xref: /openbmc/linux/fs/ext4/super.c (revision 725d26d3f09ccb5bac4b4293096b985a312a0d67)
1 /*
2  *  linux/fs/ext4/super.c
3  *
4  * Copyright (C) 1992, 1993, 1994, 1995
5  * Remy Card (card@masi.ibp.fr)
6  * Laboratoire MASI - Institut Blaise Pascal
7  * Universite Pierre et Marie Curie (Paris VI)
8  *
9  *  from
10  *
11  *  linux/fs/minix/inode.c
12  *
13  *  Copyright (C) 1991, 1992  Linus Torvalds
14  *
15  *  Big-endian to little-endian byte-swapping/bitmaps by
16  *        David S. Miller (davem@caip.rutgers.edu), 1995
17  */
18 
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/time.h>
23 #include <linux/jbd2.h>
24 #include <linux/ext4_fs.h>
25 #include <linux/ext4_jbd2.h>
26 #include <linux/slab.h>
27 #include <linux/init.h>
28 #include <linux/blkdev.h>
29 #include <linux/parser.h>
30 #include <linux/smp_lock.h>
31 #include <linux/buffer_head.h>
32 #include <linux/exportfs.h>
33 #include <linux/vfs.h>
34 #include <linux/random.h>
35 #include <linux/mount.h>
36 #include <linux/namei.h>
37 #include <linux/quotaops.h>
38 #include <linux/seq_file.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 
42 #include <asm/uaccess.h>
43 
44 #include "xattr.h"
45 #include "acl.h"
46 #include "namei.h"
47 #include "group.h"
48 
49 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
50 			     unsigned long journal_devnum);
51 static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
52 			       unsigned int);
53 static void ext4_commit_super (struct super_block * sb,
54 			       struct ext4_super_block * es,
55 			       int sync);
56 static void ext4_mark_recovery_complete(struct super_block * sb,
57 					struct ext4_super_block * es);
58 static void ext4_clear_journal_err(struct super_block * sb,
59 				   struct ext4_super_block * es);
60 static int ext4_sync_fs(struct super_block *sb, int wait);
61 static const char *ext4_decode_error(struct super_block * sb, int errno,
62 				     char nbuf[16]);
63 static int ext4_remount (struct super_block * sb, int * flags, char * data);
64 static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf);
65 static void ext4_unlockfs(struct super_block *sb);
66 static void ext4_write_super (struct super_block * sb);
67 static void ext4_write_super_lockfs(struct super_block *sb);
68 
69 
70 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
71 			       struct ext4_group_desc *bg)
72 {
73 	return le32_to_cpu(bg->bg_block_bitmap_lo) |
74 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
75 		(ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
76 }
77 
78 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
79 			       struct ext4_group_desc *bg)
80 {
81 	return le32_to_cpu(bg->bg_inode_bitmap_lo) |
82 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
83 		(ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
84 }
85 
86 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
87 			      struct ext4_group_desc *bg)
88 {
89 	return le32_to_cpu(bg->bg_inode_table_lo) |
90 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
91 		(ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
92 }
93 
94 void ext4_block_bitmap_set(struct super_block *sb,
95 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
96 {
97 	bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
98 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
99 		bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
100 }
101 
102 void ext4_inode_bitmap_set(struct super_block *sb,
103 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
104 {
105 	bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
106 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
107 		bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
108 }
109 
110 void ext4_inode_table_set(struct super_block *sb,
111 			  struct ext4_group_desc *bg, ext4_fsblk_t blk)
112 {
113 	bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
114 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
115 		bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
116 }
117 
118 /*
119  * Wrappers for jbd2_journal_start/end.
120  *
121  * The only special thing we need to do here is to make sure that all
122  * journal_end calls result in the superblock being marked dirty, so
123  * that sync() will call the filesystem's write_super callback if
124  * appropriate.
125  */
126 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
127 {
128 	journal_t *journal;
129 
130 	if (sb->s_flags & MS_RDONLY)
131 		return ERR_PTR(-EROFS);
132 
133 	/* Special case here: if the journal has aborted behind our
134 	 * backs (eg. EIO in the commit thread), then we still need to
135 	 * take the FS itself readonly cleanly. */
136 	journal = EXT4_SB(sb)->s_journal;
137 	if (is_journal_aborted(journal)) {
138 		ext4_abort(sb, __FUNCTION__,
139 			   "Detected aborted journal");
140 		return ERR_PTR(-EROFS);
141 	}
142 
143 	return jbd2_journal_start(journal, nblocks);
144 }
145 
146 /*
147  * The only special thing we need to do here is to make sure that all
148  * jbd2_journal_stop calls result in the superblock being marked dirty, so
149  * that sync() will call the filesystem's write_super callback if
150  * appropriate.
151  */
152 int __ext4_journal_stop(const char *where, handle_t *handle)
153 {
154 	struct super_block *sb;
155 	int err;
156 	int rc;
157 
158 	sb = handle->h_transaction->t_journal->j_private;
159 	err = handle->h_err;
160 	rc = jbd2_journal_stop(handle);
161 
162 	if (!err)
163 		err = rc;
164 	if (err)
165 		__ext4_std_error(sb, where, err);
166 	return err;
167 }
168 
169 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
170 		struct buffer_head *bh, handle_t *handle, int err)
171 {
172 	char nbuf[16];
173 	const char *errstr = ext4_decode_error(NULL, err, nbuf);
174 
175 	if (bh)
176 		BUFFER_TRACE(bh, "abort");
177 
178 	if (!handle->h_err)
179 		handle->h_err = err;
180 
181 	if (is_handle_aborted(handle))
182 		return;
183 
184 	printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
185 	       caller, errstr, err_fn);
186 
187 	jbd2_journal_abort_handle(handle);
188 }
189 
190 /* Deal with the reporting of failure conditions on a filesystem such as
191  * inconsistencies detected or read IO failures.
192  *
193  * On ext2, we can store the error state of the filesystem in the
194  * superblock.  That is not possible on ext4, because we may have other
195  * write ordering constraints on the superblock which prevent us from
196  * writing it out straight away; and given that the journal is about to
197  * be aborted, we can't rely on the current, or future, transactions to
198  * write out the superblock safely.
199  *
200  * We'll just use the jbd2_journal_abort() error code to record an error in
201  * the journal instead.  On recovery, the journal will compain about
202  * that error until we've noted it down and cleared it.
203  */
204 
205 static void ext4_handle_error(struct super_block *sb)
206 {
207 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
208 
209 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
210 	es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
211 
212 	if (sb->s_flags & MS_RDONLY)
213 		return;
214 
215 	if (!test_opt (sb, ERRORS_CONT)) {
216 		journal_t *journal = EXT4_SB(sb)->s_journal;
217 
218 		EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
219 		if (journal)
220 			jbd2_journal_abort(journal, -EIO);
221 	}
222 	if (test_opt (sb, ERRORS_RO)) {
223 		printk (KERN_CRIT "Remounting filesystem read-only\n");
224 		sb->s_flags |= MS_RDONLY;
225 	}
226 	ext4_commit_super(sb, es, 1);
227 	if (test_opt(sb, ERRORS_PANIC))
228 		panic("EXT4-fs (device %s): panic forced after error\n",
229 			sb->s_id);
230 }
231 
232 void ext4_error (struct super_block * sb, const char * function,
233 		 const char * fmt, ...)
234 {
235 	va_list args;
236 
237 	va_start(args, fmt);
238 	printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
239 	vprintk(fmt, args);
240 	printk("\n");
241 	va_end(args);
242 
243 	ext4_handle_error(sb);
244 }
245 
246 static const char *ext4_decode_error(struct super_block * sb, int errno,
247 				     char nbuf[16])
248 {
249 	char *errstr = NULL;
250 
251 	switch (errno) {
252 	case -EIO:
253 		errstr = "IO failure";
254 		break;
255 	case -ENOMEM:
256 		errstr = "Out of memory";
257 		break;
258 	case -EROFS:
259 		if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
260 			errstr = "Journal has aborted";
261 		else
262 			errstr = "Readonly filesystem";
263 		break;
264 	default:
265 		/* If the caller passed in an extra buffer for unknown
266 		 * errors, textualise them now.  Else we just return
267 		 * NULL. */
268 		if (nbuf) {
269 			/* Check for truncated error codes... */
270 			if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
271 				errstr = nbuf;
272 		}
273 		break;
274 	}
275 
276 	return errstr;
277 }
278 
279 /* __ext4_std_error decodes expected errors from journaling functions
280  * automatically and invokes the appropriate error response.  */
281 
282 void __ext4_std_error (struct super_block * sb, const char * function,
283 		       int errno)
284 {
285 	char nbuf[16];
286 	const char *errstr;
287 
288 	/* Special case: if the error is EROFS, and we're not already
289 	 * inside a transaction, then there's really no point in logging
290 	 * an error. */
291 	if (errno == -EROFS && journal_current_handle() == NULL &&
292 	    (sb->s_flags & MS_RDONLY))
293 		return;
294 
295 	errstr = ext4_decode_error(sb, errno, nbuf);
296 	printk (KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
297 		sb->s_id, function, errstr);
298 
299 	ext4_handle_error(sb);
300 }
301 
302 /*
303  * ext4_abort is a much stronger failure handler than ext4_error.  The
304  * abort function may be used to deal with unrecoverable failures such
305  * as journal IO errors or ENOMEM at a critical moment in log management.
306  *
307  * We unconditionally force the filesystem into an ABORT|READONLY state,
308  * unless the error response on the fs has been set to panic in which
309  * case we take the easy way out and panic immediately.
310  */
311 
312 void ext4_abort (struct super_block * sb, const char * function,
313 		 const char * fmt, ...)
314 {
315 	va_list args;
316 
317 	printk (KERN_CRIT "ext4_abort called.\n");
318 
319 	va_start(args, fmt);
320 	printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
321 	vprintk(fmt, args);
322 	printk("\n");
323 	va_end(args);
324 
325 	if (test_opt(sb, ERRORS_PANIC))
326 		panic("EXT4-fs panic from previous error\n");
327 
328 	if (sb->s_flags & MS_RDONLY)
329 		return;
330 
331 	printk(KERN_CRIT "Remounting filesystem read-only\n");
332 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
333 	sb->s_flags |= MS_RDONLY;
334 	EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
335 	jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
336 }
337 
338 void ext4_warning (struct super_block * sb, const char * function,
339 		   const char * fmt, ...)
340 {
341 	va_list args;
342 
343 	va_start(args, fmt);
344 	printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
345 	       sb->s_id, function);
346 	vprintk(fmt, args);
347 	printk("\n");
348 	va_end(args);
349 }
350 
351 void ext4_update_dynamic_rev(struct super_block *sb)
352 {
353 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
354 
355 	if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
356 		return;
357 
358 	ext4_warning(sb, __FUNCTION__,
359 		     "updating to rev %d because of new feature flag, "
360 		     "running e2fsck is recommended",
361 		     EXT4_DYNAMIC_REV);
362 
363 	es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
364 	es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
365 	es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
366 	/* leave es->s_feature_*compat flags alone */
367 	/* es->s_uuid will be set by e2fsck if empty */
368 
369 	/*
370 	 * The rest of the superblock fields should be zero, and if not it
371 	 * means they are likely already in use, so leave them alone.  We
372 	 * can leave it up to e2fsck to clean up any inconsistencies there.
373 	 */
374 }
375 
376 /*
377  * Open the external journal device
378  */
379 static struct block_device *ext4_blkdev_get(dev_t dev)
380 {
381 	struct block_device *bdev;
382 	char b[BDEVNAME_SIZE];
383 
384 	bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
385 	if (IS_ERR(bdev))
386 		goto fail;
387 	return bdev;
388 
389 fail:
390 	printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
391 			__bdevname(dev, b), PTR_ERR(bdev));
392 	return NULL;
393 }
394 
395 /*
396  * Release the journal device
397  */
398 static int ext4_blkdev_put(struct block_device *bdev)
399 {
400 	bd_release(bdev);
401 	return blkdev_put(bdev);
402 }
403 
404 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
405 {
406 	struct block_device *bdev;
407 	int ret = -ENODEV;
408 
409 	bdev = sbi->journal_bdev;
410 	if (bdev) {
411 		ret = ext4_blkdev_put(bdev);
412 		sbi->journal_bdev = NULL;
413 	}
414 	return ret;
415 }
416 
417 static inline struct inode *orphan_list_entry(struct list_head *l)
418 {
419 	return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
420 }
421 
422 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
423 {
424 	struct list_head *l;
425 
426 	printk(KERN_ERR "sb orphan head is %d\n",
427 	       le32_to_cpu(sbi->s_es->s_last_orphan));
428 
429 	printk(KERN_ERR "sb_info orphan list:\n");
430 	list_for_each(l, &sbi->s_orphan) {
431 		struct inode *inode = orphan_list_entry(l);
432 		printk(KERN_ERR "  "
433 		       "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
434 		       inode->i_sb->s_id, inode->i_ino, inode,
435 		       inode->i_mode, inode->i_nlink,
436 		       NEXT_ORPHAN(inode));
437 	}
438 }
439 
440 static void ext4_put_super (struct super_block * sb)
441 {
442 	struct ext4_sb_info *sbi = EXT4_SB(sb);
443 	struct ext4_super_block *es = sbi->s_es;
444 	int i;
445 
446 	ext4_ext_release(sb);
447 	ext4_xattr_put_super(sb);
448 	jbd2_journal_destroy(sbi->s_journal);
449 	if (!(sb->s_flags & MS_RDONLY)) {
450 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
451 		es->s_state = cpu_to_le16(sbi->s_mount_state);
452 		BUFFER_TRACE(sbi->s_sbh, "marking dirty");
453 		mark_buffer_dirty(sbi->s_sbh);
454 		ext4_commit_super(sb, es, 1);
455 	}
456 
457 	for (i = 0; i < sbi->s_gdb_count; i++)
458 		brelse(sbi->s_group_desc[i]);
459 	kfree(sbi->s_group_desc);
460 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
461 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
462 	percpu_counter_destroy(&sbi->s_dirs_counter);
463 	brelse(sbi->s_sbh);
464 #ifdef CONFIG_QUOTA
465 	for (i = 0; i < MAXQUOTAS; i++)
466 		kfree(sbi->s_qf_names[i]);
467 #endif
468 
469 	/* Debugging code just in case the in-memory inode orphan list
470 	 * isn't empty.  The on-disk one can be non-empty if we've
471 	 * detected an error and taken the fs readonly, but the
472 	 * in-memory list had better be clean by this point. */
473 	if (!list_empty(&sbi->s_orphan))
474 		dump_orphan_list(sb, sbi);
475 	J_ASSERT(list_empty(&sbi->s_orphan));
476 
477 	invalidate_bdev(sb->s_bdev);
478 	if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
479 		/*
480 		 * Invalidate the journal device's buffers.  We don't want them
481 		 * floating about in memory - the physical journal device may
482 		 * hotswapped, and it breaks the `ro-after' testing code.
483 		 */
484 		sync_blockdev(sbi->journal_bdev);
485 		invalidate_bdev(sbi->journal_bdev);
486 		ext4_blkdev_remove(sbi);
487 	}
488 	sb->s_fs_info = NULL;
489 	kfree(sbi);
490 	return;
491 }
492 
493 static struct kmem_cache *ext4_inode_cachep;
494 
495 /*
496  * Called inside transaction, so use GFP_NOFS
497  */
498 static struct inode *ext4_alloc_inode(struct super_block *sb)
499 {
500 	struct ext4_inode_info *ei;
501 
502 	ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
503 	if (!ei)
504 		return NULL;
505 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
506 	ei->i_acl = EXT4_ACL_NOT_CACHED;
507 	ei->i_default_acl = EXT4_ACL_NOT_CACHED;
508 #endif
509 	ei->i_block_alloc_info = NULL;
510 	ei->vfs_inode.i_version = 1;
511 	memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
512 	return &ei->vfs_inode;
513 }
514 
515 static void ext4_destroy_inode(struct inode *inode)
516 {
517 	if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
518 		printk("EXT4 Inode %p: orphan list check failed!\n",
519 			EXT4_I(inode));
520 		print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
521 				EXT4_I(inode), sizeof(struct ext4_inode_info),
522 				true);
523 		dump_stack();
524 	}
525 	kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
526 }
527 
528 static void init_once(struct kmem_cache *cachep, void *foo)
529 {
530 	struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
531 
532 	INIT_LIST_HEAD(&ei->i_orphan);
533 #ifdef CONFIG_EXT4DEV_FS_XATTR
534 	init_rwsem(&ei->xattr_sem);
535 #endif
536 	mutex_init(&ei->truncate_mutex);
537 	inode_init_once(&ei->vfs_inode);
538 }
539 
540 static int init_inodecache(void)
541 {
542 	ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
543 					     sizeof(struct ext4_inode_info),
544 					     0, (SLAB_RECLAIM_ACCOUNT|
545 						SLAB_MEM_SPREAD),
546 					     init_once);
547 	if (ext4_inode_cachep == NULL)
548 		return -ENOMEM;
549 	return 0;
550 }
551 
552 static void destroy_inodecache(void)
553 {
554 	kmem_cache_destroy(ext4_inode_cachep);
555 }
556 
557 static void ext4_clear_inode(struct inode *inode)
558 {
559 	struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info;
560 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
561 	if (EXT4_I(inode)->i_acl &&
562 			EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
563 		posix_acl_release(EXT4_I(inode)->i_acl);
564 		EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
565 	}
566 	if (EXT4_I(inode)->i_default_acl &&
567 			EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
568 		posix_acl_release(EXT4_I(inode)->i_default_acl);
569 		EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
570 	}
571 #endif
572 	ext4_discard_reservation(inode);
573 	EXT4_I(inode)->i_block_alloc_info = NULL;
574 	if (unlikely(rsv))
575 		kfree(rsv);
576 }
577 
578 static inline void ext4_show_quota_options(struct seq_file *seq, struct super_block *sb)
579 {
580 #if defined(CONFIG_QUOTA)
581 	struct ext4_sb_info *sbi = EXT4_SB(sb);
582 
583 	if (sbi->s_jquota_fmt)
584 		seq_printf(seq, ",jqfmt=%s",
585 		(sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
586 
587 	if (sbi->s_qf_names[USRQUOTA])
588 		seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
589 
590 	if (sbi->s_qf_names[GRPQUOTA])
591 		seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
592 
593 	if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
594 		seq_puts(seq, ",usrquota");
595 
596 	if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
597 		seq_puts(seq, ",grpquota");
598 #endif
599 }
600 
601 /*
602  * Show an option if
603  *  - it's set to a non-default value OR
604  *  - if the per-sb default is different from the global default
605  */
606 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
607 {
608 	struct super_block *sb = vfs->mnt_sb;
609 	struct ext4_sb_info *sbi = EXT4_SB(sb);
610 	struct ext4_super_block *es = sbi->s_es;
611 	unsigned long def_mount_opts;
612 
613 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
614 
615 	if (sbi->s_sb_block != 1)
616 		seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
617 	if (test_opt(sb, MINIX_DF))
618 		seq_puts(seq, ",minixdf");
619 	if (test_opt(sb, GRPID))
620 		seq_puts(seq, ",grpid");
621 	if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
622 		seq_puts(seq, ",nogrpid");
623 	if (sbi->s_resuid != EXT4_DEF_RESUID ||
624 	    le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
625 		seq_printf(seq, ",resuid=%u", sbi->s_resuid);
626 	}
627 	if (sbi->s_resgid != EXT4_DEF_RESGID ||
628 	    le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
629 		seq_printf(seq, ",resgid=%u", sbi->s_resgid);
630 	}
631 	if (test_opt(sb, ERRORS_CONT)) {
632 		int def_errors = le16_to_cpu(es->s_errors);
633 
634 		if (def_errors == EXT4_ERRORS_PANIC ||
635 		    def_errors == EXT4_ERRORS_RO) {
636 			seq_puts(seq, ",errors=continue");
637 		}
638 	}
639 	if (test_opt(sb, ERRORS_RO))
640 		seq_puts(seq, ",errors=remount-ro");
641 	if (test_opt(sb, ERRORS_PANIC))
642 		seq_puts(seq, ",errors=panic");
643 	if (test_opt(sb, NO_UID32))
644 		seq_puts(seq, ",nouid32");
645 	if (test_opt(sb, DEBUG))
646 		seq_puts(seq, ",debug");
647 	if (test_opt(sb, OLDALLOC))
648 		seq_puts(seq, ",oldalloc");
649 #ifdef CONFIG_EXT4_FS_XATTR
650 	if (test_opt(sb, XATTR_USER))
651 		seq_puts(seq, ",user_xattr");
652 	if (!test_opt(sb, XATTR_USER) &&
653 	    (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
654 		seq_puts(seq, ",nouser_xattr");
655 	}
656 #endif
657 #ifdef CONFIG_EXT4_FS_POSIX_ACL
658 	if (test_opt(sb, POSIX_ACL))
659 		seq_puts(seq, ",acl");
660 	if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
661 		seq_puts(seq, ",noacl");
662 #endif
663 	if (!test_opt(sb, RESERVATION))
664 		seq_puts(seq, ",noreservation");
665 	if (sbi->s_commit_interval) {
666 		seq_printf(seq, ",commit=%u",
667 			   (unsigned) (sbi->s_commit_interval / HZ));
668 	}
669 	if (test_opt(sb, BARRIER))
670 		seq_puts(seq, ",barrier=1");
671 	if (test_opt(sb, NOBH))
672 		seq_puts(seq, ",nobh");
673 	if (!test_opt(sb, EXTENTS))
674 		seq_puts(seq, ",noextents");
675 
676 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
677 		seq_puts(seq, ",data=journal");
678 	else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
679 		seq_puts(seq, ",data=ordered");
680 	else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
681 		seq_puts(seq, ",data=writeback");
682 
683 	ext4_show_quota_options(seq, sb);
684 
685 	return 0;
686 }
687 
688 
689 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
690 		u64 ino, u32 generation)
691 {
692 	struct inode *inode;
693 
694 	if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
695 		return ERR_PTR(-ESTALE);
696 	if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
697 		return ERR_PTR(-ESTALE);
698 
699 	/* iget isn't really right if the inode is currently unallocated!!
700 	 *
701 	 * ext4_read_inode will return a bad_inode if the inode had been
702 	 * deleted, so we should be safe.
703 	 *
704 	 * Currently we don't know the generation for parent directory, so
705 	 * a generation of 0 means "accept any"
706 	 */
707 	inode = iget(sb, ino);
708 	if (inode == NULL)
709 		return ERR_PTR(-ENOMEM);
710 	if (is_bad_inode(inode) ||
711 	    (generation && inode->i_generation != generation)) {
712 		iput(inode);
713 		return ERR_PTR(-ESTALE);
714 	}
715 
716 	return inode;
717 }
718 
719 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
720 		int fh_len, int fh_type)
721 {
722 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
723 				    ext4_nfs_get_inode);
724 }
725 
726 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
727 		int fh_len, int fh_type)
728 {
729 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
730 				    ext4_nfs_get_inode);
731 }
732 
733 #ifdef CONFIG_QUOTA
734 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
735 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
736 
737 static int ext4_dquot_initialize(struct inode *inode, int type);
738 static int ext4_dquot_drop(struct inode *inode);
739 static int ext4_write_dquot(struct dquot *dquot);
740 static int ext4_acquire_dquot(struct dquot *dquot);
741 static int ext4_release_dquot(struct dquot *dquot);
742 static int ext4_mark_dquot_dirty(struct dquot *dquot);
743 static int ext4_write_info(struct super_block *sb, int type);
744 static int ext4_quota_on(struct super_block *sb, int type, int format_id, char *path);
745 static int ext4_quota_on_mount(struct super_block *sb, int type);
746 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
747 			       size_t len, loff_t off);
748 static ssize_t ext4_quota_write(struct super_block *sb, int type,
749 				const char *data, size_t len, loff_t off);
750 
751 static struct dquot_operations ext4_quota_operations = {
752 	.initialize	= ext4_dquot_initialize,
753 	.drop		= ext4_dquot_drop,
754 	.alloc_space	= dquot_alloc_space,
755 	.alloc_inode	= dquot_alloc_inode,
756 	.free_space	= dquot_free_space,
757 	.free_inode	= dquot_free_inode,
758 	.transfer	= dquot_transfer,
759 	.write_dquot	= ext4_write_dquot,
760 	.acquire_dquot	= ext4_acquire_dquot,
761 	.release_dquot	= ext4_release_dquot,
762 	.mark_dirty	= ext4_mark_dquot_dirty,
763 	.write_info	= ext4_write_info
764 };
765 
766 static struct quotactl_ops ext4_qctl_operations = {
767 	.quota_on	= ext4_quota_on,
768 	.quota_off	= vfs_quota_off,
769 	.quota_sync	= vfs_quota_sync,
770 	.get_info	= vfs_get_dqinfo,
771 	.set_info	= vfs_set_dqinfo,
772 	.get_dqblk	= vfs_get_dqblk,
773 	.set_dqblk	= vfs_set_dqblk
774 };
775 #endif
776 
777 static const struct super_operations ext4_sops = {
778 	.alloc_inode	= ext4_alloc_inode,
779 	.destroy_inode	= ext4_destroy_inode,
780 	.read_inode	= ext4_read_inode,
781 	.write_inode	= ext4_write_inode,
782 	.dirty_inode	= ext4_dirty_inode,
783 	.delete_inode	= ext4_delete_inode,
784 	.put_super	= ext4_put_super,
785 	.write_super	= ext4_write_super,
786 	.sync_fs	= ext4_sync_fs,
787 	.write_super_lockfs = ext4_write_super_lockfs,
788 	.unlockfs	= ext4_unlockfs,
789 	.statfs		= ext4_statfs,
790 	.remount_fs	= ext4_remount,
791 	.clear_inode	= ext4_clear_inode,
792 	.show_options	= ext4_show_options,
793 #ifdef CONFIG_QUOTA
794 	.quota_read	= ext4_quota_read,
795 	.quota_write	= ext4_quota_write,
796 #endif
797 };
798 
799 static const struct export_operations ext4_export_ops = {
800 	.fh_to_dentry = ext4_fh_to_dentry,
801 	.fh_to_parent = ext4_fh_to_parent,
802 	.get_parent = ext4_get_parent,
803 };
804 
805 enum {
806 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
807 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
808 	Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
809 	Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
810 	Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
811 	Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
812 	Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
813 	Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
814 	Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
815 	Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
816 	Opt_grpquota, Opt_extents, Opt_noextents,
817 };
818 
819 static match_table_t tokens = {
820 	{Opt_bsd_df, "bsddf"},
821 	{Opt_minix_df, "minixdf"},
822 	{Opt_grpid, "grpid"},
823 	{Opt_grpid, "bsdgroups"},
824 	{Opt_nogrpid, "nogrpid"},
825 	{Opt_nogrpid, "sysvgroups"},
826 	{Opt_resgid, "resgid=%u"},
827 	{Opt_resuid, "resuid=%u"},
828 	{Opt_sb, "sb=%u"},
829 	{Opt_err_cont, "errors=continue"},
830 	{Opt_err_panic, "errors=panic"},
831 	{Opt_err_ro, "errors=remount-ro"},
832 	{Opt_nouid32, "nouid32"},
833 	{Opt_nocheck, "nocheck"},
834 	{Opt_nocheck, "check=none"},
835 	{Opt_debug, "debug"},
836 	{Opt_oldalloc, "oldalloc"},
837 	{Opt_orlov, "orlov"},
838 	{Opt_user_xattr, "user_xattr"},
839 	{Opt_nouser_xattr, "nouser_xattr"},
840 	{Opt_acl, "acl"},
841 	{Opt_noacl, "noacl"},
842 	{Opt_reservation, "reservation"},
843 	{Opt_noreservation, "noreservation"},
844 	{Opt_noload, "noload"},
845 	{Opt_nobh, "nobh"},
846 	{Opt_bh, "bh"},
847 	{Opt_commit, "commit=%u"},
848 	{Opt_journal_update, "journal=update"},
849 	{Opt_journal_inum, "journal=%u"},
850 	{Opt_journal_dev, "journal_dev=%u"},
851 	{Opt_abort, "abort"},
852 	{Opt_data_journal, "data=journal"},
853 	{Opt_data_ordered, "data=ordered"},
854 	{Opt_data_writeback, "data=writeback"},
855 	{Opt_offusrjquota, "usrjquota="},
856 	{Opt_usrjquota, "usrjquota=%s"},
857 	{Opt_offgrpjquota, "grpjquota="},
858 	{Opt_grpjquota, "grpjquota=%s"},
859 	{Opt_jqfmt_vfsold, "jqfmt=vfsold"},
860 	{Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
861 	{Opt_grpquota, "grpquota"},
862 	{Opt_noquota, "noquota"},
863 	{Opt_quota, "quota"},
864 	{Opt_usrquota, "usrquota"},
865 	{Opt_barrier, "barrier=%u"},
866 	{Opt_extents, "extents"},
867 	{Opt_noextents, "noextents"},
868 	{Opt_err, NULL},
869 	{Opt_resize, "resize"},
870 };
871 
872 static ext4_fsblk_t get_sb_block(void **data)
873 {
874 	ext4_fsblk_t	sb_block;
875 	char		*options = (char *) *data;
876 
877 	if (!options || strncmp(options, "sb=", 3) != 0)
878 		return 1;	/* Default location */
879 	options += 3;
880 	/*todo: use simple_strtoll with >32bit ext4 */
881 	sb_block = simple_strtoul(options, &options, 0);
882 	if (*options && *options != ',') {
883 		printk("EXT4-fs: Invalid sb specification: %s\n",
884 		       (char *) *data);
885 		return 1;
886 	}
887 	if (*options == ',')
888 		options++;
889 	*data = (void *) options;
890 	return sb_block;
891 }
892 
893 static int parse_options (char *options, struct super_block *sb,
894 			  unsigned int *inum, unsigned long *journal_devnum,
895 			  ext4_fsblk_t *n_blocks_count, int is_remount)
896 {
897 	struct ext4_sb_info *sbi = EXT4_SB(sb);
898 	char * p;
899 	substring_t args[MAX_OPT_ARGS];
900 	int data_opt = 0;
901 	int option;
902 #ifdef CONFIG_QUOTA
903 	int qtype;
904 	char *qname;
905 #endif
906 
907 	if (!options)
908 		return 1;
909 
910 	while ((p = strsep (&options, ",")) != NULL) {
911 		int token;
912 		if (!*p)
913 			continue;
914 
915 		token = match_token(p, tokens, args);
916 		switch (token) {
917 		case Opt_bsd_df:
918 			clear_opt (sbi->s_mount_opt, MINIX_DF);
919 			break;
920 		case Opt_minix_df:
921 			set_opt (sbi->s_mount_opt, MINIX_DF);
922 			break;
923 		case Opt_grpid:
924 			set_opt (sbi->s_mount_opt, GRPID);
925 			break;
926 		case Opt_nogrpid:
927 			clear_opt (sbi->s_mount_opt, GRPID);
928 			break;
929 		case Opt_resuid:
930 			if (match_int(&args[0], &option))
931 				return 0;
932 			sbi->s_resuid = option;
933 			break;
934 		case Opt_resgid:
935 			if (match_int(&args[0], &option))
936 				return 0;
937 			sbi->s_resgid = option;
938 			break;
939 		case Opt_sb:
940 			/* handled by get_sb_block() instead of here */
941 			/* *sb_block = match_int(&args[0]); */
942 			break;
943 		case Opt_err_panic:
944 			clear_opt (sbi->s_mount_opt, ERRORS_CONT);
945 			clear_opt (sbi->s_mount_opt, ERRORS_RO);
946 			set_opt (sbi->s_mount_opt, ERRORS_PANIC);
947 			break;
948 		case Opt_err_ro:
949 			clear_opt (sbi->s_mount_opt, ERRORS_CONT);
950 			clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
951 			set_opt (sbi->s_mount_opt, ERRORS_RO);
952 			break;
953 		case Opt_err_cont:
954 			clear_opt (sbi->s_mount_opt, ERRORS_RO);
955 			clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
956 			set_opt (sbi->s_mount_opt, ERRORS_CONT);
957 			break;
958 		case Opt_nouid32:
959 			set_opt (sbi->s_mount_opt, NO_UID32);
960 			break;
961 		case Opt_nocheck:
962 			clear_opt (sbi->s_mount_opt, CHECK);
963 			break;
964 		case Opt_debug:
965 			set_opt (sbi->s_mount_opt, DEBUG);
966 			break;
967 		case Opt_oldalloc:
968 			set_opt (sbi->s_mount_opt, OLDALLOC);
969 			break;
970 		case Opt_orlov:
971 			clear_opt (sbi->s_mount_opt, OLDALLOC);
972 			break;
973 #ifdef CONFIG_EXT4DEV_FS_XATTR
974 		case Opt_user_xattr:
975 			set_opt (sbi->s_mount_opt, XATTR_USER);
976 			break;
977 		case Opt_nouser_xattr:
978 			clear_opt (sbi->s_mount_opt, XATTR_USER);
979 			break;
980 #else
981 		case Opt_user_xattr:
982 		case Opt_nouser_xattr:
983 			printk("EXT4 (no)user_xattr options not supported\n");
984 			break;
985 #endif
986 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
987 		case Opt_acl:
988 			set_opt(sbi->s_mount_opt, POSIX_ACL);
989 			break;
990 		case Opt_noacl:
991 			clear_opt(sbi->s_mount_opt, POSIX_ACL);
992 			break;
993 #else
994 		case Opt_acl:
995 		case Opt_noacl:
996 			printk("EXT4 (no)acl options not supported\n");
997 			break;
998 #endif
999 		case Opt_reservation:
1000 			set_opt(sbi->s_mount_opt, RESERVATION);
1001 			break;
1002 		case Opt_noreservation:
1003 			clear_opt(sbi->s_mount_opt, RESERVATION);
1004 			break;
1005 		case Opt_journal_update:
1006 			/* @@@ FIXME */
1007 			/* Eventually we will want to be able to create
1008 			   a journal file here.  For now, only allow the
1009 			   user to specify an existing inode to be the
1010 			   journal file. */
1011 			if (is_remount) {
1012 				printk(KERN_ERR "EXT4-fs: cannot specify "
1013 				       "journal on remount\n");
1014 				return 0;
1015 			}
1016 			set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1017 			break;
1018 		case Opt_journal_inum:
1019 			if (is_remount) {
1020 				printk(KERN_ERR "EXT4-fs: cannot specify "
1021 				       "journal on remount\n");
1022 				return 0;
1023 			}
1024 			if (match_int(&args[0], &option))
1025 				return 0;
1026 			*inum = option;
1027 			break;
1028 		case Opt_journal_dev:
1029 			if (is_remount) {
1030 				printk(KERN_ERR "EXT4-fs: cannot specify "
1031 				       "journal on remount\n");
1032 				return 0;
1033 			}
1034 			if (match_int(&args[0], &option))
1035 				return 0;
1036 			*journal_devnum = option;
1037 			break;
1038 		case Opt_noload:
1039 			set_opt (sbi->s_mount_opt, NOLOAD);
1040 			break;
1041 		case Opt_commit:
1042 			if (match_int(&args[0], &option))
1043 				return 0;
1044 			if (option < 0)
1045 				return 0;
1046 			if (option == 0)
1047 				option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1048 			sbi->s_commit_interval = HZ * option;
1049 			break;
1050 		case Opt_data_journal:
1051 			data_opt = EXT4_MOUNT_JOURNAL_DATA;
1052 			goto datacheck;
1053 		case Opt_data_ordered:
1054 			data_opt = EXT4_MOUNT_ORDERED_DATA;
1055 			goto datacheck;
1056 		case Opt_data_writeback:
1057 			data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1058 		datacheck:
1059 			if (is_remount) {
1060 				if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1061 						!= data_opt) {
1062 					printk(KERN_ERR
1063 						"EXT4-fs: cannot change data "
1064 						"mode on remount\n");
1065 					return 0;
1066 				}
1067 			} else {
1068 				sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1069 				sbi->s_mount_opt |= data_opt;
1070 			}
1071 			break;
1072 #ifdef CONFIG_QUOTA
1073 		case Opt_usrjquota:
1074 			qtype = USRQUOTA;
1075 			goto set_qf_name;
1076 		case Opt_grpjquota:
1077 			qtype = GRPQUOTA;
1078 set_qf_name:
1079 			if (sb_any_quota_enabled(sb)) {
1080 				printk(KERN_ERR
1081 					"EXT4-fs: Cannot change journalled "
1082 					"quota options when quota turned on.\n");
1083 				return 0;
1084 			}
1085 			qname = match_strdup(&args[0]);
1086 			if (!qname) {
1087 				printk(KERN_ERR
1088 					"EXT4-fs: not enough memory for "
1089 					"storing quotafile name.\n");
1090 				return 0;
1091 			}
1092 			if (sbi->s_qf_names[qtype] &&
1093 			    strcmp(sbi->s_qf_names[qtype], qname)) {
1094 				printk(KERN_ERR
1095 					"EXT4-fs: %s quota file already "
1096 					"specified.\n", QTYPE2NAME(qtype));
1097 				kfree(qname);
1098 				return 0;
1099 			}
1100 			sbi->s_qf_names[qtype] = qname;
1101 			if (strchr(sbi->s_qf_names[qtype], '/')) {
1102 				printk(KERN_ERR
1103 					"EXT4-fs: quotafile must be on "
1104 					"filesystem root.\n");
1105 				kfree(sbi->s_qf_names[qtype]);
1106 				sbi->s_qf_names[qtype] = NULL;
1107 				return 0;
1108 			}
1109 			set_opt(sbi->s_mount_opt, QUOTA);
1110 			break;
1111 		case Opt_offusrjquota:
1112 			qtype = USRQUOTA;
1113 			goto clear_qf_name;
1114 		case Opt_offgrpjquota:
1115 			qtype = GRPQUOTA;
1116 clear_qf_name:
1117 			if (sb_any_quota_enabled(sb)) {
1118 				printk(KERN_ERR "EXT4-fs: Cannot change "
1119 					"journalled quota options when "
1120 					"quota turned on.\n");
1121 				return 0;
1122 			}
1123 			/*
1124 			 * The space will be released later when all options
1125 			 * are confirmed to be correct
1126 			 */
1127 			sbi->s_qf_names[qtype] = NULL;
1128 			break;
1129 		case Opt_jqfmt_vfsold:
1130 			sbi->s_jquota_fmt = QFMT_VFS_OLD;
1131 			break;
1132 		case Opt_jqfmt_vfsv0:
1133 			sbi->s_jquota_fmt = QFMT_VFS_V0;
1134 			break;
1135 		case Opt_quota:
1136 		case Opt_usrquota:
1137 			set_opt(sbi->s_mount_opt, QUOTA);
1138 			set_opt(sbi->s_mount_opt, USRQUOTA);
1139 			break;
1140 		case Opt_grpquota:
1141 			set_opt(sbi->s_mount_opt, QUOTA);
1142 			set_opt(sbi->s_mount_opt, GRPQUOTA);
1143 			break;
1144 		case Opt_noquota:
1145 			if (sb_any_quota_enabled(sb)) {
1146 				printk(KERN_ERR "EXT4-fs: Cannot change quota "
1147 					"options when quota turned on.\n");
1148 				return 0;
1149 			}
1150 			clear_opt(sbi->s_mount_opt, QUOTA);
1151 			clear_opt(sbi->s_mount_opt, USRQUOTA);
1152 			clear_opt(sbi->s_mount_opt, GRPQUOTA);
1153 			break;
1154 #else
1155 		case Opt_quota:
1156 		case Opt_usrquota:
1157 		case Opt_grpquota:
1158 		case Opt_usrjquota:
1159 		case Opt_grpjquota:
1160 		case Opt_offusrjquota:
1161 		case Opt_offgrpjquota:
1162 		case Opt_jqfmt_vfsold:
1163 		case Opt_jqfmt_vfsv0:
1164 			printk(KERN_ERR
1165 				"EXT4-fs: journalled quota options not "
1166 				"supported.\n");
1167 			break;
1168 		case Opt_noquota:
1169 			break;
1170 #endif
1171 		case Opt_abort:
1172 			set_opt(sbi->s_mount_opt, ABORT);
1173 			break;
1174 		case Opt_barrier:
1175 			if (match_int(&args[0], &option))
1176 				return 0;
1177 			if (option)
1178 				set_opt(sbi->s_mount_opt, BARRIER);
1179 			else
1180 				clear_opt(sbi->s_mount_opt, BARRIER);
1181 			break;
1182 		case Opt_ignore:
1183 			break;
1184 		case Opt_resize:
1185 			if (!is_remount) {
1186 				printk("EXT4-fs: resize option only available "
1187 					"for remount\n");
1188 				return 0;
1189 			}
1190 			if (match_int(&args[0], &option) != 0)
1191 				return 0;
1192 			*n_blocks_count = option;
1193 			break;
1194 		case Opt_nobh:
1195 			set_opt(sbi->s_mount_opt, NOBH);
1196 			break;
1197 		case Opt_bh:
1198 			clear_opt(sbi->s_mount_opt, NOBH);
1199 			break;
1200 		case Opt_extents:
1201 			set_opt (sbi->s_mount_opt, EXTENTS);
1202 			break;
1203 		case Opt_noextents:
1204 			clear_opt (sbi->s_mount_opt, EXTENTS);
1205 			break;
1206 		default:
1207 			printk (KERN_ERR
1208 				"EXT4-fs: Unrecognized mount option \"%s\" "
1209 				"or missing value\n", p);
1210 			return 0;
1211 		}
1212 	}
1213 #ifdef CONFIG_QUOTA
1214 	if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1215 		if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1216 		     sbi->s_qf_names[USRQUOTA])
1217 			clear_opt(sbi->s_mount_opt, USRQUOTA);
1218 
1219 		if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1220 		     sbi->s_qf_names[GRPQUOTA])
1221 			clear_opt(sbi->s_mount_opt, GRPQUOTA);
1222 
1223 		if ((sbi->s_qf_names[USRQUOTA] &&
1224 				(sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1225 		    (sbi->s_qf_names[GRPQUOTA] &&
1226 				(sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1227 			printk(KERN_ERR "EXT4-fs: old and new quota "
1228 					"format mixing.\n");
1229 			return 0;
1230 		}
1231 
1232 		if (!sbi->s_jquota_fmt) {
1233 			printk(KERN_ERR "EXT4-fs: journalled quota format "
1234 					"not specified.\n");
1235 			return 0;
1236 		}
1237 	} else {
1238 		if (sbi->s_jquota_fmt) {
1239 			printk(KERN_ERR "EXT4-fs: journalled quota format "
1240 					"specified with no journalling "
1241 					"enabled.\n");
1242 			return 0;
1243 		}
1244 	}
1245 #endif
1246 	return 1;
1247 }
1248 
1249 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1250 			    int read_only)
1251 {
1252 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1253 	int res = 0;
1254 
1255 	if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1256 		printk (KERN_ERR "EXT4-fs warning: revision level too high, "
1257 			"forcing read-only mode\n");
1258 		res = MS_RDONLY;
1259 	}
1260 	if (read_only)
1261 		return res;
1262 	if (!(sbi->s_mount_state & EXT4_VALID_FS))
1263 		printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1264 			"running e2fsck is recommended\n");
1265 	else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1266 		printk (KERN_WARNING
1267 			"EXT4-fs warning: mounting fs with errors, "
1268 			"running e2fsck is recommended\n");
1269 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1270 		 le16_to_cpu(es->s_mnt_count) >=
1271 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1272 		printk (KERN_WARNING
1273 			"EXT4-fs warning: maximal mount count reached, "
1274 			"running e2fsck is recommended\n");
1275 	else if (le32_to_cpu(es->s_checkinterval) &&
1276 		(le32_to_cpu(es->s_lastcheck) +
1277 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1278 		printk (KERN_WARNING
1279 			"EXT4-fs warning: checktime reached, "
1280 			"running e2fsck is recommended\n");
1281 #if 0
1282 		/* @@@ We _will_ want to clear the valid bit if we find
1283 		 * inconsistencies, to force a fsck at reboot.  But for
1284 		 * a plain journaled filesystem we can keep it set as
1285 		 * valid forever! :)
1286 		 */
1287 	es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT4_VALID_FS);
1288 #endif
1289 	if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1290 		es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1291 	es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
1292 	es->s_mtime = cpu_to_le32(get_seconds());
1293 	ext4_update_dynamic_rev(sb);
1294 	EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1295 
1296 	ext4_commit_super(sb, es, 1);
1297 	if (test_opt(sb, DEBUG))
1298 		printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
1299 				"bpg=%lu, ipg=%lu, mo=%04lx]\n",
1300 			sb->s_blocksize,
1301 			sbi->s_groups_count,
1302 			EXT4_BLOCKS_PER_GROUP(sb),
1303 			EXT4_INODES_PER_GROUP(sb),
1304 			sbi->s_mount_opt);
1305 
1306 	printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1307 	if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
1308 		char b[BDEVNAME_SIZE];
1309 
1310 		printk("external journal on %s\n",
1311 			bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
1312 	} else {
1313 		printk("internal journal\n");
1314 	}
1315 	return res;
1316 }
1317 
1318 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1319 			    struct ext4_group_desc *gdp)
1320 {
1321 	__u16 crc = 0;
1322 
1323 	if (sbi->s_es->s_feature_ro_compat &
1324 	    cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1325 		int offset = offsetof(struct ext4_group_desc, bg_checksum);
1326 		__le32 le_group = cpu_to_le32(block_group);
1327 
1328 		crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1329 		crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1330 		crc = crc16(crc, (__u8 *)gdp, offset);
1331 		offset += sizeof(gdp->bg_checksum); /* skip checksum */
1332 		/* for checksum of struct ext4_group_desc do the rest...*/
1333 		if ((sbi->s_es->s_feature_incompat &
1334 		     cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1335 		    offset < le16_to_cpu(sbi->s_es->s_desc_size))
1336 			crc = crc16(crc, (__u8 *)gdp + offset,
1337 				    le16_to_cpu(sbi->s_es->s_desc_size) -
1338 					offset);
1339 	}
1340 
1341 	return cpu_to_le16(crc);
1342 }
1343 
1344 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1345 				struct ext4_group_desc *gdp)
1346 {
1347 	if ((sbi->s_es->s_feature_ro_compat &
1348 	     cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1349 	    (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1350 		return 0;
1351 
1352 	return 1;
1353 }
1354 
1355 /* Called at mount-time, super-block is locked */
1356 static int ext4_check_descriptors (struct super_block * sb)
1357 {
1358 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1359 	ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1360 	ext4_fsblk_t last_block;
1361 	ext4_fsblk_t block_bitmap;
1362 	ext4_fsblk_t inode_bitmap;
1363 	ext4_fsblk_t inode_table;
1364 	struct ext4_group_desc * gdp = NULL;
1365 	int desc_block = 0;
1366 	int flexbg_flag = 0;
1367 	int i;
1368 
1369 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1370 		flexbg_flag = 1;
1371 
1372 	ext4_debug ("Checking group descriptors");
1373 
1374 	for (i = 0; i < sbi->s_groups_count; i++)
1375 	{
1376 		if (i == sbi->s_groups_count - 1 || flexbg_flag)
1377 			last_block = ext4_blocks_count(sbi->s_es) - 1;
1378 		else
1379 			last_block = first_block +
1380 				(EXT4_BLOCKS_PER_GROUP(sb) - 1);
1381 
1382 		if ((i % EXT4_DESC_PER_BLOCK(sb)) == 0)
1383 			gdp = (struct ext4_group_desc *)
1384 					sbi->s_group_desc[desc_block++]->b_data;
1385 		block_bitmap = ext4_block_bitmap(sb, gdp);
1386 		if (block_bitmap < first_block || block_bitmap > last_block)
1387 		{
1388 			ext4_error (sb, "ext4_check_descriptors",
1389 				    "Block bitmap for group %d"
1390 				    " not in group (block %llu)!",
1391 				    i, block_bitmap);
1392 			return 0;
1393 		}
1394 		inode_bitmap = ext4_inode_bitmap(sb, gdp);
1395 		if (inode_bitmap < first_block || inode_bitmap > last_block)
1396 		{
1397 			ext4_error (sb, "ext4_check_descriptors",
1398 				    "Inode bitmap for group %d"
1399 				    " not in group (block %llu)!",
1400 				    i, inode_bitmap);
1401 			return 0;
1402 		}
1403 		inode_table = ext4_inode_table(sb, gdp);
1404 		if (inode_table < first_block ||
1405 		    inode_table + sbi->s_itb_per_group - 1 > last_block)
1406 		{
1407 			ext4_error (sb, "ext4_check_descriptors",
1408 				    "Inode table for group %d"
1409 				    " not in group (block %llu)!",
1410 				    i, inode_table);
1411 			return 0;
1412 		}
1413 		if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1414 			ext4_error(sb, __FUNCTION__,
1415 				   "Checksum for group %d failed (%u!=%u)\n", i,
1416 				   le16_to_cpu(ext4_group_desc_csum(sbi, i,
1417 								    gdp)),
1418 				   le16_to_cpu(gdp->bg_checksum));
1419 			return 0;
1420 		}
1421 		if (!flexbg_flag)
1422 			first_block += EXT4_BLOCKS_PER_GROUP(sb);
1423 		gdp = (struct ext4_group_desc *)
1424 			((__u8 *)gdp + EXT4_DESC_SIZE(sb));
1425 	}
1426 
1427 	ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1428 	sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb));
1429 	return 1;
1430 }
1431 
1432 
1433 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1434  * the superblock) which were deleted from all directories, but held open by
1435  * a process at the time of a crash.  We walk the list and try to delete these
1436  * inodes at recovery time (only with a read-write filesystem).
1437  *
1438  * In order to keep the orphan inode chain consistent during traversal (in
1439  * case of crash during recovery), we link each inode into the superblock
1440  * orphan list_head and handle it the same way as an inode deletion during
1441  * normal operation (which journals the operations for us).
1442  *
1443  * We only do an iget() and an iput() on each inode, which is very safe if we
1444  * accidentally point at an in-use or already deleted inode.  The worst that
1445  * can happen in this case is that we get a "bit already cleared" message from
1446  * ext4_free_inode().  The only reason we would point at a wrong inode is if
1447  * e2fsck was run on this filesystem, and it must have already done the orphan
1448  * inode cleanup for us, so we can safely abort without any further action.
1449  */
1450 static void ext4_orphan_cleanup (struct super_block * sb,
1451 				 struct ext4_super_block * es)
1452 {
1453 	unsigned int s_flags = sb->s_flags;
1454 	int nr_orphans = 0, nr_truncates = 0;
1455 #ifdef CONFIG_QUOTA
1456 	int i;
1457 #endif
1458 	if (!es->s_last_orphan) {
1459 		jbd_debug(4, "no orphan inodes to clean up\n");
1460 		return;
1461 	}
1462 
1463 	if (bdev_read_only(sb->s_bdev)) {
1464 		printk(KERN_ERR "EXT4-fs: write access "
1465 			"unavailable, skipping orphan cleanup.\n");
1466 		return;
1467 	}
1468 
1469 	if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1470 		if (es->s_last_orphan)
1471 			jbd_debug(1, "Errors on filesystem, "
1472 				  "clearing orphan list.\n");
1473 		es->s_last_orphan = 0;
1474 		jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1475 		return;
1476 	}
1477 
1478 	if (s_flags & MS_RDONLY) {
1479 		printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1480 		       sb->s_id);
1481 		sb->s_flags &= ~MS_RDONLY;
1482 	}
1483 #ifdef CONFIG_QUOTA
1484 	/* Needed for iput() to work correctly and not trash data */
1485 	sb->s_flags |= MS_ACTIVE;
1486 	/* Turn on quotas so that they are updated correctly */
1487 	for (i = 0; i < MAXQUOTAS; i++) {
1488 		if (EXT4_SB(sb)->s_qf_names[i]) {
1489 			int ret = ext4_quota_on_mount(sb, i);
1490 			if (ret < 0)
1491 				printk(KERN_ERR
1492 					"EXT4-fs: Cannot turn on journalled "
1493 					"quota: error %d\n", ret);
1494 		}
1495 	}
1496 #endif
1497 
1498 	while (es->s_last_orphan) {
1499 		struct inode *inode;
1500 
1501 		if (!(inode =
1502 		      ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
1503 			es->s_last_orphan = 0;
1504 			break;
1505 		}
1506 
1507 		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1508 		DQUOT_INIT(inode);
1509 		if (inode->i_nlink) {
1510 			printk(KERN_DEBUG
1511 				"%s: truncating inode %lu to %Ld bytes\n",
1512 				__FUNCTION__, inode->i_ino, inode->i_size);
1513 			jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1514 				  inode->i_ino, inode->i_size);
1515 			ext4_truncate(inode);
1516 			nr_truncates++;
1517 		} else {
1518 			printk(KERN_DEBUG
1519 				"%s: deleting unreferenced inode %lu\n",
1520 				__FUNCTION__, inode->i_ino);
1521 			jbd_debug(2, "deleting unreferenced inode %lu\n",
1522 				  inode->i_ino);
1523 			nr_orphans++;
1524 		}
1525 		iput(inode);  /* The delete magic happens here! */
1526 	}
1527 
1528 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1529 
1530 	if (nr_orphans)
1531 		printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1532 		       sb->s_id, PLURAL(nr_orphans));
1533 	if (nr_truncates)
1534 		printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1535 		       sb->s_id, PLURAL(nr_truncates));
1536 #ifdef CONFIG_QUOTA
1537 	/* Turn quotas off */
1538 	for (i = 0; i < MAXQUOTAS; i++) {
1539 		if (sb_dqopt(sb)->files[i])
1540 			vfs_quota_off(sb, i);
1541 	}
1542 #endif
1543 	sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1544 }
1545 
1546 /*
1547  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
1548  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1549  * We need to be 1 filesystem block less than the 2^32 sector limit.
1550  */
1551 static loff_t ext4_max_size(int bits)
1552 {
1553 	loff_t res = EXT4_NDIR_BLOCKS;
1554 	/* This constant is calculated to be the largest file size for a
1555 	 * dense, 4k-blocksize file such that the total number of
1556 	 * sectors in the file, including data and all indirect blocks,
1557 	 * does not exceed 2^32. */
1558 	const loff_t upper_limit = 0x1ff7fffd000LL;
1559 
1560 	res += 1LL << (bits-2);
1561 	res += 1LL << (2*(bits-2));
1562 	res += 1LL << (3*(bits-2));
1563 	res <<= bits;
1564 	if (res > upper_limit)
1565 		res = upper_limit;
1566 	return res;
1567 }
1568 
1569 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1570 				ext4_fsblk_t logical_sb_block, int nr)
1571 {
1572 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1573 	unsigned long bg, first_meta_bg;
1574 	int has_super = 0;
1575 
1576 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1577 
1578 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1579 	    nr < first_meta_bg)
1580 		return logical_sb_block + nr + 1;
1581 	bg = sbi->s_desc_per_block * nr;
1582 	if (ext4_bg_has_super(sb, bg))
1583 		has_super = 1;
1584 	return (has_super + ext4_group_first_block_no(sb, bg));
1585 }
1586 
1587 
1588 static int ext4_fill_super (struct super_block *sb, void *data, int silent)
1589 {
1590 	struct buffer_head * bh;
1591 	struct ext4_super_block *es = NULL;
1592 	struct ext4_sb_info *sbi;
1593 	ext4_fsblk_t block;
1594 	ext4_fsblk_t sb_block = get_sb_block(&data);
1595 	ext4_fsblk_t logical_sb_block;
1596 	unsigned long offset = 0;
1597 	unsigned int journal_inum = 0;
1598 	unsigned long journal_devnum = 0;
1599 	unsigned long def_mount_opts;
1600 	struct inode *root;
1601 	int blocksize;
1602 	int hblock;
1603 	int db_count;
1604 	int i;
1605 	int needs_recovery;
1606 	__le32 features;
1607 	__u64 blocks_count;
1608 	int err;
1609 
1610 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1611 	if (!sbi)
1612 		return -ENOMEM;
1613 	sb->s_fs_info = sbi;
1614 	sbi->s_mount_opt = 0;
1615 	sbi->s_resuid = EXT4_DEF_RESUID;
1616 	sbi->s_resgid = EXT4_DEF_RESGID;
1617 	sbi->s_sb_block = sb_block;
1618 
1619 	unlock_kernel();
1620 
1621 	blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
1622 	if (!blocksize) {
1623 		printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
1624 		goto out_fail;
1625 	}
1626 
1627 	if (!sb_set_blocksize(sb, blocksize)) {
1628 		printk(KERN_ERR "EXT4-fs: bad blocksize %d.\n", blocksize);
1629 		goto out_fail;
1630 	}
1631 
1632 	/*
1633 	 * The ext4 superblock will not be buffer aligned for other than 1kB
1634 	 * block sizes.  We need to calculate the offset from buffer start.
1635 	 */
1636 	if (blocksize != EXT4_MIN_BLOCK_SIZE) {
1637 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1638 		offset = do_div(logical_sb_block, blocksize);
1639 	} else {
1640 		logical_sb_block = sb_block;
1641 	}
1642 
1643 	if (!(bh = sb_bread(sb, logical_sb_block))) {
1644 		printk (KERN_ERR "EXT4-fs: unable to read superblock\n");
1645 		goto out_fail;
1646 	}
1647 	/*
1648 	 * Note: s_es must be initialized as soon as possible because
1649 	 *       some ext4 macro-instructions depend on its value
1650 	 */
1651 	es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
1652 	sbi->s_es = es;
1653 	sb->s_magic = le16_to_cpu(es->s_magic);
1654 	if (sb->s_magic != EXT4_SUPER_MAGIC)
1655 		goto cantfind_ext4;
1656 
1657 	/* Set defaults before we parse the mount options */
1658 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1659 	if (def_mount_opts & EXT4_DEFM_DEBUG)
1660 		set_opt(sbi->s_mount_opt, DEBUG);
1661 	if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
1662 		set_opt(sbi->s_mount_opt, GRPID);
1663 	if (def_mount_opts & EXT4_DEFM_UID16)
1664 		set_opt(sbi->s_mount_opt, NO_UID32);
1665 #ifdef CONFIG_EXT4DEV_FS_XATTR
1666 	if (def_mount_opts & EXT4_DEFM_XATTR_USER)
1667 		set_opt(sbi->s_mount_opt, XATTR_USER);
1668 #endif
1669 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1670 	if (def_mount_opts & EXT4_DEFM_ACL)
1671 		set_opt(sbi->s_mount_opt, POSIX_ACL);
1672 #endif
1673 	if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1674 		sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1675 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1676 		sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1677 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1678 		sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1679 
1680 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
1681 		set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1682 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_RO)
1683 		set_opt(sbi->s_mount_opt, ERRORS_RO);
1684 	else
1685 		set_opt(sbi->s_mount_opt, ERRORS_CONT);
1686 
1687 	sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1688 	sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1689 
1690 	set_opt(sbi->s_mount_opt, RESERVATION);
1691 
1692 	/*
1693 	 * turn on extents feature by default in ext4 filesystem
1694 	 * User -o noextents to turn it off
1695 	 */
1696 	set_opt(sbi->s_mount_opt, EXTENTS);
1697 
1698 	if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1699 			    NULL, 0))
1700 		goto failed_mount;
1701 
1702 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1703 		((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1704 
1705 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1706 	    (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1707 	     EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1708 	     EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1709 		printk(KERN_WARNING
1710 		       "EXT4-fs warning: feature flags set on rev 0 fs, "
1711 		       "running e2fsck is recommended\n");
1712 	/*
1713 	 * Check feature flags regardless of the revision level, since we
1714 	 * previously didn't change the revision level when setting the flags,
1715 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
1716 	 */
1717 	features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
1718 	if (features) {
1719 		printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
1720 		       "unsupported optional features (%x).\n",
1721 		       sb->s_id, le32_to_cpu(features));
1722 		goto failed_mount;
1723 	}
1724 	features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
1725 	if (!(sb->s_flags & MS_RDONLY) && features) {
1726 		printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
1727 		       "unsupported optional features (%x).\n",
1728 		       sb->s_id, le32_to_cpu(features));
1729 		goto failed_mount;
1730 	}
1731 	blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1732 
1733 	if (blocksize < EXT4_MIN_BLOCK_SIZE ||
1734 	    blocksize > EXT4_MAX_BLOCK_SIZE) {
1735 		printk(KERN_ERR
1736 		       "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
1737 		       blocksize, sb->s_id);
1738 		goto failed_mount;
1739 	}
1740 
1741 	hblock = bdev_hardsect_size(sb->s_bdev);
1742 	if (sb->s_blocksize != blocksize) {
1743 		/*
1744 		 * Make sure the blocksize for the filesystem is larger
1745 		 * than the hardware sectorsize for the machine.
1746 		 */
1747 		if (blocksize < hblock) {
1748 			printk(KERN_ERR "EXT4-fs: blocksize %d too small for "
1749 			       "device blocksize %d.\n", blocksize, hblock);
1750 			goto failed_mount;
1751 		}
1752 
1753 		brelse (bh);
1754 		sb_set_blocksize(sb, blocksize);
1755 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1756 		offset = do_div(logical_sb_block, blocksize);
1757 		bh = sb_bread(sb, logical_sb_block);
1758 		if (!bh) {
1759 			printk(KERN_ERR
1760 			       "EXT4-fs: Can't read superblock on 2nd try.\n");
1761 			goto failed_mount;
1762 		}
1763 		es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
1764 		sbi->s_es = es;
1765 		if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
1766 			printk (KERN_ERR
1767 				"EXT4-fs: Magic mismatch, very weird !\n");
1768 			goto failed_mount;
1769 		}
1770 	}
1771 
1772 	sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
1773 
1774 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
1775 		sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
1776 		sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
1777 	} else {
1778 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1779 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1780 		if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1781 		    (!is_power_of_2(sbi->s_inode_size)) ||
1782 		    (sbi->s_inode_size > blocksize)) {
1783 			printk (KERN_ERR
1784 				"EXT4-fs: unsupported inode size: %d\n",
1785 				sbi->s_inode_size);
1786 			goto failed_mount;
1787 		}
1788 		if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
1789 			sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
1790 	}
1791 	sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
1792 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
1793 		if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
1794 		    sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
1795 		    !is_power_of_2(sbi->s_desc_size)) {
1796 			printk(KERN_ERR
1797 			       "EXT4-fs: unsupported descriptor size %lu\n",
1798 			       sbi->s_desc_size);
1799 			goto failed_mount;
1800 		}
1801 	} else
1802 		sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
1803 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1804 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1805 	if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
1806 		goto cantfind_ext4;
1807 	sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
1808 	if (sbi->s_inodes_per_block == 0)
1809 		goto cantfind_ext4;
1810 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
1811 					sbi->s_inodes_per_block;
1812 	sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
1813 	sbi->s_sbh = bh;
1814 	sbi->s_mount_state = le16_to_cpu(es->s_state);
1815 	sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
1816 	sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
1817 	for (i=0; i < 4; i++)
1818 		sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1819 	sbi->s_def_hash_version = es->s_def_hash_version;
1820 
1821 	if (sbi->s_blocks_per_group > blocksize * 8) {
1822 		printk (KERN_ERR
1823 			"EXT4-fs: #blocks per group too big: %lu\n",
1824 			sbi->s_blocks_per_group);
1825 		goto failed_mount;
1826 	}
1827 	if (sbi->s_inodes_per_group > blocksize * 8) {
1828 		printk (KERN_ERR
1829 			"EXT4-fs: #inodes per group too big: %lu\n",
1830 			sbi->s_inodes_per_group);
1831 		goto failed_mount;
1832 	}
1833 
1834 	if (ext4_blocks_count(es) >
1835 		    (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1836 		printk(KERN_ERR "EXT4-fs: filesystem on %s:"
1837 			" too large to mount safely\n", sb->s_id);
1838 		if (sizeof(sector_t) < 8)
1839 			printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
1840 					"enabled\n");
1841 		goto failed_mount;
1842 	}
1843 
1844 	if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
1845 		goto cantfind_ext4;
1846 	blocks_count = (ext4_blocks_count(es) -
1847 			le32_to_cpu(es->s_first_data_block) +
1848 			EXT4_BLOCKS_PER_GROUP(sb) - 1);
1849 	do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
1850 	sbi->s_groups_count = blocks_count;
1851 	db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
1852 		   EXT4_DESC_PER_BLOCK(sb);
1853 	sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1854 				    GFP_KERNEL);
1855 	if (sbi->s_group_desc == NULL) {
1856 		printk (KERN_ERR "EXT4-fs: not enough memory\n");
1857 		goto failed_mount;
1858 	}
1859 
1860 	bgl_lock_init(&sbi->s_blockgroup_lock);
1861 
1862 	for (i = 0; i < db_count; i++) {
1863 		block = descriptor_loc(sb, logical_sb_block, i);
1864 		sbi->s_group_desc[i] = sb_bread(sb, block);
1865 		if (!sbi->s_group_desc[i]) {
1866 			printk (KERN_ERR "EXT4-fs: "
1867 				"can't read group descriptor %d\n", i);
1868 			db_count = i;
1869 			goto failed_mount2;
1870 		}
1871 	}
1872 	if (!ext4_check_descriptors (sb)) {
1873 		printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
1874 		goto failed_mount2;
1875 	}
1876 	sbi->s_gdb_count = db_count;
1877 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1878 	spin_lock_init(&sbi->s_next_gen_lock);
1879 
1880 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
1881 			ext4_count_free_blocks(sb));
1882 	if (!err) {
1883 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
1884 				ext4_count_free_inodes(sb));
1885 	}
1886 	if (!err) {
1887 		err = percpu_counter_init(&sbi->s_dirs_counter,
1888 				ext4_count_dirs(sb));
1889 	}
1890 	if (err) {
1891 		printk(KERN_ERR "EXT4-fs: insufficient memory\n");
1892 		goto failed_mount3;
1893 	}
1894 
1895 	/* per fileystem reservation list head & lock */
1896 	spin_lock_init(&sbi->s_rsv_window_lock);
1897 	sbi->s_rsv_window_root = RB_ROOT;
1898 	/* Add a single, static dummy reservation to the start of the
1899 	 * reservation window list --- it gives us a placeholder for
1900 	 * append-at-start-of-list which makes the allocation logic
1901 	 * _much_ simpler. */
1902 	sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
1903 	sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
1904 	sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1905 	sbi->s_rsv_window_head.rsv_goal_size = 0;
1906 	ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
1907 
1908 	/*
1909 	 * set up enough so that it can read an inode
1910 	 */
1911 	sb->s_op = &ext4_sops;
1912 	sb->s_export_op = &ext4_export_ops;
1913 	sb->s_xattr = ext4_xattr_handlers;
1914 #ifdef CONFIG_QUOTA
1915 	sb->s_qcop = &ext4_qctl_operations;
1916 	sb->dq_op = &ext4_quota_operations;
1917 #endif
1918 	INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1919 
1920 	sb->s_root = NULL;
1921 
1922 	needs_recovery = (es->s_last_orphan != 0 ||
1923 			  EXT4_HAS_INCOMPAT_FEATURE(sb,
1924 				    EXT4_FEATURE_INCOMPAT_RECOVER));
1925 
1926 	/*
1927 	 * The first inode we look at is the journal inode.  Don't try
1928 	 * root first: it may be modified in the journal!
1929 	 */
1930 	if (!test_opt(sb, NOLOAD) &&
1931 	    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
1932 		if (ext4_load_journal(sb, es, journal_devnum))
1933 			goto failed_mount3;
1934 	} else if (journal_inum) {
1935 		if (ext4_create_journal(sb, es, journal_inum))
1936 			goto failed_mount3;
1937 	} else {
1938 		if (!silent)
1939 			printk (KERN_ERR
1940 				"ext4: No journal on filesystem on %s\n",
1941 				sb->s_id);
1942 		goto failed_mount3;
1943 	}
1944 
1945 	if (ext4_blocks_count(es) > 0xffffffffULL &&
1946 	    !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
1947 				       JBD2_FEATURE_INCOMPAT_64BIT)) {
1948 		printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
1949 		goto failed_mount4;
1950 	}
1951 
1952 	/* We have now updated the journal if required, so we can
1953 	 * validate the data journaling mode. */
1954 	switch (test_opt(sb, DATA_FLAGS)) {
1955 	case 0:
1956 		/* No mode set, assume a default based on the journal
1957 		 * capabilities: ORDERED_DATA if the journal can
1958 		 * cope, else JOURNAL_DATA
1959 		 */
1960 		if (jbd2_journal_check_available_features
1961 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
1962 			set_opt(sbi->s_mount_opt, ORDERED_DATA);
1963 		else
1964 			set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1965 		break;
1966 
1967 	case EXT4_MOUNT_ORDERED_DATA:
1968 	case EXT4_MOUNT_WRITEBACK_DATA:
1969 		if (!jbd2_journal_check_available_features
1970 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
1971 			printk(KERN_ERR "EXT4-fs: Journal does not support "
1972 			       "requested data journaling mode\n");
1973 			goto failed_mount4;
1974 		}
1975 	default:
1976 		break;
1977 	}
1978 
1979 	if (test_opt(sb, NOBH)) {
1980 		if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
1981 			printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
1982 				"its supported only with writeback mode\n");
1983 			clear_opt(sbi->s_mount_opt, NOBH);
1984 		}
1985 	}
1986 	/*
1987 	 * The jbd2_journal_load will have done any necessary log recovery,
1988 	 * so we can safely mount the rest of the filesystem now.
1989 	 */
1990 
1991 	root = iget(sb, EXT4_ROOT_INO);
1992 	sb->s_root = d_alloc_root(root);
1993 	if (!sb->s_root) {
1994 		printk(KERN_ERR "EXT4-fs: get root inode failed\n");
1995 		iput(root);
1996 		goto failed_mount4;
1997 	}
1998 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1999 		dput(sb->s_root);
2000 		sb->s_root = NULL;
2001 		printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2002 		goto failed_mount4;
2003 	}
2004 
2005 	ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
2006 
2007 	/* determine the minimum size of new large inodes, if present */
2008 	if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2009 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2010 						     EXT4_GOOD_OLD_INODE_SIZE;
2011 		if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2012 				       EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2013 			if (sbi->s_want_extra_isize <
2014 			    le16_to_cpu(es->s_want_extra_isize))
2015 				sbi->s_want_extra_isize =
2016 					le16_to_cpu(es->s_want_extra_isize);
2017 			if (sbi->s_want_extra_isize <
2018 			    le16_to_cpu(es->s_min_extra_isize))
2019 				sbi->s_want_extra_isize =
2020 					le16_to_cpu(es->s_min_extra_isize);
2021 		}
2022 	}
2023 	/* Check if enough inode space is available */
2024 	if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2025 							sbi->s_inode_size) {
2026 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2027 						       EXT4_GOOD_OLD_INODE_SIZE;
2028 		printk(KERN_INFO "EXT4-fs: required extra inode space not"
2029 			"available.\n");
2030 	}
2031 
2032 	/*
2033 	 * akpm: core read_super() calls in here with the superblock locked.
2034 	 * That deadlocks, because orphan cleanup needs to lock the superblock
2035 	 * in numerous places.  Here we just pop the lock - it's relatively
2036 	 * harmless, because we are now ready to accept write_super() requests,
2037 	 * and aviro says that's the only reason for hanging onto the
2038 	 * superblock lock.
2039 	 */
2040 	EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2041 	ext4_orphan_cleanup(sb, es);
2042 	EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2043 	if (needs_recovery)
2044 		printk (KERN_INFO "EXT4-fs: recovery complete.\n");
2045 	ext4_mark_recovery_complete(sb, es);
2046 	printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2047 		test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2048 		test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
2049 		"writeback");
2050 
2051 	ext4_ext_init(sb);
2052 
2053 	lock_kernel();
2054 	return 0;
2055 
2056 cantfind_ext4:
2057 	if (!silent)
2058 		printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2059 		       sb->s_id);
2060 	goto failed_mount;
2061 
2062 failed_mount4:
2063 	jbd2_journal_destroy(sbi->s_journal);
2064 failed_mount3:
2065 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
2066 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
2067 	percpu_counter_destroy(&sbi->s_dirs_counter);
2068 failed_mount2:
2069 	for (i = 0; i < db_count; i++)
2070 		brelse(sbi->s_group_desc[i]);
2071 	kfree(sbi->s_group_desc);
2072 failed_mount:
2073 #ifdef CONFIG_QUOTA
2074 	for (i = 0; i < MAXQUOTAS; i++)
2075 		kfree(sbi->s_qf_names[i]);
2076 #endif
2077 	ext4_blkdev_remove(sbi);
2078 	brelse(bh);
2079 out_fail:
2080 	sb->s_fs_info = NULL;
2081 	kfree(sbi);
2082 	lock_kernel();
2083 	return -EINVAL;
2084 }
2085 
2086 /*
2087  * Setup any per-fs journal parameters now.  We'll do this both on
2088  * initial mount, once the journal has been initialised but before we've
2089  * done any recovery; and again on any subsequent remount.
2090  */
2091 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2092 {
2093 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2094 
2095 	if (sbi->s_commit_interval)
2096 		journal->j_commit_interval = sbi->s_commit_interval;
2097 	/* We could also set up an ext4-specific default for the commit
2098 	 * interval here, but for now we'll just fall back to the jbd
2099 	 * default. */
2100 
2101 	spin_lock(&journal->j_state_lock);
2102 	if (test_opt(sb, BARRIER))
2103 		journal->j_flags |= JBD2_BARRIER;
2104 	else
2105 		journal->j_flags &= ~JBD2_BARRIER;
2106 	spin_unlock(&journal->j_state_lock);
2107 }
2108 
2109 static journal_t *ext4_get_journal(struct super_block *sb,
2110 				   unsigned int journal_inum)
2111 {
2112 	struct inode *journal_inode;
2113 	journal_t *journal;
2114 
2115 	/* First, test for the existence of a valid inode on disk.  Bad
2116 	 * things happen if we iget() an unused inode, as the subsequent
2117 	 * iput() will try to delete it. */
2118 
2119 	journal_inode = iget(sb, journal_inum);
2120 	if (!journal_inode) {
2121 		printk(KERN_ERR "EXT4-fs: no journal found.\n");
2122 		return NULL;
2123 	}
2124 	if (!journal_inode->i_nlink) {
2125 		make_bad_inode(journal_inode);
2126 		iput(journal_inode);
2127 		printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2128 		return NULL;
2129 	}
2130 
2131 	jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2132 		  journal_inode, journal_inode->i_size);
2133 	if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) {
2134 		printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2135 		iput(journal_inode);
2136 		return NULL;
2137 	}
2138 
2139 	journal = jbd2_journal_init_inode(journal_inode);
2140 	if (!journal) {
2141 		printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2142 		iput(journal_inode);
2143 		return NULL;
2144 	}
2145 	journal->j_private = sb;
2146 	ext4_init_journal_params(sb, journal);
2147 	return journal;
2148 }
2149 
2150 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2151 				       dev_t j_dev)
2152 {
2153 	struct buffer_head * bh;
2154 	journal_t *journal;
2155 	ext4_fsblk_t start;
2156 	ext4_fsblk_t len;
2157 	int hblock, blocksize;
2158 	ext4_fsblk_t sb_block;
2159 	unsigned long offset;
2160 	struct ext4_super_block * es;
2161 	struct block_device *bdev;
2162 
2163 	bdev = ext4_blkdev_get(j_dev);
2164 	if (bdev == NULL)
2165 		return NULL;
2166 
2167 	if (bd_claim(bdev, sb)) {
2168 		printk(KERN_ERR
2169 			"EXT4: failed to claim external journal device.\n");
2170 		blkdev_put(bdev);
2171 		return NULL;
2172 	}
2173 
2174 	blocksize = sb->s_blocksize;
2175 	hblock = bdev_hardsect_size(bdev);
2176 	if (blocksize < hblock) {
2177 		printk(KERN_ERR
2178 			"EXT4-fs: blocksize too small for journal device.\n");
2179 		goto out_bdev;
2180 	}
2181 
2182 	sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2183 	offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2184 	set_blocksize(bdev, blocksize);
2185 	if (!(bh = __bread(bdev, sb_block, blocksize))) {
2186 		printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2187 		       "external journal\n");
2188 		goto out_bdev;
2189 	}
2190 
2191 	es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2192 	if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2193 	    !(le32_to_cpu(es->s_feature_incompat) &
2194 	      EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2195 		printk(KERN_ERR "EXT4-fs: external journal has "
2196 					"bad superblock\n");
2197 		brelse(bh);
2198 		goto out_bdev;
2199 	}
2200 
2201 	if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2202 		printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2203 		brelse(bh);
2204 		goto out_bdev;
2205 	}
2206 
2207 	len = ext4_blocks_count(es);
2208 	start = sb_block + 1;
2209 	brelse(bh);	/* we're done with the superblock */
2210 
2211 	journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2212 					start, len, blocksize);
2213 	if (!journal) {
2214 		printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2215 		goto out_bdev;
2216 	}
2217 	journal->j_private = sb;
2218 	ll_rw_block(READ, 1, &journal->j_sb_buffer);
2219 	wait_on_buffer(journal->j_sb_buffer);
2220 	if (!buffer_uptodate(journal->j_sb_buffer)) {
2221 		printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2222 		goto out_journal;
2223 	}
2224 	if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2225 		printk(KERN_ERR "EXT4-fs: External journal has more than one "
2226 					"user (unsupported) - %d\n",
2227 			be32_to_cpu(journal->j_superblock->s_nr_users));
2228 		goto out_journal;
2229 	}
2230 	EXT4_SB(sb)->journal_bdev = bdev;
2231 	ext4_init_journal_params(sb, journal);
2232 	return journal;
2233 out_journal:
2234 	jbd2_journal_destroy(journal);
2235 out_bdev:
2236 	ext4_blkdev_put(bdev);
2237 	return NULL;
2238 }
2239 
2240 static int ext4_load_journal(struct super_block *sb,
2241 			     struct ext4_super_block *es,
2242 			     unsigned long journal_devnum)
2243 {
2244 	journal_t *journal;
2245 	unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2246 	dev_t journal_dev;
2247 	int err = 0;
2248 	int really_read_only;
2249 
2250 	if (journal_devnum &&
2251 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2252 		printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2253 			"numbers have changed\n");
2254 		journal_dev = new_decode_dev(journal_devnum);
2255 	} else
2256 		journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2257 
2258 	really_read_only = bdev_read_only(sb->s_bdev);
2259 
2260 	/*
2261 	 * Are we loading a blank journal or performing recovery after a
2262 	 * crash?  For recovery, we need to check in advance whether we
2263 	 * can get read-write access to the device.
2264 	 */
2265 
2266 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2267 		if (sb->s_flags & MS_RDONLY) {
2268 			printk(KERN_INFO "EXT4-fs: INFO: recovery "
2269 					"required on readonly filesystem.\n");
2270 			if (really_read_only) {
2271 				printk(KERN_ERR "EXT4-fs: write access "
2272 					"unavailable, cannot proceed.\n");
2273 				return -EROFS;
2274 			}
2275 			printk (KERN_INFO "EXT4-fs: write access will "
2276 					"be enabled during recovery.\n");
2277 		}
2278 	}
2279 
2280 	if (journal_inum && journal_dev) {
2281 		printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2282 		       "and inode journals!\n");
2283 		return -EINVAL;
2284 	}
2285 
2286 	if (journal_inum) {
2287 		if (!(journal = ext4_get_journal(sb, journal_inum)))
2288 			return -EINVAL;
2289 	} else {
2290 		if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2291 			return -EINVAL;
2292 	}
2293 
2294 	if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2295 		err = jbd2_journal_update_format(journal);
2296 		if (err)  {
2297 			printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2298 			jbd2_journal_destroy(journal);
2299 			return err;
2300 		}
2301 	}
2302 
2303 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2304 		err = jbd2_journal_wipe(journal, !really_read_only);
2305 	if (!err)
2306 		err = jbd2_journal_load(journal);
2307 
2308 	if (err) {
2309 		printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2310 		jbd2_journal_destroy(journal);
2311 		return err;
2312 	}
2313 
2314 	EXT4_SB(sb)->s_journal = journal;
2315 	ext4_clear_journal_err(sb, es);
2316 
2317 	if (journal_devnum &&
2318 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2319 		es->s_journal_dev = cpu_to_le32(journal_devnum);
2320 		sb->s_dirt = 1;
2321 
2322 		/* Make sure we flush the recovery flag to disk. */
2323 		ext4_commit_super(sb, es, 1);
2324 	}
2325 
2326 	return 0;
2327 }
2328 
2329 static int ext4_create_journal(struct super_block * sb,
2330 			       struct ext4_super_block * es,
2331 			       unsigned int journal_inum)
2332 {
2333 	journal_t *journal;
2334 	int err;
2335 
2336 	if (sb->s_flags & MS_RDONLY) {
2337 		printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
2338 				"create journal.\n");
2339 		return -EROFS;
2340 	}
2341 
2342 	journal = ext4_get_journal(sb, journal_inum);
2343 	if (!journal)
2344 		return -EINVAL;
2345 
2346 	printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
2347 	       journal_inum);
2348 
2349 	err = jbd2_journal_create(journal);
2350 	if (err) {
2351 		printk(KERN_ERR "EXT4-fs: error creating journal.\n");
2352 		jbd2_journal_destroy(journal);
2353 		return -EIO;
2354 	}
2355 
2356 	EXT4_SB(sb)->s_journal = journal;
2357 
2358 	ext4_update_dynamic_rev(sb);
2359 	EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2360 	EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
2361 
2362 	es->s_journal_inum = cpu_to_le32(journal_inum);
2363 	sb->s_dirt = 1;
2364 
2365 	/* Make sure we flush the recovery flag to disk. */
2366 	ext4_commit_super(sb, es, 1);
2367 
2368 	return 0;
2369 }
2370 
2371 static void ext4_commit_super (struct super_block * sb,
2372 			       struct ext4_super_block * es,
2373 			       int sync)
2374 {
2375 	struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2376 
2377 	if (!sbh)
2378 		return;
2379 	es->s_wtime = cpu_to_le32(get_seconds());
2380 	ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
2381 	es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
2382 	BUFFER_TRACE(sbh, "marking dirty");
2383 	mark_buffer_dirty(sbh);
2384 	if (sync)
2385 		sync_dirty_buffer(sbh);
2386 }
2387 
2388 
2389 /*
2390  * Have we just finished recovery?  If so, and if we are mounting (or
2391  * remounting) the filesystem readonly, then we will end up with a
2392  * consistent fs on disk.  Record that fact.
2393  */
2394 static void ext4_mark_recovery_complete(struct super_block * sb,
2395 					struct ext4_super_block * es)
2396 {
2397 	journal_t *journal = EXT4_SB(sb)->s_journal;
2398 
2399 	jbd2_journal_lock_updates(journal);
2400 	jbd2_journal_flush(journal);
2401 	lock_super(sb);
2402 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2403 	    sb->s_flags & MS_RDONLY) {
2404 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2405 		sb->s_dirt = 0;
2406 		ext4_commit_super(sb, es, 1);
2407 	}
2408 	unlock_super(sb);
2409 	jbd2_journal_unlock_updates(journal);
2410 }
2411 
2412 /*
2413  * If we are mounting (or read-write remounting) a filesystem whose journal
2414  * has recorded an error from a previous lifetime, move that error to the
2415  * main filesystem now.
2416  */
2417 static void ext4_clear_journal_err(struct super_block * sb,
2418 				   struct ext4_super_block * es)
2419 {
2420 	journal_t *journal;
2421 	int j_errno;
2422 	const char *errstr;
2423 
2424 	journal = EXT4_SB(sb)->s_journal;
2425 
2426 	/*
2427 	 * Now check for any error status which may have been recorded in the
2428 	 * journal by a prior ext4_error() or ext4_abort()
2429 	 */
2430 
2431 	j_errno = jbd2_journal_errno(journal);
2432 	if (j_errno) {
2433 		char nbuf[16];
2434 
2435 		errstr = ext4_decode_error(sb, j_errno, nbuf);
2436 		ext4_warning(sb, __FUNCTION__, "Filesystem error recorded "
2437 			     "from previous mount: %s", errstr);
2438 		ext4_warning(sb, __FUNCTION__, "Marking fs in need of "
2439 			     "filesystem check.");
2440 
2441 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2442 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2443 		ext4_commit_super (sb, es, 1);
2444 
2445 		jbd2_journal_clear_err(journal);
2446 	}
2447 }
2448 
2449 /*
2450  * Force the running and committing transactions to commit,
2451  * and wait on the commit.
2452  */
2453 int ext4_force_commit(struct super_block *sb)
2454 {
2455 	journal_t *journal;
2456 	int ret;
2457 
2458 	if (sb->s_flags & MS_RDONLY)
2459 		return 0;
2460 
2461 	journal = EXT4_SB(sb)->s_journal;
2462 	sb->s_dirt = 0;
2463 	ret = ext4_journal_force_commit(journal);
2464 	return ret;
2465 }
2466 
2467 /*
2468  * Ext4 always journals updates to the superblock itself, so we don't
2469  * have to propagate any other updates to the superblock on disk at this
2470  * point.  Just start an async writeback to get the buffers on their way
2471  * to the disk.
2472  *
2473  * This implicitly triggers the writebehind on sync().
2474  */
2475 
2476 static void ext4_write_super (struct super_block * sb)
2477 {
2478 	if (mutex_trylock(&sb->s_lock) != 0)
2479 		BUG();
2480 	sb->s_dirt = 0;
2481 }
2482 
2483 static int ext4_sync_fs(struct super_block *sb, int wait)
2484 {
2485 	tid_t target;
2486 
2487 	sb->s_dirt = 0;
2488 	if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
2489 		if (wait)
2490 			jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
2491 	}
2492 	return 0;
2493 }
2494 
2495 /*
2496  * LVM calls this function before a (read-only) snapshot is created.  This
2497  * gives us a chance to flush the journal completely and mark the fs clean.
2498  */
2499 static void ext4_write_super_lockfs(struct super_block *sb)
2500 {
2501 	sb->s_dirt = 0;
2502 
2503 	if (!(sb->s_flags & MS_RDONLY)) {
2504 		journal_t *journal = EXT4_SB(sb)->s_journal;
2505 
2506 		/* Now we set up the journal barrier. */
2507 		jbd2_journal_lock_updates(journal);
2508 		jbd2_journal_flush(journal);
2509 
2510 		/* Journal blocked and flushed, clear needs_recovery flag. */
2511 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2512 		ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2513 	}
2514 }
2515 
2516 /*
2517  * Called by LVM after the snapshot is done.  We need to reset the RECOVER
2518  * flag here, even though the filesystem is not technically dirty yet.
2519  */
2520 static void ext4_unlockfs(struct super_block *sb)
2521 {
2522 	if (!(sb->s_flags & MS_RDONLY)) {
2523 		lock_super(sb);
2524 		/* Reser the needs_recovery flag before the fs is unlocked. */
2525 		EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2526 		ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2527 		unlock_super(sb);
2528 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
2529 	}
2530 }
2531 
2532 static int ext4_remount (struct super_block * sb, int * flags, char * data)
2533 {
2534 	struct ext4_super_block * es;
2535 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2536 	ext4_fsblk_t n_blocks_count = 0;
2537 	unsigned long old_sb_flags;
2538 	struct ext4_mount_options old_opts;
2539 	int err;
2540 #ifdef CONFIG_QUOTA
2541 	int i;
2542 #endif
2543 
2544 	/* Store the original options */
2545 	old_sb_flags = sb->s_flags;
2546 	old_opts.s_mount_opt = sbi->s_mount_opt;
2547 	old_opts.s_resuid = sbi->s_resuid;
2548 	old_opts.s_resgid = sbi->s_resgid;
2549 	old_opts.s_commit_interval = sbi->s_commit_interval;
2550 #ifdef CONFIG_QUOTA
2551 	old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2552 	for (i = 0; i < MAXQUOTAS; i++)
2553 		old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2554 #endif
2555 
2556 	/*
2557 	 * Allow the "check" option to be passed as a remount option.
2558 	 */
2559 	if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2560 		err = -EINVAL;
2561 		goto restore_opts;
2562 	}
2563 
2564 	if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
2565 		ext4_abort(sb, __FUNCTION__, "Abort forced by user");
2566 
2567 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2568 		((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2569 
2570 	es = sbi->s_es;
2571 
2572 	ext4_init_journal_params(sb, sbi->s_journal);
2573 
2574 	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2575 		n_blocks_count > ext4_blocks_count(es)) {
2576 		if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
2577 			err = -EROFS;
2578 			goto restore_opts;
2579 		}
2580 
2581 		if (*flags & MS_RDONLY) {
2582 			/*
2583 			 * First of all, the unconditional stuff we have to do
2584 			 * to disable replay of the journal when we next remount
2585 			 */
2586 			sb->s_flags |= MS_RDONLY;
2587 
2588 			/*
2589 			 * OK, test if we are remounting a valid rw partition
2590 			 * readonly, and if so set the rdonly flag and then
2591 			 * mark the partition as valid again.
2592 			 */
2593 			if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
2594 			    (sbi->s_mount_state & EXT4_VALID_FS))
2595 				es->s_state = cpu_to_le16(sbi->s_mount_state);
2596 
2597 			/*
2598 			 * We have to unlock super so that we can wait for
2599 			 * transactions.
2600 			 */
2601 			unlock_super(sb);
2602 			ext4_mark_recovery_complete(sb, es);
2603 			lock_super(sb);
2604 		} else {
2605 			__le32 ret;
2606 			if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2607 					~EXT4_FEATURE_RO_COMPAT_SUPP))) {
2608 				printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2609 				       "remount RDWR because of unsupported "
2610 				       "optional features (%x).\n",
2611 				       sb->s_id, le32_to_cpu(ret));
2612 				err = -EROFS;
2613 				goto restore_opts;
2614 			}
2615 
2616 			/*
2617 			 * If we have an unprocessed orphan list hanging
2618 			 * around from a previously readonly bdev mount,
2619 			 * require a full umount/remount for now.
2620 			 */
2621 			if (es->s_last_orphan) {
2622 				printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2623 				       "remount RDWR because of unprocessed "
2624 				       "orphan inode list.  Please "
2625 				       "umount/remount instead.\n",
2626 				       sb->s_id);
2627 				err = -EINVAL;
2628 				goto restore_opts;
2629 			}
2630 
2631 			/*
2632 			 * Mounting a RDONLY partition read-write, so reread
2633 			 * and store the current valid flag.  (It may have
2634 			 * been changed by e2fsck since we originally mounted
2635 			 * the partition.)
2636 			 */
2637 			ext4_clear_journal_err(sb, es);
2638 			sbi->s_mount_state = le16_to_cpu(es->s_state);
2639 			if ((err = ext4_group_extend(sb, es, n_blocks_count)))
2640 				goto restore_opts;
2641 			if (!ext4_setup_super (sb, es, 0))
2642 				sb->s_flags &= ~MS_RDONLY;
2643 		}
2644 	}
2645 #ifdef CONFIG_QUOTA
2646 	/* Release old quota file names */
2647 	for (i = 0; i < MAXQUOTAS; i++)
2648 		if (old_opts.s_qf_names[i] &&
2649 		    old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2650 			kfree(old_opts.s_qf_names[i]);
2651 #endif
2652 	return 0;
2653 restore_opts:
2654 	sb->s_flags = old_sb_flags;
2655 	sbi->s_mount_opt = old_opts.s_mount_opt;
2656 	sbi->s_resuid = old_opts.s_resuid;
2657 	sbi->s_resgid = old_opts.s_resgid;
2658 	sbi->s_commit_interval = old_opts.s_commit_interval;
2659 #ifdef CONFIG_QUOTA
2660 	sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2661 	for (i = 0; i < MAXQUOTAS; i++) {
2662 		if (sbi->s_qf_names[i] &&
2663 		    old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2664 			kfree(sbi->s_qf_names[i]);
2665 		sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2666 	}
2667 #endif
2668 	return err;
2669 }
2670 
2671 static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf)
2672 {
2673 	struct super_block *sb = dentry->d_sb;
2674 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2675 	struct ext4_super_block *es = sbi->s_es;
2676 	u64 fsid;
2677 
2678 	if (test_opt(sb, MINIX_DF)) {
2679 		sbi->s_overhead_last = 0;
2680 	} else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
2681 		unsigned long ngroups = sbi->s_groups_count, i;
2682 		ext4_fsblk_t overhead = 0;
2683 		smp_rmb();
2684 
2685 		/*
2686 		 * Compute the overhead (FS structures).  This is constant
2687 		 * for a given filesystem unless the number of block groups
2688 		 * changes so we cache the previous value until it does.
2689 		 */
2690 
2691 		/*
2692 		 * All of the blocks before first_data_block are
2693 		 * overhead
2694 		 */
2695 		overhead = le32_to_cpu(es->s_first_data_block);
2696 
2697 		/*
2698 		 * Add the overhead attributed to the superblock and
2699 		 * block group descriptors.  If the sparse superblocks
2700 		 * feature is turned on, then not all groups have this.
2701 		 */
2702 		for (i = 0; i < ngroups; i++) {
2703 			overhead += ext4_bg_has_super(sb, i) +
2704 				ext4_bg_num_gdb(sb, i);
2705 			cond_resched();
2706 		}
2707 
2708 		/*
2709 		 * Every block group has an inode bitmap, a block
2710 		 * bitmap, and an inode table.
2711 		 */
2712 		overhead += ngroups * (2 + sbi->s_itb_per_group);
2713 		sbi->s_overhead_last = overhead;
2714 		smp_wmb();
2715 		sbi->s_blocks_last = ext4_blocks_count(es);
2716 	}
2717 
2718 	buf->f_type = EXT4_SUPER_MAGIC;
2719 	buf->f_bsize = sb->s_blocksize;
2720 	buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
2721 	buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2722 	ext4_free_blocks_count_set(es, buf->f_bfree);
2723 	buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
2724 	if (buf->f_bfree < ext4_r_blocks_count(es))
2725 		buf->f_bavail = 0;
2726 	buf->f_files = le32_to_cpu(es->s_inodes_count);
2727 	buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2728 	es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
2729 	buf->f_namelen = EXT4_NAME_LEN;
2730 	fsid = le64_to_cpup((void *)es->s_uuid) ^
2731 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2732 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2733 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2734 	return 0;
2735 }
2736 
2737 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2738  * is locked for write. Otherwise the are possible deadlocks:
2739  * Process 1                         Process 2
2740  * ext4_create()                     quota_sync()
2741  *   jbd2_journal_start()                   write_dquot()
2742  *   DQUOT_INIT()                        down(dqio_mutex)
2743  *     down(dqio_mutex)                    jbd2_journal_start()
2744  *
2745  */
2746 
2747 #ifdef CONFIG_QUOTA
2748 
2749 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2750 {
2751 	return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2752 }
2753 
2754 static int ext4_dquot_initialize(struct inode *inode, int type)
2755 {
2756 	handle_t *handle;
2757 	int ret, err;
2758 
2759 	/* We may create quota structure so we need to reserve enough blocks */
2760 	handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
2761 	if (IS_ERR(handle))
2762 		return PTR_ERR(handle);
2763 	ret = dquot_initialize(inode, type);
2764 	err = ext4_journal_stop(handle);
2765 	if (!ret)
2766 		ret = err;
2767 	return ret;
2768 }
2769 
2770 static int ext4_dquot_drop(struct inode *inode)
2771 {
2772 	handle_t *handle;
2773 	int ret, err;
2774 
2775 	/* We may delete quota structure so we need to reserve enough blocks */
2776 	handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
2777 	if (IS_ERR(handle))
2778 		return PTR_ERR(handle);
2779 	ret = dquot_drop(inode);
2780 	err = ext4_journal_stop(handle);
2781 	if (!ret)
2782 		ret = err;
2783 	return ret;
2784 }
2785 
2786 static int ext4_write_dquot(struct dquot *dquot)
2787 {
2788 	int ret, err;
2789 	handle_t *handle;
2790 	struct inode *inode;
2791 
2792 	inode = dquot_to_inode(dquot);
2793 	handle = ext4_journal_start(inode,
2794 					EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2795 	if (IS_ERR(handle))
2796 		return PTR_ERR(handle);
2797 	ret = dquot_commit(dquot);
2798 	err = ext4_journal_stop(handle);
2799 	if (!ret)
2800 		ret = err;
2801 	return ret;
2802 }
2803 
2804 static int ext4_acquire_dquot(struct dquot *dquot)
2805 {
2806 	int ret, err;
2807 	handle_t *handle;
2808 
2809 	handle = ext4_journal_start(dquot_to_inode(dquot),
2810 					EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2811 	if (IS_ERR(handle))
2812 		return PTR_ERR(handle);
2813 	ret = dquot_acquire(dquot);
2814 	err = ext4_journal_stop(handle);
2815 	if (!ret)
2816 		ret = err;
2817 	return ret;
2818 }
2819 
2820 static int ext4_release_dquot(struct dquot *dquot)
2821 {
2822 	int ret, err;
2823 	handle_t *handle;
2824 
2825 	handle = ext4_journal_start(dquot_to_inode(dquot),
2826 					EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2827 	if (IS_ERR(handle)) {
2828 		/* Release dquot anyway to avoid endless cycle in dqput() */
2829 		dquot_release(dquot);
2830 		return PTR_ERR(handle);
2831 	}
2832 	ret = dquot_release(dquot);
2833 	err = ext4_journal_stop(handle);
2834 	if (!ret)
2835 		ret = err;
2836 	return ret;
2837 }
2838 
2839 static int ext4_mark_dquot_dirty(struct dquot *dquot)
2840 {
2841 	/* Are we journalling quotas? */
2842 	if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2843 	    EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2844 		dquot_mark_dquot_dirty(dquot);
2845 		return ext4_write_dquot(dquot);
2846 	} else {
2847 		return dquot_mark_dquot_dirty(dquot);
2848 	}
2849 }
2850 
2851 static int ext4_write_info(struct super_block *sb, int type)
2852 {
2853 	int ret, err;
2854 	handle_t *handle;
2855 
2856 	/* Data block + inode block */
2857 	handle = ext4_journal_start(sb->s_root->d_inode, 2);
2858 	if (IS_ERR(handle))
2859 		return PTR_ERR(handle);
2860 	ret = dquot_commit_info(sb, type);
2861 	err = ext4_journal_stop(handle);
2862 	if (!ret)
2863 		ret = err;
2864 	return ret;
2865 }
2866 
2867 /*
2868  * Turn on quotas during mount time - we need to find
2869  * the quota file and such...
2870  */
2871 static int ext4_quota_on_mount(struct super_block *sb, int type)
2872 {
2873 	return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
2874 			EXT4_SB(sb)->s_jquota_fmt, type);
2875 }
2876 
2877 /*
2878  * Standard function to be called on quota_on
2879  */
2880 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
2881 			 char *path)
2882 {
2883 	int err;
2884 	struct nameidata nd;
2885 
2886 	if (!test_opt(sb, QUOTA))
2887 		return -EINVAL;
2888 	/* Not journalling quota? */
2889 	if (!EXT4_SB(sb)->s_qf_names[USRQUOTA] &&
2890 	    !EXT4_SB(sb)->s_qf_names[GRPQUOTA])
2891 		return vfs_quota_on(sb, type, format_id, path);
2892 	err = path_lookup(path, LOOKUP_FOLLOW, &nd);
2893 	if (err)
2894 		return err;
2895 	/* Quotafile not on the same filesystem? */
2896 	if (nd.mnt->mnt_sb != sb) {
2897 		path_release(&nd);
2898 		return -EXDEV;
2899 	}
2900 	/* Quotafile not of fs root? */
2901 	if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
2902 		printk(KERN_WARNING
2903 			"EXT4-fs: Quota file not on filesystem root. "
2904 			"Journalled quota will not work.\n");
2905 	path_release(&nd);
2906 	return vfs_quota_on(sb, type, format_id, path);
2907 }
2908 
2909 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2910  * acquiring the locks... As quota files are never truncated and quota code
2911  * itself serializes the operations (and noone else should touch the files)
2912  * we don't have to be afraid of races */
2913 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
2914 			       size_t len, loff_t off)
2915 {
2916 	struct inode *inode = sb_dqopt(sb)->files[type];
2917 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
2918 	int err = 0;
2919 	int offset = off & (sb->s_blocksize - 1);
2920 	int tocopy;
2921 	size_t toread;
2922 	struct buffer_head *bh;
2923 	loff_t i_size = i_size_read(inode);
2924 
2925 	if (off > i_size)
2926 		return 0;
2927 	if (off+len > i_size)
2928 		len = i_size-off;
2929 	toread = len;
2930 	while (toread > 0) {
2931 		tocopy = sb->s_blocksize - offset < toread ?
2932 				sb->s_blocksize - offset : toread;
2933 		bh = ext4_bread(NULL, inode, blk, 0, &err);
2934 		if (err)
2935 			return err;
2936 		if (!bh)	/* A hole? */
2937 			memset(data, 0, tocopy);
2938 		else
2939 			memcpy(data, bh->b_data+offset, tocopy);
2940 		brelse(bh);
2941 		offset = 0;
2942 		toread -= tocopy;
2943 		data += tocopy;
2944 		blk++;
2945 	}
2946 	return len;
2947 }
2948 
2949 /* Write to quotafile (we know the transaction is already started and has
2950  * enough credits) */
2951 static ssize_t ext4_quota_write(struct super_block *sb, int type,
2952 				const char *data, size_t len, loff_t off)
2953 {
2954 	struct inode *inode = sb_dqopt(sb)->files[type];
2955 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
2956 	int err = 0;
2957 	int offset = off & (sb->s_blocksize - 1);
2958 	int tocopy;
2959 	int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
2960 	size_t towrite = len;
2961 	struct buffer_head *bh;
2962 	handle_t *handle = journal_current_handle();
2963 
2964 	if (!handle) {
2965 		printk(KERN_WARNING "EXT4-fs: Quota write (off=%Lu, len=%Lu)"
2966 			" cancelled because transaction is not started.\n",
2967 			(unsigned long long)off, (unsigned long long)len);
2968 		return -EIO;
2969 	}
2970 	mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2971 	while (towrite > 0) {
2972 		tocopy = sb->s_blocksize - offset < towrite ?
2973 				sb->s_blocksize - offset : towrite;
2974 		bh = ext4_bread(handle, inode, blk, 1, &err);
2975 		if (!bh)
2976 			goto out;
2977 		if (journal_quota) {
2978 			err = ext4_journal_get_write_access(handle, bh);
2979 			if (err) {
2980 				brelse(bh);
2981 				goto out;
2982 			}
2983 		}
2984 		lock_buffer(bh);
2985 		memcpy(bh->b_data+offset, data, tocopy);
2986 		flush_dcache_page(bh->b_page);
2987 		unlock_buffer(bh);
2988 		if (journal_quota)
2989 			err = ext4_journal_dirty_metadata(handle, bh);
2990 		else {
2991 			/* Always do at least ordered writes for quotas */
2992 			err = ext4_journal_dirty_data(handle, bh);
2993 			mark_buffer_dirty(bh);
2994 		}
2995 		brelse(bh);
2996 		if (err)
2997 			goto out;
2998 		offset = 0;
2999 		towrite -= tocopy;
3000 		data += tocopy;
3001 		blk++;
3002 	}
3003 out:
3004 	if (len == towrite)
3005 		return err;
3006 	if (inode->i_size < off+len-towrite) {
3007 		i_size_write(inode, off+len-towrite);
3008 		EXT4_I(inode)->i_disksize = inode->i_size;
3009 	}
3010 	inode->i_version++;
3011 	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3012 	ext4_mark_inode_dirty(handle, inode);
3013 	mutex_unlock(&inode->i_mutex);
3014 	return len - towrite;
3015 }
3016 
3017 #endif
3018 
3019 static int ext4_get_sb(struct file_system_type *fs_type,
3020 	int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3021 {
3022 	return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3023 }
3024 
3025 static struct file_system_type ext4dev_fs_type = {
3026 	.owner		= THIS_MODULE,
3027 	.name		= "ext4dev",
3028 	.get_sb		= ext4_get_sb,
3029 	.kill_sb	= kill_block_super,
3030 	.fs_flags	= FS_REQUIRES_DEV,
3031 };
3032 
3033 static int __init init_ext4_fs(void)
3034 {
3035 	int err = init_ext4_xattr();
3036 	if (err)
3037 		return err;
3038 	err = init_inodecache();
3039 	if (err)
3040 		goto out1;
3041 	err = register_filesystem(&ext4dev_fs_type);
3042 	if (err)
3043 		goto out;
3044 	return 0;
3045 out:
3046 	destroy_inodecache();
3047 out1:
3048 	exit_ext4_xattr();
3049 	return err;
3050 }
3051 
3052 static void __exit exit_ext4_fs(void)
3053 {
3054 	unregister_filesystem(&ext4dev_fs_type);
3055 	destroy_inodecache();
3056 	exit_ext4_xattr();
3057 }
3058 
3059 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3060 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
3061 MODULE_LICENSE("GPL");
3062 module_init(init_ext4_fs)
3063 module_exit(exit_ext4_fs)
3064