forked from freebsd/freebsd-src
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathvm_phys.c
1894 lines (1727 loc) · 50.8 KB
/
vm_phys.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
/*-
* SPDX-License-Identifier: BSD-2-Clause-FreeBSD
*
* Copyright (c) 2002-2006 Rice University
* Copyright (c) 2007 Alan L. Cox <[email protected]>
* All rights reserved.
*
* This software was developed for the FreeBSD Project by Alan L. Cox,
* Olivier Crameri, Peter Druschel, Sitaram Iyer, and Juan Navarro.
*
* 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 COPYRIGHT HOLDERS AND CONTRIBUTORS
* ``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 COPYRIGHT
* HOLDERS OR CONTRIBUTORS 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.
*/
/*
* Physical memory system implementation
*
* Any external functions defined by this module are only to be used by the
* virtual memory system.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_ddb.h"
#include "opt_vm.h"
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/domainset.h>
#include <sys/lock.h>
#include <sys/kernel.h>
#include <sys/malloc.h>
#include <sys/mutex.h>
#include <sys/proc.h>
#include <sys/queue.h>
#include <sys/rwlock.h>
#include <sys/sbuf.h>
#include <sys/sysctl.h>
#include <sys/tree.h>
#include <sys/vmmeter.h>
#include <ddb/ddb.h>
#include <vm/vm.h>
#include <vm/vm_extern.h>
#include <vm/vm_param.h>
#include <vm/vm_kern.h>
#include <vm/vm_object.h>
#include <vm/vm_page.h>
#include <vm/vm_phys.h>
#include <vm/vm_pagequeue.h>
_Static_assert(sizeof(long) * NBBY >= VM_PHYSSEG_MAX,
"Too many physsegs.");
#ifdef NUMA
struct mem_affinity __read_mostly *mem_affinity;
int __read_mostly *mem_locality;
#endif
int __read_mostly vm_ndomains = 1;
domainset_t __read_mostly all_domains = DOMAINSET_T_INITIALIZER(0x1);
struct vm_phys_seg __read_mostly vm_phys_segs[VM_PHYSSEG_MAX];
int __read_mostly vm_phys_nsegs;
static struct vm_phys_seg vm_phys_early_segs[8];
static int vm_phys_early_nsegs;
struct vm_phys_fictitious_seg;
static int vm_phys_fictitious_cmp(struct vm_phys_fictitious_seg *,
struct vm_phys_fictitious_seg *);
RB_HEAD(fict_tree, vm_phys_fictitious_seg) vm_phys_fictitious_tree =
RB_INITIALIZER(&vm_phys_fictitious_tree);
struct vm_phys_fictitious_seg {
RB_ENTRY(vm_phys_fictitious_seg) node;
/* Memory region data */
vm_paddr_t start;
vm_paddr_t end;
vm_page_t first_page;
};
RB_GENERATE_STATIC(fict_tree, vm_phys_fictitious_seg, node,
vm_phys_fictitious_cmp);
static struct rwlock_padalign vm_phys_fictitious_reg_lock;
MALLOC_DEFINE(M_FICT_PAGES, "vm_fictitious", "Fictitious VM pages");
static struct vm_freelist __aligned(CACHE_LINE_SIZE)
vm_phys_free_queues[MAXMEMDOM][VM_NFREELIST][VM_NFREEPOOL]
[VM_NFREEORDER_MAX];
static int __read_mostly vm_nfreelists;
/*
* These "avail lists" are globals used to communicate boot-time physical
* memory layout to other parts of the kernel. Each physically contiguous
* region of memory is defined by a start address at an even index and an
* end address at the following odd index. Each list is terminated by a
* pair of zero entries.
*
* dump_avail tells the dump code what regions to include in a crash dump, and
* phys_avail is all of the remaining physical memory that is available for
* the vm system.
*
* Initially dump_avail and phys_avail are identical. Boot time memory
* allocations remove extents from phys_avail that may still be included
* in dumps.
*/
vm_paddr_t phys_avail[PHYS_AVAIL_COUNT];
vm_paddr_t dump_avail[PHYS_AVAIL_COUNT];
/*
* Provides the mapping from VM_FREELIST_* to free list indices (flind).
*/
static int __read_mostly vm_freelist_to_flind[VM_NFREELIST];
CTASSERT(VM_FREELIST_DEFAULT == 0);
#ifdef VM_FREELIST_DMA32
#define VM_DMA32_BOUNDARY ((vm_paddr_t)1 << 32)
#endif
/*
* Enforce the assumptions made by vm_phys_add_seg() and vm_phys_init() about
* the ordering of the free list boundaries.
*/
#if defined(VM_LOWMEM_BOUNDARY) && defined(VM_DMA32_BOUNDARY)
CTASSERT(VM_LOWMEM_BOUNDARY < VM_DMA32_BOUNDARY);
#endif
static int sysctl_vm_phys_free(SYSCTL_HANDLER_ARGS);
SYSCTL_OID(_vm, OID_AUTO, phys_free,
CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
sysctl_vm_phys_free, "A",
"Phys Free Info");
static int sysctl_vm_phys_segs(SYSCTL_HANDLER_ARGS);
SYSCTL_OID(_vm, OID_AUTO, phys_segs,
CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
sysctl_vm_phys_segs, "A",
"Phys Seg Info");
#ifdef NUMA
static int sysctl_vm_phys_locality(SYSCTL_HANDLER_ARGS);
SYSCTL_OID(_vm, OID_AUTO, phys_locality,
CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
sysctl_vm_phys_locality, "A",
"Phys Locality Info");
#endif
SYSCTL_INT(_vm, OID_AUTO, ndomains, CTLFLAG_RD,
&vm_ndomains, 0, "Number of physical memory domains available.");
static void _vm_phys_create_seg(vm_paddr_t start, vm_paddr_t end, int domain);
static void vm_phys_create_seg(vm_paddr_t start, vm_paddr_t end);
static void vm_phys_split_pages(vm_page_t m, int oind, struct vm_freelist *fl,
int order, int tail);
/*
* Red-black tree helpers for vm fictitious range management.
*/
static inline int
vm_phys_fictitious_in_range(struct vm_phys_fictitious_seg *p,
struct vm_phys_fictitious_seg *range)
{
KASSERT(range->start != 0 && range->end != 0,
("Invalid range passed on search for vm_fictitious page"));
if (p->start >= range->end)
return (1);
if (p->start < range->start)
return (-1);
return (0);
}
static int
vm_phys_fictitious_cmp(struct vm_phys_fictitious_seg *p1,
struct vm_phys_fictitious_seg *p2)
{
/* Check if this is a search for a page */
if (p1->end == 0)
return (vm_phys_fictitious_in_range(p1, p2));
KASSERT(p2->end != 0,
("Invalid range passed as second parameter to vm fictitious comparison"));
/* Searching to add a new range */
if (p1->end <= p2->start)
return (-1);
if (p1->start >= p2->end)
return (1);
panic("Trying to add overlapping vm fictitious ranges:\n"
"[%#jx:%#jx] and [%#jx:%#jx]", (uintmax_t)p1->start,
(uintmax_t)p1->end, (uintmax_t)p2->start, (uintmax_t)p2->end);
}
int
vm_phys_domain_match(int prefer, vm_paddr_t low, vm_paddr_t high)
{
#ifdef NUMA
domainset_t mask;
int i;
if (vm_ndomains == 1 || mem_affinity == NULL)
return (0);
DOMAINSET_ZERO(&mask);
/*
* Check for any memory that overlaps low, high.
*/
for (i = 0; mem_affinity[i].end != 0; i++)
if (mem_affinity[i].start <= high &&
mem_affinity[i].end >= low)
DOMAINSET_SET(mem_affinity[i].domain, &mask);
if (prefer != -1 && DOMAINSET_ISSET(prefer, &mask))
return (prefer);
if (DOMAINSET_EMPTY(&mask))
panic("vm_phys_domain_match: Impossible constraint");
return (DOMAINSET_FFS(&mask) - 1);
#else
return (0);
#endif
}
/*
* Outputs the state of the physical memory allocator, specifically,
* the amount of physical memory in each free list.
*/
static int
sysctl_vm_phys_free(SYSCTL_HANDLER_ARGS)
{
struct sbuf sbuf;
struct vm_freelist *fl;
int dom, error, flind, oind, pind;
error = sysctl_wire_old_buffer(req, 0);
if (error != 0)
return (error);
sbuf_new_for_sysctl(&sbuf, NULL, 128 * vm_ndomains, req);
for (dom = 0; dom < vm_ndomains; dom++) {
sbuf_printf(&sbuf,"\nDOMAIN %d:\n", dom);
for (flind = 0; flind < vm_nfreelists; flind++) {
sbuf_printf(&sbuf, "\nFREE LIST %d:\n"
"\n ORDER (SIZE) | NUMBER"
"\n ", flind);
for (pind = 0; pind < VM_NFREEPOOL; pind++)
sbuf_printf(&sbuf, " | POOL %d", pind);
sbuf_printf(&sbuf, "\n-- ");
for (pind = 0; pind < VM_NFREEPOOL; pind++)
sbuf_printf(&sbuf, "-- -- ");
sbuf_printf(&sbuf, "--\n");
for (oind = VM_NFREEORDER - 1; oind >= 0; oind--) {
sbuf_printf(&sbuf, " %2d (%6dK)", oind,
1 << (PAGE_SHIFT - 10 + oind));
for (pind = 0; pind < VM_NFREEPOOL; pind++) {
fl = vm_phys_free_queues[dom][flind][pind];
sbuf_printf(&sbuf, " | %6d",
fl[oind].lcnt);
}
sbuf_printf(&sbuf, "\n");
}
}
}
error = sbuf_finish(&sbuf);
sbuf_delete(&sbuf);
return (error);
}
/*
* Outputs the set of physical memory segments.
*/
static int
sysctl_vm_phys_segs(SYSCTL_HANDLER_ARGS)
{
struct sbuf sbuf;
struct vm_phys_seg *seg;
int error, segind;
error = sysctl_wire_old_buffer(req, 0);
if (error != 0)
return (error);
sbuf_new_for_sysctl(&sbuf, NULL, 128, req);
for (segind = 0; segind < vm_phys_nsegs; segind++) {
sbuf_printf(&sbuf, "\nSEGMENT %d:\n\n", segind);
seg = &vm_phys_segs[segind];
sbuf_printf(&sbuf, "start: %#jx\n",
(uintmax_t)seg->start);
sbuf_printf(&sbuf, "end: %#jx\n",
(uintmax_t)seg->end);
sbuf_printf(&sbuf, "domain: %d\n", seg->domain);
sbuf_printf(&sbuf, "free list: %p\n", seg->free_queues);
}
error = sbuf_finish(&sbuf);
sbuf_delete(&sbuf);
return (error);
}
/*
* Return affinity, or -1 if there's no affinity information.
*/
int
vm_phys_mem_affinity(int f, int t)
{
#ifdef NUMA
if (mem_locality == NULL)
return (-1);
if (f >= vm_ndomains || t >= vm_ndomains)
return (-1);
return (mem_locality[f * vm_ndomains + t]);
#else
return (-1);
#endif
}
#ifdef NUMA
/*
* Outputs the VM locality table.
*/
static int
sysctl_vm_phys_locality(SYSCTL_HANDLER_ARGS)
{
struct sbuf sbuf;
int error, i, j;
error = sysctl_wire_old_buffer(req, 0);
if (error != 0)
return (error);
sbuf_new_for_sysctl(&sbuf, NULL, 128, req);
sbuf_printf(&sbuf, "\n");
for (i = 0; i < vm_ndomains; i++) {
sbuf_printf(&sbuf, "%d: ", i);
for (j = 0; j < vm_ndomains; j++) {
sbuf_printf(&sbuf, "%d ", vm_phys_mem_affinity(i, j));
}
sbuf_printf(&sbuf, "\n");
}
error = sbuf_finish(&sbuf);
sbuf_delete(&sbuf);
return (error);
}
#endif
static void
vm_freelist_add(struct vm_freelist *fl, vm_page_t m, int order, int tail)
{
m->order = order;
if (tail)
TAILQ_INSERT_TAIL(&fl[order].pl, m, listq);
else
TAILQ_INSERT_HEAD(&fl[order].pl, m, listq);
fl[order].lcnt++;
}
static void
vm_freelist_rem(struct vm_freelist *fl, vm_page_t m, int order)
{
TAILQ_REMOVE(&fl[order].pl, m, listq);
fl[order].lcnt--;
m->order = VM_NFREEORDER;
}
/*
* Create a physical memory segment.
*/
static void
_vm_phys_create_seg(vm_paddr_t start, vm_paddr_t end, int domain)
{
struct vm_phys_seg *seg;
KASSERT(vm_phys_nsegs < VM_PHYSSEG_MAX,
("vm_phys_create_seg: increase VM_PHYSSEG_MAX"));
KASSERT(domain >= 0 && domain < vm_ndomains,
("vm_phys_create_seg: invalid domain provided"));
seg = &vm_phys_segs[vm_phys_nsegs++];
while (seg > vm_phys_segs && (seg - 1)->start >= end) {
*seg = *(seg - 1);
seg--;
}
seg->start = start;
seg->end = end;
seg->domain = domain;
}
static void
vm_phys_create_seg(vm_paddr_t start, vm_paddr_t end)
{
#ifdef NUMA
int i;
if (mem_affinity == NULL) {
_vm_phys_create_seg(start, end, 0);
return;
}
for (i = 0;; i++) {
if (mem_affinity[i].end == 0)
panic("Reached end of affinity info");
if (mem_affinity[i].end <= start)
continue;
if (mem_affinity[i].start > start)
panic("No affinity info for start %jx",
(uintmax_t)start);
if (mem_affinity[i].end >= end) {
_vm_phys_create_seg(start, end,
mem_affinity[i].domain);
break;
}
_vm_phys_create_seg(start, mem_affinity[i].end,
mem_affinity[i].domain);
start = mem_affinity[i].end;
}
#else
_vm_phys_create_seg(start, end, 0);
#endif
}
/*
* Add a physical memory segment.
*/
void
vm_phys_add_seg(vm_paddr_t start, vm_paddr_t end)
{
vm_paddr_t paddr;
KASSERT((start & PAGE_MASK) == 0,
("vm_phys_define_seg: start is not page aligned"));
KASSERT((end & PAGE_MASK) == 0,
("vm_phys_define_seg: end is not page aligned"));
/*
* Split the physical memory segment if it spans two or more free
* list boundaries.
*/
paddr = start;
#ifdef VM_FREELIST_LOWMEM
if (paddr < VM_LOWMEM_BOUNDARY && end > VM_LOWMEM_BOUNDARY) {
vm_phys_create_seg(paddr, VM_LOWMEM_BOUNDARY);
paddr = VM_LOWMEM_BOUNDARY;
}
#endif
#ifdef VM_FREELIST_DMA32
if (paddr < VM_DMA32_BOUNDARY && end > VM_DMA32_BOUNDARY) {
vm_phys_create_seg(paddr, VM_DMA32_BOUNDARY);
paddr = VM_DMA32_BOUNDARY;
}
#endif
vm_phys_create_seg(paddr, end);
}
/*
* Initialize the physical memory allocator.
*
* Requires that vm_page_array is initialized!
*/
void
vm_phys_init(void)
{
struct vm_freelist *fl;
struct vm_phys_seg *end_seg, *prev_seg, *seg, *tmp_seg;
#if defined(VM_DMA32_NPAGES_THRESHOLD) || defined(VM_PHYSSEG_SPARSE)
u_long npages;
#endif
int dom, flind, freelist, oind, pind, segind;
/*
* Compute the number of free lists, and generate the mapping from the
* manifest constants VM_FREELIST_* to the free list indices.
*
* Initially, the entries of vm_freelist_to_flind[] are set to either
* 0 or 1 to indicate which free lists should be created.
*/
#ifdef VM_DMA32_NPAGES_THRESHOLD
npages = 0;
#endif
for (segind = vm_phys_nsegs - 1; segind >= 0; segind--) {
seg = &vm_phys_segs[segind];
#ifdef VM_FREELIST_LOWMEM
if (seg->end <= VM_LOWMEM_BOUNDARY)
vm_freelist_to_flind[VM_FREELIST_LOWMEM] = 1;
else
#endif
#ifdef VM_FREELIST_DMA32
if (
#ifdef VM_DMA32_NPAGES_THRESHOLD
/*
* Create the DMA32 free list only if the amount of
* physical memory above physical address 4G exceeds the
* given threshold.
*/
npages > VM_DMA32_NPAGES_THRESHOLD &&
#endif
seg->end <= VM_DMA32_BOUNDARY)
vm_freelist_to_flind[VM_FREELIST_DMA32] = 1;
else
#endif
{
#ifdef VM_DMA32_NPAGES_THRESHOLD
npages += atop(seg->end - seg->start);
#endif
vm_freelist_to_flind[VM_FREELIST_DEFAULT] = 1;
}
}
/* Change each entry into a running total of the free lists. */
for (freelist = 1; freelist < VM_NFREELIST; freelist++) {
vm_freelist_to_flind[freelist] +=
vm_freelist_to_flind[freelist - 1];
}
vm_nfreelists = vm_freelist_to_flind[VM_NFREELIST - 1];
KASSERT(vm_nfreelists > 0, ("vm_phys_init: no free lists"));
/* Change each entry into a free list index. */
for (freelist = 0; freelist < VM_NFREELIST; freelist++)
vm_freelist_to_flind[freelist]--;
/*
* Initialize the first_page and free_queues fields of each physical
* memory segment.
*/
#ifdef VM_PHYSSEG_SPARSE
npages = 0;
#endif
for (segind = 0; segind < vm_phys_nsegs; segind++) {
seg = &vm_phys_segs[segind];
#ifdef VM_PHYSSEG_SPARSE
seg->first_page = &vm_page_array[npages];
npages += atop(seg->end - seg->start);
#else
seg->first_page = PHYS_TO_VM_PAGE(seg->start);
#endif
#ifdef VM_FREELIST_LOWMEM
if (seg->end <= VM_LOWMEM_BOUNDARY) {
flind = vm_freelist_to_flind[VM_FREELIST_LOWMEM];
KASSERT(flind >= 0,
("vm_phys_init: LOWMEM flind < 0"));
} else
#endif
#ifdef VM_FREELIST_DMA32
if (seg->end <= VM_DMA32_BOUNDARY) {
flind = vm_freelist_to_flind[VM_FREELIST_DMA32];
KASSERT(flind >= 0,
("vm_phys_init: DMA32 flind < 0"));
} else
#endif
{
flind = vm_freelist_to_flind[VM_FREELIST_DEFAULT];
KASSERT(flind >= 0,
("vm_phys_init: DEFAULT flind < 0"));
}
seg->free_queues = &vm_phys_free_queues[seg->domain][flind];
}
/*
* Coalesce physical memory segments that are contiguous and share the
* same per-domain free queues.
*/
prev_seg = vm_phys_segs;
seg = &vm_phys_segs[1];
end_seg = &vm_phys_segs[vm_phys_nsegs];
while (seg < end_seg) {
if (prev_seg->end == seg->start &&
prev_seg->free_queues == seg->free_queues) {
prev_seg->end = seg->end;
KASSERT(prev_seg->domain == seg->domain,
("vm_phys_init: free queues cannot span domains"));
vm_phys_nsegs--;
end_seg--;
for (tmp_seg = seg; tmp_seg < end_seg; tmp_seg++)
*tmp_seg = *(tmp_seg + 1);
} else {
prev_seg = seg;
seg++;
}
}
/*
* Initialize the free queues.
*/
for (dom = 0; dom < vm_ndomains; dom++) {
for (flind = 0; flind < vm_nfreelists; flind++) {
for (pind = 0; pind < VM_NFREEPOOL; pind++) {
fl = vm_phys_free_queues[dom][flind][pind];
for (oind = 0; oind < VM_NFREEORDER; oind++)
TAILQ_INIT(&fl[oind].pl);
}
}
}
rw_init(&vm_phys_fictitious_reg_lock, "vmfctr");
}
/*
* Register info about the NUMA topology of the system.
*
* Invoked by platform-dependent code prior to vm_phys_init().
*/
void
vm_phys_register_domains(int ndomains, struct mem_affinity *affinity,
int *locality)
{
#ifdef NUMA
int d, i;
/*
* For now the only override value that we support is 1, which
* effectively disables NUMA-awareness in the allocators.
*/
d = 0;
TUNABLE_INT_FETCH("vm.numa.disabled", &d);
if (d)
ndomains = 1;
if (ndomains > 1) {
vm_ndomains = ndomains;
mem_affinity = affinity;
mem_locality = locality;
}
for (i = 0; i < vm_ndomains; i++)
DOMAINSET_SET(i, &all_domains);
#else
(void)ndomains;
(void)affinity;
(void)locality;
#endif
}
/*
* Split a contiguous, power of two-sized set of physical pages.
*
* When this function is called by a page allocation function, the caller
* should request insertion at the head unless the order [order, oind) queues
* are known to be empty. The objective being to reduce the likelihood of
* long-term fragmentation by promoting contemporaneous allocation and
* (hopefully) deallocation.
*/
static __inline void
vm_phys_split_pages(vm_page_t m, int oind, struct vm_freelist *fl, int order,
int tail)
{
vm_page_t m_buddy;
while (oind > order) {
oind--;
m_buddy = &m[1 << oind];
KASSERT(m_buddy->order == VM_NFREEORDER,
("vm_phys_split_pages: page %p has unexpected order %d",
m_buddy, m_buddy->order));
vm_freelist_add(fl, m_buddy, oind, tail);
}
}
/*
* Add the physical pages [m, m + npages) at the end of a power-of-two aligned
* and sized set to the specified free list.
*
* When this function is called by a page allocation function, the caller
* should request insertion at the head unless the lower-order queues are
* known to be empty. The objective being to reduce the likelihood of long-
* term fragmentation by promoting contemporaneous allocation and (hopefully)
* deallocation.
*
* The physical page m's buddy must not be free.
*/
static void
vm_phys_enq_range(vm_page_t m, u_int npages, struct vm_freelist *fl, int tail)
{
u_int n;
int order;
KASSERT(npages > 0, ("vm_phys_enq_range: npages is 0"));
KASSERT(((VM_PAGE_TO_PHYS(m) + npages * PAGE_SIZE) &
((PAGE_SIZE << (fls(npages) - 1)) - 1)) == 0,
("vm_phys_enq_range: page %p and npages %u are misaligned",
m, npages));
do {
KASSERT(m->order == VM_NFREEORDER,
("vm_phys_enq_range: page %p has unexpected order %d",
m, m->order));
order = ffs(npages) - 1;
KASSERT(order < VM_NFREEORDER,
("vm_phys_enq_range: order %d is out of range", order));
vm_freelist_add(fl, m, order, tail);
n = 1 << order;
m += n;
npages -= n;
} while (npages > 0);
}
/*
* Set the pool for a contiguous, power of two-sized set of physical pages.
*/
static void
vm_phys_set_pool(int pool, vm_page_t m, int order)
{
vm_page_t m_tmp;
for (m_tmp = m; m_tmp < &m[1 << order]; m_tmp++)
m_tmp->pool = pool;
}
/*
* Tries to allocate the specified number of pages from the specified pool
* within the specified domain. Returns the actual number of allocated pages
* and a pointer to each page through the array ma[].
*
* The returned pages may not be physically contiguous. However, in contrast
* to performing multiple, back-to-back calls to vm_phys_alloc_pages(..., 0),
* calling this function once to allocate the desired number of pages will
* avoid wasted time in vm_phys_split_pages().
*
* The free page queues for the specified domain must be locked.
*/
int
vm_phys_alloc_npages(int domain, int pool, int npages, vm_page_t ma[])
{
struct vm_freelist *alt, *fl;
vm_page_t m;
int avail, end, flind, freelist, i, need, oind, pind;
KASSERT(domain >= 0 && domain < vm_ndomains,
("vm_phys_alloc_npages: domain %d is out of range", domain));
KASSERT(pool < VM_NFREEPOOL,
("vm_phys_alloc_npages: pool %d is out of range", pool));
KASSERT(npages <= 1 << (VM_NFREEORDER - 1),
("vm_phys_alloc_npages: npages %d is out of range", npages));
vm_domain_free_assert_locked(VM_DOMAIN(domain));
i = 0;
for (freelist = 0; freelist < VM_NFREELIST; freelist++) {
flind = vm_freelist_to_flind[freelist];
if (flind < 0)
continue;
fl = vm_phys_free_queues[domain][flind][pool];
for (oind = 0; oind < VM_NFREEORDER; oind++) {
while ((m = TAILQ_FIRST(&fl[oind].pl)) != NULL) {
vm_freelist_rem(fl, m, oind);
avail = 1 << oind;
need = imin(npages - i, avail);
for (end = i + need; i < end;)
ma[i++] = m++;
if (need < avail) {
/*
* Return excess pages to fl. Its
* order [0, oind) queues are empty.
*/
vm_phys_enq_range(m, avail - need, fl,
1);
return (npages);
} else if (i == npages)
return (npages);
}
}
for (oind = VM_NFREEORDER - 1; oind >= 0; oind--) {
for (pind = 0; pind < VM_NFREEPOOL; pind++) {
alt = vm_phys_free_queues[domain][flind][pind];
while ((m = TAILQ_FIRST(&alt[oind].pl)) !=
NULL) {
vm_freelist_rem(alt, m, oind);
vm_phys_set_pool(pool, m, oind);
avail = 1 << oind;
need = imin(npages - i, avail);
for (end = i + need; i < end;)
ma[i++] = m++;
if (need < avail) {
/*
* Return excess pages to fl.
* Its order [0, oind) queues
* are empty.
*/
vm_phys_enq_range(m, avail -
need, fl, 1);
return (npages);
} else if (i == npages)
return (npages);
}
}
}
}
return (i);
}
/*
* Allocate a contiguous, power of two-sized set of physical pages
* from the free lists.
*
* The free page queues must be locked.
*/
vm_page_t
vm_phys_alloc_pages(int domain, int pool, int order)
{
vm_page_t m;
int freelist;
for (freelist = 0; freelist < VM_NFREELIST; freelist++) {
m = vm_phys_alloc_freelist_pages(domain, freelist, pool, order);
if (m != NULL)
return (m);
}
return (NULL);
}
/*
* Allocate a contiguous, power of two-sized set of physical pages from the
* specified free list. The free list must be specified using one of the
* manifest constants VM_FREELIST_*.
*
* The free page queues must be locked.
*/
vm_page_t
vm_phys_alloc_freelist_pages(int domain, int freelist, int pool, int order)
{
struct vm_freelist *alt, *fl;
vm_page_t m;
int oind, pind, flind;
KASSERT(domain >= 0 && domain < vm_ndomains,
("vm_phys_alloc_freelist_pages: domain %d is out of range",
domain));
KASSERT(freelist < VM_NFREELIST,
("vm_phys_alloc_freelist_pages: freelist %d is out of range",
freelist));
KASSERT(pool < VM_NFREEPOOL,
("vm_phys_alloc_freelist_pages: pool %d is out of range", pool));
KASSERT(order < VM_NFREEORDER,
("vm_phys_alloc_freelist_pages: order %d is out of range", order));
flind = vm_freelist_to_flind[freelist];
/* Check if freelist is present */
if (flind < 0)
return (NULL);
vm_domain_free_assert_locked(VM_DOMAIN(domain));
fl = &vm_phys_free_queues[domain][flind][pool][0];
for (oind = order; oind < VM_NFREEORDER; oind++) {
m = TAILQ_FIRST(&fl[oind].pl);
if (m != NULL) {
vm_freelist_rem(fl, m, oind);
/* The order [order, oind) queues are empty. */
vm_phys_split_pages(m, oind, fl, order, 1);
return (m);
}
}
/*
* The given pool was empty. Find the largest
* contiguous, power-of-two-sized set of pages in any
* pool. Transfer these pages to the given pool, and
* use them to satisfy the allocation.
*/
for (oind = VM_NFREEORDER - 1; oind >= order; oind--) {
for (pind = 0; pind < VM_NFREEPOOL; pind++) {
alt = &vm_phys_free_queues[domain][flind][pind][0];
m = TAILQ_FIRST(&alt[oind].pl);
if (m != NULL) {
vm_freelist_rem(alt, m, oind);
vm_phys_set_pool(pool, m, oind);
/* The order [order, oind) queues are empty. */
vm_phys_split_pages(m, oind, fl, order, 1);
return (m);
}
}
}
return (NULL);
}
/*
* Find the vm_page corresponding to the given physical address.
*/
vm_page_t
vm_phys_paddr_to_vm_page(vm_paddr_t pa)
{
struct vm_phys_seg *seg;
int segind;
for (segind = 0; segind < vm_phys_nsegs; segind++) {
seg = &vm_phys_segs[segind];
if (pa >= seg->start && pa < seg->end)
return (&seg->first_page[atop(pa - seg->start)]);
}
return (NULL);
}
vm_page_t
vm_phys_fictitious_to_vm_page(vm_paddr_t pa)
{
struct vm_phys_fictitious_seg tmp, *seg;
vm_page_t m;
m = NULL;
tmp.start = pa;
tmp.end = 0;
rw_rlock(&vm_phys_fictitious_reg_lock);
seg = RB_FIND(fict_tree, &vm_phys_fictitious_tree, &tmp);
rw_runlock(&vm_phys_fictitious_reg_lock);
if (seg == NULL)
return (NULL);
m = &seg->first_page[atop(pa - seg->start)];
KASSERT((m->flags & PG_FICTITIOUS) != 0, ("%p not fictitious", m));
return (m);
}
static inline void
vm_phys_fictitious_init_range(vm_page_t range, vm_paddr_t start,
long page_count, vm_memattr_t memattr)
{
long i;
bzero(range, page_count * sizeof(*range));
for (i = 0; i < page_count; i++) {
vm_page_initfake(&range[i], start + PAGE_SIZE * i, memattr);
range[i].oflags &= ~VPO_UNMANAGED;
range[i].busy_lock = VPB_UNBUSIED;
}
}
int
vm_phys_fictitious_reg_range(vm_paddr_t start, vm_paddr_t end,
vm_memattr_t memattr)
{
struct vm_phys_fictitious_seg *seg;
vm_page_t fp;
long page_count;
#ifdef VM_PHYSSEG_DENSE
long pi, pe;
long dpage_count;
#endif
KASSERT(start < end,
("Start of segment isn't less than end (start: %jx end: %jx)",
(uintmax_t)start, (uintmax_t)end));
page_count = (end - start) / PAGE_SIZE;
#ifdef VM_PHYSSEG_DENSE
pi = atop(start);
pe = atop(end);
if (pi >= first_page && (pi - first_page) < vm_page_array_size) {
fp = &vm_page_array[pi - first_page];
if ((pe - first_page) > vm_page_array_size) {
/*
* We have a segment that starts inside
* of vm_page_array, but ends outside of it.
*
* Use vm_page_array pages for those that are
* inside of the vm_page_array range, and
* allocate the remaining ones.
*/
dpage_count = vm_page_array_size - (pi - first_page);
vm_phys_fictitious_init_range(fp, start, dpage_count,
memattr);
page_count -= dpage_count;
start += ptoa(dpage_count);
goto alloc;
}
/*
* We can allocate the full range from vm_page_array,
* so there's no need to register the range in the tree.
*/
vm_phys_fictitious_init_range(fp, start, page_count, memattr);
return (0);
} else if (pe > first_page && (pe - first_page) < vm_page_array_size) {
/*
* We have a segment that ends inside of vm_page_array,
* but starts outside of it.
*/
fp = &vm_page_array[0];
dpage_count = pe - first_page;
vm_phys_fictitious_init_range(fp, ptoa(first_page), dpage_count,
memattr);