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
|
/*-
* Copyright (c) 2013, Anish Gupta (akgupt3@gmail.com)
* 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.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/smp.h>
#include <sys/kernel.h>
#include <sys/malloc.h>
#include <sys/pcpu.h>
#include <sys/proc.h>
#include <sys/sysctl.h>
#include <vm/vm.h>
#include <vm/pmap.h>
#include <machine/cpufunc.h>
#include <machine/psl.h>
#include <machine/pmap.h>
#include <machine/md_var.h>
#include <machine/specialreg.h>
#include <machine/smp.h>
#include <machine/vmm.h>
#include <machine/vmm_dev.h>
#include <machine/vmm_instruction_emul.h>
#include "vmm_lapic.h"
#include "vmm_stat.h"
#include "vmm_ktr.h"
#include "vmm_ioport.h"
#include "vatpic.h"
#include "vlapic.h"
#include "vlapic_priv.h"
#include "x86.h"
#include "vmcb.h"
#include "svm.h"
#include "svm_softc.h"
#include "svm_msr.h"
#include "npt.h"
SYSCTL_DECL(_hw_vmm);
SYSCTL_NODE(_hw_vmm, OID_AUTO, svm, CTLFLAG_RW, NULL, NULL);
/*
* SVM CPUID function 0x8000_000A, edx bit decoding.
*/
#define AMD_CPUID_SVM_NP BIT(0) /* Nested paging or RVI */
#define AMD_CPUID_SVM_LBR BIT(1) /* Last branch virtualization */
#define AMD_CPUID_SVM_SVML BIT(2) /* SVM lock */
#define AMD_CPUID_SVM_NRIP_SAVE BIT(3) /* Next RIP is saved */
#define AMD_CPUID_SVM_TSC_RATE BIT(4) /* TSC rate control. */
#define AMD_CPUID_SVM_VMCB_CLEAN BIT(5) /* VMCB state caching */
#define AMD_CPUID_SVM_FLUSH_BY_ASID BIT(6) /* Flush by ASID */
#define AMD_CPUID_SVM_DECODE_ASSIST BIT(7) /* Decode assist */
#define AMD_CPUID_SVM_PAUSE_INC BIT(10) /* Pause intercept filter. */
#define AMD_CPUID_SVM_PAUSE_FTH BIT(12) /* Pause filter threshold */
#define VMCB_CACHE_DEFAULT (VMCB_CACHE_ASID | \
VMCB_CACHE_IOPM | \
VMCB_CACHE_I | \
VMCB_CACHE_TPR | \
VMCB_CACHE_CR2 | \
VMCB_CACHE_CR | \
VMCB_CACHE_DT | \
VMCB_CACHE_SEG | \
VMCB_CACHE_NP)
static uint32_t vmcb_clean = VMCB_CACHE_DEFAULT;
SYSCTL_INT(_hw_vmm_svm, OID_AUTO, vmcb_clean, CTLFLAG_RDTUN, &vmcb_clean,
0, NULL);
static MALLOC_DEFINE(M_SVM, "svm", "svm");
static MALLOC_DEFINE(M_SVM_VLAPIC, "svm-vlapic", "svm-vlapic");
/* Per-CPU context area. */
extern struct pcpu __pcpu[];
static uint32_t svm_feature; /* AMD SVM features. */
SYSCTL_UINT(_hw_vmm_svm, OID_AUTO, features, CTLFLAG_RD, &svm_feature, 0,
"SVM features advertised by CPUID.8000000AH:EDX");
static int disable_npf_assist;
SYSCTL_INT(_hw_vmm_svm, OID_AUTO, disable_npf_assist, CTLFLAG_RWTUN,
&disable_npf_assist, 0, NULL);
/* Maximum ASIDs supported by the processor */
static uint32_t nasid;
SYSCTL_UINT(_hw_vmm_svm, OID_AUTO, num_asids, CTLFLAG_RD, &nasid, 0,
"Number of ASIDs supported by this processor");
/* Current ASID generation for each host cpu */
static struct asid asid[MAXCPU];
/*
* SVM host state saved area of size 4KB for each core.
*/
static uint8_t hsave[MAXCPU][PAGE_SIZE] __aligned(PAGE_SIZE);
static VMM_STAT_AMD(VCPU_EXITINTINFO, "VM exits during event delivery");
static VMM_STAT_AMD(VCPU_INTINFO_INJECTED, "Events pending at VM entry");
static VMM_STAT_AMD(VMEXIT_VINTR, "VM exits due to interrupt window");
static int svm_setreg(void *arg, int vcpu, int ident, uint64_t val);
static __inline int
flush_by_asid(void)
{
return (svm_feature & AMD_CPUID_SVM_FLUSH_BY_ASID);
}
static __inline int
decode_assist(void)
{
return (svm_feature & AMD_CPUID_SVM_DECODE_ASSIST);
}
static void
svm_disable(void *arg __unused)
{
uint64_t efer;
efer = rdmsr(MSR_EFER);
efer &= ~EFER_SVM;
wrmsr(MSR_EFER, efer);
}
/*
* Disable SVM on all CPUs.
*/
static int
svm_cleanup(void)
{
smp_rendezvous(NULL, svm_disable, NULL, NULL);
return (0);
}
/*
* Verify that all the features required by bhyve are available.
*/
static int
check_svm_features(void)
{
u_int regs[4];
/* CPUID Fn8000_000A is for SVM */
do_cpuid(0x8000000A, regs);
svm_feature = regs[3];
nasid = regs[1];
KASSERT(nasid > 1, ("Insufficient ASIDs for guests: %#x", nasid));
/* bhyve requires the Nested Paging feature */
if (!(svm_feature & AMD_CPUID_SVM_NP)) {
printf("SVM: Nested Paging feature not available.\n");
return (ENXIO);
}
/* bhyve requires the NRIP Save feature */
if (!(svm_feature & AMD_CPUID_SVM_NRIP_SAVE)) {
printf("SVM: NRIP Save feature not available.\n");
return (ENXIO);
}
return (0);
}
static void
svm_enable(void *arg __unused)
{
uint64_t efer;
efer = rdmsr(MSR_EFER);
efer |= EFER_SVM;
wrmsr(MSR_EFER, efer);
wrmsr(MSR_VM_HSAVE_PA, vtophys(hsave[curcpu]));
}
/*
* Return 1 if SVM is enabled on this processor and 0 otherwise.
*/
static int
svm_available(void)
{
uint64_t msr;
/* Section 15.4 Enabling SVM from APM2. */
if ((amd_feature2 & AMDID2_SVM) == 0) {
printf("SVM: not available.\n");
return (0);
}
msr = rdmsr(MSR_VM_CR);
if ((msr & VM_CR_SVMDIS) != 0) {
printf("SVM: disabled by BIOS.\n");
return (0);
}
return (1);
}
static int
svm_init(int ipinum)
{
int error, cpu;
if (!svm_available())
return (ENXIO);
error = check_svm_features();
if (error)
return (error);
vmcb_clean &= VMCB_CACHE_DEFAULT;
for (cpu = 0; cpu < MAXCPU; cpu++) {
/*
* Initialize the host ASIDs to their "highest" valid values.
*
* The next ASID allocation will rollover both 'gen' and 'num'
* and start off the sequence at {1,1}.
*/
asid[cpu].gen = ~0UL;
asid[cpu].num = nasid - 1;
}
svm_msr_init();
svm_npt_init(ipinum);
/* Enable SVM on all CPUs */
smp_rendezvous(NULL, svm_enable, NULL, NULL);
return (0);
}
static void
svm_restore(void)
{
svm_enable(NULL);
}
/* Pentium compatible MSRs */
#define MSR_PENTIUM_START 0
#define MSR_PENTIUM_END 0x1FFF
/* AMD 6th generation and Intel compatible MSRs */
#define MSR_AMD6TH_START 0xC0000000UL
#define MSR_AMD6TH_END 0xC0001FFFUL
/* AMD 7th and 8th generation compatible MSRs */
#define MSR_AMD7TH_START 0xC0010000UL
#define MSR_AMD7TH_END 0xC0011FFFUL
/*
* Get the index and bit position for a MSR in permission bitmap.
* Two bits are used for each MSR: lower bit for read and higher bit for write.
*/
static int
svm_msr_index(uint64_t msr, int *index, int *bit)
{
uint32_t base, off;
*index = -1;
*bit = (msr % 4) * 2;
base = 0;
if (msr >= MSR_PENTIUM_START && msr <= MSR_PENTIUM_END) {
*index = msr / 4;
return (0);
}
base += (MSR_PENTIUM_END - MSR_PENTIUM_START + 1);
if (msr >= MSR_AMD6TH_START && msr <= MSR_AMD6TH_END) {
off = (msr - MSR_AMD6TH_START);
*index = (off + base) / 4;
return (0);
}
base += (MSR_AMD6TH_END - MSR_AMD6TH_START + 1);
if (msr >= MSR_AMD7TH_START && msr <= MSR_AMD7TH_END) {
off = (msr - MSR_AMD7TH_START);
*index = (off + base) / 4;
return (0);
}
return (EINVAL);
}
/*
* Allow vcpu to read or write the 'msr' without trapping into the hypervisor.
*/
static void
svm_msr_perm(uint8_t *perm_bitmap, uint64_t msr, bool read, bool write)
{
int index, bit, error;
error = svm_msr_index(msr, &index, &bit);
KASSERT(error == 0, ("%s: invalid msr %#lx", __func__, msr));
KASSERT(index >= 0 && index < SVM_MSR_BITMAP_SIZE,
("%s: invalid index %d for msr %#lx", __func__, index, msr));
KASSERT(bit >= 0 && bit <= 6, ("%s: invalid bit position %d "
"msr %#lx", __func__, bit, msr));
if (read)
perm_bitmap[index] &= ~(1UL << bit);
if (write)
perm_bitmap[index] &= ~(2UL << bit);
}
static void
svm_msr_rw_ok(uint8_t *perm_bitmap, uint64_t msr)
{
svm_msr_perm(perm_bitmap, msr, true, true);
}
static void
svm_msr_rd_ok(uint8_t *perm_bitmap, uint64_t msr)
{
svm_msr_perm(perm_bitmap, msr, true, false);
}
static __inline int
svm_get_intercept(struct svm_softc *sc, int vcpu, int idx, uint32_t bitmask)
{
struct vmcb_ctrl *ctrl;
KASSERT(idx >=0 && idx < 5, ("invalid intercept index %d", idx));
ctrl = svm_get_vmcb_ctrl(sc, vcpu);
return (ctrl->intercept[idx] & bitmask ? 1 : 0);
}
static __inline void
svm_set_intercept(struct svm_softc *sc, int vcpu, int idx, uint32_t bitmask,
int enabled)
{
struct vmcb_ctrl *ctrl;
uint32_t oldval;
KASSERT(idx >=0 && idx < 5, ("invalid intercept index %d", idx));
ctrl = svm_get_vmcb_ctrl(sc, vcpu);
oldval = ctrl->intercept[idx];
if (enabled)
ctrl->intercept[idx] |= bitmask;
else
ctrl->intercept[idx] &= ~bitmask;
if (ctrl->intercept[idx] != oldval) {
svm_set_dirty(sc, vcpu, VMCB_CACHE_I);
VCPU_CTR3(sc->vm, vcpu, "intercept[%d] modified "
"from %#x to %#x", idx, oldval, ctrl->intercept[idx]);
}
}
static __inline void
svm_disable_intercept(struct svm_softc *sc, int vcpu, int off, uint32_t bitmask)
{
svm_set_intercept(sc, vcpu, off, bitmask, 0);
}
static __inline void
svm_enable_intercept(struct svm_softc *sc, int vcpu, int off, uint32_t bitmask)
{
svm_set_intercept(sc, vcpu, off, bitmask, 1);
}
static void
vmcb_init(struct svm_softc *sc, int vcpu, uint64_t iopm_base_pa,
uint64_t msrpm_base_pa, uint64_t np_pml4)
{
struct vmcb_ctrl *ctrl;
struct vmcb_state *state;
uint32_t mask;
int n;
ctrl = svm_get_vmcb_ctrl(sc, vcpu);
state = svm_get_vmcb_state(sc, vcpu);
ctrl->iopm_base_pa = iopm_base_pa;
ctrl->msrpm_base_pa = msrpm_base_pa;
/* Enable nested paging */
ctrl->np_enable = 1;
ctrl->n_cr3 = np_pml4;
/*
* Intercept accesses to the control registers that are not shadowed
* in the VMCB - i.e. all except cr0, cr2, cr3, cr4 and cr8.
*/
for (n = 0; n < 16; n++) {
mask = (BIT(n) << 16) | BIT(n);
if (n == 0 || n == 2 || n == 3 || n == 4 || n == 8)
svm_disable_intercept(sc, vcpu, VMCB_CR_INTCPT, mask);
else
svm_enable_intercept(sc, vcpu, VMCB_CR_INTCPT, mask);
}
/*
* Intercept everything when tracing guest exceptions otherwise
* just intercept machine check exception.
*/
if (vcpu_trace_exceptions(sc->vm, vcpu)) {
for (n = 0; n < 32; n++) {
/*
* Skip unimplemented vectors in the exception bitmap.
*/
if (n == 2 || n == 9) {
continue;
}
svm_enable_intercept(sc, vcpu, VMCB_EXC_INTCPT, BIT(n));
}
} else {
svm_enable_intercept(sc, vcpu, VMCB_EXC_INTCPT, BIT(IDT_MC));
}
/* Intercept various events (for e.g. I/O, MSR and CPUID accesses) */
svm_enable_intercept(sc, vcpu, VMCB_CTRL1_INTCPT, VMCB_INTCPT_IO);
svm_enable_intercept(sc, vcpu, VMCB_CTRL1_INTCPT, VMCB_INTCPT_MSR);
svm_enable_intercept(sc, vcpu, VMCB_CTRL1_INTCPT, VMCB_INTCPT_CPUID);
svm_enable_intercept(sc, vcpu, VMCB_CTRL1_INTCPT, VMCB_INTCPT_INTR);
svm_enable_intercept(sc, vcpu, VMCB_CTRL1_INTCPT, VMCB_INTCPT_INIT);
svm_enable_intercept(sc, vcpu, VMCB_CTRL1_INTCPT, VMCB_INTCPT_NMI);
svm_enable_intercept(sc, vcpu, VMCB_CTRL1_INTCPT, VMCB_INTCPT_SMI);
svm_enable_intercept(sc, vcpu, VMCB_CTRL1_INTCPT, VMCB_INTCPT_SHUTDOWN);
svm_enable_intercept(sc, vcpu, VMCB_CTRL1_INTCPT,
VMCB_INTCPT_FERR_FREEZE);
svm_enable_intercept(sc, vcpu, VMCB_CTRL2_INTCPT, VMCB_INTCPT_MONITOR);
svm_enable_intercept(sc, vcpu, VMCB_CTRL2_INTCPT, VMCB_INTCPT_MWAIT);
/*
* From section "Canonicalization and Consistency Checks" in APMv2
* the VMRUN intercept bit must be set to pass the consistency check.
*/
svm_enable_intercept(sc, vcpu, VMCB_CTRL2_INTCPT, VMCB_INTCPT_VMRUN);
/*
* The ASID will be set to a non-zero value just before VMRUN.
*/
ctrl->asid = 0;
/*
* Section 15.21.1, Interrupt Masking in EFLAGS
* Section 15.21.2, Virtualizing APIC.TPR
*
* This must be set for %rflag and %cr8 isolation of guest and host.
*/
ctrl->v_intr_masking = 1;
/* Enable Last Branch Record aka LBR for debugging */
ctrl->lbr_virt_en = 1;
state->dbgctl = BIT(0);
/* EFER_SVM must always be set when the guest is executing */
state->efer = EFER_SVM;
/* Set up the PAT to power-on state */
state->g_pat = PAT_VALUE(0, PAT_WRITE_BACK) |
PAT_VALUE(1, PAT_WRITE_THROUGH) |
PAT_VALUE(2, PAT_UNCACHED) |
PAT_VALUE(3, PAT_UNCACHEABLE) |
PAT_VALUE(4, PAT_WRITE_BACK) |
PAT_VALUE(5, PAT_WRITE_THROUGH) |
PAT_VALUE(6, PAT_UNCACHED) |
PAT_VALUE(7, PAT_UNCACHEABLE);
}
/*
* Initialize a virtual machine.
*/
static void *
svm_vminit(struct vm *vm, pmap_t pmap)
{
struct svm_softc *svm_sc;
struct svm_vcpu *vcpu;
vm_paddr_t msrpm_pa, iopm_pa, pml4_pa;
int i;
svm_sc = malloc(sizeof (struct svm_softc), M_SVM, M_WAITOK | M_ZERO);
svm_sc->vm = vm;
svm_sc->nptp = (vm_offset_t)vtophys(pmap->pm_pml4);
/*
* Intercept read and write accesses to all MSRs.
*/
memset(svm_sc->msr_bitmap, 0xFF, sizeof(svm_sc->msr_bitmap));
/*
* Access to the following MSRs is redirected to the VMCB when the
* guest is executing. Therefore it is safe to allow the guest to
* read/write these MSRs directly without hypervisor involvement.
*/
svm_msr_rw_ok(svm_sc->msr_bitmap, MSR_GSBASE);
svm_msr_rw_ok(svm_sc->msr_bitmap, MSR_FSBASE);
svm_msr_rw_ok(svm_sc->msr_bitmap, MSR_KGSBASE);
svm_msr_rw_ok(svm_sc->msr_bitmap, MSR_STAR);
svm_msr_rw_ok(svm_sc->msr_bitmap, MSR_LSTAR);
svm_msr_rw_ok(svm_sc->msr_bitmap, MSR_CSTAR);
svm_msr_rw_ok(svm_sc->msr_bitmap, MSR_SF_MASK);
svm_msr_rw_ok(svm_sc->msr_bitmap, MSR_SYSENTER_CS_MSR);
svm_msr_rw_ok(svm_sc->msr_bitmap, MSR_SYSENTER_ESP_MSR);
svm_msr_rw_ok(svm_sc->msr_bitmap, MSR_SYSENTER_EIP_MSR);
svm_msr_rw_ok(svm_sc->msr_bitmap, MSR_PAT);
svm_msr_rd_ok(svm_sc->msr_bitmap, MSR_TSC);
/*
* Intercept writes to make sure that the EFER_SVM bit is not cleared.
*/
svm_msr_rd_ok(svm_sc->msr_bitmap, MSR_EFER);
/* Intercept access to all I/O ports. */
memset(svm_sc->iopm_bitmap, 0xFF, sizeof(svm_sc->iopm_bitmap));
iopm_pa = vtophys(svm_sc->iopm_bitmap);
msrpm_pa = vtophys(svm_sc->msr_bitmap);
pml4_pa = svm_sc->nptp;
for (i = 0; i < VM_MAXCPU; i++) {
vcpu = svm_get_vcpu(svm_sc, i);
vcpu->lastcpu = NOCPU;
vcpu->vmcb_pa = vtophys(&vcpu->vmcb);
vmcb_init(svm_sc, i, iopm_pa, msrpm_pa, pml4_pa);
svm_msr_guest_init(svm_sc, i);
}
return (svm_sc);
}
static int
svm_cpl(struct vmcb_state *state)
{
/*
* From APMv2:
* "Retrieve the CPL from the CPL field in the VMCB, not
* from any segment DPL"
*/
return (state->cpl);
}
static enum vm_cpu_mode
svm_vcpu_mode(struct vmcb *vmcb)
{
struct vmcb_segment seg;
struct vmcb_state *state;
int error;
state = &vmcb->state;
if (state->efer & EFER_LMA) {
error = vmcb_seg(vmcb, VM_REG_GUEST_CS, &seg);
KASSERT(error == 0, ("%s: vmcb_seg(cs) error %d", __func__,
error));
/*
* Section 4.8.1 for APM2, check if Code Segment has
* Long attribute set in descriptor.
*/
if (seg.attrib & VMCB_CS_ATTRIB_L)
return (CPU_MODE_64BIT);
else
return (CPU_MODE_COMPATIBILITY);
} else if (state->cr0 & CR0_PE) {
return (CPU_MODE_PROTECTED);
} else {
return (CPU_MODE_REAL);
}
}
static enum vm_paging_mode
svm_paging_mode(uint64_t cr0, uint64_t cr4, uint64_t efer)
{
if ((cr0 & CR0_PG) == 0)
return (PAGING_MODE_FLAT);
if ((cr4 & CR4_PAE) == 0)
return (PAGING_MODE_32);
if (efer & EFER_LME)
return (PAGING_MODE_64);
else
return (PAGING_MODE_PAE);
}
/*
* ins/outs utility routines
*/
static uint64_t
svm_inout_str_index(struct svm_regctx *regs, int in)
{
uint64_t val;
val = in ? regs->sctx_rdi : regs->sctx_rsi;
return (val);
}
static uint64_t
svm_inout_str_count(struct svm_regctx *regs, int rep)
{
uint64_t val;
val = rep ? regs->sctx_rcx : 1;
return (val);
}
static void
svm_inout_str_seginfo(struct svm_softc *svm_sc, int vcpu, int64_t info1,
int in, struct vm_inout_str *vis)
{
int error, s;
if (in) {
vis->seg_name = VM_REG_GUEST_ES;
} else {
/* The segment field has standard encoding */
s = (info1 >> 10) & 0x7;
vis->seg_name = vm_segment_name(s);
}
error = vmcb_getdesc(svm_sc, vcpu, vis->seg_name, &vis->seg_desc);
KASSERT(error == 0, ("%s: svm_getdesc error %d", __func__, error));
}
static int
svm_inout_str_addrsize(uint64_t info1)
{
uint32_t size;
size = (info1 >> 7) & 0x7;
switch (size) {
case 1:
return (2); /* 16 bit */
case 2:
return (4); /* 32 bit */
case 4:
return (8); /* 64 bit */
default:
panic("%s: invalid size encoding %d", __func__, size);
}
}
static void
svm_paging_info(struct vmcb *vmcb, struct vm_guest_paging *paging)
{
struct vmcb_state *state;
state = &vmcb->state;
paging->cr3 = state->cr3;
paging->cpl = svm_cpl(state);
paging->cpu_mode = svm_vcpu_mode(vmcb);
paging->paging_mode = svm_paging_mode(state->cr0, state->cr4,
state->efer);
}
#define UNHANDLED 0
/*
* Handle guest I/O intercept.
*/
static int
svm_handle_io(struct svm_softc *svm_sc, int vcpu, struct vm_exit *vmexit)
{
struct vmcb_ctrl *ctrl;
struct vmcb_state *state;
struct svm_regctx *regs;
struct vm_inout_str *vis;
uint64_t info1;
int inout_string;
state = svm_get_vmcb_state(svm_sc, vcpu);
ctrl = svm_get_vmcb_ctrl(svm_sc, vcpu);
regs = svm_get_guest_regctx(svm_sc, vcpu);
info1 = ctrl->exitinfo1;
inout_string = info1 & BIT(2) ? 1 : 0;
/*
* The effective segment number in EXITINFO1[12:10] is populated
* only if the processor has the DecodeAssist capability.
*
* XXX this is not specified explicitly in APMv2 but can be verified
* empirically.
*/
if (inout_string && !decode_assist())
return (UNHANDLED);
vmexit->exitcode = VM_EXITCODE_INOUT;
vmexit->u.inout.in = (info1 & BIT(0)) ? 1 : 0;
vmexit->u.inout.string = inout_string;
vmexit->u.inout.rep = (info1 & BIT(3)) ? 1 : 0;
vmexit->u.inout.bytes = (info1 >> 4) & 0x7;
vmexit->u.inout.port = (uint16_t)(info1 >> 16);
vmexit->u.inout.eax = (uint32_t)(state->rax);
if (inout_string) {
vmexit->exitcode = VM_EXITCODE_INOUT_STR;
vis = &vmexit->u.inout_str;
svm_paging_info(svm_get_vmcb(svm_sc, vcpu), &vis->paging);
vis->rflags = state->rflags;
vis->cr0 = state->cr0;
vis->index = svm_inout_str_index(regs, vmexit->u.inout.in);
vis->count = svm_inout_str_count(regs, vmexit->u.inout.rep);
vis->addrsize = svm_inout_str_addrsize(info1);
svm_inout_str_seginfo(svm_sc, vcpu, info1,
vmexit->u.inout.in, vis);
}
return (UNHANDLED);
}
static int
npf_fault_type(uint64_t exitinfo1)
{
if (exitinfo1 & VMCB_NPF_INFO1_W)
return (VM_PROT_WRITE);
else if (exitinfo1 & VMCB_NPF_INFO1_ID)
return (VM_PROT_EXECUTE);
else
return (VM_PROT_READ);
}
static bool
svm_npf_emul_fault(uint64_t exitinfo1)
{
if (exitinfo1 & VMCB_NPF_INFO1_ID) {
return (false);
}
if (exitinfo1 & VMCB_NPF_INFO1_GPT) {
return (false);
}
if ((exitinfo1 & VMCB_NPF_INFO1_GPA) == 0) {
return (false);
}
return (true);
}
static void
svm_handle_inst_emul(struct vmcb *vmcb, uint64_t gpa, struct vm_exit *vmexit)
{
struct vm_guest_paging *paging;
struct vmcb_segment seg;
struct vmcb_ctrl *ctrl;
char *inst_bytes;
int error, inst_len;
ctrl = &vmcb->ctrl;
paging = &vmexit->u.inst_emul.paging;
vmexit->exitcode = VM_EXITCODE_INST_EMUL;
vmexit->u.inst_emul.gpa = gpa;
vmexit->u.inst_emul.gla = VIE_INVALID_GLA;
svm_paging_info(vmcb, paging);
error = vmcb_seg(vmcb, VM_REG_GUEST_CS, &seg);
KASSERT(error == 0, ("%s: vmcb_seg(CS) error %d", __func__, error));
switch(paging->cpu_mode) {
case CPU_MODE_PROTECTED:
case CPU_MODE_COMPATIBILITY:
/*
* Section 4.8.1 of APM2, Default Operand Size or D bit.
*/
vmexit->u.inst_emul.cs_d = (seg.attrib & VMCB_CS_ATTRIB_D) ?
1 : 0;
break;
default:
vmexit->u.inst_emul.cs_d = 0;
break;
}
/*
* Copy the instruction bytes into 'vie' if available.
*/
if (decode_assist() && !disable_npf_assist) {
inst_len = ctrl->inst_len;
inst_bytes = ctrl->inst_bytes;
} else {
inst_len = 0;
inst_bytes = NULL;
}
vie_init(&vmexit->u.inst_emul.vie, inst_bytes, inst_len);
}
#ifdef KTR
static const char *
intrtype_to_str(int intr_type)
{
switch (intr_type) {
case VMCB_EVENTINJ_TYPE_INTR:
return ("hwintr");
case VMCB_EVENTINJ_TYPE_NMI:
return ("nmi");
case VMCB_EVENTINJ_TYPE_INTn:
return ("swintr");
case VMCB_EVENTINJ_TYPE_EXCEPTION:
return ("exception");
default:
panic("%s: unknown intr_type %d", __func__, intr_type);
}
}
#endif
/*
* Inject an event to vcpu as described in section 15.20, "Event injection".
*/
static void
svm_eventinject(struct svm_softc *sc, int vcpu, int intr_type, int vector,
uint32_t error, bool ec_valid)
{
struct vmcb_ctrl *ctrl;
ctrl = svm_get_vmcb_ctrl(sc, vcpu);
KASSERT((ctrl->eventinj & VMCB_EVENTINJ_VALID) == 0,
("%s: event already pending %#lx", __func__, ctrl->eventinj));
KASSERT(vector >=0 && vector <= 255, ("%s: invalid vector %d",
__func__, vector));
switch (intr_type) {
case VMCB_EVENTINJ_TYPE_INTR:
case VMCB_EVENTINJ_TYPE_NMI:
case VMCB_EVENTINJ_TYPE_INTn:
break;
case VMCB_EVENTINJ_TYPE_EXCEPTION:
if (vector >= 0 && vector <= 31 && vector != 2)
break;
/* FALLTHROUGH */
default:
panic("%s: invalid intr_type/vector: %d/%d", __func__,
intr_type, vector);
}
ctrl->eventinj = vector | (intr_type << 8) | VMCB_EVENTINJ_VALID;
if (ec_valid) {
ctrl->eventinj |= VMCB_EVENTINJ_EC_VALID;
ctrl->eventinj |= (uint64_t)error << 32;
VCPU_CTR3(sc->vm, vcpu, "Injecting %s at vector %d errcode %#x",
intrtype_to_str(intr_type), vector, error);
} else {
VCPU_CTR2(sc->vm, vcpu, "Injecting %s at vector %d",
intrtype_to_str(intr_type), vector);
}
}
static void
svm_update_virqinfo(struct svm_softc *sc, int vcpu)
{
struct vm *vm;
struct vlapic *vlapic;
struct vmcb_ctrl *ctrl;
int pending;
vm = sc->vm;
vlapic = vm_lapic(vm, vcpu);
ctrl = svm_get_vmcb_ctrl(sc, vcpu);
/* Update %cr8 in the emulated vlapic */
vlapic_set_cr8(vlapic, ctrl->v_tpr);
/*
* If V_IRQ indicates that the interrupt injection attempted on then
* last VMRUN was successful then update the vlapic accordingly.
*/
if (ctrl->v_intr_vector != 0) {
pending = ctrl->v_irq;
KASSERT(ctrl->v_intr_vector >= 16, ("%s: invalid "
"v_intr_vector %d", __func__, ctrl->v_intr_vector));
KASSERT(!ctrl->v_ign_tpr, ("%s: invalid v_ign_tpr", __func__));
VCPU_CTR2(vm, vcpu, "v_intr_vector %d %s", ctrl->v_intr_vector,
pending ? "pending" : "accepted");
if (!pending)
vlapic_intr_accepted(vlapic, ctrl->v_intr_vector);
}
}
static void
svm_save_intinfo(struct svm_softc *svm_sc, int vcpu)
{
struct vmcb_ctrl *ctrl;
uint64_t intinfo;
ctrl = svm_get_vmcb_ctrl(svm_sc, vcpu);
intinfo = ctrl->exitintinfo;
if (!VMCB_EXITINTINFO_VALID(intinfo))
return;
/*
* From APMv2, Section "Intercepts during IDT interrupt delivery"
*
* If a #VMEXIT happened during event delivery then record the event
* that was being delivered.
*/
VCPU_CTR2(svm_sc->vm, vcpu, "SVM:Pending INTINFO(0x%lx), vector=%d.\n",
intinfo, VMCB_EXITINTINFO_VECTOR(intinfo));
vmm_stat_incr(svm_sc->vm, vcpu, VCPU_EXITINTINFO, 1);
vm_exit_intinfo(svm_sc->vm, vcpu, intinfo);
}
static __inline int
vintr_intercept_enabled(struct svm_softc *sc, int vcpu)
{
return (svm_get_intercept(sc, vcpu, VMCB_CTRL1_INTCPT,
VMCB_INTCPT_VINTR));
}
static __inline void
enable_intr_window_exiting(struct svm_softc *sc, int vcpu)
{
struct vmcb_ctrl *ctrl;
ctrl = svm_get_vmcb_ctrl(sc, vcpu);
if (ctrl->v_irq && ctrl->v_intr_vector == 0) {
KASSERT(ctrl->v_ign_tpr, ("%s: invalid v_ign_tpr", __func__));
KASSERT(vintr_intercept_enabled(sc, vcpu),
("%s: vintr intercept should be enabled", __func__));
return;
}
VCPU_CTR0(sc->vm, vcpu, "Enable intr window exiting");
ctrl->v_irq = 1;
ctrl->v_ign_tpr = 1;
ctrl->v_intr_vector = 0;
svm_set_dirty(sc, vcpu, VMCB_CACHE_TPR);
svm_enable_intercept(sc, vcpu, VMCB_CTRL1_INTCPT, VMCB_INTCPT_VINTR);
}
static __inline void
disable_intr_window_exiting(struct svm_softc *sc, int vcpu)
{
struct vmcb_ctrl *ctrl;
ctrl = svm_get_vmcb_ctrl(sc, vcpu);
if (!ctrl->v_irq && ctrl->v_intr_vector == 0) {
KASSERT(!vintr_intercept_enabled(sc, vcpu),
("%s: vintr intercept should be disabled", __func__));
return;
}
#ifdef KTR
if (ctrl->v_intr_vector == 0)
VCPU_CTR0(sc->vm, vcpu, "Disable intr window exiting");
else
VCPU_CTR0(sc->vm, vcpu, "Clearing V_IRQ interrupt injection");
#endif
ctrl->v_irq = 0;
ctrl->v_intr_vector = 0;
svm_set_dirty(sc, vcpu, VMCB_CACHE_TPR);
svm_disable_intercept(sc, vcpu, VMCB_CTRL1_INTCPT, VMCB_INTCPT_VINTR);
}
static int
svm_modify_intr_shadow(struct svm_softc *sc, int vcpu, uint64_t val)
{
struct vmcb_ctrl *ctrl;
int oldval, newval;
ctrl = svm_get_vmcb_ctrl(sc, vcpu);
oldval = ctrl->intr_shadow;
newval = val ? 1 : 0;
if (newval != oldval) {
ctrl->intr_shadow = newval;
VCPU_CTR1(sc->vm, vcpu, "Setting intr_shadow to %d", newval);
}
return (0);
}
static int
svm_get_intr_shadow(struct svm_softc *sc, int vcpu, uint64_t *val)
{
struct vmcb_ctrl *ctrl;
ctrl = svm_get_vmcb_ctrl(sc, vcpu);
*val = ctrl->intr_shadow;
return (0);
}
/*
* Once an NMI is injected it blocks delivery of further NMIs until the handler
* executes an IRET. The IRET intercept is enabled when an NMI is injected to
* to track when the vcpu is done handling the NMI.
*/
static int
nmi_blocked(struct svm_softc *sc, int vcpu)
{
int blocked;
blocked = svm_get_intercept(sc, vcpu, VMCB_CTRL1_INTCPT,
VMCB_INTCPT_IRET);
return (blocked);
}
static void
enable_nmi_blocking(struct svm_softc *sc, int vcpu)
{
KASSERT(!nmi_blocked(sc, vcpu), ("vNMI already blocked"));
VCPU_CTR0(sc->vm, vcpu, "vNMI blocking enabled");
svm_enable_intercept(sc, vcpu, VMCB_CTRL1_INTCPT, VMCB_INTCPT_IRET);
}
static void
clear_nmi_blocking(struct svm_softc *sc, int vcpu)
{
int error;
KASSERT(nmi_blocked(sc, vcpu), ("vNMI already unblocked"));
VCPU_CTR0(sc->vm, vcpu, "vNMI blocking cleared");
/*
* When the IRET intercept is cleared the vcpu will attempt to execute
* the "iret" when it runs next. However, it is possible to inject
* another NMI into the vcpu before the "iret" has actually executed.
*
* For e.g. if the "iret" encounters a #NPF when accessing the stack
* it will trap back into the hypervisor. If an NMI is pending for
* the vcpu it will be injected into the guest.
*
* XXX this needs to be fixed
*/
svm_disable_intercept(sc, vcpu, VMCB_CTRL1_INTCPT, VMCB_INTCPT_IRET);
/*
* Set 'intr_shadow' to prevent an NMI from being injected on the
* immediate VMRUN.
*/
error = svm_modify_intr_shadow(sc, vcpu, 1);
KASSERT(!error, ("%s: error %d setting intr_shadow", __func__, error));
}
static int
emulate_wrmsr(struct svm_softc *sc, int vcpu, u_int num, uint64_t val,
bool *retu)
{
int error;
if (lapic_msr(num))
error = lapic_wrmsr(sc->vm, vcpu, num, val, retu);
else if (num == MSR_EFER)
error = svm_setreg(sc, vcpu, VM_REG_GUEST_EFER, val);
else
error = svm_wrmsr(sc, vcpu, num, val, retu);
return (error);
}
static int
emulate_rdmsr(struct svm_softc *sc, int vcpu, u_int num, bool *retu)
{
struct vmcb_state *state;
struct svm_regctx *ctx;
uint64_t result;
int error;
if (lapic_msr(num))
error = lapic_rdmsr(sc->vm, vcpu, num, &result, retu);
else
error = svm_rdmsr(sc, vcpu, num, &result, retu);
if (error == 0) {
state = svm_get_vmcb_state(sc, vcpu);
ctx = svm_get_guest_regctx(sc, vcpu);
state->rax = result & 0xffffffff;
ctx->sctx_rdx = result >> 32;
}
return (error);
}
#ifdef KTR
static const char *
exit_reason_to_str(uint64_t reason)
{
static char reasonbuf[32];
switch (reason) {
case VMCB_EXIT_INVALID:
return ("invalvmcb");
case VMCB_EXIT_SHUTDOWN:
return ("shutdown");
case VMCB_EXIT_NPF:
return ("nptfault");
case VMCB_EXIT_PAUSE:
return ("pause");
case VMCB_EXIT_HLT:
return ("hlt");
case VMCB_EXIT_CPUID:
return ("cpuid");
case VMCB_EXIT_IO:
return ("inout");
case VMCB_EXIT_MC:
return ("mchk");
case VMCB_EXIT_INTR:
return ("extintr");
case VMCB_EXIT_NMI:
return ("nmi");
case VMCB_EXIT_VINTR:
return ("vintr");
case VMCB_EXIT_MSR:
return ("msr");
case VMCB_EXIT_IRET:
return ("iret");
case VMCB_EXIT_MONITOR:
return ("monitor");
case VMCB_EXIT_MWAIT:
return ("mwait");
default:
snprintf(reasonbuf, sizeof(reasonbuf), "%#lx", reason);
return (reasonbuf);
}
}
#endif /* KTR */
/*
* From section "State Saved on Exit" in APMv2: nRIP is saved for all #VMEXITs
* that are due to instruction intercepts as well as MSR and IOIO intercepts
* and exceptions caused by INT3, INTO and BOUND instructions.
*
* Return 1 if the nRIP is valid and 0 otherwise.
*/
static int
nrip_valid(uint64_t exitcode)
{
switch (exitcode) {
case 0x00 ... 0x0F: /* read of CR0 through CR15 */
case 0x10 ... 0x1F: /* write of CR0 through CR15 */
case 0x20 ... 0x2F: /* read of DR0 through DR15 */
case 0x30 ... 0x3F: /* write of DR0 through DR15 */
case 0x43: /* INT3 */
case 0x44: /* INTO */
case 0x45: /* BOUND */
case 0x65 ... 0x7C: /* VMEXIT_CR0_SEL_WRITE ... VMEXIT_MSR */
case 0x80 ... 0x8D: /* VMEXIT_VMRUN ... VMEXIT_XSETBV */
return (1);
default:
return (0);
}
}
/*
* Collateral for a generic SVM VM-exit.
*/
static void
vm_exit_svm(struct vm_exit *vme, uint64_t code, uint64_t info1, uint64_t info2)
{
vme->exitcode = VM_EXITCODE_SVM;
vme->u.svm.exitcode = code;
vme->u.svm.exitinfo1 = info1;
vme->u.svm.exitinfo2 = info2;
}
static int
svm_vmexit(struct svm_softc *svm_sc, int vcpu, struct vm_exit *vmexit)
{
struct vmcb *vmcb;
struct vmcb_state *state;
struct vmcb_ctrl *ctrl;
struct svm_regctx *ctx;
struct vm_exception exception;
uint64_t code, info1, info2, val;
uint32_t eax, ecx, edx;
int error, errcode_valid, handled, idtvec, reflect;
bool retu;
ctx = svm_get_guest_regctx(svm_sc, vcpu);
vmcb = svm_get_vmcb(svm_sc, vcpu);
state = &vmcb->state;
ctrl = &vmcb->ctrl;
handled = 0;
code = ctrl->exitcode;
info1 = ctrl->exitinfo1;
info2 = ctrl->exitinfo2;
vmexit->exitcode = VM_EXITCODE_BOGUS;
vmexit->rip = state->rip;
vmexit->inst_length = nrip_valid(code) ? ctrl->nrip - state->rip : 0;
vmm_stat_incr(svm_sc->vm, vcpu, VMEXIT_COUNT, 1);
/*
* #VMEXIT(INVALID) needs to be handled early because the VMCB is
* in an inconsistent state and can trigger assertions that would
* never happen otherwise.
*/
if (code == VMCB_EXIT_INVALID) {
vm_exit_svm(vmexit, code, info1, info2);
return (0);
}
KASSERT((ctrl->eventinj & VMCB_EVENTINJ_VALID) == 0, ("%s: event "
"injection valid bit is set %#lx", __func__, ctrl->eventinj));
KASSERT(vmexit->inst_length >= 0 && vmexit->inst_length <= 15,
("invalid inst_length %d: code (%#lx), info1 (%#lx), info2 (%#lx)",
vmexit->inst_length, code, info1, info2));
svm_update_virqinfo(svm_sc, vcpu);
svm_save_intinfo(svm_sc, vcpu);
switch (code) {
case VMCB_EXIT_IRET:
/*
* Restart execution at "iret" but with the intercept cleared.
*/
vmexit->inst_length = 0;
clear_nmi_blocking(svm_sc, vcpu);
handled = 1;
break;
case VMCB_EXIT_VINTR: /* interrupt window exiting */
vmm_stat_incr(svm_sc->vm, vcpu, VMEXIT_VINTR, 1);
handled = 1;
break;
case VMCB_EXIT_INTR: /* external interrupt */
vmm_stat_incr(svm_sc->vm, vcpu, VMEXIT_EXTINT, 1);
handled = 1;
break;
case VMCB_EXIT_NMI: /* external NMI */
handled = 1;
break;
case 0x40 ... 0x5F:
vmm_stat_incr(svm_sc->vm, vcpu, VMEXIT_EXCEPTION, 1);
reflect = 1;
idtvec = code - 0x40;
switch (idtvec) {
case IDT_MC:
/*
* Call the machine check handler by hand. Also don't
* reflect the machine check back into the guest.
*/
reflect = 0;
VCPU_CTR0(svm_sc->vm, vcpu, "Vectoring to MCE handler");
__asm __volatile("int $18");
break;
case IDT_PF:
error = svm_setreg(svm_sc, vcpu, VM_REG_GUEST_CR2,
info2);
KASSERT(error == 0, ("%s: error %d updating cr2",
__func__, error));
/* fallthru */
case IDT_NP:
case IDT_SS:
case IDT_GP:
case IDT_AC:
case IDT_TS:
errcode_valid = 1;
break;
case IDT_DF:
errcode_valid = 1;
info1 = 0;
break;
case IDT_BP:
case IDT_OF:
case IDT_BR:
/*
* The 'nrip' field is populated for INT3, INTO and
* BOUND exceptions and this also implies that
* 'inst_length' is non-zero.
*
* Reset 'inst_length' to zero so the guest %rip at
* event injection is identical to what it was when
* the exception originally happened.
*/
VCPU_CTR2(svm_sc->vm, vcpu, "Reset inst_length from %d "
"to zero before injecting exception %d",
vmexit->inst_length, idtvec);
vmexit->inst_length = 0;
/* fallthru */
default:
errcode_valid = 0;
break;
}
KASSERT(vmexit->inst_length == 0, ("invalid inst_length (%d) "
"when reflecting exception %d into guest",
vmexit->inst_length, idtvec));
if (reflect) {
/* Reflect the exception back into the guest */
bzero(&exception, sizeof(struct vm_exception));
exception.vector = idtvec;
if (errcode_valid) {
exception.error_code = info1;
exception.error_code_valid = 1;
}
VCPU_CTR2(svm_sc->vm, vcpu, "Reflecting exception "
"%d/%#x into the guest", exception.vector,
exception.error_code);
error = vm_inject_exception(svm_sc->vm, vcpu,
&exception);
KASSERT(error == 0, ("%s: vm_inject_exception error %d",
__func__, error));
}
handled = 1;
break;
case VMCB_EXIT_MSR: /* MSR access. */
eax = state->rax;
ecx = ctx->sctx_rcx;
edx = ctx->sctx_rdx;
retu = false;
if (info1) {
vmm_stat_incr(svm_sc->vm, vcpu, VMEXIT_WRMSR, 1);
val = (uint64_t)edx << 32 | eax;
VCPU_CTR2(svm_sc->vm, vcpu, "wrmsr %#x val %#lx",
ecx, val);
if (emulate_wrmsr(svm_sc, vcpu, ecx, val, &retu)) {
vmexit->exitcode = VM_EXITCODE_WRMSR;
vmexit->u.msr.code = ecx;
vmexit->u.msr.wval = val;
} else if (!retu) {
handled = 1;
} else {
KASSERT(vmexit->exitcode != VM_EXITCODE_BOGUS,
("emulate_wrmsr retu with bogus exitcode"));
}
} else {
VCPU_CTR1(svm_sc->vm, vcpu, "rdmsr %#x", ecx);
vmm_stat_incr(svm_sc->vm, vcpu, VMEXIT_RDMSR, 1);
if (emulate_rdmsr(svm_sc, vcpu, ecx, &retu)) {
vmexit->exitcode = VM_EXITCODE_RDMSR;
vmexit->u.msr.code = ecx;
} else if (!retu) {
handled = 1;
} else {
KASSERT(vmexit->exitcode != VM_EXITCODE_BOGUS,
("emulate_rdmsr retu with bogus exitcode"));
}
}
break;
case VMCB_EXIT_IO:
handled = svm_handle_io(svm_sc, vcpu, vmexit);
vmm_stat_incr(svm_sc->vm, vcpu, VMEXIT_INOUT, 1);
break;
case VMCB_EXIT_CPUID:
vmm_stat_incr(svm_sc->vm, vcpu, VMEXIT_CPUID, 1);
handled = x86_emulate_cpuid(svm_sc->vm, vcpu,
(uint32_t *)&state->rax,
(uint32_t *)&ctx->sctx_rbx,
(uint32_t *)&ctx->sctx_rcx,
(uint32_t *)&ctx->sctx_rdx);
break;
case VMCB_EXIT_HLT:
vmm_stat_incr(svm_sc->vm, vcpu, VMEXIT_HLT, 1);
vmexit->exitcode = VM_EXITCODE_HLT;
vmexit->u.hlt.rflags = state->rflags;
break;
case VMCB_EXIT_PAUSE:
vmexit->exitcode = VM_EXITCODE_PAUSE;
vmm_stat_incr(svm_sc->vm, vcpu, VMEXIT_PAUSE, 1);
break;
case VMCB_EXIT_NPF:
/* EXITINFO2 contains the faulting guest physical address */
if (info1 & VMCB_NPF_INFO1_RSV) {
VCPU_CTR2(svm_sc->vm, vcpu, "nested page fault with "
"reserved bits set: info1(%#lx) info2(%#lx)",
info1, info2);
} else if (vm_mem_allocated(svm_sc->vm, info2)) {
vmexit->exitcode = VM_EXITCODE_PAGING;
vmexit->u.paging.gpa = info2;
vmexit->u.paging.fault_type = npf_fault_type(info1);
vmm_stat_incr(svm_sc->vm, vcpu, VMEXIT_NESTED_FAULT, 1);
VCPU_CTR3(svm_sc->vm, vcpu, "nested page fault "
"on gpa %#lx/%#lx at rip %#lx",
info2, info1, state->rip);
} else if (svm_npf_emul_fault(info1)) {
svm_handle_inst_emul(vmcb, info2, vmexit);
vmm_stat_incr(svm_sc->vm, vcpu, VMEXIT_INST_EMUL, 1);
VCPU_CTR3(svm_sc->vm, vcpu, "inst_emul fault "
"for gpa %#lx/%#lx at rip %#lx",
info2, info1, state->rip);
}
break;
case VMCB_EXIT_MONITOR:
vmexit->exitcode = VM_EXITCODE_MONITOR;
break;
case VMCB_EXIT_MWAIT:
vmexit->exitcode = VM_EXITCODE_MWAIT;
break;
default:
vmm_stat_incr(svm_sc->vm, vcpu, VMEXIT_UNKNOWN, 1);
break;
}
VCPU_CTR4(svm_sc->vm, vcpu, "%s %s vmexit at %#lx/%d",
handled ? "handled" : "unhandled", exit_reason_to_str(code),
vmexit->rip, vmexit->inst_length);
if (handled) {
vmexit->rip += vmexit->inst_length;
vmexit->inst_length = 0;
state->rip = vmexit->rip;
} else {
if (vmexit->exitcode == VM_EXITCODE_BOGUS) {
/*
* If this VM exit was not claimed by anybody then
* treat it as a generic SVM exit.
*/
vm_exit_svm(vmexit, code, info1, info2);
} else {
/*
* The exitcode and collateral have been populated.
* The VM exit will be processed further in userland.
*/
}
}
return (handled);
}
static void
svm_inj_intinfo(struct svm_softc *svm_sc, int vcpu)
{
uint64_t intinfo;
if (!vm_entry_intinfo(svm_sc->vm, vcpu, &intinfo))
return;
KASSERT(VMCB_EXITINTINFO_VALID(intinfo), ("%s: entry intinfo is not "
"valid: %#lx", __func__, intinfo));
svm_eventinject(svm_sc, vcpu, VMCB_EXITINTINFO_TYPE(intinfo),
VMCB_EXITINTINFO_VECTOR(intinfo),
VMCB_EXITINTINFO_EC(intinfo),
VMCB_EXITINTINFO_EC_VALID(intinfo));
vmm_stat_incr(svm_sc->vm, vcpu, VCPU_INTINFO_INJECTED, 1);
VCPU_CTR1(svm_sc->vm, vcpu, "Injected entry intinfo: %#lx", intinfo);
}
/*
* Inject event to virtual cpu.
*/
static void
svm_inj_interrupts(struct svm_softc *sc, int vcpu, struct vlapic *vlapic)
{
struct vmcb_ctrl *ctrl;
struct vmcb_state *state;
uint8_t v_tpr;
int vector, need_intr_window, pending_apic_vector;
state = svm_get_vmcb_state(sc, vcpu);
ctrl = svm_get_vmcb_ctrl(sc, vcpu);
need_intr_window = 0;
pending_apic_vector = 0;
/*
* Inject pending events or exceptions for this vcpu.
*
* An event might be pending because the previous #VMEXIT happened
* during event delivery (i.e. ctrl->exitintinfo).
*
* An event might also be pending because an exception was injected
* by the hypervisor (e.g. #PF during instruction emulation).
*/
svm_inj_intinfo(sc, vcpu);
/* NMI event has priority over interrupts. */
if (vm_nmi_pending(sc->vm, vcpu)) {
if (nmi_blocked(sc, vcpu)) {
/*
* Can't inject another NMI if the guest has not
* yet executed an "iret" after the last NMI.
*/
VCPU_CTR0(sc->vm, vcpu, "Cannot inject NMI due "
"to NMI-blocking");
} else if (ctrl->intr_shadow) {
/*
* Can't inject an NMI if the vcpu is in an intr_shadow.
*/
VCPU_CTR0(sc->vm, vcpu, "Cannot inject NMI due to "
"interrupt shadow");
need_intr_window = 1;
goto done;
} else if (ctrl->eventinj & VMCB_EVENTINJ_VALID) {
/*
* If there is already an exception/interrupt pending
* then defer the NMI until after that.
*/
VCPU_CTR1(sc->vm, vcpu, "Cannot inject NMI due to "
"eventinj %#lx", ctrl->eventinj);
/*
* Use self-IPI to trigger a VM-exit as soon as
* possible after the event injection is completed.
*
* This works only if the external interrupt exiting
* is at a lower priority than the event injection.
*
* Although not explicitly specified in APMv2 the
* relative priorities were verified empirically.
*/
ipi_cpu(curcpu, IPI_AST); /* XXX vmm_ipinum? */
} else {
vm_nmi_clear(sc->vm, vcpu);
/* Inject NMI, vector number is not used */
svm_eventinject(sc, vcpu, VMCB_EVENTINJ_TYPE_NMI,
IDT_NMI, 0, false);
/* virtual NMI blocking is now in effect */
enable_nmi_blocking(sc, vcpu);
VCPU_CTR0(sc->vm, vcpu, "Injecting vNMI");
}
}
if (!vm_extint_pending(sc->vm, vcpu)) {
/*
* APIC interrupts are delivered using the V_IRQ offload.
*
* The primary benefit is that the hypervisor doesn't need to
* deal with the various conditions that inhibit interrupts.
* It also means that TPR changes via CR8 will be handled
* without any hypervisor involvement.
*
* Note that the APIC vector must remain pending in the vIRR
* until it is confirmed that it was delivered to the guest.
* This can be confirmed based on the value of V_IRQ at the
* next #VMEXIT (1 = pending, 0 = delivered).
*
* Also note that it is possible that another higher priority
* vector can become pending before this vector is delivered
* to the guest. This is alright because vcpu_notify_event()
* will send an IPI and force the vcpu to trap back into the
* hypervisor. The higher priority vector will be injected on
* the next VMRUN.
*/
if (vlapic_pending_intr(vlapic, &vector)) {
KASSERT(vector >= 16 && vector <= 255,
("invalid vector %d from local APIC", vector));
pending_apic_vector = vector;
}
goto done;
}
/* Ask the legacy pic for a vector to inject */
vatpic_pending_intr(sc->vm, &vector);
KASSERT(vector >= 0 && vector <= 255, ("invalid vector %d from INTR",
vector));
/*
* If the guest has disabled interrupts or is in an interrupt shadow
* then we cannot inject the pending interrupt.
*/
if ((state->rflags & PSL_I) == 0) {
VCPU_CTR2(sc->vm, vcpu, "Cannot inject vector %d due to "
"rflags %#lx", vector, state->rflags);
need_intr_window = 1;
goto done;
}
if (ctrl->intr_shadow) {
VCPU_CTR1(sc->vm, vcpu, "Cannot inject vector %d due to "
"interrupt shadow", vector);
need_intr_window = 1;
goto done;
}
if (ctrl->eventinj & VMCB_EVENTINJ_VALID) {
VCPU_CTR2(sc->vm, vcpu, "Cannot inject vector %d due to "
"eventinj %#lx", vector, ctrl->eventinj);
need_intr_window = 1;
goto done;
}
/*
* Legacy PIC interrupts are delivered via the event injection
* mechanism.
*/
svm_eventinject(sc, vcpu, VMCB_EVENTINJ_TYPE_INTR, vector, 0, false);
vm_extint_clear(sc->vm, vcpu);
vatpic_intr_accepted(sc->vm, vector);
/*
* Force a VM-exit as soon as the vcpu is ready to accept another
* interrupt. This is done because the PIC might have another vector
* that it wants to inject. Also, if the APIC has a pending interrupt
* that was preempted by the ExtInt then it allows us to inject the
* APIC vector as soon as possible.
*/
need_intr_window = 1;
done:
/*
* The guest can modify the TPR by writing to %CR8. In guest mode
* the processor reflects this write to V_TPR without hypervisor
* intervention.
*
* The guest can also modify the TPR by writing to it via the memory
* mapped APIC page. In this case, the write will be emulated by the
* hypervisor. For this reason V_TPR must be updated before every
* VMRUN.
*/
v_tpr = vlapic_get_cr8(vlapic);
KASSERT(v_tpr >= 0 && v_tpr <= 15, ("invalid v_tpr %#x", v_tpr));
if (ctrl->v_tpr != v_tpr) {
VCPU_CTR2(sc->vm, vcpu, "VMCB V_TPR changed from %#x to %#x",
ctrl->v_tpr, v_tpr);
ctrl->v_tpr = v_tpr;
svm_set_dirty(sc, vcpu, VMCB_CACHE_TPR);
}
if (pending_apic_vector) {
/*
* If an APIC vector is being injected then interrupt window
* exiting is not possible on this VMRUN.
*/
KASSERT(!need_intr_window, ("intr_window exiting impossible"));
VCPU_CTR1(sc->vm, vcpu, "Injecting vector %d using V_IRQ",
pending_apic_vector);
ctrl->v_irq = 1;
ctrl->v_ign_tpr = 0;
ctrl->v_intr_vector = pending_apic_vector;
ctrl->v_intr_prio = pending_apic_vector >> 4;
svm_set_dirty(sc, vcpu, VMCB_CACHE_TPR);
} else if (need_intr_window) {
/*
* We use V_IRQ in conjunction with the VINTR intercept to
* trap into the hypervisor as soon as a virtual interrupt
* can be delivered.
*
* Since injected events are not subject to intercept checks
* we need to ensure that the V_IRQ is not actually going to
* be delivered on VM entry. The KASSERT below enforces this.
*/
KASSERT((ctrl->eventinj & VMCB_EVENTINJ_VALID) != 0 ||
(state->rflags & PSL_I) == 0 || ctrl->intr_shadow,
("Bogus intr_window_exiting: eventinj (%#lx), "
"intr_shadow (%u), rflags (%#lx)",
ctrl->eventinj, ctrl->intr_shadow, state->rflags));
enable_intr_window_exiting(sc, vcpu);
} else {
disable_intr_window_exiting(sc, vcpu);
}
}
static __inline void
restore_host_tss(void)
{
struct system_segment_descriptor *tss_sd;
/*
* The TSS descriptor was in use prior to launching the guest so it
* has been marked busy.
*
* 'ltr' requires the descriptor to be marked available so change the
* type to "64-bit available TSS".
*/
tss_sd = PCPU_GET(tss);
tss_sd->sd_type = SDT_SYSTSS;
ltr(GSEL(GPROC0_SEL, SEL_KPL));
}
static void
check_asid(struct svm_softc *sc, int vcpuid, pmap_t pmap, u_int thiscpu)
{
struct svm_vcpu *vcpustate;
struct vmcb_ctrl *ctrl;
long eptgen;
bool alloc_asid;
KASSERT(CPU_ISSET(thiscpu, &pmap->pm_active), ("%s: nested pmap not "
"active on cpu %u", __func__, thiscpu));
vcpustate = svm_get_vcpu(sc, vcpuid);
ctrl = svm_get_vmcb_ctrl(sc, vcpuid);
/*
* The TLB entries associated with the vcpu's ASID are not valid
* if either of the following conditions is true:
*
* 1. The vcpu's ASID generation is different than the host cpu's
* ASID generation. This happens when the vcpu migrates to a new
* host cpu. It can also happen when the number of vcpus executing
* on a host cpu is greater than the number of ASIDs available.
*
* 2. The pmap generation number is different than the value cached in
* the 'vcpustate'. This happens when the host invalidates pages
* belonging to the guest.
*
* asidgen eptgen Action
* mismatch mismatch
* 0 0 (a)
* 0 1 (b1) or (b2)
* 1 0 (c)
* 1 1 (d)
*
* (a) There is no mismatch in eptgen or ASID generation and therefore
* no further action is needed.
*
* (b1) If the cpu supports FlushByAsid then the vcpu's ASID is
* retained and the TLB entries associated with this ASID
* are flushed by VMRUN.
*
* (b2) If the cpu does not support FlushByAsid then a new ASID is
* allocated.
*
* (c) A new ASID is allocated.
*
* (d) A new ASID is allocated.
*/
alloc_asid = false;
eptgen = pmap->pm_eptgen;
ctrl->tlb_ctrl = VMCB_TLB_FLUSH_NOTHING;
if (vcpustate->asid.gen != asid[thiscpu].gen) {
alloc_asid = true; /* (c) and (d) */
} else if (vcpustate->eptgen != eptgen) {
if (flush_by_asid())
ctrl->tlb_ctrl = VMCB_TLB_FLUSH_GUEST; /* (b1) */
else
alloc_asid = true; /* (b2) */
} else {
/*
* This is the common case (a).
*/
KASSERT(!alloc_asid, ("ASID allocation not necessary"));
KASSERT(ctrl->tlb_ctrl == VMCB_TLB_FLUSH_NOTHING,
("Invalid VMCB tlb_ctrl: %#x", ctrl->tlb_ctrl));
}
if (alloc_asid) {
if (++asid[thiscpu].num >= nasid) {
asid[thiscpu].num = 1;
if (++asid[thiscpu].gen == 0)
asid[thiscpu].gen = 1;
/*
* If this cpu does not support "flush-by-asid"
* then flush the entire TLB on a generation
* bump. Subsequent ASID allocation in this
* generation can be done without a TLB flush.
*/
if (!flush_by_asid())
ctrl->tlb_ctrl = VMCB_TLB_FLUSH_ALL;
}
vcpustate->asid.gen = asid[thiscpu].gen;
vcpustate->asid.num = asid[thiscpu].num;
ctrl->asid = vcpustate->asid.num;
svm_set_dirty(sc, vcpuid, VMCB_CACHE_ASID);
/*
* If this cpu supports "flush-by-asid" then the TLB
* was not flushed after the generation bump. The TLB
* is flushed selectively after every new ASID allocation.
*/
if (flush_by_asid())
ctrl->tlb_ctrl = VMCB_TLB_FLUSH_GUEST;
}
vcpustate->eptgen = eptgen;
KASSERT(ctrl->asid != 0, ("Guest ASID must be non-zero"));
KASSERT(ctrl->asid == vcpustate->asid.num,
("ASID mismatch: %u/%u", ctrl->asid, vcpustate->asid.num));
}
static __inline void
disable_gintr(void)
{
__asm __volatile("clgi" : : :);
}
static __inline void
enable_gintr(void)
{
__asm __volatile("stgi" : : :);
}
/*
* Start vcpu with specified RIP.
*/
static int
svm_vmrun(void *arg, int vcpu, register_t rip, pmap_t pmap,
void *rend_cookie, void *suspended_cookie)
{
struct svm_regctx *gctx;
struct svm_softc *svm_sc;
struct svm_vcpu *vcpustate;
struct vmcb_state *state;
struct vmcb_ctrl *ctrl;
struct vm_exit *vmexit;
struct vlapic *vlapic;
struct vm *vm;
uint64_t vmcb_pa;
u_int thiscpu;
int handled;
svm_sc = arg;
vm = svm_sc->vm;
vcpustate = svm_get_vcpu(svm_sc, vcpu);
state = svm_get_vmcb_state(svm_sc, vcpu);
ctrl = svm_get_vmcb_ctrl(svm_sc, vcpu);
vmexit = vm_exitinfo(vm, vcpu);
vlapic = vm_lapic(vm, vcpu);
/*
* Stash 'curcpu' on the stack as 'thiscpu'.
*
* The per-cpu data area is not accessible until MSR_GSBASE is restored
* after the #VMEXIT. Since VMRUN is executed inside a critical section
* 'curcpu' and 'thiscpu' are guaranteed to identical.
*/
thiscpu = curcpu;
gctx = svm_get_guest_regctx(svm_sc, vcpu);
vmcb_pa = svm_sc->vcpu[vcpu].vmcb_pa;
if (vcpustate->lastcpu != thiscpu) {
/*
* Force new ASID allocation by invalidating the generation.
*/
vcpustate->asid.gen = 0;
/*
* Invalidate the VMCB state cache by marking all fields dirty.
*/
svm_set_dirty(svm_sc, vcpu, 0xffffffff);
/*
* XXX
* Setting 'vcpustate->lastcpu' here is bit premature because
* we may return from this function without actually executing
* the VMRUN instruction. This could happen if a rendezvous
* or an AST is pending on the first time through the loop.
*
* This works for now but any new side-effects of vcpu
* migration should take this case into account.
*/
vcpustate->lastcpu = thiscpu;
vmm_stat_incr(vm, vcpu, VCPU_MIGRATIONS, 1);
}
svm_msr_guest_enter(svm_sc, vcpu);
/* Update Guest RIP */
state->rip = rip;
do {
/*
* Disable global interrupts to guarantee atomicity during
* loading of guest state. This includes not only the state
* loaded by the "vmrun" instruction but also software state
* maintained by the hypervisor: suspended and rendezvous
* state, NPT generation number, vlapic interrupts etc.
*/
disable_gintr();
if (vcpu_suspended(suspended_cookie)) {
enable_gintr();
vm_exit_suspended(vm, vcpu, state->rip);
break;
}
if (vcpu_rendezvous_pending(rend_cookie)) {
enable_gintr();
vm_exit_rendezvous(vm, vcpu, state->rip);
break;
}
/* We are asked to give the cpu by scheduler. */
if (curthread->td_flags & (TDF_ASTPENDING | TDF_NEEDRESCHED)) {
enable_gintr();
vm_exit_astpending(vm, vcpu, state->rip);
break;
}
svm_inj_interrupts(svm_sc, vcpu, vlapic);
/* Activate the nested pmap on 'thiscpu' */
CPU_SET_ATOMIC_ACQ(thiscpu, &pmap->pm_active);
/*
* Check the pmap generation and the ASID generation to
* ensure that the vcpu does not use stale TLB mappings.
*/
check_asid(svm_sc, vcpu, pmap, thiscpu);
ctrl->vmcb_clean = vmcb_clean & ~vcpustate->dirty;
vcpustate->dirty = 0;
VCPU_CTR1(vm, vcpu, "vmcb clean %#x", ctrl->vmcb_clean);
/* Launch Virtual Machine. */
VCPU_CTR1(vm, vcpu, "Resume execution at %#lx", state->rip);
svm_launch(vmcb_pa, gctx);
CPU_CLR_ATOMIC(thiscpu, &pmap->pm_active);
/*
* Restore MSR_GSBASE to point to the pcpu data area.
*
* Note that accesses done via PCPU_GET/PCPU_SET will work
* only after MSR_GSBASE is restored.
*
* Also note that we don't bother restoring MSR_KGSBASE
* since it is not used in the kernel and will be restored
* when the VMRUN ioctl returns to userspace.
*/
wrmsr(MSR_GSBASE, (uint64_t)&__pcpu[thiscpu]);
KASSERT(curcpu == thiscpu, ("thiscpu/curcpu (%u/%u) mismatch",
thiscpu, curcpu));
/*
* The host GDTR and IDTR is saved by VMRUN and restored
* automatically on #VMEXIT. However, the host TSS needs
* to be restored explicitly.
*/
restore_host_tss();
/* #VMEXIT disables interrupts so re-enable them here. */
enable_gintr();
/* Handle #VMEXIT and if required return to user space. */
handled = svm_vmexit(svm_sc, vcpu, vmexit);
} while (handled);
svm_msr_guest_exit(svm_sc, vcpu);
return (0);
}
static void
svm_vmcleanup(void *arg)
{
struct svm_softc *sc = arg;
free(sc, M_SVM);
}
static register_t *
swctx_regptr(struct svm_regctx *regctx, int reg)
{
switch (reg) {
case VM_REG_GUEST_RBX:
return (®ctx->sctx_rbx);
case VM_REG_GUEST_RCX:
return (®ctx->sctx_rcx);
case VM_REG_GUEST_RDX:
return (®ctx->sctx_rdx);
case VM_REG_GUEST_RDI:
return (®ctx->sctx_rdi);
case VM_REG_GUEST_RSI:
return (®ctx->sctx_rsi);
case VM_REG_GUEST_RBP:
return (®ctx->sctx_rbp);
case VM_REG_GUEST_R8:
return (®ctx->sctx_r8);
case VM_REG_GUEST_R9:
return (®ctx->sctx_r9);
case VM_REG_GUEST_R10:
return (®ctx->sctx_r10);
case VM_REG_GUEST_R11:
return (®ctx->sctx_r11);
case VM_REG_GUEST_R12:
return (®ctx->sctx_r12);
case VM_REG_GUEST_R13:
return (®ctx->sctx_r13);
case VM_REG_GUEST_R14:
return (®ctx->sctx_r14);
case VM_REG_GUEST_R15:
return (®ctx->sctx_r15);
default:
return (NULL);
}
}
static int
svm_getreg(void *arg, int vcpu, int ident, uint64_t *val)
{
struct svm_softc *svm_sc;
register_t *reg;
svm_sc = arg;
if (ident == VM_REG_GUEST_INTR_SHADOW) {
return (svm_get_intr_shadow(svm_sc, vcpu, val));
}
if (vmcb_read(svm_sc, vcpu, ident, val) == 0) {
return (0);
}
reg = swctx_regptr(svm_get_guest_regctx(svm_sc, vcpu), ident);
if (reg != NULL) {
*val = *reg;
return (0);
}
VCPU_CTR1(svm_sc->vm, vcpu, "svm_getreg: unknown register %#x", ident);
return (EINVAL);
}
static int
svm_setreg(void *arg, int vcpu, int ident, uint64_t val)
{
struct svm_softc *svm_sc;
register_t *reg;
svm_sc = arg;
if (ident == VM_REG_GUEST_INTR_SHADOW) {
return (svm_modify_intr_shadow(svm_sc, vcpu, val));
}
if (vmcb_write(svm_sc, vcpu, ident, val) == 0) {
return (0);
}
reg = swctx_regptr(svm_get_guest_regctx(svm_sc, vcpu), ident);
if (reg != NULL) {
*reg = val;
return (0);
}
/*
* XXX deal with CR3 and invalidate TLB entries tagged with the
* vcpu's ASID. This needs to be treated differently depending on
* whether 'running' is true/false.
*/
VCPU_CTR1(svm_sc->vm, vcpu, "svm_setreg: unknown register %#x", ident);
return (EINVAL);
}
static int
svm_setcap(void *arg, int vcpu, int type, int val)
{
struct svm_softc *sc;
int error;
sc = arg;
error = 0;
switch (type) {
case VM_CAP_HALT_EXIT:
svm_set_intercept(sc, vcpu, VMCB_CTRL1_INTCPT,
VMCB_INTCPT_HLT, val);
break;
case VM_CAP_PAUSE_EXIT:
svm_set_intercept(sc, vcpu, VMCB_CTRL1_INTCPT,
VMCB_INTCPT_PAUSE, val);
break;
case VM_CAP_UNRESTRICTED_GUEST:
/* Unrestricted guest execution cannot be disabled in SVM */
if (val == 0)
error = EINVAL;
break;
default:
error = ENOENT;
break;
}
return (error);
}
static int
svm_getcap(void *arg, int vcpu, int type, int *retval)
{
struct svm_softc *sc;
int error;
sc = arg;
error = 0;
switch (type) {
case VM_CAP_HALT_EXIT:
*retval = svm_get_intercept(sc, vcpu, VMCB_CTRL1_INTCPT,
VMCB_INTCPT_HLT);
break;
case VM_CAP_PAUSE_EXIT:
*retval = svm_get_intercept(sc, vcpu, VMCB_CTRL1_INTCPT,
VMCB_INTCPT_PAUSE);
break;
case VM_CAP_UNRESTRICTED_GUEST:
*retval = 1; /* unrestricted guest is always enabled */
break;
default:
error = ENOENT;
break;
}
return (error);
}
static struct vlapic *
svm_vlapic_init(void *arg, int vcpuid)
{
struct svm_softc *svm_sc;
struct vlapic *vlapic;
svm_sc = arg;
vlapic = malloc(sizeof(struct vlapic), M_SVM_VLAPIC, M_WAITOK | M_ZERO);
vlapic->vm = svm_sc->vm;
vlapic->vcpuid = vcpuid;
vlapic->apic_page = (struct LAPIC *)&svm_sc->apic_page[vcpuid];
vlapic_init(vlapic);
return (vlapic);
}
static void
svm_vlapic_cleanup(void *arg, struct vlapic *vlapic)
{
vlapic_cleanup(vlapic);
free(vlapic, M_SVM_VLAPIC);
}
struct vmm_ops vmm_ops_amd = {
svm_init,
svm_cleanup,
svm_restore,
svm_vminit,
svm_vmrun,
svm_vmcleanup,
svm_getreg,
svm_setreg,
vmcb_getdesc,
vmcb_setdesc,
svm_getcap,
svm_setcap,
svm_npt_alloc,
svm_npt_free,
svm_vlapic_init,
svm_vlapic_cleanup
};
|