Merge tag 'regulator-v3.8-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/broon...
[GitHub/mt8127/android_kernel_alcatel_ttab.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>
dab291af 17#include <linux/jbd2.h>
ac27a0ec
DK
18#include <linux/quotaops.h>
19#include <linux/buffer_head.h>
3dcf5451
CH
20#include "ext4.h"
21#include "ext4_jbd2.h"
e21675d4 22#include "mballoc.h"
3dcf5451 23
0562e0ba
JZ
24#include <trace/events/ext4.h>
25
5f163cc7
ES
26static unsigned ext4_num_base_meta_clusters(struct super_block *sb,
27 ext4_group_t block_group);
ac27a0ec
DK
28/*
29 * balloc.c contains the blocks allocation and deallocation routines
30 */
31
72b64b59 32/*
3212a80a
TT
33 * Calculate the block group number and offset into the block/cluster
34 * allocation bitmap, given a block number
72b64b59
AM
35 */
36void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
fd2d4291 37 ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp)
72b64b59 38{
8c55e204 39 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
72b64b59
AM
40 ext4_grpblk_t offset;
41
8c55e204 42 blocknr = blocknr - le32_to_cpu(es->s_first_data_block);
3212a80a
TT
43 offset = do_div(blocknr, EXT4_BLOCKS_PER_GROUP(sb)) >>
44 EXT4_SB(sb)->s_cluster_bits;
72b64b59
AM
45 if (offsetp)
46 *offsetp = offset;
47 if (blockgrpp)
8c55e204 48 *blockgrpp = blocknr;
72b64b59
AM
49
50}
51
0bf7e837
JS
52static int ext4_block_in_group(struct super_block *sb, ext4_fsblk_t block,
53 ext4_group_t block_group)
54{
55 ext4_group_t actual_group;
7477827f 56 ext4_get_group_no_and_offset(sb, block, &actual_group, NULL);
0bf7e837
JS
57 if (actual_group == block_group)
58 return 1;
59 return 0;
60}
61
d5b8f310
TT
62/* Return the number of clusters used for file system metadata; this
63 * represents the overhead needed by the file system.
64 */
65unsigned ext4_num_overhead_clusters(struct super_block *sb,
66 ext4_group_t block_group,
67 struct ext4_group_desc *gdp)
0bf7e837 68{
d5b8f310
TT
69 unsigned num_clusters;
70 int block_cluster = -1, inode_cluster = -1, itbl_cluster = -1, i, c;
71 ext4_fsblk_t start = ext4_group_first_block_no(sb, block_group);
72 ext4_fsblk_t itbl_blk;
0bf7e837 73 struct ext4_sb_info *sbi = EXT4_SB(sb);
0bf7e837 74
d5b8f310
TT
75 /* This is the number of clusters used by the superblock,
76 * block group descriptors, and reserved block group
77 * descriptor blocks */
78 num_clusters = ext4_num_base_meta_clusters(sb, block_group);
79
80 /*
81 * For the allocation bitmaps and inode table, we first need
82 * to check to see if the block is in the block group. If it
83 * is, then check to see if the cluster is already accounted
84 * for in the clusters used for the base metadata cluster, or
85 * if we can increment the base metadata cluster to include
86 * that block. Otherwise, we will have to track the cluster
87 * used for the allocation bitmap or inode table explicitly.
88 * Normally all of these blocks are contiguous, so the special
89 * case handling shouldn't be necessary except for *very*
90 * unusual file system layouts.
91 */
92 if (ext4_block_in_group(sb, ext4_block_bitmap(sb, gdp), block_group)) {
b0dd6b70
TT
93 block_cluster = EXT4_B2C(sbi,
94 ext4_block_bitmap(sb, gdp) - start);
d5b8f310
TT
95 if (block_cluster < num_clusters)
96 block_cluster = -1;
97 else if (block_cluster == num_clusters) {
98 num_clusters++;
99 block_cluster = -1;
100 }
101 }
102
103 if (ext4_block_in_group(sb, ext4_inode_bitmap(sb, gdp), block_group)) {
104 inode_cluster = EXT4_B2C(sbi,
b0dd6b70 105 ext4_inode_bitmap(sb, gdp) - start);
d5b8f310
TT
106 if (inode_cluster < num_clusters)
107 inode_cluster = -1;
108 else if (inode_cluster == num_clusters) {
109 num_clusters++;
110 inode_cluster = -1;
111 }
112 }
113
114 itbl_blk = ext4_inode_table(sb, gdp);
115 for (i = 0; i < sbi->s_itb_per_group; i++) {
116 if (ext4_block_in_group(sb, itbl_blk + i, block_group)) {
b0dd6b70 117 c = EXT4_B2C(sbi, itbl_blk + i - start);
d5b8f310
TT
118 if ((c < num_clusters) || (c == inode_cluster) ||
119 (c == block_cluster) || (c == itbl_cluster))
120 continue;
121 if (c == num_clusters) {
122 num_clusters++;
123 continue;
124 }
125 num_clusters++;
126 itbl_cluster = c;
0bf7e837
JS
127 }
128 }
d5b8f310
TT
129
130 if (block_cluster != -1)
131 num_clusters++;
132 if (inode_cluster != -1)
133 num_clusters++;
134
135 return num_clusters;
0bf7e837 136}
c2ea3fde 137
d5b8f310
TT
138static unsigned int num_clusters_in_group(struct super_block *sb,
139 ext4_group_t block_group)
49f7f9af 140{
d5b8f310
TT
141 unsigned int blocks;
142
49f7f9af
TT
143 if (block_group == ext4_get_groups_count(sb) - 1) {
144 /*
145 * Even though mke2fs always initializes the first and
146 * last group, just in case some other tool was used,
147 * we need to make sure we calculate the right free
148 * blocks.
149 */
d5b8f310 150 blocks = ext4_blocks_count(EXT4_SB(sb)->s_es) -
49f7f9af
TT
151 ext4_group_first_block_no(sb, block_group);
152 } else
d5b8f310
TT
153 blocks = EXT4_BLOCKS_PER_GROUP(sb);
154 return EXT4_NUM_B2C(EXT4_SB(sb), blocks);
49f7f9af
TT
155}
156
fd034a84
TT
157/* Initializes an uninitialized block bitmap */
158void ext4_init_block_bitmap(struct super_block *sb, struct buffer_head *bh,
159 ext4_group_t block_group,
160 struct ext4_group_desc *gdp)
717d50e4 161{
d5b8f310 162 unsigned int bit, bit_max;
717d50e4 163 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd034a84
TT
164 ext4_fsblk_t start, tmp;
165 int flex_bg = 0;
166
167 J_ASSERT_BH(bh, buffer_locked(bh));
168
169 /* If checksum is bad mark all blocks used to prevent allocation
170 * essentially implementing a per-group read-only flag. */
feb0ab32 171 if (!ext4_group_desc_csum_verify(sb, block_group, gdp)) {
fd034a84 172 ext4_error(sb, "Checksum bad for group %u", block_group);
021b65bb 173 ext4_free_group_clusters_set(sb, gdp, 0);
fd034a84
TT
174 ext4_free_inodes_set(sb, gdp, 0);
175 ext4_itable_unused_set(sb, gdp, 0);
176 memset(bh->b_data, 0xff, sb->s_blocksize);
79f1ba49 177 ext4_block_bitmap_csum_set(sb, block_group, gdp, bh);
fd034a84 178 return;
717d50e4 179 }
fd034a84 180 memset(bh->b_data, 0, sb->s_blocksize);
717d50e4 181
d5b8f310 182 bit_max = ext4_num_base_meta_clusters(sb, block_group);
fd034a84
TT
183 for (bit = 0; bit < bit_max; bit++)
184 ext4_set_bit(bit, bh->b_data);
d00a6d7b 185
fd034a84 186 start = ext4_group_first_block_no(sb, block_group);
717d50e4 187
fd034a84
TT
188 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
189 flex_bg = 1;
717d50e4 190
fd034a84
TT
191 /* Set bits for block and inode bitmaps, and inode table */
192 tmp = ext4_block_bitmap(sb, gdp);
193 if (!flex_bg || ext4_block_in_group(sb, tmp, block_group))
d5b8f310 194 ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data);
717d50e4 195
fd034a84
TT
196 tmp = ext4_inode_bitmap(sb, gdp);
197 if (!flex_bg || ext4_block_in_group(sb, tmp, block_group))
d5b8f310 198 ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data);
0bf7e837 199
fd034a84
TT
200 tmp = ext4_inode_table(sb, gdp);
201 for (; tmp < ext4_inode_table(sb, gdp) +
202 sbi->s_itb_per_group; tmp++) {
0bf7e837 203 if (!flex_bg || ext4_block_in_group(sb, tmp, block_group))
d5b8f310 204 ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data);
717d50e4 205 }
d5b8f310 206
fd034a84
TT
207 /*
208 * Also if the number of blocks within the group is less than
209 * the blocksize * 8 ( which is the size of bitmap ), set rest
210 * of the block bitmap to 1
211 */
d5b8f310 212 ext4_mark_bitmap_end(num_clusters_in_group(sb, block_group),
fd034a84 213 sb->s_blocksize * 8, bh->b_data);
79f1ba49 214 ext4_block_bitmap_csum_set(sb, block_group, gdp, bh);
feb0ab32 215 ext4_group_desc_csum_set(sb, block_group, gdp);
717d50e4
AD
216}
217
fd034a84
TT
218/* Return the number of free blocks in a block group. It is used when
219 * the block bitmap is uninitialized, so we can't just count the bits
220 * in the bitmap. */
cff1dfd7
TT
221unsigned ext4_free_clusters_after_init(struct super_block *sb,
222 ext4_group_t block_group,
223 struct ext4_group_desc *gdp)
fd034a84 224{
d5b8f310
TT
225 return num_clusters_in_group(sb, block_group) -
226 ext4_num_overhead_clusters(sb, block_group, gdp);
fd034a84 227}
717d50e4 228
ac27a0ec
DK
229/*
230 * The free blocks are managed by bitmaps. A file system contains several
231 * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap
232 * block for inodes, N blocks for the inode table and data blocks.
233 *
234 * The file system contains group descriptors which are located after the
235 * super block. Each descriptor contains the number of the bitmap block and
236 * the free blocks count in the block. The descriptors are loaded in memory
e627432c 237 * when a file system is mounted (see ext4_fill_super).
ac27a0ec
DK
238 */
239
ac27a0ec 240/**
617ba13b 241 * ext4_get_group_desc() -- load group descriptor from disk
ac27a0ec
DK
242 * @sb: super block
243 * @block_group: given block group
244 * @bh: pointer to the buffer head to store the block
245 * group descriptor
246 */
af5bc92d 247struct ext4_group_desc * ext4_get_group_desc(struct super_block *sb,
fd2d4291 248 ext4_group_t block_group,
af5bc92d 249 struct buffer_head **bh)
ac27a0ec 250{
498e5f24
TT
251 unsigned int group_desc;
252 unsigned int offset;
8df9675f 253 ext4_group_t ngroups = ext4_get_groups_count(sb);
af5bc92d 254 struct ext4_group_desc *desc;
617ba13b 255 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 256
8df9675f 257 if (block_group >= ngroups) {
12062ddd
ES
258 ext4_error(sb, "block_group >= groups_count - block_group = %u,"
259 " groups_count = %u", block_group, ngroups);
ac27a0ec
DK
260
261 return NULL;
262 }
ac27a0ec 263
617ba13b
MC
264 group_desc = block_group >> EXT4_DESC_PER_BLOCK_BITS(sb);
265 offset = block_group & (EXT4_DESC_PER_BLOCK(sb) - 1);
ac27a0ec 266 if (!sbi->s_group_desc[group_desc]) {
12062ddd 267 ext4_error(sb, "Group descriptor not loaded - "
498e5f24 268 "block_group = %u, group_desc = %u, desc = %u",
af5bc92d 269 block_group, group_desc, offset);
ac27a0ec
DK
270 return NULL;
271 }
272
0d1ee42f
AR
273 desc = (struct ext4_group_desc *)(
274 (__u8 *)sbi->s_group_desc[group_desc]->b_data +
275 offset * EXT4_DESC_SIZE(sb));
ac27a0ec
DK
276 if (bh)
277 *bh = sbi->s_group_desc[group_desc];
0d1ee42f 278 return desc;
ac27a0ec
DK
279}
280
7a4c5de2
TT
281/*
282 * Return the block number which was discovered to be invalid, or 0 if
283 * the block bitmap is valid.
284 */
285static ext4_fsblk_t ext4_valid_block_bitmap(struct super_block *sb,
286 struct ext4_group_desc *desc,
287 unsigned int block_group,
288 struct buffer_head *bh)
abcb2947
AK
289{
290 ext4_grpblk_t offset;
291 ext4_grpblk_t next_zero_bit;
7a4c5de2 292 ext4_fsblk_t blk;
abcb2947
AK
293 ext4_fsblk_t group_first_block;
294
295 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) {
296 /* with FLEX_BG, the inode/block bitmaps and itable
297 * blocks may not be in the group at all
298 * so the bitmap validation will be skipped for those groups
299 * or it has to also read the block group where the bitmaps
300 * are located to verify they are set.
301 */
7a4c5de2 302 return 0;
abcb2947
AK
303 }
304 group_first_block = ext4_group_first_block_no(sb, block_group);
305
306 /* check whether block bitmap block number is set */
7a4c5de2
TT
307 blk = ext4_block_bitmap(sb, desc);
308 offset = blk - group_first_block;
abcb2947
AK
309 if (!ext4_test_bit(offset, bh->b_data))
310 /* bad block bitmap */
7a4c5de2 311 return blk;
abcb2947
AK
312
313 /* check whether the inode bitmap block number is set */
7a4c5de2
TT
314 blk = ext4_inode_bitmap(sb, desc);
315 offset = blk - group_first_block;
abcb2947
AK
316 if (!ext4_test_bit(offset, bh->b_data))
317 /* bad block bitmap */
7a4c5de2 318 return blk;
abcb2947
AK
319
320 /* check whether the inode table block number is set */
7a4c5de2
TT
321 blk = ext4_inode_table(sb, desc);
322 offset = blk - group_first_block;
abcb2947
AK
323 next_zero_bit = ext4_find_next_zero_bit(bh->b_data,
324 offset + EXT4_SB(sb)->s_itb_per_group,
325 offset);
7a4c5de2
TT
326 if (next_zero_bit < offset + EXT4_SB(sb)->s_itb_per_group)
327 /* bad bitmap for inode tables */
328 return blk;
abcb2947
AK
329 return 0;
330}
fa77dcfa
DW
331
332void ext4_validate_block_bitmap(struct super_block *sb,
333 struct ext4_group_desc *desc,
334 unsigned int block_group,
335 struct buffer_head *bh)
336{
7a4c5de2
TT
337 ext4_fsblk_t blk;
338
fa77dcfa
DW
339 if (buffer_verified(bh))
340 return;
341
342 ext4_lock_group(sb, block_group);
7a4c5de2
TT
343 blk = ext4_valid_block_bitmap(sb, desc, block_group, bh);
344 if (unlikely(blk != 0)) {
345 ext4_unlock_group(sb, block_group);
346 ext4_error(sb, "bg %u: block %llu: invalid block bitmap",
347 block_group, blk);
348 return;
349 }
350 if (unlikely(!ext4_block_bitmap_csum_verify(sb, block_group,
79f1ba49 351 desc, bh))) {
7a4c5de2
TT
352 ext4_unlock_group(sb, block_group);
353 ext4_error(sb, "bg %u: bad block bitmap checksum", block_group);
354 return;
355 }
356 set_buffer_verified(bh);
fa77dcfa
DW
357 ext4_unlock_group(sb, block_group);
358}
359
ac27a0ec 360/**
574ca174 361 * ext4_read_block_bitmap()
ac27a0ec
DK
362 * @sb: super block
363 * @block_group: given block group
364 *
abcb2947
AK
365 * Read the bitmap for a given block_group,and validate the
366 * bits for block/inode/inode tables are set in the bitmaps
ac27a0ec
DK
367 *
368 * Return buffer_head on success or NULL in case of failure.
369 */
717d50e4 370struct buffer_head *
813e5727 371ext4_read_block_bitmap_nowait(struct super_block *sb, ext4_group_t block_group)
ac27a0ec 372{
af5bc92d 373 struct ext4_group_desc *desc;
813e5727 374 struct buffer_head *bh;
7c9e69fa 375 ext4_fsblk_t bitmap_blk;
ac27a0ec 376
717d50e4 377 desc = ext4_get_group_desc(sb, block_group, NULL);
ac27a0ec 378 if (!desc)
7c9e69fa
AK
379 return NULL;
380 bitmap_blk = ext4_block_bitmap(sb, desc);
abcb2947
AK
381 bh = sb_getblk(sb, bitmap_blk);
382 if (unlikely(!bh)) {
813e5727
TT
383 ext4_error(sb, "Cannot get buffer for block bitmap - "
384 "block_group = %u, block_bitmap = %llu",
385 block_group, bitmap_blk);
abcb2947
AK
386 return NULL;
387 }
2ccb5fb9
AK
388
389 if (bitmap_uptodate(bh))
fa77dcfa 390 goto verify;
abcb2947 391
c806e68f 392 lock_buffer(bh);
2ccb5fb9
AK
393 if (bitmap_uptodate(bh)) {
394 unlock_buffer(bh);
fa77dcfa 395 goto verify;
2ccb5fb9 396 }
955ce5f5 397 ext4_lock_group(sb, block_group);
717d50e4 398 if (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) {
abcb2947 399 ext4_init_block_bitmap(sb, bh, block_group, desc);
2ccb5fb9 400 set_bitmap_uptodate(bh);
abcb2947 401 set_buffer_uptodate(bh);
955ce5f5 402 ext4_unlock_group(sb, block_group);
3300beda 403 unlock_buffer(bh);
abcb2947 404 return bh;
717d50e4 405 }
955ce5f5 406 ext4_unlock_group(sb, block_group);
2ccb5fb9
AK
407 if (buffer_uptodate(bh)) {
408 /*
409 * if not uninit if bh is uptodate,
410 * bitmap is also uptodate
411 */
412 set_bitmap_uptodate(bh);
413 unlock_buffer(bh);
fa77dcfa 414 goto verify;
2ccb5fb9
AK
415 }
416 /*
813e5727 417 * submit the buffer_head for reading
2ccb5fb9 418 */
813e5727 419 set_buffer_new(bh);
0562e0ba 420 trace_ext4_read_block_bitmap_load(sb, block_group);
813e5727
TT
421 bh->b_end_io = ext4_end_bitmap_read;
422 get_bh(bh);
423 submit_bh(READ, bh);
424 return bh;
fa77dcfa
DW
425verify:
426 ext4_validate_block_bitmap(sb, desc, block_group, bh);
427 return bh;
813e5727
TT
428}
429
430/* Returns 0 on success, 1 on error */
431int ext4_wait_block_bitmap(struct super_block *sb, ext4_group_t block_group,
432 struct buffer_head *bh)
433{
434 struct ext4_group_desc *desc;
435
436 if (!buffer_new(bh))
437 return 0;
438 desc = ext4_get_group_desc(sb, block_group, NULL);
439 if (!desc)
440 return 1;
441 wait_on_buffer(bh);
442 if (!buffer_uptodate(bh)) {
12062ddd 443 ext4_error(sb, "Cannot read block bitmap - "
813e5727 444 "block_group = %u, block_bitmap = %llu",
d4dc462f 445 block_group, (unsigned long long) bh->b_blocknr);
813e5727 446 return 1;
abcb2947 447 }
813e5727
TT
448 clear_buffer_new(bh);
449 /* Panic or remount fs read-only if block bitmap is invalid */
fa77dcfa 450 ext4_validate_block_bitmap(sb, desc, block_group, bh);
813e5727
TT
451 return 0;
452}
453
454struct buffer_head *
455ext4_read_block_bitmap(struct super_block *sb, ext4_group_t block_group)
456{
457 struct buffer_head *bh;
458
459 bh = ext4_read_block_bitmap_nowait(sb, block_group);
460 if (ext4_wait_block_bitmap(sb, block_group, bh)) {
461 put_bh(bh);
462 return NULL;
463 }
ac27a0ec
DK
464 return bh;
465}
ac27a0ec 466
8c3bf8a0 467/**
df55c99d 468 * ext4_has_free_clusters()
8c3bf8a0 469 * @sbi: in-core super block structure.
df55c99d
TT
470 * @nclusters: number of needed blocks
471 * @flags: flags from ext4_mb_new_blocks()
8c3bf8a0 472 *
df55c99d 473 * Check if filesystem has nclusters free & available for allocation.
8c3bf8a0
ES
474 * On success return 1, return 0 on failure.
475 */
df55c99d
TT
476static int ext4_has_free_clusters(struct ext4_sb_info *sbi,
477 s64 nclusters, unsigned int flags)
a30d542a 478{
df55c99d 479 s64 free_clusters, dirty_clusters, root_clusters;
57042651 480 struct percpu_counter *fcc = &sbi->s_freeclusters_counter;
df55c99d 481 struct percpu_counter *dcc = &sbi->s_dirtyclusters_counter;
a30d542a 482
df55c99d
TT
483 free_clusters = percpu_counter_read_positive(fcc);
484 dirty_clusters = percpu_counter_read_positive(dcc);
485 root_clusters = EXT4_B2C(sbi, ext4_r_blocks_count(sbi->s_es));
a30d542a 486
df55c99d
TT
487 if (free_clusters - (nclusters + root_clusters + dirty_clusters) <
488 EXT4_FREECLUSTERS_WATERMARK) {
489 free_clusters = EXT4_C2B(sbi, percpu_counter_sum_positive(fcc));
490 dirty_clusters = percpu_counter_sum_positive(dcc);
6bc6e63f 491 }
df55c99d
TT
492 /* Check whether we have space after accounting for current
493 * dirty clusters & root reserved clusters.
6bc6e63f 494 */
df55c99d 495 if (free_clusters >= ((root_clusters + nclusters) + dirty_clusters))
a996031c 496 return 1;
a30d542a 497
df55c99d 498 /* Hm, nope. Are (enough) root reserved clusters available? */
08cefc7a
EB
499 if (uid_eq(sbi->s_resuid, current_fsuid()) ||
500 (!gid_eq(sbi->s_resgid, GLOBAL_ROOT_GID) && in_group_p(sbi->s_resgid)) ||
55f020db
AH
501 capable(CAP_SYS_RESOURCE) ||
502 (flags & EXT4_MB_USE_ROOT_BLOCKS)) {
503
df55c99d 504 if (free_clusters >= (nclusters + dirty_clusters))
a996031c
ES
505 return 1;
506 }
507
508 return 0;
a30d542a
AK
509}
510
e7d5f315
TT
511int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
512 s64 nclusters, unsigned int flags)
ac27a0ec 513{
df55c99d 514 if (ext4_has_free_clusters(sbi, nclusters, flags)) {
e7d5f315 515 percpu_counter_add(&sbi->s_dirtyclusters_counter, nclusters);
16eb7295 516 return 0;
8c3bf8a0
ES
517 } else
518 return -ENOSPC;
a30d542a 519}
07031431 520
ac27a0ec 521/**
617ba13b 522 * ext4_should_retry_alloc()
ac27a0ec
DK
523 * @sb: super block
524 * @retries number of attemps has been made
525 *
617ba13b 526 * ext4_should_retry_alloc() is called when ENOSPC is returned, and if
ac27a0ec 527 * it is profitable to retry the operation, this function will wait
25985edc 528 * for the current or committing transaction to complete, and then
ac27a0ec
DK
529 * return TRUE.
530 *
531 * if the total number of retries exceed three times, return FALSE.
532 */
617ba13b 533int ext4_should_retry_alloc(struct super_block *sb, int *retries)
ac27a0ec 534{
df55c99d 535 if (!ext4_has_free_clusters(EXT4_SB(sb), 1, 0) ||
8f64b32e
ES
536 (*retries)++ > 3 ||
537 !EXT4_SB(sb)->s_journal)
ac27a0ec
DK
538 return 0;
539
540 jbd_debug(1, "%s: retrying operation after ENOSPC\n", sb->s_id);
541
dab291af 542 return jbd2_journal_force_commit_nested(EXT4_SB(sb)->s_journal);
ac27a0ec
DK
543}
544
654b4908 545/*
d2a17637 546 * ext4_new_meta_blocks() -- allocate block for meta data (indexing) blocks
654b4908
AK
547 *
548 * @handle: handle to this transaction
549 * @inode: file inode
550 * @goal: given target block(filesystem wide)
7b415bf6 551 * @count: pointer to total number of clusters needed
654b4908
AK
552 * @errp: error code
553 *
97df5d15 554 * Return 1st allocated block number on success, *count stores total account
d2a17637 555 * error stores in errp pointer
654b4908 556 */
d2a17637 557ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
55f020db
AH
558 ext4_fsblk_t goal, unsigned int flags,
559 unsigned long *count, int *errp)
654b4908 560{
97df5d15 561 struct ext4_allocation_request ar;
d2a17637 562 ext4_fsblk_t ret;
97df5d15
TT
563
564 memset(&ar, 0, sizeof(ar));
565 /* Fill with neighbour allocated blocks */
566 ar.inode = inode;
567 ar.goal = goal;
568 ar.len = count ? *count : 1;
55f020db 569 ar.flags = flags;
97df5d15
TT
570
571 ret = ext4_mb_new_blocks(handle, &ar, errp);
572 if (count)
573 *count = ar.len;
d2a17637 574 /*
72b8ab9d
ES
575 * Account for the allocated meta blocks. We will never
576 * fail EDQUOT for metdata, but we do account for it.
d2a17637 577 */
f2321097
TT
578 if (!(*errp) &&
579 ext4_test_inode_state(inode, EXT4_STATE_DELALLOC_RESERVED)) {
d2a17637 580 spin_lock(&EXT4_I(inode)->i_block_reservation_lock);
97df5d15 581 EXT4_I(inode)->i_allocated_meta_blocks += ar.len;
d2a17637 582 spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
7b415bf6
AK
583 dquot_alloc_block_nofail(inode,
584 EXT4_C2B(EXT4_SB(inode->i_sb), ar.len));
d2a17637
MC
585 }
586 return ret;
654b4908
AK
587}
588
ac27a0ec 589/**
5dee5437 590 * ext4_count_free_clusters() -- count filesystem free clusters
ac27a0ec
DK
591 * @sb: superblock
592 *
5dee5437 593 * Adds up the number of free clusters from each block group.
ac27a0ec 594 */
5dee5437 595ext4_fsblk_t ext4_count_free_clusters(struct super_block *sb)
ac27a0ec 596{
617ba13b
MC
597 ext4_fsblk_t desc_count;
598 struct ext4_group_desc *gdp;
fd2d4291 599 ext4_group_t i;
8df9675f 600 ext4_group_t ngroups = ext4_get_groups_count(sb);
617ba13b
MC
601#ifdef EXT4FS_DEBUG
602 struct ext4_super_block *es;
603 ext4_fsblk_t bitmap_count;
498e5f24 604 unsigned int x;
ac27a0ec
DK
605 struct buffer_head *bitmap_bh = NULL;
606
617ba13b 607 es = EXT4_SB(sb)->s_es;
ac27a0ec
DK
608 desc_count = 0;
609 bitmap_count = 0;
610 gdp = NULL;
611
ac27a0ec 612 for (i = 0; i < ngroups; i++) {
617ba13b 613 gdp = ext4_get_group_desc(sb, i, NULL);
ac27a0ec
DK
614 if (!gdp)
615 continue;
021b65bb 616 desc_count += ext4_free_group_clusters(sb, gdp);
ac27a0ec 617 brelse(bitmap_bh);
574ca174 618 bitmap_bh = ext4_read_block_bitmap(sb, i);
ac27a0ec
DK
619 if (bitmap_bh == NULL)
620 continue;
621
f6fb99ca
TT
622 x = ext4_count_free(bitmap_bh->b_data,
623 EXT4_BLOCKS_PER_GROUP(sb) / 8);
9fd9784c 624 printk(KERN_DEBUG "group %u: stored = %d, counted = %u\n",
021b65bb 625 i, ext4_free_group_clusters(sb, gdp), x);
ac27a0ec
DK
626 bitmap_count += x;
627 }
628 brelse(bitmap_bh);
5dee5437
TT
629 printk(KERN_DEBUG "ext4_count_free_clusters: stored = %llu"
630 ", computed = %llu, %llu\n",
6e58ad69 631 EXT4_B2C(EXT4_SB(sb), ext4_free_blocks_count(es)),
4776004f 632 desc_count, bitmap_count);
ac27a0ec
DK
633 return bitmap_count;
634#else
635 desc_count = 0;
ac27a0ec 636 for (i = 0; i < ngroups; i++) {
617ba13b 637 gdp = ext4_get_group_desc(sb, i, NULL);
ac27a0ec
DK
638 if (!gdp)
639 continue;
021b65bb 640 desc_count += ext4_free_group_clusters(sb, gdp);
ac27a0ec
DK
641 }
642
643 return desc_count;
644#endif
645}
646
fd2d4291 647static inline int test_root(ext4_group_t a, int b)
ac27a0ec
DK
648{
649 int num = b;
650
651 while (a > num)
652 num *= b;
653 return num == a;
654}
655
fd2d4291 656static int ext4_group_sparse(ext4_group_t group)
ac27a0ec
DK
657{
658 if (group <= 1)
659 return 1;
660 if (!(group & 1))
661 return 0;
662 return (test_root(group, 7) || test_root(group, 5) ||
663 test_root(group, 3));
664}
665
666/**
617ba13b 667 * ext4_bg_has_super - number of blocks used by the superblock in group
ac27a0ec
DK
668 * @sb: superblock for filesystem
669 * @group: group number to check
670 *
671 * Return the number of blocks used by the superblock (primary or backup)
672 * in this group. Currently this will be only 0 or 1.
673 */
fd2d4291 674int ext4_bg_has_super(struct super_block *sb, ext4_group_t group)
ac27a0ec 675{
617ba13b
MC
676 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
677 EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER) &&
678 !ext4_group_sparse(group))
ac27a0ec
DK
679 return 0;
680 return 1;
681}
682
fd2d4291
AM
683static unsigned long ext4_bg_num_gdb_meta(struct super_block *sb,
684 ext4_group_t group)
ac27a0ec 685{
617ba13b 686 unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
fd2d4291
AM
687 ext4_group_t first = metagroup * EXT4_DESC_PER_BLOCK(sb);
688 ext4_group_t last = first + EXT4_DESC_PER_BLOCK(sb) - 1;
ac27a0ec
DK
689
690 if (group == first || group == first + 1 || group == last)
691 return 1;
692 return 0;
693}
694
fd2d4291
AM
695static unsigned long ext4_bg_num_gdb_nometa(struct super_block *sb,
696 ext4_group_t group)
ac27a0ec 697{
8dadb198
TT
698 if (!ext4_bg_has_super(sb, group))
699 return 0;
700
701 if (EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG))
702 return le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg);
703 else
704 return EXT4_SB(sb)->s_gdb_count;
ac27a0ec
DK
705}
706
707/**
617ba13b 708 * ext4_bg_num_gdb - number of blocks used by the group table in group
ac27a0ec
DK
709 * @sb: superblock for filesystem
710 * @group: group number to check
711 *
712 * Return the number of blocks used by the group descriptor table
713 * (primary or backup) in this group. In the future there may be a
714 * different number of descriptor blocks in each group.
715 */
fd2d4291 716unsigned long ext4_bg_num_gdb(struct super_block *sb, ext4_group_t group)
ac27a0ec
DK
717{
718 unsigned long first_meta_bg =
617ba13b
MC
719 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg);
720 unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
ac27a0ec 721
617ba13b 722 if (!EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 723 metagroup < first_meta_bg)
af5bc92d 724 return ext4_bg_num_gdb_nometa(sb, group);
ac27a0ec 725
617ba13b 726 return ext4_bg_num_gdb_meta(sb,group);
ac27a0ec
DK
727
728}
c2ea3fde 729
49f7f9af 730/*
d5b8f310 731 * This function returns the number of file system metadata clusters at
49f7f9af
TT
732 * the beginning of a block group, including the reserved gdt blocks.
733 */
5f163cc7 734static unsigned ext4_num_base_meta_clusters(struct super_block *sb,
d5b8f310 735 ext4_group_t block_group)
49f7f9af
TT
736{
737 struct ext4_sb_info *sbi = EXT4_SB(sb);
d5b8f310 738 unsigned num;
49f7f9af
TT
739
740 /* Check for superblock and gdt backups in this group */
741 num = ext4_bg_has_super(sb, block_group);
742
743 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
744 block_group < le32_to_cpu(sbi->s_es->s_first_meta_bg) *
745 sbi->s_desc_per_block) {
746 if (num) {
747 num += ext4_bg_num_gdb(sb, block_group);
748 num += le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks);
749 }
750 } else { /* For META_BG_BLOCK_GROUPS */
751 num += ext4_bg_num_gdb(sb, block_group);
752 }
d5b8f310 753 return EXT4_NUM_B2C(sbi, num);
49f7f9af 754}
f86186b4
ES
755/**
756 * ext4_inode_to_goal_block - return a hint for block allocation
757 * @inode: inode for block allocation
758 *
759 * Return the ideal location to start allocating blocks for a
760 * newly created inode.
761 */
762ext4_fsblk_t ext4_inode_to_goal_block(struct inode *inode)
763{
764 struct ext4_inode_info *ei = EXT4_I(inode);
765 ext4_group_t block_group;
766 ext4_grpblk_t colour;
767 int flex_size = ext4_flex_bg_size(EXT4_SB(inode->i_sb));
768 ext4_fsblk_t bg_start;
769 ext4_fsblk_t last_block;
770
771 block_group = ei->i_block_group;
772 if (flex_size >= EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME) {
773 /*
774 * If there are at least EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME
775 * block groups per flexgroup, reserve the first block
776 * group for directories and special files. Regular
777 * files will start at the second block group. This
778 * tends to speed up directory access and improves
779 * fsck times.
780 */
781 block_group &= ~(flex_size-1);
782 if (S_ISREG(inode->i_mode))
783 block_group++;
784 }
785 bg_start = ext4_group_first_block_no(inode->i_sb, block_group);
786 last_block = ext4_blocks_count(EXT4_SB(inode->i_sb)->s_es) - 1;
787
788 /*
789 * If we are doing delayed allocation, we don't need take
790 * colour into account.
791 */
792 if (test_opt(inode->i_sb, DELALLOC))
793 return bg_start;
794
795 if (bg_start + EXT4_BLOCKS_PER_GROUP(inode->i_sb) <= last_block)
796 colour = (current->pid % 16) *
797 (EXT4_BLOCKS_PER_GROUP(inode->i_sb) / 16);
798 else
799 colour = (current->pid % 16) * ((last_block - bg_start) / 16);
800 return bg_start + colour;
801}
802