blob: a1b810230cc503708fa5cb2a91598684a0c3ca41 [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
Linus Torvalds1da177e2005-04-16 15:20:36 -070015#include <linux/quotaops.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070016#include <linux/random.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070017
Al Viro4613ad12012-03-29 22:30:07 -040018#include "ext3.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070019#include "xattr.h"
20#include "acl.h"
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.
34 */
35
36
37/*
38 * Read the inode allocation bitmap for a given block_group, reading
39 * into the specified slot in the superblock's bitmap cache.
40 *
41 * Return buffer_head of bitmap on success or NULL.
42 */
43static struct buffer_head *
44read_inode_bitmap(struct super_block * sb, unsigned long block_group)
45{
46 struct ext3_group_desc *desc;
47 struct buffer_head *bh = NULL;
48
49 desc = ext3_get_group_desc(sb, block_group, NULL);
50 if (!desc)
51 goto error_out;
52
53 bh = sb_bread(sb, le32_to_cpu(desc->bg_inode_bitmap));
54 if (!bh)
55 ext3_error(sb, "read_inode_bitmap",
56 "Cannot read inode bitmap - "
57 "block_group = %lu, inode_bitmap = %u",
58 block_group, le32_to_cpu(desc->bg_inode_bitmap));
59error_out:
60 return bh;
61}
62
63/*
64 * NOTE! When we get the inode, we're the only people
65 * that have access to it, and as such there are no
66 * race conditions we have to worry about. The inode
67 * is not on the hash-lists, and it cannot be reached
68 * through the filesystem because the directory entry
69 * has been deleted earlier.
70 *
71 * HOWEVER: we must make sure that we get no aliases,
72 * which means that we have to call "clear_inode()"
73 * _before_ we mark the inode not in use in the inode
74 * bitmaps. Otherwise a newly created file might use
75 * the same inode number (not actually the same pointer
76 * though), and then we'd have two inodes sharing the
77 * same inode number and space on the harddisk.
78 */
79void ext3_free_inode (handle_t *handle, struct inode * inode)
80{
81 struct super_block * sb = inode->i_sb;
82 int is_directory;
83 unsigned long ino;
84 struct buffer_head *bitmap_bh = NULL;
85 struct buffer_head *bh2;
86 unsigned long block_group;
87 unsigned long bit;
88 struct ext3_group_desc * gdp;
89 struct ext3_super_block * es;
90 struct ext3_sb_info *sbi;
91 int fatal = 0, err;
92
93 if (atomic_read(&inode->i_count) > 1) {
94 printk ("ext3_free_inode: inode has count=%d\n",
95 atomic_read(&inode->i_count));
96 return;
97 }
98 if (inode->i_nlink) {
99 printk ("ext3_free_inode: inode has nlink=%d\n",
100 inode->i_nlink);
101 return;
102 }
103 if (!sb) {
104 printk("ext3_free_inode: inode on nonexistent device\n");
105 return;
106 }
107 sbi = EXT3_SB(sb);
108
109 ino = inode->i_ino;
110 ext3_debug ("freeing inode %lu\n", ino);
Lukas Czerner785c4bc2011-05-23 18:33:01 +0200111 trace_ext3_free_inode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113 is_directory = S_ISDIR(inode->i_mode);
114
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115 es = EXT3_SB(sb)->s_es;
116 if (ino < EXT3_FIRST_INO(sb) || ino > le32_to_cpu(es->s_inodes_count)) {
117 ext3_error (sb, "ext3_free_inode",
118 "reserved or nonexistent inode %lu", ino);
119 goto error_return;
120 }
121 block_group = (ino - 1) / EXT3_INODES_PER_GROUP(sb);
122 bit = (ino - 1) % EXT3_INODES_PER_GROUP(sb);
123 bitmap_bh = read_inode_bitmap(sb, block_group);
124 if (!bitmap_bh)
125 goto error_return;
126
127 BUFFER_TRACE(bitmap_bh, "get_write_access");
128 fatal = ext3_journal_get_write_access(handle, bitmap_bh);
129 if (fatal)
130 goto error_return;
131
132 /* Ok, now we can actually update the inode bitmaps.. */
133 if (!ext3_clear_bit_atomic(sb_bgl_lock(sbi, block_group),
134 bit, bitmap_bh->b_data))
135 ext3_error (sb, "ext3_free_inode",
136 "bit already cleared for inode %lu", ino);
137 else {
138 gdp = ext3_get_group_desc (sb, block_group, &bh2);
139
140 BUFFER_TRACE(bh2, "get_write_access");
141 fatal = ext3_journal_get_write_access(handle, bh2);
142 if (fatal) goto error_return;
143
144 if (gdp) {
145 spin_lock(sb_bgl_lock(sbi, block_group));
Marcin Slusarz50e8a282008-02-08 04:20:13 -0800146 le16_add_cpu(&gdp->bg_free_inodes_count, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147 if (is_directory)
Marcin Slusarz50e8a282008-02-08 04:20:13 -0800148 le16_add_cpu(&gdp->bg_used_dirs_count, -1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149 spin_unlock(sb_bgl_lock(sbi, block_group));
150 percpu_counter_inc(&sbi->s_freeinodes_counter);
151 if (is_directory)
152 percpu_counter_dec(&sbi->s_dirs_counter);
153
154 }
155 BUFFER_TRACE(bh2, "call ext3_journal_dirty_metadata");
156 err = ext3_journal_dirty_metadata(handle, bh2);
157 if (!fatal) fatal = err;
158 }
159 BUFFER_TRACE(bitmap_bh, "call ext3_journal_dirty_metadata");
160 err = ext3_journal_dirty_metadata(handle, bitmap_bh);
161 if (!fatal)
162 fatal = err;
Christoph Hellwigca41f7b2009-04-27 09:46:42 -0400163
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164error_return:
165 brelse(bitmap_bh);
166 ext3_std_error(sb, fatal);
167}
168
169/*
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700170 * Orlov's allocator for directories.
171 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172 * We always try to spread first-level directories.
173 *
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700174 * If there are blockgroups with both free inodes and free blocks counts
175 * not worse than average we return one with smallest directory count.
176 * Otherwise we simply return a random group.
177 *
178 * For the rest rules look so:
179 *
180 * It's OK to put directory into a group unless
181 * it has too many directories already (max_dirs) or
182 * it has too few free inodes left (min_inodes) or
Akira Fujitaac0dd242012-03-27 16:09:16 +0900183 * it has too few free blocks left (min_blocks).
Benoit Boissinot1cc8dcf52008-04-21 22:45:55 +0000184 * Parent's group is preferred, if it doesn't satisfy these
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700185 * conditions we search cyclically through the rest. If none
186 * of the groups look good we just look for a group with more
187 * free inodes than average (starting at parent's group).
188 *
189 * Debt is incremented each time we allocate a directory and decremented
190 * when we allocate an inode, within 0--255.
191 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193static int find_group_orlov(struct super_block *sb, struct inode *parent)
194{
195 int parent_group = EXT3_I(parent)->i_block_group;
196 struct ext3_sb_info *sbi = EXT3_SB(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197 int ngroups = sbi->s_groups_count;
198 int inodes_per_group = EXT3_INODES_PER_GROUP(sb);
Eric Sandeeneee194e2006-09-27 01:49:30 -0700199 unsigned int freei, avefreei;
Mingming Cao1c2bf372006-06-25 05:48:06 -0700200 ext3_fsblk_t freeb, avefreeb;
Eric Sandeeneee194e2006-09-27 01:49:30 -0700201 unsigned int ndirs;
Akira Fujitaac0dd242012-03-27 16:09:16 +0900202 int max_dirs, min_inodes;
Mingming Cao1c2bf372006-06-25 05:48:06 -0700203 ext3_grpblk_t min_blocks;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204 int group = -1, i;
205 struct ext3_group_desc *desc;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206
207 freei = percpu_counter_read_positive(&sbi->s_freeinodes_counter);
208 avefreei = freei / ngroups;
209 freeb = percpu_counter_read_positive(&sbi->s_freeblocks_counter);
210 avefreeb = freeb / ngroups;
211 ndirs = percpu_counter_read_positive(&sbi->s_dirs_counter);
212
213 if ((parent == sb->s_root->d_inode) ||
214 (EXT3_I(parent)->i_flags & EXT3_TOPDIR_FL)) {
215 int best_ndir = inodes_per_group;
216 int best_group = -1;
217
ZhangZhene8781672014-02-26 10:32:41 +0800218 group = prandom_u32();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219 parent_group = (unsigned)group % ngroups;
220 for (i = 0; i < ngroups; i++) {
221 group = (parent_group + i) % ngroups;
Eric Sandeenef2fb672007-10-16 23:26:30 -0700222 desc = ext3_get_group_desc (sb, group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223 if (!desc || !desc->bg_free_inodes_count)
224 continue;
225 if (le16_to_cpu(desc->bg_used_dirs_count) >= best_ndir)
226 continue;
227 if (le16_to_cpu(desc->bg_free_inodes_count) < avefreei)
228 continue;
229 if (le16_to_cpu(desc->bg_free_blocks_count) < avefreeb)
230 continue;
231 best_group = group;
232 best_ndir = le16_to_cpu(desc->bg_used_dirs_count);
233 }
234 if (best_group >= 0)
235 return best_group;
236 goto fallback;
237 }
238
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239 max_dirs = ndirs / ngroups + inodes_per_group / 16;
240 min_inodes = avefreei - inodes_per_group / 4;
241 min_blocks = avefreeb - EXT3_BLOCKS_PER_GROUP(sb) / 4;
242
Linus Torvalds1da177e2005-04-16 15:20:36 -0700243 for (i = 0; i < ngroups; i++) {
244 group = (parent_group + i) % ngroups;
Eric Sandeenef2fb672007-10-16 23:26:30 -0700245 desc = ext3_get_group_desc (sb, group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246 if (!desc || !desc->bg_free_inodes_count)
247 continue;
248 if (le16_to_cpu(desc->bg_used_dirs_count) >= max_dirs)
249 continue;
250 if (le16_to_cpu(desc->bg_free_inodes_count) < min_inodes)
251 continue;
252 if (le16_to_cpu(desc->bg_free_blocks_count) < min_blocks)
253 continue;
254 return group;
255 }
256
257fallback:
258 for (i = 0; i < ngroups; i++) {
259 group = (parent_group + i) % ngroups;
Eric Sandeenef2fb672007-10-16 23:26:30 -0700260 desc = ext3_get_group_desc (sb, group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261 if (!desc || !desc->bg_free_inodes_count)
262 continue;
263 if (le16_to_cpu(desc->bg_free_inodes_count) >= avefreei)
264 return group;
265 }
266
267 if (avefreei) {
268 /*
269 * The free-inodes counter is approximate, and for really small
270 * filesystems the above test can fail to find any blockgroups
271 */
272 avefreei = 0;
273 goto fallback;
274 }
275
276 return -1;
277}
278
279static int find_group_other(struct super_block *sb, struct inode *parent)
280{
281 int parent_group = EXT3_I(parent)->i_block_group;
282 int ngroups = EXT3_SB(sb)->s_groups_count;
283 struct ext3_group_desc *desc;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284 int group, i;
285
286 /*
287 * Try to place the inode in its parent directory
288 */
289 group = parent_group;
Eric Sandeenef2fb672007-10-16 23:26:30 -0700290 desc = ext3_get_group_desc (sb, group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291 if (desc && le16_to_cpu(desc->bg_free_inodes_count) &&
292 le16_to_cpu(desc->bg_free_blocks_count))
293 return group;
294
295 /*
296 * We're going to place this inode in a different blockgroup from its
297 * parent. We want to cause files in a common directory to all land in
298 * the same blockgroup. But we want files which are in a different
299 * directory which shares a blockgroup with our parent to land in a
300 * different blockgroup.
301 *
302 * So add our directory's i_ino into the starting point for the hash.
303 */
304 group = (group + parent->i_ino) % ngroups;
305
306 /*
307 * Use a quadratic hash to find a group with a free inode and some free
308 * blocks.
309 */
310 for (i = 1; i < ngroups; i <<= 1) {
311 group += i;
312 if (group >= ngroups)
313 group -= ngroups;
Eric Sandeenef2fb672007-10-16 23:26:30 -0700314 desc = ext3_get_group_desc (sb, group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315 if (desc && le16_to_cpu(desc->bg_free_inodes_count) &&
316 le16_to_cpu(desc->bg_free_blocks_count))
317 return group;
318 }
319
320 /*
321 * That failed: try linear search for a free inode, even if that group
322 * has no free blocks.
323 */
324 group = parent_group;
325 for (i = 0; i < ngroups; i++) {
326 if (++group >= ngroups)
327 group = 0;
Eric Sandeenef2fb672007-10-16 23:26:30 -0700328 desc = ext3_get_group_desc (sb, group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700329 if (desc && le16_to_cpu(desc->bg_free_inodes_count))
330 return group;
331 }
332
333 return -1;
334}
335
336/*
337 * There are two policies for allocating an inode. If the new inode is
338 * a directory, then a forward search is made for a block group with both
339 * free space and a low directory-to-inode ratio; if that fails, then of
340 * the groups with above-average free space, that group with the fewest
341 * directories already is chosen.
342 *
343 * For other inodes, search forward from the parent directory's block
344 * group to find a free inode.
345 */
Eric Paris2a7dba32011-02-01 11:05:39 -0500346struct inode *ext3_new_inode(handle_t *handle, struct inode * dir,
Al Viro69b34f32011-07-26 02:46:57 -0400347 const struct qstr *qstr, umode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348{
349 struct super_block *sb;
350 struct buffer_head *bitmap_bh = NULL;
351 struct buffer_head *bh2;
352 int group;
353 unsigned long ino = 0;
354 struct inode * inode;
355 struct ext3_group_desc * gdp = NULL;
356 struct ext3_super_block * es;
357 struct ext3_inode_info *ei;
358 struct ext3_sb_info *sbi;
359 int err = 0;
360 struct inode *ret;
361 int i;
362
363 /* Cannot create files in a deleted directory */
364 if (!dir || !dir->i_nlink)
365 return ERR_PTR(-EPERM);
366
367 sb = dir->i_sb;
Lukas Czerner785c4bc2011-05-23 18:33:01 +0200368 trace_ext3_request_inode(dir, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369 inode = new_inode(sb);
370 if (!inode)
371 return ERR_PTR(-ENOMEM);
372 ei = EXT3_I(inode);
373
374 sbi = EXT3_SB(sb);
375 es = sbi->s_es;
Lukas Czernerfbc85402011-08-16 18:08:06 +0200376 if (S_ISDIR(mode))
377 group = find_group_orlov(sb, dir);
378 else
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379 group = find_group_other(sb, dir);
380
381 err = -ENOSPC;
382 if (group == -1)
383 goto out;
384
385 for (i = 0; i < sbi->s_groups_count; i++) {
386 err = -EIO;
387
388 gdp = ext3_get_group_desc(sb, group, &bh2);
389 if (!gdp)
390 goto fail;
391
392 brelse(bitmap_bh);
393 bitmap_bh = read_inode_bitmap(sb, group);
394 if (!bitmap_bh)
395 goto fail;
396
397 ino = 0;
398
399repeat_in_this_group:
400 ino = ext3_find_next_zero_bit((unsigned long *)
401 bitmap_bh->b_data, EXT3_INODES_PER_GROUP(sb), ino);
402 if (ino < EXT3_INODES_PER_GROUP(sb)) {
403
404 BUFFER_TRACE(bitmap_bh, "get_write_access");
405 err = ext3_journal_get_write_access(handle, bitmap_bh);
406 if (err)
407 goto fail;
408
409 if (!ext3_set_bit_atomic(sb_bgl_lock(sbi, group),
410 ino, bitmap_bh->b_data)) {
411 /* we won it */
412 BUFFER_TRACE(bitmap_bh,
413 "call ext3_journal_dirty_metadata");
414 err = ext3_journal_dirty_metadata(handle,
415 bitmap_bh);
416 if (err)
417 goto fail;
418 goto got;
419 }
420 /* we lost it */
421 journal_release_buffer(handle, bitmap_bh);
422
423 if (++ino < EXT3_INODES_PER_GROUP(sb))
424 goto repeat_in_this_group;
425 }
426
427 /*
428 * This case is possible in concurrent environment. It is very
429 * rare. We cannot repeat the find_group_xxx() call because
430 * that will simply return the same blockgroup, because the
431 * group descriptor metadata has not yet been updated.
432 * So we just go onto the next blockgroup.
433 */
434 if (++group == sbi->s_groups_count)
435 group = 0;
436 }
437 err = -ENOSPC;
438 goto out;
439
440got:
441 ino += group * EXT3_INODES_PER_GROUP(sb) + 1;
442 if (ino < EXT3_FIRST_INO(sb) || ino > le32_to_cpu(es->s_inodes_count)) {
443 ext3_error (sb, "ext3_new_inode",
444 "reserved inode or inode > inodes count - "
445 "block_group = %d, inode=%lu", group, ino);
446 err = -EIO;
447 goto fail;
448 }
449
450 BUFFER_TRACE(bh2, "get_write_access");
451 err = ext3_journal_get_write_access(handle, bh2);
452 if (err) goto fail;
453 spin_lock(sb_bgl_lock(sbi, group));
Marcin Slusarz50e8a282008-02-08 04:20:13 -0800454 le16_add_cpu(&gdp->bg_free_inodes_count, -1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455 if (S_ISDIR(mode)) {
Marcin Slusarz50e8a282008-02-08 04:20:13 -0800456 le16_add_cpu(&gdp->bg_used_dirs_count, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457 }
458 spin_unlock(sb_bgl_lock(sbi, group));
459 BUFFER_TRACE(bh2, "call ext3_journal_dirty_metadata");
460 err = ext3_journal_dirty_metadata(handle, bh2);
461 if (err) goto fail;
462
463 percpu_counter_dec(&sbi->s_freeinodes_counter);
464 if (S_ISDIR(mode))
465 percpu_counter_inc(&sbi->s_dirs_counter);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466
Dmitry Monakhovaab99c22010-03-04 17:31:50 +0300467
468 if (test_opt(sb, GRPID)) {
469 inode->i_mode = mode;
470 inode->i_uid = current_fsuid();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471 inode->i_gid = dir->i_gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472 } else
Dmitry Monakhovaab99c22010-03-04 17:31:50 +0300473 inode_init_owner(inode, dir, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474
475 inode->i_ino = ino;
476 /* This is the optimal IO size (for stat), not the fs block size */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477 inode->i_blocks = 0;
478 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME_SEC;
479
480 memset(ei->i_data, 0, sizeof(ei->i_data));
481 ei->i_dir_start_lookup = 0;
482 ei->i_disksize = 0;
483
Duane Griffin04143e22009-01-07 18:07:26 -0800484 ei->i_flags =
485 ext3_mask_flags(mode, EXT3_I(dir)->i_flags & EXT3_FL_INHERITED);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486#ifdef EXT3_FRAGMENTS
487 ei->i_faddr = 0;
488 ei->i_frag_no = 0;
489 ei->i_frag_size = 0;
490#endif
491 ei->i_file_acl = 0;
492 ei->i_dir_acl = 0;
493 ei->i_dtime = 0;
494 ei->i_block_alloc_info = NULL;
495 ei->i_block_group = group;
496
497 ext3_set_inode_flags(inode);
498 if (IS_DIRSYNC(inode))
499 handle->h_sync = 1;
Al Viroc38012d2008-12-30 02:02:50 -0500500 if (insert_inode_locked(inode) < 0) {
Jan Kara1415dd82011-12-08 21:13:46 +0100501 /*
502 * Likely a bitmap corruption causing inode to be allocated
503 * twice.
504 */
505 err = -EIO;
506 goto fail;
Al Viroc38012d2008-12-30 02:02:50 -0500507 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508 spin_lock(&sbi->s_next_gen_lock);
509 inode->i_generation = sbi->s_next_generation++;
510 spin_unlock(&sbi->s_next_gen_lock);
511
Linus Torvaldsde329822010-03-29 14:30:19 -0700512 ei->i_state_flags = 0;
513 ext3_set_inode_state(inode, EXT3_STATE_NEW);
514
Jan Karabeb37b82010-08-26 00:54:39 +0200515 /* See comment in ext3_iget for explanation */
516 if (ino >= EXT3_FIRST_INO(sb) + 1 &&
517 EXT3_INODE_SIZE(sb) > EXT3_GOOD_OLD_INODE_SIZE) {
518 ei->i_extra_isize =
519 sizeof(struct ext3_inode) - EXT3_GOOD_OLD_INODE_SIZE;
520 } else {
521 ei->i_extra_isize = 0;
522 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523
524 ret = inode;
Christoph Hellwig871a2932010-03-03 09:05:07 -0500525 dquot_initialize(inode);
Christoph Hellwig63936dd2010-03-03 09:05:01 -0500526 err = dquot_alloc_inode(inode);
527 if (err)
Chris Sykesdc7b5fd2005-09-27 21:45:23 -0700528 goto fail_drop;
Chris Sykesdc7b5fd2005-09-27 21:45:23 -0700529
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530 err = ext3_init_acl(handle, inode, dir);
Chris Sykesdc7b5fd2005-09-27 21:45:23 -0700531 if (err)
532 goto fail_free_drop;
533
Eric Paris2a7dba32011-02-01 11:05:39 -0500534 err = ext3_init_security(handle, inode, dir, qstr);
Chris Sykesdc7b5fd2005-09-27 21:45:23 -0700535 if (err)
536 goto fail_free_drop;
537
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538 err = ext3_mark_inode_dirty(handle, inode);
539 if (err) {
540 ext3_std_error(sb, err);
Chris Sykesdc7b5fd2005-09-27 21:45:23 -0700541 goto fail_free_drop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542 }
543
544 ext3_debug("allocating inode %lu\n", inode->i_ino);
Lukas Czerner785c4bc2011-05-23 18:33:01 +0200545 trace_ext3_allocate_inode(inode, dir, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546 goto really_out;
547fail:
548 ext3_std_error(sb, err);
549out:
550 iput(inode);
551 ret = ERR_PTR(err);
552really_out:
553 brelse(bitmap_bh);
554 return ret;
555
Chris Sykesdc7b5fd2005-09-27 21:45:23 -0700556fail_free_drop:
Christoph Hellwig63936dd2010-03-03 09:05:01 -0500557 dquot_free_inode(inode);
Chris Sykesdc7b5fd2005-09-27 21:45:23 -0700558
559fail_drop:
Christoph Hellwig9f754752010-03-03 09:05:05 -0500560 dquot_drop(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561 inode->i_flags |= S_NOQUOTA;
Miklos Szeredi6d6b77f2011-10-28 14:13:28 +0200562 clear_nlink(inode);
Al Viroc38012d2008-12-30 02:02:50 -0500563 unlock_new_inode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564 iput(inode);
565 brelse(bitmap_bh);
566 return ERR_PTR(err);
567}
568
569/* Verify that we are loading a valid orphan from disk */
570struct inode *ext3_orphan_get(struct super_block *sb, unsigned long ino)
571{
572 unsigned long max_ino = le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count);
573 unsigned long block_group;
574 int bit;
David Howells473043d2008-02-07 00:15:36 -0800575 struct buffer_head *bitmap_bh;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576 struct inode *inode = NULL;
David Howells473043d2008-02-07 00:15:36 -0800577 long err = -EIO;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578
579 /* Error cases - e2fsck has already cleaned up for us */
580 if (ino > max_ino) {
Harvey Harrisone05b6b52008-04-28 02:16:15 -0700581 ext3_warning(sb, __func__,
Glauber de Oliveira Costa9f406682006-01-08 01:03:22 -0800582 "bad orphan ino %lu! e2fsck was run?", ino);
David Howells473043d2008-02-07 00:15:36 -0800583 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584 }
585
586 block_group = (ino - 1) / EXT3_INODES_PER_GROUP(sb);
587 bit = (ino - 1) % EXT3_INODES_PER_GROUP(sb);
588 bitmap_bh = read_inode_bitmap(sb, block_group);
589 if (!bitmap_bh) {
Harvey Harrisone05b6b52008-04-28 02:16:15 -0700590 ext3_warning(sb, __func__,
Glauber de Oliveira Costa9f406682006-01-08 01:03:22 -0800591 "inode bitmap error for orphan %lu", ino);
David Howells473043d2008-02-07 00:15:36 -0800592 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593 }
594
595 /* Having the inode bit set should be a 100% indicator that this
596 * is a valid orphan (no e2fsck run on fs). Orphans also include
597 * inodes that were being truncated, so we can't check i_nlink==0.
598 */
David Howells473043d2008-02-07 00:15:36 -0800599 if (!ext3_test_bit(bit, bitmap_bh->b_data))
600 goto bad_orphan;
601
602 inode = ext3_iget(sb, ino);
603 if (IS_ERR(inode))
604 goto iget_failed;
605
Duane Griffinae76dd92008-07-25 01:46:23 -0700606 /*
607 * If the orphans has i_nlinks > 0 then it should be able to be
608 * truncated, otherwise it won't be removed from the orphan list
609 * during processing and an infinite loop will result.
610 */
611 if (inode->i_nlink && !ext3_can_truncate(inode))
612 goto bad_orphan;
613
David Howells473043d2008-02-07 00:15:36 -0800614 if (NEXT_ORPHAN(inode) > max_ino)
615 goto bad_orphan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616 brelse(bitmap_bh);
617 return inode;
David Howells473043d2008-02-07 00:15:36 -0800618
619iget_failed:
620 err = PTR_ERR(inode);
621 inode = NULL;
622bad_orphan:
Harvey Harrisone05b6b52008-04-28 02:16:15 -0700623 ext3_warning(sb, __func__,
David Howells473043d2008-02-07 00:15:36 -0800624 "bad orphan inode %lu! e2fsck was run?", ino);
625 printk(KERN_NOTICE "ext3_test_bit(bit=%d, block=%llu) = %d\n",
626 bit, (unsigned long long)bitmap_bh->b_blocknr,
627 ext3_test_bit(bit, bitmap_bh->b_data));
628 printk(KERN_NOTICE "inode=%p\n", inode);
629 if (inode) {
630 printk(KERN_NOTICE "is_bad_inode(inode)=%d\n",
631 is_bad_inode(inode));
632 printk(KERN_NOTICE "NEXT_ORPHAN(inode)=%u\n",
633 NEXT_ORPHAN(inode));
634 printk(KERN_NOTICE "max_ino=%lu\n", max_ino);
Duane Griffinae76dd92008-07-25 01:46:23 -0700635 printk(KERN_NOTICE "i_nlink=%u\n", inode->i_nlink);
David Howells473043d2008-02-07 00:15:36 -0800636 /* Avoid freeing blocks if we got a bad deleted inode */
637 if (inode->i_nlink == 0)
638 inode->i_blocks = 0;
639 iput(inode);
640 }
641 brelse(bitmap_bh);
642error:
643 return ERR_PTR(err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644}
645
646unsigned long ext3_count_free_inodes (struct super_block * sb)
647{
648 unsigned long desc_count;
649 struct ext3_group_desc *gdp;
650 int i;
651#ifdef EXT3FS_DEBUG
652 struct ext3_super_block *es;
653 unsigned long bitmap_count, x;
654 struct buffer_head *bitmap_bh = NULL;
655
Linus Torvalds1da177e2005-04-16 15:20:36 -0700656 es = EXT3_SB(sb)->s_es;
657 desc_count = 0;
658 bitmap_count = 0;
659 gdp = NULL;
660 for (i = 0; i < EXT3_SB(sb)->s_groups_count; i++) {
661 gdp = ext3_get_group_desc (sb, i, NULL);
662 if (!gdp)
663 continue;
664 desc_count += le16_to_cpu(gdp->bg_free_inodes_count);
665 brelse(bitmap_bh);
666 bitmap_bh = read_inode_bitmap(sb, i);
667 if (!bitmap_bh)
668 continue;
669
670 x = ext3_count_free(bitmap_bh, EXT3_INODES_PER_GROUP(sb) / 8);
671 printk("group %d: stored = %d, counted = %lu\n",
672 i, le16_to_cpu(gdp->bg_free_inodes_count), x);
673 bitmap_count += x;
674 }
675 brelse(bitmap_bh);
676 printk("ext3_count_free_inodes: stored = %u, computed = %lu, %lu\n",
677 le32_to_cpu(es->s_free_inodes_count), desc_count, bitmap_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678 return desc_count;
679#else
680 desc_count = 0;
681 for (i = 0; i < EXT3_SB(sb)->s_groups_count; i++) {
682 gdp = ext3_get_group_desc (sb, i, NULL);
683 if (!gdp)
684 continue;
685 desc_count += le16_to_cpu(gdp->bg_free_inodes_count);
686 cond_resched();
687 }
688 return desc_count;
689#endif
690}
691
692/* Called at mount-time, super-block is locked */
693unsigned long ext3_count_dirs (struct super_block * sb)
694{
695 unsigned long count = 0;
696 int i;
697
698 for (i = 0; i < EXT3_SB(sb)->s_groups_count; i++) {
699 struct ext3_group_desc *gdp = ext3_get_group_desc (sb, i, NULL);
700 if (!gdp)
701 continue;
702 count += le16_to_cpu(gdp->bg_used_dirs_count);
703 }
704 return count;
705}
706