Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
ac27a0ec | 2 | /* |
617ba13b | 3 | * linux/fs/ext4/balloc.c |
ac27a0ec DK |
4 | * |
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) | |
9 | * | |
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 | |
13 | */ | |
14 | ||
15 | #include <linux/time.h> | |
16 | #include <linux/capability.h> | |
17 | #include <linux/fs.h> | |
ac27a0ec DK |
18 | #include <linux/quotaops.h> |
19 | #include <linux/buffer_head.h> | |
3dcf5451 CH |
20 | #include "ext4.h" |
21 | #include "ext4_jbd2.h" | |
e21675d4 | 22 | #include "mballoc.h" |
3dcf5451 | 23 | |
0562e0ba JZ |
24 | #include <trace/events/ext4.h> |
25 | ||
5f163cc7 ES |
26 | static unsigned ext4_num_base_meta_clusters(struct super_block *sb, |
27 | ext4_group_t block_group); | |
ac27a0ec DK |
28 | /* |
29 | * balloc.c contains the blocks allocation and deallocation routines | |
30 | */ | |
31 | ||
bd86298e LC |
32 | /* |
33 | * Calculate block group number for a given block number | |
34 | */ | |
35 | ext4_group_t ext4_get_group_number(struct super_block *sb, | |
36 | ext4_fsblk_t block) | |
37 | { | |
38 | ext4_group_t group; | |
39 | ||
40 | if (test_opt2(sb, STD_GROUP_SIZE)) | |
960fd856 TT |
41 | group = (block - |
42 | le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block)) >> | |
bd86298e LC |
43 | (EXT4_BLOCK_SIZE_BITS(sb) + EXT4_CLUSTER_BITS(sb) + 3); |
44 | else | |
45 | ext4_get_group_no_and_offset(sb, block, &group, NULL); | |
46 | return group; | |
47 | } | |
48 | ||
72b64b59 | 49 | /* |
3212a80a TT |
50 | * Calculate the block group number and offset into the block/cluster |
51 | * allocation bitmap, given a block number | |
72b64b59 AM |
52 | */ |
53 | void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr, | |
fd2d4291 | 54 | ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp) |
72b64b59 | 55 | { |
8c55e204 | 56 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; |
72b64b59 AM |
57 | ext4_grpblk_t offset; |
58 | ||
8c55e204 | 59 | blocknr = blocknr - le32_to_cpu(es->s_first_data_block); |
3212a80a TT |
60 | offset = do_div(blocknr, EXT4_BLOCKS_PER_GROUP(sb)) >> |
61 | EXT4_SB(sb)->s_cluster_bits; | |
72b64b59 AM |
62 | if (offsetp) |
63 | *offsetp = offset; | |
64 | if (blockgrpp) | |
8c55e204 | 65 | *blockgrpp = blocknr; |
72b64b59 AM |
66 | |
67 | } | |
68 | ||
68911009 LC |
69 | /* |
70 | * Check whether the 'block' lives within the 'block_group'. Returns 1 if so | |
71 | * and 0 otherwise. | |
72 | */ | |
73 | static inline int ext4_block_in_group(struct super_block *sb, | |
74 | ext4_fsblk_t block, | |
75 | ext4_group_t block_group) | |
0bf7e837 JS |
76 | { |
77 | ext4_group_t actual_group; | |
68911009 | 78 | |
bd86298e | 79 | actual_group = ext4_get_group_number(sb, block); |
68911009 | 80 | return (actual_group == block_group) ? 1 : 0; |
0bf7e837 JS |
81 | } |
82 | ||
d5b8f310 TT |
83 | /* Return the number of clusters used for file system metadata; this |
84 | * represents the overhead needed by the file system. | |
85 | */ | |
c197855e SH |
86 | static unsigned ext4_num_overhead_clusters(struct super_block *sb, |
87 | ext4_group_t block_group, | |
88 | struct ext4_group_desc *gdp) | |
0bf7e837 | 89 | { |
d5b8f310 TT |
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; | |
0bf7e837 | 94 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
0bf7e837 | 95 | |
d5b8f310 TT |
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); | |
100 | ||
101 | /* | |
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. | |
112 | */ | |
113 | if (ext4_block_in_group(sb, ext4_block_bitmap(sb, gdp), block_group)) { | |
b0dd6b70 TT |
114 | block_cluster = EXT4_B2C(sbi, |
115 | ext4_block_bitmap(sb, gdp) - start); | |
d5b8f310 TT |
116 | if (block_cluster < num_clusters) |
117 | block_cluster = -1; | |
118 | else if (block_cluster == num_clusters) { | |
119 | num_clusters++; | |
120 | block_cluster = -1; | |
121 | } | |
122 | } | |
123 | ||
124 | if (ext4_block_in_group(sb, ext4_inode_bitmap(sb, gdp), block_group)) { | |
125 | inode_cluster = EXT4_B2C(sbi, | |
b0dd6b70 | 126 | ext4_inode_bitmap(sb, gdp) - start); |
d5b8f310 TT |
127 | if (inode_cluster < num_clusters) |
128 | inode_cluster = -1; | |
129 | else if (inode_cluster == num_clusters) { | |
130 | num_clusters++; | |
131 | inode_cluster = -1; | |
132 | } | |
133 | } | |
134 | ||
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)) { | |
b0dd6b70 | 138 | c = EXT4_B2C(sbi, itbl_blk + i - start); |
d5b8f310 TT |
139 | if ((c < num_clusters) || (c == inode_cluster) || |
140 | (c == block_cluster) || (c == itbl_cluster)) | |
141 | continue; | |
142 | if (c == num_clusters) { | |
143 | num_clusters++; | |
144 | continue; | |
145 | } | |
146 | num_clusters++; | |
147 | itbl_cluster = c; | |
0bf7e837 JS |
148 | } |
149 | } | |
d5b8f310 TT |
150 | |
151 | if (block_cluster != -1) | |
152 | num_clusters++; | |
153 | if (inode_cluster != -1) | |
154 | num_clusters++; | |
155 | ||
156 | return num_clusters; | |
0bf7e837 | 157 | } |
c2ea3fde | 158 | |
d5b8f310 TT |
159 | static unsigned int num_clusters_in_group(struct super_block *sb, |
160 | ext4_group_t block_group) | |
49f7f9af | 161 | { |
d5b8f310 TT |
162 | unsigned int blocks; |
163 | ||
49f7f9af TT |
164 | if (block_group == ext4_get_groups_count(sb) - 1) { |
165 | /* | |
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 | |
169 | * blocks. | |
170 | */ | |
d5b8f310 | 171 | blocks = ext4_blocks_count(EXT4_SB(sb)->s_es) - |
49f7f9af TT |
172 | ext4_group_first_block_no(sb, block_group); |
173 | } else | |
d5b8f310 TT |
174 | blocks = EXT4_BLOCKS_PER_GROUP(sb); |
175 | return EXT4_NUM_B2C(EXT4_SB(sb), blocks); | |
49f7f9af TT |
176 | } |
177 | ||
fd034a84 | 178 | /* Initializes an uninitialized block bitmap */ |
aef4885a | 179 | static int ext4_init_block_bitmap(struct super_block *sb, |
c197855e SH |
180 | struct buffer_head *bh, |
181 | ext4_group_t block_group, | |
182 | struct ext4_group_desc *gdp) | |
717d50e4 | 183 | { |
d5b8f310 | 184 | unsigned int bit, bit_max; |
717d50e4 | 185 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
fd034a84 | 186 | ext4_fsblk_t start, tmp; |
fd034a84 TT |
187 | |
188 | J_ASSERT_BH(bh, buffer_locked(bh)); | |
189 | ||
190 | /* If checksum is bad mark all blocks used to prevent allocation | |
191 | * essentially implementing a per-group read-only flag. */ | |
feb0ab32 | 192 | if (!ext4_group_desc_csum_verify(sb, block_group, gdp)) { |
db79e6d1 WS |
193 | ext4_mark_group_bitmap_corrupted(sb, block_group, |
194 | EXT4_GROUP_INFO_BBITMAP_CORRUPT | | |
195 | EXT4_GROUP_INFO_IBITMAP_CORRUPT); | |
6a797d27 | 196 | return -EFSBADCRC; |
717d50e4 | 197 | } |
fd034a84 | 198 | memset(bh->b_data, 0, sb->s_blocksize); |
717d50e4 | 199 | |
d5b8f310 | 200 | bit_max = ext4_num_base_meta_clusters(sb, block_group); |
5b9554dc TT |
201 | if ((bit_max >> 3) >= bh->b_size) |
202 | return -EFSCORRUPTED; | |
203 | ||
fd034a84 TT |
204 | for (bit = 0; bit < bit_max; bit++) |
205 | ext4_set_bit(bit, bh->b_data); | |
d00a6d7b | 206 | |
fd034a84 | 207 | start = ext4_group_first_block_no(sb, block_group); |
717d50e4 | 208 | |
fd034a84 TT |
209 | /* Set bits for block and inode bitmaps, and inode table */ |
210 | tmp = ext4_block_bitmap(sb, gdp); | |
819b23f1 | 211 | if (ext4_block_in_group(sb, tmp, block_group)) |
d5b8f310 | 212 | ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data); |
717d50e4 | 213 | |
fd034a84 | 214 | tmp = ext4_inode_bitmap(sb, gdp); |
819b23f1 | 215 | if (ext4_block_in_group(sb, tmp, block_group)) |
d5b8f310 | 216 | ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data); |
0bf7e837 | 217 | |
fd034a84 TT |
218 | tmp = ext4_inode_table(sb, gdp); |
219 | for (; tmp < ext4_inode_table(sb, gdp) + | |
220 | sbi->s_itb_per_group; tmp++) { | |
819b23f1 | 221 | if (ext4_block_in_group(sb, tmp, block_group)) |
d5b8f310 | 222 | ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data); |
717d50e4 | 223 | } |
d5b8f310 | 224 | |
fd034a84 TT |
225 | /* |
226 | * Also if the number of blocks within the group is less than | |
227 | * the blocksize * 8 ( which is the size of bitmap ), set rest | |
228 | * of the block bitmap to 1 | |
229 | */ | |
d5b8f310 | 230 | ext4_mark_bitmap_end(num_clusters_in_group(sb, block_group), |
fd034a84 | 231 | sb->s_blocksize * 8, bh->b_data); |
aef4885a | 232 | return 0; |
717d50e4 AD |
233 | } |
234 | ||
fd034a84 TT |
235 | /* Return the number of free blocks in a block group. It is used when |
236 | * the block bitmap is uninitialized, so we can't just count the bits | |
237 | * in the bitmap. */ | |
cff1dfd7 TT |
238 | unsigned ext4_free_clusters_after_init(struct super_block *sb, |
239 | ext4_group_t block_group, | |
240 | struct ext4_group_desc *gdp) | |
fd034a84 | 241 | { |
d5b8f310 TT |
242 | return num_clusters_in_group(sb, block_group) - |
243 | ext4_num_overhead_clusters(sb, block_group, gdp); | |
fd034a84 | 244 | } |
717d50e4 | 245 | |
ac27a0ec DK |
246 | /* |
247 | * The free blocks are managed by bitmaps. A file system contains several | |
248 | * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap | |
249 | * block for inodes, N blocks for the inode table and data blocks. | |
250 | * | |
251 | * The file system contains group descriptors which are located after the | |
252 | * super block. Each descriptor contains the number of the bitmap block and | |
253 | * the free blocks count in the block. The descriptors are loaded in memory | |
e627432c | 254 | * when a file system is mounted (see ext4_fill_super). |
ac27a0ec DK |
255 | */ |
256 | ||
ac27a0ec | 257 | /** |
617ba13b | 258 | * ext4_get_group_desc() -- load group descriptor from disk |
ac27a0ec DK |
259 | * @sb: super block |
260 | * @block_group: given block group | |
261 | * @bh: pointer to the buffer head to store the block | |
262 | * group descriptor | |
263 | */ | |
af5bc92d | 264 | struct ext4_group_desc * ext4_get_group_desc(struct super_block *sb, |
fd2d4291 | 265 | ext4_group_t block_group, |
af5bc92d | 266 | struct buffer_head **bh) |
ac27a0ec | 267 | { |
498e5f24 TT |
268 | unsigned int group_desc; |
269 | unsigned int offset; | |
8df9675f | 270 | ext4_group_t ngroups = ext4_get_groups_count(sb); |
af5bc92d | 271 | struct ext4_group_desc *desc; |
617ba13b | 272 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
1d0c3924 | 273 | struct buffer_head *bh_p; |
ac27a0ec | 274 | |
8df9675f | 275 | if (block_group >= ngroups) { |
12062ddd ES |
276 | ext4_error(sb, "block_group >= groups_count - block_group = %u," |
277 | " groups_count = %u", block_group, ngroups); | |
ac27a0ec DK |
278 | |
279 | return NULL; | |
280 | } | |
ac27a0ec | 281 | |
617ba13b MC |
282 | group_desc = block_group >> EXT4_DESC_PER_BLOCK_BITS(sb); |
283 | offset = block_group & (EXT4_DESC_PER_BLOCK(sb) - 1); | |
1d0c3924 TT |
284 | bh_p = sbi_array_rcu_deref(sbi, s_group_desc, group_desc); |
285 | /* | |
286 | * sbi_array_rcu_deref returns with rcu unlocked, this is ok since | |
287 | * the pointer being dereferenced won't be dereferenced again. By | |
288 | * looking at the usage in add_new_gdb() the value isn't modified, | |
289 | * just the pointer, and so it remains valid. | |
290 | */ | |
291 | if (!bh_p) { | |
12062ddd | 292 | ext4_error(sb, "Group descriptor not loaded - " |
498e5f24 | 293 | "block_group = %u, group_desc = %u, desc = %u", |
af5bc92d | 294 | block_group, group_desc, offset); |
ac27a0ec DK |
295 | return NULL; |
296 | } | |
297 | ||
0d1ee42f | 298 | desc = (struct ext4_group_desc *)( |
1d0c3924 | 299 | (__u8 *)bh_p->b_data + |
0d1ee42f | 300 | offset * EXT4_DESC_SIZE(sb)); |
ac27a0ec | 301 | if (bh) |
1d0c3924 | 302 | *bh = bh_p; |
0d1ee42f | 303 | return desc; |
ac27a0ec DK |
304 | } |
305 | ||
7a4c5de2 TT |
306 | /* |
307 | * Return the block number which was discovered to be invalid, or 0 if | |
308 | * the block bitmap is valid. | |
309 | */ | |
310 | static ext4_fsblk_t ext4_valid_block_bitmap(struct super_block *sb, | |
311 | struct ext4_group_desc *desc, | |
dbde0abe | 312 | ext4_group_t block_group, |
7a4c5de2 | 313 | struct buffer_head *bh) |
abcb2947 | 314 | { |
e674e5cb | 315 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
abcb2947 AK |
316 | ext4_grpblk_t offset; |
317 | ext4_grpblk_t next_zero_bit; | |
22be37ac | 318 | ext4_grpblk_t max_bit = EXT4_CLUSTERS_PER_GROUP(sb); |
7a4c5de2 | 319 | ext4_fsblk_t blk; |
abcb2947 AK |
320 | ext4_fsblk_t group_first_block; |
321 | ||
e2b911c5 | 322 | if (ext4_has_feature_flex_bg(sb)) { |
abcb2947 AK |
323 | /* with FLEX_BG, the inode/block bitmaps and itable |
324 | * blocks may not be in the group at all | |
325 | * so the bitmap validation will be skipped for those groups | |
326 | * or it has to also read the block group where the bitmaps | |
327 | * are located to verify they are set. | |
328 | */ | |
7a4c5de2 | 329 | return 0; |
abcb2947 AK |
330 | } |
331 | group_first_block = ext4_group_first_block_no(sb, block_group); | |
332 | ||
333 | /* check whether block bitmap block number is set */ | |
7a4c5de2 TT |
334 | blk = ext4_block_bitmap(sb, desc); |
335 | offset = blk - group_first_block; | |
22be37ac | 336 | if (offset < 0 || EXT4_B2C(sbi, offset) >= max_bit || |
7dac4a17 | 337 | !ext4_test_bit(EXT4_B2C(sbi, offset), bh->b_data)) |
abcb2947 | 338 | /* bad block bitmap */ |
7a4c5de2 | 339 | return blk; |
abcb2947 AK |
340 | |
341 | /* check whether the inode bitmap block number is set */ | |
7a4c5de2 TT |
342 | blk = ext4_inode_bitmap(sb, desc); |
343 | offset = blk - group_first_block; | |
22be37ac | 344 | if (offset < 0 || EXT4_B2C(sbi, offset) >= max_bit || |
7dac4a17 | 345 | !ext4_test_bit(EXT4_B2C(sbi, offset), bh->b_data)) |
abcb2947 | 346 | /* bad block bitmap */ |
7a4c5de2 | 347 | return blk; |
abcb2947 AK |
348 | |
349 | /* check whether the inode table block number is set */ | |
7a4c5de2 TT |
350 | blk = ext4_inode_table(sb, desc); |
351 | offset = blk - group_first_block; | |
22be37ac LC |
352 | if (offset < 0 || EXT4_B2C(sbi, offset) >= max_bit || |
353 | EXT4_B2C(sbi, offset + sbi->s_itb_per_group) >= max_bit) | |
7dac4a17 | 354 | return blk; |
abcb2947 | 355 | next_zero_bit = ext4_find_next_zero_bit(bh->b_data, |
49598e04 | 356 | EXT4_B2C(sbi, offset + sbi->s_itb_per_group), |
e674e5cb DW |
357 | EXT4_B2C(sbi, offset)); |
358 | if (next_zero_bit < | |
49598e04 | 359 | EXT4_B2C(sbi, offset + sbi->s_itb_per_group)) |
7a4c5de2 TT |
360 | /* bad bitmap for inode tables */ |
361 | return blk; | |
abcb2947 AK |
362 | return 0; |
363 | } | |
fa77dcfa | 364 | |
9008a58e DW |
365 | static int ext4_validate_block_bitmap(struct super_block *sb, |
366 | struct ext4_group_desc *desc, | |
367 | ext4_group_t block_group, | |
368 | struct buffer_head *bh) | |
fa77dcfa | 369 | { |
7a4c5de2 | 370 | ext4_fsblk_t blk; |
163a203d | 371 | struct ext4_group_info *grp = ext4_get_group_info(sb, block_group); |
7a4c5de2 | 372 | |
9008a58e DW |
373 | if (buffer_verified(bh)) |
374 | return 0; | |
375 | if (EXT4_MB_GRP_BBITMAP_CORRUPT(grp)) | |
376 | return -EFSCORRUPTED; | |
fa77dcfa DW |
377 | |
378 | ext4_lock_group(sb, block_group); | |
8d5a803c TT |
379 | if (buffer_verified(bh)) |
380 | goto verified; | |
9008a58e | 381 | if (unlikely(!ext4_block_bitmap_csum_verify(sb, block_group, |
46f870d6 TT |
382 | desc, bh) || |
383 | ext4_simulate_fail(sb, EXT4_SIM_BBITMAP_CRC))) { | |
7a4c5de2 | 384 | ext4_unlock_group(sb, block_group); |
9008a58e | 385 | ext4_error(sb, "bg %u: bad block bitmap checksum", block_group); |
db79e6d1 WS |
386 | ext4_mark_group_bitmap_corrupted(sb, block_group, |
387 | EXT4_GROUP_INFO_BBITMAP_CORRUPT); | |
9008a58e | 388 | return -EFSBADCRC; |
7a4c5de2 | 389 | } |
9008a58e DW |
390 | blk = ext4_valid_block_bitmap(sb, desc, block_group, bh); |
391 | if (unlikely(blk != 0)) { | |
7a4c5de2 | 392 | ext4_unlock_group(sb, block_group); |
9008a58e DW |
393 | ext4_error(sb, "bg %u: block %llu: invalid block bitmap", |
394 | block_group, blk); | |
db79e6d1 WS |
395 | ext4_mark_group_bitmap_corrupted(sb, block_group, |
396 | EXT4_GROUP_INFO_BBITMAP_CORRUPT); | |
9008a58e | 397 | return -EFSCORRUPTED; |
7a4c5de2 TT |
398 | } |
399 | set_buffer_verified(bh); | |
8d5a803c | 400 | verified: |
fa77dcfa | 401 | ext4_unlock_group(sb, block_group); |
9008a58e | 402 | return 0; |
fa77dcfa DW |
403 | } |
404 | ||
ac27a0ec | 405 | /** |
15b49132 | 406 | * ext4_read_block_bitmap_nowait() |
ac27a0ec DK |
407 | * @sb: super block |
408 | * @block_group: given block group | |
409 | * | |
abcb2947 AK |
410 | * Read the bitmap for a given block_group,and validate the |
411 | * bits for block/inode/inode tables are set in the bitmaps | |
ac27a0ec | 412 | * |
9033783c | 413 | * Return buffer_head on success or an ERR_PTR in case of failure. |
ac27a0ec | 414 | */ |
717d50e4 | 415 | struct buffer_head * |
cfd73237 AZ |
416 | ext4_read_block_bitmap_nowait(struct super_block *sb, ext4_group_t block_group, |
417 | bool ignore_locked) | |
ac27a0ec | 418 | { |
af5bc92d | 419 | struct ext4_group_desc *desc; |
7dac4a17 | 420 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
813e5727 | 421 | struct buffer_head *bh; |
7c9e69fa | 422 | ext4_fsblk_t bitmap_blk; |
9008a58e | 423 | int err; |
ac27a0ec | 424 | |
717d50e4 | 425 | desc = ext4_get_group_desc(sb, block_group, NULL); |
ac27a0ec | 426 | if (!desc) |
9008a58e | 427 | return ERR_PTR(-EFSCORRUPTED); |
7c9e69fa | 428 | bitmap_blk = ext4_block_bitmap(sb, desc); |
7dac4a17 TT |
429 | if ((bitmap_blk <= le32_to_cpu(sbi->s_es->s_first_data_block)) || |
430 | (bitmap_blk >= ext4_blocks_count(sbi->s_es))) { | |
431 | ext4_error(sb, "Invalid block bitmap block %llu in " | |
432 | "block_group %u", bitmap_blk, block_group); | |
736dedbb WS |
433 | ext4_mark_group_bitmap_corrupted(sb, block_group, |
434 | EXT4_GROUP_INFO_BBITMAP_CORRUPT); | |
7dac4a17 TT |
435 | return ERR_PTR(-EFSCORRUPTED); |
436 | } | |
abcb2947 AK |
437 | bh = sb_getblk(sb, bitmap_blk); |
438 | if (unlikely(!bh)) { | |
5ef2a699 WS |
439 | ext4_warning(sb, "Cannot get buffer for block bitmap - " |
440 | "block_group = %u, block_bitmap = %llu", | |
441 | block_group, bitmap_blk); | |
9008a58e | 442 | return ERR_PTR(-ENOMEM); |
abcb2947 | 443 | } |
2ccb5fb9 | 444 | |
cfd73237 AZ |
445 | if (ignore_locked && buffer_locked(bh)) { |
446 | /* buffer under IO already, return if called for prefetching */ | |
447 | put_bh(bh); | |
448 | return NULL; | |
449 | } | |
450 | ||
2ccb5fb9 | 451 | if (bitmap_uptodate(bh)) |
fa77dcfa | 452 | goto verify; |
abcb2947 | 453 | |
c806e68f | 454 | lock_buffer(bh); |
2ccb5fb9 AK |
455 | if (bitmap_uptodate(bh)) { |
456 | unlock_buffer(bh); | |
fa77dcfa | 457 | goto verify; |
2ccb5fb9 | 458 | } |
955ce5f5 | 459 | ext4_lock_group(sb, block_group); |
8844618d TT |
460 | if (ext4_has_group_desc_csum(sb) && |
461 | (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT))) { | |
462 | if (block_group == 0) { | |
463 | ext4_unlock_group(sb, block_group); | |
464 | unlock_buffer(bh); | |
465 | ext4_error(sb, "Block bitmap for bg 0 marked " | |
466 | "uninitialized"); | |
467 | err = -EFSCORRUPTED; | |
468 | goto out; | |
469 | } | |
aef4885a | 470 | err = ext4_init_block_bitmap(sb, bh, block_group, desc); |
2ccb5fb9 | 471 | set_bitmap_uptodate(bh); |
abcb2947 | 472 | set_buffer_uptodate(bh); |
044e6e3d | 473 | set_buffer_verified(bh); |
955ce5f5 | 474 | ext4_unlock_group(sb, block_group); |
3300beda | 475 | unlock_buffer(bh); |
05145bd7 JK |
476 | if (err) { |
477 | ext4_error(sb, "Failed to init block bitmap for group " | |
478 | "%u: %d", block_group, err); | |
9008a58e | 479 | goto out; |
05145bd7 | 480 | } |
41e5b7ed | 481 | goto verify; |
717d50e4 | 482 | } |
955ce5f5 | 483 | ext4_unlock_group(sb, block_group); |
2ccb5fb9 AK |
484 | if (buffer_uptodate(bh)) { |
485 | /* | |
486 | * if not uninit if bh is uptodate, | |
487 | * bitmap is also uptodate | |
488 | */ | |
489 | set_bitmap_uptodate(bh); | |
490 | unlock_buffer(bh); | |
fa77dcfa | 491 | goto verify; |
2ccb5fb9 AK |
492 | } |
493 | /* | |
813e5727 | 494 | * submit the buffer_head for reading |
2ccb5fb9 | 495 | */ |
813e5727 | 496 | set_buffer_new(bh); |
ab74c7b2 | 497 | trace_ext4_read_block_bitmap_load(sb, block_group, ignore_locked); |
2d069c08 | 498 | ext4_read_bh_nowait(bh, REQ_META | REQ_PRIO | |
499 | (ignore_locked ? REQ_RAHEAD : 0), | |
500 | ext4_end_bitmap_read); | |
813e5727 | 501 | return bh; |
fa77dcfa | 502 | verify: |
9008a58e DW |
503 | err = ext4_validate_block_bitmap(sb, desc, block_group, bh); |
504 | if (err) | |
505 | goto out; | |
506 | return bh; | |
507 | out: | |
48d9eb97 | 508 | put_bh(bh); |
9008a58e | 509 | return ERR_PTR(err); |
813e5727 TT |
510 | } |
511 | ||
9033783c | 512 | /* Returns 0 on success, -errno on error */ |
813e5727 TT |
513 | int ext4_wait_block_bitmap(struct super_block *sb, ext4_group_t block_group, |
514 | struct buffer_head *bh) | |
515 | { | |
516 | struct ext4_group_desc *desc; | |
517 | ||
518 | if (!buffer_new(bh)) | |
519 | return 0; | |
520 | desc = ext4_get_group_desc(sb, block_group, NULL); | |
521 | if (!desc) | |
9008a58e | 522 | return -EFSCORRUPTED; |
813e5727 | 523 | wait_on_buffer(bh); |
46f870d6 | 524 | ext4_simulate_fail_bh(sb, bh, EXT4_SIM_BBITMAP_EIO); |
813e5727 | 525 | if (!buffer_uptodate(bh)) { |
54d3adbc TT |
526 | ext4_error_err(sb, EIO, "Cannot read block bitmap - " |
527 | "block_group = %u, block_bitmap = %llu", | |
528 | block_group, (unsigned long long) bh->b_blocknr); | |
736dedbb WS |
529 | ext4_mark_group_bitmap_corrupted(sb, block_group, |
530 | EXT4_GROUP_INFO_BBITMAP_CORRUPT); | |
9008a58e | 531 | return -EIO; |
abcb2947 | 532 | } |
813e5727 TT |
533 | clear_buffer_new(bh); |
534 | /* Panic or remount fs read-only if block bitmap is invalid */ | |
9008a58e | 535 | return ext4_validate_block_bitmap(sb, desc, block_group, bh); |
813e5727 TT |
536 | } |
537 | ||
538 | struct buffer_head * | |
539 | ext4_read_block_bitmap(struct super_block *sb, ext4_group_t block_group) | |
540 | { | |
541 | struct buffer_head *bh; | |
9008a58e | 542 | int err; |
813e5727 | 543 | |
cfd73237 | 544 | bh = ext4_read_block_bitmap_nowait(sb, block_group, false); |
9008a58e DW |
545 | if (IS_ERR(bh)) |
546 | return bh; | |
547 | err = ext4_wait_block_bitmap(sb, block_group, bh); | |
548 | if (err) { | |
813e5727 | 549 | put_bh(bh); |
9008a58e | 550 | return ERR_PTR(err); |
813e5727 | 551 | } |
ac27a0ec DK |
552 | return bh; |
553 | } | |
ac27a0ec | 554 | |
8c3bf8a0 | 555 | /** |
df55c99d | 556 | * ext4_has_free_clusters() |
8c3bf8a0 | 557 | * @sbi: in-core super block structure. |
df55c99d TT |
558 | * @nclusters: number of needed blocks |
559 | * @flags: flags from ext4_mb_new_blocks() | |
8c3bf8a0 | 560 | * |
df55c99d | 561 | * Check if filesystem has nclusters free & available for allocation. |
8c3bf8a0 ES |
562 | * On success return 1, return 0 on failure. |
563 | */ | |
df55c99d TT |
564 | static int ext4_has_free_clusters(struct ext4_sb_info *sbi, |
565 | s64 nclusters, unsigned int flags) | |
a30d542a | 566 | { |
27dd4385 | 567 | s64 free_clusters, dirty_clusters, rsv, resv_clusters; |
57042651 | 568 | struct percpu_counter *fcc = &sbi->s_freeclusters_counter; |
df55c99d | 569 | struct percpu_counter *dcc = &sbi->s_dirtyclusters_counter; |
a30d542a | 570 | |
df55c99d TT |
571 | free_clusters = percpu_counter_read_positive(fcc); |
572 | dirty_clusters = percpu_counter_read_positive(dcc); | |
27dd4385 | 573 | resv_clusters = atomic64_read(&sbi->s_resv_clusters); |
304e220f LC |
574 | |
575 | /* | |
576 | * r_blocks_count should always be multiple of the cluster ratio so | |
577 | * we are safe to do a plane bit shift only. | |
578 | */ | |
27dd4385 LC |
579 | rsv = (ext4_r_blocks_count(sbi->s_es) >> sbi->s_cluster_bits) + |
580 | resv_clusters; | |
a30d542a | 581 | |
27dd4385 | 582 | if (free_clusters - (nclusters + rsv + dirty_clusters) < |
df55c99d | 583 | EXT4_FREECLUSTERS_WATERMARK) { |
304e220f | 584 | free_clusters = percpu_counter_sum_positive(fcc); |
df55c99d | 585 | dirty_clusters = percpu_counter_sum_positive(dcc); |
6bc6e63f | 586 | } |
df55c99d TT |
587 | /* Check whether we have space after accounting for current |
588 | * dirty clusters & root reserved clusters. | |
6bc6e63f | 589 | */ |
27dd4385 | 590 | if (free_clusters >= (rsv + nclusters + dirty_clusters)) |
a996031c | 591 | return 1; |
a30d542a | 592 | |
df55c99d | 593 | /* Hm, nope. Are (enough) root reserved clusters available? */ |
08cefc7a EB |
594 | if (uid_eq(sbi->s_resuid, current_fsuid()) || |
595 | (!gid_eq(sbi->s_resgid, GLOBAL_ROOT_GID) && in_group_p(sbi->s_resgid)) || | |
55f020db | 596 | capable(CAP_SYS_RESOURCE) || |
27dd4385 | 597 | (flags & EXT4_MB_USE_ROOT_BLOCKS)) { |
55f020db | 598 | |
27dd4385 LC |
599 | if (free_clusters >= (nclusters + dirty_clusters + |
600 | resv_clusters)) | |
601 | return 1; | |
602 | } | |
603 | /* No free blocks. Let's see if we can dip into reserved pool */ | |
604 | if (flags & EXT4_MB_USE_RESERVED) { | |
df55c99d | 605 | if (free_clusters >= (nclusters + dirty_clusters)) |
a996031c ES |
606 | return 1; |
607 | } | |
608 | ||
609 | return 0; | |
a30d542a AK |
610 | } |
611 | ||
e7d5f315 TT |
612 | int ext4_claim_free_clusters(struct ext4_sb_info *sbi, |
613 | s64 nclusters, unsigned int flags) | |
ac27a0ec | 614 | { |
df55c99d | 615 | if (ext4_has_free_clusters(sbi, nclusters, flags)) { |
e7d5f315 | 616 | percpu_counter_add(&sbi->s_dirtyclusters_counter, nclusters); |
16eb7295 | 617 | return 0; |
8c3bf8a0 ES |
618 | } else |
619 | return -ENOSPC; | |
a30d542a | 620 | } |
07031431 | 621 | |
ac27a0ec | 622 | /** |
c60990b3 | 623 | * ext4_should_retry_alloc() - check if a block allocation should be retried |
ac27a0ec | 624 | * @sb: super block |
c60990b3 | 625 | * @retries: number of attemps has been made |
ac27a0ec | 626 | * |
617ba13b | 627 | * ext4_should_retry_alloc() is called when ENOSPC is returned, and if |
ac27a0ec | 628 | * it is profitable to retry the operation, this function will wait |
25985edc | 629 | * for the current or committing transaction to complete, and then |
68fd9750 | 630 | * return TRUE. We will only retry once. |
ac27a0ec | 631 | */ |
617ba13b | 632 | int ext4_should_retry_alloc(struct super_block *sb, int *retries) |
ac27a0ec | 633 | { |
df55c99d | 634 | if (!ext4_has_free_clusters(EXT4_SB(sb), 1, 0) || |
68fd9750 | 635 | (*retries)++ > 1 || |
8f64b32e | 636 | !EXT4_SB(sb)->s_journal) |
ac27a0ec DK |
637 | return 0; |
638 | ||
d08854f5 | 639 | smp_mb(); |
68fd9750 TT |
640 | if (EXT4_SB(sb)->s_mb_free_pending == 0) |
641 | return 0; | |
642 | ||
643 | jbd_debug(1, "%s: retrying operation after ENOSPC\n", sb->s_id); | |
644 | jbd2_journal_force_commit_nested(EXT4_SB(sb)->s_journal); | |
dbc427ce | 645 | return 1; |
ac27a0ec DK |
646 | } |
647 | ||
654b4908 | 648 | /* |
d2a17637 | 649 | * ext4_new_meta_blocks() -- allocate block for meta data (indexing) blocks |
654b4908 AK |
650 | * |
651 | * @handle: handle to this transaction | |
652 | * @inode: file inode | |
653 | * @goal: given target block(filesystem wide) | |
7b415bf6 | 654 | * @count: pointer to total number of clusters needed |
654b4908 AK |
655 | * @errp: error code |
656 | * | |
97df5d15 | 657 | * Return 1st allocated block number on success, *count stores total account |
d2a17637 | 658 | * error stores in errp pointer |
654b4908 | 659 | */ |
d2a17637 | 660 | ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode, |
55f020db AH |
661 | ext4_fsblk_t goal, unsigned int flags, |
662 | unsigned long *count, int *errp) | |
654b4908 | 663 | { |
97df5d15 | 664 | struct ext4_allocation_request ar; |
d2a17637 | 665 | ext4_fsblk_t ret; |
97df5d15 TT |
666 | |
667 | memset(&ar, 0, sizeof(ar)); | |
668 | /* Fill with neighbour allocated blocks */ | |
669 | ar.inode = inode; | |
670 | ar.goal = goal; | |
671 | ar.len = count ? *count : 1; | |
55f020db | 672 | ar.flags = flags; |
97df5d15 TT |
673 | |
674 | ret = ext4_mb_new_blocks(handle, &ar, errp); | |
675 | if (count) | |
676 | *count = ar.len; | |
d2a17637 | 677 | /* |
72b8ab9d ES |
678 | * Account for the allocated meta blocks. We will never |
679 | * fail EDQUOT for metdata, but we do account for it. | |
d2a17637 | 680 | */ |
e3cf5d5d | 681 | if (!(*errp) && (flags & EXT4_MB_DELALLOC_RESERVED)) { |
7b415bf6 AK |
682 | dquot_alloc_block_nofail(inode, |
683 | EXT4_C2B(EXT4_SB(inode->i_sb), ar.len)); | |
d2a17637 MC |
684 | } |
685 | return ret; | |
654b4908 AK |
686 | } |
687 | ||
ac27a0ec | 688 | /** |
5dee5437 | 689 | * ext4_count_free_clusters() -- count filesystem free clusters |
ac27a0ec DK |
690 | * @sb: superblock |
691 | * | |
5dee5437 | 692 | * Adds up the number of free clusters from each block group. |
ac27a0ec | 693 | */ |
5dee5437 | 694 | ext4_fsblk_t ext4_count_free_clusters(struct super_block *sb) |
ac27a0ec | 695 | { |
617ba13b MC |
696 | ext4_fsblk_t desc_count; |
697 | struct ext4_group_desc *gdp; | |
fd2d4291 | 698 | ext4_group_t i; |
8df9675f | 699 | ext4_group_t ngroups = ext4_get_groups_count(sb); |
2746f7a1 | 700 | struct ext4_group_info *grp; |
617ba13b MC |
701 | #ifdef EXT4FS_DEBUG |
702 | struct ext4_super_block *es; | |
703 | ext4_fsblk_t bitmap_count; | |
498e5f24 | 704 | unsigned int x; |
ac27a0ec DK |
705 | struct buffer_head *bitmap_bh = NULL; |
706 | ||
617ba13b | 707 | es = EXT4_SB(sb)->s_es; |
ac27a0ec DK |
708 | desc_count = 0; |
709 | bitmap_count = 0; | |
710 | gdp = NULL; | |
711 | ||
ac27a0ec | 712 | for (i = 0; i < ngroups; i++) { |
617ba13b | 713 | gdp = ext4_get_group_desc(sb, i, NULL); |
ac27a0ec DK |
714 | if (!gdp) |
715 | continue; | |
2746f7a1 DW |
716 | grp = NULL; |
717 | if (EXT4_SB(sb)->s_group_info) | |
718 | grp = ext4_get_group_info(sb, i); | |
719 | if (!grp || !EXT4_MB_GRP_BBITMAP_CORRUPT(grp)) | |
720 | desc_count += ext4_free_group_clusters(sb, gdp); | |
ac27a0ec | 721 | brelse(bitmap_bh); |
574ca174 | 722 | bitmap_bh = ext4_read_block_bitmap(sb, i); |
9008a58e DW |
723 | if (IS_ERR(bitmap_bh)) { |
724 | bitmap_bh = NULL; | |
ac27a0ec | 725 | continue; |
9008a58e | 726 | } |
ac27a0ec | 727 | |
f6fb99ca | 728 | x = ext4_count_free(bitmap_bh->b_data, |
036acea2 | 729 | EXT4_CLUSTERS_PER_GROUP(sb) / 8); |
9fd9784c | 730 | printk(KERN_DEBUG "group %u: stored = %d, counted = %u\n", |
021b65bb | 731 | i, ext4_free_group_clusters(sb, gdp), x); |
ac27a0ec DK |
732 | bitmap_count += x; |
733 | } | |
734 | brelse(bitmap_bh); | |
5dee5437 TT |
735 | printk(KERN_DEBUG "ext4_count_free_clusters: stored = %llu" |
736 | ", computed = %llu, %llu\n", | |
810da240 | 737 | EXT4_NUM_B2C(EXT4_SB(sb), ext4_free_blocks_count(es)), |
4776004f | 738 | desc_count, bitmap_count); |
ac27a0ec DK |
739 | return bitmap_count; |
740 | #else | |
741 | desc_count = 0; | |
ac27a0ec | 742 | for (i = 0; i < ngroups; i++) { |
617ba13b | 743 | gdp = ext4_get_group_desc(sb, i, NULL); |
ac27a0ec DK |
744 | if (!gdp) |
745 | continue; | |
2746f7a1 DW |
746 | grp = NULL; |
747 | if (EXT4_SB(sb)->s_group_info) | |
748 | grp = ext4_get_group_info(sb, i); | |
749 | if (!grp || !EXT4_MB_GRP_BBITMAP_CORRUPT(grp)) | |
750 | desc_count += ext4_free_group_clusters(sb, gdp); | |
ac27a0ec DK |
751 | } |
752 | ||
753 | return desc_count; | |
754 | #endif | |
755 | } | |
756 | ||
fd2d4291 | 757 | static inline int test_root(ext4_group_t a, int b) |
ac27a0ec | 758 | { |
f4afb4f4 TT |
759 | while (1) { |
760 | if (a < b) | |
761 | return 0; | |
762 | if (a == b) | |
763 | return 1; | |
764 | if ((a % b) != 0) | |
765 | return 0; | |
766 | a = a / b; | |
767 | } | |
ac27a0ec DK |
768 | } |
769 | ||
ac27a0ec | 770 | /** |
617ba13b | 771 | * ext4_bg_has_super - number of blocks used by the superblock in group |
ac27a0ec DK |
772 | * @sb: superblock for filesystem |
773 | * @group: group number to check | |
774 | * | |
775 | * Return the number of blocks used by the superblock (primary or backup) | |
776 | * in this group. Currently this will be only 0 or 1. | |
777 | */ | |
fd2d4291 | 778 | int ext4_bg_has_super(struct super_block *sb, ext4_group_t group) |
ac27a0ec | 779 | { |
1beeef1b DW |
780 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; |
781 | ||
782 | if (group == 0) | |
783 | return 1; | |
e2b911c5 | 784 | if (ext4_has_feature_sparse_super2(sb)) { |
1beeef1b DW |
785 | if (group == le32_to_cpu(es->s_backup_bgs[0]) || |
786 | group == le32_to_cpu(es->s_backup_bgs[1])) | |
787 | return 1; | |
788 | return 0; | |
789 | } | |
e2b911c5 | 790 | if ((group <= 1) || !ext4_has_feature_sparse_super(sb)) |
1beeef1b DW |
791 | return 1; |
792 | if (!(group & 1)) | |
ac27a0ec | 793 | return 0; |
1beeef1b DW |
794 | if (test_root(group, 3) || (test_root(group, 5)) || |
795 | test_root(group, 7)) | |
796 | return 1; | |
797 | ||
798 | return 0; | |
ac27a0ec DK |
799 | } |
800 | ||
fd2d4291 AM |
801 | static unsigned long ext4_bg_num_gdb_meta(struct super_block *sb, |
802 | ext4_group_t group) | |
ac27a0ec | 803 | { |
617ba13b | 804 | unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb); |
fd2d4291 AM |
805 | ext4_group_t first = metagroup * EXT4_DESC_PER_BLOCK(sb); |
806 | ext4_group_t last = first + EXT4_DESC_PER_BLOCK(sb) - 1; | |
ac27a0ec DK |
807 | |
808 | if (group == first || group == first + 1 || group == last) | |
809 | return 1; | |
810 | return 0; | |
811 | } | |
812 | ||
fd2d4291 AM |
813 | static unsigned long ext4_bg_num_gdb_nometa(struct super_block *sb, |
814 | ext4_group_t group) | |
ac27a0ec | 815 | { |
8dadb198 TT |
816 | if (!ext4_bg_has_super(sb, group)) |
817 | return 0; | |
818 | ||
e2b911c5 | 819 | if (ext4_has_feature_meta_bg(sb)) |
8dadb198 TT |
820 | return le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg); |
821 | else | |
822 | return EXT4_SB(sb)->s_gdb_count; | |
ac27a0ec DK |
823 | } |
824 | ||
825 | /** | |
617ba13b | 826 | * ext4_bg_num_gdb - number of blocks used by the group table in group |
ac27a0ec DK |
827 | * @sb: superblock for filesystem |
828 | * @group: group number to check | |
829 | * | |
830 | * Return the number of blocks used by the group descriptor table | |
831 | * (primary or backup) in this group. In the future there may be a | |
832 | * different number of descriptor blocks in each group. | |
833 | */ | |
fd2d4291 | 834 | unsigned long ext4_bg_num_gdb(struct super_block *sb, ext4_group_t group) |
ac27a0ec DK |
835 | { |
836 | unsigned long first_meta_bg = | |
617ba13b MC |
837 | le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg); |
838 | unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb); | |
ac27a0ec | 839 | |
e2b911c5 | 840 | if (!ext4_has_feature_meta_bg(sb) || metagroup < first_meta_bg) |
af5bc92d | 841 | return ext4_bg_num_gdb_nometa(sb, group); |
ac27a0ec | 842 | |
617ba13b | 843 | return ext4_bg_num_gdb_meta(sb,group); |
ac27a0ec DK |
844 | |
845 | } | |
c2ea3fde | 846 | |
49f7f9af | 847 | /* |
d5b8f310 | 848 | * This function returns the number of file system metadata clusters at |
49f7f9af TT |
849 | * the beginning of a block group, including the reserved gdt blocks. |
850 | */ | |
5f163cc7 | 851 | static unsigned ext4_num_base_meta_clusters(struct super_block *sb, |
d5b8f310 | 852 | ext4_group_t block_group) |
49f7f9af TT |
853 | { |
854 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
d5b8f310 | 855 | unsigned num; |
49f7f9af TT |
856 | |
857 | /* Check for superblock and gdt backups in this group */ | |
858 | num = ext4_bg_has_super(sb, block_group); | |
859 | ||
e2b911c5 | 860 | if (!ext4_has_feature_meta_bg(sb) || |
49f7f9af TT |
861 | block_group < le32_to_cpu(sbi->s_es->s_first_meta_bg) * |
862 | sbi->s_desc_per_block) { | |
863 | if (num) { | |
864 | num += ext4_bg_num_gdb(sb, block_group); | |
865 | num += le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks); | |
866 | } | |
867 | } else { /* For META_BG_BLOCK_GROUPS */ | |
868 | num += ext4_bg_num_gdb(sb, block_group); | |
869 | } | |
d5b8f310 | 870 | return EXT4_NUM_B2C(sbi, num); |
49f7f9af | 871 | } |
f86186b4 ES |
872 | /** |
873 | * ext4_inode_to_goal_block - return a hint for block allocation | |
874 | * @inode: inode for block allocation | |
875 | * | |
876 | * Return the ideal location to start allocating blocks for a | |
877 | * newly created inode. | |
878 | */ | |
879 | ext4_fsblk_t ext4_inode_to_goal_block(struct inode *inode) | |
880 | { | |
881 | struct ext4_inode_info *ei = EXT4_I(inode); | |
882 | ext4_group_t block_group; | |
883 | ext4_grpblk_t colour; | |
884 | int flex_size = ext4_flex_bg_size(EXT4_SB(inode->i_sb)); | |
885 | ext4_fsblk_t bg_start; | |
886 | ext4_fsblk_t last_block; | |
887 | ||
888 | block_group = ei->i_block_group; | |
889 | if (flex_size >= EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME) { | |
890 | /* | |
891 | * If there are at least EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME | |
892 | * block groups per flexgroup, reserve the first block | |
893 | * group for directories and special files. Regular | |
894 | * files will start at the second block group. This | |
895 | * tends to speed up directory access and improves | |
896 | * fsck times. | |
897 | */ | |
898 | block_group &= ~(flex_size-1); | |
899 | if (S_ISREG(inode->i_mode)) | |
900 | block_group++; | |
901 | } | |
902 | bg_start = ext4_group_first_block_no(inode->i_sb, block_group); | |
903 | last_block = ext4_blocks_count(EXT4_SB(inode->i_sb)->s_es) - 1; | |
904 | ||
905 | /* | |
906 | * If we are doing delayed allocation, we don't need take | |
907 | * colour into account. | |
908 | */ | |
909 | if (test_opt(inode->i_sb, DELALLOC)) | |
910 | return bg_start; | |
911 | ||
912 | if (bg_start + EXT4_BLOCKS_PER_GROUP(inode->i_sb) <= last_block) | |
9bee5779 | 913 | colour = (task_pid_nr(current) % 16) * |
f86186b4 ES |
914 | (EXT4_BLOCKS_PER_GROUP(inode->i_sb) / 16); |
915 | else | |
9bee5779 RH |
916 | colour = (task_pid_nr(current) % 16) * |
917 | ((last_block - bg_start) / 16); | |
f86186b4 ES |
918 | return bg_start + colour; |
919 | } | |
920 |