1
2
3
4
5
6
7
8
9
10
11
12
13
14
15#include <linux/time.h>
16#include <linux/fs.h>
17#include <linux/jbd2.h>
18#include <linux/stat.h>
19#include <linux/string.h>
20#include <linux/quotaops.h>
21#include <linux/buffer_head.h>
22#include <linux/random.h>
23#include <linux/bitops.h>
24#include <linux/blkdev.h>
25#include <asm/byteorder.h>
26
27#include "ext4.h"
28#include "ext4_jbd2.h"
29#include "xattr.h"
30#include "acl.h"
31
32#include <trace/events/ext4.h>
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap)
54{
55 int i;
56
57 if (start_bit >= end_bit)
58 return;
59
60 ext4_debug("mark end bits +%d through +%d used\n", start_bit, end_bit);
61 for (i = start_bit; i < ((start_bit + 7) & ~7UL); i++)
62 ext4_set_bit(i, bitmap);
63 if (i < end_bit)
64 memset(bitmap + (i >> 3), 0xff, (end_bit - i) >> 3);
65}
66
67
68static unsigned ext4_init_inode_bitmap(struct super_block *sb,
69 struct buffer_head *bh,
70 ext4_group_t block_group,
71 struct ext4_group_desc *gdp)
72{
73 J_ASSERT_BH(bh, buffer_locked(bh));
74
75
76
77 if (!ext4_group_desc_csum_verify(sb, block_group, gdp)) {
78 ext4_error(sb, "Checksum bad for group %u", block_group);
79 ext4_free_group_clusters_set(sb, gdp, 0);
80 ext4_free_inodes_set(sb, gdp, 0);
81 ext4_itable_unused_set(sb, gdp, 0);
82 memset(bh->b_data, 0xff, sb->s_blocksize);
83 ext4_inode_bitmap_csum_set(sb, block_group, gdp, bh,
84 EXT4_INODES_PER_GROUP(sb) / 8);
85 return 0;
86 }
87
88 memset(bh->b_data, 0, (EXT4_INODES_PER_GROUP(sb) + 7) / 8);
89 ext4_mark_bitmap_end(EXT4_INODES_PER_GROUP(sb), sb->s_blocksize * 8,
90 bh->b_data);
91 ext4_inode_bitmap_csum_set(sb, block_group, gdp, bh,
92 EXT4_INODES_PER_GROUP(sb) / 8);
93 ext4_group_desc_csum_set(sb, block_group, gdp);
94
95 return EXT4_INODES_PER_GROUP(sb);
96}
97
98void ext4_end_bitmap_read(struct buffer_head *bh, int uptodate)
99{
100 if (uptodate) {
101 set_buffer_uptodate(bh);
102 set_bitmap_uptodate(bh);
103 }
104 unlock_buffer(bh);
105 put_bh(bh);
106}
107
108
109
110
111
112
113
114static struct buffer_head *
115ext4_read_inode_bitmap(struct super_block *sb, ext4_group_t block_group)
116{
117 struct ext4_group_desc *desc;
118 struct buffer_head *bh = NULL;
119 ext4_fsblk_t bitmap_blk;
120
121 desc = ext4_get_group_desc(sb, block_group, NULL);
122 if (!desc)
123 return NULL;
124
125 bitmap_blk = ext4_inode_bitmap(sb, desc);
126 bh = sb_getblk(sb, bitmap_blk);
127 if (unlikely(!bh)) {
128 ext4_error(sb, "Cannot read inode bitmap - "
129 "block_group = %u, inode_bitmap = %llu",
130 block_group, bitmap_blk);
131 return NULL;
132 }
133 if (bitmap_uptodate(bh))
134 goto verify;
135
136 lock_buffer(bh);
137 if (bitmap_uptodate(bh)) {
138 unlock_buffer(bh);
139 goto verify;
140 }
141
142 ext4_lock_group(sb, block_group);
143 if (desc->bg_flags & cpu_to_le16(EXT4_BG_INODE_UNINIT)) {
144 ext4_init_inode_bitmap(sb, bh, block_group, desc);
145 set_bitmap_uptodate(bh);
146 set_buffer_uptodate(bh);
147 set_buffer_verified(bh);
148 ext4_unlock_group(sb, block_group);
149 unlock_buffer(bh);
150 return bh;
151 }
152 ext4_unlock_group(sb, block_group);
153
154 if (buffer_uptodate(bh)) {
155
156
157
158
159 set_bitmap_uptodate(bh);
160 unlock_buffer(bh);
161 goto verify;
162 }
163
164
165
166 trace_ext4_load_inode_bitmap(sb, block_group);
167 bh->b_end_io = ext4_end_bitmap_read;
168 get_bh(bh);
169 submit_bh(READ | REQ_META | REQ_PRIO, bh);
170 wait_on_buffer(bh);
171 if (!buffer_uptodate(bh)) {
172 put_bh(bh);
173 ext4_error(sb, "Cannot read inode bitmap - "
174 "block_group = %u, inode_bitmap = %llu",
175 block_group, bitmap_blk);
176 return NULL;
177 }
178
179verify:
180 ext4_lock_group(sb, block_group);
181 if (!buffer_verified(bh) &&
182 !ext4_inode_bitmap_csum_verify(sb, block_group, desc, bh,
183 EXT4_INODES_PER_GROUP(sb) / 8)) {
184 ext4_unlock_group(sb, block_group);
185 put_bh(bh);
186 ext4_error(sb, "Corrupt inode bitmap - block_group = %u, "
187 "inode_bitmap = %llu", block_group, bitmap_blk);
188 return NULL;
189 }
190 ext4_unlock_group(sb, block_group);
191 set_buffer_verified(bh);
192 return bh;
193}
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211void ext4_free_inode(handle_t *handle, struct inode *inode)
212{
213 struct super_block *sb = inode->i_sb;
214 int is_directory;
215 unsigned long ino;
216 struct buffer_head *bitmap_bh = NULL;
217 struct buffer_head *bh2;
218 ext4_group_t block_group;
219 unsigned long bit;
220 struct ext4_group_desc *gdp;
221 struct ext4_super_block *es;
222 struct ext4_sb_info *sbi;
223 int fatal = 0, err, count, cleared;
224
225 if (!sb) {
226 printk(KERN_ERR "EXT4-fs: %s:%d: inode on "
227 "nonexistent device\n", __func__, __LINE__);
228 return;
229 }
230 if (atomic_read(&inode->i_count) > 1) {
231 ext4_msg(sb, KERN_ERR, "%s:%d: inode #%lu: count=%d",
232 __func__, __LINE__, inode->i_ino,
233 atomic_read(&inode->i_count));
234 return;
235 }
236 if (inode->i_nlink) {
237 ext4_msg(sb, KERN_ERR, "%s:%d: inode #%lu: nlink=%d\n",
238 __func__, __LINE__, inode->i_ino, inode->i_nlink);
239 return;
240 }
241 sbi = EXT4_SB(sb);
242
243 ino = inode->i_ino;
244 ext4_debug("freeing inode %lu\n", ino);
245 trace_ext4_free_inode(inode);
246
247
248
249
250
251 dquot_initialize(inode);
252 ext4_xattr_delete_inode(handle, inode);
253 dquot_free_inode(inode);
254 dquot_drop(inode);
255
256 is_directory = S_ISDIR(inode->i_mode);
257
258
259 ext4_clear_inode(inode);
260
261 es = EXT4_SB(sb)->s_es;
262 if (ino < EXT4_FIRST_INO(sb) || ino > le32_to_cpu(es->s_inodes_count)) {
263 ext4_error(sb, "reserved or nonexistent inode %lu", ino);
264 goto error_return;
265 }
266 block_group = (ino - 1) / EXT4_INODES_PER_GROUP(sb);
267 bit = (ino - 1) % EXT4_INODES_PER_GROUP(sb);
268 bitmap_bh = ext4_read_inode_bitmap(sb, block_group);
269 if (!bitmap_bh)
270 goto error_return;
271
272 BUFFER_TRACE(bitmap_bh, "get_write_access");
273 fatal = ext4_journal_get_write_access(handle, bitmap_bh);
274 if (fatal)
275 goto error_return;
276
277 fatal = -ESRCH;
278 gdp = ext4_get_group_desc(sb, block_group, &bh2);
279 if (gdp) {
280 BUFFER_TRACE(bh2, "get_write_access");
281 fatal = ext4_journal_get_write_access(handle, bh2);
282 }
283 ext4_lock_group(sb, block_group);
284 cleared = ext4_test_and_clear_bit(bit, bitmap_bh->b_data);
285 if (fatal || !cleared) {
286 ext4_unlock_group(sb, block_group);
287 goto out;
288 }
289
290 count = ext4_free_inodes_count(sb, gdp) + 1;
291 ext4_free_inodes_set(sb, gdp, count);
292 if (is_directory) {
293 count = ext4_used_dirs_count(sb, gdp) - 1;
294 ext4_used_dirs_set(sb, gdp, count);
295 percpu_counter_dec(&sbi->s_dirs_counter);
296 }
297 ext4_inode_bitmap_csum_set(sb, block_group, gdp, bitmap_bh,
298 EXT4_INODES_PER_GROUP(sb) / 8);
299 ext4_group_desc_csum_set(sb, block_group, gdp);
300 ext4_unlock_group(sb, block_group);
301
302 percpu_counter_inc(&sbi->s_freeinodes_counter);
303 if (sbi->s_log_groups_per_flex) {
304 ext4_group_t f = ext4_flex_group(sbi, block_group);
305
306 atomic_inc(&sbi->s_flex_groups[f].free_inodes);
307 if (is_directory)
308 atomic_dec(&sbi->s_flex_groups[f].used_dirs);
309 }
310 BUFFER_TRACE(bh2, "call ext4_handle_dirty_metadata");
311 fatal = ext4_handle_dirty_metadata(handle, NULL, bh2);
312out:
313 if (cleared) {
314 BUFFER_TRACE(bitmap_bh, "call ext4_handle_dirty_metadata");
315 err = ext4_handle_dirty_metadata(handle, NULL, bitmap_bh);
316 if (!fatal)
317 fatal = err;
318 } else
319 ext4_error(sb, "bit already cleared for inode %lu", ino);
320
321error_return:
322 brelse(bitmap_bh);
323 ext4_std_error(sb, fatal);
324}
325
326struct orlov_stats {
327 __u64 free_clusters;
328 __u32 free_inodes;
329 __u32 used_dirs;
330};
331
332
333
334
335
336
337static void get_orlov_stats(struct super_block *sb, ext4_group_t g,
338 int flex_size, struct orlov_stats *stats)
339{
340 struct ext4_group_desc *desc;
341 struct flex_groups *flex_group = EXT4_SB(sb)->s_flex_groups;
342
343 if (flex_size > 1) {
344 stats->free_inodes = atomic_read(&flex_group[g].free_inodes);
345 stats->free_clusters = atomic64_read(&flex_group[g].free_clusters);
346 stats->used_dirs = atomic_read(&flex_group[g].used_dirs);
347 return;
348 }
349
350 desc = ext4_get_group_desc(sb, g, NULL);
351 if (desc) {
352 stats->free_inodes = ext4_free_inodes_count(sb, desc);
353 stats->free_clusters = ext4_free_group_clusters(sb, desc);
354 stats->used_dirs = ext4_used_dirs_count(sb, desc);
355 } else {
356 stats->free_inodes = 0;
357 stats->free_clusters = 0;
358 stats->used_dirs = 0;
359 }
360}
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383static int find_group_orlov(struct super_block *sb, struct inode *parent,
384 ext4_group_t *group, umode_t mode,
385 const struct qstr *qstr)
386{
387 ext4_group_t parent_group = EXT4_I(parent)->i_block_group;
388 struct ext4_sb_info *sbi = EXT4_SB(sb);
389 ext4_group_t real_ngroups = ext4_get_groups_count(sb);
390 int inodes_per_group = EXT4_INODES_PER_GROUP(sb);
391 unsigned int freei, avefreei, grp_free;
392 ext4_fsblk_t freeb, avefreec;
393 unsigned int ndirs;
394 int max_dirs, min_inodes;
395 ext4_grpblk_t min_clusters;
396 ext4_group_t i, grp, g, ngroups;
397 struct ext4_group_desc *desc;
398 struct orlov_stats stats;
399 int flex_size = ext4_flex_bg_size(sbi);
400 struct dx_hash_info hinfo;
401
402 ngroups = real_ngroups;
403 if (flex_size > 1) {
404 ngroups = (real_ngroups + flex_size - 1) >>
405 sbi->s_log_groups_per_flex;
406 parent_group >>= sbi->s_log_groups_per_flex;
407 }
408
409 freei = percpu_counter_read_positive(&sbi->s_freeinodes_counter);
410 avefreei = freei / ngroups;
411 freeb = EXT4_C2B(sbi,
412 percpu_counter_read_positive(&sbi->s_freeclusters_counter));
413 avefreec = freeb;
414 do_div(avefreec, ngroups);
415 ndirs = percpu_counter_read_positive(&sbi->s_dirs_counter);
416
417 if (S_ISDIR(mode) &&
418 ((parent == sb->s_root->d_inode) ||
419 (ext4_test_inode_flag(parent, EXT4_INODE_TOPDIR)))) {
420 int best_ndir = inodes_per_group;
421 int ret = -1;
422
423 if (qstr) {
424 hinfo.hash_version = DX_HASH_HALF_MD4;
425 hinfo.seed = sbi->s_hash_seed;
426 ext4fs_dirhash(qstr->name, qstr->len, &hinfo);
427 grp = hinfo.hash;
428 } else
429 get_random_bytes(&grp, sizeof(grp));
430 parent_group = (unsigned)grp % ngroups;
431 for (i = 0; i < ngroups; i++) {
432 g = (parent_group + i) % ngroups;
433 get_orlov_stats(sb, g, flex_size, &stats);
434 if (!stats.free_inodes)
435 continue;
436 if (stats.used_dirs >= best_ndir)
437 continue;
438 if (stats.free_inodes < avefreei)
439 continue;
440 if (stats.free_clusters < avefreec)
441 continue;
442 grp = g;
443 ret = 0;
444 best_ndir = stats.used_dirs;
445 }
446 if (ret)
447 goto fallback;
448 found_flex_bg:
449 if (flex_size == 1) {
450 *group = grp;
451 return 0;
452 }
453
454
455
456
457
458
459
460
461 grp *= flex_size;
462 for (i = 0; i < flex_size; i++) {
463 if (grp+i >= real_ngroups)
464 break;
465 desc = ext4_get_group_desc(sb, grp+i, NULL);
466 if (desc && ext4_free_inodes_count(sb, desc)) {
467 *group = grp+i;
468 return 0;
469 }
470 }
471 goto fallback;
472 }
473
474 max_dirs = ndirs / ngroups + inodes_per_group / 16;
475 min_inodes = avefreei - inodes_per_group*flex_size / 4;
476 if (min_inodes < 1)
477 min_inodes = 1;
478 min_clusters = avefreec - EXT4_CLUSTERS_PER_GROUP(sb)*flex_size / 4;
479
480
481
482
483
484 if (EXT4_I(parent)->i_last_alloc_group != ~0) {
485 parent_group = EXT4_I(parent)->i_last_alloc_group;
486 if (flex_size > 1)
487 parent_group >>= sbi->s_log_groups_per_flex;
488 }
489
490 for (i = 0; i < ngroups; i++) {
491 grp = (parent_group + i) % ngroups;
492 get_orlov_stats(sb, grp, flex_size, &stats);
493 if (stats.used_dirs >= max_dirs)
494 continue;
495 if (stats.free_inodes < min_inodes)
496 continue;
497 if (stats.free_clusters < min_clusters)
498 continue;
499 goto found_flex_bg;
500 }
501
502fallback:
503 ngroups = real_ngroups;
504 avefreei = freei / ngroups;
505fallback_retry:
506 parent_group = EXT4_I(parent)->i_block_group;
507 for (i = 0; i < ngroups; i++) {
508 grp = (parent_group + i) % ngroups;
509 desc = ext4_get_group_desc(sb, grp, NULL);
510 if (desc) {
511 grp_free = ext4_free_inodes_count(sb, desc);
512 if (grp_free && grp_free >= avefreei) {
513 *group = grp;
514 return 0;
515 }
516 }
517 }
518
519 if (avefreei) {
520
521
522
523
524 avefreei = 0;
525 goto fallback_retry;
526 }
527
528 return -1;
529}
530
531static int find_group_other(struct super_block *sb, struct inode *parent,
532 ext4_group_t *group, umode_t mode)
533{
534 ext4_group_t parent_group = EXT4_I(parent)->i_block_group;
535 ext4_group_t i, last, ngroups = ext4_get_groups_count(sb);
536 struct ext4_group_desc *desc;
537 int flex_size = ext4_flex_bg_size(EXT4_SB(sb));
538
539
540
541
542
543
544
545
546 if (flex_size > 1) {
547 int retry = 0;
548
549 try_again:
550 parent_group &= ~(flex_size-1);
551 last = parent_group + flex_size;
552 if (last > ngroups)
553 last = ngroups;
554 for (i = parent_group; i < last; i++) {
555 desc = ext4_get_group_desc(sb, i, NULL);
556 if (desc && ext4_free_inodes_count(sb, desc)) {
557 *group = i;
558 return 0;
559 }
560 }
561 if (!retry && EXT4_I(parent)->i_last_alloc_group != ~0) {
562 retry = 1;
563 parent_group = EXT4_I(parent)->i_last_alloc_group;
564 goto try_again;
565 }
566
567
568
569
570
571 *group = parent_group + flex_size;
572 if (*group > ngroups)
573 *group = 0;
574 return find_group_orlov(sb, parent, group, mode, NULL);
575 }
576
577
578
579
580 *group = parent_group;
581 desc = ext4_get_group_desc(sb, *group, NULL);
582 if (desc && ext4_free_inodes_count(sb, desc) &&
583 ext4_free_group_clusters(sb, desc))
584 return 0;
585
586
587
588
589
590
591
592
593
594
595 *group = (*group + parent->i_ino) % ngroups;
596
597
598
599
600
601 for (i = 1; i < ngroups; i <<= 1) {
602 *group += i;
603 if (*group >= ngroups)
604 *group -= ngroups;
605 desc = ext4_get_group_desc(sb, *group, NULL);
606 if (desc && ext4_free_inodes_count(sb, desc) &&
607 ext4_free_group_clusters(sb, desc))
608 return 0;
609 }
610
611
612
613
614
615 *group = parent_group;
616 for (i = 0; i < ngroups; i++) {
617 if (++*group >= ngroups)
618 *group = 0;
619 desc = ext4_get_group_desc(sb, *group, NULL);
620 if (desc && ext4_free_inodes_count(sb, desc))
621 return 0;
622 }
623
624 return -1;
625}
626
627
628
629
630
631
632
633
634
635
636
637struct inode *__ext4_new_inode(handle_t *handle, struct inode *dir,
638 umode_t mode, const struct qstr *qstr,
639 __u32 goal, uid_t *owner, int handle_type,
640 unsigned int line_no, int nblocks)
641{
642 struct super_block *sb;
643 struct buffer_head *inode_bitmap_bh = NULL;
644 struct buffer_head *group_desc_bh;
645 ext4_group_t ngroups, group = 0;
646 unsigned long ino = 0;
647 struct inode *inode;
648 struct ext4_group_desc *gdp = NULL;
649 struct ext4_inode_info *ei;
650 struct ext4_sb_info *sbi;
651 int ret2, err = 0;
652 struct inode *ret;
653 ext4_group_t i;
654 ext4_group_t flex_group;
655
656
657 if (!dir || !dir->i_nlink)
658 return ERR_PTR(-EPERM);
659
660 sb = dir->i_sb;
661 ngroups = ext4_get_groups_count(sb);
662 trace_ext4_request_inode(dir, mode);
663 inode = new_inode(sb);
664 if (!inode)
665 return ERR_PTR(-ENOMEM);
666 ei = EXT4_I(inode);
667 sbi = EXT4_SB(sb);
668
669
670
671
672
673
674 if (owner) {
675 inode->i_mode = mode;
676 i_uid_write(inode, owner[0]);
677 i_gid_write(inode, owner[1]);
678 } else if (test_opt(sb, GRPID)) {
679 inode->i_mode = mode;
680 inode->i_uid = current_fsuid();
681 inode->i_gid = dir->i_gid;
682 } else
683 inode_init_owner(inode, dir, mode);
684 dquot_initialize(inode);
685
686 if (!goal)
687 goal = sbi->s_inode_goal;
688
689 if (goal && goal <= le32_to_cpu(sbi->s_es->s_inodes_count)) {
690 group = (goal - 1) / EXT4_INODES_PER_GROUP(sb);
691 ino = (goal - 1) % EXT4_INODES_PER_GROUP(sb);
692 ret2 = 0;
693 goto got_group;
694 }
695
696 if (S_ISDIR(mode))
697 ret2 = find_group_orlov(sb, dir, &group, mode, qstr);
698 else
699 ret2 = find_group_other(sb, dir, &group, mode);
700
701got_group:
702 EXT4_I(dir)->i_last_alloc_group = group;
703 err = -ENOSPC;
704 if (ret2 == -1)
705 goto out;
706
707
708
709
710
711
712 for (i = 0; i < ngroups; i++, ino = 0) {
713 err = -EIO;
714
715 gdp = ext4_get_group_desc(sb, group, &group_desc_bh);
716 if (!gdp)
717 goto out;
718
719
720
721
722 if (ext4_free_inodes_count(sb, gdp) == 0) {
723 if (++group == ngroups)
724 group = 0;
725 continue;
726 }
727
728 brelse(inode_bitmap_bh);
729 inode_bitmap_bh = ext4_read_inode_bitmap(sb, group);
730 if (!inode_bitmap_bh)
731 goto out;
732
733repeat_in_this_group:
734 ino = ext4_find_next_zero_bit((unsigned long *)
735 inode_bitmap_bh->b_data,
736 EXT4_INODES_PER_GROUP(sb), ino);
737 if (ino >= EXT4_INODES_PER_GROUP(sb)) {
738 if (++group == ngroups)
739 group = 0;
740 continue;
741 }
742 if (group == 0 && (ino+1) < EXT4_FIRST_INO(sb)) {
743 ext4_error(sb, "reserved inode found cleared - "
744 "inode=%lu", ino + 1);
745 continue;
746 }
747 if (!handle) {
748 BUG_ON(nblocks <= 0);
749 handle = __ext4_journal_start_sb(dir->i_sb, line_no,
750 handle_type, nblocks);
751 if (IS_ERR(handle)) {
752 err = PTR_ERR(handle);
753 ext4_std_error(sb, err);
754 goto out;
755 }
756 }
757 BUFFER_TRACE(inode_bitmap_bh, "get_write_access");
758 err = ext4_journal_get_write_access(handle, inode_bitmap_bh);
759 if (err) {
760 ext4_std_error(sb, err);
761 goto out;
762 }
763 ext4_lock_group(sb, group);
764 ret2 = ext4_test_and_set_bit(ino, inode_bitmap_bh->b_data);
765 ext4_unlock_group(sb, group);
766 ino++;
767 if (!ret2)
768 goto got;
769 if (ino < EXT4_INODES_PER_GROUP(sb))
770 goto repeat_in_this_group;
771 }
772 err = -ENOSPC;
773 goto out;
774
775got:
776 BUFFER_TRACE(inode_bitmap_bh, "call ext4_handle_dirty_metadata");
777 err = ext4_handle_dirty_metadata(handle, NULL, inode_bitmap_bh);
778 if (err) {
779 ext4_std_error(sb, err);
780 goto out;
781 }
782
783
784 if (ext4_has_group_desc_csum(sb) &&
785 gdp->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) {
786 struct buffer_head *block_bitmap_bh;
787
788 block_bitmap_bh = ext4_read_block_bitmap(sb, group);
789 BUFFER_TRACE(block_bitmap_bh, "get block bitmap access");
790 err = ext4_journal_get_write_access(handle, block_bitmap_bh);
791 if (err) {
792 brelse(block_bitmap_bh);
793 ext4_std_error(sb, err);
794 goto out;
795 }
796
797 BUFFER_TRACE(block_bitmap_bh, "dirty block bitmap");
798 err = ext4_handle_dirty_metadata(handle, NULL, block_bitmap_bh);
799
800
801 ext4_lock_group(sb, group);
802 if (gdp->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) {
803 gdp->bg_flags &= cpu_to_le16(~EXT4_BG_BLOCK_UNINIT);
804 ext4_free_group_clusters_set(sb, gdp,
805 ext4_free_clusters_after_init(sb, group, gdp));
806 ext4_block_bitmap_csum_set(sb, group, gdp,
807 block_bitmap_bh);
808 ext4_group_desc_csum_set(sb, group, gdp);
809 }
810 ext4_unlock_group(sb, group);
811 brelse(block_bitmap_bh);
812
813 if (err) {
814 ext4_std_error(sb, err);
815 goto out;
816 }
817 }
818
819 BUFFER_TRACE(group_desc_bh, "get_write_access");
820 err = ext4_journal_get_write_access(handle, group_desc_bh);
821 if (err) {
822 ext4_std_error(sb, err);
823 goto out;
824 }
825
826
827 if (ext4_has_group_desc_csum(sb)) {
828 int free;
829 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
830
831 down_read(&grp->alloc_sem);
832 ext4_lock_group(sb, group);
833 free = EXT4_INODES_PER_GROUP(sb) -
834 ext4_itable_unused_count(sb, gdp);
835 if (gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_UNINIT)) {
836 gdp->bg_flags &= cpu_to_le16(~EXT4_BG_INODE_UNINIT);
837 free = 0;
838 }
839
840
841
842
843
844 if (ino > free)
845 ext4_itable_unused_set(sb, gdp,
846 (EXT4_INODES_PER_GROUP(sb) - ino));
847 up_read(&grp->alloc_sem);
848 } else {
849 ext4_lock_group(sb, group);
850 }
851
852 ext4_free_inodes_set(sb, gdp, ext4_free_inodes_count(sb, gdp) - 1);
853 if (S_ISDIR(mode)) {
854 ext4_used_dirs_set(sb, gdp, ext4_used_dirs_count(sb, gdp) + 1);
855 if (sbi->s_log_groups_per_flex) {
856 ext4_group_t f = ext4_flex_group(sbi, group);
857
858 atomic_inc(&sbi->s_flex_groups[f].used_dirs);
859 }
860 }
861 if (ext4_has_group_desc_csum(sb)) {
862 ext4_inode_bitmap_csum_set(sb, group, gdp, inode_bitmap_bh,
863 EXT4_INODES_PER_GROUP(sb) / 8);
864 ext4_group_desc_csum_set(sb, group, gdp);
865 }
866 ext4_unlock_group(sb, group);
867
868 BUFFER_TRACE(group_desc_bh, "call ext4_handle_dirty_metadata");
869 err = ext4_handle_dirty_metadata(handle, NULL, group_desc_bh);
870 if (err) {
871 ext4_std_error(sb, err);
872 goto out;
873 }
874
875 percpu_counter_dec(&sbi->s_freeinodes_counter);
876 if (S_ISDIR(mode))
877 percpu_counter_inc(&sbi->s_dirs_counter);
878
879 if (sbi->s_log_groups_per_flex) {
880 flex_group = ext4_flex_group(sbi, group);
881 atomic_dec(&sbi->s_flex_groups[flex_group].free_inodes);
882 }
883
884 inode->i_ino = ino + group * EXT4_INODES_PER_GROUP(sb);
885
886 inode->i_blocks = 0;
887 inode->i_mtime = inode->i_atime = inode->i_ctime = ei->i_crtime =
888 ext4_current_time(inode);
889
890 memset(ei->i_data, 0, sizeof(ei->i_data));
891 ei->i_dir_start_lookup = 0;
892 ei->i_disksize = 0;
893
894
895 ei->i_flags =
896 ext4_mask_flags(mode, EXT4_I(dir)->i_flags & EXT4_FL_INHERITED);
897 ei->i_file_acl = 0;
898 ei->i_dtime = 0;
899 ei->i_block_group = group;
900 ei->i_last_alloc_group = ~0;
901
902 ext4_set_inode_flags(inode);
903 if (IS_DIRSYNC(inode))
904 ext4_handle_sync(handle);
905 if (insert_inode_locked(inode) < 0) {
906
907
908
909
910 err = -EIO;
911 ext4_error(sb, "failed to insert inode %lu: doubly allocated?",
912 inode->i_ino);
913 goto out;
914 }
915 spin_lock(&sbi->s_next_gen_lock);
916 inode->i_generation = sbi->s_next_generation++;
917 spin_unlock(&sbi->s_next_gen_lock);
918
919
920 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
921 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
922 __u32 csum;
923 __le32 inum = cpu_to_le32(inode->i_ino);
924 __le32 gen = cpu_to_le32(inode->i_generation);
925 csum = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&inum,
926 sizeof(inum));
927 ei->i_csum_seed = ext4_chksum(sbi, csum, (__u8 *)&gen,
928 sizeof(gen));
929 }
930
931 ext4_clear_state_flags(ei);
932 ext4_set_inode_state(inode, EXT4_STATE_NEW);
933
934 ei->i_extra_isize = EXT4_SB(sb)->s_want_extra_isize;
935
936 ei->i_inline_off = 0;
937 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_INLINE_DATA))
938 ext4_set_inode_state(inode, EXT4_STATE_MAY_INLINE_DATA);
939
940 ret = inode;
941 err = dquot_alloc_inode(inode);
942 if (err)
943 goto fail_drop;
944
945 err = ext4_init_acl(handle, inode, dir);
946 if (err)
947 goto fail_free_drop;
948
949 err = ext4_init_security(handle, inode, dir, qstr);
950 if (err)
951 goto fail_free_drop;
952
953 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
954
955 if (S_ISDIR(mode) || S_ISREG(mode) || S_ISLNK(mode)) {
956 ext4_set_inode_flag(inode, EXT4_INODE_EXTENTS);
957 ext4_ext_tree_init(handle, inode);
958 }
959 }
960
961 if (ext4_handle_valid(handle)) {
962 ei->i_sync_tid = handle->h_transaction->t_tid;
963 ei->i_datasync_tid = handle->h_transaction->t_tid;
964 }
965
966 err = ext4_mark_inode_dirty(handle, inode);
967 if (err) {
968 ext4_std_error(sb, err);
969 goto fail_free_drop;
970 }
971
972 ext4_debug("allocating inode %lu\n", inode->i_ino);
973 trace_ext4_allocate_inode(inode, dir, mode);
974 brelse(inode_bitmap_bh);
975 return ret;
976
977fail_free_drop:
978 dquot_free_inode(inode);
979fail_drop:
980 clear_nlink(inode);
981 unlock_new_inode(inode);
982out:
983 dquot_drop(inode);
984 inode->i_flags |= S_NOQUOTA;
985 iput(inode);
986 brelse(inode_bitmap_bh);
987 return ERR_PTR(err);
988}
989
990
991struct inode *ext4_orphan_get(struct super_block *sb, unsigned long ino)
992{
993 unsigned long max_ino = le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count);
994 ext4_group_t block_group;
995 int bit;
996 struct buffer_head *bitmap_bh;
997 struct inode *inode = NULL;
998 long err = -EIO;
999
1000
1001 if (ino > max_ino) {
1002 ext4_warning(sb, "bad orphan ino %lu! e2fsck was run?", ino);
1003 goto error;
1004 }
1005
1006 block_group = (ino - 1) / EXT4_INODES_PER_GROUP(sb);
1007 bit = (ino - 1) % EXT4_INODES_PER_GROUP(sb);
1008 bitmap_bh = ext4_read_inode_bitmap(sb, block_group);
1009 if (!bitmap_bh) {
1010 ext4_warning(sb, "inode bitmap error for orphan %lu", ino);
1011 goto error;
1012 }
1013
1014
1015
1016
1017
1018 if (!ext4_test_bit(bit, bitmap_bh->b_data))
1019 goto bad_orphan;
1020
1021 inode = ext4_iget(sb, ino);
1022 if (IS_ERR(inode))
1023 goto iget_failed;
1024
1025
1026
1027
1028
1029
1030 if (inode->i_nlink && !ext4_can_truncate(inode))
1031 goto bad_orphan;
1032
1033 if (NEXT_ORPHAN(inode) > max_ino)
1034 goto bad_orphan;
1035 brelse(bitmap_bh);
1036 return inode;
1037
1038iget_failed:
1039 err = PTR_ERR(inode);
1040 inode = NULL;
1041bad_orphan:
1042 ext4_warning(sb, "bad orphan inode %lu! e2fsck was run?", ino);
1043 printk(KERN_WARNING "ext4_test_bit(bit=%d, block=%llu) = %d\n",
1044 bit, (unsigned long long)bitmap_bh->b_blocknr,
1045 ext4_test_bit(bit, bitmap_bh->b_data));
1046 printk(KERN_WARNING "inode=%p\n", inode);
1047 if (inode) {
1048 printk(KERN_WARNING "is_bad_inode(inode)=%d\n",
1049 is_bad_inode(inode));
1050 printk(KERN_WARNING "NEXT_ORPHAN(inode)=%u\n",
1051 NEXT_ORPHAN(inode));
1052 printk(KERN_WARNING "max_ino=%lu\n", max_ino);
1053 printk(KERN_WARNING "i_nlink=%u\n", inode->i_nlink);
1054
1055 if (inode->i_nlink == 0)
1056 inode->i_blocks = 0;
1057 iput(inode);
1058 }
1059 brelse(bitmap_bh);
1060error:
1061 return ERR_PTR(err);
1062}
1063
1064unsigned long ext4_count_free_inodes(struct super_block *sb)
1065{
1066 unsigned long desc_count;
1067 struct ext4_group_desc *gdp;
1068 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
1069#ifdef EXT4FS_DEBUG
1070 struct ext4_super_block *es;
1071 unsigned long bitmap_count, x;
1072 struct buffer_head *bitmap_bh = NULL;
1073
1074 es = EXT4_SB(sb)->s_es;
1075 desc_count = 0;
1076 bitmap_count = 0;
1077 gdp = NULL;
1078 for (i = 0; i < ngroups; i++) {
1079 gdp = ext4_get_group_desc(sb, i, NULL);
1080 if (!gdp)
1081 continue;
1082 desc_count += ext4_free_inodes_count(sb, gdp);
1083 brelse(bitmap_bh);
1084 bitmap_bh = ext4_read_inode_bitmap(sb, i);
1085 if (!bitmap_bh)
1086 continue;
1087
1088 x = ext4_count_free(bitmap_bh->b_data,
1089 EXT4_INODES_PER_GROUP(sb) / 8);
1090 printk(KERN_DEBUG "group %lu: stored = %d, counted = %lu\n",
1091 (unsigned long) i, ext4_free_inodes_count(sb, gdp), x);
1092 bitmap_count += x;
1093 }
1094 brelse(bitmap_bh);
1095 printk(KERN_DEBUG "ext4_count_free_inodes: "
1096 "stored = %u, computed = %lu, %lu\n",
1097 le32_to_cpu(es->s_free_inodes_count), desc_count, bitmap_count);
1098 return desc_count;
1099#else
1100 desc_count = 0;
1101 for (i = 0; i < ngroups; i++) {
1102 gdp = ext4_get_group_desc(sb, i, NULL);
1103 if (!gdp)
1104 continue;
1105 desc_count += ext4_free_inodes_count(sb, gdp);
1106 cond_resched();
1107 }
1108 return desc_count;
1109#endif
1110}
1111
1112
1113unsigned long ext4_count_dirs(struct super_block * sb)
1114{
1115 unsigned long count = 0;
1116 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
1117
1118 for (i = 0; i < ngroups; i++) {
1119 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1120 if (!gdp)
1121 continue;
1122 count += ext4_used_dirs_count(sb, gdp);
1123 }
1124 return count;
1125}
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135int ext4_init_inode_table(struct super_block *sb, ext4_group_t group,
1136 int barrier)
1137{
1138 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
1139 struct ext4_sb_info *sbi = EXT4_SB(sb);
1140 struct ext4_group_desc *gdp = NULL;
1141 struct buffer_head *group_desc_bh;
1142 handle_t *handle;
1143 ext4_fsblk_t blk;
1144 int num, ret = 0, used_blks = 0;
1145
1146
1147 if (sb->s_flags & MS_RDONLY) {
1148 ret = 1;
1149 goto out;
1150 }
1151
1152 gdp = ext4_get_group_desc(sb, group, &group_desc_bh);
1153 if (!gdp)
1154 goto out;
1155
1156
1157
1158
1159
1160 if (gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))
1161 goto out;
1162
1163 handle = ext4_journal_start_sb(sb, EXT4_HT_MISC, 1);
1164 if (IS_ERR(handle)) {
1165 ret = PTR_ERR(handle);
1166 goto out;
1167 }
1168
1169 down_write(&grp->alloc_sem);
1170
1171
1172
1173
1174
1175 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_UNINIT)))
1176 used_blks = DIV_ROUND_UP((EXT4_INODES_PER_GROUP(sb) -
1177 ext4_itable_unused_count(sb, gdp)),
1178 sbi->s_inodes_per_block);
1179
1180 if ((used_blks < 0) || (used_blks > sbi->s_itb_per_group)) {
1181 ext4_error(sb, "Something is wrong with group %u: "
1182 "used itable blocks: %d; "
1183 "itable unused count: %u",
1184 group, used_blks,
1185 ext4_itable_unused_count(sb, gdp));
1186 ret = 1;
1187 goto err_out;
1188 }
1189
1190 blk = ext4_inode_table(sb, gdp) + used_blks;
1191 num = sbi->s_itb_per_group - used_blks;
1192
1193 BUFFER_TRACE(group_desc_bh, "get_write_access");
1194 ret = ext4_journal_get_write_access(handle,
1195 group_desc_bh);
1196 if (ret)
1197 goto err_out;
1198
1199
1200
1201
1202
1203
1204 if (unlikely(num == 0))
1205 goto skip_zeroout;
1206
1207 ext4_debug("going to zero out inode table in group %d\n",
1208 group);
1209 ret = sb_issue_zeroout(sb, blk, num, GFP_NOFS);
1210 if (ret < 0)
1211 goto err_out;
1212 if (barrier)
1213 blkdev_issue_flush(sb->s_bdev, GFP_NOFS, NULL);
1214
1215skip_zeroout:
1216 ext4_lock_group(sb, group);
1217 gdp->bg_flags |= cpu_to_le16(EXT4_BG_INODE_ZEROED);
1218 ext4_group_desc_csum_set(sb, group, gdp);
1219 ext4_unlock_group(sb, group);
1220
1221 BUFFER_TRACE(group_desc_bh,
1222 "call ext4_handle_dirty_metadata");
1223 ret = ext4_handle_dirty_metadata(handle, NULL,
1224 group_desc_bh);
1225
1226err_out:
1227 up_write(&grp->alloc_sem);
1228 ext4_journal_stop(handle);
1229out:
1230 return ret;
1231}
1232