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 | ||
49f7f9af TT |
26 | static unsigned int num_base_meta_blocks(struct super_block *sb, |
27 | ext4_group_t block_group); | |
28 | ||
ac27a0ec DK |
29 | /* |
30 | * balloc.c contains the blocks allocation and deallocation routines | |
31 | */ | |
32 | ||
72b64b59 AM |
33 | /* |
34 | * Calculate the block group number and offset, given a block number | |
35 | */ | |
36 | void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr, | |
fd2d4291 | 37 | ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp) |
72b64b59 | 38 | { |
8c55e204 | 39 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; |
72b64b59 AM |
40 | ext4_grpblk_t offset; |
41 | ||
8c55e204 | 42 | blocknr = blocknr - le32_to_cpu(es->s_first_data_block); |
f4e5bc24 | 43 | offset = do_div(blocknr, EXT4_BLOCKS_PER_GROUP(sb)); |
72b64b59 AM |
44 | if (offsetp) |
45 | *offsetp = offset; | |
46 | if (blockgrpp) | |
8c55e204 | 47 | *blockgrpp = blocknr; |
72b64b59 AM |
48 | |
49 | } | |
50 | ||
0bf7e837 JS |
51 | static int ext4_block_in_group(struct super_block *sb, ext4_fsblk_t block, |
52 | ext4_group_t block_group) | |
53 | { | |
54 | ext4_group_t actual_group; | |
7477827f | 55 | ext4_get_group_no_and_offset(sb, block, &actual_group, NULL); |
0bf7e837 JS |
56 | if (actual_group == block_group) |
57 | return 1; | |
58 | return 0; | |
59 | } | |
60 | ||
61 | static int ext4_group_used_meta_blocks(struct super_block *sb, | |
e187c658 TT |
62 | ext4_group_t block_group, |
63 | struct ext4_group_desc *gdp) | |
0bf7e837 JS |
64 | { |
65 | ext4_fsblk_t tmp; | |
66 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
67 | /* block bitmap, inode bitmap, and inode table blocks */ | |
68 | int used_blocks = sbi->s_itb_per_group + 2; | |
69 | ||
70 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) { | |
0bf7e837 JS |
71 | if (!ext4_block_in_group(sb, ext4_block_bitmap(sb, gdp), |
72 | block_group)) | |
73 | used_blocks--; | |
74 | ||
75 | if (!ext4_block_in_group(sb, ext4_inode_bitmap(sb, gdp), | |
76 | block_group)) | |
77 | used_blocks--; | |
78 | ||
79 | tmp = ext4_inode_table(sb, gdp); | |
80 | for (; tmp < ext4_inode_table(sb, gdp) + | |
81 | sbi->s_itb_per_group; tmp++) { | |
82 | if (!ext4_block_in_group(sb, tmp, block_group)) | |
83 | used_blocks -= 1; | |
84 | } | |
85 | } | |
86 | return used_blocks; | |
87 | } | |
c2ea3fde | 88 | |
49f7f9af TT |
89 | static unsigned int num_blocks_in_group(struct super_block *sb, |
90 | ext4_group_t block_group) | |
91 | { | |
92 | if (block_group == ext4_get_groups_count(sb) - 1) { | |
93 | /* | |
94 | * Even though mke2fs always initializes the first and | |
95 | * last group, just in case some other tool was used, | |
96 | * we need to make sure we calculate the right free | |
97 | * blocks. | |
98 | */ | |
99 | return ext4_blocks_count(EXT4_SB(sb)->s_es) - | |
100 | ext4_group_first_block_no(sb, block_group); | |
101 | } else | |
102 | return EXT4_BLOCKS_PER_GROUP(sb); | |
103 | } | |
104 | ||
fd034a84 TT |
105 | /* Initializes an uninitialized block bitmap */ |
106 | void ext4_init_block_bitmap(struct super_block *sb, struct buffer_head *bh, | |
107 | ext4_group_t block_group, | |
108 | struct ext4_group_desc *gdp) | |
717d50e4 | 109 | { |
49f7f9af | 110 | unsigned int bit, bit_max = num_base_meta_blocks(sb, block_group); |
717d50e4 | 111 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
fd034a84 TT |
112 | ext4_fsblk_t start, tmp; |
113 | int flex_bg = 0; | |
114 | ||
115 | J_ASSERT_BH(bh, buffer_locked(bh)); | |
116 | ||
117 | /* If checksum is bad mark all blocks used to prevent allocation | |
118 | * essentially implementing a per-group read-only flag. */ | |
119 | if (!ext4_group_desc_csum_verify(sbi, block_group, gdp)) { | |
120 | ext4_error(sb, "Checksum bad for group %u", block_group); | |
121 | ext4_free_blks_set(sb, gdp, 0); | |
122 | ext4_free_inodes_set(sb, gdp, 0); | |
123 | ext4_itable_unused_set(sb, gdp, 0); | |
124 | memset(bh->b_data, 0xff, sb->s_blocksize); | |
125 | return; | |
717d50e4 | 126 | } |
fd034a84 | 127 | memset(bh->b_data, 0, sb->s_blocksize); |
717d50e4 | 128 | |
fd034a84 TT |
129 | for (bit = 0; bit < bit_max; bit++) |
130 | ext4_set_bit(bit, bh->b_data); | |
d00a6d7b | 131 | |
fd034a84 | 132 | start = ext4_group_first_block_no(sb, block_group); |
717d50e4 | 133 | |
fd034a84 TT |
134 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) |
135 | flex_bg = 1; | |
717d50e4 | 136 | |
fd034a84 TT |
137 | /* Set bits for block and inode bitmaps, and inode table */ |
138 | tmp = ext4_block_bitmap(sb, gdp); | |
139 | if (!flex_bg || ext4_block_in_group(sb, tmp, block_group)) | |
140 | ext4_set_bit(tmp - start, bh->b_data); | |
717d50e4 | 141 | |
fd034a84 TT |
142 | tmp = ext4_inode_bitmap(sb, gdp); |
143 | if (!flex_bg || ext4_block_in_group(sb, tmp, block_group)) | |
144 | ext4_set_bit(tmp - start, bh->b_data); | |
0bf7e837 | 145 | |
fd034a84 TT |
146 | tmp = ext4_inode_table(sb, gdp); |
147 | for (; tmp < ext4_inode_table(sb, gdp) + | |
148 | sbi->s_itb_per_group; tmp++) { | |
0bf7e837 JS |
149 | if (!flex_bg || ext4_block_in_group(sb, tmp, block_group)) |
150 | ext4_set_bit(tmp - start, bh->b_data); | |
717d50e4 | 151 | } |
fd034a84 TT |
152 | /* |
153 | * Also if the number of blocks within the group is less than | |
154 | * the blocksize * 8 ( which is the size of bitmap ), set rest | |
155 | * of the block bitmap to 1 | |
156 | */ | |
157 | ext4_mark_bitmap_end(num_blocks_in_group(sb, block_group), | |
158 | sb->s_blocksize * 8, bh->b_data); | |
717d50e4 AD |
159 | } |
160 | ||
fd034a84 TT |
161 | /* Return the number of free blocks in a block group. It is used when |
162 | * the block bitmap is uninitialized, so we can't just count the bits | |
163 | * in the bitmap. */ | |
164 | unsigned ext4_free_blocks_after_init(struct super_block *sb, | |
165 | ext4_group_t block_group, | |
166 | struct ext4_group_desc *gdp) | |
167 | { | |
168 | return num_blocks_in_group(sb, block_group) - | |
169 | num_base_meta_blocks(sb, block_group) - | |
170 | ext4_group_used_meta_blocks(sb, block_group, gdp); | |
171 | } | |
717d50e4 | 172 | |
ac27a0ec DK |
173 | /* |
174 | * The free blocks are managed by bitmaps. A file system contains several | |
175 | * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap | |
176 | * block for inodes, N blocks for the inode table and data blocks. | |
177 | * | |
178 | * The file system contains group descriptors which are located after the | |
179 | * super block. Each descriptor contains the number of the bitmap block and | |
180 | * the free blocks count in the block. The descriptors are loaded in memory | |
e627432c | 181 | * when a file system is mounted (see ext4_fill_super). |
ac27a0ec DK |
182 | */ |
183 | ||
ac27a0ec | 184 | /** |
617ba13b | 185 | * ext4_get_group_desc() -- load group descriptor from disk |
ac27a0ec DK |
186 | * @sb: super block |
187 | * @block_group: given block group | |
188 | * @bh: pointer to the buffer head to store the block | |
189 | * group descriptor | |
190 | */ | |
af5bc92d | 191 | struct ext4_group_desc * ext4_get_group_desc(struct super_block *sb, |
fd2d4291 | 192 | ext4_group_t block_group, |
af5bc92d | 193 | struct buffer_head **bh) |
ac27a0ec | 194 | { |
498e5f24 TT |
195 | unsigned int group_desc; |
196 | unsigned int offset; | |
8df9675f | 197 | ext4_group_t ngroups = ext4_get_groups_count(sb); |
af5bc92d | 198 | struct ext4_group_desc *desc; |
617ba13b | 199 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec | 200 | |
8df9675f | 201 | if (block_group >= ngroups) { |
12062ddd ES |
202 | ext4_error(sb, "block_group >= groups_count - block_group = %u," |
203 | " groups_count = %u", block_group, ngroups); | |
ac27a0ec DK |
204 | |
205 | return NULL; | |
206 | } | |
ac27a0ec | 207 | |
617ba13b MC |
208 | group_desc = block_group >> EXT4_DESC_PER_BLOCK_BITS(sb); |
209 | offset = block_group & (EXT4_DESC_PER_BLOCK(sb) - 1); | |
ac27a0ec | 210 | if (!sbi->s_group_desc[group_desc]) { |
12062ddd | 211 | ext4_error(sb, "Group descriptor not loaded - " |
498e5f24 | 212 | "block_group = %u, group_desc = %u, desc = %u", |
af5bc92d | 213 | block_group, group_desc, offset); |
ac27a0ec DK |
214 | return NULL; |
215 | } | |
216 | ||
0d1ee42f AR |
217 | desc = (struct ext4_group_desc *)( |
218 | (__u8 *)sbi->s_group_desc[group_desc]->b_data + | |
219 | offset * EXT4_DESC_SIZE(sb)); | |
ac27a0ec DK |
220 | if (bh) |
221 | *bh = sbi->s_group_desc[group_desc]; | |
0d1ee42f | 222 | return desc; |
ac27a0ec DK |
223 | } |
224 | ||
abcb2947 AK |
225 | static int ext4_valid_block_bitmap(struct super_block *sb, |
226 | struct ext4_group_desc *desc, | |
227 | unsigned int block_group, | |
228 | struct buffer_head *bh) | |
229 | { | |
230 | ext4_grpblk_t offset; | |
231 | ext4_grpblk_t next_zero_bit; | |
232 | ext4_fsblk_t bitmap_blk; | |
233 | ext4_fsblk_t group_first_block; | |
234 | ||
235 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) { | |
236 | /* with FLEX_BG, the inode/block bitmaps and itable | |
237 | * blocks may not be in the group at all | |
238 | * so the bitmap validation will be skipped for those groups | |
239 | * or it has to also read the block group where the bitmaps | |
240 | * are located to verify they are set. | |
241 | */ | |
242 | return 1; | |
243 | } | |
244 | group_first_block = ext4_group_first_block_no(sb, block_group); | |
245 | ||
246 | /* check whether block bitmap block number is set */ | |
247 | bitmap_blk = ext4_block_bitmap(sb, desc); | |
248 | offset = bitmap_blk - group_first_block; | |
249 | if (!ext4_test_bit(offset, bh->b_data)) | |
250 | /* bad block bitmap */ | |
251 | goto err_out; | |
252 | ||
253 | /* check whether the inode bitmap block number is set */ | |
254 | bitmap_blk = ext4_inode_bitmap(sb, desc); | |
255 | offset = bitmap_blk - group_first_block; | |
256 | if (!ext4_test_bit(offset, bh->b_data)) | |
257 | /* bad block bitmap */ | |
258 | goto err_out; | |
259 | ||
260 | /* check whether the inode table block number is set */ | |
261 | bitmap_blk = ext4_inode_table(sb, desc); | |
262 | offset = bitmap_blk - group_first_block; | |
263 | next_zero_bit = ext4_find_next_zero_bit(bh->b_data, | |
264 | offset + EXT4_SB(sb)->s_itb_per_group, | |
265 | offset); | |
266 | if (next_zero_bit >= offset + EXT4_SB(sb)->s_itb_per_group) | |
267 | /* good bitmap for inode tables */ | |
268 | return 1; | |
269 | ||
270 | err_out: | |
12062ddd | 271 | ext4_error(sb, "Invalid block bitmap - block_group = %d, block = %llu", |
abcb2947 AK |
272 | block_group, bitmap_blk); |
273 | return 0; | |
274 | } | |
ac27a0ec | 275 | /** |
574ca174 | 276 | * ext4_read_block_bitmap() |
ac27a0ec DK |
277 | * @sb: super block |
278 | * @block_group: given block group | |
279 | * | |
abcb2947 AK |
280 | * Read the bitmap for a given block_group,and validate the |
281 | * bits for block/inode/inode tables are set in the bitmaps | |
ac27a0ec DK |
282 | * |
283 | * Return buffer_head on success or NULL in case of failure. | |
284 | */ | |
717d50e4 | 285 | struct buffer_head * |
574ca174 | 286 | ext4_read_block_bitmap(struct super_block *sb, ext4_group_t block_group) |
ac27a0ec | 287 | { |
af5bc92d TT |
288 | struct ext4_group_desc *desc; |
289 | struct buffer_head *bh = NULL; | |
7c9e69fa | 290 | ext4_fsblk_t bitmap_blk; |
ac27a0ec | 291 | |
717d50e4 | 292 | desc = ext4_get_group_desc(sb, block_group, NULL); |
ac27a0ec | 293 | if (!desc) |
7c9e69fa AK |
294 | return NULL; |
295 | bitmap_blk = ext4_block_bitmap(sb, desc); | |
abcb2947 AK |
296 | bh = sb_getblk(sb, bitmap_blk); |
297 | if (unlikely(!bh)) { | |
12062ddd | 298 | ext4_error(sb, "Cannot read block bitmap - " |
a9df9a49 | 299 | "block_group = %u, block_bitmap = %llu", |
e29d1cde | 300 | block_group, bitmap_blk); |
abcb2947 AK |
301 | return NULL; |
302 | } | |
2ccb5fb9 AK |
303 | |
304 | if (bitmap_uptodate(bh)) | |
abcb2947 AK |
305 | return bh; |
306 | ||
c806e68f | 307 | lock_buffer(bh); |
2ccb5fb9 AK |
308 | if (bitmap_uptodate(bh)) { |
309 | unlock_buffer(bh); | |
310 | return bh; | |
311 | } | |
955ce5f5 | 312 | ext4_lock_group(sb, block_group); |
717d50e4 | 313 | if (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) { |
abcb2947 | 314 | ext4_init_block_bitmap(sb, bh, block_group, desc); |
2ccb5fb9 | 315 | set_bitmap_uptodate(bh); |
abcb2947 | 316 | set_buffer_uptodate(bh); |
955ce5f5 | 317 | ext4_unlock_group(sb, block_group); |
3300beda | 318 | unlock_buffer(bh); |
abcb2947 | 319 | return bh; |
717d50e4 | 320 | } |
955ce5f5 | 321 | ext4_unlock_group(sb, block_group); |
2ccb5fb9 AK |
322 | if (buffer_uptodate(bh)) { |
323 | /* | |
324 | * if not uninit if bh is uptodate, | |
325 | * bitmap is also uptodate | |
326 | */ | |
327 | set_bitmap_uptodate(bh); | |
328 | unlock_buffer(bh); | |
329 | return bh; | |
330 | } | |
331 | /* | |
332 | * submit the buffer_head for read. We can | |
333 | * safely mark the bitmap as uptodate now. | |
334 | * We do it here so the bitmap uptodate bit | |
335 | * get set with buffer lock held. | |
336 | */ | |
0562e0ba | 337 | trace_ext4_read_block_bitmap_load(sb, block_group); |
2ccb5fb9 | 338 | set_bitmap_uptodate(bh); |
abcb2947 AK |
339 | if (bh_submit_read(bh) < 0) { |
340 | put_bh(bh); | |
12062ddd | 341 | ext4_error(sb, "Cannot read block bitmap - " |
a9df9a49 | 342 | "block_group = %u, block_bitmap = %llu", |
e29d1cde | 343 | block_group, bitmap_blk); |
abcb2947 AK |
344 | return NULL; |
345 | } | |
519deca0 AK |
346 | ext4_valid_block_bitmap(sb, desc, block_group, bh); |
347 | /* | |
348 | * file system mounted not to panic on error, | |
349 | * continue with corrupt bitmap | |
350 | */ | |
ac27a0ec DK |
351 | return bh; |
352 | } | |
ac27a0ec | 353 | |
8c3bf8a0 ES |
354 | /** |
355 | * ext4_has_free_blocks() | |
356 | * @sbi: in-core super block structure. | |
357 | * @nblocks: number of needed blocks | |
358 | * | |
359 | * Check if filesystem has nblocks free & available for allocation. | |
360 | * On success return 1, return 0 on failure. | |
361 | */ | |
55f020db AH |
362 | static int ext4_has_free_blocks(struct ext4_sb_info *sbi, |
363 | s64 nblocks, unsigned int flags) | |
a30d542a | 364 | { |
a996031c | 365 | s64 free_blocks, dirty_blocks, root_blocks; |
a30d542a | 366 | struct percpu_counter *fbc = &sbi->s_freeblocks_counter; |
6bc6e63f | 367 | struct percpu_counter *dbc = &sbi->s_dirtyblocks_counter; |
a30d542a | 368 | |
6bc6e63f AK |
369 | free_blocks = percpu_counter_read_positive(fbc); |
370 | dirty_blocks = percpu_counter_read_positive(dbc); | |
a996031c | 371 | root_blocks = ext4_r_blocks_count(sbi->s_es); |
a30d542a | 372 | |
6bc6e63f AK |
373 | if (free_blocks - (nblocks + root_blocks + dirty_blocks) < |
374 | EXT4_FREEBLOCKS_WATERMARK) { | |
02d21168 AM |
375 | free_blocks = percpu_counter_sum_positive(fbc); |
376 | dirty_blocks = percpu_counter_sum_positive(dbc); | |
6bc6e63f AK |
377 | } |
378 | /* Check whether we have space after | |
a996031c | 379 | * accounting for current dirty blocks & root reserved blocks. |
6bc6e63f | 380 | */ |
a996031c ES |
381 | if (free_blocks >= ((root_blocks + nblocks) + dirty_blocks)) |
382 | return 1; | |
a30d542a | 383 | |
a996031c | 384 | /* Hm, nope. Are (enough) root reserved blocks available? */ |
4c9c544e | 385 | if (sbi->s_resuid == current_fsuid() || |
a996031c | 386 | ((sbi->s_resgid != 0) && in_group_p(sbi->s_resgid)) || |
55f020db AH |
387 | capable(CAP_SYS_RESOURCE) || |
388 | (flags & EXT4_MB_USE_ROOT_BLOCKS)) { | |
389 | ||
a996031c ES |
390 | if (free_blocks >= (nblocks + dirty_blocks)) |
391 | return 1; | |
392 | } | |
393 | ||
394 | return 0; | |
a30d542a AK |
395 | } |
396 | ||
8c3bf8a0 | 397 | int ext4_claim_free_blocks(struct ext4_sb_info *sbi, |
55f020db | 398 | s64 nblocks, unsigned int flags) |
ac27a0ec | 399 | { |
55f020db | 400 | if (ext4_has_free_blocks(sbi, nblocks, flags)) { |
8c3bf8a0 | 401 | percpu_counter_add(&sbi->s_dirtyblocks_counter, nblocks); |
16eb7295 | 402 | return 0; |
8c3bf8a0 ES |
403 | } else |
404 | return -ENOSPC; | |
a30d542a | 405 | } |
07031431 | 406 | |
ac27a0ec | 407 | /** |
617ba13b | 408 | * ext4_should_retry_alloc() |
ac27a0ec DK |
409 | * @sb: super block |
410 | * @retries number of attemps has been made | |
411 | * | |
617ba13b | 412 | * ext4_should_retry_alloc() is called when ENOSPC is returned, and if |
ac27a0ec | 413 | * it is profitable to retry the operation, this function will wait |
25985edc | 414 | * for the current or committing transaction to complete, and then |
ac27a0ec DK |
415 | * return TRUE. |
416 | * | |
417 | * if the total number of retries exceed three times, return FALSE. | |
418 | */ | |
617ba13b | 419 | int ext4_should_retry_alloc(struct super_block *sb, int *retries) |
ac27a0ec | 420 | { |
55f020db | 421 | if (!ext4_has_free_blocks(EXT4_SB(sb), 1, 0) || |
8f64b32e ES |
422 | (*retries)++ > 3 || |
423 | !EXT4_SB(sb)->s_journal) | |
ac27a0ec DK |
424 | return 0; |
425 | ||
426 | jbd_debug(1, "%s: retrying operation after ENOSPC\n", sb->s_id); | |
427 | ||
dab291af | 428 | return jbd2_journal_force_commit_nested(EXT4_SB(sb)->s_journal); |
ac27a0ec DK |
429 | } |
430 | ||
654b4908 | 431 | /* |
d2a17637 | 432 | * ext4_new_meta_blocks() -- allocate block for meta data (indexing) blocks |
654b4908 AK |
433 | * |
434 | * @handle: handle to this transaction | |
435 | * @inode: file inode | |
436 | * @goal: given target block(filesystem wide) | |
97df5d15 | 437 | * @count: pointer to total number of blocks needed |
654b4908 AK |
438 | * @errp: error code |
439 | * | |
97df5d15 | 440 | * Return 1st allocated block number on success, *count stores total account |
d2a17637 | 441 | * error stores in errp pointer |
654b4908 | 442 | */ |
d2a17637 | 443 | ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode, |
55f020db AH |
444 | ext4_fsblk_t goal, unsigned int flags, |
445 | unsigned long *count, int *errp) | |
654b4908 | 446 | { |
97df5d15 | 447 | struct ext4_allocation_request ar; |
d2a17637 | 448 | ext4_fsblk_t ret; |
97df5d15 TT |
449 | |
450 | memset(&ar, 0, sizeof(ar)); | |
451 | /* Fill with neighbour allocated blocks */ | |
452 | ar.inode = inode; | |
453 | ar.goal = goal; | |
454 | ar.len = count ? *count : 1; | |
55f020db | 455 | ar.flags = flags; |
97df5d15 TT |
456 | |
457 | ret = ext4_mb_new_blocks(handle, &ar, errp); | |
458 | if (count) | |
459 | *count = ar.len; | |
d2a17637 | 460 | /* |
72b8ab9d ES |
461 | * Account for the allocated meta blocks. We will never |
462 | * fail EDQUOT for metdata, but we do account for it. | |
d2a17637 | 463 | */ |
f2321097 TT |
464 | if (!(*errp) && |
465 | ext4_test_inode_state(inode, EXT4_STATE_DELALLOC_RESERVED)) { | |
d2a17637 | 466 | spin_lock(&EXT4_I(inode)->i_block_reservation_lock); |
97df5d15 | 467 | EXT4_I(inode)->i_allocated_meta_blocks += ar.len; |
d2a17637 | 468 | spin_unlock(&EXT4_I(inode)->i_block_reservation_lock); |
72b8ab9d | 469 | dquot_alloc_block_nofail(inode, ar.len); |
d2a17637 MC |
470 | } |
471 | return ret; | |
654b4908 AK |
472 | } |
473 | ||
ac27a0ec | 474 | /** |
617ba13b | 475 | * ext4_count_free_blocks() -- count filesystem free blocks |
ac27a0ec DK |
476 | * @sb: superblock |
477 | * | |
478 | * Adds up the number of free blocks from each block group. | |
479 | */ | |
617ba13b | 480 | ext4_fsblk_t ext4_count_free_blocks(struct super_block *sb) |
ac27a0ec | 481 | { |
617ba13b MC |
482 | ext4_fsblk_t desc_count; |
483 | struct ext4_group_desc *gdp; | |
fd2d4291 | 484 | ext4_group_t i; |
8df9675f | 485 | ext4_group_t ngroups = ext4_get_groups_count(sb); |
617ba13b MC |
486 | #ifdef EXT4FS_DEBUG |
487 | struct ext4_super_block *es; | |
488 | ext4_fsblk_t bitmap_count; | |
498e5f24 | 489 | unsigned int x; |
ac27a0ec DK |
490 | struct buffer_head *bitmap_bh = NULL; |
491 | ||
617ba13b | 492 | es = EXT4_SB(sb)->s_es; |
ac27a0ec DK |
493 | desc_count = 0; |
494 | bitmap_count = 0; | |
495 | gdp = NULL; | |
496 | ||
ac27a0ec | 497 | for (i = 0; i < ngroups; i++) { |
617ba13b | 498 | gdp = ext4_get_group_desc(sb, i, NULL); |
ac27a0ec DK |
499 | if (!gdp) |
500 | continue; | |
9fd9784c | 501 | desc_count += ext4_free_blks_count(sb, gdp); |
ac27a0ec | 502 | brelse(bitmap_bh); |
574ca174 | 503 | bitmap_bh = ext4_read_block_bitmap(sb, i); |
ac27a0ec DK |
504 | if (bitmap_bh == NULL) |
505 | continue; | |
506 | ||
617ba13b | 507 | x = ext4_count_free(bitmap_bh, sb->s_blocksize); |
9fd9784c TLSC |
508 | printk(KERN_DEBUG "group %u: stored = %d, counted = %u\n", |
509 | i, ext4_free_blks_count(sb, gdp), x); | |
ac27a0ec DK |
510 | bitmap_count += x; |
511 | } | |
512 | brelse(bitmap_bh); | |
4776004f TT |
513 | printk(KERN_DEBUG "ext4_count_free_blocks: stored = %llu" |
514 | ", computed = %llu, %llu\n", ext4_free_blocks_count(es), | |
515 | desc_count, bitmap_count); | |
ac27a0ec DK |
516 | return bitmap_count; |
517 | #else | |
518 | desc_count = 0; | |
ac27a0ec | 519 | for (i = 0; i < ngroups; i++) { |
617ba13b | 520 | gdp = ext4_get_group_desc(sb, i, NULL); |
ac27a0ec DK |
521 | if (!gdp) |
522 | continue; | |
560671a0 | 523 | desc_count += ext4_free_blks_count(sb, gdp); |
ac27a0ec DK |
524 | } |
525 | ||
526 | return desc_count; | |
527 | #endif | |
528 | } | |
529 | ||
fd2d4291 | 530 | static inline int test_root(ext4_group_t a, int b) |
ac27a0ec DK |
531 | { |
532 | int num = b; | |
533 | ||
534 | while (a > num) | |
535 | num *= b; | |
536 | return num == a; | |
537 | } | |
538 | ||
fd2d4291 | 539 | static int ext4_group_sparse(ext4_group_t group) |
ac27a0ec DK |
540 | { |
541 | if (group <= 1) | |
542 | return 1; | |
543 | if (!(group & 1)) | |
544 | return 0; | |
545 | return (test_root(group, 7) || test_root(group, 5) || | |
546 | test_root(group, 3)); | |
547 | } | |
548 | ||
549 | /** | |
617ba13b | 550 | * ext4_bg_has_super - number of blocks used by the superblock in group |
ac27a0ec DK |
551 | * @sb: superblock for filesystem |
552 | * @group: group number to check | |
553 | * | |
554 | * Return the number of blocks used by the superblock (primary or backup) | |
555 | * in this group. Currently this will be only 0 or 1. | |
556 | */ | |
fd2d4291 | 557 | int ext4_bg_has_super(struct super_block *sb, ext4_group_t group) |
ac27a0ec | 558 | { |
617ba13b MC |
559 | if (EXT4_HAS_RO_COMPAT_FEATURE(sb, |
560 | EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER) && | |
561 | !ext4_group_sparse(group)) | |
ac27a0ec DK |
562 | return 0; |
563 | return 1; | |
564 | } | |
565 | ||
fd2d4291 AM |
566 | static unsigned long ext4_bg_num_gdb_meta(struct super_block *sb, |
567 | ext4_group_t group) | |
ac27a0ec | 568 | { |
617ba13b | 569 | unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb); |
fd2d4291 AM |
570 | ext4_group_t first = metagroup * EXT4_DESC_PER_BLOCK(sb); |
571 | ext4_group_t last = first + EXT4_DESC_PER_BLOCK(sb) - 1; | |
ac27a0ec DK |
572 | |
573 | if (group == first || group == first + 1 || group == last) | |
574 | return 1; | |
575 | return 0; | |
576 | } | |
577 | ||
fd2d4291 AM |
578 | static unsigned long ext4_bg_num_gdb_nometa(struct super_block *sb, |
579 | ext4_group_t group) | |
ac27a0ec | 580 | { |
8dadb198 TT |
581 | if (!ext4_bg_has_super(sb, group)) |
582 | return 0; | |
583 | ||
584 | if (EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG)) | |
585 | return le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg); | |
586 | else | |
587 | return EXT4_SB(sb)->s_gdb_count; | |
ac27a0ec DK |
588 | } |
589 | ||
590 | /** | |
617ba13b | 591 | * ext4_bg_num_gdb - number of blocks used by the group table in group |
ac27a0ec DK |
592 | * @sb: superblock for filesystem |
593 | * @group: group number to check | |
594 | * | |
595 | * Return the number of blocks used by the group descriptor table | |
596 | * (primary or backup) in this group. In the future there may be a | |
597 | * different number of descriptor blocks in each group. | |
598 | */ | |
fd2d4291 | 599 | unsigned long ext4_bg_num_gdb(struct super_block *sb, ext4_group_t group) |
ac27a0ec DK |
600 | { |
601 | unsigned long first_meta_bg = | |
617ba13b MC |
602 | le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg); |
603 | unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb); | |
ac27a0ec | 604 | |
617ba13b | 605 | if (!EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG) || |
ac27a0ec | 606 | metagroup < first_meta_bg) |
af5bc92d | 607 | return ext4_bg_num_gdb_nometa(sb, group); |
ac27a0ec | 608 | |
617ba13b | 609 | return ext4_bg_num_gdb_meta(sb,group); |
ac27a0ec DK |
610 | |
611 | } | |
c2ea3fde | 612 | |
49f7f9af TT |
613 | /* |
614 | * This function returns the number of file system metadata blocks at | |
615 | * the beginning of a block group, including the reserved gdt blocks. | |
616 | */ | |
617 | static unsigned int num_base_meta_blocks(struct super_block *sb, | |
618 | ext4_group_t block_group) | |
619 | { | |
620 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
621 | int num; | |
622 | ||
623 | /* Check for superblock and gdt backups in this group */ | |
624 | num = ext4_bg_has_super(sb, block_group); | |
625 | ||
626 | if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) || | |
627 | block_group < le32_to_cpu(sbi->s_es->s_first_meta_bg) * | |
628 | sbi->s_desc_per_block) { | |
629 | if (num) { | |
630 | num += ext4_bg_num_gdb(sb, block_group); | |
631 | num += le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks); | |
632 | } | |
633 | } else { /* For META_BG_BLOCK_GROUPS */ | |
634 | num += ext4_bg_num_gdb(sb, block_group); | |
635 | } | |
636 | return num; | |
637 | } | |
f86186b4 ES |
638 | /** |
639 | * ext4_inode_to_goal_block - return a hint for block allocation | |
640 | * @inode: inode for block allocation | |
641 | * | |
642 | * Return the ideal location to start allocating blocks for a | |
643 | * newly created inode. | |
644 | */ | |
645 | ext4_fsblk_t ext4_inode_to_goal_block(struct inode *inode) | |
646 | { | |
647 | struct ext4_inode_info *ei = EXT4_I(inode); | |
648 | ext4_group_t block_group; | |
649 | ext4_grpblk_t colour; | |
650 | int flex_size = ext4_flex_bg_size(EXT4_SB(inode->i_sb)); | |
651 | ext4_fsblk_t bg_start; | |
652 | ext4_fsblk_t last_block; | |
653 | ||
654 | block_group = ei->i_block_group; | |
655 | if (flex_size >= EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME) { | |
656 | /* | |
657 | * If there are at least EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME | |
658 | * block groups per flexgroup, reserve the first block | |
659 | * group for directories and special files. Regular | |
660 | * files will start at the second block group. This | |
661 | * tends to speed up directory access and improves | |
662 | * fsck times. | |
663 | */ | |
664 | block_group &= ~(flex_size-1); | |
665 | if (S_ISREG(inode->i_mode)) | |
666 | block_group++; | |
667 | } | |
668 | bg_start = ext4_group_first_block_no(inode->i_sb, block_group); | |
669 | last_block = ext4_blocks_count(EXT4_SB(inode->i_sb)->s_es) - 1; | |
670 | ||
671 | /* | |
672 | * If we are doing delayed allocation, we don't need take | |
673 | * colour into account. | |
674 | */ | |
675 | if (test_opt(inode->i_sb, DELALLOC)) | |
676 | return bg_start; | |
677 | ||
678 | if (bg_start + EXT4_BLOCKS_PER_GROUP(inode->i_sb) <= last_block) | |
679 | colour = (current->pid % 16) * | |
680 | (EXT4_BLOCKS_PER_GROUP(inode->i_sb) / 16); | |
681 | else | |
682 | colour = (current->pid % 16) * ((last_block - bg_start) / 16); | |
683 | return bg_start + colour; | |
684 | } | |
685 |