summaryrefslogtreecommitdiff
path: root/usr.sbin/vmd/vmm.c
blob: 718a2db0e8682433a0b6f589f9a57b1469f97bff (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
/*	$OpenBSD: vmm.c,v 1.5 2015/12/03 08:42:11 reyk Exp $	*/

/*
 * Copyright (c) 2015 Mike Larkin <mlarkin@openbsd.org>
 *
 * Permission to use, copy, modify, and distribute this software for any
 * purpose with or without fee is hereby granted, provided that the above
 * copyright notice and this permission notice appear in all copies.
 *
 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
 */

#include <sys/param.h>
#include <sys/ioctl.h>
#include <sys/queue.h>
#include <sys/uio.h>
#include <sys/socket.h>
#include <sys/stat.h>
#include <sys/un.h>
#include <sys/wait.h>
#include <sys/mman.h>
#include <sys/time.h>

#include <dev/ic/comreg.h>
#include <dev/ic/i8253reg.h>
#include <dev/isa/isareg.h>
#include <dev/pci/pcireg.h>

#include <machine/param.h>
#include <machine/vmmvar.h>

#include <errno.h>
#include <fcntl.h>
#include <imsg.h>
#include <limits.h>
#include <pthread.h>
#include <pwd.h>
#include <signal.h>
#include <stddef.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <syslog.h>
#include <termios.h>
#include <unistd.h>
#include <poll.h>
#include <util.h>

#include "vmd.h"
#include "loadfile.h"
#include "pci.h"
#include "virtio.h"
#include "proc.h"

/*
 * Emulated 8250 UART
 *
 */
#define COM1_DATA	0x3f8
#define COM1_IER	0x3f9
#define COM1_IIR	0x3fa
#define COM1_LCR	0x3fb
#define COM1_MCR	0x3fc
#define COM1_LSR	0x3fd
#define COM1_MSR	0x3fe
#define COM1_SCR	0x3ff

/*
 * Emulated i8253 PIT (counter)
 */
#define TIMER_BASE	0x40
#define TIMER_CTRL	0x43	/* 8253 Timer #1 */
#define NS_PER_TICK (1000000000 / TIMER_FREQ)

/* i8253 registers */
struct i8253_counter {
	struct timeval tv;	/* timer start time */
	uint16_t start;		/* starting value */
	uint16_t olatch;	/* output latch */
	uint16_t ilatch;	/* input latch */
	uint8_t last_r;		/* last read byte (MSB/LSB) */
	uint8_t last_w;		/* last written byte (MSB/LSB) */
};

/* ns8250 UART registers */
struct ns8250_regs {
	uint8_t lcr;		/* Line Control Register */
	uint8_t fcr;		/* FIFO Control Register */
	uint8_t iir;		/* Interrupt ID Register */
	uint8_t ier;		/* Interrupt Enable Register */
	uint8_t divlo;		/* Baud rate divisor low byte */
	uint8_t divhi;		/* Baud rate divisor high byte */
	uint8_t msr;		/* Modem Status Register */
	uint8_t lsr;		/* Line Status Register */
	uint8_t mcr;		/* Modem Control Register */
	uint8_t scr;		/* Scratch Register */
	uint8_t data;		/* Unread input data */
};

struct i8253_counter i8253_counter[3];
struct ns8250_regs com1_regs;

int start_client_vmd(void);
int opentap(void);
int start_vm(struct imsg *);
int terminate_vm(struct imsg *);
int get_info_vm(struct privsep *, struct imsg *);
int run_vm(int *, int *, struct vm_create_params *);
void *vcpu_run_loop(void *);
int vcpu_exit(struct vm_run_params *);
int vmm_create_vm(struct vm_create_params *);
void init_emulated_hw(struct vm_create_params *, int *, int *);
void vcpu_exit_inout(struct vm_run_params *);
uint8_t vcpu_exit_pci(struct vm_run_params *);
void vcpu_exit_i8253(union vm_exit *);
void vcpu_exit_com(struct vm_run_params *);
void vcpu_process_com_data(union vm_exit *);
void vcpu_process_com_lcr(union vm_exit *);
void vcpu_process_com_lsr(union vm_exit *);
void vcpu_process_com_ier(union vm_exit *);
void vcpu_process_com_mcr(union vm_exit *);
void vcpu_process_com_iir(union vm_exit *);
void vcpu_process_com_msr(union vm_exit *);
void vcpu_process_com_scr(union vm_exit *);

int vmm_dispatch_parent(int, struct privsep_proc *, struct imsg *);
void vmm_run(struct privsep *, struct privsep_proc *, void *);

int con_fd, vm_id;

extern struct vmd *env;

extern char *__progname;

static struct privsep_proc procs[] = {
	{ "parent",	PROC_PARENT,	vmm_dispatch_parent  },
};

pid_t
vmm(struct privsep *ps, struct privsep_proc *p)
{
	return (proc_run(ps, p, procs, nitems(procs), vmm_run, NULL));
}

void
vmm_run(struct privsep *ps, struct privsep_proc *p, void *arg)
{
	if (config_init(ps->ps_env) == -1)
		fatal("failed to initialize configuration");

#if 0
	/*
	 * pledge in the vmm process:
 	 * stdio - for malloc and basic I/O including events.
	 * XXX vmm - for the vmm ioctls and operations
	 */
	if (pledge("stdio vmm", NULL) == -1)
		fatal("pledge");
#endif
}

int
vmm_dispatch_parent(int fd, struct privsep_proc *p, struct imsg *imsg)
{
	struct privsep	*ps = p->p_ps;
	int		 res = 0, cmd = 0;

	switch (imsg->hdr.type) {
	case IMSG_VMDOP_START_VM_REQUEST:
		res = config_getvm(ps, imsg);
		if (res != 0)
			cmd = IMSG_VMDOP_START_VM_RESPONSE;
		break;
	case IMSG_VMDOP_START_VM_DISK:
		res = config_getdisk(ps, imsg);
		if (res != 0)
			cmd = IMSG_VMDOP_START_VM_RESPONSE;
		break;
	case IMSG_VMDOP_START_VM_IF:
		res = config_getif(ps, imsg);
		if (res != 0)
			cmd = IMSG_VMDOP_START_VM_RESPONSE;
		break;
	case IMSG_VMDOP_START_VM_END:
		res = start_vm(imsg);
		cmd = IMSG_VMDOP_START_VM_RESPONSE;
		break;
	case IMSG_VMDOP_TERMINATE_VM_REQUEST:
		res = terminate_vm(imsg);
		cmd = IMSG_VMDOP_TERMINATE_VM_RESPONSE;
		break;
	case IMSG_VMDOP_GET_INFO_VM_REQUEST:
		res = get_info_vm(ps, imsg);
		cmd = IMSG_VMDOP_GET_INFO_VM_END_DATA;
		break;
	default:
		return (-1);
	}

	if (cmd &&
	    proc_compose_imsg(ps, PROC_PARENT, -1, cmd, imsg->hdr.peerid, -1,
	    &res, sizeof(res)) == -1)
		return (-1);

	return (0);
}

/*
 * terminate_vm
 *
 * Requests vmm(4) to terminate the VM whose ID is provided in the
 * supplied vm_terminate_params structure (vtp->vtp_vm_id)
 *
 * Parameters
 *  imsg: The incoming imsg body whose 'data' field contains the 
 *      vm_terminate_params struct
 *
 * Return values:
 *  0: success
 *  !0 : ioctl to vmm(4) failed (eg, ENOENT if the supplied VM is not
 *      valid)
 */
int
terminate_vm(struct imsg *imsg)
{
	struct vm_terminate_params *vtp;

	vtp = (struct vm_terminate_params *)imsg->data;

	if (ioctl(env->vmd_fd, VMM_IOC_TERM, vtp) < 0)
		return (errno);

	return (0);
}

/*
 * opentap
 *
 * Opens the next available tap device, up to MAX_TAP.
 *
 * Returns a file descriptor to the tap node opened, or -1 if no tap
 * devices were available.
 */
int
opentap(void)
{
	int i, fd;
	char path[PATH_MAX];

	for (i = 0; i < MAX_TAP; i++) {
		snprintf(path, PATH_MAX, "/dev/tap%d", i);
		fd = open(path, O_RDWR | O_NONBLOCK);
		if (fd != -1)
			return (fd);
	}

	return (-1);
}

/*
 * start_vm
 *
 * Starts a new VM with the creation parameters supplied (in the incoming
 * imsg->data field). This function performs a basic sanity check on the
 * incoming parameters and then performs the following steps to complete
 * the creation of the VM:
 *
 * 1. opens the VM disk image files specified in the VM creation parameters
 * 2. opens the specified VM kernel
 * 3. creates a VM console tty pair using openpty
 * 4. forks, passing the file descriptors opened in steps 1-3 to the child
 *     vmd responsible for dropping privilege and running the VM's VCPU
 *     loops.
 *
 * Parameters:
 *  imsg: The incoming imsg body whose 'data' field is a vm_create_params
 *      struct containing the VM creation parameters.
 *
 * Return values:
 *  0: success
 *  !0 : failure - typically an errno indicating the source of the failure
 */
int
start_vm(struct imsg *imsg)
{
	struct vm_create_params	*vcp;
	struct vmd_vm		*vm;
	size_t			 i;
	int			 ret = EINVAL;

	if ((vm = vm_getbyvmid(imsg->hdr.peerid)) == NULL) {
		log_warn("%s: can't find vm", __func__);
		return (-1);
	}
	vcp = &vm->vm_params;

	if ((vm->vm_tty = imsg->fd) == -1) {
		log_warn("%s: can't get tty", __func__);
		goto err;
	}

	/* Start child vmd for this VM (fork, chroot, drop privs) */
	ret = start_client_vmd();

	/* Start child failed? - cleanup and leave */
	if (ret == -1) {
		log_warn("%s: start child failed", __func__);
		ret = EIO;
		goto err;
	}

	if (ret > 0) {
		/* Parent */
		for (i = 0 ; i < vcp->vcp_ndisks; i++) {
			close(vm->vm_disks[i]);
			vm->vm_disks[i] = -1;
		}

		for (i = 0 ; i < vcp->vcp_nnics; i++) {
			close(vm->vm_ifs[i]);
			vm->vm_ifs[i] = -1;
		}

		close(vm->vm_kernel);
		vm->vm_kernel = -1;

		close(vm->vm_tty);
		vm->vm_tty = -1;

		return (0);
	} else {
		/* Child */
		setproctitle(vcp->vcp_name);
		log_procinit(vcp->vcp_name);

		ret = vmm_create_vm(vcp);
		if (ret) {
			errno = ret;
			fatal("create vmm ioctl failed - exiting");
		}

		/* Load kernel image */
		ret = loadelf_main(vm->vm_kernel,
		    vcp->vcp_id, vcp->vcp_memory_size);
		if (ret) {
			errno = ret;
			fatal("failed to load kernel - exiting");
		}

		close(vm->vm_kernel);

		con_fd = vm->vm_tty;
		if (fcntl(con_fd, F_SETFL, O_NONBLOCK) == -1)
			fatal("failed to set nonblocking mode on console");

		/* Execute the vcpu run loop(s) for this VM */
		ret = run_vm(vm->vm_disks, vm->vm_ifs, vcp);

		_exit(ret != 0);
	}

	return (0);

 err:
	vm_remove(vm);

	return (ret);
}

/*
 * get_info_vm
 *
 * Returns a list of VMs known to vmm(4).
 *
 * Parameters:
 *  ibuf: the imsg ibuf in which to place the results. A new imsg will
 *      be created using this ibuf.
 *
 * Return values:
 *  0: success
 *  !0 : failure (eg, ENOMEM, EIO or another error code from vmm(4) ioctl)
 */
int
get_info_vm(struct privsep *ps, struct imsg *imsg)
{
	int ret;
	size_t ct, i;
	struct vm_info_params vip;
	struct vm_info_result *info;

	/*
	 * We issue the VMM_IOC_INFO ioctl twice, once with an input
	 * buffer size of 0, which results in vmm(4) returning the
	 * number of bytes required back to us in vip.vip_size,
	 * and then we call it again after malloc'ing the required
	 * number of bytes.
	 *
	 * It is possible that we could fail a second time (eg, if
	 * another VM was created in the instant between the two
	 * ioctls, but in that case the caller can just try again
	 * as vmm(4) will return a zero-sized list in that case.
	 */
	vip.vip_size = 0;
	info = NULL;
	ret = 0;

	/* First ioctl to see how many bytes needed (vip.vip_size) */
	if (ioctl(env->vmd_fd, VMM_IOC_INFO, &vip) < 0)
		return (errno);

	if (vip.vip_info_ct != 0)
		return (EIO);

	info = malloc(vip.vip_size);
	if (info == NULL)
		return (ENOMEM);

	/* Second ioctl to get the actual list */
	vip.vip_info = info;
	if (ioctl(env->vmd_fd, VMM_IOC_INFO, &vip) < 0) {
		ret = errno;
		free(info);
		return (ret);
	}

	/* Return info */
	ct = vip.vip_size / sizeof(struct vm_info_result);
	for (i = 0; i < ct; i++) {
		if (proc_compose_imsg(ps, PROC_PARENT, -1,
		    IMSG_VMDOP_GET_INFO_VM_DATA, imsg->hdr.peerid, -1,
		    &info[i], sizeof(struct vm_info_result)) == -1)
			return (EIO);
	}
	free(info);
	return (0);
}


/*
 * start_client_vmd
 *
 * forks a copy of the parent vmd, chroots to VMD_USER's home, drops
 * privileges (changes to user VMD_USER), and returns.
 * Should the fork operation succeed, but later chroot/privsep
 * fail, the child exits.
 *
 * Return values (returns to both child and parent on success):
 *  -1 : failure
 *  0: return to child vmd returns 0
 *  !0 : return to parent vmd returns the child's pid
 */
int
start_client_vmd(void)
{
	int child_pid;

	child_pid = fork();
	if (child_pid < 0)
		return (-1);

	if (!child_pid) {
		/* child, already running without privileges */
		return (0);
	}

	/* Parent */
	return (child_pid);
}

/*
 * vmm_create_vm
 *
 * Requests vmm(4) to create a new VM using the supplied creation
 * parameters. This operation results in the creation of the in-kernel
 * structures for the VM, but does not start the VM's vcpu(s).
 *
 * Parameters:
 *  vcp: vm_create_params struct containing the VM's desired creation
 *      configuration
 *
 * Return values:
 *  0: success
 *  !0 : ioctl to vmm(4) failed
 */
int
vmm_create_vm(struct vm_create_params *vcp)
{
	/* Sanity check arguments */
	if (vcp->vcp_ncpus > VMM_MAX_VCPUS_PER_VM)
		return (EINVAL);

	if (vcp->vcp_memory_size > VMM_MAX_VM_MEM_SIZE)
		return (EINVAL);

	if (vcp->vcp_ndisks > VMM_MAX_DISKS_PER_VM)
		return (EINVAL);

	if (ioctl(env->vmd_fd, VMM_IOC_CREATE, vcp) < 0)
		return (errno);

	return (0);
}

/*
 * init_emulated_hw
 *
 * Initializes the userspace hardware emulation
 */
void
init_emulated_hw(struct vm_create_params *vcp, int *child_disks,
    int *child_taps)
{
	/* Init the i8253 PIT's 3 counters */
	bzero(&i8253_counter, sizeof(struct i8253_counter) * 3);
	gettimeofday(&i8253_counter[0].tv, NULL);
	gettimeofday(&i8253_counter[1].tv, NULL);
	gettimeofday(&i8253_counter[2].tv, NULL);
	i8253_counter[0].start = TIMER_DIV(100);
	i8253_counter[1].start = TIMER_DIV(100);
	i8253_counter[2].start = TIMER_DIV(100);

	/* Init ns8250 UART */
	bzero(&com1_regs, sizeof(struct ns8250_regs));

	/* Initialize PCI */
	pci_init();

	/* Initialize virtio devices */
	virtio_init(vcp, child_disks, child_taps);
}

/*
 * run_vm
 *
 * Runs the VM whose creation parameters are specified in vcp
 *
 * Parameters:
 *  vcp: vm_create_params struct containing the VM's desired creation
 *      configuration
 *  child_disks: previously-opened child VM disk file file descriptors
 *  child_taps: previously-opened child tap file descriptors
 *
 * Return values:
 *  0: the VM exited normally
 *  !0 : the VM exited abnormally or failed to start
 */
int
run_vm(int *child_disks, int *child_taps, struct vm_create_params *vcp)
{
	size_t i;
	int ret;
	pthread_t *tid;
	void *exit_status;
	struct vm_run_params **vrp;

	ret = 0;

	/* XXX cap vcp_ncpus to avoid overflow here */
	/*
	 * XXX ensure nvcpus in vcp is same as vm, or fix vmm to return einval
	 * on bad vcpu id
	 */
	tid = malloc(sizeof(pthread_t) * vcp->vcp_ncpus);
	vrp = malloc(sizeof(struct vm_run_params *) * vcp->vcp_ncpus);
	if (tid == NULL || vrp == NULL) {
		log_warn("%s: memory allocation error - exiting.",
		    __progname);
		return (ENOMEM);
	}

	init_emulated_hw(vcp, child_disks, child_taps);

	/*
	 * Create and launch one thread for each VCPU. These threads may
	 * migrate between PCPUs over time; the need to reload CPU state
	 * in such situations is detected and performed by vmm(4) in the
	 * kernel.
	 */
	for (i = 0 ; i < vcp->vcp_ncpus; i++) {
		vrp[i] = malloc(sizeof(struct vm_run_params));
		if (vrp[i] == NULL) {
			log_warn("%s: memory allocation error - "
			    "exiting.", __progname);
			/* caller will exit, so skip free'ing */
			return (ENOMEM);
		}
		vrp[i]->vrp_exit = malloc(sizeof(union vm_exit));
		if (vrp[i]->vrp_exit == NULL) {
			log_warn("%s: memory allocation error - "
			    "exiting.", __progname);
			/* caller will exit, so skip free'ing */
			return (ENOMEM);
		}
		vrp[i]->vrp_vm_id = vcp->vcp_id;
		vrp[i]->vrp_vcpu_id = i;

		/* Start each VCPU run thread at vcpu_run_loop */
		ret = pthread_create(&tid[i], NULL, vcpu_run_loop, vrp[i]);
		if (ret) {
			/* caller will _exit after this return */
			return (ret);
		}
	}

	/* Wait for all the threads to exit */
	for (i = 0; i < vcp->vcp_ncpus; i++) {
		if (pthread_join(tid[i], &exit_status)) {
			log_warn("%s: failed to join thread %zd - "
			    "exiting", __progname, i);
			return (EIO);
		}

		if (exit_status != NULL) {
			log_warnx("%s: vm %d vcpu run thread %zd exited "
			    "abnormally", __progname, vcp->vcp_id, i);
			ret = EIO;
		}
	}

	return (ret);
}

/*
 * vcpu_run_loop
 *
 * Runs a single VCPU until vmm(4) requires help handling an exit,
 * or the VM terminates.
 *
 * Parameters:
 *  arg: vcpu_run_params for the VCPU being run by this thread
 *
 * Return values:
 *  NULL: the VCPU shutdown properly
 *  !NULL: error processing VCPU run, or the VCPU shutdown abnormally
 */
void *
vcpu_run_loop(void *arg)
{
	struct vm_run_params *vrp = (struct vm_run_params *)arg;
	intptr_t ret;

	vrp->vrp_continue = 0;
	vrp->vrp_injint = -1;

	for (;;) {
		if (ioctl(env->vmd_fd, VMM_IOC_RUN, vrp) < 0) {
			/* If run ioctl failed, exit */
			ret = errno;
			return ((void *)ret);
		}

		/* If the VM is terminating, exit normally */
		if (vrp->vrp_exit_reason == VM_EXIT_TERMINATED)
			return (NULL);

		if (vrp->vrp_exit_reason != VM_EXIT_NONE) {
			/*
			 * vmm(4) needs help handling an exit, handle in
			 * vcpu_exit.
			 */
			if (vcpu_exit(vrp))
				return ((void *)EIO);
		}
	}

	return (NULL);
}

/*
 * vcpu_exit_i8253
 *
 * Handles emulated i8253 PIT access (in/out instruction to PIT ports).
 * We don't emulate all the modes of the i8253, just the basic squarewave
 * clock.
 *
 * Parameters:
 *  vei: VM exit information from vmm(4) containing information on the in/out
 *      instruction being performed
 */
void
vcpu_exit_i8253(union vm_exit *vei)
{
	uint32_t out_data;
	uint8_t sel, rw, data;
	uint64_t ns, ticks;
	struct timeval now, delta;

	if (vei->vei.vei_port == TIMER_CTRL) {
		if (vei->vei.vei_dir == 0) { /* OUT instruction */
			out_data = vei->vei.vei_data;
			sel = out_data &
			    (TIMER_SEL0 | TIMER_SEL1 | TIMER_SEL2);
			sel = sel >> 6;
			if (sel > 2) {
				log_warnx("%s: i8253 PIT: invalid "
				    "timer selected (%d)",
				    __progname, sel);
				return;
			}

			rw = vei->vei.vei_data &
			    (TIMER_LATCH | TIMER_LSB |
			    TIMER_MSB | TIMER_16BIT);

			if (rw == TIMER_16BIT) {
				/*
				 * XXX this seems to be used on occasion, needs
				 * to be implemented
				 */
				log_warnx("%s: i8253 PIT: 16 bit "
				    "counter I/O not supported",
				    __progname);
				    return;
			}

			/*
			 * Since we don't truly emulate each tick of the PIT
			 * clock, when the guest asks for the timer to be
			 * latched, simulate what the counter would have been
			 * had we performed full emulation. We do this by
			 * calculating when the counter was reset vs how much
			 * time has elapsed, then bias by the counter tick
			 * rate.
			 */
			if (rw == TIMER_LATCH) {
				gettimeofday(&now, NULL);
				delta.tv_sec = now.tv_sec -
				    i8253_counter[sel].tv.tv_sec;
				delta.tv_usec = now.tv_usec -
				    i8253_counter[sel].tv.tv_usec;
				if (delta.tv_usec < 0) {
					delta.tv_sec--;
					delta.tv_usec += 1000000;
				}
				if (delta.tv_usec > 1000000) {
					delta.tv_sec++;
					delta.tv_usec -= 1000000;
				}
				ns = delta.tv_usec * 1000 +
				    delta.tv_sec * 1000000000;
				ticks = ns / NS_PER_TICK;
				i8253_counter[sel].olatch =
				    i8253_counter[sel].start -
				    ticks % i8253_counter[sel].start;
				return;
			}

			log_warnx("%s: i8253 PIT: unsupported rw mode "
			    "%d", __progname, rw);
			return;
		} else {
			/* XXX should this return 0xff? */
			log_warnx("%s: i8253 PIT: read from control "
			    "port unsupported", __progname);
		}
	} else {
		sel = vei->vei.vei_port - (TIMER_CNTR0 + TIMER_BASE);
		if (vei->vei.vei_dir == 0) { /* OUT instruction */
			if (i8253_counter[sel].last_w == 0) {
				out_data = vei->vei.vei_data;
				i8253_counter[sel].ilatch |= (out_data << 8);
				i8253_counter[sel].last_w = 1;
			} else {
				out_data = vei->vei.vei_data;
				i8253_counter[sel].ilatch |= out_data;
				i8253_counter[sel].start =
				    i8253_counter[sel].ilatch;
				i8253_counter[sel].last_w = 0;
			}
		} else {
			if (i8253_counter[sel].last_r == 0) {
				data = i8253_counter[sel].olatch >> 8;
				vei->vei.vei_data = data;
				i8253_counter[sel].last_w = 1;
			} else {
				data = i8253_counter[sel].olatch & 0xFF;
				vei->vei.vei_data = data;
				i8253_counter[sel].last_w = 0;
			}
		}
	}
}

/*
 * vcpu_process_com_data
 *
 * Emulate in/out instructions to the com1 (ns8250) UART data register
 *
 * Parameters:
 *  vei: vm exit information from vmm(4) containing information on the in/out
 *      instruction being performed
 */
void
vcpu_process_com_data(union vm_exit *vei)
{
	/*
	 * vei_dir == 0 : out instruction
	 *
	 * The guest wrote to the data register. Since we are emulating a
	 * no-fifo chip, write the character immediately to the pty and
	 * assert TXRDY in IIR (if the guest has requested TXRDY interrupt
	 * reporting)
	 */
	if (vei->vei.vei_dir == 0) {
		write(con_fd, &vei->vei.vei_data, 1);
		if (com1_regs.ier & 0x2) {
			/* Set TXRDY */
			com1_regs.iir |= IIR_TXRDY;
			/* Set "interrupt pending" (IIR low bit cleared) */
			com1_regs.iir &= ~0x1;
		}
	} else {
		/*
		 * vei_dir == 1 : in instruction
		 *
		 * The guest read from the data register. Check to see if
		 * there is data available (RXRDY) and if so, consume the
		 * input data and return to the guest. Also clear the
		 * interrupt info register regardless.
		 */
		if (com1_regs.lsr & LSR_RXRDY) {
			vei->vei.vei_data = com1_regs.data;
			com1_regs.data = 0x0;
			com1_regs.lsr &= ~LSR_RXRDY;
		} else {
			/* XXX should this be com1_regs.data or 0xff? */
			vei->vei.vei_data = com1_regs.data;
			log_warnx("guest reading com1 when not ready");
		}

		/* Reading the data register always clears RXRDY from IIR */
		com1_regs.iir &= ~IIR_RXRDY;

		/*
		 * Clear "interrupt pending" by setting IIR low bit to 1
		 * if no interrupt are pending
		 */
		if (com1_regs.iir == 0x0)
			com1_regs.iir = 0x1;
	}
}

/*
 * vcpu_process_com_lcr
 *
 * Emulate in/out instructions to the com1 (ns8250) UART line control register
 *
 * Paramters:
 *  vei: vm exit information from vmm(4) containing information on the in/out
 *      instruction being performed
 */
void
vcpu_process_com_lcr(union vm_exit *vei)
{
	/*
	 * vei_dir == 0 : out instruction
	 *
	 * Write content to line control register
	 */
	if (vei->vei.vei_dir == 0) {
		com1_regs.lcr = (uint8_t)vei->vei.vei_data;
	} else {
		/*
		 * vei_dir == 1 : in instruction
		 *
		 * Read line control register
		 */
		vei->vei.vei_data = com1_regs.lcr;
	}
}

/*
 * vcpu_process_com_iir
 *
 * Emulate in/out instructions to the com1 (ns8250) UART interrupt information
 * register. Note that writes to this register actually are to a different
 * register, the FCR (FIFO control register) that we don't emulate but still
 * consume the data provided.
 *
 * Parameters:
 *  vei: vm exit information from vmm(4) containing information on the in/out
 *      instruction being performed
 */
void
vcpu_process_com_iir(union vm_exit *vei)
{
	/*
	 * vei_dir == 0 : out instruction
	 *
	 * Write to FCR
	 */
	if (vei->vei.vei_dir == 0) {
		com1_regs.fcr = vei->vei.vei_data;
	} else {
		/*
		 * vei_dir == 1 : in instruction
		 *
		 * Read IIR. Reading the IIR resets the TXRDY bit in the IIR
		 * after the data is read.
		 */
		vei->vei.vei_data = com1_regs.iir;
		com1_regs.iir &= ~IIR_TXRDY;

		/*
		 * Clear "interrupt pending" by setting IIR low bit to 1
		 * if no interrupts are pending
		 */
		if (com1_regs.iir == 0x0)
			com1_regs.iir = 0x1;
	}
}

/*
 * vcpu_process_com_mcr
 *
 * Emulate in/out instructions to the com1 (ns8250) UART modem control
 * register.
 *
 * Parameters:
 *  vei: vm exit information from vmm(4) containing information on the in/out
 *      instruction being performed
 */
void
vcpu_process_com_mcr(union vm_exit *vei)
{
	/*
	 * vei_dir == 0 : out instruction
	 *
	 * Write to MCR
	 */
	if (vei->vei.vei_dir == 0) {
		com1_regs.mcr = vei->vei.vei_data;
	} else {
		/*
		 * vei_dir == 1 : in instruction
		 *
		 * Read from MCR
		 */
		vei->vei.vei_data = com1_regs.mcr;
	}
}

/*
 * vcpu_process_com_lsr
 *
 * Emulate in/out instructions to the com1 (ns8250) UART line status register.
 *
 * Parameters:
 *  vei: vm exit information from vmm(4) containing information on the in/out
 *      instruction being performed
 */
void
vcpu_process_com_lsr(union vm_exit *vei)
{
	/*
	 * vei_dir == 0 : out instruction
	 *
	 * Write to LSR. This is an illegal operation, so we just log it and
	 * continue.
	 */
	if (vei->vei.vei_dir == 0) {
		log_warnx("%s: LSR UART write 0x%x unsupported",
		    __progname, vei->vei.vei_data);
	} else {
		/*
		 * vei_dir == 1 : in instruction
		 *
		 * Read from LSR. We always report TXRDY and TSRE since we
		 * can process output characters immediately (at any time).
		 */
		vei->vei.vei_data = com1_regs.lsr | LSR_TSRE | LSR_TXRDY;
	}
}

/*
 * vcpu_process_com_msr
 *
 * Emulate in/out instructions to the com1 (ns8250) UART modem status register.
 *
 * Parameters:
 *  vei: vm exit information from vmm(4) containing information on the in/out
 *      instruction being performed
 */
void
vcpu_process_com_msr(union vm_exit *vei)
{
	/*
	 * vei_dir == 0 : out instruction
	 *
	 * Write to MSR. This is an illegal operation, so we just log it and
	 * continue.
	 */
	if (vei->vei.vei_dir == 0) {
		log_warnx("%s: MSR UART write 0x%x unsupported",
		    __progname, vei->vei.vei_data);
	} else {
		/*
		 * vei_dir == 1 : in instruction
		 *
		 * Read from MSR. We always report DCD, DSR, and CTS.
		 */
		vei->vei.vei_data = com1_regs.lsr | MSR_DCD | MSR_DSR | MSR_CTS;
	}
}

/*
 * vcpu_process_com_scr
 *
 * Emulate in/out instructions to the com1 (ns8250) UART scratch register. The
 * scratch register is sometimes used to distinguish an 8250 from a 16450,
 * and/or used to distinguish submodels of the 8250 (eg 8250A, 8250B). We
 * simulate an "original" 8250 by forcing the scratch register to return data
 * on read that is different from what was written.
 *
 * Parameters:
 *  vei: vm exit information from vmm(4) containing information on the in/out
 *      instruction being performed
 */
void
vcpu_process_com_scr(union vm_exit *vei)
{
	/*
	 * vei_dir == 0 : out instruction
	 *
	 * Write to SCR
	 */
	if (vei->vei.vei_dir == 0) {
		com1_regs.scr = vei->vei.vei_data;
	} else {
		/*
		 * vei_dir == 1 : in instruction
		 *
		 * Read from SCR. To make sure we don't accidentally simulate
		 * a real scratch register, we negate what was written on
		 * subsequent readback.
		 */
		vei->vei.vei_data = ~com1_regs.scr;
	}
}

/*
 * vcpu_process_com_ier
 *
 * Emulate in/out instructions to the com1 (ns8250) UART interrupt enable
 * register.
 *
 * Parameters:
 *  vei: vm exit information from vmm(4) containing information on the in/out
 *      instruction being performed
 */
void
vcpu_process_com_ier(union vm_exit *vei)
{
	/*
	 * vei_dir == 0 : out instruction
	 *
	 * Write to IER
	 */
	if (vei->vei.vei_dir == 0) {
		com1_regs.ier = vei->vei.vei_data;
	} else {
		/*
		 * vei_dir == 1 : in instruction
		 *
		 * Read from IER
		 */
		vei->vei.vei_data = com1_regs.ier;
	}
}

/*
 * vcpu_exit_com
 *
 * Process com1 (ns8250) UART exits. vmd handles most basic 8250
 * features with the exception of the divisor latch (eg, no baud
 * rate support)
 *
 * Parameters:
 *  vrp: vcpu run parameters containing guest state for this exit
 */
void
vcpu_exit_com(struct vm_run_params *vrp)
{
	union vm_exit *vei = vrp->vrp_exit;

	switch (vei->vei.vei_port) {
	case COM1_LCR:
		vcpu_process_com_lcr(vei);
		break;
	case COM1_IER:
		vcpu_process_com_ier(vei);
		break;
	case COM1_IIR:
		vcpu_process_com_iir(vei);
		break;
	case COM1_MCR:
		vcpu_process_com_mcr(vei);
		break;
	case COM1_LSR:
		vcpu_process_com_lsr(vei);
		break;
	case COM1_MSR:
		vcpu_process_com_msr(vei);
		break;
	case COM1_SCR:
		vcpu_process_com_scr(vei);
		break;
	case COM1_DATA:
		vcpu_process_com_data(vei);
		break;
	}
}

/*
 * vcpu_exit_pci
 *
 * Handle all I/O to the emulated PCI subsystem.
 *
 * Parameters:
 *  vrp: vcpu run paramters containing guest state for this exit
 *
 * Return values:
 *  0xff if no interrupt is required after this pci exit,
 *      or an interrupt vector otherwise
 */
uint8_t
vcpu_exit_pci(struct vm_run_params *vrp)
{
	union vm_exit *vei = vrp->vrp_exit;
	uint8_t intr;

	intr = 0xFF;

	switch (vei->vei.vei_port) {
	case PCI_MODE1_ADDRESS_REG:
		pci_handle_address_reg(vrp);
		break;
	case PCI_MODE1_DATA_REG:
		pci_handle_data_reg(vrp);
		break;
	case VMM_PCI_IO_BAR_BASE ... VMM_PCI_IO_BAR_END:
		intr = pci_handle_io(vrp);
		break;
	default:
		log_warnx("%s: unknown PCI register 0x%llx",
		    __progname, (uint64_t)vei->vei.vei_port);
		break;
	}

	return (intr);
}

/*
 * vcpu_exit_inout
 *
 * Handle all I/O exits that need to be emulated in vmd. This includes the
 * i8253 PIT and the com1 ns8250 UART.
 *
 * Parameters:
 *  vrp: vcpu run parameters containing guest state for this exit
 */
void
vcpu_exit_inout(struct vm_run_params *vrp)
{
	union vm_exit *vei = vrp->vrp_exit;
	uint8_t intr;

	switch (vei->vei.vei_port) {
	case TIMER_CTRL:
	case (TIMER_CNTR0 + TIMER_BASE):
	case (TIMER_CNTR1 + TIMER_BASE):
	case (TIMER_CNTR2 + TIMER_BASE):
		vcpu_exit_i8253(vei);
		break;
	case COM1_DATA ... COM1_SCR:
		vcpu_exit_com(vrp);
		break;
	case PCI_MODE1_ADDRESS_REG:
	case PCI_MODE1_DATA_REG:
	case VMM_PCI_IO_BAR_BASE ... VMM_PCI_IO_BAR_END:
		intr = vcpu_exit_pci(vrp);
		if (intr != 0xFF)
			vrp->vrp_injint = intr;
		else
			vrp->vrp_injint = -1;
		break;
	default:
		/* IN from unsupported port gives FFs */
		if (vei->vei.vei_dir == 1)
			vei->vei.vei_data = 0xFFFFFFFF;
		break;
	}
}

/*
 * vcpu_exit
 *
 * Handle a vcpu exit. This function is called when it is determined that
 * vmm(4) requires the assistance of vmd to support a particular guest
 * exit type (eg, accessing an I/O port or device). Guest state is contained
 * in 'vrp', and will be resent to vmm(4) on exit completion.
 *
 * Upon conclusion of handling the exit, the function determines if any
 * interrupts should be injected into the guest, and sets vrp->vrp_injint
 * to the IRQ line whose interrupt should be vectored (or -1 if no interrupt
 * is to be injected).
 *
 * Parameters:
 *  vrp: vcpu run parameters containing guest state for this exit
 *
 * Return values:
 *  0: the exit was handled successfully
 *  1: an error occurred (exit not handled)
 */
int
vcpu_exit(struct vm_run_params *vrp)
{
	ssize_t sz;
	char ch;

	switch (vrp->vrp_exit_reason) {
	case VMX_EXIT_IO:
		vcpu_exit_inout(vrp);
		break;
	case VMX_EXIT_HLT:
		/*
		 * XXX handle halted state, no reason to run this vcpu again
		 * until a vm interrupt is to be injected
		 */
		break;
	default:
		log_warnx("%s: unknown exit reason %d",
		    __progname, vrp->vrp_exit_reason);
		return (1);
	}

	/* XXX interrupt priority */
	if (vionet_process_rx())
		vrp->vrp_injint = 9;

	/*
	 * Is there a new character available on com1?
	 * If so, consume the character, buffer it into the com1 data register
	 * assert IRQ4, and set the line status register RXRDY bit.
	 *
	 * XXX - move all this com intr checking to another function
	 */
	sz = read(con_fd, &ch, sizeof(char));
	if (sz == 1) {
		com1_regs.lsr |= LSR_RXRDY;
		com1_regs.data = ch;
		/* XXX these ier and iir bits should be IER_x and IIR_x */
		if (com1_regs.ier & 0x1) {
			com1_regs.iir |= (2 << 1);
			com1_regs.iir &= ~0x1;
		}
	}

	/*
	 * Clear "interrupt pending" by setting IIR low bit to 1 if no
	 * interrupts are pending
	 */
	/* XXX these iir magic numbers should be IIR_x */
	if ((com1_regs.iir & ~0x1) == 0x0)
		com1_regs.iir = 0x1;

	/* If pending interrupt and nothing waiting to be injected, inject */
	if ((com1_regs.iir & 0x1) == 0)
		if (vrp->vrp_injint == -1)
			vrp->vrp_injint = 0x4;
	vrp->vrp_continue = 1;

	return (0);
}

/*
 * write_page
 *
 * Pushes a page of data from 'buf' into the guest VM's memory
 * at paddr 'dst'.
 *
 * Parameters:
 *  dst: the destination paddr_t in the guest VM to push into.
 *      If there is no guest paddr mapping at 'dst', a new page will be
 *      faulted in by the VMM (provided 'dst' represents a valid paddr
 *      in the guest's address space)
 *  buf: page of data to push
 *  len: size of 'buf'
 *  do_mask: 1 to mask the destination address (for kernel load), 0 to
 *      leave 'dst' unmasked
 *
 * Return values:
 *  various return values from ioctl(VMM_IOC_WRITEPAGE), or 0 if no error
 *      occurred.
 *
 * Note - this function only handles GPAs < 4GB. 
 */
int
write_page(uint32_t dst, void *buf, uint32_t len, int do_mask)
{
	struct vm_writepage_params vwp;

	/*
	 * Mask kernel load addresses to avoid uint32_t -> uint64_t cast
	 * errors
	 */
	if (do_mask)
		dst &= 0xFFFFFFF;

	vwp.vwp_paddr = (paddr_t)dst;
	vwp.vwp_data = buf;
	vwp.vwp_vm_id = vm_id;
	vwp.vwp_len = len;
	if (ioctl(env->vmd_fd, VMM_IOC_WRITEPAGE, &vwp) < 0) {
		log_warn("writepage ioctl failed");
		return (errno);
	}
	return (0);
}

/*
 * read_page
 *
 * Reads a page of memory at guest paddr 'src' into 'buf'.
 *
 * Parameters:
 *  src: the source paddr_t in the guest VM to read from.
 *  buf: destination (local) buffer
 *  len: size of 'buf'
 *  do_mask: 1 to mask the source address (for kernel load), 0 to
 *      leave 'src' unmasked
 *
 * Return values:
 *  various return values from ioctl(VMM_IOC_READPAGE), or 0 if no error
 *      occurred.
 *
 * Note - this function only handles GPAs < 4GB.
 */
int
read_page(uint32_t src, void *buf, uint32_t len, int do_mask)
{
	struct vm_readpage_params vrp;

	/*
	 * Mask kernel load addresses to avoid uint32_t -> uint64_t cast
	 * errors
	 */
	if (do_mask)
		src &= 0xFFFFFFF;

	vrp.vrp_paddr = (paddr_t)src;
	vrp.vrp_data = buf;
	vrp.vrp_vm_id = vm_id;
	vrp.vrp_len = len;
	if (ioctl(env->vmd_fd, VMM_IOC_READPAGE, &vrp) < 0) {
		log_warn("readpage ioctl failed");
		return (errno);
	}
	return (0);
}