1 // SPDX-License-Identifier: GPL-2.0 1 // SPDX-License-Identifier: GPL-2.0
2 /* 2 /*
3 * linux/fs/ext4/balloc.c 3 * linux/fs/ext4/balloc.c
4 * 4 *
5 * Copyright (C) 1992, 1993, 1994, 1995 5 * Copyright (C) 1992, 1993, 1994, 1995
6 * Remy Card (card@masi.ibp.fr) 6 * Remy Card (card@masi.ibp.fr)
7 * Laboratoire MASI - Institut Blaise Pascal 7 * Laboratoire MASI - Institut Blaise Pascal
8 * Universite Pierre et Marie Curie (Paris VI) 8 * Universite Pierre et Marie Curie (Paris VI)
9 * 9 *
10 * Enhanced block allocation by Stephen Tweed 10 * Enhanced block allocation by Stephen Tweedie (sct@redhat.com), 1993
11 * Big-endian to little-endian byte-swapping/ 11 * Big-endian to little-endian byte-swapping/bitmaps by
12 * David S. Miller (davem@caip.rutgers. 12 * David S. Miller (davem@caip.rutgers.edu), 1995
13 */ 13 */
14 14
15 #include <linux/time.h> 15 #include <linux/time.h>
16 #include <linux/capability.h> 16 #include <linux/capability.h>
17 #include <linux/fs.h> 17 #include <linux/fs.h>
18 #include <linux/quotaops.h> 18 #include <linux/quotaops.h>
19 #include <linux/buffer_head.h> 19 #include <linux/buffer_head.h>
20 #include "ext4.h" 20 #include "ext4.h"
21 #include "ext4_jbd2.h" 21 #include "ext4_jbd2.h"
22 #include "mballoc.h" 22 #include "mballoc.h"
23 23
24 #include <trace/events/ext4.h> 24 #include <trace/events/ext4.h>
25 #include <kunit/static_stub.h> <<
26 25
27 static unsigned ext4_num_base_meta_clusters(st 26 static unsigned ext4_num_base_meta_clusters(struct super_block *sb,
28 ex 27 ext4_group_t block_group);
29 /* 28 /*
30 * balloc.c contains the blocks allocation and 29 * balloc.c contains the blocks allocation and deallocation routines
31 */ 30 */
32 31
33 /* 32 /*
34 * Calculate block group number for a given bl 33 * Calculate block group number for a given block number
35 */ 34 */
36 ext4_group_t ext4_get_group_number(struct supe 35 ext4_group_t ext4_get_group_number(struct super_block *sb,
37 ext4_fsblk_ 36 ext4_fsblk_t block)
38 { 37 {
39 ext4_group_t group; 38 ext4_group_t group;
40 39
41 if (test_opt2(sb, STD_GROUP_SIZE)) 40 if (test_opt2(sb, STD_GROUP_SIZE))
42 group = (block - 41 group = (block -
43 le32_to_cpu(EXT4_SB(s 42 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block)) >>
44 (EXT4_BLOCK_SIZE_BITS( 43 (EXT4_BLOCK_SIZE_BITS(sb) + EXT4_CLUSTER_BITS(sb) + 3);
45 else 44 else
46 ext4_get_group_no_and_offset(s 45 ext4_get_group_no_and_offset(sb, block, &group, NULL);
47 return group; 46 return group;
48 } 47 }
49 48
50 /* 49 /*
51 * Calculate the block group number and offset 50 * Calculate the block group number and offset into the block/cluster
52 * allocation bitmap, given a block number 51 * allocation bitmap, given a block number
53 */ 52 */
54 void ext4_get_group_no_and_offset(struct super 53 void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
55 ext4_group_t *blockgrpp, ext4_ 54 ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp)
56 { 55 {
57 struct ext4_super_block *es = EXT4_SB( 56 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
58 ext4_grpblk_t offset; 57 ext4_grpblk_t offset;
59 58
60 blocknr = blocknr - le32_to_cpu(es->s_ 59 blocknr = blocknr - le32_to_cpu(es->s_first_data_block);
61 offset = do_div(blocknr, EXT4_BLOCKS_P 60 offset = do_div(blocknr, EXT4_BLOCKS_PER_GROUP(sb)) >>
62 EXT4_SB(sb)->s_cluster_bits; 61 EXT4_SB(sb)->s_cluster_bits;
63 if (offsetp) 62 if (offsetp)
64 *offsetp = offset; 63 *offsetp = offset;
65 if (blockgrpp) 64 if (blockgrpp)
66 *blockgrpp = blocknr; 65 *blockgrpp = blocknr;
67 66
68 } 67 }
69 68
70 /* 69 /*
71 * Check whether the 'block' lives within the 70 * Check whether the 'block' lives within the 'block_group'. Returns 1 if so
72 * and 0 otherwise. 71 * and 0 otherwise.
73 */ 72 */
74 static inline int ext4_block_in_group(struct s 73 static inline int ext4_block_in_group(struct super_block *sb,
75 ext4_fsb 74 ext4_fsblk_t block,
76 ext4_gro 75 ext4_group_t block_group)
77 { 76 {
78 ext4_group_t actual_group; 77 ext4_group_t actual_group;
79 78
80 actual_group = ext4_get_group_number(s 79 actual_group = ext4_get_group_number(sb, block);
81 return (actual_group == block_group) ? 80 return (actual_group == block_group) ? 1 : 0;
82 } 81 }
83 82
84 /* !! 83 /* 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 84 * represents the overhead needed by the file system.
87 */ 85 */
88 static unsigned ext4_num_overhead_clusters(str 86 static unsigned ext4_num_overhead_clusters(struct super_block *sb,
89 ext 87 ext4_group_t block_group,
90 str 88 struct ext4_group_desc *gdp)
91 { 89 {
92 unsigned base_clusters, num_clusters; !! 90 unsigned num_clusters;
93 int block_cluster = -1, inode_cluster; !! 91 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_ 92 ext4_fsblk_t start = ext4_group_first_block_no(sb, block_group);
96 ext4_fsblk_t end = start + EXT4_BLOCKS !! 93 ext4_fsblk_t itbl_blk;
97 ext4_fsblk_t itbl_blk_start, itbl_blk_ <<
98 struct ext4_sb_info *sbi = EXT4_SB(sb) 94 struct ext4_sb_info *sbi = EXT4_SB(sb);
99 95
100 /* This is the number of clusters used 96 /* This is the number of clusters used by the superblock,
101 * block group descriptors, and reserv 97 * block group descriptors, and reserved block group
102 * descriptor blocks */ 98 * descriptor blocks */
103 base_clusters = ext4_num_base_meta_clu !! 99 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 } <<
126 100
127 /* 101 /*
128 * For the allocation bitmaps, we firs !! 102 * For the allocation bitmaps and inode table, we first need
129 * if the block is in the block group. !! 103 * to check to see if the block is in the block group. If it
130 * to see if the cluster is already ac !! 104 * is, then check to see if the cluster is already accounted
131 * used for the base metadata cluster !! 105 * for in the clusters used for the base metadata cluster, or
>> 106 * if we can increment the base metadata cluster to include
>> 107 * that block. Otherwise, we will have to track the cluster
>> 108 * used for the allocation bitmap or inode table explicitly.
132 * Normally all of these blocks are co 109 * Normally all of these blocks are contiguous, so the special
133 * case handling shouldn't be necessar 110 * case handling shouldn't be necessary except for *very*
134 * unusual file system layouts. 111 * unusual file system layouts.
135 */ 112 */
136 if (ext4_block_in_group(sb, ext4_block 113 if (ext4_block_in_group(sb, ext4_block_bitmap(sb, gdp), block_group)) {
137 block_cluster = EXT4_B2C(sbi, 114 block_cluster = EXT4_B2C(sbi,
138 ext4_ 115 ext4_block_bitmap(sb, gdp) - start);
139 if (block_cluster >= base_clus !! 116 if (block_cluster < num_clusters)
140 (block_cluster < itbl_clus !! 117 block_cluster = -1;
141 block_cluster > itbl_clust !! 118 else if (block_cluster == num_clusters) {
142 num_clusters++; 119 num_clusters++;
>> 120 block_cluster = -1;
>> 121 }
143 } 122 }
144 123
145 if (ext4_block_in_group(sb, ext4_inode 124 if (ext4_block_in_group(sb, ext4_inode_bitmap(sb, gdp), block_group)) {
146 inode_cluster = EXT4_B2C(sbi, 125 inode_cluster = EXT4_B2C(sbi,
147 ext4_ 126 ext4_inode_bitmap(sb, gdp) - start);
148 /* !! 127 if (inode_cluster < num_clusters)
149 * Additional check if inode b !! 128 inode_cluster = -1;
150 * block_cluster !! 129 else if (inode_cluster == num_clusters) {
151 */ <<
152 if (inode_cluster != block_clu <<
153 inode_cluster >= base_clus <<
154 (inode_cluster < itbl_clus <<
155 inode_cluster > itbl_clust <<
156 num_clusters++; 130 num_clusters++;
>> 131 inode_cluster = -1;
>> 132 }
>> 133 }
>> 134
>> 135 itbl_blk = ext4_inode_table(sb, gdp);
>> 136 for (i = 0; i < sbi->s_itb_per_group; i++) {
>> 137 if (ext4_block_in_group(sb, itbl_blk + i, block_group)) {
>> 138 c = EXT4_B2C(sbi, itbl_blk + i - start);
>> 139 if ((c < num_clusters) || (c == inode_cluster) ||
>> 140 (c == block_cluster) || (c == itbl_cluster))
>> 141 continue;
>> 142 if (c == num_clusters) {
>> 143 num_clusters++;
>> 144 continue;
>> 145 }
>> 146 num_clusters++;
>> 147 itbl_cluster = c;
>> 148 }
157 } 149 }
158 150
>> 151 if (block_cluster != -1)
>> 152 num_clusters++;
>> 153 if (inode_cluster != -1)
>> 154 num_clusters++;
>> 155
159 return num_clusters; 156 return num_clusters;
160 } 157 }
161 158
162 static unsigned int num_clusters_in_group(stru 159 static unsigned int num_clusters_in_group(struct super_block *sb,
163 ext4 160 ext4_group_t block_group)
164 { 161 {
165 unsigned int blocks; 162 unsigned int blocks;
166 163
167 if (block_group == ext4_get_groups_cou 164 if (block_group == ext4_get_groups_count(sb) - 1) {
168 /* 165 /*
169 * Even though mke2fs always i 166 * Even though mke2fs always initializes the first and
170 * last group, just in case so 167 * last group, just in case some other tool was used,
171 * we need to make sure we cal 168 * we need to make sure we calculate the right free
172 * blocks. 169 * blocks.
173 */ 170 */
174 blocks = ext4_blocks_count(EXT 171 blocks = ext4_blocks_count(EXT4_SB(sb)->s_es) -
175 ext4_group_first_block 172 ext4_group_first_block_no(sb, block_group);
176 } else 173 } else
177 blocks = EXT4_BLOCKS_PER_GROUP 174 blocks = EXT4_BLOCKS_PER_GROUP(sb);
178 return EXT4_NUM_B2C(EXT4_SB(sb), block 175 return EXT4_NUM_B2C(EXT4_SB(sb), blocks);
179 } 176 }
180 177
181 /* Initializes an uninitialized block bitmap * 178 /* Initializes an uninitialized block bitmap */
182 static int ext4_init_block_bitmap(struct super 179 static int ext4_init_block_bitmap(struct super_block *sb,
183 struct buff 180 struct buffer_head *bh,
184 ext4_group_ 181 ext4_group_t block_group,
185 struct ext4 182 struct ext4_group_desc *gdp)
186 { 183 {
187 unsigned int bit, bit_max; 184 unsigned int bit, bit_max;
188 struct ext4_sb_info *sbi = EXT4_SB(sb) 185 struct ext4_sb_info *sbi = EXT4_SB(sb);
189 ext4_fsblk_t start, tmp; 186 ext4_fsblk_t start, tmp;
190 187
191 ASSERT(buffer_locked(bh)); 188 ASSERT(buffer_locked(bh));
192 189
>> 190 /* If checksum is bad mark all blocks used to prevent allocation
>> 191 * essentially implementing a per-group read-only flag. */
193 if (!ext4_group_desc_csum_verify(sb, b 192 if (!ext4_group_desc_csum_verify(sb, block_group, gdp)) {
194 ext4_mark_group_bitmap_corrupt 193 ext4_mark_group_bitmap_corrupted(sb, block_group,
195 EXT4_G 194 EXT4_GROUP_INFO_BBITMAP_CORRUPT |
196 EXT4_G 195 EXT4_GROUP_INFO_IBITMAP_CORRUPT);
197 return -EFSBADCRC; 196 return -EFSBADCRC;
198 } 197 }
199 memset(bh->b_data, 0, sb->s_blocksize) 198 memset(bh->b_data, 0, sb->s_blocksize);
200 199
201 bit_max = ext4_num_base_meta_clusters( 200 bit_max = ext4_num_base_meta_clusters(sb, block_group);
202 if ((bit_max >> 3) >= bh->b_size) 201 if ((bit_max >> 3) >= bh->b_size)
203 return -EFSCORRUPTED; 202 return -EFSCORRUPTED;
204 203
205 for (bit = 0; bit < bit_max; bit++) 204 for (bit = 0; bit < bit_max; bit++)
206 ext4_set_bit(bit, bh->b_data); 205 ext4_set_bit(bit, bh->b_data);
207 206
208 start = ext4_group_first_block_no(sb, 207 start = ext4_group_first_block_no(sb, block_group);
209 208
210 /* Set bits for block and inode bitmap 209 /* Set bits for block and inode bitmaps, and inode table */
211 tmp = ext4_block_bitmap(sb, gdp); 210 tmp = ext4_block_bitmap(sb, gdp);
212 if (ext4_block_in_group(sb, tmp, block 211 if (ext4_block_in_group(sb, tmp, block_group))
213 ext4_set_bit(EXT4_B2C(sbi, tmp 212 ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data);
214 213
215 tmp = ext4_inode_bitmap(sb, gdp); 214 tmp = ext4_inode_bitmap(sb, gdp);
216 if (ext4_block_in_group(sb, tmp, block 215 if (ext4_block_in_group(sb, tmp, block_group))
217 ext4_set_bit(EXT4_B2C(sbi, tmp 216 ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data);
218 217
219 tmp = ext4_inode_table(sb, gdp); 218 tmp = ext4_inode_table(sb, gdp);
220 for (; tmp < ext4_inode_table(sb, gdp) 219 for (; tmp < ext4_inode_table(sb, gdp) +
221 sbi->s_itb_per_group; tmp 220 sbi->s_itb_per_group; tmp++) {
222 if (ext4_block_in_group(sb, tm 221 if (ext4_block_in_group(sb, tmp, block_group))
223 ext4_set_bit(EXT4_B2C( 222 ext4_set_bit(EXT4_B2C(sbi, tmp - start), bh->b_data);
224 } 223 }
225 224
226 /* 225 /*
227 * Also if the number of blocks within 226 * Also if the number of blocks within the group is less than
228 * the blocksize * 8 ( which is the si 227 * the blocksize * 8 ( which is the size of bitmap ), set rest
229 * of the block bitmap to 1 228 * of the block bitmap to 1
230 */ 229 */
231 ext4_mark_bitmap_end(num_clusters_in_g 230 ext4_mark_bitmap_end(num_clusters_in_group(sb, block_group),
232 sb->s_blocksize * 231 sb->s_blocksize * 8, bh->b_data);
233 return 0; 232 return 0;
234 } 233 }
235 234
236 /* Return the number of free blocks in a block 235 /* Return the number of free blocks in a block group. It is used when
237 * the block bitmap is uninitialized, so we ca 236 * the block bitmap is uninitialized, so we can't just count the bits
238 * in the bitmap. */ 237 * in the bitmap. */
239 unsigned ext4_free_clusters_after_init(struct 238 unsigned ext4_free_clusters_after_init(struct super_block *sb,
240 ext4_gr 239 ext4_group_t block_group,
241 struct 240 struct ext4_group_desc *gdp)
242 { 241 {
243 return num_clusters_in_group(sb, block 242 return num_clusters_in_group(sb, block_group) -
244 ext4_num_overhead_clusters(sb, 243 ext4_num_overhead_clusters(sb, block_group, gdp);
245 } 244 }
246 245
247 /* 246 /*
248 * The free blocks are managed by bitmaps. A 247 * The free blocks are managed by bitmaps. A file system contains several
249 * blocks groups. Each group contains 1 bitma 248 * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap
250 * block for inodes, N blocks for the inode ta 249 * block for inodes, N blocks for the inode table and data blocks.
251 * 250 *
252 * The file system contains group descriptors 251 * The file system contains group descriptors which are located after the
253 * super block. Each descriptor contains the 252 * super block. Each descriptor contains the number of the bitmap block and
254 * the free blocks count in the block. The de 253 * the free blocks count in the block. The descriptors are loaded in memory
255 * when a file system is mounted (see ext4_fil 254 * when a file system is mounted (see ext4_fill_super).
256 */ 255 */
257 256
258 /** 257 /**
259 * ext4_get_group_desc() -- load group descrip 258 * ext4_get_group_desc() -- load group descriptor from disk
260 * @sb: super block 259 * @sb: super block
261 * @block_group: given block group 260 * @block_group: given block group
262 * @bh: pointer to the buffer 261 * @bh: pointer to the buffer head to store the block
263 * group descriptor 262 * group descriptor
264 */ 263 */
265 struct ext4_group_desc * ext4_get_group_desc(s 264 struct ext4_group_desc * ext4_get_group_desc(struct super_block *sb,
266 e 265 ext4_group_t block_group,
267 s 266 struct buffer_head **bh)
268 { 267 {
269 unsigned int group_desc; 268 unsigned int group_desc;
270 unsigned int offset; 269 unsigned int offset;
271 ext4_group_t ngroups = ext4_get_groups 270 ext4_group_t ngroups = ext4_get_groups_count(sb);
272 struct ext4_group_desc *desc; 271 struct ext4_group_desc *desc;
273 struct ext4_sb_info *sbi = EXT4_SB(sb) 272 struct ext4_sb_info *sbi = EXT4_SB(sb);
274 struct buffer_head *bh_p; 273 struct buffer_head *bh_p;
275 274
276 KUNIT_STATIC_STUB_REDIRECT(ext4_get_gr <<
277 sb, block_g <<
278 <<
279 if (block_group >= ngroups) { 275 if (block_group >= ngroups) {
280 ext4_error(sb, "block_group >= 276 ext4_error(sb, "block_group >= groups_count - block_group = %u,"
281 " groups_count = %u 277 " groups_count = %u", block_group, ngroups);
282 278
283 return NULL; 279 return NULL;
284 } 280 }
285 281
286 group_desc = block_group >> EXT4_DESC_ 282 group_desc = block_group >> EXT4_DESC_PER_BLOCK_BITS(sb);
287 offset = block_group & (EXT4_DESC_PER_ 283 offset = block_group & (EXT4_DESC_PER_BLOCK(sb) - 1);
288 bh_p = sbi_array_rcu_deref(sbi, s_grou 284 bh_p = sbi_array_rcu_deref(sbi, s_group_desc, group_desc);
289 /* 285 /*
290 * sbi_array_rcu_deref returns with rc 286 * sbi_array_rcu_deref returns with rcu unlocked, this is ok since
291 * the pointer being dereferenced won' 287 * the pointer being dereferenced won't be dereferenced again. By
292 * looking at the usage in add_new_gdb 288 * looking at the usage in add_new_gdb() the value isn't modified,
293 * just the pointer, and so it remains 289 * just the pointer, and so it remains valid.
294 */ 290 */
295 if (!bh_p) { 291 if (!bh_p) {
296 ext4_error(sb, "Group descript 292 ext4_error(sb, "Group descriptor not loaded - "
297 "block_group = %u, 293 "block_group = %u, group_desc = %u, desc = %u",
298 block_group, group_ 294 block_group, group_desc, offset);
299 return NULL; 295 return NULL;
300 } 296 }
301 297
302 desc = (struct ext4_group_desc *)( 298 desc = (struct ext4_group_desc *)(
303 (__u8 *)bh_p->b_data + 299 (__u8 *)bh_p->b_data +
304 offset * EXT4_DESC_SIZE(sb)); 300 offset * EXT4_DESC_SIZE(sb));
305 if (bh) 301 if (bh)
306 *bh = bh_p; 302 *bh = bh_p;
307 return desc; 303 return desc;
308 } 304 }
309 305
310 static ext4_fsblk_t ext4_valid_block_bitmap_pa 306 static ext4_fsblk_t ext4_valid_block_bitmap_padding(struct super_block *sb,
311 307 ext4_group_t block_group,
312 308 struct buffer_head *bh)
313 { 309 {
314 ext4_grpblk_t next_zero_bit; 310 ext4_grpblk_t next_zero_bit;
315 unsigned long bitmap_size = sb->s_bloc 311 unsigned long bitmap_size = sb->s_blocksize * 8;
316 unsigned int offset = num_clusters_in_ 312 unsigned int offset = num_clusters_in_group(sb, block_group);
317 313
318 if (bitmap_size <= offset) 314 if (bitmap_size <= offset)
319 return 0; 315 return 0;
320 316
321 next_zero_bit = ext4_find_next_zero_bi 317 next_zero_bit = ext4_find_next_zero_bit(bh->b_data, bitmap_size, offset);
322 318
323 return (next_zero_bit < bitmap_size ? 319 return (next_zero_bit < bitmap_size ? next_zero_bit : 0);
324 } 320 }
325 321
326 struct ext4_group_info *ext4_get_group_info(st 322 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
327 ex 323 ext4_group_t group)
328 { 324 {
329 struct ext4_group_info **grp_info; 325 struct ext4_group_info **grp_info;
330 long indexv, indexh; 326 long indexv, indexh;
331 327
332 if (unlikely(group >= EXT4_SB(sb)->s_g 328 if (unlikely(group >= EXT4_SB(sb)->s_groups_count))
333 return NULL; 329 return NULL;
334 indexv = group >> (EXT4_DESC_PER_BLOCK 330 indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
335 indexh = group & ((EXT4_DESC_PER_BLOCK 331 indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
336 grp_info = sbi_array_rcu_deref(EXT4_SB 332 grp_info = sbi_array_rcu_deref(EXT4_SB(sb), s_group_info, indexv);
337 return grp_info[indexh]; 333 return grp_info[indexh];
338 } 334 }
339 335
340 /* 336 /*
341 * Return the block number which was discovere 337 * Return the block number which was discovered to be invalid, or 0 if
342 * the block bitmap is valid. 338 * the block bitmap is valid.
343 */ 339 */
344 static ext4_fsblk_t ext4_valid_block_bitmap(st 340 static ext4_fsblk_t ext4_valid_block_bitmap(struct super_block *sb,
345 st 341 struct ext4_group_desc *desc,
346 ex 342 ext4_group_t block_group,
347 st 343 struct buffer_head *bh)
348 { 344 {
349 struct ext4_sb_info *sbi = EXT4_SB(sb) 345 struct ext4_sb_info *sbi = EXT4_SB(sb);
350 ext4_grpblk_t offset; 346 ext4_grpblk_t offset;
351 ext4_grpblk_t next_zero_bit; 347 ext4_grpblk_t next_zero_bit;
352 ext4_grpblk_t max_bit = EXT4_CLUSTERS_ 348 ext4_grpblk_t max_bit = EXT4_CLUSTERS_PER_GROUP(sb);
353 ext4_fsblk_t blk; 349 ext4_fsblk_t blk;
354 ext4_fsblk_t group_first_block; 350 ext4_fsblk_t group_first_block;
355 351
356 if (ext4_has_feature_flex_bg(sb)) { 352 if (ext4_has_feature_flex_bg(sb)) {
357 /* with FLEX_BG, the inode/blo 353 /* with FLEX_BG, the inode/block bitmaps and itable
358 * blocks may not be in the gr 354 * blocks may not be in the group at all
359 * so the bitmap validation wi 355 * so the bitmap validation will be skipped for those groups
360 * or it has to also read the 356 * or it has to also read the block group where the bitmaps
361 * are located to verify they 357 * are located to verify they are set.
362 */ 358 */
363 return 0; 359 return 0;
364 } 360 }
365 group_first_block = ext4_group_first_b 361 group_first_block = ext4_group_first_block_no(sb, block_group);
366 362
367 /* check whether block bitmap block nu 363 /* check whether block bitmap block number is set */
368 blk = ext4_block_bitmap(sb, desc); 364 blk = ext4_block_bitmap(sb, desc);
369 offset = blk - group_first_block; 365 offset = blk - group_first_block;
370 if (offset < 0 || EXT4_B2C(sbi, offset 366 if (offset < 0 || EXT4_B2C(sbi, offset) >= max_bit ||
371 !ext4_test_bit(EXT4_B2C(sbi, offse 367 !ext4_test_bit(EXT4_B2C(sbi, offset), bh->b_data))
372 /* bad block bitmap */ 368 /* bad block bitmap */
373 return blk; 369 return blk;
374 370
375 /* check whether the inode bitmap bloc 371 /* check whether the inode bitmap block number is set */
376 blk = ext4_inode_bitmap(sb, desc); 372 blk = ext4_inode_bitmap(sb, desc);
377 offset = blk - group_first_block; 373 offset = blk - group_first_block;
378 if (offset < 0 || EXT4_B2C(sbi, offset 374 if (offset < 0 || EXT4_B2C(sbi, offset) >= max_bit ||
379 !ext4_test_bit(EXT4_B2C(sbi, offse 375 !ext4_test_bit(EXT4_B2C(sbi, offset), bh->b_data))
380 /* bad block bitmap */ 376 /* bad block bitmap */
381 return blk; 377 return blk;
382 378
383 /* check whether the inode table block 379 /* check whether the inode table block number is set */
384 blk = ext4_inode_table(sb, desc); 380 blk = ext4_inode_table(sb, desc);
385 offset = blk - group_first_block; 381 offset = blk - group_first_block;
386 if (offset < 0 || EXT4_B2C(sbi, offset 382 if (offset < 0 || EXT4_B2C(sbi, offset) >= max_bit ||
387 EXT4_B2C(sbi, offset + sbi->s_itb_ !! 383 EXT4_B2C(sbi, offset + sbi->s_itb_per_group) >= max_bit)
388 return blk; 384 return blk;
389 next_zero_bit = ext4_find_next_zero_bi 385 next_zero_bit = ext4_find_next_zero_bit(bh->b_data,
390 EXT4_B2C(sbi, offset + !! 386 EXT4_B2C(sbi, offset + sbi->s_itb_per_group),
391 EXT4_B2C(sbi, offset)) 387 EXT4_B2C(sbi, offset));
392 if (next_zero_bit < 388 if (next_zero_bit <
393 EXT4_B2C(sbi, offset + sbi->s_itb_ !! 389 EXT4_B2C(sbi, offset + sbi->s_itb_per_group))
394 /* bad bitmap for inode tables 390 /* bad bitmap for inode tables */
395 return blk; 391 return blk;
396 return 0; 392 return 0;
397 } 393 }
398 394
399 static int ext4_validate_block_bitmap(struct s 395 static int ext4_validate_block_bitmap(struct super_block *sb,
400 struct e 396 struct ext4_group_desc *desc,
401 ext4_gro 397 ext4_group_t block_group,
402 struct b 398 struct buffer_head *bh)
403 { 399 {
404 ext4_fsblk_t blk; 400 ext4_fsblk_t blk;
405 struct ext4_group_info *grp; 401 struct ext4_group_info *grp;
406 402
407 if (EXT4_SB(sb)->s_mount_state & EXT4_ 403 if (EXT4_SB(sb)->s_mount_state & EXT4_FC_REPLAY)
408 return 0; 404 return 0;
409 405
410 grp = ext4_get_group_info(sb, block_gr 406 grp = ext4_get_group_info(sb, block_group);
411 407
412 if (buffer_verified(bh)) 408 if (buffer_verified(bh))
413 return 0; 409 return 0;
414 if (!grp || EXT4_MB_GRP_BBITMAP_CORRUP 410 if (!grp || EXT4_MB_GRP_BBITMAP_CORRUPT(grp))
415 return -EFSCORRUPTED; 411 return -EFSCORRUPTED;
416 412
417 ext4_lock_group(sb, block_group); 413 ext4_lock_group(sb, block_group);
418 if (buffer_verified(bh)) 414 if (buffer_verified(bh))
419 goto verified; 415 goto verified;
420 if (unlikely(!ext4_block_bitmap_csum_v !! 416 if (unlikely(!ext4_block_bitmap_csum_verify(sb, block_group,
>> 417 desc, bh) ||
421 ext4_simulate_fail(sb, EX 418 ext4_simulate_fail(sb, EXT4_SIM_BBITMAP_CRC))) {
422 ext4_unlock_group(sb, block_gr 419 ext4_unlock_group(sb, block_group);
423 ext4_error(sb, "bg %u: bad blo 420 ext4_error(sb, "bg %u: bad block bitmap checksum", block_group);
424 ext4_mark_group_bitmap_corrupt 421 ext4_mark_group_bitmap_corrupted(sb, block_group,
425 EXT4_G 422 EXT4_GROUP_INFO_BBITMAP_CORRUPT);
426 return -EFSBADCRC; 423 return -EFSBADCRC;
427 } 424 }
428 blk = ext4_valid_block_bitmap(sb, desc 425 blk = ext4_valid_block_bitmap(sb, desc, block_group, bh);
429 if (unlikely(blk != 0)) { 426 if (unlikely(blk != 0)) {
430 ext4_unlock_group(sb, block_gr 427 ext4_unlock_group(sb, block_group);
431 ext4_error(sb, "bg %u: block % 428 ext4_error(sb, "bg %u: block %llu: invalid block bitmap",
432 block_group, blk); 429 block_group, blk);
433 ext4_mark_group_bitmap_corrupt 430 ext4_mark_group_bitmap_corrupted(sb, block_group,
434 EXT4_G 431 EXT4_GROUP_INFO_BBITMAP_CORRUPT);
435 return -EFSCORRUPTED; 432 return -EFSCORRUPTED;
436 } 433 }
437 blk = ext4_valid_block_bitmap_padding( 434 blk = ext4_valid_block_bitmap_padding(sb, block_group, bh);
438 if (unlikely(blk != 0)) { 435 if (unlikely(blk != 0)) {
439 ext4_unlock_group(sb, block_gr 436 ext4_unlock_group(sb, block_group);
440 ext4_error(sb, "bg %u: block % 437 ext4_error(sb, "bg %u: block %llu: padding at end of block bitmap is not set",
441 block_group, blk); 438 block_group, blk);
442 ext4_mark_group_bitmap_corrupt 439 ext4_mark_group_bitmap_corrupted(sb, block_group,
443 440 EXT4_GROUP_INFO_BBITMAP_CORRUPT);
444 return -EFSCORRUPTED; 441 return -EFSCORRUPTED;
445 } 442 }
446 set_buffer_verified(bh); 443 set_buffer_verified(bh);
447 verified: 444 verified:
448 ext4_unlock_group(sb, block_group); 445 ext4_unlock_group(sb, block_group);
449 return 0; 446 return 0;
450 } 447 }
451 448
452 /** 449 /**
453 * ext4_read_block_bitmap_nowait() 450 * ext4_read_block_bitmap_nowait()
454 * @sb: super block 451 * @sb: super block
455 * @block_group: given block group 452 * @block_group: given block group
456 * @ignore_locked: ignore locked buffers 453 * @ignore_locked: ignore locked buffers
457 * 454 *
458 * Read the bitmap for a given block_group,and 455 * Read the bitmap for a given block_group,and validate the
459 * bits for block/inode/inode tables are set i 456 * bits for block/inode/inode tables are set in the bitmaps
460 * 457 *
461 * Return buffer_head on success or an ERR_PTR 458 * Return buffer_head on success or an ERR_PTR in case of failure.
462 */ 459 */
463 struct buffer_head * 460 struct buffer_head *
464 ext4_read_block_bitmap_nowait(struct super_blo 461 ext4_read_block_bitmap_nowait(struct super_block *sb, ext4_group_t block_group,
465 bool ignore_lock 462 bool ignore_locked)
466 { 463 {
467 struct ext4_group_desc *desc; 464 struct ext4_group_desc *desc;
468 struct ext4_sb_info *sbi = EXT4_SB(sb) 465 struct ext4_sb_info *sbi = EXT4_SB(sb);
469 struct buffer_head *bh; 466 struct buffer_head *bh;
470 ext4_fsblk_t bitmap_blk; 467 ext4_fsblk_t bitmap_blk;
471 int err; 468 int err;
472 469
473 KUNIT_STATIC_STUB_REDIRECT(ext4_read_b <<
474 sb, block_g <<
475 <<
476 desc = ext4_get_group_desc(sb, block_g 470 desc = ext4_get_group_desc(sb, block_group, NULL);
477 if (!desc) 471 if (!desc)
478 return ERR_PTR(-EFSCORRUPTED); 472 return ERR_PTR(-EFSCORRUPTED);
479 bitmap_blk = ext4_block_bitmap(sb, des 473 bitmap_blk = ext4_block_bitmap(sb, desc);
480 if ((bitmap_blk <= le32_to_cpu(sbi->s_ 474 if ((bitmap_blk <= le32_to_cpu(sbi->s_es->s_first_data_block)) ||
481 (bitmap_blk >= ext4_blocks_count(s 475 (bitmap_blk >= ext4_blocks_count(sbi->s_es))) {
482 ext4_error(sb, "Invalid block 476 ext4_error(sb, "Invalid block bitmap block %llu in "
483 "block_group %u", b 477 "block_group %u", bitmap_blk, block_group);
484 ext4_mark_group_bitmap_corrupt 478 ext4_mark_group_bitmap_corrupted(sb, block_group,
485 EXT4_G 479 EXT4_GROUP_INFO_BBITMAP_CORRUPT);
486 return ERR_PTR(-EFSCORRUPTED); 480 return ERR_PTR(-EFSCORRUPTED);
487 } 481 }
488 bh = sb_getblk(sb, bitmap_blk); 482 bh = sb_getblk(sb, bitmap_blk);
489 if (unlikely(!bh)) { 483 if (unlikely(!bh)) {
490 ext4_warning(sb, "Cannot get b 484 ext4_warning(sb, "Cannot get buffer for block bitmap - "
491 "block_group = %u 485 "block_group = %u, block_bitmap = %llu",
492 block_group, bitm 486 block_group, bitmap_blk);
493 return ERR_PTR(-ENOMEM); 487 return ERR_PTR(-ENOMEM);
494 } 488 }
495 489
496 if (ignore_locked && buffer_locked(bh) 490 if (ignore_locked && buffer_locked(bh)) {
497 /* buffer under IO already, re 491 /* buffer under IO already, return if called for prefetching */
498 put_bh(bh); 492 put_bh(bh);
499 return NULL; 493 return NULL;
500 } 494 }
501 495
502 if (bitmap_uptodate(bh)) 496 if (bitmap_uptodate(bh))
503 goto verify; 497 goto verify;
504 498
505 lock_buffer(bh); 499 lock_buffer(bh);
506 if (bitmap_uptodate(bh)) { 500 if (bitmap_uptodate(bh)) {
507 unlock_buffer(bh); 501 unlock_buffer(bh);
508 goto verify; 502 goto verify;
509 } 503 }
510 ext4_lock_group(sb, block_group); 504 ext4_lock_group(sb, block_group);
511 if (ext4_has_group_desc_csum(sb) && 505 if (ext4_has_group_desc_csum(sb) &&
512 (desc->bg_flags & cpu_to_le16(EXT4 506 (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT))) {
513 if (block_group == 0) { 507 if (block_group == 0) {
514 ext4_unlock_group(sb, 508 ext4_unlock_group(sb, block_group);
515 unlock_buffer(bh); 509 unlock_buffer(bh);
516 ext4_error(sb, "Block 510 ext4_error(sb, "Block bitmap for bg 0 marked "
517 "uninitiali 511 "uninitialized");
518 err = -EFSCORRUPTED; 512 err = -EFSCORRUPTED;
519 goto out; 513 goto out;
520 } 514 }
521 err = ext4_init_block_bitmap(s 515 err = ext4_init_block_bitmap(sb, bh, block_group, desc);
522 if (err) { <<
523 ext4_unlock_group(sb, <<
524 unlock_buffer(bh); <<
525 ext4_error(sb, "Failed <<
526 "%u: %d", b <<
527 goto out; <<
528 } <<
529 set_bitmap_uptodate(bh); 516 set_bitmap_uptodate(bh);
530 set_buffer_uptodate(bh); 517 set_buffer_uptodate(bh);
531 set_buffer_verified(bh); 518 set_buffer_verified(bh);
532 ext4_unlock_group(sb, block_gr 519 ext4_unlock_group(sb, block_group);
533 unlock_buffer(bh); 520 unlock_buffer(bh);
534 return bh; !! 521 if (err) {
>> 522 ext4_error(sb, "Failed to init block bitmap for group "
>> 523 "%u: %d", block_group, err);
>> 524 goto out;
>> 525 }
>> 526 goto verify;
535 } 527 }
536 ext4_unlock_group(sb, block_group); 528 ext4_unlock_group(sb, block_group);
537 if (buffer_uptodate(bh)) { 529 if (buffer_uptodate(bh)) {
538 /* 530 /*
539 * if not uninit if bh is upto 531 * if not uninit if bh is uptodate,
540 * bitmap is also uptodate 532 * bitmap is also uptodate
541 */ 533 */
542 set_bitmap_uptodate(bh); 534 set_bitmap_uptodate(bh);
543 unlock_buffer(bh); 535 unlock_buffer(bh);
544 goto verify; 536 goto verify;
545 } 537 }
546 /* 538 /*
547 * submit the buffer_head for reading 539 * submit the buffer_head for reading
548 */ 540 */
549 set_buffer_new(bh); 541 set_buffer_new(bh);
550 trace_ext4_read_block_bitmap_load(sb, 542 trace_ext4_read_block_bitmap_load(sb, block_group, ignore_locked);
551 ext4_read_bh_nowait(bh, REQ_META | REQ 543 ext4_read_bh_nowait(bh, REQ_META | REQ_PRIO |
552 (ignore_locked ? R 544 (ignore_locked ? REQ_RAHEAD : 0),
553 ext4_end_bitmap_re 545 ext4_end_bitmap_read);
554 return bh; 546 return bh;
555 verify: 547 verify:
556 err = ext4_validate_block_bitmap(sb, d 548 err = ext4_validate_block_bitmap(sb, desc, block_group, bh);
557 if (err) 549 if (err)
558 goto out; 550 goto out;
559 return bh; 551 return bh;
560 out: 552 out:
561 put_bh(bh); 553 put_bh(bh);
562 return ERR_PTR(err); 554 return ERR_PTR(err);
563 } 555 }
564 556
565 /* Returns 0 on success, -errno on error */ 557 /* Returns 0 on success, -errno on error */
566 int ext4_wait_block_bitmap(struct super_block 558 int ext4_wait_block_bitmap(struct super_block *sb, ext4_group_t block_group,
567 struct buffer_head 559 struct buffer_head *bh)
568 { 560 {
569 struct ext4_group_desc *desc; 561 struct ext4_group_desc *desc;
570 562
571 KUNIT_STATIC_STUB_REDIRECT(ext4_wait_b <<
572 sb, block_g <<
573 <<
574 if (!buffer_new(bh)) 563 if (!buffer_new(bh))
575 return 0; 564 return 0;
576 desc = ext4_get_group_desc(sb, block_g 565 desc = ext4_get_group_desc(sb, block_group, NULL);
577 if (!desc) 566 if (!desc)
578 return -EFSCORRUPTED; 567 return -EFSCORRUPTED;
579 wait_on_buffer(bh); 568 wait_on_buffer(bh);
580 ext4_simulate_fail_bh(sb, bh, EXT4_SIM 569 ext4_simulate_fail_bh(sb, bh, EXT4_SIM_BBITMAP_EIO);
581 if (!buffer_uptodate(bh)) { 570 if (!buffer_uptodate(bh)) {
582 ext4_error_err(sb, EIO, "Canno 571 ext4_error_err(sb, EIO, "Cannot read block bitmap - "
583 "block_group = 572 "block_group = %u, block_bitmap = %llu",
584 block_group, (u 573 block_group, (unsigned long long) bh->b_blocknr);
585 ext4_mark_group_bitmap_corrupt 574 ext4_mark_group_bitmap_corrupted(sb, block_group,
586 EXT4_G 575 EXT4_GROUP_INFO_BBITMAP_CORRUPT);
587 return -EIO; 576 return -EIO;
588 } 577 }
589 clear_buffer_new(bh); 578 clear_buffer_new(bh);
590 /* Panic or remount fs read-only if bl 579 /* Panic or remount fs read-only if block bitmap is invalid */
591 return ext4_validate_block_bitmap(sb, 580 return ext4_validate_block_bitmap(sb, desc, block_group, bh);
592 } 581 }
593 582
594 struct buffer_head * 583 struct buffer_head *
595 ext4_read_block_bitmap(struct super_block *sb, 584 ext4_read_block_bitmap(struct super_block *sb, ext4_group_t block_group)
596 { 585 {
597 struct buffer_head *bh; 586 struct buffer_head *bh;
598 int err; 587 int err;
599 588
600 bh = ext4_read_block_bitmap_nowait(sb, 589 bh = ext4_read_block_bitmap_nowait(sb, block_group, false);
601 if (IS_ERR(bh)) 590 if (IS_ERR(bh))
602 return bh; 591 return bh;
603 err = ext4_wait_block_bitmap(sb, block 592 err = ext4_wait_block_bitmap(sb, block_group, bh);
604 if (err) { 593 if (err) {
605 put_bh(bh); 594 put_bh(bh);
606 return ERR_PTR(err); 595 return ERR_PTR(err);
607 } 596 }
608 return bh; 597 return bh;
609 } 598 }
610 599
611 /** 600 /**
612 * ext4_has_free_clusters() 601 * ext4_has_free_clusters()
613 * @sbi: in-core super block structure. 602 * @sbi: in-core super block structure.
614 * @nclusters: number of needed blocks 603 * @nclusters: number of needed blocks
615 * @flags: flags from ext4_mb_new_blocks( 604 * @flags: flags from ext4_mb_new_blocks()
616 * 605 *
617 * Check if filesystem has nclusters free & av 606 * Check if filesystem has nclusters free & available for allocation.
618 * On success return 1, return 0 on failure. 607 * On success return 1, return 0 on failure.
619 */ 608 */
620 static int ext4_has_free_clusters(struct ext4_ 609 static int ext4_has_free_clusters(struct ext4_sb_info *sbi,
621 s64 ncluster 610 s64 nclusters, unsigned int flags)
622 { 611 {
623 s64 free_clusters, dirty_clusters, rsv 612 s64 free_clusters, dirty_clusters, rsv, resv_clusters;
624 struct percpu_counter *fcc = &sbi->s_f 613 struct percpu_counter *fcc = &sbi->s_freeclusters_counter;
625 struct percpu_counter *dcc = &sbi->s_d 614 struct percpu_counter *dcc = &sbi->s_dirtyclusters_counter;
626 615
627 free_clusters = percpu_counter_read_p 616 free_clusters = percpu_counter_read_positive(fcc);
628 dirty_clusters = percpu_counter_read_p 617 dirty_clusters = percpu_counter_read_positive(dcc);
629 resv_clusters = atomic64_read(&sbi->s_ 618 resv_clusters = atomic64_read(&sbi->s_resv_clusters);
630 619
631 /* 620 /*
632 * r_blocks_count should always be mul 621 * r_blocks_count should always be multiple of the cluster ratio so
633 * we are safe to do a plane bit shift 622 * we are safe to do a plane bit shift only.
634 */ 623 */
635 rsv = (ext4_r_blocks_count(sbi->s_es) 624 rsv = (ext4_r_blocks_count(sbi->s_es) >> sbi->s_cluster_bits) +
636 resv_clusters; 625 resv_clusters;
637 626
638 if (free_clusters - (nclusters + rsv + 627 if (free_clusters - (nclusters + rsv + dirty_clusters) <
639 EXT4_F 628 EXT4_FREECLUSTERS_WATERMARK) {
640 free_clusters = percpu_counte 629 free_clusters = percpu_counter_sum_positive(fcc);
641 dirty_clusters = percpu_counte 630 dirty_clusters = percpu_counter_sum_positive(dcc);
642 } 631 }
643 /* Check whether we have space after a 632 /* Check whether we have space after accounting for current
644 * dirty clusters & root reserved clus 633 * dirty clusters & root reserved clusters.
645 */ 634 */
646 if (free_clusters >= (rsv + nclusters 635 if (free_clusters >= (rsv + nclusters + dirty_clusters))
647 return 1; 636 return 1;
648 637
649 /* Hm, nope. Are (enough) root reserv 638 /* Hm, nope. Are (enough) root reserved clusters available? */
650 if (uid_eq(sbi->s_resuid, current_fsui 639 if (uid_eq(sbi->s_resuid, current_fsuid()) ||
651 (!gid_eq(sbi->s_resgid, GLOBAL_ROO 640 (!gid_eq(sbi->s_resgid, GLOBAL_ROOT_GID) && in_group_p(sbi->s_resgid)) ||
652 capable(CAP_SYS_RESOURCE) || 641 capable(CAP_SYS_RESOURCE) ||
653 (flags & EXT4_MB_USE_ROOT_BLOCKS)) 642 (flags & EXT4_MB_USE_ROOT_BLOCKS)) {
654 643
655 if (free_clusters >= (ncluster 644 if (free_clusters >= (nclusters + dirty_clusters +
656 resv_clu 645 resv_clusters))
657 return 1; 646 return 1;
658 } 647 }
659 /* No free blocks. Let's see if we can 648 /* No free blocks. Let's see if we can dip into reserved pool */
660 if (flags & EXT4_MB_USE_RESERVED) { 649 if (flags & EXT4_MB_USE_RESERVED) {
661 if (free_clusters >= (ncluster 650 if (free_clusters >= (nclusters + dirty_clusters))
662 return 1; 651 return 1;
663 } 652 }
664 653
665 return 0; 654 return 0;
666 } 655 }
667 656
668 int ext4_claim_free_clusters(struct ext4_sb_in 657 int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
669 s64 nclusters, un 658 s64 nclusters, unsigned int flags)
670 { 659 {
671 if (ext4_has_free_clusters(sbi, nclust 660 if (ext4_has_free_clusters(sbi, nclusters, flags)) {
672 percpu_counter_add(&sbi->s_dir 661 percpu_counter_add(&sbi->s_dirtyclusters_counter, nclusters);
673 return 0; 662 return 0;
674 } else 663 } else
675 return -ENOSPC; 664 return -ENOSPC;
676 } 665 }
677 666
678 /** 667 /**
679 * ext4_should_retry_alloc() - check if a bloc 668 * ext4_should_retry_alloc() - check if a block allocation should be retried
680 * @sb: superblock 669 * @sb: superblock
681 * @retries: number of retry attemp 670 * @retries: number of retry attempts made so far
682 * 671 *
683 * ext4_should_retry_alloc() is called when EN 672 * ext4_should_retry_alloc() is called when ENOSPC is returned while
684 * attempting to allocate blocks. If there's 673 * attempting to allocate blocks. If there's an indication that a pending
685 * journal transaction might free some space a 674 * journal transaction might free some space and allow another attempt to
686 * succeed, this function will wait for the cu 675 * succeed, this function will wait for the current or committing transaction
687 * to complete and then return TRUE. 676 * to complete and then return TRUE.
688 */ 677 */
689 int ext4_should_retry_alloc(struct super_block 678 int ext4_should_retry_alloc(struct super_block *sb, int *retries)
690 { 679 {
691 struct ext4_sb_info *sbi = EXT4_SB(sb) 680 struct ext4_sb_info *sbi = EXT4_SB(sb);
692 681
693 if (!sbi->s_journal) 682 if (!sbi->s_journal)
694 return 0; 683 return 0;
695 684
696 if (++(*retries) > 3) { 685 if (++(*retries) > 3) {
697 percpu_counter_inc(&sbi->s_sra 686 percpu_counter_inc(&sbi->s_sra_exceeded_retry_limit);
698 return 0; 687 return 0;
699 } 688 }
700 689
701 /* 690 /*
702 * if there's no indication that block 691 * if there's no indication that blocks are about to be freed it's
703 * possible we just missed a transacti 692 * possible we just missed a transaction commit that did so
704 */ 693 */
705 smp_mb(); 694 smp_mb();
706 if (sbi->s_mb_free_pending == 0) { 695 if (sbi->s_mb_free_pending == 0) {
707 if (test_opt(sb, DISCARD)) { 696 if (test_opt(sb, DISCARD)) {
708 atomic_inc(&sbi->s_ret 697 atomic_inc(&sbi->s_retry_alloc_pending);
709 flush_work(&sbi->s_dis 698 flush_work(&sbi->s_discard_work);
710 atomic_dec(&sbi->s_ret 699 atomic_dec(&sbi->s_retry_alloc_pending);
711 } 700 }
712 return ext4_has_free_clusters( 701 return ext4_has_free_clusters(sbi, 1, 0);
713 } 702 }
714 703
715 /* 704 /*
716 * it's possible we've just missed a t 705 * it's possible we've just missed a transaction commit here,
717 * so ignore the returned status 706 * so ignore the returned status
718 */ 707 */
719 ext4_debug("%s: retrying operation aft 708 ext4_debug("%s: retrying operation after ENOSPC\n", sb->s_id);
720 (void) jbd2_journal_force_commit_neste 709 (void) jbd2_journal_force_commit_nested(sbi->s_journal);
721 return 1; 710 return 1;
722 } 711 }
723 712
724 /* 713 /*
725 * ext4_new_meta_blocks() -- allocate block fo 714 * ext4_new_meta_blocks() -- allocate block for meta data (indexing) blocks
726 * 715 *
727 * @handle: handle to this transac 716 * @handle: handle to this transaction
728 * @inode: file inode 717 * @inode: file inode
729 * @goal: given target block(fil 718 * @goal: given target block(filesystem wide)
730 * @count: pointer to total numbe 719 * @count: pointer to total number of clusters needed
731 * @errp: error code 720 * @errp: error code
732 * 721 *
733 * Return 1st allocated block number on succes 722 * Return 1st allocated block number on success, *count stores total account
734 * error stores in errp pointer 723 * error stores in errp pointer
735 */ 724 */
736 ext4_fsblk_t ext4_new_meta_blocks(handle_t *ha 725 ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
737 ext4_fsblk_t 726 ext4_fsblk_t goal, unsigned int flags,
738 unsigned lon 727 unsigned long *count, int *errp)
739 { 728 {
740 struct ext4_allocation_request ar; 729 struct ext4_allocation_request ar;
741 ext4_fsblk_t ret; 730 ext4_fsblk_t ret;
742 731
743 memset(&ar, 0, sizeof(ar)); 732 memset(&ar, 0, sizeof(ar));
744 /* Fill with neighbour allocated block 733 /* Fill with neighbour allocated blocks */
745 ar.inode = inode; 734 ar.inode = inode;
746 ar.goal = goal; 735 ar.goal = goal;
747 ar.len = count ? *count : 1; 736 ar.len = count ? *count : 1;
748 ar.flags = flags; 737 ar.flags = flags;
749 738
750 ret = ext4_mb_new_blocks(handle, &ar, 739 ret = ext4_mb_new_blocks(handle, &ar, errp);
751 if (count) 740 if (count)
752 *count = ar.len; 741 *count = ar.len;
753 /* 742 /*
754 * Account for the allocated meta bloc 743 * Account for the allocated meta blocks. We will never
755 * fail EDQUOT for metdata, but we do 744 * fail EDQUOT for metdata, but we do account for it.
756 */ 745 */
757 if (!(*errp) && (flags & EXT4_MB_DELAL 746 if (!(*errp) && (flags & EXT4_MB_DELALLOC_RESERVED)) {
758 dquot_alloc_block_nofail(inode 747 dquot_alloc_block_nofail(inode,
759 EXT4_C2B(EXT4_ 748 EXT4_C2B(EXT4_SB(inode->i_sb), ar.len));
760 } 749 }
761 return ret; 750 return ret;
762 } 751 }
763 752
764 /** 753 /**
765 * ext4_count_free_clusters() -- count filesys 754 * ext4_count_free_clusters() -- count filesystem free clusters
766 * @sb: superblock 755 * @sb: superblock
767 * 756 *
768 * Adds up the number of free clusters from ea 757 * Adds up the number of free clusters from each block group.
769 */ 758 */
770 ext4_fsblk_t ext4_count_free_clusters(struct s 759 ext4_fsblk_t ext4_count_free_clusters(struct super_block *sb)
771 { 760 {
772 ext4_fsblk_t desc_count; 761 ext4_fsblk_t desc_count;
773 struct ext4_group_desc *gdp; 762 struct ext4_group_desc *gdp;
774 ext4_group_t i; 763 ext4_group_t i;
775 ext4_group_t ngroups = ext4_get_groups 764 ext4_group_t ngroups = ext4_get_groups_count(sb);
776 struct ext4_group_info *grp; 765 struct ext4_group_info *grp;
777 #ifdef EXT4FS_DEBUG 766 #ifdef EXT4FS_DEBUG
778 struct ext4_super_block *es; 767 struct ext4_super_block *es;
779 ext4_fsblk_t bitmap_count; 768 ext4_fsblk_t bitmap_count;
780 unsigned int x; 769 unsigned int x;
781 struct buffer_head *bitmap_bh = NULL; 770 struct buffer_head *bitmap_bh = NULL;
782 771
783 es = EXT4_SB(sb)->s_es; 772 es = EXT4_SB(sb)->s_es;
784 desc_count = 0; 773 desc_count = 0;
785 bitmap_count = 0; 774 bitmap_count = 0;
786 gdp = NULL; 775 gdp = NULL;
787 776
788 for (i = 0; i < ngroups; i++) { 777 for (i = 0; i < ngroups; i++) {
789 gdp = ext4_get_group_desc(sb, 778 gdp = ext4_get_group_desc(sb, i, NULL);
790 if (!gdp) 779 if (!gdp)
791 continue; 780 continue;
792 grp = NULL; 781 grp = NULL;
793 if (EXT4_SB(sb)->s_group_info) 782 if (EXT4_SB(sb)->s_group_info)
794 grp = ext4_get_group_i 783 grp = ext4_get_group_info(sb, i);
795 if (!grp || !EXT4_MB_GRP_BBITM 784 if (!grp || !EXT4_MB_GRP_BBITMAP_CORRUPT(grp))
796 desc_count += ext4_fre 785 desc_count += ext4_free_group_clusters(sb, gdp);
797 brelse(bitmap_bh); 786 brelse(bitmap_bh);
798 bitmap_bh = ext4_read_block_bi 787 bitmap_bh = ext4_read_block_bitmap(sb, i);
799 if (IS_ERR(bitmap_bh)) { 788 if (IS_ERR(bitmap_bh)) {
800 bitmap_bh = NULL; 789 bitmap_bh = NULL;
801 continue; 790 continue;
802 } 791 }
803 792
804 x = ext4_count_free(bitmap_bh- 793 x = ext4_count_free(bitmap_bh->b_data,
805 EXT4_CLUST 794 EXT4_CLUSTERS_PER_GROUP(sb) / 8);
806 printk(KERN_DEBUG "group %u: s 795 printk(KERN_DEBUG "group %u: stored = %d, counted = %u\n",
807 i, ext4_free_group_clu 796 i, ext4_free_group_clusters(sb, gdp), x);
808 bitmap_count += x; 797 bitmap_count += x;
809 } 798 }
810 brelse(bitmap_bh); 799 brelse(bitmap_bh);
811 printk(KERN_DEBUG "ext4_count_free_clu 800 printk(KERN_DEBUG "ext4_count_free_clusters: stored = %llu"
812 ", computed = %llu, %llu\n", 801 ", computed = %llu, %llu\n",
813 EXT4_NUM_B2C(EXT4_SB(sb), ext4_ 802 EXT4_NUM_B2C(EXT4_SB(sb), ext4_free_blocks_count(es)),
814 desc_count, bitmap_count); 803 desc_count, bitmap_count);
815 return bitmap_count; 804 return bitmap_count;
816 #else 805 #else
817 desc_count = 0; 806 desc_count = 0;
818 for (i = 0; i < ngroups; i++) { 807 for (i = 0; i < ngroups; i++) {
819 gdp = ext4_get_group_desc(sb, 808 gdp = ext4_get_group_desc(sb, i, NULL);
820 if (!gdp) 809 if (!gdp)
821 continue; 810 continue;
822 grp = NULL; 811 grp = NULL;
823 if (EXT4_SB(sb)->s_group_info) 812 if (EXT4_SB(sb)->s_group_info)
824 grp = ext4_get_group_i 813 grp = ext4_get_group_info(sb, i);
825 if (!grp || !EXT4_MB_GRP_BBITM 814 if (!grp || !EXT4_MB_GRP_BBITMAP_CORRUPT(grp))
826 desc_count += ext4_fre 815 desc_count += ext4_free_group_clusters(sb, gdp);
827 } 816 }
828 817
829 return desc_count; 818 return desc_count;
830 #endif 819 #endif
831 } 820 }
832 821
833 static inline int test_root(ext4_group_t a, in 822 static inline int test_root(ext4_group_t a, int b)
834 { 823 {
835 while (1) { 824 while (1) {
836 if (a < b) 825 if (a < b)
837 return 0; 826 return 0;
838 if (a == b) 827 if (a == b)
839 return 1; 828 return 1;
840 if ((a % b) != 0) 829 if ((a % b) != 0)
841 return 0; 830 return 0;
842 a = a / b; 831 a = a / b;
843 } 832 }
844 } 833 }
845 834
846 /** 835 /**
847 * ext4_bg_has_super - number of blocks u 836 * ext4_bg_has_super - number of blocks used by the superblock in group
848 * @sb: superblock for filesystem 837 * @sb: superblock for filesystem
849 * @group: group number to check 838 * @group: group number to check
850 * 839 *
851 * Return the number of blocks used by th 840 * Return the number of blocks used by the superblock (primary or backup)
852 * in this group. Currently this will be 841 * in this group. Currently this will be only 0 or 1.
853 */ 842 */
854 int ext4_bg_has_super(struct super_block *sb, 843 int ext4_bg_has_super(struct super_block *sb, ext4_group_t group)
855 { 844 {
856 struct ext4_super_block *es = EXT4_SB( 845 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
857 846
858 if (group == 0) 847 if (group == 0)
859 return 1; 848 return 1;
860 if (ext4_has_feature_sparse_super2(sb) 849 if (ext4_has_feature_sparse_super2(sb)) {
861 if (group == le32_to_cpu(es->s 850 if (group == le32_to_cpu(es->s_backup_bgs[0]) ||
862 group == le32_to_cpu(es->s 851 group == le32_to_cpu(es->s_backup_bgs[1]))
863 return 1; 852 return 1;
864 return 0; 853 return 0;
865 } 854 }
866 if ((group <= 1) || !ext4_has_feature_ 855 if ((group <= 1) || !ext4_has_feature_sparse_super(sb))
867 return 1; 856 return 1;
868 if (!(group & 1)) 857 if (!(group & 1))
869 return 0; 858 return 0;
870 if (test_root(group, 3) || (test_root( 859 if (test_root(group, 3) || (test_root(group, 5)) ||
871 test_root(group, 7)) 860 test_root(group, 7))
872 return 1; 861 return 1;
873 862
874 return 0; 863 return 0;
875 } 864 }
876 865
877 static unsigned long ext4_bg_num_gdb_meta(stru 866 static unsigned long ext4_bg_num_gdb_meta(struct super_block *sb,
878 ext4_g 867 ext4_group_t group)
879 { 868 {
880 unsigned long metagroup = group / EXT4 869 unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
881 ext4_group_t first = metagroup * EXT4_ 870 ext4_group_t first = metagroup * EXT4_DESC_PER_BLOCK(sb);
882 ext4_group_t last = first + EXT4_DESC_ 871 ext4_group_t last = first + EXT4_DESC_PER_BLOCK(sb) - 1;
883 872
884 if (group == first || group == first + 873 if (group == first || group == first + 1 || group == last)
885 return 1; 874 return 1;
886 return 0; 875 return 0;
887 } 876 }
888 877
889 static unsigned long ext4_bg_num_gdb_nometa(st 878 static unsigned long ext4_bg_num_gdb_nometa(struct super_block *sb,
890 ext4_g 879 ext4_group_t group)
891 { 880 {
892 if (!ext4_bg_has_super(sb, group)) 881 if (!ext4_bg_has_super(sb, group))
893 return 0; 882 return 0;
894 883
895 if (ext4_has_feature_meta_bg(sb)) 884 if (ext4_has_feature_meta_bg(sb))
896 return le32_to_cpu(EXT4_SB(sb) 885 return le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg);
897 else 886 else
898 return EXT4_SB(sb)->s_gdb_coun 887 return EXT4_SB(sb)->s_gdb_count;
899 } 888 }
900 889
901 /** 890 /**
902 * ext4_bg_num_gdb - number of blocks use 891 * ext4_bg_num_gdb - number of blocks used by the group table in group
903 * @sb: superblock for filesystem 892 * @sb: superblock for filesystem
904 * @group: group number to check 893 * @group: group number to check
905 * 894 *
906 * Return the number of blocks used by th 895 * Return the number of blocks used by the group descriptor table
907 * (primary or backup) in this group. In 896 * (primary or backup) in this group. In the future there may be a
908 * different number of descriptor blocks 897 * different number of descriptor blocks in each group.
909 */ 898 */
910 unsigned long ext4_bg_num_gdb(struct super_blo 899 unsigned long ext4_bg_num_gdb(struct super_block *sb, ext4_group_t group)
911 { 900 {
912 unsigned long first_meta_bg = 901 unsigned long first_meta_bg =
913 le32_to_cpu(EXT4_SB(sb 902 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_meta_bg);
914 unsigned long metagroup = group / EXT4 903 unsigned long metagroup = group / EXT4_DESC_PER_BLOCK(sb);
915 904
916 if (!ext4_has_feature_meta_bg(sb) || m 905 if (!ext4_has_feature_meta_bg(sb) || metagroup < first_meta_bg)
917 return ext4_bg_num_gdb_nometa( 906 return ext4_bg_num_gdb_nometa(sb, group);
918 907
919 return ext4_bg_num_gdb_meta(sb,group); 908 return ext4_bg_num_gdb_meta(sb,group);
920 909
921 } 910 }
922 911
923 /* 912 /*
924 * This function returns the number of file sy 913 * This function returns the number of file system metadata blocks at
925 * the beginning of a block group, including t 914 * the beginning of a block group, including the reserved gdt blocks.
926 */ 915 */
927 unsigned int ext4_num_base_meta_blocks(struct 916 unsigned int ext4_num_base_meta_blocks(struct super_block *sb,
928 ext4_gr 917 ext4_group_t block_group)
929 { 918 {
930 struct ext4_sb_info *sbi = EXT4_SB(sb) 919 struct ext4_sb_info *sbi = EXT4_SB(sb);
931 unsigned num; 920 unsigned num;
932 921
933 /* Check for superblock and gdt backup 922 /* Check for superblock and gdt backups in this group */
934 num = ext4_bg_has_super(sb, block_grou 923 num = ext4_bg_has_super(sb, block_group);
935 924
936 if (!ext4_has_feature_meta_bg(sb) || 925 if (!ext4_has_feature_meta_bg(sb) ||
937 block_group < le32_to_cpu(sbi->s_e 926 block_group < le32_to_cpu(sbi->s_es->s_first_meta_bg) *
938 sbi->s_desc_per_bloc 927 sbi->s_desc_per_block) {
939 if (num) { 928 if (num) {
940 num += ext4_bg_num_gdb !! 929 num += ext4_bg_num_gdb(sb, block_group);
941 num += le16_to_cpu(sbi 930 num += le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks);
942 } 931 }
943 } else { /* For META_BG_BLOCK_GROUPS * 932 } else { /* For META_BG_BLOCK_GROUPS */
944 num += ext4_bg_num_gdb_meta(sb !! 933 num += ext4_bg_num_gdb(sb, block_group);
945 } 934 }
946 return num; 935 return num;
947 } 936 }
948 937
949 static unsigned int ext4_num_base_meta_cluster 938 static unsigned int ext4_num_base_meta_clusters(struct super_block *sb,
950 939 ext4_group_t block_group)
951 { 940 {
952 return EXT4_NUM_B2C(EXT4_SB(sb), ext4_ 941 return EXT4_NUM_B2C(EXT4_SB(sb), ext4_num_base_meta_blocks(sb, block_group));
953 } 942 }
954 943
955 /** 944 /**
956 * ext4_inode_to_goal_block - return a hi 945 * ext4_inode_to_goal_block - return a hint for block allocation
957 * @inode: inode for block allocation 946 * @inode: inode for block allocation
958 * 947 *
959 * Return the ideal location to start all 948 * Return the ideal location to start allocating blocks for a
960 * newly created inode. 949 * newly created inode.
961 */ 950 */
962 ext4_fsblk_t ext4_inode_to_goal_block(struct i 951 ext4_fsblk_t ext4_inode_to_goal_block(struct inode *inode)
963 { 952 {
964 struct ext4_inode_info *ei = EXT4_I(in 953 struct ext4_inode_info *ei = EXT4_I(inode);
965 ext4_group_t block_group; 954 ext4_group_t block_group;
966 ext4_grpblk_t colour; 955 ext4_grpblk_t colour;
967 int flex_size = ext4_flex_bg_size(EXT4 956 int flex_size = ext4_flex_bg_size(EXT4_SB(inode->i_sb));
968 ext4_fsblk_t bg_start; 957 ext4_fsblk_t bg_start;
969 ext4_fsblk_t last_block; 958 ext4_fsblk_t last_block;
970 959
971 block_group = ei->i_block_group; 960 block_group = ei->i_block_group;
972 if (flex_size >= EXT4_FLEX_SIZE_DIR_AL 961 if (flex_size >= EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME) {
973 /* 962 /*
974 * If there are at least EXT4_ 963 * If there are at least EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME
975 * block groups per flexgroup, 964 * block groups per flexgroup, reserve the first block
976 * group for directories and s 965 * group for directories and special files. Regular
977 * files will start at the sec 966 * files will start at the second block group. This
978 * tends to speed up directory 967 * tends to speed up directory access and improves
979 * fsck times. 968 * fsck times.
980 */ 969 */
981 block_group &= ~(flex_size-1); 970 block_group &= ~(flex_size-1);
982 if (S_ISREG(inode->i_mode)) 971 if (S_ISREG(inode->i_mode))
983 block_group++; 972 block_group++;
984 } 973 }
985 bg_start = ext4_group_first_block_no(i 974 bg_start = ext4_group_first_block_no(inode->i_sb, block_group);
986 last_block = ext4_blocks_count(EXT4_SB 975 last_block = ext4_blocks_count(EXT4_SB(inode->i_sb)->s_es) - 1;
987 976
988 /* 977 /*
989 * If we are doing delayed allocation, 978 * If we are doing delayed allocation, we don't need take
990 * colour into account. 979 * colour into account.
991 */ 980 */
992 if (test_opt(inode->i_sb, DELALLOC)) 981 if (test_opt(inode->i_sb, DELALLOC))
993 return bg_start; 982 return bg_start;
994 983
995 if (bg_start + EXT4_BLOCKS_PER_GROUP(i 984 if (bg_start + EXT4_BLOCKS_PER_GROUP(inode->i_sb) <= last_block)
996 colour = (task_pid_nr(current) 985 colour = (task_pid_nr(current) % 16) *
997 (EXT4_BLOCKS_PER_GROUP 986 (EXT4_BLOCKS_PER_GROUP(inode->i_sb) / 16);
998 else 987 else
999 colour = (task_pid_nr(current) 988 colour = (task_pid_nr(current) % 16) *
1000 ((last_block - bg_sta 989 ((last_block - bg_start) / 16);
1001 return bg_start + colour; 990 return bg_start + colour;
1002 } 991 }
1003 992
1004 993
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.