blob: bfc2dc43681d41c6b54fe405174cc480de1663e9 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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 */
53static struct buffer_head *
54read_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));
69error_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 */
89void 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
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122 is_directory = S_ISDIR(inode->i_mode);
123
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124 es = EXT3_SB(sb)->s_es;
125 if (ino < EXT3_FIRST_INO(sb) || ino > le32_to_cpu(es->s_inodes_count)) {
126 ext3_error (sb, "ext3_free_inode",
127 "reserved or nonexistent inode %lu", ino);
128 goto error_return;
129 }
130 block_group = (ino - 1) / EXT3_INODES_PER_GROUP(sb);
131 bit = (ino - 1) % EXT3_INODES_PER_GROUP(sb);
132 bitmap_bh = read_inode_bitmap(sb, block_group);
133 if (!bitmap_bh)
134 goto error_return;
135
136 BUFFER_TRACE(bitmap_bh, "get_write_access");
137 fatal = ext3_journal_get_write_access(handle, bitmap_bh);
138 if (fatal)
139 goto error_return;
140
141 /* Ok, now we can actually update the inode bitmaps.. */
142 if (!ext3_clear_bit_atomic(sb_bgl_lock(sbi, block_group),
143 bit, bitmap_bh->b_data))
144 ext3_error (sb, "ext3_free_inode",
145 "bit already cleared for inode %lu", ino);
146 else {
147 gdp = ext3_get_group_desc (sb, block_group, &bh2);
148
149 BUFFER_TRACE(bh2, "get_write_access");
150 fatal = ext3_journal_get_write_access(handle, bh2);
151 if (fatal) goto error_return;
152
153 if (gdp) {
154 spin_lock(sb_bgl_lock(sbi, block_group));
Marcin Slusarz50e8a282008-02-08 04:20:13 -0800155 le16_add_cpu(&gdp->bg_free_inodes_count, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156 if (is_directory)
Marcin Slusarz50e8a282008-02-08 04:20:13 -0800157 le16_add_cpu(&gdp->bg_used_dirs_count, -1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158 spin_unlock(sb_bgl_lock(sbi, block_group));
159 percpu_counter_inc(&sbi->s_freeinodes_counter);
160 if (is_directory)
161 percpu_counter_dec(&sbi->s_dirs_counter);
162
163 }
164 BUFFER_TRACE(bh2, "call ext3_journal_dirty_metadata");
165 err = ext3_journal_dirty_metadata(handle, bh2);
166 if (!fatal) fatal = err;
167 }
168 BUFFER_TRACE(bitmap_bh, "call ext3_journal_dirty_metadata");
169 err = ext3_journal_dirty_metadata(handle, bitmap_bh);
170 if (!fatal)
171 fatal = err;
Christoph Hellwigca41f7b2009-04-27 09:46:42 -0400172
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173error_return:
174 brelse(bitmap_bh);
175 ext3_std_error(sb, fatal);
176}
177
178/*
179 * There are two policies for allocating an inode. If the new inode is
180 * a directory, then a forward search is made for a block group with both
181 * free space and a low directory-to-inode ratio; if that fails, then of
182 * the groups with above-average free space, that group with the fewest
183 * directories already is chosen.
184 *
185 * For other inodes, search forward from the parent directory\'s block
186 * group to find a free inode.
187 */
188static int find_group_dir(struct super_block *sb, struct inode *parent)
189{
190 int ngroups = EXT3_SB(sb)->s_groups_count;
Eric Sandeeneee194e2006-09-27 01:49:30 -0700191 unsigned int freei, avefreei;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192 struct ext3_group_desc *desc, *best_desc = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193 int group, best_group = -1;
194
195 freei = percpu_counter_read_positive(&EXT3_SB(sb)->s_freeinodes_counter);
196 avefreei = freei / ngroups;
197
198 for (group = 0; group < ngroups; group++) {
Eric Sandeenef2fb672007-10-16 23:26:30 -0700199 desc = ext3_get_group_desc (sb, group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700200 if (!desc || !desc->bg_free_inodes_count)
201 continue;
202 if (le16_to_cpu(desc->bg_free_inodes_count) < avefreei)
203 continue;
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700204 if (!best_desc ||
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205 (le16_to_cpu(desc->bg_free_blocks_count) >
206 le16_to_cpu(best_desc->bg_free_blocks_count))) {
207 best_group = group;
208 best_desc = desc;
209 }
210 }
211 return best_group;
212}
213
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700214/*
215 * Orlov's allocator for directories.
216 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217 * We always try to spread first-level directories.
218 *
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700219 * If there are blockgroups with both free inodes and free blocks counts
220 * not worse than average we return one with smallest directory count.
221 * Otherwise we simply return a random group.
222 *
223 * For the rest rules look so:
224 *
225 * It's OK to put directory into a group unless
226 * it has too many directories already (max_dirs) or
227 * it has too few free inodes left (min_inodes) or
228 * it has too few free blocks left (min_blocks) or
229 * it's already running too large debt (max_debt).
Benoit Boissinot1cc8dcf2008-04-21 22:45:55 +0000230 * Parent's group is preferred, if it doesn't satisfy these
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700231 * conditions we search cyclically through the rest. If none
232 * of the groups look good we just look for a group with more
233 * free inodes than average (starting at parent's group).
234 *
235 * Debt is incremented each time we allocate a directory and decremented
236 * when we allocate an inode, within 0--255.
237 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238
239#define INODE_COST 64
240#define BLOCK_COST 256
241
242static int find_group_orlov(struct super_block *sb, struct inode *parent)
243{
244 int parent_group = EXT3_I(parent)->i_block_group;
245 struct ext3_sb_info *sbi = EXT3_SB(sb);
246 struct ext3_super_block *es = sbi->s_es;
247 int ngroups = sbi->s_groups_count;
248 int inodes_per_group = EXT3_INODES_PER_GROUP(sb);
Eric Sandeeneee194e2006-09-27 01:49:30 -0700249 unsigned int freei, avefreei;
Mingming Cao1c2bf372006-06-25 05:48:06 -0700250 ext3_fsblk_t freeb, avefreeb;
251 ext3_fsblk_t blocks_per_dir;
Eric Sandeeneee194e2006-09-27 01:49:30 -0700252 unsigned int ndirs;
Mingming Cao1c2bf372006-06-25 05:48:06 -0700253 int max_debt, max_dirs, min_inodes;
254 ext3_grpblk_t min_blocks;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255 int group = -1, i;
256 struct ext3_group_desc *desc;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257
258 freei = percpu_counter_read_positive(&sbi->s_freeinodes_counter);
259 avefreei = freei / ngroups;
260 freeb = percpu_counter_read_positive(&sbi->s_freeblocks_counter);
261 avefreeb = freeb / ngroups;
262 ndirs = percpu_counter_read_positive(&sbi->s_dirs_counter);
263
264 if ((parent == sb->s_root->d_inode) ||
265 (EXT3_I(parent)->i_flags & EXT3_TOPDIR_FL)) {
266 int best_ndir = inodes_per_group;
267 int best_group = -1;
268
269 get_random_bytes(&group, sizeof(group));
270 parent_group = (unsigned)group % ngroups;
271 for (i = 0; i < ngroups; i++) {
272 group = (parent_group + i) % ngroups;
Eric Sandeenef2fb672007-10-16 23:26:30 -0700273 desc = ext3_get_group_desc (sb, group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274 if (!desc || !desc->bg_free_inodes_count)
275 continue;
276 if (le16_to_cpu(desc->bg_used_dirs_count) >= best_ndir)
277 continue;
278 if (le16_to_cpu(desc->bg_free_inodes_count) < avefreei)
279 continue;
280 if (le16_to_cpu(desc->bg_free_blocks_count) < avefreeb)
281 continue;
282 best_group = group;
283 best_ndir = le16_to_cpu(desc->bg_used_dirs_count);
284 }
285 if (best_group >= 0)
286 return best_group;
287 goto fallback;
288 }
289
290 blocks_per_dir = (le32_to_cpu(es->s_blocks_count) - freeb) / ndirs;
291
292 max_dirs = ndirs / ngroups + inodes_per_group / 16;
293 min_inodes = avefreei - inodes_per_group / 4;
294 min_blocks = avefreeb - EXT3_BLOCKS_PER_GROUP(sb) / 4;
295
Mingming Cao1c2bf372006-06-25 05:48:06 -0700296 max_debt = EXT3_BLOCKS_PER_GROUP(sb) / max(blocks_per_dir, (ext3_fsblk_t)BLOCK_COST);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297 if (max_debt * INODE_COST > inodes_per_group)
298 max_debt = inodes_per_group / INODE_COST;
299 if (max_debt > 255)
300 max_debt = 255;
301 if (max_debt == 0)
302 max_debt = 1;
303
304 for (i = 0; i < ngroups; i++) {
305 group = (parent_group + i) % ngroups;
Eric Sandeenef2fb672007-10-16 23:26:30 -0700306 desc = ext3_get_group_desc (sb, group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307 if (!desc || !desc->bg_free_inodes_count)
308 continue;
309 if (le16_to_cpu(desc->bg_used_dirs_count) >= max_dirs)
310 continue;
311 if (le16_to_cpu(desc->bg_free_inodes_count) < min_inodes)
312 continue;
313 if (le16_to_cpu(desc->bg_free_blocks_count) < min_blocks)
314 continue;
315 return group;
316 }
317
318fallback:
319 for (i = 0; i < ngroups; i++) {
320 group = (parent_group + i) % ngroups;
Eric Sandeenef2fb672007-10-16 23:26:30 -0700321 desc = ext3_get_group_desc (sb, group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322 if (!desc || !desc->bg_free_inodes_count)
323 continue;
324 if (le16_to_cpu(desc->bg_free_inodes_count) >= avefreei)
325 return group;
326 }
327
328 if (avefreei) {
329 /*
330 * The free-inodes counter is approximate, and for really small
331 * filesystems the above test can fail to find any blockgroups
332 */
333 avefreei = 0;
334 goto fallback;
335 }
336
337 return -1;
338}
339
340static int find_group_other(struct super_block *sb, struct inode *parent)
341{
342 int parent_group = EXT3_I(parent)->i_block_group;
343 int ngroups = EXT3_SB(sb)->s_groups_count;
344 struct ext3_group_desc *desc;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345 int group, i;
346
347 /*
348 * Try to place the inode in its parent directory
349 */
350 group = parent_group;
Eric Sandeenef2fb672007-10-16 23:26:30 -0700351 desc = ext3_get_group_desc (sb, group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352 if (desc && le16_to_cpu(desc->bg_free_inodes_count) &&
353 le16_to_cpu(desc->bg_free_blocks_count))
354 return group;
355
356 /*
357 * We're going to place this inode in a different blockgroup from its
358 * parent. We want to cause files in a common directory to all land in
359 * the same blockgroup. But we want files which are in a different
360 * directory which shares a blockgroup with our parent to land in a
361 * different blockgroup.
362 *
363 * So add our directory's i_ino into the starting point for the hash.
364 */
365 group = (group + parent->i_ino) % ngroups;
366
367 /*
368 * Use a quadratic hash to find a group with a free inode and some free
369 * blocks.
370 */
371 for (i = 1; i < ngroups; i <<= 1) {
372 group += i;
373 if (group >= ngroups)
374 group -= ngroups;
Eric Sandeenef2fb672007-10-16 23:26:30 -0700375 desc = ext3_get_group_desc (sb, group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376 if (desc && le16_to_cpu(desc->bg_free_inodes_count) &&
377 le16_to_cpu(desc->bg_free_blocks_count))
378 return group;
379 }
380
381 /*
382 * That failed: try linear search for a free inode, even if that group
383 * has no free blocks.
384 */
385 group = parent_group;
386 for (i = 0; i < ngroups; i++) {
387 if (++group >= ngroups)
388 group = 0;
Eric Sandeenef2fb672007-10-16 23:26:30 -0700389 desc = ext3_get_group_desc (sb, group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390 if (desc && le16_to_cpu(desc->bg_free_inodes_count))
391 return group;
392 }
393
394 return -1;
395}
396
397/*
398 * There are two policies for allocating an inode. If the new inode is
399 * a directory, then a forward search is made for a block group with both
400 * free space and a low directory-to-inode ratio; if that fails, then of
401 * the groups with above-average free space, that group with the fewest
402 * directories already is chosen.
403 *
404 * For other inodes, search forward from the parent directory's block
405 * group to find a free inode.
406 */
Eric Paris2a7dba32011-02-01 11:05:39 -0500407struct inode *ext3_new_inode(handle_t *handle, struct inode * dir,
408 const struct qstr *qstr, int mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409{
410 struct super_block *sb;
411 struct buffer_head *bitmap_bh = NULL;
412 struct buffer_head *bh2;
413 int group;
414 unsigned long ino = 0;
415 struct inode * inode;
416 struct ext3_group_desc * gdp = NULL;
417 struct ext3_super_block * es;
418 struct ext3_inode_info *ei;
419 struct ext3_sb_info *sbi;
420 int err = 0;
421 struct inode *ret;
422 int i;
423
424 /* Cannot create files in a deleted directory */
425 if (!dir || !dir->i_nlink)
426 return ERR_PTR(-EPERM);
427
428 sb = dir->i_sb;
429 inode = new_inode(sb);
430 if (!inode)
431 return ERR_PTR(-ENOMEM);
432 ei = EXT3_I(inode);
433
434 sbi = EXT3_SB(sb);
435 es = sbi->s_es;
436 if (S_ISDIR(mode)) {
437 if (test_opt (sb, OLDALLOC))
438 group = find_group_dir(sb, dir);
439 else
440 group = find_group_orlov(sb, dir);
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700441 } else
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442 group = find_group_other(sb, dir);
443
444 err = -ENOSPC;
445 if (group == -1)
446 goto out;
447
448 for (i = 0; i < sbi->s_groups_count; i++) {
449 err = -EIO;
450
451 gdp = ext3_get_group_desc(sb, group, &bh2);
452 if (!gdp)
453 goto fail;
454
455 brelse(bitmap_bh);
456 bitmap_bh = read_inode_bitmap(sb, group);
457 if (!bitmap_bh)
458 goto fail;
459
460 ino = 0;
461
462repeat_in_this_group:
463 ino = ext3_find_next_zero_bit((unsigned long *)
464 bitmap_bh->b_data, EXT3_INODES_PER_GROUP(sb), ino);
465 if (ino < EXT3_INODES_PER_GROUP(sb)) {
466
467 BUFFER_TRACE(bitmap_bh, "get_write_access");
468 err = ext3_journal_get_write_access(handle, bitmap_bh);
469 if (err)
470 goto fail;
471
472 if (!ext3_set_bit_atomic(sb_bgl_lock(sbi, group),
473 ino, bitmap_bh->b_data)) {
474 /* we won it */
475 BUFFER_TRACE(bitmap_bh,
476 "call ext3_journal_dirty_metadata");
477 err = ext3_journal_dirty_metadata(handle,
478 bitmap_bh);
479 if (err)
480 goto fail;
481 goto got;
482 }
483 /* we lost it */
484 journal_release_buffer(handle, bitmap_bh);
485
486 if (++ino < EXT3_INODES_PER_GROUP(sb))
487 goto repeat_in_this_group;
488 }
489
490 /*
491 * This case is possible in concurrent environment. It is very
492 * rare. We cannot repeat the find_group_xxx() call because
493 * that will simply return the same blockgroup, because the
494 * group descriptor metadata has not yet been updated.
495 * So we just go onto the next blockgroup.
496 */
497 if (++group == sbi->s_groups_count)
498 group = 0;
499 }
500 err = -ENOSPC;
501 goto out;
502
503got:
504 ino += group * EXT3_INODES_PER_GROUP(sb) + 1;
505 if (ino < EXT3_FIRST_INO(sb) || ino > le32_to_cpu(es->s_inodes_count)) {
506 ext3_error (sb, "ext3_new_inode",
507 "reserved inode or inode > inodes count - "
508 "block_group = %d, inode=%lu", group, ino);
509 err = -EIO;
510 goto fail;
511 }
512
513 BUFFER_TRACE(bh2, "get_write_access");
514 err = ext3_journal_get_write_access(handle, bh2);
515 if (err) goto fail;
516 spin_lock(sb_bgl_lock(sbi, group));
Marcin Slusarz50e8a282008-02-08 04:20:13 -0800517 le16_add_cpu(&gdp->bg_free_inodes_count, -1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518 if (S_ISDIR(mode)) {
Marcin Slusarz50e8a282008-02-08 04:20:13 -0800519 le16_add_cpu(&gdp->bg_used_dirs_count, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520 }
521 spin_unlock(sb_bgl_lock(sbi, group));
522 BUFFER_TRACE(bh2, "call ext3_journal_dirty_metadata");
523 err = ext3_journal_dirty_metadata(handle, bh2);
524 if (err) goto fail;
525
526 percpu_counter_dec(&sbi->s_freeinodes_counter);
527 if (S_ISDIR(mode))
528 percpu_counter_inc(&sbi->s_dirs_counter);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529
Dmitry Monakhovaab99c22010-03-04 17:31:50 +0300530
531 if (test_opt(sb, GRPID)) {
532 inode->i_mode = mode;
533 inode->i_uid = current_fsuid();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534 inode->i_gid = dir->i_gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535 } else
Dmitry Monakhovaab99c22010-03-04 17:31:50 +0300536 inode_init_owner(inode, dir, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537
538 inode->i_ino = ino;
539 /* This is the optimal IO size (for stat), not the fs block size */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540 inode->i_blocks = 0;
541 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME_SEC;
542
543 memset(ei->i_data, 0, sizeof(ei->i_data));
544 ei->i_dir_start_lookup = 0;
545 ei->i_disksize = 0;
546
Duane Griffin04143e22009-01-07 18:07:26 -0800547 ei->i_flags =
548 ext3_mask_flags(mode, EXT3_I(dir)->i_flags & EXT3_FL_INHERITED);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549#ifdef EXT3_FRAGMENTS
550 ei->i_faddr = 0;
551 ei->i_frag_no = 0;
552 ei->i_frag_size = 0;
553#endif
554 ei->i_file_acl = 0;
555 ei->i_dir_acl = 0;
556 ei->i_dtime = 0;
557 ei->i_block_alloc_info = NULL;
558 ei->i_block_group = group;
559
560 ext3_set_inode_flags(inode);
561 if (IS_DIRSYNC(inode))
562 handle->h_sync = 1;
Al Viroc38012d2008-12-30 02:02:50 -0500563 if (insert_inode_locked(inode) < 0) {
564 err = -EINVAL;
565 goto fail_drop;
566 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567 spin_lock(&sbi->s_next_gen_lock);
568 inode->i_generation = sbi->s_next_generation++;
569 spin_unlock(&sbi->s_next_gen_lock);
570
Linus Torvaldsde329822010-03-29 14:30:19 -0700571 ei->i_state_flags = 0;
572 ext3_set_inode_state(inode, EXT3_STATE_NEW);
573
Jan Karabeb37b82010-08-26 00:54:39 +0200574 /* See comment in ext3_iget for explanation */
575 if (ino >= EXT3_FIRST_INO(sb) + 1 &&
576 EXT3_INODE_SIZE(sb) > EXT3_GOOD_OLD_INODE_SIZE) {
577 ei->i_extra_isize =
578 sizeof(struct ext3_inode) - EXT3_GOOD_OLD_INODE_SIZE;
579 } else {
580 ei->i_extra_isize = 0;
581 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582
583 ret = inode;
Christoph Hellwig871a2932010-03-03 09:05:07 -0500584 dquot_initialize(inode);
Christoph Hellwig63936dd2010-03-03 09:05:01 -0500585 err = dquot_alloc_inode(inode);
586 if (err)
Chris Sykesdc7b5fd2005-09-27 21:45:23 -0700587 goto fail_drop;
Chris Sykesdc7b5fd2005-09-27 21:45:23 -0700588
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589 err = ext3_init_acl(handle, inode, dir);
Chris Sykesdc7b5fd2005-09-27 21:45:23 -0700590 if (err)
591 goto fail_free_drop;
592
Eric Paris2a7dba32011-02-01 11:05:39 -0500593 err = ext3_init_security(handle, inode, dir, qstr);
Chris Sykesdc7b5fd2005-09-27 21:45:23 -0700594 if (err)
595 goto fail_free_drop;
596
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597 err = ext3_mark_inode_dirty(handle, inode);
598 if (err) {
599 ext3_std_error(sb, err);
Chris Sykesdc7b5fd2005-09-27 21:45:23 -0700600 goto fail_free_drop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601 }
602
603 ext3_debug("allocating inode %lu\n", inode->i_ino);
604 goto really_out;
605fail:
606 ext3_std_error(sb, err);
607out:
608 iput(inode);
609 ret = ERR_PTR(err);
610really_out:
611 brelse(bitmap_bh);
612 return ret;
613
Chris Sykesdc7b5fd2005-09-27 21:45:23 -0700614fail_free_drop:
Christoph Hellwig63936dd2010-03-03 09:05:01 -0500615 dquot_free_inode(inode);
Chris Sykesdc7b5fd2005-09-27 21:45:23 -0700616
617fail_drop:
Christoph Hellwig9f754752010-03-03 09:05:05 -0500618 dquot_drop(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619 inode->i_flags |= S_NOQUOTA;
620 inode->i_nlink = 0;
Al Viroc38012d2008-12-30 02:02:50 -0500621 unlock_new_inode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622 iput(inode);
623 brelse(bitmap_bh);
624 return ERR_PTR(err);
625}
626
627/* Verify that we are loading a valid orphan from disk */
628struct inode *ext3_orphan_get(struct super_block *sb, unsigned long ino)
629{
630 unsigned long max_ino = le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count);
631 unsigned long block_group;
632 int bit;
David Howells473043d2008-02-07 00:15:36 -0800633 struct buffer_head *bitmap_bh;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634 struct inode *inode = NULL;
David Howells473043d2008-02-07 00:15:36 -0800635 long err = -EIO;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636
637 /* Error cases - e2fsck has already cleaned up for us */
638 if (ino > max_ino) {
Harvey Harrisone05b6b52008-04-28 02:16:15 -0700639 ext3_warning(sb, __func__,
Glauber de Oliveira Costa9f406682006-01-08 01:03:22 -0800640 "bad orphan ino %lu! e2fsck was run?", ino);
David Howells473043d2008-02-07 00:15:36 -0800641 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642 }
643
644 block_group = (ino - 1) / EXT3_INODES_PER_GROUP(sb);
645 bit = (ino - 1) % EXT3_INODES_PER_GROUP(sb);
646 bitmap_bh = read_inode_bitmap(sb, block_group);
647 if (!bitmap_bh) {
Harvey Harrisone05b6b52008-04-28 02:16:15 -0700648 ext3_warning(sb, __func__,
Glauber de Oliveira Costa9f406682006-01-08 01:03:22 -0800649 "inode bitmap error for orphan %lu", ino);
David Howells473043d2008-02-07 00:15:36 -0800650 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651 }
652
653 /* Having the inode bit set should be a 100% indicator that this
654 * is a valid orphan (no e2fsck run on fs). Orphans also include
655 * inodes that were being truncated, so we can't check i_nlink==0.
656 */
David Howells473043d2008-02-07 00:15:36 -0800657 if (!ext3_test_bit(bit, bitmap_bh->b_data))
658 goto bad_orphan;
659
660 inode = ext3_iget(sb, ino);
661 if (IS_ERR(inode))
662 goto iget_failed;
663
Duane Griffinae76dd92008-07-25 01:46:23 -0700664 /*
665 * If the orphans has i_nlinks > 0 then it should be able to be
666 * truncated, otherwise it won't be removed from the orphan list
667 * during processing and an infinite loop will result.
668 */
669 if (inode->i_nlink && !ext3_can_truncate(inode))
670 goto bad_orphan;
671
David Howells473043d2008-02-07 00:15:36 -0800672 if (NEXT_ORPHAN(inode) > max_ino)
673 goto bad_orphan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 brelse(bitmap_bh);
675 return inode;
David Howells473043d2008-02-07 00:15:36 -0800676
677iget_failed:
678 err = PTR_ERR(inode);
679 inode = NULL;
680bad_orphan:
Harvey Harrisone05b6b52008-04-28 02:16:15 -0700681 ext3_warning(sb, __func__,
David Howells473043d2008-02-07 00:15:36 -0800682 "bad orphan inode %lu! e2fsck was run?", ino);
683 printk(KERN_NOTICE "ext3_test_bit(bit=%d, block=%llu) = %d\n",
684 bit, (unsigned long long)bitmap_bh->b_blocknr,
685 ext3_test_bit(bit, bitmap_bh->b_data));
686 printk(KERN_NOTICE "inode=%p\n", inode);
687 if (inode) {
688 printk(KERN_NOTICE "is_bad_inode(inode)=%d\n",
689 is_bad_inode(inode));
690 printk(KERN_NOTICE "NEXT_ORPHAN(inode)=%u\n",
691 NEXT_ORPHAN(inode));
692 printk(KERN_NOTICE "max_ino=%lu\n", max_ino);
Duane Griffinae76dd92008-07-25 01:46:23 -0700693 printk(KERN_NOTICE "i_nlink=%u\n", inode->i_nlink);
David Howells473043d2008-02-07 00:15:36 -0800694 /* Avoid freeing blocks if we got a bad deleted inode */
695 if (inode->i_nlink == 0)
696 inode->i_blocks = 0;
697 iput(inode);
698 }
699 brelse(bitmap_bh);
700error:
701 return ERR_PTR(err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702}
703
704unsigned long ext3_count_free_inodes (struct super_block * sb)
705{
706 unsigned long desc_count;
707 struct ext3_group_desc *gdp;
708 int i;
709#ifdef EXT3FS_DEBUG
710 struct ext3_super_block *es;
711 unsigned long bitmap_count, x;
712 struct buffer_head *bitmap_bh = NULL;
713
Linus Torvalds1da177e2005-04-16 15:20:36 -0700714 es = EXT3_SB(sb)->s_es;
715 desc_count = 0;
716 bitmap_count = 0;
717 gdp = NULL;
718 for (i = 0; i < EXT3_SB(sb)->s_groups_count; i++) {
719 gdp = ext3_get_group_desc (sb, i, NULL);
720 if (!gdp)
721 continue;
722 desc_count += le16_to_cpu(gdp->bg_free_inodes_count);
723 brelse(bitmap_bh);
724 bitmap_bh = read_inode_bitmap(sb, i);
725 if (!bitmap_bh)
726 continue;
727
728 x = ext3_count_free(bitmap_bh, EXT3_INODES_PER_GROUP(sb) / 8);
729 printk("group %d: stored = %d, counted = %lu\n",
730 i, le16_to_cpu(gdp->bg_free_inodes_count), x);
731 bitmap_count += x;
732 }
733 brelse(bitmap_bh);
734 printk("ext3_count_free_inodes: stored = %u, computed = %lu, %lu\n",
735 le32_to_cpu(es->s_free_inodes_count), desc_count, bitmap_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736 return desc_count;
737#else
738 desc_count = 0;
739 for (i = 0; i < EXT3_SB(sb)->s_groups_count; i++) {
740 gdp = ext3_get_group_desc (sb, i, NULL);
741 if (!gdp)
742 continue;
743 desc_count += le16_to_cpu(gdp->bg_free_inodes_count);
744 cond_resched();
745 }
746 return desc_count;
747#endif
748}
749
750/* Called at mount-time, super-block is locked */
751unsigned long ext3_count_dirs (struct super_block * sb)
752{
753 unsigned long count = 0;
754 int i;
755
756 for (i = 0; i < EXT3_SB(sb)->s_groups_count; i++) {
757 struct ext3_group_desc *gdp = ext3_get_group_desc (sb, i, NULL);
758 if (!gdp)
759 continue;
760 count += le16_to_cpu(gdp->bg_used_dirs_count);
761 }
762 return count;
763}
764