1 // SPDX-License-Identifier: GPL-2.0 <<
2 /* 1 /*
3 * linux/fs/ext4/balloc.c 2 * linux/fs/ext4/balloc.c
4 * 3 *
5 * Copyright (C) 1992, 1993, 1994, 1995 4 * Copyright (C) 1992, 1993, 1994, 1995
6 * Remy Card (card@masi.ibp.fr) 5 * Remy Card (card@masi.ibp.fr)
7 * Laboratoire MASI - Institut Blaise Pascal 6 * Laboratoire MASI - Institut Blaise Pascal
8 * Universite Pierre et Marie Curie (Paris VI) 7 * Universite Pierre et Marie Curie (Paris VI)
9 * 8 *
10 * Enhanced block allocation by Stephen Tweed 9 * Enhanced block allocation by Stephen Tweedie (sct@redhat.com), 1993
11 * Big-endian to little-endian byte-swapping/ 10 * Big-endian to little-endian byte-swapping/bitmaps by
12 * David S. Miller (davem@caip.rutgers. 11 * David S. Miller (davem@caip.rutgers.edu), 1995
13 */ 12 */
14 13
15 #include <linux/time.h> 14 #include <linux/time.h>
16 #include <linux/capability.h> 15 #include <linux/capability.h>
17 #include <linux/fs.h> 16 #include <linux/fs.h>
18 #include <linux/quotaops.h> 17 #include <linux/quotaops.h>
19 #include <linux/buffer_head.h> 18 #include <linux/buffer_head.h>
20 #include "ext4.h" 19 #include "ext4.h"
21 #include "ext4_jbd2.h" 20 #include "ext4_jbd2.h"
22 #include "mballoc.h" 21 #include "mballoc.h"
23 22
24 #include <trace/events/ext4.h> 23 #include <trace/events/ext4.h>
25 #include <kunit/static_stub.h> <<
26 24
27 static unsigned ext4_num_base_meta_clusters(st 25 static unsigned ext4_num_base_meta_clusters(struct super_block *sb,
28 ex 26 ext4_group_t block_group);
29 /* 27 /*
30 * balloc.c contains the blocks allocation and 28 * balloc.c contains the blocks allocation and deallocation routines
31 */ 29 */
32 30
33 /* 31 /*
34 * Calculate block group number for a given bl 32 * Calculate block group number for a given block number
35 */ 33 */
36 ext4_group_t ext4_get_group_number(struct supe 34 ext4_group_t ext4_get_group_number(struct super_block *sb,
37 ext4_fsblk_ 35 ext4_fsblk_t block)
38 { 36 {
39 ext4_group_t group; 37 ext4_group_t group;
40 38
41 if (test_opt2(sb, STD_GROUP_SIZE)) 39 if (test_opt2(sb, STD_GROUP_SIZE))
42 group = (block - 40 group = (block -
43 le32_to_cpu(EXT4_SB(s 41 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block)) >>
44 (EXT4_BLOCK_SIZE_BITS( 42 (EXT4_BLOCK_SIZE_BITS(sb) + EXT4_CLUSTER_BITS(sb) + 3);
45 else 43 else
46 ext4_get_group_no_and_offset(s 44 ext4_get_group_no_and_offset(sb, block, &group, NULL);
47 return group; 45 return group;
48 } 46 }
49 47
50 /* 48 /*
51 * Calculate the block group number and offset 49 * Calculate the block group number and offset into the block/cluster
52 * allocation bitmap, given a block number 50 * allocation bitmap, given a block number
53 */ 51 */
54 void ext4_get_group_no_and_offset(struct super 52 void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
55 ext4_group_t *blockgrpp, ext4_ 53 ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp)
56 { 54 {
57 struct ext4_super_block *es = EXT4_SB( 55 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
58 ext4_grpblk_t offset; 56 ext4_grpblk_t offset;
59 57
60 blocknr = blocknr - le32_to_cpu(es->s_ 58 blocknr = blocknr - le32_to_cpu(es->s_first_data_block);
61 offset = do_div(blocknr, EXT4_BLOCKS_P 59 offset = do_div(blocknr, EXT4_BLOCKS_PER_GROUP(sb)) >>
62 EXT4_SB(sb)->s_cluster_bits; 60 EXT4_SB(sb)->s_cluster_bits;
63 if (offsetp) 61 if (offsetp)
64 *offsetp = offset; 62 *offsetp = offset;
65 if (blockgrpp) 63 if (blockgrpp)
66 *blockgrpp = blocknr; 64 *blockgrpp = blocknr;
67 65
68 } 66 }
69 67
70 /* 68 /*
71 * Check whether the 'block' lives within the 69 * Check whether the 'block' lives within the 'block_group'. Returns 1 if so
72 * and 0 otherwise. 70 * and 0 otherwise.
73 */ 71 */
74 static inline int ext4_block_in_group(struct s 72 static inline int ext4_block_in_group(struct super_block *sb,
75 ext4_fsb 73 ext4_fsblk_t block,
76 ext4_gro 74 ext4_group_t block_group)
77 { 75 {
78 ext4_group_t actual_group; 76 ext4_group_t actual_group;
79 77
80 actual_group = ext4_get_group_number(s 78 actual_group = ext4_get_group_number(sb, block);
81 return (actual_group == block_group) ? 79 return (actual_group == block_group) ? 1 : 0;
82 } 80 }
83 81
84 /* !! 82 /* Return the number of clusters used for file system metadata; this
85 * Return the number of clusters used for file <<
86 * represents the overhead needed by the file 83 * represents the overhead needed by the file system.
87 */ 84 */
88 static unsigned ext4_num_overhead_clusters(str 85 static unsigned ext4_num_overhead_clusters(struct super_block *sb,
89 ext 86 ext4_group_t block_group,
90 str 87 struct ext4_group_desc *gdp)
91 { 88 {
92 unsigned base_clusters, num_clusters; !! 89 unsigned num_clusters;
93 int block_cluster = -1, inode_cluster; !! 90 int block_cluster = -1, inode_cluster = -1, itbl_cluster = -1, i, c;
94 int itbl_cluster_start = -1, itbl_clus <<
95 ext4_fsblk_t start = ext4_group_first_ 91 ext4_fsblk_t start = ext4_group_first_block_no(sb, block_group);
96 ext4_fsblk_t end = start + EXT4_BLOCKS !! 92 ext4_fsblk_t itbl_blk;
97 ext4_fsblk_t itbl_blk_start, itbl_blk_ <<
98 struct ext4_sb_info *sbi = EXT4_SB(sb) 93 struct ext4_sb_info *sbi = EXT4_SB(sb);
99 94
100 /* This is the number of clusters used 95 /* This is the number of clusters used by the superblock,
101 * block group descriptors, and reserv 96 * block group descriptors, and reserved block group
102 * descriptor blocks */ 97 * descriptor blocks */
103 base_clusters = ext4_num_base_meta_clu !! 98 num_clusters = ext4_num_base_meta_clusters(sb, block_group);
104 num_clusters = base_clusters; <<
105 <<
106 /* <<
107 * Account and record inode table clus <<
108 * is in the block group, or inode tab <<
109 * [-1, -1] and won't overlap with blo <<
110 * accounted below. <<
111 */ <<
112 itbl_blk_start = ext4_inode_table(sb, <<
113 itbl_blk_end = itbl_blk_start + sbi->s <<
114 if (itbl_blk_start <= end && itbl_blk_ <<
115 itbl_blk_start = max(itbl_blk_ <<
116 itbl_blk_end = min(itbl_blk_en <<
117 <<
118 itbl_cluster_start = EXT4_B2C( <<
119 itbl_cluster_end = EXT4_B2C(sb <<
120 <<
121 num_clusters += itbl_cluster_e <<
122 /* check if border cluster is <<
123 if (itbl_cluster_start == base <<
124 num_clusters--; <<
125 } <<
126 99
127 /* 100 /*
128 * For the allocation bitmaps, we firs !! 101 * For the allocation bitmaps and inode table, we first need
129 * if the block is in the block group. !! 102 * to check to see if the block is in the block group. If it
130 * to see if the cluster is already ac !! 103 * is, then check to see if the cluster is already accounted
131 * used for the base metadata cluster !! 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.
132 * Normally all of these blocks are co 108 * Normally all of these blocks are contiguous, so the special
133 * case handling shouldn't be necessar 109 * case handling shouldn't be necessary except for *very*
134 * unusual file system layouts. 110 * unusual file system layouts.
135 */ 111 */
136 if (ext4_block_in_group(sb, ext4_block 112 if (ext4_block_in_group(sb, ext4_block_bitmap(sb, gdp), block_group)) {
137 block_cluster = EXT4_B2C(sbi, 113 block_cluster = EXT4_B2C(sbi,
138 ext4_ 114 ext4_block_bitmap(sb, gdp) - start);
139 if (block_cluster >= base_clus !! 115 if (block_cluster < num_clusters)
140 (block_cluster < itbl_clus !! 116 block_cluster = -1;
141 block_cluster > itbl_clust !! 117 else if (block_cluster == num_clusters) {
142 num_clusters++; 118 num_clusters++;
>> 119 block_cluster = -1;
>> 120 }
143 } 121 }
144 122
145 if (ext4_block_in_group(sb, ext4_inode 123 if (ext4_block_in_group(sb, ext4_inode_bitmap(sb, gdp), block_group)) {
146 inode_cluster = EXT4_B2C(sbi, 124 inode_cluster = EXT4_B2C(sbi,
147 ext4_ 125 ext4_inode_bitmap(sb, gdp) - start);
148 /* !! 126 if (inode_cluster < num_clusters)
149 * Additional check if inode b !! 127 inode_cluster = -1;
150 * block_cluster !! 128 else if (inode_cluster == num_clusters) {
151 */ !! 129 num_clusters++;
152 if (inode_cluster != block_clu !! 130 inode_cluster = -1;
153 inode_cluster >= base_clus !! 131 }
154 (inode_cluster < itbl_clus !! 132 }
155 inode_cluster > itbl_clust !! 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)) {
>> 137 c = EXT4_B2C(sbi, itbl_blk + i - start);
>> 138 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 }
156 num_clusters++; 145 num_clusters++;
>> 146 itbl_cluster = c;
>> 147 }
157 } 148 }
158 149
>> 150 if (block_cluster != -1)
>> 151 num_clusters++;
>> 152 if (inode_cluster != -1)
>> 153 num_clusters++;
>> 154
159 return num_clusters; 155 return num_clusters;
160 } 156 }
161 157
162 static unsigned int num_clusters_in_group(stru 158 static unsigned int num_clusters_in_group(struct super_block *sb,
163 ext4 159 ext4_group_t block_group)
164 { 160 {
165 unsigned int blocks; 161 unsigned int blocks;
166 162
167 if (block_group == ext4_get_groups_cou 163 if (block_group == ext4_get_groups_count(sb) - 1) {
168 /* 164 /*
169 * Even though mke2fs always i 165 * Even though mke2fs always initializes the first and
170 * last group, just in case so 166 * last group, just in case some other tool was used,
171 * we need to make sure we cal 167 * we need to make sure we calculate the right free
172 * blocks. 168 * blocks.
173 */ 169 */
174 blocks = ext4_blocks_count(EXT 170 blocks = ext4_blocks_count(EXT4_SB(sb)->s_es) -
175 ext4_group_first_block 171 ext4_group_first_block_no(sb, block_group);
176 } else 172 } else
177 blocks = EXT4_BLOCKS_PER_GROUP 173 blocks = EXT4_BLOCKS_PER_GROUP(sb);
178 return EXT4_NUM_B2C(EXT4_SB(sb), block 174 return EXT4_NUM_B2C(EXT4_SB(sb), blocks);
179 } 175 }
180 176
181 /* Initializes an uninitialized block bitmap * 177 /* Initializes an uninitialized block bitmap */
182 static int ext4_init_block_bitmap(struct super 178 static int ext4_init_block_bitmap(struct super_block *sb,
183 struct buff 179 struct buffer_head *bh,
184 ext4_group_ 180 ext4_group_t block_group,
185 struct ext4 181 struct ext4_group_desc *gdp)
186 { 182 {
187 unsigned int bit, bit_max; 183 unsigned int bit, bit_max;
188 struct ext4_sb_info *sbi = EXT4_SB(sb) 184 struct ext4_sb_info *sbi = EXT4_SB(sb);
189 ext4_fsblk_t start, tmp; 185 ext4_fsblk_t start, tmp;
>> 186 int flex_bg = 0;
>> 187 struct ext4_group_info *grp;
190 188
191 ASSERT(buffer_locked(bh)); !! 189 J_ASSERT_BH(bh, buffer_locked(bh));
192 190
>> 191 /* If checksum is bad mark all blocks used to prevent allocation
>> 192 * essentially implementing a per-group read-only flag. */
193 if (!ext4_group_desc_csum_verify(sb, b 193 if (!ext4_group_desc_csum_verify(sb, block_group, gdp)) {
194 ext4_mark_group_bitmap_corrupt !! 194 grp = ext4_get_group_info(sb, block_group);
195 EXT4_G !! 195 if (!EXT4_MB_GRP_BBITMAP_CORRUPT(grp))
196 EXT4_G !! 196 percpu_counter_sub(&sbi->s_freeclusters_counter,
>> 197 grp->bb_free);
>> 198 set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, &grp->bb_state);
>> 199 if (!EXT4_MB_GRP_IBITMAP_CORRUPT(grp)) {
>> 200 int count;
>> 201 count = ext4_free_inodes_count(sb, gdp);
>> 202 percpu_counter_sub(&sbi->s_freeinodes_counter,
>> 203 count);
>> 204 }
>> 205 set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, &grp->bb_state);
197 return -EFSBADCRC; 206 return -EFSBADCRC;
198 } 207 }
199 memset(bh->b_data, 0, sb->s_blocksize) 208 memset(bh->b_data, 0, sb->s_blocksize);
200 209
201 bit_max = ext4_num_base_meta_clusters( 210 bit_max = ext4_num_base_meta_clusters(sb, block_group);
202 if ((bit_max >> 3) >= bh->b_size) 211 if ((bit_max >> 3) >= bh->b_size)
203 return -EFSCORRUPTED; 212 return -EFSCORRUPTED;
204 213
205 for (bit = 0; bit < bit_max; bit++) 214 for (bit = 0; bit < bit_max; bit++)
206 ext4_set_bit(bit, bh->b_data); 215 ext4_set_bit(bit, bh->b_data);
207 216
208 start = ext4_group_first_block_no(sb, 217 start = ext4_group_first_block_no(sb, block_group);
209 218
>> 219 if (ext4_has_feature_flex_bg(sb))
>> 220 flex_bg = 1;
>> 221
210 /* Set bits for block and inode bitmap 222 /* Set bits for block and inode bitmaps, and inode table */
211 tmp = ext4_block_bitmap(sb, gdp); 223 tmp = ext4_block_bitmap(sb, gdp);
212 if (ext4_block_in_group(sb, tmp, block !! 224 if (!flex_bg || ext4_block_in_group(sb, tmp, block_group))
213 ext4_set_bit(EXT4_B2C(sbi, tmp 225 ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data);
214 226
215 tmp = ext4_inode_bitmap(sb, gdp); 227 tmp = ext4_inode_bitmap(sb, gdp);
216 if (ext4_block_in_group(sb, tmp, block !! 228 if (!flex_bg || ext4_block_in_group(sb, tmp, block_group))
217 ext4_set_bit(EXT4_B2C(sbi, tmp 229 ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data);
218 230
219 tmp = ext4_inode_table(sb, gdp); 231 tmp = ext4_inode_table(sb, gdp);
220 for (; tmp < ext4_inode_table(sb, gdp) 232 for (; tmp < ext4_inode_table(sb, gdp) +
221 sbi->s_itb_per_group; tmp 233 sbi->s_itb_per_group; tmp++) {
222 if (ext4_block_in_group(sb, tm !! 234 if (!flex_bg || ext4_block_in_group(sb, tmp, block_group))
223 ext4_set_bit(EXT4_B2C( 235 ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data);
224 } 236 }
225 237
226 /* 238 /*
227 * Also if the number of blocks within 239 * Also if the number of blocks within the group is less than
228 * the blocksize * 8 ( which is the si 240 * the blocksize * 8 ( which is the size of bitmap ), set rest
229 * of the block bitmap to 1 241 * of the block bitmap to 1
230 */ 242 */
231 ext4_mark_bitmap_end(num_clusters_in_g 243 ext4_mark_bitmap_end(num_clusters_in_group(sb, block_group),
232 sb->s_blocksize * 244 sb->s_blocksize * 8, bh->b_data);
>> 245 ext4_block_bitmap_csum_set(sb, block_group, gdp, bh);
>> 246 ext4_group_desc_csum_set(sb, block_group, gdp);
233 return 0; 247 return 0;
234 } 248 }
235 249
236 /* Return the number of free blocks in a block 250 /* Return the number of free blocks in a block group. It is used when
237 * the block bitmap is uninitialized, so we ca 251 * the block bitmap is uninitialized, so we can't just count the bits
238 * in the bitmap. */ 252 * in the bitmap. */
239 unsigned ext4_free_clusters_after_init(struct 253 unsigned ext4_free_clusters_after_init(struct super_block *sb,
240 ext4_gr 254 ext4_group_t block_group,
241 struct 255 struct ext4_group_desc *gdp)
242 { 256 {
243 return num_clusters_in_group(sb, block !! 257 return num_clusters_in_group(sb, block_group) -
244 ext4_num_overhead_clusters(sb, 258 ext4_num_overhead_clusters(sb, block_group, gdp);
245 } 259 }
246 260
247 /* 261 /*
248 * The free blocks are managed by bitmaps. A 262 * The free blocks are managed by bitmaps. A file system contains several
249 * blocks groups. Each group contains 1 bitma 263 * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap
250 * block for inodes, N blocks for the inode ta 264 * block for inodes, N blocks for the inode table and data blocks.
251 * 265 *
252 * The file system contains group descriptors 266 * The file system contains group descriptors which are located after the
253 * super block. Each descriptor contains the 267 * super block. Each descriptor contains the number of the bitmap block and
254 * the free blocks count in the block. The de 268 * the free blocks count in the block. The descriptors are loaded in memory
255 * when a file system is mounted (see ext4_fil 269 * when a file system is mounted (see ext4_fill_super).
256 */ 270 */
257 271
258 /** 272 /**
259 * ext4_get_group_desc() -- load group descrip 273 * ext4_get_group_desc() -- load group descriptor from disk
260 * @sb: super block 274 * @sb: super block
261 * @block_group: given block group 275 * @block_group: given block group
262 * @bh: pointer to the buffer 276 * @bh: pointer to the buffer head to store the block
263 * group descriptor 277 * group descriptor
264 */ 278 */
265 struct ext4_group_desc * ext4_get_group_desc(s 279 struct ext4_group_desc * ext4_get_group_desc(struct super_block *sb,
266 e 280 ext4_group_t block_group,
267 s 281 struct buffer_head **bh)
268 { 282 {
269 unsigned int group_desc; 283 unsigned int group_desc;
270 unsigned int offset; 284 unsigned int offset;
271 ext4_group_t ngroups = ext4_get_groups 285 ext4_group_t ngroups = ext4_get_groups_count(sb);
272 struct ext4_group_desc *desc; 286 struct ext4_group_desc *desc;
273 struct ext4_sb_info *sbi = EXT4_SB(sb) 287 struct ext4_sb_info *sbi = EXT4_SB(sb);
274 struct buffer_head *bh_p; <<
275 <<
276 KUNIT_STATIC_STUB_REDIRECT(ext4_get_gr <<
277 sb, block_g <<
278 288
279 if (block_group >= ngroups) { 289 if (block_group >= ngroups) {
280 ext4_error(sb, "block_group >= 290 ext4_error(sb, "block_group >= groups_count - block_group = %u,"
281 " groups_count = %u 291 " groups_count = %u", block_group, ngroups);
282 292
283 return NULL; 293 return NULL;
284 } 294 }
285 295
286 group_desc = block_group >> EXT4_DESC_ 296 group_desc = block_group >> EXT4_DESC_PER_BLOCK_BITS(sb);
287 offset = block_group & (EXT4_DESC_PER_ 297 offset = block_group & (EXT4_DESC_PER_BLOCK(sb) - 1);
288 bh_p = sbi_array_rcu_deref(sbi, s_grou !! 298 if (!sbi->s_group_desc[group_desc]) {
289 /* <<
290 * sbi_array_rcu_deref returns with rc <<
291 * the pointer being dereferenced won' <<
292 * looking at the usage in add_new_gdb <<
293 * just the pointer, and so it remains <<
294 */ <<
295 if (!bh_p) { <<
296 ext4_error(sb, "Group descript 299 ext4_error(sb, "Group descriptor not loaded - "
297 "block_group = %u, 300 "block_group = %u, group_desc = %u, desc = %u",
298 block_group, group_ 301 block_group, group_desc, offset);
299 return NULL; 302 return NULL;
300 } 303 }
301 304
302 desc = (struct ext4_group_desc *)( 305 desc = (struct ext4_group_desc *)(
303 (__u8 *)bh_p->b_data + !! 306 (__u8 *)sbi->s_group_desc[group_desc]->b_data +
304 offset * EXT4_DESC_SIZE(sb)); 307 offset * EXT4_DESC_SIZE(sb));
305 if (bh) 308 if (bh)
306 *bh = bh_p; !! 309 *bh = sbi->s_group_desc[group_desc];
307 return desc; 310 return desc;
308 } 311 }
309 312
310 static ext4_fsblk_t ext4_valid_block_bitmap_pa <<
311 <<
312 <<
313 { <<
314 ext4_grpblk_t next_zero_bit; <<
315 unsigned long bitmap_size = sb->s_bloc <<
316 unsigned int offset = num_clusters_in_ <<
317 <<
318 if (bitmap_size <= offset) <<
319 return 0; <<
320 <<
321 next_zero_bit = ext4_find_next_zero_bi <<
322 <<
323 return (next_zero_bit < bitmap_size ? <<
324 } <<
325 <<
326 struct ext4_group_info *ext4_get_group_info(st <<
327 ex <<
328 { <<
329 struct ext4_group_info **grp_info; <<
330 long indexv, indexh; <<
331 <<
332 if (unlikely(group >= EXT4_SB(sb)->s_g <<
333 return NULL; <<
334 indexv = group >> (EXT4_DESC_PER_BLOCK <<
335 indexh = group & ((EXT4_DESC_PER_BLOCK <<
336 grp_info = sbi_array_rcu_deref(EXT4_SB <<
337 return grp_info[indexh]; <<
338 } <<
339 <<
340 /* 313 /*
341 * Return the block number which was discovere 314 * Return the block number which was discovered to be invalid, or 0 if
342 * the block bitmap is valid. 315 * the block bitmap is valid.
343 */ 316 */
344 static ext4_fsblk_t ext4_valid_block_bitmap(st 317 static ext4_fsblk_t ext4_valid_block_bitmap(struct super_block *sb,
345 st 318 struct ext4_group_desc *desc,
346 ex 319 ext4_group_t block_group,
347 st 320 struct buffer_head *bh)
348 { 321 {
349 struct ext4_sb_info *sbi = EXT4_SB(sb) 322 struct ext4_sb_info *sbi = EXT4_SB(sb);
350 ext4_grpblk_t offset; 323 ext4_grpblk_t offset;
351 ext4_grpblk_t next_zero_bit; 324 ext4_grpblk_t next_zero_bit;
352 ext4_grpblk_t max_bit = EXT4_CLUSTERS_ <<
353 ext4_fsblk_t blk; 325 ext4_fsblk_t blk;
354 ext4_fsblk_t group_first_block; 326 ext4_fsblk_t group_first_block;
355 327
356 if (ext4_has_feature_flex_bg(sb)) { 328 if (ext4_has_feature_flex_bg(sb)) {
357 /* with FLEX_BG, the inode/blo 329 /* with FLEX_BG, the inode/block bitmaps and itable
358 * blocks may not be in the gr 330 * blocks may not be in the group at all
359 * so the bitmap validation wi 331 * so the bitmap validation will be skipped for those groups
360 * or it has to also read the 332 * or it has to also read the block group where the bitmaps
361 * are located to verify they 333 * are located to verify they are set.
362 */ 334 */
363 return 0; 335 return 0;
364 } 336 }
365 group_first_block = ext4_group_first_b 337 group_first_block = ext4_group_first_block_no(sb, block_group);
366 338
367 /* check whether block bitmap block nu 339 /* check whether block bitmap block number is set */
368 blk = ext4_block_bitmap(sb, desc); 340 blk = ext4_block_bitmap(sb, desc);
369 offset = blk - group_first_block; 341 offset = blk - group_first_block;
370 if (offset < 0 || EXT4_B2C(sbi, offset !! 342 if (!ext4_test_bit(EXT4_B2C(sbi, offset), bh->b_data))
371 !ext4_test_bit(EXT4_B2C(sbi, offse <<
372 /* bad block bitmap */ 343 /* bad block bitmap */
373 return blk; 344 return blk;
374 345
375 /* check whether the inode bitmap bloc 346 /* check whether the inode bitmap block number is set */
376 blk = ext4_inode_bitmap(sb, desc); 347 blk = ext4_inode_bitmap(sb, desc);
377 offset = blk - group_first_block; 348 offset = blk - group_first_block;
378 if (offset < 0 || EXT4_B2C(sbi, offset !! 349 if (!ext4_test_bit(EXT4_B2C(sbi, offset), bh->b_data))
379 !ext4_test_bit(EXT4_B2C(sbi, offse <<
380 /* bad block bitmap */ 350 /* bad block bitmap */
381 return blk; 351 return blk;
382 352
383 /* check whether the inode table block 353 /* check whether the inode table block number is set */
384 blk = ext4_inode_table(sb, desc); 354 blk = ext4_inode_table(sb, desc);
385 offset = blk - group_first_block; 355 offset = blk - group_first_block;
386 if (offset < 0 || EXT4_B2C(sbi, offset <<
387 EXT4_B2C(sbi, offset + sbi->s_itb_ <<
388 return blk; <<
389 next_zero_bit = ext4_find_next_zero_bi 356 next_zero_bit = ext4_find_next_zero_bit(bh->b_data,
390 EXT4_B2C(sbi, offset + !! 357 EXT4_B2C(sbi, offset + EXT4_SB(sb)->s_itb_per_group),
391 EXT4_B2C(sbi, offset)) 358 EXT4_B2C(sbi, offset));
392 if (next_zero_bit < 359 if (next_zero_bit <
393 EXT4_B2C(sbi, offset + sbi->s_itb_ !! 360 EXT4_B2C(sbi, offset + EXT4_SB(sb)->s_itb_per_group))
394 /* bad bitmap for inode tables 361 /* bad bitmap for inode tables */
395 return blk; 362 return blk;
396 return 0; 363 return 0;
397 } 364 }
398 365
399 static int ext4_validate_block_bitmap(struct s 366 static int ext4_validate_block_bitmap(struct super_block *sb,
400 struct e 367 struct ext4_group_desc *desc,
401 ext4_gro 368 ext4_group_t block_group,
402 struct b 369 struct buffer_head *bh)
403 { 370 {
404 ext4_fsblk_t blk; 371 ext4_fsblk_t blk;
405 struct ext4_group_info *grp; !! 372 struct ext4_group_info *grp = ext4_get_group_info(sb, block_group);
406 !! 373 struct ext4_sb_info *sbi = EXT4_SB(sb);
407 if (EXT4_SB(sb)->s_mount_state & EXT4_ <<
408 return 0; <<
409 <<
410 grp = ext4_get_group_info(sb, block_gr <<
411 374
412 if (buffer_verified(bh)) 375 if (buffer_verified(bh))
413 return 0; 376 return 0;
414 if (!grp || EXT4_MB_GRP_BBITMAP_CORRUP !! 377 if (EXT4_MB_GRP_BBITMAP_CORRUPT(grp))
415 return -EFSCORRUPTED; 378 return -EFSCORRUPTED;
416 379
417 ext4_lock_group(sb, block_group); 380 ext4_lock_group(sb, block_group);
418 if (buffer_verified(bh)) !! 381 if (unlikely(!ext4_block_bitmap_csum_verify(sb, block_group,
419 goto verified; !! 382 desc, bh))) {
420 if (unlikely(!ext4_block_bitmap_csum_v <<
421 ext4_simulate_fail(sb, EX <<
422 ext4_unlock_group(sb, block_gr 383 ext4_unlock_group(sb, block_group);
423 ext4_error(sb, "bg %u: bad blo 384 ext4_error(sb, "bg %u: bad block bitmap checksum", block_group);
424 ext4_mark_group_bitmap_corrupt !! 385 if (!EXT4_MB_GRP_BBITMAP_CORRUPT(grp))
425 EXT4_G !! 386 percpu_counter_sub(&sbi->s_freeclusters_counter,
>> 387 grp->bb_free);
>> 388 set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, &grp->bb_state);
426 return -EFSBADCRC; 389 return -EFSBADCRC;
427 } 390 }
428 blk = ext4_valid_block_bitmap(sb, desc 391 blk = ext4_valid_block_bitmap(sb, desc, block_group, bh);
429 if (unlikely(blk != 0)) { 392 if (unlikely(blk != 0)) {
430 ext4_unlock_group(sb, block_gr 393 ext4_unlock_group(sb, block_group);
431 ext4_error(sb, "bg %u: block % 394 ext4_error(sb, "bg %u: block %llu: invalid block bitmap",
432 block_group, blk); 395 block_group, blk);
433 ext4_mark_group_bitmap_corrupt !! 396 if (!EXT4_MB_GRP_BBITMAP_CORRUPT(grp))
434 EXT4_G !! 397 percpu_counter_sub(&sbi->s_freeclusters_counter,
435 return -EFSCORRUPTED; !! 398 grp->bb_free);
436 } !! 399 set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, &grp->bb_state);
437 blk = ext4_valid_block_bitmap_padding( <<
438 if (unlikely(blk != 0)) { <<
439 ext4_unlock_group(sb, block_gr <<
440 ext4_error(sb, "bg %u: block % <<
441 block_group, blk); <<
442 ext4_mark_group_bitmap_corrupt <<
443 <<
444 return -EFSCORRUPTED; 400 return -EFSCORRUPTED;
445 } 401 }
446 set_buffer_verified(bh); 402 set_buffer_verified(bh);
447 verified: <<
448 ext4_unlock_group(sb, block_group); 403 ext4_unlock_group(sb, block_group);
449 return 0; 404 return 0;
450 } 405 }
451 406
452 /** 407 /**
453 * ext4_read_block_bitmap_nowait() 408 * ext4_read_block_bitmap_nowait()
454 * @sb: super block 409 * @sb: super block
455 * @block_group: given block group 410 * @block_group: given block group
456 * @ignore_locked: ignore locked buffers <<
457 * 411 *
458 * Read the bitmap for a given block_group,and 412 * Read the bitmap for a given block_group,and validate the
459 * bits for block/inode/inode tables are set i 413 * bits for block/inode/inode tables are set in the bitmaps
460 * 414 *
461 * Return buffer_head on success or an ERR_PTR !! 415 * Return buffer_head on success or NULL in case of failure.
462 */ 416 */
463 struct buffer_head * 417 struct buffer_head *
464 ext4_read_block_bitmap_nowait(struct super_blo !! 418 ext4_read_block_bitmap_nowait(struct super_block *sb, ext4_group_t block_group)
465 bool ignore_lock <<
466 { 419 {
467 struct ext4_group_desc *desc; 420 struct ext4_group_desc *desc;
468 struct ext4_sb_info *sbi = EXT4_SB(sb) <<
469 struct buffer_head *bh; 421 struct buffer_head *bh;
470 ext4_fsblk_t bitmap_blk; 422 ext4_fsblk_t bitmap_blk;
471 int err; 423 int err;
472 424
473 KUNIT_STATIC_STUB_REDIRECT(ext4_read_b <<
474 sb, block_g <<
475 <<
476 desc = ext4_get_group_desc(sb, block_g 425 desc = ext4_get_group_desc(sb, block_group, NULL);
477 if (!desc) 426 if (!desc)
478 return ERR_PTR(-EFSCORRUPTED); 427 return ERR_PTR(-EFSCORRUPTED);
479 bitmap_blk = ext4_block_bitmap(sb, des 428 bitmap_blk = ext4_block_bitmap(sb, desc);
480 if ((bitmap_blk <= le32_to_cpu(sbi->s_ <<
481 (bitmap_blk >= ext4_blocks_count(s <<
482 ext4_error(sb, "Invalid block <<
483 "block_group %u", b <<
484 ext4_mark_group_bitmap_corrupt <<
485 EXT4_G <<
486 return ERR_PTR(-EFSCORRUPTED); <<
487 } <<
488 bh = sb_getblk(sb, bitmap_blk); 429 bh = sb_getblk(sb, bitmap_blk);
489 if (unlikely(!bh)) { 430 if (unlikely(!bh)) {
490 ext4_warning(sb, "Cannot get b !! 431 ext4_error(sb, "Cannot get buffer for block bitmap - "
491 "block_group = %u !! 432 "block_group = %u, block_bitmap = %llu",
492 block_group, bitm !! 433 block_group, bitmap_blk);
493 return ERR_PTR(-ENOMEM); 434 return ERR_PTR(-ENOMEM);
494 } 435 }
495 436
496 if (ignore_locked && buffer_locked(bh) <<
497 /* buffer under IO already, re <<
498 put_bh(bh); <<
499 return NULL; <<
500 } <<
501 <<
502 if (bitmap_uptodate(bh)) 437 if (bitmap_uptodate(bh))
503 goto verify; 438 goto verify;
504 439
505 lock_buffer(bh); 440 lock_buffer(bh);
506 if (bitmap_uptodate(bh)) { 441 if (bitmap_uptodate(bh)) {
507 unlock_buffer(bh); 442 unlock_buffer(bh);
508 goto verify; 443 goto verify;
509 } 444 }
510 ext4_lock_group(sb, block_group); 445 ext4_lock_group(sb, block_group);
511 if (ext4_has_group_desc_csum(sb) && !! 446 if (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) {
512 (desc->bg_flags & cpu_to_le16(EXT4 <<
513 if (block_group == 0) { <<
514 ext4_unlock_group(sb, <<
515 unlock_buffer(bh); <<
516 ext4_error(sb, "Block <<
517 "uninitiali <<
518 err = -EFSCORRUPTED; <<
519 goto out; <<
520 } <<
521 err = ext4_init_block_bitmap(s 447 err = ext4_init_block_bitmap(sb, bh, block_group, desc);
>> 448 set_bitmap_uptodate(bh);
>> 449 set_buffer_uptodate(bh);
>> 450 ext4_unlock_group(sb, block_group);
>> 451 unlock_buffer(bh);
522 if (err) { 452 if (err) {
523 ext4_unlock_group(sb, <<
524 unlock_buffer(bh); <<
525 ext4_error(sb, "Failed 453 ext4_error(sb, "Failed to init block bitmap for group "
526 "%u: %d", b 454 "%u: %d", block_group, err);
527 goto out; 455 goto out;
528 } 456 }
529 set_bitmap_uptodate(bh); !! 457 goto verify;
530 set_buffer_uptodate(bh); <<
531 set_buffer_verified(bh); <<
532 ext4_unlock_group(sb, block_gr <<
533 unlock_buffer(bh); <<
534 return bh; <<
535 } 458 }
536 ext4_unlock_group(sb, block_group); 459 ext4_unlock_group(sb, block_group);
537 if (buffer_uptodate(bh)) { 460 if (buffer_uptodate(bh)) {
538 /* 461 /*
539 * if not uninit if bh is upto 462 * if not uninit if bh is uptodate,
540 * bitmap is also uptodate 463 * bitmap is also uptodate
541 */ 464 */
542 set_bitmap_uptodate(bh); 465 set_bitmap_uptodate(bh);
543 unlock_buffer(bh); 466 unlock_buffer(bh);
544 goto verify; 467 goto verify;
545 } 468 }
546 /* 469 /*
547 * submit the buffer_head for reading 470 * submit the buffer_head for reading
548 */ 471 */
549 set_buffer_new(bh); 472 set_buffer_new(bh);
550 trace_ext4_read_block_bitmap_load(sb, !! 473 trace_ext4_read_block_bitmap_load(sb, block_group);
551 ext4_read_bh_nowait(bh, REQ_META | REQ !! 474 bh->b_end_io = ext4_end_bitmap_read;
552 (ignore_locked ? R !! 475 get_bh(bh);
553 ext4_end_bitmap_re !! 476 submit_bh(REQ_OP_READ, REQ_META | REQ_PRIO, bh);
554 return bh; 477 return bh;
555 verify: 478 verify:
556 err = ext4_validate_block_bitmap(sb, d 479 err = ext4_validate_block_bitmap(sb, desc, block_group, bh);
557 if (err) 480 if (err)
558 goto out; 481 goto out;
559 return bh; 482 return bh;
560 out: 483 out:
561 put_bh(bh); 484 put_bh(bh);
562 return ERR_PTR(err); 485 return ERR_PTR(err);
563 } 486 }
564 487
565 /* Returns 0 on success, -errno on error */ !! 488 /* Returns 0 on success, 1 on error */
566 int ext4_wait_block_bitmap(struct super_block 489 int ext4_wait_block_bitmap(struct super_block *sb, ext4_group_t block_group,
567 struct buffer_head 490 struct buffer_head *bh)
568 { 491 {
569 struct ext4_group_desc *desc; 492 struct ext4_group_desc *desc;
570 493
571 KUNIT_STATIC_STUB_REDIRECT(ext4_wait_b <<
572 sb, block_g <<
573 <<
574 if (!buffer_new(bh)) 494 if (!buffer_new(bh))
575 return 0; 495 return 0;
576 desc = ext4_get_group_desc(sb, block_g 496 desc = ext4_get_group_desc(sb, block_group, NULL);
577 if (!desc) 497 if (!desc)
578 return -EFSCORRUPTED; 498 return -EFSCORRUPTED;
579 wait_on_buffer(bh); 499 wait_on_buffer(bh);
580 ext4_simulate_fail_bh(sb, bh, EXT4_SIM <<
581 if (!buffer_uptodate(bh)) { 500 if (!buffer_uptodate(bh)) {
582 ext4_error_err(sb, EIO, "Canno !! 501 ext4_error(sb, "Cannot read block bitmap - "
583 "block_group = !! 502 "block_group = %u, block_bitmap = %llu",
584 block_group, (u !! 503 block_group, (unsigned long long) bh->b_blocknr);
585 ext4_mark_group_bitmap_corrupt <<
586 EXT4_G <<
587 return -EIO; 504 return -EIO;
588 } 505 }
589 clear_buffer_new(bh); 506 clear_buffer_new(bh);
590 /* Panic or remount fs read-only if bl 507 /* Panic or remount fs read-only if block bitmap is invalid */
591 return ext4_validate_block_bitmap(sb, 508 return ext4_validate_block_bitmap(sb, desc, block_group, bh);
592 } 509 }
593 510
594 struct buffer_head * 511 struct buffer_head *
595 ext4_read_block_bitmap(struct super_block *sb, 512 ext4_read_block_bitmap(struct super_block *sb, ext4_group_t block_group)
596 { 513 {
597 struct buffer_head *bh; 514 struct buffer_head *bh;
598 int err; 515 int err;
599 516
600 bh = ext4_read_block_bitmap_nowait(sb, !! 517 bh = ext4_read_block_bitmap_nowait(sb, block_group);
601 if (IS_ERR(bh)) 518 if (IS_ERR(bh))
602 return bh; 519 return bh;
603 err = ext4_wait_block_bitmap(sb, block 520 err = ext4_wait_block_bitmap(sb, block_group, bh);
604 if (err) { 521 if (err) {
605 put_bh(bh); 522 put_bh(bh);
606 return ERR_PTR(err); 523 return ERR_PTR(err);
607 } 524 }
608 return bh; 525 return bh;
609 } 526 }
610 527
611 /** 528 /**
612 * ext4_has_free_clusters() 529 * ext4_has_free_clusters()
613 * @sbi: in-core super block structure. 530 * @sbi: in-core super block structure.
614 * @nclusters: number of needed blocks 531 * @nclusters: number of needed blocks
615 * @flags: flags from ext4_mb_new_blocks( 532 * @flags: flags from ext4_mb_new_blocks()
616 * 533 *
617 * Check if filesystem has nclusters free & av 534 * Check if filesystem has nclusters free & available for allocation.
618 * On success return 1, return 0 on failure. 535 * On success return 1, return 0 on failure.
619 */ 536 */
620 static int ext4_has_free_clusters(struct ext4_ 537 static int ext4_has_free_clusters(struct ext4_sb_info *sbi,
621 s64 ncluster 538 s64 nclusters, unsigned int flags)
622 { 539 {
623 s64 free_clusters, dirty_clusters, rsv 540 s64 free_clusters, dirty_clusters, rsv, resv_clusters;
624 struct percpu_counter *fcc = &sbi->s_f 541 struct percpu_counter *fcc = &sbi->s_freeclusters_counter;
625 struct percpu_counter *dcc = &sbi->s_d 542 struct percpu_counter *dcc = &sbi->s_dirtyclusters_counter;
626 543
627 free_clusters = percpu_counter_read_p 544 free_clusters = percpu_counter_read_positive(fcc);
628 dirty_clusters = percpu_counter_read_p 545 dirty_clusters = percpu_counter_read_positive(dcc);
629 resv_clusters = atomic64_read(&sbi->s_ 546 resv_clusters = atomic64_read(&sbi->s_resv_clusters);
630 547
631 /* 548 /*
632 * r_blocks_count should always be mul 549 * r_blocks_count should always be multiple of the cluster ratio so
633 * we are safe to do a plane bit shift 550 * we are safe to do a plane bit shift only.
634 */ 551 */
635 rsv = (ext4_r_blocks_count(sbi->s_es) 552 rsv = (ext4_r_blocks_count(sbi->s_es) >> sbi->s_cluster_bits) +
636 resv_clusters; 553 resv_clusters;
637 554
638 if (free_clusters - (nclusters + rsv + 555 if (free_clusters - (nclusters + rsv + dirty_clusters) <
639 EXT4_F 556 EXT4_FREECLUSTERS_WATERMARK) {
640 free_clusters = percpu_counte 557 free_clusters = percpu_counter_sum_positive(fcc);
641 dirty_clusters = percpu_counte 558 dirty_clusters = percpu_counter_sum_positive(dcc);
642 } 559 }
643 /* Check whether we have space after a 560 /* Check whether we have space after accounting for current
644 * dirty clusters & root reserved clus 561 * dirty clusters & root reserved clusters.
645 */ 562 */
646 if (free_clusters >= (rsv + nclusters 563 if (free_clusters >= (rsv + nclusters + dirty_clusters))
647 return 1; 564 return 1;
648 565
649 /* Hm, nope. Are (enough) root reserv 566 /* Hm, nope. Are (enough) root reserved clusters available? */
650 if (uid_eq(sbi->s_resuid, current_fsui 567 if (uid_eq(sbi->s_resuid, current_fsuid()) ||
651 (!gid_eq(sbi->s_resgid, GLOBAL_ROO 568 (!gid_eq(sbi->s_resgid, GLOBAL_ROOT_GID) && in_group_p(sbi->s_resgid)) ||
652 capable(CAP_SYS_RESOURCE) || 569 capable(CAP_SYS_RESOURCE) ||
653 (flags & EXT4_MB_USE_ROOT_BLOCKS)) 570 (flags & EXT4_MB_USE_ROOT_BLOCKS)) {
654 571
655 if (free_clusters >= (ncluster 572 if (free_clusters >= (nclusters + dirty_clusters +
656 resv_clu 573 resv_clusters))
657 return 1; 574 return 1;
658 } 575 }
659 /* No free blocks. Let's see if we can 576 /* No free blocks. Let's see if we can dip into reserved pool */
660 if (flags & EXT4_MB_USE_RESERVED) { 577 if (flags & EXT4_MB_USE_RESERVED) {
661 if (free_clusters >= (ncluster 578 if (free_clusters >= (nclusters + dirty_clusters))
662 return 1; 579 return 1;
663 } 580 }
664 581
665 return 0; 582 return 0;
666 } 583 }
667 584
668 int ext4_claim_free_clusters(struct ext4_sb_in 585 int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
669 s64 nclusters, un 586 s64 nclusters, unsigned int flags)
670 { 587 {
671 if (ext4_has_free_clusters(sbi, nclust 588 if (ext4_has_free_clusters(sbi, nclusters, flags)) {
672 percpu_counter_add(&sbi->s_dir 589 percpu_counter_add(&sbi->s_dirtyclusters_counter, nclusters);
673 return 0; 590 return 0;
674 } else 591 } else
675 return -ENOSPC; 592 return -ENOSPC;
676 } 593 }
677 594
678 /** 595 /**
679 * ext4_should_retry_alloc() - check if a bloc !! 596 * ext4_should_retry_alloc()
680 * @sb: superblock !! 597 * @sb: super block
681 * @retries: number of retry attemp !! 598 * @retries number of attemps has been made
682 * !! 599 *
683 * ext4_should_retry_alloc() is called when EN !! 600 * ext4_should_retry_alloc() is called when ENOSPC is returned, and if
684 * attempting to allocate blocks. If there's !! 601 * it is profitable to retry the operation, this function will wait
685 * journal transaction might free some space a !! 602 * for the current or committing transaction to complete, and then
686 * succeed, this function will wait for the cu !! 603 * return TRUE.
687 * to complete and then return TRUE. !! 604 *
>> 605 * if the total number of retries exceed three times, return FALSE.
688 */ 606 */
689 int ext4_should_retry_alloc(struct super_block 607 int ext4_should_retry_alloc(struct super_block *sb, int *retries)
690 { 608 {
691 struct ext4_sb_info *sbi = EXT4_SB(sb) !! 609 if (!ext4_has_free_clusters(EXT4_SB(sb), 1, 0) ||
692 !! 610 (*retries)++ > 3 ||
693 if (!sbi->s_journal) !! 611 !EXT4_SB(sb)->s_journal)
694 return 0; 612 return 0;
695 613
696 if (++(*retries) > 3) { !! 614 jbd_debug(1, "%s: retrying operation after ENOSPC\n", sb->s_id);
697 percpu_counter_inc(&sbi->s_sra <<
698 return 0; <<
699 } <<
700 615
701 /* <<
702 * if there's no indication that block <<
703 * possible we just missed a transacti <<
704 */ <<
705 smp_mb(); 616 smp_mb();
706 if (sbi->s_mb_free_pending == 0) { !! 617 if (EXT4_SB(sb)->s_mb_free_pending)
707 if (test_opt(sb, DISCARD)) { !! 618 jbd2_journal_force_commit_nested(EXT4_SB(sb)->s_journal);
708 atomic_inc(&sbi->s_ret <<
709 flush_work(&sbi->s_dis <<
710 atomic_dec(&sbi->s_ret <<
711 } <<
712 return ext4_has_free_clusters( <<
713 } <<
714 <<
715 /* <<
716 * it's possible we've just missed a t <<
717 * so ignore the returned status <<
718 */ <<
719 ext4_debug("%s: retrying operation aft <<
720 (void) jbd2_journal_force_commit_neste <<
721 return 1; 619 return 1;
722 } 620 }
723 621
724 /* 622 /*
725 * ext4_new_meta_blocks() -- allocate block fo 623 * ext4_new_meta_blocks() -- allocate block for meta data (indexing) blocks
726 * 624 *
727 * @handle: handle to this transac 625 * @handle: handle to this transaction
728 * @inode: file inode 626 * @inode: file inode
729 * @goal: given target block(fil 627 * @goal: given target block(filesystem wide)
730 * @count: pointer to total numbe 628 * @count: pointer to total number of clusters needed
731 * @errp: error code 629 * @errp: error code
732 * 630 *
733 * Return 1st allocated block number on succes 631 * Return 1st allocated block number on success, *count stores total account
734 * error stores in errp pointer 632 * error stores in errp pointer
735 */ 633 */
736 ext4_fsblk_t ext4_new_meta_blocks(handle_t *ha 634 ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
737 ext4_fsblk_t 635 ext4_fsblk_t goal, unsigned int flags,
738 unsigned lon 636 unsigned long *count, int *errp)
739 { 637 {
740 struct ext4_allocation_request ar; 638 struct ext4_allocation_request ar;
741 ext4_fsblk_t ret; 639 ext4_fsblk_t ret;
742 640
743 memset(&ar, 0, sizeof(ar)); 641 memset(&ar, 0, sizeof(ar));
744 /* Fill with neighbour allocated block 642 /* Fill with neighbour allocated blocks */
745 ar.inode = inode; 643 ar.inode = inode;
746 ar.goal = goal; 644 ar.goal = goal;
747 ar.len = count ? *count : 1; 645 ar.len = count ? *count : 1;
748 ar.flags = flags; 646 ar.flags = flags;
749 647
750 ret = ext4_mb_new_blocks(handle, &ar, 648 ret = ext4_mb_new_blocks(handle, &ar, errp);
751 if (count) 649 if (count)
752 *count = ar.len; 650 *count = ar.len;
753 /* 651 /*
754 * Account for the allocated meta bloc 652 * Account for the allocated meta blocks. We will never
755 * fail EDQUOT for metdata, but we do 653 * fail EDQUOT for metdata, but we do account for it.
756 */ 654 */
757 if (!(*errp) && (flags & EXT4_MB_DELAL 655 if (!(*errp) && (flags & EXT4_MB_DELALLOC_RESERVED)) {
758 dquot_alloc_block_nofail(inode 656 dquot_alloc_block_nofail(inode,
759 EXT4_C2B(EXT4_ 657 EXT4_C2B(EXT4_SB(inode->i_sb), ar.len));
760 } 658 }
761 return ret; 659 return ret;
762 } 660 }
763 661
764 /** 662 /**
765 * ext4_count_free_clusters() -- count filesys 663 * ext4_count_free_clusters() -- count filesystem free clusters
766 * @sb: superblock 664 * @sb: superblock
767 * 665 *
768 * Adds up the number of free clusters from ea 666 * Adds up the number of free clusters from each block group.
769 */ 667 */
770 ext4_fsblk_t ext4_count_free_clusters(struct s 668 ext4_fsblk_t ext4_count_free_clusters(struct super_block *sb)
771 { 669 {
772 ext4_fsblk_t desc_count; 670 ext4_fsblk_t desc_count;
773 struct ext4_group_desc *gdp; 671 struct ext4_group_desc *gdp;
774 ext4_group_t i; 672 ext4_group_t i;
775 ext4_group_t ngroups = ext4_get_groups 673 ext4_group_t ngroups = ext4_get_groups_count(sb);
776 struct ext4_group_info *grp; 674 struct ext4_group_info *grp;
777 #ifdef EXT4FS_DEBUG 675 #ifdef EXT4FS_DEBUG
778 struct ext4_super_block *es; 676 struct ext4_super_block *es;
779 ext4_fsblk_t bitmap_count; 677 ext4_fsblk_t bitmap_count;
780 unsigned int x; 678 unsigned int x;
781 struct buffer_head *bitmap_bh = NULL; 679 struct buffer_head *bitmap_bh = NULL;
782 680
783 es = EXT4_SB(sb)->s_es; 681 es = EXT4_SB(sb)->s_es;
784 desc_count = 0; 682 desc_count = 0;
785 bitmap_count = 0; 683 bitmap_count = 0;
786 gdp = NULL; 684 gdp = NULL;
787 685
788 for (i = 0; i < ngroups; i++) { 686 for (i = 0; i < ngroups; i++) {
789 gdp = ext4_get_group_desc(sb, 687 gdp = ext4_get_group_desc(sb, i, NULL);
790 if (!gdp) 688 if (!gdp)
791 continue; 689 continue;
792 grp = NULL; 690 grp = NULL;
793 if (EXT4_SB(sb)->s_group_info) 691 if (EXT4_SB(sb)->s_group_info)
794 grp = ext4_get_group_i 692 grp = ext4_get_group_info(sb, i);
795 if (!grp || !EXT4_MB_GRP_BBITM 693 if (!grp || !EXT4_MB_GRP_BBITMAP_CORRUPT(grp))
796 desc_count += ext4_fre 694 desc_count += ext4_free_group_clusters(sb, gdp);
797 brelse(bitmap_bh); 695 brelse(bitmap_bh);
798 bitmap_bh = ext4_read_block_bi 696 bitmap_bh = ext4_read_block_bitmap(sb, i);
799 if (IS_ERR(bitmap_bh)) { 697 if (IS_ERR(bitmap_bh)) {
800 bitmap_bh = NULL; 698 bitmap_bh = NULL;
801 continue; 699 continue;
802 } 700 }
803 701
804 x = ext4_count_free(bitmap_bh- 702 x = ext4_count_free(bitmap_bh->b_data,
805 EXT4_CLUST 703 EXT4_CLUSTERS_PER_GROUP(sb) / 8);
806 printk(KERN_DEBUG "group %u: s 704 printk(KERN_DEBUG "group %u: stored = %d, counted = %u\n",
807 i, ext4_free_group_clu 705 i, ext4_free_group_clusters(sb, gdp), x);
808 bitmap_count += x; 706 bitmap_count += x;
809 } 707 }
810 brelse(bitmap_bh); 708 brelse(bitmap_bh);
811 printk(KERN_DEBUG "ext4_count_free_clu 709 printk(KERN_DEBUG "ext4_count_free_clusters: stored = %llu"
812 ", computed = %llu, %llu\n", 710 ", computed = %llu, %llu\n",
813 EXT4_NUM_B2C(EXT4_SB(sb), ext4_ 711 EXT4_NUM_B2C(EXT4_SB(sb), ext4_free_blocks_count(es)),
814 desc_count, bitmap_count); 712 desc_count, bitmap_count);
815 return bitmap_count; 713 return bitmap_count;
816 #else 714 #else
817 desc_count = 0; 715 desc_count = 0;
818 for (i = 0; i < ngroups; i++) { 716 for (i = 0; i < ngroups; i++) {
819 gdp = ext4_get_group_desc(sb, 717 gdp = ext4_get_group_desc(sb, i, NULL);
820 if (!gdp) 718 if (!gdp)
821 continue; 719 continue;
822 grp = NULL; 720 grp = NULL;
823 if (EXT4_SB(sb)->s_group_info) 721 if (EXT4_SB(sb)->s_group_info)
824 grp = ext4_get_group_i 722 grp = ext4_get_group_info(sb, i);
825 if (!grp || !EXT4_MB_GRP_BBITM 723 if (!grp || !EXT4_MB_GRP_BBITMAP_CORRUPT(grp))
826 desc_count += ext4_fre 724 desc_count += ext4_free_group_clusters(sb, gdp);
827 } 725 }
828 726
829 return desc_count; 727 return desc_count;
830 #endif 728 #endif
831 } 729 }
832 730
833 static inline int test_root(ext4_group_t a, in 731 static inline int test_root(ext4_group_t a, int b)
834 { 732 {
835 while (1) { 733 while (1) {
836 if (a < b) 734 if (a < b)
837 return 0; 735 return 0;
838 if (a == b) 736 if (a == b)
839 return 1; 737 return 1;
840 if ((a % b) != 0) 738 if ((a % b) != 0)
841 return 0; 739 return 0;
842 a = a / b; 740 a = a / b;
843 } 741 }
844 } 742 }
845 743
846 /** 744 /**
847 * ext4_bg_has_super - number of blocks u 745 * ext4_bg_has_super - number of blocks used by the superblock in group
848 * @sb: superblock for filesystem 746 * @sb: superblock for filesystem
849 * @group: group number to check 747 * @group: group number to check
850 * 748 *
851 * Return the number of blocks used by th 749 * Return the number of blocks used by the superblock (primary or backup)
852 * in this group. Currently this will be 750 * in this group. Currently this will be only 0 or 1.
853 */ 751 */
854 int ext4_bg_has_super(struct super_block *sb, 752 int ext4_bg_has_super(struct super_block *sb, ext4_group_t group)
855 { 753 {
856 struct ext4_super_block *es = EXT4_SB( 754 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
857 755
858 if (group == 0) 756 if (group == 0)
859 return 1; 757 return 1;
860 if (ext4_has_feature_sparse_super2(sb) 758 if (ext4_has_feature_sparse_super2(sb)) {
861 if (group == le32_to_cpu(es->s 759 if (group == le32_to_cpu(es->s_backup_bgs[0]) ||
862 group == le32_to_cpu(es->s 760 group == le32_to_cpu(es->s_backup_bgs[1]))
863 return 1; 761 return 1;
864 return 0; 762 return 0;
865 } 763 }
866 if ((group <= 1) || !ext4_has_feature_ 764 if ((group <= 1) || !ext4_has_feature_sparse_super(sb))
867 return 1; 765 return 1;
868 if (!(group & 1)) 766 if (!(group & 1))
869 return 0; 767 return 0;
870 if (test_root(group, 3) || (test_root( 768 if (test_root(group, 3) || (test_root(group, 5)) ||
871 test_root(group, 7)) 769 test_root(group, 7))
872 return 1; 770 return 1;
873 771
874 return 0; 772 return 0;
875 } 773 }
876 774
877 static unsigned long ext4_bg_num_gdb_meta(stru 775 static unsigned long ext4_bg_num_gdb_meta(struct super_block *sb,
878 ext4_g 776 ext4_group_t group)
879 { 777 {
880 unsigned long metagroup = group / EXT4 778 unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
881 ext4_group_t first = metagroup * EXT4_ 779 ext4_group_t first = metagroup * EXT4_DESC_PER_BLOCK(sb);
882 ext4_group_t last = first + EXT4_DESC_ 780 ext4_group_t last = first + EXT4_DESC_PER_BLOCK(sb) - 1;
883 781
884 if (group == first || group == first + 782 if (group == first || group == first + 1 || group == last)
885 return 1; 783 return 1;
886 return 0; 784 return 0;
887 } 785 }
888 786
889 static unsigned long ext4_bg_num_gdb_nometa(st 787 static unsigned long ext4_bg_num_gdb_nometa(struct super_block *sb,
890 ext4_g 788 ext4_group_t group)
891 { 789 {
892 if (!ext4_bg_has_super(sb, group)) 790 if (!ext4_bg_has_super(sb, group))
893 return 0; 791 return 0;
894 792
895 if (ext4_has_feature_meta_bg(sb)) 793 if (ext4_has_feature_meta_bg(sb))
896 return le32_to_cpu(EXT4_SB(sb) 794 return le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg);
897 else 795 else
898 return EXT4_SB(sb)->s_gdb_coun 796 return EXT4_SB(sb)->s_gdb_count;
899 } 797 }
900 798
901 /** 799 /**
902 * ext4_bg_num_gdb - number of blocks use 800 * ext4_bg_num_gdb - number of blocks used by the group table in group
903 * @sb: superblock for filesystem 801 * @sb: superblock for filesystem
904 * @group: group number to check 802 * @group: group number to check
905 * 803 *
906 * Return the number of blocks used by th 804 * Return the number of blocks used by the group descriptor table
907 * (primary or backup) in this group. In 805 * (primary or backup) in this group. In the future there may be a
908 * different number of descriptor blocks 806 * different number of descriptor blocks in each group.
909 */ 807 */
910 unsigned long ext4_bg_num_gdb(struct super_blo 808 unsigned long ext4_bg_num_gdb(struct super_block *sb, ext4_group_t group)
911 { 809 {
912 unsigned long first_meta_bg = 810 unsigned long first_meta_bg =
913 le32_to_cpu(EXT4_SB(sb 811 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg);
914 unsigned long metagroup = group / EXT4 812 unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
915 813
916 if (!ext4_has_feature_meta_bg(sb) || m 814 if (!ext4_has_feature_meta_bg(sb) || metagroup < first_meta_bg)
917 return ext4_bg_num_gdb_nometa( 815 return ext4_bg_num_gdb_nometa(sb, group);
918 816
919 return ext4_bg_num_gdb_meta(sb,group); 817 return ext4_bg_num_gdb_meta(sb,group);
920 818
921 } 819 }
922 820
923 /* 821 /*
924 * This function returns the number of file sy !! 822 * This function returns the number of file system metadata clusters at
925 * the beginning of a block group, including t 823 * the beginning of a block group, including the reserved gdt blocks.
926 */ 824 */
927 unsigned int ext4_num_base_meta_blocks(struct !! 825 static unsigned ext4_num_base_meta_clusters(struct super_block *sb,
928 ext4_gr !! 826 ext4_group_t block_group)
929 { 827 {
930 struct ext4_sb_info *sbi = EXT4_SB(sb) 828 struct ext4_sb_info *sbi = EXT4_SB(sb);
931 unsigned num; 829 unsigned num;
932 830
933 /* Check for superblock and gdt backup 831 /* Check for superblock and gdt backups in this group */
934 num = ext4_bg_has_super(sb, block_grou 832 num = ext4_bg_has_super(sb, block_group);
935 833
936 if (!ext4_has_feature_meta_bg(sb) || 834 if (!ext4_has_feature_meta_bg(sb) ||
937 block_group < le32_to_cpu(sbi->s_e 835 block_group < le32_to_cpu(sbi->s_es->s_first_meta_bg) *
938 sbi->s_desc_per_bloc 836 sbi->s_desc_per_block) {
939 if (num) { 837 if (num) {
940 num += ext4_bg_num_gdb !! 838 num += ext4_bg_num_gdb(sb, block_group);
941 num += le16_to_cpu(sbi 839 num += le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks);
942 } 840 }
943 } else { /* For META_BG_BLOCK_GROUPS * 841 } else { /* For META_BG_BLOCK_GROUPS */
944 num += ext4_bg_num_gdb_meta(sb !! 842 num += ext4_bg_num_gdb(sb, block_group);
945 } 843 }
946 return num; !! 844 return EXT4_NUM_B2C(sbi, num);
947 } 845 }
948 <<
949 static unsigned int ext4_num_base_meta_cluster <<
950 <<
951 { <<
952 return EXT4_NUM_B2C(EXT4_SB(sb), ext4_ <<
953 } <<
954 <<
955 /** 846 /**
956 * ext4_inode_to_goal_block - return a hi 847 * ext4_inode_to_goal_block - return a hint for block allocation
957 * @inode: inode for block allocation 848 * @inode: inode for block allocation
958 * 849 *
959 * Return the ideal location to start all 850 * Return the ideal location to start allocating blocks for a
960 * newly created inode. 851 * newly created inode.
961 */ 852 */
962 ext4_fsblk_t ext4_inode_to_goal_block(struct i 853 ext4_fsblk_t ext4_inode_to_goal_block(struct inode *inode)
963 { 854 {
964 struct ext4_inode_info *ei = EXT4_I(in 855 struct ext4_inode_info *ei = EXT4_I(inode);
965 ext4_group_t block_group; 856 ext4_group_t block_group;
966 ext4_grpblk_t colour; 857 ext4_grpblk_t colour;
967 int flex_size = ext4_flex_bg_size(EXT4 858 int flex_size = ext4_flex_bg_size(EXT4_SB(inode->i_sb));
968 ext4_fsblk_t bg_start; 859 ext4_fsblk_t bg_start;
969 ext4_fsblk_t last_block; 860 ext4_fsblk_t last_block;
970 861
971 block_group = ei->i_block_group; 862 block_group = ei->i_block_group;
972 if (flex_size >= EXT4_FLEX_SIZE_DIR_AL 863 if (flex_size >= EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME) {
973 /* 864 /*
974 * If there are at least EXT4_ 865 * If there are at least EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME
975 * block groups per flexgroup, 866 * block groups per flexgroup, reserve the first block
976 * group for directories and s 867 * group for directories and special files. Regular
977 * files will start at the sec 868 * files will start at the second block group. This
978 * tends to speed up directory 869 * tends to speed up directory access and improves
979 * fsck times. 870 * fsck times.
980 */ 871 */
981 block_group &= ~(flex_size-1); 872 block_group &= ~(flex_size-1);
982 if (S_ISREG(inode->i_mode)) 873 if (S_ISREG(inode->i_mode))
983 block_group++; 874 block_group++;
984 } 875 }
985 bg_start = ext4_group_first_block_no(i 876 bg_start = ext4_group_first_block_no(inode->i_sb, block_group);
986 last_block = ext4_blocks_count(EXT4_SB 877 last_block = ext4_blocks_count(EXT4_SB(inode->i_sb)->s_es) - 1;
987 878
988 /* 879 /*
989 * If we are doing delayed allocation, 880 * If we are doing delayed allocation, we don't need take
990 * colour into account. 881 * colour into account.
991 */ 882 */
992 if (test_opt(inode->i_sb, DELALLOC)) 883 if (test_opt(inode->i_sb, DELALLOC))
993 return bg_start; 884 return bg_start;
994 885
995 if (bg_start + EXT4_BLOCKS_PER_GROUP(i 886 if (bg_start + EXT4_BLOCKS_PER_GROUP(inode->i_sb) <= last_block)
996 colour = (task_pid_nr(current) !! 887 colour = (current->pid % 16) *
997 (EXT4_BLOCKS_PER_GROUP 888 (EXT4_BLOCKS_PER_GROUP(inode->i_sb) / 16);
998 else 889 else
999 colour = (task_pid_nr(current) !! 890 colour = (current->pid % 16) * ((last_block - bg_start) / 16);
1000 ((last_block - bg_sta <<
1001 return bg_start + colour; 891 return bg_start + colour;
1002 } 892 }
1003 893
1004 894
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.