summaryrefslogtreecommitdiffstats
path: root/sys/kern/kern_kse.c
blob: 10831cf2f8c600a5ed68bf038f5ba031a8555122 (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
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
/* 
 * Copyright (C) 2001 Julian Elischer <julian@freebsd.org>.
 *  All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice(s), this list of conditions and the following disclaimer as
 *    the first lines of this file unmodified other than the possible 
 *    addition of one or more copyright notices.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice(s), 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 HOLDER(S) ``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 HOLDER(S) 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.
 */

#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/lock.h>
#include <sys/malloc.h>
#include <sys/mutex.h>
#include <sys/proc.h>
#include <sys/smp.h>
#include <sys/sysctl.h>
#include <sys/sysproto.h>
#include <sys/filedesc.h>
#include <sys/sched.h>
#include <sys/signalvar.h>
#include <sys/sx.h>
#include <sys/tty.h>
#include <sys/user.h>
#include <sys/jail.h>
#include <sys/kse.h>
#include <sys/ktr.h>
#include <sys/ucontext.h>

#include <vm/vm.h>
#include <vm/vm_extern.h>
#include <vm/vm_object.h>
#include <vm/pmap.h>
#include <vm/uma.h>
#include <vm/vm_map.h>

#include <machine/frame.h>

/*
 * KSEGRP related storage.
 */
static uma_zone_t ksegrp_zone;
static uma_zone_t kse_zone;
static uma_zone_t thread_zone;
static uma_zone_t upcall_zone;

/* DEBUG ONLY */
SYSCTL_NODE(_kern, OID_AUTO, threads, CTLFLAG_RW, 0, "thread allocation");
static int thread_debug = 0;
SYSCTL_INT(_kern_threads, OID_AUTO, debug, CTLFLAG_RW,
	&thread_debug, 0, "thread debug");

static int max_threads_per_proc = 150;
SYSCTL_INT(_kern_threads, OID_AUTO, max_threads_per_proc, CTLFLAG_RW,
	&max_threads_per_proc, 0, "Limit on threads per proc");

static int max_groups_per_proc = 50;
SYSCTL_INT(_kern_threads, OID_AUTO, max_groups_per_proc, CTLFLAG_RW,
	&max_groups_per_proc, 0, "Limit on thread groups per proc");

static int max_threads_hits;
SYSCTL_INT(_kern_threads, OID_AUTO, max_threads_hits, CTLFLAG_RD,
	&max_threads_hits, 0, "");

static int virtual_cpu;

#define RANGEOF(type, start, end) (offsetof(type, end) - offsetof(type, start))

TAILQ_HEAD(, thread) zombie_threads = TAILQ_HEAD_INITIALIZER(zombie_threads);
TAILQ_HEAD(, kse) zombie_kses = TAILQ_HEAD_INITIALIZER(zombie_kses);
TAILQ_HEAD(, ksegrp) zombie_ksegrps = TAILQ_HEAD_INITIALIZER(zombie_ksegrps);
TAILQ_HEAD(, kse_upcall) zombie_upcalls = 
	TAILQ_HEAD_INITIALIZER(zombie_upcalls);
struct mtx kse_zombie_lock;
MTX_SYSINIT(kse_zombie_lock, &kse_zombie_lock, "kse zombie lock", MTX_SPIN);

static void kse_purge(struct proc *p, struct thread *td);
static void kse_purge_group(struct thread *td);
static int thread_update_usr_ticks(struct thread *td, int user);
static void thread_alloc_spare(struct thread *td, struct thread *spare);

static int
sysctl_kse_virtual_cpu(SYSCTL_HANDLER_ARGS)
{
	int error, new_val;
	int def_val;

#ifdef SMP
	def_val = mp_ncpus;
#else
	def_val = 1;
#endif
	if (virtual_cpu == 0)
		new_val = def_val;
	else
		new_val = virtual_cpu;
	error = sysctl_handle_int(oidp, &new_val, 0, req);
        if (error != 0 || req->newptr == NULL)
		return (error);
	if (new_val < 0)
		return (EINVAL);
	virtual_cpu = new_val;
	return (0);
}

/* DEBUG ONLY */
SYSCTL_PROC(_kern_threads, OID_AUTO, virtual_cpu, CTLTYPE_INT|CTLFLAG_RW,
	0, sizeof(virtual_cpu), sysctl_kse_virtual_cpu, "I",
	"debug virtual cpus");

/*
 * Prepare a thread for use.
 */
static void
thread_ctor(void *mem, int size, void *arg)
{
	struct thread	*td;

	td = (struct thread *)mem;
	td->td_state = TDS_INACTIVE;
	td->td_oncpu	= NOCPU;
	td->td_critnest = 1;
}

/*
 * Reclaim a thread after use.
 */
static void
thread_dtor(void *mem, int size, void *arg)
{
	struct thread	*td;

	td = (struct thread *)mem;

#ifdef INVARIANTS
	/* Verify that this thread is in a safe state to free. */
	switch (td->td_state) {
	case TDS_INHIBITED:
	case TDS_RUNNING:
	case TDS_CAN_RUN:
	case TDS_RUNQ:
		/*
		 * We must never unlink a thread that is in one of
		 * these states, because it is currently active.
		 */
		panic("bad state for thread unlinking");
		/* NOTREACHED */
	case TDS_INACTIVE:
		break;
	default:
		panic("bad thread state");
		/* NOTREACHED */
	}
#endif
}

/*
 * Initialize type-stable parts of a thread (when newly created).
 */
static void
thread_init(void *mem, int size)
{
	struct thread	*td;

	td = (struct thread *)mem;
	mtx_lock(&Giant);
	vm_thread_new(td, 0);
	mtx_unlock(&Giant);
	cpu_thread_setup(td);
	td->td_sched = (struct td_sched *)&td[1];
}

/*
 * Tear down type-stable parts of a thread (just before being discarded).
 */
static void
thread_fini(void *mem, int size)
{
	struct thread	*td;

	td = (struct thread *)mem;
	vm_thread_dispose(td);
}

/*
 * Initialize type-stable parts of a kse (when newly created).
 */
static void
kse_init(void *mem, int size)
{
	struct kse	*ke;

	ke = (struct kse *)mem;
	ke->ke_sched = (struct ke_sched *)&ke[1];
}

/*
 * Initialize type-stable parts of a ksegrp (when newly created).
 */
static void
ksegrp_init(void *mem, int size)
{
	struct ksegrp	*kg;

	kg = (struct ksegrp *)mem;
	kg->kg_sched = (struct kg_sched *)&kg[1];
}

/* 
 * KSE is linked into kse group.
 */
void
kse_link(struct kse *ke, struct ksegrp *kg)
{
	struct proc *p = kg->kg_proc;

	TAILQ_INSERT_HEAD(&kg->kg_kseq, ke, ke_kglist);
	kg->kg_kses++;
	ke->ke_state	= KES_UNQUEUED;
	ke->ke_proc	= p;
	ke->ke_ksegrp	= kg;
	ke->ke_thread	= NULL;
	ke->ke_oncpu	= NOCPU;
	ke->ke_flags	= 0;
}

void
kse_unlink(struct kse *ke)
{
	struct ksegrp *kg;

	mtx_assert(&sched_lock, MA_OWNED);
	kg = ke->ke_ksegrp;
	TAILQ_REMOVE(&kg->kg_kseq, ke, ke_kglist);
	if (ke->ke_state == KES_IDLE) {
		TAILQ_REMOVE(&kg->kg_iq, ke, ke_kgrlist);
		kg->kg_idle_kses--;
	}
	--kg->kg_kses;
	/*
	 * Aggregate stats from the KSE
	 */
	kse_stash(ke);
}

void
ksegrp_link(struct ksegrp *kg, struct proc *p)
{

	TAILQ_INIT(&kg->kg_threads);
	TAILQ_INIT(&kg->kg_runq);	/* links with td_runq */
	TAILQ_INIT(&kg->kg_slpq);	/* links with td_runq */
	TAILQ_INIT(&kg->kg_kseq);	/* all kses in ksegrp */
	TAILQ_INIT(&kg->kg_iq);		/* all idle kses in ksegrp */
	TAILQ_INIT(&kg->kg_upcalls);	/* all upcall structure in ksegrp */
	kg->kg_proc = p;
	/*
	 * the following counters are in the -zero- section
	 * and may not need clearing
	 */
	kg->kg_numthreads = 0;
	kg->kg_runnable   = 0;
	kg->kg_kses       = 0;
	kg->kg_runq_kses  = 0; /* XXXKSE change name */
	kg->kg_idle_kses  = 0;
	kg->kg_numupcalls = 0;
	/* link it in now that it's consistent */
	p->p_numksegrps++;
	TAILQ_INSERT_HEAD(&p->p_ksegrps, kg, kg_ksegrp);
}

void
ksegrp_unlink(struct ksegrp *kg)
{
	struct proc *p;

	mtx_assert(&sched_lock, MA_OWNED);
	KASSERT((kg->kg_numthreads == 0), ("ksegrp_unlink: residual threads"));
	KASSERT((kg->kg_kses == 0), ("ksegrp_unlink: residual kses"));
	KASSERT((kg->kg_numupcalls == 0), ("ksegrp_unlink: residual upcalls"));

	p = kg->kg_proc;
	TAILQ_REMOVE(&p->p_ksegrps, kg, kg_ksegrp);
	p->p_numksegrps--;
	/*
	 * Aggregate stats from the KSE
	 */
	ksegrp_stash(kg);
}

struct kse_upcall *
upcall_alloc(void)
{
	struct kse_upcall *ku;

	ku = uma_zalloc(upcall_zone, M_WAITOK);
	bzero(ku, sizeof(*ku));
	return (ku);
}

void
upcall_free(struct kse_upcall *ku)
{

	uma_zfree(upcall_zone, ku);
}

void
upcall_link(struct kse_upcall *ku, struct ksegrp *kg)
{

	mtx_assert(&sched_lock, MA_OWNED);
	TAILQ_INSERT_TAIL(&kg->kg_upcalls, ku, ku_link);
	ku->ku_ksegrp = kg;
	kg->kg_numupcalls++;
}

void
upcall_unlink(struct kse_upcall *ku)
{
	struct ksegrp *kg = ku->ku_ksegrp;

	mtx_assert(&sched_lock, MA_OWNED);
	KASSERT(ku->ku_owner == NULL, ("%s: have owner", __func__));
	TAILQ_REMOVE(&kg->kg_upcalls, ku, ku_link); 
	kg->kg_numupcalls--;
	upcall_stash(ku);
}

void
upcall_remove(struct thread *td)
{

	if (td->td_upcall) {
		td->td_upcall->ku_owner = NULL;
		upcall_unlink(td->td_upcall);
		td->td_upcall = 0;
	} 
}

/*
 * For a newly created process,
 * link up all the structures and its initial threads etc.
 */
void
proc_linkup(struct proc *p, struct ksegrp *kg,
	    struct kse *ke, struct thread *td)
{

	TAILQ_INIT(&p->p_ksegrps);	     /* all ksegrps in proc */
	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
	TAILQ_INIT(&p->p_suspended);	     /* Threads suspended */
	p->p_numksegrps = 0;
	p->p_numthreads = 0;

	ksegrp_link(kg, p);
	kse_link(ke, kg);
	thread_link(td, kg);
}

/*
struct kse_thr_interrupt_args {
	struct kse_thr_mailbox * tmbx;
	int cmd;
	long data;
};
*/
int
kse_thr_interrupt(struct thread *td, struct kse_thr_interrupt_args *uap)
{
	struct proc *p;
	struct thread *td2;

	p = td->td_proc;

	if (!(p->p_flag & P_SA))
		return (EINVAL);

	switch (uap->cmd) {
	case KSE_INTR_SENDSIG:
		if (uap->data < 0 || uap->data > _SIG_MAXSIG)
			return (EINVAL);
	case KSE_INTR_INTERRUPT:
	case KSE_INTR_RESTART:
		PROC_LOCK(p);
		mtx_lock_spin(&sched_lock);
		FOREACH_THREAD_IN_PROC(p, td2) {
			if (td2->td_mailbox == uap->tmbx)
				break;
		}
		if (td2 == NULL) {
			mtx_unlock_spin(&sched_lock);
			PROC_UNLOCK(p);
			return (ESRCH);
		}
		if (uap->cmd == KSE_INTR_SENDSIG) {
			if (uap->data > 0) {
				td2->td_flags &= ~TDF_INTERRUPT;
				mtx_unlock_spin(&sched_lock);
				tdsignal(td2, (int)uap->data, SIGTARGET_TD);
			} else {
				mtx_unlock_spin(&sched_lock);
			}
		} else {
			td2->td_flags |= TDF_INTERRUPT | TDF_ASTPENDING;
			if (TD_CAN_UNBIND(td2))
				td2->td_upcall->ku_flags |= KUF_DOUPCALL;
			if (uap->cmd == KSE_INTR_INTERRUPT)
				td2->td_intrval = EINTR;
			else
				td2->td_intrval = ERESTART;
			if (TD_ON_SLEEPQ(td2) && (td2->td_flags & TDF_SINTR)) {
				if (td2->td_flags & TDF_CVWAITQ)
					cv_abort(td2);
				else
					abortsleep(td2);
			}
			mtx_unlock_spin(&sched_lock);
		}
		PROC_UNLOCK(p);
		break;
	case KSE_INTR_SIGEXIT:
		if (uap->data < 1 || uap->data > _SIG_MAXSIG)
			return (EINVAL);
		PROC_LOCK(p);
		sigexit(td, (int)uap->data);
		break;
	default:
		return (EINVAL);
	}
	return (0);
}

/*
struct kse_exit_args {
	register_t dummy;
};
*/
int
kse_exit(struct thread *td, struct kse_exit_args *uap)
{
	struct proc *p;
	struct ksegrp *kg;
	struct kse *ke;
	struct kse_upcall *ku, *ku2;
	int    error, count;

	p = td->td_proc;
	if ((ku = td->td_upcall) == NULL || TD_CAN_UNBIND(td))
		return (EINVAL);
	kg = td->td_ksegrp;
	count = 0;
	PROC_LOCK(p);
	mtx_lock_spin(&sched_lock);
	FOREACH_UPCALL_IN_GROUP(kg, ku2) {
		if (ku2->ku_flags & KUF_EXITING)
			count++;
	}
	if ((kg->kg_numupcalls - count) == 1 &&
	    (kg->kg_numthreads > 1)) {
		mtx_unlock_spin(&sched_lock);
		PROC_UNLOCK(p);
		return (EDEADLK);
	}
	ku->ku_flags |= KUF_EXITING;
	mtx_unlock_spin(&sched_lock);
	PROC_UNLOCK(p);
	error = suword(&ku->ku_mailbox->km_flags, ku->ku_mflags|KMF_DONE);
	PROC_LOCK(p);
	if (error)
		psignal(p, SIGSEGV);
	mtx_lock_spin(&sched_lock);
	upcall_remove(td);
	ke = td->td_kse;
	if (p->p_numthreads == 1) {
		kse_purge(p, td);
		p->p_flag &= ~P_SA;
		mtx_unlock_spin(&sched_lock);
		PROC_UNLOCK(p);
	} else {
		if (kg->kg_numthreads == 1) { /* Shutdown a group */
			kse_purge_group(td);
			ke->ke_flags |= KEF_EXIT;
		}
		thread_stopped(p);
		thread_exit();
		/* NOTREACHED */
	}
	return (0);
}

/*
 * Either becomes an upcall or waits for an awakening event and
 * then becomes an upcall. Only error cases return.
 */
/*
struct kse_release_args {
	struct timespec *timeout;
};
*/
int
kse_release(struct thread *td, struct kse_release_args *uap)
{
	struct proc *p;
	struct ksegrp *kg;
	struct kse_upcall *ku;
	struct timespec timeout;
	struct timeval tv;
	sigset_t sigset;
	int error;

	p = td->td_proc;
	kg = td->td_ksegrp;
	if ((ku = td->td_upcall) == NULL || TD_CAN_UNBIND(td))
		return (EINVAL);
	if (uap->timeout != NULL) {
		if ((error = copyin(uap->timeout, &timeout, sizeof(timeout))))
			return (error);
		TIMESPEC_TO_TIMEVAL(&tv, &timeout);
	}
	if (td->td_flags & TDF_SA)
		td->td_pflags |= TDP_UPCALLING;
	else {
		ku->ku_mflags = fuword(&ku->ku_mailbox->km_flags);
		if (ku->ku_mflags == -1) {
			PROC_LOCK(p);
			sigexit(td, SIGSEGV);
		}
	}
	PROC_LOCK(p);
	if (ku->ku_mflags & KMF_WAITSIGEVENT) {
		/* UTS wants to wait for signal event */
		if (!(p->p_flag & P_SIGEVENT) && !(ku->ku_flags & KUF_DOUPCALL))
			error = msleep(&p->p_siglist, &p->p_mtx, PPAUSE|PCATCH,
			    "ksesigwait", (uap->timeout ? tvtohz(&tv) : 0));
		p->p_flag &= ~P_SIGEVENT;
		sigset = p->p_siglist;
		PROC_UNLOCK(p);
		error = copyout(&sigset, &ku->ku_mailbox->km_sigscaught,
		    sizeof(sigset));
	} else {
		 if (! kg->kg_completed && !(ku->ku_flags & KUF_DOUPCALL)) {
			kg->kg_upsleeps++;
			error = msleep(&kg->kg_completed, &p->p_mtx,
				PPAUSE|PCATCH, "kserel",
				(uap->timeout ? tvtohz(&tv) : 0));
			kg->kg_upsleeps--;
		}
		PROC_UNLOCK(p);
	}
	if (ku->ku_flags & KUF_DOUPCALL) {
		mtx_lock_spin(&sched_lock);
		ku->ku_flags &= ~KUF_DOUPCALL;
		mtx_unlock_spin(&sched_lock);
	}
	return (0);
}

/* struct kse_wakeup_args {
	struct kse_mailbox *mbx;
}; */
int
kse_wakeup(struct thread *td, struct kse_wakeup_args *uap)
{
	struct proc *p;
	struct ksegrp *kg;
	struct kse_upcall *ku;
	struct thread *td2;

	p = td->td_proc;
	td2 = NULL;
	ku = NULL;
	/* KSE-enabled processes only, please. */
	if (!(p->p_flag & P_SA))
		return (EINVAL);
	PROC_LOCK(p);
	mtx_lock_spin(&sched_lock);
	if (uap->mbx) {
		FOREACH_KSEGRP_IN_PROC(p, kg) {
			FOREACH_UPCALL_IN_GROUP(kg, ku) {
				if (ku->ku_mailbox == uap->mbx)
					break;
			}
			if (ku)
				break;
		}
	} else {
		kg = td->td_ksegrp;
		if (kg->kg_upsleeps) {
			wakeup_one(&kg->kg_completed);
			mtx_unlock_spin(&sched_lock);
			PROC_UNLOCK(p);
			return (0);
		}
		ku = TAILQ_FIRST(&kg->kg_upcalls);
	}
	if (ku) {
		if ((td2 = ku->ku_owner) == NULL) {
			panic("%s: no owner", __func__);
		} else if (TD_ON_SLEEPQ(td2) &&
		           ((td2->td_wchan == &kg->kg_completed) ||
			    (td2->td_wchan == &p->p_siglist &&
			     (ku->ku_mflags & KMF_WAITSIGEVENT)))) {
			abortsleep(td2);
		} else {
			ku->ku_flags |= KUF_DOUPCALL;
		}
		mtx_unlock_spin(&sched_lock);
		PROC_UNLOCK(p);
		return (0);
	}
	mtx_unlock_spin(&sched_lock);
	PROC_UNLOCK(p);
	return (ESRCH);
}

/* 
 * No new KSEG: first call: use current KSE, don't schedule an upcall
 * All other situations, do allocate max new KSEs and schedule an upcall.
 */
/* struct kse_create_args {
	struct kse_mailbox *mbx;
	int newgroup;
}; */
int
kse_create(struct thread *td, struct kse_create_args *uap)
{
	struct kse *newke;
	struct ksegrp *newkg;
	struct ksegrp *kg;
	struct proc *p;
	struct kse_mailbox mbx;
	struct kse_upcall *newku;
	int err, ncpus, sa = 0, first = 0;
	struct thread *newtd;

	p = td->td_proc;
	if ((err = copyin(uap->mbx, &mbx, sizeof(mbx))))
		return (err);

	/* Too bad, why hasn't kernel always a cpu counter !? */
#ifdef SMP
	ncpus = mp_ncpus;
#else
	ncpus = 1;
#endif
	if (virtual_cpu != 0)
		ncpus = virtual_cpu;
	if (!(mbx.km_flags & KMF_BOUND))
		sa = TDF_SA;
	else
		ncpus = 1;
	PROC_LOCK(p);
	if (!(p->p_flag & P_SA)) {
		first = 1;
		p->p_flag |= P_SA;
	}
	PROC_UNLOCK(p);
	if (!sa && !uap->newgroup && !first)
		return (EINVAL);
	kg = td->td_ksegrp;
	if (uap->newgroup) {
		/* Have race condition but it is cheap */ 
		if (p->p_numksegrps >= max_groups_per_proc)
			return (EPROCLIM);
		/* 
		 * If we want a new KSEGRP it doesn't matter whether
		 * we have already fired up KSE mode before or not.
		 * We put the process in KSE mode and create a new KSEGRP.
		 */
		newkg = ksegrp_alloc();
		bzero(&newkg->kg_startzero, RANGEOF(struct ksegrp,
		      kg_startzero, kg_endzero));
		bcopy(&kg->kg_startcopy, &newkg->kg_startcopy,
		      RANGEOF(struct ksegrp, kg_startcopy, kg_endcopy));
		PROC_LOCK(p);      
		mtx_lock_spin(&sched_lock);
		if (p->p_numksegrps >= max_groups_per_proc) {
			mtx_unlock_spin(&sched_lock);
			PROC_UNLOCK(p);
			ksegrp_free(newkg);
			return (EPROCLIM);
		}
		ksegrp_link(newkg, p);
		sched_fork_ksegrp(kg, newkg);
		mtx_unlock_spin(&sched_lock);
		PROC_UNLOCK(p);
	} else {
		if (!first && ((td->td_flags & TDF_SA) ^ sa) != 0)
			return (EINVAL);
		newkg = kg;
	}

	/*
	 * Creating upcalls more than number of physical cpu does
	 * not help performance. 
	 */
	if (newkg->kg_numupcalls >= ncpus)
		return (EPROCLIM);

	if (newkg->kg_numupcalls == 0) {
		/*
		 * Initialize KSE group
		 *
		 * For multiplxed group, create KSEs as many as physical
		 * cpus. This increases concurrent even if userland
		 * is not MP safe and can only run on single CPU.
		 * In ideal world, every physical cpu should execute a thread.
		 * If there is enough KSEs, threads in kernel can be
		 * executed parallel on different cpus with full speed, 
		 * Concurrent in kernel shouldn't be restricted by number of 
		 * upcalls userland provides. Adding more upcall structures
		 * only increases concurrent in userland.
		 *
		 * For bound thread group, because there is only thread in the
		 * group, we only create one KSE for the group. Thread in this
		 * kind of group will never schedule an upcall when blocked,
		 * this intends to simulate pthread system scope thread.
		 */
		while (newkg->kg_kses < ncpus) {
			newke = kse_alloc();
			bzero(&newke->ke_startzero, RANGEOF(struct kse,
			      ke_startzero, ke_endzero));
#if 0
			mtx_lock_spin(&sched_lock);
			bcopy(&ke->ke_startcopy, &newke->ke_startcopy,
			      RANGEOF(struct kse, ke_startcopy, ke_endcopy));
			mtx_unlock_spin(&sched_lock);
#endif
			mtx_lock_spin(&sched_lock);
			kse_link(newke, newkg);
			sched_fork_kse(td->td_kse, newke);
			/* Add engine */
			kse_reassign(newke);
			mtx_unlock_spin(&sched_lock);
		}
	}
	newku = upcall_alloc();
	newku->ku_mailbox = uap->mbx;
	newku->ku_func = mbx.km_func;
	bcopy(&mbx.km_stack, &newku->ku_stack, sizeof(stack_t));

	/* For the first call this may not have been set */
	if (td->td_standin == NULL)
		thread_alloc_spare(td, NULL);

	PROC_LOCK(p);
	if (newkg->kg_numupcalls >= ncpus) {
		PROC_UNLOCK(p);
		upcall_free(newku);
		return (EPROCLIM);
	}
	if (first && sa) {
		SIGSETOR(p->p_siglist, td->td_siglist);
		SIGEMPTYSET(td->td_siglist);
		SIGFILLSET(td->td_sigmask);
		SIG_CANTMASK(td->td_sigmask);
	}
	mtx_lock_spin(&sched_lock);
	PROC_UNLOCK(p);
	upcall_link(newku, newkg);
	if (mbx.km_quantum)
		newkg->kg_upquantum = max(1, mbx.km_quantum/tick);

	/*
	 * Each upcall structure has an owner thread, find which
	 * one owns it.
	 */
	if (uap->newgroup) {
		/* 
		 * Because new ksegrp hasn't thread,
		 * create an initial upcall thread to own it.
		 */
		newtd = thread_schedule_upcall(td, newku);
	} else {
		/*
		 * If current thread hasn't an upcall structure,
		 * just assign the upcall to it.
		 */
		if (td->td_upcall == NULL) {
			newku->ku_owner = td;
			td->td_upcall = newku;
			newtd = td;
		} else {
			/*
			 * Create a new upcall thread to own it.
			 */
			newtd = thread_schedule_upcall(td, newku);
		}
	}
	if (!sa) {
		newtd->td_mailbox = mbx.km_curthread;
		newtd->td_flags &= ~TDF_SA;
		if (newtd != td) {
			mtx_unlock_spin(&sched_lock);
			cpu_set_upcall_kse(newtd, newku);
			mtx_lock_spin(&sched_lock);
		}
	} else {
		newtd->td_flags |= TDF_SA;
	}
	if (newtd != td)
		setrunqueue(newtd);
	mtx_unlock_spin(&sched_lock);
	return (0);
}

/*
 * Initialize global thread allocation resources.
 */
void
threadinit(void)
{

	thread_zone = uma_zcreate("THREAD", sched_sizeof_thread(),
	    thread_ctor, thread_dtor, thread_init, thread_fini,
	    UMA_ALIGN_CACHE, 0);
	ksegrp_zone = uma_zcreate("KSEGRP", sched_sizeof_ksegrp(),
	    NULL, NULL, ksegrp_init, NULL,
	    UMA_ALIGN_CACHE, 0);
	kse_zone = uma_zcreate("KSE", sched_sizeof_kse(),
	    NULL, NULL, kse_init, NULL,
	    UMA_ALIGN_CACHE, 0);
	upcall_zone = uma_zcreate("UPCALL", sizeof(struct kse_upcall),
	    NULL, NULL, NULL, NULL, UMA_ALIGN_CACHE, 0);
}

/*
 * Stash an embarasingly extra thread into the zombie thread queue.
 */
void
thread_stash(struct thread *td)
{
	mtx_lock_spin(&kse_zombie_lock);
	TAILQ_INSERT_HEAD(&zombie_threads, td, td_runq);
	mtx_unlock_spin(&kse_zombie_lock);
}

/*
 * Stash an embarasingly extra kse into the zombie kse queue.
 */
void
kse_stash(struct kse *ke)
{
	mtx_lock_spin(&kse_zombie_lock);
	TAILQ_INSERT_HEAD(&zombie_kses, ke, ke_procq);
	mtx_unlock_spin(&kse_zombie_lock);
}

/*
 * Stash an embarasingly extra upcall into the zombie upcall queue.
 */

void
upcall_stash(struct kse_upcall *ku)
{
	mtx_lock_spin(&kse_zombie_lock);
	TAILQ_INSERT_HEAD(&zombie_upcalls, ku, ku_link);
	mtx_unlock_spin(&kse_zombie_lock);
}

/*
 * Stash an embarasingly extra ksegrp into the zombie ksegrp queue.
 */
void
ksegrp_stash(struct ksegrp *kg)
{
	mtx_lock_spin(&kse_zombie_lock);
	TAILQ_INSERT_HEAD(&zombie_ksegrps, kg, kg_ksegrp);
	mtx_unlock_spin(&kse_zombie_lock);
}

/*
 * Reap zombie kse resource.
 */
void
thread_reap(void)
{
	struct thread *td_first, *td_next;
	struct kse *ke_first, *ke_next;
	struct ksegrp *kg_first, * kg_next;
	struct kse_upcall *ku_first, *ku_next;

	/*
	 * Don't even bother to lock if none at this instant,
	 * we really don't care about the next instant..
	 */
	if ((!TAILQ_EMPTY(&zombie_threads))
	    || (!TAILQ_EMPTY(&zombie_kses))
	    || (!TAILQ_EMPTY(&zombie_ksegrps))
	    || (!TAILQ_EMPTY(&zombie_upcalls))) {
		mtx_lock_spin(&kse_zombie_lock);
		td_first = TAILQ_FIRST(&zombie_threads);
		ke_first = TAILQ_FIRST(&zombie_kses);
		kg_first = TAILQ_FIRST(&zombie_ksegrps);
		ku_first = TAILQ_FIRST(&zombie_upcalls);
		if (td_first)
			TAILQ_INIT(&zombie_threads);
		if (ke_first)
			TAILQ_INIT(&zombie_kses);
		if (kg_first)
			TAILQ_INIT(&zombie_ksegrps);
		if (ku_first)
			TAILQ_INIT(&zombie_upcalls);
		mtx_unlock_spin(&kse_zombie_lock);
		while (td_first) {
			td_next = TAILQ_NEXT(td_first, td_runq);
			if (td_first->td_ucred)
				crfree(td_first->td_ucred);
			thread_free(td_first);
			td_first = td_next;
		}
		while (ke_first) {
			ke_next = TAILQ_NEXT(ke_first, ke_procq);
			kse_free(ke_first);
			ke_first = ke_next;
		}
		while (kg_first) {
			kg_next = TAILQ_NEXT(kg_first, kg_ksegrp);
			ksegrp_free(kg_first);
			kg_first = kg_next;
		}
		while (ku_first) {
			ku_next = TAILQ_NEXT(ku_first, ku_link);
			upcall_free(ku_first);
			ku_first = ku_next;
		}
	}
}

/*
 * Allocate a ksegrp.
 */
struct ksegrp *
ksegrp_alloc(void)
{
	return (uma_zalloc(ksegrp_zone, M_WAITOK));
}

/*
 * Allocate a kse.
 */
struct kse *
kse_alloc(void)
{
	return (uma_zalloc(kse_zone, M_WAITOK));
}

/*
 * Allocate a thread.
 */
struct thread *
thread_alloc(void)
{
	thread_reap(); /* check if any zombies to get */
	return (uma_zalloc(thread_zone, M_WAITOK));
}

/*
 * Deallocate a ksegrp.
 */
void
ksegrp_free(struct ksegrp *td)
{
	uma_zfree(ksegrp_zone, td);
}

/*
 * Deallocate a kse.
 */
void
kse_free(struct kse *td)
{
	uma_zfree(kse_zone, td);
}

/*
 * Deallocate a thread.
 */
void
thread_free(struct thread *td)
{

	cpu_thread_clean(td);
	uma_zfree(thread_zone, td);
}

/*
 * Store the thread context in the UTS's mailbox.
 * then add the mailbox at the head of a list we are building in user space.
 * The list is anchored in the ksegrp structure.
 */
int
thread_export_context(struct thread *td, int willexit)
{
	struct proc *p;
	struct ksegrp *kg;
	uintptr_t mbx;
	void *addr;
	int error = 0, temp, sig;
	mcontext_t mc;

	p = td->td_proc;
	kg = td->td_ksegrp;

	/* Export the user/machine context. */
	get_mcontext(td, &mc, 0);
	addr = (void *)(&td->td_mailbox->tm_context.uc_mcontext);
	error = copyout(&mc, addr, sizeof(mcontext_t));
	if (error)
		goto bad;

	/* Exports clock ticks in kernel mode */
	addr = (caddr_t)(&td->td_mailbox->tm_sticks);
	temp = fuword32(addr) + td->td_usticks;
	if (suword32(addr, temp)) {
		error = EFAULT;
		goto bad;
	}

	/*
	 * Post sync signal, or process SIGKILL and SIGSTOP.
	 * For sync signal, it is only possible when the signal is not
	 * caught by userland or process is being debugged.
	 */
	PROC_LOCK(p);
	if (td->td_flags & TDF_NEEDSIGCHK) {
		mtx_lock_spin(&sched_lock);
		td->td_flags &= ~TDF_NEEDSIGCHK;
		mtx_unlock_spin(&sched_lock);
		mtx_lock(&p->p_sigacts->ps_mtx);
		while ((sig = cursig(td)) != 0)
			postsig(sig);
		mtx_unlock(&p->p_sigacts->ps_mtx);
	}
	if (willexit)
		SIGFILLSET(td->td_sigmask);
	PROC_UNLOCK(p);

	/* Get address in latest mbox of list pointer */
	addr = (void *)(&td->td_mailbox->tm_next);
	/*
	 * Put the saved address of the previous first
	 * entry into this one
	 */
	for (;;) {
		mbx = (uintptr_t)kg->kg_completed;
		if (suword(addr, mbx)) {
			error = EFAULT;
			goto bad;
		}
		PROC_LOCK(p);
		if (mbx == (uintptr_t)kg->kg_completed) {
			kg->kg_completed = td->td_mailbox;
			/*
			 * The thread context may be taken away by
			 * other upcall threads when we unlock
			 * process lock. it's no longer valid to
			 * use it again in any other places.
			 */
			td->td_mailbox = NULL;
			PROC_UNLOCK(p);
			break;
		}
		PROC_UNLOCK(p);
	}
	td->td_usticks = 0;
	return (0);

bad:
	PROC_LOCK(p);
	sigexit(td, SIGILL);
	return (error);
}

/*
 * Take the list of completed mailboxes for this KSEGRP and put them on this
 * upcall's mailbox as it's the next one going up.
 */
static int
thread_link_mboxes(struct ksegrp *kg, struct kse_upcall *ku)
{
	struct proc *p = kg->kg_proc;
	void *addr;
	uintptr_t mbx;

	addr = (void *)(&ku->ku_mailbox->km_completed);
	for (;;) {
		mbx = (uintptr_t)kg->kg_completed;
		if (suword(addr, mbx)) {
			PROC_LOCK(p);
			psignal(p, SIGSEGV);
			PROC_UNLOCK(p);
			return (EFAULT);
		}
		PROC_LOCK(p);
		if (mbx == (uintptr_t)kg->kg_completed) {
			kg->kg_completed = NULL;
			PROC_UNLOCK(p);
			break;
		}
		PROC_UNLOCK(p);
	}
	return (0);
}

/*
 * This function should be called at statclock interrupt time
 */
int
thread_statclock(int user)
{
	struct thread *td = curthread;
	struct ksegrp *kg = td->td_ksegrp;
	
	if (kg->kg_numupcalls == 0 || !(td->td_flags & TDF_SA))
		return (0);
	if (user) {
		/* Current always do via ast() */
		mtx_lock_spin(&sched_lock);
		td->td_flags |= (TDF_USTATCLOCK|TDF_ASTPENDING);
		mtx_unlock_spin(&sched_lock);
		td->td_uuticks++;
	} else {
		if (td->td_mailbox != NULL)
			td->td_usticks++;
		else {
			/* XXXKSE
		 	 * We will call thread_user_enter() for every
			 * kernel entry in future, so if the thread mailbox
			 * is NULL, it must be a UTS kernel, don't account
			 * clock ticks for it.
			 */
		}
	}
	return (0);
}

/*
 * Export state clock ticks for userland
 */
static int
thread_update_usr_ticks(struct thread *td, int user)
{
	struct proc *p = td->td_proc;
	struct kse_thr_mailbox *tmbx;
	struct kse_upcall *ku;
	struct ksegrp *kg;
	caddr_t addr;
	u_int uticks;

	if ((ku = td->td_upcall) == NULL)
		return (-1);
	
	tmbx = (void *)fuword((void *)&ku->ku_mailbox->km_curthread);
	if ((tmbx == NULL) || (tmbx == (void *)-1))
		return (-1);
	if (user) {
		uticks = td->td_uuticks;
		td->td_uuticks = 0;
		addr = (caddr_t)&tmbx->tm_uticks;
	} else {
		uticks = td->td_usticks;
		td->td_usticks = 0;
		addr = (caddr_t)&tmbx->tm_sticks;
	}
	if (uticks) {
		if (suword32(addr, uticks+fuword32(addr))) {
			PROC_LOCK(p);
			psignal(p, SIGSEGV);
			PROC_UNLOCK(p);
			return (-2);
		}
	}
	kg = td->td_ksegrp;
	if (kg->kg_upquantum && ticks >= kg->kg_nextupcall) {
		mtx_lock_spin(&sched_lock);
		td->td_upcall->ku_flags |= KUF_DOUPCALL;
		mtx_unlock_spin(&sched_lock);
	}
	return (0);
}

/*
 * Discard the current thread and exit from its context.
 *
 * Because we can't free a thread while we're operating under its context,
 * push the current thread into our CPU's deadthread holder. This means
 * we needn't worry about someone else grabbing our context before we
 * do a cpu_throw().
 */
void
thread_exit(void)
{
	struct thread *td;
	struct kse *ke;
	struct proc *p;
	struct ksegrp	*kg;

	td = curthread;
	kg = td->td_ksegrp;
	p = td->td_proc;
	ke = td->td_kse;

	mtx_assert(&sched_lock, MA_OWNED);
	KASSERT(p != NULL, ("thread exiting without a process"));
	KASSERT(ke != NULL, ("thread exiting without a kse"));
	KASSERT(kg != NULL, ("thread exiting without a kse group"));
	PROC_LOCK_ASSERT(p, MA_OWNED);
	CTR1(KTR_PROC, "thread_exit: thread %p", td);
	KASSERT(!mtx_owned(&Giant), ("dying thread owns giant"));

	if (td->td_standin != NULL) {
		thread_stash(td->td_standin);
		td->td_standin = NULL;
	}

	cpu_thread_exit(td);	/* XXXSMP */

	/*
	 * The last thread is left attached to the process
	 * So that the whole bundle gets recycled. Skip
	 * all this stuff.
	 */
	if (p->p_numthreads > 1) {
		thread_unlink(td);
		if (p->p_maxthrwaits)
			wakeup(&p->p_numthreads);
		/*
		 * The test below is NOT true if we are the
		 * sole exiting thread. P_STOPPED_SNGL is unset
		 * in exit1() after it is the only survivor.
		 */
		if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
			if (p->p_numthreads == p->p_suspcount) {
				thread_unsuspend_one(p->p_singlethread);
			}
		}

		/*
		 * Because each upcall structure has an owner thread,
		 * owner thread exits only when process is in exiting
		 * state, so upcall to userland is no longer needed,
		 * deleting upcall structure is safe here.
		 * So when all threads in a group is exited, all upcalls
		 * in the group should be automatically freed.
		 */
		if (td->td_upcall)
			upcall_remove(td);
	
		sched_exit_thread(FIRST_THREAD_IN_PROC(p), td);
		sched_exit_kse(FIRST_KSE_IN_PROC(p), ke);
		ke->ke_state = KES_UNQUEUED;
		ke->ke_thread = NULL;
		/* 
		 * Decide what to do with the KSE attached to this thread.
		 */
		if (ke->ke_flags & KEF_EXIT) {
			kse_unlink(ke);
			if (kg->kg_kses == 0) {
				sched_exit_ksegrp(FIRST_KSEGRP_IN_PROC(p), kg);
				ksegrp_unlink(kg);
			}
		}
		else
			kse_reassign(ke);
		PROC_UNLOCK(p);
		td->td_kse	= NULL;
		td->td_state	= TDS_INACTIVE;
#if 0
		td->td_proc	= NULL;
#endif
		td->td_ksegrp	= NULL;
		td->td_last_kse	= NULL;
		PCPU_SET(deadthread, td);
	} else {
		PROC_UNLOCK(p);
	}
	/* XXX Shouldn't cpu_throw() here. */
	mtx_assert(&sched_lock, MA_OWNED);
	cpu_throw(td, choosethread());
	panic("I'm a teapot!");
	/* NOTREACHED */
}

/* 
 * Do any thread specific cleanups that may be needed in wait()
 * called with Giant held, proc and schedlock not held.
 */
void
thread_wait(struct proc *p)
{
	struct thread *td;

	KASSERT((p->p_numthreads == 1), ("Muliple threads in wait1()"));
	KASSERT((p->p_numksegrps == 1), ("Muliple ksegrps in wait1()"));
	FOREACH_THREAD_IN_PROC(p, td) {
		if (td->td_standin != NULL) {
			thread_free(td->td_standin);
			td->td_standin = NULL;
		}
		cpu_thread_clean(td);
	}
	thread_reap();	/* check for zombie threads etc. */
}

/*
 * Link a thread to a process.
 * set up anything that needs to be initialized for it to
 * be used by the process.
 *
 * Note that we do not link to the proc's ucred here.
 * The thread is linked as if running but no KSE assigned.
 */
void
thread_link(struct thread *td, struct ksegrp *kg)
{
	struct proc *p;

	p = kg->kg_proc;
	td->td_state    = TDS_INACTIVE;
	td->td_proc     = p;
	td->td_ksegrp   = kg;
	td->td_last_kse = NULL;
	td->td_flags    = 0;
	td->td_kse      = NULL;

	LIST_INIT(&td->td_contested);
	callout_init(&td->td_slpcallout, CALLOUT_MPSAFE);
	TAILQ_INSERT_HEAD(&p->p_threads, td, td_plist);
	TAILQ_INSERT_HEAD(&kg->kg_threads, td, td_kglist);
	p->p_numthreads++;
	kg->kg_numthreads++;
}

void
thread_unlink(struct thread *td)
{      
	struct proc *p = td->td_proc;
	struct ksegrp *kg = td->td_ksegrp;

	mtx_assert(&sched_lock, MA_OWNED);
	TAILQ_REMOVE(&p->p_threads, td, td_plist);
	p->p_numthreads--;
	TAILQ_REMOVE(&kg->kg_threads, td, td_kglist);
	kg->kg_numthreads--;
	/* could clear a few other things here */
} 

/*
 * Purge a ksegrp resource. When a ksegrp is preparing to
 * exit, it calls this function. 
 */
static void
kse_purge_group(struct thread *td)
{
	struct ksegrp *kg;
	struct kse *ke;

	kg = td->td_ksegrp;
 	KASSERT(kg->kg_numthreads == 1, ("%s: bad thread number", __func__));
	while ((ke = TAILQ_FIRST(&kg->kg_iq)) != NULL) {
		KASSERT(ke->ke_state == KES_IDLE,
			("%s: wrong idle KSE state", __func__));
		kse_unlink(ke);
	}
	KASSERT((kg->kg_kses == 1),
		("%s: ksegrp still has %d KSEs", __func__, kg->kg_kses));
	KASSERT((kg->kg_numupcalls == 0),
	        ("%s: ksegrp still has %d upcall datas",
		__func__, kg->kg_numupcalls));
}

/*
 * Purge a process's KSE resource. When a process is preparing to 
 * exit, it calls kse_purge to release any extra KSE resources in 
 * the process.
 */
static void
kse_purge(struct proc *p, struct thread *td)
{
	struct ksegrp *kg;
	struct kse *ke;

 	KASSERT(p->p_numthreads == 1, ("bad thread number"));
	while ((kg = TAILQ_FIRST(&p->p_ksegrps)) != NULL) {
		TAILQ_REMOVE(&p->p_ksegrps, kg, kg_ksegrp);
		p->p_numksegrps--;
		/*
		 * There is no ownership for KSE, after all threads
		 * in the group exited, it is possible that some KSEs 
		 * were left in idle queue, gc them now.
		 */
		while ((ke = TAILQ_FIRST(&kg->kg_iq)) != NULL) {
			KASSERT(ke->ke_state == KES_IDLE,
			   ("%s: wrong idle KSE state", __func__));
			TAILQ_REMOVE(&kg->kg_iq, ke, ke_kgrlist);
			kg->kg_idle_kses--;
			TAILQ_REMOVE(&kg->kg_kseq, ke, ke_kglist);
			kg->kg_kses--;
			kse_stash(ke);
		}
		KASSERT(((kg->kg_kses == 0) && (kg != td->td_ksegrp)) ||
		        ((kg->kg_kses == 1) && (kg == td->td_ksegrp)),
		        ("ksegrp has wrong kg_kses: %d", kg->kg_kses));
		KASSERT((kg->kg_numupcalls == 0),
		        ("%s: ksegrp still has %d upcall datas",
			__func__, kg->kg_numupcalls));
	
		if (kg != td->td_ksegrp)
			ksegrp_stash(kg);
	}
	TAILQ_INSERT_HEAD(&p->p_ksegrps, td->td_ksegrp, kg_ksegrp);
	p->p_numksegrps++;
}

/*
 * This function is intended to be used to initialize a spare thread
 * for upcall. Initialize thread's large data area outside sched_lock
 * for thread_schedule_upcall().
 */
void
thread_alloc_spare(struct thread *td, struct thread *spare)
{
	if (td->td_standin)
		return;
	if (spare == NULL)
		spare = thread_alloc();
	td->td_standin = spare;
	bzero(&spare->td_startzero,
	    (unsigned)RANGEOF(struct thread, td_startzero, td_endzero));
	spare->td_proc = td->td_proc;
	spare->td_ucred = crhold(td->td_ucred);
}

/*
 * Create a thread and schedule it for upcall on the KSE given.
 * Use our thread's standin so that we don't have to allocate one.
 */
struct thread *
thread_schedule_upcall(struct thread *td, struct kse_upcall *ku)
{
	struct thread *td2;

	mtx_assert(&sched_lock, MA_OWNED);

	/* 
	 * Schedule an upcall thread on specified kse_upcall,
	 * the kse_upcall must be free.
	 * td must have a spare thread.
	 */
	KASSERT(ku->ku_owner == NULL, ("%s: upcall has owner", __func__));
	if ((td2 = td->td_standin) != NULL) {
		td->td_standin = NULL;
	} else {
		panic("no reserve thread when scheduling an upcall");
		return (NULL);
	}
	CTR3(KTR_PROC, "thread_schedule_upcall: thread %p (pid %d, %s)",
	     td2, td->td_proc->p_pid, td->td_proc->p_comm);
	bcopy(&td->td_startcopy, &td2->td_startcopy,
	    (unsigned) RANGEOF(struct thread, td_startcopy, td_endcopy));
	thread_link(td2, ku->ku_ksegrp);
	/* inherit blocked thread's context */
	cpu_set_upcall(td2, td);
	/* Let the new thread become owner of the upcall */
	ku->ku_owner   = td2;
	td2->td_upcall = ku;
	td2->td_flags  = TDF_SA;
	td2->td_pflags = TDP_UPCALLING;
	td2->td_kse    = NULL;
	td2->td_state  = TDS_CAN_RUN;
	td2->td_inhibitors = 0;
	SIGFILLSET(td2->td_sigmask);
	SIG_CANTMASK(td2->td_sigmask);
	sched_fork_thread(td, td2);
	return (td2);	/* bogus.. should be a void function */
}

/*
 * It is only used when thread generated a trap and process is being
 * debugged.
 */
void
thread_signal_add(struct thread *td, int sig)
{
	struct proc *p;
	siginfo_t siginfo;
	struct sigacts *ps;
	int error;

	p = td->td_proc;
	PROC_LOCK_ASSERT(p, MA_OWNED);
	ps = p->p_sigacts;
	mtx_assert(&ps->ps_mtx, MA_OWNED);

	cpu_thread_siginfo(sig, 0, &siginfo);
	mtx_unlock(&ps->ps_mtx);
	PROC_UNLOCK(p);
	error = copyout(&siginfo, &td->td_mailbox->tm_syncsig, sizeof(siginfo));
	if (error) {
		PROC_LOCK(p);
		sigexit(td, SIGILL);
	}
	PROC_LOCK(p);
	SIGADDSET(td->td_sigmask, sig);
	mtx_lock(&ps->ps_mtx);
}

void
thread_switchout(struct thread *td)
{
	struct kse_upcall *ku;
	struct thread *td2;

	mtx_assert(&sched_lock, MA_OWNED);

	/*
	 * If the outgoing thread is in threaded group and has never
	 * scheduled an upcall, decide whether this is a short
	 * or long term event and thus whether or not to schedule
	 * an upcall.
	 * If it is a short term event, just suspend it in
	 * a way that takes its KSE with it.
	 * Select the events for which we want to schedule upcalls.
	 * For now it's just sleep.
	 * XXXKSE eventually almost any inhibition could do.
	 */
	if (TD_CAN_UNBIND(td) && (td->td_standin) && TD_ON_SLEEPQ(td)) {
		/* 
		 * Release ownership of upcall, and schedule an upcall
		 * thread, this new upcall thread becomes the owner of
		 * the upcall structure.
		 */
		ku = td->td_upcall;
		ku->ku_owner = NULL;
		td->td_upcall = NULL; 
		td->td_flags &= ~TDF_CAN_UNBIND;
		td2 = thread_schedule_upcall(td, ku);
		setrunqueue(td2);
	}
}

/*
 * Setup done on the thread when it enters the kernel.
 * XXXKSE Presently only for syscalls but eventually all kernel entries.
 */
void
thread_user_enter(struct proc *p, struct thread *td)
{
	struct ksegrp *kg;
	struct kse_upcall *ku;
	struct kse_thr_mailbox *tmbx;
	uint32_t tflags;

	kg = td->td_ksegrp;

	/*
	 * First check that we shouldn't just abort.
	 * But check if we are the single thread first!
	 */
	if (p->p_flag & P_SINGLE_EXIT) {
		PROC_LOCK(p);
		mtx_lock_spin(&sched_lock);
		thread_stopped(p);
		thread_exit();
		/* NOTREACHED */
	}

	/*
	 * If we are doing a syscall in a KSE environment,
	 * note where our mailbox is. There is always the
	 * possibility that we could do this lazily (in kse_reassign()),
	 * but for now do it every time.
	 */
	kg = td->td_ksegrp;
	if (td->td_flags & TDF_SA) {
		ku = td->td_upcall;
		KASSERT(ku, ("%s: no upcall owned", __func__));
		KASSERT((ku->ku_owner == td), ("%s: wrong owner", __func__));
		KASSERT(!TD_CAN_UNBIND(td), ("%s: can unbind", __func__));
		ku->ku_mflags = fuword32((void *)&ku->ku_mailbox->km_flags);
		tmbx = (void *)fuword((void *)&ku->ku_mailbox->km_curthread);
		if ((tmbx == NULL) || (tmbx == (void *)-1L) ||
		    (ku->ku_mflags & KMF_NOUPCALL)) {
			td->td_mailbox = NULL;
		} else {
			if (td->td_standin == NULL)
				thread_alloc_spare(td, NULL);
			tflags = fuword32(&tmbx->tm_flags);
			/*
			 * On some architectures, TP register points to thread
			 * mailbox but not points to kse mailbox, and userland 
			 * can not atomically clear km_curthread, but can 
			 * use TP register, and set TMF_NOUPCALL in thread
			 * flag	to indicate a critical region.
			 */
			if (tflags & TMF_NOUPCALL) {
				td->td_mailbox = NULL;
			} else {
				td->td_mailbox = tmbx;
				mtx_lock_spin(&sched_lock);
				td->td_flags |= TDF_CAN_UNBIND;
				mtx_unlock_spin(&sched_lock);
			}
		}
	}
}

/*
 * The extra work we go through if we are a threaded process when we
 * return to userland.
 *
 * If we are a KSE process and returning to user mode, check for
 * extra work to do before we return (e.g. for more syscalls
 * to complete first).  If we were in a critical section, we should
 * just return to let it finish. Same if we were in the UTS (in
 * which case the mailbox's context's busy indicator will be set).
 * The only traps we suport will have set the mailbox.
 * We will clear it here.
 */
int
thread_userret(struct thread *td, struct trapframe *frame)
{
	int error = 0, upcalls, uts_crit;
	struct kse_upcall *ku;
	struct ksegrp *kg, *kg2;
	struct proc *p;
	struct timespec ts;

	p = td->td_proc;
	kg = td->td_ksegrp;
	ku = td->td_upcall;

	/* Nothing to do with bound thread */
	if (!(td->td_flags & TDF_SA))
		return (0);

	/*
	 * Stat clock interrupt hit in userland, it 
	 * is returning from interrupt, charge thread's
	 * userland time for UTS.
	 */
	if (td->td_flags & TDF_USTATCLOCK) {
		thread_update_usr_ticks(td, 1);
		mtx_lock_spin(&sched_lock);
		td->td_flags &= ~TDF_USTATCLOCK;
		mtx_unlock_spin(&sched_lock);
		if (kg->kg_completed ||
		    (td->td_upcall->ku_flags & KUF_DOUPCALL))
			thread_user_enter(p, td);
	}

	uts_crit = (td->td_mailbox == NULL);
	/* 
	 * Optimisation:
	 * This thread has not started any upcall.
	 * If there is no work to report other than ourself,
	 * then it can return direct to userland.
	 */
	if (TD_CAN_UNBIND(td)) {
		mtx_lock_spin(&sched_lock);
		td->td_flags &= ~TDF_CAN_UNBIND;
		if ((td->td_flags & TDF_NEEDSIGCHK) == 0 &&
		    (kg->kg_completed == NULL) &&
		    (ku->ku_flags & KUF_DOUPCALL) == 0 &&
		    (kg->kg_upquantum && ticks < kg->kg_nextupcall)) {
			mtx_unlock_spin(&sched_lock);
			thread_update_usr_ticks(td, 0);
			nanotime(&ts);
			error = copyout(&ts,
				(caddr_t)&ku->ku_mailbox->km_timeofday,
				sizeof(ts));
			td->td_mailbox = 0;
			ku->ku_mflags = 0;
			if (error)
				goto out;
			return (0);
		}
		mtx_unlock_spin(&sched_lock);
		thread_export_context(td, 0);
		/*
		 * There is something to report, and we own an upcall
		 * strucuture, we can go to userland.
		 * Turn ourself into an upcall thread.
		 */
		td->td_pflags |= TDP_UPCALLING;
	} else if (td->td_mailbox && (ku == NULL)) {
		thread_export_context(td, 1);
		PROC_LOCK(p);
		/*
		 * There are upcall threads waiting for
		 * work to do, wake one of them up.
		 * XXXKSE Maybe wake all of them up. 
		 */
		if (kg->kg_upsleeps)
			wakeup_one(&kg->kg_completed);
		mtx_lock_spin(&sched_lock);
		thread_stopped(p);
		thread_exit();
		/* NOTREACHED */
	}

	KASSERT(ku != NULL, ("upcall is NULL\n"));
	KASSERT(TD_CAN_UNBIND(td) == 0, ("can unbind"));

	if (p->p_numthreads > max_threads_per_proc) {
		max_threads_hits++;
		PROC_LOCK(p);
		mtx_lock_spin(&sched_lock);
		p->p_maxthrwaits++;
		while (p->p_numthreads > max_threads_per_proc) {
			upcalls = 0;
			FOREACH_KSEGRP_IN_PROC(p, kg2) {
				if (kg2->kg_numupcalls == 0)
					upcalls++;
				else
					upcalls += kg2->kg_numupcalls;
			}
			if (upcalls >= max_threads_per_proc)
				break;
			mtx_unlock_spin(&sched_lock);
			if (msleep(&p->p_numthreads, &p->p_mtx, PPAUSE|PCATCH,
			    "maxthreads", NULL)) {
				mtx_lock_spin(&sched_lock);
				break;
			} else {
				mtx_lock_spin(&sched_lock);
			}
		}
		p->p_maxthrwaits--;
		mtx_unlock_spin(&sched_lock);
		PROC_UNLOCK(p);
	}

	if (td->td_pflags & TDP_UPCALLING) {
		uts_crit = 0;
		kg->kg_nextupcall = ticks+kg->kg_upquantum;
		/* 
		 * There is no more work to do and we are going to ride
		 * this thread up to userland as an upcall.
		 * Do the last parts of the setup needed for the upcall.
		 */
		CTR3(KTR_PROC, "userret: upcall thread %p (pid %d, %s)",
		    td, td->td_proc->p_pid, td->td_proc->p_comm);

		td->td_pflags &= ~TDP_UPCALLING;
		if (ku->ku_flags & KUF_DOUPCALL) {
			mtx_lock_spin(&sched_lock);
			ku->ku_flags &= ~KUF_DOUPCALL;
			mtx_unlock_spin(&sched_lock);
		}
		/*
		 * Set user context to the UTS
		 */
		if (!(ku->ku_mflags & KMF_NOUPCALL)) {
			cpu_set_upcall_kse(td, ku);
			error = suword(&ku->ku_mailbox->km_curthread, 0);
			if (error)
				goto out;
		}

		/*
		 * Unhook the list of completed threads.
		 * anything that completes after this gets to 
		 * come in next time.
		 * Put the list of completed thread mailboxes on
		 * this KSE's mailbox.
		 */
		if (!(ku->ku_mflags & KMF_NOCOMPLETED) &&
		    (error = thread_link_mboxes(kg, ku)) != 0)
			goto out;
	}
	if (!uts_crit) {
		nanotime(&ts);
		error = copyout(&ts, &ku->ku_mailbox->km_timeofday, sizeof(ts));
	}

out:
	if (error) {
		/*
		 * Things are going to be so screwed we should just kill
		 * the process.
		 * how do we do that?
		 */
		PROC_LOCK(td->td_proc);
		psignal(td->td_proc, SIGSEGV);
		PROC_UNLOCK(td->td_proc);
	} else {
		/*
		 * Optimisation:
		 * Ensure that we have a spare thread available,
		 * for when we re-enter the kernel.
		 */
		if (td->td_standin == NULL)
			thread_alloc_spare(td, NULL);
	}

	ku->ku_mflags = 0;
	/*
	 * Clear thread mailbox first, then clear system tick count.
	 * The order is important because thread_statclock() use 
	 * mailbox pointer to see if it is an userland thread or
	 * an UTS kernel thread.
	 */
	td->td_mailbox = NULL;
	td->td_usticks = 0;
	return (error);	/* go sync */
}

/*
 * Enforce single-threading.
 *
 * Returns 1 if the caller must abort (another thread is waiting to
 * exit the process or similar). Process is locked!
 * Returns 0 when you are successfully the only thread running.
 * A process has successfully single threaded in the suspend mode when
 * There are no threads in user mode. Threads in the kernel must be
 * allowed to continue until they get to the user boundary. They may even
 * copy out their return values and data before suspending. They may however be
 * accellerated in reaching the user boundary as we will wake up
 * any sleeping threads that are interruptable. (PCATCH).
 */
int
thread_single(int force_exit)
{
	struct thread *td;
	struct thread *td2;
	struct proc *p;

	td = curthread;
	p = td->td_proc;
	mtx_assert(&Giant, MA_OWNED);
	PROC_LOCK_ASSERT(p, MA_OWNED);
	KASSERT((td != NULL), ("curthread is NULL"));

	if ((p->p_flag & P_SA) == 0 && p->p_numthreads == 1)
		return (0);

	/* Is someone already single threading? */
	if (p->p_singlethread) 
		return (1);

	if (force_exit == SINGLE_EXIT) {
		p->p_flag |= P_SINGLE_EXIT;
	} else
		p->p_flag &= ~P_SINGLE_EXIT;
	p->p_flag |= P_STOPPED_SINGLE;
	mtx_lock_spin(&sched_lock);
	p->p_singlethread = td;
	while ((p->p_numthreads - p->p_suspcount) != 1) {
		FOREACH_THREAD_IN_PROC(p, td2) {
			if (td2 == td)
				continue;
			td2->td_flags |= TDF_ASTPENDING;
			if (TD_IS_INHIBITED(td2)) {
				if (force_exit == SINGLE_EXIT) {
					if (TD_IS_SUSPENDED(td2)) {
						thread_unsuspend_one(td2);
					}
					if (TD_ON_SLEEPQ(td2) &&
					    (td2->td_flags & TDF_SINTR)) {
						if (td2->td_flags & TDF_CVWAITQ)
							cv_abort(td2);
						else
							abortsleep(td2);
					}
				} else {
					if (TD_IS_SUSPENDED(td2))
						continue;
					/*
					 * maybe other inhibitted states too?
					 * XXXKSE Is it totally safe to
					 * suspend a non-interruptable thread?
					 */
					if (td2->td_inhibitors &
					    (TDI_SLEEPING | TDI_SWAPPED))
						thread_suspend_one(td2);
				}
			}
		}
		/* 
		 * Maybe we suspended some threads.. was it enough? 
		 */
		if ((p->p_numthreads - p->p_suspcount) == 1)
			break;

		/*
		 * Wake us up when everyone else has suspended.
		 * In the mean time we suspend as well.
		 */
		thread_suspend_one(td);
		DROP_GIANT();
		PROC_UNLOCK(p);
		p->p_stats->p_ru.ru_nvcsw++;
		mi_switch();
		mtx_unlock_spin(&sched_lock);
		PICKUP_GIANT();
		PROC_LOCK(p);
		mtx_lock_spin(&sched_lock);
	}
	if (force_exit == SINGLE_EXIT) { 
		if (td->td_upcall)
			upcall_remove(td);
		kse_purge(p, td);
	}
	mtx_unlock_spin(&sched_lock);
	return (0);
}

/*
 * Called in from locations that can safely check to see
 * whether we have to suspend or at least throttle for a
 * single-thread event (e.g. fork).
 *
 * Such locations include userret().
 * If the "return_instead" argument is non zero, the thread must be able to
 * accept 0 (caller may continue), or 1 (caller must abort) as a result.
 *
 * The 'return_instead' argument tells the function if it may do a
 * thread_exit() or suspend, or whether the caller must abort and back
 * out instead.
 *
 * If the thread that set the single_threading request has set the
 * P_SINGLE_EXIT bit in the process flags then this call will never return
 * if 'return_instead' is false, but will exit.
 *
 * P_SINGLE_EXIT | return_instead == 0| return_instead != 0
 *---------------+--------------------+---------------------
 *       0       | returns 0          |   returns 0 or 1
 *               | when ST ends       |   immediatly
 *---------------+--------------------+---------------------
 *       1       | thread exits       |   returns 1
 *               |                    |  immediatly
 * 0 = thread_exit() or suspension ok,
 * other = return error instead of stopping the thread.
 *
 * While a full suspension is under effect, even a single threading
 * thread would be suspended if it made this call (but it shouldn't).
 * This call should only be made from places where
 * thread_exit() would be safe as that may be the outcome unless 
 * return_instead is set.
 */
int
thread_suspend_check(int return_instead)
{
	struct thread *td;
	struct proc *p;

	td = curthread;
	p = td->td_proc;
	PROC_LOCK_ASSERT(p, MA_OWNED);
	while (P_SHOULDSTOP(p)) {
		if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
			KASSERT(p->p_singlethread != NULL,
			    ("singlethread not set"));
			/*
			 * The only suspension in action is a
			 * single-threading. Single threader need not stop.
			 * XXX Should be safe to access unlocked 
			 * as it can only be set to be true by us.
			 */
			if (p->p_singlethread == td)
				return (0);	/* Exempt from stopping. */
		} 
		if (return_instead)
			return (1);

		mtx_lock_spin(&sched_lock);
		thread_stopped(p);
		/*
		 * If the process is waiting for us to exit,
		 * this thread should just suicide.
		 * Assumes that P_SINGLE_EXIT implies P_STOPPED_SINGLE.
		 */
		if ((p->p_flag & P_SINGLE_EXIT) && (p->p_singlethread != td)) {
			while (mtx_owned(&Giant))
				mtx_unlock(&Giant);
			if (p->p_flag & P_SA)
				thread_exit();
			else
				thr_exit1();
		}

		/*
		 * When a thread suspends, it just
		 * moves to the processes's suspend queue
		 * and stays there.
		 */
		thread_suspend_one(td);
		if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
			if (p->p_numthreads == p->p_suspcount) {
				thread_unsuspend_one(p->p_singlethread);
			}
		}
		DROP_GIANT();
		PROC_UNLOCK(p);
		p->p_stats->p_ru.ru_nivcsw++;
		mi_switch();
		mtx_unlock_spin(&sched_lock);
		PICKUP_GIANT();
		PROC_LOCK(p);
	}
	return (0);
}

void
thread_suspend_one(struct thread *td)
{
	struct proc *p = td->td_proc;

	mtx_assert(&sched_lock, MA_OWNED);
	PROC_LOCK_ASSERT(p, MA_OWNED);
	KASSERT(!TD_IS_SUSPENDED(td), ("already suspended"));
	p->p_suspcount++;
	TD_SET_SUSPENDED(td);
	TAILQ_INSERT_TAIL(&p->p_suspended, td, td_runq);
	/*
	 * Hack: If we are suspending but are on the sleep queue
	 * then we are in msleep or the cv equivalent. We
	 * want to look like we have two Inhibitors.
	 * May already be set.. doesn't matter.
	 */
	if (TD_ON_SLEEPQ(td))
		TD_SET_SLEEPING(td);
}

void
thread_unsuspend_one(struct thread *td)
{
	struct proc *p = td->td_proc;

	mtx_assert(&sched_lock, MA_OWNED);
	PROC_LOCK_ASSERT(p, MA_OWNED);
	TAILQ_REMOVE(&p->p_suspended, td, td_runq);
	TD_CLR_SUSPENDED(td);
	p->p_suspcount--;
	setrunnable(td);
}

/*
 * Allow all threads blocked by single threading to continue running.
 */
void
thread_unsuspend(struct proc *p)
{
	struct thread *td;

	mtx_assert(&sched_lock, MA_OWNED);
	PROC_LOCK_ASSERT(p, MA_OWNED);
	if (!P_SHOULDSTOP(p)) {
		while (( td = TAILQ_FIRST(&p->p_suspended))) {
			thread_unsuspend_one(td);
		}
	} else if ((P_SHOULDSTOP(p) == P_STOPPED_SINGLE) &&
	    (p->p_numthreads == p->p_suspcount)) {
		/*
		 * Stopping everything also did the job for the single
		 * threading request. Now we've downgraded to single-threaded,
		 * let it continue.
		 */
		thread_unsuspend_one(p->p_singlethread);
	}
}

void
thread_single_end(void)
{
	struct thread *td;
	struct proc *p;

	td = curthread;
	p = td->td_proc;
	PROC_LOCK_ASSERT(p, MA_OWNED);
	p->p_flag &= ~P_STOPPED_SINGLE;
	mtx_lock_spin(&sched_lock);
	p->p_singlethread = NULL;
	/*
	 * If there are other threads they mey now run,
	 * unless of course there is a blanket 'stop order'
	 * on the process. The single threader must be allowed
	 * to continue however as this is a bad place to stop.
	 */
	if ((p->p_numthreads != 1) && (!P_SHOULDSTOP(p))) {
		while (( td = TAILQ_FIRST(&p->p_suspended))) {
			thread_unsuspend_one(td);
		}
	}
	mtx_unlock_spin(&sched_lock);
}


OpenPOWER on IntegriCloud