bhyverun.c revision 266393
1221167Sgnn/*- 2221167Sgnn * Copyright (c) 2011 NetApp, Inc. 3221167Sgnn * All rights reserved. 4221167Sgnn * 5221167Sgnn * Redistribution and use in source and binary forms, with or without 6221167Sgnn * modification, are permitted provided that the following conditions 7221167Sgnn * are met: 8221167Sgnn * 1. Redistributions of source code must retain the above copyright 9221167Sgnn * notice, this list of conditions and the following disclaimer. 10221167Sgnn * 2. Redistributions in binary form must reproduce the above copyright 11221167Sgnn * notice, this list of conditions and the following disclaimer in the 12221167Sgnn * documentation and/or other materials provided with the distribution. 13221167Sgnn * 14221167Sgnn * THIS SOFTWARE IS PROVIDED BY NETAPP, INC ``AS IS'' AND 15221167Sgnn * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 16221167Sgnn * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 17221167Sgnn * ARE DISCLAIMED. IN NO EVENT SHALL NETAPP, INC OR CONTRIBUTORS BE LIABLE 18221167Sgnn * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 19221167Sgnn * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 20221167Sgnn * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 21221167Sgnn * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 22221167Sgnn * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 23221167Sgnn * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 24221167Sgnn * SUCH DAMAGE. 25221167Sgnn * 26221167Sgnn * $FreeBSD: stable/10/usr.sbin/bhyve/bhyverun.c 266393 2014-05-18 04:33:24Z jhb $ 27221167Sgnn */ 28221167Sgnn 29221167Sgnn#include <sys/cdefs.h> 30221167Sgnn__FBSDID("$FreeBSD: stable/10/usr.sbin/bhyve/bhyverun.c 266393 2014-05-18 04:33:24Z jhb $"); 31221167Sgnn 32221167Sgnn#include <sys/types.h> 33221167Sgnn#include <sys/mman.h> 34221167Sgnn#include <sys/time.h> 35221167Sgnn 36221167Sgnn#include <machine/atomic.h> 37221167Sgnn#include <machine/segments.h> 38221167Sgnn 39221167Sgnn#include <stdio.h> 40221167Sgnn#include <stdlib.h> 41221167Sgnn#include <string.h> 42221167Sgnn#include <err.h> 43221167Sgnn#include <libgen.h> 44221167Sgnn#include <unistd.h> 45221167Sgnn#include <assert.h> 46221167Sgnn#include <errno.h> 47221167Sgnn#include <pthread.h> 48221167Sgnn#include <pthread_np.h> 49221167Sgnn#include <sysexits.h> 50221167Sgnn 51221167Sgnn#include <machine/vmm.h> 52221167Sgnn#include <vmmapi.h> 53221167Sgnn 54221167Sgnn#include "bhyverun.h" 55221167Sgnn#include "acpi.h" 56221167Sgnn#include "inout.h" 57221167Sgnn#include "dbgport.h" 58221167Sgnn#include "legacy_irq.h" 59221167Sgnn#include "mem.h" 60221167Sgnn#include "mevent.h" 61221167Sgnn#include "mptbl.h" 62221167Sgnn#include "pci_emul.h" 63221167Sgnn#include "pci_lpc.h" 64221167Sgnn#include "xmsr.h" 65221167Sgnn#include "spinup_ap.h" 66221167Sgnn#include "rtc.h" 67221167Sgnn 68221167Sgnn#define GUEST_NIO_PORT 0x488 /* guest upcalls via i/o port */ 69221167Sgnn 70221167Sgnn#define VMEXIT_SWITCH 0 /* force vcpu switch in mux mode */ 71221167Sgnn#define VMEXIT_CONTINUE 1 /* continue from next instruction */ 72221167Sgnn#define VMEXIT_RESTART 2 /* restart current instruction */ 73221167Sgnn#define VMEXIT_ABORT 3 /* abort the vm run loop */ 74221167Sgnn#define VMEXIT_RESET 4 /* guest machine has reset */ 75221167Sgnn#define VMEXIT_POWEROFF 5 /* guest machine has powered off */ 76221167Sgnn 77221167Sgnn#define MB (1024UL * 1024) 78221167Sgnn#define GB (1024UL * MB) 79221167Sgnn 80221167Sgnntypedef int (*vmexit_handler_t)(struct vmctx *, struct vm_exit *, int *vcpu); 81221167Sgnn 82221167Sgnnchar *vmname; 83221167Sgnn 84221167Sgnnint guest_ncpus; 85221167Sgnn 86221167Sgnnstatic int pincpu = -1; 87221167Sgnnstatic int guest_vmexit_on_hlt, guest_vmexit_on_pause, disable_x2apic; 88221167Sgnnstatic int virtio_msix = 1; 89221167Sgnn 90221167Sgnnstatic int strictio; 91221167Sgnnstatic int strictmsr = 1; 92221167Sgnn 93221167Sgnnstatic int acpi; 94221167Sgnn 95221167Sgnnstatic char *progname; 96221167Sgnnstatic const int BSP = 0; 97221167Sgnn 98221167Sgnnstatic int cpumask; 99221167Sgnn 100221167Sgnnstatic void vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip); 101221167Sgnn 102221167Sgnnstruct vm_exit vmexit[VM_MAXCPU]; 103221167Sgnn 104221167Sgnnstruct bhyvestats { 105221167Sgnn uint64_t vmexit_bogus; 106221167Sgnn uint64_t vmexit_bogus_switch; 107221167Sgnn uint64_t vmexit_hlt; 108221167Sgnn uint64_t vmexit_pause; 109221167Sgnn uint64_t vmexit_mtrap; 110221167Sgnn uint64_t vmexit_inst_emul; 111221167Sgnn uint64_t cpu_switch_rotate; 112221167Sgnn uint64_t cpu_switch_direct; 113221167Sgnn int io_reset; 114221167Sgnn} stats; 115221167Sgnn 116221167Sgnnstruct mt_vmm_info { 117221167Sgnn pthread_t mt_thr; 118221167Sgnn struct vmctx *mt_ctx; 119221167Sgnn int mt_vcpu; 120221167Sgnn} mt_vmm_info[VM_MAXCPU]; 121221167Sgnn 122221167Sgnnstatic void 123221167Sgnnusage(int code) 124221167Sgnn{ 125221167Sgnn 126221167Sgnn fprintf(stderr, 127221167Sgnn "Usage: %s [-aehwAHIPW] [-g <gdb port>] [-s <pci>] [-S <pci>]\n" 128221167Sgnn " %*s [-c vcpus] [-p pincpu] [-m mem] [-l <lpc>] <vm>\n" 129221167Sgnn " -a: local apic is in XAPIC mode (default is X2APIC)\n" 130221167Sgnn " -A: create an ACPI table\n" 131221167Sgnn " -g: gdb port\n" 132221167Sgnn " -c: # cpus (default 1)\n" 133221167Sgnn " -p: pin vcpu 'n' to host cpu 'pincpu + n'\n" 134221167Sgnn " -H: vmexit from the guest on hlt\n" 135221167Sgnn " -P: vmexit from the guest on pause\n" 136221167Sgnn " -W: force virtio to use single-vector MSI\n" 137221167Sgnn " -e: exit on unhandled I/O access\n" 138221167Sgnn " -h: help\n" 139221167Sgnn " -s: <slot,driver,configinfo> PCI slot config\n" 140221167Sgnn " -S: <slot,driver,configinfo> legacy PCI slot config\n" 141221167Sgnn " -l: LPC device configuration\n" 142221167Sgnn " -m: memory size in MB\n" 143221167Sgnn " -w: ignore unimplemented MSRs\n", 144221167Sgnn progname, (int)strlen(progname), ""); 145221167Sgnn 146221167Sgnn exit(code); 147221167Sgnn} 148221167Sgnn 149221167Sgnnvoid * 150221167Sgnnpaddr_guest2host(struct vmctx *ctx, uintptr_t gaddr, size_t len) 151221167Sgnn{ 152221167Sgnn 153221167Sgnn return (vm_map_gpa(ctx, gaddr, len)); 154221167Sgnn} 155221167Sgnn 156221167Sgnnint 157221167Sgnnfbsdrun_disable_x2apic(void) 158221167Sgnn{ 159221167Sgnn 160221167Sgnn return (disable_x2apic); 161221167Sgnn} 162221167Sgnn 163221167Sgnnint 164221167Sgnnfbsdrun_vmexit_on_pause(void) 165221167Sgnn{ 166221167Sgnn 167221167Sgnn return (guest_vmexit_on_pause); 168221167Sgnn} 169221167Sgnn 170221167Sgnnint 171221167Sgnnfbsdrun_vmexit_on_hlt(void) 172221167Sgnn{ 173221167Sgnn 174221167Sgnn return (guest_vmexit_on_hlt); 175221167Sgnn} 176221167Sgnn 177221167Sgnnint 178221167Sgnnfbsdrun_virtio_msix(void) 179221167Sgnn{ 180221167Sgnn 181221167Sgnn return (virtio_msix); 182221167Sgnn} 183221167Sgnn 184221167Sgnnstatic void * 185221167Sgnnfbsdrun_start_thread(void *param) 186221167Sgnn{ 187221167Sgnn char tname[MAXCOMLEN + 1]; 188221167Sgnn struct mt_vmm_info *mtp; 189221167Sgnn int vcpu; 190221167Sgnn 191221167Sgnn mtp = param; 192221167Sgnn vcpu = mtp->mt_vcpu; 193221167Sgnn 194221167Sgnn snprintf(tname, sizeof(tname), "vcpu %d", vcpu); 195221167Sgnn pthread_set_name_np(mtp->mt_thr, tname); 196221167Sgnn 197221167Sgnn vm_loop(mtp->mt_ctx, vcpu, vmexit[vcpu].rip); 198221167Sgnn 199221167Sgnn /* not reached */ 200221167Sgnn exit(1); 201221167Sgnn return (NULL); 202221167Sgnn} 203221167Sgnn 204221167Sgnnvoid 205221167Sgnnfbsdrun_addcpu(struct vmctx *ctx, int vcpu, uint64_t rip) 206221167Sgnn{ 207221167Sgnn int error; 208221167Sgnn 209221167Sgnn if (cpumask & (1 << vcpu)) { 210221167Sgnn fprintf(stderr, "addcpu: attempting to add existing cpu %d\n", 211221167Sgnn vcpu); 212221167Sgnn exit(1); 213221167Sgnn } 214221167Sgnn 215221167Sgnn atomic_set_int(&cpumask, 1 << vcpu); 216221167Sgnn 217221167Sgnn /* 218221167Sgnn * Set up the vmexit struct to allow execution to start 219221167Sgnn * at the given RIP 220221167Sgnn */ 221221167Sgnn vmexit[vcpu].rip = rip; 222221167Sgnn vmexit[vcpu].inst_length = 0; 223221167Sgnn 224221167Sgnn mt_vmm_info[vcpu].mt_ctx = ctx; 225221167Sgnn mt_vmm_info[vcpu].mt_vcpu = vcpu; 226221167Sgnn 227221167Sgnn error = pthread_create(&mt_vmm_info[vcpu].mt_thr, NULL, 228221167Sgnn fbsdrun_start_thread, &mt_vmm_info[vcpu]); 229221167Sgnn assert(error == 0); 230221167Sgnn} 231221167Sgnn 232221167Sgnnstatic int 233221167Sgnnfbsdrun_deletecpu(struct vmctx *ctx, int vcpu) 234221167Sgnn{ 235221167Sgnn 236221167Sgnn if ((cpumask & (1 << vcpu)) == 0) { 237221167Sgnn fprintf(stderr, "addcpu: attempting to delete unknown cpu %d\n", 238221167Sgnn vcpu); 239221167Sgnn exit(1); 240221167Sgnn } 241221167Sgnn 242221167Sgnn atomic_clear_int(&cpumask, 1 << vcpu); 243221167Sgnn return (cpumask == 0); 244221167Sgnn} 245221167Sgnn 246221167Sgnnstatic int 247221167Sgnnvmexit_catch_reset(void) 248221167Sgnn{ 249221167Sgnn stats.io_reset++; 250221167Sgnn return (VMEXIT_RESET); 251221167Sgnn} 252221167Sgnn 253221167Sgnnstatic int 254221167Sgnnvmexit_catch_inout(void) 255221167Sgnn{ 256221167Sgnn return (VMEXIT_ABORT); 257221167Sgnn} 258221167Sgnn 259221167Sgnnstatic int 260221167Sgnnvmexit_handle_notify(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu, 261221167Sgnn uint32_t eax) 262221167Sgnn{ 263221167Sgnn#if BHYVE_DEBUG 264221167Sgnn /* 265221167Sgnn * put guest-driven debug here 266221167Sgnn */ 267221167Sgnn#endif 268221167Sgnn return (VMEXIT_CONTINUE); 269221167Sgnn} 270221167Sgnn 271221167Sgnnstatic int 272221167Sgnnvmexit_inout(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 273221167Sgnn{ 274221167Sgnn int error; 275221167Sgnn int bytes, port, in, out; 276221167Sgnn uint32_t eax; 277221167Sgnn int vcpu; 278221167Sgnn 279221167Sgnn vcpu = *pvcpu; 280221167Sgnn 281221167Sgnn port = vme->u.inout.port; 282221167Sgnn bytes = vme->u.inout.bytes; 283221167Sgnn eax = vme->u.inout.eax; 284221167Sgnn in = vme->u.inout.in; 285221167Sgnn out = !in; 286221167Sgnn 287221167Sgnn /* We don't deal with these */ 288221167Sgnn if (vme->u.inout.string || vme->u.inout.rep) 289221167Sgnn return (VMEXIT_ABORT); 290221167Sgnn 291221167Sgnn /* Special case of guest reset */ 292221167Sgnn if (out && port == 0x64 && (uint8_t)eax == 0xFE) 293221167Sgnn return (vmexit_catch_reset()); 294221167Sgnn 295221167Sgnn /* Extra-special case of host notifications */ 296221167Sgnn if (out && port == GUEST_NIO_PORT) 297221167Sgnn return (vmexit_handle_notify(ctx, vme, pvcpu, eax)); 298221167Sgnn 299221167Sgnn error = emulate_inout(ctx, vcpu, in, port, bytes, &eax, strictio); 300221167Sgnn if (error == INOUT_OK && in) 301221167Sgnn error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RAX, eax); 302221167Sgnn 303221167Sgnn switch (error) { 304221167Sgnn case INOUT_OK: 305221167Sgnn return (VMEXIT_CONTINUE); 306221167Sgnn case INOUT_RESET: 307221167Sgnn return (VMEXIT_RESET); 308221167Sgnn case INOUT_POWEROFF: 309221167Sgnn return (VMEXIT_POWEROFF); 310221167Sgnn default: 311221167Sgnn fprintf(stderr, "Unhandled %s%c 0x%04x\n", 312221167Sgnn in ? "in" : "out", 313221167Sgnn bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'), port); 314221167Sgnn return (vmexit_catch_inout()); 315221167Sgnn } 316221167Sgnn} 317221167Sgnn 318221167Sgnnstatic int 319221167Sgnnvmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 320221167Sgnn{ 321221167Sgnn uint64_t val; 322221167Sgnn uint32_t eax, edx; 323221167Sgnn int error; 324221167Sgnn 325221167Sgnn val = 0; 326221167Sgnn error = emulate_rdmsr(ctx, *pvcpu, vme->u.msr.code, &val); 327221167Sgnn if (error != 0) { 328221167Sgnn fprintf(stderr, "rdmsr to register %#x on vcpu %d\n", 329221167Sgnn vme->u.msr.code, *pvcpu); 330221167Sgnn if (strictmsr) 331221167Sgnn return (VMEXIT_ABORT); 332221167Sgnn } 333221167Sgnn 334221167Sgnn eax = val; 335221167Sgnn error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RAX, eax); 336221167Sgnn assert(error == 0); 337221167Sgnn 338221167Sgnn edx = val >> 32; 339221167Sgnn error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RDX, edx); 340221167Sgnn assert(error == 0); 341221167Sgnn 342221167Sgnn return (VMEXIT_CONTINUE); 343221167Sgnn} 344221167Sgnn 345221167Sgnnstatic int 346221167Sgnnvmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 347221167Sgnn{ 348221167Sgnn int error; 349221167Sgnn 350221167Sgnn error = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code, vme->u.msr.wval); 351221167Sgnn if (error != 0) { 352221167Sgnn fprintf(stderr, "wrmsr to register %#x(%#lx) on vcpu %d\n", 353221167Sgnn vme->u.msr.code, vme->u.msr.wval, *pvcpu); 354221167Sgnn if (strictmsr) 355221167Sgnn return (VMEXIT_ABORT); 356221167Sgnn } 357221167Sgnn return (VMEXIT_CONTINUE); 358221167Sgnn} 359221167Sgnn 360221167Sgnnstatic int 361221167Sgnnvmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 362221167Sgnn{ 363221167Sgnn int newcpu; 364221167Sgnn int retval = VMEXIT_CONTINUE; 365221167Sgnn 366221167Sgnn newcpu = spinup_ap(ctx, *pvcpu, 367221167Sgnn vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip); 368221167Sgnn 369221167Sgnn return (retval); 370221167Sgnn} 371221167Sgnn 372221167Sgnnstatic int 373221167Sgnnvmexit_spindown_cpu(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 374221167Sgnn{ 375221167Sgnn int lastcpu; 376221167Sgnn 377221167Sgnn lastcpu = fbsdrun_deletecpu(ctx, *pvcpu); 378221167Sgnn if (!lastcpu) 379221167Sgnn pthread_exit(NULL); 380221167Sgnn return (vmexit_catch_reset()); 381221167Sgnn} 382221167Sgnn 383221167Sgnnstatic int 384221167Sgnnvmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 385221167Sgnn{ 386221167Sgnn 387221167Sgnn fprintf(stderr, "vm exit[%d]\n", *pvcpu); 388221167Sgnn fprintf(stderr, "\treason\t\tVMX\n"); 389221167Sgnn fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip); 390221167Sgnn fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length); 391221167Sgnn fprintf(stderr, "\tstatus\t\t%d\n", vmexit->u.vmx.status); 392221167Sgnn fprintf(stderr, "\texit_reason\t%u\n", vmexit->u.vmx.exit_reason); 393221167Sgnn fprintf(stderr, "\tqualification\t0x%016lx\n", 394221167Sgnn vmexit->u.vmx.exit_qualification); 395221167Sgnn fprintf(stderr, "\tinst_type\t\t%d\n", vmexit->u.vmx.inst_type); 396221167Sgnn fprintf(stderr, "\tinst_error\t\t%d\n", vmexit->u.vmx.inst_error); 397221167Sgnn 398221167Sgnn return (VMEXIT_ABORT); 399221167Sgnn} 400221167Sgnn 401221167Sgnnstatic int 402221167Sgnnvmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 403221167Sgnn{ 404221167Sgnn 405221167Sgnn stats.vmexit_bogus++; 406221167Sgnn 407221167Sgnn return (VMEXIT_RESTART); 408221167Sgnn} 409221167Sgnn 410221167Sgnnstatic int 411221167Sgnnvmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 412221167Sgnn{ 413221167Sgnn 414221167Sgnn stats.vmexit_hlt++; 415221167Sgnn 416221167Sgnn /* 417221167Sgnn * Just continue execution with the next instruction. We use 418221167Sgnn * the HLT VM exit as a way to be friendly with the host 419221167Sgnn * scheduler. 420221167Sgnn */ 421221167Sgnn return (VMEXIT_CONTINUE); 422221167Sgnn} 423221167Sgnn 424221167Sgnnstatic int 425221167Sgnnvmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 426221167Sgnn{ 427221167Sgnn 428221167Sgnn stats.vmexit_pause++; 429221167Sgnn 430221167Sgnn return (VMEXIT_CONTINUE); 431221167Sgnn} 432221167Sgnn 433221167Sgnnstatic int 434221167Sgnnvmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 435221167Sgnn{ 436221167Sgnn 437221167Sgnn stats.vmexit_mtrap++; 438221167Sgnn 439221167Sgnn return (VMEXIT_RESTART); 440221167Sgnn} 441221167Sgnn 442221167Sgnnstatic int 443221167Sgnnvmexit_inst_emul(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 444221167Sgnn{ 445221167Sgnn int err; 446221167Sgnn stats.vmexit_inst_emul++; 447221167Sgnn 448221167Sgnn err = emulate_mem(ctx, *pvcpu, vmexit->u.inst_emul.gpa, 449221167Sgnn &vmexit->u.inst_emul.vie); 450221167Sgnn 451 if (err) { 452 if (err == EINVAL) { 453 fprintf(stderr, 454 "Failed to emulate instruction at 0x%lx\n", 455 vmexit->rip); 456 } else if (err == ESRCH) { 457 fprintf(stderr, "Unhandled memory access to 0x%lx\n", 458 vmexit->u.inst_emul.gpa); 459 } 460 461 return (VMEXIT_ABORT); 462 } 463 464 return (VMEXIT_CONTINUE); 465} 466 467static vmexit_handler_t handler[VM_EXITCODE_MAX] = { 468 [VM_EXITCODE_INOUT] = vmexit_inout, 469 [VM_EXITCODE_VMX] = vmexit_vmx, 470 [VM_EXITCODE_BOGUS] = vmexit_bogus, 471 [VM_EXITCODE_RDMSR] = vmexit_rdmsr, 472 [VM_EXITCODE_WRMSR] = vmexit_wrmsr, 473 [VM_EXITCODE_MTRAP] = vmexit_mtrap, 474 [VM_EXITCODE_INST_EMUL] = vmexit_inst_emul, 475 [VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap, 476 [VM_EXITCODE_SPINDOWN_CPU] = vmexit_spindown_cpu, 477}; 478 479static void 480vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip) 481{ 482 cpuset_t mask; 483 int error, rc, prevcpu; 484 enum vm_exitcode exitcode; 485 486 if (pincpu >= 0) { 487 CPU_ZERO(&mask); 488 CPU_SET(pincpu + vcpu, &mask); 489 error = pthread_setaffinity_np(pthread_self(), 490 sizeof(mask), &mask); 491 assert(error == 0); 492 } 493 494 while (1) { 495 error = vm_run(ctx, vcpu, rip, &vmexit[vcpu]); 496 if (error != 0) 497 break; 498 499 prevcpu = vcpu; 500 501 exitcode = vmexit[vcpu].exitcode; 502 if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) { 503 fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n", 504 exitcode); 505 exit(1); 506 } 507 508 rc = (*handler[exitcode])(ctx, &vmexit[vcpu], &vcpu); 509 510 switch (rc) { 511 case VMEXIT_CONTINUE: 512 rip = vmexit[vcpu].rip + vmexit[vcpu].inst_length; 513 break; 514 case VMEXIT_RESTART: 515 rip = vmexit[vcpu].rip; 516 break; 517 case VMEXIT_RESET: 518 exit(0); 519 default: 520 exit(1); 521 } 522 } 523 fprintf(stderr, "vm_run error %d, errno %d\n", error, errno); 524} 525 526static int 527num_vcpus_allowed(struct vmctx *ctx) 528{ 529 int tmp, error; 530 531 error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp); 532 533 /* 534 * The guest is allowed to spinup more than one processor only if the 535 * UNRESTRICTED_GUEST capability is available. 536 */ 537 if (error == 0) 538 return (VM_MAXCPU); 539 else 540 return (1); 541} 542 543void 544fbsdrun_set_capabilities(struct vmctx *ctx, int cpu) 545{ 546 int err, tmp; 547 548 if (fbsdrun_vmexit_on_hlt()) { 549 err = vm_get_capability(ctx, cpu, VM_CAP_HALT_EXIT, &tmp); 550 if (err < 0) { 551 fprintf(stderr, "VM exit on HLT not supported\n"); 552 exit(1); 553 } 554 vm_set_capability(ctx, cpu, VM_CAP_HALT_EXIT, 1); 555 if (cpu == BSP) 556 handler[VM_EXITCODE_HLT] = vmexit_hlt; 557 } 558 559 if (fbsdrun_vmexit_on_pause()) { 560 /* 561 * pause exit support required for this mode 562 */ 563 err = vm_get_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, &tmp); 564 if (err < 0) { 565 fprintf(stderr, 566 "SMP mux requested, no pause support\n"); 567 exit(1); 568 } 569 vm_set_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, 1); 570 if (cpu == BSP) 571 handler[VM_EXITCODE_PAUSE] = vmexit_pause; 572 } 573 574 if (fbsdrun_disable_x2apic()) 575 err = vm_set_x2apic_state(ctx, cpu, X2APIC_DISABLED); 576 else 577 err = vm_set_x2apic_state(ctx, cpu, X2APIC_ENABLED); 578 579 if (err) { 580 fprintf(stderr, "Unable to set x2apic state (%d)\n", err); 581 exit(1); 582 } 583 584 vm_set_capability(ctx, cpu, VM_CAP_ENABLE_INVPCID, 1); 585} 586 587int 588main(int argc, char *argv[]) 589{ 590 int c, error, gdb_port, err, bvmcons; 591 int max_vcpus; 592 struct vmctx *ctx; 593 uint64_t rip; 594 size_t memsize; 595 596 bvmcons = 0; 597 progname = basename(argv[0]); 598 gdb_port = 0; 599 guest_ncpus = 1; 600 memsize = 256 * MB; 601 602 while ((c = getopt(argc, argv, "abehwAHIPWp:g:c:s:S:m:l:")) != -1) { 603 switch (c) { 604 case 'a': 605 disable_x2apic = 1; 606 break; 607 case 'A': 608 acpi = 1; 609 break; 610 case 'b': 611 bvmcons = 1; 612 break; 613 case 'p': 614 pincpu = atoi(optarg); 615 break; 616 case 'c': 617 guest_ncpus = atoi(optarg); 618 break; 619 case 'g': 620 gdb_port = atoi(optarg); 621 break; 622 case 'l': 623 if (lpc_device_parse(optarg) != 0) { 624 errx(EX_USAGE, "invalid lpc device " 625 "configuration '%s'", optarg); 626 } 627 break; 628 case 's': 629 if (pci_parse_slot(optarg, 0) != 0) 630 exit(1); 631 else 632 break; 633 case 'S': 634 if (pci_parse_slot(optarg, 1) != 0) 635 exit(1); 636 else 637 break; 638 case 'm': 639 error = vm_parse_memsize(optarg, &memsize); 640 if (error) 641 errx(EX_USAGE, "invalid memsize '%s'", optarg); 642 break; 643 case 'H': 644 guest_vmexit_on_hlt = 1; 645 break; 646 case 'I': 647 /* 648 * The "-I" option was used to add an ioapic to the 649 * virtual machine. 650 * 651 * An ioapic is now provided unconditionally for each 652 * virtual machine and this option is now deprecated. 653 */ 654 break; 655 case 'P': 656 guest_vmexit_on_pause = 1; 657 break; 658 case 'e': 659 strictio = 1; 660 break; 661 case 'w': 662 strictmsr = 0; 663 break; 664 case 'W': 665 virtio_msix = 0; 666 break; 667 case 'h': 668 usage(0); 669 default: 670 usage(1); 671 } 672 } 673 argc -= optind; 674 argv += optind; 675 676 if (argc != 1) 677 usage(1); 678 679 vmname = argv[0]; 680 681 ctx = vm_open(vmname); 682 if (ctx == NULL) { 683 perror("vm_open"); 684 exit(1); 685 } 686 687 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); 691 exit(1); 692 } 693 694 fbsdrun_set_capabilities(ctx, BSP); 695 696 err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL); 697 if (err) { 698 fprintf(stderr, "Unable to setup memory (%d)\n", err); 699 exit(1); 700 } 701 702 init_mem(); 703 init_inout(); 704 legacy_irq_init(); 705 706 rtc_init(ctx); 707 708 /* 709 * Exit if a device emulation finds an error in it's initilization 710 */ 711 if (init_pci(ctx) != 0) 712 exit(1); 713 714 if (gdb_port != 0) 715 init_dbgport(gdb_port); 716 717 if (bvmcons) 718 init_bvmcons(); 719 720 error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip); 721 assert(error == 0); 722 723 /* 724 * build the guest tables, MP etc. 725 */ 726 mptable_build(ctx, guest_ncpus); 727 728 if (acpi) { 729 error = acpi_build(ctx, guest_ncpus); 730 assert(error == 0); 731 } 732 733 /* 734 * Change the proc title to include the VM name. 735 */ 736 setproctitle("%s", vmname); 737 738 /* 739 * Add CPU 0 740 */ 741 fbsdrun_addcpu(ctx, BSP, rip); 742 743 /* 744 * Head off to the main event dispatch loop 745 */ 746 mevent_dispatch(); 747 748 exit(1); 749} 750