xref: /openbmc/linux/arch/x86/kvm/vmx/nested.c (revision 3bf90eca)
1 // SPDX-License-Identifier: GPL-2.0
2 
3 #include <linux/objtool.h>
4 #include <linux/percpu.h>
5 
6 #include <asm/debugreg.h>
7 #include <asm/mmu_context.h>
8 
9 #include "cpuid.h"
10 #include "hyperv.h"
11 #include "mmu.h"
12 #include "nested.h"
13 #include "pmu.h"
14 #include "sgx.h"
15 #include "trace.h"
16 #include "vmx.h"
17 #include "x86.h"
18 #include "smm.h"
19 
20 static bool __read_mostly enable_shadow_vmcs = 1;
21 module_param_named(enable_shadow_vmcs, enable_shadow_vmcs, bool, S_IRUGO);
22 
23 static bool __read_mostly nested_early_check = 0;
24 module_param(nested_early_check, bool, S_IRUGO);
25 
26 #define CC KVM_NESTED_VMENTER_CONSISTENCY_CHECK
27 
28 /*
29  * Hyper-V requires all of these, so mark them as supported even though
30  * they are just treated the same as all-context.
31  */
32 #define VMX_VPID_EXTENT_SUPPORTED_MASK		\
33 	(VMX_VPID_EXTENT_INDIVIDUAL_ADDR_BIT |	\
34 	VMX_VPID_EXTENT_SINGLE_CONTEXT_BIT |	\
35 	VMX_VPID_EXTENT_GLOBAL_CONTEXT_BIT |	\
36 	VMX_VPID_EXTENT_SINGLE_NON_GLOBAL_BIT)
37 
38 #define VMX_MISC_EMULATED_PREEMPTION_TIMER_RATE 5
39 
40 enum {
41 	VMX_VMREAD_BITMAP,
42 	VMX_VMWRITE_BITMAP,
43 	VMX_BITMAP_NR
44 };
45 static unsigned long *vmx_bitmap[VMX_BITMAP_NR];
46 
47 #define vmx_vmread_bitmap                    (vmx_bitmap[VMX_VMREAD_BITMAP])
48 #define vmx_vmwrite_bitmap                   (vmx_bitmap[VMX_VMWRITE_BITMAP])
49 
50 struct shadow_vmcs_field {
51 	u16	encoding;
52 	u16	offset;
53 };
54 static struct shadow_vmcs_field shadow_read_only_fields[] = {
55 #define SHADOW_FIELD_RO(x, y) { x, offsetof(struct vmcs12, y) },
56 #include "vmcs_shadow_fields.h"
57 };
58 static int max_shadow_read_only_fields =
59 	ARRAY_SIZE(shadow_read_only_fields);
60 
61 static struct shadow_vmcs_field shadow_read_write_fields[] = {
62 #define SHADOW_FIELD_RW(x, y) { x, offsetof(struct vmcs12, y) },
63 #include "vmcs_shadow_fields.h"
64 };
65 static int max_shadow_read_write_fields =
66 	ARRAY_SIZE(shadow_read_write_fields);
67 
68 static void init_vmcs_shadow_fields(void)
69 {
70 	int i, j;
71 
72 	memset(vmx_vmread_bitmap, 0xff, PAGE_SIZE);
73 	memset(vmx_vmwrite_bitmap, 0xff, PAGE_SIZE);
74 
75 	for (i = j = 0; i < max_shadow_read_only_fields; i++) {
76 		struct shadow_vmcs_field entry = shadow_read_only_fields[i];
77 		u16 field = entry.encoding;
78 
79 		if (vmcs_field_width(field) == VMCS_FIELD_WIDTH_U64 &&
80 		    (i + 1 == max_shadow_read_only_fields ||
81 		     shadow_read_only_fields[i + 1].encoding != field + 1))
82 			pr_err("Missing field from shadow_read_only_field %x\n",
83 			       field + 1);
84 
85 		clear_bit(field, vmx_vmread_bitmap);
86 		if (field & 1)
87 #ifdef CONFIG_X86_64
88 			continue;
89 #else
90 			entry.offset += sizeof(u32);
91 #endif
92 		shadow_read_only_fields[j++] = entry;
93 	}
94 	max_shadow_read_only_fields = j;
95 
96 	for (i = j = 0; i < max_shadow_read_write_fields; i++) {
97 		struct shadow_vmcs_field entry = shadow_read_write_fields[i];
98 		u16 field = entry.encoding;
99 
100 		if (vmcs_field_width(field) == VMCS_FIELD_WIDTH_U64 &&
101 		    (i + 1 == max_shadow_read_write_fields ||
102 		     shadow_read_write_fields[i + 1].encoding != field + 1))
103 			pr_err("Missing field from shadow_read_write_field %x\n",
104 			       field + 1);
105 
106 		WARN_ONCE(field >= GUEST_ES_AR_BYTES &&
107 			  field <= GUEST_TR_AR_BYTES,
108 			  "Update vmcs12_write_any() to drop reserved bits from AR_BYTES");
109 
110 		/*
111 		 * PML and the preemption timer can be emulated, but the
112 		 * processor cannot vmwrite to fields that don't exist
113 		 * on bare metal.
114 		 */
115 		switch (field) {
116 		case GUEST_PML_INDEX:
117 			if (!cpu_has_vmx_pml())
118 				continue;
119 			break;
120 		case VMX_PREEMPTION_TIMER_VALUE:
121 			if (!cpu_has_vmx_preemption_timer())
122 				continue;
123 			break;
124 		case GUEST_INTR_STATUS:
125 			if (!cpu_has_vmx_apicv())
126 				continue;
127 			break;
128 		default:
129 			break;
130 		}
131 
132 		clear_bit(field, vmx_vmwrite_bitmap);
133 		clear_bit(field, vmx_vmread_bitmap);
134 		if (field & 1)
135 #ifdef CONFIG_X86_64
136 			continue;
137 #else
138 			entry.offset += sizeof(u32);
139 #endif
140 		shadow_read_write_fields[j++] = entry;
141 	}
142 	max_shadow_read_write_fields = j;
143 }
144 
145 /*
146  * The following 3 functions, nested_vmx_succeed()/failValid()/failInvalid(),
147  * set the success or error code of an emulated VMX instruction (as specified
148  * by Vol 2B, VMX Instruction Reference, "Conventions"), and skip the emulated
149  * instruction.
150  */
151 static int nested_vmx_succeed(struct kvm_vcpu *vcpu)
152 {
153 	vmx_set_rflags(vcpu, vmx_get_rflags(vcpu)
154 			& ~(X86_EFLAGS_CF | X86_EFLAGS_PF | X86_EFLAGS_AF |
155 			    X86_EFLAGS_ZF | X86_EFLAGS_SF | X86_EFLAGS_OF));
156 	return kvm_skip_emulated_instruction(vcpu);
157 }
158 
159 static int nested_vmx_failInvalid(struct kvm_vcpu *vcpu)
160 {
161 	vmx_set_rflags(vcpu, (vmx_get_rflags(vcpu)
162 			& ~(X86_EFLAGS_PF | X86_EFLAGS_AF | X86_EFLAGS_ZF |
163 			    X86_EFLAGS_SF | X86_EFLAGS_OF))
164 			| X86_EFLAGS_CF);
165 	return kvm_skip_emulated_instruction(vcpu);
166 }
167 
168 static int nested_vmx_failValid(struct kvm_vcpu *vcpu,
169 				u32 vm_instruction_error)
170 {
171 	vmx_set_rflags(vcpu, (vmx_get_rflags(vcpu)
172 			& ~(X86_EFLAGS_CF | X86_EFLAGS_PF | X86_EFLAGS_AF |
173 			    X86_EFLAGS_SF | X86_EFLAGS_OF))
174 			| X86_EFLAGS_ZF);
175 	get_vmcs12(vcpu)->vm_instruction_error = vm_instruction_error;
176 	/*
177 	 * We don't need to force sync to shadow VMCS because
178 	 * VM_INSTRUCTION_ERROR is not shadowed. Enlightened VMCS 'shadows' all
179 	 * fields and thus must be synced.
180 	 */
181 	if (to_vmx(vcpu)->nested.hv_evmcs_vmptr != EVMPTR_INVALID)
182 		to_vmx(vcpu)->nested.need_vmcs12_to_shadow_sync = true;
183 
184 	return kvm_skip_emulated_instruction(vcpu);
185 }
186 
187 static int nested_vmx_fail(struct kvm_vcpu *vcpu, u32 vm_instruction_error)
188 {
189 	struct vcpu_vmx *vmx = to_vmx(vcpu);
190 
191 	/*
192 	 * failValid writes the error number to the current VMCS, which
193 	 * can't be done if there isn't a current VMCS.
194 	 */
195 	if (vmx->nested.current_vmptr == INVALID_GPA &&
196 	    !evmptr_is_valid(vmx->nested.hv_evmcs_vmptr))
197 		return nested_vmx_failInvalid(vcpu);
198 
199 	return nested_vmx_failValid(vcpu, vm_instruction_error);
200 }
201 
202 static void nested_vmx_abort(struct kvm_vcpu *vcpu, u32 indicator)
203 {
204 	/* TODO: not to reset guest simply here. */
205 	kvm_make_request(KVM_REQ_TRIPLE_FAULT, vcpu);
206 	pr_debug_ratelimited("kvm: nested vmx abort, indicator %d\n", indicator);
207 }
208 
209 static inline bool vmx_control_verify(u32 control, u32 low, u32 high)
210 {
211 	return fixed_bits_valid(control, low, high);
212 }
213 
214 static inline u64 vmx_control_msr(u32 low, u32 high)
215 {
216 	return low | ((u64)high << 32);
217 }
218 
219 static void vmx_disable_shadow_vmcs(struct vcpu_vmx *vmx)
220 {
221 	secondary_exec_controls_clearbit(vmx, SECONDARY_EXEC_SHADOW_VMCS);
222 	vmcs_write64(VMCS_LINK_POINTER, INVALID_GPA);
223 	vmx->nested.need_vmcs12_to_shadow_sync = false;
224 }
225 
226 static inline void nested_release_evmcs(struct kvm_vcpu *vcpu)
227 {
228 	struct kvm_vcpu_hv *hv_vcpu = to_hv_vcpu(vcpu);
229 	struct vcpu_vmx *vmx = to_vmx(vcpu);
230 
231 	if (evmptr_is_valid(vmx->nested.hv_evmcs_vmptr)) {
232 		kvm_vcpu_unmap(vcpu, &vmx->nested.hv_evmcs_map, true);
233 		vmx->nested.hv_evmcs = NULL;
234 	}
235 
236 	vmx->nested.hv_evmcs_vmptr = EVMPTR_INVALID;
237 
238 	if (hv_vcpu) {
239 		hv_vcpu->nested.pa_page_gpa = INVALID_GPA;
240 		hv_vcpu->nested.vm_id = 0;
241 		hv_vcpu->nested.vp_id = 0;
242 	}
243 }
244 
245 static void vmx_sync_vmcs_host_state(struct vcpu_vmx *vmx,
246 				     struct loaded_vmcs *prev)
247 {
248 	struct vmcs_host_state *dest, *src;
249 
250 	if (unlikely(!vmx->guest_state_loaded))
251 		return;
252 
253 	src = &prev->host_state;
254 	dest = &vmx->loaded_vmcs->host_state;
255 
256 	vmx_set_host_fs_gs(dest, src->fs_sel, src->gs_sel, src->fs_base, src->gs_base);
257 	dest->ldt_sel = src->ldt_sel;
258 #ifdef CONFIG_X86_64
259 	dest->ds_sel = src->ds_sel;
260 	dest->es_sel = src->es_sel;
261 #endif
262 }
263 
264 static void vmx_switch_vmcs(struct kvm_vcpu *vcpu, struct loaded_vmcs *vmcs)
265 {
266 	struct vcpu_vmx *vmx = to_vmx(vcpu);
267 	struct loaded_vmcs *prev;
268 	int cpu;
269 
270 	if (WARN_ON_ONCE(vmx->loaded_vmcs == vmcs))
271 		return;
272 
273 	cpu = get_cpu();
274 	prev = vmx->loaded_vmcs;
275 	vmx->loaded_vmcs = vmcs;
276 	vmx_vcpu_load_vmcs(vcpu, cpu, prev);
277 	vmx_sync_vmcs_host_state(vmx, prev);
278 	put_cpu();
279 
280 	vcpu->arch.regs_avail = ~VMX_REGS_LAZY_LOAD_SET;
281 
282 	/*
283 	 * All lazily updated registers will be reloaded from VMCS12 on both
284 	 * vmentry and vmexit.
285 	 */
286 	vcpu->arch.regs_dirty = 0;
287 }
288 
289 /*
290  * Free whatever needs to be freed from vmx->nested when L1 goes down, or
291  * just stops using VMX.
292  */
293 static void free_nested(struct kvm_vcpu *vcpu)
294 {
295 	struct vcpu_vmx *vmx = to_vmx(vcpu);
296 
297 	if (WARN_ON_ONCE(vmx->loaded_vmcs != &vmx->vmcs01))
298 		vmx_switch_vmcs(vcpu, &vmx->vmcs01);
299 
300 	if (!vmx->nested.vmxon && !vmx->nested.smm.vmxon)
301 		return;
302 
303 	kvm_clear_request(KVM_REQ_GET_NESTED_STATE_PAGES, vcpu);
304 
305 	vmx->nested.vmxon = false;
306 	vmx->nested.smm.vmxon = false;
307 	vmx->nested.vmxon_ptr = INVALID_GPA;
308 	free_vpid(vmx->nested.vpid02);
309 	vmx->nested.posted_intr_nv = -1;
310 	vmx->nested.current_vmptr = INVALID_GPA;
311 	if (enable_shadow_vmcs) {
312 		vmx_disable_shadow_vmcs(vmx);
313 		vmcs_clear(vmx->vmcs01.shadow_vmcs);
314 		free_vmcs(vmx->vmcs01.shadow_vmcs);
315 		vmx->vmcs01.shadow_vmcs = NULL;
316 	}
317 	kfree(vmx->nested.cached_vmcs12);
318 	vmx->nested.cached_vmcs12 = NULL;
319 	kfree(vmx->nested.cached_shadow_vmcs12);
320 	vmx->nested.cached_shadow_vmcs12 = NULL;
321 	/*
322 	 * Unpin physical memory we referred to in the vmcs02.  The APIC access
323 	 * page's backing page (yeah, confusing) shouldn't actually be accessed,
324 	 * and if it is written, the contents are irrelevant.
325 	 */
326 	kvm_vcpu_unmap(vcpu, &vmx->nested.apic_access_page_map, false);
327 	kvm_vcpu_unmap(vcpu, &vmx->nested.virtual_apic_map, true);
328 	kvm_vcpu_unmap(vcpu, &vmx->nested.pi_desc_map, true);
329 	vmx->nested.pi_desc = NULL;
330 
331 	kvm_mmu_free_roots(vcpu->kvm, &vcpu->arch.guest_mmu, KVM_MMU_ROOTS_ALL);
332 
333 	nested_release_evmcs(vcpu);
334 
335 	free_loaded_vmcs(&vmx->nested.vmcs02);
336 }
337 
338 /*
339  * Ensure that the current vmcs of the logical processor is the
340  * vmcs01 of the vcpu before calling free_nested().
341  */
342 void nested_vmx_free_vcpu(struct kvm_vcpu *vcpu)
343 {
344 	vcpu_load(vcpu);
345 	vmx_leave_nested(vcpu);
346 	vcpu_put(vcpu);
347 }
348 
349 #define EPTP_PA_MASK   GENMASK_ULL(51, 12)
350 
351 static bool nested_ept_root_matches(hpa_t root_hpa, u64 root_eptp, u64 eptp)
352 {
353 	return VALID_PAGE(root_hpa) &&
354 	       ((root_eptp & EPTP_PA_MASK) == (eptp & EPTP_PA_MASK));
355 }
356 
357 static void nested_ept_invalidate_addr(struct kvm_vcpu *vcpu, gpa_t eptp,
358 				       gpa_t addr)
359 {
360 	uint i;
361 	struct kvm_mmu_root_info *cached_root;
362 
363 	WARN_ON_ONCE(!mmu_is_nested(vcpu));
364 
365 	for (i = 0; i < KVM_MMU_NUM_PREV_ROOTS; i++) {
366 		cached_root = &vcpu->arch.mmu->prev_roots[i];
367 
368 		if (nested_ept_root_matches(cached_root->hpa, cached_root->pgd,
369 					    eptp))
370 			vcpu->arch.mmu->invlpg(vcpu, addr, cached_root->hpa);
371 	}
372 }
373 
374 static void nested_ept_inject_page_fault(struct kvm_vcpu *vcpu,
375 		struct x86_exception *fault)
376 {
377 	struct vmcs12 *vmcs12 = get_vmcs12(vcpu);
378 	struct vcpu_vmx *vmx = to_vmx(vcpu);
379 	u32 vm_exit_reason;
380 	unsigned long exit_qualification = vcpu->arch.exit_qualification;
381 
382 	if (vmx->nested.pml_full) {
383 		vm_exit_reason = EXIT_REASON_PML_FULL;
384 		vmx->nested.pml_full = false;
385 		exit_qualification &= INTR_INFO_UNBLOCK_NMI;
386 	} else {
387 		if (fault->error_code & PFERR_RSVD_MASK)
388 			vm_exit_reason = EXIT_REASON_EPT_MISCONFIG;
389 		else
390 			vm_exit_reason = EXIT_REASON_EPT_VIOLATION;
391 
392 		/*
393 		 * Although the caller (kvm_inject_emulated_page_fault) would
394 		 * have already synced the faulting address in the shadow EPT
395 		 * tables for the current EPTP12, we also need to sync it for
396 		 * any other cached EPTP02s based on the same EP4TA, since the
397 		 * TLB associates mappings to the EP4TA rather than the full EPTP.
398 		 */
399 		nested_ept_invalidate_addr(vcpu, vmcs12->ept_pointer,
400 					   fault->address);
401 	}
402 
403 	nested_vmx_vmexit(vcpu, vm_exit_reason, 0, exit_qualification);
404 	vmcs12->guest_physical_address = fault->address;
405 }
406 
407 static void nested_ept_new_eptp(struct kvm_vcpu *vcpu)
408 {
409 	struct vcpu_vmx *vmx = to_vmx(vcpu);
410 	bool execonly = vmx->nested.msrs.ept_caps & VMX_EPT_EXECUTE_ONLY_BIT;
411 	int ept_lpage_level = ept_caps_to_lpage_level(vmx->nested.msrs.ept_caps);
412 
413 	kvm_init_shadow_ept_mmu(vcpu, execonly, ept_lpage_level,
414 				nested_ept_ad_enabled(vcpu),
415 				nested_ept_get_eptp(vcpu));
416 }
417 
418 static void nested_ept_init_mmu_context(struct kvm_vcpu *vcpu)
419 {
420 	WARN_ON(mmu_is_nested(vcpu));
421 
422 	vcpu->arch.mmu = &vcpu->arch.guest_mmu;
423 	nested_ept_new_eptp(vcpu);
424 	vcpu->arch.mmu->get_guest_pgd     = nested_ept_get_eptp;
425 	vcpu->arch.mmu->inject_page_fault = nested_ept_inject_page_fault;
426 	vcpu->arch.mmu->get_pdptr         = kvm_pdptr_read;
427 
428 	vcpu->arch.walk_mmu              = &vcpu->arch.nested_mmu;
429 }
430 
431 static void nested_ept_uninit_mmu_context(struct kvm_vcpu *vcpu)
432 {
433 	vcpu->arch.mmu = &vcpu->arch.root_mmu;
434 	vcpu->arch.walk_mmu = &vcpu->arch.root_mmu;
435 }
436 
437 static bool nested_vmx_is_page_fault_vmexit(struct vmcs12 *vmcs12,
438 					    u16 error_code)
439 {
440 	bool inequality, bit;
441 
442 	bit = (vmcs12->exception_bitmap & (1u << PF_VECTOR)) != 0;
443 	inequality =
444 		(error_code & vmcs12->page_fault_error_code_mask) !=
445 		 vmcs12->page_fault_error_code_match;
446 	return inequality ^ bit;
447 }
448 
449 static bool nested_vmx_is_exception_vmexit(struct kvm_vcpu *vcpu, u8 vector,
450 					   u32 error_code)
451 {
452 	struct vmcs12 *vmcs12 = get_vmcs12(vcpu);
453 
454 	/*
455 	 * Drop bits 31:16 of the error code when performing the #PF mask+match
456 	 * check.  All VMCS fields involved are 32 bits, but Intel CPUs never
457 	 * set bits 31:16 and VMX disallows setting bits 31:16 in the injected
458 	 * error code.  Including the to-be-dropped bits in the check might
459 	 * result in an "impossible" or missed exit from L1's perspective.
460 	 */
461 	if (vector == PF_VECTOR)
462 		return nested_vmx_is_page_fault_vmexit(vmcs12, (u16)error_code);
463 
464 	return (vmcs12->exception_bitmap & (1u << vector));
465 }
466 
467 static int nested_vmx_check_io_bitmap_controls(struct kvm_vcpu *vcpu,
468 					       struct vmcs12 *vmcs12)
469 {
470 	if (!nested_cpu_has(vmcs12, CPU_BASED_USE_IO_BITMAPS))
471 		return 0;
472 
473 	if (CC(!page_address_valid(vcpu, vmcs12->io_bitmap_a)) ||
474 	    CC(!page_address_valid(vcpu, vmcs12->io_bitmap_b)))
475 		return -EINVAL;
476 
477 	return 0;
478 }
479 
480 static int nested_vmx_check_msr_bitmap_controls(struct kvm_vcpu *vcpu,
481 						struct vmcs12 *vmcs12)
482 {
483 	if (!nested_cpu_has(vmcs12, CPU_BASED_USE_MSR_BITMAPS))
484 		return 0;
485 
486 	if (CC(!page_address_valid(vcpu, vmcs12->msr_bitmap)))
487 		return -EINVAL;
488 
489 	return 0;
490 }
491 
492 static int nested_vmx_check_tpr_shadow_controls(struct kvm_vcpu *vcpu,
493 						struct vmcs12 *vmcs12)
494 {
495 	if (!nested_cpu_has(vmcs12, CPU_BASED_TPR_SHADOW))
496 		return 0;
497 
498 	if (CC(!page_address_valid(vcpu, vmcs12->virtual_apic_page_addr)))
499 		return -EINVAL;
500 
501 	return 0;
502 }
503 
504 /*
505  * For x2APIC MSRs, ignore the vmcs01 bitmap.  L1 can enable x2APIC without L1
506  * itself utilizing x2APIC.  All MSRs were previously set to be intercepted,
507  * only the "disable intercept" case needs to be handled.
508  */
509 static void nested_vmx_disable_intercept_for_x2apic_msr(unsigned long *msr_bitmap_l1,
510 							unsigned long *msr_bitmap_l0,
511 							u32 msr, int type)
512 {
513 	if (type & MSR_TYPE_R && !vmx_test_msr_bitmap_read(msr_bitmap_l1, msr))
514 		vmx_clear_msr_bitmap_read(msr_bitmap_l0, msr);
515 
516 	if (type & MSR_TYPE_W && !vmx_test_msr_bitmap_write(msr_bitmap_l1, msr))
517 		vmx_clear_msr_bitmap_write(msr_bitmap_l0, msr);
518 }
519 
520 static inline void enable_x2apic_msr_intercepts(unsigned long *msr_bitmap)
521 {
522 	int msr;
523 
524 	for (msr = 0x800; msr <= 0x8ff; msr += BITS_PER_LONG) {
525 		unsigned word = msr / BITS_PER_LONG;
526 
527 		msr_bitmap[word] = ~0;
528 		msr_bitmap[word + (0x800 / sizeof(long))] = ~0;
529 	}
530 }
531 
532 #define BUILD_NVMX_MSR_INTERCEPT_HELPER(rw)					\
533 static inline									\
534 void nested_vmx_set_msr_##rw##_intercept(struct vcpu_vmx *vmx,			\
535 					 unsigned long *msr_bitmap_l1,		\
536 					 unsigned long *msr_bitmap_l0, u32 msr)	\
537 {										\
538 	if (vmx_test_msr_bitmap_##rw(vmx->vmcs01.msr_bitmap, msr) ||		\
539 	    vmx_test_msr_bitmap_##rw(msr_bitmap_l1, msr))			\
540 		vmx_set_msr_bitmap_##rw(msr_bitmap_l0, msr);			\
541 	else									\
542 		vmx_clear_msr_bitmap_##rw(msr_bitmap_l0, msr);			\
543 }
544 BUILD_NVMX_MSR_INTERCEPT_HELPER(read)
545 BUILD_NVMX_MSR_INTERCEPT_HELPER(write)
546 
547 static inline void nested_vmx_set_intercept_for_msr(struct vcpu_vmx *vmx,
548 						    unsigned long *msr_bitmap_l1,
549 						    unsigned long *msr_bitmap_l0,
550 						    u32 msr, int types)
551 {
552 	if (types & MSR_TYPE_R)
553 		nested_vmx_set_msr_read_intercept(vmx, msr_bitmap_l1,
554 						  msr_bitmap_l0, msr);
555 	if (types & MSR_TYPE_W)
556 		nested_vmx_set_msr_write_intercept(vmx, msr_bitmap_l1,
557 						   msr_bitmap_l0, msr);
558 }
559 
560 /*
561  * Merge L0's and L1's MSR bitmap, return false to indicate that
562  * we do not use the hardware.
563  */
564 static inline bool nested_vmx_prepare_msr_bitmap(struct kvm_vcpu *vcpu,
565 						 struct vmcs12 *vmcs12)
566 {
567 	struct vcpu_vmx *vmx = to_vmx(vcpu);
568 	int msr;
569 	unsigned long *msr_bitmap_l1;
570 	unsigned long *msr_bitmap_l0 = vmx->nested.vmcs02.msr_bitmap;
571 	struct hv_enlightened_vmcs *evmcs = vmx->nested.hv_evmcs;
572 	struct kvm_host_map *map = &vmx->nested.msr_bitmap_map;
573 
574 	/* Nothing to do if the MSR bitmap is not in use.  */
575 	if (!cpu_has_vmx_msr_bitmap() ||
576 	    !nested_cpu_has(vmcs12, CPU_BASED_USE_MSR_BITMAPS))
577 		return false;
578 
579 	/*
580 	 * MSR bitmap update can be skipped when:
581 	 * - MSR bitmap for L1 hasn't changed.
582 	 * - Nested hypervisor (L1) is attempting to launch the same L2 as
583 	 *   before.
584 	 * - Nested hypervisor (L1) has enabled 'Enlightened MSR Bitmap' feature
585 	 *   and tells KVM (L0) there were no changes in MSR bitmap for L2.
586 	 */
587 	if (!vmx->nested.force_msr_bitmap_recalc && evmcs &&
588 	    evmcs->hv_enlightenments_control.msr_bitmap &&
589 	    evmcs->hv_clean_fields & HV_VMX_ENLIGHTENED_CLEAN_FIELD_MSR_BITMAP)
590 		return true;
591 
592 	if (kvm_vcpu_map(vcpu, gpa_to_gfn(vmcs12->msr_bitmap), map))
593 		return false;
594 
595 	msr_bitmap_l1 = (unsigned long *)map->hva;
596 
597 	/*
598 	 * To keep the control flow simple, pay eight 8-byte writes (sixteen
599 	 * 4-byte writes on 32-bit systems) up front to enable intercepts for
600 	 * the x2APIC MSR range and selectively toggle those relevant to L2.
601 	 */
602 	enable_x2apic_msr_intercepts(msr_bitmap_l0);
603 
604 	if (nested_cpu_has_virt_x2apic_mode(vmcs12)) {
605 		if (nested_cpu_has_apic_reg_virt(vmcs12)) {
606 			/*
607 			 * L0 need not intercept reads for MSRs between 0x800
608 			 * and 0x8ff, it just lets the processor take the value
609 			 * from the virtual-APIC page; take those 256 bits
610 			 * directly from the L1 bitmap.
611 			 */
612 			for (msr = 0x800; msr <= 0x8ff; msr += BITS_PER_LONG) {
613 				unsigned word = msr / BITS_PER_LONG;
614 
615 				msr_bitmap_l0[word] = msr_bitmap_l1[word];
616 			}
617 		}
618 
619 		nested_vmx_disable_intercept_for_x2apic_msr(
620 			msr_bitmap_l1, msr_bitmap_l0,
621 			X2APIC_MSR(APIC_TASKPRI),
622 			MSR_TYPE_R | MSR_TYPE_W);
623 
624 		if (nested_cpu_has_vid(vmcs12)) {
625 			nested_vmx_disable_intercept_for_x2apic_msr(
626 				msr_bitmap_l1, msr_bitmap_l0,
627 				X2APIC_MSR(APIC_EOI),
628 				MSR_TYPE_W);
629 			nested_vmx_disable_intercept_for_x2apic_msr(
630 				msr_bitmap_l1, msr_bitmap_l0,
631 				X2APIC_MSR(APIC_SELF_IPI),
632 				MSR_TYPE_W);
633 		}
634 	}
635 
636 	/*
637 	 * Always check vmcs01's bitmap to honor userspace MSR filters and any
638 	 * other runtime changes to vmcs01's bitmap, e.g. dynamic pass-through.
639 	 */
640 #ifdef CONFIG_X86_64
641 	nested_vmx_set_intercept_for_msr(vmx, msr_bitmap_l1, msr_bitmap_l0,
642 					 MSR_FS_BASE, MSR_TYPE_RW);
643 
644 	nested_vmx_set_intercept_for_msr(vmx, msr_bitmap_l1, msr_bitmap_l0,
645 					 MSR_GS_BASE, MSR_TYPE_RW);
646 
647 	nested_vmx_set_intercept_for_msr(vmx, msr_bitmap_l1, msr_bitmap_l0,
648 					 MSR_KERNEL_GS_BASE, MSR_TYPE_RW);
649 #endif
650 	nested_vmx_set_intercept_for_msr(vmx, msr_bitmap_l1, msr_bitmap_l0,
651 					 MSR_IA32_SPEC_CTRL, MSR_TYPE_RW);
652 
653 	nested_vmx_set_intercept_for_msr(vmx, msr_bitmap_l1, msr_bitmap_l0,
654 					 MSR_IA32_PRED_CMD, MSR_TYPE_W);
655 
656 	kvm_vcpu_unmap(vcpu, &vmx->nested.msr_bitmap_map, false);
657 
658 	vmx->nested.force_msr_bitmap_recalc = false;
659 
660 	return true;
661 }
662 
663 static void nested_cache_shadow_vmcs12(struct kvm_vcpu *vcpu,
664 				       struct vmcs12 *vmcs12)
665 {
666 	struct vcpu_vmx *vmx = to_vmx(vcpu);
667 	struct gfn_to_hva_cache *ghc = &vmx->nested.shadow_vmcs12_cache;
668 
669 	if (!nested_cpu_has_shadow_vmcs(vmcs12) ||
670 	    vmcs12->vmcs_link_pointer == INVALID_GPA)
671 		return;
672 
673 	if (ghc->gpa != vmcs12->vmcs_link_pointer &&
674 	    kvm_gfn_to_hva_cache_init(vcpu->kvm, ghc,
675 				      vmcs12->vmcs_link_pointer, VMCS12_SIZE))
676 		return;
677 
678 	kvm_read_guest_cached(vmx->vcpu.kvm, ghc, get_shadow_vmcs12(vcpu),
679 			      VMCS12_SIZE);
680 }
681 
682 static void nested_flush_cached_shadow_vmcs12(struct kvm_vcpu *vcpu,
683 					      struct vmcs12 *vmcs12)
684 {
685 	struct vcpu_vmx *vmx = to_vmx(vcpu);
686 	struct gfn_to_hva_cache *ghc = &vmx->nested.shadow_vmcs12_cache;
687 
688 	if (!nested_cpu_has_shadow_vmcs(vmcs12) ||
689 	    vmcs12->vmcs_link_pointer == INVALID_GPA)
690 		return;
691 
692 	if (ghc->gpa != vmcs12->vmcs_link_pointer &&
693 	    kvm_gfn_to_hva_cache_init(vcpu->kvm, ghc,
694 				      vmcs12->vmcs_link_pointer, VMCS12_SIZE))
695 		return;
696 
697 	kvm_write_guest_cached(vmx->vcpu.kvm, ghc, get_shadow_vmcs12(vcpu),
698 			       VMCS12_SIZE);
699 }
700 
701 /*
702  * In nested virtualization, check if L1 has set
703  * VM_EXIT_ACK_INTR_ON_EXIT
704  */
705 static bool nested_exit_intr_ack_set(struct kvm_vcpu *vcpu)
706 {
707 	return get_vmcs12(vcpu)->vm_exit_controls &
708 		VM_EXIT_ACK_INTR_ON_EXIT;
709 }
710 
711 static int nested_vmx_check_apic_access_controls(struct kvm_vcpu *vcpu,
712 					  struct vmcs12 *vmcs12)
713 {
714 	if (nested_cpu_has2(vmcs12, SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES) &&
715 	    CC(!page_address_valid(vcpu, vmcs12->apic_access_addr)))
716 		return -EINVAL;
717 	else
718 		return 0;
719 }
720 
721 static int nested_vmx_check_apicv_controls(struct kvm_vcpu *vcpu,
722 					   struct vmcs12 *vmcs12)
723 {
724 	if (!nested_cpu_has_virt_x2apic_mode(vmcs12) &&
725 	    !nested_cpu_has_apic_reg_virt(vmcs12) &&
726 	    !nested_cpu_has_vid(vmcs12) &&
727 	    !nested_cpu_has_posted_intr(vmcs12))
728 		return 0;
729 
730 	/*
731 	 * If virtualize x2apic mode is enabled,
732 	 * virtualize apic access must be disabled.
733 	 */
734 	if (CC(nested_cpu_has_virt_x2apic_mode(vmcs12) &&
735 	       nested_cpu_has2(vmcs12, SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES)))
736 		return -EINVAL;
737 
738 	/*
739 	 * If virtual interrupt delivery is enabled,
740 	 * we must exit on external interrupts.
741 	 */
742 	if (CC(nested_cpu_has_vid(vmcs12) && !nested_exit_on_intr(vcpu)))
743 		return -EINVAL;
744 
745 	/*
746 	 * bits 15:8 should be zero in posted_intr_nv,
747 	 * the descriptor address has been already checked
748 	 * in nested_get_vmcs12_pages.
749 	 *
750 	 * bits 5:0 of posted_intr_desc_addr should be zero.
751 	 */
752 	if (nested_cpu_has_posted_intr(vmcs12) &&
753 	   (CC(!nested_cpu_has_vid(vmcs12)) ||
754 	    CC(!nested_exit_intr_ack_set(vcpu)) ||
755 	    CC((vmcs12->posted_intr_nv & 0xff00)) ||
756 	    CC(!kvm_vcpu_is_legal_aligned_gpa(vcpu, vmcs12->posted_intr_desc_addr, 64))))
757 		return -EINVAL;
758 
759 	/* tpr shadow is needed by all apicv features. */
760 	if (CC(!nested_cpu_has(vmcs12, CPU_BASED_TPR_SHADOW)))
761 		return -EINVAL;
762 
763 	return 0;
764 }
765 
766 static int nested_vmx_check_msr_switch(struct kvm_vcpu *vcpu,
767 				       u32 count, u64 addr)
768 {
769 	if (count == 0)
770 		return 0;
771 
772 	if (!kvm_vcpu_is_legal_aligned_gpa(vcpu, addr, 16) ||
773 	    !kvm_vcpu_is_legal_gpa(vcpu, (addr + count * sizeof(struct vmx_msr_entry) - 1)))
774 		return -EINVAL;
775 
776 	return 0;
777 }
778 
779 static int nested_vmx_check_exit_msr_switch_controls(struct kvm_vcpu *vcpu,
780 						     struct vmcs12 *vmcs12)
781 {
782 	if (CC(nested_vmx_check_msr_switch(vcpu,
783 					   vmcs12->vm_exit_msr_load_count,
784 					   vmcs12->vm_exit_msr_load_addr)) ||
785 	    CC(nested_vmx_check_msr_switch(vcpu,
786 					   vmcs12->vm_exit_msr_store_count,
787 					   vmcs12->vm_exit_msr_store_addr)))
788 		return -EINVAL;
789 
790 	return 0;
791 }
792 
793 static int nested_vmx_check_entry_msr_switch_controls(struct kvm_vcpu *vcpu,
794                                                       struct vmcs12 *vmcs12)
795 {
796 	if (CC(nested_vmx_check_msr_switch(vcpu,
797 					   vmcs12->vm_entry_msr_load_count,
798 					   vmcs12->vm_entry_msr_load_addr)))
799                 return -EINVAL;
800 
801 	return 0;
802 }
803 
804 static int nested_vmx_check_pml_controls(struct kvm_vcpu *vcpu,
805 					 struct vmcs12 *vmcs12)
806 {
807 	if (!nested_cpu_has_pml(vmcs12))
808 		return 0;
809 
810 	if (CC(!nested_cpu_has_ept(vmcs12)) ||
811 	    CC(!page_address_valid(vcpu, vmcs12->pml_address)))
812 		return -EINVAL;
813 
814 	return 0;
815 }
816 
817 static int nested_vmx_check_unrestricted_guest_controls(struct kvm_vcpu *vcpu,
818 							struct vmcs12 *vmcs12)
819 {
820 	if (CC(nested_cpu_has2(vmcs12, SECONDARY_EXEC_UNRESTRICTED_GUEST) &&
821 	       !nested_cpu_has_ept(vmcs12)))
822 		return -EINVAL;
823 	return 0;
824 }
825 
826 static int nested_vmx_check_mode_based_ept_exec_controls(struct kvm_vcpu *vcpu,
827 							 struct vmcs12 *vmcs12)
828 {
829 	if (CC(nested_cpu_has2(vmcs12, SECONDARY_EXEC_MODE_BASED_EPT_EXEC) &&
830 	       !nested_cpu_has_ept(vmcs12)))
831 		return -EINVAL;
832 	return 0;
833 }
834 
835 static int nested_vmx_check_shadow_vmcs_controls(struct kvm_vcpu *vcpu,
836 						 struct vmcs12 *vmcs12)
837 {
838 	if (!nested_cpu_has_shadow_vmcs(vmcs12))
839 		return 0;
840 
841 	if (CC(!page_address_valid(vcpu, vmcs12->vmread_bitmap)) ||
842 	    CC(!page_address_valid(vcpu, vmcs12->vmwrite_bitmap)))
843 		return -EINVAL;
844 
845 	return 0;
846 }
847 
848 static int nested_vmx_msr_check_common(struct kvm_vcpu *vcpu,
849 				       struct vmx_msr_entry *e)
850 {
851 	/* x2APIC MSR accesses are not allowed */
852 	if (CC(vcpu->arch.apic_base & X2APIC_ENABLE && e->index >> 8 == 0x8))
853 		return -EINVAL;
854 	if (CC(e->index == MSR_IA32_UCODE_WRITE) || /* SDM Table 35-2 */
855 	    CC(e->index == MSR_IA32_UCODE_REV))
856 		return -EINVAL;
857 	if (CC(e->reserved != 0))
858 		return -EINVAL;
859 	return 0;
860 }
861 
862 static int nested_vmx_load_msr_check(struct kvm_vcpu *vcpu,
863 				     struct vmx_msr_entry *e)
864 {
865 	if (CC(e->index == MSR_FS_BASE) ||
866 	    CC(e->index == MSR_GS_BASE) ||
867 	    CC(e->index == MSR_IA32_SMM_MONITOR_CTL) || /* SMM is not supported */
868 	    nested_vmx_msr_check_common(vcpu, e))
869 		return -EINVAL;
870 	return 0;
871 }
872 
873 static int nested_vmx_store_msr_check(struct kvm_vcpu *vcpu,
874 				      struct vmx_msr_entry *e)
875 {
876 	if (CC(e->index == MSR_IA32_SMBASE) || /* SMM is not supported */
877 	    nested_vmx_msr_check_common(vcpu, e))
878 		return -EINVAL;
879 	return 0;
880 }
881 
882 static u32 nested_vmx_max_atomic_switch_msrs(struct kvm_vcpu *vcpu)
883 {
884 	struct vcpu_vmx *vmx = to_vmx(vcpu);
885 	u64 vmx_misc = vmx_control_msr(vmx->nested.msrs.misc_low,
886 				       vmx->nested.msrs.misc_high);
887 
888 	return (vmx_misc_max_msr(vmx_misc) + 1) * VMX_MISC_MSR_LIST_MULTIPLIER;
889 }
890 
891 /*
892  * Load guest's/host's msr at nested entry/exit.
893  * return 0 for success, entry index for failure.
894  *
895  * One of the failure modes for MSR load/store is when a list exceeds the
896  * virtual hardware's capacity. To maintain compatibility with hardware inasmuch
897  * as possible, process all valid entries before failing rather than precheck
898  * for a capacity violation.
899  */
900 static u32 nested_vmx_load_msr(struct kvm_vcpu *vcpu, u64 gpa, u32 count)
901 {
902 	u32 i;
903 	struct vmx_msr_entry e;
904 	u32 max_msr_list_size = nested_vmx_max_atomic_switch_msrs(vcpu);
905 
906 	for (i = 0; i < count; i++) {
907 		if (unlikely(i >= max_msr_list_size))
908 			goto fail;
909 
910 		if (kvm_vcpu_read_guest(vcpu, gpa + i * sizeof(e),
911 					&e, sizeof(e))) {
912 			pr_debug_ratelimited(
913 				"%s cannot read MSR entry (%u, 0x%08llx)\n",
914 				__func__, i, gpa + i * sizeof(e));
915 			goto fail;
916 		}
917 		if (nested_vmx_load_msr_check(vcpu, &e)) {
918 			pr_debug_ratelimited(
919 				"%s check failed (%u, 0x%x, 0x%x)\n",
920 				__func__, i, e.index, e.reserved);
921 			goto fail;
922 		}
923 		if (kvm_set_msr(vcpu, e.index, e.value)) {
924 			pr_debug_ratelimited(
925 				"%s cannot write MSR (%u, 0x%x, 0x%llx)\n",
926 				__func__, i, e.index, e.value);
927 			goto fail;
928 		}
929 	}
930 	return 0;
931 fail:
932 	/* Note, max_msr_list_size is at most 4096, i.e. this can't wrap. */
933 	return i + 1;
934 }
935 
936 static bool nested_vmx_get_vmexit_msr_value(struct kvm_vcpu *vcpu,
937 					    u32 msr_index,
938 					    u64 *data)
939 {
940 	struct vcpu_vmx *vmx = to_vmx(vcpu);
941 
942 	/*
943 	 * If the L0 hypervisor stored a more accurate value for the TSC that
944 	 * does not include the time taken for emulation of the L2->L1
945 	 * VM-exit in L0, use the more accurate value.
946 	 */
947 	if (msr_index == MSR_IA32_TSC) {
948 		int i = vmx_find_loadstore_msr_slot(&vmx->msr_autostore.guest,
949 						    MSR_IA32_TSC);
950 
951 		if (i >= 0) {
952 			u64 val = vmx->msr_autostore.guest.val[i].value;
953 
954 			*data = kvm_read_l1_tsc(vcpu, val);
955 			return true;
956 		}
957 	}
958 
959 	if (kvm_get_msr(vcpu, msr_index, data)) {
960 		pr_debug_ratelimited("%s cannot read MSR (0x%x)\n", __func__,
961 			msr_index);
962 		return false;
963 	}
964 	return true;
965 }
966 
967 static bool read_and_check_msr_entry(struct kvm_vcpu *vcpu, u64 gpa, int i,
968 				     struct vmx_msr_entry *e)
969 {
970 	if (kvm_vcpu_read_guest(vcpu,
971 				gpa + i * sizeof(*e),
972 				e, 2 * sizeof(u32))) {
973 		pr_debug_ratelimited(
974 			"%s cannot read MSR entry (%u, 0x%08llx)\n",
975 			__func__, i, gpa + i * sizeof(*e));
976 		return false;
977 	}
978 	if (nested_vmx_store_msr_check(vcpu, e)) {
979 		pr_debug_ratelimited(
980 			"%s check failed (%u, 0x%x, 0x%x)\n",
981 			__func__, i, e->index, e->reserved);
982 		return false;
983 	}
984 	return true;
985 }
986 
987 static int nested_vmx_store_msr(struct kvm_vcpu *vcpu, u64 gpa, u32 count)
988 {
989 	u64 data;
990 	u32 i;
991 	struct vmx_msr_entry e;
992 	u32 max_msr_list_size = nested_vmx_max_atomic_switch_msrs(vcpu);
993 
994 	for (i = 0; i < count; i++) {
995 		if (unlikely(i >= max_msr_list_size))
996 			return -EINVAL;
997 
998 		if (!read_and_check_msr_entry(vcpu, gpa, i, &e))
999 			return -EINVAL;
1000 
1001 		if (!nested_vmx_get_vmexit_msr_value(vcpu, e.index, &data))
1002 			return -EINVAL;
1003 
1004 		if (kvm_vcpu_write_guest(vcpu,
1005 					 gpa + i * sizeof(e) +
1006 					     offsetof(struct vmx_msr_entry, value),
1007 					 &data, sizeof(data))) {
1008 			pr_debug_ratelimited(
1009 				"%s cannot write MSR (%u, 0x%x, 0x%llx)\n",
1010 				__func__, i, e.index, data);
1011 			return -EINVAL;
1012 		}
1013 	}
1014 	return 0;
1015 }
1016 
1017 static bool nested_msr_store_list_has_msr(struct kvm_vcpu *vcpu, u32 msr_index)
1018 {
1019 	struct vmcs12 *vmcs12 = get_vmcs12(vcpu);
1020 	u32 count = vmcs12->vm_exit_msr_store_count;
1021 	u64 gpa = vmcs12->vm_exit_msr_store_addr;
1022 	struct vmx_msr_entry e;
1023 	u32 i;
1024 
1025 	for (i = 0; i < count; i++) {
1026 		if (!read_and_check_msr_entry(vcpu, gpa, i, &e))
1027 			return false;
1028 
1029 		if (e.index == msr_index)
1030 			return true;
1031 	}
1032 	return false;
1033 }
1034 
1035 static void prepare_vmx_msr_autostore_list(struct kvm_vcpu *vcpu,
1036 					   u32 msr_index)
1037 {
1038 	struct vcpu_vmx *vmx = to_vmx(vcpu);
1039 	struct vmx_msrs *autostore = &vmx->msr_autostore.guest;
1040 	bool in_vmcs12_store_list;
1041 	int msr_autostore_slot;
1042 	bool in_autostore_list;
1043 	int last;
1044 
1045 	msr_autostore_slot = vmx_find_loadstore_msr_slot(autostore, msr_index);
1046 	in_autostore_list = msr_autostore_slot >= 0;
1047 	in_vmcs12_store_list = nested_msr_store_list_has_msr(vcpu, msr_index);
1048 
1049 	if (in_vmcs12_store_list && !in_autostore_list) {
1050 		if (autostore->nr == MAX_NR_LOADSTORE_MSRS) {
1051 			/*
1052 			 * Emulated VMEntry does not fail here.  Instead a less
1053 			 * accurate value will be returned by
1054 			 * nested_vmx_get_vmexit_msr_value() using kvm_get_msr()
1055 			 * instead of reading the value from the vmcs02 VMExit
1056 			 * MSR-store area.
1057 			 */
1058 			pr_warn_ratelimited(
1059 				"Not enough msr entries in msr_autostore.  Can't add msr %x\n",
1060 				msr_index);
1061 			return;
1062 		}
1063 		last = autostore->nr++;
1064 		autostore->val[last].index = msr_index;
1065 	} else if (!in_vmcs12_store_list && in_autostore_list) {
1066 		last = --autostore->nr;
1067 		autostore->val[msr_autostore_slot] = autostore->val[last];
1068 	}
1069 }
1070 
1071 /*
1072  * Load guest's/host's cr3 at nested entry/exit.  @nested_ept is true if we are
1073  * emulating VM-Entry into a guest with EPT enabled.  On failure, the expected
1074  * Exit Qualification (for a VM-Entry consistency check VM-Exit) is assigned to
1075  * @entry_failure_code.
1076  */
1077 static int nested_vmx_load_cr3(struct kvm_vcpu *vcpu, unsigned long cr3,
1078 			       bool nested_ept, bool reload_pdptrs,
1079 			       enum vm_entry_failure_code *entry_failure_code)
1080 {
1081 	if (CC(kvm_vcpu_is_illegal_gpa(vcpu, cr3))) {
1082 		*entry_failure_code = ENTRY_FAIL_DEFAULT;
1083 		return -EINVAL;
1084 	}
1085 
1086 	/*
1087 	 * If PAE paging and EPT are both on, CR3 is not used by the CPU and
1088 	 * must not be dereferenced.
1089 	 */
1090 	if (reload_pdptrs && !nested_ept && is_pae_paging(vcpu) &&
1091 	    CC(!load_pdptrs(vcpu, cr3))) {
1092 		*entry_failure_code = ENTRY_FAIL_PDPTE;
1093 		return -EINVAL;
1094 	}
1095 
1096 	vcpu->arch.cr3 = cr3;
1097 	kvm_register_mark_dirty(vcpu, VCPU_EXREG_CR3);
1098 
1099 	/* Re-initialize the MMU, e.g. to pick up CR4 MMU role changes. */
1100 	kvm_init_mmu(vcpu);
1101 
1102 	if (!nested_ept)
1103 		kvm_mmu_new_pgd(vcpu, cr3);
1104 
1105 	return 0;
1106 }
1107 
1108 /*
1109  * Returns if KVM is able to config CPU to tag TLB entries
1110  * populated by L2 differently than TLB entries populated
1111  * by L1.
1112  *
1113  * If L0 uses EPT, L1 and L2 run with different EPTP because
1114  * guest_mode is part of kvm_mmu_page_role. Thus, TLB entries
1115  * are tagged with different EPTP.
1116  *
1117  * If L1 uses VPID and we allocated a vpid02, TLB entries are tagged
1118  * with different VPID (L1 entries are tagged with vmx->vpid
1119  * while L2 entries are tagged with vmx->nested.vpid02).
1120  */
1121 static bool nested_has_guest_tlb_tag(struct kvm_vcpu *vcpu)
1122 {
1123 	struct vmcs12 *vmcs12 = get_vmcs12(vcpu);
1124 
1125 	return enable_ept ||
1126 	       (nested_cpu_has_vpid(vmcs12) && to_vmx(vcpu)->nested.vpid02);
1127 }
1128 
1129 static void nested_vmx_transition_tlb_flush(struct kvm_vcpu *vcpu,
1130 					    struct vmcs12 *vmcs12,
1131 					    bool is_vmenter)
1132 {
1133 	struct vcpu_vmx *vmx = to_vmx(vcpu);
1134 
1135 	/*
1136 	 * KVM_REQ_HV_TLB_FLUSH flushes entries from either L1's VP_ID or
1137 	 * L2's VP_ID upon request from the guest. Make sure we check for
1138 	 * pending entries in the right FIFO upon L1/L2 transition as these
1139 	 * requests are put by other vCPUs asynchronously.
1140 	 */
1141 	if (to_hv_vcpu(vcpu) && enable_ept)
1142 		kvm_make_request(KVM_REQ_HV_TLB_FLUSH, vcpu);
1143 
1144 	/*
1145 	 * If vmcs12 doesn't use VPID, L1 expects linear and combined mappings
1146 	 * for *all* contexts to be flushed on VM-Enter/VM-Exit, i.e. it's a
1147 	 * full TLB flush from the guest's perspective.  This is required even
1148 	 * if VPID is disabled in the host as KVM may need to synchronize the
1149 	 * MMU in response to the guest TLB flush.
1150 	 *
1151 	 * Note, using TLB_FLUSH_GUEST is correct even if nested EPT is in use.
1152 	 * EPT is a special snowflake, as guest-physical mappings aren't
1153 	 * flushed on VPID invalidations, including VM-Enter or VM-Exit with
1154 	 * VPID disabled.  As a result, KVM _never_ needs to sync nEPT
1155 	 * entries on VM-Enter because L1 can't rely on VM-Enter to flush
1156 	 * those mappings.
1157 	 */
1158 	if (!nested_cpu_has_vpid(vmcs12)) {
1159 		kvm_make_request(KVM_REQ_TLB_FLUSH_GUEST, vcpu);
1160 		return;
1161 	}
1162 
1163 	/* L2 should never have a VPID if VPID is disabled. */
1164 	WARN_ON(!enable_vpid);
1165 
1166 	/*
1167 	 * VPID is enabled and in use by vmcs12.  If vpid12 is changing, then
1168 	 * emulate a guest TLB flush as KVM does not track vpid12 history nor
1169 	 * is the VPID incorporated into the MMU context.  I.e. KVM must assume
1170 	 * that the new vpid12 has never been used and thus represents a new
1171 	 * guest ASID that cannot have entries in the TLB.
1172 	 */
1173 	if (is_vmenter && vmcs12->virtual_processor_id != vmx->nested.last_vpid) {
1174 		vmx->nested.last_vpid = vmcs12->virtual_processor_id;
1175 		kvm_make_request(KVM_REQ_TLB_FLUSH_GUEST, vcpu);
1176 		return;
1177 	}
1178 
1179 	/*
1180 	 * If VPID is enabled, used by vmc12, and vpid12 is not changing but
1181 	 * does not have a unique TLB tag (ASID), i.e. EPT is disabled and
1182 	 * KVM was unable to allocate a VPID for L2, flush the current context
1183 	 * as the effective ASID is common to both L1 and L2.
1184 	 */
1185 	if (!nested_has_guest_tlb_tag(vcpu))
1186 		kvm_make_request(KVM_REQ_TLB_FLUSH_CURRENT, vcpu);
1187 }
1188 
1189 static bool is_bitwise_subset(u64 superset, u64 subset, u64 mask)
1190 {
1191 	superset &= mask;
1192 	subset &= mask;
1193 
1194 	return (superset | subset) == superset;
1195 }
1196 
1197 static int vmx_restore_vmx_basic(struct vcpu_vmx *vmx, u64 data)
1198 {
1199 	const u64 feature_and_reserved =
1200 		/* feature (except bit 48; see below) */
1201 		BIT_ULL(49) | BIT_ULL(54) | BIT_ULL(55) |
1202 		/* reserved */
1203 		BIT_ULL(31) | GENMASK_ULL(47, 45) | GENMASK_ULL(63, 56);
1204 	u64 vmx_basic = vmcs_config.nested.basic;
1205 
1206 	if (!is_bitwise_subset(vmx_basic, data, feature_and_reserved))
1207 		return -EINVAL;
1208 
1209 	/*
1210 	 * KVM does not emulate a version of VMX that constrains physical
1211 	 * addresses of VMX structures (e.g. VMCS) to 32-bits.
1212 	 */
1213 	if (data & BIT_ULL(48))
1214 		return -EINVAL;
1215 
1216 	if (vmx_basic_vmcs_revision_id(vmx_basic) !=
1217 	    vmx_basic_vmcs_revision_id(data))
1218 		return -EINVAL;
1219 
1220 	if (vmx_basic_vmcs_size(vmx_basic) > vmx_basic_vmcs_size(data))
1221 		return -EINVAL;
1222 
1223 	vmx->nested.msrs.basic = data;
1224 	return 0;
1225 }
1226 
1227 static void vmx_get_control_msr(struct nested_vmx_msrs *msrs, u32 msr_index,
1228 				u32 **low, u32 **high)
1229 {
1230 	switch (msr_index) {
1231 	case MSR_IA32_VMX_TRUE_PINBASED_CTLS:
1232 		*low = &msrs->pinbased_ctls_low;
1233 		*high = &msrs->pinbased_ctls_high;
1234 		break;
1235 	case MSR_IA32_VMX_TRUE_PROCBASED_CTLS:
1236 		*low = &msrs->procbased_ctls_low;
1237 		*high = &msrs->procbased_ctls_high;
1238 		break;
1239 	case MSR_IA32_VMX_TRUE_EXIT_CTLS:
1240 		*low = &msrs->exit_ctls_low;
1241 		*high = &msrs->exit_ctls_high;
1242 		break;
1243 	case MSR_IA32_VMX_TRUE_ENTRY_CTLS:
1244 		*low = &msrs->entry_ctls_low;
1245 		*high = &msrs->entry_ctls_high;
1246 		break;
1247 	case MSR_IA32_VMX_PROCBASED_CTLS2:
1248 		*low = &msrs->secondary_ctls_low;
1249 		*high = &msrs->secondary_ctls_high;
1250 		break;
1251 	default:
1252 		BUG();
1253 	}
1254 }
1255 
1256 static int
1257 vmx_restore_control_msr(struct vcpu_vmx *vmx, u32 msr_index, u64 data)
1258 {
1259 	u32 *lowp, *highp;
1260 	u64 supported;
1261 
1262 	vmx_get_control_msr(&vmcs_config.nested, msr_index, &lowp, &highp);
1263 
1264 	supported = vmx_control_msr(*lowp, *highp);
1265 
1266 	/* Check must-be-1 bits are still 1. */
1267 	if (!is_bitwise_subset(data, supported, GENMASK_ULL(31, 0)))
1268 		return -EINVAL;
1269 
1270 	/* Check must-be-0 bits are still 0. */
1271 	if (!is_bitwise_subset(supported, data, GENMASK_ULL(63, 32)))
1272 		return -EINVAL;
1273 
1274 	vmx_get_control_msr(&vmx->nested.msrs, msr_index, &lowp, &highp);
1275 	*lowp = data;
1276 	*highp = data >> 32;
1277 	return 0;
1278 }
1279 
1280 static int vmx_restore_vmx_misc(struct vcpu_vmx *vmx, u64 data)
1281 {
1282 	const u64 feature_and_reserved_bits =
1283 		/* feature */
1284 		BIT_ULL(5) | GENMASK_ULL(8, 6) | BIT_ULL(14) | BIT_ULL(15) |
1285 		BIT_ULL(28) | BIT_ULL(29) | BIT_ULL(30) |
1286 		/* reserved */
1287 		GENMASK_ULL(13, 9) | BIT_ULL(31);
1288 	u64 vmx_misc = vmx_control_msr(vmcs_config.nested.misc_low,
1289 				       vmcs_config.nested.misc_high);
1290 
1291 	if (!is_bitwise_subset(vmx_misc, data, feature_and_reserved_bits))
1292 		return -EINVAL;
1293 
1294 	if ((vmx->nested.msrs.pinbased_ctls_high &
1295 	     PIN_BASED_VMX_PREEMPTION_TIMER) &&
1296 	    vmx_misc_preemption_timer_rate(data) !=
1297 	    vmx_misc_preemption_timer_rate(vmx_misc))
1298 		return -EINVAL;
1299 
1300 	if (vmx_misc_cr3_count(data) > vmx_misc_cr3_count(vmx_misc))
1301 		return -EINVAL;
1302 
1303 	if (vmx_misc_max_msr(data) > vmx_misc_max_msr(vmx_misc))
1304 		return -EINVAL;
1305 
1306 	if (vmx_misc_mseg_revid(data) != vmx_misc_mseg_revid(vmx_misc))
1307 		return -EINVAL;
1308 
1309 	vmx->nested.msrs.misc_low = data;
1310 	vmx->nested.msrs.misc_high = data >> 32;
1311 
1312 	return 0;
1313 }
1314 
1315 static int vmx_restore_vmx_ept_vpid_cap(struct vcpu_vmx *vmx, u64 data)
1316 {
1317 	u64 vmx_ept_vpid_cap = vmx_control_msr(vmcs_config.nested.ept_caps,
1318 					       vmcs_config.nested.vpid_caps);
1319 
1320 	/* Every bit is either reserved or a feature bit. */
1321 	if (!is_bitwise_subset(vmx_ept_vpid_cap, data, -1ULL))
1322 		return -EINVAL;
1323 
1324 	vmx->nested.msrs.ept_caps = data;
1325 	vmx->nested.msrs.vpid_caps = data >> 32;
1326 	return 0;
1327 }
1328 
1329 static u64 *vmx_get_fixed0_msr(struct nested_vmx_msrs *msrs, u32 msr_index)
1330 {
1331 	switch (msr_index) {
1332 	case MSR_IA32_VMX_CR0_FIXED0:
1333 		return &msrs->cr0_fixed0;
1334 	case MSR_IA32_VMX_CR4_FIXED0:
1335 		return &msrs->cr4_fixed0;
1336 	default:
1337 		BUG();
1338 	}
1339 }
1340 
1341 static int vmx_restore_fixed0_msr(struct vcpu_vmx *vmx, u32 msr_index, u64 data)
1342 {
1343 	const u64 *msr = vmx_get_fixed0_msr(&vmcs_config.nested, msr_index);
1344 
1345 	/*
1346 	 * 1 bits (which indicates bits which "must-be-1" during VMX operation)
1347 	 * must be 1 in the restored value.
1348 	 */
1349 	if (!is_bitwise_subset(data, *msr, -1ULL))
1350 		return -EINVAL;
1351 
1352 	*vmx_get_fixed0_msr(&vmx->nested.msrs, msr_index) = data;
1353 	return 0;
1354 }
1355 
1356 /*
1357  * Called when userspace is restoring VMX MSRs.
1358  *
1359  * Returns 0 on success, non-0 otherwise.
1360  */
1361 int vmx_set_vmx_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 data)
1362 {
1363 	struct vcpu_vmx *vmx = to_vmx(vcpu);
1364 
1365 	/*
1366 	 * Don't allow changes to the VMX capability MSRs while the vCPU
1367 	 * is in VMX operation.
1368 	 */
1369 	if (vmx->nested.vmxon)
1370 		return -EBUSY;
1371 
1372 	switch (msr_index) {
1373 	case MSR_IA32_VMX_BASIC:
1374 		return vmx_restore_vmx_basic(vmx, data);
1375 	case MSR_IA32_VMX_PINBASED_CTLS:
1376 	case MSR_IA32_VMX_PROCBASED_CTLS:
1377 	case MSR_IA32_VMX_EXIT_CTLS:
1378 	case MSR_IA32_VMX_ENTRY_CTLS:
1379 		/*
1380 		 * The "non-true" VMX capability MSRs are generated from the
1381 		 * "true" MSRs, so we do not support restoring them directly.
1382 		 *
1383 		 * If userspace wants to emulate VMX_BASIC[55]=0, userspace
1384 		 * should restore the "true" MSRs with the must-be-1 bits
1385 		 * set according to the SDM Vol 3. A.2 "RESERVED CONTROLS AND
1386 		 * DEFAULT SETTINGS".
1387 		 */
1388 		return -EINVAL;
1389 	case MSR_IA32_VMX_TRUE_PINBASED_CTLS:
1390 	case MSR_IA32_VMX_TRUE_PROCBASED_CTLS:
1391 	case MSR_IA32_VMX_TRUE_EXIT_CTLS:
1392 	case MSR_IA32_VMX_TRUE_ENTRY_CTLS:
1393 	case MSR_IA32_VMX_PROCBASED_CTLS2:
1394 		return vmx_restore_control_msr(vmx, msr_index, data);
1395 	case MSR_IA32_VMX_MISC:
1396 		return vmx_restore_vmx_misc(vmx, data);
1397 	case MSR_IA32_VMX_CR0_FIXED0:
1398 	case MSR_IA32_VMX_CR4_FIXED0:
1399 		return vmx_restore_fixed0_msr(vmx, msr_index, data);
1400 	case MSR_IA32_VMX_CR0_FIXED1:
1401 	case MSR_IA32_VMX_CR4_FIXED1:
1402 		/*
1403 		 * These MSRs are generated based on the vCPU's CPUID, so we
1404 		 * do not support restoring them directly.
1405 		 */
1406 		return -EINVAL;
1407 	case MSR_IA32_VMX_EPT_VPID_CAP:
1408 		return vmx_restore_vmx_ept_vpid_cap(vmx, data);
1409 	case MSR_IA32_VMX_VMCS_ENUM:
1410 		vmx->nested.msrs.vmcs_enum = data;
1411 		return 0;
1412 	case MSR_IA32_VMX_VMFUNC:
1413 		if (data & ~vmcs_config.nested.vmfunc_controls)
1414 			return -EINVAL;
1415 		vmx->nested.msrs.vmfunc_controls = data;
1416 		return 0;
1417 	default:
1418 		/*
1419 		 * The rest of the VMX capability MSRs do not support restore.
1420 		 */
1421 		return -EINVAL;
1422 	}
1423 }
1424 
1425 /* Returns 0 on success, non-0 otherwise. */
1426 int vmx_get_vmx_msr(struct nested_vmx_msrs *msrs, u32 msr_index, u64 *pdata)
1427 {
1428 	switch (msr_index) {
1429 	case MSR_IA32_VMX_BASIC:
1430 		*pdata = msrs->basic;
1431 		break;
1432 	case MSR_IA32_VMX_TRUE_PINBASED_CTLS:
1433 	case MSR_IA32_VMX_PINBASED_CTLS:
1434 		*pdata = vmx_control_msr(
1435 			msrs->pinbased_ctls_low,
1436 			msrs->pinbased_ctls_high);
1437 		if (msr_index == MSR_IA32_VMX_PINBASED_CTLS)
1438 			*pdata |= PIN_BASED_ALWAYSON_WITHOUT_TRUE_MSR;
1439 		break;
1440 	case MSR_IA32_VMX_TRUE_PROCBASED_CTLS:
1441 	case MSR_IA32_VMX_PROCBASED_CTLS:
1442 		*pdata = vmx_control_msr(
1443 			msrs->procbased_ctls_low,
1444 			msrs->procbased_ctls_high);
1445 		if (msr_index == MSR_IA32_VMX_PROCBASED_CTLS)
1446 			*pdata |= CPU_BASED_ALWAYSON_WITHOUT_TRUE_MSR;
1447 		break;
1448 	case MSR_IA32_VMX_TRUE_EXIT_CTLS:
1449 	case MSR_IA32_VMX_EXIT_CTLS:
1450 		*pdata = vmx_control_msr(
1451 			msrs->exit_ctls_low,
1452 			msrs->exit_ctls_high);
1453 		if (msr_index == MSR_IA32_VMX_EXIT_CTLS)
1454 			*pdata |= VM_EXIT_ALWAYSON_WITHOUT_TRUE_MSR;
1455 		break;
1456 	case MSR_IA32_VMX_TRUE_ENTRY_CTLS:
1457 	case MSR_IA32_VMX_ENTRY_CTLS:
1458 		*pdata = vmx_control_msr(
1459 			msrs->entry_ctls_low,
1460 			msrs->entry_ctls_high);
1461 		if (msr_index == MSR_IA32_VMX_ENTRY_CTLS)
1462 			*pdata |= VM_ENTRY_ALWAYSON_WITHOUT_TRUE_MSR;
1463 		break;
1464 	case MSR_IA32_VMX_MISC:
1465 		*pdata = vmx_control_msr(
1466 			msrs->misc_low,
1467 			msrs->misc_high);
1468 		break;
1469 	case MSR_IA32_VMX_CR0_FIXED0:
1470 		*pdata = msrs->cr0_fixed0;
1471 		break;
1472 	case MSR_IA32_VMX_CR0_FIXED1:
1473 		*pdata = msrs->cr0_fixed1;
1474 		break;
1475 	case MSR_IA32_VMX_CR4_FIXED0:
1476 		*pdata = msrs->cr4_fixed0;
1477 		break;
1478 	case MSR_IA32_VMX_CR4_FIXED1:
1479 		*pdata = msrs->cr4_fixed1;
1480 		break;
1481 	case MSR_IA32_VMX_VMCS_ENUM:
1482 		*pdata = msrs->vmcs_enum;
1483 		break;
1484 	case MSR_IA32_VMX_PROCBASED_CTLS2:
1485 		*pdata = vmx_control_msr(
1486 			msrs->secondary_ctls_low,
1487 			msrs->secondary_ctls_high);
1488 		break;
1489 	case MSR_IA32_VMX_EPT_VPID_CAP:
1490 		*pdata = msrs->ept_caps |
1491 			((u64)msrs->vpid_caps << 32);
1492 		break;
1493 	case MSR_IA32_VMX_VMFUNC:
1494 		*pdata = msrs->vmfunc_controls;
1495 		break;
1496 	default:
1497 		return 1;
1498 	}
1499 
1500 	return 0;
1501 }
1502 
1503 /*
1504  * Copy the writable VMCS shadow fields back to the VMCS12, in case they have
1505  * been modified by the L1 guest.  Note, "writable" in this context means
1506  * "writable by the guest", i.e. tagged SHADOW_FIELD_RW; the set of
1507  * fields tagged SHADOW_FIELD_RO may or may not align with the "read-only"
1508  * VM-exit information fields (which are actually writable if the vCPU is
1509  * configured to support "VMWRITE to any supported field in the VMCS").
1510  */
1511 static void copy_shadow_to_vmcs12(struct vcpu_vmx *vmx)
1512 {
1513 	struct vmcs *shadow_vmcs = vmx->vmcs01.shadow_vmcs;
1514 	struct vmcs12 *vmcs12 = get_vmcs12(&vmx->vcpu);
1515 	struct shadow_vmcs_field field;
1516 	unsigned long val;
1517 	int i;
1518 
1519 	if (WARN_ON(!shadow_vmcs))
1520 		return;
1521 
1522 	preempt_disable();
1523 
1524 	vmcs_load(shadow_vmcs);
1525 
1526 	for (i = 0; i < max_shadow_read_write_fields; i++) {
1527 		field = shadow_read_write_fields[i];
1528 		val = __vmcs_readl(field.encoding);
1529 		vmcs12_write_any(vmcs12, field.encoding, field.offset, val);
1530 	}
1531 
1532 	vmcs_clear(shadow_vmcs);
1533 	vmcs_load(vmx->loaded_vmcs->vmcs);
1534 
1535 	preempt_enable();
1536 }
1537 
1538 static void copy_vmcs12_to_shadow(struct vcpu_vmx *vmx)
1539 {
1540 	const struct shadow_vmcs_field *fields[] = {
1541 		shadow_read_write_fields,
1542 		shadow_read_only_fields
1543 	};
1544 	const int max_fields[] = {
1545 		max_shadow_read_write_fields,
1546 		max_shadow_read_only_fields
1547 	};
1548 	struct vmcs *shadow_vmcs = vmx->vmcs01.shadow_vmcs;
1549 	struct vmcs12 *vmcs12 = get_vmcs12(&vmx->vcpu);
1550 	struct shadow_vmcs_field field;
1551 	unsigned long val;
1552 	int i, q;
1553 
1554 	if (WARN_ON(!shadow_vmcs))
1555 		return;
1556 
1557 	vmcs_load(shadow_vmcs);
1558 
1559 	for (q = 0; q < ARRAY_SIZE(fields); q++) {
1560 		for (i = 0; i < max_fields[q]; i++) {
1561 			field = fields[q][i];
1562 			val = vmcs12_read_any(vmcs12, field.encoding,
1563 					      field.offset);
1564 			__vmcs_writel(field.encoding, val);
1565 		}
1566 	}
1567 
1568 	vmcs_clear(shadow_vmcs);
1569 	vmcs_load(vmx->loaded_vmcs->vmcs);
1570 }
1571 
1572 static void copy_enlightened_to_vmcs12(struct vcpu_vmx *vmx, u32 hv_clean_fields)
1573 {
1574 	struct vmcs12 *vmcs12 = vmx->nested.cached_vmcs12;
1575 	struct hv_enlightened_vmcs *evmcs = vmx->nested.hv_evmcs;
1576 	struct kvm_vcpu_hv *hv_vcpu = to_hv_vcpu(&vmx->vcpu);
1577 
1578 	/* HV_VMX_ENLIGHTENED_CLEAN_FIELD_NONE */
1579 	vmcs12->tpr_threshold = evmcs->tpr_threshold;
1580 	vmcs12->guest_rip = evmcs->guest_rip;
1581 
1582 	if (unlikely(!(hv_clean_fields &
1583 		       HV_VMX_ENLIGHTENED_CLEAN_FIELD_ENLIGHTENMENTSCONTROL))) {
1584 		hv_vcpu->nested.pa_page_gpa = evmcs->partition_assist_page;
1585 		hv_vcpu->nested.vm_id = evmcs->hv_vm_id;
1586 		hv_vcpu->nested.vp_id = evmcs->hv_vp_id;
1587 	}
1588 
1589 	if (unlikely(!(hv_clean_fields &
1590 		       HV_VMX_ENLIGHTENED_CLEAN_FIELD_GUEST_BASIC))) {
1591 		vmcs12->guest_rsp = evmcs->guest_rsp;
1592 		vmcs12->guest_rflags = evmcs->guest_rflags;
1593 		vmcs12->guest_interruptibility_info =
1594 			evmcs->guest_interruptibility_info;
1595 		/*
1596 		 * Not present in struct vmcs12:
1597 		 * vmcs12->guest_ssp = evmcs->guest_ssp;
1598 		 */
1599 	}
1600 
1601 	if (unlikely(!(hv_clean_fields &
1602 		       HV_VMX_ENLIGHTENED_CLEAN_FIELD_CONTROL_PROC))) {
1603 		vmcs12->cpu_based_vm_exec_control =
1604 			evmcs->cpu_based_vm_exec_control;
1605 	}
1606 
1607 	if (unlikely(!(hv_clean_fields &
1608 		       HV_VMX_ENLIGHTENED_CLEAN_FIELD_CONTROL_EXCPN))) {
1609 		vmcs12->exception_bitmap = evmcs->exception_bitmap;
1610 	}
1611 
1612 	if (unlikely(!(hv_clean_fields &
1613 		       HV_VMX_ENLIGHTENED_CLEAN_FIELD_CONTROL_ENTRY))) {
1614 		vmcs12->vm_entry_controls = evmcs->vm_entry_controls;
1615 	}
1616 
1617 	if (unlikely(!(hv_clean_fields &
1618 		       HV_VMX_ENLIGHTENED_CLEAN_FIELD_CONTROL_EVENT))) {
1619 		vmcs12->vm_entry_intr_info_field =
1620 			evmcs->vm_entry_intr_info_field;
1621 		vmcs12->vm_entry_exception_error_code =
1622 			evmcs->vm_entry_exception_error_code;
1623 		vmcs12->vm_entry_instruction_len =
1624 			evmcs->vm_entry_instruction_len;
1625 	}
1626 
1627 	if (unlikely(!(hv_clean_fields &
1628 		       HV_VMX_ENLIGHTENED_CLEAN_FIELD_HOST_GRP1))) {
1629 		vmcs12->host_ia32_pat = evmcs->host_ia32_pat;
1630 		vmcs12->host_ia32_efer = evmcs->host_ia32_efer;
1631 		vmcs12->host_cr0 = evmcs->host_cr0;
1632 		vmcs12->host_cr3 = evmcs->host_cr3;
1633 		vmcs12->host_cr4 = evmcs->host_cr4;
1634 		vmcs12->host_ia32_sysenter_esp = evmcs->host_ia32_sysenter_esp;
1635 		vmcs12->host_ia32_sysenter_eip = evmcs->host_ia32_sysenter_eip;
1636 		vmcs12->host_rip = evmcs->host_rip;
1637 		vmcs12->host_ia32_sysenter_cs = evmcs->host_ia32_sysenter_cs;
1638 		vmcs12->host_es_selector = evmcs->host_es_selector;
1639 		vmcs12->host_cs_selector = evmcs->host_cs_selector;
1640 		vmcs12->host_ss_selector = evmcs->host_ss_selector;
1641 		vmcs12->host_ds_selector = evmcs->host_ds_selector;
1642 		vmcs12->host_fs_selector = evmcs->host_fs_selector;
1643 		vmcs12->host_gs_selector = evmcs->host_gs_selector;
1644 		vmcs12->host_tr_selector = evmcs->host_tr_selector;
1645 		vmcs12->host_ia32_perf_global_ctrl = evmcs->host_ia32_perf_global_ctrl;
1646 		/*
1647 		 * Not present in struct vmcs12:
1648 		 * vmcs12->host_ia32_s_cet = evmcs->host_ia32_s_cet;
1649 		 * vmcs12->host_ssp = evmcs->host_ssp;
1650 		 * vmcs12->host_ia32_int_ssp_table_addr = evmcs->host_ia32_int_ssp_table_addr;
1651 		 */
1652 	}
1653 
1654 	if (unlikely(!(hv_clean_fields &
1655 		       HV_VMX_ENLIGHTENED_CLEAN_FIELD_CONTROL_GRP1))) {
1656 		vmcs12->pin_based_vm_exec_control =
1657 			evmcs->pin_based_vm_exec_control;
1658 		vmcs12->vm_exit_controls = evmcs->vm_exit_controls;
1659 		vmcs12->secondary_vm_exec_control =
1660 			evmcs->secondary_vm_exec_control;
1661 	}
1662 
1663 	if (unlikely(!(hv_clean_fields &
1664 		       HV_VMX_ENLIGHTENED_CLEAN_FIELD_IO_BITMAP))) {
1665 		vmcs12->io_bitmap_a = evmcs->io_bitmap_a;
1666 		vmcs12->io_bitmap_b = evmcs->io_bitmap_b;
1667 	}
1668 
1669 	if (unlikely(!(hv_clean_fields &
1670 		       HV_VMX_ENLIGHTENED_CLEAN_FIELD_MSR_BITMAP))) {
1671 		vmcs12->msr_bitmap = evmcs->msr_bitmap;
1672 	}
1673 
1674 	if (unlikely(!(hv_clean_fields &
1675 		       HV_VMX_ENLIGHTENED_CLEAN_FIELD_GUEST_GRP2))) {
1676 		vmcs12->guest_es_base = evmcs->guest_es_base;
1677 		vmcs12->guest_cs_base = evmcs->guest_cs_base;
1678 		vmcs12->guest_ss_base = evmcs->guest_ss_base;
1679 		vmcs12->guest_ds_base = evmcs->guest_ds_base;
1680 		vmcs12->guest_fs_base = evmcs->guest_fs_base;
1681 		vmcs12->guest_gs_base = evmcs->guest_gs_base;
1682 		vmcs12->guest_ldtr_base = evmcs->guest_ldtr_base;
1683 		vmcs12->guest_tr_base = evmcs->guest_tr_base;
1684 		vmcs12->guest_gdtr_base = evmcs->guest_gdtr_base;
1685 		vmcs12->guest_idtr_base = evmcs->guest_idtr_base;
1686 		vmcs12->guest_es_limit = evmcs->guest_es_limit;
1687 		vmcs12->guest_cs_limit = evmcs->guest_cs_limit;
1688 		vmcs12->guest_ss_limit = evmcs->guest_ss_limit;
1689 		vmcs12->guest_ds_limit = evmcs->guest_ds_limit;
1690 		vmcs12->guest_fs_limit = evmcs->guest_fs_limit;
1691 		vmcs12->guest_gs_limit = evmcs->guest_gs_limit;
1692 		vmcs12->guest_ldtr_limit = evmcs->guest_ldtr_limit;
1693 		vmcs12->guest_tr_limit = evmcs->guest_tr_limit;
1694 		vmcs12->guest_gdtr_limit = evmcs->guest_gdtr_limit;
1695 		vmcs12->guest_idtr_limit = evmcs->guest_idtr_limit;
1696 		vmcs12->guest_es_ar_bytes = evmcs->guest_es_ar_bytes;
1697 		vmcs12->guest_cs_ar_bytes = evmcs->guest_cs_ar_bytes;
1698 		vmcs12->guest_ss_ar_bytes = evmcs->guest_ss_ar_bytes;
1699 		vmcs12->guest_ds_ar_bytes = evmcs->guest_ds_ar_bytes;
1700 		vmcs12->guest_fs_ar_bytes = evmcs->guest_fs_ar_bytes;
1701 		vmcs12->guest_gs_ar_bytes = evmcs->guest_gs_ar_bytes;
1702 		vmcs12->guest_ldtr_ar_bytes = evmcs->guest_ldtr_ar_bytes;
1703 		vmcs12->guest_tr_ar_bytes = evmcs->guest_tr_ar_bytes;
1704 		vmcs12->guest_es_selector = evmcs->guest_es_selector;
1705 		vmcs12->guest_cs_selector = evmcs->guest_cs_selector;
1706 		vmcs12->guest_ss_selector = evmcs->guest_ss_selector;
1707 		vmcs12->guest_ds_selector = evmcs->guest_ds_selector;
1708 		vmcs12->guest_fs_selector = evmcs->guest_fs_selector;
1709 		vmcs12->guest_gs_selector = evmcs->guest_gs_selector;
1710 		vmcs12->guest_ldtr_selector = evmcs->guest_ldtr_selector;
1711 		vmcs12->guest_tr_selector = evmcs->guest_tr_selector;
1712 	}
1713 
1714 	if (unlikely(!(hv_clean_fields &
1715 		       HV_VMX_ENLIGHTENED_CLEAN_FIELD_CONTROL_GRP2))) {
1716 		vmcs12->tsc_offset = evmcs->tsc_offset;
1717 		vmcs12->virtual_apic_page_addr = evmcs->virtual_apic_page_addr;
1718 		vmcs12->xss_exit_bitmap = evmcs->xss_exit_bitmap;
1719 		vmcs12->encls_exiting_bitmap = evmcs->encls_exiting_bitmap;
1720 		vmcs12->tsc_multiplier = evmcs->tsc_multiplier;
1721 	}
1722 
1723 	if (unlikely(!(hv_clean_fields &
1724 		       HV_VMX_ENLIGHTENED_CLEAN_FIELD_CRDR))) {
1725 		vmcs12->cr0_guest_host_mask = evmcs->cr0_guest_host_mask;
1726 		vmcs12->cr4_guest_host_mask = evmcs->cr4_guest_host_mask;
1727 		vmcs12->cr0_read_shadow = evmcs->cr0_read_shadow;
1728 		vmcs12->cr4_read_shadow = evmcs->cr4_read_shadow;
1729 		vmcs12->guest_cr0 = evmcs->guest_cr0;
1730 		vmcs12->guest_cr3 = evmcs->guest_cr3;
1731 		vmcs12->guest_cr4 = evmcs->guest_cr4;
1732 		vmcs12->guest_dr7 = evmcs->guest_dr7;
1733 	}
1734 
1735 	if (unlikely(!(hv_clean_fields &
1736 		       HV_VMX_ENLIGHTENED_CLEAN_FIELD_HOST_POINTER))) {
1737 		vmcs12->host_fs_base = evmcs->host_fs_base;
1738 		vmcs12->host_gs_base = evmcs->host_gs_base;
1739 		vmcs12->host_tr_base = evmcs->host_tr_base;
1740 		vmcs12->host_gdtr_base = evmcs->host_gdtr_base;
1741 		vmcs12->host_idtr_base = evmcs->host_idtr_base;
1742 		vmcs12->host_rsp = evmcs->host_rsp;
1743 	}
1744 
1745 	if (unlikely(!(hv_clean_fields &
1746 		       HV_VMX_ENLIGHTENED_CLEAN_FIELD_CONTROL_XLAT))) {
1747 		vmcs12->ept_pointer = evmcs->ept_pointer;
1748 		vmcs12->virtual_processor_id = evmcs->virtual_processor_id;
1749 	}
1750 
1751 	if (unlikely(!(hv_clean_fields &
1752 		       HV_VMX_ENLIGHTENED_CLEAN_FIELD_GUEST_GRP1))) {
1753 		vmcs12->vmcs_link_pointer = evmcs->vmcs_link_pointer;
1754 		vmcs12->guest_ia32_debugctl = evmcs->guest_ia32_debugctl;
1755 		vmcs12->guest_ia32_pat = evmcs->guest_ia32_pat;
1756 		vmcs12->guest_ia32_efer = evmcs->guest_ia32_efer;
1757 		vmcs12->guest_pdptr0 = evmcs->guest_pdptr0;
1758 		vmcs12->guest_pdptr1 = evmcs->guest_pdptr1;
1759 		vmcs12->guest_pdptr2 = evmcs->guest_pdptr2;
1760 		vmcs12->guest_pdptr3 = evmcs->guest_pdptr3;
1761 		vmcs12->guest_pending_dbg_exceptions =
1762 			evmcs->guest_pending_dbg_exceptions;
1763 		vmcs12->guest_sysenter_esp = evmcs->guest_sysenter_esp;
1764 		vmcs12->guest_sysenter_eip = evmcs->guest_sysenter_eip;
1765 		vmcs12->guest_bndcfgs = evmcs->guest_bndcfgs;
1766 		vmcs12->guest_activity_state = evmcs->guest_activity_state;
1767 		vmcs12->guest_sysenter_cs = evmcs->guest_sysenter_cs;
1768 		vmcs12->guest_ia32_perf_global_ctrl = evmcs->guest_ia32_perf_global_ctrl;
1769 		/*
1770 		 * Not present in struct vmcs12:
1771 		 * vmcs12->guest_ia32_s_cet = evmcs->guest_ia32_s_cet;
1772 		 * vmcs12->guest_ia32_lbr_ctl = evmcs->guest_ia32_lbr_ctl;
1773 		 * vmcs12->guest_ia32_int_ssp_table_addr = evmcs->guest_ia32_int_ssp_table_addr;
1774 		 */
1775 	}
1776 
1777 	/*
1778 	 * Not used?
1779 	 * vmcs12->vm_exit_msr_store_addr = evmcs->vm_exit_msr_store_addr;
1780 	 * vmcs12->vm_exit_msr_load_addr = evmcs->vm_exit_msr_load_addr;
1781 	 * vmcs12->vm_entry_msr_load_addr = evmcs->vm_entry_msr_load_addr;
1782 	 * vmcs12->page_fault_error_code_mask =
1783 	 *		evmcs->page_fault_error_code_mask;
1784 	 * vmcs12->page_fault_error_code_match =
1785 	 *		evmcs->page_fault_error_code_match;
1786 	 * vmcs12->cr3_target_count = evmcs->cr3_target_count;
1787 	 * vmcs12->vm_exit_msr_store_count = evmcs->vm_exit_msr_store_count;
1788 	 * vmcs12->vm_exit_msr_load_count = evmcs->vm_exit_msr_load_count;
1789 	 * vmcs12->vm_entry_msr_load_count = evmcs->vm_entry_msr_load_count;
1790 	 */
1791 
1792 	/*
1793 	 * Read only fields:
1794 	 * vmcs12->guest_physical_address = evmcs->guest_physical_address;
1795 	 * vmcs12->vm_instruction_error = evmcs->vm_instruction_error;
1796 	 * vmcs12->vm_exit_reason = evmcs->vm_exit_reason;
1797 	 * vmcs12->vm_exit_intr_info = evmcs->vm_exit_intr_info;
1798 	 * vmcs12->vm_exit_intr_error_code = evmcs->vm_exit_intr_error_code;
1799 	 * vmcs12->idt_vectoring_info_field = evmcs->idt_vectoring_info_field;
1800 	 * vmcs12->idt_vectoring_error_code = evmcs->idt_vectoring_error_code;
1801 	 * vmcs12->vm_exit_instruction_len = evmcs->vm_exit_instruction_len;
1802 	 * vmcs12->vmx_instruction_info = evmcs->vmx_instruction_info;
1803 	 * vmcs12->exit_qualification = evmcs->exit_qualification;
1804 	 * vmcs12->guest_linear_address = evmcs->guest_linear_address;
1805 	 *
1806 	 * Not present in struct vmcs12:
1807 	 * vmcs12->exit_io_instruction_ecx = evmcs->exit_io_instruction_ecx;
1808 	 * vmcs12->exit_io_instruction_esi = evmcs->exit_io_instruction_esi;
1809 	 * vmcs12->exit_io_instruction_edi = evmcs->exit_io_instruction_edi;
1810 	 * vmcs12->exit_io_instruction_eip = evmcs->exit_io_instruction_eip;
1811 	 */
1812 
1813 	return;
1814 }
1815 
1816 static void copy_vmcs12_to_enlightened(struct vcpu_vmx *vmx)
1817 {
1818 	struct vmcs12 *vmcs12 = vmx->nested.cached_vmcs12;
1819 	struct hv_enlightened_vmcs *evmcs = vmx->nested.hv_evmcs;
1820 
1821 	/*
1822 	 * Should not be changed by KVM:
1823 	 *
1824 	 * evmcs->host_es_selector = vmcs12->host_es_selector;
1825 	 * evmcs->host_cs_selector = vmcs12->host_cs_selector;
1826 	 * evmcs->host_ss_selector = vmcs12->host_ss_selector;
1827 	 * evmcs->host_ds_selector = vmcs12->host_ds_selector;
1828 	 * evmcs->host_fs_selector = vmcs12->host_fs_selector;
1829 	 * evmcs->host_gs_selector = vmcs12->host_gs_selector;
1830 	 * evmcs->host_tr_selector = vmcs12->host_tr_selector;
1831 	 * evmcs->host_ia32_pat = vmcs12->host_ia32_pat;
1832 	 * evmcs->host_ia32_efer = vmcs12->host_ia32_efer;
1833 	 * evmcs->host_cr0 = vmcs12->host_cr0;
1834 	 * evmcs->host_cr3 = vmcs12->host_cr3;
1835 	 * evmcs->host_cr4 = vmcs12->host_cr4;
1836 	 * evmcs->host_ia32_sysenter_esp = vmcs12->host_ia32_sysenter_esp;
1837 	 * evmcs->host_ia32_sysenter_eip = vmcs12->host_ia32_sysenter_eip;
1838 	 * evmcs->host_rip = vmcs12->host_rip;
1839 	 * evmcs->host_ia32_sysenter_cs = vmcs12->host_ia32_sysenter_cs;
1840 	 * evmcs->host_fs_base = vmcs12->host_fs_base;
1841 	 * evmcs->host_gs_base = vmcs12->host_gs_base;
1842 	 * evmcs->host_tr_base = vmcs12->host_tr_base;
1843 	 * evmcs->host_gdtr_base = vmcs12->host_gdtr_base;
1844 	 * evmcs->host_idtr_base = vmcs12->host_idtr_base;
1845 	 * evmcs->host_rsp = vmcs12->host_rsp;
1846 	 * sync_vmcs02_to_vmcs12() doesn't read these:
1847 	 * evmcs->io_bitmap_a = vmcs12->io_bitmap_a;
1848 	 * evmcs->io_bitmap_b = vmcs12->io_bitmap_b;
1849 	 * evmcs->msr_bitmap = vmcs12->msr_bitmap;
1850 	 * evmcs->ept_pointer = vmcs12->ept_pointer;
1851 	 * evmcs->xss_exit_bitmap = vmcs12->xss_exit_bitmap;
1852 	 * evmcs->vm_exit_msr_store_addr = vmcs12->vm_exit_msr_store_addr;
1853 	 * evmcs->vm_exit_msr_load_addr = vmcs12->vm_exit_msr_load_addr;
1854 	 * evmcs->vm_entry_msr_load_addr = vmcs12->vm_entry_msr_load_addr;
1855 	 * evmcs->tpr_threshold = vmcs12->tpr_threshold;
1856 	 * evmcs->virtual_processor_id = vmcs12->virtual_processor_id;
1857 	 * evmcs->exception_bitmap = vmcs12->exception_bitmap;
1858 	 * evmcs->vmcs_link_pointer = vmcs12->vmcs_link_pointer;
1859 	 * evmcs->pin_based_vm_exec_control = vmcs12->pin_based_vm_exec_control;
1860 	 * evmcs->vm_exit_controls = vmcs12->vm_exit_controls;
1861 	 * evmcs->secondary_vm_exec_control = vmcs12->secondary_vm_exec_control;
1862 	 * evmcs->page_fault_error_code_mask =
1863 	 *		vmcs12->page_fault_error_code_mask;
1864 	 * evmcs->page_fault_error_code_match =
1865 	 *		vmcs12->page_fault_error_code_match;
1866 	 * evmcs->cr3_target_count = vmcs12->cr3_target_count;
1867 	 * evmcs->virtual_apic_page_addr = vmcs12->virtual_apic_page_addr;
1868 	 * evmcs->tsc_offset = vmcs12->tsc_offset;
1869 	 * evmcs->guest_ia32_debugctl = vmcs12->guest_ia32_debugctl;
1870 	 * evmcs->cr0_guest_host_mask = vmcs12->cr0_guest_host_mask;
1871 	 * evmcs->cr4_guest_host_mask = vmcs12->cr4_guest_host_mask;
1872 	 * evmcs->cr0_read_shadow = vmcs12->cr0_read_shadow;
1873 	 * evmcs->cr4_read_shadow = vmcs12->cr4_read_shadow;
1874 	 * evmcs->vm_exit_msr_store_count = vmcs12->vm_exit_msr_store_count;
1875 	 * evmcs->vm_exit_msr_load_count = vmcs12->vm_exit_msr_load_count;
1876 	 * evmcs->vm_entry_msr_load_count = vmcs12->vm_entry_msr_load_count;
1877 	 * evmcs->guest_ia32_perf_global_ctrl = vmcs12->guest_ia32_perf_global_ctrl;
1878 	 * evmcs->host_ia32_perf_global_ctrl = vmcs12->host_ia32_perf_global_ctrl;
1879 	 * evmcs->encls_exiting_bitmap = vmcs12->encls_exiting_bitmap;
1880 	 * evmcs->tsc_multiplier = vmcs12->tsc_multiplier;
1881 	 *
1882 	 * Not present in struct vmcs12:
1883 	 * evmcs->exit_io_instruction_ecx = vmcs12->exit_io_instruction_ecx;
1884 	 * evmcs->exit_io_instruction_esi = vmcs12->exit_io_instruction_esi;
1885 	 * evmcs->exit_io_instruction_edi = vmcs12->exit_io_instruction_edi;
1886 	 * evmcs->exit_io_instruction_eip = vmcs12->exit_io_instruction_eip;
1887 	 * evmcs->host_ia32_s_cet = vmcs12->host_ia32_s_cet;
1888 	 * evmcs->host_ssp = vmcs12->host_ssp;
1889 	 * evmcs->host_ia32_int_ssp_table_addr = vmcs12->host_ia32_int_ssp_table_addr;
1890 	 * evmcs->guest_ia32_s_cet = vmcs12->guest_ia32_s_cet;
1891 	 * evmcs->guest_ia32_lbr_ctl = vmcs12->guest_ia32_lbr_ctl;
1892 	 * evmcs->guest_ia32_int_ssp_table_addr = vmcs12->guest_ia32_int_ssp_table_addr;
1893 	 * evmcs->guest_ssp = vmcs12->guest_ssp;
1894 	 */
1895 
1896 	evmcs->guest_es_selector = vmcs12->guest_es_selector;
1897 	evmcs->guest_cs_selector = vmcs12->guest_cs_selector;
1898 	evmcs->guest_ss_selector = vmcs12->guest_ss_selector;
1899 	evmcs->guest_ds_selector = vmcs12->guest_ds_selector;
1900 	evmcs->guest_fs_selector = vmcs12->guest_fs_selector;
1901 	evmcs->guest_gs_selector = vmcs12->guest_gs_selector;
1902 	evmcs->guest_ldtr_selector = vmcs12->guest_ldtr_selector;
1903 	evmcs->guest_tr_selector = vmcs12->guest_tr_selector;
1904 
1905 	evmcs->guest_es_limit = vmcs12->guest_es_limit;
1906 	evmcs->guest_cs_limit = vmcs12->guest_cs_limit;
1907 	evmcs->guest_ss_limit = vmcs12->guest_ss_limit;
1908 	evmcs->guest_ds_limit = vmcs12->guest_ds_limit;
1909 	evmcs->guest_fs_limit = vmcs12->guest_fs_limit;
1910 	evmcs->guest_gs_limit = vmcs12->guest_gs_limit;
1911 	evmcs->guest_ldtr_limit = vmcs12->guest_ldtr_limit;
1912 	evmcs->guest_tr_limit = vmcs12->guest_tr_limit;
1913 	evmcs->guest_gdtr_limit = vmcs12->guest_gdtr_limit;
1914 	evmcs->guest_idtr_limit = vmcs12->guest_idtr_limit;
1915 
1916 	evmcs->guest_es_ar_bytes = vmcs12->guest_es_ar_bytes;
1917 	evmcs->guest_cs_ar_bytes = vmcs12->guest_cs_ar_bytes;
1918 	evmcs->guest_ss_ar_bytes = vmcs12->guest_ss_ar_bytes;
1919 	evmcs->guest_ds_ar_bytes = vmcs12->guest_ds_ar_bytes;
1920 	evmcs->guest_fs_ar_bytes = vmcs12->guest_fs_ar_bytes;
1921 	evmcs->guest_gs_ar_bytes = vmcs12->guest_gs_ar_bytes;
1922 	evmcs->guest_ldtr_ar_bytes = vmcs12->guest_ldtr_ar_bytes;
1923 	evmcs->guest_tr_ar_bytes = vmcs12->guest_tr_ar_bytes;
1924 
1925 	evmcs->guest_es_base = vmcs12->guest_es_base;
1926 	evmcs->guest_cs_base = vmcs12->guest_cs_base;
1927 	evmcs->guest_ss_base = vmcs12->guest_ss_base;
1928 	evmcs->guest_ds_base = vmcs12->guest_ds_base;
1929 	evmcs->guest_fs_base = vmcs12->guest_fs_base;
1930 	evmcs->guest_gs_base = vmcs12->guest_gs_base;
1931 	evmcs->guest_ldtr_base = vmcs12->guest_ldtr_base;
1932 	evmcs->guest_tr_base = vmcs12->guest_tr_base;
1933 	evmcs->guest_gdtr_base = vmcs12->guest_gdtr_base;
1934 	evmcs->guest_idtr_base = vmcs12->guest_idtr_base;
1935 
1936 	evmcs->guest_ia32_pat = vmcs12->guest_ia32_pat;
1937 	evmcs->guest_ia32_efer = vmcs12->guest_ia32_efer;
1938 
1939 	evmcs->guest_pdptr0 = vmcs12->guest_pdptr0;
1940 	evmcs->guest_pdptr1 = vmcs12->guest_pdptr1;
1941 	evmcs->guest_pdptr2 = vmcs12->guest_pdptr2;
1942 	evmcs->guest_pdptr3 = vmcs12->guest_pdptr3;
1943 
1944 	evmcs->guest_pending_dbg_exceptions =
1945 		vmcs12->guest_pending_dbg_exceptions;
1946 	evmcs->guest_sysenter_esp = vmcs12->guest_sysenter_esp;
1947 	evmcs->guest_sysenter_eip = vmcs12->guest_sysenter_eip;
1948 
1949 	evmcs->guest_activity_state = vmcs12->guest_activity_state;
1950 	evmcs->guest_sysenter_cs = vmcs12->guest_sysenter_cs;
1951 
1952 	evmcs->guest_cr0 = vmcs12->guest_cr0;
1953 	evmcs->guest_cr3 = vmcs12->guest_cr3;
1954 	evmcs->guest_cr4 = vmcs12->guest_cr4;
1955 	evmcs->guest_dr7 = vmcs12->guest_dr7;
1956 
1957 	evmcs->guest_physical_address = vmcs12->guest_physical_address;
1958 
1959 	evmcs->vm_instruction_error = vmcs12->vm_instruction_error;
1960 	evmcs->vm_exit_reason = vmcs12->vm_exit_reason;
1961 	evmcs->vm_exit_intr_info = vmcs12->vm_exit_intr_info;
1962 	evmcs->vm_exit_intr_error_code = vmcs12->vm_exit_intr_error_code;
1963 	evmcs->idt_vectoring_info_field = vmcs12->idt_vectoring_info_field;
1964 	evmcs->idt_vectoring_error_code = vmcs12->idt_vectoring_error_code;
1965 	evmcs->vm_exit_instruction_len = vmcs12->vm_exit_instruction_len;
1966 	evmcs->vmx_instruction_info = vmcs12->vmx_instruction_info;
1967 
1968 	evmcs->exit_qualification = vmcs12->exit_qualification;
1969 
1970 	evmcs->guest_linear_address = vmcs12->guest_linear_address;
1971 	evmcs->guest_rsp = vmcs12->guest_rsp;
1972 	evmcs->guest_rflags = vmcs12->guest_rflags;
1973 
1974 	evmcs->guest_interruptibility_info =
1975 		vmcs12->guest_interruptibility_info;
1976 	evmcs->cpu_based_vm_exec_control = vmcs12->cpu_based_vm_exec_control;
1977 	evmcs->vm_entry_controls = vmcs12->vm_entry_controls;
1978 	evmcs->vm_entry_intr_info_field = vmcs12->vm_entry_intr_info_field;
1979 	evmcs->vm_entry_exception_error_code =
1980 		vmcs12->vm_entry_exception_error_code;
1981 	evmcs->vm_entry_instruction_len = vmcs12->vm_entry_instruction_len;
1982 
1983 	evmcs->guest_rip = vmcs12->guest_rip;
1984 
1985 	evmcs->guest_bndcfgs = vmcs12->guest_bndcfgs;
1986 
1987 	return;
1988 }
1989 
1990 /*
1991  * This is an equivalent of the nested hypervisor executing the vmptrld
1992  * instruction.
1993  */
1994 static enum nested_evmptrld_status nested_vmx_handle_enlightened_vmptrld(
1995 	struct kvm_vcpu *vcpu, bool from_launch)
1996 {
1997 	struct vcpu_vmx *vmx = to_vmx(vcpu);
1998 	bool evmcs_gpa_changed = false;
1999 	u64 evmcs_gpa;
2000 
2001 	if (likely(!guest_cpuid_has_evmcs(vcpu)))
2002 		return EVMPTRLD_DISABLED;
2003 
2004 	evmcs_gpa = nested_get_evmptr(vcpu);
2005 	if (!evmptr_is_valid(evmcs_gpa)) {
2006 		nested_release_evmcs(vcpu);
2007 		return EVMPTRLD_DISABLED;
2008 	}
2009 
2010 	if (unlikely(evmcs_gpa != vmx->nested.hv_evmcs_vmptr)) {
2011 		vmx->nested.current_vmptr = INVALID_GPA;
2012 
2013 		nested_release_evmcs(vcpu);
2014 
2015 		if (kvm_vcpu_map(vcpu, gpa_to_gfn(evmcs_gpa),
2016 				 &vmx->nested.hv_evmcs_map))
2017 			return EVMPTRLD_ERROR;
2018 
2019 		vmx->nested.hv_evmcs = vmx->nested.hv_evmcs_map.hva;
2020 
2021 		/*
2022 		 * Currently, KVM only supports eVMCS version 1
2023 		 * (== KVM_EVMCS_VERSION) and thus we expect guest to set this
2024 		 * value to first u32 field of eVMCS which should specify eVMCS
2025 		 * VersionNumber.
2026 		 *
2027 		 * Guest should be aware of supported eVMCS versions by host by
2028 		 * examining CPUID.0x4000000A.EAX[0:15]. Host userspace VMM is
2029 		 * expected to set this CPUID leaf according to the value
2030 		 * returned in vmcs_version from nested_enable_evmcs().
2031 		 *
2032 		 * However, it turns out that Microsoft Hyper-V fails to comply
2033 		 * to their own invented interface: When Hyper-V use eVMCS, it
2034 		 * just sets first u32 field of eVMCS to revision_id specified
2035 		 * in MSR_IA32_VMX_BASIC. Instead of used eVMCS version number
2036 		 * which is one of the supported versions specified in
2037 		 * CPUID.0x4000000A.EAX[0:15].
2038 		 *
2039 		 * To overcome Hyper-V bug, we accept here either a supported
2040 		 * eVMCS version or VMCS12 revision_id as valid values for first
2041 		 * u32 field of eVMCS.
2042 		 */
2043 		if ((vmx->nested.hv_evmcs->revision_id != KVM_EVMCS_VERSION) &&
2044 		    (vmx->nested.hv_evmcs->revision_id != VMCS12_REVISION)) {
2045 			nested_release_evmcs(vcpu);
2046 			return EVMPTRLD_VMFAIL;
2047 		}
2048 
2049 		vmx->nested.hv_evmcs_vmptr = evmcs_gpa;
2050 
2051 		evmcs_gpa_changed = true;
2052 		/*
2053 		 * Unlike normal vmcs12, enlightened vmcs12 is not fully
2054 		 * reloaded from guest's memory (read only fields, fields not
2055 		 * present in struct hv_enlightened_vmcs, ...). Make sure there
2056 		 * are no leftovers.
2057 		 */
2058 		if (from_launch) {
2059 			struct vmcs12 *vmcs12 = get_vmcs12(vcpu);
2060 			memset(vmcs12, 0, sizeof(*vmcs12));
2061 			vmcs12->hdr.revision_id = VMCS12_REVISION;
2062 		}
2063 
2064 	}
2065 
2066 	/*
2067 	 * Clean fields data can't be used on VMLAUNCH and when we switch
2068 	 * between different L2 guests as KVM keeps a single VMCS12 per L1.
2069 	 */
2070 	if (from_launch || evmcs_gpa_changed) {
2071 		vmx->nested.hv_evmcs->hv_clean_fields &=
2072 			~HV_VMX_ENLIGHTENED_CLEAN_FIELD_ALL;
2073 
2074 		vmx->nested.force_msr_bitmap_recalc = true;
2075 	}
2076 
2077 	return EVMPTRLD_SUCCEEDED;
2078 }
2079 
2080 void nested_sync_vmcs12_to_shadow(struct kvm_vcpu *vcpu)
2081 {
2082 	struct vcpu_vmx *vmx = to_vmx(vcpu);
2083 
2084 	if (evmptr_is_valid(vmx->nested.hv_evmcs_vmptr))
2085 		copy_vmcs12_to_enlightened(vmx);
2086 	else
2087 		copy_vmcs12_to_shadow(vmx);
2088 
2089 	vmx->nested.need_vmcs12_to_shadow_sync = false;
2090 }
2091 
2092 static enum hrtimer_restart vmx_preemption_timer_fn(struct hrtimer *timer)
2093 {
2094 	struct vcpu_vmx *vmx =
2095 		container_of(timer, struct vcpu_vmx, nested.preemption_timer);
2096 
2097 	vmx->nested.preemption_timer_expired = true;
2098 	kvm_make_request(KVM_REQ_EVENT, &vmx->vcpu);
2099 	kvm_vcpu_kick(&vmx->vcpu);
2100 
2101 	return HRTIMER_NORESTART;
2102 }
2103 
2104 static u64 vmx_calc_preemption_timer_value(struct kvm_vcpu *vcpu)
2105 {
2106 	struct vcpu_vmx *vmx = to_vmx(vcpu);
2107 	struct vmcs12 *vmcs12 = get_vmcs12(vcpu);
2108 
2109 	u64 l1_scaled_tsc = kvm_read_l1_tsc(vcpu, rdtsc()) >>
2110 			    VMX_MISC_EMULATED_PREEMPTION_TIMER_RATE;
2111 
2112 	if (!vmx->nested.has_preemption_timer_deadline) {
2113 		vmx->nested.preemption_timer_deadline =
2114 			vmcs12->vmx_preemption_timer_value + l1_scaled_tsc;
2115 		vmx->nested.has_preemption_timer_deadline = true;
2116 	}
2117 	return vmx->nested.preemption_timer_deadline - l1_scaled_tsc;
2118 }
2119 
2120 static void vmx_start_preemption_timer(struct kvm_vcpu *vcpu,
2121 					u64 preemption_timeout)
2122 {
2123 	struct vcpu_vmx *vmx = to_vmx(vcpu);
2124 
2125 	/*
2126 	 * A timer value of zero is architecturally guaranteed to cause
2127 	 * a VMExit prior to executing any instructions in the guest.
2128 	 */
2129 	if (preemption_timeout == 0) {
2130 		vmx_preemption_timer_fn(&vmx->nested.preemption_timer);
2131 		return;
2132 	}
2133 
2134 	if (vcpu->arch.virtual_tsc_khz == 0)
2135 		return;
2136 
2137 	preemption_timeout <<= VMX_MISC_EMULATED_PREEMPTION_TIMER_RATE;
2138 	preemption_timeout *= 1000000;
2139 	do_div(preemption_timeout, vcpu->arch.virtual_tsc_khz);
2140 	hrtimer_start(&vmx->nested.preemption_timer,
2141 		      ktime_add_ns(ktime_get(), preemption_timeout),
2142 		      HRTIMER_MODE_ABS_PINNED);
2143 }
2144 
2145 static u64 nested_vmx_calc_efer(struct vcpu_vmx *vmx, struct vmcs12 *vmcs12)
2146 {
2147 	if (vmx->nested.nested_run_pending &&
2148 	    (vmcs12->vm_entry_controls & VM_ENTRY_LOAD_IA32_EFER))
2149 		return vmcs12->guest_ia32_efer;
2150 	else if (vmcs12->vm_entry_controls & VM_ENTRY_IA32E_MODE)
2151 		return vmx->vcpu.arch.efer | (EFER_LMA | EFER_LME);
2152 	else
2153 		return vmx->vcpu.arch.efer & ~(EFER_LMA | EFER_LME);
2154 }
2155 
2156 static void prepare_vmcs02_constant_state(struct vcpu_vmx *vmx)
2157 {
2158 	struct kvm *kvm = vmx->vcpu.kvm;
2159 
2160 	/*
2161 	 * If vmcs02 hasn't been initialized, set the constant vmcs02 state
2162 	 * according to L0's settings (vmcs12 is irrelevant here).  Host
2163 	 * fields that come from L0 and are not constant, e.g. HOST_CR3,
2164 	 * will be set as needed prior to VMLAUNCH/VMRESUME.
2165 	 */
2166 	if (vmx->nested.vmcs02_initialized)
2167 		return;
2168 	vmx->nested.vmcs02_initialized = true;
2169 
2170 	/*
2171 	 * We don't care what the EPTP value is we just need to guarantee
2172 	 * it's valid so we don't get a false positive when doing early
2173 	 * consistency checks.
2174 	 */
2175 	if (enable_ept && nested_early_check)
2176 		vmcs_write64(EPT_POINTER,
2177 			     construct_eptp(&vmx->vcpu, 0, PT64_ROOT_4LEVEL));
2178 
2179 	/* All VMFUNCs are currently emulated through L0 vmexits.  */
2180 	if (cpu_has_vmx_vmfunc())
2181 		vmcs_write64(VM_FUNCTION_CONTROL, 0);
2182 
2183 	if (cpu_has_vmx_posted_intr())
2184 		vmcs_write16(POSTED_INTR_NV, POSTED_INTR_NESTED_VECTOR);
2185 
2186 	if (cpu_has_vmx_msr_bitmap())
2187 		vmcs_write64(MSR_BITMAP, __pa(vmx->nested.vmcs02.msr_bitmap));
2188 
2189 	/*
2190 	 * PML is emulated for L2, but never enabled in hardware as the MMU
2191 	 * handles A/D emulation.  Disabling PML for L2 also avoids having to
2192 	 * deal with filtering out L2 GPAs from the buffer.
2193 	 */
2194 	if (enable_pml) {
2195 		vmcs_write64(PML_ADDRESS, 0);
2196 		vmcs_write16(GUEST_PML_INDEX, -1);
2197 	}
2198 
2199 	if (cpu_has_vmx_encls_vmexit())
2200 		vmcs_write64(ENCLS_EXITING_BITMAP, INVALID_GPA);
2201 
2202 	if (kvm_notify_vmexit_enabled(kvm))
2203 		vmcs_write32(NOTIFY_WINDOW, kvm->arch.notify_window);
2204 
2205 	/*
2206 	 * Set the MSR load/store lists to match L0's settings.  Only the
2207 	 * addresses are constant (for vmcs02), the counts can change based
2208 	 * on L2's behavior, e.g. switching to/from long mode.
2209 	 */
2210 	vmcs_write64(VM_EXIT_MSR_STORE_ADDR, __pa(vmx->msr_autostore.guest.val));
2211 	vmcs_write64(VM_EXIT_MSR_LOAD_ADDR, __pa(vmx->msr_autoload.host.val));
2212 	vmcs_write64(VM_ENTRY_MSR_LOAD_ADDR, __pa(vmx->msr_autoload.guest.val));
2213 
2214 	vmx_set_constant_host_state(vmx);
2215 }
2216 
2217 static void prepare_vmcs02_early_rare(struct vcpu_vmx *vmx,
2218 				      struct vmcs12 *vmcs12)
2219 {
2220 	prepare_vmcs02_constant_state(vmx);
2221 
2222 	vmcs_write64(VMCS_LINK_POINTER, INVALID_GPA);
2223 
2224 	if (enable_vpid) {
2225 		if (nested_cpu_has_vpid(vmcs12) && vmx->nested.vpid02)
2226 			vmcs_write16(VIRTUAL_PROCESSOR_ID, vmx->nested.vpid02);
2227 		else
2228 			vmcs_write16(VIRTUAL_PROCESSOR_ID, vmx->vpid);
2229 	}
2230 }
2231 
2232 static void prepare_vmcs02_early(struct vcpu_vmx *vmx, struct loaded_vmcs *vmcs01,
2233 				 struct vmcs12 *vmcs12)
2234 {
2235 	u32 exec_control;
2236 	u64 guest_efer = nested_vmx_calc_efer(vmx, vmcs12);
2237 
2238 	if (vmx->nested.dirty_vmcs12 || evmptr_is_valid(vmx->nested.hv_evmcs_vmptr))
2239 		prepare_vmcs02_early_rare(vmx, vmcs12);
2240 
2241 	/*
2242 	 * PIN CONTROLS
2243 	 */
2244 	exec_control = __pin_controls_get(vmcs01);
2245 	exec_control |= (vmcs12->pin_based_vm_exec_control &
2246 			 ~PIN_BASED_VMX_PREEMPTION_TIMER);
2247 
2248 	/* Posted interrupts setting is only taken from vmcs12.  */
2249 	vmx->nested.pi_pending = false;
2250 	if (nested_cpu_has_posted_intr(vmcs12))
2251 		vmx->nested.posted_intr_nv = vmcs12->posted_intr_nv;
2252 	else
2253 		exec_control &= ~PIN_BASED_POSTED_INTR;
2254 	pin_controls_set(vmx, exec_control);
2255 
2256 	/*
2257 	 * EXEC CONTROLS
2258 	 */
2259 	exec_control = __exec_controls_get(vmcs01); /* L0's desires */
2260 	exec_control &= ~CPU_BASED_INTR_WINDOW_EXITING;
2261 	exec_control &= ~CPU_BASED_NMI_WINDOW_EXITING;
2262 	exec_control &= ~CPU_BASED_TPR_SHADOW;
2263 	exec_control |= vmcs12->cpu_based_vm_exec_control;
2264 
2265 	vmx->nested.l1_tpr_threshold = -1;
2266 	if (exec_control & CPU_BASED_TPR_SHADOW)
2267 		vmcs_write32(TPR_THRESHOLD, vmcs12->tpr_threshold);
2268 #ifdef CONFIG_X86_64
2269 	else
2270 		exec_control |= CPU_BASED_CR8_LOAD_EXITING |
2271 				CPU_BASED_CR8_STORE_EXITING;
2272 #endif
2273 
2274 	/*
2275 	 * A vmexit (to either L1 hypervisor or L0 userspace) is always needed
2276 	 * for I/O port accesses.
2277 	 */
2278 	exec_control |= CPU_BASED_UNCOND_IO_EXITING;
2279 	exec_control &= ~CPU_BASED_USE_IO_BITMAPS;
2280 
2281 	/*
2282 	 * This bit will be computed in nested_get_vmcs12_pages, because
2283 	 * we do not have access to L1's MSR bitmap yet.  For now, keep
2284 	 * the same bit as before, hoping to avoid multiple VMWRITEs that
2285 	 * only set/clear this bit.
2286 	 */
2287 	exec_control &= ~CPU_BASED_USE_MSR_BITMAPS;
2288 	exec_control |= exec_controls_get(vmx) & CPU_BASED_USE_MSR_BITMAPS;
2289 
2290 	exec_controls_set(vmx, exec_control);
2291 
2292 	/*
2293 	 * SECONDARY EXEC CONTROLS
2294 	 */
2295 	if (cpu_has_secondary_exec_ctrls()) {
2296 		exec_control = __secondary_exec_controls_get(vmcs01);
2297 
2298 		/* Take the following fields only from vmcs12 */
2299 		exec_control &= ~(SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES |
2300 				  SECONDARY_EXEC_VIRTUALIZE_X2APIC_MODE |
2301 				  SECONDARY_EXEC_ENABLE_INVPCID |
2302 				  SECONDARY_EXEC_ENABLE_RDTSCP |
2303 				  SECONDARY_EXEC_XSAVES |
2304 				  SECONDARY_EXEC_ENABLE_USR_WAIT_PAUSE |
2305 				  SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY |
2306 				  SECONDARY_EXEC_APIC_REGISTER_VIRT |
2307 				  SECONDARY_EXEC_ENABLE_VMFUNC |
2308 				  SECONDARY_EXEC_DESC);
2309 
2310 		if (nested_cpu_has(vmcs12,
2311 				   CPU_BASED_ACTIVATE_SECONDARY_CONTROLS))
2312 			exec_control |= vmcs12->secondary_vm_exec_control;
2313 
2314 		/* PML is emulated and never enabled in hardware for L2. */
2315 		exec_control &= ~SECONDARY_EXEC_ENABLE_PML;
2316 
2317 		/* VMCS shadowing for L2 is emulated for now */
2318 		exec_control &= ~SECONDARY_EXEC_SHADOW_VMCS;
2319 
2320 		/*
2321 		 * Preset *DT exiting when emulating UMIP, so that vmx_set_cr4()
2322 		 * will not have to rewrite the controls just for this bit.
2323 		 */
2324 		if (!boot_cpu_has(X86_FEATURE_UMIP) && vmx_umip_emulated() &&
2325 		    (vmcs12->guest_cr4 & X86_CR4_UMIP))
2326 			exec_control |= SECONDARY_EXEC_DESC;
2327 
2328 		if (exec_control & SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY)
2329 			vmcs_write16(GUEST_INTR_STATUS,
2330 				vmcs12->guest_intr_status);
2331 
2332 		if (!nested_cpu_has2(vmcs12, SECONDARY_EXEC_UNRESTRICTED_GUEST))
2333 		    exec_control &= ~SECONDARY_EXEC_UNRESTRICTED_GUEST;
2334 
2335 		if (exec_control & SECONDARY_EXEC_ENCLS_EXITING)
2336 			vmx_write_encls_bitmap(&vmx->vcpu, vmcs12);
2337 
2338 		secondary_exec_controls_set(vmx, exec_control);
2339 	}
2340 
2341 	/*
2342 	 * ENTRY CONTROLS
2343 	 *
2344 	 * vmcs12's VM_{ENTRY,EXIT}_LOAD_IA32_EFER and VM_ENTRY_IA32E_MODE
2345 	 * are emulated by vmx_set_efer() in prepare_vmcs02(), but speculate
2346 	 * on the related bits (if supported by the CPU) in the hope that
2347 	 * we can avoid VMWrites during vmx_set_efer().
2348 	 *
2349 	 * Similarly, take vmcs01's PERF_GLOBAL_CTRL in the hope that if KVM is
2350 	 * loading PERF_GLOBAL_CTRL via the VMCS for L1, then KVM will want to
2351 	 * do the same for L2.
2352 	 */
2353 	exec_control = __vm_entry_controls_get(vmcs01);
2354 	exec_control |= (vmcs12->vm_entry_controls &
2355 			 ~VM_ENTRY_LOAD_IA32_PERF_GLOBAL_CTRL);
2356 	exec_control &= ~(VM_ENTRY_IA32E_MODE | VM_ENTRY_LOAD_IA32_EFER);
2357 	if (cpu_has_load_ia32_efer()) {
2358 		if (guest_efer & EFER_LMA)
2359 			exec_control |= VM_ENTRY_IA32E_MODE;
2360 		if (guest_efer != host_efer)
2361 			exec_control |= VM_ENTRY_LOAD_IA32_EFER;
2362 	}
2363 	vm_entry_controls_set(vmx, exec_control);
2364 
2365 	/*
2366 	 * EXIT CONTROLS
2367 	 *
2368 	 * L2->L1 exit controls are emulated - the hardware exit is to L0 so
2369 	 * we should use its exit controls. Note that VM_EXIT_LOAD_IA32_EFER
2370 	 * bits may be modified by vmx_set_efer() in prepare_vmcs02().
2371 	 */
2372 	exec_control = __vm_exit_controls_get(vmcs01);
2373 	if (cpu_has_load_ia32_efer() && guest_efer != host_efer)
2374 		exec_control |= VM_EXIT_LOAD_IA32_EFER;
2375 	else
2376 		exec_control &= ~VM_EXIT_LOAD_IA32_EFER;
2377 	vm_exit_controls_set(vmx, exec_control);
2378 
2379 	/*
2380 	 * Interrupt/Exception Fields
2381 	 */
2382 	if (vmx->nested.nested_run_pending) {
2383 		vmcs_write32(VM_ENTRY_INTR_INFO_FIELD,
2384 			     vmcs12->vm_entry_intr_info_field);
2385 		vmcs_write32(VM_ENTRY_EXCEPTION_ERROR_CODE,
2386 			     vmcs12->vm_entry_exception_error_code);
2387 		vmcs_write32(VM_ENTRY_INSTRUCTION_LEN,
2388 			     vmcs12->vm_entry_instruction_len);
2389 		vmcs_write32(GUEST_INTERRUPTIBILITY_INFO,
2390 			     vmcs12->guest_interruptibility_info);
2391 		vmx->loaded_vmcs->nmi_known_unmasked =
2392 			!(vmcs12->guest_interruptibility_info & GUEST_INTR_STATE_NMI);
2393 	} else {
2394 		vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, 0);
2395 	}
2396 }
2397 
2398 static void prepare_vmcs02_rare(struct vcpu_vmx *vmx, struct vmcs12 *vmcs12)
2399 {
2400 	struct hv_enlightened_vmcs *hv_evmcs = vmx->nested.hv_evmcs;
2401 
2402 	if (!hv_evmcs || !(hv_evmcs->hv_clean_fields &
2403 			   HV_VMX_ENLIGHTENED_CLEAN_FIELD_GUEST_GRP2)) {
2404 		vmcs_write16(GUEST_ES_SELECTOR, vmcs12->guest_es_selector);
2405 		vmcs_write16(GUEST_CS_SELECTOR, vmcs12->guest_cs_selector);
2406 		vmcs_write16(GUEST_SS_SELECTOR, vmcs12->guest_ss_selector);
2407 		vmcs_write16(GUEST_DS_SELECTOR, vmcs12->guest_ds_selector);
2408 		vmcs_write16(GUEST_FS_SELECTOR, vmcs12->guest_fs_selector);
2409 		vmcs_write16(GUEST_GS_SELECTOR, vmcs12->guest_gs_selector);
2410 		vmcs_write16(GUEST_LDTR_SELECTOR, vmcs12->guest_ldtr_selector);
2411 		vmcs_write16(GUEST_TR_SELECTOR, vmcs12->guest_tr_selector);
2412 		vmcs_write32(GUEST_ES_LIMIT, vmcs12->guest_es_limit);
2413 		vmcs_write32(GUEST_CS_LIMIT, vmcs12->guest_cs_limit);
2414 		vmcs_write32(GUEST_SS_LIMIT, vmcs12->guest_ss_limit);
2415 		vmcs_write32(GUEST_DS_LIMIT, vmcs12->guest_ds_limit);
2416 		vmcs_write32(GUEST_FS_LIMIT, vmcs12->guest_fs_limit);
2417 		vmcs_write32(GUEST_GS_LIMIT, vmcs12->guest_gs_limit);
2418 		vmcs_write32(GUEST_LDTR_LIMIT, vmcs12->guest_ldtr_limit);
2419 		vmcs_write32(GUEST_TR_LIMIT, vmcs12->guest_tr_limit);
2420 		vmcs_write32(GUEST_GDTR_LIMIT, vmcs12->guest_gdtr_limit);
2421 		vmcs_write32(GUEST_IDTR_LIMIT, vmcs12->guest_idtr_limit);
2422 		vmcs_write32(GUEST_CS_AR_BYTES, vmcs12->guest_cs_ar_bytes);
2423 		vmcs_write32(GUEST_SS_AR_BYTES, vmcs12->guest_ss_ar_bytes);
2424 		vmcs_write32(GUEST_ES_AR_BYTES, vmcs12->guest_es_ar_bytes);
2425 		vmcs_write32(GUEST_DS_AR_BYTES, vmcs12->guest_ds_ar_bytes);
2426 		vmcs_write32(GUEST_FS_AR_BYTES, vmcs12->guest_fs_ar_bytes);
2427 		vmcs_write32(GUEST_GS_AR_BYTES, vmcs12->guest_gs_ar_bytes);
2428 		vmcs_write32(GUEST_LDTR_AR_BYTES, vmcs12->guest_ldtr_ar_bytes);
2429 		vmcs_write32(GUEST_TR_AR_BYTES, vmcs12->guest_tr_ar_bytes);
2430 		vmcs_writel(GUEST_ES_BASE, vmcs12->guest_es_base);
2431 		vmcs_writel(GUEST_CS_BASE, vmcs12->guest_cs_base);
2432 		vmcs_writel(GUEST_SS_BASE, vmcs12->guest_ss_base);
2433 		vmcs_writel(GUEST_DS_BASE, vmcs12->guest_ds_base);
2434 		vmcs_writel(GUEST_FS_BASE, vmcs12->guest_fs_base);
2435 		vmcs_writel(GUEST_GS_BASE, vmcs12->guest_gs_base);
2436 		vmcs_writel(GUEST_LDTR_BASE, vmcs12->guest_ldtr_base);
2437 		vmcs_writel(GUEST_TR_BASE, vmcs12->guest_tr_base);
2438 		vmcs_writel(GUEST_GDTR_BASE, vmcs12->guest_gdtr_base);
2439 		vmcs_writel(GUEST_IDTR_BASE, vmcs12->guest_idtr_base);
2440 
2441 		vmx->segment_cache.bitmask = 0;
2442 	}
2443 
2444 	if (!hv_evmcs || !(hv_evmcs->hv_clean_fields &
2445 			   HV_VMX_ENLIGHTENED_CLEAN_FIELD_GUEST_GRP1)) {
2446 		vmcs_write32(GUEST_SYSENTER_CS, vmcs12->guest_sysenter_cs);
2447 		vmcs_writel(GUEST_PENDING_DBG_EXCEPTIONS,
2448 			    vmcs12->guest_pending_dbg_exceptions);
2449 		vmcs_writel(GUEST_SYSENTER_ESP, vmcs12->guest_sysenter_esp);
2450 		vmcs_writel(GUEST_SYSENTER_EIP, vmcs12->guest_sysenter_eip);
2451 
2452 		/*
2453 		 * L1 may access the L2's PDPTR, so save them to construct
2454 		 * vmcs12
2455 		 */
2456 		if (enable_ept) {
2457 			vmcs_write64(GUEST_PDPTR0, vmcs12->guest_pdptr0);
2458 			vmcs_write64(GUEST_PDPTR1, vmcs12->guest_pdptr1);
2459 			vmcs_write64(GUEST_PDPTR2, vmcs12->guest_pdptr2);
2460 			vmcs_write64(GUEST_PDPTR3, vmcs12->guest_pdptr3);
2461 		}
2462 
2463 		if (kvm_mpx_supported() && vmx->nested.nested_run_pending &&
2464 		    (vmcs12->vm_entry_controls & VM_ENTRY_LOAD_BNDCFGS))
2465 			vmcs_write64(GUEST_BNDCFGS, vmcs12->guest_bndcfgs);
2466 	}
2467 
2468 	if (nested_cpu_has_xsaves(vmcs12))
2469 		vmcs_write64(XSS_EXIT_BITMAP, vmcs12->xss_exit_bitmap);
2470 
2471 	/*
2472 	 * Whether page-faults are trapped is determined by a combination of
2473 	 * 3 settings: PFEC_MASK, PFEC_MATCH and EXCEPTION_BITMAP.PF.  If L0
2474 	 * doesn't care about page faults then we should set all of these to
2475 	 * L1's desires. However, if L0 does care about (some) page faults, it
2476 	 * is not easy (if at all possible?) to merge L0 and L1's desires, we
2477 	 * simply ask to exit on each and every L2 page fault. This is done by
2478 	 * setting MASK=MATCH=0 and (see below) EB.PF=1.
2479 	 * Note that below we don't need special code to set EB.PF beyond the
2480 	 * "or"ing of the EB of vmcs01 and vmcs12, because when enable_ept,
2481 	 * vmcs01's EB.PF is 0 so the "or" will take vmcs12's value, and when
2482 	 * !enable_ept, EB.PF is 1, so the "or" will always be 1.
2483 	 */
2484 	if (vmx_need_pf_intercept(&vmx->vcpu)) {
2485 		/*
2486 		 * TODO: if both L0 and L1 need the same MASK and MATCH,
2487 		 * go ahead and use it?
2488 		 */
2489 		vmcs_write32(PAGE_FAULT_ERROR_CODE_MASK, 0);
2490 		vmcs_write32(PAGE_FAULT_ERROR_CODE_MATCH, 0);
2491 	} else {
2492 		vmcs_write32(PAGE_FAULT_ERROR_CODE_MASK, vmcs12->page_fault_error_code_mask);
2493 		vmcs_write32(PAGE_FAULT_ERROR_CODE_MATCH, vmcs12->page_fault_error_code_match);
2494 	}
2495 
2496 	if (cpu_has_vmx_apicv()) {
2497 		vmcs_write64(EOI_EXIT_BITMAP0, vmcs12->eoi_exit_bitmap0);
2498 		vmcs_write64(EOI_EXIT_BITMAP1, vmcs12->eoi_exit_bitmap1);
2499 		vmcs_write64(EOI_EXIT_BITMAP2, vmcs12->eoi_exit_bitmap2);
2500 		vmcs_write64(EOI_EXIT_BITMAP3, vmcs12->eoi_exit_bitmap3);
2501 	}
2502 
2503 	/*
2504 	 * Make sure the msr_autostore list is up to date before we set the
2505 	 * count in the vmcs02.
2506 	 */
2507 	prepare_vmx_msr_autostore_list(&vmx->vcpu, MSR_IA32_TSC);
2508 
2509 	vmcs_write32(VM_EXIT_MSR_STORE_COUNT, vmx->msr_autostore.guest.nr);
2510 	vmcs_write32(VM_EXIT_MSR_LOAD_COUNT, vmx->msr_autoload.host.nr);
2511 	vmcs_write32(VM_ENTRY_MSR_LOAD_COUNT, vmx->msr_autoload.guest.nr);
2512 
2513 	set_cr4_guest_host_mask(vmx);
2514 }
2515 
2516 /*
2517  * prepare_vmcs02 is called when the L1 guest hypervisor runs its nested
2518  * L2 guest. L1 has a vmcs for L2 (vmcs12), and this function "merges" it
2519  * with L0's requirements for its guest (a.k.a. vmcs01), so we can run the L2
2520  * guest in a way that will both be appropriate to L1's requests, and our
2521  * needs. In addition to modifying the active vmcs (which is vmcs02), this
2522  * function also has additional necessary side-effects, like setting various
2523  * vcpu->arch fields.
2524  * Returns 0 on success, 1 on failure. Invalid state exit qualification code
2525  * is assigned to entry_failure_code on failure.
2526  */
2527 static int prepare_vmcs02(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12,
2528 			  bool from_vmentry,
2529 			  enum vm_entry_failure_code *entry_failure_code)
2530 {
2531 	struct vcpu_vmx *vmx = to_vmx(vcpu);
2532 	bool load_guest_pdptrs_vmcs12 = false;
2533 
2534 	if (vmx->nested.dirty_vmcs12 || evmptr_is_valid(vmx->nested.hv_evmcs_vmptr)) {
2535 		prepare_vmcs02_rare(vmx, vmcs12);
2536 		vmx->nested.dirty_vmcs12 = false;
2537 
2538 		load_guest_pdptrs_vmcs12 = !evmptr_is_valid(vmx->nested.hv_evmcs_vmptr) ||
2539 			!(vmx->nested.hv_evmcs->hv_clean_fields &
2540 			  HV_VMX_ENLIGHTENED_CLEAN_FIELD_GUEST_GRP1);
2541 	}
2542 
2543 	if (vmx->nested.nested_run_pending &&
2544 	    (vmcs12->vm_entry_controls & VM_ENTRY_LOAD_DEBUG_CONTROLS)) {
2545 		kvm_set_dr(vcpu, 7, vmcs12->guest_dr7);
2546 		vmcs_write64(GUEST_IA32_DEBUGCTL, vmcs12->guest_ia32_debugctl);
2547 	} else {
2548 		kvm_set_dr(vcpu, 7, vcpu->arch.dr7);
2549 		vmcs_write64(GUEST_IA32_DEBUGCTL, vmx->nested.pre_vmenter_debugctl);
2550 	}
2551 	if (kvm_mpx_supported() && (!vmx->nested.nested_run_pending ||
2552 	    !(vmcs12->vm_entry_controls & VM_ENTRY_LOAD_BNDCFGS)))
2553 		vmcs_write64(GUEST_BNDCFGS, vmx->nested.pre_vmenter_bndcfgs);
2554 	vmx_set_rflags(vcpu, vmcs12->guest_rflags);
2555 
2556 	/* EXCEPTION_BITMAP and CR0_GUEST_HOST_MASK should basically be the
2557 	 * bitwise-or of what L1 wants to trap for L2, and what we want to
2558 	 * trap. Note that CR0.TS also needs updating - we do this later.
2559 	 */
2560 	vmx_update_exception_bitmap(vcpu);
2561 	vcpu->arch.cr0_guest_owned_bits &= ~vmcs12->cr0_guest_host_mask;
2562 	vmcs_writel(CR0_GUEST_HOST_MASK, ~vcpu->arch.cr0_guest_owned_bits);
2563 
2564 	if (vmx->nested.nested_run_pending &&
2565 	    (vmcs12->vm_entry_controls & VM_ENTRY_LOAD_IA32_PAT)) {
2566 		vmcs_write64(GUEST_IA32_PAT, vmcs12->guest_ia32_pat);
2567 		vcpu->arch.pat = vmcs12->guest_ia32_pat;
2568 	} else if (vmcs_config.vmentry_ctrl & VM_ENTRY_LOAD_IA32_PAT) {
2569 		vmcs_write64(GUEST_IA32_PAT, vmx->vcpu.arch.pat);
2570 	}
2571 
2572 	vcpu->arch.tsc_offset = kvm_calc_nested_tsc_offset(
2573 			vcpu->arch.l1_tsc_offset,
2574 			vmx_get_l2_tsc_offset(vcpu),
2575 			vmx_get_l2_tsc_multiplier(vcpu));
2576 
2577 	vcpu->arch.tsc_scaling_ratio = kvm_calc_nested_tsc_multiplier(
2578 			vcpu->arch.l1_tsc_scaling_ratio,
2579 			vmx_get_l2_tsc_multiplier(vcpu));
2580 
2581 	vmcs_write64(TSC_OFFSET, vcpu->arch.tsc_offset);
2582 	if (kvm_caps.has_tsc_control)
2583 		vmcs_write64(TSC_MULTIPLIER, vcpu->arch.tsc_scaling_ratio);
2584 
2585 	nested_vmx_transition_tlb_flush(vcpu, vmcs12, true);
2586 
2587 	if (nested_cpu_has_ept(vmcs12))
2588 		nested_ept_init_mmu_context(vcpu);
2589 
2590 	/*
2591 	 * Override the CR0/CR4 read shadows after setting the effective guest
2592 	 * CR0/CR4.  The common helpers also set the shadows, but they don't
2593 	 * account for vmcs12's cr0/4_guest_host_mask.
2594 	 */
2595 	vmx_set_cr0(vcpu, vmcs12->guest_cr0);
2596 	vmcs_writel(CR0_READ_SHADOW, nested_read_cr0(vmcs12));
2597 
2598 	vmx_set_cr4(vcpu, vmcs12->guest_cr4);
2599 	vmcs_writel(CR4_READ_SHADOW, nested_read_cr4(vmcs12));
2600 
2601 	vcpu->arch.efer = nested_vmx_calc_efer(vmx, vmcs12);
2602 	/* Note: may modify VM_ENTRY/EXIT_CONTROLS and GUEST/HOST_IA32_EFER */
2603 	vmx_set_efer(vcpu, vcpu->arch.efer);
2604 
2605 	/*
2606 	 * Guest state is invalid and unrestricted guest is disabled,
2607 	 * which means L1 attempted VMEntry to L2 with invalid state.
2608 	 * Fail the VMEntry.
2609 	 *
2610 	 * However when force loading the guest state (SMM exit or
2611 	 * loading nested state after migration, it is possible to
2612 	 * have invalid guest state now, which will be later fixed by
2613 	 * restoring L2 register state
2614 	 */
2615 	if (CC(from_vmentry && !vmx_guest_state_valid(vcpu))) {
2616 		*entry_failure_code = ENTRY_FAIL_DEFAULT;
2617 		return -EINVAL;
2618 	}
2619 
2620 	/* Shadow page tables on either EPT or shadow page tables. */
2621 	if (nested_vmx_load_cr3(vcpu, vmcs12->guest_cr3, nested_cpu_has_ept(vmcs12),
2622 				from_vmentry, entry_failure_code))
2623 		return -EINVAL;
2624 
2625 	/*
2626 	 * Immediately write vmcs02.GUEST_CR3.  It will be propagated to vmcs12
2627 	 * on nested VM-Exit, which can occur without actually running L2 and
2628 	 * thus without hitting vmx_load_mmu_pgd(), e.g. if L1 is entering L2 with
2629 	 * vmcs12.GUEST_ACTIVITYSTATE=HLT, in which case KVM will intercept the
2630 	 * transition to HLT instead of running L2.
2631 	 */
2632 	if (enable_ept)
2633 		vmcs_writel(GUEST_CR3, vmcs12->guest_cr3);
2634 
2635 	/* Late preparation of GUEST_PDPTRs now that EFER and CRs are set. */
2636 	if (load_guest_pdptrs_vmcs12 && nested_cpu_has_ept(vmcs12) &&
2637 	    is_pae_paging(vcpu)) {
2638 		vmcs_write64(GUEST_PDPTR0, vmcs12->guest_pdptr0);
2639 		vmcs_write64(GUEST_PDPTR1, vmcs12->guest_pdptr1);
2640 		vmcs_write64(GUEST_PDPTR2, vmcs12->guest_pdptr2);
2641 		vmcs_write64(GUEST_PDPTR3, vmcs12->guest_pdptr3);
2642 	}
2643 
2644 	if ((vmcs12->vm_entry_controls & VM_ENTRY_LOAD_IA32_PERF_GLOBAL_CTRL) &&
2645 	    intel_pmu_has_perf_global_ctrl(vcpu_to_pmu(vcpu)) &&
2646 	    WARN_ON_ONCE(kvm_set_msr(vcpu, MSR_CORE_PERF_GLOBAL_CTRL,
2647 				     vmcs12->guest_ia32_perf_global_ctrl))) {
2648 		*entry_failure_code = ENTRY_FAIL_DEFAULT;
2649 		return -EINVAL;
2650 	}
2651 
2652 	kvm_rsp_write(vcpu, vmcs12->guest_rsp);
2653 	kvm_rip_write(vcpu, vmcs12->guest_rip);
2654 
2655 	/*
2656 	 * It was observed that genuine Hyper-V running in L1 doesn't reset
2657 	 * 'hv_clean_fields' by itself, it only sets the corresponding dirty
2658 	 * bits when it changes a field in eVMCS. Mark all fields as clean
2659 	 * here.
2660 	 */
2661 	if (evmptr_is_valid(vmx->nested.hv_evmcs_vmptr))
2662 		vmx->nested.hv_evmcs->hv_clean_fields |=
2663 			HV_VMX_ENLIGHTENED_CLEAN_FIELD_ALL;
2664 
2665 	return 0;
2666 }
2667 
2668 static int nested_vmx_check_nmi_controls(struct vmcs12 *vmcs12)
2669 {
2670 	if (CC(!nested_cpu_has_nmi_exiting(vmcs12) &&
2671 	       nested_cpu_has_virtual_nmis(vmcs12)))
2672 		return -EINVAL;
2673 
2674 	if (CC(!nested_cpu_has_virtual_nmis(vmcs12) &&
2675 	       nested_cpu_has(vmcs12, CPU_BASED_NMI_WINDOW_EXITING)))
2676 		return -EINVAL;
2677 
2678 	return 0;
2679 }
2680 
2681 static bool nested_vmx_check_eptp(struct kvm_vcpu *vcpu, u64 new_eptp)
2682 {
2683 	struct vcpu_vmx *vmx = to_vmx(vcpu);
2684 
2685 	/* Check for memory type validity */
2686 	switch (new_eptp & VMX_EPTP_MT_MASK) {
2687 	case VMX_EPTP_MT_UC:
2688 		if (CC(!(vmx->nested.msrs.ept_caps & VMX_EPTP_UC_BIT)))
2689 			return false;
2690 		break;
2691 	case VMX_EPTP_MT_WB:
2692 		if (CC(!(vmx->nested.msrs.ept_caps & VMX_EPTP_WB_BIT)))
2693 			return false;
2694 		break;
2695 	default:
2696 		return false;
2697 	}
2698 
2699 	/* Page-walk levels validity. */
2700 	switch (new_eptp & VMX_EPTP_PWL_MASK) {
2701 	case VMX_EPTP_PWL_5:
2702 		if (CC(!(vmx->nested.msrs.ept_caps & VMX_EPT_PAGE_WALK_5_BIT)))
2703 			return false;
2704 		break;
2705 	case VMX_EPTP_PWL_4:
2706 		if (CC(!(vmx->nested.msrs.ept_caps & VMX_EPT_PAGE_WALK_4_BIT)))
2707 			return false;
2708 		break;
2709 	default:
2710 		return false;
2711 	}
2712 
2713 	/* Reserved bits should not be set */
2714 	if (CC(kvm_vcpu_is_illegal_gpa(vcpu, new_eptp) || ((new_eptp >> 7) & 0x1f)))
2715 		return false;
2716 
2717 	/* AD, if set, should be supported */
2718 	if (new_eptp & VMX_EPTP_AD_ENABLE_BIT) {
2719 		if (CC(!(vmx->nested.msrs.ept_caps & VMX_EPT_AD_BIT)))
2720 			return false;
2721 	}
2722 
2723 	return true;
2724 }
2725 
2726 /*
2727  * Checks related to VM-Execution Control Fields
2728  */
2729 static int nested_check_vm_execution_controls(struct kvm_vcpu *vcpu,
2730                                               struct vmcs12 *vmcs12)
2731 {
2732 	struct vcpu_vmx *vmx = to_vmx(vcpu);
2733 
2734 	if (CC(!vmx_control_verify(vmcs12->pin_based_vm_exec_control,
2735 				   vmx->nested.msrs.pinbased_ctls_low,
2736 				   vmx->nested.msrs.pinbased_ctls_high)) ||
2737 	    CC(!vmx_control_verify(vmcs12->cpu_based_vm_exec_control,
2738 				   vmx->nested.msrs.procbased_ctls_low,
2739 				   vmx->nested.msrs.procbased_ctls_high)))
2740 		return -EINVAL;
2741 
2742 	if (nested_cpu_has(vmcs12, CPU_BASED_ACTIVATE_SECONDARY_CONTROLS) &&
2743 	    CC(!vmx_control_verify(vmcs12->secondary_vm_exec_control,
2744 				   vmx->nested.msrs.secondary_ctls_low,
2745 				   vmx->nested.msrs.secondary_ctls_high)))
2746 		return -EINVAL;
2747 
2748 	if (CC(vmcs12->cr3_target_count > nested_cpu_vmx_misc_cr3_count(vcpu)) ||
2749 	    nested_vmx_check_io_bitmap_controls(vcpu, vmcs12) ||
2750 	    nested_vmx_check_msr_bitmap_controls(vcpu, vmcs12) ||
2751 	    nested_vmx_check_tpr_shadow_controls(vcpu, vmcs12) ||
2752 	    nested_vmx_check_apic_access_controls(vcpu, vmcs12) ||
2753 	    nested_vmx_check_apicv_controls(vcpu, vmcs12) ||
2754 	    nested_vmx_check_nmi_controls(vmcs12) ||
2755 	    nested_vmx_check_pml_controls(vcpu, vmcs12) ||
2756 	    nested_vmx_check_unrestricted_guest_controls(vcpu, vmcs12) ||
2757 	    nested_vmx_check_mode_based_ept_exec_controls(vcpu, vmcs12) ||
2758 	    nested_vmx_check_shadow_vmcs_controls(vcpu, vmcs12) ||
2759 	    CC(nested_cpu_has_vpid(vmcs12) && !vmcs12->virtual_processor_id))
2760 		return -EINVAL;
2761 
2762 	if (!nested_cpu_has_preemption_timer(vmcs12) &&
2763 	    nested_cpu_has_save_preemption_timer(vmcs12))
2764 		return -EINVAL;
2765 
2766 	if (nested_cpu_has_ept(vmcs12) &&
2767 	    CC(!nested_vmx_check_eptp(vcpu, vmcs12->ept_pointer)))
2768 		return -EINVAL;
2769 
2770 	if (nested_cpu_has_vmfunc(vmcs12)) {
2771 		if (CC(vmcs12->vm_function_control &
2772 		       ~vmx->nested.msrs.vmfunc_controls))
2773 			return -EINVAL;
2774 
2775 		if (nested_cpu_has_eptp_switching(vmcs12)) {
2776 			if (CC(!nested_cpu_has_ept(vmcs12)) ||
2777 			    CC(!page_address_valid(vcpu, vmcs12->eptp_list_address)))
2778 				return -EINVAL;
2779 		}
2780 	}
2781 
2782 	return 0;
2783 }
2784 
2785 /*
2786  * Checks related to VM-Exit Control Fields
2787  */
2788 static int nested_check_vm_exit_controls(struct kvm_vcpu *vcpu,
2789                                          struct vmcs12 *vmcs12)
2790 {
2791 	struct vcpu_vmx *vmx = to_vmx(vcpu);
2792 
2793 	if (CC(!vmx_control_verify(vmcs12->vm_exit_controls,
2794 				    vmx->nested.msrs.exit_ctls_low,
2795 				    vmx->nested.msrs.exit_ctls_high)) ||
2796 	    CC(nested_vmx_check_exit_msr_switch_controls(vcpu, vmcs12)))
2797 		return -EINVAL;
2798 
2799 	return 0;
2800 }
2801 
2802 /*
2803  * Checks related to VM-Entry Control Fields
2804  */
2805 static int nested_check_vm_entry_controls(struct kvm_vcpu *vcpu,
2806 					  struct vmcs12 *vmcs12)
2807 {
2808 	struct vcpu_vmx *vmx = to_vmx(vcpu);
2809 
2810 	if (CC(!vmx_control_verify(vmcs12->vm_entry_controls,
2811 				    vmx->nested.msrs.entry_ctls_low,
2812 				    vmx->nested.msrs.entry_ctls_high)))
2813 		return -EINVAL;
2814 
2815 	/*
2816 	 * From the Intel SDM, volume 3:
2817 	 * Fields relevant to VM-entry event injection must be set properly.
2818 	 * These fields are the VM-entry interruption-information field, the
2819 	 * VM-entry exception error code, and the VM-entry instruction length.
2820 	 */
2821 	if (vmcs12->vm_entry_intr_info_field & INTR_INFO_VALID_MASK) {
2822 		u32 intr_info = vmcs12->vm_entry_intr_info_field;
2823 		u8 vector = intr_info & INTR_INFO_VECTOR_MASK;
2824 		u32 intr_type = intr_info & INTR_INFO_INTR_TYPE_MASK;
2825 		bool has_error_code = intr_info & INTR_INFO_DELIVER_CODE_MASK;
2826 		bool should_have_error_code;
2827 		bool urg = nested_cpu_has2(vmcs12,
2828 					   SECONDARY_EXEC_UNRESTRICTED_GUEST);
2829 		bool prot_mode = !urg || vmcs12->guest_cr0 & X86_CR0_PE;
2830 
2831 		/* VM-entry interruption-info field: interruption type */
2832 		if (CC(intr_type == INTR_TYPE_RESERVED) ||
2833 		    CC(intr_type == INTR_TYPE_OTHER_EVENT &&
2834 		       !nested_cpu_supports_monitor_trap_flag(vcpu)))
2835 			return -EINVAL;
2836 
2837 		/* VM-entry interruption-info field: vector */
2838 		if (CC(intr_type == INTR_TYPE_NMI_INTR && vector != NMI_VECTOR) ||
2839 		    CC(intr_type == INTR_TYPE_HARD_EXCEPTION && vector > 31) ||
2840 		    CC(intr_type == INTR_TYPE_OTHER_EVENT && vector != 0))
2841 			return -EINVAL;
2842 
2843 		/* VM-entry interruption-info field: deliver error code */
2844 		should_have_error_code =
2845 			intr_type == INTR_TYPE_HARD_EXCEPTION && prot_mode &&
2846 			x86_exception_has_error_code(vector);
2847 		if (CC(has_error_code != should_have_error_code))
2848 			return -EINVAL;
2849 
2850 		/* VM-entry exception error code */
2851 		if (CC(has_error_code &&
2852 		       vmcs12->vm_entry_exception_error_code & GENMASK(31, 16)))
2853 			return -EINVAL;
2854 
2855 		/* VM-entry interruption-info field: reserved bits */
2856 		if (CC(intr_info & INTR_INFO_RESVD_BITS_MASK))
2857 			return -EINVAL;
2858 
2859 		/* VM-entry instruction length */
2860 		switch (intr_type) {
2861 		case INTR_TYPE_SOFT_EXCEPTION:
2862 		case INTR_TYPE_SOFT_INTR:
2863 		case INTR_TYPE_PRIV_SW_EXCEPTION:
2864 			if (CC(vmcs12->vm_entry_instruction_len > 15) ||
2865 			    CC(vmcs12->vm_entry_instruction_len == 0 &&
2866 			    CC(!nested_cpu_has_zero_length_injection(vcpu))))
2867 				return -EINVAL;
2868 		}
2869 	}
2870 
2871 	if (nested_vmx_check_entry_msr_switch_controls(vcpu, vmcs12))
2872 		return -EINVAL;
2873 
2874 	return 0;
2875 }
2876 
2877 static int nested_vmx_check_controls(struct kvm_vcpu *vcpu,
2878 				     struct vmcs12 *vmcs12)
2879 {
2880 	if (nested_check_vm_execution_controls(vcpu, vmcs12) ||
2881 	    nested_check_vm_exit_controls(vcpu, vmcs12) ||
2882 	    nested_check_vm_entry_controls(vcpu, vmcs12))
2883 		return -EINVAL;
2884 
2885 	if (guest_cpuid_has_evmcs(vcpu))
2886 		return nested_evmcs_check_controls(vmcs12);
2887 
2888 	return 0;
2889 }
2890 
2891 static int nested_vmx_check_address_space_size(struct kvm_vcpu *vcpu,
2892 				       struct vmcs12 *vmcs12)
2893 {
2894 #ifdef CONFIG_X86_64
2895 	if (CC(!!(vmcs12->vm_exit_controls & VM_EXIT_HOST_ADDR_SPACE_SIZE) !=
2896 		!!(vcpu->arch.efer & EFER_LMA)))
2897 		return -EINVAL;
2898 #endif
2899 	return 0;
2900 }
2901 
2902 static int nested_vmx_check_host_state(struct kvm_vcpu *vcpu,
2903 				       struct vmcs12 *vmcs12)
2904 {
2905 	bool ia32e;
2906 
2907 	if (CC(!nested_host_cr0_valid(vcpu, vmcs12->host_cr0)) ||
2908 	    CC(!nested_host_cr4_valid(vcpu, vmcs12->host_cr4)) ||
2909 	    CC(kvm_vcpu_is_illegal_gpa(vcpu, vmcs12->host_cr3)))
2910 		return -EINVAL;
2911 
2912 	if (CC(is_noncanonical_address(vmcs12->host_ia32_sysenter_esp, vcpu)) ||
2913 	    CC(is_noncanonical_address(vmcs12->host_ia32_sysenter_eip, vcpu)))
2914 		return -EINVAL;
2915 
2916 	if ((vmcs12->vm_exit_controls & VM_EXIT_LOAD_IA32_PAT) &&
2917 	    CC(!kvm_pat_valid(vmcs12->host_ia32_pat)))
2918 		return -EINVAL;
2919 
2920 	if ((vmcs12->vm_exit_controls & VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL) &&
2921 	    CC(!kvm_valid_perf_global_ctrl(vcpu_to_pmu(vcpu),
2922 					   vmcs12->host_ia32_perf_global_ctrl)))
2923 		return -EINVAL;
2924 
2925 #ifdef CONFIG_X86_64
2926 	ia32e = !!(vmcs12->vm_exit_controls & VM_EXIT_HOST_ADDR_SPACE_SIZE);
2927 #else
2928 	ia32e = false;
2929 #endif
2930 
2931 	if (ia32e) {
2932 		if (CC(!(vmcs12->host_cr4 & X86_CR4_PAE)))
2933 			return -EINVAL;
2934 	} else {
2935 		if (CC(vmcs12->vm_entry_controls & VM_ENTRY_IA32E_MODE) ||
2936 		    CC(vmcs12->host_cr4 & X86_CR4_PCIDE) ||
2937 		    CC((vmcs12->host_rip) >> 32))
2938 			return -EINVAL;
2939 	}
2940 
2941 	if (CC(vmcs12->host_cs_selector & (SEGMENT_RPL_MASK | SEGMENT_TI_MASK)) ||
2942 	    CC(vmcs12->host_ss_selector & (SEGMENT_RPL_MASK | SEGMENT_TI_MASK)) ||
2943 	    CC(vmcs12->host_ds_selector & (SEGMENT_RPL_MASK | SEGMENT_TI_MASK)) ||
2944 	    CC(vmcs12->host_es_selector & (SEGMENT_RPL_MASK | SEGMENT_TI_MASK)) ||
2945 	    CC(vmcs12->host_fs_selector & (SEGMENT_RPL_MASK | SEGMENT_TI_MASK)) ||
2946 	    CC(vmcs12->host_gs_selector & (SEGMENT_RPL_MASK | SEGMENT_TI_MASK)) ||
2947 	    CC(vmcs12->host_tr_selector & (SEGMENT_RPL_MASK | SEGMENT_TI_MASK)) ||
2948 	    CC(vmcs12->host_cs_selector == 0) ||
2949 	    CC(vmcs12->host_tr_selector == 0) ||
2950 	    CC(vmcs12->host_ss_selector == 0 && !ia32e))
2951 		return -EINVAL;
2952 
2953 	if (CC(is_noncanonical_address(vmcs12->host_fs_base, vcpu)) ||
2954 	    CC(is_noncanonical_address(vmcs12->host_gs_base, vcpu)) ||
2955 	    CC(is_noncanonical_address(vmcs12->host_gdtr_base, vcpu)) ||
2956 	    CC(is_noncanonical_address(vmcs12->host_idtr_base, vcpu)) ||
2957 	    CC(is_noncanonical_address(vmcs12->host_tr_base, vcpu)) ||
2958 	    CC(is_noncanonical_address(vmcs12->host_rip, vcpu)))
2959 		return -EINVAL;
2960 
2961 	/*
2962 	 * If the load IA32_EFER VM-exit control is 1, bits reserved in the
2963 	 * IA32_EFER MSR must be 0 in the field for that register. In addition,
2964 	 * the values of the LMA and LME bits in the field must each be that of
2965 	 * the host address-space size VM-exit control.
2966 	 */
2967 	if (vmcs12->vm_exit_controls & VM_EXIT_LOAD_IA32_EFER) {
2968 		if (CC(!kvm_valid_efer(vcpu, vmcs12->host_ia32_efer)) ||
2969 		    CC(ia32e != !!(vmcs12->host_ia32_efer & EFER_LMA)) ||
2970 		    CC(ia32e != !!(vmcs12->host_ia32_efer & EFER_LME)))
2971 			return -EINVAL;
2972 	}
2973 
2974 	return 0;
2975 }
2976 
2977 static int nested_vmx_check_vmcs_link_ptr(struct kvm_vcpu *vcpu,
2978 					  struct vmcs12 *vmcs12)
2979 {
2980 	struct vcpu_vmx *vmx = to_vmx(vcpu);
2981 	struct gfn_to_hva_cache *ghc = &vmx->nested.shadow_vmcs12_cache;
2982 	struct vmcs_hdr hdr;
2983 
2984 	if (vmcs12->vmcs_link_pointer == INVALID_GPA)
2985 		return 0;
2986 
2987 	if (CC(!page_address_valid(vcpu, vmcs12->vmcs_link_pointer)))
2988 		return -EINVAL;
2989 
2990 	if (ghc->gpa != vmcs12->vmcs_link_pointer &&
2991 	    CC(kvm_gfn_to_hva_cache_init(vcpu->kvm, ghc,
2992 					 vmcs12->vmcs_link_pointer, VMCS12_SIZE)))
2993                 return -EINVAL;
2994 
2995 	if (CC(kvm_read_guest_offset_cached(vcpu->kvm, ghc, &hdr,
2996 					    offsetof(struct vmcs12, hdr),
2997 					    sizeof(hdr))))
2998 		return -EINVAL;
2999 
3000 	if (CC(hdr.revision_id != VMCS12_REVISION) ||
3001 	    CC(hdr.shadow_vmcs != nested_cpu_has_shadow_vmcs(vmcs12)))
3002 		return -EINVAL;
3003 
3004 	return 0;
3005 }
3006 
3007 /*
3008  * Checks related to Guest Non-register State
3009  */
3010 static int nested_check_guest_non_reg_state(struct vmcs12 *vmcs12)
3011 {
3012 	if (CC(vmcs12->guest_activity_state != GUEST_ACTIVITY_ACTIVE &&
3013 	       vmcs12->guest_activity_state != GUEST_ACTIVITY_HLT &&
3014 	       vmcs12->guest_activity_state != GUEST_ACTIVITY_WAIT_SIPI))
3015 		return -EINVAL;
3016 
3017 	return 0;
3018 }
3019 
3020 static int nested_vmx_check_guest_state(struct kvm_vcpu *vcpu,
3021 					struct vmcs12 *vmcs12,
3022 					enum vm_entry_failure_code *entry_failure_code)
3023 {
3024 	bool ia32e;
3025 
3026 	*entry_failure_code = ENTRY_FAIL_DEFAULT;
3027 
3028 	if (CC(!nested_guest_cr0_valid(vcpu, vmcs12->guest_cr0)) ||
3029 	    CC(!nested_guest_cr4_valid(vcpu, vmcs12->guest_cr4)))
3030 		return -EINVAL;
3031 
3032 	if ((vmcs12->vm_entry_controls & VM_ENTRY_LOAD_DEBUG_CONTROLS) &&
3033 	    CC(!kvm_dr7_valid(vmcs12->guest_dr7)))
3034 		return -EINVAL;
3035 
3036 	if ((vmcs12->vm_entry_controls & VM_ENTRY_LOAD_IA32_PAT) &&
3037 	    CC(!kvm_pat_valid(vmcs12->guest_ia32_pat)))
3038 		return -EINVAL;
3039 
3040 	if (nested_vmx_check_vmcs_link_ptr(vcpu, vmcs12)) {
3041 		*entry_failure_code = ENTRY_FAIL_VMCS_LINK_PTR;
3042 		return -EINVAL;
3043 	}
3044 
3045 	if ((vmcs12->vm_entry_controls & VM_ENTRY_LOAD_IA32_PERF_GLOBAL_CTRL) &&
3046 	    CC(!kvm_valid_perf_global_ctrl(vcpu_to_pmu(vcpu),
3047 					   vmcs12->guest_ia32_perf_global_ctrl)))
3048 		return -EINVAL;
3049 
3050 	/*
3051 	 * If the load IA32_EFER VM-entry control is 1, the following checks
3052 	 * are performed on the field for the IA32_EFER MSR:
3053 	 * - Bits reserved in the IA32_EFER MSR must be 0.
3054 	 * - Bit 10 (corresponding to IA32_EFER.LMA) must equal the value of
3055 	 *   the IA-32e mode guest VM-exit control. It must also be identical
3056 	 *   to bit 8 (LME) if bit 31 in the CR0 field (corresponding to
3057 	 *   CR0.PG) is 1.
3058 	 */
3059 	if (to_vmx(vcpu)->nested.nested_run_pending &&
3060 	    (vmcs12->vm_entry_controls & VM_ENTRY_LOAD_IA32_EFER)) {
3061 		ia32e = (vmcs12->vm_entry_controls & VM_ENTRY_IA32E_MODE) != 0;
3062 		if (CC(!kvm_valid_efer(vcpu, vmcs12->guest_ia32_efer)) ||
3063 		    CC(ia32e != !!(vmcs12->guest_ia32_efer & EFER_LMA)) ||
3064 		    CC(((vmcs12->guest_cr0 & X86_CR0_PG) &&
3065 		     ia32e != !!(vmcs12->guest_ia32_efer & EFER_LME))))
3066 			return -EINVAL;
3067 	}
3068 
3069 	if ((vmcs12->vm_entry_controls & VM_ENTRY_LOAD_BNDCFGS) &&
3070 	    (CC(is_noncanonical_address(vmcs12->guest_bndcfgs & PAGE_MASK, vcpu)) ||
3071 	     CC((vmcs12->guest_bndcfgs & MSR_IA32_BNDCFGS_RSVD))))
3072 		return -EINVAL;
3073 
3074 	if (nested_check_guest_non_reg_state(vmcs12))
3075 		return -EINVAL;
3076 
3077 	return 0;
3078 }
3079 
3080 static int nested_vmx_check_vmentry_hw(struct kvm_vcpu *vcpu)
3081 {
3082 	struct vcpu_vmx *vmx = to_vmx(vcpu);
3083 	unsigned long cr3, cr4;
3084 	bool vm_fail;
3085 
3086 	if (!nested_early_check)
3087 		return 0;
3088 
3089 	if (vmx->msr_autoload.host.nr)
3090 		vmcs_write32(VM_EXIT_MSR_LOAD_COUNT, 0);
3091 	if (vmx->msr_autoload.guest.nr)
3092 		vmcs_write32(VM_ENTRY_MSR_LOAD_COUNT, 0);
3093 
3094 	preempt_disable();
3095 
3096 	vmx_prepare_switch_to_guest(vcpu);
3097 
3098 	/*
3099 	 * Induce a consistency check VMExit by clearing bit 1 in GUEST_RFLAGS,
3100 	 * which is reserved to '1' by hardware.  GUEST_RFLAGS is guaranteed to
3101 	 * be written (by prepare_vmcs02()) before the "real" VMEnter, i.e.
3102 	 * there is no need to preserve other bits or save/restore the field.
3103 	 */
3104 	vmcs_writel(GUEST_RFLAGS, 0);
3105 
3106 	cr3 = __get_current_cr3_fast();
3107 	if (unlikely(cr3 != vmx->loaded_vmcs->host_state.cr3)) {
3108 		vmcs_writel(HOST_CR3, cr3);
3109 		vmx->loaded_vmcs->host_state.cr3 = cr3;
3110 	}
3111 
3112 	cr4 = cr4_read_shadow();
3113 	if (unlikely(cr4 != vmx->loaded_vmcs->host_state.cr4)) {
3114 		vmcs_writel(HOST_CR4, cr4);
3115 		vmx->loaded_vmcs->host_state.cr4 = cr4;
3116 	}
3117 
3118 	vm_fail = __vmx_vcpu_run(vmx, (unsigned long *)&vcpu->arch.regs,
3119 				 __vmx_vcpu_run_flags(vmx));
3120 
3121 	if (vmx->msr_autoload.host.nr)
3122 		vmcs_write32(VM_EXIT_MSR_LOAD_COUNT, vmx->msr_autoload.host.nr);
3123 	if (vmx->msr_autoload.guest.nr)
3124 		vmcs_write32(VM_ENTRY_MSR_LOAD_COUNT, vmx->msr_autoload.guest.nr);
3125 
3126 	if (vm_fail) {
3127 		u32 error = vmcs_read32(VM_INSTRUCTION_ERROR);
3128 
3129 		preempt_enable();
3130 
3131 		trace_kvm_nested_vmenter_failed(
3132 			"early hardware check VM-instruction error: ", error);
3133 		WARN_ON_ONCE(error != VMXERR_ENTRY_INVALID_CONTROL_FIELD);
3134 		return 1;
3135 	}
3136 
3137 	/*
3138 	 * VMExit clears RFLAGS.IF and DR7, even on a consistency check.
3139 	 */
3140 	if (hw_breakpoint_active())
3141 		set_debugreg(__this_cpu_read(cpu_dr7), 7);
3142 	local_irq_enable();
3143 	preempt_enable();
3144 
3145 	/*
3146 	 * A non-failing VMEntry means we somehow entered guest mode with
3147 	 * an illegal RIP, and that's just the tip of the iceberg.  There
3148 	 * is no telling what memory has been modified or what state has
3149 	 * been exposed to unknown code.  Hitting this all but guarantees
3150 	 * a (very critical) hardware issue.
3151 	 */
3152 	WARN_ON(!(vmcs_read32(VM_EXIT_REASON) &
3153 		VMX_EXIT_REASONS_FAILED_VMENTRY));
3154 
3155 	return 0;
3156 }
3157 
3158 static bool nested_get_evmcs_page(struct kvm_vcpu *vcpu)
3159 {
3160 	struct vcpu_vmx *vmx = to_vmx(vcpu);
3161 
3162 	/*
3163 	 * hv_evmcs may end up being not mapped after migration (when
3164 	 * L2 was running), map it here to make sure vmcs12 changes are
3165 	 * properly reflected.
3166 	 */
3167 	if (guest_cpuid_has_evmcs(vcpu) &&
3168 	    vmx->nested.hv_evmcs_vmptr == EVMPTR_MAP_PENDING) {
3169 		enum nested_evmptrld_status evmptrld_status =
3170 			nested_vmx_handle_enlightened_vmptrld(vcpu, false);
3171 
3172 		if (evmptrld_status == EVMPTRLD_VMFAIL ||
3173 		    evmptrld_status == EVMPTRLD_ERROR)
3174 			return false;
3175 
3176 		/*
3177 		 * Post migration VMCS12 always provides the most actual
3178 		 * information, copy it to eVMCS upon entry.
3179 		 */
3180 		vmx->nested.need_vmcs12_to_shadow_sync = true;
3181 	}
3182 
3183 	return true;
3184 }
3185 
3186 static bool nested_get_vmcs12_pages(struct kvm_vcpu *vcpu)
3187 {
3188 	struct vmcs12 *vmcs12 = get_vmcs12(vcpu);
3189 	struct vcpu_vmx *vmx = to_vmx(vcpu);
3190 	struct kvm_host_map *map;
3191 
3192 	if (!vcpu->arch.pdptrs_from_userspace &&
3193 	    !nested_cpu_has_ept(vmcs12) && is_pae_paging(vcpu)) {
3194 		/*
3195 		 * Reload the guest's PDPTRs since after a migration
3196 		 * the guest CR3 might be restored prior to setting the nested
3197 		 * state which can lead to a load of wrong PDPTRs.
3198 		 */
3199 		if (CC(!load_pdptrs(vcpu, vcpu->arch.cr3)))
3200 			return false;
3201 	}
3202 
3203 
3204 	if (nested_cpu_has2(vmcs12, SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES)) {
3205 		map = &vmx->nested.apic_access_page_map;
3206 
3207 		if (!kvm_vcpu_map(vcpu, gpa_to_gfn(vmcs12->apic_access_addr), map)) {
3208 			vmcs_write64(APIC_ACCESS_ADDR, pfn_to_hpa(map->pfn));
3209 		} else {
3210 			pr_debug_ratelimited("%s: no backing for APIC-access address in vmcs12\n",
3211 					     __func__);
3212 			vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
3213 			vcpu->run->internal.suberror =
3214 				KVM_INTERNAL_ERROR_EMULATION;
3215 			vcpu->run->internal.ndata = 0;
3216 			return false;
3217 		}
3218 	}
3219 
3220 	if (nested_cpu_has(vmcs12, CPU_BASED_TPR_SHADOW)) {
3221 		map = &vmx->nested.virtual_apic_map;
3222 
3223 		if (!kvm_vcpu_map(vcpu, gpa_to_gfn(vmcs12->virtual_apic_page_addr), map)) {
3224 			vmcs_write64(VIRTUAL_APIC_PAGE_ADDR, pfn_to_hpa(map->pfn));
3225 		} else if (nested_cpu_has(vmcs12, CPU_BASED_CR8_LOAD_EXITING) &&
3226 		           nested_cpu_has(vmcs12, CPU_BASED_CR8_STORE_EXITING) &&
3227 			   !nested_cpu_has2(vmcs12, SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES)) {
3228 			/*
3229 			 * The processor will never use the TPR shadow, simply
3230 			 * clear the bit from the execution control.  Such a
3231 			 * configuration is useless, but it happens in tests.
3232 			 * For any other configuration, failing the vm entry is
3233 			 * _not_ what the processor does but it's basically the
3234 			 * only possibility we have.
3235 			 */
3236 			exec_controls_clearbit(vmx, CPU_BASED_TPR_SHADOW);
3237 		} else {
3238 			/*
3239 			 * Write an illegal value to VIRTUAL_APIC_PAGE_ADDR to
3240 			 * force VM-Entry to fail.
3241 			 */
3242 			vmcs_write64(VIRTUAL_APIC_PAGE_ADDR, INVALID_GPA);
3243 		}
3244 	}
3245 
3246 	if (nested_cpu_has_posted_intr(vmcs12)) {
3247 		map = &vmx->nested.pi_desc_map;
3248 
3249 		if (!kvm_vcpu_map(vcpu, gpa_to_gfn(vmcs12->posted_intr_desc_addr), map)) {
3250 			vmx->nested.pi_desc =
3251 				(struct pi_desc *)(((void *)map->hva) +
3252 				offset_in_page(vmcs12->posted_intr_desc_addr));
3253 			vmcs_write64(POSTED_INTR_DESC_ADDR,
3254 				     pfn_to_hpa(map->pfn) + offset_in_page(vmcs12->posted_intr_desc_addr));
3255 		} else {
3256 			/*
3257 			 * Defer the KVM_INTERNAL_EXIT until KVM tries to
3258 			 * access the contents of the VMCS12 posted interrupt
3259 			 * descriptor. (Note that KVM may do this when it
3260 			 * should not, per the architectural specification.)
3261 			 */
3262 			vmx->nested.pi_desc = NULL;
3263 			pin_controls_clearbit(vmx, PIN_BASED_POSTED_INTR);
3264 		}
3265 	}
3266 	if (nested_vmx_prepare_msr_bitmap(vcpu, vmcs12))
3267 		exec_controls_setbit(vmx, CPU_BASED_USE_MSR_BITMAPS);
3268 	else
3269 		exec_controls_clearbit(vmx, CPU_BASED_USE_MSR_BITMAPS);
3270 
3271 	return true;
3272 }
3273 
3274 static bool vmx_get_nested_state_pages(struct kvm_vcpu *vcpu)
3275 {
3276 	/*
3277 	 * Note: nested_get_evmcs_page() also updates 'vp_assist_page' copy
3278 	 * in 'struct kvm_vcpu_hv' in case eVMCS is in use, this is mandatory
3279 	 * to make nested_evmcs_l2_tlb_flush_enabled() work correctly post
3280 	 * migration.
3281 	 */
3282 	if (!nested_get_evmcs_page(vcpu)) {
3283 		pr_debug_ratelimited("%s: enlightened vmptrld failed\n",
3284 				     __func__);
3285 		vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
3286 		vcpu->run->internal.suberror =
3287 			KVM_INTERNAL_ERROR_EMULATION;
3288 		vcpu->run->internal.ndata = 0;
3289 
3290 		return false;
3291 	}
3292 
3293 	if (is_guest_mode(vcpu) && !nested_get_vmcs12_pages(vcpu))
3294 		return false;
3295 
3296 	return true;
3297 }
3298 
3299 static int nested_vmx_write_pml_buffer(struct kvm_vcpu *vcpu, gpa_t gpa)
3300 {
3301 	struct vmcs12 *vmcs12;
3302 	struct vcpu_vmx *vmx = to_vmx(vcpu);
3303 	gpa_t dst;
3304 
3305 	if (WARN_ON_ONCE(!is_guest_mode(vcpu)))
3306 		return 0;
3307 
3308 	if (WARN_ON_ONCE(vmx->nested.pml_full))
3309 		return 1;
3310 
3311 	/*
3312 	 * Check if PML is enabled for the nested guest. Whether eptp bit 6 is
3313 	 * set is already checked as part of A/D emulation.
3314 	 */
3315 	vmcs12 = get_vmcs12(vcpu);
3316 	if (!nested_cpu_has_pml(vmcs12))
3317 		return 0;
3318 
3319 	if (vmcs12->guest_pml_index >= PML_ENTITY_NUM) {
3320 		vmx->nested.pml_full = true;
3321 		return 1;
3322 	}
3323 
3324 	gpa &= ~0xFFFull;
3325 	dst = vmcs12->pml_address + sizeof(u64) * vmcs12->guest_pml_index;
3326 
3327 	if (kvm_write_guest_page(vcpu->kvm, gpa_to_gfn(dst), &gpa,
3328 				 offset_in_page(dst), sizeof(gpa)))
3329 		return 0;
3330 
3331 	vmcs12->guest_pml_index--;
3332 
3333 	return 0;
3334 }
3335 
3336 /*
3337  * Intel's VMX Instruction Reference specifies a common set of prerequisites
3338  * for running VMX instructions (except VMXON, whose prerequisites are
3339  * slightly different). It also specifies what exception to inject otherwise.
3340  * Note that many of these exceptions have priority over VM exits, so they
3341  * don't have to be checked again here.
3342  */
3343 static int nested_vmx_check_permission(struct kvm_vcpu *vcpu)
3344 {
3345 	if (!to_vmx(vcpu)->nested.vmxon) {
3346 		kvm_queue_exception(vcpu, UD_VECTOR);
3347 		return 0;
3348 	}
3349 
3350 	if (vmx_get_cpl(vcpu)) {
3351 		kvm_inject_gp(vcpu, 0);
3352 		return 0;
3353 	}
3354 
3355 	return 1;
3356 }
3357 
3358 static u8 vmx_has_apicv_interrupt(struct kvm_vcpu *vcpu)
3359 {
3360 	u8 rvi = vmx_get_rvi();
3361 	u8 vppr = kvm_lapic_get_reg(vcpu->arch.apic, APIC_PROCPRI);
3362 
3363 	return ((rvi & 0xf0) > (vppr & 0xf0));
3364 }
3365 
3366 static void load_vmcs12_host_state(struct kvm_vcpu *vcpu,
3367 				   struct vmcs12 *vmcs12);
3368 
3369 /*
3370  * If from_vmentry is false, this is being called from state restore (either RSM
3371  * or KVM_SET_NESTED_STATE).  Otherwise it's called from vmlaunch/vmresume.
3372  *
3373  * Returns:
3374  *	NVMX_VMENTRY_SUCCESS: Entered VMX non-root mode
3375  *	NVMX_VMENTRY_VMFAIL:  Consistency check VMFail
3376  *	NVMX_VMENTRY_VMEXIT:  Consistency check VMExit
3377  *	NVMX_VMENTRY_KVM_INTERNAL_ERROR: KVM internal error
3378  */
3379 enum nvmx_vmentry_status nested_vmx_enter_non_root_mode(struct kvm_vcpu *vcpu,
3380 							bool from_vmentry)
3381 {
3382 	struct vcpu_vmx *vmx = to_vmx(vcpu);
3383 	struct vmcs12 *vmcs12 = get_vmcs12(vcpu);
3384 	enum vm_entry_failure_code entry_failure_code;
3385 	bool evaluate_pending_interrupts;
3386 	union vmx_exit_reason exit_reason = {
3387 		.basic = EXIT_REASON_INVALID_STATE,
3388 		.failed_vmentry = 1,
3389 	};
3390 	u32 failed_index;
3391 
3392 	trace_kvm_nested_vmenter(kvm_rip_read(vcpu),
3393 				 vmx->nested.current_vmptr,
3394 				 vmcs12->guest_rip,
3395 				 vmcs12->guest_intr_status,
3396 				 vmcs12->vm_entry_intr_info_field,
3397 				 vmcs12->secondary_vm_exec_control & SECONDARY_EXEC_ENABLE_EPT,
3398 				 vmcs12->ept_pointer,
3399 				 vmcs12->guest_cr3,
3400 				 KVM_ISA_VMX);
3401 
3402 	kvm_service_local_tlb_flush_requests(vcpu);
3403 
3404 	evaluate_pending_interrupts = exec_controls_get(vmx) &
3405 		(CPU_BASED_INTR_WINDOW_EXITING | CPU_BASED_NMI_WINDOW_EXITING);
3406 	if (likely(!evaluate_pending_interrupts) && kvm_vcpu_apicv_active(vcpu))
3407 		evaluate_pending_interrupts |= vmx_has_apicv_interrupt(vcpu);
3408 	if (!evaluate_pending_interrupts)
3409 		evaluate_pending_interrupts |= kvm_apic_has_pending_init_or_sipi(vcpu);
3410 
3411 	if (!vmx->nested.nested_run_pending ||
3412 	    !(vmcs12->vm_entry_controls & VM_ENTRY_LOAD_DEBUG_CONTROLS))
3413 		vmx->nested.pre_vmenter_debugctl = vmcs_read64(GUEST_IA32_DEBUGCTL);
3414 	if (kvm_mpx_supported() &&
3415 	    (!vmx->nested.nested_run_pending ||
3416 	     !(vmcs12->vm_entry_controls & VM_ENTRY_LOAD_BNDCFGS)))
3417 		vmx->nested.pre_vmenter_bndcfgs = vmcs_read64(GUEST_BNDCFGS);
3418 
3419 	/*
3420 	 * Overwrite vmcs01.GUEST_CR3 with L1's CR3 if EPT is disabled *and*
3421 	 * nested early checks are disabled.  In the event of a "late" VM-Fail,
3422 	 * i.e. a VM-Fail detected by hardware but not KVM, KVM must unwind its
3423 	 * software model to the pre-VMEntry host state.  When EPT is disabled,
3424 	 * GUEST_CR3 holds KVM's shadow CR3, not L1's "real" CR3, which causes
3425 	 * nested_vmx_restore_host_state() to corrupt vcpu->arch.cr3.  Stuffing
3426 	 * vmcs01.GUEST_CR3 results in the unwind naturally setting arch.cr3 to
3427 	 * the correct value.  Smashing vmcs01.GUEST_CR3 is safe because nested
3428 	 * VM-Exits, and the unwind, reset KVM's MMU, i.e. vmcs01.GUEST_CR3 is
3429 	 * guaranteed to be overwritten with a shadow CR3 prior to re-entering
3430 	 * L1.  Don't stuff vmcs01.GUEST_CR3 when using nested early checks as
3431 	 * KVM modifies vcpu->arch.cr3 if and only if the early hardware checks
3432 	 * pass, and early VM-Fails do not reset KVM's MMU, i.e. the VM-Fail
3433 	 * path would need to manually save/restore vmcs01.GUEST_CR3.
3434 	 */
3435 	if (!enable_ept && !nested_early_check)
3436 		vmcs_writel(GUEST_CR3, vcpu->arch.cr3);
3437 
3438 	vmx_switch_vmcs(vcpu, &vmx->nested.vmcs02);
3439 
3440 	prepare_vmcs02_early(vmx, &vmx->vmcs01, vmcs12);
3441 
3442 	if (from_vmentry) {
3443 		if (unlikely(!nested_get_vmcs12_pages(vcpu))) {
3444 			vmx_switch_vmcs(vcpu, &vmx->vmcs01);
3445 			return NVMX_VMENTRY_KVM_INTERNAL_ERROR;
3446 		}
3447 
3448 		if (nested_vmx_check_vmentry_hw(vcpu)) {
3449 			vmx_switch_vmcs(vcpu, &vmx->vmcs01);
3450 			return NVMX_VMENTRY_VMFAIL;
3451 		}
3452 
3453 		if (nested_vmx_check_guest_state(vcpu, vmcs12,
3454 						 &entry_failure_code)) {
3455 			exit_reason.basic = EXIT_REASON_INVALID_STATE;
3456 			vmcs12->exit_qualification = entry_failure_code;
3457 			goto vmentry_fail_vmexit;
3458 		}
3459 	}
3460 
3461 	enter_guest_mode(vcpu);
3462 
3463 	if (prepare_vmcs02(vcpu, vmcs12, from_vmentry, &entry_failure_code)) {
3464 		exit_reason.basic = EXIT_REASON_INVALID_STATE;
3465 		vmcs12->exit_qualification = entry_failure_code;
3466 		goto vmentry_fail_vmexit_guest_mode;
3467 	}
3468 
3469 	if (from_vmentry) {
3470 		failed_index = nested_vmx_load_msr(vcpu,
3471 						   vmcs12->vm_entry_msr_load_addr,
3472 						   vmcs12->vm_entry_msr_load_count);
3473 		if (failed_index) {
3474 			exit_reason.basic = EXIT_REASON_MSR_LOAD_FAIL;
3475 			vmcs12->exit_qualification = failed_index;
3476 			goto vmentry_fail_vmexit_guest_mode;
3477 		}
3478 	} else {
3479 		/*
3480 		 * The MMU is not initialized to point at the right entities yet and
3481 		 * "get pages" would need to read data from the guest (i.e. we will
3482 		 * need to perform gpa to hpa translation). Request a call
3483 		 * to nested_get_vmcs12_pages before the next VM-entry.  The MSRs
3484 		 * have already been set at vmentry time and should not be reset.
3485 		 */
3486 		kvm_make_request(KVM_REQ_GET_NESTED_STATE_PAGES, vcpu);
3487 	}
3488 
3489 	/*
3490 	 * Re-evaluate pending events if L1 had a pending IRQ/NMI/INIT/SIPI
3491 	 * when it executed VMLAUNCH/VMRESUME, as entering non-root mode can
3492 	 * effectively unblock various events, e.g. INIT/SIPI cause VM-Exit
3493 	 * unconditionally.
3494 	 */
3495 	if (unlikely(evaluate_pending_interrupts))
3496 		kvm_make_request(KVM_REQ_EVENT, vcpu);
3497 
3498 	/*
3499 	 * Do not start the preemption timer hrtimer until after we know
3500 	 * we are successful, so that only nested_vmx_vmexit needs to cancel
3501 	 * the timer.
3502 	 */
3503 	vmx->nested.preemption_timer_expired = false;
3504 	if (nested_cpu_has_preemption_timer(vmcs12)) {
3505 		u64 timer_value = vmx_calc_preemption_timer_value(vcpu);
3506 		vmx_start_preemption_timer(vcpu, timer_value);
3507 	}
3508 
3509 	/*
3510 	 * Note no nested_vmx_succeed or nested_vmx_fail here. At this point
3511 	 * we are no longer running L1, and VMLAUNCH/VMRESUME has not yet
3512 	 * returned as far as L1 is concerned. It will only return (and set
3513 	 * the success flag) when L2 exits (see nested_vmx_vmexit()).
3514 	 */
3515 	return NVMX_VMENTRY_SUCCESS;
3516 
3517 	/*
3518 	 * A failed consistency check that leads to a VMExit during L1's
3519 	 * VMEnter to L2 is a variation of a normal VMexit, as explained in
3520 	 * 26.7 "VM-entry failures during or after loading guest state".
3521 	 */
3522 vmentry_fail_vmexit_guest_mode:
3523 	if (vmcs12->cpu_based_vm_exec_control & CPU_BASED_USE_TSC_OFFSETTING)
3524 		vcpu->arch.tsc_offset -= vmcs12->tsc_offset;
3525 	leave_guest_mode(vcpu);
3526 
3527 vmentry_fail_vmexit:
3528 	vmx_switch_vmcs(vcpu, &vmx->vmcs01);
3529 
3530 	if (!from_vmentry)
3531 		return NVMX_VMENTRY_VMEXIT;
3532 
3533 	load_vmcs12_host_state(vcpu, vmcs12);
3534 	vmcs12->vm_exit_reason = exit_reason.full;
3535 	if (enable_shadow_vmcs || evmptr_is_valid(vmx->nested.hv_evmcs_vmptr))
3536 		vmx->nested.need_vmcs12_to_shadow_sync = true;
3537 	return NVMX_VMENTRY_VMEXIT;
3538 }
3539 
3540 /*
3541  * nested_vmx_run() handles a nested entry, i.e., a VMLAUNCH or VMRESUME on L1
3542  * for running an L2 nested guest.
3543  */
3544 static int nested_vmx_run(struct kvm_vcpu *vcpu, bool launch)
3545 {
3546 	struct vmcs12 *vmcs12;
3547 	enum nvmx_vmentry_status status;
3548 	struct vcpu_vmx *vmx = to_vmx(vcpu);
3549 	u32 interrupt_shadow = vmx_get_interrupt_shadow(vcpu);
3550 	enum nested_evmptrld_status evmptrld_status;
3551 
3552 	if (!nested_vmx_check_permission(vcpu))
3553 		return 1;
3554 
3555 	evmptrld_status = nested_vmx_handle_enlightened_vmptrld(vcpu, launch);
3556 	if (evmptrld_status == EVMPTRLD_ERROR) {
3557 		kvm_queue_exception(vcpu, UD_VECTOR);
3558 		return 1;
3559 	}
3560 
3561 	kvm_pmu_trigger_event(vcpu, PERF_COUNT_HW_BRANCH_INSTRUCTIONS);
3562 
3563 	if (CC(evmptrld_status == EVMPTRLD_VMFAIL))
3564 		return nested_vmx_failInvalid(vcpu);
3565 
3566 	if (CC(!evmptr_is_valid(vmx->nested.hv_evmcs_vmptr) &&
3567 	       vmx->nested.current_vmptr == INVALID_GPA))
3568 		return nested_vmx_failInvalid(vcpu);
3569 
3570 	vmcs12 = get_vmcs12(vcpu);
3571 
3572 	/*
3573 	 * Can't VMLAUNCH or VMRESUME a shadow VMCS. Despite the fact
3574 	 * that there *is* a valid VMCS pointer, RFLAGS.CF is set
3575 	 * rather than RFLAGS.ZF, and no error number is stored to the
3576 	 * VM-instruction error field.
3577 	 */
3578 	if (CC(vmcs12->hdr.shadow_vmcs))
3579 		return nested_vmx_failInvalid(vcpu);
3580 
3581 	if (evmptr_is_valid(vmx->nested.hv_evmcs_vmptr)) {
3582 		copy_enlightened_to_vmcs12(vmx, vmx->nested.hv_evmcs->hv_clean_fields);
3583 		/* Enlightened VMCS doesn't have launch state */
3584 		vmcs12->launch_state = !launch;
3585 	} else if (enable_shadow_vmcs) {
3586 		copy_shadow_to_vmcs12(vmx);
3587 	}
3588 
3589 	/*
3590 	 * The nested entry process starts with enforcing various prerequisites
3591 	 * on vmcs12 as required by the Intel SDM, and act appropriately when
3592 	 * they fail: As the SDM explains, some conditions should cause the
3593 	 * instruction to fail, while others will cause the instruction to seem
3594 	 * to succeed, but return an EXIT_REASON_INVALID_STATE.
3595 	 * To speed up the normal (success) code path, we should avoid checking
3596 	 * for misconfigurations which will anyway be caught by the processor
3597 	 * when using the merged vmcs02.
3598 	 */
3599 	if (CC(interrupt_shadow & KVM_X86_SHADOW_INT_MOV_SS))
3600 		return nested_vmx_fail(vcpu, VMXERR_ENTRY_EVENTS_BLOCKED_BY_MOV_SS);
3601 
3602 	if (CC(vmcs12->launch_state == launch))
3603 		return nested_vmx_fail(vcpu,
3604 			launch ? VMXERR_VMLAUNCH_NONCLEAR_VMCS
3605 			       : VMXERR_VMRESUME_NONLAUNCHED_VMCS);
3606 
3607 	if (nested_vmx_check_controls(vcpu, vmcs12))
3608 		return nested_vmx_fail(vcpu, VMXERR_ENTRY_INVALID_CONTROL_FIELD);
3609 
3610 	if (nested_vmx_check_address_space_size(vcpu, vmcs12))
3611 		return nested_vmx_fail(vcpu, VMXERR_ENTRY_INVALID_HOST_STATE_FIELD);
3612 
3613 	if (nested_vmx_check_host_state(vcpu, vmcs12))
3614 		return nested_vmx_fail(vcpu, VMXERR_ENTRY_INVALID_HOST_STATE_FIELD);
3615 
3616 	/*
3617 	 * We're finally done with prerequisite checking, and can start with
3618 	 * the nested entry.
3619 	 */
3620 	vmx->nested.nested_run_pending = 1;
3621 	vmx->nested.has_preemption_timer_deadline = false;
3622 	status = nested_vmx_enter_non_root_mode(vcpu, true);
3623 	if (unlikely(status != NVMX_VMENTRY_SUCCESS))
3624 		goto vmentry_failed;
3625 
3626 	/* Emulate processing of posted interrupts on VM-Enter. */
3627 	if (nested_cpu_has_posted_intr(vmcs12) &&
3628 	    kvm_apic_has_interrupt(vcpu) == vmx->nested.posted_intr_nv) {
3629 		vmx->nested.pi_pending = true;
3630 		kvm_make_request(KVM_REQ_EVENT, vcpu);
3631 		kvm_apic_clear_irr(vcpu, vmx->nested.posted_intr_nv);
3632 	}
3633 
3634 	/* Hide L1D cache contents from the nested guest.  */
3635 	vmx->vcpu.arch.l1tf_flush_l1d = true;
3636 
3637 	/*
3638 	 * Must happen outside of nested_vmx_enter_non_root_mode() as it will
3639 	 * also be used as part of restoring nVMX state for
3640 	 * snapshot restore (migration).
3641 	 *
3642 	 * In this flow, it is assumed that vmcs12 cache was
3643 	 * transferred as part of captured nVMX state and should
3644 	 * therefore not be read from guest memory (which may not
3645 	 * exist on destination host yet).
3646 	 */
3647 	nested_cache_shadow_vmcs12(vcpu, vmcs12);
3648 
3649 	switch (vmcs12->guest_activity_state) {
3650 	case GUEST_ACTIVITY_HLT:
3651 		/*
3652 		 * If we're entering a halted L2 vcpu and the L2 vcpu won't be
3653 		 * awakened by event injection or by an NMI-window VM-exit or
3654 		 * by an interrupt-window VM-exit, halt the vcpu.
3655 		 */
3656 		if (!(vmcs12->vm_entry_intr_info_field & INTR_INFO_VALID_MASK) &&
3657 		    !nested_cpu_has(vmcs12, CPU_BASED_NMI_WINDOW_EXITING) &&
3658 		    !(nested_cpu_has(vmcs12, CPU_BASED_INTR_WINDOW_EXITING) &&
3659 		      (vmcs12->guest_rflags & X86_EFLAGS_IF))) {
3660 			vmx->nested.nested_run_pending = 0;
3661 			return kvm_emulate_halt_noskip(vcpu);
3662 		}
3663 		break;
3664 	case GUEST_ACTIVITY_WAIT_SIPI:
3665 		vmx->nested.nested_run_pending = 0;
3666 		vcpu->arch.mp_state = KVM_MP_STATE_INIT_RECEIVED;
3667 		break;
3668 	default:
3669 		break;
3670 	}
3671 
3672 	return 1;
3673 
3674 vmentry_failed:
3675 	vmx->nested.nested_run_pending = 0;
3676 	if (status == NVMX_VMENTRY_KVM_INTERNAL_ERROR)
3677 		return 0;
3678 	if (status == NVMX_VMENTRY_VMEXIT)
3679 		return 1;
3680 	WARN_ON_ONCE(status != NVMX_VMENTRY_VMFAIL);
3681 	return nested_vmx_fail(vcpu, VMXERR_ENTRY_INVALID_CONTROL_FIELD);
3682 }
3683 
3684 /*
3685  * On a nested exit from L2 to L1, vmcs12.guest_cr0 might not be up-to-date
3686  * because L2 may have changed some cr0 bits directly (CR0_GUEST_HOST_MASK).
3687  * This function returns the new value we should put in vmcs12.guest_cr0.
3688  * It's not enough to just return the vmcs02 GUEST_CR0. Rather,
3689  *  1. Bits that neither L0 nor L1 trapped, were set directly by L2 and are now
3690  *     available in vmcs02 GUEST_CR0. (Note: It's enough to check that L0
3691  *     didn't trap the bit, because if L1 did, so would L0).
3692  *  2. Bits that L1 asked to trap (and therefore L0 also did) could not have
3693  *     been modified by L2, and L1 knows it. So just leave the old value of
3694  *     the bit from vmcs12.guest_cr0. Note that the bit from vmcs02 GUEST_CR0
3695  *     isn't relevant, because if L0 traps this bit it can set it to anything.
3696  *  3. Bits that L1 didn't trap, but L0 did. L1 believes the guest could have
3697  *     changed these bits, and therefore they need to be updated, but L0
3698  *     didn't necessarily allow them to be changed in GUEST_CR0 - and rather
3699  *     put them in vmcs02 CR0_READ_SHADOW. So take these bits from there.
3700  */
3701 static inline unsigned long
3702 vmcs12_guest_cr0(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12)
3703 {
3704 	return
3705 	/*1*/	(vmcs_readl(GUEST_CR0) & vcpu->arch.cr0_guest_owned_bits) |
3706 	/*2*/	(vmcs12->guest_cr0 & vmcs12->cr0_guest_host_mask) |
3707 	/*3*/	(vmcs_readl(CR0_READ_SHADOW) & ~(vmcs12->cr0_guest_host_mask |
3708 			vcpu->arch.cr0_guest_owned_bits));
3709 }
3710 
3711 static inline unsigned long
3712 vmcs12_guest_cr4(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12)
3713 {
3714 	return
3715 	/*1*/	(vmcs_readl(GUEST_CR4) & vcpu->arch.cr4_guest_owned_bits) |
3716 	/*2*/	(vmcs12->guest_cr4 & vmcs12->cr4_guest_host_mask) |
3717 	/*3*/	(vmcs_readl(CR4_READ_SHADOW) & ~(vmcs12->cr4_guest_host_mask |
3718 			vcpu->arch.cr4_guest_owned_bits));
3719 }
3720 
3721 static void vmcs12_save_pending_event(struct kvm_vcpu *vcpu,
3722 				      struct vmcs12 *vmcs12,
3723 				      u32 vm_exit_reason, u32 exit_intr_info)
3724 {
3725 	u32 idt_vectoring;
3726 	unsigned int nr;
3727 
3728 	/*
3729 	 * Per the SDM, VM-Exits due to double and triple faults are never
3730 	 * considered to occur during event delivery, even if the double/triple
3731 	 * fault is the result of an escalating vectoring issue.
3732 	 *
3733 	 * Note, the SDM qualifies the double fault behavior with "The original
3734 	 * event results in a double-fault exception".  It's unclear why the
3735 	 * qualification exists since exits due to double fault can occur only
3736 	 * while vectoring a different exception (injected events are never
3737 	 * subject to interception), i.e. there's _always_ an original event.
3738 	 *
3739 	 * The SDM also uses NMI as a confusing example for the "original event
3740 	 * causes the VM exit directly" clause.  NMI isn't special in any way,
3741 	 * the same rule applies to all events that cause an exit directly.
3742 	 * NMI is an odd choice for the example because NMIs can only occur on
3743 	 * instruction boundaries, i.e. they _can't_ occur during vectoring.
3744 	 */
3745 	if ((u16)vm_exit_reason == EXIT_REASON_TRIPLE_FAULT ||
3746 	    ((u16)vm_exit_reason == EXIT_REASON_EXCEPTION_NMI &&
3747 	     is_double_fault(exit_intr_info))) {
3748 		vmcs12->idt_vectoring_info_field = 0;
3749 	} else if (vcpu->arch.exception.injected) {
3750 		nr = vcpu->arch.exception.vector;
3751 		idt_vectoring = nr | VECTORING_INFO_VALID_MASK;
3752 
3753 		if (kvm_exception_is_soft(nr)) {
3754 			vmcs12->vm_exit_instruction_len =
3755 				vcpu->arch.event_exit_inst_len;
3756 			idt_vectoring |= INTR_TYPE_SOFT_EXCEPTION;
3757 		} else
3758 			idt_vectoring |= INTR_TYPE_HARD_EXCEPTION;
3759 
3760 		if (vcpu->arch.exception.has_error_code) {
3761 			idt_vectoring |= VECTORING_INFO_DELIVER_CODE_MASK;
3762 			vmcs12->idt_vectoring_error_code =
3763 				vcpu->arch.exception.error_code;
3764 		}
3765 
3766 		vmcs12->idt_vectoring_info_field = idt_vectoring;
3767 	} else if (vcpu->arch.nmi_injected) {
3768 		vmcs12->idt_vectoring_info_field =
3769 			INTR_TYPE_NMI_INTR | INTR_INFO_VALID_MASK | NMI_VECTOR;
3770 	} else if (vcpu->arch.interrupt.injected) {
3771 		nr = vcpu->arch.interrupt.nr;
3772 		idt_vectoring = nr | VECTORING_INFO_VALID_MASK;
3773 
3774 		if (vcpu->arch.interrupt.soft) {
3775 			idt_vectoring |= INTR_TYPE_SOFT_INTR;
3776 			vmcs12->vm_entry_instruction_len =
3777 				vcpu->arch.event_exit_inst_len;
3778 		} else
3779 			idt_vectoring |= INTR_TYPE_EXT_INTR;
3780 
3781 		vmcs12->idt_vectoring_info_field = idt_vectoring;
3782 	} else {
3783 		vmcs12->idt_vectoring_info_field = 0;
3784 	}
3785 }
3786 
3787 
3788 void nested_mark_vmcs12_pages_dirty(struct kvm_vcpu *vcpu)
3789 {
3790 	struct vmcs12 *vmcs12 = get_vmcs12(vcpu);
3791 	gfn_t gfn;
3792 
3793 	/*
3794 	 * Don't need to mark the APIC access page dirty; it is never
3795 	 * written to by the CPU during APIC virtualization.
3796 	 */
3797 
3798 	if (nested_cpu_has(vmcs12, CPU_BASED_TPR_SHADOW)) {
3799 		gfn = vmcs12->virtual_apic_page_addr >> PAGE_SHIFT;
3800 		kvm_vcpu_mark_page_dirty(vcpu, gfn);
3801 	}
3802 
3803 	if (nested_cpu_has_posted_intr(vmcs12)) {
3804 		gfn = vmcs12->posted_intr_desc_addr >> PAGE_SHIFT;
3805 		kvm_vcpu_mark_page_dirty(vcpu, gfn);
3806 	}
3807 }
3808 
3809 static int vmx_complete_nested_posted_interrupt(struct kvm_vcpu *vcpu)
3810 {
3811 	struct vcpu_vmx *vmx = to_vmx(vcpu);
3812 	int max_irr;
3813 	void *vapic_page;
3814 	u16 status;
3815 
3816 	if (!vmx->nested.pi_pending)
3817 		return 0;
3818 
3819 	if (!vmx->nested.pi_desc)
3820 		goto mmio_needed;
3821 
3822 	vmx->nested.pi_pending = false;
3823 
3824 	if (!pi_test_and_clear_on(vmx->nested.pi_desc))
3825 		return 0;
3826 
3827 	max_irr = find_last_bit((unsigned long *)vmx->nested.pi_desc->pir, 256);
3828 	if (max_irr != 256) {
3829 		vapic_page = vmx->nested.virtual_apic_map.hva;
3830 		if (!vapic_page)
3831 			goto mmio_needed;
3832 
3833 		__kvm_apic_update_irr(vmx->nested.pi_desc->pir,
3834 			vapic_page, &max_irr);
3835 		status = vmcs_read16(GUEST_INTR_STATUS);
3836 		if ((u8)max_irr > ((u8)status & 0xff)) {
3837 			status &= ~0xff;
3838 			status |= (u8)max_irr;
3839 			vmcs_write16(GUEST_INTR_STATUS, status);
3840 		}
3841 	}
3842 
3843 	nested_mark_vmcs12_pages_dirty(vcpu);
3844 	return 0;
3845 
3846 mmio_needed:
3847 	kvm_handle_memory_failure(vcpu, X86EMUL_IO_NEEDED, NULL);
3848 	return -ENXIO;
3849 }
3850 
3851 static void nested_vmx_inject_exception_vmexit(struct kvm_vcpu *vcpu)
3852 {
3853 	struct kvm_queued_exception *ex = &vcpu->arch.exception_vmexit;
3854 	u32 intr_info = ex->vector | INTR_INFO_VALID_MASK;
3855 	struct vmcs12 *vmcs12 = get_vmcs12(vcpu);
3856 	unsigned long exit_qual;
3857 
3858 	if (ex->has_payload) {
3859 		exit_qual = ex->payload;
3860 	} else if (ex->vector == PF_VECTOR) {
3861 		exit_qual = vcpu->arch.cr2;
3862 	} else if (ex->vector == DB_VECTOR) {
3863 		exit_qual = vcpu->arch.dr6;
3864 		exit_qual &= ~DR6_BT;
3865 		exit_qual ^= DR6_ACTIVE_LOW;
3866 	} else {
3867 		exit_qual = 0;
3868 	}
3869 
3870 	if (ex->has_error_code) {
3871 		/*
3872 		 * Intel CPUs do not generate error codes with bits 31:16 set,
3873 		 * and more importantly VMX disallows setting bits 31:16 in the
3874 		 * injected error code for VM-Entry.  Drop the bits to mimic
3875 		 * hardware and avoid inducing failure on nested VM-Entry if L1
3876 		 * chooses to inject the exception back to L2.  AMD CPUs _do_
3877 		 * generate "full" 32-bit error codes, so KVM allows userspace
3878 		 * to inject exception error codes with bits 31:16 set.
3879 		 */
3880 		vmcs12->vm_exit_intr_error_code = (u16)ex->error_code;
3881 		intr_info |= INTR_INFO_DELIVER_CODE_MASK;
3882 	}
3883 
3884 	if (kvm_exception_is_soft(ex->vector))
3885 		intr_info |= INTR_TYPE_SOFT_EXCEPTION;
3886 	else
3887 		intr_info |= INTR_TYPE_HARD_EXCEPTION;
3888 
3889 	if (!(vmcs12->idt_vectoring_info_field & VECTORING_INFO_VALID_MASK) &&
3890 	    vmx_get_nmi_mask(vcpu))
3891 		intr_info |= INTR_INFO_UNBLOCK_NMI;
3892 
3893 	nested_vmx_vmexit(vcpu, EXIT_REASON_EXCEPTION_NMI, intr_info, exit_qual);
3894 }
3895 
3896 /*
3897  * Returns true if a debug trap is (likely) pending delivery.  Infer the class
3898  * of a #DB (trap-like vs. fault-like) from the exception payload (to-be-DR6).
3899  * Using the payload is flawed because code breakpoints (fault-like) and data
3900  * breakpoints (trap-like) set the same bits in DR6 (breakpoint detected), i.e.
3901  * this will return false positives if a to-be-injected code breakpoint #DB is
3902  * pending (from KVM's perspective, but not "pending" across an instruction
3903  * boundary).  ICEBP, a.k.a. INT1, is also not reflected here even though it
3904  * too is trap-like.
3905  *
3906  * KVM "works" despite these flaws as ICEBP isn't currently supported by the
3907  * emulator, Monitor Trap Flag is not marked pending on intercepted #DBs (the
3908  * #DB has already happened), and MTF isn't marked pending on code breakpoints
3909  * from the emulator (because such #DBs are fault-like and thus don't trigger
3910  * actions that fire on instruction retire).
3911  */
3912 static unsigned long vmx_get_pending_dbg_trap(struct kvm_queued_exception *ex)
3913 {
3914 	if (!ex->pending || ex->vector != DB_VECTOR)
3915 		return 0;
3916 
3917 	/* General Detect #DBs are always fault-like. */
3918 	return ex->payload & ~DR6_BD;
3919 }
3920 
3921 /*
3922  * Returns true if there's a pending #DB exception that is lower priority than
3923  * a pending Monitor Trap Flag VM-Exit.  TSS T-flag #DBs are not emulated by
3924  * KVM, but could theoretically be injected by userspace.  Note, this code is
3925  * imperfect, see above.
3926  */
3927 static bool vmx_is_low_priority_db_trap(struct kvm_queued_exception *ex)
3928 {
3929 	return vmx_get_pending_dbg_trap(ex) & ~DR6_BT;
3930 }
3931 
3932 /*
3933  * Certain VM-exits set the 'pending debug exceptions' field to indicate a
3934  * recognized #DB (data or single-step) that has yet to be delivered. Since KVM
3935  * represents these debug traps with a payload that is said to be compatible
3936  * with the 'pending debug exceptions' field, write the payload to the VMCS
3937  * field if a VM-exit is delivered before the debug trap.
3938  */
3939 static void nested_vmx_update_pending_dbg(struct kvm_vcpu *vcpu)
3940 {
3941 	unsigned long pending_dbg;
3942 
3943 	pending_dbg = vmx_get_pending_dbg_trap(&vcpu->arch.exception);
3944 	if (pending_dbg)
3945 		vmcs_writel(GUEST_PENDING_DBG_EXCEPTIONS, pending_dbg);
3946 }
3947 
3948 static bool nested_vmx_preemption_timer_pending(struct kvm_vcpu *vcpu)
3949 {
3950 	return nested_cpu_has_preemption_timer(get_vmcs12(vcpu)) &&
3951 	       to_vmx(vcpu)->nested.preemption_timer_expired;
3952 }
3953 
3954 static bool vmx_has_nested_events(struct kvm_vcpu *vcpu)
3955 {
3956 	return nested_vmx_preemption_timer_pending(vcpu) ||
3957 	       to_vmx(vcpu)->nested.mtf_pending;
3958 }
3959 
3960 /*
3961  * Per the Intel SDM's table "Priority Among Concurrent Events", with minor
3962  * edits to fill in missing examples, e.g. #DB due to split-lock accesses,
3963  * and less minor edits to splice in the priority of VMX Non-Root specific
3964  * events, e.g. MTF and NMI/INTR-window exiting.
3965  *
3966  * 1 Hardware Reset and Machine Checks
3967  *	- RESET
3968  *	- Machine Check
3969  *
3970  * 2 Trap on Task Switch
3971  *	- T flag in TSS is set (on task switch)
3972  *
3973  * 3 External Hardware Interventions
3974  *	- FLUSH
3975  *	- STOPCLK
3976  *	- SMI
3977  *	- INIT
3978  *
3979  * 3.5 Monitor Trap Flag (MTF) VM-exit[1]
3980  *
3981  * 4 Traps on Previous Instruction
3982  *	- Breakpoints
3983  *	- Trap-class Debug Exceptions (#DB due to TF flag set, data/I-O
3984  *	  breakpoint, or #DB due to a split-lock access)
3985  *
3986  * 4.3	VMX-preemption timer expired VM-exit
3987  *
3988  * 4.6	NMI-window exiting VM-exit[2]
3989  *
3990  * 5 Nonmaskable Interrupts (NMI)
3991  *
3992  * 5.5 Interrupt-window exiting VM-exit and Virtual-interrupt delivery
3993  *
3994  * 6 Maskable Hardware Interrupts
3995  *
3996  * 7 Code Breakpoint Fault
3997  *
3998  * 8 Faults from Fetching Next Instruction
3999  *	- Code-Segment Limit Violation
4000  *	- Code Page Fault
4001  *	- Control protection exception (missing ENDBRANCH at target of indirect
4002  *					call or jump)
4003  *
4004  * 9 Faults from Decoding Next Instruction
4005  *	- Instruction length > 15 bytes
4006  *	- Invalid Opcode
4007  *	- Coprocessor Not Available
4008  *
4009  *10 Faults on Executing Instruction
4010  *	- Overflow
4011  *	- Bound error
4012  *	- Invalid TSS
4013  *	- Segment Not Present
4014  *	- Stack fault
4015  *	- General Protection
4016  *	- Data Page Fault
4017  *	- Alignment Check
4018  *	- x86 FPU Floating-point exception
4019  *	- SIMD floating-point exception
4020  *	- Virtualization exception
4021  *	- Control protection exception
4022  *
4023  * [1] Per the "Monitor Trap Flag" section: System-management interrupts (SMIs),
4024  *     INIT signals, and higher priority events take priority over MTF VM exits.
4025  *     MTF VM exits take priority over debug-trap exceptions and lower priority
4026  *     events.
4027  *
4028  * [2] Debug-trap exceptions and higher priority events take priority over VM exits
4029  *     caused by the VMX-preemption timer.  VM exits caused by the VMX-preemption
4030  *     timer take priority over VM exits caused by the "NMI-window exiting"
4031  *     VM-execution control and lower priority events.
4032  *
4033  * [3] Debug-trap exceptions and higher priority events take priority over VM exits
4034  *     caused by "NMI-window exiting".  VM exits caused by this control take
4035  *     priority over non-maskable interrupts (NMIs) and lower priority events.
4036  *
4037  * [4] Virtual-interrupt delivery has the same priority as that of VM exits due to
4038  *     the 1-setting of the "interrupt-window exiting" VM-execution control.  Thus,
4039  *     non-maskable interrupts (NMIs) and higher priority events take priority over
4040  *     delivery of a virtual interrupt; delivery of a virtual interrupt takes
4041  *     priority over external interrupts and lower priority events.
4042  */
4043 static int vmx_check_nested_events(struct kvm_vcpu *vcpu)
4044 {
4045 	struct kvm_lapic *apic = vcpu->arch.apic;
4046 	struct vcpu_vmx *vmx = to_vmx(vcpu);
4047 	/*
4048 	 * Only a pending nested run blocks a pending exception.  If there is a
4049 	 * previously injected event, the pending exception occurred while said
4050 	 * event was being delivered and thus needs to be handled.
4051 	 */
4052 	bool block_nested_exceptions = vmx->nested.nested_run_pending;
4053 	/*
4054 	 * New events (not exceptions) are only recognized at instruction
4055 	 * boundaries.  If an event needs reinjection, then KVM is handling a
4056 	 * VM-Exit that occurred _during_ instruction execution; new events are
4057 	 * blocked until the instruction completes.
4058 	 */
4059 	bool block_nested_events = block_nested_exceptions ||
4060 				   kvm_event_needs_reinjection(vcpu);
4061 
4062 	if (lapic_in_kernel(vcpu) &&
4063 		test_bit(KVM_APIC_INIT, &apic->pending_events)) {
4064 		if (block_nested_events)
4065 			return -EBUSY;
4066 		nested_vmx_update_pending_dbg(vcpu);
4067 		clear_bit(KVM_APIC_INIT, &apic->pending_events);
4068 		if (vcpu->arch.mp_state != KVM_MP_STATE_INIT_RECEIVED)
4069 			nested_vmx_vmexit(vcpu, EXIT_REASON_INIT_SIGNAL, 0, 0);
4070 
4071 		/* MTF is discarded if the vCPU is in WFS. */
4072 		vmx->nested.mtf_pending = false;
4073 		return 0;
4074 	}
4075 
4076 	if (lapic_in_kernel(vcpu) &&
4077 	    test_bit(KVM_APIC_SIPI, &apic->pending_events)) {
4078 		if (block_nested_events)
4079 			return -EBUSY;
4080 
4081 		clear_bit(KVM_APIC_SIPI, &apic->pending_events);
4082 		if (vcpu->arch.mp_state == KVM_MP_STATE_INIT_RECEIVED) {
4083 			nested_vmx_vmexit(vcpu, EXIT_REASON_SIPI_SIGNAL, 0,
4084 						apic->sipi_vector & 0xFFUL);
4085 			return 0;
4086 		}
4087 		/* Fallthrough, the SIPI is completely ignored. */
4088 	}
4089 
4090 	/*
4091 	 * Process exceptions that are higher priority than Monitor Trap Flag:
4092 	 * fault-like exceptions, TSS T flag #DB (not emulated by KVM, but
4093 	 * could theoretically come in from userspace), and ICEBP (INT1).
4094 	 *
4095 	 * TODO: SMIs have higher priority than MTF and trap-like #DBs (except
4096 	 * for TSS T flag #DBs).  KVM also doesn't save/restore pending MTF
4097 	 * across SMI/RSM as it should; that needs to be addressed in order to
4098 	 * prioritize SMI over MTF and trap-like #DBs.
4099 	 */
4100 	if (vcpu->arch.exception_vmexit.pending &&
4101 	    !vmx_is_low_priority_db_trap(&vcpu->arch.exception_vmexit)) {
4102 		if (block_nested_exceptions)
4103 			return -EBUSY;
4104 
4105 		nested_vmx_inject_exception_vmexit(vcpu);
4106 		return 0;
4107 	}
4108 
4109 	if (vcpu->arch.exception.pending &&
4110 	    !vmx_is_low_priority_db_trap(&vcpu->arch.exception)) {
4111 		if (block_nested_exceptions)
4112 			return -EBUSY;
4113 		goto no_vmexit;
4114 	}
4115 
4116 	if (vmx->nested.mtf_pending) {
4117 		if (block_nested_events)
4118 			return -EBUSY;
4119 		nested_vmx_update_pending_dbg(vcpu);
4120 		nested_vmx_vmexit(vcpu, EXIT_REASON_MONITOR_TRAP_FLAG, 0, 0);
4121 		return 0;
4122 	}
4123 
4124 	if (vcpu->arch.exception_vmexit.pending) {
4125 		if (block_nested_exceptions)
4126 			return -EBUSY;
4127 
4128 		nested_vmx_inject_exception_vmexit(vcpu);
4129 		return 0;
4130 	}
4131 
4132 	if (vcpu->arch.exception.pending) {
4133 		if (block_nested_exceptions)
4134 			return -EBUSY;
4135 		goto no_vmexit;
4136 	}
4137 
4138 	if (nested_vmx_preemption_timer_pending(vcpu)) {
4139 		if (block_nested_events)
4140 			return -EBUSY;
4141 		nested_vmx_vmexit(vcpu, EXIT_REASON_PREEMPTION_TIMER, 0, 0);
4142 		return 0;
4143 	}
4144 
4145 	if (vcpu->arch.smi_pending && !is_smm(vcpu)) {
4146 		if (block_nested_events)
4147 			return -EBUSY;
4148 		goto no_vmexit;
4149 	}
4150 
4151 	if (vcpu->arch.nmi_pending && !vmx_nmi_blocked(vcpu)) {
4152 		if (block_nested_events)
4153 			return -EBUSY;
4154 		if (!nested_exit_on_nmi(vcpu))
4155 			goto no_vmexit;
4156 
4157 		nested_vmx_vmexit(vcpu, EXIT_REASON_EXCEPTION_NMI,
4158 				  NMI_VECTOR | INTR_TYPE_NMI_INTR |
4159 				  INTR_INFO_VALID_MASK, 0);
4160 		/*
4161 		 * The NMI-triggered VM exit counts as injection:
4162 		 * clear this one and block further NMIs.
4163 		 */
4164 		vcpu->arch.nmi_pending = 0;
4165 		vmx_set_nmi_mask(vcpu, true);
4166 		return 0;
4167 	}
4168 
4169 	if (kvm_cpu_has_interrupt(vcpu) && !vmx_interrupt_blocked(vcpu)) {
4170 		if (block_nested_events)
4171 			return -EBUSY;
4172 		if (!nested_exit_on_intr(vcpu))
4173 			goto no_vmexit;
4174 		nested_vmx_vmexit(vcpu, EXIT_REASON_EXTERNAL_INTERRUPT, 0, 0);
4175 		return 0;
4176 	}
4177 
4178 no_vmexit:
4179 	return vmx_complete_nested_posted_interrupt(vcpu);
4180 }
4181 
4182 static u32 vmx_get_preemption_timer_value(struct kvm_vcpu *vcpu)
4183 {
4184 	ktime_t remaining =
4185 		hrtimer_get_remaining(&to_vmx(vcpu)->nested.preemption_timer);
4186 	u64 value;
4187 
4188 	if (ktime_to_ns(remaining) <= 0)
4189 		return 0;
4190 
4191 	value = ktime_to_ns(remaining) * vcpu->arch.virtual_tsc_khz;
4192 	do_div(value, 1000000);
4193 	return value >> VMX_MISC_EMULATED_PREEMPTION_TIMER_RATE;
4194 }
4195 
4196 static bool is_vmcs12_ext_field(unsigned long field)
4197 {
4198 	switch (field) {
4199 	case GUEST_ES_SELECTOR:
4200 	case GUEST_CS_SELECTOR:
4201 	case GUEST_SS_SELECTOR:
4202 	case GUEST_DS_SELECTOR:
4203 	case GUEST_FS_SELECTOR:
4204 	case GUEST_GS_SELECTOR:
4205 	case GUEST_LDTR_SELECTOR:
4206 	case GUEST_TR_SELECTOR:
4207 	case GUEST_ES_LIMIT:
4208 	case GUEST_CS_LIMIT:
4209 	case GUEST_SS_LIMIT:
4210 	case GUEST_DS_LIMIT:
4211 	case GUEST_FS_LIMIT:
4212 	case GUEST_GS_LIMIT:
4213 	case GUEST_LDTR_LIMIT:
4214 	case GUEST_TR_LIMIT:
4215 	case GUEST_GDTR_LIMIT:
4216 	case GUEST_IDTR_LIMIT:
4217 	case GUEST_ES_AR_BYTES:
4218 	case GUEST_DS_AR_BYTES:
4219 	case GUEST_FS_AR_BYTES:
4220 	case GUEST_GS_AR_BYTES:
4221 	case GUEST_LDTR_AR_BYTES:
4222 	case GUEST_TR_AR_BYTES:
4223 	case GUEST_ES_BASE:
4224 	case GUEST_CS_BASE:
4225 	case GUEST_SS_BASE:
4226 	case GUEST_DS_BASE:
4227 	case GUEST_FS_BASE:
4228 	case GUEST_GS_BASE:
4229 	case GUEST_LDTR_BASE:
4230 	case GUEST_TR_BASE:
4231 	case GUEST_GDTR_BASE:
4232 	case GUEST_IDTR_BASE:
4233 	case GUEST_PENDING_DBG_EXCEPTIONS:
4234 	case GUEST_BNDCFGS:
4235 		return true;
4236 	default:
4237 		break;
4238 	}
4239 
4240 	return false;
4241 }
4242 
4243 static void sync_vmcs02_to_vmcs12_rare(struct kvm_vcpu *vcpu,
4244 				       struct vmcs12 *vmcs12)
4245 {
4246 	struct vcpu_vmx *vmx = to_vmx(vcpu);
4247 
4248 	vmcs12->guest_es_selector = vmcs_read16(GUEST_ES_SELECTOR);
4249 	vmcs12->guest_cs_selector = vmcs_read16(GUEST_CS_SELECTOR);
4250 	vmcs12->guest_ss_selector = vmcs_read16(GUEST_SS_SELECTOR);
4251 	vmcs12->guest_ds_selector = vmcs_read16(GUEST_DS_SELECTOR);
4252 	vmcs12->guest_fs_selector = vmcs_read16(GUEST_FS_SELECTOR);
4253 	vmcs12->guest_gs_selector = vmcs_read16(GUEST_GS_SELECTOR);
4254 	vmcs12->guest_ldtr_selector = vmcs_read16(GUEST_LDTR_SELECTOR);
4255 	vmcs12->guest_tr_selector = vmcs_read16(GUEST_TR_SELECTOR);
4256 	vmcs12->guest_es_limit = vmcs_read32(GUEST_ES_LIMIT);
4257 	vmcs12->guest_cs_limit = vmcs_read32(GUEST_CS_LIMIT);
4258 	vmcs12->guest_ss_limit = vmcs_read32(GUEST_SS_LIMIT);
4259 	vmcs12->guest_ds_limit = vmcs_read32(GUEST_DS_LIMIT);
4260 	vmcs12->guest_fs_limit = vmcs_read32(GUEST_FS_LIMIT);
4261 	vmcs12->guest_gs_limit = vmcs_read32(GUEST_GS_LIMIT);
4262 	vmcs12->guest_ldtr_limit = vmcs_read32(GUEST_LDTR_LIMIT);
4263 	vmcs12->guest_tr_limit = vmcs_read32(GUEST_TR_LIMIT);
4264 	vmcs12->guest_gdtr_limit = vmcs_read32(GUEST_GDTR_LIMIT);
4265 	vmcs12->guest_idtr_limit = vmcs_read32(GUEST_IDTR_LIMIT);
4266 	vmcs12->guest_es_ar_bytes = vmcs_read32(GUEST_ES_AR_BYTES);
4267 	vmcs12->guest_ds_ar_bytes = vmcs_read32(GUEST_DS_AR_BYTES);
4268 	vmcs12->guest_fs_ar_bytes = vmcs_read32(GUEST_FS_AR_BYTES);
4269 	vmcs12->guest_gs_ar_bytes = vmcs_read32(GUEST_GS_AR_BYTES);
4270 	vmcs12->guest_ldtr_ar_bytes = vmcs_read32(GUEST_LDTR_AR_BYTES);
4271 	vmcs12->guest_tr_ar_bytes = vmcs_read32(GUEST_TR_AR_BYTES);
4272 	vmcs12->guest_es_base = vmcs_readl(GUEST_ES_BASE);
4273 	vmcs12->guest_cs_base = vmcs_readl(GUEST_CS_BASE);
4274 	vmcs12->guest_ss_base = vmcs_readl(GUEST_SS_BASE);
4275 	vmcs12->guest_ds_base = vmcs_readl(GUEST_DS_BASE);
4276 	vmcs12->guest_fs_base = vmcs_readl(GUEST_FS_BASE);
4277 	vmcs12->guest_gs_base = vmcs_readl(GUEST_GS_BASE);
4278 	vmcs12->guest_ldtr_base = vmcs_readl(GUEST_LDTR_BASE);
4279 	vmcs12->guest_tr_base = vmcs_readl(GUEST_TR_BASE);
4280 	vmcs12->guest_gdtr_base = vmcs_readl(GUEST_GDTR_BASE);
4281 	vmcs12->guest_idtr_base = vmcs_readl(GUEST_IDTR_BASE);
4282 	vmcs12->guest_pending_dbg_exceptions =
4283 		vmcs_readl(GUEST_PENDING_DBG_EXCEPTIONS);
4284 
4285 	vmx->nested.need_sync_vmcs02_to_vmcs12_rare = false;
4286 }
4287 
4288 static void copy_vmcs02_to_vmcs12_rare(struct kvm_vcpu *vcpu,
4289 				       struct vmcs12 *vmcs12)
4290 {
4291 	struct vcpu_vmx *vmx = to_vmx(vcpu);
4292 	int cpu;
4293 
4294 	if (!vmx->nested.need_sync_vmcs02_to_vmcs12_rare)
4295 		return;
4296 
4297 
4298 	WARN_ON_ONCE(vmx->loaded_vmcs != &vmx->vmcs01);
4299 
4300 	cpu = get_cpu();
4301 	vmx->loaded_vmcs = &vmx->nested.vmcs02;
4302 	vmx_vcpu_load_vmcs(vcpu, cpu, &vmx->vmcs01);
4303 
4304 	sync_vmcs02_to_vmcs12_rare(vcpu, vmcs12);
4305 
4306 	vmx->loaded_vmcs = &vmx->vmcs01;
4307 	vmx_vcpu_load_vmcs(vcpu, cpu, &vmx->nested.vmcs02);
4308 	put_cpu();
4309 }
4310 
4311 /*
4312  * Update the guest state fields of vmcs12 to reflect changes that
4313  * occurred while L2 was running. (The "IA-32e mode guest" bit of the
4314  * VM-entry controls is also updated, since this is really a guest
4315  * state bit.)
4316  */
4317 static void sync_vmcs02_to_vmcs12(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12)
4318 {
4319 	struct vcpu_vmx *vmx = to_vmx(vcpu);
4320 
4321 	if (evmptr_is_valid(vmx->nested.hv_evmcs_vmptr))
4322 		sync_vmcs02_to_vmcs12_rare(vcpu, vmcs12);
4323 
4324 	vmx->nested.need_sync_vmcs02_to_vmcs12_rare =
4325 		!evmptr_is_valid(vmx->nested.hv_evmcs_vmptr);
4326 
4327 	vmcs12->guest_cr0 = vmcs12_guest_cr0(vcpu, vmcs12);
4328 	vmcs12->guest_cr4 = vmcs12_guest_cr4(vcpu, vmcs12);
4329 
4330 	vmcs12->guest_rsp = kvm_rsp_read(vcpu);
4331 	vmcs12->guest_rip = kvm_rip_read(vcpu);
4332 	vmcs12->guest_rflags = vmcs_readl(GUEST_RFLAGS);
4333 
4334 	vmcs12->guest_cs_ar_bytes = vmcs_read32(GUEST_CS_AR_BYTES);
4335 	vmcs12->guest_ss_ar_bytes = vmcs_read32(GUEST_SS_AR_BYTES);
4336 
4337 	vmcs12->guest_interruptibility_info =
4338 		vmcs_read32(GUEST_INTERRUPTIBILITY_INFO);
4339 
4340 	if (vcpu->arch.mp_state == KVM_MP_STATE_HALTED)
4341 		vmcs12->guest_activity_state = GUEST_ACTIVITY_HLT;
4342 	else if (vcpu->arch.mp_state == KVM_MP_STATE_INIT_RECEIVED)
4343 		vmcs12->guest_activity_state = GUEST_ACTIVITY_WAIT_SIPI;
4344 	else
4345 		vmcs12->guest_activity_state = GUEST_ACTIVITY_ACTIVE;
4346 
4347 	if (nested_cpu_has_preemption_timer(vmcs12) &&
4348 	    vmcs12->vm_exit_controls & VM_EXIT_SAVE_VMX_PREEMPTION_TIMER &&
4349 	    !vmx->nested.nested_run_pending)
4350 		vmcs12->vmx_preemption_timer_value =
4351 			vmx_get_preemption_timer_value(vcpu);
4352 
4353 	/*
4354 	 * In some cases (usually, nested EPT), L2 is allowed to change its
4355 	 * own CR3 without exiting. If it has changed it, we must keep it.
4356 	 * Of course, if L0 is using shadow page tables, GUEST_CR3 was defined
4357 	 * by L0, not L1 or L2, so we mustn't unconditionally copy it to vmcs12.
4358 	 *
4359 	 * Additionally, restore L2's PDPTR to vmcs12.
4360 	 */
4361 	if (enable_ept) {
4362 		vmcs12->guest_cr3 = vmcs_readl(GUEST_CR3);
4363 		if (nested_cpu_has_ept(vmcs12) && is_pae_paging(vcpu)) {
4364 			vmcs12->guest_pdptr0 = vmcs_read64(GUEST_PDPTR0);
4365 			vmcs12->guest_pdptr1 = vmcs_read64(GUEST_PDPTR1);
4366 			vmcs12->guest_pdptr2 = vmcs_read64(GUEST_PDPTR2);
4367 			vmcs12->guest_pdptr3 = vmcs_read64(GUEST_PDPTR3);
4368 		}
4369 	}
4370 
4371 	vmcs12->guest_linear_address = vmcs_readl(GUEST_LINEAR_ADDRESS);
4372 
4373 	if (nested_cpu_has_vid(vmcs12))
4374 		vmcs12->guest_intr_status = vmcs_read16(GUEST_INTR_STATUS);
4375 
4376 	vmcs12->vm_entry_controls =
4377 		(vmcs12->vm_entry_controls & ~VM_ENTRY_IA32E_MODE) |
4378 		(vm_entry_controls_get(to_vmx(vcpu)) & VM_ENTRY_IA32E_MODE);
4379 
4380 	if (vmcs12->vm_exit_controls & VM_EXIT_SAVE_DEBUG_CONTROLS)
4381 		kvm_get_dr(vcpu, 7, (unsigned long *)&vmcs12->guest_dr7);
4382 
4383 	if (vmcs12->vm_exit_controls & VM_EXIT_SAVE_IA32_EFER)
4384 		vmcs12->guest_ia32_efer = vcpu->arch.efer;
4385 }
4386 
4387 /*
4388  * prepare_vmcs12 is part of what we need to do when the nested L2 guest exits
4389  * and we want to prepare to run its L1 parent. L1 keeps a vmcs for L2 (vmcs12),
4390  * and this function updates it to reflect the changes to the guest state while
4391  * L2 was running (and perhaps made some exits which were handled directly by L0
4392  * without going back to L1), and to reflect the exit reason.
4393  * Note that we do not have to copy here all VMCS fields, just those that
4394  * could have changed by the L2 guest or the exit - i.e., the guest-state and
4395  * exit-information fields only. Other fields are modified by L1 with VMWRITE,
4396  * which already writes to vmcs12 directly.
4397  */
4398 static void prepare_vmcs12(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12,
4399 			   u32 vm_exit_reason, u32 exit_intr_info,
4400 			   unsigned long exit_qualification)
4401 {
4402 	/* update exit information fields: */
4403 	vmcs12->vm_exit_reason = vm_exit_reason;
4404 	if (to_vmx(vcpu)->exit_reason.enclave_mode)
4405 		vmcs12->vm_exit_reason |= VMX_EXIT_REASONS_SGX_ENCLAVE_MODE;
4406 	vmcs12->exit_qualification = exit_qualification;
4407 
4408 	/*
4409 	 * On VM-Exit due to a failed VM-Entry, the VMCS isn't marked launched
4410 	 * and only EXIT_REASON and EXIT_QUALIFICATION are updated, all other
4411 	 * exit info fields are unmodified.
4412 	 */
4413 	if (!(vmcs12->vm_exit_reason & VMX_EXIT_REASONS_FAILED_VMENTRY)) {
4414 		vmcs12->launch_state = 1;
4415 
4416 		/* vm_entry_intr_info_field is cleared on exit. Emulate this
4417 		 * instead of reading the real value. */
4418 		vmcs12->vm_entry_intr_info_field &= ~INTR_INFO_VALID_MASK;
4419 
4420 		/*
4421 		 * Transfer the event that L0 or L1 may wanted to inject into
4422 		 * L2 to IDT_VECTORING_INFO_FIELD.
4423 		 */
4424 		vmcs12_save_pending_event(vcpu, vmcs12,
4425 					  vm_exit_reason, exit_intr_info);
4426 
4427 		vmcs12->vm_exit_intr_info = exit_intr_info;
4428 		vmcs12->vm_exit_instruction_len = vmcs_read32(VM_EXIT_INSTRUCTION_LEN);
4429 		vmcs12->vmx_instruction_info = vmcs_read32(VMX_INSTRUCTION_INFO);
4430 
4431 		/*
4432 		 * According to spec, there's no need to store the guest's
4433 		 * MSRs if the exit is due to a VM-entry failure that occurs
4434 		 * during or after loading the guest state. Since this exit
4435 		 * does not fall in that category, we need to save the MSRs.
4436 		 */
4437 		if (nested_vmx_store_msr(vcpu,
4438 					 vmcs12->vm_exit_msr_store_addr,
4439 					 vmcs12->vm_exit_msr_store_count))
4440 			nested_vmx_abort(vcpu,
4441 					 VMX_ABORT_SAVE_GUEST_MSR_FAIL);
4442 	}
4443 }
4444 
4445 /*
4446  * A part of what we need to when the nested L2 guest exits and we want to
4447  * run its L1 parent, is to reset L1's guest state to the host state specified
4448  * in vmcs12.
4449  * This function is to be called not only on normal nested exit, but also on
4450  * a nested entry failure, as explained in Intel's spec, 3B.23.7 ("VM-Entry
4451  * Failures During or After Loading Guest State").
4452  * This function should be called when the active VMCS is L1's (vmcs01).
4453  */
4454 static void load_vmcs12_host_state(struct kvm_vcpu *vcpu,
4455 				   struct vmcs12 *vmcs12)
4456 {
4457 	enum vm_entry_failure_code ignored;
4458 	struct kvm_segment seg;
4459 
4460 	if (vmcs12->vm_exit_controls & VM_EXIT_LOAD_IA32_EFER)
4461 		vcpu->arch.efer = vmcs12->host_ia32_efer;
4462 	else if (vmcs12->vm_exit_controls & VM_EXIT_HOST_ADDR_SPACE_SIZE)
4463 		vcpu->arch.efer |= (EFER_LMA | EFER_LME);
4464 	else
4465 		vcpu->arch.efer &= ~(EFER_LMA | EFER_LME);
4466 	vmx_set_efer(vcpu, vcpu->arch.efer);
4467 
4468 	kvm_rsp_write(vcpu, vmcs12->host_rsp);
4469 	kvm_rip_write(vcpu, vmcs12->host_rip);
4470 	vmx_set_rflags(vcpu, X86_EFLAGS_FIXED);
4471 	vmx_set_interrupt_shadow(vcpu, 0);
4472 
4473 	/*
4474 	 * Note that calling vmx_set_cr0 is important, even if cr0 hasn't
4475 	 * actually changed, because vmx_set_cr0 refers to efer set above.
4476 	 *
4477 	 * CR0_GUEST_HOST_MASK is already set in the original vmcs01
4478 	 * (KVM doesn't change it);
4479 	 */
4480 	vcpu->arch.cr0_guest_owned_bits = KVM_POSSIBLE_CR0_GUEST_BITS;
4481 	vmx_set_cr0(vcpu, vmcs12->host_cr0);
4482 
4483 	/* Same as above - no reason to call set_cr4_guest_host_mask().  */
4484 	vcpu->arch.cr4_guest_owned_bits = ~vmcs_readl(CR4_GUEST_HOST_MASK);
4485 	vmx_set_cr4(vcpu, vmcs12->host_cr4);
4486 
4487 	nested_ept_uninit_mmu_context(vcpu);
4488 
4489 	/*
4490 	 * Only PDPTE load can fail as the value of cr3 was checked on entry and
4491 	 * couldn't have changed.
4492 	 */
4493 	if (nested_vmx_load_cr3(vcpu, vmcs12->host_cr3, false, true, &ignored))
4494 		nested_vmx_abort(vcpu, VMX_ABORT_LOAD_HOST_PDPTE_FAIL);
4495 
4496 	nested_vmx_transition_tlb_flush(vcpu, vmcs12, false);
4497 
4498 	vmcs_write32(GUEST_SYSENTER_CS, vmcs12->host_ia32_sysenter_cs);
4499 	vmcs_writel(GUEST_SYSENTER_ESP, vmcs12->host_ia32_sysenter_esp);
4500 	vmcs_writel(GUEST_SYSENTER_EIP, vmcs12->host_ia32_sysenter_eip);
4501 	vmcs_writel(GUEST_IDTR_BASE, vmcs12->host_idtr_base);
4502 	vmcs_writel(GUEST_GDTR_BASE, vmcs12->host_gdtr_base);
4503 	vmcs_write32(GUEST_IDTR_LIMIT, 0xFFFF);
4504 	vmcs_write32(GUEST_GDTR_LIMIT, 0xFFFF);
4505 
4506 	/* If not VM_EXIT_CLEAR_BNDCFGS, the L2 value propagates to L1.  */
4507 	if (vmcs12->vm_exit_controls & VM_EXIT_CLEAR_BNDCFGS)
4508 		vmcs_write64(GUEST_BNDCFGS, 0);
4509 
4510 	if (vmcs12->vm_exit_controls & VM_EXIT_LOAD_IA32_PAT) {
4511 		vmcs_write64(GUEST_IA32_PAT, vmcs12->host_ia32_pat);
4512 		vcpu->arch.pat = vmcs12->host_ia32_pat;
4513 	}
4514 	if ((vmcs12->vm_exit_controls & VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL) &&
4515 	    intel_pmu_has_perf_global_ctrl(vcpu_to_pmu(vcpu)))
4516 		WARN_ON_ONCE(kvm_set_msr(vcpu, MSR_CORE_PERF_GLOBAL_CTRL,
4517 					 vmcs12->host_ia32_perf_global_ctrl));
4518 
4519 	/* Set L1 segment info according to Intel SDM
4520 	    27.5.2 Loading Host Segment and Descriptor-Table Registers */
4521 	seg = (struct kvm_segment) {
4522 		.base = 0,
4523 		.limit = 0xFFFFFFFF,
4524 		.selector = vmcs12->host_cs_selector,
4525 		.type = 11,
4526 		.present = 1,
4527 		.s = 1,
4528 		.g = 1
4529 	};
4530 	if (vmcs12->vm_exit_controls & VM_EXIT_HOST_ADDR_SPACE_SIZE)
4531 		seg.l = 1;
4532 	else
4533 		seg.db = 1;
4534 	__vmx_set_segment(vcpu, &seg, VCPU_SREG_CS);
4535 	seg = (struct kvm_segment) {
4536 		.base = 0,
4537 		.limit = 0xFFFFFFFF,
4538 		.type = 3,
4539 		.present = 1,
4540 		.s = 1,
4541 		.db = 1,
4542 		.g = 1
4543 	};
4544 	seg.selector = vmcs12->host_ds_selector;
4545 	__vmx_set_segment(vcpu, &seg, VCPU_SREG_DS);
4546 	seg.selector = vmcs12->host_es_selector;
4547 	__vmx_set_segment(vcpu, &seg, VCPU_SREG_ES);
4548 	seg.selector = vmcs12->host_ss_selector;
4549 	__vmx_set_segment(vcpu, &seg, VCPU_SREG_SS);
4550 	seg.selector = vmcs12->host_fs_selector;
4551 	seg.base = vmcs12->host_fs_base;
4552 	__vmx_set_segment(vcpu, &seg, VCPU_SREG_FS);
4553 	seg.selector = vmcs12->host_gs_selector;
4554 	seg.base = vmcs12->host_gs_base;
4555 	__vmx_set_segment(vcpu, &seg, VCPU_SREG_GS);
4556 	seg = (struct kvm_segment) {
4557 		.base = vmcs12->host_tr_base,
4558 		.limit = 0x67,
4559 		.selector = vmcs12->host_tr_selector,
4560 		.type = 11,
4561 		.present = 1
4562 	};
4563 	__vmx_set_segment(vcpu, &seg, VCPU_SREG_TR);
4564 
4565 	memset(&seg, 0, sizeof(seg));
4566 	seg.unusable = 1;
4567 	__vmx_set_segment(vcpu, &seg, VCPU_SREG_LDTR);
4568 
4569 	kvm_set_dr(vcpu, 7, 0x400);
4570 	vmcs_write64(GUEST_IA32_DEBUGCTL, 0);
4571 
4572 	if (nested_vmx_load_msr(vcpu, vmcs12->vm_exit_msr_load_addr,
4573 				vmcs12->vm_exit_msr_load_count))
4574 		nested_vmx_abort(vcpu, VMX_ABORT_LOAD_HOST_MSR_FAIL);
4575 
4576 	to_vmx(vcpu)->emulation_required = vmx_emulation_required(vcpu);
4577 }
4578 
4579 static inline u64 nested_vmx_get_vmcs01_guest_efer(struct vcpu_vmx *vmx)
4580 {
4581 	struct vmx_uret_msr *efer_msr;
4582 	unsigned int i;
4583 
4584 	if (vm_entry_controls_get(vmx) & VM_ENTRY_LOAD_IA32_EFER)
4585 		return vmcs_read64(GUEST_IA32_EFER);
4586 
4587 	if (cpu_has_load_ia32_efer())
4588 		return host_efer;
4589 
4590 	for (i = 0; i < vmx->msr_autoload.guest.nr; ++i) {
4591 		if (vmx->msr_autoload.guest.val[i].index == MSR_EFER)
4592 			return vmx->msr_autoload.guest.val[i].value;
4593 	}
4594 
4595 	efer_msr = vmx_find_uret_msr(vmx, MSR_EFER);
4596 	if (efer_msr)
4597 		return efer_msr->data;
4598 
4599 	return host_efer;
4600 }
4601 
4602 static void nested_vmx_restore_host_state(struct kvm_vcpu *vcpu)
4603 {
4604 	struct vmcs12 *vmcs12 = get_vmcs12(vcpu);
4605 	struct vcpu_vmx *vmx = to_vmx(vcpu);
4606 	struct vmx_msr_entry g, h;
4607 	gpa_t gpa;
4608 	u32 i, j;
4609 
4610 	vcpu->arch.pat = vmcs_read64(GUEST_IA32_PAT);
4611 
4612 	if (vmcs12->vm_entry_controls & VM_ENTRY_LOAD_DEBUG_CONTROLS) {
4613 		/*
4614 		 * L1's host DR7 is lost if KVM_GUESTDBG_USE_HW_BP is set
4615 		 * as vmcs01.GUEST_DR7 contains a userspace defined value
4616 		 * and vcpu->arch.dr7 is not squirreled away before the
4617 		 * nested VMENTER (not worth adding a variable in nested_vmx).
4618 		 */
4619 		if (vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP)
4620 			kvm_set_dr(vcpu, 7, DR7_FIXED_1);
4621 		else
4622 			WARN_ON(kvm_set_dr(vcpu, 7, vmcs_readl(GUEST_DR7)));
4623 	}
4624 
4625 	/*
4626 	 * Note that calling vmx_set_{efer,cr0,cr4} is important as they
4627 	 * handle a variety of side effects to KVM's software model.
4628 	 */
4629 	vmx_set_efer(vcpu, nested_vmx_get_vmcs01_guest_efer(vmx));
4630 
4631 	vcpu->arch.cr0_guest_owned_bits = KVM_POSSIBLE_CR0_GUEST_BITS;
4632 	vmx_set_cr0(vcpu, vmcs_readl(CR0_READ_SHADOW));
4633 
4634 	vcpu->arch.cr4_guest_owned_bits = ~vmcs_readl(CR4_GUEST_HOST_MASK);
4635 	vmx_set_cr4(vcpu, vmcs_readl(CR4_READ_SHADOW));
4636 
4637 	nested_ept_uninit_mmu_context(vcpu);
4638 	vcpu->arch.cr3 = vmcs_readl(GUEST_CR3);
4639 	kvm_register_mark_available(vcpu, VCPU_EXREG_CR3);
4640 
4641 	/*
4642 	 * Use ept_save_pdptrs(vcpu) to load the MMU's cached PDPTRs
4643 	 * from vmcs01 (if necessary).  The PDPTRs are not loaded on
4644 	 * VMFail, like everything else we just need to ensure our
4645 	 * software model is up-to-date.
4646 	 */
4647 	if (enable_ept && is_pae_paging(vcpu))
4648 		ept_save_pdptrs(vcpu);
4649 
4650 	kvm_mmu_reset_context(vcpu);
4651 
4652 	/*
4653 	 * This nasty bit of open coding is a compromise between blindly
4654 	 * loading L1's MSRs using the exit load lists (incorrect emulation
4655 	 * of VMFail), leaving the nested VM's MSRs in the software model
4656 	 * (incorrect behavior) and snapshotting the modified MSRs (too
4657 	 * expensive since the lists are unbound by hardware).  For each
4658 	 * MSR that was (prematurely) loaded from the nested VMEntry load
4659 	 * list, reload it from the exit load list if it exists and differs
4660 	 * from the guest value.  The intent is to stuff host state as
4661 	 * silently as possible, not to fully process the exit load list.
4662 	 */
4663 	for (i = 0; i < vmcs12->vm_entry_msr_load_count; i++) {
4664 		gpa = vmcs12->vm_entry_msr_load_addr + (i * sizeof(g));
4665 		if (kvm_vcpu_read_guest(vcpu, gpa, &g, sizeof(g))) {
4666 			pr_debug_ratelimited(
4667 				"%s read MSR index failed (%u, 0x%08llx)\n",
4668 				__func__, i, gpa);
4669 			goto vmabort;
4670 		}
4671 
4672 		for (j = 0; j < vmcs12->vm_exit_msr_load_count; j++) {
4673 			gpa = vmcs12->vm_exit_msr_load_addr + (j * sizeof(h));
4674 			if (kvm_vcpu_read_guest(vcpu, gpa, &h, sizeof(h))) {
4675 				pr_debug_ratelimited(
4676 					"%s read MSR failed (%u, 0x%08llx)\n",
4677 					__func__, j, gpa);
4678 				goto vmabort;
4679 			}
4680 			if (h.index != g.index)
4681 				continue;
4682 			if (h.value == g.value)
4683 				break;
4684 
4685 			if (nested_vmx_load_msr_check(vcpu, &h)) {
4686 				pr_debug_ratelimited(
4687 					"%s check failed (%u, 0x%x, 0x%x)\n",
4688 					__func__, j, h.index, h.reserved);
4689 				goto vmabort;
4690 			}
4691 
4692 			if (kvm_set_msr(vcpu, h.index, h.value)) {
4693 				pr_debug_ratelimited(
4694 					"%s WRMSR failed (%u, 0x%x, 0x%llx)\n",
4695 					__func__, j, h.index, h.value);
4696 				goto vmabort;
4697 			}
4698 		}
4699 	}
4700 
4701 	return;
4702 
4703 vmabort:
4704 	nested_vmx_abort(vcpu, VMX_ABORT_LOAD_HOST_MSR_FAIL);
4705 }
4706 
4707 /*
4708  * Emulate an exit from nested guest (L2) to L1, i.e., prepare to run L1
4709  * and modify vmcs12 to make it see what it would expect to see there if
4710  * L2 was its real guest. Must only be called when in L2 (is_guest_mode())
4711  */
4712 void nested_vmx_vmexit(struct kvm_vcpu *vcpu, u32 vm_exit_reason,
4713 		       u32 exit_intr_info, unsigned long exit_qualification)
4714 {
4715 	struct vcpu_vmx *vmx = to_vmx(vcpu);
4716 	struct vmcs12 *vmcs12 = get_vmcs12(vcpu);
4717 
4718 	/* Pending MTF traps are discarded on VM-Exit. */
4719 	vmx->nested.mtf_pending = false;
4720 
4721 	/* trying to cancel vmlaunch/vmresume is a bug */
4722 	WARN_ON_ONCE(vmx->nested.nested_run_pending);
4723 
4724 	if (kvm_check_request(KVM_REQ_GET_NESTED_STATE_PAGES, vcpu)) {
4725 		/*
4726 		 * KVM_REQ_GET_NESTED_STATE_PAGES is also used to map
4727 		 * Enlightened VMCS after migration and we still need to
4728 		 * do that when something is forcing L2->L1 exit prior to
4729 		 * the first L2 run.
4730 		 */
4731 		(void)nested_get_evmcs_page(vcpu);
4732 	}
4733 
4734 	/* Service pending TLB flush requests for L2 before switching to L1. */
4735 	kvm_service_local_tlb_flush_requests(vcpu);
4736 
4737 	/*
4738 	 * VCPU_EXREG_PDPTR will be clobbered in arch/x86/kvm/vmx/vmx.h between
4739 	 * now and the new vmentry.  Ensure that the VMCS02 PDPTR fields are
4740 	 * up-to-date before switching to L1.
4741 	 */
4742 	if (enable_ept && is_pae_paging(vcpu))
4743 		vmx_ept_load_pdptrs(vcpu);
4744 
4745 	leave_guest_mode(vcpu);
4746 
4747 	if (nested_cpu_has_preemption_timer(vmcs12))
4748 		hrtimer_cancel(&to_vmx(vcpu)->nested.preemption_timer);
4749 
4750 	if (nested_cpu_has(vmcs12, CPU_BASED_USE_TSC_OFFSETTING)) {
4751 		vcpu->arch.tsc_offset = vcpu->arch.l1_tsc_offset;
4752 		if (nested_cpu_has2(vmcs12, SECONDARY_EXEC_TSC_SCALING))
4753 			vcpu->arch.tsc_scaling_ratio = vcpu->arch.l1_tsc_scaling_ratio;
4754 	}
4755 
4756 	if (likely(!vmx->fail)) {
4757 		sync_vmcs02_to_vmcs12(vcpu, vmcs12);
4758 
4759 		if (vm_exit_reason != -1)
4760 			prepare_vmcs12(vcpu, vmcs12, vm_exit_reason,
4761 				       exit_intr_info, exit_qualification);
4762 
4763 		/*
4764 		 * Must happen outside of sync_vmcs02_to_vmcs12() as it will
4765 		 * also be used to capture vmcs12 cache as part of
4766 		 * capturing nVMX state for snapshot (migration).
4767 		 *
4768 		 * Otherwise, this flush will dirty guest memory at a
4769 		 * point it is already assumed by user-space to be
4770 		 * immutable.
4771 		 */
4772 		nested_flush_cached_shadow_vmcs12(vcpu, vmcs12);
4773 	} else {
4774 		/*
4775 		 * The only expected VM-instruction error is "VM entry with
4776 		 * invalid control field(s)." Anything else indicates a
4777 		 * problem with L0.  And we should never get here with a
4778 		 * VMFail of any type if early consistency checks are enabled.
4779 		 */
4780 		WARN_ON_ONCE(vmcs_read32(VM_INSTRUCTION_ERROR) !=
4781 			     VMXERR_ENTRY_INVALID_CONTROL_FIELD);
4782 		WARN_ON_ONCE(nested_early_check);
4783 	}
4784 
4785 	/*
4786 	 * Drop events/exceptions that were queued for re-injection to L2
4787 	 * (picked up via vmx_complete_interrupts()), as well as exceptions
4788 	 * that were pending for L2.  Note, this must NOT be hoisted above
4789 	 * prepare_vmcs12(), events/exceptions queued for re-injection need to
4790 	 * be captured in vmcs12 (see vmcs12_save_pending_event()).
4791 	 */
4792 	vcpu->arch.nmi_injected = false;
4793 	kvm_clear_exception_queue(vcpu);
4794 	kvm_clear_interrupt_queue(vcpu);
4795 
4796 	vmx_switch_vmcs(vcpu, &vmx->vmcs01);
4797 
4798 	/*
4799 	 * If IBRS is advertised to the vCPU, KVM must flush the indirect
4800 	 * branch predictors when transitioning from L2 to L1, as L1 expects
4801 	 * hardware (KVM in this case) to provide separate predictor modes.
4802 	 * Bare metal isolates VMX root (host) from VMX non-root (guest), but
4803 	 * doesn't isolate different VMCSs, i.e. in this case, doesn't provide
4804 	 * separate modes for L2 vs L1.
4805 	 */
4806 	if (guest_cpuid_has(vcpu, X86_FEATURE_SPEC_CTRL))
4807 		indirect_branch_prediction_barrier();
4808 
4809 	/* Update any VMCS fields that might have changed while L2 ran */
4810 	vmcs_write32(VM_EXIT_MSR_LOAD_COUNT, vmx->msr_autoload.host.nr);
4811 	vmcs_write32(VM_ENTRY_MSR_LOAD_COUNT, vmx->msr_autoload.guest.nr);
4812 	vmcs_write64(TSC_OFFSET, vcpu->arch.tsc_offset);
4813 	if (kvm_caps.has_tsc_control)
4814 		vmcs_write64(TSC_MULTIPLIER, vcpu->arch.tsc_scaling_ratio);
4815 
4816 	if (vmx->nested.l1_tpr_threshold != -1)
4817 		vmcs_write32(TPR_THRESHOLD, vmx->nested.l1_tpr_threshold);
4818 
4819 	if (vmx->nested.change_vmcs01_virtual_apic_mode) {
4820 		vmx->nested.change_vmcs01_virtual_apic_mode = false;
4821 		vmx_set_virtual_apic_mode(vcpu);
4822 	}
4823 
4824 	if (vmx->nested.update_vmcs01_cpu_dirty_logging) {
4825 		vmx->nested.update_vmcs01_cpu_dirty_logging = false;
4826 		vmx_update_cpu_dirty_logging(vcpu);
4827 	}
4828 
4829 	/* Unpin physical memory we referred to in vmcs02 */
4830 	kvm_vcpu_unmap(vcpu, &vmx->nested.apic_access_page_map, false);
4831 	kvm_vcpu_unmap(vcpu, &vmx->nested.virtual_apic_map, true);
4832 	kvm_vcpu_unmap(vcpu, &vmx->nested.pi_desc_map, true);
4833 	vmx->nested.pi_desc = NULL;
4834 
4835 	if (vmx->nested.reload_vmcs01_apic_access_page) {
4836 		vmx->nested.reload_vmcs01_apic_access_page = false;
4837 		kvm_make_request(KVM_REQ_APIC_PAGE_RELOAD, vcpu);
4838 	}
4839 
4840 	if (vmx->nested.update_vmcs01_apicv_status) {
4841 		vmx->nested.update_vmcs01_apicv_status = false;
4842 		kvm_make_request(KVM_REQ_APICV_UPDATE, vcpu);
4843 	}
4844 
4845 	if ((vm_exit_reason != -1) &&
4846 	    (enable_shadow_vmcs || evmptr_is_valid(vmx->nested.hv_evmcs_vmptr)))
4847 		vmx->nested.need_vmcs12_to_shadow_sync = true;
4848 
4849 	/* in case we halted in L2 */
4850 	vcpu->arch.mp_state = KVM_MP_STATE_RUNNABLE;
4851 
4852 	if (likely(!vmx->fail)) {
4853 		if ((u16)vm_exit_reason == EXIT_REASON_EXTERNAL_INTERRUPT &&
4854 		    nested_exit_intr_ack_set(vcpu)) {
4855 			int irq = kvm_cpu_get_interrupt(vcpu);
4856 			WARN_ON(irq < 0);
4857 			vmcs12->vm_exit_intr_info = irq |
4858 				INTR_INFO_VALID_MASK | INTR_TYPE_EXT_INTR;
4859 		}
4860 
4861 		if (vm_exit_reason != -1)
4862 			trace_kvm_nested_vmexit_inject(vmcs12->vm_exit_reason,
4863 						       vmcs12->exit_qualification,
4864 						       vmcs12->idt_vectoring_info_field,
4865 						       vmcs12->vm_exit_intr_info,
4866 						       vmcs12->vm_exit_intr_error_code,
4867 						       KVM_ISA_VMX);
4868 
4869 		load_vmcs12_host_state(vcpu, vmcs12);
4870 
4871 		return;
4872 	}
4873 
4874 	/*
4875 	 * After an early L2 VM-entry failure, we're now back
4876 	 * in L1 which thinks it just finished a VMLAUNCH or
4877 	 * VMRESUME instruction, so we need to set the failure
4878 	 * flag and the VM-instruction error field of the VMCS
4879 	 * accordingly, and skip the emulated instruction.
4880 	 */
4881 	(void)nested_vmx_fail(vcpu, VMXERR_ENTRY_INVALID_CONTROL_FIELD);
4882 
4883 	/*
4884 	 * Restore L1's host state to KVM's software model.  We're here
4885 	 * because a consistency check was caught by hardware, which
4886 	 * means some amount of guest state has been propagated to KVM's
4887 	 * model and needs to be unwound to the host's state.
4888 	 */
4889 	nested_vmx_restore_host_state(vcpu);
4890 
4891 	vmx->fail = 0;
4892 }
4893 
4894 static void nested_vmx_triple_fault(struct kvm_vcpu *vcpu)
4895 {
4896 	kvm_clear_request(KVM_REQ_TRIPLE_FAULT, vcpu);
4897 	nested_vmx_vmexit(vcpu, EXIT_REASON_TRIPLE_FAULT, 0, 0);
4898 }
4899 
4900 /*
4901  * Decode the memory-address operand of a vmx instruction, as recorded on an
4902  * exit caused by such an instruction (run by a guest hypervisor).
4903  * On success, returns 0. When the operand is invalid, returns 1 and throws
4904  * #UD, #GP, or #SS.
4905  */
4906 int get_vmx_mem_address(struct kvm_vcpu *vcpu, unsigned long exit_qualification,
4907 			u32 vmx_instruction_info, bool wr, int len, gva_t *ret)
4908 {
4909 	gva_t off;
4910 	bool exn;
4911 	struct kvm_segment s;
4912 
4913 	/*
4914 	 * According to Vol. 3B, "Information for VM Exits Due to Instruction
4915 	 * Execution", on an exit, vmx_instruction_info holds most of the
4916 	 * addressing components of the operand. Only the displacement part
4917 	 * is put in exit_qualification (see 3B, "Basic VM-Exit Information").
4918 	 * For how an actual address is calculated from all these components,
4919 	 * refer to Vol. 1, "Operand Addressing".
4920 	 */
4921 	int  scaling = vmx_instruction_info & 3;
4922 	int  addr_size = (vmx_instruction_info >> 7) & 7;
4923 	bool is_reg = vmx_instruction_info & (1u << 10);
4924 	int  seg_reg = (vmx_instruction_info >> 15) & 7;
4925 	int  index_reg = (vmx_instruction_info >> 18) & 0xf;
4926 	bool index_is_valid = !(vmx_instruction_info & (1u << 22));
4927 	int  base_reg       = (vmx_instruction_info >> 23) & 0xf;
4928 	bool base_is_valid  = !(vmx_instruction_info & (1u << 27));
4929 
4930 	if (is_reg) {
4931 		kvm_queue_exception(vcpu, UD_VECTOR);
4932 		return 1;
4933 	}
4934 
4935 	/* Addr = segment_base + offset */
4936 	/* offset = base + [index * scale] + displacement */
4937 	off = exit_qualification; /* holds the displacement */
4938 	if (addr_size == 1)
4939 		off = (gva_t)sign_extend64(off, 31);
4940 	else if (addr_size == 0)
4941 		off = (gva_t)sign_extend64(off, 15);
4942 	if (base_is_valid)
4943 		off += kvm_register_read(vcpu, base_reg);
4944 	if (index_is_valid)
4945 		off += kvm_register_read(vcpu, index_reg) << scaling;
4946 	vmx_get_segment(vcpu, &s, seg_reg);
4947 
4948 	/*
4949 	 * The effective address, i.e. @off, of a memory operand is truncated
4950 	 * based on the address size of the instruction.  Note that this is
4951 	 * the *effective address*, i.e. the address prior to accounting for
4952 	 * the segment's base.
4953 	 */
4954 	if (addr_size == 1) /* 32 bit */
4955 		off &= 0xffffffff;
4956 	else if (addr_size == 0) /* 16 bit */
4957 		off &= 0xffff;
4958 
4959 	/* Checks for #GP/#SS exceptions. */
4960 	exn = false;
4961 	if (is_long_mode(vcpu)) {
4962 		/*
4963 		 * The virtual/linear address is never truncated in 64-bit
4964 		 * mode, e.g. a 32-bit address size can yield a 64-bit virtual
4965 		 * address when using FS/GS with a non-zero base.
4966 		 */
4967 		if (seg_reg == VCPU_SREG_FS || seg_reg == VCPU_SREG_GS)
4968 			*ret = s.base + off;
4969 		else
4970 			*ret = off;
4971 
4972 		/* Long mode: #GP(0)/#SS(0) if the memory address is in a
4973 		 * non-canonical form. This is the only check on the memory
4974 		 * destination for long mode!
4975 		 */
4976 		exn = is_noncanonical_address(*ret, vcpu);
4977 	} else {
4978 		/*
4979 		 * When not in long mode, the virtual/linear address is
4980 		 * unconditionally truncated to 32 bits regardless of the
4981 		 * address size.
4982 		 */
4983 		*ret = (s.base + off) & 0xffffffff;
4984 
4985 		/* Protected mode: apply checks for segment validity in the
4986 		 * following order:
4987 		 * - segment type check (#GP(0) may be thrown)
4988 		 * - usability check (#GP(0)/#SS(0))
4989 		 * - limit check (#GP(0)/#SS(0))
4990 		 */
4991 		if (wr)
4992 			/* #GP(0) if the destination operand is located in a
4993 			 * read-only data segment or any code segment.
4994 			 */
4995 			exn = ((s.type & 0xa) == 0 || (s.type & 8));
4996 		else
4997 			/* #GP(0) if the source operand is located in an
4998 			 * execute-only code segment
4999 			 */
5000 			exn = ((s.type & 0xa) == 8);
5001 		if (exn) {
5002 			kvm_queue_exception_e(vcpu, GP_VECTOR, 0);
5003 			return 1;
5004 		}
5005 		/* Protected mode: #GP(0)/#SS(0) if the segment is unusable.
5006 		 */
5007 		exn = (s.unusable != 0);
5008 
5009 		/*
5010 		 * Protected mode: #GP(0)/#SS(0) if the memory operand is
5011 		 * outside the segment limit.  All CPUs that support VMX ignore
5012 		 * limit checks for flat segments, i.e. segments with base==0,
5013 		 * limit==0xffffffff and of type expand-up data or code.
5014 		 */
5015 		if (!(s.base == 0 && s.limit == 0xffffffff &&
5016 		     ((s.type & 8) || !(s.type & 4))))
5017 			exn = exn || ((u64)off + len - 1 > s.limit);
5018 	}
5019 	if (exn) {
5020 		kvm_queue_exception_e(vcpu,
5021 				      seg_reg == VCPU_SREG_SS ?
5022 						SS_VECTOR : GP_VECTOR,
5023 				      0);
5024 		return 1;
5025 	}
5026 
5027 	return 0;
5028 }
5029 
5030 static int nested_vmx_get_vmptr(struct kvm_vcpu *vcpu, gpa_t *vmpointer,
5031 				int *ret)
5032 {
5033 	gva_t gva;
5034 	struct x86_exception e;
5035 	int r;
5036 
5037 	if (get_vmx_mem_address(vcpu, vmx_get_exit_qual(vcpu),
5038 				vmcs_read32(VMX_INSTRUCTION_INFO), false,
5039 				sizeof(*vmpointer), &gva)) {
5040 		*ret = 1;
5041 		return -EINVAL;
5042 	}
5043 
5044 	r = kvm_read_guest_virt(vcpu, gva, vmpointer, sizeof(*vmpointer), &e);
5045 	if (r != X86EMUL_CONTINUE) {
5046 		*ret = kvm_handle_memory_failure(vcpu, r, &e);
5047 		return -EINVAL;
5048 	}
5049 
5050 	return 0;
5051 }
5052 
5053 /*
5054  * Allocate a shadow VMCS and associate it with the currently loaded
5055  * VMCS, unless such a shadow VMCS already exists. The newly allocated
5056  * VMCS is also VMCLEARed, so that it is ready for use.
5057  */
5058 static struct vmcs *alloc_shadow_vmcs(struct kvm_vcpu *vcpu)
5059 {
5060 	struct vcpu_vmx *vmx = to_vmx(vcpu);
5061 	struct loaded_vmcs *loaded_vmcs = vmx->loaded_vmcs;
5062 
5063 	/*
5064 	 * KVM allocates a shadow VMCS only when L1 executes VMXON and frees it
5065 	 * when L1 executes VMXOFF or the vCPU is forced out of nested
5066 	 * operation.  VMXON faults if the CPU is already post-VMXON, so it
5067 	 * should be impossible to already have an allocated shadow VMCS.  KVM
5068 	 * doesn't support virtualization of VMCS shadowing, so vmcs01 should
5069 	 * always be the loaded VMCS.
5070 	 */
5071 	if (WARN_ON(loaded_vmcs != &vmx->vmcs01 || loaded_vmcs->shadow_vmcs))
5072 		return loaded_vmcs->shadow_vmcs;
5073 
5074 	loaded_vmcs->shadow_vmcs = alloc_vmcs(true);
5075 	if (loaded_vmcs->shadow_vmcs)
5076 		vmcs_clear(loaded_vmcs->shadow_vmcs);
5077 
5078 	return loaded_vmcs->shadow_vmcs;
5079 }
5080 
5081 static int enter_vmx_operation(struct kvm_vcpu *vcpu)
5082 {
5083 	struct vcpu_vmx *vmx = to_vmx(vcpu);
5084 	int r;
5085 
5086 	r = alloc_loaded_vmcs(&vmx->nested.vmcs02);
5087 	if (r < 0)
5088 		goto out_vmcs02;
5089 
5090 	vmx->nested.cached_vmcs12 = kzalloc(VMCS12_SIZE, GFP_KERNEL_ACCOUNT);
5091 	if (!vmx->nested.cached_vmcs12)
5092 		goto out_cached_vmcs12;
5093 
5094 	vmx->nested.shadow_vmcs12_cache.gpa = INVALID_GPA;
5095 	vmx->nested.cached_shadow_vmcs12 = kzalloc(VMCS12_SIZE, GFP_KERNEL_ACCOUNT);
5096 	if (!vmx->nested.cached_shadow_vmcs12)
5097 		goto out_cached_shadow_vmcs12;
5098 
5099 	if (enable_shadow_vmcs && !alloc_shadow_vmcs(vcpu))
5100 		goto out_shadow_vmcs;
5101 
5102 	hrtimer_init(&vmx->nested.preemption_timer, CLOCK_MONOTONIC,
5103 		     HRTIMER_MODE_ABS_PINNED);
5104 	vmx->nested.preemption_timer.function = vmx_preemption_timer_fn;
5105 
5106 	vmx->nested.vpid02 = allocate_vpid();
5107 
5108 	vmx->nested.vmcs02_initialized = false;
5109 	vmx->nested.vmxon = true;
5110 
5111 	if (vmx_pt_mode_is_host_guest()) {
5112 		vmx->pt_desc.guest.ctl = 0;
5113 		pt_update_intercept_for_msr(vcpu);
5114 	}
5115 
5116 	return 0;
5117 
5118 out_shadow_vmcs:
5119 	kfree(vmx->nested.cached_shadow_vmcs12);
5120 
5121 out_cached_shadow_vmcs12:
5122 	kfree(vmx->nested.cached_vmcs12);
5123 
5124 out_cached_vmcs12:
5125 	free_loaded_vmcs(&vmx->nested.vmcs02);
5126 
5127 out_vmcs02:
5128 	return -ENOMEM;
5129 }
5130 
5131 /* Emulate the VMXON instruction. */
5132 static int handle_vmxon(struct kvm_vcpu *vcpu)
5133 {
5134 	int ret;
5135 	gpa_t vmptr;
5136 	uint32_t revision;
5137 	struct vcpu_vmx *vmx = to_vmx(vcpu);
5138 	const u64 VMXON_NEEDED_FEATURES = FEAT_CTL_LOCKED
5139 		| FEAT_CTL_VMX_ENABLED_OUTSIDE_SMX;
5140 
5141 	/*
5142 	 * Manually check CR4.VMXE checks, KVM must force CR4.VMXE=1 to enter
5143 	 * the guest and so cannot rely on hardware to perform the check,
5144 	 * which has higher priority than VM-Exit (see Intel SDM's pseudocode
5145 	 * for VMXON).
5146 	 *
5147 	 * Rely on hardware for the other pre-VM-Exit checks, CR0.PE=1, !VM86
5148 	 * and !COMPATIBILITY modes.  For an unrestricted guest, KVM doesn't
5149 	 * force any of the relevant guest state.  For a restricted guest, KVM
5150 	 * does force CR0.PE=1, but only to also force VM86 in order to emulate
5151 	 * Real Mode, and so there's no need to check CR0.PE manually.
5152 	 */
5153 	if (!kvm_read_cr4_bits(vcpu, X86_CR4_VMXE)) {
5154 		kvm_queue_exception(vcpu, UD_VECTOR);
5155 		return 1;
5156 	}
5157 
5158 	/*
5159 	 * The CPL is checked for "not in VMX operation" and for "in VMX root",
5160 	 * and has higher priority than the VM-Fail due to being post-VMXON,
5161 	 * i.e. VMXON #GPs outside of VMX non-root if CPL!=0.  In VMX non-root,
5162 	 * VMXON causes VM-Exit and KVM unconditionally forwards VMXON VM-Exits
5163 	 * from L2 to L1, i.e. there's no need to check for the vCPU being in
5164 	 * VMX non-root.
5165 	 *
5166 	 * Forwarding the VM-Exit unconditionally, i.e. without performing the
5167 	 * #UD checks (see above), is functionally ok because KVM doesn't allow
5168 	 * L1 to run L2 without CR4.VMXE=0, and because KVM never modifies L2's
5169 	 * CR0 or CR4, i.e. it's L2's responsibility to emulate #UDs that are
5170 	 * missed by hardware due to shadowing CR0 and/or CR4.
5171 	 */
5172 	if (vmx_get_cpl(vcpu)) {
5173 		kvm_inject_gp(vcpu, 0);
5174 		return 1;
5175 	}
5176 
5177 	if (vmx->nested.vmxon)
5178 		return nested_vmx_fail(vcpu, VMXERR_VMXON_IN_VMX_ROOT_OPERATION);
5179 
5180 	/*
5181 	 * Invalid CR0/CR4 generates #GP.  These checks are performed if and
5182 	 * only if the vCPU isn't already in VMX operation, i.e. effectively
5183 	 * have lower priority than the VM-Fail above.
5184 	 */
5185 	if (!nested_host_cr0_valid(vcpu, kvm_read_cr0(vcpu)) ||
5186 	    !nested_host_cr4_valid(vcpu, kvm_read_cr4(vcpu))) {
5187 		kvm_inject_gp(vcpu, 0);
5188 		return 1;
5189 	}
5190 
5191 	if ((vmx->msr_ia32_feature_control & VMXON_NEEDED_FEATURES)
5192 			!= VMXON_NEEDED_FEATURES) {
5193 		kvm_inject_gp(vcpu, 0);
5194 		return 1;
5195 	}
5196 
5197 	if (nested_vmx_get_vmptr(vcpu, &vmptr, &ret))
5198 		return ret;
5199 
5200 	/*
5201 	 * SDM 3: 24.11.5
5202 	 * The first 4 bytes of VMXON region contain the supported
5203 	 * VMCS revision identifier
5204 	 *
5205 	 * Note - IA32_VMX_BASIC[48] will never be 1 for the nested case;
5206 	 * which replaces physical address width with 32
5207 	 */
5208 	if (!page_address_valid(vcpu, vmptr))
5209 		return nested_vmx_failInvalid(vcpu);
5210 
5211 	if (kvm_read_guest(vcpu->kvm, vmptr, &revision, sizeof(revision)) ||
5212 	    revision != VMCS12_REVISION)
5213 		return nested_vmx_failInvalid(vcpu);
5214 
5215 	vmx->nested.vmxon_ptr = vmptr;
5216 	ret = enter_vmx_operation(vcpu);
5217 	if (ret)
5218 		return ret;
5219 
5220 	return nested_vmx_succeed(vcpu);
5221 }
5222 
5223 static inline void nested_release_vmcs12(struct kvm_vcpu *vcpu)
5224 {
5225 	struct vcpu_vmx *vmx = to_vmx(vcpu);
5226 
5227 	if (vmx->nested.current_vmptr == INVALID_GPA)
5228 		return;
5229 
5230 	copy_vmcs02_to_vmcs12_rare(vcpu, get_vmcs12(vcpu));
5231 
5232 	if (enable_shadow_vmcs) {
5233 		/* copy to memory all shadowed fields in case
5234 		   they were modified */
5235 		copy_shadow_to_vmcs12(vmx);
5236 		vmx_disable_shadow_vmcs(vmx);
5237 	}
5238 	vmx->nested.posted_intr_nv = -1;
5239 
5240 	/* Flush VMCS12 to guest memory */
5241 	kvm_vcpu_write_guest_page(vcpu,
5242 				  vmx->nested.current_vmptr >> PAGE_SHIFT,
5243 				  vmx->nested.cached_vmcs12, 0, VMCS12_SIZE);
5244 
5245 	kvm_mmu_free_roots(vcpu->kvm, &vcpu->arch.guest_mmu, KVM_MMU_ROOTS_ALL);
5246 
5247 	vmx->nested.current_vmptr = INVALID_GPA;
5248 }
5249 
5250 /* Emulate the VMXOFF instruction */
5251 static int handle_vmxoff(struct kvm_vcpu *vcpu)
5252 {
5253 	if (!nested_vmx_check_permission(vcpu))
5254 		return 1;
5255 
5256 	free_nested(vcpu);
5257 
5258 	if (kvm_apic_has_pending_init_or_sipi(vcpu))
5259 		kvm_make_request(KVM_REQ_EVENT, vcpu);
5260 
5261 	return nested_vmx_succeed(vcpu);
5262 }
5263 
5264 /* Emulate the VMCLEAR instruction */
5265 static int handle_vmclear(struct kvm_vcpu *vcpu)
5266 {
5267 	struct vcpu_vmx *vmx = to_vmx(vcpu);
5268 	u32 zero = 0;
5269 	gpa_t vmptr;
5270 	int r;
5271 
5272 	if (!nested_vmx_check_permission(vcpu))
5273 		return 1;
5274 
5275 	if (nested_vmx_get_vmptr(vcpu, &vmptr, &r))
5276 		return r;
5277 
5278 	if (!page_address_valid(vcpu, vmptr))
5279 		return nested_vmx_fail(vcpu, VMXERR_VMCLEAR_INVALID_ADDRESS);
5280 
5281 	if (vmptr == vmx->nested.vmxon_ptr)
5282 		return nested_vmx_fail(vcpu, VMXERR_VMCLEAR_VMXON_POINTER);
5283 
5284 	/*
5285 	 * When Enlightened VMEntry is enabled on the calling CPU we treat
5286 	 * memory area pointer by vmptr as Enlightened VMCS (as there's no good
5287 	 * way to distinguish it from VMCS12) and we must not corrupt it by
5288 	 * writing to the non-existent 'launch_state' field. The area doesn't
5289 	 * have to be the currently active EVMCS on the calling CPU and there's
5290 	 * nothing KVM has to do to transition it from 'active' to 'non-active'
5291 	 * state. It is possible that the area will stay mapped as
5292 	 * vmx->nested.hv_evmcs but this shouldn't be a problem.
5293 	 */
5294 	if (likely(!guest_cpuid_has_evmcs(vcpu) ||
5295 		   !evmptr_is_valid(nested_get_evmptr(vcpu)))) {
5296 		if (vmptr == vmx->nested.current_vmptr)
5297 			nested_release_vmcs12(vcpu);
5298 
5299 		kvm_vcpu_write_guest(vcpu,
5300 				     vmptr + offsetof(struct vmcs12,
5301 						      launch_state),
5302 				     &zero, sizeof(zero));
5303 	} else if (vmx->nested.hv_evmcs && vmptr == vmx->nested.hv_evmcs_vmptr) {
5304 		nested_release_evmcs(vcpu);
5305 	}
5306 
5307 	return nested_vmx_succeed(vcpu);
5308 }
5309 
5310 /* Emulate the VMLAUNCH instruction */
5311 static int handle_vmlaunch(struct kvm_vcpu *vcpu)
5312 {
5313 	return nested_vmx_run(vcpu, true);
5314 }
5315 
5316 /* Emulate the VMRESUME instruction */
5317 static int handle_vmresume(struct kvm_vcpu *vcpu)
5318 {
5319 
5320 	return nested_vmx_run(vcpu, false);
5321 }
5322 
5323 static int handle_vmread(struct kvm_vcpu *vcpu)
5324 {
5325 	struct vmcs12 *vmcs12 = is_guest_mode(vcpu) ? get_shadow_vmcs12(vcpu)
5326 						    : get_vmcs12(vcpu);
5327 	unsigned long exit_qualification = vmx_get_exit_qual(vcpu);
5328 	u32 instr_info = vmcs_read32(VMX_INSTRUCTION_INFO);
5329 	struct vcpu_vmx *vmx = to_vmx(vcpu);
5330 	struct x86_exception e;
5331 	unsigned long field;
5332 	u64 value;
5333 	gva_t gva = 0;
5334 	short offset;
5335 	int len, r;
5336 
5337 	if (!nested_vmx_check_permission(vcpu))
5338 		return 1;
5339 
5340 	/* Decode instruction info and find the field to read */
5341 	field = kvm_register_read(vcpu, (((instr_info) >> 28) & 0xf));
5342 
5343 	if (!evmptr_is_valid(vmx->nested.hv_evmcs_vmptr)) {
5344 		/*
5345 		 * In VMX non-root operation, when the VMCS-link pointer is INVALID_GPA,
5346 		 * any VMREAD sets the ALU flags for VMfailInvalid.
5347 		 */
5348 		if (vmx->nested.current_vmptr == INVALID_GPA ||
5349 		    (is_guest_mode(vcpu) &&
5350 		     get_vmcs12(vcpu)->vmcs_link_pointer == INVALID_GPA))
5351 			return nested_vmx_failInvalid(vcpu);
5352 
5353 		offset = get_vmcs12_field_offset(field);
5354 		if (offset < 0)
5355 			return nested_vmx_fail(vcpu, VMXERR_UNSUPPORTED_VMCS_COMPONENT);
5356 
5357 		if (!is_guest_mode(vcpu) && is_vmcs12_ext_field(field))
5358 			copy_vmcs02_to_vmcs12_rare(vcpu, vmcs12);
5359 
5360 		/* Read the field, zero-extended to a u64 value */
5361 		value = vmcs12_read_any(vmcs12, field, offset);
5362 	} else {
5363 		/*
5364 		 * Hyper-V TLFS (as of 6.0b) explicitly states, that while an
5365 		 * enlightened VMCS is active VMREAD/VMWRITE instructions are
5366 		 * unsupported. Unfortunately, certain versions of Windows 11
5367 		 * don't comply with this requirement which is not enforced in
5368 		 * genuine Hyper-V. Allow VMREAD from an enlightened VMCS as a
5369 		 * workaround, as misbehaving guests will panic on VM-Fail.
5370 		 * Note, enlightened VMCS is incompatible with shadow VMCS so
5371 		 * all VMREADs from L2 should go to L1.
5372 		 */
5373 		if (WARN_ON_ONCE(is_guest_mode(vcpu)))
5374 			return nested_vmx_failInvalid(vcpu);
5375 
5376 		offset = evmcs_field_offset(field, NULL);
5377 		if (offset < 0)
5378 			return nested_vmx_fail(vcpu, VMXERR_UNSUPPORTED_VMCS_COMPONENT);
5379 
5380 		/* Read the field, zero-extended to a u64 value */
5381 		value = evmcs_read_any(vmx->nested.hv_evmcs, field, offset);
5382 	}
5383 
5384 	/*
5385 	 * Now copy part of this value to register or memory, as requested.
5386 	 * Note that the number of bits actually copied is 32 or 64 depending
5387 	 * on the guest's mode (32 or 64 bit), not on the given field's length.
5388 	 */
5389 	if (instr_info & BIT(10)) {
5390 		kvm_register_write(vcpu, (((instr_info) >> 3) & 0xf), value);
5391 	} else {
5392 		len = is_64_bit_mode(vcpu) ? 8 : 4;
5393 		if (get_vmx_mem_address(vcpu, exit_qualification,
5394 					instr_info, true, len, &gva))
5395 			return 1;
5396 		/* _system ok, nested_vmx_check_permission has verified cpl=0 */
5397 		r = kvm_write_guest_virt_system(vcpu, gva, &value, len, &e);
5398 		if (r != X86EMUL_CONTINUE)
5399 			return kvm_handle_memory_failure(vcpu, r, &e);
5400 	}
5401 
5402 	return nested_vmx_succeed(vcpu);
5403 }
5404 
5405 static bool is_shadow_field_rw(unsigned long field)
5406 {
5407 	switch (field) {
5408 #define SHADOW_FIELD_RW(x, y) case x:
5409 #include "vmcs_shadow_fields.h"
5410 		return true;
5411 	default:
5412 		break;
5413 	}
5414 	return false;
5415 }
5416 
5417 static bool is_shadow_field_ro(unsigned long field)
5418 {
5419 	switch (field) {
5420 #define SHADOW_FIELD_RO(x, y) case x:
5421 #include "vmcs_shadow_fields.h"
5422 		return true;
5423 	default:
5424 		break;
5425 	}
5426 	return false;
5427 }
5428 
5429 static int handle_vmwrite(struct kvm_vcpu *vcpu)
5430 {
5431 	struct vmcs12 *vmcs12 = is_guest_mode(vcpu) ? get_shadow_vmcs12(vcpu)
5432 						    : get_vmcs12(vcpu);
5433 	unsigned long exit_qualification = vmx_get_exit_qual(vcpu);
5434 	u32 instr_info = vmcs_read32(VMX_INSTRUCTION_INFO);
5435 	struct vcpu_vmx *vmx = to_vmx(vcpu);
5436 	struct x86_exception e;
5437 	unsigned long field;
5438 	short offset;
5439 	gva_t gva;
5440 	int len, r;
5441 
5442 	/*
5443 	 * The value to write might be 32 or 64 bits, depending on L1's long
5444 	 * mode, and eventually we need to write that into a field of several
5445 	 * possible lengths. The code below first zero-extends the value to 64
5446 	 * bit (value), and then copies only the appropriate number of
5447 	 * bits into the vmcs12 field.
5448 	 */
5449 	u64 value = 0;
5450 
5451 	if (!nested_vmx_check_permission(vcpu))
5452 		return 1;
5453 
5454 	/*
5455 	 * In VMX non-root operation, when the VMCS-link pointer is INVALID_GPA,
5456 	 * any VMWRITE sets the ALU flags for VMfailInvalid.
5457 	 */
5458 	if (vmx->nested.current_vmptr == INVALID_GPA ||
5459 	    (is_guest_mode(vcpu) &&
5460 	     get_vmcs12(vcpu)->vmcs_link_pointer == INVALID_GPA))
5461 		return nested_vmx_failInvalid(vcpu);
5462 
5463 	if (instr_info & BIT(10))
5464 		value = kvm_register_read(vcpu, (((instr_info) >> 3) & 0xf));
5465 	else {
5466 		len = is_64_bit_mode(vcpu) ? 8 : 4;
5467 		if (get_vmx_mem_address(vcpu, exit_qualification,
5468 					instr_info, false, len, &gva))
5469 			return 1;
5470 		r = kvm_read_guest_virt(vcpu, gva, &value, len, &e);
5471 		if (r != X86EMUL_CONTINUE)
5472 			return kvm_handle_memory_failure(vcpu, r, &e);
5473 	}
5474 
5475 	field = kvm_register_read(vcpu, (((instr_info) >> 28) & 0xf));
5476 
5477 	offset = get_vmcs12_field_offset(field);
5478 	if (offset < 0)
5479 		return nested_vmx_fail(vcpu, VMXERR_UNSUPPORTED_VMCS_COMPONENT);
5480 
5481 	/*
5482 	 * If the vCPU supports "VMWRITE to any supported field in the
5483 	 * VMCS," then the "read-only" fields are actually read/write.
5484 	 */
5485 	if (vmcs_field_readonly(field) &&
5486 	    !nested_cpu_has_vmwrite_any_field(vcpu))
5487 		return nested_vmx_fail(vcpu, VMXERR_VMWRITE_READ_ONLY_VMCS_COMPONENT);
5488 
5489 	/*
5490 	 * Ensure vmcs12 is up-to-date before any VMWRITE that dirties
5491 	 * vmcs12, else we may crush a field or consume a stale value.
5492 	 */
5493 	if (!is_guest_mode(vcpu) && !is_shadow_field_rw(field))
5494 		copy_vmcs02_to_vmcs12_rare(vcpu, vmcs12);
5495 
5496 	/*
5497 	 * Some Intel CPUs intentionally drop the reserved bits of the AR byte
5498 	 * fields on VMWRITE.  Emulate this behavior to ensure consistent KVM
5499 	 * behavior regardless of the underlying hardware, e.g. if an AR_BYTE
5500 	 * field is intercepted for VMWRITE but not VMREAD (in L1), then VMREAD
5501 	 * from L1 will return a different value than VMREAD from L2 (L1 sees
5502 	 * the stripped down value, L2 sees the full value as stored by KVM).
5503 	 */
5504 	if (field >= GUEST_ES_AR_BYTES && field <= GUEST_TR_AR_BYTES)
5505 		value &= 0x1f0ff;
5506 
5507 	vmcs12_write_any(vmcs12, field, offset, value);
5508 
5509 	/*
5510 	 * Do not track vmcs12 dirty-state if in guest-mode as we actually
5511 	 * dirty shadow vmcs12 instead of vmcs12.  Fields that can be updated
5512 	 * by L1 without a vmexit are always updated in the vmcs02, i.e. don't
5513 	 * "dirty" vmcs12, all others go down the prepare_vmcs02() slow path.
5514 	 */
5515 	if (!is_guest_mode(vcpu) && !is_shadow_field_rw(field)) {
5516 		/*
5517 		 * L1 can read these fields without exiting, ensure the
5518 		 * shadow VMCS is up-to-date.
5519 		 */
5520 		if (enable_shadow_vmcs && is_shadow_field_ro(field)) {
5521 			preempt_disable();
5522 			vmcs_load(vmx->vmcs01.shadow_vmcs);
5523 
5524 			__vmcs_writel(field, value);
5525 
5526 			vmcs_clear(vmx->vmcs01.shadow_vmcs);
5527 			vmcs_load(vmx->loaded_vmcs->vmcs);
5528 			preempt_enable();
5529 		}
5530 		vmx->nested.dirty_vmcs12 = true;
5531 	}
5532 
5533 	return nested_vmx_succeed(vcpu);
5534 }
5535 
5536 static void set_current_vmptr(struct vcpu_vmx *vmx, gpa_t vmptr)
5537 {
5538 	vmx->nested.current_vmptr = vmptr;
5539 	if (enable_shadow_vmcs) {
5540 		secondary_exec_controls_setbit(vmx, SECONDARY_EXEC_SHADOW_VMCS);
5541 		vmcs_write64(VMCS_LINK_POINTER,
5542 			     __pa(vmx->vmcs01.shadow_vmcs));
5543 		vmx->nested.need_vmcs12_to_shadow_sync = true;
5544 	}
5545 	vmx->nested.dirty_vmcs12 = true;
5546 	vmx->nested.force_msr_bitmap_recalc = true;
5547 }
5548 
5549 /* Emulate the VMPTRLD instruction */
5550 static int handle_vmptrld(struct kvm_vcpu *vcpu)
5551 {
5552 	struct vcpu_vmx *vmx = to_vmx(vcpu);
5553 	gpa_t vmptr;
5554 	int r;
5555 
5556 	if (!nested_vmx_check_permission(vcpu))
5557 		return 1;
5558 
5559 	if (nested_vmx_get_vmptr(vcpu, &vmptr, &r))
5560 		return r;
5561 
5562 	if (!page_address_valid(vcpu, vmptr))
5563 		return nested_vmx_fail(vcpu, VMXERR_VMPTRLD_INVALID_ADDRESS);
5564 
5565 	if (vmptr == vmx->nested.vmxon_ptr)
5566 		return nested_vmx_fail(vcpu, VMXERR_VMPTRLD_VMXON_POINTER);
5567 
5568 	/* Forbid normal VMPTRLD if Enlightened version was used */
5569 	if (evmptr_is_valid(vmx->nested.hv_evmcs_vmptr))
5570 		return 1;
5571 
5572 	if (vmx->nested.current_vmptr != vmptr) {
5573 		struct gfn_to_hva_cache *ghc = &vmx->nested.vmcs12_cache;
5574 		struct vmcs_hdr hdr;
5575 
5576 		if (kvm_gfn_to_hva_cache_init(vcpu->kvm, ghc, vmptr, VMCS12_SIZE)) {
5577 			/*
5578 			 * Reads from an unbacked page return all 1s,
5579 			 * which means that the 32 bits located at the
5580 			 * given physical address won't match the required
5581 			 * VMCS12_REVISION identifier.
5582 			 */
5583 			return nested_vmx_fail(vcpu,
5584 				VMXERR_VMPTRLD_INCORRECT_VMCS_REVISION_ID);
5585 		}
5586 
5587 		if (kvm_read_guest_offset_cached(vcpu->kvm, ghc, &hdr,
5588 						 offsetof(struct vmcs12, hdr),
5589 						 sizeof(hdr))) {
5590 			return nested_vmx_fail(vcpu,
5591 				VMXERR_VMPTRLD_INCORRECT_VMCS_REVISION_ID);
5592 		}
5593 
5594 		if (hdr.revision_id != VMCS12_REVISION ||
5595 		    (hdr.shadow_vmcs &&
5596 		     !nested_cpu_has_vmx_shadow_vmcs(vcpu))) {
5597 			return nested_vmx_fail(vcpu,
5598 				VMXERR_VMPTRLD_INCORRECT_VMCS_REVISION_ID);
5599 		}
5600 
5601 		nested_release_vmcs12(vcpu);
5602 
5603 		/*
5604 		 * Load VMCS12 from guest memory since it is not already
5605 		 * cached.
5606 		 */
5607 		if (kvm_read_guest_cached(vcpu->kvm, ghc, vmx->nested.cached_vmcs12,
5608 					  VMCS12_SIZE)) {
5609 			return nested_vmx_fail(vcpu,
5610 				VMXERR_VMPTRLD_INCORRECT_VMCS_REVISION_ID);
5611 		}
5612 
5613 		set_current_vmptr(vmx, vmptr);
5614 	}
5615 
5616 	return nested_vmx_succeed(vcpu);
5617 }
5618 
5619 /* Emulate the VMPTRST instruction */
5620 static int handle_vmptrst(struct kvm_vcpu *vcpu)
5621 {
5622 	unsigned long exit_qual = vmx_get_exit_qual(vcpu);
5623 	u32 instr_info = vmcs_read32(VMX_INSTRUCTION_INFO);
5624 	gpa_t current_vmptr = to_vmx(vcpu)->nested.current_vmptr;
5625 	struct x86_exception e;
5626 	gva_t gva;
5627 	int r;
5628 
5629 	if (!nested_vmx_check_permission(vcpu))
5630 		return 1;
5631 
5632 	if (unlikely(evmptr_is_valid(to_vmx(vcpu)->nested.hv_evmcs_vmptr)))
5633 		return 1;
5634 
5635 	if (get_vmx_mem_address(vcpu, exit_qual, instr_info,
5636 				true, sizeof(gpa_t), &gva))
5637 		return 1;
5638 	/* *_system ok, nested_vmx_check_permission has verified cpl=0 */
5639 	r = kvm_write_guest_virt_system(vcpu, gva, (void *)&current_vmptr,
5640 					sizeof(gpa_t), &e);
5641 	if (r != X86EMUL_CONTINUE)
5642 		return kvm_handle_memory_failure(vcpu, r, &e);
5643 
5644 	return nested_vmx_succeed(vcpu);
5645 }
5646 
5647 /* Emulate the INVEPT instruction */
5648 static int handle_invept(struct kvm_vcpu *vcpu)
5649 {
5650 	struct vcpu_vmx *vmx = to_vmx(vcpu);
5651 	u32 vmx_instruction_info, types;
5652 	unsigned long type, roots_to_free;
5653 	struct kvm_mmu *mmu;
5654 	gva_t gva;
5655 	struct x86_exception e;
5656 	struct {
5657 		u64 eptp, gpa;
5658 	} operand;
5659 	int i, r, gpr_index;
5660 
5661 	if (!(vmx->nested.msrs.secondary_ctls_high &
5662 	      SECONDARY_EXEC_ENABLE_EPT) ||
5663 	    !(vmx->nested.msrs.ept_caps & VMX_EPT_INVEPT_BIT)) {
5664 		kvm_queue_exception(vcpu, UD_VECTOR);
5665 		return 1;
5666 	}
5667 
5668 	if (!nested_vmx_check_permission(vcpu))
5669 		return 1;
5670 
5671 	vmx_instruction_info = vmcs_read32(VMX_INSTRUCTION_INFO);
5672 	gpr_index = vmx_get_instr_info_reg2(vmx_instruction_info);
5673 	type = kvm_register_read(vcpu, gpr_index);
5674 
5675 	types = (vmx->nested.msrs.ept_caps >> VMX_EPT_EXTENT_SHIFT) & 6;
5676 
5677 	if (type >= 32 || !(types & (1 << type)))
5678 		return nested_vmx_fail(vcpu, VMXERR_INVALID_OPERAND_TO_INVEPT_INVVPID);
5679 
5680 	/* According to the Intel VMX instruction reference, the memory
5681 	 * operand is read even if it isn't needed (e.g., for type==global)
5682 	 */
5683 	if (get_vmx_mem_address(vcpu, vmx_get_exit_qual(vcpu),
5684 			vmx_instruction_info, false, sizeof(operand), &gva))
5685 		return 1;
5686 	r = kvm_read_guest_virt(vcpu, gva, &operand, sizeof(operand), &e);
5687 	if (r != X86EMUL_CONTINUE)
5688 		return kvm_handle_memory_failure(vcpu, r, &e);
5689 
5690 	/*
5691 	 * Nested EPT roots are always held through guest_mmu,
5692 	 * not root_mmu.
5693 	 */
5694 	mmu = &vcpu->arch.guest_mmu;
5695 
5696 	switch (type) {
5697 	case VMX_EPT_EXTENT_CONTEXT:
5698 		if (!nested_vmx_check_eptp(vcpu, operand.eptp))
5699 			return nested_vmx_fail(vcpu,
5700 				VMXERR_INVALID_OPERAND_TO_INVEPT_INVVPID);
5701 
5702 		roots_to_free = 0;
5703 		if (nested_ept_root_matches(mmu->root.hpa, mmu->root.pgd,
5704 					    operand.eptp))
5705 			roots_to_free |= KVM_MMU_ROOT_CURRENT;
5706 
5707 		for (i = 0; i < KVM_MMU_NUM_PREV_ROOTS; i++) {
5708 			if (nested_ept_root_matches(mmu->prev_roots[i].hpa,
5709 						    mmu->prev_roots[i].pgd,
5710 						    operand.eptp))
5711 				roots_to_free |= KVM_MMU_ROOT_PREVIOUS(i);
5712 		}
5713 		break;
5714 	case VMX_EPT_EXTENT_GLOBAL:
5715 		roots_to_free = KVM_MMU_ROOTS_ALL;
5716 		break;
5717 	default:
5718 		BUG();
5719 		break;
5720 	}
5721 
5722 	if (roots_to_free)
5723 		kvm_mmu_free_roots(vcpu->kvm, mmu, roots_to_free);
5724 
5725 	return nested_vmx_succeed(vcpu);
5726 }
5727 
5728 static int handle_invvpid(struct kvm_vcpu *vcpu)
5729 {
5730 	struct vcpu_vmx *vmx = to_vmx(vcpu);
5731 	u32 vmx_instruction_info;
5732 	unsigned long type, types;
5733 	gva_t gva;
5734 	struct x86_exception e;
5735 	struct {
5736 		u64 vpid;
5737 		u64 gla;
5738 	} operand;
5739 	u16 vpid02;
5740 	int r, gpr_index;
5741 
5742 	if (!(vmx->nested.msrs.secondary_ctls_high &
5743 	      SECONDARY_EXEC_ENABLE_VPID) ||
5744 			!(vmx->nested.msrs.vpid_caps & VMX_VPID_INVVPID_BIT)) {
5745 		kvm_queue_exception(vcpu, UD_VECTOR);
5746 		return 1;
5747 	}
5748 
5749 	if (!nested_vmx_check_permission(vcpu))
5750 		return 1;
5751 
5752 	vmx_instruction_info = vmcs_read32(VMX_INSTRUCTION_INFO);
5753 	gpr_index = vmx_get_instr_info_reg2(vmx_instruction_info);
5754 	type = kvm_register_read(vcpu, gpr_index);
5755 
5756 	types = (vmx->nested.msrs.vpid_caps &
5757 			VMX_VPID_EXTENT_SUPPORTED_MASK) >> 8;
5758 
5759 	if (type >= 32 || !(types & (1 << type)))
5760 		return nested_vmx_fail(vcpu,
5761 			VMXERR_INVALID_OPERAND_TO_INVEPT_INVVPID);
5762 
5763 	/* according to the intel vmx instruction reference, the memory
5764 	 * operand is read even if it isn't needed (e.g., for type==global)
5765 	 */
5766 	if (get_vmx_mem_address(vcpu, vmx_get_exit_qual(vcpu),
5767 			vmx_instruction_info, false, sizeof(operand), &gva))
5768 		return 1;
5769 	r = kvm_read_guest_virt(vcpu, gva, &operand, sizeof(operand), &e);
5770 	if (r != X86EMUL_CONTINUE)
5771 		return kvm_handle_memory_failure(vcpu, r, &e);
5772 
5773 	if (operand.vpid >> 16)
5774 		return nested_vmx_fail(vcpu,
5775 			VMXERR_INVALID_OPERAND_TO_INVEPT_INVVPID);
5776 
5777 	vpid02 = nested_get_vpid02(vcpu);
5778 	switch (type) {
5779 	case VMX_VPID_EXTENT_INDIVIDUAL_ADDR:
5780 		if (!operand.vpid ||
5781 		    is_noncanonical_address(operand.gla, vcpu))
5782 			return nested_vmx_fail(vcpu,
5783 				VMXERR_INVALID_OPERAND_TO_INVEPT_INVVPID);
5784 		vpid_sync_vcpu_addr(vpid02, operand.gla);
5785 		break;
5786 	case VMX_VPID_EXTENT_SINGLE_CONTEXT:
5787 	case VMX_VPID_EXTENT_SINGLE_NON_GLOBAL:
5788 		if (!operand.vpid)
5789 			return nested_vmx_fail(vcpu,
5790 				VMXERR_INVALID_OPERAND_TO_INVEPT_INVVPID);
5791 		vpid_sync_context(vpid02);
5792 		break;
5793 	case VMX_VPID_EXTENT_ALL_CONTEXT:
5794 		vpid_sync_context(vpid02);
5795 		break;
5796 	default:
5797 		WARN_ON_ONCE(1);
5798 		return kvm_skip_emulated_instruction(vcpu);
5799 	}
5800 
5801 	/*
5802 	 * Sync the shadow page tables if EPT is disabled, L1 is invalidating
5803 	 * linear mappings for L2 (tagged with L2's VPID).  Free all guest
5804 	 * roots as VPIDs are not tracked in the MMU role.
5805 	 *
5806 	 * Note, this operates on root_mmu, not guest_mmu, as L1 and L2 share
5807 	 * an MMU when EPT is disabled.
5808 	 *
5809 	 * TODO: sync only the affected SPTEs for INVDIVIDUAL_ADDR.
5810 	 */
5811 	if (!enable_ept)
5812 		kvm_mmu_free_guest_mode_roots(vcpu->kvm, &vcpu->arch.root_mmu);
5813 
5814 	return nested_vmx_succeed(vcpu);
5815 }
5816 
5817 static int nested_vmx_eptp_switching(struct kvm_vcpu *vcpu,
5818 				     struct vmcs12 *vmcs12)
5819 {
5820 	u32 index = kvm_rcx_read(vcpu);
5821 	u64 new_eptp;
5822 
5823 	if (WARN_ON_ONCE(!nested_cpu_has_ept(vmcs12)))
5824 		return 1;
5825 	if (index >= VMFUNC_EPTP_ENTRIES)
5826 		return 1;
5827 
5828 	if (kvm_vcpu_read_guest_page(vcpu, vmcs12->eptp_list_address >> PAGE_SHIFT,
5829 				     &new_eptp, index * 8, 8))
5830 		return 1;
5831 
5832 	/*
5833 	 * If the (L2) guest does a vmfunc to the currently
5834 	 * active ept pointer, we don't have to do anything else
5835 	 */
5836 	if (vmcs12->ept_pointer != new_eptp) {
5837 		if (!nested_vmx_check_eptp(vcpu, new_eptp))
5838 			return 1;
5839 
5840 		vmcs12->ept_pointer = new_eptp;
5841 		nested_ept_new_eptp(vcpu);
5842 
5843 		if (!nested_cpu_has_vpid(vmcs12))
5844 			kvm_make_request(KVM_REQ_TLB_FLUSH_GUEST, vcpu);
5845 	}
5846 
5847 	return 0;
5848 }
5849 
5850 static int handle_vmfunc(struct kvm_vcpu *vcpu)
5851 {
5852 	struct vcpu_vmx *vmx = to_vmx(vcpu);
5853 	struct vmcs12 *vmcs12;
5854 	u32 function = kvm_rax_read(vcpu);
5855 
5856 	/*
5857 	 * VMFUNC is only supported for nested guests, but we always enable the
5858 	 * secondary control for simplicity; for non-nested mode, fake that we
5859 	 * didn't by injecting #UD.
5860 	 */
5861 	if (!is_guest_mode(vcpu)) {
5862 		kvm_queue_exception(vcpu, UD_VECTOR);
5863 		return 1;
5864 	}
5865 
5866 	vmcs12 = get_vmcs12(vcpu);
5867 
5868 	/*
5869 	 * #UD on out-of-bounds function has priority over VM-Exit, and VMFUNC
5870 	 * is enabled in vmcs02 if and only if it's enabled in vmcs12.
5871 	 */
5872 	if (WARN_ON_ONCE((function > 63) || !nested_cpu_has_vmfunc(vmcs12))) {
5873 		kvm_queue_exception(vcpu, UD_VECTOR);
5874 		return 1;
5875 	}
5876 
5877 	if (!(vmcs12->vm_function_control & BIT_ULL(function)))
5878 		goto fail;
5879 
5880 	switch (function) {
5881 	case 0:
5882 		if (nested_vmx_eptp_switching(vcpu, vmcs12))
5883 			goto fail;
5884 		break;
5885 	default:
5886 		goto fail;
5887 	}
5888 	return kvm_skip_emulated_instruction(vcpu);
5889 
5890 fail:
5891 	/*
5892 	 * This is effectively a reflected VM-Exit, as opposed to a synthesized
5893 	 * nested VM-Exit.  Pass the original exit reason, i.e. don't hardcode
5894 	 * EXIT_REASON_VMFUNC as the exit reason.
5895 	 */
5896 	nested_vmx_vmexit(vcpu, vmx->exit_reason.full,
5897 			  vmx_get_intr_info(vcpu),
5898 			  vmx_get_exit_qual(vcpu));
5899 	return 1;
5900 }
5901 
5902 /*
5903  * Return true if an IO instruction with the specified port and size should cause
5904  * a VM-exit into L1.
5905  */
5906 bool nested_vmx_check_io_bitmaps(struct kvm_vcpu *vcpu, unsigned int port,
5907 				 int size)
5908 {
5909 	struct vmcs12 *vmcs12 = get_vmcs12(vcpu);
5910 	gpa_t bitmap, last_bitmap;
5911 	u8 b;
5912 
5913 	last_bitmap = INVALID_GPA;
5914 	b = -1;
5915 
5916 	while (size > 0) {
5917 		if (port < 0x8000)
5918 			bitmap = vmcs12->io_bitmap_a;
5919 		else if (port < 0x10000)
5920 			bitmap = vmcs12->io_bitmap_b;
5921 		else
5922 			return true;
5923 		bitmap += (port & 0x7fff) / 8;
5924 
5925 		if (last_bitmap != bitmap)
5926 			if (kvm_vcpu_read_guest(vcpu, bitmap, &b, 1))
5927 				return true;
5928 		if (b & (1 << (port & 7)))
5929 			return true;
5930 
5931 		port++;
5932 		size--;
5933 		last_bitmap = bitmap;
5934 	}
5935 
5936 	return false;
5937 }
5938 
5939 static bool nested_vmx_exit_handled_io(struct kvm_vcpu *vcpu,
5940 				       struct vmcs12 *vmcs12)
5941 {
5942 	unsigned long exit_qualification;
5943 	unsigned short port;
5944 	int size;
5945 
5946 	if (!nested_cpu_has(vmcs12, CPU_BASED_USE_IO_BITMAPS))
5947 		return nested_cpu_has(vmcs12, CPU_BASED_UNCOND_IO_EXITING);
5948 
5949 	exit_qualification = vmx_get_exit_qual(vcpu);
5950 
5951 	port = exit_qualification >> 16;
5952 	size = (exit_qualification & 7) + 1;
5953 
5954 	return nested_vmx_check_io_bitmaps(vcpu, port, size);
5955 }
5956 
5957 /*
5958  * Return 1 if we should exit from L2 to L1 to handle an MSR access,
5959  * rather than handle it ourselves in L0. I.e., check whether L1 expressed
5960  * disinterest in the current event (read or write a specific MSR) by using an
5961  * MSR bitmap. This may be the case even when L0 doesn't use MSR bitmaps.
5962  */
5963 static bool nested_vmx_exit_handled_msr(struct kvm_vcpu *vcpu,
5964 					struct vmcs12 *vmcs12,
5965 					union vmx_exit_reason exit_reason)
5966 {
5967 	u32 msr_index = kvm_rcx_read(vcpu);
5968 	gpa_t bitmap;
5969 
5970 	if (!nested_cpu_has(vmcs12, CPU_BASED_USE_MSR_BITMAPS))
5971 		return true;
5972 
5973 	/*
5974 	 * The MSR_BITMAP page is divided into four 1024-byte bitmaps,
5975 	 * for the four combinations of read/write and low/high MSR numbers.
5976 	 * First we need to figure out which of the four to use:
5977 	 */
5978 	bitmap = vmcs12->msr_bitmap;
5979 	if (exit_reason.basic == EXIT_REASON_MSR_WRITE)
5980 		bitmap += 2048;
5981 	if (msr_index >= 0xc0000000) {
5982 		msr_index -= 0xc0000000;
5983 		bitmap += 1024;
5984 	}
5985 
5986 	/* Then read the msr_index'th bit from this bitmap: */
5987 	if (msr_index < 1024*8) {
5988 		unsigned char b;
5989 		if (kvm_vcpu_read_guest(vcpu, bitmap + msr_index/8, &b, 1))
5990 			return true;
5991 		return 1 & (b >> (msr_index & 7));
5992 	} else
5993 		return true; /* let L1 handle the wrong parameter */
5994 }
5995 
5996 /*
5997  * Return 1 if we should exit from L2 to L1 to handle a CR access exit,
5998  * rather than handle it ourselves in L0. I.e., check if L1 wanted to
5999  * intercept (via guest_host_mask etc.) the current event.
6000  */
6001 static bool nested_vmx_exit_handled_cr(struct kvm_vcpu *vcpu,
6002 	struct vmcs12 *vmcs12)
6003 {
6004 	unsigned long exit_qualification = vmx_get_exit_qual(vcpu);
6005 	int cr = exit_qualification & 15;
6006 	int reg;
6007 	unsigned long val;
6008 
6009 	switch ((exit_qualification >> 4) & 3) {
6010 	case 0: /* mov to cr */
6011 		reg = (exit_qualification >> 8) & 15;
6012 		val = kvm_register_read(vcpu, reg);
6013 		switch (cr) {
6014 		case 0:
6015 			if (vmcs12->cr0_guest_host_mask &
6016 			    (val ^ vmcs12->cr0_read_shadow))
6017 				return true;
6018 			break;
6019 		case 3:
6020 			if (nested_cpu_has(vmcs12, CPU_BASED_CR3_LOAD_EXITING))
6021 				return true;
6022 			break;
6023 		case 4:
6024 			if (vmcs12->cr4_guest_host_mask &
6025 			    (vmcs12->cr4_read_shadow ^ val))
6026 				return true;
6027 			break;
6028 		case 8:
6029 			if (nested_cpu_has(vmcs12, CPU_BASED_CR8_LOAD_EXITING))
6030 				return true;
6031 			break;
6032 		}
6033 		break;
6034 	case 2: /* clts */
6035 		if ((vmcs12->cr0_guest_host_mask & X86_CR0_TS) &&
6036 		    (vmcs12->cr0_read_shadow & X86_CR0_TS))
6037 			return true;
6038 		break;
6039 	case 1: /* mov from cr */
6040 		switch (cr) {
6041 		case 3:
6042 			if (vmcs12->cpu_based_vm_exec_control &
6043 			    CPU_BASED_CR3_STORE_EXITING)
6044 				return true;
6045 			break;
6046 		case 8:
6047 			if (vmcs12->cpu_based_vm_exec_control &
6048 			    CPU_BASED_CR8_STORE_EXITING)
6049 				return true;
6050 			break;
6051 		}
6052 		break;
6053 	case 3: /* lmsw */
6054 		/*
6055 		 * lmsw can change bits 1..3 of cr0, and only set bit 0 of
6056 		 * cr0. Other attempted changes are ignored, with no exit.
6057 		 */
6058 		val = (exit_qualification >> LMSW_SOURCE_DATA_SHIFT) & 0x0f;
6059 		if (vmcs12->cr0_guest_host_mask & 0xe &
6060 		    (val ^ vmcs12->cr0_read_shadow))
6061 			return true;
6062 		if ((vmcs12->cr0_guest_host_mask & 0x1) &&
6063 		    !(vmcs12->cr0_read_shadow & 0x1) &&
6064 		    (val & 0x1))
6065 			return true;
6066 		break;
6067 	}
6068 	return false;
6069 }
6070 
6071 static bool nested_vmx_exit_handled_encls(struct kvm_vcpu *vcpu,
6072 					  struct vmcs12 *vmcs12)
6073 {
6074 	u32 encls_leaf;
6075 
6076 	if (!guest_cpuid_has(vcpu, X86_FEATURE_SGX) ||
6077 	    !nested_cpu_has2(vmcs12, SECONDARY_EXEC_ENCLS_EXITING))
6078 		return false;
6079 
6080 	encls_leaf = kvm_rax_read(vcpu);
6081 	if (encls_leaf > 62)
6082 		encls_leaf = 63;
6083 	return vmcs12->encls_exiting_bitmap & BIT_ULL(encls_leaf);
6084 }
6085 
6086 static bool nested_vmx_exit_handled_vmcs_access(struct kvm_vcpu *vcpu,
6087 	struct vmcs12 *vmcs12, gpa_t bitmap)
6088 {
6089 	u32 vmx_instruction_info;
6090 	unsigned long field;
6091 	u8 b;
6092 
6093 	if (!nested_cpu_has_shadow_vmcs(vmcs12))
6094 		return true;
6095 
6096 	/* Decode instruction info and find the field to access */
6097 	vmx_instruction_info = vmcs_read32(VMX_INSTRUCTION_INFO);
6098 	field = kvm_register_read(vcpu, (((vmx_instruction_info) >> 28) & 0xf));
6099 
6100 	/* Out-of-range fields always cause a VM exit from L2 to L1 */
6101 	if (field >> 15)
6102 		return true;
6103 
6104 	if (kvm_vcpu_read_guest(vcpu, bitmap + field/8, &b, 1))
6105 		return true;
6106 
6107 	return 1 & (b >> (field & 7));
6108 }
6109 
6110 static bool nested_vmx_exit_handled_mtf(struct vmcs12 *vmcs12)
6111 {
6112 	u32 entry_intr_info = vmcs12->vm_entry_intr_info_field;
6113 
6114 	if (nested_cpu_has_mtf(vmcs12))
6115 		return true;
6116 
6117 	/*
6118 	 * An MTF VM-exit may be injected into the guest by setting the
6119 	 * interruption-type to 7 (other event) and the vector field to 0. Such
6120 	 * is the case regardless of the 'monitor trap flag' VM-execution
6121 	 * control.
6122 	 */
6123 	return entry_intr_info == (INTR_INFO_VALID_MASK
6124 				   | INTR_TYPE_OTHER_EVENT);
6125 }
6126 
6127 /*
6128  * Return true if L0 wants to handle an exit from L2 regardless of whether or not
6129  * L1 wants the exit.  Only call this when in is_guest_mode (L2).
6130  */
6131 static bool nested_vmx_l0_wants_exit(struct kvm_vcpu *vcpu,
6132 				     union vmx_exit_reason exit_reason)
6133 {
6134 	u32 intr_info;
6135 
6136 	switch ((u16)exit_reason.basic) {
6137 	case EXIT_REASON_EXCEPTION_NMI:
6138 		intr_info = vmx_get_intr_info(vcpu);
6139 		if (is_nmi(intr_info))
6140 			return true;
6141 		else if (is_page_fault(intr_info))
6142 			return vcpu->arch.apf.host_apf_flags ||
6143 			       vmx_need_pf_intercept(vcpu);
6144 		else if (is_debug(intr_info) &&
6145 			 vcpu->guest_debug &
6146 			 (KVM_GUESTDBG_SINGLESTEP | KVM_GUESTDBG_USE_HW_BP))
6147 			return true;
6148 		else if (is_breakpoint(intr_info) &&
6149 			 vcpu->guest_debug & KVM_GUESTDBG_USE_SW_BP)
6150 			return true;
6151 		else if (is_alignment_check(intr_info) &&
6152 			 !vmx_guest_inject_ac(vcpu))
6153 			return true;
6154 		return false;
6155 	case EXIT_REASON_EXTERNAL_INTERRUPT:
6156 		return true;
6157 	case EXIT_REASON_MCE_DURING_VMENTRY:
6158 		return true;
6159 	case EXIT_REASON_EPT_VIOLATION:
6160 		/*
6161 		 * L0 always deals with the EPT violation. If nested EPT is
6162 		 * used, and the nested mmu code discovers that the address is
6163 		 * missing in the guest EPT table (EPT12), the EPT violation
6164 		 * will be injected with nested_ept_inject_page_fault()
6165 		 */
6166 		return true;
6167 	case EXIT_REASON_EPT_MISCONFIG:
6168 		/*
6169 		 * L2 never uses directly L1's EPT, but rather L0's own EPT
6170 		 * table (shadow on EPT) or a merged EPT table that L0 built
6171 		 * (EPT on EPT). So any problems with the structure of the
6172 		 * table is L0's fault.
6173 		 */
6174 		return true;
6175 	case EXIT_REASON_PREEMPTION_TIMER:
6176 		return true;
6177 	case EXIT_REASON_PML_FULL:
6178 		/*
6179 		 * PML is emulated for an L1 VMM and should never be enabled in
6180 		 * vmcs02, always "handle" PML_FULL by exiting to userspace.
6181 		 */
6182 		return true;
6183 	case EXIT_REASON_VMFUNC:
6184 		/* VM functions are emulated through L2->L0 vmexits. */
6185 		return true;
6186 	case EXIT_REASON_BUS_LOCK:
6187 		/*
6188 		 * At present, bus lock VM exit is never exposed to L1.
6189 		 * Handle L2's bus locks in L0 directly.
6190 		 */
6191 		return true;
6192 	case EXIT_REASON_VMCALL:
6193 		/* Hyper-V L2 TLB flush hypercall is handled by L0 */
6194 		return guest_hv_cpuid_has_l2_tlb_flush(vcpu) &&
6195 			nested_evmcs_l2_tlb_flush_enabled(vcpu) &&
6196 			kvm_hv_is_tlb_flush_hcall(vcpu);
6197 	default:
6198 		break;
6199 	}
6200 	return false;
6201 }
6202 
6203 /*
6204  * Return 1 if L1 wants to intercept an exit from L2.  Only call this when in
6205  * is_guest_mode (L2).
6206  */
6207 static bool nested_vmx_l1_wants_exit(struct kvm_vcpu *vcpu,
6208 				     union vmx_exit_reason exit_reason)
6209 {
6210 	struct vmcs12 *vmcs12 = get_vmcs12(vcpu);
6211 	u32 intr_info;
6212 
6213 	switch ((u16)exit_reason.basic) {
6214 	case EXIT_REASON_EXCEPTION_NMI:
6215 		intr_info = vmx_get_intr_info(vcpu);
6216 		if (is_nmi(intr_info))
6217 			return true;
6218 		else if (is_page_fault(intr_info))
6219 			return true;
6220 		return vmcs12->exception_bitmap &
6221 				(1u << (intr_info & INTR_INFO_VECTOR_MASK));
6222 	case EXIT_REASON_EXTERNAL_INTERRUPT:
6223 		return nested_exit_on_intr(vcpu);
6224 	case EXIT_REASON_TRIPLE_FAULT:
6225 		return true;
6226 	case EXIT_REASON_INTERRUPT_WINDOW:
6227 		return nested_cpu_has(vmcs12, CPU_BASED_INTR_WINDOW_EXITING);
6228 	case EXIT_REASON_NMI_WINDOW:
6229 		return nested_cpu_has(vmcs12, CPU_BASED_NMI_WINDOW_EXITING);
6230 	case EXIT_REASON_TASK_SWITCH:
6231 		return true;
6232 	case EXIT_REASON_CPUID:
6233 		return true;
6234 	case EXIT_REASON_HLT:
6235 		return nested_cpu_has(vmcs12, CPU_BASED_HLT_EXITING);
6236 	case EXIT_REASON_INVD:
6237 		return true;
6238 	case EXIT_REASON_INVLPG:
6239 		return nested_cpu_has(vmcs12, CPU_BASED_INVLPG_EXITING);
6240 	case EXIT_REASON_RDPMC:
6241 		return nested_cpu_has(vmcs12, CPU_BASED_RDPMC_EXITING);
6242 	case EXIT_REASON_RDRAND:
6243 		return nested_cpu_has2(vmcs12, SECONDARY_EXEC_RDRAND_EXITING);
6244 	case EXIT_REASON_RDSEED:
6245 		return nested_cpu_has2(vmcs12, SECONDARY_EXEC_RDSEED_EXITING);
6246 	case EXIT_REASON_RDTSC: case EXIT_REASON_RDTSCP:
6247 		return nested_cpu_has(vmcs12, CPU_BASED_RDTSC_EXITING);
6248 	case EXIT_REASON_VMREAD:
6249 		return nested_vmx_exit_handled_vmcs_access(vcpu, vmcs12,
6250 			vmcs12->vmread_bitmap);
6251 	case EXIT_REASON_VMWRITE:
6252 		return nested_vmx_exit_handled_vmcs_access(vcpu, vmcs12,
6253 			vmcs12->vmwrite_bitmap);
6254 	case EXIT_REASON_VMCALL: case EXIT_REASON_VMCLEAR:
6255 	case EXIT_REASON_VMLAUNCH: case EXIT_REASON_VMPTRLD:
6256 	case EXIT_REASON_VMPTRST: case EXIT_REASON_VMRESUME:
6257 	case EXIT_REASON_VMOFF: case EXIT_REASON_VMON:
6258 	case EXIT_REASON_INVEPT: case EXIT_REASON_INVVPID:
6259 		/*
6260 		 * VMX instructions trap unconditionally. This allows L1 to
6261 		 * emulate them for its L2 guest, i.e., allows 3-level nesting!
6262 		 */
6263 		return true;
6264 	case EXIT_REASON_CR_ACCESS:
6265 		return nested_vmx_exit_handled_cr(vcpu, vmcs12);
6266 	case EXIT_REASON_DR_ACCESS:
6267 		return nested_cpu_has(vmcs12, CPU_BASED_MOV_DR_EXITING);
6268 	case EXIT_REASON_IO_INSTRUCTION:
6269 		return nested_vmx_exit_handled_io(vcpu, vmcs12);
6270 	case EXIT_REASON_GDTR_IDTR: case EXIT_REASON_LDTR_TR:
6271 		return nested_cpu_has2(vmcs12, SECONDARY_EXEC_DESC);
6272 	case EXIT_REASON_MSR_READ:
6273 	case EXIT_REASON_MSR_WRITE:
6274 		return nested_vmx_exit_handled_msr(vcpu, vmcs12, exit_reason);
6275 	case EXIT_REASON_INVALID_STATE:
6276 		return true;
6277 	case EXIT_REASON_MWAIT_INSTRUCTION:
6278 		return nested_cpu_has(vmcs12, CPU_BASED_MWAIT_EXITING);
6279 	case EXIT_REASON_MONITOR_TRAP_FLAG:
6280 		return nested_vmx_exit_handled_mtf(vmcs12);
6281 	case EXIT_REASON_MONITOR_INSTRUCTION:
6282 		return nested_cpu_has(vmcs12, CPU_BASED_MONITOR_EXITING);
6283 	case EXIT_REASON_PAUSE_INSTRUCTION:
6284 		return nested_cpu_has(vmcs12, CPU_BASED_PAUSE_EXITING) ||
6285 			nested_cpu_has2(vmcs12,
6286 				SECONDARY_EXEC_PAUSE_LOOP_EXITING);
6287 	case EXIT_REASON_MCE_DURING_VMENTRY:
6288 		return true;
6289 	case EXIT_REASON_TPR_BELOW_THRESHOLD:
6290 		return nested_cpu_has(vmcs12, CPU_BASED_TPR_SHADOW);
6291 	case EXIT_REASON_APIC_ACCESS:
6292 	case EXIT_REASON_APIC_WRITE:
6293 	case EXIT_REASON_EOI_INDUCED:
6294 		/*
6295 		 * The controls for "virtualize APIC accesses," "APIC-
6296 		 * register virtualization," and "virtual-interrupt
6297 		 * delivery" only come from vmcs12.
6298 		 */
6299 		return true;
6300 	case EXIT_REASON_INVPCID:
6301 		return
6302 			nested_cpu_has2(vmcs12, SECONDARY_EXEC_ENABLE_INVPCID) &&
6303 			nested_cpu_has(vmcs12, CPU_BASED_INVLPG_EXITING);
6304 	case EXIT_REASON_WBINVD:
6305 		return nested_cpu_has2(vmcs12, SECONDARY_EXEC_WBINVD_EXITING);
6306 	case EXIT_REASON_XSETBV:
6307 		return true;
6308 	case EXIT_REASON_XSAVES: case EXIT_REASON_XRSTORS:
6309 		/*
6310 		 * This should never happen, since it is not possible to
6311 		 * set XSS to a non-zero value---neither in L1 nor in L2.
6312 		 * If if it were, XSS would have to be checked against
6313 		 * the XSS exit bitmap in vmcs12.
6314 		 */
6315 		return nested_cpu_has2(vmcs12, SECONDARY_EXEC_XSAVES);
6316 	case EXIT_REASON_UMWAIT:
6317 	case EXIT_REASON_TPAUSE:
6318 		return nested_cpu_has2(vmcs12,
6319 			SECONDARY_EXEC_ENABLE_USR_WAIT_PAUSE);
6320 	case EXIT_REASON_ENCLS:
6321 		return nested_vmx_exit_handled_encls(vcpu, vmcs12);
6322 	case EXIT_REASON_NOTIFY:
6323 		/* Notify VM exit is not exposed to L1 */
6324 		return false;
6325 	default:
6326 		return true;
6327 	}
6328 }
6329 
6330 /*
6331  * Conditionally reflect a VM-Exit into L1.  Returns %true if the VM-Exit was
6332  * reflected into L1.
6333  */
6334 bool nested_vmx_reflect_vmexit(struct kvm_vcpu *vcpu)
6335 {
6336 	struct vcpu_vmx *vmx = to_vmx(vcpu);
6337 	union vmx_exit_reason exit_reason = vmx->exit_reason;
6338 	unsigned long exit_qual;
6339 	u32 exit_intr_info;
6340 
6341 	WARN_ON_ONCE(vmx->nested.nested_run_pending);
6342 
6343 	/*
6344 	 * Late nested VM-Fail shares the same flow as nested VM-Exit since KVM
6345 	 * has already loaded L2's state.
6346 	 */
6347 	if (unlikely(vmx->fail)) {
6348 		trace_kvm_nested_vmenter_failed(
6349 			"hardware VM-instruction error: ",
6350 			vmcs_read32(VM_INSTRUCTION_ERROR));
6351 		exit_intr_info = 0;
6352 		exit_qual = 0;
6353 		goto reflect_vmexit;
6354 	}
6355 
6356 	trace_kvm_nested_vmexit(vcpu, KVM_ISA_VMX);
6357 
6358 	/* If L0 (KVM) wants the exit, it trumps L1's desires. */
6359 	if (nested_vmx_l0_wants_exit(vcpu, exit_reason))
6360 		return false;
6361 
6362 	/* If L1 doesn't want the exit, handle it in L0. */
6363 	if (!nested_vmx_l1_wants_exit(vcpu, exit_reason))
6364 		return false;
6365 
6366 	/*
6367 	 * vmcs.VM_EXIT_INTR_INFO is only valid for EXCEPTION_NMI exits.  For
6368 	 * EXTERNAL_INTERRUPT, the value for vmcs12->vm_exit_intr_info would
6369 	 * need to be synthesized by querying the in-kernel LAPIC, but external
6370 	 * interrupts are never reflected to L1 so it's a non-issue.
6371 	 */
6372 	exit_intr_info = vmx_get_intr_info(vcpu);
6373 	if (is_exception_with_error_code(exit_intr_info)) {
6374 		struct vmcs12 *vmcs12 = get_vmcs12(vcpu);
6375 
6376 		vmcs12->vm_exit_intr_error_code =
6377 			vmcs_read32(VM_EXIT_INTR_ERROR_CODE);
6378 	}
6379 	exit_qual = vmx_get_exit_qual(vcpu);
6380 
6381 reflect_vmexit:
6382 	nested_vmx_vmexit(vcpu, exit_reason.full, exit_intr_info, exit_qual);
6383 	return true;
6384 }
6385 
6386 static int vmx_get_nested_state(struct kvm_vcpu *vcpu,
6387 				struct kvm_nested_state __user *user_kvm_nested_state,
6388 				u32 user_data_size)
6389 {
6390 	struct vcpu_vmx *vmx;
6391 	struct vmcs12 *vmcs12;
6392 	struct kvm_nested_state kvm_state = {
6393 		.flags = 0,
6394 		.format = KVM_STATE_NESTED_FORMAT_VMX,
6395 		.size = sizeof(kvm_state),
6396 		.hdr.vmx.flags = 0,
6397 		.hdr.vmx.vmxon_pa = INVALID_GPA,
6398 		.hdr.vmx.vmcs12_pa = INVALID_GPA,
6399 		.hdr.vmx.preemption_timer_deadline = 0,
6400 	};
6401 	struct kvm_vmx_nested_state_data __user *user_vmx_nested_state =
6402 		&user_kvm_nested_state->data.vmx[0];
6403 
6404 	if (!vcpu)
6405 		return kvm_state.size + sizeof(*user_vmx_nested_state);
6406 
6407 	vmx = to_vmx(vcpu);
6408 	vmcs12 = get_vmcs12(vcpu);
6409 
6410 	if (nested_vmx_allowed(vcpu) &&
6411 	    (vmx->nested.vmxon || vmx->nested.smm.vmxon)) {
6412 		kvm_state.hdr.vmx.vmxon_pa = vmx->nested.vmxon_ptr;
6413 		kvm_state.hdr.vmx.vmcs12_pa = vmx->nested.current_vmptr;
6414 
6415 		if (vmx_has_valid_vmcs12(vcpu)) {
6416 			kvm_state.size += sizeof(user_vmx_nested_state->vmcs12);
6417 
6418 			/* 'hv_evmcs_vmptr' can also be EVMPTR_MAP_PENDING here */
6419 			if (vmx->nested.hv_evmcs_vmptr != EVMPTR_INVALID)
6420 				kvm_state.flags |= KVM_STATE_NESTED_EVMCS;
6421 
6422 			if (is_guest_mode(vcpu) &&
6423 			    nested_cpu_has_shadow_vmcs(vmcs12) &&
6424 			    vmcs12->vmcs_link_pointer != INVALID_GPA)
6425 				kvm_state.size += sizeof(user_vmx_nested_state->shadow_vmcs12);
6426 		}
6427 
6428 		if (vmx->nested.smm.vmxon)
6429 			kvm_state.hdr.vmx.smm.flags |= KVM_STATE_NESTED_SMM_VMXON;
6430 
6431 		if (vmx->nested.smm.guest_mode)
6432 			kvm_state.hdr.vmx.smm.flags |= KVM_STATE_NESTED_SMM_GUEST_MODE;
6433 
6434 		if (is_guest_mode(vcpu)) {
6435 			kvm_state.flags |= KVM_STATE_NESTED_GUEST_MODE;
6436 
6437 			if (vmx->nested.nested_run_pending)
6438 				kvm_state.flags |= KVM_STATE_NESTED_RUN_PENDING;
6439 
6440 			if (vmx->nested.mtf_pending)
6441 				kvm_state.flags |= KVM_STATE_NESTED_MTF_PENDING;
6442 
6443 			if (nested_cpu_has_preemption_timer(vmcs12) &&
6444 			    vmx->nested.has_preemption_timer_deadline) {
6445 				kvm_state.hdr.vmx.flags |=
6446 					KVM_STATE_VMX_PREEMPTION_TIMER_DEADLINE;
6447 				kvm_state.hdr.vmx.preemption_timer_deadline =
6448 					vmx->nested.preemption_timer_deadline;
6449 			}
6450 		}
6451 	}
6452 
6453 	if (user_data_size < kvm_state.size)
6454 		goto out;
6455 
6456 	if (copy_to_user(user_kvm_nested_state, &kvm_state, sizeof(kvm_state)))
6457 		return -EFAULT;
6458 
6459 	if (!vmx_has_valid_vmcs12(vcpu))
6460 		goto out;
6461 
6462 	/*
6463 	 * When running L2, the authoritative vmcs12 state is in the
6464 	 * vmcs02. When running L1, the authoritative vmcs12 state is
6465 	 * in the shadow or enlightened vmcs linked to vmcs01, unless
6466 	 * need_vmcs12_to_shadow_sync is set, in which case, the authoritative
6467 	 * vmcs12 state is in the vmcs12 already.
6468 	 */
6469 	if (is_guest_mode(vcpu)) {
6470 		sync_vmcs02_to_vmcs12(vcpu, vmcs12);
6471 		sync_vmcs02_to_vmcs12_rare(vcpu, vmcs12);
6472 	} else  {
6473 		copy_vmcs02_to_vmcs12_rare(vcpu, get_vmcs12(vcpu));
6474 		if (!vmx->nested.need_vmcs12_to_shadow_sync) {
6475 			if (evmptr_is_valid(vmx->nested.hv_evmcs_vmptr))
6476 				/*
6477 				 * L1 hypervisor is not obliged to keep eVMCS
6478 				 * clean fields data always up-to-date while
6479 				 * not in guest mode, 'hv_clean_fields' is only
6480 				 * supposed to be actual upon vmentry so we need
6481 				 * to ignore it here and do full copy.
6482 				 */
6483 				copy_enlightened_to_vmcs12(vmx, 0);
6484 			else if (enable_shadow_vmcs)
6485 				copy_shadow_to_vmcs12(vmx);
6486 		}
6487 	}
6488 
6489 	BUILD_BUG_ON(sizeof(user_vmx_nested_state->vmcs12) < VMCS12_SIZE);
6490 	BUILD_BUG_ON(sizeof(user_vmx_nested_state->shadow_vmcs12) < VMCS12_SIZE);
6491 
6492 	/*
6493 	 * Copy over the full allocated size of vmcs12 rather than just the size
6494 	 * of the struct.
6495 	 */
6496 	if (copy_to_user(user_vmx_nested_state->vmcs12, vmcs12, VMCS12_SIZE))
6497 		return -EFAULT;
6498 
6499 	if (nested_cpu_has_shadow_vmcs(vmcs12) &&
6500 	    vmcs12->vmcs_link_pointer != INVALID_GPA) {
6501 		if (copy_to_user(user_vmx_nested_state->shadow_vmcs12,
6502 				 get_shadow_vmcs12(vcpu), VMCS12_SIZE))
6503 			return -EFAULT;
6504 	}
6505 out:
6506 	return kvm_state.size;
6507 }
6508 
6509 void vmx_leave_nested(struct kvm_vcpu *vcpu)
6510 {
6511 	if (is_guest_mode(vcpu)) {
6512 		to_vmx(vcpu)->nested.nested_run_pending = 0;
6513 		nested_vmx_vmexit(vcpu, -1, 0, 0);
6514 	}
6515 	free_nested(vcpu);
6516 }
6517 
6518 static int vmx_set_nested_state(struct kvm_vcpu *vcpu,
6519 				struct kvm_nested_state __user *user_kvm_nested_state,
6520 				struct kvm_nested_state *kvm_state)
6521 {
6522 	struct vcpu_vmx *vmx = to_vmx(vcpu);
6523 	struct vmcs12 *vmcs12;
6524 	enum vm_entry_failure_code ignored;
6525 	struct kvm_vmx_nested_state_data __user *user_vmx_nested_state =
6526 		&user_kvm_nested_state->data.vmx[0];
6527 	int ret;
6528 
6529 	if (kvm_state->format != KVM_STATE_NESTED_FORMAT_VMX)
6530 		return -EINVAL;
6531 
6532 	if (kvm_state->hdr.vmx.vmxon_pa == INVALID_GPA) {
6533 		if (kvm_state->hdr.vmx.smm.flags)
6534 			return -EINVAL;
6535 
6536 		if (kvm_state->hdr.vmx.vmcs12_pa != INVALID_GPA)
6537 			return -EINVAL;
6538 
6539 		/*
6540 		 * KVM_STATE_NESTED_EVMCS used to signal that KVM should
6541 		 * enable eVMCS capability on vCPU. However, since then
6542 		 * code was changed such that flag signals vmcs12 should
6543 		 * be copied into eVMCS in guest memory.
6544 		 *
6545 		 * To preserve backwards compatability, allow user
6546 		 * to set this flag even when there is no VMXON region.
6547 		 */
6548 		if (kvm_state->flags & ~KVM_STATE_NESTED_EVMCS)
6549 			return -EINVAL;
6550 	} else {
6551 		if (!nested_vmx_allowed(vcpu))
6552 			return -EINVAL;
6553 
6554 		if (!page_address_valid(vcpu, kvm_state->hdr.vmx.vmxon_pa))
6555 			return -EINVAL;
6556 	}
6557 
6558 	if ((kvm_state->hdr.vmx.smm.flags & KVM_STATE_NESTED_SMM_GUEST_MODE) &&
6559 	    (kvm_state->flags & KVM_STATE_NESTED_GUEST_MODE))
6560 		return -EINVAL;
6561 
6562 	if (kvm_state->hdr.vmx.smm.flags &
6563 	    ~(KVM_STATE_NESTED_SMM_GUEST_MODE | KVM_STATE_NESTED_SMM_VMXON))
6564 		return -EINVAL;
6565 
6566 	if (kvm_state->hdr.vmx.flags & ~KVM_STATE_VMX_PREEMPTION_TIMER_DEADLINE)
6567 		return -EINVAL;
6568 
6569 	/*
6570 	 * SMM temporarily disables VMX, so we cannot be in guest mode,
6571 	 * nor can VMLAUNCH/VMRESUME be pending.  Outside SMM, SMM flags
6572 	 * must be zero.
6573 	 */
6574 	if (is_smm(vcpu) ?
6575 		(kvm_state->flags &
6576 		 (KVM_STATE_NESTED_GUEST_MODE | KVM_STATE_NESTED_RUN_PENDING))
6577 		: kvm_state->hdr.vmx.smm.flags)
6578 		return -EINVAL;
6579 
6580 	if ((kvm_state->hdr.vmx.smm.flags & KVM_STATE_NESTED_SMM_GUEST_MODE) &&
6581 	    !(kvm_state->hdr.vmx.smm.flags & KVM_STATE_NESTED_SMM_VMXON))
6582 		return -EINVAL;
6583 
6584 	if ((kvm_state->flags & KVM_STATE_NESTED_EVMCS) &&
6585 		(!nested_vmx_allowed(vcpu) || !vmx->nested.enlightened_vmcs_enabled))
6586 			return -EINVAL;
6587 
6588 	vmx_leave_nested(vcpu);
6589 
6590 	if (kvm_state->hdr.vmx.vmxon_pa == INVALID_GPA)
6591 		return 0;
6592 
6593 	vmx->nested.vmxon_ptr = kvm_state->hdr.vmx.vmxon_pa;
6594 	ret = enter_vmx_operation(vcpu);
6595 	if (ret)
6596 		return ret;
6597 
6598 	/* Empty 'VMXON' state is permitted if no VMCS loaded */
6599 	if (kvm_state->size < sizeof(*kvm_state) + sizeof(*vmcs12)) {
6600 		/* See vmx_has_valid_vmcs12.  */
6601 		if ((kvm_state->flags & KVM_STATE_NESTED_GUEST_MODE) ||
6602 		    (kvm_state->flags & KVM_STATE_NESTED_EVMCS) ||
6603 		    (kvm_state->hdr.vmx.vmcs12_pa != INVALID_GPA))
6604 			return -EINVAL;
6605 		else
6606 			return 0;
6607 	}
6608 
6609 	if (kvm_state->hdr.vmx.vmcs12_pa != INVALID_GPA) {
6610 		if (kvm_state->hdr.vmx.vmcs12_pa == kvm_state->hdr.vmx.vmxon_pa ||
6611 		    !page_address_valid(vcpu, kvm_state->hdr.vmx.vmcs12_pa))
6612 			return -EINVAL;
6613 
6614 		set_current_vmptr(vmx, kvm_state->hdr.vmx.vmcs12_pa);
6615 	} else if (kvm_state->flags & KVM_STATE_NESTED_EVMCS) {
6616 		/*
6617 		 * nested_vmx_handle_enlightened_vmptrld() cannot be called
6618 		 * directly from here as HV_X64_MSR_VP_ASSIST_PAGE may not be
6619 		 * restored yet. EVMCS will be mapped from
6620 		 * nested_get_vmcs12_pages().
6621 		 */
6622 		vmx->nested.hv_evmcs_vmptr = EVMPTR_MAP_PENDING;
6623 		kvm_make_request(KVM_REQ_GET_NESTED_STATE_PAGES, vcpu);
6624 	} else {
6625 		return -EINVAL;
6626 	}
6627 
6628 	if (kvm_state->hdr.vmx.smm.flags & KVM_STATE_NESTED_SMM_VMXON) {
6629 		vmx->nested.smm.vmxon = true;
6630 		vmx->nested.vmxon = false;
6631 
6632 		if (kvm_state->hdr.vmx.smm.flags & KVM_STATE_NESTED_SMM_GUEST_MODE)
6633 			vmx->nested.smm.guest_mode = true;
6634 	}
6635 
6636 	vmcs12 = get_vmcs12(vcpu);
6637 	if (copy_from_user(vmcs12, user_vmx_nested_state->vmcs12, sizeof(*vmcs12)))
6638 		return -EFAULT;
6639 
6640 	if (vmcs12->hdr.revision_id != VMCS12_REVISION)
6641 		return -EINVAL;
6642 
6643 	if (!(kvm_state->flags & KVM_STATE_NESTED_GUEST_MODE))
6644 		return 0;
6645 
6646 	vmx->nested.nested_run_pending =
6647 		!!(kvm_state->flags & KVM_STATE_NESTED_RUN_PENDING);
6648 
6649 	vmx->nested.mtf_pending =
6650 		!!(kvm_state->flags & KVM_STATE_NESTED_MTF_PENDING);
6651 
6652 	ret = -EINVAL;
6653 	if (nested_cpu_has_shadow_vmcs(vmcs12) &&
6654 	    vmcs12->vmcs_link_pointer != INVALID_GPA) {
6655 		struct vmcs12 *shadow_vmcs12 = get_shadow_vmcs12(vcpu);
6656 
6657 		if (kvm_state->size <
6658 		    sizeof(*kvm_state) +
6659 		    sizeof(user_vmx_nested_state->vmcs12) + sizeof(*shadow_vmcs12))
6660 			goto error_guest_mode;
6661 
6662 		if (copy_from_user(shadow_vmcs12,
6663 				   user_vmx_nested_state->shadow_vmcs12,
6664 				   sizeof(*shadow_vmcs12))) {
6665 			ret = -EFAULT;
6666 			goto error_guest_mode;
6667 		}
6668 
6669 		if (shadow_vmcs12->hdr.revision_id != VMCS12_REVISION ||
6670 		    !shadow_vmcs12->hdr.shadow_vmcs)
6671 			goto error_guest_mode;
6672 	}
6673 
6674 	vmx->nested.has_preemption_timer_deadline = false;
6675 	if (kvm_state->hdr.vmx.flags & KVM_STATE_VMX_PREEMPTION_TIMER_DEADLINE) {
6676 		vmx->nested.has_preemption_timer_deadline = true;
6677 		vmx->nested.preemption_timer_deadline =
6678 			kvm_state->hdr.vmx.preemption_timer_deadline;
6679 	}
6680 
6681 	if (nested_vmx_check_controls(vcpu, vmcs12) ||
6682 	    nested_vmx_check_host_state(vcpu, vmcs12) ||
6683 	    nested_vmx_check_guest_state(vcpu, vmcs12, &ignored))
6684 		goto error_guest_mode;
6685 
6686 	vmx->nested.dirty_vmcs12 = true;
6687 	vmx->nested.force_msr_bitmap_recalc = true;
6688 	ret = nested_vmx_enter_non_root_mode(vcpu, false);
6689 	if (ret)
6690 		goto error_guest_mode;
6691 
6692 	if (vmx->nested.mtf_pending)
6693 		kvm_make_request(KVM_REQ_EVENT, vcpu);
6694 
6695 	return 0;
6696 
6697 error_guest_mode:
6698 	vmx->nested.nested_run_pending = 0;
6699 	return ret;
6700 }
6701 
6702 void nested_vmx_set_vmcs_shadowing_bitmap(void)
6703 {
6704 	if (enable_shadow_vmcs) {
6705 		vmcs_write64(VMREAD_BITMAP, __pa(vmx_vmread_bitmap));
6706 		vmcs_write64(VMWRITE_BITMAP, __pa(vmx_vmwrite_bitmap));
6707 	}
6708 }
6709 
6710 /*
6711  * Indexing into the vmcs12 uses the VMCS encoding rotated left by 6.  Undo
6712  * that madness to get the encoding for comparison.
6713  */
6714 #define VMCS12_IDX_TO_ENC(idx) ((u16)(((u16)(idx) >> 6) | ((u16)(idx) << 10)))
6715 
6716 static u64 nested_vmx_calc_vmcs_enum_msr(void)
6717 {
6718 	/*
6719 	 * Note these are the so called "index" of the VMCS field encoding, not
6720 	 * the index into vmcs12.
6721 	 */
6722 	unsigned int max_idx, idx;
6723 	int i;
6724 
6725 	/*
6726 	 * For better or worse, KVM allows VMREAD/VMWRITE to all fields in
6727 	 * vmcs12, regardless of whether or not the associated feature is
6728 	 * exposed to L1.  Simply find the field with the highest index.
6729 	 */
6730 	max_idx = 0;
6731 	for (i = 0; i < nr_vmcs12_fields; i++) {
6732 		/* The vmcs12 table is very, very sparsely populated. */
6733 		if (!vmcs12_field_offsets[i])
6734 			continue;
6735 
6736 		idx = vmcs_field_index(VMCS12_IDX_TO_ENC(i));
6737 		if (idx > max_idx)
6738 			max_idx = idx;
6739 	}
6740 
6741 	return (u64)max_idx << VMCS_FIELD_INDEX_SHIFT;
6742 }
6743 
6744 /*
6745  * nested_vmx_setup_ctls_msrs() sets up variables containing the values to be
6746  * returned for the various VMX controls MSRs when nested VMX is enabled.
6747  * The same values should also be used to verify that vmcs12 control fields are
6748  * valid during nested entry from L1 to L2.
6749  * Each of these control msrs has a low and high 32-bit half: A low bit is on
6750  * if the corresponding bit in the (32-bit) control field *must* be on, and a
6751  * bit in the high half is on if the corresponding bit in the control field
6752  * may be on. See also vmx_control_verify().
6753  */
6754 void nested_vmx_setup_ctls_msrs(struct vmcs_config *vmcs_conf, u32 ept_caps)
6755 {
6756 	struct nested_vmx_msrs *msrs = &vmcs_conf->nested;
6757 
6758 	/*
6759 	 * Note that as a general rule, the high half of the MSRs (bits in
6760 	 * the control fields which may be 1) should be initialized by the
6761 	 * intersection of the underlying hardware's MSR (i.e., features which
6762 	 * can be supported) and the list of features we want to expose -
6763 	 * because they are known to be properly supported in our code.
6764 	 * Also, usually, the low half of the MSRs (bits which must be 1) can
6765 	 * be set to 0, meaning that L1 may turn off any of these bits. The
6766 	 * reason is that if one of these bits is necessary, it will appear
6767 	 * in vmcs01 and prepare_vmcs02, when it bitwise-or's the control
6768 	 * fields of vmcs01 and vmcs02, will turn these bits off - and
6769 	 * nested_vmx_l1_wants_exit() will not pass related exits to L1.
6770 	 * These rules have exceptions below.
6771 	 */
6772 
6773 	/* pin-based controls */
6774 	msrs->pinbased_ctls_low =
6775 		PIN_BASED_ALWAYSON_WITHOUT_TRUE_MSR;
6776 
6777 	msrs->pinbased_ctls_high = vmcs_conf->pin_based_exec_ctrl;
6778 	msrs->pinbased_ctls_high &=
6779 		PIN_BASED_EXT_INTR_MASK |
6780 		PIN_BASED_NMI_EXITING |
6781 		PIN_BASED_VIRTUAL_NMIS |
6782 		(enable_apicv ? PIN_BASED_POSTED_INTR : 0);
6783 	msrs->pinbased_ctls_high |=
6784 		PIN_BASED_ALWAYSON_WITHOUT_TRUE_MSR |
6785 		PIN_BASED_VMX_PREEMPTION_TIMER;
6786 
6787 	/* exit controls */
6788 	msrs->exit_ctls_low =
6789 		VM_EXIT_ALWAYSON_WITHOUT_TRUE_MSR;
6790 
6791 	msrs->exit_ctls_high = vmcs_conf->vmexit_ctrl;
6792 	msrs->exit_ctls_high &=
6793 #ifdef CONFIG_X86_64
6794 		VM_EXIT_HOST_ADDR_SPACE_SIZE |
6795 #endif
6796 		VM_EXIT_LOAD_IA32_PAT | VM_EXIT_SAVE_IA32_PAT |
6797 		VM_EXIT_CLEAR_BNDCFGS;
6798 	msrs->exit_ctls_high |=
6799 		VM_EXIT_ALWAYSON_WITHOUT_TRUE_MSR |
6800 		VM_EXIT_LOAD_IA32_EFER | VM_EXIT_SAVE_IA32_EFER |
6801 		VM_EXIT_SAVE_VMX_PREEMPTION_TIMER | VM_EXIT_ACK_INTR_ON_EXIT |
6802 		VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL;
6803 
6804 	/* We support free control of debug control saving. */
6805 	msrs->exit_ctls_low &= ~VM_EXIT_SAVE_DEBUG_CONTROLS;
6806 
6807 	/* entry controls */
6808 	msrs->entry_ctls_low =
6809 		VM_ENTRY_ALWAYSON_WITHOUT_TRUE_MSR;
6810 
6811 	msrs->entry_ctls_high = vmcs_conf->vmentry_ctrl;
6812 	msrs->entry_ctls_high &=
6813 #ifdef CONFIG_X86_64
6814 		VM_ENTRY_IA32E_MODE |
6815 #endif
6816 		VM_ENTRY_LOAD_IA32_PAT | VM_ENTRY_LOAD_BNDCFGS;
6817 	msrs->entry_ctls_high |=
6818 		(VM_ENTRY_ALWAYSON_WITHOUT_TRUE_MSR | VM_ENTRY_LOAD_IA32_EFER |
6819 		 VM_ENTRY_LOAD_IA32_PERF_GLOBAL_CTRL);
6820 
6821 	/* We support free control of debug control loading. */
6822 	msrs->entry_ctls_low &= ~VM_ENTRY_LOAD_DEBUG_CONTROLS;
6823 
6824 	/* cpu-based controls */
6825 	msrs->procbased_ctls_low =
6826 		CPU_BASED_ALWAYSON_WITHOUT_TRUE_MSR;
6827 
6828 	msrs->procbased_ctls_high = vmcs_conf->cpu_based_exec_ctrl;
6829 	msrs->procbased_ctls_high &=
6830 		CPU_BASED_INTR_WINDOW_EXITING |
6831 		CPU_BASED_NMI_WINDOW_EXITING | CPU_BASED_USE_TSC_OFFSETTING |
6832 		CPU_BASED_HLT_EXITING | CPU_BASED_INVLPG_EXITING |
6833 		CPU_BASED_MWAIT_EXITING | CPU_BASED_CR3_LOAD_EXITING |
6834 		CPU_BASED_CR3_STORE_EXITING |
6835 #ifdef CONFIG_X86_64
6836 		CPU_BASED_CR8_LOAD_EXITING | CPU_BASED_CR8_STORE_EXITING |
6837 #endif
6838 		CPU_BASED_MOV_DR_EXITING | CPU_BASED_UNCOND_IO_EXITING |
6839 		CPU_BASED_USE_IO_BITMAPS | CPU_BASED_MONITOR_TRAP_FLAG |
6840 		CPU_BASED_MONITOR_EXITING | CPU_BASED_RDPMC_EXITING |
6841 		CPU_BASED_RDTSC_EXITING | CPU_BASED_PAUSE_EXITING |
6842 		CPU_BASED_TPR_SHADOW | CPU_BASED_ACTIVATE_SECONDARY_CONTROLS;
6843 	/*
6844 	 * We can allow some features even when not supported by the
6845 	 * hardware. For example, L1 can specify an MSR bitmap - and we
6846 	 * can use it to avoid exits to L1 - even when L0 runs L2
6847 	 * without MSR bitmaps.
6848 	 */
6849 	msrs->procbased_ctls_high |=
6850 		CPU_BASED_ALWAYSON_WITHOUT_TRUE_MSR |
6851 		CPU_BASED_USE_MSR_BITMAPS;
6852 
6853 	/* We support free control of CR3 access interception. */
6854 	msrs->procbased_ctls_low &=
6855 		~(CPU_BASED_CR3_LOAD_EXITING | CPU_BASED_CR3_STORE_EXITING);
6856 
6857 	/*
6858 	 * secondary cpu-based controls.  Do not include those that
6859 	 * depend on CPUID bits, they are added later by
6860 	 * vmx_vcpu_after_set_cpuid.
6861 	 */
6862 	msrs->secondary_ctls_low = 0;
6863 
6864 	msrs->secondary_ctls_high = vmcs_conf->cpu_based_2nd_exec_ctrl;
6865 	msrs->secondary_ctls_high &=
6866 		SECONDARY_EXEC_DESC |
6867 		SECONDARY_EXEC_ENABLE_RDTSCP |
6868 		SECONDARY_EXEC_VIRTUALIZE_X2APIC_MODE |
6869 		SECONDARY_EXEC_WBINVD_EXITING |
6870 		SECONDARY_EXEC_APIC_REGISTER_VIRT |
6871 		SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY |
6872 		SECONDARY_EXEC_RDRAND_EXITING |
6873 		SECONDARY_EXEC_ENABLE_INVPCID |
6874 		SECONDARY_EXEC_RDSEED_EXITING |
6875 		SECONDARY_EXEC_XSAVES |
6876 		SECONDARY_EXEC_TSC_SCALING;
6877 
6878 	/*
6879 	 * We can emulate "VMCS shadowing," even if the hardware
6880 	 * doesn't support it.
6881 	 */
6882 	msrs->secondary_ctls_high |=
6883 		SECONDARY_EXEC_SHADOW_VMCS;
6884 
6885 	if (enable_ept) {
6886 		/* nested EPT: emulate EPT also to L1 */
6887 		msrs->secondary_ctls_high |=
6888 			SECONDARY_EXEC_ENABLE_EPT;
6889 		msrs->ept_caps =
6890 			VMX_EPT_PAGE_WALK_4_BIT |
6891 			VMX_EPT_PAGE_WALK_5_BIT |
6892 			VMX_EPTP_WB_BIT |
6893 			VMX_EPT_INVEPT_BIT |
6894 			VMX_EPT_EXECUTE_ONLY_BIT;
6895 
6896 		msrs->ept_caps &= ept_caps;
6897 		msrs->ept_caps |= VMX_EPT_EXTENT_GLOBAL_BIT |
6898 			VMX_EPT_EXTENT_CONTEXT_BIT | VMX_EPT_2MB_PAGE_BIT |
6899 			VMX_EPT_1GB_PAGE_BIT;
6900 		if (enable_ept_ad_bits) {
6901 			msrs->secondary_ctls_high |=
6902 				SECONDARY_EXEC_ENABLE_PML;
6903 			msrs->ept_caps |= VMX_EPT_AD_BIT;
6904 		}
6905 	}
6906 
6907 	if (cpu_has_vmx_vmfunc()) {
6908 		msrs->secondary_ctls_high |=
6909 			SECONDARY_EXEC_ENABLE_VMFUNC;
6910 		/*
6911 		 * Advertise EPTP switching unconditionally
6912 		 * since we emulate it
6913 		 */
6914 		if (enable_ept)
6915 			msrs->vmfunc_controls =
6916 				VMX_VMFUNC_EPTP_SWITCHING;
6917 	}
6918 
6919 	/*
6920 	 * Old versions of KVM use the single-context version without
6921 	 * checking for support, so declare that it is supported even
6922 	 * though it is treated as global context.  The alternative is
6923 	 * not failing the single-context invvpid, and it is worse.
6924 	 */
6925 	if (enable_vpid) {
6926 		msrs->secondary_ctls_high |=
6927 			SECONDARY_EXEC_ENABLE_VPID;
6928 		msrs->vpid_caps = VMX_VPID_INVVPID_BIT |
6929 			VMX_VPID_EXTENT_SUPPORTED_MASK;
6930 	}
6931 
6932 	if (enable_unrestricted_guest)
6933 		msrs->secondary_ctls_high |=
6934 			SECONDARY_EXEC_UNRESTRICTED_GUEST;
6935 
6936 	if (flexpriority_enabled)
6937 		msrs->secondary_ctls_high |=
6938 			SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES;
6939 
6940 	if (enable_sgx)
6941 		msrs->secondary_ctls_high |= SECONDARY_EXEC_ENCLS_EXITING;
6942 
6943 	/* miscellaneous data */
6944 	msrs->misc_low = (u32)vmcs_conf->misc & VMX_MISC_SAVE_EFER_LMA;
6945 	msrs->misc_low |=
6946 		MSR_IA32_VMX_MISC_VMWRITE_SHADOW_RO_FIELDS |
6947 		VMX_MISC_EMULATED_PREEMPTION_TIMER_RATE |
6948 		VMX_MISC_ACTIVITY_HLT |
6949 		VMX_MISC_ACTIVITY_WAIT_SIPI;
6950 	msrs->misc_high = 0;
6951 
6952 	/*
6953 	 * This MSR reports some information about VMX support. We
6954 	 * should return information about the VMX we emulate for the
6955 	 * guest, and the VMCS structure we give it - not about the
6956 	 * VMX support of the underlying hardware.
6957 	 */
6958 	msrs->basic =
6959 		VMCS12_REVISION |
6960 		VMX_BASIC_TRUE_CTLS |
6961 		((u64)VMCS12_SIZE << VMX_BASIC_VMCS_SIZE_SHIFT) |
6962 		(VMX_BASIC_MEM_TYPE_WB << VMX_BASIC_MEM_TYPE_SHIFT);
6963 
6964 	if (cpu_has_vmx_basic_inout())
6965 		msrs->basic |= VMX_BASIC_INOUT;
6966 
6967 	/*
6968 	 * These MSRs specify bits which the guest must keep fixed on
6969 	 * while L1 is in VMXON mode (in L1's root mode, or running an L2).
6970 	 * We picked the standard core2 setting.
6971 	 */
6972 #define VMXON_CR0_ALWAYSON     (X86_CR0_PE | X86_CR0_PG | X86_CR0_NE)
6973 #define VMXON_CR4_ALWAYSON     X86_CR4_VMXE
6974 	msrs->cr0_fixed0 = VMXON_CR0_ALWAYSON;
6975 	msrs->cr4_fixed0 = VMXON_CR4_ALWAYSON;
6976 
6977 	/* These MSRs specify bits which the guest must keep fixed off. */
6978 	rdmsrl(MSR_IA32_VMX_CR0_FIXED1, msrs->cr0_fixed1);
6979 	rdmsrl(MSR_IA32_VMX_CR4_FIXED1, msrs->cr4_fixed1);
6980 
6981 	if (vmx_umip_emulated())
6982 		msrs->cr4_fixed1 |= X86_CR4_UMIP;
6983 
6984 	msrs->vmcs_enum = nested_vmx_calc_vmcs_enum_msr();
6985 }
6986 
6987 void nested_vmx_hardware_unsetup(void)
6988 {
6989 	int i;
6990 
6991 	if (enable_shadow_vmcs) {
6992 		for (i = 0; i < VMX_BITMAP_NR; i++)
6993 			free_page((unsigned long)vmx_bitmap[i]);
6994 	}
6995 }
6996 
6997 __init int nested_vmx_hardware_setup(int (*exit_handlers[])(struct kvm_vcpu *))
6998 {
6999 	int i;
7000 
7001 	if (!cpu_has_vmx_shadow_vmcs())
7002 		enable_shadow_vmcs = 0;
7003 	if (enable_shadow_vmcs) {
7004 		for (i = 0; i < VMX_BITMAP_NR; i++) {
7005 			/*
7006 			 * The vmx_bitmap is not tied to a VM and so should
7007 			 * not be charged to a memcg.
7008 			 */
7009 			vmx_bitmap[i] = (unsigned long *)
7010 				__get_free_page(GFP_KERNEL);
7011 			if (!vmx_bitmap[i]) {
7012 				nested_vmx_hardware_unsetup();
7013 				return -ENOMEM;
7014 			}
7015 		}
7016 
7017 		init_vmcs_shadow_fields();
7018 	}
7019 
7020 	exit_handlers[EXIT_REASON_VMCLEAR]	= handle_vmclear;
7021 	exit_handlers[EXIT_REASON_VMLAUNCH]	= handle_vmlaunch;
7022 	exit_handlers[EXIT_REASON_VMPTRLD]	= handle_vmptrld;
7023 	exit_handlers[EXIT_REASON_VMPTRST]	= handle_vmptrst;
7024 	exit_handlers[EXIT_REASON_VMREAD]	= handle_vmread;
7025 	exit_handlers[EXIT_REASON_VMRESUME]	= handle_vmresume;
7026 	exit_handlers[EXIT_REASON_VMWRITE]	= handle_vmwrite;
7027 	exit_handlers[EXIT_REASON_VMOFF]	= handle_vmxoff;
7028 	exit_handlers[EXIT_REASON_VMON]		= handle_vmxon;
7029 	exit_handlers[EXIT_REASON_INVEPT]	= handle_invept;
7030 	exit_handlers[EXIT_REASON_INVVPID]	= handle_invvpid;
7031 	exit_handlers[EXIT_REASON_VMFUNC]	= handle_vmfunc;
7032 
7033 	return 0;
7034 }
7035 
7036 struct kvm_x86_nested_ops vmx_nested_ops = {
7037 	.leave_nested = vmx_leave_nested,
7038 	.is_exception_vmexit = nested_vmx_is_exception_vmexit,
7039 	.check_events = vmx_check_nested_events,
7040 	.has_events = vmx_has_nested_events,
7041 	.triple_fault = nested_vmx_triple_fault,
7042 	.get_state = vmx_get_nested_state,
7043 	.set_state = vmx_set_nested_state,
7044 	.get_nested_state_pages = vmx_get_nested_state_pages,
7045 	.write_log_dirty = nested_vmx_write_pml_buffer,
7046 	.enable_evmcs = nested_enable_evmcs,
7047 	.get_evmcs_version = nested_get_evmcs_version,
7048 	.hv_inject_synthetic_vmexit_post_tlb_flush = vmx_hv_inject_synthetic_vmexit_post_tlb_flush,
7049 };
7050