aboutsummaryrefslogtreecommitdiff
path: root/sys/vm/vm_reserv.c
blob: c088951641717503086fca5a4f3c742d28241359 (plain) (blame)
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
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
/*-
 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
 *
 * Copyright (c) 2002-2006 Rice University
 * Copyright (c) 2007-2011 Alan L. Cox <alc@cs.rice.edu>
 * 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.
 */

/*
 *	Superpage reservation management module
 *
 * 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_vm.h"

#include <sys/param.h>
#include <sys/kernel.h>
#include <sys/lock.h>
#include <sys/malloc.h>
#include <sys/mutex.h>
#include <sys/queue.h>
#include <sys/rwlock.h>
#include <sys/sbuf.h>
#include <sys/sysctl.h>
#include <sys/systm.h>
#include <sys/counter.h>
#include <sys/ktr.h>
#include <sys/vmmeter.h>
#include <sys/smp.h>

#include <vm/vm.h>
#include <vm/vm_param.h>
#include <vm/vm_object.h>
#include <vm/vm_page.h>
#include <vm/vm_pageout.h>
#include <vm/vm_phys.h>
#include <vm/vm_pagequeue.h>
#include <vm/vm_radix.h>
#include <vm/vm_reserv.h>

/*
 * The reservation system supports the speculative allocation of large physical
 * pages ("superpages").  Speculative allocation enables the fully automatic
 * utilization of superpages by the virtual memory system.  In other words, no
 * programmatic directives are required to use superpages.
 */

#if VM_NRESERVLEVEL > 0

#ifndef VM_LEVEL_0_ORDER_MAX
#define	VM_LEVEL_0_ORDER_MAX	VM_LEVEL_0_ORDER
#endif

/*
 * The number of small pages that are contained in a level 0 reservation
 */
#define	VM_LEVEL_0_NPAGES	(1 << VM_LEVEL_0_ORDER)
#define	VM_LEVEL_0_NPAGES_MAX	(1 << VM_LEVEL_0_ORDER_MAX)

/*
 * The number of bits by which a physical address is shifted to obtain the
 * reservation number
 */
#define	VM_LEVEL_0_SHIFT	(VM_LEVEL_0_ORDER + PAGE_SHIFT)

/*
 * The size of a level 0 reservation in bytes
 */
#define	VM_LEVEL_0_SIZE		(1 << VM_LEVEL_0_SHIFT)

/*
 * Computes the index of the small page underlying the given (object, pindex)
 * within the reservation's array of small pages.
 */
#define	VM_RESERV_INDEX(object, pindex)	\
    (((object)->pg_color + (pindex)) & (VM_LEVEL_0_NPAGES - 1))

/*
 * The size of a population map entry
 */
typedef	u_long		popmap_t;

/*
 * The number of bits in a population map entry
 */
#define	NBPOPMAP	(NBBY * sizeof(popmap_t))

/*
 * The number of population map entries in a reservation
 */
#define	NPOPMAP		howmany(VM_LEVEL_0_NPAGES, NBPOPMAP)
#define	NPOPMAP_MAX	howmany(VM_LEVEL_0_NPAGES_MAX, NBPOPMAP)

/*
 * Number of elapsed ticks before we update the LRU queue position.  Used
 * to reduce contention and churn on the list.
 */
#define	PARTPOPSLOP	1

/*
 * Clear a bit in the population map.
 */
static __inline void
popmap_clear(popmap_t popmap[], int i)
{

	popmap[i / NBPOPMAP] &= ~(1UL << (i % NBPOPMAP));
}

/*
 * Set a bit in the population map.
 */
static __inline void
popmap_set(popmap_t popmap[], int i)
{

	popmap[i / NBPOPMAP] |= 1UL << (i % NBPOPMAP);
}

/*
 * Is a bit in the population map clear?
 */
static __inline boolean_t
popmap_is_clear(popmap_t popmap[], int i)
{

	return ((popmap[i / NBPOPMAP] & (1UL << (i % NBPOPMAP))) == 0);
}

/*
 * Is a bit in the population map set?
 */
static __inline boolean_t
popmap_is_set(popmap_t popmap[], int i)
{

	return ((popmap[i / NBPOPMAP] & (1UL << (i % NBPOPMAP))) != 0);
}

/*
 * The reservation structure
 *
 * A reservation structure is constructed whenever a large physical page is
 * speculatively allocated to an object.  The reservation provides the small
 * physical pages for the range [pindex, pindex + VM_LEVEL_0_NPAGES) of offsets
 * within that object.  The reservation's "popcnt" tracks the number of these
 * small physical pages that are in use at any given time.  When and if the
 * reservation is not fully utilized, it appears in the queue of partially
 * populated reservations.  The reservation always appears on the containing
 * object's list of reservations.
 *
 * A partially populated reservation can be broken and reclaimed at any time.
 *
 * c - constant after boot
 * d - vm_reserv_domain_lock
 * o - vm_reserv_object_lock
 * r - vm_reserv_lock
 * s - vm_reserv_domain_scan_lock
 */
struct vm_reserv {
	struct mtx	lock;			/* reservation lock. */
	TAILQ_ENTRY(vm_reserv) partpopq;	/* (d, r) per-domain queue. */
	LIST_ENTRY(vm_reserv) objq;		/* (o, r) object queue */
	vm_object_t	object;			/* (o, r) containing object */
	vm_pindex_t	pindex;			/* (o, r) offset in object */
	vm_page_t	pages;			/* (c) first page  */
	uint16_t	popcnt;			/* (r) # of pages in use */
	uint8_t		domain;			/* (c) NUMA domain. */
	char		inpartpopq;		/* (d, r) */
	int		lasttick;		/* (r) last pop update tick. */
	popmap_t	popmap[NPOPMAP_MAX];	/* (r) bit vector, used pages */
};

TAILQ_HEAD(vm_reserv_queue, vm_reserv);

#define	vm_reserv_lockptr(rv)		(&(rv)->lock)
#define	vm_reserv_assert_locked(rv)					\
	    mtx_assert(vm_reserv_lockptr(rv), MA_OWNED)
#define	vm_reserv_lock(rv)		mtx_lock(vm_reserv_lockptr(rv))
#define	vm_reserv_trylock(rv)		mtx_trylock(vm_reserv_lockptr(rv))
#define	vm_reserv_unlock(rv)		mtx_unlock(vm_reserv_lockptr(rv))

/*
 * The reservation array
 *
 * This array is analoguous in function to vm_page_array.  It differs in the
 * respect that it may contain a greater number of useful reservation
 * structures than there are (physical) superpages.  These "invalid"
 * reservation structures exist to trade-off space for time in the
 * implementation of vm_reserv_from_page().  Invalid reservation structures are
 * distinguishable from "valid" reservation structures by inspecting the
 * reservation's "pages" field.  Invalid reservation structures have a NULL
 * "pages" field.
 *
 * vm_reserv_from_page() maps a small (physical) page to an element of this
 * array by computing a physical reservation number from the page's physical
 * address.  The physical reservation number is used as the array index.
 *
 * An "active" reservation is a valid reservation structure that has a non-NULL
 * "object" field and a non-zero "popcnt" field.  In other words, every active
 * reservation belongs to a particular object.  Moreover, every active
 * reservation has an entry in the containing object's list of reservations.  
 */
static vm_reserv_t vm_reserv_array;

/*
 * The per-domain partially populated reservation queues
 *
 * These queues enable the fast recovery of an unused free small page from a
 * partially populated reservation.  The reservation at the head of a queue
 * is the least recently changed, partially populated reservation.
 *
 * Access to this queue is synchronized by the per-domain reservation lock.
 * Threads reclaiming free pages from the queue must hold the per-domain scan
 * lock.
 */
struct vm_reserv_domain {
	struct mtx 		lock;
	struct vm_reserv_queue	partpop;	/* (d) */
	struct vm_reserv	marker;		/* (d, s) scan marker/lock */
} __aligned(CACHE_LINE_SIZE);

static struct vm_reserv_domain vm_rvd[MAXMEMDOM];

#define	vm_reserv_domain_lockptr(d)	(&vm_rvd[(d)].lock)
#define	vm_reserv_domain_assert_locked(d)	\
	mtx_assert(vm_reserv_domain_lockptr(d), MA_OWNED)
#define	vm_reserv_domain_lock(d)	mtx_lock(vm_reserv_domain_lockptr(d))
#define	vm_reserv_domain_unlock(d)	mtx_unlock(vm_reserv_domain_lockptr(d))

#define	vm_reserv_domain_scan_lock(d)	mtx_lock(&vm_rvd[(d)].marker.lock)
#define	vm_reserv_domain_scan_unlock(d)	mtx_unlock(&vm_rvd[(d)].marker.lock)

static SYSCTL_NODE(_vm, OID_AUTO, reserv, CTLFLAG_RD | CTLFLAG_MPSAFE, 0,
    "Reservation Info");

static COUNTER_U64_DEFINE_EARLY(vm_reserv_broken);
SYSCTL_COUNTER_U64(_vm_reserv, OID_AUTO, broken, CTLFLAG_RD,
    &vm_reserv_broken, "Cumulative number of broken reservations");

static COUNTER_U64_DEFINE_EARLY(vm_reserv_freed);
SYSCTL_COUNTER_U64(_vm_reserv, OID_AUTO, freed, CTLFLAG_RD,
    &vm_reserv_freed, "Cumulative number of freed reservations");

static int sysctl_vm_reserv_fullpop(SYSCTL_HANDLER_ARGS);

SYSCTL_PROC(_vm_reserv, OID_AUTO, fullpop, CTLTYPE_INT | CTLFLAG_MPSAFE | CTLFLAG_RD,
    NULL, 0, sysctl_vm_reserv_fullpop, "I", "Current number of full reservations");

static int sysctl_vm_reserv_partpopq(SYSCTL_HANDLER_ARGS);

SYSCTL_OID(_vm_reserv, OID_AUTO, partpopq,
    CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NEEDGIANT, NULL, 0,
    sysctl_vm_reserv_partpopq, "A",
    "Partially populated reservation queues");

static COUNTER_U64_DEFINE_EARLY(vm_reserv_reclaimed);
SYSCTL_COUNTER_U64(_vm_reserv, OID_AUTO, reclaimed, CTLFLAG_RD,
    &vm_reserv_reclaimed, "Cumulative number of reclaimed reservations");

/*
 * The object lock pool is used to synchronize the rvq.  We can not use a
 * pool mutex because it is required before malloc works.
 *
 * The "hash" function could be made faster without divide and modulo.
 */
#define	VM_RESERV_OBJ_LOCK_COUNT	MAXCPU

struct mtx_padalign vm_reserv_object_mtx[VM_RESERV_OBJ_LOCK_COUNT];

#define	vm_reserv_object_lock_idx(object)			\
	    (((uintptr_t)object / sizeof(*object)) % VM_RESERV_OBJ_LOCK_COUNT)
#define	vm_reserv_object_lock_ptr(object)			\
	    &vm_reserv_object_mtx[vm_reserv_object_lock_idx((object))]
#define	vm_reserv_object_lock(object)				\
	    mtx_lock(vm_reserv_object_lock_ptr((object)))
#define	vm_reserv_object_unlock(object)				\
	    mtx_unlock(vm_reserv_object_lock_ptr((object)))

static void		vm_reserv_break(vm_reserv_t rv);
static void		vm_reserv_depopulate(vm_reserv_t rv, int index);
static vm_reserv_t	vm_reserv_from_page(vm_page_t m);
static boolean_t	vm_reserv_has_pindex(vm_reserv_t rv,
			    vm_pindex_t pindex);
static void		vm_reserv_populate(vm_reserv_t rv, int index);
static void		vm_reserv_reclaim(vm_reserv_t rv);

/*
 * Returns the current number of full reservations.
 *
 * Since the number of full reservations is computed without acquiring any
 * locks, the returned value is inexact.
 */
static int
sysctl_vm_reserv_fullpop(SYSCTL_HANDLER_ARGS)
{
	vm_paddr_t paddr;
	struct vm_phys_seg *seg;
	vm_reserv_t rv;
	int fullpop, segind;

	fullpop = 0;
	for (segind = 0; segind < vm_phys_nsegs; segind++) {
		seg = &vm_phys_segs[segind];
		paddr = roundup2(seg->start, VM_LEVEL_0_SIZE);
		while (paddr + VM_LEVEL_0_SIZE > paddr && paddr +
		    VM_LEVEL_0_SIZE <= seg->end) {
			rv = &vm_reserv_array[paddr >> VM_LEVEL_0_SHIFT];
			fullpop += rv->popcnt == VM_LEVEL_0_NPAGES;
			paddr += VM_LEVEL_0_SIZE;
		}
	}
	return (sysctl_handle_int(oidp, &fullpop, 0, req));
}

/*
 * Describes the current state of the partially populated reservation queue.
 */
static int
sysctl_vm_reserv_partpopq(SYSCTL_HANDLER_ARGS)
{
	struct sbuf sbuf;
	vm_reserv_t rv;
	int counter, error, domain, level, unused_pages;

	error = sysctl_wire_old_buffer(req, 0);
	if (error != 0)
		return (error);
	sbuf_new_for_sysctl(&sbuf, NULL, 128, req);
	sbuf_printf(&sbuf, "\nDOMAIN    LEVEL     SIZE  NUMBER\n\n");
	for (domain = 0; domain < vm_ndomains; domain++) {
		for (level = -1; level <= VM_NRESERVLEVEL - 2; level++) {
			counter = 0;
			unused_pages = 0;
			vm_reserv_domain_lock(domain);
			TAILQ_FOREACH(rv, &vm_rvd[domain].partpop, partpopq) {
				if (rv == &vm_rvd[domain].marker)
					continue;
				counter++;
				unused_pages += VM_LEVEL_0_NPAGES - rv->popcnt;
			}
			vm_reserv_domain_unlock(domain);
			sbuf_printf(&sbuf, "%6d, %7d, %6dK, %6d\n",
			    domain, level,
			    unused_pages * ((int)PAGE_SIZE / 1024), counter);
		}
	}
	error = sbuf_finish(&sbuf);
	sbuf_delete(&sbuf);
	return (error);
}

/*
 * Remove a reservation from the object's objq.
 */
static void
vm_reserv_remove(vm_reserv_t rv)
{
	vm_object_t object;

	vm_reserv_assert_locked(rv);
	CTR5(KTR_VM, "%s: rv %p object %p popcnt %d inpartpop %d",
	    __FUNCTION__, rv, rv->object, rv->popcnt, rv->inpartpopq);
	KASSERT(rv->object != NULL,
	    ("vm_reserv_remove: reserv %p is free", rv));
	KASSERT(!rv->inpartpopq,
	    ("vm_reserv_remove: reserv %p's inpartpopq is TRUE", rv));
	object = rv->object;
	vm_reserv_object_lock(object);
	LIST_REMOVE(rv, objq);
	rv->object = NULL;
	vm_reserv_object_unlock(object);
}

/*
 * Insert a new reservation into the object's objq.
 */
static void
vm_reserv_insert(vm_reserv_t rv, vm_object_t object, vm_pindex_t pindex)
{
	int i;

	vm_reserv_assert_locked(rv);
	CTR6(KTR_VM,
	    "%s: rv %p(%p) object %p new %p popcnt %d",
	    __FUNCTION__, rv, rv->pages, rv->object, object,
	   rv->popcnt);
	KASSERT(rv->object == NULL,
	    ("vm_reserv_insert: reserv %p isn't free", rv));
	KASSERT(rv->popcnt == 0,
	    ("vm_reserv_insert: reserv %p's popcnt is corrupted", rv));
	KASSERT(!rv->inpartpopq,
	    ("vm_reserv_insert: reserv %p's inpartpopq is TRUE", rv));
	for (i = 0; i < NPOPMAP; i++)
		KASSERT(rv->popmap[i] == 0,
		    ("vm_reserv_insert: reserv %p's popmap is corrupted", rv));
	vm_reserv_object_lock(object);
	rv->pindex = pindex;
	rv->object = object;
	rv->lasttick = ticks;
	LIST_INSERT_HEAD(&object->rvq, rv, objq);
	vm_reserv_object_unlock(object);
}

/*
 * Reduces the given reservation's population count.  If the population count
 * becomes zero, the reservation is destroyed.  Additionally, moves the
 * reservation to the tail of the partially populated reservation queue if the
 * population count is non-zero.
 */
static void
vm_reserv_depopulate(vm_reserv_t rv, int index)
{
	struct vm_domain *vmd;

	vm_reserv_assert_locked(rv);
	CTR5(KTR_VM, "%s: rv %p object %p popcnt %d inpartpop %d",
	    __FUNCTION__, rv, rv->object, rv->popcnt, rv->inpartpopq);
	KASSERT(rv->object != NULL,
	    ("vm_reserv_depopulate: reserv %p is free", rv));
	KASSERT(popmap_is_set(rv->popmap, index),
	    ("vm_reserv_depopulate: reserv %p's popmap[%d] is clear", rv,
	    index));
	KASSERT(rv->popcnt > 0,
	    ("vm_reserv_depopulate: reserv %p's popcnt is corrupted", rv));
	KASSERT(rv->domain < vm_ndomains,
	    ("vm_reserv_depopulate: reserv %p's domain is corrupted %d",
	    rv, rv->domain));
	if (rv->popcnt == VM_LEVEL_0_NPAGES) {
		KASSERT(rv->pages->psind == 1,
		    ("vm_reserv_depopulate: reserv %p is already demoted",
		    rv));
		rv->pages->psind = 0;
	}
	popmap_clear(rv->popmap, index);
	rv->popcnt--;
	if ((unsigned)(ticks - rv->lasttick) >= PARTPOPSLOP ||
	    rv->popcnt == 0) {
		vm_reserv_domain_lock(rv->domain);
		if (rv->inpartpopq) {
			TAILQ_REMOVE(&vm_rvd[rv->domain].partpop, rv, partpopq);
			rv->inpartpopq = FALSE;
		}
		if (rv->popcnt != 0) {
			rv->inpartpopq = TRUE;
			TAILQ_INSERT_TAIL(&vm_rvd[rv->domain].partpop, rv,
			    partpopq);
		}
		vm_reserv_domain_unlock(rv->domain);
		rv->lasttick = ticks;
	}
	vmd = VM_DOMAIN(rv->domain);
	if (rv->popcnt == 0) {
		vm_reserv_remove(rv);
		vm_domain_free_lock(vmd);
		vm_phys_free_pages(rv->pages, VM_LEVEL_0_ORDER);
		vm_domain_free_unlock(vmd);
		counter_u64_add(vm_reserv_freed, 1);
	}
	vm_domain_freecnt_inc(vmd, 1);
}

/*
 * Returns the reservation to which the given page might belong.
 */
static __inline vm_reserv_t
vm_reserv_from_page(vm_page_t m)
{

	return (&vm_reserv_array[VM_PAGE_TO_PHYS(m) >> VM_LEVEL_0_SHIFT]);
}

/*
 * Returns an existing reservation or NULL and initialized successor pointer.
 */
static vm_reserv_t
vm_reserv_from_object(vm_object_t object, vm_pindex_t pindex,
    vm_page_t mpred, vm_page_t *msuccp)
{
	vm_reserv_t rv;
	vm_page_t msucc;

	msucc = NULL;
	if (mpred != NULL) {
		KASSERT(mpred->object == object,
		    ("vm_reserv_from_object: object doesn't contain mpred"));
		KASSERT(mpred->pindex < pindex,
		    ("vm_reserv_from_object: mpred doesn't precede pindex"));
		rv = vm_reserv_from_page(mpred);
		if (rv->object == object && vm_reserv_has_pindex(rv, pindex))
			goto found;
		msucc = TAILQ_NEXT(mpred, listq);
	} else
		msucc = TAILQ_FIRST(&object->memq);
	if (msucc != NULL) {
		KASSERT(msucc->pindex > pindex,
		    ("vm_reserv_from_object: msucc doesn't succeed pindex"));
		rv = vm_reserv_from_page(msucc);
		if (rv->object == object && vm_reserv_has_pindex(rv, pindex))
			goto found;
	}
	rv = NULL;

found:
	*msuccp = msucc;

	return (rv);
}

/*
 * Returns TRUE if the given reservation contains the given page index and
 * FALSE otherwise.
 */
static __inline boolean_t
vm_reserv_has_pindex(vm_reserv_t rv, vm_pindex_t pindex)
{

	return (((pindex - rv->pindex) & ~(VM_LEVEL_0_NPAGES - 1)) == 0);
}

/*
 * Increases the given reservation's population count.  Moves the reservation
 * to the tail of the partially populated reservation queue.
 */
static void
vm_reserv_populate(vm_reserv_t rv, int index)
{

	vm_reserv_assert_locked(rv);
	CTR5(KTR_VM, "%s: rv %p object %p popcnt %d inpartpop %d",
	    __FUNCTION__, rv, rv->object, rv->popcnt, rv->inpartpopq);
	KASSERT(rv->object != NULL,
	    ("vm_reserv_populate: reserv %p is free", rv));
	KASSERT(popmap_is_clear(rv->popmap, index),
	    ("vm_reserv_populate: reserv %p's popmap[%d] is set", rv,
	    index));
	KASSERT(rv->popcnt < VM_LEVEL_0_NPAGES,
	    ("vm_reserv_populate: reserv %p is already full", rv));
	KASSERT(rv->pages->psind == 0,
	    ("vm_reserv_populate: reserv %p is already promoted", rv));
	KASSERT(rv->domain < vm_ndomains,
	    ("vm_reserv_populate: reserv %p's domain is corrupted %d",
	    rv, rv->domain));
	popmap_set(rv->popmap, index);
	rv->popcnt++;
	if ((unsigned)(ticks - rv->lasttick) < PARTPOPSLOP &&
	    rv->inpartpopq && rv->popcnt != VM_LEVEL_0_NPAGES)
		return;
	rv->lasttick = ticks;
	vm_reserv_domain_lock(rv->domain);
	if (rv->inpartpopq) {
		TAILQ_REMOVE(&vm_rvd[rv->domain].partpop, rv, partpopq);
		rv->inpartpopq = FALSE;
	}
	if (rv->popcnt < VM_LEVEL_0_NPAGES) {
		rv->inpartpopq = TRUE;
		TAILQ_INSERT_TAIL(&vm_rvd[rv->domain].partpop, rv, partpopq);
	} else {
		KASSERT(rv->pages->psind == 0,
		    ("vm_reserv_populate: reserv %p is already promoted",
		    rv));
		rv->pages->psind = 1;
	}
	vm_reserv_domain_unlock(rv->domain);
}

/*
 * Allocates a contiguous set of physical pages of the given size "npages"
 * from existing or newly created reservations.  All of the physical pages
 * must be at or above the given physical address "low" and below the given
 * physical address "high".  The given value "alignment" determines the
 * alignment of the first physical page in the set.  If the given value
 * "boundary" is non-zero, then the set of physical pages cannot cross any
 * physical address boundary that is a multiple of that value.  Both
 * "alignment" and "boundary" must be a power of two.
 *
 * The page "mpred" must immediately precede the offset "pindex" within the
 * specified object.
 *
 * The object must be locked.
 */
vm_page_t
vm_reserv_alloc_contig(vm_object_t object, vm_pindex_t pindex, int domain,
    int req, vm_page_t mpred, u_long npages, vm_paddr_t low, vm_paddr_t high,
    u_long alignment, vm_paddr_t boundary)
{
	struct vm_domain *vmd;
	vm_paddr_t pa, size;
	vm_page_t m, m_ret, msucc;
	vm_pindex_t first, leftcap, rightcap;
	vm_reserv_t rv;
	u_long allocpages, maxpages, minpages;
	int i, index, n;

	VM_OBJECT_ASSERT_WLOCKED(object);
	KASSERT(npages != 0, ("vm_reserv_alloc_contig: npages is 0"));

	/*
	 * Is a reservation fundamentally impossible?
	 */
	if (pindex < VM_RESERV_INDEX(object, pindex) ||
	    pindex + npages > object->size)
		return (NULL);

	/*
	 * All reservations of a particular size have the same alignment.
	 * Assuming that the first page is allocated from a reservation, the
	 * least significant bits of its physical address can be determined
	 * from its offset from the beginning of the reservation and the size
	 * of the reservation.
	 *
	 * Could the specified index within a reservation of the smallest
	 * possible size satisfy the alignment and boundary requirements?
	 */
	pa = VM_RESERV_INDEX(object, pindex) << PAGE_SHIFT;
	if ((pa & (alignment - 1)) != 0)
		return (NULL);
	size = npages << PAGE_SHIFT;
	if (((pa ^ (pa + size - 1)) & ~(boundary - 1)) != 0)
		return (NULL);

	/*
	 * Look for an existing reservation.
	 */
	rv = vm_reserv_from_object(object, pindex, mpred, &msucc);
	if (rv != NULL) {
		KASSERT(object != kernel_object || rv->domain == domain,
		    ("vm_reserv_alloc_contig: domain mismatch"));
		index = VM_RESERV_INDEX(object, pindex);
		/* Does the allocation fit within the reservation? */
		if (index + npages > VM_LEVEL_0_NPAGES)
			return (NULL);
		domain = rv->domain;
		vmd = VM_DOMAIN(domain);
		vm_reserv_lock(rv);
		/* Handle reclaim race. */
		if (rv->object != object)
			goto out;
		m = &rv->pages[index];
		pa = VM_PAGE_TO_PHYS(m);
		if (pa < low || pa + size > high ||
		    (pa & (alignment - 1)) != 0 ||
		    ((pa ^ (pa + size - 1)) & ~(boundary - 1)) != 0)
			goto out;
		/* Handle vm_page_rename(m, new_object, ...). */
		for (i = 0; i < npages; i++)
			if (popmap_is_set(rv->popmap, index + i))
				goto out;
		if (!vm_domain_allocate(vmd, req, npages))
			goto out;
		for (i = 0; i < npages; i++)
			vm_reserv_populate(rv, index + i);
		vm_reserv_unlock(rv);
		return (m);
out:
		vm_reserv_unlock(rv);
		return (NULL);
	}

	/*
	 * Could at least one reservation fit between the first index to the
	 * left that can be used ("leftcap") and the first index to the right
	 * that cannot be used ("rightcap")?
	 *
	 * We must synchronize with the reserv object lock to protect the
	 * pindex/object of the resulting reservations against rename while
	 * we are inspecting.
	 */
	first = pindex - VM_RESERV_INDEX(object, pindex);
	minpages = VM_RESERV_INDEX(object, pindex) + npages;
	maxpages = roundup2(minpages, VM_LEVEL_0_NPAGES);
	allocpages = maxpages;
	vm_reserv_object_lock(object);
	if (mpred != NULL) {
		if ((rv = vm_reserv_from_page(mpred))->object != object)
			leftcap = mpred->pindex + 1;
		else
			leftcap = rv->pindex + VM_LEVEL_0_NPAGES;
		if (leftcap > first) {
			vm_reserv_object_unlock(object);
			return (NULL);
		}
	}
	if (msucc != NULL) {
		if ((rv = vm_reserv_from_page(msucc))->object != object)
			rightcap = msucc->pindex;
		else
			rightcap = rv->pindex;
		if (first + maxpages > rightcap) {
			if (maxpages == VM_LEVEL_0_NPAGES) {
				vm_reserv_object_unlock(object);
				return (NULL);
			}

			/*
			 * At least one reservation will fit between "leftcap"
			 * and "rightcap".  However, a reservation for the
			 * last of the requested pages will not fit.  Reduce
			 * the size of the upcoming allocation accordingly.
			 */
			allocpages = minpages;
		}
	}
	vm_reserv_object_unlock(object);

	/*
	 * Would the last new reservation extend past the end of the object?
	 *
	 * If the object is unlikely to grow don't allocate a reservation for
	 * the tail.
	 */
	if ((object->flags & OBJ_ANON) == 0 &&
	    first + maxpages > object->size) {
		if (maxpages == VM_LEVEL_0_NPAGES)
			return (NULL);
		allocpages = minpages;
	}

	/*
	 * Allocate the physical pages.  The alignment and boundary specified
	 * for this allocation may be different from the alignment and
	 * boundary specified for the requested pages.  For instance, the
	 * specified index may not be the first page within the first new
	 * reservation.
	 */
	m = NULL;
	vmd = VM_DOMAIN(domain);
	if (vm_domain_allocate(vmd, req, npages)) {
		vm_domain_free_lock(vmd);
		m = vm_phys_alloc_contig(domain, allocpages, low, high,
		    ulmax(alignment, VM_LEVEL_0_SIZE),
		    boundary > VM_LEVEL_0_SIZE ? boundary : 0);
		vm_domain_free_unlock(vmd);
		if (m == NULL) {
			vm_domain_freecnt_inc(vmd, npages);
			return (NULL);
		}
	} else
		return (NULL);
	KASSERT(vm_phys_domain(m) == domain,
	    ("vm_reserv_alloc_contig: Page domain does not match requested."));

	/*
	 * The allocated physical pages always begin at a reservation
	 * boundary, but they do not always end at a reservation boundary.
	 * Initialize every reservation that is completely covered by the
	 * allocated physical pages.
	 */
	m_ret = NULL;
	index = VM_RESERV_INDEX(object, pindex);
	do {
		rv = vm_reserv_from_page(m);
		KASSERT(rv->pages == m,
		    ("vm_reserv_alloc_contig: reserv %p's pages is corrupted",
		    rv));
		vm_reserv_lock(rv);
		vm_reserv_insert(rv, object, first);
		n = ulmin(VM_LEVEL_0_NPAGES - index, npages);
		for (i = 0; i < n; i++)
			vm_reserv_populate(rv, index + i);
		npages -= n;
		if (m_ret == NULL) {
			m_ret = &rv->pages[index];
			index = 0;
		}
		vm_reserv_unlock(rv);
		m += VM_LEVEL_0_NPAGES;
		first += VM_LEVEL_0_NPAGES;
		allocpages -= VM_LEVEL_0_NPAGES;
	} while (allocpages >= VM_LEVEL_0_NPAGES);
	return (m_ret);
}

/*
 * Allocate a physical page from an existing or newly created reservation.
 *
 * The page "mpred" must immediately precede the offset "pindex" within the
 * specified object.
 *
 * The object must be locked.
 */
vm_page_t
vm_reserv_alloc_page(vm_object_t object, vm_pindex_t pindex, int domain,
    int req, vm_page_t mpred)
{
	struct vm_domain *vmd;
	vm_page_t m, msucc;
	vm_pindex_t first, leftcap, rightcap;
	vm_reserv_t rv;
	int index;

	VM_OBJECT_ASSERT_WLOCKED(object);

	/*
	 * Is a reservation fundamentally impossible?
	 */
	if (pindex < VM_RESERV_INDEX(object, pindex) ||
	    pindex >= object->size)
		return (NULL);

	/*
	 * Look for an existing reservation.
	 */
	rv = vm_reserv_from_object(object, pindex, mpred, &msucc);
	if (rv != NULL) {
		KASSERT(object != kernel_object || rv->domain == domain,
		    ("vm_reserv_alloc_page: domain mismatch"));
		domain = rv->domain;
		vmd = VM_DOMAIN(domain);
		index = VM_RESERV_INDEX(object, pindex);
		m = &rv->pages[index];
		vm_reserv_lock(rv);
		/* Handle reclaim race. */
		if (rv->object != object ||
		    /* Handle vm_page_rename(m, new_object, ...). */
		    popmap_is_set(rv->popmap, index)) {
			m = NULL;
			goto out;
		}
		if (vm_domain_allocate(vmd, req, 1) == 0)
			m = NULL;
		else
			vm_reserv_populate(rv, index);
out:
		vm_reserv_unlock(rv);
		return (m);
	}

	/*
	 * Could a reservation fit between the first index to the left that
	 * can be used and the first index to the right that cannot be used?
	 *
	 * We must synchronize with the reserv object lock to protect the
	 * pindex/object of the resulting reservations against rename while
	 * we are inspecting.
	 */
	first = pindex - VM_RESERV_INDEX(object, pindex);
	vm_reserv_object_lock(object);
	if (mpred != NULL) {
		if ((rv = vm_reserv_from_page(mpred))->object != object)
			leftcap = mpred->pindex + 1;
		else
			leftcap = rv->pindex + VM_LEVEL_0_NPAGES;
		if (leftcap > first) {
			vm_reserv_object_unlock(object);
			return (NULL);
		}
	}
	if (msucc != NULL) {
		if ((rv = vm_reserv_from_page(msucc))->object != object)
			rightcap = msucc->pindex;
		else
			rightcap = rv->pindex;
		if (first + VM_LEVEL_0_NPAGES > rightcap) {
			vm_reserv_object_unlock(object);
			return (NULL);
		}
	}
	vm_reserv_object_unlock(object);

	/*
	 * Would the last new reservation extend past the end of the object?
	 *
	 * If the object is unlikely to grow don't allocate a reservation for
	 * the tail.
	 */
	if ((object->flags & OBJ_ANON) == 0 &&
	    first + VM_LEVEL_0_NPAGES > object->size)
		return (NULL);

	/*
	 * Allocate and populate the new reservation.
	 */
	m = NULL;
	vmd = VM_DOMAIN(domain);
	if (vm_domain_allocate(vmd, req, 1)) {
		vm_domain_free_lock(vmd);
		m = vm_phys_alloc_pages(domain, VM_FREEPOOL_DEFAULT,
		    VM_LEVEL_0_ORDER);
		vm_domain_free_unlock(vmd);
		if (m == NULL) {
			vm_domain_freecnt_inc(vmd, 1);
			return (NULL);
		}
	} else
		return (NULL);
	rv = vm_reserv_from_page(m);
	vm_reserv_lock(rv);
	KASSERT(rv->pages == m,
	    ("vm_reserv_alloc_page: reserv %p's pages is corrupted", rv));
	vm_reserv_insert(rv, object, first);
	index = VM_RESERV_INDEX(object, pindex);
	vm_reserv_populate(rv, index);
	vm_reserv_unlock(rv);

	return (&rv->pages[index]);
}

/*
 * Breaks the given reservation.  All free pages in the reservation
 * are returned to the physical memory allocator.  The reservation's
 * population count and map are reset to their initial state.
 *
 * The given reservation must not be in the partially populated reservation
 * queue.
 */
static void
vm_reserv_break(vm_reserv_t rv)
{
	u_long changes;
	int bitpos, hi, i, lo;

	vm_reserv_assert_locked(rv);
	CTR5(KTR_VM, "%s: rv %p object %p popcnt %d inpartpop %d",
	    __FUNCTION__, rv, rv->object, rv->popcnt, rv->inpartpopq);
	vm_reserv_remove(rv);
	rv->pages->psind = 0;
	hi = lo = -1;
	for (i = 0; i <= NPOPMAP; i++) {
		/*
		 * "changes" is a bitmask that marks where a new sequence of
		 * 0s or 1s begins in popmap[i], with last bit in popmap[i-1]
		 * considered to be 1 if and only if lo == hi.  The bits of
		 * popmap[-1] and popmap[NPOPMAP] are considered all 1s.
		 */
		if (i == NPOPMAP)
			changes = lo != hi;
		else {
			changes = rv->popmap[i];
			changes ^= (changes << 1) | (lo == hi);
			rv->popmap[i] = 0;
		}
		while (changes != 0) {
			/*
			 * If the next change marked begins a run of 0s, set
			 * lo to mark that position.  Otherwise set hi and
			 * free pages from lo up to hi.
			 */
			bitpos = ffsl(changes) - 1;
			changes ^= 1UL << bitpos;
			if (lo == hi)
				lo = NBPOPMAP * i + bitpos;
			else {
				hi = NBPOPMAP * i + bitpos;
				vm_domain_free_lock(VM_DOMAIN(rv->domain));
				vm_phys_enqueue_contig(&rv->pages[lo], hi - lo);
				vm_domain_free_unlock(VM_DOMAIN(rv->domain));
				lo = hi;
			}
		}
	}
	rv->popcnt = 0;
	counter_u64_add(vm_reserv_broken, 1);
}

/*
 * Breaks all reservations belonging to the given object.
 */
void
vm_reserv_break_all(vm_object_t object)
{
	vm_reserv_t rv;

	/*
	 * This access of object->rvq is unsynchronized so that the
	 * object rvq lock can nest after the domain_free lock.  We
	 * must check for races in the results.  However, the object
	 * lock prevents new additions, so we are guaranteed that when
	 * it returns NULL the object is properly empty.
	 */
	while ((rv = LIST_FIRST(&object->rvq)) != NULL) {
		vm_reserv_lock(rv);
		/* Reclaim race. */
		if (rv->object != object) {
			vm_reserv_unlock(rv);
			continue;
		}
		vm_reserv_domain_lock(rv->domain);
		if (rv->inpartpopq) {
			TAILQ_REMOVE(&vm_rvd[rv->domain].partpop, rv, partpopq);
			rv->inpartpopq = FALSE;
		}
		vm_reserv_domain_unlock(rv->domain);
		vm_reserv_break(rv);
		vm_reserv_unlock(rv);
	}
}

/*
 * Frees the given page if it belongs to a reservation.  Returns TRUE if the
 * page is freed and FALSE otherwise.
 */
boolean_t
vm_reserv_free_page(vm_page_t m)
{
	vm_reserv_t rv;
	boolean_t ret;

	rv = vm_reserv_from_page(m);
	if (rv->object == NULL)
		return (FALSE);
	vm_reserv_lock(rv);
	/* Re-validate after lock. */
	if (rv->object != NULL) {
		vm_reserv_depopulate(rv, m - rv->pages);
		ret = TRUE;
	} else
		ret = FALSE;
	vm_reserv_unlock(rv);

	return (ret);
}

/*
 * Initializes the reservation management system.  Specifically, initializes
 * the reservation array.
 *
 * Requires that vm_page_array and first_page are initialized!
 */
void
vm_reserv_init(void)
{
	vm_paddr_t paddr;
	struct vm_phys_seg *seg;
	struct vm_reserv *rv;
	struct vm_reserv_domain *rvd;
	int i, j, segind;

	/*
	 * Initialize the reservation array.  Specifically, initialize the
	 * "pages" field for every element that has an underlying superpage.
	 */
	for (segind = 0; segind < vm_phys_nsegs; segind++) {
		seg = &vm_phys_segs[segind];
		paddr = roundup2(seg->start, VM_LEVEL_0_SIZE);
		while (paddr + VM_LEVEL_0_SIZE > paddr && paddr +
		    VM_LEVEL_0_SIZE <= seg->end) {
			rv = &vm_reserv_array[paddr >> VM_LEVEL_0_SHIFT];
			rv->pages = PHYS_TO_VM_PAGE(paddr);
			rv->domain = seg->domain;
			mtx_init(&rv->lock, "vm reserv", NULL, MTX_DEF);
			paddr += VM_LEVEL_0_SIZE;
		}
	}
	for (i = 0; i < MAXMEMDOM; i++) {
		rvd = &vm_rvd[i];
		mtx_init(&rvd->lock, "vm reserv domain", NULL, MTX_DEF);
		TAILQ_INIT(&rvd->partpop);
		mtx_init(&rvd->marker.lock, "vm reserv marker", NULL, MTX_DEF);

		/*
		 * Fully populated reservations should never be present in the
		 * partially populated reservation queues.
		 */
		rvd->marker.popcnt = VM_LEVEL_0_NPAGES;
		for (j = 0; j < NBPOPMAP; j++)
			popmap_set(rvd->marker.popmap, j);
	}

	for (i = 0; i < VM_RESERV_OBJ_LOCK_COUNT; i++)
		mtx_init(&vm_reserv_object_mtx[i], "resv obj lock", NULL,
		    MTX_DEF);
}

/*
 * Returns true if the given page belongs to a reservation and that page is
 * free.  Otherwise, returns false.
 */
bool
vm_reserv_is_page_free(vm_page_t m)
{
	vm_reserv_t rv;

	rv = vm_reserv_from_page(m);
	if (rv->object == NULL)
		return (false);
	return (popmap_is_clear(rv->popmap, m - rv->pages));
}

/*
 * If the given page belongs to a reservation, returns the level of that
 * reservation.  Otherwise, returns -1.
 */
int
vm_reserv_level(vm_page_t m)
{
	vm_reserv_t rv;

	rv = vm_reserv_from_page(m);
	return (rv->object != NULL ? 0 : -1);
}

/*
 * Returns a reservation level if the given page belongs to a fully populated
 * reservation and -1 otherwise.
 */
int
vm_reserv_level_iffullpop(vm_page_t m)
{
	vm_reserv_t rv;

	rv = vm_reserv_from_page(m);
	return (rv->popcnt == VM_LEVEL_0_NPAGES ? 0 : -1);
}

/*
 * Remove a partially populated reservation from the queue.
 */
static void
vm_reserv_dequeue(vm_reserv_t rv)
{

	vm_reserv_domain_assert_locked(rv->domain);
	vm_reserv_assert_locked(rv);
	CTR5(KTR_VM, "%s: rv %p object %p popcnt %d inpartpop %d",
	    __FUNCTION__, rv, rv->object, rv->popcnt, rv->inpartpopq);
	KASSERT(rv->inpartpopq,
	    ("vm_reserv_reclaim: reserv %p's inpartpopq is FALSE", rv));

	TAILQ_REMOVE(&vm_rvd[rv->domain].partpop, rv, partpopq);
	rv->inpartpopq = FALSE;
}

/*
 * Breaks the given partially populated reservation, releasing its free pages
 * to the physical memory allocator.
 */
static void
vm_reserv_reclaim(vm_reserv_t rv)
{

	vm_reserv_assert_locked(rv);
	CTR5(KTR_VM, "%s: rv %p object %p popcnt %d inpartpop %d",
	    __FUNCTION__, rv, rv->object, rv->popcnt, rv->inpartpopq);
	if (rv->inpartpopq) {
		vm_reserv_domain_lock(rv->domain);
		vm_reserv_dequeue(rv);
		vm_reserv_domain_unlock(rv->domain);
	}
	vm_reserv_break(rv);
	counter_u64_add(vm_reserv_reclaimed, 1);
}

/*
 * Breaks a reservation near the head of the partially populated reservation
 * queue, releasing its free pages to the physical memory allocator.  Returns
 * TRUE if a reservation is broken and FALSE otherwise.
 */
bool
vm_reserv_reclaim_inactive(int domain)
{
	vm_reserv_t rv;

	vm_reserv_domain_lock(domain);
	TAILQ_FOREACH(rv, &vm_rvd[domain].partpop, partpopq) {
		/*
		 * A locked reservation is likely being updated or reclaimed,
		 * so just skip ahead.
		 */
		if (rv != &vm_rvd[domain].marker && vm_reserv_trylock(rv)) {
			vm_reserv_dequeue(rv);
			break;
		}
	}
	vm_reserv_domain_unlock(domain);
	if (rv != NULL) {
		vm_reserv_reclaim(rv);
		vm_reserv_unlock(rv);
		return (true);
	}
	return (false);
}

/*
 * Determine whether this reservation has free pages that satisfy the given
 * request for contiguous physical memory.  Start searching from the lower
 * bound, defined by low_index.
 */
static bool
vm_reserv_test_contig(vm_reserv_t rv, u_long npages, vm_paddr_t low,
    vm_paddr_t high, u_long alignment, vm_paddr_t boundary)
{
	vm_paddr_t pa, size;
	u_long changes;
	int bitpos, bits_left, i, hi, lo, n;

	vm_reserv_assert_locked(rv);
	size = npages << PAGE_SHIFT;
	pa = VM_PAGE_TO_PHYS(&rv->pages[0]);
	lo = (pa < low) ?
	    ((low + PAGE_MASK - pa) >> PAGE_SHIFT) : 0;
	i = lo / NBPOPMAP;
	changes = rv->popmap[i] | ((1UL << (lo % NBPOPMAP)) - 1);
	hi = (pa + VM_LEVEL_0_SIZE > high) ?
	    ((high + PAGE_MASK - pa) >> PAGE_SHIFT) : VM_LEVEL_0_NPAGES;
	n = hi / NBPOPMAP;
	bits_left = hi % NBPOPMAP;
	hi = lo = -1;
	for (;;) {
		/*
		 * "changes" is a bitmask that marks where a new sequence of
		 * 0s or 1s begins in popmap[i], with last bit in popmap[i-1]
		 * considered to be 1 if and only if lo == hi.  The bits of
		 * popmap[-1] and popmap[NPOPMAP] are considered all 1s.
		 */
		changes ^= (changes << 1) | (lo == hi);
		while (changes != 0) {
			/*
			 * If the next change marked begins a run of 0s, set
			 * lo to mark that position.  Otherwise set hi and
			 * look for a satisfactory first page from lo up to hi.
			 */
			bitpos = ffsl(changes) - 1;
			changes ^= 1UL << bitpos;
			if (lo == hi) {
				lo = NBPOPMAP * i + bitpos;
				continue;
			}
			hi = NBPOPMAP * i + bitpos;
			pa = VM_PAGE_TO_PHYS(&rv->pages[lo]);
			if ((pa & (alignment - 1)) != 0) {
				/* Skip to next aligned page. */
				lo += (((pa - 1) | (alignment - 1)) + 1) >>
				    PAGE_SHIFT;
				if (lo >= VM_LEVEL_0_NPAGES)
					return (false);
				pa = VM_PAGE_TO_PHYS(&rv->pages[lo]);
			}
			if (((pa ^ (pa + size - 1)) & ~(boundary - 1)) != 0) {
				/* Skip to next boundary-matching page. */
				lo += (((pa - 1) | (boundary - 1)) + 1) >>
				    PAGE_SHIFT;
				if (lo >= VM_LEVEL_0_NPAGES)
					return (false);
				pa = VM_PAGE_TO_PHYS(&rv->pages[lo]);
			}
			if (lo * PAGE_SIZE + size <= hi * PAGE_SIZE)
				return (true);
			lo = hi;
		}
		if (++i < n)
			changes = rv->popmap[i];
		else if (i == n)
			changes = bits_left == 0 ? -1UL :
			    (rv->popmap[n] | (-1UL << bits_left));
		else
			return (false);
	}
}

/*
 * Searches the partially populated reservation queue for the least recently
 * changed reservation with free pages that satisfy the given request for
 * contiguous physical memory.  If a satisfactory reservation is found, it is
 * broken.  Returns true if a reservation is broken and false otherwise.
 */
bool
vm_reserv_reclaim_contig(int domain, u_long npages, vm_paddr_t low,
    vm_paddr_t high, u_long alignment, vm_paddr_t boundary)
{
	struct vm_reserv_queue *queue;
	vm_paddr_t pa, size;
	vm_reserv_t marker, rv, rvn;

	if (npages > VM_LEVEL_0_NPAGES - 1)
		return (false);
	marker = &vm_rvd[domain].marker;
	queue = &vm_rvd[domain].partpop;
	size = npages << PAGE_SHIFT;

	vm_reserv_domain_scan_lock(domain);
	vm_reserv_domain_lock(domain);
	TAILQ_FOREACH_SAFE(rv, queue, partpopq, rvn) {
		pa = VM_PAGE_TO_PHYS(&rv->pages[0]);
		if (pa + VM_LEVEL_0_SIZE - size < low) {
			/* This entire reservation is too low; go to next. */
			continue;
		}
		if (pa + size > high) {
			/* This entire reservation is too high; go to next. */
			continue;
		}

		if (vm_reserv_trylock(rv) == 0) {
			TAILQ_INSERT_AFTER(queue, rv, marker, partpopq);
			vm_reserv_domain_unlock(domain);
			vm_reserv_lock(rv);
			if (!rv->inpartpopq ||
			    TAILQ_NEXT(rv, partpopq) != marker) {
				vm_reserv_unlock(rv);
				vm_reserv_domain_lock(domain);
				rvn = TAILQ_NEXT(marker, partpopq);
				TAILQ_REMOVE(queue, marker, partpopq);
				continue;
			}
			vm_reserv_domain_lock(domain);
			TAILQ_REMOVE(queue, marker, partpopq);
		}
		vm_reserv_domain_unlock(domain);
		if (vm_reserv_test_contig(rv, npages, low, high,
		    alignment, boundary)) {
			vm_reserv_domain_scan_unlock(domain);
			vm_reserv_reclaim(rv);
			vm_reserv_unlock(rv);
			return (true);
		}
		vm_reserv_unlock(rv);
		vm_reserv_domain_lock(domain);
	}
	vm_reserv_domain_unlock(domain);
	vm_reserv_domain_scan_unlock(domain);
	return (false);
}

/*
 * Transfers the reservation underlying the given page to a new object.
 *
 * The object must be locked.
 */
void
vm_reserv_rename(vm_page_t m, vm_object_t new_object, vm_object_t old_object,
    vm_pindex_t old_object_offset)
{
	vm_reserv_t rv;

	VM_OBJECT_ASSERT_WLOCKED(new_object);
	rv = vm_reserv_from_page(m);
	if (rv->object == old_object) {
		vm_reserv_lock(rv);
		CTR6(KTR_VM,
		    "%s: rv %p object %p new %p popcnt %d inpartpop %d",
		    __FUNCTION__, rv, rv->object, new_object, rv->popcnt,
		    rv->inpartpopq);
		if (rv->object == old_object) {
			vm_reserv_object_lock(old_object);
			rv->object = NULL;
			LIST_REMOVE(rv, objq);
			vm_reserv_object_unlock(old_object);
			vm_reserv_object_lock(new_object);
			rv->object = new_object;
			rv->pindex -= old_object_offset;
			LIST_INSERT_HEAD(&new_object->rvq, rv, objq);
			vm_reserv_object_unlock(new_object);
		}
		vm_reserv_unlock(rv);
	}
}

/*
 * Returns the size (in bytes) of a reservation of the specified level.
 */
int
vm_reserv_size(int level)
{

	switch (level) {
	case 0:
		return (VM_LEVEL_0_SIZE);
	case -1:
		return (PAGE_SIZE);
	default:
		return (0);
	}
}

/*
 * Allocates the virtual and physical memory required by the reservation
 * management system's data structures, in particular, the reservation array.
 */
vm_paddr_t
vm_reserv_startup(vm_offset_t *vaddr, vm_paddr_t end)
{
	vm_paddr_t new_end, high_water;
	size_t size;
	int i;

	high_water = phys_avail[1];
	for (i = 0; i < vm_phys_nsegs; i++) {
		if (vm_phys_segs[i].end > high_water)
			high_water = vm_phys_segs[i].end;
	}

	/* Skip the first chunk.  It is already accounted for. */
	for (i = 2; phys_avail[i + 1] != 0; i += 2) {
		if (phys_avail[i + 1] > high_water)
			high_water = phys_avail[i + 1];
	}

	/*
	 * Calculate the size (in bytes) of the reservation array.  Round up
	 * from "high_water" because every small page is mapped to an element
	 * in the reservation array based on its physical address.  Thus, the
	 * number of elements in the reservation array can be greater than the
	 * number of superpages. 
	 */
	size = howmany(high_water, VM_LEVEL_0_SIZE) * sizeof(struct vm_reserv);

	/*
	 * Allocate and map the physical memory for the reservation array.  The
	 * next available virtual address is returned by reference.
	 */
	new_end = end - round_page(size);
	vm_reserv_array = (void *)(uintptr_t)pmap_map(vaddr, new_end, end,
	    VM_PROT_READ | VM_PROT_WRITE);
	bzero(vm_reserv_array, size);

	/*
	 * Return the next available physical address.
	 */
	return (new_end);
}

/*
 * Returns the superpage containing the given page.
 */
vm_page_t
vm_reserv_to_superpage(vm_page_t m)
{
	vm_reserv_t rv;

	VM_OBJECT_ASSERT_LOCKED(m->object);
	rv = vm_reserv_from_page(m);
	if (rv->object == m->object && rv->popcnt == VM_LEVEL_0_NPAGES)
		m = rv->pages;
	else
		m = NULL;

	return (m);
}

#endif	/* VM_NRESERVLEVEL > 0 */