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
|
/* $OpenBSD: uvm_amap.c,v 1.16 2001/11/28 19:28:14 art Exp $ */
/* $NetBSD: uvm_amap.c,v 1.32 2001/06/02 18:09:25 chs Exp $ */
/*
*
* Copyright (c) 1997 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.
*/
/*
* uvm_amap.c: amap operations
*/
/*
* this file contains functions that perform operations on amaps. see
* uvm_amap.h for a brief explanation of the role of amaps in uvm.
*/
#undef UVM_AMAP_INLINE /* enable/disable amap inlines */
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/proc.h>
#include <sys/malloc.h>
#include <sys/kernel.h>
#include <sys/pool.h>
#define UVM_AMAP_C /* ensure disabled inlines are in */
#include <uvm/uvm.h>
#include <uvm/uvm_swap.h>
/*
* pool for allocation of vm_map structures. note that the pool has
* its own simplelock for its protection. also note that in order to
* avoid an endless loop, the amap pool's allocator cannot allocate
* memory from an amap (it currently goes through the kernel uobj, so
* we are ok).
*/
struct pool uvm_amap_pool;
/*
* local functions
*/
static struct vm_amap *amap_alloc1 __P((int, int, int));
#ifdef UVM_AMAP_PPREF
/*
* what is ppref? ppref is an _optional_ amap feature which is used
* to keep track of reference counts on a per-page basis. it is enabled
* when UVM_AMAP_PPREF is defined.
*
* when enabled, an array of ints is allocated for the pprefs. this
* array is allocated only when a partial reference is added to the
* map (either by unmapping part of the amap, or gaining a reference
* to only a part of an amap). if the malloc of the array fails
* (M_NOWAIT), then we set the array pointer to PPREF_NONE to indicate
* that we tried to do ppref's but couldn't alloc the array so just
* give up (after all, this is an optional feature!).
*
* the array is divided into page sized "chunks." for chunks of length 1,
* the chunk reference count plus one is stored in that chunk's slot.
* for chunks of length > 1 the first slot contains (the reference count
* plus one) * -1. [the negative value indicates that the length is
* greater than one.] the second slot of the chunk contains the length
* of the chunk. here is an example:
*
* actual REFS: 2 2 2 2 3 1 1 0 0 0 4 4 0 1 1 1
* ppref: -3 4 x x 4 -2 2 -1 3 x -5 2 1 -2 3 x
* <----------><-><----><-------><----><-><------->
* (x = don't care)
*
* this allows us to allow one int to contain the ref count for the whole
* chunk. note that the "plus one" part is needed because a reference
* count of zero is neither positive or negative (need a way to tell
* if we've got one zero or a bunch of them).
*
* here are some in-line functions to help us.
*/
static __inline void pp_getreflen __P((int *, int, int *, int *));
static __inline void pp_setreflen __P((int *, int, int, int));
/*
* pp_getreflen: get the reference and length for a specific offset
*
* => ppref's amap must be locked
*/
static __inline void
pp_getreflen(ppref, offset, refp, lenp)
int *ppref, offset, *refp, *lenp;
{
if (ppref[offset] > 0) { /* chunk size must be 1 */
*refp = ppref[offset] - 1; /* don't forget to adjust */
*lenp = 1;
} else {
*refp = (ppref[offset] * -1) - 1;
*lenp = ppref[offset+1];
}
}
/*
* pp_setreflen: set the reference and length for a specific offset
*
* => ppref's amap must be locked
*/
static __inline void
pp_setreflen(ppref, offset, ref, len)
int *ppref, offset, ref, len;
{
if (len == 1) {
ppref[offset] = ref + 1;
} else {
ppref[offset] = (ref + 1) * -1;
ppref[offset+1] = len;
}
}
#endif
/*
* amap_init: called at boot time to init global amap data structures
*/
void
amap_init()
{
/*
* Initialize the vm_amap pool.
*/
pool_init(&uvm_amap_pool, sizeof(struct vm_amap), 0, 0, 0,
"amappl", 0, pool_page_alloc_nointr, pool_page_free_nointr,
M_UVMAMAP);
}
/*
* amap_alloc1: internal function that allocates an amap, but does not
* init the overlay.
*
* => lock on returned amap is init'd
*/
static inline struct vm_amap *
amap_alloc1(slots, padslots, waitf)
int slots, padslots, waitf;
{
struct vm_amap *amap;
int totalslots = slots + padslots;
amap = pool_get(&uvm_amap_pool, (waitf == M_WAITOK) ? PR_WAITOK : 0);
if (amap == NULL)
return(NULL);
simple_lock_init(&amap->am_l);
amap->am_ref = 1;
amap->am_flags = 0;
#ifdef UVM_AMAP_PPREF
amap->am_ppref = NULL;
#endif
amap->am_maxslot = totalslots;
amap->am_nslot = slots;
amap->am_nused = 0;
amap->am_slots = malloc(totalslots * sizeof(int), M_UVMAMAP,
waitf);
if (amap->am_slots == NULL)
goto fail1;
amap->am_bckptr = malloc(totalslots * sizeof(int), M_UVMAMAP, waitf);
if (amap->am_bckptr == NULL)
goto fail2;
amap->am_anon = malloc(totalslots * sizeof(struct vm_anon *),
M_UVMAMAP, waitf);
if (amap->am_anon == NULL)
goto fail3;
return(amap);
fail3:
free(amap->am_bckptr, M_UVMAMAP);
fail2:
free(amap->am_slots, M_UVMAMAP);
fail1:
pool_put(&uvm_amap_pool, amap);
return (NULL);
}
/*
* amap_alloc: allocate an amap to manage "sz" bytes of anonymous VM
*
* => caller should ensure sz is a multiple of PAGE_SIZE
* => reference count to new amap is set to one
* => new amap is returned unlocked
*/
struct vm_amap *
amap_alloc(sz, padsz, waitf)
vaddr_t sz, padsz;
int waitf;
{
struct vm_amap *amap;
int slots, padslots;
UVMHIST_FUNC("amap_alloc"); UVMHIST_CALLED(maphist);
AMAP_B2SLOT(slots, sz); /* load slots */
AMAP_B2SLOT(padslots, padsz);
amap = amap_alloc1(slots, padslots, waitf);
if (amap)
memset(amap->am_anon, 0, (slots + padslots) * sizeof(struct vm_anon *));
UVMHIST_LOG(maphist,"<- done, amap = 0x%x, sz=%d", amap, sz, 0, 0);
return(amap);
}
/*
* amap_free: free an amap
*
* => the amap must be locked (mainly for simplelock accounting)
* => the amap should have a zero reference count and be empty
*/
void
amap_free(amap)
struct vm_amap *amap;
{
UVMHIST_FUNC("amap_free"); UVMHIST_CALLED(maphist);
KASSERT(amap->am_ref == 0 && amap->am_nused == 0);
LOCK_ASSERT(simple_lock_held(&amap->am_l));
free(amap->am_slots, M_UVMAMAP);
free(amap->am_bckptr, M_UVMAMAP);
free(amap->am_anon, M_UVMAMAP);
#ifdef UVM_AMAP_PPREF
if (amap->am_ppref && amap->am_ppref != PPREF_NONE)
free(amap->am_ppref, M_UVMAMAP);
#endif
amap_unlock(amap); /* mainly for lock debugging */
pool_put(&uvm_amap_pool, amap);
UVMHIST_LOG(maphist,"<- done, freed amap = 0x%x", amap, 0, 0, 0);
}
/*
* amap_extend: extend the size of an amap (if needed)
*
* => called from uvm_map when we want to extend an amap to cover
* a new mapping (rather than allocate a new one)
* => amap should be unlocked (we will lock it)
* => to safely extend an amap it should have a reference count of
* one (thus it can't be shared)
* => XXXCDC: needs a waitflag or failure return value?
* => XXXCDC: support padding at this level?
*/
void
amap_extend(entry, addsize)
struct vm_map_entry *entry;
vsize_t addsize;
{
struct vm_amap *amap = entry->aref.ar_amap;
int slotoff = entry->aref.ar_pageoff;
int slotmapped, slotadd, slotneed;
#ifdef UVM_AMAP_PPREF
int *newppref, *oldppref;
#endif
u_int *newsl, *newbck, *oldsl, *oldbck;
struct vm_anon **newover, **oldover;
int slotadded;
UVMHIST_FUNC("amap_extend"); UVMHIST_CALLED(maphist);
UVMHIST_LOG(maphist, " (entry=0x%x, addsize=0x%x)", entry,addsize,0,0);
/*
* first, determine how many slots we need in the amap. don't
* forget that ar_pageoff could be non-zero: this means that
* there are some unused slots before us in the amap.
*/
amap_lock(amap); /* lock! */
AMAP_B2SLOT(slotmapped, entry->end - entry->start); /* slots mapped */
AMAP_B2SLOT(slotadd, addsize); /* slots to add */
slotneed = slotoff + slotmapped + slotadd;
/*
* case 1: we already have enough slots in the map and thus
* only need to bump the reference counts on the slots we are
* adding.
*/
if (amap->am_nslot >= slotneed) {
#ifdef UVM_AMAP_PPREF
if (amap->am_ppref && amap->am_ppref != PPREF_NONE) {
amap_pp_adjref(amap, slotoff + slotmapped, slotadd, 1);
}
#endif
amap_unlock(amap);
UVMHIST_LOG(maphist,"<- done (case 1), amap = 0x%x, sltneed=%d",
amap, slotneed, 0, 0);
return; /* done! */
}
/*
* case 2: we pre-allocated slots for use and we just need to
* bump nslot up to take account for these slots.
*/
if (amap->am_maxslot >= slotneed) {
#ifdef UVM_AMAP_PPREF
if (amap->am_ppref && amap->am_ppref != PPREF_NONE) {
if ((slotoff + slotmapped) < amap->am_nslot)
amap_pp_adjref(amap, slotoff + slotmapped,
(amap->am_nslot - (slotoff + slotmapped)),
1);
pp_setreflen(amap->am_ppref, amap->am_nslot, 1,
slotneed - amap->am_nslot);
}
#endif
amap->am_nslot = slotneed;
amap_unlock(amap);
/*
* no need to zero am_anon since that was done at
* alloc time and we never shrink an allocation.
*/
UVMHIST_LOG(maphist,"<- done (case 2), amap = 0x%x, slotneed=%d",
amap, slotneed, 0, 0);
return;
}
/*
* case 3: we need to malloc a new amap and copy all the amap
* data over from old amap to the new one.
*
* XXXCDC: could we take advantage of a kernel realloc()?
*/
amap_unlock(amap); /* unlock in case we sleep in malloc */
#ifdef UVM_AMAP_PPREF
newppref = NULL;
if (amap->am_ppref && amap->am_ppref != PPREF_NONE) {
newppref = malloc(slotneed * sizeof(int), M_UVMAMAP, M_NOWAIT);
if (newppref == NULL) {
/* give up if malloc fails */
free(amap->am_ppref, M_UVMAMAP);
amap->am_ppref = PPREF_NONE;
}
}
#endif
newsl = malloc(slotneed * sizeof(int), M_UVMAMAP, M_WAITOK);
newbck = malloc(slotneed * sizeof(int), M_UVMAMAP, M_WAITOK);
newover = malloc(slotneed * sizeof(struct vm_anon *),
M_UVMAMAP, M_WAITOK);
amap_lock(amap); /* re-lock! */
KASSERT(amap->am_maxslot < slotneed);
/*
* now copy everything over to new malloc'd areas...
*/
slotadded = slotneed - amap->am_nslot;
/* do am_slots */
oldsl = amap->am_slots;
memcpy(newsl, oldsl, sizeof(int) * amap->am_nused);
amap->am_slots = newsl;
/* do am_anon */
oldover = amap->am_anon;
memcpy(newover, oldover, sizeof(struct vm_anon *) * amap->am_nslot);
memset(newover + amap->am_nslot, 0, sizeof(struct vm_anon *) * slotadded);
amap->am_anon = newover;
/* do am_bckptr */
oldbck = amap->am_bckptr;
memcpy(newbck, oldbck, sizeof(int) * amap->am_nslot);
memset(newbck + amap->am_nslot, 0, sizeof(int) * slotadded); /* XXX: needed? */
amap->am_bckptr = newbck;
#ifdef UVM_AMAP_PPREF
/* do ppref */
oldppref = amap->am_ppref;
if (newppref) {
memcpy(newppref, oldppref, sizeof(int) * amap->am_nslot);
memset(newppref + amap->am_nslot, 0, sizeof(int) * slotadded);
amap->am_ppref = newppref;
if ((slotoff + slotmapped) < amap->am_nslot)
amap_pp_adjref(amap, slotoff + slotmapped,
(amap->am_nslot - (slotoff + slotmapped)), 1);
pp_setreflen(newppref, amap->am_nslot, 1, slotadded);
}
#endif
/* update master values */
amap->am_nslot = slotneed;
amap->am_maxslot = slotneed;
/* unlock */
amap_unlock(amap);
/* and free */
free(oldsl, M_UVMAMAP);
free(oldbck, M_UVMAMAP);
free(oldover, M_UVMAMAP);
#ifdef UVM_AMAP_PPREF
if (oldppref && oldppref != PPREF_NONE)
free(oldppref, M_UVMAMAP);
#endif
UVMHIST_LOG(maphist,"<- done (case 3), amap = 0x%x, slotneed=%d",
amap, slotneed, 0, 0);
}
/*
* amap_share_protect: change protection of anons in a shared amap
*
* for shared amaps, given the current data structure layout, it is
* not possible for us to directly locate all maps referencing the
* shared anon (to change the protection). in order to protect data
* in shared maps we use pmap_page_protect(). [this is useful for IPC
* mechanisms like map entry passing that may want to write-protect
* all mappings of a shared amap.] we traverse am_anon or am_slots
* depending on the current state of the amap.
*
* => entry's map and amap must be locked by the caller
*/
void
amap_share_protect(entry, prot)
struct vm_map_entry *entry;
vm_prot_t prot;
{
struct vm_amap *amap = entry->aref.ar_amap;
int slots, lcv, slot, stop;
LOCK_ASSERT(simple_lock_held(&amap->am_l));
AMAP_B2SLOT(slots, (entry->end - entry->start));
stop = entry->aref.ar_pageoff + slots;
if (slots < amap->am_nused) {
/* cheaper to traverse am_anon */
for (lcv = entry->aref.ar_pageoff ; lcv < stop ; lcv++) {
if (amap->am_anon[lcv] == NULL)
continue;
if (amap->am_anon[lcv]->u.an_page != NULL)
pmap_page_protect(amap->am_anon[lcv]->u.an_page,
prot);
}
return;
}
/* cheaper to traverse am_slots */
for (lcv = 0 ; lcv < amap->am_nused ; lcv++) {
slot = amap->am_slots[lcv];
if (slot < entry->aref.ar_pageoff || slot >= stop)
continue;
if (amap->am_anon[slot]->u.an_page != NULL)
pmap_page_protect(amap->am_anon[slot]->u.an_page, prot);
}
return;
}
/*
* amap_wipeout: wipeout all anon's in an amap; then free the amap!
*
* => called from amap_unref when the final reference to an amap is
* discarded (i.e. when reference count == 1)
* => the amap should be locked (by the caller)
*/
void
amap_wipeout(amap)
struct vm_amap *amap;
{
int lcv, slot;
struct vm_anon *anon;
UVMHIST_FUNC("amap_wipeout"); UVMHIST_CALLED(maphist);
UVMHIST_LOG(maphist,"(amap=0x%x)", amap, 0,0,0);
LOCK_ASSERT(simple_lock_held(&amap->am_l));
for (lcv = 0 ; lcv < amap->am_nused ; lcv++) {
int refs;
slot = amap->am_slots[lcv];
anon = amap->am_anon[slot];
if (anon == NULL || anon->an_ref == 0)
panic("amap_wipeout: corrupt amap");
simple_lock(&anon->an_lock); /* lock anon */
UVMHIST_LOG(maphist," processing anon 0x%x, ref=%d", anon,
anon->an_ref, 0, 0);
refs = --anon->an_ref;
simple_unlock(&anon->an_lock);
if (refs == 0) {
/*
* we had the last reference to a vm_anon. free it.
*/
uvm_anfree(anon);
}
}
/*
* now we free the map
*/
amap->am_ref = 0; /* ... was one */
amap->am_nused = 0;
amap_free(amap); /* will unlock and free amap */
UVMHIST_LOG(maphist,"<- done!", 0,0,0,0);
}
/*
* amap_copy: ensure that a map entry's "needs_copy" flag is false
* by copying the amap if necessary.
*
* => an entry with a null amap pointer will get a new (blank) one.
* => the map that the map entry belongs to must be locked by caller.
* => the amap currently attached to "entry" (if any) must be unlocked.
* => if canchunk is true, then we may clip the entry into a chunk
* => "startva" and "endva" are used only if canchunk is true. they are
* used to limit chunking (e.g. if you have a large space that you
* know you are going to need to allocate amaps for, there is no point
* in allowing that to be chunked)
*/
void
amap_copy(map, entry, waitf, canchunk, startva, endva)
struct vm_map *map;
struct vm_map_entry *entry;
int waitf;
boolean_t canchunk;
vaddr_t startva, endva;
{
struct vm_amap *amap, *srcamap;
int slots, lcv;
vaddr_t chunksize;
UVMHIST_FUNC("amap_copy"); UVMHIST_CALLED(maphist);
UVMHIST_LOG(maphist, " (map=%p, entry=%p, waitf=%d)",
map, entry, waitf, 0);
/*
* is there a map to copy? if not, create one from scratch.
*/
if (entry->aref.ar_amap == NULL) {
/*
* check to see if we have a large amap that we can
* chunk. we align startva/endva to chunk-sized
* boundaries and then clip to them.
*/
if (canchunk && atop(entry->end - entry->start) >=
UVM_AMAP_LARGE) {
/* convert slots to bytes */
chunksize = UVM_AMAP_CHUNK << PAGE_SHIFT;
startva = (startva / chunksize) * chunksize;
endva = roundup(endva, chunksize);
UVMHIST_LOG(maphist, " chunk amap ==> clip 0x%x->0x%x"
"to 0x%x->0x%x", entry->start, entry->end, startva,
endva);
UVM_MAP_CLIP_START(map, entry, startva);
/* watch out for endva wrap-around! */
if (endva >= startva)
UVM_MAP_CLIP_END(map, entry, endva);
}
UVMHIST_LOG(maphist, "<- done [creating new amap 0x%x->0x%x]",
entry->start, entry->end, 0, 0);
entry->aref.ar_pageoff = 0;
entry->aref.ar_amap = amap_alloc(entry->end - entry->start, 0,
waitf);
if (entry->aref.ar_amap != NULL)
entry->etype &= ~UVM_ET_NEEDSCOPY;
return;
}
/*
* first check and see if we are the only map entry
* referencing the amap we currently have. if so, then we can
* just take it over rather than copying it. note that we are
* reading am_ref with the amap unlocked... the value can only
* be one if we have the only reference to the amap (via our
* locked map). if we are greater than one we fall through to
* the next case (where we double check the value).
*/
if (entry->aref.ar_amap->am_ref == 1) {
entry->etype &= ~UVM_ET_NEEDSCOPY;
UVMHIST_LOG(maphist, "<- done [ref cnt = 1, took it over]",
0, 0, 0, 0);
return;
}
/*
* looks like we need to copy the map.
*/
UVMHIST_LOG(maphist," amap=%p, ref=%d, must copy it",
entry->aref.ar_amap, entry->aref.ar_amap->am_ref, 0, 0);
AMAP_B2SLOT(slots, entry->end - entry->start);
amap = amap_alloc1(slots, 0, waitf);
if (amap == NULL) {
UVMHIST_LOG(maphist, " amap_alloc1 failed", 0,0,0,0);
return;
}
srcamap = entry->aref.ar_amap;
amap_lock(srcamap);
/*
* need to double check reference count now that we've got the
* src amap locked down. the reference count could have
* changed while we were in malloc. if the reference count
* dropped down to one we take over the old map rather than
* copying the amap.
*/
if (srcamap->am_ref == 1) { /* take it over? */
entry->etype &= ~UVM_ET_NEEDSCOPY;
amap->am_ref--; /* drop final reference to map */
amap_free(amap); /* dispose of new (unused) amap */
amap_unlock(srcamap);
return;
}
/*
* we must copy it now.
*/
UVMHIST_LOG(maphist, " copying amap now",0, 0, 0, 0);
for (lcv = 0 ; lcv < slots; lcv++) {
amap->am_anon[lcv] =
srcamap->am_anon[entry->aref.ar_pageoff + lcv];
if (amap->am_anon[lcv] == NULL)
continue;
simple_lock(&amap->am_anon[lcv]->an_lock);
amap->am_anon[lcv]->an_ref++;
simple_unlock(&amap->am_anon[lcv]->an_lock);
amap->am_bckptr[lcv] = amap->am_nused;
amap->am_slots[amap->am_nused] = lcv;
amap->am_nused++;
}
/*
* drop our reference to the old amap (srcamap) and unlock.
* we know that the reference count on srcamap is greater than
* one (we checked above), so there is no way we could drop
* the count to zero. [and no need to worry about freeing it]
*/
srcamap->am_ref--;
if (srcamap->am_ref == 1 && (srcamap->am_flags & AMAP_SHARED) != 0)
srcamap->am_flags &= ~AMAP_SHARED; /* clear shared flag */
#ifdef UVM_AMAP_PPREF
if (srcamap->am_ppref && srcamap->am_ppref != PPREF_NONE) {
amap_pp_adjref(srcamap, entry->aref.ar_pageoff,
(entry->end - entry->start) >> PAGE_SHIFT, -1);
}
#endif
amap_unlock(srcamap);
/*
* install new amap.
*/
entry->aref.ar_pageoff = 0;
entry->aref.ar_amap = amap;
entry->etype &= ~UVM_ET_NEEDSCOPY;
/*
* done!
*/
UVMHIST_LOG(maphist, "<- done",0, 0, 0, 0);
}
/*
* amap_cow_now: resolve all copy-on-write faults in an amap now for fork(2)
*
* called during fork(2) when the parent process has a wired map
* entry. in that case we want to avoid write-protecting pages
* in the parent's map (e.g. like what you'd do for a COW page)
* so we resolve the COW here.
*
* => assume parent's entry was wired, thus all pages are resident.
* => assume pages that are loaned out (loan_count) are already mapped
* read-only in all maps, and thus no need for us to worry about them
* => assume both parent and child vm_map's are locked
* => caller passes child's map/entry in to us
* => if we run out of memory we will unlock the amap and sleep _with_ the
* parent and child vm_map's locked(!). we have to do this since
* we are in the middle of a fork(2) and we can't let the parent
* map change until we are done copying all the map entrys.
* => XXXCDC: out of memory should cause fork to fail, but there is
* currently no easy way to do this (needs fix)
* => page queues must be unlocked (we may lock them)
*/
void
amap_cow_now(map, entry)
struct vm_map *map;
struct vm_map_entry *entry;
{
struct vm_amap *amap = entry->aref.ar_amap;
int lcv, slot;
struct vm_anon *anon, *nanon;
struct vm_page *pg, *npg;
/*
* note that if we unlock the amap then we must ReStart the "lcv" for
* loop because some other process could reorder the anon's in the
* am_anon[] array on us while the lock is dropped.
*/
ReStart:
amap_lock(amap);
for (lcv = 0 ; lcv < amap->am_nused ; lcv++) {
/*
* get the page
*/
slot = amap->am_slots[lcv];
anon = amap->am_anon[slot];
simple_lock(&anon->an_lock);
pg = anon->u.an_page;
/*
* page must be resident since parent is wired
*/
if (pg == NULL)
panic("amap_cow_now: non-resident wired page in anon %p",
anon);
/*
* if the anon ref count is one and the page is not loaned,
* then we are safe (the child has exclusive access to the
* page). if the page is loaned, then it must already be
* mapped read-only.
*
* we only need to get involved when these are not true.
* [note: if loan_count == 0, then the anon must own the page]
*/
if (anon->an_ref > 1 && pg->loan_count == 0) {
/*
* if the page is busy then we have to unlock, wait for
* it and then restart.
*/
if (pg->flags & PG_BUSY) {
pg->flags |= PG_WANTED;
amap_unlock(amap);
UVM_UNLOCK_AND_WAIT(pg, &anon->an_lock, FALSE,
"cownow", 0);
goto ReStart;
}
/*
* ok, time to do a copy-on-write to a new anon
*/
nanon = uvm_analloc();
if (nanon) {
/* nanon is locked! */
npg = uvm_pagealloc(NULL, 0, nanon, 0);
} else
npg = NULL; /* XXX: quiet gcc warning */
if (nanon == NULL || npg == NULL) {
/* out of memory */
/*
* XXXCDC: we should cause fork to fail, but
* we can't ...
*/
if (nanon) {
nanon->an_ref--;
simple_unlock(&nanon->an_lock);
uvm_anfree(nanon);
}
simple_unlock(&anon->an_lock);
amap_unlock(amap);
uvm_wait("cownowpage");
goto ReStart;
}
/*
* got it... now we can copy the data and replace anon
* with our new one...
*/
uvm_pagecopy(pg, npg); /* old -> new */
anon->an_ref--; /* can't drop to zero */
amap->am_anon[slot] = nanon; /* replace */
/*
* drop PG_BUSY on new page ... since we have had it's
* owner locked the whole time it can't be
* PG_RELEASED | PG_WANTED.
*/
npg->flags &= ~(PG_BUSY|PG_FAKE);
UVM_PAGE_OWN(npg, NULL);
uvm_lock_pageq();
uvm_pageactivate(npg);
uvm_unlock_pageq();
simple_unlock(&nanon->an_lock);
}
simple_unlock(&anon->an_lock);
/*
* done with this anon, next ...!
*/
} /* end of 'for' loop */
amap_unlock(amap);
}
/*
* amap_splitref: split a single reference into two separate references
*
* => called from uvm_map's clip routines
* => origref's map should be locked
* => origref->ar_amap should be unlocked (we will lock)
*/
void
amap_splitref(origref, splitref, offset)
struct vm_aref *origref, *splitref;
vaddr_t offset;
{
int leftslots;
AMAP_B2SLOT(leftslots, offset);
if (leftslots == 0)
panic("amap_splitref: split at zero offset");
/*
* lock the amap
*/
amap_lock(origref->ar_amap);
/*
* now: amap is locked and we have a valid am_mapped array.
*/
if (origref->ar_amap->am_nslot - origref->ar_pageoff - leftslots <= 0)
panic("amap_splitref: map size check failed");
#ifdef UVM_AMAP_PPREF
/*
* establish ppref before we add a duplicate reference to the amap
*/
if (origref->ar_amap->am_ppref == NULL)
amap_pp_establish(origref->ar_amap);
#endif
splitref->ar_amap = origref->ar_amap;
splitref->ar_amap->am_ref++; /* not a share reference */
splitref->ar_pageoff = origref->ar_pageoff + leftslots;
amap_unlock(origref->ar_amap);
}
#ifdef UVM_AMAP_PPREF
/*
* amap_pp_establish: add a ppref array to an amap, if possible
*
* => amap locked by caller
*/
void
amap_pp_establish(amap)
struct vm_amap *amap;
{
amap->am_ppref = malloc(sizeof(int) * amap->am_maxslot,
M_UVMAMAP, M_NOWAIT);
/*
* if we fail then we just won't use ppref for this amap
*/
if (amap->am_ppref == NULL) {
amap->am_ppref = PPREF_NONE; /* not using it */
return;
}
/*
* init ppref
*/
memset(amap->am_ppref, 0, sizeof(int) * amap->am_maxslot);
pp_setreflen(amap->am_ppref, 0, amap->am_ref, amap->am_nslot);
return;
}
/*
* amap_pp_adjref: adjust reference count to a part of an amap using the
* per-page reference count array.
*
* => map and amap locked by caller
* => caller must check that ppref != PPREF_NONE before calling
*/
void
amap_pp_adjref(amap, curslot, slotlen, adjval)
struct vm_amap *amap;
int curslot;
vsize_t slotlen;
int adjval;
{
int stopslot, *ppref, lcv;
int ref, len;
/*
* get init values
*/
stopslot = curslot + slotlen;
ppref = amap->am_ppref;
/*
* first advance to the correct place in the ppref array, fragment
* if needed.
*/
for (lcv = 0 ; lcv < curslot ; lcv += len) {
pp_getreflen(ppref, lcv, &ref, &len);
if (lcv + len > curslot) { /* goes past start? */
pp_setreflen(ppref, lcv, ref, curslot - lcv);
pp_setreflen(ppref, curslot, ref, len - (curslot -lcv));
len = curslot - lcv; /* new length of entry @ lcv */
}
}
/*
* now adjust reference counts in range (make sure we dont overshoot)
*/
if (lcv != curslot)
panic("amap_pp_adjref: overshot target");
for (/* lcv already set */; lcv < stopslot ; lcv += len) {
pp_getreflen(ppref, lcv, &ref, &len);
if (lcv + len > stopslot) { /* goes past end? */
pp_setreflen(ppref, lcv, ref, stopslot - lcv);
pp_setreflen(ppref, stopslot, ref,
len - (stopslot - lcv));
len = stopslot - lcv;
}
ref = ref + adjval; /* ADJUST! */
if (ref < 0)
panic("amap_pp_adjref: negative reference count");
pp_setreflen(ppref, lcv, ref, len);
if (ref == 0)
amap_wiperange(amap, lcv, len);
}
}
/*
* amap_wiperange: wipe out a range of an amap
* [different from amap_wipeout because the amap is kept intact]
*
* => both map and amap must be locked by caller.
*/
void
amap_wiperange(amap, slotoff, slots)
struct vm_amap *amap;
int slotoff, slots;
{
int byanon, lcv, stop, curslot, ptr;
struct vm_anon *anon;
/*
* we can either traverse the amap by am_anon or by am_slots depending
* on which is cheaper. decide now.
*/
if (slots < amap->am_nused) {
byanon = TRUE;
lcv = slotoff;
stop = slotoff + slots;
} else {
byanon = FALSE;
lcv = 0;
stop = amap->am_nused;
}
/*
* ok, now do it!
*/
for (; lcv < stop; lcv++) {
int refs;
/*
* verify the anon is ok.
*/
if (byanon) {
if (amap->am_anon[lcv] == NULL)
continue;
curslot = lcv;
} else {
curslot = amap->am_slots[lcv];
if (curslot < slotoff || curslot >= stop)
continue;
}
anon = amap->am_anon[curslot];
/*
* remove it from the amap
*/
amap->am_anon[curslot] = NULL;
ptr = amap->am_bckptr[curslot];
if (ptr != (amap->am_nused - 1)) {
amap->am_slots[ptr] =
amap->am_slots[amap->am_nused - 1];
amap->am_bckptr[amap->am_slots[ptr]] =
ptr; /* back ptr. */
}
amap->am_nused--;
/*
* drop anon reference count
*/
simple_lock(&anon->an_lock);
refs = --anon->an_ref;
simple_unlock(&anon->an_lock);
if (refs == 0) {
/*
* we just eliminated the last reference to an anon.
* free it.
*/
uvm_anfree(anon);
}
}
}
#endif
|