summaryrefslogtreecommitdiffstats
path: root/sys/vm/swap_pager.c
blob: f4c598eb11689764079fb0b0422d858779a32737 (plain)
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
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
/*
 * Copyright (c) 1998 Matthew Dillon,
 * Copyright (c) 1994 John S. Dyson
 * Copyright (c) 1990 University of Utah.
 * Copyright (c) 1991, 1993
 *	The Regents of the University of California.  All rights reserved.
 *
 * This code is derived from software contributed to Berkeley by
 * the Systems Programming Group of the University of Utah Computer
 * Science Department.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. All advertising materials mentioning features or use of this software
 *    must display the following acknowledgement:
 *	This product includes software developed by the University of
 *	California, Berkeley and its contributors.
 * 4. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS 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 REGENTS 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.
 *
 *				New Swap System
 *				Matthew Dillon
 *
 * Radix Bitmap 'blists'.
 *
 *	- The new swapper uses the new radix bitmap code.  This should scale
 *	  to arbitrarily small or arbitrarily large swap spaces and an almost
 *	  arbitrary degree of fragmentation.
 *
 * Features:
 *
 *	- on the fly reallocation of swap during putpages.  The new system
 *	  does not try to keep previously allocated swap blocks for dirty
 *	  pages.  
 *
 *	- on the fly deallocation of swap
 *
 *	- No more garbage collection required.  Unnecessarily allocated swap
 *	  blocks only exist for dirty vm_page_t's now and these are already
 *	  cycled (in a high-load system) by the pager.  We also do on-the-fly
 *	  removal of invalidated swap blocks when a page is destroyed
 *	  or renamed.
 *
 * from: Utah $Hdr: swap_pager.c 1.4 91/04/30$
 *
 *	@(#)swap_pager.c	8.9 (Berkeley) 3/21/94
 *
 * $FreeBSD$
 */

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/conf.h>
#include <sys/kernel.h>
#include <sys/proc.h>
#include <sys/buf.h>
#include <sys/vnode.h>
#include <sys/malloc.h>
#include <sys/vmmeter.h>
#include <sys/sysctl.h>
#include <sys/blist.h>
#include <sys/lock.h>

#ifndef MAX_PAGEOUT_CLUSTER
#define MAX_PAGEOUT_CLUSTER 16
#endif

#define SWB_NPAGES	MAX_PAGEOUT_CLUSTER

#include "opt_swap.h"
#include <vm/vm.h>
#include <vm/vm_object.h>
#include <vm/vm_page.h>
#include <vm/vm_pager.h>
#include <vm/vm_pageout.h>
#include <vm/swap_pager.h>
#include <vm/vm_extern.h>
#include <vm/vm_zone.h>

#define SWM_FREE	0x02	/* free, period			*/
#define SWM_POP		0x04	/* pop out			*/

/*
 * vm_swap_size is in page-sized chunks now.  It was DEV_BSIZE'd chunks
 * in the old system.
 */

extern int vm_swap_size;	/* number of free swap blocks, in pages */

int swap_pager_full;		/* swap space exhaustion (task killing) */
static int swap_pager_almost_full; /* swap space exhaustion (w/ hysteresis)*/
static int nsw_rcount;		/* free read buffers			*/
static int nsw_wcount_sync;	/* limit write buffers / synchronous	*/
static int nsw_wcount_async;	/* limit write buffers / asynchronous	*/
static int nsw_wcount_async_max;/* assigned maximum			*/
static int nsw_cluster_max;	/* maximum VOP I/O allowed		*/
static int sw_alloc_interlock;	/* swap pager allocation interlock	*/

struct blist *swapblist;
static struct swblock **swhash;
static int swhash_mask;
static int swap_async_max = 4;	/* maximum in-progress async I/O's	*/

extern struct vnode *swapdev_vp;	/* from vm_swap.c */

SYSCTL_INT(_vm, OID_AUTO, swap_async_max,
        CTLFLAG_RW, &swap_async_max, 0, "Maximum running async swap ops");

/*
 * "named" and "unnamed" anon region objects.  Try to reduce the overhead
 * of searching a named list by hashing it just a little.
 */

#define NOBJLISTS		8

#define NOBJLIST(handle)	\
	(&swap_pager_object_list[((int)(intptr_t)handle >> 4) & (NOBJLISTS-1)])

static struct pagerlst	swap_pager_object_list[NOBJLISTS];
struct pagerlst		swap_pager_un_object_list;
vm_zone_t		swap_zone;

/*
 * pagerops for OBJT_SWAP - "swap pager".  Some ops are also global procedure
 * calls hooked from other parts of the VM system and do not appear here.
 * (see vm/swap_pager.h).
 */

static vm_object_t
		swap_pager_alloc __P((void *handle, vm_ooffset_t size,
				      vm_prot_t prot, vm_ooffset_t offset));
static void	swap_pager_dealloc __P((vm_object_t object));
static int	swap_pager_getpages __P((vm_object_t, vm_page_t *, int, int));
static void	swap_pager_init __P((void));
static void	swap_pager_unswapped __P((vm_page_t));
static void	swap_pager_strategy __P((vm_object_t, struct buf *));

struct pagerops swappagerops = {
	swap_pager_init,	/* early system initialization of pager	*/
	swap_pager_alloc,	/* allocate an OBJT_SWAP object		*/
	swap_pager_dealloc,	/* deallocate an OBJT_SWAP object	*/
	swap_pager_getpages,	/* pagein				*/
	swap_pager_putpages,	/* pageout				*/
	swap_pager_haspage,	/* get backing store status for page	*/
	swap_pager_unswapped,	/* remove swap related to page		*/
	swap_pager_strategy	/* pager strategy call			*/
};

/*
 * dmmax is in page-sized chunks with the new swap system.  It was
 * dev-bsized chunks in the old.
 *
 * swap_*() routines are externally accessible.  swp_*() routines are
 * internal.
 */

int dmmax;
static int dmmax_mask;
int nswap_lowat = 128;		/* in pages, swap_pager_almost_full warn */
int nswap_hiwat = 512;		/* in pages, swap_pager_almost_full warn */

static __inline void	swp_sizecheck __P((void));
static void	swp_pager_sync_iodone __P((struct buf *bp));
static void	swp_pager_async_iodone __P((struct buf *bp));

/*
 * Swap bitmap functions
 */

static __inline void	swp_pager_freeswapspace __P((daddr_t blk, int npages));
static __inline daddr_t	swp_pager_getswapspace __P((int npages));

/*
 * Metadata functions
 */

static void swp_pager_meta_build __P((vm_object_t, vm_pindex_t, daddr_t));
static void swp_pager_meta_free __P((vm_object_t, vm_pindex_t, daddr_t));
static void swp_pager_meta_free_all __P((vm_object_t));
static daddr_t swp_pager_meta_ctl __P((vm_object_t, vm_pindex_t, int));

/*
 * SWP_SIZECHECK() -	update swap_pager_full indication
 *	
 *	update the swap_pager_almost_full indication and warn when we are
 *	about to run out of swap space, using lowat/hiwat hysteresis.
 *
 *	Clear swap_pager_full ( task killing ) indication when lowat is met.
 *
 *	No restrictions on call
 *	This routine may not block.
 *	This routine must be called at splvm()
 */

static __inline void
swp_sizecheck()
{
	if (vm_swap_size < nswap_lowat) {
		if (swap_pager_almost_full == 0) {
			printf("swap_pager: out of swap space\n");
			swap_pager_almost_full = 1;
		}
	} else {
		swap_pager_full = 0;
		if (vm_swap_size > nswap_hiwat)
			swap_pager_almost_full = 0;
	}
}

/*
 * SWAP_PAGER_INIT() -	initialize the swap pager!
 *
 *	Expected to be started from system init.  NOTE:  This code is run 
 *	before much else so be careful what you depend on.  Most of the VM
 *	system has yet to be initialized at this point.
 */

static void
swap_pager_init()
{
	/*
	 * Initialize object lists
	 */
	int i;

	for (i = 0; i < NOBJLISTS; ++i)
		TAILQ_INIT(&swap_pager_object_list[i]);
	TAILQ_INIT(&swap_pager_un_object_list);

	/*
	 * Device Stripe, in PAGE_SIZE'd blocks
	 */

	dmmax = SWB_NPAGES * 2;
	dmmax_mask = ~(dmmax - 1);
}

/*
 * SWAP_PAGER_SWAP_INIT() - swap pager initialization from pageout process
 *
 *	Expected to be started from pageout process once, prior to entering
 *	its main loop.
 */

void
swap_pager_swap_init()
{
	int n;

	/*
	 * Number of in-transit swap bp operations.  Don't
	 * exhaust the pbufs completely.  Make sure we
	 * initialize workable values (0 will work for hysteresis
	 * but it isn't very efficient).
	 *
	 * The nsw_cluster_max is constrained by the bp->b_pages[]
	 * array (MAXPHYS/PAGE_SIZE) and our locally defined
	 * MAX_PAGEOUT_CLUSTER.   Also be aware that swap ops are
	 * constrained by the swap device interleave stripe size.
	 *
	 * Currently we hardwire nsw_wcount_async to 4.  This limit is 
	 * designed to prevent other I/O from having high latencies due to
	 * our pageout I/O.  The value 4 works well for one or two active swap
	 * devices but is probably a little low if you have more.  Even so,
	 * a higher value would probably generate only a limited improvement
	 * with three or four active swap devices since the system does not
	 * typically have to pageout at extreme bandwidths.   We will want
	 * at least 2 per swap devices, and 4 is a pretty good value if you
	 * have one NFS swap device due to the command/ack latency over NFS.
	 * So it all works out pretty well.
	 */

	nsw_cluster_max = min((MAXPHYS/PAGE_SIZE), MAX_PAGEOUT_CLUSTER);

	nsw_rcount = (nswbuf + 1) / 2;
	nsw_wcount_sync = (nswbuf + 3) / 4;
	nsw_wcount_async = 4;
	nsw_wcount_async_max = nsw_wcount_async;

	/*
	 * Initialize our zone.  Right now I'm just guessing on the number
	 * we need based on the number of pages in the system.  Each swblock
	 * can hold 16 pages, so this is probably overkill.
	 */

	n = cnt.v_page_count * 2;

	swap_zone = zinit(
	    "SWAPMETA", 
	    sizeof(struct swblock), 
	    n,
	    ZONE_INTERRUPT, 
	    1
	);

	/*
	 * Initialize our meta-data hash table.  The swapper does not need to
	 * be quite as efficient as the VM system, so we do not use an 
	 * oversized hash table.
	 *
	 * 	n: 		size of hash table, must be power of 2
	 *	swhash_mask:	hash table index mask
	 */

	for (n = 1; n < cnt.v_page_count / 4; n <<= 1)
		;

	swhash = malloc(sizeof(struct swblock *) * n, M_VMPGDATA, M_WAITOK);
	bzero(swhash, sizeof(struct swblock *) * n);

	swhash_mask = n - 1;
}

/*
 * SWAP_PAGER_ALLOC() -	allocate a new OBJT_SWAP VM object and instantiate
 *			its metadata structures.
 *
 *	This routine is called from the mmap and fork code to create a new
 *	OBJT_SWAP object.  We do this by creating an OBJT_DEFAULT object
 *	and then converting it with swp_pager_meta_build().
 *
 *	This routine may block in vm_object_allocate() and create a named
 *	object lookup race, so we must interlock.   We must also run at
 *	splvm() for the object lookup to handle races with interrupts, but
 *	we do not have to maintain splvm() in between the lookup and the
 *	add because (I believe) it is not possible to attempt to create
 *	a new swap object w/handle when a default object with that handle
 *	already exists.
 */

static vm_object_t
swap_pager_alloc(void *handle, vm_ooffset_t size, vm_prot_t prot,
		 vm_ooffset_t offset)
{
	vm_object_t object;

	if (handle) {
		/*
		 * Reference existing named region or allocate new one.  There
		 * should not be a race here against swp_pager_meta_build()
		 * as called from vm_page_remove() in regards to the lookup
		 * of the handle.
		 */

		while (sw_alloc_interlock) {
			sw_alloc_interlock = -1;
			tsleep(&sw_alloc_interlock, PVM, "swpalc", 0);
		}
		sw_alloc_interlock = 1;

		object = vm_pager_object_lookup(NOBJLIST(handle), handle);

		if (object != NULL) {
			vm_object_reference(object);
		} else {
			object = vm_object_allocate(OBJT_DEFAULT,
				OFF_TO_IDX(offset + PAGE_MASK + size));
			object->handle = handle;

			swp_pager_meta_build(object, 0, SWAPBLK_NONE);
		}

		if (sw_alloc_interlock < 0)
			wakeup(&sw_alloc_interlock);

		sw_alloc_interlock = 0;
	} else {
		object = vm_object_allocate(OBJT_DEFAULT,
			OFF_TO_IDX(offset + PAGE_MASK + size));

		swp_pager_meta_build(object, 0, SWAPBLK_NONE);
	}

	return (object);
}

/*
 * SWAP_PAGER_DEALLOC() -	remove swap metadata from object
 *
 *	The swap backing for the object is destroyed.  The code is 
 *	designed such that we can reinstantiate it later, but this
 *	routine is typically called only when the entire object is
 *	about to be destroyed.
 *
 *	This routine may block, but no longer does. 
 *
 *	The object must be locked or unreferenceable.
 */

static void
swap_pager_dealloc(object)
	vm_object_t object;
{
	int s;

	/*
	 * Remove from list right away so lookups will fail if we block for
	 * pageout completion.
	 */

	if (object->handle == NULL) {
		TAILQ_REMOVE(&swap_pager_un_object_list, object, pager_object_list);
	} else {
		TAILQ_REMOVE(NOBJLIST(object->handle), object, pager_object_list);
	}

	vm_object_pip_wait(object, "swpdea");

	/*
	 * Free all remaining metadata.  We only bother to free it from 
	 * the swap meta data.  We do not attempt to free swapblk's still
	 * associated with vm_page_t's for this object.  We do not care
	 * if paging is still in progress on some objects.
	 */
	s = splvm();
	swp_pager_meta_free_all(object);
	splx(s);
}

/************************************************************************
 *			SWAP PAGER BITMAP ROUTINES			*
 ************************************************************************/

/*
 * SWP_PAGER_GETSWAPSPACE() -	allocate raw swap space
 *
 *	Allocate swap for the requested number of pages.  The starting
 *	swap block number (a page index) is returned or SWAPBLK_NONE
 *	if the allocation failed.
 *
 *	Also has the side effect of advising that somebody made a mistake
 *	when they configured swap and didn't configure enough.
 *
 *	Must be called at splvm() to avoid races with bitmap frees from
 *	vm_page_remove() aka swap_pager_page_removed().
 *
 *	This routine may not block
 *	This routine must be called at splvm().
 */

static __inline daddr_t
swp_pager_getswapspace(npages)
	int npages;
{
	daddr_t blk;

	if ((blk = blist_alloc(swapblist, npages)) == SWAPBLK_NONE) {
		if (swap_pager_full != 2) {
			printf("swap_pager_getswapspace: failed\n");
			swap_pager_full = 2;
			swap_pager_almost_full = 1;
		}
	} else {
		vm_swap_size -= npages;
		swp_sizecheck();
	}
	return(blk);
}

/*
 * SWP_PAGER_FREESWAPSPACE() -	free raw swap space 
 *
 *	This routine returns the specified swap blocks back to the bitmap.
 *
 *	Note:  This routine may not block (it could in the old swap code),
 *	and through the use of the new blist routines it does not block.
 *
 *	We must be called at splvm() to avoid races with bitmap frees from
 *	vm_page_remove() aka swap_pager_page_removed().
 *
 *	This routine may not block
 *	This routine must be called at splvm().
 */

static __inline void
swp_pager_freeswapspace(blk, npages)
	daddr_t blk;
	int npages;
{
	blist_free(swapblist, blk, npages);
	vm_swap_size += npages;
	swp_sizecheck();
}

/*
 * SWAP_PAGER_FREESPACE() -	frees swap blocks associated with a page
 *				range within an object.
 *
 *	This is a globally accessible routine.
 *
 *	This routine removes swapblk assignments from swap metadata.
 *
 *	The external callers of this routine typically have already destroyed 
 *	or renamed vm_page_t's associated with this range in the object so 
 *	we should be ok.
 *
 *	This routine may be called at any spl.  We up our spl to splvm temporarily
 *	in order to perform the metadata removal.
 */

void
swap_pager_freespace(object, start, size)
	vm_object_t object;
	vm_pindex_t start;
	vm_size_t size;
{
	int s = splvm();
	swp_pager_meta_free(object, start, size);
	splx(s);
}

/*
 * SWAP_PAGER_RESERVE() - reserve swap blocks in object
 *
 *	Assigns swap blocks to the specified range within the object.  The 
 *	swap blocks are not zerod.  Any previous swap assignment is destroyed.
 *
 *	Returns 0 on success, -1 on failure.
 */

int
swap_pager_reserve(vm_object_t object, vm_pindex_t start, vm_size_t size)
{
	int s;
	int n = 0;
	daddr_t blk = SWAPBLK_NONE;
	vm_pindex_t beg = start;	/* save start index */

	s = splvm();
	while (size) {
		if (n == 0) {
			n = BLIST_MAX_ALLOC;
			while ((blk = swp_pager_getswapspace(n)) == SWAPBLK_NONE) {
				n >>= 1;
				if (n == 0) {
					swp_pager_meta_free(object, beg, start - beg);
					splx(s);
					return(-1);
				}
			}
		}
		swp_pager_meta_build(object, start, blk);
		--size;
		++start;
		++blk;
		--n;
	}
	swp_pager_meta_free(object, start, n);
	splx(s);
	return(0);
}

/*
 * SWAP_PAGER_COPY() -  copy blocks from source pager to destination pager
 *			and destroy the source.
 *
 *	Copy any valid swapblks from the source to the destination.  In
 *	cases where both the source and destination have a valid swapblk,
 *	we keep the destination's.
 *
 *	This routine is allowed to block.  It may block allocating metadata
 *	indirectly through swp_pager_meta_build() or if paging is still in
 *	progress on the source. 
 *
 *	This routine can be called at any spl
 *
 *	XXX vm_page_collapse() kinda expects us not to block because we 
 *	supposedly do not need to allocate memory, but for the moment we
 *	*may* have to get a little memory from the zone allocator, but
 *	it is taken from the interrupt memory.  We should be ok. 
 *
 *	The source object contains no vm_page_t's (which is just as well)
 *
 *	The source object is of type OBJT_SWAP.
 *
 *	The source and destination objects must be locked or 
 *	inaccessible (XXX are they ?)
 */

void
swap_pager_copy(srcobject, dstobject, offset, destroysource)
	vm_object_t srcobject;
	vm_object_t dstobject;
	vm_pindex_t offset;
	int destroysource;
{
	vm_pindex_t i;
	int s;

	s = splvm();

	/*
	 * If destroysource is set, we remove the source object from the 
	 * swap_pager internal queue now. 
	 */

	if (destroysource) {
		if (srcobject->handle == NULL) {
			TAILQ_REMOVE(
			    &swap_pager_un_object_list, 
			    srcobject, 
			    pager_object_list
			);
		} else {
			TAILQ_REMOVE(
			    NOBJLIST(srcobject->handle),
			    srcobject,
			    pager_object_list
			);
		}
	}

	/*
	 * transfer source to destination.
	 */

	for (i = 0; i < dstobject->size; ++i) {
		daddr_t dstaddr;

		/*
		 * Locate (without changing) the swapblk on the destination,
		 * unless it is invalid in which case free it silently, or
		 * if the destination is a resident page, in which case the
		 * source is thrown away.
		 */

		dstaddr = swp_pager_meta_ctl(dstobject, i, 0);

		if (dstaddr == SWAPBLK_NONE) {
			/*
			 * Destination has no swapblk and is not resident,
			 * copy source.
			 */
			daddr_t srcaddr;

			srcaddr = swp_pager_meta_ctl(
			    srcobject, 
			    i + offset,
			    SWM_POP
			);

			if (srcaddr != SWAPBLK_NONE)
				swp_pager_meta_build(dstobject, i, srcaddr);
		} else {
			/*
			 * Destination has valid swapblk or it is represented
			 * by a resident page.  We destroy the sourceblock.
			 */
			
			swp_pager_meta_ctl(srcobject, i + offset, SWM_FREE);
		}
	}

	/*
	 * Free left over swap blocks in source.
	 *
	 * We have to revert the type to OBJT_DEFAULT so we do not accidently
	 * double-remove the object from the swap queues.
	 */

	if (destroysource) {
		swp_pager_meta_free_all(srcobject);
		/*
		 * Reverting the type is not necessary, the caller is going
		 * to destroy srcobject directly, but I'm doing it here
		 * for consistency since we've removed the object from its
		 * queues.
		 */
		srcobject->type = OBJT_DEFAULT;
	}
	splx(s);
}

/*
 * SWAP_PAGER_HASPAGE() -	determine if we have good backing store for
 *				the requested page.
 *
 *	We determine whether good backing store exists for the requested
 *	page and return TRUE if it does, FALSE if it doesn't.
 *
 *	If TRUE, we also try to determine how much valid, contiguous backing
 *	store exists before and after the requested page within a reasonable
 *	distance.  We do not try to restrict it to the swap device stripe
 *	(that is handled in getpages/putpages).  It probably isn't worth
 *	doing here.
 */

boolean_t
swap_pager_haspage(object, pindex, before, after)
	vm_object_t object;
	vm_pindex_t pindex;
	int *before;
	int *after;
{
	daddr_t blk0;
	int s;

	/*
	 * do we have good backing store at the requested index ?
	 */

	s = splvm();
	blk0 = swp_pager_meta_ctl(object, pindex, 0);

	if (blk0 == SWAPBLK_NONE) {
		splx(s);
		if (before)
			*before = 0;
		if (after)
			*after = 0;
		return (FALSE);
	}

	/*
	 * find backwards-looking contiguous good backing store
	 */

	if (before != NULL) {
		int i;

		for (i = 1; i < (SWB_NPAGES/2); ++i) {
			daddr_t blk;

			if (i > pindex)
				break;
			blk = swp_pager_meta_ctl(object, pindex - i, 0);
			if (blk != blk0 - i)
				break;
		}
		*before = (i - 1);
	}

	/*
	 * find forward-looking contiguous good backing store
	 */

	if (after != NULL) {
		int i;

		for (i = 1; i < (SWB_NPAGES/2); ++i) {
			daddr_t blk;

			blk = swp_pager_meta_ctl(object, pindex + i, 0);
			if (blk != blk0 + i)
				break;
		}
		*after = (i - 1);
	}
	splx(s);
	return (TRUE);
}

/*
 * SWAP_PAGER_PAGE_UNSWAPPED() - remove swap backing store related to page
 *
 *	This removes any associated swap backing store, whether valid or
 *	not, from the page.  
 *
 *	This routine is typically called when a page is made dirty, at
 *	which point any associated swap can be freed.  MADV_FREE also
 *	calls us in a special-case situation
 *
 *	NOTE!!!  If the page is clean and the swap was valid, the caller
 *	should make the page dirty before calling this routine.  This routine
 *	does NOT change the m->dirty status of the page.  Also: MADV_FREE
 *	depends on it.
 *
 *	This routine may not block
 *	This routine must be called at splvm()
 */

static void
swap_pager_unswapped(m)
	vm_page_t m;
{
	swp_pager_meta_ctl(m->object, m->pindex, SWM_FREE);
}

/*
 * SWAP_PAGER_STRATEGY() - read, write, free blocks
 *
 *	This implements the vm_pager_strategy() interface to swap and allows
 *	other parts of the system to directly access swap as backing store
 *	through vm_objects of type OBJT_SWAP.  This is intended to be a 
 *	cacheless interface ( i.e. caching occurs at higher levels ).
 *	Therefore we do not maintain any resident pages.  All I/O goes
 *	directly to and from the swap device.
 *	
 *	Note that b_blkno is scaled for PAGE_SIZE
 *
 *	We currently attempt to run I/O synchronously or asynchronously as
 *	the caller requests.  This isn't perfect because we loose error
 *	sequencing when we run multiple ops in parallel to satisfy a request.
 *	But this is swap, so we let it all hang out.
 */

static void	
swap_pager_strategy(vm_object_t object, struct buf *bp)
{
	vm_pindex_t start;
	int count;
	int s;
	char *data;
	struct buf *nbp = NULL;

	if (bp->b_bcount & PAGE_MASK) {
		bp->b_error = EINVAL;
		bp->b_ioflags |= BIO_ERROR;
		bp->b_flags |= B_INVAL;
		biodone(bp);
		printf("swap_pager_strategy: bp %p b_vp %p blk %d size %d, not page bounded\n", bp, bp->b_vp, (int)bp->b_pblkno, (int)bp->b_bcount);
		return;
	}

	/*
	 * Clear error indication, initialize page index, count, data pointer.
	 */

	bp->b_error = 0;
	bp->b_ioflags &= ~BIO_ERROR;
	bp->b_resid = bp->b_bcount;

	start = bp->b_pblkno;
	count = howmany(bp->b_bcount, PAGE_SIZE);
	data = bp->b_data;

	s = splvm();

	/*
	 * Deal with BIO_DELETE
	 */

	if (bp->b_iocmd == BIO_DELETE) {
		/*
		 * FREE PAGE(s) - destroy underlying swap that is no longer
		 *		  needed.
		 */
		swp_pager_meta_free(object, start, count);
		splx(s);
		bp->b_resid = 0;
		biodone(bp);
		return;
	}

	/*
	 * Execute read or write
	 */

	while (count > 0) {
		daddr_t blk;

		/*
		 * Obtain block.  If block not found and writing, allocate a
		 * new block and build it into the object.
		 */

		blk = swp_pager_meta_ctl(object, start, 0);
		if ((blk == SWAPBLK_NONE) && (bp->b_iocmd == BIO_WRITE)) {
			blk = swp_pager_getswapspace(1);
			if (blk == SWAPBLK_NONE) {
				bp->b_error = ENOMEM;
				bp->b_ioflags |= BIO_ERROR;
				break;
			}
			swp_pager_meta_build(object, start, blk);
		}
			
		/*
		 * Do we have to flush our current collection?  Yes if:
		 *
		 *	- no swap block at this index
		 *	- swap block is not contiguous
		 *	- we cross a physical disk boundry in the
		 *	  stripe.
		 */

		if (
		    nbp && (nbp->b_blkno + btoc(nbp->b_bcount) != blk ||
		     ((nbp->b_blkno ^ blk) & dmmax_mask)
		    )
		) {
			splx(s);
			if (bp->b_iocmd == BIO_READ) {
				++cnt.v_swapin;
				cnt.v_swappgsin += btoc(nbp->b_bcount);
			} else {
				++cnt.v_swapout;
				cnt.v_swappgsout += btoc(nbp->b_bcount);
				nbp->b_dirtyend = nbp->b_bcount;
			}
			flushchainbuf(nbp);
			s = splvm();
			nbp = NULL;
		}

		/*
		 * Add new swapblk to nbp, instantiating nbp if necessary.
		 * Zero-fill reads are able to take a shortcut.
		 */

		if (blk == SWAPBLK_NONE) {
			/*
			 * We can only get here if we are reading.  Since
			 * we are at splvm() we can safely modify b_resid,
			 * even if chain ops are in progress.
			 */
			bzero(data, PAGE_SIZE);
			bp->b_resid -= PAGE_SIZE;
		} else {
			if (nbp == NULL) {
				nbp = getchainbuf(bp, swapdev_vp, (bp->b_iocmd == BIO_READ) | B_ASYNC);
				nbp->b_blkno = blk;
				nbp->b_bcount = 0;
				nbp->b_data = data;
			}
			nbp->b_bcount += PAGE_SIZE;
		}
		--count;
		++start;
		data += PAGE_SIZE;
	}

	/*
	 *  Flush out last buffer
	 */

	splx(s);

	if (nbp) {
		if ((bp->b_flags & B_ASYNC) == 0)
			nbp->b_flags &= ~B_ASYNC;
		if (nbp->b_iocmd == BIO_READ) {
			++cnt.v_swapin;
			cnt.v_swappgsin += btoc(nbp->b_bcount);
		} else {
			++cnt.v_swapout;
			cnt.v_swappgsout += btoc(nbp->b_bcount);
			nbp->b_dirtyend = nbp->b_bcount;
		}
		flushchainbuf(nbp);
		/* nbp = NULL; */
	}

	/*
	 * Wait for completion.
	 */

	if (bp->b_flags & B_ASYNC) {
		autochaindone(bp);
	} else {
		waitchainbuf(bp, 0, 1);
	}
}

/*
 * SWAP_PAGER_GETPAGES() - bring pages in from swap
 *
 *	Attempt to retrieve (m, count) pages from backing store, but make
 *	sure we retrieve at least m[reqpage].  We try to load in as large
 *	a chunk surrounding m[reqpage] as is contiguous in swap and which
 *	belongs to the same object.
 *
 *	The code is designed for asynchronous operation and 
 *	immediate-notification of 'reqpage' but tends not to be
 *	used that way.  Please do not optimize-out this algorithmic
 *	feature, I intend to improve on it in the future.
 *
 *	The parent has a single vm_object_pip_add() reference prior to
 *	calling us and we should return with the same.
 *
 *	The parent has BUSY'd the pages.  We should return with 'm'
 *	left busy, but the others adjusted.
 */

static int
swap_pager_getpages(object, m, count, reqpage)
	vm_object_t object;
	vm_page_t *m;
	int count, reqpage;
{
	struct buf *bp;
	vm_page_t mreq;
	int s;
	int i;
	int j;
	daddr_t blk;
	vm_offset_t kva;
	vm_pindex_t lastpindex;

	mreq = m[reqpage];

	if (mreq->object != object) {
		panic("swap_pager_getpages: object mismatch %p/%p", 
		    object, 
		    mreq->object
		);
	}
	/*
	 * Calculate range to retrieve.  The pages have already been assigned
	 * their swapblks.  We require a *contiguous* range that falls entirely
	 * within a single device stripe.   If we do not supply it, bad things
	 * happen.  Note that blk, iblk & jblk can be SWAPBLK_NONE, but the 
	 * loops are set up such that the case(s) are handled implicitly.
	 *
	 * The swp_*() calls must be made at splvm().  vm_page_free() does
	 * not need to be, but it will go a little faster if it is.
	 */

	s = splvm();
	blk = swp_pager_meta_ctl(mreq->object, mreq->pindex, 0);

	for (i = reqpage - 1; i >= 0; --i) {
		daddr_t iblk;

		iblk = swp_pager_meta_ctl(m[i]->object, m[i]->pindex, 0);
		if (blk != iblk + (reqpage - i))
			break;
		if ((blk ^ iblk) & dmmax_mask)
			break;
	}
	++i;

	for (j = reqpage + 1; j < count; ++j) {
		daddr_t jblk;

		jblk = swp_pager_meta_ctl(m[j]->object, m[j]->pindex, 0);
		if (blk != jblk - (j - reqpage))
			break;
		if ((blk ^ jblk) & dmmax_mask)
			break;
	}

	/*
	 * free pages outside our collection range.   Note: we never free
	 * mreq, it must remain busy throughout.
	 */

	{
		int k;

		for (k = 0; k < i; ++k)
			vm_page_free(m[k]);
		for (k = j; k < count; ++k)
			vm_page_free(m[k]);
	}
	splx(s);


	/*
	 * Return VM_PAGER_FAIL if we have nothing to do.  Return mreq 
	 * still busy, but the others unbusied.
	 */

	if (blk == SWAPBLK_NONE)
		return(VM_PAGER_FAIL);

	/*
	 * Get a swap buffer header to perform the IO
	 */

	bp = getpbuf(&nsw_rcount);
	kva = (vm_offset_t) bp->b_data;

	/*
	 * map our page(s) into kva for input
	 *
	 * NOTE: B_PAGING is set by pbgetvp()
	 */

	pmap_qenter(kva, m + i, j - i);

	bp->b_iocmd = BIO_READ;
	bp->b_iodone = swp_pager_async_iodone;
	bp->b_rcred = bp->b_wcred = proc0.p_ucred;
	bp->b_data = (caddr_t) kva;
	crhold(bp->b_rcred);
	crhold(bp->b_wcred);
	bp->b_blkno = blk - (reqpage - i);
	bp->b_bcount = PAGE_SIZE * (j - i);
	bp->b_bufsize = PAGE_SIZE * (j - i);
	bp->b_pager.pg_reqpage = reqpage - i;

	{
		int k;

		for (k = i; k < j; ++k) {
			bp->b_pages[k - i] = m[k];
			vm_page_flag_set(m[k], PG_SWAPINPROG);
		}
	}
	bp->b_npages = j - i;

	pbgetvp(swapdev_vp, bp);

	cnt.v_swapin++;
	cnt.v_swappgsin += bp->b_npages;

	/*
	 * We still hold the lock on mreq, and our automatic completion routine
	 * does not remove it.
	 */

	vm_object_pip_add(mreq->object, bp->b_npages);
	lastpindex = m[j-1]->pindex;

	/*
	 * perform the I/O.  NOTE!!!  bp cannot be considered valid after
	 * this point because we automatically release it on completion.
	 * Instead, we look at the one page we are interested in which we
	 * still hold a lock on even through the I/O completion.
	 *
	 * The other pages in our m[] array are also released on completion,
	 * so we cannot assume they are valid anymore either.
	 *
	 * NOTE: b_blkno is destroyed by the call to VOP_STRATEGY
	 */

	BUF_KERNPROC(bp);
	BUF_STRATEGY(bp);

	/*
	 * wait for the page we want to complete.  PG_SWAPINPROG is always
	 * cleared on completion.  If an I/O error occurs, SWAPBLK_NONE
	 * is set in the meta-data.
	 */

	s = splvm();

	while ((mreq->flags & PG_SWAPINPROG) != 0) {
		vm_page_flag_set(mreq, PG_WANTED | PG_REFERENCED);
		cnt.v_intrans++;
		if (tsleep(mreq, PSWP, "swread", hz*20)) {
			printf(
			    "swap_pager: indefinite wait buffer: device:"
				" %s, blkno: %ld, size: %ld\n",
			    devtoname(bp->b_dev), (long)bp->b_blkno,
			    bp->b_bcount
			);
		}
	}

	splx(s);

	/*
	 * mreq is left bussied after completion, but all the other pages
	 * are freed.  If we had an unrecoverable read error the page will
	 * not be valid.
	 */

	if (mreq->valid != VM_PAGE_BITS_ALL) {
		return(VM_PAGER_ERROR);
	} else {
		return(VM_PAGER_OK);
	}

	/*
	 * A final note: in a low swap situation, we cannot deallocate swap
	 * and mark a page dirty here because the caller is likely to mark
	 * the page clean when we return, causing the page to possibly revert 
	 * to all-zero's later.
	 */
}

/*
 *	swap_pager_putpages: 
 *
 *	Assign swap (if necessary) and initiate I/O on the specified pages.
 *
 *	We support both OBJT_DEFAULT and OBJT_SWAP objects.  DEFAULT objects
 *	are automatically converted to SWAP objects.
 *
 *	In a low memory situation we may block in VOP_STRATEGY(), but the new 
 *	vm_page reservation system coupled with properly written VFS devices 
 *	should ensure that no low-memory deadlock occurs.  This is an area
 *	which needs work.
 *
 *	The parent has N vm_object_pip_add() references prior to
 *	calling us and will remove references for rtvals[] that are
 *	not set to VM_PAGER_PEND.  We need to remove the rest on I/O
 *	completion.
 *
 *	The parent has soft-busy'd the pages it passes us and will unbusy
 *	those whos rtvals[] entry is not set to VM_PAGER_PEND on return.
 *	We need to unbusy the rest on I/O completion.
 */

void
swap_pager_putpages(object, m, count, sync, rtvals)
	vm_object_t object;
	vm_page_t *m;
	int count;
	boolean_t sync;
	int *rtvals;
{
	int i;
	int n = 0;

	if (count && m[0]->object != object) {
		panic("swap_pager_getpages: object mismatch %p/%p", 
		    object, 
		    m[0]->object
		);
	}
	/*
	 * Step 1
	 *
	 * Turn object into OBJT_SWAP
	 * check for bogus sysops
	 * force sync if not pageout process
	 */

	if (object->type != OBJT_SWAP)
		swp_pager_meta_build(object, 0, SWAPBLK_NONE);

	if (curproc != pageproc)
		sync = TRUE;

	/*
	 * Step 2
	 *
	 * Update nsw parameters from swap_async_max sysctl values.  
	 * Do not let the sysop crash the machine with bogus numbers.
	 */

	if (swap_async_max != nsw_wcount_async_max) {
		int n;
		int s;

		/*
		 * limit range
		 */
		if ((n = swap_async_max) > nswbuf / 2)
			n = nswbuf / 2;
		if (n < 1)
			n = 1;
		swap_async_max = n;

		/*
		 * Adjust difference ( if possible ).  If the current async
		 * count is too low, we may not be able to make the adjustment
		 * at this time.
		 */
		s = splvm();
		n -= nsw_wcount_async_max;
		if (nsw_wcount_async + n >= 0) {
			nsw_wcount_async += n;
			nsw_wcount_async_max += n;
			wakeup(&nsw_wcount_async);
		}
		splx(s);
	}

	/*
	 * Step 3
	 *
	 * Assign swap blocks and issue I/O.  We reallocate swap on the fly.
	 * The page is left dirty until the pageout operation completes
	 * successfully.
	 */

	for (i = 0; i < count; i += n) {
		int s;
		int j;
		struct buf *bp;
		daddr_t blk;

		/*
		 * Maximum I/O size is limited by a number of factors.
		 */

		n = min(BLIST_MAX_ALLOC, count - i);
		n = min(n, nsw_cluster_max);

		s = splvm();

		/*
		 * Get biggest block of swap we can.  If we fail, fall
		 * back and try to allocate a smaller block.  Don't go
		 * overboard trying to allocate space if it would overly
		 * fragment swap.
		 */
		while (
		    (blk = swp_pager_getswapspace(n)) == SWAPBLK_NONE &&
		    n > 4
		) {
			n >>= 1;
		}
		if (blk == SWAPBLK_NONE) {
			for (j = 0; j < n; ++j)
				rtvals[i+j] = VM_PAGER_FAIL;
			splx(s);
			continue;
		}

		/*
		 * The I/O we are constructing cannot cross a physical
		 * disk boundry in the swap stripe.  Note: we are still
		 * at splvm().
		 */
		if ((blk ^ (blk + n)) & dmmax_mask) {
			j = ((blk + dmmax) & dmmax_mask) - blk;
			swp_pager_freeswapspace(blk + j, n - j);
			n = j;
		}

		/*
		 * All I/O parameters have been satisfied, build the I/O
		 * request and assign the swap space.
		 *
		 * NOTE: B_PAGING is set by pbgetvp()
		 */

		if (sync == TRUE) {
			bp = getpbuf(&nsw_wcount_sync);
		} else {
			bp = getpbuf(&nsw_wcount_async);
			bp->b_flags = B_ASYNC;
		}
		bp->b_iocmd = BIO_WRITE;
		bp->b_spc = NULL;	/* not used, but NULL-out anyway */

		pmap_qenter((vm_offset_t)bp->b_data, &m[i], n);

		bp->b_rcred = bp->b_wcred = proc0.p_ucred;
		bp->b_bcount = PAGE_SIZE * n;
		bp->b_bufsize = PAGE_SIZE * n;
		bp->b_blkno = blk;

		crhold(bp->b_rcred);
		crhold(bp->b_wcred);

		pbgetvp(swapdev_vp, bp);

		for (j = 0; j < n; ++j) {
			vm_page_t mreq = m[i+j];

			swp_pager_meta_build(
			    mreq->object, 
			    mreq->pindex,
			    blk + j
			);
			vm_page_dirty(mreq);
			rtvals[i+j] = VM_PAGER_OK;

			vm_page_flag_set(mreq, PG_SWAPINPROG);
			bp->b_pages[j] = mreq;
		}
		bp->b_npages = n;
		/*
		 * Must set dirty range for NFS to work.
		 */
		bp->b_dirtyoff = 0;
		bp->b_dirtyend = bp->b_bcount;

		cnt.v_swapout++;
		cnt.v_swappgsout += bp->b_npages;
		swapdev_vp->v_numoutput++;

		splx(s);

		/*
		 * asynchronous
		 *
		 * NOTE: b_blkno is destroyed by the call to VOP_STRATEGY
		 */

		if (sync == FALSE) {
			bp->b_iodone = swp_pager_async_iodone;
			BUF_KERNPROC(bp);
			BUF_STRATEGY(bp);

			for (j = 0; j < n; ++j)
				rtvals[i+j] = VM_PAGER_PEND;
			continue;
		}

		/*
		 * synchronous
		 *
		 * NOTE: b_blkno is destroyed by the call to VOP_STRATEGY
		 */

		bp->b_iodone = swp_pager_sync_iodone;
		BUF_STRATEGY(bp);

		/*
		 * Wait for the sync I/O to complete, then update rtvals.
		 * We just set the rtvals[] to VM_PAGER_PEND so we can call
		 * our async completion routine at the end, thus avoiding a
		 * double-free.
		 */
		s = splbio();

		while ((bp->b_flags & B_DONE) == 0) {
			tsleep(bp, PVM, "swwrt", 0);
		}

		for (j = 0; j < n; ++j)
			rtvals[i+j] = VM_PAGER_PEND;

		/*
		 * Now that we are through with the bp, we can call the
		 * normal async completion, which frees everything up.
		 */

		swp_pager_async_iodone(bp);

		splx(s);
	}
}

/*
 *	swap_pager_sync_iodone:
 *
 *	Completion routine for synchronous reads and writes from/to swap.
 *	We just mark the bp is complete and wake up anyone waiting on it.
 *
 *	This routine may not block.  This routine is called at splbio() or better.
 */

static void
swp_pager_sync_iodone(bp)
	struct buf *bp;
{
	bp->b_flags |= B_DONE;
	bp->b_flags &= ~B_ASYNC;
	wakeup(bp);
}

/*
 *	swp_pager_async_iodone:
 *
 *	Completion routine for asynchronous reads and writes from/to swap.
 *	Also called manually by synchronous code to finish up a bp.
 *
 *	For READ operations, the pages are PG_BUSY'd.  For WRITE operations, 
 *	the pages are vm_page_t->busy'd.  For READ operations, we PG_BUSY 
 *	unbusy all pages except the 'main' request page.  For WRITE 
 *	operations, we vm_page_t->busy'd unbusy all pages ( we can do this 
 *	because we marked them all VM_PAGER_PEND on return from putpages ).
 *
 *	This routine may not block.
 *	This routine is called at splbio() or better
 *
 *	We up ourselves to splvm() as required for various vm_page related
 *	calls.
 */

static void
swp_pager_async_iodone(bp)
	register struct buf *bp;
{
	int s;
	int i;
	vm_object_t object = NULL;

	bp->b_flags |= B_DONE;

	/*
	 * report error
	 */

	if (bp->b_ioflags & BIO_ERROR) {
		printf(
		    "swap_pager: I/O error - %s failed; blkno %ld,"
			"size %ld, error %d\n",
		    ((bp->b_iocmd == BIO_READ) ? "pagein" : "pageout"),
		    (long)bp->b_blkno, 
		    (long)bp->b_bcount,
		    bp->b_error
		);
	}

	/*
	 * set object, raise to splvm().
	 */

	if (bp->b_npages)
		object = bp->b_pages[0]->object;
	s = splvm();

	/*
	 * remove the mapping for kernel virtual
	 */

	pmap_qremove((vm_offset_t)bp->b_data, bp->b_npages);

	/*
	 * cleanup pages.  If an error occurs writing to swap, we are in
	 * very serious trouble.  If it happens to be a disk error, though,
	 * we may be able to recover by reassigning the swap later on.  So
	 * in this case we remove the m->swapblk assignment for the page 
	 * but do not free it in the rlist.  The errornous block(s) are thus
	 * never reallocated as swap.  Redirty the page and continue.
	 */

	for (i = 0; i < bp->b_npages; ++i) {
		vm_page_t m = bp->b_pages[i];

		vm_page_flag_clear(m, PG_SWAPINPROG);

		if (bp->b_ioflags & BIO_ERROR) {
			/*
			 * If an error occurs I'd love to throw the swapblk
			 * away without freeing it back to swapspace, so it
			 * can never be used again.  But I can't from an 
			 * interrupt.
			 */

			if (bp->b_iocmd == BIO_READ) {
				/*
				 * When reading, reqpage needs to stay
				 * locked for the parent, but all other
				 * pages can be freed.  We still want to
				 * wakeup the parent waiting on the page,
				 * though.  ( also: pg_reqpage can be -1 and 
				 * not match anything ).
				 *
				 * We have to wake specifically requested pages
				 * up too because we cleared PG_SWAPINPROG and
				 * someone may be waiting for that.
				 *
				 * NOTE: for reads, m->dirty will probably
				 * be overridden by the original caller of
				 * getpages so don't play cute tricks here.
				 *
				 * XXX it may not be legal to free the page
				 * here as this messes with the object->memq's.
				 */

				m->valid = 0;
				vm_page_flag_clear(m, PG_ZERO);

				if (i != bp->b_pager.pg_reqpage)
					vm_page_free(m);
				else
					vm_page_flash(m);
				/*
				 * If i == bp->b_pager.pg_reqpage, do not wake 
				 * the page up.  The caller needs to.
				 */
			} else {
				/*
				 * If a write error occurs, reactivate page
				 * so it doesn't clog the inactive list,
				 * then finish the I/O.
				 */
				vm_page_dirty(m);
				vm_page_activate(m);
				vm_page_io_finish(m);
			}
		} else if (bp->b_iocmd == BIO_READ) {
			/*
			 * For read success, clear dirty bits.  Nobody should
			 * have this page mapped but don't take any chances,
			 * make sure the pmap modify bits are also cleared.
			 *
			 * NOTE: for reads, m->dirty will probably be 
			 * overridden by the original caller of getpages so
			 * we cannot set them in order to free the underlying
			 * swap in a low-swap situation.  I don't think we'd
			 * want to do that anyway, but it was an optimization
			 * that existed in the old swapper for a time before
			 * it got ripped out due to precisely this problem.
			 *
			 * clear PG_ZERO in page.
			 *
			 * If not the requested page then deactivate it.
			 *
			 * Note that the requested page, reqpage, is left
			 * busied, but we still have to wake it up.  The
			 * other pages are released (unbusied) by 
			 * vm_page_wakeup().  We do not set reqpage's
			 * valid bits here, it is up to the caller.
			 */

			pmap_clear_modify(VM_PAGE_TO_PHYS(m));
			m->valid = VM_PAGE_BITS_ALL;
			vm_page_undirty(m);
			vm_page_flag_clear(m, PG_ZERO);

			/*
			 * We have to wake specifically requested pages
			 * up too because we cleared PG_SWAPINPROG and
			 * could be waiting for it in getpages.  However,
			 * be sure to not unbusy getpages specifically
			 * requested page - getpages expects it to be 
			 * left busy.
			 */
			if (i != bp->b_pager.pg_reqpage) {
				vm_page_deactivate(m);
				vm_page_wakeup(m);
			} else {
				vm_page_flash(m);
			}
		} else {
			/*
			 * For write success, clear the modify and dirty 
			 * status, then finish the I/O ( which decrements the 
			 * busy count and possibly wakes waiter's up ).
			 */
			vm_page_protect(m, VM_PROT_READ);
			pmap_clear_modify(VM_PAGE_TO_PHYS(m));
			vm_page_undirty(m);
			vm_page_io_finish(m);
		}
	}

	/*
	 * adjust pip.  NOTE: the original parent may still have its own
	 * pip refs on the object.
	 */

	if (object)
		vm_object_pip_wakeupn(object, bp->b_npages);

	/*
	 * release the physical I/O buffer
	 */

	relpbuf(
	    bp, 
	    ((bp->b_iocmd == BIO_READ) ? &nsw_rcount : 
		((bp->b_flags & B_ASYNC) ? 
		    &nsw_wcount_async : 
		    &nsw_wcount_sync
		)
	    )
	);
	splx(s);
}

/************************************************************************
 *				SWAP META DATA 				*
 ************************************************************************
 *
 *	These routines manipulate the swap metadata stored in the 
 *	OBJT_SWAP object.  All swp_*() routines must be called at
 *	splvm() because swap can be freed up by the low level vm_page
 *	code which might be called from interrupts beyond what splbio() covers.
 *
 *	Swap metadata is implemented with a global hash and not directly
 *	linked into the object.  Instead the object simply contains
 *	appropriate tracking counters.
 */

/*
 * SWP_PAGER_HASH() -	hash swap meta data
 *
 *	This is an inline helper function which hashes the swapblk given
 *	the object and page index.  It returns a pointer to a pointer
 *	to the object, or a pointer to a NULL pointer if it could not
 *	find a swapblk.
 *
 *	This routine must be called at splvm().
 */

static __inline struct swblock **
swp_pager_hash(vm_object_t object, vm_pindex_t index)
{
	struct swblock **pswap;
	struct swblock *swap;

	index &= ~SWAP_META_MASK;
	pswap = &swhash[(index ^ (int)(intptr_t)object) & swhash_mask];

	while ((swap = *pswap) != NULL) {
		if (swap->swb_object == object &&
		    swap->swb_index == index
		) {
			break;
		}
		pswap = &swap->swb_hnext;
	}
	return(pswap);
}

/*
 * SWP_PAGER_META_BUILD() -	add swap block to swap meta data for object
 *
 *	We first convert the object to a swap object if it is a default
 *	object.
 *
 *	The specified swapblk is added to the object's swap metadata.  If
 *	the swapblk is not valid, it is freed instead.  Any previously
 *	assigned swapblk is freed.
 *
 *	This routine must be called at splvm(), except when used to convert
 *	an OBJT_DEFAULT object into an OBJT_SWAP object.

 */

static void
swp_pager_meta_build(
	vm_object_t object, 
	vm_pindex_t index,
	daddr_t swapblk
) {
	struct swblock *swap;
	struct swblock **pswap;

	/*
	 * Convert default object to swap object if necessary
	 */

	if (object->type != OBJT_SWAP) {
		object->type = OBJT_SWAP;
		object->un_pager.swp.swp_bcount = 0;

		if (object->handle != NULL) {
			TAILQ_INSERT_TAIL(
			    NOBJLIST(object->handle),
			    object, 
			    pager_object_list
			);
		} else {
			TAILQ_INSERT_TAIL(
			    &swap_pager_un_object_list,
			    object, 
			    pager_object_list
			);
		}
	}
	
	/*
	 * Locate hash entry.  If not found create, but if we aren't adding
	 * anything just return.  If we run out of space in the map we wait
	 * and, since the hash table may have changed, retry.
	 */

retry:
	pswap = swp_pager_hash(object, index);

	if ((swap = *pswap) == NULL) {
		int i;

		if (swapblk == SWAPBLK_NONE)
			return;

		swap = *pswap = zalloc(swap_zone);
		if (swap == NULL) {
			VM_WAIT;
			goto retry;
		}
		swap->swb_hnext = NULL;
		swap->swb_object = object;
		swap->swb_index = index & ~SWAP_META_MASK;
		swap->swb_count = 0;

		++object->un_pager.swp.swp_bcount;

		for (i = 0; i < SWAP_META_PAGES; ++i)
			swap->swb_pages[i] = SWAPBLK_NONE;
	}

	/*
	 * Delete prior contents of metadata
	 */

	index &= SWAP_META_MASK;

	if (swap->swb_pages[index] != SWAPBLK_NONE) {
		swp_pager_freeswapspace(swap->swb_pages[index], 1);
		--swap->swb_count;
	}

	/*
	 * Enter block into metadata
	 */

	swap->swb_pages[index] = swapblk;
	if (swapblk != SWAPBLK_NONE)
		++swap->swb_count;
}

/*
 * SWP_PAGER_META_FREE() - free a range of blocks in the object's swap metadata
 *
 *	The requested range of blocks is freed, with any associated swap 
 *	returned to the swap bitmap.
 *
 *	This routine will free swap metadata structures as they are cleaned 
 *	out.  This routine does *NOT* operate on swap metadata associated
 *	with resident pages.
 *
 *	This routine must be called at splvm()
 */

static void
swp_pager_meta_free(vm_object_t object, vm_pindex_t index, daddr_t count)
{
	if (object->type != OBJT_SWAP)
		return;

	while (count > 0) {
		struct swblock **pswap;
		struct swblock *swap;

		pswap = swp_pager_hash(object, index);

		if ((swap = *pswap) != NULL) {
			daddr_t v = swap->swb_pages[index & SWAP_META_MASK];

			if (v != SWAPBLK_NONE) {
				swp_pager_freeswapspace(v, 1);
				swap->swb_pages[index & SWAP_META_MASK] =
					SWAPBLK_NONE;
				if (--swap->swb_count == 0) {
					*pswap = swap->swb_hnext;
					zfree(swap_zone, swap);
					--object->un_pager.swp.swp_bcount;
				}
			}
			--count;
			++index;
		} else {
			int n = SWAP_META_PAGES - (index & SWAP_META_MASK);
			count -= n;
			index += n;
		}
	}
}

/*
 * SWP_PAGER_META_FREE_ALL() - destroy all swap metadata associated with object
 *
 *	This routine locates and destroys all swap metadata associated with
 *	an object.
 *
 *	This routine must be called at splvm()
 */

static void
swp_pager_meta_free_all(vm_object_t object)
{
	daddr_t index = 0;

	if (object->type != OBJT_SWAP)
		return;

	while (object->un_pager.swp.swp_bcount) {
		struct swblock **pswap;
		struct swblock *swap;

		pswap = swp_pager_hash(object, index);
		if ((swap = *pswap) != NULL) {
			int i;

			for (i = 0; i < SWAP_META_PAGES; ++i) {
				daddr_t v = swap->swb_pages[i];
				if (v != SWAPBLK_NONE) {
					--swap->swb_count;
					swp_pager_freeswapspace(v, 1);
				}
			}
			if (swap->swb_count != 0)
				panic("swap_pager_meta_free_all: swb_count != 0");
			*pswap = swap->swb_hnext;
			zfree(swap_zone, swap);
			--object->un_pager.swp.swp_bcount;
		}
		index += SWAP_META_PAGES;
		if (index > 0x20000000)
			panic("swp_pager_meta_free_all: failed to locate all swap meta blocks");
	}
}

/*
 * SWP_PAGER_METACTL() -  misc control of swap and vm_page_t meta data.
 *
 *	This routine is capable of looking up, popping, or freeing
 *	swapblk assignments in the swap meta data or in the vm_page_t.
 *	The routine typically returns the swapblk being looked-up, or popped,
 *	or SWAPBLK_NONE if the block was freed, or SWAPBLK_NONE if the block
 *	was invalid.  This routine will automatically free any invalid 
 *	meta-data swapblks.
 *
 *	It is not possible to store invalid swapblks in the swap meta data
 *	(other then a literal 'SWAPBLK_NONE'), so we don't bother checking.
 *
 *	When acting on a busy resident page and paging is in progress, we 
 *	have to wait until paging is complete but otherwise can act on the 
 *	busy page.
 *
 *	This routine must be called at splvm().
 *
 *	SWM_FREE	remove and free swap block from metadata
 *	SWM_POP		remove from meta data but do not free.. pop it out
 */

static daddr_t
swp_pager_meta_ctl(
	vm_object_t object,
	vm_pindex_t index,
	int flags
) {
	struct swblock **pswap;
	struct swblock *swap;
	daddr_t r1;

	/*
	 * The meta data only exists of the object is OBJT_SWAP 
	 * and even then might not be allocated yet.
	 */

	if (object->type != OBJT_SWAP)
		return(SWAPBLK_NONE);

	r1 = SWAPBLK_NONE;
	pswap = swp_pager_hash(object, index);

	if ((swap = *pswap) != NULL) {
		index &= SWAP_META_MASK;
		r1 = swap->swb_pages[index];

		if (r1 != SWAPBLK_NONE) {
			if (flags & SWM_FREE) {
				swp_pager_freeswapspace(r1, 1);
				r1 = SWAPBLK_NONE;
			}
			if (flags & (SWM_FREE|SWM_POP)) {
				swap->swb_pages[index] = SWAPBLK_NONE;
				if (--swap->swb_count == 0) {
					*pswap = swap->swb_hnext;
					zfree(swap_zone, swap);
					--object->un_pager.swp.swp_bcount;
				}
			} 
		}
	}
	return(r1);
}

OpenPOWER on IntegriCloud