usr.sbin/bhyve/bhyverun.c

   1 /*-
   2  * Copyright (c) 2011 NetApp, Inc.
   3  * All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions
   7  * are met:
   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.
  13  *
  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
  24  * SUCH DAMAGE.
  25  *
  26  * $FreeBSD$
  27  */
  28
  29 #include <sys/cdefs.h>
  30 __FBSDID("$FreeBSD$");
  31
  32 #include <sys/types.h>
  33 #include <sys/mman.h>
  34 #include <sys/time.h>
  35
  36 #include <machine/atomic.h>
  37 #include <machine/segments.h>
  38
  39 #include <stdio.h>
  40 #include <stdlib.h>
  41 #include <string.h>
  42 #include <err.h>
  43 #include <libgen.h>
  44 #include <unistd.h>
  45 #include <assert.h>
  46 #include <errno.h>
  47 #include <pthread.h>
  48 #include <pthread_np.h>
  49 #include <sysexits.h>
  50 #include <stdbool.h>
  51
  52 #include <machine/vmm.h>
  53 #include <vmmapi.h>
  54
  55 #include "bhyverun.h"
  56 #include "acpi.h"
  57 #include "inout.h"
  58 #include "dbgport.h"
  59 #include "fwctl.h"
  60 #include "ioapic.h"
  61 #include "mem.h"
  62 #include "mevent.h"
  63 #include "mptbl.h"
  64 #include "pci_emul.h"
  65 #include "pci_irq.h"
  66 #include "pci_lpc.h"
  67 #include "smbiostbl.h"
  68 #include "xmsr.h"
  69 #include "spinup_ap.h"
  70 #include "rtc.h"
  71
  72 #define GUEST_NIO_PORT          0x488   /* guest upcalls via i/o port */
  73
  74 #define MB              (1024UL * 1024)
  75 #define GB              (1024UL * MB)
  76
  77 typedef int (*vmexit_handler_t)(struct vmctx *, struct vm_exit *, int *vcpu);
  78 extern int vmexit_task_switch(struct vmctx *, struct vm_exit *, int *vcpu);
  79
  80 char *vmname;
  81
  82 int guest_ncpus;
  83 char *guest_uuid_str;
  84
  85 static int guest_vmexit_on_hlt, guest_vmexit_on_pause;
  86 static int virtio_msix = 1;
  87 static int x2apic_mode = 0;     /* default is xAPIC */
  88
  89 static int strictio;
  90 static int strictmsr = 1;
  91
  92 static int acpi;
  93
  94 static char *progname;
  95 static const int BSP = 0;
  96
  97 static cpuset_t cpumask;
  98
  99 static void vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip);
 100
 101 static struct vm_exit vmexit[VM_MAXCPU];
 102
 103 struct bhyvestats {
 104         uint64_t        vmexit_bogus;
 105         uint64_t        vmexit_reqidle;
 106         uint64_t        vmexit_hlt;
 107         uint64_t        vmexit_pause;
 108         uint64_t        vmexit_mtrap;
 109         uint64_t        vmexit_inst_emul;
 110         uint64_t        cpu_switch_rotate;
 111         uint64_t        cpu_switch_direct;
 112 } stats;
 113
 114 struct mt_vmm_info {
 115         pthread_t       mt_thr;
 116         struct vmctx    *mt_ctx;
 117         int             mt_vcpu;
 118 } mt_vmm_info[VM_MAXCPU];
 119
 120 static cpuset_t *vcpumap[VM_MAXCPU] = { NULL };
 121
 122 static void
 123 usage(int code)
 124 {
 125
 126         fprintf(stderr,
 127                 "Usage: %s [-abehuwxACHPSWY] [-c vcpus] [-g <gdb port>] [-l <lpc>]\n"
 128                 "       %*s [-m mem] [-p vcpu:hostcpu] [-s <pci>] [-U uuid] <vm>\n"
 129                 "       -a: local apic is in xAPIC mode (deprecated)\n"
 130                 "       -A: create ACPI tables\n"
 131                 "       -c: # cpus (default 1)\n"
 132                 "       -C: include guest memory in core file\n"
 133                 "       -e: exit on unhandled I/O access\n"
 134                 "       -g: gdb port\n"
 135                 "       -h: help\n"
 136                 "       -H: vmexit from the guest on hlt\n"
 137                 "       -l: LPC device configuration\n"
 138                 "       -m: memory size in MB\n"
 139                 "       -p: pin 'vcpu' to 'hostcpu'\n"
 140                 "       -P: vmexit from the guest on pause\n"
 141                 "       -s: <slot,driver,configinfo> PCI slot config\n"
 142                 "       -S: guest memory cannot be swapped\n"
 143                 "       -u: RTC keeps UTC time\n"
 144                 "       -U: uuid\n"
 145                 "       -w: ignore unimplemented MSRs\n"
 146                 "       -W: force virtio to use single-vector MSI\n"
 147                 "       -x: local apic is in x2APIC mode\n"
 148                 "       -Y: disable MPtable generation\n",
 149                 progname, (int)strlen(progname), "");
 150
 151         exit(code);
 152 }
 153
 154 static int
 155 pincpu_parse(const char *opt)
 156 {
 157         int vcpu, pcpu;
 158
 159         if (sscanf(opt, "%d:%d", &vcpu, &pcpu) != 2) {
 160                 fprintf(stderr, "invalid format: %s\n", opt);
 161                 return (-1);
 162         }
 163
 164         if (vcpu < 0 || vcpu >= VM_MAXCPU) {
 165                 fprintf(stderr, "vcpu '%d' outside valid range from 0 to %d\n",
 166                     vcpu, VM_MAXCPU - 1);
 167                 return (-1);
 168         }
 169
 170         if (pcpu < 0 || pcpu >= CPU_SETSIZE) {
 171                 fprintf(stderr, "hostcpu '%d' outside valid range from "
 172                     "0 to %d\n", pcpu, CPU_SETSIZE - 1);
 173                 return (-1);
 174         }
 175
 176         if (vcpumap[vcpu] == NULL) {
 177                 if ((vcpumap[vcpu] = malloc(sizeof(cpuset_t))) == NULL) {
 178                         perror("malloc");
 179                         return (-1);
 180                 }
 181                 CPU_ZERO(vcpumap[vcpu]);
 182         }
 183         CPU_SET(pcpu, vcpumap[vcpu]);
 184         return (0);
 185 }
 186
 187 void
 188 vm_inject_fault(void *arg, int vcpu, int vector, int errcode_valid,
 189     int errcode)
 190 {
 191         struct vmctx *ctx;
 192         int error, restart_instruction;
 193
 194         ctx = arg;
 195         restart_instruction = 1;
 196
 197         error = vm_inject_exception(ctx, vcpu, vector, errcode_valid, errcode,
 198             restart_instruction);
 199         assert(error == 0);
 200 }
 201
 202 void *
 203 paddr_guest2host(struct vmctx *ctx, uintptr_t gaddr, size_t len)
 204 {
 205
 206         return (vm_map_gpa(ctx, gaddr, len));
 207 }
 208
 209 int
 210 fbsdrun_vmexit_on_pause(void)
 211 {
 212
 213         return (guest_vmexit_on_pause);
 214 }
 215
 216 int
 217 fbsdrun_vmexit_on_hlt(void)
 218 {
 219
 220         return (guest_vmexit_on_hlt);
 221 }
 222
 223 int
 224 fbsdrun_virtio_msix(void)
 225 {
 226
 227         return (virtio_msix);
 228 }
 229
 230 static void *
 231 fbsdrun_start_thread(void *param)
 232 {
 233         char tname[MAXCOMLEN + 1];
 234         struct mt_vmm_info *mtp;
 235         int vcpu;
 236
 237         mtp = param;
 238         vcpu = mtp->mt_vcpu;
 239
 240         snprintf(tname, sizeof(tname), "vcpu %d", vcpu);
 241         pthread_set_name_np(mtp->mt_thr, tname);
 242
 243         vm_loop(mtp->mt_ctx, vcpu, vmexit[vcpu].rip);
 244
 245         /* not reached */
 246         exit(1);
 247         return (NULL);
 248 }
 249
 250 void
 251 fbsdrun_addcpu(struct vmctx *ctx, int fromcpu, int newcpu, uint64_t rip)
 252 {
 253         int error;
 254
 255         assert(fromcpu == BSP);
 256
 257         /*
 258          * The 'newcpu' must be activated in the context of 'fromcpu'. If
 259          * vm_activate_cpu() is delayed until newcpu's pthread starts running
 260          * then vmm.ko is out-of-sync with bhyve and this can create a race
 261          * with vm_suspend().
 262          */
 263         error = vm_activate_cpu(ctx, newcpu);
 264         if (error != 0)
 265                 err(EX_OSERR, "could not activate CPU %d", newcpu);
 266
 267         CPU_SET_ATOMIC(newcpu, &cpumask);
 268
 269         /*
 270          * Set up the vmexit struct to allow execution to start
 271          * at the given RIP
 272          */
 273         vmexit[newcpu].rip = rip;
 274         vmexit[newcpu].inst_length = 0;
 275
 276         mt_vmm_info[newcpu].mt_ctx = ctx;
 277         mt_vmm_info[newcpu].mt_vcpu = newcpu;
 278
 279         error = pthread_create(&mt_vmm_info[newcpu].mt_thr, NULL,
 280             fbsdrun_start_thread, &mt_vmm_info[newcpu]);
 281         assert(error == 0);
 282 }
 283
 284 static int
 285 fbsdrun_deletecpu(struct vmctx *ctx, int vcpu)
 286 {
 287
 288         if (!CPU_ISSET(vcpu, &cpumask)) {
 289                 fprintf(stderr, "Attempting to delete unknown cpu %d\n", vcpu);
 290                 exit(1);
 291         }
 292
 293         CPU_CLR_ATOMIC(vcpu, &cpumask);
 294         return (CPU_EMPTY(&cpumask));
 295 }
 296
 297 static int
 298 vmexit_handle_notify(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu,
 299                      uint32_t eax)
 300 {
 301 #if BHYVE_DEBUG
 302         /*
 303          * put guest-driven debug here
 304          */
 305 #endif
 306         return (VMEXIT_CONTINUE);
 307 }
 308
 309 static int
 310 vmexit_inout(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
 311 {
 312         int error;
 313         int bytes, port, in, out;
 314         int vcpu;
 315
 316         vcpu = *pvcpu;
 317
 318         port = vme->u.inout.port;
 319         bytes = vme->u.inout.bytes;
 320         in = vme->u.inout.in;
 321         out = !in;
 322
 323         /* Extra-special case of host notifications */
 324         if (out && port == GUEST_NIO_PORT) {
 325                 error = vmexit_handle_notify(ctx, vme, pvcpu, vme->u.inout.eax);
 326                 return (error);
 327         }
 328
 329         error = emulate_inout(ctx, vcpu, vme, strictio);
 330         if (error) {
 331                 fprintf(stderr, "Unhandled %s%c 0x%04x at 0x%lx\n",
 332                     in ? "in" : "out",
 333                     bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'),
 334                     port, vmexit->rip);
 335                 return (VMEXIT_ABORT);
 336         } else {
 337                 return (VMEXIT_CONTINUE);
 338         }
 339 }
 340
 341 static int
 342 vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
 343 {
 344         uint64_t val;
 345         uint32_t eax, edx;
 346         int error;
 347
 348         val = 0;
 349         error = emulate_rdmsr(ctx, *pvcpu, vme->u.msr.code, &val);
 350         if (error != 0) {
 351                 fprintf(stderr, "rdmsr to register %#x on vcpu %d\n",
 352                     vme->u.msr.code, *pvcpu);
 353                 if (strictmsr) {
 354                         vm_inject_gp(ctx, *pvcpu);
 355                         return (VMEXIT_CONTINUE);
 356                 }
 357         }
 358
 359         eax = val;
 360         error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RAX, eax);
 361         assert(error == 0);
 362
 363         edx = val >> 32;
 364         error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RDX, edx);
 365         assert(error == 0);
 366
 367         return (VMEXIT_CONTINUE);
 368 }
 369
 370 static int
 371 vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
 372 {
 373         int error;
 374
 375         error = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code, vme->u.msr.wval);
 376         if (error != 0) {
 377                 fprintf(stderr, "wrmsr to register %#x(%#lx) on vcpu %d\n",
 378                     vme->u.msr.code, vme->u.msr.wval, *pvcpu);
 379                 if (strictmsr) {
 380                         vm_inject_gp(ctx, *pvcpu);
 381                         return (VMEXIT_CONTINUE);
 382                 }
 383         }
 384         return (VMEXIT_CONTINUE);
 385 }
 386
 387 static int
 388 vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
 389 {
 390         int newcpu;
 391         int retval = VMEXIT_CONTINUE;
 392
 393         newcpu = spinup_ap(ctx, *pvcpu,
 394                            vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip);
 395
 396         return (retval);
 397 }
 398
 399 #define DEBUG_EPT_MISCONFIG
 400 #ifdef DEBUG_EPT_MISCONFIG
 401 #define EXIT_REASON_EPT_MISCONFIG       49
 402 #define VMCS_GUEST_PHYSICAL_ADDRESS     0x00002400
 403 #define VMCS_IDENT(x)                   ((x) | 0x80000000)
 404
 405 static uint64_t ept_misconfig_gpa, ept_misconfig_pte[4];
 406 static int ept_misconfig_ptenum;
 407 #endif
 408
 409 static int
 410 vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
 411 {
 412
 413         fprintf(stderr, "vm exit[%d]\n", *pvcpu);
 414         fprintf(stderr, "\treason\t\tVMX\n");
 415         fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
 416         fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
 417         fprintf(stderr, "\tstatus\t\t%d\n", vmexit->u.vmx.status);
 418         fprintf(stderr, "\texit_reason\t%u\n", vmexit->u.vmx.exit_reason);
 419         fprintf(stderr, "\tqualification\t0x%016lx\n",
 420             vmexit->u.vmx.exit_qualification);
 421         fprintf(stderr, "\tinst_type\t\t%d\n", vmexit->u.vmx.inst_type);
 422         fprintf(stderr, "\tinst_error\t\t%d\n", vmexit->u.vmx.inst_error);
 423 #ifdef DEBUG_EPT_MISCONFIG
 424         if (vmexit->u.vmx.exit_reason == EXIT_REASON_EPT_MISCONFIG) {
 425                 vm_get_register(ctx, *pvcpu,
 426                     VMCS_IDENT(VMCS_GUEST_PHYSICAL_ADDRESS),
 427                     &ept_misconfig_gpa);
 428                 vm_get_gpa_pmap(ctx, ept_misconfig_gpa, ept_misconfig_pte,
 429                     &ept_misconfig_ptenum);
 430                 fprintf(stderr, "\tEPT misconfiguration:\n");
 431                 fprintf(stderr, "\t\tGPA: %#lx\n", ept_misconfig_gpa);
 432                 fprintf(stderr, "\t\tPTE(%d): %#lx %#lx %#lx %#lx\n",
 433                     ept_misconfig_ptenum, ept_misconfig_pte[0],
 434                     ept_misconfig_pte[1], ept_misconfig_pte[2],
 435                     ept_misconfig_pte[3]);
 436         }
 437 #endif  /* DEBUG_EPT_MISCONFIG */
 438         return (VMEXIT_ABORT);
 439 }
 440
 441 static int
 442 vmexit_svm(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
 443 {
 444
 445         fprintf(stderr, "vm exit[%d]\n", *pvcpu);
 446         fprintf(stderr, "\treason\t\tSVM\n");
 447         fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
 448         fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
 449         fprintf(stderr, "\texitcode\t%#lx\n", vmexit->u.svm.exitcode);
 450         fprintf(stderr, "\texitinfo1\t%#lx\n", vmexit->u.svm.exitinfo1);
 451         fprintf(stderr, "\texitinfo2\t%#lx\n", vmexit->u.svm.exitinfo2);
 452         return (VMEXIT_ABORT);
 453 }
 454
 455 static int
 456 vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
 457 {
 458
 459         assert(vmexit->inst_length == 0);
 460
 461         stats.vmexit_bogus++;
 462
 463         return (VMEXIT_CONTINUE);
 464 }
 465
 466 static int
 467 vmexit_reqidle(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
 468 {
 469
 470         assert(vmexit->inst_length == 0);
 471
 472         stats.vmexit_reqidle++;
 473
 474         return (VMEXIT_CONTINUE);
 475 }
 476
 477 static int
 478 vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
 479 {
 480
 481         stats.vmexit_hlt++;
 482
 483         /*
 484          * Just continue execution with the next instruction. We use
 485          * the HLT VM exit as a way to be friendly with the host
 486          * scheduler.
 487          */
 488         return (VMEXIT_CONTINUE);
 489 }
 490
 491 static int
 492 vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
 493 {
 494
 495         stats.vmexit_pause++;
 496
 497         return (VMEXIT_CONTINUE);
 498 }
 499
 500 static int
 501 vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
 502 {
 503
 504         assert(vmexit->inst_length == 0);
 505
 506         stats.vmexit_mtrap++;
 507
 508         return (VMEXIT_CONTINUE);
 509 }
 510
 511 static int
 512 vmexit_inst_emul(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
 513 {
 514         int err, i;
 515         struct vie *vie;
 516
 517         stats.vmexit_inst_emul++;
 518
 519         vie = &vmexit->u.inst_emul.vie;
 520         err = emulate_mem(ctx, *pvcpu, vmexit->u.inst_emul.gpa,
 521             vie, &vmexit->u.inst_emul.paging);
 522
 523         if (err) {
 524                 if (err == ESRCH) {
 525                         fprintf(stderr, "Unhandled memory access to 0x%lx\n",
 526                             vmexit->u.inst_emul.gpa);
 527                 }
 528
 529                 fprintf(stderr, "Failed to emulate instruction [");
 530                 for (i = 0; i < vie->num_valid; i++) {
 531                         fprintf(stderr, "0x%02x%s", vie->inst[i],
 532                             i != (vie->num_valid - 1) ? " " : "");
 533                 }
 534                 fprintf(stderr, "] at 0x%lx\n", vmexit->rip);
 535                 return (VMEXIT_ABORT);
 536         }
 537
 538         return (VMEXIT_CONTINUE);
 539 }
 540
 541 static pthread_mutex_t resetcpu_mtx = PTHREAD_MUTEX_INITIALIZER;
 542 static pthread_cond_t resetcpu_cond = PTHREAD_COND_INITIALIZER;
 543
 544 static int
 545 vmexit_suspend(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
 546 {
 547         enum vm_suspend_how how;
 548
 549         how = vmexit->u.suspended.how;
 550
 551         fbsdrun_deletecpu(ctx, *pvcpu);
 552
 553         if (*pvcpu != BSP) {
 554                 pthread_mutex_lock(&resetcpu_mtx);
 555                 pthread_cond_signal(&resetcpu_cond);
 556                 pthread_mutex_unlock(&resetcpu_mtx);
 557                 pthread_exit(NULL);
 558         }
 559
 560         pthread_mutex_lock(&resetcpu_mtx);
 561         while (!CPU_EMPTY(&cpumask)) {
 562                 pthread_cond_wait(&resetcpu_cond, &resetcpu_mtx);
 563         }
 564         pthread_mutex_unlock(&resetcpu_mtx);
 565
 566         switch (how) {
 567         case VM_SUSPEND_RESET:
 568                 exit(0);
 569         case VM_SUSPEND_POWEROFF:
 570                 exit(1);
 571         case VM_SUSPEND_HALT:
 572                 exit(2);
 573         case VM_SUSPEND_TRIPLEFAULT:
 574                 exit(3);
 575         default:
 576                 fprintf(stderr, "vmexit_suspend: invalid reason %d\n", how);
 577                 exit(100);
 578         }
 579         return (0);     /* NOTREACHED */
 580 }
 581
 582 static vmexit_handler_t handler[VM_EXITCODE_MAX] = {
 583         [VM_EXITCODE_INOUT]  = vmexit_inout,
 584         [VM_EXITCODE_INOUT_STR]  = vmexit_inout,
 585         [VM_EXITCODE_VMX]    = vmexit_vmx,
 586         [VM_EXITCODE_SVM]    = vmexit_svm,
 587         [VM_EXITCODE_BOGUS]  = vmexit_bogus,
 588         [VM_EXITCODE_REQIDLE] = vmexit_reqidle,
 589         [VM_EXITCODE_RDMSR]  = vmexit_rdmsr,
 590         [VM_EXITCODE_WRMSR]  = vmexit_wrmsr,
 591         [VM_EXITCODE_MTRAP]  = vmexit_mtrap,
 592         [VM_EXITCODE_INST_EMUL] = vmexit_inst_emul,
 593         [VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap,
 594         [VM_EXITCODE_SUSPENDED] = vmexit_suspend,
 595         [VM_EXITCODE_TASK_SWITCH] = vmexit_task_switch,
 596 };
 597
 598 static void
 599 vm_loop(struct vmctx *ctx, int vcpu, uint64_t startrip)
 600 {
 601         int error, rc;
 602         enum vm_exitcode exitcode;
 603         cpuset_t active_cpus;
 604
 605         if (vcpumap[vcpu] != NULL) {
 606                 error = pthread_setaffinity_np(pthread_self(),
 607                     sizeof(cpuset_t), vcpumap[vcpu]);
 608                 assert(error == 0);
 609         }
 610
 611         error = vm_active_cpus(ctx, &active_cpus);
 612         assert(CPU_ISSET(vcpu, &active_cpus));
 613
 614         error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RIP, startrip);
 615         assert(error == 0);
 616
 617         while (1) {
 618                 error = vm_run(ctx, vcpu, &vmexit[vcpu]);
 619                 if (error != 0)
 620                         break;
 621
 622                 exitcode = vmexit[vcpu].exitcode;
 623                 if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) {
 624                         fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n",
 625                             exitcode);
 626                         exit(1);
 627                 }
 628
 629                 rc = (*handler[exitcode])(ctx, &vmexit[vcpu], &vcpu);
 630
 631                 switch (rc) {
 632                 case VMEXIT_CONTINUE:
 633                         break;
 634                 case VMEXIT_ABORT:
 635                         abort();
 636                 default:
 637                         exit(1);
 638                 }
 639         }
 640         fprintf(stderr, "vm_run error %d, errno %d\n", error, errno);
 641 }
 642
 643 static int
 644 num_vcpus_allowed(struct vmctx *ctx)
 645 {
 646         int tmp, error;
 647
 648         error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp);
 649
 650         /*
 651          * The guest is allowed to spinup more than one processor only if the
 652          * UNRESTRICTED_GUEST capability is available.
 653          */
 654         if (error == 0)
 655                 return (VM_MAXCPU);
 656         else
 657                 return (1);
 658 }
 659
 660 void
 661 fbsdrun_set_capabilities(struct vmctx *ctx, int cpu)
 662 {
 663         int err, tmp;
 664
 665         if (fbsdrun_vmexit_on_hlt()) {
 666                 err = vm_get_capability(ctx, cpu, VM_CAP_HALT_EXIT, &tmp);
 667                 if (err < 0) {
 668                         fprintf(stderr, "VM exit on HLT not supported\n");
 669                         exit(1);
 670                 }
 671                 vm_set_capability(ctx, cpu, VM_CAP_HALT_EXIT, 1);
 672                 if (cpu == BSP)
 673                         handler[VM_EXITCODE_HLT] = vmexit_hlt;
 674         }
 675
 676         if (fbsdrun_vmexit_on_pause()) {
 677                 /*
 678                  * pause exit support required for this mode
 679                  */
 680                 err = vm_get_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, &tmp);
 681                 if (err < 0) {
 682                         fprintf(stderr,
 683                             "SMP mux requested, no pause support\n");
 684                         exit(1);
 685                 }
 686                 vm_set_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, 1);
 687                 if (cpu == BSP)
 688                         handler[VM_EXITCODE_PAUSE] = vmexit_pause;
 689         }
 690
 691         if (x2apic_mode)
 692                 err = vm_set_x2apic_state(ctx, cpu, X2APIC_ENABLED);
 693         else
 694                 err = vm_set_x2apic_state(ctx, cpu, X2APIC_DISABLED);
 695
 696         if (err) {
 697                 fprintf(stderr, "Unable to set x2apic state (%d)\n", err);
 698                 exit(1);
 699         }
 700
 701         vm_set_capability(ctx, cpu, VM_CAP_ENABLE_INVPCID, 1);
 702 }
 703
 704 static struct vmctx *
 705 do_open(const char *vmname)
 706 {
 707         struct vmctx *ctx;
 708         int error;
 709         bool reinit, romboot;
 710
 711         reinit = romboot = false;
 712
 713         if (lpc_bootrom())
 714                 romboot = true;
 715
 716         error = vm_create(vmname);
 717         if (error) {
 718                 if (errno == EEXIST) {
 719                         if (romboot) {
 720                                 reinit = true;
 721                         } else {
 722                                 /*
 723                                  * The virtual machine has been setup by the
 724                                  * userspace bootloader.
 725                                  */
 726                         }
 727                 } else {
 728                         perror("vm_create");
 729                         exit(1);
 730                 }
 731         } else {
 732                 if (!romboot) {
 733                         /*
 734                          * If the virtual machine was just created then a
 735                          * bootrom must be configured to boot it.
 736                          */
 737                         fprintf(stderr, "virtual machine cannot be booted\n");
 738                         exit(1);
 739                 }
 740         }
 741
 742         ctx = vm_open(vmname);
 743         if (ctx == NULL) {
 744                 perror("vm_open");
 745                 exit(1);
 746         }
 747
 748         if (reinit) {
 749                 error = vm_reinit(ctx);
 750                 if (error) {
 751                         perror("vm_reinit");
 752                         exit(1);
 753                 }
 754         }
 755         return (ctx);
 756 }
 757
 758 int
 759 main(int argc, char *argv[])
 760 {
 761         int c, error, gdb_port, err, bvmcons;
 762         int max_vcpus, mptgen, memflags;
 763         int rtc_localtime;
 764         struct vmctx *ctx;
 765         uint64_t rip;
 766         size_t memsize;
 767         char *optstr;
 768
 769         bvmcons = 0;
 770         progname = basename(argv[0]);
 771         gdb_port = 0;
 772         guest_ncpus = 1;
 773         memsize = 256 * MB;
 774         mptgen = 1;
 775         rtc_localtime = 1;
 776         memflags = 0;
 777
 778         optstr = "abehuwxACHIPSWYp:g:c:s:m:l:U:";
 779         while ((c = getopt(argc, argv, optstr)) != -1) {
 780                 switch (c) {
 781                 case 'a':
 782                         x2apic_mode = 0;
 783                         break;
 784                 case 'A':
 785                         acpi = 1;
 786                         break;
 787                 case 'b':
 788                         bvmcons = 1;
 789                         break;
 790                 case 'p':
 791                         if (pincpu_parse(optarg) != 0) {
 792                             errx(EX_USAGE, "invalid vcpu pinning "
 793                                  "configuration '%s'", optarg);
 794                         }
 795                         break;
 796                 case 'c':
 797                         guest_ncpus = atoi(optarg);
 798                         break;
 799                 case 'C':
 800                         memflags |= VM_MEM_F_INCORE;
 801                         break;
 802                 case 'g':
 803                         gdb_port = atoi(optarg);
 804                         break;
 805                 case 'l':
 806                         if (lpc_device_parse(optarg) != 0) {
 807                                 errx(EX_USAGE, "invalid lpc device "
 808                                     "configuration '%s'", optarg);
 809                         }
 810                         break;
 811                 case 's':
 812                         if (pci_parse_slot(optarg) != 0)
 813                                 exit(1);
 814                         else
 815                                 break;
 816                 case 'S':
 817                         memflags |= VM_MEM_F_WIRED;
 818                         break;
 819                 case 'm':
 820                         error = vm_parse_memsize(optarg, &memsize);
 821                         if (error)
 822                                 errx(EX_USAGE, "invalid memsize '%s'", optarg);
 823                         break;
 824                 case 'H':
 825                         guest_vmexit_on_hlt = 1;
 826                         break;
 827                 case 'I':
 828                         /*
 829                          * The "-I" option was used to add an ioapic to the
 830                          * virtual machine.
 831                          *
 832                          * An ioapic is now provided unconditionally for each
 833                          * virtual machine and this option is now deprecated.
 834                          */
 835                         break;
 836                 case 'P':
 837                         guest_vmexit_on_pause = 1;
 838                         break;
 839                 case 'e':
 840                         strictio = 1;
 841                         break;
 842                 case 'u':
 843                         rtc_localtime = 0;
 844                         break;
 845                 case 'U':
 846                         guest_uuid_str = optarg;
 847                         break;
 848                 case 'w':
 849                         strictmsr = 0;
 850                         break;
 851                 case 'W':
 852                         virtio_msix = 0;
 853                         break;
 854                 case 'x':
 855                         x2apic_mode = 1;
 856                         break;
 857                 case 'Y':
 858                         mptgen = 0;
 859                         break;
 860                 case 'h':
 861                         usage(0);
 862                 default:
 863                         usage(1);
 864                 }
 865         }
 866         argc -= optind;
 867         argv += optind;
 868
 869         if (argc != 1)
 870                 usage(1);
 871
 872         vmname = argv[0];
 873         ctx = do_open(vmname);
 874
 875         if (guest_ncpus < 1) {
 876                 fprintf(stderr, "Invalid guest vCPUs (%d)\n", guest_ncpus);
 877                 exit(1);
 878         }
 879
 880         max_vcpus = num_vcpus_allowed(ctx);
 881         if (guest_ncpus > max_vcpus) {
 882                 fprintf(stderr, "%d vCPUs requested but only %d available\n",
 883                         guest_ncpus, max_vcpus);
 884                 exit(1);
 885         }
 886
 887         fbsdrun_set_capabilities(ctx, BSP);
 888
 889         vm_set_memflags(ctx, memflags);
 890         err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL);
 891         if (err) {
 892                 fprintf(stderr, "Unable to setup memory (%d)\n", errno);
 893                 exit(1);
 894         }
 895
 896         error = init_msr();
 897         if (error) {
 898                 fprintf(stderr, "init_msr error %d", error);
 899                 exit(1);
 900         }
 901
 902         init_mem();
 903         init_inout();
 904         pci_irq_init(ctx);
 905         ioapic_init(ctx);
 906
 907         rtc_init(ctx, rtc_localtime);
 908         sci_init(ctx);
 909
 910         /*
 911          * Exit if a device emulation finds an error in it's initilization
 912          */
 913         if (init_pci(ctx) != 0)
 914                 exit(1);
 915
 916         if (gdb_port != 0)
 917                 init_dbgport(gdb_port);
 918
 919         if (bvmcons)
 920                 init_bvmcons();
 921
 922         if (lpc_bootrom()) {
 923                 if (vm_set_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, 1)) {
 924                         fprintf(stderr, "ROM boot failed: unrestricted guest "
 925                             "capability not available\n");
 926                         exit(1);
 927                 }
 928                 error = vcpu_reset(ctx, BSP);
 929                 assert(error == 0);
 930         }
 931
 932         error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip);
 933         assert(error == 0);
 934
 935         /*
 936          * build the guest tables, MP etc.
 937          */
 938         if (mptgen) {
 939                 error = mptable_build(ctx, guest_ncpus);
 940                 if (error)
 941                         exit(1);
 942         }
 943
 944         error = smbios_build(ctx);
 945         assert(error == 0);
 946
 947         if (acpi) {
 948                 error = acpi_build(ctx, guest_ncpus);
 949                 assert(error == 0);
 950         }
 951
 952         if (lpc_bootrom())
 953                 fwctl_init();
 954
 955         /*
 956          * Change the proc title to include the VM name.
 957          */
 958         setproctitle("%s", vmname);
 959
 960         /*
 961          * Add CPU 0
 962          */
 963         fbsdrun_addcpu(ctx, BSP, BSP, rip);
 964
 965         /*
 966          * Head off to the main event dispatch loop
 967          */
 968         mevent_dispatch();
 969
 970         exit(1);
 971 }