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
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
|
/* $OpenBSD: uvm_aobj.c,v 1.39 2009/05/08 13:50:15 ariane Exp $ */
/* $NetBSD: uvm_aobj.c,v 1.39 2001/02/18 21:19:08 chs Exp $ */
/*
* Copyright (c) 1998 Chuck Silvers, Charles D. Cranor and
* Washington University.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by Charles D. Cranor and
* Washington University.
* 4. The name of the author may not be used to endorse or promote products
* derived from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
* IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
* THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*
* from: Id: uvm_aobj.c,v 1.1.2.5 1998/02/06 05:14:38 chs Exp
*/
/*
* uvm_aobj.c: anonymous memory uvm_object pager
*
* author: Chuck Silvers <chuq@chuq.com>
* started: Jan-1998
*
* - design mostly from Chuck Cranor
*/
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/proc.h>
#include <sys/malloc.h>
#include <sys/kernel.h>
#include <sys/pool.h>
#include <sys/kernel.h>
#include <uvm/uvm.h>
/*
* an aobj manages anonymous-memory backed uvm_objects. in addition
* to keeping the list of resident pages, it also keeps a list of
* allocated swap blocks. depending on the size of the aobj this list
* of allocated swap blocks is either stored in an array (small objects)
* or in a hash table (large objects).
*/
/*
* local structures
*/
/*
* for hash tables, we break the address space of the aobj into blocks
* of UAO_SWHASH_CLUSTER_SIZE pages. we require the cluster size to
* be a power of two.
*/
#define UAO_SWHASH_CLUSTER_SHIFT 4
#define UAO_SWHASH_CLUSTER_SIZE (1 << UAO_SWHASH_CLUSTER_SHIFT)
/* get the "tag" for this page index */
#define UAO_SWHASH_ELT_TAG(PAGEIDX) \
((PAGEIDX) >> UAO_SWHASH_CLUSTER_SHIFT)
/* given an ELT and a page index, find the swap slot */
#define UAO_SWHASH_ELT_PAGESLOT(ELT, PAGEIDX) \
((ELT)->slots[(PAGEIDX) & (UAO_SWHASH_CLUSTER_SIZE - 1)])
/* given an ELT, return its pageidx base */
#define UAO_SWHASH_ELT_PAGEIDX_BASE(ELT) \
((ELT)->tag << UAO_SWHASH_CLUSTER_SHIFT)
/*
* the swhash hash function
*/
#define UAO_SWHASH_HASH(AOBJ, PAGEIDX) \
(&(AOBJ)->u_swhash[(((PAGEIDX) >> UAO_SWHASH_CLUSTER_SHIFT) \
& (AOBJ)->u_swhashmask)])
/*
* the swhash threshold determines if we will use an array or a
* hash table to store the list of allocated swap blocks.
*/
#define UAO_SWHASH_THRESHOLD (UAO_SWHASH_CLUSTER_SIZE * 4)
#define UAO_USES_SWHASH(AOBJ) \
((AOBJ)->u_pages > UAO_SWHASH_THRESHOLD) /* use hash? */
/*
* the number of buckets in a swhash, with an upper bound
*/
#define UAO_SWHASH_MAXBUCKETS 256
#define UAO_SWHASH_BUCKETS(AOBJ) \
(min((AOBJ)->u_pages >> UAO_SWHASH_CLUSTER_SHIFT, \
UAO_SWHASH_MAXBUCKETS))
/*
* uao_swhash_elt: when a hash table is being used, this structure defines
* the format of an entry in the bucket list.
*/
struct uao_swhash_elt {
LIST_ENTRY(uao_swhash_elt) list; /* the hash list */
voff_t tag; /* our 'tag' */
int count; /* our number of active slots */
int slots[UAO_SWHASH_CLUSTER_SIZE]; /* the slots */
};
/*
* uao_swhash: the swap hash table structure
*/
LIST_HEAD(uao_swhash, uao_swhash_elt);
/*
* uao_swhash_elt_pool: pool of uao_swhash_elt structures
*/
struct pool uao_swhash_elt_pool;
/*
* uvm_aobj: the actual anon-backed uvm_object
*
* => the uvm_object is at the top of the structure, this allows
* (struct uvm_aobj *) == (struct uvm_object *)
* => only one of u_swslots and u_swhash is used in any given aobj
*/
struct uvm_aobj {
struct uvm_object u_obj; /* has: lock, pgops, memq, #pages, #refs */
int u_pages; /* number of pages in entire object */
int u_flags; /* the flags (see uvm_aobj.h) */
int *u_swslots; /* array of offset->swapslot mappings */
/*
* hashtable of offset->swapslot mappings
* (u_swhash is an array of bucket heads)
*/
struct uao_swhash *u_swhash;
u_long u_swhashmask; /* mask for hashtable */
LIST_ENTRY(uvm_aobj) u_list; /* global list of aobjs */
};
/*
* uvm_aobj_pool: pool of uvm_aobj structures
*/
struct pool uvm_aobj_pool;
/*
* local functions
*/
struct uao_swhash_elt *uao_find_swhash_elt(struct uvm_aobj *, int,
boolean_t);
static int uao_find_swslot(struct uvm_aobj *, int);
boolean_t uao_flush(struct uvm_object *, voff_t, voff_t, int);
void uao_free(struct uvm_aobj *);
int uao_get(struct uvm_object *, voff_t, vm_page_t *,
int *, int, vm_prot_t, int, int);
boolean_t uao_releasepg(struct vm_page *, struct vm_page **);
boolean_t uao_pagein(struct uvm_aobj *, int, int);
boolean_t uao_pagein_page(struct uvm_aobj *, int);
/*
* aobj_pager
*
* note that some functions (e.g. put) are handled elsewhere
*/
struct uvm_pagerops aobj_pager = {
NULL, /* init */
uao_reference, /* reference */
uao_detach, /* detach */
NULL, /* fault */
uao_flush, /* flush */
uao_get, /* get */
NULL, /* put (done by pagedaemon) */
NULL, /* cluster */
NULL, /* mk_pcluster */
uao_releasepg /* releasepg */
};
/*
* uao_list: global list of active aobjs, locked by uao_list_lock
*/
static LIST_HEAD(aobjlist, uvm_aobj) uao_list;
static simple_lock_data_t uao_list_lock;
/*
* functions
*/
/*
* hash table/array related functions
*/
/*
* uao_find_swhash_elt: find (or create) a hash table entry for a page
* offset.
*
* => the object should be locked by the caller
*/
struct uao_swhash_elt *
uao_find_swhash_elt(struct uvm_aobj *aobj, int pageidx, boolean_t create)
{
struct uao_swhash *swhash;
struct uao_swhash_elt *elt;
voff_t page_tag;
swhash = UAO_SWHASH_HASH(aobj, pageidx); /* first hash to get bucket */
page_tag = UAO_SWHASH_ELT_TAG(pageidx); /* tag to search for */
/*
* now search the bucket for the requested tag
*/
LIST_FOREACH(elt, swhash, list) {
if (elt->tag == page_tag)
return(elt);
}
/* fail now if we are not allowed to create a new entry in the bucket */
if (!create)
return NULL;
/*
* allocate a new entry for the bucket and init/insert it in
*/
elt = pool_get(&uao_swhash_elt_pool, PR_WAITOK | PR_ZERO);
LIST_INSERT_HEAD(swhash, elt, list);
elt->tag = page_tag;
return(elt);
}
/*
* uao_find_swslot: find the swap slot number for an aobj/pageidx
*
* => object must be locked by caller
*/
static __inline int
uao_find_swslot(struct uvm_aobj *aobj, int pageidx)
{
/*
* if noswap flag is set, then we never return a slot
*/
if (aobj->u_flags & UAO_FLAG_NOSWAP)
return(0);
/*
* if hashing, look in hash table.
*/
if (UAO_USES_SWHASH(aobj)) {
struct uao_swhash_elt *elt =
uao_find_swhash_elt(aobj, pageidx, FALSE);
if (elt)
return(UAO_SWHASH_ELT_PAGESLOT(elt, pageidx));
else
return(0);
}
/*
* otherwise, look in the array
*/
return(aobj->u_swslots[pageidx]);
}
/*
* uao_set_swslot: set the swap slot for a page in an aobj.
*
* => setting a slot to zero frees the slot
* => object must be locked by caller
*/
int
uao_set_swslot(struct uvm_object *uobj, int pageidx, int slot)
{
struct uvm_aobj *aobj = (struct uvm_aobj *)uobj;
int oldslot;
UVMHIST_FUNC("uao_set_swslot"); UVMHIST_CALLED(pdhist);
UVMHIST_LOG(pdhist, "aobj %p pageidx %ld slot %ld",
aobj, pageidx, slot, 0);
/*
* if noswap flag is set, then we can't set a slot
*/
if (aobj->u_flags & UAO_FLAG_NOSWAP) {
if (slot == 0)
return(0); /* a clear is ok */
/* but a set is not */
printf("uao_set_swslot: uobj = %p\n", uobj);
panic("uao_set_swslot: attempt to set a slot on a NOSWAP object");
}
/*
* are we using a hash table? if so, add it in the hash.
*/
if (UAO_USES_SWHASH(aobj)) {
/*
* Avoid allocating an entry just to free it again if
* the page had not swap slot in the first place, and
* we are freeing.
*/
struct uao_swhash_elt *elt =
uao_find_swhash_elt(aobj, pageidx, slot ? TRUE : FALSE);
if (elt == NULL) {
KASSERT(slot == 0);
return (0);
}
oldslot = UAO_SWHASH_ELT_PAGESLOT(elt, pageidx);
UAO_SWHASH_ELT_PAGESLOT(elt, pageidx) = slot;
/*
* now adjust the elt's reference counter and free it if we've
* dropped it to zero.
*/
/* an allocation? */
if (slot) {
if (oldslot == 0)
elt->count++;
} else { /* freeing slot ... */
if (oldslot) /* to be safe */
elt->count--;
if (elt->count == 0) {
LIST_REMOVE(elt, list);
pool_put(&uao_swhash_elt_pool, elt);
}
}
} else {
/* we are using an array */
oldslot = aobj->u_swslots[pageidx];
aobj->u_swslots[pageidx] = slot;
}
return (oldslot);
}
/*
* end of hash/array functions
*/
/*
* uao_free: free all resources held by an aobj, and then free the aobj
*
* => the aobj should be dead
*/
void
uao_free(struct uvm_aobj *aobj)
{
simple_unlock(&aobj->u_obj.vmobjlock);
if (UAO_USES_SWHASH(aobj)) {
int i, hashbuckets = aobj->u_swhashmask + 1;
/*
* free the swslots from each hash bucket,
* then the hash bucket, and finally the hash table itself.
*/
for (i = 0; i < hashbuckets; i++) {
struct uao_swhash_elt *elt, *next;
for (elt = LIST_FIRST(&aobj->u_swhash[i]);
elt != NULL;
elt = next) {
int j;
for (j = 0; j < UAO_SWHASH_CLUSTER_SIZE; j++) {
int slot = elt->slots[j];
if (slot == 0) {
continue;
}
uvm_swap_free(slot, 1);
/*
* this page is no longer
* only in swap.
*/
simple_lock(&uvm.swap_data_lock);
uvmexp.swpgonly--;
simple_unlock(&uvm.swap_data_lock);
}
next = LIST_NEXT(elt, list);
pool_put(&uao_swhash_elt_pool, elt);
}
}
free(aobj->u_swhash, M_UVMAOBJ);
} else {
int i;
/*
* free the array
*/
for (i = 0; i < aobj->u_pages; i++) {
int slot = aobj->u_swslots[i];
if (slot) {
uvm_swap_free(slot, 1);
/* this page is no longer only in swap. */
simple_lock(&uvm.swap_data_lock);
uvmexp.swpgonly--;
simple_unlock(&uvm.swap_data_lock);
}
}
free(aobj->u_swslots, M_UVMAOBJ);
}
/*
* finally free the aobj itself
*/
pool_put(&uvm_aobj_pool, aobj);
}
/*
* pager functions
*/
/*
* uao_create: create an aobj of the given size and return its uvm_object.
*
* => for normal use, flags are always zero
* => for the kernel object, the flags are:
* UAO_FLAG_KERNOBJ - allocate the kernel object (can only happen once)
* UAO_FLAG_KERNSWAP - enable swapping of kernel object (" ")
*/
struct uvm_object *
uao_create(vsize_t size, int flags)
{
static struct uvm_aobj kernel_object_store; /* home of kernel_object */
static int kobj_alloced = 0; /* not allocated yet */
int pages = round_page(size) >> PAGE_SHIFT;
struct uvm_aobj *aobj;
/*
* malloc a new aobj unless we are asked for the kernel object
*/
if (flags & UAO_FLAG_KERNOBJ) { /* want kernel object? */
if (kobj_alloced)
panic("uao_create: kernel object already allocated");
aobj = &kernel_object_store;
aobj->u_pages = pages;
aobj->u_flags = UAO_FLAG_NOSWAP; /* no swap to start */
/* we are special, we never die */
aobj->u_obj.uo_refs = UVM_OBJ_KERN;
kobj_alloced = UAO_FLAG_KERNOBJ;
} else if (flags & UAO_FLAG_KERNSWAP) {
aobj = &kernel_object_store;
if (kobj_alloced != UAO_FLAG_KERNOBJ)
panic("uao_create: asked to enable swap on kernel object");
kobj_alloced = UAO_FLAG_KERNSWAP;
} else { /* normal object */
aobj = pool_get(&uvm_aobj_pool, PR_WAITOK);
aobj->u_pages = pages;
aobj->u_flags = 0; /* normal object */
aobj->u_obj.uo_refs = 1; /* start with 1 reference */
}
/*
* allocate hash/array if necessary
*
* note: in the KERNSWAP case no need to worry about locking since
* we are still booting we should be the only thread around.
*/
if (flags == 0 || (flags & UAO_FLAG_KERNSWAP) != 0) {
int mflags = (flags & UAO_FLAG_KERNSWAP) != 0 ?
M_NOWAIT : M_WAITOK;
/* allocate hash table or array depending on object size */
if (UAO_USES_SWHASH(aobj)) {
aobj->u_swhash = hashinit(UAO_SWHASH_BUCKETS(aobj),
M_UVMAOBJ, mflags, &aobj->u_swhashmask);
if (aobj->u_swhash == NULL)
panic("uao_create: hashinit swhash failed");
} else {
aobj->u_swslots = malloc(pages * sizeof(int),
M_UVMAOBJ, mflags|M_ZERO);
if (aobj->u_swslots == NULL)
panic("uao_create: malloc swslots failed");
}
if (flags) {
aobj->u_flags &= ~UAO_FLAG_NOSWAP; /* clear noswap */
return(&aobj->u_obj);
/* done! */
}
}
/*
* init aobj fields
*/
simple_lock_init(&aobj->u_obj.vmobjlock);
aobj->u_obj.pgops = &aobj_pager;
TAILQ_INIT(&aobj->u_obj.memq);
aobj->u_obj.uo_npages = 0;
/*
* now that aobj is ready, add it to the global list
*/
simple_lock(&uao_list_lock);
LIST_INSERT_HEAD(&uao_list, aobj, u_list);
simple_unlock(&uao_list_lock);
/*
* done!
*/
return(&aobj->u_obj);
}
/*
* uao_init: set up aobj pager subsystem
*
* => called at boot time from uvm_pager_init()
*/
void
uao_init(void)
{
static int uao_initialized;
if (uao_initialized)
return;
uao_initialized = TRUE;
LIST_INIT(&uao_list);
simple_lock_init(&uao_list_lock);
/*
* NOTE: Pages for this pool must not come from a pageable
* kernel map!
*/
pool_init(&uao_swhash_elt_pool, sizeof(struct uao_swhash_elt),
0, 0, 0, "uaoeltpl", &pool_allocator_nointr);
pool_init(&uvm_aobj_pool, sizeof(struct uvm_aobj), 0, 0, 0,
"aobjpl", &pool_allocator_nointr);
}
/*
* uao_reference: add a ref to an aobj
*
* => aobj must be unlocked
* => just lock it and call the locked version
*/
void
uao_reference(struct uvm_object *uobj)
{
simple_lock(&uobj->vmobjlock);
uao_reference_locked(uobj);
simple_unlock(&uobj->vmobjlock);
}
/*
* uao_reference_locked: add a ref to an aobj that is already locked
*
* => aobj must be locked
* this needs to be separate from the normal routine
* since sometimes we need to add a reference to an aobj when
* it's already locked.
*/
void
uao_reference_locked(struct uvm_object *uobj)
{
UVMHIST_FUNC("uao_reference"); UVMHIST_CALLED(maphist);
/*
* kernel_object already has plenty of references, leave it alone.
*/
if (UVM_OBJ_IS_KERN_OBJECT(uobj))
return;
uobj->uo_refs++; /* bump! */
UVMHIST_LOG(maphist, "<- done (uobj=%p, ref = %ld)",
uobj, uobj->uo_refs,0,0);
}
/*
* uao_detach: drop a reference to an aobj
*
* => aobj must be unlocked
* => just lock it and call the locked version
*/
void
uao_detach(struct uvm_object *uobj)
{
simple_lock(&uobj->vmobjlock);
uao_detach_locked(uobj);
}
/*
* uao_detach_locked: drop a reference to an aobj
*
* => aobj must be locked, and is unlocked (or freed) upon return.
* this needs to be separate from the normal routine
* since sometimes we need to detach from an aobj when
* it's already locked.
*/
void
uao_detach_locked(struct uvm_object *uobj)
{
struct uvm_aobj *aobj = (struct uvm_aobj *)uobj;
struct vm_page *pg;
UVMHIST_FUNC("uao_detach"); UVMHIST_CALLED(maphist);
/*
* detaching from kernel_object is a noop.
*/
if (UVM_OBJ_IS_KERN_OBJECT(uobj)) {
simple_unlock(&uobj->vmobjlock);
return;
}
UVMHIST_LOG(maphist," (uobj=%p) ref=%ld", uobj,uobj->uo_refs,0,0);
uobj->uo_refs--; /* drop ref! */
if (uobj->uo_refs) { /* still more refs? */
simple_unlock(&uobj->vmobjlock);
UVMHIST_LOG(maphist, "<- done (rc>0)", 0,0,0,0);
return;
}
/*
* remove the aobj from the global list.
*/
simple_lock(&uao_list_lock);
LIST_REMOVE(aobj, u_list);
simple_unlock(&uao_list_lock);
/*
* Free all pages left in the object. If they're busy, wait
* for them to become available before we kill it.
* Release swap resources then free the page.
*/
uvm_lock_pageq();
while((pg = TAILQ_FIRST(&uobj->memq)) != NULL) {
if (pg->pg_flags & PG_BUSY) {
atomic_setbits_int(&pg->pg_flags, PG_WANTED);
uvm_unlock_pageq();
UVM_UNLOCK_AND_WAIT(pg, &uobj->vmobjlock, 0,
"uao_det", 0);
simple_lock(&uobj->vmobjlock);
uvm_lock_pageq();
continue;
}
pmap_page_protect(pg, VM_PROT_NONE);
uao_dropswap(&aobj->u_obj, pg->offset >> PAGE_SHIFT);
uvm_pagefree(pg);
}
uvm_unlock_pageq();
/*
* finally, free the rest.
*/
uao_free(aobj);
}
/*
* uao_flush: "flush" pages out of a uvm object
*
* => object should be locked by caller. we may _unlock_ the object
* if (and only if) we need to clean a page (PGO_CLEANIT).
* XXXJRT Currently, however, we don't. In the case of cleaning
* XXXJRT a page, we simply just deactivate it. Should probably
* XXXJRT handle this better, in the future (although "flushing"
* XXXJRT anonymous memory isn't terribly important).
* => if PGO_CLEANIT is not set, then we will neither unlock the object
* or block.
* => if PGO_ALLPAGE is set, then all pages in the object are valid targets
* for flushing.
* => NOTE: we rely on the fact that the object's memq is a TAILQ and
* that new pages are inserted on the tail end of the list. thus,
* we can make a complete pass through the object in one go by starting
* at the head and working towards the tail (new pages are put in
* front of us).
* => NOTE: we are allowed to lock the page queues, so the caller
* must not be holding the lock on them [e.g. pagedaemon had
* better not call us with the queues locked]
* => we return TRUE unless we encountered some sort of I/O error
* XXXJRT currently never happens, as we never directly initiate
* XXXJRT I/O
*/
#define UAO_HASH_PENALTY 4 /* XXX: a guess */
boolean_t
uao_flush(struct uvm_object *uobj, voff_t start, voff_t stop, int flags)
{
struct uvm_aobj *aobj = (struct uvm_aobj *) uobj;
struct vm_page *pp;
voff_t curoff;
UVMHIST_FUNC("uao_flush"); UVMHIST_CALLED(maphist);
if (flags & PGO_ALLPAGES) {
start = 0;
stop = aobj->u_pages << PAGE_SHIFT;
} else {
start = trunc_page(start);
stop = round_page(stop);
if (stop > (aobj->u_pages << PAGE_SHIFT)) {
printf("uao_flush: strange, got an out of range "
"flush (fixed)\n");
stop = aobj->u_pages << PAGE_SHIFT;
}
}
UVMHIST_LOG(maphist, " flush start=0x%lx, stop=0x%lx, flags=0x%lx",
(u_long)start, (u_long)stop, flags, 0);
/*
* Don't need to do any work here if we're not freeing
* or deactivating pages.
*/
if ((flags & (PGO_DEACTIVATE|PGO_FREE)) == 0) {
UVMHIST_LOG(maphist,
"<- done (no work to do)",0,0,0,0);
return (TRUE);
}
/* locked: uobj */
curoff = start;
for (;;) {
if (curoff < stop) {
pp = uvm_pagelookup(uobj, curoff);
curoff += PAGE_SIZE;
if (pp == NULL)
continue;
} else {
break;
}
/* Make sure page is unbusy, else wait for it. */
if (pp->pg_flags & PG_BUSY) {
atomic_setbits_int(&pp->pg_flags, PG_WANTED);
UVM_UNLOCK_AND_WAIT(pp, &uobj->vmobjlock, 0,
"uaoflsh", 0);
simple_lock(&uobj->vmobjlock);
curoff -= PAGE_SIZE;
continue;
}
switch (flags & (PGO_CLEANIT|PGO_FREE|PGO_DEACTIVATE)) {
/*
* XXX In these first 3 cases, we always just
* XXX deactivate the page. We may want to
* XXX handle the different cases more specifically
* XXX in the future.
*/
case PGO_CLEANIT|PGO_FREE:
/* FALLTHROUGH */
case PGO_CLEANIT|PGO_DEACTIVATE:
/* FALLTHROUGH */
case PGO_DEACTIVATE:
deactivate_it:
/* skip the page if it's loaned or wired */
if (pp->loan_count != 0 ||
pp->wire_count != 0)
continue;
uvm_lock_pageq();
/* zap all mappings for the page. */
pmap_page_protect(pp, VM_PROT_NONE);
/* ...and deactivate the page. */
uvm_pagedeactivate(pp);
uvm_unlock_pageq();
continue;
case PGO_FREE:
/*
* If there are multiple references to
* the object, just deactivate the page.
*/
if (uobj->uo_refs > 1)
goto deactivate_it;
/* XXX skip the page if it's loaned or wired */
if (pp->loan_count != 0 ||
pp->wire_count != 0)
continue;
/* zap all mappings for the page. */
pmap_page_protect(pp, VM_PROT_NONE);
uao_dropswap(uobj, pp->offset >> PAGE_SHIFT);
uvm_lock_pageq();
uvm_pagefree(pp);
uvm_unlock_pageq();
continue;
default:
panic("uao_flush: weird flags");
}
}
UVMHIST_LOG(maphist,
"<- done, rv=%ld",retval,0,0,0);
return (TRUE);
}
/*
* uao_get: fetch me a page
*
* we have three cases:
* 1: page is resident -> just return the page.
* 2: page is zero-fill -> allocate a new page and zero it.
* 3: page is swapped out -> fetch the page from swap.
*
* cases 1 and 2 can be handled with PGO_LOCKED, case 3 cannot.
* so, if the "center" page hits case 3 (or any page, with PGO_ALLPAGES),
* then we will need to return VM_PAGER_UNLOCK.
*
* => prefer map unlocked (not required)
* => object must be locked! we will _unlock_ it before starting any I/O.
* => flags: PGO_ALLPAGES: get all of the pages
* PGO_LOCKED: fault data structures are locked
* => NOTE: offset is the offset of pps[0], _NOT_ pps[centeridx]
* => NOTE: caller must check for released pages!!
*/
int
uao_get(struct uvm_object *uobj, voff_t offset, struct vm_page **pps,
int *npagesp, int centeridx, vm_prot_t access_type, int advice, int flags)
{
struct uvm_aobj *aobj = (struct uvm_aobj *)uobj;
voff_t current_offset;
vm_page_t ptmp;
int lcv, gotpages, maxpages, swslot, rv, pageidx;
boolean_t done;
UVMHIST_FUNC("uao_get"); UVMHIST_CALLED(pdhist);
UVMHIST_LOG(pdhist, "aobj=%p offset=%ld, flags=%ld",
aobj, (u_long)offset, flags,0);
/*
* get number of pages
*/
maxpages = *npagesp;
/*
* step 1: handled the case where fault data structures are locked.
*/
if (flags & PGO_LOCKED) {
/*
* step 1a: get pages that are already resident. only do
* this if the data structures are locked (i.e. the first
* time through).
*/
done = TRUE; /* be optimistic */
gotpages = 0; /* # of pages we got so far */
for (lcv = 0, current_offset = offset ; lcv < maxpages ;
lcv++, current_offset += PAGE_SIZE) {
/* do we care about this page? if not, skip it */
if (pps[lcv] == PGO_DONTCARE)
continue;
ptmp = uvm_pagelookup(uobj, current_offset);
/*
* if page is new, attempt to allocate the page,
* zero-fill'd.
*/
if (ptmp == NULL && uao_find_swslot(aobj,
current_offset >> PAGE_SHIFT) == 0) {
ptmp = uvm_pagealloc(uobj, current_offset,
NULL, UVM_PGA_ZERO);
if (ptmp) {
/* new page */
atomic_clearbits_int(&ptmp->pg_flags,
PG_BUSY|PG_FAKE);
atomic_setbits_int(&ptmp->pg_flags,
PQ_AOBJ);
UVM_PAGE_OWN(ptmp, NULL);
}
}
/*
* to be useful must get a non-busy page
*/
if (ptmp == NULL ||
(ptmp->pg_flags & PG_BUSY) != 0) {
if (lcv == centeridx ||
(flags & PGO_ALLPAGES) != 0)
/* need to do a wait or I/O! */
done = FALSE;
continue;
}
/*
* useful page: busy/lock it and plug it in our
* result array
*/
/* caller must un-busy this page */
atomic_setbits_int(&ptmp->pg_flags, PG_BUSY);
UVM_PAGE_OWN(ptmp, "uao_get1");
pps[lcv] = ptmp;
gotpages++;
} /* "for" lcv loop */
/*
* step 1b: now we've either done everything needed or we
* to unlock and do some waiting or I/O.
*/
UVMHIST_LOG(pdhist, "<- done (done=%ld)", done, 0,0,0);
*npagesp = gotpages;
if (done)
/* bingo! */
return(VM_PAGER_OK);
else
/* EEK! Need to unlock and I/O */
return(VM_PAGER_UNLOCK);
}
/*
* step 2: get non-resident or busy pages.
* object is locked. data structures are unlocked.
*/
for (lcv = 0, current_offset = offset ; lcv < maxpages ;
lcv++, current_offset += PAGE_SIZE) {
/*
* - skip over pages we've already gotten or don't want
* - skip over pages we don't _have_ to get
*/
if (pps[lcv] != NULL ||
(lcv != centeridx && (flags & PGO_ALLPAGES) == 0))
continue;
pageidx = current_offset >> PAGE_SHIFT;
/*
* we have yet to locate the current page (pps[lcv]). we
* first look for a page that is already at the current offset.
* if we find a page, we check to see if it is busy or
* released. if that is the case, then we sleep on the page
* until it is no longer busy or released and repeat the lookup.
* if the page we found is neither busy nor released, then we
* busy it (so we own it) and plug it into pps[lcv]. this
* 'break's the following while loop and indicates we are
* ready to move on to the next page in the "lcv" loop above.
*
* if we exit the while loop with pps[lcv] still set to NULL,
* then it means that we allocated a new busy/fake/clean page
* ptmp in the object and we need to do I/O to fill in the data.
*/
/* top of "pps" while loop */
while (pps[lcv] == NULL) {
/* look for a resident page */
ptmp = uvm_pagelookup(uobj, current_offset);
/* not resident? allocate one now (if we can) */
if (ptmp == NULL) {
ptmp = uvm_pagealloc(uobj, current_offset,
NULL, 0);
/* out of RAM? */
if (ptmp == NULL) {
simple_unlock(&uobj->vmobjlock);
UVMHIST_LOG(pdhist,
"sleeping, ptmp == NULL\n",0,0,0,0);
uvm_wait("uao_getpage");
simple_lock(&uobj->vmobjlock);
/* goto top of pps while loop */
continue;
}
/*
* safe with PQ's unlocked: because we just
* alloc'd the page
*/
atomic_setbits_int(&ptmp->pg_flags, PQ_AOBJ);
/*
* got new page ready for I/O. break pps while
* loop. pps[lcv] is still NULL.
*/
break;
}
/* page is there, see if we need to wait on it */
if ((ptmp->pg_flags & PG_BUSY) != 0) {
atomic_setbits_int(&ptmp->pg_flags, PG_WANTED);
UVMHIST_LOG(pdhist,
"sleeping, ptmp->flags 0x%lx\n",
ptmp->pg_flags,0,0,0);
UVM_UNLOCK_AND_WAIT(ptmp, &uobj->vmobjlock,
FALSE, "uao_get", 0);
simple_lock(&uobj->vmobjlock);
continue; /* goto top of pps while loop */
}
/*
* if we get here then the page has become resident and
* unbusy between steps 1 and 2. we busy it now (so we
* own it) and set pps[lcv] (so that we exit the while
* loop).
*/
/* we own it, caller must un-busy */
atomic_setbits_int(&ptmp->pg_flags, PG_BUSY);
UVM_PAGE_OWN(ptmp, "uao_get2");
pps[lcv] = ptmp;
}
/*
* if we own the valid page at the correct offset, pps[lcv] will
* point to it. nothing more to do except go to the next page.
*/
if (pps[lcv])
continue; /* next lcv */
/*
* we have a "fake/busy/clean" page that we just allocated.
* do the needed "i/o", either reading from swap or zeroing.
*/
swslot = uao_find_swslot(aobj, pageidx);
/*
* just zero the page if there's nothing in swap.
*/
if (swslot == 0) {
/*
* page hasn't existed before, just zero it.
*/
uvm_pagezero(ptmp);
} else {
UVMHIST_LOG(pdhist, "pagein from swslot %ld",
swslot, 0,0,0);
/*
* page in the swapped-out page.
* unlock object for i/o, relock when done.
*/
simple_unlock(&uobj->vmobjlock);
rv = uvm_swap_get(ptmp, swslot, PGO_SYNCIO);
simple_lock(&uobj->vmobjlock);
/*
* I/O done. check for errors.
*/
if (rv != VM_PAGER_OK)
{
UVMHIST_LOG(pdhist, "<- done (error=%ld)",
rv,0,0,0);
if (ptmp->pg_flags & PG_WANTED)
wakeup(ptmp);
/*
* remove the swap slot from the aobj
* and mark the aobj as having no real slot.
* don't free the swap slot, thus preventing
* it from being used again.
*/
swslot = uao_set_swslot(&aobj->u_obj, pageidx,
SWSLOT_BAD);
uvm_swap_markbad(swslot, 1);
atomic_clearbits_int(&ptmp->pg_flags,
PG_WANTED|PG_BUSY);
UVM_PAGE_OWN(ptmp, NULL);
uvm_lock_pageq();
uvm_pagefree(ptmp);
uvm_unlock_pageq();
simple_unlock(&uobj->vmobjlock);
return (rv);
}
}
/*
* we got the page! clear the fake flag (indicates valid
* data now in page) and plug into our result array. note
* that page is still busy.
*
* it is the callers job to:
* => check if the page is released
* => unbusy the page
* => activate the page
*/
/* data is valid ... */
atomic_clearbits_int(&ptmp->pg_flags, PG_FAKE);
pmap_clear_modify(ptmp); /* ... and clean */
pps[lcv] = ptmp;
} /* lcv loop */
/*
* finally, unlock object and return.
*/
simple_unlock(&uobj->vmobjlock);
UVMHIST_LOG(pdhist, "<- done (OK)",0,0,0,0);
return(VM_PAGER_OK);
}
/*
* uao_releasepg: handle released page in an aobj
*
* => "pg" is a PG_BUSY [caller owns it], PG_RELEASED page that we need
* to dispose of.
* => caller must handle PG_WANTED case
* => called with page's object locked, pageq's unlocked
* => returns TRUE if page's object is still alive, FALSE if we
* killed the page's object. if we return TRUE, then we
* return with the object locked.
* => if (nextpgp != NULL) => we return the next page on the queue, and return
* with the page queues locked [for pagedaemon]
* => if (nextpgp == NULL) => we return with page queues unlocked [normal case]
* => we kill the aobj if it is not referenced and we are suppose to
* kill it ("KILLME").
*/
boolean_t
uao_releasepg(struct vm_page *pg, struct vm_page **nextpgp /* OUT */)
{
struct uvm_aobj *aobj = (struct uvm_aobj *) pg->uobject;
KASSERT(pg->pg_flags & PG_RELEASED);
/*
* dispose of the page [caller handles PG_WANTED] and swap slot.
*/
pmap_page_protect(pg, VM_PROT_NONE);
uao_dropswap(&aobj->u_obj, pg->offset >> PAGE_SHIFT);
uvm_lock_pageq();
if (nextpgp)
*nextpgp = TAILQ_NEXT(pg, pageq); /* next page for daemon */
uvm_pagefree(pg);
if (!nextpgp)
uvm_unlock_pageq(); /* keep locked for daemon */
return TRUE;
}
/*
* uao_dropswap: release any swap resources from this aobj page.
*
* => aobj must be locked or have a reference count of 0.
*/
void
uao_dropswap(struct uvm_object *uobj, int pageidx)
{
int slot;
slot = uao_set_swslot(uobj, pageidx, 0);
if (slot) {
uvm_swap_free(slot, 1);
}
}
/*
* page in every page in every aobj that is paged-out to a range of swslots.
*
* => nothing should be locked.
* => returns TRUE if pagein was aborted due to lack of memory.
*/
boolean_t
uao_swap_off(int startslot, int endslot)
{
struct uvm_aobj *aobj, *nextaobj;
/*
* walk the list of all aobjs.
*/
restart:
simple_lock(&uao_list_lock);
for (aobj = LIST_FIRST(&uao_list);
aobj != NULL;
aobj = nextaobj) {
boolean_t rv;
/*
* try to get the object lock,
* start all over if we fail.
* most of the time we'll get the aobj lock,
* so this should be a rare case.
*/
if (!simple_lock_try(&aobj->u_obj.vmobjlock)) {
simple_unlock(&uao_list_lock);
goto restart;
}
/*
* add a ref to the aobj so it doesn't disappear
* while we're working.
*/
uao_reference_locked(&aobj->u_obj);
/*
* now it's safe to unlock the uao list.
*/
simple_unlock(&uao_list_lock);
/*
* page in any pages in the swslot range.
* if there's an error, abort and return the error.
*/
rv = uao_pagein(aobj, startslot, endslot);
if (rv) {
uao_detach_locked(&aobj->u_obj);
return rv;
}
/*
* we're done with this aobj.
* relock the list and drop our ref on the aobj.
*/
simple_lock(&uao_list_lock);
nextaobj = LIST_NEXT(aobj, u_list);
uao_detach_locked(&aobj->u_obj);
}
/*
* done with traversal, unlock the list
*/
simple_unlock(&uao_list_lock);
return FALSE;
}
/*
* page in any pages from aobj in the given range.
*
* => aobj must be locked and is returned locked.
* => returns TRUE if pagein was aborted due to lack of memory.
*/
boolean_t
uao_pagein(struct uvm_aobj *aobj, int startslot, int endslot)
{
boolean_t rv;
if (UAO_USES_SWHASH(aobj)) {
struct uao_swhash_elt *elt;
int bucket;
restart:
for (bucket = aobj->u_swhashmask; bucket >= 0; bucket--) {
for (elt = LIST_FIRST(&aobj->u_swhash[bucket]);
elt != NULL;
elt = LIST_NEXT(elt, list)) {
int i;
for (i = 0; i < UAO_SWHASH_CLUSTER_SIZE; i++) {
int slot = elt->slots[i];
/*
* if the slot isn't in range, skip it.
*/
if (slot < startslot ||
slot >= endslot) {
continue;
}
/*
* process the page,
* the start over on this object
* since the swhash elt
* may have been freed.
*/
rv = uao_pagein_page(aobj,
UAO_SWHASH_ELT_PAGEIDX_BASE(elt) + i);
if (rv) {
return rv;
}
goto restart;
}
}
}
} else {
int i;
for (i = 0; i < aobj->u_pages; i++) {
int slot = aobj->u_swslots[i];
/*
* if the slot isn't in range, skip it
*/
if (slot < startslot || slot >= endslot) {
continue;
}
/*
* process the page.
*/
rv = uao_pagein_page(aobj, i);
if (rv) {
return rv;
}
}
}
return FALSE;
}
/*
* page in a page from an aobj. used for swap_off.
* returns TRUE if pagein was aborted due to lack of memory.
*
* => aobj must be locked and is returned locked.
*/
boolean_t
uao_pagein_page(struct uvm_aobj *aobj, int pageidx)
{
struct vm_page *pg;
int rv, slot, npages;
pg = NULL;
npages = 1;
/* locked: aobj */
rv = uao_get(&aobj->u_obj, pageidx << PAGE_SHIFT,
&pg, &npages, 0, VM_PROT_READ|VM_PROT_WRITE, 0, 0);
/* unlocked: aobj */
/*
* relock and finish up.
*/
simple_lock(&aobj->u_obj.vmobjlock);
switch (rv) {
case VM_PAGER_OK:
break;
case VM_PAGER_ERROR:
case VM_PAGER_REFAULT:
/*
* nothing more to do on errors.
* VM_PAGER_REFAULT can only mean that the anon was freed,
* so again there's nothing to do.
*/
return FALSE;
}
/*
* ok, we've got the page now.
* mark it as dirty, clear its swslot and un-busy it.
*/
slot = uao_set_swslot(&aobj->u_obj, pageidx, 0);
uvm_swap_free(slot, 1);
atomic_clearbits_int(&pg->pg_flags, PG_BUSY|PG_CLEAN|PG_FAKE);
UVM_PAGE_OWN(pg, NULL);
/*
* deactivate the page (to put it on a page queue).
*/
pmap_clear_reference(pg);
#ifndef UBC
pmap_page_protect(pg, VM_PROT_NONE);
#endif
uvm_lock_pageq();
uvm_pagedeactivate(pg);
uvm_unlock_pageq();
return FALSE;
}
|