]>
Commit | Line | Data |
---|---|---|
ac27a0ec | 1 | /* |
617ba13b | 2 | * linux/fs/ext4/balloc.c |
ac27a0ec DK |
3 | * |
4 | * Copyright (C) 1992, 1993, 1994, 1995 | |
5 | * Remy Card (card@masi.ibp.fr) | |
6 | * Laboratoire MASI - Institut Blaise Pascal | |
7 | * Universite Pierre et Marie Curie (Paris VI) | |
8 | * | |
9 | * Enhanced block allocation by Stephen Tweedie (sct@redhat.com), 1993 | |
10 | * Big-endian to little-endian byte-swapping/bitmaps by | |
11 | * David S. Miller (davem@caip.rutgers.edu), 1995 | |
12 | */ | |
13 | ||
14 | #include <linux/time.h> | |
15 | #include <linux/capability.h> | |
16 | #include <linux/fs.h> | |
dab291af | 17 | #include <linux/jbd2.h> |
ac27a0ec DK |
18 | #include <linux/quotaops.h> |
19 | #include <linux/buffer_head.h> | |
3dcf5451 CH |
20 | #include "ext4.h" |
21 | #include "ext4_jbd2.h" | |
e21675d4 | 22 | #include "mballoc.h" |
3dcf5451 | 23 | |
0562e0ba JZ |
24 | #include <trace/events/ext4.h> |
25 | ||
ac27a0ec DK |
26 | /* |
27 | * balloc.c contains the blocks allocation and deallocation routines | |
28 | */ | |
29 | ||
72b64b59 AM |
30 | /* |
31 | * Calculate the block group number and offset, given a block number | |
32 | */ | |
33 | void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr, | |
fd2d4291 | 34 | ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp) |
72b64b59 | 35 | { |
8c55e204 | 36 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; |
72b64b59 AM |
37 | ext4_grpblk_t offset; |
38 | ||
8c55e204 | 39 | blocknr = blocknr - le32_to_cpu(es->s_first_data_block); |
f4e5bc24 | 40 | offset = do_div(blocknr, EXT4_BLOCKS_PER_GROUP(sb)); |
72b64b59 AM |
41 | if (offsetp) |
42 | *offsetp = offset; | |
43 | if (blockgrpp) | |
8c55e204 | 44 | *blockgrpp = blocknr; |
72b64b59 AM |
45 | |
46 | } | |
47 | ||
0bf7e837 JS |
48 | static int ext4_block_in_group(struct super_block *sb, ext4_fsblk_t block, |
49 | ext4_group_t block_group) | |
50 | { | |
51 | ext4_group_t actual_group; | |
7477827f | 52 | ext4_get_group_no_and_offset(sb, block, &actual_group, NULL); |
0bf7e837 JS |
53 | if (actual_group == block_group) |
54 | return 1; | |
55 | return 0; | |
56 | } | |
57 | ||
58 | static int ext4_group_used_meta_blocks(struct super_block *sb, | |
e187c658 TT |
59 | ext4_group_t block_group, |
60 | struct ext4_group_desc *gdp) | |
0bf7e837 JS |
61 | { |
62 | ext4_fsblk_t tmp; | |
63 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
64 | /* block bitmap, inode bitmap, and inode table blocks */ | |
65 | int used_blocks = sbi->s_itb_per_group + 2; | |
66 | ||
67 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) { | |
0bf7e837 JS |
68 | if (!ext4_block_in_group(sb, ext4_block_bitmap(sb, gdp), |
69 | block_group)) | |
70 | used_blocks--; | |
71 | ||
72 | if (!ext4_block_in_group(sb, ext4_inode_bitmap(sb, gdp), | |
73 | block_group)) | |
74 | used_blocks--; | |
75 | ||
76 | tmp = ext4_inode_table(sb, gdp); | |
77 | for (; tmp < ext4_inode_table(sb, gdp) + | |
78 | sbi->s_itb_per_group; tmp++) { | |
79 | if (!ext4_block_in_group(sb, tmp, block_group)) | |
80 | used_blocks -= 1; | |
81 | } | |
82 | } | |
83 | return used_blocks; | |
84 | } | |
c2ea3fde | 85 | |
717d50e4 AD |
86 | /* Initializes an uninitialized block bitmap if given, and returns the |
87 | * number of blocks free in the group. */ | |
88 | unsigned ext4_init_block_bitmap(struct super_block *sb, struct buffer_head *bh, | |
fd2d4291 | 89 | ext4_group_t block_group, struct ext4_group_desc *gdp) |
717d50e4 | 90 | { |
717d50e4 | 91 | int bit, bit_max; |
8df9675f | 92 | ext4_group_t ngroups = ext4_get_groups_count(sb); |
717d50e4 AD |
93 | unsigned free_blocks, group_blocks; |
94 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
95 | ||
96 | if (bh) { | |
97 | J_ASSERT_BH(bh, buffer_locked(bh)); | |
98 | ||
99 | /* If checksum is bad mark all blocks used to prevent allocation | |
100 | * essentially implementing a per-group read-only flag. */ | |
101 | if (!ext4_group_desc_csum_verify(sbi, block_group, gdp)) { | |
12062ddd ES |
102 | ext4_error(sb, "Checksum bad for group %u", |
103 | block_group); | |
560671a0 AK |
104 | ext4_free_blks_set(sb, gdp, 0); |
105 | ext4_free_inodes_set(sb, gdp, 0); | |
106 | ext4_itable_unused_set(sb, gdp, 0); | |
717d50e4 AD |
107 | memset(bh->b_data, 0xff, sb->s_blocksize); |
108 | return 0; | |
109 | } | |
110 | memset(bh->b_data, 0, sb->s_blocksize); | |
111 | } | |
112 | ||
113 | /* Check for superblock and gdt backups in this group */ | |
114 | bit_max = ext4_bg_has_super(sb, block_group); | |
115 | ||
116 | if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) || | |
117 | block_group < le32_to_cpu(sbi->s_es->s_first_meta_bg) * | |
118 | sbi->s_desc_per_block) { | |
119 | if (bit_max) { | |
120 | bit_max += ext4_bg_num_gdb(sb, block_group); | |
121 | bit_max += | |
122 | le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks); | |
123 | } | |
124 | } else { /* For META_BG_BLOCK_GROUPS */ | |
6afd6707 | 125 | bit_max += ext4_bg_num_gdb(sb, block_group); |
717d50e4 AD |
126 | } |
127 | ||
8df9675f | 128 | if (block_group == ngroups - 1) { |
717d50e4 AD |
129 | /* |
130 | * Even though mke2fs always initialize first and last group | |
131 | * if some other tool enabled the EXT4_BG_BLOCK_UNINIT we need | |
132 | * to make sure we calculate the right free blocks | |
133 | */ | |
134 | group_blocks = ext4_blocks_count(sbi->s_es) - | |
5661bd68 | 135 | ext4_group_first_block_no(sb, ngroups - 1); |
717d50e4 AD |
136 | } else { |
137 | group_blocks = EXT4_BLOCKS_PER_GROUP(sb); | |
138 | } | |
139 | ||
140 | free_blocks = group_blocks - bit_max; | |
141 | ||
142 | if (bh) { | |
0bf7e837 JS |
143 | ext4_fsblk_t start, tmp; |
144 | int flex_bg = 0; | |
d00a6d7b | 145 | |
717d50e4 AD |
146 | for (bit = 0; bit < bit_max; bit++) |
147 | ext4_set_bit(bit, bh->b_data); | |
148 | ||
d00a6d7b | 149 | start = ext4_group_first_block_no(sb, block_group); |
717d50e4 | 150 | |
0bf7e837 JS |
151 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, |
152 | EXT4_FEATURE_INCOMPAT_FLEX_BG)) | |
153 | flex_bg = 1; | |
717d50e4 | 154 | |
0bf7e837 JS |
155 | /* Set bits for block and inode bitmaps, and inode table */ |
156 | tmp = ext4_block_bitmap(sb, gdp); | |
157 | if (!flex_bg || ext4_block_in_group(sb, tmp, block_group)) | |
158 | ext4_set_bit(tmp - start, bh->b_data); | |
159 | ||
160 | tmp = ext4_inode_bitmap(sb, gdp); | |
161 | if (!flex_bg || ext4_block_in_group(sb, tmp, block_group)) | |
162 | ext4_set_bit(tmp - start, bh->b_data); | |
163 | ||
164 | tmp = ext4_inode_table(sb, gdp); | |
165 | for (; tmp < ext4_inode_table(sb, gdp) + | |
166 | sbi->s_itb_per_group; tmp++) { | |
167 | if (!flex_bg || | |
168 | ext4_block_in_group(sb, tmp, block_group)) | |
169 | ext4_set_bit(tmp - start, bh->b_data); | |
170 | } | |
717d50e4 AD |
171 | /* |
172 | * Also if the number of blocks within the group is | |
173 | * less than the blocksize * 8 ( which is the size | |
174 | * of bitmap ), set rest of the block bitmap to 1 | |
175 | */ | |
61d08673 TT |
176 | ext4_mark_bitmap_end(group_blocks, sb->s_blocksize * 8, |
177 | bh->b_data); | |
717d50e4 | 178 | } |
e187c658 | 179 | return free_blocks - ext4_group_used_meta_blocks(sb, block_group, gdp); |
717d50e4 AD |
180 | } |
181 | ||
182 | ||
ac27a0ec DK |
183 | /* |
184 | * The free blocks are managed by bitmaps. A file system contains several | |
185 | * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap | |
186 | * block for inodes, N blocks for the inode table and data blocks. | |
187 | * | |
188 | * The file system contains group descriptors which are located after the | |
189 | * super block. Each descriptor contains the number of the bitmap block and | |
190 | * the free blocks count in the block. The descriptors are loaded in memory | |
e627432c | 191 | * when a file system is mounted (see ext4_fill_super). |
ac27a0ec DK |
192 | */ |
193 | ||
ac27a0ec | 194 | /** |
617ba13b | 195 | * ext4_get_group_desc() -- load group descriptor from disk |
ac27a0ec DK |
196 | * @sb: super block |
197 | * @block_group: given block group | |
198 | * @bh: pointer to the buffer head to store the block | |
199 | * group descriptor | |
200 | */ | |
af5bc92d | 201 | struct ext4_group_desc * ext4_get_group_desc(struct super_block *sb, |
fd2d4291 | 202 | ext4_group_t block_group, |
af5bc92d | 203 | struct buffer_head **bh) |
ac27a0ec | 204 | { |
498e5f24 TT |
205 | unsigned int group_desc; |
206 | unsigned int offset; | |
8df9675f | 207 | ext4_group_t ngroups = ext4_get_groups_count(sb); |
af5bc92d | 208 | struct ext4_group_desc *desc; |
617ba13b | 209 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec | 210 | |
8df9675f | 211 | if (block_group >= ngroups) { |
12062ddd ES |
212 | ext4_error(sb, "block_group >= groups_count - block_group = %u," |
213 | " groups_count = %u", block_group, ngroups); | |
ac27a0ec DK |
214 | |
215 | return NULL; | |
216 | } | |
ac27a0ec | 217 | |
617ba13b MC |
218 | group_desc = block_group >> EXT4_DESC_PER_BLOCK_BITS(sb); |
219 | offset = block_group & (EXT4_DESC_PER_BLOCK(sb) - 1); | |
ac27a0ec | 220 | if (!sbi->s_group_desc[group_desc]) { |
12062ddd | 221 | ext4_error(sb, "Group descriptor not loaded - " |
498e5f24 | 222 | "block_group = %u, group_desc = %u, desc = %u", |
af5bc92d | 223 | block_group, group_desc, offset); |
ac27a0ec DK |
224 | return NULL; |
225 | } | |
226 | ||
0d1ee42f AR |
227 | desc = (struct ext4_group_desc *)( |
228 | (__u8 *)sbi->s_group_desc[group_desc]->b_data + | |
229 | offset * EXT4_DESC_SIZE(sb)); | |
ac27a0ec DK |
230 | if (bh) |
231 | *bh = sbi->s_group_desc[group_desc]; | |
0d1ee42f | 232 | return desc; |
ac27a0ec DK |
233 | } |
234 | ||
abcb2947 AK |
235 | static int ext4_valid_block_bitmap(struct super_block *sb, |
236 | struct ext4_group_desc *desc, | |
237 | unsigned int block_group, | |
238 | struct buffer_head *bh) | |
239 | { | |
240 | ext4_grpblk_t offset; | |
241 | ext4_grpblk_t next_zero_bit; | |
242 | ext4_fsblk_t bitmap_blk; | |
243 | ext4_fsblk_t group_first_block; | |
244 | ||
245 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) { | |
246 | /* with FLEX_BG, the inode/block bitmaps and itable | |
247 | * blocks may not be in the group at all | |
248 | * so the bitmap validation will be skipped for those groups | |
249 | * or it has to also read the block group where the bitmaps | |
250 | * are located to verify they are set. | |
251 | */ | |
252 | return 1; | |
253 | } | |
254 | group_first_block = ext4_group_first_block_no(sb, block_group); | |
255 | ||
256 | /* check whether block bitmap block number is set */ | |
257 | bitmap_blk = ext4_block_bitmap(sb, desc); | |
258 | offset = bitmap_blk - group_first_block; | |
259 | if (!ext4_test_bit(offset, bh->b_data)) | |
260 | /* bad block bitmap */ | |
261 | goto err_out; | |
262 | ||
263 | /* check whether the inode bitmap block number is set */ | |
264 | bitmap_blk = ext4_inode_bitmap(sb, desc); | |
265 | offset = bitmap_blk - group_first_block; | |
266 | if (!ext4_test_bit(offset, bh->b_data)) | |
267 | /* bad block bitmap */ | |
268 | goto err_out; | |
269 | ||
270 | /* check whether the inode table block number is set */ | |
271 | bitmap_blk = ext4_inode_table(sb, desc); | |
272 | offset = bitmap_blk - group_first_block; | |
273 | next_zero_bit = ext4_find_next_zero_bit(bh->b_data, | |
274 | offset + EXT4_SB(sb)->s_itb_per_group, | |
275 | offset); | |
276 | if (next_zero_bit >= offset + EXT4_SB(sb)->s_itb_per_group) | |
277 | /* good bitmap for inode tables */ | |
278 | return 1; | |
279 | ||
280 | err_out: | |
12062ddd | 281 | ext4_error(sb, "Invalid block bitmap - block_group = %d, block = %llu", |
abcb2947 AK |
282 | block_group, bitmap_blk); |
283 | return 0; | |
284 | } | |
ac27a0ec | 285 | /** |
574ca174 | 286 | * ext4_read_block_bitmap() |
ac27a0ec DK |
287 | * @sb: super block |
288 | * @block_group: given block group | |
289 | * | |
abcb2947 AK |
290 | * Read the bitmap for a given block_group,and validate the |
291 | * bits for block/inode/inode tables are set in the bitmaps | |
ac27a0ec DK |
292 | * |
293 | * Return buffer_head on success or NULL in case of failure. | |
294 | */ | |
717d50e4 | 295 | struct buffer_head * |
574ca174 | 296 | ext4_read_block_bitmap(struct super_block *sb, ext4_group_t block_group) |
ac27a0ec | 297 | { |
af5bc92d TT |
298 | struct ext4_group_desc *desc; |
299 | struct buffer_head *bh = NULL; | |
7c9e69fa | 300 | ext4_fsblk_t bitmap_blk; |
ac27a0ec | 301 | |
717d50e4 | 302 | desc = ext4_get_group_desc(sb, block_group, NULL); |
ac27a0ec | 303 | if (!desc) |
7c9e69fa AK |
304 | return NULL; |
305 | bitmap_blk = ext4_block_bitmap(sb, desc); | |
abcb2947 AK |
306 | bh = sb_getblk(sb, bitmap_blk); |
307 | if (unlikely(!bh)) { | |
12062ddd | 308 | ext4_error(sb, "Cannot read block bitmap - " |
a9df9a49 | 309 | "block_group = %u, block_bitmap = %llu", |
e29d1cde | 310 | block_group, bitmap_blk); |
abcb2947 AK |
311 | return NULL; |
312 | } | |
2ccb5fb9 AK |
313 | |
314 | if (bitmap_uptodate(bh)) | |
abcb2947 AK |
315 | return bh; |
316 | ||
c806e68f | 317 | lock_buffer(bh); |
2ccb5fb9 AK |
318 | if (bitmap_uptodate(bh)) { |
319 | unlock_buffer(bh); | |
320 | return bh; | |
321 | } | |
955ce5f5 | 322 | ext4_lock_group(sb, block_group); |
717d50e4 | 323 | if (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) { |
abcb2947 | 324 | ext4_init_block_bitmap(sb, bh, block_group, desc); |
2ccb5fb9 | 325 | set_bitmap_uptodate(bh); |
abcb2947 | 326 | set_buffer_uptodate(bh); |
955ce5f5 | 327 | ext4_unlock_group(sb, block_group); |
3300beda | 328 | unlock_buffer(bh); |
abcb2947 | 329 | return bh; |
717d50e4 | 330 | } |
955ce5f5 | 331 | ext4_unlock_group(sb, block_group); |
2ccb5fb9 AK |
332 | if (buffer_uptodate(bh)) { |
333 | /* | |
334 | * if not uninit if bh is uptodate, | |
335 | * bitmap is also uptodate | |
336 | */ | |
337 | set_bitmap_uptodate(bh); | |
338 | unlock_buffer(bh); | |
339 | return bh; | |
340 | } | |
341 | /* | |
342 | * submit the buffer_head for read. We can | |
343 | * safely mark the bitmap as uptodate now. | |
344 | * We do it here so the bitmap uptodate bit | |
345 | * get set with buffer lock held. | |
346 | */ | |
0562e0ba | 347 | trace_ext4_read_block_bitmap_load(sb, block_group); |
2ccb5fb9 | 348 | set_bitmap_uptodate(bh); |
abcb2947 AK |
349 | if (bh_submit_read(bh) < 0) { |
350 | put_bh(bh); | |
12062ddd | 351 | ext4_error(sb, "Cannot read block bitmap - " |
a9df9a49 | 352 | "block_group = %u, block_bitmap = %llu", |
e29d1cde | 353 | block_group, bitmap_blk); |
abcb2947 AK |
354 | return NULL; |
355 | } | |
519deca0 AK |
356 | ext4_valid_block_bitmap(sb, desc, block_group, bh); |
357 | /* | |
358 | * file system mounted not to panic on error, | |
359 | * continue with corrupt bitmap | |
360 | */ | |
ac27a0ec DK |
361 | return bh; |
362 | } | |
ac27a0ec DK |
363 | |
364 | /** | |
e21675d4 | 365 | * ext4_add_groupblocks() -- Add given blocks to an existing group |
ac27a0ec DK |
366 | * @handle: handle to this transaction |
367 | * @sb: super block | |
e21675d4 | 368 | * @block: start physcial block to add to the block group |
ac27a0ec | 369 | * @count: number of blocks to free |
c2ea3fde | 370 | * |
e21675d4 AK |
371 | * This marks the blocks as free in the bitmap. We ask the |
372 | * mballoc to reload the buddy after this by setting group | |
373 | * EXT4_GROUP_INFO_NEED_INIT_BIT flag | |
ac27a0ec | 374 | */ |
e21675d4 AK |
375 | void ext4_add_groupblocks(handle_t *handle, struct super_block *sb, |
376 | ext4_fsblk_t block, unsigned long count) | |
ac27a0ec DK |
377 | { |
378 | struct buffer_head *bitmap_bh = NULL; | |
379 | struct buffer_head *gd_bh; | |
fd2d4291 | 380 | ext4_group_t block_group; |
617ba13b | 381 | ext4_grpblk_t bit; |
498e5f24 | 382 | unsigned int i; |
af5bc92d | 383 | struct ext4_group_desc *desc; |
a0375156 | 384 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
560671a0 | 385 | int err = 0, ret, blk_free_count; |
e21675d4 AK |
386 | ext4_grpblk_t blocks_freed; |
387 | struct ext4_group_info *grp; | |
ac27a0ec | 388 | |
e21675d4 | 389 | ext4_debug("Adding block(s) %llu-%llu\n", block, block + count - 1); |
ac27a0ec | 390 | |
3a5b2ecd | 391 | ext4_get_group_no_and_offset(sb, block, &block_group, &bit); |
920313a7 | 392 | grp = ext4_get_group_info(sb, block_group); |
ac27a0ec DK |
393 | /* |
394 | * Check to see if we are freeing blocks across a group | |
395 | * boundary. | |
396 | */ | |
617ba13b | 397 | if (bit + count > EXT4_BLOCKS_PER_GROUP(sb)) { |
e21675d4 | 398 | goto error_return; |
ac27a0ec | 399 | } |
574ca174 | 400 | bitmap_bh = ext4_read_block_bitmap(sb, block_group); |
ac27a0ec DK |
401 | if (!bitmap_bh) |
402 | goto error_return; | |
af5bc92d | 403 | desc = ext4_get_group_desc(sb, block_group, &gd_bh); |
ac27a0ec DK |
404 | if (!desc) |
405 | goto error_return; | |
406 | ||
8fadc143 AR |
407 | if (in_range(ext4_block_bitmap(sb, desc), block, count) || |
408 | in_range(ext4_inode_bitmap(sb, desc), block, count) || | |
409 | in_range(block, ext4_inode_table(sb, desc), sbi->s_itb_per_group) || | |
410 | in_range(block + count - 1, ext4_inode_table(sb, desc), | |
cb47dce7 | 411 | sbi->s_itb_per_group)) { |
12062ddd | 412 | ext4_error(sb, "Adding blocks in system zones - " |
af5bc92d TT |
413 | "Block = %llu, count = %lu", |
414 | block, count); | |
cb47dce7 AK |
415 | goto error_return; |
416 | } | |
ac27a0ec DK |
417 | |
418 | /* | |
e21675d4 | 419 | * We are about to add blocks to the bitmap, |
ac27a0ec DK |
420 | * so we need undo access. |
421 | */ | |
ac27a0ec | 422 | BUFFER_TRACE(bitmap_bh, "getting undo access"); |
617ba13b | 423 | err = ext4_journal_get_undo_access(handle, bitmap_bh); |
ac27a0ec DK |
424 | if (err) |
425 | goto error_return; | |
426 | ||
427 | /* | |
428 | * We are about to modify some metadata. Call the journal APIs | |
429 | * to unshare ->b_data if a currently-committing transaction is | |
430 | * using it | |
431 | */ | |
432 | BUFFER_TRACE(gd_bh, "get_write_access"); | |
617ba13b | 433 | err = ext4_journal_get_write_access(handle, gd_bh); |
ac27a0ec DK |
434 | if (err) |
435 | goto error_return; | |
920313a7 AK |
436 | /* |
437 | * make sure we don't allow a parallel init on other groups in the | |
438 | * same buddy cache | |
439 | */ | |
440 | down_write(&grp->alloc_sem); | |
e21675d4 | 441 | for (i = 0, blocks_freed = 0; i < count; i++) { |
ac27a0ec | 442 | BUFFER_TRACE(bitmap_bh, "clear bit"); |
955ce5f5 | 443 | if (!ext4_clear_bit_atomic(ext4_group_lock_ptr(sb, block_group), |
ac27a0ec | 444 | bit + i, bitmap_bh->b_data)) { |
12062ddd | 445 | ext4_error(sb, "bit already cleared for block %llu", |
bd81d8ee | 446 | (ext4_fsblk_t)(block + i)); |
ac27a0ec DK |
447 | BUFFER_TRACE(bitmap_bh, "bit already cleared"); |
448 | } else { | |
e21675d4 | 449 | blocks_freed++; |
ac27a0ec DK |
450 | } |
451 | } | |
955ce5f5 | 452 | ext4_lock_group(sb, block_group); |
560671a0 AK |
453 | blk_free_count = blocks_freed + ext4_free_blks_count(sb, desc); |
454 | ext4_free_blks_set(sb, desc, blk_free_count); | |
717d50e4 | 455 | desc->bg_checksum = ext4_group_desc_csum(sbi, block_group, desc); |
955ce5f5 | 456 | ext4_unlock_group(sb, block_group); |
e21675d4 | 457 | percpu_counter_add(&sbi->s_freeblocks_counter, blocks_freed); |
ac27a0ec | 458 | |
772cb7c8 JS |
459 | if (sbi->s_log_groups_per_flex) { |
460 | ext4_group_t flex_group = ext4_flex_group(sbi, block_group); | |
9f24e420 TT |
461 | atomic_add(blocks_freed, |
462 | &sbi->s_flex_groups[flex_group].free_blocks); | |
772cb7c8 | 463 | } |
920313a7 AK |
464 | /* |
465 | * request to reload the buddy with the | |
466 | * new bitmap information | |
467 | */ | |
468 | set_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &(grp->bb_state)); | |
0373130d | 469 | grp->bb_free += blocks_freed; |
920313a7 | 470 | up_write(&grp->alloc_sem); |
772cb7c8 | 471 | |
ac27a0ec DK |
472 | /* We dirtied the bitmap block */ |
473 | BUFFER_TRACE(bitmap_bh, "dirtied bitmap block"); | |
0390131b | 474 | err = ext4_handle_dirty_metadata(handle, NULL, bitmap_bh); |
ac27a0ec DK |
475 | |
476 | /* And the group descriptor block */ | |
477 | BUFFER_TRACE(gd_bh, "dirtied group descriptor block"); | |
0390131b | 478 | ret = ext4_handle_dirty_metadata(handle, NULL, gd_bh); |
e21675d4 AK |
479 | if (!err) |
480 | err = ret; | |
e21675d4 | 481 | |
ac27a0ec DK |
482 | error_return: |
483 | brelse(bitmap_bh); | |
617ba13b | 484 | ext4_std_error(sb, err); |
ac27a0ec DK |
485 | return; |
486 | } | |
487 | ||
8c3bf8a0 ES |
488 | /** |
489 | * ext4_has_free_blocks() | |
490 | * @sbi: in-core super block structure. | |
491 | * @nblocks: number of needed blocks | |
492 | * | |
493 | * Check if filesystem has nblocks free & available for allocation. | |
494 | * On success return 1, return 0 on failure. | |
495 | */ | |
1f109d5a | 496 | static int ext4_has_free_blocks(struct ext4_sb_info *sbi, s64 nblocks) |
a30d542a | 497 | { |
a996031c | 498 | s64 free_blocks, dirty_blocks, root_blocks; |
a30d542a | 499 | struct percpu_counter *fbc = &sbi->s_freeblocks_counter; |
6bc6e63f | 500 | struct percpu_counter *dbc = &sbi->s_dirtyblocks_counter; |
a30d542a | 501 | |
6bc6e63f AK |
502 | free_blocks = percpu_counter_read_positive(fbc); |
503 | dirty_blocks = percpu_counter_read_positive(dbc); | |
a996031c | 504 | root_blocks = ext4_r_blocks_count(sbi->s_es); |
a30d542a | 505 | |
6bc6e63f AK |
506 | if (free_blocks - (nblocks + root_blocks + dirty_blocks) < |
507 | EXT4_FREEBLOCKS_WATERMARK) { | |
02d21168 AM |
508 | free_blocks = percpu_counter_sum_positive(fbc); |
509 | dirty_blocks = percpu_counter_sum_positive(dbc); | |
6bc6e63f AK |
510 | if (dirty_blocks < 0) { |
511 | printk(KERN_CRIT "Dirty block accounting " | |
512 | "went wrong %lld\n", | |
8f72fbdf | 513 | (long long)dirty_blocks); |
6bc6e63f AK |
514 | } |
515 | } | |
516 | /* Check whether we have space after | |
a996031c | 517 | * accounting for current dirty blocks & root reserved blocks. |
6bc6e63f | 518 | */ |
a996031c ES |
519 | if (free_blocks >= ((root_blocks + nblocks) + dirty_blocks)) |
520 | return 1; | |
a30d542a | 521 | |
a996031c | 522 | /* Hm, nope. Are (enough) root reserved blocks available? */ |
4c9c544e | 523 | if (sbi->s_resuid == current_fsuid() || |
a996031c ES |
524 | ((sbi->s_resgid != 0) && in_group_p(sbi->s_resgid)) || |
525 | capable(CAP_SYS_RESOURCE)) { | |
526 | if (free_blocks >= (nblocks + dirty_blocks)) | |
527 | return 1; | |
528 | } | |
529 | ||
530 | return 0; | |
a30d542a AK |
531 | } |
532 | ||
8c3bf8a0 | 533 | int ext4_claim_free_blocks(struct ext4_sb_info *sbi, |
5c791616 | 534 | s64 nblocks) |
ac27a0ec | 535 | { |
8c3bf8a0 ES |
536 | if (ext4_has_free_blocks(sbi, nblocks)) { |
537 | percpu_counter_add(&sbi->s_dirtyblocks_counter, nblocks); | |
16eb7295 | 538 | return 0; |
8c3bf8a0 ES |
539 | } else |
540 | return -ENOSPC; | |
a30d542a | 541 | } |
07031431 | 542 | |
ac27a0ec | 543 | /** |
617ba13b | 544 | * ext4_should_retry_alloc() |
ac27a0ec DK |
545 | * @sb: super block |
546 | * @retries number of attemps has been made | |
547 | * | |
617ba13b | 548 | * ext4_should_retry_alloc() is called when ENOSPC is returned, and if |
ac27a0ec DK |
549 | * it is profitable to retry the operation, this function will wait |
550 | * for the current or commiting transaction to complete, and then | |
551 | * return TRUE. | |
552 | * | |
553 | * if the total number of retries exceed three times, return FALSE. | |
554 | */ | |
617ba13b | 555 | int ext4_should_retry_alloc(struct super_block *sb, int *retries) |
ac27a0ec | 556 | { |
8f64b32e ES |
557 | if (!ext4_has_free_blocks(EXT4_SB(sb), 1) || |
558 | (*retries)++ > 3 || | |
559 | !EXT4_SB(sb)->s_journal) | |
ac27a0ec DK |
560 | return 0; |
561 | ||
562 | jbd_debug(1, "%s: retrying operation after ENOSPC\n", sb->s_id); | |
563 | ||
dab291af | 564 | return jbd2_journal_force_commit_nested(EXT4_SB(sb)->s_journal); |
ac27a0ec DK |
565 | } |
566 | ||
654b4908 | 567 | /* |
d2a17637 | 568 | * ext4_new_meta_blocks() -- allocate block for meta data (indexing) blocks |
654b4908 AK |
569 | * |
570 | * @handle: handle to this transaction | |
571 | * @inode: file inode | |
572 | * @goal: given target block(filesystem wide) | |
97df5d15 | 573 | * @count: pointer to total number of blocks needed |
654b4908 AK |
574 | * @errp: error code |
575 | * | |
97df5d15 | 576 | * Return 1st allocated block number on success, *count stores total account |
d2a17637 | 577 | * error stores in errp pointer |
654b4908 | 578 | */ |
d2a17637 MC |
579 | ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode, |
580 | ext4_fsblk_t goal, unsigned long *count, int *errp) | |
654b4908 | 581 | { |
97df5d15 | 582 | struct ext4_allocation_request ar; |
d2a17637 | 583 | ext4_fsblk_t ret; |
97df5d15 TT |
584 | |
585 | memset(&ar, 0, sizeof(ar)); | |
586 | /* Fill with neighbour allocated blocks */ | |
587 | ar.inode = inode; | |
588 | ar.goal = goal; | |
589 | ar.len = count ? *count : 1; | |
590 | ||
591 | ret = ext4_mb_new_blocks(handle, &ar, errp); | |
592 | if (count) | |
593 | *count = ar.len; | |
d2a17637 | 594 | /* |
72b8ab9d ES |
595 | * Account for the allocated meta blocks. We will never |
596 | * fail EDQUOT for metdata, but we do account for it. | |
d2a17637 | 597 | */ |
f2321097 TT |
598 | if (!(*errp) && |
599 | ext4_test_inode_state(inode, EXT4_STATE_DELALLOC_RESERVED)) { | |
d2a17637 | 600 | spin_lock(&EXT4_I(inode)->i_block_reservation_lock); |
97df5d15 | 601 | EXT4_I(inode)->i_allocated_meta_blocks += ar.len; |
d2a17637 | 602 | spin_unlock(&EXT4_I(inode)->i_block_reservation_lock); |
72b8ab9d | 603 | dquot_alloc_block_nofail(inode, ar.len); |
d2a17637 MC |
604 | } |
605 | return ret; | |
654b4908 AK |
606 | } |
607 | ||
ac27a0ec | 608 | /** |
617ba13b | 609 | * ext4_count_free_blocks() -- count filesystem free blocks |
ac27a0ec DK |
610 | * @sb: superblock |
611 | * | |
612 | * Adds up the number of free blocks from each block group. | |
613 | */ | |
617ba13b | 614 | ext4_fsblk_t ext4_count_free_blocks(struct super_block *sb) |
ac27a0ec | 615 | { |
617ba13b MC |
616 | ext4_fsblk_t desc_count; |
617 | struct ext4_group_desc *gdp; | |
fd2d4291 | 618 | ext4_group_t i; |
8df9675f | 619 | ext4_group_t ngroups = ext4_get_groups_count(sb); |
617ba13b MC |
620 | #ifdef EXT4FS_DEBUG |
621 | struct ext4_super_block *es; | |
622 | ext4_fsblk_t bitmap_count; | |
498e5f24 | 623 | unsigned int x; |
ac27a0ec DK |
624 | struct buffer_head *bitmap_bh = NULL; |
625 | ||
617ba13b | 626 | es = EXT4_SB(sb)->s_es; |
ac27a0ec DK |
627 | desc_count = 0; |
628 | bitmap_count = 0; | |
629 | gdp = NULL; | |
630 | ||
ac27a0ec | 631 | for (i = 0; i < ngroups; i++) { |
617ba13b | 632 | gdp = ext4_get_group_desc(sb, i, NULL); |
ac27a0ec DK |
633 | if (!gdp) |
634 | continue; | |
9fd9784c | 635 | desc_count += ext4_free_blks_count(sb, gdp); |
ac27a0ec | 636 | brelse(bitmap_bh); |
574ca174 | 637 | bitmap_bh = ext4_read_block_bitmap(sb, i); |
ac27a0ec DK |
638 | if (bitmap_bh == NULL) |
639 | continue; | |
640 | ||
617ba13b | 641 | x = ext4_count_free(bitmap_bh, sb->s_blocksize); |
9fd9784c TLSC |
642 | printk(KERN_DEBUG "group %u: stored = %d, counted = %u\n", |
643 | i, ext4_free_blks_count(sb, gdp), x); | |
ac27a0ec DK |
644 | bitmap_count += x; |
645 | } | |
646 | brelse(bitmap_bh); | |
4776004f TT |
647 | printk(KERN_DEBUG "ext4_count_free_blocks: stored = %llu" |
648 | ", computed = %llu, %llu\n", ext4_free_blocks_count(es), | |
649 | desc_count, bitmap_count); | |
ac27a0ec DK |
650 | return bitmap_count; |
651 | #else | |
652 | desc_count = 0; | |
ac27a0ec | 653 | for (i = 0; i < ngroups; i++) { |
617ba13b | 654 | gdp = ext4_get_group_desc(sb, i, NULL); |
ac27a0ec DK |
655 | if (!gdp) |
656 | continue; | |
560671a0 | 657 | desc_count += ext4_free_blks_count(sb, gdp); |
ac27a0ec DK |
658 | } |
659 | ||
660 | return desc_count; | |
661 | #endif | |
662 | } | |
663 | ||
fd2d4291 | 664 | static inline int test_root(ext4_group_t a, int b) |
ac27a0ec DK |
665 | { |
666 | int num = b; | |
667 | ||
668 | while (a > num) | |
669 | num *= b; | |
670 | return num == a; | |
671 | } | |
672 | ||
fd2d4291 | 673 | static int ext4_group_sparse(ext4_group_t group) |
ac27a0ec DK |
674 | { |
675 | if (group <= 1) | |
676 | return 1; | |
677 | if (!(group & 1)) | |
678 | return 0; | |
679 | return (test_root(group, 7) || test_root(group, 5) || | |
680 | test_root(group, 3)); | |
681 | } | |
682 | ||
683 | /** | |
617ba13b | 684 | * ext4_bg_has_super - number of blocks used by the superblock in group |
ac27a0ec DK |
685 | * @sb: superblock for filesystem |
686 | * @group: group number to check | |
687 | * | |
688 | * Return the number of blocks used by the superblock (primary or backup) | |
689 | * in this group. Currently this will be only 0 or 1. | |
690 | */ | |
fd2d4291 | 691 | int ext4_bg_has_super(struct super_block *sb, ext4_group_t group) |
ac27a0ec | 692 | { |
617ba13b MC |
693 | if (EXT4_HAS_RO_COMPAT_FEATURE(sb, |
694 | EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER) && | |
695 | !ext4_group_sparse(group)) | |
ac27a0ec DK |
696 | return 0; |
697 | return 1; | |
698 | } | |
699 | ||
fd2d4291 AM |
700 | static unsigned long ext4_bg_num_gdb_meta(struct super_block *sb, |
701 | ext4_group_t group) | |
ac27a0ec | 702 | { |
617ba13b | 703 | unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb); |
fd2d4291 AM |
704 | ext4_group_t first = metagroup * EXT4_DESC_PER_BLOCK(sb); |
705 | ext4_group_t last = first + EXT4_DESC_PER_BLOCK(sb) - 1; | |
ac27a0ec DK |
706 | |
707 | if (group == first || group == first + 1 || group == last) | |
708 | return 1; | |
709 | return 0; | |
710 | } | |
711 | ||
fd2d4291 AM |
712 | static unsigned long ext4_bg_num_gdb_nometa(struct super_block *sb, |
713 | ext4_group_t group) | |
ac27a0ec | 714 | { |
8dadb198 TT |
715 | if (!ext4_bg_has_super(sb, group)) |
716 | return 0; | |
717 | ||
718 | if (EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG)) | |
719 | return le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg); | |
720 | else | |
721 | return EXT4_SB(sb)->s_gdb_count; | |
ac27a0ec DK |
722 | } |
723 | ||
724 | /** | |
617ba13b | 725 | * ext4_bg_num_gdb - number of blocks used by the group table in group |
ac27a0ec DK |
726 | * @sb: superblock for filesystem |
727 | * @group: group number to check | |
728 | * | |
729 | * Return the number of blocks used by the group descriptor table | |
730 | * (primary or backup) in this group. In the future there may be a | |
731 | * different number of descriptor blocks in each group. | |
732 | */ | |
fd2d4291 | 733 | unsigned long ext4_bg_num_gdb(struct super_block *sb, ext4_group_t group) |
ac27a0ec DK |
734 | { |
735 | unsigned long first_meta_bg = | |
617ba13b MC |
736 | le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg); |
737 | unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb); | |
ac27a0ec | 738 | |
617ba13b | 739 | if (!EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG) || |
ac27a0ec | 740 | metagroup < first_meta_bg) |
af5bc92d | 741 | return ext4_bg_num_gdb_nometa(sb, group); |
ac27a0ec | 742 | |
617ba13b | 743 | return ext4_bg_num_gdb_meta(sb,group); |
ac27a0ec DK |
744 | |
745 | } | |
c2ea3fde | 746 |