blob: b430afe77f0cec7bb550fb6466bd878945572aed [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
Eric Sandeen5f163cc2012-01-04 22:33:28 -050026static unsigned ext4_num_base_meta_clusters(struct super_block *sb,
27 ext4_group_t block_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -070028/*
29 * balloc.c contains the blocks allocation and deallocation routines
30 */
31
32/*
Lukas Czernerbd862982013-04-03 23:32:34 -040033 * Calculate block group number for a given block number
34 */
35ext4_group_t ext4_get_group_number(struct super_block *sb,
36 ext4_fsblk_t block)
37{
38 ext4_group_t group;
39
40 if (test_opt2(sb, STD_GROUP_SIZE))
Theodore Ts'o960fd8562013-07-05 23:11:16 -040041 group = (block -
42 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block)) >>
Lukas Czernerbd862982013-04-03 23:32:34 -040043 (EXT4_BLOCK_SIZE_BITS(sb) + EXT4_CLUSTER_BITS(sb) + 3);
44 else
45 ext4_get_group_no_and_offset(sb, block, &group, NULL);
46 return group;
47}
48
49/*
Theodore Ts'o3212a802011-09-09 18:46:51 -040050 * Calculate the block group number and offset into the block/cluster
51 * allocation bitmap, given a block number
Andrew Morton72b64b52006-10-11 01:21:18 -070052 */
53void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
Avantika Mathurfd2d4292008-01-28 23:58:27 -050054 ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp)
Andrew Morton72b64b52006-10-11 01:21:18 -070055{
Dave Kleikamp8c55e202007-05-24 13:04:54 -040056 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Andrew Morton72b64b52006-10-11 01:21:18 -070057 ext4_grpblk_t offset;
58
Dave Kleikamp8c55e202007-05-24 13:04:54 -040059 blocknr = blocknr - le32_to_cpu(es->s_first_data_block);
Theodore Ts'o3212a802011-09-09 18:46:51 -040060 offset = do_div(blocknr, EXT4_BLOCKS_PER_GROUP(sb)) >>
61 EXT4_SB(sb)->s_cluster_bits;
Andrew Morton72b64b52006-10-11 01:21:18 -070062 if (offsetp)
63 *offsetp = offset;
64 if (blockgrpp)
Dave Kleikamp8c55e202007-05-24 13:04:54 -040065 *blockgrpp = blocknr;
Andrew Morton72b64b52006-10-11 01:21:18 -070066
67}
68
Lukas Czerner68911002013-04-03 22:12:52 -040069/*
70 * Check whether the 'block' lives within the 'block_group'. Returns 1 if so
71 * and 0 otherwise.
72 */
73static inline int ext4_block_in_group(struct super_block *sb,
74 ext4_fsblk_t block,
75 ext4_group_t block_group)
Jose R. Santos0bf7e832008-06-03 14:07:29 -040076{
77 ext4_group_t actual_group;
Lukas Czerner68911002013-04-03 22:12:52 -040078
Lukas Czernerbd862982013-04-03 23:32:34 -040079 actual_group = ext4_get_group_number(sb, block);
Lukas Czerner68911002013-04-03 22:12:52 -040080 return (actual_group == block_group) ? 1 : 0;
Jose R. Santos0bf7e832008-06-03 14:07:29 -040081}
82
Theodore Ts'od5b8f312011-09-09 18:44:51 -040083/* Return the number of clusters used for file system metadata; this
84 * represents the overhead needed by the file system.
85 */
86unsigned ext4_num_overhead_clusters(struct super_block *sb,
87 ext4_group_t block_group,
88 struct ext4_group_desc *gdp)
Jose R. Santos0bf7e832008-06-03 14:07:29 -040089{
Theodore Ts'od5b8f312011-09-09 18:44:51 -040090 unsigned num_clusters;
91 int block_cluster = -1, inode_cluster = -1, itbl_cluster = -1, i, c;
92 ext4_fsblk_t start = ext4_group_first_block_no(sb, block_group);
93 ext4_fsblk_t itbl_blk;
Jose R. Santos0bf7e832008-06-03 14:07:29 -040094 struct ext4_sb_info *sbi = EXT4_SB(sb);
Jose R. Santos0bf7e832008-06-03 14:07:29 -040095
Theodore Ts'od5b8f312011-09-09 18:44:51 -040096 /* This is the number of clusters used by the superblock,
97 * block group descriptors, and reserved block group
98 * descriptor blocks */
99 num_clusters = ext4_num_base_meta_clusters(sb, block_group);
Jose R. Santos0bf7e832008-06-03 14:07:29 -0400100
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400101 /*
102 * For the allocation bitmaps and inode table, we first need
103 * to check to see if the block is in the block group. If it
104 * is, then check to see if the cluster is already accounted
105 * for in the clusters used for the base metadata cluster, or
106 * if we can increment the base metadata cluster to include
107 * that block. Otherwise, we will have to track the cluster
108 * used for the allocation bitmap or inode table explicitly.
109 * Normally all of these blocks are contiguous, so the special
110 * case handling shouldn't be necessary except for *very*
111 * unusual file system layouts.
112 */
113 if (ext4_block_in_group(sb, ext4_block_bitmap(sb, gdp), block_group)) {
Theodore Ts'ob0dd6b72012-06-07 18:56:06 -0400114 block_cluster = EXT4_B2C(sbi,
115 ext4_block_bitmap(sb, gdp) - start);
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400116 if (block_cluster < num_clusters)
117 block_cluster = -1;
118 else if (block_cluster == num_clusters) {
119 num_clusters++;
120 block_cluster = -1;
Jose R. Santos0bf7e832008-06-03 14:07:29 -0400121 }
122 }
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400123
124 if (ext4_block_in_group(sb, ext4_inode_bitmap(sb, gdp), block_group)) {
125 inode_cluster = EXT4_B2C(sbi,
Theodore Ts'ob0dd6b72012-06-07 18:56:06 -0400126 ext4_inode_bitmap(sb, gdp) - start);
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400127 if (inode_cluster < num_clusters)
128 inode_cluster = -1;
129 else if (inode_cluster == num_clusters) {
130 num_clusters++;
131 inode_cluster = -1;
132 }
133 }
134
135 itbl_blk = ext4_inode_table(sb, gdp);
136 for (i = 0; i < sbi->s_itb_per_group; i++) {
137 if (ext4_block_in_group(sb, itbl_blk + i, block_group)) {
Theodore Ts'ob0dd6b72012-06-07 18:56:06 -0400138 c = EXT4_B2C(sbi, itbl_blk + i - start);
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400139 if ((c < num_clusters) || (c == inode_cluster) ||
140 (c == block_cluster) || (c == itbl_cluster))
141 continue;
142 if (c == num_clusters) {
143 num_clusters++;
144 continue;
145 }
146 num_clusters++;
147 itbl_cluster = c;
148 }
149 }
150
151 if (block_cluster != -1)
152 num_clusters++;
153 if (inode_cluster != -1)
154 num_clusters++;
155
156 return num_clusters;
Jose R. Santos0bf7e832008-06-03 14:07:29 -0400157}
Theodore Ts'oc2ea3fd2008-10-10 09:40:52 -0400158
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400159static unsigned int num_clusters_in_group(struct super_block *sb,
160 ext4_group_t block_group)
Theodore Ts'o49f7f9a2011-09-09 18:40:51 -0400161{
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400162 unsigned int blocks;
163
Theodore Ts'o49f7f9a2011-09-09 18:40:51 -0400164 if (block_group == ext4_get_groups_count(sb) - 1) {
165 /*
166 * Even though mke2fs always initializes the first and
167 * last group, just in case some other tool was used,
168 * we need to make sure we calculate the right free
169 * blocks.
170 */
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400171 blocks = ext4_blocks_count(EXT4_SB(sb)->s_es) -
Theodore Ts'o49f7f9a2011-09-09 18:40:51 -0400172 ext4_group_first_block_no(sb, block_group);
173 } else
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400174 blocks = EXT4_BLOCKS_PER_GROUP(sb);
175 return EXT4_NUM_B2C(EXT4_SB(sb), blocks);
Theodore Ts'o49f7f9a2011-09-09 18:40:51 -0400176}
177
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400178/* Initializes an uninitialized block bitmap */
179void ext4_init_block_bitmap(struct super_block *sb, struct buffer_head *bh,
180 ext4_group_t block_group,
181 struct ext4_group_desc *gdp)
Andreas Dilger717d50e2007-10-16 18:38:25 -0400182{
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400183 unsigned int bit, bit_max;
Andreas Dilger717d50e2007-10-16 18:38:25 -0400184 struct ext4_sb_info *sbi = EXT4_SB(sb);
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400185 ext4_fsblk_t start, tmp;
186 int flex_bg = 0;
Andreas Dilger717d50e2007-10-16 18:38:25 -0400187
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400188 J_ASSERT_BH(bh, buffer_locked(bh));
Andreas Dilger717d50e2007-10-16 18:38:25 -0400189
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400190 /* If checksum is bad mark all blocks used to prevent allocation
191 * essentially implementing a per-group read-only flag. */
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -0400192 if (!ext4_group_desc_csum_verify(sb, block_group, gdp)) {
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400193 ext4_error(sb, "Checksum bad for group %u", block_group);
Theodore Ts'o021b65b2011-09-09 19:08:51 -0400194 ext4_free_group_clusters_set(sb, gdp, 0);
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400195 ext4_free_inodes_set(sb, gdp, 0);
196 ext4_itable_unused_set(sb, gdp, 0);
197 memset(bh->b_data, 0xff, sb->s_blocksize);
Tao Ma79f1ba42012-10-22 00:34:32 -0400198 ext4_block_bitmap_csum_set(sb, block_group, gdp, bh);
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400199 return;
Andreas Dilger717d50e2007-10-16 18:38:25 -0400200 }
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400201 memset(bh->b_data, 0, sb->s_blocksize);
Andreas Dilger717d50e2007-10-16 18:38:25 -0400202
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400203 bit_max = ext4_num_base_meta_clusters(sb, block_group);
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400204 for (bit = 0; bit < bit_max; bit++)
205 ext4_set_bit(bit, bh->b_data);
Akinobu Mitad00a6d72008-04-17 10:38:59 -0400206
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400207 start = ext4_group_first_block_no(sb, block_group);
Andreas Dilger717d50e2007-10-16 18:38:25 -0400208
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400209 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
210 flex_bg = 1;
Andreas Dilger717d50e2007-10-16 18:38:25 -0400211
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400212 /* Set bits for block and inode bitmaps, and inode table */
213 tmp = ext4_block_bitmap(sb, gdp);
214 if (!flex_bg || ext4_block_in_group(sb, tmp, block_group))
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400215 ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data);
Andreas Dilger717d50e2007-10-16 18:38:25 -0400216
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400217 tmp = ext4_inode_bitmap(sb, gdp);
218 if (!flex_bg || ext4_block_in_group(sb, tmp, block_group))
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400219 ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data);
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400220
221 tmp = ext4_inode_table(sb, gdp);
222 for (; tmp < ext4_inode_table(sb, gdp) +
223 sbi->s_itb_per_group; tmp++) {
Jose R. Santos0bf7e832008-06-03 14:07:29 -0400224 if (!flex_bg || ext4_block_in_group(sb, tmp, block_group))
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400225 ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data);
Andreas Dilger717d50e2007-10-16 18:38:25 -0400226 }
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400227
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400228 /*
229 * Also if the number of blocks within the group is less than
230 * the blocksize * 8 ( which is the size of bitmap ), set rest
231 * of the block bitmap to 1
232 */
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400233 ext4_mark_bitmap_end(num_clusters_in_group(sb, block_group),
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400234 sb->s_blocksize * 8, bh->b_data);
Tao Ma79f1ba42012-10-22 00:34:32 -0400235 ext4_block_bitmap_csum_set(sb, block_group, gdp, bh);
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -0400236 ext4_group_desc_csum_set(sb, block_group, gdp);
Andreas Dilger717d50e2007-10-16 18:38:25 -0400237}
238
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400239/* Return the number of free blocks in a block group. It is used when
240 * the block bitmap is uninitialized, so we can't just count the bits
241 * in the bitmap. */
Theodore Ts'ocff1dfd72011-09-09 19:12:51 -0400242unsigned ext4_free_clusters_after_init(struct super_block *sb,
243 ext4_group_t block_group,
244 struct ext4_group_desc *gdp)
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400245{
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400246 return num_clusters_in_group(sb, block_group) -
247 ext4_num_overhead_clusters(sb, block_group, gdp);
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400248}
Andreas Dilger717d50e2007-10-16 18:38:25 -0400249
Andrew Morton72b64b52006-10-11 01:21:18 -0700250/*
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700251 * The free blocks are managed by bitmaps. A file system contains several
252 * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap
253 * block for inodes, N blocks for the inode table and data blocks.
254 *
255 * The file system contains group descriptors which are located after the
256 * super block. Each descriptor contains the number of the bitmap block and
257 * the free blocks count in the block. The descriptors are loaded in memory
Aneesh Kumar K.Ve6274322007-02-20 13:57:58 -0800258 * when a file system is mounted (see ext4_fill_super).
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700259 */
260
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700261/**
Mingming Cao617ba132006-10-11 01:20:53 -0700262 * ext4_get_group_desc() -- load group descriptor from disk
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700263 * @sb: super block
264 * @block_group: given block group
265 * @bh: pointer to the buffer head to store the block
266 * group descriptor
267 */
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400268struct ext4_group_desc * ext4_get_group_desc(struct super_block *sb,
Avantika Mathurfd2d4292008-01-28 23:58:27 -0500269 ext4_group_t block_group,
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400270 struct buffer_head **bh)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700271{
Theodore Ts'o498e5f22008-11-05 00:14:04 -0500272 unsigned int group_desc;
273 unsigned int offset;
Theodore Ts'o8df96752009-05-01 08:50:38 -0400274 ext4_group_t ngroups = ext4_get_groups_count(sb);
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400275 struct ext4_group_desc *desc;
Mingming Cao617ba132006-10-11 01:20:53 -0700276 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700277
Theodore Ts'o8df96752009-05-01 08:50:38 -0400278 if (block_group >= ngroups) {
Eric Sandeen12062dd2010-02-15 14:19:27 -0500279 ext4_error(sb, "block_group >= groups_count - block_group = %u,"
280 " groups_count = %u", block_group, ngroups);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700281
282 return NULL;
283 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700284
Mingming Cao617ba132006-10-11 01:20:53 -0700285 group_desc = block_group >> EXT4_DESC_PER_BLOCK_BITS(sb);
286 offset = block_group & (EXT4_DESC_PER_BLOCK(sb) - 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700287 if (!sbi->s_group_desc[group_desc]) {
Eric Sandeen12062dd2010-02-15 14:19:27 -0500288 ext4_error(sb, "Group descriptor not loaded - "
Theodore Ts'o498e5f22008-11-05 00:14:04 -0500289 "block_group = %u, group_desc = %u, desc = %u",
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400290 block_group, group_desc, offset);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700291 return NULL;
292 }
293
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -0700294 desc = (struct ext4_group_desc *)(
295 (__u8 *)sbi->s_group_desc[group_desc]->b_data +
296 offset * EXT4_DESC_SIZE(sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700297 if (bh)
298 *bh = sbi->s_group_desc[group_desc];
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -0700299 return desc;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700300}
301
Theodore Ts'o7a4c5de2012-08-10 13:57:52 -0400302/*
303 * Return the block number which was discovered to be invalid, or 0 if
304 * the block bitmap is valid.
305 */
306static ext4_fsblk_t ext4_valid_block_bitmap(struct super_block *sb,
307 struct ext4_group_desc *desc,
308 unsigned int block_group,
309 struct buffer_head *bh)
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500310{
311 ext4_grpblk_t offset;
312 ext4_grpblk_t next_zero_bit;
Theodore Ts'o7a4c5de2012-08-10 13:57:52 -0400313 ext4_fsblk_t blk;
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500314 ext4_fsblk_t group_first_block;
315
316 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) {
317 /* with FLEX_BG, the inode/block bitmaps and itable
318 * blocks may not be in the group at all
319 * so the bitmap validation will be skipped for those groups
320 * or it has to also read the block group where the bitmaps
321 * are located to verify they are set.
322 */
Theodore Ts'o7a4c5de2012-08-10 13:57:52 -0400323 return 0;
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500324 }
325 group_first_block = ext4_group_first_block_no(sb, block_group);
326
327 /* check whether block bitmap block number is set */
Theodore Ts'o7a4c5de2012-08-10 13:57:52 -0400328 blk = ext4_block_bitmap(sb, desc);
329 offset = blk - group_first_block;
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500330 if (!ext4_test_bit(offset, bh->b_data))
331 /* bad block bitmap */
Theodore Ts'o7a4c5de2012-08-10 13:57:52 -0400332 return blk;
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500333
334 /* check whether the inode bitmap block number is set */
Theodore Ts'o7a4c5de2012-08-10 13:57:52 -0400335 blk = ext4_inode_bitmap(sb, desc);
336 offset = blk - group_first_block;
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500337 if (!ext4_test_bit(offset, bh->b_data))
338 /* bad block bitmap */
Theodore Ts'o7a4c5de2012-08-10 13:57:52 -0400339 return blk;
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500340
341 /* check whether the inode table block number is set */
Theodore Ts'o7a4c5de2012-08-10 13:57:52 -0400342 blk = ext4_inode_table(sb, desc);
343 offset = blk - group_first_block;
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500344 next_zero_bit = ext4_find_next_zero_bit(bh->b_data,
345 offset + EXT4_SB(sb)->s_itb_per_group,
346 offset);
Theodore Ts'o7a4c5de2012-08-10 13:57:52 -0400347 if (next_zero_bit < offset + EXT4_SB(sb)->s_itb_per_group)
348 /* bad bitmap for inode tables */
349 return blk;
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500350 return 0;
351}
Darrick J. Wongfa77dcf2012-04-29 18:35:10 -0400352
353void ext4_validate_block_bitmap(struct super_block *sb,
354 struct ext4_group_desc *desc,
355 unsigned int block_group,
356 struct buffer_head *bh)
357{
Theodore Ts'o7a4c5de2012-08-10 13:57:52 -0400358 ext4_fsblk_t blk;
359
Darrick J. Wongfa77dcf2012-04-29 18:35:10 -0400360 if (buffer_verified(bh))
361 return;
362
363 ext4_lock_group(sb, block_group);
Theodore Ts'o7a4c5de2012-08-10 13:57:52 -0400364 blk = ext4_valid_block_bitmap(sb, desc, block_group, bh);
365 if (unlikely(blk != 0)) {
366 ext4_unlock_group(sb, block_group);
367 ext4_error(sb, "bg %u: block %llu: invalid block bitmap",
368 block_group, blk);
369 return;
370 }
371 if (unlikely(!ext4_block_bitmap_csum_verify(sb, block_group,
Tao Ma79f1ba42012-10-22 00:34:32 -0400372 desc, bh))) {
Theodore Ts'o7a4c5de2012-08-10 13:57:52 -0400373 ext4_unlock_group(sb, block_group);
374 ext4_error(sb, "bg %u: bad block bitmap checksum", block_group);
375 return;
376 }
377 set_buffer_verified(bh);
Darrick J. Wongfa77dcf2012-04-29 18:35:10 -0400378 ext4_unlock_group(sb, block_group);
379}
380
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700381/**
Eryu Guan15b49132013-01-12 16:33:25 -0500382 * ext4_read_block_bitmap_nowait()
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700383 * @sb: super block
384 * @block_group: given block group
385 *
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500386 * Read the bitmap for a given block_group,and validate the
387 * bits for block/inode/inode tables are set in the bitmaps
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700388 *
389 * Return buffer_head on success or NULL in case of failure.
390 */
Andreas Dilger717d50e2007-10-16 18:38:25 -0400391struct buffer_head *
Theodore Ts'o813e5722012-02-20 17:52:46 -0500392ext4_read_block_bitmap_nowait(struct super_block *sb, ext4_group_t block_group)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700393{
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400394 struct ext4_group_desc *desc;
Theodore Ts'o813e5722012-02-20 17:52:46 -0500395 struct buffer_head *bh;
Aneesh Kumar K.V7c9e69f2007-10-16 23:27:02 -0700396 ext4_fsblk_t bitmap_blk;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700397
Andreas Dilger717d50e2007-10-16 18:38:25 -0400398 desc = ext4_get_group_desc(sb, block_group, NULL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700399 if (!desc)
Aneesh Kumar K.V7c9e69f2007-10-16 23:27:02 -0700400 return NULL;
401 bitmap_blk = ext4_block_bitmap(sb, desc);
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500402 bh = sb_getblk(sb, bitmap_blk);
403 if (unlikely(!bh)) {
Theodore Ts'o813e5722012-02-20 17:52:46 -0500404 ext4_error(sb, "Cannot get buffer for block bitmap - "
405 "block_group = %u, block_bitmap = %llu",
406 block_group, bitmap_blk);
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500407 return NULL;
408 }
Aneesh Kumar K.V2ccb5fb2009-01-05 21:49:55 -0500409
410 if (bitmap_uptodate(bh))
Darrick J. Wongfa77dcf2012-04-29 18:35:10 -0400411 goto verify;
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500412
Frederic Bohec806e682008-10-10 08:09:18 -0400413 lock_buffer(bh);
Aneesh Kumar K.V2ccb5fb2009-01-05 21:49:55 -0500414 if (bitmap_uptodate(bh)) {
415 unlock_buffer(bh);
Darrick J. Wongfa77dcf2012-04-29 18:35:10 -0400416 goto verify;
Aneesh Kumar K.V2ccb5fb2009-01-05 21:49:55 -0500417 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -0400418 ext4_lock_group(sb, block_group);
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500419 if (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) {
420 ext4_init_block_bitmap(sb, bh, block_group, desc);
Aneesh Kumar K.V2ccb5fb2009-01-05 21:49:55 -0500421 set_bitmap_uptodate(bh);
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500422 set_buffer_uptodate(bh);
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -0400423 ext4_unlock_group(sb, block_group);
Aneesh Kumar K.V3300bed2009-01-03 22:33:39 -0500424 unlock_buffer(bh);
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500425 return bh;
426 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -0400427 ext4_unlock_group(sb, block_group);
Aneesh Kumar K.V2ccb5fb2009-01-05 21:49:55 -0500428 if (buffer_uptodate(bh)) {
429 /*
430 * if not uninit if bh is uptodate,
431 * bitmap is also uptodate
432 */
433 set_bitmap_uptodate(bh);
434 unlock_buffer(bh);
Darrick J. Wongfa77dcf2012-04-29 18:35:10 -0400435 goto verify;
Aneesh Kumar K.V2ccb5fb2009-01-05 21:49:55 -0500436 }
437 /*
Theodore Ts'o813e5722012-02-20 17:52:46 -0500438 * submit the buffer_head for reading
Aneesh Kumar K.V2ccb5fb2009-01-05 21:49:55 -0500439 */
Theodore Ts'o813e5722012-02-20 17:52:46 -0500440 set_buffer_new(bh);
Jiaying Zhang0562e0b2011-03-21 21:38:05 -0400441 trace_ext4_read_block_bitmap_load(sb, block_group);
Theodore Ts'o813e5722012-02-20 17:52:46 -0500442 bh->b_end_io = ext4_end_bitmap_read;
443 get_bh(bh);
Theodore Ts'o9f203502013-04-20 15:46:17 -0400444 submit_bh(READ | REQ_META | REQ_PRIO, bh);
Theodore Ts'o813e5722012-02-20 17:52:46 -0500445 return bh;
Darrick J. Wongfa77dcf2012-04-29 18:35:10 -0400446verify:
447 ext4_validate_block_bitmap(sb, desc, block_group, bh);
Darrick J. Wong48d9eb92013-08-28 15:35:27 -0400448 if (buffer_verified(bh))
449 return bh;
450 put_bh(bh);
451 return NULL;
Theodore Ts'o813e5722012-02-20 17:52:46 -0500452}
453
454/* Returns 0 on success, 1 on error */
455int ext4_wait_block_bitmap(struct super_block *sb, ext4_group_t block_group,
456 struct buffer_head *bh)
457{
458 struct ext4_group_desc *desc;
459
460 if (!buffer_new(bh))
461 return 0;
462 desc = ext4_get_group_desc(sb, block_group, NULL);
463 if (!desc)
464 return 1;
465 wait_on_buffer(bh);
466 if (!buffer_uptodate(bh)) {
Eric Sandeen12062dd2010-02-15 14:19:27 -0500467 ext4_error(sb, "Cannot read block bitmap - "
Theodore Ts'o813e5722012-02-20 17:52:46 -0500468 "block_group = %u, block_bitmap = %llu",
Heiko Carstensd4dc4622012-02-20 17:57:24 -0500469 block_group, (unsigned long long) bh->b_blocknr);
Theodore Ts'o813e5722012-02-20 17:52:46 -0500470 return 1;
471 }
472 clear_buffer_new(bh);
473 /* Panic or remount fs read-only if block bitmap is invalid */
Darrick J. Wongfa77dcf2012-04-29 18:35:10 -0400474 ext4_validate_block_bitmap(sb, desc, block_group, bh);
Darrick J. Wong48d9eb92013-08-28 15:35:27 -0400475 /* ...but check for error just in case errors=continue. */
476 return !buffer_verified(bh);
Theodore Ts'o813e5722012-02-20 17:52:46 -0500477}
478
479struct buffer_head *
480ext4_read_block_bitmap(struct super_block *sb, ext4_group_t block_group)
481{
482 struct buffer_head *bh;
483
484 bh = ext4_read_block_bitmap_nowait(sb, block_group);
Eryu Guan15b49132013-01-12 16:33:25 -0500485 if (!bh)
486 return NULL;
Theodore Ts'o813e5722012-02-20 17:52:46 -0500487 if (ext4_wait_block_bitmap(sb, block_group, bh)) {
488 put_bh(bh);
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500489 return NULL;
490 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700491 return bh;
492}
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700493
494/**
Theodore Ts'odf55c992011-09-09 19:16:51 -0400495 * ext4_has_free_clusters()
Eric Sandeen8c3bf8a02008-10-28 00:08:12 -0400496 * @sbi: in-core super block structure.
Theodore Ts'odf55c992011-09-09 19:16:51 -0400497 * @nclusters: number of needed blocks
498 * @flags: flags from ext4_mb_new_blocks()
Eric Sandeen8c3bf8a02008-10-28 00:08:12 -0400499 *
Theodore Ts'odf55c992011-09-09 19:16:51 -0400500 * Check if filesystem has nclusters free & available for allocation.
Eric Sandeen8c3bf8a02008-10-28 00:08:12 -0400501 * On success return 1, return 0 on failure.
502 */
Theodore Ts'odf55c992011-09-09 19:16:51 -0400503static int ext4_has_free_clusters(struct ext4_sb_info *sbi,
504 s64 nclusters, unsigned int flags)
Aneesh Kumar K.Va30d542a2008-10-09 10:56:23 -0400505{
Lukas Czerner27dd4382013-04-09 22:11:22 -0400506 s64 free_clusters, dirty_clusters, rsv, resv_clusters;
Theodore Ts'o57042652011-09-09 18:56:51 -0400507 struct percpu_counter *fcc = &sbi->s_freeclusters_counter;
Theodore Ts'odf55c992011-09-09 19:16:51 -0400508 struct percpu_counter *dcc = &sbi->s_dirtyclusters_counter;
Aneesh Kumar K.Va30d542a2008-10-09 10:56:23 -0400509
Theodore Ts'odf55c992011-09-09 19:16:51 -0400510 free_clusters = percpu_counter_read_positive(fcc);
511 dirty_clusters = percpu_counter_read_positive(dcc);
Lukas Czerner27dd4382013-04-09 22:11:22 -0400512 resv_clusters = atomic64_read(&sbi->s_resv_clusters);
Lukas Czerner304e2202013-02-22 15:27:52 -0500513
514 /*
515 * r_blocks_count should always be multiple of the cluster ratio so
516 * we are safe to do a plane bit shift only.
517 */
Lukas Czerner27dd4382013-04-09 22:11:22 -0400518 rsv = (ext4_r_blocks_count(sbi->s_es) >> sbi->s_cluster_bits) +
519 resv_clusters;
Aneesh Kumar K.Va30d542a2008-10-09 10:56:23 -0400520
Lukas Czerner27dd4382013-04-09 22:11:22 -0400521 if (free_clusters - (nclusters + rsv + dirty_clusters) <
Theodore Ts'odf55c992011-09-09 19:16:51 -0400522 EXT4_FREECLUSTERS_WATERMARK) {
Lukas Czerner304e2202013-02-22 15:27:52 -0500523 free_clusters = percpu_counter_sum_positive(fcc);
Theodore Ts'odf55c992011-09-09 19:16:51 -0400524 dirty_clusters = percpu_counter_sum_positive(dcc);
Aneesh Kumar K.V6bc6e632008-10-10 09:39:00 -0400525 }
Theodore Ts'odf55c992011-09-09 19:16:51 -0400526 /* Check whether we have space after accounting for current
527 * dirty clusters & root reserved clusters.
Aneesh Kumar K.V6bc6e632008-10-10 09:39:00 -0400528 */
Lukas Czerner27dd4382013-04-09 22:11:22 -0400529 if (free_clusters >= (rsv + nclusters + dirty_clusters))
Eric Sandeena9960312008-10-28 00:08:17 -0400530 return 1;
Aneesh Kumar K.V5c791612008-10-08 23:12:24 -0400531
Theodore Ts'odf55c992011-09-09 19:16:51 -0400532 /* Hm, nope. Are (enough) root reserved clusters available? */
Eric W. Biederman08cefc72012-02-07 15:41:49 -0800533 if (uid_eq(sbi->s_resuid, current_fsuid()) ||
534 (!gid_eq(sbi->s_resgid, GLOBAL_ROOT_GID) && in_group_p(sbi->s_resgid)) ||
Allison Henderson55f020d2011-05-25 07:41:26 -0400535 capable(CAP_SYS_RESOURCE) ||
Lukas Czerner27dd4382013-04-09 22:11:22 -0400536 (flags & EXT4_MB_USE_ROOT_BLOCKS)) {
Allison Henderson55f020d2011-05-25 07:41:26 -0400537
Lukas Czerner27dd4382013-04-09 22:11:22 -0400538 if (free_clusters >= (nclusters + dirty_clusters +
539 resv_clusters))
540 return 1;
541 }
542 /* No free blocks. Let's see if we can dip into reserved pool */
543 if (flags & EXT4_MB_USE_RESERVED) {
Theodore Ts'odf55c992011-09-09 19:16:51 -0400544 if (free_clusters >= (nclusters + dirty_clusters))
Eric Sandeena9960312008-10-28 00:08:17 -0400545 return 1;
546 }
547
548 return 0;
Aneesh Kumar K.Va30d542a2008-10-09 10:56:23 -0400549}
Mingming Cao07031432008-07-11 19:27:31 -0400550
Theodore Ts'oe7d5f312011-09-09 19:14:51 -0400551int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
552 s64 nclusters, unsigned int flags)
Eric Sandeen8c3bf8a02008-10-28 00:08:12 -0400553{
Theodore Ts'odf55c992011-09-09 19:16:51 -0400554 if (ext4_has_free_clusters(sbi, nclusters, flags)) {
Theodore Ts'oe7d5f312011-09-09 19:14:51 -0400555 percpu_counter_add(&sbi->s_dirtyclusters_counter, nclusters);
Eric Sandeen8c3bf8a02008-10-28 00:08:12 -0400556 return 0;
557 } else
558 return -ENOSPC;
559}
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700560
561/**
Mingming Cao617ba132006-10-11 01:20:53 -0700562 * ext4_should_retry_alloc()
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700563 * @sb: super block
564 * @retries number of attemps has been made
565 *
Mingming Cao617ba132006-10-11 01:20:53 -0700566 * ext4_should_retry_alloc() is called when ENOSPC is returned, and if
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700567 * it is profitable to retry the operation, this function will wait
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300568 * for the current or committing transaction to complete, and then
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700569 * return TRUE.
570 *
571 * if the total number of retries exceed three times, return FALSE.
572 */
Mingming Cao617ba132006-10-11 01:20:53 -0700573int ext4_should_retry_alloc(struct super_block *sb, int *retries)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700574{
Theodore Ts'odf55c992011-09-09 19:16:51 -0400575 if (!ext4_has_free_clusters(EXT4_SB(sb), 1, 0) ||
Eric Sandeen8f64b322009-02-26 00:57:35 -0500576 (*retries)++ > 3 ||
577 !EXT4_SB(sb)->s_journal)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700578 return 0;
579
580 jbd_debug(1, "%s: retrying operation after ENOSPC\n", sb->s_id);
581
Mingming Caodab291a2006-10-11 01:21:01 -0700582 return jbd2_journal_force_commit_nested(EXT4_SB(sb)->s_journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700583}
584
Aneesh Kumar K.V654b4902008-07-11 19:27:31 -0400585/*
Aneesh Kumar K.V654b4902008-07-11 19:27:31 -0400586 * ext4_new_meta_blocks() -- allocate block for meta data (indexing) blocks
587 *
588 * @handle: handle to this transaction
589 * @inode: file inode
590 * @goal: given target block(filesystem wide)
Aditya Kali7b415bf2011-09-09 19:04:51 -0400591 * @count: pointer to total number of clusters needed
Aneesh Kumar K.V654b4902008-07-11 19:27:31 -0400592 * @errp: error code
593 *
Theodore Ts'o97df5d12008-12-12 12:41:28 -0500594 * Return 1st allocated block number on success, *count stores total account
Aneesh Kumar K.V654b4902008-07-11 19:27:31 -0400595 * error stores in errp pointer
596 */
597ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
Allison Henderson55f020d2011-05-25 07:41:26 -0400598 ext4_fsblk_t goal, unsigned int flags,
599 unsigned long *count, int *errp)
Aneesh Kumar K.V654b4902008-07-11 19:27:31 -0400600{
Theodore Ts'o97df5d12008-12-12 12:41:28 -0500601 struct ext4_allocation_request ar;
Mingming Caod2a17632008-07-14 17:52:37 -0400602 ext4_fsblk_t ret;
Theodore Ts'o97df5d12008-12-12 12:41:28 -0500603
604 memset(&ar, 0, sizeof(ar));
605 /* Fill with neighbour allocated blocks */
606 ar.inode = inode;
607 ar.goal = goal;
608 ar.len = count ? *count : 1;
Allison Henderson55f020d2011-05-25 07:41:26 -0400609 ar.flags = flags;
Theodore Ts'o97df5d12008-12-12 12:41:28 -0500610
611 ret = ext4_mb_new_blocks(handle, &ar, errp);
612 if (count)
613 *count = ar.len;
Mingming Caod2a17632008-07-14 17:52:37 -0400614 /*
Eric Sandeen72b8ab92010-05-16 11:00:00 -0400615 * Account for the allocated meta blocks. We will never
616 * fail EDQUOT for metdata, but we do account for it.
Mingming Caod2a17632008-07-14 17:52:37 -0400617 */
Theodore Ts'of2321092011-01-10 12:12:36 -0500618 if (!(*errp) &&
619 ext4_test_inode_state(inode, EXT4_STATE_DELALLOC_RESERVED)) {
Mingming Caod2a17632008-07-14 17:52:37 -0400620 spin_lock(&EXT4_I(inode)->i_block_reservation_lock);
Theodore Ts'o97df5d12008-12-12 12:41:28 -0500621 EXT4_I(inode)->i_allocated_meta_blocks += ar.len;
Mingming Caod2a17632008-07-14 17:52:37 -0400622 spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
Aditya Kali7b415bf2011-09-09 19:04:51 -0400623 dquot_alloc_block_nofail(inode,
624 EXT4_C2B(EXT4_SB(inode->i_sb), ar.len));
Mingming Caod2a17632008-07-14 17:52:37 -0400625 }
626 return ret;
627}
628
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700629/**
Theodore Ts'o5dee5432011-09-09 19:10:51 -0400630 * ext4_count_free_clusters() -- count filesystem free clusters
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700631 * @sb: superblock
632 *
Theodore Ts'o5dee5432011-09-09 19:10:51 -0400633 * Adds up the number of free clusters from each block group.
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700634 */
Theodore Ts'o5dee5432011-09-09 19:10:51 -0400635ext4_fsblk_t ext4_count_free_clusters(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700636{
Mingming Cao617ba132006-10-11 01:20:53 -0700637 ext4_fsblk_t desc_count;
638 struct ext4_group_desc *gdp;
Avantika Mathurfd2d4292008-01-28 23:58:27 -0500639 ext4_group_t i;
Theodore Ts'o8df96752009-05-01 08:50:38 -0400640 ext4_group_t ngroups = ext4_get_groups_count(sb);
Mingming Cao617ba132006-10-11 01:20:53 -0700641#ifdef EXT4FS_DEBUG
642 struct ext4_super_block *es;
643 ext4_fsblk_t bitmap_count;
Theodore Ts'o498e5f22008-11-05 00:14:04 -0500644 unsigned int x;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700645 struct buffer_head *bitmap_bh = NULL;
646
Mingming Cao617ba132006-10-11 01:20:53 -0700647 es = EXT4_SB(sb)->s_es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700648 desc_count = 0;
649 bitmap_count = 0;
650 gdp = NULL;
651
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700652 for (i = 0; i < ngroups; i++) {
Mingming Cao617ba132006-10-11 01:20:53 -0700653 gdp = ext4_get_group_desc(sb, i, NULL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700654 if (!gdp)
655 continue;
Theodore Ts'o021b65b2011-09-09 19:08:51 -0400656 desc_count += ext4_free_group_clusters(sb, gdp);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700657 brelse(bitmap_bh);
Theodore Ts'o574ca172008-07-11 19:27:31 -0400658 bitmap_bh = ext4_read_block_bitmap(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700659 if (bitmap_bh == NULL)
660 continue;
661
Theodore Ts'of6fb99c2012-06-30 19:14:57 -0400662 x = ext4_count_free(bitmap_bh->b_data,
663 EXT4_BLOCKS_PER_GROUP(sb) / 8);
Thadeu Lima de Souza Cascardo9fd97842009-01-26 19:26:26 -0500664 printk(KERN_DEBUG "group %u: stored = %d, counted = %u\n",
Theodore Ts'o021b65b2011-09-09 19:08:51 -0400665 i, ext4_free_group_clusters(sb, gdp), x);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700666 bitmap_count += x;
667 }
668 brelse(bitmap_bh);
Theodore Ts'o5dee5432011-09-09 19:10:51 -0400669 printk(KERN_DEBUG "ext4_count_free_clusters: stored = %llu"
670 ", computed = %llu, %llu\n",
Lukas Czerner810da242013-03-02 17:18:58 -0500671 EXT4_NUM_B2C(EXT4_SB(sb), ext4_free_blocks_count(es)),
Theodore Ts'o4776004f2008-09-08 23:00:52 -0400672 desc_count, bitmap_count);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700673 return bitmap_count;
674#else
675 desc_count = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700676 for (i = 0; i < ngroups; i++) {
Mingming Cao617ba132006-10-11 01:20:53 -0700677 gdp = ext4_get_group_desc(sb, i, NULL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700678 if (!gdp)
679 continue;
Theodore Ts'o021b65b2011-09-09 19:08:51 -0400680 desc_count += ext4_free_group_clusters(sb, gdp);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700681 }
682
683 return desc_count;
684#endif
685}
686
Avantika Mathurfd2d4292008-01-28 23:58:27 -0500687static inline int test_root(ext4_group_t a, int b)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700688{
Theodore Ts'of4afb4f2013-06-06 11:40:37 -0400689 while (1) {
690 if (a < b)
691 return 0;
692 if (a == b)
693 return 1;
694 if ((a % b) != 0)
695 return 0;
696 a = a / b;
697 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700698}
699
Avantika Mathurfd2d4292008-01-28 23:58:27 -0500700static int ext4_group_sparse(ext4_group_t group)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700701{
702 if (group <= 1)
703 return 1;
704 if (!(group & 1))
705 return 0;
706 return (test_root(group, 7) || test_root(group, 5) ||
707 test_root(group, 3));
708}
709
710/**
Mingming Cao617ba132006-10-11 01:20:53 -0700711 * ext4_bg_has_super - number of blocks used by the superblock in group
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700712 * @sb: superblock for filesystem
713 * @group: group number to check
714 *
715 * Return the number of blocks used by the superblock (primary or backup)
716 * in this group. Currently this will be only 0 or 1.
717 */
Avantika Mathurfd2d4292008-01-28 23:58:27 -0500718int ext4_bg_has_super(struct super_block *sb, ext4_group_t group)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700719{
Mingming Cao617ba132006-10-11 01:20:53 -0700720 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
721 EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER) &&
722 !ext4_group_sparse(group))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700723 return 0;
724 return 1;
725}
726
Avantika Mathurfd2d4292008-01-28 23:58:27 -0500727static unsigned long ext4_bg_num_gdb_meta(struct super_block *sb,
728 ext4_group_t group)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700729{
Mingming Cao617ba132006-10-11 01:20:53 -0700730 unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
Avantika Mathurfd2d4292008-01-28 23:58:27 -0500731 ext4_group_t first = metagroup * EXT4_DESC_PER_BLOCK(sb);
732 ext4_group_t last = first + EXT4_DESC_PER_BLOCK(sb) - 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700733
734 if (group == first || group == first + 1 || group == last)
735 return 1;
736 return 0;
737}
738
Avantika Mathurfd2d4292008-01-28 23:58:27 -0500739static unsigned long ext4_bg_num_gdb_nometa(struct super_block *sb,
740 ext4_group_t group)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700741{
Theodore Ts'o8dadb192009-11-23 07:24:38 -0500742 if (!ext4_bg_has_super(sb, group))
743 return 0;
744
745 if (EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG))
746 return le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg);
747 else
748 return EXT4_SB(sb)->s_gdb_count;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700749}
750
751/**
Mingming Cao617ba132006-10-11 01:20:53 -0700752 * ext4_bg_num_gdb - number of blocks used by the group table in group
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700753 * @sb: superblock for filesystem
754 * @group: group number to check
755 *
756 * Return the number of blocks used by the group descriptor table
757 * (primary or backup) in this group. In the future there may be a
758 * different number of descriptor blocks in each group.
759 */
Avantika Mathurfd2d4292008-01-28 23:58:27 -0500760unsigned long ext4_bg_num_gdb(struct super_block *sb, ext4_group_t group)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700761{
762 unsigned long first_meta_bg =
Mingming Cao617ba132006-10-11 01:20:53 -0700763 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg);
764 unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700765
Mingming Cao617ba132006-10-11 01:20:53 -0700766 if (!EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700767 metagroup < first_meta_bg)
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400768 return ext4_bg_num_gdb_nometa(sb, group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700769
Mingming Cao617ba132006-10-11 01:20:53 -0700770 return ext4_bg_num_gdb_meta(sb,group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700771
772}
Theodore Ts'oc2ea3fd2008-10-10 09:40:52 -0400773
Theodore Ts'o49f7f9a2011-09-09 18:40:51 -0400774/*
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400775 * This function returns the number of file system metadata clusters at
Theodore Ts'o49f7f9a2011-09-09 18:40:51 -0400776 * the beginning of a block group, including the reserved gdt blocks.
777 */
Eric Sandeen5f163cc2012-01-04 22:33:28 -0500778static unsigned ext4_num_base_meta_clusters(struct super_block *sb,
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400779 ext4_group_t block_group)
Theodore Ts'o49f7f9a2011-09-09 18:40:51 -0400780{
781 struct ext4_sb_info *sbi = EXT4_SB(sb);
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400782 unsigned num;
Theodore Ts'o49f7f9a2011-09-09 18:40:51 -0400783
784 /* Check for superblock and gdt backups in this group */
785 num = ext4_bg_has_super(sb, block_group);
786
787 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
788 block_group < le32_to_cpu(sbi->s_es->s_first_meta_bg) *
789 sbi->s_desc_per_block) {
790 if (num) {
791 num += ext4_bg_num_gdb(sb, block_group);
792 num += le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks);
793 }
794 } else { /* For META_BG_BLOCK_GROUPS */
795 num += ext4_bg_num_gdb(sb, block_group);
796 }
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400797 return EXT4_NUM_B2C(sbi, num);
Theodore Ts'o49f7f9a2011-09-09 18:40:51 -0400798}
Eric Sandeenf86186b2011-06-28 10:01:31 -0400799/**
800 * ext4_inode_to_goal_block - return a hint for block allocation
801 * @inode: inode for block allocation
802 *
803 * Return the ideal location to start allocating blocks for a
804 * newly created inode.
805 */
806ext4_fsblk_t ext4_inode_to_goal_block(struct inode *inode)
807{
808 struct ext4_inode_info *ei = EXT4_I(inode);
809 ext4_group_t block_group;
810 ext4_grpblk_t colour;
811 int flex_size = ext4_flex_bg_size(EXT4_SB(inode->i_sb));
812 ext4_fsblk_t bg_start;
813 ext4_fsblk_t last_block;
814
815 block_group = ei->i_block_group;
816 if (flex_size >= EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME) {
817 /*
818 * If there are at least EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME
819 * block groups per flexgroup, reserve the first block
820 * group for directories and special files. Regular
821 * files will start at the second block group. This
822 * tends to speed up directory access and improves
823 * fsck times.
824 */
825 block_group &= ~(flex_size-1);
826 if (S_ISREG(inode->i_mode))
827 block_group++;
828 }
829 bg_start = ext4_group_first_block_no(inode->i_sb, block_group);
830 last_block = ext4_blocks_count(EXT4_SB(inode->i_sb)->s_es) - 1;
831
832 /*
833 * If we are doing delayed allocation, we don't need take
834 * colour into account.
835 */
836 if (test_opt(inode->i_sb, DELALLOC))
837 return bg_start;
838
839 if (bg_start + EXT4_BLOCKS_PER_GROUP(inode->i_sb) <= last_block)
840 colour = (current->pid % 16) *
841 (EXT4_BLOCKS_PER_GROUP(inode->i_sb) / 16);
842 else
843 colour = (current->pid % 16) * ((last_block - bg_start) / 16);
844 return bg_start + colour;
845}
846