2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
4 * Copyright (c) 2011 NetApp, Inc.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
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.
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
31 #include <sys/cdefs.h>
32 __FBSDID("$FreeBSD$");
34 #include <sys/types.h>
35 #ifndef WITHOUT_CAPSICUM
36 #include <sys/capsicum.h>
41 #include <amd64/vmm/intel/vmcs.h>
43 #include <machine/atomic.h>
44 #include <machine/segments.h>
46 #ifndef WITHOUT_CAPSICUM
47 #include <capsicum_helpers.h>
58 #include <pthread_np.h>
63 #include <machine/vmm.h>
64 #ifndef WITHOUT_CAPSICUM
65 #include <machine/vmm_dev.h>
83 #include "smbiostbl.h"
85 #include "spinup_ap.h"
88 #define GUEST_NIO_PORT 0x488 /* guest upcalls via i/o port */
90 #define MB (1024UL * 1024)
91 #define GB (1024UL * MB)
93 static 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"
160 typedef int (*vmexit_handler_t)(struct vmctx *, struct vm_exit *, int *vcpu);
161 extern int vmexit_task_switch(struct vmctx *, struct vm_exit *, int *vcpu);
166 uint16_t cores, maxcpus, sockets, threads;
168 char *guest_uuid_str;
170 static int gdb_port = 0;
171 static int guest_vmexit_on_hlt, guest_vmexit_on_pause;
172 static int virtio_msix = 1;
173 static int x2apic_mode = 0; /* default is xAPIC */
176 static int strictmsr = 1;
180 static char *progname;
181 static const int BSP = 0;
183 static cpuset_t cpumask;
185 static void vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip);
187 static struct vm_exit vmexit[VM_MAXCPU];
190 uint64_t vmexit_bogus;
191 uint64_t vmexit_reqidle;
193 uint64_t vmexit_pause;
194 uint64_t vmexit_mtrap;
195 uint64_t vmexit_inst_emul;
196 uint64_t cpu_switch_rotate;
197 uint64_t cpu_switch_direct;
202 struct vmctx *mt_ctx;
204 } mt_vmm_info[VM_MAXCPU];
206 static cpuset_t *vcpumap[VM_MAXCPU] = { NULL };
213 "Usage: %s [-abehuwxACHPSWY]\n"
214 " %*s [-c [[cpus=]numcpus][,sockets=n][,cores=n][,threads=n]]\n"
215 " %*s [-g <gdb port>] [-l <lpc>]\n"
216 " %*s [-m mem] [-p vcpu:hostcpu] [-s <pci>] [-U uuid] <vm>\n"
217 " -a: local apic is in xAPIC mode (deprecated)\n"
218 " -A: create ACPI tables\n"
219 " -c: number of cpus and/or topology specification\n"
220 " -C: include guest memory in core file\n"
221 " -e: exit on unhandled I/O access\n"
224 " -H: vmexit from the guest on hlt\n"
225 " -l: LPC device configuration\n"
226 " -m: memory size in MB\n"
227 " -p: pin 'vcpu' to 'hostcpu'\n"
228 " -P: vmexit from the guest on pause\n"
229 " -s: <slot,driver,configinfo> PCI slot config\n"
230 " -S: guest memory cannot be swapped\n"
231 " -u: RTC keeps UTC time\n"
233 " -w: ignore unimplemented MSRs\n"
234 " -W: force virtio to use single-vector MSI\n"
235 " -x: local apic is in x2APIC mode\n"
236 " -Y: disable MPtable generation\n",
237 progname, (int)strlen(progname), "", (int)strlen(progname), "",
238 (int)strlen(progname), "");
244 * XXX This parser is known to have the following issues:
245 * 1. It accepts null key=value tokens ",,".
246 * 2. It accepts whitespace after = and before value.
247 * 3. Values out of range of INT are silently wrapped.
248 * 4. It doesn't check non-final values.
249 * 5. The apparently bogus limits of UINT16_MAX are for future expansion.
251 * The acceptance of a null specification ('-c ""') is by design to match the
252 * manual page syntax specification, this results in a topology of 1 vCPU.
255 topology_parse(const char *opt)
258 int c, chk, n, s, t, tmp;
262 c = 1, n = 1, s = 1, t = 1;
263 ns = false, scts = false;
268 while ((cp = strsep(&str, ",")) != NULL) {
269 if (sscanf(cp, "%i%n", &tmp, &chk) == 1) {
272 } else if (sscanf(cp, "cpus=%i%n", &tmp, &chk) == 1) {
275 } else if (sscanf(cp, "sockets=%i%n", &tmp, &chk) == 1) {
278 } else if (sscanf(cp, "cores=%i%n", &tmp, &chk) == 1) {
281 } else if (sscanf(cp, "threads=%i%n", &tmp, &chk) == 1) {
284 #ifdef notyet /* Do not expose this until vmm.ko implements it */
285 } else if (sscanf(cp, "maxcpus=%i%n", &tmp, &chk) == 1) {
288 /* Skip the empty argument case from -c "" */
289 } else if (cp[0] == '\0')
293 /* Any trailing garbage causes an error */
301 * Range check 1 <= n <= UINT16_MAX all values
303 if (n < 1 || s < 1 || c < 1 || t < 1 ||
304 n > UINT16_MAX || s > UINT16_MAX || c > UINT16_MAX ||
308 /* If only the cpus was specified, use that as sockets */
312 * Compute sockets * cores * threads avoiding overflow
313 * The range check above insures these are 16 bit values
314 * If n was specified check it against computed ncpus
316 ncpus = (uint64_t)s * c * t;
317 if (ncpus > UINT16_MAX || (ns && n != ncpus))
332 pincpu_parse(const char *opt)
336 if (sscanf(opt, "%d:%d", &vcpu, &pcpu) != 2) {
337 fprintf(stderr, "invalid format: %s\n", opt);
341 if (vcpu < 0 || vcpu >= VM_MAXCPU) {
342 fprintf(stderr, "vcpu '%d' outside valid range from 0 to %d\n",
343 vcpu, VM_MAXCPU - 1);
347 if (pcpu < 0 || pcpu >= CPU_SETSIZE) {
348 fprintf(stderr, "hostcpu '%d' outside valid range from "
349 "0 to %d\n", pcpu, CPU_SETSIZE - 1);
353 if (vcpumap[vcpu] == NULL) {
354 if ((vcpumap[vcpu] = malloc(sizeof(cpuset_t))) == NULL) {
358 CPU_ZERO(vcpumap[vcpu]);
360 CPU_SET(pcpu, vcpumap[vcpu]);
365 vm_inject_fault(void *arg, int vcpu, int vector, int errcode_valid,
369 int error, restart_instruction;
372 restart_instruction = 1;
374 error = vm_inject_exception(ctx, vcpu, vector, errcode_valid, errcode,
375 restart_instruction);
380 paddr_guest2host(struct vmctx *ctx, uintptr_t gaddr, size_t len)
383 return (vm_map_gpa(ctx, gaddr, len));
387 fbsdrun_vmexit_on_pause(void)
390 return (guest_vmexit_on_pause);
394 fbsdrun_vmexit_on_hlt(void)
397 return (guest_vmexit_on_hlt);
401 fbsdrun_virtio_msix(void)
404 return (virtio_msix);
408 fbsdrun_start_thread(void *param)
410 char tname[MAXCOMLEN + 1];
411 struct mt_vmm_info *mtp;
417 snprintf(tname, sizeof(tname), "vcpu %d", vcpu);
418 pthread_set_name_np(mtp->mt_thr, tname);
423 vm_loop(mtp->mt_ctx, vcpu, vmexit[vcpu].rip);
431 fbsdrun_addcpu(struct vmctx *ctx, int fromcpu, int newcpu, uint64_t rip)
435 assert(fromcpu == BSP);
438 * The 'newcpu' must be activated in the context of 'fromcpu'. If
439 * vm_activate_cpu() is delayed until newcpu's pthread starts running
440 * then vmm.ko is out-of-sync with bhyve and this can create a race
443 error = vm_activate_cpu(ctx, newcpu);
445 err(EX_OSERR, "could not activate CPU %d", newcpu);
447 CPU_SET_ATOMIC(newcpu, &cpumask);
450 * Set up the vmexit struct to allow execution to start
453 vmexit[newcpu].rip = rip;
454 vmexit[newcpu].inst_length = 0;
456 mt_vmm_info[newcpu].mt_ctx = ctx;
457 mt_vmm_info[newcpu].mt_vcpu = newcpu;
459 error = pthread_create(&mt_vmm_info[newcpu].mt_thr, NULL,
460 fbsdrun_start_thread, &mt_vmm_info[newcpu]);
465 fbsdrun_deletecpu(struct vmctx *ctx, int vcpu)
468 if (!CPU_ISSET(vcpu, &cpumask)) {
469 fprintf(stderr, "Attempting to delete unknown cpu %d\n", vcpu);
473 CPU_CLR_ATOMIC(vcpu, &cpumask);
474 return (CPU_EMPTY(&cpumask));
478 vmexit_handle_notify(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu,
483 * put guest-driven debug here
486 return (VMEXIT_CONTINUE);
490 vmexit_inout(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
493 int bytes, port, in, out;
498 port = vme->u.inout.port;
499 bytes = vme->u.inout.bytes;
500 in = vme->u.inout.in;
503 /* Extra-special case of host notifications */
504 if (out && port == GUEST_NIO_PORT) {
505 error = vmexit_handle_notify(ctx, vme, pvcpu, vme->u.inout.eax);
509 error = emulate_inout(ctx, vcpu, vme, strictio);
511 fprintf(stderr, "Unhandled %s%c 0x%04x at 0x%lx\n",
513 bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'),
515 return (VMEXIT_ABORT);
517 return (VMEXIT_CONTINUE);
522 vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
529 error = emulate_rdmsr(ctx, *pvcpu, vme->u.msr.code, &val);
531 fprintf(stderr, "rdmsr to register %#x on vcpu %d\n",
532 vme->u.msr.code, *pvcpu);
534 vm_inject_gp(ctx, *pvcpu);
535 return (VMEXIT_CONTINUE);
540 error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RAX, eax);
544 error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RDX, edx);
547 return (VMEXIT_CONTINUE);
551 vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
555 error = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code, vme->u.msr.wval);
557 fprintf(stderr, "wrmsr to register %#x(%#lx) on vcpu %d\n",
558 vme->u.msr.code, vme->u.msr.wval, *pvcpu);
560 vm_inject_gp(ctx, *pvcpu);
561 return (VMEXIT_CONTINUE);
564 return (VMEXIT_CONTINUE);
568 vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
571 (void)spinup_ap(ctx, *pvcpu,
572 vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip);
574 return (VMEXIT_CONTINUE);
577 #define DEBUG_EPT_MISCONFIG
578 #ifdef DEBUG_EPT_MISCONFIG
579 #define VMCS_GUEST_PHYSICAL_ADDRESS 0x00002400
581 static uint64_t ept_misconfig_gpa, ept_misconfig_pte[4];
582 static int ept_misconfig_ptenum;
586 vmexit_vmx_desc(uint32_t exit_reason)
589 if (exit_reason >= nitems(vmx_exit_reason_desc) ||
590 vmx_exit_reason_desc[exit_reason] == NULL)
592 return (vmx_exit_reason_desc[exit_reason]);
596 vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
599 fprintf(stderr, "vm exit[%d]\n", *pvcpu);
600 fprintf(stderr, "\treason\t\tVMX\n");
601 fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
602 fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
603 fprintf(stderr, "\tstatus\t\t%d\n", vmexit->u.vmx.status);
604 fprintf(stderr, "\texit_reason\t%u (%s)\n", vmexit->u.vmx.exit_reason,
605 vmexit_vmx_desc(vmexit->u.vmx.exit_reason));
606 fprintf(stderr, "\tqualification\t0x%016lx\n",
607 vmexit->u.vmx.exit_qualification);
608 fprintf(stderr, "\tinst_type\t\t%d\n", vmexit->u.vmx.inst_type);
609 fprintf(stderr, "\tinst_error\t\t%d\n", vmexit->u.vmx.inst_error);
610 #ifdef DEBUG_EPT_MISCONFIG
611 if (vmexit->u.vmx.exit_reason == EXIT_REASON_EPT_MISCONFIG) {
612 vm_get_register(ctx, *pvcpu,
613 VMCS_IDENT(VMCS_GUEST_PHYSICAL_ADDRESS),
615 vm_get_gpa_pmap(ctx, ept_misconfig_gpa, ept_misconfig_pte,
616 &ept_misconfig_ptenum);
617 fprintf(stderr, "\tEPT misconfiguration:\n");
618 fprintf(stderr, "\t\tGPA: %#lx\n", ept_misconfig_gpa);
619 fprintf(stderr, "\t\tPTE(%d): %#lx %#lx %#lx %#lx\n",
620 ept_misconfig_ptenum, ept_misconfig_pte[0],
621 ept_misconfig_pte[1], ept_misconfig_pte[2],
622 ept_misconfig_pte[3]);
624 #endif /* DEBUG_EPT_MISCONFIG */
625 return (VMEXIT_ABORT);
629 vmexit_svm(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
632 fprintf(stderr, "vm exit[%d]\n", *pvcpu);
633 fprintf(stderr, "\treason\t\tSVM\n");
634 fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
635 fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
636 fprintf(stderr, "\texitcode\t%#lx\n", vmexit->u.svm.exitcode);
637 fprintf(stderr, "\texitinfo1\t%#lx\n", vmexit->u.svm.exitinfo1);
638 fprintf(stderr, "\texitinfo2\t%#lx\n", vmexit->u.svm.exitinfo2);
639 return (VMEXIT_ABORT);
643 vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
646 assert(vmexit->inst_length == 0);
648 stats.vmexit_bogus++;
650 return (VMEXIT_CONTINUE);
654 vmexit_reqidle(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
657 assert(vmexit->inst_length == 0);
659 stats.vmexit_reqidle++;
661 return (VMEXIT_CONTINUE);
665 vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
671 * Just continue execution with the next instruction. We use
672 * the HLT VM exit as a way to be friendly with the host
675 return (VMEXIT_CONTINUE);
679 vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
682 stats.vmexit_pause++;
684 return (VMEXIT_CONTINUE);
688 vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
691 assert(vmexit->inst_length == 0);
693 stats.vmexit_mtrap++;
696 fprintf(stderr, "vm_loop: unexpected VMEXIT_MTRAP\n");
699 gdb_cpu_mtrap(*pvcpu);
700 return (VMEXIT_CONTINUE);
704 vmexit_inst_emul(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
709 stats.vmexit_inst_emul++;
711 vie = &vmexit->u.inst_emul.vie;
712 err = emulate_mem(ctx, *pvcpu, vmexit->u.inst_emul.gpa,
713 vie, &vmexit->u.inst_emul.paging);
717 fprintf(stderr, "Unhandled memory access to 0x%lx\n",
718 vmexit->u.inst_emul.gpa);
721 fprintf(stderr, "Failed to emulate instruction [");
722 for (i = 0; i < vie->num_valid; i++) {
723 fprintf(stderr, "0x%02x%s", vie->inst[i],
724 i != (vie->num_valid - 1) ? " " : "");
726 fprintf(stderr, "] at 0x%lx\n", vmexit->rip);
727 return (VMEXIT_ABORT);
730 return (VMEXIT_CONTINUE);
733 static pthread_mutex_t resetcpu_mtx = PTHREAD_MUTEX_INITIALIZER;
734 static pthread_cond_t resetcpu_cond = PTHREAD_COND_INITIALIZER;
737 vmexit_suspend(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
739 enum vm_suspend_how how;
741 how = vmexit->u.suspended.how;
743 fbsdrun_deletecpu(ctx, *pvcpu);
746 pthread_mutex_lock(&resetcpu_mtx);
747 pthread_cond_signal(&resetcpu_cond);
748 pthread_mutex_unlock(&resetcpu_mtx);
752 pthread_mutex_lock(&resetcpu_mtx);
753 while (!CPU_EMPTY(&cpumask)) {
754 pthread_cond_wait(&resetcpu_cond, &resetcpu_mtx);
756 pthread_mutex_unlock(&resetcpu_mtx);
759 case VM_SUSPEND_RESET:
761 case VM_SUSPEND_POWEROFF:
763 case VM_SUSPEND_HALT:
765 case VM_SUSPEND_TRIPLEFAULT:
768 fprintf(stderr, "vmexit_suspend: invalid reason %d\n", how);
771 return (0); /* NOTREACHED */
775 vmexit_debug(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
779 fprintf(stderr, "vm_loop: unexpected VMEXIT_DEBUG\n");
782 gdb_cpu_suspend(*pvcpu);
783 return (VMEXIT_CONTINUE);
787 vmexit_breakpoint(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
791 fprintf(stderr, "vm_loop: unexpected VMEXIT_DEBUG\n");
794 gdb_cpu_breakpoint(*pvcpu, vmexit);
795 return (VMEXIT_CONTINUE);
798 static vmexit_handler_t handler[VM_EXITCODE_MAX] = {
799 [VM_EXITCODE_INOUT] = vmexit_inout,
800 [VM_EXITCODE_INOUT_STR] = vmexit_inout,
801 [VM_EXITCODE_VMX] = vmexit_vmx,
802 [VM_EXITCODE_SVM] = vmexit_svm,
803 [VM_EXITCODE_BOGUS] = vmexit_bogus,
804 [VM_EXITCODE_REQIDLE] = vmexit_reqidle,
805 [VM_EXITCODE_RDMSR] = vmexit_rdmsr,
806 [VM_EXITCODE_WRMSR] = vmexit_wrmsr,
807 [VM_EXITCODE_MTRAP] = vmexit_mtrap,
808 [VM_EXITCODE_INST_EMUL] = vmexit_inst_emul,
809 [VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap,
810 [VM_EXITCODE_SUSPENDED] = vmexit_suspend,
811 [VM_EXITCODE_TASK_SWITCH] = vmexit_task_switch,
812 [VM_EXITCODE_DEBUG] = vmexit_debug,
813 [VM_EXITCODE_BPT] = vmexit_breakpoint,
817 vm_loop(struct vmctx *ctx, int vcpu, uint64_t startrip)
820 enum vm_exitcode exitcode;
821 cpuset_t active_cpus;
823 if (vcpumap[vcpu] != NULL) {
824 error = pthread_setaffinity_np(pthread_self(),
825 sizeof(cpuset_t), vcpumap[vcpu]);
829 error = vm_active_cpus(ctx, &active_cpus);
830 assert(CPU_ISSET(vcpu, &active_cpus));
832 error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RIP, startrip);
836 error = vm_run(ctx, vcpu, &vmexit[vcpu]);
840 exitcode = vmexit[vcpu].exitcode;
841 if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) {
842 fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n",
847 rc = (*handler[exitcode])(ctx, &vmexit[vcpu], &vcpu);
850 case VMEXIT_CONTINUE:
858 fprintf(stderr, "vm_run error %d, errno %d\n", error, errno);
862 num_vcpus_allowed(struct vmctx *ctx)
866 error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp);
869 * The guest is allowed to spinup more than one processor only if the
870 * UNRESTRICTED_GUEST capability is available.
879 fbsdrun_set_capabilities(struct vmctx *ctx, int cpu)
883 if (fbsdrun_vmexit_on_hlt()) {
884 err = vm_get_capability(ctx, cpu, VM_CAP_HALT_EXIT, &tmp);
886 fprintf(stderr, "VM exit on HLT not supported\n");
889 vm_set_capability(ctx, cpu, VM_CAP_HALT_EXIT, 1);
891 handler[VM_EXITCODE_HLT] = vmexit_hlt;
894 if (fbsdrun_vmexit_on_pause()) {
896 * pause exit support required for this mode
898 err = vm_get_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, &tmp);
901 "SMP mux requested, no pause support\n");
904 vm_set_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, 1);
906 handler[VM_EXITCODE_PAUSE] = vmexit_pause;
910 err = vm_set_x2apic_state(ctx, cpu, X2APIC_ENABLED);
912 err = vm_set_x2apic_state(ctx, cpu, X2APIC_DISABLED);
915 fprintf(stderr, "Unable to set x2apic state (%d)\n", err);
919 vm_set_capability(ctx, cpu, VM_CAP_ENABLE_INVPCID, 1);
922 static struct vmctx *
923 do_open(const char *vmname)
927 bool reinit, romboot;
928 #ifndef WITHOUT_CAPSICUM
930 const cap_ioctl_t *cmds;
934 reinit = romboot = false;
939 error = vm_create(vmname);
941 if (errno == EEXIST) {
946 * The virtual machine has been setup by the
947 * userspace bootloader.
957 * If the virtual machine was just created then a
958 * bootrom must be configured to boot it.
960 fprintf(stderr, "virtual machine cannot be booted\n");
965 ctx = vm_open(vmname);
971 #ifndef WITHOUT_CAPSICUM
972 cap_rights_init(&rights, CAP_IOCTL, CAP_MMAP_RW);
973 if (caph_rights_limit(vm_get_device_fd(ctx), &rights) == -1)
974 errx(EX_OSERR, "Unable to apply rights for sandbox");
975 vm_get_ioctls(&ncmds);
976 cmds = vm_get_ioctls(NULL);
978 errx(EX_OSERR, "out of memory");
979 if (caph_ioctls_limit(vm_get_device_fd(ctx), cmds, ncmds) == -1)
980 errx(EX_OSERR, "Unable to apply rights for sandbox");
981 free((cap_ioctl_t *)cmds);
985 error = vm_reinit(ctx);
991 error = vm_set_topology(ctx, sockets, cores, threads, maxcpus);
993 errx(EX_OSERR, "vm_set_topology");
998 main(int argc, char *argv[])
1000 int c, error, dbg_port, err, bvmcons;
1001 int max_vcpus, mptgen, memflags;
1010 progname = basename(argv[0]);
1014 sockets = cores = threads = 1;
1021 optstr = "abehuwxACHIPSWYp:g:G:c:s:m:l:U:";
1022 while ((c = getopt(argc, argv, optstr)) != -1) {
1034 if (pincpu_parse(optarg) != 0) {
1035 errx(EX_USAGE, "invalid vcpu pinning "
1036 "configuration '%s'", optarg);
1040 if (topology_parse(optarg) != 0) {
1041 errx(EX_USAGE, "invalid cpu topology "
1046 memflags |= VM_MEM_F_INCORE;
1049 dbg_port = atoi(optarg);
1052 if (optarg[0] == 'w') {
1056 gdb_port = atoi(optarg);
1059 if (strncmp(optarg, "help", strlen(optarg)) == 0) {
1060 lpc_print_supported_devices();
1062 } else if (lpc_device_parse(optarg) != 0) {
1063 errx(EX_USAGE, "invalid lpc device "
1064 "configuration '%s'", optarg);
1068 if (strncmp(optarg, "help", strlen(optarg)) == 0) {
1069 pci_print_supported_devices();
1071 } else if (pci_parse_slot(optarg) != 0)
1076 memflags |= VM_MEM_F_WIRED;
1079 error = vm_parse_memsize(optarg, &memsize);
1081 errx(EX_USAGE, "invalid memsize '%s'", optarg);
1084 guest_vmexit_on_hlt = 1;
1088 * The "-I" option was used to add an ioapic to the
1091 * An ioapic is now provided unconditionally for each
1092 * virtual machine and this option is now deprecated.
1096 guest_vmexit_on_pause = 1;
1105 guest_uuid_str = optarg;
1132 ctx = do_open(vmname);
1134 max_vcpus = num_vcpus_allowed(ctx);
1135 if (guest_ncpus > max_vcpus) {
1136 fprintf(stderr, "%d vCPUs requested but only %d available\n",
1137 guest_ncpus, max_vcpus);
1141 fbsdrun_set_capabilities(ctx, BSP);
1143 vm_set_memflags(ctx, memflags);
1144 err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL);
1146 fprintf(stderr, "Unable to setup memory (%d)\n", errno);
1152 fprintf(stderr, "init_msr error %d", error);
1162 rtc_init(ctx, rtc_localtime);
1166 * Exit if a device emulation finds an error in its initilization
1168 if (init_pci(ctx) != 0) {
1169 perror("device emulation initialization error");
1174 init_dbgport(dbg_port);
1177 init_gdb(ctx, gdb_port, gdb_stop);
1182 if (lpc_bootrom()) {
1183 if (vm_set_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, 1)) {
1184 fprintf(stderr, "ROM boot failed: unrestricted guest "
1185 "capability not available\n");
1188 error = vcpu_reset(ctx, BSP);
1192 error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip);
1196 * build the guest tables, MP etc.
1199 error = mptable_build(ctx, guest_ncpus);
1201 perror("error to build the guest tables");
1206 error = smbios_build(ctx);
1210 error = acpi_build(ctx, guest_ncpus);
1218 * Change the proc title to include the VM name.
1220 setproctitle("%s", vmname);
1222 #ifndef WITHOUT_CAPSICUM
1223 caph_cache_catpages();
1225 if (caph_limit_stdout() == -1 || caph_limit_stderr() == -1)
1226 errx(EX_OSERR, "Unable to apply rights for sandbox");
1228 if (caph_enter() == -1)
1229 errx(EX_OSERR, "cap_enter() failed");
1235 fbsdrun_addcpu(ctx, BSP, BSP, rip);
1238 * Head off to the main event dispatch loop