]> bbs.cooldavid.org Git - net-next-2.6.git/blame - fs/ext4/balloc.c
ext4: make various ext4 functions be static
[net-next-2.6.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
ac27a0ec
DK
24/*
25 * balloc.c contains the blocks allocation and deallocation routines
26 */
27
72b64b59
AM
28/*
29 * Calculate the block group number and offset, given a block number
30 */
31void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
fd2d4291 32 ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp)
72b64b59 33{
8c55e204 34 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
72b64b59
AM
35 ext4_grpblk_t offset;
36
8c55e204 37 blocknr = blocknr - le32_to_cpu(es->s_first_data_block);
f4e5bc24 38 offset = do_div(blocknr, EXT4_BLOCKS_PER_GROUP(sb));
72b64b59
AM
39 if (offsetp)
40 *offsetp = offset;
41 if (blockgrpp)
8c55e204 42 *blockgrpp = blocknr;
72b64b59
AM
43
44}
45
0bf7e837
JS
46static int ext4_block_in_group(struct super_block *sb, ext4_fsblk_t block,
47 ext4_group_t block_group)
48{
49 ext4_group_t actual_group;
7477827f 50 ext4_get_group_no_and_offset(sb, block, &actual_group, NULL);
0bf7e837
JS
51 if (actual_group == block_group)
52 return 1;
53 return 0;
54}
55
56static int ext4_group_used_meta_blocks(struct super_block *sb,
e187c658
TT
57 ext4_group_t block_group,
58 struct ext4_group_desc *gdp)
0bf7e837
JS
59{
60 ext4_fsblk_t tmp;
61 struct ext4_sb_info *sbi = EXT4_SB(sb);
62 /* block bitmap, inode bitmap, and inode table blocks */
63 int used_blocks = sbi->s_itb_per_group + 2;
64
65 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) {
0bf7e837
JS
66 if (!ext4_block_in_group(sb, ext4_block_bitmap(sb, gdp),
67 block_group))
68 used_blocks--;
69
70 if (!ext4_block_in_group(sb, ext4_inode_bitmap(sb, gdp),
71 block_group))
72 used_blocks--;
73
74 tmp = ext4_inode_table(sb, gdp);
75 for (; tmp < ext4_inode_table(sb, gdp) +
76 sbi->s_itb_per_group; tmp++) {
77 if (!ext4_block_in_group(sb, tmp, block_group))
78 used_blocks -= 1;
79 }
80 }
81 return used_blocks;
82}
c2ea3fde 83
717d50e4
AD
84/* Initializes an uninitialized block bitmap if given, and returns the
85 * number of blocks free in the group. */
86unsigned ext4_init_block_bitmap(struct super_block *sb, struct buffer_head *bh,
fd2d4291 87 ext4_group_t block_group, struct ext4_group_desc *gdp)
717d50e4 88{
717d50e4 89 int bit, bit_max;
8df9675f 90 ext4_group_t ngroups = ext4_get_groups_count(sb);
717d50e4
AD
91 unsigned free_blocks, group_blocks;
92 struct ext4_sb_info *sbi = EXT4_SB(sb);
93
94 if (bh) {
95 J_ASSERT_BH(bh, buffer_locked(bh));
96
97 /* If checksum is bad mark all blocks used to prevent allocation
98 * essentially implementing a per-group read-only flag. */
99 if (!ext4_group_desc_csum_verify(sbi, block_group, gdp)) {
12062ddd
ES
100 ext4_error(sb, "Checksum bad for group %u",
101 block_group);
560671a0
AK
102 ext4_free_blks_set(sb, gdp, 0);
103 ext4_free_inodes_set(sb, gdp, 0);
104 ext4_itable_unused_set(sb, gdp, 0);
717d50e4
AD
105 memset(bh->b_data, 0xff, sb->s_blocksize);
106 return 0;
107 }
108 memset(bh->b_data, 0, sb->s_blocksize);
109 }
110
111 /* Check for superblock and gdt backups in this group */
112 bit_max = ext4_bg_has_super(sb, block_group);
113
114 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
115 block_group < le32_to_cpu(sbi->s_es->s_first_meta_bg) *
116 sbi->s_desc_per_block) {
117 if (bit_max) {
118 bit_max += ext4_bg_num_gdb(sb, block_group);
119 bit_max +=
120 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks);
121 }
122 } else { /* For META_BG_BLOCK_GROUPS */
6afd6707 123 bit_max += ext4_bg_num_gdb(sb, block_group);
717d50e4
AD
124 }
125
8df9675f 126 if (block_group == ngroups - 1) {
717d50e4
AD
127 /*
128 * Even though mke2fs always initialize first and last group
129 * if some other tool enabled the EXT4_BG_BLOCK_UNINIT we need
130 * to make sure we calculate the right free blocks
131 */
132 group_blocks = ext4_blocks_count(sbi->s_es) -
5661bd68 133 ext4_group_first_block_no(sb, ngroups - 1);
717d50e4
AD
134 } else {
135 group_blocks = EXT4_BLOCKS_PER_GROUP(sb);
136 }
137
138 free_blocks = group_blocks - bit_max;
139
140 if (bh) {
0bf7e837
JS
141 ext4_fsblk_t start, tmp;
142 int flex_bg = 0;
d00a6d7b 143
717d50e4
AD
144 for (bit = 0; bit < bit_max; bit++)
145 ext4_set_bit(bit, bh->b_data);
146
d00a6d7b 147 start = ext4_group_first_block_no(sb, block_group);
717d50e4 148
0bf7e837
JS
149 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
150 EXT4_FEATURE_INCOMPAT_FLEX_BG))
151 flex_bg = 1;
717d50e4 152
0bf7e837
JS
153 /* Set bits for block and inode bitmaps, and inode table */
154 tmp = ext4_block_bitmap(sb, gdp);
155 if (!flex_bg || ext4_block_in_group(sb, tmp, block_group))
156 ext4_set_bit(tmp - start, bh->b_data);
157
158 tmp = ext4_inode_bitmap(sb, gdp);
159 if (!flex_bg || ext4_block_in_group(sb, tmp, block_group))
160 ext4_set_bit(tmp - start, bh->b_data);
161
162 tmp = ext4_inode_table(sb, gdp);
163 for (; tmp < ext4_inode_table(sb, gdp) +
164 sbi->s_itb_per_group; tmp++) {
165 if (!flex_bg ||
166 ext4_block_in_group(sb, tmp, block_group))
167 ext4_set_bit(tmp - start, bh->b_data);
168 }
717d50e4
AD
169 /*
170 * Also if the number of blocks within the group is
171 * less than the blocksize * 8 ( which is the size
172 * of bitmap ), set rest of the block bitmap to 1
173 */
174 mark_bitmap_end(group_blocks, sb->s_blocksize * 8, bh->b_data);
175 }
e187c658 176 return free_blocks - ext4_group_used_meta_blocks(sb, block_group, gdp);
717d50e4
AD
177}
178
179
ac27a0ec
DK
180/*
181 * The free blocks are managed by bitmaps. A file system contains several
182 * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap
183 * block for inodes, N blocks for the inode table and data blocks.
184 *
185 * The file system contains group descriptors which are located after the
186 * super block. Each descriptor contains the number of the bitmap block and
187 * the free blocks count in the block. The descriptors are loaded in memory
e627432c 188 * when a file system is mounted (see ext4_fill_super).
ac27a0ec
DK
189 */
190
ac27a0ec 191/**
617ba13b 192 * ext4_get_group_desc() -- load group descriptor from disk
ac27a0ec
DK
193 * @sb: super block
194 * @block_group: given block group
195 * @bh: pointer to the buffer head to store the block
196 * group descriptor
197 */
af5bc92d 198struct ext4_group_desc * ext4_get_group_desc(struct super_block *sb,
fd2d4291 199 ext4_group_t block_group,
af5bc92d 200 struct buffer_head **bh)
ac27a0ec 201{
498e5f24
TT
202 unsigned int group_desc;
203 unsigned int offset;
8df9675f 204 ext4_group_t ngroups = ext4_get_groups_count(sb);
af5bc92d 205 struct ext4_group_desc *desc;
617ba13b 206 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 207
8df9675f 208 if (block_group >= ngroups) {
12062ddd
ES
209 ext4_error(sb, "block_group >= groups_count - block_group = %u,"
210 " groups_count = %u", block_group, ngroups);
ac27a0ec
DK
211
212 return NULL;
213 }
ac27a0ec 214
617ba13b
MC
215 group_desc = block_group >> EXT4_DESC_PER_BLOCK_BITS(sb);
216 offset = block_group & (EXT4_DESC_PER_BLOCK(sb) - 1);
ac27a0ec 217 if (!sbi->s_group_desc[group_desc]) {
12062ddd 218 ext4_error(sb, "Group descriptor not loaded - "
498e5f24 219 "block_group = %u, group_desc = %u, desc = %u",
af5bc92d 220 block_group, group_desc, offset);
ac27a0ec
DK
221 return NULL;
222 }
223
0d1ee42f
AR
224 desc = (struct ext4_group_desc *)(
225 (__u8 *)sbi->s_group_desc[group_desc]->b_data +
226 offset * EXT4_DESC_SIZE(sb));
ac27a0ec
DK
227 if (bh)
228 *bh = sbi->s_group_desc[group_desc];
0d1ee42f 229 return desc;
ac27a0ec
DK
230}
231
abcb2947
AK
232static int ext4_valid_block_bitmap(struct super_block *sb,
233 struct ext4_group_desc *desc,
234 unsigned int block_group,
235 struct buffer_head *bh)
236{
237 ext4_grpblk_t offset;
238 ext4_grpblk_t next_zero_bit;
239 ext4_fsblk_t bitmap_blk;
240 ext4_fsblk_t group_first_block;
241
242 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) {
243 /* with FLEX_BG, the inode/block bitmaps and itable
244 * blocks may not be in the group at all
245 * so the bitmap validation will be skipped for those groups
246 * or it has to also read the block group where the bitmaps
247 * are located to verify they are set.
248 */
249 return 1;
250 }
251 group_first_block = ext4_group_first_block_no(sb, block_group);
252
253 /* check whether block bitmap block number is set */
254 bitmap_blk = ext4_block_bitmap(sb, desc);
255 offset = bitmap_blk - group_first_block;
256 if (!ext4_test_bit(offset, bh->b_data))
257 /* bad block bitmap */
258 goto err_out;
259
260 /* check whether the inode bitmap block number is set */
261 bitmap_blk = ext4_inode_bitmap(sb, desc);
262 offset = bitmap_blk - group_first_block;
263 if (!ext4_test_bit(offset, bh->b_data))
264 /* bad block bitmap */
265 goto err_out;
266
267 /* check whether the inode table block number is set */
268 bitmap_blk = ext4_inode_table(sb, desc);
269 offset = bitmap_blk - group_first_block;
270 next_zero_bit = ext4_find_next_zero_bit(bh->b_data,
271 offset + EXT4_SB(sb)->s_itb_per_group,
272 offset);
273 if (next_zero_bit >= offset + EXT4_SB(sb)->s_itb_per_group)
274 /* good bitmap for inode tables */
275 return 1;
276
277err_out:
12062ddd 278 ext4_error(sb, "Invalid block bitmap - block_group = %d, block = %llu",
abcb2947
AK
279 block_group, bitmap_blk);
280 return 0;
281}
ac27a0ec 282/**
574ca174 283 * ext4_read_block_bitmap()
ac27a0ec
DK
284 * @sb: super block
285 * @block_group: given block group
286 *
abcb2947
AK
287 * Read the bitmap for a given block_group,and validate the
288 * bits for block/inode/inode tables are set in the bitmaps
ac27a0ec
DK
289 *
290 * Return buffer_head on success or NULL in case of failure.
291 */
717d50e4 292struct buffer_head *
574ca174 293ext4_read_block_bitmap(struct super_block *sb, ext4_group_t block_group)
ac27a0ec 294{
af5bc92d
TT
295 struct ext4_group_desc *desc;
296 struct buffer_head *bh = NULL;
7c9e69fa 297 ext4_fsblk_t bitmap_blk;
ac27a0ec 298
717d50e4 299 desc = ext4_get_group_desc(sb, block_group, NULL);
ac27a0ec 300 if (!desc)
7c9e69fa
AK
301 return NULL;
302 bitmap_blk = ext4_block_bitmap(sb, desc);
abcb2947
AK
303 bh = sb_getblk(sb, bitmap_blk);
304 if (unlikely(!bh)) {
12062ddd 305 ext4_error(sb, "Cannot read block bitmap - "
a9df9a49 306 "block_group = %u, block_bitmap = %llu",
e29d1cde 307 block_group, bitmap_blk);
abcb2947
AK
308 return NULL;
309 }
2ccb5fb9
AK
310
311 if (bitmap_uptodate(bh))
abcb2947
AK
312 return bh;
313
c806e68f 314 lock_buffer(bh);
2ccb5fb9
AK
315 if (bitmap_uptodate(bh)) {
316 unlock_buffer(bh);
317 return bh;
318 }
955ce5f5 319 ext4_lock_group(sb, block_group);
717d50e4 320 if (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) {
abcb2947 321 ext4_init_block_bitmap(sb, bh, block_group, desc);
2ccb5fb9 322 set_bitmap_uptodate(bh);
abcb2947 323 set_buffer_uptodate(bh);
955ce5f5 324 ext4_unlock_group(sb, block_group);
3300beda 325 unlock_buffer(bh);
abcb2947 326 return bh;
717d50e4 327 }
955ce5f5 328 ext4_unlock_group(sb, block_group);
2ccb5fb9
AK
329 if (buffer_uptodate(bh)) {
330 /*
331 * if not uninit if bh is uptodate,
332 * bitmap is also uptodate
333 */
334 set_bitmap_uptodate(bh);
335 unlock_buffer(bh);
336 return bh;
337 }
338 /*
339 * submit the buffer_head for read. We can
340 * safely mark the bitmap as uptodate now.
341 * We do it here so the bitmap uptodate bit
342 * get set with buffer lock held.
343 */
344 set_bitmap_uptodate(bh);
abcb2947
AK
345 if (bh_submit_read(bh) < 0) {
346 put_bh(bh);
12062ddd 347 ext4_error(sb, "Cannot read block bitmap - "
a9df9a49 348 "block_group = %u, block_bitmap = %llu",
e29d1cde 349 block_group, bitmap_blk);
abcb2947
AK
350 return NULL;
351 }
519deca0
AK
352 ext4_valid_block_bitmap(sb, desc, block_group, bh);
353 /*
354 * file system mounted not to panic on error,
355 * continue with corrupt bitmap
356 */
ac27a0ec
DK
357 return bh;
358}
ac27a0ec
DK
359
360/**
e21675d4 361 * ext4_add_groupblocks() -- Add given blocks to an existing group
ac27a0ec
DK
362 * @handle: handle to this transaction
363 * @sb: super block
e21675d4 364 * @block: start physcial block to add to the block group
ac27a0ec 365 * @count: number of blocks to free
c2ea3fde 366 *
e21675d4
AK
367 * This marks the blocks as free in the bitmap. We ask the
368 * mballoc to reload the buddy after this by setting group
369 * EXT4_GROUP_INFO_NEED_INIT_BIT flag
ac27a0ec 370 */
e21675d4
AK
371void ext4_add_groupblocks(handle_t *handle, struct super_block *sb,
372 ext4_fsblk_t block, unsigned long count)
ac27a0ec
DK
373{
374 struct buffer_head *bitmap_bh = NULL;
375 struct buffer_head *gd_bh;
fd2d4291 376 ext4_group_t block_group;
617ba13b 377 ext4_grpblk_t bit;
498e5f24 378 unsigned int i;
af5bc92d 379 struct ext4_group_desc *desc;
a0375156 380 struct ext4_sb_info *sbi = EXT4_SB(sb);
560671a0 381 int err = 0, ret, blk_free_count;
e21675d4
AK
382 ext4_grpblk_t blocks_freed;
383 struct ext4_group_info *grp;
ac27a0ec 384
e21675d4 385 ext4_debug("Adding block(s) %llu-%llu\n", block, block + count - 1);
ac27a0ec 386
3a5b2ecd 387 ext4_get_group_no_and_offset(sb, block, &block_group, &bit);
920313a7 388 grp = ext4_get_group_info(sb, block_group);
ac27a0ec
DK
389 /*
390 * Check to see if we are freeing blocks across a group
391 * boundary.
392 */
617ba13b 393 if (bit + count > EXT4_BLOCKS_PER_GROUP(sb)) {
e21675d4 394 goto error_return;
ac27a0ec 395 }
574ca174 396 bitmap_bh = ext4_read_block_bitmap(sb, block_group);
ac27a0ec
DK
397 if (!bitmap_bh)
398 goto error_return;
af5bc92d 399 desc = ext4_get_group_desc(sb, block_group, &gd_bh);
ac27a0ec
DK
400 if (!desc)
401 goto error_return;
402
8fadc143
AR
403 if (in_range(ext4_block_bitmap(sb, desc), block, count) ||
404 in_range(ext4_inode_bitmap(sb, desc), block, count) ||
405 in_range(block, ext4_inode_table(sb, desc), sbi->s_itb_per_group) ||
406 in_range(block + count - 1, ext4_inode_table(sb, desc),
cb47dce7 407 sbi->s_itb_per_group)) {
12062ddd 408 ext4_error(sb, "Adding blocks in system zones - "
af5bc92d
TT
409 "Block = %llu, count = %lu",
410 block, count);
cb47dce7
AK
411 goto error_return;
412 }
ac27a0ec
DK
413
414 /*
e21675d4 415 * We are about to add blocks to the bitmap,
ac27a0ec
DK
416 * so we need undo access.
417 */
ac27a0ec 418 BUFFER_TRACE(bitmap_bh, "getting undo access");
617ba13b 419 err = ext4_journal_get_undo_access(handle, bitmap_bh);
ac27a0ec
DK
420 if (err)
421 goto error_return;
422
423 /*
424 * We are about to modify some metadata. Call the journal APIs
425 * to unshare ->b_data if a currently-committing transaction is
426 * using it
427 */
428 BUFFER_TRACE(gd_bh, "get_write_access");
617ba13b 429 err = ext4_journal_get_write_access(handle, gd_bh);
ac27a0ec
DK
430 if (err)
431 goto error_return;
920313a7
AK
432 /*
433 * make sure we don't allow a parallel init on other groups in the
434 * same buddy cache
435 */
436 down_write(&grp->alloc_sem);
e21675d4 437 for (i = 0, blocks_freed = 0; i < count; i++) {
ac27a0ec 438 BUFFER_TRACE(bitmap_bh, "clear bit");
955ce5f5 439 if (!ext4_clear_bit_atomic(ext4_group_lock_ptr(sb, block_group),
ac27a0ec 440 bit + i, bitmap_bh->b_data)) {
12062ddd 441 ext4_error(sb, "bit already cleared for block %llu",
bd81d8ee 442 (ext4_fsblk_t)(block + i));
ac27a0ec
DK
443 BUFFER_TRACE(bitmap_bh, "bit already cleared");
444 } else {
e21675d4 445 blocks_freed++;
ac27a0ec
DK
446 }
447 }
955ce5f5 448 ext4_lock_group(sb, block_group);
560671a0
AK
449 blk_free_count = blocks_freed + ext4_free_blks_count(sb, desc);
450 ext4_free_blks_set(sb, desc, blk_free_count);
717d50e4 451 desc->bg_checksum = ext4_group_desc_csum(sbi, block_group, desc);
955ce5f5 452 ext4_unlock_group(sb, block_group);
e21675d4 453 percpu_counter_add(&sbi->s_freeblocks_counter, blocks_freed);
ac27a0ec 454
772cb7c8
JS
455 if (sbi->s_log_groups_per_flex) {
456 ext4_group_t flex_group = ext4_flex_group(sbi, block_group);
9f24e420
TT
457 atomic_add(blocks_freed,
458 &sbi->s_flex_groups[flex_group].free_blocks);
772cb7c8 459 }
920313a7
AK
460 /*
461 * request to reload the buddy with the
462 * new bitmap information
463 */
464 set_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &(grp->bb_state));
0373130d 465 grp->bb_free += blocks_freed;
920313a7 466 up_write(&grp->alloc_sem);
772cb7c8 467
ac27a0ec
DK
468 /* We dirtied the bitmap block */
469 BUFFER_TRACE(bitmap_bh, "dirtied bitmap block");
0390131b 470 err = ext4_handle_dirty_metadata(handle, NULL, bitmap_bh);
ac27a0ec
DK
471
472 /* And the group descriptor block */
473 BUFFER_TRACE(gd_bh, "dirtied group descriptor block");
0390131b 474 ret = ext4_handle_dirty_metadata(handle, NULL, gd_bh);
e21675d4
AK
475 if (!err)
476 err = ret;
e21675d4 477
ac27a0ec
DK
478error_return:
479 brelse(bitmap_bh);
617ba13b 480 ext4_std_error(sb, err);
ac27a0ec
DK
481 return;
482}
483
8c3bf8a0
ES
484/**
485 * ext4_has_free_blocks()
486 * @sbi: in-core super block structure.
487 * @nblocks: number of needed blocks
488 *
489 * Check if filesystem has nblocks free & available for allocation.
490 * On success return 1, return 0 on failure.
491 */
1f109d5a 492static int ext4_has_free_blocks(struct ext4_sb_info *sbi, s64 nblocks)
a30d542a 493{
a996031c 494 s64 free_blocks, dirty_blocks, root_blocks;
a30d542a 495 struct percpu_counter *fbc = &sbi->s_freeblocks_counter;
6bc6e63f 496 struct percpu_counter *dbc = &sbi->s_dirtyblocks_counter;
a30d542a 497
6bc6e63f
AK
498 free_blocks = percpu_counter_read_positive(fbc);
499 dirty_blocks = percpu_counter_read_positive(dbc);
a996031c 500 root_blocks = ext4_r_blocks_count(sbi->s_es);
a30d542a 501
6bc6e63f
AK
502 if (free_blocks - (nblocks + root_blocks + dirty_blocks) <
503 EXT4_FREEBLOCKS_WATERMARK) {
02d21168
AM
504 free_blocks = percpu_counter_sum_positive(fbc);
505 dirty_blocks = percpu_counter_sum_positive(dbc);
6bc6e63f
AK
506 if (dirty_blocks < 0) {
507 printk(KERN_CRIT "Dirty block accounting "
508 "went wrong %lld\n",
8f72fbdf 509 (long long)dirty_blocks);
6bc6e63f
AK
510 }
511 }
512 /* Check whether we have space after
a996031c 513 * accounting for current dirty blocks & root reserved blocks.
6bc6e63f 514 */
a996031c
ES
515 if (free_blocks >= ((root_blocks + nblocks) + dirty_blocks))
516 return 1;
a30d542a 517
a996031c 518 /* Hm, nope. Are (enough) root reserved blocks available? */
4c9c544e 519 if (sbi->s_resuid == current_fsuid() ||
a996031c
ES
520 ((sbi->s_resgid != 0) && in_group_p(sbi->s_resgid)) ||
521 capable(CAP_SYS_RESOURCE)) {
522 if (free_blocks >= (nblocks + dirty_blocks))
523 return 1;
524 }
525
526 return 0;
a30d542a
AK
527}
528
8c3bf8a0 529int ext4_claim_free_blocks(struct ext4_sb_info *sbi,
5c791616 530 s64 nblocks)
ac27a0ec 531{
8c3bf8a0
ES
532 if (ext4_has_free_blocks(sbi, nblocks)) {
533 percpu_counter_add(&sbi->s_dirtyblocks_counter, nblocks);
16eb7295 534 return 0;
8c3bf8a0
ES
535 } else
536 return -ENOSPC;
a30d542a 537}
07031431 538
ac27a0ec 539/**
617ba13b 540 * ext4_should_retry_alloc()
ac27a0ec
DK
541 * @sb: super block
542 * @retries number of attemps has been made
543 *
617ba13b 544 * ext4_should_retry_alloc() is called when ENOSPC is returned, and if
ac27a0ec
DK
545 * it is profitable to retry the operation, this function will wait
546 * for the current or commiting transaction to complete, and then
547 * return TRUE.
548 *
549 * if the total number of retries exceed three times, return FALSE.
550 */
617ba13b 551int ext4_should_retry_alloc(struct super_block *sb, int *retries)
ac27a0ec 552{
8f64b32e
ES
553 if (!ext4_has_free_blocks(EXT4_SB(sb), 1) ||
554 (*retries)++ > 3 ||
555 !EXT4_SB(sb)->s_journal)
ac27a0ec
DK
556 return 0;
557
558 jbd_debug(1, "%s: retrying operation after ENOSPC\n", sb->s_id);
559
dab291af 560 return jbd2_journal_force_commit_nested(EXT4_SB(sb)->s_journal);
ac27a0ec
DK
561}
562
654b4908 563/*
d2a17637 564 * ext4_new_meta_blocks() -- allocate block for meta data (indexing) blocks
654b4908
AK
565 *
566 * @handle: handle to this transaction
567 * @inode: file inode
568 * @goal: given target block(filesystem wide)
97df5d15 569 * @count: pointer to total number of blocks needed
654b4908
AK
570 * @errp: error code
571 *
97df5d15 572 * Return 1st allocated block number on success, *count stores total account
d2a17637 573 * error stores in errp pointer
654b4908 574 */
d2a17637
MC
575ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
576 ext4_fsblk_t goal, unsigned long *count, int *errp)
654b4908 577{
97df5d15 578 struct ext4_allocation_request ar;
d2a17637 579 ext4_fsblk_t ret;
97df5d15
TT
580
581 memset(&ar, 0, sizeof(ar));
582 /* Fill with neighbour allocated blocks */
583 ar.inode = inode;
584 ar.goal = goal;
585 ar.len = count ? *count : 1;
586
587 ret = ext4_mb_new_blocks(handle, &ar, errp);
588 if (count)
589 *count = ar.len;
d2a17637 590 /*
72b8ab9d
ES
591 * Account for the allocated meta blocks. We will never
592 * fail EDQUOT for metdata, but we do account for it.
d2a17637 593 */
166348dd 594 if (!(*errp) && EXT4_I(inode)->i_delalloc_reserved_flag) {
d2a17637 595 spin_lock(&EXT4_I(inode)->i_block_reservation_lock);
97df5d15 596 EXT4_I(inode)->i_allocated_meta_blocks += ar.len;
d2a17637 597 spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
72b8ab9d 598 dquot_alloc_block_nofail(inode, ar.len);
d2a17637
MC
599 }
600 return ret;
654b4908
AK
601}
602
ac27a0ec 603/**
617ba13b 604 * ext4_count_free_blocks() -- count filesystem free blocks
ac27a0ec
DK
605 * @sb: superblock
606 *
607 * Adds up the number of free blocks from each block group.
608 */
617ba13b 609ext4_fsblk_t ext4_count_free_blocks(struct super_block *sb)
ac27a0ec 610{
617ba13b
MC
611 ext4_fsblk_t desc_count;
612 struct ext4_group_desc *gdp;
fd2d4291 613 ext4_group_t i;
8df9675f 614 ext4_group_t ngroups = ext4_get_groups_count(sb);
617ba13b
MC
615#ifdef EXT4FS_DEBUG
616 struct ext4_super_block *es;
617 ext4_fsblk_t bitmap_count;
498e5f24 618 unsigned int x;
ac27a0ec
DK
619 struct buffer_head *bitmap_bh = NULL;
620
617ba13b 621 es = EXT4_SB(sb)->s_es;
ac27a0ec
DK
622 desc_count = 0;
623 bitmap_count = 0;
624 gdp = NULL;
625
ac27a0ec 626 for (i = 0; i < ngroups; i++) {
617ba13b 627 gdp = ext4_get_group_desc(sb, i, NULL);
ac27a0ec
DK
628 if (!gdp)
629 continue;
9fd9784c 630 desc_count += ext4_free_blks_count(sb, gdp);
ac27a0ec 631 brelse(bitmap_bh);
574ca174 632 bitmap_bh = ext4_read_block_bitmap(sb, i);
ac27a0ec
DK
633 if (bitmap_bh == NULL)
634 continue;
635
617ba13b 636 x = ext4_count_free(bitmap_bh, sb->s_blocksize);
9fd9784c
TLSC
637 printk(KERN_DEBUG "group %u: stored = %d, counted = %u\n",
638 i, ext4_free_blks_count(sb, gdp), x);
ac27a0ec
DK
639 bitmap_count += x;
640 }
641 brelse(bitmap_bh);
4776004f
TT
642 printk(KERN_DEBUG "ext4_count_free_blocks: stored = %llu"
643 ", computed = %llu, %llu\n", ext4_free_blocks_count(es),
644 desc_count, bitmap_count);
ac27a0ec
DK
645 return bitmap_count;
646#else
647 desc_count = 0;
ac27a0ec 648 for (i = 0; i < ngroups; i++) {
617ba13b 649 gdp = ext4_get_group_desc(sb, i, NULL);
ac27a0ec
DK
650 if (!gdp)
651 continue;
560671a0 652 desc_count += ext4_free_blks_count(sb, gdp);
ac27a0ec
DK
653 }
654
655 return desc_count;
656#endif
657}
658
fd2d4291 659static inline int test_root(ext4_group_t a, int b)
ac27a0ec
DK
660{
661 int num = b;
662
663 while (a > num)
664 num *= b;
665 return num == a;
666}
667
fd2d4291 668static int ext4_group_sparse(ext4_group_t group)
ac27a0ec
DK
669{
670 if (group <= 1)
671 return 1;
672 if (!(group & 1))
673 return 0;
674 return (test_root(group, 7) || test_root(group, 5) ||
675 test_root(group, 3));
676}
677
678/**
617ba13b 679 * ext4_bg_has_super - number of blocks used by the superblock in group
ac27a0ec
DK
680 * @sb: superblock for filesystem
681 * @group: group number to check
682 *
683 * Return the number of blocks used by the superblock (primary or backup)
684 * in this group. Currently this will be only 0 or 1.
685 */
fd2d4291 686int ext4_bg_has_super(struct super_block *sb, ext4_group_t group)
ac27a0ec 687{
617ba13b
MC
688 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
689 EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER) &&
690 !ext4_group_sparse(group))
ac27a0ec
DK
691 return 0;
692 return 1;
693}
694
fd2d4291
AM
695static unsigned long ext4_bg_num_gdb_meta(struct super_block *sb,
696 ext4_group_t group)
ac27a0ec 697{
617ba13b 698 unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
fd2d4291
AM
699 ext4_group_t first = metagroup * EXT4_DESC_PER_BLOCK(sb);
700 ext4_group_t last = first + EXT4_DESC_PER_BLOCK(sb) - 1;
ac27a0ec
DK
701
702 if (group == first || group == first + 1 || group == last)
703 return 1;
704 return 0;
705}
706
fd2d4291
AM
707static unsigned long ext4_bg_num_gdb_nometa(struct super_block *sb,
708 ext4_group_t group)
ac27a0ec 709{
8dadb198
TT
710 if (!ext4_bg_has_super(sb, group))
711 return 0;
712
713 if (EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG))
714 return le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg);
715 else
716 return EXT4_SB(sb)->s_gdb_count;
ac27a0ec
DK
717}
718
719/**
617ba13b 720 * ext4_bg_num_gdb - number of blocks used by the group table in group
ac27a0ec
DK
721 * @sb: superblock for filesystem
722 * @group: group number to check
723 *
724 * Return the number of blocks used by the group descriptor table
725 * (primary or backup) in this group. In the future there may be a
726 * different number of descriptor blocks in each group.
727 */
fd2d4291 728unsigned long ext4_bg_num_gdb(struct super_block *sb, ext4_group_t group)
ac27a0ec
DK
729{
730 unsigned long first_meta_bg =
617ba13b
MC
731 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg);
732 unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
ac27a0ec 733
617ba13b 734 if (!EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 735 metagroup < first_meta_bg)
af5bc92d 736 return ext4_bg_num_gdb_nometa(sb, group);
ac27a0ec 737
617ba13b 738 return ext4_bg_num_gdb_meta(sb,group);
ac27a0ec
DK
739
740}
c2ea3fde 741