aboutsummaryrefslogtreecommitdiff
path: root/sys/kern/sched_ule.c
blob: 3bb73d64a70c6046ace5365af9b945bf5d750889 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
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
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
/*-
 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
 *
 * Copyright (c) 2002-2007, Jeffrey Roberson <jeff@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 unmodified, this list of conditions, and the following
 *    disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 *
 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 */

/*
 * This file implements the ULE scheduler.  ULE supports independent CPU
 * run queues and fine grain locking.  It has superior interactive
 * performance under load even on uni-processor systems.
 *
 * etymology:
 *   ULE is the last three letters in schedule.  It owes its name to a
 * generic user created for a scheduling system by Paul Mikesell at
 * Isilon Systems and a general lack of creativity on the part of the author.
 */

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

#include "opt_hwpmc_hooks.h"
#include "opt_sched.h"

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kdb.h>
#include <sys/kernel.h>
#include <sys/ktr.h>
#include <sys/limits.h>
#include <sys/lock.h>
#include <sys/mutex.h>
#include <sys/proc.h>
#include <sys/resource.h>
#include <sys/resourcevar.h>
#include <sys/sched.h>
#include <sys/sdt.h>
#include <sys/smp.h>
#include <sys/sx.h>
#include <sys/sysctl.h>
#include <sys/sysproto.h>
#include <sys/turnstile.h>
#include <sys/umtx.h>
#include <sys/vmmeter.h>
#include <sys/cpuset.h>
#include <sys/sbuf.h>

#ifdef HWPMC_HOOKS
#include <sys/pmckern.h>
#endif

#ifdef KDTRACE_HOOKS
#include <sys/dtrace_bsd.h>
int __read_mostly		dtrace_vtime_active;
dtrace_vtime_switch_func_t	dtrace_vtime_switch_func;
#endif

#include <machine/cpu.h>
#include <machine/smp.h>

#define	KTR_ULE	0

#define	TS_NAME_LEN (MAXCOMLEN + sizeof(" td ") + sizeof(__XSTRING(UINT_MAX)))
#define	TDQ_NAME_LEN	(sizeof("sched lock ") + sizeof(__XSTRING(MAXCPU)))
#define	TDQ_LOADNAME_LEN	(sizeof("CPU ") + sizeof(__XSTRING(MAXCPU)) - 1 + sizeof(" load"))

/*
 * Thread scheduler specific section.  All fields are protected
 * by the thread lock.
 */
struct td_sched {	
	struct runq	*ts_runq;	/* Run-queue we're queued on. */
	short		ts_flags;	/* TSF_* flags. */
	int		ts_cpu;		/* CPU that we have affinity for. */
	int		ts_rltick;	/* Real last tick, for affinity. */
	int		ts_slice;	/* Ticks of slice remaining. */
	u_int		ts_slptime;	/* Number of ticks we vol. slept */
	u_int		ts_runtime;	/* Number of ticks we were running */
	int		ts_ltick;	/* Last tick that we were running on */
	int		ts_ftick;	/* First tick that we were running on */
	int		ts_ticks;	/* Tick count */
#ifdef KTR
	char		ts_name[TS_NAME_LEN];
#endif
};
/* flags kept in ts_flags */
#define	TSF_BOUND	0x0001		/* Thread can not migrate. */
#define	TSF_XFERABLE	0x0002		/* Thread was added as transferable. */

#define	THREAD_CAN_MIGRATE(td)	((td)->td_pinned == 0)
#define	THREAD_CAN_SCHED(td, cpu)	\
    CPU_ISSET((cpu), &(td)->td_cpuset->cs_mask)

_Static_assert(sizeof(struct thread) + sizeof(struct td_sched) <=
    sizeof(struct thread0_storage),
    "increase struct thread0_storage.t0st_sched size");

/*
 * Priority ranges used for interactive and non-interactive timeshare
 * threads.  The timeshare priorities are split up into four ranges.
 * The first range handles interactive threads.  The last three ranges
 * (NHALF, x, and NHALF) handle non-interactive threads with the outer
 * ranges supporting nice values.
 */
#define	PRI_TIMESHARE_RANGE	(PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE + 1)
#define	PRI_INTERACT_RANGE	((PRI_TIMESHARE_RANGE - SCHED_PRI_NRESV) / 2)
#define	PRI_BATCH_RANGE		(PRI_TIMESHARE_RANGE - PRI_INTERACT_RANGE)

#define	PRI_MIN_INTERACT	PRI_MIN_TIMESHARE
#define	PRI_MAX_INTERACT	(PRI_MIN_TIMESHARE + PRI_INTERACT_RANGE - 1)
#define	PRI_MIN_BATCH		(PRI_MIN_TIMESHARE + PRI_INTERACT_RANGE)
#define	PRI_MAX_BATCH		PRI_MAX_TIMESHARE

/*
 * Cpu percentage computation macros and defines.
 *
 * SCHED_TICK_SECS:	Number of seconds to average the cpu usage across.
 * SCHED_TICK_TARG:	Number of hz ticks to average the cpu usage across.
 * SCHED_TICK_MAX:	Maximum number of ticks before scaling back.
 * SCHED_TICK_SHIFT:	Shift factor to avoid rounding away results.
 * SCHED_TICK_HZ:	Compute the number of hz ticks for a given ticks count.
 * SCHED_TICK_TOTAL:	Gives the amount of time we've been recording ticks.
 */
#define	SCHED_TICK_SECS		10
#define	SCHED_TICK_TARG		(hz * SCHED_TICK_SECS)
#define	SCHED_TICK_MAX		(SCHED_TICK_TARG + hz)
#define	SCHED_TICK_SHIFT	10
#define	SCHED_TICK_HZ(ts)	((ts)->ts_ticks >> SCHED_TICK_SHIFT)
#define	SCHED_TICK_TOTAL(ts)	(max((ts)->ts_ltick - (ts)->ts_ftick, hz))

/*
 * These macros determine priorities for non-interactive threads.  They are
 * assigned a priority based on their recent cpu utilization as expressed
 * by the ratio of ticks to the tick total.  NHALF priorities at the start
 * and end of the MIN to MAX timeshare range are only reachable with negative
 * or positive nice respectively.
 *
 * PRI_RANGE:	Priority range for utilization dependent priorities.
 * PRI_NRESV:	Number of nice values.
 * PRI_TICKS:	Compute a priority in PRI_RANGE from the ticks count and total.
 * PRI_NICE:	Determines the part of the priority inherited from nice.
 */
#define	SCHED_PRI_NRESV		(PRIO_MAX - PRIO_MIN)
#define	SCHED_PRI_NHALF		(SCHED_PRI_NRESV / 2)
#define	SCHED_PRI_MIN		(PRI_MIN_BATCH + SCHED_PRI_NHALF)
#define	SCHED_PRI_MAX		(PRI_MAX_BATCH - SCHED_PRI_NHALF)
#define	SCHED_PRI_RANGE		(SCHED_PRI_MAX - SCHED_PRI_MIN + 1)
#define	SCHED_PRI_TICKS(ts)						\
    (SCHED_TICK_HZ((ts)) /						\
    (roundup(SCHED_TICK_TOTAL((ts)), SCHED_PRI_RANGE) / SCHED_PRI_RANGE))
#define	SCHED_PRI_NICE(nice)	(nice)

/*
 * These determine the interactivity of a process.  Interactivity differs from
 * cpu utilization in that it expresses the voluntary time slept vs time ran
 * while cpu utilization includes all time not running.  This more accurately
 * models the intent of the thread.
 *
 * SLP_RUN_MAX:	Maximum amount of sleep time + run time we'll accumulate
 *		before throttling back.
 * SLP_RUN_FORK:	Maximum slp+run time to inherit at fork time.
 * INTERACT_MAX:	Maximum interactivity value.  Smaller is better.
 * INTERACT_THRESH:	Threshold for placement on the current runq.
 */
#define	SCHED_SLP_RUN_MAX	((hz * 5) << SCHED_TICK_SHIFT)
#define	SCHED_SLP_RUN_FORK	((hz / 2) << SCHED_TICK_SHIFT)
#define	SCHED_INTERACT_MAX	(100)
#define	SCHED_INTERACT_HALF	(SCHED_INTERACT_MAX / 2)
#define	SCHED_INTERACT_THRESH	(30)

/*
 * These parameters determine the slice behavior for batch work.
 */
#define	SCHED_SLICE_DEFAULT_DIVISOR	10	/* ~94 ms, 12 stathz ticks. */
#define	SCHED_SLICE_MIN_DIVISOR		6	/* DEFAULT/MIN = ~16 ms. */

/* Flags kept in td_flags. */
#define	TDF_SLICEEND	TDF_SCHED2	/* Thread time slice is over. */

/*
 * tickincr:		Converts a stathz tick into a hz domain scaled by
 *			the shift factor.  Without the shift the error rate
 *			due to rounding would be unacceptably high.
 * realstathz:		stathz is sometimes 0 and run off of hz.
 * sched_slice:		Runtime of each thread before rescheduling.
 * preempt_thresh:	Priority threshold for preemption and remote IPIs.
 */
static int __read_mostly sched_interact = SCHED_INTERACT_THRESH;
static int __read_mostly tickincr = 8 << SCHED_TICK_SHIFT;
static int __read_mostly realstathz = 127;	/* reset during boot. */
static int __read_mostly sched_slice = 10;	/* reset during boot. */
static int __read_mostly sched_slice_min = 1;	/* reset during boot. */
#ifdef PREEMPTION
#ifdef FULL_PREEMPTION
static int __read_mostly preempt_thresh = PRI_MAX_IDLE;
#else
static int __read_mostly preempt_thresh = PRI_MIN_KERN;
#endif
#else 
static int __read_mostly preempt_thresh = 0;
#endif
static int __read_mostly static_boost = PRI_MIN_BATCH;
static int __read_mostly sched_idlespins = 10000;
static int __read_mostly sched_idlespinthresh = -1;

/*
 * tdq - per processor runqs and statistics.  All fields are protected by the
 * tdq_lock.  The load and lowpri may be accessed without to avoid excess
 * locking in sched_pickcpu();
 */
struct tdq {
	/* 
	 * Ordered to improve efficiency of cpu_search() and switch().
	 * tdq_lock is padded to avoid false sharing with tdq_load and
	 * tdq_cpu_idle.
	 */
	struct mtx_padalign tdq_lock;		/* run queue lock. */
	struct cpu_group *tdq_cg;		/* Pointer to cpu topology. */
	volatile int	tdq_load;		/* Aggregate load. */
	volatile int	tdq_cpu_idle;		/* cpu_idle() is active. */
	int		tdq_sysload;		/* For loadavg, !ITHD load. */
	volatile int	tdq_transferable;	/* Transferable thread count. */
	volatile short	tdq_switchcnt;		/* Switches this tick. */
	volatile short	tdq_oldswitchcnt;	/* Switches last tick. */
	u_char		tdq_lowpri;		/* Lowest priority thread. */
	u_char		tdq_owepreempt;		/* Remote preemption pending. */
	u_char		tdq_idx;		/* Current insert index. */
	u_char		tdq_ridx;		/* Current removal index. */
	int		tdq_id;			/* cpuid. */
	struct runq	tdq_realtime;		/* real-time run queue. */
	struct runq	tdq_timeshare;		/* timeshare run queue. */
	struct runq	tdq_idle;		/* Queue of IDLE threads. */
	char		tdq_name[TDQ_NAME_LEN];
#ifdef KTR
	char		tdq_loadname[TDQ_LOADNAME_LEN];
#endif
} __aligned(64);

/* Idle thread states and config. */
#define	TDQ_RUNNING	1
#define	TDQ_IDLE	2

#ifdef SMP
struct cpu_group __read_mostly *cpu_top;		/* CPU topology */

#define	SCHED_AFFINITY_DEFAULT	(max(1, hz / 1000))
#define	SCHED_AFFINITY(ts, t)	((ts)->ts_rltick > ticks - ((t) * affinity))

/*
 * Run-time tunables.
 */
static int rebalance = 1;
static int balance_interval = 128;	/* Default set in sched_initticks(). */
static int __read_mostly affinity;
static int __read_mostly steal_idle = 1;
static int __read_mostly steal_thresh = 2;
static int __read_mostly always_steal = 0;
static int __read_mostly trysteal_limit = 2;

/*
 * One thread queue per processor.
 */
static struct tdq __read_mostly *balance_tdq;
static int balance_ticks;
DPCPU_DEFINE_STATIC(struct tdq, tdq);
DPCPU_DEFINE_STATIC(uint32_t, randomval);

#define	TDQ_SELF()	((struct tdq *)PCPU_GET(sched))
#define	TDQ_CPU(x)	(DPCPU_ID_PTR((x), tdq))
#define	TDQ_ID(x)	((x)->tdq_id)
#else	/* !SMP */
static struct tdq	tdq_cpu;

#define	TDQ_ID(x)	(0)
#define	TDQ_SELF()	(&tdq_cpu)
#define	TDQ_CPU(x)	(&tdq_cpu)
#endif

#define	TDQ_LOCK_ASSERT(t, type)	mtx_assert(TDQ_LOCKPTR((t)), (type))
#define	TDQ_LOCK(t)		mtx_lock_spin(TDQ_LOCKPTR((t)))
#define	TDQ_LOCK_FLAGS(t, f)	mtx_lock_spin_flags(TDQ_LOCKPTR((t)), (f))
#define	TDQ_UNLOCK(t)		mtx_unlock_spin(TDQ_LOCKPTR((t)))
#define	TDQ_LOCKPTR(t)		((struct mtx *)(&(t)->tdq_lock))

static void sched_priority(struct thread *);
static void sched_thread_priority(struct thread *, u_char);
static int sched_interact_score(struct thread *);
static void sched_interact_update(struct thread *);
static void sched_interact_fork(struct thread *);
static void sched_pctcpu_update(struct td_sched *, int);

/* Operations on per processor queues */
static struct thread *tdq_choose(struct tdq *);
static void tdq_setup(struct tdq *, int i);
static void tdq_load_add(struct tdq *, struct thread *);
static void tdq_load_rem(struct tdq *, struct thread *);
static __inline void tdq_runq_add(struct tdq *, struct thread *, int);
static __inline void tdq_runq_rem(struct tdq *, struct thread *);
static inline int sched_shouldpreempt(int, int, int);
void tdq_print(int cpu);
static void runq_print(struct runq *rq);
static void tdq_add(struct tdq *, struct thread *, int);
#ifdef SMP
static struct thread *tdq_move(struct tdq *, struct tdq *);
static int tdq_idled(struct tdq *);
static void tdq_notify(struct tdq *, struct thread *);
static struct thread *tdq_steal(struct tdq *, int);
static struct thread *runq_steal(struct runq *, int);
static int sched_pickcpu(struct thread *, int);
static void sched_balance(void);
static int sched_balance_pair(struct tdq *, struct tdq *);
static inline struct tdq *sched_setcpu(struct thread *, int, int);
static inline void thread_unblock_switch(struct thread *, struct mtx *);
static int sysctl_kern_sched_topology_spec(SYSCTL_HANDLER_ARGS);
static int sysctl_kern_sched_topology_spec_internal(struct sbuf *sb, 
    struct cpu_group *cg, int indent);
#endif

static void sched_setup(void *dummy);
SYSINIT(sched_setup, SI_SUB_RUN_QUEUE, SI_ORDER_FIRST, sched_setup, NULL);

static void sched_initticks(void *dummy);
SYSINIT(sched_initticks, SI_SUB_CLOCKS, SI_ORDER_THIRD, sched_initticks,
    NULL);

SDT_PROVIDER_DEFINE(sched);

SDT_PROBE_DEFINE3(sched, , , change__pri, "struct thread *", 
    "struct proc *", "uint8_t");
SDT_PROBE_DEFINE3(sched, , , dequeue, "struct thread *", 
    "struct proc *", "void *");
SDT_PROBE_DEFINE4(sched, , , enqueue, "struct thread *", 
    "struct proc *", "void *", "int");
SDT_PROBE_DEFINE4(sched, , , lend__pri, "struct thread *", 
    "struct proc *", "uint8_t", "struct thread *");
SDT_PROBE_DEFINE2(sched, , , load__change, "int", "int");
SDT_PROBE_DEFINE2(sched, , , off__cpu, "struct thread *", 
    "struct proc *");
SDT_PROBE_DEFINE(sched, , , on__cpu);
SDT_PROBE_DEFINE(sched, , , remain__cpu);
SDT_PROBE_DEFINE2(sched, , , surrender, "struct thread *", 
    "struct proc *");

/*
 * Print the threads waiting on a run-queue.
 */
static void
runq_print(struct runq *rq)
{
	struct rqhead *rqh;
	struct thread *td;
	int pri;
	int j;
	int i;

	for (i = 0; i < RQB_LEN; i++) {
		printf("\t\trunq bits %d 0x%zx\n",
		    i, rq->rq_status.rqb_bits[i]);
		for (j = 0; j < RQB_BPW; j++)
			if (rq->rq_status.rqb_bits[i] & (1ul << j)) {
				pri = j + (i << RQB_L2BPW);
				rqh = &rq->rq_queues[pri];
				TAILQ_FOREACH(td, rqh, td_runq) {
					printf("\t\t\ttd %p(%s) priority %d rqindex %d pri %d\n",
					    td, td->td_name, td->td_priority,
					    td->td_rqindex, pri);
				}
			}
	}
}

/*
 * Print the status of a per-cpu thread queue.  Should be a ddb show cmd.
 */
void
tdq_print(int cpu)
{
	struct tdq *tdq;

	tdq = TDQ_CPU(cpu);

	printf("tdq %d:\n", TDQ_ID(tdq));
	printf("\tlock            %p\n", TDQ_LOCKPTR(tdq));
	printf("\tLock name:      %s\n", tdq->tdq_name);
	printf("\tload:           %d\n", tdq->tdq_load);
	printf("\tswitch cnt:     %d\n", tdq->tdq_switchcnt);
	printf("\told switch cnt: %d\n", tdq->tdq_oldswitchcnt);
	printf("\ttimeshare idx:  %d\n", tdq->tdq_idx);
	printf("\ttimeshare ridx: %d\n", tdq->tdq_ridx);
	printf("\tload transferable: %d\n", tdq->tdq_transferable);
	printf("\tlowest priority:   %d\n", tdq->tdq_lowpri);
	printf("\trealtime runq:\n");
	runq_print(&tdq->tdq_realtime);
	printf("\ttimeshare runq:\n");
	runq_print(&tdq->tdq_timeshare);
	printf("\tidle runq:\n");
	runq_print(&tdq->tdq_idle);
}

static inline int
sched_shouldpreempt(int pri, int cpri, int remote)
{
	/*
	 * If the new priority is not better than the current priority there is
	 * nothing to do.
	 */
	if (pri >= cpri)
		return (0);
	/*
	 * Always preempt idle.
	 */
	if (cpri >= PRI_MIN_IDLE)
		return (1);
	/*
	 * If preemption is disabled don't preempt others.
	 */
	if (preempt_thresh == 0)
		return (0);
	/*
	 * Preempt if we exceed the threshold.
	 */
	if (pri <= preempt_thresh)
		return (1);
	/*
	 * If we're interactive or better and there is non-interactive
	 * or worse running preempt only remote processors.
	 */
	if (remote && pri <= PRI_MAX_INTERACT && cpri > PRI_MAX_INTERACT)
		return (1);
	return (0);
}

/*
 * Add a thread to the actual run-queue.  Keeps transferable counts up to
 * date with what is actually on the run-queue.  Selects the correct
 * queue position for timeshare threads.
 */
static __inline void
tdq_runq_add(struct tdq *tdq, struct thread *td, int flags)
{
	struct td_sched *ts;
	u_char pri;

	TDQ_LOCK_ASSERT(tdq, MA_OWNED);
	THREAD_LOCK_BLOCKED_ASSERT(td, MA_OWNED);

	pri = td->td_priority;
	ts = td_get_sched(td);
	TD_SET_RUNQ(td);
	if (THREAD_CAN_MIGRATE(td)) {
		tdq->tdq_transferable++;
		ts->ts_flags |= TSF_XFERABLE;
	}
	if (pri < PRI_MIN_BATCH) {
		ts->ts_runq = &tdq->tdq_realtime;
	} else if (pri <= PRI_MAX_BATCH) {
		ts->ts_runq = &tdq->tdq_timeshare;
		KASSERT(pri <= PRI_MAX_BATCH && pri >= PRI_MIN_BATCH,
			("Invalid priority %d on timeshare runq", pri));
		/*
		 * This queue contains only priorities between MIN and MAX
		 * realtime.  Use the whole queue to represent these values.
		 */
		if ((flags & (SRQ_BORROWING|SRQ_PREEMPTED)) == 0) {
			pri = RQ_NQS * (pri - PRI_MIN_BATCH) / PRI_BATCH_RANGE;
			pri = (pri + tdq->tdq_idx) % RQ_NQS;
			/*
			 * This effectively shortens the queue by one so we
			 * can have a one slot difference between idx and
			 * ridx while we wait for threads to drain.
			 */
			if (tdq->tdq_ridx != tdq->tdq_idx &&
			    pri == tdq->tdq_ridx)
				pri = (unsigned char)(pri - 1) % RQ_NQS;
		} else
			pri = tdq->tdq_ridx;
		runq_add_pri(ts->ts_runq, td, pri, flags);
		return;
	} else
		ts->ts_runq = &tdq->tdq_idle;
	runq_add(ts->ts_runq, td, flags);
}

/* 
 * Remove a thread from a run-queue.  This typically happens when a thread
 * is selected to run.  Running threads are not on the queue and the
 * transferable count does not reflect them.
 */
static __inline void
tdq_runq_rem(struct tdq *tdq, struct thread *td)
{
	struct td_sched *ts;

	ts = td_get_sched(td);
	TDQ_LOCK_ASSERT(tdq, MA_OWNED);
	THREAD_LOCK_BLOCKED_ASSERT(td, MA_OWNED);
	KASSERT(ts->ts_runq != NULL,
	    ("tdq_runq_remove: thread %p null ts_runq", td));
	if (ts->ts_flags & TSF_XFERABLE) {
		tdq->tdq_transferable--;
		ts->ts_flags &= ~TSF_XFERABLE;
	}
	if (ts->ts_runq == &tdq->tdq_timeshare) {
		if (tdq->tdq_idx != tdq->tdq_ridx)
			runq_remove_idx(ts->ts_runq, td, &tdq->tdq_ridx);
		else
			runq_remove_idx(ts->ts_runq, td, NULL);
	} else
		runq_remove(ts->ts_runq, td);
}

/*
 * Load is maintained for all threads RUNNING and ON_RUNQ.  Add the load
 * for this thread to the referenced thread queue.
 */
static void
tdq_load_add(struct tdq *tdq, struct thread *td)
{

	TDQ_LOCK_ASSERT(tdq, MA_OWNED);
	THREAD_LOCK_BLOCKED_ASSERT(td, MA_OWNED);

	tdq->tdq_load++;
	if ((td->td_flags & TDF_NOLOAD) == 0)
		tdq->tdq_sysload++;
	KTR_COUNTER0(KTR_SCHED, "load", tdq->tdq_loadname, tdq->tdq_load);
	SDT_PROBE2(sched, , , load__change, (int)TDQ_ID(tdq), tdq->tdq_load);
}

/*
 * Remove the load from a thread that is transitioning to a sleep state or
 * exiting.
 */
static void
tdq_load_rem(struct tdq *tdq, struct thread *td)
{

	TDQ_LOCK_ASSERT(tdq, MA_OWNED);
	THREAD_LOCK_BLOCKED_ASSERT(td, MA_OWNED);
	KASSERT(tdq->tdq_load != 0,
	    ("tdq_load_rem: Removing with 0 load on queue %d", TDQ_ID(tdq)));

	tdq->tdq_load--;
	if ((td->td_flags & TDF_NOLOAD) == 0)
		tdq->tdq_sysload--;
	KTR_COUNTER0(KTR_SCHED, "load", tdq->tdq_loadname, tdq->tdq_load);
	SDT_PROBE2(sched, , , load__change, (int)TDQ_ID(tdq), tdq->tdq_load);
}

/*
 * Bound timeshare latency by decreasing slice size as load increases.  We
 * consider the maximum latency as the sum of the threads waiting to run
 * aside from curthread and target no more than sched_slice latency but
 * no less than sched_slice_min runtime.
 */
static inline int
tdq_slice(struct tdq *tdq)
{
	int load;

	/*
	 * It is safe to use sys_load here because this is called from
	 * contexts where timeshare threads are running and so there
	 * cannot be higher priority load in the system.
	 */
	load = tdq->tdq_sysload - 1;
	if (load >= SCHED_SLICE_MIN_DIVISOR)
		return (sched_slice_min);
	if (load <= 1)
		return (sched_slice);
	return (sched_slice / load);
}

/*
 * Set lowpri to its exact value by searching the run-queue and
 * evaluating curthread.  curthread may be passed as an optimization.
 */
static void
tdq_setlowpri(struct tdq *tdq, struct thread *ctd)
{
	struct thread *td;

	TDQ_LOCK_ASSERT(tdq, MA_OWNED);
	if (ctd == NULL)
		ctd = pcpu_find(TDQ_ID(tdq))->pc_curthread;
	td = tdq_choose(tdq);
	if (td == NULL || td->td_priority > ctd->td_priority)
		tdq->tdq_lowpri = ctd->td_priority;
	else
		tdq->tdq_lowpri = td->td_priority;
}

#ifdef SMP
/*
 * We need some randomness. Implement a classic Linear Congruential
 * Generator X_{n+1}=(aX_n+c) mod m. These values are optimized for
 * m = 2^32, a = 69069 and c = 5. We only return the upper 16 bits
 * of the random state (in the low bits of our answer) to keep
 * the maximum randomness.
 */
static uint32_t
sched_random(void)
{
	uint32_t *rndptr;

	rndptr = DPCPU_PTR(randomval);
	*rndptr = *rndptr * 69069 + 5;

	return (*rndptr >> 16);
}

struct cpu_search {
	cpuset_t *cs_mask;
	u_int	cs_prefer;
	int	cs_pri;		/* Min priority for low. */
	int	cs_limit;	/* Max load for low, min load for high. */
};

struct cpu_search_res {
	int	cs_cpu;
	int	cs_load;
};

/*
 * Search the tree of cpu_groups for the lowest or highest loaded CPU.
 * These routines actually compare the load on all paths through the tree
 * and find the least loaded cpu on the least loaded path, which may differ
 * from the least loaded cpu in the system.  This balances work among caches
 * and buses.
 */
static int
cpu_search_lowest(const struct cpu_group *cg, const struct cpu_search *s,
    struct cpu_search_res *r)
{
	struct cpu_search_res lr;
	struct tdq *tdq;
	int c, bload, l, load, total;

	total = 0;
	bload = INT_MAX;
	r->cs_cpu = -1;

	/* Loop through children CPU groups if there are any. */
	if (cg->cg_children > 0) {
		for (c = cg->cg_children - 1; c >= 0; c--) {
			load = cpu_search_lowest(&cg->cg_child[c], s, &lr);
			total += load;
			if (lr.cs_cpu >= 0 && (load < bload ||
			    (load == bload && lr.cs_load < r->cs_load))) {
				bload = load;
				r->cs_cpu = lr.cs_cpu;
				r->cs_load = lr.cs_load;
			}
		}
		return (total);
	}

	/* Loop through children CPUs otherwise. */
	for (c = cg->cg_last; c >= cg->cg_first; c--) {
		if (!CPU_ISSET(c, &cg->cg_mask))
			continue;
		tdq = TDQ_CPU(c);
		l = tdq->tdq_load;
		load = l * 256;
		if (c == s->cs_prefer)
			load -= 128;
		total += load;
		if (l > s->cs_limit || tdq->tdq_lowpri <= s->cs_pri ||
		    !CPU_ISSET(c, s->cs_mask))
			continue;
		load -= sched_random() % 128;
		if (load < bload) {
			bload = load;
			r->cs_cpu = c;
		}
	}
	r->cs_load = bload;
	return (total);
}

static int
cpu_search_highest(const struct cpu_group *cg, const struct cpu_search *s,
    struct cpu_search_res *r)
{
	struct cpu_search_res lr;
	struct tdq *tdq;
	int c, bload, l, load, total;

	total = 0;
	bload = INT_MIN;
	r->cs_cpu = -1;

	/* Loop through children CPU groups if there are any. */
	if (cg->cg_children > 0) {
		for (c = cg->cg_children - 1; c >= 0; c--) {
			load = cpu_search_highest(&cg->cg_child[c], s, &lr);
			total += load;
			if (lr.cs_cpu >= 0 && (load > bload ||
			    (load == bload && lr.cs_load > r->cs_load))) {
				bload = load;
				r->cs_cpu = lr.cs_cpu;
				r->cs_load = lr.cs_load;
			}
		}
		return (total);
	}

	/* Loop through children CPUs otherwise. */
	for (c = cg->cg_last; c >= cg->cg_first; c--) {
		if (!CPU_ISSET(c, &cg->cg_mask))
			continue;
		tdq = TDQ_CPU(c);
		l = tdq->tdq_load;
		load = l * 256;
		total += load;
		if (l < s->cs_limit || !tdq->tdq_transferable ||
		    !CPU_ISSET(c, s->cs_mask))
			continue;
		load -= sched_random() % 128;
		if (load > bload) {
			bload = load;
			r->cs_cpu = c;
		}
	}
	r->cs_load = bload;
	return (total);
}

/*
 * Find the cpu with the least load via the least loaded path that has a
 * lowpri greater than pri  pri.  A pri of -1 indicates any priority is
 * acceptable.
 */
static inline int
sched_lowest(const struct cpu_group *cg, cpuset_t *mask, int pri, int maxload,
    int prefer)
{
	struct cpu_search s;
	struct cpu_search_res r;

	s.cs_prefer = prefer;
	s.cs_mask = mask;
	s.cs_pri = pri;
	s.cs_limit = maxload;
	cpu_search_lowest(cg, &s, &r);
	return (r.cs_cpu);
}

/*
 * Find the cpu with the highest load via the highest loaded path.
 */
static inline int
sched_highest(const struct cpu_group *cg, cpuset_t *mask, int minload)
{
	struct cpu_search s;
	struct cpu_search_res r;

	s.cs_mask = mask;
	s.cs_limit = minload;
	cpu_search_highest(cg, &s, &r);
	return (r.cs_cpu);
}

static void
sched_balance_group(struct cpu_group *cg)
{
	struct tdq *tdq;
	cpuset_t hmask, lmask;
	int high, low, anylow;

	CPU_FILL(&hmask);
	for (;;) {
		high = sched_highest(cg, &hmask, 2);
		/* Stop if there is no more CPU with transferrable threads. */
		if (high == -1)
			break;
		CPU_CLR(high, &hmask);
		CPU_COPY(&hmask, &lmask);
		/* Stop if there is no more CPU left for low. */
		if (CPU_EMPTY(&lmask))
			break;
		anylow = 1;
		tdq = TDQ_CPU(high);
nextlow:
		low = sched_lowest(cg, &lmask, -1, tdq->tdq_load - 1, high);
		/* Stop if we looked well and found no less loaded CPU. */
		if (anylow && low == -1)
			break;
		/* Go to next high if we found no less loaded CPU. */
		if (low == -1)
			continue;
		/* Transfer thread from high to low. */
		if (sched_balance_pair(tdq, TDQ_CPU(low))) {
			/* CPU that got thread can no longer be a donor. */
			CPU_CLR(low, &hmask);
		} else {
			/*
			 * If failed, then there is no threads on high
			 * that can run on this low. Drop low from low
			 * mask and look for different one.
			 */
			CPU_CLR(low, &lmask);
			anylow = 0;
			goto nextlow;
		}
	}
}

static void
sched_balance(void)
{
	struct tdq *tdq;

	balance_ticks = max(balance_interval / 2, 1) +
	    (sched_random() % balance_interval);
	tdq = TDQ_SELF();
	TDQ_UNLOCK(tdq);
	sched_balance_group(cpu_top);
	TDQ_LOCK(tdq);
}

/*
 * Lock two thread queues using their address to maintain lock order.
 */
static void
tdq_lock_pair(struct tdq *one, struct tdq *two)
{
	if (one < two) {
		TDQ_LOCK(one);
		TDQ_LOCK_FLAGS(two, MTX_DUPOK);
	} else {
		TDQ_LOCK(two);
		TDQ_LOCK_FLAGS(one, MTX_DUPOK);
	}
}

/*
 * Unlock two thread queues.  Order is not important here.
 */
static void
tdq_unlock_pair(struct tdq *one, struct tdq *two)
{
	TDQ_UNLOCK(one);
	TDQ_UNLOCK(two);
}

/*
 * Transfer load between two imbalanced thread queues.
 */
static int
sched_balance_pair(struct tdq *high, struct tdq *low)
{
	struct thread *td;
	int cpu;

	tdq_lock_pair(high, low);
	td = NULL;
	/*
	 * Transfer a thread from high to low.
	 */
	if (high->tdq_transferable != 0 && high->tdq_load > low->tdq_load &&
	    (td = tdq_move(high, low)) != NULL) {
		/*
		 * In case the target isn't the current cpu notify it of the
		 * new load, possibly sending an IPI to force it to reschedule.
		 */
		cpu = TDQ_ID(low);
		if (cpu != PCPU_GET(cpuid))
			tdq_notify(low, td);
	}
	tdq_unlock_pair(high, low);
	return (td != NULL);
}

/*
 * Move a thread from one thread queue to another.
 */
static struct thread *
tdq_move(struct tdq *from, struct tdq *to)
{
	struct thread *td;
	struct tdq *tdq;
	int cpu;

	TDQ_LOCK_ASSERT(from, MA_OWNED);
	TDQ_LOCK_ASSERT(to, MA_OWNED);

	tdq = from;
	cpu = TDQ_ID(to);
	td = tdq_steal(tdq, cpu);
	if (td == NULL)
		return (NULL);

	/*
	 * Although the run queue is locked the thread may be
	 * blocked.  We can not set the lock until it is unblocked.
	 */
	thread_lock_block_wait(td);
	sched_rem(td);
	THREAD_LOCKPTR_ASSERT(td, TDQ_LOCKPTR(from));
	td->td_lock = TDQ_LOCKPTR(to);
	td_get_sched(td)->ts_cpu = cpu;
	tdq_add(to, td, SRQ_YIELDING);

	return (td);
}

/*
 * This tdq has idled.  Try to steal a thread from another cpu and switch
 * to it.
 */
static int
tdq_idled(struct tdq *tdq)
{
	struct cpu_group *cg;
	struct tdq *steal;
	cpuset_t mask;
	int cpu, switchcnt;

	if (smp_started == 0 || steal_idle == 0 || tdq->tdq_cg == NULL)
		return (1);
	CPU_FILL(&mask);
	CPU_CLR(PCPU_GET(cpuid), &mask);
    restart:
	switchcnt = tdq->tdq_switchcnt + tdq->tdq_oldswitchcnt;
	for (cg = tdq->tdq_cg; ; ) {
		cpu = sched_highest(cg, &mask, steal_thresh);
		/*
		 * We were assigned a thread but not preempted.  Returning
		 * 0 here will cause our caller to switch to it.
		 */
		if (tdq->tdq_load)
			return (0);
		if (cpu == -1) {
			cg = cg->cg_parent;
			if (cg == NULL)
				return (1);
			continue;
		}
		steal = TDQ_CPU(cpu);
		/*
		 * The data returned by sched_highest() is stale and
		 * the chosen CPU no longer has an eligible thread.
		 *
		 * Testing this ahead of tdq_lock_pair() only catches
		 * this situation about 20% of the time on an 8 core
		 * 16 thread Ryzen 7, but it still helps performance.
		 */
		if (steal->tdq_load < steal_thresh ||
		    steal->tdq_transferable == 0)
			goto restart;
		tdq_lock_pair(tdq, steal);
		/*
		 * We were assigned a thread while waiting for the locks.
		 * Switch to it now instead of stealing a thread.
		 */
		if (tdq->tdq_load)
			break;
		/*
		 * The data returned by sched_highest() is stale and
		 * the chosen CPU no longer has an eligible thread, or
		 * we were preempted and the CPU loading info may be out
		 * of date.  The latter is rare.  In either case restart
		 * the search.
		 */
		if (steal->tdq_load < steal_thresh ||
		    steal->tdq_transferable == 0 ||
		    switchcnt != tdq->tdq_switchcnt + tdq->tdq_oldswitchcnt) {
			tdq_unlock_pair(tdq, steal);
			goto restart;
		}
		/*
		 * Steal the thread and switch to it.
		 */
		if (tdq_move(steal, tdq) != NULL)
			break;
		/*
		 * We failed to acquire a thread even though it looked
		 * like one was available.  This could be due to affinity
		 * restrictions or for other reasons.  Loop again after
		 * removing this CPU from the set.  The restart logic
		 * above does not restore this CPU to the set due to the
		 * likelyhood of failing here again.
		 */
		CPU_CLR(cpu, &mask);
		tdq_unlock_pair(tdq, steal);
	}
	TDQ_UNLOCK(steal);
	mi_switch(SW_VOL | SWT_IDLE);
	return (0);
}

/*
 * Notify a remote cpu of new work.  Sends an IPI if criteria are met.
 */
static void
tdq_notify(struct tdq *tdq, struct thread *td)
{
	struct thread *ctd;
	int pri;
	int cpu;

	if (tdq->tdq_owepreempt)
		return;
	cpu = td_get_sched(td)->ts_cpu;
	pri = td->td_priority;
	ctd = pcpu_find(cpu)->pc_curthread;
	if (!sched_shouldpreempt(pri, ctd->td_priority, 1))
		return;

	/*
	 * Make sure that our caller's earlier update to tdq_load is
	 * globally visible before we read tdq_cpu_idle.  Idle thread
	 * accesses both of them without locks, and the order is important.
	 */
	atomic_thread_fence_seq_cst();

	if (TD_IS_IDLETHREAD(ctd)) {
		/*
		 * If the MD code has an idle wakeup routine try that before
		 * falling back to IPI.
		 */
		if (!tdq->tdq_cpu_idle || cpu_idle_wakeup(cpu))
			return;
	}

	/*
	 * The run queues have been updated, so any switch on the remote CPU
	 * will satisfy the preemption request.
	 */
	tdq->tdq_owepreempt = 1;
	ipi_cpu(cpu, IPI_PREEMPT);
}

/*
 * Steals load from a timeshare queue.  Honors the rotating queue head
 * index.
 */
static struct thread *
runq_steal_from(struct runq *rq, int cpu, u_char start)
{
	struct rqbits *rqb;
	struct rqhead *rqh;
	struct thread *td, *first;
	int bit;
	int i;

	rqb = &rq->rq_status;
	bit = start & (RQB_BPW -1);
	first = NULL;
again:
	for (i = RQB_WORD(start); i < RQB_LEN; bit = 0, i++) {
		if (rqb->rqb_bits[i] == 0)
			continue;
		if (bit == 0)
			bit = RQB_FFS(rqb->rqb_bits[i]);
		for (; bit < RQB_BPW; bit++) {
			if ((rqb->rqb_bits[i] & (1ul << bit)) == 0)
				continue;
			rqh = &rq->rq_queues[bit + (i << RQB_L2BPW)];
			TAILQ_FOREACH(td, rqh, td_runq) {
				if (first && THREAD_CAN_MIGRATE(td) &&
				    THREAD_CAN_SCHED(td, cpu))
					return (td);
				first = td;
			}
		}
	}
	if (start != 0) {
		start = 0;
		goto again;
	}

	if (first && THREAD_CAN_MIGRATE(first) &&
	    THREAD_CAN_SCHED(first, cpu))
		return (first);
	return (NULL);
}

/*
 * Steals load from a standard linear queue.
 */
static struct thread *
runq_steal(struct runq *rq, int cpu)
{
	struct rqhead *rqh;
	struct rqbits *rqb;
	struct thread *td;
	int word;
	int bit;

	rqb = &rq->rq_status;
	for (word = 0; word < RQB_LEN; word++) {
		if (rqb->rqb_bits[word] == 0)
			continue;
		for (bit = 0; bit < RQB_BPW; bit++) {
			if ((rqb->rqb_bits[word] & (1ul << bit)) == 0)
				continue;
			rqh = &rq->rq_queues[bit + (word << RQB_L2BPW)];
			TAILQ_FOREACH(td, rqh, td_runq)
				if (THREAD_CAN_MIGRATE(td) &&
				    THREAD_CAN_SCHED(td, cpu))
					return (td);
		}
	}
	return (NULL);
}

/*
 * Attempt to steal a thread in priority order from a thread queue.
 */
static struct thread *
tdq_steal(struct tdq *tdq, int cpu)
{
	struct thread *td;

	TDQ_LOCK_ASSERT(tdq, MA_OWNED);
	if ((td = runq_steal(&tdq->tdq_realtime, cpu)) != NULL)
		return (td);
	if ((td = runq_steal_from(&tdq->tdq_timeshare,
	    cpu, tdq->tdq_ridx)) != NULL)
		return (td);
	return (runq_steal(&tdq->tdq_idle, cpu));
}

/*
 * Sets the thread lock and ts_cpu to match the requested cpu.  Unlocks the
 * current lock and returns with the assigned queue locked.
 */
static inline struct tdq *
sched_setcpu(struct thread *td, int cpu, int flags)
{

	struct tdq *tdq;
	struct mtx *mtx;

	THREAD_LOCK_ASSERT(td, MA_OWNED);
	tdq = TDQ_CPU(cpu);
	td_get_sched(td)->ts_cpu = cpu;
	/*
	 * If the lock matches just return the queue.
	 */
	if (td->td_lock == TDQ_LOCKPTR(tdq)) {
		KASSERT((flags & SRQ_HOLD) == 0,
		    ("sched_setcpu: Invalid lock for SRQ_HOLD"));
		return (tdq);
	}

	/*
	 * The hard case, migration, we need to block the thread first to
	 * prevent order reversals with other cpus locks.
	 */
	spinlock_enter();
	mtx = thread_lock_block(td);
	if ((flags & SRQ_HOLD) == 0)
		mtx_unlock_spin(mtx);
	TDQ_LOCK(tdq);
	thread_lock_unblock(td, TDQ_LOCKPTR(tdq));
	spinlock_exit();
	return (tdq);
}

SCHED_STAT_DEFINE(pickcpu_intrbind, "Soft interrupt binding");
SCHED_STAT_DEFINE(pickcpu_idle_affinity, "Picked idle cpu based on affinity");
SCHED_STAT_DEFINE(pickcpu_affinity, "Picked cpu based on affinity");
SCHED_STAT_DEFINE(pickcpu_lowest, "Selected lowest load");
SCHED_STAT_DEFINE(pickcpu_local, "Migrated to current cpu");
SCHED_STAT_DEFINE(pickcpu_migration, "Selection may have caused migration");

static int
sched_pickcpu(struct thread *td, int flags)
{
	struct cpu_group *cg, *ccg;
	struct td_sched *ts;
	struct tdq *tdq;
	cpuset_t *mask;
	int cpu, pri, self, intr;

	self = PCPU_GET(cpuid);
	ts = td_get_sched(td);
	KASSERT(!CPU_ABSENT(ts->ts_cpu), ("sched_pickcpu: Start scheduler on "
	    "absent CPU %d for thread %s.", ts->ts_cpu, td->td_name));
	if (smp_started == 0)
		return (self);
	/*
	 * Don't migrate a running thread from sched_switch().
	 */
	if ((flags & SRQ_OURSELF) || !THREAD_CAN_MIGRATE(td))
		return (ts->ts_cpu);
	/*
	 * Prefer to run interrupt threads on the processors that generate
	 * the interrupt.
	 */
	if (td->td_priority <= PRI_MAX_ITHD && THREAD_CAN_SCHED(td, self) &&
	    curthread->td_intr_nesting_level) {
		tdq = TDQ_SELF();
		if (tdq->tdq_lowpri >= PRI_MIN_IDLE) {
			SCHED_STAT_INC(pickcpu_idle_affinity);
			return (self);
		}
		ts->ts_cpu = self;
		intr = 1;
		cg = tdq->tdq_cg;
		goto llc;
	} else {
		intr = 0;
		tdq = TDQ_CPU(ts->ts_cpu);
		cg = tdq->tdq_cg;
	}
	/*
	 * If the thread can run on the last cpu and the affinity has not
	 * expired and it is idle, run it there.
	 */
	if (THREAD_CAN_SCHED(td, ts->ts_cpu) &&
	    tdq->tdq_lowpri >= PRI_MIN_IDLE &&
	    SCHED_AFFINITY(ts, CG_SHARE_L2)) {
		if (cg->cg_flags & CG_FLAG_THREAD) {
			/* Check all SMT threads for being idle. */
			for (cpu = cg->cg_first; cpu <= cg->cg_last; cpu++) {
				if (CPU_ISSET(cpu, &cg->cg_mask) &&
				    TDQ_CPU(cpu)->tdq_lowpri < PRI_MIN_IDLE)
					break;
			}
			if (cpu > cg->cg_last) {
				SCHED_STAT_INC(pickcpu_idle_affinity);
				return (ts->ts_cpu);
			}
		} else {
			SCHED_STAT_INC(pickcpu_idle_affinity);
			return (ts->ts_cpu);
		}
	}
llc:
	/*
	 * Search for the last level cache CPU group in the tree.
	 * Skip SMT, identical groups and caches with expired affinity.
	 * Interrupt threads affinity is explicit and never expires.
	 */
	for (ccg = NULL; cg != NULL; cg = cg->cg_parent) {
		if (cg->cg_flags & CG_FLAG_THREAD)
			continue;
		if (cg->cg_children == 1 || cg->cg_count == 1)
			continue;
		if (cg->cg_level == CG_SHARE_NONE ||
		    (!intr && !SCHED_AFFINITY(ts, cg->cg_level)))
			continue;
		ccg = cg;
	}
	/* Found LLC shared by all CPUs, so do a global search. */
	if (ccg == cpu_top)
		ccg = NULL;
	cpu = -1;
	mask = &td->td_cpuset->cs_mask;
	pri = td->td_priority;
	/*
	 * Try hard to keep interrupts within found LLC.  Search the LLC for
	 * the least loaded CPU we can run now.  For NUMA systems it should
	 * be within target domain, and it also reduces scheduling overhead.
	 */
	if (ccg != NULL && intr) {
		cpu = sched_lowest(ccg, mask, pri, INT_MAX, ts->ts_cpu);
		if (cpu >= 0)
			SCHED_STAT_INC(pickcpu_intrbind);
	} else
	/* Search the LLC for the least loaded idle CPU we can run now. */
	if (ccg != NULL) {
		cpu = sched_lowest(ccg, mask, max(pri, PRI_MAX_TIMESHARE),
		    INT_MAX, ts->ts_cpu);
		if (cpu >= 0)
			SCHED_STAT_INC(pickcpu_affinity);
	}
	/* Search globally for the least loaded CPU we can run now. */
	if (cpu < 0) {
		cpu = sched_lowest(cpu_top, mask, pri, INT_MAX, ts->ts_cpu);
		if (cpu >= 0)
			SCHED_STAT_INC(pickcpu_lowest);
	}
	/* Search globally for the least loaded CPU. */
	if (cpu < 0) {
		cpu = sched_lowest(cpu_top, mask, -1, INT_MAX, ts->ts_cpu);
		if (cpu >= 0)
			SCHED_STAT_INC(pickcpu_lowest);
	}
	KASSERT(cpu >= 0, ("sched_pickcpu: Failed to find a cpu."));
	KASSERT(!CPU_ABSENT(cpu), ("sched_pickcpu: Picked absent CPU %d.", cpu));
	/*
	 * Compare the lowest loaded cpu to current cpu.
	 */
	tdq = TDQ_CPU(cpu);
	if (THREAD_CAN_SCHED(td, self) && TDQ_SELF()->tdq_lowpri > pri &&
	    tdq->tdq_lowpri < PRI_MIN_IDLE &&
	    TDQ_SELF()->tdq_load <= tdq->tdq_load + 1) {
		SCHED_STAT_INC(pickcpu_local);
		cpu = self;
	}
	if (cpu != ts->ts_cpu)
		SCHED_STAT_INC(pickcpu_migration);
	return (cpu);
}
#endif

/*
 * Pick the highest priority task we have and return it.
 */
static struct thread *
tdq_choose(struct tdq *tdq)
{
	struct thread *td;

	TDQ_LOCK_ASSERT(tdq, MA_OWNED);
	td = runq_choose(&tdq->tdq_realtime);
	if (td != NULL)
		return (td);
	td = runq_choose_from(&tdq->tdq_timeshare, tdq->tdq_ridx);
	if (td != NULL) {
		KASSERT(td->td_priority >= PRI_MIN_BATCH,
		    ("tdq_choose: Invalid priority on timeshare queue %d",
		    td->td_priority));
		return (td);
	}
	td = runq_choose(&tdq->tdq_idle);
	if (td != NULL) {
		KASSERT(td->td_priority >= PRI_MIN_IDLE,
		    ("tdq_choose: Invalid priority on idle queue %d",
		    td->td_priority));
		return (td);
	}

	return (NULL);
}

/*
 * Initialize a thread queue.
 */
static void
tdq_setup(struct tdq *tdq, int id)
{

	if (bootverbose)
		printf("ULE: setup cpu %d\n", id);
	runq_init(&tdq->tdq_realtime);
	runq_init(&tdq->tdq_timeshare);
	runq_init(&tdq->tdq_idle);
	tdq->tdq_id = id;
	snprintf(tdq->tdq_name, sizeof(tdq->tdq_name),
	    "sched lock %d", (int)TDQ_ID(tdq));
	mtx_init(&tdq->tdq_lock, tdq->tdq_name, "sched lock", MTX_SPIN);
#ifdef KTR
	snprintf(tdq->tdq_loadname, sizeof(tdq->tdq_loadname),
	    "CPU %d load", (int)TDQ_ID(tdq));
#endif
}

#ifdef SMP
static void
sched_setup_smp(void)
{
	struct tdq *tdq;
	int i;

	cpu_top = smp_topo();
	CPU_FOREACH(i) {
		tdq = DPCPU_ID_PTR(i, tdq);
		tdq_setup(tdq, i);
		tdq->tdq_cg = smp_topo_find(cpu_top, i);
		if (tdq->tdq_cg == NULL)
			panic("Can't find cpu group for %d\n", i);
	}
	PCPU_SET(sched, DPCPU_PTR(tdq));
	balance_tdq = TDQ_SELF();
}
#endif

/*
 * Setup the thread queues and initialize the topology based on MD
 * information.
 */
static void
sched_setup(void *dummy)
{
	struct tdq *tdq;

#ifdef SMP
	sched_setup_smp();
#else
	tdq_setup(TDQ_SELF(), 0);
#endif
	tdq = TDQ_SELF();

	/* Add thread0's load since it's running. */
	TDQ_LOCK(tdq);
	thread0.td_lock = TDQ_LOCKPTR(tdq);
	tdq_load_add(tdq, &thread0);
	tdq->tdq_lowpri = thread0.td_priority;
	TDQ_UNLOCK(tdq);
}

/*
 * This routine determines time constants after stathz and hz are setup.
 */
/* ARGSUSED */
static void
sched_initticks(void *dummy)
{
	int incr;

	realstathz = stathz ? stathz : hz;
	sched_slice = realstathz / SCHED_SLICE_DEFAULT_DIVISOR;
	sched_slice_min = sched_slice / SCHED_SLICE_MIN_DIVISOR;
	hogticks = imax(1, (2 * hz * sched_slice + realstathz / 2) /
	    realstathz);

	/*
	 * tickincr is shifted out by 10 to avoid rounding errors due to
	 * hz not being evenly divisible by stathz on all platforms.
	 */
	incr = (hz << SCHED_TICK_SHIFT) / realstathz;
	/*
	 * This does not work for values of stathz that are more than
	 * 1 << SCHED_TICK_SHIFT * hz.  In practice this does not happen.
	 */
	if (incr == 0)
		incr = 1;
	tickincr = incr;
#ifdef SMP
	/*
	 * Set the default balance interval now that we know
	 * what realstathz is.
	 */
	balance_interval = realstathz;
	balance_ticks = balance_interval;
	affinity = SCHED_AFFINITY_DEFAULT;
#endif
	if (sched_idlespinthresh < 0)
		sched_idlespinthresh = 2 * max(10000, 6 * hz) / realstathz;
}

/*
 * This is the core of the interactivity algorithm.  Determines a score based
 * on past behavior.  It is the ratio of sleep time to run time scaled to
 * a [0, 100] integer.  This is the voluntary sleep time of a process, which
 * differs from the cpu usage because it does not account for time spent
 * waiting on a run-queue.  Would be prettier if we had floating point.
 *
 * When a thread's sleep time is greater than its run time the
 * calculation is:
 *
 *                           scaling factor
 * interactivity score =  ---------------------
 *                        sleep time / run time
 *
 *
 * When a thread's run time is greater than its sleep time the
 * calculation is:
 *
 *                                                 scaling factor
 * interactivity score = 2 * scaling factor  -  ---------------------
 *                                              run time / sleep time
 */
static int
sched_interact_score(struct thread *td)
{
	struct td_sched *ts;
	int div;

	ts = td_get_sched(td);
	/*
	 * The score is only needed if this is likely to be an interactive
	 * task.  Don't go through the expense of computing it if there's
	 * no chance.
	 */
	if (sched_interact <= SCHED_INTERACT_HALF &&
		ts->ts_runtime >= ts->ts_slptime)
			return (SCHED_INTERACT_HALF);

	if (ts->ts_runtime > ts->ts_slptime) {
		div = max(1, ts->ts_runtime / SCHED_INTERACT_HALF);
		return (SCHED_INTERACT_HALF +
		    (SCHED_INTERACT_HALF - (ts->ts_slptime / div)));
	}
	if (ts->ts_slptime > ts->ts_runtime) {
		div = max(1, ts->ts_slptime / SCHED_INTERACT_HALF);
		return (ts->ts_runtime / div);
	}
	/* runtime == slptime */
	if (ts->ts_runtime)
		return (SCHED_INTERACT_HALF);

	/*
	 * This can happen if slptime and runtime are 0.
	 */
	return (0);

}

/*
 * Scale the scheduling priority according to the "interactivity" of this
 * process.
 */
static void
sched_priority(struct thread *td)
{
	int score;
	int pri;

	if (PRI_BASE(td->td_pri_class) != PRI_TIMESHARE)
		return;
	/*
	 * If the score is interactive we place the thread in the realtime
	 * queue with a priority that is less than kernel and interrupt
	 * priorities.  These threads are not subject to nice restrictions.
	 *
	 * Scores greater than this are placed on the normal timeshare queue
	 * where the priority is partially decided by the most recent cpu
	 * utilization and the rest is decided by nice value.
	 *
	 * The nice value of the process has a linear effect on the calculated
	 * score.  Negative nice values make it easier for a thread to be
	 * considered interactive.
	 */
	score = imax(0, sched_interact_score(td) + td->td_proc->p_nice);
	if (score < sched_interact) {
		pri = PRI_MIN_INTERACT;
		pri += ((PRI_MAX_INTERACT - PRI_MIN_INTERACT + 1) /
		    sched_interact) * score;
		KASSERT(pri >= PRI_MIN_INTERACT && pri <= PRI_MAX_INTERACT,
		    ("sched_priority: invalid interactive priority %d score %d",
		    pri, score));
	} else {
		pri = SCHED_PRI_MIN;
		if (td_get_sched(td)->ts_ticks)
			pri += min(SCHED_PRI_TICKS(td_get_sched(td)),
			    SCHED_PRI_RANGE - 1);
		pri += SCHED_PRI_NICE(td->td_proc->p_nice);
		KASSERT(pri >= PRI_MIN_BATCH && pri <= PRI_MAX_BATCH,
		    ("sched_priority: invalid priority %d: nice %d, " 
		    "ticks %d ftick %d ltick %d tick pri %d",
		    pri, td->td_proc->p_nice, td_get_sched(td)->ts_ticks,
		    td_get_sched(td)->ts_ftick, td_get_sched(td)->ts_ltick,
		    SCHED_PRI_TICKS(td_get_sched(td))));
	}
	sched_user_prio(td, pri);

	return;
}

/*
 * This routine enforces a maximum limit on the amount of scheduling history
 * kept.  It is called after either the slptime or runtime is adjusted.  This
 * function is ugly due to integer math.
 */
static void
sched_interact_update(struct thread *td)
{
	struct td_sched *ts;
	u_int sum;

	ts = td_get_sched(td);
	sum = ts->ts_runtime + ts->ts_slptime;
	if (sum < SCHED_SLP_RUN_MAX)
		return;
	/*
	 * This only happens from two places:
	 * 1) We have added an unusual amount of run time from fork_exit.
	 * 2) We have added an unusual amount of sleep time from sched_sleep().
	 */
	if (sum > SCHED_SLP_RUN_MAX * 2) {
		if (ts->ts_runtime > ts->ts_slptime) {
			ts->ts_runtime = SCHED_SLP_RUN_MAX;
			ts->ts_slptime = 1;
		} else {
			ts->ts_slptime = SCHED_SLP_RUN_MAX;
			ts->ts_runtime = 1;
		}
		return;
	}
	/*
	 * If we have exceeded by more than 1/5th then the algorithm below
	 * will not bring us back into range.  Dividing by two here forces
	 * us into the range of [4/5 * SCHED_INTERACT_MAX, SCHED_INTERACT_MAX]
	 */
	if (sum > (SCHED_SLP_RUN_MAX / 5) * 6) {
		ts->ts_runtime /= 2;
		ts->ts_slptime /= 2;
		return;
	}
	ts->ts_runtime = (ts->ts_runtime / 5) * 4;
	ts->ts_slptime = (ts->ts_slptime / 5) * 4;
}

/*
 * Scale back the interactivity history when a child thread is created.  The
 * history is inherited from the parent but the thread may behave totally
 * differently.  For example, a shell spawning a compiler process.  We want
 * to learn that the compiler is behaving badly very quickly.
 */
static void
sched_interact_fork(struct thread *td)
{
	struct td_sched *ts;
	int ratio;
	int sum;

	ts = td_get_sched(td);
	sum = ts->ts_runtime + ts->ts_slptime;
	if (sum > SCHED_SLP_RUN_FORK) {
		ratio = sum / SCHED_SLP_RUN_FORK;
		ts->ts_runtime /= ratio;
		ts->ts_slptime /= ratio;
	}
}

/*
 * Called from proc0_init() to setup the scheduler fields.
 */
void
schedinit(void)
{
	struct td_sched *ts0;

	/*
	 * Set up the scheduler specific parts of thread0.
	 */
	ts0 = td_get_sched(&thread0);
	ts0->ts_ltick = ticks;
	ts0->ts_ftick = ticks;
	ts0->ts_slice = 0;
	ts0->ts_cpu = curcpu;	/* set valid CPU number */
}

/*
 * This is only somewhat accurate since given many processes of the same
 * priority they will switch when their slices run out, which will be
 * at most sched_slice stathz ticks.
 */
int
sched_rr_interval(void)
{

	/* Convert sched_slice from stathz to hz. */
	return (imax(1, (sched_slice * hz + realstathz / 2) / realstathz));
}

/*
 * Update the percent cpu tracking information when it is requested or
 * the total history exceeds the maximum.  We keep a sliding history of
 * tick counts that slowly decays.  This is less precise than the 4BSD
 * mechanism since it happens with less regular and frequent events.
 */
static void
sched_pctcpu_update(struct td_sched *ts, int run)
{
	int t = ticks;

	/*
	 * The signed difference may be negative if the thread hasn't run for
	 * over half of the ticks rollover period.
	 */
	if ((u_int)(t - ts->ts_ltick) >= SCHED_TICK_TARG) {
		ts->ts_ticks = 0;
		ts->ts_ftick = t - SCHED_TICK_TARG;
	} else if (t - ts->ts_ftick >= SCHED_TICK_MAX) {
		ts->ts_ticks = (ts->ts_ticks / (ts->ts_ltick - ts->ts_ftick)) *
		    (ts->ts_ltick - (t - SCHED_TICK_TARG));
		ts->ts_ftick = t - SCHED_TICK_TARG;
	}
	if (run)
		ts->ts_ticks += (t - ts->ts_ltick) << SCHED_TICK_SHIFT;
	ts->ts_ltick = t;
}

/*
 * Adjust the priority of a thread.  Move it to the appropriate run-queue
 * if necessary.  This is the back-end for several priority related
 * functions.
 */
static void
sched_thread_priority(struct thread *td, u_char prio)
{
	struct td_sched *ts;
	struct tdq *tdq;
	int oldpri;

	KTR_POINT3(KTR_SCHED, "thread", sched_tdname(td), "prio",
	    "prio:%d", td->td_priority, "new prio:%d", prio,
	    KTR_ATTR_LINKED, sched_tdname(curthread));
	SDT_PROBE3(sched, , , change__pri, td, td->td_proc, prio);
	if (td != curthread && prio < td->td_priority) {
		KTR_POINT3(KTR_SCHED, "thread", sched_tdname(curthread),
		    "lend prio", "prio:%d", td->td_priority, "new prio:%d",
		    prio, KTR_ATTR_LINKED, sched_tdname(td));
		SDT_PROBE4(sched, , , lend__pri, td, td->td_proc, prio, 
		    curthread);
	} 
	ts = td_get_sched(td);
	THREAD_LOCK_ASSERT(td, MA_OWNED);
	if (td->td_priority == prio)
		return;
	/*
	 * If the priority has been elevated due to priority
	 * propagation, we may have to move ourselves to a new
	 * queue.  This could be optimized to not re-add in some
	 * cases.
	 */
	if (TD_ON_RUNQ(td) && prio < td->td_priority) {
		sched_rem(td);
		td->td_priority = prio;
		sched_add(td, SRQ_BORROWING | SRQ_HOLDTD);
		return;
	}
	/*
	 * If the thread is currently running we may have to adjust the lowpri
	 * information so other cpus are aware of our current priority.
	 */
	if (TD_IS_RUNNING(td)) {
		tdq = TDQ_CPU(ts->ts_cpu);
		oldpri = td->td_priority;
		td->td_priority = prio;
		if (prio < tdq->tdq_lowpri)
			tdq->tdq_lowpri = prio;
		else if (tdq->tdq_lowpri == oldpri)
			tdq_setlowpri(tdq, td);
		return;
	}
	td->td_priority = prio;
}

/*
 * Update a thread's priority when it is lent another thread's
 * priority.
 */
void
sched_lend_prio(struct thread *td, u_char prio)
{

	td->td_flags |= TDF_BORROWING;
	sched_thread_priority(td, prio);
}

/*
 * Restore a thread's priority when priority propagation is
 * over.  The prio argument is the minimum priority the thread
 * needs to have to satisfy other possible priority lending
 * requests.  If the thread's regular priority is less
 * important than prio, the thread will keep a priority boost
 * of prio.
 */
void
sched_unlend_prio(struct thread *td, u_char prio)
{
	u_char base_pri;

	if (td->td_base_pri >= PRI_MIN_TIMESHARE &&
	    td->td_base_pri <= PRI_MAX_TIMESHARE)
		base_pri = td->td_user_pri;
	else
		base_pri = td->td_base_pri;
	if (prio >= base_pri) {
		td->td_flags &= ~TDF_BORROWING;
		sched_thread_priority(td, base_pri);
	} else
		sched_lend_prio(td, prio);
}

/*
 * Standard entry for setting the priority to an absolute value.
 */
void
sched_prio(struct thread *td, u_char prio)
{
	u_char oldprio;

	/* First, update the base priority. */
	td->td_base_pri = prio;

	/*
	 * If the thread is borrowing another thread's priority, don't
	 * ever lower the priority.
	 */
	if (td->td_flags & TDF_BORROWING && td->td_priority < prio)
		return;

	/* Change the real priority. */
	oldprio = td->td_priority;
	sched_thread_priority(td, prio);

	/*
	 * If the thread is on a turnstile, then let the turnstile update
	 * its state.
	 */
	if (TD_ON_LOCK(td) && oldprio != prio)
		turnstile_adjust(td, oldprio);
}

/*
 * Set the base user priority, does not effect current running priority.
 */
void
sched_user_prio(struct thread *td, u_char prio)
{

	td->td_base_user_pri = prio;
	if (td->td_lend_user_pri <= prio)
		return;
	td->td_user_pri = prio;
}

void
sched_lend_user_prio(struct thread *td, u_char prio)
{

	THREAD_LOCK_ASSERT(td, MA_OWNED);
	td->td_lend_user_pri = prio;
	td->td_user_pri = min(prio, td->td_base_user_pri);
	if (td->td_priority > td->td_user_pri)
		sched_prio(td, td->td_user_pri);
	else if (td->td_priority != td->td_user_pri)
		td->td_flags |= TDF_NEEDRESCHED;
}

/*
 * Like the above but first check if there is anything to do.
 */
void
sched_lend_user_prio_cond(struct thread *td, u_char prio)
{

	if (td->td_lend_user_pri != prio)
		goto lend;
	if (td->td_user_pri != min(prio, td->td_base_user_pri))
		goto lend;
	if (td->td_priority != td->td_user_pri)
		goto lend;
	return;

lend:
	thread_lock(td);
	sched_lend_user_prio(td, prio);
	thread_unlock(td);
}

#ifdef SMP
/*
 * This tdq is about to idle.  Try to steal a thread from another CPU before
 * choosing the idle thread.
 */
static void
tdq_trysteal(struct tdq *tdq)
{
	struct cpu_group *cg;
	struct tdq *steal;
	cpuset_t mask;
	int cpu, i;

	if (smp_started == 0 || trysteal_limit == 0 || tdq->tdq_cg == NULL)
		return;
	CPU_FILL(&mask);
	CPU_CLR(PCPU_GET(cpuid), &mask);
	/* We don't want to be preempted while we're iterating. */
	spinlock_enter();
	TDQ_UNLOCK(tdq);
	for (i = 1, cg = tdq->tdq_cg; ; ) {
		cpu = sched_highest(cg, &mask, steal_thresh);
		/*
		 * If a thread was added while interrupts were disabled don't
		 * steal one here.
		 */
		if (tdq->tdq_load > 0) {
			TDQ_LOCK(tdq);
			break;
		}
		if (cpu == -1) {
			i++;
			cg = cg->cg_parent;
			if (cg == NULL || i > trysteal_limit) {
				TDQ_LOCK(tdq);
				break;
			}
			continue;
		}
		steal = TDQ_CPU(cpu);
		/*
		 * The data returned by sched_highest() is stale and
                 * the chosen CPU no longer has an eligible thread.
		 */
		if (steal->tdq_load < steal_thresh ||
		    steal->tdq_transferable == 0)
			continue;
		tdq_lock_pair(tdq, steal);
		/*
		 * If we get to this point, unconditonally exit the loop
		 * to bound the time spent in the critcal section.
		 *
		 * If a thread was added while interrupts were disabled don't
		 * steal one here.
		 */
		if (tdq->tdq_load > 0) {
			TDQ_UNLOCK(steal);
			break;
		}
		/*
		 * The data returned by sched_highest() is stale and
                 * the chosen CPU no longer has an eligible thread.
		 */
		if (steal->tdq_load < steal_thresh ||
		    steal->tdq_transferable == 0) {
			TDQ_UNLOCK(steal);
			break;
		}
		/*
		 * If we fail to acquire one due to affinity restrictions,
		 * bail out and let the idle thread to a more complete search
		 * outside of a critical section.
		 */
		if (tdq_move(steal, tdq) == NULL) {
			TDQ_UNLOCK(steal);
			break;
		}
		TDQ_UNLOCK(steal);
		break;
	}
	spinlock_exit();
}
#endif

/*
 * Handle migration from sched_switch().  This happens only for
 * cpu binding.
 */
static struct mtx *
sched_switch_migrate(struct tdq *tdq, struct thread *td, int flags)
{
	struct tdq *tdn;

	KASSERT(THREAD_CAN_MIGRATE(td) ||
	    (td_get_sched(td)->ts_flags & TSF_BOUND) != 0,
	    ("Thread %p shouldn't migrate", td));
	KASSERT(!CPU_ABSENT(td_get_sched(td)->ts_cpu), ("sched_switch_migrate: "
	    "thread %s queued on absent CPU %d.", td->td_name,
	    td_get_sched(td)->ts_cpu));
	tdn = TDQ_CPU(td_get_sched(td)->ts_cpu);
#ifdef SMP
	tdq_load_rem(tdq, td);
	/*
	 * Do the lock dance required to avoid LOR.  We have an 
	 * extra spinlock nesting from sched_switch() which will
	 * prevent preemption while we're holding neither run-queue lock.
	 */
	TDQ_UNLOCK(tdq);
	TDQ_LOCK(tdn);
	tdq_add(tdn, td, flags);
	tdq_notify(tdn, td);
	TDQ_UNLOCK(tdn);
	TDQ_LOCK(tdq);
#endif
	return (TDQ_LOCKPTR(tdn));
}

/*
 * thread_lock_unblock() that does not assume td_lock is blocked.
 */
static inline void
thread_unblock_switch(struct thread *td, struct mtx *mtx)
{
	atomic_store_rel_ptr((volatile uintptr_t *)&td->td_lock,
	    (uintptr_t)mtx);
}

/*
 * Switch threads.  This function has to handle threads coming in while
 * blocked for some reason, running, or idle.  It also must deal with
 * migrating a thread from one queue to another as running threads may
 * be assigned elsewhere via binding.
 */
void
sched_switch(struct thread *td, int flags)
{
	struct thread *newtd;
	struct tdq *tdq;
	struct td_sched *ts;
	struct mtx *mtx;
	int srqflag;
	int cpuid, preempted;

	THREAD_LOCK_ASSERT(td, MA_OWNED);

	cpuid = PCPU_GET(cpuid);
	tdq = TDQ_SELF();
	ts = td_get_sched(td);
	sched_pctcpu_update(ts, 1);
	ts->ts_rltick = ticks;
	td->td_lastcpu = td->td_oncpu;
	preempted = (td->td_flags & TDF_SLICEEND) == 0 &&
	    (flags & SW_PREEMPT) != 0;
	td->td_flags &= ~(TDF_NEEDRESCHED | TDF_SLICEEND);
	td->td_owepreempt = 0;
	tdq->tdq_owepreempt = 0;
	if (!TD_IS_IDLETHREAD(td))
		tdq->tdq_switchcnt++;

	/*
	 * Always block the thread lock so we can drop the tdq lock early.
	 */
	mtx = thread_lock_block(td);
	spinlock_enter();
	if (TD_IS_IDLETHREAD(td)) {
		MPASS(mtx == TDQ_LOCKPTR(tdq));
		TD_SET_CAN_RUN(td);
	} else if (TD_IS_RUNNING(td)) {
		MPASS(mtx == TDQ_LOCKPTR(tdq));
		srqflag = preempted ?
		    SRQ_OURSELF|SRQ_YIELDING|SRQ_PREEMPTED :
		    SRQ_OURSELF|SRQ_YIELDING;
#ifdef SMP
		if (THREAD_CAN_MIGRATE(td) && !THREAD_CAN_SCHED(td, ts->ts_cpu))
			ts->ts_cpu = sched_pickcpu(td, 0);
#endif
		if (ts->ts_cpu == cpuid)
			tdq_runq_add(tdq, td, srqflag);
		else
			mtx = sched_switch_migrate(tdq, td, srqflag);
	} else {
		/* This thread must be going to sleep. */
		if (mtx != TDQ_LOCKPTR(tdq)) {
			mtx_unlock_spin(mtx);
			TDQ_LOCK(tdq);
		}
		tdq_load_rem(tdq, td);
#ifdef SMP
		if (tdq->tdq_load == 0)
			tdq_trysteal(tdq);
#endif
	}

#if (KTR_COMPILE & KTR_SCHED) != 0
	if (TD_IS_IDLETHREAD(td))
		KTR_STATE1(KTR_SCHED, "thread", sched_tdname(td), "idle",
		    "prio:%d", td->td_priority);
	else
		KTR_STATE3(KTR_SCHED, "thread", sched_tdname(td), KTDSTATE(td),
		    "prio:%d", td->td_priority, "wmesg:\"%s\"", td->td_wmesg,
		    "lockname:\"%s\"", td->td_lockname);
#endif

	/*
	 * We enter here with the thread blocked and assigned to the
	 * appropriate cpu run-queue or sleep-queue and with the current
	 * thread-queue locked.
	 */
	TDQ_LOCK_ASSERT(tdq, MA_OWNED | MA_NOTRECURSED);
	newtd = choosethread();
	sched_pctcpu_update(td_get_sched(newtd), 0);
	TDQ_UNLOCK(tdq);

	/*
	 * Call the MD code to switch contexts if necessary.
	 */
	if (td != newtd) {
#ifdef	HWPMC_HOOKS
		if (PMC_PROC_IS_USING_PMCS(td->td_proc))
			PMC_SWITCH_CONTEXT(td, PMC_FN_CSW_OUT);
#endif
		SDT_PROBE2(sched, , , off__cpu, newtd, newtd->td_proc);

#ifdef KDTRACE_HOOKS
		/*
		 * If DTrace has set the active vtime enum to anything
		 * other than INACTIVE (0), then it should have set the
		 * function to call.
		 */
		if (dtrace_vtime_active)
			(*dtrace_vtime_switch_func)(newtd);
#endif
		td->td_oncpu = NOCPU;
		cpu_switch(td, newtd, mtx);
		cpuid = td->td_oncpu = PCPU_GET(cpuid);

		SDT_PROBE0(sched, , , on__cpu);
#ifdef	HWPMC_HOOKS
		if (PMC_PROC_IS_USING_PMCS(td->td_proc))
			PMC_SWITCH_CONTEXT(td, PMC_FN_CSW_IN);
#endif
	} else {
		thread_unblock_switch(td, mtx);
		SDT_PROBE0(sched, , , remain__cpu);
	}
	KASSERT(curthread->td_md.md_spinlock_count == 1,
	    ("invalid count %d", curthread->td_md.md_spinlock_count));

	KTR_STATE1(KTR_SCHED, "thread", sched_tdname(td), "running",
	    "prio:%d", td->td_priority);
}

/*
 * Adjust thread priorities as a result of a nice request.
 */
void
sched_nice(struct proc *p, int nice)
{
	struct thread *td;

	PROC_LOCK_ASSERT(p, MA_OWNED);

	p->p_nice = nice;
	FOREACH_THREAD_IN_PROC(p, td) {
		thread_lock(td);
		sched_priority(td);
		sched_prio(td, td->td_base_user_pri);
		thread_unlock(td);
	}
}

/*
 * Record the sleep time for the interactivity scorer.
 */
void
sched_sleep(struct thread *td, int prio)
{

	THREAD_LOCK_ASSERT(td, MA_OWNED);

	td->td_slptick = ticks;
	if (TD_IS_SUSPENDED(td) || prio >= PSOCK)
		td->td_flags |= TDF_CANSWAP;
	if (PRI_BASE(td->td_pri_class) != PRI_TIMESHARE)
		return;
	if (static_boost == 1 && prio)
		sched_prio(td, prio);
	else if (static_boost && td->td_priority > static_boost)
		sched_prio(td, static_boost);
}

/*
 * Schedule a thread to resume execution and record how long it voluntarily
 * slept.  We also update the pctcpu, interactivity, and priority.
 *
 * Requires the thread lock on entry, drops on exit.
 */
void
sched_wakeup(struct thread *td, int srqflags)
{
	struct td_sched *ts;
	int slptick;

	THREAD_LOCK_ASSERT(td, MA_OWNED);
	ts = td_get_sched(td);
	td->td_flags &= ~TDF_CANSWAP;

	/*
	 * If we slept for more than a tick update our interactivity and
	 * priority.
	 */
	slptick = td->td_slptick;
	td->td_slptick = 0;
	if (slptick && slptick != ticks) {
		ts->ts_slptime += (ticks - slptick) << SCHED_TICK_SHIFT;
		sched_interact_update(td);
		sched_pctcpu_update(ts, 0);
	}
	/*
	 * Reset the slice value since we slept and advanced the round-robin.
	 */
	ts->ts_slice = 0;
	sched_add(td, SRQ_BORING | srqflags);
}

/*
 * Penalize the parent for creating a new child and initialize the child's
 * priority.
 */
void
sched_fork(struct thread *td, struct thread *child)
{
	THREAD_LOCK_ASSERT(td, MA_OWNED);
	sched_pctcpu_update(td_get_sched(td), 1);
	sched_fork_thread(td, child);
	/*
	 * Penalize the parent and child for forking.
	 */
	sched_interact_fork(child);
	sched_priority(child);
	td_get_sched(td)->ts_runtime += tickincr;
	sched_interact_update(td);
	sched_priority(td);
}

/*
 * Fork a new thread, may be within the same process.
 */
void
sched_fork_thread(struct thread *td, struct thread *child)
{
	struct td_sched *ts;
	struct td_sched *ts2;
	struct tdq *tdq;

	tdq = TDQ_SELF();
	THREAD_LOCK_ASSERT(td, MA_OWNED);
	/*
	 * Initialize child.
	 */
	ts = td_get_sched(td);
	ts2 = td_get_sched(child);
	child->td_oncpu = NOCPU;
	child->td_lastcpu = NOCPU;
	child->td_lock = TDQ_LOCKPTR(tdq);
	child->td_cpuset = cpuset_ref(td->td_cpuset);
	child->td_domain.dr_policy = td->td_cpuset->cs_domain;
	ts2->ts_cpu = ts->ts_cpu;
	ts2->ts_flags = 0;
	/*
	 * Grab our parents cpu estimation information.
	 */
	ts2->ts_ticks = ts->ts_ticks;
	ts2->ts_ltick = ts->ts_ltick;
	ts2->ts_ftick = ts->ts_ftick;
	/*
	 * Do not inherit any borrowed priority from the parent.
	 */
	child->td_priority = child->td_base_pri;
	/*
	 * And update interactivity score.
	 */
	ts2->ts_slptime = ts->ts_slptime;
	ts2->ts_runtime = ts->ts_runtime;
	/* Attempt to quickly learn interactivity. */
	ts2->ts_slice = tdq_slice(tdq) - sched_slice_min;
#ifdef KTR
	bzero(ts2->ts_name, sizeof(ts2->ts_name));
#endif
}

/*
 * Adjust the priority class of a thread.
 */
void
sched_class(struct thread *td, int class)
{

	THREAD_LOCK_ASSERT(td, MA_OWNED);
	if (td->td_pri_class == class)
		return;
	td->td_pri_class = class;
}

/*
 * Return some of the child's priority and interactivity to the parent.
 */
void
sched_exit(struct proc *p, struct thread *child)
{
	struct thread *td;

	KTR_STATE1(KTR_SCHED, "thread", sched_tdname(child), "proc exit",
	    "prio:%d", child->td_priority);
	PROC_LOCK_ASSERT(p, MA_OWNED);
	td = FIRST_THREAD_IN_PROC(p);
	sched_exit_thread(td, child);
}

/*
 * Penalize another thread for the time spent on this one.  This helps to
 * worsen the priority and interactivity of processes which schedule batch
 * jobs such as make.  This has little effect on the make process itself but
 * causes new processes spawned by it to receive worse scores immediately.
 */
void
sched_exit_thread(struct thread *td, struct thread *child)
{

	KTR_STATE1(KTR_SCHED, "thread", sched_tdname(child), "thread exit",
	    "prio:%d", child->td_priority);
	/*
	 * Give the child's runtime to the parent without returning the
	 * sleep time as a penalty to the parent.  This causes shells that
	 * launch expensive things to mark their children as expensive.
	 */
	thread_lock(td);
	td_get_sched(td)->ts_runtime += td_get_sched(child)->ts_runtime;
	sched_interact_update(td);
	sched_priority(td);
	thread_unlock(td);
}

void
sched_preempt(struct thread *td)
{
	struct tdq *tdq;
	int flags;

	SDT_PROBE2(sched, , , surrender, td, td->td_proc);

	thread_lock(td);
	tdq = TDQ_SELF();
	TDQ_LOCK_ASSERT(tdq, MA_OWNED);
	if (td->td_priority > tdq->tdq_lowpri) {
		if (td->td_critnest == 1) {
			flags = SW_INVOL | SW_PREEMPT;
			flags |= TD_IS_IDLETHREAD(td) ? SWT_REMOTEWAKEIDLE :
			    SWT_REMOTEPREEMPT;
			mi_switch(flags);
			/* Switch dropped thread lock. */
			return;
		}
		td->td_owepreempt = 1;
	} else {
		tdq->tdq_owepreempt = 0;
	}
	thread_unlock(td);
}

/*
 * Fix priorities on return to user-space.  Priorities may be elevated due
 * to static priorities in msleep() or similar.
 */
void
sched_userret_slowpath(struct thread *td)
{

	thread_lock(td);
	td->td_priority = td->td_user_pri;
	td->td_base_pri = td->td_user_pri;
	tdq_setlowpri(TDQ_SELF(), td);
	thread_unlock(td);
}

/*
 * Handle a stathz tick.  This is really only relevant for timeshare
 * threads.
 */
void
sched_clock(struct thread *td, int cnt)
{
	struct tdq *tdq;
	struct td_sched *ts;

	THREAD_LOCK_ASSERT(td, MA_OWNED);
	tdq = TDQ_SELF();
#ifdef SMP
	/*
	 * We run the long term load balancer infrequently on the first cpu.
	 */
	if (balance_tdq == tdq && smp_started != 0 && rebalance != 0 &&
	    balance_ticks != 0) {
		balance_ticks -= cnt;
		if (balance_ticks <= 0)
			sched_balance();
	}
#endif
	/*
	 * Save the old switch count so we have a record of the last ticks
	 * activity.   Initialize the new switch count based on our load.
	 * If there is some activity seed it to reflect that.
	 */
	tdq->tdq_oldswitchcnt = tdq->tdq_switchcnt;
	tdq->tdq_switchcnt = tdq->tdq_load;
	/*
	 * Advance the insert index once for each tick to ensure that all
	 * threads get a chance to run.
	 */
	if (tdq->tdq_idx == tdq->tdq_ridx) {
		tdq->tdq_idx = (tdq->tdq_idx + 1) % RQ_NQS;
		if (TAILQ_EMPTY(&tdq->tdq_timeshare.rq_queues[tdq->tdq_ridx]))
			tdq->tdq_ridx = tdq->tdq_idx;
	}
	ts = td_get_sched(td);
	sched_pctcpu_update(ts, 1);
	if ((td->td_pri_class & PRI_FIFO_BIT) || TD_IS_IDLETHREAD(td))
		return;

	if (PRI_BASE(td->td_pri_class) == PRI_TIMESHARE) {
		/*
		 * We used a tick; charge it to the thread so
		 * that we can compute our interactivity.
		 */
		td_get_sched(td)->ts_runtime += tickincr * cnt;
		sched_interact_update(td);
		sched_priority(td);
	}

	/*
	 * Force a context switch if the current thread has used up a full
	 * time slice (default is 100ms).
	 */
	ts->ts_slice += cnt;
	if (ts->ts_slice >= tdq_slice(tdq)) {
		ts->ts_slice = 0;
		td->td_flags |= TDF_NEEDRESCHED | TDF_SLICEEND;
	}
}

u_int
sched_estcpu(struct thread *td __unused)
{

	return (0);
}

/*
 * Return whether the current CPU has runnable tasks.  Used for in-kernel
 * cooperative idle threads.
 */
int
sched_runnable(void)
{
	struct tdq *tdq;
	int load;

	load = 1;

	tdq = TDQ_SELF();
	if ((curthread->td_flags & TDF_IDLETD) != 0) {
		if (tdq->tdq_load > 0)
			goto out;
	} else
		if (tdq->tdq_load - 1 > 0)
			goto out;
	load = 0;
out:
	return (load);
}

/*
 * Choose the highest priority thread to run.  The thread is removed from
 * the run-queue while running however the load remains.  For SMP we set
 * the tdq in the global idle bitmask if it idles here.
 */
struct thread *
sched_choose(void)
{
	struct thread *td;
	struct tdq *tdq;

	tdq = TDQ_SELF();
	TDQ_LOCK_ASSERT(tdq, MA_OWNED);
	td = tdq_choose(tdq);
	if (td) {
		tdq_runq_rem(tdq, td);
		tdq->tdq_lowpri = td->td_priority;
		return (td);
	}
	tdq->tdq_lowpri = PRI_MAX_IDLE;
	return (PCPU_GET(idlethread));
}

/*
 * Set owepreempt if necessary.  Preemption never happens directly in ULE,
 * we always request it once we exit a critical section.
 */
static inline void
sched_setpreempt(struct thread *td)
{
	struct thread *ctd;
	int cpri;
	int pri;

	THREAD_LOCK_ASSERT(curthread, MA_OWNED);

	ctd = curthread;
	pri = td->td_priority;
	cpri = ctd->td_priority;
	if (pri < cpri)
		ctd->td_flags |= TDF_NEEDRESCHED;
	if (KERNEL_PANICKED() || pri >= cpri || cold || TD_IS_INHIBITED(ctd))
		return;
	if (!sched_shouldpreempt(pri, cpri, 0))
		return;
	ctd->td_owepreempt = 1;
}

/*
 * Add a thread to a thread queue.  Select the appropriate runq and add the
 * thread to it.  This is the internal function called when the tdq is
 * predetermined.
 */
void
tdq_add(struct tdq *tdq, struct thread *td, int flags)
{

	TDQ_LOCK_ASSERT(tdq, MA_OWNED);
	THREAD_LOCK_BLOCKED_ASSERT(td, MA_OWNED);
	KASSERT((td->td_inhibitors == 0),
	    ("sched_add: trying to run inhibited thread"));
	KASSERT((TD_CAN_RUN(td) || TD_IS_RUNNING(td)),
	    ("sched_add: bad thread state"));
	KASSERT(td->td_flags & TDF_INMEM,
	    ("sched_add: thread swapped out"));

	if (td->td_priority < tdq->tdq_lowpri)
		tdq->tdq_lowpri = td->td_priority;
	tdq_runq_add(tdq, td, flags);
	tdq_load_add(tdq, td);
}

/*
 * Select the target thread queue and add a thread to it.  Request
 * preemption or IPI a remote processor if required.
 *
 * Requires the thread lock on entry, drops on exit.
 */
void
sched_add(struct thread *td, int flags)
{
	struct tdq *tdq;
#ifdef SMP
	int cpu;
#endif

	KTR_STATE2(KTR_SCHED, "thread", sched_tdname(td), "runq add",
	    "prio:%d", td->td_priority, KTR_ATTR_LINKED,
	    sched_tdname(curthread));
	KTR_POINT1(KTR_SCHED, "thread", sched_tdname(curthread), "wokeup",
	    KTR_ATTR_LINKED, sched_tdname(td));
	SDT_PROBE4(sched, , , enqueue, td, td->td_proc, NULL, 
	    flags & SRQ_PREEMPTED);
	THREAD_LOCK_ASSERT(td, MA_OWNED);
	/*
	 * Recalculate the priority before we select the target cpu or
	 * run-queue.
	 */
	if (PRI_BASE(td->td_pri_class) == PRI_TIMESHARE)
		sched_priority(td);
#ifdef SMP
	/*
	 * Pick the destination cpu and if it isn't ours transfer to the
	 * target cpu.
	 */
	cpu = sched_pickcpu(td, flags);
	tdq = sched_setcpu(td, cpu, flags);
	tdq_add(tdq, td, flags);
	if (cpu != PCPU_GET(cpuid))
		tdq_notify(tdq, td);
	else if (!(flags & SRQ_YIELDING))
		sched_setpreempt(td);
#else
	tdq = TDQ_SELF();
	/*
	 * Now that the thread is moving to the run-queue, set the lock
	 * to the scheduler's lock.
	 */
	if (td->td_lock != TDQ_LOCKPTR(tdq)) {
		TDQ_LOCK(tdq);
		if ((flags & SRQ_HOLD) != 0)
			td->td_lock = TDQ_LOCKPTR(tdq);
		else
			thread_lock_set(td, TDQ_LOCKPTR(tdq));
	}
	tdq_add(tdq, td, flags);
	if (!(flags & SRQ_YIELDING))
		sched_setpreempt(td);
#endif
	if (!(flags & SRQ_HOLDTD))
		thread_unlock(td);
}

/*
 * Remove a thread from a run-queue without running it.  This is used
 * when we're stealing a thread from a remote queue.  Otherwise all threads
 * exit by calling sched_exit_thread() and sched_throw() themselves.
 */
void
sched_rem(struct thread *td)
{
	struct tdq *tdq;

	KTR_STATE1(KTR_SCHED, "thread", sched_tdname(td), "runq rem",
	    "prio:%d", td->td_priority);
	SDT_PROBE3(sched, , , dequeue, td, td->td_proc, NULL);
	tdq = TDQ_CPU(td_get_sched(td)->ts_cpu);
	TDQ_LOCK_ASSERT(tdq, MA_OWNED);
	MPASS(td->td_lock == TDQ_LOCKPTR(tdq));
	KASSERT(TD_ON_RUNQ(td),
	    ("sched_rem: thread not on run queue"));
	tdq_runq_rem(tdq, td);
	tdq_load_rem(tdq, td);
	TD_SET_CAN_RUN(td);
	if (td->td_priority == tdq->tdq_lowpri)
		tdq_setlowpri(tdq, NULL);
}

/*
 * Fetch cpu utilization information.  Updates on demand.
 */
fixpt_t
sched_pctcpu(struct thread *td)
{
	fixpt_t pctcpu;
	struct td_sched *ts;

	pctcpu = 0;
	ts = td_get_sched(td);

	THREAD_LOCK_ASSERT(td, MA_OWNED);
	sched_pctcpu_update(ts, TD_IS_RUNNING(td));
	if (ts->ts_ticks) {
		int rtick;

		/* How many rtick per second ? */
		rtick = min(SCHED_TICK_HZ(ts) / SCHED_TICK_SECS, hz);
		pctcpu = (FSCALE * ((FSCALE * rtick)/hz)) >> FSHIFT;
	}

	return (pctcpu);
}

/*
 * Enforce affinity settings for a thread.  Called after adjustments to
 * cpumask.
 */
void
sched_affinity(struct thread *td)
{
#ifdef SMP
	struct td_sched *ts;

	THREAD_LOCK_ASSERT(td, MA_OWNED);
	ts = td_get_sched(td);
	if (THREAD_CAN_SCHED(td, ts->ts_cpu))
		return;
	if (TD_ON_RUNQ(td)) {
		sched_rem(td);
		sched_add(td, SRQ_BORING | SRQ_HOLDTD);
		return;
	}
	if (!TD_IS_RUNNING(td))
		return;
	/*
	 * Force a switch before returning to userspace.  If the
	 * target thread is not running locally send an ipi to force
	 * the issue.
	 */
	td->td_flags |= TDF_NEEDRESCHED;
	if (td != curthread)
		ipi_cpu(ts->ts_cpu, IPI_PREEMPT);
#endif
}

/*
 * Bind a thread to a target cpu.
 */
void
sched_bind(struct thread *td, int cpu)
{
	struct td_sched *ts;

	THREAD_LOCK_ASSERT(td, MA_OWNED|MA_NOTRECURSED);
	KASSERT(td == curthread, ("sched_bind: can only bind curthread"));
	ts = td_get_sched(td);
	if (ts->ts_flags & TSF_BOUND)
		sched_unbind(td);
	KASSERT(THREAD_CAN_MIGRATE(td), ("%p must be migratable", td));
	ts->ts_flags |= TSF_BOUND;
	sched_pin();
	if (PCPU_GET(cpuid) == cpu)
		return;
	ts->ts_cpu = cpu;
	/* When we return from mi_switch we'll be on the correct cpu. */
	mi_switch(SW_VOL);
	thread_lock(td);
}

/*
 * Release a bound thread.
 */
void
sched_unbind(struct thread *td)
{
	struct td_sched *ts;

	THREAD_LOCK_ASSERT(td, MA_OWNED);
	KASSERT(td == curthread, ("sched_unbind: can only bind curthread"));
	ts = td_get_sched(td);
	if ((ts->ts_flags & TSF_BOUND) == 0)
		return;
	ts->ts_flags &= ~TSF_BOUND;
	sched_unpin();
}

int
sched_is_bound(struct thread *td)
{
	THREAD_LOCK_ASSERT(td, MA_OWNED);
	return (td_get_sched(td)->ts_flags & TSF_BOUND);
}

/*
 * Basic yield call.
 */
void
sched_relinquish(struct thread *td)
{
	thread_lock(td);
	mi_switch(SW_VOL | SWT_RELINQUISH);
}

/*
 * Return the total system load.
 */
int
sched_load(void)
{
#ifdef SMP
	int total;
	int i;

	total = 0;
	CPU_FOREACH(i)
		total += TDQ_CPU(i)->tdq_sysload;
	return (total);
#else
	return (TDQ_SELF()->tdq_sysload);
#endif
}

int
sched_sizeof_proc(void)
{
	return (sizeof(struct proc));
}

int
sched_sizeof_thread(void)
{
	return (sizeof(struct thread) + sizeof(struct td_sched));
}

#ifdef SMP
#define	TDQ_IDLESPIN(tdq)						\
    ((tdq)->tdq_cg != NULL && ((tdq)->tdq_cg->cg_flags & CG_FLAG_THREAD) == 0)
#else
#define	TDQ_IDLESPIN(tdq)	1
#endif

/*
 * The actual idle process.
 */
void
sched_idletd(void *dummy)
{
	struct thread *td;
	struct tdq *tdq;
	int oldswitchcnt, switchcnt;
	int i;

	mtx_assert(&Giant, MA_NOTOWNED);
	td = curthread;
	tdq = TDQ_SELF();
	THREAD_NO_SLEEPING();
	oldswitchcnt = -1;
	for (;;) {
		if (tdq->tdq_load) {
			thread_lock(td);
			mi_switch(SW_VOL | SWT_IDLE);
		}
		switchcnt = tdq->tdq_switchcnt + tdq->tdq_oldswitchcnt;
#ifdef SMP
		if (always_steal || switchcnt != oldswitchcnt) {
			oldswitchcnt = switchcnt;
			if (tdq_idled(tdq) == 0)
				continue;
		}
		switchcnt = tdq->tdq_switchcnt + tdq->tdq_oldswitchcnt;
#else
		oldswitchcnt = switchcnt;
#endif
		/*
		 * If we're switching very frequently, spin while checking
		 * for load rather than entering a low power state that 
		 * may require an IPI.  However, don't do any busy
		 * loops while on SMT machines as this simply steals
		 * cycles from cores doing useful work.
		 */
		if (TDQ_IDLESPIN(tdq) && switchcnt > sched_idlespinthresh) {
			for (i = 0; i < sched_idlespins; i++) {
				if (tdq->tdq_load)
					break;
				cpu_spinwait();
			}
		}

		/* If there was context switch during spin, restart it. */
		switchcnt = tdq->tdq_switchcnt + tdq->tdq_oldswitchcnt;
		if (tdq->tdq_load != 0 || switchcnt != oldswitchcnt)
			continue;

		/* Run main MD idle handler. */
		tdq->tdq_cpu_idle = 1;
		/*
		 * Make sure that tdq_cpu_idle update is globally visible
		 * before cpu_idle() read tdq_load.  The order is important
		 * to avoid race with tdq_notify.
		 */
		atomic_thread_fence_seq_cst();
		/*
		 * Checking for again after the fence picks up assigned
		 * threads often enough to make it worthwhile to do so in
		 * order to avoid calling cpu_idle().
		 */
		if (tdq->tdq_load != 0) {
			tdq->tdq_cpu_idle = 0;
			continue;
		}
		cpu_idle(switchcnt * 4 > sched_idlespinthresh);
		tdq->tdq_cpu_idle = 0;

		/*
		 * Account thread-less hardware interrupts and
		 * other wakeup reasons equal to context switches.
		 */
		switchcnt = tdq->tdq_switchcnt + tdq->tdq_oldswitchcnt;
		if (switchcnt != oldswitchcnt)
			continue;
		tdq->tdq_switchcnt++;
		oldswitchcnt++;
	}
}

/*
 * A CPU is entering for the first time or a thread is exiting.
 */
void
sched_throw(struct thread *td)
{
	struct thread *newtd;
	struct tdq *tdq;

	if (__predict_false(td == NULL)) {
#ifdef SMP
		PCPU_SET(sched, DPCPU_PTR(tdq));
#endif
		/* Correct spinlock nesting and acquire the correct lock. */
		tdq = TDQ_SELF();
		TDQ_LOCK(tdq);
		spinlock_exit();
		PCPU_SET(switchtime, cpu_ticks());
		PCPU_SET(switchticks, ticks);
		PCPU_GET(idlethread)->td_lock = TDQ_LOCKPTR(tdq);
	} else {
		tdq = TDQ_SELF();
		THREAD_LOCK_ASSERT(td, MA_OWNED);
		THREAD_LOCKPTR_ASSERT(td, TDQ_LOCKPTR(tdq));
		tdq_load_rem(tdq, td);
		td->td_lastcpu = td->td_oncpu;
		td->td_oncpu = NOCPU;
		thread_lock_block(td);
	}
	newtd = choosethread();
	spinlock_enter();
	TDQ_UNLOCK(tdq);
	KASSERT(curthread->td_md.md_spinlock_count == 1,
	    ("invalid count %d", curthread->td_md.md_spinlock_count));
	/* doesn't return */
	if (__predict_false(td == NULL))
		cpu_throw(td, newtd);		/* doesn't return */
	else
		cpu_switch(td, newtd, TDQ_LOCKPTR(tdq));
}

/*
 * This is called from fork_exit().  Just acquire the correct locks and
 * let fork do the rest of the work.
 */
void
sched_fork_exit(struct thread *td)
{
	struct tdq *tdq;
	int cpuid;

	/*
	 * Finish setting up thread glue so that it begins execution in a
	 * non-nested critical section with the scheduler lock held.
	 */
	KASSERT(curthread->td_md.md_spinlock_count == 1,
	    ("invalid count %d", curthread->td_md.md_spinlock_count));
	cpuid = PCPU_GET(cpuid);
	tdq = TDQ_SELF();
	TDQ_LOCK(tdq);
	spinlock_exit();
	MPASS(td->td_lock == TDQ_LOCKPTR(tdq));
	td->td_oncpu = cpuid;
	KTR_STATE1(KTR_SCHED, "thread", sched_tdname(td), "running",
	    "prio:%d", td->td_priority);
	SDT_PROBE0(sched, , , on__cpu);
}

/*
 * Create on first use to catch odd startup conditons.
 */
char *
sched_tdname(struct thread *td)
{
#ifdef KTR
	struct td_sched *ts;

	ts = td_get_sched(td);
	if (ts->ts_name[0] == '\0')
		snprintf(ts->ts_name, sizeof(ts->ts_name),
		    "%s tid %d", td->td_name, td->td_tid);
	return (ts->ts_name);
#else
	return (td->td_name);
#endif
}

#ifdef KTR
void
sched_clear_tdname(struct thread *td)
{
	struct td_sched *ts;

	ts = td_get_sched(td);
	ts->ts_name[0] = '\0';
}
#endif

#ifdef SMP

/*
 * Build the CPU topology dump string. Is recursively called to collect
 * the topology tree.
 */
static int
sysctl_kern_sched_topology_spec_internal(struct sbuf *sb, struct cpu_group *cg,
    int indent)
{
	char cpusetbuf[CPUSETBUFSIZ];
	int i, first;

	sbuf_printf(sb, "%*s<group level=\"%d\" cache-level=\"%d\">\n", indent,
	    "", 1 + indent / 2, cg->cg_level);
	sbuf_printf(sb, "%*s <cpu count=\"%d\" mask=\"%s\">", indent, "",
	    cg->cg_count, cpusetobj_strprint(cpusetbuf, &cg->cg_mask));
	first = TRUE;
	for (i = cg->cg_first; i <= cg->cg_last; i++) {
		if (CPU_ISSET(i, &cg->cg_mask)) {
			if (!first)
				sbuf_printf(sb, ", ");
			else
				first = FALSE;
			sbuf_printf(sb, "%d", i);
		}
	}
	sbuf_printf(sb, "</cpu>\n");

	if (cg->cg_flags != 0) {
		sbuf_printf(sb, "%*s <flags>", indent, "");
		if ((cg->cg_flags & CG_FLAG_HTT) != 0)
			sbuf_printf(sb, "<flag name=\"HTT\">HTT group</flag>");
		if ((cg->cg_flags & CG_FLAG_THREAD) != 0)
			sbuf_printf(sb, "<flag name=\"THREAD\">THREAD group</flag>");
		if ((cg->cg_flags & CG_FLAG_SMT) != 0)
			sbuf_printf(sb, "<flag name=\"SMT\">SMT group</flag>");
		sbuf_printf(sb, "</flags>\n");
	}

	if (cg->cg_children > 0) {
		sbuf_printf(sb, "%*s <children>\n", indent, "");
		for (i = 0; i < cg->cg_children; i++)
			sysctl_kern_sched_topology_spec_internal(sb, 
			    &cg->cg_child[i], indent+2);
		sbuf_printf(sb, "%*s </children>\n", indent, "");
	}
	sbuf_printf(sb, "%*s</group>\n", indent, "");
	return (0);
}

/*
 * Sysctl handler for retrieving topology dump. It's a wrapper for
 * the recursive sysctl_kern_smp_topology_spec_internal().
 */
static int
sysctl_kern_sched_topology_spec(SYSCTL_HANDLER_ARGS)
{
	struct sbuf *topo;
	int err;

	KASSERT(cpu_top != NULL, ("cpu_top isn't initialized"));

	topo = sbuf_new_for_sysctl(NULL, NULL, 512, req);
	if (topo == NULL)
		return (ENOMEM);

	sbuf_printf(topo, "<groups>\n");
	err = sysctl_kern_sched_topology_spec_internal(topo, cpu_top, 1);
	sbuf_printf(topo, "</groups>\n");

	if (err == 0) {
		err = sbuf_finish(topo);
	}
	sbuf_delete(topo);
	return (err);
}

#endif

static int
sysctl_kern_quantum(SYSCTL_HANDLER_ARGS)
{
	int error, new_val, period;

	period = 1000000 / realstathz;
	new_val = period * sched_slice;
	error = sysctl_handle_int(oidp, &new_val, 0, req);
	if (error != 0 || req->newptr == NULL)
		return (error);
	if (new_val <= 0)
		return (EINVAL);
	sched_slice = imax(1, (new_val + period / 2) / period);
	sched_slice_min = sched_slice / SCHED_SLICE_MIN_DIVISOR;
	hogticks = imax(1, (2 * hz * sched_slice + realstathz / 2) /
	    realstathz);
	return (0);
}

SYSCTL_NODE(_kern, OID_AUTO, sched, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
    "Scheduler");
SYSCTL_STRING(_kern_sched, OID_AUTO, name, CTLFLAG_RD, "ULE", 0,
    "Scheduler name");
SYSCTL_PROC(_kern_sched, OID_AUTO, quantum,
    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, NULL, 0,
    sysctl_kern_quantum, "I",
    "Quantum for timeshare threads in microseconds");
SYSCTL_INT(_kern_sched, OID_AUTO, slice, CTLFLAG_RW, &sched_slice, 0,
    "Quantum for timeshare threads in stathz ticks");
SYSCTL_INT(_kern_sched, OID_AUTO, interact, CTLFLAG_RW, &sched_interact, 0,
    "Interactivity score threshold");
SYSCTL_INT(_kern_sched, OID_AUTO, preempt_thresh, CTLFLAG_RW,
    &preempt_thresh, 0,
    "Maximal (lowest) priority for preemption");
SYSCTL_INT(_kern_sched, OID_AUTO, static_boost, CTLFLAG_RW, &static_boost, 0,
    "Assign static kernel priorities to sleeping threads");
SYSCTL_INT(_kern_sched, OID_AUTO, idlespins, CTLFLAG_RW, &sched_idlespins, 0,
    "Number of times idle thread will spin waiting for new work");
SYSCTL_INT(_kern_sched, OID_AUTO, idlespinthresh, CTLFLAG_RW,
    &sched_idlespinthresh, 0,
    "Threshold before we will permit idle thread spinning");
#ifdef SMP
SYSCTL_INT(_kern_sched, OID_AUTO, affinity, CTLFLAG_RW, &affinity, 0,
    "Number of hz ticks to keep thread affinity for");
SYSCTL_INT(_kern_sched, OID_AUTO, balance, CTLFLAG_RW, &rebalance, 0,
    "Enables the long-term load balancer");
SYSCTL_INT(_kern_sched, OID_AUTO, balance_interval, CTLFLAG_RW,
    &balance_interval, 0,
    "Average period in stathz ticks to run the long-term balancer");
SYSCTL_INT(_kern_sched, OID_AUTO, steal_idle, CTLFLAG_RW, &steal_idle, 0,
    "Attempts to steal work from other cores before idling");
SYSCTL_INT(_kern_sched, OID_AUTO, steal_thresh, CTLFLAG_RW, &steal_thresh, 0,
    "Minimum load on remote CPU before we'll steal");
SYSCTL_INT(_kern_sched, OID_AUTO, trysteal_limit, CTLFLAG_RW, &trysteal_limit,
    0, "Topological distance limit for stealing threads in sched_switch()");
SYSCTL_INT(_kern_sched, OID_AUTO, always_steal, CTLFLAG_RW, &always_steal, 0,
    "Always run the stealer from the idle thread");
SYSCTL_PROC(_kern_sched, OID_AUTO, topology_spec, CTLTYPE_STRING |
    CTLFLAG_MPSAFE | CTLFLAG_RD, NULL, 0, sysctl_kern_sched_topology_spec, "A",
    "XML dump of detected CPU topology");
#endif

/* ps compat.  All cpu percentages from ULE are weighted. */
static int ccpu = 0;
SYSCTL_INT(_kern, OID_AUTO, ccpu, CTLFLAG_RD, &ccpu, 0,
    "Decay factor used for updating %CPU in 4BSD scheduler");