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 <machine/atomic.h>
42 #include <machine/segments.h>
44 #ifndef WITHOUT_CAPSICUM
45 #include <capsicum_helpers.h>
57 #include <pthread_np.h>
62 #include <machine/vmm.h>
63 #ifndef WITHOUT_CAPSICUM
64 #include <machine/vmm_dev.h>
82 #include "smbiostbl.h"
84 #include "spinup_ap.h"
87 #define GUEST_NIO_PORT 0x488 /* guest upcalls via i/o port */
89 #define MB (1024UL * 1024)
90 #define GB (1024UL * MB)
92 typedef int (*vmexit_handler_t)(struct vmctx *, struct vm_exit *, int *vcpu);
93 extern int vmexit_task_switch(struct vmctx *, struct vm_exit *, int *vcpu);
98 uint16_t cores, maxcpus, sockets, threads;
100 char *guest_uuid_str;
102 static int guest_vmexit_on_hlt, guest_vmexit_on_pause;
103 static int virtio_msix = 1;
104 static int x2apic_mode = 0; /* default is xAPIC */
107 static int strictmsr = 1;
111 static char *progname;
112 static const int BSP = 0;
114 static cpuset_t cpumask;
116 static void vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip);
118 static struct vm_exit vmexit[VM_MAXCPU];
121 uint64_t vmexit_bogus;
122 uint64_t vmexit_reqidle;
124 uint64_t vmexit_pause;
125 uint64_t vmexit_mtrap;
126 uint64_t vmexit_inst_emul;
127 uint64_t cpu_switch_rotate;
128 uint64_t cpu_switch_direct;
133 struct vmctx *mt_ctx;
135 } mt_vmm_info[VM_MAXCPU];
137 static cpuset_t *vcpumap[VM_MAXCPU] = { NULL };
144 "Usage: %s [-abehuwxACHPSWY]\n"
145 " %*s [-c [[cpus=]numcpus][,sockets=n][,cores=n][,threads=n]]\n"
146 " %*s [-g <gdb port>] [-l <lpc>]\n"
147 " %*s [-m mem] [-p vcpu:hostcpu] [-s <pci>] [-U uuid] <vm>\n"
148 " -a: local apic is in xAPIC mode (deprecated)\n"
149 " -A: create ACPI tables\n"
150 " -c: number of cpus and/or topology specification"
151 " -C: include guest memory in core file\n"
152 " -e: exit on unhandled I/O access\n"
155 " -H: vmexit from the guest on hlt\n"
156 " -l: LPC device configuration\n"
157 " -m: memory size in MB\n"
158 " -p: pin 'vcpu' to 'hostcpu'\n"
159 " -P: vmexit from the guest on pause\n"
160 " -s: <slot,driver,configinfo> PCI slot config\n"
161 " -S: guest memory cannot be swapped\n"
162 " -u: RTC keeps UTC time\n"
164 " -w: ignore unimplemented MSRs\n"
165 " -W: force virtio to use single-vector MSI\n"
166 " -x: local apic is in x2APIC mode\n"
167 " -Y: disable MPtable generation\n",
168 progname, (int)strlen(progname), "", (int)strlen(progname), "",
169 (int)strlen(progname), "");
175 * XXX This parser is known to have the following issues:
176 * 1. It accepts null key=value tokens ",,".
177 * 2. It accepts whitespace after = and before value.
178 * 3. Values out of range of INT are silently wrapped.
179 * 4. It doesn't check non-final values.
180 * 5. The apparently bogus limits of UINT16_MAX are for future expansion.
182 * The acceptance of a null specification ('-c ""') is by design to match the
183 * manual page syntax specification, this results in a topology of 1 vCPU.
186 topology_parse(const char *opt)
189 int c, chk, n, s, t, tmp;
193 c = 1, n = 1, s = 1, t = 1;
194 ns = false, scts = false;
197 while ((cp = strsep(&str, ",")) != NULL) {
198 if (sscanf(cp, "%i%n", &tmp, &chk) == 1) {
201 } else if (sscanf(cp, "cpus=%i%n", &tmp, &chk) == 1) {
204 } else if (sscanf(cp, "sockets=%i%n", &tmp, &chk) == 1) {
207 } else if (sscanf(cp, "cores=%i%n", &tmp, &chk) == 1) {
210 } else if (sscanf(cp, "threads=%i%n", &tmp, &chk) == 1) {
213 #ifdef notyet /* Do not expose this until vmm.ko implements it */
214 } else if (sscanf(cp, "maxcpus=%i%n", &tmp, &chk) == 1) {
217 /* Skip the empty argument case from -c "" */
218 } else if (cp[0] == '\0')
222 /* Any trailing garbage causes an error */
227 * Range check 1 <= n <= UINT16_MAX all values
229 if (n < 1 || s < 1 || c < 1 || t < 1 ||
230 n > UINT16_MAX || s > UINT16_MAX || c > UINT16_MAX ||
234 /* If only the cpus was specified, use that as sockets */
238 * Compute sockets * cores * threads avoiding overflow
239 * The range check above insures these are 16 bit values
240 * If n was specified check it against computed ncpus
242 ncpus = (uint64_t)s * c * t;
243 if (ncpus > UINT16_MAX || (ns && n != ncpus))
254 pincpu_parse(const char *opt)
258 if (sscanf(opt, "%d:%d", &vcpu, &pcpu) != 2) {
259 fprintf(stderr, "invalid format: %s\n", opt);
263 if (vcpu < 0 || vcpu >= VM_MAXCPU) {
264 fprintf(stderr, "vcpu '%d' outside valid range from 0 to %d\n",
265 vcpu, VM_MAXCPU - 1);
269 if (pcpu < 0 || pcpu >= CPU_SETSIZE) {
270 fprintf(stderr, "hostcpu '%d' outside valid range from "
271 "0 to %d\n", pcpu, CPU_SETSIZE - 1);
275 if (vcpumap[vcpu] == NULL) {
276 if ((vcpumap[vcpu] = malloc(sizeof(cpuset_t))) == NULL) {
280 CPU_ZERO(vcpumap[vcpu]);
282 CPU_SET(pcpu, vcpumap[vcpu]);
287 vm_inject_fault(void *arg, int vcpu, int vector, int errcode_valid,
291 int error, restart_instruction;
294 restart_instruction = 1;
296 error = vm_inject_exception(ctx, vcpu, vector, errcode_valid, errcode,
297 restart_instruction);
302 paddr_guest2host(struct vmctx *ctx, uintptr_t gaddr, size_t len)
305 return (vm_map_gpa(ctx, gaddr, len));
309 fbsdrun_vmexit_on_pause(void)
312 return (guest_vmexit_on_pause);
316 fbsdrun_vmexit_on_hlt(void)
319 return (guest_vmexit_on_hlt);
323 fbsdrun_virtio_msix(void)
326 return (virtio_msix);
330 fbsdrun_start_thread(void *param)
332 char tname[MAXCOMLEN + 1];
333 struct mt_vmm_info *mtp;
339 snprintf(tname, sizeof(tname), "vcpu %d", vcpu);
340 pthread_set_name_np(mtp->mt_thr, tname);
344 vm_loop(mtp->mt_ctx, vcpu, vmexit[vcpu].rip);
352 fbsdrun_addcpu(struct vmctx *ctx, int fromcpu, int newcpu, uint64_t rip)
356 assert(fromcpu == BSP);
359 * The 'newcpu' must be activated in the context of 'fromcpu'. If
360 * vm_activate_cpu() is delayed until newcpu's pthread starts running
361 * then vmm.ko is out-of-sync with bhyve and this can create a race
364 error = vm_activate_cpu(ctx, newcpu);
366 err(EX_OSERR, "could not activate CPU %d", newcpu);
368 CPU_SET_ATOMIC(newcpu, &cpumask);
371 * Set up the vmexit struct to allow execution to start
374 vmexit[newcpu].rip = rip;
375 vmexit[newcpu].inst_length = 0;
377 mt_vmm_info[newcpu].mt_ctx = ctx;
378 mt_vmm_info[newcpu].mt_vcpu = newcpu;
380 error = pthread_create(&mt_vmm_info[newcpu].mt_thr, NULL,
381 fbsdrun_start_thread, &mt_vmm_info[newcpu]);
386 fbsdrun_deletecpu(struct vmctx *ctx, int vcpu)
389 if (!CPU_ISSET(vcpu, &cpumask)) {
390 fprintf(stderr, "Attempting to delete unknown cpu %d\n", vcpu);
394 CPU_CLR_ATOMIC(vcpu, &cpumask);
395 return (CPU_EMPTY(&cpumask));
399 vmexit_handle_notify(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu,
404 * put guest-driven debug here
407 return (VMEXIT_CONTINUE);
411 vmexit_inout(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
414 int bytes, port, in, out;
419 port = vme->u.inout.port;
420 bytes = vme->u.inout.bytes;
421 in = vme->u.inout.in;
424 /* Extra-special case of host notifications */
425 if (out && port == GUEST_NIO_PORT) {
426 error = vmexit_handle_notify(ctx, vme, pvcpu, vme->u.inout.eax);
430 error = emulate_inout(ctx, vcpu, vme, strictio);
432 fprintf(stderr, "Unhandled %s%c 0x%04x at 0x%lx\n",
434 bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'),
436 return (VMEXIT_ABORT);
438 return (VMEXIT_CONTINUE);
443 vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
450 error = emulate_rdmsr(ctx, *pvcpu, vme->u.msr.code, &val);
452 fprintf(stderr, "rdmsr to register %#x on vcpu %d\n",
453 vme->u.msr.code, *pvcpu);
455 vm_inject_gp(ctx, *pvcpu);
456 return (VMEXIT_CONTINUE);
461 error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RAX, eax);
465 error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RDX, edx);
468 return (VMEXIT_CONTINUE);
472 vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
476 error = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code, vme->u.msr.wval);
478 fprintf(stderr, "wrmsr to register %#x(%#lx) on vcpu %d\n",
479 vme->u.msr.code, vme->u.msr.wval, *pvcpu);
481 vm_inject_gp(ctx, *pvcpu);
482 return (VMEXIT_CONTINUE);
485 return (VMEXIT_CONTINUE);
489 vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
492 (void)spinup_ap(ctx, *pvcpu,
493 vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip);
495 return (VMEXIT_CONTINUE);
498 #define DEBUG_EPT_MISCONFIG
499 #ifdef DEBUG_EPT_MISCONFIG
500 #define EXIT_REASON_EPT_MISCONFIG 49
501 #define VMCS_GUEST_PHYSICAL_ADDRESS 0x00002400
502 #define VMCS_IDENT(x) ((x) | 0x80000000)
504 static uint64_t ept_misconfig_gpa, ept_misconfig_pte[4];
505 static int ept_misconfig_ptenum;
509 vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
512 fprintf(stderr, "vm exit[%d]\n", *pvcpu);
513 fprintf(stderr, "\treason\t\tVMX\n");
514 fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
515 fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
516 fprintf(stderr, "\tstatus\t\t%d\n", vmexit->u.vmx.status);
517 fprintf(stderr, "\texit_reason\t%u\n", vmexit->u.vmx.exit_reason);
518 fprintf(stderr, "\tqualification\t0x%016lx\n",
519 vmexit->u.vmx.exit_qualification);
520 fprintf(stderr, "\tinst_type\t\t%d\n", vmexit->u.vmx.inst_type);
521 fprintf(stderr, "\tinst_error\t\t%d\n", vmexit->u.vmx.inst_error);
522 #ifdef DEBUG_EPT_MISCONFIG
523 if (vmexit->u.vmx.exit_reason == EXIT_REASON_EPT_MISCONFIG) {
524 vm_get_register(ctx, *pvcpu,
525 VMCS_IDENT(VMCS_GUEST_PHYSICAL_ADDRESS),
527 vm_get_gpa_pmap(ctx, ept_misconfig_gpa, ept_misconfig_pte,
528 &ept_misconfig_ptenum);
529 fprintf(stderr, "\tEPT misconfiguration:\n");
530 fprintf(stderr, "\t\tGPA: %#lx\n", ept_misconfig_gpa);
531 fprintf(stderr, "\t\tPTE(%d): %#lx %#lx %#lx %#lx\n",
532 ept_misconfig_ptenum, ept_misconfig_pte[0],
533 ept_misconfig_pte[1], ept_misconfig_pte[2],
534 ept_misconfig_pte[3]);
536 #endif /* DEBUG_EPT_MISCONFIG */
537 return (VMEXIT_ABORT);
541 vmexit_svm(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
544 fprintf(stderr, "vm exit[%d]\n", *pvcpu);
545 fprintf(stderr, "\treason\t\tSVM\n");
546 fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
547 fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
548 fprintf(stderr, "\texitcode\t%#lx\n", vmexit->u.svm.exitcode);
549 fprintf(stderr, "\texitinfo1\t%#lx\n", vmexit->u.svm.exitinfo1);
550 fprintf(stderr, "\texitinfo2\t%#lx\n", vmexit->u.svm.exitinfo2);
551 return (VMEXIT_ABORT);
555 vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
558 assert(vmexit->inst_length == 0);
560 stats.vmexit_bogus++;
562 return (VMEXIT_CONTINUE);
566 vmexit_reqidle(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
569 assert(vmexit->inst_length == 0);
571 stats.vmexit_reqidle++;
573 return (VMEXIT_CONTINUE);
577 vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
583 * Just continue execution with the next instruction. We use
584 * the HLT VM exit as a way to be friendly with the host
587 return (VMEXIT_CONTINUE);
591 vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
594 stats.vmexit_pause++;
596 return (VMEXIT_CONTINUE);
600 vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
603 assert(vmexit->inst_length == 0);
605 stats.vmexit_mtrap++;
607 gdb_cpu_mtrap(*pvcpu);
609 return (VMEXIT_CONTINUE);
613 vmexit_inst_emul(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
618 stats.vmexit_inst_emul++;
620 vie = &vmexit->u.inst_emul.vie;
621 err = emulate_mem(ctx, *pvcpu, vmexit->u.inst_emul.gpa,
622 vie, &vmexit->u.inst_emul.paging);
626 fprintf(stderr, "Unhandled memory access to 0x%lx\n",
627 vmexit->u.inst_emul.gpa);
630 fprintf(stderr, "Failed to emulate instruction [");
631 for (i = 0; i < vie->num_valid; i++) {
632 fprintf(stderr, "0x%02x%s", vie->inst[i],
633 i != (vie->num_valid - 1) ? " " : "");
635 fprintf(stderr, "] at 0x%lx\n", vmexit->rip);
636 return (VMEXIT_ABORT);
639 return (VMEXIT_CONTINUE);
642 static pthread_mutex_t resetcpu_mtx = PTHREAD_MUTEX_INITIALIZER;
643 static pthread_cond_t resetcpu_cond = PTHREAD_COND_INITIALIZER;
646 vmexit_suspend(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
648 enum vm_suspend_how how;
650 how = vmexit->u.suspended.how;
652 fbsdrun_deletecpu(ctx, *pvcpu);
655 pthread_mutex_lock(&resetcpu_mtx);
656 pthread_cond_signal(&resetcpu_cond);
657 pthread_mutex_unlock(&resetcpu_mtx);
661 pthread_mutex_lock(&resetcpu_mtx);
662 while (!CPU_EMPTY(&cpumask)) {
663 pthread_cond_wait(&resetcpu_cond, &resetcpu_mtx);
665 pthread_mutex_unlock(&resetcpu_mtx);
668 case VM_SUSPEND_RESET:
670 case VM_SUSPEND_POWEROFF:
672 case VM_SUSPEND_HALT:
674 case VM_SUSPEND_TRIPLEFAULT:
677 fprintf(stderr, "vmexit_suspend: invalid reason %d\n", how);
680 return (0); /* NOTREACHED */
684 vmexit_debug(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
687 gdb_cpu_suspend(*pvcpu);
688 return (VMEXIT_CONTINUE);
691 static vmexit_handler_t handler[VM_EXITCODE_MAX] = {
692 [VM_EXITCODE_INOUT] = vmexit_inout,
693 [VM_EXITCODE_INOUT_STR] = vmexit_inout,
694 [VM_EXITCODE_VMX] = vmexit_vmx,
695 [VM_EXITCODE_SVM] = vmexit_svm,
696 [VM_EXITCODE_BOGUS] = vmexit_bogus,
697 [VM_EXITCODE_REQIDLE] = vmexit_reqidle,
698 [VM_EXITCODE_RDMSR] = vmexit_rdmsr,
699 [VM_EXITCODE_WRMSR] = vmexit_wrmsr,
700 [VM_EXITCODE_MTRAP] = vmexit_mtrap,
701 [VM_EXITCODE_INST_EMUL] = vmexit_inst_emul,
702 [VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap,
703 [VM_EXITCODE_SUSPENDED] = vmexit_suspend,
704 [VM_EXITCODE_TASK_SWITCH] = vmexit_task_switch,
705 [VM_EXITCODE_DEBUG] = vmexit_debug,
709 vm_loop(struct vmctx *ctx, int vcpu, uint64_t startrip)
712 enum vm_exitcode exitcode;
713 cpuset_t active_cpus;
715 if (vcpumap[vcpu] != NULL) {
716 error = pthread_setaffinity_np(pthread_self(),
717 sizeof(cpuset_t), vcpumap[vcpu]);
721 error = vm_active_cpus(ctx, &active_cpus);
722 assert(CPU_ISSET(vcpu, &active_cpus));
724 error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RIP, startrip);
728 error = vm_run(ctx, vcpu, &vmexit[vcpu]);
732 exitcode = vmexit[vcpu].exitcode;
733 if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) {
734 fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n",
739 rc = (*handler[exitcode])(ctx, &vmexit[vcpu], &vcpu);
742 case VMEXIT_CONTINUE:
750 fprintf(stderr, "vm_run error %d, errno %d\n", error, errno);
754 num_vcpus_allowed(struct vmctx *ctx)
758 error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp);
761 * The guest is allowed to spinup more than one processor only if the
762 * UNRESTRICTED_GUEST capability is available.
771 fbsdrun_set_capabilities(struct vmctx *ctx, int cpu)
775 if (fbsdrun_vmexit_on_hlt()) {
776 err = vm_get_capability(ctx, cpu, VM_CAP_HALT_EXIT, &tmp);
778 fprintf(stderr, "VM exit on HLT not supported\n");
781 vm_set_capability(ctx, cpu, VM_CAP_HALT_EXIT, 1);
783 handler[VM_EXITCODE_HLT] = vmexit_hlt;
786 if (fbsdrun_vmexit_on_pause()) {
788 * pause exit support required for this mode
790 err = vm_get_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, &tmp);
793 "SMP mux requested, no pause support\n");
796 vm_set_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, 1);
798 handler[VM_EXITCODE_PAUSE] = vmexit_pause;
802 err = vm_set_x2apic_state(ctx, cpu, X2APIC_ENABLED);
804 err = vm_set_x2apic_state(ctx, cpu, X2APIC_DISABLED);
807 fprintf(stderr, "Unable to set x2apic state (%d)\n", err);
811 vm_set_capability(ctx, cpu, VM_CAP_ENABLE_INVPCID, 1);
814 static struct vmctx *
815 do_open(const char *vmname)
819 bool reinit, romboot;
820 #ifndef WITHOUT_CAPSICUM
822 const cap_ioctl_t *cmds;
826 reinit = romboot = false;
831 error = vm_create(vmname);
833 if (errno == EEXIST) {
838 * The virtual machine has been setup by the
839 * userspace bootloader.
849 * If the virtual machine was just created then a
850 * bootrom must be configured to boot it.
852 fprintf(stderr, "virtual machine cannot be booted\n");
857 ctx = vm_open(vmname);
863 #ifndef WITHOUT_CAPSICUM
864 cap_rights_init(&rights, CAP_IOCTL, CAP_MMAP_RW);
865 if (cap_rights_limit(vm_get_device_fd(ctx), &rights) == -1 &&
867 errx(EX_OSERR, "Unable to apply rights for sandbox");
868 vm_get_ioctls(&ncmds);
869 cmds = vm_get_ioctls(NULL);
871 errx(EX_OSERR, "out of memory");
872 if (cap_ioctls_limit(vm_get_device_fd(ctx), cmds, ncmds) == -1 &&
874 errx(EX_OSERR, "Unable to apply rights for sandbox");
875 free((cap_ioctl_t *)cmds);
879 error = vm_reinit(ctx);
885 error = vm_set_topology(ctx, sockets, cores, threads, maxcpus);
887 errx(EX_OSERR, "vm_set_topology");
892 main(int argc, char *argv[])
894 int c, error, dbg_port, gdb_port, err, bvmcons;
895 int max_vcpus, mptgen, memflags;
904 progname = basename(argv[0]);
909 sockets = cores = threads = 1;
916 optstr = "abehuwxACHIPSWYp:g:G:c:s:m:l:U:";
917 while ((c = getopt(argc, argv, optstr)) != -1) {
929 if (pincpu_parse(optarg) != 0) {
930 errx(EX_USAGE, "invalid vcpu pinning "
931 "configuration '%s'", optarg);
935 if (topology_parse(optarg) != 0) {
936 errx(EX_USAGE, "invalid cpu topology "
941 memflags |= VM_MEM_F_INCORE;
944 dbg_port = atoi(optarg);
947 if (optarg[0] == 'w') {
951 gdb_port = atoi(optarg);
954 if (lpc_device_parse(optarg) != 0) {
955 errx(EX_USAGE, "invalid lpc device "
956 "configuration '%s'", optarg);
960 if (pci_parse_slot(optarg) != 0)
965 memflags |= VM_MEM_F_WIRED;
968 error = vm_parse_memsize(optarg, &memsize);
970 errx(EX_USAGE, "invalid memsize '%s'", optarg);
973 guest_vmexit_on_hlt = 1;
977 * The "-I" option was used to add an ioapic to the
980 * An ioapic is now provided unconditionally for each
981 * virtual machine and this option is now deprecated.
985 guest_vmexit_on_pause = 1;
994 guest_uuid_str = optarg;
1021 ctx = do_open(vmname);
1023 max_vcpus = num_vcpus_allowed(ctx);
1024 if (guest_ncpus > max_vcpus) {
1025 fprintf(stderr, "%d vCPUs requested but only %d available\n",
1026 guest_ncpus, max_vcpus);
1030 fbsdrun_set_capabilities(ctx, BSP);
1032 vm_set_memflags(ctx, memflags);
1033 err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL);
1035 fprintf(stderr, "Unable to setup memory (%d)\n", errno);
1041 fprintf(stderr, "init_msr error %d", error);
1051 rtc_init(ctx, rtc_localtime);
1055 * Exit if a device emulation finds an error in its initilization
1057 if (init_pci(ctx) != 0)
1061 init_dbgport(dbg_port);
1064 init_gdb(ctx, gdb_port, gdb_stop);
1069 if (lpc_bootrom()) {
1070 if (vm_set_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, 1)) {
1071 fprintf(stderr, "ROM boot failed: unrestricted guest "
1072 "capability not available\n");
1075 error = vcpu_reset(ctx, BSP);
1079 error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip);
1083 * build the guest tables, MP etc.
1086 error = mptable_build(ctx, guest_ncpus);
1091 error = smbios_build(ctx);
1095 error = acpi_build(ctx, guest_ncpus);
1102 #ifndef WITHOUT_CAPSICUM
1103 caph_cache_catpages();
1105 if (caph_limit_stdout() == -1 || caph_limit_stderr() == -1)
1106 errx(EX_OSERR, "Unable to apply rights for sandbox");
1108 if (cap_enter() == -1 && errno != ENOSYS)
1109 errx(EX_OSERR, "cap_enter() failed");
1113 * Change the proc title to include the VM name.
1115 setproctitle("%s", vmname);
1120 fbsdrun_addcpu(ctx, BSP, BSP, rip);
1123 * Head off to the main event dispatch loop