2 * linux/fs/ext4/ialloc.c
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)
9 * BSD ufs-inspired inode and directory allocation by
10 * Stephen Tweedie (sct@redhat.com), 1993
11 * Big-endian to little-endian byte-swapping/bitmaps by
12 * David S. Miller (davem@caip.rutgers.edu), 1995
15 #include <linux/time.h>
17 #include <linux/stat.h>
18 #include <linux/string.h>
19 #include <linux/quotaops.h>
20 #include <linux/buffer_head.h>
21 #include <linux/random.h>
22 #include <linux/bitops.h>
23 #include <linux/blkdev.h>
24 #include <linux/cred.h>
26 #include <asm/byteorder.h>
29 #include "ext4_jbd2.h"
33 #include <trace/events/ext4.h>
36 * ialloc.c contains the inodes allocation and deallocation routines
40 * The free inodes are managed by bitmaps. A file system contains several
41 * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap
42 * block for inodes, N blocks for the inode table and data blocks.
44 * The file system contains group descriptors which are located after the
45 * super block. Each descriptor contains the number of the bitmap block and
46 * the free blocks count in the block.
50 * To avoid calling the atomic setbit hundreds or thousands of times, we only
51 * need to use it within a single byte (to ensure we get endianness right).
52 * We can use memset for the rest of the bitmap as there are no other users.
54 void ext4_mark_bitmap_end(int start_bit
, int end_bit
, char *bitmap
)
58 if (start_bit
>= end_bit
)
61 ext4_debug("mark end bits +%d through +%d used\n", start_bit
, end_bit
);
62 for (i
= start_bit
; i
< ((start_bit
+ 7) & ~7UL); i
++)
63 ext4_set_bit(i
, bitmap
);
65 memset(bitmap
+ (i
>> 3), 0xff, (end_bit
- i
) >> 3);
68 /* Initializes an uninitialized inode bitmap */
69 static int ext4_init_inode_bitmap(struct super_block
*sb
,
70 struct buffer_head
*bh
,
71 ext4_group_t block_group
,
72 struct ext4_group_desc
*gdp
)
74 struct ext4_group_info
*grp
;
75 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
76 J_ASSERT_BH(bh
, buffer_locked(bh
));
78 /* If checksum is bad mark all blocks and inodes use to prevent
79 * allocation, essentially implementing a per-group read-only flag. */
80 if (!ext4_group_desc_csum_verify(sb
, block_group
, gdp
)) {
81 grp
= ext4_get_group_info(sb
, block_group
);
82 if (!EXT4_MB_GRP_BBITMAP_CORRUPT(grp
))
83 percpu_counter_sub(&sbi
->s_freeclusters_counter
,
85 set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT
, &grp
->bb_state
);
86 if (!EXT4_MB_GRP_IBITMAP_CORRUPT(grp
)) {
88 count
= ext4_free_inodes_count(sb
, gdp
);
89 percpu_counter_sub(&sbi
->s_freeinodes_counter
,
92 set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT
, &grp
->bb_state
);
96 memset(bh
->b_data
, 0, (EXT4_INODES_PER_GROUP(sb
) + 7) / 8);
97 ext4_mark_bitmap_end(EXT4_INODES_PER_GROUP(sb
), sb
->s_blocksize
* 8,
99 ext4_inode_bitmap_csum_set(sb
, block_group
, gdp
, bh
,
100 EXT4_INODES_PER_GROUP(sb
) / 8);
101 ext4_group_desc_csum_set(sb
, block_group
, gdp
);
106 void ext4_end_bitmap_read(struct buffer_head
*bh
, int uptodate
)
109 set_buffer_uptodate(bh
);
110 set_bitmap_uptodate(bh
);
116 static int ext4_validate_inode_bitmap(struct super_block
*sb
,
117 struct ext4_group_desc
*desc
,
118 ext4_group_t block_group
,
119 struct buffer_head
*bh
)
122 struct ext4_group_info
*grp
= ext4_get_group_info(sb
, block_group
);
123 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
125 if (buffer_verified(bh
))
127 if (EXT4_MB_GRP_IBITMAP_CORRUPT(grp
))
128 return -EFSCORRUPTED
;
130 ext4_lock_group(sb
, block_group
);
131 blk
= ext4_inode_bitmap(sb
, desc
);
132 if (!ext4_inode_bitmap_csum_verify(sb
, block_group
, desc
, bh
,
133 EXT4_INODES_PER_GROUP(sb
) / 8)) {
134 ext4_unlock_group(sb
, block_group
);
135 ext4_error(sb
, "Corrupt inode bitmap - block_group = %u, "
136 "inode_bitmap = %llu", block_group
, blk
);
137 grp
= ext4_get_group_info(sb
, block_group
);
138 if (!EXT4_MB_GRP_IBITMAP_CORRUPT(grp
)) {
140 count
= ext4_free_inodes_count(sb
, desc
);
141 percpu_counter_sub(&sbi
->s_freeinodes_counter
,
144 set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT
, &grp
->bb_state
);
147 set_buffer_verified(bh
);
148 ext4_unlock_group(sb
, block_group
);
153 * Read the inode allocation bitmap for a given block_group, reading
154 * into the specified slot in the superblock's bitmap cache.
156 * Return buffer_head of bitmap on success or NULL.
158 static struct buffer_head
*
159 ext4_read_inode_bitmap(struct super_block
*sb
, ext4_group_t block_group
)
161 struct ext4_group_desc
*desc
;
162 struct buffer_head
*bh
= NULL
;
163 ext4_fsblk_t bitmap_blk
;
166 desc
= ext4_get_group_desc(sb
, block_group
, NULL
);
168 return ERR_PTR(-EFSCORRUPTED
);
170 bitmap_blk
= ext4_inode_bitmap(sb
, desc
);
171 bh
= sb_getblk(sb
, bitmap_blk
);
173 ext4_error(sb
, "Cannot read inode bitmap - "
174 "block_group = %u, inode_bitmap = %llu",
175 block_group
, bitmap_blk
);
176 return ERR_PTR(-EIO
);
178 if (bitmap_uptodate(bh
))
182 if (bitmap_uptodate(bh
)) {
187 ext4_lock_group(sb
, block_group
);
188 if (desc
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_UNINIT
)) {
189 err
= ext4_init_inode_bitmap(sb
, bh
, block_group
, desc
);
190 set_bitmap_uptodate(bh
);
191 set_buffer_uptodate(bh
);
192 set_buffer_verified(bh
);
193 ext4_unlock_group(sb
, block_group
);
196 ext4_error(sb
, "Failed to init inode bitmap for group "
197 "%u: %d", block_group
, err
);
202 ext4_unlock_group(sb
, block_group
);
204 if (buffer_uptodate(bh
)) {
206 * if not uninit if bh is uptodate,
207 * bitmap is also uptodate
209 set_bitmap_uptodate(bh
);
214 * submit the buffer_head for reading
216 trace_ext4_load_inode_bitmap(sb
, block_group
);
217 bh
->b_end_io
= ext4_end_bitmap_read
;
219 submit_bh(REQ_OP_READ
, REQ_META
| REQ_PRIO
, bh
);
221 if (!buffer_uptodate(bh
)) {
223 ext4_error(sb
, "Cannot read inode bitmap - "
224 "block_group = %u, inode_bitmap = %llu",
225 block_group
, bitmap_blk
);
226 return ERR_PTR(-EIO
);
230 err
= ext4_validate_inode_bitmap(sb
, desc
, block_group
, bh
);
240 * NOTE! When we get the inode, we're the only people
241 * that have access to it, and as such there are no
242 * race conditions we have to worry about. The inode
243 * is not on the hash-lists, and it cannot be reached
244 * through the filesystem because the directory entry
245 * has been deleted earlier.
247 * HOWEVER: we must make sure that we get no aliases,
248 * which means that we have to call "clear_inode()"
249 * _before_ we mark the inode not in use in the inode
250 * bitmaps. Otherwise a newly created file might use
251 * the same inode number (not actually the same pointer
252 * though), and then we'd have two inodes sharing the
253 * same inode number and space on the harddisk.
255 void ext4_free_inode(handle_t
*handle
, struct inode
*inode
)
257 struct super_block
*sb
= inode
->i_sb
;
260 struct buffer_head
*bitmap_bh
= NULL
;
261 struct buffer_head
*bh2
;
262 ext4_group_t block_group
;
264 struct ext4_group_desc
*gdp
;
265 struct ext4_super_block
*es
;
266 struct ext4_sb_info
*sbi
;
267 int fatal
= 0, err
, count
, cleared
;
268 struct ext4_group_info
*grp
;
271 printk(KERN_ERR
"EXT4-fs: %s:%d: inode on "
272 "nonexistent device\n", __func__
, __LINE__
);
275 if (atomic_read(&inode
->i_count
) > 1) {
276 ext4_msg(sb
, KERN_ERR
, "%s:%d: inode #%lu: count=%d",
277 __func__
, __LINE__
, inode
->i_ino
,
278 atomic_read(&inode
->i_count
));
281 if (inode
->i_nlink
) {
282 ext4_msg(sb
, KERN_ERR
, "%s:%d: inode #%lu: nlink=%d\n",
283 __func__
, __LINE__
, inode
->i_ino
, inode
->i_nlink
);
289 ext4_debug("freeing inode %lu\n", ino
);
290 trace_ext4_free_inode(inode
);
293 * Note: we must free any quota before locking the superblock,
294 * as writing the quota to disk may need the lock as well.
296 dquot_initialize(inode
);
297 dquot_free_inode(inode
);
300 is_directory
= S_ISDIR(inode
->i_mode
);
302 /* Do this BEFORE marking the inode not in use or returning an error */
303 ext4_clear_inode(inode
);
305 es
= EXT4_SB(sb
)->s_es
;
306 if (ino
< EXT4_FIRST_INO(sb
) || ino
> le32_to_cpu(es
->s_inodes_count
)) {
307 ext4_error(sb
, "reserved or nonexistent inode %lu", ino
);
310 block_group
= (ino
- 1) / EXT4_INODES_PER_GROUP(sb
);
311 bit
= (ino
- 1) % EXT4_INODES_PER_GROUP(sb
);
312 bitmap_bh
= ext4_read_inode_bitmap(sb
, block_group
);
313 /* Don't bother if the inode bitmap is corrupt. */
314 grp
= ext4_get_group_info(sb
, block_group
);
315 if (IS_ERR(bitmap_bh
)) {
316 fatal
= PTR_ERR(bitmap_bh
);
320 if (unlikely(EXT4_MB_GRP_IBITMAP_CORRUPT(grp
))) {
321 fatal
= -EFSCORRUPTED
;
325 BUFFER_TRACE(bitmap_bh
, "get_write_access");
326 fatal
= ext4_journal_get_write_access(handle
, bitmap_bh
);
331 gdp
= ext4_get_group_desc(sb
, block_group
, &bh2
);
333 BUFFER_TRACE(bh2
, "get_write_access");
334 fatal
= ext4_journal_get_write_access(handle
, bh2
);
336 ext4_lock_group(sb
, block_group
);
337 cleared
= ext4_test_and_clear_bit(bit
, bitmap_bh
->b_data
);
338 if (fatal
|| !cleared
) {
339 ext4_unlock_group(sb
, block_group
);
343 count
= ext4_free_inodes_count(sb
, gdp
) + 1;
344 ext4_free_inodes_set(sb
, gdp
, count
);
346 count
= ext4_used_dirs_count(sb
, gdp
) - 1;
347 ext4_used_dirs_set(sb
, gdp
, count
);
348 percpu_counter_dec(&sbi
->s_dirs_counter
);
350 ext4_inode_bitmap_csum_set(sb
, block_group
, gdp
, bitmap_bh
,
351 EXT4_INODES_PER_GROUP(sb
) / 8);
352 ext4_group_desc_csum_set(sb
, block_group
, gdp
);
353 ext4_unlock_group(sb
, block_group
);
355 percpu_counter_inc(&sbi
->s_freeinodes_counter
);
356 if (sbi
->s_log_groups_per_flex
) {
357 ext4_group_t f
= ext4_flex_group(sbi
, block_group
);
359 atomic_inc(&sbi
->s_flex_groups
[f
].free_inodes
);
361 atomic_dec(&sbi
->s_flex_groups
[f
].used_dirs
);
363 BUFFER_TRACE(bh2
, "call ext4_handle_dirty_metadata");
364 fatal
= ext4_handle_dirty_metadata(handle
, NULL
, bh2
);
367 BUFFER_TRACE(bitmap_bh
, "call ext4_handle_dirty_metadata");
368 err
= ext4_handle_dirty_metadata(handle
, NULL
, bitmap_bh
);
372 ext4_error(sb
, "bit already cleared for inode %lu", ino
);
373 if (gdp
&& !EXT4_MB_GRP_IBITMAP_CORRUPT(grp
)) {
375 count
= ext4_free_inodes_count(sb
, gdp
);
376 percpu_counter_sub(&sbi
->s_freeinodes_counter
,
379 set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT
, &grp
->bb_state
);
384 ext4_std_error(sb
, fatal
);
394 * Helper function for Orlov's allocator; returns critical information
395 * for a particular block group or flex_bg. If flex_size is 1, then g
396 * is a block group number; otherwise it is flex_bg number.
398 static void get_orlov_stats(struct super_block
*sb
, ext4_group_t g
,
399 int flex_size
, struct orlov_stats
*stats
)
401 struct ext4_group_desc
*desc
;
402 struct flex_groups
*flex_group
= EXT4_SB(sb
)->s_flex_groups
;
405 stats
->free_inodes
= atomic_read(&flex_group
[g
].free_inodes
);
406 stats
->free_clusters
= atomic64_read(&flex_group
[g
].free_clusters
);
407 stats
->used_dirs
= atomic_read(&flex_group
[g
].used_dirs
);
411 desc
= ext4_get_group_desc(sb
, g
, NULL
);
413 stats
->free_inodes
= ext4_free_inodes_count(sb
, desc
);
414 stats
->free_clusters
= ext4_free_group_clusters(sb
, desc
);
415 stats
->used_dirs
= ext4_used_dirs_count(sb
, desc
);
417 stats
->free_inodes
= 0;
418 stats
->free_clusters
= 0;
419 stats
->used_dirs
= 0;
424 * Orlov's allocator for directories.
426 * We always try to spread first-level directories.
428 * If there are blockgroups with both free inodes and free blocks counts
429 * not worse than average we return one with smallest directory count.
430 * Otherwise we simply return a random group.
432 * For the rest rules look so:
434 * It's OK to put directory into a group unless
435 * it has too many directories already (max_dirs) or
436 * it has too few free inodes left (min_inodes) or
437 * it has too few free blocks left (min_blocks) or
438 * Parent's group is preferred, if it doesn't satisfy these
439 * conditions we search cyclically through the rest. If none
440 * of the groups look good we just look for a group with more
441 * free inodes than average (starting at parent's group).
444 static int find_group_orlov(struct super_block
*sb
, struct inode
*parent
,
445 ext4_group_t
*group
, umode_t mode
,
446 const struct qstr
*qstr
)
448 ext4_group_t parent_group
= EXT4_I(parent
)->i_block_group
;
449 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
450 ext4_group_t real_ngroups
= ext4_get_groups_count(sb
);
451 int inodes_per_group
= EXT4_INODES_PER_GROUP(sb
);
452 unsigned int freei
, avefreei
, grp_free
;
453 ext4_fsblk_t freeb
, avefreec
;
455 int max_dirs
, min_inodes
;
456 ext4_grpblk_t min_clusters
;
457 ext4_group_t i
, grp
, g
, ngroups
;
458 struct ext4_group_desc
*desc
;
459 struct orlov_stats stats
;
460 int flex_size
= ext4_flex_bg_size(sbi
);
461 struct dx_hash_info hinfo
;
463 ngroups
= real_ngroups
;
465 ngroups
= (real_ngroups
+ flex_size
- 1) >>
466 sbi
->s_log_groups_per_flex
;
467 parent_group
>>= sbi
->s_log_groups_per_flex
;
470 freei
= percpu_counter_read_positive(&sbi
->s_freeinodes_counter
);
471 avefreei
= freei
/ ngroups
;
472 freeb
= EXT4_C2B(sbi
,
473 percpu_counter_read_positive(&sbi
->s_freeclusters_counter
));
475 do_div(avefreec
, ngroups
);
476 ndirs
= percpu_counter_read_positive(&sbi
->s_dirs_counter
);
479 ((parent
== d_inode(sb
->s_root
)) ||
480 (ext4_test_inode_flag(parent
, EXT4_INODE_TOPDIR
)))) {
481 int best_ndir
= inodes_per_group
;
485 hinfo
.hash_version
= DX_HASH_HALF_MD4
;
486 hinfo
.seed
= sbi
->s_hash_seed
;
487 ext4fs_dirhash(qstr
->name
, qstr
->len
, &hinfo
);
491 parent_group
= (unsigned)grp
% ngroups
;
492 for (i
= 0; i
< ngroups
; i
++) {
493 g
= (parent_group
+ i
) % ngroups
;
494 get_orlov_stats(sb
, g
, flex_size
, &stats
);
495 if (!stats
.free_inodes
)
497 if (stats
.used_dirs
>= best_ndir
)
499 if (stats
.free_inodes
< avefreei
)
501 if (stats
.free_clusters
< avefreec
)
505 best_ndir
= stats
.used_dirs
;
510 if (flex_size
== 1) {
516 * We pack inodes at the beginning of the flexgroup's
517 * inode tables. Block allocation decisions will do
518 * something similar, although regular files will
519 * start at 2nd block group of the flexgroup. See
520 * ext4_ext_find_goal() and ext4_find_near().
523 for (i
= 0; i
< flex_size
; i
++) {
524 if (grp
+i
>= real_ngroups
)
526 desc
= ext4_get_group_desc(sb
, grp
+i
, NULL
);
527 if (desc
&& ext4_free_inodes_count(sb
, desc
)) {
535 max_dirs
= ndirs
/ ngroups
+ inodes_per_group
/ 16;
536 min_inodes
= avefreei
- inodes_per_group
*flex_size
/ 4;
539 min_clusters
= avefreec
- EXT4_CLUSTERS_PER_GROUP(sb
)*flex_size
/ 4;
542 * Start looking in the flex group where we last allocated an
543 * inode for this parent directory
545 if (EXT4_I(parent
)->i_last_alloc_group
!= ~0) {
546 parent_group
= EXT4_I(parent
)->i_last_alloc_group
;
548 parent_group
>>= sbi
->s_log_groups_per_flex
;
551 for (i
= 0; i
< ngroups
; i
++) {
552 grp
= (parent_group
+ i
) % ngroups
;
553 get_orlov_stats(sb
, grp
, flex_size
, &stats
);
554 if (stats
.used_dirs
>= max_dirs
)
556 if (stats
.free_inodes
< min_inodes
)
558 if (stats
.free_clusters
< min_clusters
)
564 ngroups
= real_ngroups
;
565 avefreei
= freei
/ ngroups
;
567 parent_group
= EXT4_I(parent
)->i_block_group
;
568 for (i
= 0; i
< ngroups
; i
++) {
569 grp
= (parent_group
+ i
) % ngroups
;
570 desc
= ext4_get_group_desc(sb
, grp
, NULL
);
572 grp_free
= ext4_free_inodes_count(sb
, desc
);
573 if (grp_free
&& grp_free
>= avefreei
) {
582 * The free-inodes counter is approximate, and for really small
583 * filesystems the above test can fail to find any blockgroups
592 static int find_group_other(struct super_block
*sb
, struct inode
*parent
,
593 ext4_group_t
*group
, umode_t mode
)
595 ext4_group_t parent_group
= EXT4_I(parent
)->i_block_group
;
596 ext4_group_t i
, last
, ngroups
= ext4_get_groups_count(sb
);
597 struct ext4_group_desc
*desc
;
598 int flex_size
= ext4_flex_bg_size(EXT4_SB(sb
));
601 * Try to place the inode is the same flex group as its
602 * parent. If we can't find space, use the Orlov algorithm to
603 * find another flex group, and store that information in the
604 * parent directory's inode information so that use that flex
605 * group for future allocations.
611 parent_group
&= ~(flex_size
-1);
612 last
= parent_group
+ flex_size
;
615 for (i
= parent_group
; i
< last
; i
++) {
616 desc
= ext4_get_group_desc(sb
, i
, NULL
);
617 if (desc
&& ext4_free_inodes_count(sb
, desc
)) {
622 if (!retry
&& EXT4_I(parent
)->i_last_alloc_group
!= ~0) {
624 parent_group
= EXT4_I(parent
)->i_last_alloc_group
;
628 * If this didn't work, use the Orlov search algorithm
629 * to find a new flex group; we pass in the mode to
630 * avoid the topdir algorithms.
632 *group
= parent_group
+ flex_size
;
633 if (*group
> ngroups
)
635 return find_group_orlov(sb
, parent
, group
, mode
, NULL
);
639 * Try to place the inode in its parent directory
641 *group
= parent_group
;
642 desc
= ext4_get_group_desc(sb
, *group
, NULL
);
643 if (desc
&& ext4_free_inodes_count(sb
, desc
) &&
644 ext4_free_group_clusters(sb
, desc
))
648 * We're going to place this inode in a different blockgroup from its
649 * parent. We want to cause files in a common directory to all land in
650 * the same blockgroup. But we want files which are in a different
651 * directory which shares a blockgroup with our parent to land in a
652 * different blockgroup.
654 * So add our directory's i_ino into the starting point for the hash.
656 *group
= (*group
+ parent
->i_ino
) % ngroups
;
659 * Use a quadratic hash to find a group with a free inode and some free
662 for (i
= 1; i
< ngroups
; i
<<= 1) {
664 if (*group
>= ngroups
)
666 desc
= ext4_get_group_desc(sb
, *group
, NULL
);
667 if (desc
&& ext4_free_inodes_count(sb
, desc
) &&
668 ext4_free_group_clusters(sb
, desc
))
673 * That failed: try linear search for a free inode, even if that group
674 * has no free blocks.
676 *group
= parent_group
;
677 for (i
= 0; i
< ngroups
; i
++) {
678 if (++*group
>= ngroups
)
680 desc
= ext4_get_group_desc(sb
, *group
, NULL
);
681 if (desc
&& ext4_free_inodes_count(sb
, desc
))
689 * In no journal mode, if an inode has recently been deleted, we want
690 * to avoid reusing it until we're reasonably sure the inode table
691 * block has been written back to disk. (Yes, these values are
692 * somewhat arbitrary...)
694 #define RECENTCY_MIN 5
695 #define RECENTCY_DIRTY 30
697 static int recently_deleted(struct super_block
*sb
, ext4_group_t group
, int ino
)
699 struct ext4_group_desc
*gdp
;
700 struct ext4_inode
*raw_inode
;
701 struct buffer_head
*bh
;
702 unsigned long dtime
, now
;
703 int inodes_per_block
= EXT4_SB(sb
)->s_inodes_per_block
;
704 int offset
, ret
= 0, recentcy
= RECENTCY_MIN
;
706 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
710 bh
= sb_getblk(sb
, ext4_inode_table(sb
, gdp
) +
711 (ino
/ inodes_per_block
));
712 if (unlikely(!bh
) || !buffer_uptodate(bh
))
714 * If the block is not in the buffer cache, then it
715 * must have been written out.
719 offset
= (ino
% inodes_per_block
) * EXT4_INODE_SIZE(sb
);
720 raw_inode
= (struct ext4_inode
*) (bh
->b_data
+ offset
);
721 dtime
= le32_to_cpu(raw_inode
->i_dtime
);
723 if (buffer_dirty(bh
))
724 recentcy
+= RECENTCY_DIRTY
;
726 if (dtime
&& (dtime
< now
) && (now
< dtime
+ recentcy
))
734 * There are two policies for allocating an inode. If the new inode is
735 * a directory, then a forward search is made for a block group with both
736 * free space and a low directory-to-inode ratio; if that fails, then of
737 * the groups with above-average free space, that group with the fewest
738 * directories already is chosen.
740 * For other inodes, search forward from the parent directory's block
741 * group to find a free inode.
743 struct inode
*__ext4_new_inode(handle_t
*handle
, struct inode
*dir
,
744 umode_t mode
, const struct qstr
*qstr
,
745 __u32 goal
, uid_t
*owner
, __u32 i_flags
,
746 int handle_type
, unsigned int line_no
,
749 struct super_block
*sb
;
750 struct buffer_head
*inode_bitmap_bh
= NULL
;
751 struct buffer_head
*group_desc_bh
;
752 ext4_group_t ngroups
, group
= 0;
753 unsigned long ino
= 0;
755 struct ext4_group_desc
*gdp
= NULL
;
756 struct ext4_inode_info
*ei
;
757 struct ext4_sb_info
*sbi
;
761 ext4_group_t flex_group
;
762 struct ext4_group_info
*grp
;
765 /* Cannot create files in a deleted directory */
766 if (!dir
|| !dir
->i_nlink
)
767 return ERR_PTR(-EPERM
);
772 if (unlikely(ext4_forced_shutdown(sbi
)))
773 return ERR_PTR(-EIO
);
775 /* Supplied owner must be valid */
776 if (owner
&& (owner
[0] == (uid_t
)-1 || owner
[1] == (uid_t
)-1))
777 return ERR_PTR(-EOVERFLOW
);
779 if ((ext4_encrypted_inode(dir
) || DUMMY_ENCRYPTION_ENABLED(sbi
)) &&
780 (S_ISREG(mode
) || S_ISDIR(mode
) || S_ISLNK(mode
)) &&
781 !(i_flags
& EXT4_EA_INODE_FL
)) {
782 err
= fscrypt_get_encryption_info(dir
);
785 if (!fscrypt_has_encryption_key(dir
))
786 return ERR_PTR(-ENOKEY
);
790 if (!handle
&& sbi
->s_journal
&& !(i_flags
& EXT4_EA_INODE_FL
)) {
791 #ifdef CONFIG_EXT4_FS_POSIX_ACL
792 struct posix_acl
*p
= get_acl(dir
, ACL_TYPE_DEFAULT
);
795 int acl_size
= p
->a_count
* sizeof(ext4_acl_entry
);
797 nblocks
+= (S_ISDIR(mode
) ? 2 : 1) *
798 __ext4_xattr_set_credits(sb
, NULL
/* inode */,
799 NULL
/* block_bh */, acl_size
,
800 true /* is_create */);
801 posix_acl_release(p
);
805 #ifdef CONFIG_SECURITY
807 int num_security_xattrs
= 1;
809 #ifdef CONFIG_INTEGRITY
810 num_security_xattrs
++;
813 * We assume that security xattrs are never
814 * more than 1k. In practice they are under
817 nblocks
+= num_security_xattrs
*
818 __ext4_xattr_set_credits(sb
, NULL
/* inode */,
819 NULL
/* block_bh */, 1024,
820 true /* is_create */);
824 nblocks
+= __ext4_xattr_set_credits(sb
,
825 NULL
/* inode */, NULL
/* block_bh */,
826 FSCRYPT_SET_CONTEXT_MAX_SIZE
,
827 true /* is_create */);
830 ngroups
= ext4_get_groups_count(sb
);
831 trace_ext4_request_inode(dir
, mode
);
832 inode
= new_inode(sb
);
834 return ERR_PTR(-ENOMEM
);
838 * Initialize owners and quota early so that we don't have to account
839 * for quota initialization worst case in standard inode creating
843 inode
->i_mode
= mode
;
844 i_uid_write(inode
, owner
[0]);
845 i_gid_write(inode
, owner
[1]);
846 } else if (test_opt(sb
, GRPID
)) {
847 inode
->i_mode
= mode
;
848 inode
->i_uid
= current_fsuid();
849 inode
->i_gid
= dir
->i_gid
;
851 inode_init_owner(inode
, dir
, mode
);
853 if (ext4_has_feature_project(sb
) &&
854 ext4_test_inode_flag(dir
, EXT4_INODE_PROJINHERIT
))
855 ei
->i_projid
= EXT4_I(dir
)->i_projid
;
857 ei
->i_projid
= make_kprojid(sb
->s_user_ns
, EXT4_DEF_PROJID
);
859 err
= dquot_initialize(inode
);
864 goal
= sbi
->s_inode_goal
;
866 if (goal
&& goal
<= le32_to_cpu(sbi
->s_es
->s_inodes_count
)) {
867 group
= (goal
- 1) / EXT4_INODES_PER_GROUP(sb
);
868 ino
= (goal
- 1) % EXT4_INODES_PER_GROUP(sb
);
874 ret2
= find_group_orlov(sb
, dir
, &group
, mode
, qstr
);
876 ret2
= find_group_other(sb
, dir
, &group
, mode
);
879 EXT4_I(dir
)->i_last_alloc_group
= group
;
885 * Normally we will only go through one pass of this loop,
886 * unless we get unlucky and it turns out the group we selected
887 * had its last inode grabbed by someone else.
889 for (i
= 0; i
< ngroups
; i
++, ino
= 0) {
892 gdp
= ext4_get_group_desc(sb
, group
, &group_desc_bh
);
897 * Check free inodes count before loading bitmap.
899 if (ext4_free_inodes_count(sb
, gdp
) == 0) {
900 if (++group
== ngroups
)
905 grp
= ext4_get_group_info(sb
, group
);
906 /* Skip groups with already-known suspicious inode tables */
907 if (EXT4_MB_GRP_IBITMAP_CORRUPT(grp
)) {
908 if (++group
== ngroups
)
913 brelse(inode_bitmap_bh
);
914 inode_bitmap_bh
= ext4_read_inode_bitmap(sb
, group
);
915 /* Skip groups with suspicious inode tables */
916 if (EXT4_MB_GRP_IBITMAP_CORRUPT(grp
) ||
917 IS_ERR(inode_bitmap_bh
)) {
918 inode_bitmap_bh
= NULL
;
919 if (++group
== ngroups
)
924 repeat_in_this_group
:
925 ino
= ext4_find_next_zero_bit((unsigned long *)
926 inode_bitmap_bh
->b_data
,
927 EXT4_INODES_PER_GROUP(sb
), ino
);
928 if (ino
>= EXT4_INODES_PER_GROUP(sb
))
930 if (group
== 0 && (ino
+1) < EXT4_FIRST_INO(sb
)) {
931 ext4_error(sb
, "reserved inode found cleared - "
932 "inode=%lu", ino
+ 1);
935 if ((EXT4_SB(sb
)->s_journal
== NULL
) &&
936 recently_deleted(sb
, group
, ino
)) {
941 BUG_ON(nblocks
<= 0);
942 handle
= __ext4_journal_start_sb(dir
->i_sb
, line_no
,
943 handle_type
, nblocks
,
945 if (IS_ERR(handle
)) {
946 err
= PTR_ERR(handle
);
947 ext4_std_error(sb
, err
);
951 BUFFER_TRACE(inode_bitmap_bh
, "get_write_access");
952 err
= ext4_journal_get_write_access(handle
, inode_bitmap_bh
);
954 ext4_std_error(sb
, err
);
957 ext4_lock_group(sb
, group
);
958 ret2
= ext4_test_and_set_bit(ino
, inode_bitmap_bh
->b_data
);
959 ext4_unlock_group(sb
, group
);
960 ino
++; /* the inode bitmap is zero-based */
962 goto got
; /* we grabbed the inode! */
964 if (ino
< EXT4_INODES_PER_GROUP(sb
))
965 goto repeat_in_this_group
;
967 if (++group
== ngroups
)
974 BUFFER_TRACE(inode_bitmap_bh
, "call ext4_handle_dirty_metadata");
975 err
= ext4_handle_dirty_metadata(handle
, NULL
, inode_bitmap_bh
);
977 ext4_std_error(sb
, err
);
981 BUFFER_TRACE(group_desc_bh
, "get_write_access");
982 err
= ext4_journal_get_write_access(handle
, group_desc_bh
);
984 ext4_std_error(sb
, err
);
988 /* We may have to initialize the block bitmap if it isn't already */
989 if (ext4_has_group_desc_csum(sb
) &&
990 gdp
->bg_flags
& cpu_to_le16(EXT4_BG_BLOCK_UNINIT
)) {
991 struct buffer_head
*block_bitmap_bh
;
993 block_bitmap_bh
= ext4_read_block_bitmap(sb
, group
);
994 if (IS_ERR(block_bitmap_bh
)) {
995 err
= PTR_ERR(block_bitmap_bh
);
998 BUFFER_TRACE(block_bitmap_bh
, "get block bitmap access");
999 err
= ext4_journal_get_write_access(handle
, block_bitmap_bh
);
1001 brelse(block_bitmap_bh
);
1002 ext4_std_error(sb
, err
);
1006 BUFFER_TRACE(block_bitmap_bh
, "dirty block bitmap");
1007 err
= ext4_handle_dirty_metadata(handle
, NULL
, block_bitmap_bh
);
1009 /* recheck and clear flag under lock if we still need to */
1010 ext4_lock_group(sb
, group
);
1011 if (gdp
->bg_flags
& cpu_to_le16(EXT4_BG_BLOCK_UNINIT
)) {
1012 gdp
->bg_flags
&= cpu_to_le16(~EXT4_BG_BLOCK_UNINIT
);
1013 ext4_free_group_clusters_set(sb
, gdp
,
1014 ext4_free_clusters_after_init(sb
, group
, gdp
));
1015 ext4_block_bitmap_csum_set(sb
, group
, gdp
,
1017 ext4_group_desc_csum_set(sb
, group
, gdp
);
1019 ext4_unlock_group(sb
, group
);
1020 brelse(block_bitmap_bh
);
1023 ext4_std_error(sb
, err
);
1028 /* Update the relevant bg descriptor fields */
1029 if (ext4_has_group_desc_csum(sb
)) {
1031 struct ext4_group_info
*grp
= ext4_get_group_info(sb
, group
);
1033 down_read(&grp
->alloc_sem
); /* protect vs itable lazyinit */
1034 ext4_lock_group(sb
, group
); /* while we modify the bg desc */
1035 free
= EXT4_INODES_PER_GROUP(sb
) -
1036 ext4_itable_unused_count(sb
, gdp
);
1037 if (gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_UNINIT
)) {
1038 gdp
->bg_flags
&= cpu_to_le16(~EXT4_BG_INODE_UNINIT
);
1042 * Check the relative inode number against the last used
1043 * relative inode number in this group. if it is greater
1044 * we need to update the bg_itable_unused count
1047 ext4_itable_unused_set(sb
, gdp
,
1048 (EXT4_INODES_PER_GROUP(sb
) - ino
));
1049 up_read(&grp
->alloc_sem
);
1051 ext4_lock_group(sb
, group
);
1054 ext4_free_inodes_set(sb
, gdp
, ext4_free_inodes_count(sb
, gdp
) - 1);
1055 if (S_ISDIR(mode
)) {
1056 ext4_used_dirs_set(sb
, gdp
, ext4_used_dirs_count(sb
, gdp
) + 1);
1057 if (sbi
->s_log_groups_per_flex
) {
1058 ext4_group_t f
= ext4_flex_group(sbi
, group
);
1060 atomic_inc(&sbi
->s_flex_groups
[f
].used_dirs
);
1063 if (ext4_has_group_desc_csum(sb
)) {
1064 ext4_inode_bitmap_csum_set(sb
, group
, gdp
, inode_bitmap_bh
,
1065 EXT4_INODES_PER_GROUP(sb
) / 8);
1066 ext4_group_desc_csum_set(sb
, group
, gdp
);
1068 ext4_unlock_group(sb
, group
);
1070 BUFFER_TRACE(group_desc_bh
, "call ext4_handle_dirty_metadata");
1071 err
= ext4_handle_dirty_metadata(handle
, NULL
, group_desc_bh
);
1073 ext4_std_error(sb
, err
);
1077 percpu_counter_dec(&sbi
->s_freeinodes_counter
);
1079 percpu_counter_inc(&sbi
->s_dirs_counter
);
1081 if (sbi
->s_log_groups_per_flex
) {
1082 flex_group
= ext4_flex_group(sbi
, group
);
1083 atomic_dec(&sbi
->s_flex_groups
[flex_group
].free_inodes
);
1086 inode
->i_ino
= ino
+ group
* EXT4_INODES_PER_GROUP(sb
);
1087 /* This is the optimal IO size (for stat), not the fs block size */
1088 inode
->i_blocks
= 0;
1089 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= ei
->i_crtime
=
1090 current_time(inode
);
1092 memset(ei
->i_data
, 0, sizeof(ei
->i_data
));
1093 ei
->i_dir_start_lookup
= 0;
1096 /* Don't inherit extent flag from directory, amongst others. */
1098 ext4_mask_flags(mode
, EXT4_I(dir
)->i_flags
& EXT4_FL_INHERITED
);
1099 ei
->i_flags
|= i_flags
;
1102 ei
->i_block_group
= group
;
1103 ei
->i_last_alloc_group
= ~0;
1105 ext4_set_inode_flags(inode
);
1106 if (IS_DIRSYNC(inode
))
1107 ext4_handle_sync(handle
);
1108 if (insert_inode_locked(inode
) < 0) {
1110 * Likely a bitmap corruption causing inode to be allocated
1114 ext4_error(sb
, "failed to insert inode %lu: doubly allocated?",
1118 spin_lock(&sbi
->s_next_gen_lock
);
1119 inode
->i_generation
= sbi
->s_next_generation
++;
1120 spin_unlock(&sbi
->s_next_gen_lock
);
1122 /* Precompute checksum seed for inode metadata */
1123 if (ext4_has_metadata_csum(sb
)) {
1125 __le32 inum
= cpu_to_le32(inode
->i_ino
);
1126 __le32 gen
= cpu_to_le32(inode
->i_generation
);
1127 csum
= ext4_chksum(sbi
, sbi
->s_csum_seed
, (__u8
*)&inum
,
1129 ei
->i_csum_seed
= ext4_chksum(sbi
, csum
, (__u8
*)&gen
,
1133 ext4_clear_state_flags(ei
); /* Only relevant on 32-bit archs */
1134 ext4_set_inode_state(inode
, EXT4_STATE_NEW
);
1136 ei
->i_extra_isize
= EXT4_SB(sb
)->s_want_extra_isize
;
1137 ei
->i_inline_off
= 0;
1138 if (ext4_has_feature_inline_data(sb
))
1139 ext4_set_inode_state(inode
, EXT4_STATE_MAY_INLINE_DATA
);
1141 err
= dquot_alloc_inode(inode
);
1146 * Since the encryption xattr will always be unique, create it first so
1147 * that it's less likely to end up in an external xattr block and
1148 * prevent its deduplication.
1151 err
= fscrypt_inherit_context(dir
, inode
, handle
, true);
1153 goto fail_free_drop
;
1156 if (!(ei
->i_flags
& EXT4_EA_INODE_FL
)) {
1157 err
= ext4_init_acl(handle
, inode
, dir
);
1159 goto fail_free_drop
;
1161 err
= ext4_init_security(handle
, inode
, dir
, qstr
);
1163 goto fail_free_drop
;
1166 if (ext4_has_feature_extents(sb
)) {
1167 /* set extent flag only for directory, file and normal symlink*/
1168 if (S_ISDIR(mode
) || S_ISREG(mode
) || S_ISLNK(mode
)) {
1169 ext4_set_inode_flag(inode
, EXT4_INODE_EXTENTS
);
1170 ext4_ext_tree_init(handle
, inode
);
1174 if (ext4_handle_valid(handle
)) {
1175 ei
->i_sync_tid
= handle
->h_transaction
->t_tid
;
1176 ei
->i_datasync_tid
= handle
->h_transaction
->t_tid
;
1179 err
= ext4_mark_inode_dirty(handle
, inode
);
1181 ext4_std_error(sb
, err
);
1182 goto fail_free_drop
;
1185 ext4_debug("allocating inode %lu\n", inode
->i_ino
);
1186 trace_ext4_allocate_inode(inode
, dir
, mode
);
1187 brelse(inode_bitmap_bh
);
1191 dquot_free_inode(inode
);
1194 unlock_new_inode(inode
);
1197 inode
->i_flags
|= S_NOQUOTA
;
1199 brelse(inode_bitmap_bh
);
1200 return ERR_PTR(err
);
1203 /* Verify that we are loading a valid orphan from disk */
1204 struct inode
*ext4_orphan_get(struct super_block
*sb
, unsigned long ino
)
1206 unsigned long max_ino
= le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
);
1207 ext4_group_t block_group
;
1209 struct buffer_head
*bitmap_bh
= NULL
;
1210 struct inode
*inode
= NULL
;
1211 int err
= -EFSCORRUPTED
;
1213 if (ino
< EXT4_FIRST_INO(sb
) || ino
> max_ino
)
1216 block_group
= (ino
- 1) / EXT4_INODES_PER_GROUP(sb
);
1217 bit
= (ino
- 1) % EXT4_INODES_PER_GROUP(sb
);
1218 bitmap_bh
= ext4_read_inode_bitmap(sb
, block_group
);
1219 if (IS_ERR(bitmap_bh
)) {
1220 ext4_error(sb
, "inode bitmap error %ld for orphan %lu",
1221 ino
, PTR_ERR(bitmap_bh
));
1222 return (struct inode
*) bitmap_bh
;
1225 /* Having the inode bit set should be a 100% indicator that this
1226 * is a valid orphan (no e2fsck run on fs). Orphans also include
1227 * inodes that were being truncated, so we can't check i_nlink==0.
1229 if (!ext4_test_bit(bit
, bitmap_bh
->b_data
))
1232 inode
= ext4_iget(sb
, ino
);
1233 if (IS_ERR(inode
)) {
1234 err
= PTR_ERR(inode
);
1235 ext4_error(sb
, "couldn't read orphan inode %lu (err %d)",
1241 * If the orphans has i_nlinks > 0 then it should be able to
1242 * be truncated, otherwise it won't be removed from the orphan
1243 * list during processing and an infinite loop will result.
1244 * Similarly, it must not be a bad inode.
1246 if ((inode
->i_nlink
&& !ext4_can_truncate(inode
)) ||
1247 is_bad_inode(inode
))
1250 if (NEXT_ORPHAN(inode
) > max_ino
)
1256 ext4_error(sb
, "bad orphan inode %lu", ino
);
1258 printk(KERN_ERR
"ext4_test_bit(bit=%d, block=%llu) = %d\n",
1259 bit
, (unsigned long long)bitmap_bh
->b_blocknr
,
1260 ext4_test_bit(bit
, bitmap_bh
->b_data
));
1262 printk(KERN_ERR
"is_bad_inode(inode)=%d\n",
1263 is_bad_inode(inode
));
1264 printk(KERN_ERR
"NEXT_ORPHAN(inode)=%u\n",
1265 NEXT_ORPHAN(inode
));
1266 printk(KERN_ERR
"max_ino=%lu\n", max_ino
);
1267 printk(KERN_ERR
"i_nlink=%u\n", inode
->i_nlink
);
1268 /* Avoid freeing blocks if we got a bad deleted inode */
1269 if (inode
->i_nlink
== 0)
1270 inode
->i_blocks
= 0;
1274 return ERR_PTR(err
);
1277 unsigned long ext4_count_free_inodes(struct super_block
*sb
)
1279 unsigned long desc_count
;
1280 struct ext4_group_desc
*gdp
;
1281 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
1283 struct ext4_super_block
*es
;
1284 unsigned long bitmap_count
, x
;
1285 struct buffer_head
*bitmap_bh
= NULL
;
1287 es
= EXT4_SB(sb
)->s_es
;
1291 for (i
= 0; i
< ngroups
; i
++) {
1292 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1295 desc_count
+= ext4_free_inodes_count(sb
, gdp
);
1297 bitmap_bh
= ext4_read_inode_bitmap(sb
, i
);
1298 if (IS_ERR(bitmap_bh
)) {
1303 x
= ext4_count_free(bitmap_bh
->b_data
,
1304 EXT4_INODES_PER_GROUP(sb
) / 8);
1305 printk(KERN_DEBUG
"group %lu: stored = %d, counted = %lu\n",
1306 (unsigned long) i
, ext4_free_inodes_count(sb
, gdp
), x
);
1310 printk(KERN_DEBUG
"ext4_count_free_inodes: "
1311 "stored = %u, computed = %lu, %lu\n",
1312 le32_to_cpu(es
->s_free_inodes_count
), desc_count
, bitmap_count
);
1316 for (i
= 0; i
< ngroups
; i
++) {
1317 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1320 desc_count
+= ext4_free_inodes_count(sb
, gdp
);
1327 /* Called at mount-time, super-block is locked */
1328 unsigned long ext4_count_dirs(struct super_block
* sb
)
1330 unsigned long count
= 0;
1331 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
1333 for (i
= 0; i
< ngroups
; i
++) {
1334 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1337 count
+= ext4_used_dirs_count(sb
, gdp
);
1343 * Zeroes not yet zeroed inode table - just write zeroes through the whole
1344 * inode table. Must be called without any spinlock held. The only place
1345 * where it is called from on active part of filesystem is ext4lazyinit
1346 * thread, so we do not need any special locks, however we have to prevent
1347 * inode allocation from the current group, so we take alloc_sem lock, to
1348 * block ext4_new_inode() until we are finished.
1350 int ext4_init_inode_table(struct super_block
*sb
, ext4_group_t group
,
1353 struct ext4_group_info
*grp
= ext4_get_group_info(sb
, group
);
1354 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1355 struct ext4_group_desc
*gdp
= NULL
;
1356 struct buffer_head
*group_desc_bh
;
1359 int num
, ret
= 0, used_blks
= 0;
1361 /* This should not happen, but just to be sure check this */
1362 if (sb
->s_flags
& MS_RDONLY
) {
1367 gdp
= ext4_get_group_desc(sb
, group
, &group_desc_bh
);
1372 * We do not need to lock this, because we are the only one
1373 * handling this flag.
1375 if (gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
))
1378 handle
= ext4_journal_start_sb(sb
, EXT4_HT_MISC
, 1);
1379 if (IS_ERR(handle
)) {
1380 ret
= PTR_ERR(handle
);
1384 down_write(&grp
->alloc_sem
);
1386 * If inode bitmap was already initialized there may be some
1387 * used inodes so we need to skip blocks with used inodes in
1390 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_UNINIT
)))
1391 used_blks
= DIV_ROUND_UP((EXT4_INODES_PER_GROUP(sb
) -
1392 ext4_itable_unused_count(sb
, gdp
)),
1393 sbi
->s_inodes_per_block
);
1395 if ((used_blks
< 0) || (used_blks
> sbi
->s_itb_per_group
)) {
1396 ext4_error(sb
, "Something is wrong with group %u: "
1397 "used itable blocks: %d; "
1398 "itable unused count: %u",
1400 ext4_itable_unused_count(sb
, gdp
));
1405 blk
= ext4_inode_table(sb
, gdp
) + used_blks
;
1406 num
= sbi
->s_itb_per_group
- used_blks
;
1408 BUFFER_TRACE(group_desc_bh
, "get_write_access");
1409 ret
= ext4_journal_get_write_access(handle
,
1415 * Skip zeroout if the inode table is full. But we set the ZEROED
1416 * flag anyway, because obviously, when it is full it does not need
1419 if (unlikely(num
== 0))
1422 ext4_debug("going to zero out inode table in group %d\n",
1424 ret
= sb_issue_zeroout(sb
, blk
, num
, GFP_NOFS
);
1428 blkdev_issue_flush(sb
->s_bdev
, GFP_NOFS
, NULL
);
1431 ext4_lock_group(sb
, group
);
1432 gdp
->bg_flags
|= cpu_to_le16(EXT4_BG_INODE_ZEROED
);
1433 ext4_group_desc_csum_set(sb
, group
, gdp
);
1434 ext4_unlock_group(sb
, group
);
1436 BUFFER_TRACE(group_desc_bh
,
1437 "call ext4_handle_dirty_metadata");
1438 ret
= ext4_handle_dirty_metadata(handle
, NULL
,
1442 up_write(&grp
->alloc_sem
);
1443 ext4_journal_stop(handle
);