blob: b2d10da505eff34d7516e307b664a44555b22658 [file] [log] [blame]
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001/*
Mingming Cao617ba132006-10-11 01:20:53 -07002 * linux/fs/ext4/balloc.c
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003 *
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>
Mingming Caodab291a2006-10-11 01:21:01 -070017#include <linux/jbd2.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070018#include <linux/quotaops.h>
19#include <linux/buffer_head.h>
Christoph Hellwig3dcf5452008-04-29 18:13:32 -040020#include "ext4.h"
21#include "ext4_jbd2.h"
Aneesh Kumar K.Ve21675d2009-01-05 21:36:02 -050022#include "mballoc.h"
Christoph Hellwig3dcf5452008-04-29 18:13:32 -040023
Jiaying Zhang0562e0b2011-03-21 21:38:05 -040024#include <trace/events/ext4.h>
25
Dave Kleikampac27a0e2006-10-11 01:20:50 -070026/*
27 * balloc.c contains the blocks allocation and deallocation routines
28 */
29
30/*
Andrew Morton72b64b52006-10-11 01:21:18 -070031 * Calculate the block group number and offset, given a block number
32 */
33void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
Avantika Mathurfd2d4292008-01-28 23:58:27 -050034 ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp)
Andrew Morton72b64b52006-10-11 01:21:18 -070035{
Dave Kleikamp8c55e202007-05-24 13:04:54 -040036 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Andrew Morton72b64b52006-10-11 01:21:18 -070037 ext4_grpblk_t offset;
38
Dave Kleikamp8c55e202007-05-24 13:04:54 -040039 blocknr = blocknr - le32_to_cpu(es->s_first_data_block);
Andrew Mortonf4e5bc22006-10-11 01:21:19 -070040 offset = do_div(blocknr, EXT4_BLOCKS_PER_GROUP(sb));
Andrew Morton72b64b52006-10-11 01:21:18 -070041 if (offsetp)
42 *offsetp = offset;
43 if (blockgrpp)
Dave Kleikamp8c55e202007-05-24 13:04:54 -040044 *blockgrpp = blocknr;
Andrew Morton72b64b52006-10-11 01:21:18 -070045
46}
47
Jose R. Santos0bf7e832008-06-03 14:07:29 -040048static int ext4_block_in_group(struct super_block *sb, ext4_fsblk_t block,
49 ext4_group_t block_group)
50{
51 ext4_group_t actual_group;
Aneesh Kumar K.V74778272008-07-11 19:27:31 -040052 ext4_get_group_no_and_offset(sb, block, &actual_group, NULL);
Jose R. Santos0bf7e832008-06-03 14:07:29 -040053 if (actual_group == block_group)
54 return 1;
55 return 0;
56}
57
58static int ext4_group_used_meta_blocks(struct super_block *sb,
Theodore Ts'oe187c652009-02-06 16:23:37 -050059 ext4_group_t block_group,
60 struct ext4_group_desc *gdp)
Jose R. Santos0bf7e832008-06-03 14:07:29 -040061{
62 ext4_fsblk_t tmp;
63 struct ext4_sb_info *sbi = EXT4_SB(sb);
64 /* block bitmap, inode bitmap, and inode table blocks */
65 int used_blocks = sbi->s_itb_per_group + 2;
66
67 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) {
Jose R. Santos0bf7e832008-06-03 14:07:29 -040068 if (!ext4_block_in_group(sb, ext4_block_bitmap(sb, gdp),
69 block_group))
70 used_blocks--;
71
72 if (!ext4_block_in_group(sb, ext4_inode_bitmap(sb, gdp),
73 block_group))
74 used_blocks--;
75
76 tmp = ext4_inode_table(sb, gdp);
77 for (; tmp < ext4_inode_table(sb, gdp) +
78 sbi->s_itb_per_group; tmp++) {
79 if (!ext4_block_in_group(sb, tmp, block_group))
80 used_blocks -= 1;
81 }
82 }
83 return used_blocks;
84}
Theodore Ts'oc2ea3fd2008-10-10 09:40:52 -040085
Andreas Dilger717d50e2007-10-16 18:38:25 -040086/* Initializes an uninitialized block bitmap if given, and returns the
87 * number of blocks free in the group. */
88unsigned ext4_init_block_bitmap(struct super_block *sb, struct buffer_head *bh,
Avantika Mathurfd2d4292008-01-28 23:58:27 -050089 ext4_group_t block_group, struct ext4_group_desc *gdp)
Andreas Dilger717d50e2007-10-16 18:38:25 -040090{
Andreas Dilger717d50e2007-10-16 18:38:25 -040091 int bit, bit_max;
Theodore Ts'o8df96752009-05-01 08:50:38 -040092 ext4_group_t ngroups = ext4_get_groups_count(sb);
Andreas Dilger717d50e2007-10-16 18:38:25 -040093 unsigned free_blocks, group_blocks;
94 struct ext4_sb_info *sbi = EXT4_SB(sb);
95
96 if (bh) {
97 J_ASSERT_BH(bh, buffer_locked(bh));
98
99 /* If checksum is bad mark all blocks used to prevent allocation
100 * essentially implementing a per-group read-only flag. */
101 if (!ext4_group_desc_csum_verify(sbi, block_group, gdp)) {
Eric Sandeen12062dd2010-02-15 14:19:27 -0500102 ext4_error(sb, "Checksum bad for group %u",
103 block_group);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500104 ext4_free_blks_set(sb, gdp, 0);
105 ext4_free_inodes_set(sb, gdp, 0);
106 ext4_itable_unused_set(sb, gdp, 0);
Andreas Dilger717d50e2007-10-16 18:38:25 -0400107 memset(bh->b_data, 0xff, sb->s_blocksize);
108 return 0;
109 }
110 memset(bh->b_data, 0, sb->s_blocksize);
111 }
112
113 /* Check for superblock and gdt backups in this group */
114 bit_max = ext4_bg_has_super(sb, block_group);
115
116 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
117 block_group < le32_to_cpu(sbi->s_es->s_first_meta_bg) *
118 sbi->s_desc_per_block) {
119 if (bit_max) {
120 bit_max += ext4_bg_num_gdb(sb, block_group);
121 bit_max +=
122 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks);
123 }
124 } else { /* For META_BG_BLOCK_GROUPS */
Akinobu Mita6afd6702008-07-11 19:27:31 -0400125 bit_max += ext4_bg_num_gdb(sb, block_group);
Andreas Dilger717d50e2007-10-16 18:38:25 -0400126 }
127
Theodore Ts'o8df96752009-05-01 08:50:38 -0400128 if (block_group == ngroups - 1) {
Andreas Dilger717d50e2007-10-16 18:38:25 -0400129 /*
130 * Even though mke2fs always initialize first and last group
131 * if some other tool enabled the EXT4_BG_BLOCK_UNINIT we need
132 * to make sure we calculate the right free blocks
133 */
134 group_blocks = ext4_blocks_count(sbi->s_es) -
Akinobu Mita5661bd62010-03-03 23:53:39 -0500135 ext4_group_first_block_no(sb, ngroups - 1);
Andreas Dilger717d50e2007-10-16 18:38:25 -0400136 } else {
137 group_blocks = EXT4_BLOCKS_PER_GROUP(sb);
138 }
139
140 free_blocks = group_blocks - bit_max;
141
142 if (bh) {
Jose R. Santos0bf7e832008-06-03 14:07:29 -0400143 ext4_fsblk_t start, tmp;
144 int flex_bg = 0;
Akinobu Mitad00a6d72008-04-17 10:38:59 -0400145
Andreas Dilger717d50e2007-10-16 18:38:25 -0400146 for (bit = 0; bit < bit_max; bit++)
147 ext4_set_bit(bit, bh->b_data);
148
Akinobu Mitad00a6d72008-04-17 10:38:59 -0400149 start = ext4_group_first_block_no(sb, block_group);
Andreas Dilger717d50e2007-10-16 18:38:25 -0400150
Jose R. Santos0bf7e832008-06-03 14:07:29 -0400151 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
152 EXT4_FEATURE_INCOMPAT_FLEX_BG))
153 flex_bg = 1;
Andreas Dilger717d50e2007-10-16 18:38:25 -0400154
Jose R. Santos0bf7e832008-06-03 14:07:29 -0400155 /* Set bits for block and inode bitmaps, and inode table */
156 tmp = ext4_block_bitmap(sb, gdp);
157 if (!flex_bg || ext4_block_in_group(sb, tmp, block_group))
158 ext4_set_bit(tmp - start, bh->b_data);
159
160 tmp = ext4_inode_bitmap(sb, gdp);
161 if (!flex_bg || ext4_block_in_group(sb, tmp, block_group))
162 ext4_set_bit(tmp - start, bh->b_data);
163
164 tmp = ext4_inode_table(sb, gdp);
165 for (; tmp < ext4_inode_table(sb, gdp) +
166 sbi->s_itb_per_group; tmp++) {
167 if (!flex_bg ||
168 ext4_block_in_group(sb, tmp, block_group))
169 ext4_set_bit(tmp - start, bh->b_data);
170 }
Andreas Dilger717d50e2007-10-16 18:38:25 -0400171 /*
172 * Also if the number of blocks within the group is
173 * less than the blocksize * 8 ( which is the size
174 * of bitmap ), set rest of the block bitmap to 1
175 */
Theodore Ts'o61d08672010-10-27 21:30:15 -0400176 ext4_mark_bitmap_end(group_blocks, sb->s_blocksize * 8,
177 bh->b_data);
Andreas Dilger717d50e2007-10-16 18:38:25 -0400178 }
Theodore Ts'oe187c652009-02-06 16:23:37 -0500179 return free_blocks - ext4_group_used_meta_blocks(sb, block_group, gdp);
Andreas Dilger717d50e2007-10-16 18:38:25 -0400180}
181
182
Andrew Morton72b64b52006-10-11 01:21:18 -0700183/*
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700184 * The free blocks are managed by bitmaps. A file system contains several
185 * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap
186 * block for inodes, N blocks for the inode table and data blocks.
187 *
188 * The file system contains group descriptors which are located after the
189 * super block. Each descriptor contains the number of the bitmap block and
190 * the free blocks count in the block. The descriptors are loaded in memory
Aneesh Kumar K.Ve6274322007-02-20 13:57:58 -0800191 * when a file system is mounted (see ext4_fill_super).
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700192 */
193
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700194/**
Mingming Cao617ba132006-10-11 01:20:53 -0700195 * ext4_get_group_desc() -- load group descriptor from disk
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700196 * @sb: super block
197 * @block_group: given block group
198 * @bh: pointer to the buffer head to store the block
199 * group descriptor
200 */
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400201struct ext4_group_desc * ext4_get_group_desc(struct super_block *sb,
Avantika Mathurfd2d4292008-01-28 23:58:27 -0500202 ext4_group_t block_group,
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400203 struct buffer_head **bh)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700204{
Theodore Ts'o498e5f22008-11-05 00:14:04 -0500205 unsigned int group_desc;
206 unsigned int offset;
Theodore Ts'o8df96752009-05-01 08:50:38 -0400207 ext4_group_t ngroups = ext4_get_groups_count(sb);
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400208 struct ext4_group_desc *desc;
Mingming Cao617ba132006-10-11 01:20:53 -0700209 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700210
Theodore Ts'o8df96752009-05-01 08:50:38 -0400211 if (block_group >= ngroups) {
Eric Sandeen12062dd2010-02-15 14:19:27 -0500212 ext4_error(sb, "block_group >= groups_count - block_group = %u,"
213 " groups_count = %u", block_group, ngroups);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700214
215 return NULL;
216 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700217
Mingming Cao617ba132006-10-11 01:20:53 -0700218 group_desc = block_group >> EXT4_DESC_PER_BLOCK_BITS(sb);
219 offset = block_group & (EXT4_DESC_PER_BLOCK(sb) - 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700220 if (!sbi->s_group_desc[group_desc]) {
Eric Sandeen12062dd2010-02-15 14:19:27 -0500221 ext4_error(sb, "Group descriptor not loaded - "
Theodore Ts'o498e5f22008-11-05 00:14:04 -0500222 "block_group = %u, group_desc = %u, desc = %u",
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400223 block_group, group_desc, offset);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700224 return NULL;
225 }
226
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -0700227 desc = (struct ext4_group_desc *)(
228 (__u8 *)sbi->s_group_desc[group_desc]->b_data +
229 offset * EXT4_DESC_SIZE(sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700230 if (bh)
231 *bh = sbi->s_group_desc[group_desc];
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -0700232 return desc;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700233}
234
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500235static int ext4_valid_block_bitmap(struct super_block *sb,
236 struct ext4_group_desc *desc,
237 unsigned int block_group,
238 struct buffer_head *bh)
239{
240 ext4_grpblk_t offset;
241 ext4_grpblk_t next_zero_bit;
242 ext4_fsblk_t bitmap_blk;
243 ext4_fsblk_t group_first_block;
244
245 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) {
246 /* with FLEX_BG, the inode/block bitmaps and itable
247 * blocks may not be in the group at all
248 * so the bitmap validation will be skipped for those groups
249 * or it has to also read the block group where the bitmaps
250 * are located to verify they are set.
251 */
252 return 1;
253 }
254 group_first_block = ext4_group_first_block_no(sb, block_group);
255
256 /* check whether block bitmap block number is set */
257 bitmap_blk = ext4_block_bitmap(sb, desc);
258 offset = bitmap_blk - group_first_block;
259 if (!ext4_test_bit(offset, bh->b_data))
260 /* bad block bitmap */
261 goto err_out;
262
263 /* check whether the inode bitmap block number is set */
264 bitmap_blk = ext4_inode_bitmap(sb, desc);
265 offset = bitmap_blk - group_first_block;
266 if (!ext4_test_bit(offset, bh->b_data))
267 /* bad block bitmap */
268 goto err_out;
269
270 /* check whether the inode table block number is set */
271 bitmap_blk = ext4_inode_table(sb, desc);
272 offset = bitmap_blk - group_first_block;
273 next_zero_bit = ext4_find_next_zero_bit(bh->b_data,
274 offset + EXT4_SB(sb)->s_itb_per_group,
275 offset);
276 if (next_zero_bit >= offset + EXT4_SB(sb)->s_itb_per_group)
277 /* good bitmap for inode tables */
278 return 1;
279
280err_out:
Eric Sandeen12062dd2010-02-15 14:19:27 -0500281 ext4_error(sb, "Invalid block bitmap - block_group = %d, block = %llu",
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500282 block_group, bitmap_blk);
283 return 0;
284}
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700285/**
Theodore Ts'o574ca172008-07-11 19:27:31 -0400286 * ext4_read_block_bitmap()
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700287 * @sb: super block
288 * @block_group: given block group
289 *
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500290 * Read the bitmap for a given block_group,and validate the
291 * bits for block/inode/inode tables are set in the bitmaps
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700292 *
293 * Return buffer_head on success or NULL in case of failure.
294 */
Andreas Dilger717d50e2007-10-16 18:38:25 -0400295struct buffer_head *
Theodore Ts'o574ca172008-07-11 19:27:31 -0400296ext4_read_block_bitmap(struct super_block *sb, ext4_group_t block_group)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700297{
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400298 struct ext4_group_desc *desc;
299 struct buffer_head *bh = NULL;
Aneesh Kumar K.V7c9e69f2007-10-16 23:27:02 -0700300 ext4_fsblk_t bitmap_blk;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700301
Andreas Dilger717d50e2007-10-16 18:38:25 -0400302 desc = ext4_get_group_desc(sb, block_group, NULL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700303 if (!desc)
Aneesh Kumar K.V7c9e69f2007-10-16 23:27:02 -0700304 return NULL;
305 bitmap_blk = ext4_block_bitmap(sb, desc);
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500306 bh = sb_getblk(sb, bitmap_blk);
307 if (unlikely(!bh)) {
Eric Sandeen12062dd2010-02-15 14:19:27 -0500308 ext4_error(sb, "Cannot read block bitmap - "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -0500309 "block_group = %u, block_bitmap = %llu",
Eric Sandeene29d1cd2008-08-02 21:21:02 -0400310 block_group, bitmap_blk);
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500311 return NULL;
312 }
Aneesh Kumar K.V2ccb5fb2009-01-05 21:49:55 -0500313
314 if (bitmap_uptodate(bh))
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500315 return bh;
316
Frederic Bohec806e682008-10-10 08:09:18 -0400317 lock_buffer(bh);
Aneesh Kumar K.V2ccb5fb2009-01-05 21:49:55 -0500318 if (bitmap_uptodate(bh)) {
319 unlock_buffer(bh);
320 return bh;
321 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -0400322 ext4_lock_group(sb, block_group);
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500323 if (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) {
324 ext4_init_block_bitmap(sb, bh, block_group, desc);
Aneesh Kumar K.V2ccb5fb2009-01-05 21:49:55 -0500325 set_bitmap_uptodate(bh);
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500326 set_buffer_uptodate(bh);
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -0400327 ext4_unlock_group(sb, block_group);
Aneesh Kumar K.V3300bed2009-01-03 22:33:39 -0500328 unlock_buffer(bh);
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500329 return bh;
330 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -0400331 ext4_unlock_group(sb, block_group);
Aneesh Kumar K.V2ccb5fb2009-01-05 21:49:55 -0500332 if (buffer_uptodate(bh)) {
333 /*
334 * if not uninit if bh is uptodate,
335 * bitmap is also uptodate
336 */
337 set_bitmap_uptodate(bh);
338 unlock_buffer(bh);
339 return bh;
340 }
341 /*
342 * submit the buffer_head for read. We can
343 * safely mark the bitmap as uptodate now.
344 * We do it here so the bitmap uptodate bit
345 * get set with buffer lock held.
346 */
Jiaying Zhang0562e0b2011-03-21 21:38:05 -0400347 trace_ext4_read_block_bitmap_load(sb, block_group);
Aneesh Kumar K.V2ccb5fb2009-01-05 21:49:55 -0500348 set_bitmap_uptodate(bh);
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500349 if (bh_submit_read(bh) < 0) {
350 put_bh(bh);
Eric Sandeen12062dd2010-02-15 14:19:27 -0500351 ext4_error(sb, "Cannot read block bitmap - "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -0500352 "block_group = %u, block_bitmap = %llu",
Eric Sandeene29d1cd2008-08-02 21:21:02 -0400353 block_group, bitmap_blk);
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500354 return NULL;
355 }
Aneesh Kumar K.V519deca02008-05-15 14:43:20 -0400356 ext4_valid_block_bitmap(sb, desc, block_group, bh);
357 /*
358 * file system mounted not to panic on error,
359 * continue with corrupt bitmap
360 */
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700361 return bh;
362}
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700363
364/**
Eric Sandeen8c3bf8a02008-10-28 00:08:12 -0400365 * ext4_has_free_blocks()
366 * @sbi: in-core super block structure.
367 * @nblocks: number of needed blocks
368 *
369 * Check if filesystem has nblocks free & available for allocation.
370 * On success return 1, return 0 on failure.
371 */
Theodore Ts'o1f109d52010-10-27 21:30:14 -0400372static int ext4_has_free_blocks(struct ext4_sb_info *sbi, s64 nblocks)
Aneesh Kumar K.Va30d542a2008-10-09 10:56:23 -0400373{
Eric Sandeena9960312008-10-28 00:08:17 -0400374 s64 free_blocks, dirty_blocks, root_blocks;
Aneesh Kumar K.Va30d542a2008-10-09 10:56:23 -0400375 struct percpu_counter *fbc = &sbi->s_freeblocks_counter;
Aneesh Kumar K.V6bc6e632008-10-10 09:39:00 -0400376 struct percpu_counter *dbc = &sbi->s_dirtyblocks_counter;
Aneesh Kumar K.Va30d542a2008-10-09 10:56:23 -0400377
Aneesh Kumar K.V6bc6e632008-10-10 09:39:00 -0400378 free_blocks = percpu_counter_read_positive(fbc);
379 dirty_blocks = percpu_counter_read_positive(dbc);
Eric Sandeena9960312008-10-28 00:08:17 -0400380 root_blocks = ext4_r_blocks_count(sbi->s_es);
Aneesh Kumar K.Va30d542a2008-10-09 10:56:23 -0400381
Aneesh Kumar K.V6bc6e632008-10-10 09:39:00 -0400382 if (free_blocks - (nblocks + root_blocks + dirty_blocks) <
383 EXT4_FREEBLOCKS_WATERMARK) {
Andrew Morton02d21162008-12-09 13:14:14 -0800384 free_blocks = percpu_counter_sum_positive(fbc);
385 dirty_blocks = percpu_counter_sum_positive(dbc);
Aneesh Kumar K.V6bc6e632008-10-10 09:39:00 -0400386 }
387 /* Check whether we have space after
Eric Sandeena9960312008-10-28 00:08:17 -0400388 * accounting for current dirty blocks & root reserved blocks.
Aneesh Kumar K.V6bc6e632008-10-10 09:39:00 -0400389 */
Eric Sandeena9960312008-10-28 00:08:17 -0400390 if (free_blocks >= ((root_blocks + nblocks) + dirty_blocks))
391 return 1;
Aneesh Kumar K.V5c791612008-10-08 23:12:24 -0400392
Eric Sandeena9960312008-10-28 00:08:17 -0400393 /* Hm, nope. Are (enough) root reserved blocks available? */
David Howells4c9c5442008-11-14 10:38:51 +1100394 if (sbi->s_resuid == current_fsuid() ||
Eric Sandeena9960312008-10-28 00:08:17 -0400395 ((sbi->s_resgid != 0) && in_group_p(sbi->s_resgid)) ||
396 capable(CAP_SYS_RESOURCE)) {
397 if (free_blocks >= (nblocks + dirty_blocks))
398 return 1;
399 }
400
401 return 0;
Aneesh Kumar K.Va30d542a2008-10-09 10:56:23 -0400402}
Mingming Cao07031432008-07-11 19:27:31 -0400403
Eric Sandeen8c3bf8a02008-10-28 00:08:12 -0400404int ext4_claim_free_blocks(struct ext4_sb_info *sbi,
405 s64 nblocks)
406{
407 if (ext4_has_free_blocks(sbi, nblocks)) {
408 percpu_counter_add(&sbi->s_dirtyblocks_counter, nblocks);
409 return 0;
410 } else
411 return -ENOSPC;
412}
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700413
414/**
Mingming Cao617ba132006-10-11 01:20:53 -0700415 * ext4_should_retry_alloc()
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700416 * @sb: super block
417 * @retries number of attemps has been made
418 *
Mingming Cao617ba132006-10-11 01:20:53 -0700419 * ext4_should_retry_alloc() is called when ENOSPC is returned, and if
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700420 * it is profitable to retry the operation, this function will wait
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300421 * for the current or committing transaction to complete, and then
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700422 * return TRUE.
423 *
424 * if the total number of retries exceed three times, return FALSE.
425 */
Mingming Cao617ba132006-10-11 01:20:53 -0700426int ext4_should_retry_alloc(struct super_block *sb, int *retries)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700427{
Eric Sandeen8f64b322009-02-26 00:57:35 -0500428 if (!ext4_has_free_blocks(EXT4_SB(sb), 1) ||
429 (*retries)++ > 3 ||
430 !EXT4_SB(sb)->s_journal)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700431 return 0;
432
433 jbd_debug(1, "%s: retrying operation after ENOSPC\n", sb->s_id);
434
Mingming Caodab291a2006-10-11 01:21:01 -0700435 return jbd2_journal_force_commit_nested(EXT4_SB(sb)->s_journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700436}
437
Aneesh Kumar K.V654b4902008-07-11 19:27:31 -0400438/*
Aneesh Kumar K.V654b4902008-07-11 19:27:31 -0400439 * ext4_new_meta_blocks() -- allocate block for meta data (indexing) blocks
440 *
441 * @handle: handle to this transaction
442 * @inode: file inode
443 * @goal: given target block(filesystem wide)
Theodore Ts'o97df5d12008-12-12 12:41:28 -0500444 * @count: pointer to total number of blocks needed
Aneesh Kumar K.V654b4902008-07-11 19:27:31 -0400445 * @errp: error code
446 *
Theodore Ts'o97df5d12008-12-12 12:41:28 -0500447 * Return 1st allocated block number on success, *count stores total account
Aneesh Kumar K.V654b4902008-07-11 19:27:31 -0400448 * error stores in errp pointer
449 */
450ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
451 ext4_fsblk_t goal, unsigned long *count, int *errp)
452{
Theodore Ts'o97df5d12008-12-12 12:41:28 -0500453 struct ext4_allocation_request ar;
Mingming Caod2a17632008-07-14 17:52:37 -0400454 ext4_fsblk_t ret;
Theodore Ts'o97df5d12008-12-12 12:41:28 -0500455
456 memset(&ar, 0, sizeof(ar));
457 /* Fill with neighbour allocated blocks */
458 ar.inode = inode;
459 ar.goal = goal;
460 ar.len = count ? *count : 1;
461
462 ret = ext4_mb_new_blocks(handle, &ar, errp);
463 if (count)
464 *count = ar.len;
Mingming Caod2a17632008-07-14 17:52:37 -0400465 /*
Eric Sandeen72b8ab92010-05-16 11:00:00 -0400466 * Account for the allocated meta blocks. We will never
467 * fail EDQUOT for metdata, but we do account for it.
Mingming Caod2a17632008-07-14 17:52:37 -0400468 */
Theodore Ts'of2321092011-01-10 12:12:36 -0500469 if (!(*errp) &&
470 ext4_test_inode_state(inode, EXT4_STATE_DELALLOC_RESERVED)) {
Mingming Caod2a17632008-07-14 17:52:37 -0400471 spin_lock(&EXT4_I(inode)->i_block_reservation_lock);
Theodore Ts'o97df5d12008-12-12 12:41:28 -0500472 EXT4_I(inode)->i_allocated_meta_blocks += ar.len;
Mingming Caod2a17632008-07-14 17:52:37 -0400473 spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
Eric Sandeen72b8ab92010-05-16 11:00:00 -0400474 dquot_alloc_block_nofail(inode, ar.len);
Mingming Caod2a17632008-07-14 17:52:37 -0400475 }
476 return ret;
477}
478
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700479/**
Mingming Cao617ba132006-10-11 01:20:53 -0700480 * ext4_count_free_blocks() -- count filesystem free blocks
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700481 * @sb: superblock
482 *
483 * Adds up the number of free blocks from each block group.
484 */
Mingming Cao617ba132006-10-11 01:20:53 -0700485ext4_fsblk_t ext4_count_free_blocks(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700486{
Mingming Cao617ba132006-10-11 01:20:53 -0700487 ext4_fsblk_t desc_count;
488 struct ext4_group_desc *gdp;
Avantika Mathurfd2d4292008-01-28 23:58:27 -0500489 ext4_group_t i;
Theodore Ts'o8df96752009-05-01 08:50:38 -0400490 ext4_group_t ngroups = ext4_get_groups_count(sb);
Mingming Cao617ba132006-10-11 01:20:53 -0700491#ifdef EXT4FS_DEBUG
492 struct ext4_super_block *es;
493 ext4_fsblk_t bitmap_count;
Theodore Ts'o498e5f22008-11-05 00:14:04 -0500494 unsigned int x;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700495 struct buffer_head *bitmap_bh = NULL;
496
Mingming Cao617ba132006-10-11 01:20:53 -0700497 es = EXT4_SB(sb)->s_es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700498 desc_count = 0;
499 bitmap_count = 0;
500 gdp = NULL;
501
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700502 for (i = 0; i < ngroups; i++) {
Mingming Cao617ba132006-10-11 01:20:53 -0700503 gdp = ext4_get_group_desc(sb, i, NULL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700504 if (!gdp)
505 continue;
Thadeu Lima de Souza Cascardo9fd97842009-01-26 19:26:26 -0500506 desc_count += ext4_free_blks_count(sb, gdp);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700507 brelse(bitmap_bh);
Theodore Ts'o574ca172008-07-11 19:27:31 -0400508 bitmap_bh = ext4_read_block_bitmap(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700509 if (bitmap_bh == NULL)
510 continue;
511
Mingming Cao617ba132006-10-11 01:20:53 -0700512 x = ext4_count_free(bitmap_bh, sb->s_blocksize);
Thadeu Lima de Souza Cascardo9fd97842009-01-26 19:26:26 -0500513 printk(KERN_DEBUG "group %u: stored = %d, counted = %u\n",
514 i, ext4_free_blks_count(sb, gdp), x);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700515 bitmap_count += x;
516 }
517 brelse(bitmap_bh);
Theodore Ts'o4776004f2008-09-08 23:00:52 -0400518 printk(KERN_DEBUG "ext4_count_free_blocks: stored = %llu"
519 ", computed = %llu, %llu\n", ext4_free_blocks_count(es),
520 desc_count, bitmap_count);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700521 return bitmap_count;
522#else
523 desc_count = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700524 for (i = 0; i < ngroups; i++) {
Mingming Cao617ba132006-10-11 01:20:53 -0700525 gdp = ext4_get_group_desc(sb, i, NULL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700526 if (!gdp)
527 continue;
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500528 desc_count += ext4_free_blks_count(sb, gdp);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700529 }
530
531 return desc_count;
532#endif
533}
534
Avantika Mathurfd2d4292008-01-28 23:58:27 -0500535static inline int test_root(ext4_group_t a, int b)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700536{
537 int num = b;
538
539 while (a > num)
540 num *= b;
541 return num == a;
542}
543
Avantika Mathurfd2d4292008-01-28 23:58:27 -0500544static int ext4_group_sparse(ext4_group_t group)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700545{
546 if (group <= 1)
547 return 1;
548 if (!(group & 1))
549 return 0;
550 return (test_root(group, 7) || test_root(group, 5) ||
551 test_root(group, 3));
552}
553
554/**
Mingming Cao617ba132006-10-11 01:20:53 -0700555 * ext4_bg_has_super - number of blocks used by the superblock in group
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700556 * @sb: superblock for filesystem
557 * @group: group number to check
558 *
559 * Return the number of blocks used by the superblock (primary or backup)
560 * in this group. Currently this will be only 0 or 1.
561 */
Avantika Mathurfd2d4292008-01-28 23:58:27 -0500562int ext4_bg_has_super(struct super_block *sb, ext4_group_t group)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700563{
Mingming Cao617ba132006-10-11 01:20:53 -0700564 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
565 EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER) &&
566 !ext4_group_sparse(group))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700567 return 0;
568 return 1;
569}
570
Avantika Mathurfd2d4292008-01-28 23:58:27 -0500571static unsigned long ext4_bg_num_gdb_meta(struct super_block *sb,
572 ext4_group_t group)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700573{
Mingming Cao617ba132006-10-11 01:20:53 -0700574 unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
Avantika Mathurfd2d4292008-01-28 23:58:27 -0500575 ext4_group_t first = metagroup * EXT4_DESC_PER_BLOCK(sb);
576 ext4_group_t last = first + EXT4_DESC_PER_BLOCK(sb) - 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700577
578 if (group == first || group == first + 1 || group == last)
579 return 1;
580 return 0;
581}
582
Avantika Mathurfd2d4292008-01-28 23:58:27 -0500583static unsigned long ext4_bg_num_gdb_nometa(struct super_block *sb,
584 ext4_group_t group)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700585{
Theodore Ts'o8dadb192009-11-23 07:24:38 -0500586 if (!ext4_bg_has_super(sb, group))
587 return 0;
588
589 if (EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG))
590 return le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg);
591 else
592 return EXT4_SB(sb)->s_gdb_count;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700593}
594
595/**
Mingming Cao617ba132006-10-11 01:20:53 -0700596 * ext4_bg_num_gdb - number of blocks used by the group table in group
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700597 * @sb: superblock for filesystem
598 * @group: group number to check
599 *
600 * Return the number of blocks used by the group descriptor table
601 * (primary or backup) in this group. In the future there may be a
602 * different number of descriptor blocks in each group.
603 */
Avantika Mathurfd2d4292008-01-28 23:58:27 -0500604unsigned long ext4_bg_num_gdb(struct super_block *sb, ext4_group_t group)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700605{
606 unsigned long first_meta_bg =
Mingming Cao617ba132006-10-11 01:20:53 -0700607 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg);
608 unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700609
Mingming Cao617ba132006-10-11 01:20:53 -0700610 if (!EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700611 metagroup < first_meta_bg)
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400612 return ext4_bg_num_gdb_nometa(sb, group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700613
Mingming Cao617ba132006-10-11 01:20:53 -0700614 return ext4_bg_num_gdb_meta(sb,group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700615
616}
Theodore Ts'oc2ea3fd2008-10-10 09:40:52 -0400617