]>
Commit | Line | Data |
---|---|---|
ac27a0ec | 1 | /* |
617ba13b | 2 | * linux/fs/ext4/balloc.c |
ac27a0ec DK |
3 | * |
4 | * Copyright (C) 1992, 1993, 1994, 1995 | |
5 | * Remy Card (card@masi.ibp.fr) | |
6 | * Laboratoire MASI - Institut Blaise Pascal | |
7 | * Universite Pierre et Marie Curie (Paris VI) | |
8 | * | |
9 | * 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 | |
12 | */ | |
13 | ||
14 | #include <linux/time.h> | |
15 | #include <linux/capability.h> | |
16 | #include <linux/fs.h> | |
dab291af | 17 | #include <linux/jbd2.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 | ||
ac27a0ec DK |
26 | /* |
27 | * balloc.c contains the blocks allocation and deallocation routines | |
28 | */ | |
29 | ||
72b64b59 AM |
30 | /* |
31 | * Calculate the block group number and offset, given a block number | |
32 | */ | |
33 | void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr, | |
fd2d4291 | 34 | ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp) |
72b64b59 | 35 | { |
8c55e204 | 36 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; |
72b64b59 AM |
37 | ext4_grpblk_t offset; |
38 | ||
8c55e204 | 39 | blocknr = blocknr - le32_to_cpu(es->s_first_data_block); |
f4e5bc24 | 40 | offset = do_div(blocknr, EXT4_BLOCKS_PER_GROUP(sb)); |
72b64b59 AM |
41 | if (offsetp) |
42 | *offsetp = offset; | |
43 | if (blockgrpp) | |
8c55e204 | 44 | *blockgrpp = blocknr; |
72b64b59 AM |
45 | |
46 | } | |
47 | ||
0bf7e837 JS |
48 | static int ext4_block_in_group(struct super_block *sb, ext4_fsblk_t block, |
49 | ext4_group_t block_group) | |
50 | { | |
51 | ext4_group_t actual_group; | |
7477827f | 52 | ext4_get_group_no_and_offset(sb, block, &actual_group, NULL); |
0bf7e837 JS |
53 | if (actual_group == block_group) |
54 | return 1; | |
55 | return 0; | |
56 | } | |
57 | ||
58 | static int ext4_group_used_meta_blocks(struct super_block *sb, | |
e187c658 TT |
59 | ext4_group_t block_group, |
60 | struct ext4_group_desc *gdp) | |
0bf7e837 JS |
61 | { |
62 | ext4_fsblk_t tmp; | |
63 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
64 | /* block bitmap, inode bitmap, and inode table blocks */ | |
65 | int used_blocks = sbi->s_itb_per_group + 2; | |
66 | ||
67 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) { | |
0bf7e837 JS |
68 | if (!ext4_block_in_group(sb, ext4_block_bitmap(sb, gdp), |
69 | block_group)) | |
70 | used_blocks--; | |
71 | ||
72 | if (!ext4_block_in_group(sb, ext4_inode_bitmap(sb, gdp), | |
73 | block_group)) | |
74 | used_blocks--; | |
75 | ||
76 | tmp = ext4_inode_table(sb, gdp); | |
77 | for (; tmp < ext4_inode_table(sb, gdp) + | |
78 | sbi->s_itb_per_group; tmp++) { | |
79 | if (!ext4_block_in_group(sb, tmp, block_group)) | |
80 | used_blocks -= 1; | |
81 | } | |
82 | } | |
83 | return used_blocks; | |
84 | } | |
c2ea3fde | 85 | |
717d50e4 AD |
86 | /* Initializes an uninitialized block bitmap if given, and returns the |
87 | * number of blocks free in the group. */ | |
88 | unsigned ext4_init_block_bitmap(struct super_block *sb, struct buffer_head *bh, | |
fd2d4291 | 89 | ext4_group_t block_group, struct ext4_group_desc *gdp) |
717d50e4 | 90 | { |
717d50e4 | 91 | int bit, bit_max; |
8df9675f | 92 | ext4_group_t ngroups = ext4_get_groups_count(sb); |
717d50e4 AD |
93 | unsigned free_blocks, group_blocks; |
94 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
95 | ||
96 | if (bh) { | |
97 | J_ASSERT_BH(bh, buffer_locked(bh)); | |
98 | ||
99 | /* If checksum is bad mark all blocks used to prevent allocation | |
100 | * essentially implementing a per-group read-only flag. */ | |
101 | if (!ext4_group_desc_csum_verify(sbi, block_group, gdp)) { | |
12062ddd ES |
102 | ext4_error(sb, "Checksum bad for group %u", |
103 | block_group); | |
560671a0 AK |
104 | ext4_free_blks_set(sb, gdp, 0); |
105 | ext4_free_inodes_set(sb, gdp, 0); | |
106 | ext4_itable_unused_set(sb, gdp, 0); | |
717d50e4 AD |
107 | memset(bh->b_data, 0xff, sb->s_blocksize); |
108 | return 0; | |
109 | } | |
110 | memset(bh->b_data, 0, sb->s_blocksize); | |
111 | } | |
112 | ||
113 | /* Check for superblock and gdt backups in this group */ | |
114 | bit_max = ext4_bg_has_super(sb, block_group); | |
115 | ||
116 | if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) || | |
117 | block_group < le32_to_cpu(sbi->s_es->s_first_meta_bg) * | |
118 | sbi->s_desc_per_block) { | |
119 | if (bit_max) { | |
120 | bit_max += ext4_bg_num_gdb(sb, block_group); | |
121 | bit_max += | |
122 | le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks); | |
123 | } | |
124 | } else { /* For META_BG_BLOCK_GROUPS */ | |
6afd6707 | 125 | bit_max += ext4_bg_num_gdb(sb, block_group); |
717d50e4 AD |
126 | } |
127 | ||
8df9675f | 128 | if (block_group == ngroups - 1) { |
717d50e4 AD |
129 | /* |
130 | * Even though mke2fs always initialize first and last group | |
131 | * if some other tool enabled the EXT4_BG_BLOCK_UNINIT we need | |
132 | * to make sure we calculate the right free blocks | |
133 | */ | |
134 | group_blocks = ext4_blocks_count(sbi->s_es) - | |
5661bd68 | 135 | ext4_group_first_block_no(sb, ngroups - 1); |
717d50e4 AD |
136 | } else { |
137 | group_blocks = EXT4_BLOCKS_PER_GROUP(sb); | |
138 | } | |
139 | ||
140 | free_blocks = group_blocks - bit_max; | |
141 | ||
142 | if (bh) { | |
0bf7e837 JS |
143 | ext4_fsblk_t start, tmp; |
144 | int flex_bg = 0; | |
d00a6d7b | 145 | |
717d50e4 AD |
146 | for (bit = 0; bit < bit_max; bit++) |
147 | ext4_set_bit(bit, bh->b_data); | |
148 | ||
d00a6d7b | 149 | start = ext4_group_first_block_no(sb, block_group); |
717d50e4 | 150 | |
0bf7e837 JS |
151 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, |
152 | EXT4_FEATURE_INCOMPAT_FLEX_BG)) | |
153 | flex_bg = 1; | |
717d50e4 | 154 | |
0bf7e837 JS |
155 | /* Set bits for block and inode bitmaps, and inode table */ |
156 | tmp = ext4_block_bitmap(sb, gdp); | |
157 | if (!flex_bg || ext4_block_in_group(sb, tmp, block_group)) | |
158 | ext4_set_bit(tmp - start, bh->b_data); | |
159 | ||
160 | tmp = ext4_inode_bitmap(sb, gdp); | |
161 | if (!flex_bg || ext4_block_in_group(sb, tmp, block_group)) | |
162 | ext4_set_bit(tmp - start, bh->b_data); | |
163 | ||
164 | tmp = ext4_inode_table(sb, gdp); | |
165 | for (; tmp < ext4_inode_table(sb, gdp) + | |
166 | sbi->s_itb_per_group; tmp++) { | |
167 | if (!flex_bg || | |
168 | ext4_block_in_group(sb, tmp, block_group)) | |
169 | ext4_set_bit(tmp - start, bh->b_data); | |
170 | } | |
717d50e4 AD |
171 | /* |
172 | * Also if the number of blocks within the group is | |
173 | * less than the blocksize * 8 ( which is the size | |
174 | * of bitmap ), set rest of the block bitmap to 1 | |
175 | */ | |
61d08673 TT |
176 | ext4_mark_bitmap_end(group_blocks, sb->s_blocksize * 8, |
177 | bh->b_data); | |
717d50e4 | 178 | } |
e187c658 | 179 | return free_blocks - ext4_group_used_meta_blocks(sb, block_group, gdp); |
717d50e4 AD |
180 | } |
181 | ||
182 | ||
ac27a0ec DK |
183 | /* |
184 | * The free blocks are managed by bitmaps. A file system contains several | |
185 | * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap | |
186 | * block for inodes, N blocks for the inode table and data blocks. | |
187 | * | |
188 | * The file system contains group descriptors which are located after the | |
189 | * super block. Each descriptor contains the number of the bitmap block and | |
190 | * the free blocks count in the block. The descriptors are loaded in memory | |
e627432c | 191 | * when a file system is mounted (see ext4_fill_super). |
ac27a0ec DK |
192 | */ |
193 | ||
ac27a0ec | 194 | /** |
617ba13b | 195 | * ext4_get_group_desc() -- load group descriptor from disk |
ac27a0ec DK |
196 | * @sb: super block |
197 | * @block_group: given block group | |
198 | * @bh: pointer to the buffer head to store the block | |
199 | * group descriptor | |
200 | */ | |
af5bc92d | 201 | struct ext4_group_desc * ext4_get_group_desc(struct super_block *sb, |
fd2d4291 | 202 | ext4_group_t block_group, |
af5bc92d | 203 | struct buffer_head **bh) |
ac27a0ec | 204 | { |
498e5f24 TT |
205 | unsigned int group_desc; |
206 | unsigned int offset; | |
8df9675f | 207 | ext4_group_t ngroups = ext4_get_groups_count(sb); |
af5bc92d | 208 | struct ext4_group_desc *desc; |
617ba13b | 209 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec | 210 | |
8df9675f | 211 | if (block_group >= ngroups) { |
12062ddd ES |
212 | ext4_error(sb, "block_group >= groups_count - block_group = %u," |
213 | " groups_count = %u", block_group, ngroups); | |
ac27a0ec DK |
214 | |
215 | return NULL; | |
216 | } | |
ac27a0ec | 217 | |
617ba13b MC |
218 | group_desc = block_group >> EXT4_DESC_PER_BLOCK_BITS(sb); |
219 | offset = block_group & (EXT4_DESC_PER_BLOCK(sb) - 1); | |
ac27a0ec | 220 | if (!sbi->s_group_desc[group_desc]) { |
12062ddd | 221 | ext4_error(sb, "Group descriptor not loaded - " |
498e5f24 | 222 | "block_group = %u, group_desc = %u, desc = %u", |
af5bc92d | 223 | block_group, group_desc, offset); |
ac27a0ec DK |
224 | return NULL; |
225 | } | |
226 | ||
0d1ee42f AR |
227 | desc = (struct ext4_group_desc *)( |
228 | (__u8 *)sbi->s_group_desc[group_desc]->b_data + | |
229 | offset * EXT4_DESC_SIZE(sb)); | |
ac27a0ec DK |
230 | if (bh) |
231 | *bh = sbi->s_group_desc[group_desc]; | |
0d1ee42f | 232 | return desc; |
ac27a0ec DK |
233 | } |
234 | ||
abcb2947 AK |
235 | static int ext4_valid_block_bitmap(struct super_block *sb, |
236 | struct ext4_group_desc *desc, | |
237 | unsigned int block_group, | |
238 | struct buffer_head *bh) | |
239 | { | |
240 | ext4_grpblk_t offset; | |
241 | ext4_grpblk_t next_zero_bit; | |
242 | ext4_fsblk_t bitmap_blk; | |
243 | ext4_fsblk_t group_first_block; | |
244 | ||
245 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) { | |
246 | /* with FLEX_BG, the inode/block bitmaps and itable | |
247 | * blocks may not be in the group at all | |
248 | * so the bitmap validation will be skipped for those groups | |
249 | * or it has to also read the block group where the bitmaps | |
250 | * are located to verify they are set. | |
251 | */ | |
252 | return 1; | |
253 | } | |
254 | group_first_block = ext4_group_first_block_no(sb, block_group); | |
255 | ||
256 | /* check whether block bitmap block number is set */ | |
257 | bitmap_blk = ext4_block_bitmap(sb, desc); | |
258 | offset = bitmap_blk - group_first_block; | |
259 | if (!ext4_test_bit(offset, bh->b_data)) | |
260 | /* bad block bitmap */ | |
261 | goto err_out; | |
262 | ||
263 | /* check whether the inode bitmap block number is set */ | |
264 | bitmap_blk = ext4_inode_bitmap(sb, desc); | |
265 | offset = bitmap_blk - group_first_block; | |
266 | if (!ext4_test_bit(offset, bh->b_data)) | |
267 | /* bad block bitmap */ | |
268 | goto err_out; | |
269 | ||
270 | /* check whether the inode table block number is set */ | |
271 | bitmap_blk = ext4_inode_table(sb, desc); | |
272 | offset = bitmap_blk - group_first_block; | |
273 | next_zero_bit = ext4_find_next_zero_bit(bh->b_data, | |
274 | offset + EXT4_SB(sb)->s_itb_per_group, | |
275 | offset); | |
276 | if (next_zero_bit >= offset + EXT4_SB(sb)->s_itb_per_group) | |
277 | /* good bitmap for inode tables */ | |
278 | return 1; | |
279 | ||
280 | err_out: | |
12062ddd | 281 | ext4_error(sb, "Invalid block bitmap - block_group = %d, block = %llu", |
abcb2947 AK |
282 | block_group, bitmap_blk); |
283 | return 0; | |
284 | } | |
ac27a0ec | 285 | /** |
574ca174 | 286 | * ext4_read_block_bitmap() |
ac27a0ec DK |
287 | * @sb: super block |
288 | * @block_group: given block group | |
289 | * | |
abcb2947 AK |
290 | * Read the bitmap for a given block_group,and validate the |
291 | * bits for block/inode/inode tables are set in the bitmaps | |
ac27a0ec DK |
292 | * |
293 | * Return buffer_head on success or NULL in case of failure. | |
294 | */ | |
717d50e4 | 295 | struct buffer_head * |
574ca174 | 296 | ext4_read_block_bitmap(struct super_block *sb, ext4_group_t block_group) |
ac27a0ec | 297 | { |
af5bc92d TT |
298 | struct ext4_group_desc *desc; |
299 | struct buffer_head *bh = NULL; | |
7c9e69fa | 300 | ext4_fsblk_t bitmap_blk; |
ac27a0ec | 301 | |
717d50e4 | 302 | desc = ext4_get_group_desc(sb, block_group, NULL); |
ac27a0ec | 303 | if (!desc) |
7c9e69fa AK |
304 | return NULL; |
305 | bitmap_blk = ext4_block_bitmap(sb, desc); | |
abcb2947 AK |
306 | bh = sb_getblk(sb, bitmap_blk); |
307 | if (unlikely(!bh)) { | |
12062ddd | 308 | ext4_error(sb, "Cannot read block bitmap - " |
a9df9a49 | 309 | "block_group = %u, block_bitmap = %llu", |
e29d1cde | 310 | block_group, bitmap_blk); |
abcb2947 AK |
311 | return NULL; |
312 | } | |
2ccb5fb9 AK |
313 | |
314 | if (bitmap_uptodate(bh)) | |
abcb2947 AK |
315 | return bh; |
316 | ||
c806e68f | 317 | lock_buffer(bh); |
2ccb5fb9 AK |
318 | if (bitmap_uptodate(bh)) { |
319 | unlock_buffer(bh); | |
320 | return bh; | |
321 | } | |
955ce5f5 | 322 | ext4_lock_group(sb, block_group); |
717d50e4 | 323 | if (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) { |
abcb2947 | 324 | ext4_init_block_bitmap(sb, bh, block_group, desc); |
2ccb5fb9 | 325 | set_bitmap_uptodate(bh); |
abcb2947 | 326 | set_buffer_uptodate(bh); |
955ce5f5 | 327 | ext4_unlock_group(sb, block_group); |
3300beda | 328 | unlock_buffer(bh); |
abcb2947 | 329 | return bh; |
717d50e4 | 330 | } |
955ce5f5 | 331 | ext4_unlock_group(sb, block_group); |
2ccb5fb9 AK |
332 | if (buffer_uptodate(bh)) { |
333 | /* | |
334 | * if not uninit if bh is uptodate, | |
335 | * bitmap is also uptodate | |
336 | */ | |
337 | set_bitmap_uptodate(bh); | |
338 | unlock_buffer(bh); | |
339 | return bh; | |
340 | } | |
341 | /* | |
342 | * submit the buffer_head for read. We can | |
343 | * safely mark the bitmap as uptodate now. | |
344 | * We do it here so the bitmap uptodate bit | |
345 | * get set with buffer lock held. | |
346 | */ | |
0562e0ba | 347 | trace_ext4_read_block_bitmap_load(sb, block_group); |
2ccb5fb9 | 348 | set_bitmap_uptodate(bh); |
abcb2947 AK |
349 | if (bh_submit_read(bh) < 0) { |
350 | put_bh(bh); | |
12062ddd | 351 | ext4_error(sb, "Cannot read block bitmap - " |
a9df9a49 | 352 | "block_group = %u, block_bitmap = %llu", |
e29d1cde | 353 | block_group, bitmap_blk); |
abcb2947 AK |
354 | return NULL; |
355 | } | |
519deca0 AK |
356 | ext4_valid_block_bitmap(sb, desc, block_group, bh); |
357 | /* | |
358 | * file system mounted not to panic on error, | |
359 | * continue with corrupt bitmap | |
360 | */ | |
ac27a0ec DK |
361 | return bh; |
362 | } | |
ac27a0ec | 363 | |
8c3bf8a0 ES |
364 | /** |
365 | * ext4_has_free_blocks() | |
366 | * @sbi: in-core super block structure. | |
367 | * @nblocks: number of needed blocks | |
368 | * | |
369 | * Check if filesystem has nblocks free & available for allocation. | |
370 | * On success return 1, return 0 on failure. | |
371 | */ | |
55f020db AH |
372 | static int ext4_has_free_blocks(struct ext4_sb_info *sbi, |
373 | s64 nblocks, unsigned int flags) | |
a30d542a | 374 | { |
a996031c | 375 | s64 free_blocks, dirty_blocks, root_blocks; |
a30d542a | 376 | struct percpu_counter *fbc = &sbi->s_freeblocks_counter; |
6bc6e63f | 377 | struct percpu_counter *dbc = &sbi->s_dirtyblocks_counter; |
a30d542a | 378 | |
6bc6e63f AK |
379 | free_blocks = percpu_counter_read_positive(fbc); |
380 | dirty_blocks = percpu_counter_read_positive(dbc); | |
a996031c | 381 | root_blocks = ext4_r_blocks_count(sbi->s_es); |
a30d542a | 382 | |
6bc6e63f AK |
383 | if (free_blocks - (nblocks + root_blocks + dirty_blocks) < |
384 | EXT4_FREEBLOCKS_WATERMARK) { | |
02d21168 AM |
385 | free_blocks = percpu_counter_sum_positive(fbc); |
386 | dirty_blocks = percpu_counter_sum_positive(dbc); | |
6bc6e63f AK |
387 | } |
388 | /* Check whether we have space after | |
a996031c | 389 | * accounting for current dirty blocks & root reserved blocks. |
6bc6e63f | 390 | */ |
a996031c ES |
391 | if (free_blocks >= ((root_blocks + nblocks) + dirty_blocks)) |
392 | return 1; | |
a30d542a | 393 | |
a996031c | 394 | /* Hm, nope. Are (enough) root reserved blocks available? */ |
4c9c544e | 395 | if (sbi->s_resuid == current_fsuid() || |
a996031c | 396 | ((sbi->s_resgid != 0) && in_group_p(sbi->s_resgid)) || |
55f020db AH |
397 | capable(CAP_SYS_RESOURCE) || |
398 | (flags & EXT4_MB_USE_ROOT_BLOCKS)) { | |
399 | ||
a996031c ES |
400 | if (free_blocks >= (nblocks + dirty_blocks)) |
401 | return 1; | |
402 | } | |
403 | ||
404 | return 0; | |
a30d542a AK |
405 | } |
406 | ||
8c3bf8a0 | 407 | int ext4_claim_free_blocks(struct ext4_sb_info *sbi, |
55f020db | 408 | s64 nblocks, unsigned int flags) |
ac27a0ec | 409 | { |
55f020db | 410 | if (ext4_has_free_blocks(sbi, nblocks, flags)) { |
8c3bf8a0 | 411 | percpu_counter_add(&sbi->s_dirtyblocks_counter, nblocks); |
16eb7295 | 412 | return 0; |
8c3bf8a0 ES |
413 | } else |
414 | return -ENOSPC; | |
a30d542a | 415 | } |
07031431 | 416 | |
ac27a0ec | 417 | /** |
617ba13b | 418 | * ext4_should_retry_alloc() |
ac27a0ec DK |
419 | * @sb: super block |
420 | * @retries number of attemps has been made | |
421 | * | |
617ba13b | 422 | * ext4_should_retry_alloc() is called when ENOSPC is returned, and if |
ac27a0ec | 423 | * it is profitable to retry the operation, this function will wait |
25985edc | 424 | * for the current or committing transaction to complete, and then |
ac27a0ec DK |
425 | * return TRUE. |
426 | * | |
427 | * if the total number of retries exceed three times, return FALSE. | |
428 | */ | |
617ba13b | 429 | int ext4_should_retry_alloc(struct super_block *sb, int *retries) |
ac27a0ec | 430 | { |
55f020db | 431 | if (!ext4_has_free_blocks(EXT4_SB(sb), 1, 0) || |
8f64b32e ES |
432 | (*retries)++ > 3 || |
433 | !EXT4_SB(sb)->s_journal) | |
ac27a0ec DK |
434 | return 0; |
435 | ||
436 | jbd_debug(1, "%s: retrying operation after ENOSPC\n", sb->s_id); | |
437 | ||
dab291af | 438 | return jbd2_journal_force_commit_nested(EXT4_SB(sb)->s_journal); |
ac27a0ec DK |
439 | } |
440 | ||
654b4908 | 441 | /* |
d2a17637 | 442 | * ext4_new_meta_blocks() -- allocate block for meta data (indexing) blocks |
654b4908 AK |
443 | * |
444 | * @handle: handle to this transaction | |
445 | * @inode: file inode | |
446 | * @goal: given target block(filesystem wide) | |
97df5d15 | 447 | * @count: pointer to total number of blocks needed |
654b4908 AK |
448 | * @errp: error code |
449 | * | |
97df5d15 | 450 | * Return 1st allocated block number on success, *count stores total account |
d2a17637 | 451 | * error stores in errp pointer |
654b4908 | 452 | */ |
d2a17637 | 453 | ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode, |
55f020db AH |
454 | ext4_fsblk_t goal, unsigned int flags, |
455 | unsigned long *count, int *errp) | |
654b4908 | 456 | { |
97df5d15 | 457 | struct ext4_allocation_request ar; |
d2a17637 | 458 | ext4_fsblk_t ret; |
97df5d15 TT |
459 | |
460 | memset(&ar, 0, sizeof(ar)); | |
461 | /* Fill with neighbour allocated blocks */ | |
462 | ar.inode = inode; | |
463 | ar.goal = goal; | |
464 | ar.len = count ? *count : 1; | |
55f020db | 465 | ar.flags = flags; |
97df5d15 TT |
466 | |
467 | ret = ext4_mb_new_blocks(handle, &ar, errp); | |
468 | if (count) | |
469 | *count = ar.len; | |
d2a17637 | 470 | /* |
72b8ab9d ES |
471 | * Account for the allocated meta blocks. We will never |
472 | * fail EDQUOT for metdata, but we do account for it. | |
d2a17637 | 473 | */ |
f2321097 TT |
474 | if (!(*errp) && |
475 | ext4_test_inode_state(inode, EXT4_STATE_DELALLOC_RESERVED)) { | |
d2a17637 | 476 | spin_lock(&EXT4_I(inode)->i_block_reservation_lock); |
97df5d15 | 477 | EXT4_I(inode)->i_allocated_meta_blocks += ar.len; |
d2a17637 | 478 | spin_unlock(&EXT4_I(inode)->i_block_reservation_lock); |
72b8ab9d | 479 | dquot_alloc_block_nofail(inode, ar.len); |
d2a17637 MC |
480 | } |
481 | return ret; | |
654b4908 AK |
482 | } |
483 | ||
ac27a0ec | 484 | /** |
617ba13b | 485 | * ext4_count_free_blocks() -- count filesystem free blocks |
ac27a0ec DK |
486 | * @sb: superblock |
487 | * | |
488 | * Adds up the number of free blocks from each block group. | |
489 | */ | |
617ba13b | 490 | ext4_fsblk_t ext4_count_free_blocks(struct super_block *sb) |
ac27a0ec | 491 | { |
617ba13b MC |
492 | ext4_fsblk_t desc_count; |
493 | struct ext4_group_desc *gdp; | |
fd2d4291 | 494 | ext4_group_t i; |
8df9675f | 495 | ext4_group_t ngroups = ext4_get_groups_count(sb); |
617ba13b MC |
496 | #ifdef EXT4FS_DEBUG |
497 | struct ext4_super_block *es; | |
498 | ext4_fsblk_t bitmap_count; | |
498e5f24 | 499 | unsigned int x; |
ac27a0ec DK |
500 | struct buffer_head *bitmap_bh = NULL; |
501 | ||
617ba13b | 502 | es = EXT4_SB(sb)->s_es; |
ac27a0ec DK |
503 | desc_count = 0; |
504 | bitmap_count = 0; | |
505 | gdp = NULL; | |
506 | ||
ac27a0ec | 507 | for (i = 0; i < ngroups; i++) { |
617ba13b | 508 | gdp = ext4_get_group_desc(sb, i, NULL); |
ac27a0ec DK |
509 | if (!gdp) |
510 | continue; | |
9fd9784c | 511 | desc_count += ext4_free_blks_count(sb, gdp); |
ac27a0ec | 512 | brelse(bitmap_bh); |
574ca174 | 513 | bitmap_bh = ext4_read_block_bitmap(sb, i); |
ac27a0ec DK |
514 | if (bitmap_bh == NULL) |
515 | continue; | |
516 | ||
617ba13b | 517 | x = ext4_count_free(bitmap_bh, sb->s_blocksize); |
9fd9784c TLSC |
518 | printk(KERN_DEBUG "group %u: stored = %d, counted = %u\n", |
519 | i, ext4_free_blks_count(sb, gdp), x); | |
ac27a0ec DK |
520 | bitmap_count += x; |
521 | } | |
522 | brelse(bitmap_bh); | |
4776004f TT |
523 | printk(KERN_DEBUG "ext4_count_free_blocks: stored = %llu" |
524 | ", computed = %llu, %llu\n", ext4_free_blocks_count(es), | |
525 | desc_count, bitmap_count); | |
ac27a0ec DK |
526 | return bitmap_count; |
527 | #else | |
528 | desc_count = 0; | |
ac27a0ec | 529 | for (i = 0; i < ngroups; i++) { |
617ba13b | 530 | gdp = ext4_get_group_desc(sb, i, NULL); |
ac27a0ec DK |
531 | if (!gdp) |
532 | continue; | |
560671a0 | 533 | desc_count += ext4_free_blks_count(sb, gdp); |
ac27a0ec DK |
534 | } |
535 | ||
536 | return desc_count; | |
537 | #endif | |
538 | } | |
539 | ||
fd2d4291 | 540 | static inline int test_root(ext4_group_t a, int b) |
ac27a0ec DK |
541 | { |
542 | int num = b; | |
543 | ||
544 | while (a > num) | |
545 | num *= b; | |
546 | return num == a; | |
547 | } | |
548 | ||
fd2d4291 | 549 | static int ext4_group_sparse(ext4_group_t group) |
ac27a0ec DK |
550 | { |
551 | if (group <= 1) | |
552 | return 1; | |
553 | if (!(group & 1)) | |
554 | return 0; | |
555 | return (test_root(group, 7) || test_root(group, 5) || | |
556 | test_root(group, 3)); | |
557 | } | |
558 | ||
559 | /** | |
617ba13b | 560 | * ext4_bg_has_super - number of blocks used by the superblock in group |
ac27a0ec DK |
561 | * @sb: superblock for filesystem |
562 | * @group: group number to check | |
563 | * | |
564 | * Return the number of blocks used by the superblock (primary or backup) | |
565 | * in this group. Currently this will be only 0 or 1. | |
566 | */ | |
fd2d4291 | 567 | int ext4_bg_has_super(struct super_block *sb, ext4_group_t group) |
ac27a0ec | 568 | { |
617ba13b MC |
569 | if (EXT4_HAS_RO_COMPAT_FEATURE(sb, |
570 | EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER) && | |
571 | !ext4_group_sparse(group)) | |
ac27a0ec DK |
572 | return 0; |
573 | return 1; | |
574 | } | |
575 | ||
fd2d4291 AM |
576 | static unsigned long ext4_bg_num_gdb_meta(struct super_block *sb, |
577 | ext4_group_t group) | |
ac27a0ec | 578 | { |
617ba13b | 579 | unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb); |
fd2d4291 AM |
580 | ext4_group_t first = metagroup * EXT4_DESC_PER_BLOCK(sb); |
581 | ext4_group_t last = first + EXT4_DESC_PER_BLOCK(sb) - 1; | |
ac27a0ec DK |
582 | |
583 | if (group == first || group == first + 1 || group == last) | |
584 | return 1; | |
585 | return 0; | |
586 | } | |
587 | ||
fd2d4291 AM |
588 | static unsigned long ext4_bg_num_gdb_nometa(struct super_block *sb, |
589 | ext4_group_t group) | |
ac27a0ec | 590 | { |
8dadb198 TT |
591 | if (!ext4_bg_has_super(sb, group)) |
592 | return 0; | |
593 | ||
594 | if (EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG)) | |
595 | return le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg); | |
596 | else | |
597 | return EXT4_SB(sb)->s_gdb_count; | |
ac27a0ec DK |
598 | } |
599 | ||
600 | /** | |
617ba13b | 601 | * ext4_bg_num_gdb - number of blocks used by the group table in group |
ac27a0ec DK |
602 | * @sb: superblock for filesystem |
603 | * @group: group number to check | |
604 | * | |
605 | * Return the number of blocks used by the group descriptor table | |
606 | * (primary or backup) in this group. In the future there may be a | |
607 | * different number of descriptor blocks in each group. | |
608 | */ | |
fd2d4291 | 609 | unsigned long ext4_bg_num_gdb(struct super_block *sb, ext4_group_t group) |
ac27a0ec DK |
610 | { |
611 | unsigned long first_meta_bg = | |
617ba13b MC |
612 | le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg); |
613 | unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb); | |
ac27a0ec | 614 | |
617ba13b | 615 | if (!EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG) || |
ac27a0ec | 616 | metagroup < first_meta_bg) |
af5bc92d | 617 | return ext4_bg_num_gdb_nometa(sb, group); |
ac27a0ec | 618 | |
617ba13b | 619 | return ext4_bg_num_gdb_meta(sb,group); |
ac27a0ec DK |
620 | |
621 | } | |
c2ea3fde | 622 | |
f86186b4 ES |
623 | /** |
624 | * ext4_inode_to_goal_block - return a hint for block allocation | |
625 | * @inode: inode for block allocation | |
626 | * | |
627 | * Return the ideal location to start allocating blocks for a | |
628 | * newly created inode. | |
629 | */ | |
630 | ext4_fsblk_t ext4_inode_to_goal_block(struct inode *inode) | |
631 | { | |
632 | struct ext4_inode_info *ei = EXT4_I(inode); | |
633 | ext4_group_t block_group; | |
634 | ext4_grpblk_t colour; | |
635 | int flex_size = ext4_flex_bg_size(EXT4_SB(inode->i_sb)); | |
636 | ext4_fsblk_t bg_start; | |
637 | ext4_fsblk_t last_block; | |
638 | ||
639 | block_group = ei->i_block_group; | |
640 | if (flex_size >= EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME) { | |
641 | /* | |
642 | * If there are at least EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME | |
643 | * block groups per flexgroup, reserve the first block | |
644 | * group for directories and special files. Regular | |
645 | * files will start at the second block group. This | |
646 | * tends to speed up directory access and improves | |
647 | * fsck times. | |
648 | */ | |
649 | block_group &= ~(flex_size-1); | |
650 | if (S_ISREG(inode->i_mode)) | |
651 | block_group++; | |
652 | } | |
653 | bg_start = ext4_group_first_block_no(inode->i_sb, block_group); | |
654 | last_block = ext4_blocks_count(EXT4_SB(inode->i_sb)->s_es) - 1; | |
655 | ||
656 | /* | |
657 | * If we are doing delayed allocation, we don't need take | |
658 | * colour into account. | |
659 | */ | |
660 | if (test_opt(inode->i_sb, DELALLOC)) | |
661 | return bg_start; | |
662 | ||
663 | if (bg_start + EXT4_BLOCKS_PER_GROUP(inode->i_sb) <= last_block) | |
664 | colour = (current->pid % 16) * | |
665 | (EXT4_BLOCKS_PER_GROUP(inode->i_sb) / 16); | |
666 | else | |
667 | colour = (current->pid % 16) * ((last_block - bg_start) / 16); | |
668 | return bg_start + colour; | |
669 | } | |
670 |