/
uvm_amap.c
1381 lines (1193 loc) · 35.1 KB
/
uvm_amap.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
/* $OpenBSD: uvm_amap.c,v 1.92 2023/04/11 00:45:09 jsg Exp $ */
/* $NetBSD: uvm_amap.c,v 1.27 2000/11/25 06:27:59 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.
*
* 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.
*/
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/malloc.h>
#include <sys/kernel.h>
#include <sys/pool.h>
#include <sys/atomic.h>
#include <uvm/uvm.h>
#include <uvm/uvm_swap.h>
/*
* pools for allocation of vm_amap structures. 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;
struct pool uvm_small_amap_pool[UVM_AMAP_CHUNK];
struct pool uvm_amap_chunk_pool;
LIST_HEAD(, vm_amap) amap_list;
struct rwlock amap_list_lock = RWLOCK_INITIALIZER("amaplstlk");
#define amap_lock_list() rw_enter_write(&amap_list_lock)
#define amap_unlock_list() rw_exit_write(&amap_list_lock)
static char amap_small_pool_names[UVM_AMAP_CHUNK][9];
/*
* local functions
*/
static struct vm_amap *amap_alloc1(int, int, int);
static inline void amap_list_insert(struct vm_amap *);
static inline void amap_list_remove(struct vm_amap *);
struct vm_amap_chunk *amap_chunk_get(struct vm_amap *, int, int, int);
void amap_chunk_free(struct vm_amap *, struct vm_amap_chunk *);
/*
* if we enable PPREF, then we have a couple of extra functions that
* we need to prototype here...
*/
#ifdef UVM_AMAP_PPREF
#define PPREF_NONE ((int *) -1) /* not using ppref */
void amap_pp_adjref(struct vm_amap *, int, vsize_t, int);
void amap_pp_establish(struct vm_amap *);
void amap_wiperange_chunk(struct vm_amap *, struct vm_amap_chunk *, int,
int);
void amap_wiperange(struct vm_amap *, int, int);
#endif /* UVM_AMAP_PPREF */
static inline void
amap_list_insert(struct vm_amap *amap)
{
amap_lock_list();
LIST_INSERT_HEAD(&amap_list, amap, am_list);
amap_unlock_list();
}
static inline void
amap_list_remove(struct vm_amap *amap)
{
amap_lock_list();
LIST_REMOVE(amap, am_list);
amap_unlock_list();
}
/*
* amap_chunk_get: lookup a chunk for slot. if create is non-zero,
* the chunk is created if it does not yet exist.
*
* => returns the chunk on success or NULL on error
*/
struct vm_amap_chunk *
amap_chunk_get(struct vm_amap *amap, int slot, int create, int waitf)
{
int bucket = UVM_AMAP_BUCKET(amap, slot);
int baseslot = AMAP_BASE_SLOT(slot);
int n;
struct vm_amap_chunk *chunk, *newchunk, *pchunk = NULL;
if (UVM_AMAP_SMALL(amap))
return &amap->am_small;
for (chunk = amap->am_buckets[bucket]; chunk != NULL;
chunk = TAILQ_NEXT(chunk, ac_list)) {
if (UVM_AMAP_BUCKET(amap, chunk->ac_baseslot) != bucket)
break;
if (chunk->ac_baseslot == baseslot)
return chunk;
pchunk = chunk;
}
if (!create)
return NULL;
if (amap->am_nslot - baseslot >= UVM_AMAP_CHUNK)
n = UVM_AMAP_CHUNK;
else
n = amap->am_nslot - baseslot;
newchunk = pool_get(&uvm_amap_chunk_pool, waitf | PR_ZERO);
if (newchunk == NULL)
return NULL;
if (pchunk == NULL) {
TAILQ_INSERT_TAIL(&amap->am_chunks, newchunk, ac_list);
KASSERT(amap->am_buckets[bucket] == NULL);
amap->am_buckets[bucket] = newchunk;
} else
TAILQ_INSERT_AFTER(&amap->am_chunks, pchunk, newchunk,
ac_list);
amap->am_ncused++;
newchunk->ac_baseslot = baseslot;
newchunk->ac_nslot = n;
return newchunk;
}
void
amap_chunk_free(struct vm_amap *amap, struct vm_amap_chunk *chunk)
{
int bucket = UVM_AMAP_BUCKET(amap, chunk->ac_baseslot);
struct vm_amap_chunk *nchunk;
if (UVM_AMAP_SMALL(amap))
return;
nchunk = TAILQ_NEXT(chunk, ac_list);
TAILQ_REMOVE(&amap->am_chunks, chunk, ac_list);
if (amap->am_buckets[bucket] == chunk) {
if (nchunk != NULL &&
UVM_AMAP_BUCKET(amap, nchunk->ac_baseslot) == bucket)
amap->am_buckets[bucket] = nchunk;
else
amap->am_buckets[bucket] = NULL;
}
pool_put(&uvm_amap_chunk_pool, chunk);
amap->am_ncused--;
}
#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 allocation 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.
*/
/*
* pp_getreflen: get the reference and length for a specific offset
*
* => ppref's amap must be locked
*/
static inline void
pp_getreflen(int *ppref, int offset, int *refp, int *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(int *ppref, int offset, int ref, int len)
{
if (len == 1) {
ppref[offset] = ref + 1;
} else {
ppref[offset] = (ref + 1) * -1;
ppref[offset+1] = len;
}
}
#endif /* UVM_AMAP_PPREF */
/*
* amap_init: called at boot time to init global amap data structures
*/
void
amap_init(void)
{
int i;
size_t size;
/* Initialize the vm_amap pool. */
pool_init(&uvm_amap_pool, sizeof(struct vm_amap),
0, IPL_MPFLOOR, PR_WAITOK, "amappl", NULL);
pool_sethiwat(&uvm_amap_pool, 4096);
/* initialize small amap pools */
for (i = 0; i < nitems(uvm_small_amap_pool); i++) {
snprintf(amap_small_pool_names[i],
sizeof(amap_small_pool_names[0]), "amappl%d", i + 1);
size = offsetof(struct vm_amap, am_small.ac_anon) +
(i + 1) * sizeof(struct vm_anon *);
pool_init(&uvm_small_amap_pool[i], size, 0, IPL_MPFLOOR,
PR_WAITOK, amap_small_pool_names[i], NULL);
}
pool_init(&uvm_amap_chunk_pool, sizeof(struct vm_amap_chunk) +
UVM_AMAP_CHUNK * sizeof(struct vm_anon *),
0, IPL_MPFLOOR, PR_WAITOK, "amapchunkpl", NULL);
pool_sethiwat(&uvm_amap_chunk_pool, 4096);
}
/*
* amap_alloc1: allocate an amap, but do not initialise the overlay.
*
* => Note: lock is not set.
*/
static inline struct vm_amap *
amap_alloc1(int slots, int waitf, int lazyalloc)
{
struct vm_amap *amap;
struct vm_amap_chunk *chunk, *tmp;
int chunks, log_chunks, chunkperbucket = 1, hashshift = 0;
int buckets, i, n;
int pwaitf = (waitf & M_WAITOK) ? PR_WAITOK : PR_NOWAIT;
KASSERT(slots > 0);
/*
* Cast to unsigned so that rounding up cannot cause integer overflow
* if slots is large.
*/
chunks = roundup((unsigned int)slots, UVM_AMAP_CHUNK) / UVM_AMAP_CHUNK;
if (lazyalloc) {
/*
* Basically, the amap is a hash map where the number of
* buckets is fixed. We select the number of buckets using the
* following strategy:
*
* 1. The maximal number of entries to search in a bucket upon
* a collision should be less than or equal to
* log2(slots / UVM_AMAP_CHUNK). This is the worst-case number
* of lookups we would have if we could chunk the amap. The
* log2(n) comes from the fact that amaps are chunked by
* splitting up their vm_map_entries and organizing those
* in a binary search tree.
*
* 2. The maximal number of entries in a bucket must be a
* power of two.
*
* The maximal number of entries per bucket is used to hash
* a slot to a bucket.
*
* In the future, this strategy could be refined to make it
* even harder/impossible that the total amount of KVA needed
* for the hash buckets of all amaps to exceed the maximal
* amount of KVA memory reserved for amaps.
*/
for (log_chunks = 1; (chunks >> log_chunks) > 0; log_chunks++)
continue;
chunkperbucket = 1 << hashshift;
while (chunkperbucket + 1 < log_chunks) {
hashshift++;
chunkperbucket = 1 << hashshift;
}
}
if (slots > UVM_AMAP_CHUNK)
amap = pool_get(&uvm_amap_pool, pwaitf);
else
amap = pool_get(&uvm_small_amap_pool[slots - 1],
pwaitf | PR_ZERO);
if (amap == NULL)
return NULL;
amap->am_lock = NULL;
amap->am_ref = 1;
amap->am_flags = 0;
#ifdef UVM_AMAP_PPREF
amap->am_ppref = NULL;
#endif
amap->am_nslot = slots;
amap->am_nused = 0;
if (UVM_AMAP_SMALL(amap)) {
amap->am_small.ac_nslot = slots;
return amap;
}
amap->am_ncused = 0;
TAILQ_INIT(&amap->am_chunks);
amap->am_hashshift = hashshift;
amap->am_buckets = NULL;
buckets = howmany(chunks, chunkperbucket);
amap->am_buckets = mallocarray(buckets, sizeof(*amap->am_buckets),
M_UVMAMAP, waitf | (lazyalloc ? M_ZERO : 0));
if (amap->am_buckets == NULL)
goto fail1;
amap->am_nbuckets = buckets;
if (!lazyalloc) {
for (i = 0; i < buckets; i++) {
if (i == buckets - 1) {
n = slots % UVM_AMAP_CHUNK;
if (n == 0)
n = UVM_AMAP_CHUNK;
} else
n = UVM_AMAP_CHUNK;
chunk = pool_get(&uvm_amap_chunk_pool,
PR_ZERO | pwaitf);
if (chunk == NULL)
goto fail1;
amap->am_buckets[i] = chunk;
amap->am_ncused++;
chunk->ac_baseslot = i * UVM_AMAP_CHUNK;
chunk->ac_nslot = n;
TAILQ_INSERT_TAIL(&amap->am_chunks, chunk, ac_list);
}
}
return amap;
fail1:
free(amap->am_buckets, M_UVMAMAP, buckets * sizeof(*amap->am_buckets));
TAILQ_FOREACH_SAFE(chunk, &amap->am_chunks, ac_list, tmp)
pool_put(&uvm_amap_chunk_pool, chunk);
pool_put(&uvm_amap_pool, amap);
return NULL;
}
static void
amap_lock_alloc(struct vm_amap *amap)
{
rw_obj_alloc(&amap->am_lock, "amaplk");
}
/*
* 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(vaddr_t sz, int waitf, int lazyalloc)
{
struct vm_amap *amap;
size_t slots;
AMAP_B2SLOT(slots, sz); /* load slots */
if (slots > INT_MAX)
return NULL;
amap = amap_alloc1(slots, waitf, lazyalloc);
if (amap != NULL) {
amap_lock_alloc(amap);
amap_list_insert(amap);
}
return amap;
}
/*
* amap_free: free an amap
*
* => the amap must be unlocked
* => the amap should have a zero reference count and be empty
*/
void
amap_free(struct vm_amap *amap)
{
struct vm_amap_chunk *chunk, *tmp;
KASSERT(amap->am_ref == 0 && amap->am_nused == 0);
KASSERT((amap->am_flags & AMAP_SWAPOFF) == 0);
if (amap->am_lock != NULL) {
KASSERT(amap->am_lock == NULL || !rw_write_held(amap->am_lock));
rw_obj_free(amap->am_lock);
}
#ifdef UVM_AMAP_PPREF
if (amap->am_ppref && amap->am_ppref != PPREF_NONE)
free(amap->am_ppref, M_UVMAMAP, amap->am_nslot * sizeof(int));
#endif
if (UVM_AMAP_SMALL(amap))
pool_put(&uvm_small_amap_pool[amap->am_nslot - 1], amap);
else {
TAILQ_FOREACH_SAFE(chunk, &amap->am_chunks, ac_list, tmp)
pool_put(&uvm_amap_chunk_pool, chunk);
free(amap->am_buckets, M_UVMAMAP,
amap->am_nbuckets * sizeof(*amap->am_buckets));
pool_put(&uvm_amap_pool, amap);
}
}
/*
* amap_wipeout: wipeout all anon's in an amap; then free the amap!
*
* => Called from amap_unref(), when reference count drops to zero.
* => amap must be locked.
*/
void
amap_wipeout(struct vm_amap *amap)
{
int slot;
struct vm_anon *anon;
struct vm_amap_chunk *chunk;
struct pglist pgl;
KASSERT(rw_write_held(amap->am_lock));
KASSERT(amap->am_ref == 0);
if (__predict_false((amap->am_flags & AMAP_SWAPOFF) != 0)) {
/*
* Note: amap_swap_off() will call us again.
*/
amap_unlock(amap);
return;
}
TAILQ_INIT(&pgl);
amap_list_remove(amap);
AMAP_CHUNK_FOREACH(chunk, amap) {
int i, refs, map = chunk->ac_usedmap;
for (i = ffs(map); i != 0; i = ffs(map)) {
slot = i - 1;
map ^= 1 << slot;
anon = chunk->ac_anon[slot];
if (anon == NULL || anon->an_ref == 0)
panic("amap_wipeout: corrupt amap");
KASSERT(anon->an_lock == amap->am_lock);
/*
* Drop the reference.
*/
refs = --anon->an_ref;
if (refs == 0) {
uvm_anfree_list(anon, &pgl);
}
}
}
/* free the pages */
uvm_pglistfree(&pgl);
/*
* Finally, destroy the amap.
*/
amap->am_ref = 0; /* ... was one */
amap->am_nused = 0;
amap_unlock(amap);
amap_free(amap);
}
/*
* 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(struct vm_map *map, struct vm_map_entry *entry, int waitf,
boolean_t canchunk, vaddr_t startva, vaddr_t endva)
{
struct vm_amap *amap, *srcamap;
int slots, lcv, lazyalloc = 0;
vaddr_t chunksize;
int i, j, k, n, srcslot;
struct vm_amap_chunk *chunk = NULL, *srcchunk = NULL;
struct vm_anon *anon;
KASSERT(map != kernel_map); /* we use sleeping locks */
/*
* Is there an amap to copy? If not, create one.
*/
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 we cannot chunk the amap, allocate it in a way
* that makes it grow or shrink dynamically with
* the number of slots.
*/
if (atop(entry->end - entry->start) >= UVM_AMAP_LARGE) {
if (canchunk) {
/* convert slots to bytes */
chunksize = UVM_AMAP_CHUNK << PAGE_SHIFT;
startva = (startva / chunksize) * chunksize;
endva = roundup(endva, chunksize);
UVM_MAP_CLIP_START(map, entry, startva);
/* watch out for endva wrap-around! */
if (endva >= startva)
UVM_MAP_CLIP_END(map, entry, endva);
} else
lazyalloc = 1;
}
entry->aref.ar_pageoff = 0;
entry->aref.ar_amap = amap_alloc(entry->end - entry->start,
waitf, lazyalloc);
if (entry->aref.ar_amap != NULL)
entry->etype &= ~UVM_ET_NEEDSCOPY;
return;
}
/*
* First check and see if we are the only map entry referencing
* he amap we currently have. If so, then just take it over instead
* of copying it. Note that we are reading am_ref without lock held
* as the value can only be one if we have the only reference
* to the amap (via our locked map). If the value is greater than
* one, then allocate amap and re-check the value.
*/
if (entry->aref.ar_amap->am_ref == 1) {
entry->etype &= ~UVM_ET_NEEDSCOPY;
return;
}
/*
* Allocate a new amap (note: not initialised, etc).
*/
AMAP_B2SLOT(slots, entry->end - entry->start);
if (!UVM_AMAP_SMALL(entry->aref.ar_amap) &&
entry->aref.ar_amap->am_hashshift != 0)
lazyalloc = 1;
amap = amap_alloc1(slots, waitf, lazyalloc);
if (amap == NULL)
return;
srcamap = entry->aref.ar_amap;
/*
* Make the new amap share the source amap's lock, and then lock
* both.
*/
amap->am_lock = srcamap->am_lock;
rw_obj_hold(amap->am_lock);
amap_lock(srcamap);
/*
* Re-check the reference count with the lock held. If it has
* dropped to one - we can take over the existing map.
*/
if (srcamap->am_ref == 1) {
/* Just take over the existing amap. */
entry->etype &= ~UVM_ET_NEEDSCOPY;
amap_unlock(srcamap);
/* Destroy the new (unused) amap. */
amap->am_ref--;
amap_free(amap);
return;
}
/*
* Copy the slots.
*/
for (lcv = 0; lcv < slots; lcv += n) {
srcslot = entry->aref.ar_pageoff + lcv;
i = UVM_AMAP_SLOTIDX(lcv);
j = UVM_AMAP_SLOTIDX(srcslot);
n = UVM_AMAP_CHUNK;
if (i > j)
n -= i;
else
n -= j;
if (lcv + n > slots)
n = slots - lcv;
srcchunk = amap_chunk_get(srcamap, srcslot, 0, PR_NOWAIT);
if (srcchunk == NULL)
continue;
chunk = amap_chunk_get(amap, lcv, 1, PR_NOWAIT);
if (chunk == NULL) {
/* amap_wipeout() releases the lock. */
amap->am_ref = 0;
amap_wipeout(amap);
return;
}
for (k = 0; k < n; i++, j++, k++) {
chunk->ac_anon[i] = anon = srcchunk->ac_anon[j];
if (anon == NULL)
continue;
KASSERT(anon->an_lock == srcamap->am_lock);
KASSERT(anon->an_ref > 0);
chunk->ac_usedmap |= (1 << i);
anon->an_ref++;
amap->am_nused++;
}
}
/*
* Drop our reference to the old amap (srcamap) and unlock.
* Since the reference count on srcamap is greater than one,
* (we checked above), it cannot drop to zero while it is locked.
*/
srcamap->am_ref--;
KASSERT(srcamap->am_ref > 0);
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
/*
* If we referenced any anons, then share the source amap's lock.
* Otherwise, we have nothing in common, so allocate a new one.
*/
KASSERT(amap->am_lock == srcamap->am_lock);
if (amap->am_nused == 0) {
rw_obj_free(amap->am_lock);
amap->am_lock = NULL;
}
amap_unlock(srcamap);
if (amap->am_lock == NULL)
amap_lock_alloc(amap);
/*
* Install new amap.
*/
entry->aref.ar_pageoff = 0;
entry->aref.ar_amap = amap;
entry->etype &= ~UVM_ET_NEEDSCOPY;
amap_list_insert(amap);
}
/*
* 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.
* => the parent and child vm_map must both be locked.
* => caller passes child's map/entry in to us
* => XXXCDC: out of memory should cause fork to fail, but there is
* currently no easy way to do this (needs fix)
*/
void
amap_cow_now(struct vm_map *map, struct vm_map_entry *entry)
{
struct vm_amap *amap = entry->aref.ar_amap;
int slot;
struct vm_anon *anon, *nanon;
struct vm_page *pg, *npg;
struct vm_amap_chunk *chunk;
/*
* 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);
AMAP_CHUNK_FOREACH(chunk, amap) {
int i, map = chunk->ac_usedmap;
for (i = ffs(map); i != 0; i = ffs(map)) {
slot = i - 1;
map ^= 1 << slot;
anon = chunk->ac_anon[slot];
pg = anon->an_page;
KASSERT(anon->an_lock == amap->am_lock);
/*
* The old page must be resident since the parent is
* wired.
*/
KASSERT(pg != NULL);
/*
* if the anon ref count is one, we are safe (the child
* has exclusive access to the page).
*/
if (anon->an_ref <= 1)
continue;
/*
* If the page is busy, then we have to unlock, wait for
* it and then restart.
*/
if (pg->pg_flags & PG_BUSY) {
uvm_pagewait(pg, amap->am_lock, "cownow");
goto ReStart;
}
/*
* Perform a copy-on-write.
* First - get a new anon and a page.
*/
nanon = uvm_analloc();
if (nanon != NULL) {
/* the new anon will share the amap's lock */
nanon->an_lock = amap->am_lock;
npg = uvm_pagealloc(NULL, 0, nanon, 0);
} else
npg = NULL; /* XXX: quiet gcc warning */
if (nanon == NULL || npg == NULL) {
/* out of memory */
amap_unlock(amap);
if (nanon != NULL) {
nanon->an_lock = NULL;
nanon->an_ref--;
KASSERT(nanon->an_ref == 0);
uvm_anfree(nanon);
}
uvm_wait("cownowpage");
goto ReStart;
}
/*
* Copy the data and replace anon with the new one.
* Also, setup its lock (share the with amap's lock).
*/
uvm_pagecopy(pg, npg);
anon->an_ref--;
KASSERT(anon->an_ref > 0);
chunk->ac_anon[slot] = nanon;
/*
* Drop PG_BUSY on new page. Since its owner was write
* locked all this time - it cannot be PG_RELEASED or
* PG_WANTED.
*/
atomic_clearbits_int(&npg->pg_flags, PG_BUSY|PG_FAKE);
UVM_PAGE_OWN(npg, NULL);
uvm_lock_pageq();
uvm_pageactivate(npg);
uvm_unlock_pageq();
}
}
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(struct vm_aref *origref, struct vm_aref *splitref, vaddr_t offset)
{
struct vm_amap *amap = origref->ar_amap;
int leftslots;
KASSERT(splitref->ar_amap == amap);
AMAP_B2SLOT(leftslots, offset);
if (leftslots == 0)
panic("amap_splitref: split at zero offset");
amap_lock(amap);
if (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 (amap->am_ppref == NULL)
amap_pp_establish(amap);
#endif
/* Note: not a share reference. */
amap->am_ref++;
splitref->ar_amap = amap;
splitref->ar_pageoff = origref->ar_pageoff + leftslots;
amap_unlock(amap);
}
#ifdef UVM_AMAP_PPREF
/*
* amap_pp_establish: add a ppref array to an amap, if possible.
*
* => amap should be locked by caller* => amap should be locked by caller
*/
void
amap_pp_establish(struct vm_amap *amap)
{
KASSERT(rw_write_held(amap->am_lock));
amap->am_ppref = mallocarray(amap->am_nslot, sizeof(int),
M_UVMAMAP, M_NOWAIT|M_ZERO);
if (amap->am_ppref == NULL) {
/* Failure - just do not use ppref. */
amap->am_ppref = PPREF_NONE;
return;
}
pp_setreflen(amap->am_ppref, 0, amap->am_ref, amap->am_nslot);
}
/*
* amap_pp_adjref: adjust reference count to a part of an amap using the
* per-page reference count array.
*
* => caller must check that ppref != PPREF_NONE before calling.
* => map and amap must be locked.
*/
void
amap_pp_adjref(struct vm_amap *amap, int curslot, vsize_t slotlen, int adjval)
{
int stopslot, *ppref, lcv, prevlcv;
int ref, len, prevref, prevlen;
KASSERT(rw_write_held(amap->am_lock));
stopslot = curslot + slotlen;
ppref = amap->am_ppref;
prevlcv = 0;
/*
* Advance to the correct place in the 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 */
}
prevlcv = lcv;
}
if (lcv != 0)
pp_getreflen(ppref, prevlcv, &prevref, &prevlen);
else {
/*
* Ensure that the "prevref == ref" test below always
* fails, since we are starting from the beginning of
* the ppref array; that is, there is no previous chunk.
*/
prevref = -1;
prevlen = 0;
}
/*
* Now adjust reference counts in range. Merge the first
* changed entry with the last unchanged entry if possible.
*/
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 += adjval;
if (ref < 0)
panic("amap_pp_adjref: negative reference count");
if (lcv == prevlcv + prevlen && ref == prevref) {
pp_setreflen(ppref, prevlcv, ref, prevlen + len);
} else {
pp_setreflen(ppref, lcv, ref, len);
}
if (ref == 0)
amap_wiperange(amap, lcv, len);
}
}
void
amap_wiperange_chunk(struct vm_amap *amap, struct vm_amap_chunk *chunk,
int slotoff, int slots)
{
int curslot, i, map;
int startbase, endbase;
struct vm_anon *anon;
startbase = AMAP_BASE_SLOT(slotoff);
endbase = AMAP_BASE_SLOT(slotoff + slots - 1);
map = chunk->ac_usedmap;
if (startbase == chunk->ac_baseslot)
map &= ~((1 << (slotoff - startbase)) - 1);
if (endbase == chunk->ac_baseslot)
map &= (1 << (slotoff + slots - endbase)) - 1;
for (i = ffs(map); i != 0; i = ffs(map)) {
int refs;
curslot = i - 1;
map ^= 1 << curslot;
chunk->ac_usedmap ^= 1 << curslot;
anon = chunk->ac_anon[curslot];
KASSERT(anon->an_lock == amap->am_lock);
/* remove it from the amap */
chunk->ac_anon[curslot] = NULL;
amap->am_nused--;
/* drop anon reference count */