vmmapi.c revision 268889
1133808Spjd/*- 2156878Spjd * Copyright (c) 2011 NetApp, Inc. 3133808Spjd * All rights reserved. 4133808Spjd * 5133808Spjd * Redistribution and use in source and binary forms, with or without 6133808Spjd * modification, are permitted provided that the following conditions 7133808Spjd * are met: 8133808Spjd * 1. Redistributions of source code must retain the above copyright 9133808Spjd * notice, this list of conditions and the following disclaimer. 10133808Spjd * 2. Redistributions in binary form must reproduce the above copyright 11133808Spjd * notice, this list of conditions and the following disclaimer in the 12133808Spjd * documentation and/or other materials provided with the distribution. 13155174Spjd * 14133808Spjd * THIS SOFTWARE IS PROVIDED BY NETAPP, INC ``AS IS'' AND 15133808Spjd * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 16133808Spjd * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 17133808Spjd * ARE DISCLAIMED. IN NO EVENT SHALL NETAPP, INC OR CONTRIBUTORS BE LIABLE 18133808Spjd * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 19133808Spjd * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 20133808Spjd * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 21133808Spjd * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 22133808Spjd * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 23133808Spjd * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 24133808Spjd * SUCH DAMAGE. 25133808Spjd * 26133808Spjd * $FreeBSD: head/lib/libvmmapi/vmmapi.c 268889 2014-07-19 20:59:08Z neel $ 27133808Spjd */ 28133808Spjd 29133808Spjd#include <sys/cdefs.h> 30133808Spjd__FBSDID("$FreeBSD: head/lib/libvmmapi/vmmapi.c 268889 2014-07-19 20:59:08Z neel $"); 31133808Spjd 32133808Spjd#include <sys/param.h> 33133808Spjd#include <sys/sysctl.h> 34133808Spjd#include <sys/ioctl.h> 35133808Spjd#include <sys/mman.h> 36133808Spjd#include <sys/_iovec.h> 37133808Spjd#include <sys/cpuset.h> 38133808Spjd 39133808Spjd#include <machine/specialreg.h> 40133808Spjd#include <machine/param.h> 41133808Spjd 42133808Spjd#include <stdio.h> 43133808Spjd#include <stdlib.h> 44133808Spjd#include <assert.h> 45133808Spjd#include <string.h> 46133808Spjd#include <fcntl.h> 47133808Spjd#include <unistd.h> 48133808Spjd 49133808Spjd#include <libutil.h> 50133808Spjd 51133808Spjd#include <machine/vmm.h> 52133808Spjd#include <machine/vmm_dev.h> 53133808Spjd 54156612Spjd#include "vmmapi.h" 55133808Spjd 56133808Spjd#define MB (1024 * 1024UL) 57133808Spjd#define GB (1024 * 1024 * 1024UL) 58133808Spjd 59133808Spjdstruct vmctx { 60133808Spjd int fd; 61133808Spjd uint32_t lowmem_limit; 62133808Spjd enum vm_mmap_style vms; 63156612Spjd int memflags; 64156612Spjd size_t lowmem; 65133808Spjd char *lowmem_addr; 66133808Spjd size_t highmem; 67133808Spjd char *highmem_addr; 68156612Spjd char *name; 69133808Spjd}; 70133808Spjd 71133808Spjd#define CREATE(x) sysctlbyname("hw.vmm.create", NULL, NULL, (x), strlen((x))) 72133808Spjd#define DESTROY(x) sysctlbyname("hw.vmm.destroy", NULL, NULL, (x), strlen((x))) 73133808Spjd 74133808Spjdstatic int 75133808Spjdvm_device_open(const char *name) 76133808Spjd{ 77133808Spjd int fd, len; 78156612Spjd char *vmfile; 79133808Spjd 80133808Spjd len = strlen("/dev/vmm/") + strlen(name) + 1; 81133808Spjd vmfile = malloc(len); 82133808Spjd assert(vmfile != NULL); 83133808Spjd snprintf(vmfile, len, "/dev/vmm/%s", name); 84133808Spjd 85133808Spjd /* Open the device file */ 86133808Spjd fd = open(vmfile, O_RDWR, 0); 87133808Spjd 88133808Spjd free(vmfile); 89133808Spjd return (fd); 90133808Spjd} 91133808Spjd 92133808Spjdint 93133808Spjdvm_create(const char *name) 94133808Spjd{ 95133808Spjd 96133808Spjd return (CREATE((char *)name)); 97133808Spjd} 98133808Spjd 99134124Spjdstruct vmctx * 100134168Spjdvm_open(const char *name) 101134168Spjd{ 102134168Spjd struct vmctx *vm; 103133808Spjd 104133808Spjd vm = malloc(sizeof(struct vmctx) + strlen(name) + 1); 105133808Spjd assert(vm != NULL); 106133808Spjd 107144142Spjd vm->fd = -1; 108144142Spjd vm->memflags = 0; 109144142Spjd vm->lowmem_limit = 3 * GB; 110144142Spjd vm->name = (char *)(vm + 1); 111133808Spjd strcpy(vm->name, name); 112133808Spjd 113133808Spjd if ((vm->fd = vm_device_open(vm->name)) < 0) 114133808Spjd goto err; 115133808Spjd 116133808Spjd return (vm); 117133808Spjderr: 118133808Spjd vm_destroy(vm); 119133808Spjd return (NULL); 120133808Spjd} 121133808Spjd 122133808Spjdvoid 123133808Spjdvm_destroy(struct vmctx *vm) 124133808Spjd{ 125133808Spjd assert(vm != NULL); 126133808Spjd 127133808Spjd if (vm->fd >= 0) 128133808Spjd close(vm->fd); 129133808Spjd DESTROY(vm->name); 130134124Spjd 131134124Spjd free(vm); 132134124Spjd} 133134124Spjd 134134124Spjdint 135134124Spjdvm_parse_memsize(const char *optarg, size_t *ret_memsize) 136134124Spjd{ 137134124Spjd char *endptr; 138134124Spjd size_t optval; 139134124Spjd int error; 140134124Spjd 141134124Spjd optval = strtoul(optarg, &endptr, 0); 142134124Spjd if (*optarg != '\0' && *endptr == '\0') { 143134124Spjd /* 144134124Spjd * For the sake of backward compatibility if the memory size 145134124Spjd * specified on the command line is less than a megabyte then 146134168Spjd * it is interpreted as being in units of MB. 147134168Spjd */ 148134168Spjd if (optval < MB) 149134168Spjd optval *= MB; 150134168Spjd *ret_memsize = optval; 151134168Spjd error = 0; 152134168Spjd } else 153134168Spjd error = expand_number(optarg, ret_memsize); 154134168Spjd 155134168Spjd return (error); 156134168Spjd} 157134168Spjd 158134168Spjdint 159134168Spjdvm_get_memory_seg(struct vmctx *ctx, vm_paddr_t gpa, size_t *ret_len, 160134168Spjd int *wired) 161134168Spjd{ 162134168Spjd int error; 163134124Spjd struct vm_memory_segment seg; 164134124Spjd 165134124Spjd bzero(&seg, sizeof(seg)); 166156612Spjd seg.gpa = gpa; 167156612Spjd error = ioctl(ctx->fd, VM_GET_MEMORY_SEG, &seg); 168156612Spjd *ret_len = seg.len; 169156612Spjd if (wired != NULL) 170156612Spjd *wired = seg.wired; 171156612Spjd return (error); 172156612Spjd} 173156612Spjd 174156612Spjduint32_t 175156612Spjdvm_get_lowmem_limit(struct vmctx *ctx) 176156612Spjd{ 177156612Spjd 178156612Spjd return (ctx->lowmem_limit); 179156612Spjd} 180156612Spjd 181133808Spjdvoid 182133808Spjdvm_set_lowmem_limit(struct vmctx *ctx, uint32_t limit) 183133808Spjd{ 184133808Spjd 185133808Spjd ctx->lowmem_limit = limit; 186133808Spjd} 187133808Spjd 188133808Spjdvoid 189133808Spjdvm_set_memflags(struct vmctx *ctx, int flags) 190134168Spjd{ 191134168Spjd 192134168Spjd ctx->memflags = flags; 193134168Spjd} 194134168Spjd 195134168Spjdstatic int 196134168Spjdsetup_memory_segment(struct vmctx *ctx, vm_paddr_t gpa, size_t len, char **addr) 197134124Spjd{ 198134124Spjd int error, mmap_flags; 199134124Spjd struct vm_memory_segment seg; 200134124Spjd 201134124Spjd /* 202134124Spjd * Create and optionally map 'len' bytes of memory at guest 203134124Spjd * physical address 'gpa' 204134168Spjd */ 205134168Spjd bzero(&seg, sizeof(seg)); 206134168Spjd seg.gpa = gpa; 207134168Spjd seg.len = len; 208134168Spjd error = ioctl(ctx->fd, VM_MAP_MEMORY, &seg); 209134168Spjd if (error == 0 && addr != NULL) { 210134168Spjd mmap_flags = MAP_SHARED; 211133808Spjd if ((ctx->memflags & VM_MEM_F_INCORE) == 0) 212133808Spjd mmap_flags |= MAP_NOCORE; 213133808Spjd *addr = mmap(NULL, len, PROT_READ | PROT_WRITE, mmap_flags, 214133808Spjd ctx->fd, gpa); 215133808Spjd } 216133808Spjd return (error); 217133808Spjd} 218133808Spjd 219133808Spjdint 220133808Spjdvm_setup_memory(struct vmctx *ctx, size_t memsize, enum vm_mmap_style vms) 221133808Spjd{ 222133808Spjd char **addr; 223133808Spjd int error; 224133808Spjd 225133808Spjd /* XXX VM_MMAP_SPARSE not implemented yet */ 226133808Spjd assert(vms == VM_MMAP_NONE || vms == VM_MMAP_ALL); 227133808Spjd ctx->vms = vms; 228133808Spjd 229133808Spjd /* 230156612Spjd * If 'memsize' cannot fit entirely in the 'lowmem' segment then 231133808Spjd * create another 'highmem' segment above 4GB for the remainder. 232133808Spjd */ 233133808Spjd if (memsize > ctx->lowmem_limit) { 234133808Spjd ctx->lowmem = ctx->lowmem_limit; 235133808Spjd ctx->highmem = memsize - ctx->lowmem; 236139671Spjd } else { 237133808Spjd ctx->lowmem = memsize; 238133808Spjd ctx->highmem = 0; 239139671Spjd } 240133808Spjd 241139671Spjd if (ctx->lowmem > 0) { 242133808Spjd addr = (vms == VM_MMAP_ALL) ? &ctx->lowmem_addr : NULL; 243133808Spjd error = setup_memory_segment(ctx, 0, ctx->lowmem, addr); 244133808Spjd if (error) 245133808Spjd return (error); 246133808Spjd } 247133808Spjd 248133808Spjd if (ctx->highmem > 0) { 249133808Spjd addr = (vms == VM_MMAP_ALL) ? &ctx->highmem_addr : NULL; 250133808Spjd error = setup_memory_segment(ctx, 4*GB, ctx->highmem, addr); 251133808Spjd if (error) 252133808Spjd return (error); 253133808Spjd } 254133808Spjd 255133808Spjd return (0); 256133808Spjd} 257133808Spjd 258133808Spjdvoid * 259133808Spjdvm_map_gpa(struct vmctx *ctx, vm_paddr_t gaddr, size_t len) 260133808Spjd{ 261133808Spjd 262133808Spjd /* XXX VM_MMAP_SPARSE not implemented yet */ 263133808Spjd assert(ctx->vms == VM_MMAP_ALL); 264133808Spjd 265156612Spjd if (gaddr < ctx->lowmem && gaddr + len <= ctx->lowmem) 266133808Spjd return ((void *)(ctx->lowmem_addr + gaddr)); 267133808Spjd 268133808Spjd if (gaddr >= 4*GB) { 269133808Spjd gaddr -= 4*GB; 270133808Spjd if (gaddr < ctx->highmem && gaddr + len <= ctx->highmem) 271156612Spjd return ((void *)(ctx->highmem_addr + gaddr)); 272133808Spjd } 273133808Spjd 274133808Spjd return (NULL); 275133808Spjd} 276133808Spjd 277156612Spjdsize_t 278133808Spjdvm_get_lowmem_size(struct vmctx *ctx) 279133808Spjd{ 280133808Spjd 281133808Spjd return (ctx->lowmem); 282133808Spjd} 283133808Spjd 284133808Spjdsize_t 285133808Spjdvm_get_highmem_size(struct vmctx *ctx) 286133808Spjd{ 287133808Spjd 288139671Spjd return (ctx->highmem); 289156612Spjd} 290139671Spjd 291156612Spjdint 292133808Spjdvm_set_desc(struct vmctx *ctx, int vcpu, int reg, 293133808Spjd uint64_t base, uint32_t limit, uint32_t access) 294139671Spjd{ 295139671Spjd int error; 296156612Spjd struct vm_seg_desc vmsegdesc; 297139671Spjd 298139671Spjd bzero(&vmsegdesc, sizeof(vmsegdesc)); 299139671Spjd vmsegdesc.cpuid = vcpu; 300156612Spjd vmsegdesc.regnum = reg; 301139671Spjd vmsegdesc.desc.base = base; 302156612Spjd vmsegdesc.desc.limit = limit; 303133808Spjd vmsegdesc.desc.access = access; 304133808Spjd 305133808Spjd error = ioctl(ctx->fd, VM_SET_SEGMENT_DESCRIPTOR, &vmsegdesc); 306133808Spjd return (error); 307133808Spjd} 308133808Spjd 309133808Spjdint 310133808Spjdvm_get_desc(struct vmctx *ctx, int vcpu, int reg, 311133808Spjd uint64_t *base, uint32_t *limit, uint32_t *access) 312133808Spjd{ 313133808Spjd int error; 314133808Spjd struct vm_seg_desc vmsegdesc; 315133808Spjd 316133808Spjd bzero(&vmsegdesc, sizeof(vmsegdesc)); 317133808Spjd vmsegdesc.cpuid = vcpu; 318133808Spjd vmsegdesc.regnum = reg; 319133808Spjd 320133808Spjd error = ioctl(ctx->fd, VM_GET_SEGMENT_DESCRIPTOR, &vmsegdesc); 321133808Spjd if (error == 0) { 322133808Spjd *base = vmsegdesc.desc.base; 323133808Spjd *limit = vmsegdesc.desc.limit; 324133808Spjd *access = vmsegdesc.desc.access; 325133808Spjd } 326133808Spjd return (error); 327133808Spjd} 328133808Spjd 329133808Spjdint 330133808Spjdvm_set_register(struct vmctx *ctx, int vcpu, int reg, uint64_t val) 331133808Spjd{ 332133808Spjd int error; 333133808Spjd struct vm_register vmreg; 334133808Spjd 335133808Spjd bzero(&vmreg, sizeof(vmreg)); 336133808Spjd vmreg.cpuid = vcpu; 337133808Spjd vmreg.regnum = reg; 338133808Spjd vmreg.regval = val; 339133808Spjd 340133808Spjd error = ioctl(ctx->fd, VM_SET_REGISTER, &vmreg); 341133808Spjd return (error); 342133808Spjd} 343133808Spjd 344133808Spjdint 345156612Spjdvm_get_register(struct vmctx *ctx, int vcpu, int reg, uint64_t *ret_val) 346133808Spjd{ 347133808Spjd int error; 348156612Spjd struct vm_register vmreg; 349133808Spjd 350133808Spjd bzero(&vmreg, sizeof(vmreg)); 351133808Spjd vmreg.cpuid = vcpu; 352133808Spjd vmreg.regnum = reg; 353133808Spjd 354133808Spjd error = ioctl(ctx->fd, VM_GET_REGISTER, &vmreg); 355133808Spjd *ret_val = vmreg.regval; 356133808Spjd return (error); 357133808Spjd} 358133808Spjd 359133808Spjdint 360133808Spjdvm_run(struct vmctx *ctx, int vcpu, uint64_t rip, struct vm_exit *vmexit) 361133808Spjd{ 362133808Spjd int error; 363133808Spjd struct vm_run vmrun; 364133808Spjd 365133808Spjd bzero(&vmrun, sizeof(vmrun)); 366133808Spjd vmrun.cpuid = vcpu; 367133808Spjd vmrun.rip = rip; 368133808Spjd 369133808Spjd error = ioctl(ctx->fd, VM_RUN, &vmrun); 370133808Spjd bcopy(&vmrun.vm_exit, vmexit, sizeof(struct vm_exit)); 371134420Spjd return (error); 372133808Spjd} 373133808Spjd 374133808Spjdint 375133808Spjdvm_suspend(struct vmctx *ctx, enum vm_suspend_how how) 376133808Spjd{ 377133808Spjd struct vm_suspend vmsuspend; 378133808Spjd 379133808Spjd bzero(&vmsuspend, sizeof(vmsuspend)); 380133808Spjd vmsuspend.how = how; 381133808Spjd return (ioctl(ctx->fd, VM_SUSPEND, &vmsuspend)); 382133808Spjd} 383133808Spjd 384156612Spjdint 385156612Spjdvm_reinit(struct vmctx *ctx) 386156612Spjd{ 387133808Spjd 388133808Spjd return (ioctl(ctx->fd, VM_REINIT, 0)); 389156612Spjd} 390156612Spjd 391156612Spjdstatic int 392133808Spjdvm_inject_exception_real(struct vmctx *ctx, int vcpu, int vector, 393133808Spjd int error_code, int error_code_valid) 394133808Spjd{ 395133808Spjd struct vm_exception exc; 396133808Spjd 397133808Spjd bzero(&exc, sizeof(exc)); 398133808Spjd exc.cpuid = vcpu; 399156612Spjd exc.vector = vector; 400156612Spjd exc.error_code = error_code; 401156612Spjd exc.error_code_valid = error_code_valid; 402156612Spjd 403156612Spjd return (ioctl(ctx->fd, VM_INJECT_EXCEPTION, &exc)); 404156612Spjd} 405156612Spjd 406156612Spjdint 407156612Spjdvm_inject_exception(struct vmctx *ctx, int vcpu, int vector) 408156612Spjd{ 409156612Spjd 410156612Spjd return (vm_inject_exception_real(ctx, vcpu, vector, 0, 0)); 411156612Spjd} 412156612Spjd 413156612Spjdint 414156612Spjdvm_inject_exception2(struct vmctx *ctx, int vcpu, int vector, int errcode) 415156612Spjd{ 416156612Spjd 417156612Spjd return (vm_inject_exception_real(ctx, vcpu, vector, errcode, 1)); 418156612Spjd} 419156612Spjd 420156612Spjdint 421156612Spjdvm_apicid2vcpu(struct vmctx *ctx, int apicid) 422156612Spjd{ 423156612Spjd /* 424156612Spjd * The apic id associated with the 'vcpu' has the same numerical value 425156612Spjd * as the 'vcpu' itself. 426156612Spjd */ 427156612Spjd return (apicid); 428156612Spjd} 429156612Spjd 430156612Spjdint 431156612Spjdvm_lapic_irq(struct vmctx *ctx, int vcpu, int vector) 432133808Spjd{ 433156612Spjd struct vm_lapic_irq vmirq; 434133808Spjd 435156612Spjd bzero(&vmirq, sizeof(vmirq)); 436133808Spjd vmirq.cpuid = vcpu; 437133808Spjd vmirq.vector = vector; 438133808Spjd 439156612Spjd return (ioctl(ctx->fd, VM_LAPIC_IRQ, &vmirq)); 440133808Spjd} 441156612Spjd 442133808Spjdint 443133808Spjdvm_lapic_local_irq(struct vmctx *ctx, int vcpu, int vector) 444156612Spjd{ 445133808Spjd struct vm_lapic_irq vmirq; 446133808Spjd 447133808Spjd bzero(&vmirq, sizeof(vmirq)); 448156612Spjd vmirq.cpuid = vcpu; 449133808Spjd vmirq.vector = vector; 450134420Spjd 451134420Spjd return (ioctl(ctx->fd, VM_LAPIC_LOCAL_IRQ, &vmirq)); 452156612Spjd} 453134420Spjd 454156612Spjdint 455134420Spjdvm_lapic_msi(struct vmctx *ctx, uint64_t addr, uint64_t msg) 456134420Spjd{ 457134420Spjd struct vm_lapic_msi vmmsi; 458134420Spjd 459134420Spjd bzero(&vmmsi, sizeof(vmmsi)); 460134420Spjd vmmsi.addr = addr; 461134420Spjd vmmsi.msg = msg; 462133808Spjd 463156612Spjd return (ioctl(ctx->fd, VM_LAPIC_MSI, &vmmsi)); 464133808Spjd} 465133808Spjd 466133808Spjdint 467133808Spjdvm_ioapic_assert_irq(struct vmctx *ctx, int irq) 468133808Spjd{ 469133808Spjd struct vm_ioapic_irq ioapic_irq; 470156527Spjd 471133808Spjd bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq)); 472133808Spjd ioapic_irq.irq = irq; 473133808Spjd 474133808Spjd return (ioctl(ctx->fd, VM_IOAPIC_ASSERT_IRQ, &ioapic_irq)); 475133808Spjd} 476133808Spjd 477133808Spjdint 478133808Spjdvm_ioapic_deassert_irq(struct vmctx *ctx, int irq) 479156612Spjd{ 480146118Spjd struct vm_ioapic_irq ioapic_irq; 481146118Spjd 482146118Spjd bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq)); 483146118Spjd ioapic_irq.irq = irq; 484146118Spjd 485146117Spjd return (ioctl(ctx->fd, VM_IOAPIC_DEASSERT_IRQ, &ioapic_irq)); 486156612Spjd} 487133808Spjd 488133808Spjdint 489133808Spjdvm_ioapic_pulse_irq(struct vmctx *ctx, int irq) 490133808Spjd{ 491133808Spjd struct vm_ioapic_irq ioapic_irq; 492133808Spjd 493133808Spjd bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq)); 494133808Spjd ioapic_irq.irq = irq; 495133808Spjd 496133808Spjd return (ioctl(ctx->fd, VM_IOAPIC_PULSE_IRQ, &ioapic_irq)); 497133808Spjd} 498133808Spjd 499133808Spjdint 500133808Spjdvm_ioapic_pincount(struct vmctx *ctx, int *pincount) 501133808Spjd{ 502133808Spjd 503133808Spjd return (ioctl(ctx->fd, VM_IOAPIC_PINCOUNT, pincount)); 504133808Spjd} 505133808Spjd 506133808Spjdint 507156612Spjdvm_isa_assert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq) 508156612Spjd{ 509156612Spjd struct vm_isa_irq isa_irq; 510156612Spjd 511156612Spjd bzero(&isa_irq, sizeof(struct vm_isa_irq)); 512133808Spjd isa_irq.atpic_irq = atpic_irq; 513133808Spjd isa_irq.ioapic_irq = ioapic_irq; 514133808Spjd 515133808Spjd return (ioctl(ctx->fd, VM_ISA_ASSERT_IRQ, &isa_irq)); 516133808Spjd} 517133808Spjd 518133808Spjdint 519133808Spjdvm_isa_deassert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq) 520133808Spjd{ 521133808Spjd struct vm_isa_irq isa_irq; 522133808Spjd 523156612Spjd bzero(&isa_irq, sizeof(struct vm_isa_irq)); 524133808Spjd isa_irq.atpic_irq = atpic_irq; 525133808Spjd isa_irq.ioapic_irq = ioapic_irq; 526133808Spjd 527133808Spjd return (ioctl(ctx->fd, VM_ISA_DEASSERT_IRQ, &isa_irq)); 528133808Spjd} 529133808Spjd 530133808Spjdint 531133808Spjdvm_isa_pulse_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq) 532133808Spjd{ 533133808Spjd struct vm_isa_irq isa_irq; 534133808Spjd 535133808Spjd bzero(&isa_irq, sizeof(struct vm_isa_irq)); 536133808Spjd isa_irq.atpic_irq = atpic_irq; 537133808Spjd isa_irq.ioapic_irq = ioapic_irq; 538156612Spjd 539133808Spjd return (ioctl(ctx->fd, VM_ISA_PULSE_IRQ, &isa_irq)); 540133808Spjd} 541133808Spjd 542133808Spjdint 543133808Spjdvm_isa_set_irq_trigger(struct vmctx *ctx, int atpic_irq, 544133808Spjd enum vm_intr_trigger trigger) 545133808Spjd{ 546139295Spjd struct vm_isa_irq_trigger isa_irq_trigger; 547139295Spjd 548139295Spjd bzero(&isa_irq_trigger, sizeof(struct vm_isa_irq_trigger)); 549156612Spjd isa_irq_trigger.atpic_irq = atpic_irq; 550133808Spjd isa_irq_trigger.trigger = trigger; 551133808Spjd 552133808Spjd return (ioctl(ctx->fd, VM_ISA_SET_IRQ_TRIGGER, &isa_irq_trigger)); 553133808Spjd} 554133808Spjd 555133808Spjdint 556156612Spjdvm_inject_nmi(struct vmctx *ctx, int vcpu) 557133808Spjd{ 558156612Spjd struct vm_nmi vmnmi; 559133808Spjd 560133808Spjd bzero(&vmnmi, sizeof(vmnmi)); 561133808Spjd vmnmi.cpuid = vcpu; 562133808Spjd 563133808Spjd return (ioctl(ctx->fd, VM_INJECT_NMI, &vmnmi)); 564133808Spjd} 565133808Spjd 566133808Spjdstatic struct { 567133808Spjd const char *name; 568133808Spjd int type; 569133808Spjd} capstrmap[] = { 570133808Spjd { "hlt_exit", VM_CAP_HALT_EXIT }, 571133808Spjd { "mtrap_exit", VM_CAP_MTRAP_EXIT }, 572133808Spjd { "pause_exit", VM_CAP_PAUSE_EXIT }, 573133808Spjd { "unrestricted_guest", VM_CAP_UNRESTRICTED_GUEST }, 574133808Spjd { "enable_invpcid", VM_CAP_ENABLE_INVPCID }, 575133808Spjd { 0 } 576133808Spjd}; 577133808Spjd 578156612Spjdint 579133808Spjdvm_capability_name2type(const char *capname) 580133808Spjd{ 581133808Spjd int i; 582133808Spjd 583133808Spjd for (i = 0; capstrmap[i].name != NULL && capname != NULL; i++) { 584133808Spjd if (strcmp(capstrmap[i].name, capname) == 0) 585133808Spjd return (capstrmap[i].type); 586133808Spjd } 587133808Spjd 588133808Spjd return (-1); 589133808Spjd} 590133808Spjd 591156612Spjdconst char * 592133808Spjdvm_capability_type2name(int type) 593{ 594 int i; 595 596 for (i = 0; capstrmap[i].name != NULL; i++) { 597 if (capstrmap[i].type == type) 598 return (capstrmap[i].name); 599 } 600 601 return (NULL); 602} 603 604int 605vm_get_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap, 606 int *retval) 607{ 608 int error; 609 struct vm_capability vmcap; 610 611 bzero(&vmcap, sizeof(vmcap)); 612 vmcap.cpuid = vcpu; 613 vmcap.captype = cap; 614 615 error = ioctl(ctx->fd, VM_GET_CAPABILITY, &vmcap); 616 *retval = vmcap.capval; 617 return (error); 618} 619 620int 621vm_set_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap, int val) 622{ 623 struct vm_capability vmcap; 624 625 bzero(&vmcap, sizeof(vmcap)); 626 vmcap.cpuid = vcpu; 627 vmcap.captype = cap; 628 vmcap.capval = val; 629 630 return (ioctl(ctx->fd, VM_SET_CAPABILITY, &vmcap)); 631} 632 633int 634vm_assign_pptdev(struct vmctx *ctx, int bus, int slot, int func) 635{ 636 struct vm_pptdev pptdev; 637 638 bzero(&pptdev, sizeof(pptdev)); 639 pptdev.bus = bus; 640 pptdev.slot = slot; 641 pptdev.func = func; 642 643 return (ioctl(ctx->fd, VM_BIND_PPTDEV, &pptdev)); 644} 645 646int 647vm_unassign_pptdev(struct vmctx *ctx, int bus, int slot, int func) 648{ 649 struct vm_pptdev pptdev; 650 651 bzero(&pptdev, sizeof(pptdev)); 652 pptdev.bus = bus; 653 pptdev.slot = slot; 654 pptdev.func = func; 655 656 return (ioctl(ctx->fd, VM_UNBIND_PPTDEV, &pptdev)); 657} 658 659int 660vm_map_pptdev_mmio(struct vmctx *ctx, int bus, int slot, int func, 661 vm_paddr_t gpa, size_t len, vm_paddr_t hpa) 662{ 663 struct vm_pptdev_mmio pptmmio; 664 665 bzero(&pptmmio, sizeof(pptmmio)); 666 pptmmio.bus = bus; 667 pptmmio.slot = slot; 668 pptmmio.func = func; 669 pptmmio.gpa = gpa; 670 pptmmio.len = len; 671 pptmmio.hpa = hpa; 672 673 return (ioctl(ctx->fd, VM_MAP_PPTDEV_MMIO, &pptmmio)); 674} 675 676int 677vm_setup_pptdev_msi(struct vmctx *ctx, int vcpu, int bus, int slot, int func, 678 uint64_t addr, uint64_t msg, int numvec) 679{ 680 struct vm_pptdev_msi pptmsi; 681 682 bzero(&pptmsi, sizeof(pptmsi)); 683 pptmsi.vcpu = vcpu; 684 pptmsi.bus = bus; 685 pptmsi.slot = slot; 686 pptmsi.func = func; 687 pptmsi.msg = msg; 688 pptmsi.addr = addr; 689 pptmsi.numvec = numvec; 690 691 return (ioctl(ctx->fd, VM_PPTDEV_MSI, &pptmsi)); 692} 693 694int 695vm_setup_pptdev_msix(struct vmctx *ctx, int vcpu, int bus, int slot, int func, 696 int idx, uint64_t addr, uint64_t msg, uint32_t vector_control) 697{ 698 struct vm_pptdev_msix pptmsix; 699 700 bzero(&pptmsix, sizeof(pptmsix)); 701 pptmsix.vcpu = vcpu; 702 pptmsix.bus = bus; 703 pptmsix.slot = slot; 704 pptmsix.func = func; 705 pptmsix.idx = idx; 706 pptmsix.msg = msg; 707 pptmsix.addr = addr; 708 pptmsix.vector_control = vector_control; 709 710 return ioctl(ctx->fd, VM_PPTDEV_MSIX, &pptmsix); 711} 712 713uint64_t * 714vm_get_stats(struct vmctx *ctx, int vcpu, struct timeval *ret_tv, 715 int *ret_entries) 716{ 717 int error; 718 719 static struct vm_stats vmstats; 720 721 vmstats.cpuid = vcpu; 722 723 error = ioctl(ctx->fd, VM_STATS, &vmstats); 724 if (error == 0) { 725 if (ret_entries) 726 *ret_entries = vmstats.num_entries; 727 if (ret_tv) 728 *ret_tv = vmstats.tv; 729 return (vmstats.statbuf); 730 } else 731 return (NULL); 732} 733 734const char * 735vm_get_stat_desc(struct vmctx *ctx, int index) 736{ 737 static struct vm_stat_desc statdesc; 738 739 statdesc.index = index; 740 if (ioctl(ctx->fd, VM_STAT_DESC, &statdesc) == 0) 741 return (statdesc.desc); 742 else 743 return (NULL); 744} 745 746int 747vm_get_x2apic_state(struct vmctx *ctx, int vcpu, enum x2apic_state *state) 748{ 749 int error; 750 struct vm_x2apic x2apic; 751 752 bzero(&x2apic, sizeof(x2apic)); 753 x2apic.cpuid = vcpu; 754 755 error = ioctl(ctx->fd, VM_GET_X2APIC_STATE, &x2apic); 756 *state = x2apic.state; 757 return (error); 758} 759 760int 761vm_set_x2apic_state(struct vmctx *ctx, int vcpu, enum x2apic_state state) 762{ 763 int error; 764 struct vm_x2apic x2apic; 765 766 bzero(&x2apic, sizeof(x2apic)); 767 x2apic.cpuid = vcpu; 768 x2apic.state = state; 769 770 error = ioctl(ctx->fd, VM_SET_X2APIC_STATE, &x2apic); 771 772 return (error); 773} 774 775/* 776 * From Intel Vol 3a: 777 * Table 9-1. IA-32 Processor States Following Power-up, Reset or INIT 778 */ 779int 780vcpu_reset(struct vmctx *vmctx, int vcpu) 781{ 782 int error; 783 uint64_t rflags, rip, cr0, cr4, zero, desc_base, rdx; 784 uint32_t desc_access, desc_limit; 785 uint16_t sel; 786 787 zero = 0; 788 789 rflags = 0x2; 790 error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RFLAGS, rflags); 791 if (error) 792 goto done; 793 794 rip = 0xfff0; 795 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RIP, rip)) != 0) 796 goto done; 797 798 cr0 = CR0_NE; 799 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR0, cr0)) != 0) 800 goto done; 801 802 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR3, zero)) != 0) 803 goto done; 804 805 cr4 = 0; 806 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR4, cr4)) != 0) 807 goto done; 808 809 /* 810 * CS: present, r/w, accessed, 16-bit, byte granularity, usable 811 */ 812 desc_base = 0xffff0000; 813 desc_limit = 0xffff; 814 desc_access = 0x0093; 815 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_CS, 816 desc_base, desc_limit, desc_access); 817 if (error) 818 goto done; 819 820 sel = 0xf000; 821 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CS, sel)) != 0) 822 goto done; 823 824 /* 825 * SS,DS,ES,FS,GS: present, r/w, accessed, 16-bit, byte granularity 826 */ 827 desc_base = 0; 828 desc_limit = 0xffff; 829 desc_access = 0x0093; 830 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_SS, 831 desc_base, desc_limit, desc_access); 832 if (error) 833 goto done; 834 835 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_DS, 836 desc_base, desc_limit, desc_access); 837 if (error) 838 goto done; 839 840 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_ES, 841 desc_base, desc_limit, desc_access); 842 if (error) 843 goto done; 844 845 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_FS, 846 desc_base, desc_limit, desc_access); 847 if (error) 848 goto done; 849 850 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GS, 851 desc_base, desc_limit, desc_access); 852 if (error) 853 goto done; 854 855 sel = 0; 856 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_SS, sel)) != 0) 857 goto done; 858 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_DS, sel)) != 0) 859 goto done; 860 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_ES, sel)) != 0) 861 goto done; 862 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_FS, sel)) != 0) 863 goto done; 864 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_GS, sel)) != 0) 865 goto done; 866 867 /* General purpose registers */ 868 rdx = 0xf00; 869 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RAX, zero)) != 0) 870 goto done; 871 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBX, zero)) != 0) 872 goto done; 873 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RCX, zero)) != 0) 874 goto done; 875 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDX, rdx)) != 0) 876 goto done; 877 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSI, zero)) != 0) 878 goto done; 879 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDI, zero)) != 0) 880 goto done; 881 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBP, zero)) != 0) 882 goto done; 883 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSP, zero)) != 0) 884 goto done; 885 886 /* GDTR, IDTR */ 887 desc_base = 0; 888 desc_limit = 0xffff; 889 desc_access = 0; 890 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GDTR, 891 desc_base, desc_limit, desc_access); 892 if (error != 0) 893 goto done; 894 895 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_IDTR, 896 desc_base, desc_limit, desc_access); 897 if (error != 0) 898 goto done; 899 900 /* TR */ 901 desc_base = 0; 902 desc_limit = 0xffff; 903 desc_access = 0x0000008b; 904 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_TR, 0, 0, desc_access); 905 if (error) 906 goto done; 907 908 sel = 0; 909 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_TR, sel)) != 0) 910 goto done; 911 912 /* LDTR */ 913 desc_base = 0; 914 desc_limit = 0xffff; 915 desc_access = 0x00000082; 916 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_LDTR, desc_base, 917 desc_limit, desc_access); 918 if (error) 919 goto done; 920 921 sel = 0; 922 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_LDTR, 0)) != 0) 923 goto done; 924 925 /* XXX cr2, debug registers */ 926 927 error = 0; 928done: 929 return (error); 930} 931 932int 933vm_get_gpa_pmap(struct vmctx *ctx, uint64_t gpa, uint64_t *pte, int *num) 934{ 935 int error, i; 936 struct vm_gpa_pte gpapte; 937 938 bzero(&gpapte, sizeof(gpapte)); 939 gpapte.gpa = gpa; 940 941 error = ioctl(ctx->fd, VM_GET_GPA_PMAP, &gpapte); 942 943 if (error == 0) { 944 *num = gpapte.ptenum; 945 for (i = 0; i < gpapte.ptenum; i++) 946 pte[i] = gpapte.pte[i]; 947 } 948 949 return (error); 950} 951 952int 953vm_get_hpet_capabilities(struct vmctx *ctx, uint32_t *capabilities) 954{ 955 int error; 956 struct vm_hpet_cap cap; 957 958 bzero(&cap, sizeof(struct vm_hpet_cap)); 959 error = ioctl(ctx->fd, VM_GET_HPET_CAPABILITIES, &cap); 960 if (capabilities != NULL) 961 *capabilities = cap.capabilities; 962 return (error); 963} 964 965static int 966gla2gpa(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging, 967 uint64_t gla, int prot, int *fault, uint64_t *gpa) 968{ 969 struct vm_gla2gpa gg; 970 int error; 971 972 bzero(&gg, sizeof(struct vm_gla2gpa)); 973 gg.vcpuid = vcpu; 974 gg.prot = prot; 975 gg.gla = gla; 976 gg.paging = *paging; 977 978 error = ioctl(ctx->fd, VM_GLA2GPA, &gg); 979 if (error == 0) { 980 *fault = gg.fault; 981 *gpa = gg.gpa; 982 } 983 return (error); 984} 985 986#ifndef min 987#define min(a,b) (((a) < (b)) ? (a) : (b)) 988#endif 989 990int 991vm_gla2gpa(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging, 992 uint64_t gla, size_t len, int prot, struct iovec *iov, int iovcnt) 993{ 994 uint64_t gpa; 995 int error, fault, i, n, off; 996 997 for (i = 0; i < iovcnt; i++) { 998 iov[i].iov_base = 0; 999 iov[i].iov_len = 0; 1000 } 1001 1002 while (len) { 1003 assert(iovcnt > 0); 1004 error = gla2gpa(ctx, vcpu, paging, gla, prot, &fault, &gpa); 1005 if (error) 1006 return (-1); 1007 if (fault) 1008 return (1); 1009 1010 off = gpa & PAGE_MASK; 1011 n = min(len, PAGE_SIZE - off); 1012 1013 iov->iov_base = (void *)gpa; 1014 iov->iov_len = n; 1015 iov++; 1016 iovcnt--; 1017 1018 gla += n; 1019 len -= n; 1020 } 1021 return (0); 1022} 1023 1024void 1025vm_copyin(struct vmctx *ctx, int vcpu, struct iovec *iov, void *vp, size_t len) 1026{ 1027 const char *src; 1028 char *dst; 1029 uint64_t gpa; 1030 size_t n; 1031 1032 dst = vp; 1033 while (len) { 1034 assert(iov->iov_len); 1035 gpa = (uint64_t)iov->iov_base; 1036 n = min(len, iov->iov_len); 1037 src = vm_map_gpa(ctx, gpa, n); 1038 bcopy(src, dst, n); 1039 1040 iov++; 1041 dst += n; 1042 len -= n; 1043 } 1044} 1045 1046void 1047vm_copyout(struct vmctx *ctx, int vcpu, const void *vp, struct iovec *iov, 1048 size_t len) 1049{ 1050 const char *src; 1051 char *dst; 1052 uint64_t gpa; 1053 size_t n; 1054 1055 src = vp; 1056 while (len) { 1057 assert(iov->iov_len); 1058 gpa = (uint64_t)iov->iov_base; 1059 n = min(len, iov->iov_len); 1060 dst = vm_map_gpa(ctx, gpa, n); 1061 bcopy(src, dst, n); 1062 1063 iov++; 1064 src += n; 1065 len -= n; 1066 } 1067} 1068 1069static int 1070vm_get_cpus(struct vmctx *ctx, int which, cpuset_t *cpus) 1071{ 1072 struct vm_cpuset vm_cpuset; 1073 int error; 1074 1075 bzero(&vm_cpuset, sizeof(struct vm_cpuset)); 1076 vm_cpuset.which = which; 1077 vm_cpuset.cpusetsize = sizeof(cpuset_t); 1078 vm_cpuset.cpus = cpus; 1079 1080 error = ioctl(ctx->fd, VM_GET_CPUS, &vm_cpuset); 1081 return (error); 1082} 1083 1084int 1085vm_active_cpus(struct vmctx *ctx, cpuset_t *cpus) 1086{ 1087 1088 return (vm_get_cpus(ctx, VM_ACTIVE_CPUS, cpus)); 1089} 1090 1091int 1092vm_suspended_cpus(struct vmctx *ctx, cpuset_t *cpus) 1093{ 1094 1095 return (vm_get_cpus(ctx, VM_SUSPENDED_CPUS, cpus)); 1096} 1097 1098int 1099vm_activate_cpu(struct vmctx *ctx, int vcpu) 1100{ 1101 struct vm_activate_cpu ac; 1102 int error; 1103 1104 bzero(&ac, sizeof(struct vm_activate_cpu)); 1105 ac.vcpuid = vcpu; 1106 error = ioctl(ctx->fd, VM_ACTIVATE_CPU, &ac); 1107 return (error); 1108} 1109 1110int 1111vm_get_intinfo(struct vmctx *ctx, int vcpu, uint64_t *info1, uint64_t *info2) 1112{ 1113 struct vm_intinfo vmii; 1114 int error; 1115 1116 bzero(&vmii, sizeof(struct vm_intinfo)); 1117 vmii.vcpuid = vcpu; 1118 error = ioctl(ctx->fd, VM_GET_INTINFO, &vmii); 1119 if (error == 0) { 1120 *info1 = vmii.info1; 1121 *info2 = vmii.info2; 1122 } 1123 return (error); 1124} 1125 1126int 1127vm_set_intinfo(struct vmctx *ctx, int vcpu, uint64_t info1) 1128{ 1129 struct vm_intinfo vmii; 1130 int error; 1131 1132 bzero(&vmii, sizeof(struct vm_intinfo)); 1133 vmii.vcpuid = vcpu; 1134 vmii.info1 = info1; 1135 error = ioctl(ctx->fd, VM_SET_INTINFO, &vmii); 1136 return (error); 1137} 1138