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 <asm/msr-index.h>
9 
10 #include "test_util.h"
11 #include "kvm_util.h"
12 #include "processor.h"
13 #include "vmx.h"
14 
15 #define PAGE_SHIFT_4K  12
16 
17 #define KVM_EPT_PAGE_TABLE_MIN_PADDR 0x1c0000
18 
19 bool enable_evmcs;
20 
21 struct hv_enlightened_vmcs *current_evmcs;
22 struct hv_vp_assist_page *current_vp_assist;
23 
24 struct eptPageTableEntry {
25 	uint64_t readable:1;
26 	uint64_t writable:1;
27 	uint64_t executable:1;
28 	uint64_t memory_type:3;
29 	uint64_t ignore_pat:1;
30 	uint64_t page_size:1;
31 	uint64_t accessed:1;
32 	uint64_t dirty:1;
33 	uint64_t ignored_11_10:2;
34 	uint64_t address:40;
35 	uint64_t ignored_62_52:11;
36 	uint64_t suppress_ve:1;
37 };
38 
39 struct eptPageTablePointer {
40 	uint64_t memory_type:3;
41 	uint64_t page_walk_length:3;
42 	uint64_t ad_enabled:1;
43 	uint64_t reserved_11_07:5;
44 	uint64_t address:40;
45 	uint64_t reserved_63_52:12;
46 };
47 int vcpu_enable_evmcs(struct kvm_vcpu *vcpu)
48 {
49 	uint16_t evmcs_ver;
50 
51 	vcpu_enable_cap(vcpu, KVM_CAP_HYPERV_ENLIGHTENED_VMCS,
52 			(unsigned long)&evmcs_ver);
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_page(vm);
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_page(vm);
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_page(vm);
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_page(vm);
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_page(vm);
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_page(vm);
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_page(vm);
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_page(vm);
114 	vmx->vp_assist_hva = addr_gva2hva(vm, (uintptr_t)vmx->vp_assist);
115 	vmx->vp_assist_gpa = addr_gva2gpa(vm, (uintptr_t)vmx->vp_assist);
116 
117 	/* Setup of a region of guest memory for the enlightened VMCS. */
118 	vmx->enlightened_vmcs = (void *)vm_vaddr_alloc_page(vm);
119 	vmx->enlightened_vmcs_hva =
120 		addr_gva2hva(vm, (uintptr_t)vmx->enlightened_vmcs);
121 	vmx->enlightened_vmcs_gpa =
122 		addr_gva2gpa(vm, (uintptr_t)vmx->enlightened_vmcs);
123 
124 	*p_vmx_gva = vmx_gva;
125 	return vmx;
126 }
127 
128 bool prepare_for_vmx_operation(struct vmx_pages *vmx)
129 {
130 	uint64_t feature_control;
131 	uint64_t required;
132 	unsigned long cr0;
133 	unsigned long cr4;
134 
135 	/*
136 	 * Ensure bits in CR0 and CR4 are valid in VMX operation:
137 	 * - Bit X is 1 in _FIXED0: bit X is fixed to 1 in CRx.
138 	 * - Bit X is 0 in _FIXED1: bit X is fixed to 0 in CRx.
139 	 */
140 	__asm__ __volatile__("mov %%cr0, %0" : "=r"(cr0) : : "memory");
141 	cr0 &= rdmsr(MSR_IA32_VMX_CR0_FIXED1);
142 	cr0 |= rdmsr(MSR_IA32_VMX_CR0_FIXED0);
143 	__asm__ __volatile__("mov %0, %%cr0" : : "r"(cr0) : "memory");
144 
145 	__asm__ __volatile__("mov %%cr4, %0" : "=r"(cr4) : : "memory");
146 	cr4 &= rdmsr(MSR_IA32_VMX_CR4_FIXED1);
147 	cr4 |= rdmsr(MSR_IA32_VMX_CR4_FIXED0);
148 	/* Enable VMX operation */
149 	cr4 |= X86_CR4_VMXE;
150 	__asm__ __volatile__("mov %0, %%cr4" : : "r"(cr4) : "memory");
151 
152 	/*
153 	 * Configure IA32_FEATURE_CONTROL MSR to allow VMXON:
154 	 *  Bit 0: Lock bit. If clear, VMXON causes a #GP.
155 	 *  Bit 2: Enables VMXON outside of SMX operation. If clear, VMXON
156 	 *    outside of SMX causes a #GP.
157 	 */
158 	required = FEAT_CTL_VMX_ENABLED_OUTSIDE_SMX;
159 	required |= FEAT_CTL_LOCKED;
160 	feature_control = rdmsr(MSR_IA32_FEAT_CTL);
161 	if ((feature_control & required) != required)
162 		wrmsr(MSR_IA32_FEAT_CTL, feature_control | required);
163 
164 	/* Enter VMX root operation. */
165 	*(uint32_t *)(vmx->vmxon) = vmcs_revision();
166 	if (vmxon(vmx->vmxon_gpa))
167 		return false;
168 
169 	return true;
170 }
171 
172 bool load_vmcs(struct vmx_pages *vmx)
173 {
174 	if (!enable_evmcs) {
175 		/* Load a VMCS. */
176 		*(uint32_t *)(vmx->vmcs) = vmcs_revision();
177 		if (vmclear(vmx->vmcs_gpa))
178 			return false;
179 
180 		if (vmptrld(vmx->vmcs_gpa))
181 			return false;
182 
183 		/* Setup shadow VMCS, do not load it yet. */
184 		*(uint32_t *)(vmx->shadow_vmcs) =
185 			vmcs_revision() | 0x80000000ul;
186 		if (vmclear(vmx->shadow_vmcs_gpa))
187 			return false;
188 	} else {
189 		if (evmcs_vmptrld(vmx->enlightened_vmcs_gpa,
190 				  vmx->enlightened_vmcs))
191 			return false;
192 		current_evmcs->revision_id = EVMCS_VERSION;
193 	}
194 
195 	return true;
196 }
197 
198 static bool ept_vpid_cap_supported(uint64_t mask)
199 {
200 	return rdmsr(MSR_IA32_VMX_EPT_VPID_CAP) & mask;
201 }
202 
203 bool ept_1g_pages_supported(void)
204 {
205 	return ept_vpid_cap_supported(VMX_EPT_VPID_CAP_1G_PAGES);
206 }
207 
208 /*
209  * Initialize the control fields to the most basic settings possible.
210  */
211 static inline void init_vmcs_control_fields(struct vmx_pages *vmx)
212 {
213 	uint32_t sec_exec_ctl = 0;
214 
215 	vmwrite(VIRTUAL_PROCESSOR_ID, 0);
216 	vmwrite(POSTED_INTR_NV, 0);
217 
218 	vmwrite(PIN_BASED_VM_EXEC_CONTROL, rdmsr(MSR_IA32_VMX_TRUE_PINBASED_CTLS));
219 
220 	if (vmx->eptp_gpa) {
221 		uint64_t ept_paddr;
222 		struct eptPageTablePointer eptp = {
223 			.memory_type = VMX_BASIC_MEM_TYPE_WB,
224 			.page_walk_length = 3, /* + 1 */
225 			.ad_enabled = ept_vpid_cap_supported(VMX_EPT_VPID_CAP_AD_BITS),
226 			.address = vmx->eptp_gpa >> PAGE_SHIFT_4K,
227 		};
228 
229 		memcpy(&ept_paddr, &eptp, sizeof(ept_paddr));
230 		vmwrite(EPT_POINTER, ept_paddr);
231 		sec_exec_ctl |= SECONDARY_EXEC_ENABLE_EPT;
232 	}
233 
234 	if (!vmwrite(SECONDARY_VM_EXEC_CONTROL, sec_exec_ctl))
235 		vmwrite(CPU_BASED_VM_EXEC_CONTROL,
236 			rdmsr(MSR_IA32_VMX_TRUE_PROCBASED_CTLS) | CPU_BASED_ACTIVATE_SECONDARY_CONTROLS);
237 	else {
238 		vmwrite(CPU_BASED_VM_EXEC_CONTROL, rdmsr(MSR_IA32_VMX_TRUE_PROCBASED_CTLS));
239 		GUEST_ASSERT(!sec_exec_ctl);
240 	}
241 
242 	vmwrite(EXCEPTION_BITMAP, 0);
243 	vmwrite(PAGE_FAULT_ERROR_CODE_MASK, 0);
244 	vmwrite(PAGE_FAULT_ERROR_CODE_MATCH, -1); /* Never match */
245 	vmwrite(CR3_TARGET_COUNT, 0);
246 	vmwrite(VM_EXIT_CONTROLS, rdmsr(MSR_IA32_VMX_EXIT_CTLS) |
247 		VM_EXIT_HOST_ADDR_SPACE_SIZE);	  /* 64-bit host */
248 	vmwrite(VM_EXIT_MSR_STORE_COUNT, 0);
249 	vmwrite(VM_EXIT_MSR_LOAD_COUNT, 0);
250 	vmwrite(VM_ENTRY_CONTROLS, rdmsr(MSR_IA32_VMX_ENTRY_CTLS) |
251 		VM_ENTRY_IA32E_MODE);		  /* 64-bit guest */
252 	vmwrite(VM_ENTRY_MSR_LOAD_COUNT, 0);
253 	vmwrite(VM_ENTRY_INTR_INFO_FIELD, 0);
254 	vmwrite(TPR_THRESHOLD, 0);
255 
256 	vmwrite(CR0_GUEST_HOST_MASK, 0);
257 	vmwrite(CR4_GUEST_HOST_MASK, 0);
258 	vmwrite(CR0_READ_SHADOW, get_cr0());
259 	vmwrite(CR4_READ_SHADOW, get_cr4());
260 
261 	vmwrite(MSR_BITMAP, vmx->msr_gpa);
262 	vmwrite(VMREAD_BITMAP, vmx->vmread_gpa);
263 	vmwrite(VMWRITE_BITMAP, vmx->vmwrite_gpa);
264 }
265 
266 /*
267  * Initialize the host state fields based on the current host state, with
268  * the exception of HOST_RSP and HOST_RIP, which should be set by vmlaunch
269  * or vmresume.
270  */
271 static inline void init_vmcs_host_state(void)
272 {
273 	uint32_t exit_controls = vmreadz(VM_EXIT_CONTROLS);
274 
275 	vmwrite(HOST_ES_SELECTOR, get_es());
276 	vmwrite(HOST_CS_SELECTOR, get_cs());
277 	vmwrite(HOST_SS_SELECTOR, get_ss());
278 	vmwrite(HOST_DS_SELECTOR, get_ds());
279 	vmwrite(HOST_FS_SELECTOR, get_fs());
280 	vmwrite(HOST_GS_SELECTOR, get_gs());
281 	vmwrite(HOST_TR_SELECTOR, get_tr());
282 
283 	if (exit_controls & VM_EXIT_LOAD_IA32_PAT)
284 		vmwrite(HOST_IA32_PAT, rdmsr(MSR_IA32_CR_PAT));
285 	if (exit_controls & VM_EXIT_LOAD_IA32_EFER)
286 		vmwrite(HOST_IA32_EFER, rdmsr(MSR_EFER));
287 	if (exit_controls & VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL)
288 		vmwrite(HOST_IA32_PERF_GLOBAL_CTRL,
289 			rdmsr(MSR_CORE_PERF_GLOBAL_CTRL));
290 
291 	vmwrite(HOST_IA32_SYSENTER_CS, rdmsr(MSR_IA32_SYSENTER_CS));
292 
293 	vmwrite(HOST_CR0, get_cr0());
294 	vmwrite(HOST_CR3, get_cr3());
295 	vmwrite(HOST_CR4, get_cr4());
296 	vmwrite(HOST_FS_BASE, rdmsr(MSR_FS_BASE));
297 	vmwrite(HOST_GS_BASE, rdmsr(MSR_GS_BASE));
298 	vmwrite(HOST_TR_BASE,
299 		get_desc64_base((struct desc64 *)(get_gdt().address + get_tr())));
300 	vmwrite(HOST_GDTR_BASE, get_gdt().address);
301 	vmwrite(HOST_IDTR_BASE, get_idt().address);
302 	vmwrite(HOST_IA32_SYSENTER_ESP, rdmsr(MSR_IA32_SYSENTER_ESP));
303 	vmwrite(HOST_IA32_SYSENTER_EIP, rdmsr(MSR_IA32_SYSENTER_EIP));
304 }
305 
306 /*
307  * Initialize the guest state fields essentially as a clone of
308  * the host state fields. Some host state fields have fixed
309  * values, and we set the corresponding guest state fields accordingly.
310  */
311 static inline void init_vmcs_guest_state(void *rip, void *rsp)
312 {
313 	vmwrite(GUEST_ES_SELECTOR, vmreadz(HOST_ES_SELECTOR));
314 	vmwrite(GUEST_CS_SELECTOR, vmreadz(HOST_CS_SELECTOR));
315 	vmwrite(GUEST_SS_SELECTOR, vmreadz(HOST_SS_SELECTOR));
316 	vmwrite(GUEST_DS_SELECTOR, vmreadz(HOST_DS_SELECTOR));
317 	vmwrite(GUEST_FS_SELECTOR, vmreadz(HOST_FS_SELECTOR));
318 	vmwrite(GUEST_GS_SELECTOR, vmreadz(HOST_GS_SELECTOR));
319 	vmwrite(GUEST_LDTR_SELECTOR, 0);
320 	vmwrite(GUEST_TR_SELECTOR, vmreadz(HOST_TR_SELECTOR));
321 	vmwrite(GUEST_INTR_STATUS, 0);
322 	vmwrite(GUEST_PML_INDEX, 0);
323 
324 	vmwrite(VMCS_LINK_POINTER, -1ll);
325 	vmwrite(GUEST_IA32_DEBUGCTL, 0);
326 	vmwrite(GUEST_IA32_PAT, vmreadz(HOST_IA32_PAT));
327 	vmwrite(GUEST_IA32_EFER, vmreadz(HOST_IA32_EFER));
328 	vmwrite(GUEST_IA32_PERF_GLOBAL_CTRL,
329 		vmreadz(HOST_IA32_PERF_GLOBAL_CTRL));
330 
331 	vmwrite(GUEST_ES_LIMIT, -1);
332 	vmwrite(GUEST_CS_LIMIT, -1);
333 	vmwrite(GUEST_SS_LIMIT, -1);
334 	vmwrite(GUEST_DS_LIMIT, -1);
335 	vmwrite(GUEST_FS_LIMIT, -1);
336 	vmwrite(GUEST_GS_LIMIT, -1);
337 	vmwrite(GUEST_LDTR_LIMIT, -1);
338 	vmwrite(GUEST_TR_LIMIT, 0x67);
339 	vmwrite(GUEST_GDTR_LIMIT, 0xffff);
340 	vmwrite(GUEST_IDTR_LIMIT, 0xffff);
341 	vmwrite(GUEST_ES_AR_BYTES,
342 		vmreadz(GUEST_ES_SELECTOR) == 0 ? 0x10000 : 0xc093);
343 	vmwrite(GUEST_CS_AR_BYTES, 0xa09b);
344 	vmwrite(GUEST_SS_AR_BYTES, 0xc093);
345 	vmwrite(GUEST_DS_AR_BYTES,
346 		vmreadz(GUEST_DS_SELECTOR) == 0 ? 0x10000 : 0xc093);
347 	vmwrite(GUEST_FS_AR_BYTES,
348 		vmreadz(GUEST_FS_SELECTOR) == 0 ? 0x10000 : 0xc093);
349 	vmwrite(GUEST_GS_AR_BYTES,
350 		vmreadz(GUEST_GS_SELECTOR) == 0 ? 0x10000 : 0xc093);
351 	vmwrite(GUEST_LDTR_AR_BYTES, 0x10000);
352 	vmwrite(GUEST_TR_AR_BYTES, 0x8b);
353 	vmwrite(GUEST_INTERRUPTIBILITY_INFO, 0);
354 	vmwrite(GUEST_ACTIVITY_STATE, 0);
355 	vmwrite(GUEST_SYSENTER_CS, vmreadz(HOST_IA32_SYSENTER_CS));
356 	vmwrite(VMX_PREEMPTION_TIMER_VALUE, 0);
357 
358 	vmwrite(GUEST_CR0, vmreadz(HOST_CR0));
359 	vmwrite(GUEST_CR3, vmreadz(HOST_CR3));
360 	vmwrite(GUEST_CR4, vmreadz(HOST_CR4));
361 	vmwrite(GUEST_ES_BASE, 0);
362 	vmwrite(GUEST_CS_BASE, 0);
363 	vmwrite(GUEST_SS_BASE, 0);
364 	vmwrite(GUEST_DS_BASE, 0);
365 	vmwrite(GUEST_FS_BASE, vmreadz(HOST_FS_BASE));
366 	vmwrite(GUEST_GS_BASE, vmreadz(HOST_GS_BASE));
367 	vmwrite(GUEST_LDTR_BASE, 0);
368 	vmwrite(GUEST_TR_BASE, vmreadz(HOST_TR_BASE));
369 	vmwrite(GUEST_GDTR_BASE, vmreadz(HOST_GDTR_BASE));
370 	vmwrite(GUEST_IDTR_BASE, vmreadz(HOST_IDTR_BASE));
371 	vmwrite(GUEST_DR7, 0x400);
372 	vmwrite(GUEST_RSP, (uint64_t)rsp);
373 	vmwrite(GUEST_RIP, (uint64_t)rip);
374 	vmwrite(GUEST_RFLAGS, 2);
375 	vmwrite(GUEST_PENDING_DBG_EXCEPTIONS, 0);
376 	vmwrite(GUEST_SYSENTER_ESP, vmreadz(HOST_IA32_SYSENTER_ESP));
377 	vmwrite(GUEST_SYSENTER_EIP, vmreadz(HOST_IA32_SYSENTER_EIP));
378 }
379 
380 void prepare_vmcs(struct vmx_pages *vmx, void *guest_rip, void *guest_rsp)
381 {
382 	init_vmcs_control_fields(vmx);
383 	init_vmcs_host_state();
384 	init_vmcs_guest_state(guest_rip, guest_rsp);
385 }
386 
387 static void nested_create_pte(struct kvm_vm *vm,
388 			      struct eptPageTableEntry *pte,
389 			      uint64_t nested_paddr,
390 			      uint64_t paddr,
391 			      int current_level,
392 			      int target_level)
393 {
394 	if (!pte->readable) {
395 		pte->writable = true;
396 		pte->readable = true;
397 		pte->executable = true;
398 		pte->page_size = (current_level == target_level);
399 		if (pte->page_size)
400 			pte->address = paddr >> vm->page_shift;
401 		else
402 			pte->address = vm_alloc_page_table(vm) >> vm->page_shift;
403 	} else {
404 		/*
405 		 * Entry already present.  Assert that the caller doesn't want
406 		 * a hugepage at this level, and that there isn't a hugepage at
407 		 * this level.
408 		 */
409 		TEST_ASSERT(current_level != target_level,
410 			    "Cannot create hugepage at level: %u, nested_paddr: 0x%lx\n",
411 			    current_level, nested_paddr);
412 		TEST_ASSERT(!pte->page_size,
413 			    "Cannot create page table at level: %u, nested_paddr: 0x%lx\n",
414 			    current_level, nested_paddr);
415 	}
416 }
417 
418 
419 void __nested_pg_map(struct vmx_pages *vmx, struct kvm_vm *vm,
420 		     uint64_t nested_paddr, uint64_t paddr, int target_level)
421 {
422 	const uint64_t page_size = PG_LEVEL_SIZE(target_level);
423 	struct eptPageTableEntry *pt = vmx->eptp_hva, *pte;
424 	uint16_t index;
425 
426 	TEST_ASSERT(vm->mode == VM_MODE_PXXV48_4K, "Attempt to use "
427 		    "unknown or unsupported guest mode, mode: 0x%x", vm->mode);
428 
429 	TEST_ASSERT((nested_paddr >> 48) == 0,
430 		    "Nested physical address 0x%lx requires 5-level paging",
431 		    nested_paddr);
432 	TEST_ASSERT((nested_paddr % page_size) == 0,
433 		    "Nested physical address not on page boundary,\n"
434 		    "  nested_paddr: 0x%lx page_size: 0x%lx",
435 		    nested_paddr, page_size);
436 	TEST_ASSERT((nested_paddr >> vm->page_shift) <= vm->max_gfn,
437 		    "Physical address beyond beyond maximum supported,\n"
438 		    "  nested_paddr: 0x%lx vm->max_gfn: 0x%lx vm->page_size: 0x%x",
439 		    paddr, vm->max_gfn, vm->page_size);
440 	TEST_ASSERT((paddr % page_size) == 0,
441 		    "Physical address not on page boundary,\n"
442 		    "  paddr: 0x%lx page_size: 0x%lx",
443 		    paddr, page_size);
444 	TEST_ASSERT((paddr >> vm->page_shift) <= vm->max_gfn,
445 		    "Physical address beyond beyond maximum supported,\n"
446 		    "  paddr: 0x%lx vm->max_gfn: 0x%lx vm->page_size: 0x%x",
447 		    paddr, vm->max_gfn, vm->page_size);
448 
449 	for (int level = PG_LEVEL_512G; level >= PG_LEVEL_4K; level--) {
450 		index = (nested_paddr >> PG_LEVEL_SHIFT(level)) & 0x1ffu;
451 		pte = &pt[index];
452 
453 		nested_create_pte(vm, pte, nested_paddr, paddr, level, target_level);
454 
455 		if (pte->page_size)
456 			break;
457 
458 		pt = addr_gpa2hva(vm, pte->address * vm->page_size);
459 	}
460 
461 	/*
462 	 * For now mark these as accessed and dirty because the only
463 	 * testcase we have needs that.  Can be reconsidered later.
464 	 */
465 	pte->accessed = true;
466 	pte->dirty = true;
467 
468 }
469 
470 void nested_pg_map(struct vmx_pages *vmx, struct kvm_vm *vm,
471 		   uint64_t nested_paddr, uint64_t paddr)
472 {
473 	__nested_pg_map(vmx, vm, nested_paddr, paddr, PG_LEVEL_4K);
474 }
475 
476 /*
477  * Map a range of EPT guest physical addresses to the VM's physical address
478  *
479  * Input Args:
480  *   vm - Virtual Machine
481  *   nested_paddr - Nested guest physical address to map
482  *   paddr - VM Physical Address
483  *   size - The size of the range to map
484  *   level - The level at which to map the range
485  *
486  * Output Args: None
487  *
488  * Return: None
489  *
490  * Within the VM given by vm, creates a nested guest translation for the
491  * page range starting at nested_paddr to the page range starting at paddr.
492  */
493 void __nested_map(struct vmx_pages *vmx, struct kvm_vm *vm,
494 		  uint64_t nested_paddr, uint64_t paddr, uint64_t size,
495 		  int level)
496 {
497 	size_t page_size = PG_LEVEL_SIZE(level);
498 	size_t npages = size / page_size;
499 
500 	TEST_ASSERT(nested_paddr + size > nested_paddr, "Vaddr overflow");
501 	TEST_ASSERT(paddr + size > paddr, "Paddr overflow");
502 
503 	while (npages--) {
504 		__nested_pg_map(vmx, vm, nested_paddr, paddr, level);
505 		nested_paddr += page_size;
506 		paddr += page_size;
507 	}
508 }
509 
510 void nested_map(struct vmx_pages *vmx, struct kvm_vm *vm,
511 		uint64_t nested_paddr, uint64_t paddr, uint64_t size)
512 {
513 	__nested_map(vmx, vm, nested_paddr, paddr, size, PG_LEVEL_4K);
514 }
515 
516 /* Prepare an identity extended page table that maps all the
517  * physical pages in VM.
518  */
519 void nested_map_memslot(struct vmx_pages *vmx, struct kvm_vm *vm,
520 			uint32_t memslot)
521 {
522 	sparsebit_idx_t i, last;
523 	struct userspace_mem_region *region =
524 		memslot2region(vm, memslot);
525 
526 	i = (region->region.guest_phys_addr >> vm->page_shift) - 1;
527 	last = i + (region->region.memory_size >> vm->page_shift);
528 	for (;;) {
529 		i = sparsebit_next_clear(region->unused_phy_pages, i);
530 		if (i > last)
531 			break;
532 
533 		nested_map(vmx, vm,
534 			   (uint64_t)i << vm->page_shift,
535 			   (uint64_t)i << vm->page_shift,
536 			   1 << vm->page_shift);
537 	}
538 }
539 
540 /* Identity map a region with 1GiB Pages. */
541 void nested_identity_map_1g(struct vmx_pages *vmx, struct kvm_vm *vm,
542 			    uint64_t addr, uint64_t size)
543 {
544 	__nested_map(vmx, vm, addr, addr, size, PG_LEVEL_1G);
545 }
546 
547 bool kvm_vm_has_ept(struct kvm_vm *vm)
548 {
549 	struct kvm_vcpu *vcpu;
550 	uint64_t ctrl;
551 
552 	vcpu = list_first_entry(&vm->vcpus, struct kvm_vcpu, list);
553 	TEST_ASSERT(vcpu, "Cannot determine EPT support without vCPUs.\n");
554 
555 	ctrl = vcpu_get_msr(vcpu, MSR_IA32_VMX_TRUE_PROCBASED_CTLS) >> 32;
556 	if (!(ctrl & CPU_BASED_ACTIVATE_SECONDARY_CONTROLS))
557 		return false;
558 
559 	ctrl = vcpu_get_msr(vcpu, MSR_IA32_VMX_PROCBASED_CTLS2) >> 32;
560 	return ctrl & SECONDARY_EXEC_ENABLE_EPT;
561 }
562 
563 void prepare_eptp(struct vmx_pages *vmx, struct kvm_vm *vm,
564 		  uint32_t eptp_memslot)
565 {
566 	TEST_REQUIRE(kvm_vm_has_ept(vm));
567 
568 	vmx->eptp = (void *)vm_vaddr_alloc_page(vm);
569 	vmx->eptp_hva = addr_gva2hva(vm, (uintptr_t)vmx->eptp);
570 	vmx->eptp_gpa = addr_gva2gpa(vm, (uintptr_t)vmx->eptp);
571 }
572 
573 void prepare_virtualize_apic_accesses(struct vmx_pages *vmx, struct kvm_vm *vm)
574 {
575 	vmx->apic_access = (void *)vm_vaddr_alloc_page(vm);
576 	vmx->apic_access_hva = addr_gva2hva(vm, (uintptr_t)vmx->apic_access);
577 	vmx->apic_access_gpa = addr_gva2gpa(vm, (uintptr_t)vmx->apic_access);
578 }
579