vmmapi.c revision 268976
1169691Skan/*- 2169691Skan * Copyright (c) 2011 NetApp, Inc. 3169691Skan * All rights reserved. 4169691Skan * 5169691Skan * Redistribution and use in source and binary forms, with or without 6169691Skan * modification, are permitted provided that the following conditions 7169691Skan * are met: 8169691Skan * 1. Redistributions of source code must retain the above copyright 9169691Skan * notice, this list of conditions and the following disclaimer. 10169691Skan * 2. Redistributions in binary form must reproduce the above copyright 11169691Skan * notice, this list of conditions and the following disclaimer in the 12169691Skan * documentation and/or other materials provided with the distribution. 13169691Skan * 14169691Skan * THIS SOFTWARE IS PROVIDED BY NETAPP, INC ``AS IS'' AND 15169691Skan * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 16169691Skan * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 17169691Skan * ARE DISCLAIMED. IN NO EVENT SHALL NETAPP, INC OR CONTRIBUTORS BE LIABLE 18169691Skan * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 19169691Skan * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 20169691Skan * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 21169691Skan * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 22169691Skan * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 23169691Skan * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 24169691Skan * SUCH DAMAGE. 25169691Skan * 26169691Skan * $FreeBSD: stable/10/lib/libvmmapi/vmmapi.c 268976 2014-07-22 04:39:16Z jhb $ 27169691Skan */ 28169691Skan 29169691Skan#include <sys/cdefs.h> 30169691Skan__FBSDID("$FreeBSD: stable/10/lib/libvmmapi/vmmapi.c 268976 2014-07-22 04:39:16Z jhb $"); 31169691Skan 32169691Skan#include <sys/types.h> 33169691Skan#include <sys/sysctl.h> 34169691Skan#include <sys/ioctl.h> 35169691Skan#include <sys/mman.h> 36169691Skan#include <sys/_iovec.h> 37169691Skan 38169691Skan#include <machine/specialreg.h> 39169691Skan#include <machine/param.h> 40169691Skan 41169691Skan#include <stdio.h> 42169691Skan#include <stdlib.h> 43169691Skan#include <assert.h> 44169691Skan#include <string.h> 45169691Skan#include <fcntl.h> 46169691Skan#include <unistd.h> 47169691Skan 48169691Skan#include <libutil.h> 49169691Skan 50169691Skan#include <machine/vmm.h> 51169691Skan#include <machine/vmm_dev.h> 52169691Skan 53169691Skan#include "vmmapi.h" 54169691Skan 55169691Skan#define MB (1024 * 1024UL) 56169691Skan#define GB (1024 * 1024 * 1024UL) 57169691Skan 58169691Skanstruct vmctx { 59169691Skan int fd; 60169691Skan uint32_t lowmem_limit; 61169691Skan enum vm_mmap_style vms; 62169691Skan int memflags; 63169691Skan size_t lowmem; 64169691Skan char *lowmem_addr; 65169691Skan size_t highmem; 66169691Skan char *highmem_addr; 67169691Skan char *name; 68169691Skan}; 69169691Skan 70169691Skan#define CREATE(x) sysctlbyname("hw.vmm.create", NULL, NULL, (x), strlen((x))) 71169691Skan#define DESTROY(x) sysctlbyname("hw.vmm.destroy", NULL, NULL, (x), strlen((x))) 72169691Skan 73169691Skanstatic int 74169691Skanvm_device_open(const char *name) 75169691Skan{ 76169691Skan int fd, len; 77169691Skan char *vmfile; 78169691Skan 79169691Skan len = strlen("/dev/vmm/") + strlen(name) + 1; 80169691Skan vmfile = malloc(len); 81169691Skan assert(vmfile != NULL); 82169691Skan snprintf(vmfile, len, "/dev/vmm/%s", name); 83169691Skan 84169691Skan /* Open the device file */ 85169691Skan fd = open(vmfile, O_RDWR, 0); 86169691Skan 87169691Skan free(vmfile); 88169691Skan return (fd); 89169691Skan} 90169691Skan 91169691Skanint 92169691Skanvm_create(const char *name) 93169691Skan{ 94169691Skan 95169691Skan return (CREATE((char *)name)); 96169691Skan} 97169691Skan 98169691Skanstruct vmctx * 99169691Skanvm_open(const char *name) 100169691Skan{ 101169691Skan struct vmctx *vm; 102169691Skan 103169691Skan vm = malloc(sizeof(struct vmctx) + strlen(name) + 1); 104169691Skan assert(vm != NULL); 105169691Skan 106169691Skan vm->fd = -1; 107169691Skan vm->memflags = 0; 108169691Skan vm->lowmem_limit = 3 * GB; 109169691Skan vm->name = (char *)(vm + 1); 110169691Skan strcpy(vm->name, name); 111169691Skan 112169691Skan if ((vm->fd = vm_device_open(vm->name)) < 0) 113169691Skan goto err; 114169691Skan 115169691Skan return (vm); 116169691Skanerr: 117169691Skan vm_destroy(vm); 118169691Skan return (NULL); 119169691Skan} 120169691Skan 121169691Skanvoid 122169691Skanvm_destroy(struct vmctx *vm) 123169691Skan{ 124169691Skan assert(vm != NULL); 125169691Skan 126169691Skan if (vm->fd >= 0) 127169691Skan close(vm->fd); 128169691Skan DESTROY(vm->name); 129169691Skan 130169691Skan free(vm); 131169691Skan} 132169691Skan 133169691Skanint 134169691Skanvm_parse_memsize(const char *optarg, size_t *ret_memsize) 135169691Skan{ 136169691Skan char *endptr; 137169691Skan size_t optval; 138169691Skan int error; 139169691Skan 140169691Skan optval = strtoul(optarg, &endptr, 0); 141169691Skan if (*optarg != '\0' && *endptr == '\0') { 142169691Skan /* 143169691Skan * For the sake of backward compatibility if the memory size 144169691Skan * specified on the command line is less than a megabyte then 145169691Skan * it is interpreted as being in units of MB. 146169691Skan */ 147169691Skan if (optval < MB) 148169691Skan optval *= MB; 149169691Skan *ret_memsize = optval; 150169691Skan error = 0; 151169691Skan } else 152169691Skan error = expand_number(optarg, ret_memsize); 153169691Skan 154169691Skan return (error); 155169691Skan} 156169691Skan 157169691Skanint 158169691Skanvm_get_memory_seg(struct vmctx *ctx, vm_paddr_t gpa, size_t *ret_len, 159169691Skan int *wired) 160169691Skan{ 161169691Skan int error; 162169691Skan struct vm_memory_segment seg; 163169691Skan 164169691Skan bzero(&seg, sizeof(seg)); 165169691Skan seg.gpa = gpa; 166169691Skan error = ioctl(ctx->fd, VM_GET_MEMORY_SEG, &seg); 167169691Skan *ret_len = seg.len; 168169691Skan if (wired != NULL) 169169691Skan *wired = seg.wired; 170169691Skan return (error); 171169691Skan} 172169691Skan 173169691Skanuint32_t 174169691Skanvm_get_lowmem_limit(struct vmctx *ctx) 175169691Skan{ 176169691Skan 177169691Skan return (ctx->lowmem_limit); 178169691Skan} 179169691Skan 180169691Skanvoid 181169691Skanvm_set_lowmem_limit(struct vmctx *ctx, uint32_t limit) 182169691Skan{ 183169691Skan 184169691Skan ctx->lowmem_limit = limit; 185169691Skan} 186169691Skan 187169691Skanvoid 188169691Skanvm_set_memflags(struct vmctx *ctx, int flags) 189169691Skan{ 190169691Skan 191169691Skan ctx->memflags = flags; 192169691Skan} 193169691Skan 194169691Skanstatic int 195169691Skansetup_memory_segment(struct vmctx *ctx, vm_paddr_t gpa, size_t len, char **addr) 196169691Skan{ 197169691Skan int error, mmap_flags; 198169691Skan struct vm_memory_segment seg; 199169691Skan 200169691Skan /* 201169691Skan * Create and optionally map 'len' bytes of memory at guest 202169691Skan * physical address 'gpa' 203169691Skan */ 204169691Skan bzero(&seg, sizeof(seg)); 205169691Skan seg.gpa = gpa; 206169691Skan seg.len = len; 207169691Skan error = ioctl(ctx->fd, VM_MAP_MEMORY, &seg); 208169691Skan if (error == 0 && addr != NULL) { 209169691Skan mmap_flags = MAP_SHARED; 210169691Skan if ((ctx->memflags & VM_MEM_F_INCORE) == 0) 211169691Skan mmap_flags |= MAP_NOCORE; 212169691Skan *addr = mmap(NULL, len, PROT_READ | PROT_WRITE, mmap_flags, 213169691Skan ctx->fd, gpa); 214169691Skan } 215169691Skan return (error); 216169691Skan} 217169691Skan 218169691Skanint 219169691Skanvm_setup_memory(struct vmctx *ctx, size_t memsize, enum vm_mmap_style vms) 220169691Skan{ 221169691Skan char **addr; 222169691Skan int error; 223169691Skan 224169691Skan /* XXX VM_MMAP_SPARSE not implemented yet */ 225169691Skan assert(vms == VM_MMAP_NONE || vms == VM_MMAP_ALL); 226169691Skan ctx->vms = vms; 227169691Skan 228169691Skan /* 229169691Skan * If 'memsize' cannot fit entirely in the 'lowmem' segment then 230169691Skan * create another 'highmem' segment above 4GB for the remainder. 231169691Skan */ 232169691Skan if (memsize > ctx->lowmem_limit) { 233169691Skan ctx->lowmem = ctx->lowmem_limit; 234169691Skan ctx->highmem = memsize - ctx->lowmem; 235169691Skan } else { 236169691Skan ctx->lowmem = memsize; 237169691Skan ctx->highmem = 0; 238169691Skan } 239169691Skan 240169691Skan if (ctx->lowmem > 0) { 241169691Skan addr = (vms == VM_MMAP_ALL) ? &ctx->lowmem_addr : NULL; 242169691Skan error = setup_memory_segment(ctx, 0, ctx->lowmem, addr); 243169691Skan if (error) 244169691Skan return (error); 245169691Skan } 246169691Skan 247169691Skan if (ctx->highmem > 0) { 248169691Skan addr = (vms == VM_MMAP_ALL) ? &ctx->highmem_addr : NULL; 249169691Skan error = setup_memory_segment(ctx, 4*GB, ctx->highmem, addr); 250169691Skan if (error) 251169691Skan return (error); 252169691Skan } 253169691Skan 254169691Skan return (0); 255169691Skan} 256169691Skan 257169691Skanvoid * 258169691Skanvm_map_gpa(struct vmctx *ctx, vm_paddr_t gaddr, size_t len) 259169691Skan{ 260169691Skan 261 /* XXX VM_MMAP_SPARSE not implemented yet */ 262 assert(ctx->vms == VM_MMAP_ALL); 263 264 if (gaddr < ctx->lowmem && gaddr + len <= ctx->lowmem) 265 return ((void *)(ctx->lowmem_addr + gaddr)); 266 267 if (gaddr >= 4*GB) { 268 gaddr -= 4*GB; 269 if (gaddr < ctx->highmem && gaddr + len <= ctx->highmem) 270 return ((void *)(ctx->highmem_addr + gaddr)); 271 } 272 273 return (NULL); 274} 275 276int 277vm_set_desc(struct vmctx *ctx, int vcpu, int reg, 278 uint64_t base, uint32_t limit, uint32_t access) 279{ 280 int error; 281 struct vm_seg_desc vmsegdesc; 282 283 bzero(&vmsegdesc, sizeof(vmsegdesc)); 284 vmsegdesc.cpuid = vcpu; 285 vmsegdesc.regnum = reg; 286 vmsegdesc.desc.base = base; 287 vmsegdesc.desc.limit = limit; 288 vmsegdesc.desc.access = access; 289 290 error = ioctl(ctx->fd, VM_SET_SEGMENT_DESCRIPTOR, &vmsegdesc); 291 return (error); 292} 293 294int 295vm_get_desc(struct vmctx *ctx, int vcpu, int reg, 296 uint64_t *base, uint32_t *limit, uint32_t *access) 297{ 298 int error; 299 struct vm_seg_desc vmsegdesc; 300 301 bzero(&vmsegdesc, sizeof(vmsegdesc)); 302 vmsegdesc.cpuid = vcpu; 303 vmsegdesc.regnum = reg; 304 305 error = ioctl(ctx->fd, VM_GET_SEGMENT_DESCRIPTOR, &vmsegdesc); 306 if (error == 0) { 307 *base = vmsegdesc.desc.base; 308 *limit = vmsegdesc.desc.limit; 309 *access = vmsegdesc.desc.access; 310 } 311 return (error); 312} 313 314int 315vm_set_register(struct vmctx *ctx, int vcpu, int reg, uint64_t val) 316{ 317 int error; 318 struct vm_register vmreg; 319 320 bzero(&vmreg, sizeof(vmreg)); 321 vmreg.cpuid = vcpu; 322 vmreg.regnum = reg; 323 vmreg.regval = val; 324 325 error = ioctl(ctx->fd, VM_SET_REGISTER, &vmreg); 326 return (error); 327} 328 329int 330vm_get_register(struct vmctx *ctx, int vcpu, int reg, uint64_t *ret_val) 331{ 332 int error; 333 struct vm_register vmreg; 334 335 bzero(&vmreg, sizeof(vmreg)); 336 vmreg.cpuid = vcpu; 337 vmreg.regnum = reg; 338 339 error = ioctl(ctx->fd, VM_GET_REGISTER, &vmreg); 340 *ret_val = vmreg.regval; 341 return (error); 342} 343 344int 345vm_run(struct vmctx *ctx, int vcpu, uint64_t rip, struct vm_exit *vmexit) 346{ 347 int error; 348 struct vm_run vmrun; 349 350 bzero(&vmrun, sizeof(vmrun)); 351 vmrun.cpuid = vcpu; 352 vmrun.rip = rip; 353 354 error = ioctl(ctx->fd, VM_RUN, &vmrun); 355 bcopy(&vmrun.vm_exit, vmexit, sizeof(struct vm_exit)); 356 return (error); 357} 358 359int 360vm_suspend(struct vmctx *ctx, enum vm_suspend_how how) 361{ 362 struct vm_suspend vmsuspend; 363 364 bzero(&vmsuspend, sizeof(vmsuspend)); 365 vmsuspend.how = how; 366 return (ioctl(ctx->fd, VM_SUSPEND, &vmsuspend)); 367} 368 369static int 370vm_inject_exception_real(struct vmctx *ctx, int vcpu, int vector, 371 int error_code, int error_code_valid) 372{ 373 struct vm_exception exc; 374 375 bzero(&exc, sizeof(exc)); 376 exc.cpuid = vcpu; 377 exc.vector = vector; 378 exc.error_code = error_code; 379 exc.error_code_valid = error_code_valid; 380 381 return (ioctl(ctx->fd, VM_INJECT_EXCEPTION, &exc)); 382} 383 384int 385vm_inject_exception(struct vmctx *ctx, int vcpu, int vector) 386{ 387 388 return (vm_inject_exception_real(ctx, vcpu, vector, 0, 0)); 389} 390 391int 392vm_inject_exception2(struct vmctx *ctx, int vcpu, int vector, int errcode) 393{ 394 395 return (vm_inject_exception_real(ctx, vcpu, vector, errcode, 1)); 396} 397 398int 399vm_apicid2vcpu(struct vmctx *ctx, int apicid) 400{ 401 /* 402 * The apic id associated with the 'vcpu' has the same numerical value 403 * as the 'vcpu' itself. 404 */ 405 return (apicid); 406} 407 408int 409vm_lapic_irq(struct vmctx *ctx, int vcpu, int vector) 410{ 411 struct vm_lapic_irq vmirq; 412 413 bzero(&vmirq, sizeof(vmirq)); 414 vmirq.cpuid = vcpu; 415 vmirq.vector = vector; 416 417 return (ioctl(ctx->fd, VM_LAPIC_IRQ, &vmirq)); 418} 419 420int 421vm_lapic_local_irq(struct vmctx *ctx, int vcpu, int vector) 422{ 423 struct vm_lapic_irq vmirq; 424 425 bzero(&vmirq, sizeof(vmirq)); 426 vmirq.cpuid = vcpu; 427 vmirq.vector = vector; 428 429 return (ioctl(ctx->fd, VM_LAPIC_LOCAL_IRQ, &vmirq)); 430} 431 432int 433vm_lapic_msi(struct vmctx *ctx, uint64_t addr, uint64_t msg) 434{ 435 struct vm_lapic_msi vmmsi; 436 437 bzero(&vmmsi, sizeof(vmmsi)); 438 vmmsi.addr = addr; 439 vmmsi.msg = msg; 440 441 return (ioctl(ctx->fd, VM_LAPIC_MSI, &vmmsi)); 442} 443 444int 445vm_ioapic_assert_irq(struct vmctx *ctx, int irq) 446{ 447 struct vm_ioapic_irq ioapic_irq; 448 449 bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq)); 450 ioapic_irq.irq = irq; 451 452 return (ioctl(ctx->fd, VM_IOAPIC_ASSERT_IRQ, &ioapic_irq)); 453} 454 455int 456vm_ioapic_deassert_irq(struct vmctx *ctx, int irq) 457{ 458 struct vm_ioapic_irq ioapic_irq; 459 460 bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq)); 461 ioapic_irq.irq = irq; 462 463 return (ioctl(ctx->fd, VM_IOAPIC_DEASSERT_IRQ, &ioapic_irq)); 464} 465 466int 467vm_ioapic_pulse_irq(struct vmctx *ctx, int irq) 468{ 469 struct vm_ioapic_irq ioapic_irq; 470 471 bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq)); 472 ioapic_irq.irq = irq; 473 474 return (ioctl(ctx->fd, VM_IOAPIC_PULSE_IRQ, &ioapic_irq)); 475} 476 477int 478vm_ioapic_pincount(struct vmctx *ctx, int *pincount) 479{ 480 481 return (ioctl(ctx->fd, VM_IOAPIC_PINCOUNT, pincount)); 482} 483 484int 485vm_isa_assert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq) 486{ 487 struct vm_isa_irq isa_irq; 488 489 bzero(&isa_irq, sizeof(struct vm_isa_irq)); 490 isa_irq.atpic_irq = atpic_irq; 491 isa_irq.ioapic_irq = ioapic_irq; 492 493 return (ioctl(ctx->fd, VM_ISA_ASSERT_IRQ, &isa_irq)); 494} 495 496int 497vm_isa_deassert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq) 498{ 499 struct vm_isa_irq isa_irq; 500 501 bzero(&isa_irq, sizeof(struct vm_isa_irq)); 502 isa_irq.atpic_irq = atpic_irq; 503 isa_irq.ioapic_irq = ioapic_irq; 504 505 return (ioctl(ctx->fd, VM_ISA_DEASSERT_IRQ, &isa_irq)); 506} 507 508int 509vm_isa_pulse_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq) 510{ 511 struct vm_isa_irq isa_irq; 512 513 bzero(&isa_irq, sizeof(struct vm_isa_irq)); 514 isa_irq.atpic_irq = atpic_irq; 515 isa_irq.ioapic_irq = ioapic_irq; 516 517 return (ioctl(ctx->fd, VM_ISA_PULSE_IRQ, &isa_irq)); 518} 519 520int 521vm_isa_set_irq_trigger(struct vmctx *ctx, int atpic_irq, 522 enum vm_intr_trigger trigger) 523{ 524 struct vm_isa_irq_trigger isa_irq_trigger; 525 526 bzero(&isa_irq_trigger, sizeof(struct vm_isa_irq_trigger)); 527 isa_irq_trigger.atpic_irq = atpic_irq; 528 isa_irq_trigger.trigger = trigger; 529 530 return (ioctl(ctx->fd, VM_ISA_SET_IRQ_TRIGGER, &isa_irq_trigger)); 531} 532 533int 534vm_inject_nmi(struct vmctx *ctx, int vcpu) 535{ 536 struct vm_nmi vmnmi; 537 538 bzero(&vmnmi, sizeof(vmnmi)); 539 vmnmi.cpuid = vcpu; 540 541 return (ioctl(ctx->fd, VM_INJECT_NMI, &vmnmi)); 542} 543 544static struct { 545 const char *name; 546 int type; 547} capstrmap[] = { 548 { "hlt_exit", VM_CAP_HALT_EXIT }, 549 { "mtrap_exit", VM_CAP_MTRAP_EXIT }, 550 { "pause_exit", VM_CAP_PAUSE_EXIT }, 551 { "unrestricted_guest", VM_CAP_UNRESTRICTED_GUEST }, 552 { "enable_invpcid", VM_CAP_ENABLE_INVPCID }, 553 { 0 } 554}; 555 556int 557vm_capability_name2type(const char *capname) 558{ 559 int i; 560 561 for (i = 0; capstrmap[i].name != NULL && capname != NULL; i++) { 562 if (strcmp(capstrmap[i].name, capname) == 0) 563 return (capstrmap[i].type); 564 } 565 566 return (-1); 567} 568 569const char * 570vm_capability_type2name(int type) 571{ 572 int i; 573 574 for (i = 0; capstrmap[i].name != NULL; i++) { 575 if (capstrmap[i].type == type) 576 return (capstrmap[i].name); 577 } 578 579 return (NULL); 580} 581 582int 583vm_get_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap, 584 int *retval) 585{ 586 int error; 587 struct vm_capability vmcap; 588 589 bzero(&vmcap, sizeof(vmcap)); 590 vmcap.cpuid = vcpu; 591 vmcap.captype = cap; 592 593 error = ioctl(ctx->fd, VM_GET_CAPABILITY, &vmcap); 594 *retval = vmcap.capval; 595 return (error); 596} 597 598int 599vm_set_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap, int val) 600{ 601 struct vm_capability vmcap; 602 603 bzero(&vmcap, sizeof(vmcap)); 604 vmcap.cpuid = vcpu; 605 vmcap.captype = cap; 606 vmcap.capval = val; 607 608 return (ioctl(ctx->fd, VM_SET_CAPABILITY, &vmcap)); 609} 610 611int 612vm_assign_pptdev(struct vmctx *ctx, int bus, int slot, int func) 613{ 614 struct vm_pptdev pptdev; 615 616 bzero(&pptdev, sizeof(pptdev)); 617 pptdev.bus = bus; 618 pptdev.slot = slot; 619 pptdev.func = func; 620 621 return (ioctl(ctx->fd, VM_BIND_PPTDEV, &pptdev)); 622} 623 624int 625vm_unassign_pptdev(struct vmctx *ctx, int bus, int slot, int func) 626{ 627 struct vm_pptdev pptdev; 628 629 bzero(&pptdev, sizeof(pptdev)); 630 pptdev.bus = bus; 631 pptdev.slot = slot; 632 pptdev.func = func; 633 634 return (ioctl(ctx->fd, VM_UNBIND_PPTDEV, &pptdev)); 635} 636 637int 638vm_map_pptdev_mmio(struct vmctx *ctx, int bus, int slot, int func, 639 vm_paddr_t gpa, size_t len, vm_paddr_t hpa) 640{ 641 struct vm_pptdev_mmio pptmmio; 642 643 bzero(&pptmmio, sizeof(pptmmio)); 644 pptmmio.bus = bus; 645 pptmmio.slot = slot; 646 pptmmio.func = func; 647 pptmmio.gpa = gpa; 648 pptmmio.len = len; 649 pptmmio.hpa = hpa; 650 651 return (ioctl(ctx->fd, VM_MAP_PPTDEV_MMIO, &pptmmio)); 652} 653 654int 655vm_setup_pptdev_msi(struct vmctx *ctx, int vcpu, int bus, int slot, int func, 656 uint64_t addr, uint64_t msg, int numvec) 657{ 658 struct vm_pptdev_msi pptmsi; 659 660 bzero(&pptmsi, sizeof(pptmsi)); 661 pptmsi.vcpu = vcpu; 662 pptmsi.bus = bus; 663 pptmsi.slot = slot; 664 pptmsi.func = func; 665 pptmsi.msg = msg; 666 pptmsi.addr = addr; 667 pptmsi.numvec = numvec; 668 669 return (ioctl(ctx->fd, VM_PPTDEV_MSI, &pptmsi)); 670} 671 672int 673vm_setup_pptdev_msix(struct vmctx *ctx, int vcpu, int bus, int slot, int func, 674 int idx, uint64_t addr, uint64_t msg, uint32_t vector_control) 675{ 676 struct vm_pptdev_msix pptmsix; 677 678 bzero(&pptmsix, sizeof(pptmsix)); 679 pptmsix.vcpu = vcpu; 680 pptmsix.bus = bus; 681 pptmsix.slot = slot; 682 pptmsix.func = func; 683 pptmsix.idx = idx; 684 pptmsix.msg = msg; 685 pptmsix.addr = addr; 686 pptmsix.vector_control = vector_control; 687 688 return ioctl(ctx->fd, VM_PPTDEV_MSIX, &pptmsix); 689} 690 691uint64_t * 692vm_get_stats(struct vmctx *ctx, int vcpu, struct timeval *ret_tv, 693 int *ret_entries) 694{ 695 int error; 696 697 static struct vm_stats vmstats; 698 699 vmstats.cpuid = vcpu; 700 701 error = ioctl(ctx->fd, VM_STATS, &vmstats); 702 if (error == 0) { 703 if (ret_entries) 704 *ret_entries = vmstats.num_entries; 705 if (ret_tv) 706 *ret_tv = vmstats.tv; 707 return (vmstats.statbuf); 708 } else 709 return (NULL); 710} 711 712const char * 713vm_get_stat_desc(struct vmctx *ctx, int index) 714{ 715 static struct vm_stat_desc statdesc; 716 717 statdesc.index = index; 718 if (ioctl(ctx->fd, VM_STAT_DESC, &statdesc) == 0) 719 return (statdesc.desc); 720 else 721 return (NULL); 722} 723 724int 725vm_get_x2apic_state(struct vmctx *ctx, int vcpu, enum x2apic_state *state) 726{ 727 int error; 728 struct vm_x2apic x2apic; 729 730 bzero(&x2apic, sizeof(x2apic)); 731 x2apic.cpuid = vcpu; 732 733 error = ioctl(ctx->fd, VM_GET_X2APIC_STATE, &x2apic); 734 *state = x2apic.state; 735 return (error); 736} 737 738int 739vm_set_x2apic_state(struct vmctx *ctx, int vcpu, enum x2apic_state state) 740{ 741 int error; 742 struct vm_x2apic x2apic; 743 744 bzero(&x2apic, sizeof(x2apic)); 745 x2apic.cpuid = vcpu; 746 x2apic.state = state; 747 748 error = ioctl(ctx->fd, VM_SET_X2APIC_STATE, &x2apic); 749 750 return (error); 751} 752 753/* 754 * From Intel Vol 3a: 755 * Table 9-1. IA-32 Processor States Following Power-up, Reset or INIT 756 */ 757int 758vcpu_reset(struct vmctx *vmctx, int vcpu) 759{ 760 int error; 761 uint64_t rflags, rip, cr0, cr4, zero, desc_base, rdx; 762 uint32_t desc_access, desc_limit; 763 uint16_t sel; 764 765 zero = 0; 766 767 rflags = 0x2; 768 error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RFLAGS, rflags); 769 if (error) 770 goto done; 771 772 rip = 0xfff0; 773 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RIP, rip)) != 0) 774 goto done; 775 776 cr0 = CR0_NE; 777 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR0, cr0)) != 0) 778 goto done; 779 780 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR3, zero)) != 0) 781 goto done; 782 783 cr4 = 0; 784 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR4, cr4)) != 0) 785 goto done; 786 787 /* 788 * CS: present, r/w, accessed, 16-bit, byte granularity, usable 789 */ 790 desc_base = 0xffff0000; 791 desc_limit = 0xffff; 792 desc_access = 0x0093; 793 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_CS, 794 desc_base, desc_limit, desc_access); 795 if (error) 796 goto done; 797 798 sel = 0xf000; 799 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CS, sel)) != 0) 800 goto done; 801 802 /* 803 * SS,DS,ES,FS,GS: present, r/w, accessed, 16-bit, byte granularity 804 */ 805 desc_base = 0; 806 desc_limit = 0xffff; 807 desc_access = 0x0093; 808 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_SS, 809 desc_base, desc_limit, desc_access); 810 if (error) 811 goto done; 812 813 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_DS, 814 desc_base, desc_limit, desc_access); 815 if (error) 816 goto done; 817 818 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_ES, 819 desc_base, desc_limit, desc_access); 820 if (error) 821 goto done; 822 823 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_FS, 824 desc_base, desc_limit, desc_access); 825 if (error) 826 goto done; 827 828 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GS, 829 desc_base, desc_limit, desc_access); 830 if (error) 831 goto done; 832 833 sel = 0; 834 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_SS, sel)) != 0) 835 goto done; 836 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_DS, sel)) != 0) 837 goto done; 838 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_ES, sel)) != 0) 839 goto done; 840 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_FS, sel)) != 0) 841 goto done; 842 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_GS, sel)) != 0) 843 goto done; 844 845 /* General purpose registers */ 846 rdx = 0xf00; 847 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RAX, zero)) != 0) 848 goto done; 849 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBX, zero)) != 0) 850 goto done; 851 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RCX, zero)) != 0) 852 goto done; 853 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDX, rdx)) != 0) 854 goto done; 855 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSI, zero)) != 0) 856 goto done; 857 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDI, zero)) != 0) 858 goto done; 859 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBP, zero)) != 0) 860 goto done; 861 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSP, zero)) != 0) 862 goto done; 863 864 /* GDTR, IDTR */ 865 desc_base = 0; 866 desc_limit = 0xffff; 867 desc_access = 0; 868 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GDTR, 869 desc_base, desc_limit, desc_access); 870 if (error != 0) 871 goto done; 872 873 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_IDTR, 874 desc_base, desc_limit, desc_access); 875 if (error != 0) 876 goto done; 877 878 /* TR */ 879 desc_base = 0; 880 desc_limit = 0xffff; 881 desc_access = 0x0000008b; 882 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_TR, 0, 0, desc_access); 883 if (error) 884 goto done; 885 886 sel = 0; 887 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_TR, sel)) != 0) 888 goto done; 889 890 /* LDTR */ 891 desc_base = 0; 892 desc_limit = 0xffff; 893 desc_access = 0x00000082; 894 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_LDTR, desc_base, 895 desc_limit, desc_access); 896 if (error) 897 goto done; 898 899 sel = 0; 900 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_LDTR, 0)) != 0) 901 goto done; 902 903 /* XXX cr2, debug registers */ 904 905 error = 0; 906done: 907 return (error); 908} 909 910int 911vm_get_gpa_pmap(struct vmctx *ctx, uint64_t gpa, uint64_t *pte, int *num) 912{ 913 int error, i; 914 struct vm_gpa_pte gpapte; 915 916 bzero(&gpapte, sizeof(gpapte)); 917 gpapte.gpa = gpa; 918 919 error = ioctl(ctx->fd, VM_GET_GPA_PMAP, &gpapte); 920 921 if (error == 0) { 922 *num = gpapte.ptenum; 923 for (i = 0; i < gpapte.ptenum; i++) 924 pte[i] = gpapte.pte[i]; 925 } 926 927 return (error); 928} 929 930int 931vm_get_hpet_capabilities(struct vmctx *ctx, uint32_t *capabilities) 932{ 933 int error; 934 struct vm_hpet_cap cap; 935 936 bzero(&cap, sizeof(struct vm_hpet_cap)); 937 error = ioctl(ctx->fd, VM_GET_HPET_CAPABILITIES, &cap); 938 if (capabilities != NULL) 939 *capabilities = cap.capabilities; 940 return (error); 941} 942 943static int 944gla2gpa(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging, 945 uint64_t gla, int prot, int *fault, uint64_t *gpa) 946{ 947 struct vm_gla2gpa gg; 948 int error; 949 950 bzero(&gg, sizeof(struct vm_gla2gpa)); 951 gg.vcpuid = vcpu; 952 gg.prot = prot; 953 gg.gla = gla; 954 gg.paging = *paging; 955 956 error = ioctl(ctx->fd, VM_GLA2GPA, &gg); 957 if (error == 0) { 958 *fault = gg.fault; 959 *gpa = gg.gpa; 960 } 961 return (error); 962} 963 964#ifndef min 965#define min(a,b) (((a) < (b)) ? (a) : (b)) 966#endif 967 968int 969vm_gla2gpa(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging, 970 uint64_t gla, size_t len, int prot, struct iovec *iov, int iovcnt) 971{ 972 uint64_t gpa; 973 int error, fault, i, n, off; 974 975 for (i = 0; i < iovcnt; i++) { 976 iov[i].iov_base = 0; 977 iov[i].iov_len = 0; 978 } 979 980 while (len) { 981 assert(iovcnt > 0); 982 error = gla2gpa(ctx, vcpu, paging, gla, prot, &fault, &gpa); 983 if (error) 984 return (-1); 985 if (fault) 986 return (1); 987 988 off = gpa & PAGE_MASK; 989 n = min(len, PAGE_SIZE - off); 990 991 iov->iov_base = (void *)gpa; 992 iov->iov_len = n; 993 iov++; 994 iovcnt--; 995 996 gla += n; 997 len -= n; 998 } 999 return (0); 1000} 1001 1002void 1003vm_copyin(struct vmctx *ctx, int vcpu, struct iovec *iov, void *vp, size_t len) 1004{ 1005 const char *src; 1006 char *dst; 1007 uint64_t gpa; 1008 size_t n; 1009 1010 dst = vp; 1011 while (len) { 1012 assert(iov->iov_len); 1013 gpa = (uint64_t)iov->iov_base; 1014 n = min(len, iov->iov_len); 1015 src = vm_map_gpa(ctx, gpa, n); 1016 bcopy(src, dst, n); 1017 1018 iov++; 1019 dst += n; 1020 len -= n; 1021 } 1022} 1023 1024void 1025vm_copyout(struct vmctx *ctx, int vcpu, const void *vp, struct iovec *iov, 1026 size_t len) 1027{ 1028 const char *src; 1029 char *dst; 1030 uint64_t gpa; 1031 size_t n; 1032 1033 src = vp; 1034 while (len) { 1035 assert(iov->iov_len); 1036 gpa = (uint64_t)iov->iov_base; 1037 n = min(len, iov->iov_len); 1038 dst = vm_map_gpa(ctx, gpa, n); 1039 bcopy(src, dst, n); 1040 1041 iov++; 1042 src += n; 1043 len -= n; 1044 } 1045} 1046