xref: /openbmc/linux/arch/x86/kvm/vmx/vmx.h (revision 2f4e3926)
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 
19 #define MSR_TYPE_R	1
20 #define MSR_TYPE_W	2
21 #define MSR_TYPE_RW	3
22 
23 #define X2APIC_MSR(r) (APIC_BASE_MSR + ((r) >> 4))
24 
25 #ifdef CONFIG_X86_64
26 #define MAX_NR_USER_RETURN_MSRS	7
27 #else
28 #define MAX_NR_USER_RETURN_MSRS	4
29 #endif
30 
31 #define MAX_NR_LOADSTORE_MSRS	8
32 
33 struct vmx_msrs {
34 	unsigned int		nr;
35 	struct vmx_msr_entry	val[MAX_NR_LOADSTORE_MSRS];
36 };
37 
38 struct vmx_uret_msr {
39 	bool load_into_hardware;
40 	u64 data;
41 	u64 mask;
42 };
43 
44 enum segment_cache_field {
45 	SEG_FIELD_SEL = 0,
46 	SEG_FIELD_BASE = 1,
47 	SEG_FIELD_LIMIT = 2,
48 	SEG_FIELD_AR = 3,
49 
50 	SEG_FIELD_NR = 4
51 };
52 
53 #define RTIT_ADDR_RANGE		4
54 
55 struct pt_ctx {
56 	u64 ctl;
57 	u64 status;
58 	u64 output_base;
59 	u64 output_mask;
60 	u64 cr3_match;
61 	u64 addr_a[RTIT_ADDR_RANGE];
62 	u64 addr_b[RTIT_ADDR_RANGE];
63 };
64 
65 struct pt_desc {
66 	u64 ctl_bitmask;
67 	u32 num_address_ranges;
68 	u32 caps[PT_CPUID_REGS_NUM * PT_CPUID_LEAVES];
69 	struct pt_ctx host;
70 	struct pt_ctx guest;
71 };
72 
73 union vmx_exit_reason {
74 	struct {
75 		u32	basic			: 16;
76 		u32	reserved16		: 1;
77 		u32	reserved17		: 1;
78 		u32	reserved18		: 1;
79 		u32	reserved19		: 1;
80 		u32	reserved20		: 1;
81 		u32	reserved21		: 1;
82 		u32	reserved22		: 1;
83 		u32	reserved23		: 1;
84 		u32	reserved24		: 1;
85 		u32	reserved25		: 1;
86 		u32	bus_lock_detected	: 1;
87 		u32	enclave_mode		: 1;
88 		u32	smi_pending_mtf		: 1;
89 		u32	smi_from_vmx_root	: 1;
90 		u32	reserved30		: 1;
91 		u32	failed_vmentry		: 1;
92 	};
93 	u32 full;
94 };
95 
96 struct lbr_desc {
97 	/* Basic info about guest LBR records. */
98 	struct x86_pmu_lbr records;
99 
100 	/*
101 	 * Emulate LBR feature via passthrough LBR registers when the
102 	 * per-vcpu guest LBR event is scheduled on the current pcpu.
103 	 *
104 	 * The records may be inaccurate if the host reclaims the LBR.
105 	 */
106 	struct perf_event *event;
107 
108 	/* True if LBRs are marked as not intercepted in the MSR bitmap */
109 	bool msr_passthrough;
110 };
111 
112 /*
113  * The nested_vmx structure is part of vcpu_vmx, and holds information we need
114  * for correct emulation of VMX (i.e., nested VMX) on this vcpu.
115  */
116 struct nested_vmx {
117 	/* Has the level1 guest done vmxon? */
118 	bool vmxon;
119 	gpa_t vmxon_ptr;
120 	bool pml_full;
121 
122 	/* The guest-physical address of the current VMCS L1 keeps for L2 */
123 	gpa_t current_vmptr;
124 	/*
125 	 * Cache of the guest's VMCS, existing outside of guest memory.
126 	 * Loaded from guest memory during VMPTRLD. Flushed to guest
127 	 * memory during VMCLEAR and VMPTRLD.
128 	 */
129 	struct vmcs12 *cached_vmcs12;
130 	/*
131 	 * Cache of the guest's shadow VMCS, existing outside of guest
132 	 * memory. Loaded from guest memory during VM entry. Flushed
133 	 * to guest memory during VM exit.
134 	 */
135 	struct vmcs12 *cached_shadow_vmcs12;
136 
137 	/*
138 	 * GPA to HVA cache for accessing vmcs12->vmcs_link_pointer
139 	 */
140 	struct gfn_to_hva_cache shadow_vmcs12_cache;
141 
142 	/*
143 	 * GPA to HVA cache for VMCS12
144 	 */
145 	struct gfn_to_hva_cache vmcs12_cache;
146 
147 	/*
148 	 * Indicates if the shadow vmcs or enlightened vmcs must be updated
149 	 * with the data held by struct vmcs12.
150 	 */
151 	bool need_vmcs12_to_shadow_sync;
152 	bool dirty_vmcs12;
153 
154 	/*
155 	 * Indicates whether MSR bitmap for L2 needs to be rebuilt due to
156 	 * changes in MSR bitmap for L1 or switching to a different L2. Note,
157 	 * this flag can only be used reliably in conjunction with a paravirt L1
158 	 * which informs L0 whether any changes to MSR bitmap for L2 were done
159 	 * on its side.
160 	 */
161 	bool force_msr_bitmap_recalc;
162 
163 	/*
164 	 * Indicates lazily loaded guest state has not yet been decached from
165 	 * vmcs02.
166 	 */
167 	bool need_sync_vmcs02_to_vmcs12_rare;
168 
169 	/*
170 	 * vmcs02 has been initialized, i.e. state that is constant for
171 	 * vmcs02 has been written to the backing VMCS.  Initialization
172 	 * is delayed until L1 actually attempts to run a nested VM.
173 	 */
174 	bool vmcs02_initialized;
175 
176 	bool change_vmcs01_virtual_apic_mode;
177 	bool reload_vmcs01_apic_access_page;
178 	bool update_vmcs01_cpu_dirty_logging;
179 	bool update_vmcs01_apicv_status;
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 kvm_host_map msr_bitmap_map;
205 
206 	struct pi_desc *pi_desc;
207 	bool pi_pending;
208 	u16 posted_intr_nv;
209 
210 	struct hrtimer preemption_timer;
211 	u64 preemption_timer_deadline;
212 	bool has_preemption_timer_deadline;
213 	bool preemption_timer_expired;
214 
215 	/*
216 	 * Used to snapshot MSRs that are conditionally loaded on VM-Enter in
217 	 * order to propagate the guest's pre-VM-Enter value into vmcs02.  For
218 	 * emulation of VMLAUNCH/VMRESUME, the snapshot will be of L1's value.
219 	 * For KVM_SET_NESTED_STATE, the snapshot is of L2's value, _if_
220 	 * userspace restores MSRs before nested state.  If userspace restores
221 	 * MSRs after nested state, the snapshot holds garbage, but KVM can't
222 	 * detect that, and the garbage value in vmcs02 will be overwritten by
223 	 * MSR restoration in any case.
224 	 */
225 	u64 pre_vmenter_debugctl;
226 	u64 pre_vmenter_bndcfgs;
227 
228 	/* to migrate it to L1 if L2 writes to L1's CR8 directly */
229 	int l1_tpr_threshold;
230 
231 	u16 vpid02;
232 	u16 last_vpid;
233 
234 	struct nested_vmx_msrs msrs;
235 
236 	/* SMM related state */
237 	struct {
238 		/* in VMX operation on SMM entry? */
239 		bool vmxon;
240 		/* in guest mode on SMM entry? */
241 		bool guest_mode;
242 	} smm;
243 
244 	gpa_t hv_evmcs_vmptr;
245 	struct kvm_host_map hv_evmcs_map;
246 	struct hv_enlightened_vmcs *hv_evmcs;
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 	bool req_immediate_exit;
334 
335 	/* Support for PML */
336 #define PML_ENTITY_NUM		512
337 	struct page *pml_pg;
338 
339 	/* apic deadline value in host tsc */
340 	u64 hv_deadline_tsc;
341 
342 	unsigned long host_debugctlmsr;
343 
344 	/*
345 	 * Only bits masked by msr_ia32_feature_control_valid_bits can be set in
346 	 * msr_ia32_feature_control. FEAT_CTL_LOCKED is always included
347 	 * in msr_ia32_feature_control_valid_bits.
348 	 */
349 	u64 msr_ia32_feature_control;
350 	u64 msr_ia32_feature_control_valid_bits;
351 	/* SGX Launch Control public key hash */
352 	u64 msr_ia32_sgxlepubkeyhash[4];
353 	u64 msr_ia32_mcu_opt_ctrl;
354 	bool disable_fb_clear;
355 
356 	struct pt_desc pt_desc;
357 	struct lbr_desc lbr_desc;
358 
359 	/* Save desired MSR intercept (read: pass-through) state */
360 #define MAX_POSSIBLE_PASSTHROUGH_MSRS	16
361 	struct {
362 		DECLARE_BITMAP(read, MAX_POSSIBLE_PASSTHROUGH_MSRS);
363 		DECLARE_BITMAP(write, MAX_POSSIBLE_PASSTHROUGH_MSRS);
364 	} shadow_msr_intercept;
365 };
366 
367 struct kvm_vmx {
368 	struct kvm kvm;
369 
370 	unsigned int tss_addr;
371 	bool ept_identity_pagetable_done;
372 	gpa_t ept_identity_map_addr;
373 	/* Posted Interrupt Descriptor (PID) table for IPI virtualization */
374 	u64 *pid_table;
375 };
376 
377 void vmx_vcpu_load_vmcs(struct kvm_vcpu *vcpu, int cpu,
378 			struct loaded_vmcs *buddy);
379 int allocate_vpid(void);
380 void free_vpid(int vpid);
381 void vmx_set_constant_host_state(struct vcpu_vmx *vmx);
382 void vmx_prepare_switch_to_guest(struct kvm_vcpu *vcpu);
383 void vmx_set_host_fs_gs(struct vmcs_host_state *host, u16 fs_sel, u16 gs_sel,
384 			unsigned long fs_base, unsigned long gs_base);
385 int vmx_get_cpl(struct kvm_vcpu *vcpu);
386 bool vmx_emulation_required(struct kvm_vcpu *vcpu);
387 unsigned long vmx_get_rflags(struct kvm_vcpu *vcpu);
388 void vmx_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
389 u32 vmx_get_interrupt_shadow(struct kvm_vcpu *vcpu);
390 void vmx_set_interrupt_shadow(struct kvm_vcpu *vcpu, int mask);
391 int vmx_set_efer(struct kvm_vcpu *vcpu, u64 efer);
392 void vmx_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
393 void vmx_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
394 void set_cr4_guest_host_mask(struct vcpu_vmx *vmx);
395 void ept_save_pdptrs(struct kvm_vcpu *vcpu);
396 void vmx_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
397 void __vmx_set_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
398 u64 construct_eptp(struct kvm_vcpu *vcpu, hpa_t root_hpa, int root_level);
399 
400 bool vmx_guest_inject_ac(struct kvm_vcpu *vcpu);
401 void vmx_update_exception_bitmap(struct kvm_vcpu *vcpu);
402 bool vmx_nmi_blocked(struct kvm_vcpu *vcpu);
403 bool __vmx_interrupt_blocked(struct kvm_vcpu *vcpu);
404 bool vmx_interrupt_blocked(struct kvm_vcpu *vcpu);
405 bool vmx_get_nmi_mask(struct kvm_vcpu *vcpu);
406 void vmx_set_nmi_mask(struct kvm_vcpu *vcpu, bool masked);
407 void vmx_set_virtual_apic_mode(struct kvm_vcpu *vcpu);
408 struct vmx_uret_msr *vmx_find_uret_msr(struct vcpu_vmx *vmx, u32 msr);
409 void pt_update_intercept_for_msr(struct kvm_vcpu *vcpu);
410 void vmx_update_host_rsp(struct vcpu_vmx *vmx, unsigned long host_rsp);
411 void vmx_spec_ctrl_restore_host(struct vcpu_vmx *vmx, unsigned int flags);
412 unsigned int __vmx_vcpu_run_flags(struct vcpu_vmx *vmx);
413 bool __vmx_vcpu_run(struct vcpu_vmx *vmx, unsigned long *regs,
414 		    unsigned int flags);
415 int vmx_find_loadstore_msr_slot(struct vmx_msrs *m, u32 msr);
416 void vmx_ept_load_pdptrs(struct kvm_vcpu *vcpu);
417 
418 void vmx_disable_intercept_for_msr(struct kvm_vcpu *vcpu, u32 msr, int type);
419 void vmx_enable_intercept_for_msr(struct kvm_vcpu *vcpu, u32 msr, int type);
420 
421 u64 vmx_get_l2_tsc_offset(struct kvm_vcpu *vcpu);
422 u64 vmx_get_l2_tsc_multiplier(struct kvm_vcpu *vcpu);
423 
424 static inline void vmx_set_intercept_for_msr(struct kvm_vcpu *vcpu, u32 msr,
425 					     int type, bool value)
426 {
427 	if (value)
428 		vmx_enable_intercept_for_msr(vcpu, msr, type);
429 	else
430 		vmx_disable_intercept_for_msr(vcpu, msr, type);
431 }
432 
433 void vmx_update_cpu_dirty_logging(struct kvm_vcpu *vcpu);
434 
435 /*
436  * Note, early Intel manuals have the write-low and read-high bitmap offsets
437  * the wrong way round.  The bitmaps control MSRs 0x00000000-0x00001fff and
438  * 0xc0000000-0xc0001fff.  The former (low) uses bytes 0-0x3ff for reads and
439  * 0x800-0xbff for writes.  The latter (high) uses 0x400-0x7ff for reads and
440  * 0xc00-0xfff for writes.  MSRs not covered by either of the ranges always
441  * VM-Exit.
442  */
443 #define __BUILD_VMX_MSR_BITMAP_HELPER(rtype, action, bitop, access, base)      \
444 static inline rtype vmx_##action##_msr_bitmap_##access(unsigned long *bitmap,  \
445 						       u32 msr)		       \
446 {									       \
447 	int f = sizeof(unsigned long);					       \
448 									       \
449 	if (msr <= 0x1fff)						       \
450 		return bitop##_bit(msr, bitmap + base / f);		       \
451 	else if ((msr >= 0xc0000000) && (msr <= 0xc0001fff))		       \
452 		return bitop##_bit(msr & 0x1fff, bitmap + (base + 0x400) / f); \
453 	return (rtype)true;						       \
454 }
455 #define BUILD_VMX_MSR_BITMAP_HELPERS(ret_type, action, bitop)		       \
456 	__BUILD_VMX_MSR_BITMAP_HELPER(ret_type, action, bitop, read,  0x0)     \
457 	__BUILD_VMX_MSR_BITMAP_HELPER(ret_type, action, bitop, write, 0x800)
458 
459 BUILD_VMX_MSR_BITMAP_HELPERS(bool, test, test)
460 BUILD_VMX_MSR_BITMAP_HELPERS(void, clear, __clear)
461 BUILD_VMX_MSR_BITMAP_HELPERS(void, set, __set)
462 
463 static inline u8 vmx_get_rvi(void)
464 {
465 	return vmcs_read16(GUEST_INTR_STATUS) & 0xff;
466 }
467 
468 #define __KVM_REQUIRED_VMX_VM_ENTRY_CONTROLS				\
469 	(VM_ENTRY_LOAD_DEBUG_CONTROLS)
470 #ifdef CONFIG_X86_64
471 	#define KVM_REQUIRED_VMX_VM_ENTRY_CONTROLS			\
472 		(__KVM_REQUIRED_VMX_VM_ENTRY_CONTROLS |			\
473 		 VM_ENTRY_IA32E_MODE)
474 #else
475 	#define KVM_REQUIRED_VMX_VM_ENTRY_CONTROLS			\
476 		__KVM_REQUIRED_VMX_VM_ENTRY_CONTROLS
477 #endif
478 #define KVM_OPTIONAL_VMX_VM_ENTRY_CONTROLS				\
479 	(VM_ENTRY_LOAD_IA32_PERF_GLOBAL_CTRL |				\
480 	 VM_ENTRY_LOAD_IA32_PAT |					\
481 	 VM_ENTRY_LOAD_IA32_EFER |					\
482 	 VM_ENTRY_LOAD_BNDCFGS |					\
483 	 VM_ENTRY_PT_CONCEAL_PIP |					\
484 	 VM_ENTRY_LOAD_IA32_RTIT_CTL)
485 
486 #define __KVM_REQUIRED_VMX_VM_EXIT_CONTROLS				\
487 	(VM_EXIT_SAVE_DEBUG_CONTROLS |					\
488 	 VM_EXIT_ACK_INTR_ON_EXIT)
489 #ifdef CONFIG_X86_64
490 	#define KVM_REQUIRED_VMX_VM_EXIT_CONTROLS			\
491 		(__KVM_REQUIRED_VMX_VM_EXIT_CONTROLS |			\
492 		 VM_EXIT_HOST_ADDR_SPACE_SIZE)
493 #else
494 	#define KVM_REQUIRED_VMX_VM_EXIT_CONTROLS			\
495 		__KVM_REQUIRED_VMX_VM_EXIT_CONTROLS
496 #endif
497 #define KVM_OPTIONAL_VMX_VM_EXIT_CONTROLS				\
498 	      (VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL |			\
499 	       VM_EXIT_SAVE_IA32_PAT |					\
500 	       VM_EXIT_LOAD_IA32_PAT |					\
501 	       VM_EXIT_SAVE_IA32_EFER |					\
502 	       VM_EXIT_SAVE_VMX_PREEMPTION_TIMER |			\
503 	       VM_EXIT_LOAD_IA32_EFER |					\
504 	       VM_EXIT_CLEAR_BNDCFGS |					\
505 	       VM_EXIT_PT_CONCEAL_PIP |					\
506 	       VM_EXIT_CLEAR_IA32_RTIT_CTL)
507 
508 #define KVM_REQUIRED_VMX_PIN_BASED_VM_EXEC_CONTROL			\
509 	(PIN_BASED_EXT_INTR_MASK |					\
510 	 PIN_BASED_NMI_EXITING)
511 #define KVM_OPTIONAL_VMX_PIN_BASED_VM_EXEC_CONTROL			\
512 	(PIN_BASED_VIRTUAL_NMIS |					\
513 	 PIN_BASED_POSTED_INTR |					\
514 	 PIN_BASED_VMX_PREEMPTION_TIMER)
515 
516 #define __KVM_REQUIRED_VMX_CPU_BASED_VM_EXEC_CONTROL			\
517 	(CPU_BASED_HLT_EXITING |					\
518 	 CPU_BASED_CR3_LOAD_EXITING |					\
519 	 CPU_BASED_CR3_STORE_EXITING |					\
520 	 CPU_BASED_UNCOND_IO_EXITING |					\
521 	 CPU_BASED_MOV_DR_EXITING |					\
522 	 CPU_BASED_USE_TSC_OFFSETTING |					\
523 	 CPU_BASED_MWAIT_EXITING |					\
524 	 CPU_BASED_MONITOR_EXITING |					\
525 	 CPU_BASED_INVLPG_EXITING |					\
526 	 CPU_BASED_RDPMC_EXITING |					\
527 	 CPU_BASED_INTR_WINDOW_EXITING)
528 
529 #ifdef CONFIG_X86_64
530 	#define KVM_REQUIRED_VMX_CPU_BASED_VM_EXEC_CONTROL		\
531 		(__KVM_REQUIRED_VMX_CPU_BASED_VM_EXEC_CONTROL |		\
532 		 CPU_BASED_CR8_LOAD_EXITING |				\
533 		 CPU_BASED_CR8_STORE_EXITING)
534 #else
535 	#define KVM_REQUIRED_VMX_CPU_BASED_VM_EXEC_CONTROL		\
536 		__KVM_REQUIRED_VMX_CPU_BASED_VM_EXEC_CONTROL
537 #endif
538 
539 #define KVM_OPTIONAL_VMX_CPU_BASED_VM_EXEC_CONTROL			\
540 	(CPU_BASED_RDTSC_EXITING |					\
541 	 CPU_BASED_TPR_SHADOW |						\
542 	 CPU_BASED_USE_IO_BITMAPS |					\
543 	 CPU_BASED_MONITOR_TRAP_FLAG |					\
544 	 CPU_BASED_USE_MSR_BITMAPS |					\
545 	 CPU_BASED_NMI_WINDOW_EXITING |					\
546 	 CPU_BASED_PAUSE_EXITING |					\
547 	 CPU_BASED_ACTIVATE_SECONDARY_CONTROLS |			\
548 	 CPU_BASED_ACTIVATE_TERTIARY_CONTROLS)
549 
550 #define KVM_REQUIRED_VMX_SECONDARY_VM_EXEC_CONTROL 0
551 #define KVM_OPTIONAL_VMX_SECONDARY_VM_EXEC_CONTROL			\
552 	(SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES |			\
553 	 SECONDARY_EXEC_VIRTUALIZE_X2APIC_MODE |			\
554 	 SECONDARY_EXEC_WBINVD_EXITING |				\
555 	 SECONDARY_EXEC_ENABLE_VPID |					\
556 	 SECONDARY_EXEC_ENABLE_EPT |					\
557 	 SECONDARY_EXEC_UNRESTRICTED_GUEST |				\
558 	 SECONDARY_EXEC_PAUSE_LOOP_EXITING |				\
559 	 SECONDARY_EXEC_DESC |						\
560 	 SECONDARY_EXEC_ENABLE_RDTSCP |					\
561 	 SECONDARY_EXEC_ENABLE_INVPCID |				\
562 	 SECONDARY_EXEC_APIC_REGISTER_VIRT |				\
563 	 SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY |				\
564 	 SECONDARY_EXEC_SHADOW_VMCS |					\
565 	 SECONDARY_EXEC_ENABLE_XSAVES |					\
566 	 SECONDARY_EXEC_RDSEED_EXITING |				\
567 	 SECONDARY_EXEC_RDRAND_EXITING |				\
568 	 SECONDARY_EXEC_ENABLE_PML |					\
569 	 SECONDARY_EXEC_TSC_SCALING |					\
570 	 SECONDARY_EXEC_ENABLE_USR_WAIT_PAUSE |				\
571 	 SECONDARY_EXEC_PT_USE_GPA |					\
572 	 SECONDARY_EXEC_PT_CONCEAL_VMX |				\
573 	 SECONDARY_EXEC_ENABLE_VMFUNC |					\
574 	 SECONDARY_EXEC_BUS_LOCK_DETECTION |				\
575 	 SECONDARY_EXEC_NOTIFY_VM_EXITING |				\
576 	 SECONDARY_EXEC_ENCLS_EXITING)
577 
578 #define KVM_REQUIRED_VMX_TERTIARY_VM_EXEC_CONTROL 0
579 #define KVM_OPTIONAL_VMX_TERTIARY_VM_EXEC_CONTROL			\
580 	(TERTIARY_EXEC_IPI_VIRT)
581 
582 #define BUILD_CONTROLS_SHADOW(lname, uname, bits)						\
583 static inline void lname##_controls_set(struct vcpu_vmx *vmx, u##bits val)			\
584 {												\
585 	if (vmx->loaded_vmcs->controls_shadow.lname != val) {					\
586 		vmcs_write##bits(uname, val);							\
587 		vmx->loaded_vmcs->controls_shadow.lname = val;					\
588 	}											\
589 }												\
590 static inline u##bits __##lname##_controls_get(struct loaded_vmcs *vmcs)			\
591 {												\
592 	return vmcs->controls_shadow.lname;							\
593 }												\
594 static inline u##bits lname##_controls_get(struct vcpu_vmx *vmx)				\
595 {												\
596 	return __##lname##_controls_get(vmx->loaded_vmcs);					\
597 }												\
598 static __always_inline void lname##_controls_setbit(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_clearbit(struct vcpu_vmx *vmx, u##bits val)	\
604 {												\
605 	BUILD_BUG_ON(!(val & (KVM_REQUIRED_VMX_##uname | KVM_OPTIONAL_VMX_##uname)));		\
606 	lname##_controls_set(vmx, lname##_controls_get(vmx) & ~val);				\
607 }
608 BUILD_CONTROLS_SHADOW(vm_entry, VM_ENTRY_CONTROLS, 32)
609 BUILD_CONTROLS_SHADOW(vm_exit, VM_EXIT_CONTROLS, 32)
610 BUILD_CONTROLS_SHADOW(pin, PIN_BASED_VM_EXEC_CONTROL, 32)
611 BUILD_CONTROLS_SHADOW(exec, CPU_BASED_VM_EXEC_CONTROL, 32)
612 BUILD_CONTROLS_SHADOW(secondary_exec, SECONDARY_VM_EXEC_CONTROL, 32)
613 BUILD_CONTROLS_SHADOW(tertiary_exec, TERTIARY_VM_EXEC_CONTROL, 64)
614 
615 /*
616  * VMX_REGS_LAZY_LOAD_SET - The set of registers that will be updated in the
617  * cache on demand.  Other registers not listed here are synced to
618  * the cache immediately after VM-Exit.
619  */
620 #define VMX_REGS_LAZY_LOAD_SET	((1 << VCPU_REGS_RIP) |         \
621 				(1 << VCPU_REGS_RSP) |          \
622 				(1 << VCPU_EXREG_RFLAGS) |      \
623 				(1 << VCPU_EXREG_PDPTR) |       \
624 				(1 << VCPU_EXREG_SEGMENTS) |    \
625 				(1 << VCPU_EXREG_CR0) |         \
626 				(1 << VCPU_EXREG_CR3) |         \
627 				(1 << VCPU_EXREG_CR4) |         \
628 				(1 << VCPU_EXREG_EXIT_INFO_1) | \
629 				(1 << VCPU_EXREG_EXIT_INFO_2))
630 
631 static inline unsigned long vmx_l1_guest_owned_cr0_bits(void)
632 {
633 	unsigned long bits = KVM_POSSIBLE_CR0_GUEST_BITS;
634 
635 	/*
636 	 * CR0.WP needs to be intercepted when KVM is shadowing legacy paging
637 	 * in order to construct shadow PTEs with the correct protections.
638 	 * Note!  CR0.WP technically can be passed through to the guest if
639 	 * paging is disabled, but checking CR0.PG would generate a cyclical
640 	 * dependency of sorts due to forcing the caller to ensure CR0 holds
641 	 * the correct value prior to determining which CR0 bits can be owned
642 	 * by L1.  Keep it simple and limit the optimization to EPT.
643 	 */
644 	if (!enable_ept)
645 		bits &= ~X86_CR0_WP;
646 	return bits;
647 }
648 
649 static __always_inline struct kvm_vmx *to_kvm_vmx(struct kvm *kvm)
650 {
651 	return container_of(kvm, struct kvm_vmx, kvm);
652 }
653 
654 static __always_inline struct vcpu_vmx *to_vmx(struct kvm_vcpu *vcpu)
655 {
656 	return container_of(vcpu, struct vcpu_vmx, vcpu);
657 }
658 
659 static inline struct lbr_desc *vcpu_to_lbr_desc(struct kvm_vcpu *vcpu)
660 {
661 	return &to_vmx(vcpu)->lbr_desc;
662 }
663 
664 static inline struct x86_pmu_lbr *vcpu_to_lbr_records(struct kvm_vcpu *vcpu)
665 {
666 	return &vcpu_to_lbr_desc(vcpu)->records;
667 }
668 
669 static inline bool intel_pmu_lbr_is_enabled(struct kvm_vcpu *vcpu)
670 {
671 	return !!vcpu_to_lbr_records(vcpu)->nr;
672 }
673 
674 void intel_pmu_cross_mapped_check(struct kvm_pmu *pmu);
675 int intel_pmu_create_guest_lbr_event(struct kvm_vcpu *vcpu);
676 void vmx_passthrough_lbr_msrs(struct kvm_vcpu *vcpu);
677 
678 static __always_inline unsigned long vmx_get_exit_qual(struct kvm_vcpu *vcpu)
679 {
680 	struct vcpu_vmx *vmx = to_vmx(vcpu);
681 
682 	if (!kvm_register_test_and_mark_available(vcpu, VCPU_EXREG_EXIT_INFO_1))
683 		vmx->exit_qualification = vmcs_readl(EXIT_QUALIFICATION);
684 
685 	return vmx->exit_qualification;
686 }
687 
688 static __always_inline u32 vmx_get_intr_info(struct kvm_vcpu *vcpu)
689 {
690 	struct vcpu_vmx *vmx = to_vmx(vcpu);
691 
692 	if (!kvm_register_test_and_mark_available(vcpu, VCPU_EXREG_EXIT_INFO_2))
693 		vmx->exit_intr_info = vmcs_read32(VM_EXIT_INTR_INFO);
694 
695 	return vmx->exit_intr_info;
696 }
697 
698 struct vmcs *alloc_vmcs_cpu(bool shadow, int cpu, gfp_t flags);
699 void free_vmcs(struct vmcs *vmcs);
700 int alloc_loaded_vmcs(struct loaded_vmcs *loaded_vmcs);
701 void free_loaded_vmcs(struct loaded_vmcs *loaded_vmcs);
702 void loaded_vmcs_clear(struct loaded_vmcs *loaded_vmcs);
703 
704 static inline struct vmcs *alloc_vmcs(bool shadow)
705 {
706 	return alloc_vmcs_cpu(shadow, raw_smp_processor_id(),
707 			      GFP_KERNEL_ACCOUNT);
708 }
709 
710 static inline bool vmx_has_waitpkg(struct vcpu_vmx *vmx)
711 {
712 	return secondary_exec_controls_get(vmx) &
713 		SECONDARY_EXEC_ENABLE_USR_WAIT_PAUSE;
714 }
715 
716 static inline bool vmx_need_pf_intercept(struct kvm_vcpu *vcpu)
717 {
718 	if (!enable_ept)
719 		return true;
720 
721 	return allow_smaller_maxphyaddr && cpuid_maxphyaddr(vcpu) < boot_cpu_data.x86_phys_bits;
722 }
723 
724 static inline bool is_unrestricted_guest(struct kvm_vcpu *vcpu)
725 {
726 	return enable_unrestricted_guest && (!is_guest_mode(vcpu) ||
727 	    (secondary_exec_controls_get(to_vmx(vcpu)) &
728 	    SECONDARY_EXEC_UNRESTRICTED_GUEST));
729 }
730 
731 bool __vmx_guest_state_valid(struct kvm_vcpu *vcpu);
732 static inline bool vmx_guest_state_valid(struct kvm_vcpu *vcpu)
733 {
734 	return is_unrestricted_guest(vcpu) || __vmx_guest_state_valid(vcpu);
735 }
736 
737 void dump_vmcs(struct kvm_vcpu *vcpu);
738 
739 static inline int vmx_get_instr_info_reg2(u32 vmx_instr_info)
740 {
741 	return (vmx_instr_info >> 28) & 0xf;
742 }
743 
744 static inline bool vmx_can_use_ipiv(struct kvm_vcpu *vcpu)
745 {
746 	return  lapic_in_kernel(vcpu) && enable_ipiv;
747 }
748 
749 static inline bool guest_cpuid_has_evmcs(struct kvm_vcpu *vcpu)
750 {
751 	/*
752 	 * eVMCS is exposed to the guest if Hyper-V is enabled in CPUID and
753 	 * eVMCS has been explicitly enabled by userspace.
754 	 */
755 	return vcpu->arch.hyperv_enabled &&
756 	       to_vmx(vcpu)->nested.enlightened_vmcs_enabled;
757 }
758 
759 #endif /* __KVM_X86_VMX_H */
760