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 | ||
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 | */ | |
86 | 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 TT |
178 | /* Initializes an uninitialized block bitmap */ |
179 | void ext4_init_block_bitmap(struct super_block *sb, struct buffer_head *bh, | |
180 | ext4_group_t block_group, | |
181 | struct ext4_group_desc *gdp) | |
717d50e4 | 182 | { |
d5b8f310 | 183 | unsigned int bit, bit_max; |
717d50e4 | 184 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
fd034a84 TT |
185 | ext4_fsblk_t start, tmp; |
186 | int flex_bg = 0; | |
bdfb6ff4 | 187 | struct ext4_group_info *grp; |
fd034a84 TT |
188 | |
189 | J_ASSERT_BH(bh, buffer_locked(bh)); | |
190 | ||
191 | /* If checksum is bad mark all blocks used to prevent allocation | |
192 | * essentially implementing a per-group read-only flag. */ | |
feb0ab32 | 193 | if (!ext4_group_desc_csum_verify(sb, block_group, gdp)) { |
fd034a84 | 194 | ext4_error(sb, "Checksum bad for group %u", block_group); |
bdfb6ff4 DW |
195 | grp = ext4_get_group_info(sb, block_group); |
196 | set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, &grp->bb_state); | |
197 | set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, &grp->bb_state); | |
fd034a84 | 198 | return; |
717d50e4 | 199 | } |
fd034a84 | 200 | memset(bh->b_data, 0, sb->s_blocksize); |
717d50e4 | 201 | |
d5b8f310 | 202 | bit_max = ext4_num_base_meta_clusters(sb, block_group); |
fd034a84 TT |
203 | for (bit = 0; bit < bit_max; bit++) |
204 | ext4_set_bit(bit, bh->b_data); | |
d00a6d7b | 205 | |
fd034a84 | 206 | start = ext4_group_first_block_no(sb, block_group); |
717d50e4 | 207 | |
fd034a84 TT |
208 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) |
209 | flex_bg = 1; | |
717d50e4 | 210 | |
fd034a84 TT |
211 | /* Set bits for block and inode bitmaps, and inode table */ |
212 | tmp = ext4_block_bitmap(sb, gdp); | |
213 | if (!flex_bg || ext4_block_in_group(sb, tmp, block_group)) | |
d5b8f310 | 214 | ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data); |
717d50e4 | 215 | |
fd034a84 TT |
216 | tmp = ext4_inode_bitmap(sb, gdp); |
217 | if (!flex_bg || ext4_block_in_group(sb, tmp, block_group)) | |
d5b8f310 | 218 | ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data); |
0bf7e837 | 219 | |
fd034a84 TT |
220 | tmp = ext4_inode_table(sb, gdp); |
221 | for (; tmp < ext4_inode_table(sb, gdp) + | |
222 | sbi->s_itb_per_group; tmp++) { | |
0bf7e837 | 223 | if (!flex_bg || ext4_block_in_group(sb, tmp, block_group)) |
d5b8f310 | 224 | ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data); |
717d50e4 | 225 | } |
d5b8f310 | 226 | |
fd034a84 TT |
227 | /* |
228 | * Also if the number of blocks within the group is less than | |
229 | * the blocksize * 8 ( which is the size of bitmap ), set rest | |
230 | * of the block bitmap to 1 | |
231 | */ | |
d5b8f310 | 232 | ext4_mark_bitmap_end(num_clusters_in_group(sb, block_group), |
fd034a84 | 233 | sb->s_blocksize * 8, bh->b_data); |
79f1ba49 | 234 | ext4_block_bitmap_csum_set(sb, block_group, gdp, bh); |
feb0ab32 | 235 | ext4_group_desc_csum_set(sb, block_group, gdp); |
717d50e4 AD |
236 | } |
237 | ||
fd034a84 TT |
238 | /* Return the number of free blocks in a block group. It is used when |
239 | * the block bitmap is uninitialized, so we can't just count the bits | |
240 | * in the bitmap. */ | |
cff1dfd7 TT |
241 | unsigned ext4_free_clusters_after_init(struct super_block *sb, |
242 | ext4_group_t block_group, | |
243 | struct ext4_group_desc *gdp) | |
fd034a84 | 244 | { |
d5b8f310 TT |
245 | return num_clusters_in_group(sb, block_group) - |
246 | ext4_num_overhead_clusters(sb, block_group, gdp); | |
fd034a84 | 247 | } |
717d50e4 | 248 | |
ac27a0ec DK |
249 | /* |
250 | * The free blocks are managed by bitmaps. A file system contains several | |
251 | * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap | |
252 | * block for inodes, N blocks for the inode table and data blocks. | |
253 | * | |
254 | * The file system contains group descriptors which are located after the | |
255 | * super block. Each descriptor contains the number of the bitmap block and | |
256 | * the free blocks count in the block. The descriptors are loaded in memory | |
e627432c | 257 | * when a file system is mounted (see ext4_fill_super). |
ac27a0ec DK |
258 | */ |
259 | ||
ac27a0ec | 260 | /** |
617ba13b | 261 | * ext4_get_group_desc() -- load group descriptor from disk |
ac27a0ec DK |
262 | * @sb: super block |
263 | * @block_group: given block group | |
264 | * @bh: pointer to the buffer head to store the block | |
265 | * group descriptor | |
266 | */ | |
af5bc92d | 267 | struct ext4_group_desc * ext4_get_group_desc(struct super_block *sb, |
fd2d4291 | 268 | ext4_group_t block_group, |
af5bc92d | 269 | struct buffer_head **bh) |
ac27a0ec | 270 | { |
498e5f24 TT |
271 | unsigned int group_desc; |
272 | unsigned int offset; | |
8df9675f | 273 | ext4_group_t ngroups = ext4_get_groups_count(sb); |
af5bc92d | 274 | struct ext4_group_desc *desc; |
617ba13b | 275 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec | 276 | |
8df9675f | 277 | if (block_group >= ngroups) { |
12062ddd ES |
278 | ext4_error(sb, "block_group >= groups_count - block_group = %u," |
279 | " groups_count = %u", block_group, ngroups); | |
ac27a0ec DK |
280 | |
281 | return NULL; | |
282 | } | |
ac27a0ec | 283 | |
617ba13b MC |
284 | group_desc = block_group >> EXT4_DESC_PER_BLOCK_BITS(sb); |
285 | offset = block_group & (EXT4_DESC_PER_BLOCK(sb) - 1); | |
ac27a0ec | 286 | if (!sbi->s_group_desc[group_desc]) { |
12062ddd | 287 | ext4_error(sb, "Group descriptor not loaded - " |
498e5f24 | 288 | "block_group = %u, group_desc = %u, desc = %u", |
af5bc92d | 289 | block_group, group_desc, offset); |
ac27a0ec DK |
290 | return NULL; |
291 | } | |
292 | ||
0d1ee42f AR |
293 | desc = (struct ext4_group_desc *)( |
294 | (__u8 *)sbi->s_group_desc[group_desc]->b_data + | |
295 | offset * EXT4_DESC_SIZE(sb)); | |
ac27a0ec DK |
296 | if (bh) |
297 | *bh = sbi->s_group_desc[group_desc]; | |
0d1ee42f | 298 | return desc; |
ac27a0ec DK |
299 | } |
300 | ||
7a4c5de2 TT |
301 | /* |
302 | * Return the block number which was discovered to be invalid, or 0 if | |
303 | * the block bitmap is valid. | |
304 | */ | |
305 | static ext4_fsblk_t ext4_valid_block_bitmap(struct super_block *sb, | |
306 | struct ext4_group_desc *desc, | |
dbde0abe | 307 | ext4_group_t block_group, |
7a4c5de2 | 308 | struct buffer_head *bh) |
abcb2947 AK |
309 | { |
310 | ext4_grpblk_t offset; | |
311 | ext4_grpblk_t next_zero_bit; | |
7a4c5de2 | 312 | ext4_fsblk_t blk; |
abcb2947 AK |
313 | ext4_fsblk_t group_first_block; |
314 | ||
315 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) { | |
316 | /* with FLEX_BG, the inode/block bitmaps and itable | |
317 | * blocks may not be in the group at all | |
318 | * so the bitmap validation will be skipped for those groups | |
319 | * or it has to also read the block group where the bitmaps | |
320 | * are located to verify they are set. | |
321 | */ | |
7a4c5de2 | 322 | return 0; |
abcb2947 AK |
323 | } |
324 | group_first_block = ext4_group_first_block_no(sb, block_group); | |
325 | ||
326 | /* check whether block bitmap block number is set */ | |
7a4c5de2 TT |
327 | blk = ext4_block_bitmap(sb, desc); |
328 | offset = blk - group_first_block; | |
abcb2947 AK |
329 | if (!ext4_test_bit(offset, bh->b_data)) |
330 | /* bad block bitmap */ | |
7a4c5de2 | 331 | return blk; |
abcb2947 AK |
332 | |
333 | /* check whether the inode bitmap block number is set */ | |
7a4c5de2 TT |
334 | blk = ext4_inode_bitmap(sb, desc); |
335 | offset = blk - group_first_block; | |
abcb2947 AK |
336 | if (!ext4_test_bit(offset, bh->b_data)) |
337 | /* bad block bitmap */ | |
7a4c5de2 | 338 | return blk; |
abcb2947 AK |
339 | |
340 | /* check whether the inode table block number is set */ | |
7a4c5de2 TT |
341 | blk = ext4_inode_table(sb, desc); |
342 | offset = blk - group_first_block; | |
abcb2947 AK |
343 | next_zero_bit = ext4_find_next_zero_bit(bh->b_data, |
344 | offset + EXT4_SB(sb)->s_itb_per_group, | |
345 | offset); | |
7a4c5de2 TT |
346 | if (next_zero_bit < offset + EXT4_SB(sb)->s_itb_per_group) |
347 | /* bad bitmap for inode tables */ | |
348 | return blk; | |
abcb2947 AK |
349 | return 0; |
350 | } | |
fa77dcfa DW |
351 | |
352 | void ext4_validate_block_bitmap(struct super_block *sb, | |
353 | struct ext4_group_desc *desc, | |
dbde0abe | 354 | ext4_group_t block_group, |
fa77dcfa DW |
355 | struct buffer_head *bh) |
356 | { | |
7a4c5de2 | 357 | ext4_fsblk_t blk; |
163a203d | 358 | struct ext4_group_info *grp = ext4_get_group_info(sb, block_group); |
7a4c5de2 | 359 | |
fa77dcfa DW |
360 | if (buffer_verified(bh)) |
361 | return; | |
362 | ||
363 | ext4_lock_group(sb, block_group); | |
7a4c5de2 TT |
364 | blk = ext4_valid_block_bitmap(sb, desc, block_group, bh); |
365 | if (unlikely(blk != 0)) { | |
366 | ext4_unlock_group(sb, block_group); | |
367 | ext4_error(sb, "bg %u: block %llu: invalid block bitmap", | |
368 | block_group, blk); | |
163a203d | 369 | set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, &grp->bb_state); |
7a4c5de2 TT |
370 | return; |
371 | } | |
372 | if (unlikely(!ext4_block_bitmap_csum_verify(sb, block_group, | |
79f1ba49 | 373 | desc, bh))) { |
7a4c5de2 TT |
374 | ext4_unlock_group(sb, block_group); |
375 | ext4_error(sb, "bg %u: bad block bitmap checksum", block_group); | |
163a203d | 376 | set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, &grp->bb_state); |
7a4c5de2 TT |
377 | return; |
378 | } | |
379 | set_buffer_verified(bh); | |
fa77dcfa DW |
380 | ext4_unlock_group(sb, block_group); |
381 | } | |
382 | ||
ac27a0ec | 383 | /** |
15b49132 | 384 | * ext4_read_block_bitmap_nowait() |
ac27a0ec DK |
385 | * @sb: super block |
386 | * @block_group: given block group | |
387 | * | |
abcb2947 AK |
388 | * Read the bitmap for a given block_group,and validate the |
389 | * bits for block/inode/inode tables are set in the bitmaps | |
ac27a0ec DK |
390 | * |
391 | * Return buffer_head on success or NULL in case of failure. | |
392 | */ | |
717d50e4 | 393 | struct buffer_head * |
813e5727 | 394 | ext4_read_block_bitmap_nowait(struct super_block *sb, ext4_group_t block_group) |
ac27a0ec | 395 | { |
af5bc92d | 396 | struct ext4_group_desc *desc; |
813e5727 | 397 | struct buffer_head *bh; |
7c9e69fa | 398 | ext4_fsblk_t bitmap_blk; |
ac27a0ec | 399 | |
717d50e4 | 400 | desc = ext4_get_group_desc(sb, block_group, NULL); |
ac27a0ec | 401 | if (!desc) |
7c9e69fa AK |
402 | return NULL; |
403 | bitmap_blk = ext4_block_bitmap(sb, desc); | |
abcb2947 AK |
404 | bh = sb_getblk(sb, bitmap_blk); |
405 | if (unlikely(!bh)) { | |
813e5727 TT |
406 | ext4_error(sb, "Cannot get buffer for block bitmap - " |
407 | "block_group = %u, block_bitmap = %llu", | |
408 | block_group, bitmap_blk); | |
abcb2947 AK |
409 | return NULL; |
410 | } | |
2ccb5fb9 AK |
411 | |
412 | if (bitmap_uptodate(bh)) | |
fa77dcfa | 413 | goto verify; |
abcb2947 | 414 | |
c806e68f | 415 | lock_buffer(bh); |
2ccb5fb9 AK |
416 | if (bitmap_uptodate(bh)) { |
417 | unlock_buffer(bh); | |
fa77dcfa | 418 | goto verify; |
2ccb5fb9 | 419 | } |
955ce5f5 | 420 | ext4_lock_group(sb, block_group); |
717d50e4 | 421 | if (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) { |
abcb2947 | 422 | ext4_init_block_bitmap(sb, bh, block_group, desc); |
2ccb5fb9 | 423 | set_bitmap_uptodate(bh); |
abcb2947 | 424 | set_buffer_uptodate(bh); |
955ce5f5 | 425 | ext4_unlock_group(sb, block_group); |
3300beda | 426 | unlock_buffer(bh); |
abcb2947 | 427 | return bh; |
717d50e4 | 428 | } |
955ce5f5 | 429 | ext4_unlock_group(sb, block_group); |
2ccb5fb9 AK |
430 | if (buffer_uptodate(bh)) { |
431 | /* | |
432 | * if not uninit if bh is uptodate, | |
433 | * bitmap is also uptodate | |
434 | */ | |
435 | set_bitmap_uptodate(bh); | |
436 | unlock_buffer(bh); | |
fa77dcfa | 437 | goto verify; |
2ccb5fb9 AK |
438 | } |
439 | /* | |
813e5727 | 440 | * submit the buffer_head for reading |
2ccb5fb9 | 441 | */ |
813e5727 | 442 | set_buffer_new(bh); |
0562e0ba | 443 | trace_ext4_read_block_bitmap_load(sb, block_group); |
813e5727 TT |
444 | bh->b_end_io = ext4_end_bitmap_read; |
445 | get_bh(bh); | |
9f203507 | 446 | submit_bh(READ | REQ_META | REQ_PRIO, bh); |
813e5727 | 447 | return bh; |
fa77dcfa DW |
448 | verify: |
449 | ext4_validate_block_bitmap(sb, desc, block_group, bh); | |
48d9eb97 DW |
450 | if (buffer_verified(bh)) |
451 | return bh; | |
452 | put_bh(bh); | |
453 | return NULL; | |
813e5727 TT |
454 | } |
455 | ||
456 | /* Returns 0 on success, 1 on error */ | |
457 | int ext4_wait_block_bitmap(struct super_block *sb, ext4_group_t block_group, | |
458 | struct buffer_head *bh) | |
459 | { | |
460 | struct ext4_group_desc *desc; | |
461 | ||
462 | if (!buffer_new(bh)) | |
463 | return 0; | |
464 | desc = ext4_get_group_desc(sb, block_group, NULL); | |
465 | if (!desc) | |
466 | return 1; | |
467 | wait_on_buffer(bh); | |
468 | if (!buffer_uptodate(bh)) { | |
12062ddd | 469 | ext4_error(sb, "Cannot read block bitmap - " |
813e5727 | 470 | "block_group = %u, block_bitmap = %llu", |
d4dc462f | 471 | block_group, (unsigned long long) bh->b_blocknr); |
813e5727 | 472 | return 1; |
abcb2947 | 473 | } |
813e5727 TT |
474 | clear_buffer_new(bh); |
475 | /* Panic or remount fs read-only if block bitmap is invalid */ | |
fa77dcfa | 476 | ext4_validate_block_bitmap(sb, desc, block_group, bh); |
48d9eb97 DW |
477 | /* ...but check for error just in case errors=continue. */ |
478 | return !buffer_verified(bh); | |
813e5727 TT |
479 | } |
480 | ||
481 | struct buffer_head * | |
482 | ext4_read_block_bitmap(struct super_block *sb, ext4_group_t block_group) | |
483 | { | |
484 | struct buffer_head *bh; | |
485 | ||
486 | bh = ext4_read_block_bitmap_nowait(sb, block_group); | |
15b49132 EG |
487 | if (!bh) |
488 | return NULL; | |
813e5727 TT |
489 | if (ext4_wait_block_bitmap(sb, block_group, bh)) { |
490 | put_bh(bh); | |
491 | return NULL; | |
492 | } | |
ac27a0ec DK |
493 | return bh; |
494 | } | |
ac27a0ec | 495 | |
8c3bf8a0 | 496 | /** |
df55c99d | 497 | * ext4_has_free_clusters() |
8c3bf8a0 | 498 | * @sbi: in-core super block structure. |
df55c99d TT |
499 | * @nclusters: number of needed blocks |
500 | * @flags: flags from ext4_mb_new_blocks() | |
8c3bf8a0 | 501 | * |
df55c99d | 502 | * Check if filesystem has nclusters free & available for allocation. |
8c3bf8a0 ES |
503 | * On success return 1, return 0 on failure. |
504 | */ | |
df55c99d TT |
505 | static int ext4_has_free_clusters(struct ext4_sb_info *sbi, |
506 | s64 nclusters, unsigned int flags) | |
a30d542a | 507 | { |
27dd4385 | 508 | s64 free_clusters, dirty_clusters, rsv, resv_clusters; |
57042651 | 509 | struct percpu_counter *fcc = &sbi->s_freeclusters_counter; |
df55c99d | 510 | struct percpu_counter *dcc = &sbi->s_dirtyclusters_counter; |
a30d542a | 511 | |
df55c99d TT |
512 | free_clusters = percpu_counter_read_positive(fcc); |
513 | dirty_clusters = percpu_counter_read_positive(dcc); | |
27dd4385 | 514 | resv_clusters = atomic64_read(&sbi->s_resv_clusters); |
304e220f LC |
515 | |
516 | /* | |
517 | * r_blocks_count should always be multiple of the cluster ratio so | |
518 | * we are safe to do a plane bit shift only. | |
519 | */ | |
27dd4385 LC |
520 | rsv = (ext4_r_blocks_count(sbi->s_es) >> sbi->s_cluster_bits) + |
521 | resv_clusters; | |
a30d542a | 522 | |
27dd4385 | 523 | if (free_clusters - (nclusters + rsv + dirty_clusters) < |
df55c99d | 524 | EXT4_FREECLUSTERS_WATERMARK) { |
304e220f | 525 | free_clusters = percpu_counter_sum_positive(fcc); |
df55c99d | 526 | dirty_clusters = percpu_counter_sum_positive(dcc); |
6bc6e63f | 527 | } |
df55c99d TT |
528 | /* Check whether we have space after accounting for current |
529 | * dirty clusters & root reserved clusters. | |
6bc6e63f | 530 | */ |
27dd4385 | 531 | if (free_clusters >= (rsv + nclusters + dirty_clusters)) |
a996031c | 532 | return 1; |
a30d542a | 533 | |
df55c99d | 534 | /* Hm, nope. Are (enough) root reserved clusters available? */ |
08cefc7a EB |
535 | if (uid_eq(sbi->s_resuid, current_fsuid()) || |
536 | (!gid_eq(sbi->s_resgid, GLOBAL_ROOT_GID) && in_group_p(sbi->s_resgid)) || | |
55f020db | 537 | capable(CAP_SYS_RESOURCE) || |
27dd4385 | 538 | (flags & EXT4_MB_USE_ROOT_BLOCKS)) { |
55f020db | 539 | |
27dd4385 LC |
540 | if (free_clusters >= (nclusters + dirty_clusters + |
541 | resv_clusters)) | |
542 | return 1; | |
543 | } | |
544 | /* No free blocks. Let's see if we can dip into reserved pool */ | |
545 | if (flags & EXT4_MB_USE_RESERVED) { | |
df55c99d | 546 | if (free_clusters >= (nclusters + dirty_clusters)) |
a996031c ES |
547 | return 1; |
548 | } | |
549 | ||
550 | return 0; | |
a30d542a AK |
551 | } |
552 | ||
e7d5f315 TT |
553 | int ext4_claim_free_clusters(struct ext4_sb_info *sbi, |
554 | s64 nclusters, unsigned int flags) | |
ac27a0ec | 555 | { |
df55c99d | 556 | if (ext4_has_free_clusters(sbi, nclusters, flags)) { |
e7d5f315 | 557 | percpu_counter_add(&sbi->s_dirtyclusters_counter, nclusters); |
16eb7295 | 558 | return 0; |
8c3bf8a0 ES |
559 | } else |
560 | return -ENOSPC; | |
a30d542a | 561 | } |
07031431 | 562 | |
ac27a0ec | 563 | /** |
617ba13b | 564 | * ext4_should_retry_alloc() |
ac27a0ec DK |
565 | * @sb: super block |
566 | * @retries number of attemps has been made | |
567 | * | |
617ba13b | 568 | * ext4_should_retry_alloc() is called when ENOSPC is returned, and if |
ac27a0ec | 569 | * it is profitable to retry the operation, this function will wait |
25985edc | 570 | * for the current or committing transaction to complete, and then |
ac27a0ec DK |
571 | * return TRUE. |
572 | * | |
573 | * if the total number of retries exceed three times, return FALSE. | |
574 | */ | |
617ba13b | 575 | int ext4_should_retry_alloc(struct super_block *sb, int *retries) |
ac27a0ec | 576 | { |
df55c99d | 577 | if (!ext4_has_free_clusters(EXT4_SB(sb), 1, 0) || |
8f64b32e ES |
578 | (*retries)++ > 3 || |
579 | !EXT4_SB(sb)->s_journal) | |
ac27a0ec DK |
580 | return 0; |
581 | ||
582 | jbd_debug(1, "%s: retrying operation after ENOSPC\n", sb->s_id); | |
583 | ||
dab291af | 584 | return jbd2_journal_force_commit_nested(EXT4_SB(sb)->s_journal); |
ac27a0ec DK |
585 | } |
586 | ||
654b4908 | 587 | /* |
d2a17637 | 588 | * ext4_new_meta_blocks() -- allocate block for meta data (indexing) blocks |
654b4908 AK |
589 | * |
590 | * @handle: handle to this transaction | |
591 | * @inode: file inode | |
592 | * @goal: given target block(filesystem wide) | |
7b415bf6 | 593 | * @count: pointer to total number of clusters needed |
654b4908 AK |
594 | * @errp: error code |
595 | * | |
97df5d15 | 596 | * Return 1st allocated block number on success, *count stores total account |
d2a17637 | 597 | * error stores in errp pointer |
654b4908 | 598 | */ |
d2a17637 | 599 | ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode, |
55f020db AH |
600 | ext4_fsblk_t goal, unsigned int flags, |
601 | unsigned long *count, int *errp) | |
654b4908 | 602 | { |
97df5d15 | 603 | struct ext4_allocation_request ar; |
d2a17637 | 604 | ext4_fsblk_t ret; |
97df5d15 TT |
605 | |
606 | memset(&ar, 0, sizeof(ar)); | |
607 | /* Fill with neighbour allocated blocks */ | |
608 | ar.inode = inode; | |
609 | ar.goal = goal; | |
610 | ar.len = count ? *count : 1; | |
55f020db | 611 | ar.flags = flags; |
97df5d15 TT |
612 | |
613 | ret = ext4_mb_new_blocks(handle, &ar, errp); | |
614 | if (count) | |
615 | *count = ar.len; | |
d2a17637 | 616 | /* |
72b8ab9d ES |
617 | * Account for the allocated meta blocks. We will never |
618 | * fail EDQUOT for metdata, but we do account for it. | |
d2a17637 | 619 | */ |
f2321097 TT |
620 | if (!(*errp) && |
621 | ext4_test_inode_state(inode, EXT4_STATE_DELALLOC_RESERVED)) { | |
d2a17637 | 622 | spin_lock(&EXT4_I(inode)->i_block_reservation_lock); |
97df5d15 | 623 | EXT4_I(inode)->i_allocated_meta_blocks += ar.len; |
d2a17637 | 624 | spin_unlock(&EXT4_I(inode)->i_block_reservation_lock); |
7b415bf6 AK |
625 | dquot_alloc_block_nofail(inode, |
626 | EXT4_C2B(EXT4_SB(inode->i_sb), ar.len)); | |
d2a17637 MC |
627 | } |
628 | return ret; | |
654b4908 AK |
629 | } |
630 | ||
ac27a0ec | 631 | /** |
5dee5437 | 632 | * ext4_count_free_clusters() -- count filesystem free clusters |
ac27a0ec DK |
633 | * @sb: superblock |
634 | * | |
5dee5437 | 635 | * Adds up the number of free clusters from each block group. |
ac27a0ec | 636 | */ |
5dee5437 | 637 | ext4_fsblk_t ext4_count_free_clusters(struct super_block *sb) |
ac27a0ec | 638 | { |
617ba13b MC |
639 | ext4_fsblk_t desc_count; |
640 | struct ext4_group_desc *gdp; | |
fd2d4291 | 641 | ext4_group_t i; |
8df9675f | 642 | ext4_group_t ngroups = ext4_get_groups_count(sb); |
2746f7a1 | 643 | struct ext4_group_info *grp; |
617ba13b MC |
644 | #ifdef EXT4FS_DEBUG |
645 | struct ext4_super_block *es; | |
646 | ext4_fsblk_t bitmap_count; | |
498e5f24 | 647 | unsigned int x; |
ac27a0ec DK |
648 | struct buffer_head *bitmap_bh = NULL; |
649 | ||
617ba13b | 650 | es = EXT4_SB(sb)->s_es; |
ac27a0ec DK |
651 | desc_count = 0; |
652 | bitmap_count = 0; | |
653 | gdp = NULL; | |
654 | ||
ac27a0ec | 655 | for (i = 0; i < ngroups; i++) { |
617ba13b | 656 | gdp = ext4_get_group_desc(sb, i, NULL); |
ac27a0ec DK |
657 | if (!gdp) |
658 | continue; | |
2746f7a1 DW |
659 | grp = NULL; |
660 | if (EXT4_SB(sb)->s_group_info) | |
661 | grp = ext4_get_group_info(sb, i); | |
662 | if (!grp || !EXT4_MB_GRP_BBITMAP_CORRUPT(grp)) | |
663 | desc_count += ext4_free_group_clusters(sb, gdp); | |
ac27a0ec | 664 | brelse(bitmap_bh); |
574ca174 | 665 | bitmap_bh = ext4_read_block_bitmap(sb, i); |
ac27a0ec DK |
666 | if (bitmap_bh == NULL) |
667 | continue; | |
668 | ||
f6fb99ca TT |
669 | x = ext4_count_free(bitmap_bh->b_data, |
670 | EXT4_BLOCKS_PER_GROUP(sb) / 8); | |
9fd9784c | 671 | printk(KERN_DEBUG "group %u: stored = %d, counted = %u\n", |
021b65bb | 672 | i, ext4_free_group_clusters(sb, gdp), x); |
ac27a0ec DK |
673 | bitmap_count += x; |
674 | } | |
675 | brelse(bitmap_bh); | |
5dee5437 TT |
676 | printk(KERN_DEBUG "ext4_count_free_clusters: stored = %llu" |
677 | ", computed = %llu, %llu\n", | |
810da240 | 678 | EXT4_NUM_B2C(EXT4_SB(sb), ext4_free_blocks_count(es)), |
4776004f | 679 | desc_count, bitmap_count); |
ac27a0ec DK |
680 | return bitmap_count; |
681 | #else | |
682 | desc_count = 0; | |
ac27a0ec | 683 | for (i = 0; i < ngroups; i++) { |
617ba13b | 684 | gdp = ext4_get_group_desc(sb, i, NULL); |
ac27a0ec DK |
685 | if (!gdp) |
686 | continue; | |
2746f7a1 DW |
687 | grp = NULL; |
688 | if (EXT4_SB(sb)->s_group_info) | |
689 | grp = ext4_get_group_info(sb, i); | |
690 | if (!grp || !EXT4_MB_GRP_BBITMAP_CORRUPT(grp)) | |
691 | desc_count += ext4_free_group_clusters(sb, gdp); | |
ac27a0ec DK |
692 | } |
693 | ||
694 | return desc_count; | |
695 | #endif | |
696 | } | |
697 | ||
fd2d4291 | 698 | static inline int test_root(ext4_group_t a, int b) |
ac27a0ec | 699 | { |
f4afb4f4 TT |
700 | while (1) { |
701 | if (a < b) | |
702 | return 0; | |
703 | if (a == b) | |
704 | return 1; | |
705 | if ((a % b) != 0) | |
706 | return 0; | |
707 | a = a / b; | |
708 | } | |
ac27a0ec DK |
709 | } |
710 | ||
fd2d4291 | 711 | static int ext4_group_sparse(ext4_group_t group) |
ac27a0ec DK |
712 | { |
713 | if (group <= 1) | |
714 | return 1; | |
715 | if (!(group & 1)) | |
716 | return 0; | |
717 | return (test_root(group, 7) || test_root(group, 5) || | |
718 | test_root(group, 3)); | |
719 | } | |
720 | ||
721 | /** | |
617ba13b | 722 | * ext4_bg_has_super - number of blocks used by the superblock in group |
ac27a0ec DK |
723 | * @sb: superblock for filesystem |
724 | * @group: group number to check | |
725 | * | |
726 | * Return the number of blocks used by the superblock (primary or backup) | |
727 | * in this group. Currently this will be only 0 or 1. | |
728 | */ | |
fd2d4291 | 729 | int ext4_bg_has_super(struct super_block *sb, ext4_group_t group) |
ac27a0ec | 730 | { |
617ba13b MC |
731 | if (EXT4_HAS_RO_COMPAT_FEATURE(sb, |
732 | EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER) && | |
733 | !ext4_group_sparse(group)) | |
ac27a0ec DK |
734 | return 0; |
735 | return 1; | |
736 | } | |
737 | ||
fd2d4291 AM |
738 | static unsigned long ext4_bg_num_gdb_meta(struct super_block *sb, |
739 | ext4_group_t group) | |
ac27a0ec | 740 | { |
617ba13b | 741 | unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb); |
fd2d4291 AM |
742 | ext4_group_t first = metagroup * EXT4_DESC_PER_BLOCK(sb); |
743 | ext4_group_t last = first + EXT4_DESC_PER_BLOCK(sb) - 1; | |
ac27a0ec DK |
744 | |
745 | if (group == first || group == first + 1 || group == last) | |
746 | return 1; | |
747 | return 0; | |
748 | } | |
749 | ||
fd2d4291 AM |
750 | static unsigned long ext4_bg_num_gdb_nometa(struct super_block *sb, |
751 | ext4_group_t group) | |
ac27a0ec | 752 | { |
8dadb198 TT |
753 | if (!ext4_bg_has_super(sb, group)) |
754 | return 0; | |
755 | ||
756 | if (EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG)) | |
757 | return le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg); | |
758 | else | |
759 | return EXT4_SB(sb)->s_gdb_count; | |
ac27a0ec DK |
760 | } |
761 | ||
762 | /** | |
617ba13b | 763 | * ext4_bg_num_gdb - number of blocks used by the group table in group |
ac27a0ec DK |
764 | * @sb: superblock for filesystem |
765 | * @group: group number to check | |
766 | * | |
767 | * Return the number of blocks used by the group descriptor table | |
768 | * (primary or backup) in this group. In the future there may be a | |
769 | * different number of descriptor blocks in each group. | |
770 | */ | |
fd2d4291 | 771 | unsigned long ext4_bg_num_gdb(struct super_block *sb, ext4_group_t group) |
ac27a0ec DK |
772 | { |
773 | unsigned long first_meta_bg = | |
617ba13b MC |
774 | le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg); |
775 | unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb); | |
ac27a0ec | 776 | |
617ba13b | 777 | if (!EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG) || |
ac27a0ec | 778 | metagroup < first_meta_bg) |
af5bc92d | 779 | return ext4_bg_num_gdb_nometa(sb, group); |
ac27a0ec | 780 | |
617ba13b | 781 | return ext4_bg_num_gdb_meta(sb,group); |
ac27a0ec DK |
782 | |
783 | } | |
c2ea3fde | 784 | |
49f7f9af | 785 | /* |
d5b8f310 | 786 | * This function returns the number of file system metadata clusters at |
49f7f9af TT |
787 | * the beginning of a block group, including the reserved gdt blocks. |
788 | */ | |
5f163cc7 | 789 | static unsigned ext4_num_base_meta_clusters(struct super_block *sb, |
d5b8f310 | 790 | ext4_group_t block_group) |
49f7f9af TT |
791 | { |
792 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
d5b8f310 | 793 | unsigned num; |
49f7f9af TT |
794 | |
795 | /* Check for superblock and gdt backups in this group */ | |
796 | num = ext4_bg_has_super(sb, block_group); | |
797 | ||
798 | if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) || | |
799 | block_group < le32_to_cpu(sbi->s_es->s_first_meta_bg) * | |
800 | sbi->s_desc_per_block) { | |
801 | if (num) { | |
802 | num += ext4_bg_num_gdb(sb, block_group); | |
803 | num += le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks); | |
804 | } | |
805 | } else { /* For META_BG_BLOCK_GROUPS */ | |
806 | num += ext4_bg_num_gdb(sb, block_group); | |
807 | } | |
d5b8f310 | 808 | return EXT4_NUM_B2C(sbi, num); |
49f7f9af | 809 | } |
f86186b4 ES |
810 | /** |
811 | * ext4_inode_to_goal_block - return a hint for block allocation | |
812 | * @inode: inode for block allocation | |
813 | * | |
814 | * Return the ideal location to start allocating blocks for a | |
815 | * newly created inode. | |
816 | */ | |
817 | ext4_fsblk_t ext4_inode_to_goal_block(struct inode *inode) | |
818 | { | |
819 | struct ext4_inode_info *ei = EXT4_I(inode); | |
820 | ext4_group_t block_group; | |
821 | ext4_grpblk_t colour; | |
822 | int flex_size = ext4_flex_bg_size(EXT4_SB(inode->i_sb)); | |
823 | ext4_fsblk_t bg_start; | |
824 | ext4_fsblk_t last_block; | |
825 | ||
826 | block_group = ei->i_block_group; | |
827 | if (flex_size >= EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME) { | |
828 | /* | |
829 | * If there are at least EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME | |
830 | * block groups per flexgroup, reserve the first block | |
831 | * group for directories and special files. Regular | |
832 | * files will start at the second block group. This | |
833 | * tends to speed up directory access and improves | |
834 | * fsck times. | |
835 | */ | |
836 | block_group &= ~(flex_size-1); | |
837 | if (S_ISREG(inode->i_mode)) | |
838 | block_group++; | |
839 | } | |
840 | bg_start = ext4_group_first_block_no(inode->i_sb, block_group); | |
841 | last_block = ext4_blocks_count(EXT4_SB(inode->i_sb)->s_es) - 1; | |
842 | ||
843 | /* | |
844 | * If we are doing delayed allocation, we don't need take | |
845 | * colour into account. | |
846 | */ | |
847 | if (test_opt(inode->i_sb, DELALLOC)) | |
848 | return bg_start; | |
849 | ||
850 | if (bg_start + EXT4_BLOCKS_PER_GROUP(inode->i_sb) <= last_block) | |
851 | colour = (current->pid % 16) * | |
852 | (EXT4_BLOCKS_PER_GROUP(inode->i_sb) / 16); | |
853 | else | |
854 | colour = (current->pid % 16) * ((last_block - bg_start) / 16); | |
855 | return bg_start + colour; | |
856 | } | |
857 |