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_bogus_switch;
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\n", in ? "in" : "out",
329 bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'), port);
330 return (VMEXIT_ABORT);
332 return (VMEXIT_CONTINUE);
337 vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
344 error = emulate_rdmsr(ctx, *pvcpu, vme->u.msr.code, &val);
346 fprintf(stderr, "rdmsr to register %#x on vcpu %d\n",
347 vme->u.msr.code, *pvcpu);
349 vm_inject_gp(ctx, *pvcpu);
350 return (VMEXIT_CONTINUE);
355 error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RAX, eax);
359 error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RDX, edx);
362 return (VMEXIT_CONTINUE);
366 vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
370 error = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code, vme->u.msr.wval);
372 fprintf(stderr, "wrmsr to register %#x(%#lx) on vcpu %d\n",
373 vme->u.msr.code, vme->u.msr.wval, *pvcpu);
375 vm_inject_gp(ctx, *pvcpu);
376 return (VMEXIT_CONTINUE);
379 return (VMEXIT_CONTINUE);
383 vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
386 int retval = VMEXIT_CONTINUE;
388 newcpu = spinup_ap(ctx, *pvcpu,
389 vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip);
394 #define DEBUG_EPT_MISCONFIG
395 #ifdef DEBUG_EPT_MISCONFIG
396 #define EXIT_REASON_EPT_MISCONFIG 49
397 #define VMCS_GUEST_PHYSICAL_ADDRESS 0x00002400
398 #define VMCS_IDENT(x) ((x) | 0x80000000)
400 static uint64_t ept_misconfig_gpa, ept_misconfig_pte[4];
401 static int ept_misconfig_ptenum;
405 vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
408 fprintf(stderr, "vm exit[%d]\n", *pvcpu);
409 fprintf(stderr, "\treason\t\tVMX\n");
410 fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
411 fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
412 fprintf(stderr, "\tstatus\t\t%d\n", vmexit->u.vmx.status);
413 fprintf(stderr, "\texit_reason\t%u\n", vmexit->u.vmx.exit_reason);
414 fprintf(stderr, "\tqualification\t0x%016lx\n",
415 vmexit->u.vmx.exit_qualification);
416 fprintf(stderr, "\tinst_type\t\t%d\n", vmexit->u.vmx.inst_type);
417 fprintf(stderr, "\tinst_error\t\t%d\n", vmexit->u.vmx.inst_error);
418 #ifdef DEBUG_EPT_MISCONFIG
419 if (vmexit->u.vmx.exit_reason == EXIT_REASON_EPT_MISCONFIG) {
420 vm_get_register(ctx, *pvcpu,
421 VMCS_IDENT(VMCS_GUEST_PHYSICAL_ADDRESS),
423 vm_get_gpa_pmap(ctx, ept_misconfig_gpa, ept_misconfig_pte,
424 &ept_misconfig_ptenum);
425 fprintf(stderr, "\tEPT misconfiguration:\n");
426 fprintf(stderr, "\t\tGPA: %#lx\n", ept_misconfig_gpa);
427 fprintf(stderr, "\t\tPTE(%d): %#lx %#lx %#lx %#lx\n",
428 ept_misconfig_ptenum, ept_misconfig_pte[0],
429 ept_misconfig_pte[1], ept_misconfig_pte[2],
430 ept_misconfig_pte[3]);
432 #endif /* DEBUG_EPT_MISCONFIG */
433 return (VMEXIT_ABORT);
437 vmexit_svm(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
440 fprintf(stderr, "vm exit[%d]\n", *pvcpu);
441 fprintf(stderr, "\treason\t\tSVM\n");
442 fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
443 fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
444 fprintf(stderr, "\texitcode\t%#lx\n", vmexit->u.svm.exitcode);
445 fprintf(stderr, "\texitinfo1\t%#lx\n", vmexit->u.svm.exitinfo1);
446 fprintf(stderr, "\texitinfo2\t%#lx\n", vmexit->u.svm.exitinfo2);
447 return (VMEXIT_ABORT);
451 vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
454 assert(vmexit->inst_length == 0);
456 stats.vmexit_bogus++;
458 return (VMEXIT_CONTINUE);
462 vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
468 * Just continue execution with the next instruction. We use
469 * the HLT VM exit as a way to be friendly with the host
472 return (VMEXIT_CONTINUE);
476 vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
479 stats.vmexit_pause++;
481 return (VMEXIT_CONTINUE);
485 vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
488 assert(vmexit->inst_length == 0);
490 stats.vmexit_mtrap++;
492 return (VMEXIT_CONTINUE);
496 vmexit_inst_emul(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
499 stats.vmexit_inst_emul++;
501 err = emulate_mem(ctx, *pvcpu, vmexit->u.inst_emul.gpa,
502 &vmexit->u.inst_emul.vie, &vmexit->u.inst_emul.paging);
507 "Failed to emulate instruction at 0x%lx\n",
509 } else if (err == ESRCH) {
510 fprintf(stderr, "Unhandled memory access to 0x%lx\n",
511 vmexit->u.inst_emul.gpa);
514 return (VMEXIT_ABORT);
517 return (VMEXIT_CONTINUE);
520 static pthread_mutex_t resetcpu_mtx = PTHREAD_MUTEX_INITIALIZER;
521 static pthread_cond_t resetcpu_cond = PTHREAD_COND_INITIALIZER;
524 vmexit_suspend(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
526 enum vm_suspend_how how;
528 how = vmexit->u.suspended.how;
530 fbsdrun_deletecpu(ctx, *pvcpu);
533 pthread_mutex_lock(&resetcpu_mtx);
534 pthread_cond_signal(&resetcpu_cond);
535 pthread_mutex_unlock(&resetcpu_mtx);
539 pthread_mutex_lock(&resetcpu_mtx);
540 while (!CPU_EMPTY(&cpumask)) {
541 pthread_cond_wait(&resetcpu_cond, &resetcpu_mtx);
543 pthread_mutex_unlock(&resetcpu_mtx);
546 case VM_SUSPEND_RESET:
548 case VM_SUSPEND_POWEROFF:
550 case VM_SUSPEND_HALT:
552 case VM_SUSPEND_TRIPLEFAULT:
555 fprintf(stderr, "vmexit_suspend: invalid reason %d\n", how);
558 return (0); /* NOTREACHED */
561 static vmexit_handler_t handler[VM_EXITCODE_MAX] = {
562 [VM_EXITCODE_INOUT] = vmexit_inout,
563 [VM_EXITCODE_INOUT_STR] = vmexit_inout,
564 [VM_EXITCODE_VMX] = vmexit_vmx,
565 [VM_EXITCODE_SVM] = vmexit_svm,
566 [VM_EXITCODE_BOGUS] = vmexit_bogus,
567 [VM_EXITCODE_RDMSR] = vmexit_rdmsr,
568 [VM_EXITCODE_WRMSR] = vmexit_wrmsr,
569 [VM_EXITCODE_MTRAP] = vmexit_mtrap,
570 [VM_EXITCODE_INST_EMUL] = vmexit_inst_emul,
571 [VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap,
572 [VM_EXITCODE_SUSPENDED] = vmexit_suspend,
573 [VM_EXITCODE_TASK_SWITCH] = vmexit_task_switch,
577 vm_loop(struct vmctx *ctx, int vcpu, uint64_t startrip)
579 int error, rc, prevcpu;
580 enum vm_exitcode exitcode;
581 cpuset_t active_cpus;
583 if (vcpumap[vcpu] != NULL) {
584 error = pthread_setaffinity_np(pthread_self(),
585 sizeof(cpuset_t), vcpumap[vcpu]);
589 error = vm_active_cpus(ctx, &active_cpus);
590 assert(CPU_ISSET(vcpu, &active_cpus));
592 error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RIP, startrip);
596 error = vm_run(ctx, vcpu, &vmexit[vcpu]);
602 exitcode = vmexit[vcpu].exitcode;
603 if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) {
604 fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n",
609 rc = (*handler[exitcode])(ctx, &vmexit[vcpu], &vcpu);
612 case VMEXIT_CONTINUE:
620 fprintf(stderr, "vm_run error %d, errno %d\n", error, errno);
624 num_vcpus_allowed(struct vmctx *ctx)
628 error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp);
631 * The guest is allowed to spinup more than one processor only if the
632 * UNRESTRICTED_GUEST capability is available.
641 fbsdrun_set_capabilities(struct vmctx *ctx, int cpu)
645 if (fbsdrun_vmexit_on_hlt()) {
646 err = vm_get_capability(ctx, cpu, VM_CAP_HALT_EXIT, &tmp);
648 fprintf(stderr, "VM exit on HLT not supported\n");
651 vm_set_capability(ctx, cpu, VM_CAP_HALT_EXIT, 1);
653 handler[VM_EXITCODE_HLT] = vmexit_hlt;
656 if (fbsdrun_vmexit_on_pause()) {
658 * pause exit support required for this mode
660 err = vm_get_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, &tmp);
663 "SMP mux requested, no pause support\n");
666 vm_set_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, 1);
668 handler[VM_EXITCODE_PAUSE] = vmexit_pause;
672 err = vm_set_x2apic_state(ctx, cpu, X2APIC_ENABLED);
674 err = vm_set_x2apic_state(ctx, cpu, X2APIC_DISABLED);
677 fprintf(stderr, "Unable to set x2apic state (%d)\n", err);
681 vm_set_capability(ctx, cpu, VM_CAP_ENABLE_INVPCID, 1);
685 main(int argc, char *argv[])
687 int c, error, gdb_port, err, bvmcons;
688 int dump_guest_memory, max_vcpus, mptgen;
695 dump_guest_memory = 0;
696 progname = basename(argv[0]);
703 while ((c = getopt(argc, argv, "abehuwxACHIPWYp:g:c:s:m:l:U:")) != -1) {
715 if (pincpu_parse(optarg) != 0) {
716 errx(EX_USAGE, "invalid vcpu pinning "
717 "configuration '%s'", optarg);
721 guest_ncpus = atoi(optarg);
724 dump_guest_memory = 1;
727 gdb_port = atoi(optarg);
730 if (lpc_device_parse(optarg) != 0) {
731 errx(EX_USAGE, "invalid lpc device "
732 "configuration '%s'", optarg);
736 if (pci_parse_slot(optarg) != 0)
741 error = vm_parse_memsize(optarg, &memsize);
743 errx(EX_USAGE, "invalid memsize '%s'", optarg);
746 guest_vmexit_on_hlt = 1;
750 * The "-I" option was used to add an ioapic to the
753 * An ioapic is now provided unconditionally for each
754 * virtual machine and this option is now deprecated.
758 guest_vmexit_on_pause = 1;
767 guest_uuid_str = optarg;
795 ctx = vm_open(vmname);
801 max_vcpus = num_vcpus_allowed(ctx);
802 if (guest_ncpus > max_vcpus) {
803 fprintf(stderr, "%d vCPUs requested but only %d available\n",
804 guest_ncpus, max_vcpus);
808 fbsdrun_set_capabilities(ctx, BSP);
810 if (dump_guest_memory)
811 vm_set_memflags(ctx, VM_MEM_F_INCORE);
812 err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL);
814 fprintf(stderr, "Unable to setup memory (%d)\n", err);
820 fprintf(stderr, "init_msr error %d", error);
829 rtc_init(ctx, rtc_localtime);
833 * Exit if a device emulation finds an error in it's initilization
835 if (init_pci(ctx) != 0)
839 init_dbgport(gdb_port);
844 error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip);
848 * build the guest tables, MP etc.
851 error = mptable_build(ctx, guest_ncpus);
856 error = smbios_build(ctx);
860 error = acpi_build(ctx, guest_ncpus);
865 * Change the proc title to include the VM name.
867 setproctitle("%s", vmname);
872 fbsdrun_addcpu(ctx, BSP, BSP, rip);
875 * Head off to the main event dispatch loop