1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD 3 * 4 * Copyright (c) 2011 NetApp, Inc. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY NETAPP, INC ``AS IS'' AND 17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 19 * ARE DISCLAIMED. IN NO EVENT SHALL NETAPP, INC OR CONTRIBUTORS BE LIABLE 20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 26 * SUCH DAMAGE. 27 * 28 * $FreeBSD$ 29 */ 30 /* 31 * This file and its contents are supplied under the terms of the 32 * Common Development and Distribution License ("CDDL"), version 1.0. 33 * You may only use this file in accordance with the terms of version 34 * 1.0 of the CDDL. 35 * 36 * A full copy of the text of the CDDL should have accompanied this 37 * source. A copy of the CDDL is also available via the Internet at 38 * http://www.illumos.org/license/CDDL. 39 * 40 * Copyright 2015 Pluribus Networks Inc. 41 * Copyright 2019 Joyent, Inc. 42 */ 43 44 #include <sys/cdefs.h> 45 __FBSDID("$FreeBSD$"); 46 47 #include <sys/param.h> 48 #include <sys/sysctl.h> 49 #include <sys/ioctl.h> 50 #include <sys/mman.h> 51 #include <sys/_iovec.h> 52 #include <sys/cpuset.h> 53 54 #include <x86/segments.h> 55 #include <machine/specialreg.h> 56 57 #include <errno.h> 58 #include <stdio.h> 59 #include <stdlib.h> 60 #include <assert.h> 61 #include <string.h> 62 #include <fcntl.h> 63 #include <unistd.h> 64 65 #include <libutil.h> 66 67 #include <machine/vmm.h> 68 #include <machine/vmm_dev.h> 69 #ifndef __FreeBSD__ 70 #include <sys/vmm_impl.h> 71 #endif 72 73 #include "vmmapi.h" 74 75 #define MB (1024 * 1024UL) 76 #define GB (1024 * 1024 * 1024UL) 77 78 #ifndef __FreeBSD__ 79 /* shim to no-op for now */ 80 #define MAP_NOCORE 0 81 #define MAP_ALIGNED_SUPER 0 82 83 /* Rely on PROT_NONE for guard purposes */ 84 #define MAP_GUARD (MAP_PRIVATE | MAP_ANON | MAP_NORESERVE) 85 #endif 86 87 /* 88 * Size of the guard region before and after the virtual address space 89 * mapping the guest physical memory. This must be a multiple of the 90 * superpage size for performance reasons. 91 */ 92 #define VM_MMAP_GUARD_SIZE (4 * MB) 93 94 #define PROT_RW (PROT_READ | PROT_WRITE) 95 #define PROT_ALL (PROT_READ | PROT_WRITE | PROT_EXEC) 96 97 struct vmctx { 98 int fd; 99 uint32_t lowmem_limit; 100 int memflags; 101 size_t lowmem; 102 size_t highmem; 103 char *baseaddr; 104 char *name; 105 }; 106 107 #ifdef __FreeBSD__ 108 #define CREATE(x) sysctlbyname("hw.vmm.create", NULL, NULL, (x), strlen((x))) 109 #define DESTROY(x) sysctlbyname("hw.vmm.destroy", NULL, NULL, (x), strlen((x))) 110 #else 111 #define CREATE(x) vm_do_ctl(VMM_CREATE_VM, (x)) 112 #define DESTROY(x) vm_do_ctl(VMM_DESTROY_VM, (x)) 113 114 static int 115 vm_do_ctl(int cmd, const char *name) 116 { 117 int ctl_fd; 118 119 ctl_fd = open(VMM_CTL_DEV, O_EXCL | O_RDWR); 120 if (ctl_fd < 0) { 121 return (-1); 122 } 123 124 if (ioctl(ctl_fd, cmd, name) == -1) { 125 int err = errno; 126 127 /* Do not lose ioctl errno through the close(2) */ 128 (void) close(ctl_fd); 129 errno = err; 130 return (-1); 131 } 132 (void) close(ctl_fd); 133 134 return (0); 135 } 136 #endif 137 138 static int 139 vm_device_open(const char *name) 140 { 141 int fd, len; 142 char *vmfile; 143 144 len = strlen("/dev/vmm/") + strlen(name) + 1; 145 vmfile = malloc(len); 146 assert(vmfile != NULL); 147 snprintf(vmfile, len, "/dev/vmm/%s", name); 148 149 /* Open the device file */ 150 fd = open(vmfile, O_RDWR, 0); 151 152 free(vmfile); 153 return (fd); 154 } 155 156 int 157 vm_create(const char *name) 158 { 159 160 return (CREATE((char *)name)); 161 } 162 163 struct vmctx * 164 vm_open(const char *name) 165 { 166 struct vmctx *vm; 167 168 vm = malloc(sizeof(struct vmctx) + strlen(name) + 1); 169 assert(vm != NULL); 170 171 vm->fd = -1; 172 vm->memflags = 0; 173 vm->lowmem_limit = 3 * GB; 174 vm->name = (char *)(vm + 1); 175 strcpy(vm->name, name); 176 177 if ((vm->fd = vm_device_open(vm->name)) < 0) 178 goto err; 179 180 return (vm); 181 err: 182 vm_destroy(vm); 183 return (NULL); 184 } 185 186 void 187 vm_destroy(struct vmctx *vm) 188 { 189 assert(vm != NULL); 190 191 if (vm->fd >= 0) 192 close(vm->fd); 193 DESTROY(vm->name); 194 195 free(vm); 196 } 197 198 int 199 vm_parse_memsize(const char *optarg, size_t *ret_memsize) 200 { 201 char *endptr; 202 size_t optval; 203 int error; 204 205 optval = strtoul(optarg, &endptr, 0); 206 if (*optarg != '\0' && *endptr == '\0') { 207 /* 208 * For the sake of backward compatibility if the memory size 209 * specified on the command line is less than a megabyte then 210 * it is interpreted as being in units of MB. 211 */ 212 if (optval < MB) 213 optval *= MB; 214 *ret_memsize = optval; 215 error = 0; 216 } else 217 error = expand_number(optarg, ret_memsize); 218 219 return (error); 220 } 221 222 uint32_t 223 vm_get_lowmem_limit(struct vmctx *ctx) 224 { 225 226 return (ctx->lowmem_limit); 227 } 228 229 void 230 vm_set_lowmem_limit(struct vmctx *ctx, uint32_t limit) 231 { 232 233 ctx->lowmem_limit = limit; 234 } 235 236 void 237 vm_set_memflags(struct vmctx *ctx, int flags) 238 { 239 240 ctx->memflags = flags; 241 } 242 243 int 244 vm_get_memflags(struct vmctx *ctx) 245 { 246 247 return (ctx->memflags); 248 } 249 250 /* 251 * Map segment 'segid' starting at 'off' into guest address range [gpa,gpa+len). 252 */ 253 int 254 vm_mmap_memseg(struct vmctx *ctx, vm_paddr_t gpa, int segid, vm_ooffset_t off, 255 size_t len, int prot) 256 { 257 struct vm_memmap memmap; 258 int error, flags; 259 260 memmap.gpa = gpa; 261 memmap.segid = segid; 262 memmap.segoff = off; 263 memmap.len = len; 264 memmap.prot = prot; 265 memmap.flags = 0; 266 267 if (ctx->memflags & VM_MEM_F_WIRED) 268 memmap.flags |= VM_MEMMAP_F_WIRED; 269 270 /* 271 * If this mapping already exists then don't create it again. This 272 * is the common case for SYSMEM mappings created by bhyveload(8). 273 */ 274 error = vm_mmap_getnext(ctx, &gpa, &segid, &off, &len, &prot, &flags); 275 if (error == 0 && gpa == memmap.gpa) { 276 if (segid != memmap.segid || off != memmap.segoff || 277 prot != memmap.prot || flags != memmap.flags) { 278 errno = EEXIST; 279 return (-1); 280 } else { 281 return (0); 282 } 283 } 284 285 error = ioctl(ctx->fd, VM_MMAP_MEMSEG, &memmap); 286 return (error); 287 } 288 289 int 290 vm_mmap_getnext(struct vmctx *ctx, vm_paddr_t *gpa, int *segid, 291 vm_ooffset_t *segoff, size_t *len, int *prot, int *flags) 292 { 293 struct vm_memmap memmap; 294 int error; 295 296 bzero(&memmap, sizeof(struct vm_memmap)); 297 memmap.gpa = *gpa; 298 error = ioctl(ctx->fd, VM_MMAP_GETNEXT, &memmap); 299 if (error == 0) { 300 *gpa = memmap.gpa; 301 *segid = memmap.segid; 302 *segoff = memmap.segoff; 303 *len = memmap.len; 304 *prot = memmap.prot; 305 *flags = memmap.flags; 306 } 307 return (error); 308 } 309 310 /* 311 * Return 0 if the segments are identical and non-zero otherwise. 312 * 313 * This is slightly complicated by the fact that only device memory segments 314 * are named. 315 */ 316 static int 317 cmpseg(size_t len, const char *str, size_t len2, const char *str2) 318 { 319 320 if (len == len2) { 321 if ((!str && !str2) || (str && str2 && !strcmp(str, str2))) 322 return (0); 323 } 324 return (-1); 325 } 326 327 static int 328 vm_alloc_memseg(struct vmctx *ctx, int segid, size_t len, const char *name) 329 { 330 struct vm_memseg memseg; 331 size_t n; 332 int error; 333 334 /* 335 * If the memory segment has already been created then just return. 336 * This is the usual case for the SYSMEM segment created by userspace 337 * loaders like bhyveload(8). 338 */ 339 error = vm_get_memseg(ctx, segid, &memseg.len, memseg.name, 340 sizeof(memseg.name)); 341 if (error) 342 return (error); 343 344 if (memseg.len != 0) { 345 if (cmpseg(len, name, memseg.len, VM_MEMSEG_NAME(&memseg))) { 346 errno = EINVAL; 347 return (-1); 348 } else { 349 return (0); 350 } 351 } 352 353 bzero(&memseg, sizeof(struct vm_memseg)); 354 memseg.segid = segid; 355 memseg.len = len; 356 if (name != NULL) { 357 n = strlcpy(memseg.name, name, sizeof(memseg.name)); 358 if (n >= sizeof(memseg.name)) { 359 errno = ENAMETOOLONG; 360 return (-1); 361 } 362 } 363 364 error = ioctl(ctx->fd, VM_ALLOC_MEMSEG, &memseg); 365 return (error); 366 } 367 368 int 369 vm_get_memseg(struct vmctx *ctx, int segid, size_t *lenp, char *namebuf, 370 size_t bufsize) 371 { 372 struct vm_memseg memseg; 373 size_t n; 374 int error; 375 376 memseg.segid = segid; 377 error = ioctl(ctx->fd, VM_GET_MEMSEG, &memseg); 378 if (error == 0) { 379 *lenp = memseg.len; 380 n = strlcpy(namebuf, memseg.name, bufsize); 381 if (n >= bufsize) { 382 errno = ENAMETOOLONG; 383 error = -1; 384 } 385 } 386 return (error); 387 } 388 389 static int 390 #ifdef __FreeBSD__ 391 setup_memory_segment(struct vmctx *ctx, vm_paddr_t gpa, size_t len, char *base) 392 #else 393 setup_memory_segment(struct vmctx *ctx, int segid, vm_paddr_t gpa, size_t len, 394 char *base) 395 #endif 396 { 397 char *ptr; 398 int error, flags; 399 400 /* Map 'len' bytes starting at 'gpa' in the guest address space */ 401 #ifdef __FreeBSD__ 402 error = vm_mmap_memseg(ctx, gpa, VM_SYSMEM, gpa, len, PROT_ALL); 403 #else 404 /* 405 * As we use two segments for lowmem/highmem the offset within the 406 * segment is 0 on illumos. 407 */ 408 error = vm_mmap_memseg(ctx, gpa, segid, 0, len, PROT_ALL); 409 #endif 410 if (error) 411 return (error); 412 413 flags = MAP_SHARED | MAP_FIXED; 414 if ((ctx->memflags & VM_MEM_F_INCORE) == 0) 415 flags |= MAP_NOCORE; 416 417 /* mmap into the process address space on the host */ 418 ptr = mmap(base + gpa, len, PROT_RW, flags, ctx->fd, gpa); 419 if (ptr == MAP_FAILED) 420 return (-1); 421 422 return (0); 423 } 424 425 int 426 vm_setup_memory(struct vmctx *ctx, size_t memsize, enum vm_mmap_style vms) 427 { 428 size_t objsize, len; 429 vm_paddr_t gpa; 430 char *baseaddr, *ptr; 431 int error; 432 433 assert(vms == VM_MMAP_ALL); 434 435 /* 436 * If 'memsize' cannot fit entirely in the 'lowmem' segment then 437 * create another 'highmem' segment above 4GB for the remainder. 438 */ 439 if (memsize > ctx->lowmem_limit) { 440 ctx->lowmem = ctx->lowmem_limit; 441 ctx->highmem = memsize - ctx->lowmem_limit; 442 objsize = 4*GB + ctx->highmem; 443 } else { 444 ctx->lowmem = memsize; 445 ctx->highmem = 0; 446 objsize = ctx->lowmem; 447 } 448 449 #ifdef __FreeBSD__ 450 error = vm_alloc_memseg(ctx, VM_SYSMEM, objsize, NULL); 451 if (error) 452 return (error); 453 #endif 454 455 /* 456 * Stake out a contiguous region covering the guest physical memory 457 * and the adjoining guard regions. 458 */ 459 len = VM_MMAP_GUARD_SIZE + objsize + VM_MMAP_GUARD_SIZE; 460 ptr = mmap(NULL, len, PROT_NONE, MAP_GUARD | MAP_ALIGNED_SUPER, -1, 0); 461 if (ptr == MAP_FAILED) 462 return (-1); 463 464 baseaddr = ptr + VM_MMAP_GUARD_SIZE; 465 466 #ifdef __FreeBSD__ 467 if (ctx->highmem > 0) { 468 gpa = 4*GB; 469 len = ctx->highmem; 470 error = setup_memory_segment(ctx, gpa, len, baseaddr); 471 if (error) 472 return (error); 473 } 474 475 if (ctx->lowmem > 0) { 476 gpa = 0; 477 len = ctx->lowmem; 478 error = setup_memory_segment(ctx, gpa, len, baseaddr); 479 if (error) 480 return (error); 481 } 482 #else 483 if (ctx->highmem > 0) { 484 error = vm_alloc_memseg(ctx, VM_HIGHMEM, ctx->highmem, NULL); 485 if (error) 486 return (error); 487 gpa = 4*GB; 488 len = ctx->highmem; 489 error = setup_memory_segment(ctx, VM_HIGHMEM, gpa, len, baseaddr); 490 if (error) 491 return (error); 492 } 493 494 if (ctx->lowmem > 0) { 495 error = vm_alloc_memseg(ctx, VM_LOWMEM, ctx->lowmem, NULL); 496 if (error) 497 return (error); 498 gpa = 0; 499 len = ctx->lowmem; 500 error = setup_memory_segment(ctx, VM_LOWMEM, gpa, len, baseaddr); 501 if (error) 502 return (error); 503 } 504 #endif 505 506 ctx->baseaddr = baseaddr; 507 508 return (0); 509 } 510 511 /* 512 * Returns a non-NULL pointer if [gaddr, gaddr+len) is entirely contained in 513 * the lowmem or highmem regions. 514 * 515 * In particular return NULL if [gaddr, gaddr+len) falls in guest MMIO region. 516 * The instruction emulation code depends on this behavior. 517 */ 518 void * 519 vm_map_gpa(struct vmctx *ctx, vm_paddr_t gaddr, size_t len) 520 { 521 522 if (ctx->lowmem > 0) { 523 if (gaddr < ctx->lowmem && len <= ctx->lowmem && 524 gaddr + len <= ctx->lowmem) 525 return (ctx->baseaddr + gaddr); 526 } 527 528 if (ctx->highmem > 0) { 529 if (gaddr >= 4*GB) { 530 if (gaddr < 4*GB + ctx->highmem && 531 len <= ctx->highmem && 532 gaddr + len <= 4*GB + ctx->highmem) 533 return (ctx->baseaddr + gaddr); 534 } 535 } 536 537 return (NULL); 538 } 539 540 size_t 541 vm_get_lowmem_size(struct vmctx *ctx) 542 { 543 544 return (ctx->lowmem); 545 } 546 547 size_t 548 vm_get_highmem_size(struct vmctx *ctx) 549 { 550 551 return (ctx->highmem); 552 } 553 554 void * 555 vm_create_devmem(struct vmctx *ctx, int segid, const char *name, size_t len) 556 { 557 #ifdef __FreeBSD__ 558 char pathname[MAXPATHLEN]; 559 #endif 560 size_t len2; 561 char *base, *ptr; 562 int fd, error, flags; 563 off_t mapoff; 564 565 fd = -1; 566 ptr = MAP_FAILED; 567 if (name == NULL || strlen(name) == 0) { 568 errno = EINVAL; 569 goto done; 570 } 571 572 error = vm_alloc_memseg(ctx, segid, len, name); 573 if (error) 574 goto done; 575 576 #ifdef __FreeBSD__ 577 strlcpy(pathname, "/dev/vmm.io/", sizeof(pathname)); 578 strlcat(pathname, ctx->name, sizeof(pathname)); 579 strlcat(pathname, ".", sizeof(pathname)); 580 strlcat(pathname, name, sizeof(pathname)); 581 582 fd = open(pathname, O_RDWR); 583 if (fd < 0) 584 goto done; 585 #else 586 { 587 struct vm_devmem_offset vdo; 588 589 vdo.segid = segid; 590 error = ioctl(ctx->fd, VM_DEVMEM_GETOFFSET, &vdo); 591 if (error == 0) { 592 mapoff = vdo.offset; 593 } else { 594 goto done; 595 } 596 } 597 #endif 598 599 /* 600 * Stake out a contiguous region covering the device memory and the 601 * adjoining guard regions. 602 */ 603 len2 = VM_MMAP_GUARD_SIZE + len + VM_MMAP_GUARD_SIZE; 604 base = mmap(NULL, len2, PROT_NONE, MAP_GUARD | MAP_ALIGNED_SUPER, -1, 605 0); 606 if (base == MAP_FAILED) 607 goto done; 608 609 flags = MAP_SHARED | MAP_FIXED; 610 if ((ctx->memflags & VM_MEM_F_INCORE) == 0) 611 flags |= MAP_NOCORE; 612 613 #ifdef __FreeBSD__ 614 /* mmap the devmem region in the host address space */ 615 ptr = mmap(base + VM_MMAP_GUARD_SIZE, len, PROT_RW, flags, fd, 0); 616 #else 617 /* mmap the devmem region in the host address space */ 618 ptr = mmap(base + VM_MMAP_GUARD_SIZE, len, PROT_RW, flags, ctx->fd, 619 mapoff); 620 #endif 621 done: 622 if (fd >= 0) 623 close(fd); 624 return (ptr); 625 } 626 627 int 628 vm_set_desc(struct vmctx *ctx, int vcpu, int reg, 629 uint64_t base, uint32_t limit, uint32_t access) 630 { 631 int error; 632 struct vm_seg_desc vmsegdesc; 633 634 bzero(&vmsegdesc, sizeof(vmsegdesc)); 635 vmsegdesc.cpuid = vcpu; 636 vmsegdesc.regnum = reg; 637 vmsegdesc.desc.base = base; 638 vmsegdesc.desc.limit = limit; 639 vmsegdesc.desc.access = access; 640 641 error = ioctl(ctx->fd, VM_SET_SEGMENT_DESCRIPTOR, &vmsegdesc); 642 return (error); 643 } 644 645 int 646 vm_get_desc(struct vmctx *ctx, int vcpu, int reg, 647 uint64_t *base, uint32_t *limit, uint32_t *access) 648 { 649 int error; 650 struct vm_seg_desc vmsegdesc; 651 652 bzero(&vmsegdesc, sizeof(vmsegdesc)); 653 vmsegdesc.cpuid = vcpu; 654 vmsegdesc.regnum = reg; 655 656 error = ioctl(ctx->fd, VM_GET_SEGMENT_DESCRIPTOR, &vmsegdesc); 657 if (error == 0) { 658 *base = vmsegdesc.desc.base; 659 *limit = vmsegdesc.desc.limit; 660 *access = vmsegdesc.desc.access; 661 } 662 return (error); 663 } 664 665 int 666 vm_get_seg_desc(struct vmctx *ctx, int vcpu, int reg, struct seg_desc *seg_desc) 667 { 668 int error; 669 670 error = vm_get_desc(ctx, vcpu, reg, &seg_desc->base, &seg_desc->limit, 671 &seg_desc->access); 672 return (error); 673 } 674 675 int 676 vm_set_register(struct vmctx *ctx, int vcpu, int reg, uint64_t val) 677 { 678 int error; 679 struct vm_register vmreg; 680 681 bzero(&vmreg, sizeof(vmreg)); 682 vmreg.cpuid = vcpu; 683 vmreg.regnum = reg; 684 vmreg.regval = val; 685 686 error = ioctl(ctx->fd, VM_SET_REGISTER, &vmreg); 687 return (error); 688 } 689 690 int 691 vm_get_register(struct vmctx *ctx, int vcpu, int reg, uint64_t *ret_val) 692 { 693 int error; 694 struct vm_register vmreg; 695 696 bzero(&vmreg, sizeof(vmreg)); 697 vmreg.cpuid = vcpu; 698 vmreg.regnum = reg; 699 700 error = ioctl(ctx->fd, VM_GET_REGISTER, &vmreg); 701 *ret_val = vmreg.regval; 702 return (error); 703 } 704 705 int 706 vm_set_register_set(struct vmctx *ctx, int vcpu, unsigned int count, 707 const int *regnums, uint64_t *regvals) 708 { 709 int error; 710 struct vm_register_set vmregset; 711 712 bzero(&vmregset, sizeof(vmregset)); 713 vmregset.cpuid = vcpu; 714 vmregset.count = count; 715 vmregset.regnums = regnums; 716 vmregset.regvals = regvals; 717 718 error = ioctl(ctx->fd, VM_SET_REGISTER_SET, &vmregset); 719 return (error); 720 } 721 722 int 723 vm_get_register_set(struct vmctx *ctx, int vcpu, unsigned int count, 724 const int *regnums, uint64_t *regvals) 725 { 726 int error; 727 struct vm_register_set vmregset; 728 729 bzero(&vmregset, sizeof(vmregset)); 730 vmregset.cpuid = vcpu; 731 vmregset.count = count; 732 vmregset.regnums = regnums; 733 vmregset.regvals = regvals; 734 735 error = ioctl(ctx->fd, VM_GET_REGISTER_SET, &vmregset); 736 return (error); 737 } 738 739 int 740 vm_run(struct vmctx *ctx, int vcpu, struct vm_exit *vmexit) 741 { 742 int error; 743 struct vm_run vmrun; 744 745 bzero(&vmrun, sizeof(vmrun)); 746 vmrun.cpuid = vcpu; 747 748 error = ioctl(ctx->fd, VM_RUN, &vmrun); 749 bcopy(&vmrun.vm_exit, vmexit, sizeof(struct vm_exit)); 750 return (error); 751 } 752 753 int 754 vm_suspend(struct vmctx *ctx, enum vm_suspend_how how) 755 { 756 struct vm_suspend vmsuspend; 757 758 bzero(&vmsuspend, sizeof(vmsuspend)); 759 vmsuspend.how = how; 760 return (ioctl(ctx->fd, VM_SUSPEND, &vmsuspend)); 761 } 762 763 int 764 vm_reinit(struct vmctx *ctx) 765 { 766 767 return (ioctl(ctx->fd, VM_REINIT, 0)); 768 } 769 770 int 771 vm_inject_exception(struct vmctx *ctx, int vcpu, int vector, int errcode_valid, 772 uint32_t errcode, int restart_instruction) 773 { 774 struct vm_exception exc; 775 776 exc.cpuid = vcpu; 777 exc.vector = vector; 778 exc.error_code = errcode; 779 exc.error_code_valid = errcode_valid; 780 exc.restart_instruction = restart_instruction; 781 782 return (ioctl(ctx->fd, VM_INJECT_EXCEPTION, &exc)); 783 } 784 785 int 786 vm_apicid2vcpu(struct vmctx *ctx, int apicid) 787 { 788 /* 789 * The apic id associated with the 'vcpu' has the same numerical value 790 * as the 'vcpu' itself. 791 */ 792 return (apicid); 793 } 794 795 int 796 vm_lapic_irq(struct vmctx *ctx, int vcpu, int vector) 797 { 798 struct vm_lapic_irq vmirq; 799 800 bzero(&vmirq, sizeof(vmirq)); 801 vmirq.cpuid = vcpu; 802 vmirq.vector = vector; 803 804 return (ioctl(ctx->fd, VM_LAPIC_IRQ, &vmirq)); 805 } 806 807 int 808 vm_lapic_local_irq(struct vmctx *ctx, int vcpu, int vector) 809 { 810 struct vm_lapic_irq vmirq; 811 812 bzero(&vmirq, sizeof(vmirq)); 813 vmirq.cpuid = vcpu; 814 vmirq.vector = vector; 815 816 return (ioctl(ctx->fd, VM_LAPIC_LOCAL_IRQ, &vmirq)); 817 } 818 819 int 820 vm_lapic_msi(struct vmctx *ctx, uint64_t addr, uint64_t msg) 821 { 822 struct vm_lapic_msi vmmsi; 823 824 bzero(&vmmsi, sizeof(vmmsi)); 825 vmmsi.addr = addr; 826 vmmsi.msg = msg; 827 828 return (ioctl(ctx->fd, VM_LAPIC_MSI, &vmmsi)); 829 } 830 831 int 832 vm_ioapic_assert_irq(struct vmctx *ctx, int irq) 833 { 834 struct vm_ioapic_irq ioapic_irq; 835 836 bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq)); 837 ioapic_irq.irq = irq; 838 839 return (ioctl(ctx->fd, VM_IOAPIC_ASSERT_IRQ, &ioapic_irq)); 840 } 841 842 int 843 vm_ioapic_deassert_irq(struct vmctx *ctx, int irq) 844 { 845 struct vm_ioapic_irq ioapic_irq; 846 847 bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq)); 848 ioapic_irq.irq = irq; 849 850 return (ioctl(ctx->fd, VM_IOAPIC_DEASSERT_IRQ, &ioapic_irq)); 851 } 852 853 int 854 vm_ioapic_pulse_irq(struct vmctx *ctx, int irq) 855 { 856 struct vm_ioapic_irq ioapic_irq; 857 858 bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq)); 859 ioapic_irq.irq = irq; 860 861 return (ioctl(ctx->fd, VM_IOAPIC_PULSE_IRQ, &ioapic_irq)); 862 } 863 864 int 865 vm_ioapic_pincount(struct vmctx *ctx, int *pincount) 866 { 867 868 return (ioctl(ctx->fd, VM_IOAPIC_PINCOUNT, pincount)); 869 } 870 871 int 872 vm_isa_assert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq) 873 { 874 struct vm_isa_irq isa_irq; 875 876 bzero(&isa_irq, sizeof(struct vm_isa_irq)); 877 isa_irq.atpic_irq = atpic_irq; 878 isa_irq.ioapic_irq = ioapic_irq; 879 880 return (ioctl(ctx->fd, VM_ISA_ASSERT_IRQ, &isa_irq)); 881 } 882 883 int 884 vm_isa_deassert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq) 885 { 886 struct vm_isa_irq isa_irq; 887 888 bzero(&isa_irq, sizeof(struct vm_isa_irq)); 889 isa_irq.atpic_irq = atpic_irq; 890 isa_irq.ioapic_irq = ioapic_irq; 891 892 return (ioctl(ctx->fd, VM_ISA_DEASSERT_IRQ, &isa_irq)); 893 } 894 895 int 896 vm_isa_pulse_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq) 897 { 898 struct vm_isa_irq isa_irq; 899 900 bzero(&isa_irq, sizeof(struct vm_isa_irq)); 901 isa_irq.atpic_irq = atpic_irq; 902 isa_irq.ioapic_irq = ioapic_irq; 903 904 return (ioctl(ctx->fd, VM_ISA_PULSE_IRQ, &isa_irq)); 905 } 906 907 int 908 vm_isa_set_irq_trigger(struct vmctx *ctx, int atpic_irq, 909 enum vm_intr_trigger trigger) 910 { 911 struct vm_isa_irq_trigger isa_irq_trigger; 912 913 bzero(&isa_irq_trigger, sizeof(struct vm_isa_irq_trigger)); 914 isa_irq_trigger.atpic_irq = atpic_irq; 915 isa_irq_trigger.trigger = trigger; 916 917 return (ioctl(ctx->fd, VM_ISA_SET_IRQ_TRIGGER, &isa_irq_trigger)); 918 } 919 920 int 921 vm_inject_nmi(struct vmctx *ctx, int vcpu) 922 { 923 struct vm_nmi vmnmi; 924 925 bzero(&vmnmi, sizeof(vmnmi)); 926 vmnmi.cpuid = vcpu; 927 928 return (ioctl(ctx->fd, VM_INJECT_NMI, &vmnmi)); 929 } 930 931 static struct { 932 const char *name; 933 int type; 934 } capstrmap[] = { 935 { "hlt_exit", VM_CAP_HALT_EXIT }, 936 { "mtrap_exit", VM_CAP_MTRAP_EXIT }, 937 { "pause_exit", VM_CAP_PAUSE_EXIT }, 938 { "unrestricted_guest", VM_CAP_UNRESTRICTED_GUEST }, 939 { "enable_invpcid", VM_CAP_ENABLE_INVPCID }, 940 { 0 } 941 }; 942 943 int 944 vm_capability_name2type(const char *capname) 945 { 946 int i; 947 948 for (i = 0; capstrmap[i].name != NULL && capname != NULL; i++) { 949 if (strcmp(capstrmap[i].name, capname) == 0) 950 return (capstrmap[i].type); 951 } 952 953 return (-1); 954 } 955 956 const char * 957 vm_capability_type2name(int type) 958 { 959 int i; 960 961 for (i = 0; capstrmap[i].name != NULL; i++) { 962 if (capstrmap[i].type == type) 963 return (capstrmap[i].name); 964 } 965 966 return (NULL); 967 } 968 969 int 970 vm_get_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap, 971 int *retval) 972 { 973 int error; 974 struct vm_capability vmcap; 975 976 bzero(&vmcap, sizeof(vmcap)); 977 vmcap.cpuid = vcpu; 978 vmcap.captype = cap; 979 980 error = ioctl(ctx->fd, VM_GET_CAPABILITY, &vmcap); 981 *retval = vmcap.capval; 982 return (error); 983 } 984 985 int 986 vm_set_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap, int val) 987 { 988 struct vm_capability vmcap; 989 990 bzero(&vmcap, sizeof(vmcap)); 991 vmcap.cpuid = vcpu; 992 vmcap.captype = cap; 993 vmcap.capval = val; 994 995 return (ioctl(ctx->fd, VM_SET_CAPABILITY, &vmcap)); 996 } 997 998 int 999 vm_assign_pptdev(struct vmctx *ctx, int bus, int slot, int func) 1000 { 1001 struct vm_pptdev pptdev; 1002 1003 bzero(&pptdev, sizeof(pptdev)); 1004 pptdev.bus = bus; 1005 pptdev.slot = slot; 1006 pptdev.func = func; 1007 1008 return (ioctl(ctx->fd, VM_BIND_PPTDEV, &pptdev)); 1009 } 1010 1011 int 1012 vm_unassign_pptdev(struct vmctx *ctx, int bus, int slot, int func) 1013 { 1014 struct vm_pptdev pptdev; 1015 1016 bzero(&pptdev, sizeof(pptdev)); 1017 pptdev.bus = bus; 1018 pptdev.slot = slot; 1019 pptdev.func = func; 1020 1021 return (ioctl(ctx->fd, VM_UNBIND_PPTDEV, &pptdev)); 1022 } 1023 1024 int 1025 vm_map_pptdev_mmio(struct vmctx *ctx, int bus, int slot, int func, 1026 vm_paddr_t gpa, size_t len, vm_paddr_t hpa) 1027 { 1028 struct vm_pptdev_mmio pptmmio; 1029 1030 bzero(&pptmmio, sizeof(pptmmio)); 1031 pptmmio.bus = bus; 1032 pptmmio.slot = slot; 1033 pptmmio.func = func; 1034 pptmmio.gpa = gpa; 1035 pptmmio.len = len; 1036 pptmmio.hpa = hpa; 1037 1038 return (ioctl(ctx->fd, VM_MAP_PPTDEV_MMIO, &pptmmio)); 1039 } 1040 1041 int 1042 vm_setup_pptdev_msi(struct vmctx *ctx, int vcpu, int bus, int slot, int func, 1043 uint64_t addr, uint64_t msg, int numvec) 1044 { 1045 struct vm_pptdev_msi pptmsi; 1046 1047 bzero(&pptmsi, sizeof(pptmsi)); 1048 pptmsi.vcpu = vcpu; 1049 pptmsi.bus = bus; 1050 pptmsi.slot = slot; 1051 pptmsi.func = func; 1052 pptmsi.msg = msg; 1053 pptmsi.addr = addr; 1054 pptmsi.numvec = numvec; 1055 1056 return (ioctl(ctx->fd, VM_PPTDEV_MSI, &pptmsi)); 1057 } 1058 1059 int 1060 vm_setup_pptdev_msix(struct vmctx *ctx, int vcpu, int bus, int slot, int func, 1061 int idx, uint64_t addr, uint64_t msg, uint32_t vector_control) 1062 { 1063 struct vm_pptdev_msix pptmsix; 1064 1065 bzero(&pptmsix, sizeof(pptmsix)); 1066 pptmsix.vcpu = vcpu; 1067 pptmsix.bus = bus; 1068 pptmsix.slot = slot; 1069 pptmsix.func = func; 1070 pptmsix.idx = idx; 1071 pptmsix.msg = msg; 1072 pptmsix.addr = addr; 1073 pptmsix.vector_control = vector_control; 1074 1075 return ioctl(ctx->fd, VM_PPTDEV_MSIX, &pptmsix); 1076 } 1077 1078 uint64_t * 1079 vm_get_stats(struct vmctx *ctx, int vcpu, struct timeval *ret_tv, 1080 int *ret_entries) 1081 { 1082 int error; 1083 1084 static struct vm_stats vmstats; 1085 1086 vmstats.cpuid = vcpu; 1087 1088 error = ioctl(ctx->fd, VM_STATS_IOC, &vmstats); 1089 if (error == 0) { 1090 if (ret_entries) 1091 *ret_entries = vmstats.num_entries; 1092 if (ret_tv) 1093 *ret_tv = vmstats.tv; 1094 return (vmstats.statbuf); 1095 } else 1096 return (NULL); 1097 } 1098 1099 const char * 1100 vm_get_stat_desc(struct vmctx *ctx, int index) 1101 { 1102 static struct vm_stat_desc statdesc; 1103 1104 statdesc.index = index; 1105 if (ioctl(ctx->fd, VM_STAT_DESC, &statdesc) == 0) 1106 return (statdesc.desc); 1107 else 1108 return (NULL); 1109 } 1110 1111 int 1112 vm_get_x2apic_state(struct vmctx *ctx, int vcpu, enum x2apic_state *state) 1113 { 1114 int error; 1115 struct vm_x2apic x2apic; 1116 1117 bzero(&x2apic, sizeof(x2apic)); 1118 x2apic.cpuid = vcpu; 1119 1120 error = ioctl(ctx->fd, VM_GET_X2APIC_STATE, &x2apic); 1121 *state = x2apic.state; 1122 return (error); 1123 } 1124 1125 int 1126 vm_set_x2apic_state(struct vmctx *ctx, int vcpu, enum x2apic_state state) 1127 { 1128 int error; 1129 struct vm_x2apic x2apic; 1130 1131 bzero(&x2apic, sizeof(x2apic)); 1132 x2apic.cpuid = vcpu; 1133 x2apic.state = state; 1134 1135 error = ioctl(ctx->fd, VM_SET_X2APIC_STATE, &x2apic); 1136 1137 return (error); 1138 } 1139 1140 /* 1141 * From Intel Vol 3a: 1142 * Table 9-1. IA-32 Processor States Following Power-up, Reset or INIT 1143 */ 1144 int 1145 vcpu_reset(struct vmctx *vmctx, int vcpu) 1146 { 1147 int error; 1148 uint64_t rflags, rip, cr0, cr4, zero, desc_base, rdx; 1149 uint32_t desc_access, desc_limit; 1150 uint16_t sel; 1151 1152 zero = 0; 1153 1154 rflags = 0x2; 1155 error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RFLAGS, rflags); 1156 if (error) 1157 goto done; 1158 1159 rip = 0xfff0; 1160 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RIP, rip)) != 0) 1161 goto done; 1162 1163 cr0 = CR0_NE; 1164 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR0, cr0)) != 0) 1165 goto done; 1166 1167 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR3, zero)) != 0) 1168 goto done; 1169 1170 cr4 = 0; 1171 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR4, cr4)) != 0) 1172 goto done; 1173 1174 /* 1175 * CS: present, r/w, accessed, 16-bit, byte granularity, usable 1176 */ 1177 desc_base = 0xffff0000; 1178 desc_limit = 0xffff; 1179 desc_access = 0x0093; 1180 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_CS, 1181 desc_base, desc_limit, desc_access); 1182 if (error) 1183 goto done; 1184 1185 sel = 0xf000; 1186 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CS, sel)) != 0) 1187 goto done; 1188 1189 /* 1190 * SS,DS,ES,FS,GS: present, r/w, accessed, 16-bit, byte granularity 1191 */ 1192 desc_base = 0; 1193 desc_limit = 0xffff; 1194 desc_access = 0x0093; 1195 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_SS, 1196 desc_base, desc_limit, desc_access); 1197 if (error) 1198 goto done; 1199 1200 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_DS, 1201 desc_base, desc_limit, desc_access); 1202 if (error) 1203 goto done; 1204 1205 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_ES, 1206 desc_base, desc_limit, desc_access); 1207 if (error) 1208 goto done; 1209 1210 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_FS, 1211 desc_base, desc_limit, desc_access); 1212 if (error) 1213 goto done; 1214 1215 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GS, 1216 desc_base, desc_limit, desc_access); 1217 if (error) 1218 goto done; 1219 1220 sel = 0; 1221 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_SS, sel)) != 0) 1222 goto done; 1223 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_DS, sel)) != 0) 1224 goto done; 1225 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_ES, sel)) != 0) 1226 goto done; 1227 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_FS, sel)) != 0) 1228 goto done; 1229 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_GS, sel)) != 0) 1230 goto done; 1231 1232 /* General purpose registers */ 1233 rdx = 0xf00; 1234 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RAX, zero)) != 0) 1235 goto done; 1236 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBX, zero)) != 0) 1237 goto done; 1238 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RCX, zero)) != 0) 1239 goto done; 1240 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDX, rdx)) != 0) 1241 goto done; 1242 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSI, zero)) != 0) 1243 goto done; 1244 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDI, zero)) != 0) 1245 goto done; 1246 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBP, zero)) != 0) 1247 goto done; 1248 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSP, zero)) != 0) 1249 goto done; 1250 1251 /* GDTR, IDTR */ 1252 desc_base = 0; 1253 desc_limit = 0xffff; 1254 desc_access = 0; 1255 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GDTR, 1256 desc_base, desc_limit, desc_access); 1257 if (error != 0) 1258 goto done; 1259 1260 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_IDTR, 1261 desc_base, desc_limit, desc_access); 1262 if (error != 0) 1263 goto done; 1264 1265 /* TR */ 1266 desc_base = 0; 1267 desc_limit = 0xffff; 1268 desc_access = 0x0000008b; 1269 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_TR, 0, 0, desc_access); 1270 if (error) 1271 goto done; 1272 1273 sel = 0; 1274 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_TR, sel)) != 0) 1275 goto done; 1276 1277 /* LDTR */ 1278 desc_base = 0; 1279 desc_limit = 0xffff; 1280 desc_access = 0x00000082; 1281 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_LDTR, desc_base, 1282 desc_limit, desc_access); 1283 if (error) 1284 goto done; 1285 1286 sel = 0; 1287 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_LDTR, 0)) != 0) 1288 goto done; 1289 1290 /* XXX cr2, debug registers */ 1291 1292 error = 0; 1293 done: 1294 return (error); 1295 } 1296 1297 int 1298 vm_get_gpa_pmap(struct vmctx *ctx, uint64_t gpa, uint64_t *pte, int *num) 1299 { 1300 int error, i; 1301 struct vm_gpa_pte gpapte; 1302 1303 bzero(&gpapte, sizeof(gpapte)); 1304 gpapte.gpa = gpa; 1305 1306 error = ioctl(ctx->fd, VM_GET_GPA_PMAP, &gpapte); 1307 1308 if (error == 0) { 1309 *num = gpapte.ptenum; 1310 for (i = 0; i < gpapte.ptenum; i++) 1311 pte[i] = gpapte.pte[i]; 1312 } 1313 1314 return (error); 1315 } 1316 1317 int 1318 vm_get_hpet_capabilities(struct vmctx *ctx, uint32_t *capabilities) 1319 { 1320 int error; 1321 struct vm_hpet_cap cap; 1322 1323 bzero(&cap, sizeof(struct vm_hpet_cap)); 1324 error = ioctl(ctx->fd, VM_GET_HPET_CAPABILITIES, &cap); 1325 if (capabilities != NULL) 1326 *capabilities = cap.capabilities; 1327 return (error); 1328 } 1329 1330 int 1331 vm_gla2gpa(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging, 1332 uint64_t gla, int prot, uint64_t *gpa, int *fault) 1333 { 1334 struct vm_gla2gpa gg; 1335 int error; 1336 1337 bzero(&gg, sizeof(struct vm_gla2gpa)); 1338 gg.vcpuid = vcpu; 1339 gg.prot = prot; 1340 gg.gla = gla; 1341 gg.paging = *paging; 1342 1343 error = ioctl(ctx->fd, VM_GLA2GPA, &gg); 1344 if (error == 0) { 1345 *fault = gg.fault; 1346 *gpa = gg.gpa; 1347 } 1348 return (error); 1349 } 1350 1351 int 1352 vm_gla2gpa_nofault(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging, 1353 uint64_t gla, int prot, uint64_t *gpa, int *fault) 1354 { 1355 struct vm_gla2gpa gg; 1356 int error; 1357 1358 bzero(&gg, sizeof(struct vm_gla2gpa)); 1359 gg.vcpuid = vcpu; 1360 gg.prot = prot; 1361 gg.gla = gla; 1362 gg.paging = *paging; 1363 1364 error = ioctl(ctx->fd, VM_GLA2GPA_NOFAULT, &gg); 1365 if (error == 0) { 1366 *fault = gg.fault; 1367 *gpa = gg.gpa; 1368 } 1369 return (error); 1370 } 1371 1372 #ifndef min 1373 #define min(a,b) (((a) < (b)) ? (a) : (b)) 1374 #endif 1375 1376 int 1377 vm_copy_setup(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging, 1378 uint64_t gla, size_t len, int prot, struct iovec *iov, int iovcnt, 1379 int *fault) 1380 { 1381 void *va; 1382 uint64_t gpa; 1383 int error, i, n, off; 1384 1385 for (i = 0; i < iovcnt; i++) { 1386 iov[i].iov_base = 0; 1387 iov[i].iov_len = 0; 1388 } 1389 1390 while (len) { 1391 assert(iovcnt > 0); 1392 error = vm_gla2gpa(ctx, vcpu, paging, gla, prot, &gpa, fault); 1393 if (error || *fault) 1394 return (error); 1395 1396 off = gpa & PAGE_MASK; 1397 n = min(len, PAGE_SIZE - off); 1398 1399 va = vm_map_gpa(ctx, gpa, n); 1400 if (va == NULL) 1401 return (EFAULT); 1402 1403 iov->iov_base = va; 1404 iov->iov_len = n; 1405 iov++; 1406 iovcnt--; 1407 1408 gla += n; 1409 len -= n; 1410 } 1411 return (0); 1412 } 1413 1414 void 1415 vm_copy_teardown(struct vmctx *ctx, int vcpu, struct iovec *iov, int iovcnt) 1416 { 1417 1418 return; 1419 } 1420 1421 void 1422 vm_copyin(struct vmctx *ctx, int vcpu, struct iovec *iov, void *vp, size_t len) 1423 { 1424 const char *src; 1425 char *dst; 1426 size_t n; 1427 1428 dst = vp; 1429 while (len) { 1430 assert(iov->iov_len); 1431 n = min(len, iov->iov_len); 1432 src = iov->iov_base; 1433 bcopy(src, dst, n); 1434 1435 iov++; 1436 dst += n; 1437 len -= n; 1438 } 1439 } 1440 1441 void 1442 vm_copyout(struct vmctx *ctx, int vcpu, const void *vp, struct iovec *iov, 1443 size_t len) 1444 { 1445 const char *src; 1446 char *dst; 1447 size_t n; 1448 1449 src = vp; 1450 while (len) { 1451 assert(iov->iov_len); 1452 n = min(len, iov->iov_len); 1453 dst = iov->iov_base; 1454 bcopy(src, dst, n); 1455 1456 iov++; 1457 src += n; 1458 len -= n; 1459 } 1460 } 1461 1462 static int 1463 vm_get_cpus(struct vmctx *ctx, int which, cpuset_t *cpus) 1464 { 1465 struct vm_cpuset vm_cpuset; 1466 int error; 1467 1468 bzero(&vm_cpuset, sizeof(struct vm_cpuset)); 1469 vm_cpuset.which = which; 1470 vm_cpuset.cpusetsize = sizeof(cpuset_t); 1471 vm_cpuset.cpus = cpus; 1472 1473 error = ioctl(ctx->fd, VM_GET_CPUS, &vm_cpuset); 1474 return (error); 1475 } 1476 1477 int 1478 vm_active_cpus(struct vmctx *ctx, cpuset_t *cpus) 1479 { 1480 1481 return (vm_get_cpus(ctx, VM_ACTIVE_CPUS, cpus)); 1482 } 1483 1484 int 1485 vm_suspended_cpus(struct vmctx *ctx, cpuset_t *cpus) 1486 { 1487 1488 return (vm_get_cpus(ctx, VM_SUSPENDED_CPUS, cpus)); 1489 } 1490 1491 int 1492 vm_debug_cpus(struct vmctx *ctx, cpuset_t *cpus) 1493 { 1494 1495 return (vm_get_cpus(ctx, VM_DEBUG_CPUS, cpus)); 1496 } 1497 1498 int 1499 vm_activate_cpu(struct vmctx *ctx, int vcpu) 1500 { 1501 struct vm_activate_cpu ac; 1502 int error; 1503 1504 bzero(&ac, sizeof(struct vm_activate_cpu)); 1505 ac.vcpuid = vcpu; 1506 error = ioctl(ctx->fd, VM_ACTIVATE_CPU, &ac); 1507 return (error); 1508 } 1509 1510 int 1511 vm_suspend_cpu(struct vmctx *ctx, int vcpu) 1512 { 1513 struct vm_activate_cpu ac; 1514 int error; 1515 1516 bzero(&ac, sizeof(struct vm_activate_cpu)); 1517 ac.vcpuid = vcpu; 1518 error = ioctl(ctx->fd, VM_SUSPEND_CPU, &ac); 1519 return (error); 1520 } 1521 1522 int 1523 vm_resume_cpu(struct vmctx *ctx, int vcpu) 1524 { 1525 struct vm_activate_cpu ac; 1526 int error; 1527 1528 bzero(&ac, sizeof(struct vm_activate_cpu)); 1529 ac.vcpuid = vcpu; 1530 error = ioctl(ctx->fd, VM_RESUME_CPU, &ac); 1531 return (error); 1532 } 1533 1534 int 1535 vm_get_intinfo(struct vmctx *ctx, int vcpu, uint64_t *info1, uint64_t *info2) 1536 { 1537 struct vm_intinfo vmii; 1538 int error; 1539 1540 bzero(&vmii, sizeof(struct vm_intinfo)); 1541 vmii.vcpuid = vcpu; 1542 error = ioctl(ctx->fd, VM_GET_INTINFO, &vmii); 1543 if (error == 0) { 1544 *info1 = vmii.info1; 1545 *info2 = vmii.info2; 1546 } 1547 return (error); 1548 } 1549 1550 int 1551 vm_set_intinfo(struct vmctx *ctx, int vcpu, uint64_t info1) 1552 { 1553 struct vm_intinfo vmii; 1554 int error; 1555 1556 bzero(&vmii, sizeof(struct vm_intinfo)); 1557 vmii.vcpuid = vcpu; 1558 vmii.info1 = info1; 1559 error = ioctl(ctx->fd, VM_SET_INTINFO, &vmii); 1560 return (error); 1561 } 1562 1563 int 1564 vm_rtc_write(struct vmctx *ctx, int offset, uint8_t value) 1565 { 1566 struct vm_rtc_data rtcdata; 1567 int error; 1568 1569 bzero(&rtcdata, sizeof(struct vm_rtc_data)); 1570 rtcdata.offset = offset; 1571 rtcdata.value = value; 1572 error = ioctl(ctx->fd, VM_RTC_WRITE, &rtcdata); 1573 return (error); 1574 } 1575 1576 int 1577 vm_rtc_read(struct vmctx *ctx, int offset, uint8_t *retval) 1578 { 1579 struct vm_rtc_data rtcdata; 1580 int error; 1581 1582 bzero(&rtcdata, sizeof(struct vm_rtc_data)); 1583 rtcdata.offset = offset; 1584 error = ioctl(ctx->fd, VM_RTC_READ, &rtcdata); 1585 if (error == 0) 1586 *retval = rtcdata.value; 1587 return (error); 1588 } 1589 1590 int 1591 vm_rtc_settime(struct vmctx *ctx, time_t secs) 1592 { 1593 struct vm_rtc_time rtctime; 1594 int error; 1595 1596 bzero(&rtctime, sizeof(struct vm_rtc_time)); 1597 rtctime.secs = secs; 1598 error = ioctl(ctx->fd, VM_RTC_SETTIME, &rtctime); 1599 return (error); 1600 } 1601 1602 int 1603 vm_rtc_gettime(struct vmctx *ctx, time_t *secs) 1604 { 1605 struct vm_rtc_time rtctime; 1606 int error; 1607 1608 bzero(&rtctime, sizeof(struct vm_rtc_time)); 1609 error = ioctl(ctx->fd, VM_RTC_GETTIME, &rtctime); 1610 if (error == 0) 1611 *secs = rtctime.secs; 1612 return (error); 1613 } 1614 1615 int 1616 vm_restart_instruction(void *arg, int vcpu) 1617 { 1618 struct vmctx *ctx = arg; 1619 1620 return (ioctl(ctx->fd, VM_RESTART_INSTRUCTION, &vcpu)); 1621 } 1622 1623 int 1624 vm_set_topology(struct vmctx *ctx, 1625 uint16_t sockets, uint16_t cores, uint16_t threads, uint16_t maxcpus) 1626 { 1627 struct vm_cpu_topology topology; 1628 1629 bzero(&topology, sizeof (struct vm_cpu_topology)); 1630 topology.sockets = sockets; 1631 topology.cores = cores; 1632 topology.threads = threads; 1633 topology.maxcpus = maxcpus; 1634 return (ioctl(ctx->fd, VM_SET_TOPOLOGY, &topology)); 1635 } 1636 1637 int 1638 vm_get_topology(struct vmctx *ctx, 1639 uint16_t *sockets, uint16_t *cores, uint16_t *threads, uint16_t *maxcpus) 1640 { 1641 struct vm_cpu_topology topology; 1642 int error; 1643 1644 bzero(&topology, sizeof (struct vm_cpu_topology)); 1645 error = ioctl(ctx->fd, VM_GET_TOPOLOGY, &topology); 1646 if (error == 0) { 1647 *sockets = topology.sockets; 1648 *cores = topology.cores; 1649 *threads = topology.threads; 1650 *maxcpus = topology.maxcpus; 1651 } 1652 return (error); 1653 } 1654 1655 int 1656 vm_get_device_fd(struct vmctx *ctx) 1657 { 1658 1659 return (ctx->fd); 1660 } 1661 1662 #ifndef __FreeBSD__ 1663 int 1664 vm_wrlock_cycle(struct vmctx *ctx) 1665 { 1666 if (ioctl(ctx->fd, VM_WRLOCK_CYCLE, 0) != 0) { 1667 return (errno); 1668 } 1669 return (0); 1670 } 1671 #endif /* __FreeBSD__ */ 1672 1673 #ifdef __FreeBSD__ 1674 const cap_ioctl_t * 1675 vm_get_ioctls(size_t *len) 1676 { 1677 cap_ioctl_t *cmds; 1678 /* keep in sync with machine/vmm_dev.h */ 1679 static const cap_ioctl_t vm_ioctl_cmds[] = { VM_RUN, VM_SUSPEND, VM_REINIT, 1680 VM_ALLOC_MEMSEG, VM_GET_MEMSEG, VM_MMAP_MEMSEG, VM_MMAP_MEMSEG, 1681 VM_MMAP_GETNEXT, VM_SET_REGISTER, VM_GET_REGISTER, 1682 VM_SET_SEGMENT_DESCRIPTOR, VM_GET_SEGMENT_DESCRIPTOR, 1683 VM_SET_REGISTER_SET, VM_GET_REGISTER_SET, 1684 VM_INJECT_EXCEPTION, VM_LAPIC_IRQ, VM_LAPIC_LOCAL_IRQ, 1685 VM_LAPIC_MSI, VM_IOAPIC_ASSERT_IRQ, VM_IOAPIC_DEASSERT_IRQ, 1686 VM_IOAPIC_PULSE_IRQ, VM_IOAPIC_PINCOUNT, VM_ISA_ASSERT_IRQ, 1687 VM_ISA_DEASSERT_IRQ, VM_ISA_PULSE_IRQ, VM_ISA_SET_IRQ_TRIGGER, 1688 VM_SET_CAPABILITY, VM_GET_CAPABILITY, VM_BIND_PPTDEV, 1689 VM_UNBIND_PPTDEV, VM_MAP_PPTDEV_MMIO, VM_PPTDEV_MSI, 1690 VM_PPTDEV_MSIX, VM_INJECT_NMI, VM_STATS, VM_STAT_DESC, 1691 VM_SET_X2APIC_STATE, VM_GET_X2APIC_STATE, 1692 VM_GET_HPET_CAPABILITIES, VM_GET_GPA_PMAP, VM_GLA2GPA, 1693 VM_GLA2GPA_NOFAULT, 1694 VM_ACTIVATE_CPU, VM_GET_CPUS, VM_SUSPEND_CPU, VM_RESUME_CPU, 1695 VM_SET_INTINFO, VM_GET_INTINFO, 1696 VM_RTC_WRITE, VM_RTC_READ, VM_RTC_SETTIME, VM_RTC_GETTIME, 1697 VM_RESTART_INSTRUCTION, VM_SET_TOPOLOGY, VM_GET_TOPOLOGY }; 1698 1699 if (len == NULL) { 1700 cmds = malloc(sizeof(vm_ioctl_cmds)); 1701 if (cmds == NULL) 1702 return (NULL); 1703 bcopy(vm_ioctl_cmds, cmds, sizeof(vm_ioctl_cmds)); 1704 return (cmds); 1705 } 1706 1707 *len = nitems(vm_ioctl_cmds); 1708 return (NULL); 1709 } 1710 #endif /* __FreeBSD__ */ 1711