]>
Commit | Line | Data |
---|---|---|
ac27a0ec | 1 | /* |
617ba13b | 2 | * linux/fs/ext4/balloc.c |
ac27a0ec DK |
3 | * |
4 | * Copyright (C) 1992, 1993, 1994, 1995 | |
5 | * Remy Card (card@masi.ibp.fr) | |
6 | * Laboratoire MASI - Institut Blaise Pascal | |
7 | * Universite Pierre et Marie Curie (Paris VI) | |
8 | * | |
9 | * Enhanced block allocation by Stephen Tweedie (sct@redhat.com), 1993 | |
10 | * Big-endian to little-endian byte-swapping/bitmaps by | |
11 | * David S. Miller (davem@caip.rutgers.edu), 1995 | |
12 | */ | |
13 | ||
14 | #include <linux/time.h> | |
15 | #include <linux/capability.h> | |
16 | #include <linux/fs.h> | |
dab291af | 17 | #include <linux/jbd2.h> |
ac27a0ec DK |
18 | #include <linux/quotaops.h> |
19 | #include <linux/buffer_head.h> | |
3dcf5451 CH |
20 | #include "ext4.h" |
21 | #include "ext4_jbd2.h" | |
e21675d4 | 22 | #include "mballoc.h" |
3dcf5451 | 23 | |
0562e0ba JZ |
24 | #include <trace/events/ext4.h> |
25 | ||
ac27a0ec DK |
26 | /* |
27 | * balloc.c contains the blocks allocation and deallocation routines | |
28 | */ | |
29 | ||
72b64b59 | 30 | /* |
3212a80a TT |
31 | * Calculate the block group number and offset into the block/cluster |
32 | * allocation bitmap, given a block number | |
72b64b59 AM |
33 | */ |
34 | void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr, | |
fd2d4291 | 35 | ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp) |
72b64b59 | 36 | { |
8c55e204 | 37 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; |
72b64b59 AM |
38 | ext4_grpblk_t offset; |
39 | ||
8c55e204 | 40 | blocknr = blocknr - le32_to_cpu(es->s_first_data_block); |
3212a80a TT |
41 | offset = do_div(blocknr, EXT4_BLOCKS_PER_GROUP(sb)) >> |
42 | EXT4_SB(sb)->s_cluster_bits; | |
72b64b59 AM |
43 | if (offsetp) |
44 | *offsetp = offset; | |
45 | if (blockgrpp) | |
8c55e204 | 46 | *blockgrpp = blocknr; |
72b64b59 AM |
47 | |
48 | } | |
49 | ||
0bf7e837 JS |
50 | static int ext4_block_in_group(struct super_block *sb, ext4_fsblk_t block, |
51 | ext4_group_t block_group) | |
52 | { | |
53 | ext4_group_t actual_group; | |
7477827f | 54 | ext4_get_group_no_and_offset(sb, block, &actual_group, NULL); |
0bf7e837 JS |
55 | if (actual_group == block_group) |
56 | return 1; | |
57 | return 0; | |
58 | } | |
59 | ||
d5b8f310 TT |
60 | /* Return the number of clusters used for file system metadata; this |
61 | * represents the overhead needed by the file system. | |
62 | */ | |
63 | unsigned ext4_num_overhead_clusters(struct super_block *sb, | |
64 | ext4_group_t block_group, | |
65 | struct ext4_group_desc *gdp) | |
0bf7e837 | 66 | { |
d5b8f310 TT |
67 | unsigned num_clusters; |
68 | int block_cluster = -1, inode_cluster = -1, itbl_cluster = -1, i, c; | |
69 | ext4_fsblk_t start = ext4_group_first_block_no(sb, block_group); | |
70 | ext4_fsblk_t itbl_blk; | |
0bf7e837 | 71 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
0bf7e837 | 72 | |
d5b8f310 TT |
73 | /* This is the number of clusters used by the superblock, |
74 | * block group descriptors, and reserved block group | |
75 | * descriptor blocks */ | |
76 | num_clusters = ext4_num_base_meta_clusters(sb, block_group); | |
77 | ||
78 | /* | |
79 | * For the allocation bitmaps and inode table, we first need | |
80 | * to check to see if the block is in the block group. If it | |
81 | * is, then check to see if the cluster is already accounted | |
82 | * for in the clusters used for the base metadata cluster, or | |
83 | * if we can increment the base metadata cluster to include | |
84 | * that block. Otherwise, we will have to track the cluster | |
85 | * used for the allocation bitmap or inode table explicitly. | |
86 | * Normally all of these blocks are contiguous, so the special | |
87 | * case handling shouldn't be necessary except for *very* | |
88 | * unusual file system layouts. | |
89 | */ | |
90 | if (ext4_block_in_group(sb, ext4_block_bitmap(sb, gdp), block_group)) { | |
91 | block_cluster = EXT4_B2C(sbi, (start - | |
92 | ext4_block_bitmap(sb, gdp))); | |
93 | if (block_cluster < num_clusters) | |
94 | block_cluster = -1; | |
95 | else if (block_cluster == num_clusters) { | |
96 | num_clusters++; | |
97 | block_cluster = -1; | |
98 | } | |
99 | } | |
100 | ||
101 | if (ext4_block_in_group(sb, ext4_inode_bitmap(sb, gdp), block_group)) { | |
102 | inode_cluster = EXT4_B2C(sbi, | |
103 | start - ext4_inode_bitmap(sb, gdp)); | |
104 | if (inode_cluster < num_clusters) | |
105 | inode_cluster = -1; | |
106 | else if (inode_cluster == num_clusters) { | |
107 | num_clusters++; | |
108 | inode_cluster = -1; | |
109 | } | |
110 | } | |
111 | ||
112 | itbl_blk = ext4_inode_table(sb, gdp); | |
113 | for (i = 0; i < sbi->s_itb_per_group; i++) { | |
114 | if (ext4_block_in_group(sb, itbl_blk + i, block_group)) { | |
115 | c = EXT4_B2C(sbi, start - itbl_blk + i); | |
116 | if ((c < num_clusters) || (c == inode_cluster) || | |
117 | (c == block_cluster) || (c == itbl_cluster)) | |
118 | continue; | |
119 | if (c == num_clusters) { | |
120 | num_clusters++; | |
121 | continue; | |
122 | } | |
123 | num_clusters++; | |
124 | itbl_cluster = c; | |
0bf7e837 JS |
125 | } |
126 | } | |
d5b8f310 TT |
127 | |
128 | if (block_cluster != -1) | |
129 | num_clusters++; | |
130 | if (inode_cluster != -1) | |
131 | num_clusters++; | |
132 | ||
133 | return num_clusters; | |
0bf7e837 | 134 | } |
c2ea3fde | 135 | |
d5b8f310 TT |
136 | static unsigned int num_clusters_in_group(struct super_block *sb, |
137 | ext4_group_t block_group) | |
49f7f9af | 138 | { |
d5b8f310 TT |
139 | unsigned int blocks; |
140 | ||
49f7f9af TT |
141 | if (block_group == ext4_get_groups_count(sb) - 1) { |
142 | /* | |
143 | * Even though mke2fs always initializes the first and | |
144 | * last group, just in case some other tool was used, | |
145 | * we need to make sure we calculate the right free | |
146 | * blocks. | |
147 | */ | |
d5b8f310 | 148 | blocks = ext4_blocks_count(EXT4_SB(sb)->s_es) - |
49f7f9af TT |
149 | ext4_group_first_block_no(sb, block_group); |
150 | } else | |
d5b8f310 TT |
151 | blocks = EXT4_BLOCKS_PER_GROUP(sb); |
152 | return EXT4_NUM_B2C(EXT4_SB(sb), blocks); | |
49f7f9af TT |
153 | } |
154 | ||
fd034a84 TT |
155 | /* Initializes an uninitialized block bitmap */ |
156 | void ext4_init_block_bitmap(struct super_block *sb, struct buffer_head *bh, | |
157 | ext4_group_t block_group, | |
158 | struct ext4_group_desc *gdp) | |
717d50e4 | 159 | { |
d5b8f310 | 160 | unsigned int bit, bit_max; |
717d50e4 | 161 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
fd034a84 TT |
162 | ext4_fsblk_t start, tmp; |
163 | int flex_bg = 0; | |
164 | ||
165 | J_ASSERT_BH(bh, buffer_locked(bh)); | |
166 | ||
167 | /* If checksum is bad mark all blocks used to prevent allocation | |
168 | * essentially implementing a per-group read-only flag. */ | |
169 | if (!ext4_group_desc_csum_verify(sbi, block_group, gdp)) { | |
170 | ext4_error(sb, "Checksum bad for group %u", block_group); | |
021b65bb | 171 | ext4_free_group_clusters_set(sb, gdp, 0); |
fd034a84 TT |
172 | ext4_free_inodes_set(sb, gdp, 0); |
173 | ext4_itable_unused_set(sb, gdp, 0); | |
174 | memset(bh->b_data, 0xff, sb->s_blocksize); | |
175 | return; | |
717d50e4 | 176 | } |
fd034a84 | 177 | memset(bh->b_data, 0, sb->s_blocksize); |
717d50e4 | 178 | |
d5b8f310 | 179 | bit_max = ext4_num_base_meta_clusters(sb, block_group); |
fd034a84 TT |
180 | for (bit = 0; bit < bit_max; bit++) |
181 | ext4_set_bit(bit, bh->b_data); | |
d00a6d7b | 182 | |
fd034a84 | 183 | start = ext4_group_first_block_no(sb, block_group); |
717d50e4 | 184 | |
fd034a84 TT |
185 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) |
186 | flex_bg = 1; | |
717d50e4 | 187 | |
fd034a84 TT |
188 | /* Set bits for block and inode bitmaps, and inode table */ |
189 | tmp = ext4_block_bitmap(sb, gdp); | |
190 | if (!flex_bg || ext4_block_in_group(sb, tmp, block_group)) | |
d5b8f310 | 191 | ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data); |
717d50e4 | 192 | |
fd034a84 TT |
193 | tmp = ext4_inode_bitmap(sb, gdp); |
194 | if (!flex_bg || ext4_block_in_group(sb, tmp, block_group)) | |
d5b8f310 | 195 | ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data); |
0bf7e837 | 196 | |
fd034a84 TT |
197 | tmp = ext4_inode_table(sb, gdp); |
198 | for (; tmp < ext4_inode_table(sb, gdp) + | |
199 | sbi->s_itb_per_group; tmp++) { | |
0bf7e837 | 200 | if (!flex_bg || ext4_block_in_group(sb, tmp, block_group)) |
d5b8f310 | 201 | ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data); |
717d50e4 | 202 | } |
d5b8f310 | 203 | |
fd034a84 TT |
204 | /* |
205 | * Also if the number of blocks within the group is less than | |
206 | * the blocksize * 8 ( which is the size of bitmap ), set rest | |
207 | * of the block bitmap to 1 | |
208 | */ | |
d5b8f310 | 209 | ext4_mark_bitmap_end(num_clusters_in_group(sb, block_group), |
fd034a84 | 210 | sb->s_blocksize * 8, bh->b_data); |
717d50e4 AD |
211 | } |
212 | ||
fd034a84 TT |
213 | /* Return the number of free blocks in a block group. It is used when |
214 | * the block bitmap is uninitialized, so we can't just count the bits | |
215 | * in the bitmap. */ | |
cff1dfd7 TT |
216 | unsigned ext4_free_clusters_after_init(struct super_block *sb, |
217 | ext4_group_t block_group, | |
218 | struct ext4_group_desc *gdp) | |
fd034a84 | 219 | { |
d5b8f310 TT |
220 | return num_clusters_in_group(sb, block_group) - |
221 | ext4_num_overhead_clusters(sb, block_group, gdp); | |
fd034a84 | 222 | } |
717d50e4 | 223 | |
ac27a0ec DK |
224 | /* |
225 | * The free blocks are managed by bitmaps. A file system contains several | |
226 | * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap | |
227 | * block for inodes, N blocks for the inode table and data blocks. | |
228 | * | |
229 | * The file system contains group descriptors which are located after the | |
230 | * super block. Each descriptor contains the number of the bitmap block and | |
231 | * the free blocks count in the block. The descriptors are loaded in memory | |
e627432c | 232 | * when a file system is mounted (see ext4_fill_super). |
ac27a0ec DK |
233 | */ |
234 | ||
ac27a0ec | 235 | /** |
617ba13b | 236 | * ext4_get_group_desc() -- load group descriptor from disk |
ac27a0ec DK |
237 | * @sb: super block |
238 | * @block_group: given block group | |
239 | * @bh: pointer to the buffer head to store the block | |
240 | * group descriptor | |
241 | */ | |
af5bc92d | 242 | struct ext4_group_desc * ext4_get_group_desc(struct super_block *sb, |
fd2d4291 | 243 | ext4_group_t block_group, |
af5bc92d | 244 | struct buffer_head **bh) |
ac27a0ec | 245 | { |
498e5f24 TT |
246 | unsigned int group_desc; |
247 | unsigned int offset; | |
8df9675f | 248 | ext4_group_t ngroups = ext4_get_groups_count(sb); |
af5bc92d | 249 | struct ext4_group_desc *desc; |
617ba13b | 250 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec | 251 | |
8df9675f | 252 | if (block_group >= ngroups) { |
12062ddd ES |
253 | ext4_error(sb, "block_group >= groups_count - block_group = %u," |
254 | " groups_count = %u", block_group, ngroups); | |
ac27a0ec DK |
255 | |
256 | return NULL; | |
257 | } | |
ac27a0ec | 258 | |
617ba13b MC |
259 | group_desc = block_group >> EXT4_DESC_PER_BLOCK_BITS(sb); |
260 | offset = block_group & (EXT4_DESC_PER_BLOCK(sb) - 1); | |
ac27a0ec | 261 | if (!sbi->s_group_desc[group_desc]) { |
12062ddd | 262 | ext4_error(sb, "Group descriptor not loaded - " |
498e5f24 | 263 | "block_group = %u, group_desc = %u, desc = %u", |
af5bc92d | 264 | block_group, group_desc, offset); |
ac27a0ec DK |
265 | return NULL; |
266 | } | |
267 | ||
0d1ee42f AR |
268 | desc = (struct ext4_group_desc *)( |
269 | (__u8 *)sbi->s_group_desc[group_desc]->b_data + | |
270 | offset * EXT4_DESC_SIZE(sb)); | |
ac27a0ec DK |
271 | if (bh) |
272 | *bh = sbi->s_group_desc[group_desc]; | |
0d1ee42f | 273 | return desc; |
ac27a0ec DK |
274 | } |
275 | ||
abcb2947 AK |
276 | static int ext4_valid_block_bitmap(struct super_block *sb, |
277 | struct ext4_group_desc *desc, | |
278 | unsigned int block_group, | |
279 | struct buffer_head *bh) | |
280 | { | |
281 | ext4_grpblk_t offset; | |
282 | ext4_grpblk_t next_zero_bit; | |
283 | ext4_fsblk_t bitmap_blk; | |
284 | ext4_fsblk_t group_first_block; | |
285 | ||
286 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) { | |
287 | /* with FLEX_BG, the inode/block bitmaps and itable | |
288 | * blocks may not be in the group at all | |
289 | * so the bitmap validation will be skipped for those groups | |
290 | * or it has to also read the block group where the bitmaps | |
291 | * are located to verify they are set. | |
292 | */ | |
293 | return 1; | |
294 | } | |
295 | group_first_block = ext4_group_first_block_no(sb, block_group); | |
296 | ||
297 | /* check whether block bitmap block number is set */ | |
298 | bitmap_blk = ext4_block_bitmap(sb, desc); | |
299 | offset = bitmap_blk - group_first_block; | |
300 | if (!ext4_test_bit(offset, bh->b_data)) | |
301 | /* bad block bitmap */ | |
302 | goto err_out; | |
303 | ||
304 | /* check whether the inode bitmap block number is set */ | |
305 | bitmap_blk = ext4_inode_bitmap(sb, desc); | |
306 | offset = bitmap_blk - group_first_block; | |
307 | if (!ext4_test_bit(offset, bh->b_data)) | |
308 | /* bad block bitmap */ | |
309 | goto err_out; | |
310 | ||
311 | /* check whether the inode table block number is set */ | |
312 | bitmap_blk = ext4_inode_table(sb, desc); | |
313 | offset = bitmap_blk - group_first_block; | |
314 | next_zero_bit = ext4_find_next_zero_bit(bh->b_data, | |
315 | offset + EXT4_SB(sb)->s_itb_per_group, | |
316 | offset); | |
317 | if (next_zero_bit >= offset + EXT4_SB(sb)->s_itb_per_group) | |
318 | /* good bitmap for inode tables */ | |
319 | return 1; | |
320 | ||
321 | err_out: | |
12062ddd | 322 | ext4_error(sb, "Invalid block bitmap - block_group = %d, block = %llu", |
abcb2947 AK |
323 | block_group, bitmap_blk); |
324 | return 0; | |
325 | } | |
ac27a0ec | 326 | /** |
574ca174 | 327 | * ext4_read_block_bitmap() |
ac27a0ec DK |
328 | * @sb: super block |
329 | * @block_group: given block group | |
330 | * | |
abcb2947 AK |
331 | * Read the bitmap for a given block_group,and validate the |
332 | * bits for block/inode/inode tables are set in the bitmaps | |
ac27a0ec DK |
333 | * |
334 | * Return buffer_head on success or NULL in case of failure. | |
335 | */ | |
717d50e4 | 336 | struct buffer_head * |
574ca174 | 337 | ext4_read_block_bitmap(struct super_block *sb, ext4_group_t block_group) |
ac27a0ec | 338 | { |
af5bc92d TT |
339 | struct ext4_group_desc *desc; |
340 | struct buffer_head *bh = NULL; | |
7c9e69fa | 341 | ext4_fsblk_t bitmap_blk; |
ac27a0ec | 342 | |
717d50e4 | 343 | desc = ext4_get_group_desc(sb, block_group, NULL); |
ac27a0ec | 344 | if (!desc) |
7c9e69fa AK |
345 | return NULL; |
346 | bitmap_blk = ext4_block_bitmap(sb, desc); | |
abcb2947 AK |
347 | bh = sb_getblk(sb, bitmap_blk); |
348 | if (unlikely(!bh)) { | |
12062ddd | 349 | ext4_error(sb, "Cannot read block bitmap - " |
a9df9a49 | 350 | "block_group = %u, block_bitmap = %llu", |
e29d1cde | 351 | block_group, bitmap_blk); |
abcb2947 AK |
352 | return NULL; |
353 | } | |
2ccb5fb9 AK |
354 | |
355 | if (bitmap_uptodate(bh)) | |
abcb2947 AK |
356 | return bh; |
357 | ||
c806e68f | 358 | lock_buffer(bh); |
2ccb5fb9 AK |
359 | if (bitmap_uptodate(bh)) { |
360 | unlock_buffer(bh); | |
361 | return bh; | |
362 | } | |
955ce5f5 | 363 | ext4_lock_group(sb, block_group); |
717d50e4 | 364 | if (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) { |
abcb2947 | 365 | ext4_init_block_bitmap(sb, bh, block_group, desc); |
2ccb5fb9 | 366 | set_bitmap_uptodate(bh); |
abcb2947 | 367 | set_buffer_uptodate(bh); |
955ce5f5 | 368 | ext4_unlock_group(sb, block_group); |
3300beda | 369 | unlock_buffer(bh); |
abcb2947 | 370 | return bh; |
717d50e4 | 371 | } |
955ce5f5 | 372 | ext4_unlock_group(sb, block_group); |
2ccb5fb9 AK |
373 | if (buffer_uptodate(bh)) { |
374 | /* | |
375 | * if not uninit if bh is uptodate, | |
376 | * bitmap is also uptodate | |
377 | */ | |
378 | set_bitmap_uptodate(bh); | |
379 | unlock_buffer(bh); | |
380 | return bh; | |
381 | } | |
382 | /* | |
383 | * submit the buffer_head for read. We can | |
384 | * safely mark the bitmap as uptodate now. | |
385 | * We do it here so the bitmap uptodate bit | |
386 | * get set with buffer lock held. | |
387 | */ | |
0562e0ba | 388 | trace_ext4_read_block_bitmap_load(sb, block_group); |
2ccb5fb9 | 389 | set_bitmap_uptodate(bh); |
abcb2947 AK |
390 | if (bh_submit_read(bh) < 0) { |
391 | put_bh(bh); | |
12062ddd | 392 | ext4_error(sb, "Cannot read block bitmap - " |
a9df9a49 | 393 | "block_group = %u, block_bitmap = %llu", |
e29d1cde | 394 | block_group, bitmap_blk); |
abcb2947 AK |
395 | return NULL; |
396 | } | |
519deca0 AK |
397 | ext4_valid_block_bitmap(sb, desc, block_group, bh); |
398 | /* | |
399 | * file system mounted not to panic on error, | |
400 | * continue with corrupt bitmap | |
401 | */ | |
ac27a0ec DK |
402 | return bh; |
403 | } | |
ac27a0ec | 404 | |
8c3bf8a0 | 405 | /** |
df55c99d | 406 | * ext4_has_free_clusters() |
8c3bf8a0 | 407 | * @sbi: in-core super block structure. |
df55c99d TT |
408 | * @nclusters: number of needed blocks |
409 | * @flags: flags from ext4_mb_new_blocks() | |
8c3bf8a0 | 410 | * |
df55c99d | 411 | * Check if filesystem has nclusters free & available for allocation. |
8c3bf8a0 ES |
412 | * On success return 1, return 0 on failure. |
413 | */ | |
df55c99d TT |
414 | static int ext4_has_free_clusters(struct ext4_sb_info *sbi, |
415 | s64 nclusters, unsigned int flags) | |
a30d542a | 416 | { |
df55c99d | 417 | s64 free_clusters, dirty_clusters, root_clusters; |
57042651 | 418 | struct percpu_counter *fcc = &sbi->s_freeclusters_counter; |
df55c99d | 419 | struct percpu_counter *dcc = &sbi->s_dirtyclusters_counter; |
a30d542a | 420 | |
df55c99d TT |
421 | free_clusters = percpu_counter_read_positive(fcc); |
422 | dirty_clusters = percpu_counter_read_positive(dcc); | |
423 | root_clusters = EXT4_B2C(sbi, ext4_r_blocks_count(sbi->s_es)); | |
a30d542a | 424 | |
df55c99d TT |
425 | if (free_clusters - (nclusters + root_clusters + dirty_clusters) < |
426 | EXT4_FREECLUSTERS_WATERMARK) { | |
427 | free_clusters = EXT4_C2B(sbi, percpu_counter_sum_positive(fcc)); | |
428 | dirty_clusters = percpu_counter_sum_positive(dcc); | |
6bc6e63f | 429 | } |
df55c99d TT |
430 | /* Check whether we have space after accounting for current |
431 | * dirty clusters & root reserved clusters. | |
6bc6e63f | 432 | */ |
df55c99d | 433 | if (free_clusters >= ((root_clusters + nclusters) + dirty_clusters)) |
a996031c | 434 | return 1; |
a30d542a | 435 | |
df55c99d | 436 | /* Hm, nope. Are (enough) root reserved clusters available? */ |
4c9c544e | 437 | if (sbi->s_resuid == current_fsuid() || |
a996031c | 438 | ((sbi->s_resgid != 0) && in_group_p(sbi->s_resgid)) || |
55f020db AH |
439 | capable(CAP_SYS_RESOURCE) || |
440 | (flags & EXT4_MB_USE_ROOT_BLOCKS)) { | |
441 | ||
df55c99d | 442 | if (free_clusters >= (nclusters + dirty_clusters)) |
a996031c ES |
443 | return 1; |
444 | } | |
445 | ||
446 | return 0; | |
a30d542a AK |
447 | } |
448 | ||
e7d5f315 TT |
449 | int ext4_claim_free_clusters(struct ext4_sb_info *sbi, |
450 | s64 nclusters, unsigned int flags) | |
ac27a0ec | 451 | { |
df55c99d | 452 | if (ext4_has_free_clusters(sbi, nclusters, flags)) { |
e7d5f315 | 453 | percpu_counter_add(&sbi->s_dirtyclusters_counter, nclusters); |
16eb7295 | 454 | return 0; |
8c3bf8a0 ES |
455 | } else |
456 | return -ENOSPC; | |
a30d542a | 457 | } |
07031431 | 458 | |
ac27a0ec | 459 | /** |
617ba13b | 460 | * ext4_should_retry_alloc() |
ac27a0ec DK |
461 | * @sb: super block |
462 | * @retries number of attemps has been made | |
463 | * | |
617ba13b | 464 | * ext4_should_retry_alloc() is called when ENOSPC is returned, and if |
ac27a0ec | 465 | * it is profitable to retry the operation, this function will wait |
25985edc | 466 | * for the current or committing transaction to complete, and then |
ac27a0ec DK |
467 | * return TRUE. |
468 | * | |
469 | * if the total number of retries exceed three times, return FALSE. | |
470 | */ | |
617ba13b | 471 | int ext4_should_retry_alloc(struct super_block *sb, int *retries) |
ac27a0ec | 472 | { |
df55c99d | 473 | if (!ext4_has_free_clusters(EXT4_SB(sb), 1, 0) || |
8f64b32e ES |
474 | (*retries)++ > 3 || |
475 | !EXT4_SB(sb)->s_journal) | |
ac27a0ec DK |
476 | return 0; |
477 | ||
478 | jbd_debug(1, "%s: retrying operation after ENOSPC\n", sb->s_id); | |
479 | ||
dab291af | 480 | return jbd2_journal_force_commit_nested(EXT4_SB(sb)->s_journal); |
ac27a0ec DK |
481 | } |
482 | ||
654b4908 | 483 | /* |
d2a17637 | 484 | * ext4_new_meta_blocks() -- allocate block for meta data (indexing) blocks |
654b4908 AK |
485 | * |
486 | * @handle: handle to this transaction | |
487 | * @inode: file inode | |
488 | * @goal: given target block(filesystem wide) | |
7b415bf6 | 489 | * @count: pointer to total number of clusters needed |
654b4908 AK |
490 | * @errp: error code |
491 | * | |
97df5d15 | 492 | * Return 1st allocated block number on success, *count stores total account |
d2a17637 | 493 | * error stores in errp pointer |
654b4908 | 494 | */ |
d2a17637 | 495 | ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode, |
55f020db AH |
496 | ext4_fsblk_t goal, unsigned int flags, |
497 | unsigned long *count, int *errp) | |
654b4908 | 498 | { |
97df5d15 | 499 | struct ext4_allocation_request ar; |
d2a17637 | 500 | ext4_fsblk_t ret; |
97df5d15 TT |
501 | |
502 | memset(&ar, 0, sizeof(ar)); | |
503 | /* Fill with neighbour allocated blocks */ | |
504 | ar.inode = inode; | |
505 | ar.goal = goal; | |
506 | ar.len = count ? *count : 1; | |
55f020db | 507 | ar.flags = flags; |
97df5d15 TT |
508 | |
509 | ret = ext4_mb_new_blocks(handle, &ar, errp); | |
510 | if (count) | |
511 | *count = ar.len; | |
d2a17637 | 512 | /* |
72b8ab9d ES |
513 | * Account for the allocated meta blocks. We will never |
514 | * fail EDQUOT for metdata, but we do account for it. | |
d2a17637 | 515 | */ |
f2321097 TT |
516 | if (!(*errp) && |
517 | ext4_test_inode_state(inode, EXT4_STATE_DELALLOC_RESERVED)) { | |
d2a17637 | 518 | spin_lock(&EXT4_I(inode)->i_block_reservation_lock); |
97df5d15 | 519 | EXT4_I(inode)->i_allocated_meta_blocks += ar.len; |
d2a17637 | 520 | spin_unlock(&EXT4_I(inode)->i_block_reservation_lock); |
7b415bf6 AK |
521 | dquot_alloc_block_nofail(inode, |
522 | EXT4_C2B(EXT4_SB(inode->i_sb), ar.len)); | |
d2a17637 MC |
523 | } |
524 | return ret; | |
654b4908 AK |
525 | } |
526 | ||
ac27a0ec | 527 | /** |
5dee5437 | 528 | * ext4_count_free_clusters() -- count filesystem free clusters |
ac27a0ec DK |
529 | * @sb: superblock |
530 | * | |
5dee5437 | 531 | * Adds up the number of free clusters from each block group. |
ac27a0ec | 532 | */ |
5dee5437 | 533 | ext4_fsblk_t ext4_count_free_clusters(struct super_block *sb) |
ac27a0ec | 534 | { |
617ba13b MC |
535 | ext4_fsblk_t desc_count; |
536 | struct ext4_group_desc *gdp; | |
fd2d4291 | 537 | ext4_group_t i; |
8df9675f | 538 | ext4_group_t ngroups = ext4_get_groups_count(sb); |
617ba13b MC |
539 | #ifdef EXT4FS_DEBUG |
540 | struct ext4_super_block *es; | |
541 | ext4_fsblk_t bitmap_count; | |
498e5f24 | 542 | unsigned int x; |
ac27a0ec DK |
543 | struct buffer_head *bitmap_bh = NULL; |
544 | ||
617ba13b | 545 | es = EXT4_SB(sb)->s_es; |
ac27a0ec DK |
546 | desc_count = 0; |
547 | bitmap_count = 0; | |
548 | gdp = NULL; | |
549 | ||
ac27a0ec | 550 | for (i = 0; i < ngroups; i++) { |
617ba13b | 551 | gdp = ext4_get_group_desc(sb, i, NULL); |
ac27a0ec DK |
552 | if (!gdp) |
553 | continue; | |
021b65bb | 554 | desc_count += ext4_free_group_clusters(sb, gdp); |
ac27a0ec | 555 | brelse(bitmap_bh); |
574ca174 | 556 | bitmap_bh = ext4_read_block_bitmap(sb, i); |
ac27a0ec DK |
557 | if (bitmap_bh == NULL) |
558 | continue; | |
559 | ||
617ba13b | 560 | x = ext4_count_free(bitmap_bh, sb->s_blocksize); |
9fd9784c | 561 | printk(KERN_DEBUG "group %u: stored = %d, counted = %u\n", |
021b65bb | 562 | i, ext4_free_group_clusters(sb, gdp), x); |
ac27a0ec DK |
563 | bitmap_count += x; |
564 | } | |
565 | brelse(bitmap_bh); | |
5dee5437 TT |
566 | printk(KERN_DEBUG "ext4_count_free_clusters: stored = %llu" |
567 | ", computed = %llu, %llu\n", | |
568 | EXT4_B2C(sbi, ext4_free_blocks_count(es)), | |
4776004f | 569 | desc_count, bitmap_count); |
ac27a0ec DK |
570 | return bitmap_count; |
571 | #else | |
572 | desc_count = 0; | |
ac27a0ec | 573 | for (i = 0; i < ngroups; i++) { |
617ba13b | 574 | gdp = ext4_get_group_desc(sb, i, NULL); |
ac27a0ec DK |
575 | if (!gdp) |
576 | continue; | |
021b65bb | 577 | desc_count += ext4_free_group_clusters(sb, gdp); |
ac27a0ec DK |
578 | } |
579 | ||
580 | return desc_count; | |
581 | #endif | |
582 | } | |
583 | ||
fd2d4291 | 584 | static inline int test_root(ext4_group_t a, int b) |
ac27a0ec DK |
585 | { |
586 | int num = b; | |
587 | ||
588 | while (a > num) | |
589 | num *= b; | |
590 | return num == a; | |
591 | } | |
592 | ||
fd2d4291 | 593 | static int ext4_group_sparse(ext4_group_t group) |
ac27a0ec DK |
594 | { |
595 | if (group <= 1) | |
596 | return 1; | |
597 | if (!(group & 1)) | |
598 | return 0; | |
599 | return (test_root(group, 7) || test_root(group, 5) || | |
600 | test_root(group, 3)); | |
601 | } | |
602 | ||
603 | /** | |
617ba13b | 604 | * ext4_bg_has_super - number of blocks used by the superblock in group |
ac27a0ec DK |
605 | * @sb: superblock for filesystem |
606 | * @group: group number to check | |
607 | * | |
608 | * Return the number of blocks used by the superblock (primary or backup) | |
609 | * in this group. Currently this will be only 0 or 1. | |
610 | */ | |
fd2d4291 | 611 | int ext4_bg_has_super(struct super_block *sb, ext4_group_t group) |
ac27a0ec | 612 | { |
617ba13b MC |
613 | if (EXT4_HAS_RO_COMPAT_FEATURE(sb, |
614 | EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER) && | |
615 | !ext4_group_sparse(group)) | |
ac27a0ec DK |
616 | return 0; |
617 | return 1; | |
618 | } | |
619 | ||
fd2d4291 AM |
620 | static unsigned long ext4_bg_num_gdb_meta(struct super_block *sb, |
621 | ext4_group_t group) | |
ac27a0ec | 622 | { |
617ba13b | 623 | unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb); |
fd2d4291 AM |
624 | ext4_group_t first = metagroup * EXT4_DESC_PER_BLOCK(sb); |
625 | ext4_group_t last = first + EXT4_DESC_PER_BLOCK(sb) - 1; | |
ac27a0ec DK |
626 | |
627 | if (group == first || group == first + 1 || group == last) | |
628 | return 1; | |
629 | return 0; | |
630 | } | |
631 | ||
fd2d4291 AM |
632 | static unsigned long ext4_bg_num_gdb_nometa(struct super_block *sb, |
633 | ext4_group_t group) | |
ac27a0ec | 634 | { |
8dadb198 TT |
635 | if (!ext4_bg_has_super(sb, group)) |
636 | return 0; | |
637 | ||
638 | if (EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG)) | |
639 | return le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg); | |
640 | else | |
641 | return EXT4_SB(sb)->s_gdb_count; | |
ac27a0ec DK |
642 | } |
643 | ||
644 | /** | |
617ba13b | 645 | * ext4_bg_num_gdb - number of blocks used by the group table in group |
ac27a0ec DK |
646 | * @sb: superblock for filesystem |
647 | * @group: group number to check | |
648 | * | |
649 | * Return the number of blocks used by the group descriptor table | |
650 | * (primary or backup) in this group. In the future there may be a | |
651 | * different number of descriptor blocks in each group. | |
652 | */ | |
fd2d4291 | 653 | unsigned long ext4_bg_num_gdb(struct super_block *sb, ext4_group_t group) |
ac27a0ec DK |
654 | { |
655 | unsigned long first_meta_bg = | |
617ba13b MC |
656 | le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg); |
657 | unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb); | |
ac27a0ec | 658 | |
617ba13b | 659 | if (!EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG) || |
ac27a0ec | 660 | metagroup < first_meta_bg) |
af5bc92d | 661 | return ext4_bg_num_gdb_nometa(sb, group); |
ac27a0ec | 662 | |
617ba13b | 663 | return ext4_bg_num_gdb_meta(sb,group); |
ac27a0ec DK |
664 | |
665 | } | |
c2ea3fde | 666 | |
49f7f9af | 667 | /* |
d5b8f310 | 668 | * This function returns the number of file system metadata clusters at |
49f7f9af TT |
669 | * the beginning of a block group, including the reserved gdt blocks. |
670 | */ | |
d5b8f310 TT |
671 | unsigned ext4_num_base_meta_clusters(struct super_block *sb, |
672 | ext4_group_t block_group) | |
49f7f9af TT |
673 | { |
674 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
d5b8f310 | 675 | unsigned num; |
49f7f9af TT |
676 | |
677 | /* Check for superblock and gdt backups in this group */ | |
678 | num = ext4_bg_has_super(sb, block_group); | |
679 | ||
680 | if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) || | |
681 | block_group < le32_to_cpu(sbi->s_es->s_first_meta_bg) * | |
682 | sbi->s_desc_per_block) { | |
683 | if (num) { | |
684 | num += ext4_bg_num_gdb(sb, block_group); | |
685 | num += le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks); | |
686 | } | |
687 | } else { /* For META_BG_BLOCK_GROUPS */ | |
688 | num += ext4_bg_num_gdb(sb, block_group); | |
689 | } | |
d5b8f310 | 690 | return EXT4_NUM_B2C(sbi, num); |
49f7f9af | 691 | } |
f86186b4 ES |
692 | /** |
693 | * ext4_inode_to_goal_block - return a hint for block allocation | |
694 | * @inode: inode for block allocation | |
695 | * | |
696 | * Return the ideal location to start allocating blocks for a | |
697 | * newly created inode. | |
698 | */ | |
699 | ext4_fsblk_t ext4_inode_to_goal_block(struct inode *inode) | |
700 | { | |
701 | struct ext4_inode_info *ei = EXT4_I(inode); | |
702 | ext4_group_t block_group; | |
703 | ext4_grpblk_t colour; | |
704 | int flex_size = ext4_flex_bg_size(EXT4_SB(inode->i_sb)); | |
705 | ext4_fsblk_t bg_start; | |
706 | ext4_fsblk_t last_block; | |
707 | ||
708 | block_group = ei->i_block_group; | |
709 | if (flex_size >= EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME) { | |
710 | /* | |
711 | * If there are at least EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME | |
712 | * block groups per flexgroup, reserve the first block | |
713 | * group for directories and special files. Regular | |
714 | * files will start at the second block group. This | |
715 | * tends to speed up directory access and improves | |
716 | * fsck times. | |
717 | */ | |
718 | block_group &= ~(flex_size-1); | |
719 | if (S_ISREG(inode->i_mode)) | |
720 | block_group++; | |
721 | } | |
722 | bg_start = ext4_group_first_block_no(inode->i_sb, block_group); | |
723 | last_block = ext4_blocks_count(EXT4_SB(inode->i_sb)->s_es) - 1; | |
724 | ||
725 | /* | |
726 | * If we are doing delayed allocation, we don't need take | |
727 | * colour into account. | |
728 | */ | |
729 | if (test_opt(inode->i_sb, DELALLOC)) | |
730 | return bg_start; | |
731 | ||
732 | if (bg_start + EXT4_BLOCKS_PER_GROUP(inode->i_sb) <= last_block) | |
733 | colour = (current->pid % 16) * | |
734 | (EXT4_BLOCKS_PER_GROUP(inode->i_sb) / 16); | |
735 | else | |
736 | colour = (current->pid % 16) * ((last_block - bg_start) / 16); | |
737 | return bg_start + colour; | |
738 | } | |
739 |