tools/testing/selftests/kvm/lib/x86_64/vmx.c

   1 // SPDX-License-Identifier: GPL-2.0-only
   2 /*
   3  * tools/testing/selftests/kvm/lib/x86_64/vmx.c
   4  *
   5  * Copyright (C) 2018, Google LLC.
   6  */
   7
   8 #include "test_util.h"
   9 #include "kvm_util.h"
  10 #include "../kvm_util_internal.h"
  11 #include "processor.h"
  12 #include "vmx.h"
  13
  14 #define PAGE_SHIFT_4K  12
  15
  16 #define KVM_EPT_PAGE_TABLE_MIN_PADDR 0x1c0000
  17
  18 bool enable_evmcs;
  19
  20 struct eptPageTableEntry {
  21         uint64_t readable:1;
  22         uint64_t writable:1;
  23         uint64_t executable:1;
  24         uint64_t memory_type:3;
  25         uint64_t ignore_pat:1;
  26         uint64_t page_size:1;
  27         uint64_t accessed:1;
  28         uint64_t dirty:1;
  29         uint64_t ignored_11_10:2;
  30         uint64_t address:40;
  31         uint64_t ignored_62_52:11;
  32         uint64_t suppress_ve:1;
  33 };
  34
  35 struct eptPageTablePointer {
  36         uint64_t memory_type:3;
  37         uint64_t page_walk_length:3;
  38         uint64_t ad_enabled:1;
  39         uint64_t reserved_11_07:5;
  40         uint64_t address:40;
  41         uint64_t reserved_63_52:12;
  42 };
  43 int vcpu_enable_evmcs(struct kvm_vm *vm, int vcpu_id)
  44 {
  45         uint16_t evmcs_ver;
  46
  47         struct kvm_enable_cap enable_evmcs_cap = {
  48                 .cap = KVM_CAP_HYPERV_ENLIGHTENED_VMCS,
  49                  .args[0] = (unsigned long)&evmcs_ver
  50         };
  51
  52         vcpu_ioctl(vm, vcpu_id, KVM_ENABLE_CAP, &enable_evmcs_cap);
  53
  54         /* KVM should return supported EVMCS version range */
  55         TEST_ASSERT(((evmcs_ver >> 8) >= (evmcs_ver & 0xff)) &&
  56                     (evmcs_ver & 0xff) > 0,
  57                     "Incorrect EVMCS version range: %x:%x\n",
  58                     evmcs_ver & 0xff, evmcs_ver >> 8);
  59
  60         return evmcs_ver;
  61 }
  62
  63 /* Allocate memory regions for nested VMX tests.
  64  *
  65  * Input Args:
  66  *   vm - The VM to allocate guest-virtual addresses in.
  67  *
  68  * Output Args:
  69  *   p_vmx_gva - The guest virtual address for the struct vmx_pages.
  70  *
  71  * Return:
  72  *   Pointer to structure with the addresses of the VMX areas.
  73  */
  74 struct vmx_pages *
  75 vcpu_alloc_vmx(struct kvm_vm *vm, vm_vaddr_t *p_vmx_gva)
  76 {
  77         vm_vaddr_t vmx_gva = vm_vaddr_alloc(vm, getpagesize(), 0x10000, 0, 0);
  78         struct vmx_pages *vmx = addr_gva2hva(vm, vmx_gva);
  79
  80         /* Setup of a region of guest memory for the vmxon region. */
  81         vmx->vmxon = (void *)vm_vaddr_alloc(vm, getpagesize(), 0x10000, 0, 0);
  82         vmx->vmxon_hva = addr_gva2hva(vm, (uintptr_t)vmx->vmxon);
  83         vmx->vmxon_gpa = addr_gva2gpa(vm, (uintptr_t)vmx->vmxon);
  84
  85         /* Setup of a region of guest memory for a vmcs. */
  86         vmx->vmcs = (void *)vm_vaddr_alloc(vm, getpagesize(), 0x10000, 0, 0);
  87         vmx->vmcs_hva = addr_gva2hva(vm, (uintptr_t)vmx->vmcs);
  88         vmx->vmcs_gpa = addr_gva2gpa(vm, (uintptr_t)vmx->vmcs);
  89
  90         /* Setup of a region of guest memory for the MSR bitmap. */
  91         vmx->msr = (void *)vm_vaddr_alloc(vm, getpagesize(), 0x10000, 0, 0);
  92         vmx->msr_hva = addr_gva2hva(vm, (uintptr_t)vmx->msr);
  93         vmx->msr_gpa = addr_gva2gpa(vm, (uintptr_t)vmx->msr);
  94         memset(vmx->msr_hva, 0, getpagesize());
  95
  96         /* Setup of a region of guest memory for the shadow VMCS. */
  97         vmx->shadow_vmcs = (void *)vm_vaddr_alloc(vm, getpagesize(), 0x10000, 0, 0);
  98         vmx->shadow_vmcs_hva = addr_gva2hva(vm, (uintptr_t)vmx->shadow_vmcs);
  99         vmx->shadow_vmcs_gpa = addr_gva2gpa(vm, (uintptr_t)vmx->shadow_vmcs);
 100
 101         /* Setup of a region of guest memory for the VMREAD and VMWRITE bitmaps. */
 102         vmx->vmread = (void *)vm_vaddr_alloc(vm, getpagesize(), 0x10000, 0, 0);
 103         vmx->vmread_hva = addr_gva2hva(vm, (uintptr_t)vmx->vmread);
 104         vmx->vmread_gpa = addr_gva2gpa(vm, (uintptr_t)vmx->vmread);
 105         memset(vmx->vmread_hva, 0, getpagesize());
 106
 107         vmx->vmwrite = (void *)vm_vaddr_alloc(vm, getpagesize(), 0x10000, 0, 0);
 108         vmx->vmwrite_hva = addr_gva2hva(vm, (uintptr_t)vmx->vmwrite);
 109         vmx->vmwrite_gpa = addr_gva2gpa(vm, (uintptr_t)vmx->vmwrite);
 110         memset(vmx->vmwrite_hva, 0, getpagesize());
 111
 112         /* Setup of a region of guest memory for the VP Assist page. */
 113         vmx->vp_assist = (void *)vm_vaddr_alloc(vm, getpagesize(),
 114                                                 0x10000, 0, 0);
 115         vmx->vp_assist_hva = addr_gva2hva(vm, (uintptr_t)vmx->vp_assist);
 116         vmx->vp_assist_gpa = addr_gva2gpa(vm, (uintptr_t)vmx->vp_assist);
 117
 118         /* Setup of a region of guest memory for the enlightened VMCS. */
 119         vmx->enlightened_vmcs = (void *)vm_vaddr_alloc(vm, getpagesize(),
 120                                                        0x10000, 0, 0);
 121         vmx->enlightened_vmcs_hva =
 122                 addr_gva2hva(vm, (uintptr_t)vmx->enlightened_vmcs);
 123         vmx->enlightened_vmcs_gpa =
 124                 addr_gva2gpa(vm, (uintptr_t)vmx->enlightened_vmcs);
 125
 126         *p_vmx_gva = vmx_gva;
 127         return vmx;
 128 }
 129
 130 bool prepare_for_vmx_operation(struct vmx_pages *vmx)
 131 {
 132         uint64_t feature_control;
 133         uint64_t required;
 134         unsigned long cr0;
 135         unsigned long cr4;
 136
 137         /*
 138          * Ensure bits in CR0 and CR4 are valid in VMX operation:
 139          * - Bit X is 1 in _FIXED0: bit X is fixed to 1 in CRx.
 140          * - Bit X is 0 in _FIXED1: bit X is fixed to 0 in CRx.
 141          */
 142         __asm__ __volatile__("mov %%cr0, %0" : "=r"(cr0) : : "memory");
 143         cr0 &= rdmsr(MSR_IA32_VMX_CR0_FIXED1);
 144         cr0 |= rdmsr(MSR_IA32_VMX_CR0_FIXED0);
 145         __asm__ __volatile__("mov %0, %%cr0" : : "r"(cr0) : "memory");
 146
 147         __asm__ __volatile__("mov %%cr4, %0" : "=r"(cr4) : : "memory");
 148         cr4 &= rdmsr(MSR_IA32_VMX_CR4_FIXED1);
 149         cr4 |= rdmsr(MSR_IA32_VMX_CR4_FIXED0);
 150         /* Enable VMX operation */
 151         cr4 |= X86_CR4_VMXE;
 152         __asm__ __volatile__("mov %0, %%cr4" : : "r"(cr4) : "memory");
 153
 154         /*
 155          * Configure IA32_FEATURE_CONTROL MSR to allow VMXON:
 156          *  Bit 0: Lock bit. If clear, VMXON causes a #GP.
 157          *  Bit 2: Enables VMXON outside of SMX operation. If clear, VMXON
 158          *    outside of SMX causes a #GP.
 159          */
 160         required = FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX;
 161         required |= FEATURE_CONTROL_LOCKED;
 162         feature_control = rdmsr(MSR_IA32_FEATURE_CONTROL);
 163         if ((feature_control & required) != required)
 164                 wrmsr(MSR_IA32_FEATURE_CONTROL, feature_control | required);
 165
 166         /* Enter VMX root operation. */
 167         *(uint32_t *)(vmx->vmxon) = vmcs_revision();
 168         if (vmxon(vmx->vmxon_gpa))
 169                 return false;
 170
 171         return true;
 172 }
 173
 174 bool load_vmcs(struct vmx_pages *vmx)
 175 {
 176         if (!enable_evmcs) {
 177                 /* Load a VMCS. */
 178                 *(uint32_t *)(vmx->vmcs) = vmcs_revision();
 179                 if (vmclear(vmx->vmcs_gpa))
 180                         return false;
 181
 182                 if (vmptrld(vmx->vmcs_gpa))
 183                         return false;
 184
 185                 /* Setup shadow VMCS, do not load it yet. */
 186                 *(uint32_t *)(vmx->shadow_vmcs) =
 187                         vmcs_revision() | 0x80000000ul;
 188                 if (vmclear(vmx->shadow_vmcs_gpa))
 189                         return false;
 190         } else {
 191                 if (evmcs_vmptrld(vmx->enlightened_vmcs_gpa,
 192                                   vmx->enlightened_vmcs))
 193                         return false;
 194                 current_evmcs->revision_id = vmcs_revision();
 195         }
 196
 197         return true;
 198 }
 199
 200 /*
 201  * Initialize the control fields to the most basic settings possible.
 202  */
 203 static inline void init_vmcs_control_fields(struct vmx_pages *vmx)
 204 {
 205         uint32_t sec_exec_ctl = 0;
 206
 207         vmwrite(VIRTUAL_PROCESSOR_ID, 0);
 208         vmwrite(POSTED_INTR_NV, 0);
 209
 210         vmwrite(PIN_BASED_VM_EXEC_CONTROL, rdmsr(MSR_IA32_VMX_TRUE_PINBASED_CTLS));
 211
 212         if (vmx->eptp_gpa) {
 213                 uint64_t ept_paddr;
 214                 struct eptPageTablePointer eptp = {
 215                         .memory_type = VMX_BASIC_MEM_TYPE_WB,
 216                         .page_walk_length = 3, /* + 1 */
 217                         .ad_enabled = !!(rdmsr(MSR_IA32_VMX_EPT_VPID_CAP) & VMX_EPT_VPID_CAP_AD_BITS),
 218                         .address = vmx->eptp_gpa >> PAGE_SHIFT_4K,
 219                 };
 220
 221                 memcpy(&ept_paddr, &eptp, sizeof(ept_paddr));
 222                 vmwrite(EPT_POINTER, ept_paddr);
 223                 sec_exec_ctl |= SECONDARY_EXEC_ENABLE_EPT;
 224         }
 225
 226         if (!vmwrite(SECONDARY_VM_EXEC_CONTROL, sec_exec_ctl))
 227                 vmwrite(CPU_BASED_VM_EXEC_CONTROL,
 228                         rdmsr(MSR_IA32_VMX_TRUE_PROCBASED_CTLS) | CPU_BASED_ACTIVATE_SECONDARY_CONTROLS);
 229         else {
 230                 vmwrite(CPU_BASED_VM_EXEC_CONTROL, rdmsr(MSR_IA32_VMX_TRUE_PROCBASED_CTLS));
 231                 GUEST_ASSERT(!sec_exec_ctl);
 232         }
 233
 234         vmwrite(EXCEPTION_BITMAP, 0);
 235         vmwrite(PAGE_FAULT_ERROR_CODE_MASK, 0);
 236         vmwrite(PAGE_FAULT_ERROR_CODE_MATCH, -1); /* Never match */
 237         vmwrite(CR3_TARGET_COUNT, 0);
 238         vmwrite(VM_EXIT_CONTROLS, rdmsr(MSR_IA32_VMX_EXIT_CTLS) |
 239                 VM_EXIT_HOST_ADDR_SPACE_SIZE);    /* 64-bit host */
 240         vmwrite(VM_EXIT_MSR_STORE_COUNT, 0);
 241         vmwrite(VM_EXIT_MSR_LOAD_COUNT, 0);
 242         vmwrite(VM_ENTRY_CONTROLS, rdmsr(MSR_IA32_VMX_ENTRY_CTLS) |
 243                 VM_ENTRY_IA32E_MODE);             /* 64-bit guest */
 244         vmwrite(VM_ENTRY_MSR_LOAD_COUNT, 0);
 245         vmwrite(VM_ENTRY_INTR_INFO_FIELD, 0);
 246         vmwrite(TPR_THRESHOLD, 0);
 247
 248         vmwrite(CR0_GUEST_HOST_MASK, 0);
 249         vmwrite(CR4_GUEST_HOST_MASK, 0);
 250         vmwrite(CR0_READ_SHADOW, get_cr0());
 251         vmwrite(CR4_READ_SHADOW, get_cr4());
 252
 253         vmwrite(MSR_BITMAP, vmx->msr_gpa);
 254         vmwrite(VMREAD_BITMAP, vmx->vmread_gpa);
 255         vmwrite(VMWRITE_BITMAP, vmx->vmwrite_gpa);
 256 }
 257
 258 /*
 259  * Initialize the host state fields based on the current host state, with
 260  * the exception of HOST_RSP and HOST_RIP, which should be set by vmlaunch
 261  * or vmresume.
 262  */
 263 static inline void init_vmcs_host_state(void)
 264 {
 265         uint32_t exit_controls = vmreadz(VM_EXIT_CONTROLS);
 266
 267         vmwrite(HOST_ES_SELECTOR, get_es());
 268         vmwrite(HOST_CS_SELECTOR, get_cs());
 269         vmwrite(HOST_SS_SELECTOR, get_ss());
 270         vmwrite(HOST_DS_SELECTOR, get_ds());
 271         vmwrite(HOST_FS_SELECTOR, get_fs());
 272         vmwrite(HOST_GS_SELECTOR, get_gs());
 273         vmwrite(HOST_TR_SELECTOR, get_tr());
 274
 275         if (exit_controls & VM_EXIT_LOAD_IA32_PAT)
 276                 vmwrite(HOST_IA32_PAT, rdmsr(MSR_IA32_CR_PAT));
 277         if (exit_controls & VM_EXIT_LOAD_IA32_EFER)
 278                 vmwrite(HOST_IA32_EFER, rdmsr(MSR_EFER));
 279         if (exit_controls & VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL)
 280                 vmwrite(HOST_IA32_PERF_GLOBAL_CTRL,
 281                         rdmsr(MSR_CORE_PERF_GLOBAL_CTRL));
 282
 283         vmwrite(HOST_IA32_SYSENTER_CS, rdmsr(MSR_IA32_SYSENTER_CS));
 284
 285         vmwrite(HOST_CR0, get_cr0());
 286         vmwrite(HOST_CR3, get_cr3());
 287         vmwrite(HOST_CR4, get_cr4());
 288         vmwrite(HOST_FS_BASE, rdmsr(MSR_FS_BASE));
 289         vmwrite(HOST_GS_BASE, rdmsr(MSR_GS_BASE));
 290         vmwrite(HOST_TR_BASE,
 291                 get_desc64_base((struct desc64 *)(get_gdt_base() + get_tr())));
 292         vmwrite(HOST_GDTR_BASE, get_gdt_base());
 293         vmwrite(HOST_IDTR_BASE, get_idt_base());
 294         vmwrite(HOST_IA32_SYSENTER_ESP, rdmsr(MSR_IA32_SYSENTER_ESP));
 295         vmwrite(HOST_IA32_SYSENTER_EIP, rdmsr(MSR_IA32_SYSENTER_EIP));
 296 }
 297
 298 /*
 299  * Initialize the guest state fields essentially as a clone of
 300  * the host state fields. Some host state fields have fixed
 301  * values, and we set the corresponding guest state fields accordingly.
 302  */
 303 static inline void init_vmcs_guest_state(void *rip, void *rsp)
 304 {
 305         vmwrite(GUEST_ES_SELECTOR, vmreadz(HOST_ES_SELECTOR));
 306         vmwrite(GUEST_CS_SELECTOR, vmreadz(HOST_CS_SELECTOR));
 307         vmwrite(GUEST_SS_SELECTOR, vmreadz(HOST_SS_SELECTOR));
 308         vmwrite(GUEST_DS_SELECTOR, vmreadz(HOST_DS_SELECTOR));
 309         vmwrite(GUEST_FS_SELECTOR, vmreadz(HOST_FS_SELECTOR));
 310         vmwrite(GUEST_GS_SELECTOR, vmreadz(HOST_GS_SELECTOR));
 311         vmwrite(GUEST_LDTR_SELECTOR, 0);
 312         vmwrite(GUEST_TR_SELECTOR, vmreadz(HOST_TR_SELECTOR));
 313         vmwrite(GUEST_INTR_STATUS, 0);
 314         vmwrite(GUEST_PML_INDEX, 0);
 315
 316         vmwrite(VMCS_LINK_POINTER, -1ll);
 317         vmwrite(GUEST_IA32_DEBUGCTL, 0);
 318         vmwrite(GUEST_IA32_PAT, vmreadz(HOST_IA32_PAT));
 319         vmwrite(GUEST_IA32_EFER, vmreadz(HOST_IA32_EFER));
 320         vmwrite(GUEST_IA32_PERF_GLOBAL_CTRL,
 321                 vmreadz(HOST_IA32_PERF_GLOBAL_CTRL));
 322
 323         vmwrite(GUEST_ES_LIMIT, -1);
 324         vmwrite(GUEST_CS_LIMIT, -1);
 325         vmwrite(GUEST_SS_LIMIT, -1);
 326         vmwrite(GUEST_DS_LIMIT, -1);
 327         vmwrite(GUEST_FS_LIMIT, -1);
 328         vmwrite(GUEST_GS_LIMIT, -1);
 329         vmwrite(GUEST_LDTR_LIMIT, -1);
 330         vmwrite(GUEST_TR_LIMIT, 0x67);
 331         vmwrite(GUEST_GDTR_LIMIT, 0xffff);
 332         vmwrite(GUEST_IDTR_LIMIT, 0xffff);
 333         vmwrite(GUEST_ES_AR_BYTES,
 334                 vmreadz(GUEST_ES_SELECTOR) == 0 ? 0x10000 : 0xc093);
 335         vmwrite(GUEST_CS_AR_BYTES, 0xa09b);
 336         vmwrite(GUEST_SS_AR_BYTES, 0xc093);
 337         vmwrite(GUEST_DS_AR_BYTES,
 338                 vmreadz(GUEST_DS_SELECTOR) == 0 ? 0x10000 : 0xc093);
 339         vmwrite(GUEST_FS_AR_BYTES,
 340                 vmreadz(GUEST_FS_SELECTOR) == 0 ? 0x10000 : 0xc093);
 341         vmwrite(GUEST_GS_AR_BYTES,
 342                 vmreadz(GUEST_GS_SELECTOR) == 0 ? 0x10000 : 0xc093);
 343         vmwrite(GUEST_LDTR_AR_BYTES, 0x10000);
 344         vmwrite(GUEST_TR_AR_BYTES, 0x8b);
 345         vmwrite(GUEST_INTERRUPTIBILITY_INFO, 0);
 346         vmwrite(GUEST_ACTIVITY_STATE, 0);
 347         vmwrite(GUEST_SYSENTER_CS, vmreadz(HOST_IA32_SYSENTER_CS));
 348         vmwrite(VMX_PREEMPTION_TIMER_VALUE, 0);
 349
 350         vmwrite(GUEST_CR0, vmreadz(HOST_CR0));
 351         vmwrite(GUEST_CR3, vmreadz(HOST_CR3));
 352         vmwrite(GUEST_CR4, vmreadz(HOST_CR4));
 353         vmwrite(GUEST_ES_BASE, 0);
 354         vmwrite(GUEST_CS_BASE, 0);
 355         vmwrite(GUEST_SS_BASE, 0);
 356         vmwrite(GUEST_DS_BASE, 0);
 357         vmwrite(GUEST_FS_BASE, vmreadz(HOST_FS_BASE));
 358         vmwrite(GUEST_GS_BASE, vmreadz(HOST_GS_BASE));
 359         vmwrite(GUEST_LDTR_BASE, 0);
 360         vmwrite(GUEST_TR_BASE, vmreadz(HOST_TR_BASE));
 361         vmwrite(GUEST_GDTR_BASE, vmreadz(HOST_GDTR_BASE));
 362         vmwrite(GUEST_IDTR_BASE, vmreadz(HOST_IDTR_BASE));
 363         vmwrite(GUEST_DR7, 0x400);
 364         vmwrite(GUEST_RSP, (uint64_t)rsp);
 365         vmwrite(GUEST_RIP, (uint64_t)rip);
 366         vmwrite(GUEST_RFLAGS, 2);
 367         vmwrite(GUEST_PENDING_DBG_EXCEPTIONS, 0);
 368         vmwrite(GUEST_SYSENTER_ESP, vmreadz(HOST_IA32_SYSENTER_ESP));
 369         vmwrite(GUEST_SYSENTER_EIP, vmreadz(HOST_IA32_SYSENTER_EIP));
 370 }
 371
 372 void prepare_vmcs(struct vmx_pages *vmx, void *guest_rip, void *guest_rsp)
 373 {
 374         init_vmcs_control_fields(vmx);
 375         init_vmcs_host_state();
 376         init_vmcs_guest_state(guest_rip, guest_rsp);
 377 }
 378
 379 void nested_vmx_check_supported(void)
 380 {
 381         struct kvm_cpuid_entry2 *entry = kvm_get_supported_cpuid_entry(1);
 382
 383         if (!(entry->ecx & CPUID_VMX)) {
 384                 fprintf(stderr, "nested VMX not enabled, skipping test\n");
 385                 exit(KSFT_SKIP);
 386         }
 387 }
 388
 389 void nested_pg_map(struct vmx_pages *vmx, struct kvm_vm *vm,
 390                    uint64_t nested_paddr, uint64_t paddr, uint32_t eptp_memslot)
 391 {
 392         uint16_t index[4];
 393         struct eptPageTableEntry *pml4e;
 394
 395         TEST_ASSERT(vm->mode == VM_MODE_PXXV48_4K, "Attempt to use "
 396                     "unknown or unsupported guest mode, mode: 0x%x", vm->mode);
 397
 398         TEST_ASSERT((nested_paddr % vm->page_size) == 0,
 399                     "Nested physical address not on page boundary,\n"
 400                     "  nested_paddr: 0x%lx vm->page_size: 0x%x",
 401                     nested_paddr, vm->page_size);
 402         TEST_ASSERT((nested_paddr >> vm->page_shift) <= vm->max_gfn,
 403                     "Physical address beyond beyond maximum supported,\n"
 404                     "  nested_paddr: 0x%lx vm->max_gfn: 0x%lx vm->page_size: 0x%x",
 405                     paddr, vm->max_gfn, vm->page_size);
 406         TEST_ASSERT((paddr % vm->page_size) == 0,
 407                     "Physical address not on page boundary,\n"
 408                     "  paddr: 0x%lx vm->page_size: 0x%x",
 409                     paddr, vm->page_size);
 410         TEST_ASSERT((paddr >> vm->page_shift) <= vm->max_gfn,
 411                     "Physical address beyond beyond maximum supported,\n"
 412                     "  paddr: 0x%lx vm->max_gfn: 0x%lx vm->page_size: 0x%x",
 413                     paddr, vm->max_gfn, vm->page_size);
 414
 415         index[0] = (nested_paddr >> 12) & 0x1ffu;
 416         index[1] = (nested_paddr >> 21) & 0x1ffu;
 417         index[2] = (nested_paddr >> 30) & 0x1ffu;
 418         index[3] = (nested_paddr >> 39) & 0x1ffu;
 419
 420         /* Allocate page directory pointer table if not present. */
 421         pml4e = vmx->eptp_hva;
 422         if (!pml4e[index[3]].readable) {
 423                 pml4e[index[3]].address = vm_phy_page_alloc(vm,
 424                           KVM_EPT_PAGE_TABLE_MIN_PADDR, eptp_memslot)
 425                         >> vm->page_shift;
 426                 pml4e[index[3]].writable = true;
 427                 pml4e[index[3]].readable = true;
 428                 pml4e[index[3]].executable = true;
 429         }
 430
 431         /* Allocate page directory table if not present. */
 432         struct eptPageTableEntry *pdpe;
 433         pdpe = addr_gpa2hva(vm, pml4e[index[3]].address * vm->page_size);
 434         if (!pdpe[index[2]].readable) {
 435                 pdpe[index[2]].address = vm_phy_page_alloc(vm,
 436                           KVM_EPT_PAGE_TABLE_MIN_PADDR, eptp_memslot)
 437                         >> vm->page_shift;
 438                 pdpe[index[2]].writable = true;
 439                 pdpe[index[2]].readable = true;
 440                 pdpe[index[2]].executable = true;
 441         }
 442
 443         /* Allocate page table if not present. */
 444         struct eptPageTableEntry *pde;
 445         pde = addr_gpa2hva(vm, pdpe[index[2]].address * vm->page_size);
 446         if (!pde[index[1]].readable) {
 447                 pde[index[1]].address = vm_phy_page_alloc(vm,
 448                           KVM_EPT_PAGE_TABLE_MIN_PADDR, eptp_memslot)
 449                         >> vm->page_shift;
 450                 pde[index[1]].writable = true;
 451                 pde[index[1]].readable = true;
 452                 pde[index[1]].executable = true;
 453         }
 454
 455         /* Fill in page table entry. */
 456         struct eptPageTableEntry *pte;
 457         pte = addr_gpa2hva(vm, pde[index[1]].address * vm->page_size);
 458         pte[index[0]].address = paddr >> vm->page_shift;
 459         pte[index[0]].writable = true;
 460         pte[index[0]].readable = true;
 461         pte[index[0]].executable = true;
 462
 463         /*
 464          * For now mark these as accessed and dirty because the only
 465          * testcase we have needs that.  Can be reconsidered later.
 466          */
 467         pte[index[0]].accessed = true;
 468         pte[index[0]].dirty = true;
 469 }
 470
 471 /*
 472  * Map a range of EPT guest physical addresses to the VM's physical address
 473  *
 474  * Input Args:
 475  *   vm - Virtual Machine
 476  *   nested_paddr - Nested guest physical address to map
 477  *   paddr - VM Physical Address
 478  *   size - The size of the range to map
 479  *   eptp_memslot - Memory region slot for new virtual translation tables
 480  *
 481  * Output Args: None
 482  *
 483  * Return: None
 484  *
 485  * Within the VM given by vm, creates a nested guest translation for the
 486  * page range starting at nested_paddr to the page range starting at paddr.
 487  */
 488 void nested_map(struct vmx_pages *vmx, struct kvm_vm *vm,
 489                 uint64_t nested_paddr, uint64_t paddr, uint64_t size,
 490                 uint32_t eptp_memslot)
 491 {
 492         size_t page_size = vm->page_size;
 493         size_t npages = size / page_size;
 494
 495         TEST_ASSERT(nested_paddr + size > nested_paddr, "Vaddr overflow");
 496         TEST_ASSERT(paddr + size > paddr, "Paddr overflow");
 497
 498         while (npages--) {
 499                 nested_pg_map(vmx, vm, nested_paddr, paddr, eptp_memslot);
 500                 nested_paddr += page_size;
 501                 paddr += page_size;
 502         }
 503 }
 504
 505 /* Prepare an identity extended page table that maps all the
 506  * physical pages in VM.
 507  */
 508 void nested_map_memslot(struct vmx_pages *vmx, struct kvm_vm *vm,
 509                         uint32_t memslot, uint32_t eptp_memslot)
 510 {
 511         sparsebit_idx_t i, last;
 512         struct userspace_mem_region *region =
 513                 memslot2region(vm, memslot);
 514
 515         i = (region->region.guest_phys_addr >> vm->page_shift) - 1;
 516         last = i + (region->region.memory_size >> vm->page_shift);
 517         for (;;) {
 518                 i = sparsebit_next_clear(region->unused_phy_pages, i);
 519                 if (i > last)
 520                         break;
 521
 522                 nested_map(vmx, vm,
 523                            (uint64_t)i << vm->page_shift,
 524                            (uint64_t)i << vm->page_shift,
 525                            1 << vm->page_shift,
 526                            eptp_memslot);
 527         }
 528 }
 529
 530 void prepare_eptp(struct vmx_pages *vmx, struct kvm_vm *vm,
 531                   uint32_t eptp_memslot)
 532 {
 533         vmx->eptp = (void *)vm_vaddr_alloc(vm, getpagesize(), 0x10000, 0, 0);
 534         vmx->eptp_hva = addr_gva2hva(vm, (uintptr_t)vmx->eptp);
 535         vmx->eptp_gpa = addr_gva2gpa(vm, (uintptr_t)vmx->eptp);
 536 }