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 | */ | |
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 TT |
186 | ext4_fsblk_t start, tmp; |
187 | int flex_bg = 0; | |
bdfb6ff4 | 188 | struct ext4_group_info *grp; |
fd034a84 TT |
189 | |
190 | J_ASSERT_BH(bh, buffer_locked(bh)); | |
191 | ||
192 | /* If checksum is bad mark all blocks used to prevent allocation | |
193 | * essentially implementing a per-group read-only flag. */ | |
feb0ab32 | 194 | if (!ext4_group_desc_csum_verify(sb, block_group, gdp)) { |
bdfb6ff4 | 195 | grp = ext4_get_group_info(sb, block_group); |
e43bb4e6 NJ |
196 | if (!EXT4_MB_GRP_BBITMAP_CORRUPT(grp)) |
197 | percpu_counter_sub(&sbi->s_freeclusters_counter, | |
198 | grp->bb_free); | |
bdfb6ff4 | 199 | set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, &grp->bb_state); |
e43bb4e6 NJ |
200 | if (!EXT4_MB_GRP_IBITMAP_CORRUPT(grp)) { |
201 | int count; | |
202 | count = ext4_free_inodes_count(sb, gdp); | |
203 | percpu_counter_sub(&sbi->s_freeinodes_counter, | |
204 | count); | |
205 | } | |
bdfb6ff4 | 206 | set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, &grp->bb_state); |
aef4885a | 207 | return -EIO; |
717d50e4 | 208 | } |
fd034a84 | 209 | memset(bh->b_data, 0, sb->s_blocksize); |
717d50e4 | 210 | |
d5b8f310 | 211 | bit_max = ext4_num_base_meta_clusters(sb, block_group); |
fd034a84 TT |
212 | for (bit = 0; bit < bit_max; bit++) |
213 | ext4_set_bit(bit, bh->b_data); | |
d00a6d7b | 214 | |
fd034a84 | 215 | start = ext4_group_first_block_no(sb, block_group); |
717d50e4 | 216 | |
fd034a84 TT |
217 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) |
218 | flex_bg = 1; | |
717d50e4 | 219 | |
fd034a84 TT |
220 | /* Set bits for block and inode bitmaps, and inode table */ |
221 | tmp = ext4_block_bitmap(sb, gdp); | |
222 | if (!flex_bg || ext4_block_in_group(sb, tmp, block_group)) | |
d5b8f310 | 223 | ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data); |
717d50e4 | 224 | |
fd034a84 TT |
225 | tmp = ext4_inode_bitmap(sb, gdp); |
226 | if (!flex_bg || ext4_block_in_group(sb, tmp, block_group)) | |
d5b8f310 | 227 | ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data); |
0bf7e837 | 228 | |
fd034a84 TT |
229 | tmp = ext4_inode_table(sb, gdp); |
230 | for (; tmp < ext4_inode_table(sb, gdp) + | |
231 | sbi->s_itb_per_group; tmp++) { | |
0bf7e837 | 232 | if (!flex_bg || ext4_block_in_group(sb, tmp, block_group)) |
d5b8f310 | 233 | ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data); |
717d50e4 | 234 | } |
d5b8f310 | 235 | |
fd034a84 TT |
236 | /* |
237 | * Also if the number of blocks within the group is less than | |
238 | * the blocksize * 8 ( which is the size of bitmap ), set rest | |
239 | * of the block bitmap to 1 | |
240 | */ | |
d5b8f310 | 241 | ext4_mark_bitmap_end(num_clusters_in_group(sb, block_group), |
fd034a84 | 242 | sb->s_blocksize * 8, bh->b_data); |
79f1ba49 | 243 | ext4_block_bitmap_csum_set(sb, block_group, gdp, bh); |
feb0ab32 | 244 | ext4_group_desc_csum_set(sb, block_group, gdp); |
aef4885a | 245 | return 0; |
717d50e4 AD |
246 | } |
247 | ||
fd034a84 TT |
248 | /* Return the number of free blocks in a block group. It is used when |
249 | * the block bitmap is uninitialized, so we can't just count the bits | |
250 | * in the bitmap. */ | |
cff1dfd7 TT |
251 | unsigned ext4_free_clusters_after_init(struct super_block *sb, |
252 | ext4_group_t block_group, | |
253 | struct ext4_group_desc *gdp) | |
fd034a84 | 254 | { |
d5b8f310 TT |
255 | return num_clusters_in_group(sb, block_group) - |
256 | ext4_num_overhead_clusters(sb, block_group, gdp); | |
fd034a84 | 257 | } |
717d50e4 | 258 | |
ac27a0ec DK |
259 | /* |
260 | * The free blocks are managed by bitmaps. A file system contains several | |
261 | * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap | |
262 | * block for inodes, N blocks for the inode table and data blocks. | |
263 | * | |
264 | * The file system contains group descriptors which are located after the | |
265 | * super block. Each descriptor contains the number of the bitmap block and | |
266 | * the free blocks count in the block. The descriptors are loaded in memory | |
e627432c | 267 | * when a file system is mounted (see ext4_fill_super). |
ac27a0ec DK |
268 | */ |
269 | ||
ac27a0ec | 270 | /** |
617ba13b | 271 | * ext4_get_group_desc() -- load group descriptor from disk |
ac27a0ec DK |
272 | * @sb: super block |
273 | * @block_group: given block group | |
274 | * @bh: pointer to the buffer head to store the block | |
275 | * group descriptor | |
276 | */ | |
af5bc92d | 277 | struct ext4_group_desc * ext4_get_group_desc(struct super_block *sb, |
fd2d4291 | 278 | ext4_group_t block_group, |
af5bc92d | 279 | struct buffer_head **bh) |
ac27a0ec | 280 | { |
498e5f24 TT |
281 | unsigned int group_desc; |
282 | unsigned int offset; | |
8df9675f | 283 | ext4_group_t ngroups = ext4_get_groups_count(sb); |
af5bc92d | 284 | struct ext4_group_desc *desc; |
617ba13b | 285 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec | 286 | |
8df9675f | 287 | if (block_group >= ngroups) { |
12062ddd ES |
288 | ext4_error(sb, "block_group >= groups_count - block_group = %u," |
289 | " groups_count = %u", block_group, ngroups); | |
ac27a0ec DK |
290 | |
291 | return NULL; | |
292 | } | |
ac27a0ec | 293 | |
617ba13b MC |
294 | group_desc = block_group >> EXT4_DESC_PER_BLOCK_BITS(sb); |
295 | offset = block_group & (EXT4_DESC_PER_BLOCK(sb) - 1); | |
ac27a0ec | 296 | if (!sbi->s_group_desc[group_desc]) { |
12062ddd | 297 | ext4_error(sb, "Group descriptor not loaded - " |
498e5f24 | 298 | "block_group = %u, group_desc = %u, desc = %u", |
af5bc92d | 299 | block_group, group_desc, offset); |
ac27a0ec DK |
300 | return NULL; |
301 | } | |
302 | ||
0d1ee42f AR |
303 | desc = (struct ext4_group_desc *)( |
304 | (__u8 *)sbi->s_group_desc[group_desc]->b_data + | |
305 | offset * EXT4_DESC_SIZE(sb)); | |
ac27a0ec DK |
306 | if (bh) |
307 | *bh = sbi->s_group_desc[group_desc]; | |
0d1ee42f | 308 | return desc; |
ac27a0ec DK |
309 | } |
310 | ||
7a4c5de2 TT |
311 | /* |
312 | * Return the block number which was discovered to be invalid, or 0 if | |
313 | * the block bitmap is valid. | |
314 | */ | |
315 | static ext4_fsblk_t ext4_valid_block_bitmap(struct super_block *sb, | |
316 | struct ext4_group_desc *desc, | |
dbde0abe | 317 | ext4_group_t block_group, |
7a4c5de2 | 318 | struct buffer_head *bh) |
abcb2947 | 319 | { |
e674e5cb | 320 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
abcb2947 AK |
321 | ext4_grpblk_t offset; |
322 | ext4_grpblk_t next_zero_bit; | |
7a4c5de2 | 323 | ext4_fsblk_t blk; |
abcb2947 AK |
324 | ext4_fsblk_t group_first_block; |
325 | ||
326 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) { | |
327 | /* with FLEX_BG, the inode/block bitmaps and itable | |
328 | * blocks may not be in the group at all | |
329 | * so the bitmap validation will be skipped for those groups | |
330 | * or it has to also read the block group where the bitmaps | |
331 | * are located to verify they are set. | |
332 | */ | |
7a4c5de2 | 333 | return 0; |
abcb2947 AK |
334 | } |
335 | group_first_block = ext4_group_first_block_no(sb, block_group); | |
336 | ||
337 | /* check whether block bitmap block number is set */ | |
7a4c5de2 TT |
338 | blk = ext4_block_bitmap(sb, desc); |
339 | offset = blk - group_first_block; | |
e674e5cb | 340 | if (!ext4_test_bit(EXT4_B2C(sbi, offset), bh->b_data)) |
abcb2947 | 341 | /* bad block bitmap */ |
7a4c5de2 | 342 | return blk; |
abcb2947 AK |
343 | |
344 | /* check whether the inode bitmap block number is set */ | |
7a4c5de2 TT |
345 | blk = ext4_inode_bitmap(sb, desc); |
346 | offset = blk - group_first_block; | |
e674e5cb | 347 | if (!ext4_test_bit(EXT4_B2C(sbi, offset), bh->b_data)) |
abcb2947 | 348 | /* bad block bitmap */ |
7a4c5de2 | 349 | return blk; |
abcb2947 AK |
350 | |
351 | /* check whether the inode table block number is set */ | |
7a4c5de2 TT |
352 | blk = ext4_inode_table(sb, desc); |
353 | offset = blk - group_first_block; | |
abcb2947 | 354 | next_zero_bit = ext4_find_next_zero_bit(bh->b_data, |
e674e5cb DW |
355 | EXT4_B2C(sbi, offset + EXT4_SB(sb)->s_itb_per_group), |
356 | EXT4_B2C(sbi, offset)); | |
357 | if (next_zero_bit < | |
358 | EXT4_B2C(sbi, offset + EXT4_SB(sb)->s_itb_per_group)) | |
7a4c5de2 TT |
359 | /* bad bitmap for inode tables */ |
360 | return blk; | |
abcb2947 AK |
361 | return 0; |
362 | } | |
fa77dcfa | 363 | |
c197855e SH |
364 | static void ext4_validate_block_bitmap(struct super_block *sb, |
365 | struct ext4_group_desc *desc, | |
366 | ext4_group_t block_group, | |
367 | struct buffer_head *bh) | |
fa77dcfa | 368 | { |
7a4c5de2 | 369 | ext4_fsblk_t blk; |
163a203d | 370 | struct ext4_group_info *grp = ext4_get_group_info(sb, block_group); |
e43bb4e6 | 371 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
7a4c5de2 | 372 | |
fa77dcfa DW |
373 | if (buffer_verified(bh)) |
374 | return; | |
375 | ||
376 | ext4_lock_group(sb, block_group); | |
7a4c5de2 TT |
377 | blk = ext4_valid_block_bitmap(sb, desc, block_group, bh); |
378 | if (unlikely(blk != 0)) { | |
379 | ext4_unlock_group(sb, block_group); | |
380 | ext4_error(sb, "bg %u: block %llu: invalid block bitmap", | |
381 | block_group, blk); | |
e43bb4e6 NJ |
382 | if (!EXT4_MB_GRP_BBITMAP_CORRUPT(grp)) |
383 | percpu_counter_sub(&sbi->s_freeclusters_counter, | |
384 | grp->bb_free); | |
163a203d | 385 | set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, &grp->bb_state); |
7a4c5de2 TT |
386 | return; |
387 | } | |
388 | if (unlikely(!ext4_block_bitmap_csum_verify(sb, block_group, | |
79f1ba49 | 389 | desc, bh))) { |
7a4c5de2 TT |
390 | ext4_unlock_group(sb, block_group); |
391 | ext4_error(sb, "bg %u: bad block bitmap checksum", block_group); | |
e43bb4e6 NJ |
392 | if (!EXT4_MB_GRP_BBITMAP_CORRUPT(grp)) |
393 | percpu_counter_sub(&sbi->s_freeclusters_counter, | |
394 | grp->bb_free); | |
163a203d | 395 | set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, &grp->bb_state); |
7a4c5de2 TT |
396 | return; |
397 | } | |
398 | set_buffer_verified(bh); | |
fa77dcfa DW |
399 | ext4_unlock_group(sb, block_group); |
400 | } | |
401 | ||
ac27a0ec | 402 | /** |
15b49132 | 403 | * ext4_read_block_bitmap_nowait() |
ac27a0ec DK |
404 | * @sb: super block |
405 | * @block_group: given block group | |
406 | * | |
abcb2947 AK |
407 | * Read the bitmap for a given block_group,and validate the |
408 | * bits for block/inode/inode tables are set in the bitmaps | |
ac27a0ec DK |
409 | * |
410 | * Return buffer_head on success or NULL in case of failure. | |
411 | */ | |
717d50e4 | 412 | struct buffer_head * |
813e5727 | 413 | ext4_read_block_bitmap_nowait(struct super_block *sb, ext4_group_t block_group) |
ac27a0ec | 414 | { |
af5bc92d | 415 | struct ext4_group_desc *desc; |
813e5727 | 416 | struct buffer_head *bh; |
7c9e69fa | 417 | ext4_fsblk_t bitmap_blk; |
ac27a0ec | 418 | |
717d50e4 | 419 | desc = ext4_get_group_desc(sb, block_group, NULL); |
ac27a0ec | 420 | if (!desc) |
7c9e69fa AK |
421 | return NULL; |
422 | bitmap_blk = ext4_block_bitmap(sb, desc); | |
abcb2947 AK |
423 | bh = sb_getblk(sb, bitmap_blk); |
424 | if (unlikely(!bh)) { | |
813e5727 TT |
425 | ext4_error(sb, "Cannot get buffer for block bitmap - " |
426 | "block_group = %u, block_bitmap = %llu", | |
427 | block_group, bitmap_blk); | |
abcb2947 AK |
428 | return NULL; |
429 | } | |
2ccb5fb9 AK |
430 | |
431 | if (bitmap_uptodate(bh)) | |
fa77dcfa | 432 | goto verify; |
abcb2947 | 433 | |
c806e68f | 434 | lock_buffer(bh); |
2ccb5fb9 AK |
435 | if (bitmap_uptodate(bh)) { |
436 | unlock_buffer(bh); | |
fa77dcfa | 437 | goto verify; |
2ccb5fb9 | 438 | } |
955ce5f5 | 439 | ext4_lock_group(sb, block_group); |
717d50e4 | 440 | if (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) { |
aef4885a DM |
441 | int err; |
442 | ||
443 | err = ext4_init_block_bitmap(sb, bh, block_group, desc); | |
2ccb5fb9 | 444 | set_bitmap_uptodate(bh); |
abcb2947 | 445 | set_buffer_uptodate(bh); |
955ce5f5 | 446 | ext4_unlock_group(sb, block_group); |
3300beda | 447 | unlock_buffer(bh); |
aef4885a DM |
448 | if (err) |
449 | ext4_error(sb, "Checksum bad for grp %u", block_group); | |
abcb2947 | 450 | return bh; |
717d50e4 | 451 | } |
955ce5f5 | 452 | ext4_unlock_group(sb, block_group); |
2ccb5fb9 AK |
453 | if (buffer_uptodate(bh)) { |
454 | /* | |
455 | * if not uninit if bh is uptodate, | |
456 | * bitmap is also uptodate | |
457 | */ | |
458 | set_bitmap_uptodate(bh); | |
459 | unlock_buffer(bh); | |
fa77dcfa | 460 | goto verify; |
2ccb5fb9 AK |
461 | } |
462 | /* | |
813e5727 | 463 | * submit the buffer_head for reading |
2ccb5fb9 | 464 | */ |
813e5727 | 465 | set_buffer_new(bh); |
0562e0ba | 466 | trace_ext4_read_block_bitmap_load(sb, block_group); |
813e5727 TT |
467 | bh->b_end_io = ext4_end_bitmap_read; |
468 | get_bh(bh); | |
9f203507 | 469 | submit_bh(READ | REQ_META | REQ_PRIO, bh); |
813e5727 | 470 | return bh; |
fa77dcfa DW |
471 | verify: |
472 | ext4_validate_block_bitmap(sb, desc, block_group, bh); | |
48d9eb97 DW |
473 | if (buffer_verified(bh)) |
474 | return bh; | |
475 | put_bh(bh); | |
476 | return NULL; | |
813e5727 TT |
477 | } |
478 | ||
479 | /* Returns 0 on success, 1 on error */ | |
480 | int ext4_wait_block_bitmap(struct super_block *sb, ext4_group_t block_group, | |
481 | struct buffer_head *bh) | |
482 | { | |
483 | struct ext4_group_desc *desc; | |
484 | ||
485 | if (!buffer_new(bh)) | |
486 | return 0; | |
487 | desc = ext4_get_group_desc(sb, block_group, NULL); | |
488 | if (!desc) | |
489 | return 1; | |
490 | wait_on_buffer(bh); | |
491 | if (!buffer_uptodate(bh)) { | |
12062ddd | 492 | ext4_error(sb, "Cannot read block bitmap - " |
813e5727 | 493 | "block_group = %u, block_bitmap = %llu", |
d4dc462f | 494 | block_group, (unsigned long long) bh->b_blocknr); |
813e5727 | 495 | return 1; |
abcb2947 | 496 | } |
813e5727 TT |
497 | clear_buffer_new(bh); |
498 | /* Panic or remount fs read-only if block bitmap is invalid */ | |
fa77dcfa | 499 | ext4_validate_block_bitmap(sb, desc, block_group, bh); |
48d9eb97 DW |
500 | /* ...but check for error just in case errors=continue. */ |
501 | return !buffer_verified(bh); | |
813e5727 TT |
502 | } |
503 | ||
504 | struct buffer_head * | |
505 | ext4_read_block_bitmap(struct super_block *sb, ext4_group_t block_group) | |
506 | { | |
507 | struct buffer_head *bh; | |
508 | ||
509 | bh = ext4_read_block_bitmap_nowait(sb, block_group); | |
15b49132 EG |
510 | if (!bh) |
511 | return NULL; | |
813e5727 TT |
512 | if (ext4_wait_block_bitmap(sb, block_group, bh)) { |
513 | put_bh(bh); | |
514 | return NULL; | |
515 | } | |
ac27a0ec DK |
516 | return bh; |
517 | } | |
ac27a0ec | 518 | |
8c3bf8a0 | 519 | /** |
df55c99d | 520 | * ext4_has_free_clusters() |
8c3bf8a0 | 521 | * @sbi: in-core super block structure. |
df55c99d TT |
522 | * @nclusters: number of needed blocks |
523 | * @flags: flags from ext4_mb_new_blocks() | |
8c3bf8a0 | 524 | * |
df55c99d | 525 | * Check if filesystem has nclusters free & available for allocation. |
8c3bf8a0 ES |
526 | * On success return 1, return 0 on failure. |
527 | */ | |
df55c99d TT |
528 | static int ext4_has_free_clusters(struct ext4_sb_info *sbi, |
529 | s64 nclusters, unsigned int flags) | |
a30d542a | 530 | { |
27dd4385 | 531 | s64 free_clusters, dirty_clusters, rsv, resv_clusters; |
57042651 | 532 | struct percpu_counter *fcc = &sbi->s_freeclusters_counter; |
df55c99d | 533 | struct percpu_counter *dcc = &sbi->s_dirtyclusters_counter; |
a30d542a | 534 | |
df55c99d TT |
535 | free_clusters = percpu_counter_read_positive(fcc); |
536 | dirty_clusters = percpu_counter_read_positive(dcc); | |
27dd4385 | 537 | resv_clusters = atomic64_read(&sbi->s_resv_clusters); |
304e220f LC |
538 | |
539 | /* | |
540 | * r_blocks_count should always be multiple of the cluster ratio so | |
541 | * we are safe to do a plane bit shift only. | |
542 | */ | |
27dd4385 LC |
543 | rsv = (ext4_r_blocks_count(sbi->s_es) >> sbi->s_cluster_bits) + |
544 | resv_clusters; | |
a30d542a | 545 | |
27dd4385 | 546 | if (free_clusters - (nclusters + rsv + dirty_clusters) < |
df55c99d | 547 | EXT4_FREECLUSTERS_WATERMARK) { |
304e220f | 548 | free_clusters = percpu_counter_sum_positive(fcc); |
df55c99d | 549 | dirty_clusters = percpu_counter_sum_positive(dcc); |
6bc6e63f | 550 | } |
df55c99d TT |
551 | /* Check whether we have space after accounting for current |
552 | * dirty clusters & root reserved clusters. | |
6bc6e63f | 553 | */ |
27dd4385 | 554 | if (free_clusters >= (rsv + nclusters + dirty_clusters)) |
a996031c | 555 | return 1; |
a30d542a | 556 | |
df55c99d | 557 | /* Hm, nope. Are (enough) root reserved clusters available? */ |
08cefc7a EB |
558 | if (uid_eq(sbi->s_resuid, current_fsuid()) || |
559 | (!gid_eq(sbi->s_resgid, GLOBAL_ROOT_GID) && in_group_p(sbi->s_resgid)) || | |
55f020db | 560 | capable(CAP_SYS_RESOURCE) || |
27dd4385 | 561 | (flags & EXT4_MB_USE_ROOT_BLOCKS)) { |
55f020db | 562 | |
27dd4385 LC |
563 | if (free_clusters >= (nclusters + dirty_clusters + |
564 | resv_clusters)) | |
565 | return 1; | |
566 | } | |
567 | /* No free blocks. Let's see if we can dip into reserved pool */ | |
568 | if (flags & EXT4_MB_USE_RESERVED) { | |
df55c99d | 569 | if (free_clusters >= (nclusters + dirty_clusters)) |
a996031c ES |
570 | return 1; |
571 | } | |
572 | ||
573 | return 0; | |
a30d542a AK |
574 | } |
575 | ||
e7d5f315 TT |
576 | int ext4_claim_free_clusters(struct ext4_sb_info *sbi, |
577 | s64 nclusters, unsigned int flags) | |
ac27a0ec | 578 | { |
df55c99d | 579 | if (ext4_has_free_clusters(sbi, nclusters, flags)) { |
e7d5f315 | 580 | percpu_counter_add(&sbi->s_dirtyclusters_counter, nclusters); |
16eb7295 | 581 | return 0; |
8c3bf8a0 ES |
582 | } else |
583 | return -ENOSPC; | |
a30d542a | 584 | } |
07031431 | 585 | |
ac27a0ec | 586 | /** |
617ba13b | 587 | * ext4_should_retry_alloc() |
ac27a0ec DK |
588 | * @sb: super block |
589 | * @retries number of attemps has been made | |
590 | * | |
617ba13b | 591 | * ext4_should_retry_alloc() is called when ENOSPC is returned, and if |
ac27a0ec | 592 | * it is profitable to retry the operation, this function will wait |
25985edc | 593 | * for the current or committing transaction to complete, and then |
ac27a0ec DK |
594 | * return TRUE. |
595 | * | |
596 | * if the total number of retries exceed three times, return FALSE. | |
597 | */ | |
617ba13b | 598 | int ext4_should_retry_alloc(struct super_block *sb, int *retries) |
ac27a0ec | 599 | { |
df55c99d | 600 | if (!ext4_has_free_clusters(EXT4_SB(sb), 1, 0) || |
8f64b32e ES |
601 | (*retries)++ > 3 || |
602 | !EXT4_SB(sb)->s_journal) | |
ac27a0ec DK |
603 | return 0; |
604 | ||
605 | jbd_debug(1, "%s: retrying operation after ENOSPC\n", sb->s_id); | |
606 | ||
dab291af | 607 | return jbd2_journal_force_commit_nested(EXT4_SB(sb)->s_journal); |
ac27a0ec DK |
608 | } |
609 | ||
654b4908 | 610 | /* |
d2a17637 | 611 | * ext4_new_meta_blocks() -- allocate block for meta data (indexing) blocks |
654b4908 AK |
612 | * |
613 | * @handle: handle to this transaction | |
614 | * @inode: file inode | |
615 | * @goal: given target block(filesystem wide) | |
7b415bf6 | 616 | * @count: pointer to total number of clusters needed |
654b4908 AK |
617 | * @errp: error code |
618 | * | |
97df5d15 | 619 | * Return 1st allocated block number on success, *count stores total account |
d2a17637 | 620 | * error stores in errp pointer |
654b4908 | 621 | */ |
d2a17637 | 622 | ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode, |
55f020db AH |
623 | ext4_fsblk_t goal, unsigned int flags, |
624 | unsigned long *count, int *errp) | |
654b4908 | 625 | { |
97df5d15 | 626 | struct ext4_allocation_request ar; |
d2a17637 | 627 | ext4_fsblk_t ret; |
97df5d15 TT |
628 | |
629 | memset(&ar, 0, sizeof(ar)); | |
630 | /* Fill with neighbour allocated blocks */ | |
631 | ar.inode = inode; | |
632 | ar.goal = goal; | |
633 | ar.len = count ? *count : 1; | |
55f020db | 634 | ar.flags = flags; |
97df5d15 TT |
635 | |
636 | ret = ext4_mb_new_blocks(handle, &ar, errp); | |
637 | if (count) | |
638 | *count = ar.len; | |
d2a17637 | 639 | /* |
72b8ab9d ES |
640 | * Account for the allocated meta blocks. We will never |
641 | * fail EDQUOT for metdata, but we do account for it. | |
d2a17637 | 642 | */ |
e3cf5d5d | 643 | if (!(*errp) && (flags & EXT4_MB_DELALLOC_RESERVED)) { |
d2a17637 | 644 | spin_lock(&EXT4_I(inode)->i_block_reservation_lock); |
d2a17637 | 645 | spin_unlock(&EXT4_I(inode)->i_block_reservation_lock); |
7b415bf6 AK |
646 | dquot_alloc_block_nofail(inode, |
647 | EXT4_C2B(EXT4_SB(inode->i_sb), ar.len)); | |
d2a17637 MC |
648 | } |
649 | return ret; | |
654b4908 AK |
650 | } |
651 | ||
ac27a0ec | 652 | /** |
5dee5437 | 653 | * ext4_count_free_clusters() -- count filesystem free clusters |
ac27a0ec DK |
654 | * @sb: superblock |
655 | * | |
5dee5437 | 656 | * Adds up the number of free clusters from each block group. |
ac27a0ec | 657 | */ |
5dee5437 | 658 | ext4_fsblk_t ext4_count_free_clusters(struct super_block *sb) |
ac27a0ec | 659 | { |
617ba13b MC |
660 | ext4_fsblk_t desc_count; |
661 | struct ext4_group_desc *gdp; | |
fd2d4291 | 662 | ext4_group_t i; |
8df9675f | 663 | ext4_group_t ngroups = ext4_get_groups_count(sb); |
2746f7a1 | 664 | struct ext4_group_info *grp; |
617ba13b MC |
665 | #ifdef EXT4FS_DEBUG |
666 | struct ext4_super_block *es; | |
667 | ext4_fsblk_t bitmap_count; | |
498e5f24 | 668 | unsigned int x; |
ac27a0ec DK |
669 | struct buffer_head *bitmap_bh = NULL; |
670 | ||
617ba13b | 671 | es = EXT4_SB(sb)->s_es; |
ac27a0ec DK |
672 | desc_count = 0; |
673 | bitmap_count = 0; | |
674 | gdp = NULL; | |
675 | ||
ac27a0ec | 676 | for (i = 0; i < ngroups; i++) { |
617ba13b | 677 | gdp = ext4_get_group_desc(sb, i, NULL); |
ac27a0ec DK |
678 | if (!gdp) |
679 | continue; | |
2746f7a1 DW |
680 | grp = NULL; |
681 | if (EXT4_SB(sb)->s_group_info) | |
682 | grp = ext4_get_group_info(sb, i); | |
683 | if (!grp || !EXT4_MB_GRP_BBITMAP_CORRUPT(grp)) | |
684 | desc_count += ext4_free_group_clusters(sb, gdp); | |
ac27a0ec | 685 | brelse(bitmap_bh); |
574ca174 | 686 | bitmap_bh = ext4_read_block_bitmap(sb, i); |
ac27a0ec DK |
687 | if (bitmap_bh == NULL) |
688 | continue; | |
689 | ||
f6fb99ca | 690 | x = ext4_count_free(bitmap_bh->b_data, |
036acea2 | 691 | EXT4_CLUSTERS_PER_GROUP(sb) / 8); |
9fd9784c | 692 | printk(KERN_DEBUG "group %u: stored = %d, counted = %u\n", |
021b65bb | 693 | i, ext4_free_group_clusters(sb, gdp), x); |
ac27a0ec DK |
694 | bitmap_count += x; |
695 | } | |
696 | brelse(bitmap_bh); | |
5dee5437 TT |
697 | printk(KERN_DEBUG "ext4_count_free_clusters: stored = %llu" |
698 | ", computed = %llu, %llu\n", | |
810da240 | 699 | EXT4_NUM_B2C(EXT4_SB(sb), ext4_free_blocks_count(es)), |
4776004f | 700 | desc_count, bitmap_count); |
ac27a0ec DK |
701 | return bitmap_count; |
702 | #else | |
703 | desc_count = 0; | |
ac27a0ec | 704 | for (i = 0; i < ngroups; i++) { |
617ba13b | 705 | gdp = ext4_get_group_desc(sb, i, NULL); |
ac27a0ec DK |
706 | if (!gdp) |
707 | continue; | |
2746f7a1 DW |
708 | grp = NULL; |
709 | if (EXT4_SB(sb)->s_group_info) | |
710 | grp = ext4_get_group_info(sb, i); | |
711 | if (!grp || !EXT4_MB_GRP_BBITMAP_CORRUPT(grp)) | |
712 | desc_count += ext4_free_group_clusters(sb, gdp); | |
ac27a0ec DK |
713 | } |
714 | ||
715 | return desc_count; | |
716 | #endif | |
717 | } | |
718 | ||
fd2d4291 | 719 | static inline int test_root(ext4_group_t a, int b) |
ac27a0ec | 720 | { |
f4afb4f4 TT |
721 | while (1) { |
722 | if (a < b) | |
723 | return 0; | |
724 | if (a == b) | |
725 | return 1; | |
726 | if ((a % b) != 0) | |
727 | return 0; | |
728 | a = a / b; | |
729 | } | |
ac27a0ec DK |
730 | } |
731 | ||
ac27a0ec | 732 | /** |
617ba13b | 733 | * ext4_bg_has_super - number of blocks used by the superblock in group |
ac27a0ec DK |
734 | * @sb: superblock for filesystem |
735 | * @group: group number to check | |
736 | * | |
737 | * Return the number of blocks used by the superblock (primary or backup) | |
738 | * in this group. Currently this will be only 0 or 1. | |
739 | */ | |
fd2d4291 | 740 | int ext4_bg_has_super(struct super_block *sb, ext4_group_t group) |
ac27a0ec | 741 | { |
1beeef1b DW |
742 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; |
743 | ||
744 | if (group == 0) | |
745 | return 1; | |
746 | if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_SPARSE_SUPER2)) { | |
747 | if (group == le32_to_cpu(es->s_backup_bgs[0]) || | |
748 | group == le32_to_cpu(es->s_backup_bgs[1])) | |
749 | return 1; | |
750 | return 0; | |
751 | } | |
752 | if ((group <= 1) || !EXT4_HAS_RO_COMPAT_FEATURE(sb, | |
753 | EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER)) | |
754 | return 1; | |
755 | if (!(group & 1)) | |
ac27a0ec | 756 | return 0; |
1beeef1b DW |
757 | if (test_root(group, 3) || (test_root(group, 5)) || |
758 | test_root(group, 7)) | |
759 | return 1; | |
760 | ||
761 | return 0; | |
ac27a0ec DK |
762 | } |
763 | ||
fd2d4291 AM |
764 | static unsigned long ext4_bg_num_gdb_meta(struct super_block *sb, |
765 | ext4_group_t group) | |
ac27a0ec | 766 | { |
617ba13b | 767 | unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb); |
fd2d4291 AM |
768 | ext4_group_t first = metagroup * EXT4_DESC_PER_BLOCK(sb); |
769 | ext4_group_t last = first + EXT4_DESC_PER_BLOCK(sb) - 1; | |
ac27a0ec DK |
770 | |
771 | if (group == first || group == first + 1 || group == last) | |
772 | return 1; | |
773 | return 0; | |
774 | } | |
775 | ||
fd2d4291 AM |
776 | static unsigned long ext4_bg_num_gdb_nometa(struct super_block *sb, |
777 | ext4_group_t group) | |
ac27a0ec | 778 | { |
8dadb198 TT |
779 | if (!ext4_bg_has_super(sb, group)) |
780 | return 0; | |
781 | ||
782 | if (EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG)) | |
783 | return le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg); | |
784 | else | |
785 | return EXT4_SB(sb)->s_gdb_count; | |
ac27a0ec DK |
786 | } |
787 | ||
788 | /** | |
617ba13b | 789 | * ext4_bg_num_gdb - number of blocks used by the group table in group |
ac27a0ec DK |
790 | * @sb: superblock for filesystem |
791 | * @group: group number to check | |
792 | * | |
793 | * Return the number of blocks used by the group descriptor table | |
794 | * (primary or backup) in this group. In the future there may be a | |
795 | * different number of descriptor blocks in each group. | |
796 | */ | |
fd2d4291 | 797 | unsigned long ext4_bg_num_gdb(struct super_block *sb, ext4_group_t group) |
ac27a0ec DK |
798 | { |
799 | unsigned long first_meta_bg = | |
617ba13b MC |
800 | le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg); |
801 | unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb); | |
ac27a0ec | 802 | |
617ba13b | 803 | if (!EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG) || |
ac27a0ec | 804 | metagroup < first_meta_bg) |
af5bc92d | 805 | return ext4_bg_num_gdb_nometa(sb, group); |
ac27a0ec | 806 | |
617ba13b | 807 | return ext4_bg_num_gdb_meta(sb,group); |
ac27a0ec DK |
808 | |
809 | } | |
c2ea3fde | 810 | |
49f7f9af | 811 | /* |
d5b8f310 | 812 | * This function returns the number of file system metadata clusters at |
49f7f9af TT |
813 | * the beginning of a block group, including the reserved gdt blocks. |
814 | */ | |
5f163cc7 | 815 | static unsigned ext4_num_base_meta_clusters(struct super_block *sb, |
d5b8f310 | 816 | ext4_group_t block_group) |
49f7f9af TT |
817 | { |
818 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
d5b8f310 | 819 | unsigned num; |
49f7f9af TT |
820 | |
821 | /* Check for superblock and gdt backups in this group */ | |
822 | num = ext4_bg_has_super(sb, block_group); | |
823 | ||
824 | if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) || | |
825 | block_group < le32_to_cpu(sbi->s_es->s_first_meta_bg) * | |
826 | sbi->s_desc_per_block) { | |
827 | if (num) { | |
828 | num += ext4_bg_num_gdb(sb, block_group); | |
829 | num += le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks); | |
830 | } | |
831 | } else { /* For META_BG_BLOCK_GROUPS */ | |
832 | num += ext4_bg_num_gdb(sb, block_group); | |
833 | } | |
d5b8f310 | 834 | return EXT4_NUM_B2C(sbi, num); |
49f7f9af | 835 | } |
f86186b4 ES |
836 | /** |
837 | * ext4_inode_to_goal_block - return a hint for block allocation | |
838 | * @inode: inode for block allocation | |
839 | * | |
840 | * Return the ideal location to start allocating blocks for a | |
841 | * newly created inode. | |
842 | */ | |
843 | ext4_fsblk_t ext4_inode_to_goal_block(struct inode *inode) | |
844 | { | |
845 | struct ext4_inode_info *ei = EXT4_I(inode); | |
846 | ext4_group_t block_group; | |
847 | ext4_grpblk_t colour; | |
848 | int flex_size = ext4_flex_bg_size(EXT4_SB(inode->i_sb)); | |
849 | ext4_fsblk_t bg_start; | |
850 | ext4_fsblk_t last_block; | |
851 | ||
852 | block_group = ei->i_block_group; | |
853 | if (flex_size >= EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME) { | |
854 | /* | |
855 | * If there are at least EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME | |
856 | * block groups per flexgroup, reserve the first block | |
857 | * group for directories and special files. Regular | |
858 | * files will start at the second block group. This | |
859 | * tends to speed up directory access and improves | |
860 | * fsck times. | |
861 | */ | |
862 | block_group &= ~(flex_size-1); | |
863 | if (S_ISREG(inode->i_mode)) | |
864 | block_group++; | |
865 | } | |
866 | bg_start = ext4_group_first_block_no(inode->i_sb, block_group); | |
867 | last_block = ext4_blocks_count(EXT4_SB(inode->i_sb)->s_es) - 1; | |
868 | ||
869 | /* | |
870 | * If we are doing delayed allocation, we don't need take | |
871 | * colour into account. | |
872 | */ | |
873 | if (test_opt(inode->i_sb, DELALLOC)) | |
874 | return bg_start; | |
875 | ||
876 | if (bg_start + EXT4_BLOCKS_PER_GROUP(inode->i_sb) <= last_block) | |
877 | colour = (current->pid % 16) * | |
878 | (EXT4_BLOCKS_PER_GROUP(inode->i_sb) / 16); | |
879 | else | |
880 | colour = (current->pid % 16) * ((last_block - bg_start) / 16); | |
881 | return bg_start + colour; | |
882 | } | |
883 |