xref: /openbmc/linux/fs/ext4/super.c (revision 22246614)
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/log2.h>
38 #include <linux/crc16.h>
39 #include <asm/uaccess.h>
40 
41 #include "ext4.h"
42 #include "ext4_jbd2.h"
43 #include "xattr.h"
44 #include "acl.h"
45 #include "namei.h"
46 #include "group.h"
47 
48 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
49 			     unsigned long journal_devnum);
50 static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
51 			       unsigned int);
52 static void ext4_commit_super (struct super_block * sb,
53 			       struct ext4_super_block * es,
54 			       int sync);
55 static void ext4_mark_recovery_complete(struct super_block * sb,
56 					struct ext4_super_block * es);
57 static void ext4_clear_journal_err(struct super_block * sb,
58 				   struct ext4_super_block * es);
59 static int ext4_sync_fs(struct super_block *sb, int wait);
60 static const char *ext4_decode_error(struct super_block * sb, int errno,
61 				     char nbuf[16]);
62 static int ext4_remount (struct super_block * sb, int * flags, char * data);
63 static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf);
64 static void ext4_unlockfs(struct super_block *sb);
65 static void ext4_write_super (struct super_block * sb);
66 static void ext4_write_super_lockfs(struct super_block *sb);
67 
68 
69 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
70 			       struct ext4_group_desc *bg)
71 {
72 	return le32_to_cpu(bg->bg_block_bitmap_lo) |
73 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
74 		(ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
75 }
76 
77 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
78 			       struct ext4_group_desc *bg)
79 {
80 	return le32_to_cpu(bg->bg_inode_bitmap_lo) |
81 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
82 		(ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
83 }
84 
85 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
86 			      struct ext4_group_desc *bg)
87 {
88 	return le32_to_cpu(bg->bg_inode_table_lo) |
89 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
90 		(ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
91 }
92 
93 void ext4_block_bitmap_set(struct super_block *sb,
94 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
95 {
96 	bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
97 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
98 		bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
99 }
100 
101 void ext4_inode_bitmap_set(struct super_block *sb,
102 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
103 {
104 	bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
105 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
106 		bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
107 }
108 
109 void ext4_inode_table_set(struct super_block *sb,
110 			  struct ext4_group_desc *bg, ext4_fsblk_t blk)
111 {
112 	bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
113 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
114 		bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
115 }
116 
117 /*
118  * Wrappers for jbd2_journal_start/end.
119  *
120  * The only special thing we need to do here is to make sure that all
121  * journal_end calls result in the superblock being marked dirty, so
122  * that sync() will call the filesystem's write_super callback if
123  * appropriate.
124  */
125 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
126 {
127 	journal_t *journal;
128 
129 	if (sb->s_flags & MS_RDONLY)
130 		return ERR_PTR(-EROFS);
131 
132 	/* Special case here: if the journal has aborted behind our
133 	 * backs (eg. EIO in the commit thread), then we still need to
134 	 * take the FS itself readonly cleanly. */
135 	journal = EXT4_SB(sb)->s_journal;
136 	if (is_journal_aborted(journal)) {
137 		ext4_abort(sb, __func__,
138 			   "Detected aborted journal");
139 		return ERR_PTR(-EROFS);
140 	}
141 
142 	return jbd2_journal_start(journal, nblocks);
143 }
144 
145 /*
146  * The only special thing we need to do here is to make sure that all
147  * jbd2_journal_stop calls result in the superblock being marked dirty, so
148  * that sync() will call the filesystem's write_super callback if
149  * appropriate.
150  */
151 int __ext4_journal_stop(const char *where, handle_t *handle)
152 {
153 	struct super_block *sb;
154 	int err;
155 	int rc;
156 
157 	sb = handle->h_transaction->t_journal->j_private;
158 	err = handle->h_err;
159 	rc = jbd2_journal_stop(handle);
160 
161 	if (!err)
162 		err = rc;
163 	if (err)
164 		__ext4_std_error(sb, where, err);
165 	return err;
166 }
167 
168 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
169 		struct buffer_head *bh, handle_t *handle, int err)
170 {
171 	char nbuf[16];
172 	const char *errstr = ext4_decode_error(NULL, err, nbuf);
173 
174 	if (bh)
175 		BUFFER_TRACE(bh, "abort");
176 
177 	if (!handle->h_err)
178 		handle->h_err = err;
179 
180 	if (is_handle_aborted(handle))
181 		return;
182 
183 	printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
184 	       caller, errstr, err_fn);
185 
186 	jbd2_journal_abort_handle(handle);
187 }
188 
189 /* Deal with the reporting of failure conditions on a filesystem such as
190  * inconsistencies detected or read IO failures.
191  *
192  * On ext2, we can store the error state of the filesystem in the
193  * superblock.  That is not possible on ext4, because we may have other
194  * write ordering constraints on the superblock which prevent us from
195  * writing it out straight away; and given that the journal is about to
196  * be aborted, we can't rely on the current, or future, transactions to
197  * write out the superblock safely.
198  *
199  * We'll just use the jbd2_journal_abort() error code to record an error in
200  * the journal instead.  On recovery, the journal will compain about
201  * that error until we've noted it down and cleared it.
202  */
203 
204 static void ext4_handle_error(struct super_block *sb)
205 {
206 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
207 
208 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
209 	es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
210 
211 	if (sb->s_flags & MS_RDONLY)
212 		return;
213 
214 	if (!test_opt (sb, ERRORS_CONT)) {
215 		journal_t *journal = EXT4_SB(sb)->s_journal;
216 
217 		EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
218 		if (journal)
219 			jbd2_journal_abort(journal, -EIO);
220 	}
221 	if (test_opt (sb, ERRORS_RO)) {
222 		printk (KERN_CRIT "Remounting filesystem read-only\n");
223 		sb->s_flags |= MS_RDONLY;
224 	}
225 	ext4_commit_super(sb, es, 1);
226 	if (test_opt(sb, ERRORS_PANIC))
227 		panic("EXT4-fs (device %s): panic forced after error\n",
228 			sb->s_id);
229 }
230 
231 void ext4_error (struct super_block * sb, const char * function,
232 		 const char * fmt, ...)
233 {
234 	va_list args;
235 
236 	va_start(args, fmt);
237 	printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
238 	vprintk(fmt, args);
239 	printk("\n");
240 	va_end(args);
241 
242 	ext4_handle_error(sb);
243 }
244 
245 static const char *ext4_decode_error(struct super_block * sb, int errno,
246 				     char nbuf[16])
247 {
248 	char *errstr = NULL;
249 
250 	switch (errno) {
251 	case -EIO:
252 		errstr = "IO failure";
253 		break;
254 	case -ENOMEM:
255 		errstr = "Out of memory";
256 		break;
257 	case -EROFS:
258 		if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
259 			errstr = "Journal has aborted";
260 		else
261 			errstr = "Readonly filesystem";
262 		break;
263 	default:
264 		/* If the caller passed in an extra buffer for unknown
265 		 * errors, textualise them now.  Else we just return
266 		 * NULL. */
267 		if (nbuf) {
268 			/* Check for truncated error codes... */
269 			if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
270 				errstr = nbuf;
271 		}
272 		break;
273 	}
274 
275 	return errstr;
276 }
277 
278 /* __ext4_std_error decodes expected errors from journaling functions
279  * automatically and invokes the appropriate error response.  */
280 
281 void __ext4_std_error (struct super_block * sb, const char * function,
282 		       int errno)
283 {
284 	char nbuf[16];
285 	const char *errstr;
286 
287 	/* Special case: if the error is EROFS, and we're not already
288 	 * inside a transaction, then there's really no point in logging
289 	 * an error. */
290 	if (errno == -EROFS && journal_current_handle() == NULL &&
291 	    (sb->s_flags & MS_RDONLY))
292 		return;
293 
294 	errstr = ext4_decode_error(sb, errno, nbuf);
295 	printk (KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
296 		sb->s_id, function, errstr);
297 
298 	ext4_handle_error(sb);
299 }
300 
301 /*
302  * ext4_abort is a much stronger failure handler than ext4_error.  The
303  * abort function may be used to deal with unrecoverable failures such
304  * as journal IO errors or ENOMEM at a critical moment in log management.
305  *
306  * We unconditionally force the filesystem into an ABORT|READONLY state,
307  * unless the error response on the fs has been set to panic in which
308  * case we take the easy way out and panic immediately.
309  */
310 
311 void ext4_abort (struct super_block * sb, const char * function,
312 		 const char * fmt, ...)
313 {
314 	va_list args;
315 
316 	printk (KERN_CRIT "ext4_abort called.\n");
317 
318 	va_start(args, fmt);
319 	printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
320 	vprintk(fmt, args);
321 	printk("\n");
322 	va_end(args);
323 
324 	if (test_opt(sb, ERRORS_PANIC))
325 		panic("EXT4-fs panic from previous error\n");
326 
327 	if (sb->s_flags & MS_RDONLY)
328 		return;
329 
330 	printk(KERN_CRIT "Remounting filesystem read-only\n");
331 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
332 	sb->s_flags |= MS_RDONLY;
333 	EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
334 	jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
335 }
336 
337 void ext4_warning (struct super_block * sb, const char * function,
338 		   const char * fmt, ...)
339 {
340 	va_list args;
341 
342 	va_start(args, fmt);
343 	printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
344 	       sb->s_id, function);
345 	vprintk(fmt, args);
346 	printk("\n");
347 	va_end(args);
348 }
349 
350 void ext4_update_dynamic_rev(struct super_block *sb)
351 {
352 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
353 
354 	if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
355 		return;
356 
357 	ext4_warning(sb, __func__,
358 		     "updating to rev %d because of new feature flag, "
359 		     "running e2fsck is recommended",
360 		     EXT4_DYNAMIC_REV);
361 
362 	es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
363 	es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
364 	es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
365 	/* leave es->s_feature_*compat flags alone */
366 	/* es->s_uuid will be set by e2fsck if empty */
367 
368 	/*
369 	 * The rest of the superblock fields should be zero, and if not it
370 	 * means they are likely already in use, so leave them alone.  We
371 	 * can leave it up to e2fsck to clean up any inconsistencies there.
372 	 */
373 }
374 
375 int ext4_update_compat_feature(handle_t *handle,
376 					struct super_block *sb, __u32 compat)
377 {
378 	int err = 0;
379 	if (!EXT4_HAS_COMPAT_FEATURE(sb, compat)) {
380 		err = ext4_journal_get_write_access(handle,
381 				EXT4_SB(sb)->s_sbh);
382 		if (err)
383 			return err;
384 		EXT4_SET_COMPAT_FEATURE(sb, compat);
385 		sb->s_dirt = 1;
386 		handle->h_sync = 1;
387 		BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
388 					"call ext4_journal_dirty_met adata");
389 		err = ext4_journal_dirty_metadata(handle,
390 				EXT4_SB(sb)->s_sbh);
391 	}
392 	return err;
393 }
394 
395 int ext4_update_rocompat_feature(handle_t *handle,
396 					struct super_block *sb, __u32 rocompat)
397 {
398 	int err = 0;
399 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, rocompat)) {
400 		err = ext4_journal_get_write_access(handle,
401 				EXT4_SB(sb)->s_sbh);
402 		if (err)
403 			return err;
404 		EXT4_SET_RO_COMPAT_FEATURE(sb, rocompat);
405 		sb->s_dirt = 1;
406 		handle->h_sync = 1;
407 		BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
408 					"call ext4_journal_dirty_met adata");
409 		err = ext4_journal_dirty_metadata(handle,
410 				EXT4_SB(sb)->s_sbh);
411 	}
412 	return err;
413 }
414 
415 int ext4_update_incompat_feature(handle_t *handle,
416 					struct super_block *sb, __u32 incompat)
417 {
418 	int err = 0;
419 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, incompat)) {
420 		err = ext4_journal_get_write_access(handle,
421 				EXT4_SB(sb)->s_sbh);
422 		if (err)
423 			return err;
424 		EXT4_SET_INCOMPAT_FEATURE(sb, incompat);
425 		sb->s_dirt = 1;
426 		handle->h_sync = 1;
427 		BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
428 					"call ext4_journal_dirty_met adata");
429 		err = ext4_journal_dirty_metadata(handle,
430 				EXT4_SB(sb)->s_sbh);
431 	}
432 	return err;
433 }
434 
435 /*
436  * Open the external journal device
437  */
438 static struct block_device *ext4_blkdev_get(dev_t dev)
439 {
440 	struct block_device *bdev;
441 	char b[BDEVNAME_SIZE];
442 
443 	bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
444 	if (IS_ERR(bdev))
445 		goto fail;
446 	return bdev;
447 
448 fail:
449 	printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
450 			__bdevname(dev, b), PTR_ERR(bdev));
451 	return NULL;
452 }
453 
454 /*
455  * Release the journal device
456  */
457 static int ext4_blkdev_put(struct block_device *bdev)
458 {
459 	bd_release(bdev);
460 	return blkdev_put(bdev);
461 }
462 
463 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
464 {
465 	struct block_device *bdev;
466 	int ret = -ENODEV;
467 
468 	bdev = sbi->journal_bdev;
469 	if (bdev) {
470 		ret = ext4_blkdev_put(bdev);
471 		sbi->journal_bdev = NULL;
472 	}
473 	return ret;
474 }
475 
476 static inline struct inode *orphan_list_entry(struct list_head *l)
477 {
478 	return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
479 }
480 
481 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
482 {
483 	struct list_head *l;
484 
485 	printk(KERN_ERR "sb orphan head is %d\n",
486 	       le32_to_cpu(sbi->s_es->s_last_orphan));
487 
488 	printk(KERN_ERR "sb_info orphan list:\n");
489 	list_for_each(l, &sbi->s_orphan) {
490 		struct inode *inode = orphan_list_entry(l);
491 		printk(KERN_ERR "  "
492 		       "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
493 		       inode->i_sb->s_id, inode->i_ino, inode,
494 		       inode->i_mode, inode->i_nlink,
495 		       NEXT_ORPHAN(inode));
496 	}
497 }
498 
499 static void ext4_put_super (struct super_block * sb)
500 {
501 	struct ext4_sb_info *sbi = EXT4_SB(sb);
502 	struct ext4_super_block *es = sbi->s_es;
503 	int i;
504 
505 	ext4_mb_release(sb);
506 	ext4_ext_release(sb);
507 	ext4_xattr_put_super(sb);
508 	jbd2_journal_destroy(sbi->s_journal);
509 	if (!(sb->s_flags & MS_RDONLY)) {
510 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
511 		es->s_state = cpu_to_le16(sbi->s_mount_state);
512 		BUFFER_TRACE(sbi->s_sbh, "marking dirty");
513 		mark_buffer_dirty(sbi->s_sbh);
514 		ext4_commit_super(sb, es, 1);
515 	}
516 
517 	for (i = 0; i < sbi->s_gdb_count; i++)
518 		brelse(sbi->s_group_desc[i]);
519 	kfree(sbi->s_group_desc);
520 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
521 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
522 	percpu_counter_destroy(&sbi->s_dirs_counter);
523 	brelse(sbi->s_sbh);
524 #ifdef CONFIG_QUOTA
525 	for (i = 0; i < MAXQUOTAS; i++)
526 		kfree(sbi->s_qf_names[i]);
527 #endif
528 
529 	/* Debugging code just in case the in-memory inode orphan list
530 	 * isn't empty.  The on-disk one can be non-empty if we've
531 	 * detected an error and taken the fs readonly, but the
532 	 * in-memory list had better be clean by this point. */
533 	if (!list_empty(&sbi->s_orphan))
534 		dump_orphan_list(sb, sbi);
535 	J_ASSERT(list_empty(&sbi->s_orphan));
536 
537 	invalidate_bdev(sb->s_bdev);
538 	if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
539 		/*
540 		 * Invalidate the journal device's buffers.  We don't want them
541 		 * floating about in memory - the physical journal device may
542 		 * hotswapped, and it breaks the `ro-after' testing code.
543 		 */
544 		sync_blockdev(sbi->journal_bdev);
545 		invalidate_bdev(sbi->journal_bdev);
546 		ext4_blkdev_remove(sbi);
547 	}
548 	sb->s_fs_info = NULL;
549 	kfree(sbi);
550 	return;
551 }
552 
553 static struct kmem_cache *ext4_inode_cachep;
554 
555 /*
556  * Called inside transaction, so use GFP_NOFS
557  */
558 static struct inode *ext4_alloc_inode(struct super_block *sb)
559 {
560 	struct ext4_inode_info *ei;
561 
562 	ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
563 	if (!ei)
564 		return NULL;
565 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
566 	ei->i_acl = EXT4_ACL_NOT_CACHED;
567 	ei->i_default_acl = EXT4_ACL_NOT_CACHED;
568 #endif
569 	ei->i_block_alloc_info = NULL;
570 	ei->vfs_inode.i_version = 1;
571 	memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
572 	INIT_LIST_HEAD(&ei->i_prealloc_list);
573 	spin_lock_init(&ei->i_prealloc_lock);
574 	return &ei->vfs_inode;
575 }
576 
577 static void ext4_destroy_inode(struct inode *inode)
578 {
579 	if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
580 		printk("EXT4 Inode %p: orphan list check failed!\n",
581 			EXT4_I(inode));
582 		print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
583 				EXT4_I(inode), sizeof(struct ext4_inode_info),
584 				true);
585 		dump_stack();
586 	}
587 	kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
588 }
589 
590 static void init_once(struct kmem_cache *cachep, void *foo)
591 {
592 	struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
593 
594 	INIT_LIST_HEAD(&ei->i_orphan);
595 #ifdef CONFIG_EXT4DEV_FS_XATTR
596 	init_rwsem(&ei->xattr_sem);
597 #endif
598 	init_rwsem(&ei->i_data_sem);
599 	inode_init_once(&ei->vfs_inode);
600 }
601 
602 static int init_inodecache(void)
603 {
604 	ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
605 					     sizeof(struct ext4_inode_info),
606 					     0, (SLAB_RECLAIM_ACCOUNT|
607 						SLAB_MEM_SPREAD),
608 					     init_once);
609 	if (ext4_inode_cachep == NULL)
610 		return -ENOMEM;
611 	return 0;
612 }
613 
614 static void destroy_inodecache(void)
615 {
616 	kmem_cache_destroy(ext4_inode_cachep);
617 }
618 
619 static void ext4_clear_inode(struct inode *inode)
620 {
621 	struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info;
622 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
623 	if (EXT4_I(inode)->i_acl &&
624 			EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
625 		posix_acl_release(EXT4_I(inode)->i_acl);
626 		EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
627 	}
628 	if (EXT4_I(inode)->i_default_acl &&
629 			EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
630 		posix_acl_release(EXT4_I(inode)->i_default_acl);
631 		EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
632 	}
633 #endif
634 	ext4_discard_reservation(inode);
635 	EXT4_I(inode)->i_block_alloc_info = NULL;
636 	if (unlikely(rsv))
637 		kfree(rsv);
638 }
639 
640 static inline void ext4_show_quota_options(struct seq_file *seq, struct super_block *sb)
641 {
642 #if defined(CONFIG_QUOTA)
643 	struct ext4_sb_info *sbi = EXT4_SB(sb);
644 
645 	if (sbi->s_jquota_fmt)
646 		seq_printf(seq, ",jqfmt=%s",
647 		(sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
648 
649 	if (sbi->s_qf_names[USRQUOTA])
650 		seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
651 
652 	if (sbi->s_qf_names[GRPQUOTA])
653 		seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
654 
655 	if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
656 		seq_puts(seq, ",usrquota");
657 
658 	if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
659 		seq_puts(seq, ",grpquota");
660 #endif
661 }
662 
663 /*
664  * Show an option if
665  *  - it's set to a non-default value OR
666  *  - if the per-sb default is different from the global default
667  */
668 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
669 {
670 	int def_errors;
671 	unsigned long def_mount_opts;
672 	struct super_block *sb = vfs->mnt_sb;
673 	struct ext4_sb_info *sbi = EXT4_SB(sb);
674 	struct ext4_super_block *es = sbi->s_es;
675 
676 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
677 	def_errors     = le16_to_cpu(es->s_errors);
678 
679 	if (sbi->s_sb_block != 1)
680 		seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
681 	if (test_opt(sb, MINIX_DF))
682 		seq_puts(seq, ",minixdf");
683 	if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
684 		seq_puts(seq, ",grpid");
685 	if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
686 		seq_puts(seq, ",nogrpid");
687 	if (sbi->s_resuid != EXT4_DEF_RESUID ||
688 	    le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
689 		seq_printf(seq, ",resuid=%u", sbi->s_resuid);
690 	}
691 	if (sbi->s_resgid != EXT4_DEF_RESGID ||
692 	    le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
693 		seq_printf(seq, ",resgid=%u", sbi->s_resgid);
694 	}
695 	if (test_opt(sb, ERRORS_RO)) {
696 		if (def_errors == EXT4_ERRORS_PANIC ||
697 		    def_errors == EXT4_ERRORS_CONTINUE) {
698 			seq_puts(seq, ",errors=remount-ro");
699 		}
700 	}
701 	if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
702 		seq_puts(seq, ",errors=continue");
703 	if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
704 		seq_puts(seq, ",errors=panic");
705 	if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
706 		seq_puts(seq, ",nouid32");
707 	if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
708 		seq_puts(seq, ",debug");
709 	if (test_opt(sb, OLDALLOC))
710 		seq_puts(seq, ",oldalloc");
711 #ifdef CONFIG_EXT4DEV_FS_XATTR
712 	if (test_opt(sb, XATTR_USER) &&
713 		!(def_mount_opts & EXT4_DEFM_XATTR_USER))
714 		seq_puts(seq, ",user_xattr");
715 	if (!test_opt(sb, XATTR_USER) &&
716 	    (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
717 		seq_puts(seq, ",nouser_xattr");
718 	}
719 #endif
720 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
721 	if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
722 		seq_puts(seq, ",acl");
723 	if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
724 		seq_puts(seq, ",noacl");
725 #endif
726 	if (!test_opt(sb, RESERVATION))
727 		seq_puts(seq, ",noreservation");
728 	if (sbi->s_commit_interval) {
729 		seq_printf(seq, ",commit=%u",
730 			   (unsigned) (sbi->s_commit_interval / HZ));
731 	}
732 	if (test_opt(sb, BARRIER))
733 		seq_puts(seq, ",barrier=1");
734 	if (test_opt(sb, NOBH))
735 		seq_puts(seq, ",nobh");
736 	if (!test_opt(sb, EXTENTS))
737 		seq_puts(seq, ",noextents");
738 	if (!test_opt(sb, MBALLOC))
739 		seq_puts(seq, ",nomballoc");
740 	if (test_opt(sb, I_VERSION))
741 		seq_puts(seq, ",i_version");
742 
743 	if (sbi->s_stripe)
744 		seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
745 	/*
746 	 * journal mode get enabled in different ways
747 	 * So just print the value even if we didn't specify it
748 	 */
749 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
750 		seq_puts(seq, ",data=journal");
751 	else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
752 		seq_puts(seq, ",data=ordered");
753 	else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
754 		seq_puts(seq, ",data=writeback");
755 
756 	ext4_show_quota_options(seq, sb);
757 	return 0;
758 }
759 
760 
761 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
762 		u64 ino, u32 generation)
763 {
764 	struct inode *inode;
765 
766 	if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
767 		return ERR_PTR(-ESTALE);
768 	if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
769 		return ERR_PTR(-ESTALE);
770 
771 	/* iget isn't really right if the inode is currently unallocated!!
772 	 *
773 	 * ext4_read_inode will return a bad_inode if the inode had been
774 	 * deleted, so we should be safe.
775 	 *
776 	 * Currently we don't know the generation for parent directory, so
777 	 * a generation of 0 means "accept any"
778 	 */
779 	inode = ext4_iget(sb, ino);
780 	if (IS_ERR(inode))
781 		return ERR_CAST(inode);
782 	if (generation && inode->i_generation != generation) {
783 		iput(inode);
784 		return ERR_PTR(-ESTALE);
785 	}
786 
787 	return inode;
788 }
789 
790 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
791 		int fh_len, int fh_type)
792 {
793 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
794 				    ext4_nfs_get_inode);
795 }
796 
797 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
798 		int fh_len, int fh_type)
799 {
800 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
801 				    ext4_nfs_get_inode);
802 }
803 
804 #ifdef CONFIG_QUOTA
805 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
806 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
807 
808 static int ext4_dquot_initialize(struct inode *inode, int type);
809 static int ext4_dquot_drop(struct inode *inode);
810 static int ext4_write_dquot(struct dquot *dquot);
811 static int ext4_acquire_dquot(struct dquot *dquot);
812 static int ext4_release_dquot(struct dquot *dquot);
813 static int ext4_mark_dquot_dirty(struct dquot *dquot);
814 static int ext4_write_info(struct super_block *sb, int type);
815 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
816 				char *path, int remount);
817 static int ext4_quota_on_mount(struct super_block *sb, int type);
818 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
819 			       size_t len, loff_t off);
820 static ssize_t ext4_quota_write(struct super_block *sb, int type,
821 				const char *data, size_t len, loff_t off);
822 
823 static struct dquot_operations ext4_quota_operations = {
824 	.initialize	= ext4_dquot_initialize,
825 	.drop		= ext4_dquot_drop,
826 	.alloc_space	= dquot_alloc_space,
827 	.alloc_inode	= dquot_alloc_inode,
828 	.free_space	= dquot_free_space,
829 	.free_inode	= dquot_free_inode,
830 	.transfer	= dquot_transfer,
831 	.write_dquot	= ext4_write_dquot,
832 	.acquire_dquot	= ext4_acquire_dquot,
833 	.release_dquot	= ext4_release_dquot,
834 	.mark_dirty	= ext4_mark_dquot_dirty,
835 	.write_info	= ext4_write_info
836 };
837 
838 static struct quotactl_ops ext4_qctl_operations = {
839 	.quota_on	= ext4_quota_on,
840 	.quota_off	= vfs_quota_off,
841 	.quota_sync	= vfs_quota_sync,
842 	.get_info	= vfs_get_dqinfo,
843 	.set_info	= vfs_set_dqinfo,
844 	.get_dqblk	= vfs_get_dqblk,
845 	.set_dqblk	= vfs_set_dqblk
846 };
847 #endif
848 
849 static const struct super_operations ext4_sops = {
850 	.alloc_inode	= ext4_alloc_inode,
851 	.destroy_inode	= ext4_destroy_inode,
852 	.write_inode	= ext4_write_inode,
853 	.dirty_inode	= ext4_dirty_inode,
854 	.delete_inode	= ext4_delete_inode,
855 	.put_super	= ext4_put_super,
856 	.write_super	= ext4_write_super,
857 	.sync_fs	= ext4_sync_fs,
858 	.write_super_lockfs = ext4_write_super_lockfs,
859 	.unlockfs	= ext4_unlockfs,
860 	.statfs		= ext4_statfs,
861 	.remount_fs	= ext4_remount,
862 	.clear_inode	= ext4_clear_inode,
863 	.show_options	= ext4_show_options,
864 #ifdef CONFIG_QUOTA
865 	.quota_read	= ext4_quota_read,
866 	.quota_write	= ext4_quota_write,
867 #endif
868 };
869 
870 static const struct export_operations ext4_export_ops = {
871 	.fh_to_dentry = ext4_fh_to_dentry,
872 	.fh_to_parent = ext4_fh_to_parent,
873 	.get_parent = ext4_get_parent,
874 };
875 
876 enum {
877 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
878 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
879 	Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
880 	Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
881 	Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
882 	Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
883 	Opt_journal_checksum, Opt_journal_async_commit,
884 	Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
885 	Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
886 	Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
887 	Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
888 	Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
889 	Opt_mballoc, Opt_nomballoc, Opt_stripe,
890 };
891 
892 static match_table_t tokens = {
893 	{Opt_bsd_df, "bsddf"},
894 	{Opt_minix_df, "minixdf"},
895 	{Opt_grpid, "grpid"},
896 	{Opt_grpid, "bsdgroups"},
897 	{Opt_nogrpid, "nogrpid"},
898 	{Opt_nogrpid, "sysvgroups"},
899 	{Opt_resgid, "resgid=%u"},
900 	{Opt_resuid, "resuid=%u"},
901 	{Opt_sb, "sb=%u"},
902 	{Opt_err_cont, "errors=continue"},
903 	{Opt_err_panic, "errors=panic"},
904 	{Opt_err_ro, "errors=remount-ro"},
905 	{Opt_nouid32, "nouid32"},
906 	{Opt_nocheck, "nocheck"},
907 	{Opt_nocheck, "check=none"},
908 	{Opt_debug, "debug"},
909 	{Opt_oldalloc, "oldalloc"},
910 	{Opt_orlov, "orlov"},
911 	{Opt_user_xattr, "user_xattr"},
912 	{Opt_nouser_xattr, "nouser_xattr"},
913 	{Opt_acl, "acl"},
914 	{Opt_noacl, "noacl"},
915 	{Opt_reservation, "reservation"},
916 	{Opt_noreservation, "noreservation"},
917 	{Opt_noload, "noload"},
918 	{Opt_nobh, "nobh"},
919 	{Opt_bh, "bh"},
920 	{Opt_commit, "commit=%u"},
921 	{Opt_journal_update, "journal=update"},
922 	{Opt_journal_inum, "journal=%u"},
923 	{Opt_journal_dev, "journal_dev=%u"},
924 	{Opt_journal_checksum, "journal_checksum"},
925 	{Opt_journal_async_commit, "journal_async_commit"},
926 	{Opt_abort, "abort"},
927 	{Opt_data_journal, "data=journal"},
928 	{Opt_data_ordered, "data=ordered"},
929 	{Opt_data_writeback, "data=writeback"},
930 	{Opt_offusrjquota, "usrjquota="},
931 	{Opt_usrjquota, "usrjquota=%s"},
932 	{Opt_offgrpjquota, "grpjquota="},
933 	{Opt_grpjquota, "grpjquota=%s"},
934 	{Opt_jqfmt_vfsold, "jqfmt=vfsold"},
935 	{Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
936 	{Opt_grpquota, "grpquota"},
937 	{Opt_noquota, "noquota"},
938 	{Opt_quota, "quota"},
939 	{Opt_usrquota, "usrquota"},
940 	{Opt_barrier, "barrier=%u"},
941 	{Opt_extents, "extents"},
942 	{Opt_noextents, "noextents"},
943 	{Opt_i_version, "i_version"},
944 	{Opt_mballoc, "mballoc"},
945 	{Opt_nomballoc, "nomballoc"},
946 	{Opt_stripe, "stripe=%u"},
947 	{Opt_resize, "resize"},
948 	{Opt_err, NULL},
949 };
950 
951 static ext4_fsblk_t get_sb_block(void **data)
952 {
953 	ext4_fsblk_t	sb_block;
954 	char		*options = (char *) *data;
955 
956 	if (!options || strncmp(options, "sb=", 3) != 0)
957 		return 1;	/* Default location */
958 	options += 3;
959 	/*todo: use simple_strtoll with >32bit ext4 */
960 	sb_block = simple_strtoul(options, &options, 0);
961 	if (*options && *options != ',') {
962 		printk("EXT4-fs: Invalid sb specification: %s\n",
963 		       (char *) *data);
964 		return 1;
965 	}
966 	if (*options == ',')
967 		options++;
968 	*data = (void *) options;
969 	return sb_block;
970 }
971 
972 static int parse_options (char *options, struct super_block *sb,
973 			  unsigned int *inum, unsigned long *journal_devnum,
974 			  ext4_fsblk_t *n_blocks_count, int is_remount)
975 {
976 	struct ext4_sb_info *sbi = EXT4_SB(sb);
977 	char * p;
978 	substring_t args[MAX_OPT_ARGS];
979 	int data_opt = 0;
980 	int option;
981 #ifdef CONFIG_QUOTA
982 	int qtype;
983 	char *qname;
984 #endif
985 
986 	if (!options)
987 		return 1;
988 
989 	while ((p = strsep (&options, ",")) != NULL) {
990 		int token;
991 		if (!*p)
992 			continue;
993 
994 		token = match_token(p, tokens, args);
995 		switch (token) {
996 		case Opt_bsd_df:
997 			clear_opt (sbi->s_mount_opt, MINIX_DF);
998 			break;
999 		case Opt_minix_df:
1000 			set_opt (sbi->s_mount_opt, MINIX_DF);
1001 			break;
1002 		case Opt_grpid:
1003 			set_opt (sbi->s_mount_opt, GRPID);
1004 			break;
1005 		case Opt_nogrpid:
1006 			clear_opt (sbi->s_mount_opt, GRPID);
1007 			break;
1008 		case Opt_resuid:
1009 			if (match_int(&args[0], &option))
1010 				return 0;
1011 			sbi->s_resuid = option;
1012 			break;
1013 		case Opt_resgid:
1014 			if (match_int(&args[0], &option))
1015 				return 0;
1016 			sbi->s_resgid = option;
1017 			break;
1018 		case Opt_sb:
1019 			/* handled by get_sb_block() instead of here */
1020 			/* *sb_block = match_int(&args[0]); */
1021 			break;
1022 		case Opt_err_panic:
1023 			clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1024 			clear_opt (sbi->s_mount_opt, ERRORS_RO);
1025 			set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1026 			break;
1027 		case Opt_err_ro:
1028 			clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1029 			clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1030 			set_opt (sbi->s_mount_opt, ERRORS_RO);
1031 			break;
1032 		case Opt_err_cont:
1033 			clear_opt (sbi->s_mount_opt, ERRORS_RO);
1034 			clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1035 			set_opt (sbi->s_mount_opt, ERRORS_CONT);
1036 			break;
1037 		case Opt_nouid32:
1038 			set_opt (sbi->s_mount_opt, NO_UID32);
1039 			break;
1040 		case Opt_nocheck:
1041 			clear_opt (sbi->s_mount_opt, CHECK);
1042 			break;
1043 		case Opt_debug:
1044 			set_opt (sbi->s_mount_opt, DEBUG);
1045 			break;
1046 		case Opt_oldalloc:
1047 			set_opt (sbi->s_mount_opt, OLDALLOC);
1048 			break;
1049 		case Opt_orlov:
1050 			clear_opt (sbi->s_mount_opt, OLDALLOC);
1051 			break;
1052 #ifdef CONFIG_EXT4DEV_FS_XATTR
1053 		case Opt_user_xattr:
1054 			set_opt (sbi->s_mount_opt, XATTR_USER);
1055 			break;
1056 		case Opt_nouser_xattr:
1057 			clear_opt (sbi->s_mount_opt, XATTR_USER);
1058 			break;
1059 #else
1060 		case Opt_user_xattr:
1061 		case Opt_nouser_xattr:
1062 			printk("EXT4 (no)user_xattr options not supported\n");
1063 			break;
1064 #endif
1065 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1066 		case Opt_acl:
1067 			set_opt(sbi->s_mount_opt, POSIX_ACL);
1068 			break;
1069 		case Opt_noacl:
1070 			clear_opt(sbi->s_mount_opt, POSIX_ACL);
1071 			break;
1072 #else
1073 		case Opt_acl:
1074 		case Opt_noacl:
1075 			printk("EXT4 (no)acl options not supported\n");
1076 			break;
1077 #endif
1078 		case Opt_reservation:
1079 			set_opt(sbi->s_mount_opt, RESERVATION);
1080 			break;
1081 		case Opt_noreservation:
1082 			clear_opt(sbi->s_mount_opt, RESERVATION);
1083 			break;
1084 		case Opt_journal_update:
1085 			/* @@@ FIXME */
1086 			/* Eventually we will want to be able to create
1087 			   a journal file here.  For now, only allow the
1088 			   user to specify an existing inode to be the
1089 			   journal file. */
1090 			if (is_remount) {
1091 				printk(KERN_ERR "EXT4-fs: cannot specify "
1092 				       "journal on remount\n");
1093 				return 0;
1094 			}
1095 			set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1096 			break;
1097 		case Opt_journal_inum:
1098 			if (is_remount) {
1099 				printk(KERN_ERR "EXT4-fs: cannot specify "
1100 				       "journal on remount\n");
1101 				return 0;
1102 			}
1103 			if (match_int(&args[0], &option))
1104 				return 0;
1105 			*inum = option;
1106 			break;
1107 		case Opt_journal_dev:
1108 			if (is_remount) {
1109 				printk(KERN_ERR "EXT4-fs: cannot specify "
1110 				       "journal on remount\n");
1111 				return 0;
1112 			}
1113 			if (match_int(&args[0], &option))
1114 				return 0;
1115 			*journal_devnum = option;
1116 			break;
1117 		case Opt_journal_checksum:
1118 			set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1119 			break;
1120 		case Opt_journal_async_commit:
1121 			set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1122 			set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1123 			break;
1124 		case Opt_noload:
1125 			set_opt (sbi->s_mount_opt, NOLOAD);
1126 			break;
1127 		case Opt_commit:
1128 			if (match_int(&args[0], &option))
1129 				return 0;
1130 			if (option < 0)
1131 				return 0;
1132 			if (option == 0)
1133 				option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1134 			sbi->s_commit_interval = HZ * option;
1135 			break;
1136 		case Opt_data_journal:
1137 			data_opt = EXT4_MOUNT_JOURNAL_DATA;
1138 			goto datacheck;
1139 		case Opt_data_ordered:
1140 			data_opt = EXT4_MOUNT_ORDERED_DATA;
1141 			goto datacheck;
1142 		case Opt_data_writeback:
1143 			data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1144 		datacheck:
1145 			if (is_remount) {
1146 				if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1147 						!= data_opt) {
1148 					printk(KERN_ERR
1149 						"EXT4-fs: cannot change data "
1150 						"mode on remount\n");
1151 					return 0;
1152 				}
1153 			} else {
1154 				sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1155 				sbi->s_mount_opt |= data_opt;
1156 			}
1157 			break;
1158 #ifdef CONFIG_QUOTA
1159 		case Opt_usrjquota:
1160 			qtype = USRQUOTA;
1161 			goto set_qf_name;
1162 		case Opt_grpjquota:
1163 			qtype = GRPQUOTA;
1164 set_qf_name:
1165 			if (sb_any_quota_enabled(sb)) {
1166 				printk(KERN_ERR
1167 					"EXT4-fs: Cannot change journalled "
1168 					"quota options when quota turned on.\n");
1169 				return 0;
1170 			}
1171 			qname = match_strdup(&args[0]);
1172 			if (!qname) {
1173 				printk(KERN_ERR
1174 					"EXT4-fs: not enough memory for "
1175 					"storing quotafile name.\n");
1176 				return 0;
1177 			}
1178 			if (sbi->s_qf_names[qtype] &&
1179 			    strcmp(sbi->s_qf_names[qtype], qname)) {
1180 				printk(KERN_ERR
1181 					"EXT4-fs: %s quota file already "
1182 					"specified.\n", QTYPE2NAME(qtype));
1183 				kfree(qname);
1184 				return 0;
1185 			}
1186 			sbi->s_qf_names[qtype] = qname;
1187 			if (strchr(sbi->s_qf_names[qtype], '/')) {
1188 				printk(KERN_ERR
1189 					"EXT4-fs: quotafile must be on "
1190 					"filesystem root.\n");
1191 				kfree(sbi->s_qf_names[qtype]);
1192 				sbi->s_qf_names[qtype] = NULL;
1193 				return 0;
1194 			}
1195 			set_opt(sbi->s_mount_opt, QUOTA);
1196 			break;
1197 		case Opt_offusrjquota:
1198 			qtype = USRQUOTA;
1199 			goto clear_qf_name;
1200 		case Opt_offgrpjquota:
1201 			qtype = GRPQUOTA;
1202 clear_qf_name:
1203 			if (sb_any_quota_enabled(sb)) {
1204 				printk(KERN_ERR "EXT4-fs: Cannot change "
1205 					"journalled quota options when "
1206 					"quota turned on.\n");
1207 				return 0;
1208 			}
1209 			/*
1210 			 * The space will be released later when all options
1211 			 * are confirmed to be correct
1212 			 */
1213 			sbi->s_qf_names[qtype] = NULL;
1214 			break;
1215 		case Opt_jqfmt_vfsold:
1216 			sbi->s_jquota_fmt = QFMT_VFS_OLD;
1217 			break;
1218 		case Opt_jqfmt_vfsv0:
1219 			sbi->s_jquota_fmt = QFMT_VFS_V0;
1220 			break;
1221 		case Opt_quota:
1222 		case Opt_usrquota:
1223 			set_opt(sbi->s_mount_opt, QUOTA);
1224 			set_opt(sbi->s_mount_opt, USRQUOTA);
1225 			break;
1226 		case Opt_grpquota:
1227 			set_opt(sbi->s_mount_opt, QUOTA);
1228 			set_opt(sbi->s_mount_opt, GRPQUOTA);
1229 			break;
1230 		case Opt_noquota:
1231 			if (sb_any_quota_enabled(sb)) {
1232 				printk(KERN_ERR "EXT4-fs: Cannot change quota "
1233 					"options when quota turned on.\n");
1234 				return 0;
1235 			}
1236 			clear_opt(sbi->s_mount_opt, QUOTA);
1237 			clear_opt(sbi->s_mount_opt, USRQUOTA);
1238 			clear_opt(sbi->s_mount_opt, GRPQUOTA);
1239 			break;
1240 #else
1241 		case Opt_quota:
1242 		case Opt_usrquota:
1243 		case Opt_grpquota:
1244 		case Opt_usrjquota:
1245 		case Opt_grpjquota:
1246 		case Opt_offusrjquota:
1247 		case Opt_offgrpjquota:
1248 		case Opt_jqfmt_vfsold:
1249 		case Opt_jqfmt_vfsv0:
1250 			printk(KERN_ERR
1251 				"EXT4-fs: journalled quota options not "
1252 				"supported.\n");
1253 			break;
1254 		case Opt_noquota:
1255 			break;
1256 #endif
1257 		case Opt_abort:
1258 			set_opt(sbi->s_mount_opt, ABORT);
1259 			break;
1260 		case Opt_barrier:
1261 			if (match_int(&args[0], &option))
1262 				return 0;
1263 			if (option)
1264 				set_opt(sbi->s_mount_opt, BARRIER);
1265 			else
1266 				clear_opt(sbi->s_mount_opt, BARRIER);
1267 			break;
1268 		case Opt_ignore:
1269 			break;
1270 		case Opt_resize:
1271 			if (!is_remount) {
1272 				printk("EXT4-fs: resize option only available "
1273 					"for remount\n");
1274 				return 0;
1275 			}
1276 			if (match_int(&args[0], &option) != 0)
1277 				return 0;
1278 			*n_blocks_count = option;
1279 			break;
1280 		case Opt_nobh:
1281 			set_opt(sbi->s_mount_opt, NOBH);
1282 			break;
1283 		case Opt_bh:
1284 			clear_opt(sbi->s_mount_opt, NOBH);
1285 			break;
1286 		case Opt_extents:
1287 			set_opt (sbi->s_mount_opt, EXTENTS);
1288 			break;
1289 		case Opt_noextents:
1290 			clear_opt (sbi->s_mount_opt, EXTENTS);
1291 			break;
1292 		case Opt_i_version:
1293 			set_opt(sbi->s_mount_opt, I_VERSION);
1294 			sb->s_flags |= MS_I_VERSION;
1295 			break;
1296 		case Opt_mballoc:
1297 			set_opt(sbi->s_mount_opt, MBALLOC);
1298 			break;
1299 		case Opt_nomballoc:
1300 			clear_opt(sbi->s_mount_opt, MBALLOC);
1301 			break;
1302 		case Opt_stripe:
1303 			if (match_int(&args[0], &option))
1304 				return 0;
1305 			if (option < 0)
1306 				return 0;
1307 			sbi->s_stripe = option;
1308 			break;
1309 		default:
1310 			printk (KERN_ERR
1311 				"EXT4-fs: Unrecognized mount option \"%s\" "
1312 				"or missing value\n", p);
1313 			return 0;
1314 		}
1315 	}
1316 #ifdef CONFIG_QUOTA
1317 	if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1318 		if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1319 		     sbi->s_qf_names[USRQUOTA])
1320 			clear_opt(sbi->s_mount_opt, USRQUOTA);
1321 
1322 		if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1323 		     sbi->s_qf_names[GRPQUOTA])
1324 			clear_opt(sbi->s_mount_opt, GRPQUOTA);
1325 
1326 		if ((sbi->s_qf_names[USRQUOTA] &&
1327 				(sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1328 		    (sbi->s_qf_names[GRPQUOTA] &&
1329 				(sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1330 			printk(KERN_ERR "EXT4-fs: old and new quota "
1331 					"format mixing.\n");
1332 			return 0;
1333 		}
1334 
1335 		if (!sbi->s_jquota_fmt) {
1336 			printk(KERN_ERR "EXT4-fs: journalled quota format "
1337 					"not specified.\n");
1338 			return 0;
1339 		}
1340 	} else {
1341 		if (sbi->s_jquota_fmt) {
1342 			printk(KERN_ERR "EXT4-fs: journalled quota format "
1343 					"specified with no journalling "
1344 					"enabled.\n");
1345 			return 0;
1346 		}
1347 	}
1348 #endif
1349 	return 1;
1350 }
1351 
1352 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1353 			    int read_only)
1354 {
1355 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1356 	int res = 0;
1357 
1358 	if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1359 		printk (KERN_ERR "EXT4-fs warning: revision level too high, "
1360 			"forcing read-only mode\n");
1361 		res = MS_RDONLY;
1362 	}
1363 	if (read_only)
1364 		return res;
1365 	if (!(sbi->s_mount_state & EXT4_VALID_FS))
1366 		printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1367 			"running e2fsck is recommended\n");
1368 	else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1369 		printk (KERN_WARNING
1370 			"EXT4-fs warning: mounting fs with errors, "
1371 			"running e2fsck is recommended\n");
1372 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1373 		 le16_to_cpu(es->s_mnt_count) >=
1374 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1375 		printk (KERN_WARNING
1376 			"EXT4-fs warning: maximal mount count reached, "
1377 			"running e2fsck is recommended\n");
1378 	else if (le32_to_cpu(es->s_checkinterval) &&
1379 		(le32_to_cpu(es->s_lastcheck) +
1380 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1381 		printk (KERN_WARNING
1382 			"EXT4-fs warning: checktime reached, "
1383 			"running e2fsck is recommended\n");
1384 #if 0
1385 		/* @@@ We _will_ want to clear the valid bit if we find
1386 		 * inconsistencies, to force a fsck at reboot.  But for
1387 		 * a plain journaled filesystem we can keep it set as
1388 		 * valid forever! :)
1389 		 */
1390 	es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1391 #endif
1392 	if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1393 		es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1394 	le16_add_cpu(&es->s_mnt_count, 1);
1395 	es->s_mtime = cpu_to_le32(get_seconds());
1396 	ext4_update_dynamic_rev(sb);
1397 	EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1398 
1399 	ext4_commit_super(sb, es, 1);
1400 	if (test_opt(sb, DEBUG))
1401 		printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
1402 				"bpg=%lu, ipg=%lu, mo=%04lx]\n",
1403 			sb->s_blocksize,
1404 			sbi->s_groups_count,
1405 			EXT4_BLOCKS_PER_GROUP(sb),
1406 			EXT4_INODES_PER_GROUP(sb),
1407 			sbi->s_mount_opt);
1408 
1409 	printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1410 	if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
1411 		char b[BDEVNAME_SIZE];
1412 
1413 		printk("external journal on %s\n",
1414 			bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
1415 	} else {
1416 		printk("internal journal\n");
1417 	}
1418 	return res;
1419 }
1420 
1421 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1422 			    struct ext4_group_desc *gdp)
1423 {
1424 	__u16 crc = 0;
1425 
1426 	if (sbi->s_es->s_feature_ro_compat &
1427 	    cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1428 		int offset = offsetof(struct ext4_group_desc, bg_checksum);
1429 		__le32 le_group = cpu_to_le32(block_group);
1430 
1431 		crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1432 		crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1433 		crc = crc16(crc, (__u8 *)gdp, offset);
1434 		offset += sizeof(gdp->bg_checksum); /* skip checksum */
1435 		/* for checksum of struct ext4_group_desc do the rest...*/
1436 		if ((sbi->s_es->s_feature_incompat &
1437 		     cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1438 		    offset < le16_to_cpu(sbi->s_es->s_desc_size))
1439 			crc = crc16(crc, (__u8 *)gdp + offset,
1440 				    le16_to_cpu(sbi->s_es->s_desc_size) -
1441 					offset);
1442 	}
1443 
1444 	return cpu_to_le16(crc);
1445 }
1446 
1447 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1448 				struct ext4_group_desc *gdp)
1449 {
1450 	if ((sbi->s_es->s_feature_ro_compat &
1451 	     cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1452 	    (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1453 		return 0;
1454 
1455 	return 1;
1456 }
1457 
1458 /* Called at mount-time, super-block is locked */
1459 static int ext4_check_descriptors(struct super_block *sb)
1460 {
1461 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1462 	ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1463 	ext4_fsblk_t last_block;
1464 	ext4_fsblk_t block_bitmap;
1465 	ext4_fsblk_t inode_bitmap;
1466 	ext4_fsblk_t inode_table;
1467 	int flexbg_flag = 0;
1468 	ext4_group_t i;
1469 
1470 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1471 		flexbg_flag = 1;
1472 
1473 	ext4_debug ("Checking group descriptors");
1474 
1475 	for (i = 0; i < sbi->s_groups_count; i++) {
1476 		struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1477 
1478 		if (i == sbi->s_groups_count - 1 || flexbg_flag)
1479 			last_block = ext4_blocks_count(sbi->s_es) - 1;
1480 		else
1481 			last_block = first_block +
1482 				(EXT4_BLOCKS_PER_GROUP(sb) - 1);
1483 
1484 		block_bitmap = ext4_block_bitmap(sb, gdp);
1485 		if (block_bitmap < first_block || block_bitmap > last_block)
1486 		{
1487 			printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1488 			       "Block bitmap for group %lu not in group "
1489 			       "(block %llu)!", i, block_bitmap);
1490 			return 0;
1491 		}
1492 		inode_bitmap = ext4_inode_bitmap(sb, gdp);
1493 		if (inode_bitmap < first_block || inode_bitmap > last_block)
1494 		{
1495 			printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1496 			       "Inode bitmap for group %lu not in group "
1497 			       "(block %llu)!", i, inode_bitmap);
1498 			return 0;
1499 		}
1500 		inode_table = ext4_inode_table(sb, gdp);
1501 		if (inode_table < first_block ||
1502 		    inode_table + sbi->s_itb_per_group - 1 > last_block)
1503 		{
1504 			printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1505 			       "Inode table for group %lu not in group "
1506 			       "(block %llu)!", i, inode_table);
1507 			return 0;
1508 		}
1509 		if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1510 			printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1511 			       "Checksum for group %lu failed (%u!=%u)\n",
1512 			       i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1513 			       gdp)), le16_to_cpu(gdp->bg_checksum));
1514 			return 0;
1515 		}
1516 		if (!flexbg_flag)
1517 			first_block += EXT4_BLOCKS_PER_GROUP(sb);
1518 	}
1519 
1520 	ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1521 	sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb));
1522 	return 1;
1523 }
1524 
1525 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1526  * the superblock) which were deleted from all directories, but held open by
1527  * a process at the time of a crash.  We walk the list and try to delete these
1528  * inodes at recovery time (only with a read-write filesystem).
1529  *
1530  * In order to keep the orphan inode chain consistent during traversal (in
1531  * case of crash during recovery), we link each inode into the superblock
1532  * orphan list_head and handle it the same way as an inode deletion during
1533  * normal operation (which journals the operations for us).
1534  *
1535  * We only do an iget() and an iput() on each inode, which is very safe if we
1536  * accidentally point at an in-use or already deleted inode.  The worst that
1537  * can happen in this case is that we get a "bit already cleared" message from
1538  * ext4_free_inode().  The only reason we would point at a wrong inode is if
1539  * e2fsck was run on this filesystem, and it must have already done the orphan
1540  * inode cleanup for us, so we can safely abort without any further action.
1541  */
1542 static void ext4_orphan_cleanup (struct super_block * sb,
1543 				 struct ext4_super_block * es)
1544 {
1545 	unsigned int s_flags = sb->s_flags;
1546 	int nr_orphans = 0, nr_truncates = 0;
1547 #ifdef CONFIG_QUOTA
1548 	int i;
1549 #endif
1550 	if (!es->s_last_orphan) {
1551 		jbd_debug(4, "no orphan inodes to clean up\n");
1552 		return;
1553 	}
1554 
1555 	if (bdev_read_only(sb->s_bdev)) {
1556 		printk(KERN_ERR "EXT4-fs: write access "
1557 			"unavailable, skipping orphan cleanup.\n");
1558 		return;
1559 	}
1560 
1561 	if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1562 		if (es->s_last_orphan)
1563 			jbd_debug(1, "Errors on filesystem, "
1564 				  "clearing orphan list.\n");
1565 		es->s_last_orphan = 0;
1566 		jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1567 		return;
1568 	}
1569 
1570 	if (s_flags & MS_RDONLY) {
1571 		printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1572 		       sb->s_id);
1573 		sb->s_flags &= ~MS_RDONLY;
1574 	}
1575 #ifdef CONFIG_QUOTA
1576 	/* Needed for iput() to work correctly and not trash data */
1577 	sb->s_flags |= MS_ACTIVE;
1578 	/* Turn on quotas so that they are updated correctly */
1579 	for (i = 0; i < MAXQUOTAS; i++) {
1580 		if (EXT4_SB(sb)->s_qf_names[i]) {
1581 			int ret = ext4_quota_on_mount(sb, i);
1582 			if (ret < 0)
1583 				printk(KERN_ERR
1584 					"EXT4-fs: Cannot turn on journalled "
1585 					"quota: error %d\n", ret);
1586 		}
1587 	}
1588 #endif
1589 
1590 	while (es->s_last_orphan) {
1591 		struct inode *inode;
1592 
1593 		inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1594 		if (IS_ERR(inode)) {
1595 			es->s_last_orphan = 0;
1596 			break;
1597 		}
1598 
1599 		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1600 		DQUOT_INIT(inode);
1601 		if (inode->i_nlink) {
1602 			printk(KERN_DEBUG
1603 				"%s: truncating inode %lu to %Ld bytes\n",
1604 				__func__, inode->i_ino, inode->i_size);
1605 			jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1606 				  inode->i_ino, inode->i_size);
1607 			ext4_truncate(inode);
1608 			nr_truncates++;
1609 		} else {
1610 			printk(KERN_DEBUG
1611 				"%s: deleting unreferenced inode %lu\n",
1612 				__func__, inode->i_ino);
1613 			jbd_debug(2, "deleting unreferenced inode %lu\n",
1614 				  inode->i_ino);
1615 			nr_orphans++;
1616 		}
1617 		iput(inode);  /* The delete magic happens here! */
1618 	}
1619 
1620 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1621 
1622 	if (nr_orphans)
1623 		printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1624 		       sb->s_id, PLURAL(nr_orphans));
1625 	if (nr_truncates)
1626 		printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1627 		       sb->s_id, PLURAL(nr_truncates));
1628 #ifdef CONFIG_QUOTA
1629 	/* Turn quotas off */
1630 	for (i = 0; i < MAXQUOTAS; i++) {
1631 		if (sb_dqopt(sb)->files[i])
1632 			vfs_quota_off(sb, i, 0);
1633 	}
1634 #endif
1635 	sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1636 }
1637 /*
1638  * Maximal extent format file size.
1639  * Resulting logical blkno at s_maxbytes must fit in our on-disk
1640  * extent format containers, within a sector_t, and within i_blocks
1641  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
1642  * so that won't be a limiting factor.
1643  *
1644  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1645  */
1646 static loff_t ext4_max_size(int blkbits)
1647 {
1648 	loff_t res;
1649 	loff_t upper_limit = MAX_LFS_FILESIZE;
1650 
1651 	/* small i_blocks in vfs inode? */
1652 	if (sizeof(blkcnt_t) < sizeof(u64)) {
1653 		/*
1654 		 * CONFIG_LSF is not enabled implies the inode
1655 		 * i_block represent total blocks in 512 bytes
1656 		 * 32 == size of vfs inode i_blocks * 8
1657 		 */
1658 		upper_limit = (1LL << 32) - 1;
1659 
1660 		/* total blocks in file system block size */
1661 		upper_limit >>= (blkbits - 9);
1662 		upper_limit <<= blkbits;
1663 	}
1664 
1665 	/* 32-bit extent-start container, ee_block */
1666 	res = 1LL << 32;
1667 	res <<= blkbits;
1668 	res -= 1;
1669 
1670 	/* Sanity check against vm- & vfs- imposed limits */
1671 	if (res > upper_limit)
1672 		res = upper_limit;
1673 
1674 	return res;
1675 }
1676 
1677 /*
1678  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
1679  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1680  * We need to be 1 filesystem block less than the 2^48 sector limit.
1681  */
1682 static loff_t ext4_max_bitmap_size(int bits)
1683 {
1684 	loff_t res = EXT4_NDIR_BLOCKS;
1685 	int meta_blocks;
1686 	loff_t upper_limit;
1687 	/* This is calculated to be the largest file size for a
1688 	 * dense, bitmapped file such that the total number of
1689 	 * sectors in the file, including data and all indirect blocks,
1690 	 * does not exceed 2^48 -1
1691 	 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1692 	 * total number of  512 bytes blocks of the file
1693 	 */
1694 
1695 	if (sizeof(blkcnt_t) < sizeof(u64)) {
1696 		/*
1697 		 * CONFIG_LSF is not enabled implies the inode
1698 		 * i_block represent total blocks in 512 bytes
1699 		 * 32 == size of vfs inode i_blocks * 8
1700 		 */
1701 		upper_limit = (1LL << 32) - 1;
1702 
1703 		/* total blocks in file system block size */
1704 		upper_limit >>= (bits - 9);
1705 
1706 	} else {
1707 		/*
1708 		 * We use 48 bit ext4_inode i_blocks
1709 		 * With EXT4_HUGE_FILE_FL set the i_blocks
1710 		 * represent total number of blocks in
1711 		 * file system block size
1712 		 */
1713 		upper_limit = (1LL << 48) - 1;
1714 
1715 	}
1716 
1717 	/* indirect blocks */
1718 	meta_blocks = 1;
1719 	/* double indirect blocks */
1720 	meta_blocks += 1 + (1LL << (bits-2));
1721 	/* tripple indirect blocks */
1722 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1723 
1724 	upper_limit -= meta_blocks;
1725 	upper_limit <<= bits;
1726 
1727 	res += 1LL << (bits-2);
1728 	res += 1LL << (2*(bits-2));
1729 	res += 1LL << (3*(bits-2));
1730 	res <<= bits;
1731 	if (res > upper_limit)
1732 		res = upper_limit;
1733 
1734 	if (res > MAX_LFS_FILESIZE)
1735 		res = MAX_LFS_FILESIZE;
1736 
1737 	return res;
1738 }
1739 
1740 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1741 				ext4_fsblk_t logical_sb_block, int nr)
1742 {
1743 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1744 	ext4_group_t bg, first_meta_bg;
1745 	int has_super = 0;
1746 
1747 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1748 
1749 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1750 	    nr < first_meta_bg)
1751 		return logical_sb_block + nr + 1;
1752 	bg = sbi->s_desc_per_block * nr;
1753 	if (ext4_bg_has_super(sb, bg))
1754 		has_super = 1;
1755 	return (has_super + ext4_group_first_block_no(sb, bg));
1756 }
1757 
1758 /**
1759  * ext4_get_stripe_size: Get the stripe size.
1760  * @sbi: In memory super block info
1761  *
1762  * If we have specified it via mount option, then
1763  * use the mount option value. If the value specified at mount time is
1764  * greater than the blocks per group use the super block value.
1765  * If the super block value is greater than blocks per group return 0.
1766  * Allocator needs it be less than blocks per group.
1767  *
1768  */
1769 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1770 {
1771 	unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1772 	unsigned long stripe_width =
1773 			le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1774 
1775 	if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1776 		return sbi->s_stripe;
1777 
1778 	if (stripe_width <= sbi->s_blocks_per_group)
1779 		return stripe_width;
1780 
1781 	if (stride <= sbi->s_blocks_per_group)
1782 		return stride;
1783 
1784 	return 0;
1785 }
1786 
1787 static int ext4_fill_super (struct super_block *sb, void *data, int silent)
1788 				__releases(kernel_sem)
1789 				__acquires(kernel_sem)
1790 
1791 {
1792 	struct buffer_head * bh;
1793 	struct ext4_super_block *es = NULL;
1794 	struct ext4_sb_info *sbi;
1795 	ext4_fsblk_t block;
1796 	ext4_fsblk_t sb_block = get_sb_block(&data);
1797 	ext4_fsblk_t logical_sb_block;
1798 	unsigned long offset = 0;
1799 	unsigned int journal_inum = 0;
1800 	unsigned long journal_devnum = 0;
1801 	unsigned long def_mount_opts;
1802 	struct inode *root;
1803 	int ret = -EINVAL;
1804 	int blocksize;
1805 	int db_count;
1806 	int i;
1807 	int needs_recovery;
1808 	__le32 features;
1809 	__u64 blocks_count;
1810 	int err;
1811 
1812 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1813 	if (!sbi)
1814 		return -ENOMEM;
1815 	sb->s_fs_info = sbi;
1816 	sbi->s_mount_opt = 0;
1817 	sbi->s_resuid = EXT4_DEF_RESUID;
1818 	sbi->s_resgid = EXT4_DEF_RESGID;
1819 	sbi->s_sb_block = sb_block;
1820 
1821 	unlock_kernel();
1822 
1823 	blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
1824 	if (!blocksize) {
1825 		printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
1826 		goto out_fail;
1827 	}
1828 
1829 	if (!sb_set_blocksize(sb, blocksize)) {
1830 		printk(KERN_ERR "EXT4-fs: bad blocksize %d.\n", blocksize);
1831 		goto out_fail;
1832 	}
1833 
1834 	/*
1835 	 * The ext4 superblock will not be buffer aligned for other than 1kB
1836 	 * block sizes.  We need to calculate the offset from buffer start.
1837 	 */
1838 	if (blocksize != EXT4_MIN_BLOCK_SIZE) {
1839 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1840 		offset = do_div(logical_sb_block, blocksize);
1841 	} else {
1842 		logical_sb_block = sb_block;
1843 	}
1844 
1845 	if (!(bh = sb_bread(sb, logical_sb_block))) {
1846 		printk (KERN_ERR "EXT4-fs: unable to read superblock\n");
1847 		goto out_fail;
1848 	}
1849 	/*
1850 	 * Note: s_es must be initialized as soon as possible because
1851 	 *       some ext4 macro-instructions depend on its value
1852 	 */
1853 	es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
1854 	sbi->s_es = es;
1855 	sb->s_magic = le16_to_cpu(es->s_magic);
1856 	if (sb->s_magic != EXT4_SUPER_MAGIC)
1857 		goto cantfind_ext4;
1858 
1859 	/* Set defaults before we parse the mount options */
1860 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1861 	if (def_mount_opts & EXT4_DEFM_DEBUG)
1862 		set_opt(sbi->s_mount_opt, DEBUG);
1863 	if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
1864 		set_opt(sbi->s_mount_opt, GRPID);
1865 	if (def_mount_opts & EXT4_DEFM_UID16)
1866 		set_opt(sbi->s_mount_opt, NO_UID32);
1867 #ifdef CONFIG_EXT4DEV_FS_XATTR
1868 	if (def_mount_opts & EXT4_DEFM_XATTR_USER)
1869 		set_opt(sbi->s_mount_opt, XATTR_USER);
1870 #endif
1871 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1872 	if (def_mount_opts & EXT4_DEFM_ACL)
1873 		set_opt(sbi->s_mount_opt, POSIX_ACL);
1874 #endif
1875 	if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1876 		sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1877 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1878 		sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1879 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1880 		sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1881 
1882 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
1883 		set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1884 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
1885 		set_opt(sbi->s_mount_opt, ERRORS_CONT);
1886 	else
1887 		set_opt(sbi->s_mount_opt, ERRORS_RO);
1888 
1889 	sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1890 	sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1891 
1892 	set_opt(sbi->s_mount_opt, RESERVATION);
1893 
1894 	/*
1895 	 * turn on extents feature by default in ext4 filesystem
1896 	 * User -o noextents to turn it off
1897 	 */
1898 	set_opt(sbi->s_mount_opt, EXTENTS);
1899 	/*
1900 	 * turn on mballoc feature by default in ext4 filesystem
1901 	 * User -o nomballoc to turn it off
1902 	 */
1903 	set_opt(sbi->s_mount_opt, MBALLOC);
1904 
1905 	if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1906 			    NULL, 0))
1907 		goto failed_mount;
1908 
1909 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1910 		((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1911 
1912 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1913 	    (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1914 	     EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1915 	     EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1916 		printk(KERN_WARNING
1917 		       "EXT4-fs warning: feature flags set on rev 0 fs, "
1918 		       "running e2fsck is recommended\n");
1919 
1920 	/*
1921 	 * Since ext4 is still considered development code, we require
1922 	 * that the TEST_FILESYS flag in s->flags be set.
1923 	 */
1924 	if (!(le32_to_cpu(es->s_flags) & EXT2_FLAGS_TEST_FILESYS)) {
1925 		printk(KERN_WARNING "EXT4-fs: %s: not marked "
1926 		       "OK to use with test code.\n", sb->s_id);
1927 		goto failed_mount;
1928 	}
1929 
1930 	/*
1931 	 * Check feature flags regardless of the revision level, since we
1932 	 * previously didn't change the revision level when setting the flags,
1933 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
1934 	 */
1935 	features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
1936 	if (features) {
1937 		printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
1938 		       "unsupported optional features (%x).\n",
1939 		       sb->s_id, le32_to_cpu(features));
1940 		goto failed_mount;
1941 	}
1942 	features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
1943 	if (!(sb->s_flags & MS_RDONLY) && features) {
1944 		printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
1945 		       "unsupported optional features (%x).\n",
1946 		       sb->s_id, le32_to_cpu(features));
1947 		goto failed_mount;
1948 	}
1949 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
1950 		/*
1951 		 * Large file size enabled file system can only be
1952 		 * mount if kernel is build with CONFIG_LSF
1953 		 */
1954 		if (sizeof(root->i_blocks) < sizeof(u64) &&
1955 				!(sb->s_flags & MS_RDONLY)) {
1956 			printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
1957 					"files cannot be mounted read-write "
1958 					"without CONFIG_LSF.\n", sb->s_id);
1959 			goto failed_mount;
1960 		}
1961 	}
1962 	blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1963 
1964 	if (blocksize < EXT4_MIN_BLOCK_SIZE ||
1965 	    blocksize > EXT4_MAX_BLOCK_SIZE) {
1966 		printk(KERN_ERR
1967 		       "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
1968 		       blocksize, sb->s_id);
1969 		goto failed_mount;
1970 	}
1971 
1972 	if (sb->s_blocksize != blocksize) {
1973 
1974 		/* Validate the filesystem blocksize */
1975 		if (!sb_set_blocksize(sb, blocksize)) {
1976 			printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
1977 					blocksize);
1978 			goto failed_mount;
1979 		}
1980 
1981 		brelse (bh);
1982 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1983 		offset = do_div(logical_sb_block, blocksize);
1984 		bh = sb_bread(sb, logical_sb_block);
1985 		if (!bh) {
1986 			printk(KERN_ERR
1987 			       "EXT4-fs: Can't read superblock on 2nd try.\n");
1988 			goto failed_mount;
1989 		}
1990 		es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
1991 		sbi->s_es = es;
1992 		if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
1993 			printk (KERN_ERR
1994 				"EXT4-fs: Magic mismatch, very weird !\n");
1995 			goto failed_mount;
1996 		}
1997 	}
1998 
1999 	sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits);
2000 	sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
2001 
2002 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2003 		sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2004 		sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2005 	} else {
2006 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2007 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2008 		if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2009 		    (!is_power_of_2(sbi->s_inode_size)) ||
2010 		    (sbi->s_inode_size > blocksize)) {
2011 			printk (KERN_ERR
2012 				"EXT4-fs: unsupported inode size: %d\n",
2013 				sbi->s_inode_size);
2014 			goto failed_mount;
2015 		}
2016 		if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2017 			sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2018 	}
2019 	sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2020 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2021 		if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2022 		    sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2023 		    !is_power_of_2(sbi->s_desc_size)) {
2024 			printk(KERN_ERR
2025 			       "EXT4-fs: unsupported descriptor size %lu\n",
2026 			       sbi->s_desc_size);
2027 			goto failed_mount;
2028 		}
2029 	} else
2030 		sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2031 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2032 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2033 	if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2034 		goto cantfind_ext4;
2035 	sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2036 	if (sbi->s_inodes_per_block == 0)
2037 		goto cantfind_ext4;
2038 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
2039 					sbi->s_inodes_per_block;
2040 	sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2041 	sbi->s_sbh = bh;
2042 	sbi->s_mount_state = le16_to_cpu(es->s_state);
2043 	sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2044 	sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2045 	for (i=0; i < 4; i++)
2046 		sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2047 	sbi->s_def_hash_version = es->s_def_hash_version;
2048 
2049 	if (sbi->s_blocks_per_group > blocksize * 8) {
2050 		printk (KERN_ERR
2051 			"EXT4-fs: #blocks per group too big: %lu\n",
2052 			sbi->s_blocks_per_group);
2053 		goto failed_mount;
2054 	}
2055 	if (sbi->s_inodes_per_group > blocksize * 8) {
2056 		printk (KERN_ERR
2057 			"EXT4-fs: #inodes per group too big: %lu\n",
2058 			sbi->s_inodes_per_group);
2059 		goto failed_mount;
2060 	}
2061 
2062 	if (ext4_blocks_count(es) >
2063 		    (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2064 		printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2065 			" too large to mount safely\n", sb->s_id);
2066 		if (sizeof(sector_t) < 8)
2067 			printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2068 					"enabled\n");
2069 		goto failed_mount;
2070 	}
2071 
2072 	if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2073 		goto cantfind_ext4;
2074 
2075 	/* ensure blocks_count calculation below doesn't sign-extend */
2076 	if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2077 	    le32_to_cpu(es->s_first_data_block) + 1) {
2078 		printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2079 		       "first data block %u, blocks per group %lu\n",
2080 			ext4_blocks_count(es),
2081 			le32_to_cpu(es->s_first_data_block),
2082 			EXT4_BLOCKS_PER_GROUP(sb));
2083 		goto failed_mount;
2084 	}
2085 	blocks_count = (ext4_blocks_count(es) -
2086 			le32_to_cpu(es->s_first_data_block) +
2087 			EXT4_BLOCKS_PER_GROUP(sb) - 1);
2088 	do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2089 	sbi->s_groups_count = blocks_count;
2090 	db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2091 		   EXT4_DESC_PER_BLOCK(sb);
2092 	sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
2093 				    GFP_KERNEL);
2094 	if (sbi->s_group_desc == NULL) {
2095 		printk (KERN_ERR "EXT4-fs: not enough memory\n");
2096 		goto failed_mount;
2097 	}
2098 
2099 	bgl_lock_init(&sbi->s_blockgroup_lock);
2100 
2101 	for (i = 0; i < db_count; i++) {
2102 		block = descriptor_loc(sb, logical_sb_block, i);
2103 		sbi->s_group_desc[i] = sb_bread(sb, block);
2104 		if (!sbi->s_group_desc[i]) {
2105 			printk (KERN_ERR "EXT4-fs: "
2106 				"can't read group descriptor %d\n", i);
2107 			db_count = i;
2108 			goto failed_mount2;
2109 		}
2110 	}
2111 	if (!ext4_check_descriptors (sb)) {
2112 		printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2113 		goto failed_mount2;
2114 	}
2115 	sbi->s_gdb_count = db_count;
2116 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2117 	spin_lock_init(&sbi->s_next_gen_lock);
2118 
2119 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
2120 			ext4_count_free_blocks(sb));
2121 	if (!err) {
2122 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
2123 				ext4_count_free_inodes(sb));
2124 	}
2125 	if (!err) {
2126 		err = percpu_counter_init(&sbi->s_dirs_counter,
2127 				ext4_count_dirs(sb));
2128 	}
2129 	if (err) {
2130 		printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2131 		goto failed_mount3;
2132 	}
2133 
2134 	/* per fileystem reservation list head & lock */
2135 	spin_lock_init(&sbi->s_rsv_window_lock);
2136 	sbi->s_rsv_window_root = RB_ROOT;
2137 	/* Add a single, static dummy reservation to the start of the
2138 	 * reservation window list --- it gives us a placeholder for
2139 	 * append-at-start-of-list which makes the allocation logic
2140 	 * _much_ simpler. */
2141 	sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2142 	sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2143 	sbi->s_rsv_window_head.rsv_alloc_hit = 0;
2144 	sbi->s_rsv_window_head.rsv_goal_size = 0;
2145 	ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
2146 
2147 	sbi->s_stripe = ext4_get_stripe_size(sbi);
2148 
2149 	/*
2150 	 * set up enough so that it can read an inode
2151 	 */
2152 	sb->s_op = &ext4_sops;
2153 	sb->s_export_op = &ext4_export_ops;
2154 	sb->s_xattr = ext4_xattr_handlers;
2155 #ifdef CONFIG_QUOTA
2156 	sb->s_qcop = &ext4_qctl_operations;
2157 	sb->dq_op = &ext4_quota_operations;
2158 #endif
2159 	INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2160 
2161 	sb->s_root = NULL;
2162 
2163 	needs_recovery = (es->s_last_orphan != 0 ||
2164 			  EXT4_HAS_INCOMPAT_FEATURE(sb,
2165 				    EXT4_FEATURE_INCOMPAT_RECOVER));
2166 
2167 	/*
2168 	 * The first inode we look at is the journal inode.  Don't try
2169 	 * root first: it may be modified in the journal!
2170 	 */
2171 	if (!test_opt(sb, NOLOAD) &&
2172 	    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2173 		if (ext4_load_journal(sb, es, journal_devnum))
2174 			goto failed_mount3;
2175 	} else if (journal_inum) {
2176 		if (ext4_create_journal(sb, es, journal_inum))
2177 			goto failed_mount3;
2178 	} else {
2179 		if (!silent)
2180 			printk (KERN_ERR
2181 				"ext4: No journal on filesystem on %s\n",
2182 				sb->s_id);
2183 		goto failed_mount3;
2184 	}
2185 
2186 	if (ext4_blocks_count(es) > 0xffffffffULL &&
2187 	    !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2188 				       JBD2_FEATURE_INCOMPAT_64BIT)) {
2189 		printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2190 		goto failed_mount4;
2191 	}
2192 
2193 	if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2194 		jbd2_journal_set_features(sbi->s_journal,
2195 				JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2196 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2197 	} else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2198 		jbd2_journal_set_features(sbi->s_journal,
2199 				JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2200 		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2201 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2202 	} else {
2203 		jbd2_journal_clear_features(sbi->s_journal,
2204 				JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2205 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2206 	}
2207 
2208 	/* We have now updated the journal if required, so we can
2209 	 * validate the data journaling mode. */
2210 	switch (test_opt(sb, DATA_FLAGS)) {
2211 	case 0:
2212 		/* No mode set, assume a default based on the journal
2213 		 * capabilities: ORDERED_DATA if the journal can
2214 		 * cope, else JOURNAL_DATA
2215 		 */
2216 		if (jbd2_journal_check_available_features
2217 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2218 			set_opt(sbi->s_mount_opt, ORDERED_DATA);
2219 		else
2220 			set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2221 		break;
2222 
2223 	case EXT4_MOUNT_ORDERED_DATA:
2224 	case EXT4_MOUNT_WRITEBACK_DATA:
2225 		if (!jbd2_journal_check_available_features
2226 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2227 			printk(KERN_ERR "EXT4-fs: Journal does not support "
2228 			       "requested data journaling mode\n");
2229 			goto failed_mount4;
2230 		}
2231 	default:
2232 		break;
2233 	}
2234 
2235 	if (test_opt(sb, NOBH)) {
2236 		if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2237 			printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2238 				"its supported only with writeback mode\n");
2239 			clear_opt(sbi->s_mount_opt, NOBH);
2240 		}
2241 	}
2242 	/*
2243 	 * The jbd2_journal_load will have done any necessary log recovery,
2244 	 * so we can safely mount the rest of the filesystem now.
2245 	 */
2246 
2247 	root = ext4_iget(sb, EXT4_ROOT_INO);
2248 	if (IS_ERR(root)) {
2249 		printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2250 		ret = PTR_ERR(root);
2251 		goto failed_mount4;
2252 	}
2253 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2254 		iput(root);
2255 		printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2256 		goto failed_mount4;
2257 	}
2258 	sb->s_root = d_alloc_root(root);
2259 	if (!sb->s_root) {
2260 		printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2261 		iput(root);
2262 		ret = -ENOMEM;
2263 		goto failed_mount4;
2264 	}
2265 
2266 	ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
2267 
2268 	/* determine the minimum size of new large inodes, if present */
2269 	if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2270 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2271 						     EXT4_GOOD_OLD_INODE_SIZE;
2272 		if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2273 				       EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2274 			if (sbi->s_want_extra_isize <
2275 			    le16_to_cpu(es->s_want_extra_isize))
2276 				sbi->s_want_extra_isize =
2277 					le16_to_cpu(es->s_want_extra_isize);
2278 			if (sbi->s_want_extra_isize <
2279 			    le16_to_cpu(es->s_min_extra_isize))
2280 				sbi->s_want_extra_isize =
2281 					le16_to_cpu(es->s_min_extra_isize);
2282 		}
2283 	}
2284 	/* Check if enough inode space is available */
2285 	if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2286 							sbi->s_inode_size) {
2287 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2288 						       EXT4_GOOD_OLD_INODE_SIZE;
2289 		printk(KERN_INFO "EXT4-fs: required extra inode space not"
2290 			"available.\n");
2291 	}
2292 
2293 	/*
2294 	 * akpm: core read_super() calls in here with the superblock locked.
2295 	 * That deadlocks, because orphan cleanup needs to lock the superblock
2296 	 * in numerous places.  Here we just pop the lock - it's relatively
2297 	 * harmless, because we are now ready to accept write_super() requests,
2298 	 * and aviro says that's the only reason for hanging onto the
2299 	 * superblock lock.
2300 	 */
2301 	EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2302 	ext4_orphan_cleanup(sb, es);
2303 	EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2304 	if (needs_recovery)
2305 		printk (KERN_INFO "EXT4-fs: recovery complete.\n");
2306 	ext4_mark_recovery_complete(sb, es);
2307 	printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2308 		test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2309 		test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
2310 		"writeback");
2311 
2312 	ext4_ext_init(sb);
2313 	ext4_mb_init(sb, needs_recovery);
2314 
2315 	lock_kernel();
2316 	return 0;
2317 
2318 cantfind_ext4:
2319 	if (!silent)
2320 		printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2321 		       sb->s_id);
2322 	goto failed_mount;
2323 
2324 failed_mount4:
2325 	jbd2_journal_destroy(sbi->s_journal);
2326 failed_mount3:
2327 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
2328 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
2329 	percpu_counter_destroy(&sbi->s_dirs_counter);
2330 failed_mount2:
2331 	for (i = 0; i < db_count; i++)
2332 		brelse(sbi->s_group_desc[i]);
2333 	kfree(sbi->s_group_desc);
2334 failed_mount:
2335 #ifdef CONFIG_QUOTA
2336 	for (i = 0; i < MAXQUOTAS; i++)
2337 		kfree(sbi->s_qf_names[i]);
2338 #endif
2339 	ext4_blkdev_remove(sbi);
2340 	brelse(bh);
2341 out_fail:
2342 	sb->s_fs_info = NULL;
2343 	kfree(sbi);
2344 	lock_kernel();
2345 	return ret;
2346 }
2347 
2348 /*
2349  * Setup any per-fs journal parameters now.  We'll do this both on
2350  * initial mount, once the journal has been initialised but before we've
2351  * done any recovery; and again on any subsequent remount.
2352  */
2353 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2354 {
2355 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2356 
2357 	if (sbi->s_commit_interval)
2358 		journal->j_commit_interval = sbi->s_commit_interval;
2359 	/* We could also set up an ext4-specific default for the commit
2360 	 * interval here, but for now we'll just fall back to the jbd
2361 	 * default. */
2362 
2363 	spin_lock(&journal->j_state_lock);
2364 	if (test_opt(sb, BARRIER))
2365 		journal->j_flags |= JBD2_BARRIER;
2366 	else
2367 		journal->j_flags &= ~JBD2_BARRIER;
2368 	spin_unlock(&journal->j_state_lock);
2369 }
2370 
2371 static journal_t *ext4_get_journal(struct super_block *sb,
2372 				   unsigned int journal_inum)
2373 {
2374 	struct inode *journal_inode;
2375 	journal_t *journal;
2376 
2377 	/* First, test for the existence of a valid inode on disk.  Bad
2378 	 * things happen if we iget() an unused inode, as the subsequent
2379 	 * iput() will try to delete it. */
2380 
2381 	journal_inode = ext4_iget(sb, journal_inum);
2382 	if (IS_ERR(journal_inode)) {
2383 		printk(KERN_ERR "EXT4-fs: no journal found.\n");
2384 		return NULL;
2385 	}
2386 	if (!journal_inode->i_nlink) {
2387 		make_bad_inode(journal_inode);
2388 		iput(journal_inode);
2389 		printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2390 		return NULL;
2391 	}
2392 
2393 	jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2394 		  journal_inode, journal_inode->i_size);
2395 	if (!S_ISREG(journal_inode->i_mode)) {
2396 		printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2397 		iput(journal_inode);
2398 		return NULL;
2399 	}
2400 
2401 	journal = jbd2_journal_init_inode(journal_inode);
2402 	if (!journal) {
2403 		printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2404 		iput(journal_inode);
2405 		return NULL;
2406 	}
2407 	journal->j_private = sb;
2408 	ext4_init_journal_params(sb, journal);
2409 	return journal;
2410 }
2411 
2412 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2413 				       dev_t j_dev)
2414 {
2415 	struct buffer_head * bh;
2416 	journal_t *journal;
2417 	ext4_fsblk_t start;
2418 	ext4_fsblk_t len;
2419 	int hblock, blocksize;
2420 	ext4_fsblk_t sb_block;
2421 	unsigned long offset;
2422 	struct ext4_super_block * es;
2423 	struct block_device *bdev;
2424 
2425 	bdev = ext4_blkdev_get(j_dev);
2426 	if (bdev == NULL)
2427 		return NULL;
2428 
2429 	if (bd_claim(bdev, sb)) {
2430 		printk(KERN_ERR
2431 			"EXT4: failed to claim external journal device.\n");
2432 		blkdev_put(bdev);
2433 		return NULL;
2434 	}
2435 
2436 	blocksize = sb->s_blocksize;
2437 	hblock = bdev_hardsect_size(bdev);
2438 	if (blocksize < hblock) {
2439 		printk(KERN_ERR
2440 			"EXT4-fs: blocksize too small for journal device.\n");
2441 		goto out_bdev;
2442 	}
2443 
2444 	sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2445 	offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2446 	set_blocksize(bdev, blocksize);
2447 	if (!(bh = __bread(bdev, sb_block, blocksize))) {
2448 		printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2449 		       "external journal\n");
2450 		goto out_bdev;
2451 	}
2452 
2453 	es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2454 	if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2455 	    !(le32_to_cpu(es->s_feature_incompat) &
2456 	      EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2457 		printk(KERN_ERR "EXT4-fs: external journal has "
2458 					"bad superblock\n");
2459 		brelse(bh);
2460 		goto out_bdev;
2461 	}
2462 
2463 	if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2464 		printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2465 		brelse(bh);
2466 		goto out_bdev;
2467 	}
2468 
2469 	len = ext4_blocks_count(es);
2470 	start = sb_block + 1;
2471 	brelse(bh);	/* we're done with the superblock */
2472 
2473 	journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2474 					start, len, blocksize);
2475 	if (!journal) {
2476 		printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2477 		goto out_bdev;
2478 	}
2479 	journal->j_private = sb;
2480 	ll_rw_block(READ, 1, &journal->j_sb_buffer);
2481 	wait_on_buffer(journal->j_sb_buffer);
2482 	if (!buffer_uptodate(journal->j_sb_buffer)) {
2483 		printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2484 		goto out_journal;
2485 	}
2486 	if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2487 		printk(KERN_ERR "EXT4-fs: External journal has more than one "
2488 					"user (unsupported) - %d\n",
2489 			be32_to_cpu(journal->j_superblock->s_nr_users));
2490 		goto out_journal;
2491 	}
2492 	EXT4_SB(sb)->journal_bdev = bdev;
2493 	ext4_init_journal_params(sb, journal);
2494 	return journal;
2495 out_journal:
2496 	jbd2_journal_destroy(journal);
2497 out_bdev:
2498 	ext4_blkdev_put(bdev);
2499 	return NULL;
2500 }
2501 
2502 static int ext4_load_journal(struct super_block *sb,
2503 			     struct ext4_super_block *es,
2504 			     unsigned long journal_devnum)
2505 {
2506 	journal_t *journal;
2507 	unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2508 	dev_t journal_dev;
2509 	int err = 0;
2510 	int really_read_only;
2511 
2512 	if (journal_devnum &&
2513 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2514 		printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2515 			"numbers have changed\n");
2516 		journal_dev = new_decode_dev(journal_devnum);
2517 	} else
2518 		journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2519 
2520 	really_read_only = bdev_read_only(sb->s_bdev);
2521 
2522 	/*
2523 	 * Are we loading a blank journal or performing recovery after a
2524 	 * crash?  For recovery, we need to check in advance whether we
2525 	 * can get read-write access to the device.
2526 	 */
2527 
2528 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2529 		if (sb->s_flags & MS_RDONLY) {
2530 			printk(KERN_INFO "EXT4-fs: INFO: recovery "
2531 					"required on readonly filesystem.\n");
2532 			if (really_read_only) {
2533 				printk(KERN_ERR "EXT4-fs: write access "
2534 					"unavailable, cannot proceed.\n");
2535 				return -EROFS;
2536 			}
2537 			printk (KERN_INFO "EXT4-fs: write access will "
2538 					"be enabled during recovery.\n");
2539 		}
2540 	}
2541 
2542 	if (journal_inum && journal_dev) {
2543 		printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2544 		       "and inode journals!\n");
2545 		return -EINVAL;
2546 	}
2547 
2548 	if (journal_inum) {
2549 		if (!(journal = ext4_get_journal(sb, journal_inum)))
2550 			return -EINVAL;
2551 	} else {
2552 		if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2553 			return -EINVAL;
2554 	}
2555 
2556 	if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2557 		err = jbd2_journal_update_format(journal);
2558 		if (err)  {
2559 			printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2560 			jbd2_journal_destroy(journal);
2561 			return err;
2562 		}
2563 	}
2564 
2565 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2566 		err = jbd2_journal_wipe(journal, !really_read_only);
2567 	if (!err)
2568 		err = jbd2_journal_load(journal);
2569 
2570 	if (err) {
2571 		printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2572 		jbd2_journal_destroy(journal);
2573 		return err;
2574 	}
2575 
2576 	EXT4_SB(sb)->s_journal = journal;
2577 	ext4_clear_journal_err(sb, es);
2578 
2579 	if (journal_devnum &&
2580 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2581 		es->s_journal_dev = cpu_to_le32(journal_devnum);
2582 		sb->s_dirt = 1;
2583 
2584 		/* Make sure we flush the recovery flag to disk. */
2585 		ext4_commit_super(sb, es, 1);
2586 	}
2587 
2588 	return 0;
2589 }
2590 
2591 static int ext4_create_journal(struct super_block * sb,
2592 			       struct ext4_super_block * es,
2593 			       unsigned int journal_inum)
2594 {
2595 	journal_t *journal;
2596 	int err;
2597 
2598 	if (sb->s_flags & MS_RDONLY) {
2599 		printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
2600 				"create journal.\n");
2601 		return -EROFS;
2602 	}
2603 
2604 	journal = ext4_get_journal(sb, journal_inum);
2605 	if (!journal)
2606 		return -EINVAL;
2607 
2608 	printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
2609 	       journal_inum);
2610 
2611 	err = jbd2_journal_create(journal);
2612 	if (err) {
2613 		printk(KERN_ERR "EXT4-fs: error creating journal.\n");
2614 		jbd2_journal_destroy(journal);
2615 		return -EIO;
2616 	}
2617 
2618 	EXT4_SB(sb)->s_journal = journal;
2619 
2620 	ext4_update_dynamic_rev(sb);
2621 	EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2622 	EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
2623 
2624 	es->s_journal_inum = cpu_to_le32(journal_inum);
2625 	sb->s_dirt = 1;
2626 
2627 	/* Make sure we flush the recovery flag to disk. */
2628 	ext4_commit_super(sb, es, 1);
2629 
2630 	return 0;
2631 }
2632 
2633 static void ext4_commit_super (struct super_block * sb,
2634 			       struct ext4_super_block * es,
2635 			       int sync)
2636 {
2637 	struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2638 
2639 	if (!sbh)
2640 		return;
2641 	es->s_wtime = cpu_to_le32(get_seconds());
2642 	ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
2643 	es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
2644 	BUFFER_TRACE(sbh, "marking dirty");
2645 	mark_buffer_dirty(sbh);
2646 	if (sync)
2647 		sync_dirty_buffer(sbh);
2648 }
2649 
2650 
2651 /*
2652  * Have we just finished recovery?  If so, and if we are mounting (or
2653  * remounting) the filesystem readonly, then we will end up with a
2654  * consistent fs on disk.  Record that fact.
2655  */
2656 static void ext4_mark_recovery_complete(struct super_block * sb,
2657 					struct ext4_super_block * es)
2658 {
2659 	journal_t *journal = EXT4_SB(sb)->s_journal;
2660 
2661 	jbd2_journal_lock_updates(journal);
2662 	jbd2_journal_flush(journal);
2663 	lock_super(sb);
2664 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2665 	    sb->s_flags & MS_RDONLY) {
2666 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2667 		sb->s_dirt = 0;
2668 		ext4_commit_super(sb, es, 1);
2669 	}
2670 	unlock_super(sb);
2671 	jbd2_journal_unlock_updates(journal);
2672 }
2673 
2674 /*
2675  * If we are mounting (or read-write remounting) a filesystem whose journal
2676  * has recorded an error from a previous lifetime, move that error to the
2677  * main filesystem now.
2678  */
2679 static void ext4_clear_journal_err(struct super_block * sb,
2680 				   struct ext4_super_block * es)
2681 {
2682 	journal_t *journal;
2683 	int j_errno;
2684 	const char *errstr;
2685 
2686 	journal = EXT4_SB(sb)->s_journal;
2687 
2688 	/*
2689 	 * Now check for any error status which may have been recorded in the
2690 	 * journal by a prior ext4_error() or ext4_abort()
2691 	 */
2692 
2693 	j_errno = jbd2_journal_errno(journal);
2694 	if (j_errno) {
2695 		char nbuf[16];
2696 
2697 		errstr = ext4_decode_error(sb, j_errno, nbuf);
2698 		ext4_warning(sb, __func__, "Filesystem error recorded "
2699 			     "from previous mount: %s", errstr);
2700 		ext4_warning(sb, __func__, "Marking fs in need of "
2701 			     "filesystem check.");
2702 
2703 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2704 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2705 		ext4_commit_super (sb, es, 1);
2706 
2707 		jbd2_journal_clear_err(journal);
2708 	}
2709 }
2710 
2711 /*
2712  * Force the running and committing transactions to commit,
2713  * and wait on the commit.
2714  */
2715 int ext4_force_commit(struct super_block *sb)
2716 {
2717 	journal_t *journal;
2718 	int ret;
2719 
2720 	if (sb->s_flags & MS_RDONLY)
2721 		return 0;
2722 
2723 	journal = EXT4_SB(sb)->s_journal;
2724 	sb->s_dirt = 0;
2725 	ret = ext4_journal_force_commit(journal);
2726 	return ret;
2727 }
2728 
2729 /*
2730  * Ext4 always journals updates to the superblock itself, so we don't
2731  * have to propagate any other updates to the superblock on disk at this
2732  * point.  Just start an async writeback to get the buffers on their way
2733  * to the disk.
2734  *
2735  * This implicitly triggers the writebehind on sync().
2736  */
2737 
2738 static void ext4_write_super (struct super_block * sb)
2739 {
2740 	if (mutex_trylock(&sb->s_lock) != 0)
2741 		BUG();
2742 	sb->s_dirt = 0;
2743 }
2744 
2745 static int ext4_sync_fs(struct super_block *sb, int wait)
2746 {
2747 	tid_t target;
2748 
2749 	sb->s_dirt = 0;
2750 	if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
2751 		if (wait)
2752 			jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
2753 	}
2754 	return 0;
2755 }
2756 
2757 /*
2758  * LVM calls this function before a (read-only) snapshot is created.  This
2759  * gives us a chance to flush the journal completely and mark the fs clean.
2760  */
2761 static void ext4_write_super_lockfs(struct super_block *sb)
2762 {
2763 	sb->s_dirt = 0;
2764 
2765 	if (!(sb->s_flags & MS_RDONLY)) {
2766 		journal_t *journal = EXT4_SB(sb)->s_journal;
2767 
2768 		/* Now we set up the journal barrier. */
2769 		jbd2_journal_lock_updates(journal);
2770 		jbd2_journal_flush(journal);
2771 
2772 		/* Journal blocked and flushed, clear needs_recovery flag. */
2773 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2774 		ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2775 	}
2776 }
2777 
2778 /*
2779  * Called by LVM after the snapshot is done.  We need to reset the RECOVER
2780  * flag here, even though the filesystem is not technically dirty yet.
2781  */
2782 static void ext4_unlockfs(struct super_block *sb)
2783 {
2784 	if (!(sb->s_flags & MS_RDONLY)) {
2785 		lock_super(sb);
2786 		/* Reser the needs_recovery flag before the fs is unlocked. */
2787 		EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2788 		ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2789 		unlock_super(sb);
2790 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
2791 	}
2792 }
2793 
2794 static int ext4_remount (struct super_block * sb, int * flags, char * data)
2795 {
2796 	struct ext4_super_block * es;
2797 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2798 	ext4_fsblk_t n_blocks_count = 0;
2799 	unsigned long old_sb_flags;
2800 	struct ext4_mount_options old_opts;
2801 	int err;
2802 #ifdef CONFIG_QUOTA
2803 	int i;
2804 #endif
2805 
2806 	/* Store the original options */
2807 	old_sb_flags = sb->s_flags;
2808 	old_opts.s_mount_opt = sbi->s_mount_opt;
2809 	old_opts.s_resuid = sbi->s_resuid;
2810 	old_opts.s_resgid = sbi->s_resgid;
2811 	old_opts.s_commit_interval = sbi->s_commit_interval;
2812 #ifdef CONFIG_QUOTA
2813 	old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2814 	for (i = 0; i < MAXQUOTAS; i++)
2815 		old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2816 #endif
2817 
2818 	/*
2819 	 * Allow the "check" option to be passed as a remount option.
2820 	 */
2821 	if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2822 		err = -EINVAL;
2823 		goto restore_opts;
2824 	}
2825 
2826 	if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
2827 		ext4_abort(sb, __func__, "Abort forced by user");
2828 
2829 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2830 		((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2831 
2832 	es = sbi->s_es;
2833 
2834 	ext4_init_journal_params(sb, sbi->s_journal);
2835 
2836 	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2837 		n_blocks_count > ext4_blocks_count(es)) {
2838 		if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
2839 			err = -EROFS;
2840 			goto restore_opts;
2841 		}
2842 
2843 		if (*flags & MS_RDONLY) {
2844 			/*
2845 			 * First of all, the unconditional stuff we have to do
2846 			 * to disable replay of the journal when we next remount
2847 			 */
2848 			sb->s_flags |= MS_RDONLY;
2849 
2850 			/*
2851 			 * OK, test if we are remounting a valid rw partition
2852 			 * readonly, and if so set the rdonly flag and then
2853 			 * mark the partition as valid again.
2854 			 */
2855 			if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
2856 			    (sbi->s_mount_state & EXT4_VALID_FS))
2857 				es->s_state = cpu_to_le16(sbi->s_mount_state);
2858 
2859 			/*
2860 			 * We have to unlock super so that we can wait for
2861 			 * transactions.
2862 			 */
2863 			unlock_super(sb);
2864 			ext4_mark_recovery_complete(sb, es);
2865 			lock_super(sb);
2866 		} else {
2867 			__le32 ret;
2868 			if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2869 					~EXT4_FEATURE_RO_COMPAT_SUPP))) {
2870 				printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2871 				       "remount RDWR because of unsupported "
2872 				       "optional features (%x).\n",
2873 				       sb->s_id, le32_to_cpu(ret));
2874 				err = -EROFS;
2875 				goto restore_opts;
2876 			}
2877 
2878 			/*
2879 			 * If we have an unprocessed orphan list hanging
2880 			 * around from a previously readonly bdev mount,
2881 			 * require a full umount/remount for now.
2882 			 */
2883 			if (es->s_last_orphan) {
2884 				printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2885 				       "remount RDWR because of unprocessed "
2886 				       "orphan inode list.  Please "
2887 				       "umount/remount instead.\n",
2888 				       sb->s_id);
2889 				err = -EINVAL;
2890 				goto restore_opts;
2891 			}
2892 
2893 			/*
2894 			 * Mounting a RDONLY partition read-write, so reread
2895 			 * and store the current valid flag.  (It may have
2896 			 * been changed by e2fsck since we originally mounted
2897 			 * the partition.)
2898 			 */
2899 			ext4_clear_journal_err(sb, es);
2900 			sbi->s_mount_state = le16_to_cpu(es->s_state);
2901 			if ((err = ext4_group_extend(sb, es, n_blocks_count)))
2902 				goto restore_opts;
2903 			if (!ext4_setup_super (sb, es, 0))
2904 				sb->s_flags &= ~MS_RDONLY;
2905 		}
2906 	}
2907 #ifdef CONFIG_QUOTA
2908 	/* Release old quota file names */
2909 	for (i = 0; i < MAXQUOTAS; i++)
2910 		if (old_opts.s_qf_names[i] &&
2911 		    old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2912 			kfree(old_opts.s_qf_names[i]);
2913 #endif
2914 	return 0;
2915 restore_opts:
2916 	sb->s_flags = old_sb_flags;
2917 	sbi->s_mount_opt = old_opts.s_mount_opt;
2918 	sbi->s_resuid = old_opts.s_resuid;
2919 	sbi->s_resgid = old_opts.s_resgid;
2920 	sbi->s_commit_interval = old_opts.s_commit_interval;
2921 #ifdef CONFIG_QUOTA
2922 	sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2923 	for (i = 0; i < MAXQUOTAS; i++) {
2924 		if (sbi->s_qf_names[i] &&
2925 		    old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2926 			kfree(sbi->s_qf_names[i]);
2927 		sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2928 	}
2929 #endif
2930 	return err;
2931 }
2932 
2933 static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf)
2934 {
2935 	struct super_block *sb = dentry->d_sb;
2936 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2937 	struct ext4_super_block *es = sbi->s_es;
2938 	u64 fsid;
2939 
2940 	if (test_opt(sb, MINIX_DF)) {
2941 		sbi->s_overhead_last = 0;
2942 	} else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
2943 		ext4_group_t ngroups = sbi->s_groups_count, i;
2944 		ext4_fsblk_t overhead = 0;
2945 		smp_rmb();
2946 
2947 		/*
2948 		 * Compute the overhead (FS structures).  This is constant
2949 		 * for a given filesystem unless the number of block groups
2950 		 * changes so we cache the previous value until it does.
2951 		 */
2952 
2953 		/*
2954 		 * All of the blocks before first_data_block are
2955 		 * overhead
2956 		 */
2957 		overhead = le32_to_cpu(es->s_first_data_block);
2958 
2959 		/*
2960 		 * Add the overhead attributed to the superblock and
2961 		 * block group descriptors.  If the sparse superblocks
2962 		 * feature is turned on, then not all groups have this.
2963 		 */
2964 		for (i = 0; i < ngroups; i++) {
2965 			overhead += ext4_bg_has_super(sb, i) +
2966 				ext4_bg_num_gdb(sb, i);
2967 			cond_resched();
2968 		}
2969 
2970 		/*
2971 		 * Every block group has an inode bitmap, a block
2972 		 * bitmap, and an inode table.
2973 		 */
2974 		overhead += ngroups * (2 + sbi->s_itb_per_group);
2975 		sbi->s_overhead_last = overhead;
2976 		smp_wmb();
2977 		sbi->s_blocks_last = ext4_blocks_count(es);
2978 	}
2979 
2980 	buf->f_type = EXT4_SUPER_MAGIC;
2981 	buf->f_bsize = sb->s_blocksize;
2982 	buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
2983 	buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2984 	ext4_free_blocks_count_set(es, buf->f_bfree);
2985 	buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
2986 	if (buf->f_bfree < ext4_r_blocks_count(es))
2987 		buf->f_bavail = 0;
2988 	buf->f_files = le32_to_cpu(es->s_inodes_count);
2989 	buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2990 	es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
2991 	buf->f_namelen = EXT4_NAME_LEN;
2992 	fsid = le64_to_cpup((void *)es->s_uuid) ^
2993 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2994 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2995 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2996 	return 0;
2997 }
2998 
2999 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3000  * is locked for write. Otherwise the are possible deadlocks:
3001  * Process 1                         Process 2
3002  * ext4_create()                     quota_sync()
3003  *   jbd2_journal_start()                   write_dquot()
3004  *   DQUOT_INIT()                        down(dqio_mutex)
3005  *     down(dqio_mutex)                    jbd2_journal_start()
3006  *
3007  */
3008 
3009 #ifdef CONFIG_QUOTA
3010 
3011 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3012 {
3013 	return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3014 }
3015 
3016 static int ext4_dquot_initialize(struct inode *inode, int type)
3017 {
3018 	handle_t *handle;
3019 	int ret, err;
3020 
3021 	/* We may create quota structure so we need to reserve enough blocks */
3022 	handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
3023 	if (IS_ERR(handle))
3024 		return PTR_ERR(handle);
3025 	ret = dquot_initialize(inode, type);
3026 	err = ext4_journal_stop(handle);
3027 	if (!ret)
3028 		ret = err;
3029 	return ret;
3030 }
3031 
3032 static int ext4_dquot_drop(struct inode *inode)
3033 {
3034 	handle_t *handle;
3035 	int ret, err;
3036 
3037 	/* We may delete quota structure so we need to reserve enough blocks */
3038 	handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
3039 	if (IS_ERR(handle)) {
3040 		/*
3041 		 * We call dquot_drop() anyway to at least release references
3042 		 * to quota structures so that umount does not hang.
3043 		 */
3044 		dquot_drop(inode);
3045 		return PTR_ERR(handle);
3046 	}
3047 	ret = dquot_drop(inode);
3048 	err = ext4_journal_stop(handle);
3049 	if (!ret)
3050 		ret = err;
3051 	return ret;
3052 }
3053 
3054 static int ext4_write_dquot(struct dquot *dquot)
3055 {
3056 	int ret, err;
3057 	handle_t *handle;
3058 	struct inode *inode;
3059 
3060 	inode = dquot_to_inode(dquot);
3061 	handle = ext4_journal_start(inode,
3062 					EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3063 	if (IS_ERR(handle))
3064 		return PTR_ERR(handle);
3065 	ret = dquot_commit(dquot);
3066 	err = ext4_journal_stop(handle);
3067 	if (!ret)
3068 		ret = err;
3069 	return ret;
3070 }
3071 
3072 static int ext4_acquire_dquot(struct dquot *dquot)
3073 {
3074 	int ret, err;
3075 	handle_t *handle;
3076 
3077 	handle = ext4_journal_start(dquot_to_inode(dquot),
3078 					EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3079 	if (IS_ERR(handle))
3080 		return PTR_ERR(handle);
3081 	ret = dquot_acquire(dquot);
3082 	err = ext4_journal_stop(handle);
3083 	if (!ret)
3084 		ret = err;
3085 	return ret;
3086 }
3087 
3088 static int ext4_release_dquot(struct dquot *dquot)
3089 {
3090 	int ret, err;
3091 	handle_t *handle;
3092 
3093 	handle = ext4_journal_start(dquot_to_inode(dquot),
3094 					EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3095 	if (IS_ERR(handle)) {
3096 		/* Release dquot anyway to avoid endless cycle in dqput() */
3097 		dquot_release(dquot);
3098 		return PTR_ERR(handle);
3099 	}
3100 	ret = dquot_release(dquot);
3101 	err = ext4_journal_stop(handle);
3102 	if (!ret)
3103 		ret = err;
3104 	return ret;
3105 }
3106 
3107 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3108 {
3109 	/* Are we journalling quotas? */
3110 	if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3111 	    EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3112 		dquot_mark_dquot_dirty(dquot);
3113 		return ext4_write_dquot(dquot);
3114 	} else {
3115 		return dquot_mark_dquot_dirty(dquot);
3116 	}
3117 }
3118 
3119 static int ext4_write_info(struct super_block *sb, int type)
3120 {
3121 	int ret, err;
3122 	handle_t *handle;
3123 
3124 	/* Data block + inode block */
3125 	handle = ext4_journal_start(sb->s_root->d_inode, 2);
3126 	if (IS_ERR(handle))
3127 		return PTR_ERR(handle);
3128 	ret = dquot_commit_info(sb, type);
3129 	err = ext4_journal_stop(handle);
3130 	if (!ret)
3131 		ret = err;
3132 	return ret;
3133 }
3134 
3135 /*
3136  * Turn on quotas during mount time - we need to find
3137  * the quota file and such...
3138  */
3139 static int ext4_quota_on_mount(struct super_block *sb, int type)
3140 {
3141 	return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3142 			EXT4_SB(sb)->s_jquota_fmt, type);
3143 }
3144 
3145 /*
3146  * Standard function to be called on quota_on
3147  */
3148 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3149 			 char *path, int remount)
3150 {
3151 	int err;
3152 	struct nameidata nd;
3153 
3154 	if (!test_opt(sb, QUOTA))
3155 		return -EINVAL;
3156 	/* Not journalling quota? */
3157 	if ((!EXT4_SB(sb)->s_qf_names[USRQUOTA] &&
3158 	    !EXT4_SB(sb)->s_qf_names[GRPQUOTA]) || remount)
3159 		return vfs_quota_on(sb, type, format_id, path, remount);
3160 	err = path_lookup(path, LOOKUP_FOLLOW, &nd);
3161 	if (err)
3162 		return err;
3163 	/* Quotafile not on the same filesystem? */
3164 	if (nd.path.mnt->mnt_sb != sb) {
3165 		path_put(&nd.path);
3166 		return -EXDEV;
3167 	}
3168 	/* Quotafile not of fs root? */
3169 	if (nd.path.dentry->d_parent->d_inode != sb->s_root->d_inode)
3170 		printk(KERN_WARNING
3171 			"EXT4-fs: Quota file not on filesystem root. "
3172 			"Journalled quota will not work.\n");
3173 	path_put(&nd.path);
3174 	return vfs_quota_on(sb, type, format_id, path, remount);
3175 }
3176 
3177 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3178  * acquiring the locks... As quota files are never truncated and quota code
3179  * itself serializes the operations (and noone else should touch the files)
3180  * we don't have to be afraid of races */
3181 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3182 			       size_t len, loff_t off)
3183 {
3184 	struct inode *inode = sb_dqopt(sb)->files[type];
3185 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3186 	int err = 0;
3187 	int offset = off & (sb->s_blocksize - 1);
3188 	int tocopy;
3189 	size_t toread;
3190 	struct buffer_head *bh;
3191 	loff_t i_size = i_size_read(inode);
3192 
3193 	if (off > i_size)
3194 		return 0;
3195 	if (off+len > i_size)
3196 		len = i_size-off;
3197 	toread = len;
3198 	while (toread > 0) {
3199 		tocopy = sb->s_blocksize - offset < toread ?
3200 				sb->s_blocksize - offset : toread;
3201 		bh = ext4_bread(NULL, inode, blk, 0, &err);
3202 		if (err)
3203 			return err;
3204 		if (!bh)	/* A hole? */
3205 			memset(data, 0, tocopy);
3206 		else
3207 			memcpy(data, bh->b_data+offset, tocopy);
3208 		brelse(bh);
3209 		offset = 0;
3210 		toread -= tocopy;
3211 		data += tocopy;
3212 		blk++;
3213 	}
3214 	return len;
3215 }
3216 
3217 /* Write to quotafile (we know the transaction is already started and has
3218  * enough credits) */
3219 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3220 				const char *data, size_t len, loff_t off)
3221 {
3222 	struct inode *inode = sb_dqopt(sb)->files[type];
3223 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3224 	int err = 0;
3225 	int offset = off & (sb->s_blocksize - 1);
3226 	int tocopy;
3227 	int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3228 	size_t towrite = len;
3229 	struct buffer_head *bh;
3230 	handle_t *handle = journal_current_handle();
3231 
3232 	if (!handle) {
3233 		printk(KERN_WARNING "EXT4-fs: Quota write (off=%Lu, len=%Lu)"
3234 			" cancelled because transaction is not started.\n",
3235 			(unsigned long long)off, (unsigned long long)len);
3236 		return -EIO;
3237 	}
3238 	mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3239 	while (towrite > 0) {
3240 		tocopy = sb->s_blocksize - offset < towrite ?
3241 				sb->s_blocksize - offset : towrite;
3242 		bh = ext4_bread(handle, inode, blk, 1, &err);
3243 		if (!bh)
3244 			goto out;
3245 		if (journal_quota) {
3246 			err = ext4_journal_get_write_access(handle, bh);
3247 			if (err) {
3248 				brelse(bh);
3249 				goto out;
3250 			}
3251 		}
3252 		lock_buffer(bh);
3253 		memcpy(bh->b_data+offset, data, tocopy);
3254 		flush_dcache_page(bh->b_page);
3255 		unlock_buffer(bh);
3256 		if (journal_quota)
3257 			err = ext4_journal_dirty_metadata(handle, bh);
3258 		else {
3259 			/* Always do at least ordered writes for quotas */
3260 			err = ext4_journal_dirty_data(handle, bh);
3261 			mark_buffer_dirty(bh);
3262 		}
3263 		brelse(bh);
3264 		if (err)
3265 			goto out;
3266 		offset = 0;
3267 		towrite -= tocopy;
3268 		data += tocopy;
3269 		blk++;
3270 	}
3271 out:
3272 	if (len == towrite)
3273 		return err;
3274 	if (inode->i_size < off+len-towrite) {
3275 		i_size_write(inode, off+len-towrite);
3276 		EXT4_I(inode)->i_disksize = inode->i_size;
3277 	}
3278 	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3279 	ext4_mark_inode_dirty(handle, inode);
3280 	mutex_unlock(&inode->i_mutex);
3281 	return len - towrite;
3282 }
3283 
3284 #endif
3285 
3286 static int ext4_get_sb(struct file_system_type *fs_type,
3287 	int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3288 {
3289 	return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3290 }
3291 
3292 static struct file_system_type ext4dev_fs_type = {
3293 	.owner		= THIS_MODULE,
3294 	.name		= "ext4dev",
3295 	.get_sb		= ext4_get_sb,
3296 	.kill_sb	= kill_block_super,
3297 	.fs_flags	= FS_REQUIRES_DEV,
3298 };
3299 
3300 static int __init init_ext4_fs(void)
3301 {
3302 	int err;
3303 
3304 	err = init_ext4_mballoc();
3305 	if (err)
3306 		return err;
3307 
3308 	err = init_ext4_xattr();
3309 	if (err)
3310 		goto out2;
3311 	err = init_inodecache();
3312 	if (err)
3313 		goto out1;
3314 	err = register_filesystem(&ext4dev_fs_type);
3315 	if (err)
3316 		goto out;
3317 	return 0;
3318 out:
3319 	destroy_inodecache();
3320 out1:
3321 	exit_ext4_xattr();
3322 out2:
3323 	exit_ext4_mballoc();
3324 	return err;
3325 }
3326 
3327 static void __exit exit_ext4_fs(void)
3328 {
3329 	unregister_filesystem(&ext4dev_fs_type);
3330 	destroy_inodecache();
3331 	exit_ext4_xattr();
3332 	exit_ext4_mballoc();
3333 }
3334 
3335 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3336 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
3337 MODULE_LICENSE("GPL");
3338 module_init(init_ext4_fs)
3339 module_exit(exit_ext4_fs)
3340