1/*-
2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3 *
4 * Copyright (c) 2011 NetApp, Inc.
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 *    notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 *    notice, this list of conditions and the following disclaimer in the
14 *    documentation and/or other materials provided with the distribution.
15 *
16 * THIS SOFTWARE IS PROVIDED BY NETAPP, INC ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED.  IN NO EVENT SHALL NETAPP, INC OR CONTRIBUTORS BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26 * SUCH DAMAGE.
27 *
28 * $FreeBSD$
29 */
30
31#include <sys/cdefs.h>
32__FBSDID("$FreeBSD$");
33
34#include <sys/types.h>
35#ifndef WITHOUT_CAPSICUM
36#include <sys/capsicum.h>
37#endif
38#include <sys/mman.h>
39#include <sys/time.h>
40
41#include <amd64/vmm/intel/vmcs.h>
42
43#include <machine/atomic.h>
44#include <machine/segments.h>
45
46#ifndef WITHOUT_CAPSICUM
47#include <capsicum_helpers.h>
48#endif
49#include <stdio.h>
50#include <stdlib.h>
51#include <string.h>
52#include <err.h>
53#include <errno.h>
54#include <libgen.h>
55#include <unistd.h>
56#include <assert.h>
57#include <pthread.h>
58#include <pthread_np.h>
59#include <sysexits.h>
60#include <stdbool.h>
61#include <stdint.h>
62
63#include <machine/vmm.h>
64#ifndef WITHOUT_CAPSICUM
65#include <machine/vmm_dev.h>
66#endif
67#include <vmmapi.h>
68
69#include "bhyverun.h"
70#include "acpi.h"
71#include "atkbdc.h"
72#include "inout.h"
73#include "dbgport.h"
74#include "fwctl.h"
75#include "gdb.h"
76#include "ioapic.h"
77#include "mem.h"
78#include "mevent.h"
79#include "mptbl.h"
80#include "pci_emul.h"
81#include "pci_irq.h"
82#include "pci_lpc.h"
83#include "smbiostbl.h"
84#include "xmsr.h"
85#include "spinup_ap.h"
86#include "rtc.h"
87
88#define GUEST_NIO_PORT		0x488	/* guest upcalls via i/o port */
89
90#define MB		(1024UL * 1024)
91#define GB		(1024UL * MB)
92
93static const char * const vmx_exit_reason_desc[] = {
94	[EXIT_REASON_EXCEPTION] = "Exception or non-maskable interrupt (NMI)",
95	[EXIT_REASON_EXT_INTR] = "External interrupt",
96	[EXIT_REASON_TRIPLE_FAULT] = "Triple fault",
97	[EXIT_REASON_INIT] = "INIT signal",
98	[EXIT_REASON_SIPI] = "Start-up IPI (SIPI)",
99	[EXIT_REASON_IO_SMI] = "I/O system-management interrupt (SMI)",
100	[EXIT_REASON_SMI] = "Other SMI",
101	[EXIT_REASON_INTR_WINDOW] = "Interrupt window",
102	[EXIT_REASON_NMI_WINDOW] = "NMI window",
103	[EXIT_REASON_TASK_SWITCH] = "Task switch",
104	[EXIT_REASON_CPUID] = "CPUID",
105	[EXIT_REASON_GETSEC] = "GETSEC",
106	[EXIT_REASON_HLT] = "HLT",
107	[EXIT_REASON_INVD] = "INVD",
108	[EXIT_REASON_INVLPG] = "INVLPG",
109	[EXIT_REASON_RDPMC] = "RDPMC",
110	[EXIT_REASON_RDTSC] = "RDTSC",
111	[EXIT_REASON_RSM] = "RSM",
112	[EXIT_REASON_VMCALL] = "VMCALL",
113	[EXIT_REASON_VMCLEAR] = "VMCLEAR",
114	[EXIT_REASON_VMLAUNCH] = "VMLAUNCH",
115	[EXIT_REASON_VMPTRLD] = "VMPTRLD",
116	[EXIT_REASON_VMPTRST] = "VMPTRST",
117	[EXIT_REASON_VMREAD] = "VMREAD",
118	[EXIT_REASON_VMRESUME] = "VMRESUME",
119	[EXIT_REASON_VMWRITE] = "VMWRITE",
120	[EXIT_REASON_VMXOFF] = "VMXOFF",
121	[EXIT_REASON_VMXON] = "VMXON",
122	[EXIT_REASON_CR_ACCESS] = "Control-register accesses",
123	[EXIT_REASON_DR_ACCESS] = "MOV DR",
124	[EXIT_REASON_INOUT] = "I/O instruction",
125	[EXIT_REASON_RDMSR] = "RDMSR",
126	[EXIT_REASON_WRMSR] = "WRMSR",
127	[EXIT_REASON_INVAL_VMCS] =
128	    "VM-entry failure due to invalid guest state",
129	[EXIT_REASON_INVAL_MSR] = "VM-entry failure due to MSR loading",
130	[EXIT_REASON_MWAIT] = "MWAIT",
131	[EXIT_REASON_MTF] = "Monitor trap flag",
132	[EXIT_REASON_MONITOR] = "MONITOR",
133	[EXIT_REASON_PAUSE] = "PAUSE",
134	[EXIT_REASON_MCE_DURING_ENTRY] =
135	    "VM-entry failure due to machine-check event",
136	[EXIT_REASON_TPR] = "TPR below threshold",
137	[EXIT_REASON_APIC_ACCESS] = "APIC access",
138	[EXIT_REASON_VIRTUALIZED_EOI] = "Virtualized EOI",
139	[EXIT_REASON_GDTR_IDTR] = "Access to GDTR or IDTR",
140	[EXIT_REASON_LDTR_TR] = "Access to LDTR or TR",
141	[EXIT_REASON_EPT_FAULT] = "EPT violation",
142	[EXIT_REASON_EPT_MISCONFIG] = "EPT misconfiguration",
143	[EXIT_REASON_INVEPT] = "INVEPT",
144	[EXIT_REASON_RDTSCP] = "RDTSCP",
145	[EXIT_REASON_VMX_PREEMPT] = "VMX-preemption timer expired",
146	[EXIT_REASON_INVVPID] = "INVVPID",
147	[EXIT_REASON_WBINVD] = "WBINVD",
148	[EXIT_REASON_XSETBV] = "XSETBV",
149	[EXIT_REASON_APIC_WRITE] = "APIC write",
150	[EXIT_REASON_RDRAND] = "RDRAND",
151	[EXIT_REASON_INVPCID] = "INVPCID",
152	[EXIT_REASON_VMFUNC] = "VMFUNC",
153	[EXIT_REASON_ENCLS] = "ENCLS",
154	[EXIT_REASON_RDSEED] = "RDSEED",
155	[EXIT_REASON_PM_LOG_FULL] = "Page-modification log full",
156	[EXIT_REASON_XSAVES] = "XSAVES",
157	[EXIT_REASON_XRSTORS] = "XRSTORS"
158};
159
160typedef int (*vmexit_handler_t)(struct vmctx *, struct vm_exit *, int *vcpu);
161extern int vmexit_task_switch(struct vmctx *, struct vm_exit *, int *vcpu);
162
163char *vmname;
164
165int guest_ncpus;
166uint16_t cores, maxcpus, sockets, threads;
167
168char *guest_uuid_str;
169
170int raw_stdio = 0;
171
172static int gdb_port = 0;
173static int guest_vmexit_on_hlt, guest_vmexit_on_pause;
174static int virtio_msix = 1;
175static int x2apic_mode = 0;	/* default is xAPIC */
176static int destroy_on_poweroff = 0;
177
178static int strictio;
179static int strictmsr = 1;
180
181static int acpi;
182
183static char *progname;
184static const int BSP = 0;
185
186static cpuset_t cpumask;
187
188static void vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip);
189
190static struct vm_exit vmexit[VM_MAXCPU];
191
192struct bhyvestats {
193	uint64_t	vmexit_bogus;
194	uint64_t	vmexit_reqidle;
195	uint64_t	vmexit_hlt;
196	uint64_t	vmexit_pause;
197	uint64_t	vmexit_mtrap;
198	uint64_t	vmexit_inst_emul;
199	uint64_t	cpu_switch_rotate;
200	uint64_t	cpu_switch_direct;
201} stats;
202
203struct mt_vmm_info {
204	pthread_t	mt_thr;
205	struct vmctx	*mt_ctx;
206	int		mt_vcpu;
207} mt_vmm_info[VM_MAXCPU];
208
209static cpuset_t *vcpumap[VM_MAXCPU] = { NULL };
210
211static void
212usage(int code)
213{
214
215        fprintf(stderr,
216		"Usage: %s [-abehuwxACDHPSWY]\n"
217		"       %*s [-c [[cpus=]numcpus][,sockets=n][,cores=n][,threads=n]]\n"
218		"       %*s [-g <gdb port>] [-l <lpc>]\n"
219		"       %*s [-m mem] [-p vcpu:hostcpu] [-s <pci>] [-U uuid] <vm>\n"
220		"       -a: local apic is in xAPIC mode (deprecated)\n"
221		"       -A: create ACPI tables\n"
222		"       -c: number of cpus and/or topology specification\n"
223		"       -C: include guest memory in core file\n"
224		"       -D: destroy on power-off\n"
225		"       -e: exit on unhandled I/O access\n"
226		"       -g: gdb port\n"
227		"       -h: help\n"
228		"       -H: vmexit from the guest on hlt\n"
229		"       -l: LPC device configuration\n"
230		"       -m: memory size in MB\n"
231		"       -p: pin 'vcpu' to 'hostcpu'\n"
232		"       -P: vmexit from the guest on pause\n"
233		"       -s: <slot,driver,configinfo> PCI slot config\n"
234		"       -S: guest memory cannot be swapped\n"
235		"       -u: RTC keeps UTC time\n"
236		"       -U: uuid\n"
237		"       -w: ignore unimplemented MSRs\n"
238		"       -W: force virtio to use single-vector MSI\n"
239		"       -x: local apic is in x2APIC mode\n"
240		"       -Y: disable MPtable generation\n",
241		progname, (int)strlen(progname), "", (int)strlen(progname), "",
242		(int)strlen(progname), "");
243
244	exit(code);
245}
246
247/*
248 * XXX This parser is known to have the following issues:
249 * 1.  It accepts null key=value tokens ",,".
250 * 2.  It accepts whitespace after = and before value.
251 * 3.  Values out of range of INT are silently wrapped.
252 * 4.  It doesn't check non-final values.
253 * 5.  The apparently bogus limits of UINT16_MAX are for future expansion.
254 *
255 * The acceptance of a null specification ('-c ""') is by design to match the
256 * manual page syntax specification, this results in a topology of 1 vCPU.
257 */
258static int
259topology_parse(const char *opt)
260{
261	uint64_t ncpus;
262	int c, chk, n, s, t, tmp;
263	char *cp, *str;
264	bool ns, scts;
265
266	c = 1, n = 1, s = 1, t = 1;
267	ns = false, scts = false;
268	str = strdup(opt);
269	if (str == NULL)
270		goto out;
271
272	while ((cp = strsep(&str, ",")) != NULL) {
273		if (sscanf(cp, "%i%n", &tmp, &chk) == 1) {
274			n = tmp;
275			ns = true;
276		} else if (sscanf(cp, "cpus=%i%n", &tmp, &chk) == 1) {
277			n = tmp;
278			ns = true;
279		} else if (sscanf(cp, "sockets=%i%n", &tmp, &chk) == 1) {
280			s = tmp;
281			scts = true;
282		} else if (sscanf(cp, "cores=%i%n", &tmp, &chk) == 1) {
283			c = tmp;
284			scts = true;
285		} else if (sscanf(cp, "threads=%i%n", &tmp, &chk) == 1) {
286			t = tmp;
287			scts = true;
288#ifdef notyet  /* Do not expose this until vmm.ko implements it */
289		} else if (sscanf(cp, "maxcpus=%i%n", &tmp, &chk) == 1) {
290			m = tmp;
291#endif
292		/* Skip the empty argument case from -c "" */
293		} else if (cp[0] == '\0')
294			continue;
295		else
296			goto out;
297		/* Any trailing garbage causes an error */
298		if (cp[chk] != '\0')
299			goto out;
300	}
301	free(str);
302	str = NULL;
303
304	/*
305	 * Range check 1 <= n <= UINT16_MAX all values
306	 */
307	if (n < 1 || s < 1 || c < 1 || t < 1 ||
308	    n > UINT16_MAX || s > UINT16_MAX || c > UINT16_MAX  ||
309	    t > UINT16_MAX)
310		return (-1);
311
312	/* If only the cpus was specified, use that as sockets */
313	if (!scts)
314		s = n;
315	/*
316	 * Compute sockets * cores * threads avoiding overflow
317	 * The range check above insures these are 16 bit values
318	 * If n was specified check it against computed ncpus
319	 */
320	ncpus = (uint64_t)s * c * t;
321	if (ncpus > UINT16_MAX || (ns && n != ncpus))
322		return (-1);
323
324	guest_ncpus = ncpus;
325	sockets = s;
326	cores = c;
327	threads = t;
328	return(0);
329
330out:
331	free(str);
332	return (-1);
333}
334
335static int
336pincpu_parse(const char *opt)
337{
338	int vcpu, pcpu;
339
340	if (sscanf(opt, "%d:%d", &vcpu, &pcpu) != 2) {
341		fprintf(stderr, "invalid format: %s\n", opt);
342		return (-1);
343	}
344
345	if (vcpu < 0 || vcpu >= VM_MAXCPU) {
346		fprintf(stderr, "vcpu '%d' outside valid range from 0 to %d\n",
347		    vcpu, VM_MAXCPU - 1);
348		return (-1);
349	}
350
351	if (pcpu < 0 || pcpu >= CPU_SETSIZE) {
352		fprintf(stderr, "hostcpu '%d' outside valid range from "
353		    "0 to %d\n", pcpu, CPU_SETSIZE - 1);
354		return (-1);
355	}
356
357	if (vcpumap[vcpu] == NULL) {
358		if ((vcpumap[vcpu] = malloc(sizeof(cpuset_t))) == NULL) {
359			perror("malloc");
360			return (-1);
361		}
362		CPU_ZERO(vcpumap[vcpu]);
363	}
364	CPU_SET(pcpu, vcpumap[vcpu]);
365	return (0);
366}
367
368void
369vm_inject_fault(void *arg, int vcpu, int vector, int errcode_valid,
370    int errcode)
371{
372	struct vmctx *ctx;
373	int error, restart_instruction;
374
375	ctx = arg;
376	restart_instruction = 1;
377
378	error = vm_inject_exception(ctx, vcpu, vector, errcode_valid, errcode,
379	    restart_instruction);
380	assert(error == 0);
381}
382
383void *
384paddr_guest2host(struct vmctx *ctx, uintptr_t gaddr, size_t len)
385{
386
387	return (vm_map_gpa(ctx, gaddr, len));
388}
389
390int
391fbsdrun_vmexit_on_pause(void)
392{
393
394	return (guest_vmexit_on_pause);
395}
396
397int
398fbsdrun_vmexit_on_hlt(void)
399{
400
401	return (guest_vmexit_on_hlt);
402}
403
404int
405fbsdrun_virtio_msix(void)
406{
407
408	return (virtio_msix);
409}
410
411static void *
412fbsdrun_start_thread(void *param)
413{
414	char tname[MAXCOMLEN + 1];
415	struct mt_vmm_info *mtp;
416	int vcpu;
417
418	mtp = param;
419	vcpu = mtp->mt_vcpu;
420
421	snprintf(tname, sizeof(tname), "vcpu %d", vcpu);
422	pthread_set_name_np(mtp->mt_thr, tname);
423
424	if (gdb_port != 0)
425		gdb_cpu_add(vcpu);
426
427	vm_loop(mtp->mt_ctx, vcpu, vmexit[vcpu].rip);
428
429	/* not reached */
430	exit(1);
431	return (NULL);
432}
433
434void
435fbsdrun_addcpu(struct vmctx *ctx, int fromcpu, int newcpu, uint64_t rip)
436{
437	int error;
438
439	assert(fromcpu == BSP);
440
441	/*
442	 * The 'newcpu' must be activated in the context of 'fromcpu'. If
443	 * vm_activate_cpu() is delayed until newcpu's pthread starts running
444	 * then vmm.ko is out-of-sync with bhyve and this can create a race
445	 * with vm_suspend().
446	 */
447	error = vm_activate_cpu(ctx, newcpu);
448	if (error != 0)
449		err(EX_OSERR, "could not activate CPU %d", newcpu);
450
451	CPU_SET_ATOMIC(newcpu, &cpumask);
452
453	/*
454	 * Set up the vmexit struct to allow execution to start
455	 * at the given RIP
456	 */
457	vmexit[newcpu].rip = rip;
458	vmexit[newcpu].inst_length = 0;
459
460	mt_vmm_info[newcpu].mt_ctx = ctx;
461	mt_vmm_info[newcpu].mt_vcpu = newcpu;
462
463	error = pthread_create(&mt_vmm_info[newcpu].mt_thr, NULL,
464	    fbsdrun_start_thread, &mt_vmm_info[newcpu]);
465	assert(error == 0);
466}
467
468static int
469fbsdrun_deletecpu(struct vmctx *ctx, int vcpu)
470{
471
472	if (!CPU_ISSET(vcpu, &cpumask)) {
473		fprintf(stderr, "Attempting to delete unknown cpu %d\n", vcpu);
474		exit(4);
475	}
476
477	CPU_CLR_ATOMIC(vcpu, &cpumask);
478	return (CPU_EMPTY(&cpumask));
479}
480
481static int
482vmexit_handle_notify(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu,
483		     uint32_t eax)
484{
485#if BHYVE_DEBUG
486	/*
487	 * put guest-driven debug here
488	 */
489#endif
490	return (VMEXIT_CONTINUE);
491}
492
493static int
494vmexit_inout(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
495{
496	int error;
497	int bytes, port, in, out;
498	int vcpu;
499
500	vcpu = *pvcpu;
501
502	port = vme->u.inout.port;
503	bytes = vme->u.inout.bytes;
504	in = vme->u.inout.in;
505	out = !in;
506
507        /* Extra-special case of host notifications */
508        if (out && port == GUEST_NIO_PORT) {
509                error = vmexit_handle_notify(ctx, vme, pvcpu, vme->u.inout.eax);
510		return (error);
511	}
512
513	error = emulate_inout(ctx, vcpu, vme, strictio);
514	if (error) {
515		fprintf(stderr, "Unhandled %s%c 0x%04x at 0x%lx\n",
516		    in ? "in" : "out",
517		    bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'),
518		    port, vmexit->rip);
519		return (VMEXIT_ABORT);
520	} else {
521		return (VMEXIT_CONTINUE);
522	}
523}
524
525static int
526vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
527{
528	uint64_t val;
529	uint32_t eax, edx;
530	int error;
531
532	val = 0;
533	error = emulate_rdmsr(ctx, *pvcpu, vme->u.msr.code, &val);
534	if (error != 0) {
535		fprintf(stderr, "rdmsr to register %#x on vcpu %d\n",
536		    vme->u.msr.code, *pvcpu);
537		if (strictmsr) {
538			vm_inject_gp(ctx, *pvcpu);
539			return (VMEXIT_CONTINUE);
540		}
541	}
542
543	eax = val;
544	error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RAX, eax);
545	assert(error == 0);
546
547	edx = val >> 32;
548	error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RDX, edx);
549	assert(error == 0);
550
551	return (VMEXIT_CONTINUE);
552}
553
554static int
555vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
556{
557	int error;
558
559	error = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code, vme->u.msr.wval);
560	if (error != 0) {
561		fprintf(stderr, "wrmsr to register %#x(%#lx) on vcpu %d\n",
562		    vme->u.msr.code, vme->u.msr.wval, *pvcpu);
563		if (strictmsr) {
564			vm_inject_gp(ctx, *pvcpu);
565			return (VMEXIT_CONTINUE);
566		}
567	}
568	return (VMEXIT_CONTINUE);
569}
570
571static int
572vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
573{
574
575	(void)spinup_ap(ctx, *pvcpu,
576		    vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip);
577
578	return (VMEXIT_CONTINUE);
579}
580
581#define	DEBUG_EPT_MISCONFIG
582#ifdef DEBUG_EPT_MISCONFIG
583#define	VMCS_GUEST_PHYSICAL_ADDRESS	0x00002400
584
585static uint64_t ept_misconfig_gpa, ept_misconfig_pte[4];
586static int ept_misconfig_ptenum;
587#endif
588
589static const char *
590vmexit_vmx_desc(uint32_t exit_reason)
591{
592
593	if (exit_reason >= nitems(vmx_exit_reason_desc) ||
594	    vmx_exit_reason_desc[exit_reason] == NULL)
595		return ("Unknown");
596	return (vmx_exit_reason_desc[exit_reason]);
597}
598
599static int
600vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
601{
602
603	fprintf(stderr, "vm exit[%d]\n", *pvcpu);
604	fprintf(stderr, "\treason\t\tVMX\n");
605	fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
606	fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
607	fprintf(stderr, "\tstatus\t\t%d\n", vmexit->u.vmx.status);
608	fprintf(stderr, "\texit_reason\t%u (%s)\n", vmexit->u.vmx.exit_reason,
609	    vmexit_vmx_desc(vmexit->u.vmx.exit_reason));
610	fprintf(stderr, "\tqualification\t0x%016lx\n",
611	    vmexit->u.vmx.exit_qualification);
612	fprintf(stderr, "\tinst_type\t\t%d\n", vmexit->u.vmx.inst_type);
613	fprintf(stderr, "\tinst_error\t\t%d\n", vmexit->u.vmx.inst_error);
614#ifdef DEBUG_EPT_MISCONFIG
615	if (vmexit->u.vmx.exit_reason == EXIT_REASON_EPT_MISCONFIG) {
616		vm_get_register(ctx, *pvcpu,
617		    VMCS_IDENT(VMCS_GUEST_PHYSICAL_ADDRESS),
618		    &ept_misconfig_gpa);
619		vm_get_gpa_pmap(ctx, ept_misconfig_gpa, ept_misconfig_pte,
620		    &ept_misconfig_ptenum);
621		fprintf(stderr, "\tEPT misconfiguration:\n");
622		fprintf(stderr, "\t\tGPA: %#lx\n", ept_misconfig_gpa);
623		fprintf(stderr, "\t\tPTE(%d): %#lx %#lx %#lx %#lx\n",
624		    ept_misconfig_ptenum, ept_misconfig_pte[0],
625		    ept_misconfig_pte[1], ept_misconfig_pte[2],
626		    ept_misconfig_pte[3]);
627	}
628#endif	/* DEBUG_EPT_MISCONFIG */
629	return (VMEXIT_ABORT);
630}
631
632static int
633vmexit_svm(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
634{
635
636	fprintf(stderr, "vm exit[%d]\n", *pvcpu);
637	fprintf(stderr, "\treason\t\tSVM\n");
638	fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
639	fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
640	fprintf(stderr, "\texitcode\t%#lx\n", vmexit->u.svm.exitcode);
641	fprintf(stderr, "\texitinfo1\t%#lx\n", vmexit->u.svm.exitinfo1);
642	fprintf(stderr, "\texitinfo2\t%#lx\n", vmexit->u.svm.exitinfo2);
643	return (VMEXIT_ABORT);
644}
645
646static int
647vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
648{
649
650	assert(vmexit->inst_length == 0);
651
652	stats.vmexit_bogus++;
653
654	return (VMEXIT_CONTINUE);
655}
656
657static int
658vmexit_reqidle(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
659{
660
661	assert(vmexit->inst_length == 0);
662
663	stats.vmexit_reqidle++;
664
665	return (VMEXIT_CONTINUE);
666}
667
668static int
669vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
670{
671
672	stats.vmexit_hlt++;
673
674	/*
675	 * Just continue execution with the next instruction. We use
676	 * the HLT VM exit as a way to be friendly with the host
677	 * scheduler.
678	 */
679	return (VMEXIT_CONTINUE);
680}
681
682static int
683vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
684{
685
686	stats.vmexit_pause++;
687
688	return (VMEXIT_CONTINUE);
689}
690
691static int
692vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
693{
694
695	assert(vmexit->inst_length == 0);
696
697	stats.vmexit_mtrap++;
698
699	if (gdb_port == 0) {
700		fprintf(stderr, "vm_loop: unexpected VMEXIT_MTRAP\n");
701		exit(4);
702	}
703	gdb_cpu_mtrap(*pvcpu);
704	return (VMEXIT_CONTINUE);
705}
706
707static int
708vmexit_inst_emul(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
709{
710	int err, i;
711	struct vie *vie;
712
713	stats.vmexit_inst_emul++;
714
715	vie = &vmexit->u.inst_emul.vie;
716	err = emulate_mem(ctx, *pvcpu, vmexit->u.inst_emul.gpa,
717	    vie, &vmexit->u.inst_emul.paging);
718
719	if (err) {
720		if (err == ESRCH) {
721			fprintf(stderr, "Unhandled memory access to 0x%lx\n",
722			    vmexit->u.inst_emul.gpa);
723		}
724
725		fprintf(stderr, "Failed to emulate instruction [");
726		for (i = 0; i < vie->num_valid; i++) {
727			fprintf(stderr, "0x%02x%s", vie->inst[i],
728			    i != (vie->num_valid - 1) ? " " : "");
729		}
730		fprintf(stderr, "] at 0x%lx\n", vmexit->rip);
731		return (VMEXIT_ABORT);
732	}
733
734	return (VMEXIT_CONTINUE);
735}
736
737static pthread_mutex_t resetcpu_mtx = PTHREAD_MUTEX_INITIALIZER;
738static pthread_cond_t resetcpu_cond = PTHREAD_COND_INITIALIZER;
739
740static int
741vmexit_suspend(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
742{
743	enum vm_suspend_how how;
744
745	how = vmexit->u.suspended.how;
746
747	fbsdrun_deletecpu(ctx, *pvcpu);
748
749	if (*pvcpu != BSP) {
750		pthread_mutex_lock(&resetcpu_mtx);
751		pthread_cond_signal(&resetcpu_cond);
752		pthread_mutex_unlock(&resetcpu_mtx);
753		pthread_exit(NULL);
754	}
755
756	pthread_mutex_lock(&resetcpu_mtx);
757	while (!CPU_EMPTY(&cpumask)) {
758		pthread_cond_wait(&resetcpu_cond, &resetcpu_mtx);
759	}
760	pthread_mutex_unlock(&resetcpu_mtx);
761
762	switch (how) {
763	case VM_SUSPEND_RESET:
764		exit(0);
765	case VM_SUSPEND_POWEROFF:
766		if (destroy_on_poweroff)
767			vm_destroy(ctx);
768		exit(1);
769	case VM_SUSPEND_HALT:
770		exit(2);
771	case VM_SUSPEND_TRIPLEFAULT:
772		exit(3);
773	default:
774		fprintf(stderr, "vmexit_suspend: invalid reason %d\n", how);
775		exit(100);
776	}
777	return (0);	/* NOTREACHED */
778}
779
780static int
781vmexit_debug(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
782{
783
784	if (gdb_port == 0) {
785		fprintf(stderr, "vm_loop: unexpected VMEXIT_DEBUG\n");
786		exit(4);
787	}
788	gdb_cpu_suspend(*pvcpu);
789	return (VMEXIT_CONTINUE);
790}
791
792static int
793vmexit_breakpoint(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
794{
795
796	if (gdb_port == 0) {
797		fprintf(stderr, "vm_loop: unexpected VMEXIT_DEBUG\n");
798		exit(4);
799	}
800	gdb_cpu_breakpoint(*pvcpu, vmexit);
801	return (VMEXIT_CONTINUE);
802}
803
804static vmexit_handler_t handler[VM_EXITCODE_MAX] = {
805	[VM_EXITCODE_INOUT]  = vmexit_inout,
806	[VM_EXITCODE_INOUT_STR]  = vmexit_inout,
807	[VM_EXITCODE_VMX]    = vmexit_vmx,
808	[VM_EXITCODE_SVM]    = vmexit_svm,
809	[VM_EXITCODE_BOGUS]  = vmexit_bogus,
810	[VM_EXITCODE_REQIDLE] = vmexit_reqidle,
811	[VM_EXITCODE_RDMSR]  = vmexit_rdmsr,
812	[VM_EXITCODE_WRMSR]  = vmexit_wrmsr,
813	[VM_EXITCODE_MTRAP]  = vmexit_mtrap,
814	[VM_EXITCODE_INST_EMUL] = vmexit_inst_emul,
815	[VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap,
816	[VM_EXITCODE_SUSPENDED] = vmexit_suspend,
817	[VM_EXITCODE_TASK_SWITCH] = vmexit_task_switch,
818	[VM_EXITCODE_DEBUG] = vmexit_debug,
819	[VM_EXITCODE_BPT] = vmexit_breakpoint,
820};
821
822static void
823vm_loop(struct vmctx *ctx, int vcpu, uint64_t startrip)
824{
825	int error, rc;
826	enum vm_exitcode exitcode;
827	cpuset_t active_cpus;
828
829	if (vcpumap[vcpu] != NULL) {
830		error = pthread_setaffinity_np(pthread_self(),
831		    sizeof(cpuset_t), vcpumap[vcpu]);
832		assert(error == 0);
833	}
834
835	error = vm_active_cpus(ctx, &active_cpus);
836	assert(CPU_ISSET(vcpu, &active_cpus));
837
838	error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RIP, startrip);
839	assert(error == 0);
840
841	while (1) {
842		error = vm_run(ctx, vcpu, &vmexit[vcpu]);
843		if (error != 0)
844			break;
845
846		exitcode = vmexit[vcpu].exitcode;
847		if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) {
848			fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n",
849			    exitcode);
850			exit(4);
851		}
852
853		rc = (*handler[exitcode])(ctx, &vmexit[vcpu], &vcpu);
854
855		switch (rc) {
856		case VMEXIT_CONTINUE:
857			break;
858		case VMEXIT_ABORT:
859			abort();
860		default:
861			exit(4);
862		}
863	}
864	fprintf(stderr, "vm_run error %d, errno %d\n", error, errno);
865}
866
867static int
868num_vcpus_allowed(struct vmctx *ctx)
869{
870	int tmp, error;
871
872	error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp);
873
874	/*
875	 * The guest is allowed to spinup more than one processor only if the
876	 * UNRESTRICTED_GUEST capability is available.
877	 */
878	if (error == 0)
879		return (VM_MAXCPU);
880	else
881		return (1);
882}
883
884void
885fbsdrun_set_capabilities(struct vmctx *ctx, int cpu)
886{
887	int err, tmp;
888
889	if (fbsdrun_vmexit_on_hlt()) {
890		err = vm_get_capability(ctx, cpu, VM_CAP_HALT_EXIT, &tmp);
891		if (err < 0) {
892			fprintf(stderr, "VM exit on HLT not supported\n");
893			exit(4);
894		}
895		vm_set_capability(ctx, cpu, VM_CAP_HALT_EXIT, 1);
896		if (cpu == BSP)
897			handler[VM_EXITCODE_HLT] = vmexit_hlt;
898	}
899
900        if (fbsdrun_vmexit_on_pause()) {
901		/*
902		 * pause exit support required for this mode
903		 */
904		err = vm_get_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, &tmp);
905		if (err < 0) {
906			fprintf(stderr,
907			    "SMP mux requested, no pause support\n");
908			exit(4);
909		}
910		vm_set_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, 1);
911		if (cpu == BSP)
912			handler[VM_EXITCODE_PAUSE] = vmexit_pause;
913        }
914
915	if (x2apic_mode)
916		err = vm_set_x2apic_state(ctx, cpu, X2APIC_ENABLED);
917	else
918		err = vm_set_x2apic_state(ctx, cpu, X2APIC_DISABLED);
919
920	if (err) {
921		fprintf(stderr, "Unable to set x2apic state (%d)\n", err);
922		exit(4);
923	}
924
925	vm_set_capability(ctx, cpu, VM_CAP_ENABLE_INVPCID, 1);
926}
927
928static struct vmctx *
929do_open(const char *vmname)
930{
931	struct vmctx *ctx;
932	int error;
933	bool reinit, romboot;
934#ifndef WITHOUT_CAPSICUM
935	cap_rights_t rights;
936	const cap_ioctl_t *cmds;
937	size_t ncmds;
938#endif
939
940	reinit = romboot = false;
941
942	if (lpc_bootrom())
943		romboot = true;
944
945	error = vm_create(vmname);
946	if (error) {
947		if (errno == EEXIST) {
948			if (romboot) {
949				reinit = true;
950			} else {
951				/*
952				 * The virtual machine has been setup by the
953				 * userspace bootloader.
954				 */
955			}
956		} else {
957			perror("vm_create");
958			exit(4);
959		}
960	} else {
961		if (!romboot) {
962			/*
963			 * If the virtual machine was just created then a
964			 * bootrom must be configured to boot it.
965			 */
966			fprintf(stderr, "virtual machine cannot be booted\n");
967			exit(4);
968		}
969	}
970
971	ctx = vm_open(vmname);
972	if (ctx == NULL) {
973		perror("vm_open");
974		exit(4);
975	}
976
977#ifndef WITHOUT_CAPSICUM
978	cap_rights_init(&rights, CAP_IOCTL, CAP_MMAP_RW);
979	if (caph_rights_limit(vm_get_device_fd(ctx), &rights) == -1)
980		errx(EX_OSERR, "Unable to apply rights for sandbox");
981	vm_get_ioctls(&ncmds);
982	cmds = vm_get_ioctls(NULL);
983	if (cmds == NULL)
984		errx(EX_OSERR, "out of memory");
985	if (caph_ioctls_limit(vm_get_device_fd(ctx), cmds, ncmds) == -1)
986		errx(EX_OSERR, "Unable to apply rights for sandbox");
987	free((cap_ioctl_t *)cmds);
988#endif
989
990	if (reinit) {
991		error = vm_reinit(ctx);
992		if (error) {
993			perror("vm_reinit");
994			exit(4);
995		}
996	}
997	error = vm_set_topology(ctx, sockets, cores, threads, maxcpus);
998	if (error)
999		errx(EX_OSERR, "vm_set_topology");
1000	return (ctx);
1001}
1002
1003int
1004main(int argc, char *argv[])
1005{
1006	int c, error, dbg_port, err, bvmcons;
1007	int max_vcpus, mptgen, memflags;
1008	int rtc_localtime;
1009	bool gdb_stop;
1010	struct vmctx *ctx;
1011	uint64_t rip;
1012	size_t memsize;
1013	char *optstr;
1014
1015	bvmcons = 0;
1016	progname = basename(argv[0]);
1017	dbg_port = 0;
1018	gdb_stop = false;
1019	guest_ncpus = 1;
1020	sockets = cores = threads = 1;
1021	maxcpus = 0;
1022	memsize = 256 * MB;
1023	mptgen = 1;
1024	rtc_localtime = 1;
1025	memflags = 0;
1026
1027	optstr = "abehuwxACDHIPSWYp:g:G:c:s:m:l:U:";
1028	while ((c = getopt(argc, argv, optstr)) != -1) {
1029		switch (c) {
1030		case 'a':
1031			x2apic_mode = 0;
1032			break;
1033		case 'A':
1034			acpi = 1;
1035			break;
1036		case 'b':
1037			warnx("-b flag is deprecated and will be removed in FreeBSD 13.0");
1038			bvmcons = 1;
1039			break;
1040		case 'D':
1041			destroy_on_poweroff = 1;
1042			break;
1043		case 'p':
1044                        if (pincpu_parse(optarg) != 0) {
1045                            errx(EX_USAGE, "invalid vcpu pinning "
1046                                 "configuration '%s'", optarg);
1047                        }
1048			break;
1049                case 'c':
1050			if (topology_parse(optarg) != 0) {
1051			    errx(EX_USAGE, "invalid cpu topology "
1052				"'%s'", optarg);
1053			}
1054			break;
1055		case 'C':
1056			memflags |= VM_MEM_F_INCORE;
1057			break;
1058		case 'g':
1059			warnx("-g flag is deprecated and will be removed in FreeBSD 13.0");
1060			dbg_port = atoi(optarg);
1061			break;
1062		case 'G':
1063			if (optarg[0] == 'w') {
1064				gdb_stop = true;
1065				optarg++;
1066			}
1067			gdb_port = atoi(optarg);
1068			break;
1069		case 'l':
1070			if (strncmp(optarg, "help", strlen(optarg)) == 0) {
1071				lpc_print_supported_devices();
1072				exit(0);
1073			} else if (lpc_device_parse(optarg) != 0) {
1074				errx(EX_USAGE, "invalid lpc device "
1075				    "configuration '%s'", optarg);
1076			}
1077			break;
1078		case 's':
1079			if (strncmp(optarg, "help", strlen(optarg)) == 0) {
1080				pci_print_supported_devices();
1081				exit(0);
1082			} else if (pci_parse_slot(optarg) != 0)
1083				exit(4);
1084			else
1085				break;
1086		case 'S':
1087			memflags |= VM_MEM_F_WIRED;
1088			break;
1089                case 'm':
1090			error = vm_parse_memsize(optarg, &memsize);
1091			if (error)
1092				errx(EX_USAGE, "invalid memsize '%s'", optarg);
1093			break;
1094		case 'H':
1095			guest_vmexit_on_hlt = 1;
1096			break;
1097		case 'I':
1098			/*
1099			 * The "-I" option was used to add an ioapic to the
1100			 * virtual machine.
1101			 *
1102			 * An ioapic is now provided unconditionally for each
1103			 * virtual machine and this option is now deprecated.
1104			 */
1105			break;
1106		case 'P':
1107			guest_vmexit_on_pause = 1;
1108			break;
1109		case 'e':
1110			strictio = 1;
1111			break;
1112		case 'u':
1113			rtc_localtime = 0;
1114			break;
1115		case 'U':
1116			guest_uuid_str = optarg;
1117			break;
1118		case 'w':
1119			strictmsr = 0;
1120			break;
1121		case 'W':
1122			virtio_msix = 0;
1123			break;
1124		case 'x':
1125			x2apic_mode = 1;
1126			break;
1127		case 'Y':
1128			mptgen = 0;
1129			break;
1130		case 'h':
1131			usage(0);
1132		default:
1133			usage(1);
1134		}
1135	}
1136	argc -= optind;
1137	argv += optind;
1138
1139	if (argc != 1)
1140		usage(1);
1141
1142	vmname = argv[0];
1143	ctx = do_open(vmname);
1144
1145	max_vcpus = num_vcpus_allowed(ctx);
1146	if (guest_ncpus > max_vcpus) {
1147		fprintf(stderr, "%d vCPUs requested but only %d available\n",
1148			guest_ncpus, max_vcpus);
1149		exit(4);
1150	}
1151
1152	fbsdrun_set_capabilities(ctx, BSP);
1153
1154	vm_set_memflags(ctx, memflags);
1155	err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL);
1156	if (err) {
1157		fprintf(stderr, "Unable to setup memory (%d)\n", errno);
1158		exit(4);
1159	}
1160
1161	error = init_msr();
1162	if (error) {
1163		fprintf(stderr, "init_msr error %d", error);
1164		exit(4);
1165	}
1166
1167	init_mem();
1168	init_inout();
1169	atkbdc_init(ctx);
1170	pci_irq_init(ctx);
1171	ioapic_init(ctx);
1172
1173	rtc_init(ctx, rtc_localtime);
1174	sci_init(ctx);
1175
1176	/*
1177	 * Exit if a device emulation finds an error in its initilization
1178	 */
1179	if (init_pci(ctx) != 0) {
1180		perror("device emulation initialization error");
1181		exit(4);
1182	}
1183
1184	if (dbg_port != 0)
1185		init_dbgport(dbg_port);
1186
1187	if (gdb_port != 0)
1188		init_gdb(ctx, gdb_port, gdb_stop);
1189
1190	if (bvmcons)
1191		init_bvmcons();
1192
1193	if (lpc_bootrom()) {
1194		if (vm_set_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, 1)) {
1195			fprintf(stderr, "ROM boot failed: unrestricted guest "
1196			    "capability not available\n");
1197			exit(4);
1198		}
1199		error = vcpu_reset(ctx, BSP);
1200		assert(error == 0);
1201	}
1202
1203	error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip);
1204	assert(error == 0);
1205
1206	/*
1207	 * build the guest tables, MP etc.
1208	 */
1209	if (mptgen) {
1210		error = mptable_build(ctx, guest_ncpus);
1211		if (error) {
1212			perror("error to build the guest tables");
1213			exit(4);
1214		}
1215	}
1216
1217	error = smbios_build(ctx);
1218	assert(error == 0);
1219
1220	if (acpi) {
1221		error = acpi_build(ctx, guest_ncpus);
1222		assert(error == 0);
1223	}
1224
1225	if (lpc_bootrom())
1226		fwctl_init();
1227
1228	/*
1229	 * Change the proc title to include the VM name.
1230	 */
1231	setproctitle("%s", vmname);
1232
1233#ifndef WITHOUT_CAPSICUM
1234	caph_cache_catpages();
1235
1236	if (caph_limit_stdout() == -1 || caph_limit_stderr() == -1)
1237		errx(EX_OSERR, "Unable to apply rights for sandbox");
1238
1239	if (caph_enter() == -1)
1240		errx(EX_OSERR, "cap_enter() failed");
1241#endif
1242
1243	/*
1244	 * Add CPU 0
1245	 */
1246	fbsdrun_addcpu(ctx, BSP, BSP, rip);
1247
1248	/*
1249	 * Head off to the main event dispatch loop
1250	 */
1251	mevent_dispatch();
1252
1253	exit(4);
1254}
1255