blob: 1cde28438014bfaccd9b1e2b27785044d9f26aec [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/*
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700181 * Orlov's allocator for directories.
182 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183 * We always try to spread first-level directories.
184 *
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700185 * If there are blockgroups with both free inodes and free blocks counts
186 * not worse than average we return one with smallest directory count.
187 * Otherwise we simply return a random group.
188 *
189 * For the rest rules look so:
190 *
191 * It's OK to put directory into a group unless
192 * it has too many directories already (max_dirs) or
193 * it has too few free inodes left (min_inodes) or
194 * it has too few free blocks left (min_blocks) or
195 * it's already running too large debt (max_debt).
Benoit Boissinot1cc8dcf2008-04-21 22:45:55 +0000196 * Parent's group is preferred, if it doesn't satisfy these
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700197 * conditions we search cyclically through the rest. If none
198 * of the groups look good we just look for a group with more
199 * free inodes than average (starting at parent's group).
200 *
201 * Debt is incremented each time we allocate a directory and decremented
202 * when we allocate an inode, within 0--255.
203 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204
205#define INODE_COST 64
206#define BLOCK_COST 256
207
208static int find_group_orlov(struct super_block *sb, struct inode *parent)
209{
210 int parent_group = EXT3_I(parent)->i_block_group;
211 struct ext3_sb_info *sbi = EXT3_SB(sb);
212 struct ext3_super_block *es = sbi->s_es;
213 int ngroups = sbi->s_groups_count;
214 int inodes_per_group = EXT3_INODES_PER_GROUP(sb);
Eric Sandeeneee194e2006-09-27 01:49:30 -0700215 unsigned int freei, avefreei;
Mingming Cao1c2bf372006-06-25 05:48:06 -0700216 ext3_fsblk_t freeb, avefreeb;
217 ext3_fsblk_t blocks_per_dir;
Eric Sandeeneee194e2006-09-27 01:49:30 -0700218 unsigned int ndirs;
Mingming Cao1c2bf372006-06-25 05:48:06 -0700219 int max_debt, max_dirs, min_inodes;
220 ext3_grpblk_t min_blocks;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221 int group = -1, i;
222 struct ext3_group_desc *desc;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223
224 freei = percpu_counter_read_positive(&sbi->s_freeinodes_counter);
225 avefreei = freei / ngroups;
226 freeb = percpu_counter_read_positive(&sbi->s_freeblocks_counter);
227 avefreeb = freeb / ngroups;
228 ndirs = percpu_counter_read_positive(&sbi->s_dirs_counter);
229
230 if ((parent == sb->s_root->d_inode) ||
231 (EXT3_I(parent)->i_flags & EXT3_TOPDIR_FL)) {
232 int best_ndir = inodes_per_group;
233 int best_group = -1;
234
235 get_random_bytes(&group, sizeof(group));
236 parent_group = (unsigned)group % ngroups;
237 for (i = 0; i < ngroups; i++) {
238 group = (parent_group + i) % ngroups;
Eric Sandeenef2fb672007-10-16 23:26:30 -0700239 desc = ext3_get_group_desc (sb, group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240 if (!desc || !desc->bg_free_inodes_count)
241 continue;
242 if (le16_to_cpu(desc->bg_used_dirs_count) >= best_ndir)
243 continue;
244 if (le16_to_cpu(desc->bg_free_inodes_count) < avefreei)
245 continue;
246 if (le16_to_cpu(desc->bg_free_blocks_count) < avefreeb)
247 continue;
248 best_group = group;
249 best_ndir = le16_to_cpu(desc->bg_used_dirs_count);
250 }
251 if (best_group >= 0)
252 return best_group;
253 goto fallback;
254 }
255
256 blocks_per_dir = (le32_to_cpu(es->s_blocks_count) - freeb) / ndirs;
257
258 max_dirs = ndirs / ngroups + inodes_per_group / 16;
259 min_inodes = avefreei - inodes_per_group / 4;
260 min_blocks = avefreeb - EXT3_BLOCKS_PER_GROUP(sb) / 4;
261
Mingming Cao1c2bf372006-06-25 05:48:06 -0700262 max_debt = EXT3_BLOCKS_PER_GROUP(sb) / max(blocks_per_dir, (ext3_fsblk_t)BLOCK_COST);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263 if (max_debt * INODE_COST > inodes_per_group)
264 max_debt = inodes_per_group / INODE_COST;
265 if (max_debt > 255)
266 max_debt = 255;
267 if (max_debt == 0)
268 max_debt = 1;
269
270 for (i = 0; i < ngroups; i++) {
271 group = (parent_group + i) % ngroups;
Eric Sandeenef2fb672007-10-16 23:26:30 -0700272 desc = ext3_get_group_desc (sb, group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273 if (!desc || !desc->bg_free_inodes_count)
274 continue;
275 if (le16_to_cpu(desc->bg_used_dirs_count) >= max_dirs)
276 continue;
277 if (le16_to_cpu(desc->bg_free_inodes_count) < min_inodes)
278 continue;
279 if (le16_to_cpu(desc->bg_free_blocks_count) < min_blocks)
280 continue;
281 return group;
282 }
283
284fallback:
285 for (i = 0; i < ngroups; i++) {
286 group = (parent_group + i) % ngroups;
Eric Sandeenef2fb672007-10-16 23:26:30 -0700287 desc = ext3_get_group_desc (sb, group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288 if (!desc || !desc->bg_free_inodes_count)
289 continue;
290 if (le16_to_cpu(desc->bg_free_inodes_count) >= avefreei)
291 return group;
292 }
293
294 if (avefreei) {
295 /*
296 * The free-inodes counter is approximate, and for really small
297 * filesystems the above test can fail to find any blockgroups
298 */
299 avefreei = 0;
300 goto fallback;
301 }
302
303 return -1;
304}
305
306static int find_group_other(struct super_block *sb, struct inode *parent)
307{
308 int parent_group = EXT3_I(parent)->i_block_group;
309 int ngroups = EXT3_SB(sb)->s_groups_count;
310 struct ext3_group_desc *desc;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311 int group, i;
312
313 /*
314 * Try to place the inode in its parent directory
315 */
316 group = parent_group;
Eric Sandeenef2fb672007-10-16 23:26:30 -0700317 desc = ext3_get_group_desc (sb, group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700318 if (desc && le16_to_cpu(desc->bg_free_inodes_count) &&
319 le16_to_cpu(desc->bg_free_blocks_count))
320 return group;
321
322 /*
323 * We're going to place this inode in a different blockgroup from its
324 * parent. We want to cause files in a common directory to all land in
325 * the same blockgroup. But we want files which are in a different
326 * directory which shares a blockgroup with our parent to land in a
327 * different blockgroup.
328 *
329 * So add our directory's i_ino into the starting point for the hash.
330 */
331 group = (group + parent->i_ino) % ngroups;
332
333 /*
334 * Use a quadratic hash to find a group with a free inode and some free
335 * blocks.
336 */
337 for (i = 1; i < ngroups; i <<= 1) {
338 group += i;
339 if (group >= ngroups)
340 group -= ngroups;
Eric Sandeenef2fb672007-10-16 23:26:30 -0700341 desc = ext3_get_group_desc (sb, group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342 if (desc && le16_to_cpu(desc->bg_free_inodes_count) &&
343 le16_to_cpu(desc->bg_free_blocks_count))
344 return group;
345 }
346
347 /*
348 * That failed: try linear search for a free inode, even if that group
349 * has no free blocks.
350 */
351 group = parent_group;
352 for (i = 0; i < ngroups; i++) {
353 if (++group >= ngroups)
354 group = 0;
Eric Sandeenef2fb672007-10-16 23:26:30 -0700355 desc = ext3_get_group_desc (sb, group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356 if (desc && le16_to_cpu(desc->bg_free_inodes_count))
357 return group;
358 }
359
360 return -1;
361}
362
363/*
364 * There are two policies for allocating an inode. If the new inode is
365 * a directory, then a forward search is made for a block group with both
366 * free space and a low directory-to-inode ratio; if that fails, then of
367 * the groups with above-average free space, that group with the fewest
368 * directories already is chosen.
369 *
370 * For other inodes, search forward from the parent directory's block
371 * group to find a free inode.
372 */
Eric Paris2a7dba32011-02-01 11:05:39 -0500373struct inode *ext3_new_inode(handle_t *handle, struct inode * dir,
Al Viro69b34f32011-07-26 02:46:57 -0400374 const struct qstr *qstr, umode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700375{
376 struct super_block *sb;
377 struct buffer_head *bitmap_bh = NULL;
378 struct buffer_head *bh2;
379 int group;
380 unsigned long ino = 0;
381 struct inode * inode;
382 struct ext3_group_desc * gdp = NULL;
383 struct ext3_super_block * es;
384 struct ext3_inode_info *ei;
385 struct ext3_sb_info *sbi;
386 int err = 0;
387 struct inode *ret;
388 int i;
389
390 /* Cannot create files in a deleted directory */
391 if (!dir || !dir->i_nlink)
392 return ERR_PTR(-EPERM);
393
394 sb = dir->i_sb;
Lukas Czerner785c4bc2011-05-23 18:33:01 +0200395 trace_ext3_request_inode(dir, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396 inode = new_inode(sb);
397 if (!inode)
398 return ERR_PTR(-ENOMEM);
399 ei = EXT3_I(inode);
400
401 sbi = EXT3_SB(sb);
402 es = sbi->s_es;
Lukas Czernerfbc85402011-08-16 18:08:06 +0200403 if (S_ISDIR(mode))
404 group = find_group_orlov(sb, dir);
405 else
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406 group = find_group_other(sb, dir);
407
408 err = -ENOSPC;
409 if (group == -1)
410 goto out;
411
412 for (i = 0; i < sbi->s_groups_count; i++) {
413 err = -EIO;
414
415 gdp = ext3_get_group_desc(sb, group, &bh2);
416 if (!gdp)
417 goto fail;
418
419 brelse(bitmap_bh);
420 bitmap_bh = read_inode_bitmap(sb, group);
421 if (!bitmap_bh)
422 goto fail;
423
424 ino = 0;
425
426repeat_in_this_group:
427 ino = ext3_find_next_zero_bit((unsigned long *)
428 bitmap_bh->b_data, EXT3_INODES_PER_GROUP(sb), ino);
429 if (ino < EXT3_INODES_PER_GROUP(sb)) {
430
431 BUFFER_TRACE(bitmap_bh, "get_write_access");
432 err = ext3_journal_get_write_access(handle, bitmap_bh);
433 if (err)
434 goto fail;
435
436 if (!ext3_set_bit_atomic(sb_bgl_lock(sbi, group),
437 ino, bitmap_bh->b_data)) {
438 /* we won it */
439 BUFFER_TRACE(bitmap_bh,
440 "call ext3_journal_dirty_metadata");
441 err = ext3_journal_dirty_metadata(handle,
442 bitmap_bh);
443 if (err)
444 goto fail;
445 goto got;
446 }
447 /* we lost it */
448 journal_release_buffer(handle, bitmap_bh);
449
450 if (++ino < EXT3_INODES_PER_GROUP(sb))
451 goto repeat_in_this_group;
452 }
453
454 /*
455 * This case is possible in concurrent environment. It is very
456 * rare. We cannot repeat the find_group_xxx() call because
457 * that will simply return the same blockgroup, because the
458 * group descriptor metadata has not yet been updated.
459 * So we just go onto the next blockgroup.
460 */
461 if (++group == sbi->s_groups_count)
462 group = 0;
463 }
464 err = -ENOSPC;
465 goto out;
466
467got:
468 ino += group * EXT3_INODES_PER_GROUP(sb) + 1;
469 if (ino < EXT3_FIRST_INO(sb) || ino > le32_to_cpu(es->s_inodes_count)) {
470 ext3_error (sb, "ext3_new_inode",
471 "reserved inode or inode > inodes count - "
472 "block_group = %d, inode=%lu", group, ino);
473 err = -EIO;
474 goto fail;
475 }
476
477 BUFFER_TRACE(bh2, "get_write_access");
478 err = ext3_journal_get_write_access(handle, bh2);
479 if (err) goto fail;
480 spin_lock(sb_bgl_lock(sbi, group));
Marcin Slusarz50e8a282008-02-08 04:20:13 -0800481 le16_add_cpu(&gdp->bg_free_inodes_count, -1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482 if (S_ISDIR(mode)) {
Marcin Slusarz50e8a282008-02-08 04:20:13 -0800483 le16_add_cpu(&gdp->bg_used_dirs_count, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484 }
485 spin_unlock(sb_bgl_lock(sbi, group));
486 BUFFER_TRACE(bh2, "call ext3_journal_dirty_metadata");
487 err = ext3_journal_dirty_metadata(handle, bh2);
488 if (err) goto fail;
489
490 percpu_counter_dec(&sbi->s_freeinodes_counter);
491 if (S_ISDIR(mode))
492 percpu_counter_inc(&sbi->s_dirs_counter);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493
Dmitry Monakhovaab99c22010-03-04 17:31:50 +0300494
495 if (test_opt(sb, GRPID)) {
496 inode->i_mode = mode;
497 inode->i_uid = current_fsuid();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498 inode->i_gid = dir->i_gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499 } else
Dmitry Monakhovaab99c22010-03-04 17:31:50 +0300500 inode_init_owner(inode, dir, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501
502 inode->i_ino = ino;
503 /* This is the optimal IO size (for stat), not the fs block size */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504 inode->i_blocks = 0;
505 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME_SEC;
506
507 memset(ei->i_data, 0, sizeof(ei->i_data));
508 ei->i_dir_start_lookup = 0;
509 ei->i_disksize = 0;
510
Duane Griffin04143e22009-01-07 18:07:26 -0800511 ei->i_flags =
512 ext3_mask_flags(mode, EXT3_I(dir)->i_flags & EXT3_FL_INHERITED);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513#ifdef EXT3_FRAGMENTS
514 ei->i_faddr = 0;
515 ei->i_frag_no = 0;
516 ei->i_frag_size = 0;
517#endif
518 ei->i_file_acl = 0;
519 ei->i_dir_acl = 0;
520 ei->i_dtime = 0;
521 ei->i_block_alloc_info = NULL;
522 ei->i_block_group = group;
523
524 ext3_set_inode_flags(inode);
525 if (IS_DIRSYNC(inode))
526 handle->h_sync = 1;
Al Viroc38012d2008-12-30 02:02:50 -0500527 if (insert_inode_locked(inode) < 0) {
Jan Kara1415dd82011-12-08 21:13:46 +0100528 /*
529 * Likely a bitmap corruption causing inode to be allocated
530 * twice.
531 */
532 err = -EIO;
533 goto fail;
Al Viroc38012d2008-12-30 02:02:50 -0500534 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535 spin_lock(&sbi->s_next_gen_lock);
536 inode->i_generation = sbi->s_next_generation++;
537 spin_unlock(&sbi->s_next_gen_lock);
538
Linus Torvaldsde329822010-03-29 14:30:19 -0700539 ei->i_state_flags = 0;
540 ext3_set_inode_state(inode, EXT3_STATE_NEW);
541
Jan Karabeb37b82010-08-26 00:54:39 +0200542 /* See comment in ext3_iget for explanation */
543 if (ino >= EXT3_FIRST_INO(sb) + 1 &&
544 EXT3_INODE_SIZE(sb) > EXT3_GOOD_OLD_INODE_SIZE) {
545 ei->i_extra_isize =
546 sizeof(struct ext3_inode) - EXT3_GOOD_OLD_INODE_SIZE;
547 } else {
548 ei->i_extra_isize = 0;
549 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550
551 ret = inode;
Christoph Hellwig871a2932010-03-03 09:05:07 -0500552 dquot_initialize(inode);
Christoph Hellwig63936dd2010-03-03 09:05:01 -0500553 err = dquot_alloc_inode(inode);
554 if (err)
Chris Sykesdc7b5fd2005-09-27 21:45:23 -0700555 goto fail_drop;
Chris Sykesdc7b5fd2005-09-27 21:45:23 -0700556
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557 err = ext3_init_acl(handle, inode, dir);
Chris Sykesdc7b5fd2005-09-27 21:45:23 -0700558 if (err)
559 goto fail_free_drop;
560
Eric Paris2a7dba32011-02-01 11:05:39 -0500561 err = ext3_init_security(handle, inode, dir, qstr);
Chris Sykesdc7b5fd2005-09-27 21:45:23 -0700562 if (err)
563 goto fail_free_drop;
564
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565 err = ext3_mark_inode_dirty(handle, inode);
566 if (err) {
567 ext3_std_error(sb, err);
Chris Sykesdc7b5fd2005-09-27 21:45:23 -0700568 goto fail_free_drop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569 }
570
571 ext3_debug("allocating inode %lu\n", inode->i_ino);
Lukas Czerner785c4bc2011-05-23 18:33:01 +0200572 trace_ext3_allocate_inode(inode, dir, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573 goto really_out;
574fail:
575 ext3_std_error(sb, err);
576out:
577 iput(inode);
578 ret = ERR_PTR(err);
579really_out:
580 brelse(bitmap_bh);
581 return ret;
582
Chris Sykesdc7b5fd2005-09-27 21:45:23 -0700583fail_free_drop:
Christoph Hellwig63936dd2010-03-03 09:05:01 -0500584 dquot_free_inode(inode);
Chris Sykesdc7b5fd2005-09-27 21:45:23 -0700585
586fail_drop:
Christoph Hellwig9f754752010-03-03 09:05:05 -0500587 dquot_drop(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588 inode->i_flags |= S_NOQUOTA;
Miklos Szeredi6d6b77f2011-10-28 14:13:28 +0200589 clear_nlink(inode);
Al Viroc38012d2008-12-30 02:02:50 -0500590 unlock_new_inode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591 iput(inode);
592 brelse(bitmap_bh);
593 return ERR_PTR(err);
594}
595
596/* Verify that we are loading a valid orphan from disk */
597struct inode *ext3_orphan_get(struct super_block *sb, unsigned long ino)
598{
599 unsigned long max_ino = le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count);
600 unsigned long block_group;
601 int bit;
David Howells473043d2008-02-07 00:15:36 -0800602 struct buffer_head *bitmap_bh;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603 struct inode *inode = NULL;
David Howells473043d2008-02-07 00:15:36 -0800604 long err = -EIO;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700605
606 /* Error cases - e2fsck has already cleaned up for us */
607 if (ino > max_ino) {
Harvey Harrisone05b6b52008-04-28 02:16:15 -0700608 ext3_warning(sb, __func__,
Glauber de Oliveira Costa9f406682006-01-08 01:03:22 -0800609 "bad orphan ino %lu! e2fsck was run?", ino);
David Howells473043d2008-02-07 00:15:36 -0800610 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611 }
612
613 block_group = (ino - 1) / EXT3_INODES_PER_GROUP(sb);
614 bit = (ino - 1) % EXT3_INODES_PER_GROUP(sb);
615 bitmap_bh = read_inode_bitmap(sb, block_group);
616 if (!bitmap_bh) {
Harvey Harrisone05b6b52008-04-28 02:16:15 -0700617 ext3_warning(sb, __func__,
Glauber de Oliveira Costa9f406682006-01-08 01:03:22 -0800618 "inode bitmap error for orphan %lu", ino);
David Howells473043d2008-02-07 00:15:36 -0800619 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700620 }
621
622 /* Having the inode bit set should be a 100% indicator that this
623 * is a valid orphan (no e2fsck run on fs). Orphans also include
624 * inodes that were being truncated, so we can't check i_nlink==0.
625 */
David Howells473043d2008-02-07 00:15:36 -0800626 if (!ext3_test_bit(bit, bitmap_bh->b_data))
627 goto bad_orphan;
628
629 inode = ext3_iget(sb, ino);
630 if (IS_ERR(inode))
631 goto iget_failed;
632
Duane Griffinae76dd92008-07-25 01:46:23 -0700633 /*
634 * If the orphans has i_nlinks > 0 then it should be able to be
635 * truncated, otherwise it won't be removed from the orphan list
636 * during processing and an infinite loop will result.
637 */
638 if (inode->i_nlink && !ext3_can_truncate(inode))
639 goto bad_orphan;
640
David Howells473043d2008-02-07 00:15:36 -0800641 if (NEXT_ORPHAN(inode) > max_ino)
642 goto bad_orphan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643 brelse(bitmap_bh);
644 return inode;
David Howells473043d2008-02-07 00:15:36 -0800645
646iget_failed:
647 err = PTR_ERR(inode);
648 inode = NULL;
649bad_orphan:
Harvey Harrisone05b6b52008-04-28 02:16:15 -0700650 ext3_warning(sb, __func__,
David Howells473043d2008-02-07 00:15:36 -0800651 "bad orphan inode %lu! e2fsck was run?", ino);
652 printk(KERN_NOTICE "ext3_test_bit(bit=%d, block=%llu) = %d\n",
653 bit, (unsigned long long)bitmap_bh->b_blocknr,
654 ext3_test_bit(bit, bitmap_bh->b_data));
655 printk(KERN_NOTICE "inode=%p\n", inode);
656 if (inode) {
657 printk(KERN_NOTICE "is_bad_inode(inode)=%d\n",
658 is_bad_inode(inode));
659 printk(KERN_NOTICE "NEXT_ORPHAN(inode)=%u\n",
660 NEXT_ORPHAN(inode));
661 printk(KERN_NOTICE "max_ino=%lu\n", max_ino);
Duane Griffinae76dd92008-07-25 01:46:23 -0700662 printk(KERN_NOTICE "i_nlink=%u\n", inode->i_nlink);
David Howells473043d2008-02-07 00:15:36 -0800663 /* Avoid freeing blocks if we got a bad deleted inode */
664 if (inode->i_nlink == 0)
665 inode->i_blocks = 0;
666 iput(inode);
667 }
668 brelse(bitmap_bh);
669error:
670 return ERR_PTR(err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671}
672
673unsigned long ext3_count_free_inodes (struct super_block * sb)
674{
675 unsigned long desc_count;
676 struct ext3_group_desc *gdp;
677 int i;
678#ifdef EXT3FS_DEBUG
679 struct ext3_super_block *es;
680 unsigned long bitmap_count, x;
681 struct buffer_head *bitmap_bh = NULL;
682
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683 es = EXT3_SB(sb)->s_es;
684 desc_count = 0;
685 bitmap_count = 0;
686 gdp = NULL;
687 for (i = 0; i < EXT3_SB(sb)->s_groups_count; i++) {
688 gdp = ext3_get_group_desc (sb, i, NULL);
689 if (!gdp)
690 continue;
691 desc_count += le16_to_cpu(gdp->bg_free_inodes_count);
692 brelse(bitmap_bh);
693 bitmap_bh = read_inode_bitmap(sb, i);
694 if (!bitmap_bh)
695 continue;
696
697 x = ext3_count_free(bitmap_bh, EXT3_INODES_PER_GROUP(sb) / 8);
698 printk("group %d: stored = %d, counted = %lu\n",
699 i, le16_to_cpu(gdp->bg_free_inodes_count), x);
700 bitmap_count += x;
701 }
702 brelse(bitmap_bh);
703 printk("ext3_count_free_inodes: stored = %u, computed = %lu, %lu\n",
704 le32_to_cpu(es->s_free_inodes_count), desc_count, bitmap_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705 return desc_count;
706#else
707 desc_count = 0;
708 for (i = 0; i < EXT3_SB(sb)->s_groups_count; i++) {
709 gdp = ext3_get_group_desc (sb, i, NULL);
710 if (!gdp)
711 continue;
712 desc_count += le16_to_cpu(gdp->bg_free_inodes_count);
713 cond_resched();
714 }
715 return desc_count;
716#endif
717}
718
719/* Called at mount-time, super-block is locked */
720unsigned long ext3_count_dirs (struct super_block * sb)
721{
722 unsigned long count = 0;
723 int i;
724
725 for (i = 0; i < EXT3_SB(sb)->s_groups_count; i++) {
726 struct ext3_group_desc *gdp = ext3_get_group_desc (sb, i, NULL);
727 if (!gdp)
728 continue;
729 count += le16_to_cpu(gdp->bg_used_dirs_count);
730 }
731 return count;
732}
733