1 // SPDX-License-Identifier: GPL-2.0
3 * linux/fs/ext4/balloc.c
5 * Copyright (C) 1992, 1993, 1994, 1995
6 * Remy Card (card@masi.ibp.fr)
7 * Laboratoire MASI - Institut Blaise Pascal
8 * Universite Pierre et Marie Curie (Paris VI)
10 * Enhanced block allocation by 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>
16 #include <linux/capability.h>
18 #include <linux/quotaops.h>
19 #include <linux/buffer_head.h>
21 #include "ext4_jbd2.h"
24 #include <trace/events/ext4.h>
26 static unsigned ext4_num_base_meta_clusters(struct super_block
*sb
,
27 ext4_group_t block_group
);
29 * balloc.c contains the blocks allocation and deallocation routines
33 * Calculate block group number for a given block number
35 ext4_group_t
ext4_get_group_number(struct super_block
*sb
,
40 if (test_opt2(sb
, STD_GROUP_SIZE
))
42 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_first_data_block
)) >>
43 (EXT4_BLOCK_SIZE_BITS(sb
) + EXT4_CLUSTER_BITS(sb
) + 3);
45 ext4_get_group_no_and_offset(sb
, block
, &group
, NULL
);
50 * Calculate the block group number and offset into the block/cluster
51 * allocation bitmap, given a block number
53 void ext4_get_group_no_and_offset(struct super_block
*sb
, ext4_fsblk_t blocknr
,
54 ext4_group_t
*blockgrpp
, ext4_grpblk_t
*offsetp
)
56 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
59 blocknr
= blocknr
- le32_to_cpu(es
->s_first_data_block
);
60 offset
= do_div(blocknr
, EXT4_BLOCKS_PER_GROUP(sb
)) >>
61 EXT4_SB(sb
)->s_cluster_bits
;
70 * Check whether the 'block' lives within the 'block_group'. Returns 1 if so
73 static inline int ext4_block_in_group(struct super_block
*sb
,
75 ext4_group_t block_group
)
77 ext4_group_t actual_group
;
79 actual_group
= ext4_get_group_number(sb
, block
);
80 return (actual_group
== block_group
) ? 1 : 0;
83 /* Return the number of clusters used for file system metadata; this
84 * represents the overhead needed by the file system.
86 static unsigned ext4_num_overhead_clusters(struct super_block
*sb
,
87 ext4_group_t block_group
,
88 struct ext4_group_desc
*gdp
)
90 unsigned num_clusters
;
91 int block_cluster
= -1, inode_cluster
= -1, itbl_cluster
= -1, i
, c
;
92 ext4_fsblk_t start
= ext4_group_first_block_no(sb
, block_group
);
93 ext4_fsblk_t itbl_blk
;
94 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
96 /* This is the number of clusters used by the superblock,
97 * block group descriptors, and reserved block group
98 * descriptor blocks */
99 num_clusters
= ext4_num_base_meta_clusters(sb
, block_group
);
102 * For the allocation bitmaps and inode table, we first need
103 * to check to see if the block is in the block group. If it
104 * is, then check to see if the cluster is already accounted
105 * for in the clusters used for the base metadata cluster, or
106 * if we can increment the base metadata cluster to include
107 * that block. Otherwise, we will have to track the cluster
108 * used for the allocation bitmap or inode table explicitly.
109 * Normally all of these blocks are contiguous, so the special
110 * case handling shouldn't be necessary except for *very*
111 * unusual file system layouts.
113 if (ext4_block_in_group(sb
, ext4_block_bitmap(sb
, gdp
), block_group
)) {
114 block_cluster
= EXT4_B2C(sbi
,
115 ext4_block_bitmap(sb
, gdp
) - start
);
116 if (block_cluster
< num_clusters
)
118 else if (block_cluster
== num_clusters
) {
124 if (ext4_block_in_group(sb
, ext4_inode_bitmap(sb
, gdp
), block_group
)) {
125 inode_cluster
= EXT4_B2C(sbi
,
126 ext4_inode_bitmap(sb
, gdp
) - start
);
127 if (inode_cluster
< num_clusters
)
129 else if (inode_cluster
== num_clusters
) {
135 itbl_blk
= ext4_inode_table(sb
, gdp
);
136 for (i
= 0; i
< sbi
->s_itb_per_group
; i
++) {
137 if (ext4_block_in_group(sb
, itbl_blk
+ i
, block_group
)) {
138 c
= EXT4_B2C(sbi
, itbl_blk
+ i
- start
);
139 if ((c
< num_clusters
) || (c
== inode_cluster
) ||
140 (c
== block_cluster
) || (c
== itbl_cluster
))
142 if (c
== num_clusters
) {
151 if (block_cluster
!= -1)
153 if (inode_cluster
!= -1)
159 static unsigned int num_clusters_in_group(struct super_block
*sb
,
160 ext4_group_t block_group
)
164 if (block_group
== ext4_get_groups_count(sb
) - 1) {
166 * Even though mke2fs always initializes the first and
167 * last group, just in case some other tool was used,
168 * we need to make sure we calculate the right free
171 blocks
= ext4_blocks_count(EXT4_SB(sb
)->s_es
) -
172 ext4_group_first_block_no(sb
, block_group
);
174 blocks
= EXT4_BLOCKS_PER_GROUP(sb
);
175 return EXT4_NUM_B2C(EXT4_SB(sb
), blocks
);
178 /* Initializes an uninitialized block bitmap */
179 static int ext4_init_block_bitmap(struct super_block
*sb
,
180 struct buffer_head
*bh
,
181 ext4_group_t block_group
,
182 struct ext4_group_desc
*gdp
)
184 unsigned int bit
, bit_max
;
185 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
186 ext4_fsblk_t start
, tmp
;
188 struct ext4_group_info
*grp
;
190 J_ASSERT_BH(bh
, buffer_locked(bh
));
192 /* If checksum is bad mark all blocks used to prevent allocation
193 * essentially implementing a per-group read-only flag. */
194 if (!ext4_group_desc_csum_verify(sb
, block_group
, gdp
)) {
195 grp
= ext4_get_group_info(sb
, block_group
);
196 if (!EXT4_MB_GRP_BBITMAP_CORRUPT(grp
))
197 percpu_counter_sub(&sbi
->s_freeclusters_counter
,
199 set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT
, &grp
->bb_state
);
200 if (!EXT4_MB_GRP_IBITMAP_CORRUPT(grp
)) {
202 count
= ext4_free_inodes_count(sb
, gdp
);
203 percpu_counter_sub(&sbi
->s_freeinodes_counter
,
206 set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT
, &grp
->bb_state
);
209 memset(bh
->b_data
, 0, sb
->s_blocksize
);
211 bit_max
= ext4_num_base_meta_clusters(sb
, block_group
);
212 if ((bit_max
>> 3) >= bh
->b_size
)
213 return -EFSCORRUPTED
;
215 for (bit
= 0; bit
< bit_max
; bit
++)
216 ext4_set_bit(bit
, bh
->b_data
);
218 start
= ext4_group_first_block_no(sb
, block_group
);
220 if (ext4_has_feature_flex_bg(sb
))
223 /* Set bits for block and inode bitmaps, and inode table */
224 tmp
= ext4_block_bitmap(sb
, gdp
);
225 if (!flex_bg
|| ext4_block_in_group(sb
, tmp
, block_group
))
226 ext4_set_bit(EXT4_B2C(sbi
, tmp
- start
), bh
->b_data
);
228 tmp
= ext4_inode_bitmap(sb
, gdp
);
229 if (!flex_bg
|| ext4_block_in_group(sb
, tmp
, block_group
))
230 ext4_set_bit(EXT4_B2C(sbi
, tmp
- start
), bh
->b_data
);
232 tmp
= ext4_inode_table(sb
, gdp
);
233 for (; tmp
< ext4_inode_table(sb
, gdp
) +
234 sbi
->s_itb_per_group
; tmp
++) {
235 if (!flex_bg
|| ext4_block_in_group(sb
, tmp
, block_group
))
236 ext4_set_bit(EXT4_B2C(sbi
, tmp
- start
), bh
->b_data
);
240 * Also if the number of blocks within the group is less than
241 * the blocksize * 8 ( which is the size of bitmap ), set rest
242 * of the block bitmap to 1
244 ext4_mark_bitmap_end(num_clusters_in_group(sb
, block_group
),
245 sb
->s_blocksize
* 8, bh
->b_data
);
249 /* Return the number of free blocks in a block group. It is used when
250 * the block bitmap is uninitialized, so we can't just count the bits
252 unsigned ext4_free_clusters_after_init(struct super_block
*sb
,
253 ext4_group_t block_group
,
254 struct ext4_group_desc
*gdp
)
256 return num_clusters_in_group(sb
, block_group
) -
257 ext4_num_overhead_clusters(sb
, block_group
, gdp
);
261 * The free blocks are managed by bitmaps. A file system contains several
262 * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap
263 * block for inodes, N blocks for the inode table and data blocks.
265 * The file system contains group descriptors which are located after the
266 * super block. Each descriptor contains the number of the bitmap block and
267 * the free blocks count in the block. The descriptors are loaded in memory
268 * when a file system is mounted (see ext4_fill_super).
272 * ext4_get_group_desc() -- load group descriptor from disk
274 * @block_group: given block group
275 * @bh: pointer to the buffer head to store the block
278 struct ext4_group_desc
* ext4_get_group_desc(struct super_block
*sb
,
279 ext4_group_t block_group
,
280 struct buffer_head
**bh
)
282 unsigned int group_desc
;
284 ext4_group_t ngroups
= ext4_get_groups_count(sb
);
285 struct ext4_group_desc
*desc
;
286 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
288 if (block_group
>= ngroups
) {
289 ext4_error(sb
, "block_group >= groups_count - block_group = %u,"
290 " groups_count = %u", block_group
, ngroups
);
295 group_desc
= block_group
>> EXT4_DESC_PER_BLOCK_BITS(sb
);
296 offset
= block_group
& (EXT4_DESC_PER_BLOCK(sb
) - 1);
297 if (!sbi
->s_group_desc
[group_desc
]) {
298 ext4_error(sb
, "Group descriptor not loaded - "
299 "block_group = %u, group_desc = %u, desc = %u",
300 block_group
, group_desc
, offset
);
304 desc
= (struct ext4_group_desc
*)(
305 (__u8
*)sbi
->s_group_desc
[group_desc
]->b_data
+
306 offset
* EXT4_DESC_SIZE(sb
));
308 *bh
= sbi
->s_group_desc
[group_desc
];
313 * Return the block number which was discovered to be invalid, or 0 if
314 * the block bitmap is valid.
316 static ext4_fsblk_t
ext4_valid_block_bitmap(struct super_block
*sb
,
317 struct ext4_group_desc
*desc
,
318 ext4_group_t block_group
,
319 struct buffer_head
*bh
)
321 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
322 ext4_grpblk_t offset
;
323 ext4_grpblk_t next_zero_bit
;
325 ext4_fsblk_t group_first_block
;
327 if (ext4_has_feature_flex_bg(sb
)) {
328 /* with FLEX_BG, the inode/block bitmaps and itable
329 * blocks may not be in the group at all
330 * so the bitmap validation will be skipped for those groups
331 * or it has to also read the block group where the bitmaps
332 * are located to verify they are set.
336 group_first_block
= ext4_group_first_block_no(sb
, block_group
);
338 /* check whether block bitmap block number is set */
339 blk
= ext4_block_bitmap(sb
, desc
);
340 offset
= blk
- group_first_block
;
341 if (offset
< 0 || EXT4_B2C(sbi
, offset
) >= sb
->s_blocksize
||
342 !ext4_test_bit(EXT4_B2C(sbi
, offset
), bh
->b_data
))
343 /* bad block bitmap */
346 /* check whether the inode bitmap block number is set */
347 blk
= ext4_inode_bitmap(sb
, desc
);
348 offset
= blk
- group_first_block
;
349 if (offset
< 0 || EXT4_B2C(sbi
, offset
) >= sb
->s_blocksize
||
350 !ext4_test_bit(EXT4_B2C(sbi
, offset
), bh
->b_data
))
351 /* bad block bitmap */
354 /* check whether the inode table block number is set */
355 blk
= ext4_inode_table(sb
, desc
);
356 offset
= blk
- group_first_block
;
357 if (offset
< 0 || EXT4_B2C(sbi
, offset
) >= sb
->s_blocksize
||
358 EXT4_B2C(sbi
, offset
+ sbi
->s_itb_per_group
) >= sb
->s_blocksize
)
360 next_zero_bit
= ext4_find_next_zero_bit(bh
->b_data
,
361 EXT4_B2C(sbi
, offset
+ EXT4_SB(sb
)->s_itb_per_group
),
362 EXT4_B2C(sbi
, offset
));
364 EXT4_B2C(sbi
, offset
+ EXT4_SB(sb
)->s_itb_per_group
))
365 /* bad bitmap for inode tables */
370 static int ext4_validate_block_bitmap(struct super_block
*sb
,
371 struct ext4_group_desc
*desc
,
372 ext4_group_t block_group
,
373 struct buffer_head
*bh
)
376 struct ext4_group_info
*grp
= ext4_get_group_info(sb
, block_group
);
377 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
379 if (buffer_verified(bh
))
381 if (EXT4_MB_GRP_BBITMAP_CORRUPT(grp
))
382 return -EFSCORRUPTED
;
384 ext4_lock_group(sb
, block_group
);
385 if (buffer_verified(bh
))
387 if (unlikely(!ext4_block_bitmap_csum_verify(sb
, block_group
,
389 ext4_unlock_group(sb
, block_group
);
390 ext4_error(sb
, "bg %u: bad block bitmap checksum", block_group
);
391 if (!EXT4_MB_GRP_BBITMAP_CORRUPT(grp
))
392 percpu_counter_sub(&sbi
->s_freeclusters_counter
,
394 set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT
, &grp
->bb_state
);
397 blk
= ext4_valid_block_bitmap(sb
, desc
, block_group
, bh
);
398 if (unlikely(blk
!= 0)) {
399 ext4_unlock_group(sb
, block_group
);
400 ext4_error(sb
, "bg %u: block %llu: invalid block bitmap",
402 if (!EXT4_MB_GRP_BBITMAP_CORRUPT(grp
))
403 percpu_counter_sub(&sbi
->s_freeclusters_counter
,
405 set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT
, &grp
->bb_state
);
406 return -EFSCORRUPTED
;
408 set_buffer_verified(bh
);
410 ext4_unlock_group(sb
, block_group
);
415 * ext4_read_block_bitmap_nowait()
417 * @block_group: given block group
419 * Read the bitmap for a given block_group,and validate the
420 * bits for block/inode/inode tables are set in the bitmaps
422 * Return buffer_head on success or NULL in case of failure.
425 ext4_read_block_bitmap_nowait(struct super_block
*sb
, ext4_group_t block_group
)
427 struct ext4_group_desc
*desc
;
428 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
429 struct buffer_head
*bh
;
430 ext4_fsblk_t bitmap_blk
;
433 desc
= ext4_get_group_desc(sb
, block_group
, NULL
);
435 return ERR_PTR(-EFSCORRUPTED
);
436 bitmap_blk
= ext4_block_bitmap(sb
, desc
);
437 if ((bitmap_blk
<= le32_to_cpu(sbi
->s_es
->s_first_data_block
)) ||
438 (bitmap_blk
>= ext4_blocks_count(sbi
->s_es
))) {
439 ext4_error(sb
, "Invalid block bitmap block %llu in "
440 "block_group %u", bitmap_blk
, block_group
);
441 return ERR_PTR(-EFSCORRUPTED
);
443 bh
= sb_getblk(sb
, bitmap_blk
);
445 ext4_error(sb
, "Cannot get buffer for block bitmap - "
446 "block_group = %u, block_bitmap = %llu",
447 block_group
, bitmap_blk
);
448 return ERR_PTR(-ENOMEM
);
451 if (bitmap_uptodate(bh
))
455 if (bitmap_uptodate(bh
)) {
459 ext4_lock_group(sb
, block_group
);
460 if (desc
->bg_flags
& cpu_to_le16(EXT4_BG_BLOCK_UNINIT
)) {
461 err
= ext4_init_block_bitmap(sb
, bh
, block_group
, desc
);
462 set_bitmap_uptodate(bh
);
463 set_buffer_uptodate(bh
);
464 set_buffer_verified(bh
);
465 ext4_unlock_group(sb
, block_group
);
468 ext4_error(sb
, "Failed to init block bitmap for group "
469 "%u: %d", block_group
, err
);
474 ext4_unlock_group(sb
, block_group
);
475 if (buffer_uptodate(bh
)) {
477 * if not uninit if bh is uptodate,
478 * bitmap is also uptodate
480 set_bitmap_uptodate(bh
);
485 * submit the buffer_head for reading
488 trace_ext4_read_block_bitmap_load(sb
, block_group
);
489 bh
->b_end_io
= ext4_end_bitmap_read
;
491 submit_bh(REQ_OP_READ
, REQ_META
| REQ_PRIO
, bh
);
494 err
= ext4_validate_block_bitmap(sb
, desc
, block_group
, bh
);
503 /* Returns 0 on success, 1 on error */
504 int ext4_wait_block_bitmap(struct super_block
*sb
, ext4_group_t block_group
,
505 struct buffer_head
*bh
)
507 struct ext4_group_desc
*desc
;
511 desc
= ext4_get_group_desc(sb
, block_group
, NULL
);
513 return -EFSCORRUPTED
;
515 if (!buffer_uptodate(bh
)) {
516 ext4_error(sb
, "Cannot read block bitmap - "
517 "block_group = %u, block_bitmap = %llu",
518 block_group
, (unsigned long long) bh
->b_blocknr
);
521 clear_buffer_new(bh
);
522 /* Panic or remount fs read-only if block bitmap is invalid */
523 return ext4_validate_block_bitmap(sb
, desc
, block_group
, bh
);
527 ext4_read_block_bitmap(struct super_block
*sb
, ext4_group_t block_group
)
529 struct buffer_head
*bh
;
532 bh
= ext4_read_block_bitmap_nowait(sb
, block_group
);
535 err
= ext4_wait_block_bitmap(sb
, block_group
, bh
);
544 * ext4_has_free_clusters()
545 * @sbi: in-core super block structure.
546 * @nclusters: number of needed blocks
547 * @flags: flags from ext4_mb_new_blocks()
549 * Check if filesystem has nclusters free & available for allocation.
550 * On success return 1, return 0 on failure.
552 static int ext4_has_free_clusters(struct ext4_sb_info
*sbi
,
553 s64 nclusters
, unsigned int flags
)
555 s64 free_clusters
, dirty_clusters
, rsv
, resv_clusters
;
556 struct percpu_counter
*fcc
= &sbi
->s_freeclusters_counter
;
557 struct percpu_counter
*dcc
= &sbi
->s_dirtyclusters_counter
;
559 free_clusters
= percpu_counter_read_positive(fcc
);
560 dirty_clusters
= percpu_counter_read_positive(dcc
);
561 resv_clusters
= atomic64_read(&sbi
->s_resv_clusters
);
564 * r_blocks_count should always be multiple of the cluster ratio so
565 * we are safe to do a plane bit shift only.
567 rsv
= (ext4_r_blocks_count(sbi
->s_es
) >> sbi
->s_cluster_bits
) +
570 if (free_clusters
- (nclusters
+ rsv
+ dirty_clusters
) <
571 EXT4_FREECLUSTERS_WATERMARK
) {
572 free_clusters
= percpu_counter_sum_positive(fcc
);
573 dirty_clusters
= percpu_counter_sum_positive(dcc
);
575 /* Check whether we have space after accounting for current
576 * dirty clusters & root reserved clusters.
578 if (free_clusters
>= (rsv
+ nclusters
+ dirty_clusters
))
581 /* Hm, nope. Are (enough) root reserved clusters available? */
582 if (uid_eq(sbi
->s_resuid
, current_fsuid()) ||
583 (!gid_eq(sbi
->s_resgid
, make_kgid(sbi
->s_sb
->s_user_ns
, 0)) && in_group_p(sbi
->s_resgid
)) ||
584 ns_capable(sbi
->s_sb
->s_user_ns
, CAP_SYS_RESOURCE
) ||
585 (flags
& EXT4_MB_USE_ROOT_BLOCKS
)) {
587 if (free_clusters
>= (nclusters
+ dirty_clusters
+
591 /* No free blocks. Let's see if we can dip into reserved pool */
592 if (flags
& EXT4_MB_USE_RESERVED
) {
593 if (free_clusters
>= (nclusters
+ dirty_clusters
))
600 int ext4_claim_free_clusters(struct ext4_sb_info
*sbi
,
601 s64 nclusters
, unsigned int flags
)
603 if (ext4_has_free_clusters(sbi
, nclusters
, flags
)) {
604 percpu_counter_add(&sbi
->s_dirtyclusters_counter
, nclusters
);
611 * ext4_should_retry_alloc()
613 * @retries number of attemps has been made
615 * ext4_should_retry_alloc() is called when ENOSPC is returned, and if
616 * it is profitable to retry the operation, this function will wait
617 * for the current or committing transaction to complete, and then
618 * return TRUE. We will only retry once.
620 int ext4_should_retry_alloc(struct super_block
*sb
, int *retries
)
622 if (!ext4_has_free_clusters(EXT4_SB(sb
), 1, 0) ||
624 !EXT4_SB(sb
)->s_journal
)
628 if (EXT4_SB(sb
)->s_mb_free_pending
== 0)
631 jbd_debug(1, "%s: retrying operation after ENOSPC\n", sb
->s_id
);
632 jbd2_journal_force_commit_nested(EXT4_SB(sb
)->s_journal
);
637 * ext4_new_meta_blocks() -- allocate block for meta data (indexing) blocks
639 * @handle: handle to this transaction
641 * @goal: given target block(filesystem wide)
642 * @count: pointer to total number of clusters needed
645 * Return 1st allocated block number on success, *count stores total account
646 * error stores in errp pointer
648 ext4_fsblk_t
ext4_new_meta_blocks(handle_t
*handle
, struct inode
*inode
,
649 ext4_fsblk_t goal
, unsigned int flags
,
650 unsigned long *count
, int *errp
)
652 struct ext4_allocation_request ar
;
655 memset(&ar
, 0, sizeof(ar
));
656 /* Fill with neighbour allocated blocks */
659 ar
.len
= count
? *count
: 1;
662 ret
= ext4_mb_new_blocks(handle
, &ar
, errp
);
666 * Account for the allocated meta blocks. We will never
667 * fail EDQUOT for metdata, but we do account for it.
669 if (!(*errp
) && (flags
& EXT4_MB_DELALLOC_RESERVED
)) {
670 dquot_alloc_block_nofail(inode
,
671 EXT4_C2B(EXT4_SB(inode
->i_sb
), ar
.len
));
677 * ext4_count_free_clusters() -- count filesystem free clusters
680 * Adds up the number of free clusters from each block group.
682 ext4_fsblk_t
ext4_count_free_clusters(struct super_block
*sb
)
684 ext4_fsblk_t desc_count
;
685 struct ext4_group_desc
*gdp
;
687 ext4_group_t ngroups
= ext4_get_groups_count(sb
);
688 struct ext4_group_info
*grp
;
690 struct ext4_super_block
*es
;
691 ext4_fsblk_t bitmap_count
;
693 struct buffer_head
*bitmap_bh
= NULL
;
695 es
= EXT4_SB(sb
)->s_es
;
700 for (i
= 0; i
< ngroups
; i
++) {
701 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
705 if (EXT4_SB(sb
)->s_group_info
)
706 grp
= ext4_get_group_info(sb
, i
);
707 if (!grp
|| !EXT4_MB_GRP_BBITMAP_CORRUPT(grp
))
708 desc_count
+= ext4_free_group_clusters(sb
, gdp
);
710 bitmap_bh
= ext4_read_block_bitmap(sb
, i
);
711 if (IS_ERR(bitmap_bh
)) {
716 x
= ext4_count_free(bitmap_bh
->b_data
,
717 EXT4_CLUSTERS_PER_GROUP(sb
) / 8);
718 printk(KERN_DEBUG
"group %u: stored = %d, counted = %u\n",
719 i
, ext4_free_group_clusters(sb
, gdp
), x
);
723 printk(KERN_DEBUG
"ext4_count_free_clusters: stored = %llu"
724 ", computed = %llu, %llu\n",
725 EXT4_NUM_B2C(EXT4_SB(sb
), ext4_free_blocks_count(es
)),
726 desc_count
, bitmap_count
);
730 for (i
= 0; i
< ngroups
; i
++) {
731 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
735 if (EXT4_SB(sb
)->s_group_info
)
736 grp
= ext4_get_group_info(sb
, i
);
737 if (!grp
|| !EXT4_MB_GRP_BBITMAP_CORRUPT(grp
))
738 desc_count
+= ext4_free_group_clusters(sb
, gdp
);
745 static inline int test_root(ext4_group_t a
, int b
)
759 * ext4_bg_has_super - number of blocks used by the superblock in group
760 * @sb: superblock for filesystem
761 * @group: group number to check
763 * Return the number of blocks used by the superblock (primary or backup)
764 * in this group. Currently this will be only 0 or 1.
766 int ext4_bg_has_super(struct super_block
*sb
, ext4_group_t group
)
768 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
772 if (ext4_has_feature_sparse_super2(sb
)) {
773 if (group
== le32_to_cpu(es
->s_backup_bgs
[0]) ||
774 group
== le32_to_cpu(es
->s_backup_bgs
[1]))
778 if ((group
<= 1) || !ext4_has_feature_sparse_super(sb
))
782 if (test_root(group
, 3) || (test_root(group
, 5)) ||
789 static unsigned long ext4_bg_num_gdb_meta(struct super_block
*sb
,
792 unsigned long metagroup
= group
/ EXT4_DESC_PER_BLOCK(sb
);
793 ext4_group_t first
= metagroup
* EXT4_DESC_PER_BLOCK(sb
);
794 ext4_group_t last
= first
+ EXT4_DESC_PER_BLOCK(sb
) - 1;
796 if (group
== first
|| group
== first
+ 1 || group
== last
)
801 static unsigned long ext4_bg_num_gdb_nometa(struct super_block
*sb
,
804 if (!ext4_bg_has_super(sb
, group
))
807 if (ext4_has_feature_meta_bg(sb
))
808 return le32_to_cpu(EXT4_SB(sb
)->s_es
->s_first_meta_bg
);
810 return EXT4_SB(sb
)->s_gdb_count
;
814 * ext4_bg_num_gdb - number of blocks used by the group table in group
815 * @sb: superblock for filesystem
816 * @group: group number to check
818 * Return the number of blocks used by the group descriptor table
819 * (primary or backup) in this group. In the future there may be a
820 * different number of descriptor blocks in each group.
822 unsigned long ext4_bg_num_gdb(struct super_block
*sb
, ext4_group_t group
)
824 unsigned long first_meta_bg
=
825 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_first_meta_bg
);
826 unsigned long metagroup
= group
/ EXT4_DESC_PER_BLOCK(sb
);
828 if (!ext4_has_feature_meta_bg(sb
) || metagroup
< first_meta_bg
)
829 return ext4_bg_num_gdb_nometa(sb
, group
);
831 return ext4_bg_num_gdb_meta(sb
,group
);
836 * This function returns the number of file system metadata clusters at
837 * the beginning of a block group, including the reserved gdt blocks.
839 static unsigned ext4_num_base_meta_clusters(struct super_block
*sb
,
840 ext4_group_t block_group
)
842 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
845 /* Check for superblock and gdt backups in this group */
846 num
= ext4_bg_has_super(sb
, block_group
);
848 if (!ext4_has_feature_meta_bg(sb
) ||
849 block_group
< le32_to_cpu(sbi
->s_es
->s_first_meta_bg
) *
850 sbi
->s_desc_per_block
) {
852 num
+= ext4_bg_num_gdb(sb
, block_group
);
853 num
+= le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
);
855 } else { /* For META_BG_BLOCK_GROUPS */
856 num
+= ext4_bg_num_gdb(sb
, block_group
);
858 return EXT4_NUM_B2C(sbi
, num
);
861 * ext4_inode_to_goal_block - return a hint for block allocation
862 * @inode: inode for block allocation
864 * Return the ideal location to start allocating blocks for a
865 * newly created inode.
867 ext4_fsblk_t
ext4_inode_to_goal_block(struct inode
*inode
)
869 struct ext4_inode_info
*ei
= EXT4_I(inode
);
870 ext4_group_t block_group
;
871 ext4_grpblk_t colour
;
872 int flex_size
= ext4_flex_bg_size(EXT4_SB(inode
->i_sb
));
873 ext4_fsblk_t bg_start
;
874 ext4_fsblk_t last_block
;
876 block_group
= ei
->i_block_group
;
877 if (flex_size
>= EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME
) {
879 * If there are at least EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME
880 * block groups per flexgroup, reserve the first block
881 * group for directories and special files. Regular
882 * files will start at the second block group. This
883 * tends to speed up directory access and improves
886 block_group
&= ~(flex_size
-1);
887 if (S_ISREG(inode
->i_mode
))
890 bg_start
= ext4_group_first_block_no(inode
->i_sb
, block_group
);
891 last_block
= ext4_blocks_count(EXT4_SB(inode
->i_sb
)->s_es
) - 1;
894 * If we are doing delayed allocation, we don't need take
895 * colour into account.
897 if (test_opt(inode
->i_sb
, DELALLOC
))
900 if (bg_start
+ EXT4_BLOCKS_PER_GROUP(inode
->i_sb
) <= last_block
)
901 colour
= (current
->pid
% 16) *
902 (EXT4_BLOCKS_PER_GROUP(inode
->i_sb
) / 16);
904 colour
= (current
->pid
% 16) * ((last_block
- bg_start
) / 16);
905 return bg_start
+ colour
;