blob: de601f3c023d8b01fcc06f3ca5a209fe9d325e06 [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>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070017#include <linux/quotaops.h>
18#include <linux/buffer_head.h>
Christoph Hellwig3dcf5452008-04-29 18:13:32 -040019#include "ext4.h"
20#include "ext4_jbd2.h"
Aneesh Kumar K.Ve21675d2009-01-05 21:36:02 -050021#include "mballoc.h"
Christoph Hellwig3dcf5452008-04-29 18:13:32 -040022
Jiaying Zhang0562e0b2011-03-21 21:38:05 -040023#include <trace/events/ext4.h>
24
Eric Sandeen5f163cc2012-01-04 22:33:28 -050025static unsigned ext4_num_base_meta_clusters(struct super_block *sb,
26 ext4_group_t block_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -070027/*
28 * balloc.c contains the blocks allocation and deallocation routines
29 */
30
31/*
Lukas Czernerbd862982013-04-03 23:32:34 -040032 * Calculate block group number for a given block number
33 */
34ext4_group_t ext4_get_group_number(struct super_block *sb,
35 ext4_fsblk_t block)
36{
37 ext4_group_t group;
38
39 if (test_opt2(sb, STD_GROUP_SIZE))
Theodore Ts'o960fd8562013-07-05 23:11:16 -040040 group = (block -
41 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block)) >>
Lukas Czernerbd862982013-04-03 23:32:34 -040042 (EXT4_BLOCK_SIZE_BITS(sb) + EXT4_CLUSTER_BITS(sb) + 3);
43 else
44 ext4_get_group_no_and_offset(sb, block, &group, NULL);
45 return group;
46}
47
48/*
Theodore Ts'o3212a802011-09-09 18:46:51 -040049 * Calculate the block group number and offset into the block/cluster
50 * allocation bitmap, given a block number
Andrew Morton72b64b52006-10-11 01:21:18 -070051 */
52void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
Avantika Mathurfd2d4292008-01-28 23:58:27 -050053 ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp)
Andrew Morton72b64b52006-10-11 01:21:18 -070054{
Dave Kleikamp8c55e202007-05-24 13:04:54 -040055 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Andrew Morton72b64b52006-10-11 01:21:18 -070056 ext4_grpblk_t offset;
57
Dave Kleikamp8c55e202007-05-24 13:04:54 -040058 blocknr = blocknr - le32_to_cpu(es->s_first_data_block);
Theodore Ts'o3212a802011-09-09 18:46:51 -040059 offset = do_div(blocknr, EXT4_BLOCKS_PER_GROUP(sb)) >>
60 EXT4_SB(sb)->s_cluster_bits;
Andrew Morton72b64b52006-10-11 01:21:18 -070061 if (offsetp)
62 *offsetp = offset;
63 if (blockgrpp)
Dave Kleikamp8c55e202007-05-24 13:04:54 -040064 *blockgrpp = blocknr;
Andrew Morton72b64b52006-10-11 01:21:18 -070065
66}
67
Lukas Czerner68911002013-04-03 22:12:52 -040068/*
69 * Check whether the 'block' lives within the 'block_group'. Returns 1 if so
70 * and 0 otherwise.
71 */
72static inline int ext4_block_in_group(struct super_block *sb,
73 ext4_fsblk_t block,
74 ext4_group_t block_group)
Jose R. Santos0bf7e832008-06-03 14:07:29 -040075{
76 ext4_group_t actual_group;
Lukas Czerner68911002013-04-03 22:12:52 -040077
Lukas Czernerbd862982013-04-03 23:32:34 -040078 actual_group = ext4_get_group_number(sb, block);
Lukas Czerner68911002013-04-03 22:12:52 -040079 return (actual_group == block_group) ? 1 : 0;
Jose R. Santos0bf7e832008-06-03 14:07:29 -040080}
81
Theodore Ts'od5b8f312011-09-09 18:44:51 -040082/* Return the number of clusters used for file system metadata; this
83 * represents the overhead needed by the file system.
84 */
Stephen Hemmingerc1978552014-05-12 10:50:23 -040085static unsigned ext4_num_overhead_clusters(struct super_block *sb,
86 ext4_group_t block_group,
87 struct ext4_group_desc *gdp)
Jose R. Santos0bf7e832008-06-03 14:07:29 -040088{
Theodore Ts'od5b8f312011-09-09 18:44:51 -040089 unsigned num_clusters;
90 int block_cluster = -1, inode_cluster = -1, itbl_cluster = -1, i, c;
91 ext4_fsblk_t start = ext4_group_first_block_no(sb, block_group);
92 ext4_fsblk_t itbl_blk;
Jose R. Santos0bf7e832008-06-03 14:07:29 -040093 struct ext4_sb_info *sbi = EXT4_SB(sb);
Jose R. Santos0bf7e832008-06-03 14:07:29 -040094
Theodore Ts'od5b8f312011-09-09 18:44:51 -040095 /* This is the number of clusters used by the superblock,
96 * block group descriptors, and reserved block group
97 * descriptor blocks */
98 num_clusters = ext4_num_base_meta_clusters(sb, block_group);
Jose R. Santos0bf7e832008-06-03 14:07:29 -040099
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400100 /*
101 * For the allocation bitmaps and inode table, we first need
102 * to check to see if the block is in the block group. If it
103 * is, then check to see if the cluster is already accounted
104 * for in the clusters used for the base metadata cluster, or
105 * if we can increment the base metadata cluster to include
106 * that block. Otherwise, we will have to track the cluster
107 * used for the allocation bitmap or inode table explicitly.
108 * Normally all of these blocks are contiguous, so the special
109 * case handling shouldn't be necessary except for *very*
110 * unusual file system layouts.
111 */
112 if (ext4_block_in_group(sb, ext4_block_bitmap(sb, gdp), block_group)) {
Theodore Ts'ob0dd6b72012-06-07 18:56:06 -0400113 block_cluster = EXT4_B2C(sbi,
114 ext4_block_bitmap(sb, gdp) - start);
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400115 if (block_cluster < num_clusters)
116 block_cluster = -1;
117 else if (block_cluster == num_clusters) {
118 num_clusters++;
119 block_cluster = -1;
Jose R. Santos0bf7e832008-06-03 14:07:29 -0400120 }
121 }
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400122
123 if (ext4_block_in_group(sb, ext4_inode_bitmap(sb, gdp), block_group)) {
124 inode_cluster = EXT4_B2C(sbi,
Theodore Ts'ob0dd6b72012-06-07 18:56:06 -0400125 ext4_inode_bitmap(sb, gdp) - start);
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400126 if (inode_cluster < num_clusters)
127 inode_cluster = -1;
128 else if (inode_cluster == num_clusters) {
129 num_clusters++;
130 inode_cluster = -1;
131 }
132 }
133
134 itbl_blk = ext4_inode_table(sb, gdp);
135 for (i = 0; i < sbi->s_itb_per_group; i++) {
136 if (ext4_block_in_group(sb, itbl_blk + i, block_group)) {
Theodore Ts'ob0dd6b72012-06-07 18:56:06 -0400137 c = EXT4_B2C(sbi, itbl_blk + i - start);
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400138 if ((c < num_clusters) || (c == inode_cluster) ||
139 (c == block_cluster) || (c == itbl_cluster))
140 continue;
141 if (c == num_clusters) {
142 num_clusters++;
143 continue;
144 }
145 num_clusters++;
146 itbl_cluster = c;
147 }
148 }
149
150 if (block_cluster != -1)
151 num_clusters++;
152 if (inode_cluster != -1)
153 num_clusters++;
154
155 return num_clusters;
Jose R. Santos0bf7e832008-06-03 14:07:29 -0400156}
Theodore Ts'oc2ea3fd2008-10-10 09:40:52 -0400157
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400158static unsigned int num_clusters_in_group(struct super_block *sb,
159 ext4_group_t block_group)
Theodore Ts'o49f7f9a2011-09-09 18:40:51 -0400160{
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400161 unsigned int blocks;
162
Theodore Ts'o49f7f9a2011-09-09 18:40:51 -0400163 if (block_group == ext4_get_groups_count(sb) - 1) {
164 /*
165 * Even though mke2fs always initializes the first and
166 * last group, just in case some other tool was used,
167 * we need to make sure we calculate the right free
168 * blocks.
169 */
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400170 blocks = ext4_blocks_count(EXT4_SB(sb)->s_es) -
Theodore Ts'o49f7f9a2011-09-09 18:40:51 -0400171 ext4_group_first_block_no(sb, block_group);
172 } else
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400173 blocks = EXT4_BLOCKS_PER_GROUP(sb);
174 return EXT4_NUM_B2C(EXT4_SB(sb), blocks);
Theodore Ts'o49f7f9a2011-09-09 18:40:51 -0400175}
176
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400177/* Initializes an uninitialized block bitmap */
Dmitry Monakhovaef48852014-10-13 03:42:12 -0400178static int ext4_init_block_bitmap(struct super_block *sb,
Stephen Hemmingerc1978552014-05-12 10:50:23 -0400179 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;
Darrick J. Wongbdfb6ff2013-08-28 18:46:56 -0400186 struct ext4_group_info *grp;
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)) {
Darrick J. Wongbdfb6ff2013-08-28 18:46:56 -0400193 grp = ext4_get_group_info(sb, block_group);
Namjae Jeone43bb4e2014-06-26 10:11:53 -0400194 if (!EXT4_MB_GRP_BBITMAP_CORRUPT(grp))
195 percpu_counter_sub(&sbi->s_freeclusters_counter,
196 grp->bb_free);
Darrick J. Wongbdfb6ff2013-08-28 18:46:56 -0400197 set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, &grp->bb_state);
Namjae Jeone43bb4e2014-06-26 10:11:53 -0400198 if (!EXT4_MB_GRP_IBITMAP_CORRUPT(grp)) {
199 int count;
200 count = ext4_free_inodes_count(sb, gdp);
201 percpu_counter_sub(&sbi->s_freeinodes_counter,
202 count);
203 }
Darrick J. Wongbdfb6ff2013-08-28 18:46:56 -0400204 set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, &grp->bb_state);
Darrick J. Wong6a797d22015-10-17 16:16:04 -0400205 return -EFSBADCRC;
Andreas Dilger717d50e2007-10-16 18:38:25 -0400206 }
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400207 memset(bh->b_data, 0, sb->s_blocksize);
Andreas Dilger717d50e2007-10-16 18:38:25 -0400208
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400209 bit_max = ext4_num_base_meta_clusters(sb, block_group);
Theodore Ts'o5b9554d2016-07-05 20:01:52 -0400210 if ((bit_max >> 3) >= bh->b_size)
211 return -EFSCORRUPTED;
212
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400213 for (bit = 0; bit < bit_max; bit++)
214 ext4_set_bit(bit, bh->b_data);
Akinobu Mitad00a6d72008-04-17 10:38:59 -0400215
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400216 start = ext4_group_first_block_no(sb, block_group);
Andreas Dilger717d50e2007-10-16 18:38:25 -0400217
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400218 /* Set bits for block and inode bitmaps, and inode table */
219 tmp = ext4_block_bitmap(sb, gdp);
Theodore Ts'ocdde8762018-06-13 23:00:48 -0400220 if (ext4_block_in_group(sb, tmp, block_group))
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400221 ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data);
Andreas Dilger717d50e2007-10-16 18:38:25 -0400222
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400223 tmp = ext4_inode_bitmap(sb, gdp);
Theodore Ts'ocdde8762018-06-13 23:00:48 -0400224 if (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);
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400226
227 tmp = ext4_inode_table(sb, gdp);
228 for (; tmp < ext4_inode_table(sb, gdp) +
229 sbi->s_itb_per_group; tmp++) {
Theodore Ts'ocdde8762018-06-13 23:00:48 -0400230 if (ext4_block_in_group(sb, tmp, block_group))
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400231 ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data);
Andreas Dilger717d50e2007-10-16 18:38:25 -0400232 }
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400233
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400234 /*
235 * Also if the number of blocks within the group is less than
236 * the blocksize * 8 ( which is the size of bitmap ), set rest
237 * of the block bitmap to 1
238 */
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400239 ext4_mark_bitmap_end(num_clusters_in_group(sb, block_group),
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400240 sb->s_blocksize * 8, bh->b_data);
Dmitry Monakhovaef48852014-10-13 03:42:12 -0400241 return 0;
Andreas Dilger717d50e2007-10-16 18:38:25 -0400242}
243
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400244/* Return the number of free blocks in a block group. It is used when
245 * the block bitmap is uninitialized, so we can't just count the bits
246 * in the bitmap. */
Theodore Ts'ocff1dfd72011-09-09 19:12:51 -0400247unsigned ext4_free_clusters_after_init(struct super_block *sb,
248 ext4_group_t block_group,
249 struct ext4_group_desc *gdp)
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400250{
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400251 return num_clusters_in_group(sb, block_group) -
252 ext4_num_overhead_clusters(sb, block_group, gdp);
Theodore Ts'ofd034a82011-09-09 18:42:51 -0400253}
Andreas Dilger717d50e2007-10-16 18:38:25 -0400254
Andrew Morton72b64b52006-10-11 01:21:18 -0700255/*
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700256 * The free blocks are managed by bitmaps. A file system contains several
257 * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap
258 * block for inodes, N blocks for the inode table and data blocks.
259 *
260 * The file system contains group descriptors which are located after the
261 * super block. Each descriptor contains the number of the bitmap block and
262 * the free blocks count in the block. The descriptors are loaded in memory
Aneesh Kumar K.Ve6274322007-02-20 13:57:58 -0800263 * when a file system is mounted (see ext4_fill_super).
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700264 */
265
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700266/**
Mingming Cao617ba132006-10-11 01:20:53 -0700267 * ext4_get_group_desc() -- load group descriptor from disk
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700268 * @sb: super block
269 * @block_group: given block group
270 * @bh: pointer to the buffer head to store the block
271 * group descriptor
272 */
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400273struct ext4_group_desc * ext4_get_group_desc(struct super_block *sb,
Avantika Mathurfd2d4292008-01-28 23:58:27 -0500274 ext4_group_t block_group,
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400275 struct buffer_head **bh)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700276{
Theodore Ts'o498e5f22008-11-05 00:14:04 -0500277 unsigned int group_desc;
278 unsigned int offset;
Theodore Ts'o8df96752009-05-01 08:50:38 -0400279 ext4_group_t ngroups = ext4_get_groups_count(sb);
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400280 struct ext4_group_desc *desc;
Mingming Cao617ba132006-10-11 01:20:53 -0700281 struct ext4_sb_info *sbi = EXT4_SB(sb);
Theodore Ts'oe9f47ea2020-02-28 16:51:17 -0800282 struct buffer_head *bh_p;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700283
Theodore Ts'o8df96752009-05-01 08:50:38 -0400284 if (block_group >= ngroups) {
Eric Sandeen12062dd2010-02-15 14:19:27 -0500285 ext4_error(sb, "block_group >= groups_count - block_group = %u,"
286 " groups_count = %u", block_group, ngroups);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700287
288 return NULL;
289 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700290
Mingming Cao617ba132006-10-11 01:20:53 -0700291 group_desc = block_group >> EXT4_DESC_PER_BLOCK_BITS(sb);
292 offset = block_group & (EXT4_DESC_PER_BLOCK(sb) - 1);
Theodore Ts'oe9f47ea2020-02-28 16:51:17 -0800293 bh_p = sbi_array_rcu_deref(sbi, s_group_desc, group_desc);
294 /*
295 * sbi_array_rcu_deref returns with rcu unlocked, this is ok since
296 * the pointer being dereferenced won't be dereferenced again. By
297 * looking at the usage in add_new_gdb() the value isn't modified,
298 * just the pointer, and so it remains valid.
299 */
300 if (!bh_p) {
Eric Sandeen12062dd2010-02-15 14:19:27 -0500301 ext4_error(sb, "Group descriptor not loaded - "
Theodore Ts'o498e5f22008-11-05 00:14:04 -0500302 "block_group = %u, group_desc = %u, desc = %u",
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400303 block_group, group_desc, offset);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700304 return NULL;
305 }
306
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -0700307 desc = (struct ext4_group_desc *)(
Theodore Ts'oe9f47ea2020-02-28 16:51:17 -0800308 (__u8 *)bh_p->b_data +
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -0700309 offset * EXT4_DESC_SIZE(sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700310 if (bh)
Theodore Ts'oe9f47ea2020-02-28 16:51:17 -0800311 *bh = bh_p;
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -0700312 return desc;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700313}
314
Theodore Ts'o7a4c5de2012-08-10 13:57:52 -0400315/*
316 * Return the block number which was discovered to be invalid, or 0 if
317 * the block bitmap is valid.
318 */
319static ext4_fsblk_t ext4_valid_block_bitmap(struct super_block *sb,
320 struct ext4_group_desc *desc,
Darrick J. Wongdbde0ab2013-08-28 15:59:51 -0400321 ext4_group_t block_group,
Theodore Ts'o7a4c5de2012-08-10 13:57:52 -0400322 struct buffer_head *bh)
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500323{
Darrick J. Wonge674e5c2014-05-12 10:17:55 -0400324 struct ext4_sb_info *sbi = EXT4_SB(sb);
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500325 ext4_grpblk_t offset;
326 ext4_grpblk_t next_zero_bit;
Lukas Czerner1fd7c772018-04-24 11:31:44 -0400327 ext4_grpblk_t max_bit = EXT4_CLUSTERS_PER_GROUP(sb);
Theodore Ts'o7a4c5de2012-08-10 13:57:52 -0400328 ext4_fsblk_t blk;
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500329 ext4_fsblk_t group_first_block;
330
Darrick J. Wonge2b911c2015-10-17 16:18:43 -0400331 if (ext4_has_feature_flex_bg(sb)) {
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500332 /* with FLEX_BG, the inode/block bitmaps and itable
333 * blocks may not be in the group at all
334 * so the bitmap validation will be skipped for those groups
335 * or it has to also read the block group where the bitmaps
336 * are located to verify they are set.
337 */
Theodore Ts'o7a4c5de2012-08-10 13:57:52 -0400338 return 0;
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500339 }
340 group_first_block = ext4_group_first_block_no(sb, block_group);
341
342 /* check whether block bitmap block number is set */
Theodore Ts'o7a4c5de2012-08-10 13:57:52 -0400343 blk = ext4_block_bitmap(sb, desc);
344 offset = blk - group_first_block;
Lukas Czerner1fd7c772018-04-24 11:31:44 -0400345 if (offset < 0 || EXT4_B2C(sbi, offset) >= max_bit ||
Theodore Ts'o76964812018-03-26 23:54:10 -0400346 !ext4_test_bit(EXT4_B2C(sbi, offset), bh->b_data))
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500347 /* bad block bitmap */
Theodore Ts'o7a4c5de2012-08-10 13:57:52 -0400348 return blk;
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500349
350 /* check whether the inode bitmap block number is set */
Theodore Ts'o7a4c5de2012-08-10 13:57:52 -0400351 blk = ext4_inode_bitmap(sb, desc);
352 offset = blk - group_first_block;
Lukas Czerner1fd7c772018-04-24 11:31:44 -0400353 if (offset < 0 || EXT4_B2C(sbi, offset) >= max_bit ||
Theodore Ts'o76964812018-03-26 23:54:10 -0400354 !ext4_test_bit(EXT4_B2C(sbi, offset), bh->b_data))
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500355 /* bad block bitmap */
Theodore Ts'o7a4c5de2012-08-10 13:57:52 -0400356 return blk;
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500357
358 /* check whether the inode table block number is set */
Theodore Ts'o7a4c5de2012-08-10 13:57:52 -0400359 blk = ext4_inode_table(sb, desc);
360 offset = blk - group_first_block;
Lukas Czerner1fd7c772018-04-24 11:31:44 -0400361 if (offset < 0 || EXT4_B2C(sbi, offset) >= max_bit ||
362 EXT4_B2C(sbi, offset + sbi->s_itb_per_group) >= max_bit)
Theodore Ts'o76964812018-03-26 23:54:10 -0400363 return blk;
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500364 next_zero_bit = ext4_find_next_zero_bit(bh->b_data,
Darrick J. Wonge674e5c2014-05-12 10:17:55 -0400365 EXT4_B2C(sbi, offset + EXT4_SB(sb)->s_itb_per_group),
366 EXT4_B2C(sbi, offset));
367 if (next_zero_bit <
368 EXT4_B2C(sbi, offset + EXT4_SB(sb)->s_itb_per_group))
Theodore Ts'o7a4c5de2012-08-10 13:57:52 -0400369 /* bad bitmap for inode tables */
370 return blk;
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500371 return 0;
372}
Darrick J. Wongfa77dcf2012-04-29 18:35:10 -0400373
Darrick J. Wong9008a582015-10-17 21:33:24 -0400374static int ext4_validate_block_bitmap(struct super_block *sb,
375 struct ext4_group_desc *desc,
376 ext4_group_t block_group,
377 struct buffer_head *bh)
Darrick J. Wongfa77dcf2012-04-29 18:35:10 -0400378{
Theodore Ts'o7a4c5de2012-08-10 13:57:52 -0400379 ext4_fsblk_t blk;
Darrick J. Wong163a2032013-08-28 17:35:51 -0400380 struct ext4_group_info *grp = ext4_get_group_info(sb, block_group);
Namjae Jeone43bb4e2014-06-26 10:11:53 -0400381 struct ext4_sb_info *sbi = EXT4_SB(sb);
Theodore Ts'o7a4c5de2012-08-10 13:57:52 -0400382
Darrick J. Wong9008a582015-10-17 21:33:24 -0400383 if (buffer_verified(bh))
384 return 0;
385 if (EXT4_MB_GRP_BBITMAP_CORRUPT(grp))
386 return -EFSCORRUPTED;
Darrick J. Wongfa77dcf2012-04-29 18:35:10 -0400387
388 ext4_lock_group(sb, block_group);
Theodore Ts'o262a62c2018-07-12 19:08:05 -0400389 if (buffer_verified(bh))
390 goto verified;
Darrick J. Wong9008a582015-10-17 21:33:24 -0400391 if (unlikely(!ext4_block_bitmap_csum_verify(sb, block_group,
392 desc, bh))) {
393 ext4_unlock_group(sb, block_group);
394 ext4_error(sb, "bg %u: bad block bitmap checksum", block_group);
395 if (!EXT4_MB_GRP_BBITMAP_CORRUPT(grp))
396 percpu_counter_sub(&sbi->s_freeclusters_counter,
397 grp->bb_free);
398 set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, &grp->bb_state);
399 return -EFSBADCRC;
400 }
Theodore Ts'o7a4c5de2012-08-10 13:57:52 -0400401 blk = ext4_valid_block_bitmap(sb, desc, block_group, bh);
402 if (unlikely(blk != 0)) {
403 ext4_unlock_group(sb, block_group);
404 ext4_error(sb, "bg %u: block %llu: invalid block bitmap",
405 block_group, blk);
Namjae Jeone43bb4e2014-06-26 10:11:53 -0400406 if (!EXT4_MB_GRP_BBITMAP_CORRUPT(grp))
407 percpu_counter_sub(&sbi->s_freeclusters_counter,
408 grp->bb_free);
Darrick J. Wong163a2032013-08-28 17:35:51 -0400409 set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, &grp->bb_state);
Darrick J. Wong9008a582015-10-17 21:33:24 -0400410 return -EFSCORRUPTED;
Theodore Ts'o7a4c5de2012-08-10 13:57:52 -0400411 }
412 set_buffer_verified(bh);
Theodore Ts'o262a62c2018-07-12 19:08:05 -0400413verified:
Darrick J. Wongfa77dcf2012-04-29 18:35:10 -0400414 ext4_unlock_group(sb, block_group);
Darrick J. Wong9008a582015-10-17 21:33:24 -0400415 return 0;
Darrick J. Wongfa77dcf2012-04-29 18:35:10 -0400416}
417
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700418/**
Eryu Guan15b49132013-01-12 16:33:25 -0500419 * ext4_read_block_bitmap_nowait()
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700420 * @sb: super block
421 * @block_group: given block group
422 *
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500423 * Read the bitmap for a given block_group,and validate the
424 * bits for block/inode/inode tables are set in the bitmaps
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700425 *
426 * Return buffer_head on success or NULL in case of failure.
427 */
Andreas Dilger717d50e2007-10-16 18:38:25 -0400428struct buffer_head *
Theodore Ts'o813e5722012-02-20 17:52:46 -0500429ext4_read_block_bitmap_nowait(struct super_block *sb, ext4_group_t block_group)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700430{
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400431 struct ext4_group_desc *desc;
Theodore Ts'o76964812018-03-26 23:54:10 -0400432 struct ext4_sb_info *sbi = EXT4_SB(sb);
Theodore Ts'o813e5722012-02-20 17:52:46 -0500433 struct buffer_head *bh;
Aneesh Kumar K.V7c9e69f2007-10-16 23:27:02 -0700434 ext4_fsblk_t bitmap_blk;
Darrick J. Wong9008a582015-10-17 21:33:24 -0400435 int err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700436
Andreas Dilger717d50e2007-10-16 18:38:25 -0400437 desc = ext4_get_group_desc(sb, block_group, NULL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700438 if (!desc)
Darrick J. Wong9008a582015-10-17 21:33:24 -0400439 return ERR_PTR(-EFSCORRUPTED);
Aneesh Kumar K.V7c9e69f2007-10-16 23:27:02 -0700440 bitmap_blk = ext4_block_bitmap(sb, desc);
Theodore Ts'o76964812018-03-26 23:54:10 -0400441 if ((bitmap_blk <= le32_to_cpu(sbi->s_es->s_first_data_block)) ||
442 (bitmap_blk >= ext4_blocks_count(sbi->s_es))) {
443 ext4_error(sb, "Invalid block bitmap block %llu in "
444 "block_group %u", bitmap_blk, block_group);
445 return ERR_PTR(-EFSCORRUPTED);
446 }
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500447 bh = sb_getblk(sb, bitmap_blk);
448 if (unlikely(!bh)) {
Theodore Ts'o813e5722012-02-20 17:52:46 -0500449 ext4_error(sb, "Cannot get buffer for block bitmap - "
450 "block_group = %u, block_bitmap = %llu",
451 block_group, bitmap_blk);
Darrick J. Wong9008a582015-10-17 21:33:24 -0400452 return ERR_PTR(-ENOMEM);
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500453 }
Aneesh Kumar K.V2ccb5fb2009-01-05 21:49:55 -0500454
455 if (bitmap_uptodate(bh))
Darrick J. Wongfa77dcf2012-04-29 18:35:10 -0400456 goto verify;
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500457
Frederic Bohec806e682008-10-10 08:09:18 -0400458 lock_buffer(bh);
Aneesh Kumar K.V2ccb5fb2009-01-05 21:49:55 -0500459 if (bitmap_uptodate(bh)) {
460 unlock_buffer(bh);
Darrick J. Wongfa77dcf2012-04-29 18:35:10 -0400461 goto verify;
Aneesh Kumar K.V2ccb5fb2009-01-05 21:49:55 -0500462 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -0400463 ext4_lock_group(sb, block_group);
Theodore Ts'o5ae57322018-06-14 00:58:00 -0400464 if (ext4_has_group_desc_csum(sb) &&
465 (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT))) {
466 if (block_group == 0) {
467 ext4_unlock_group(sb, block_group);
468 unlock_buffer(bh);
469 ext4_error(sb, "Block bitmap for bg 0 marked "
470 "uninitialized");
471 err = -EFSCORRUPTED;
472 goto out;
473 }
Dmitry Monakhovaef48852014-10-13 03:42:12 -0400474 err = ext4_init_block_bitmap(sb, bh, block_group, desc);
Aneesh Kumar K.V2ccb5fb2009-01-05 21:49:55 -0500475 set_bitmap_uptodate(bh);
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500476 set_buffer_uptodate(bh);
Theodore Ts'oa06b7982018-02-19 14:16:47 -0500477 set_buffer_verified(bh);
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -0400478 ext4_unlock_group(sb, block_group);
Aneesh Kumar K.V3300bed2009-01-03 22:33:39 -0500479 unlock_buffer(bh);
Jan Kara05145bd2016-02-11 23:15:12 -0500480 if (err) {
481 ext4_error(sb, "Failed to init block bitmap for group "
482 "%u: %d", block_group, err);
Darrick J. Wong9008a582015-10-17 21:33:24 -0400483 goto out;
Jan Kara05145bd2016-02-11 23:15:12 -0500484 }
Lukas Czerner41e5b7e2015-06-08 11:18:52 -0400485 goto verify;
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500486 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -0400487 ext4_unlock_group(sb, block_group);
Aneesh Kumar K.V2ccb5fb2009-01-05 21:49:55 -0500488 if (buffer_uptodate(bh)) {
489 /*
490 * if not uninit if bh is uptodate,
491 * bitmap is also uptodate
492 */
493 set_bitmap_uptodate(bh);
494 unlock_buffer(bh);
Darrick J. Wongfa77dcf2012-04-29 18:35:10 -0400495 goto verify;
Aneesh Kumar K.V2ccb5fb2009-01-05 21:49:55 -0500496 }
497 /*
Theodore Ts'o813e5722012-02-20 17:52:46 -0500498 * submit the buffer_head for reading
Aneesh Kumar K.V2ccb5fb2009-01-05 21:49:55 -0500499 */
Theodore Ts'o813e5722012-02-20 17:52:46 -0500500 set_buffer_new(bh);
Jiaying Zhang0562e0b2011-03-21 21:38:05 -0400501 trace_ext4_read_block_bitmap_load(sb, block_group);
Theodore Ts'o813e5722012-02-20 17:52:46 -0500502 bh->b_end_io = ext4_end_bitmap_read;
503 get_bh(bh);
Mike Christie2a222ca2016-06-05 14:31:43 -0500504 submit_bh(REQ_OP_READ, REQ_META | REQ_PRIO, bh);
Theodore Ts'o813e5722012-02-20 17:52:46 -0500505 return bh;
Darrick J. Wongfa77dcf2012-04-29 18:35:10 -0400506verify:
Darrick J. Wong9008a582015-10-17 21:33:24 -0400507 err = ext4_validate_block_bitmap(sb, desc, block_group, bh);
508 if (err)
509 goto out;
510 return bh;
511out:
Darrick J. Wong48d9eb92013-08-28 15:35:27 -0400512 put_bh(bh);
Darrick J. Wong9008a582015-10-17 21:33:24 -0400513 return ERR_PTR(err);
Theodore Ts'o813e5722012-02-20 17:52:46 -0500514}
515
516/* Returns 0 on success, 1 on error */
517int ext4_wait_block_bitmap(struct super_block *sb, ext4_group_t block_group,
518 struct buffer_head *bh)
519{
520 struct ext4_group_desc *desc;
521
522 if (!buffer_new(bh))
523 return 0;
524 desc = ext4_get_group_desc(sb, block_group, NULL);
525 if (!desc)
Darrick J. Wong9008a582015-10-17 21:33:24 -0400526 return -EFSCORRUPTED;
Theodore Ts'o813e5722012-02-20 17:52:46 -0500527 wait_on_buffer(bh);
528 if (!buffer_uptodate(bh)) {
Eric Sandeen12062dd2010-02-15 14:19:27 -0500529 ext4_error(sb, "Cannot read block bitmap - "
Theodore Ts'o813e5722012-02-20 17:52:46 -0500530 "block_group = %u, block_bitmap = %llu",
Heiko Carstensd4dc4622012-02-20 17:57:24 -0500531 block_group, (unsigned long long) bh->b_blocknr);
Darrick J. Wong9008a582015-10-17 21:33:24 -0400532 return -EIO;
Theodore Ts'o813e5722012-02-20 17:52:46 -0500533 }
534 clear_buffer_new(bh);
535 /* Panic or remount fs read-only if block bitmap is invalid */
Darrick J. Wong9008a582015-10-17 21:33:24 -0400536 return ext4_validate_block_bitmap(sb, desc, block_group, bh);
Theodore Ts'o813e5722012-02-20 17:52:46 -0500537}
538
539struct buffer_head *
540ext4_read_block_bitmap(struct super_block *sb, ext4_group_t block_group)
541{
542 struct buffer_head *bh;
Darrick J. Wong9008a582015-10-17 21:33:24 -0400543 int err;
Theodore Ts'o813e5722012-02-20 17:52:46 -0500544
545 bh = ext4_read_block_bitmap_nowait(sb, block_group);
Darrick J. Wong9008a582015-10-17 21:33:24 -0400546 if (IS_ERR(bh))
547 return bh;
548 err = ext4_wait_block_bitmap(sb, block_group, bh);
549 if (err) {
Theodore Ts'o813e5722012-02-20 17:52:46 -0500550 put_bh(bh);
Darrick J. Wong9008a582015-10-17 21:33:24 -0400551 return ERR_PTR(err);
Aneesh Kumar K.Vabcb2942008-01-28 23:58:27 -0500552 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700553 return bh;
554}
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700555
556/**
Theodore Ts'odf55c992011-09-09 19:16:51 -0400557 * ext4_has_free_clusters()
Eric Sandeen8c3bf8a02008-10-28 00:08:12 -0400558 * @sbi: in-core super block structure.
Theodore Ts'odf55c992011-09-09 19:16:51 -0400559 * @nclusters: number of needed blocks
560 * @flags: flags from ext4_mb_new_blocks()
Eric Sandeen8c3bf8a02008-10-28 00:08:12 -0400561 *
Theodore Ts'odf55c992011-09-09 19:16:51 -0400562 * Check if filesystem has nclusters free & available for allocation.
Eric Sandeen8c3bf8a02008-10-28 00:08:12 -0400563 * On success return 1, return 0 on failure.
564 */
Theodore Ts'odf55c992011-09-09 19:16:51 -0400565static int ext4_has_free_clusters(struct ext4_sb_info *sbi,
566 s64 nclusters, unsigned int flags)
Aneesh Kumar K.Va30d542a2008-10-09 10:56:23 -0400567{
Lukas Czerner27dd4382013-04-09 22:11:22 -0400568 s64 free_clusters, dirty_clusters, rsv, resv_clusters;
Theodore Ts'o57042652011-09-09 18:56:51 -0400569 struct percpu_counter *fcc = &sbi->s_freeclusters_counter;
Theodore Ts'odf55c992011-09-09 19:16:51 -0400570 struct percpu_counter *dcc = &sbi->s_dirtyclusters_counter;
Aneesh Kumar K.Va30d542a2008-10-09 10:56:23 -0400571
Theodore Ts'odf55c992011-09-09 19:16:51 -0400572 free_clusters = percpu_counter_read_positive(fcc);
573 dirty_clusters = percpu_counter_read_positive(dcc);
Lukas Czerner27dd4382013-04-09 22:11:22 -0400574 resv_clusters = atomic64_read(&sbi->s_resv_clusters);
Lukas Czerner304e2202013-02-22 15:27:52 -0500575
576 /*
577 * r_blocks_count should always be multiple of the cluster ratio so
578 * we are safe to do a plane bit shift only.
579 */
Lukas Czerner27dd4382013-04-09 22:11:22 -0400580 rsv = (ext4_r_blocks_count(sbi->s_es) >> sbi->s_cluster_bits) +
581 resv_clusters;
Aneesh Kumar K.Va30d542a2008-10-09 10:56:23 -0400582
Lukas Czerner27dd4382013-04-09 22:11:22 -0400583 if (free_clusters - (nclusters + rsv + dirty_clusters) <
Theodore Ts'odf55c992011-09-09 19:16:51 -0400584 EXT4_FREECLUSTERS_WATERMARK) {
Lukas Czerner304e2202013-02-22 15:27:52 -0500585 free_clusters = percpu_counter_sum_positive(fcc);
Theodore Ts'odf55c992011-09-09 19:16:51 -0400586 dirty_clusters = percpu_counter_sum_positive(dcc);
Aneesh Kumar K.V6bc6e632008-10-10 09:39:00 -0400587 }
Theodore Ts'odf55c992011-09-09 19:16:51 -0400588 /* Check whether we have space after accounting for current
589 * dirty clusters & root reserved clusters.
Aneesh Kumar K.V6bc6e632008-10-10 09:39:00 -0400590 */
Lukas Czerner27dd4382013-04-09 22:11:22 -0400591 if (free_clusters >= (rsv + nclusters + dirty_clusters))
Eric Sandeena9960312008-10-28 00:08:17 -0400592 return 1;
Aneesh Kumar K.V5c791612008-10-08 23:12:24 -0400593
Theodore Ts'odf55c992011-09-09 19:16:51 -0400594 /* Hm, nope. Are (enough) root reserved clusters available? */
Eric W. Biederman08cefc72012-02-07 15:41:49 -0800595 if (uid_eq(sbi->s_resuid, current_fsuid()) ||
596 (!gid_eq(sbi->s_resgid, GLOBAL_ROOT_GID) && in_group_p(sbi->s_resgid)) ||
Allison Henderson55f020d2011-05-25 07:41:26 -0400597 capable(CAP_SYS_RESOURCE) ||
Lukas Czerner27dd4382013-04-09 22:11:22 -0400598 (flags & EXT4_MB_USE_ROOT_BLOCKS)) {
Allison Henderson55f020d2011-05-25 07:41:26 -0400599
Lukas Czerner27dd4382013-04-09 22:11:22 -0400600 if (free_clusters >= (nclusters + dirty_clusters +
601 resv_clusters))
602 return 1;
603 }
604 /* No free blocks. Let's see if we can dip into reserved pool */
605 if (flags & EXT4_MB_USE_RESERVED) {
Theodore Ts'odf55c992011-09-09 19:16:51 -0400606 if (free_clusters >= (nclusters + dirty_clusters))
Eric Sandeena9960312008-10-28 00:08:17 -0400607 return 1;
608 }
609
610 return 0;
Aneesh Kumar K.Va30d542a2008-10-09 10:56:23 -0400611}
Mingming Cao07031432008-07-11 19:27:31 -0400612
Theodore Ts'oe7d5f312011-09-09 19:14:51 -0400613int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
614 s64 nclusters, unsigned int flags)
Eric Sandeen8c3bf8a02008-10-28 00:08:12 -0400615{
Theodore Ts'odf55c992011-09-09 19:16:51 -0400616 if (ext4_has_free_clusters(sbi, nclusters, flags)) {
Theodore Ts'oe7d5f312011-09-09 19:14:51 -0400617 percpu_counter_add(&sbi->s_dirtyclusters_counter, nclusters);
Eric Sandeen8c3bf8a02008-10-28 00:08:12 -0400618 return 0;
619 } else
620 return -ENOSPC;
621}
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700622
623/**
Mingming Cao617ba132006-10-11 01:20:53 -0700624 * ext4_should_retry_alloc()
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700625 * @sb: super block
626 * @retries number of attemps has been made
627 *
Mingming Cao617ba132006-10-11 01:20:53 -0700628 * ext4_should_retry_alloc() is called when ENOSPC is returned, and if
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700629 * it is profitable to retry the operation, this function will wait
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300630 * for the current or committing transaction to complete, and then
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700631 * return TRUE.
632 *
633 * if the total number of retries exceed three times, return FALSE.
634 */
Mingming Cao617ba132006-10-11 01:20:53 -0700635int ext4_should_retry_alloc(struct super_block *sb, int *retries)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700636{
Theodore Ts'odf55c992011-09-09 19:16:51 -0400637 if (!ext4_has_free_clusters(EXT4_SB(sb), 1, 0) ||
Eric Sandeen8f64b322009-02-26 00:57:35 -0500638 (*retries)++ > 3 ||
639 !EXT4_SB(sb)->s_journal)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700640 return 0;
641
642 jbd_debug(1, "%s: retrying operation after ENOSPC\n", sb->s_id);
643
Theodore Ts'od08854f2016-06-26 18:24:01 -0400644 smp_mb();
645 if (EXT4_SB(sb)->s_mb_free_pending)
646 jbd2_journal_force_commit_nested(EXT4_SB(sb)->s_journal);
Jan Karadbc427c2016-05-13 00:42:40 -0400647 return 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700648}
649
Aneesh Kumar K.V654b4902008-07-11 19:27:31 -0400650/*
Aneesh Kumar K.V654b4902008-07-11 19:27:31 -0400651 * ext4_new_meta_blocks() -- allocate block for meta data (indexing) blocks
652 *
653 * @handle: handle to this transaction
654 * @inode: file inode
655 * @goal: given target block(filesystem wide)
Aditya Kali7b415bf2011-09-09 19:04:51 -0400656 * @count: pointer to total number of clusters needed
Aneesh Kumar K.V654b4902008-07-11 19:27:31 -0400657 * @errp: error code
658 *
Theodore Ts'o97df5d12008-12-12 12:41:28 -0500659 * Return 1st allocated block number on success, *count stores total account
Aneesh Kumar K.V654b4902008-07-11 19:27:31 -0400660 * error stores in errp pointer
661 */
662ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
Allison Henderson55f020d2011-05-25 07:41:26 -0400663 ext4_fsblk_t goal, unsigned int flags,
664 unsigned long *count, int *errp)
Aneesh Kumar K.V654b4902008-07-11 19:27:31 -0400665{
Theodore Ts'o97df5d12008-12-12 12:41:28 -0500666 struct ext4_allocation_request ar;
Mingming Caod2a17632008-07-14 17:52:37 -0400667 ext4_fsblk_t ret;
Theodore Ts'o97df5d12008-12-12 12:41:28 -0500668
669 memset(&ar, 0, sizeof(ar));
670 /* Fill with neighbour allocated blocks */
671 ar.inode = inode;
672 ar.goal = goal;
673 ar.len = count ? *count : 1;
Allison Henderson55f020d2011-05-25 07:41:26 -0400674 ar.flags = flags;
Theodore Ts'o97df5d12008-12-12 12:41:28 -0500675
676 ret = ext4_mb_new_blocks(handle, &ar, errp);
677 if (count)
678 *count = ar.len;
Mingming Caod2a17632008-07-14 17:52:37 -0400679 /*
Eric Sandeen72b8ab92010-05-16 11:00:00 -0400680 * Account for the allocated meta blocks. We will never
681 * fail EDQUOT for metdata, but we do account for it.
Mingming Caod2a17632008-07-14 17:52:37 -0400682 */
Theodore Ts'oe3cf5d52014-09-04 18:07:25 -0400683 if (!(*errp) && (flags & EXT4_MB_DELALLOC_RESERVED)) {
Aditya Kali7b415bf2011-09-09 19:04:51 -0400684 dquot_alloc_block_nofail(inode,
685 EXT4_C2B(EXT4_SB(inode->i_sb), ar.len));
Mingming Caod2a17632008-07-14 17:52:37 -0400686 }
687 return ret;
688}
689
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700690/**
Theodore Ts'o5dee5432011-09-09 19:10:51 -0400691 * ext4_count_free_clusters() -- count filesystem free clusters
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700692 * @sb: superblock
693 *
Theodore Ts'o5dee5432011-09-09 19:10:51 -0400694 * Adds up the number of free clusters from each block group.
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700695 */
Theodore Ts'o5dee5432011-09-09 19:10:51 -0400696ext4_fsblk_t ext4_count_free_clusters(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700697{
Mingming Cao617ba132006-10-11 01:20:53 -0700698 ext4_fsblk_t desc_count;
699 struct ext4_group_desc *gdp;
Avantika Mathurfd2d4292008-01-28 23:58:27 -0500700 ext4_group_t i;
Theodore Ts'o8df96752009-05-01 08:50:38 -0400701 ext4_group_t ngroups = ext4_get_groups_count(sb);
Darrick J. Wong2746f7a2013-10-31 11:46:31 -0400702 struct ext4_group_info *grp;
Mingming Cao617ba132006-10-11 01:20:53 -0700703#ifdef EXT4FS_DEBUG
704 struct ext4_super_block *es;
705 ext4_fsblk_t bitmap_count;
Theodore Ts'o498e5f22008-11-05 00:14:04 -0500706 unsigned int x;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700707 struct buffer_head *bitmap_bh = NULL;
708
Mingming Cao617ba132006-10-11 01:20:53 -0700709 es = EXT4_SB(sb)->s_es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700710 desc_count = 0;
711 bitmap_count = 0;
712 gdp = NULL;
713
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700714 for (i = 0; i < ngroups; i++) {
Mingming Cao617ba132006-10-11 01:20:53 -0700715 gdp = ext4_get_group_desc(sb, i, NULL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700716 if (!gdp)
717 continue;
Darrick J. Wong2746f7a2013-10-31 11:46:31 -0400718 grp = NULL;
719 if (EXT4_SB(sb)->s_group_info)
720 grp = ext4_get_group_info(sb, i);
721 if (!grp || !EXT4_MB_GRP_BBITMAP_CORRUPT(grp))
722 desc_count += ext4_free_group_clusters(sb, gdp);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700723 brelse(bitmap_bh);
Theodore Ts'o574ca172008-07-11 19:27:31 -0400724 bitmap_bh = ext4_read_block_bitmap(sb, i);
Darrick J. Wong9008a582015-10-17 21:33:24 -0400725 if (IS_ERR(bitmap_bh)) {
726 bitmap_bh = NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700727 continue;
Darrick J. Wong9008a582015-10-17 21:33:24 -0400728 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700729
Theodore Ts'of6fb99c2012-06-30 19:14:57 -0400730 x = ext4_count_free(bitmap_bh->b_data,
Azat Khuzhin036acea2014-04-14 23:36:15 -0400731 EXT4_CLUSTERS_PER_GROUP(sb) / 8);
Thadeu Lima de Souza Cascardo9fd97842009-01-26 19:26:26 -0500732 printk(KERN_DEBUG "group %u: stored = %d, counted = %u\n",
Theodore Ts'o021b65b2011-09-09 19:08:51 -0400733 i, ext4_free_group_clusters(sb, gdp), x);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700734 bitmap_count += x;
735 }
736 brelse(bitmap_bh);
Theodore Ts'o5dee5432011-09-09 19:10:51 -0400737 printk(KERN_DEBUG "ext4_count_free_clusters: stored = %llu"
738 ", computed = %llu, %llu\n",
Lukas Czerner810da242013-03-02 17:18:58 -0500739 EXT4_NUM_B2C(EXT4_SB(sb), ext4_free_blocks_count(es)),
Theodore Ts'o4776004f2008-09-08 23:00:52 -0400740 desc_count, bitmap_count);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700741 return bitmap_count;
742#else
743 desc_count = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700744 for (i = 0; i < ngroups; i++) {
Mingming Cao617ba132006-10-11 01:20:53 -0700745 gdp = ext4_get_group_desc(sb, i, NULL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700746 if (!gdp)
747 continue;
Darrick J. Wong2746f7a2013-10-31 11:46:31 -0400748 grp = NULL;
749 if (EXT4_SB(sb)->s_group_info)
750 grp = ext4_get_group_info(sb, i);
751 if (!grp || !EXT4_MB_GRP_BBITMAP_CORRUPT(grp))
752 desc_count += ext4_free_group_clusters(sb, gdp);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700753 }
754
755 return desc_count;
756#endif
757}
758
Avantika Mathurfd2d4292008-01-28 23:58:27 -0500759static inline int test_root(ext4_group_t a, int b)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700760{
Theodore Ts'of4afb4f2013-06-06 11:40:37 -0400761 while (1) {
762 if (a < b)
763 return 0;
764 if (a == b)
765 return 1;
766 if ((a % b) != 0)
767 return 0;
768 a = a / b;
769 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700770}
771
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700772/**
Mingming Cao617ba132006-10-11 01:20:53 -0700773 * ext4_bg_has_super - number of blocks used by the superblock in group
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700774 * @sb: superblock for filesystem
775 * @group: group number to check
776 *
777 * Return the number of blocks used by the superblock (primary or backup)
778 * in this group. Currently this will be only 0 or 1.
779 */
Avantika Mathurfd2d4292008-01-28 23:58:27 -0500780int ext4_bg_has_super(struct super_block *sb, ext4_group_t group)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700781{
Darrick J. Wong1beeef12014-05-12 10:16:06 -0400782 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
783
784 if (group == 0)
785 return 1;
Darrick J. Wonge2b911c2015-10-17 16:18:43 -0400786 if (ext4_has_feature_sparse_super2(sb)) {
Darrick J. Wong1beeef12014-05-12 10:16:06 -0400787 if (group == le32_to_cpu(es->s_backup_bgs[0]) ||
788 group == le32_to_cpu(es->s_backup_bgs[1]))
789 return 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700790 return 0;
Darrick J. Wong1beeef12014-05-12 10:16:06 -0400791 }
Darrick J. Wonge2b911c2015-10-17 16:18:43 -0400792 if ((group <= 1) || !ext4_has_feature_sparse_super(sb))
Darrick J. Wong1beeef12014-05-12 10:16:06 -0400793 return 1;
794 if (!(group & 1))
795 return 0;
796 if (test_root(group, 3) || (test_root(group, 5)) ||
797 test_root(group, 7))
798 return 1;
799
800 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700801}
802
Avantika Mathurfd2d4292008-01-28 23:58:27 -0500803static unsigned long ext4_bg_num_gdb_meta(struct super_block *sb,
804 ext4_group_t group)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700805{
Mingming Cao617ba132006-10-11 01:20:53 -0700806 unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
Avantika Mathurfd2d4292008-01-28 23:58:27 -0500807 ext4_group_t first = metagroup * EXT4_DESC_PER_BLOCK(sb);
808 ext4_group_t last = first + EXT4_DESC_PER_BLOCK(sb) - 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700809
810 if (group == first || group == first + 1 || group == last)
811 return 1;
812 return 0;
813}
814
Avantika Mathurfd2d4292008-01-28 23:58:27 -0500815static unsigned long ext4_bg_num_gdb_nometa(struct super_block *sb,
816 ext4_group_t group)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700817{
Theodore Ts'o8dadb192009-11-23 07:24:38 -0500818 if (!ext4_bg_has_super(sb, group))
819 return 0;
820
Darrick J. Wonge2b911c2015-10-17 16:18:43 -0400821 if (ext4_has_feature_meta_bg(sb))
Theodore Ts'o8dadb192009-11-23 07:24:38 -0500822 return le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg);
823 else
824 return EXT4_SB(sb)->s_gdb_count;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700825}
826
827/**
Mingming Cao617ba132006-10-11 01:20:53 -0700828 * ext4_bg_num_gdb - number of blocks used by the group table in group
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700829 * @sb: superblock for filesystem
830 * @group: group number to check
831 *
832 * Return the number of blocks used by the group descriptor table
833 * (primary or backup) in this group. In the future there may be a
834 * different number of descriptor blocks in each group.
835 */
Avantika Mathurfd2d4292008-01-28 23:58:27 -0500836unsigned long ext4_bg_num_gdb(struct super_block *sb, ext4_group_t group)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700837{
838 unsigned long first_meta_bg =
Mingming Cao617ba132006-10-11 01:20:53 -0700839 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg);
840 unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700841
Darrick J. Wonge2b911c2015-10-17 16:18:43 -0400842 if (!ext4_has_feature_meta_bg(sb) || metagroup < first_meta_bg)
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400843 return ext4_bg_num_gdb_nometa(sb, group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700844
Mingming Cao617ba132006-10-11 01:20:53 -0700845 return ext4_bg_num_gdb_meta(sb,group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700846
847}
Theodore Ts'oc2ea3fd2008-10-10 09:40:52 -0400848
Theodore Ts'o49f7f9a2011-09-09 18:40:51 -0400849/*
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400850 * This function returns the number of file system metadata clusters at
Theodore Ts'o49f7f9a2011-09-09 18:40:51 -0400851 * the beginning of a block group, including the reserved gdt blocks.
852 */
Eric Sandeen5f163cc2012-01-04 22:33:28 -0500853static unsigned ext4_num_base_meta_clusters(struct super_block *sb,
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400854 ext4_group_t block_group)
Theodore Ts'o49f7f9a2011-09-09 18:40:51 -0400855{
856 struct ext4_sb_info *sbi = EXT4_SB(sb);
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400857 unsigned num;
Theodore Ts'o49f7f9a2011-09-09 18:40:51 -0400858
859 /* Check for superblock and gdt backups in this group */
860 num = ext4_bg_has_super(sb, block_group);
861
Darrick J. Wonge2b911c2015-10-17 16:18:43 -0400862 if (!ext4_has_feature_meta_bg(sb) ||
Theodore Ts'o49f7f9a2011-09-09 18:40:51 -0400863 block_group < le32_to_cpu(sbi->s_es->s_first_meta_bg) *
864 sbi->s_desc_per_block) {
865 if (num) {
866 num += ext4_bg_num_gdb(sb, block_group);
867 num += le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks);
868 }
869 } else { /* For META_BG_BLOCK_GROUPS */
870 num += ext4_bg_num_gdb(sb, block_group);
871 }
Theodore Ts'od5b8f312011-09-09 18:44:51 -0400872 return EXT4_NUM_B2C(sbi, num);
Theodore Ts'o49f7f9a2011-09-09 18:40:51 -0400873}
Eric Sandeenf86186b2011-06-28 10:01:31 -0400874/**
875 * ext4_inode_to_goal_block - return a hint for block allocation
876 * @inode: inode for block allocation
877 *
878 * Return the ideal location to start allocating blocks for a
879 * newly created inode.
880 */
881ext4_fsblk_t ext4_inode_to_goal_block(struct inode *inode)
882{
883 struct ext4_inode_info *ei = EXT4_I(inode);
884 ext4_group_t block_group;
885 ext4_grpblk_t colour;
886 int flex_size = ext4_flex_bg_size(EXT4_SB(inode->i_sb));
887 ext4_fsblk_t bg_start;
888 ext4_fsblk_t last_block;
889
890 block_group = ei->i_block_group;
891 if (flex_size >= EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME) {
892 /*
893 * If there are at least EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME
894 * block groups per flexgroup, reserve the first block
895 * group for directories and special files. Regular
896 * files will start at the second block group. This
897 * tends to speed up directory access and improves
898 * fsck times.
899 */
900 block_group &= ~(flex_size-1);
901 if (S_ISREG(inode->i_mode))
902 block_group++;
903 }
904 bg_start = ext4_group_first_block_no(inode->i_sb, block_group);
905 last_block = ext4_blocks_count(EXT4_SB(inode->i_sb)->s_es) - 1;
906
907 /*
908 * If we are doing delayed allocation, we don't need take
909 * colour into account.
910 */
911 if (test_opt(inode->i_sb, DELALLOC))
912 return bg_start;
913
914 if (bg_start + EXT4_BLOCKS_PER_GROUP(inode->i_sb) <= last_block)
915 colour = (current->pid % 16) *
916 (EXT4_BLOCKS_PER_GROUP(inode->i_sb) / 16);
917 else
918 colour = (current->pid % 16) * ((last_block - bg_start) / 16);
919 return bg_start + colour;
920}
921