forked from naver/arcus-memcached
-
Notifications
You must be signed in to change notification settings - Fork 0
/
slabs.c
1211 lines (1079 loc) · 41.9 KB
/
slabs.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
997
998
999
1000
/* -*- Mode: C; tab-width: 4; c-basic-offset: 4; indent-tabs-mode: nil -*- */
/*
* arcus-memcached - Arcus memory cache server
* Copyright 2010-2014 NAVER Corp.
* Copyright 2014-2015 JaM2in Co., Ltd.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
/*
* Slabs memory allocation, based on powers-of-N. Slabs are up to 1MB in size
* and are divided into chunks. The chunk sizes start off at the size of the
* "item" structure plus space for a small key and value. They increase by
* a multiplier factor from there, up to half the maximum slab size. The last
* slab size is always 1MB, since that's the maximum item size allowed by the
* memcached protocol.
*/
#include "config.h"
#include <fcntl.h>
#include <errno.h>
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <assert.h>
#include <pthread.h>
#include <inttypes.h>
#include <stdarg.h>
#include "default_engine.h"
#define RESERVED_SLABS 4
#define RESERVED_SLAB_RATIO 5
#define MAX_SPACE_SHORTAGE_LEVEL 100
#define SSL_FOR_BACKGROUND_EVICT 10 /* space shortage level for background evict */
#define SSL_CHECK_BY_MEM_REQUEST 100 /* space shortage level check tick by slab*/
/* should be computed manually */
#define SMMGR_NUM_CLASSES 1025
//#define SMMGR_NUM_CLASSES 81
//#define SMMGR_NUM_CLASSES 9
//#define SMMGR_FREE_MINSIZE 32
//#define SMMGR_NUM_CLASSES 8
//#define SMMGR_FREE_MINSIZE 64
//#define SMMGR_NUM_CLASSES 5
//#define SMMGR_FREE_MINSIZE 512
/* sm slot head */
typedef struct _sm_slot {
uint32_t status; /* 0: free slot */
uint16_t offset;
uint16_t length;
struct _sm_slot *prev;
struct _sm_slot *next;
} sm_slot_t;
/* sm slot tail */
typedef struct _sm_tail {
uint16_t offset;
uint16_t length; /* 0: free slot */
} sm_tail_t;
/* sm block */
typedef struct _sm_blck {
struct _sm_blck *prev;
struct _sm_blck *next;
uint32_t frspc; /* (NOT USED) free slot space in each block */
uint32_t frcnt; /* (NOT USED) free slot count in each block */
} sm_blck_t;
/* sm slot list */
typedef struct _sm_slist {
sm_slot_t *head;
sm_slot_t *tail;
uint64_t space;
uint64_t count;
} sm_slist_t;
/* sm block list */
typedef struct _sm_blist {
sm_blck_t *head;
sm_blck_t *tail;
uint64_t count;
} sm_blist_t;
typedef struct _sm_anchor {
int space_shortage_level; /* 0, 1 ~ 100 */
int blck_clsid; /* slab class id of block */
int blck_tsize; /* block total size */
int blck_bsize; /* block body size */
int used_num_classes; /* # of used slot classes */
int free_num_classes; /* # of free slot classes */
int used_minid; /* min sm classid of used slots */
int used_maxid; /* max sm classid of used slots */
int free_minid; /* max sm classid of free slots */
int free_maxid; /* max sm classid of free slots (excluding the largest free slot) */
sm_blist_t used_blist; /* used block list */
sm_slist_t free_slist[SMMGR_NUM_CLASSES]; /* free slot list */
sm_slist_t used_slist[SMMGR_NUM_CLASSES]; /* used slot info */
uint64_t used_total_space; /* the amount of used space */
uint64_t free_small_space; /* the amount of free space that can't be used */
uint64_t free_avail_space; /* the amount of free space that can be used */
uint64_t free_chunk_space; /* the amount of free chunk space */
uint64_t free_limit_space; /* the amount of minimum free space that must be maintained */
uint32_t num_smmgr_request; /* the number that do_smmgr_alloc/do_smmgr_free are invoked */
} sm_anchor_t;
#define SMMGR_BLOCK_SIZE (64*1024)
#define SMMGR_SLOT_SIZE(size) (((((size)+sizeof(sm_tail_t)-1) / 8) + 1) * 8)
#define SMMGR_MIN_SLOT_SIZE 32
#define SMMGR_MAX_SLOT_SIZE SMMGR_SLOT_SIZE(MAX_SM_VALUE_SIZE)
static sm_anchor_t sm_anchor;
static EXTENSION_LOGGER_DESCRIPTOR *logger;
/*
* Forward Declarations
*/
static void *do_slabs_alloc(struct default_engine *engine, const size_t size, unsigned int id);
static void do_slabs_free(struct default_engine *engine, void *ptr, const size_t size, unsigned int id);
static int do_slabs_newslab(struct default_engine *engine, const unsigned int id);
static void *memory_allocate(struct default_engine *engine, size_t size);
#ifndef DONT_PREALLOC_SLABS
/* Preallocate as many slab pages as possible (called from slabs_init)
on start-up, so users don't get confused out-of-memory errors when
they do have free (in-slab) space, but no space to make new slabs.
if maxslabs is 18 (POWER_LARGEST - POWER_SMALLEST + 1), then all
slab types can be made. if max memory is less than 18 MB, only the
smaller ones will be made. */
static void slabs_preallocate (const unsigned int maxslabs);
#endif
/*
* Figures out which slab class (chunk size) is required to store an item of
* a given size.
*
* Given object size, return id to use when allocating/freeing memory for object
* 0 means error: can't store such a large object
*/
unsigned int slabs_clsid(struct default_engine *engine, const size_t size)
{
int res = POWER_SMALLEST;
if (size == 0)
return 0;
while (size > engine->slabs.slabclass[res].size)
if (res++ == engine->slabs.power_largest) /* won't fit in the biggest slab */
return 0;
return res;
}
static void do_slabs_check_space_shortage_level(struct default_engine *engine)
{
uint64_t curr_avail_space = sm_anchor.free_chunk_space
+ sm_anchor.free_avail_space;
if (curr_avail_space < sm_anchor.free_limit_space) {
/* How do we compute the space_shortage_level ?
* Use the following formula.
* => (free_limit_space / (curr_avail_space / N)) - (N-1)
*/
int ssl; /* space shortage level */
int num=10; /* N */
if ((curr_avail_space / num) > 0) {
ssl = (sm_anchor.free_limit_space / (curr_avail_space / num)) - (num-1);
if (ssl > MAX_SPACE_SHORTAGE_LEVEL) ssl = MAX_SPACE_SHORTAGE_LEVEL;
} else {
ssl = MAX_SPACE_SHORTAGE_LEVEL;
}
sm_anchor.space_shortage_level = ssl;
} else {
sm_anchor.space_shortage_level = 0;
}
}
static void do_smmgr_init(struct default_engine *engine)
{
/* small memory allocator */
memset(&sm_anchor, 0, sizeof(sm_anchor_t));
sm_anchor.blck_tsize = SMMGR_BLOCK_SIZE;
sm_anchor.blck_bsize = sm_anchor.blck_tsize - sizeof(sm_blck_t);
sm_anchor.used_minid = SMMGR_NUM_CLASSES;
sm_anchor.used_maxid = -1;
sm_anchor.free_minid = SMMGR_NUM_CLASSES;
sm_anchor.free_maxid = -1;
/* slab allocator */
/* slab class 0 is used for collection items ans small-sized kv items */
engine->slabs.slabclass[0].size = sm_anchor.blck_tsize;
engine->slabs.slabclass[0].perslab = engine->config.item_size_max / engine->slabs.slabclass[0].size;
engine->slabs.slabclass[0].rsvd_slabs = 0; // undefined
}
static inline int do_smmgr_memid(size_t size)
{
if (size < 8192) return (int)(size/8);
else return SMMGR_NUM_CLASSES-1;
#if 0
int memid = -1;
int limit = SMMGR_FREE_MINSIZE;
while (size >= limit && memid < (SMMGR_NUM_CLASSES-1)) {
limit *= 2;
memid += 1;
}
assert(memid < SMMGR_NUM_CLASSES);
return memid;
#endif
#if 0
int memid;
if (size < 2048) {
if (size < 1024) memid = ( 0 + ((size ) / 32));
else memid = (32 + ((size-1024) / 64));
} else if (size < 8192) {
if (size < 4096) memid = (48 + ((size-2048) / 128));
else memid = (64 + ((size-4096) / 256));
} else {
memid = SMMGR_NUM_CLASSES-1;
}
assert(memid < SMMGR_NUM_CLASSES);
return memid;
#endif
#if 0
int memid;
if (size < 2048) {
if (size < 1024) memid = ( 0 + ((size ) / 16));
else memid = ( 64 + ((size-1024) / 32));
} else if (size < 8192) {
if (size < 4096) memid = ( 96 + ((size-2048) / 64));
else memid = (128 + ((size-4096) / 128));
} else {
memid = SMMGR_NUM_CLASSES-1;
}
assert(memid < SMMGR_NUM_CLASSES);
return memid;
#endif
#if 0
int memid;
if (size <= 2048) {
if (size <= 512) memid = (0 + ((size -1) / 16));
else if (size <= 1024) memid = (32 + ((size- 512-1) / 32));
else memid = (48 + ((size-1024-1) / 64));
} else {
if (size <= 4096) memid = (64 + ((size-2048-1) / 128));
else if (size <= 8192) memid = (80 + ((size-4096-1) / 256));
else memid = 96;
}
assert(memid < SMMGR_NUM_CLASSES);
return memid;
#endif
}
static void do_smmgr_used_slot_list_add(int targ)
{
if (targ < sm_anchor.used_minid) {
sm_anchor.used_minid = targ;
}
if (targ > sm_anchor.used_maxid) {
/* adjust free_space small & avail */
int smid = (sm_anchor.used_maxid < 0 ? 0 : sm_anchor.used_maxid);
for ( ; smid < targ; smid++) {
if (sm_anchor.free_slist[smid].space > 0) {
sm_anchor.free_small_space += sm_anchor.free_slist[smid].space;
sm_anchor.free_avail_space -= sm_anchor.free_slist[smid].space;
}
}
sm_anchor.used_maxid = targ;
}
sm_anchor.used_num_classes += 1;
}
static void do_smmgr_used_slot_list_del(int targ)
{
int smid;
if (targ == sm_anchor.used_minid) {
if (sm_anchor.used_total_space > 0) {
for (smid = sm_anchor.used_minid+1; smid <= sm_anchor.used_maxid; smid++) {
if (sm_anchor.used_slist[smid].count > 0) break;
}
sm_anchor.used_minid = smid;
} else {
sm_anchor.used_minid = SMMGR_NUM_CLASSES;
}
}
if (targ == sm_anchor.used_maxid) {
if (sm_anchor.used_total_space > 0) {
for (smid = sm_anchor.used_maxid-1; smid >= sm_anchor.used_minid; smid--) {
if (sm_anchor.used_slist[smid].count > 0) break;
}
sm_anchor.used_maxid = smid;
} else {
sm_anchor.used_maxid = -1;
}
/* adjust free_space small & avail */
smid = (sm_anchor.used_maxid < 0 ? 0 : sm_anchor.used_maxid);
for ( ; smid < targ; smid++) {
if (sm_anchor.free_slist[smid].space > 0) {
sm_anchor.free_small_space -= sm_anchor.free_slist[smid].space;
sm_anchor.free_avail_space += sm_anchor.free_slist[smid].space;
}
}
}
sm_anchor.used_num_classes -= 1;
}
static void do_smmgr_free_slot_list_add(int targ)
{
if (targ < (SMMGR_NUM_CLASSES-1)) {
if (sm_anchor.free_minid > targ) {
sm_anchor.free_minid = targ;
}
if (sm_anchor.free_maxid < targ) {
sm_anchor.free_maxid = targ;
}
sm_anchor.free_num_classes += 1;
}
/*****
if (sm_anchor.free_maxid < targ) {
sm_anchor.free_maxid = targ;
}
*****/
}
static void do_smmgr_free_slot_list_del(int targ)
{
if (targ < (SMMGR_NUM_CLASSES-1)) {
int smid;
if (targ == sm_anchor.free_minid) {
if (sm_anchor.free_minid < sm_anchor.free_maxid) {
for (smid = sm_anchor.free_minid+1; smid <= sm_anchor.free_maxid; smid++) {
if (sm_anchor.free_slist[smid].head != NULL) break;
}
sm_anchor.free_minid = smid;
} else {
sm_anchor.free_minid = SMMGR_NUM_CLASSES;
}
}
if (targ == sm_anchor.free_maxid) {
if (sm_anchor.free_maxid > sm_anchor.free_minid) {
for (smid = sm_anchor.free_maxid-1; smid >= sm_anchor.free_minid; smid--) {
if (sm_anchor.free_slist[smid].head != NULL) break;
}
sm_anchor.free_maxid = smid;
} else {
sm_anchor.free_maxid = -1;
}
}
sm_anchor.free_num_classes -= 1;
}
}
static void do_smmgr_used_slot_init(sm_slot_t *slot, uint16_t offset, uint16_t length)
{
sm_tail_t *tail = (sm_tail_t*)((char*)slot + length - sizeof(sm_tail_t));
tail->offset = offset;
tail->length = length; /* used slot */
/* Set it as used slot.
* In eager invalidation, incomplete slot can be inspected.
* At that time, the slot must be regared as an used slot.
*/
slot->status = (uint32_t)-1; /* used slot */
}
static void do_smmgr_free_slot_init(sm_slot_t *slot, uint16_t offset, uint16_t length)
{
sm_tail_t *tail = (sm_tail_t*)((char*)slot + length - sizeof(sm_tail_t));
tail->offset = offset;
tail->length = 0; /* free slot */
slot->status = 0; /* free slot */
slot->offset = offset;
slot->length = length;
}
static void do_smmgr_free_slot_link(sm_slot_t *slot)
{
sm_slist_t *list;
int smid;
if (slot->length < SMMGR_MIN_SLOT_SIZE) {
sm_anchor.free_small_space += slot->length;
return;
}
smid = do_smmgr_memid(slot->length);
list = &sm_anchor.free_slist[smid];
if (list->head == NULL) {
assert(list->tail == NULL && list->count == 0 && list->space == 0);
slot->prev = NULL;
slot->next = NULL;
list->head = slot;
list->tail = slot;
list->space = slot->length;
list->count = 1;
do_smmgr_free_slot_list_add(smid);
} else {
assert(list->tail != NULL && list->count > 0 && list->space > 0);
slot->prev = list->tail;
slot->next = NULL;
slot->prev->next = slot;
list->tail = slot;
list->space += slot->length;
list->count += 1;
}
if (smid < sm_anchor.used_maxid) {
sm_anchor.free_small_space += slot->length;
} else {
sm_anchor.free_avail_space += slot->length;
}
}
static void do_smmgr_free_slot_unlink(sm_slot_t *slot)
{
sm_slist_t *list;
int smid;
if (slot->length < SMMGR_MIN_SLOT_SIZE) {
sm_anchor.free_small_space -= slot->length;
return;
}
smid = do_smmgr_memid(slot->length);
if (smid < sm_anchor.used_maxid) {
sm_anchor.free_small_space -= slot->length;
} else {
sm_anchor.free_avail_space -= slot->length;
}
list = &sm_anchor.free_slist[smid];
if (list->count == 1) {
assert(list->space == slot->length && list->head == slot && list->tail == slot);
list->head = NULL;
list->tail = NULL;
list->space = 0;
list->count = 0;
do_smmgr_free_slot_list_del(smid);
} else {
assert(list->count > 1 && list->space > slot->length &&
list->head != NULL && list->tail != NULL);
if (slot->prev != NULL) slot->prev->next = slot->next;
if (slot->next != NULL) slot->next->prev = slot->prev;
if (list->head == slot) list->head = slot->next;
if (list->tail == slot) list->tail = slot->prev;
list->space -= slot->length;
list->count -= 1;
}
}
static void do_smmgr_free_slot_replace(sm_slot_t *old_slot, sm_slot_t *new_slot, int smid)
{
sm_slist_t *list = &sm_anchor.free_slist[smid];
assert(list->head == old_slot);
uint16_t diff_leng = old_slot->length - new_slot->length;
new_slot->prev = NULL;
if (old_slot->next == NULL) {
new_slot->next = NULL;
list->tail = new_slot;
} else {
new_slot->next = old_slot->next;
new_slot->next->prev = new_slot;
}
list->head = new_slot;
list->space -= diff_leng;
if (smid < sm_anchor.used_maxid) {
sm_anchor.free_small_space -= diff_leng;
} else {
sm_anchor.free_avail_space -= diff_leng;
}
}
#if 0 // can be used later
static void do_smmgr_used_blck_check(void)
{
sm_blck_t *blck;
sm_slot_t *slot;
sm_tail_t *tail;
uint64_t blck_count = 0;
uint64_t used_count = 0;
uint64_t free_count = 0;
uint32_t comp_length;
blck = sm_anchor.used_blist.head;
while (blck != NULL) {
blck_count += 1;
tail = (sm_tail_t*)((char*)blck + sm_anchor.blck_tsize - sizeof(sm_tail_t));
while (((char*)tail - (char*)blck) > sizeof(sm_blck_t)) {
slot = (sm_slot_t*)((char*)blck + tail->offset);
if (tail->length > 8) { /* used slot */
used_count += 1;
assert(slot->status != 0);
} else { /* free slot */
free_count += 1;
comp_length = (char*)tail - (char*)slot + sizeof(sm_tail_t);
assert(slot->status == 0);
assert(slot->offset == tail->offset);
assert(slot->length == comp_length);
}
tail = (sm_tail_t*)((char*)slot - sizeof(sm_tail_t));
}
blck = blck->next;
}
assert(blck_count == sm_anchor.used_blist.count);
}
#endif
static void do_smmgr_used_blck_link(sm_blck_t *blck)
{
blck->frspc = 0;
blck->frcnt = 0;
blck->prev = sm_anchor.used_blist.tail;
blck->next = NULL;
if (sm_anchor.used_blist.head == NULL) {
sm_anchor.used_blist.head = blck;
sm_anchor.used_blist.tail = blck;
} else {
blck->prev->next = blck;
sm_anchor.used_blist.tail = blck;
}
sm_anchor.used_blist.count += 1;
}
static void do_smmgr_used_blck_unlink(sm_blck_t *blck)
{
if (blck->prev != NULL) blck->prev->next = blck->next;
if (blck->next != NULL) blck->next->prev = blck->prev;
if (sm_anchor.used_blist.head == blck) sm_anchor.used_blist.head = blck->next;
if (sm_anchor.used_blist.tail == blck) sm_anchor.used_blist.tail = blck->prev;
sm_anchor.used_blist.count -= 1;
}
static sm_blck_t *do_smmgr_blck_alloc(struct default_engine *engine)
{
sm_blck_t *blck = (sm_blck_t *)do_slabs_alloc(engine, sm_anchor.blck_tsize, sm_anchor.blck_clsid);
if (blck != NULL) {
if (sm_anchor.free_limit_space > 0) {
sm_anchor.free_chunk_space -= sm_anchor.blck_tsize;
}
do_smmgr_used_blck_link(blck);
} else {
logger->log(EXTENSION_LOG_INFO, NULL, "no more small memory chunk\n");
}
return blck;
}
static void do_smmgr_blck_free(struct default_engine *engine, sm_blck_t *blck)
{
do_smmgr_used_blck_unlink(blck);
do_slabs_free(engine, blck, sm_anchor.blck_tsize, sm_anchor.blck_clsid);
if (sm_anchor.free_limit_space > 0) {
sm_anchor.free_chunk_space += sm_anchor.blck_tsize;
}
}
static void *do_smmgr_alloc(struct default_engine *engine, const size_t size)
{
sm_slot_t *cur_slot = NULL;
int smid, targ;
int slen;
if (sm_anchor.free_limit_space > 0) {
if ((++sm_anchor.num_smmgr_request) > SSL_CHECK_BY_MEM_REQUEST) {
sm_anchor.num_smmgr_request = 0;
do_slabs_check_space_shortage_level(engine);
if (sm_anchor.space_shortage_level >= SSL_FOR_BACKGROUND_EVICT)
coll_del_thread_wakeup(engine);
}
}
slen = SMMGR_SLOT_SIZE(size);
if (slen < SMMGR_MIN_SLOT_SIZE)
slen = SMMGR_MIN_SLOT_SIZE;
targ = do_smmgr_memid(slen);
/* allocate slot from free slot list */
#if 0 // check
if (targ <= sm_anchor.free_maxid) {
for (smid = targ; smid <= sm_anchor.free_maxid; smid++) {
if (sm_anchor.free_slist[smid].head != NULL) break;
}
} else {
smid = SMMGR_NUM_CLASSES-1;
}
cur_slot = sm_anchor.free_slist[smid].head;
#endif
#if 0 // check
if (targ <= sm_anchor.free_maxid) {
if (sm_anchor.free_slist[targ].head != NULL) smid = targ;
else smid = sm_anchor.free_maxid;
} else {
smid = SMMGR_NUM_CLASSES-1;
}
cur_slot = sm_anchor.free_slist[smid].head;
#endif
#if 0
for (smid = targ; smid < sm_anchor.free_maxid; smid += targ) {
if (sm_anchor.free_slist[smid].head != NULL) {
cur_slot = sm_anchor.free_slist[smid].head; break;
}
}
if (cur_slot == NULL && targ < sm_anchor.free_maxid) {
smid = sm_anchor.free_maxid;
cur_slot = sm_anchor.free_slist[smid].head;
assert(cur_slot != NULL);
}
#endif
if (targ <= sm_anchor.free_maxid) {
if (sm_anchor.free_slist[targ].head != NULL) {
smid = targ;
} else if ((targ*2) <= sm_anchor.free_maxid) {
for (smid = targ*2; smid <= sm_anchor.free_maxid; smid++) {
if (sm_anchor.free_slist[smid].head != NULL) break;
}
} else {
smid = sm_anchor.free_maxid;
}
} else {
smid = SMMGR_NUM_CLASSES-1;
}
cur_slot = sm_anchor.free_slist[smid].head;
if (cur_slot == NULL) {
sm_blck_t *blck = do_smmgr_blck_alloc(engine);
if (blck == NULL) return NULL;
cur_slot = (sm_slot_t*)((char*)blck + sizeof(sm_blck_t) + slen);
do_smmgr_free_slot_init(cur_slot, sizeof(sm_blck_t) + slen, sm_anchor.blck_bsize - slen);
do_smmgr_free_slot_link(cur_slot);
cur_slot = (sm_slot_t*)((char*)blck + sizeof(sm_blck_t));
do_smmgr_used_slot_init(cur_slot, sizeof(sm_blck_t), slen);
} else {
if (cur_slot->length > slen) {
sm_slot_t *nxt_slot = (sm_slot_t*)((char*)cur_slot + slen);
if (smid != do_smmgr_memid(cur_slot->length - slen)) {
do_smmgr_free_slot_unlink(cur_slot);
do_smmgr_free_slot_init(nxt_slot, cur_slot->offset + slen, cur_slot->length - slen);
do_smmgr_free_slot_link(nxt_slot);
} else {
do_smmgr_free_slot_init(nxt_slot, cur_slot->offset + slen, cur_slot->length - slen);
do_smmgr_free_slot_replace(cur_slot, nxt_slot, smid);
}
} else {
do_smmgr_free_slot_unlink(cur_slot);
}
do_smmgr_used_slot_init(cur_slot, cur_slot->offset, slen);
}
/* used slot stats */
sm_anchor.used_total_space += slen;
sm_anchor.used_slist[targ].space += slen;
sm_anchor.used_slist[targ].count += 1;
if (sm_anchor.used_slist[targ].count == 1) {
do_smmgr_used_slot_list_add(targ);
}
//do_smmgr_used_blck_check();
return (void*)cur_slot;
}
static void do_smmgr_free(struct default_engine *engine, void *ptr, const size_t size)
{
sm_blck_t *cur_blck;
sm_slot_t *cur_slot;
sm_tail_t *cur_tail;
int targ;
int slen;
if (sm_anchor.free_limit_space > 0) {
if ((++sm_anchor.num_smmgr_request) > SSL_CHECK_BY_MEM_REQUEST) {
sm_anchor.num_smmgr_request = 0;
do_slabs_check_space_shortage_level(engine);
if (sm_anchor.space_shortage_level >= SSL_FOR_BACKGROUND_EVICT)
coll_del_thread_wakeup(engine);
}
}
slen = SMMGR_SLOT_SIZE(size);
if (slen < SMMGR_MIN_SLOT_SIZE)
slen = SMMGR_MIN_SLOT_SIZE;
targ = do_smmgr_memid(slen);
//do_smmgr_used_blck_check();
cur_tail = (sm_tail_t*)((char*)ptr + slen - sizeof(sm_tail_t));
assert(cur_tail->length == slen);
cur_blck = (sm_blck_t*)((char*)ptr - cur_tail->offset);
/* check if prev slot is in freed state. if then, merge with that */
if (cur_tail->offset > sizeof(sm_blck_t)) {
sm_slot_t *prv_slot;
sm_tail_t *prv_tail = (sm_tail_t*)((char*)ptr - sizeof(sm_tail_t));
if (prv_tail->length <= 8) { /* free slot */
prv_slot = (sm_slot_t*)((char*)cur_blck + prv_tail->offset);
assert(prv_slot->offset == prv_tail->offset);
do_smmgr_free_slot_unlink(prv_slot);
cur_tail->offset = prv_slot->offset;
cur_tail->length += prv_slot->length;
}
}
/* check if next slot is in freed state. if then, merge with that */
if ((cur_tail->offset + cur_tail->length) < sm_anchor.blck_tsize) {
sm_slot_t *nxt_slot = (sm_slot_t*)((char*)cur_tail + sizeof(sm_tail_t));
sm_tail_t *nxt_tail;
if (nxt_slot->status == 0) { /* free slot */
nxt_tail = (sm_tail_t*)((char*)nxt_slot + nxt_slot->length - sizeof(sm_tail_t));
assert(nxt_tail->offset == nxt_slot->offset && nxt_tail->length <= 8);
do_smmgr_free_slot_unlink(nxt_slot);
nxt_tail->offset = cur_tail->offset;
nxt_tail->length = cur_tail->length + nxt_slot->length;
cur_tail = nxt_tail;
}
}
if (cur_tail->offset > sizeof(sm_blck_t) || cur_tail->length < sm_anchor.blck_bsize) {
cur_slot = (sm_slot_t*)((char*)cur_tail - cur_tail->length + sizeof(sm_tail_t));
do_smmgr_free_slot_init(cur_slot, cur_tail->offset, cur_tail->length);
do_smmgr_free_slot_link(cur_slot);
} else {
do_smmgr_blck_free(engine, cur_blck);
}
/* used slot stats */
assert(sm_anchor.used_slist[targ].count >= 1);
sm_anchor.used_total_space -= slen;
sm_anchor.used_slist[targ].space -= slen;
sm_anchor.used_slist[targ].count -= 1;
if (sm_anchor.used_slist[targ].count == 0) {
do_smmgr_used_slot_list_del(targ);
}
}
unsigned int slabs_space_size(struct default_engine *engine, const size_t size)
{
if (size <= MAX_SM_VALUE_SIZE) {
return SMMGR_SLOT_SIZE(size);
}
int clsid = slabs_clsid(engine, size);
if (clsid == 0)
return 0;
else
return engine->slabs.slabclass[clsid].size;
}
int slabs_space_shortage_level(void)
{
return sm_anchor.space_shortage_level;
}
/**
* Determines the chunk sizes and initializes the slab class descriptors
* accordingly.
*/
ENGINE_ERROR_CODE slabs_init(struct default_engine *engine,
const size_t limit, const double factor, const bool prealloc)
{
int i = POWER_SMALLEST - 1;
unsigned int size = sizeof(hash_item) + engine->config.chunk_size;
logger = engine->server.log->get_logger();
engine->slabs.mem_limit = limit;
engine->slabs.mem_reserved = (limit / 100) * RESERVED_SLAB_RATIO;
if (engine->slabs.mem_reserved < (RESERVED_SLABS*engine->config.item_size_max))
engine->slabs.mem_reserved = (RESERVED_SLABS*engine->config.item_size_max);
if (prealloc) {
/* Allocate everything in a big chunk with malloc */
engine->slabs.mem_base = malloc(engine->slabs.mem_limit);
if (engine->slabs.mem_base != NULL) {
engine->slabs.mem_current = engine->slabs.mem_base;
engine->slabs.mem_avail = engine->slabs.mem_limit;
} else {
return ENGINE_ENOMEM;
}
} else {
engine->slabs.mem_base = NULL;
engine->slabs.mem_current = NULL;
engine->slabs.mem_avail = 0;
}
memset(engine->slabs.slabclass, 0, sizeof(engine->slabs.slabclass));
while (++i < POWER_LARGEST && size <= engine->config.item_size_max / factor) {
/* Make sure items are always n-byte aligned */
if (size % CHUNK_ALIGN_BYTES)
size += CHUNK_ALIGN_BYTES - (size % CHUNK_ALIGN_BYTES);
engine->slabs.slabclass[i].size = size;
engine->slabs.slabclass[i].perslab = engine->config.item_size_max / engine->slabs.slabclass[i].size;
engine->slabs.slabclass[i].rsvd_slabs = RESERVED_SLABS;
size *= factor;
if (engine->config.verbose > 1) {
fprintf(stderr, "slab class %3d: chunk size %9u perslab %7u\n",
i, engine->slabs.slabclass[i].size, engine->slabs.slabclass[i].perslab);
}
}
engine->slabs.power_largest = i;
engine->slabs.slabclass[engine->slabs.power_largest].size = engine->config.item_size_max;
engine->slabs.slabclass[engine->slabs.power_largest].perslab = 1;
engine->slabs.slabclass[engine->slabs.power_largest].rsvd_slabs = RESERVED_SLABS;
if (engine->config.verbose > 1) {
fprintf(stderr, "slab class %3d: chunk size %9u perslab %7u\n",
i, engine->slabs.slabclass[i].size, engine->slabs.slabclass[i].perslab);
}
/* for the test suite: faking of how much we've already malloc'd */
{
char *t_initial_malloc = getenv("T_MEMD_INITIAL_MALLOC");
if (t_initial_malloc) {
engine->slabs.mem_malloced = (size_t)atol(t_initial_malloc);
}
}
#ifndef DONT_PREALLOC_SLABS
{
char *pre_alloc = getenv("T_MEMD_SLABS_ALLOC");
if (pre_alloc == NULL || atoi(pre_alloc) != 0) {
slabs_preallocate(power_largest);
}
}
#endif
do_smmgr_init(engine);
logger->log(EXTENSION_LOG_INFO, NULL, "SLABS module initialized.\n");
return ENGINE_SUCCESS;
}
void slabs_final(struct default_engine *engine)
{
/* Free memory allocated. */
/* Do nothing, currently. */
logger->log(EXTENSION_LOG_INFO, NULL, "SLABS module destroyed.\n");
}
#ifndef DONT_PREALLOC_SLABS
static void slabs_preallocate(const unsigned int maxslabs)
{
int i;
unsigned int prealloc = 0;
/* pre-allocate a 1MB slab in every size class so people don't get
confused by non-intuitive "SERVER_ERROR out of memory"
messages. this is the most common question on the mailing
list. if you really don't want this, you can rebuild without
these three lines. */
for (i = POWER_SMALLEST; i <= POWER_LARGEST; i++) {
if (++prealloc > maxslabs)
return;
do_slabs_newslab(i);
}
/* slab class 0 is used for collection items and small-size kv items */
do_slabs_newslab(0);
}
#endif
static int grow_slab_list (struct default_engine *engine, const unsigned int id)
{
slabclass_t *p = &engine->slabs.slabclass[id];
if (p->slabs == p->list_size) {
size_t new_size = (p->list_size != 0) ? p->list_size * 2 : 16;
void *new_list = realloc(p->slab_list, new_size * sizeof(void *));
if (new_list == 0) return 0;
p->list_size = new_size;
p->slab_list = new_list;
}
return 1;
}
static int do_slabs_newslab(struct default_engine *engine, const unsigned int id)
{
slabclass_t *p = &engine->slabs.slabclass[id];
int len = p->size * p->perslab;
char *ptr;
if ((engine->slabs.mem_limit && engine->slabs.mem_malloced + len > engine->slabs.mem_limit && p->slabs >= p->rsvd_slabs) ||
(grow_slab_list(engine, id) == 0) ||
((ptr = memory_allocate(engine, (size_t)len)) == 0)) {
MEMCACHED_SLABS_SLABCLASS_ALLOCATE_FAILED(id);
return 0;
}
memset(ptr, 0, (size_t)len);
p->end_page_ptr = ptr;
p->end_page_free = p->perslab;
p->slab_list[p->slabs++] = ptr;
engine->slabs.mem_malloced += len;
if ((engine->slabs.mem_limit <= engine->slabs.mem_malloced) ||
((engine->slabs.mem_limit - engine->slabs.mem_malloced) < engine->slabs.mem_reserved))
{
if (engine->slabs.slabclass[sm_anchor.blck_clsid].rsvd_slabs == 0) { /* undefined */
/* define the reserved slab count of slab class 0 */
slabclass_t *z = &engine->slabs.slabclass[sm_anchor.blck_clsid];
unsigned int additional_slabs = (z->slabs/100) * RESERVED_SLAB_RATIO;
if (additional_slabs < RESERVED_SLABS)
additional_slabs = RESERVED_SLABS;
z->rsvd_slabs = z->slabs + additional_slabs;
sm_anchor.free_limit_space = (additional_slabs * z->perslab) * sm_anchor.blck_tsize;
sm_anchor.free_chunk_space = sm_anchor.free_limit_space
+ (z->sl_curr + z->end_page_free) * sm_anchor.blck_tsize;
}
}
MEMCACHED_SLABS_SLABCLASS_ALLOCATE(id);
return 1;
}
/*@null@*/
static void *do_slabs_alloc(struct default_engine *engine, const size_t size, unsigned int id)
{
slabclass_t *p;
void *ret = NULL;
if (size <= MAX_SM_VALUE_SIZE) {
return do_smmgr_alloc(engine, size);
}
p = &engine->slabs.slabclass[id];
#ifdef USE_SYSTEM_MALLOC
if (engine->slabs.mem_limit && engine->slabs.mem_malloced + size > engine->slabs.mem_limit) {
MEMCACHED_SLABS_ALLOCATE_FAILED(size, id);
return 0;
}
engine->slabs.mem_malloced += size;
ret = malloc(size);
MEMCACHED_SLABS_ALLOCATE(size, id, 0, ret);
return ret;
#endif
/* fail unless we have space at the end of a recently allocated page,
we have something on our freelist, or we could allocate a new page */
if (! (p->end_page_ptr != 0 || p->sl_curr != 0 ||
do_slabs_newslab(engine, id) != 0)) {
/* We don't have more memory available */
ret = NULL;
} else if (p->sl_curr != 0) {
/* return off our freelist */
ret = p->slots[--p->sl_curr];
} else {
/* if we recently allocated a whole page, return from that */
assert(p->end_page_ptr != NULL);
ret = p->end_page_ptr;
if (--p->end_page_free != 0) {
p->end_page_ptr = ((caddr_t)p->end_page_ptr) + p->size;
} else {
p->end_page_ptr = 0;
}
}
if (ret) {
p->requested += size;
MEMCACHED_SLABS_ALLOCATE(size, id, p->size, ret);
} else {
MEMCACHED_SLABS_ALLOCATE_FAILED(size, id);
}
return ret;
}
static void do_slabs_free(struct default_engine *engine, void *ptr, const size_t size, unsigned int id)
{
slabclass_t *p;
if (size <= MAX_SM_VALUE_SIZE) {
do_smmgr_free(engine, ptr, size);
return;
}
MEMCACHED_SLABS_FREE(size, id, ptr);
p = &engine->slabs.slabclass[id];
#ifdef USE_SYSTEM_MALLOC
engine->slabs.mem_malloced -= size;
free(ptr);
return;
#endif