blob: bf09cbf938cc155c3328b6cc6d6a8a6f1bed637c [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>
Lukas Czerner785c4bc2011-05-23 18:33:01 +020026#include <trace/events/ext3.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027
28#include <asm/byteorder.h>
29
30#include "xattr.h"
31#include "acl.h"
32
33/*
34 * ialloc.c contains the inodes allocation and deallocation routines
35 */
36
37/*
38 * The free inodes are managed by bitmaps. A file system contains several
39 * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap
40 * block for inodes, N blocks for the inode table and data blocks.
41 *
42 * The file system contains group descriptors which are located after the
43 * super block. Each descriptor contains the number of the bitmap block and
44 * the free blocks count in the block.
45 */
46
47
48/*
49 * Read the inode allocation bitmap for a given block_group, reading
50 * into the specified slot in the superblock's bitmap cache.
51 *
52 * Return buffer_head of bitmap on success or NULL.
53 */
54static struct buffer_head *
55read_inode_bitmap(struct super_block * sb, unsigned long block_group)
56{
57 struct ext3_group_desc *desc;
58 struct buffer_head *bh = NULL;
59
60 desc = ext3_get_group_desc(sb, block_group, NULL);
61 if (!desc)
62 goto error_out;
63
64 bh = sb_bread(sb, le32_to_cpu(desc->bg_inode_bitmap));
65 if (!bh)
66 ext3_error(sb, "read_inode_bitmap",
67 "Cannot read inode bitmap - "
68 "block_group = %lu, inode_bitmap = %u",
69 block_group, le32_to_cpu(desc->bg_inode_bitmap));
70error_out:
71 return bh;
72}
73
74/*
75 * NOTE! When we get the inode, we're the only people
76 * that have access to it, and as such there are no
77 * race conditions we have to worry about. The inode
78 * is not on the hash-lists, and it cannot be reached
79 * through the filesystem because the directory entry
80 * has been deleted earlier.
81 *
82 * HOWEVER: we must make sure that we get no aliases,
83 * which means that we have to call "clear_inode()"
84 * _before_ we mark the inode not in use in the inode
85 * bitmaps. Otherwise a newly created file might use
86 * the same inode number (not actually the same pointer
87 * though), and then we'd have two inodes sharing the
88 * same inode number and space on the harddisk.
89 */
90void ext3_free_inode (handle_t *handle, struct inode * inode)
91{
92 struct super_block * sb = inode->i_sb;
93 int is_directory;
94 unsigned long ino;
95 struct buffer_head *bitmap_bh = NULL;
96 struct buffer_head *bh2;
97 unsigned long block_group;
98 unsigned long bit;
99 struct ext3_group_desc * gdp;
100 struct ext3_super_block * es;
101 struct ext3_sb_info *sbi;
102 int fatal = 0, err;
103
104 if (atomic_read(&inode->i_count) > 1) {
105 printk ("ext3_free_inode: inode has count=%d\n",
106 atomic_read(&inode->i_count));
107 return;
108 }
109 if (inode->i_nlink) {
110 printk ("ext3_free_inode: inode has nlink=%d\n",
111 inode->i_nlink);
112 return;
113 }
114 if (!sb) {
115 printk("ext3_free_inode: inode on nonexistent device\n");
116 return;
117 }
118 sbi = EXT3_SB(sb);
119
120 ino = inode->i_ino;
121 ext3_debug ("freeing inode %lu\n", ino);
Lukas Czerner785c4bc2011-05-23 18:33:01 +0200122 trace_ext3_free_inode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124 is_directory = S_ISDIR(inode->i_mode);
125
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126 es = EXT3_SB(sb)->s_es;
127 if (ino < EXT3_FIRST_INO(sb) || ino > le32_to_cpu(es->s_inodes_count)) {
128 ext3_error (sb, "ext3_free_inode",
129 "reserved or nonexistent inode %lu", ino);
130 goto error_return;
131 }
132 block_group = (ino - 1) / EXT3_INODES_PER_GROUP(sb);
133 bit = (ino - 1) % EXT3_INODES_PER_GROUP(sb);
134 bitmap_bh = read_inode_bitmap(sb, block_group);
135 if (!bitmap_bh)
136 goto error_return;
137
138 BUFFER_TRACE(bitmap_bh, "get_write_access");
139 fatal = ext3_journal_get_write_access(handle, bitmap_bh);
140 if (fatal)
141 goto error_return;
142
143 /* Ok, now we can actually update the inode bitmaps.. */
144 if (!ext3_clear_bit_atomic(sb_bgl_lock(sbi, block_group),
145 bit, bitmap_bh->b_data))
146 ext3_error (sb, "ext3_free_inode",
147 "bit already cleared for inode %lu", ino);
148 else {
149 gdp = ext3_get_group_desc (sb, block_group, &bh2);
150
151 BUFFER_TRACE(bh2, "get_write_access");
152 fatal = ext3_journal_get_write_access(handle, bh2);
153 if (fatal) goto error_return;
154
155 if (gdp) {
156 spin_lock(sb_bgl_lock(sbi, block_group));
Marcin Slusarz50e8a282008-02-08 04:20:13 -0800157 le16_add_cpu(&gdp->bg_free_inodes_count, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158 if (is_directory)
Marcin Slusarz50e8a282008-02-08 04:20:13 -0800159 le16_add_cpu(&gdp->bg_used_dirs_count, -1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700160 spin_unlock(sb_bgl_lock(sbi, block_group));
161 percpu_counter_inc(&sbi->s_freeinodes_counter);
162 if (is_directory)
163 percpu_counter_dec(&sbi->s_dirs_counter);
164
165 }
166 BUFFER_TRACE(bh2, "call ext3_journal_dirty_metadata");
167 err = ext3_journal_dirty_metadata(handle, bh2);
168 if (!fatal) fatal = err;
169 }
170 BUFFER_TRACE(bitmap_bh, "call ext3_journal_dirty_metadata");
171 err = ext3_journal_dirty_metadata(handle, bitmap_bh);
172 if (!fatal)
173 fatal = err;
Christoph Hellwigca41f7b2009-04-27 09:46:42 -0400174
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175error_return:
176 brelse(bitmap_bh);
177 ext3_std_error(sb, fatal);
178}
179
180/*
181 * There are two policies for allocating an inode. If the new inode is
182 * a directory, then a forward search is made for a block group with both
183 * free space and a low directory-to-inode ratio; if that fails, then of
184 * the groups with above-average free space, that group with the fewest
185 * directories already is chosen.
186 *
187 * For other inodes, search forward from the parent directory\'s block
188 * group to find a free inode.
189 */
190static int find_group_dir(struct super_block *sb, struct inode *parent)
191{
192 int ngroups = EXT3_SB(sb)->s_groups_count;
Eric Sandeeneee194e2006-09-27 01:49:30 -0700193 unsigned int freei, avefreei;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194 struct ext3_group_desc *desc, *best_desc = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195 int group, best_group = -1;
196
197 freei = percpu_counter_read_positive(&EXT3_SB(sb)->s_freeinodes_counter);
198 avefreei = freei / ngroups;
199
200 for (group = 0; group < ngroups; group++) {
Eric Sandeenef2fb672007-10-16 23:26:30 -0700201 desc = ext3_get_group_desc (sb, group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202 if (!desc || !desc->bg_free_inodes_count)
203 continue;
204 if (le16_to_cpu(desc->bg_free_inodes_count) < avefreei)
205 continue;
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700206 if (!best_desc ||
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207 (le16_to_cpu(desc->bg_free_blocks_count) >
208 le16_to_cpu(best_desc->bg_free_blocks_count))) {
209 best_group = group;
210 best_desc = desc;
211 }
212 }
213 return best_group;
214}
215
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700216/*
217 * Orlov's allocator for directories.
218 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219 * We always try to spread first-level directories.
220 *
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700221 * If there are blockgroups with both free inodes and free blocks counts
222 * not worse than average we return one with smallest directory count.
223 * Otherwise we simply return a random group.
224 *
225 * For the rest rules look so:
226 *
227 * It's OK to put directory into a group unless
228 * it has too many directories already (max_dirs) or
229 * it has too few free inodes left (min_inodes) or
230 * it has too few free blocks left (min_blocks) or
231 * it's already running too large debt (max_debt).
Benoit Boissinot1cc8dcf2008-04-21 22:45:55 +0000232 * Parent's group is preferred, if it doesn't satisfy these
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700233 * conditions we search cyclically through the rest. If none
234 * of the groups look good we just look for a group with more
235 * free inodes than average (starting at parent's group).
236 *
237 * Debt is incremented each time we allocate a directory and decremented
238 * when we allocate an inode, within 0--255.
239 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240
241#define INODE_COST 64
242#define BLOCK_COST 256
243
244static int find_group_orlov(struct super_block *sb, struct inode *parent)
245{
246 int parent_group = EXT3_I(parent)->i_block_group;
247 struct ext3_sb_info *sbi = EXT3_SB(sb);
248 struct ext3_super_block *es = sbi->s_es;
249 int ngroups = sbi->s_groups_count;
250 int inodes_per_group = EXT3_INODES_PER_GROUP(sb);
Eric Sandeeneee194e2006-09-27 01:49:30 -0700251 unsigned int freei, avefreei;
Mingming Cao1c2bf372006-06-25 05:48:06 -0700252 ext3_fsblk_t freeb, avefreeb;
253 ext3_fsblk_t blocks_per_dir;
Eric Sandeeneee194e2006-09-27 01:49:30 -0700254 unsigned int ndirs;
Mingming Cao1c2bf372006-06-25 05:48:06 -0700255 int max_debt, max_dirs, min_inodes;
256 ext3_grpblk_t min_blocks;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257 int group = -1, i;
258 struct ext3_group_desc *desc;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259
260 freei = percpu_counter_read_positive(&sbi->s_freeinodes_counter);
261 avefreei = freei / ngroups;
262 freeb = percpu_counter_read_positive(&sbi->s_freeblocks_counter);
263 avefreeb = freeb / ngroups;
264 ndirs = percpu_counter_read_positive(&sbi->s_dirs_counter);
265
266 if ((parent == sb->s_root->d_inode) ||
267 (EXT3_I(parent)->i_flags & EXT3_TOPDIR_FL)) {
268 int best_ndir = inodes_per_group;
269 int best_group = -1;
270
271 get_random_bytes(&group, sizeof(group));
272 parent_group = (unsigned)group % ngroups;
273 for (i = 0; i < ngroups; i++) {
274 group = (parent_group + i) % ngroups;
Eric Sandeenef2fb672007-10-16 23:26:30 -0700275 desc = ext3_get_group_desc (sb, group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700276 if (!desc || !desc->bg_free_inodes_count)
277 continue;
278 if (le16_to_cpu(desc->bg_used_dirs_count) >= best_ndir)
279 continue;
280 if (le16_to_cpu(desc->bg_free_inodes_count) < avefreei)
281 continue;
282 if (le16_to_cpu(desc->bg_free_blocks_count) < avefreeb)
283 continue;
284 best_group = group;
285 best_ndir = le16_to_cpu(desc->bg_used_dirs_count);
286 }
287 if (best_group >= 0)
288 return best_group;
289 goto fallback;
290 }
291
292 blocks_per_dir = (le32_to_cpu(es->s_blocks_count) - freeb) / ndirs;
293
294 max_dirs = ndirs / ngroups + inodes_per_group / 16;
295 min_inodes = avefreei - inodes_per_group / 4;
296 min_blocks = avefreeb - EXT3_BLOCKS_PER_GROUP(sb) / 4;
297
Mingming Cao1c2bf372006-06-25 05:48:06 -0700298 max_debt = EXT3_BLOCKS_PER_GROUP(sb) / max(blocks_per_dir, (ext3_fsblk_t)BLOCK_COST);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299 if (max_debt * INODE_COST > inodes_per_group)
300 max_debt = inodes_per_group / INODE_COST;
301 if (max_debt > 255)
302 max_debt = 255;
303 if (max_debt == 0)
304 max_debt = 1;
305
306 for (i = 0; i < ngroups; i++) {
307 group = (parent_group + i) % ngroups;
Eric Sandeenef2fb672007-10-16 23:26:30 -0700308 desc = ext3_get_group_desc (sb, group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309 if (!desc || !desc->bg_free_inodes_count)
310 continue;
311 if (le16_to_cpu(desc->bg_used_dirs_count) >= max_dirs)
312 continue;
313 if (le16_to_cpu(desc->bg_free_inodes_count) < min_inodes)
314 continue;
315 if (le16_to_cpu(desc->bg_free_blocks_count) < min_blocks)
316 continue;
317 return group;
318 }
319
320fallback:
321 for (i = 0; i < ngroups; i++) {
322 group = (parent_group + i) % ngroups;
Eric Sandeenef2fb672007-10-16 23:26:30 -0700323 desc = ext3_get_group_desc (sb, group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324 if (!desc || !desc->bg_free_inodes_count)
325 continue;
326 if (le16_to_cpu(desc->bg_free_inodes_count) >= avefreei)
327 return group;
328 }
329
330 if (avefreei) {
331 /*
332 * The free-inodes counter is approximate, and for really small
333 * filesystems the above test can fail to find any blockgroups
334 */
335 avefreei = 0;
336 goto fallback;
337 }
338
339 return -1;
340}
341
342static int find_group_other(struct super_block *sb, struct inode *parent)
343{
344 int parent_group = EXT3_I(parent)->i_block_group;
345 int ngroups = EXT3_SB(sb)->s_groups_count;
346 struct ext3_group_desc *desc;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347 int group, i;
348
349 /*
350 * Try to place the inode in its parent directory
351 */
352 group = parent_group;
Eric Sandeenef2fb672007-10-16 23:26:30 -0700353 desc = ext3_get_group_desc (sb, group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354 if (desc && le16_to_cpu(desc->bg_free_inodes_count) &&
355 le16_to_cpu(desc->bg_free_blocks_count))
356 return group;
357
358 /*
359 * We're going to place this inode in a different blockgroup from its
360 * parent. We want to cause files in a common directory to all land in
361 * the same blockgroup. But we want files which are in a different
362 * directory which shares a blockgroup with our parent to land in a
363 * different blockgroup.
364 *
365 * So add our directory's i_ino into the starting point for the hash.
366 */
367 group = (group + parent->i_ino) % ngroups;
368
369 /*
370 * Use a quadratic hash to find a group with a free inode and some free
371 * blocks.
372 */
373 for (i = 1; i < ngroups; i <<= 1) {
374 group += i;
375 if (group >= ngroups)
376 group -= ngroups;
Eric Sandeenef2fb672007-10-16 23:26:30 -0700377 desc = ext3_get_group_desc (sb, group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378 if (desc && le16_to_cpu(desc->bg_free_inodes_count) &&
379 le16_to_cpu(desc->bg_free_blocks_count))
380 return group;
381 }
382
383 /*
384 * That failed: try linear search for a free inode, even if that group
385 * has no free blocks.
386 */
387 group = parent_group;
388 for (i = 0; i < ngroups; i++) {
389 if (++group >= ngroups)
390 group = 0;
Eric Sandeenef2fb672007-10-16 23:26:30 -0700391 desc = ext3_get_group_desc (sb, group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392 if (desc && le16_to_cpu(desc->bg_free_inodes_count))
393 return group;
394 }
395
396 return -1;
397}
398
399/*
400 * There are two policies for allocating an inode. If the new inode is
401 * a directory, then a forward search is made for a block group with both
402 * free space and a low directory-to-inode ratio; if that fails, then of
403 * the groups with above-average free space, that group with the fewest
404 * directories already is chosen.
405 *
406 * For other inodes, search forward from the parent directory's block
407 * group to find a free inode.
408 */
Eric Paris2a7dba32011-02-01 11:05:39 -0500409struct inode *ext3_new_inode(handle_t *handle, struct inode * dir,
410 const struct qstr *qstr, int mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411{
412 struct super_block *sb;
413 struct buffer_head *bitmap_bh = NULL;
414 struct buffer_head *bh2;
415 int group;
416 unsigned long ino = 0;
417 struct inode * inode;
418 struct ext3_group_desc * gdp = NULL;
419 struct ext3_super_block * es;
420 struct ext3_inode_info *ei;
421 struct ext3_sb_info *sbi;
422 int err = 0;
423 struct inode *ret;
424 int i;
425
426 /* Cannot create files in a deleted directory */
427 if (!dir || !dir->i_nlink)
428 return ERR_PTR(-EPERM);
429
430 sb = dir->i_sb;
Lukas Czerner785c4bc2011-05-23 18:33:01 +0200431 trace_ext3_request_inode(dir, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432 inode = new_inode(sb);
433 if (!inode)
434 return ERR_PTR(-ENOMEM);
435 ei = EXT3_I(inode);
436
437 sbi = EXT3_SB(sb);
438 es = sbi->s_es;
439 if (S_ISDIR(mode)) {
440 if (test_opt (sb, OLDALLOC))
441 group = find_group_dir(sb, dir);
442 else
443 group = find_group_orlov(sb, dir);
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700444 } else
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445 group = find_group_other(sb, dir);
446
447 err = -ENOSPC;
448 if (group == -1)
449 goto out;
450
451 for (i = 0; i < sbi->s_groups_count; i++) {
452 err = -EIO;
453
454 gdp = ext3_get_group_desc(sb, group, &bh2);
455 if (!gdp)
456 goto fail;
457
458 brelse(bitmap_bh);
459 bitmap_bh = read_inode_bitmap(sb, group);
460 if (!bitmap_bh)
461 goto fail;
462
463 ino = 0;
464
465repeat_in_this_group:
466 ino = ext3_find_next_zero_bit((unsigned long *)
467 bitmap_bh->b_data, EXT3_INODES_PER_GROUP(sb), ino);
468 if (ino < EXT3_INODES_PER_GROUP(sb)) {
469
470 BUFFER_TRACE(bitmap_bh, "get_write_access");
471 err = ext3_journal_get_write_access(handle, bitmap_bh);
472 if (err)
473 goto fail;
474
475 if (!ext3_set_bit_atomic(sb_bgl_lock(sbi, group),
476 ino, bitmap_bh->b_data)) {
477 /* we won it */
478 BUFFER_TRACE(bitmap_bh,
479 "call ext3_journal_dirty_metadata");
480 err = ext3_journal_dirty_metadata(handle,
481 bitmap_bh);
482 if (err)
483 goto fail;
484 goto got;
485 }
486 /* we lost it */
487 journal_release_buffer(handle, bitmap_bh);
488
489 if (++ino < EXT3_INODES_PER_GROUP(sb))
490 goto repeat_in_this_group;
491 }
492
493 /*
494 * This case is possible in concurrent environment. It is very
495 * rare. We cannot repeat the find_group_xxx() call because
496 * that will simply return the same blockgroup, because the
497 * group descriptor metadata has not yet been updated.
498 * So we just go onto the next blockgroup.
499 */
500 if (++group == sbi->s_groups_count)
501 group = 0;
502 }
503 err = -ENOSPC;
504 goto out;
505
506got:
507 ino += group * EXT3_INODES_PER_GROUP(sb) + 1;
508 if (ino < EXT3_FIRST_INO(sb) || ino > le32_to_cpu(es->s_inodes_count)) {
509 ext3_error (sb, "ext3_new_inode",
510 "reserved inode or inode > inodes count - "
511 "block_group = %d, inode=%lu", group, ino);
512 err = -EIO;
513 goto fail;
514 }
515
516 BUFFER_TRACE(bh2, "get_write_access");
517 err = ext3_journal_get_write_access(handle, bh2);
518 if (err) goto fail;
519 spin_lock(sb_bgl_lock(sbi, group));
Marcin Slusarz50e8a282008-02-08 04:20:13 -0800520 le16_add_cpu(&gdp->bg_free_inodes_count, -1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521 if (S_ISDIR(mode)) {
Marcin Slusarz50e8a282008-02-08 04:20:13 -0800522 le16_add_cpu(&gdp->bg_used_dirs_count, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523 }
524 spin_unlock(sb_bgl_lock(sbi, group));
525 BUFFER_TRACE(bh2, "call ext3_journal_dirty_metadata");
526 err = ext3_journal_dirty_metadata(handle, bh2);
527 if (err) goto fail;
528
529 percpu_counter_dec(&sbi->s_freeinodes_counter);
530 if (S_ISDIR(mode))
531 percpu_counter_inc(&sbi->s_dirs_counter);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532
Dmitry Monakhovaab99c22010-03-04 17:31:50 +0300533
534 if (test_opt(sb, GRPID)) {
535 inode->i_mode = mode;
536 inode->i_uid = current_fsuid();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537 inode->i_gid = dir->i_gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538 } else
Dmitry Monakhovaab99c22010-03-04 17:31:50 +0300539 inode_init_owner(inode, dir, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540
541 inode->i_ino = ino;
542 /* This is the optimal IO size (for stat), not the fs block size */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 inode->i_blocks = 0;
544 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME_SEC;
545
546 memset(ei->i_data, 0, sizeof(ei->i_data));
547 ei->i_dir_start_lookup = 0;
548 ei->i_disksize = 0;
549
Duane Griffin04143e22009-01-07 18:07:26 -0800550 ei->i_flags =
551 ext3_mask_flags(mode, EXT3_I(dir)->i_flags & EXT3_FL_INHERITED);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552#ifdef EXT3_FRAGMENTS
553 ei->i_faddr = 0;
554 ei->i_frag_no = 0;
555 ei->i_frag_size = 0;
556#endif
557 ei->i_file_acl = 0;
558 ei->i_dir_acl = 0;
559 ei->i_dtime = 0;
560 ei->i_block_alloc_info = NULL;
561 ei->i_block_group = group;
562
563 ext3_set_inode_flags(inode);
564 if (IS_DIRSYNC(inode))
565 handle->h_sync = 1;
Al Viroc38012d2008-12-30 02:02:50 -0500566 if (insert_inode_locked(inode) < 0) {
567 err = -EINVAL;
568 goto fail_drop;
569 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700570 spin_lock(&sbi->s_next_gen_lock);
571 inode->i_generation = sbi->s_next_generation++;
572 spin_unlock(&sbi->s_next_gen_lock);
573
Linus Torvaldsde329822010-03-29 14:30:19 -0700574 ei->i_state_flags = 0;
575 ext3_set_inode_state(inode, EXT3_STATE_NEW);
576
Jan Karabeb37b82010-08-26 00:54:39 +0200577 /* See comment in ext3_iget for explanation */
578 if (ino >= EXT3_FIRST_INO(sb) + 1 &&
579 EXT3_INODE_SIZE(sb) > EXT3_GOOD_OLD_INODE_SIZE) {
580 ei->i_extra_isize =
581 sizeof(struct ext3_inode) - EXT3_GOOD_OLD_INODE_SIZE;
582 } else {
583 ei->i_extra_isize = 0;
584 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585
586 ret = inode;
Christoph Hellwig871a2932010-03-03 09:05:07 -0500587 dquot_initialize(inode);
Christoph Hellwig63936dd2010-03-03 09:05:01 -0500588 err = dquot_alloc_inode(inode);
589 if (err)
Chris Sykesdc7b5fd2005-09-27 21:45:23 -0700590 goto fail_drop;
Chris Sykesdc7b5fd2005-09-27 21:45:23 -0700591
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592 err = ext3_init_acl(handle, inode, dir);
Chris Sykesdc7b5fd2005-09-27 21:45:23 -0700593 if (err)
594 goto fail_free_drop;
595
Eric Paris2a7dba32011-02-01 11:05:39 -0500596 err = ext3_init_security(handle, inode, dir, qstr);
Chris Sykesdc7b5fd2005-09-27 21:45:23 -0700597 if (err)
598 goto fail_free_drop;
599
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600 err = ext3_mark_inode_dirty(handle, inode);
601 if (err) {
602 ext3_std_error(sb, err);
Chris Sykesdc7b5fd2005-09-27 21:45:23 -0700603 goto fail_free_drop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604 }
605
606 ext3_debug("allocating inode %lu\n", inode->i_ino);
Lukas Czerner785c4bc2011-05-23 18:33:01 +0200607 trace_ext3_allocate_inode(inode, dir, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608 goto really_out;
609fail:
610 ext3_std_error(sb, err);
611out:
612 iput(inode);
613 ret = ERR_PTR(err);
614really_out:
615 brelse(bitmap_bh);
616 return ret;
617
Chris Sykesdc7b5fd2005-09-27 21:45:23 -0700618fail_free_drop:
Christoph Hellwig63936dd2010-03-03 09:05:01 -0500619 dquot_free_inode(inode);
Chris Sykesdc7b5fd2005-09-27 21:45:23 -0700620
621fail_drop:
Christoph Hellwig9f754752010-03-03 09:05:05 -0500622 dquot_drop(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623 inode->i_flags |= S_NOQUOTA;
624 inode->i_nlink = 0;
Al Viroc38012d2008-12-30 02:02:50 -0500625 unlock_new_inode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626 iput(inode);
627 brelse(bitmap_bh);
628 return ERR_PTR(err);
629}
630
631/* Verify that we are loading a valid orphan from disk */
632struct inode *ext3_orphan_get(struct super_block *sb, unsigned long ino)
633{
634 unsigned long max_ino = le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count);
635 unsigned long block_group;
636 int bit;
David Howells473043d2008-02-07 00:15:36 -0800637 struct buffer_head *bitmap_bh;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638 struct inode *inode = NULL;
David Howells473043d2008-02-07 00:15:36 -0800639 long err = -EIO;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640
641 /* Error cases - e2fsck has already cleaned up for us */
642 if (ino > max_ino) {
Harvey Harrisone05b6b52008-04-28 02:16:15 -0700643 ext3_warning(sb, __func__,
Glauber de Oliveira Costa9f406682006-01-08 01:03:22 -0800644 "bad orphan ino %lu! e2fsck was run?", ino);
David Howells473043d2008-02-07 00:15:36 -0800645 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646 }
647
648 block_group = (ino - 1) / EXT3_INODES_PER_GROUP(sb);
649 bit = (ino - 1) % EXT3_INODES_PER_GROUP(sb);
650 bitmap_bh = read_inode_bitmap(sb, block_group);
651 if (!bitmap_bh) {
Harvey Harrisone05b6b52008-04-28 02:16:15 -0700652 ext3_warning(sb, __func__,
Glauber de Oliveira Costa9f406682006-01-08 01:03:22 -0800653 "inode bitmap error for orphan %lu", ino);
David Howells473043d2008-02-07 00:15:36 -0800654 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655 }
656
657 /* Having the inode bit set should be a 100% indicator that this
658 * is a valid orphan (no e2fsck run on fs). Orphans also include
659 * inodes that were being truncated, so we can't check i_nlink==0.
660 */
David Howells473043d2008-02-07 00:15:36 -0800661 if (!ext3_test_bit(bit, bitmap_bh->b_data))
662 goto bad_orphan;
663
664 inode = ext3_iget(sb, ino);
665 if (IS_ERR(inode))
666 goto iget_failed;
667
Duane Griffinae76dd92008-07-25 01:46:23 -0700668 /*
669 * If the orphans has i_nlinks > 0 then it should be able to be
670 * truncated, otherwise it won't be removed from the orphan list
671 * during processing and an infinite loop will result.
672 */
673 if (inode->i_nlink && !ext3_can_truncate(inode))
674 goto bad_orphan;
675
David Howells473043d2008-02-07 00:15:36 -0800676 if (NEXT_ORPHAN(inode) > max_ino)
677 goto bad_orphan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678 brelse(bitmap_bh);
679 return inode;
David Howells473043d2008-02-07 00:15:36 -0800680
681iget_failed:
682 err = PTR_ERR(inode);
683 inode = NULL;
684bad_orphan:
Harvey Harrisone05b6b52008-04-28 02:16:15 -0700685 ext3_warning(sb, __func__,
David Howells473043d2008-02-07 00:15:36 -0800686 "bad orphan inode %lu! e2fsck was run?", ino);
687 printk(KERN_NOTICE "ext3_test_bit(bit=%d, block=%llu) = %d\n",
688 bit, (unsigned long long)bitmap_bh->b_blocknr,
689 ext3_test_bit(bit, bitmap_bh->b_data));
690 printk(KERN_NOTICE "inode=%p\n", inode);
691 if (inode) {
692 printk(KERN_NOTICE "is_bad_inode(inode)=%d\n",
693 is_bad_inode(inode));
694 printk(KERN_NOTICE "NEXT_ORPHAN(inode)=%u\n",
695 NEXT_ORPHAN(inode));
696 printk(KERN_NOTICE "max_ino=%lu\n", max_ino);
Duane Griffinae76dd92008-07-25 01:46:23 -0700697 printk(KERN_NOTICE "i_nlink=%u\n", inode->i_nlink);
David Howells473043d2008-02-07 00:15:36 -0800698 /* Avoid freeing blocks if we got a bad deleted inode */
699 if (inode->i_nlink == 0)
700 inode->i_blocks = 0;
701 iput(inode);
702 }
703 brelse(bitmap_bh);
704error:
705 return ERR_PTR(err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706}
707
708unsigned long ext3_count_free_inodes (struct super_block * sb)
709{
710 unsigned long desc_count;
711 struct ext3_group_desc *gdp;
712 int i;
713#ifdef EXT3FS_DEBUG
714 struct ext3_super_block *es;
715 unsigned long bitmap_count, x;
716 struct buffer_head *bitmap_bh = NULL;
717
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 es = EXT3_SB(sb)->s_es;
719 desc_count = 0;
720 bitmap_count = 0;
721 gdp = NULL;
722 for (i = 0; i < EXT3_SB(sb)->s_groups_count; i++) {
723 gdp = ext3_get_group_desc (sb, i, NULL);
724 if (!gdp)
725 continue;
726 desc_count += le16_to_cpu(gdp->bg_free_inodes_count);
727 brelse(bitmap_bh);
728 bitmap_bh = read_inode_bitmap(sb, i);
729 if (!bitmap_bh)
730 continue;
731
732 x = ext3_count_free(bitmap_bh, EXT3_INODES_PER_GROUP(sb) / 8);
733 printk("group %d: stored = %d, counted = %lu\n",
734 i, le16_to_cpu(gdp->bg_free_inodes_count), x);
735 bitmap_count += x;
736 }
737 brelse(bitmap_bh);
738 printk("ext3_count_free_inodes: stored = %u, computed = %lu, %lu\n",
739 le32_to_cpu(es->s_free_inodes_count), desc_count, bitmap_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740 return desc_count;
741#else
742 desc_count = 0;
743 for (i = 0; i < EXT3_SB(sb)->s_groups_count; i++) {
744 gdp = ext3_get_group_desc (sb, i, NULL);
745 if (!gdp)
746 continue;
747 desc_count += le16_to_cpu(gdp->bg_free_inodes_count);
748 cond_resched();
749 }
750 return desc_count;
751#endif
752}
753
754/* Called at mount-time, super-block is locked */
755unsigned long ext3_count_dirs (struct super_block * sb)
756{
757 unsigned long count = 0;
758 int i;
759
760 for (i = 0; i < EXT3_SB(sb)->s_groups_count; i++) {
761 struct ext3_group_desc *gdp = ext3_get_group_desc (sb, i, NULL);
762 if (!gdp)
763 continue;
764 count += le16_to_cpu(gdp->bg_used_dirs_count);
765 }
766 return count;
767}
768