Merge tag 'pm-4.8-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
[deliverable/linux.git] / fs / ext4 / balloc.c
CommitLineData
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>
ac27a0ec
DK
17#include <linux/quotaops.h>
18#include <linux/buffer_head.h>
3dcf5451
CH
19#include "ext4.h"
20#include "ext4_jbd2.h"
e21675d4 21#include "mballoc.h"
3dcf5451 22
0562e0ba
JZ
23#include <trace/events/ext4.h>
24
5f163cc7
ES
25static unsigned ext4_num_base_meta_clusters(struct super_block *sb,
26 ext4_group_t block_group);
ac27a0ec
DK
27/*
28 * balloc.c contains the blocks allocation and deallocation routines
29 */
30
bd86298e
LC
31/*
32 * Calculate block group number for a given block number
33 */
34ext4_group_t ext4_get_group_number(struct super_block *sb,
35 ext4_fsblk_t block)
36{
37 ext4_group_t group;
38
39 if (test_opt2(sb, STD_GROUP_SIZE))
960fd856
TT
40 group = (block -
41 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block)) >>
bd86298e
LC
42 (EXT4_BLOCK_SIZE_BITS(sb) + EXT4_CLUSTER_BITS(sb) + 3);
43 else
44 ext4_get_group_no_and_offset(sb, block, &group, NULL);
45 return group;
46}
47
72b64b59 48/*
3212a80a
TT
49 * Calculate the block group number and offset into the block/cluster
50 * allocation bitmap, given a block number
72b64b59
AM
51 */
52void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
fd2d4291 53 ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp)
72b64b59 54{
8c55e204 55 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
72b64b59
AM
56 ext4_grpblk_t offset;
57
8c55e204 58 blocknr = blocknr - le32_to_cpu(es->s_first_data_block);
3212a80a
TT
59 offset = do_div(blocknr, EXT4_BLOCKS_PER_GROUP(sb)) >>
60 EXT4_SB(sb)->s_cluster_bits;
72b64b59
AM
61 if (offsetp)
62 *offsetp = offset;
63 if (blockgrpp)
8c55e204 64 *blockgrpp = blocknr;
72b64b59
AM
65
66}
67
68911009
LC
68/*
69 * Check whether the 'block' lives within the 'block_group'. Returns 1 if so
70 * and 0 otherwise.
71 */
72static inline int ext4_block_in_group(struct super_block *sb,
73 ext4_fsblk_t block,
74 ext4_group_t block_group)
0bf7e837
JS
75{
76 ext4_group_t actual_group;
68911009 77
bd86298e 78 actual_group = ext4_get_group_number(sb, block);
68911009 79 return (actual_group == block_group) ? 1 : 0;
0bf7e837
JS
80}
81
d5b8f310
TT
82/* Return the number of clusters used for file system metadata; this
83 * represents the overhead needed by the file system.
84 */
c197855e
SH
85static unsigned ext4_num_overhead_clusters(struct super_block *sb,
86 ext4_group_t block_group,
87 struct ext4_group_desc *gdp)
0bf7e837 88{
d5b8f310
TT
89 unsigned num_clusters;
90 int block_cluster = -1, inode_cluster = -1, itbl_cluster = -1, i, c;
91 ext4_fsblk_t start = ext4_group_first_block_no(sb, block_group);
92 ext4_fsblk_t itbl_blk;
0bf7e837 93 struct ext4_sb_info *sbi = EXT4_SB(sb);
0bf7e837 94
d5b8f310
TT
95 /* This is the number of clusters used by the superblock,
96 * block group descriptors, and reserved block group
97 * descriptor blocks */
98 num_clusters = ext4_num_base_meta_clusters(sb, block_group);
99
100 /*
101 * For the allocation bitmaps and inode table, we first need
102 * to check to see if the block is in the block group. If it
103 * is, then check to see if the cluster is already accounted
104 * for in the clusters used for the base metadata cluster, or
105 * if we can increment the base metadata cluster to include
106 * that block. Otherwise, we will have to track the cluster
107 * used for the allocation bitmap or inode table explicitly.
108 * Normally all of these blocks are contiguous, so the special
109 * case handling shouldn't be necessary except for *very*
110 * unusual file system layouts.
111 */
112 if (ext4_block_in_group(sb, ext4_block_bitmap(sb, gdp), block_group)) {
b0dd6b70
TT
113 block_cluster = EXT4_B2C(sbi,
114 ext4_block_bitmap(sb, gdp) - start);
d5b8f310
TT
115 if (block_cluster < num_clusters)
116 block_cluster = -1;
117 else if (block_cluster == num_clusters) {
118 num_clusters++;
119 block_cluster = -1;
120 }
121 }
122
123 if (ext4_block_in_group(sb, ext4_inode_bitmap(sb, gdp), block_group)) {
124 inode_cluster = EXT4_B2C(sbi,
b0dd6b70 125 ext4_inode_bitmap(sb, gdp) - start);
d5b8f310
TT
126 if (inode_cluster < num_clusters)
127 inode_cluster = -1;
128 else if (inode_cluster == num_clusters) {
129 num_clusters++;
130 inode_cluster = -1;
131 }
132 }
133
134 itbl_blk = ext4_inode_table(sb, gdp);
135 for (i = 0; i < sbi->s_itb_per_group; i++) {
136 if (ext4_block_in_group(sb, itbl_blk + i, block_group)) {
b0dd6b70 137 c = EXT4_B2C(sbi, itbl_blk + i - start);
d5b8f310
TT
138 if ((c < num_clusters) || (c == inode_cluster) ||
139 (c == block_cluster) || (c == itbl_cluster))
140 continue;
141 if (c == num_clusters) {
142 num_clusters++;
143 continue;
144 }
145 num_clusters++;
146 itbl_cluster = c;
0bf7e837
JS
147 }
148 }
d5b8f310
TT
149
150 if (block_cluster != -1)
151 num_clusters++;
152 if (inode_cluster != -1)
153 num_clusters++;
154
155 return num_clusters;
0bf7e837 156}
c2ea3fde 157
d5b8f310
TT
158static unsigned int num_clusters_in_group(struct super_block *sb,
159 ext4_group_t block_group)
49f7f9af 160{
d5b8f310
TT
161 unsigned int blocks;
162
49f7f9af
TT
163 if (block_group == ext4_get_groups_count(sb) - 1) {
164 /*
165 * Even though mke2fs always initializes the first and
166 * last group, just in case some other tool was used,
167 * we need to make sure we calculate the right free
168 * blocks.
169 */
d5b8f310 170 blocks = ext4_blocks_count(EXT4_SB(sb)->s_es) -
49f7f9af
TT
171 ext4_group_first_block_no(sb, block_group);
172 } else
d5b8f310
TT
173 blocks = EXT4_BLOCKS_PER_GROUP(sb);
174 return EXT4_NUM_B2C(EXT4_SB(sb), blocks);
49f7f9af
TT
175}
176
fd034a84 177/* Initializes an uninitialized block bitmap */
aef4885a 178static int ext4_init_block_bitmap(struct super_block *sb,
c197855e
SH
179 struct buffer_head *bh,
180 ext4_group_t block_group,
181 struct ext4_group_desc *gdp)
717d50e4 182{
d5b8f310 183 unsigned int bit, bit_max;
717d50e4 184 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd034a84
TT
185 ext4_fsblk_t start, tmp;
186 int flex_bg = 0;
bdfb6ff4 187 struct ext4_group_info *grp;
fd034a84
TT
188
189 J_ASSERT_BH(bh, buffer_locked(bh));
190
191 /* If checksum is bad mark all blocks used to prevent allocation
192 * essentially implementing a per-group read-only flag. */
feb0ab32 193 if (!ext4_group_desc_csum_verify(sb, block_group, gdp)) {
bdfb6ff4 194 grp = ext4_get_group_info(sb, block_group);
e43bb4e6
NJ
195 if (!EXT4_MB_GRP_BBITMAP_CORRUPT(grp))
196 percpu_counter_sub(&sbi->s_freeclusters_counter,
197 grp->bb_free);
bdfb6ff4 198 set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, &grp->bb_state);
e43bb4e6
NJ
199 if (!EXT4_MB_GRP_IBITMAP_CORRUPT(grp)) {
200 int count;
201 count = ext4_free_inodes_count(sb, gdp);
202 percpu_counter_sub(&sbi->s_freeinodes_counter,
203 count);
204 }
bdfb6ff4 205 set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, &grp->bb_state);
6a797d27 206 return -EFSBADCRC;
717d50e4 207 }
fd034a84 208 memset(bh->b_data, 0, sb->s_blocksize);
717d50e4 209
d5b8f310 210 bit_max = ext4_num_base_meta_clusters(sb, block_group);
fd034a84
TT
211 for (bit = 0; bit < bit_max; bit++)
212 ext4_set_bit(bit, bh->b_data);
d00a6d7b 213
fd034a84 214 start = ext4_group_first_block_no(sb, block_group);
717d50e4 215
e2b911c5 216 if (ext4_has_feature_flex_bg(sb))
fd034a84 217 flex_bg = 1;
717d50e4 218
fd034a84
TT
219 /* Set bits for block and inode bitmaps, and inode table */
220 tmp = ext4_block_bitmap(sb, gdp);
221 if (!flex_bg || ext4_block_in_group(sb, tmp, block_group))
d5b8f310 222 ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data);
717d50e4 223
fd034a84
TT
224 tmp = ext4_inode_bitmap(sb, gdp);
225 if (!flex_bg || ext4_block_in_group(sb, tmp, block_group))
d5b8f310 226 ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data);
0bf7e837 227
fd034a84
TT
228 tmp = ext4_inode_table(sb, gdp);
229 for (; tmp < ext4_inode_table(sb, gdp) +
230 sbi->s_itb_per_group; tmp++) {
0bf7e837 231 if (!flex_bg || ext4_block_in_group(sb, tmp, block_group))
d5b8f310 232 ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data);
717d50e4 233 }
d5b8f310 234
fd034a84
TT
235 /*
236 * Also if the number of blocks within the group is less than
237 * the blocksize * 8 ( which is the size of bitmap ), set rest
238 * of the block bitmap to 1
239 */
d5b8f310 240 ext4_mark_bitmap_end(num_clusters_in_group(sb, block_group),
fd034a84 241 sb->s_blocksize * 8, bh->b_data);
79f1ba49 242 ext4_block_bitmap_csum_set(sb, block_group, gdp, bh);
feb0ab32 243 ext4_group_desc_csum_set(sb, block_group, gdp);
aef4885a 244 return 0;
717d50e4
AD
245}
246
fd034a84
TT
247/* Return the number of free blocks in a block group. It is used when
248 * the block bitmap is uninitialized, so we can't just count the bits
249 * in the bitmap. */
cff1dfd7
TT
250unsigned ext4_free_clusters_after_init(struct super_block *sb,
251 ext4_group_t block_group,
252 struct ext4_group_desc *gdp)
fd034a84 253{
d5b8f310
TT
254 return num_clusters_in_group(sb, block_group) -
255 ext4_num_overhead_clusters(sb, block_group, gdp);
fd034a84 256}
717d50e4 257
ac27a0ec
DK
258/*
259 * The free blocks are managed by bitmaps. A file system contains several
260 * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap
261 * block for inodes, N blocks for the inode table and data blocks.
262 *
263 * The file system contains group descriptors which are located after the
264 * super block. Each descriptor contains the number of the bitmap block and
265 * the free blocks count in the block. The descriptors are loaded in memory
e627432c 266 * when a file system is mounted (see ext4_fill_super).
ac27a0ec
DK
267 */
268
ac27a0ec 269/**
617ba13b 270 * ext4_get_group_desc() -- load group descriptor from disk
ac27a0ec
DK
271 * @sb: super block
272 * @block_group: given block group
273 * @bh: pointer to the buffer head to store the block
274 * group descriptor
275 */
af5bc92d 276struct ext4_group_desc * ext4_get_group_desc(struct super_block *sb,
fd2d4291 277 ext4_group_t block_group,
af5bc92d 278 struct buffer_head **bh)
ac27a0ec 279{
498e5f24
TT
280 unsigned int group_desc;
281 unsigned int offset;
8df9675f 282 ext4_group_t ngroups = ext4_get_groups_count(sb);
af5bc92d 283 struct ext4_group_desc *desc;
617ba13b 284 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 285
8df9675f 286 if (block_group >= ngroups) {
12062ddd
ES
287 ext4_error(sb, "block_group >= groups_count - block_group = %u,"
288 " groups_count = %u", block_group, ngroups);
ac27a0ec
DK
289
290 return NULL;
291 }
ac27a0ec 292
617ba13b
MC
293 group_desc = block_group >> EXT4_DESC_PER_BLOCK_BITS(sb);
294 offset = block_group & (EXT4_DESC_PER_BLOCK(sb) - 1);
ac27a0ec 295 if (!sbi->s_group_desc[group_desc]) {
12062ddd 296 ext4_error(sb, "Group descriptor not loaded - "
498e5f24 297 "block_group = %u, group_desc = %u, desc = %u",
af5bc92d 298 block_group, group_desc, offset);
ac27a0ec
DK
299 return NULL;
300 }
301
0d1ee42f
AR
302 desc = (struct ext4_group_desc *)(
303 (__u8 *)sbi->s_group_desc[group_desc]->b_data +
304 offset * EXT4_DESC_SIZE(sb));
ac27a0ec
DK
305 if (bh)
306 *bh = sbi->s_group_desc[group_desc];
0d1ee42f 307 return desc;
ac27a0ec
DK
308}
309
7a4c5de2
TT
310/*
311 * Return the block number which was discovered to be invalid, or 0 if
312 * the block bitmap is valid.
313 */
314static ext4_fsblk_t ext4_valid_block_bitmap(struct super_block *sb,
315 struct ext4_group_desc *desc,
dbde0abe 316 ext4_group_t block_group,
7a4c5de2 317 struct buffer_head *bh)
abcb2947 318{
e674e5cb 319 struct ext4_sb_info *sbi = EXT4_SB(sb);
abcb2947
AK
320 ext4_grpblk_t offset;
321 ext4_grpblk_t next_zero_bit;
7a4c5de2 322 ext4_fsblk_t blk;
abcb2947
AK
323 ext4_fsblk_t group_first_block;
324
e2b911c5 325 if (ext4_has_feature_flex_bg(sb)) {
abcb2947
AK
326 /* with FLEX_BG, the inode/block bitmaps and itable
327 * blocks may not be in the group at all
328 * so the bitmap validation will be skipped for those groups
329 * or it has to also read the block group where the bitmaps
330 * are located to verify they are set.
331 */
7a4c5de2 332 return 0;
abcb2947
AK
333 }
334 group_first_block = ext4_group_first_block_no(sb, block_group);
335
336 /* check whether block bitmap block number is set */
7a4c5de2
TT
337 blk = ext4_block_bitmap(sb, desc);
338 offset = blk - group_first_block;
e674e5cb 339 if (!ext4_test_bit(EXT4_B2C(sbi, offset), bh->b_data))
abcb2947 340 /* bad block bitmap */
7a4c5de2 341 return blk;
abcb2947
AK
342
343 /* check whether the inode bitmap block number is set */
7a4c5de2
TT
344 blk = ext4_inode_bitmap(sb, desc);
345 offset = blk - group_first_block;
e674e5cb 346 if (!ext4_test_bit(EXT4_B2C(sbi, offset), bh->b_data))
abcb2947 347 /* bad block bitmap */
7a4c5de2 348 return blk;
abcb2947
AK
349
350 /* check whether the inode table block number is set */
7a4c5de2
TT
351 blk = ext4_inode_table(sb, desc);
352 offset = blk - group_first_block;
abcb2947 353 next_zero_bit = ext4_find_next_zero_bit(bh->b_data,
e674e5cb
DW
354 EXT4_B2C(sbi, offset + EXT4_SB(sb)->s_itb_per_group),
355 EXT4_B2C(sbi, offset));
356 if (next_zero_bit <
357 EXT4_B2C(sbi, offset + EXT4_SB(sb)->s_itb_per_group))
7a4c5de2
TT
358 /* bad bitmap for inode tables */
359 return blk;
abcb2947
AK
360 return 0;
361}
fa77dcfa 362
9008a58e
DW
363static int ext4_validate_block_bitmap(struct super_block *sb,
364 struct ext4_group_desc *desc,
365 ext4_group_t block_group,
366 struct buffer_head *bh)
fa77dcfa 367{
7a4c5de2 368 ext4_fsblk_t blk;
163a203d 369 struct ext4_group_info *grp = ext4_get_group_info(sb, block_group);
e43bb4e6 370 struct ext4_sb_info *sbi = EXT4_SB(sb);
7a4c5de2 371
9008a58e
DW
372 if (buffer_verified(bh))
373 return 0;
374 if (EXT4_MB_GRP_BBITMAP_CORRUPT(grp))
375 return -EFSCORRUPTED;
fa77dcfa
DW
376
377 ext4_lock_group(sb, block_group);
9008a58e
DW
378 if (unlikely(!ext4_block_bitmap_csum_verify(sb, block_group,
379 desc, bh))) {
7a4c5de2 380 ext4_unlock_group(sb, block_group);
9008a58e 381 ext4_error(sb, "bg %u: bad block bitmap checksum", block_group);
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);
9008a58e 386 return -EFSBADCRC;
7a4c5de2 387 }
9008a58e
DW
388 blk = ext4_valid_block_bitmap(sb, desc, block_group, bh);
389 if (unlikely(blk != 0)) {
7a4c5de2 390 ext4_unlock_group(sb, block_group);
9008a58e
DW
391 ext4_error(sb, "bg %u: block %llu: invalid block bitmap",
392 block_group, blk);
e43bb4e6
NJ
393 if (!EXT4_MB_GRP_BBITMAP_CORRUPT(grp))
394 percpu_counter_sub(&sbi->s_freeclusters_counter,
395 grp->bb_free);
163a203d 396 set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, &grp->bb_state);
9008a58e 397 return -EFSCORRUPTED;
7a4c5de2
TT
398 }
399 set_buffer_verified(bh);
fa77dcfa 400 ext4_unlock_group(sb, block_group);
9008a58e 401 return 0;
fa77dcfa
DW
402}
403
ac27a0ec 404/**
15b49132 405 * ext4_read_block_bitmap_nowait()
ac27a0ec
DK
406 * @sb: super block
407 * @block_group: given block group
408 *
abcb2947
AK
409 * Read the bitmap for a given block_group,and validate the
410 * bits for block/inode/inode tables are set in the bitmaps
ac27a0ec
DK
411 *
412 * Return buffer_head on success or NULL in case of failure.
413 */
717d50e4 414struct buffer_head *
813e5727 415ext4_read_block_bitmap_nowait(struct super_block *sb, ext4_group_t block_group)
ac27a0ec 416{
af5bc92d 417 struct ext4_group_desc *desc;
813e5727 418 struct buffer_head *bh;
7c9e69fa 419 ext4_fsblk_t bitmap_blk;
9008a58e 420 int err;
ac27a0ec 421
717d50e4 422 desc = ext4_get_group_desc(sb, block_group, NULL);
ac27a0ec 423 if (!desc)
9008a58e 424 return ERR_PTR(-EFSCORRUPTED);
7c9e69fa 425 bitmap_blk = ext4_block_bitmap(sb, desc);
abcb2947
AK
426 bh = sb_getblk(sb, bitmap_blk);
427 if (unlikely(!bh)) {
813e5727
TT
428 ext4_error(sb, "Cannot get buffer for block bitmap - "
429 "block_group = %u, block_bitmap = %llu",
430 block_group, bitmap_blk);
9008a58e 431 return ERR_PTR(-ENOMEM);
abcb2947 432 }
2ccb5fb9
AK
433
434 if (bitmap_uptodate(bh))
fa77dcfa 435 goto verify;
abcb2947 436
c806e68f 437 lock_buffer(bh);
2ccb5fb9
AK
438 if (bitmap_uptodate(bh)) {
439 unlock_buffer(bh);
fa77dcfa 440 goto verify;
2ccb5fb9 441 }
955ce5f5 442 ext4_lock_group(sb, block_group);
717d50e4 443 if (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) {
aef4885a 444 err = ext4_init_block_bitmap(sb, bh, block_group, desc);
2ccb5fb9 445 set_bitmap_uptodate(bh);
abcb2947 446 set_buffer_uptodate(bh);
955ce5f5 447 ext4_unlock_group(sb, block_group);
3300beda 448 unlock_buffer(bh);
05145bd7
JK
449 if (err) {
450 ext4_error(sb, "Failed to init block bitmap for group "
451 "%u: %d", block_group, err);
9008a58e 452 goto out;
05145bd7 453 }
41e5b7ed 454 goto verify;
717d50e4 455 }
955ce5f5 456 ext4_unlock_group(sb, block_group);
2ccb5fb9
AK
457 if (buffer_uptodate(bh)) {
458 /*
459 * if not uninit if bh is uptodate,
460 * bitmap is also uptodate
461 */
462 set_bitmap_uptodate(bh);
463 unlock_buffer(bh);
fa77dcfa 464 goto verify;
2ccb5fb9
AK
465 }
466 /*
813e5727 467 * submit the buffer_head for reading
2ccb5fb9 468 */
813e5727 469 set_buffer_new(bh);
0562e0ba 470 trace_ext4_read_block_bitmap_load(sb, block_group);
813e5727
TT
471 bh->b_end_io = ext4_end_bitmap_read;
472 get_bh(bh);
2a222ca9 473 submit_bh(REQ_OP_READ, REQ_META | REQ_PRIO, bh);
813e5727 474 return bh;
fa77dcfa 475verify:
9008a58e
DW
476 err = ext4_validate_block_bitmap(sb, desc, block_group, bh);
477 if (err)
478 goto out;
479 return bh;
480out:
48d9eb97 481 put_bh(bh);
9008a58e 482 return ERR_PTR(err);
813e5727
TT
483}
484
485/* Returns 0 on success, 1 on error */
486int ext4_wait_block_bitmap(struct super_block *sb, ext4_group_t block_group,
487 struct buffer_head *bh)
488{
489 struct ext4_group_desc *desc;
490
491 if (!buffer_new(bh))
492 return 0;
493 desc = ext4_get_group_desc(sb, block_group, NULL);
494 if (!desc)
9008a58e 495 return -EFSCORRUPTED;
813e5727
TT
496 wait_on_buffer(bh);
497 if (!buffer_uptodate(bh)) {
12062ddd 498 ext4_error(sb, "Cannot read block bitmap - "
813e5727 499 "block_group = %u, block_bitmap = %llu",
d4dc462f 500 block_group, (unsigned long long) bh->b_blocknr);
9008a58e 501 return -EIO;
abcb2947 502 }
813e5727
TT
503 clear_buffer_new(bh);
504 /* Panic or remount fs read-only if block bitmap is invalid */
9008a58e 505 return ext4_validate_block_bitmap(sb, desc, block_group, bh);
813e5727
TT
506}
507
508struct buffer_head *
509ext4_read_block_bitmap(struct super_block *sb, ext4_group_t block_group)
510{
511 struct buffer_head *bh;
9008a58e 512 int err;
813e5727
TT
513
514 bh = ext4_read_block_bitmap_nowait(sb, block_group);
9008a58e
DW
515 if (IS_ERR(bh))
516 return bh;
517 err = ext4_wait_block_bitmap(sb, block_group, bh);
518 if (err) {
813e5727 519 put_bh(bh);
9008a58e 520 return ERR_PTR(err);
813e5727 521 }
ac27a0ec
DK
522 return bh;
523}
ac27a0ec 524
8c3bf8a0 525/**
df55c99d 526 * ext4_has_free_clusters()
8c3bf8a0 527 * @sbi: in-core super block structure.
df55c99d
TT
528 * @nclusters: number of needed blocks
529 * @flags: flags from ext4_mb_new_blocks()
8c3bf8a0 530 *
df55c99d 531 * Check if filesystem has nclusters free & available for allocation.
8c3bf8a0
ES
532 * On success return 1, return 0 on failure.
533 */
df55c99d
TT
534static int ext4_has_free_clusters(struct ext4_sb_info *sbi,
535 s64 nclusters, unsigned int flags)
a30d542a 536{
27dd4385 537 s64 free_clusters, dirty_clusters, rsv, resv_clusters;
57042651 538 struct percpu_counter *fcc = &sbi->s_freeclusters_counter;
df55c99d 539 struct percpu_counter *dcc = &sbi->s_dirtyclusters_counter;
a30d542a 540
df55c99d
TT
541 free_clusters = percpu_counter_read_positive(fcc);
542 dirty_clusters = percpu_counter_read_positive(dcc);
27dd4385 543 resv_clusters = atomic64_read(&sbi->s_resv_clusters);
304e220f
LC
544
545 /*
546 * r_blocks_count should always be multiple of the cluster ratio so
547 * we are safe to do a plane bit shift only.
548 */
27dd4385
LC
549 rsv = (ext4_r_blocks_count(sbi->s_es) >> sbi->s_cluster_bits) +
550 resv_clusters;
a30d542a 551
27dd4385 552 if (free_clusters - (nclusters + rsv + dirty_clusters) <
df55c99d 553 EXT4_FREECLUSTERS_WATERMARK) {
304e220f 554 free_clusters = percpu_counter_sum_positive(fcc);
df55c99d 555 dirty_clusters = percpu_counter_sum_positive(dcc);
6bc6e63f 556 }
df55c99d
TT
557 /* Check whether we have space after accounting for current
558 * dirty clusters & root reserved clusters.
6bc6e63f 559 */
27dd4385 560 if (free_clusters >= (rsv + nclusters + dirty_clusters))
a996031c 561 return 1;
a30d542a 562
df55c99d 563 /* Hm, nope. Are (enough) root reserved clusters available? */
08cefc7a
EB
564 if (uid_eq(sbi->s_resuid, current_fsuid()) ||
565 (!gid_eq(sbi->s_resgid, GLOBAL_ROOT_GID) && in_group_p(sbi->s_resgid)) ||
55f020db 566 capable(CAP_SYS_RESOURCE) ||
27dd4385 567 (flags & EXT4_MB_USE_ROOT_BLOCKS)) {
55f020db 568
27dd4385
LC
569 if (free_clusters >= (nclusters + dirty_clusters +
570 resv_clusters))
571 return 1;
572 }
573 /* No free blocks. Let's see if we can dip into reserved pool */
574 if (flags & EXT4_MB_USE_RESERVED) {
df55c99d 575 if (free_clusters >= (nclusters + dirty_clusters))
a996031c
ES
576 return 1;
577 }
578
579 return 0;
a30d542a
AK
580}
581
e7d5f315
TT
582int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
583 s64 nclusters, unsigned int flags)
ac27a0ec 584{
df55c99d 585 if (ext4_has_free_clusters(sbi, nclusters, flags)) {
e7d5f315 586 percpu_counter_add(&sbi->s_dirtyclusters_counter, nclusters);
16eb7295 587 return 0;
8c3bf8a0
ES
588 } else
589 return -ENOSPC;
a30d542a 590}
07031431 591
ac27a0ec 592/**
617ba13b 593 * ext4_should_retry_alloc()
ac27a0ec
DK
594 * @sb: super block
595 * @retries number of attemps has been made
596 *
617ba13b 597 * ext4_should_retry_alloc() is called when ENOSPC is returned, and if
ac27a0ec 598 * it is profitable to retry the operation, this function will wait
25985edc 599 * for the current or committing transaction to complete, and then
ac27a0ec
DK
600 * return TRUE.
601 *
602 * if the total number of retries exceed three times, return FALSE.
603 */
617ba13b 604int ext4_should_retry_alloc(struct super_block *sb, int *retries)
ac27a0ec 605{
df55c99d 606 if (!ext4_has_free_clusters(EXT4_SB(sb), 1, 0) ||
8f64b32e
ES
607 (*retries)++ > 3 ||
608 !EXT4_SB(sb)->s_journal)
ac27a0ec
DK
609 return 0;
610
611 jbd_debug(1, "%s: retrying operation after ENOSPC\n", sb->s_id);
612
dbc427ce
JK
613 jbd2_journal_force_commit_nested(EXT4_SB(sb)->s_journal);
614 return 1;
ac27a0ec
DK
615}
616
654b4908 617/*
d2a17637 618 * ext4_new_meta_blocks() -- allocate block for meta data (indexing) blocks
654b4908
AK
619 *
620 * @handle: handle to this transaction
621 * @inode: file inode
622 * @goal: given target block(filesystem wide)
7b415bf6 623 * @count: pointer to total number of clusters needed
654b4908
AK
624 * @errp: error code
625 *
97df5d15 626 * Return 1st allocated block number on success, *count stores total account
d2a17637 627 * error stores in errp pointer
654b4908 628 */
d2a17637 629ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
55f020db
AH
630 ext4_fsblk_t goal, unsigned int flags,
631 unsigned long *count, int *errp)
654b4908 632{
97df5d15 633 struct ext4_allocation_request ar;
d2a17637 634 ext4_fsblk_t ret;
97df5d15
TT
635
636 memset(&ar, 0, sizeof(ar));
637 /* Fill with neighbour allocated blocks */
638 ar.inode = inode;
639 ar.goal = goal;
640 ar.len = count ? *count : 1;
55f020db 641 ar.flags = flags;
97df5d15
TT
642
643 ret = ext4_mb_new_blocks(handle, &ar, errp);
644 if (count)
645 *count = ar.len;
d2a17637 646 /*
72b8ab9d
ES
647 * Account for the allocated meta blocks. We will never
648 * fail EDQUOT for metdata, but we do account for it.
d2a17637 649 */
e3cf5d5d 650 if (!(*errp) && (flags & EXT4_MB_DELALLOC_RESERVED)) {
7b415bf6
AK
651 dquot_alloc_block_nofail(inode,
652 EXT4_C2B(EXT4_SB(inode->i_sb), ar.len));
d2a17637
MC
653 }
654 return ret;
654b4908
AK
655}
656
ac27a0ec 657/**
5dee5437 658 * ext4_count_free_clusters() -- count filesystem free clusters
ac27a0ec
DK
659 * @sb: superblock
660 *
5dee5437 661 * Adds up the number of free clusters from each block group.
ac27a0ec 662 */
5dee5437 663ext4_fsblk_t ext4_count_free_clusters(struct super_block *sb)
ac27a0ec 664{
617ba13b
MC
665 ext4_fsblk_t desc_count;
666 struct ext4_group_desc *gdp;
fd2d4291 667 ext4_group_t i;
8df9675f 668 ext4_group_t ngroups = ext4_get_groups_count(sb);
2746f7a1 669 struct ext4_group_info *grp;
617ba13b
MC
670#ifdef EXT4FS_DEBUG
671 struct ext4_super_block *es;
672 ext4_fsblk_t bitmap_count;
498e5f24 673 unsigned int x;
ac27a0ec
DK
674 struct buffer_head *bitmap_bh = NULL;
675
617ba13b 676 es = EXT4_SB(sb)->s_es;
ac27a0ec
DK
677 desc_count = 0;
678 bitmap_count = 0;
679 gdp = NULL;
680
ac27a0ec 681 for (i = 0; i < ngroups; i++) {
617ba13b 682 gdp = ext4_get_group_desc(sb, i, NULL);
ac27a0ec
DK
683 if (!gdp)
684 continue;
2746f7a1
DW
685 grp = NULL;
686 if (EXT4_SB(sb)->s_group_info)
687 grp = ext4_get_group_info(sb, i);
688 if (!grp || !EXT4_MB_GRP_BBITMAP_CORRUPT(grp))
689 desc_count += ext4_free_group_clusters(sb, gdp);
ac27a0ec 690 brelse(bitmap_bh);
574ca174 691 bitmap_bh = ext4_read_block_bitmap(sb, i);
9008a58e
DW
692 if (IS_ERR(bitmap_bh)) {
693 bitmap_bh = NULL;
ac27a0ec 694 continue;
9008a58e 695 }
ac27a0ec 696
f6fb99ca 697 x = ext4_count_free(bitmap_bh->b_data,
036acea2 698 EXT4_CLUSTERS_PER_GROUP(sb) / 8);
9fd9784c 699 printk(KERN_DEBUG "group %u: stored = %d, counted = %u\n",
021b65bb 700 i, ext4_free_group_clusters(sb, gdp), x);
ac27a0ec
DK
701 bitmap_count += x;
702 }
703 brelse(bitmap_bh);
5dee5437
TT
704 printk(KERN_DEBUG "ext4_count_free_clusters: stored = %llu"
705 ", computed = %llu, %llu\n",
810da240 706 EXT4_NUM_B2C(EXT4_SB(sb), ext4_free_blocks_count(es)),
4776004f 707 desc_count, bitmap_count);
ac27a0ec
DK
708 return bitmap_count;
709#else
710 desc_count = 0;
ac27a0ec 711 for (i = 0; i < ngroups; i++) {
617ba13b 712 gdp = ext4_get_group_desc(sb, i, NULL);
ac27a0ec
DK
713 if (!gdp)
714 continue;
2746f7a1
DW
715 grp = NULL;
716 if (EXT4_SB(sb)->s_group_info)
717 grp = ext4_get_group_info(sb, i);
718 if (!grp || !EXT4_MB_GRP_BBITMAP_CORRUPT(grp))
719 desc_count += ext4_free_group_clusters(sb, gdp);
ac27a0ec
DK
720 }
721
722 return desc_count;
723#endif
724}
725
fd2d4291 726static inline int test_root(ext4_group_t a, int b)
ac27a0ec 727{
f4afb4f4
TT
728 while (1) {
729 if (a < b)
730 return 0;
731 if (a == b)
732 return 1;
733 if ((a % b) != 0)
734 return 0;
735 a = a / b;
736 }
ac27a0ec
DK
737}
738
ac27a0ec 739/**
617ba13b 740 * ext4_bg_has_super - number of blocks used by the superblock in group
ac27a0ec
DK
741 * @sb: superblock for filesystem
742 * @group: group number to check
743 *
744 * Return the number of blocks used by the superblock (primary or backup)
745 * in this group. Currently this will be only 0 or 1.
746 */
fd2d4291 747int ext4_bg_has_super(struct super_block *sb, ext4_group_t group)
ac27a0ec 748{
1beeef1b
DW
749 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
750
751 if (group == 0)
752 return 1;
e2b911c5 753 if (ext4_has_feature_sparse_super2(sb)) {
1beeef1b
DW
754 if (group == le32_to_cpu(es->s_backup_bgs[0]) ||
755 group == le32_to_cpu(es->s_backup_bgs[1]))
756 return 1;
757 return 0;
758 }
e2b911c5 759 if ((group <= 1) || !ext4_has_feature_sparse_super(sb))
1beeef1b
DW
760 return 1;
761 if (!(group & 1))
ac27a0ec 762 return 0;
1beeef1b
DW
763 if (test_root(group, 3) || (test_root(group, 5)) ||
764 test_root(group, 7))
765 return 1;
766
767 return 0;
ac27a0ec
DK
768}
769
fd2d4291
AM
770static unsigned long ext4_bg_num_gdb_meta(struct super_block *sb,
771 ext4_group_t group)
ac27a0ec 772{
617ba13b 773 unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
fd2d4291
AM
774 ext4_group_t first = metagroup * EXT4_DESC_PER_BLOCK(sb);
775 ext4_group_t last = first + EXT4_DESC_PER_BLOCK(sb) - 1;
ac27a0ec
DK
776
777 if (group == first || group == first + 1 || group == last)
778 return 1;
779 return 0;
780}
781
fd2d4291
AM
782static unsigned long ext4_bg_num_gdb_nometa(struct super_block *sb,
783 ext4_group_t group)
ac27a0ec 784{
8dadb198
TT
785 if (!ext4_bg_has_super(sb, group))
786 return 0;
787
e2b911c5 788 if (ext4_has_feature_meta_bg(sb))
8dadb198
TT
789 return le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg);
790 else
791 return EXT4_SB(sb)->s_gdb_count;
ac27a0ec
DK
792}
793
794/**
617ba13b 795 * ext4_bg_num_gdb - number of blocks used by the group table in group
ac27a0ec
DK
796 * @sb: superblock for filesystem
797 * @group: group number to check
798 *
799 * Return the number of blocks used by the group descriptor table
800 * (primary or backup) in this group. In the future there may be a
801 * different number of descriptor blocks in each group.
802 */
fd2d4291 803unsigned long ext4_bg_num_gdb(struct super_block *sb, ext4_group_t group)
ac27a0ec
DK
804{
805 unsigned long first_meta_bg =
617ba13b
MC
806 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg);
807 unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
ac27a0ec 808
e2b911c5 809 if (!ext4_has_feature_meta_bg(sb) || metagroup < first_meta_bg)
af5bc92d 810 return ext4_bg_num_gdb_nometa(sb, group);
ac27a0ec 811
617ba13b 812 return ext4_bg_num_gdb_meta(sb,group);
ac27a0ec
DK
813
814}
c2ea3fde 815
49f7f9af 816/*
d5b8f310 817 * This function returns the number of file system metadata clusters at
49f7f9af
TT
818 * the beginning of a block group, including the reserved gdt blocks.
819 */
5f163cc7 820static unsigned ext4_num_base_meta_clusters(struct super_block *sb,
d5b8f310 821 ext4_group_t block_group)
49f7f9af
TT
822{
823 struct ext4_sb_info *sbi = EXT4_SB(sb);
d5b8f310 824 unsigned num;
49f7f9af
TT
825
826 /* Check for superblock and gdt backups in this group */
827 num = ext4_bg_has_super(sb, block_group);
828
e2b911c5 829 if (!ext4_has_feature_meta_bg(sb) ||
49f7f9af
TT
830 block_group < le32_to_cpu(sbi->s_es->s_first_meta_bg) *
831 sbi->s_desc_per_block) {
832 if (num) {
833 num += ext4_bg_num_gdb(sb, block_group);
834 num += le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks);
835 }
836 } else { /* For META_BG_BLOCK_GROUPS */
837 num += ext4_bg_num_gdb(sb, block_group);
838 }
d5b8f310 839 return EXT4_NUM_B2C(sbi, num);
49f7f9af 840}
f86186b4
ES
841/**
842 * ext4_inode_to_goal_block - return a hint for block allocation
843 * @inode: inode for block allocation
844 *
845 * Return the ideal location to start allocating blocks for a
846 * newly created inode.
847 */
848ext4_fsblk_t ext4_inode_to_goal_block(struct inode *inode)
849{
850 struct ext4_inode_info *ei = EXT4_I(inode);
851 ext4_group_t block_group;
852 ext4_grpblk_t colour;
853 int flex_size = ext4_flex_bg_size(EXT4_SB(inode->i_sb));
854 ext4_fsblk_t bg_start;
855 ext4_fsblk_t last_block;
856
857 block_group = ei->i_block_group;
858 if (flex_size >= EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME) {
859 /*
860 * If there are at least EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME
861 * block groups per flexgroup, reserve the first block
862 * group for directories and special files. Regular
863 * files will start at the second block group. This
864 * tends to speed up directory access and improves
865 * fsck times.
866 */
867 block_group &= ~(flex_size-1);
868 if (S_ISREG(inode->i_mode))
869 block_group++;
870 }
871 bg_start = ext4_group_first_block_no(inode->i_sb, block_group);
872 last_block = ext4_blocks_count(EXT4_SB(inode->i_sb)->s_es) - 1;
873
874 /*
875 * If we are doing delayed allocation, we don't need take
876 * colour into account.
877 */
878 if (test_opt(inode->i_sb, DELALLOC))
879 return bg_start;
880
881 if (bg_start + EXT4_BLOCKS_PER_GROUP(inode->i_sb) <= last_block)
882 colour = (current->pid % 16) *
883 (EXT4_BLOCKS_PER_GROUP(inode->i_sb) / 16);
884 else
885 colour = (current->pid % 16) * ((last_block - bg_start) / 16);
886 return bg_start + colour;
887}
888
This page took 0.583125 seconds and 5 git commands to generate.