xref: /linux/arch/x86/kvm/vmx/vmx.h (revision 53597deca0e38c30e6cd4ba2114fa42d2bcd85bb)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef __KVM_X86_VMX_H
3 #define __KVM_X86_VMX_H
4 
5 #include <linux/kvm_host.h>
6 
7 #include <asm/kvm.h>
8 #include <asm/intel_pt.h>
9 #include <asm/perf_event.h>
10 #include <asm/posted_intr.h>
11 
12 #include "capabilities.h"
13 #include "../kvm_cache_regs.h"
14 #include "pmu_intel.h"
15 #include "vmcs.h"
16 #include "vmx_ops.h"
17 #include "../cpuid.h"
18 #include "run_flags.h"
19 #include "../mmu.h"
20 #include "common.h"
21 
22 #ifdef CONFIG_X86_64
23 #define MAX_NR_USER_RETURN_MSRS	7
24 #else
25 #define MAX_NR_USER_RETURN_MSRS	4
26 #endif
27 
28 #define MAX_NR_LOADSTORE_MSRS	8
29 
30 struct vmx_msrs {
31 	unsigned int		nr;
32 	struct vmx_msr_entry	val[MAX_NR_LOADSTORE_MSRS];
33 };
34 
35 struct vmx_uret_msr {
36 	bool load_into_hardware;
37 	u64 data;
38 	u64 mask;
39 };
40 
41 enum segment_cache_field {
42 	SEG_FIELD_SEL = 0,
43 	SEG_FIELD_BASE = 1,
44 	SEG_FIELD_LIMIT = 2,
45 	SEG_FIELD_AR = 3,
46 
47 	SEG_FIELD_NR = 4
48 };
49 
50 #define RTIT_ADDR_RANGE		4
51 
52 struct pt_ctx {
53 	u64 ctl;
54 	u64 status;
55 	u64 output_base;
56 	u64 output_mask;
57 	u64 cr3_match;
58 	u64 addr_a[RTIT_ADDR_RANGE];
59 	u64 addr_b[RTIT_ADDR_RANGE];
60 };
61 
62 struct pt_desc {
63 	u64 ctl_bitmask;
64 	u32 num_address_ranges;
65 	u32 caps[PT_CPUID_REGS_NUM * PT_CPUID_LEAVES];
66 	struct pt_ctx host;
67 	struct pt_ctx guest;
68 };
69 
70 /*
71  * The nested_vmx structure is part of vcpu_vmx, and holds information we need
72  * for correct emulation of VMX (i.e., nested VMX) on this vcpu.
73  */
74 struct nested_vmx {
75 	/* Has the level1 guest done vmxon? */
76 	bool vmxon;
77 	gpa_t vmxon_ptr;
78 	bool pml_full;
79 
80 	/* The guest-physical address of the current VMCS L1 keeps for L2 */
81 	gpa_t current_vmptr;
82 	/*
83 	 * Cache of the guest's VMCS, existing outside of guest memory.
84 	 * Loaded from guest memory during VMPTRLD. Flushed to guest
85 	 * memory during VMCLEAR and VMPTRLD.
86 	 */
87 	struct vmcs12 *cached_vmcs12;
88 	/*
89 	 * Cache of the guest's shadow VMCS, existing outside of guest
90 	 * memory. Loaded from guest memory during VM entry. Flushed
91 	 * to guest memory during VM exit.
92 	 */
93 	struct vmcs12 *cached_shadow_vmcs12;
94 
95 	/*
96 	 * GPA to HVA cache for accessing vmcs12->vmcs_link_pointer
97 	 */
98 	struct gfn_to_hva_cache shadow_vmcs12_cache;
99 
100 	/*
101 	 * GPA to HVA cache for VMCS12
102 	 */
103 	struct gfn_to_hva_cache vmcs12_cache;
104 
105 	/*
106 	 * Indicates if the shadow vmcs or enlightened vmcs must be updated
107 	 * with the data held by struct vmcs12.
108 	 */
109 	bool need_vmcs12_to_shadow_sync;
110 	bool dirty_vmcs12;
111 
112 	/*
113 	 * Indicates whether MSR bitmap for L2 needs to be rebuilt due to
114 	 * changes in MSR bitmap for L1 or switching to a different L2. Note,
115 	 * this flag can only be used reliably in conjunction with a paravirt L1
116 	 * which informs L0 whether any changes to MSR bitmap for L2 were done
117 	 * on its side.
118 	 */
119 	bool force_msr_bitmap_recalc;
120 
121 	/*
122 	 * Indicates lazily loaded guest state has not yet been decached from
123 	 * vmcs02.
124 	 */
125 	bool need_sync_vmcs02_to_vmcs12_rare;
126 
127 	/*
128 	 * vmcs02 has been initialized, i.e. state that is constant for
129 	 * vmcs02 has been written to the backing VMCS.  Initialization
130 	 * is delayed until L1 actually attempts to run a nested VM.
131 	 */
132 	bool vmcs02_initialized;
133 
134 	/*
135 	 * Enlightened VMCS has been enabled. It does not mean that L1 has to
136 	 * use it. However, VMX features available to L1 will be limited based
137 	 * on what the enlightened VMCS supports.
138 	 */
139 	bool enlightened_vmcs_enabled;
140 
141 	/* Pending MTF VM-exit into L1.  */
142 	bool mtf_pending;
143 
144 	struct loaded_vmcs vmcs02;
145 
146 	/*
147 	 * Guest pages referred to in the vmcs02 with host-physical
148 	 * pointers, so we must keep them pinned while L2 runs.
149 	 */
150 	struct kvm_host_map apic_access_page_map;
151 	struct kvm_host_map virtual_apic_map;
152 	struct kvm_host_map pi_desc_map;
153 
154 	struct pi_desc *pi_desc;
155 	bool pi_pending;
156 	u16 posted_intr_nv;
157 
158 	struct hrtimer preemption_timer;
159 	u64 preemption_timer_deadline;
160 	bool has_preemption_timer_deadline;
161 	bool preemption_timer_expired;
162 
163 	/*
164 	 * Used to snapshot MSRs that are conditionally loaded on VM-Enter in
165 	 * order to propagate the guest's pre-VM-Enter value into vmcs02.  For
166 	 * emulation of VMLAUNCH/VMRESUME, the snapshot will be of L1's value.
167 	 * For KVM_SET_NESTED_STATE, the snapshot is of L2's value, _if_
168 	 * userspace restores MSRs before nested state.  If userspace restores
169 	 * MSRs after nested state, the snapshot holds garbage, but KVM can't
170 	 * detect that, and the garbage value in vmcs02 will be overwritten by
171 	 * MSR restoration in any case.
172 	 */
173 	u64 pre_vmenter_debugctl;
174 	u64 pre_vmenter_bndcfgs;
175 	u64 pre_vmenter_s_cet;
176 	u64 pre_vmenter_ssp;
177 	u64 pre_vmenter_ssp_tbl;
178 
179 	u16 vpid02;
180 	u16 last_vpid;
181 
182 	int tsc_autostore_slot;
183 	struct nested_vmx_msrs msrs;
184 
185 	/* SMM related state */
186 	struct {
187 		/* in VMX operation on SMM entry? */
188 		bool vmxon;
189 		/* in guest mode on SMM entry? */
190 		bool guest_mode;
191 	} smm;
192 
193 #ifdef CONFIG_KVM_HYPERV
194 	gpa_t hv_evmcs_vmptr;
195 	struct kvm_host_map hv_evmcs_map;
196 	struct hv_enlightened_vmcs *hv_evmcs;
197 #endif
198 };
199 
200 struct vcpu_vmx {
201 	struct kvm_vcpu       vcpu;
202 	struct vcpu_vt	      vt;
203 	u8                    fail;
204 	u8		      x2apic_msr_bitmap_mode;
205 
206 	u32                   idt_vectoring_info;
207 	ulong                 rflags;
208 
209 	/*
210 	 * User return MSRs are always emulated when enabled in the guest, but
211 	 * only loaded into hardware when necessary, e.g. SYSCALL #UDs outside
212 	 * of 64-bit mode or if EFER.SCE=1, thus the SYSCALL MSRs don't need to
213 	 * be loaded into hardware if those conditions aren't met.
214 	 */
215 	struct vmx_uret_msr   guest_uret_msrs[MAX_NR_USER_RETURN_MSRS];
216 	bool                  guest_uret_msrs_loaded;
217 #ifdef CONFIG_X86_64
218 	u64		      msr_guest_kernel_gs_base;
219 #endif
220 
221 	u64		      spec_ctrl;
222 	u32		      msr_ia32_umwait_control;
223 
224 	/*
225 	 * loaded_vmcs points to the VMCS currently used in this vcpu. For a
226 	 * non-nested (L1) guest, it always points to vmcs01. For a nested
227 	 * guest (L2), it points to a different VMCS.
228 	 */
229 	struct loaded_vmcs    vmcs01;
230 	struct loaded_vmcs   *loaded_vmcs;
231 
232 	struct msr_autoload {
233 		struct vmx_msrs guest;
234 		struct vmx_msrs host;
235 	} msr_autoload;
236 
237 	struct vmx_msrs msr_autostore;
238 
239 	struct {
240 		int vm86_active;
241 		ulong save_rflags;
242 		struct kvm_segment segs[8];
243 	} rmode;
244 	struct {
245 		u32 bitmask; /* 4 bits per segment (1 bit per field) */
246 		struct kvm_save_segment {
247 			u16 selector;
248 			unsigned long base;
249 			u32 limit;
250 			u32 ar;
251 		} seg[8];
252 	} segment_cache;
253 	int vpid;
254 
255 	/* Support for a guest hypervisor (nested VMX) */
256 	struct nested_vmx nested;
257 
258 	/* Dynamic PLE window. */
259 	unsigned int ple_window;
260 	bool ple_window_dirty;
261 
262 	/* Support for PML */
263 #define PML_LOG_NR_ENTRIES	512
264 	/* PML is written backwards: this is the first entry written by the CPU */
265 #define PML_HEAD_INDEX		(PML_LOG_NR_ENTRIES-1)
266 
267 	struct page *pml_pg;
268 
269 	/* apic deadline value in host tsc */
270 	u64 hv_deadline_tsc;
271 
272 	/*
273 	 * Only bits masked by msr_ia32_feature_control_valid_bits can be set in
274 	 * msr_ia32_feature_control. FEAT_CTL_LOCKED is always included
275 	 * in msr_ia32_feature_control_valid_bits.
276 	 */
277 	u64 msr_ia32_feature_control;
278 	u64 msr_ia32_feature_control_valid_bits;
279 	/* SGX Launch Control public key hash */
280 	u64 msr_ia32_sgxlepubkeyhash[4];
281 	u64 msr_ia32_mcu_opt_ctrl;
282 	bool disable_fb_clear;
283 
284 	struct pt_desc pt_desc;
285 	struct lbr_desc lbr_desc;
286 
287 	/* ve_info must be page aligned. */
288 	struct vmx_ve_information *ve_info;
289 };
290 
291 struct kvm_vmx {
292 	struct kvm kvm;
293 
294 	unsigned int tss_addr;
295 	bool ept_identity_pagetable_done;
296 	gpa_t ept_identity_map_addr;
297 	/* Posted Interrupt Descriptor (PID) table for IPI virtualization */
298 	u64 *pid_table;
299 };
300 
301 static __always_inline struct vcpu_vt *to_vt(struct kvm_vcpu *vcpu)
302 {
303 	return &(container_of(vcpu, struct vcpu_vmx, vcpu)->vt);
304 }
305 
306 static __always_inline struct kvm_vcpu *vt_to_vcpu(struct vcpu_vt *vt)
307 {
308 	return &(container_of(vt, struct vcpu_vmx, vt)->vcpu);
309 }
310 
311 static __always_inline union vmx_exit_reason vmx_get_exit_reason(struct kvm_vcpu *vcpu)
312 {
313 	return to_vt(vcpu)->exit_reason;
314 }
315 
316 static __always_inline unsigned long vmx_get_exit_qual(struct kvm_vcpu *vcpu)
317 {
318 	struct vcpu_vt *vt = to_vt(vcpu);
319 
320 	if (!kvm_register_test_and_mark_available(vcpu, VCPU_EXREG_EXIT_INFO_1) &&
321 	    !WARN_ON_ONCE(is_td_vcpu(vcpu)))
322 		vt->exit_qualification = vmcs_readl(EXIT_QUALIFICATION);
323 
324 	return vt->exit_qualification;
325 }
326 
327 static __always_inline u32 vmx_get_intr_info(struct kvm_vcpu *vcpu)
328 {
329 	struct vcpu_vt *vt = to_vt(vcpu);
330 
331 	if (!kvm_register_test_and_mark_available(vcpu, VCPU_EXREG_EXIT_INFO_2) &&
332 	    !WARN_ON_ONCE(is_td_vcpu(vcpu)))
333 		vt->exit_intr_info = vmcs_read32(VM_EXIT_INTR_INFO);
334 
335 	return vt->exit_intr_info;
336 }
337 
338 void vmx_vcpu_load_vmcs(struct kvm_vcpu *vcpu, int cpu);
339 int allocate_vpid(void);
340 void free_vpid(int vpid);
341 void vmx_set_constant_host_state(struct vcpu_vmx *vmx);
342 void vmx_prepare_switch_to_guest(struct kvm_vcpu *vcpu);
343 void vmx_set_host_fs_gs(struct vmcs_host_state *host, u16 fs_sel, u16 gs_sel,
344 			unsigned long fs_base, unsigned long gs_base);
345 int vmx_get_cpl(struct kvm_vcpu *vcpu);
346 int vmx_get_cpl_no_cache(struct kvm_vcpu *vcpu);
347 bool vmx_emulation_required(struct kvm_vcpu *vcpu);
348 unsigned long vmx_get_rflags(struct kvm_vcpu *vcpu);
349 void vmx_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
350 u32 vmx_get_interrupt_shadow(struct kvm_vcpu *vcpu);
351 void vmx_set_interrupt_shadow(struct kvm_vcpu *vcpu, int mask);
352 int vmx_set_efer(struct kvm_vcpu *vcpu, u64 efer);
353 void vmx_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
354 void vmx_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
355 void set_cr4_guest_host_mask(struct vcpu_vmx *vmx);
356 void ept_save_pdptrs(struct kvm_vcpu *vcpu);
357 void vmx_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
358 void __vmx_set_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
359 
360 bool vmx_guest_inject_ac(struct kvm_vcpu *vcpu);
361 void vmx_update_exception_bitmap(struct kvm_vcpu *vcpu);
362 bool vmx_nmi_blocked(struct kvm_vcpu *vcpu);
363 bool __vmx_interrupt_blocked(struct kvm_vcpu *vcpu);
364 bool vmx_interrupt_blocked(struct kvm_vcpu *vcpu);
365 bool vmx_get_nmi_mask(struct kvm_vcpu *vcpu);
366 void vmx_set_nmi_mask(struct kvm_vcpu *vcpu, bool masked);
367 void vmx_set_virtual_apic_mode(struct kvm_vcpu *vcpu);
368 struct vmx_uret_msr *vmx_find_uret_msr(struct vcpu_vmx *vmx, u32 msr);
369 void pt_update_intercept_for_msr(struct kvm_vcpu *vcpu);
370 void vmx_update_host_rsp(struct vcpu_vmx *vmx, unsigned long host_rsp);
371 void vmx_spec_ctrl_restore_host(struct vcpu_vmx *vmx, unsigned int flags);
372 unsigned int __vmx_vcpu_run_flags(struct vcpu_vmx *vmx);
373 bool __vmx_vcpu_run(struct vcpu_vmx *vmx, unsigned long *regs,
374 		    unsigned int flags);
375 void vmx_ept_load_pdptrs(struct kvm_vcpu *vcpu);
376 
377 void vmx_set_intercept_for_msr(struct kvm_vcpu *vcpu, u32 msr, int type, bool set);
378 
379 static inline void vmx_disable_intercept_for_msr(struct kvm_vcpu *vcpu,
380 						 u32 msr, int type)
381 {
382 	vmx_set_intercept_for_msr(vcpu, msr, type, false);
383 }
384 
385 static inline void vmx_enable_intercept_for_msr(struct kvm_vcpu *vcpu,
386 						u32 msr, int type)
387 {
388 	vmx_set_intercept_for_msr(vcpu, msr, type, true);
389 }
390 
391 u64 vmx_get_l2_tsc_offset(struct kvm_vcpu *vcpu);
392 u64 vmx_get_l2_tsc_multiplier(struct kvm_vcpu *vcpu);
393 
394 gva_t vmx_get_untagged_addr(struct kvm_vcpu *vcpu, gva_t gva, unsigned int flags);
395 
396 void vmx_update_cpu_dirty_logging(struct kvm_vcpu *vcpu);
397 
398 u64 vmx_get_supported_debugctl(struct kvm_vcpu *vcpu, bool host_initiated);
399 bool vmx_is_valid_debugctl(struct kvm_vcpu *vcpu, u64 data, bool host_initiated);
400 
401 #define VMX_HOST_OWNED_DEBUGCTL_BITS	(DEBUGCTLMSR_FREEZE_IN_SMM)
402 
403 static inline void vmx_guest_debugctl_write(struct kvm_vcpu *vcpu, u64 val)
404 {
405 	WARN_ON_ONCE(val & VMX_HOST_OWNED_DEBUGCTL_BITS);
406 
407 	val |= vcpu->arch.host_debugctl & VMX_HOST_OWNED_DEBUGCTL_BITS;
408 	vmcs_write64(GUEST_IA32_DEBUGCTL, val);
409 }
410 
411 static inline u64 vmx_guest_debugctl_read(void)
412 {
413 	return vmcs_read64(GUEST_IA32_DEBUGCTL) & ~VMX_HOST_OWNED_DEBUGCTL_BITS;
414 }
415 
416 static inline void vmx_reload_guest_debugctl(struct kvm_vcpu *vcpu)
417 {
418 	u64 val = vmcs_read64(GUEST_IA32_DEBUGCTL);
419 
420 	if (!((val ^ vcpu->arch.host_debugctl) & VMX_HOST_OWNED_DEBUGCTL_BITS))
421 		return;
422 
423 	vmx_guest_debugctl_write(vcpu, val & ~VMX_HOST_OWNED_DEBUGCTL_BITS);
424 }
425 
426 /*
427  * Note, early Intel manuals have the write-low and read-high bitmap offsets
428  * the wrong way round.  The bitmaps control MSRs 0x00000000-0x00001fff and
429  * 0xc0000000-0xc0001fff.  The former (low) uses bytes 0-0x3ff for reads and
430  * 0x800-0xbff for writes.  The latter (high) uses 0x400-0x7ff for reads and
431  * 0xc00-0xfff for writes.  MSRs not covered by either of the ranges always
432  * VM-Exit.
433  */
434 #define __BUILD_VMX_MSR_BITMAP_HELPER(rtype, action, bitop, access, base)      \
435 static inline rtype vmx_##action##_msr_bitmap_##access(unsigned long *bitmap,  \
436 						       u32 msr)		       \
437 {									       \
438 	int f = sizeof(unsigned long);					       \
439 									       \
440 	if (msr <= 0x1fff)						       \
441 		return bitop##_bit(msr, bitmap + base / f);		       \
442 	else if ((msr >= 0xc0000000) && (msr <= 0xc0001fff))		       \
443 		return bitop##_bit(msr & 0x1fff, bitmap + (base + 0x400) / f); \
444 	return (rtype)true;						       \
445 }
446 #define BUILD_VMX_MSR_BITMAP_HELPERS(ret_type, action, bitop)		       \
447 	__BUILD_VMX_MSR_BITMAP_HELPER(ret_type, action, bitop, read,  0x0)     \
448 	__BUILD_VMX_MSR_BITMAP_HELPER(ret_type, action, bitop, write, 0x800)
449 
450 BUILD_VMX_MSR_BITMAP_HELPERS(bool, test, test)
451 BUILD_VMX_MSR_BITMAP_HELPERS(void, clear, __clear)
452 BUILD_VMX_MSR_BITMAP_HELPERS(void, set, __set)
453 
454 static inline u8 vmx_get_rvi(void)
455 {
456 	return vmcs_read16(GUEST_INTR_STATUS) & 0xff;
457 }
458 
459 #define __KVM_REQUIRED_VMX_VM_ENTRY_CONTROLS				\
460 	(VM_ENTRY_LOAD_DEBUG_CONTROLS)
461 #ifdef CONFIG_X86_64
462 	#define KVM_REQUIRED_VMX_VM_ENTRY_CONTROLS			\
463 		(__KVM_REQUIRED_VMX_VM_ENTRY_CONTROLS |			\
464 		 VM_ENTRY_IA32E_MODE)
465 #else
466 	#define KVM_REQUIRED_VMX_VM_ENTRY_CONTROLS			\
467 		__KVM_REQUIRED_VMX_VM_ENTRY_CONTROLS
468 #endif
469 #define KVM_OPTIONAL_VMX_VM_ENTRY_CONTROLS				\
470 	(VM_ENTRY_LOAD_IA32_PERF_GLOBAL_CTRL |				\
471 	 VM_ENTRY_LOAD_IA32_PAT |					\
472 	 VM_ENTRY_LOAD_IA32_EFER |					\
473 	 VM_ENTRY_LOAD_BNDCFGS |					\
474 	 VM_ENTRY_PT_CONCEAL_PIP |					\
475 	 VM_ENTRY_LOAD_IA32_RTIT_CTL |					\
476 	 VM_ENTRY_LOAD_CET_STATE)
477 
478 #define __KVM_REQUIRED_VMX_VM_EXIT_CONTROLS				\
479 	(VM_EXIT_SAVE_DEBUG_CONTROLS |					\
480 	 VM_EXIT_ACK_INTR_ON_EXIT)
481 #ifdef CONFIG_X86_64
482 	#define KVM_REQUIRED_VMX_VM_EXIT_CONTROLS			\
483 		(__KVM_REQUIRED_VMX_VM_EXIT_CONTROLS |			\
484 		 VM_EXIT_HOST_ADDR_SPACE_SIZE)
485 #else
486 	#define KVM_REQUIRED_VMX_VM_EXIT_CONTROLS			\
487 		__KVM_REQUIRED_VMX_VM_EXIT_CONTROLS
488 #endif
489 #define KVM_OPTIONAL_VMX_VM_EXIT_CONTROLS				\
490 	      (VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL |			\
491 	       VM_EXIT_SAVE_IA32_PAT |					\
492 	       VM_EXIT_LOAD_IA32_PAT |					\
493 	       VM_EXIT_SAVE_IA32_EFER |					\
494 	       VM_EXIT_SAVE_VMX_PREEMPTION_TIMER |			\
495 	       VM_EXIT_LOAD_IA32_EFER |					\
496 	       VM_EXIT_CLEAR_BNDCFGS |					\
497 	       VM_EXIT_PT_CONCEAL_PIP |					\
498 	       VM_EXIT_CLEAR_IA32_RTIT_CTL |				\
499 	       VM_EXIT_LOAD_CET_STATE |					\
500 	       VM_EXIT_SAVE_IA32_PERF_GLOBAL_CTRL)
501 
502 #define KVM_REQUIRED_VMX_PIN_BASED_VM_EXEC_CONTROL			\
503 	(PIN_BASED_EXT_INTR_MASK |					\
504 	 PIN_BASED_NMI_EXITING)
505 #define KVM_OPTIONAL_VMX_PIN_BASED_VM_EXEC_CONTROL			\
506 	(PIN_BASED_VIRTUAL_NMIS |					\
507 	 PIN_BASED_POSTED_INTR |					\
508 	 PIN_BASED_VMX_PREEMPTION_TIMER)
509 
510 #define __KVM_REQUIRED_VMX_CPU_BASED_VM_EXEC_CONTROL			\
511 	(CPU_BASED_HLT_EXITING |					\
512 	 CPU_BASED_CR3_LOAD_EXITING |					\
513 	 CPU_BASED_CR3_STORE_EXITING |					\
514 	 CPU_BASED_UNCOND_IO_EXITING |					\
515 	 CPU_BASED_MOV_DR_EXITING |					\
516 	 CPU_BASED_USE_TSC_OFFSETTING |					\
517 	 CPU_BASED_MWAIT_EXITING |					\
518 	 CPU_BASED_MONITOR_EXITING |					\
519 	 CPU_BASED_INVLPG_EXITING |					\
520 	 CPU_BASED_RDPMC_EXITING |					\
521 	 CPU_BASED_INTR_WINDOW_EXITING)
522 
523 #ifdef CONFIG_X86_64
524 	#define KVM_REQUIRED_VMX_CPU_BASED_VM_EXEC_CONTROL		\
525 		(__KVM_REQUIRED_VMX_CPU_BASED_VM_EXEC_CONTROL |		\
526 		 CPU_BASED_CR8_LOAD_EXITING |				\
527 		 CPU_BASED_CR8_STORE_EXITING)
528 #else
529 	#define KVM_REQUIRED_VMX_CPU_BASED_VM_EXEC_CONTROL		\
530 		__KVM_REQUIRED_VMX_CPU_BASED_VM_EXEC_CONTROL
531 #endif
532 
533 #define KVM_OPTIONAL_VMX_CPU_BASED_VM_EXEC_CONTROL			\
534 	(CPU_BASED_RDTSC_EXITING |					\
535 	 CPU_BASED_TPR_SHADOW |						\
536 	 CPU_BASED_USE_IO_BITMAPS |					\
537 	 CPU_BASED_MONITOR_TRAP_FLAG |					\
538 	 CPU_BASED_USE_MSR_BITMAPS |					\
539 	 CPU_BASED_NMI_WINDOW_EXITING |					\
540 	 CPU_BASED_PAUSE_EXITING |					\
541 	 CPU_BASED_ACTIVATE_SECONDARY_CONTROLS |			\
542 	 CPU_BASED_ACTIVATE_TERTIARY_CONTROLS)
543 
544 #define KVM_REQUIRED_VMX_SECONDARY_VM_EXEC_CONTROL 0
545 #define KVM_OPTIONAL_VMX_SECONDARY_VM_EXEC_CONTROL			\
546 	(SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES |			\
547 	 SECONDARY_EXEC_VIRTUALIZE_X2APIC_MODE |			\
548 	 SECONDARY_EXEC_WBINVD_EXITING |				\
549 	 SECONDARY_EXEC_ENABLE_VPID |					\
550 	 SECONDARY_EXEC_ENABLE_EPT |					\
551 	 SECONDARY_EXEC_UNRESTRICTED_GUEST |				\
552 	 SECONDARY_EXEC_PAUSE_LOOP_EXITING |				\
553 	 SECONDARY_EXEC_DESC |						\
554 	 SECONDARY_EXEC_ENABLE_RDTSCP |					\
555 	 SECONDARY_EXEC_ENABLE_INVPCID |				\
556 	 SECONDARY_EXEC_APIC_REGISTER_VIRT |				\
557 	 SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY |				\
558 	 SECONDARY_EXEC_SHADOW_VMCS |					\
559 	 SECONDARY_EXEC_ENABLE_XSAVES |					\
560 	 SECONDARY_EXEC_RDSEED_EXITING |				\
561 	 SECONDARY_EXEC_RDRAND_EXITING |				\
562 	 SECONDARY_EXEC_ENABLE_PML |					\
563 	 SECONDARY_EXEC_TSC_SCALING |					\
564 	 SECONDARY_EXEC_ENABLE_USR_WAIT_PAUSE |				\
565 	 SECONDARY_EXEC_PT_USE_GPA |					\
566 	 SECONDARY_EXEC_PT_CONCEAL_VMX |				\
567 	 SECONDARY_EXEC_ENABLE_VMFUNC |					\
568 	 SECONDARY_EXEC_BUS_LOCK_DETECTION |				\
569 	 SECONDARY_EXEC_NOTIFY_VM_EXITING |				\
570 	 SECONDARY_EXEC_ENCLS_EXITING |					\
571 	 SECONDARY_EXEC_EPT_VIOLATION_VE)
572 
573 #define KVM_REQUIRED_VMX_TERTIARY_VM_EXEC_CONTROL 0
574 #define KVM_OPTIONAL_VMX_TERTIARY_VM_EXEC_CONTROL			\
575 	(TERTIARY_EXEC_IPI_VIRT)
576 
577 #define BUILD_CONTROLS_SHADOW(lname, uname, bits)						\
578 static inline void lname##_controls_set(struct vcpu_vmx *vmx, u##bits val)			\
579 {												\
580 	if (vmx->loaded_vmcs->controls_shadow.lname != val) {					\
581 		vmcs_write##bits(uname, val);							\
582 		vmx->loaded_vmcs->controls_shadow.lname = val;					\
583 	}											\
584 }												\
585 static inline u##bits __##lname##_controls_get(struct loaded_vmcs *vmcs)			\
586 {												\
587 	return vmcs->controls_shadow.lname;							\
588 }												\
589 static inline u##bits lname##_controls_get(struct vcpu_vmx *vmx)				\
590 {												\
591 	return __##lname##_controls_get(vmx->loaded_vmcs);					\
592 }												\
593 static __always_inline void lname##_controls_setbit(struct vcpu_vmx *vmx, u##bits val)		\
594 {												\
595 	BUILD_BUG_ON(!(val & (KVM_REQUIRED_VMX_##uname | KVM_OPTIONAL_VMX_##uname)));		\
596 	lname##_controls_set(vmx, lname##_controls_get(vmx) | val);				\
597 }												\
598 static __always_inline void lname##_controls_clearbit(struct vcpu_vmx *vmx, u##bits val)	\
599 {												\
600 	BUILD_BUG_ON(!(val & (KVM_REQUIRED_VMX_##uname | KVM_OPTIONAL_VMX_##uname)));		\
601 	lname##_controls_set(vmx, lname##_controls_get(vmx) & ~val);				\
602 }												\
603 static __always_inline void lname##_controls_changebit(struct vcpu_vmx *vmx, u##bits val,	\
604 						       bool set)				\
605 {												\
606 	if (set)										\
607 		lname##_controls_setbit(vmx, val);						\
608 	else											\
609 		lname##_controls_clearbit(vmx, val);						\
610 }
611 BUILD_CONTROLS_SHADOW(vm_entry, VM_ENTRY_CONTROLS, 32)
612 BUILD_CONTROLS_SHADOW(vm_exit, VM_EXIT_CONTROLS, 32)
613 BUILD_CONTROLS_SHADOW(pin, PIN_BASED_VM_EXEC_CONTROL, 32)
614 BUILD_CONTROLS_SHADOW(exec, CPU_BASED_VM_EXEC_CONTROL, 32)
615 BUILD_CONTROLS_SHADOW(secondary_exec, SECONDARY_VM_EXEC_CONTROL, 32)
616 BUILD_CONTROLS_SHADOW(tertiary_exec, TERTIARY_VM_EXEC_CONTROL, 64)
617 
618 /*
619  * VMX_REGS_LAZY_LOAD_SET - The set of registers that will be updated in the
620  * cache on demand.  Other registers not listed here are synced to
621  * the cache immediately after VM-Exit.
622  */
623 #define VMX_REGS_LAZY_LOAD_SET	((1 << VCPU_REGS_RIP) |         \
624 				(1 << VCPU_REGS_RSP) |          \
625 				(1 << VCPU_EXREG_RFLAGS) |      \
626 				(1 << VCPU_EXREG_PDPTR) |       \
627 				(1 << VCPU_EXREG_SEGMENTS) |    \
628 				(1 << VCPU_EXREG_CR0) |         \
629 				(1 << VCPU_EXREG_CR3) |         \
630 				(1 << VCPU_EXREG_CR4) |         \
631 				(1 << VCPU_EXREG_EXIT_INFO_1) | \
632 				(1 << VCPU_EXREG_EXIT_INFO_2))
633 
634 static inline unsigned long vmx_l1_guest_owned_cr0_bits(void)
635 {
636 	unsigned long bits = KVM_POSSIBLE_CR0_GUEST_BITS;
637 
638 	/*
639 	 * CR0.WP needs to be intercepted when KVM is shadowing legacy paging
640 	 * in order to construct shadow PTEs with the correct protections.
641 	 * Note!  CR0.WP technically can be passed through to the guest if
642 	 * paging is disabled, but checking CR0.PG would generate a cyclical
643 	 * dependency of sorts due to forcing the caller to ensure CR0 holds
644 	 * the correct value prior to determining which CR0 bits can be owned
645 	 * by L1.  Keep it simple and limit the optimization to EPT.
646 	 */
647 	if (!enable_ept)
648 		bits &= ~X86_CR0_WP;
649 	return bits;
650 }
651 
652 static __always_inline struct kvm_vmx *to_kvm_vmx(struct kvm *kvm)
653 {
654 	return container_of(kvm, struct kvm_vmx, kvm);
655 }
656 
657 static __always_inline struct vcpu_vmx *to_vmx(struct kvm_vcpu *vcpu)
658 {
659 	return container_of(vcpu, struct vcpu_vmx, vcpu);
660 }
661 
662 void intel_pmu_cross_mapped_check(struct kvm_pmu *pmu);
663 int intel_pmu_create_guest_lbr_event(struct kvm_vcpu *vcpu);
664 void vmx_passthrough_lbr_msrs(struct kvm_vcpu *vcpu);
665 
666 struct vmcs *alloc_vmcs_cpu(bool shadow, int cpu, gfp_t flags);
667 void free_vmcs(struct vmcs *vmcs);
668 int alloc_loaded_vmcs(struct loaded_vmcs *loaded_vmcs);
669 void free_loaded_vmcs(struct loaded_vmcs *loaded_vmcs);
670 
671 static inline struct vmcs *alloc_vmcs(bool shadow)
672 {
673 	return alloc_vmcs_cpu(shadow, raw_smp_processor_id(),
674 			      GFP_KERNEL_ACCOUNT);
675 }
676 
677 static inline bool vmx_has_waitpkg(struct vcpu_vmx *vmx)
678 {
679 	return secondary_exec_controls_get(vmx) &
680 		SECONDARY_EXEC_ENABLE_USR_WAIT_PAUSE;
681 }
682 
683 static inline bool vmx_need_pf_intercept(struct kvm_vcpu *vcpu)
684 {
685 	if (!enable_ept)
686 		return true;
687 
688 	return allow_smaller_maxphyaddr &&
689 	       cpuid_maxphyaddr(vcpu) < kvm_host.maxphyaddr;
690 }
691 
692 static inline bool is_unrestricted_guest(struct kvm_vcpu *vcpu)
693 {
694 	return enable_unrestricted_guest && (!is_guest_mode(vcpu) ||
695 	    (secondary_exec_controls_get(to_vmx(vcpu)) &
696 	    SECONDARY_EXEC_UNRESTRICTED_GUEST));
697 }
698 
699 bool __vmx_guest_state_valid(struct kvm_vcpu *vcpu);
700 static inline bool vmx_guest_state_valid(struct kvm_vcpu *vcpu)
701 {
702 	return is_unrestricted_guest(vcpu) || __vmx_guest_state_valid(vcpu);
703 }
704 
705 void dump_vmcs(struct kvm_vcpu *vcpu);
706 
707 static inline int vmx_get_instr_info_reg(u32 vmx_instr_info)
708 {
709 	return (vmx_instr_info >> 3) & 0xf;
710 }
711 
712 static inline int vmx_get_instr_info_reg2(u32 vmx_instr_info)
713 {
714 	return (vmx_instr_info >> 28) & 0xf;
715 }
716 
717 static inline bool vmx_can_use_ipiv(struct kvm_vcpu *vcpu)
718 {
719 	return  lapic_in_kernel(vcpu) && enable_ipiv;
720 }
721 
722 static inline void vmx_segment_cache_clear(struct vcpu_vmx *vmx)
723 {
724 	vmx->segment_cache.bitmask = 0;
725 }
726 
727 int vmx_init(void);
728 void vmx_exit(void);
729 
730 #endif /* __KVM_X86_VMX_H */
731