2 * linux/fs/ext4/balloc.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 * Enhanced block allocation by Stephen Tweedie (sct@redhat.com), 1993
10 * Big-endian to little-endian byte-swapping/bitmaps by
11 * David S. Miller (davem@caip.rutgers.edu), 1995
14 #include <linux/time.h>
15 #include <linux/capability.h>
17 #include <linux/jbd2.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 for (bit
= 0; bit
< bit_max
; bit
++)
213 ext4_set_bit(bit
, bh
->b_data
);
215 start
= ext4_group_first_block_no(sb
, block_group
);
217 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
220 /* Set bits for block and inode bitmaps, and inode table */
221 tmp
= ext4_block_bitmap(sb
, gdp
);
222 if (!flex_bg
|| ext4_block_in_group(sb
, tmp
, block_group
))
223 ext4_set_bit(EXT4_B2C(sbi
, tmp
- start
), bh
->b_data
);
225 tmp
= ext4_inode_bitmap(sb
, gdp
);
226 if (!flex_bg
|| ext4_block_in_group(sb
, tmp
, block_group
))
227 ext4_set_bit(EXT4_B2C(sbi
, tmp
- start
), bh
->b_data
);
229 tmp
= ext4_inode_table(sb
, gdp
);
230 for (; tmp
< ext4_inode_table(sb
, gdp
) +
231 sbi
->s_itb_per_group
; tmp
++) {
232 if (!flex_bg
|| ext4_block_in_group(sb
, tmp
, block_group
))
233 ext4_set_bit(EXT4_B2C(sbi
, tmp
- start
), bh
->b_data
);
237 * Also if the number of blocks within the group is less than
238 * the blocksize * 8 ( which is the size of bitmap ), set rest
239 * of the block bitmap to 1
241 ext4_mark_bitmap_end(num_clusters_in_group(sb
, block_group
),
242 sb
->s_blocksize
* 8, bh
->b_data
);
243 ext4_block_bitmap_csum_set(sb
, block_group
, gdp
, bh
);
244 ext4_group_desc_csum_set(sb
, block_group
, gdp
);
248 /* Return the number of free blocks in a block group. It is used when
249 * the block bitmap is uninitialized, so we can't just count the bits
251 unsigned ext4_free_clusters_after_init(struct super_block
*sb
,
252 ext4_group_t block_group
,
253 struct ext4_group_desc
*gdp
)
255 return num_clusters_in_group(sb
, block_group
) -
256 ext4_num_overhead_clusters(sb
, block_group
, gdp
);
260 * The free blocks are managed by bitmaps. A file system contains several
261 * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap
262 * block for inodes, N blocks for the inode table and data blocks.
264 * The file system contains group descriptors which are located after the
265 * super block. Each descriptor contains the number of the bitmap block and
266 * the free blocks count in the block. The descriptors are loaded in memory
267 * when a file system is mounted (see ext4_fill_super).
271 * ext4_get_group_desc() -- load group descriptor from disk
273 * @block_group: given block group
274 * @bh: pointer to the buffer head to store the block
277 struct ext4_group_desc
* ext4_get_group_desc(struct super_block
*sb
,
278 ext4_group_t block_group
,
279 struct buffer_head
**bh
)
281 unsigned int group_desc
;
283 ext4_group_t ngroups
= ext4_get_groups_count(sb
);
284 struct ext4_group_desc
*desc
;
285 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
287 if (block_group
>= ngroups
) {
288 ext4_error(sb
, "block_group >= groups_count - block_group = %u,"
289 " groups_count = %u", block_group
, ngroups
);
294 group_desc
= block_group
>> EXT4_DESC_PER_BLOCK_BITS(sb
);
295 offset
= block_group
& (EXT4_DESC_PER_BLOCK(sb
) - 1);
296 if (!sbi
->s_group_desc
[group_desc
]) {
297 ext4_error(sb
, "Group descriptor not loaded - "
298 "block_group = %u, group_desc = %u, desc = %u",
299 block_group
, group_desc
, offset
);
303 desc
= (struct ext4_group_desc
*)(
304 (__u8
*)sbi
->s_group_desc
[group_desc
]->b_data
+
305 offset
* EXT4_DESC_SIZE(sb
));
307 *bh
= sbi
->s_group_desc
[group_desc
];
312 * Return the block number which was discovered to be invalid, or 0 if
313 * the block bitmap is valid.
315 static ext4_fsblk_t
ext4_valid_block_bitmap(struct super_block
*sb
,
316 struct ext4_group_desc
*desc
,
317 ext4_group_t block_group
,
318 struct buffer_head
*bh
)
320 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
321 ext4_grpblk_t offset
;
322 ext4_grpblk_t next_zero_bit
;
324 ext4_fsblk_t group_first_block
;
326 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
)) {
327 /* with FLEX_BG, the inode/block bitmaps and itable
328 * blocks may not be in the group at all
329 * so the bitmap validation will be skipped for those groups
330 * or it has to also read the block group where the bitmaps
331 * are located to verify they are set.
335 group_first_block
= ext4_group_first_block_no(sb
, block_group
);
337 /* check whether block bitmap block number is set */
338 blk
= ext4_block_bitmap(sb
, desc
);
339 offset
= blk
- group_first_block
;
340 if (!ext4_test_bit(EXT4_B2C(sbi
, offset
), bh
->b_data
))
341 /* bad block bitmap */
344 /* check whether the inode bitmap block number is set */
345 blk
= ext4_inode_bitmap(sb
, desc
);
346 offset
= blk
- group_first_block
;
347 if (!ext4_test_bit(EXT4_B2C(sbi
, offset
), bh
->b_data
))
348 /* bad block bitmap */
351 /* check whether the inode table block number is set */
352 blk
= ext4_inode_table(sb
, desc
);
353 offset
= blk
- group_first_block
;
354 next_zero_bit
= ext4_find_next_zero_bit(bh
->b_data
,
355 EXT4_B2C(sbi
, offset
+ EXT4_SB(sb
)->s_itb_per_group
),
356 EXT4_B2C(sbi
, offset
));
358 EXT4_B2C(sbi
, offset
+ EXT4_SB(sb
)->s_itb_per_group
))
359 /* bad bitmap for inode tables */
364 static void ext4_validate_block_bitmap(struct super_block
*sb
,
365 struct ext4_group_desc
*desc
,
366 ext4_group_t block_group
,
367 struct buffer_head
*bh
)
370 struct ext4_group_info
*grp
= ext4_get_group_info(sb
, block_group
);
371 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
373 if (buffer_verified(bh
))
376 ext4_lock_group(sb
, block_group
);
377 blk
= ext4_valid_block_bitmap(sb
, desc
, block_group
, bh
);
378 if (unlikely(blk
!= 0)) {
379 ext4_unlock_group(sb
, block_group
);
380 ext4_error(sb
, "bg %u: block %llu: invalid block bitmap",
382 if (!EXT4_MB_GRP_BBITMAP_CORRUPT(grp
))
383 percpu_counter_sub(&sbi
->s_freeclusters_counter
,
385 set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT
, &grp
->bb_state
);
388 if (unlikely(!ext4_block_bitmap_csum_verify(sb
, block_group
,
390 ext4_unlock_group(sb
, block_group
);
391 ext4_error(sb
, "bg %u: bad block bitmap checksum", block_group
);
392 if (!EXT4_MB_GRP_BBITMAP_CORRUPT(grp
))
393 percpu_counter_sub(&sbi
->s_freeclusters_counter
,
395 set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT
, &grp
->bb_state
);
398 set_buffer_verified(bh
);
399 ext4_unlock_group(sb
, block_group
);
403 * ext4_read_block_bitmap_nowait()
405 * @block_group: given block group
407 * Read the bitmap for a given block_group,and validate the
408 * bits for block/inode/inode tables are set in the bitmaps
410 * Return buffer_head on success or NULL in case of failure.
413 ext4_read_block_bitmap_nowait(struct super_block
*sb
, ext4_group_t block_group
)
415 struct ext4_group_desc
*desc
;
416 struct buffer_head
*bh
;
417 ext4_fsblk_t bitmap_blk
;
419 desc
= ext4_get_group_desc(sb
, block_group
, NULL
);
422 bitmap_blk
= ext4_block_bitmap(sb
, desc
);
423 bh
= sb_getblk(sb
, bitmap_blk
);
425 ext4_error(sb
, "Cannot get buffer for block bitmap - "
426 "block_group = %u, block_bitmap = %llu",
427 block_group
, bitmap_blk
);
431 if (bitmap_uptodate(bh
))
435 if (bitmap_uptodate(bh
)) {
439 ext4_lock_group(sb
, block_group
);
440 if (desc
->bg_flags
& cpu_to_le16(EXT4_BG_BLOCK_UNINIT
)) {
443 err
= ext4_init_block_bitmap(sb
, bh
, block_group
, desc
);
444 set_bitmap_uptodate(bh
);
445 set_buffer_uptodate(bh
);
446 ext4_unlock_group(sb
, block_group
);
449 ext4_error(sb
, "Checksum bad for grp %u", block_group
);
452 ext4_unlock_group(sb
, block_group
);
453 if (buffer_uptodate(bh
)) {
455 * if not uninit if bh is uptodate,
456 * bitmap is also uptodate
458 set_bitmap_uptodate(bh
);
463 * submit the buffer_head for reading
466 trace_ext4_read_block_bitmap_load(sb
, block_group
);
467 bh
->b_end_io
= ext4_end_bitmap_read
;
469 submit_bh(READ
| REQ_META
| REQ_PRIO
, bh
);
472 ext4_validate_block_bitmap(sb
, desc
, block_group
, bh
);
473 if (buffer_verified(bh
))
479 /* Returns 0 on success, 1 on error */
480 int ext4_wait_block_bitmap(struct super_block
*sb
, ext4_group_t block_group
,
481 struct buffer_head
*bh
)
483 struct ext4_group_desc
*desc
;
487 desc
= ext4_get_group_desc(sb
, block_group
, NULL
);
491 if (!buffer_uptodate(bh
)) {
492 ext4_error(sb
, "Cannot read block bitmap - "
493 "block_group = %u, block_bitmap = %llu",
494 block_group
, (unsigned long long) bh
->b_blocknr
);
497 clear_buffer_new(bh
);
498 /* Panic or remount fs read-only if block bitmap is invalid */
499 ext4_validate_block_bitmap(sb
, desc
, block_group
, bh
);
500 /* ...but check for error just in case errors=continue. */
501 return !buffer_verified(bh
);
505 ext4_read_block_bitmap(struct super_block
*sb
, ext4_group_t block_group
)
507 struct buffer_head
*bh
;
509 bh
= ext4_read_block_bitmap_nowait(sb
, block_group
);
512 if (ext4_wait_block_bitmap(sb
, block_group
, bh
)) {
520 * ext4_has_free_clusters()
521 * @sbi: in-core super block structure.
522 * @nclusters: number of needed blocks
523 * @flags: flags from ext4_mb_new_blocks()
525 * Check if filesystem has nclusters free & available for allocation.
526 * On success return 1, return 0 on failure.
528 static int ext4_has_free_clusters(struct ext4_sb_info
*sbi
,
529 s64 nclusters
, unsigned int flags
)
531 s64 free_clusters
, dirty_clusters
, rsv
, resv_clusters
;
532 struct percpu_counter
*fcc
= &sbi
->s_freeclusters_counter
;
533 struct percpu_counter
*dcc
= &sbi
->s_dirtyclusters_counter
;
535 free_clusters
= percpu_counter_read_positive(fcc
);
536 dirty_clusters
= percpu_counter_read_positive(dcc
);
537 resv_clusters
= atomic64_read(&sbi
->s_resv_clusters
);
540 * r_blocks_count should always be multiple of the cluster ratio so
541 * we are safe to do a plane bit shift only.
543 rsv
= (ext4_r_blocks_count(sbi
->s_es
) >> sbi
->s_cluster_bits
) +
546 if (free_clusters
- (nclusters
+ rsv
+ dirty_clusters
) <
547 EXT4_FREECLUSTERS_WATERMARK
) {
548 free_clusters
= percpu_counter_sum_positive(fcc
);
549 dirty_clusters
= percpu_counter_sum_positive(dcc
);
551 /* Check whether we have space after accounting for current
552 * dirty clusters & root reserved clusters.
554 if (free_clusters
>= (rsv
+ nclusters
+ dirty_clusters
))
557 /* Hm, nope. Are (enough) root reserved clusters available? */
558 if (uid_eq(sbi
->s_resuid
, current_fsuid()) ||
559 (!gid_eq(sbi
->s_resgid
, GLOBAL_ROOT_GID
) && in_group_p(sbi
->s_resgid
)) ||
560 capable(CAP_SYS_RESOURCE
) ||
561 (flags
& EXT4_MB_USE_ROOT_BLOCKS
)) {
563 if (free_clusters
>= (nclusters
+ dirty_clusters
+
567 /* No free blocks. Let's see if we can dip into reserved pool */
568 if (flags
& EXT4_MB_USE_RESERVED
) {
569 if (free_clusters
>= (nclusters
+ dirty_clusters
))
576 int ext4_claim_free_clusters(struct ext4_sb_info
*sbi
,
577 s64 nclusters
, unsigned int flags
)
579 if (ext4_has_free_clusters(sbi
, nclusters
, flags
)) {
580 percpu_counter_add(&sbi
->s_dirtyclusters_counter
, nclusters
);
587 * ext4_should_retry_alloc()
589 * @retries number of attemps has been made
591 * ext4_should_retry_alloc() is called when ENOSPC is returned, and if
592 * it is profitable to retry the operation, this function will wait
593 * for the current or committing transaction to complete, and then
596 * if the total number of retries exceed three times, return FALSE.
598 int ext4_should_retry_alloc(struct super_block
*sb
, int *retries
)
600 if (!ext4_has_free_clusters(EXT4_SB(sb
), 1, 0) ||
602 !EXT4_SB(sb
)->s_journal
)
605 jbd_debug(1, "%s: retrying operation after ENOSPC\n", sb
->s_id
);
607 return jbd2_journal_force_commit_nested(EXT4_SB(sb
)->s_journal
);
611 * ext4_new_meta_blocks() -- allocate block for meta data (indexing) blocks
613 * @handle: handle to this transaction
615 * @goal: given target block(filesystem wide)
616 * @count: pointer to total number of clusters needed
619 * Return 1st allocated block number on success, *count stores total account
620 * error stores in errp pointer
622 ext4_fsblk_t
ext4_new_meta_blocks(handle_t
*handle
, struct inode
*inode
,
623 ext4_fsblk_t goal
, unsigned int flags
,
624 unsigned long *count
, int *errp
)
626 struct ext4_allocation_request ar
;
629 memset(&ar
, 0, sizeof(ar
));
630 /* Fill with neighbour allocated blocks */
633 ar
.len
= count
? *count
: 1;
636 ret
= ext4_mb_new_blocks(handle
, &ar
, errp
);
640 * Account for the allocated meta blocks. We will never
641 * fail EDQUOT for metdata, but we do account for it.
643 if (!(*errp
) && (flags
& EXT4_MB_DELALLOC_RESERVED
)) {
644 spin_lock(&EXT4_I(inode
)->i_block_reservation_lock
);
645 spin_unlock(&EXT4_I(inode
)->i_block_reservation_lock
);
646 dquot_alloc_block_nofail(inode
,
647 EXT4_C2B(EXT4_SB(inode
->i_sb
), ar
.len
));
653 * ext4_count_free_clusters() -- count filesystem free clusters
656 * Adds up the number of free clusters from each block group.
658 ext4_fsblk_t
ext4_count_free_clusters(struct super_block
*sb
)
660 ext4_fsblk_t desc_count
;
661 struct ext4_group_desc
*gdp
;
663 ext4_group_t ngroups
= ext4_get_groups_count(sb
);
664 struct ext4_group_info
*grp
;
666 struct ext4_super_block
*es
;
667 ext4_fsblk_t bitmap_count
;
669 struct buffer_head
*bitmap_bh
= NULL
;
671 es
= EXT4_SB(sb
)->s_es
;
676 for (i
= 0; i
< ngroups
; i
++) {
677 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
681 if (EXT4_SB(sb
)->s_group_info
)
682 grp
= ext4_get_group_info(sb
, i
);
683 if (!grp
|| !EXT4_MB_GRP_BBITMAP_CORRUPT(grp
))
684 desc_count
+= ext4_free_group_clusters(sb
, gdp
);
686 bitmap_bh
= ext4_read_block_bitmap(sb
, i
);
687 if (bitmap_bh
== NULL
)
690 x
= ext4_count_free(bitmap_bh
->b_data
,
691 EXT4_CLUSTERS_PER_GROUP(sb
) / 8);
692 printk(KERN_DEBUG
"group %u: stored = %d, counted = %u\n",
693 i
, ext4_free_group_clusters(sb
, gdp
), x
);
697 printk(KERN_DEBUG
"ext4_count_free_clusters: stored = %llu"
698 ", computed = %llu, %llu\n",
699 EXT4_NUM_B2C(EXT4_SB(sb
), ext4_free_blocks_count(es
)),
700 desc_count
, bitmap_count
);
704 for (i
= 0; i
< ngroups
; i
++) {
705 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
709 if (EXT4_SB(sb
)->s_group_info
)
710 grp
= ext4_get_group_info(sb
, i
);
711 if (!grp
|| !EXT4_MB_GRP_BBITMAP_CORRUPT(grp
))
712 desc_count
+= ext4_free_group_clusters(sb
, gdp
);
719 static inline int test_root(ext4_group_t a
, int b
)
733 * ext4_bg_has_super - number of blocks used by the superblock in group
734 * @sb: superblock for filesystem
735 * @group: group number to check
737 * Return the number of blocks used by the superblock (primary or backup)
738 * in this group. Currently this will be only 0 or 1.
740 int ext4_bg_has_super(struct super_block
*sb
, ext4_group_t group
)
742 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
746 if (EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_SPARSE_SUPER2
)) {
747 if (group
== le32_to_cpu(es
->s_backup_bgs
[0]) ||
748 group
== le32_to_cpu(es
->s_backup_bgs
[1]))
752 if ((group
<= 1) || !EXT4_HAS_RO_COMPAT_FEATURE(sb
,
753 EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER
))
757 if (test_root(group
, 3) || (test_root(group
, 5)) ||
764 static unsigned long ext4_bg_num_gdb_meta(struct super_block
*sb
,
767 unsigned long metagroup
= group
/ EXT4_DESC_PER_BLOCK(sb
);
768 ext4_group_t first
= metagroup
* EXT4_DESC_PER_BLOCK(sb
);
769 ext4_group_t last
= first
+ EXT4_DESC_PER_BLOCK(sb
) - 1;
771 if (group
== first
|| group
== first
+ 1 || group
== last
)
776 static unsigned long ext4_bg_num_gdb_nometa(struct super_block
*sb
,
779 if (!ext4_bg_has_super(sb
, group
))
782 if (EXT4_HAS_INCOMPAT_FEATURE(sb
,EXT4_FEATURE_INCOMPAT_META_BG
))
783 return le32_to_cpu(EXT4_SB(sb
)->s_es
->s_first_meta_bg
);
785 return EXT4_SB(sb
)->s_gdb_count
;
789 * ext4_bg_num_gdb - number of blocks used by the group table in group
790 * @sb: superblock for filesystem
791 * @group: group number to check
793 * Return the number of blocks used by the group descriptor table
794 * (primary or backup) in this group. In the future there may be a
795 * different number of descriptor blocks in each group.
797 unsigned long ext4_bg_num_gdb(struct super_block
*sb
, ext4_group_t group
)
799 unsigned long first_meta_bg
=
800 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_first_meta_bg
);
801 unsigned long metagroup
= group
/ EXT4_DESC_PER_BLOCK(sb
);
803 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
,EXT4_FEATURE_INCOMPAT_META_BG
) ||
804 metagroup
< first_meta_bg
)
805 return ext4_bg_num_gdb_nometa(sb
, group
);
807 return ext4_bg_num_gdb_meta(sb
,group
);
812 * This function returns the number of file system metadata clusters at
813 * the beginning of a block group, including the reserved gdt blocks.
815 static unsigned ext4_num_base_meta_clusters(struct super_block
*sb
,
816 ext4_group_t block_group
)
818 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
821 /* Check for superblock and gdt backups in this group */
822 num
= ext4_bg_has_super(sb
, block_group
);
824 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_META_BG
) ||
825 block_group
< le32_to_cpu(sbi
->s_es
->s_first_meta_bg
) *
826 sbi
->s_desc_per_block
) {
828 num
+= ext4_bg_num_gdb(sb
, block_group
);
829 num
+= le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
);
831 } else { /* For META_BG_BLOCK_GROUPS */
832 num
+= ext4_bg_num_gdb(sb
, block_group
);
834 return EXT4_NUM_B2C(sbi
, num
);
837 * ext4_inode_to_goal_block - return a hint for block allocation
838 * @inode: inode for block allocation
840 * Return the ideal location to start allocating blocks for a
841 * newly created inode.
843 ext4_fsblk_t
ext4_inode_to_goal_block(struct inode
*inode
)
845 struct ext4_inode_info
*ei
= EXT4_I(inode
);
846 ext4_group_t block_group
;
847 ext4_grpblk_t colour
;
848 int flex_size
= ext4_flex_bg_size(EXT4_SB(inode
->i_sb
));
849 ext4_fsblk_t bg_start
;
850 ext4_fsblk_t last_block
;
852 block_group
= ei
->i_block_group
;
853 if (flex_size
>= EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME
) {
855 * If there are at least EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME
856 * block groups per flexgroup, reserve the first block
857 * group for directories and special files. Regular
858 * files will start at the second block group. This
859 * tends to speed up directory access and improves
862 block_group
&= ~(flex_size
-1);
863 if (S_ISREG(inode
->i_mode
))
866 bg_start
= ext4_group_first_block_no(inode
->i_sb
, block_group
);
867 last_block
= ext4_blocks_count(EXT4_SB(inode
->i_sb
)->s_es
) - 1;
870 * If we are doing delayed allocation, we don't need take
871 * colour into account.
873 if (test_opt(inode
->i_sb
, DELALLOC
))
876 if (bg_start
+ EXT4_BLOCKS_PER_GROUP(inode
->i_sb
) <= last_block
)
877 colour
= (current
->pid
% 16) *
878 (EXT4_BLOCKS_PER_GROUP(inode
->i_sb
) / 16);
880 colour
= (current
->pid
% 16) * ((last_block
- bg_start
) / 16);
881 return bg_start
+ colour
;