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>
46 #include <pthread_np.h>
49 #include <machine/vmm.h>
62 #include "spinup_ap.h"
65 #define GUEST_NIO_PORT 0x488 /* guest upcalls via i/o port */
67 #define VMEXIT_SWITCH 0 /* force vcpu switch in mux mode */
68 #define VMEXIT_CONTINUE 1 /* continue from next instruction */
69 #define VMEXIT_RESTART 2 /* restart current instruction */
70 #define VMEXIT_ABORT 3 /* abort the vm run loop */
71 #define VMEXIT_RESET 4 /* guest machine has reset */
73 #define MB (1024UL * 1024)
74 #define GB (1024UL * MB)
76 typedef int (*vmexit_handler_t)(struct vmctx *, struct vm_exit *, int *vcpu);
82 static int pincpu = -1;
83 static int guest_vmexit_on_hlt, guest_vmexit_on_pause, disable_x2apic;
84 static int virtio_msix = 1;
92 static char *progname;
93 static const int BSP = 0;
97 static void vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip);
99 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;
115 struct vmctx *mt_ctx;
117 } mt_vmm_info[VM_MAXCPU];
124 "Usage: %s [-aehAHIPW][-g <gdb port>][-s <pci>][-S <pci>]"
125 "[-c vcpus][-p pincpu][-m mem]"
127 " -a: local apic is in XAPIC mode (default is X2APIC)\n"
128 " -A: create an ACPI table\n"
130 " -c: # cpus (default 1)\n"
131 " -p: pin vcpu 'n' to host cpu 'pincpu + n'\n"
132 " -H: vmexit from the guest on hlt\n"
133 " -I: present an ioapic to the guest\n"
134 " -P: vmexit from the guest on pause\n"
135 " -W: force virtio to use single-vector MSI\n"
136 " -e: exit on unhandled i/o access\n"
138 " -s: <slot,driver,configinfo> PCI slot config\n"
139 " -S: <slot,driver,configinfo> legacy PCI slot config\n"
140 " -m: memory size in MB\n",
147 paddr_guest2host(struct vmctx *ctx, uintptr_t gaddr, size_t len)
150 return (vm_map_gpa(ctx, gaddr, len));
154 fbsdrun_disable_x2apic(void)
157 return (disable_x2apic);
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), "%s vcpu %d", vmname, 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 vcpu, uint64_t rip)
206 if (cpumask & (1 << vcpu)) {
207 fprintf(stderr, "addcpu: attempting to add existing cpu %d\n",
212 cpumask |= 1 << vcpu;
216 * Set up the vmexit struct to allow execution to start
219 vmexit[vcpu].rip = rip;
220 vmexit[vcpu].inst_length = 0;
222 mt_vmm_info[vcpu].mt_ctx = ctx;
223 mt_vmm_info[vcpu].mt_vcpu = vcpu;
225 error = pthread_create(&mt_vmm_info[vcpu].mt_thr, NULL,
226 fbsdrun_start_thread, &mt_vmm_info[vcpu]);
231 vmexit_catch_reset(void)
234 return (VMEXIT_RESET);
238 vmexit_catch_inout(void)
240 return (VMEXIT_ABORT);
244 vmexit_handle_notify(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu,
249 * put guest-driven debug here
252 return (VMEXIT_CONTINUE);
256 vmexit_inout(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
259 int bytes, port, in, out;
265 port = vme->u.inout.port;
266 bytes = vme->u.inout.bytes;
267 eax = vme->u.inout.eax;
268 in = vme->u.inout.in;
271 /* We don't deal with these */
272 if (vme->u.inout.string || vme->u.inout.rep)
273 return (VMEXIT_ABORT);
275 /* Special case of guest reset */
276 if (out && port == 0x64 && (uint8_t)eax == 0xFE)
277 return (vmexit_catch_reset());
279 /* Extra-special case of host notifications */
280 if (out && port == GUEST_NIO_PORT)
281 return (vmexit_handle_notify(ctx, vme, pvcpu, eax));
283 error = emulate_inout(ctx, vcpu, in, port, bytes, &eax, strictio);
284 if (error == 0 && in)
285 error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RAX, eax);
288 return (VMEXIT_CONTINUE);
290 fprintf(stderr, "Unhandled %s%c 0x%04x\n",
292 bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'), port);
293 return (vmexit_catch_inout());
298 vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
300 fprintf(stderr, "vm exit rdmsr 0x%x, cpu %d\n", vme->u.msr.code,
302 return (VMEXIT_ABORT);
306 vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
309 int retval = VMEXIT_CONTINUE;
311 newcpu = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code,vme->u.msr.wval);
317 vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
320 int retval = VMEXIT_CONTINUE;
322 newcpu = spinup_ap(ctx, *pvcpu,
323 vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip);
329 vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
332 fprintf(stderr, "vm exit[%d]\n", *pvcpu);
333 fprintf(stderr, "\treason\t\tVMX\n");
334 fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
335 fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
336 fprintf(stderr, "\terror\t\t%d\n", vmexit->u.vmx.error);
337 fprintf(stderr, "\texit_reason\t%u\n", vmexit->u.vmx.exit_reason);
338 fprintf(stderr, "\tqualification\t0x%016lx\n",
339 vmexit->u.vmx.exit_qualification);
341 return (VMEXIT_ABORT);
345 vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
348 stats.vmexit_bogus++;
350 return (VMEXIT_RESTART);
354 vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
360 * Just continue execution with the next instruction. We use
361 * the HLT VM exit as a way to be friendly with the host
364 return (VMEXIT_CONTINUE);
368 vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
371 stats.vmexit_pause++;
373 return (VMEXIT_CONTINUE);
377 vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
380 stats.vmexit_mtrap++;
382 return (VMEXIT_RESTART);
386 vmexit_inst_emul(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
389 stats.vmexit_inst_emul++;
391 err = emulate_mem(ctx, *pvcpu, vmexit->u.inst_emul.gpa,
392 &vmexit->u.inst_emul.vie);
397 "Failed to emulate instruction at 0x%lx\n",
399 } else if (err == ESRCH) {
400 fprintf(stderr, "Unhandled memory access to 0x%lx\n",
401 vmexit->u.inst_emul.gpa);
404 return (VMEXIT_ABORT);
407 return (VMEXIT_CONTINUE);
410 static vmexit_handler_t handler[VM_EXITCODE_MAX] = {
411 [VM_EXITCODE_INOUT] = vmexit_inout,
412 [VM_EXITCODE_VMX] = vmexit_vmx,
413 [VM_EXITCODE_BOGUS] = vmexit_bogus,
414 [VM_EXITCODE_RDMSR] = vmexit_rdmsr,
415 [VM_EXITCODE_WRMSR] = vmexit_wrmsr,
416 [VM_EXITCODE_MTRAP] = vmexit_mtrap,
417 [VM_EXITCODE_INST_EMUL] = vmexit_inst_emul,
418 [VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap,
422 vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip)
425 int error, rc, prevcpu;
426 enum vm_exitcode exitcode;
430 CPU_SET(pincpu + vcpu, &mask);
431 error = pthread_setaffinity_np(pthread_self(),
432 sizeof(mask), &mask);
437 error = vm_run(ctx, vcpu, rip, &vmexit[vcpu]);
440 * It is possible that 'vmmctl' or some other process
441 * has transitioned the vcpu to CANNOT_RUN state right
442 * before we tried to transition it to RUNNING.
444 * This is expected to be temporary so just retry.
454 exitcode = vmexit[vcpu].exitcode;
455 if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) {
456 fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n",
461 rc = (*handler[exitcode])(ctx, &vmexit[vcpu], &vcpu);
464 case VMEXIT_CONTINUE:
465 rip = vmexit[vcpu].rip + vmexit[vcpu].inst_length;
468 rip = vmexit[vcpu].rip;
476 fprintf(stderr, "vm_run error %d, errno %d\n", error, errno);
480 num_vcpus_allowed(struct vmctx *ctx)
484 error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp);
487 * The guest is allowed to spinup more than one processor only if the
488 * UNRESTRICTED_GUEST capability is available.
497 fbsdrun_set_capabilities(struct vmctx *ctx, int cpu)
501 if (fbsdrun_vmexit_on_hlt()) {
502 err = vm_get_capability(ctx, cpu, VM_CAP_HALT_EXIT, &tmp);
504 fprintf(stderr, "VM exit on HLT not supported\n");
507 vm_set_capability(ctx, cpu, VM_CAP_HALT_EXIT, 1);
509 handler[VM_EXITCODE_HLT] = vmexit_hlt;
512 if (fbsdrun_vmexit_on_pause()) {
514 * pause exit support required for this mode
516 err = vm_get_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, &tmp);
519 "SMP mux requested, no pause support\n");
522 vm_set_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, 1);
524 handler[VM_EXITCODE_PAUSE] = vmexit_pause;
527 if (fbsdrun_disable_x2apic())
528 err = vm_set_x2apic_state(ctx, cpu, X2APIC_DISABLED);
530 err = vm_set_x2apic_state(ctx, cpu, X2APIC_ENABLED);
533 fprintf(stderr, "Unable to set x2apic state (%d)\n", err);
537 vm_set_capability(ctx, cpu, VM_CAP_ENABLE_INVPCID, 1);
541 main(int argc, char *argv[])
543 int c, error, gdb_port, err, ioapic, bvmcons;
550 progname = basename(argv[0]);
556 while ((c = getopt(argc, argv, "abehAHIPWp:g:c:s:S:m:")) != -1) {
568 pincpu = atoi(optarg);
571 guest_ncpus = atoi(optarg);
574 gdb_port = atoi(optarg);
577 if (pci_parse_slot(optarg, 0) != 0)
582 if (pci_parse_slot(optarg, 1) != 0)
587 error = vm_parse_memsize(optarg, &memsize);
589 errx(EX_USAGE, "invalid memsize '%s'", optarg);
592 guest_vmexit_on_hlt = 1;
598 guest_vmexit_on_pause = 1;
620 ctx = vm_open(vmname);
626 max_vcpus = num_vcpus_allowed(ctx);
627 if (guest_ncpus > max_vcpus) {
628 fprintf(stderr, "%d vCPUs requested but only %d available\n",
629 guest_ncpus, max_vcpus);
633 fbsdrun_set_capabilities(ctx, BSP);
635 err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL);
637 fprintf(stderr, "Unable to setup memory (%d)\n", err);
647 * Exit if a device emulation finds an error in it's initilization
649 if (init_pci(ctx) != 0)
656 init_dbgport(gdb_port);
661 error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip);
665 * build the guest tables, MP etc.
667 mptable_build(ctx, guest_ncpus, ioapic);
670 error = acpi_build(ctx, guest_ncpus, ioapic);
677 fbsdrun_addcpu(ctx, BSP, rip);
680 * Head off to the main event dispatch loop