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/segments.h>
47 #include <pthread_np.h>
50 #include <machine/vmm.h>
57 #include "legacy_irq.h"
65 #include "spinup_ap.h"
68 #define GUEST_NIO_PORT 0x488 /* guest upcalls via i/o port */
70 #define VMEXIT_SWITCH 0 /* force vcpu switch in mux mode */
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 */
76 #define MB (1024UL * 1024)
77 #define GB (1024UL * MB)
79 typedef int (*vmexit_handler_t)(struct vmctx *, struct vm_exit *, int *vcpu);
85 static int pincpu = -1;
86 static int guest_vmexit_on_hlt, guest_vmexit_on_pause, disable_x2apic;
87 static int virtio_msix = 1;
95 static char *progname;
96 static const int BSP = 0;
100 static void vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip);
102 struct vm_exit vmexit[VM_MAXCPU];
105 uint64_t vmexit_bogus;
106 uint64_t vmexit_bogus_switch;
108 uint64_t vmexit_pause;
109 uint64_t vmexit_mtrap;
110 uint64_t vmexit_inst_emul;
111 uint64_t cpu_switch_rotate;
112 uint64_t cpu_switch_direct;
118 struct vmctx *mt_ctx;
120 } mt_vmm_info[VM_MAXCPU];
127 "Usage: %s [-aehAHIPW] [-g <gdb port>] [-s <pci>] [-S <pci>]\n"
128 " %*s [-c vcpus] [-p pincpu] [-m mem] [-l <lpc>] <vm>\n"
129 " -a: local apic is in XAPIC mode (default is X2APIC)\n"
130 " -A: create an ACPI table\n"
132 " -c: # cpus (default 1)\n"
133 " -p: pin vcpu 'n' to host cpu 'pincpu + n'\n"
134 " -H: vmexit from the guest on hlt\n"
135 " -P: vmexit from the guest on pause\n"
136 " -W: force virtio to use single-vector MSI\n"
137 " -e: exit on unhandled I/O access\n"
139 " -s: <slot,driver,configinfo> PCI slot config\n"
140 " -S: <slot,driver,configinfo> legacy PCI slot config\n"
141 " -l: LPC device configuration\n"
142 " -m: memory size in MB\n",
143 progname, (int)strlen(progname), "");
149 paddr_guest2host(struct vmctx *ctx, uintptr_t gaddr, size_t len)
152 return (vm_map_gpa(ctx, gaddr, len));
156 fbsdrun_disable_x2apic(void)
159 return (disable_x2apic);
163 fbsdrun_vmexit_on_pause(void)
166 return (guest_vmexit_on_pause);
170 fbsdrun_vmexit_on_hlt(void)
173 return (guest_vmexit_on_hlt);
177 fbsdrun_virtio_msix(void)
180 return (virtio_msix);
184 fbsdrun_start_thread(void *param)
186 char tname[MAXCOMLEN + 1];
187 struct mt_vmm_info *mtp;
193 snprintf(tname, sizeof(tname), "vcpu %d", vcpu);
194 pthread_set_name_np(mtp->mt_thr, tname);
196 vm_loop(mtp->mt_ctx, vcpu, vmexit[vcpu].rip);
204 fbsdrun_addcpu(struct vmctx *ctx, int vcpu, uint64_t rip)
208 if (cpumask & (1 << vcpu)) {
209 fprintf(stderr, "addcpu: attempting to add existing cpu %d\n",
214 cpumask |= 1 << vcpu;
218 * Set up the vmexit struct to allow execution to start
221 vmexit[vcpu].rip = rip;
222 vmexit[vcpu].inst_length = 0;
224 mt_vmm_info[vcpu].mt_ctx = ctx;
225 mt_vmm_info[vcpu].mt_vcpu = vcpu;
227 error = pthread_create(&mt_vmm_info[vcpu].mt_thr, NULL,
228 fbsdrun_start_thread, &mt_vmm_info[vcpu]);
233 vmexit_catch_reset(void)
236 return (VMEXIT_RESET);
240 vmexit_catch_inout(void)
242 return (VMEXIT_ABORT);
246 vmexit_handle_notify(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu,
251 * put guest-driven debug here
254 return (VMEXIT_CONTINUE);
258 vmexit_inout(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
261 int bytes, port, in, out;
267 port = vme->u.inout.port;
268 bytes = vme->u.inout.bytes;
269 eax = vme->u.inout.eax;
270 in = vme->u.inout.in;
273 /* We don't deal with these */
274 if (vme->u.inout.string || vme->u.inout.rep)
275 return (VMEXIT_ABORT);
277 /* Special case of guest reset */
278 if (out && port == 0x64 && (uint8_t)eax == 0xFE)
279 return (vmexit_catch_reset());
281 /* Extra-special case of host notifications */
282 if (out && port == GUEST_NIO_PORT)
283 return (vmexit_handle_notify(ctx, vme, pvcpu, eax));
285 error = emulate_inout(ctx, vcpu, in, port, bytes, &eax, strictio);
286 if (error == 0 && in)
287 error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RAX, eax);
290 return (VMEXIT_CONTINUE);
292 fprintf(stderr, "Unhandled %s%c 0x%04x\n",
294 bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'), port);
295 return (vmexit_catch_inout());
300 vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
302 fprintf(stderr, "vm exit rdmsr 0x%x, cpu %d\n", vme->u.msr.code,
304 return (VMEXIT_ABORT);
308 vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
311 int retval = VMEXIT_CONTINUE;
313 newcpu = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code,vme->u.msr.wval);
319 vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
322 int retval = VMEXIT_CONTINUE;
324 newcpu = spinup_ap(ctx, *pvcpu,
325 vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip);
331 vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
334 fprintf(stderr, "vm exit[%d]\n", *pvcpu);
335 fprintf(stderr, "\treason\t\tVMX\n");
336 fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
337 fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
338 fprintf(stderr, "\terror\t\t%d\n", vmexit->u.vmx.error);
339 fprintf(stderr, "\texit_reason\t%u\n", vmexit->u.vmx.exit_reason);
340 fprintf(stderr, "\tqualification\t0x%016lx\n",
341 vmexit->u.vmx.exit_qualification);
343 return (VMEXIT_ABORT);
347 vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
350 stats.vmexit_bogus++;
352 return (VMEXIT_RESTART);
356 vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
362 * Just continue execution with the next instruction. We use
363 * the HLT VM exit as a way to be friendly with the host
366 return (VMEXIT_CONTINUE);
370 vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
373 stats.vmexit_pause++;
375 return (VMEXIT_CONTINUE);
379 vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
382 stats.vmexit_mtrap++;
384 return (VMEXIT_RESTART);
388 vmexit_inst_emul(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
391 stats.vmexit_inst_emul++;
393 err = emulate_mem(ctx, *pvcpu, vmexit->u.inst_emul.gpa,
394 &vmexit->u.inst_emul.vie);
399 "Failed to emulate instruction at 0x%lx\n",
401 } else if (err == ESRCH) {
402 fprintf(stderr, "Unhandled memory access to 0x%lx\n",
403 vmexit->u.inst_emul.gpa);
406 return (VMEXIT_ABORT);
409 return (VMEXIT_CONTINUE);
412 static vmexit_handler_t handler[VM_EXITCODE_MAX] = {
413 [VM_EXITCODE_INOUT] = vmexit_inout,
414 [VM_EXITCODE_VMX] = vmexit_vmx,
415 [VM_EXITCODE_BOGUS] = vmexit_bogus,
416 [VM_EXITCODE_RDMSR] = vmexit_rdmsr,
417 [VM_EXITCODE_WRMSR] = vmexit_wrmsr,
418 [VM_EXITCODE_MTRAP] = vmexit_mtrap,
419 [VM_EXITCODE_INST_EMUL] = vmexit_inst_emul,
420 [VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap,
424 vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip)
427 int error, rc, prevcpu;
428 enum vm_exitcode exitcode;
432 CPU_SET(pincpu + vcpu, &mask);
433 error = pthread_setaffinity_np(pthread_self(),
434 sizeof(mask), &mask);
439 error = vm_run(ctx, vcpu, rip, &vmexit[vcpu]);
442 * It is possible that 'vmmctl' or some other process
443 * has transitioned the vcpu to CANNOT_RUN state right
444 * before we tried to transition it to RUNNING.
446 * This is expected to be temporary so just retry.
456 exitcode = vmexit[vcpu].exitcode;
457 if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) {
458 fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n",
463 rc = (*handler[exitcode])(ctx, &vmexit[vcpu], &vcpu);
466 case VMEXIT_CONTINUE:
467 rip = vmexit[vcpu].rip + vmexit[vcpu].inst_length;
470 rip = vmexit[vcpu].rip;
478 fprintf(stderr, "vm_run error %d, errno %d\n", error, errno);
482 num_vcpus_allowed(struct vmctx *ctx)
486 error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp);
489 * The guest is allowed to spinup more than one processor only if the
490 * UNRESTRICTED_GUEST capability is available.
499 fbsdrun_set_capabilities(struct vmctx *ctx, int cpu)
503 if (fbsdrun_vmexit_on_hlt()) {
504 err = vm_get_capability(ctx, cpu, VM_CAP_HALT_EXIT, &tmp);
506 fprintf(stderr, "VM exit on HLT not supported\n");
509 vm_set_capability(ctx, cpu, VM_CAP_HALT_EXIT, 1);
511 handler[VM_EXITCODE_HLT] = vmexit_hlt;
514 if (fbsdrun_vmexit_on_pause()) {
516 * pause exit support required for this mode
518 err = vm_get_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, &tmp);
521 "SMP mux requested, no pause support\n");
524 vm_set_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, 1);
526 handler[VM_EXITCODE_PAUSE] = vmexit_pause;
529 if (fbsdrun_disable_x2apic())
530 err = vm_set_x2apic_state(ctx, cpu, X2APIC_DISABLED);
532 err = vm_set_x2apic_state(ctx, cpu, X2APIC_ENABLED);
535 fprintf(stderr, "Unable to set x2apic state (%d)\n", err);
539 vm_set_capability(ctx, cpu, VM_CAP_ENABLE_INVPCID, 1);
543 main(int argc, char *argv[])
545 int c, error, gdb_port, err, bvmcons;
552 progname = basename(argv[0]);
557 while ((c = getopt(argc, argv, "abehAHIPWp:g:c:s:S:m:l:")) != -1) {
569 pincpu = atoi(optarg);
572 guest_ncpus = atoi(optarg);
575 gdb_port = atoi(optarg);
578 if (lpc_device_parse(optarg) != 0) {
579 errx(EX_USAGE, "invalid lpc device "
580 "configuration '%s'", optarg);
584 if (pci_parse_slot(optarg, 0) != 0)
589 if (pci_parse_slot(optarg, 1) != 0)
594 error = vm_parse_memsize(optarg, &memsize);
596 errx(EX_USAGE, "invalid memsize '%s'", optarg);
599 guest_vmexit_on_hlt = 1;
603 * The "-I" option was used to add an ioapic to the
606 * An ioapic is now provided unconditionally for each
607 * virtual machine and this option is now deprecated.
611 guest_vmexit_on_pause = 1;
633 ctx = vm_open(vmname);
639 max_vcpus = num_vcpus_allowed(ctx);
640 if (guest_ncpus > max_vcpus) {
641 fprintf(stderr, "%d vCPUs requested but only %d available\n",
642 guest_ncpus, max_vcpus);
646 fbsdrun_set_capabilities(ctx, BSP);
648 err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL);
650 fprintf(stderr, "Unable to setup memory (%d)\n", err);
661 * Exit if a device emulation finds an error in it's initilization
663 if (init_pci(ctx) != 0)
669 init_dbgport(gdb_port);
674 error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip);
678 * build the guest tables, MP etc.
680 mptable_build(ctx, guest_ncpus);
683 error = acpi_build(ctx, guest_ncpus);
688 * Change the proc title to include the VM name.
690 setproctitle("%s", vmname);
695 fbsdrun_addcpu(ctx, BSP, rip);
698 * Head off to the main event dispatch loop