xref: /openbmc/linux/fs/ext4/super.c (revision 9ac8d3fb)
1 /*
2  *  linux/fs/ext4/super.c
3  *
4  * Copyright (C) 1992, 1993, 1994, 1995
5  * Remy Card (card@masi.ibp.fr)
6  * Laboratoire MASI - Institut Blaise Pascal
7  * Universite Pierre et Marie Curie (Paris VI)
8  *
9  *  from
10  *
11  *  linux/fs/minix/inode.c
12  *
13  *  Copyright (C) 1991, 1992  Linus Torvalds
14  *
15  *  Big-endian to little-endian byte-swapping/bitmaps by
16  *        David S. Miller (davem@caip.rutgers.edu), 1995
17  */
18 
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/time.h>
23 #include <linux/jbd2.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/parser.h>
28 #include <linux/smp_lock.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/marker.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <asm/uaccess.h>
42 
43 #include "ext4.h"
44 #include "ext4_jbd2.h"
45 #include "xattr.h"
46 #include "acl.h"
47 #include "namei.h"
48 #include "group.h"
49 
50 struct proc_dir_entry *ext4_proc_root;
51 
52 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
53 			     unsigned long journal_devnum);
54 static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
55 			       unsigned int);
56 static void ext4_commit_super(struct super_block *sb,
57 			      struct ext4_super_block *es, int sync);
58 static void ext4_mark_recovery_complete(struct super_block *sb,
59 					struct ext4_super_block *es);
60 static void ext4_clear_journal_err(struct super_block *sb,
61 				   struct ext4_super_block *es);
62 static int ext4_sync_fs(struct super_block *sb, int wait);
63 static const char *ext4_decode_error(struct super_block *sb, int errno,
64 				     char nbuf[16]);
65 static int ext4_remount(struct super_block *sb, int *flags, char *data);
66 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
67 static void ext4_unlockfs(struct super_block *sb);
68 static void ext4_write_super(struct super_block *sb);
69 static void ext4_write_super_lockfs(struct super_block *sb);
70 
71 
72 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
73 			       struct ext4_group_desc *bg)
74 {
75 	return le32_to_cpu(bg->bg_block_bitmap_lo) |
76 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
77 		(ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
78 }
79 
80 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
81 			       struct ext4_group_desc *bg)
82 {
83 	return le32_to_cpu(bg->bg_inode_bitmap_lo) |
84 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
85 		(ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
86 }
87 
88 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
89 			      struct ext4_group_desc *bg)
90 {
91 	return le32_to_cpu(bg->bg_inode_table_lo) |
92 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
93 		(ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
94 }
95 
96 void ext4_block_bitmap_set(struct super_block *sb,
97 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
98 {
99 	bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
100 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
101 		bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
102 }
103 
104 void ext4_inode_bitmap_set(struct super_block *sb,
105 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
106 {
107 	bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
108 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
109 		bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
110 }
111 
112 void ext4_inode_table_set(struct super_block *sb,
113 			  struct ext4_group_desc *bg, ext4_fsblk_t blk)
114 {
115 	bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
116 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
117 		bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
118 }
119 
120 /*
121  * Wrappers for jbd2_journal_start/end.
122  *
123  * The only special thing we need to do here is to make sure that all
124  * journal_end calls result in the superblock being marked dirty, so
125  * that sync() will call the filesystem's write_super callback if
126  * appropriate.
127  */
128 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
129 {
130 	journal_t *journal;
131 
132 	if (sb->s_flags & MS_RDONLY)
133 		return ERR_PTR(-EROFS);
134 
135 	/* Special case here: if the journal has aborted behind our
136 	 * backs (eg. EIO in the commit thread), then we still need to
137 	 * take the FS itself readonly cleanly. */
138 	journal = EXT4_SB(sb)->s_journal;
139 	if (is_journal_aborted(journal)) {
140 		ext4_abort(sb, __func__,
141 			   "Detected aborted journal");
142 		return ERR_PTR(-EROFS);
143 	}
144 
145 	return jbd2_journal_start(journal, nblocks);
146 }
147 
148 /*
149  * The only special thing we need to do here is to make sure that all
150  * jbd2_journal_stop calls result in the superblock being marked dirty, so
151  * that sync() will call the filesystem's write_super callback if
152  * appropriate.
153  */
154 int __ext4_journal_stop(const char *where, handle_t *handle)
155 {
156 	struct super_block *sb;
157 	int err;
158 	int rc;
159 
160 	sb = handle->h_transaction->t_journal->j_private;
161 	err = handle->h_err;
162 	rc = jbd2_journal_stop(handle);
163 
164 	if (!err)
165 		err = rc;
166 	if (err)
167 		__ext4_std_error(sb, where, err);
168 	return err;
169 }
170 
171 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
172 		struct buffer_head *bh, handle_t *handle, int err)
173 {
174 	char nbuf[16];
175 	const char *errstr = ext4_decode_error(NULL, err, nbuf);
176 
177 	if (bh)
178 		BUFFER_TRACE(bh, "abort");
179 
180 	if (!handle->h_err)
181 		handle->h_err = err;
182 
183 	if (is_handle_aborted(handle))
184 		return;
185 
186 	printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
187 	       caller, errstr, err_fn);
188 
189 	jbd2_journal_abort_handle(handle);
190 }
191 
192 /* Deal with the reporting of failure conditions on a filesystem such as
193  * inconsistencies detected or read IO failures.
194  *
195  * On ext2, we can store the error state of the filesystem in the
196  * superblock.  That is not possible on ext4, because we may have other
197  * write ordering constraints on the superblock which prevent us from
198  * writing it out straight away; and given that the journal is about to
199  * be aborted, we can't rely on the current, or future, transactions to
200  * write out the superblock safely.
201  *
202  * We'll just use the jbd2_journal_abort() error code to record an error in
203  * the journal instead.  On recovery, the journal will compain about
204  * that error until we've noted it down and cleared it.
205  */
206 
207 static void ext4_handle_error(struct super_block *sb)
208 {
209 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
210 
211 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
212 	es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
213 
214 	if (sb->s_flags & MS_RDONLY)
215 		return;
216 
217 	if (!test_opt(sb, ERRORS_CONT)) {
218 		journal_t *journal = EXT4_SB(sb)->s_journal;
219 
220 		EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
221 		if (journal)
222 			jbd2_journal_abort(journal, -EIO);
223 	}
224 	if (test_opt(sb, ERRORS_RO)) {
225 		printk(KERN_CRIT "Remounting filesystem read-only\n");
226 		sb->s_flags |= MS_RDONLY;
227 	}
228 	ext4_commit_super(sb, es, 1);
229 	if (test_opt(sb, ERRORS_PANIC))
230 		panic("EXT4-fs (device %s): panic forced after error\n",
231 			sb->s_id);
232 }
233 
234 void ext4_error(struct super_block *sb, const char *function,
235 		const char *fmt, ...)
236 {
237 	va_list args;
238 
239 	va_start(args, fmt);
240 	printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
241 	vprintk(fmt, args);
242 	printk("\n");
243 	va_end(args);
244 
245 	ext4_handle_error(sb);
246 }
247 
248 static const char *ext4_decode_error(struct super_block *sb, int errno,
249 				     char nbuf[16])
250 {
251 	char *errstr = NULL;
252 
253 	switch (errno) {
254 	case -EIO:
255 		errstr = "IO failure";
256 		break;
257 	case -ENOMEM:
258 		errstr = "Out of memory";
259 		break;
260 	case -EROFS:
261 		if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
262 			errstr = "Journal has aborted";
263 		else
264 			errstr = "Readonly filesystem";
265 		break;
266 	default:
267 		/* If the caller passed in an extra buffer for unknown
268 		 * errors, textualise them now.  Else we just return
269 		 * NULL. */
270 		if (nbuf) {
271 			/* Check for truncated error codes... */
272 			if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
273 				errstr = nbuf;
274 		}
275 		break;
276 	}
277 
278 	return errstr;
279 }
280 
281 /* __ext4_std_error decodes expected errors from journaling functions
282  * automatically and invokes the appropriate error response.  */
283 
284 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
285 {
286 	char nbuf[16];
287 	const char *errstr;
288 
289 	/* Special case: if the error is EROFS, and we're not already
290 	 * inside a transaction, then there's really no point in logging
291 	 * an error. */
292 	if (errno == -EROFS && journal_current_handle() == NULL &&
293 	    (sb->s_flags & MS_RDONLY))
294 		return;
295 
296 	errstr = ext4_decode_error(sb, errno, nbuf);
297 	printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
298 	       sb->s_id, function, errstr);
299 
300 	ext4_handle_error(sb);
301 }
302 
303 /*
304  * ext4_abort is a much stronger failure handler than ext4_error.  The
305  * abort function may be used to deal with unrecoverable failures such
306  * as journal IO errors or ENOMEM at a critical moment in log management.
307  *
308  * We unconditionally force the filesystem into an ABORT|READONLY state,
309  * unless the error response on the fs has been set to panic in which
310  * case we take the easy way out and panic immediately.
311  */
312 
313 void ext4_abort(struct super_block *sb, const char *function,
314 		const char *fmt, ...)
315 {
316 	va_list args;
317 
318 	printk(KERN_CRIT "ext4_abort called.\n");
319 
320 	va_start(args, fmt);
321 	printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
322 	vprintk(fmt, args);
323 	printk("\n");
324 	va_end(args);
325 
326 	if (test_opt(sb, ERRORS_PANIC))
327 		panic("EXT4-fs panic from previous error\n");
328 
329 	if (sb->s_flags & MS_RDONLY)
330 		return;
331 
332 	printk(KERN_CRIT "Remounting filesystem read-only\n");
333 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
334 	sb->s_flags |= MS_RDONLY;
335 	EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
336 	if (EXT4_SB(sb)->s_journal)
337 		jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
338 }
339 
340 void ext4_warning(struct super_block *sb, const char *function,
341 		  const char *fmt, ...)
342 {
343 	va_list args;
344 
345 	va_start(args, fmt);
346 	printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
347 	       sb->s_id, function);
348 	vprintk(fmt, args);
349 	printk("\n");
350 	va_end(args);
351 }
352 
353 void ext4_update_dynamic_rev(struct super_block *sb)
354 {
355 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
356 
357 	if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
358 		return;
359 
360 	ext4_warning(sb, __func__,
361 		     "updating to rev %d because of new feature flag, "
362 		     "running e2fsck is recommended",
363 		     EXT4_DYNAMIC_REV);
364 
365 	es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
366 	es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
367 	es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
368 	/* leave es->s_feature_*compat flags alone */
369 	/* es->s_uuid will be set by e2fsck if empty */
370 
371 	/*
372 	 * The rest of the superblock fields should be zero, and if not it
373 	 * means they are likely already in use, so leave them alone.  We
374 	 * can leave it up to e2fsck to clean up any inconsistencies there.
375 	 */
376 }
377 
378 /*
379  * Open the external journal device
380  */
381 static struct block_device *ext4_blkdev_get(dev_t dev)
382 {
383 	struct block_device *bdev;
384 	char b[BDEVNAME_SIZE];
385 
386 	bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
387 	if (IS_ERR(bdev))
388 		goto fail;
389 	return bdev;
390 
391 fail:
392 	printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
393 			__bdevname(dev, b), PTR_ERR(bdev));
394 	return NULL;
395 }
396 
397 /*
398  * Release the journal device
399  */
400 static int ext4_blkdev_put(struct block_device *bdev)
401 {
402 	bd_release(bdev);
403 	return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
404 }
405 
406 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
407 {
408 	struct block_device *bdev;
409 	int ret = -ENODEV;
410 
411 	bdev = sbi->journal_bdev;
412 	if (bdev) {
413 		ret = ext4_blkdev_put(bdev);
414 		sbi->journal_bdev = NULL;
415 	}
416 	return ret;
417 }
418 
419 static inline struct inode *orphan_list_entry(struct list_head *l)
420 {
421 	return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
422 }
423 
424 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
425 {
426 	struct list_head *l;
427 
428 	printk(KERN_ERR "sb orphan head is %d\n",
429 	       le32_to_cpu(sbi->s_es->s_last_orphan));
430 
431 	printk(KERN_ERR "sb_info orphan list:\n");
432 	list_for_each(l, &sbi->s_orphan) {
433 		struct inode *inode = orphan_list_entry(l);
434 		printk(KERN_ERR "  "
435 		       "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
436 		       inode->i_sb->s_id, inode->i_ino, inode,
437 		       inode->i_mode, inode->i_nlink,
438 		       NEXT_ORPHAN(inode));
439 	}
440 }
441 
442 static void ext4_put_super(struct super_block *sb)
443 {
444 	struct ext4_sb_info *sbi = EXT4_SB(sb);
445 	struct ext4_super_block *es = sbi->s_es;
446 	int i, err;
447 
448 	ext4_mb_release(sb);
449 	ext4_ext_release(sb);
450 	ext4_xattr_put_super(sb);
451 	err = jbd2_journal_destroy(sbi->s_journal);
452 	sbi->s_journal = NULL;
453 	if (err < 0)
454 		ext4_abort(sb, __func__, "Couldn't clean up the journal");
455 
456 	if (!(sb->s_flags & MS_RDONLY)) {
457 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
458 		es->s_state = cpu_to_le16(sbi->s_mount_state);
459 		ext4_commit_super(sb, es, 1);
460 	}
461 	if (sbi->s_proc) {
462 		remove_proc_entry("inode_readahead_blks", sbi->s_proc);
463 		remove_proc_entry(sb->s_id, ext4_proc_root);
464 	}
465 
466 	for (i = 0; i < sbi->s_gdb_count; i++)
467 		brelse(sbi->s_group_desc[i]);
468 	kfree(sbi->s_group_desc);
469 	kfree(sbi->s_flex_groups);
470 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
471 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
472 	percpu_counter_destroy(&sbi->s_dirs_counter);
473 	percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
474 	brelse(sbi->s_sbh);
475 #ifdef CONFIG_QUOTA
476 	for (i = 0; i < MAXQUOTAS; i++)
477 		kfree(sbi->s_qf_names[i]);
478 #endif
479 
480 	/* Debugging code just in case the in-memory inode orphan list
481 	 * isn't empty.  The on-disk one can be non-empty if we've
482 	 * detected an error and taken the fs readonly, but the
483 	 * in-memory list had better be clean by this point. */
484 	if (!list_empty(&sbi->s_orphan))
485 		dump_orphan_list(sb, sbi);
486 	J_ASSERT(list_empty(&sbi->s_orphan));
487 
488 	invalidate_bdev(sb->s_bdev);
489 	if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
490 		/*
491 		 * Invalidate the journal device's buffers.  We don't want them
492 		 * floating about in memory - the physical journal device may
493 		 * hotswapped, and it breaks the `ro-after' testing code.
494 		 */
495 		sync_blockdev(sbi->journal_bdev);
496 		invalidate_bdev(sbi->journal_bdev);
497 		ext4_blkdev_remove(sbi);
498 	}
499 	sb->s_fs_info = NULL;
500 	kfree(sbi);
501 	return;
502 }
503 
504 static struct kmem_cache *ext4_inode_cachep;
505 
506 /*
507  * Called inside transaction, so use GFP_NOFS
508  */
509 static struct inode *ext4_alloc_inode(struct super_block *sb)
510 {
511 	struct ext4_inode_info *ei;
512 
513 	ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
514 	if (!ei)
515 		return NULL;
516 #ifdef CONFIG_EXT4_FS_POSIX_ACL
517 	ei->i_acl = EXT4_ACL_NOT_CACHED;
518 	ei->i_default_acl = EXT4_ACL_NOT_CACHED;
519 #endif
520 	ei->vfs_inode.i_version = 1;
521 	ei->vfs_inode.i_data.writeback_index = 0;
522 	memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
523 	INIT_LIST_HEAD(&ei->i_prealloc_list);
524 	spin_lock_init(&ei->i_prealloc_lock);
525 	jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
526 	ei->i_reserved_data_blocks = 0;
527 	ei->i_reserved_meta_blocks = 0;
528 	ei->i_allocated_meta_blocks = 0;
529 	ei->i_delalloc_reserved_flag = 0;
530 	spin_lock_init(&(ei->i_block_reservation_lock));
531 	return &ei->vfs_inode;
532 }
533 
534 static void ext4_destroy_inode(struct inode *inode)
535 {
536 	if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
537 		printk("EXT4 Inode %p: orphan list check failed!\n",
538 			EXT4_I(inode));
539 		print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
540 				EXT4_I(inode), sizeof(struct ext4_inode_info),
541 				true);
542 		dump_stack();
543 	}
544 	kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
545 }
546 
547 static void init_once(void *foo)
548 {
549 	struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
550 
551 	INIT_LIST_HEAD(&ei->i_orphan);
552 #ifdef CONFIG_EXT4_FS_XATTR
553 	init_rwsem(&ei->xattr_sem);
554 #endif
555 	init_rwsem(&ei->i_data_sem);
556 	inode_init_once(&ei->vfs_inode);
557 }
558 
559 static int init_inodecache(void)
560 {
561 	ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
562 					     sizeof(struct ext4_inode_info),
563 					     0, (SLAB_RECLAIM_ACCOUNT|
564 						SLAB_MEM_SPREAD),
565 					     init_once);
566 	if (ext4_inode_cachep == NULL)
567 		return -ENOMEM;
568 	return 0;
569 }
570 
571 static void destroy_inodecache(void)
572 {
573 	kmem_cache_destroy(ext4_inode_cachep);
574 }
575 
576 static void ext4_clear_inode(struct inode *inode)
577 {
578 #ifdef CONFIG_EXT4_FS_POSIX_ACL
579 	if (EXT4_I(inode)->i_acl &&
580 			EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
581 		posix_acl_release(EXT4_I(inode)->i_acl);
582 		EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
583 	}
584 	if (EXT4_I(inode)->i_default_acl &&
585 			EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
586 		posix_acl_release(EXT4_I(inode)->i_default_acl);
587 		EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
588 	}
589 #endif
590 	ext4_discard_preallocations(inode);
591 	jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
592 				       &EXT4_I(inode)->jinode);
593 }
594 
595 static inline void ext4_show_quota_options(struct seq_file *seq,
596 					   struct super_block *sb)
597 {
598 #if defined(CONFIG_QUOTA)
599 	struct ext4_sb_info *sbi = EXT4_SB(sb);
600 
601 	if (sbi->s_jquota_fmt)
602 		seq_printf(seq, ",jqfmt=%s",
603 		(sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
604 
605 	if (sbi->s_qf_names[USRQUOTA])
606 		seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
607 
608 	if (sbi->s_qf_names[GRPQUOTA])
609 		seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
610 
611 	if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
612 		seq_puts(seq, ",usrquota");
613 
614 	if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
615 		seq_puts(seq, ",grpquota");
616 #endif
617 }
618 
619 /*
620  * Show an option if
621  *  - it's set to a non-default value OR
622  *  - if the per-sb default is different from the global default
623  */
624 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
625 {
626 	int def_errors;
627 	unsigned long def_mount_opts;
628 	struct super_block *sb = vfs->mnt_sb;
629 	struct ext4_sb_info *sbi = EXT4_SB(sb);
630 	struct ext4_super_block *es = sbi->s_es;
631 
632 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
633 	def_errors     = le16_to_cpu(es->s_errors);
634 
635 	if (sbi->s_sb_block != 1)
636 		seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
637 	if (test_opt(sb, MINIX_DF))
638 		seq_puts(seq, ",minixdf");
639 	if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
640 		seq_puts(seq, ",grpid");
641 	if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
642 		seq_puts(seq, ",nogrpid");
643 	if (sbi->s_resuid != EXT4_DEF_RESUID ||
644 	    le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
645 		seq_printf(seq, ",resuid=%u", sbi->s_resuid);
646 	}
647 	if (sbi->s_resgid != EXT4_DEF_RESGID ||
648 	    le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
649 		seq_printf(seq, ",resgid=%u", sbi->s_resgid);
650 	}
651 	if (test_opt(sb, ERRORS_RO)) {
652 		if (def_errors == EXT4_ERRORS_PANIC ||
653 		    def_errors == EXT4_ERRORS_CONTINUE) {
654 			seq_puts(seq, ",errors=remount-ro");
655 		}
656 	}
657 	if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
658 		seq_puts(seq, ",errors=continue");
659 	if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
660 		seq_puts(seq, ",errors=panic");
661 	if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
662 		seq_puts(seq, ",nouid32");
663 	if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
664 		seq_puts(seq, ",debug");
665 	if (test_opt(sb, OLDALLOC))
666 		seq_puts(seq, ",oldalloc");
667 #ifdef CONFIG_EXT4_FS_XATTR
668 	if (test_opt(sb, XATTR_USER) &&
669 		!(def_mount_opts & EXT4_DEFM_XATTR_USER))
670 		seq_puts(seq, ",user_xattr");
671 	if (!test_opt(sb, XATTR_USER) &&
672 	    (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
673 		seq_puts(seq, ",nouser_xattr");
674 	}
675 #endif
676 #ifdef CONFIG_EXT4_FS_POSIX_ACL
677 	if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
678 		seq_puts(seq, ",acl");
679 	if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
680 		seq_puts(seq, ",noacl");
681 #endif
682 	if (!test_opt(sb, RESERVATION))
683 		seq_puts(seq, ",noreservation");
684 	if (sbi->s_commit_interval) {
685 		seq_printf(seq, ",commit=%u",
686 			   (unsigned) (sbi->s_commit_interval / HZ));
687 	}
688 	/*
689 	 * We're changing the default of barrier mount option, so
690 	 * let's always display its mount state so it's clear what its
691 	 * status is.
692 	 */
693 	seq_puts(seq, ",barrier=");
694 	seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
695 	if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
696 		seq_puts(seq, ",journal_async_commit");
697 	if (test_opt(sb, NOBH))
698 		seq_puts(seq, ",nobh");
699 	if (!test_opt(sb, EXTENTS))
700 		seq_puts(seq, ",noextents");
701 	if (test_opt(sb, I_VERSION))
702 		seq_puts(seq, ",i_version");
703 	if (!test_opt(sb, DELALLOC))
704 		seq_puts(seq, ",nodelalloc");
705 
706 
707 	if (sbi->s_stripe)
708 		seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
709 	/*
710 	 * journal mode get enabled in different ways
711 	 * So just print the value even if we didn't specify it
712 	 */
713 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
714 		seq_puts(seq, ",data=journal");
715 	else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
716 		seq_puts(seq, ",data=ordered");
717 	else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
718 		seq_puts(seq, ",data=writeback");
719 
720 	if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
721 		seq_printf(seq, ",inode_readahead_blks=%u",
722 			   sbi->s_inode_readahead_blks);
723 
724 	if (test_opt(sb, DATA_ERR_ABORT))
725 		seq_puts(seq, ",data_err=abort");
726 
727 	ext4_show_quota_options(seq, sb);
728 	return 0;
729 }
730 
731 
732 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
733 		u64 ino, u32 generation)
734 {
735 	struct inode *inode;
736 
737 	if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
738 		return ERR_PTR(-ESTALE);
739 	if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
740 		return ERR_PTR(-ESTALE);
741 
742 	/* iget isn't really right if the inode is currently unallocated!!
743 	 *
744 	 * ext4_read_inode will return a bad_inode if the inode had been
745 	 * deleted, so we should be safe.
746 	 *
747 	 * Currently we don't know the generation for parent directory, so
748 	 * a generation of 0 means "accept any"
749 	 */
750 	inode = ext4_iget(sb, ino);
751 	if (IS_ERR(inode))
752 		return ERR_CAST(inode);
753 	if (generation && inode->i_generation != generation) {
754 		iput(inode);
755 		return ERR_PTR(-ESTALE);
756 	}
757 
758 	return inode;
759 }
760 
761 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
762 		int fh_len, int fh_type)
763 {
764 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
765 				    ext4_nfs_get_inode);
766 }
767 
768 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
769 		int fh_len, int fh_type)
770 {
771 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
772 				    ext4_nfs_get_inode);
773 }
774 
775 #ifdef CONFIG_QUOTA
776 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
777 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
778 
779 static int ext4_dquot_initialize(struct inode *inode, int type);
780 static int ext4_dquot_drop(struct inode *inode);
781 static int ext4_write_dquot(struct dquot *dquot);
782 static int ext4_acquire_dquot(struct dquot *dquot);
783 static int ext4_release_dquot(struct dquot *dquot);
784 static int ext4_mark_dquot_dirty(struct dquot *dquot);
785 static int ext4_write_info(struct super_block *sb, int type);
786 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
787 				char *path, int remount);
788 static int ext4_quota_on_mount(struct super_block *sb, int type);
789 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
790 			       size_t len, loff_t off);
791 static ssize_t ext4_quota_write(struct super_block *sb, int type,
792 				const char *data, size_t len, loff_t off);
793 
794 static struct dquot_operations ext4_quota_operations = {
795 	.initialize	= ext4_dquot_initialize,
796 	.drop		= ext4_dquot_drop,
797 	.alloc_space	= dquot_alloc_space,
798 	.alloc_inode	= dquot_alloc_inode,
799 	.free_space	= dquot_free_space,
800 	.free_inode	= dquot_free_inode,
801 	.transfer	= dquot_transfer,
802 	.write_dquot	= ext4_write_dquot,
803 	.acquire_dquot	= ext4_acquire_dquot,
804 	.release_dquot	= ext4_release_dquot,
805 	.mark_dirty	= ext4_mark_dquot_dirty,
806 	.write_info	= ext4_write_info
807 };
808 
809 static struct quotactl_ops ext4_qctl_operations = {
810 	.quota_on	= ext4_quota_on,
811 	.quota_off	= vfs_quota_off,
812 	.quota_sync	= vfs_quota_sync,
813 	.get_info	= vfs_get_dqinfo,
814 	.set_info	= vfs_set_dqinfo,
815 	.get_dqblk	= vfs_get_dqblk,
816 	.set_dqblk	= vfs_set_dqblk
817 };
818 #endif
819 
820 static const struct super_operations ext4_sops = {
821 	.alloc_inode	= ext4_alloc_inode,
822 	.destroy_inode	= ext4_destroy_inode,
823 	.write_inode	= ext4_write_inode,
824 	.dirty_inode	= ext4_dirty_inode,
825 	.delete_inode	= ext4_delete_inode,
826 	.put_super	= ext4_put_super,
827 	.write_super	= ext4_write_super,
828 	.sync_fs	= ext4_sync_fs,
829 	.write_super_lockfs = ext4_write_super_lockfs,
830 	.unlockfs	= ext4_unlockfs,
831 	.statfs		= ext4_statfs,
832 	.remount_fs	= ext4_remount,
833 	.clear_inode	= ext4_clear_inode,
834 	.show_options	= ext4_show_options,
835 #ifdef CONFIG_QUOTA
836 	.quota_read	= ext4_quota_read,
837 	.quota_write	= ext4_quota_write,
838 #endif
839 };
840 
841 static const struct export_operations ext4_export_ops = {
842 	.fh_to_dentry = ext4_fh_to_dentry,
843 	.fh_to_parent = ext4_fh_to_parent,
844 	.get_parent = ext4_get_parent,
845 };
846 
847 enum {
848 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
849 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
850 	Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
851 	Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
852 	Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
853 	Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
854 	Opt_journal_checksum, Opt_journal_async_commit,
855 	Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
856 	Opt_data_err_abort, Opt_data_err_ignore,
857 	Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
858 	Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
859 	Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
860 	Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
861 	Opt_stripe, Opt_delalloc, Opt_nodelalloc,
862 	Opt_inode_readahead_blks
863 };
864 
865 static const match_table_t tokens = {
866 	{Opt_bsd_df, "bsddf"},
867 	{Opt_minix_df, "minixdf"},
868 	{Opt_grpid, "grpid"},
869 	{Opt_grpid, "bsdgroups"},
870 	{Opt_nogrpid, "nogrpid"},
871 	{Opt_nogrpid, "sysvgroups"},
872 	{Opt_resgid, "resgid=%u"},
873 	{Opt_resuid, "resuid=%u"},
874 	{Opt_sb, "sb=%u"},
875 	{Opt_err_cont, "errors=continue"},
876 	{Opt_err_panic, "errors=panic"},
877 	{Opt_err_ro, "errors=remount-ro"},
878 	{Opt_nouid32, "nouid32"},
879 	{Opt_debug, "debug"},
880 	{Opt_oldalloc, "oldalloc"},
881 	{Opt_orlov, "orlov"},
882 	{Opt_user_xattr, "user_xattr"},
883 	{Opt_nouser_xattr, "nouser_xattr"},
884 	{Opt_acl, "acl"},
885 	{Opt_noacl, "noacl"},
886 	{Opt_reservation, "reservation"},
887 	{Opt_noreservation, "noreservation"},
888 	{Opt_noload, "noload"},
889 	{Opt_nobh, "nobh"},
890 	{Opt_bh, "bh"},
891 	{Opt_commit, "commit=%u"},
892 	{Opt_journal_update, "journal=update"},
893 	{Opt_journal_inum, "journal=%u"},
894 	{Opt_journal_dev, "journal_dev=%u"},
895 	{Opt_journal_checksum, "journal_checksum"},
896 	{Opt_journal_async_commit, "journal_async_commit"},
897 	{Opt_abort, "abort"},
898 	{Opt_data_journal, "data=journal"},
899 	{Opt_data_ordered, "data=ordered"},
900 	{Opt_data_writeback, "data=writeback"},
901 	{Opt_data_err_abort, "data_err=abort"},
902 	{Opt_data_err_ignore, "data_err=ignore"},
903 	{Opt_offusrjquota, "usrjquota="},
904 	{Opt_usrjquota, "usrjquota=%s"},
905 	{Opt_offgrpjquota, "grpjquota="},
906 	{Opt_grpjquota, "grpjquota=%s"},
907 	{Opt_jqfmt_vfsold, "jqfmt=vfsold"},
908 	{Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
909 	{Opt_grpquota, "grpquota"},
910 	{Opt_noquota, "noquota"},
911 	{Opt_quota, "quota"},
912 	{Opt_usrquota, "usrquota"},
913 	{Opt_barrier, "barrier=%u"},
914 	{Opt_extents, "extents"},
915 	{Opt_noextents, "noextents"},
916 	{Opt_i_version, "i_version"},
917 	{Opt_stripe, "stripe=%u"},
918 	{Opt_resize, "resize"},
919 	{Opt_delalloc, "delalloc"},
920 	{Opt_nodelalloc, "nodelalloc"},
921 	{Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
922 	{Opt_err, NULL},
923 };
924 
925 static ext4_fsblk_t get_sb_block(void **data)
926 {
927 	ext4_fsblk_t	sb_block;
928 	char		*options = (char *) *data;
929 
930 	if (!options || strncmp(options, "sb=", 3) != 0)
931 		return 1;	/* Default location */
932 	options += 3;
933 	/*todo: use simple_strtoll with >32bit ext4 */
934 	sb_block = simple_strtoul(options, &options, 0);
935 	if (*options && *options != ',') {
936 		printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
937 		       (char *) *data);
938 		return 1;
939 	}
940 	if (*options == ',')
941 		options++;
942 	*data = (void *) options;
943 	return sb_block;
944 }
945 
946 static int parse_options(char *options, struct super_block *sb,
947 			 unsigned int *inum, unsigned long *journal_devnum,
948 			 ext4_fsblk_t *n_blocks_count, int is_remount)
949 {
950 	struct ext4_sb_info *sbi = EXT4_SB(sb);
951 	char *p;
952 	substring_t args[MAX_OPT_ARGS];
953 	int data_opt = 0;
954 	int option;
955 #ifdef CONFIG_QUOTA
956 	int qtype, qfmt;
957 	char *qname;
958 #endif
959 	ext4_fsblk_t last_block;
960 
961 	if (!options)
962 		return 1;
963 
964 	while ((p = strsep(&options, ",")) != NULL) {
965 		int token;
966 		if (!*p)
967 			continue;
968 
969 		token = match_token(p, tokens, args);
970 		switch (token) {
971 		case Opt_bsd_df:
972 			clear_opt(sbi->s_mount_opt, MINIX_DF);
973 			break;
974 		case Opt_minix_df:
975 			set_opt(sbi->s_mount_opt, MINIX_DF);
976 			break;
977 		case Opt_grpid:
978 			set_opt(sbi->s_mount_opt, GRPID);
979 			break;
980 		case Opt_nogrpid:
981 			clear_opt(sbi->s_mount_opt, GRPID);
982 			break;
983 		case Opt_resuid:
984 			if (match_int(&args[0], &option))
985 				return 0;
986 			sbi->s_resuid = option;
987 			break;
988 		case Opt_resgid:
989 			if (match_int(&args[0], &option))
990 				return 0;
991 			sbi->s_resgid = option;
992 			break;
993 		case Opt_sb:
994 			/* handled by get_sb_block() instead of here */
995 			/* *sb_block = match_int(&args[0]); */
996 			break;
997 		case Opt_err_panic:
998 			clear_opt(sbi->s_mount_opt, ERRORS_CONT);
999 			clear_opt(sbi->s_mount_opt, ERRORS_RO);
1000 			set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1001 			break;
1002 		case Opt_err_ro:
1003 			clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1004 			clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1005 			set_opt(sbi->s_mount_opt, ERRORS_RO);
1006 			break;
1007 		case Opt_err_cont:
1008 			clear_opt(sbi->s_mount_opt, ERRORS_RO);
1009 			clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1010 			set_opt(sbi->s_mount_opt, ERRORS_CONT);
1011 			break;
1012 		case Opt_nouid32:
1013 			set_opt(sbi->s_mount_opt, NO_UID32);
1014 			break;
1015 		case Opt_debug:
1016 			set_opt(sbi->s_mount_opt, DEBUG);
1017 			break;
1018 		case Opt_oldalloc:
1019 			set_opt(sbi->s_mount_opt, OLDALLOC);
1020 			break;
1021 		case Opt_orlov:
1022 			clear_opt(sbi->s_mount_opt, OLDALLOC);
1023 			break;
1024 #ifdef CONFIG_EXT4_FS_XATTR
1025 		case Opt_user_xattr:
1026 			set_opt(sbi->s_mount_opt, XATTR_USER);
1027 			break;
1028 		case Opt_nouser_xattr:
1029 			clear_opt(sbi->s_mount_opt, XATTR_USER);
1030 			break;
1031 #else
1032 		case Opt_user_xattr:
1033 		case Opt_nouser_xattr:
1034 			printk(KERN_ERR "EXT4 (no)user_xattr options "
1035 			       "not supported\n");
1036 			break;
1037 #endif
1038 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1039 		case Opt_acl:
1040 			set_opt(sbi->s_mount_opt, POSIX_ACL);
1041 			break;
1042 		case Opt_noacl:
1043 			clear_opt(sbi->s_mount_opt, POSIX_ACL);
1044 			break;
1045 #else
1046 		case Opt_acl:
1047 		case Opt_noacl:
1048 			printk(KERN_ERR "EXT4 (no)acl options "
1049 			       "not supported\n");
1050 			break;
1051 #endif
1052 		case Opt_reservation:
1053 			set_opt(sbi->s_mount_opt, RESERVATION);
1054 			break;
1055 		case Opt_noreservation:
1056 			clear_opt(sbi->s_mount_opt, RESERVATION);
1057 			break;
1058 		case Opt_journal_update:
1059 			/* @@@ FIXME */
1060 			/* Eventually we will want to be able to create
1061 			   a journal file here.  For now, only allow the
1062 			   user to specify an existing inode to be the
1063 			   journal file. */
1064 			if (is_remount) {
1065 				printk(KERN_ERR "EXT4-fs: cannot specify "
1066 				       "journal on remount\n");
1067 				return 0;
1068 			}
1069 			set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1070 			break;
1071 		case Opt_journal_inum:
1072 			if (is_remount) {
1073 				printk(KERN_ERR "EXT4-fs: cannot specify "
1074 				       "journal on remount\n");
1075 				return 0;
1076 			}
1077 			if (match_int(&args[0], &option))
1078 				return 0;
1079 			*inum = option;
1080 			break;
1081 		case Opt_journal_dev:
1082 			if (is_remount) {
1083 				printk(KERN_ERR "EXT4-fs: cannot specify "
1084 				       "journal on remount\n");
1085 				return 0;
1086 			}
1087 			if (match_int(&args[0], &option))
1088 				return 0;
1089 			*journal_devnum = option;
1090 			break;
1091 		case Opt_journal_checksum:
1092 			set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1093 			break;
1094 		case Opt_journal_async_commit:
1095 			set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1096 			set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1097 			break;
1098 		case Opt_noload:
1099 			set_opt(sbi->s_mount_opt, NOLOAD);
1100 			break;
1101 		case Opt_commit:
1102 			if (match_int(&args[0], &option))
1103 				return 0;
1104 			if (option < 0)
1105 				return 0;
1106 			if (option == 0)
1107 				option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1108 			sbi->s_commit_interval = HZ * option;
1109 			break;
1110 		case Opt_data_journal:
1111 			data_opt = EXT4_MOUNT_JOURNAL_DATA;
1112 			goto datacheck;
1113 		case Opt_data_ordered:
1114 			data_opt = EXT4_MOUNT_ORDERED_DATA;
1115 			goto datacheck;
1116 		case Opt_data_writeback:
1117 			data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1118 		datacheck:
1119 			if (is_remount) {
1120 				if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1121 						!= data_opt) {
1122 					printk(KERN_ERR
1123 						"EXT4-fs: cannot change data "
1124 						"mode on remount\n");
1125 					return 0;
1126 				}
1127 			} else {
1128 				sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1129 				sbi->s_mount_opt |= data_opt;
1130 			}
1131 			break;
1132 		case Opt_data_err_abort:
1133 			set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1134 			break;
1135 		case Opt_data_err_ignore:
1136 			clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1137 			break;
1138 #ifdef CONFIG_QUOTA
1139 		case Opt_usrjquota:
1140 			qtype = USRQUOTA;
1141 			goto set_qf_name;
1142 		case Opt_grpjquota:
1143 			qtype = GRPQUOTA;
1144 set_qf_name:
1145 			if ((sb_any_quota_enabled(sb) ||
1146 			     sb_any_quota_suspended(sb)) &&
1147 			    !sbi->s_qf_names[qtype]) {
1148 				printk(KERN_ERR
1149 				       "EXT4-fs: Cannot change journaled "
1150 				       "quota options when quota turned on.\n");
1151 				return 0;
1152 			}
1153 			qname = match_strdup(&args[0]);
1154 			if (!qname) {
1155 				printk(KERN_ERR
1156 					"EXT4-fs: not enough memory for "
1157 					"storing quotafile name.\n");
1158 				return 0;
1159 			}
1160 			if (sbi->s_qf_names[qtype] &&
1161 			    strcmp(sbi->s_qf_names[qtype], qname)) {
1162 				printk(KERN_ERR
1163 					"EXT4-fs: %s quota file already "
1164 					"specified.\n", QTYPE2NAME(qtype));
1165 				kfree(qname);
1166 				return 0;
1167 			}
1168 			sbi->s_qf_names[qtype] = qname;
1169 			if (strchr(sbi->s_qf_names[qtype], '/')) {
1170 				printk(KERN_ERR
1171 					"EXT4-fs: quotafile must be on "
1172 					"filesystem root.\n");
1173 				kfree(sbi->s_qf_names[qtype]);
1174 				sbi->s_qf_names[qtype] = NULL;
1175 				return 0;
1176 			}
1177 			set_opt(sbi->s_mount_opt, QUOTA);
1178 			break;
1179 		case Opt_offusrjquota:
1180 			qtype = USRQUOTA;
1181 			goto clear_qf_name;
1182 		case Opt_offgrpjquota:
1183 			qtype = GRPQUOTA;
1184 clear_qf_name:
1185 			if ((sb_any_quota_enabled(sb) ||
1186 			     sb_any_quota_suspended(sb)) &&
1187 			    sbi->s_qf_names[qtype]) {
1188 				printk(KERN_ERR "EXT4-fs: Cannot change "
1189 					"journaled quota options when "
1190 					"quota turned on.\n");
1191 				return 0;
1192 			}
1193 			/*
1194 			 * The space will be released later when all options
1195 			 * are confirmed to be correct
1196 			 */
1197 			sbi->s_qf_names[qtype] = NULL;
1198 			break;
1199 		case Opt_jqfmt_vfsold:
1200 			qfmt = QFMT_VFS_OLD;
1201 			goto set_qf_format;
1202 		case Opt_jqfmt_vfsv0:
1203 			qfmt = QFMT_VFS_V0;
1204 set_qf_format:
1205 			if ((sb_any_quota_enabled(sb) ||
1206 			     sb_any_quota_suspended(sb)) &&
1207 			    sbi->s_jquota_fmt != qfmt) {
1208 				printk(KERN_ERR "EXT4-fs: Cannot change "
1209 					"journaled quota options when "
1210 					"quota turned on.\n");
1211 				return 0;
1212 			}
1213 			sbi->s_jquota_fmt = qfmt;
1214 			break;
1215 		case Opt_quota:
1216 		case Opt_usrquota:
1217 			set_opt(sbi->s_mount_opt, QUOTA);
1218 			set_opt(sbi->s_mount_opt, USRQUOTA);
1219 			break;
1220 		case Opt_grpquota:
1221 			set_opt(sbi->s_mount_opt, QUOTA);
1222 			set_opt(sbi->s_mount_opt, GRPQUOTA);
1223 			break;
1224 		case Opt_noquota:
1225 			if (sb_any_quota_enabled(sb)) {
1226 				printk(KERN_ERR "EXT4-fs: Cannot change quota "
1227 					"options when quota turned on.\n");
1228 				return 0;
1229 			}
1230 			clear_opt(sbi->s_mount_opt, QUOTA);
1231 			clear_opt(sbi->s_mount_opt, USRQUOTA);
1232 			clear_opt(sbi->s_mount_opt, GRPQUOTA);
1233 			break;
1234 #else
1235 		case Opt_quota:
1236 		case Opt_usrquota:
1237 		case Opt_grpquota:
1238 			printk(KERN_ERR
1239 				"EXT4-fs: quota options not supported.\n");
1240 			break;
1241 		case Opt_usrjquota:
1242 		case Opt_grpjquota:
1243 		case Opt_offusrjquota:
1244 		case Opt_offgrpjquota:
1245 		case Opt_jqfmt_vfsold:
1246 		case Opt_jqfmt_vfsv0:
1247 			printk(KERN_ERR
1248 				"EXT4-fs: journaled quota options not "
1249 				"supported.\n");
1250 			break;
1251 		case Opt_noquota:
1252 			break;
1253 #endif
1254 		case Opt_abort:
1255 			set_opt(sbi->s_mount_opt, ABORT);
1256 			break;
1257 		case Opt_barrier:
1258 			if (match_int(&args[0], &option))
1259 				return 0;
1260 			if (option)
1261 				set_opt(sbi->s_mount_opt, BARRIER);
1262 			else
1263 				clear_opt(sbi->s_mount_opt, BARRIER);
1264 			break;
1265 		case Opt_ignore:
1266 			break;
1267 		case Opt_resize:
1268 			if (!is_remount) {
1269 				printk("EXT4-fs: resize option only available "
1270 					"for remount\n");
1271 				return 0;
1272 			}
1273 			if (match_int(&args[0], &option) != 0)
1274 				return 0;
1275 			*n_blocks_count = option;
1276 			break;
1277 		case Opt_nobh:
1278 			set_opt(sbi->s_mount_opt, NOBH);
1279 			break;
1280 		case Opt_bh:
1281 			clear_opt(sbi->s_mount_opt, NOBH);
1282 			break;
1283 		case Opt_extents:
1284 			if (!EXT4_HAS_INCOMPAT_FEATURE(sb,
1285 					EXT4_FEATURE_INCOMPAT_EXTENTS)) {
1286 				ext4_warning(sb, __func__,
1287 					"extents feature not enabled "
1288 					"on this filesystem, use tune2fs\n");
1289 				return 0;
1290 			}
1291 			set_opt(sbi->s_mount_opt, EXTENTS);
1292 			break;
1293 		case Opt_noextents:
1294 			/*
1295 			 * When e2fsprogs support resizing an already existing
1296 			 * ext3 file system to greater than 2**32 we need to
1297 			 * add support to block allocator to handle growing
1298 			 * already existing block  mapped inode so that blocks
1299 			 * allocated for them fall within 2**32
1300 			 */
1301 			last_block = ext4_blocks_count(sbi->s_es) - 1;
1302 			if (last_block  > 0xffffffffULL) {
1303 				printk(KERN_ERR "EXT4-fs: Filesystem too "
1304 						"large to mount with "
1305 						"-o noextents options\n");
1306 				return 0;
1307 			}
1308 			clear_opt(sbi->s_mount_opt, EXTENTS);
1309 			break;
1310 		case Opt_i_version:
1311 			set_opt(sbi->s_mount_opt, I_VERSION);
1312 			sb->s_flags |= MS_I_VERSION;
1313 			break;
1314 		case Opt_nodelalloc:
1315 			clear_opt(sbi->s_mount_opt, DELALLOC);
1316 			break;
1317 		case Opt_stripe:
1318 			if (match_int(&args[0], &option))
1319 				return 0;
1320 			if (option < 0)
1321 				return 0;
1322 			sbi->s_stripe = option;
1323 			break;
1324 		case Opt_delalloc:
1325 			set_opt(sbi->s_mount_opt, DELALLOC);
1326 			break;
1327 		case Opt_inode_readahead_blks:
1328 			if (match_int(&args[0], &option))
1329 				return 0;
1330 			if (option < 0 || option > (1 << 30))
1331 				return 0;
1332 			sbi->s_inode_readahead_blks = option;
1333 			break;
1334 		default:
1335 			printk(KERN_ERR
1336 			       "EXT4-fs: Unrecognized mount option \"%s\" "
1337 			       "or missing value\n", p);
1338 			return 0;
1339 		}
1340 	}
1341 #ifdef CONFIG_QUOTA
1342 	if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1343 		if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1344 		     sbi->s_qf_names[USRQUOTA])
1345 			clear_opt(sbi->s_mount_opt, USRQUOTA);
1346 
1347 		if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1348 		     sbi->s_qf_names[GRPQUOTA])
1349 			clear_opt(sbi->s_mount_opt, GRPQUOTA);
1350 
1351 		if ((sbi->s_qf_names[USRQUOTA] &&
1352 				(sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1353 		    (sbi->s_qf_names[GRPQUOTA] &&
1354 				(sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1355 			printk(KERN_ERR "EXT4-fs: old and new quota "
1356 					"format mixing.\n");
1357 			return 0;
1358 		}
1359 
1360 		if (!sbi->s_jquota_fmt) {
1361 			printk(KERN_ERR "EXT4-fs: journaled quota format "
1362 					"not specified.\n");
1363 			return 0;
1364 		}
1365 	} else {
1366 		if (sbi->s_jquota_fmt) {
1367 			printk(KERN_ERR "EXT4-fs: journaled quota format "
1368 					"specified with no journaling "
1369 					"enabled.\n");
1370 			return 0;
1371 		}
1372 	}
1373 #endif
1374 	return 1;
1375 }
1376 
1377 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1378 			    int read_only)
1379 {
1380 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1381 	int res = 0;
1382 
1383 	if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1384 		printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1385 		       "forcing read-only mode\n");
1386 		res = MS_RDONLY;
1387 	}
1388 	if (read_only)
1389 		return res;
1390 	if (!(sbi->s_mount_state & EXT4_VALID_FS))
1391 		printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1392 		       "running e2fsck is recommended\n");
1393 	else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1394 		printk(KERN_WARNING
1395 		       "EXT4-fs warning: mounting fs with errors, "
1396 		       "running e2fsck is recommended\n");
1397 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1398 		 le16_to_cpu(es->s_mnt_count) >=
1399 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1400 		printk(KERN_WARNING
1401 		       "EXT4-fs warning: maximal mount count reached, "
1402 		       "running e2fsck is recommended\n");
1403 	else if (le32_to_cpu(es->s_checkinterval) &&
1404 		(le32_to_cpu(es->s_lastcheck) +
1405 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1406 		printk(KERN_WARNING
1407 		       "EXT4-fs warning: checktime reached, "
1408 		       "running e2fsck is recommended\n");
1409 #if 0
1410 		/* @@@ We _will_ want to clear the valid bit if we find
1411 		 * inconsistencies, to force a fsck at reboot.  But for
1412 		 * a plain journaled filesystem we can keep it set as
1413 		 * valid forever! :)
1414 		 */
1415 	es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1416 #endif
1417 	if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1418 		es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1419 	le16_add_cpu(&es->s_mnt_count, 1);
1420 	es->s_mtime = cpu_to_le32(get_seconds());
1421 	ext4_update_dynamic_rev(sb);
1422 	EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1423 
1424 	ext4_commit_super(sb, es, 1);
1425 	if (test_opt(sb, DEBUG))
1426 		printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
1427 				"bpg=%lu, ipg=%lu, mo=%04lx]\n",
1428 			sb->s_blocksize,
1429 			sbi->s_groups_count,
1430 			EXT4_BLOCKS_PER_GROUP(sb),
1431 			EXT4_INODES_PER_GROUP(sb),
1432 			sbi->s_mount_opt);
1433 
1434 	printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1435 	       sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1436 	       "external", EXT4_SB(sb)->s_journal->j_devname);
1437 	return res;
1438 }
1439 
1440 static int ext4_fill_flex_info(struct super_block *sb)
1441 {
1442 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1443 	struct ext4_group_desc *gdp = NULL;
1444 	struct buffer_head *bh;
1445 	ext4_group_t flex_group_count;
1446 	ext4_group_t flex_group;
1447 	int groups_per_flex = 0;
1448 	__u64 block_bitmap = 0;
1449 	int i;
1450 
1451 	if (!sbi->s_es->s_log_groups_per_flex) {
1452 		sbi->s_log_groups_per_flex = 0;
1453 		return 1;
1454 	}
1455 
1456 	sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1457 	groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1458 
1459 	/* We allocate both existing and potentially added groups */
1460 	flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1461 			    ((sbi->s_es->s_reserved_gdt_blocks +1 ) <<
1462 			      EXT4_DESC_PER_BLOCK_BITS(sb))) /
1463 			   groups_per_flex;
1464 	sbi->s_flex_groups = kzalloc(flex_group_count *
1465 				     sizeof(struct flex_groups), GFP_KERNEL);
1466 	if (sbi->s_flex_groups == NULL) {
1467 		printk(KERN_ERR "EXT4-fs: not enough memory for "
1468 				"%lu flex groups\n", flex_group_count);
1469 		goto failed;
1470 	}
1471 
1472 	gdp = ext4_get_group_desc(sb, 1, &bh);
1473 	block_bitmap = ext4_block_bitmap(sb, gdp) - 1;
1474 
1475 	for (i = 0; i < sbi->s_groups_count; i++) {
1476 		gdp = ext4_get_group_desc(sb, i, &bh);
1477 
1478 		flex_group = ext4_flex_group(sbi, i);
1479 		sbi->s_flex_groups[flex_group].free_inodes +=
1480 			le16_to_cpu(gdp->bg_free_inodes_count);
1481 		sbi->s_flex_groups[flex_group].free_blocks +=
1482 			le16_to_cpu(gdp->bg_free_blocks_count);
1483 	}
1484 
1485 	return 1;
1486 failed:
1487 	return 0;
1488 }
1489 
1490 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1491 			    struct ext4_group_desc *gdp)
1492 {
1493 	__u16 crc = 0;
1494 
1495 	if (sbi->s_es->s_feature_ro_compat &
1496 	    cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1497 		int offset = offsetof(struct ext4_group_desc, bg_checksum);
1498 		__le32 le_group = cpu_to_le32(block_group);
1499 
1500 		crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1501 		crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1502 		crc = crc16(crc, (__u8 *)gdp, offset);
1503 		offset += sizeof(gdp->bg_checksum); /* skip checksum */
1504 		/* for checksum of struct ext4_group_desc do the rest...*/
1505 		if ((sbi->s_es->s_feature_incompat &
1506 		     cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1507 		    offset < le16_to_cpu(sbi->s_es->s_desc_size))
1508 			crc = crc16(crc, (__u8 *)gdp + offset,
1509 				    le16_to_cpu(sbi->s_es->s_desc_size) -
1510 					offset);
1511 	}
1512 
1513 	return cpu_to_le16(crc);
1514 }
1515 
1516 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1517 				struct ext4_group_desc *gdp)
1518 {
1519 	if ((sbi->s_es->s_feature_ro_compat &
1520 	     cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1521 	    (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1522 		return 0;
1523 
1524 	return 1;
1525 }
1526 
1527 /* Called at mount-time, super-block is locked */
1528 static int ext4_check_descriptors(struct super_block *sb)
1529 {
1530 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1531 	ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1532 	ext4_fsblk_t last_block;
1533 	ext4_fsblk_t block_bitmap;
1534 	ext4_fsblk_t inode_bitmap;
1535 	ext4_fsblk_t inode_table;
1536 	int flexbg_flag = 0;
1537 	ext4_group_t i;
1538 
1539 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1540 		flexbg_flag = 1;
1541 
1542 	ext4_debug("Checking group descriptors");
1543 
1544 	for (i = 0; i < sbi->s_groups_count; i++) {
1545 		struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1546 
1547 		if (i == sbi->s_groups_count - 1 || flexbg_flag)
1548 			last_block = ext4_blocks_count(sbi->s_es) - 1;
1549 		else
1550 			last_block = first_block +
1551 				(EXT4_BLOCKS_PER_GROUP(sb) - 1);
1552 
1553 		block_bitmap = ext4_block_bitmap(sb, gdp);
1554 		if (block_bitmap < first_block || block_bitmap > last_block) {
1555 			printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1556 			       "Block bitmap for group %lu not in group "
1557 			       "(block %llu)!\n", i, block_bitmap);
1558 			return 0;
1559 		}
1560 		inode_bitmap = ext4_inode_bitmap(sb, gdp);
1561 		if (inode_bitmap < first_block || inode_bitmap > last_block) {
1562 			printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1563 			       "Inode bitmap for group %lu not in group "
1564 			       "(block %llu)!\n", i, inode_bitmap);
1565 			return 0;
1566 		}
1567 		inode_table = ext4_inode_table(sb, gdp);
1568 		if (inode_table < first_block ||
1569 		    inode_table + sbi->s_itb_per_group - 1 > last_block) {
1570 			printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1571 			       "Inode table for group %lu not in group "
1572 			       "(block %llu)!\n", i, inode_table);
1573 			return 0;
1574 		}
1575 		spin_lock(sb_bgl_lock(sbi, i));
1576 		if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1577 			printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1578 			       "Checksum for group %lu failed (%u!=%u)\n",
1579 			       i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1580 			       gdp)), le16_to_cpu(gdp->bg_checksum));
1581 			if (!(sb->s_flags & MS_RDONLY)) {
1582 				spin_unlock(sb_bgl_lock(sbi, i));
1583 				return 0;
1584 			}
1585 		}
1586 		spin_unlock(sb_bgl_lock(sbi, i));
1587 		if (!flexbg_flag)
1588 			first_block += EXT4_BLOCKS_PER_GROUP(sb);
1589 	}
1590 
1591 	ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1592 	sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
1593 	return 1;
1594 }
1595 
1596 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1597  * the superblock) which were deleted from all directories, but held open by
1598  * a process at the time of a crash.  We walk the list and try to delete these
1599  * inodes at recovery time (only with a read-write filesystem).
1600  *
1601  * In order to keep the orphan inode chain consistent during traversal (in
1602  * case of crash during recovery), we link each inode into the superblock
1603  * orphan list_head and handle it the same way as an inode deletion during
1604  * normal operation (which journals the operations for us).
1605  *
1606  * We only do an iget() and an iput() on each inode, which is very safe if we
1607  * accidentally point at an in-use or already deleted inode.  The worst that
1608  * can happen in this case is that we get a "bit already cleared" message from
1609  * ext4_free_inode().  The only reason we would point at a wrong inode is if
1610  * e2fsck was run on this filesystem, and it must have already done the orphan
1611  * inode cleanup for us, so we can safely abort without any further action.
1612  */
1613 static void ext4_orphan_cleanup(struct super_block *sb,
1614 				struct ext4_super_block *es)
1615 {
1616 	unsigned int s_flags = sb->s_flags;
1617 	int nr_orphans = 0, nr_truncates = 0;
1618 #ifdef CONFIG_QUOTA
1619 	int i;
1620 #endif
1621 	if (!es->s_last_orphan) {
1622 		jbd_debug(4, "no orphan inodes to clean up\n");
1623 		return;
1624 	}
1625 
1626 	if (bdev_read_only(sb->s_bdev)) {
1627 		printk(KERN_ERR "EXT4-fs: write access "
1628 			"unavailable, skipping orphan cleanup.\n");
1629 		return;
1630 	}
1631 
1632 	if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1633 		if (es->s_last_orphan)
1634 			jbd_debug(1, "Errors on filesystem, "
1635 				  "clearing orphan list.\n");
1636 		es->s_last_orphan = 0;
1637 		jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1638 		return;
1639 	}
1640 
1641 	if (s_flags & MS_RDONLY) {
1642 		printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1643 		       sb->s_id);
1644 		sb->s_flags &= ~MS_RDONLY;
1645 	}
1646 #ifdef CONFIG_QUOTA
1647 	/* Needed for iput() to work correctly and not trash data */
1648 	sb->s_flags |= MS_ACTIVE;
1649 	/* Turn on quotas so that they are updated correctly */
1650 	for (i = 0; i < MAXQUOTAS; i++) {
1651 		if (EXT4_SB(sb)->s_qf_names[i]) {
1652 			int ret = ext4_quota_on_mount(sb, i);
1653 			if (ret < 0)
1654 				printk(KERN_ERR
1655 					"EXT4-fs: Cannot turn on journaled "
1656 					"quota: error %d\n", ret);
1657 		}
1658 	}
1659 #endif
1660 
1661 	while (es->s_last_orphan) {
1662 		struct inode *inode;
1663 
1664 		inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1665 		if (IS_ERR(inode)) {
1666 			es->s_last_orphan = 0;
1667 			break;
1668 		}
1669 
1670 		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1671 		DQUOT_INIT(inode);
1672 		if (inode->i_nlink) {
1673 			printk(KERN_DEBUG
1674 				"%s: truncating inode %lu to %lld bytes\n",
1675 				__func__, inode->i_ino, inode->i_size);
1676 			jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1677 				  inode->i_ino, inode->i_size);
1678 			ext4_truncate(inode);
1679 			nr_truncates++;
1680 		} else {
1681 			printk(KERN_DEBUG
1682 				"%s: deleting unreferenced inode %lu\n",
1683 				__func__, inode->i_ino);
1684 			jbd_debug(2, "deleting unreferenced inode %lu\n",
1685 				  inode->i_ino);
1686 			nr_orphans++;
1687 		}
1688 		iput(inode);  /* The delete magic happens here! */
1689 	}
1690 
1691 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1692 
1693 	if (nr_orphans)
1694 		printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1695 		       sb->s_id, PLURAL(nr_orphans));
1696 	if (nr_truncates)
1697 		printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1698 		       sb->s_id, PLURAL(nr_truncates));
1699 #ifdef CONFIG_QUOTA
1700 	/* Turn quotas off */
1701 	for (i = 0; i < MAXQUOTAS; i++) {
1702 		if (sb_dqopt(sb)->files[i])
1703 			vfs_quota_off(sb, i, 0);
1704 	}
1705 #endif
1706 	sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1707 }
1708 /*
1709  * Maximal extent format file size.
1710  * Resulting logical blkno at s_maxbytes must fit in our on-disk
1711  * extent format containers, within a sector_t, and within i_blocks
1712  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
1713  * so that won't be a limiting factor.
1714  *
1715  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1716  */
1717 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1718 {
1719 	loff_t res;
1720 	loff_t upper_limit = MAX_LFS_FILESIZE;
1721 
1722 	/* small i_blocks in vfs inode? */
1723 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1724 		/*
1725 		 * CONFIG_LSF is not enabled implies the inode
1726 		 * i_block represent total blocks in 512 bytes
1727 		 * 32 == size of vfs inode i_blocks * 8
1728 		 */
1729 		upper_limit = (1LL << 32) - 1;
1730 
1731 		/* total blocks in file system block size */
1732 		upper_limit >>= (blkbits - 9);
1733 		upper_limit <<= blkbits;
1734 	}
1735 
1736 	/* 32-bit extent-start container, ee_block */
1737 	res = 1LL << 32;
1738 	res <<= blkbits;
1739 	res -= 1;
1740 
1741 	/* Sanity check against vm- & vfs- imposed limits */
1742 	if (res > upper_limit)
1743 		res = upper_limit;
1744 
1745 	return res;
1746 }
1747 
1748 /*
1749  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
1750  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1751  * We need to be 1 filesystem block less than the 2^48 sector limit.
1752  */
1753 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1754 {
1755 	loff_t res = EXT4_NDIR_BLOCKS;
1756 	int meta_blocks;
1757 	loff_t upper_limit;
1758 	/* This is calculated to be the largest file size for a
1759 	 * dense, bitmapped file such that the total number of
1760 	 * sectors in the file, including data and all indirect blocks,
1761 	 * does not exceed 2^48 -1
1762 	 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1763 	 * total number of  512 bytes blocks of the file
1764 	 */
1765 
1766 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1767 		/*
1768 		 * !has_huge_files or CONFIG_LSF is not enabled
1769 		 * implies the inode i_block represent total blocks in
1770 		 * 512 bytes 32 == size of vfs inode i_blocks * 8
1771 		 */
1772 		upper_limit = (1LL << 32) - 1;
1773 
1774 		/* total blocks in file system block size */
1775 		upper_limit >>= (bits - 9);
1776 
1777 	} else {
1778 		/*
1779 		 * We use 48 bit ext4_inode i_blocks
1780 		 * With EXT4_HUGE_FILE_FL set the i_blocks
1781 		 * represent total number of blocks in
1782 		 * file system block size
1783 		 */
1784 		upper_limit = (1LL << 48) - 1;
1785 
1786 	}
1787 
1788 	/* indirect blocks */
1789 	meta_blocks = 1;
1790 	/* double indirect blocks */
1791 	meta_blocks += 1 + (1LL << (bits-2));
1792 	/* tripple indirect blocks */
1793 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1794 
1795 	upper_limit -= meta_blocks;
1796 	upper_limit <<= bits;
1797 
1798 	res += 1LL << (bits-2);
1799 	res += 1LL << (2*(bits-2));
1800 	res += 1LL << (3*(bits-2));
1801 	res <<= bits;
1802 	if (res > upper_limit)
1803 		res = upper_limit;
1804 
1805 	if (res > MAX_LFS_FILESIZE)
1806 		res = MAX_LFS_FILESIZE;
1807 
1808 	return res;
1809 }
1810 
1811 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1812 				ext4_fsblk_t logical_sb_block, int nr)
1813 {
1814 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1815 	ext4_group_t bg, first_meta_bg;
1816 	int has_super = 0;
1817 
1818 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1819 
1820 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1821 	    nr < first_meta_bg)
1822 		return logical_sb_block + nr + 1;
1823 	bg = sbi->s_desc_per_block * nr;
1824 	if (ext4_bg_has_super(sb, bg))
1825 		has_super = 1;
1826 	return (has_super + ext4_group_first_block_no(sb, bg));
1827 }
1828 
1829 /**
1830  * ext4_get_stripe_size: Get the stripe size.
1831  * @sbi: In memory super block info
1832  *
1833  * If we have specified it via mount option, then
1834  * use the mount option value. If the value specified at mount time is
1835  * greater than the blocks per group use the super block value.
1836  * If the super block value is greater than blocks per group return 0.
1837  * Allocator needs it be less than blocks per group.
1838  *
1839  */
1840 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1841 {
1842 	unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1843 	unsigned long stripe_width =
1844 			le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1845 
1846 	if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1847 		return sbi->s_stripe;
1848 
1849 	if (stripe_width <= sbi->s_blocks_per_group)
1850 		return stripe_width;
1851 
1852 	if (stride <= sbi->s_blocks_per_group)
1853 		return stride;
1854 
1855 	return 0;
1856 }
1857 
1858 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
1859 				__releases(kernel_lock)
1860 				__acquires(kernel_lock)
1861 
1862 {
1863 	struct buffer_head *bh;
1864 	struct ext4_super_block *es = NULL;
1865 	struct ext4_sb_info *sbi;
1866 	ext4_fsblk_t block;
1867 	ext4_fsblk_t sb_block = get_sb_block(&data);
1868 	ext4_fsblk_t logical_sb_block;
1869 	unsigned long offset = 0;
1870 	unsigned int journal_inum = 0;
1871 	unsigned long journal_devnum = 0;
1872 	unsigned long def_mount_opts;
1873 	struct inode *root;
1874 	char *cp;
1875 	int ret = -EINVAL;
1876 	int blocksize;
1877 	int db_count;
1878 	int i;
1879 	int needs_recovery, has_huge_files;
1880 	__le32 features;
1881 	__u64 blocks_count;
1882 	int err;
1883 
1884 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1885 	if (!sbi)
1886 		return -ENOMEM;
1887 	sb->s_fs_info = sbi;
1888 	sbi->s_mount_opt = 0;
1889 	sbi->s_resuid = EXT4_DEF_RESUID;
1890 	sbi->s_resgid = EXT4_DEF_RESGID;
1891 	sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
1892 	sbi->s_sb_block = sb_block;
1893 
1894 	unlock_kernel();
1895 
1896 	/* Cleanup superblock name */
1897 	for (cp = sb->s_id; (cp = strchr(cp, '/'));)
1898 		*cp = '!';
1899 
1900 	blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
1901 	if (!blocksize) {
1902 		printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
1903 		goto out_fail;
1904 	}
1905 
1906 	/*
1907 	 * The ext4 superblock will not be buffer aligned for other than 1kB
1908 	 * block sizes.  We need to calculate the offset from buffer start.
1909 	 */
1910 	if (blocksize != EXT4_MIN_BLOCK_SIZE) {
1911 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1912 		offset = do_div(logical_sb_block, blocksize);
1913 	} else {
1914 		logical_sb_block = sb_block;
1915 	}
1916 
1917 	if (!(bh = sb_bread(sb, logical_sb_block))) {
1918 		printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
1919 		goto out_fail;
1920 	}
1921 	/*
1922 	 * Note: s_es must be initialized as soon as possible because
1923 	 *       some ext4 macro-instructions depend on its value
1924 	 */
1925 	es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
1926 	sbi->s_es = es;
1927 	sb->s_magic = le16_to_cpu(es->s_magic);
1928 	if (sb->s_magic != EXT4_SUPER_MAGIC)
1929 		goto cantfind_ext4;
1930 
1931 	/* Set defaults before we parse the mount options */
1932 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1933 	if (def_mount_opts & EXT4_DEFM_DEBUG)
1934 		set_opt(sbi->s_mount_opt, DEBUG);
1935 	if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
1936 		set_opt(sbi->s_mount_opt, GRPID);
1937 	if (def_mount_opts & EXT4_DEFM_UID16)
1938 		set_opt(sbi->s_mount_opt, NO_UID32);
1939 #ifdef CONFIG_EXT4_FS_XATTR
1940 	if (def_mount_opts & EXT4_DEFM_XATTR_USER)
1941 		set_opt(sbi->s_mount_opt, XATTR_USER);
1942 #endif
1943 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1944 	if (def_mount_opts & EXT4_DEFM_ACL)
1945 		set_opt(sbi->s_mount_opt, POSIX_ACL);
1946 #endif
1947 	if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1948 		sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1949 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1950 		sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1951 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1952 		sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1953 
1954 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
1955 		set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1956 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
1957 		set_opt(sbi->s_mount_opt, ERRORS_CONT);
1958 	else
1959 		set_opt(sbi->s_mount_opt, ERRORS_RO);
1960 
1961 	sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1962 	sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1963 
1964 	set_opt(sbi->s_mount_opt, RESERVATION);
1965 	set_opt(sbi->s_mount_opt, BARRIER);
1966 
1967 	/*
1968 	 * turn on extents feature by default in ext4 filesystem
1969 	 * only if feature flag already set by mkfs or tune2fs.
1970 	 * Use -o noextents to turn it off
1971 	 */
1972 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
1973 		set_opt(sbi->s_mount_opt, EXTENTS);
1974 	else
1975 		ext4_warning(sb, __func__,
1976 			"extents feature not enabled on this filesystem, "
1977 			"use tune2fs.\n");
1978 
1979 	/*
1980 	 * enable delayed allocation by default
1981 	 * Use -o nodelalloc to turn it off
1982 	 */
1983 	set_opt(sbi->s_mount_opt, DELALLOC);
1984 
1985 
1986 	if (!parse_options((char *) data, sb, &journal_inum, &journal_devnum,
1987 			   NULL, 0))
1988 		goto failed_mount;
1989 
1990 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1991 		((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1992 
1993 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1994 	    (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1995 	     EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1996 	     EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1997 		printk(KERN_WARNING
1998 		       "EXT4-fs warning: feature flags set on rev 0 fs, "
1999 		       "running e2fsck is recommended\n");
2000 
2001 	/*
2002 	 * Check feature flags regardless of the revision level, since we
2003 	 * previously didn't change the revision level when setting the flags,
2004 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
2005 	 */
2006 	features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2007 	if (features) {
2008 		printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
2009 		       "unsupported optional features (%x).\n",
2010 		       sb->s_id, le32_to_cpu(features));
2011 		goto failed_mount;
2012 	}
2013 	features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2014 	if (!(sb->s_flags & MS_RDONLY) && features) {
2015 		printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
2016 		       "unsupported optional features (%x).\n",
2017 		       sb->s_id, le32_to_cpu(features));
2018 		goto failed_mount;
2019 	}
2020 	has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2021 				    EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2022 	if (has_huge_files) {
2023 		/*
2024 		 * Large file size enabled file system can only be
2025 		 * mount if kernel is build with CONFIG_LSF
2026 		 */
2027 		if (sizeof(root->i_blocks) < sizeof(u64) &&
2028 				!(sb->s_flags & MS_RDONLY)) {
2029 			printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2030 					"files cannot be mounted read-write "
2031 					"without CONFIG_LSF.\n", sb->s_id);
2032 			goto failed_mount;
2033 		}
2034 	}
2035 	blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2036 
2037 	if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2038 	    blocksize > EXT4_MAX_BLOCK_SIZE) {
2039 		printk(KERN_ERR
2040 		       "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2041 		       blocksize, sb->s_id);
2042 		goto failed_mount;
2043 	}
2044 
2045 	if (sb->s_blocksize != blocksize) {
2046 
2047 		/* Validate the filesystem blocksize */
2048 		if (!sb_set_blocksize(sb, blocksize)) {
2049 			printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2050 					blocksize);
2051 			goto failed_mount;
2052 		}
2053 
2054 		brelse(bh);
2055 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2056 		offset = do_div(logical_sb_block, blocksize);
2057 		bh = sb_bread(sb, logical_sb_block);
2058 		if (!bh) {
2059 			printk(KERN_ERR
2060 			       "EXT4-fs: Can't read superblock on 2nd try.\n");
2061 			goto failed_mount;
2062 		}
2063 		es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2064 		sbi->s_es = es;
2065 		if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2066 			printk(KERN_ERR
2067 			       "EXT4-fs: Magic mismatch, very weird !\n");
2068 			goto failed_mount;
2069 		}
2070 	}
2071 
2072 	sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2073 						      has_huge_files);
2074 	sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2075 
2076 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2077 		sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2078 		sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2079 	} else {
2080 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2081 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2082 		if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2083 		    (!is_power_of_2(sbi->s_inode_size)) ||
2084 		    (sbi->s_inode_size > blocksize)) {
2085 			printk(KERN_ERR
2086 			       "EXT4-fs: unsupported inode size: %d\n",
2087 			       sbi->s_inode_size);
2088 			goto failed_mount;
2089 		}
2090 		if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2091 			sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2092 	}
2093 	sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2094 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2095 		if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2096 		    sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2097 		    !is_power_of_2(sbi->s_desc_size)) {
2098 			printk(KERN_ERR
2099 			       "EXT4-fs: unsupported descriptor size %lu\n",
2100 			       sbi->s_desc_size);
2101 			goto failed_mount;
2102 		}
2103 	} else
2104 		sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2105 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2106 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2107 	if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2108 		goto cantfind_ext4;
2109 	sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2110 	if (sbi->s_inodes_per_block == 0)
2111 		goto cantfind_ext4;
2112 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
2113 					sbi->s_inodes_per_block;
2114 	sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2115 	sbi->s_sbh = bh;
2116 	sbi->s_mount_state = le16_to_cpu(es->s_state);
2117 	sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2118 	sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2119 	for (i = 0; i < 4; i++)
2120 		sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2121 	sbi->s_def_hash_version = es->s_def_hash_version;
2122 
2123 	if (sbi->s_blocks_per_group > blocksize * 8) {
2124 		printk(KERN_ERR
2125 		       "EXT4-fs: #blocks per group too big: %lu\n",
2126 		       sbi->s_blocks_per_group);
2127 		goto failed_mount;
2128 	}
2129 	if (sbi->s_inodes_per_group > blocksize * 8) {
2130 		printk(KERN_ERR
2131 		       "EXT4-fs: #inodes per group too big: %lu\n",
2132 		       sbi->s_inodes_per_group);
2133 		goto failed_mount;
2134 	}
2135 
2136 	if (ext4_blocks_count(es) >
2137 		    (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2138 		printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2139 			" too large to mount safely\n", sb->s_id);
2140 		if (sizeof(sector_t) < 8)
2141 			printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2142 					"enabled\n");
2143 		goto failed_mount;
2144 	}
2145 
2146 	if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2147 		goto cantfind_ext4;
2148 
2149 	/* ensure blocks_count calculation below doesn't sign-extend */
2150 	if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2151 	    le32_to_cpu(es->s_first_data_block) + 1) {
2152 		printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2153 		       "first data block %u, blocks per group %lu\n",
2154 			ext4_blocks_count(es),
2155 			le32_to_cpu(es->s_first_data_block),
2156 			EXT4_BLOCKS_PER_GROUP(sb));
2157 		goto failed_mount;
2158 	}
2159 	blocks_count = (ext4_blocks_count(es) -
2160 			le32_to_cpu(es->s_first_data_block) +
2161 			EXT4_BLOCKS_PER_GROUP(sb) - 1);
2162 	do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2163 	sbi->s_groups_count = blocks_count;
2164 	db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2165 		   EXT4_DESC_PER_BLOCK(sb);
2166 	sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2167 				    GFP_KERNEL);
2168 	if (sbi->s_group_desc == NULL) {
2169 		printk(KERN_ERR "EXT4-fs: not enough memory\n");
2170 		goto failed_mount;
2171 	}
2172 
2173 #ifdef CONFIG_PROC_FS
2174 	if (ext4_proc_root)
2175 		sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2176 
2177 	if (sbi->s_proc)
2178 		proc_create_data("inode_readahead_blks", 0644, sbi->s_proc,
2179 				 &ext4_ui_proc_fops,
2180 				 &sbi->s_inode_readahead_blks);
2181 #endif
2182 
2183 	bgl_lock_init(&sbi->s_blockgroup_lock);
2184 
2185 	for (i = 0; i < db_count; i++) {
2186 		block = descriptor_loc(sb, logical_sb_block, i);
2187 		sbi->s_group_desc[i] = sb_bread(sb, block);
2188 		if (!sbi->s_group_desc[i]) {
2189 			printk(KERN_ERR "EXT4-fs: "
2190 			       "can't read group descriptor %d\n", i);
2191 			db_count = i;
2192 			goto failed_mount2;
2193 		}
2194 	}
2195 	if (!ext4_check_descriptors(sb)) {
2196 		printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2197 		goto failed_mount2;
2198 	}
2199 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2200 		if (!ext4_fill_flex_info(sb)) {
2201 			printk(KERN_ERR
2202 			       "EXT4-fs: unable to initialize "
2203 			       "flex_bg meta info!\n");
2204 			goto failed_mount2;
2205 		}
2206 
2207 	sbi->s_gdb_count = db_count;
2208 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2209 	spin_lock_init(&sbi->s_next_gen_lock);
2210 
2211 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
2212 			ext4_count_free_blocks(sb));
2213 	if (!err) {
2214 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
2215 				ext4_count_free_inodes(sb));
2216 	}
2217 	if (!err) {
2218 		err = percpu_counter_init(&sbi->s_dirs_counter,
2219 				ext4_count_dirs(sb));
2220 	}
2221 	if (!err) {
2222 		err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2223 	}
2224 	if (err) {
2225 		printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2226 		goto failed_mount3;
2227 	}
2228 
2229 	sbi->s_stripe = ext4_get_stripe_size(sbi);
2230 
2231 	/*
2232 	 * set up enough so that it can read an inode
2233 	 */
2234 	sb->s_op = &ext4_sops;
2235 	sb->s_export_op = &ext4_export_ops;
2236 	sb->s_xattr = ext4_xattr_handlers;
2237 #ifdef CONFIG_QUOTA
2238 	sb->s_qcop = &ext4_qctl_operations;
2239 	sb->dq_op = &ext4_quota_operations;
2240 #endif
2241 	INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2242 
2243 	sb->s_root = NULL;
2244 
2245 	needs_recovery = (es->s_last_orphan != 0 ||
2246 			  EXT4_HAS_INCOMPAT_FEATURE(sb,
2247 				    EXT4_FEATURE_INCOMPAT_RECOVER));
2248 
2249 	/*
2250 	 * The first inode we look at is the journal inode.  Don't try
2251 	 * root first: it may be modified in the journal!
2252 	 */
2253 	if (!test_opt(sb, NOLOAD) &&
2254 	    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2255 		if (ext4_load_journal(sb, es, journal_devnum))
2256 			goto failed_mount3;
2257 		if (!(sb->s_flags & MS_RDONLY) &&
2258 		    EXT4_SB(sb)->s_journal->j_failed_commit) {
2259 			printk(KERN_CRIT "EXT4-fs error (device %s): "
2260 			       "ext4_fill_super: Journal transaction "
2261 			       "%u is corrupt\n", sb->s_id,
2262 			       EXT4_SB(sb)->s_journal->j_failed_commit);
2263 			if (test_opt(sb, ERRORS_RO)) {
2264 				printk(KERN_CRIT
2265 				       "Mounting filesystem read-only\n");
2266 				sb->s_flags |= MS_RDONLY;
2267 				EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2268 				es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2269 			}
2270 			if (test_opt(sb, ERRORS_PANIC)) {
2271 				EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2272 				es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2273 				ext4_commit_super(sb, es, 1);
2274 				printk(KERN_CRIT
2275 				       "EXT4-fs (device %s): mount failed\n",
2276 				      sb->s_id);
2277 				goto failed_mount4;
2278 			}
2279 		}
2280 	} else if (journal_inum) {
2281 		if (ext4_create_journal(sb, es, journal_inum))
2282 			goto failed_mount3;
2283 	} else {
2284 		if (!silent)
2285 			printk(KERN_ERR
2286 			       "ext4: No journal on filesystem on %s\n",
2287 			       sb->s_id);
2288 		goto failed_mount3;
2289 	}
2290 
2291 	if (ext4_blocks_count(es) > 0xffffffffULL &&
2292 	    !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2293 				       JBD2_FEATURE_INCOMPAT_64BIT)) {
2294 		printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2295 		goto failed_mount4;
2296 	}
2297 
2298 	if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2299 		jbd2_journal_set_features(sbi->s_journal,
2300 				JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2301 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2302 	} else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2303 		jbd2_journal_set_features(sbi->s_journal,
2304 				JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2305 		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2306 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2307 	} else {
2308 		jbd2_journal_clear_features(sbi->s_journal,
2309 				JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2310 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2311 	}
2312 
2313 	/* We have now updated the journal if required, so we can
2314 	 * validate the data journaling mode. */
2315 	switch (test_opt(sb, DATA_FLAGS)) {
2316 	case 0:
2317 		/* No mode set, assume a default based on the journal
2318 		 * capabilities: ORDERED_DATA if the journal can
2319 		 * cope, else JOURNAL_DATA
2320 		 */
2321 		if (jbd2_journal_check_available_features
2322 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2323 			set_opt(sbi->s_mount_opt, ORDERED_DATA);
2324 		else
2325 			set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2326 		break;
2327 
2328 	case EXT4_MOUNT_ORDERED_DATA:
2329 	case EXT4_MOUNT_WRITEBACK_DATA:
2330 		if (!jbd2_journal_check_available_features
2331 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2332 			printk(KERN_ERR "EXT4-fs: Journal does not support "
2333 			       "requested data journaling mode\n");
2334 			goto failed_mount4;
2335 		}
2336 	default:
2337 		break;
2338 	}
2339 
2340 	if (test_opt(sb, NOBH)) {
2341 		if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2342 			printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2343 				"its supported only with writeback mode\n");
2344 			clear_opt(sbi->s_mount_opt, NOBH);
2345 		}
2346 	}
2347 	/*
2348 	 * The jbd2_journal_load will have done any necessary log recovery,
2349 	 * so we can safely mount the rest of the filesystem now.
2350 	 */
2351 
2352 	root = ext4_iget(sb, EXT4_ROOT_INO);
2353 	if (IS_ERR(root)) {
2354 		printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2355 		ret = PTR_ERR(root);
2356 		goto failed_mount4;
2357 	}
2358 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2359 		iput(root);
2360 		printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2361 		goto failed_mount4;
2362 	}
2363 	sb->s_root = d_alloc_root(root);
2364 	if (!sb->s_root) {
2365 		printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2366 		iput(root);
2367 		ret = -ENOMEM;
2368 		goto failed_mount4;
2369 	}
2370 
2371 	ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2372 
2373 	/* determine the minimum size of new large inodes, if present */
2374 	if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2375 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2376 						     EXT4_GOOD_OLD_INODE_SIZE;
2377 		if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2378 				       EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2379 			if (sbi->s_want_extra_isize <
2380 			    le16_to_cpu(es->s_want_extra_isize))
2381 				sbi->s_want_extra_isize =
2382 					le16_to_cpu(es->s_want_extra_isize);
2383 			if (sbi->s_want_extra_isize <
2384 			    le16_to_cpu(es->s_min_extra_isize))
2385 				sbi->s_want_extra_isize =
2386 					le16_to_cpu(es->s_min_extra_isize);
2387 		}
2388 	}
2389 	/* Check if enough inode space is available */
2390 	if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2391 							sbi->s_inode_size) {
2392 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2393 						       EXT4_GOOD_OLD_INODE_SIZE;
2394 		printk(KERN_INFO "EXT4-fs: required extra inode space not"
2395 			"available.\n");
2396 	}
2397 
2398 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2399 		printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2400 				"requested data journaling mode\n");
2401 		clear_opt(sbi->s_mount_opt, DELALLOC);
2402 	} else if (test_opt(sb, DELALLOC))
2403 		printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2404 
2405 	ext4_ext_init(sb);
2406 	err = ext4_mb_init(sb, needs_recovery);
2407 	if (err) {
2408 		printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2409 		       err);
2410 		goto failed_mount4;
2411 	}
2412 
2413 	/*
2414 	 * akpm: core read_super() calls in here with the superblock locked.
2415 	 * That deadlocks, because orphan cleanup needs to lock the superblock
2416 	 * in numerous places.  Here we just pop the lock - it's relatively
2417 	 * harmless, because we are now ready to accept write_super() requests,
2418 	 * and aviro says that's the only reason for hanging onto the
2419 	 * superblock lock.
2420 	 */
2421 	EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2422 	ext4_orphan_cleanup(sb, es);
2423 	EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2424 	if (needs_recovery)
2425 		printk(KERN_INFO "EXT4-fs: recovery complete.\n");
2426 	ext4_mark_recovery_complete(sb, es);
2427 	printk(KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2428 	       test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2429 	       test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
2430 	       "writeback");
2431 
2432 	lock_kernel();
2433 	return 0;
2434 
2435 cantfind_ext4:
2436 	if (!silent)
2437 		printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2438 		       sb->s_id);
2439 	goto failed_mount;
2440 
2441 failed_mount4:
2442 	jbd2_journal_destroy(sbi->s_journal);
2443 	sbi->s_journal = NULL;
2444 failed_mount3:
2445 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
2446 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
2447 	percpu_counter_destroy(&sbi->s_dirs_counter);
2448 	percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2449 failed_mount2:
2450 	for (i = 0; i < db_count; i++)
2451 		brelse(sbi->s_group_desc[i]);
2452 	kfree(sbi->s_group_desc);
2453 failed_mount:
2454 	if (sbi->s_proc) {
2455 		remove_proc_entry("inode_readahead_blks", sbi->s_proc);
2456 		remove_proc_entry(sb->s_id, ext4_proc_root);
2457 	}
2458 #ifdef CONFIG_QUOTA
2459 	for (i = 0; i < MAXQUOTAS; i++)
2460 		kfree(sbi->s_qf_names[i]);
2461 #endif
2462 	ext4_blkdev_remove(sbi);
2463 	brelse(bh);
2464 out_fail:
2465 	sb->s_fs_info = NULL;
2466 	kfree(sbi);
2467 	lock_kernel();
2468 	return ret;
2469 }
2470 
2471 /*
2472  * Setup any per-fs journal parameters now.  We'll do this both on
2473  * initial mount, once the journal has been initialised but before we've
2474  * done any recovery; and again on any subsequent remount.
2475  */
2476 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2477 {
2478 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2479 
2480 	if (sbi->s_commit_interval)
2481 		journal->j_commit_interval = sbi->s_commit_interval;
2482 	/* We could also set up an ext4-specific default for the commit
2483 	 * interval here, but for now we'll just fall back to the jbd
2484 	 * default. */
2485 
2486 	spin_lock(&journal->j_state_lock);
2487 	if (test_opt(sb, BARRIER))
2488 		journal->j_flags |= JBD2_BARRIER;
2489 	else
2490 		journal->j_flags &= ~JBD2_BARRIER;
2491 	if (test_opt(sb, DATA_ERR_ABORT))
2492 		journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2493 	else
2494 		journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2495 	spin_unlock(&journal->j_state_lock);
2496 }
2497 
2498 static journal_t *ext4_get_journal(struct super_block *sb,
2499 				   unsigned int journal_inum)
2500 {
2501 	struct inode *journal_inode;
2502 	journal_t *journal;
2503 
2504 	/* First, test for the existence of a valid inode on disk.  Bad
2505 	 * things happen if we iget() an unused inode, as the subsequent
2506 	 * iput() will try to delete it. */
2507 
2508 	journal_inode = ext4_iget(sb, journal_inum);
2509 	if (IS_ERR(journal_inode)) {
2510 		printk(KERN_ERR "EXT4-fs: no journal found.\n");
2511 		return NULL;
2512 	}
2513 	if (!journal_inode->i_nlink) {
2514 		make_bad_inode(journal_inode);
2515 		iput(journal_inode);
2516 		printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2517 		return NULL;
2518 	}
2519 
2520 	jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2521 		  journal_inode, journal_inode->i_size);
2522 	if (!S_ISREG(journal_inode->i_mode)) {
2523 		printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2524 		iput(journal_inode);
2525 		return NULL;
2526 	}
2527 
2528 	journal = jbd2_journal_init_inode(journal_inode);
2529 	if (!journal) {
2530 		printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2531 		iput(journal_inode);
2532 		return NULL;
2533 	}
2534 	journal->j_private = sb;
2535 	ext4_init_journal_params(sb, journal);
2536 	return journal;
2537 }
2538 
2539 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2540 				       dev_t j_dev)
2541 {
2542 	struct buffer_head *bh;
2543 	journal_t *journal;
2544 	ext4_fsblk_t start;
2545 	ext4_fsblk_t len;
2546 	int hblock, blocksize;
2547 	ext4_fsblk_t sb_block;
2548 	unsigned long offset;
2549 	struct ext4_super_block *es;
2550 	struct block_device *bdev;
2551 
2552 	bdev = ext4_blkdev_get(j_dev);
2553 	if (bdev == NULL)
2554 		return NULL;
2555 
2556 	if (bd_claim(bdev, sb)) {
2557 		printk(KERN_ERR
2558 			"EXT4: failed to claim external journal device.\n");
2559 		blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2560 		return NULL;
2561 	}
2562 
2563 	blocksize = sb->s_blocksize;
2564 	hblock = bdev_hardsect_size(bdev);
2565 	if (blocksize < hblock) {
2566 		printk(KERN_ERR
2567 			"EXT4-fs: blocksize too small for journal device.\n");
2568 		goto out_bdev;
2569 	}
2570 
2571 	sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2572 	offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2573 	set_blocksize(bdev, blocksize);
2574 	if (!(bh = __bread(bdev, sb_block, blocksize))) {
2575 		printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2576 		       "external journal\n");
2577 		goto out_bdev;
2578 	}
2579 
2580 	es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2581 	if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2582 	    !(le32_to_cpu(es->s_feature_incompat) &
2583 	      EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2584 		printk(KERN_ERR "EXT4-fs: external journal has "
2585 					"bad superblock\n");
2586 		brelse(bh);
2587 		goto out_bdev;
2588 	}
2589 
2590 	if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2591 		printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2592 		brelse(bh);
2593 		goto out_bdev;
2594 	}
2595 
2596 	len = ext4_blocks_count(es);
2597 	start = sb_block + 1;
2598 	brelse(bh);	/* we're done with the superblock */
2599 
2600 	journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2601 					start, len, blocksize);
2602 	if (!journal) {
2603 		printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2604 		goto out_bdev;
2605 	}
2606 	journal->j_private = sb;
2607 	ll_rw_block(READ, 1, &journal->j_sb_buffer);
2608 	wait_on_buffer(journal->j_sb_buffer);
2609 	if (!buffer_uptodate(journal->j_sb_buffer)) {
2610 		printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2611 		goto out_journal;
2612 	}
2613 	if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2614 		printk(KERN_ERR "EXT4-fs: External journal has more than one "
2615 					"user (unsupported) - %d\n",
2616 			be32_to_cpu(journal->j_superblock->s_nr_users));
2617 		goto out_journal;
2618 	}
2619 	EXT4_SB(sb)->journal_bdev = bdev;
2620 	ext4_init_journal_params(sb, journal);
2621 	return journal;
2622 out_journal:
2623 	jbd2_journal_destroy(journal);
2624 out_bdev:
2625 	ext4_blkdev_put(bdev);
2626 	return NULL;
2627 }
2628 
2629 static int ext4_load_journal(struct super_block *sb,
2630 			     struct ext4_super_block *es,
2631 			     unsigned long journal_devnum)
2632 {
2633 	journal_t *journal;
2634 	unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2635 	dev_t journal_dev;
2636 	int err = 0;
2637 	int really_read_only;
2638 
2639 	if (journal_devnum &&
2640 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2641 		printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2642 			"numbers have changed\n");
2643 		journal_dev = new_decode_dev(journal_devnum);
2644 	} else
2645 		journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2646 
2647 	really_read_only = bdev_read_only(sb->s_bdev);
2648 
2649 	/*
2650 	 * Are we loading a blank journal or performing recovery after a
2651 	 * crash?  For recovery, we need to check in advance whether we
2652 	 * can get read-write access to the device.
2653 	 */
2654 
2655 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2656 		if (sb->s_flags & MS_RDONLY) {
2657 			printk(KERN_INFO "EXT4-fs: INFO: recovery "
2658 					"required on readonly filesystem.\n");
2659 			if (really_read_only) {
2660 				printk(KERN_ERR "EXT4-fs: write access "
2661 					"unavailable, cannot proceed.\n");
2662 				return -EROFS;
2663 			}
2664 			printk(KERN_INFO "EXT4-fs: write access will "
2665 			       "be enabled during recovery.\n");
2666 		}
2667 	}
2668 
2669 	if (journal_inum && journal_dev) {
2670 		printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2671 		       "and inode journals!\n");
2672 		return -EINVAL;
2673 	}
2674 
2675 	if (journal_inum) {
2676 		if (!(journal = ext4_get_journal(sb, journal_inum)))
2677 			return -EINVAL;
2678 	} else {
2679 		if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2680 			return -EINVAL;
2681 	}
2682 
2683 	if (journal->j_flags & JBD2_BARRIER)
2684 		printk(KERN_INFO "EXT4-fs: barriers enabled\n");
2685 	else
2686 		printk(KERN_INFO "EXT4-fs: barriers disabled\n");
2687 
2688 	if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2689 		err = jbd2_journal_update_format(journal);
2690 		if (err)  {
2691 			printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2692 			jbd2_journal_destroy(journal);
2693 			return err;
2694 		}
2695 	}
2696 
2697 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2698 		err = jbd2_journal_wipe(journal, !really_read_only);
2699 	if (!err)
2700 		err = jbd2_journal_load(journal);
2701 
2702 	if (err) {
2703 		printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2704 		jbd2_journal_destroy(journal);
2705 		return err;
2706 	}
2707 
2708 	EXT4_SB(sb)->s_journal = journal;
2709 	ext4_clear_journal_err(sb, es);
2710 
2711 	if (journal_devnum &&
2712 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2713 		es->s_journal_dev = cpu_to_le32(journal_devnum);
2714 		sb->s_dirt = 1;
2715 
2716 		/* Make sure we flush the recovery flag to disk. */
2717 		ext4_commit_super(sb, es, 1);
2718 	}
2719 
2720 	return 0;
2721 }
2722 
2723 static int ext4_create_journal(struct super_block *sb,
2724 			       struct ext4_super_block *es,
2725 			       unsigned int journal_inum)
2726 {
2727 	journal_t *journal;
2728 	int err;
2729 
2730 	if (sb->s_flags & MS_RDONLY) {
2731 		printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
2732 				"create journal.\n");
2733 		return -EROFS;
2734 	}
2735 
2736 	journal = ext4_get_journal(sb, journal_inum);
2737 	if (!journal)
2738 		return -EINVAL;
2739 
2740 	printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
2741 	       journal_inum);
2742 
2743 	err = jbd2_journal_create(journal);
2744 	if (err) {
2745 		printk(KERN_ERR "EXT4-fs: error creating journal.\n");
2746 		jbd2_journal_destroy(journal);
2747 		return -EIO;
2748 	}
2749 
2750 	EXT4_SB(sb)->s_journal = journal;
2751 
2752 	ext4_update_dynamic_rev(sb);
2753 	EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2754 	EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
2755 
2756 	es->s_journal_inum = cpu_to_le32(journal_inum);
2757 	sb->s_dirt = 1;
2758 
2759 	/* Make sure we flush the recovery flag to disk. */
2760 	ext4_commit_super(sb, es, 1);
2761 
2762 	return 0;
2763 }
2764 
2765 static void ext4_commit_super(struct super_block *sb,
2766 			      struct ext4_super_block *es, int sync)
2767 {
2768 	struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2769 
2770 	if (!sbh)
2771 		return;
2772 	if (buffer_write_io_error(sbh)) {
2773 		/*
2774 		 * Oh, dear.  A previous attempt to write the
2775 		 * superblock failed.  This could happen because the
2776 		 * USB device was yanked out.  Or it could happen to
2777 		 * be a transient write error and maybe the block will
2778 		 * be remapped.  Nothing we can do but to retry the
2779 		 * write and hope for the best.
2780 		 */
2781 		printk(KERN_ERR "ext4: previous I/O error to "
2782 		       "superblock detected for %s.\n", sb->s_id);
2783 		clear_buffer_write_io_error(sbh);
2784 		set_buffer_uptodate(sbh);
2785 	}
2786 	es->s_wtime = cpu_to_le32(get_seconds());
2787 	ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
2788 	es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
2789 	BUFFER_TRACE(sbh, "marking dirty");
2790 	mark_buffer_dirty(sbh);
2791 	if (sync) {
2792 		sync_dirty_buffer(sbh);
2793 		if (buffer_write_io_error(sbh)) {
2794 			printk(KERN_ERR "ext4: I/O error while writing "
2795 			       "superblock for %s.\n", sb->s_id);
2796 			clear_buffer_write_io_error(sbh);
2797 			set_buffer_uptodate(sbh);
2798 		}
2799 	}
2800 }
2801 
2802 
2803 /*
2804  * Have we just finished recovery?  If so, and if we are mounting (or
2805  * remounting) the filesystem readonly, then we will end up with a
2806  * consistent fs on disk.  Record that fact.
2807  */
2808 static void ext4_mark_recovery_complete(struct super_block *sb,
2809 					struct ext4_super_block *es)
2810 {
2811 	journal_t *journal = EXT4_SB(sb)->s_journal;
2812 
2813 	jbd2_journal_lock_updates(journal);
2814 	if (jbd2_journal_flush(journal) < 0)
2815 		goto out;
2816 
2817 	lock_super(sb);
2818 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2819 	    sb->s_flags & MS_RDONLY) {
2820 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2821 		sb->s_dirt = 0;
2822 		ext4_commit_super(sb, es, 1);
2823 	}
2824 	unlock_super(sb);
2825 
2826 out:
2827 	jbd2_journal_unlock_updates(journal);
2828 }
2829 
2830 /*
2831  * If we are mounting (or read-write remounting) a filesystem whose journal
2832  * has recorded an error from a previous lifetime, move that error to the
2833  * main filesystem now.
2834  */
2835 static void ext4_clear_journal_err(struct super_block *sb,
2836 				   struct ext4_super_block *es)
2837 {
2838 	journal_t *journal;
2839 	int j_errno;
2840 	const char *errstr;
2841 
2842 	journal = EXT4_SB(sb)->s_journal;
2843 
2844 	/*
2845 	 * Now check for any error status which may have been recorded in the
2846 	 * journal by a prior ext4_error() or ext4_abort()
2847 	 */
2848 
2849 	j_errno = jbd2_journal_errno(journal);
2850 	if (j_errno) {
2851 		char nbuf[16];
2852 
2853 		errstr = ext4_decode_error(sb, j_errno, nbuf);
2854 		ext4_warning(sb, __func__, "Filesystem error recorded "
2855 			     "from previous mount: %s", errstr);
2856 		ext4_warning(sb, __func__, "Marking fs in need of "
2857 			     "filesystem check.");
2858 
2859 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2860 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2861 		ext4_commit_super(sb, es, 1);
2862 
2863 		jbd2_journal_clear_err(journal);
2864 	}
2865 }
2866 
2867 /*
2868  * Force the running and committing transactions to commit,
2869  * and wait on the commit.
2870  */
2871 int ext4_force_commit(struct super_block *sb)
2872 {
2873 	journal_t *journal;
2874 	int ret;
2875 
2876 	if (sb->s_flags & MS_RDONLY)
2877 		return 0;
2878 
2879 	journal = EXT4_SB(sb)->s_journal;
2880 	sb->s_dirt = 0;
2881 	ret = ext4_journal_force_commit(journal);
2882 	return ret;
2883 }
2884 
2885 /*
2886  * Ext4 always journals updates to the superblock itself, so we don't
2887  * have to propagate any other updates to the superblock on disk at this
2888  * point.  Just start an async writeback to get the buffers on their way
2889  * to the disk.
2890  *
2891  * This implicitly triggers the writebehind on sync().
2892  */
2893 
2894 static void ext4_write_super(struct super_block *sb)
2895 {
2896 	if (mutex_trylock(&sb->s_lock) != 0)
2897 		BUG();
2898 	sb->s_dirt = 0;
2899 }
2900 
2901 static int ext4_sync_fs(struct super_block *sb, int wait)
2902 {
2903 	tid_t target;
2904 
2905 	trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
2906 	sb->s_dirt = 0;
2907 	if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
2908 		if (wait)
2909 			jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
2910 	}
2911 	return 0;
2912 }
2913 
2914 /*
2915  * LVM calls this function before a (read-only) snapshot is created.  This
2916  * gives us a chance to flush the journal completely and mark the fs clean.
2917  */
2918 static void ext4_write_super_lockfs(struct super_block *sb)
2919 {
2920 	sb->s_dirt = 0;
2921 
2922 	if (!(sb->s_flags & MS_RDONLY)) {
2923 		journal_t *journal = EXT4_SB(sb)->s_journal;
2924 
2925 		/* Now we set up the journal barrier. */
2926 		jbd2_journal_lock_updates(journal);
2927 
2928 		/*
2929 		 * We don't want to clear needs_recovery flag when we failed
2930 		 * to flush the journal.
2931 		 */
2932 		if (jbd2_journal_flush(journal) < 0)
2933 			return;
2934 
2935 		/* Journal blocked and flushed, clear needs_recovery flag. */
2936 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2937 		ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2938 	}
2939 }
2940 
2941 /*
2942  * Called by LVM after the snapshot is done.  We need to reset the RECOVER
2943  * flag here, even though the filesystem is not technically dirty yet.
2944  */
2945 static void ext4_unlockfs(struct super_block *sb)
2946 {
2947 	if (!(sb->s_flags & MS_RDONLY)) {
2948 		lock_super(sb);
2949 		/* Reser the needs_recovery flag before the fs is unlocked. */
2950 		EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2951 		ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2952 		unlock_super(sb);
2953 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
2954 	}
2955 }
2956 
2957 static int ext4_remount(struct super_block *sb, int *flags, char *data)
2958 {
2959 	struct ext4_super_block *es;
2960 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2961 	ext4_fsblk_t n_blocks_count = 0;
2962 	unsigned long old_sb_flags;
2963 	struct ext4_mount_options old_opts;
2964 	ext4_group_t g;
2965 	int err;
2966 #ifdef CONFIG_QUOTA
2967 	int i;
2968 #endif
2969 
2970 	/* Store the original options */
2971 	old_sb_flags = sb->s_flags;
2972 	old_opts.s_mount_opt = sbi->s_mount_opt;
2973 	old_opts.s_resuid = sbi->s_resuid;
2974 	old_opts.s_resgid = sbi->s_resgid;
2975 	old_opts.s_commit_interval = sbi->s_commit_interval;
2976 #ifdef CONFIG_QUOTA
2977 	old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2978 	for (i = 0; i < MAXQUOTAS; i++)
2979 		old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2980 #endif
2981 
2982 	/*
2983 	 * Allow the "check" option to be passed as a remount option.
2984 	 */
2985 	if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2986 		err = -EINVAL;
2987 		goto restore_opts;
2988 	}
2989 
2990 	if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
2991 		ext4_abort(sb, __func__, "Abort forced by user");
2992 
2993 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2994 		((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2995 
2996 	es = sbi->s_es;
2997 
2998 	ext4_init_journal_params(sb, sbi->s_journal);
2999 
3000 	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3001 		n_blocks_count > ext4_blocks_count(es)) {
3002 		if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
3003 			err = -EROFS;
3004 			goto restore_opts;
3005 		}
3006 
3007 		if (*flags & MS_RDONLY) {
3008 			/*
3009 			 * First of all, the unconditional stuff we have to do
3010 			 * to disable replay of the journal when we next remount
3011 			 */
3012 			sb->s_flags |= MS_RDONLY;
3013 
3014 			/*
3015 			 * OK, test if we are remounting a valid rw partition
3016 			 * readonly, and if so set the rdonly flag and then
3017 			 * mark the partition as valid again.
3018 			 */
3019 			if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3020 			    (sbi->s_mount_state & EXT4_VALID_FS))
3021 				es->s_state = cpu_to_le16(sbi->s_mount_state);
3022 
3023 			/*
3024 			 * We have to unlock super so that we can wait for
3025 			 * transactions.
3026 			 */
3027 			unlock_super(sb);
3028 			ext4_mark_recovery_complete(sb, es);
3029 			lock_super(sb);
3030 		} else {
3031 			__le32 ret;
3032 			if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3033 					~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3034 				printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3035 				       "remount RDWR because of unsupported "
3036 				       "optional features (%x).\n",
3037 				       sb->s_id, le32_to_cpu(ret));
3038 				err = -EROFS;
3039 				goto restore_opts;
3040 			}
3041 
3042 			/*
3043 			 * Make sure the group descriptor checksums
3044 			 * are sane.  If they aren't, refuse to
3045 			 * remount r/w.
3046 			 */
3047 			for (g = 0; g < sbi->s_groups_count; g++) {
3048 				struct ext4_group_desc *gdp =
3049 					ext4_get_group_desc(sb, g, NULL);
3050 
3051 				if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3052 					printk(KERN_ERR
3053 	       "EXT4-fs: ext4_remount: "
3054 		"Checksum for group %lu failed (%u!=%u)\n",
3055 		g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3056 					       le16_to_cpu(gdp->bg_checksum));
3057 					err = -EINVAL;
3058 					goto restore_opts;
3059 				}
3060 			}
3061 
3062 			/*
3063 			 * If we have an unprocessed orphan list hanging
3064 			 * around from a previously readonly bdev mount,
3065 			 * require a full umount/remount for now.
3066 			 */
3067 			if (es->s_last_orphan) {
3068 				printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3069 				       "remount RDWR because of unprocessed "
3070 				       "orphan inode list.  Please "
3071 				       "umount/remount instead.\n",
3072 				       sb->s_id);
3073 				err = -EINVAL;
3074 				goto restore_opts;
3075 			}
3076 
3077 			/*
3078 			 * Mounting a RDONLY partition read-write, so reread
3079 			 * and store the current valid flag.  (It may have
3080 			 * been changed by e2fsck since we originally mounted
3081 			 * the partition.)
3082 			 */
3083 			ext4_clear_journal_err(sb, es);
3084 			sbi->s_mount_state = le16_to_cpu(es->s_state);
3085 			if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3086 				goto restore_opts;
3087 			if (!ext4_setup_super(sb, es, 0))
3088 				sb->s_flags &= ~MS_RDONLY;
3089 		}
3090 	}
3091 #ifdef CONFIG_QUOTA
3092 	/* Release old quota file names */
3093 	for (i = 0; i < MAXQUOTAS; i++)
3094 		if (old_opts.s_qf_names[i] &&
3095 		    old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3096 			kfree(old_opts.s_qf_names[i]);
3097 #endif
3098 	return 0;
3099 restore_opts:
3100 	sb->s_flags = old_sb_flags;
3101 	sbi->s_mount_opt = old_opts.s_mount_opt;
3102 	sbi->s_resuid = old_opts.s_resuid;
3103 	sbi->s_resgid = old_opts.s_resgid;
3104 	sbi->s_commit_interval = old_opts.s_commit_interval;
3105 #ifdef CONFIG_QUOTA
3106 	sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3107 	for (i = 0; i < MAXQUOTAS; i++) {
3108 		if (sbi->s_qf_names[i] &&
3109 		    old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3110 			kfree(sbi->s_qf_names[i]);
3111 		sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3112 	}
3113 #endif
3114 	return err;
3115 }
3116 
3117 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3118 {
3119 	struct super_block *sb = dentry->d_sb;
3120 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3121 	struct ext4_super_block *es = sbi->s_es;
3122 	u64 fsid;
3123 
3124 	if (test_opt(sb, MINIX_DF)) {
3125 		sbi->s_overhead_last = 0;
3126 	} else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3127 		ext4_group_t ngroups = sbi->s_groups_count, i;
3128 		ext4_fsblk_t overhead = 0;
3129 		smp_rmb();
3130 
3131 		/*
3132 		 * Compute the overhead (FS structures).  This is constant
3133 		 * for a given filesystem unless the number of block groups
3134 		 * changes so we cache the previous value until it does.
3135 		 */
3136 
3137 		/*
3138 		 * All of the blocks before first_data_block are
3139 		 * overhead
3140 		 */
3141 		overhead = le32_to_cpu(es->s_first_data_block);
3142 
3143 		/*
3144 		 * Add the overhead attributed to the superblock and
3145 		 * block group descriptors.  If the sparse superblocks
3146 		 * feature is turned on, then not all groups have this.
3147 		 */
3148 		for (i = 0; i < ngroups; i++) {
3149 			overhead += ext4_bg_has_super(sb, i) +
3150 				ext4_bg_num_gdb(sb, i);
3151 			cond_resched();
3152 		}
3153 
3154 		/*
3155 		 * Every block group has an inode bitmap, a block
3156 		 * bitmap, and an inode table.
3157 		 */
3158 		overhead += ngroups * (2 + sbi->s_itb_per_group);
3159 		sbi->s_overhead_last = overhead;
3160 		smp_wmb();
3161 		sbi->s_blocks_last = ext4_blocks_count(es);
3162 	}
3163 
3164 	buf->f_type = EXT4_SUPER_MAGIC;
3165 	buf->f_bsize = sb->s_blocksize;
3166 	buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3167 	buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3168 		       percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3169 	ext4_free_blocks_count_set(es, buf->f_bfree);
3170 	buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3171 	if (buf->f_bfree < ext4_r_blocks_count(es))
3172 		buf->f_bavail = 0;
3173 	buf->f_files = le32_to_cpu(es->s_inodes_count);
3174 	buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3175 	es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3176 	buf->f_namelen = EXT4_NAME_LEN;
3177 	fsid = le64_to_cpup((void *)es->s_uuid) ^
3178 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3179 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3180 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3181 	return 0;
3182 }
3183 
3184 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3185  * is locked for write. Otherwise the are possible deadlocks:
3186  * Process 1                         Process 2
3187  * ext4_create()                     quota_sync()
3188  *   jbd2_journal_start()                   write_dquot()
3189  *   DQUOT_INIT()                        down(dqio_mutex)
3190  *     down(dqio_mutex)                    jbd2_journal_start()
3191  *
3192  */
3193 
3194 #ifdef CONFIG_QUOTA
3195 
3196 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3197 {
3198 	return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3199 }
3200 
3201 static int ext4_dquot_initialize(struct inode *inode, int type)
3202 {
3203 	handle_t *handle;
3204 	int ret, err;
3205 
3206 	/* We may create quota structure so we need to reserve enough blocks */
3207 	handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
3208 	if (IS_ERR(handle))
3209 		return PTR_ERR(handle);
3210 	ret = dquot_initialize(inode, type);
3211 	err = ext4_journal_stop(handle);
3212 	if (!ret)
3213 		ret = err;
3214 	return ret;
3215 }
3216 
3217 static int ext4_dquot_drop(struct inode *inode)
3218 {
3219 	handle_t *handle;
3220 	int ret, err;
3221 
3222 	/* We may delete quota structure so we need to reserve enough blocks */
3223 	handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
3224 	if (IS_ERR(handle)) {
3225 		/*
3226 		 * We call dquot_drop() anyway to at least release references
3227 		 * to quota structures so that umount does not hang.
3228 		 */
3229 		dquot_drop(inode);
3230 		return PTR_ERR(handle);
3231 	}
3232 	ret = dquot_drop(inode);
3233 	err = ext4_journal_stop(handle);
3234 	if (!ret)
3235 		ret = err;
3236 	return ret;
3237 }
3238 
3239 static int ext4_write_dquot(struct dquot *dquot)
3240 {
3241 	int ret, err;
3242 	handle_t *handle;
3243 	struct inode *inode;
3244 
3245 	inode = dquot_to_inode(dquot);
3246 	handle = ext4_journal_start(inode,
3247 					EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3248 	if (IS_ERR(handle))
3249 		return PTR_ERR(handle);
3250 	ret = dquot_commit(dquot);
3251 	err = ext4_journal_stop(handle);
3252 	if (!ret)
3253 		ret = err;
3254 	return ret;
3255 }
3256 
3257 static int ext4_acquire_dquot(struct dquot *dquot)
3258 {
3259 	int ret, err;
3260 	handle_t *handle;
3261 
3262 	handle = ext4_journal_start(dquot_to_inode(dquot),
3263 					EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3264 	if (IS_ERR(handle))
3265 		return PTR_ERR(handle);
3266 	ret = dquot_acquire(dquot);
3267 	err = ext4_journal_stop(handle);
3268 	if (!ret)
3269 		ret = err;
3270 	return ret;
3271 }
3272 
3273 static int ext4_release_dquot(struct dquot *dquot)
3274 {
3275 	int ret, err;
3276 	handle_t *handle;
3277 
3278 	handle = ext4_journal_start(dquot_to_inode(dquot),
3279 					EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3280 	if (IS_ERR(handle)) {
3281 		/* Release dquot anyway to avoid endless cycle in dqput() */
3282 		dquot_release(dquot);
3283 		return PTR_ERR(handle);
3284 	}
3285 	ret = dquot_release(dquot);
3286 	err = ext4_journal_stop(handle);
3287 	if (!ret)
3288 		ret = err;
3289 	return ret;
3290 }
3291 
3292 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3293 {
3294 	/* Are we journaling quotas? */
3295 	if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3296 	    EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3297 		dquot_mark_dquot_dirty(dquot);
3298 		return ext4_write_dquot(dquot);
3299 	} else {
3300 		return dquot_mark_dquot_dirty(dquot);
3301 	}
3302 }
3303 
3304 static int ext4_write_info(struct super_block *sb, int type)
3305 {
3306 	int ret, err;
3307 	handle_t *handle;
3308 
3309 	/* Data block + inode block */
3310 	handle = ext4_journal_start(sb->s_root->d_inode, 2);
3311 	if (IS_ERR(handle))
3312 		return PTR_ERR(handle);
3313 	ret = dquot_commit_info(sb, type);
3314 	err = ext4_journal_stop(handle);
3315 	if (!ret)
3316 		ret = err;
3317 	return ret;
3318 }
3319 
3320 /*
3321  * Turn on quotas during mount time - we need to find
3322  * the quota file and such...
3323  */
3324 static int ext4_quota_on_mount(struct super_block *sb, int type)
3325 {
3326 	return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3327 			EXT4_SB(sb)->s_jquota_fmt, type);
3328 }
3329 
3330 /*
3331  * Standard function to be called on quota_on
3332  */
3333 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3334 			 char *name, int remount)
3335 {
3336 	int err;
3337 	struct path path;
3338 
3339 	if (!test_opt(sb, QUOTA))
3340 		return -EINVAL;
3341 	/* When remounting, no checks are needed and in fact, name is NULL */
3342 	if (remount)
3343 		return vfs_quota_on(sb, type, format_id, name, remount);
3344 
3345 	err = kern_path(name, LOOKUP_FOLLOW, &path);
3346 	if (err)
3347 		return err;
3348 
3349 	/* Quotafile not on the same filesystem? */
3350 	if (path.mnt->mnt_sb != sb) {
3351 		path_put(&path);
3352 		return -EXDEV;
3353 	}
3354 	/* Journaling quota? */
3355 	if (EXT4_SB(sb)->s_qf_names[type]) {
3356 		/* Quotafile not in fs root? */
3357 		if (path.dentry->d_parent != sb->s_root)
3358 			printk(KERN_WARNING
3359 				"EXT4-fs: Quota file not on filesystem root. "
3360 				"Journaled quota will not work.\n");
3361 	}
3362 
3363 	/*
3364 	 * When we journal data on quota file, we have to flush journal to see
3365 	 * all updates to the file when we bypass pagecache...
3366 	 */
3367 	if (ext4_should_journal_data(path.dentry->d_inode)) {
3368 		/*
3369 		 * We don't need to lock updates but journal_flush() could
3370 		 * otherwise be livelocked...
3371 		 */
3372 		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3373 		err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3374 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3375 		if (err) {
3376 			path_put(&path);
3377 			return err;
3378 		}
3379 	}
3380 
3381 	err = vfs_quota_on_path(sb, type, format_id, &path);
3382 	path_put(&path);
3383 	return err;
3384 }
3385 
3386 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3387  * acquiring the locks... As quota files are never truncated and quota code
3388  * itself serializes the operations (and noone else should touch the files)
3389  * we don't have to be afraid of races */
3390 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3391 			       size_t len, loff_t off)
3392 {
3393 	struct inode *inode = sb_dqopt(sb)->files[type];
3394 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3395 	int err = 0;
3396 	int offset = off & (sb->s_blocksize - 1);
3397 	int tocopy;
3398 	size_t toread;
3399 	struct buffer_head *bh;
3400 	loff_t i_size = i_size_read(inode);
3401 
3402 	if (off > i_size)
3403 		return 0;
3404 	if (off+len > i_size)
3405 		len = i_size-off;
3406 	toread = len;
3407 	while (toread > 0) {
3408 		tocopy = sb->s_blocksize - offset < toread ?
3409 				sb->s_blocksize - offset : toread;
3410 		bh = ext4_bread(NULL, inode, blk, 0, &err);
3411 		if (err)
3412 			return err;
3413 		if (!bh)	/* A hole? */
3414 			memset(data, 0, tocopy);
3415 		else
3416 			memcpy(data, bh->b_data+offset, tocopy);
3417 		brelse(bh);
3418 		offset = 0;
3419 		toread -= tocopy;
3420 		data += tocopy;
3421 		blk++;
3422 	}
3423 	return len;
3424 }
3425 
3426 /* Write to quotafile (we know the transaction is already started and has
3427  * enough credits) */
3428 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3429 				const char *data, size_t len, loff_t off)
3430 {
3431 	struct inode *inode = sb_dqopt(sb)->files[type];
3432 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3433 	int err = 0;
3434 	int offset = off & (sb->s_blocksize - 1);
3435 	int tocopy;
3436 	int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3437 	size_t towrite = len;
3438 	struct buffer_head *bh;
3439 	handle_t *handle = journal_current_handle();
3440 
3441 	if (!handle) {
3442 		printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
3443 			" cancelled because transaction is not started.\n",
3444 			(unsigned long long)off, (unsigned long long)len);
3445 		return -EIO;
3446 	}
3447 	mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3448 	while (towrite > 0) {
3449 		tocopy = sb->s_blocksize - offset < towrite ?
3450 				sb->s_blocksize - offset : towrite;
3451 		bh = ext4_bread(handle, inode, blk, 1, &err);
3452 		if (!bh)
3453 			goto out;
3454 		if (journal_quota) {
3455 			err = ext4_journal_get_write_access(handle, bh);
3456 			if (err) {
3457 				brelse(bh);
3458 				goto out;
3459 			}
3460 		}
3461 		lock_buffer(bh);
3462 		memcpy(bh->b_data+offset, data, tocopy);
3463 		flush_dcache_page(bh->b_page);
3464 		unlock_buffer(bh);
3465 		if (journal_quota)
3466 			err = ext4_journal_dirty_metadata(handle, bh);
3467 		else {
3468 			/* Always do at least ordered writes for quotas */
3469 			err = ext4_jbd2_file_inode(handle, inode);
3470 			mark_buffer_dirty(bh);
3471 		}
3472 		brelse(bh);
3473 		if (err)
3474 			goto out;
3475 		offset = 0;
3476 		towrite -= tocopy;
3477 		data += tocopy;
3478 		blk++;
3479 	}
3480 out:
3481 	if (len == towrite) {
3482 		mutex_unlock(&inode->i_mutex);
3483 		return err;
3484 	}
3485 	if (inode->i_size < off+len-towrite) {
3486 		i_size_write(inode, off+len-towrite);
3487 		EXT4_I(inode)->i_disksize = inode->i_size;
3488 	}
3489 	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3490 	ext4_mark_inode_dirty(handle, inode);
3491 	mutex_unlock(&inode->i_mutex);
3492 	return len - towrite;
3493 }
3494 
3495 #endif
3496 
3497 static int ext4_get_sb(struct file_system_type *fs_type,
3498 	int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3499 {
3500 	return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3501 }
3502 
3503 #ifdef CONFIG_PROC_FS
3504 static int ext4_ui_proc_show(struct seq_file *m, void *v)
3505 {
3506 	unsigned int *p = m->private;
3507 
3508 	seq_printf(m, "%u\n", *p);
3509 	return 0;
3510 }
3511 
3512 static int ext4_ui_proc_open(struct inode *inode, struct file *file)
3513 {
3514 	return single_open(file, ext4_ui_proc_show, PDE(inode)->data);
3515 }
3516 
3517 static ssize_t ext4_ui_proc_write(struct file *file, const char __user *buf,
3518 			       size_t cnt, loff_t *ppos)
3519 {
3520 	unsigned int *p = PDE(file->f_path.dentry->d_inode)->data;
3521 	char str[32];
3522 	unsigned long value;
3523 
3524 	if (cnt >= sizeof(str))
3525 		return -EINVAL;
3526 	if (copy_from_user(str, buf, cnt))
3527 		return -EFAULT;
3528 	value = simple_strtol(str, NULL, 0);
3529 	if (value < 0)
3530 		return -ERANGE;
3531 	*p = value;
3532 	return cnt;
3533 }
3534 
3535 const struct file_operations ext4_ui_proc_fops = {
3536 	.owner		= THIS_MODULE,
3537 	.open		= ext4_ui_proc_open,
3538 	.read		= seq_read,
3539 	.llseek		= seq_lseek,
3540 	.release	= single_release,
3541 	.write		= ext4_ui_proc_write,
3542 };
3543 #endif
3544 
3545 static struct file_system_type ext4_fs_type = {
3546 	.owner		= THIS_MODULE,
3547 	.name		= "ext4",
3548 	.get_sb		= ext4_get_sb,
3549 	.kill_sb	= kill_block_super,
3550 	.fs_flags	= FS_REQUIRES_DEV,
3551 };
3552 
3553 #ifdef CONFIG_EXT4DEV_COMPAT
3554 static int ext4dev_get_sb(struct file_system_type *fs_type,
3555 	int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3556 {
3557 	printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3558 	       "to mount using ext4\n");
3559 	printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3560 	       "will go away by 2.6.31\n");
3561 	return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3562 }
3563 
3564 static struct file_system_type ext4dev_fs_type = {
3565 	.owner		= THIS_MODULE,
3566 	.name		= "ext4dev",
3567 	.get_sb		= ext4dev_get_sb,
3568 	.kill_sb	= kill_block_super,
3569 	.fs_flags	= FS_REQUIRES_DEV,
3570 };
3571 MODULE_ALIAS("ext4dev");
3572 #endif
3573 
3574 static int __init init_ext4_fs(void)
3575 {
3576 	int err;
3577 
3578 	ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3579 	err = init_ext4_mballoc();
3580 	if (err)
3581 		return err;
3582 
3583 	err = init_ext4_xattr();
3584 	if (err)
3585 		goto out2;
3586 	err = init_inodecache();
3587 	if (err)
3588 		goto out1;
3589 	err = register_filesystem(&ext4_fs_type);
3590 	if (err)
3591 		goto out;
3592 #ifdef CONFIG_EXT4DEV_COMPAT
3593 	err = register_filesystem(&ext4dev_fs_type);
3594 	if (err) {
3595 		unregister_filesystem(&ext4_fs_type);
3596 		goto out;
3597 	}
3598 #endif
3599 	return 0;
3600 out:
3601 	destroy_inodecache();
3602 out1:
3603 	exit_ext4_xattr();
3604 out2:
3605 	exit_ext4_mballoc();
3606 	return err;
3607 }
3608 
3609 static void __exit exit_ext4_fs(void)
3610 {
3611 	unregister_filesystem(&ext4_fs_type);
3612 #ifdef CONFIG_EXT4DEV_COMPAT
3613 	unregister_filesystem(&ext4dev_fs_type);
3614 #endif
3615 	destroy_inodecache();
3616 	exit_ext4_xattr();
3617 	exit_ext4_mballoc();
3618 	remove_proc_entry("fs/ext4", NULL);
3619 }
3620 
3621 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3622 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
3623 MODULE_LICENSE("GPL");
3624 module_init(init_ext4_fs)
3625 module_exit(exit_ext4_fs)
3626