xref: /linux/fs/ext4/super.c (revision c80544dc0b87bb65038355e7aafdc30be16b26ab)
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 dentry *ext4_get_dentry(struct super_block *sb, void *vobjp)
690 {
691 	__u32 *objp = vobjp;
692 	unsigned long ino = objp[0];
693 	__u32 generation = objp[1];
694 	struct inode *inode;
695 	struct dentry *result;
696 
697 	if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
698 		return ERR_PTR(-ESTALE);
699 	if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
700 		return ERR_PTR(-ESTALE);
701 
702 	/* iget isn't really right if the inode is currently unallocated!!
703 	 *
704 	 * ext4_read_inode will return a bad_inode if the inode had been
705 	 * deleted, so we should be safe.
706 	 *
707 	 * Currently we don't know the generation for parent directory, so
708 	 * a generation of 0 means "accept any"
709 	 */
710 	inode = iget(sb, ino);
711 	if (inode == NULL)
712 		return ERR_PTR(-ENOMEM);
713 	if (is_bad_inode(inode) ||
714 	    (generation && inode->i_generation != generation)) {
715 		iput(inode);
716 		return ERR_PTR(-ESTALE);
717 	}
718 	/* now to find a dentry.
719 	 * If possible, get a well-connected one
720 	 */
721 	result = d_alloc_anon(inode);
722 	if (!result) {
723 		iput(inode);
724 		return ERR_PTR(-ENOMEM);
725 	}
726 	return result;
727 }
728 
729 #ifdef CONFIG_QUOTA
730 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
731 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
732 
733 static int ext4_dquot_initialize(struct inode *inode, int type);
734 static int ext4_dquot_drop(struct inode *inode);
735 static int ext4_write_dquot(struct dquot *dquot);
736 static int ext4_acquire_dquot(struct dquot *dquot);
737 static int ext4_release_dquot(struct dquot *dquot);
738 static int ext4_mark_dquot_dirty(struct dquot *dquot);
739 static int ext4_write_info(struct super_block *sb, int type);
740 static int ext4_quota_on(struct super_block *sb, int type, int format_id, char *path);
741 static int ext4_quota_on_mount(struct super_block *sb, int type);
742 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
743 			       size_t len, loff_t off);
744 static ssize_t ext4_quota_write(struct super_block *sb, int type,
745 				const char *data, size_t len, loff_t off);
746 
747 static struct dquot_operations ext4_quota_operations = {
748 	.initialize	= ext4_dquot_initialize,
749 	.drop		= ext4_dquot_drop,
750 	.alloc_space	= dquot_alloc_space,
751 	.alloc_inode	= dquot_alloc_inode,
752 	.free_space	= dquot_free_space,
753 	.free_inode	= dquot_free_inode,
754 	.transfer	= dquot_transfer,
755 	.write_dquot	= ext4_write_dquot,
756 	.acquire_dquot	= ext4_acquire_dquot,
757 	.release_dquot	= ext4_release_dquot,
758 	.mark_dirty	= ext4_mark_dquot_dirty,
759 	.write_info	= ext4_write_info
760 };
761 
762 static struct quotactl_ops ext4_qctl_operations = {
763 	.quota_on	= ext4_quota_on,
764 	.quota_off	= vfs_quota_off,
765 	.quota_sync	= vfs_quota_sync,
766 	.get_info	= vfs_get_dqinfo,
767 	.set_info	= vfs_set_dqinfo,
768 	.get_dqblk	= vfs_get_dqblk,
769 	.set_dqblk	= vfs_set_dqblk
770 };
771 #endif
772 
773 static const struct super_operations ext4_sops = {
774 	.alloc_inode	= ext4_alloc_inode,
775 	.destroy_inode	= ext4_destroy_inode,
776 	.read_inode	= ext4_read_inode,
777 	.write_inode	= ext4_write_inode,
778 	.dirty_inode	= ext4_dirty_inode,
779 	.delete_inode	= ext4_delete_inode,
780 	.put_super	= ext4_put_super,
781 	.write_super	= ext4_write_super,
782 	.sync_fs	= ext4_sync_fs,
783 	.write_super_lockfs = ext4_write_super_lockfs,
784 	.unlockfs	= ext4_unlockfs,
785 	.statfs		= ext4_statfs,
786 	.remount_fs	= ext4_remount,
787 	.clear_inode	= ext4_clear_inode,
788 	.show_options	= ext4_show_options,
789 #ifdef CONFIG_QUOTA
790 	.quota_read	= ext4_quota_read,
791 	.quota_write	= ext4_quota_write,
792 #endif
793 };
794 
795 static struct export_operations ext4_export_ops = {
796 	.get_parent = ext4_get_parent,
797 	.get_dentry = ext4_get_dentry,
798 };
799 
800 enum {
801 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
802 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
803 	Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
804 	Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
805 	Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
806 	Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
807 	Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
808 	Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
809 	Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
810 	Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
811 	Opt_grpquota, Opt_extents, Opt_noextents,
812 };
813 
814 static match_table_t tokens = {
815 	{Opt_bsd_df, "bsddf"},
816 	{Opt_minix_df, "minixdf"},
817 	{Opt_grpid, "grpid"},
818 	{Opt_grpid, "bsdgroups"},
819 	{Opt_nogrpid, "nogrpid"},
820 	{Opt_nogrpid, "sysvgroups"},
821 	{Opt_resgid, "resgid=%u"},
822 	{Opt_resuid, "resuid=%u"},
823 	{Opt_sb, "sb=%u"},
824 	{Opt_err_cont, "errors=continue"},
825 	{Opt_err_panic, "errors=panic"},
826 	{Opt_err_ro, "errors=remount-ro"},
827 	{Opt_nouid32, "nouid32"},
828 	{Opt_nocheck, "nocheck"},
829 	{Opt_nocheck, "check=none"},
830 	{Opt_debug, "debug"},
831 	{Opt_oldalloc, "oldalloc"},
832 	{Opt_orlov, "orlov"},
833 	{Opt_user_xattr, "user_xattr"},
834 	{Opt_nouser_xattr, "nouser_xattr"},
835 	{Opt_acl, "acl"},
836 	{Opt_noacl, "noacl"},
837 	{Opt_reservation, "reservation"},
838 	{Opt_noreservation, "noreservation"},
839 	{Opt_noload, "noload"},
840 	{Opt_nobh, "nobh"},
841 	{Opt_bh, "bh"},
842 	{Opt_commit, "commit=%u"},
843 	{Opt_journal_update, "journal=update"},
844 	{Opt_journal_inum, "journal=%u"},
845 	{Opt_journal_dev, "journal_dev=%u"},
846 	{Opt_abort, "abort"},
847 	{Opt_data_journal, "data=journal"},
848 	{Opt_data_ordered, "data=ordered"},
849 	{Opt_data_writeback, "data=writeback"},
850 	{Opt_offusrjquota, "usrjquota="},
851 	{Opt_usrjquota, "usrjquota=%s"},
852 	{Opt_offgrpjquota, "grpjquota="},
853 	{Opt_grpjquota, "grpjquota=%s"},
854 	{Opt_jqfmt_vfsold, "jqfmt=vfsold"},
855 	{Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
856 	{Opt_grpquota, "grpquota"},
857 	{Opt_noquota, "noquota"},
858 	{Opt_quota, "quota"},
859 	{Opt_usrquota, "usrquota"},
860 	{Opt_barrier, "barrier=%u"},
861 	{Opt_extents, "extents"},
862 	{Opt_noextents, "noextents"},
863 	{Opt_err, NULL},
864 	{Opt_resize, "resize"},
865 };
866 
867 static ext4_fsblk_t get_sb_block(void **data)
868 {
869 	ext4_fsblk_t	sb_block;
870 	char		*options = (char *) *data;
871 
872 	if (!options || strncmp(options, "sb=", 3) != 0)
873 		return 1;	/* Default location */
874 	options += 3;
875 	/*todo: use simple_strtoll with >32bit ext4 */
876 	sb_block = simple_strtoul(options, &options, 0);
877 	if (*options && *options != ',') {
878 		printk("EXT4-fs: Invalid sb specification: %s\n",
879 		       (char *) *data);
880 		return 1;
881 	}
882 	if (*options == ',')
883 		options++;
884 	*data = (void *) options;
885 	return sb_block;
886 }
887 
888 static int parse_options (char *options, struct super_block *sb,
889 			  unsigned int *inum, unsigned long *journal_devnum,
890 			  ext4_fsblk_t *n_blocks_count, int is_remount)
891 {
892 	struct ext4_sb_info *sbi = EXT4_SB(sb);
893 	char * p;
894 	substring_t args[MAX_OPT_ARGS];
895 	int data_opt = 0;
896 	int option;
897 #ifdef CONFIG_QUOTA
898 	int qtype;
899 	char *qname;
900 #endif
901 
902 	if (!options)
903 		return 1;
904 
905 	while ((p = strsep (&options, ",")) != NULL) {
906 		int token;
907 		if (!*p)
908 			continue;
909 
910 		token = match_token(p, tokens, args);
911 		switch (token) {
912 		case Opt_bsd_df:
913 			clear_opt (sbi->s_mount_opt, MINIX_DF);
914 			break;
915 		case Opt_minix_df:
916 			set_opt (sbi->s_mount_opt, MINIX_DF);
917 			break;
918 		case Opt_grpid:
919 			set_opt (sbi->s_mount_opt, GRPID);
920 			break;
921 		case Opt_nogrpid:
922 			clear_opt (sbi->s_mount_opt, GRPID);
923 			break;
924 		case Opt_resuid:
925 			if (match_int(&args[0], &option))
926 				return 0;
927 			sbi->s_resuid = option;
928 			break;
929 		case Opt_resgid:
930 			if (match_int(&args[0], &option))
931 				return 0;
932 			sbi->s_resgid = option;
933 			break;
934 		case Opt_sb:
935 			/* handled by get_sb_block() instead of here */
936 			/* *sb_block = match_int(&args[0]); */
937 			break;
938 		case Opt_err_panic:
939 			clear_opt (sbi->s_mount_opt, ERRORS_CONT);
940 			clear_opt (sbi->s_mount_opt, ERRORS_RO);
941 			set_opt (sbi->s_mount_opt, ERRORS_PANIC);
942 			break;
943 		case Opt_err_ro:
944 			clear_opt (sbi->s_mount_opt, ERRORS_CONT);
945 			clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
946 			set_opt (sbi->s_mount_opt, ERRORS_RO);
947 			break;
948 		case Opt_err_cont:
949 			clear_opt (sbi->s_mount_opt, ERRORS_RO);
950 			clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
951 			set_opt (sbi->s_mount_opt, ERRORS_CONT);
952 			break;
953 		case Opt_nouid32:
954 			set_opt (sbi->s_mount_opt, NO_UID32);
955 			break;
956 		case Opt_nocheck:
957 			clear_opt (sbi->s_mount_opt, CHECK);
958 			break;
959 		case Opt_debug:
960 			set_opt (sbi->s_mount_opt, DEBUG);
961 			break;
962 		case Opt_oldalloc:
963 			set_opt (sbi->s_mount_opt, OLDALLOC);
964 			break;
965 		case Opt_orlov:
966 			clear_opt (sbi->s_mount_opt, OLDALLOC);
967 			break;
968 #ifdef CONFIG_EXT4DEV_FS_XATTR
969 		case Opt_user_xattr:
970 			set_opt (sbi->s_mount_opt, XATTR_USER);
971 			break;
972 		case Opt_nouser_xattr:
973 			clear_opt (sbi->s_mount_opt, XATTR_USER);
974 			break;
975 #else
976 		case Opt_user_xattr:
977 		case Opt_nouser_xattr:
978 			printk("EXT4 (no)user_xattr options not supported\n");
979 			break;
980 #endif
981 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
982 		case Opt_acl:
983 			set_opt(sbi->s_mount_opt, POSIX_ACL);
984 			break;
985 		case Opt_noacl:
986 			clear_opt(sbi->s_mount_opt, POSIX_ACL);
987 			break;
988 #else
989 		case Opt_acl:
990 		case Opt_noacl:
991 			printk("EXT4 (no)acl options not supported\n");
992 			break;
993 #endif
994 		case Opt_reservation:
995 			set_opt(sbi->s_mount_opt, RESERVATION);
996 			break;
997 		case Opt_noreservation:
998 			clear_opt(sbi->s_mount_opt, RESERVATION);
999 			break;
1000 		case Opt_journal_update:
1001 			/* @@@ FIXME */
1002 			/* Eventually we will want to be able to create
1003 			   a journal file here.  For now, only allow the
1004 			   user to specify an existing inode to be the
1005 			   journal file. */
1006 			if (is_remount) {
1007 				printk(KERN_ERR "EXT4-fs: cannot specify "
1008 				       "journal on remount\n");
1009 				return 0;
1010 			}
1011 			set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1012 			break;
1013 		case Opt_journal_inum:
1014 			if (is_remount) {
1015 				printk(KERN_ERR "EXT4-fs: cannot specify "
1016 				       "journal on remount\n");
1017 				return 0;
1018 			}
1019 			if (match_int(&args[0], &option))
1020 				return 0;
1021 			*inum = option;
1022 			break;
1023 		case Opt_journal_dev:
1024 			if (is_remount) {
1025 				printk(KERN_ERR "EXT4-fs: cannot specify "
1026 				       "journal on remount\n");
1027 				return 0;
1028 			}
1029 			if (match_int(&args[0], &option))
1030 				return 0;
1031 			*journal_devnum = option;
1032 			break;
1033 		case Opt_noload:
1034 			set_opt (sbi->s_mount_opt, NOLOAD);
1035 			break;
1036 		case Opt_commit:
1037 			if (match_int(&args[0], &option))
1038 				return 0;
1039 			if (option < 0)
1040 				return 0;
1041 			if (option == 0)
1042 				option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1043 			sbi->s_commit_interval = HZ * option;
1044 			break;
1045 		case Opt_data_journal:
1046 			data_opt = EXT4_MOUNT_JOURNAL_DATA;
1047 			goto datacheck;
1048 		case Opt_data_ordered:
1049 			data_opt = EXT4_MOUNT_ORDERED_DATA;
1050 			goto datacheck;
1051 		case Opt_data_writeback:
1052 			data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1053 		datacheck:
1054 			if (is_remount) {
1055 				if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1056 						!= data_opt) {
1057 					printk(KERN_ERR
1058 						"EXT4-fs: cannot change data "
1059 						"mode on remount\n");
1060 					return 0;
1061 				}
1062 			} else {
1063 				sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1064 				sbi->s_mount_opt |= data_opt;
1065 			}
1066 			break;
1067 #ifdef CONFIG_QUOTA
1068 		case Opt_usrjquota:
1069 			qtype = USRQUOTA;
1070 			goto set_qf_name;
1071 		case Opt_grpjquota:
1072 			qtype = GRPQUOTA;
1073 set_qf_name:
1074 			if (sb_any_quota_enabled(sb)) {
1075 				printk(KERN_ERR
1076 					"EXT4-fs: Cannot change journalled "
1077 					"quota options when quota turned on.\n");
1078 				return 0;
1079 			}
1080 			qname = match_strdup(&args[0]);
1081 			if (!qname) {
1082 				printk(KERN_ERR
1083 					"EXT4-fs: not enough memory for "
1084 					"storing quotafile name.\n");
1085 				return 0;
1086 			}
1087 			if (sbi->s_qf_names[qtype] &&
1088 			    strcmp(sbi->s_qf_names[qtype], qname)) {
1089 				printk(KERN_ERR
1090 					"EXT4-fs: %s quota file already "
1091 					"specified.\n", QTYPE2NAME(qtype));
1092 				kfree(qname);
1093 				return 0;
1094 			}
1095 			sbi->s_qf_names[qtype] = qname;
1096 			if (strchr(sbi->s_qf_names[qtype], '/')) {
1097 				printk(KERN_ERR
1098 					"EXT4-fs: quotafile must be on "
1099 					"filesystem root.\n");
1100 				kfree(sbi->s_qf_names[qtype]);
1101 				sbi->s_qf_names[qtype] = NULL;
1102 				return 0;
1103 			}
1104 			set_opt(sbi->s_mount_opt, QUOTA);
1105 			break;
1106 		case Opt_offusrjquota:
1107 			qtype = USRQUOTA;
1108 			goto clear_qf_name;
1109 		case Opt_offgrpjquota:
1110 			qtype = GRPQUOTA;
1111 clear_qf_name:
1112 			if (sb_any_quota_enabled(sb)) {
1113 				printk(KERN_ERR "EXT4-fs: Cannot change "
1114 					"journalled quota options when "
1115 					"quota turned on.\n");
1116 				return 0;
1117 			}
1118 			/*
1119 			 * The space will be released later when all options
1120 			 * are confirmed to be correct
1121 			 */
1122 			sbi->s_qf_names[qtype] = NULL;
1123 			break;
1124 		case Opt_jqfmt_vfsold:
1125 			sbi->s_jquota_fmt = QFMT_VFS_OLD;
1126 			break;
1127 		case Opt_jqfmt_vfsv0:
1128 			sbi->s_jquota_fmt = QFMT_VFS_V0;
1129 			break;
1130 		case Opt_quota:
1131 		case Opt_usrquota:
1132 			set_opt(sbi->s_mount_opt, QUOTA);
1133 			set_opt(sbi->s_mount_opt, USRQUOTA);
1134 			break;
1135 		case Opt_grpquota:
1136 			set_opt(sbi->s_mount_opt, QUOTA);
1137 			set_opt(sbi->s_mount_opt, GRPQUOTA);
1138 			break;
1139 		case Opt_noquota:
1140 			if (sb_any_quota_enabled(sb)) {
1141 				printk(KERN_ERR "EXT4-fs: Cannot change quota "
1142 					"options when quota turned on.\n");
1143 				return 0;
1144 			}
1145 			clear_opt(sbi->s_mount_opt, QUOTA);
1146 			clear_opt(sbi->s_mount_opt, USRQUOTA);
1147 			clear_opt(sbi->s_mount_opt, GRPQUOTA);
1148 			break;
1149 #else
1150 		case Opt_quota:
1151 		case Opt_usrquota:
1152 		case Opt_grpquota:
1153 		case Opt_usrjquota:
1154 		case Opt_grpjquota:
1155 		case Opt_offusrjquota:
1156 		case Opt_offgrpjquota:
1157 		case Opt_jqfmt_vfsold:
1158 		case Opt_jqfmt_vfsv0:
1159 			printk(KERN_ERR
1160 				"EXT4-fs: journalled quota options not "
1161 				"supported.\n");
1162 			break;
1163 		case Opt_noquota:
1164 			break;
1165 #endif
1166 		case Opt_abort:
1167 			set_opt(sbi->s_mount_opt, ABORT);
1168 			break;
1169 		case Opt_barrier:
1170 			if (match_int(&args[0], &option))
1171 				return 0;
1172 			if (option)
1173 				set_opt(sbi->s_mount_opt, BARRIER);
1174 			else
1175 				clear_opt(sbi->s_mount_opt, BARRIER);
1176 			break;
1177 		case Opt_ignore:
1178 			break;
1179 		case Opt_resize:
1180 			if (!is_remount) {
1181 				printk("EXT4-fs: resize option only available "
1182 					"for remount\n");
1183 				return 0;
1184 			}
1185 			if (match_int(&args[0], &option) != 0)
1186 				return 0;
1187 			*n_blocks_count = option;
1188 			break;
1189 		case Opt_nobh:
1190 			set_opt(sbi->s_mount_opt, NOBH);
1191 			break;
1192 		case Opt_bh:
1193 			clear_opt(sbi->s_mount_opt, NOBH);
1194 			break;
1195 		case Opt_extents:
1196 			set_opt (sbi->s_mount_opt, EXTENTS);
1197 			break;
1198 		case Opt_noextents:
1199 			clear_opt (sbi->s_mount_opt, EXTENTS);
1200 			break;
1201 		default:
1202 			printk (KERN_ERR
1203 				"EXT4-fs: Unrecognized mount option \"%s\" "
1204 				"or missing value\n", p);
1205 			return 0;
1206 		}
1207 	}
1208 #ifdef CONFIG_QUOTA
1209 	if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1210 		if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1211 		     sbi->s_qf_names[USRQUOTA])
1212 			clear_opt(sbi->s_mount_opt, USRQUOTA);
1213 
1214 		if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1215 		     sbi->s_qf_names[GRPQUOTA])
1216 			clear_opt(sbi->s_mount_opt, GRPQUOTA);
1217 
1218 		if ((sbi->s_qf_names[USRQUOTA] &&
1219 				(sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1220 		    (sbi->s_qf_names[GRPQUOTA] &&
1221 				(sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1222 			printk(KERN_ERR "EXT4-fs: old and new quota "
1223 					"format mixing.\n");
1224 			return 0;
1225 		}
1226 
1227 		if (!sbi->s_jquota_fmt) {
1228 			printk(KERN_ERR "EXT4-fs: journalled quota format "
1229 					"not specified.\n");
1230 			return 0;
1231 		}
1232 	} else {
1233 		if (sbi->s_jquota_fmt) {
1234 			printk(KERN_ERR "EXT4-fs: journalled quota format "
1235 					"specified with no journalling "
1236 					"enabled.\n");
1237 			return 0;
1238 		}
1239 	}
1240 #endif
1241 	return 1;
1242 }
1243 
1244 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1245 			    int read_only)
1246 {
1247 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1248 	int res = 0;
1249 
1250 	if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1251 		printk (KERN_ERR "EXT4-fs warning: revision level too high, "
1252 			"forcing read-only mode\n");
1253 		res = MS_RDONLY;
1254 	}
1255 	if (read_only)
1256 		return res;
1257 	if (!(sbi->s_mount_state & EXT4_VALID_FS))
1258 		printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1259 			"running e2fsck is recommended\n");
1260 	else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1261 		printk (KERN_WARNING
1262 			"EXT4-fs warning: mounting fs with errors, "
1263 			"running e2fsck is recommended\n");
1264 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1265 		 le16_to_cpu(es->s_mnt_count) >=
1266 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1267 		printk (KERN_WARNING
1268 			"EXT4-fs warning: maximal mount count reached, "
1269 			"running e2fsck is recommended\n");
1270 	else if (le32_to_cpu(es->s_checkinterval) &&
1271 		(le32_to_cpu(es->s_lastcheck) +
1272 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1273 		printk (KERN_WARNING
1274 			"EXT4-fs warning: checktime reached, "
1275 			"running e2fsck is recommended\n");
1276 #if 0
1277 		/* @@@ We _will_ want to clear the valid bit if we find
1278 		 * inconsistencies, to force a fsck at reboot.  But for
1279 		 * a plain journaled filesystem we can keep it set as
1280 		 * valid forever! :)
1281 		 */
1282 	es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT4_VALID_FS);
1283 #endif
1284 	if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1285 		es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1286 	es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
1287 	es->s_mtime = cpu_to_le32(get_seconds());
1288 	ext4_update_dynamic_rev(sb);
1289 	EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1290 
1291 	ext4_commit_super(sb, es, 1);
1292 	if (test_opt(sb, DEBUG))
1293 		printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
1294 				"bpg=%lu, ipg=%lu, mo=%04lx]\n",
1295 			sb->s_blocksize,
1296 			sbi->s_groups_count,
1297 			EXT4_BLOCKS_PER_GROUP(sb),
1298 			EXT4_INODES_PER_GROUP(sb),
1299 			sbi->s_mount_opt);
1300 
1301 	printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1302 	if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
1303 		char b[BDEVNAME_SIZE];
1304 
1305 		printk("external journal on %s\n",
1306 			bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
1307 	} else {
1308 		printk("internal journal\n");
1309 	}
1310 	return res;
1311 }
1312 
1313 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1314 			    struct ext4_group_desc *gdp)
1315 {
1316 	__u16 crc = 0;
1317 
1318 	if (sbi->s_es->s_feature_ro_compat &
1319 	    cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1320 		int offset = offsetof(struct ext4_group_desc, bg_checksum);
1321 		__le32 le_group = cpu_to_le32(block_group);
1322 
1323 		crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1324 		crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1325 		crc = crc16(crc, (__u8 *)gdp, offset);
1326 		offset += sizeof(gdp->bg_checksum); /* skip checksum */
1327 		/* for checksum of struct ext4_group_desc do the rest...*/
1328 		if ((sbi->s_es->s_feature_incompat &
1329 		     cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1330 		    offset < le16_to_cpu(sbi->s_es->s_desc_size))
1331 			crc = crc16(crc, (__u8 *)gdp + offset,
1332 				    le16_to_cpu(sbi->s_es->s_desc_size) -
1333 					offset);
1334 	}
1335 
1336 	return cpu_to_le16(crc);
1337 }
1338 
1339 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1340 				struct ext4_group_desc *gdp)
1341 {
1342 	if ((sbi->s_es->s_feature_ro_compat &
1343 	     cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1344 	    (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1345 		return 0;
1346 
1347 	return 1;
1348 }
1349 
1350 /* Called at mount-time, super-block is locked */
1351 static int ext4_check_descriptors (struct super_block * sb)
1352 {
1353 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1354 	ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1355 	ext4_fsblk_t last_block;
1356 	ext4_fsblk_t block_bitmap;
1357 	ext4_fsblk_t inode_bitmap;
1358 	ext4_fsblk_t inode_table;
1359 	struct ext4_group_desc * gdp = NULL;
1360 	int desc_block = 0;
1361 	int flexbg_flag = 0;
1362 	int i;
1363 
1364 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1365 		flexbg_flag = 1;
1366 
1367 	ext4_debug ("Checking group descriptors");
1368 
1369 	for (i = 0; i < sbi->s_groups_count; i++)
1370 	{
1371 		if (i == sbi->s_groups_count - 1 || flexbg_flag)
1372 			last_block = ext4_blocks_count(sbi->s_es) - 1;
1373 		else
1374 			last_block = first_block +
1375 				(EXT4_BLOCKS_PER_GROUP(sb) - 1);
1376 
1377 		if ((i % EXT4_DESC_PER_BLOCK(sb)) == 0)
1378 			gdp = (struct ext4_group_desc *)
1379 					sbi->s_group_desc[desc_block++]->b_data;
1380 		block_bitmap = ext4_block_bitmap(sb, gdp);
1381 		if (block_bitmap < first_block || block_bitmap > last_block)
1382 		{
1383 			ext4_error (sb, "ext4_check_descriptors",
1384 				    "Block bitmap for group %d"
1385 				    " not in group (block %llu)!",
1386 				    i, block_bitmap);
1387 			return 0;
1388 		}
1389 		inode_bitmap = ext4_inode_bitmap(sb, gdp);
1390 		if (inode_bitmap < first_block || inode_bitmap > last_block)
1391 		{
1392 			ext4_error (sb, "ext4_check_descriptors",
1393 				    "Inode bitmap for group %d"
1394 				    " not in group (block %llu)!",
1395 				    i, inode_bitmap);
1396 			return 0;
1397 		}
1398 		inode_table = ext4_inode_table(sb, gdp);
1399 		if (inode_table < first_block ||
1400 		    inode_table + sbi->s_itb_per_group - 1 > last_block)
1401 		{
1402 			ext4_error (sb, "ext4_check_descriptors",
1403 				    "Inode table for group %d"
1404 				    " not in group (block %llu)!",
1405 				    i, inode_table);
1406 			return 0;
1407 		}
1408 		if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1409 			ext4_error(sb, __FUNCTION__,
1410 				   "Checksum for group %d failed (%u!=%u)\n", i,
1411 				   le16_to_cpu(ext4_group_desc_csum(sbi, i,
1412 								    gdp)),
1413 				   le16_to_cpu(gdp->bg_checksum));
1414 			return 0;
1415 		}
1416 		if (!flexbg_flag)
1417 			first_block += EXT4_BLOCKS_PER_GROUP(sb);
1418 		gdp = (struct ext4_group_desc *)
1419 			((__u8 *)gdp + EXT4_DESC_SIZE(sb));
1420 	}
1421 
1422 	ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1423 	sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb));
1424 	return 1;
1425 }
1426 
1427 
1428 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1429  * the superblock) which were deleted from all directories, but held open by
1430  * a process at the time of a crash.  We walk the list and try to delete these
1431  * inodes at recovery time (only with a read-write filesystem).
1432  *
1433  * In order to keep the orphan inode chain consistent during traversal (in
1434  * case of crash during recovery), we link each inode into the superblock
1435  * orphan list_head and handle it the same way as an inode deletion during
1436  * normal operation (which journals the operations for us).
1437  *
1438  * We only do an iget() and an iput() on each inode, which is very safe if we
1439  * accidentally point at an in-use or already deleted inode.  The worst that
1440  * can happen in this case is that we get a "bit already cleared" message from
1441  * ext4_free_inode().  The only reason we would point at a wrong inode is if
1442  * e2fsck was run on this filesystem, and it must have already done the orphan
1443  * inode cleanup for us, so we can safely abort without any further action.
1444  */
1445 static void ext4_orphan_cleanup (struct super_block * sb,
1446 				 struct ext4_super_block * es)
1447 {
1448 	unsigned int s_flags = sb->s_flags;
1449 	int nr_orphans = 0, nr_truncates = 0;
1450 #ifdef CONFIG_QUOTA
1451 	int i;
1452 #endif
1453 	if (!es->s_last_orphan) {
1454 		jbd_debug(4, "no orphan inodes to clean up\n");
1455 		return;
1456 	}
1457 
1458 	if (bdev_read_only(sb->s_bdev)) {
1459 		printk(KERN_ERR "EXT4-fs: write access "
1460 			"unavailable, skipping orphan cleanup.\n");
1461 		return;
1462 	}
1463 
1464 	if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1465 		if (es->s_last_orphan)
1466 			jbd_debug(1, "Errors on filesystem, "
1467 				  "clearing orphan list.\n");
1468 		es->s_last_orphan = 0;
1469 		jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1470 		return;
1471 	}
1472 
1473 	if (s_flags & MS_RDONLY) {
1474 		printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1475 		       sb->s_id);
1476 		sb->s_flags &= ~MS_RDONLY;
1477 	}
1478 #ifdef CONFIG_QUOTA
1479 	/* Needed for iput() to work correctly and not trash data */
1480 	sb->s_flags |= MS_ACTIVE;
1481 	/* Turn on quotas so that they are updated correctly */
1482 	for (i = 0; i < MAXQUOTAS; i++) {
1483 		if (EXT4_SB(sb)->s_qf_names[i]) {
1484 			int ret = ext4_quota_on_mount(sb, i);
1485 			if (ret < 0)
1486 				printk(KERN_ERR
1487 					"EXT4-fs: Cannot turn on journalled "
1488 					"quota: error %d\n", ret);
1489 		}
1490 	}
1491 #endif
1492 
1493 	while (es->s_last_orphan) {
1494 		struct inode *inode;
1495 
1496 		if (!(inode =
1497 		      ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
1498 			es->s_last_orphan = 0;
1499 			break;
1500 		}
1501 
1502 		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1503 		DQUOT_INIT(inode);
1504 		if (inode->i_nlink) {
1505 			printk(KERN_DEBUG
1506 				"%s: truncating inode %lu to %Ld bytes\n",
1507 				__FUNCTION__, inode->i_ino, inode->i_size);
1508 			jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1509 				  inode->i_ino, inode->i_size);
1510 			ext4_truncate(inode);
1511 			nr_truncates++;
1512 		} else {
1513 			printk(KERN_DEBUG
1514 				"%s: deleting unreferenced inode %lu\n",
1515 				__FUNCTION__, inode->i_ino);
1516 			jbd_debug(2, "deleting unreferenced inode %lu\n",
1517 				  inode->i_ino);
1518 			nr_orphans++;
1519 		}
1520 		iput(inode);  /* The delete magic happens here! */
1521 	}
1522 
1523 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1524 
1525 	if (nr_orphans)
1526 		printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1527 		       sb->s_id, PLURAL(nr_orphans));
1528 	if (nr_truncates)
1529 		printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1530 		       sb->s_id, PLURAL(nr_truncates));
1531 #ifdef CONFIG_QUOTA
1532 	/* Turn quotas off */
1533 	for (i = 0; i < MAXQUOTAS; i++) {
1534 		if (sb_dqopt(sb)->files[i])
1535 			vfs_quota_off(sb, i);
1536 	}
1537 #endif
1538 	sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1539 }
1540 
1541 /*
1542  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
1543  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1544  * We need to be 1 filesystem block less than the 2^32 sector limit.
1545  */
1546 static loff_t ext4_max_size(int bits)
1547 {
1548 	loff_t res = EXT4_NDIR_BLOCKS;
1549 	/* This constant is calculated to be the largest file size for a
1550 	 * dense, 4k-blocksize file such that the total number of
1551 	 * sectors in the file, including data and all indirect blocks,
1552 	 * does not exceed 2^32. */
1553 	const loff_t upper_limit = 0x1ff7fffd000LL;
1554 
1555 	res += 1LL << (bits-2);
1556 	res += 1LL << (2*(bits-2));
1557 	res += 1LL << (3*(bits-2));
1558 	res <<= bits;
1559 	if (res > upper_limit)
1560 		res = upper_limit;
1561 	return res;
1562 }
1563 
1564 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1565 				ext4_fsblk_t logical_sb_block, int nr)
1566 {
1567 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1568 	unsigned long bg, first_meta_bg;
1569 	int has_super = 0;
1570 
1571 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1572 
1573 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1574 	    nr < first_meta_bg)
1575 		return logical_sb_block + nr + 1;
1576 	bg = sbi->s_desc_per_block * nr;
1577 	if (ext4_bg_has_super(sb, bg))
1578 		has_super = 1;
1579 	return (has_super + ext4_group_first_block_no(sb, bg));
1580 }
1581 
1582 
1583 static int ext4_fill_super (struct super_block *sb, void *data, int silent)
1584 {
1585 	struct buffer_head * bh;
1586 	struct ext4_super_block *es = NULL;
1587 	struct ext4_sb_info *sbi;
1588 	ext4_fsblk_t block;
1589 	ext4_fsblk_t sb_block = get_sb_block(&data);
1590 	ext4_fsblk_t logical_sb_block;
1591 	unsigned long offset = 0;
1592 	unsigned int journal_inum = 0;
1593 	unsigned long journal_devnum = 0;
1594 	unsigned long def_mount_opts;
1595 	struct inode *root;
1596 	int blocksize;
1597 	int hblock;
1598 	int db_count;
1599 	int i;
1600 	int needs_recovery;
1601 	__le32 features;
1602 	__u64 blocks_count;
1603 	int err;
1604 
1605 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1606 	if (!sbi)
1607 		return -ENOMEM;
1608 	sb->s_fs_info = sbi;
1609 	sbi->s_mount_opt = 0;
1610 	sbi->s_resuid = EXT4_DEF_RESUID;
1611 	sbi->s_resgid = EXT4_DEF_RESGID;
1612 	sbi->s_sb_block = sb_block;
1613 
1614 	unlock_kernel();
1615 
1616 	blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
1617 	if (!blocksize) {
1618 		printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
1619 		goto out_fail;
1620 	}
1621 
1622 	/*
1623 	 * The ext4 superblock will not be buffer aligned for other than 1kB
1624 	 * block sizes.  We need to calculate the offset from buffer start.
1625 	 */
1626 	if (blocksize != EXT4_MIN_BLOCK_SIZE) {
1627 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1628 		offset = do_div(logical_sb_block, blocksize);
1629 	} else {
1630 		logical_sb_block = sb_block;
1631 	}
1632 
1633 	if (!(bh = sb_bread(sb, logical_sb_block))) {
1634 		printk (KERN_ERR "EXT4-fs: unable to read superblock\n");
1635 		goto out_fail;
1636 	}
1637 	/*
1638 	 * Note: s_es must be initialized as soon as possible because
1639 	 *       some ext4 macro-instructions depend on its value
1640 	 */
1641 	es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
1642 	sbi->s_es = es;
1643 	sb->s_magic = le16_to_cpu(es->s_magic);
1644 	if (sb->s_magic != EXT4_SUPER_MAGIC)
1645 		goto cantfind_ext4;
1646 
1647 	/* Set defaults before we parse the mount options */
1648 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1649 	if (def_mount_opts & EXT4_DEFM_DEBUG)
1650 		set_opt(sbi->s_mount_opt, DEBUG);
1651 	if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
1652 		set_opt(sbi->s_mount_opt, GRPID);
1653 	if (def_mount_opts & EXT4_DEFM_UID16)
1654 		set_opt(sbi->s_mount_opt, NO_UID32);
1655 #ifdef CONFIG_EXT4DEV_FS_XATTR
1656 	if (def_mount_opts & EXT4_DEFM_XATTR_USER)
1657 		set_opt(sbi->s_mount_opt, XATTR_USER);
1658 #endif
1659 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1660 	if (def_mount_opts & EXT4_DEFM_ACL)
1661 		set_opt(sbi->s_mount_opt, POSIX_ACL);
1662 #endif
1663 	if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1664 		sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1665 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1666 		sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1667 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1668 		sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1669 
1670 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
1671 		set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1672 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_RO)
1673 		set_opt(sbi->s_mount_opt, ERRORS_RO);
1674 	else
1675 		set_opt(sbi->s_mount_opt, ERRORS_CONT);
1676 
1677 	sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1678 	sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1679 
1680 	set_opt(sbi->s_mount_opt, RESERVATION);
1681 
1682 	/*
1683 	 * turn on extents feature by default in ext4 filesystem
1684 	 * User -o noextents to turn it off
1685 	 */
1686 	set_opt(sbi->s_mount_opt, EXTENTS);
1687 
1688 	if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1689 			    NULL, 0))
1690 		goto failed_mount;
1691 
1692 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1693 		((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1694 
1695 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1696 	    (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1697 	     EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1698 	     EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1699 		printk(KERN_WARNING
1700 		       "EXT4-fs warning: feature flags set on rev 0 fs, "
1701 		       "running e2fsck is recommended\n");
1702 	/*
1703 	 * Check feature flags regardless of the revision level, since we
1704 	 * previously didn't change the revision level when setting the flags,
1705 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
1706 	 */
1707 	features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
1708 	if (features) {
1709 		printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
1710 		       "unsupported optional features (%x).\n",
1711 		       sb->s_id, le32_to_cpu(features));
1712 		goto failed_mount;
1713 	}
1714 	features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
1715 	if (!(sb->s_flags & MS_RDONLY) && features) {
1716 		printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
1717 		       "unsupported optional features (%x).\n",
1718 		       sb->s_id, le32_to_cpu(features));
1719 		goto failed_mount;
1720 	}
1721 	blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1722 
1723 	if (blocksize < EXT4_MIN_BLOCK_SIZE ||
1724 	    blocksize > EXT4_MAX_BLOCK_SIZE) {
1725 		printk(KERN_ERR
1726 		       "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
1727 		       blocksize, sb->s_id);
1728 		goto failed_mount;
1729 	}
1730 
1731 	hblock = bdev_hardsect_size(sb->s_bdev);
1732 	if (sb->s_blocksize != blocksize) {
1733 		/*
1734 		 * Make sure the blocksize for the filesystem is larger
1735 		 * than the hardware sectorsize for the machine.
1736 		 */
1737 		if (blocksize < hblock) {
1738 			printk(KERN_ERR "EXT4-fs: blocksize %d too small for "
1739 			       "device blocksize %d.\n", blocksize, hblock);
1740 			goto failed_mount;
1741 		}
1742 
1743 		brelse (bh);
1744 		sb_set_blocksize(sb, blocksize);
1745 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1746 		offset = do_div(logical_sb_block, blocksize);
1747 		bh = sb_bread(sb, logical_sb_block);
1748 		if (!bh) {
1749 			printk(KERN_ERR
1750 			       "EXT4-fs: Can't read superblock on 2nd try.\n");
1751 			goto failed_mount;
1752 		}
1753 		es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
1754 		sbi->s_es = es;
1755 		if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
1756 			printk (KERN_ERR
1757 				"EXT4-fs: Magic mismatch, very weird !\n");
1758 			goto failed_mount;
1759 		}
1760 	}
1761 
1762 	sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
1763 
1764 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
1765 		sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
1766 		sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
1767 	} else {
1768 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1769 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1770 		if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1771 		    (!is_power_of_2(sbi->s_inode_size)) ||
1772 		    (sbi->s_inode_size > blocksize)) {
1773 			printk (KERN_ERR
1774 				"EXT4-fs: unsupported inode size: %d\n",
1775 				sbi->s_inode_size);
1776 			goto failed_mount;
1777 		}
1778 		if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
1779 			sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
1780 	}
1781 	sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
1782 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
1783 		if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
1784 		    sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
1785 		    !is_power_of_2(sbi->s_desc_size)) {
1786 			printk(KERN_ERR
1787 			       "EXT4-fs: unsupported descriptor size %lu\n",
1788 			       sbi->s_desc_size);
1789 			goto failed_mount;
1790 		}
1791 	} else
1792 		sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
1793 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1794 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1795 	if (EXT4_INODE_SIZE(sb) == 0)
1796 		goto cantfind_ext4;
1797 	sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
1798 	if (sbi->s_inodes_per_block == 0)
1799 		goto cantfind_ext4;
1800 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
1801 					sbi->s_inodes_per_block;
1802 	sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
1803 	sbi->s_sbh = bh;
1804 	sbi->s_mount_state = le16_to_cpu(es->s_state);
1805 	sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
1806 	sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
1807 	for (i=0; i < 4; i++)
1808 		sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1809 	sbi->s_def_hash_version = es->s_def_hash_version;
1810 
1811 	if (sbi->s_blocks_per_group > blocksize * 8) {
1812 		printk (KERN_ERR
1813 			"EXT4-fs: #blocks per group too big: %lu\n",
1814 			sbi->s_blocks_per_group);
1815 		goto failed_mount;
1816 	}
1817 	if (sbi->s_inodes_per_group > blocksize * 8) {
1818 		printk (KERN_ERR
1819 			"EXT4-fs: #inodes per group too big: %lu\n",
1820 			sbi->s_inodes_per_group);
1821 		goto failed_mount;
1822 	}
1823 
1824 	if (ext4_blocks_count(es) >
1825 		    (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1826 		printk(KERN_ERR "EXT4-fs: filesystem on %s:"
1827 			" too large to mount safely\n", sb->s_id);
1828 		if (sizeof(sector_t) < 8)
1829 			printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
1830 					"enabled\n");
1831 		goto failed_mount;
1832 	}
1833 
1834 	if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
1835 		goto cantfind_ext4;
1836 	blocks_count = (ext4_blocks_count(es) -
1837 			le32_to_cpu(es->s_first_data_block) +
1838 			EXT4_BLOCKS_PER_GROUP(sb) - 1);
1839 	do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
1840 	sbi->s_groups_count = blocks_count;
1841 	db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
1842 		   EXT4_DESC_PER_BLOCK(sb);
1843 	sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1844 				    GFP_KERNEL);
1845 	if (sbi->s_group_desc == NULL) {
1846 		printk (KERN_ERR "EXT4-fs: not enough memory\n");
1847 		goto failed_mount;
1848 	}
1849 
1850 	bgl_lock_init(&sbi->s_blockgroup_lock);
1851 
1852 	for (i = 0; i < db_count; i++) {
1853 		block = descriptor_loc(sb, logical_sb_block, i);
1854 		sbi->s_group_desc[i] = sb_bread(sb, block);
1855 		if (!sbi->s_group_desc[i]) {
1856 			printk (KERN_ERR "EXT4-fs: "
1857 				"can't read group descriptor %d\n", i);
1858 			db_count = i;
1859 			goto failed_mount2;
1860 		}
1861 	}
1862 	if (!ext4_check_descriptors (sb)) {
1863 		printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
1864 		goto failed_mount2;
1865 	}
1866 	sbi->s_gdb_count = db_count;
1867 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1868 	spin_lock_init(&sbi->s_next_gen_lock);
1869 
1870 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
1871 			ext4_count_free_blocks(sb));
1872 	if (!err) {
1873 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
1874 				ext4_count_free_inodes(sb));
1875 	}
1876 	if (!err) {
1877 		err = percpu_counter_init(&sbi->s_dirs_counter,
1878 				ext4_count_dirs(sb));
1879 	}
1880 	if (err) {
1881 		printk(KERN_ERR "EXT4-fs: insufficient memory\n");
1882 		goto failed_mount3;
1883 	}
1884 
1885 	/* per fileystem reservation list head & lock */
1886 	spin_lock_init(&sbi->s_rsv_window_lock);
1887 	sbi->s_rsv_window_root = RB_ROOT;
1888 	/* Add a single, static dummy reservation to the start of the
1889 	 * reservation window list --- it gives us a placeholder for
1890 	 * append-at-start-of-list which makes the allocation logic
1891 	 * _much_ simpler. */
1892 	sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
1893 	sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
1894 	sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1895 	sbi->s_rsv_window_head.rsv_goal_size = 0;
1896 	ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
1897 
1898 	/*
1899 	 * set up enough so that it can read an inode
1900 	 */
1901 	sb->s_op = &ext4_sops;
1902 	sb->s_export_op = &ext4_export_ops;
1903 	sb->s_xattr = ext4_xattr_handlers;
1904 #ifdef CONFIG_QUOTA
1905 	sb->s_qcop = &ext4_qctl_operations;
1906 	sb->dq_op = &ext4_quota_operations;
1907 #endif
1908 	INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1909 
1910 	sb->s_root = NULL;
1911 
1912 	needs_recovery = (es->s_last_orphan != 0 ||
1913 			  EXT4_HAS_INCOMPAT_FEATURE(sb,
1914 				    EXT4_FEATURE_INCOMPAT_RECOVER));
1915 
1916 	/*
1917 	 * The first inode we look at is the journal inode.  Don't try
1918 	 * root first: it may be modified in the journal!
1919 	 */
1920 	if (!test_opt(sb, NOLOAD) &&
1921 	    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
1922 		if (ext4_load_journal(sb, es, journal_devnum))
1923 			goto failed_mount3;
1924 	} else if (journal_inum) {
1925 		if (ext4_create_journal(sb, es, journal_inum))
1926 			goto failed_mount3;
1927 	} else {
1928 		if (!silent)
1929 			printk (KERN_ERR
1930 				"ext4: No journal on filesystem on %s\n",
1931 				sb->s_id);
1932 		goto failed_mount3;
1933 	}
1934 
1935 	if (ext4_blocks_count(es) > 0xffffffffULL &&
1936 	    !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
1937 				       JBD2_FEATURE_INCOMPAT_64BIT)) {
1938 		printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
1939 		goto failed_mount4;
1940 	}
1941 
1942 	/* We have now updated the journal if required, so we can
1943 	 * validate the data journaling mode. */
1944 	switch (test_opt(sb, DATA_FLAGS)) {
1945 	case 0:
1946 		/* No mode set, assume a default based on the journal
1947 		 * capabilities: ORDERED_DATA if the journal can
1948 		 * cope, else JOURNAL_DATA
1949 		 */
1950 		if (jbd2_journal_check_available_features
1951 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
1952 			set_opt(sbi->s_mount_opt, ORDERED_DATA);
1953 		else
1954 			set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1955 		break;
1956 
1957 	case EXT4_MOUNT_ORDERED_DATA:
1958 	case EXT4_MOUNT_WRITEBACK_DATA:
1959 		if (!jbd2_journal_check_available_features
1960 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
1961 			printk(KERN_ERR "EXT4-fs: Journal does not support "
1962 			       "requested data journaling mode\n");
1963 			goto failed_mount4;
1964 		}
1965 	default:
1966 		break;
1967 	}
1968 
1969 	if (test_opt(sb, NOBH)) {
1970 		if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
1971 			printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
1972 				"its supported only with writeback mode\n");
1973 			clear_opt(sbi->s_mount_opt, NOBH);
1974 		}
1975 	}
1976 	/*
1977 	 * The jbd2_journal_load will have done any necessary log recovery,
1978 	 * so we can safely mount the rest of the filesystem now.
1979 	 */
1980 
1981 	root = iget(sb, EXT4_ROOT_INO);
1982 	sb->s_root = d_alloc_root(root);
1983 	if (!sb->s_root) {
1984 		printk(KERN_ERR "EXT4-fs: get root inode failed\n");
1985 		iput(root);
1986 		goto failed_mount4;
1987 	}
1988 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1989 		dput(sb->s_root);
1990 		sb->s_root = NULL;
1991 		printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
1992 		goto failed_mount4;
1993 	}
1994 
1995 	ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1996 
1997 	/* determine the minimum size of new large inodes, if present */
1998 	if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
1999 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2000 						     EXT4_GOOD_OLD_INODE_SIZE;
2001 		if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2002 				       EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2003 			if (sbi->s_want_extra_isize <
2004 			    le16_to_cpu(es->s_want_extra_isize))
2005 				sbi->s_want_extra_isize =
2006 					le16_to_cpu(es->s_want_extra_isize);
2007 			if (sbi->s_want_extra_isize <
2008 			    le16_to_cpu(es->s_min_extra_isize))
2009 				sbi->s_want_extra_isize =
2010 					le16_to_cpu(es->s_min_extra_isize);
2011 		}
2012 	}
2013 	/* Check if enough inode space is available */
2014 	if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2015 							sbi->s_inode_size) {
2016 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2017 						       EXT4_GOOD_OLD_INODE_SIZE;
2018 		printk(KERN_INFO "EXT4-fs: required extra inode space not"
2019 			"available.\n");
2020 	}
2021 
2022 	/*
2023 	 * akpm: core read_super() calls in here with the superblock locked.
2024 	 * That deadlocks, because orphan cleanup needs to lock the superblock
2025 	 * in numerous places.  Here we just pop the lock - it's relatively
2026 	 * harmless, because we are now ready to accept write_super() requests,
2027 	 * and aviro says that's the only reason for hanging onto the
2028 	 * superblock lock.
2029 	 */
2030 	EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2031 	ext4_orphan_cleanup(sb, es);
2032 	EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2033 	if (needs_recovery)
2034 		printk (KERN_INFO "EXT4-fs: recovery complete.\n");
2035 	ext4_mark_recovery_complete(sb, es);
2036 	printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2037 		test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2038 		test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
2039 		"writeback");
2040 
2041 	ext4_ext_init(sb);
2042 
2043 	lock_kernel();
2044 	return 0;
2045 
2046 cantfind_ext4:
2047 	if (!silent)
2048 		printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2049 		       sb->s_id);
2050 	goto failed_mount;
2051 
2052 failed_mount4:
2053 	jbd2_journal_destroy(sbi->s_journal);
2054 failed_mount3:
2055 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
2056 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
2057 	percpu_counter_destroy(&sbi->s_dirs_counter);
2058 failed_mount2:
2059 	for (i = 0; i < db_count; i++)
2060 		brelse(sbi->s_group_desc[i]);
2061 	kfree(sbi->s_group_desc);
2062 failed_mount:
2063 #ifdef CONFIG_QUOTA
2064 	for (i = 0; i < MAXQUOTAS; i++)
2065 		kfree(sbi->s_qf_names[i]);
2066 #endif
2067 	ext4_blkdev_remove(sbi);
2068 	brelse(bh);
2069 out_fail:
2070 	sb->s_fs_info = NULL;
2071 	kfree(sbi);
2072 	lock_kernel();
2073 	return -EINVAL;
2074 }
2075 
2076 /*
2077  * Setup any per-fs journal parameters now.  We'll do this both on
2078  * initial mount, once the journal has been initialised but before we've
2079  * done any recovery; and again on any subsequent remount.
2080  */
2081 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2082 {
2083 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2084 
2085 	if (sbi->s_commit_interval)
2086 		journal->j_commit_interval = sbi->s_commit_interval;
2087 	/* We could also set up an ext4-specific default for the commit
2088 	 * interval here, but for now we'll just fall back to the jbd
2089 	 * default. */
2090 
2091 	spin_lock(&journal->j_state_lock);
2092 	if (test_opt(sb, BARRIER))
2093 		journal->j_flags |= JBD2_BARRIER;
2094 	else
2095 		journal->j_flags &= ~JBD2_BARRIER;
2096 	spin_unlock(&journal->j_state_lock);
2097 }
2098 
2099 static journal_t *ext4_get_journal(struct super_block *sb,
2100 				   unsigned int journal_inum)
2101 {
2102 	struct inode *journal_inode;
2103 	journal_t *journal;
2104 
2105 	/* First, test for the existence of a valid inode on disk.  Bad
2106 	 * things happen if we iget() an unused inode, as the subsequent
2107 	 * iput() will try to delete it. */
2108 
2109 	journal_inode = iget(sb, journal_inum);
2110 	if (!journal_inode) {
2111 		printk(KERN_ERR "EXT4-fs: no journal found.\n");
2112 		return NULL;
2113 	}
2114 	if (!journal_inode->i_nlink) {
2115 		make_bad_inode(journal_inode);
2116 		iput(journal_inode);
2117 		printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2118 		return NULL;
2119 	}
2120 
2121 	jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2122 		  journal_inode, journal_inode->i_size);
2123 	if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) {
2124 		printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2125 		iput(journal_inode);
2126 		return NULL;
2127 	}
2128 
2129 	journal = jbd2_journal_init_inode(journal_inode);
2130 	if (!journal) {
2131 		printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2132 		iput(journal_inode);
2133 		return NULL;
2134 	}
2135 	journal->j_private = sb;
2136 	ext4_init_journal_params(sb, journal);
2137 	return journal;
2138 }
2139 
2140 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2141 				       dev_t j_dev)
2142 {
2143 	struct buffer_head * bh;
2144 	journal_t *journal;
2145 	ext4_fsblk_t start;
2146 	ext4_fsblk_t len;
2147 	int hblock, blocksize;
2148 	ext4_fsblk_t sb_block;
2149 	unsigned long offset;
2150 	struct ext4_super_block * es;
2151 	struct block_device *bdev;
2152 
2153 	bdev = ext4_blkdev_get(j_dev);
2154 	if (bdev == NULL)
2155 		return NULL;
2156 
2157 	if (bd_claim(bdev, sb)) {
2158 		printk(KERN_ERR
2159 			"EXT4: failed to claim external journal device.\n");
2160 		blkdev_put(bdev);
2161 		return NULL;
2162 	}
2163 
2164 	blocksize = sb->s_blocksize;
2165 	hblock = bdev_hardsect_size(bdev);
2166 	if (blocksize < hblock) {
2167 		printk(KERN_ERR
2168 			"EXT4-fs: blocksize too small for journal device.\n");
2169 		goto out_bdev;
2170 	}
2171 
2172 	sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2173 	offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2174 	set_blocksize(bdev, blocksize);
2175 	if (!(bh = __bread(bdev, sb_block, blocksize))) {
2176 		printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2177 		       "external journal\n");
2178 		goto out_bdev;
2179 	}
2180 
2181 	es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2182 	if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2183 	    !(le32_to_cpu(es->s_feature_incompat) &
2184 	      EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2185 		printk(KERN_ERR "EXT4-fs: external journal has "
2186 					"bad superblock\n");
2187 		brelse(bh);
2188 		goto out_bdev;
2189 	}
2190 
2191 	if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2192 		printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2193 		brelse(bh);
2194 		goto out_bdev;
2195 	}
2196 
2197 	len = ext4_blocks_count(es);
2198 	start = sb_block + 1;
2199 	brelse(bh);	/* we're done with the superblock */
2200 
2201 	journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2202 					start, len, blocksize);
2203 	if (!journal) {
2204 		printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2205 		goto out_bdev;
2206 	}
2207 	journal->j_private = sb;
2208 	ll_rw_block(READ, 1, &journal->j_sb_buffer);
2209 	wait_on_buffer(journal->j_sb_buffer);
2210 	if (!buffer_uptodate(journal->j_sb_buffer)) {
2211 		printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2212 		goto out_journal;
2213 	}
2214 	if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2215 		printk(KERN_ERR "EXT4-fs: External journal has more than one "
2216 					"user (unsupported) - %d\n",
2217 			be32_to_cpu(journal->j_superblock->s_nr_users));
2218 		goto out_journal;
2219 	}
2220 	EXT4_SB(sb)->journal_bdev = bdev;
2221 	ext4_init_journal_params(sb, journal);
2222 	return journal;
2223 out_journal:
2224 	jbd2_journal_destroy(journal);
2225 out_bdev:
2226 	ext4_blkdev_put(bdev);
2227 	return NULL;
2228 }
2229 
2230 static int ext4_load_journal(struct super_block *sb,
2231 			     struct ext4_super_block *es,
2232 			     unsigned long journal_devnum)
2233 {
2234 	journal_t *journal;
2235 	unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2236 	dev_t journal_dev;
2237 	int err = 0;
2238 	int really_read_only;
2239 
2240 	if (journal_devnum &&
2241 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2242 		printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2243 			"numbers have changed\n");
2244 		journal_dev = new_decode_dev(journal_devnum);
2245 	} else
2246 		journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2247 
2248 	really_read_only = bdev_read_only(sb->s_bdev);
2249 
2250 	/*
2251 	 * Are we loading a blank journal or performing recovery after a
2252 	 * crash?  For recovery, we need to check in advance whether we
2253 	 * can get read-write access to the device.
2254 	 */
2255 
2256 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2257 		if (sb->s_flags & MS_RDONLY) {
2258 			printk(KERN_INFO "EXT4-fs: INFO: recovery "
2259 					"required on readonly filesystem.\n");
2260 			if (really_read_only) {
2261 				printk(KERN_ERR "EXT4-fs: write access "
2262 					"unavailable, cannot proceed.\n");
2263 				return -EROFS;
2264 			}
2265 			printk (KERN_INFO "EXT4-fs: write access will "
2266 					"be enabled during recovery.\n");
2267 		}
2268 	}
2269 
2270 	if (journal_inum && journal_dev) {
2271 		printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2272 		       "and inode journals!\n");
2273 		return -EINVAL;
2274 	}
2275 
2276 	if (journal_inum) {
2277 		if (!(journal = ext4_get_journal(sb, journal_inum)))
2278 			return -EINVAL;
2279 	} else {
2280 		if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2281 			return -EINVAL;
2282 	}
2283 
2284 	if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2285 		err = jbd2_journal_update_format(journal);
2286 		if (err)  {
2287 			printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2288 			jbd2_journal_destroy(journal);
2289 			return err;
2290 		}
2291 	}
2292 
2293 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2294 		err = jbd2_journal_wipe(journal, !really_read_only);
2295 	if (!err)
2296 		err = jbd2_journal_load(journal);
2297 
2298 	if (err) {
2299 		printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2300 		jbd2_journal_destroy(journal);
2301 		return err;
2302 	}
2303 
2304 	EXT4_SB(sb)->s_journal = journal;
2305 	ext4_clear_journal_err(sb, es);
2306 
2307 	if (journal_devnum &&
2308 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2309 		es->s_journal_dev = cpu_to_le32(journal_devnum);
2310 		sb->s_dirt = 1;
2311 
2312 		/* Make sure we flush the recovery flag to disk. */
2313 		ext4_commit_super(sb, es, 1);
2314 	}
2315 
2316 	return 0;
2317 }
2318 
2319 static int ext4_create_journal(struct super_block * sb,
2320 			       struct ext4_super_block * es,
2321 			       unsigned int journal_inum)
2322 {
2323 	journal_t *journal;
2324 	int err;
2325 
2326 	if (sb->s_flags & MS_RDONLY) {
2327 		printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
2328 				"create journal.\n");
2329 		return -EROFS;
2330 	}
2331 
2332 	journal = ext4_get_journal(sb, journal_inum);
2333 	if (!journal)
2334 		return -EINVAL;
2335 
2336 	printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
2337 	       journal_inum);
2338 
2339 	err = jbd2_journal_create(journal);
2340 	if (err) {
2341 		printk(KERN_ERR "EXT4-fs: error creating journal.\n");
2342 		jbd2_journal_destroy(journal);
2343 		return -EIO;
2344 	}
2345 
2346 	EXT4_SB(sb)->s_journal = journal;
2347 
2348 	ext4_update_dynamic_rev(sb);
2349 	EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2350 	EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
2351 
2352 	es->s_journal_inum = cpu_to_le32(journal_inum);
2353 	sb->s_dirt = 1;
2354 
2355 	/* Make sure we flush the recovery flag to disk. */
2356 	ext4_commit_super(sb, es, 1);
2357 
2358 	return 0;
2359 }
2360 
2361 static void ext4_commit_super (struct super_block * sb,
2362 			       struct ext4_super_block * es,
2363 			       int sync)
2364 {
2365 	struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2366 
2367 	if (!sbh)
2368 		return;
2369 	es->s_wtime = cpu_to_le32(get_seconds());
2370 	ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
2371 	es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
2372 	BUFFER_TRACE(sbh, "marking dirty");
2373 	mark_buffer_dirty(sbh);
2374 	if (sync)
2375 		sync_dirty_buffer(sbh);
2376 }
2377 
2378 
2379 /*
2380  * Have we just finished recovery?  If so, and if we are mounting (or
2381  * remounting) the filesystem readonly, then we will end up with a
2382  * consistent fs on disk.  Record that fact.
2383  */
2384 static void ext4_mark_recovery_complete(struct super_block * sb,
2385 					struct ext4_super_block * es)
2386 {
2387 	journal_t *journal = EXT4_SB(sb)->s_journal;
2388 
2389 	jbd2_journal_lock_updates(journal);
2390 	jbd2_journal_flush(journal);
2391 	lock_super(sb);
2392 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2393 	    sb->s_flags & MS_RDONLY) {
2394 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2395 		sb->s_dirt = 0;
2396 		ext4_commit_super(sb, es, 1);
2397 	}
2398 	unlock_super(sb);
2399 	jbd2_journal_unlock_updates(journal);
2400 }
2401 
2402 /*
2403  * If we are mounting (or read-write remounting) a filesystem whose journal
2404  * has recorded an error from a previous lifetime, move that error to the
2405  * main filesystem now.
2406  */
2407 static void ext4_clear_journal_err(struct super_block * sb,
2408 				   struct ext4_super_block * es)
2409 {
2410 	journal_t *journal;
2411 	int j_errno;
2412 	const char *errstr;
2413 
2414 	journal = EXT4_SB(sb)->s_journal;
2415 
2416 	/*
2417 	 * Now check for any error status which may have been recorded in the
2418 	 * journal by a prior ext4_error() or ext4_abort()
2419 	 */
2420 
2421 	j_errno = jbd2_journal_errno(journal);
2422 	if (j_errno) {
2423 		char nbuf[16];
2424 
2425 		errstr = ext4_decode_error(sb, j_errno, nbuf);
2426 		ext4_warning(sb, __FUNCTION__, "Filesystem error recorded "
2427 			     "from previous mount: %s", errstr);
2428 		ext4_warning(sb, __FUNCTION__, "Marking fs in need of "
2429 			     "filesystem check.");
2430 
2431 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2432 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2433 		ext4_commit_super (sb, es, 1);
2434 
2435 		jbd2_journal_clear_err(journal);
2436 	}
2437 }
2438 
2439 /*
2440  * Force the running and committing transactions to commit,
2441  * and wait on the commit.
2442  */
2443 int ext4_force_commit(struct super_block *sb)
2444 {
2445 	journal_t *journal;
2446 	int ret;
2447 
2448 	if (sb->s_flags & MS_RDONLY)
2449 		return 0;
2450 
2451 	journal = EXT4_SB(sb)->s_journal;
2452 	sb->s_dirt = 0;
2453 	ret = ext4_journal_force_commit(journal);
2454 	return ret;
2455 }
2456 
2457 /*
2458  * Ext4 always journals updates to the superblock itself, so we don't
2459  * have to propagate any other updates to the superblock on disk at this
2460  * point.  Just start an async writeback to get the buffers on their way
2461  * to the disk.
2462  *
2463  * This implicitly triggers the writebehind on sync().
2464  */
2465 
2466 static void ext4_write_super (struct super_block * sb)
2467 {
2468 	if (mutex_trylock(&sb->s_lock) != 0)
2469 		BUG();
2470 	sb->s_dirt = 0;
2471 }
2472 
2473 static int ext4_sync_fs(struct super_block *sb, int wait)
2474 {
2475 	tid_t target;
2476 
2477 	sb->s_dirt = 0;
2478 	if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
2479 		if (wait)
2480 			jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
2481 	}
2482 	return 0;
2483 }
2484 
2485 /*
2486  * LVM calls this function before a (read-only) snapshot is created.  This
2487  * gives us a chance to flush the journal completely and mark the fs clean.
2488  */
2489 static void ext4_write_super_lockfs(struct super_block *sb)
2490 {
2491 	sb->s_dirt = 0;
2492 
2493 	if (!(sb->s_flags & MS_RDONLY)) {
2494 		journal_t *journal = EXT4_SB(sb)->s_journal;
2495 
2496 		/* Now we set up the journal barrier. */
2497 		jbd2_journal_lock_updates(journal);
2498 		jbd2_journal_flush(journal);
2499 
2500 		/* Journal blocked and flushed, clear needs_recovery flag. */
2501 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2502 		ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2503 	}
2504 }
2505 
2506 /*
2507  * Called by LVM after the snapshot is done.  We need to reset the RECOVER
2508  * flag here, even though the filesystem is not technically dirty yet.
2509  */
2510 static void ext4_unlockfs(struct super_block *sb)
2511 {
2512 	if (!(sb->s_flags & MS_RDONLY)) {
2513 		lock_super(sb);
2514 		/* Reser the needs_recovery flag before the fs is unlocked. */
2515 		EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2516 		ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2517 		unlock_super(sb);
2518 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
2519 	}
2520 }
2521 
2522 static int ext4_remount (struct super_block * sb, int * flags, char * data)
2523 {
2524 	struct ext4_super_block * es;
2525 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2526 	ext4_fsblk_t n_blocks_count = 0;
2527 	unsigned long old_sb_flags;
2528 	struct ext4_mount_options old_opts;
2529 	int err;
2530 #ifdef CONFIG_QUOTA
2531 	int i;
2532 #endif
2533 
2534 	/* Store the original options */
2535 	old_sb_flags = sb->s_flags;
2536 	old_opts.s_mount_opt = sbi->s_mount_opt;
2537 	old_opts.s_resuid = sbi->s_resuid;
2538 	old_opts.s_resgid = sbi->s_resgid;
2539 	old_opts.s_commit_interval = sbi->s_commit_interval;
2540 #ifdef CONFIG_QUOTA
2541 	old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2542 	for (i = 0; i < MAXQUOTAS; i++)
2543 		old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2544 #endif
2545 
2546 	/*
2547 	 * Allow the "check" option to be passed as a remount option.
2548 	 */
2549 	if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2550 		err = -EINVAL;
2551 		goto restore_opts;
2552 	}
2553 
2554 	if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
2555 		ext4_abort(sb, __FUNCTION__, "Abort forced by user");
2556 
2557 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2558 		((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2559 
2560 	es = sbi->s_es;
2561 
2562 	ext4_init_journal_params(sb, sbi->s_journal);
2563 
2564 	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2565 		n_blocks_count > ext4_blocks_count(es)) {
2566 		if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
2567 			err = -EROFS;
2568 			goto restore_opts;
2569 		}
2570 
2571 		if (*flags & MS_RDONLY) {
2572 			/*
2573 			 * First of all, the unconditional stuff we have to do
2574 			 * to disable replay of the journal when we next remount
2575 			 */
2576 			sb->s_flags |= MS_RDONLY;
2577 
2578 			/*
2579 			 * OK, test if we are remounting a valid rw partition
2580 			 * readonly, and if so set the rdonly flag and then
2581 			 * mark the partition as valid again.
2582 			 */
2583 			if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
2584 			    (sbi->s_mount_state & EXT4_VALID_FS))
2585 				es->s_state = cpu_to_le16(sbi->s_mount_state);
2586 
2587 			/*
2588 			 * We have to unlock super so that we can wait for
2589 			 * transactions.
2590 			 */
2591 			unlock_super(sb);
2592 			ext4_mark_recovery_complete(sb, es);
2593 			lock_super(sb);
2594 		} else {
2595 			__le32 ret;
2596 			if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2597 					~EXT4_FEATURE_RO_COMPAT_SUPP))) {
2598 				printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2599 				       "remount RDWR because of unsupported "
2600 				       "optional features (%x).\n",
2601 				       sb->s_id, le32_to_cpu(ret));
2602 				err = -EROFS;
2603 				goto restore_opts;
2604 			}
2605 
2606 			/*
2607 			 * If we have an unprocessed orphan list hanging
2608 			 * around from a previously readonly bdev mount,
2609 			 * require a full umount/remount for now.
2610 			 */
2611 			if (es->s_last_orphan) {
2612 				printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2613 				       "remount RDWR because of unprocessed "
2614 				       "orphan inode list.  Please "
2615 				       "umount/remount instead.\n",
2616 				       sb->s_id);
2617 				err = -EINVAL;
2618 				goto restore_opts;
2619 			}
2620 
2621 			/*
2622 			 * Mounting a RDONLY partition read-write, so reread
2623 			 * and store the current valid flag.  (It may have
2624 			 * been changed by e2fsck since we originally mounted
2625 			 * the partition.)
2626 			 */
2627 			ext4_clear_journal_err(sb, es);
2628 			sbi->s_mount_state = le16_to_cpu(es->s_state);
2629 			if ((err = ext4_group_extend(sb, es, n_blocks_count)))
2630 				goto restore_opts;
2631 			if (!ext4_setup_super (sb, es, 0))
2632 				sb->s_flags &= ~MS_RDONLY;
2633 		}
2634 	}
2635 #ifdef CONFIG_QUOTA
2636 	/* Release old quota file names */
2637 	for (i = 0; i < MAXQUOTAS; i++)
2638 		if (old_opts.s_qf_names[i] &&
2639 		    old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2640 			kfree(old_opts.s_qf_names[i]);
2641 #endif
2642 	return 0;
2643 restore_opts:
2644 	sb->s_flags = old_sb_flags;
2645 	sbi->s_mount_opt = old_opts.s_mount_opt;
2646 	sbi->s_resuid = old_opts.s_resuid;
2647 	sbi->s_resgid = old_opts.s_resgid;
2648 	sbi->s_commit_interval = old_opts.s_commit_interval;
2649 #ifdef CONFIG_QUOTA
2650 	sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2651 	for (i = 0; i < MAXQUOTAS; i++) {
2652 		if (sbi->s_qf_names[i] &&
2653 		    old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2654 			kfree(sbi->s_qf_names[i]);
2655 		sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2656 	}
2657 #endif
2658 	return err;
2659 }
2660 
2661 static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf)
2662 {
2663 	struct super_block *sb = dentry->d_sb;
2664 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2665 	struct ext4_super_block *es = sbi->s_es;
2666 	u64 fsid;
2667 
2668 	if (test_opt(sb, MINIX_DF)) {
2669 		sbi->s_overhead_last = 0;
2670 	} else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
2671 		unsigned long ngroups = sbi->s_groups_count, i;
2672 		ext4_fsblk_t overhead = 0;
2673 		smp_rmb();
2674 
2675 		/*
2676 		 * Compute the overhead (FS structures).  This is constant
2677 		 * for a given filesystem unless the number of block groups
2678 		 * changes so we cache the previous value until it does.
2679 		 */
2680 
2681 		/*
2682 		 * All of the blocks before first_data_block are
2683 		 * overhead
2684 		 */
2685 		overhead = le32_to_cpu(es->s_first_data_block);
2686 
2687 		/*
2688 		 * Add the overhead attributed to the superblock and
2689 		 * block group descriptors.  If the sparse superblocks
2690 		 * feature is turned on, then not all groups have this.
2691 		 */
2692 		for (i = 0; i < ngroups; i++) {
2693 			overhead += ext4_bg_has_super(sb, i) +
2694 				ext4_bg_num_gdb(sb, i);
2695 			cond_resched();
2696 		}
2697 
2698 		/*
2699 		 * Every block group has an inode bitmap, a block
2700 		 * bitmap, and an inode table.
2701 		 */
2702 		overhead += ngroups * (2 + sbi->s_itb_per_group);
2703 		sbi->s_overhead_last = overhead;
2704 		smp_wmb();
2705 		sbi->s_blocks_last = ext4_blocks_count(es);
2706 	}
2707 
2708 	buf->f_type = EXT4_SUPER_MAGIC;
2709 	buf->f_bsize = sb->s_blocksize;
2710 	buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
2711 	buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2712 	ext4_free_blocks_count_set(es, buf->f_bfree);
2713 	buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
2714 	if (buf->f_bfree < ext4_r_blocks_count(es))
2715 		buf->f_bavail = 0;
2716 	buf->f_files = le32_to_cpu(es->s_inodes_count);
2717 	buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2718 	es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
2719 	buf->f_namelen = EXT4_NAME_LEN;
2720 	fsid = le64_to_cpup((void *)es->s_uuid) ^
2721 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2722 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2723 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2724 	return 0;
2725 }
2726 
2727 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2728  * is locked for write. Otherwise the are possible deadlocks:
2729  * Process 1                         Process 2
2730  * ext4_create()                     quota_sync()
2731  *   jbd2_journal_start()                   write_dquot()
2732  *   DQUOT_INIT()                        down(dqio_mutex)
2733  *     down(dqio_mutex)                    jbd2_journal_start()
2734  *
2735  */
2736 
2737 #ifdef CONFIG_QUOTA
2738 
2739 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2740 {
2741 	return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2742 }
2743 
2744 static int ext4_dquot_initialize(struct inode *inode, int type)
2745 {
2746 	handle_t *handle;
2747 	int ret, err;
2748 
2749 	/* We may create quota structure so we need to reserve enough blocks */
2750 	handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
2751 	if (IS_ERR(handle))
2752 		return PTR_ERR(handle);
2753 	ret = dquot_initialize(inode, type);
2754 	err = ext4_journal_stop(handle);
2755 	if (!ret)
2756 		ret = err;
2757 	return ret;
2758 }
2759 
2760 static int ext4_dquot_drop(struct inode *inode)
2761 {
2762 	handle_t *handle;
2763 	int ret, err;
2764 
2765 	/* We may delete quota structure so we need to reserve enough blocks */
2766 	handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
2767 	if (IS_ERR(handle))
2768 		return PTR_ERR(handle);
2769 	ret = dquot_drop(inode);
2770 	err = ext4_journal_stop(handle);
2771 	if (!ret)
2772 		ret = err;
2773 	return ret;
2774 }
2775 
2776 static int ext4_write_dquot(struct dquot *dquot)
2777 {
2778 	int ret, err;
2779 	handle_t *handle;
2780 	struct inode *inode;
2781 
2782 	inode = dquot_to_inode(dquot);
2783 	handle = ext4_journal_start(inode,
2784 					EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2785 	if (IS_ERR(handle))
2786 		return PTR_ERR(handle);
2787 	ret = dquot_commit(dquot);
2788 	err = ext4_journal_stop(handle);
2789 	if (!ret)
2790 		ret = err;
2791 	return ret;
2792 }
2793 
2794 static int ext4_acquire_dquot(struct dquot *dquot)
2795 {
2796 	int ret, err;
2797 	handle_t *handle;
2798 
2799 	handle = ext4_journal_start(dquot_to_inode(dquot),
2800 					EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2801 	if (IS_ERR(handle))
2802 		return PTR_ERR(handle);
2803 	ret = dquot_acquire(dquot);
2804 	err = ext4_journal_stop(handle);
2805 	if (!ret)
2806 		ret = err;
2807 	return ret;
2808 }
2809 
2810 static int ext4_release_dquot(struct dquot *dquot)
2811 {
2812 	int ret, err;
2813 	handle_t *handle;
2814 
2815 	handle = ext4_journal_start(dquot_to_inode(dquot),
2816 					EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2817 	if (IS_ERR(handle)) {
2818 		/* Release dquot anyway to avoid endless cycle in dqput() */
2819 		dquot_release(dquot);
2820 		return PTR_ERR(handle);
2821 	}
2822 	ret = dquot_release(dquot);
2823 	err = ext4_journal_stop(handle);
2824 	if (!ret)
2825 		ret = err;
2826 	return ret;
2827 }
2828 
2829 static int ext4_mark_dquot_dirty(struct dquot *dquot)
2830 {
2831 	/* Are we journalling quotas? */
2832 	if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2833 	    EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2834 		dquot_mark_dquot_dirty(dquot);
2835 		return ext4_write_dquot(dquot);
2836 	} else {
2837 		return dquot_mark_dquot_dirty(dquot);
2838 	}
2839 }
2840 
2841 static int ext4_write_info(struct super_block *sb, int type)
2842 {
2843 	int ret, err;
2844 	handle_t *handle;
2845 
2846 	/* Data block + inode block */
2847 	handle = ext4_journal_start(sb->s_root->d_inode, 2);
2848 	if (IS_ERR(handle))
2849 		return PTR_ERR(handle);
2850 	ret = dquot_commit_info(sb, type);
2851 	err = ext4_journal_stop(handle);
2852 	if (!ret)
2853 		ret = err;
2854 	return ret;
2855 }
2856 
2857 /*
2858  * Turn on quotas during mount time - we need to find
2859  * the quota file and such...
2860  */
2861 static int ext4_quota_on_mount(struct super_block *sb, int type)
2862 {
2863 	return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
2864 			EXT4_SB(sb)->s_jquota_fmt, type);
2865 }
2866 
2867 /*
2868  * Standard function to be called on quota_on
2869  */
2870 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
2871 			 char *path)
2872 {
2873 	int err;
2874 	struct nameidata nd;
2875 
2876 	if (!test_opt(sb, QUOTA))
2877 		return -EINVAL;
2878 	/* Not journalling quota? */
2879 	if (!EXT4_SB(sb)->s_qf_names[USRQUOTA] &&
2880 	    !EXT4_SB(sb)->s_qf_names[GRPQUOTA])
2881 		return vfs_quota_on(sb, type, format_id, path);
2882 	err = path_lookup(path, LOOKUP_FOLLOW, &nd);
2883 	if (err)
2884 		return err;
2885 	/* Quotafile not on the same filesystem? */
2886 	if (nd.mnt->mnt_sb != sb) {
2887 		path_release(&nd);
2888 		return -EXDEV;
2889 	}
2890 	/* Quotafile not of fs root? */
2891 	if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
2892 		printk(KERN_WARNING
2893 			"EXT4-fs: Quota file not on filesystem root. "
2894 			"Journalled quota will not work.\n");
2895 	path_release(&nd);
2896 	return vfs_quota_on(sb, type, format_id, path);
2897 }
2898 
2899 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2900  * acquiring the locks... As quota files are never truncated and quota code
2901  * itself serializes the operations (and noone else should touch the files)
2902  * we don't have to be afraid of races */
2903 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
2904 			       size_t len, loff_t off)
2905 {
2906 	struct inode *inode = sb_dqopt(sb)->files[type];
2907 	sector_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
2908 	int err = 0;
2909 	int offset = off & (sb->s_blocksize - 1);
2910 	int tocopy;
2911 	size_t toread;
2912 	struct buffer_head *bh;
2913 	loff_t i_size = i_size_read(inode);
2914 
2915 	if (off > i_size)
2916 		return 0;
2917 	if (off+len > i_size)
2918 		len = i_size-off;
2919 	toread = len;
2920 	while (toread > 0) {
2921 		tocopy = sb->s_blocksize - offset < toread ?
2922 				sb->s_blocksize - offset : toread;
2923 		bh = ext4_bread(NULL, inode, blk, 0, &err);
2924 		if (err)
2925 			return err;
2926 		if (!bh)	/* A hole? */
2927 			memset(data, 0, tocopy);
2928 		else
2929 			memcpy(data, bh->b_data+offset, tocopy);
2930 		brelse(bh);
2931 		offset = 0;
2932 		toread -= tocopy;
2933 		data += tocopy;
2934 		blk++;
2935 	}
2936 	return len;
2937 }
2938 
2939 /* Write to quotafile (we know the transaction is already started and has
2940  * enough credits) */
2941 static ssize_t ext4_quota_write(struct super_block *sb, int type,
2942 				const char *data, size_t len, loff_t off)
2943 {
2944 	struct inode *inode = sb_dqopt(sb)->files[type];
2945 	sector_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
2946 	int err = 0;
2947 	int offset = off & (sb->s_blocksize - 1);
2948 	int tocopy;
2949 	int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
2950 	size_t towrite = len;
2951 	struct buffer_head *bh;
2952 	handle_t *handle = journal_current_handle();
2953 
2954 	if (!handle) {
2955 		printk(KERN_WARNING "EXT4-fs: Quota write (off=%Lu, len=%Lu)"
2956 			" cancelled because transaction is not started.\n",
2957 			(unsigned long long)off, (unsigned long long)len);
2958 		return -EIO;
2959 	}
2960 	mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2961 	while (towrite > 0) {
2962 		tocopy = sb->s_blocksize - offset < towrite ?
2963 				sb->s_blocksize - offset : towrite;
2964 		bh = ext4_bread(handle, inode, blk, 1, &err);
2965 		if (!bh)
2966 			goto out;
2967 		if (journal_quota) {
2968 			err = ext4_journal_get_write_access(handle, bh);
2969 			if (err) {
2970 				brelse(bh);
2971 				goto out;
2972 			}
2973 		}
2974 		lock_buffer(bh);
2975 		memcpy(bh->b_data+offset, data, tocopy);
2976 		flush_dcache_page(bh->b_page);
2977 		unlock_buffer(bh);
2978 		if (journal_quota)
2979 			err = ext4_journal_dirty_metadata(handle, bh);
2980 		else {
2981 			/* Always do at least ordered writes for quotas */
2982 			err = ext4_journal_dirty_data(handle, bh);
2983 			mark_buffer_dirty(bh);
2984 		}
2985 		brelse(bh);
2986 		if (err)
2987 			goto out;
2988 		offset = 0;
2989 		towrite -= tocopy;
2990 		data += tocopy;
2991 		blk++;
2992 	}
2993 out:
2994 	if (len == towrite)
2995 		return err;
2996 	if (inode->i_size < off+len-towrite) {
2997 		i_size_write(inode, off+len-towrite);
2998 		EXT4_I(inode)->i_disksize = inode->i_size;
2999 	}
3000 	inode->i_version++;
3001 	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3002 	ext4_mark_inode_dirty(handle, inode);
3003 	mutex_unlock(&inode->i_mutex);
3004 	return len - towrite;
3005 }
3006 
3007 #endif
3008 
3009 static int ext4_get_sb(struct file_system_type *fs_type,
3010 	int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3011 {
3012 	return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3013 }
3014 
3015 static struct file_system_type ext4dev_fs_type = {
3016 	.owner		= THIS_MODULE,
3017 	.name		= "ext4dev",
3018 	.get_sb		= ext4_get_sb,
3019 	.kill_sb	= kill_block_super,
3020 	.fs_flags	= FS_REQUIRES_DEV,
3021 };
3022 
3023 static int __init init_ext4_fs(void)
3024 {
3025 	int err = init_ext4_xattr();
3026 	if (err)
3027 		return err;
3028 	err = init_inodecache();
3029 	if (err)
3030 		goto out1;
3031 	err = register_filesystem(&ext4dev_fs_type);
3032 	if (err)
3033 		goto out;
3034 	return 0;
3035 out:
3036 	destroy_inodecache();
3037 out1:
3038 	exit_ext4_xattr();
3039 	return err;
3040 }
3041 
3042 static void __exit exit_ext4_fs(void)
3043 {
3044 	unregister_filesystem(&ext4dev_fs_type);
3045 	destroy_inodecache();
3046 	exit_ext4_xattr();
3047 }
3048 
3049 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3050 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
3051 MODULE_LICENSE("GPL");
3052 module_init(init_ext4_fs)
3053 module_exit(exit_ext4_fs)
3054