forked from markfasheh/duperemove
-
Notifications
You must be signed in to change notification settings - Fork 0
/
run_dedupe.c
996 lines (866 loc) · 24.4 KB
/
run_dedupe.c
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
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
/*
* run_dedupe.c
*
* Implements dedupe of duplicate extents from our results tree
*
* Copyright (C) 2014 SUSE. All rights reserved.
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public
* License version 2 as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* General Public License for more details.
*
* Authors: Mark Fasheh <[email protected]>
*/
#include <unistd.h>
#include <stdio.h>
#include <stdlib.h>
#include <stdint.h>
#include <stddef.h>
#include <errno.h>
#include <string.h>
#include <inttypes.h>
#include <glib.h>
#include "rbtree.h"
#include "list.h"
#include "csum.h"
#include "filerec.h"
#include "hash-tree.h"
#include "results-tree.h"
#include "dedupe.h"
#include "util.h"
#include "memstats.h"
#include "debug.h"
#include "run_dedupe.h"
extern int block_dedupe;
extern int dedupe_same_file;
extern int fiemap_during_dedupe;
static GMutex mutex;
static GMutex console_mutex;
static struct results_tree *results_tree;
static volatile unsigned long long total_dedupe_passes;
static volatile unsigned long long curr_dedupe_pass;
static unsigned int leading_spaces;
void print_dupes_table(struct results_tree *res)
{
struct rb_root *root = &res->root;
struct rb_node *node = rb_first(root);
struct dupe_extents *dext;
struct extent *extent;
printf("Simple read and compare of file data found %u instances of "
"extents that might benefit from deduplication.\n",
res->num_dupes);
if (res->num_dupes == 0)
return;
while (1) {
if (node == NULL)
break;
dext = rb_entry(node, struct dupe_extents, de_node);
printf("Showing %u identical extents of length %s with id ",
dext->de_num_dupes, pretty_size(dext->de_len));
debug_print_digest_short(stdout, dext->de_hash);
printf("\n");
printf("Start\t\tFilename\n");
list_for_each_entry(extent, &dext->de_extents, e_list) {
printf("%s\t\"%s\"\n",
pretty_size(extent->e_loff),
extent->e_file->filename);
}
node = rb_next(node);
}
}
static void process_dedupe_results(struct dedupe_ctxt *ctxt,
uint64_t *kern_bytes)
{
int done = 0;
int target_status;
uint64_t target_loff, target_bytes;
struct filerec *f;
const char *status_str = "[unknown status]";
while (!done) {
done = pop_one_dedupe_result(ctxt, &target_status, &target_loff,
&target_bytes, &f);
if (kern_bytes)
*kern_bytes += target_bytes;
/*
* Only print in case of error.
*
* Kernels older than 4.2 can't handle the target and
* dedupe files being the same and -EINVAL in that
* case. Don't bubble it up so as to avoid user
* confusion.
*/
if (target_status == 0 ||
(target_status == -EINVAL && f == ctxt->ioctl_file))
continue;
if (target_status == BTRFS_SAME_DATA_DIFFERS)
status_str = "data changed";
else if (target_status < 0)
status_str = strerror(-target_status);
printf("[%p] Dedupe for file \"%s\" had status (%d) "
"\"%s\".\n",
g_thread_self(), f->filename, target_status,
status_str);
}
}
static void get_extent_info(struct dupe_extents *dext)
{
int ret = 0;
struct extent *extent;
struct filerec *file;
list_for_each_entry(extent, &dext->de_extents, e_list) {
file = extent->e_file;
if (filerec_open(file, 0))
continue;
extent_shared_bytes(extent) = 0;
ret = filerec_count_shared(file, extent->e_loff, dext->de_len,
&extent_shared_bytes(extent),
&extent_poff(extent),
&extent_plen(extent));
if (ret) {
fprintf(stderr, "%s: fiemap error %d: %s\n",
extent->e_file->filename, ret, strerror(ret));
}
filerec_close(file);
}
}
static void add_shared_extents(struct dupe_extents *dext, uint64_t *shared)
{
struct extent *extent;
list_for_each_entry(extent, &dext->de_extents, e_list)
*shared += extent_shared_bytes(extent);
}
static int disk_extent_grew(struct dupe_extents *dext, struct extent *extent)
{
/*
* Check length of the virtual extent versus that of the 1st
* physical extent in our range.
*
* If the physical extent is smaller than our virtual
* (duplicate) extent, we want to go ahead and dedupe in order
* to catch two cases:
*
* - The files were appended to (separately) with duplicate
* data - this will result in a pair of new extents on each
* file that can be deduped.
*
* - Kernels before 4.2 rejected unaligned lengths, so we can
* have a residual tail extent to dedupe.
*/
if (extent_plen(extent) < dext->de_len)
return 1;
return 0;
}
/*
* Removes extents which it believes have already been deduped. We err
* on the side of more deduping here.
*/
static void clean_deduped(struct dupe_extents **ret_dext)
{
int left;
int extents_kept = 0;
int first = 1;
struct dupe_extents *dext = *ret_dext;
struct rb_node *inner, *outer;
struct extent *inner_extent, *outer_extent;
if (!dext || dext->de_num_dupes == 0)
return;
if (block_dedupe)
return;
outer = rb_first(&dext->de_extents_root);
while (outer) {
outer_extent = rb_entry(outer, struct extent, e_node);
/*
* First extent will not be considered for removal
* below, which is fine as remove_extent() handles the
* case of only 1 extent left on the dext for us.
*
* Replicate the checks though and count it as kept if
* we don't want it deleted. That will trigger the
* logic below to save the dext if we should wind up
* throwing everything else out.
*/
if (first &&
(extent_poff(outer_extent) == 0 ||
disk_extent_grew(dext, outer_extent)))
extents_kept++;
first = 0;
inner = rb_next(outer);
while (inner) {
inner_extent = rb_entry(inner, struct extent, e_node);
inner = rb_next(inner);
/*
* Track if any extents have survived the
* culling. If we're down to the last two and
* at least one of them was deemed worthy,
* exit here so that he may be deduped.
*/
if (dext->de_num_dupes == 2 && extents_kept)
return;
/*
* e_poff could be zero if fiemap from
* add_shared_extents fails. In that case,
* skip the extent (it might want to be
* deduped).
*/
if (extent_poff(inner_extent)
&& extent_poff(outer_extent) == extent_poff(inner_extent)
&& !disk_extent_grew(dext, inner_extent)) {
dprintf("Remove extent "
"(\"%s\", %"PRIu64", %"PRIu64")\n",
inner_extent->e_file->filename,
extent_poff(inner_extent),
extent_plen(inner_extent));
g_mutex_lock(&mutex);
left = remove_extent(results_tree,
inner_extent);
g_mutex_unlock(&mutex);
if (left == 0) {
*ret_dext = dext = NULL;
return;
}
} else
extents_kept++;
}
outer = rb_next(outer);
}
}
#define DEDUPE_EXTENTS_CLEANED (-1)
static int dedupe_extent_list(struct dupe_extents *dext, uint64_t *fiemap_bytes,
uint64_t *kern_bytes, unsigned long long passno)
{
int ret = 0;
int last = 0;
int rc;
uint64_t shared_prev, shared_post;
struct extent *extent;
struct dedupe_ctxt *ctxt = NULL;
uint64_t len = dext->de_len;
OPEN_ONCE(open_files);
struct extent *tgt_extent = NULL;
abort_on(dext->de_num_dupes < 2);
/* Dedupe extents with id %s*/
g_mutex_lock(&console_mutex);
printf("[%p] (%0*llu/%llu) Try to dedupe extents with id ",
g_thread_self(), leading_spaces, passno, total_dedupe_passes);
debug_print_digest_short(stdout, dext->de_hash);
printf("\n");
g_mutex_unlock(&console_mutex);
shared_prev = shared_post = 0ULL;
/*
* Remove any extents which have already been deduped. This
* will free dext for us if the number of available extents
* goes below 2. If that happens, we return a special value so
* the caller knows not to reference dext any more.
*/
if (fiemap_during_dedupe) {
ret = init_all_extent_dedupe_info(dext);
if (ret)
goto out;
get_extent_info(dext);
clean_deduped(&dext);
if (!dext) {
printf("[%p] Skipping - extents are already deduped.\n",
g_thread_self());
return DEDUPE_EXTENTS_CLEANED;
}
/*
* Do this after clean_deduped as we may have removed some
* extents.
*/
add_shared_extents(dext, &shared_prev);
}
list_for_each_entry(extent, &dext->de_extents, e_list) {
if (list_is_last(&extent->e_list, &dext->de_extents))
last = 1;
ret = filerec_open_once(extent->e_file, target_rw, &open_files);
if (ret) {
fprintf(stderr, "%s: Skipping dedupe.\n",
extent->e_file->filename);
/*
* If this was our last duplicate extent in
* the list, and we added dupes from a
* previous iteration of the loop we need to
* run dedupe before exiting.
*/
if (ctxt && last)
goto run_dedupe;
continue;
}
vprintf("[%p] Add extent for file \"%s\" at offset %s (%d)\n",
g_thread_self(), extent->e_file->filename,
pretty_size(extent->e_loff), extent->e_file->fd);
if (ctxt == NULL) {
if (tgt_extent == NULL) {
/*
* We had some errors adding files
* previously and are down to the last
* dedupe candidate. Proceed only if
* we can guarantee two extents for
* dedupe (target, and this file).
*/
if (last)
goto close_files;
tgt_extent = extent;
}
ctxt = new_dedupe_ctxt(dext->de_num_dupes,
tgt_extent->e_loff, len,
tgt_extent->e_file);
if (ctxt == NULL) {
fprintf(stderr, "Out of memory while "
"allocating dedupe context.\n");
ret = ENOMEM;
goto out;
}
/*
* If we just picked the target, it got added
* with the new context. Otherwise fall
* through to let other extents onto the
* dedupe ctxt.
*/
if (tgt_extent == extent)
continue;
}
rc = add_extent_to_dedupe(ctxt, extent->e_loff, extent->e_file);
if (rc) {
if (rc < 0) {
/* This can only be ENOMEM. */
fprintf(stderr, "%s: Request not queued.\n",
extent->e_file->filename);
ret = ENOMEM;
goto out;
}
if (!last)
continue;
}
run_dedupe:
/*
* We can get here with only the target extent (0
* queued) for many reasons. Skip the dedupe in that
* case but always do cleanup.
*/
if (ctxt->num_queued) {
g_mutex_lock(&console_mutex);
printf("[%p] Dedupe %u extents (id: ", g_thread_self(),
ctxt->num_queued);
debug_print_digest_short(stdout, dext->de_hash);
printf(") with target: (%s, %s), "
"\"%s\"\n",
pretty_size(ctxt->orig_file_off),
pretty_size(ctxt->orig_len),
ctxt->ioctl_file->filename);
g_mutex_unlock(&console_mutex);
ret = dedupe_extents(ctxt);
if (ret == 0) {
process_dedupe_results(ctxt, kern_bytes);
} else {
ret = errno;
fprintf(stderr,
"FAILURE: Dedupe ioctl returns %d: %s\n",
ret, strerror(ret));
}
}
close_files:
filerec_close_open_list(&open_files);
free_dedupe_ctxt(ctxt);
ctxt = NULL;
if (!last) {
/* reopen target file as it got closed above */
ret = filerec_open_once(tgt_extent->e_file, target_rw,
&open_files);
if (ret) {
fprintf(stderr,
"%s: Could not re-open as target.\n",
extent->e_file->filename);
break;
}
}
}
abort_on(ctxt != NULL);
abort_on(!RB_EMPTY_ROOT(&open_files.root));
if (fiemap_during_dedupe) {
get_extent_info(dext);
add_shared_extents(dext, &shared_post);
}
/*
* It's entirely possible that some other process is
* manipulating files underneath us. Take care not to
* report some randomly enormous 64 bit value.
*/
if (shared_prev < shared_post)
*fiemap_bytes += shared_post - shared_prev;
/* The only error we want to bubble up is ENOMEM */
ret = 0;
out:
/*
* ENOMEM error during context allocation may have caused open
* files to stay in our list.
*/
filerec_close_open_list(&open_files);
/*
* We might have allocated a context above but not
* filled it with any extents, make sure to free it
* here.
*/
free_dedupe_ctxt(ctxt);
abort_on(!RB_EMPTY_ROOT(&open_files.root));
return ret;
}
static GMutex dedupe_counts_mutex;
struct dedupe_counts {
uint64_t kern_bytes;
uint64_t fiemap_bytes;
};
static int extent_dedupe_worker(struct dupe_extents *dext,
uint64_t *fiemap_bytes, uint64_t *kern_bytes)
{
int ret;
unsigned long long passno = __sync_add_and_fetch(&curr_dedupe_pass, 1);
ret = dedupe_extent_list(dext, fiemap_bytes, kern_bytes, passno);
if (ret) {
if (ret == DEDUPE_EXTENTS_CLEANED)
return 0;
/* dedupe_extent_list already printed to stderr for us */
return ret;
}
if (!list_empty(&dext->de_extents)) {
g_mutex_lock(&mutex);
dupe_extents_free(dext, results_tree);
g_mutex_unlock(&mutex);
}
return 0;
}
struct block_dedupe_list
{
unsigned char bd_hash[DIGEST_LEN_MAX];
struct list_head bd_block_list;
struct hash_tree *bd_hash_tree;
};
static void free_bdl(struct block_dedupe_list *bdl);
/*
* tgt_file/tgt_off here are used only when we are asked not to dedupe
* within the same file - see block_dedupe_nosame(). Otherwise this is
* very straight-forward. We walk the dupe blocks list and insert into
* the result tree then run that through dedupe_extent_list().
*/
static int __block_dedupe(struct block_dedupe_list *bdl,
struct results_tree *res,
struct filerec *tgt_file,
uint64_t tgt_off, uint64_t tgt_len,
uint64_t *fiemap_bytes,
uint64_t *kern_bytes, unsigned long long passno)
{
int ret, one_old = 0;
struct dupe_extents *dext = NULL;
struct file_block *block;
if (tgt_file) {
/* Insert this first so it gets picked as target. */
ret = insert_one_result(res, bdl->bd_hash, tgt_file, tgt_off,
tgt_len);
if (ret)
return ret;
if (filerec_deduped(tgt_file))
one_old = 1;
}
list_for_each_entry(block, &bdl->bd_block_list, b_list) {
if (block->b_file == tgt_file)
continue;
if (filerec_deduped(block->b_file)) {
/*
* Take one old result to dedupe from. That
* way we don't get into a situation where
* we've thrown out all blocks except a single
* rescanned one.
*/
if (one_old)
continue;
one_old = 1;
}
ret = insert_one_result(res, bdl->bd_hash, block->b_file,
block->b_loff, block_len(block));
if (ret)
goto out;
}
dext = rb_entry(rb_first(&res->root), struct dupe_extents, de_node);
abort_on(!dext);
if (dext->de_num_dupes >= 2) {
ret = dedupe_extent_list(dext, fiemap_bytes, kern_bytes,
passno);
if (ret == DEDUPE_EXTENTS_CLEANED)
ret = 0;
}
out:
dext = rb_entry(rb_first(&res->root), struct dupe_extents, de_node);
if (dext) {
g_mutex_lock(&mutex);
dupe_extents_free(dext, res);
g_mutex_unlock(&mutex);
}
return ret;
}
/*
* Newer kernels (Linux v4.2+) support dedupe of the target file
* (dedupe with a file). We don't have a way of testing that yet so
* for now assume we can't dedupe with the same file. Extent dedupe
* handles this by never comparing a file to itself. We don't have
* that luxury.
*
* Walk the dupe list once until we find two files which are
* different. This is dumb but easy on memory and probably fast
* enough.
*/
static void pick_target_files(struct block_dedupe_list *bdl,
struct filerec **tgt1, uint64_t *loff1,
uint64_t *tlen1, struct filerec **tgt2,
uint64_t *loff2, uint64_t *tlen2)
{
struct filerec *file1, *file2;
uint64_t off1, off2, len1, len2;
struct file_block *block;
file1 = file2 = NULL;
off1 = off2 = len1 = len2 = 0;
list_for_each_entry(block, &bdl->bd_block_list, b_list) {
if (!file1) {
file1 = block->b_file;
off1 = block->b_loff;
len1 = block_len(block);
} else if (file1 != block->b_file) {
file2 = block->b_file;
off2 = block->b_loff;
len2 = block_len(block);
}
if (file1 && file2)
break;
}
*tgt1 = file1;
*tgt2 = file2;
*loff1 = off1;
*loff2 = off2;
*tlen1 = len1;
*tlen2 = len2;
return;
}
/* for kernels without same file dedupe (< v4.2) */
static int block_dedupe_nosame(struct block_dedupe_list *bdl,
struct results_tree *res, uint64_t *fiemap_bytes,
uint64_t *kern_bytes, unsigned long long passno)
{
int ret;
struct filerec *tgt_file1, *tgt_file2;
uint64_t tgt_off1, tgt_off2, tgt_len1, tgt_len2;
tgt_off1 = tgt_off2 = 0;
pick_target_files(bdl, &tgt_file1, &tgt_off1, &tgt_len1,
&tgt_file2, &tgt_off2, &tgt_len2);
if (!tgt_file2) {
/* Can't dedupe this in nosame mode */
if (verbose) {
printf("[%p] Can not dedupe hash ", g_thread_self());
debug_print_digest_short(stdout, bdl->bd_hash);
printf(" in nosame mode - all hashes belong to %s.\n",
tgt_file1->filename);
}
return 0;
}
ret = __block_dedupe(bdl, res, tgt_file1, tgt_off1, tgt_len1,
fiemap_bytes, kern_bytes, passno);
if (!ret)
ret = __block_dedupe(bdl, res, tgt_file2, tgt_off2, tgt_len2,
fiemap_bytes, kern_bytes, passno);
return ret;
}
static int block_dedupe_worker(struct block_dedupe_list *bdl,
uint64_t *fiemap_bytes, uint64_t *kern_bytes)
{
int ret;
struct results_tree res;
unsigned long long passno = __sync_add_and_fetch(&curr_dedupe_pass, 1);
init_results_tree(&res);
if (!dedupe_same_file)
ret = block_dedupe_nosame(bdl, &res, fiemap_bytes,
kern_bytes, passno);
else
ret = __block_dedupe(bdl, &res, NULL, 0, 0, fiemap_bytes,
kern_bytes, passno);
free_bdl(bdl);
return ret;
}
static int dedupe_worker(void *priv, struct dedupe_counts *counts)
{
int ret;
uint64_t fiemap_bytes = 0ULL;
uint64_t kern_bytes = 0ULL;
if (block_dedupe)
ret = block_dedupe_worker(priv, &fiemap_bytes, &kern_bytes);
else
ret = extent_dedupe_worker(priv, &fiemap_bytes, &kern_bytes);
g_mutex_lock(&dedupe_counts_mutex);
counts->fiemap_bytes += fiemap_bytes;
counts->kern_bytes += kern_bytes;
g_mutex_unlock(&dedupe_counts_mutex);
return ret;
}
static GThreadPool *dedupe_pool = NULL;
struct block_dedupe_list *alloc_bdl(struct hash_tree *tree,
struct dupe_blocks_list *dups)
{
struct block_dedupe_list *bdl = malloc(sizeof(*bdl));
if (bdl) {
bdl->bd_hash_tree = tree;
INIT_LIST_HEAD(&bdl->bd_block_list);
memcpy(&bdl->bd_hash, &dups->dl_hash, DIGEST_LEN_MAX);
}
return bdl;
}
static void free_bdl(struct block_dedupe_list *bdl)
{
struct file_block *block, *tmp;
if (bdl) {
list_for_each_entry_safe(block, tmp, &bdl->bd_block_list, b_list) {
list_del_init(&block->b_list);
g_mutex_lock(&mutex);
remove_hashed_block(bdl->bd_hash_tree, block);
g_mutex_unlock(&mutex);
}
free(bdl);
}
}
static int push_bdl(struct block_dedupe_list *bdl)
{
GError *err = NULL;
g_thread_pool_push(dedupe_pool, bdl, &err);
if (err) {
fprintf(stderr, "Fatal error while deduping: %s\n",
err->message);
g_error_free(err);
return 1;
}
return 0;
}
/*
* Don't queue more than DEDUPE_MAX dedupes on any one thread. We want
* this because large lists (in the multiple millions) can wind up
* taking longer on one thread than all the other lists. This happens
* because a common pattern is only one or two extremely large buckets
* (think zeros) with the rest having a drastically smaller number of
* duplicates.
*/
#define DEDUPE_MAX (1000000)
static int __push_blocks(struct hash_tree *hashes,
struct dupe_blocks_list *dups)
{
int i, ret = 0;
unsigned long long j;
struct block_dedupe_list *bdl;
struct file_block *block;
if (dups->dl_num_elem < DEDUPE_MAX) {
/* the easy way */
bdl = alloc_bdl(hashes, dups);
if (!bdl) {
fprintf(stderr, "Fatal: out of memory while deduping\n");
ret = ENOMEM;
goto out;
}
list_splice_init(&dups->dl_list, &bdl->bd_block_list);
if (push_bdl(bdl))
goto out;
} else {
unsigned long long smax = dups->dl_num_elem / io_threads;
if (verbose) {
printf("Hash ");
debug_print_digest_short(stdout, dups->dl_hash);
printf(" has %llu elements. It will be spread amongst "
"all dedupe threads.\n", dups->dl_num_elem);
}
for (i = 0; i < io_threads; i++) {
j = 0;
list_for_each_entry(block, &dups->dl_list, b_list) {
j++;
/*
* Give the last thread any leftovers
* even if that goes above our soft
* max.
*/
if (list_is_last(&block->b_list, &dups->dl_list)
|| (j == smax && i != (io_threads - 1))) {
bdl = alloc_bdl(hashes, dups);
if (!bdl) {
fprintf(stderr,
"Fatal: out of memory"
" while deduping\n");
ret = ENOMEM;
goto out;
}
list_cut_position(&bdl->bd_block_list,
&dups->dl_list,
&block->b_list);
if (push_bdl(bdl))
goto out;
bdl = NULL;
__sync_add_and_fetch(
&total_dedupe_passes, 1);
break;
}
}
}
}
bdl = NULL; /* freed by thread */
out:
free_bdl(bdl);
return ret;
}
static int push_blocks(struct hash_tree *hashes)
{
struct dupe_blocks_list *dups, *tmp;
list_for_each_entry_safe(dups, tmp, &hashes->size_list, dl_size_list) {
if (dups->dl_num_elem < 2)
continue;
/*
* Do this here because we don't want one of the
* worker threads removing it from the size list for
* us when they free the dups structure.
*/
list_del_init(&dups->dl_size_list);
if (__push_blocks(hashes, dups))
return 1;
}
return 0;
}
/* Errors from this function are fatal. */
static int push_extents(struct results_tree *res)
{
struct rb_root *root = &res->root;
struct rb_node *node = rb_first(root);
struct dupe_extents *dext;
GError *err = NULL;
while (node) {
dext = rb_entry(node, struct dupe_extents, de_node);
/*
* dext may be free'd by the dedupe threads, so get
* the next node now. In addition we want to lock
* around the rbtree code here so rb_erase doesn't
* change the tree underneath us.
*/
g_mutex_lock(&mutex);
node = rb_next(node);
g_mutex_unlock(&mutex);
g_thread_pool_push(dedupe_pool, dext, &err);
if (err) {
fprintf(stderr, "Fatal error while deduping: %s\n",
err->message);
g_error_free(err);
return 1;
}
}
return 0;
}
void dedupe_results(struct results_tree *res, struct hash_tree *hashes)
{
int ret;
struct dedupe_counts counts = { 0ULL, };
GError *err = NULL;
if (!block_dedupe) {
results_tree = res;
print_dupes_table(res);
if (RB_EMPTY_ROOT(&res->root)) {
printf("Nothing to dedupe.\n");
return;
}
}
printf("Using %u threads for dedupe phase\n", io_threads);
dedupe_pool = g_thread_pool_new((GFunc) dedupe_worker, &counts,
io_threads, TRUE, &err);
if (err) {
fprintf(stderr, "Unable to create dedupe thread pool: %s\n",
err->message);
g_error_free(err);
return;
}
if (block_dedupe) {
total_dedupe_passes = hashes->num_hashes;
leading_spaces = num_digits(total_dedupe_passes);
ret = push_blocks(hashes);
} else {
total_dedupe_passes = res->num_dupes;
leading_spaces = num_digits(total_dedupe_passes);
ret = push_extents(res);
}
if (ret) {
fprintf(stderr, "Fatal error while deduping: %s\n",
err->message);
g_error_free(err);
}
g_thread_pool_free(dedupe_pool, FALSE, TRUE);
if (ret == 0) {
if (fiemap_during_dedupe) {
printf("Kernel processed data (excludes target files): "
"%s\nComparison of extent info shows a net "
"change in shared extents of: %s\n",
pretty_size(counts.kern_bytes),
pretty_size(counts.fiemap_bytes));
} else {
printf("Kernel processed data (excludes target files): "
"%s\n", pretty_size(counts.kern_bytes));
}
}
}
int fdupes_dedupe(void)
{
int ret;
struct filerec *file;
struct dedupe_ctxt *ctxt = NULL;
uint64_t bytes = 0;
OPEN_ONCE(open_files);
list_for_each_entry(file, &filerec_list, rec_list) {
ret = filerec_open_once(file, 0, &open_files);
if (ret) {
fprintf(stderr, "%s: Skipping dedupe.\n",
file->filename);
continue;
}
printf("Queue entire file for dedupe: %s\n", file->filename);
if (ctxt == NULL) {
ctxt = new_dedupe_ctxt(MAX_DEDUPES_PER_IOCTL,
0, file->size, file);
if (ctxt == NULL) {
fprintf(stderr, "Out of memory while "
"allocating dedupe context.\n");
ret = ENOMEM;
goto out;
}
continue;
}
ret = add_extent_to_dedupe(ctxt, 0, file);
if (ret < 0) {
fprintf(stderr, "%s: Request not queued.\n",
file->filename);
ret = ENOMEM;
goto out;
} else if (ret == 0 ||
list_is_last(&file->rec_list, &filerec_list)) {
ret = dedupe_extents(ctxt);
if (ret) {
ret = errno;
fprintf(stderr,
"FAILURE: Dedupe ioctl returns %d: %s\n",
ret, strerror(ret));
goto out;
}
filerec_close_open_list(&open_files);
process_dedupe_results(ctxt, &bytes);
free_dedupe_ctxt(ctxt);
ctxt = NULL;
printf("Dedupe pass on %llu files completed\n",
num_filerecs);
}
}
ret = 0;
out:
filerec_close_open_list(&open_files);
free_dedupe_ctxt(ctxt);
free_all_filerecs();
return ret;
}