xref: /linux/arch/x86/kvm/vmx/vmx.h (revision fe7fad476ec8153a8b8767a08114e3e4a58a837e)
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 "vmcs.h"
15 #include "vmx_ops.h"
16 #include "../cpuid.h"
17 #include "run_flags.h"
18 #include "../mmu.h"
19 
20 #define X2APIC_MSR(r) (APIC_BASE_MSR + ((r) >> 4))
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 union vmx_exit_reason {
71 	struct {
72 		u32	basic			: 16;
73 		u32	reserved16		: 1;
74 		u32	reserved17		: 1;
75 		u32	reserved18		: 1;
76 		u32	reserved19		: 1;
77 		u32	reserved20		: 1;
78 		u32	reserved21		: 1;
79 		u32	reserved22		: 1;
80 		u32	reserved23		: 1;
81 		u32	reserved24		: 1;
82 		u32	reserved25		: 1;
83 		u32	bus_lock_detected	: 1;
84 		u32	enclave_mode		: 1;
85 		u32	smi_pending_mtf		: 1;
86 		u32	smi_from_vmx_root	: 1;
87 		u32	reserved30		: 1;
88 		u32	failed_vmentry		: 1;
89 	};
90 	u32 full;
91 };
92 
93 struct lbr_desc {
94 	/* Basic info about guest LBR records. */
95 	struct x86_pmu_lbr records;
96 
97 	/*
98 	 * Emulate LBR feature via passthrough LBR registers when the
99 	 * per-vcpu guest LBR event is scheduled on the current pcpu.
100 	 *
101 	 * The records may be inaccurate if the host reclaims the LBR.
102 	 */
103 	struct perf_event *event;
104 
105 	/* True if LBRs are marked as not intercepted in the MSR bitmap */
106 	bool msr_passthrough;
107 };
108 
109 extern struct x86_pmu_lbr vmx_lbr_caps;
110 
111 /*
112  * The nested_vmx structure is part of vcpu_vmx, and holds information we need
113  * for correct emulation of VMX (i.e., nested VMX) on this vcpu.
114  */
115 struct nested_vmx {
116 	/* Has the level1 guest done vmxon? */
117 	bool vmxon;
118 	gpa_t vmxon_ptr;
119 	bool pml_full;
120 
121 	/* The guest-physical address of the current VMCS L1 keeps for L2 */
122 	gpa_t current_vmptr;
123 	/*
124 	 * Cache of the guest's VMCS, existing outside of guest memory.
125 	 * Loaded from guest memory during VMPTRLD. Flushed to guest
126 	 * memory during VMCLEAR and VMPTRLD.
127 	 */
128 	struct vmcs12 *cached_vmcs12;
129 	/*
130 	 * Cache of the guest's shadow VMCS, existing outside of guest
131 	 * memory. Loaded from guest memory during VM entry. Flushed
132 	 * to guest memory during VM exit.
133 	 */
134 	struct vmcs12 *cached_shadow_vmcs12;
135 
136 	/*
137 	 * GPA to HVA cache for accessing vmcs12->vmcs_link_pointer
138 	 */
139 	struct gfn_to_hva_cache shadow_vmcs12_cache;
140 
141 	/*
142 	 * GPA to HVA cache for VMCS12
143 	 */
144 	struct gfn_to_hva_cache vmcs12_cache;
145 
146 	/*
147 	 * Indicates if the shadow vmcs or enlightened vmcs must be updated
148 	 * with the data held by struct vmcs12.
149 	 */
150 	bool need_vmcs12_to_shadow_sync;
151 	bool dirty_vmcs12;
152 
153 	/*
154 	 * Indicates whether MSR bitmap for L2 needs to be rebuilt due to
155 	 * changes in MSR bitmap for L1 or switching to a different L2. Note,
156 	 * this flag can only be used reliably in conjunction with a paravirt L1
157 	 * which informs L0 whether any changes to MSR bitmap for L2 were done
158 	 * on its side.
159 	 */
160 	bool force_msr_bitmap_recalc;
161 
162 	/*
163 	 * Indicates lazily loaded guest state has not yet been decached from
164 	 * vmcs02.
165 	 */
166 	bool need_sync_vmcs02_to_vmcs12_rare;
167 
168 	/*
169 	 * vmcs02 has been initialized, i.e. state that is constant for
170 	 * vmcs02 has been written to the backing VMCS.  Initialization
171 	 * is delayed until L1 actually attempts to run a nested VM.
172 	 */
173 	bool vmcs02_initialized;
174 
175 	bool change_vmcs01_virtual_apic_mode;
176 	bool reload_vmcs01_apic_access_page;
177 	bool update_vmcs01_cpu_dirty_logging;
178 	bool update_vmcs01_apicv_status;
179 	bool update_vmcs01_hwapic_isr;
180 
181 	/*
182 	 * Enlightened VMCS has been enabled. It does not mean that L1 has to
183 	 * use it. However, VMX features available to L1 will be limited based
184 	 * on what the enlightened VMCS supports.
185 	 */
186 	bool enlightened_vmcs_enabled;
187 
188 	/* L2 must run next, and mustn't decide to exit to L1. */
189 	bool nested_run_pending;
190 
191 	/* Pending MTF VM-exit into L1.  */
192 	bool mtf_pending;
193 
194 	struct loaded_vmcs vmcs02;
195 
196 	/*
197 	 * Guest pages referred to in the vmcs02 with host-physical
198 	 * pointers, so we must keep them pinned while L2 runs.
199 	 */
200 	struct kvm_host_map apic_access_page_map;
201 	struct kvm_host_map virtual_apic_map;
202 	struct kvm_host_map pi_desc_map;
203 
204 	struct pi_desc *pi_desc;
205 	bool pi_pending;
206 	u16 posted_intr_nv;
207 
208 	struct hrtimer preemption_timer;
209 	u64 preemption_timer_deadline;
210 	bool has_preemption_timer_deadline;
211 	bool preemption_timer_expired;
212 
213 	/*
214 	 * Used to snapshot MSRs that are conditionally loaded on VM-Enter in
215 	 * order to propagate the guest's pre-VM-Enter value into vmcs02.  For
216 	 * emulation of VMLAUNCH/VMRESUME, the snapshot will be of L1's value.
217 	 * For KVM_SET_NESTED_STATE, the snapshot is of L2's value, _if_
218 	 * userspace restores MSRs before nested state.  If userspace restores
219 	 * MSRs after nested state, the snapshot holds garbage, but KVM can't
220 	 * detect that, and the garbage value in vmcs02 will be overwritten by
221 	 * MSR restoration in any case.
222 	 */
223 	u64 pre_vmenter_debugctl;
224 	u64 pre_vmenter_bndcfgs;
225 
226 	/* to migrate it to L1 if L2 writes to L1's CR8 directly */
227 	int l1_tpr_threshold;
228 
229 	u16 vpid02;
230 	u16 last_vpid;
231 
232 	struct nested_vmx_msrs msrs;
233 
234 	/* SMM related state */
235 	struct {
236 		/* in VMX operation on SMM entry? */
237 		bool vmxon;
238 		/* in guest mode on SMM entry? */
239 		bool guest_mode;
240 	} smm;
241 
242 #ifdef CONFIG_KVM_HYPERV
243 	gpa_t hv_evmcs_vmptr;
244 	struct kvm_host_map hv_evmcs_map;
245 	struct hv_enlightened_vmcs *hv_evmcs;
246 #endif
247 };
248 
249 struct vcpu_vmx {
250 	struct kvm_vcpu       vcpu;
251 	u8                    fail;
252 	u8		      x2apic_msr_bitmap_mode;
253 
254 	/*
255 	 * If true, host state has been stored in vmx->loaded_vmcs for
256 	 * the CPU registers that only need to be switched when transitioning
257 	 * to/from the kernel, and the registers have been loaded with guest
258 	 * values.  If false, host state is loaded in the CPU registers
259 	 * and vmx->loaded_vmcs->host_state is invalid.
260 	 */
261 	bool		      guest_state_loaded;
262 
263 	unsigned long         exit_qualification;
264 	u32                   exit_intr_info;
265 	u32                   idt_vectoring_info;
266 	ulong                 rflags;
267 
268 	/*
269 	 * User return MSRs are always emulated when enabled in the guest, but
270 	 * only loaded into hardware when necessary, e.g. SYSCALL #UDs outside
271 	 * of 64-bit mode or if EFER.SCE=1, thus the SYSCALL MSRs don't need to
272 	 * be loaded into hardware if those conditions aren't met.
273 	 */
274 	struct vmx_uret_msr   guest_uret_msrs[MAX_NR_USER_RETURN_MSRS];
275 	bool                  guest_uret_msrs_loaded;
276 #ifdef CONFIG_X86_64
277 	u64		      msr_host_kernel_gs_base;
278 	u64		      msr_guest_kernel_gs_base;
279 #endif
280 
281 	u64		      spec_ctrl;
282 	u32		      msr_ia32_umwait_control;
283 
284 	/*
285 	 * loaded_vmcs points to the VMCS currently used in this vcpu. For a
286 	 * non-nested (L1) guest, it always points to vmcs01. For a nested
287 	 * guest (L2), it points to a different VMCS.
288 	 */
289 	struct loaded_vmcs    vmcs01;
290 	struct loaded_vmcs   *loaded_vmcs;
291 
292 	struct msr_autoload {
293 		struct vmx_msrs guest;
294 		struct vmx_msrs host;
295 	} msr_autoload;
296 
297 	struct msr_autostore {
298 		struct vmx_msrs guest;
299 	} msr_autostore;
300 
301 	struct {
302 		int vm86_active;
303 		ulong save_rflags;
304 		struct kvm_segment segs[8];
305 	} rmode;
306 	struct {
307 		u32 bitmask; /* 4 bits per segment (1 bit per field) */
308 		struct kvm_save_segment {
309 			u16 selector;
310 			unsigned long base;
311 			u32 limit;
312 			u32 ar;
313 		} seg[8];
314 	} segment_cache;
315 	int vpid;
316 	bool emulation_required;
317 
318 	union vmx_exit_reason exit_reason;
319 
320 	/* Posted interrupt descriptor */
321 	struct pi_desc pi_desc;
322 
323 	/* Used if this vCPU is waiting for PI notification wakeup. */
324 	struct list_head pi_wakeup_list;
325 
326 	/* Support for a guest hypervisor (nested VMX) */
327 	struct nested_vmx nested;
328 
329 	/* Dynamic PLE window. */
330 	unsigned int ple_window;
331 	bool ple_window_dirty;
332 
333 	/* Support for PML */
334 #define PML_LOG_NR_ENTRIES	512
335 	/* PML is written backwards: this is the first entry written by the CPU */
336 #define PML_HEAD_INDEX		(PML_LOG_NR_ENTRIES-1)
337 
338 	struct page *pml_pg;
339 
340 	/* apic deadline value in host tsc */
341 	u64 hv_deadline_tsc;
342 
343 	unsigned long host_debugctlmsr;
344 
345 	/*
346 	 * Only bits masked by msr_ia32_feature_control_valid_bits can be set in
347 	 * msr_ia32_feature_control. FEAT_CTL_LOCKED is always included
348 	 * in msr_ia32_feature_control_valid_bits.
349 	 */
350 	u64 msr_ia32_feature_control;
351 	u64 msr_ia32_feature_control_valid_bits;
352 	/* SGX Launch Control public key hash */
353 	u64 msr_ia32_sgxlepubkeyhash[4];
354 	u64 msr_ia32_mcu_opt_ctrl;
355 	bool disable_fb_clear;
356 
357 	struct pt_desc pt_desc;
358 	struct lbr_desc lbr_desc;
359 
360 	/* Save desired MSR intercept (read: pass-through) state */
361 #define MAX_POSSIBLE_PASSTHROUGH_MSRS	16
362 	struct {
363 		DECLARE_BITMAP(read, MAX_POSSIBLE_PASSTHROUGH_MSRS);
364 		DECLARE_BITMAP(write, MAX_POSSIBLE_PASSTHROUGH_MSRS);
365 	} shadow_msr_intercept;
366 
367 	/* ve_info must be page aligned. */
368 	struct vmx_ve_information *ve_info;
369 };
370 
371 struct kvm_vmx {
372 	struct kvm kvm;
373 
374 	unsigned int tss_addr;
375 	bool ept_identity_pagetable_done;
376 	gpa_t ept_identity_map_addr;
377 	/* Posted Interrupt Descriptor (PID) table for IPI virtualization */
378 	u64 *pid_table;
379 };
380 
381 void vmx_vcpu_load_vmcs(struct kvm_vcpu *vcpu, int cpu,
382 			struct loaded_vmcs *buddy);
383 int allocate_vpid(void);
384 void free_vpid(int vpid);
385 void vmx_set_constant_host_state(struct vcpu_vmx *vmx);
386 void vmx_prepare_switch_to_guest(struct kvm_vcpu *vcpu);
387 void vmx_set_host_fs_gs(struct vmcs_host_state *host, u16 fs_sel, u16 gs_sel,
388 			unsigned long fs_base, unsigned long gs_base);
389 int vmx_get_cpl(struct kvm_vcpu *vcpu);
390 int vmx_get_cpl_no_cache(struct kvm_vcpu *vcpu);
391 bool vmx_emulation_required(struct kvm_vcpu *vcpu);
392 unsigned long vmx_get_rflags(struct kvm_vcpu *vcpu);
393 void vmx_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
394 u32 vmx_get_interrupt_shadow(struct kvm_vcpu *vcpu);
395 void vmx_set_interrupt_shadow(struct kvm_vcpu *vcpu, int mask);
396 int vmx_set_efer(struct kvm_vcpu *vcpu, u64 efer);
397 void vmx_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
398 void vmx_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
399 void set_cr4_guest_host_mask(struct vcpu_vmx *vmx);
400 void ept_save_pdptrs(struct kvm_vcpu *vcpu);
401 void vmx_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
402 void __vmx_set_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
403 u64 construct_eptp(struct kvm_vcpu *vcpu, hpa_t root_hpa, int root_level);
404 
405 bool vmx_guest_inject_ac(struct kvm_vcpu *vcpu);
406 void vmx_update_exception_bitmap(struct kvm_vcpu *vcpu);
407 bool vmx_nmi_blocked(struct kvm_vcpu *vcpu);
408 bool __vmx_interrupt_blocked(struct kvm_vcpu *vcpu);
409 bool vmx_interrupt_blocked(struct kvm_vcpu *vcpu);
410 bool vmx_get_nmi_mask(struct kvm_vcpu *vcpu);
411 void vmx_set_nmi_mask(struct kvm_vcpu *vcpu, bool masked);
412 void vmx_set_virtual_apic_mode(struct kvm_vcpu *vcpu);
413 struct vmx_uret_msr *vmx_find_uret_msr(struct vcpu_vmx *vmx, u32 msr);
414 void pt_update_intercept_for_msr(struct kvm_vcpu *vcpu);
415 void vmx_update_host_rsp(struct vcpu_vmx *vmx, unsigned long host_rsp);
416 void vmx_spec_ctrl_restore_host(struct vcpu_vmx *vmx, unsigned int flags);
417 unsigned int __vmx_vcpu_run_flags(struct vcpu_vmx *vmx);
418 bool __vmx_vcpu_run(struct vcpu_vmx *vmx, unsigned long *regs,
419 		    unsigned int flags);
420 int vmx_find_loadstore_msr_slot(struct vmx_msrs *m, u32 msr);
421 void vmx_ept_load_pdptrs(struct kvm_vcpu *vcpu);
422 
423 void vmx_disable_intercept_for_msr(struct kvm_vcpu *vcpu, u32 msr, int type);
424 void vmx_enable_intercept_for_msr(struct kvm_vcpu *vcpu, u32 msr, int type);
425 
426 u64 vmx_get_l2_tsc_offset(struct kvm_vcpu *vcpu);
427 u64 vmx_get_l2_tsc_multiplier(struct kvm_vcpu *vcpu);
428 
429 gva_t vmx_get_untagged_addr(struct kvm_vcpu *vcpu, gva_t gva, unsigned int flags);
430 
431 static inline void vmx_set_intercept_for_msr(struct kvm_vcpu *vcpu, u32 msr,
432 					     int type, bool value)
433 {
434 	if (value)
435 		vmx_enable_intercept_for_msr(vcpu, msr, type);
436 	else
437 		vmx_disable_intercept_for_msr(vcpu, msr, type);
438 }
439 
440 void vmx_update_cpu_dirty_logging(struct kvm_vcpu *vcpu);
441 
442 /*
443  * Note, early Intel manuals have the write-low and read-high bitmap offsets
444  * the wrong way round.  The bitmaps control MSRs 0x00000000-0x00001fff and
445  * 0xc0000000-0xc0001fff.  The former (low) uses bytes 0-0x3ff for reads and
446  * 0x800-0xbff for writes.  The latter (high) uses 0x400-0x7ff for reads and
447  * 0xc00-0xfff for writes.  MSRs not covered by either of the ranges always
448  * VM-Exit.
449  */
450 #define __BUILD_VMX_MSR_BITMAP_HELPER(rtype, action, bitop, access, base)      \
451 static inline rtype vmx_##action##_msr_bitmap_##access(unsigned long *bitmap,  \
452 						       u32 msr)		       \
453 {									       \
454 	int f = sizeof(unsigned long);					       \
455 									       \
456 	if (msr <= 0x1fff)						       \
457 		return bitop##_bit(msr, bitmap + base / f);		       \
458 	else if ((msr >= 0xc0000000) && (msr <= 0xc0001fff))		       \
459 		return bitop##_bit(msr & 0x1fff, bitmap + (base + 0x400) / f); \
460 	return (rtype)true;						       \
461 }
462 #define BUILD_VMX_MSR_BITMAP_HELPERS(ret_type, action, bitop)		       \
463 	__BUILD_VMX_MSR_BITMAP_HELPER(ret_type, action, bitop, read,  0x0)     \
464 	__BUILD_VMX_MSR_BITMAP_HELPER(ret_type, action, bitop, write, 0x800)
465 
466 BUILD_VMX_MSR_BITMAP_HELPERS(bool, test, test)
467 BUILD_VMX_MSR_BITMAP_HELPERS(void, clear, __clear)
468 BUILD_VMX_MSR_BITMAP_HELPERS(void, set, __set)
469 
470 static inline u8 vmx_get_rvi(void)
471 {
472 	return vmcs_read16(GUEST_INTR_STATUS) & 0xff;
473 }
474 
475 #define __KVM_REQUIRED_VMX_VM_ENTRY_CONTROLS				\
476 	(VM_ENTRY_LOAD_DEBUG_CONTROLS)
477 #ifdef CONFIG_X86_64
478 	#define KVM_REQUIRED_VMX_VM_ENTRY_CONTROLS			\
479 		(__KVM_REQUIRED_VMX_VM_ENTRY_CONTROLS |			\
480 		 VM_ENTRY_IA32E_MODE)
481 #else
482 	#define KVM_REQUIRED_VMX_VM_ENTRY_CONTROLS			\
483 		__KVM_REQUIRED_VMX_VM_ENTRY_CONTROLS
484 #endif
485 #define KVM_OPTIONAL_VMX_VM_ENTRY_CONTROLS				\
486 	(VM_ENTRY_LOAD_IA32_PERF_GLOBAL_CTRL |				\
487 	 VM_ENTRY_LOAD_IA32_PAT |					\
488 	 VM_ENTRY_LOAD_IA32_EFER |					\
489 	 VM_ENTRY_LOAD_BNDCFGS |					\
490 	 VM_ENTRY_PT_CONCEAL_PIP |					\
491 	 VM_ENTRY_LOAD_IA32_RTIT_CTL)
492 
493 #define __KVM_REQUIRED_VMX_VM_EXIT_CONTROLS				\
494 	(VM_EXIT_SAVE_DEBUG_CONTROLS |					\
495 	 VM_EXIT_ACK_INTR_ON_EXIT)
496 #ifdef CONFIG_X86_64
497 	#define KVM_REQUIRED_VMX_VM_EXIT_CONTROLS			\
498 		(__KVM_REQUIRED_VMX_VM_EXIT_CONTROLS |			\
499 		 VM_EXIT_HOST_ADDR_SPACE_SIZE)
500 #else
501 	#define KVM_REQUIRED_VMX_VM_EXIT_CONTROLS			\
502 		__KVM_REQUIRED_VMX_VM_EXIT_CONTROLS
503 #endif
504 #define KVM_OPTIONAL_VMX_VM_EXIT_CONTROLS				\
505 	      (VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL |			\
506 	       VM_EXIT_SAVE_IA32_PAT |					\
507 	       VM_EXIT_LOAD_IA32_PAT |					\
508 	       VM_EXIT_SAVE_IA32_EFER |					\
509 	       VM_EXIT_SAVE_VMX_PREEMPTION_TIMER |			\
510 	       VM_EXIT_LOAD_IA32_EFER |					\
511 	       VM_EXIT_CLEAR_BNDCFGS |					\
512 	       VM_EXIT_PT_CONCEAL_PIP |					\
513 	       VM_EXIT_CLEAR_IA32_RTIT_CTL)
514 
515 #define KVM_REQUIRED_VMX_PIN_BASED_VM_EXEC_CONTROL			\
516 	(PIN_BASED_EXT_INTR_MASK |					\
517 	 PIN_BASED_NMI_EXITING)
518 #define KVM_OPTIONAL_VMX_PIN_BASED_VM_EXEC_CONTROL			\
519 	(PIN_BASED_VIRTUAL_NMIS |					\
520 	 PIN_BASED_POSTED_INTR |					\
521 	 PIN_BASED_VMX_PREEMPTION_TIMER)
522 
523 #define __KVM_REQUIRED_VMX_CPU_BASED_VM_EXEC_CONTROL			\
524 	(CPU_BASED_HLT_EXITING |					\
525 	 CPU_BASED_CR3_LOAD_EXITING |					\
526 	 CPU_BASED_CR3_STORE_EXITING |					\
527 	 CPU_BASED_UNCOND_IO_EXITING |					\
528 	 CPU_BASED_MOV_DR_EXITING |					\
529 	 CPU_BASED_USE_TSC_OFFSETTING |					\
530 	 CPU_BASED_MWAIT_EXITING |					\
531 	 CPU_BASED_MONITOR_EXITING |					\
532 	 CPU_BASED_INVLPG_EXITING |					\
533 	 CPU_BASED_RDPMC_EXITING |					\
534 	 CPU_BASED_INTR_WINDOW_EXITING)
535 
536 #ifdef CONFIG_X86_64
537 	#define KVM_REQUIRED_VMX_CPU_BASED_VM_EXEC_CONTROL		\
538 		(__KVM_REQUIRED_VMX_CPU_BASED_VM_EXEC_CONTROL |		\
539 		 CPU_BASED_CR8_LOAD_EXITING |				\
540 		 CPU_BASED_CR8_STORE_EXITING)
541 #else
542 	#define KVM_REQUIRED_VMX_CPU_BASED_VM_EXEC_CONTROL		\
543 		__KVM_REQUIRED_VMX_CPU_BASED_VM_EXEC_CONTROL
544 #endif
545 
546 #define KVM_OPTIONAL_VMX_CPU_BASED_VM_EXEC_CONTROL			\
547 	(CPU_BASED_RDTSC_EXITING |					\
548 	 CPU_BASED_TPR_SHADOW |						\
549 	 CPU_BASED_USE_IO_BITMAPS |					\
550 	 CPU_BASED_MONITOR_TRAP_FLAG |					\
551 	 CPU_BASED_USE_MSR_BITMAPS |					\
552 	 CPU_BASED_NMI_WINDOW_EXITING |					\
553 	 CPU_BASED_PAUSE_EXITING |					\
554 	 CPU_BASED_ACTIVATE_SECONDARY_CONTROLS |			\
555 	 CPU_BASED_ACTIVATE_TERTIARY_CONTROLS)
556 
557 #define KVM_REQUIRED_VMX_SECONDARY_VM_EXEC_CONTROL 0
558 #define KVM_OPTIONAL_VMX_SECONDARY_VM_EXEC_CONTROL			\
559 	(SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES |			\
560 	 SECONDARY_EXEC_VIRTUALIZE_X2APIC_MODE |			\
561 	 SECONDARY_EXEC_WBINVD_EXITING |				\
562 	 SECONDARY_EXEC_ENABLE_VPID |					\
563 	 SECONDARY_EXEC_ENABLE_EPT |					\
564 	 SECONDARY_EXEC_UNRESTRICTED_GUEST |				\
565 	 SECONDARY_EXEC_PAUSE_LOOP_EXITING |				\
566 	 SECONDARY_EXEC_DESC |						\
567 	 SECONDARY_EXEC_ENABLE_RDTSCP |					\
568 	 SECONDARY_EXEC_ENABLE_INVPCID |				\
569 	 SECONDARY_EXEC_APIC_REGISTER_VIRT |				\
570 	 SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY |				\
571 	 SECONDARY_EXEC_SHADOW_VMCS |					\
572 	 SECONDARY_EXEC_ENABLE_XSAVES |					\
573 	 SECONDARY_EXEC_RDSEED_EXITING |				\
574 	 SECONDARY_EXEC_RDRAND_EXITING |				\
575 	 SECONDARY_EXEC_ENABLE_PML |					\
576 	 SECONDARY_EXEC_TSC_SCALING |					\
577 	 SECONDARY_EXEC_ENABLE_USR_WAIT_PAUSE |				\
578 	 SECONDARY_EXEC_PT_USE_GPA |					\
579 	 SECONDARY_EXEC_PT_CONCEAL_VMX |				\
580 	 SECONDARY_EXEC_ENABLE_VMFUNC |					\
581 	 SECONDARY_EXEC_BUS_LOCK_DETECTION |				\
582 	 SECONDARY_EXEC_NOTIFY_VM_EXITING |				\
583 	 SECONDARY_EXEC_ENCLS_EXITING |					\
584 	 SECONDARY_EXEC_EPT_VIOLATION_VE)
585 
586 #define KVM_REQUIRED_VMX_TERTIARY_VM_EXEC_CONTROL 0
587 #define KVM_OPTIONAL_VMX_TERTIARY_VM_EXEC_CONTROL			\
588 	(TERTIARY_EXEC_IPI_VIRT)
589 
590 #define BUILD_CONTROLS_SHADOW(lname, uname, bits)						\
591 static inline void lname##_controls_set(struct vcpu_vmx *vmx, u##bits val)			\
592 {												\
593 	if (vmx->loaded_vmcs->controls_shadow.lname != val) {					\
594 		vmcs_write##bits(uname, val);							\
595 		vmx->loaded_vmcs->controls_shadow.lname = val;					\
596 	}											\
597 }												\
598 static inline u##bits __##lname##_controls_get(struct loaded_vmcs *vmcs)			\
599 {												\
600 	return vmcs->controls_shadow.lname;							\
601 }												\
602 static inline u##bits lname##_controls_get(struct vcpu_vmx *vmx)				\
603 {												\
604 	return __##lname##_controls_get(vmx->loaded_vmcs);					\
605 }												\
606 static __always_inline void lname##_controls_setbit(struct vcpu_vmx *vmx, u##bits val)		\
607 {												\
608 	BUILD_BUG_ON(!(val & (KVM_REQUIRED_VMX_##uname | KVM_OPTIONAL_VMX_##uname)));		\
609 	lname##_controls_set(vmx, lname##_controls_get(vmx) | val);				\
610 }												\
611 static __always_inline void lname##_controls_clearbit(struct vcpu_vmx *vmx, u##bits val)	\
612 {												\
613 	BUILD_BUG_ON(!(val & (KVM_REQUIRED_VMX_##uname | KVM_OPTIONAL_VMX_##uname)));		\
614 	lname##_controls_set(vmx, lname##_controls_get(vmx) & ~val);				\
615 }
616 BUILD_CONTROLS_SHADOW(vm_entry, VM_ENTRY_CONTROLS, 32)
617 BUILD_CONTROLS_SHADOW(vm_exit, VM_EXIT_CONTROLS, 32)
618 BUILD_CONTROLS_SHADOW(pin, PIN_BASED_VM_EXEC_CONTROL, 32)
619 BUILD_CONTROLS_SHADOW(exec, CPU_BASED_VM_EXEC_CONTROL, 32)
620 BUILD_CONTROLS_SHADOW(secondary_exec, SECONDARY_VM_EXEC_CONTROL, 32)
621 BUILD_CONTROLS_SHADOW(tertiary_exec, TERTIARY_VM_EXEC_CONTROL, 64)
622 
623 /*
624  * VMX_REGS_LAZY_LOAD_SET - The set of registers that will be updated in the
625  * cache on demand.  Other registers not listed here are synced to
626  * the cache immediately after VM-Exit.
627  */
628 #define VMX_REGS_LAZY_LOAD_SET	((1 << VCPU_REGS_RIP) |         \
629 				(1 << VCPU_REGS_RSP) |          \
630 				(1 << VCPU_EXREG_RFLAGS) |      \
631 				(1 << VCPU_EXREG_PDPTR) |       \
632 				(1 << VCPU_EXREG_SEGMENTS) |    \
633 				(1 << VCPU_EXREG_CR0) |         \
634 				(1 << VCPU_EXREG_CR3) |         \
635 				(1 << VCPU_EXREG_CR4) |         \
636 				(1 << VCPU_EXREG_EXIT_INFO_1) | \
637 				(1 << VCPU_EXREG_EXIT_INFO_2))
638 
639 static inline unsigned long vmx_l1_guest_owned_cr0_bits(void)
640 {
641 	unsigned long bits = KVM_POSSIBLE_CR0_GUEST_BITS;
642 
643 	/*
644 	 * CR0.WP needs to be intercepted when KVM is shadowing legacy paging
645 	 * in order to construct shadow PTEs with the correct protections.
646 	 * Note!  CR0.WP technically can be passed through to the guest if
647 	 * paging is disabled, but checking CR0.PG would generate a cyclical
648 	 * dependency of sorts due to forcing the caller to ensure CR0 holds
649 	 * the correct value prior to determining which CR0 bits can be owned
650 	 * by L1.  Keep it simple and limit the optimization to EPT.
651 	 */
652 	if (!enable_ept)
653 		bits &= ~X86_CR0_WP;
654 	return bits;
655 }
656 
657 static __always_inline struct kvm_vmx *to_kvm_vmx(struct kvm *kvm)
658 {
659 	return container_of(kvm, struct kvm_vmx, kvm);
660 }
661 
662 static __always_inline struct vcpu_vmx *to_vmx(struct kvm_vcpu *vcpu)
663 {
664 	return container_of(vcpu, struct vcpu_vmx, vcpu);
665 }
666 
667 static inline struct lbr_desc *vcpu_to_lbr_desc(struct kvm_vcpu *vcpu)
668 {
669 	return &to_vmx(vcpu)->lbr_desc;
670 }
671 
672 static inline struct x86_pmu_lbr *vcpu_to_lbr_records(struct kvm_vcpu *vcpu)
673 {
674 	return &vcpu_to_lbr_desc(vcpu)->records;
675 }
676 
677 static inline bool intel_pmu_lbr_is_enabled(struct kvm_vcpu *vcpu)
678 {
679 	return !!vcpu_to_lbr_records(vcpu)->nr;
680 }
681 
682 void intel_pmu_cross_mapped_check(struct kvm_pmu *pmu);
683 int intel_pmu_create_guest_lbr_event(struct kvm_vcpu *vcpu);
684 void vmx_passthrough_lbr_msrs(struct kvm_vcpu *vcpu);
685 
686 static __always_inline unsigned long vmx_get_exit_qual(struct kvm_vcpu *vcpu)
687 {
688 	struct vcpu_vmx *vmx = to_vmx(vcpu);
689 
690 	if (!kvm_register_test_and_mark_available(vcpu, VCPU_EXREG_EXIT_INFO_1))
691 		vmx->exit_qualification = vmcs_readl(EXIT_QUALIFICATION);
692 
693 	return vmx->exit_qualification;
694 }
695 
696 static __always_inline u32 vmx_get_intr_info(struct kvm_vcpu *vcpu)
697 {
698 	struct vcpu_vmx *vmx = to_vmx(vcpu);
699 
700 	if (!kvm_register_test_and_mark_available(vcpu, VCPU_EXREG_EXIT_INFO_2))
701 		vmx->exit_intr_info = vmcs_read32(VM_EXIT_INTR_INFO);
702 
703 	return vmx->exit_intr_info;
704 }
705 
706 struct vmcs *alloc_vmcs_cpu(bool shadow, int cpu, gfp_t flags);
707 void free_vmcs(struct vmcs *vmcs);
708 int alloc_loaded_vmcs(struct loaded_vmcs *loaded_vmcs);
709 void free_loaded_vmcs(struct loaded_vmcs *loaded_vmcs);
710 void loaded_vmcs_clear(struct loaded_vmcs *loaded_vmcs);
711 
712 static inline struct vmcs *alloc_vmcs(bool shadow)
713 {
714 	return alloc_vmcs_cpu(shadow, raw_smp_processor_id(),
715 			      GFP_KERNEL_ACCOUNT);
716 }
717 
718 static inline bool vmx_has_waitpkg(struct vcpu_vmx *vmx)
719 {
720 	return secondary_exec_controls_get(vmx) &
721 		SECONDARY_EXEC_ENABLE_USR_WAIT_PAUSE;
722 }
723 
724 static inline bool vmx_need_pf_intercept(struct kvm_vcpu *vcpu)
725 {
726 	if (!enable_ept)
727 		return true;
728 
729 	return allow_smaller_maxphyaddr &&
730 	       cpuid_maxphyaddr(vcpu) < kvm_host.maxphyaddr;
731 }
732 
733 static inline bool is_unrestricted_guest(struct kvm_vcpu *vcpu)
734 {
735 	return enable_unrestricted_guest && (!is_guest_mode(vcpu) ||
736 	    (secondary_exec_controls_get(to_vmx(vcpu)) &
737 	    SECONDARY_EXEC_UNRESTRICTED_GUEST));
738 }
739 
740 bool __vmx_guest_state_valid(struct kvm_vcpu *vcpu);
741 static inline bool vmx_guest_state_valid(struct kvm_vcpu *vcpu)
742 {
743 	return is_unrestricted_guest(vcpu) || __vmx_guest_state_valid(vcpu);
744 }
745 
746 void dump_vmcs(struct kvm_vcpu *vcpu);
747 
748 static inline int vmx_get_instr_info_reg2(u32 vmx_instr_info)
749 {
750 	return (vmx_instr_info >> 28) & 0xf;
751 }
752 
753 static inline bool vmx_can_use_ipiv(struct kvm_vcpu *vcpu)
754 {
755 	return  lapic_in_kernel(vcpu) && enable_ipiv;
756 }
757 
758 static inline void vmx_segment_cache_clear(struct vcpu_vmx *vmx)
759 {
760 	vmx->segment_cache.bitmask = 0;
761 }
762 
763 #endif /* __KVM_X86_VMX_H */
764