1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
|
/*
* fs/logfs/dir.c - directory-related code
*
* As should be obvious for Linux kernel code, license is GPLv2
*
* Copyright (c) 2005-2008 Joern Engel <joern@logfs.org>
*/
#include "logfs.h"
#include <linux/slab.h>
/*
* Atomic dir operations
*
* Directory operations are by default not atomic. Dentries and Inodes are
* created/removed/altered in separate operations. Therefore we need to do
* a small amount of journaling.
*
* Create, link, mkdir, mknod and symlink all share the same function to do
* the work: __logfs_create. This function works in two atomic steps:
* 1. allocate inode (remember in journal)
* 2. allocate dentry (clear journal)
*
* As we can only get interrupted between the two, when the inode we just
* created is simply stored in the anchor. On next mount, if we were
* interrupted, we delete the inode. From a users point of view the
* operation never happened.
*
* Unlink and rmdir also share the same function: unlink. Again, this
* function works in two atomic steps
* 1. remove dentry (remember inode in journal)
* 2. unlink inode (clear journal)
*
* And again, on the next mount, if we were interrupted, we delete the inode.
* From a users point of view the operation succeeded.
*
* Rename is the real pain to deal with, harder than all the other methods
* combined. Depending on the circumstances we can run into three cases.
* A "target rename" where the target dentry already existed, a "local
* rename" where both parent directories are identical or a "cross-directory
* rename" in the remaining case.
*
* Local rename is atomic, as the old dentry is simply rewritten with a new
* name.
*
* Cross-directory rename works in two steps, similar to __logfs_create and
* logfs_unlink:
* 1. Write new dentry (remember old dentry in journal)
* 2. Remove old dentry (clear journal)
*
* Here we remember a dentry instead of an inode. On next mount, if we were
* interrupted, we delete the dentry. From a users point of view, the
* operation succeeded.
*
* Target rename works in three atomic steps:
* 1. Attach old inode to new dentry (remember old dentry and new inode)
* 2. Remove old dentry (still remember the new inode)
* 3. Remove victim inode
*
* Here we remember both an inode an a dentry. If we get interrupted
* between steps 1 and 2, we delete both the dentry and the inode. If
* we get interrupted between steps 2 and 3, we delete just the inode.
* In either case, the remaining objects are deleted on next mount. From
* a users point of view, the operation succeeded.
*/
static int write_dir(struct inode *dir, struct logfs_disk_dentry *dd,
loff_t pos)
{
return logfs_inode_write(dir, dd, sizeof(*dd), pos, WF_LOCK, NULL);
}
static int write_inode(struct inode *inode)
{
return __logfs_write_inode(inode, NULL, WF_LOCK);
}
static s64 dir_seek_data(struct inode *inode, s64 pos)
{
s64 new_pos = logfs_seek_data(inode, pos);
return max(pos, new_pos - 1);
}
static int beyond_eof(struct inode *inode, loff_t bix)
{
loff_t pos = bix << inode->i_sb->s_blocksize_bits;
return pos >= i_size_read(inode);
}
/*
* Prime value was chosen to be roughly 256 + 26. r5 hash uses 11,
* so short names (len <= 9) don't even occupy the complete 32bit name
* space. A prime >256 ensures short names quickly spread the 32bit
* name space. Add about 26 for the estimated amount of information
* of each character and pick a prime nearby, preferably a bit-sparse
* one.
*/
static u32 logfs_hash_32(const char *s, int len, u32 seed)
{
u32 hash = seed;
int i;
for (i = 0; i < len; i++)
hash = hash * 293 + s[i];
return hash;
}
/*
* We have to satisfy several conflicting requirements here. Small
* directories should stay fairly compact and not require too many
* indirect blocks. The number of possible locations for a given hash
* should be small to make lookup() fast. And we should try hard not
* to overflow the 32bit name space or nfs and 32bit host systems will
* be unhappy.
*
* So we use the following scheme. First we reduce the hash to 0..15
* and try a direct block. If that is occupied we reduce the hash to
* 16..255 and try an indirect block. Same for 2x and 3x indirect
* blocks. Lastly we reduce the hash to 0x800_0000 .. 0xffff_ffff,
* but use buckets containing eight entries instead of a single one.
*
* Using 16 entries should allow for a reasonable amount of hash
* collisions, so the 32bit name space can be packed fairly tight
* before overflowing. Oh and currently we don't overflow but return
* and error.
*
* How likely are collisions? Doing the appropriate math is beyond me
* and the Bronstein textbook. But running a test program to brute
* force collisions for a couple of days showed that on average the
* first collision occurs after 598M entries, with 290M being the
* smallest result. Obviously 21 entries could already cause a
* collision if all entries are carefully chosen.
*/
static pgoff_t hash_index(u32 hash, int round)
{
u32 i0_blocks = I0_BLOCKS;
u32 i1_blocks = I1_BLOCKS;
u32 i2_blocks = I2_BLOCKS;
u32 i3_blocks = I3_BLOCKS;
switch (round) {
case 0:
return hash % i0_blocks;
case 1:
return i0_blocks + hash % (i1_blocks - i0_blocks);
case 2:
return i1_blocks + hash % (i2_blocks - i1_blocks);
case 3:
return i2_blocks + hash % (i3_blocks - i2_blocks);
case 4 ... 19:
return i3_blocks + 16 * (hash % (((1<<31) - i3_blocks) / 16))
+ round - 4;
}
BUG();
}
static struct page *logfs_get_dd_page(struct inode *dir, struct dentry *dentry)
{
struct qstr *name = &dentry->d_name;
struct page *page;
struct logfs_disk_dentry *dd;
u32 hash = logfs_hash_32(name->name, name->len, 0);
pgoff_t index;
int round;
if (name->len > LOGFS_MAX_NAMELEN)
return ERR_PTR(-ENAMETOOLONG);
for (round = 0; round < 20; round++) {
index = hash_index(hash, round);
if (beyond_eof(dir, index))
return NULL;
if (!logfs_exist_block(dir, index))
continue;
page = read_cache_page(dir->i_mapping, index,
(filler_t *)logfs_readpage, NULL);
if (IS_ERR(page))
return page;
dd = kmap_atomic(page);
BUG_ON(dd->namelen == 0);
if (name->len != be16_to_cpu(dd->namelen) ||
memcmp(name->name, dd->name, name->len)) {
kunmap_atomic(dd);
put_page(page);
continue;
}
kunmap_atomic(dd);
return page;
}
return NULL;
}
static int logfs_remove_inode(struct inode *inode)
{
int ret;
drop_nlink(inode);
ret = write_inode(inode);
LOGFS_BUG_ON(ret, inode->i_sb);
return ret;
}
static void abort_transaction(struct inode *inode, struct logfs_transaction *ta)
{
if (logfs_inode(inode)->li_block)
logfs_inode(inode)->li_block->ta = NULL;
kfree(ta);
}
static int logfs_unlink(struct inode *dir, struct dentry *dentry)
{
struct logfs_super *super = logfs_super(dir->i_sb
|