_journal_create(journal);
2741 	if (err) {
2742 		printk(KERN_ERR "EXT4-fs: error creating journal.\n");
2743 		jbd2_journal_destroy(journal);
2744 		return -EIO;
2745 	}
2746 
2747 	EXT4_SB(sb)->s_journal = journal;
2748 
2749 	ext4_update_dynamic_rev(sb);
2750 	EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2751 	EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
2752 
2753 	es->s_journal_inum = cpu_to_le32(journal_inum);
2754 	sb->s_dirt = 1;
2755 
2756 	/* Make sure we flush the recovery flag to disk. */
2757 	ext4_commit_super(sb, es, 1);
2758 
2759 	return 0;
2760 }
2761 
2762 static void ext4_commit_super(struct super_block *sb,
2763 			      struct ext4_super_block *es, int sync)
2764 {
2765 	struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2766 
2767 	if (!sbh)
2768 		return;
2769 	if (buffer_write_io_error(sbh)) {
2770 		/*
2771 		 * Oh, dear.  A previous attempt to write the
2772 		 * superblock failed.  This could happen because the
2773 		 * USB device was yanked out.  Or it could happen to
2774 		 * be a transient write error and maybe the block will
2775 		 * be remapped.  Nothing we can do but to retry the
2776 		 * write and hope for the best.
2777 		 */
2778 		printk(KERN_ERR "ext4: previous I/O error to "
2779 		       "superblock detected for %s.\n", sb->s_id);
2780 		clear_buffer_write_io_error(sbh);
2781 		set_buffer_uptodate(sbh);
2782 	}
2783 	es->s_wtime = cpu_to_le32(get_seconds());
2784 	ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
2785 	es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
2786 	BUFFER_TRACE(sbh, "marking dirty");
2787 	mark_buffer_dirty(sbh);
2788 	if (sync) {
2789 		sync_dirty_buffer(sbh);
2790 		if (buffer_write_io_error(sbh)) {
2791 			printk(KERN_ERR "ext4: I/O error while writing "
2792 			       "superblock for %s.\n", sb->s_id);
2793 			clear_buffer_write_io_error(sbh);
2794 			set_buffer_uptodate(sbh);
2795 		}
2796 	}
2797 }
2798 
2799 
2800 /*
2801  * Have we just finished recovery?  If so, and if we are mounting (or
2802  * remounting) the filesystem readonly, then we will end up with a
2803  * consistent fs on disk.  Record that fact.
2804  */
2805 static void ext4_mark_recovery_complete(struct super_block *sb,
2806 					struct ext4_super_block *es)
2807 {
2808 	journal_t *journal = EXT4_SB(sb)->s_journal;
2809 
2810 	jbd2_journal_lock_updates(journal);
2811 	if (jbd2_journal_flush(journal) < 0)
2812 		goto out;
2813 
2814 	lock_super(sb);
2815 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2816 	    sb->s_flags & MS_RDONLY) {
2817 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2818 		sb->s_dirt = 0;
2819 		ext4_commit_super(sb, es, 1);
2820 	}
2821 	unlock_super(sb);
2822 
2823 out:
2824 	jbd2_journal_unlock_updates(journal);
2825 }
2826 
2827 /*
2828  * If we are mounting (or read-write remounting) a filesystem whose journal
2829  * has recorded an error from a previous lifetime, move that error to the
2830  * main filesystem now.
2831  */
2832 static void ext4_clear_journal_err(struct super_block *sb,
2833 				   struct ext4_super_block *es)
2834 {
2835 	journal_t *journal;
2836 	int j_errno;
2837 	const char *errstr;
2838 
2839 	journal = EXT4_SB(sb)->s_journal;
2840 
2841 	/*
2842 	 * Now check for any error status which may have been recorded in the
2843 	 * journal by a prior ext4_error() or ext4_abort()
2844 	 */
2845 
2846 	j_errno = jbd2_journal_errno(journal);
2847 	if (j_errno) {
2848 		char nbuf[16];
2849 
2850 		errstr = ext4_decode_error(sb, j_errno, nbuf);
2851 		ext4_warning(sb, __func__, "Filesystem error recorded "
2852 			     "from previous mount: %s", errstr);
2853 		ext4_warning(sb, __func__, "Marking fs in need of "
2854 			     "filesystem check.");
2855 
2856 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2857 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2858 		ext4_commit_super(sb, es, 1);
2859 
2860 		jbd2_journal_clear_err(journal);
2861 	}
2862 }
2863 
2864 /*
2865  * Force the running and committing transactions to commit,
2866  * and wait on the commit.
2867  */
2868 int ext4_force_commit(struct super_block *sb)
2869 {
2870 	journal_t *journal;
2871 	int ret;
2872 
2873 	if (sb->s_flags & MS_RDONLY)
2874 		return 0;
2875 
2876 	journal = EXT4_SB(sb)->s_journal;
2877 	sb->s_dirt = 0;
2878 	ret = ext4_journal_force_commit(journal);
2879 	return ret;
2880 }
2881 
2882 /*
2883  * Ext4 always journals updates to the superblock itself, so we don't
2884  * have to propagate any other updates to the superblock on disk at this
2885  * point.  Just start an async writeback to get the buffers on their way
2886  * to the disk.
2887  *
2888  * This implicitly triggers the writebehind on sync().
2889  */
2890 
2891 static void ext4_write_super(struct super_block *sb)
2892 {
2893 	if (mutex_trylock(&sb->s_lock) != 0)
2894 		BUG();
2895 	sb->s_dirt = 0;
2896 }
2897 
2898 static int ext4_sync_fs(struct super_block *sb, int wait)
2899 {
2900 	tid_t target;
2901 
2902 	trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
2903 	sb->s_dirt = 0;
2904 	if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
2905 		if (wait)
2906 			jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
2907 	}
2908 	return 0;
2909 }
2910 
2911 /*
2912  * LVM calls this function before a (read-only) snapshot is created.  This
2913  * gives us a chance to flush the journal completely and mark the fs clean.
2914  */
2915 static void ext4_write_super_lockfs(struct super_block *sb)
2916 {
2917 	sb->s_dirt = 0;
2918 
2919 	if (!(sb->s_flags & MS_RDONLY)) {
2920 		journal_t *journal = EXT4_SB(sb)->s_journal;
2921 
2922 		/* Now we set up the journal barrier. */
2923 		jbd2_journal_lock_updates(journal);
2924 
2925 		/*
2926 		 * We don't want to clear needs_recovery flag when we failed
2927 		 * to flush the journal.
2928 		 */
2929 		if (jbd2_journal_flush(journal) < 0)
2930 			return;
2931 
2932 		/* Journal blocked and flushed, clear needs_recovery flag. */
2933 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2934 		ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2935 	}
2936 }
2937 
2938 /*
2939  * Called by LVM after the snapshot is done.  We need to reset the RECOVER
2940  * flag here, even though the filesystem is not technically dirty yet.
2941  */
2942 static void ext4_unlockfs(struct super_block *sb)
2943 {
2944 	if (!(sb->s_flags & MS_RDONLY)) {
2945 		lock_super(sb);
2946 		/* Reser the needs_recovery flag before the fs is unlocked. */
2947 		EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2948 		ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2949 		unlock_super(sb);
2950 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
2951 	}
2952 }
2953 
2954 static int ext4_remount(struct super_block *sb, int *flags, char *data)
2955 {
2956 	struct ext4_super_block *es;
2957 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2958 	ext4_fsblk_t n_blocks_count = 0;
2959 	unsigned long old_sb_flags;
2960 	struct ext4_mount_options old_opts;
2961 	ext4_group_t g;
2962 	int err;
2963 #ifdef CONFIG_QUOTA
2964 	int i;
2965 #endif
2966 
2967 	/* Store the original options */
2968 	old_sb_flags = sb->s_flags;
2969 	old_opts.s_mount_opt = sbi->s_mount_opt;
2970 	old_opts.s_resuid = sbi->s_resuid;
2971 	old_opts.s_resgid = sbi->s_resgid;
2972 	old_opts.s_commit_interval = sbi->s_commit_interval;
2973 #ifdef CONFIG_QUOTA
2974 	old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2975 	for (i = 0; i < MAXQUOTAS; i++)
2976 		old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2977 #endif
2978 
2979 	/*
2980 	 * Allow the "check" option to be passed as a remount option.
2981 	 */
2982 	if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2983 		err = -EINVAL;
2984 		goto restore_opts;
2985 	}
2986 
2987 	if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
2988 		ext4_abort(sb, __func__, "Abort forced by user");
2989 
2990 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2991 		((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2992 
2993 	es = sbi->s_es;
2994 
2995 	ext4_init_journal_params(sb, sbi->s_journal);
2996 
2997 	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2998 		n_blocks_count > ext4_blocks_count(es)) {
2999 		if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
3000 			err = -EROFS;
3001 			goto restore_opts;
3002 		}
3003 
3004 		if (*flags & MS_RDONLY) {
3005 			/*
3006 			 * First of all, the unconditional stuff we have to do
3007 			 * to disable replay of the journal when we next remount
3008 			 */
3009 			sb->s_flags |= MS_RDONLY;
3010 
3011 			/*
3012 			 * OK, test if we are remounting a valid rw partition
3013 			 * readonly, and if so set the rdonly flag and then
3014 			 * mark the partition as valid again.
3015 			 */
3016 			if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3017 			    (sbi->s_mount_state & EXT4_VALID_FS))
3018 				es->s_state = cpu_to_le16(sbi->s_mount_state);
3019 
3020 			/*
3021 			 * We have to unlock super so that we can wait for
3022 			 * transactions.
3023 			 */
3024 			unlock_super(sb);
3025 			ext4_mark_recovery_complete(sb, es);
3026 			lock_super(sb);
3027 		} else {
3028 			__le32 ret;
3029 			if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3030 					~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3031 				printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3032 				       "remount RDWR because of unsupported "
3033 				       "optional features (%x).\n",
3034 				       sb->s_id, le32_to_cpu(ret));
3035 				err = -EROFS;
3036 				goto restore_opts;
3037 			}
3038 
3039 			/*
3040 			 * Make sure the group descriptor checksums
3041 			 * are sane.  If they aren't, refuse to
3042 			 * remount r/w.
3043 			 */
3044 			for (g = 0; g < sbi->s_groups_count; g++) {
3045 				struct ext4_group_desc *gdp =
3046 					ext4_get_group_desc(sb, g, NULL);
3047 
3048 				if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3049 					printk(KERN_ERR
3050 	       "EXT4-fs: ext4_remount: "
3051 		"Checksum for group %lu failed (%u!=%u)\n",
3052 		g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3053 					       le16_to_cpu(gdp->bg_checksum));
3054 					err = -EINVAL;
3055 					goto restore_opts;
3056 				}
3057 			}
3058 
3059 			/*
3060 			 * If we have an unprocessed orphan list hanging
3061 			 * around from a previously readonly bdev mount,
3062 			 * require a full umount/remount for now.
3063 			 */
3064 			if (es->s_last_orphan) {
3065 				printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3066 				       "remount RDWR because of unprocessed "
3067 				       "orphan inode list.  Please "
3068 				       "umount/remount instead.\n",
3069 				       sb->s_id);
3070 				err = -EINVAL;
3071 				goto restore_opts;
3072 			}
3073 
3074 			/*
3075 			 * Mounting a RDONLY partition read-write, so reread
3076 			 * and store the current valid flag.  (It may have
3077 			 * been changed by e2fsck since we originally mounted
3078 			 * the partition.)
3079 			 */
3080 			ext4_clear_journal_err(sb, es);
3081 			sbi->s_mount_state = le16_to_cpu(es->s_state);
3082 			if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3083 				goto restore_opts;
3084 			if (!ext4_setup_super(sb, es, 0))
3085 				sb->s_flags &= ~MS_RDONLY;
3086 		}
3087 	}
3088 #ifdef CONFIG_QUOTA
3089 	/* Release old quota file names */
3090 	for (i = 0; i < MAXQUOTAS; i++)
3091 		if (old_opts.s_qf_names[i] &&
3092 		    old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3093 			kfree(old_opts.s_qf_names[i]);
3094 #endif
3095 	return 0;
3096 restore_opts:
3097 	sb->s_flags = old_sb_flags;
3098 	sbi->s_mount_opt = old_opts.s_mount_opt;
3099 	sbi->s_resuid = old_opts.s_resuid;
3100 	sbi->s_resgid = old_opts.s_resgid;
3101 	sbi->s_commit_interval = old_opts.s_commit_interval;
3102 #ifdef CONFIG_QUOTA
3103 	sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3104 	for (i = 0; i < MAXQUOTAS; i++) {
3105 		if (sbi->s_qf_names[i] &&
3106 		    old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3107 			kfree(sbi->s_qf_names[i]);
3108 		sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3109 	}
3110 #endif
3111 	return err;
3112 }
3113 
3114 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3115 {
3116 	struct super_block *sb = dentry->d_sb;
3117 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3118 	struct ext4_super_block *es = sbi->s_es;
3119 	u64 fsid;
3120 
3121 	if (test_opt(sb, MINIX_DF)) {
3122 		sbi->s_overhead_last = 0;
3123 	} else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3124 		ext4_group_t ngroups = sbi->s_groups_count, i;
3125 		ext4_fsblk_t overhead = 0;
3126 		smp_rmb();
3127 
3128 		/*
3129 		 * Compute the overhead (FS structures).  This is constant
3130 		 * for a given filesystem unless the number of block groups
3131 		 * changes so we cache the previous value until it does.
3132 		 */
3133 
3134 		/*
3135 		 * All of the blocks before first_data_block are
3136 		 * overhead
3137 		 */
3138 		overhead = le32_to_cpu(es->s_first_data_block);
3139 
3140 		/*
3141 		 * Add the overhead attributed to the superblock and
3142 		 * block group descriptors.  If the sparse superblocks
3143 		 * feature is turned on, then not all groups have this.
3144 		 */
3145 		for (i = 0; i < ngroups; i++) {
3146 			overhead += ext4_bg_has_super(sb, i) +
3147 				ext4_bg_num_gdb(sb, i);
3148 			cond_resched();
3149 		}
3150 
3151 		/*
3152 		 * Every block group has an inode bitmap, a block
3153 		 * bitmap, and an inode table.
3154 		 */
3155 		overhead += ngroups * (2 + sbi->s_itb_per_group);
3156 		sbi->s_overhead_last = overhead;
3157 		smp_wmb();
3158 		sbi->s_blocks_last = ext4_blocks_count(es);
3159 	}
3160 
3161 	buf->f_type = EXT4_SUPER_MAGIC;
3162 	buf->f_bsize = sb->s_blocksize;
3163 	buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3164 	buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3165 		       percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3166 	ext4_free_blocks_count_set(es, buf->f_bfree);
3167 	buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3168 	if (buf->f_bfree < ext4_r_blocks_count(es))
3169 		buf->f_bavail = 0;
3170 	buf->f_files = le32_to_cpu(es->s_inodes_count);
3171 	buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3172 	es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3173 	buf->f_namelen = EXT4_NAME_LEN;
3174 	fsid = le64_to_cpup((void *)es->s_uuid) ^
3175 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3176 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3177 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3178 	return 0;
3179 }
3180 
3181 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3182  * is locked for write. Otherwise the are possible deadlocks:
3183  * Process 1                         Process 2
3184  * ext4_create()                     quota_sync()
3185  *   jbd2_journal_start()                   write_dquot()
3186  *   DQUOT_INIT()                        down(dqio_mutex)
3187  *     down(dqio_mutex)                    jbd2_journal_start()
3188  *
3189  */
3190 
3191 #ifdef CONFIG_QUOTA
3192 
3193 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3194 {
3195 	return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3196 }
3197 
3198 static int ext4_dquot_initialize(struct inode *inode, int type)
3199 {
3200 	handle_t *handle;
3201 	int ret, err;
3202 
3203 	/* We may create quota structure so we need to reserve enough blocks */
3204 	handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
3205 	if (IS_ERR(handle))
3206 		return PTR_ERR(handle);
3207 	ret = dquot_initialize(inode, type);
3208 	err = ext4_journal_stop(handle);
3209 	if (!ret)
3210 		ret = err;
3211 	return ret;
3212 }
3213 
3214 static int ext4_dquot_drop(struct inode *inode)
3215 {
3216 	handle_t *handle;
3217 	int ret, err;
3218 
3219 	/* We may delete quota structure so we need to reserve enough blocks */
3220 	handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
3221 	if (IS_ERR(handle)) {
3222 		/*
3223 		 * We call dquot_drop() anyway to at least release references
3224 		 * to quota structures so that umount does not hang.
3225 		 */
3226 		dquot_drop(inode);
3227 		return PTR_ERR(handle);
3228 	}
3229 	ret = dquot_drop(inode);
3230 	err = ext4_journal_stop(handle);
3231 	if (!ret)
3232 		ret = err;
3233 	return ret;
3234 }
3235 
3236 static int ext4_write_dquot(struct dquot *dquot)
3237 {
3238 	int ret, err;
3239 	handle_t *handle;
3240 	struct inode *inode;
3241 
3242 	inode = dquot_to_inode(dquot);
3243 	handle = ext4_journal_start(inode,
3244 					EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3245 	if (IS_ERR(handle))
3246 		return PTR_ERR(handle);
3247 	ret = dquot_commit(dquot);
3248 	err = ext4_journal_stop(handle);
3249 	if (!ret)
3250 		ret = err;
3251 	return ret;
3252 }
3253 
3254 static int ext4_acquire_dquot(struct dquot *dquot)
3255 {
3256 	int ret, err;
3257 	handle_t *handle;
3258 
3259 	handle = ext4_journal_start(dquot_to_inode(dquot),
3260 					EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3261 	if (IS_ERR(handle))
3262 		return PTR_ERR(handle);
3263 	ret = dquot_acquire(dquot);
3264 	err = ext4_journal_stop(handle);
3265 	if (!ret)
3266 		ret = err;
3267 	return ret;
3268 }
3269 
3270 static int ext4_release_dquot(struct dquot *dquot)
3271 {
3272 	int ret, err;
3273 	handle_t *handle;
3274 
3275 	handle = ext4_journal_start(dquot_to_inode(dquot),
3276 					EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3277 	if (IS_ERR(handle)) {
3278 		/* Release dquot anyway to avoid endless cycle in dqput() */
3279 		dquot_release(dquot);
3280 		return PTR_ERR(handle);
3281 	}
3282 	ret = dquot_release(dquot);
3283 	err = ext4_journal_stop(handle);
3284 	if (!ret)
3285 		ret = err;
3286 	return ret;
3287 }
3288 
3289 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3290 {
3291 	/* Are we journaling quotas? */
3292 	if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3293 	    EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3294 		dquot_mark_dquot_dirty(dquot);
3295 		return ext4_write_dquot(dquot);
3296 	} else {
3297 		return dquot_mark_dquot_dirty(dquot);
3298 	}
3299 }
3300 
3301 static int ext4_write_info(struct super_block *sb, int type)
3302 {
3303 	int ret, err;
3304 	handle_t *handle;
3305 
3306 	/* Data block + inode block */
3307 	handle = ext4_journal_start(sb->s_root->d_inode, 2);
3308 	if (IS_ERR(handle))
3309 		return PTR_ERR(handle);
3310 	ret = dquot_commit_info(sb, type);
3311 	err = ext4_journal_stop(handle);
3312 	if (!ret)
3313 		ret = err;
3314 	return ret;
3315 }
3316 
3317 /*
3318  * Turn on quotas during mount time - we need to find
3319  * the quota file and such...
3320  */
3321 static int ext4_quota_on_mount(struct super_block *sb, int type)
3322 {
3323 	return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3324 			EXT4_SB(sb)->s_jquota_fmt, type);
3325 }
3326 
3327 /*
3328  * Standard function to be called on quota_on
3329  */
3330 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3331 			 char *name, int remount)
3332 {
3333 	int err;
3334 	struct path path;
3335 
3336 	if (!test_opt(sb, QUOTA))
3337 		return -EINVAL;
3338 	/* When remounting, no checks are needed and in fact, name is NULL */
3339 	if (remount)
3340 		return vfs_quota_on(sb, type, format_id, name, remount);
3341 
3342 	err = kern_path(name, LOOKUP_FOLLOW, &path);
3343 	if (err)
3344 		return err;
3345 
3346 	/* Quotafile not on the same filesystem? */
3347 	if (path.mnt->mnt_sb != sb) {
3348 		path_put(&path);
3349 		return -EXDEV;
3350 	}
3351 	/* Journaling quota? */
3352 	if (EXT4_SB(sb)->s_qf_names[type]) {
3353 		/* Quotafile not in fs root? */
3354 		if (path.dentry->d_parent != sb->s_root)
3355 			printk(KERN_WARNING
3356 				"EXT4-fs: Quota file not on filesystem root. "
3357 				"Journaled quota will not work.\n");
3358 	}
3359 
3360 	/*
3361 	 * When we journal data on quota file, we have to flush journal to see
3362 	 * all updates to the file when we bypass pagecache...
3363 	 */
3364 	if (ext4_should_journal_data(path.dentry->d_inode)) {
3365 		/*
3366 		 * We don't need to lock updates but journal_flush() could
3367 		 * otherwise be livelocked...
3368 		 */
3369 		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3370 		err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3371 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3372 		if (err) {
3373 			path_put(&path);
3374 			return err;
3375 		}
3376 	}
3377 
3378 	err = vfs_quota_on_path(sb, type, format_id, &path);
3379 	path_put(&path);
3380 	return err;
3381 }
3382 
3383 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3384  * acquiring the locks... As quota files are never truncated and quota code
3385  * itself serializes the operations (and noone else should touch the files)
3386  * we don't have to be afraid of races */
3387 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3388 			       size_t len, loff_t off)
3389 {
3390 	struct inode *inode = sb_dqopt(sb)->files[type];
3391 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3392 	int err = 0;
3393 	int offset = off & (sb->s_blocksize - 1);
3394 	int tocopy;
3395 	size_t toread;
3396 	struct buffer_head *bh;
3397 	loff_t i_size = i_size_read(inode);
3398 
3399 	if (off > i_size)
3400 		return 0;
3401 	if (off+len > i_size)
3402 		len = i_size-off;
3403 	toread = len;
3404 	while (toread > 0) {
3405 		tocopy = sb->s_blocksize - offset < toread ?
3406 				sb->s_blocksize - offset : toread;
3407 		bh = ext4_bread(NULL, inode, blk, 0, &err);
3408 		if (err)
3409 			return err;
3410 		if (!bh)	/* A hole? */
3411 			memset(data, 0, tocopy);
3412 		else
3413 			memcpy(data, bh->b_data+offset, tocopy);
3414 		brelse(bh);
3415 		offset = 0;
3416 		toread -= tocopy;
3417 		data += tocopy;
3418 		blk++;
3419 	}
3420 	return len;
3421 }
3422 
3423 /* Write to quotafile (we know the transaction is already started and has
3424  * enough credits) */
3425 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3426 				const char *data, size_t len, loff_t off)
3427 {
3428 	struct inode *inode = sb_dqopt(sb)->files[type];
3429 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3430 	int err = 0;
3431 	int offset = off & (sb->s_blocksize - 1);
3432 	int tocopy;
3433 	int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3434 	size_t towrite = len;
3435 	struct buffer_head *bh;
3436 	handle_t *handle = journal_current_handle();
3437 
3438 	if (!handle) {
3439 		printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
3440 			" cancelled because transaction is not started.\n",
3441 			(unsigned long long)off, (unsigned long long)len);
3442 		return -EIO;
3443 	}
3444 	mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3445 	while (towrite > 0) {
3446 		tocopy = sb->s_blocksize - offset < towrite ?
3447 				sb->s_blocksize - offset : towrite;
3448 		bh = ext4_bread(handle, inode, blk, 1, &err);
3449 		if (!bh)
3450 			goto out;
3451 		if (journal_quota) {
3452 			err = ext4_journal_get_write_access(handle, bh);
3453 			if (err) {
3454 				brelse(bh);
3455 				goto out;
3456 			}
3457 		}
3458 		lock_buffer(bh);
3459 		memcpy(bh->b_data+offset, data, tocopy);
3460 		flush_dcache_page(bh->b_page);
3461 		unlock_buffer(bh);
3462 		if (journal_quota)
3463 			err = ext4_journal_dirty_metadata(handle, bh);
3464 		else {
3465 			/* Always do at least ordered writes for quotas */
3466 			err = ext4_jbd2_file_inode(handle, inode);
3467 			mark_buffer_dirty(bh);
3468 		}
3469 		brelse(bh);
3470 		if (err)
3471 			goto out;
3472 		offset = 0;
3473 		towrite -= tocopy;
3474 		data += tocopy;
3475 		blk++;
3476 	}
3477 out:
3478 	if (len == towrite) {
3479 		mutex_unlock(&inode->i_mutex);
3480 		return err;
3481 	}
3482 	if (inode->i_size < off+len-towrite) {
3483 		i_size_write(inode, off+len-towrite);
3484 		EXT4_I(inode)->i_disksize = inode->i_size;
3485 	}
3486 	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3487 	ext4_mark_inode_dirty(handle, inode);
3488 	mutex_unlock(&inode->i_mutex);
3489 	return len - towrite;
3490 }
3491 
3492 #endif
3493 
3494 static int ext4_get_sb(struct file_system_type *fs_type,
3495 	int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3496 {
3497 	return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3498 }
3499 
3500 #ifdef CONFIG_PROC_FS
3501 static int ext4_ui_proc_show(struct seq_file *m, void *v)
3502 {
3503 	unsigned int *p = m->private;
3504 
3505 	seq_printf(m, "%u\n", *p);
3506 	return 0;
3507 }
3508 
3509 static int ext4_ui_proc_open(struct inode *inode, struct file *file)
3510 {
3511 	return single_open(file, ext4_ui_proc_show, PDE(inode)->data);
3512 }
3513 
3514 static ssize_t ext4_ui_proc_write(struct file *file, const char __user *buf,
3515 			       size_t cnt, loff_t *ppos)
3516 {
3517 	unsigned int *p = PDE(file->f_path.dentry->d_inode)->data;
3518 	char str[32];
3519 	unsigned long value;
3520 
3521 	if (cnt >= sizeof(str))
3522 		return -EINVAL;
3523 	if (copy_from_user(str, buf, cnt))
3524 		return -EFAULT;
3525 	value = simple_strtol(str, NULL, 0);
3526 	if (value < 0)
3527 		return -ERANGE;
3528 	*p = value;
3529 	return cnt;
3530 }
3531 
3532 const struct file_operations ext4_ui_proc_fops = {
3533 	.owner		= THIS_MODULE,
3534 	.open		= ext4_ui_proc_open,
3535 	.read		= seq_read,
3536 	.llseek		= seq_lseek,
3537 	.release	= single_release,
3538 	.write		= ext4_ui_proc_write,
3539 };
3540 #endif
3541 
3542 static struct file_system_type ext4_fs_type = {
3543 	.owner		= THIS_MODULE,
3544 	.name		= "ext4",
3545 	.get_sb		= ext4_get_sb,
3546 	.kill_sb	= kill_block_super,
3547 	.fs_flags	= FS_REQUIRES_DEV,
3548 };
3549 
3550 #ifdef CONFIG_EXT4DEV_COMPAT
3551 static int ext4dev_get_sb(struct file_system_type *fs_type,
3552 	int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3553 {
3554 	printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3555 	       "to mount using ext4\n");
3556 	printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3557 	       "will go away by 2.6.31\n");
3558 	return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3559 }
3560 
3561 static struct file_system_type ext4dev_fs_type = {
3562 	.owner		= THIS_MODULE,
3563 	.name		= "ext4dev",
3564 	.get_sb		= ext4dev_get_sb,
3565 	.kill_sb	= kill_block_super,
3566 	.fs_flags	= FS_REQUIRES_DEV,
3567 };
3568 MODULE_ALIAS("ext4dev");
3569 #endif
3570 
3571 static int __init init_ext4_fs(void)
3572 {
3573 	int err;
3574 
3575 	ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3576 	err = init_ext4_mballoc();
3577 	if (err)
3578 		return err;
3579 
3580 	err = init_ext4_xattr();
3581 	if (err)
3582 		goto out2;
3583 	err = init_inodecache();
3584 	if (err)
3585 		goto out1;
3586 	err = register_filesystem(&ext4_fs_type);
3587 	if (err)
3588 		goto out;
3589 #ifdef CONFIG_EXT4DEV_COMPAT
3590 	err = register_filesystem(&ext4dev_fs_type);
3591 	if (err) {
3592 		unregister_filesystem(&ext4_fs_type);
3593 		goto out;
3594 	}
3595 #endif
3596 	return 0;
3597 out:
3598 	destroy_inodecache();
3599 out1:
3600 	exit_ext4_xattr();
3601 out2:
3602 	exit_ext4_mballoc();
3603 	return err;
3604 }
3605 
3606 static void __exit exit_ext4_fs(void)
3607 {
3608 	unregister_filesystem(&ext4_fs_type);
3609 #ifdef CONFIG_EXT4DEV_COMPAT
3610 	unregister_filesystem(&ext4dev_fs_type);
3611 #endif
3612 	destroy_inodecache();
3613 	exit_ext4_xattr();
3614 	exit_ext4_mballoc();
3615 	remove_proc_entry("fs/ext4", NULL);
3616 }
3617 
3618 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3619 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
3620 MODULE_LICENSE("GPL");
3621 module_init(init_ext4_fs)
3622 module_exit(exit_ext4_fs)
3623