File: /usr/src/linux/fs/ext2/ialloc.c
1 /*
2 * linux/fs/ext2/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@dcs.ed.ac.uk), 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/config.h>
16 #include <linux/fs.h>
17 #include <linux/ext2_fs.h>
18 #include <linux/locks.h>
19 #include <linux/quotaops.h>
20
21
22 /*
23 * ialloc.c contains the inodes allocation and deallocation routines
24 */
25
26 /*
27 * The free inodes are managed by bitmaps. A file system contains several
28 * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap
29 * block for inodes, N blocks for the inode table and data blocks.
30 *
31 * The file system contains group descriptors which are located after the
32 * super block. Each descriptor contains the number of the bitmap block and
33 * the free blocks count in the block. The descriptors are loaded in memory
34 * when a file system is mounted (see ext2_read_super).
35 */
36
37
38 /*
39 * Read the inode allocation bitmap for a given block_group, reading
40 * into the specified slot in the superblock's bitmap cache.
41 *
42 * Return >=0 on success or a -ve error code.
43 */
44 static int read_inode_bitmap (struct super_block * sb,
45 unsigned long block_group,
46 unsigned int bitmap_nr)
47 {
48 struct ext2_group_desc * gdp;
49 struct buffer_head * bh = NULL;
50 int retval = 0;
51
52 gdp = ext2_get_group_desc (sb, block_group, NULL);
53 if (!gdp) {
54 retval = -EIO;
55 goto error_out;
56 }
57 bh = bread (sb->s_dev, le32_to_cpu(gdp->bg_inode_bitmap), sb->s_blocksize);
58 if (!bh) {
59 ext2_error (sb, "read_inode_bitmap",
60 "Cannot read inode bitmap - "
61 "block_group = %lu, inode_bitmap = %lu",
62 block_group, (unsigned long) gdp->bg_inode_bitmap);
63 retval = -EIO;
64 }
65 /*
66 * On IO error, just leave a zero in the superblock's block pointer for
67 * this group. The IO will be retried next time.
68 */
69 error_out:
70 sb->u.ext2_sb.s_inode_bitmap_number[bitmap_nr] = block_group;
71 sb->u.ext2_sb.s_inode_bitmap[bitmap_nr] = bh;
72 return retval;
73 }
74
75 /*
76 * load_inode_bitmap loads the inode bitmap for a blocks group
77 *
78 * It maintains a cache for the last bitmaps loaded. This cache is managed
79 * with a LRU algorithm.
80 *
81 * Notes:
82 * 1/ There is one cache per mounted file system.
83 * 2/ If the file system contains less than EXT2_MAX_GROUP_LOADED groups,
84 * this function reads the bitmap without maintaining a LRU cache.
85 *
86 * Return the slot used to store the bitmap, or a -ve error code.
87 */
88 static int load_inode_bitmap (struct super_block * sb,
89 unsigned int block_group)
90 {
91 int i, j, retval = 0;
92 unsigned long inode_bitmap_number;
93 struct buffer_head * inode_bitmap;
94
95 if (block_group >= sb->u.ext2_sb.s_groups_count)
96 ext2_panic (sb, "load_inode_bitmap",
97 "block_group >= groups_count - "
98 "block_group = %d, groups_count = %lu",
99 block_group, sb->u.ext2_sb.s_groups_count);
100 if (sb->u.ext2_sb.s_loaded_inode_bitmaps > 0 &&
101 sb->u.ext2_sb.s_inode_bitmap_number[0] == block_group &&
102 sb->u.ext2_sb.s_inode_bitmap[0] != NULL)
103 return 0;
104 if (sb->u.ext2_sb.s_groups_count <= EXT2_MAX_GROUP_LOADED) {
105 if (sb->u.ext2_sb.s_inode_bitmap[block_group]) {
106 if (sb->u.ext2_sb.s_inode_bitmap_number[block_group] != block_group)
107 ext2_panic (sb, "load_inode_bitmap",
108 "block_group != inode_bitmap_number");
109 else
110 return block_group;
111 } else {
112 retval = read_inode_bitmap (sb, block_group,
113 block_group);
114 if (retval < 0)
115 return retval;
116 return block_group;
117 }
118 }
119
120 for (i = 0; i < sb->u.ext2_sb.s_loaded_inode_bitmaps &&
121 sb->u.ext2_sb.s_inode_bitmap_number[i] != block_group;
122 i++)
123 ;
124 if (i < sb->u.ext2_sb.s_loaded_inode_bitmaps &&
125 sb->u.ext2_sb.s_inode_bitmap_number[i] == block_group) {
126 inode_bitmap_number = sb->u.ext2_sb.s_inode_bitmap_number[i];
127 inode_bitmap = sb->u.ext2_sb.s_inode_bitmap[i];
128 for (j = i; j > 0; j--) {
129 sb->u.ext2_sb.s_inode_bitmap_number[j] =
130 sb->u.ext2_sb.s_inode_bitmap_number[j - 1];
131 sb->u.ext2_sb.s_inode_bitmap[j] =
132 sb->u.ext2_sb.s_inode_bitmap[j - 1];
133 }
134 sb->u.ext2_sb.s_inode_bitmap_number[0] = inode_bitmap_number;
135 sb->u.ext2_sb.s_inode_bitmap[0] = inode_bitmap;
136
137 /*
138 * There's still one special case here --- if inode_bitmap == 0
139 * then our last attempt to read the bitmap failed and we have
140 * just ended up caching that failure. Try again to read it.
141 */
142 if (!inode_bitmap)
143 retval = read_inode_bitmap (sb, block_group, 0);
144
145 } else {
146 if (sb->u.ext2_sb.s_loaded_inode_bitmaps < EXT2_MAX_GROUP_LOADED)
147 sb->u.ext2_sb.s_loaded_inode_bitmaps++;
148 else
149 brelse (sb->u.ext2_sb.s_inode_bitmap[EXT2_MAX_GROUP_LOADED - 1]);
150 for (j = sb->u.ext2_sb.s_loaded_inode_bitmaps - 1; j > 0; j--) {
151 sb->u.ext2_sb.s_inode_bitmap_number[j] =
152 sb->u.ext2_sb.s_inode_bitmap_number[j - 1];
153 sb->u.ext2_sb.s_inode_bitmap[j] =
154 sb->u.ext2_sb.s_inode_bitmap[j - 1];
155 }
156 retval = read_inode_bitmap (sb, block_group, 0);
157 }
158 return retval;
159 }
160
161 /*
162 * NOTE! When we get the inode, we're the only people
163 * that have access to it, and as such there are no
164 * race conditions we have to worry about. The inode
165 * is not on the hash-lists, and it cannot be reached
166 * through the filesystem because the directory entry
167 * has been deleted earlier.
168 *
169 * HOWEVER: we must make sure that we get no aliases,
170 * which means that we have to call "clear_inode()"
171 * _before_ we mark the inode not in use in the inode
172 * bitmaps. Otherwise a newly created file might use
173 * the same inode number (not actually the same pointer
174 * though), and then we'd have two inodes sharing the
175 * same inode number and space on the harddisk.
176 */
177 void ext2_free_inode (struct inode * inode)
178 {
179 struct super_block * sb = inode->i_sb;
180 int is_directory;
181 unsigned long ino;
182 struct buffer_head * bh;
183 struct buffer_head * bh2;
184 unsigned long block_group;
185 unsigned long bit;
186 int bitmap_nr;
187 struct ext2_group_desc * gdp;
188 struct ext2_super_block * es;
189
190 ino = inode->i_ino;
191 ext2_debug ("freeing inode %lu\n", ino);
192
193 /*
194 * Note: we must free any quota before locking the superblock,
195 * as writing the quota to disk may need the lock as well.
196 */
197 if (!is_bad_inode(inode)) {
198 /* Quota is already initialized in iput() */
199 DQUOT_FREE_INODE(sb, inode);
200 DQUOT_DROP(inode);
201 }
202
203 lock_super (sb);
204 es = sb->u.ext2_sb.s_es;
205 is_directory = S_ISDIR(inode->i_mode);
206
207 /* Do this BEFORE marking the inode not in use or returning an error */
208 clear_inode (inode);
209
210 if (ino < EXT2_FIRST_INO(sb) ||
211 ino > le32_to_cpu(es->s_inodes_count)) {
212 ext2_error (sb, "ext2_free_inode",
213 "reserved or nonexistent inode %lu", ino);
214 goto error_return;
215 }
216 block_group = (ino - 1) / EXT2_INODES_PER_GROUP(sb);
217 bit = (ino - 1) % EXT2_INODES_PER_GROUP(sb);
218 bitmap_nr = load_inode_bitmap (sb, block_group);
219 if (bitmap_nr < 0)
220 goto error_return;
221
222 bh = sb->u.ext2_sb.s_inode_bitmap[bitmap_nr];
223
224 /* Ok, now we can actually update the inode bitmaps.. */
225 if (!ext2_clear_bit (bit, bh->b_data))
226 ext2_error (sb, "ext2_free_inode",
227 "bit already cleared for inode %lu", ino);
228 else {
229 gdp = ext2_get_group_desc (sb, block_group, &bh2);
230 if (gdp) {
231 gdp->bg_free_inodes_count =
232 cpu_to_le16(le16_to_cpu(gdp->bg_free_inodes_count) + 1);
233 if (is_directory)
234 gdp->bg_used_dirs_count =
235 cpu_to_le16(le16_to_cpu(gdp->bg_used_dirs_count) - 1);
236 }
237 mark_buffer_dirty(bh2);
238 es->s_free_inodes_count =
239 cpu_to_le32(le32_to_cpu(es->s_free_inodes_count) + 1);
240 mark_buffer_dirty(sb->u.ext2_sb.s_sbh);
241 }
242 mark_buffer_dirty(bh);
243 if (sb->s_flags & MS_SYNCHRONOUS) {
244 ll_rw_block (WRITE, 1, &bh);
245 wait_on_buffer (bh);
246 }
247 sb->s_dirt = 1;
248 error_return:
249 unlock_super (sb);
250 }
251
252 /*
253 * There are two policies for allocating an inode. If the new inode is
254 * a directory, then a forward search is made for a block group with both
255 * free space and a low directory-to-inode ratio; if that fails, then of
256 * the groups with above-average free space, that group with the fewest
257 * directories already is chosen.
258 *
259 * For other inodes, search forward from the parent directory\'s block
260 * group to find a free inode.
261 */
262 struct inode * ext2_new_inode (const struct inode * dir, int mode)
263 {
264 struct super_block * sb;
265 struct buffer_head * bh;
266 struct buffer_head * bh2;
267 int i, j, avefreei;
268 struct inode * inode;
269 int bitmap_nr;
270 struct ext2_group_desc * gdp;
271 struct ext2_group_desc * tmp;
272 struct ext2_super_block * es;
273 int err;
274
275 /* Cannot create files in a deleted directory */
276 if (!dir || !dir->i_nlink)
277 return ERR_PTR(-EPERM);
278
279 sb = dir->i_sb;
280 inode = new_inode(sb);
281 if (!inode)
282 return ERR_PTR(-ENOMEM);
283
284 lock_super (sb);
285 es = sb->u.ext2_sb.s_es;
286 repeat:
287 gdp = NULL; i=0;
288
289 if (S_ISDIR(mode)) {
290 avefreei = le32_to_cpu(es->s_free_inodes_count) /
291 sb->u.ext2_sb.s_groups_count;
292 /* I am not yet convinced that this next bit is necessary.
293 i = dir->u.ext2_i.i_block_group;
294 for (j = 0; j < sb->u.ext2_sb.s_groups_count; j++) {
295 tmp = ext2_get_group_desc (sb, i, &bh2);
296 if (tmp &&
297 (le16_to_cpu(tmp->bg_used_dirs_count) << 8) <
298 le16_to_cpu(tmp->bg_free_inodes_count)) {
299 gdp = tmp;
300 break;
301 }
302 else
303 i = ++i % sb->u.ext2_sb.s_groups_count;
304 }
305 */
306 if (!gdp) {
307 for (j = 0; j < sb->u.ext2_sb.s_groups_count; j++) {
308 tmp = ext2_get_group_desc (sb, j, &bh2);
309 if (tmp &&
310 le16_to_cpu(tmp->bg_free_inodes_count) &&
311 le16_to_cpu(tmp->bg_free_inodes_count) >= avefreei) {
312 if (!gdp ||
313 (le16_to_cpu(tmp->bg_free_blocks_count) >
314 le16_to_cpu(gdp->bg_free_blocks_count))) {
315 i = j;
316 gdp = tmp;
317 }
318 }
319 }
320 }
321 }
322 else
323 {
324 /*
325 * Try to place the inode in its parent directory
326 */
327 i = dir->u.ext2_i.i_block_group;
328 tmp = ext2_get_group_desc (sb, i, &bh2);
329 if (tmp && le16_to_cpu(tmp->bg_free_inodes_count))
330 gdp = tmp;
331 else
332 {
333 /*
334 * Use a quadratic hash to find a group with a
335 * free inode
336 */
337 for (j = 1; j < sb->u.ext2_sb.s_groups_count; j <<= 1) {
338 i += j;
339 if (i >= sb->u.ext2_sb.s_groups_count)
340 i -= sb->u.ext2_sb.s_groups_count;
341 tmp = ext2_get_group_desc (sb, i, &bh2);
342 if (tmp &&
343 le16_to_cpu(tmp->bg_free_inodes_count)) {
344 gdp = tmp;
345 break;
346 }
347 }
348 }
349 if (!gdp) {
350 /*
351 * That failed: try linear search for a free inode
352 */
353 i = dir->u.ext2_i.i_block_group + 1;
354 for (j = 2; j < sb->u.ext2_sb.s_groups_count; j++) {
355 if (++i >= sb->u.ext2_sb.s_groups_count)
356 i = 0;
357 tmp = ext2_get_group_desc (sb, i, &bh2);
358 if (tmp &&
359 le16_to_cpu(tmp->bg_free_inodes_count)) {
360 gdp = tmp;
361 break;
362 }
363 }
364 }
365 }
366
367 err = -ENOSPC;
368 if (!gdp)
369 goto fail;
370
371 err = -EIO;
372 bitmap_nr = load_inode_bitmap (sb, i);
373 if (bitmap_nr < 0)
374 goto fail;
375
376 bh = sb->u.ext2_sb.s_inode_bitmap[bitmap_nr];
377 if ((j = ext2_find_first_zero_bit ((unsigned long *) bh->b_data,
378 EXT2_INODES_PER_GROUP(sb))) <
379 EXT2_INODES_PER_GROUP(sb)) {
380 if (ext2_set_bit (j, bh->b_data)) {
381 ext2_error (sb, "ext2_new_inode",
382 "bit already set for inode %d", j);
383 goto repeat;
384 }
385 mark_buffer_dirty(bh);
386 if (sb->s_flags & MS_SYNCHRONOUS) {
387 ll_rw_block (WRITE, 1, &bh);
388 wait_on_buffer (bh);
389 }
390 } else {
391 if (le16_to_cpu(gdp->bg_free_inodes_count) != 0) {
392 ext2_error (sb, "ext2_new_inode",
393 "Free inodes count corrupted in group %d",
394 i);
395 /* Is it really ENOSPC? */
396 err = -ENOSPC;
397 if (sb->s_flags & MS_RDONLY)
398 goto fail;
399
400 gdp->bg_free_inodes_count = 0;
401 mark_buffer_dirty(bh2);
402 }
403 goto repeat;
404 }
405 j += i * EXT2_INODES_PER_GROUP(sb) + 1;
406 if (j < EXT2_FIRST_INO(sb) || j > le32_to_cpu(es->s_inodes_count)) {
407 ext2_error (sb, "ext2_new_inode",
408 "reserved inode or inode > inodes count - "
409 "block_group = %d,inode=%d", i, j);
410 err = -EIO;
411 goto fail;
412 }
413 gdp->bg_free_inodes_count =
414 cpu_to_le16(le16_to_cpu(gdp->bg_free_inodes_count) - 1);
415 if (S_ISDIR(mode))
416 gdp->bg_used_dirs_count =
417 cpu_to_le16(le16_to_cpu(gdp->bg_used_dirs_count) + 1);
418 mark_buffer_dirty(bh2);
419 es->s_free_inodes_count =
420 cpu_to_le32(le32_to_cpu(es->s_free_inodes_count) - 1);
421 mark_buffer_dirty(sb->u.ext2_sb.s_sbh);
422 sb->s_dirt = 1;
423 inode->i_uid = current->fsuid;
424 if (test_opt (sb, GRPID))
425 inode->i_gid = dir->i_gid;
426 else if (dir->i_mode & S_ISGID) {
427 inode->i_gid = dir->i_gid;
428 if (S_ISDIR(mode))
429 mode |= S_ISGID;
430 } else
431 inode->i_gid = current->fsgid;
432 inode->i_mode = mode;
433
434 inode->i_ino = j;
435 inode->i_blksize = PAGE_SIZE; /* This is the optimal IO size (for stat), not the fs block size */
436 inode->i_blocks = 0;
437 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
438 inode->u.ext2_i.i_new_inode = 1;
439 inode->u.ext2_i.i_flags = dir->u.ext2_i.i_flags;
440 if (S_ISLNK(mode))
441 inode->u.ext2_i.i_flags &= ~(EXT2_IMMUTABLE_FL | EXT2_APPEND_FL);
442 inode->u.ext2_i.i_faddr = 0;
443 inode->u.ext2_i.i_frag_no = 0;
444 inode->u.ext2_i.i_frag_size = 0;
445 inode->u.ext2_i.i_file_acl = 0;
446 inode->u.ext2_i.i_dir_acl = 0;
447 inode->u.ext2_i.i_dtime = 0;
448 inode->u.ext2_i.i_prealloc_count = 0;
449 inode->u.ext2_i.i_block_group = i;
450 if (inode->u.ext2_i.i_flags & EXT2_SYNC_FL)
451 inode->i_flags |= S_SYNC;
452 insert_inode_hash(inode);
453 inode->i_generation = event++;
454 mark_inode_dirty(inode);
455
456 unlock_super (sb);
457 if(DQUOT_ALLOC_INODE(sb, inode)) {
458 DQUOT_DROP(inode);
459 inode->i_flags |= S_NOQUOTA;
460 inode->i_nlink = 0;
461 iput(inode);
462 return ERR_PTR(-EDQUOT);
463 }
464 ext2_debug ("allocating inode %lu\n", inode->i_ino);
465 return inode;
466
467 fail:
468 unlock_super(sb);
469 make_bad_inode(inode);
470 iput(inode);
471 return ERR_PTR(err);
472 }
473
474 unsigned long ext2_count_free_inodes (struct super_block * sb)
475 {
476 #ifdef EXT2FS_DEBUG
477 struct ext2_super_block * es;
478 unsigned long desc_count, bitmap_count, x;
479 int bitmap_nr;
480 struct ext2_group_desc * gdp;
481 int i;
482
483 lock_super (sb);
484 es = sb->u.ext2_sb.s_es;
485 desc_count = 0;
486 bitmap_count = 0;
487 gdp = NULL;
488 for (i = 0; i < sb->u.ext2_sb.s_groups_count; i++) {
489 gdp = ext2_get_group_desc (sb, i, NULL);
490 if (!gdp)
491 continue;
492 desc_count += le16_to_cpu(gdp->bg_free_inodes_count);
493 bitmap_nr = load_inode_bitmap (sb, i);
494 if (bitmap_nr < 0)
495 continue;
496
497 x = ext2_count_free (sb->u.ext2_sb.s_inode_bitmap[bitmap_nr],
498 EXT2_INODES_PER_GROUP(sb) / 8);
499 printk ("group %d: stored = %d, counted = %lu\n",
500 i, le16_to_cpu(gdp->bg_free_inodes_count), x);
501 bitmap_count += x;
502 }
503 printk("ext2_count_free_inodes: stored = %lu, computed = %lu, %lu\n",
504 le32_to_cpu(es->s_free_inodes_count), desc_count, bitmap_count);
505 unlock_super (sb);
506 return desc_count;
507 #else
508 return le32_to_cpu(sb->u.ext2_sb.s_es->s_free_inodes_count);
509 #endif
510 }
511
512 #ifdef CONFIG_EXT2_CHECK
513 /* Called at mount-time, super-block is locked */
514 void ext2_check_inodes_bitmap (struct super_block * sb)
515 {
516 struct ext2_super_block * es;
517 unsigned long desc_count, bitmap_count, x;
518 int bitmap_nr;
519 struct ext2_group_desc * gdp;
520 int i;
521
522 es = sb->u.ext2_sb.s_es;
523 desc_count = 0;
524 bitmap_count = 0;
525 gdp = NULL;
526 for (i = 0; i < sb->u.ext2_sb.s_groups_count; i++) {
527 gdp = ext2_get_group_desc (sb, i, NULL);
528 if (!gdp)
529 continue;
530 desc_count += le16_to_cpu(gdp->bg_free_inodes_count);
531 bitmap_nr = load_inode_bitmap (sb, i);
532 if (bitmap_nr < 0)
533 continue;
534
535 x = ext2_count_free (sb->u.ext2_sb.s_inode_bitmap[bitmap_nr],
536 EXT2_INODES_PER_GROUP(sb) / 8);
537 if (le16_to_cpu(gdp->bg_free_inodes_count) != x)
538 ext2_error (sb, "ext2_check_inodes_bitmap",
539 "Wrong free inodes count in group %d, "
540 "stored = %d, counted = %lu", i,
541 le16_to_cpu(gdp->bg_free_inodes_count), x);
542 bitmap_count += x;
543 }
544 if (le32_to_cpu(es->s_free_inodes_count) != bitmap_count)
545 ext2_error (sb, "ext2_check_inodes_bitmap",
546 "Wrong free inodes count in super block, "
547 "stored = %lu, counted = %lu",
548 (unsigned long) le32_to_cpu(es->s_free_inodes_count),
549 bitmap_count);
550 }
551 #endif
552