]>
Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
ac27a0ec | 2 | /* |
617ba13b | 3 | * linux/fs/ext4/ialloc.c |
ac27a0ec DK |
4 | * |
5 | * Copyright (C) 1992, 1993, 1994, 1995 | |
6 | * Remy Card (card@masi.ibp.fr) | |
7 | * Laboratoire MASI - Institut Blaise Pascal | |
8 | * Universite Pierre et Marie Curie (Paris VI) | |
9 | * | |
10 | * BSD ufs-inspired inode and directory allocation by | |
11 | * Stephen Tweedie (sct@redhat.com), 1993 | |
12 | * Big-endian to little-endian byte-swapping/bitmaps by | |
13 | * David S. Miller (davem@caip.rutgers.edu), 1995 | |
14 | */ | |
15 | ||
16 | #include <linux/time.h> | |
17 | #include <linux/fs.h> | |
ac27a0ec DK |
18 | #include <linux/stat.h> |
19 | #include <linux/string.h> | |
20 | #include <linux/quotaops.h> | |
21 | #include <linux/buffer_head.h> | |
22 | #include <linux/random.h> | |
23 | #include <linux/bitops.h> | |
3a5b2ecd | 24 | #include <linux/blkdev.h> |
5b825c3a IM |
25 | #include <linux/cred.h> |
26 | ||
ac27a0ec | 27 | #include <asm/byteorder.h> |
9bffad1e | 28 | |
3dcf5451 CH |
29 | #include "ext4.h" |
30 | #include "ext4_jbd2.h" | |
ac27a0ec DK |
31 | #include "xattr.h" |
32 | #include "acl.h" | |
33 | ||
9bffad1e TT |
34 | #include <trace/events/ext4.h> |
35 | ||
ac27a0ec DK |
36 | /* |
37 | * ialloc.c contains the inodes allocation and deallocation routines | |
38 | */ | |
39 | ||
40 | /* | |
41 | * The free inodes are managed by bitmaps. A file system contains several | |
42 | * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap | |
43 | * block for inodes, N blocks for the inode table and data blocks. | |
44 | * | |
45 | * The file system contains group descriptors which are located after the | |
46 | * super block. Each descriptor contains the number of the bitmap block and | |
47 | * the free blocks count in the block. | |
48 | */ | |
49 | ||
717d50e4 AD |
50 | /* |
51 | * To avoid calling the atomic setbit hundreds or thousands of times, we only | |
52 | * need to use it within a single byte (to ensure we get endianness right). | |
53 | * We can use memset for the rest of the bitmap as there are no other users. | |
54 | */ | |
61d08673 | 55 | void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap) |
717d50e4 AD |
56 | { |
57 | int i; | |
58 | ||
59 | if (start_bit >= end_bit) | |
60 | return; | |
61 | ||
62 | ext4_debug("mark end bits +%d through +%d used\n", start_bit, end_bit); | |
63 | for (i = start_bit; i < ((start_bit + 7) & ~7UL); i++) | |
64 | ext4_set_bit(i, bitmap); | |
65 | if (i < end_bit) | |
66 | memset(bitmap + (i >> 3), 0xff, (end_bit - i) >> 3); | |
67 | } | |
68 | ||
813e5727 TT |
69 | void ext4_end_bitmap_read(struct buffer_head *bh, int uptodate) |
70 | { | |
71 | if (uptodate) { | |
72 | set_buffer_uptodate(bh); | |
73 | set_bitmap_uptodate(bh); | |
74 | } | |
75 | unlock_buffer(bh); | |
76 | put_bh(bh); | |
77 | } | |
78 | ||
9008a58e DW |
79 | static int ext4_validate_inode_bitmap(struct super_block *sb, |
80 | struct ext4_group_desc *desc, | |
81 | ext4_group_t block_group, | |
82 | struct buffer_head *bh) | |
83 | { | |
84 | ext4_fsblk_t blk; | |
8016e29f HS |
85 | struct ext4_group_info *grp; |
86 | ||
87 | if (EXT4_SB(sb)->s_mount_state & EXT4_FC_REPLAY) | |
88 | return 0; | |
89 | ||
90 | grp = ext4_get_group_info(sb, block_group); | |
9008a58e DW |
91 | |
92 | if (buffer_verified(bh)) | |
93 | return 0; | |
94 | if (EXT4_MB_GRP_IBITMAP_CORRUPT(grp)) | |
95 | return -EFSCORRUPTED; | |
96 | ||
97 | ext4_lock_group(sb, block_group); | |
8d5a803c TT |
98 | if (buffer_verified(bh)) |
99 | goto verified; | |
9008a58e DW |
100 | blk = ext4_inode_bitmap(sb, desc); |
101 | if (!ext4_inode_bitmap_csum_verify(sb, block_group, desc, bh, | |
46f870d6 TT |
102 | EXT4_INODES_PER_GROUP(sb) / 8) || |
103 | ext4_simulate_fail(sb, EXT4_SIM_IBITMAP_CRC)) { | |
9008a58e DW |
104 | ext4_unlock_group(sb, block_group); |
105 | ext4_error(sb, "Corrupt inode bitmap - block_group = %u, " | |
106 | "inode_bitmap = %llu", block_group, blk); | |
db79e6d1 WS |
107 | ext4_mark_group_bitmap_corrupted(sb, block_group, |
108 | EXT4_GROUP_INFO_IBITMAP_CORRUPT); | |
9008a58e DW |
109 | return -EFSBADCRC; |
110 | } | |
111 | set_buffer_verified(bh); | |
8d5a803c | 112 | verified: |
9008a58e DW |
113 | ext4_unlock_group(sb, block_group); |
114 | return 0; | |
115 | } | |
116 | ||
ac27a0ec DK |
117 | /* |
118 | * Read the inode allocation bitmap for a given block_group, reading | |
119 | * into the specified slot in the superblock's bitmap cache. | |
120 | * | |
9033783c | 121 | * Return buffer_head of bitmap on success, or an ERR_PTR on error. |
ac27a0ec DK |
122 | */ |
123 | static struct buffer_head * | |
e29d1cde | 124 | ext4_read_inode_bitmap(struct super_block *sb, ext4_group_t block_group) |
ac27a0ec | 125 | { |
617ba13b | 126 | struct ext4_group_desc *desc; |
7dac4a17 | 127 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec | 128 | struct buffer_head *bh = NULL; |
e29d1cde | 129 | ext4_fsblk_t bitmap_blk; |
9008a58e | 130 | int err; |
ac27a0ec | 131 | |
617ba13b | 132 | desc = ext4_get_group_desc(sb, block_group, NULL); |
ac27a0ec | 133 | if (!desc) |
9008a58e | 134 | return ERR_PTR(-EFSCORRUPTED); |
bfff6873 | 135 | |
e29d1cde | 136 | bitmap_blk = ext4_inode_bitmap(sb, desc); |
7dac4a17 TT |
137 | if ((bitmap_blk <= le32_to_cpu(sbi->s_es->s_first_data_block)) || |
138 | (bitmap_blk >= ext4_blocks_count(sbi->s_es))) { | |
139 | ext4_error(sb, "Invalid inode bitmap blk %llu in " | |
140 | "block_group %u", bitmap_blk, block_group); | |
206f6d55 WS |
141 | ext4_mark_group_bitmap_corrupted(sb, block_group, |
142 | EXT4_GROUP_INFO_IBITMAP_CORRUPT); | |
7dac4a17 TT |
143 | return ERR_PTR(-EFSCORRUPTED); |
144 | } | |
e29d1cde ES |
145 | bh = sb_getblk(sb, bitmap_blk); |
146 | if (unlikely(!bh)) { | |
5ef2a699 WS |
147 | ext4_warning(sb, "Cannot read inode bitmap - " |
148 | "block_group = %u, inode_bitmap = %llu", | |
149 | block_group, bitmap_blk); | |
0db9fdeb | 150 | return ERR_PTR(-ENOMEM); |
e29d1cde | 151 | } |
2ccb5fb9 | 152 | if (bitmap_uptodate(bh)) |
41a246d1 | 153 | goto verify; |
e29d1cde | 154 | |
c806e68f | 155 | lock_buffer(bh); |
2ccb5fb9 AK |
156 | if (bitmap_uptodate(bh)) { |
157 | unlock_buffer(bh); | |
41a246d1 | 158 | goto verify; |
2ccb5fb9 | 159 | } |
bfff6873 | 160 | |
955ce5f5 | 161 | ext4_lock_group(sb, block_group); |
8844618d TT |
162 | if (ext4_has_group_desc_csum(sb) && |
163 | (desc->bg_flags & cpu_to_le16(EXT4_BG_INODE_UNINIT))) { | |
164 | if (block_group == 0) { | |
165 | ext4_unlock_group(sb, block_group); | |
166 | unlock_buffer(bh); | |
167 | ext4_error(sb, "Inode bitmap for bg 0 marked " | |
168 | "uninitialized"); | |
169 | err = -EFSCORRUPTED; | |
170 | goto out; | |
171 | } | |
044e6e3d TT |
172 | memset(bh->b_data, 0, (EXT4_INODES_PER_GROUP(sb) + 7) / 8); |
173 | ext4_mark_bitmap_end(EXT4_INODES_PER_GROUP(sb), | |
174 | sb->s_blocksize * 8, bh->b_data); | |
2ccb5fb9 | 175 | set_bitmap_uptodate(bh); |
e29d1cde | 176 | set_buffer_uptodate(bh); |
41a246d1 | 177 | set_buffer_verified(bh); |
955ce5f5 | 178 | ext4_unlock_group(sb, block_group); |
3300beda | 179 | unlock_buffer(bh); |
e29d1cde | 180 | return bh; |
717d50e4 | 181 | } |
955ce5f5 | 182 | ext4_unlock_group(sb, block_group); |
bfff6873 | 183 | |
2ccb5fb9 AK |
184 | if (buffer_uptodate(bh)) { |
185 | /* | |
186 | * if not uninit if bh is uptodate, | |
187 | * bitmap is also uptodate | |
188 | */ | |
189 | set_bitmap_uptodate(bh); | |
190 | unlock_buffer(bh); | |
41a246d1 | 191 | goto verify; |
2ccb5fb9 AK |
192 | } |
193 | /* | |
813e5727 | 194 | * submit the buffer_head for reading |
2ccb5fb9 | 195 | */ |
0562e0ba | 196 | trace_ext4_load_inode_bitmap(sb, block_group); |
2d069c08 | 197 | ext4_read_bh(bh, REQ_META | REQ_PRIO, ext4_end_bitmap_read); |
46f870d6 | 198 | ext4_simulate_fail_bh(sb, bh, EXT4_SIM_IBITMAP_EIO); |
813e5727 | 199 | if (!buffer_uptodate(bh)) { |
e29d1cde | 200 | put_bh(bh); |
54d3adbc TT |
201 | ext4_error_err(sb, EIO, "Cannot read inode bitmap - " |
202 | "block_group = %u, inode_bitmap = %llu", | |
203 | block_group, bitmap_blk); | |
206f6d55 WS |
204 | ext4_mark_group_bitmap_corrupted(sb, block_group, |
205 | EXT4_GROUP_INFO_IBITMAP_CORRUPT); | |
9008a58e | 206 | return ERR_PTR(-EIO); |
e29d1cde | 207 | } |
41a246d1 DW |
208 | |
209 | verify: | |
9008a58e DW |
210 | err = ext4_validate_inode_bitmap(sb, desc, block_group, bh); |
211 | if (err) | |
212 | goto out; | |
ac27a0ec | 213 | return bh; |
9008a58e DW |
214 | out: |
215 | put_bh(bh); | |
216 | return ERR_PTR(err); | |
ac27a0ec DK |
217 | } |
218 | ||
219 | /* | |
220 | * NOTE! When we get the inode, we're the only people | |
221 | * that have access to it, and as such there are no | |
222 | * race conditions we have to worry about. The inode | |
223 | * is not on the hash-lists, and it cannot be reached | |
224 | * through the filesystem because the directory entry | |
225 | * has been deleted earlier. | |
226 | * | |
227 | * HOWEVER: we must make sure that we get no aliases, | |
228 | * which means that we have to call "clear_inode()" | |
229 | * _before_ we mark the inode not in use in the inode | |
230 | * bitmaps. Otherwise a newly created file might use | |
231 | * the same inode number (not actually the same pointer | |
232 | * though), and then we'd have two inodes sharing the | |
233 | * same inode number and space on the harddisk. | |
234 | */ | |
af5bc92d | 235 | void ext4_free_inode(handle_t *handle, struct inode *inode) |
ac27a0ec | 236 | { |
af5bc92d | 237 | struct super_block *sb = inode->i_sb; |
ac27a0ec DK |
238 | int is_directory; |
239 | unsigned long ino; | |
240 | struct buffer_head *bitmap_bh = NULL; | |
241 | struct buffer_head *bh2; | |
fd2d4291 | 242 | ext4_group_t block_group; |
ac27a0ec | 243 | unsigned long bit; |
af5bc92d TT |
244 | struct ext4_group_desc *gdp; |
245 | struct ext4_super_block *es; | |
617ba13b | 246 | struct ext4_sb_info *sbi; |
7ce9d5d1 | 247 | int fatal = 0, err, count, cleared; |
87a39389 | 248 | struct ext4_group_info *grp; |
ac27a0ec | 249 | |
92b97816 TT |
250 | if (!sb) { |
251 | printk(KERN_ERR "EXT4-fs: %s:%d: inode on " | |
252 | "nonexistent device\n", __func__, __LINE__); | |
ac27a0ec DK |
253 | return; |
254 | } | |
92b97816 TT |
255 | if (atomic_read(&inode->i_count) > 1) { |
256 | ext4_msg(sb, KERN_ERR, "%s:%d: inode #%lu: count=%d", | |
257 | __func__, __LINE__, inode->i_ino, | |
258 | atomic_read(&inode->i_count)); | |
ac27a0ec DK |
259 | return; |
260 | } | |
92b97816 TT |
261 | if (inode->i_nlink) { |
262 | ext4_msg(sb, KERN_ERR, "%s:%d: inode #%lu: nlink=%d\n", | |
263 | __func__, __LINE__, inode->i_ino, inode->i_nlink); | |
ac27a0ec DK |
264 | return; |
265 | } | |
617ba13b | 266 | sbi = EXT4_SB(sb); |
ac27a0ec DK |
267 | |
268 | ino = inode->i_ino; | |
af5bc92d | 269 | ext4_debug("freeing inode %lu\n", ino); |
9bffad1e | 270 | trace_ext4_free_inode(inode); |
ac27a0ec | 271 | |
871a2931 | 272 | dquot_initialize(inode); |
63936dda | 273 | dquot_free_inode(inode); |
ac27a0ec DK |
274 | |
275 | is_directory = S_ISDIR(inode->i_mode); | |
276 | ||
277 | /* Do this BEFORE marking the inode not in use or returning an error */ | |
0930fcc1 | 278 | ext4_clear_inode(inode); |
ac27a0ec | 279 | |
49598e04 | 280 | es = sbi->s_es; |
617ba13b | 281 | if (ino < EXT4_FIRST_INO(sb) || ino > le32_to_cpu(es->s_inodes_count)) { |
12062ddd | 282 | ext4_error(sb, "reserved or nonexistent inode %lu", ino); |
ac27a0ec DK |
283 | goto error_return; |
284 | } | |
617ba13b MC |
285 | block_group = (ino - 1) / EXT4_INODES_PER_GROUP(sb); |
286 | bit = (ino - 1) % EXT4_INODES_PER_GROUP(sb); | |
e29d1cde | 287 | bitmap_bh = ext4_read_inode_bitmap(sb, block_group); |
87a39389 | 288 | /* Don't bother if the inode bitmap is corrupt. */ |
9008a58e DW |
289 | if (IS_ERR(bitmap_bh)) { |
290 | fatal = PTR_ERR(bitmap_bh); | |
291 | bitmap_bh = NULL; | |
292 | goto error_return; | |
293 | } | |
8016e29f HS |
294 | if (!(sbi->s_mount_state & EXT4_FC_REPLAY)) { |
295 | grp = ext4_get_group_info(sb, block_group); | |
296 | if (unlikely(EXT4_MB_GRP_IBITMAP_CORRUPT(grp))) { | |
297 | fatal = -EFSCORRUPTED; | |
298 | goto error_return; | |
299 | } | |
9008a58e | 300 | } |
ac27a0ec DK |
301 | |
302 | BUFFER_TRACE(bitmap_bh, "get_write_access"); | |
617ba13b | 303 | fatal = ext4_journal_get_write_access(handle, bitmap_bh); |
ac27a0ec DK |
304 | if (fatal) |
305 | goto error_return; | |
306 | ||
d17413c0 DM |
307 | fatal = -ESRCH; |
308 | gdp = ext4_get_group_desc(sb, block_group, &bh2); | |
309 | if (gdp) { | |
ac27a0ec | 310 | BUFFER_TRACE(bh2, "get_write_access"); |
617ba13b | 311 | fatal = ext4_journal_get_write_access(handle, bh2); |
d17413c0 DM |
312 | } |
313 | ext4_lock_group(sb, block_group); | |
597d508c | 314 | cleared = ext4_test_and_clear_bit(bit, bitmap_bh->b_data); |
d17413c0 DM |
315 | if (fatal || !cleared) { |
316 | ext4_unlock_group(sb, block_group); | |
317 | goto out; | |
318 | } | |
7d39db14 | 319 | |
d17413c0 DM |
320 | count = ext4_free_inodes_count(sb, gdp) + 1; |
321 | ext4_free_inodes_set(sb, gdp, count); | |
322 | if (is_directory) { | |
323 | count = ext4_used_dirs_count(sb, gdp) - 1; | |
324 | ext4_used_dirs_set(sb, gdp, count); | |
b45f189a RH |
325 | if (percpu_counter_initialized(&sbi->s_dirs_counter)) |
326 | percpu_counter_dec(&sbi->s_dirs_counter); | |
ac27a0ec | 327 | } |
41a246d1 DW |
328 | ext4_inode_bitmap_csum_set(sb, block_group, gdp, bitmap_bh, |
329 | EXT4_INODES_PER_GROUP(sb) / 8); | |
feb0ab32 | 330 | ext4_group_desc_csum_set(sb, block_group, gdp); |
d17413c0 | 331 | ext4_unlock_group(sb, block_group); |
ac27a0ec | 332 | |
b45f189a RH |
333 | if (percpu_counter_initialized(&sbi->s_freeinodes_counter)) |
334 | percpu_counter_inc(&sbi->s_freeinodes_counter); | |
d17413c0 | 335 | if (sbi->s_log_groups_per_flex) { |
7c990728 | 336 | struct flex_groups *fg; |
9f24e420 | 337 | |
7c990728 SJS |
338 | fg = sbi_array_rcu_deref(sbi, s_flex_groups, |
339 | ext4_flex_group(sbi, block_group)); | |
340 | atomic_inc(&fg->free_inodes); | |
d17413c0 | 341 | if (is_directory) |
7c990728 | 342 | atomic_dec(&fg->used_dirs); |
ac27a0ec | 343 | } |
d17413c0 DM |
344 | BUFFER_TRACE(bh2, "call ext4_handle_dirty_metadata"); |
345 | fatal = ext4_handle_dirty_metadata(handle, NULL, bh2); | |
346 | out: | |
347 | if (cleared) { | |
348 | BUFFER_TRACE(bitmap_bh, "call ext4_handle_dirty_metadata"); | |
349 | err = ext4_handle_dirty_metadata(handle, NULL, bitmap_bh); | |
350 | if (!fatal) | |
351 | fatal = err; | |
87a39389 | 352 | } else { |
d17413c0 | 353 | ext4_error(sb, "bit already cleared for inode %lu", ino); |
db79e6d1 WS |
354 | ext4_mark_group_bitmap_corrupted(sb, block_group, |
355 | EXT4_GROUP_INFO_IBITMAP_CORRUPT); | |
87a39389 | 356 | } |
d17413c0 | 357 | |
ac27a0ec DK |
358 | error_return: |
359 | brelse(bitmap_bh); | |
617ba13b | 360 | ext4_std_error(sb, fatal); |
ac27a0ec DK |
361 | } |
362 | ||
a4912123 | 363 | struct orlov_stats { |
90ba983f | 364 | __u64 free_clusters; |
a4912123 | 365 | __u32 free_inodes; |
a4912123 TT |
366 | __u32 used_dirs; |
367 | }; | |
368 | ||
369 | /* | |
370 | * Helper function for Orlov's allocator; returns critical information | |
371 | * for a particular block group or flex_bg. If flex_size is 1, then g | |
372 | * is a block group number; otherwise it is flex_bg number. | |
373 | */ | |
1f109d5a TT |
374 | static void get_orlov_stats(struct super_block *sb, ext4_group_t g, |
375 | int flex_size, struct orlov_stats *stats) | |
a4912123 TT |
376 | { |
377 | struct ext4_group_desc *desc; | |
a4912123 | 378 | |
7d39db14 | 379 | if (flex_size > 1) { |
7c990728 SJS |
380 | struct flex_groups *fg = sbi_array_rcu_deref(EXT4_SB(sb), |
381 | s_flex_groups, g); | |
382 | stats->free_inodes = atomic_read(&fg->free_inodes); | |
383 | stats->free_clusters = atomic64_read(&fg->free_clusters); | |
384 | stats->used_dirs = atomic_read(&fg->used_dirs); | |
7d39db14 TT |
385 | return; |
386 | } | |
a4912123 | 387 | |
7d39db14 TT |
388 | desc = ext4_get_group_desc(sb, g, NULL); |
389 | if (desc) { | |
390 | stats->free_inodes = ext4_free_inodes_count(sb, desc); | |
021b65bb | 391 | stats->free_clusters = ext4_free_group_clusters(sb, desc); |
7d39db14 TT |
392 | stats->used_dirs = ext4_used_dirs_count(sb, desc); |
393 | } else { | |
394 | stats->free_inodes = 0; | |
24aaa8ef | 395 | stats->free_clusters = 0; |
7d39db14 | 396 | stats->used_dirs = 0; |
a4912123 TT |
397 | } |
398 | } | |
399 | ||
ac27a0ec DK |
400 | /* |
401 | * Orlov's allocator for directories. | |
402 | * | |
403 | * We always try to spread first-level directories. | |
404 | * | |
071ce6d3 | 405 | * If there are blockgroups with both free inodes and free clusters counts |
ac27a0ec DK |
406 | * not worse than average we return one with smallest directory count. |
407 | * Otherwise we simply return a random group. | |
408 | * | |
409 | * For the rest rules look so: | |
410 | * | |
411 | * It's OK to put directory into a group unless | |
412 | * it has too many directories already (max_dirs) or | |
413 | * it has too few free inodes left (min_inodes) or | |
071ce6d3 | 414 | * it has too few free clusters left (min_clusters) or |
1cc8dcf5 | 415 | * Parent's group is preferred, if it doesn't satisfy these |
ac27a0ec DK |
416 | * conditions we search cyclically through the rest. If none |
417 | * of the groups look good we just look for a group with more | |
418 | * free inodes than average (starting at parent's group). | |
ac27a0ec DK |
419 | */ |
420 | ||
2aa9fc4c | 421 | static int find_group_orlov(struct super_block *sb, struct inode *parent, |
dcca3fec | 422 | ext4_group_t *group, umode_t mode, |
f157a4aa | 423 | const struct qstr *qstr) |
ac27a0ec | 424 | { |
fd2d4291 | 425 | ext4_group_t parent_group = EXT4_I(parent)->i_block_group; |
617ba13b | 426 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
8df9675f | 427 | ext4_group_t real_ngroups = ext4_get_groups_count(sb); |
617ba13b | 428 | int inodes_per_group = EXT4_INODES_PER_GROUP(sb); |
14c83c9f | 429 | unsigned int freei, avefreei, grp_free; |
071ce6d3 | 430 | ext4_fsblk_t freec, avefreec; |
ac27a0ec | 431 | unsigned int ndirs; |
a4912123 | 432 | int max_dirs, min_inodes; |
24aaa8ef | 433 | ext4_grpblk_t min_clusters; |
8df9675f | 434 | ext4_group_t i, grp, g, ngroups; |
617ba13b | 435 | struct ext4_group_desc *desc; |
a4912123 TT |
436 | struct orlov_stats stats; |
437 | int flex_size = ext4_flex_bg_size(sbi); | |
f157a4aa | 438 | struct dx_hash_info hinfo; |
a4912123 | 439 | |
8df9675f | 440 | ngroups = real_ngroups; |
a4912123 | 441 | if (flex_size > 1) { |
8df9675f | 442 | ngroups = (real_ngroups + flex_size - 1) >> |
a4912123 TT |
443 | sbi->s_log_groups_per_flex; |
444 | parent_group >>= sbi->s_log_groups_per_flex; | |
445 | } | |
ac27a0ec DK |
446 | |
447 | freei = percpu_counter_read_positive(&sbi->s_freeinodes_counter); | |
448 | avefreei = freei / ngroups; | |
071ce6d3 PD |
449 | freec = percpu_counter_read_positive(&sbi->s_freeclusters_counter); |
450 | avefreec = freec; | |
24aaa8ef | 451 | do_div(avefreec, ngroups); |
ac27a0ec DK |
452 | ndirs = percpu_counter_read_positive(&sbi->s_dirs_counter); |
453 | ||
a4912123 | 454 | if (S_ISDIR(mode) && |
2b0143b5 | 455 | ((parent == d_inode(sb->s_root)) || |
12e9b892 | 456 | (ext4_test_inode_flag(parent, EXT4_INODE_TOPDIR)))) { |
ac27a0ec | 457 | int best_ndir = inodes_per_group; |
2aa9fc4c | 458 | int ret = -1; |
ac27a0ec | 459 | |
f157a4aa TT |
460 | if (qstr) { |
461 | hinfo.hash_version = DX_HASH_HALF_MD4; | |
462 | hinfo.seed = sbi->s_hash_seed; | |
b886ee3e | 463 | ext4fs_dirhash(parent, qstr->name, qstr->len, &hinfo); |
f157a4aa TT |
464 | grp = hinfo.hash; |
465 | } else | |
dd1f723b | 466 | grp = prandom_u32(); |
2aa9fc4c | 467 | parent_group = (unsigned)grp % ngroups; |
ac27a0ec | 468 | for (i = 0; i < ngroups; i++) { |
a4912123 TT |
469 | g = (parent_group + i) % ngroups; |
470 | get_orlov_stats(sb, g, flex_size, &stats); | |
471 | if (!stats.free_inodes) | |
ac27a0ec | 472 | continue; |
a4912123 | 473 | if (stats.used_dirs >= best_ndir) |
ac27a0ec | 474 | continue; |
a4912123 | 475 | if (stats.free_inodes < avefreei) |
ac27a0ec | 476 | continue; |
24aaa8ef | 477 | if (stats.free_clusters < avefreec) |
ac27a0ec | 478 | continue; |
a4912123 | 479 | grp = g; |
2aa9fc4c | 480 | ret = 0; |
a4912123 TT |
481 | best_ndir = stats.used_dirs; |
482 | } | |
483 | if (ret) | |
484 | goto fallback; | |
485 | found_flex_bg: | |
486 | if (flex_size == 1) { | |
487 | *group = grp; | |
488 | return 0; | |
489 | } | |
490 | ||
491 | /* | |
492 | * We pack inodes at the beginning of the flexgroup's | |
493 | * inode tables. Block allocation decisions will do | |
494 | * something similar, although regular files will | |
495 | * start at 2nd block group of the flexgroup. See | |
496 | * ext4_ext_find_goal() and ext4_find_near(). | |
497 | */ | |
498 | grp *= flex_size; | |
499 | for (i = 0; i < flex_size; i++) { | |
8df9675f | 500 | if (grp+i >= real_ngroups) |
a4912123 TT |
501 | break; |
502 | desc = ext4_get_group_desc(sb, grp+i, NULL); | |
503 | if (desc && ext4_free_inodes_count(sb, desc)) { | |
504 | *group = grp+i; | |
505 | return 0; | |
506 | } | |
ac27a0ec | 507 | } |
ac27a0ec DK |
508 | goto fallback; |
509 | } | |
510 | ||
ac27a0ec | 511 | max_dirs = ndirs / ngroups + inodes_per_group / 16; |
a4912123 TT |
512 | min_inodes = avefreei - inodes_per_group*flex_size / 4; |
513 | if (min_inodes < 1) | |
514 | min_inodes = 1; | |
24aaa8ef | 515 | min_clusters = avefreec - EXT4_CLUSTERS_PER_GROUP(sb)*flex_size / 4; |
a4912123 TT |
516 | |
517 | /* | |
518 | * Start looking in the flex group where we last allocated an | |
519 | * inode for this parent directory | |
520 | */ | |
521 | if (EXT4_I(parent)->i_last_alloc_group != ~0) { | |
522 | parent_group = EXT4_I(parent)->i_last_alloc_group; | |
523 | if (flex_size > 1) | |
524 | parent_group >>= sbi->s_log_groups_per_flex; | |
525 | } | |
ac27a0ec DK |
526 | |
527 | for (i = 0; i < ngroups; i++) { | |
a4912123 TT |
528 | grp = (parent_group + i) % ngroups; |
529 | get_orlov_stats(sb, grp, flex_size, &stats); | |
530 | if (stats.used_dirs >= max_dirs) | |
ac27a0ec | 531 | continue; |
a4912123 | 532 | if (stats.free_inodes < min_inodes) |
ac27a0ec | 533 | continue; |
24aaa8ef | 534 | if (stats.free_clusters < min_clusters) |
ac27a0ec | 535 | continue; |
a4912123 | 536 | goto found_flex_bg; |
ac27a0ec DK |
537 | } |
538 | ||
539 | fallback: | |
8df9675f | 540 | ngroups = real_ngroups; |
a4912123 | 541 | avefreei = freei / ngroups; |
b5451f7b | 542 | fallback_retry: |
a4912123 | 543 | parent_group = EXT4_I(parent)->i_block_group; |
ac27a0ec | 544 | for (i = 0; i < ngroups; i++) { |
a4912123 TT |
545 | grp = (parent_group + i) % ngroups; |
546 | desc = ext4_get_group_desc(sb, grp, NULL); | |
bb3d132a DC |
547 | if (desc) { |
548 | grp_free = ext4_free_inodes_count(sb, desc); | |
549 | if (grp_free && grp_free >= avefreei) { | |
550 | *group = grp; | |
551 | return 0; | |
552 | } | |
a4912123 | 553 | } |
ac27a0ec DK |
554 | } |
555 | ||
556 | if (avefreei) { | |
557 | /* | |
558 | * The free-inodes counter is approximate, and for really small | |
559 | * filesystems the above test can fail to find any blockgroups | |
560 | */ | |
561 | avefreei = 0; | |
b5451f7b | 562 | goto fallback_retry; |
ac27a0ec DK |
563 | } |
564 | ||
565 | return -1; | |
566 | } | |
567 | ||
2aa9fc4c | 568 | static int find_group_other(struct super_block *sb, struct inode *parent, |
dcca3fec | 569 | ext4_group_t *group, umode_t mode) |
ac27a0ec | 570 | { |
fd2d4291 | 571 | ext4_group_t parent_group = EXT4_I(parent)->i_block_group; |
8df9675f | 572 | ext4_group_t i, last, ngroups = ext4_get_groups_count(sb); |
617ba13b | 573 | struct ext4_group_desc *desc; |
a4912123 TT |
574 | int flex_size = ext4_flex_bg_size(EXT4_SB(sb)); |
575 | ||
576 | /* | |
577 | * Try to place the inode is the same flex group as its | |
578 | * parent. If we can't find space, use the Orlov algorithm to | |
579 | * find another flex group, and store that information in the | |
580 | * parent directory's inode information so that use that flex | |
581 | * group for future allocations. | |
582 | */ | |
583 | if (flex_size > 1) { | |
584 | int retry = 0; | |
585 | ||
586 | try_again: | |
587 | parent_group &= ~(flex_size-1); | |
588 | last = parent_group + flex_size; | |
589 | if (last > ngroups) | |
590 | last = ngroups; | |
591 | for (i = parent_group; i < last; i++) { | |
592 | desc = ext4_get_group_desc(sb, i, NULL); | |
593 | if (desc && ext4_free_inodes_count(sb, desc)) { | |
594 | *group = i; | |
595 | return 0; | |
596 | } | |
597 | } | |
598 | if (!retry && EXT4_I(parent)->i_last_alloc_group != ~0) { | |
599 | retry = 1; | |
600 | parent_group = EXT4_I(parent)->i_last_alloc_group; | |
601 | goto try_again; | |
602 | } | |
603 | /* | |
604 | * If this didn't work, use the Orlov search algorithm | |
605 | * to find a new flex group; we pass in the mode to | |
606 | * avoid the topdir algorithms. | |
607 | */ | |
608 | *group = parent_group + flex_size; | |
609 | if (*group > ngroups) | |
610 | *group = 0; | |
7dc57615 | 611 | return find_group_orlov(sb, parent, group, mode, NULL); |
a4912123 | 612 | } |
ac27a0ec DK |
613 | |
614 | /* | |
615 | * Try to place the inode in its parent directory | |
616 | */ | |
2aa9fc4c AM |
617 | *group = parent_group; |
618 | desc = ext4_get_group_desc(sb, *group, NULL); | |
560671a0 | 619 | if (desc && ext4_free_inodes_count(sb, desc) && |
021b65bb | 620 | ext4_free_group_clusters(sb, desc)) |
2aa9fc4c | 621 | return 0; |
ac27a0ec DK |
622 | |
623 | /* | |
624 | * We're going to place this inode in a different blockgroup from its | |
625 | * parent. We want to cause files in a common directory to all land in | |
626 | * the same blockgroup. But we want files which are in a different | |
627 | * directory which shares a blockgroup with our parent to land in a | |
628 | * different blockgroup. | |
629 | * | |
630 | * So add our directory's i_ino into the starting point for the hash. | |
631 | */ | |
2aa9fc4c | 632 | *group = (*group + parent->i_ino) % ngroups; |
ac27a0ec DK |
633 | |
634 | /* | |
635 | * Use a quadratic hash to find a group with a free inode and some free | |
636 | * blocks. | |
637 | */ | |
638 | for (i = 1; i < ngroups; i <<= 1) { | |
2aa9fc4c AM |
639 | *group += i; |
640 | if (*group >= ngroups) | |
641 | *group -= ngroups; | |
642 | desc = ext4_get_group_desc(sb, *group, NULL); | |
560671a0 | 643 | if (desc && ext4_free_inodes_count(sb, desc) && |
021b65bb | 644 | ext4_free_group_clusters(sb, desc)) |
2aa9fc4c | 645 | return 0; |
ac27a0ec DK |
646 | } |
647 | ||
648 | /* | |
649 | * That failed: try linear search for a free inode, even if that group | |
650 | * has no free blocks. | |
651 | */ | |
2aa9fc4c | 652 | *group = parent_group; |
ac27a0ec | 653 | for (i = 0; i < ngroups; i++) { |
2aa9fc4c AM |
654 | if (++*group >= ngroups) |
655 | *group = 0; | |
656 | desc = ext4_get_group_desc(sb, *group, NULL); | |
560671a0 | 657 | if (desc && ext4_free_inodes_count(sb, desc)) |
2aa9fc4c | 658 | return 0; |
ac27a0ec DK |
659 | } |
660 | ||
661 | return -1; | |
662 | } | |
663 | ||
19883bd9 TT |
664 | /* |
665 | * In no journal mode, if an inode has recently been deleted, we want | |
666 | * to avoid reusing it until we're reasonably sure the inode table | |
667 | * block has been written back to disk. (Yes, these values are | |
668 | * somewhat arbitrary...) | |
669 | */ | |
a17a9d93 | 670 | #define RECENTCY_MIN 60 |
b5f51573 | 671 | #define RECENTCY_DIRTY 300 |
19883bd9 TT |
672 | |
673 | static int recently_deleted(struct super_block *sb, ext4_group_t group, int ino) | |
674 | { | |
675 | struct ext4_group_desc *gdp; | |
676 | struct ext4_inode *raw_inode; | |
677 | struct buffer_head *bh; | |
b5f51573 AD |
678 | int inodes_per_block = EXT4_SB(sb)->s_inodes_per_block; |
679 | int offset, ret = 0; | |
680 | int recentcy = RECENTCY_MIN; | |
681 | u32 dtime, now; | |
19883bd9 TT |
682 | |
683 | gdp = ext4_get_group_desc(sb, group, NULL); | |
684 | if (unlikely(!gdp)) | |
685 | return 0; | |
686 | ||
4f9d956d | 687 | bh = sb_find_get_block(sb, ext4_inode_table(sb, gdp) + |
19883bd9 | 688 | (ino / inodes_per_block)); |
4f9d956d | 689 | if (!bh || !buffer_uptodate(bh)) |
19883bd9 TT |
690 | /* |
691 | * If the block is not in the buffer cache, then it | |
692 | * must have been written out. | |
693 | */ | |
694 | goto out; | |
695 | ||
696 | offset = (ino % inodes_per_block) * EXT4_INODE_SIZE(sb); | |
697 | raw_inode = (struct ext4_inode *) (bh->b_data + offset); | |
b5f51573 AD |
698 | |
699 | /* i_dtime is only 32 bits on disk, but we only care about relative | |
700 | * times in the range of a few minutes (i.e. long enough to sync a | |
701 | * recently-deleted inode to disk), so using the low 32 bits of the | |
702 | * clock (a 68 year range) is enough, see time_before32() */ | |
19883bd9 | 703 | dtime = le32_to_cpu(raw_inode->i_dtime); |
b5f51573 | 704 | now = ktime_get_real_seconds(); |
19883bd9 TT |
705 | if (buffer_dirty(bh)) |
706 | recentcy += RECENTCY_DIRTY; | |
707 | ||
b5f51573 AD |
708 | if (dtime && time_before32(dtime, now) && |
709 | time_before32(now, dtime + recentcy)) | |
19883bd9 TT |
710 | ret = 1; |
711 | out: | |
712 | brelse(bh); | |
713 | return ret; | |
714 | } | |
715 | ||
901ed070 WS |
716 | static int find_inode_bit(struct super_block *sb, ext4_group_t group, |
717 | struct buffer_head *bitmap, unsigned long *ino) | |
718 | { | |
d05466b2 JK |
719 | bool check_recently_deleted = EXT4_SB(sb)->s_journal == NULL; |
720 | unsigned long recently_deleted_ino = EXT4_INODES_PER_GROUP(sb); | |
721 | ||
901ed070 WS |
722 | next: |
723 | *ino = ext4_find_next_zero_bit((unsigned long *) | |
724 | bitmap->b_data, | |
725 | EXT4_INODES_PER_GROUP(sb), *ino); | |
726 | if (*ino >= EXT4_INODES_PER_GROUP(sb)) | |
d05466b2 | 727 | goto not_found; |
901ed070 | 728 | |
d05466b2 JK |
729 | if (check_recently_deleted && recently_deleted(sb, group, *ino)) { |
730 | recently_deleted_ino = *ino; | |
901ed070 WS |
731 | *ino = *ino + 1; |
732 | if (*ino < EXT4_INODES_PER_GROUP(sb)) | |
733 | goto next; | |
d05466b2 | 734 | goto not_found; |
901ed070 | 735 | } |
d05466b2 JK |
736 | return 1; |
737 | not_found: | |
738 | if (recently_deleted_ino >= EXT4_INODES_PER_GROUP(sb)) | |
739 | return 0; | |
740 | /* | |
741 | * Not reusing recently deleted inodes is mostly a preference. We don't | |
742 | * want to report ENOSPC or skew allocation patterns because of that. | |
743 | * So return even recently deleted inode if we could find better in the | |
744 | * given range. | |
745 | */ | |
746 | *ino = recently_deleted_ino; | |
901ed070 WS |
747 | return 1; |
748 | } | |
749 | ||
8016e29f HS |
750 | int ext4_mark_inode_used(struct super_block *sb, int ino) |
751 | { | |
752 | unsigned long max_ino = le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count); | |
753 | struct buffer_head *inode_bitmap_bh = NULL, *group_desc_bh = NULL; | |
754 | struct ext4_group_desc *gdp; | |
755 | ext4_group_t group; | |
756 | int bit; | |
757 | int err = -EFSCORRUPTED; | |
758 | ||
759 | if (ino < EXT4_FIRST_INO(sb) || ino > max_ino) | |
760 | goto out; | |
761 | ||
762 | group = (ino - 1) / EXT4_INODES_PER_GROUP(sb); | |
763 | bit = (ino - 1) % EXT4_INODES_PER_GROUP(sb); | |
764 | inode_bitmap_bh = ext4_read_inode_bitmap(sb, group); | |
765 | if (IS_ERR(inode_bitmap_bh)) | |
766 | return PTR_ERR(inode_bitmap_bh); | |
767 | ||
768 | if (ext4_test_bit(bit, inode_bitmap_bh->b_data)) { | |
769 | err = 0; | |
770 | goto out; | |
771 | } | |
772 | ||
773 | gdp = ext4_get_group_desc(sb, group, &group_desc_bh); | |
774 | if (!gdp || !group_desc_bh) { | |
775 | err = -EINVAL; | |
776 | goto out; | |
777 | } | |
778 | ||
779 | ext4_set_bit(bit, inode_bitmap_bh->b_data); | |
780 | ||
781 | BUFFER_TRACE(inode_bitmap_bh, "call ext4_handle_dirty_metadata"); | |
782 | err = ext4_handle_dirty_metadata(NULL, NULL, inode_bitmap_bh); | |
783 | if (err) { | |
784 | ext4_std_error(sb, err); | |
785 | goto out; | |
786 | } | |
787 | err = sync_dirty_buffer(inode_bitmap_bh); | |
788 | if (err) { | |
789 | ext4_std_error(sb, err); | |
790 | goto out; | |
791 | } | |
792 | ||
793 | /* We may have to initialize the block bitmap if it isn't already */ | |
794 | if (ext4_has_group_desc_csum(sb) && | |
795 | gdp->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) { | |
796 | struct buffer_head *block_bitmap_bh; | |
797 | ||
798 | block_bitmap_bh = ext4_read_block_bitmap(sb, group); | |
799 | if (IS_ERR(block_bitmap_bh)) { | |
800 | err = PTR_ERR(block_bitmap_bh); | |
801 | goto out; | |
802 | } | |
803 | ||
804 | BUFFER_TRACE(block_bitmap_bh, "dirty block bitmap"); | |
805 | err = ext4_handle_dirty_metadata(NULL, NULL, block_bitmap_bh); | |
806 | sync_dirty_buffer(block_bitmap_bh); | |
807 | ||
808 | /* recheck and clear flag under lock if we still need to */ | |
809 | ext4_lock_group(sb, group); | |
810 | if (ext4_has_group_desc_csum(sb) && | |
811 | (gdp->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT))) { | |
812 | gdp->bg_flags &= cpu_to_le16(~EXT4_BG_BLOCK_UNINIT); | |
813 | ext4_free_group_clusters_set(sb, gdp, | |
814 | ext4_free_clusters_after_init(sb, group, gdp)); | |
815 | ext4_block_bitmap_csum_set(sb, group, gdp, | |
816 | block_bitmap_bh); | |
817 | ext4_group_desc_csum_set(sb, group, gdp); | |
818 | } | |
819 | ext4_unlock_group(sb, group); | |
820 | brelse(block_bitmap_bh); | |
821 | ||
822 | if (err) { | |
823 | ext4_std_error(sb, err); | |
824 | goto out; | |
825 | } | |
826 | } | |
827 | ||
828 | /* Update the relevant bg descriptor fields */ | |
829 | if (ext4_has_group_desc_csum(sb)) { | |
830 | int free; | |
831 | ||
832 | ext4_lock_group(sb, group); /* while we modify the bg desc */ | |
833 | free = EXT4_INODES_PER_GROUP(sb) - | |
834 | ext4_itable_unused_count(sb, gdp); | |
835 | if (gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_UNINIT)) { | |
836 | gdp->bg_flags &= cpu_to_le16(~EXT4_BG_INODE_UNINIT); | |
837 | free = 0; | |
838 | } | |
839 | ||
840 | /* | |
841 | * Check the relative inode number against the last used | |
842 | * relative inode number in this group. if it is greater | |
843 | * we need to update the bg_itable_unused count | |
844 | */ | |
845 | if (bit >= free) | |
846 | ext4_itable_unused_set(sb, gdp, | |
847 | (EXT4_INODES_PER_GROUP(sb) - bit - 1)); | |
848 | } else { | |
849 | ext4_lock_group(sb, group); | |
850 | } | |
851 | ||
852 | ext4_free_inodes_set(sb, gdp, ext4_free_inodes_count(sb, gdp) - 1); | |
853 | if (ext4_has_group_desc_csum(sb)) { | |
854 | ext4_inode_bitmap_csum_set(sb, group, gdp, inode_bitmap_bh, | |
855 | EXT4_INODES_PER_GROUP(sb) / 8); | |
856 | ext4_group_desc_csum_set(sb, group, gdp); | |
857 | } | |
858 | ||
859 | ext4_unlock_group(sb, group); | |
860 | err = ext4_handle_dirty_metadata(NULL, NULL, group_desc_bh); | |
861 | sync_dirty_buffer(group_desc_bh); | |
862 | out: | |
863 | return err; | |
864 | } | |
865 | ||
177cc0e7 EB |
866 | static int ext4_xattr_credits_for_new_inode(struct inode *dir, mode_t mode, |
867 | bool encrypt) | |
868 | { | |
869 | struct super_block *sb = dir->i_sb; | |
870 | int nblocks = 0; | |
871 | #ifdef CONFIG_EXT4_FS_POSIX_ACL | |
872 | struct posix_acl *p = get_acl(dir, ACL_TYPE_DEFAULT); | |
873 | ||
874 | if (IS_ERR(p)) | |
875 | return PTR_ERR(p); | |
876 | if (p) { | |
877 | int acl_size = p->a_count * sizeof(ext4_acl_entry); | |
878 | ||
879 | nblocks += (S_ISDIR(mode) ? 2 : 1) * | |
880 | __ext4_xattr_set_credits(sb, NULL /* inode */, | |
881 | NULL /* block_bh */, acl_size, | |
882 | true /* is_create */); | |
883 | posix_acl_release(p); | |
884 | } | |
885 | #endif | |
886 | ||
887 | #ifdef CONFIG_SECURITY | |
888 | { | |
889 | int num_security_xattrs = 1; | |
890 | ||
891 | #ifdef CONFIG_INTEGRITY | |
892 | num_security_xattrs++; | |
893 | #endif | |
894 | /* | |
895 | * We assume that security xattrs are never more than 1k. | |
896 | * In practice they are under 128 bytes. | |
897 | */ | |
898 | nblocks += num_security_xattrs * | |
899 | __ext4_xattr_set_credits(sb, NULL /* inode */, | |
900 | NULL /* block_bh */, 1024, | |
901 | true /* is_create */); | |
902 | } | |
903 | #endif | |
904 | if (encrypt) | |
905 | nblocks += __ext4_xattr_set_credits(sb, | |
906 | NULL /* inode */, | |
907 | NULL /* block_bh */, | |
908 | FSCRYPT_SET_CONTEXT_MAX_SIZE, | |
909 | true /* is_create */); | |
910 | return nblocks; | |
911 | } | |
912 | ||
ac27a0ec DK |
913 | /* |
914 | * There are two policies for allocating an inode. If the new inode is | |
915 | * a directory, then a forward search is made for a block group with both | |
916 | * free space and a low directory-to-inode ratio; if that fails, then of | |
917 | * the groups with above-average free space, that group with the fewest | |
918 | * directories already is chosen. | |
919 | * | |
920 | * For other inodes, search forward from the parent directory's block | |
921 | * group to find a free inode. | |
922 | */ | |
14f3db55 CB |
923 | struct inode *__ext4_new_inode(struct user_namespace *mnt_userns, |
924 | handle_t *handle, struct inode *dir, | |
1139575a | 925 | umode_t mode, const struct qstr *qstr, |
1b917ed8 TE |
926 | __u32 goal, uid_t *owner, __u32 i_flags, |
927 | int handle_type, unsigned int line_no, | |
928 | int nblocks) | |
ac27a0ec DK |
929 | { |
930 | struct super_block *sb; | |
3300beda AK |
931 | struct buffer_head *inode_bitmap_bh = NULL; |
932 | struct buffer_head *group_desc_bh; | |
8df9675f | 933 | ext4_group_t ngroups, group = 0; |
ac27a0ec | 934 | unsigned long ino = 0; |
af5bc92d TT |
935 | struct inode *inode; |
936 | struct ext4_group_desc *gdp = NULL; | |
617ba13b MC |
937 | struct ext4_inode_info *ei; |
938 | struct ext4_sb_info *sbi; | |
a7cdadee | 939 | int ret2, err; |
ac27a0ec | 940 | struct inode *ret; |
2aa9fc4c | 941 | ext4_group_t i; |
772cb7c8 | 942 | ext4_group_t flex_group; |
8016e29f | 943 | struct ext4_group_info *grp = NULL; |
02ce5316 | 944 | bool encrypt = false; |
ac27a0ec DK |
945 | |
946 | /* Cannot create files in a deleted directory */ | |
947 | if (!dir || !dir->i_nlink) | |
948 | return ERR_PTR(-EPERM); | |
949 | ||
af65207c TE |
950 | sb = dir->i_sb; |
951 | sbi = EXT4_SB(sb); | |
952 | ||
953 | if (unlikely(ext4_forced_shutdown(sbi))) | |
0db1ff22 TT |
954 | return ERR_PTR(-EIO); |
955 | ||
8df9675f | 956 | ngroups = ext4_get_groups_count(sb); |
9bffad1e | 957 | trace_ext4_request_inode(dir, mode); |
ac27a0ec DK |
958 | inode = new_inode(sb); |
959 | if (!inode) | |
960 | return ERR_PTR(-ENOMEM); | |
617ba13b | 961 | ei = EXT4_I(inode); |
772cb7c8 | 962 | |
eb9cc7e1 | 963 | /* |
b8a07463 | 964 | * Initialize owners and quota early so that we don't have to account |
eb9cc7e1 JK |
965 | * for quota initialization worst case in standard inode creating |
966 | * transaction | |
967 | */ | |
968 | if (owner) { | |
969 | inode->i_mode = mode; | |
970 | i_uid_write(inode, owner[0]); | |
971 | i_gid_write(inode, owner[1]); | |
972 | } else if (test_opt(sb, GRPID)) { | |
973 | inode->i_mode = mode; | |
db998553 | 974 | inode_fsuid_set(inode, mnt_userns); |
eb9cc7e1 JK |
975 | inode->i_gid = dir->i_gid; |
976 | } else | |
14f3db55 | 977 | inode_init_owner(mnt_userns, inode, dir, mode); |
040cb378 | 978 | |
0b7b7779 | 979 | if (ext4_has_feature_project(sb) && |
040cb378 LX |
980 | ext4_test_inode_flag(dir, EXT4_INODE_PROJINHERIT)) |
981 | ei->i_projid = EXT4_I(dir)->i_projid; | |
982 | else | |
983 | ei->i_projid = make_kprojid(&init_user_ns, EXT4_DEF_PROJID); | |
984 | ||
02ce5316 EB |
985 | if (!(i_flags & EXT4_EA_INODE_FL)) { |
986 | err = fscrypt_prepare_new_inode(dir, inode, &encrypt); | |
987 | if (err) | |
988 | goto out; | |
989 | } | |
990 | ||
a7cdadee JK |
991 | err = dquot_initialize(inode); |
992 | if (err) | |
993 | goto out; | |
eb9cc7e1 | 994 | |
02ce5316 EB |
995 | if (!handle && sbi->s_journal && !(i_flags & EXT4_EA_INODE_FL)) { |
996 | ret2 = ext4_xattr_credits_for_new_inode(dir, mode, encrypt); | |
997 | if (ret2 < 0) { | |
998 | err = ret2; | |
999 | goto out; | |
1000 | } | |
1001 | nblocks += ret2; | |
1002 | } | |
1003 | ||
11013911 AD |
1004 | if (!goal) |
1005 | goal = sbi->s_inode_goal; | |
1006 | ||
e6462869 | 1007 | if (goal && goal <= le32_to_cpu(sbi->s_es->s_inodes_count)) { |
11013911 AD |
1008 | group = (goal - 1) / EXT4_INODES_PER_GROUP(sb); |
1009 | ino = (goal - 1) % EXT4_INODES_PER_GROUP(sb); | |
1010 | ret2 = 0; | |
1011 | goto got_group; | |
1012 | } | |
1013 | ||
4113c4ca LC |
1014 | if (S_ISDIR(mode)) |
1015 | ret2 = find_group_orlov(sb, dir, &group, mode, qstr); | |
1016 | else | |
a4912123 | 1017 | ret2 = find_group_other(sb, dir, &group, mode); |
ac27a0ec | 1018 | |
772cb7c8 | 1019 | got_group: |
a4912123 | 1020 | EXT4_I(dir)->i_last_alloc_group = group; |
ac27a0ec | 1021 | err = -ENOSPC; |
2aa9fc4c | 1022 | if (ret2 == -1) |
ac27a0ec DK |
1023 | goto out; |
1024 | ||
119c0d44 TT |
1025 | /* |
1026 | * Normally we will only go through one pass of this loop, | |
1027 | * unless we get unlucky and it turns out the group we selected | |
1028 | * had its last inode grabbed by someone else. | |
1029 | */ | |
11013911 | 1030 | for (i = 0; i < ngroups; i++, ino = 0) { |
ac27a0ec DK |
1031 | err = -EIO; |
1032 | ||
3300beda | 1033 | gdp = ext4_get_group_desc(sb, group, &group_desc_bh); |
ac27a0ec | 1034 | if (!gdp) |
eb9cc7e1 | 1035 | goto out; |
ac27a0ec | 1036 | |
f2a09af6 YY |
1037 | /* |
1038 | * Check free inodes count before loading bitmap. | |
1039 | */ | |
2fe435d8 WS |
1040 | if (ext4_free_inodes_count(sb, gdp) == 0) |
1041 | goto next_group; | |
f2a09af6 | 1042 | |
8016e29f HS |
1043 | if (!(sbi->s_mount_state & EXT4_FC_REPLAY)) { |
1044 | grp = ext4_get_group_info(sb, group); | |
1045 | /* | |
1046 | * Skip groups with already-known suspicious inode | |
1047 | * tables | |
1048 | */ | |
1049 | if (EXT4_MB_GRP_IBITMAP_CORRUPT(grp)) | |
1050 | goto next_group; | |
1051 | } | |
87a39389 | 1052 | |
3300beda AK |
1053 | brelse(inode_bitmap_bh); |
1054 | inode_bitmap_bh = ext4_read_inode_bitmap(sb, group); | |
87a39389 | 1055 | /* Skip groups with suspicious inode tables */ |
8016e29f HS |
1056 | if (((!(sbi->s_mount_state & EXT4_FC_REPLAY)) |
1057 | && EXT4_MB_GRP_IBITMAP_CORRUPT(grp)) || | |
9008a58e DW |
1058 | IS_ERR(inode_bitmap_bh)) { |
1059 | inode_bitmap_bh = NULL; | |
2fe435d8 | 1060 | goto next_group; |
87a39389 | 1061 | } |
ac27a0ec | 1062 | |
ac27a0ec | 1063 | repeat_in_this_group: |
901ed070 WS |
1064 | ret2 = find_inode_bit(sb, group, inode_bitmap_bh, &ino); |
1065 | if (!ret2) | |
a34eb503 | 1066 | goto next_group; |
901ed070 WS |
1067 | |
1068 | if (group == 0 && (ino + 1) < EXT4_FIRST_INO(sb)) { | |
119c0d44 TT |
1069 | ext4_error(sb, "reserved inode found cleared - " |
1070 | "inode=%lu", ino + 1); | |
206f6d55 WS |
1071 | ext4_mark_group_bitmap_corrupted(sb, group, |
1072 | EXT4_GROUP_INFO_IBITMAP_CORRUPT); | |
2fe435d8 | 1073 | goto next_group; |
119c0d44 | 1074 | } |
901ed070 | 1075 | |
8016e29f | 1076 | if ((!(sbi->s_mount_state & EXT4_FC_REPLAY)) && !handle) { |
1139575a TT |
1077 | BUG_ON(nblocks <= 0); |
1078 | handle = __ext4_journal_start_sb(dir->i_sb, line_no, | |
a70fd5ac | 1079 | handle_type, nblocks, 0, |
1080 | ext4_trans_default_revoke_credits(sb)); | |
1139575a TT |
1081 | if (IS_ERR(handle)) { |
1082 | err = PTR_ERR(handle); | |
eb9cc7e1 JK |
1083 | ext4_std_error(sb, err); |
1084 | goto out; | |
1139575a TT |
1085 | } |
1086 | } | |
ffb5387e ES |
1087 | BUFFER_TRACE(inode_bitmap_bh, "get_write_access"); |
1088 | err = ext4_journal_get_write_access(handle, inode_bitmap_bh); | |
eb9cc7e1 JK |
1089 | if (err) { |
1090 | ext4_std_error(sb, err); | |
1091 | goto out; | |
1092 | } | |
119c0d44 TT |
1093 | ext4_lock_group(sb, group); |
1094 | ret2 = ext4_test_and_set_bit(ino, inode_bitmap_bh->b_data); | |
901ed070 WS |
1095 | if (ret2) { |
1096 | /* Someone already took the bit. Repeat the search | |
1097 | * with lock held. | |
1098 | */ | |
1099 | ret2 = find_inode_bit(sb, group, inode_bitmap_bh, &ino); | |
1100 | if (ret2) { | |
1101 | ext4_set_bit(ino, inode_bitmap_bh->b_data); | |
1102 | ret2 = 0; | |
1103 | } else { | |
1104 | ret2 = 1; /* we didn't grab the inode */ | |
1105 | } | |
1106 | } | |
119c0d44 TT |
1107 | ext4_unlock_group(sb, group); |
1108 | ino++; /* the inode bitmap is zero-based */ | |
1109 | if (!ret2) | |
1110 | goto got; /* we grabbed the inode! */ | |
901ed070 | 1111 | |
119c0d44 TT |
1112 | if (ino < EXT4_INODES_PER_GROUP(sb)) |
1113 | goto repeat_in_this_group; | |
a34eb503 TT |
1114 | next_group: |
1115 | if (++group == ngroups) | |
1116 | group = 0; | |
ac27a0ec DK |
1117 | } |
1118 | err = -ENOSPC; | |
1119 | goto out; | |
1120 | ||
1121 | got: | |
ffb5387e ES |
1122 | BUFFER_TRACE(inode_bitmap_bh, "call ext4_handle_dirty_metadata"); |
1123 | err = ext4_handle_dirty_metadata(handle, NULL, inode_bitmap_bh); | |
eb9cc7e1 JK |
1124 | if (err) { |
1125 | ext4_std_error(sb, err); | |
1126 | goto out; | |
1127 | } | |
ffb5387e | 1128 | |
61c219f5 TT |
1129 | BUFFER_TRACE(group_desc_bh, "get_write_access"); |
1130 | err = ext4_journal_get_write_access(handle, group_desc_bh); | |
1131 | if (err) { | |
1132 | ext4_std_error(sb, err); | |
1133 | goto out; | |
1134 | } | |
1135 | ||
717d50e4 | 1136 | /* We may have to initialize the block bitmap if it isn't already */ |
feb0ab32 | 1137 | if (ext4_has_group_desc_csum(sb) && |
717d50e4 | 1138 | gdp->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) { |
3300beda | 1139 | struct buffer_head *block_bitmap_bh; |
717d50e4 | 1140 | |
3300beda | 1141 | block_bitmap_bh = ext4_read_block_bitmap(sb, group); |
9008a58e DW |
1142 | if (IS_ERR(block_bitmap_bh)) { |
1143 | err = PTR_ERR(block_bitmap_bh); | |
599a9b77 JK |
1144 | goto out; |
1145 | } | |
3300beda AK |
1146 | BUFFER_TRACE(block_bitmap_bh, "get block bitmap access"); |
1147 | err = ext4_journal_get_write_access(handle, block_bitmap_bh); | |
717d50e4 | 1148 | if (err) { |
3300beda | 1149 | brelse(block_bitmap_bh); |
eb9cc7e1 JK |
1150 | ext4_std_error(sb, err); |
1151 | goto out; | |
717d50e4 AD |
1152 | } |
1153 | ||
fd034a84 TT |
1154 | BUFFER_TRACE(block_bitmap_bh, "dirty block bitmap"); |
1155 | err = ext4_handle_dirty_metadata(handle, NULL, block_bitmap_bh); | |
fd034a84 | 1156 | |
717d50e4 | 1157 | /* recheck and clear flag under lock if we still need to */ |
fd034a84 | 1158 | ext4_lock_group(sb, group); |
8844618d TT |
1159 | if (ext4_has_group_desc_csum(sb) && |
1160 | (gdp->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT))) { | |
3300beda | 1161 | gdp->bg_flags &= cpu_to_le16(~EXT4_BG_BLOCK_UNINIT); |
021b65bb | 1162 | ext4_free_group_clusters_set(sb, gdp, |
cff1dfd7 | 1163 | ext4_free_clusters_after_init(sb, group, gdp)); |
fa77dcfa | 1164 | ext4_block_bitmap_csum_set(sb, group, gdp, |
79f1ba49 | 1165 | block_bitmap_bh); |
feb0ab32 | 1166 | ext4_group_desc_csum_set(sb, group, gdp); |
717d50e4 | 1167 | } |
955ce5f5 | 1168 | ext4_unlock_group(sb, group); |
aeb1e5d6 | 1169 | brelse(block_bitmap_bh); |
717d50e4 | 1170 | |
eb9cc7e1 JK |
1171 | if (err) { |
1172 | ext4_std_error(sb, err); | |
1173 | goto out; | |
1174 | } | |
717d50e4 | 1175 | } |
119c0d44 | 1176 | |
119c0d44 | 1177 | /* Update the relevant bg descriptor fields */ |
41a246d1 | 1178 | if (ext4_has_group_desc_csum(sb)) { |
119c0d44 | 1179 | int free; |
8016e29f HS |
1180 | struct ext4_group_info *grp = NULL; |
1181 | ||
1182 | if (!(sbi->s_mount_state & EXT4_FC_REPLAY)) { | |
1183 | grp = ext4_get_group_info(sb, group); | |
1184 | down_read(&grp->alloc_sem); /* | |
1185 | * protect vs itable | |
1186 | * lazyinit | |
1187 | */ | |
1188 | } | |
119c0d44 TT |
1189 | ext4_lock_group(sb, group); /* while we modify the bg desc */ |
1190 | free = EXT4_INODES_PER_GROUP(sb) - | |
1191 | ext4_itable_unused_count(sb, gdp); | |
1192 | if (gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_UNINIT)) { | |
1193 | gdp->bg_flags &= cpu_to_le16(~EXT4_BG_INODE_UNINIT); | |
1194 | free = 0; | |
1195 | } | |
1196 | /* | |
1197 | * Check the relative inode number against the last used | |
1198 | * relative inode number in this group. if it is greater | |
1199 | * we need to update the bg_itable_unused count | |
1200 | */ | |
1201 | if (ino > free) | |
1202 | ext4_itable_unused_set(sb, gdp, | |
1203 | (EXT4_INODES_PER_GROUP(sb) - ino)); | |
8016e29f HS |
1204 | if (!(sbi->s_mount_state & EXT4_FC_REPLAY)) |
1205 | up_read(&grp->alloc_sem); | |
6f2e9f0e TM |
1206 | } else { |
1207 | ext4_lock_group(sb, group); | |
119c0d44 | 1208 | } |
6f2e9f0e | 1209 | |
119c0d44 TT |
1210 | ext4_free_inodes_set(sb, gdp, ext4_free_inodes_count(sb, gdp) - 1); |
1211 | if (S_ISDIR(mode)) { | |
1212 | ext4_used_dirs_set(sb, gdp, ext4_used_dirs_count(sb, gdp) + 1); | |
1213 | if (sbi->s_log_groups_per_flex) { | |
1214 | ext4_group_t f = ext4_flex_group(sbi, group); | |
1215 | ||
7c990728 SJS |
1216 | atomic_inc(&sbi_array_rcu_deref(sbi, s_flex_groups, |
1217 | f)->used_dirs); | |
119c0d44 TT |
1218 | } |
1219 | } | |
41a246d1 DW |
1220 | if (ext4_has_group_desc_csum(sb)) { |
1221 | ext4_inode_bitmap_csum_set(sb, group, gdp, inode_bitmap_bh, | |
1222 | EXT4_INODES_PER_GROUP(sb) / 8); | |
feb0ab32 | 1223 | ext4_group_desc_csum_set(sb, group, gdp); |
119c0d44 | 1224 | } |
6f2e9f0e | 1225 | ext4_unlock_group(sb, group); |
119c0d44 | 1226 | |
3300beda AK |
1227 | BUFFER_TRACE(group_desc_bh, "call ext4_handle_dirty_metadata"); |
1228 | err = ext4_handle_dirty_metadata(handle, NULL, group_desc_bh); | |
eb9cc7e1 JK |
1229 | if (err) { |
1230 | ext4_std_error(sb, err); | |
1231 | goto out; | |
1232 | } | |
ac27a0ec DK |
1233 | |
1234 | percpu_counter_dec(&sbi->s_freeinodes_counter); | |
1235 | if (S_ISDIR(mode)) | |
1236 | percpu_counter_inc(&sbi->s_dirs_counter); | |
ac27a0ec | 1237 | |
772cb7c8 JS |
1238 | if (sbi->s_log_groups_per_flex) { |
1239 | flex_group = ext4_flex_group(sbi, group); | |
7c990728 SJS |
1240 | atomic_dec(&sbi_array_rcu_deref(sbi, s_flex_groups, |
1241 | flex_group)->free_inodes); | |
772cb7c8 | 1242 | } |
ac27a0ec | 1243 | |
717d50e4 | 1244 | inode->i_ino = ino + group * EXT4_INODES_PER_GROUP(sb); |
ac27a0ec DK |
1245 | /* This is the optimal IO size (for stat), not the fs block size */ |
1246 | inode->i_blocks = 0; | |
95582b00 | 1247 | inode->i_mtime = inode->i_atime = inode->i_ctime = current_time(inode); |
7b62b293 | 1248 | ei->i_crtime = inode->i_mtime; |
ac27a0ec DK |
1249 | |
1250 | memset(ei->i_data, 0, sizeof(ei->i_data)); | |
1251 | ei->i_dir_start_lookup = 0; | |
1252 | ei->i_disksize = 0; | |
1253 | ||
4af83508 | 1254 | /* Don't inherit extent flag from directory, amongst others. */ |
2dc6b0d4 DG |
1255 | ei->i_flags = |
1256 | ext4_mask_flags(mode, EXT4_I(dir)->i_flags & EXT4_FL_INHERITED); | |
1b917ed8 | 1257 | ei->i_flags |= i_flags; |
ac27a0ec | 1258 | ei->i_file_acl = 0; |
ac27a0ec | 1259 | ei->i_dtime = 0; |
ac27a0ec | 1260 | ei->i_block_group = group; |
a4912123 | 1261 | ei->i_last_alloc_group = ~0; |
ac27a0ec | 1262 | |
043546e4 | 1263 | ext4_set_inode_flags(inode, true); |
ac27a0ec | 1264 | if (IS_DIRSYNC(inode)) |
0390131b | 1265 | ext4_handle_sync(handle); |
6b38e842 | 1266 | if (insert_inode_locked(inode) < 0) { |
acd6ad83 JK |
1267 | /* |
1268 | * Likely a bitmap corruption causing inode to be allocated | |
1269 | * twice. | |
1270 | */ | |
1271 | err = -EIO; | |
eb9cc7e1 JK |
1272 | ext4_error(sb, "failed to insert inode %lu: doubly allocated?", |
1273 | inode->i_ino); | |
206f6d55 WS |
1274 | ext4_mark_group_bitmap_corrupted(sb, group, |
1275 | EXT4_GROUP_INFO_IBITMAP_CORRUPT); | |
eb9cc7e1 | 1276 | goto out; |
6b38e842 | 1277 | } |
23253068 | 1278 | inode->i_generation = prandom_u32(); |
ac27a0ec | 1279 | |
814525f4 | 1280 | /* Precompute checksum seed for inode metadata */ |
9aa5d32b | 1281 | if (ext4_has_metadata_csum(sb)) { |
814525f4 | 1282 | __u32 csum; |
814525f4 DW |
1283 | __le32 inum = cpu_to_le32(inode->i_ino); |
1284 | __le32 gen = cpu_to_le32(inode->i_generation); | |
1285 | csum = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&inum, | |
1286 | sizeof(inum)); | |
1287 | ei->i_csum_seed = ext4_chksum(sbi, csum, (__u8 *)&gen, | |
1288 | sizeof(gen)); | |
1289 | } | |
1290 | ||
353eb83c | 1291 | ext4_clear_state_flags(ei); /* Only relevant on 32-bit archs */ |
19f5fb7a | 1292 | ext4_set_inode_state(inode, EXT4_STATE_NEW); |
ef7f3835 | 1293 | |
49598e04 | 1294 | ei->i_extra_isize = sbi->s_want_extra_isize; |
f08225d1 | 1295 | ei->i_inline_off = 0; |
4811d992 TT |
1296 | if (ext4_has_feature_inline_data(sb) && |
1297 | (!(ei->i_flags & EXT4_DAX_FL) || S_ISDIR(mode))) | |
f08225d1 | 1298 | ext4_set_inode_state(inode, EXT4_STATE_MAY_INLINE_DATA); |
ac27a0ec | 1299 | ret = inode; |
63936dda CH |
1300 | err = dquot_alloc_inode(inode); |
1301 | if (err) | |
ac27a0ec | 1302 | goto fail_drop; |
ac27a0ec | 1303 | |
aa1dca3b EB |
1304 | /* |
1305 | * Since the encryption xattr will always be unique, create it first so | |
1306 | * that it's less likely to end up in an external xattr block and | |
1307 | * prevent its deduplication. | |
1308 | */ | |
1309 | if (encrypt) { | |
02ce5316 | 1310 | err = fscrypt_set_context(inode, handle); |
aa1dca3b EB |
1311 | if (err) |
1312 | goto fail_free_drop; | |
1313 | } | |
1314 | ||
1b917ed8 TE |
1315 | if (!(ei->i_flags & EXT4_EA_INODE_FL)) { |
1316 | err = ext4_init_acl(handle, inode, dir); | |
1317 | if (err) | |
1318 | goto fail_free_drop; | |
ac27a0ec | 1319 | |
ad47f953 TE |
1320 | err = ext4_init_security(handle, inode, dir, qstr); |
1321 | if (err) | |
1322 | goto fail_free_drop; | |
1323 | } | |
ac27a0ec | 1324 | |
e2b911c5 | 1325 | if (ext4_has_feature_extents(sb)) { |
e4079a11 | 1326 | /* set extent flag only for directory, file and normal symlink*/ |
e65187e6 | 1327 | if (S_ISDIR(mode) || S_ISREG(mode) || S_ISLNK(mode)) { |
12e9b892 | 1328 | ext4_set_inode_flag(inode, EXT4_INODE_EXTENTS); |
42bf0383 | 1329 | ext4_ext_tree_init(handle, inode); |
42bf0383 | 1330 | } |
a86c6181 | 1331 | } |
ac27a0ec | 1332 | |
688f869c TT |
1333 | if (ext4_handle_valid(handle)) { |
1334 | ei->i_sync_tid = handle->h_transaction->t_tid; | |
1335 | ei->i_datasync_tid = handle->h_transaction->t_tid; | |
1336 | } | |
1337 | ||
8753e88f AK |
1338 | err = ext4_mark_inode_dirty(handle, inode); |
1339 | if (err) { | |
1340 | ext4_std_error(sb, err); | |
1341 | goto fail_free_drop; | |
1342 | } | |
1343 | ||
617ba13b | 1344 | ext4_debug("allocating inode %lu\n", inode->i_ino); |
9bffad1e | 1345 | trace_ext4_allocate_inode(inode, dir, mode); |
3300beda | 1346 | brelse(inode_bitmap_bh); |
ac27a0ec DK |
1347 | return ret; |
1348 | ||
1349 | fail_free_drop: | |
63936dda | 1350 | dquot_free_inode(inode); |
ac27a0ec | 1351 | fail_drop: |
6d6b77f1 | 1352 | clear_nlink(inode); |
6b38e842 | 1353 | unlock_new_inode(inode); |
eb9cc7e1 JK |
1354 | out: |
1355 | dquot_drop(inode); | |
1356 | inode->i_flags |= S_NOQUOTA; | |
ac27a0ec | 1357 | iput(inode); |
3300beda | 1358 | brelse(inode_bitmap_bh); |
ac27a0ec DK |
1359 | return ERR_PTR(err); |
1360 | } | |
1361 | ||
1362 | /* Verify that we are loading a valid orphan from disk */ | |
617ba13b | 1363 | struct inode *ext4_orphan_get(struct super_block *sb, unsigned long ino) |
ac27a0ec | 1364 | { |
617ba13b | 1365 | unsigned long max_ino = le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count); |
fd2d4291 | 1366 | ext4_group_t block_group; |
ac27a0ec | 1367 | int bit; |
7827a7f6 | 1368 | struct buffer_head *bitmap_bh = NULL; |
ac27a0ec | 1369 | struct inode *inode = NULL; |
7827a7f6 | 1370 | int err = -EFSCORRUPTED; |
ac27a0ec | 1371 | |
7827a7f6 TT |
1372 | if (ino < EXT4_FIRST_INO(sb) || ino > max_ino) |
1373 | goto bad_orphan; | |
ac27a0ec | 1374 | |
617ba13b MC |
1375 | block_group = (ino - 1) / EXT4_INODES_PER_GROUP(sb); |
1376 | bit = (ino - 1) % EXT4_INODES_PER_GROUP(sb); | |
e29d1cde | 1377 | bitmap_bh = ext4_read_inode_bitmap(sb, block_group); |
206f6d55 | 1378 | if (IS_ERR(bitmap_bh)) |
e884bce1 | 1379 | return ERR_CAST(bitmap_bh); |
ac27a0ec DK |
1380 | |
1381 | /* Having the inode bit set should be a 100% indicator that this | |
1382 | * is a valid orphan (no e2fsck run on fs). Orphans also include | |
1383 | * inodes that were being truncated, so we can't check i_nlink==0. | |
1384 | */ | |
1d1fe1ee DH |
1385 | if (!ext4_test_bit(bit, bitmap_bh->b_data)) |
1386 | goto bad_orphan; | |
1387 | ||
8a363970 | 1388 | inode = ext4_iget(sb, ino, EXT4_IGET_NORMAL); |
7827a7f6 TT |
1389 | if (IS_ERR(inode)) { |
1390 | err = PTR_ERR(inode); | |
54d3adbc TT |
1391 | ext4_error_err(sb, -err, |
1392 | "couldn't read orphan inode %lu (err %d)", | |
1393 | ino, err); | |
3bbd0ef2 | 1394 | brelse(bitmap_bh); |
7827a7f6 TT |
1395 | return inode; |
1396 | } | |
1d1fe1ee | 1397 | |
91ef4caf | 1398 | /* |
c9eb13a9 TT |
1399 | * If the orphans has i_nlinks > 0 then it should be able to |
1400 | * be truncated, otherwise it won't be removed from the orphan | |
1401 | * list during processing and an infinite loop will result. | |
1402 | * Similarly, it must not be a bad inode. | |
91ef4caf | 1403 | */ |
c9eb13a9 TT |
1404 | if ((inode->i_nlink && !ext4_can_truncate(inode)) || |
1405 | is_bad_inode(inode)) | |
91ef4caf DG |
1406 | goto bad_orphan; |
1407 | ||
1d1fe1ee DH |
1408 | if (NEXT_ORPHAN(inode) > max_ino) |
1409 | goto bad_orphan; | |
1410 | brelse(bitmap_bh); | |
1411 | return inode; | |
1412 | ||
1d1fe1ee | 1413 | bad_orphan: |
7827a7f6 TT |
1414 | ext4_error(sb, "bad orphan inode %lu", ino); |
1415 | if (bitmap_bh) | |
1416 | printk(KERN_ERR "ext4_test_bit(bit=%d, block=%llu) = %d\n", | |
1417 | bit, (unsigned long long)bitmap_bh->b_blocknr, | |
1418 | ext4_test_bit(bit, bitmap_bh->b_data)); | |
1d1fe1ee | 1419 | if (inode) { |
7827a7f6 | 1420 | printk(KERN_ERR "is_bad_inode(inode)=%d\n", |
1d1fe1ee | 1421 | is_bad_inode(inode)); |
7827a7f6 | 1422 | printk(KERN_ERR "NEXT_ORPHAN(inode)=%u\n", |
1d1fe1ee | 1423 | NEXT_ORPHAN(inode)); |
7827a7f6 TT |
1424 | printk(KERN_ERR "max_ino=%lu\n", max_ino); |
1425 | printk(KERN_ERR "i_nlink=%u\n", inode->i_nlink); | |
ac27a0ec | 1426 | /* Avoid freeing blocks if we got a bad deleted inode */ |
1d1fe1ee | 1427 | if (inode->i_nlink == 0) |
ac27a0ec DK |
1428 | inode->i_blocks = 0; |
1429 | iput(inode); | |
ac27a0ec | 1430 | } |
ac27a0ec | 1431 | brelse(bitmap_bh); |
1d1fe1ee | 1432 | return ERR_PTR(err); |
ac27a0ec DK |
1433 | } |
1434 | ||
af5bc92d | 1435 | unsigned long ext4_count_free_inodes(struct super_block *sb) |
ac27a0ec DK |
1436 | { |
1437 | unsigned long desc_count; | |
617ba13b | 1438 | struct ext4_group_desc *gdp; |
8df9675f | 1439 | ext4_group_t i, ngroups = ext4_get_groups_count(sb); |
617ba13b MC |
1440 | #ifdef EXT4FS_DEBUG |
1441 | struct ext4_super_block *es; | |
ac27a0ec DK |
1442 | unsigned long bitmap_count, x; |
1443 | struct buffer_head *bitmap_bh = NULL; | |
1444 | ||
617ba13b | 1445 | es = EXT4_SB(sb)->s_es; |
ac27a0ec DK |
1446 | desc_count = 0; |
1447 | bitmap_count = 0; | |
1448 | gdp = NULL; | |
8df9675f | 1449 | for (i = 0; i < ngroups; i++) { |
af5bc92d | 1450 | gdp = ext4_get_group_desc(sb, i, NULL); |
ac27a0ec DK |
1451 | if (!gdp) |
1452 | continue; | |
560671a0 | 1453 | desc_count += ext4_free_inodes_count(sb, gdp); |
ac27a0ec | 1454 | brelse(bitmap_bh); |
e29d1cde | 1455 | bitmap_bh = ext4_read_inode_bitmap(sb, i); |
9008a58e DW |
1456 | if (IS_ERR(bitmap_bh)) { |
1457 | bitmap_bh = NULL; | |
ac27a0ec | 1458 | continue; |
9008a58e | 1459 | } |
ac27a0ec | 1460 | |
f6fb99ca TT |
1461 | x = ext4_count_free(bitmap_bh->b_data, |
1462 | EXT4_INODES_PER_GROUP(sb) / 8); | |
c549a95d | 1463 | printk(KERN_DEBUG "group %lu: stored = %d, counted = %lu\n", |
785b4b3a | 1464 | (unsigned long) i, ext4_free_inodes_count(sb, gdp), x); |
ac27a0ec DK |
1465 | bitmap_count += x; |
1466 | } | |
1467 | brelse(bitmap_bh); | |
4776004f TT |
1468 | printk(KERN_DEBUG "ext4_count_free_inodes: " |
1469 | "stored = %u, computed = %lu, %lu\n", | |
1470 | le32_to_cpu(es->s_free_inodes_count), desc_count, bitmap_count); | |
ac27a0ec DK |
1471 | return desc_count; |
1472 | #else | |
1473 | desc_count = 0; | |
8df9675f | 1474 | for (i = 0; i < ngroups; i++) { |
af5bc92d | 1475 | gdp = ext4_get_group_desc(sb, i, NULL); |
ac27a0ec DK |
1476 | if (!gdp) |
1477 | continue; | |
560671a0 | 1478 | desc_count += ext4_free_inodes_count(sb, gdp); |
ac27a0ec DK |
1479 | cond_resched(); |
1480 | } | |
1481 | return desc_count; | |
1482 | #endif | |
1483 | } | |
1484 | ||
1485 | /* Called at mount-time, super-block is locked */ | |
af5bc92d | 1486 | unsigned long ext4_count_dirs(struct super_block * sb) |
ac27a0ec DK |
1487 | { |
1488 | unsigned long count = 0; | |
8df9675f | 1489 | ext4_group_t i, ngroups = ext4_get_groups_count(sb); |
ac27a0ec | 1490 | |
8df9675f | 1491 | for (i = 0; i < ngroups; i++) { |
af5bc92d | 1492 | struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); |
ac27a0ec DK |
1493 | if (!gdp) |
1494 | continue; | |
560671a0 | 1495 | count += ext4_used_dirs_count(sb, gdp); |
ac27a0ec DK |
1496 | } |
1497 | return count; | |
1498 | } | |
bfff6873 LC |
1499 | |
1500 | /* | |
1501 | * Zeroes not yet zeroed inode table - just write zeroes through the whole | |
1502 | * inode table. Must be called without any spinlock held. The only place | |
1503 | * where it is called from on active part of filesystem is ext4lazyinit | |
1504 | * thread, so we do not need any special locks, however we have to prevent | |
1505 | * inode allocation from the current group, so we take alloc_sem lock, to | |
119c0d44 | 1506 | * block ext4_new_inode() until we are finished. |
bfff6873 | 1507 | */ |
e0cbee3e | 1508 | int ext4_init_inode_table(struct super_block *sb, ext4_group_t group, |
bfff6873 LC |
1509 | int barrier) |
1510 | { | |
1511 | struct ext4_group_info *grp = ext4_get_group_info(sb, group); | |
1512 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
1513 | struct ext4_group_desc *gdp = NULL; | |
1514 | struct buffer_head *group_desc_bh; | |
1515 | handle_t *handle; | |
1516 | ext4_fsblk_t blk; | |
1517 | int num, ret = 0, used_blks = 0; | |
a149d2a5 | 1518 | unsigned long used_inos = 0; |
bfff6873 LC |
1519 | |
1520 | /* This should not happen, but just to be sure check this */ | |
bc98a42c | 1521 | if (sb_rdonly(sb)) { |
bfff6873 LC |
1522 | ret = 1; |
1523 | goto out; | |
1524 | } | |
1525 | ||
1526 | gdp = ext4_get_group_desc(sb, group, &group_desc_bh); | |
1527 | if (!gdp) | |
1528 | goto out; | |
1529 | ||
1530 | /* | |
1531 | * We do not need to lock this, because we are the only one | |
1532 | * handling this flag. | |
1533 | */ | |
1534 | if (gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)) | |
1535 | goto out; | |
1536 | ||
9924a92a | 1537 | handle = ext4_journal_start_sb(sb, EXT4_HT_MISC, 1); |
bfff6873 LC |
1538 | if (IS_ERR(handle)) { |
1539 | ret = PTR_ERR(handle); | |
1540 | goto out; | |
1541 | } | |
1542 | ||
1543 | down_write(&grp->alloc_sem); | |
1544 | /* | |
1545 | * If inode bitmap was already initialized there may be some | |
1546 | * used inodes so we need to skip blocks with used inodes in | |
1547 | * inode table. | |
1548 | */ | |
a149d2a5 ZY |
1549 | if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_UNINIT))) { |
1550 | used_inos = EXT4_INODES_PER_GROUP(sb) - | |
1551 | ext4_itable_unused_count(sb, gdp); | |
1552 | used_blks = DIV_ROUND_UP(used_inos, sbi->s_inodes_per_block); | |
1553 | ||
1554 | /* Bogus inode unused count? */ | |
1555 | if (used_blks < 0 || used_blks > sbi->s_itb_per_group) { | |
1556 | ext4_error(sb, "Something is wrong with group %u: " | |
1557 | "used itable blocks: %d; " | |
1558 | "itable unused count: %u", | |
1559 | group, used_blks, | |
1560 | ext4_itable_unused_count(sb, gdp)); | |
1561 | ret = 1; | |
1562 | goto err_out; | |
1563 | } | |
1564 | ||
1565 | used_inos += group * EXT4_INODES_PER_GROUP(sb); | |
1566 | /* | |
1567 | * Are there some uninitialized inodes in the inode table | |
1568 | * before the first normal inode? | |
1569 | */ | |
1570 | if ((used_blks != sbi->s_itb_per_group) && | |
1571 | (used_inos < EXT4_FIRST_INO(sb))) { | |
1572 | ext4_error(sb, "Something is wrong with group %u: " | |
1573 | "itable unused count: %u; " | |
1574 | "itables initialized count: %ld", | |
1575 | group, ext4_itable_unused_count(sb, gdp), | |
1576 | used_inos); | |
1577 | ret = 1; | |
1578 | goto err_out; | |
1579 | } | |
857ac889 LC |
1580 | } |
1581 | ||
bfff6873 LC |
1582 | blk = ext4_inode_table(sb, gdp) + used_blks; |
1583 | num = sbi->s_itb_per_group - used_blks; | |
1584 | ||
1585 | BUFFER_TRACE(group_desc_bh, "get_write_access"); | |
1586 | ret = ext4_journal_get_write_access(handle, | |
1587 | group_desc_bh); | |
1588 | if (ret) | |
1589 | goto err_out; | |
1590 | ||
bfff6873 LC |
1591 | /* |
1592 | * Skip zeroout if the inode table is full. But we set the ZEROED | |
1593 | * flag anyway, because obviously, when it is full it does not need | |
1594 | * further zeroing. | |
1595 | */ | |
1596 | if (unlikely(num == 0)) | |
1597 | goto skip_zeroout; | |
1598 | ||
1599 | ext4_debug("going to zero out inode table in group %d\n", | |
1600 | group); | |
a107e5a3 | 1601 | ret = sb_issue_zeroout(sb, blk, num, GFP_NOFS); |
bfff6873 LC |
1602 | if (ret < 0) |
1603 | goto err_out; | |
a107e5a3 | 1604 | if (barrier) |
c6bf3f0e | 1605 | blkdev_issue_flush(sb->s_bdev); |
bfff6873 LC |
1606 | |
1607 | skip_zeroout: | |
1608 | ext4_lock_group(sb, group); | |
1609 | gdp->bg_flags |= cpu_to_le16(EXT4_BG_INODE_ZEROED); | |
feb0ab32 | 1610 | ext4_group_desc_csum_set(sb, group, gdp); |
bfff6873 LC |
1611 | ext4_unlock_group(sb, group); |
1612 | ||
1613 | BUFFER_TRACE(group_desc_bh, | |
1614 | "call ext4_handle_dirty_metadata"); | |
1615 | ret = ext4_handle_dirty_metadata(handle, NULL, | |
1616 | group_desc_bh); | |
1617 | ||
1618 | err_out: | |
1619 | up_write(&grp->alloc_sem); | |
1620 | ext4_journal_stop(handle); | |
1621 | out: | |
1622 | return ret; | |
1623 | } |