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>
64 #include "smbiostbl.h"
66 #include "spinup_ap.h"
69 #define GUEST_NIO_PORT 0x488 /* guest upcalls via i/o port */
71 #define VMEXIT_CONTINUE 1 /* continue from next instruction */
72 #define VMEXIT_RESTART 2 /* restart current instruction */
73 #define VMEXIT_ABORT 3 /* abort the vm run loop */
74 #define VMEXIT_RESET 4 /* guest machine has reset */
75 #define VMEXIT_POWEROFF 5 /* guest machine has powered off */
77 #define MB (1024UL * 1024)
78 #define GB (1024UL * MB)
80 typedef int (*vmexit_handler_t)(struct vmctx *, struct vm_exit *, int *vcpu);
87 static int guest_vmexit_on_hlt, guest_vmexit_on_pause;
88 static int virtio_msix = 1;
89 static int x2apic_mode = 0; /* default is xAPIC */
92 static int strictmsr = 1;
96 static char *progname;
97 static const int BSP = 0;
99 static cpuset_t cpumask;
101 static void vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip);
103 struct vm_exit vmexit[VM_MAXCPU];
106 uint64_t vmexit_bogus;
107 uint64_t vmexit_bogus_switch;
109 uint64_t vmexit_pause;
110 uint64_t vmexit_mtrap;
111 uint64_t vmexit_inst_emul;
112 uint64_t cpu_switch_rotate;
113 uint64_t cpu_switch_direct;
120 struct vmctx *mt_ctx;
122 } mt_vmm_info[VM_MAXCPU];
124 static cpuset_t *vcpumap[VM_MAXCPU] = { NULL };
131 "Usage: %s [-aehwAHIPW] [-g <gdb port>] [-s <pci>] [-c vcpus]\n"
132 " %*s [-p vcpu:hostcpu] [-m mem] [-l <lpc>] <vm>\n"
133 " -a: local apic is in xAPIC mode (deprecated)\n"
134 " -A: create an ACPI table\n"
136 " -c: # cpus (default 1)\n"
137 " -C: include guest memory in core file\n"
138 " -p: pin 'vcpu' to 'hostcpu'\n"
139 " -H: vmexit from the guest on hlt\n"
140 " -P: vmexit from the guest on pause\n"
141 " -W: force virtio to use single-vector MSI\n"
142 " -e: exit on unhandled I/O access\n"
144 " -s: <slot,driver,configinfo> PCI slot config\n"
145 " -l: LPC device configuration\n"
146 " -m: memory size in MB\n"
147 " -w: ignore unimplemented MSRs\n"
148 " -x: local apic is in x2APIC mode\n"
149 " -Y: disable MPtable generation\n"
151 progname, (int)strlen(progname), "");
157 pincpu_parse(const char *opt)
161 if (sscanf(opt, "%d:%d", &vcpu, &pcpu) != 2) {
162 fprintf(stderr, "invalid format: %s\n", opt);
166 if (vcpu < 0 || vcpu >= VM_MAXCPU) {
167 fprintf(stderr, "vcpu '%d' outside valid range from 0 to %d\n",
168 vcpu, VM_MAXCPU - 1);
172 if (pcpu < 0 || pcpu >= CPU_SETSIZE) {
173 fprintf(stderr, "hostcpu '%d' outside valid range from "
174 "0 to %d\n", pcpu, CPU_SETSIZE - 1);
178 if (vcpumap[vcpu] == NULL) {
179 if ((vcpumap[vcpu] = malloc(sizeof(cpuset_t))) == NULL) {
183 CPU_ZERO(vcpumap[vcpu]);
185 CPU_SET(pcpu, vcpumap[vcpu]);
190 paddr_guest2host(struct vmctx *ctx, uintptr_t gaddr, size_t len)
193 return (vm_map_gpa(ctx, gaddr, len));
197 fbsdrun_vmexit_on_pause(void)
200 return (guest_vmexit_on_pause);
204 fbsdrun_vmexit_on_hlt(void)
207 return (guest_vmexit_on_hlt);
211 fbsdrun_virtio_msix(void)
214 return (virtio_msix);
218 fbsdrun_start_thread(void *param)
220 char tname[MAXCOMLEN + 1];
221 struct mt_vmm_info *mtp;
227 snprintf(tname, sizeof(tname), "vcpu %d", vcpu);
228 pthread_set_name_np(mtp->mt_thr, tname);
230 vm_loop(mtp->mt_ctx, vcpu, vmexit[vcpu].rip);
238 fbsdrun_addcpu(struct vmctx *ctx, int fromcpu, int newcpu, uint64_t rip)
242 assert(fromcpu == BSP);
244 CPU_SET_ATOMIC(newcpu, &cpumask);
247 * Set up the vmexit struct to allow execution to start
250 vmexit[newcpu].rip = rip;
251 vmexit[newcpu].inst_length = 0;
253 mt_vmm_info[newcpu].mt_ctx = ctx;
254 mt_vmm_info[newcpu].mt_vcpu = newcpu;
256 error = pthread_create(&mt_vmm_info[newcpu].mt_thr, NULL,
257 fbsdrun_start_thread, &mt_vmm_info[newcpu]);
262 fbsdrun_deletecpu(struct vmctx *ctx, int vcpu)
265 if (!CPU_ISSET(vcpu, &cpumask)) {
266 fprintf(stderr, "Attempting to delete unknown cpu %d\n", vcpu);
270 CPU_CLR_ATOMIC(vcpu, &cpumask);
271 return (CPU_EMPTY(&cpumask));
275 vmexit_handle_notify(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu,
280 * put guest-driven debug here
283 return (VMEXIT_CONTINUE);
287 vmexit_inout(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
290 int bytes, port, in, out;
296 port = vme->u.inout.port;
297 bytes = vme->u.inout.bytes;
298 eax = vme->u.inout.eax;
299 in = vme->u.inout.in;
302 /* We don't deal with these */
303 if (vme->u.inout.string || vme->u.inout.rep)
304 return (VMEXIT_ABORT);
306 /* Extra-special case of host notifications */
307 if (out && port == GUEST_NIO_PORT)
308 return (vmexit_handle_notify(ctx, vme, pvcpu, eax));
310 error = emulate_inout(ctx, vcpu, in, port, bytes, &eax, strictio);
311 if (error == INOUT_OK && in)
312 error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RAX, eax);
316 return (VMEXIT_CONTINUE);
319 return (VMEXIT_RESET);
322 return (VMEXIT_POWEROFF);
324 fprintf(stderr, "Unhandled %s%c 0x%04x\n",
326 bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'), port);
327 return (VMEXIT_ABORT);
332 vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
339 error = emulate_rdmsr(ctx, *pvcpu, vme->u.msr.code, &val);
341 fprintf(stderr, "rdmsr to register %#x on vcpu %d\n",
342 vme->u.msr.code, *pvcpu);
344 error = vm_inject_exception2(ctx, *pvcpu, IDT_GP, 0);
346 return (VMEXIT_RESTART);
351 error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RAX, eax);
355 error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RDX, edx);
358 return (VMEXIT_CONTINUE);
362 vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
366 error = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code, vme->u.msr.wval);
368 fprintf(stderr, "wrmsr to register %#x(%#lx) on vcpu %d\n",
369 vme->u.msr.code, vme->u.msr.wval, *pvcpu);
371 error = vm_inject_exception2(ctx, *pvcpu, IDT_GP, 0);
373 return (VMEXIT_RESTART);
376 return (VMEXIT_CONTINUE);
380 vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
383 int retval = VMEXIT_CONTINUE;
385 newcpu = spinup_ap(ctx, *pvcpu,
386 vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip);
392 vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
395 fprintf(stderr, "vm exit[%d]\n", *pvcpu);
396 fprintf(stderr, "\treason\t\tVMX\n");
397 fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
398 fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
399 fprintf(stderr, "\tstatus\t\t%d\n", vmexit->u.vmx.status);
400 fprintf(stderr, "\texit_reason\t%u\n", vmexit->u.vmx.exit_reason);
401 fprintf(stderr, "\tqualification\t0x%016lx\n",
402 vmexit->u.vmx.exit_qualification);
403 fprintf(stderr, "\tinst_type\t\t%d\n", vmexit->u.vmx.inst_type);
404 fprintf(stderr, "\tinst_error\t\t%d\n", vmexit->u.vmx.inst_error);
406 return (VMEXIT_ABORT);
410 vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
413 stats.vmexit_bogus++;
415 return (VMEXIT_RESTART);
419 vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
425 * Just continue execution with the next instruction. We use
426 * the HLT VM exit as a way to be friendly with the host
429 return (VMEXIT_CONTINUE);
433 vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
436 stats.vmexit_pause++;
438 return (VMEXIT_CONTINUE);
442 vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
445 stats.vmexit_mtrap++;
447 return (VMEXIT_RESTART);
451 vmexit_inst_emul(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
454 stats.vmexit_inst_emul++;
456 err = emulate_mem(ctx, *pvcpu, vmexit->u.inst_emul.gpa,
457 &vmexit->u.inst_emul.vie);
462 "Failed to emulate instruction at 0x%lx\n",
464 } else if (err == ESRCH) {
465 fprintf(stderr, "Unhandled memory access to 0x%lx\n",
466 vmexit->u.inst_emul.gpa);
469 return (VMEXIT_ABORT);
472 return (VMEXIT_CONTINUE);
475 static pthread_mutex_t resetcpu_mtx = PTHREAD_MUTEX_INITIALIZER;
476 static pthread_cond_t resetcpu_cond = PTHREAD_COND_INITIALIZER;
479 vmexit_suspend(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
481 enum vm_suspend_how how;
483 how = vmexit->u.suspended.how;
485 fbsdrun_deletecpu(ctx, *pvcpu);
488 pthread_mutex_lock(&resetcpu_mtx);
489 pthread_cond_signal(&resetcpu_cond);
490 pthread_mutex_unlock(&resetcpu_mtx);
494 pthread_mutex_lock(&resetcpu_mtx);
495 while (!CPU_EMPTY(&cpumask)) {
496 pthread_cond_wait(&resetcpu_cond, &resetcpu_mtx);
498 pthread_mutex_unlock(&resetcpu_mtx);
501 case VM_SUSPEND_RESET:
503 case VM_SUSPEND_POWEROFF:
505 case VM_SUSPEND_HALT:
508 fprintf(stderr, "vmexit_suspend: invalid reason %d\n", how);
511 return (0); /* NOTREACHED */
514 static vmexit_handler_t handler[VM_EXITCODE_MAX] = {
515 [VM_EXITCODE_INOUT] = vmexit_inout,
516 [VM_EXITCODE_VMX] = vmexit_vmx,
517 [VM_EXITCODE_BOGUS] = vmexit_bogus,
518 [VM_EXITCODE_RDMSR] = vmexit_rdmsr,
519 [VM_EXITCODE_WRMSR] = vmexit_wrmsr,
520 [VM_EXITCODE_MTRAP] = vmexit_mtrap,
521 [VM_EXITCODE_INST_EMUL] = vmexit_inst_emul,
522 [VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap,
523 [VM_EXITCODE_SUSPENDED] = vmexit_suspend
527 vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip)
529 int error, rc, prevcpu;
530 enum vm_exitcode exitcode;
531 enum vm_suspend_how how;
533 if (vcpumap[vcpu] != NULL) {
534 error = pthread_setaffinity_np(pthread_self(),
535 sizeof(cpuset_t), vcpumap[vcpu]);
540 error = vm_run(ctx, vcpu, rip, &vmexit[vcpu]);
546 exitcode = vmexit[vcpu].exitcode;
547 if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) {
548 fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n",
553 rc = (*handler[exitcode])(ctx, &vmexit[vcpu], &vcpu);
556 case VMEXIT_CONTINUE:
557 rip = vmexit[vcpu].rip + vmexit[vcpu].inst_length;
560 rip = vmexit[vcpu].rip;
563 case VMEXIT_POWEROFF:
564 if (rc == VMEXIT_RESET)
565 how = VM_SUSPEND_RESET;
567 how = VM_SUSPEND_POWEROFF;
568 error = vm_suspend(ctx, how);
569 assert(error == 0 || errno == EALREADY);
570 rip = vmexit[vcpu].rip + vmexit[vcpu].inst_length;
578 fprintf(stderr, "vm_run error %d, errno %d\n", error, errno);
582 num_vcpus_allowed(struct vmctx *ctx)
586 error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp);
589 * The guest is allowed to spinup more than one processor only if the
590 * UNRESTRICTED_GUEST capability is available.
599 fbsdrun_set_capabilities(struct vmctx *ctx, int cpu)
603 if (fbsdrun_vmexit_on_hlt()) {
604 err = vm_get_capability(ctx, cpu, VM_CAP_HALT_EXIT, &tmp);
606 fprintf(stderr, "VM exit on HLT not supported\n");
609 vm_set_capability(ctx, cpu, VM_CAP_HALT_EXIT, 1);
611 handler[VM_EXITCODE_HLT] = vmexit_hlt;
614 if (fbsdrun_vmexit_on_pause()) {
616 * pause exit support required for this mode
618 err = vm_get_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, &tmp);
621 "SMP mux requested, no pause support\n");
624 vm_set_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, 1);
626 handler[VM_EXITCODE_PAUSE] = vmexit_pause;
630 err = vm_set_x2apic_state(ctx, cpu, X2APIC_ENABLED);
632 err = vm_set_x2apic_state(ctx, cpu, X2APIC_DISABLED);
635 fprintf(stderr, "Unable to set x2apic state (%d)\n", err);
639 vm_set_capability(ctx, cpu, VM_CAP_ENABLE_INVPCID, 1);
643 main(int argc, char *argv[])
645 int c, error, gdb_port, err, bvmcons;
646 int dump_guest_memory, max_vcpus, mptgen;
652 dump_guest_memory = 0;
653 progname = basename(argv[0]);
659 while ((c = getopt(argc, argv, "abehwxACHIPWYp:g:c:s:m:l:U:")) != -1) {
671 if (pincpu_parse(optarg) != 0) {
672 errx(EX_USAGE, "invalid vcpu pinning "
673 "configuration '%s'", optarg);
677 guest_ncpus = atoi(optarg);
680 dump_guest_memory = 1;
683 gdb_port = atoi(optarg);
686 if (lpc_device_parse(optarg) != 0) {
687 errx(EX_USAGE, "invalid lpc device "
688 "configuration '%s'", optarg);
692 if (pci_parse_slot(optarg) != 0)
697 error = vm_parse_memsize(optarg, &memsize);
699 errx(EX_USAGE, "invalid memsize '%s'", optarg);
702 guest_vmexit_on_hlt = 1;
706 * The "-I" option was used to add an ioapic to the
709 * An ioapic is now provided unconditionally for each
710 * virtual machine and this option is now deprecated.
714 guest_vmexit_on_pause = 1;
720 guest_uuid_str = optarg;
748 ctx = vm_open(vmname);
754 max_vcpus = num_vcpus_allowed(ctx);
755 if (guest_ncpus > max_vcpus) {
756 fprintf(stderr, "%d vCPUs requested but only %d available\n",
757 guest_ncpus, max_vcpus);
761 fbsdrun_set_capabilities(ctx, BSP);
763 if (dump_guest_memory)
764 vm_set_memflags(ctx, VM_MEM_F_INCORE);
765 err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL);
767 fprintf(stderr, "Unable to setup memory (%d)\n", err);
778 * Exit if a device emulation finds an error in it's initilization
780 if (init_pci(ctx) != 0)
784 init_dbgport(gdb_port);
789 error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip);
793 * build the guest tables, MP etc.
796 error = mptable_build(ctx, guest_ncpus);
801 error = smbios_build(ctx);
805 error = acpi_build(ctx, guest_ncpus);
810 * Change the proc title to include the VM name.
812 setproctitle("%s", vmname);
817 fbsdrun_addcpu(ctx, BSP, BSP, rip);
820 * Head off to the main event dispatch loop