2 * Copyright (c) 2011 NetApp, Inc.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
14 * THIS SOFTWARE IS PROVIDED BY NETAPP, INC ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED. IN NO EVENT SHALL NETAPP, INC OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
32 #include <sys/types.h>
36 #include <machine/atomic.h>
37 #include <machine/segments.h>
48 #include <pthread_np.h>
51 #include <machine/vmm.h>
65 #include "smbiostbl.h"
67 #include "spinup_ap.h"
70 #define GUEST_NIO_PORT 0x488 /* guest upcalls via i/o port */
72 #define MB (1024UL * 1024)
73 #define GB (1024UL * MB)
75 typedef int (*vmexit_handler_t)(struct vmctx *, struct vm_exit *, int *vcpu);
76 extern int vmexit_task_switch(struct vmctx *, struct vm_exit *, int *vcpu);
83 static int guest_vmexit_on_hlt, guest_vmexit_on_pause;
84 static int virtio_msix = 1;
85 static int x2apic_mode = 0; /* default is xAPIC */
88 static int strictmsr = 1;
92 static char *progname;
93 static const int BSP = 0;
95 static cpuset_t cpumask;
97 static void vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip);
99 static struct vm_exit vmexit[VM_MAXCPU];
102 uint64_t vmexit_bogus;
103 uint64_t vmexit_reqidle;
105 uint64_t vmexit_pause;
106 uint64_t vmexit_mtrap;
107 uint64_t vmexit_inst_emul;
108 uint64_t cpu_switch_rotate;
109 uint64_t cpu_switch_direct;
114 struct vmctx *mt_ctx;
116 } mt_vmm_info[VM_MAXCPU];
118 static cpuset_t *vcpumap[VM_MAXCPU] = { NULL };
125 "Usage: %s [-abehuwxACHPWY] [-c vcpus] [-g <gdb port>] [-l <lpc>]\n"
126 " %*s [-m mem] [-p vcpu:hostcpu] [-s <pci>] [-U uuid] <vm>\n"
127 " -a: local apic is in xAPIC mode (deprecated)\n"
128 " -A: create ACPI tables\n"
129 " -c: # cpus (default 1)\n"
130 " -C: include guest memory in core file\n"
131 " -e: exit on unhandled I/O access\n"
134 " -H: vmexit from the guest on hlt\n"
135 " -l: LPC device configuration\n"
136 " -m: memory size in MB\n"
137 " -p: pin 'vcpu' to 'hostcpu'\n"
138 " -P: vmexit from the guest on pause\n"
139 " -s: <slot,driver,configinfo> PCI slot config\n"
140 " -u: RTC keeps UTC time\n"
142 " -w: ignore unimplemented MSRs\n"
143 " -W: force virtio to use single-vector MSI\n"
144 " -x: local apic is in x2APIC mode\n"
145 " -Y: disable MPtable generation\n",
146 progname, (int)strlen(progname), "");
152 pincpu_parse(const char *opt)
156 if (sscanf(opt, "%d:%d", &vcpu, &pcpu) != 2) {
157 fprintf(stderr, "invalid format: %s\n", opt);
161 if (vcpu < 0 || vcpu >= VM_MAXCPU) {
162 fprintf(stderr, "vcpu '%d' outside valid range from 0 to %d\n",
163 vcpu, VM_MAXCPU - 1);
167 if (pcpu < 0 || pcpu >= CPU_SETSIZE) {
168 fprintf(stderr, "hostcpu '%d' outside valid range from "
169 "0 to %d\n", pcpu, CPU_SETSIZE - 1);
173 if (vcpumap[vcpu] == NULL) {
174 if ((vcpumap[vcpu] = malloc(sizeof(cpuset_t))) == NULL) {
178 CPU_ZERO(vcpumap[vcpu]);
180 CPU_SET(pcpu, vcpumap[vcpu]);
185 vm_inject_fault(void *arg, int vcpu, int vector, int errcode_valid,
189 int error, restart_instruction;
192 restart_instruction = 1;
194 error = vm_inject_exception(ctx, vcpu, vector, errcode_valid, errcode,
195 restart_instruction);
200 paddr_guest2host(struct vmctx *ctx, uintptr_t gaddr, size_t len)
203 return (vm_map_gpa(ctx, gaddr, len));
207 fbsdrun_vmexit_on_pause(void)
210 return (guest_vmexit_on_pause);
214 fbsdrun_vmexit_on_hlt(void)
217 return (guest_vmexit_on_hlt);
221 fbsdrun_virtio_msix(void)
224 return (virtio_msix);
228 fbsdrun_start_thread(void *param)
230 char tname[MAXCOMLEN + 1];
231 struct mt_vmm_info *mtp;
237 snprintf(tname, sizeof(tname), "vcpu %d", vcpu);
238 pthread_set_name_np(mtp->mt_thr, tname);
240 vm_loop(mtp->mt_ctx, vcpu, vmexit[vcpu].rip);
248 fbsdrun_addcpu(struct vmctx *ctx, int fromcpu, int newcpu, uint64_t rip)
252 assert(fromcpu == BSP);
255 * The 'newcpu' must be activated in the context of 'fromcpu'. If
256 * vm_activate_cpu() is delayed until newcpu's pthread starts running
257 * then vmm.ko is out-of-sync with bhyve and this can create a race
260 error = vm_activate_cpu(ctx, newcpu);
263 CPU_SET_ATOMIC(newcpu, &cpumask);
266 * Set up the vmexit struct to allow execution to start
269 vmexit[newcpu].rip = rip;
270 vmexit[newcpu].inst_length = 0;
272 mt_vmm_info[newcpu].mt_ctx = ctx;
273 mt_vmm_info[newcpu].mt_vcpu = newcpu;
275 error = pthread_create(&mt_vmm_info[newcpu].mt_thr, NULL,
276 fbsdrun_start_thread, &mt_vmm_info[newcpu]);
281 fbsdrun_deletecpu(struct vmctx *ctx, int vcpu)
284 if (!CPU_ISSET(vcpu, &cpumask)) {
285 fprintf(stderr, "Attempting to delete unknown cpu %d\n", vcpu);
289 CPU_CLR_ATOMIC(vcpu, &cpumask);
290 return (CPU_EMPTY(&cpumask));
294 vmexit_handle_notify(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu,
299 * put guest-driven debug here
302 return (VMEXIT_CONTINUE);
306 vmexit_inout(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
309 int bytes, port, in, out, string;
314 port = vme->u.inout.port;
315 bytes = vme->u.inout.bytes;
316 string = vme->u.inout.string;
317 in = vme->u.inout.in;
320 /* Extra-special case of host notifications */
321 if (out && port == GUEST_NIO_PORT) {
322 error = vmexit_handle_notify(ctx, vme, pvcpu, vme->u.inout.eax);
326 error = emulate_inout(ctx, vcpu, vme, strictio);
328 fprintf(stderr, "Unhandled %s%c 0x%04x at 0x%lx\n",
330 bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'),
332 return (VMEXIT_ABORT);
334 return (VMEXIT_CONTINUE);
339 vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
346 error = emulate_rdmsr(ctx, *pvcpu, vme->u.msr.code, &val);
348 fprintf(stderr, "rdmsr to register %#x on vcpu %d\n",
349 vme->u.msr.code, *pvcpu);
351 vm_inject_gp(ctx, *pvcpu);
352 return (VMEXIT_CONTINUE);
357 error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RAX, eax);
361 error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RDX, edx);
364 return (VMEXIT_CONTINUE);
368 vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
372 error = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code, vme->u.msr.wval);
374 fprintf(stderr, "wrmsr to register %#x(%#lx) on vcpu %d\n",
375 vme->u.msr.code, vme->u.msr.wval, *pvcpu);
377 vm_inject_gp(ctx, *pvcpu);
378 return (VMEXIT_CONTINUE);
381 return (VMEXIT_CONTINUE);
385 vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
388 int retval = VMEXIT_CONTINUE;
390 newcpu = spinup_ap(ctx, *pvcpu,
391 vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip);
396 #define DEBUG_EPT_MISCONFIG
397 #ifdef DEBUG_EPT_MISCONFIG
398 #define EXIT_REASON_EPT_MISCONFIG 49
399 #define VMCS_GUEST_PHYSICAL_ADDRESS 0x00002400
400 #define VMCS_IDENT(x) ((x) | 0x80000000)
402 static uint64_t ept_misconfig_gpa, ept_misconfig_pte[4];
403 static int ept_misconfig_ptenum;
407 vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
410 fprintf(stderr, "vm exit[%d]\n", *pvcpu);
411 fprintf(stderr, "\treason\t\tVMX\n");
412 fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
413 fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
414 fprintf(stderr, "\tstatus\t\t%d\n", vmexit->u.vmx.status);
415 fprintf(stderr, "\texit_reason\t%u\n", vmexit->u.vmx.exit_reason);
416 fprintf(stderr, "\tqualification\t0x%016lx\n",
417 vmexit->u.vmx.exit_qualification);
418 fprintf(stderr, "\tinst_type\t\t%d\n", vmexit->u.vmx.inst_type);
419 fprintf(stderr, "\tinst_error\t\t%d\n", vmexit->u.vmx.inst_error);
420 #ifdef DEBUG_EPT_MISCONFIG
421 if (vmexit->u.vmx.exit_reason == EXIT_REASON_EPT_MISCONFIG) {
422 vm_get_register(ctx, *pvcpu,
423 VMCS_IDENT(VMCS_GUEST_PHYSICAL_ADDRESS),
425 vm_get_gpa_pmap(ctx, ept_misconfig_gpa, ept_misconfig_pte,
426 &ept_misconfig_ptenum);
427 fprintf(stderr, "\tEPT misconfiguration:\n");
428 fprintf(stderr, "\t\tGPA: %#lx\n", ept_misconfig_gpa);
429 fprintf(stderr, "\t\tPTE(%d): %#lx %#lx %#lx %#lx\n",
430 ept_misconfig_ptenum, ept_misconfig_pte[0],
431 ept_misconfig_pte[1], ept_misconfig_pte[2],
432 ept_misconfig_pte[3]);
434 #endif /* DEBUG_EPT_MISCONFIG */
435 return (VMEXIT_ABORT);
439 vmexit_svm(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
442 fprintf(stderr, "vm exit[%d]\n", *pvcpu);
443 fprintf(stderr, "\treason\t\tSVM\n");
444 fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
445 fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
446 fprintf(stderr, "\texitcode\t%#lx\n", vmexit->u.svm.exitcode);
447 fprintf(stderr, "\texitinfo1\t%#lx\n", vmexit->u.svm.exitinfo1);
448 fprintf(stderr, "\texitinfo2\t%#lx\n", vmexit->u.svm.exitinfo2);
449 return (VMEXIT_ABORT);
453 vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
456 assert(vmexit->inst_length == 0);
458 stats.vmexit_bogus++;
460 return (VMEXIT_CONTINUE);
464 vmexit_reqidle(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
467 assert(vmexit->inst_length == 0);
469 stats.vmexit_reqidle++;
471 return (VMEXIT_CONTINUE);
475 vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
481 * Just continue execution with the next instruction. We use
482 * the HLT VM exit as a way to be friendly with the host
485 return (VMEXIT_CONTINUE);
489 vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
492 stats.vmexit_pause++;
494 return (VMEXIT_CONTINUE);
498 vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
501 assert(vmexit->inst_length == 0);
503 stats.vmexit_mtrap++;
505 return (VMEXIT_CONTINUE);
509 vmexit_inst_emul(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
514 stats.vmexit_inst_emul++;
516 vie = &vmexit->u.inst_emul.vie;
517 err = emulate_mem(ctx, *pvcpu, vmexit->u.inst_emul.gpa,
518 vie, &vmexit->u.inst_emul.paging);
522 fprintf(stderr, "Unhandled memory access to 0x%lx\n",
523 vmexit->u.inst_emul.gpa);
526 fprintf(stderr, "Failed to emulate instruction [");
527 for (i = 0; i < vie->num_valid; i++) {
528 fprintf(stderr, "0x%02x%s", vie->inst[i],
529 i != (vie->num_valid - 1) ? " " : "");
531 fprintf(stderr, "] at 0x%lx\n", vmexit->rip);
532 return (VMEXIT_ABORT);
535 return (VMEXIT_CONTINUE);
538 static pthread_mutex_t resetcpu_mtx = PTHREAD_MUTEX_INITIALIZER;
539 static pthread_cond_t resetcpu_cond = PTHREAD_COND_INITIALIZER;
542 vmexit_suspend(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
544 enum vm_suspend_how how;
546 how = vmexit->u.suspended.how;
548 fbsdrun_deletecpu(ctx, *pvcpu);
551 pthread_mutex_lock(&resetcpu_mtx);
552 pthread_cond_signal(&resetcpu_cond);
553 pthread_mutex_unlock(&resetcpu_mtx);
557 pthread_mutex_lock(&resetcpu_mtx);
558 while (!CPU_EMPTY(&cpumask)) {
559 pthread_cond_wait(&resetcpu_cond, &resetcpu_mtx);
561 pthread_mutex_unlock(&resetcpu_mtx);
564 case VM_SUSPEND_RESET:
566 case VM_SUSPEND_POWEROFF:
568 case VM_SUSPEND_HALT:
570 case VM_SUSPEND_TRIPLEFAULT:
573 fprintf(stderr, "vmexit_suspend: invalid reason %d\n", how);
576 return (0); /* NOTREACHED */
579 static vmexit_handler_t handler[VM_EXITCODE_MAX] = {
580 [VM_EXITCODE_INOUT] = vmexit_inout,
581 [VM_EXITCODE_INOUT_STR] = vmexit_inout,
582 [VM_EXITCODE_VMX] = vmexit_vmx,
583 [VM_EXITCODE_SVM] = vmexit_svm,
584 [VM_EXITCODE_BOGUS] = vmexit_bogus,
585 [VM_EXITCODE_REQIDLE] = vmexit_reqidle,
586 [VM_EXITCODE_RDMSR] = vmexit_rdmsr,
587 [VM_EXITCODE_WRMSR] = vmexit_wrmsr,
588 [VM_EXITCODE_MTRAP] = vmexit_mtrap,
589 [VM_EXITCODE_INST_EMUL] = vmexit_inst_emul,
590 [VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap,
591 [VM_EXITCODE_SUSPENDED] = vmexit_suspend,
592 [VM_EXITCODE_TASK_SWITCH] = vmexit_task_switch,
596 vm_loop(struct vmctx *ctx, int vcpu, uint64_t startrip)
598 int error, rc, prevcpu;
599 enum vm_exitcode exitcode;
600 cpuset_t active_cpus;
602 if (vcpumap[vcpu] != NULL) {
603 error = pthread_setaffinity_np(pthread_self(),
604 sizeof(cpuset_t), vcpumap[vcpu]);
608 error = vm_active_cpus(ctx, &active_cpus);
609 assert(CPU_ISSET(vcpu, &active_cpus));
611 error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RIP, startrip);
615 error = vm_run(ctx, vcpu, &vmexit[vcpu]);
621 exitcode = vmexit[vcpu].exitcode;
622 if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) {
623 fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n",
628 rc = (*handler[exitcode])(ctx, &vmexit[vcpu], &vcpu);
631 case VMEXIT_CONTINUE:
639 fprintf(stderr, "vm_run error %d, errno %d\n", error, errno);
643 num_vcpus_allowed(struct vmctx *ctx)
647 error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp);
650 * The guest is allowed to spinup more than one processor only if the
651 * UNRESTRICTED_GUEST capability is available.
660 fbsdrun_set_capabilities(struct vmctx *ctx, int cpu)
664 if (fbsdrun_vmexit_on_hlt()) {
665 err = vm_get_capability(ctx, cpu, VM_CAP_HALT_EXIT, &tmp);
667 fprintf(stderr, "VM exit on HLT not supported\n");
670 vm_set_capability(ctx, cpu, VM_CAP_HALT_EXIT, 1);
672 handler[VM_EXITCODE_HLT] = vmexit_hlt;
675 if (fbsdrun_vmexit_on_pause()) {
677 * pause exit support required for this mode
679 err = vm_get_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, &tmp);
682 "SMP mux requested, no pause support\n");
685 vm_set_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, 1);
687 handler[VM_EXITCODE_PAUSE] = vmexit_pause;
691 err = vm_set_x2apic_state(ctx, cpu, X2APIC_ENABLED);
693 err = vm_set_x2apic_state(ctx, cpu, X2APIC_DISABLED);
696 fprintf(stderr, "Unable to set x2apic state (%d)\n", err);
700 vm_set_capability(ctx, cpu, VM_CAP_ENABLE_INVPCID, 1);
704 main(int argc, char *argv[])
706 int c, error, gdb_port, err, bvmcons;
707 int dump_guest_memory, max_vcpus, mptgen;
714 dump_guest_memory = 0;
715 progname = basename(argv[0]);
722 while ((c = getopt(argc, argv, "abehuwxACHIPWYp:g:c:s:m:l:U:")) != -1) {
734 if (pincpu_parse(optarg) != 0) {
735 errx(EX_USAGE, "invalid vcpu pinning "
736 "configuration '%s'", optarg);
740 guest_ncpus = atoi(optarg);
743 dump_guest_memory = 1;
746 gdb_port = atoi(optarg);
749 if (lpc_device_parse(optarg) != 0) {
750 errx(EX_USAGE, "invalid lpc device "
751 "configuration '%s'", optarg);
755 if (pci_parse_slot(optarg) != 0)
760 error = vm_parse_memsize(optarg, &memsize);
762 errx(EX_USAGE, "invalid memsize '%s'", optarg);
765 guest_vmexit_on_hlt = 1;
769 * The "-I" option was used to add an ioapic to the
772 * An ioapic is now provided unconditionally for each
773 * virtual machine and this option is now deprecated.
777 guest_vmexit_on_pause = 1;
786 guest_uuid_str = optarg;
814 ctx = vm_open(vmname);
820 if (guest_ncpus < 1) {
821 fprintf(stderr, "Invalid guest vCPUs (%d)\n", guest_ncpus);
825 max_vcpus = num_vcpus_allowed(ctx);
826 if (guest_ncpus > max_vcpus) {
827 fprintf(stderr, "%d vCPUs requested but only %d available\n",
828 guest_ncpus, max_vcpus);
832 fbsdrun_set_capabilities(ctx, BSP);
834 if (dump_guest_memory)
835 vm_set_memflags(ctx, VM_MEM_F_INCORE);
836 err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL);
838 fprintf(stderr, "Unable to setup memory (%d)\n", err);
844 fprintf(stderr, "init_msr error %d", error);
853 rtc_init(ctx, rtc_localtime);
857 * Exit if a device emulation finds an error in it's initilization
859 if (init_pci(ctx) != 0)
863 init_dbgport(gdb_port);
868 error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip);
872 * build the guest tables, MP etc.
875 error = mptable_build(ctx, guest_ncpus);
880 error = smbios_build(ctx);
884 error = acpi_build(ctx, guest_ncpus);
889 * Change the proc title to include the VM name.
891 setproctitle("%s", vmname);
896 fbsdrun_addcpu(ctx, BSP, BSP, rip);
899 * Head off to the main event dispatch loop