diff options
author | Linus Torvalds <torvalds@ppc970.osdl.org> | 2005-04-16 18:20:36 -0400 |
---|---|---|
committer | Linus Torvalds <torvalds@ppc970.osdl.org> | 2005-04-16 18:20:36 -0400 |
commit | 1da177e4c3f41524e886b7f1b8a0c1fc7321cac2 (patch) | |
tree | 0bba044c4ce775e45a88a51686b5d9f90697ea9d /fs/ext3/ialloc.c |
Linux-2.6.12-rc2v2.6.12-rc2
Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.
Let it rip!
Diffstat (limited to 'fs/ext3/ialloc.c')
-rw-r--r-- | fs/ext3/ialloc.c | 794 |
1 files changed, 794 insertions, 0 deletions
diff --git a/fs/ext3/ialloc.c b/fs/ext3/ialloc.c new file mode 100644 index 000000000000..1e6f3ea28713 --- /dev/null +++ b/fs/ext3/ialloc.c | |||
@@ -0,0 +1,794 @@ | |||
1 | /* | ||
2 | * linux/fs/ext3/ialloc.c | ||
3 | * | ||
4 | * Copyright (C) 1992, 1993, 1994, 1995 | ||
5 | * Remy Card (card@masi.ibp.fr) | ||
6 | * Laboratoire MASI - Institut Blaise Pascal | ||
7 | * Universite Pierre et Marie Curie (Paris VI) | ||
8 | * | ||
9 | * BSD ufs-inspired inode and directory allocation by | ||
10 | * Stephen Tweedie (sct@redhat.com), 1993 | ||
11 | * Big-endian to little-endian byte-swapping/bitmaps by | ||
12 | * David S. Miller (davem@caip.rutgers.edu), 1995 | ||
13 | */ | ||
14 | |||
15 | #include <linux/time.h> | ||
16 | #include <linux/fs.h> | ||
17 | #include <linux/jbd.h> | ||
18 | #include <linux/ext3_fs.h> | ||
19 | #include <linux/ext3_jbd.h> | ||
20 | #include <linux/stat.h> | ||
21 | #include <linux/string.h> | ||
22 | #include <linux/quotaops.h> | ||
23 | #include <linux/buffer_head.h> | ||
24 | #include <linux/random.h> | ||
25 | #include <linux/bitops.h> | ||
26 | |||
27 | #include <asm/byteorder.h> | ||
28 | |||
29 | #include "xattr.h" | ||
30 | #include "acl.h" | ||
31 | |||
32 | /* | ||
33 | * ialloc.c contains the inodes allocation and deallocation routines | ||
34 | */ | ||
35 | |||
36 | /* | ||
37 | * The free inodes are managed by bitmaps. A file system contains several | ||
38 | * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap | ||
39 | * block for inodes, N blocks for the inode table and data blocks. | ||
40 | * | ||
41 | * The file system contains group descriptors which are located after the | ||
42 | * super block. Each descriptor contains the number of the bitmap block and | ||
43 | * the free blocks count in the block. | ||
44 | */ | ||
45 | |||
46 | |||
47 | /* | ||
48 | * Read the inode allocation bitmap for a given block_group, reading | ||
49 | * into the specified slot in the superblock's bitmap cache. | ||
50 | * | ||
51 | * Return buffer_head of bitmap on success or NULL. | ||
52 | */ | ||
53 | static struct buffer_head * | ||
54 | read_inode_bitmap(struct super_block * sb, unsigned long block_group) | ||
55 | { | ||
56 | struct ext3_group_desc *desc; | ||
57 | struct buffer_head *bh = NULL; | ||
58 | |||
59 | desc = ext3_get_group_desc(sb, block_group, NULL); | ||
60 | if (!desc) | ||
61 | goto error_out; | ||
62 | |||
63 | bh = sb_bread(sb, le32_to_cpu(desc->bg_inode_bitmap)); | ||
64 | if (!bh) | ||
65 | ext3_error(sb, "read_inode_bitmap", | ||
66 | "Cannot read inode bitmap - " | ||
67 | "block_group = %lu, inode_bitmap = %u", | ||
68 | block_group, le32_to_cpu(desc->bg_inode_bitmap)); | ||
69 | error_out: | ||
70 | return bh; | ||
71 | } | ||
72 | |||
73 | /* | ||
74 | * NOTE! When we get the inode, we're the only people | ||
75 | * that have access to it, and as such there are no | ||
76 | * race conditions we have to worry about. The inode | ||
77 | * is not on the hash-lists, and it cannot be reached | ||
78 | * through the filesystem because the directory entry | ||
79 | * has been deleted earlier. | ||
80 | * | ||
81 | * HOWEVER: we must make sure that we get no aliases, | ||
82 | * which means that we have to call "clear_inode()" | ||
83 | * _before_ we mark the inode not in use in the inode | ||
84 | * bitmaps. Otherwise a newly created file might use | ||
85 | * the same inode number (not actually the same pointer | ||
86 | * though), and then we'd have two inodes sharing the | ||
87 | * same inode number and space on the harddisk. | ||
88 | */ | ||
89 | void ext3_free_inode (handle_t *handle, struct inode * inode) | ||
90 | { | ||
91 | struct super_block * sb = inode->i_sb; | ||
92 | int is_directory; | ||
93 | unsigned long ino; | ||
94 | struct buffer_head *bitmap_bh = NULL; | ||
95 | struct buffer_head *bh2; | ||
96 | unsigned long block_group; | ||
97 | unsigned long bit; | ||
98 | struct ext3_group_desc * gdp; | ||
99 | struct ext3_super_block * es; | ||
100 | struct ext3_sb_info *sbi; | ||
101 | int fatal = 0, err; | ||
102 | |||
103 | if (atomic_read(&inode->i_count) > 1) { | ||
104 | printk ("ext3_free_inode: inode has count=%d\n", | ||
105 | atomic_read(&inode->i_count)); | ||
106 | return; | ||
107 | } | ||
108 | if (inode->i_nlink) { | ||
109 | printk ("ext3_free_inode: inode has nlink=%d\n", | ||
110 | inode->i_nlink); | ||
111 | return; | ||
112 | } | ||
113 | if (!sb) { | ||
114 | printk("ext3_free_inode: inode on nonexistent device\n"); | ||
115 | return; | ||
116 | } | ||
117 | sbi = EXT3_SB(sb); | ||
118 | |||
119 | ino = inode->i_ino; | ||
120 | ext3_debug ("freeing inode %lu\n", ino); | ||
121 | |||
122 | /* | ||
123 | * Note: we must free any quota before locking the superblock, | ||
124 | * as writing the quota to disk may need the lock as well. | ||
125 | */ | ||
126 | DQUOT_INIT(inode); | ||
127 | ext3_xattr_delete_inode(handle, inode); | ||
128 | DQUOT_FREE_INODE(inode); | ||
129 | DQUOT_DROP(inode); | ||
130 | |||
131 | is_directory = S_ISDIR(inode->i_mode); | ||
132 | |||
133 | /* Do this BEFORE marking the inode not in use or returning an error */ | ||
134 | clear_inode (inode); | ||
135 | |||
136 | es = EXT3_SB(sb)->s_es; | ||
137 | if (ino < EXT3_FIRST_INO(sb) || ino > le32_to_cpu(es->s_inodes_count)) { | ||
138 | ext3_error (sb, "ext3_free_inode", | ||
139 | "reserved or nonexistent inode %lu", ino); | ||
140 | goto error_return; | ||
141 | } | ||
142 | block_group = (ino - 1) / EXT3_INODES_PER_GROUP(sb); | ||
143 | bit = (ino - 1) % EXT3_INODES_PER_GROUP(sb); | ||
144 | bitmap_bh = read_inode_bitmap(sb, block_group); | ||
145 | if (!bitmap_bh) | ||
146 | goto error_return; | ||
147 | |||
148 | BUFFER_TRACE(bitmap_bh, "get_write_access"); | ||
149 | fatal = ext3_journal_get_write_access(handle, bitmap_bh); | ||
150 | if (fatal) | ||
151 | goto error_return; | ||
152 | |||
153 | /* Ok, now we can actually update the inode bitmaps.. */ | ||
154 | if (!ext3_clear_bit_atomic(sb_bgl_lock(sbi, block_group), | ||
155 | bit, bitmap_bh->b_data)) | ||
156 | ext3_error (sb, "ext3_free_inode", | ||
157 | "bit already cleared for inode %lu", ino); | ||
158 | else { | ||
159 | gdp = ext3_get_group_desc (sb, block_group, &bh2); | ||
160 | |||
161 | BUFFER_TRACE(bh2, "get_write_access"); | ||
162 | fatal = ext3_journal_get_write_access(handle, bh2); | ||
163 | if (fatal) goto error_return; | ||
164 | |||
165 | if (gdp) { | ||
166 | spin_lock(sb_bgl_lock(sbi, block_group)); | ||
167 | gdp->bg_free_inodes_count = cpu_to_le16( | ||
168 | le16_to_cpu(gdp->bg_free_inodes_count) + 1); | ||
169 | if (is_directory) | ||
170 | gdp->bg_used_dirs_count = cpu_to_le16( | ||
171 | le16_to_cpu(gdp->bg_used_dirs_count) - 1); | ||
172 | spin_unlock(sb_bgl_lock(sbi, block_group)); | ||
173 | percpu_counter_inc(&sbi->s_freeinodes_counter); | ||
174 | if (is_directory) | ||
175 | percpu_counter_dec(&sbi->s_dirs_counter); | ||
176 | |||
177 | } | ||
178 | BUFFER_TRACE(bh2, "call ext3_journal_dirty_metadata"); | ||
179 | err = ext3_journal_dirty_metadata(handle, bh2); | ||
180 | if (!fatal) fatal = err; | ||
181 | } | ||
182 | BUFFER_TRACE(bitmap_bh, "call ext3_journal_dirty_metadata"); | ||
183 | err = ext3_journal_dirty_metadata(handle, bitmap_bh); | ||
184 | if (!fatal) | ||
185 | fatal = err; | ||
186 | sb->s_dirt = 1; | ||
187 | error_return: | ||
188 | brelse(bitmap_bh); | ||
189 | ext3_std_error(sb, fatal); | ||
190 | } | ||
191 | |||
192 | /* | ||
193 | * There are two policies for allocating an inode. If the new inode is | ||
194 | * a directory, then a forward search is made for a block group with both | ||
195 | * free space and a low directory-to-inode ratio; if that fails, then of | ||
196 | * the groups with above-average free space, that group with the fewest | ||
197 | * directories already is chosen. | ||
198 | * | ||
199 | * For other inodes, search forward from the parent directory\'s block | ||
200 | * group to find a free inode. | ||
201 | */ | ||
202 | static int find_group_dir(struct super_block *sb, struct inode *parent) | ||
203 | { | ||
204 | int ngroups = EXT3_SB(sb)->s_groups_count; | ||
205 | int freei, avefreei; | ||
206 | struct ext3_group_desc *desc, *best_desc = NULL; | ||
207 | struct buffer_head *bh; | ||
208 | int group, best_group = -1; | ||
209 | |||
210 | freei = percpu_counter_read_positive(&EXT3_SB(sb)->s_freeinodes_counter); | ||
211 | avefreei = freei / ngroups; | ||
212 | |||
213 | for (group = 0; group < ngroups; group++) { | ||
214 | desc = ext3_get_group_desc (sb, group, &bh); | ||
215 | if (!desc || !desc->bg_free_inodes_count) | ||
216 | continue; | ||
217 | if (le16_to_cpu(desc->bg_free_inodes_count) < avefreei) | ||
218 | continue; | ||
219 | if (!best_desc || | ||
220 | (le16_to_cpu(desc->bg_free_blocks_count) > | ||
221 | le16_to_cpu(best_desc->bg_free_blocks_count))) { | ||
222 | best_group = group; | ||
223 | best_desc = desc; | ||
224 | } | ||
225 | } | ||
226 | return best_group; | ||
227 | } | ||
228 | |||
229 | /* | ||
230 | * Orlov's allocator for directories. | ||
231 | * | ||
232 | * We always try to spread first-level directories. | ||
233 | * | ||
234 | * If there are blockgroups with both free inodes and free blocks counts | ||
235 | * not worse than average we return one with smallest directory count. | ||
236 | * Otherwise we simply return a random group. | ||
237 | * | ||
238 | * For the rest rules look so: | ||
239 | * | ||
240 | * It's OK to put directory into a group unless | ||
241 | * it has too many directories already (max_dirs) or | ||
242 | * it has too few free inodes left (min_inodes) or | ||
243 | * it has too few free blocks left (min_blocks) or | ||
244 | * it's already running too large debt (max_debt). | ||
245 | * Parent's group is prefered, if it doesn't satisfy these | ||
246 | * conditions we search cyclically through the rest. If none | ||
247 | * of the groups look good we just look for a group with more | ||
248 | * free inodes than average (starting at parent's group). | ||
249 | * | ||
250 | * Debt is incremented each time we allocate a directory and decremented | ||
251 | * when we allocate an inode, within 0--255. | ||
252 | */ | ||
253 | |||
254 | #define INODE_COST 64 | ||
255 | #define BLOCK_COST 256 | ||
256 | |||
257 | static int find_group_orlov(struct super_block *sb, struct inode *parent) | ||
258 | { | ||
259 | int parent_group = EXT3_I(parent)->i_block_group; | ||
260 | struct ext3_sb_info *sbi = EXT3_SB(sb); | ||
261 | struct ext3_super_block *es = sbi->s_es; | ||
262 | int ngroups = sbi->s_groups_count; | ||
263 | int inodes_per_group = EXT3_INODES_PER_GROUP(sb); | ||
264 | int freei, avefreei; | ||
265 | int freeb, avefreeb; | ||
266 | int blocks_per_dir, ndirs; | ||
267 | int max_debt, max_dirs, min_blocks, min_inodes; | ||
268 | int group = -1, i; | ||
269 | struct ext3_group_desc *desc; | ||
270 | struct buffer_head *bh; | ||
271 | |||
272 | freei = percpu_counter_read_positive(&sbi->s_freeinodes_counter); | ||
273 | avefreei = freei / ngroups; | ||
274 | freeb = percpu_counter_read_positive(&sbi->s_freeblocks_counter); | ||
275 | avefreeb = freeb / ngroups; | ||
276 | ndirs = percpu_counter_read_positive(&sbi->s_dirs_counter); | ||
277 | |||
278 | if ((parent == sb->s_root->d_inode) || | ||
279 | (EXT3_I(parent)->i_flags & EXT3_TOPDIR_FL)) { | ||
280 | int best_ndir = inodes_per_group; | ||
281 | int best_group = -1; | ||
282 | |||
283 | get_random_bytes(&group, sizeof(group)); | ||
284 | parent_group = (unsigned)group % ngroups; | ||
285 | for (i = 0; i < ngroups; i++) { | ||
286 | group = (parent_group + i) % ngroups; | ||
287 | desc = ext3_get_group_desc (sb, group, &bh); | ||
288 | if (!desc || !desc->bg_free_inodes_count) | ||
289 | continue; | ||
290 | if (le16_to_cpu(desc->bg_used_dirs_count) >= best_ndir) | ||
291 | continue; | ||
292 | if (le16_to_cpu(desc->bg_free_inodes_count) < avefreei) | ||
293 | continue; | ||
294 | if (le16_to_cpu(desc->bg_free_blocks_count) < avefreeb) | ||
295 | continue; | ||
296 | best_group = group; | ||
297 | best_ndir = le16_to_cpu(desc->bg_used_dirs_count); | ||
298 | } | ||
299 | if (best_group >= 0) | ||
300 | return best_group; | ||
301 | goto fallback; | ||
302 | } | ||
303 | |||
304 | blocks_per_dir = (le32_to_cpu(es->s_blocks_count) - freeb) / ndirs; | ||
305 | |||
306 | max_dirs = ndirs / ngroups + inodes_per_group / 16; | ||
307 | min_inodes = avefreei - inodes_per_group / 4; | ||
308 | min_blocks = avefreeb - EXT3_BLOCKS_PER_GROUP(sb) / 4; | ||
309 | |||
310 | max_debt = EXT3_BLOCKS_PER_GROUP(sb) / max(blocks_per_dir, BLOCK_COST); | ||
311 | if (max_debt * INODE_COST > inodes_per_group) | ||
312 | max_debt = inodes_per_group / INODE_COST; | ||
313 | if (max_debt > 255) | ||
314 | max_debt = 255; | ||
315 | if (max_debt == 0) | ||
316 | max_debt = 1; | ||
317 | |||
318 | for (i = 0; i < ngroups; i++) { | ||
319 | group = (parent_group + i) % ngroups; | ||
320 | desc = ext3_get_group_desc (sb, group, &bh); | ||
321 | if (!desc || !desc->bg_free_inodes_count) | ||
322 | continue; | ||
323 | if (le16_to_cpu(desc->bg_used_dirs_count) >= max_dirs) | ||
324 | continue; | ||
325 | if (le16_to_cpu(desc->bg_free_inodes_count) < min_inodes) | ||
326 | continue; | ||
327 | if (le16_to_cpu(desc->bg_free_blocks_count) < min_blocks) | ||
328 | continue; | ||
329 | return group; | ||
330 | } | ||
331 | |||
332 | fallback: | ||
333 | for (i = 0; i < ngroups; i++) { | ||
334 | group = (parent_group + i) % ngroups; | ||
335 | desc = ext3_get_group_desc (sb, group, &bh); | ||
336 | if (!desc || !desc->bg_free_inodes_count) | ||
337 | continue; | ||
338 | if (le16_to_cpu(desc->bg_free_inodes_count) >= avefreei) | ||
339 | return group; | ||
340 | } | ||
341 | |||
342 | if (avefreei) { | ||
343 | /* | ||
344 | * The free-inodes counter is approximate, and for really small | ||
345 | * filesystems the above test can fail to find any blockgroups | ||
346 | */ | ||
347 | avefreei = 0; | ||
348 | goto fallback; | ||
349 | } | ||
350 | |||
351 | return -1; | ||
352 | } | ||
353 | |||
354 | static int find_group_other(struct super_block *sb, struct inode *parent) | ||
355 | { | ||
356 | int parent_group = EXT3_I(parent)->i_block_group; | ||
357 | int ngroups = EXT3_SB(sb)->s_groups_count; | ||
358 | struct ext3_group_desc *desc; | ||
359 | struct buffer_head *bh; | ||
360 | int group, i; | ||
361 | |||
362 | /* | ||
363 | * Try to place the inode in its parent directory | ||
364 | */ | ||
365 | group = parent_group; | ||
366 | desc = ext3_get_group_desc (sb, group, &bh); | ||
367 | if (desc && le16_to_cpu(desc->bg_free_inodes_count) && | ||
368 | le16_to_cpu(desc->bg_free_blocks_count)) | ||
369 | return group; | ||
370 | |||
371 | /* | ||
372 | * We're going to place this inode in a different blockgroup from its | ||
373 | * parent. We want to cause files in a common directory to all land in | ||
374 | * the same blockgroup. But we want files which are in a different | ||
375 | * directory which shares a blockgroup with our parent to land in a | ||
376 | * different blockgroup. | ||
377 | * | ||
378 | * So add our directory's i_ino into the starting point for the hash. | ||
379 | */ | ||
380 | group = (group + parent->i_ino) % ngroups; | ||
381 | |||
382 | /* | ||
383 | * Use a quadratic hash to find a group with a free inode and some free | ||
384 | * blocks. | ||
385 | */ | ||
386 | for (i = 1; i < ngroups; i <<= 1) { | ||
387 | group += i; | ||
388 | if (group >= ngroups) | ||
389 | group -= ngroups; | ||
390 | desc = ext3_get_group_desc (sb, group, &bh); | ||
391 | if (desc && le16_to_cpu(desc->bg_free_inodes_count) && | ||
392 | le16_to_cpu(desc->bg_free_blocks_count)) | ||
393 | return group; | ||
394 | } | ||
395 | |||
396 | /* | ||
397 | * That failed: try linear search for a free inode, even if that group | ||
398 | * has no free blocks. | ||
399 | */ | ||
400 | group = parent_group; | ||
401 | for (i = 0; i < ngroups; i++) { | ||
402 | if (++group >= ngroups) | ||
403 | group = 0; | ||
404 | desc = ext3_get_group_desc (sb, group, &bh); | ||
405 | if (desc && le16_to_cpu(desc->bg_free_inodes_count)) | ||
406 | return group; | ||
407 | } | ||
408 | |||
409 | return -1; | ||
410 | } | ||
411 | |||
412 | /* | ||
413 | * There are two policies for allocating an inode. If the new inode is | ||
414 | * a directory, then a forward search is made for a block group with both | ||
415 | * free space and a low directory-to-inode ratio; if that fails, then of | ||
416 | * the groups with above-average free space, that group with the fewest | ||
417 | * directories already is chosen. | ||
418 | * | ||
419 | * For other inodes, search forward from the parent directory's block | ||
420 | * group to find a free inode. | ||
421 | */ | ||
422 | struct inode *ext3_new_inode(handle_t *handle, struct inode * dir, int mode) | ||
423 | { | ||
424 | struct super_block *sb; | ||
425 | struct buffer_head *bitmap_bh = NULL; | ||
426 | struct buffer_head *bh2; | ||
427 | int group; | ||
428 | unsigned long ino = 0; | ||
429 | struct inode * inode; | ||
430 | struct ext3_group_desc * gdp = NULL; | ||
431 | struct ext3_super_block * es; | ||
432 | struct ext3_inode_info *ei; | ||
433 | struct ext3_sb_info *sbi; | ||
434 | int err = 0; | ||
435 | struct inode *ret; | ||
436 | int i; | ||
437 | |||
438 | /* Cannot create files in a deleted directory */ | ||
439 | if (!dir || !dir->i_nlink) | ||
440 | return ERR_PTR(-EPERM); | ||
441 | |||
442 | sb = dir->i_sb; | ||
443 | inode = new_inode(sb); | ||
444 | if (!inode) | ||
445 | return ERR_PTR(-ENOMEM); | ||
446 | ei = EXT3_I(inode); | ||
447 | |||
448 | sbi = EXT3_SB(sb); | ||
449 | es = sbi->s_es; | ||
450 | if (S_ISDIR(mode)) { | ||
451 | if (test_opt (sb, OLDALLOC)) | ||
452 | group = find_group_dir(sb, dir); | ||
453 | else | ||
454 | group = find_group_orlov(sb, dir); | ||
455 | } else | ||
456 | group = find_group_other(sb, dir); | ||
457 | |||
458 | err = -ENOSPC; | ||
459 | if (group == -1) | ||
460 | goto out; | ||
461 | |||
462 | for (i = 0; i < sbi->s_groups_count; i++) { | ||
463 | err = -EIO; | ||
464 | |||
465 | gdp = ext3_get_group_desc(sb, group, &bh2); | ||
466 | if (!gdp) | ||
467 | goto fail; | ||
468 | |||
469 | brelse(bitmap_bh); | ||
470 | bitmap_bh = read_inode_bitmap(sb, group); | ||
471 | if (!bitmap_bh) | ||
472 | goto fail; | ||
473 | |||
474 | ino = 0; | ||
475 | |||
476 | repeat_in_this_group: | ||
477 | ino = ext3_find_next_zero_bit((unsigned long *) | ||
478 | bitmap_bh->b_data, EXT3_INODES_PER_GROUP(sb), ino); | ||
479 | if (ino < EXT3_INODES_PER_GROUP(sb)) { | ||
480 | |||
481 | BUFFER_TRACE(bitmap_bh, "get_write_access"); | ||
482 | err = ext3_journal_get_write_access(handle, bitmap_bh); | ||
483 | if (err) | ||
484 | goto fail; | ||
485 | |||
486 | if (!ext3_set_bit_atomic(sb_bgl_lock(sbi, group), | ||
487 | ino, bitmap_bh->b_data)) { | ||
488 | /* we won it */ | ||
489 | BUFFER_TRACE(bitmap_bh, | ||
490 | "call ext3_journal_dirty_metadata"); | ||
491 | err = ext3_journal_dirty_metadata(handle, | ||
492 | bitmap_bh); | ||
493 | if (err) | ||
494 | goto fail; | ||
495 | goto got; | ||
496 | } | ||
497 | /* we lost it */ | ||
498 | journal_release_buffer(handle, bitmap_bh); | ||
499 | |||
500 | if (++ino < EXT3_INODES_PER_GROUP(sb)) | ||
501 | goto repeat_in_this_group; | ||
502 | } | ||
503 | |||
504 | /* | ||
505 | * This case is possible in concurrent environment. It is very | ||
506 | * rare. We cannot repeat the find_group_xxx() call because | ||
507 | * that will simply return the same blockgroup, because the | ||
508 | * group descriptor metadata has not yet been updated. | ||
509 | * So we just go onto the next blockgroup. | ||
510 | */ | ||
511 | if (++group == sbi->s_groups_count) | ||
512 | group = 0; | ||
513 | } | ||
514 | err = -ENOSPC; | ||
515 | goto out; | ||
516 | |||
517 | got: | ||
518 | ino += group * EXT3_INODES_PER_GROUP(sb) + 1; | ||
519 | if (ino < EXT3_FIRST_INO(sb) || ino > le32_to_cpu(es->s_inodes_count)) { | ||
520 | ext3_error (sb, "ext3_new_inode", | ||
521 | "reserved inode or inode > inodes count - " | ||
522 | "block_group = %d, inode=%lu", group, ino); | ||
523 | err = -EIO; | ||
524 | goto fail; | ||
525 | } | ||
526 | |||
527 | BUFFER_TRACE(bh2, "get_write_access"); | ||
528 | err = ext3_journal_get_write_access(handle, bh2); | ||
529 | if (err) goto fail; | ||
530 | spin_lock(sb_bgl_lock(sbi, group)); | ||
531 | gdp->bg_free_inodes_count = | ||
532 | cpu_to_le16(le16_to_cpu(gdp->bg_free_inodes_count) - 1); | ||
533 | if (S_ISDIR(mode)) { | ||
534 | gdp->bg_used_dirs_count = | ||
535 | cpu_to_le16(le16_to_cpu(gdp->bg_used_dirs_count) + 1); | ||
536 | } | ||
537 | spin_unlock(sb_bgl_lock(sbi, group)); | ||
538 | BUFFER_TRACE(bh2, "call ext3_journal_dirty_metadata"); | ||
539 | err = ext3_journal_dirty_metadata(handle, bh2); | ||
540 | if (err) goto fail; | ||
541 | |||
542 | percpu_counter_dec(&sbi->s_freeinodes_counter); | ||
543 | if (S_ISDIR(mode)) | ||
544 | percpu_counter_inc(&sbi->s_dirs_counter); | ||
545 | sb->s_dirt = 1; | ||
546 | |||
547 | inode->i_uid = current->fsuid; | ||
548 | if (test_opt (sb, GRPID)) | ||
549 | inode->i_gid = dir->i_gid; | ||
550 | else if (dir->i_mode & S_ISGID) { | ||
551 | inode->i_gid = dir->i_gid; | ||
552 | if (S_ISDIR(mode)) | ||
553 | mode |= S_ISGID; | ||
554 | } else | ||
555 | inode->i_gid = current->fsgid; | ||
556 | inode->i_mode = mode; | ||
557 | |||
558 | inode->i_ino = ino; | ||
559 | /* This is the optimal IO size (for stat), not the fs block size */ | ||
560 | inode->i_blksize = PAGE_SIZE; | ||
561 | inode->i_blocks = 0; | ||
562 | inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME_SEC; | ||
563 | |||
564 | memset(ei->i_data, 0, sizeof(ei->i_data)); | ||
565 | ei->i_dir_start_lookup = 0; | ||
566 | ei->i_disksize = 0; | ||
567 | |||
568 | ei->i_flags = EXT3_I(dir)->i_flags & ~EXT3_INDEX_FL; | ||
569 | if (S_ISLNK(mode)) | ||
570 | ei->i_flags &= ~(EXT3_IMMUTABLE_FL|EXT3_APPEND_FL); | ||
571 | /* dirsync only applies to directories */ | ||
572 | if (!S_ISDIR(mode)) | ||
573 | ei->i_flags &= ~EXT3_DIRSYNC_FL; | ||
574 | #ifdef EXT3_FRAGMENTS | ||
575 | ei->i_faddr = 0; | ||
576 | ei->i_frag_no = 0; | ||
577 | ei->i_frag_size = 0; | ||
578 | #endif | ||
579 | ei->i_file_acl = 0; | ||
580 | ei->i_dir_acl = 0; | ||
581 | ei->i_dtime = 0; | ||
582 | ei->i_block_alloc_info = NULL; | ||
583 | ei->i_block_group = group; | ||
584 | |||
585 | ext3_set_inode_flags(inode); | ||
586 | if (IS_DIRSYNC(inode)) | ||
587 | handle->h_sync = 1; | ||
588 | insert_inode_hash(inode); | ||
589 | spin_lock(&sbi->s_next_gen_lock); | ||
590 | inode->i_generation = sbi->s_next_generation++; | ||
591 | spin_unlock(&sbi->s_next_gen_lock); | ||
592 | |||
593 | ei->i_state = EXT3_STATE_NEW; | ||
594 | ei->i_extra_isize = | ||
595 | (EXT3_INODE_SIZE(inode->i_sb) > EXT3_GOOD_OLD_INODE_SIZE) ? | ||
596 | sizeof(struct ext3_inode) - EXT3_GOOD_OLD_INODE_SIZE : 0; | ||
597 | |||
598 | ret = inode; | ||
599 | if(DQUOT_ALLOC_INODE(inode)) { | ||
600 | DQUOT_DROP(inode); | ||
601 | err = -EDQUOT; | ||
602 | goto fail2; | ||
603 | } | ||
604 | err = ext3_init_acl(handle, inode, dir); | ||
605 | if (err) { | ||
606 | DQUOT_FREE_INODE(inode); | ||
607 | goto fail2; | ||
608 | } | ||
609 | err = ext3_mark_inode_dirty(handle, inode); | ||
610 | if (err) { | ||
611 | ext3_std_error(sb, err); | ||
612 | DQUOT_FREE_INODE(inode); | ||
613 | goto fail2; | ||
614 | } | ||
615 | |||
616 | ext3_debug("allocating inode %lu\n", inode->i_ino); | ||
617 | goto really_out; | ||
618 | fail: | ||
619 | ext3_std_error(sb, err); | ||
620 | out: | ||
621 | iput(inode); | ||
622 | ret = ERR_PTR(err); | ||
623 | really_out: | ||
624 | brelse(bitmap_bh); | ||
625 | return ret; | ||
626 | |||
627 | fail2: | ||
628 | inode->i_flags |= S_NOQUOTA; | ||
629 | inode->i_nlink = 0; | ||
630 | iput(inode); | ||
631 | brelse(bitmap_bh); | ||
632 | return ERR_PTR(err); | ||
633 | } | ||
634 | |||
635 | /* Verify that we are loading a valid orphan from disk */ | ||
636 | struct inode *ext3_orphan_get(struct super_block *sb, unsigned long ino) | ||
637 | { | ||
638 | unsigned long max_ino = le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count); | ||
639 | unsigned long block_group; | ||
640 | int bit; | ||
641 | struct buffer_head *bitmap_bh = NULL; | ||
642 | struct inode *inode = NULL; | ||
643 | |||
644 | /* Error cases - e2fsck has already cleaned up for us */ | ||
645 | if (ino > max_ino) { | ||
646 | ext3_warning(sb, __FUNCTION__, | ||
647 | "bad orphan ino %lu! e2fsck was run?\n", ino); | ||
648 | goto out; | ||
649 | } | ||
650 | |||
651 | block_group = (ino - 1) / EXT3_INODES_PER_GROUP(sb); | ||
652 | bit = (ino - 1) % EXT3_INODES_PER_GROUP(sb); | ||
653 | bitmap_bh = read_inode_bitmap(sb, block_group); | ||
654 | if (!bitmap_bh) { | ||
655 | ext3_warning(sb, __FUNCTION__, | ||
656 | "inode bitmap error for orphan %lu\n", ino); | ||
657 | goto out; | ||
658 | } | ||
659 | |||
660 | /* Having the inode bit set should be a 100% indicator that this | ||
661 | * is a valid orphan (no e2fsck run on fs). Orphans also include | ||
662 | * inodes that were being truncated, so we can't check i_nlink==0. | ||
663 | */ | ||
664 | if (!ext3_test_bit(bit, bitmap_bh->b_data) || | ||
665 | !(inode = iget(sb, ino)) || is_bad_inode(inode) || | ||
666 | NEXT_ORPHAN(inode) > max_ino) { | ||
667 | ext3_warning(sb, __FUNCTION__, | ||
668 | "bad orphan inode %lu! e2fsck was run?\n", ino); | ||
669 | printk(KERN_NOTICE "ext3_test_bit(bit=%d, block=%llu) = %d\n", | ||
670 | bit, (unsigned long long)bitmap_bh->b_blocknr, | ||
671 | ext3_test_bit(bit, bitmap_bh->b_data)); | ||
672 | printk(KERN_NOTICE "inode=%p\n", inode); | ||
673 | if (inode) { | ||
674 | printk(KERN_NOTICE "is_bad_inode(inode)=%d\n", | ||
675 | is_bad_inode(inode)); | ||
676 | printk(KERN_NOTICE "NEXT_ORPHAN(inode)=%u\n", | ||
677 | NEXT_ORPHAN(inode)); | ||
678 | printk(KERN_NOTICE "max_ino=%lu\n", max_ino); | ||
679 | } | ||
680 | /* Avoid freeing blocks if we got a bad deleted inode */ | ||
681 | if (inode && inode->i_nlink == 0) | ||
682 | inode->i_blocks = 0; | ||
683 | iput(inode); | ||
684 | inode = NULL; | ||
685 | } | ||
686 | out: | ||
687 | brelse(bitmap_bh); | ||
688 | return inode; | ||
689 | } | ||
690 | |||
691 | unsigned long ext3_count_free_inodes (struct super_block * sb) | ||
692 | { | ||
693 | unsigned long desc_count; | ||
694 | struct ext3_group_desc *gdp; | ||
695 | int i; | ||
696 | #ifdef EXT3FS_DEBUG | ||
697 | struct ext3_super_block *es; | ||
698 | unsigned long bitmap_count, x; | ||
699 | struct buffer_head *bitmap_bh = NULL; | ||
700 | |||
701 | lock_super (sb); | ||
702 | es = EXT3_SB(sb)->s_es; | ||
703 | desc_count = 0; | ||
704 | bitmap_count = 0; | ||
705 | gdp = NULL; | ||
706 | for (i = 0; i < EXT3_SB(sb)->s_groups_count; i++) { | ||
707 | gdp = ext3_get_group_desc (sb, i, NULL); | ||
708 | if (!gdp) | ||
709 | continue; | ||
710 | desc_count += le16_to_cpu(gdp->bg_free_inodes_count); | ||
711 | brelse(bitmap_bh); | ||
712 | bitmap_bh = read_inode_bitmap(sb, i); | ||
713 | if (!bitmap_bh) | ||
714 | continue; | ||
715 | |||
716 | x = ext3_count_free(bitmap_bh, EXT3_INODES_PER_GROUP(sb) / 8); | ||
717 | printk("group %d: stored = %d, counted = %lu\n", | ||
718 | i, le16_to_cpu(gdp->bg_free_inodes_count), x); | ||
719 | bitmap_count += x; | ||
720 | } | ||
721 | brelse(bitmap_bh); | ||
722 | printk("ext3_count_free_inodes: stored = %u, computed = %lu, %lu\n", | ||
723 | le32_to_cpu(es->s_free_inodes_count), desc_count, bitmap_count); | ||
724 | unlock_super(sb); | ||
725 | return desc_count; | ||
726 | #else | ||
727 | desc_count = 0; | ||
728 | for (i = 0; i < EXT3_SB(sb)->s_groups_count; i++) { | ||
729 | gdp = ext3_get_group_desc (sb, i, NULL); | ||
730 | if (!gdp) | ||
731 | continue; | ||
732 | desc_count += le16_to_cpu(gdp->bg_free_inodes_count); | ||
733 | cond_resched(); | ||
734 | } | ||
735 | return desc_count; | ||
736 | #endif | ||
737 | } | ||
738 | |||
739 | /* Called at mount-time, super-block is locked */ | ||
740 | unsigned long ext3_count_dirs (struct super_block * sb) | ||
741 | { | ||
742 | unsigned long count = 0; | ||
743 | int i; | ||
744 | |||
745 | for (i = 0; i < EXT3_SB(sb)->s_groups_count; i++) { | ||
746 | struct ext3_group_desc *gdp = ext3_get_group_desc (sb, i, NULL); | ||
747 | if (!gdp) | ||
748 | continue; | ||
749 | count += le16_to_cpu(gdp->bg_used_dirs_count); | ||
750 | } | ||
751 | return count; | ||
752 | } | ||
753 | |||
754 | #ifdef CONFIG_EXT3_CHECK | ||
755 | /* Called at mount-time, super-block is locked */ | ||
756 | void ext3_check_inodes_bitmap (struct super_block * sb) | ||
757 | { | ||
758 | struct ext3_super_block * es; | ||
759 | unsigned long desc_count, bitmap_count, x; | ||
760 | struct buffer_head *bitmap_bh = NULL; | ||
761 | struct ext3_group_desc * gdp; | ||
762 | int i; | ||
763 | |||
764 | es = EXT3_SB(sb)->s_es; | ||
765 | desc_count = 0; | ||
766 | bitmap_count = 0; | ||
767 | gdp = NULL; | ||
768 | for (i = 0; i < EXT3_SB(sb)->s_groups_count; i++) { | ||
769 | gdp = ext3_get_group_desc (sb, i, NULL); | ||
770 | if (!gdp) | ||
771 | continue; | ||
772 | desc_count += le16_to_cpu(gdp->bg_free_inodes_count); | ||
773 | brelse(bitmap_bh); | ||
774 | bitmap_bh = read_inode_bitmap(sb, i); | ||
775 | if (!bitmap_bh) | ||
776 | continue; | ||
777 | |||
778 | x = ext3_count_free(bitmap_bh, EXT3_INODES_PER_GROUP(sb) / 8); | ||
779 | if (le16_to_cpu(gdp->bg_free_inodes_count) != x) | ||
780 | ext3_error (sb, "ext3_check_inodes_bitmap", | ||
781 | "Wrong free inodes count in group %d, " | ||
782 | "stored = %d, counted = %lu", i, | ||
783 | le16_to_cpu(gdp->bg_free_inodes_count), x); | ||
784 | bitmap_count += x; | ||
785 | } | ||
786 | brelse(bitmap_bh); | ||
787 | if (le32_to_cpu(es->s_free_inodes_count) != bitmap_count) | ||
788 | ext3_error (sb, "ext3_check_inodes_bitmap", | ||
789 | "Wrong free inodes count in super block, " | ||
790 | "stored = %lu, counted = %lu", | ||
791 | (unsigned long)le32_to_cpu(es->s_free_inodes_count), | ||
792 | bitmap_count); | ||
793 | } | ||
794 | #endif | ||