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_SWITCH 0 /* force vcpu switch in mux mode */
72 #define VMEXIT_CONTINUE 1 /* continue from next instruction */
73 #define VMEXIT_RESTART 2 /* restart current instruction */
74 #define VMEXIT_ABORT 3 /* abort the vm run loop */
75 #define VMEXIT_RESET 4 /* guest machine has reset */
76 #define VMEXIT_POWEROFF 5 /* guest machine has powered off */
78 #define MB (1024UL * 1024)
79 #define GB (1024UL * MB)
81 typedef int (*vmexit_handler_t)(struct vmctx *, struct vm_exit *, int *vcpu);
88 static int pincpu = -1;
89 static int guest_vmexit_on_hlt, guest_vmexit_on_pause;
90 static int virtio_msix = 1;
91 static int x2apic_mode = 0; /* default is xAPIC */
94 static int strictmsr = 1;
98 static char *progname;
99 static const int BSP = 0;
101 static cpuset_t cpumask;
103 static void vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip);
105 struct vm_exit vmexit[VM_MAXCPU];
108 uint64_t vmexit_bogus;
109 uint64_t vmexit_bogus_switch;
111 uint64_t vmexit_pause;
112 uint64_t vmexit_mtrap;
113 uint64_t vmexit_inst_emul;
114 uint64_t cpu_switch_rotate;
115 uint64_t cpu_switch_direct;
121 struct vmctx *mt_ctx;
123 } mt_vmm_info[VM_MAXCPU];
130 "Usage: %s [-aehwAHIPW] [-g <gdb port>] [-s <pci>]\n"
131 " %*s [-c vcpus] [-p pincpu] [-m mem] [-l <lpc>] <vm>\n"
132 " -a: local apic is in xAPIC mode (deprecated)\n"
133 " -A: create an ACPI table\n"
135 " -c: # cpus (default 1)\n"
136 " -p: pin vcpu 'n' to host cpu 'pincpu + n'\n"
137 " -H: vmexit from the guest on hlt\n"
138 " -P: vmexit from the guest on pause\n"
139 " -W: force virtio to use single-vector MSI\n"
140 " -e: exit on unhandled I/O access\n"
142 " -s: <slot,driver,configinfo> PCI slot config\n"
143 " -l: LPC device configuration\n"
144 " -m: memory size in MB\n"
145 " -w: ignore unimplemented MSRs\n"
146 " -x: local apic is in x2APIC mode\n"
148 progname, (int)strlen(progname), "");
154 paddr_guest2host(struct vmctx *ctx, uintptr_t gaddr, size_t len)
157 return (vm_map_gpa(ctx, gaddr, len));
161 fbsdrun_vmexit_on_pause(void)
164 return (guest_vmexit_on_pause);
168 fbsdrun_vmexit_on_hlt(void)
171 return (guest_vmexit_on_hlt);
175 fbsdrun_virtio_msix(void)
178 return (virtio_msix);
182 fbsdrun_start_thread(void *param)
184 char tname[MAXCOMLEN + 1];
185 struct mt_vmm_info *mtp;
191 snprintf(tname, sizeof(tname), "vcpu %d", vcpu);
192 pthread_set_name_np(mtp->mt_thr, tname);
194 vm_loop(mtp->mt_ctx, vcpu, vmexit[vcpu].rip);
202 fbsdrun_addcpu(struct vmctx *ctx, int fromcpu, int newcpu, uint64_t rip)
206 assert(fromcpu == BSP);
208 CPU_SET_ATOMIC(newcpu, &cpumask);
211 * Set up the vmexit struct to allow execution to start
214 vmexit[newcpu].rip = rip;
215 vmexit[newcpu].inst_length = 0;
217 mt_vmm_info[newcpu].mt_ctx = ctx;
218 mt_vmm_info[newcpu].mt_vcpu = newcpu;
220 error = pthread_create(&mt_vmm_info[newcpu].mt_thr, NULL,
221 fbsdrun_start_thread, &mt_vmm_info[newcpu]);
226 fbsdrun_deletecpu(struct vmctx *ctx, int vcpu)
229 if (!CPU_ISSET(vcpu, &cpumask)) {
230 fprintf(stderr, "addcpu: attempting to delete unknown cpu %d\n",
235 CPU_CLR_ATOMIC(vcpu, &cpumask);
236 return (CPU_EMPTY(&cpumask));
240 vmexit_catch_reset(void)
243 return (VMEXIT_RESET);
247 vmexit_catch_inout(void)
249 return (VMEXIT_ABORT);
253 vmexit_handle_notify(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu,
258 * put guest-driven debug here
261 return (VMEXIT_CONTINUE);
265 vmexit_inout(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
268 int bytes, port, in, out;
274 port = vme->u.inout.port;
275 bytes = vme->u.inout.bytes;
276 eax = vme->u.inout.eax;
277 in = vme->u.inout.in;
280 /* We don't deal with these */
281 if (vme->u.inout.string || vme->u.inout.rep)
282 return (VMEXIT_ABORT);
284 /* Special case of guest reset */
285 if (out && port == 0x64 && (uint8_t)eax == 0xFE)
286 return (vmexit_catch_reset());
288 /* Extra-special case of host notifications */
289 if (out && port == GUEST_NIO_PORT)
290 return (vmexit_handle_notify(ctx, vme, pvcpu, eax));
292 error = emulate_inout(ctx, vcpu, in, port, bytes, &eax, strictio);
293 if (error == INOUT_OK && in)
294 error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RAX, eax);
298 return (VMEXIT_CONTINUE);
300 return (VMEXIT_RESET);
302 return (VMEXIT_POWEROFF);
304 fprintf(stderr, "Unhandled %s%c 0x%04x\n",
306 bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'), port);
307 return (vmexit_catch_inout());
312 vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
319 error = emulate_rdmsr(ctx, *pvcpu, vme->u.msr.code, &val);
321 fprintf(stderr, "rdmsr to register %#x on vcpu %d\n",
322 vme->u.msr.code, *pvcpu);
324 error = vm_inject_exception2(ctx, *pvcpu, IDT_GP, 0);
326 return (VMEXIT_RESTART);
331 error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RAX, eax);
335 error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RDX, edx);
338 return (VMEXIT_CONTINUE);
342 vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
346 error = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code, vme->u.msr.wval);
348 fprintf(stderr, "wrmsr to register %#x(%#lx) on vcpu %d\n",
349 vme->u.msr.code, vme->u.msr.wval, *pvcpu);
351 error = vm_inject_exception2(ctx, *pvcpu, IDT_GP, 0);
353 return (VMEXIT_RESTART);
356 return (VMEXIT_CONTINUE);
360 vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
363 int retval = VMEXIT_CONTINUE;
365 newcpu = spinup_ap(ctx, *pvcpu,
366 vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip);
372 vmexit_spindown_cpu(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
376 lastcpu = fbsdrun_deletecpu(ctx, *pvcpu);
379 return (vmexit_catch_reset());
383 vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
386 fprintf(stderr, "vm exit[%d]\n", *pvcpu);
387 fprintf(stderr, "\treason\t\tVMX\n");
388 fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
389 fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
390 fprintf(stderr, "\tstatus\t\t%d\n", vmexit->u.vmx.status);
391 fprintf(stderr, "\texit_reason\t%u\n", vmexit->u.vmx.exit_reason);
392 fprintf(stderr, "\tqualification\t0x%016lx\n",
393 vmexit->u.vmx.exit_qualification);
394 fprintf(stderr, "\tinst_type\t\t%d\n", vmexit->u.vmx.inst_type);
395 fprintf(stderr, "\tinst_error\t\t%d\n", vmexit->u.vmx.inst_error);
397 return (VMEXIT_ABORT);
401 vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
404 stats.vmexit_bogus++;
406 return (VMEXIT_RESTART);
410 vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
416 * Just continue execution with the next instruction. We use
417 * the HLT VM exit as a way to be friendly with the host
420 return (VMEXIT_CONTINUE);
424 vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
427 stats.vmexit_pause++;
429 return (VMEXIT_CONTINUE);
433 vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
436 stats.vmexit_mtrap++;
438 return (VMEXIT_RESTART);
442 vmexit_inst_emul(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
445 stats.vmexit_inst_emul++;
447 err = emulate_mem(ctx, *pvcpu, vmexit->u.inst_emul.gpa,
448 &vmexit->u.inst_emul.vie);
453 "Failed to emulate instruction at 0x%lx\n",
455 } else if (err == ESRCH) {
456 fprintf(stderr, "Unhandled memory access to 0x%lx\n",
457 vmexit->u.inst_emul.gpa);
460 return (VMEXIT_ABORT);
463 return (VMEXIT_CONTINUE);
466 static vmexit_handler_t handler[VM_EXITCODE_MAX] = {
467 [VM_EXITCODE_INOUT] = vmexit_inout,
468 [VM_EXITCODE_VMX] = vmexit_vmx,
469 [VM_EXITCODE_BOGUS] = vmexit_bogus,
470 [VM_EXITCODE_RDMSR] = vmexit_rdmsr,
471 [VM_EXITCODE_WRMSR] = vmexit_wrmsr,
472 [VM_EXITCODE_MTRAP] = vmexit_mtrap,
473 [VM_EXITCODE_INST_EMUL] = vmexit_inst_emul,
474 [VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap,
475 [VM_EXITCODE_SPINDOWN_CPU] = vmexit_spindown_cpu,
479 vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip)
482 int error, rc, prevcpu;
483 enum vm_exitcode exitcode;
487 CPU_SET(pincpu + vcpu, &mask);
488 error = pthread_setaffinity_np(pthread_self(),
489 sizeof(mask), &mask);
494 error = vm_run(ctx, vcpu, rip, &vmexit[vcpu]);
500 exitcode = vmexit[vcpu].exitcode;
501 if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) {
502 fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n",
507 rc = (*handler[exitcode])(ctx, &vmexit[vcpu], &vcpu);
510 case VMEXIT_CONTINUE:
511 rip = vmexit[vcpu].rip + vmexit[vcpu].inst_length;
514 rip = vmexit[vcpu].rip;
522 fprintf(stderr, "vm_run error %d, errno %d\n", error, errno);
526 num_vcpus_allowed(struct vmctx *ctx)
530 error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp);
533 * The guest is allowed to spinup more than one processor only if the
534 * UNRESTRICTED_GUEST capability is available.
543 fbsdrun_set_capabilities(struct vmctx *ctx, int cpu)
547 if (fbsdrun_vmexit_on_hlt()) {
548 err = vm_get_capability(ctx, cpu, VM_CAP_HALT_EXIT, &tmp);
550 fprintf(stderr, "VM exit on HLT not supported\n");
553 vm_set_capability(ctx, cpu, VM_CAP_HALT_EXIT, 1);
555 handler[VM_EXITCODE_HLT] = vmexit_hlt;
558 if (fbsdrun_vmexit_on_pause()) {
560 * pause exit support required for this mode
562 err = vm_get_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, &tmp);
565 "SMP mux requested, no pause support\n");
568 vm_set_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, 1);
570 handler[VM_EXITCODE_PAUSE] = vmexit_pause;
574 err = vm_set_x2apic_state(ctx, cpu, X2APIC_ENABLED);
576 err = vm_set_x2apic_state(ctx, cpu, X2APIC_DISABLED);
579 fprintf(stderr, "Unable to set x2apic state (%d)\n", err);
583 vm_set_capability(ctx, cpu, VM_CAP_ENABLE_INVPCID, 1);
587 main(int argc, char *argv[])
589 int c, error, gdb_port, err, bvmcons;
596 progname = basename(argv[0]);
601 while ((c = getopt(argc, argv, "abehwxAHIPWp:g:c:s:m:l:U:")) != -1) {
613 pincpu = atoi(optarg);
616 guest_ncpus = atoi(optarg);
619 gdb_port = atoi(optarg);
622 if (lpc_device_parse(optarg) != 0) {
623 errx(EX_USAGE, "invalid lpc device "
624 "configuration '%s'", optarg);
628 if (pci_parse_slot(optarg) != 0)
633 error = vm_parse_memsize(optarg, &memsize);
635 errx(EX_USAGE, "invalid memsize '%s'", optarg);
638 guest_vmexit_on_hlt = 1;
642 * The "-I" option was used to add an ioapic to the
645 * An ioapic is now provided unconditionally for each
646 * virtual machine and this option is now deprecated.
650 guest_vmexit_on_pause = 1;
656 guest_uuid_str = optarg;
681 ctx = vm_open(vmname);
687 max_vcpus = num_vcpus_allowed(ctx);
688 if (guest_ncpus > max_vcpus) {
689 fprintf(stderr, "%d vCPUs requested but only %d available\n",
690 guest_ncpus, max_vcpus);
694 fbsdrun_set_capabilities(ctx, BSP);
696 err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL);
698 fprintf(stderr, "Unable to setup memory (%d)\n", err);
709 * Exit if a device emulation finds an error in it's initilization
711 if (init_pci(ctx) != 0)
715 init_dbgport(gdb_port);
720 error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip);
724 * build the guest tables, MP etc.
726 mptable_build(ctx, guest_ncpus);
728 error = smbios_build(ctx);
732 error = acpi_build(ctx, guest_ncpus);
737 * Change the proc title to include the VM name.
739 setproctitle("%s", vmname);
744 fbsdrun_addcpu(ctx, BSP, BSP, rip);
747 * Head off to the main event dispatch loop