]> bbs.cooldavid.org Git - net-next-2.6.git/blame - fs/ext4/balloc.c
xps: Transmit Packet Steering
[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 */
61d08673
TT
174 ext4_mark_bitmap_end(group_blocks, sb->s_blocksize * 8,
175 bh->b_data);
717d50e4 176 }
e187c658 177 return free_blocks - ext4_group_used_meta_blocks(sb, block_group, gdp);
717d50e4
AD
178}
179
180
ac27a0ec
DK
181/*
182 * The free blocks are managed by bitmaps. A file system contains several
183 * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap
184 * block for inodes, N blocks for the inode table and data blocks.
185 *
186 * The file system contains group descriptors which are located after the
187 * super block. Each descriptor contains the number of the bitmap block and
188 * the free blocks count in the block. The descriptors are loaded in memory
e627432c 189 * when a file system is mounted (see ext4_fill_super).
ac27a0ec
DK
190 */
191
ac27a0ec 192/**
617ba13b 193 * ext4_get_group_desc() -- load group descriptor from disk
ac27a0ec
DK
194 * @sb: super block
195 * @block_group: given block group
196 * @bh: pointer to the buffer head to store the block
197 * group descriptor
198 */
af5bc92d 199struct ext4_group_desc * ext4_get_group_desc(struct super_block *sb,
fd2d4291 200 ext4_group_t block_group,
af5bc92d 201 struct buffer_head **bh)
ac27a0ec 202{
498e5f24
TT
203 unsigned int group_desc;
204 unsigned int offset;
8df9675f 205 ext4_group_t ngroups = ext4_get_groups_count(sb);
af5bc92d 206 struct ext4_group_desc *desc;
617ba13b 207 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 208
8df9675f 209 if (block_group >= ngroups) {
12062ddd
ES
210 ext4_error(sb, "block_group >= groups_count - block_group = %u,"
211 " groups_count = %u", block_group, ngroups);
ac27a0ec
DK
212
213 return NULL;
214 }
ac27a0ec 215
617ba13b
MC
216 group_desc = block_group >> EXT4_DESC_PER_BLOCK_BITS(sb);
217 offset = block_group & (EXT4_DESC_PER_BLOCK(sb) - 1);
ac27a0ec 218 if (!sbi->s_group_desc[group_desc]) {
12062ddd 219 ext4_error(sb, "Group descriptor not loaded - "
498e5f24 220 "block_group = %u, group_desc = %u, desc = %u",
af5bc92d 221 block_group, group_desc, offset);
ac27a0ec
DK
222 return NULL;
223 }
224
0d1ee42f
AR
225 desc = (struct ext4_group_desc *)(
226 (__u8 *)sbi->s_group_desc[group_desc]->b_data +
227 offset * EXT4_DESC_SIZE(sb));
ac27a0ec
DK
228 if (bh)
229 *bh = sbi->s_group_desc[group_desc];
0d1ee42f 230 return desc;
ac27a0ec
DK
231}
232
abcb2947
AK
233static int ext4_valid_block_bitmap(struct super_block *sb,
234 struct ext4_group_desc *desc,
235 unsigned int block_group,
236 struct buffer_head *bh)
237{
238 ext4_grpblk_t offset;
239 ext4_grpblk_t next_zero_bit;
240 ext4_fsblk_t bitmap_blk;
241 ext4_fsblk_t group_first_block;
242
243 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) {
244 /* with FLEX_BG, the inode/block bitmaps and itable
245 * blocks may not be in the group at all
246 * so the bitmap validation will be skipped for those groups
247 * or it has to also read the block group where the bitmaps
248 * are located to verify they are set.
249 */
250 return 1;
251 }
252 group_first_block = ext4_group_first_block_no(sb, block_group);
253
254 /* check whether block bitmap block number is set */
255 bitmap_blk = ext4_block_bitmap(sb, desc);
256 offset = bitmap_blk - group_first_block;
257 if (!ext4_test_bit(offset, bh->b_data))
258 /* bad block bitmap */
259 goto err_out;
260
261 /* check whether the inode bitmap block number is set */
262 bitmap_blk = ext4_inode_bitmap(sb, desc);
263 offset = bitmap_blk - group_first_block;
264 if (!ext4_test_bit(offset, bh->b_data))
265 /* bad block bitmap */
266 goto err_out;
267
268 /* check whether the inode table block number is set */
269 bitmap_blk = ext4_inode_table(sb, desc);
270 offset = bitmap_blk - group_first_block;
271 next_zero_bit = ext4_find_next_zero_bit(bh->b_data,
272 offset + EXT4_SB(sb)->s_itb_per_group,
273 offset);
274 if (next_zero_bit >= offset + EXT4_SB(sb)->s_itb_per_group)
275 /* good bitmap for inode tables */
276 return 1;
277
278err_out:
12062ddd 279 ext4_error(sb, "Invalid block bitmap - block_group = %d, block = %llu",
abcb2947
AK
280 block_group, bitmap_blk);
281 return 0;
282}
ac27a0ec 283/**
574ca174 284 * ext4_read_block_bitmap()
ac27a0ec
DK
285 * @sb: super block
286 * @block_group: given block group
287 *
abcb2947
AK
288 * Read the bitmap for a given block_group,and validate the
289 * bits for block/inode/inode tables are set in the bitmaps
ac27a0ec
DK
290 *
291 * Return buffer_head on success or NULL in case of failure.
292 */
717d50e4 293struct buffer_head *
574ca174 294ext4_read_block_bitmap(struct super_block *sb, ext4_group_t block_group)
ac27a0ec 295{
af5bc92d
TT
296 struct ext4_group_desc *desc;
297 struct buffer_head *bh = NULL;
7c9e69fa 298 ext4_fsblk_t bitmap_blk;
ac27a0ec 299
717d50e4 300 desc = ext4_get_group_desc(sb, block_group, NULL);
ac27a0ec 301 if (!desc)
7c9e69fa
AK
302 return NULL;
303 bitmap_blk = ext4_block_bitmap(sb, desc);
abcb2947
AK
304 bh = sb_getblk(sb, bitmap_blk);
305 if (unlikely(!bh)) {
12062ddd 306 ext4_error(sb, "Cannot read block bitmap - "
a9df9a49 307 "block_group = %u, block_bitmap = %llu",
e29d1cde 308 block_group, bitmap_blk);
abcb2947
AK
309 return NULL;
310 }
2ccb5fb9
AK
311
312 if (bitmap_uptodate(bh))
abcb2947
AK
313 return bh;
314
c806e68f 315 lock_buffer(bh);
2ccb5fb9
AK
316 if (bitmap_uptodate(bh)) {
317 unlock_buffer(bh);
318 return bh;
319 }
955ce5f5 320 ext4_lock_group(sb, block_group);
717d50e4 321 if (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) {
abcb2947 322 ext4_init_block_bitmap(sb, bh, block_group, desc);
2ccb5fb9 323 set_bitmap_uptodate(bh);
abcb2947 324 set_buffer_uptodate(bh);
955ce5f5 325 ext4_unlock_group(sb, block_group);
3300beda 326 unlock_buffer(bh);
abcb2947 327 return bh;
717d50e4 328 }
955ce5f5 329 ext4_unlock_group(sb, block_group);
2ccb5fb9
AK
330 if (buffer_uptodate(bh)) {
331 /*
332 * if not uninit if bh is uptodate,
333 * bitmap is also uptodate
334 */
335 set_bitmap_uptodate(bh);
336 unlock_buffer(bh);
337 return bh;
338 }
339 /*
340 * submit the buffer_head for read. We can
341 * safely mark the bitmap as uptodate now.
342 * We do it here so the bitmap uptodate bit
343 * get set with buffer lock held.
344 */
345 set_bitmap_uptodate(bh);
abcb2947
AK
346 if (bh_submit_read(bh) < 0) {
347 put_bh(bh);
12062ddd 348 ext4_error(sb, "Cannot read block bitmap - "
a9df9a49 349 "block_group = %u, block_bitmap = %llu",
e29d1cde 350 block_group, bitmap_blk);
abcb2947
AK
351 return NULL;
352 }
519deca0
AK
353 ext4_valid_block_bitmap(sb, desc, block_group, bh);
354 /*
355 * file system mounted not to panic on error,
356 * continue with corrupt bitmap
357 */
ac27a0ec
DK
358 return bh;
359}
ac27a0ec
DK
360
361/**
e21675d4 362 * ext4_add_groupblocks() -- Add given blocks to an existing group
ac27a0ec
DK
363 * @handle: handle to this transaction
364 * @sb: super block
e21675d4 365 * @block: start physcial block to add to the block group
ac27a0ec 366 * @count: number of blocks to free
c2ea3fde 367 *
e21675d4
AK
368 * This marks the blocks as free in the bitmap. We ask the
369 * mballoc to reload the buddy after this by setting group
370 * EXT4_GROUP_INFO_NEED_INIT_BIT flag
ac27a0ec 371 */
e21675d4
AK
372void ext4_add_groupblocks(handle_t *handle, struct super_block *sb,
373 ext4_fsblk_t block, unsigned long count)
ac27a0ec
DK
374{
375 struct buffer_head *bitmap_bh = NULL;
376 struct buffer_head *gd_bh;
fd2d4291 377 ext4_group_t block_group;
617ba13b 378 ext4_grpblk_t bit;
498e5f24 379 unsigned int i;
af5bc92d 380 struct ext4_group_desc *desc;
a0375156 381 struct ext4_sb_info *sbi = EXT4_SB(sb);
560671a0 382 int err = 0, ret, blk_free_count;
e21675d4
AK
383 ext4_grpblk_t blocks_freed;
384 struct ext4_group_info *grp;
ac27a0ec 385
e21675d4 386 ext4_debug("Adding block(s) %llu-%llu\n", block, block + count - 1);
ac27a0ec 387
3a5b2ecd 388 ext4_get_group_no_and_offset(sb, block, &block_group, &bit);
920313a7 389 grp = ext4_get_group_info(sb, block_group);
ac27a0ec
DK
390 /*
391 * Check to see if we are freeing blocks across a group
392 * boundary.
393 */
617ba13b 394 if (bit + count > EXT4_BLOCKS_PER_GROUP(sb)) {
e21675d4 395 goto error_return;
ac27a0ec 396 }
574ca174 397 bitmap_bh = ext4_read_block_bitmap(sb, block_group);
ac27a0ec
DK
398 if (!bitmap_bh)
399 goto error_return;
af5bc92d 400 desc = ext4_get_group_desc(sb, block_group, &gd_bh);
ac27a0ec
DK
401 if (!desc)
402 goto error_return;
403
8fadc143
AR
404 if (in_range(ext4_block_bitmap(sb, desc), block, count) ||
405 in_range(ext4_inode_bitmap(sb, desc), block, count) ||
406 in_range(block, ext4_inode_table(sb, desc), sbi->s_itb_per_group) ||
407 in_range(block + count - 1, ext4_inode_table(sb, desc),
cb47dce7 408 sbi->s_itb_per_group)) {
12062ddd 409 ext4_error(sb, "Adding blocks in system zones - "
af5bc92d
TT
410 "Block = %llu, count = %lu",
411 block, count);
cb47dce7
AK
412 goto error_return;
413 }
ac27a0ec
DK
414
415 /*
e21675d4 416 * We are about to add blocks to the bitmap,
ac27a0ec
DK
417 * so we need undo access.
418 */
ac27a0ec 419 BUFFER_TRACE(bitmap_bh, "getting undo access");
617ba13b 420 err = ext4_journal_get_undo_access(handle, bitmap_bh);
ac27a0ec
DK
421 if (err)
422 goto error_return;
423
424 /*
425 * We are about to modify some metadata. Call the journal APIs
426 * to unshare ->b_data if a currently-committing transaction is
427 * using it
428 */
429 BUFFER_TRACE(gd_bh, "get_write_access");
617ba13b 430 err = ext4_journal_get_write_access(handle, gd_bh);
ac27a0ec
DK
431 if (err)
432 goto error_return;
920313a7
AK
433 /*
434 * make sure we don't allow a parallel init on other groups in the
435 * same buddy cache
436 */
437 down_write(&grp->alloc_sem);
e21675d4 438 for (i = 0, blocks_freed = 0; i < count; i++) {
ac27a0ec 439 BUFFER_TRACE(bitmap_bh, "clear bit");
955ce5f5 440 if (!ext4_clear_bit_atomic(ext4_group_lock_ptr(sb, block_group),
ac27a0ec 441 bit + i, bitmap_bh->b_data)) {
12062ddd 442 ext4_error(sb, "bit already cleared for block %llu",
bd81d8ee 443 (ext4_fsblk_t)(block + i));
ac27a0ec
DK
444 BUFFER_TRACE(bitmap_bh, "bit already cleared");
445 } else {
e21675d4 446 blocks_freed++;
ac27a0ec
DK
447 }
448 }
955ce5f5 449 ext4_lock_group(sb, block_group);
560671a0
AK
450 blk_free_count = blocks_freed + ext4_free_blks_count(sb, desc);
451 ext4_free_blks_set(sb, desc, blk_free_count);
717d50e4 452 desc->bg_checksum = ext4_group_desc_csum(sbi, block_group, desc);
955ce5f5 453 ext4_unlock_group(sb, block_group);
e21675d4 454 percpu_counter_add(&sbi->s_freeblocks_counter, blocks_freed);
ac27a0ec 455
772cb7c8
JS
456 if (sbi->s_log_groups_per_flex) {
457 ext4_group_t flex_group = ext4_flex_group(sbi, block_group);
9f24e420
TT
458 atomic_add(blocks_freed,
459 &sbi->s_flex_groups[flex_group].free_blocks);
772cb7c8 460 }
920313a7
AK
461 /*
462 * request to reload the buddy with the
463 * new bitmap information
464 */
465 set_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &(grp->bb_state));
0373130d 466 grp->bb_free += blocks_freed;
920313a7 467 up_write(&grp->alloc_sem);
772cb7c8 468
ac27a0ec
DK
469 /* We dirtied the bitmap block */
470 BUFFER_TRACE(bitmap_bh, "dirtied bitmap block");
0390131b 471 err = ext4_handle_dirty_metadata(handle, NULL, bitmap_bh);
ac27a0ec
DK
472
473 /* And the group descriptor block */
474 BUFFER_TRACE(gd_bh, "dirtied group descriptor block");
0390131b 475 ret = ext4_handle_dirty_metadata(handle, NULL, gd_bh);
e21675d4
AK
476 if (!err)
477 err = ret;
e21675d4 478
ac27a0ec
DK
479error_return:
480 brelse(bitmap_bh);
617ba13b 481 ext4_std_error(sb, err);
ac27a0ec
DK
482 return;
483}
484
8c3bf8a0
ES
485/**
486 * ext4_has_free_blocks()
487 * @sbi: in-core super block structure.
488 * @nblocks: number of needed blocks
489 *
490 * Check if filesystem has nblocks free & available for allocation.
491 * On success return 1, return 0 on failure.
492 */
1f109d5a 493static int ext4_has_free_blocks(struct ext4_sb_info *sbi, s64 nblocks)
a30d542a 494{
a996031c 495 s64 free_blocks, dirty_blocks, root_blocks;
a30d542a 496 struct percpu_counter *fbc = &sbi->s_freeblocks_counter;
6bc6e63f 497 struct percpu_counter *dbc = &sbi->s_dirtyblocks_counter;
a30d542a 498
6bc6e63f
AK
499 free_blocks = percpu_counter_read_positive(fbc);
500 dirty_blocks = percpu_counter_read_positive(dbc);
a996031c 501 root_blocks = ext4_r_blocks_count(sbi->s_es);
a30d542a 502
6bc6e63f
AK
503 if (free_blocks - (nblocks + root_blocks + dirty_blocks) <
504 EXT4_FREEBLOCKS_WATERMARK) {
02d21168
AM
505 free_blocks = percpu_counter_sum_positive(fbc);
506 dirty_blocks = percpu_counter_sum_positive(dbc);
6bc6e63f
AK
507 if (dirty_blocks < 0) {
508 printk(KERN_CRIT "Dirty block accounting "
509 "went wrong %lld\n",
8f72fbdf 510 (long long)dirty_blocks);
6bc6e63f
AK
511 }
512 }
513 /* Check whether we have space after
a996031c 514 * accounting for current dirty blocks & root reserved blocks.
6bc6e63f 515 */
a996031c
ES
516 if (free_blocks >= ((root_blocks + nblocks) + dirty_blocks))
517 return 1;
a30d542a 518
a996031c 519 /* Hm, nope. Are (enough) root reserved blocks available? */
4c9c544e 520 if (sbi->s_resuid == current_fsuid() ||
a996031c
ES
521 ((sbi->s_resgid != 0) && in_group_p(sbi->s_resgid)) ||
522 capable(CAP_SYS_RESOURCE)) {
523 if (free_blocks >= (nblocks + dirty_blocks))
524 return 1;
525 }
526
527 return 0;
a30d542a
AK
528}
529
8c3bf8a0 530int ext4_claim_free_blocks(struct ext4_sb_info *sbi,
5c791616 531 s64 nblocks)
ac27a0ec 532{
8c3bf8a0
ES
533 if (ext4_has_free_blocks(sbi, nblocks)) {
534 percpu_counter_add(&sbi->s_dirtyblocks_counter, nblocks);
16eb7295 535 return 0;
8c3bf8a0
ES
536 } else
537 return -ENOSPC;
a30d542a 538}
07031431 539
ac27a0ec 540/**
617ba13b 541 * ext4_should_retry_alloc()
ac27a0ec
DK
542 * @sb: super block
543 * @retries number of attemps has been made
544 *
617ba13b 545 * ext4_should_retry_alloc() is called when ENOSPC is returned, and if
ac27a0ec
DK
546 * it is profitable to retry the operation, this function will wait
547 * for the current or commiting transaction to complete, and then
548 * return TRUE.
549 *
550 * if the total number of retries exceed three times, return FALSE.
551 */
617ba13b 552int ext4_should_retry_alloc(struct super_block *sb, int *retries)
ac27a0ec 553{
8f64b32e
ES
554 if (!ext4_has_free_blocks(EXT4_SB(sb), 1) ||
555 (*retries)++ > 3 ||
556 !EXT4_SB(sb)->s_journal)
ac27a0ec
DK
557 return 0;
558
559 jbd_debug(1, "%s: retrying operation after ENOSPC\n", sb->s_id);
560
dab291af 561 return jbd2_journal_force_commit_nested(EXT4_SB(sb)->s_journal);
ac27a0ec
DK
562}
563
654b4908 564/*
d2a17637 565 * ext4_new_meta_blocks() -- allocate block for meta data (indexing) blocks
654b4908
AK
566 *
567 * @handle: handle to this transaction
568 * @inode: file inode
569 * @goal: given target block(filesystem wide)
97df5d15 570 * @count: pointer to total number of blocks needed
654b4908
AK
571 * @errp: error code
572 *
97df5d15 573 * Return 1st allocated block number on success, *count stores total account
d2a17637 574 * error stores in errp pointer
654b4908 575 */
d2a17637
MC
576ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
577 ext4_fsblk_t goal, unsigned long *count, int *errp)
654b4908 578{
97df5d15 579 struct ext4_allocation_request ar;
d2a17637 580 ext4_fsblk_t ret;
97df5d15
TT
581
582 memset(&ar, 0, sizeof(ar));
583 /* Fill with neighbour allocated blocks */
584 ar.inode = inode;
585 ar.goal = goal;
586 ar.len = count ? *count : 1;
587
588 ret = ext4_mb_new_blocks(handle, &ar, errp);
589 if (count)
590 *count = ar.len;
d2a17637 591 /*
72b8ab9d
ES
592 * Account for the allocated meta blocks. We will never
593 * fail EDQUOT for metdata, but we do account for it.
d2a17637 594 */
166348dd 595 if (!(*errp) && EXT4_I(inode)->i_delalloc_reserved_flag) {
d2a17637 596 spin_lock(&EXT4_I(inode)->i_block_reservation_lock);
97df5d15 597 EXT4_I(inode)->i_allocated_meta_blocks += ar.len;
d2a17637 598 spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
72b8ab9d 599 dquot_alloc_block_nofail(inode, ar.len);
d2a17637
MC
600 }
601 return ret;
654b4908
AK
602}
603
ac27a0ec 604/**
617ba13b 605 * ext4_count_free_blocks() -- count filesystem free blocks
ac27a0ec
DK
606 * @sb: superblock
607 *
608 * Adds up the number of free blocks from each block group.
609 */
617ba13b 610ext4_fsblk_t ext4_count_free_blocks(struct super_block *sb)
ac27a0ec 611{
617ba13b
MC
612 ext4_fsblk_t desc_count;
613 struct ext4_group_desc *gdp;
fd2d4291 614 ext4_group_t i;
8df9675f 615 ext4_group_t ngroups = ext4_get_groups_count(sb);
617ba13b
MC
616#ifdef EXT4FS_DEBUG
617 struct ext4_super_block *es;
618 ext4_fsblk_t bitmap_count;
498e5f24 619 unsigned int x;
ac27a0ec
DK
620 struct buffer_head *bitmap_bh = NULL;
621
617ba13b 622 es = EXT4_SB(sb)->s_es;
ac27a0ec
DK
623 desc_count = 0;
624 bitmap_count = 0;
625 gdp = NULL;
626
ac27a0ec 627 for (i = 0; i < ngroups; i++) {
617ba13b 628 gdp = ext4_get_group_desc(sb, i, NULL);
ac27a0ec
DK
629 if (!gdp)
630 continue;
9fd9784c 631 desc_count += ext4_free_blks_count(sb, gdp);
ac27a0ec 632 brelse(bitmap_bh);
574ca174 633 bitmap_bh = ext4_read_block_bitmap(sb, i);
ac27a0ec
DK
634 if (bitmap_bh == NULL)
635 continue;
636
617ba13b 637 x = ext4_count_free(bitmap_bh, sb->s_blocksize);
9fd9784c
TLSC
638 printk(KERN_DEBUG "group %u: stored = %d, counted = %u\n",
639 i, ext4_free_blks_count(sb, gdp), x);
ac27a0ec
DK
640 bitmap_count += x;
641 }
642 brelse(bitmap_bh);
4776004f
TT
643 printk(KERN_DEBUG "ext4_count_free_blocks: stored = %llu"
644 ", computed = %llu, %llu\n", ext4_free_blocks_count(es),
645 desc_count, bitmap_count);
ac27a0ec
DK
646 return bitmap_count;
647#else
648 desc_count = 0;
ac27a0ec 649 for (i = 0; i < ngroups; i++) {
617ba13b 650 gdp = ext4_get_group_desc(sb, i, NULL);
ac27a0ec
DK
651 if (!gdp)
652 continue;
560671a0 653 desc_count += ext4_free_blks_count(sb, gdp);
ac27a0ec
DK
654 }
655
656 return desc_count;
657#endif
658}
659
fd2d4291 660static inline int test_root(ext4_group_t a, int b)
ac27a0ec
DK
661{
662 int num = b;
663
664 while (a > num)
665 num *= b;
666 return num == a;
667}
668
fd2d4291 669static int ext4_group_sparse(ext4_group_t group)
ac27a0ec
DK
670{
671 if (group <= 1)
672 return 1;
673 if (!(group & 1))
674 return 0;
675 return (test_root(group, 7) || test_root(group, 5) ||
676 test_root(group, 3));
677}
678
679/**
617ba13b 680 * ext4_bg_has_super - number of blocks used by the superblock in group
ac27a0ec
DK
681 * @sb: superblock for filesystem
682 * @group: group number to check
683 *
684 * Return the number of blocks used by the superblock (primary or backup)
685 * in this group. Currently this will be only 0 or 1.
686 */
fd2d4291 687int ext4_bg_has_super(struct super_block *sb, ext4_group_t group)
ac27a0ec 688{
617ba13b
MC
689 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
690 EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER) &&
691 !ext4_group_sparse(group))
ac27a0ec
DK
692 return 0;
693 return 1;
694}
695
fd2d4291
AM
696static unsigned long ext4_bg_num_gdb_meta(struct super_block *sb,
697 ext4_group_t group)
ac27a0ec 698{
617ba13b 699 unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
fd2d4291
AM
700 ext4_group_t first = metagroup * EXT4_DESC_PER_BLOCK(sb);
701 ext4_group_t last = first + EXT4_DESC_PER_BLOCK(sb) - 1;
ac27a0ec
DK
702
703 if (group == first || group == first + 1 || group == last)
704 return 1;
705 return 0;
706}
707
fd2d4291
AM
708static unsigned long ext4_bg_num_gdb_nometa(struct super_block *sb,
709 ext4_group_t group)
ac27a0ec 710{
8dadb198
TT
711 if (!ext4_bg_has_super(sb, group))
712 return 0;
713
714 if (EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG))
715 return le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg);
716 else
717 return EXT4_SB(sb)->s_gdb_count;
ac27a0ec
DK
718}
719
720/**
617ba13b 721 * ext4_bg_num_gdb - number of blocks used by the group table in group
ac27a0ec
DK
722 * @sb: superblock for filesystem
723 * @group: group number to check
724 *
725 * Return the number of blocks used by the group descriptor table
726 * (primary or backup) in this group. In the future there may be a
727 * different number of descriptor blocks in each group.
728 */
fd2d4291 729unsigned long ext4_bg_num_gdb(struct super_block *sb, ext4_group_t group)
ac27a0ec
DK
730{
731 unsigned long first_meta_bg =
617ba13b
MC
732 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg);
733 unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
ac27a0ec 734
617ba13b 735 if (!EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 736 metagroup < first_meta_bg)
af5bc92d 737 return ext4_bg_num_gdb_nometa(sb, group);
ac27a0ec 738
617ba13b 739 return ext4_bg_num_gdb_meta(sb,group);
ac27a0ec
DK
740
741}
c2ea3fde 742