/
dm-stripe.c
1644 lines (1449 loc) · 49 KB
/
dm-stripe.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
/*
* Copyright (C) 2001-2003 Sistina Software (UK) Limited.
*
* This file is released under the GPL.
*/
#include "dm.h"
#include <linux/device-mapper.h>
#include <linux/module.h>
#include <linux/init.h>
#include <linux/blkdev.h>
#include <linux/blk-mq.h>
#include <scsi/scsi_device.h>
#include <scsi/scsi_cmnd.h>
#include <scsi/scsi_driver.h>
#include <linux/bio.h>
#include <linux/slab.h>
#include <linux/log2.h>
#include <linux/kthread.h>
#include <linux/delay.h>
#include <linux/vmalloc.h>
#include <linux/dm-io.h>
#include <linux/spinlock.h>
#include <linux/rcupdate.h>
#define DM_MSG_PREFIX "striped"
#define DM_IO_ERROR_THRESHOLD 15
#define minor_shift 4
#define num_flag_per_page (4096/sizeof(struct flag_nodes))
#define gc_buffer_size 50
#define GC_Weight 4
#define Targeted_Weight 3
#define Writing_Weight 2
#define Writed_Weight 1
#define Clean_Weight 0
#define winnowing 0
#define WEATHERING_RATIO 30
#define KIZIL_POLICY 0
#define Return_Weight() if(!KIZIL_POLICY && vc->gp_list[tp] == Targeted_Weight) vc->gp_list[tp] -= Targeted_Weight;\
else if(!KIZIL_POLICY && vc->gp_list[tp] == Writing_Weight) vc->gp_list[tp] -= Writing_Weight;\
struct frc{
char* buf;
unsigned long long msector;
};
struct reverse_nodes{
sector_t index;
unsigned char dirty;
};
struct flag_nodes{
sector_t msector;
unsigned int wp;
unsigned int num_moved;
};
struct flag_set{
struct flag_nodes** table;
struct kmem_cache* node_buf;
struct reverse_nodes** reverse_table;
};
struct buf_set{
char *buf;
unsigned long long index;
unsigned long long sector;
char size;
unsigned int target_wp;
};
struct gc_set{
unsigned char set_num;
struct task_struct *r_id;
struct task_struct *w_id;
struct buf_set *bs;
struct dm_target *ti;
struct mutex *gc_lock;
unsigned int tp;
unsigned int gp;
sector_t tp_io_sector;
unsigned int ptr_ovflw_size;
char *kijil_map;
unsigned long long tp_table_size;
unsigned long long kijil_size;
char phase_flag;
};
struct vm {
struct dm_dev *dev;
sector_t physical_start;
sector_t end_sector;
unsigned int main_dev;
unsigned int maj_dev;
unsigned long long num_dirty;
unsigned char gen;
atomic_t error_count;
};
struct vm_c {
uint32_t vms;
int vms_shift;
/* The size of this target / num. stripes */
sector_t vm_width;
uint32_t chunk_size;
int chunk_size_shift;
/* Needed for handling events */
struct dm_target *ti;
/* Work struct used for triggering events*/
struct work_struct trigger_event;
/* volume manager variable*/
unsigned int wp;//device Write pointer
unsigned char *gp_list; //need do gc device
unsigned long long *ws;//in device Write sector pointer
unsigned long long *d_num;
unsigned long long num_entry;// number of table's entry
unsigned char mig_flag;
unsigned int num_map_block;
unsigned int num_gp;
unsigned char overload;
struct flag_set* fs;
struct gc_set* gs;
struct mutex lock;
struct mutex gc_lock;
unsigned long long read_index;
unsigned long long cur_sector;
unsigned char gc_flag;
struct dm_io_client *io_client;
unsigned int debug;
struct vm vm[0];
};
static int read_job(struct gc_set *);
static int write_job(struct gc_set *);
/*static struct flag_nodes* vm_lfs_map_sector(struct vm_c *vc, sector_t target_sector,
unsigned int wp, sector_t *write_sector, struct block_device **bdev, unsigned long bi_rw);*/
/*
* An event is triggered whenever a drive
* drops out of a stripe volume.
*/
static int atoj(const char *name){//ascii to major number
int val;
for(val=0;;name++){
val = 10 *val + (*name - '0');
if(*name == ':'){
break;
}
}
return val;
}
static int atom(const char *name){//ascii to minor number
int val;
for(;;name++){
if(*name == ':'){
name++;
break;
}
}
for(val=0;;name++){
switch(*name){
case '0'...'9':
val = 10 *val + (*name - '0');
break;
default:
return val;
}
}
return val;
}
static void trigger_event(struct work_struct *work)
{
struct vm_c *vc = container_of(work, struct vm_c,
trigger_event);
printk("trigger event\n");
dm_table_event(vc->ti->table);
}
static inline struct vm_c *alloc_context(unsigned int vms)
{
size_t len;
if (dm_array_too_big(sizeof(struct vm_c), sizeof(struct vm),
vms))
return NULL;
len = sizeof(struct vm_c) + (sizeof(struct vm) * vms);
//return kmalloc(len, GFP_NOFS);
return kmalloc(len, GFP_KERNEL);
}
/*
* Parse a single <dev> <sector> pair
*/
static int get_vm(struct dm_target *ti, struct vm_c *vc,
unsigned int vm, char **argv)
{
unsigned long long start;
char dummy;
if (sscanf(argv[1], "%llu%c", &start, &dummy) != 1)
return -EINVAL;
if (dm_get_device(ti, argv[0], dm_table_get_mode(ti->table),
&vc->vm[vm].dev))
return -ENXIO;
vc->vm[vm].physical_start = start;
return 0;
}
/*
* Construct a striped mapping.
* <number of stripes> <chunk size> [<dev_path> <offset>]+
*/
static int vm_ctr(struct dm_target *ti, unsigned int argc, char **argv)
{
struct vm_c *vc;
sector_t width, tmp_len;
uint32_t vms;
uint32_t chunk_size;
int r;
unsigned long long i;
if (argc < 2) {
ti->error = "Not enough arguments";
return -EINVAL;
}
if (kstrtouint(argv[0], 10, &vms) || !vms) {
ti->error = "Invalid stripe count";
return -EINVAL;
}
if (kstrtouint(argv[1], 10, &chunk_size) || !chunk_size) {
ti->error = "Invalid chunk_size";
return -EINVAL;
}
width = ti->len;
if (sector_div(width, vms)) {
ti->error = "Target length not divisible by "
"number of stripes";
return -EINVAL;
}
tmp_len = width;
if (sector_div(tmp_len, chunk_size)) {
ti->error = "Target length not divisible by "
"chunk size";
return -EINVAL;
}
/*
* Do we have enough arguments for that many stripes ?
*/
if (argc != (2 + 2 * vms)) {
ti->error = "Not enough destinations "
"specified";
return -EINVAL;
}
vc = alloc_context(vms);
if (!vc) {
ti->error = "Memory allocation for striped context "
"failed";
return -ENOMEM;
}
INIT_WORK(&vc->trigger_event, trigger_event);
/* Set pointer to dm target; used in trigger_event */
vc->ti = ti;
vc->vms = vms;
vc->vm_width = width;
if (vms & (vms - 1))
vc->vms_shift = -1;
else
vc->vms_shift = __ffs(vms);
r = dm_set_target_max_io_len(ti, chunk_size);
if (r) {
kfree(vc);
return r;
}
ti->num_flush_bios = vms;
ti->num_discard_bios = vms;
ti->num_write_same_bios = vms;
vc->chunk_size = chunk_size;
if (chunk_size & (chunk_size - 1))
vc->chunk_size_shift = -1;
else
vc->chunk_size_shift = __ffs(chunk_size);
/*
* Get the stripe destinations.
*/
for (i = 0; i < vms; i++) {
argv += 2;
r = get_vm(ti, vc, i, argv);
if (r < 0) {
ti->error = "Couldn't parse stripe destination";
while (i--)
dm_put_device(ti, vc->vm[i].dev);
kfree(vc);
return r;
}
atomic_set(&(vc->vm[i].error_count), 0);
}
/*volume manager initialize*/
vc->wp = 0;//////current 0 is NVMe
//vc->wp = 1;
vc->ws = kmalloc(sizeof(unsigned long long) * vc->vms, GFP_KERNEL);
for(i = 0; i<vc->vms; i++)
vc->ws[i] = 0;
vc->gp_list = kmalloc(sizeof(char) * vc->vms, GFP_KERNEL);
vc->num_gp = 0;
vc->io_client = dm_io_client_create();
vc->gs = NULL;
vc->overload = 0;
for(i=0; i<vc->vms; i++)
vc->gp_list[i] = Clean_Weight;//0 is clean
vc->gp_list[vc->wp] = Writing_Weight;
{
unsigned long long tem, disk_size;
tem = 0;
for(i = 0; i<vms; i++){
struct block_device *cur_bdev = vc->vm[i].dev->bdev;
vc->vm[i].end_sector = i_size_read(cur_bdev->bd_inode)>>9;//unit of sector
printk("vm%llu start_sector %llu, end_sector %llu, target_offset %llu\n",
i, (unsigned long long) vc->vm[i].physical_start, (unsigned long long) vc->vm[i].end_sector, (unsigned long long)dm_target_offset(ti, vc->ws[i]));
disk_size = vc->vm[i].end_sector * 512;
do_div(disk_size, (unsigned long long) vc->vm[i].dev->bdev->bd_block_size);
tem += disk_size;
}
vc->num_entry = tem;//num entry is blk num
}
printk("num entry is %llu, node size is %lu, req mem is %llu\n", vc->num_entry, sizeof(struct flag_nodes), sizeof(struct flag_nodes) * vc->num_entry);
//flag set initialize
vc->fs = (struct flag_set *) kmalloc(sizeof(struct flag_set), GFP_KERNEL);
vc->fs->node_buf = kmem_cache_create("dirty_data_buf", sizeof(struct flag_nodes),
0, (SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD), NULL);
vc->fs->table = (struct flag_nodes **)vmalloc(sizeof(struct flag_nodes*) * vc->num_entry);
for(i=0; i<vc->num_entry; i++){
//vc->fs->table[i] = NULL;//late alloc code
vc->fs->table[i] = kmem_cache_alloc(vc->fs->node_buf, GFP_KERNEL);//pre alloc start
vc->fs->table[i]->msector = -1;
vc->fs->table[i]->wp = -1;//pre alloc end
if(KIZIL_POLICY) vc->fs->table[i]->num_moved = 0;
}
vc->num_map_block = 0;//vc->num_entry * sizeof(struct flag_nodes) / 4096;
//vc->ws[0] += vc->num_map_block;
vc->fs->reverse_table = vmalloc(sizeof(struct reverse_nodes*) * vc->vms);
vc->d_num = kmalloc(sizeof(unsigned long long) * vc->vms, GFP_KERNEL);
for(i=0; i<vc->vms; i++){
unsigned long long j;
unsigned long long r_table_size = (vc->vm[i].end_sector + 7);
unsigned long long phy_sect = vc->vm[i].physical_start;
do_div(phy_sect, 8);
do_div(r_table_size, 8);
printk("r_table_size = %llu\n", r_table_size);
vc->vm[i].num_dirty = r_table_size - phy_sect;
vc->d_num[i] = vc->vm[i].num_dirty;
vc->fs->reverse_table[i] = vmalloc(sizeof(struct reverse_nodes) * r_table_size);
for(j=0; j<r_table_size; j++){
vc->fs->reverse_table[i][j].index = -1;
vc->fs->reverse_table[i][j].dirty = 1;
}
//printk("%u's first ptr is %p, final ptr is %p\n", i, &(vc->fs->reverse_table[i][0]), &(vc->fs->reverse_table[i][j]));
}
for(i=0; i<vc->vms; i++){
unsigned int minor = atom(vc->vm[i].dev->name);
unsigned int major = atoj(vc->vm[i].dev->name);
printk("dev name is %s\t", vc->vm[i].dev->name);
if(major != 2600) vc->vm[i].main_dev = minor >> minor_shift;
else vc->vm[i].main_dev = minor - 1;
vc->vm[i].maj_dev = major;
printk("main %u, maj %u\n", vc->vm[i].main_dev, vc->vm[i].maj_dev);
}
vc->mig_flag = 0;
mutex_init(&vc->lock);
mutex_init(&vc->gc_lock);
if(KIZIL_POLICY){
unsigned int i, j, r_size, l_size, gen;
r_size = vc->vms; l_size = vc->vms;
j = 0; gen = 0;
while(l_size){
if(l_size/2>0)
r_size = l_size/2;
else
r_size = 1;
//l_size/2>0 ? r_size = l_size/2 : r_size = 1;
//printk("l_size %u, r_size %u, gen %u, j %u\n", l_size, r_size, gen, j);
for(i=0; i<r_size; i++){
vc->vm[j].gen = gen;
j++;
}
gen++;
l_size-= r_size;
}
for(i=0; i<vc->vms; i++)
printk("%u's gen is %u\t", i, vc->vm[i].gen);
printk("\n");
}
else vc->vm[i].gen = 0;
ti->private = vc;
vc->gs = kmalloc(sizeof(struct gc_set) * gc_buffer_size, GFP_KERNEL);
for(i=0; i<gc_buffer_size; i++){
vc->gs[i].set_num = i;
vc->gs[i].ti = ti;
vc->gs[i].gc_lock = &vc->gc_lock;
vc->gs[i].kijil_map = NULL;
vc->gs[i].bs = kmalloc(sizeof(struct buf_set), GFP_KERNEL);
vc->gs[i].bs->buf = vmalloc(4096*127);
vc->gs[i].r_id = kthread_run((void*)read_job, &vc->gs[i], "read_th");
vc->gs[i].w_id = kthread_run((void*)write_job, &vc->gs[i], "write_th");
vc->gs[i].phase_flag = -1;
}
vc->gc_flag = 0;
ti->discards_supported = true;
return 0;
}
static void vm_dtr(struct dm_target *ti)
{
unsigned int i;
struct vm_c *vc = (struct vm_c *) ti->private;
for (i = 0; i < vc->vms; i++)
dm_put_device(ti, vc->vm[i].dev);
flush_work(&vc->trigger_event);
/*for(i=0;i<vc->vms;i++){
temp = vc->fs->wp[i];
if(temp == NULL) continue;
do{
struct flag_node* del_node = temp;
temp = temp->next;
flag_erase(vc->fs, del_node);
}while(temp->next != NULL);
vc->fs->wp[i] = NULL;
}*/
/*if(vc->th_id[0]){
kthread_stop(vc->th_id[0]);
vc->th_id[0] = NULL;
}
if(vc->th_id[1]){
kthread_stop(vc->th_id[1]);
vc->th_id[1] = NULL;
}*/
/*if(vc->th_id){
kthread_stop(vc->th_id);
vc->th_id = NULL;
}*/
/*if(vc->th_id){
kthread_stop(vc->th_id);
vc->th_id = NULL;
}
if(vc->gs){
kfree(vc->gs);
vc->gs = NULL;
}*/
vfree(vc->fs->table);
kfree(vc);
}
inline int do_kijil(struct vm_c* vc, int gp){
unsigned long long disk_block_size = vc->vm[gp].end_sector+7;// - vc->vm[gp].physical_start;//initialize for do_div
signed char num_count = 0;
unsigned long long i = 0;
//unsigned long long j, remainder;
//unsigned long long temp;
char* kijil_map = vmalloc(disk_block_size);
struct reverse_nodes* gp_reverse_table = vc->fs->reverse_table[gp];
int kijil_size = 0;
do_div(disk_block_size, 8);
///why kijil grain is 1 byte?? more coars grain??
//printk("rv table print start\n");
//for(i=0; i<disk_block_size; i++){
// printk("%llu:%u ", i, gp_reverse_table[i].dirty);
// if(i !=0 && i%30 == 0)
// printk("\n");
//}
//printk("rv table print end\n");
/*i = vc->vm[gp].physical_start;
do_div(i, 8);
j=i; remainder = do_div(j, 127);
printk("physical start 'j' is %llu\n", j);
for(temp = 0;temp < j; temp++){
kijil_map[kijil_size] = -127;
kijil_size++;
}*/
if(gp_reverse_table[i].dirty == 0) num_count = 1;
else if(gp_reverse_table[i].dirty == 1) num_count = -1;
for(i=1; i<disk_block_size; i++){///already check 0 index, modified to (i=0)
if(num_count > 0){
if(num_count == 127){//range over
kijil_map[kijil_size] = num_count;
kijil_size++;
num_count = 0;
}
if(gp_reverse_table[i].dirty == 0) num_count++; //continuous valid blk
else{//valid is end
kijil_map[kijil_size] = num_count;
kijil_size++;
num_count = -1;
}
}
else if(num_count < 0){
if(num_count == -127){//range over
kijil_map[kijil_size] = num_count;//recording count
kijil_size++;
num_count = 0;
}
if(gp_reverse_table[i].dirty == 1) num_count--;//continuous invalid blk
else{//invalid is end
kijil_map[kijil_size] = num_count;
kijil_size++;
num_count = 1;
}
}
else printk("unknown else error\n");
}
kijil_map[kijil_size++] = num_count;
/*kijil_size = 0;
for(i=0; i<disk_block_size; i++){
//if(gp_reverse_table[i].size == -1){
// printk("??in kijil, size error...sector %llu\n", i);
//}
printk("sector dirty %u, sector size %u, kijil_size %d, sector %llu, disk_size %llu\n", gp_reverse_table[i].dirty,
gp_reverse_table[i].size, kijil_size, i, disk_block_size);
if(gp_reverse_table[i].dirty == 0){//this is valid data
kijil_map[kijil_size] = gp_reverse_table[i].size;
}
else{//this is dirty data
kijil_map[kijil_size] = -(gp_reverse_table[i].size);
}
kijil_size++;
i+= gp_reverse_table[i].size;
}//*/
//printk("kijil_loop_end\n");
//end doing kijil
//printk("\n");
//for(i=0; i<kijil_size; i++){//Printing kijil_map
// if(kijil_map[i] <0)
// printk("%llu:%d ", i, kijil_map[i]);
// else
// printk("%llu:+%d ", i, kijil_map[i]);
// if(i != 0 && i%30 == 0)
// printk("\n");
//}
//printk("kijil_map print end\n");
vc->gs[0].kijil_map = vmalloc(kijil_size);
memcpy(vc->gs[0].kijil_map, kijil_map, kijil_size);
for(i=1; i<gc_buffer_size; i++)
vc->gs[i].kijil_map = vc->gs[0].kijil_map;
vfree(kijil_map); kijil_map = NULL;
return kijil_size;
}
inline char point_targeting(struct vm_c *vc, int *r_tp, int *r_gp){//r_tp, r_gp is return_tp, return gp
unsigned int tp, i, wp_main_dev, min, min_weight;
unsigned int wp_maj_dev;
unsigned long long percent_of_dirtied;// = vc->vm[gp].num_dirty - vc->d_num[gp] * 100;/////thisthis
printk("pt targeting \n");
if(*r_gp == -1){
for(i=0; i<vc->vms; i++){
if(vc->gp_list[i] == GC_Weight){
*r_gp = i;
break;
}
else if(i == vc->vms-1 && vc->gp_list[i] != GC_Weight){
printk("no existence gc ssd\n");
vc->mig_flag = 0;
return 0;
}
}
}
tp = *r_gp;
wp_main_dev = vc->vm[vc->wp].main_dev;
wp_maj_dev = vc->vm[vc->wp].maj_dev;
min = tp; min_weight = -1;
printk("for...\n");
for(i=0; i<vc->vms; i++){
unsigned weight = 0;
percent_of_dirtied = (vc->vm[i].num_dirty - vc->d_num[i]) * 100;
do_div(percent_of_dirtied, vc->vm[i].num_dirty);
//printk("%u's valid ratio is %llu(%llu)\t", i, percent_of_dirtied, vc->d_num[i]);
printk("%u's(%u) valid ratio is %llu(%llu) ", i, vc->gp_list[i], percent_of_dirtied, vc->d_num[i]);
tp = (tp + 1) % vc->vms;
if(vc->vm[tp].maj_dev == wp_maj_dev && vc->vm[tp].main_dev == wp_main_dev)
weight = 5;
weight += vc->gp_list[tp];
printk("weight 1?\n");
if(weight == 1){
unsigned long long percent_of_ptr_location;
percent_of_ptr_location = (vc->ws[tp] + vc->vm[tp].physical_start) * 100;
do_div(percent_of_ptr_location, vc->vm[tp].end_sector);
if(percent_of_dirtied + percent_of_ptr_location < 100){
min = tp;
break;
}
}
printk("weight == writed_weight\n");
if(weight == Writed_Weight){
unsigned long long percent_of_ptr_location;
percent_of_ptr_location = (vc->ws[tp] + vc->vm[tp].physical_start) * 100;
do_div(percent_of_ptr_location, vc->vm[tp].end_sector);
if(percent_of_dirtied + percent_of_ptr_location < 100){
min = tp;
break;
}
}
printk("min>weight?\n");
if(min_weight > weight){//search target device by minimal weight
min = tp;
min_weight = weight;
wp_maj_dev = vc->vm[min].maj_dev;
wp_main_dev = vc->vm[min].main_dev;
}
weight = 0;
}
printk("\n");
//printk("in ptr targeting, gp_list ++ %u\n", vc->gp_list[min]);
*r_tp = min;
vc->gp_list[min] = Targeted_Weight;///target pointer is 2
return 1;
}
inline char weathering_check(struct vm_c *vc){
if(vc->num_gp >= 1){
unsigned int i;
unsigned int tp = 0, gp = -1;
unsigned long long kijil_size = 1;
unsigned long long percent_of_dirtied = 0;
unsigned int min = 0;
unsigned int min_percent = 100;
/*
printk("weight cal in weathering\n");
for(i=0; i<vc->vms; i++){
if(vc->gp_list[i] != GC_Weight) continue;
percent_of_dirtied = (vc->vm[i].num_dirty - vc->d_num[i]) * 100;
do_div(percent_of_dirtied, vc->vm[i].num_dirty);
if(percent_of_dirtied < min_percent){
min = i; min_percent = percent_of_dirtied;
}
}//switched sequence (select minimum gp alg.) for aging or debugging
gp = min;
printk("weathering start\n");
if(point_targeting(vc, &tp, &gp) == 0){
//if(false){
Return_Weight();
return 0;
}
printk(" minp > Wratio? %d\n", min_percent);
if(min_percent > WEATHERING_RATIO){
//if(vc->vms - vc->num_gp <= 2){
if(false){///for debug
Return_Weight();
if(vc->gp_list[min] != GC_Weight)
return 0;
gp = min;/////////////////????why...?
if(point_targeting(vc, &tp, &gp) == 0){
Return_Weight();
return 0;
}
}
else{
printk("not yet\n");
Return_Weight();
return 0;
}
}
else{
printk("weathering success\n");
}
printk("dirty ratio ??\n");
percent_of_dirtied = (vc->vm[gp].num_dirty - vc->d_num[gp]) * 100;
do_div(percent_of_dirtied, vc->vm[gp].num_dirty);
printk("dirty_ratio is %llu\n", percent_of_dirtied);
printk("mig is start\n");
for(i=0; i<gc_buffer_size; i++){
vc->gs[i].ptr_ovflw_size = 0;
vc->gs[i].tp_io_sector = 0;
vc->gs[i].tp_table_size = 0;
vc->gs[i].tp = tp;
vc->gs[i].gp = gp;
//vc->gs[i].kijil_size = kijil_size;
vc->gs[i].phase_flag = -1;
}
vc->read_index = 0;
//vc->cur_sector = vc->vm[gp].physical_start;
vc->cur_sector = 0;
printk("gp_count %u, gp %u, tp %u\n", vc->num_gp, gp, tp);
return 1;
*/
}
return 0;
}
inline void map_store(struct vm_c *vc){
struct dm_io_region io;
struct dm_io_request io_req;
sector_t map_ptr = 0;
char* buf_for_store = vmalloc(4096);
struct flag_nodes *table = vmalloc(sizeof(struct flag_nodes) * num_flag_per_page);
table[0].msector = 52;
table[0].wp = 0;
table[1].msector = 33;
table[1].wp = 0;
table[2].msector = 21;
table[2].wp = 1;
memcpy(buf_for_store, (char*) table, 4096);
io_req.bi_op = REQ_OP_WRITE; io_req.mem.type = DM_IO_VMA;
io_req.mem.ptr.vma = buf_for_store;
io_req.client = vc->io_client;
io_req.notify.fn = NULL;
io.bdev = vc->vm[0].dev->bdev;
io.sector = vc->vm[0].physical_start + map_ptr * 8;
io.count = 8;
printk("store start\n");
dm_io(&io_req, 1, &io, NULL);
printk("store end\n");
vfree(buf_for_store);
buf_for_store = vmalloc(4096);
io_req.bi_op = REQ_OP_READ; io_req.mem.type = DM_IO_VMA;
io_req.mem.ptr.vma = buf_for_store;
io_req.client = vc->io_client;
io_req.notify.fn = NULL;
io.bdev = vc->vm[0].dev->bdev;
io.sector = vc->vm[0].physical_start + map_ptr * 8;
io.count = 8;
printk("load start\n");
dm_io(&io_req, 1, &io, NULL);
printk("load end\n");
printk("0 sector %llu, wp %u, 1 sector %llu, wp %u, 2 sector %llu, wp %u\n",
(unsigned long long) table[0].msector, table[0].wp, (unsigned long long) table[1].msector, table[1].wp, (unsigned long long) table[2].msector, table[2].wp);
}
static int write_job(struct gc_set* gs){
struct dm_target *ti = gs->ti;
struct vm_c *vc = ti->private;
unsigned long long write_index, cur_sector;
struct dm_io_region io;
struct dm_io_request io_req;
unsigned int i, size;
struct reverse_nodes* tp_reverse_table = NULL;
struct reverse_nodes* gp_reverse_table = NULL;
io_req.bi_op = REQ_OP_WRITE; io_req.mem.type = DM_IO_VMA;
io_req.mem.ptr.vma = gs->bs->buf; io_req.notify.fn = NULL;
io_req.client = vc->io_client;
while(1){
if(vc->mig_flag == 1){
//printk("gs check\n");
//if(gs->kijil_map != NULL){//outer gc is now started.
if(vc->gc_flag & 1){//outer gc is now started.
//printk("ogc is start!\n");
tp_reverse_table = vc->fs->reverse_table[gs->tp];
gp_reverse_table = vc->fs->reverse_table[gs->gp];
write_index = 0;
cur_sector = vc->vm[gs->tp].physical_start;
while(1){//...this condition is ... may have problem...
if(gs->phase_flag == 1){//write is able.
struct buf_set *c_bs = gs->bs;
unsigned long long g_tis;
//size = gs->kijil_map[c_bs->index];
size = c_bs->size;
//cur_sector = c_bs->sector;// - vc->vm[gs->gp].physical_start;
gs->tp_table_size = vc->vm[gs->tp].end_sector + 7;
//do_div(cur_sector, 8);
mutex_lock(&vc->lock);{//modified reverse_table information
gs->tp_io_sector = vc->ws[gs->tp] + vc->vm[gs->tp].physical_start;
g_tis = vc->ws[gs->tp];
do_div(g_tis, 8);
for(i=0; i<size; i++){
//unsigned int j;
//unsigned int next_tp = (gs->tp+1) % vc->vms;
if(vc->ws[gs->tp] + vc->vm[gs->tp].physical_start + 8 > vc->vm[gs->tp].end_sector){
unsigned int next_point, gp_main_dev, gp_maj_dev, min, min_weight, weight;
vc->gp_list[gs->tp] = GC_Weight;
vc->num_gp++;
next_point = gs->tp;
gp_main_dev = vc->vm[gs->tp].main_dev;
gp_maj_dev = vc->vm[gs->tp].maj_dev;
min = next_point; min_weight = -1;
weight = 0;
for(i = 0; i < vc->vms; i++){
next_point = (next_point + 1) % vc->vms;
if(vc->vm[next_point].maj_dev == gp_maj_dev && vc->vm[next_point].main_dev == gp_main_dev)
weight = 5;
weight+= vc->gp_list[next_point];
if(min_weight > weight){
min = next_point;
min_weight = weight;
gp_maj_dev = vc->vm[min].maj_dev;
gp_main_dev = vc->vm[min].main_dev;
}
weight = 0;
}
if(min_weight != 0) vc->overload = 1;
vc->gp_list[min] = Targeted_Weight;
printk("over flow!!! next tp is %d, %s\n", min, vc->vm[min].dev->name);
gs->tp = min;
if(vc->mig_flag == 0) vc->mig_flag = 1;
break;
}
tp_reverse_table[g_tis + i].index = gp_reverse_table[c_bs->index + i].index;
tp_reverse_table[g_tis + i].dirty = gp_reverse_table[c_bs->index + i].dirty;
vc->ws[gs->tp] += 8;
if(g_tis+i < vc->vm[gs->tp].end_sector + 7 &&
c_bs->index + i < vc->vm[gs->gp].end_sector + 7){
if(tp_reverse_table[g_tis+i].index != -1){
if(tp_reverse_table[g_tis+i].dirty == 0){//if it is valid data
//printk("gp dirty %u, tp dirty %u, gp index %llu, tp index %llu, gp msector %llu, tp msector %llu\n",
vc->d_num[gs->tp]--;////valid data must sub d_num;
}
//printk("%llu's gp dirty %u\n", c_bs->index + i, gp_reverse_table[c_bs->index+i].dirty);
vc->fs->table[tp_reverse_table[g_tis + i].index]->msector = -1;
vc->fs->table[tp_reverse_table[g_tis + i].index]->wp = -1;
gp_reverse_table[c_bs->index + i].index = -1;
gp_reverse_table[c_bs->index + i].dirty = 1;
}
}
}
}mutex_unlock(&vc->lock);
if(KIZIL_POLICY) io.bdev = vc->vm[gs->bs->target_wp].dev->bdev;
else io.bdev = vc->vm[gs->tp].dev->bdev;///need to modify
//if overflow occur, then tp is need to change
io.sector = gs->tp_io_sector;
io.count = (c_bs->size * 8) - gs->ptr_ovflw_size;//(gs->kijil_map[c_bs->index] - gs->ptr_ovflw_size) * 8;
//printk("%d's write index %llu, sector %llu, real sector %llu, size %llu\n", gs->set_num, c_bs->index, c_bs->sector, (unsigned long long)io.sector, (unsigned long long)io.count);
dm_io(&io_req, 1, &io, NULL);
//sync io is finished.
size-= gs->ptr_ovflw_size;
if(size != 0){
for(i=0; i<size; i++){
struct reverse_nodes *rn;
if(g_tis + i > gs->tp_table_size) break;
rn = &(tp_reverse_table[g_tis + i]);
if(rn->index == -1)//size -1 is a linked block
continue;//index -1 is a non writed sector
mutex_lock(&vc->lock);//is this overhead??
vc->fs->table[rn->index]->msector = gs->tp_io_sector + (i * 8);
vc->fs->table[rn->index]->wp = gs->tp;
mutex_unlock(&vc->lock);
if(KIZIL_POLICY && vc->gp_list[gs->bs->target_wp] == Writed_Weight)
vc->gp_list[gs->bs->target_wp] -= Writed_Weight;
}
}
if(gs->ptr_ovflw_size != 0){//need to verify...
unsigned int j, next_tp;
printk("oGC's write overflow occur\n");
gs->kijil_map[c_bs->index] = size;
gs->ptr_ovflw_size = 0;
mutex_lock(&vc->lock);
next_tp = (gs->tp + 1) % vc->vms;//need to apply point targeting algorithms
for(j=0; j<gc_buffer_size; j++)
vc->gs[j].tp = next_tp;
mutex_unlock(&vc->lock);
}
///judge to next operation
if(!(vc->gc_flag & 2)){///all read job is not end.
gs->phase_flag = 0;//this operations means ready to read job
}
else{//read job is end..
//printk("0. %d's write job is end\n", gs->set_num);
if(gs->phase_flag != -2)
gs->phase_flag = -2;
break;//my write job is end
}
///////if phase_flag == -2, then all thread's read job is end.
//////therefore my write job is endest write job.
//////if phase_flag != -2, then read job is not end. therefore continue for Outer GC's read job
}
else if(gs->phase_flag == -2) break;
else{//holding!!
msleep(1);
}
}
///this code section is escape loop(write is finished).
//therefore can trimming SSD
{//this section is waiting for all write job is end.
char wait_flag = 1;
if(gs->set_num == 0){
//printk("0's wait start\n");
while(wait_flag){//if wait_flag >= 1, infinite loop
msleep(5);
for(i=0; i<gc_buffer_size; i++){
if(vc->gs[i].phase_flag == -2){
//printk("end set is %d\n", i);
wait_flag+= 1;// flag is at least one of not ended-SSD is exist.
}
}
/*if(i == gc_buffer_size && vc->gs[gc_buffer_size-1].phase_flag == -2){
//is a all -2
wait_flag = 0;
}*/
if(wait_flag != 1+gc_buffer_size) wait_flag = 1;
else wait_flag = 0;
}
vc->gc_flag |= 4;
}
}
//if(gs->set_num == 0){
// printk("end set check start\n");
// for(i=0; i<gc_buffer_size; i++){
// if(vc->gs[i].phase_flag != -2){
// printk("not end set is %d\t", i);
// }
// }
// printk("\n");
//}
//and... we discard all data in GC SSD
if(vc->gc_flag & 4 && gs->set_num == 0){//TRIM command perform only 0 GC set.
//io_req.bi_rw = REQ_WRITE | REQ_DISCARD;
//io_req.mem.ptr.vma = gs->bs->buf;
//io.bdev = vc->vm[gs->gp].dev->bdev;
//io.sector = vc->vm[gs->gp].physical_start;
//io.count = vc->vm[gs->gp].end_sector - 1 - vc->vm[gs->gp].physical_start;
//dm_io(&io_req, 1, &io, NULL);////discard by DM_IO
blkdev_issue_discard(vc->vm[gs->gp].dev->bdev, vc->vm[gs->gp].physical_start,
vc->vm[gs->gp].end_sector - 1 - vc->vm[gs->gp].physical_start, GFP_NOFS, 0);
//discard is finished.
printk("dirty_num is %llu\n", vc->d_num[gs->gp]);