xref: /openbmc/linux/arch/x86/include/asm/kvm_host.h (revision 413d6ed3)
1 /* SPDX-License-Identifier: GPL-2.0-only */
2 /*
3  * Kernel-based Virtual Machine driver for Linux
4  *
5  * This header defines architecture specific interfaces, x86 version
6  */
7 
8 #ifndef _ASM_X86_KVM_HOST_H
9 #define _ASM_X86_KVM_HOST_H
10 
11 #include <linux/types.h>
12 #include <linux/mm.h>
13 #include <linux/mmu_notifier.h>
14 #include <linux/tracepoint.h>
15 #include <linux/cpumask.h>
16 #include <linux/irq_work.h>
17 #include <linux/irq.h>
18 
19 #include <linux/kvm.h>
20 #include <linux/kvm_para.h>
21 #include <linux/kvm_types.h>
22 #include <linux/perf_event.h>
23 #include <linux/pvclock_gtod.h>
24 #include <linux/clocksource.h>
25 #include <linux/irqbypass.h>
26 #include <linux/hyperv.h>
27 
28 #include <asm/apic.h>
29 #include <asm/pvclock-abi.h>
30 #include <asm/desc.h>
31 #include <asm/mtrr.h>
32 #include <asm/msr-index.h>
33 #include <asm/asm.h>
34 #include <asm/kvm_page_track.h>
35 #include <asm/kvm_vcpu_regs.h>
36 #include <asm/hyperv-tlfs.h>
37 
38 #define __KVM_HAVE_ARCH_VCPU_DEBUGFS
39 
40 #define KVM_MAX_VCPUS 288
41 #define KVM_SOFT_MAX_VCPUS 240
42 #define KVM_MAX_VCPU_ID 1023
43 /* memory slots that are not exposed to userspace */
44 #define KVM_PRIVATE_MEM_SLOTS 3
45 
46 #define KVM_HALT_POLL_NS_DEFAULT 200000
47 
48 #define KVM_IRQCHIP_NUM_PINS  KVM_IOAPIC_NUM_PINS
49 
50 #define KVM_DIRTY_LOG_MANUAL_CAPS   (KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE | \
51 					KVM_DIRTY_LOG_INITIALLY_SET)
52 
53 #define KVM_BUS_LOCK_DETECTION_VALID_MODE	(KVM_BUS_LOCK_DETECTION_OFF | \
54 						 KVM_BUS_LOCK_DETECTION_EXIT)
55 
56 /* x86-specific vcpu->requests bit members */
57 #define KVM_REQ_MIGRATE_TIMER		KVM_ARCH_REQ(0)
58 #define KVM_REQ_REPORT_TPR_ACCESS	KVM_ARCH_REQ(1)
59 #define KVM_REQ_TRIPLE_FAULT		KVM_ARCH_REQ(2)
60 #define KVM_REQ_MMU_SYNC		KVM_ARCH_REQ(3)
61 #define KVM_REQ_CLOCK_UPDATE		KVM_ARCH_REQ(4)
62 #define KVM_REQ_LOAD_MMU_PGD		KVM_ARCH_REQ(5)
63 #define KVM_REQ_EVENT			KVM_ARCH_REQ(6)
64 #define KVM_REQ_APF_HALT		KVM_ARCH_REQ(7)
65 #define KVM_REQ_STEAL_UPDATE		KVM_ARCH_REQ(8)
66 #define KVM_REQ_NMI			KVM_ARCH_REQ(9)
67 #define KVM_REQ_PMU			KVM_ARCH_REQ(10)
68 #define KVM_REQ_PMI			KVM_ARCH_REQ(11)
69 #define KVM_REQ_SMI			KVM_ARCH_REQ(12)
70 #define KVM_REQ_MASTERCLOCK_UPDATE	KVM_ARCH_REQ(13)
71 #define KVM_REQ_MCLOCK_INPROGRESS \
72 	KVM_ARCH_REQ_FLAGS(14, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
73 #define KVM_REQ_SCAN_IOAPIC \
74 	KVM_ARCH_REQ_FLAGS(15, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
75 #define KVM_REQ_GLOBAL_CLOCK_UPDATE	KVM_ARCH_REQ(16)
76 #define KVM_REQ_APIC_PAGE_RELOAD \
77 	KVM_ARCH_REQ_FLAGS(17, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
78 #define KVM_REQ_HV_CRASH		KVM_ARCH_REQ(18)
79 #define KVM_REQ_IOAPIC_EOI_EXIT		KVM_ARCH_REQ(19)
80 #define KVM_REQ_HV_RESET		KVM_ARCH_REQ(20)
81 #define KVM_REQ_HV_EXIT			KVM_ARCH_REQ(21)
82 #define KVM_REQ_HV_STIMER		KVM_ARCH_REQ(22)
83 #define KVM_REQ_LOAD_EOI_EXITMAP	KVM_ARCH_REQ(23)
84 #define KVM_REQ_GET_NESTED_STATE_PAGES	KVM_ARCH_REQ(24)
85 #define KVM_REQ_APICV_UPDATE \
86 	KVM_ARCH_REQ_FLAGS(25, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
87 #define KVM_REQ_TLB_FLUSH_CURRENT	KVM_ARCH_REQ(26)
88 #define KVM_REQ_HV_TLB_FLUSH \
89 	KVM_ARCH_REQ_FLAGS(27, KVM_REQUEST_NO_WAKEUP)
90 #define KVM_REQ_APF_READY		KVM_ARCH_REQ(28)
91 #define KVM_REQ_MSR_FILTER_CHANGED	KVM_ARCH_REQ(29)
92 #define KVM_REQ_UPDATE_CPU_DIRTY_LOGGING \
93 	KVM_ARCH_REQ_FLAGS(30, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
94 
95 #define CR0_RESERVED_BITS                                               \
96 	(~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \
97 			  | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \
98 			  | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG))
99 
100 #define CR4_RESERVED_BITS                                               \
101 	(~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\
102 			  | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE     \
103 			  | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR | X86_CR4_PCIDE \
104 			  | X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_FSGSBASE \
105 			  | X86_CR4_OSXMMEXCPT | X86_CR4_LA57 | X86_CR4_VMXE \
106 			  | X86_CR4_SMAP | X86_CR4_PKE | X86_CR4_UMIP))
107 
108 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR)
109 
110 
111 
112 #define INVALID_PAGE (~(hpa_t)0)
113 #define VALID_PAGE(x) ((x) != INVALID_PAGE)
114 
115 #define UNMAPPED_GVA (~(gpa_t)0)
116 #define INVALID_GPA (~(gpa_t)0)
117 
118 /* KVM Hugepage definitions for x86 */
119 #define KVM_MAX_HUGEPAGE_LEVEL	PG_LEVEL_1G
120 #define KVM_NR_PAGE_SIZES	(KVM_MAX_HUGEPAGE_LEVEL - PG_LEVEL_4K + 1)
121 #define KVM_HPAGE_GFN_SHIFT(x)	(((x) - 1) * 9)
122 #define KVM_HPAGE_SHIFT(x)	(PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
123 #define KVM_HPAGE_SIZE(x)	(1UL << KVM_HPAGE_SHIFT(x))
124 #define KVM_HPAGE_MASK(x)	(~(KVM_HPAGE_SIZE(x) - 1))
125 #define KVM_PAGES_PER_HPAGE(x)	(KVM_HPAGE_SIZE(x) / PAGE_SIZE)
126 
127 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
128 {
129 	/* KVM_HPAGE_GFN_SHIFT(PG_LEVEL_4K) must be 0. */
130 	return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
131 		(base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
132 }
133 
134 #define KVM_PERMILLE_MMU_PAGES 20
135 #define KVM_MIN_ALLOC_MMU_PAGES 64UL
136 #define KVM_MMU_HASH_SHIFT 12
137 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
138 #define KVM_MIN_FREE_MMU_PAGES 5
139 #define KVM_REFILL_PAGES 25
140 #define KVM_MAX_CPUID_ENTRIES 256
141 #define KVM_NR_FIXED_MTRR_REGION 88
142 #define KVM_NR_VAR_MTRR 8
143 
144 #define ASYNC_PF_PER_VCPU 64
145 
146 enum kvm_reg {
147 	VCPU_REGS_RAX = __VCPU_REGS_RAX,
148 	VCPU_REGS_RCX = __VCPU_REGS_RCX,
149 	VCPU_REGS_RDX = __VCPU_REGS_RDX,
150 	VCPU_REGS_RBX = __VCPU_REGS_RBX,
151 	VCPU_REGS_RSP = __VCPU_REGS_RSP,
152 	VCPU_REGS_RBP = __VCPU_REGS_RBP,
153 	VCPU_REGS_RSI = __VCPU_REGS_RSI,
154 	VCPU_REGS_RDI = __VCPU_REGS_RDI,
155 #ifdef CONFIG_X86_64
156 	VCPU_REGS_R8  = __VCPU_REGS_R8,
157 	VCPU_REGS_R9  = __VCPU_REGS_R9,
158 	VCPU_REGS_R10 = __VCPU_REGS_R10,
159 	VCPU_REGS_R11 = __VCPU_REGS_R11,
160 	VCPU_REGS_R12 = __VCPU_REGS_R12,
161 	VCPU_REGS_R13 = __VCPU_REGS_R13,
162 	VCPU_REGS_R14 = __VCPU_REGS_R14,
163 	VCPU_REGS_R15 = __VCPU_REGS_R15,
164 #endif
165 	VCPU_REGS_RIP,
166 	NR_VCPU_REGS,
167 
168 	VCPU_EXREG_PDPTR = NR_VCPU_REGS,
169 	VCPU_EXREG_CR0,
170 	VCPU_EXREG_CR3,
171 	VCPU_EXREG_CR4,
172 	VCPU_EXREG_RFLAGS,
173 	VCPU_EXREG_SEGMENTS,
174 	VCPU_EXREG_EXIT_INFO_1,
175 	VCPU_EXREG_EXIT_INFO_2,
176 };
177 
178 enum {
179 	VCPU_SREG_ES,
180 	VCPU_SREG_CS,
181 	VCPU_SREG_SS,
182 	VCPU_SREG_DS,
183 	VCPU_SREG_FS,
184 	VCPU_SREG_GS,
185 	VCPU_SREG_TR,
186 	VCPU_SREG_LDTR,
187 };
188 
189 enum exit_fastpath_completion {
190 	EXIT_FASTPATH_NONE,
191 	EXIT_FASTPATH_REENTER_GUEST,
192 	EXIT_FASTPATH_EXIT_HANDLED,
193 };
194 typedef enum exit_fastpath_completion fastpath_t;
195 
196 struct x86_emulate_ctxt;
197 struct x86_exception;
198 enum x86_intercept;
199 enum x86_intercept_stage;
200 
201 #define KVM_NR_DB_REGS	4
202 
203 #define DR6_BUS_LOCK   (1 << 11)
204 #define DR6_BD		(1 << 13)
205 #define DR6_BS		(1 << 14)
206 #define DR6_BT		(1 << 15)
207 #define DR6_RTM		(1 << 16)
208 /*
209  * DR6_ACTIVE_LOW combines fixed-1 and active-low bits.
210  * We can regard all the bits in DR6_FIXED_1 as active_low bits;
211  * they will never be 0 for now, but when they are defined
212  * in the future it will require no code change.
213  *
214  * DR6_ACTIVE_LOW is also used as the init/reset value for DR6.
215  */
216 #define DR6_ACTIVE_LOW	0xffff0ff0
217 #define DR6_VOLATILE	0x0001e80f
218 #define DR6_FIXED_1	(DR6_ACTIVE_LOW & ~DR6_VOLATILE)
219 
220 #define DR7_BP_EN_MASK	0x000000ff
221 #define DR7_GE		(1 << 9)
222 #define DR7_GD		(1 << 13)
223 #define DR7_FIXED_1	0x00000400
224 #define DR7_VOLATILE	0xffff2bff
225 
226 #define KVM_GUESTDBG_VALID_MASK \
227 	(KVM_GUESTDBG_ENABLE | \
228 	KVM_GUESTDBG_SINGLESTEP | \
229 	KVM_GUESTDBG_USE_HW_BP | \
230 	KVM_GUESTDBG_USE_SW_BP | \
231 	KVM_GUESTDBG_INJECT_BP | \
232 	KVM_GUESTDBG_INJECT_DB)
233 
234 
235 #define PFERR_PRESENT_BIT 0
236 #define PFERR_WRITE_BIT 1
237 #define PFERR_USER_BIT 2
238 #define PFERR_RSVD_BIT 3
239 #define PFERR_FETCH_BIT 4
240 #define PFERR_PK_BIT 5
241 #define PFERR_SGX_BIT 15
242 #define PFERR_GUEST_FINAL_BIT 32
243 #define PFERR_GUEST_PAGE_BIT 33
244 
245 #define PFERR_PRESENT_MASK (1U << PFERR_PRESENT_BIT)
246 #define PFERR_WRITE_MASK (1U << PFERR_WRITE_BIT)
247 #define PFERR_USER_MASK (1U << PFERR_USER_BIT)
248 #define PFERR_RSVD_MASK (1U << PFERR_RSVD_BIT)
249 #define PFERR_FETCH_MASK (1U << PFERR_FETCH_BIT)
250 #define PFERR_PK_MASK (1U << PFERR_PK_BIT)
251 #define PFERR_SGX_MASK (1U << PFERR_SGX_BIT)
252 #define PFERR_GUEST_FINAL_MASK (1ULL << PFERR_GUEST_FINAL_BIT)
253 #define PFERR_GUEST_PAGE_MASK (1ULL << PFERR_GUEST_PAGE_BIT)
254 
255 #define PFERR_NESTED_GUEST_PAGE (PFERR_GUEST_PAGE_MASK |	\
256 				 PFERR_WRITE_MASK |		\
257 				 PFERR_PRESENT_MASK)
258 
259 /* apic attention bits */
260 #define KVM_APIC_CHECK_VAPIC	0
261 /*
262  * The following bit is set with PV-EOI, unset on EOI.
263  * We detect PV-EOI changes by guest by comparing
264  * this bit with PV-EOI in guest memory.
265  * See the implementation in apic_update_pv_eoi.
266  */
267 #define KVM_APIC_PV_EOI_PENDING	1
268 
269 struct kvm_kernel_irq_routing_entry;
270 
271 /*
272  * the pages used as guest page table on soft mmu are tracked by
273  * kvm_memory_slot.arch.gfn_track which is 16 bits, so the role bits used
274  * by indirect shadow page can not be more than 15 bits.
275  *
276  * Currently, we used 14 bits that are @level, @gpte_is_8_bytes, @quadrant, @access,
277  * @nxe, @cr0_wp, @smep_andnot_wp and @smap_andnot_wp.
278  */
279 union kvm_mmu_page_role {
280 	u32 word;
281 	struct {
282 		unsigned level:4;
283 		unsigned gpte_is_8_bytes:1;
284 		unsigned quadrant:2;
285 		unsigned direct:1;
286 		unsigned access:3;
287 		unsigned invalid:1;
288 		unsigned nxe:1;
289 		unsigned cr0_wp:1;
290 		unsigned smep_andnot_wp:1;
291 		unsigned smap_andnot_wp:1;
292 		unsigned ad_disabled:1;
293 		unsigned guest_mode:1;
294 		unsigned :6;
295 
296 		/*
297 		 * This is left at the top of the word so that
298 		 * kvm_memslots_for_spte_role can extract it with a
299 		 * simple shift.  While there is room, give it a whole
300 		 * byte so it is also faster to load it from memory.
301 		 */
302 		unsigned smm:8;
303 	};
304 };
305 
306 union kvm_mmu_extended_role {
307 /*
308  * This structure complements kvm_mmu_page_role caching everything needed for
309  * MMU configuration. If nothing in both these structures changed, MMU
310  * re-configuration can be skipped. @valid bit is set on first usage so we don't
311  * treat all-zero structure as valid data.
312  */
313 	u32 word;
314 	struct {
315 		unsigned int valid:1;
316 		unsigned int execonly:1;
317 		unsigned int cr0_pg:1;
318 		unsigned int cr4_pae:1;
319 		unsigned int cr4_pse:1;
320 		unsigned int cr4_pke:1;
321 		unsigned int cr4_smap:1;
322 		unsigned int cr4_smep:1;
323 		unsigned int maxphyaddr:6;
324 	};
325 };
326 
327 union kvm_mmu_role {
328 	u64 as_u64;
329 	struct {
330 		union kvm_mmu_page_role base;
331 		union kvm_mmu_extended_role ext;
332 	};
333 };
334 
335 struct kvm_rmap_head {
336 	unsigned long val;
337 };
338 
339 struct kvm_pio_request {
340 	unsigned long linear_rip;
341 	unsigned long count;
342 	int in;
343 	int port;
344 	int size;
345 };
346 
347 #define PT64_ROOT_MAX_LEVEL 5
348 
349 struct rsvd_bits_validate {
350 	u64 rsvd_bits_mask[2][PT64_ROOT_MAX_LEVEL];
351 	u64 bad_mt_xwr;
352 };
353 
354 struct kvm_mmu_root_info {
355 	gpa_t pgd;
356 	hpa_t hpa;
357 };
358 
359 #define KVM_MMU_ROOT_INFO_INVALID \
360 	((struct kvm_mmu_root_info) { .pgd = INVALID_PAGE, .hpa = INVALID_PAGE })
361 
362 #define KVM_MMU_NUM_PREV_ROOTS 3
363 
364 #define KVM_HAVE_MMU_RWLOCK
365 
366 struct kvm_mmu_page;
367 
368 /*
369  * x86 supports 4 paging modes (5-level 64-bit, 4-level 64-bit, 3-level 32-bit,
370  * and 2-level 32-bit).  The kvm_mmu structure abstracts the details of the
371  * current mmu mode.
372  */
373 struct kvm_mmu {
374 	unsigned long (*get_guest_pgd)(struct kvm_vcpu *vcpu);
375 	u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
376 	int (*page_fault)(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u32 err,
377 			  bool prefault);
378 	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
379 				  struct x86_exception *fault);
380 	gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gpa_t gva_or_gpa,
381 			    u32 access, struct x86_exception *exception);
382 	gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
383 			       struct x86_exception *exception);
384 	int (*sync_page)(struct kvm_vcpu *vcpu,
385 			 struct kvm_mmu_page *sp);
386 	void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva, hpa_t root_hpa);
387 	hpa_t root_hpa;
388 	gpa_t root_pgd;
389 	union kvm_mmu_role mmu_role;
390 	u8 root_level;
391 	u8 shadow_root_level;
392 	u8 ept_ad;
393 	bool direct_map;
394 	struct kvm_mmu_root_info prev_roots[KVM_MMU_NUM_PREV_ROOTS];
395 
396 	/*
397 	 * Bitmap; bit set = permission fault
398 	 * Byte index: page fault error code [4:1]
399 	 * Bit index: pte permissions in ACC_* format
400 	 */
401 	u8 permissions[16];
402 
403 	/*
404 	* The pkru_mask indicates if protection key checks are needed.  It
405 	* consists of 16 domains indexed by page fault error code bits [4:1],
406 	* with PFEC.RSVD replaced by ACC_USER_MASK from the page tables.
407 	* Each domain has 2 bits which are ANDed with AD and WD from PKRU.
408 	*/
409 	u32 pkru_mask;
410 
411 	u64 *pae_root;
412 	u64 *pml4_root;
413 
414 	/*
415 	 * check zero bits on shadow page table entries, these
416 	 * bits include not only hardware reserved bits but also
417 	 * the bits spte never used.
418 	 */
419 	struct rsvd_bits_validate shadow_zero_check;
420 
421 	struct rsvd_bits_validate guest_rsvd_check;
422 
423 	/* Can have large pages at levels 2..last_nonleaf_level-1. */
424 	u8 last_nonleaf_level;
425 
426 	bool nx;
427 
428 	u64 pdptrs[4]; /* pae */
429 };
430 
431 struct kvm_tlb_range {
432 	u64 start_gfn;
433 	u64 pages;
434 };
435 
436 enum pmc_type {
437 	KVM_PMC_GP = 0,
438 	KVM_PMC_FIXED,
439 };
440 
441 struct kvm_pmc {
442 	enum pmc_type type;
443 	u8 idx;
444 	u64 counter;
445 	u64 eventsel;
446 	struct perf_event *perf_event;
447 	struct kvm_vcpu *vcpu;
448 	/*
449 	 * eventsel value for general purpose counters,
450 	 * ctrl value for fixed counters.
451 	 */
452 	u64 current_config;
453 };
454 
455 struct kvm_pmu {
456 	unsigned nr_arch_gp_counters;
457 	unsigned nr_arch_fixed_counters;
458 	unsigned available_event_types;
459 	u64 fixed_ctr_ctrl;
460 	u64 global_ctrl;
461 	u64 global_status;
462 	u64 global_ovf_ctrl;
463 	u64 counter_bitmask[2];
464 	u64 global_ctrl_mask;
465 	u64 global_ovf_ctrl_mask;
466 	u64 reserved_bits;
467 	u8 version;
468 	struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
469 	struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED];
470 	struct irq_work irq_work;
471 	DECLARE_BITMAP(reprogram_pmi, X86_PMC_IDX_MAX);
472 	DECLARE_BITMAP(all_valid_pmc_idx, X86_PMC_IDX_MAX);
473 	DECLARE_BITMAP(pmc_in_use, X86_PMC_IDX_MAX);
474 
475 	/*
476 	 * The gate to release perf_events not marked in
477 	 * pmc_in_use only once in a vcpu time slice.
478 	 */
479 	bool need_cleanup;
480 
481 	/*
482 	 * The total number of programmed perf_events and it helps to avoid
483 	 * redundant check before cleanup if guest don't use vPMU at all.
484 	 */
485 	u8 event_count;
486 };
487 
488 struct kvm_pmu_ops;
489 
490 enum {
491 	KVM_DEBUGREG_BP_ENABLED = 1,
492 	KVM_DEBUGREG_WONT_EXIT = 2,
493 	KVM_DEBUGREG_RELOAD = 4,
494 };
495 
496 struct kvm_mtrr_range {
497 	u64 base;
498 	u64 mask;
499 	struct list_head node;
500 };
501 
502 struct kvm_mtrr {
503 	struct kvm_mtrr_range var_ranges[KVM_NR_VAR_MTRR];
504 	mtrr_type fixed_ranges[KVM_NR_FIXED_MTRR_REGION];
505 	u64 deftype;
506 
507 	struct list_head head;
508 };
509 
510 /* Hyper-V SynIC timer */
511 struct kvm_vcpu_hv_stimer {
512 	struct hrtimer timer;
513 	int index;
514 	union hv_stimer_config config;
515 	u64 count;
516 	u64 exp_time;
517 	struct hv_message msg;
518 	bool msg_pending;
519 };
520 
521 /* Hyper-V synthetic interrupt controller (SynIC)*/
522 struct kvm_vcpu_hv_synic {
523 	u64 version;
524 	u64 control;
525 	u64 msg_page;
526 	u64 evt_page;
527 	atomic64_t sint[HV_SYNIC_SINT_COUNT];
528 	atomic_t sint_to_gsi[HV_SYNIC_SINT_COUNT];
529 	DECLARE_BITMAP(auto_eoi_bitmap, 256);
530 	DECLARE_BITMAP(vec_bitmap, 256);
531 	bool active;
532 	bool dont_zero_synic_pages;
533 };
534 
535 /* Hyper-V per vcpu emulation context */
536 struct kvm_vcpu_hv {
537 	struct kvm_vcpu *vcpu;
538 	u32 vp_index;
539 	u64 hv_vapic;
540 	s64 runtime_offset;
541 	struct kvm_vcpu_hv_synic synic;
542 	struct kvm_hyperv_exit exit;
543 	struct kvm_vcpu_hv_stimer stimer[HV_SYNIC_STIMER_COUNT];
544 	DECLARE_BITMAP(stimer_pending_bitmap, HV_SYNIC_STIMER_COUNT);
545 	cpumask_t tlb_flush;
546 };
547 
548 /* Xen HVM per vcpu emulation context */
549 struct kvm_vcpu_xen {
550 	u64 hypercall_rip;
551 	u32 current_runstate;
552 	bool vcpu_info_set;
553 	bool vcpu_time_info_set;
554 	bool runstate_set;
555 	struct gfn_to_hva_cache vcpu_info_cache;
556 	struct gfn_to_hva_cache vcpu_time_info_cache;
557 	struct gfn_to_hva_cache runstate_cache;
558 	u64 last_steal;
559 	u64 runstate_entry_time;
560 	u64 runstate_times[4];
561 };
562 
563 struct kvm_vcpu_arch {
564 	/*
565 	 * rip and regs accesses must go through
566 	 * kvm_{register,rip}_{read,write} functions.
567 	 */
568 	unsigned long regs[NR_VCPU_REGS];
569 	u32 regs_avail;
570 	u32 regs_dirty;
571 
572 	unsigned long cr0;
573 	unsigned long cr0_guest_owned_bits;
574 	unsigned long cr2;
575 	unsigned long cr3;
576 	unsigned long cr4;
577 	unsigned long cr4_guest_owned_bits;
578 	unsigned long cr4_guest_rsvd_bits;
579 	unsigned long cr8;
580 	u32 host_pkru;
581 	u32 pkru;
582 	u32 hflags;
583 	u64 efer;
584 	u64 apic_base;
585 	struct kvm_lapic *apic;    /* kernel irqchip context */
586 	bool apicv_active;
587 	bool load_eoi_exitmap_pending;
588 	DECLARE_BITMAP(ioapic_handled_vectors, 256);
589 	unsigned long apic_attention;
590 	int32_t apic_arb_prio;
591 	int mp_state;
592 	u64 ia32_misc_enable_msr;
593 	u64 smbase;
594 	u64 smi_count;
595 	bool tpr_access_reporting;
596 	bool xsaves_enabled;
597 	u64 ia32_xss;
598 	u64 microcode_version;
599 	u64 arch_capabilities;
600 	u64 perf_capabilities;
601 
602 	/*
603 	 * Paging state of the vcpu
604 	 *
605 	 * If the vcpu runs in guest mode with two level paging this still saves
606 	 * the paging mode of the l1 guest. This context is always used to
607 	 * handle faults.
608 	 */
609 	struct kvm_mmu *mmu;
610 
611 	/* Non-nested MMU for L1 */
612 	struct kvm_mmu root_mmu;
613 
614 	/* L1 MMU when running nested */
615 	struct kvm_mmu guest_mmu;
616 
617 	/*
618 	 * Paging state of an L2 guest (used for nested npt)
619 	 *
620 	 * This context will save all necessary information to walk page tables
621 	 * of an L2 guest. This context is only initialized for page table
622 	 * walking and not for faulting since we never handle l2 page faults on
623 	 * the host.
624 	 */
625 	struct kvm_mmu nested_mmu;
626 
627 	/*
628 	 * Pointer to the mmu context currently used for
629 	 * gva_to_gpa translations.
630 	 */
631 	struct kvm_mmu *walk_mmu;
632 
633 	struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
634 	struct kvm_mmu_memory_cache mmu_shadow_page_cache;
635 	struct kvm_mmu_memory_cache mmu_gfn_array_cache;
636 	struct kvm_mmu_memory_cache mmu_page_header_cache;
637 
638 	/*
639 	 * QEMU userspace and the guest each have their own FPU state.
640 	 * In vcpu_run, we switch between the user and guest FPU contexts.
641 	 * While running a VCPU, the VCPU thread will have the guest FPU
642 	 * context.
643 	 *
644 	 * Note that while the PKRU state lives inside the fpu registers,
645 	 * it is switched out separately at VMENTER and VMEXIT time. The
646 	 * "guest_fpu" state here contains the guest FPU context, with the
647 	 * host PRKU bits.
648 	 */
649 	struct fpu *user_fpu;
650 	struct fpu *guest_fpu;
651 
652 	u64 xcr0;
653 	u64 guest_supported_xcr0;
654 
655 	struct kvm_pio_request pio;
656 	void *pio_data;
657 	void *guest_ins_data;
658 
659 	u8 event_exit_inst_len;
660 
661 	struct kvm_queued_exception {
662 		bool pending;
663 		bool injected;
664 		bool has_error_code;
665 		u8 nr;
666 		u32 error_code;
667 		unsigned long payload;
668 		bool has_payload;
669 		u8 nested_apf;
670 	} exception;
671 
672 	struct kvm_queued_interrupt {
673 		bool injected;
674 		bool soft;
675 		u8 nr;
676 	} interrupt;
677 
678 	int halt_request; /* real mode on Intel only */
679 
680 	int cpuid_nent;
681 	struct kvm_cpuid_entry2 *cpuid_entries;
682 
683 	u64 reserved_gpa_bits;
684 	int maxphyaddr;
685 	int max_tdp_level;
686 
687 	/* emulate context */
688 
689 	struct x86_emulate_ctxt *emulate_ctxt;
690 	bool emulate_regs_need_sync_to_vcpu;
691 	bool emulate_regs_need_sync_from_vcpu;
692 	int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
693 
694 	gpa_t time;
695 	struct pvclock_vcpu_time_info hv_clock;
696 	unsigned int hw_tsc_khz;
697 	struct gfn_to_hva_cache pv_time;
698 	bool pv_time_enabled;
699 	/* set guest stopped flag in pvclock flags field */
700 	bool pvclock_set_guest_stopped_request;
701 
702 	struct {
703 		u8 preempted;
704 		u64 msr_val;
705 		u64 last_steal;
706 		struct gfn_to_pfn_cache cache;
707 	} st;
708 
709 	u64 l1_tsc_offset;
710 	u64 tsc_offset;
711 	u64 last_guest_tsc;
712 	u64 last_host_tsc;
713 	u64 tsc_offset_adjustment;
714 	u64 this_tsc_nsec;
715 	u64 this_tsc_write;
716 	u64 this_tsc_generation;
717 	bool tsc_catchup;
718 	bool tsc_always_catchup;
719 	s8 virtual_tsc_shift;
720 	u32 virtual_tsc_mult;
721 	u32 virtual_tsc_khz;
722 	s64 ia32_tsc_adjust_msr;
723 	u64 msr_ia32_power_ctl;
724 	u64 tsc_scaling_ratio;
725 
726 	atomic_t nmi_queued;  /* unprocessed asynchronous NMIs */
727 	unsigned nmi_pending; /* NMI queued after currently running handler */
728 	bool nmi_injected;    /* Trying to inject an NMI this entry */
729 	bool smi_pending;    /* SMI queued after currently running handler */
730 
731 	struct kvm_mtrr mtrr_state;
732 	u64 pat;
733 
734 	unsigned switch_db_regs;
735 	unsigned long db[KVM_NR_DB_REGS];
736 	unsigned long dr6;
737 	unsigned long dr7;
738 	unsigned long eff_db[KVM_NR_DB_REGS];
739 	unsigned long guest_debug_dr7;
740 	u64 msr_platform_info;
741 	u64 msr_misc_features_enables;
742 
743 	u64 mcg_cap;
744 	u64 mcg_status;
745 	u64 mcg_ctl;
746 	u64 mcg_ext_ctl;
747 	u64 *mce_banks;
748 
749 	/* Cache MMIO info */
750 	u64 mmio_gva;
751 	unsigned mmio_access;
752 	gfn_t mmio_gfn;
753 	u64 mmio_gen;
754 
755 	struct kvm_pmu pmu;
756 
757 	/* used for guest single stepping over the given code position */
758 	unsigned long singlestep_rip;
759 
760 	bool hyperv_enabled;
761 	struct kvm_vcpu_hv *hyperv;
762 	struct kvm_vcpu_xen xen;
763 
764 	cpumask_var_t wbinvd_dirty_mask;
765 
766 	unsigned long last_retry_eip;
767 	unsigned long last_retry_addr;
768 
769 	struct {
770 		bool halted;
771 		gfn_t gfns[ASYNC_PF_PER_VCPU];
772 		struct gfn_to_hva_cache data;
773 		u64 msr_en_val; /* MSR_KVM_ASYNC_PF_EN */
774 		u64 msr_int_val; /* MSR_KVM_ASYNC_PF_INT */
775 		u16 vec;
776 		u32 id;
777 		bool send_user_only;
778 		u32 host_apf_flags;
779 		unsigned long nested_apf_token;
780 		bool delivery_as_pf_vmexit;
781 		bool pageready_pending;
782 	} apf;
783 
784 	/* OSVW MSRs (AMD only) */
785 	struct {
786 		u64 length;
787 		u64 status;
788 	} osvw;
789 
790 	struct {
791 		u64 msr_val;
792 		struct gfn_to_hva_cache data;
793 	} pv_eoi;
794 
795 	u64 msr_kvm_poll_control;
796 
797 	/*
798 	 * Indicates the guest is trying to write a gfn that contains one or
799 	 * more of the PTEs used to translate the write itself, i.e. the access
800 	 * is changing its own translation in the guest page tables.  KVM exits
801 	 * to userspace if emulation of the faulting instruction fails and this
802 	 * flag is set, as KVM cannot make forward progress.
803 	 *
804 	 * If emulation fails for a write to guest page tables, KVM unprotects
805 	 * (zaps) the shadow page for the target gfn and resumes the guest to
806 	 * retry the non-emulatable instruction (on hardware).  Unprotecting the
807 	 * gfn doesn't allow forward progress for a self-changing access because
808 	 * doing so also zaps the translation for the gfn, i.e. retrying the
809 	 * instruction will hit a !PRESENT fault, which results in a new shadow
810 	 * page and sends KVM back to square one.
811 	 */
812 	bool write_fault_to_shadow_pgtable;
813 
814 	/* set at EPT violation at this point */
815 	unsigned long exit_qualification;
816 
817 	/* pv related host specific info */
818 	struct {
819 		bool pv_unhalted;
820 	} pv;
821 
822 	int pending_ioapic_eoi;
823 	int pending_external_vector;
824 
825 	/* be preempted when it's in kernel-mode(cpl=0) */
826 	bool preempted_in_kernel;
827 
828 	/* Flush the L1 Data cache for L1TF mitigation on VMENTER */
829 	bool l1tf_flush_l1d;
830 
831 	/* Host CPU on which VM-entry was most recently attempted */
832 	unsigned int last_vmentry_cpu;
833 
834 	/* AMD MSRC001_0015 Hardware Configuration */
835 	u64 msr_hwcr;
836 
837 	/* pv related cpuid info */
838 	struct {
839 		/*
840 		 * value of the eax register in the KVM_CPUID_FEATURES CPUID
841 		 * leaf.
842 		 */
843 		u32 features;
844 
845 		/*
846 		 * indicates whether pv emulation should be disabled if features
847 		 * are not present in the guest's cpuid
848 		 */
849 		bool enforce;
850 	} pv_cpuid;
851 
852 	/* Protected Guests */
853 	bool guest_state_protected;
854 };
855 
856 struct kvm_lpage_info {
857 	int disallow_lpage;
858 };
859 
860 struct kvm_arch_memory_slot {
861 	struct kvm_rmap_head *rmap[KVM_NR_PAGE_SIZES];
862 	struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
863 	unsigned short *gfn_track[KVM_PAGE_TRACK_MAX];
864 };
865 
866 /*
867  * We use as the mode the number of bits allocated in the LDR for the
868  * logical processor ID.  It happens that these are all powers of two.
869  * This makes it is very easy to detect cases where the APICs are
870  * configured for multiple modes; in that case, we cannot use the map and
871  * hence cannot use kvm_irq_delivery_to_apic_fast either.
872  */
873 #define KVM_APIC_MODE_XAPIC_CLUSTER          4
874 #define KVM_APIC_MODE_XAPIC_FLAT             8
875 #define KVM_APIC_MODE_X2APIC                16
876 
877 struct kvm_apic_map {
878 	struct rcu_head rcu;
879 	u8 mode;
880 	u32 max_apic_id;
881 	union {
882 		struct kvm_lapic *xapic_flat_map[8];
883 		struct kvm_lapic *xapic_cluster_map[16][4];
884 	};
885 	struct kvm_lapic *phys_map[];
886 };
887 
888 /* Hyper-V synthetic debugger (SynDbg)*/
889 struct kvm_hv_syndbg {
890 	struct {
891 		u64 control;
892 		u64 status;
893 		u64 send_page;
894 		u64 recv_page;
895 		u64 pending_page;
896 	} control;
897 	u64 options;
898 };
899 
900 /* Current state of Hyper-V TSC page clocksource */
901 enum hv_tsc_page_status {
902 	/* TSC page was not set up or disabled */
903 	HV_TSC_PAGE_UNSET = 0,
904 	/* TSC page MSR was written by the guest, update pending */
905 	HV_TSC_PAGE_GUEST_CHANGED,
906 	/* TSC page MSR was written by KVM userspace, update pending */
907 	HV_TSC_PAGE_HOST_CHANGED,
908 	/* TSC page was properly set up and is currently active  */
909 	HV_TSC_PAGE_SET,
910 	/* TSC page is currently being updated and therefore is inactive */
911 	HV_TSC_PAGE_UPDATING,
912 	/* TSC page was set up with an inaccessible GPA */
913 	HV_TSC_PAGE_BROKEN,
914 };
915 
916 /* Hyper-V emulation context */
917 struct kvm_hv {
918 	struct mutex hv_lock;
919 	u64 hv_guest_os_id;
920 	u64 hv_hypercall;
921 	u64 hv_tsc_page;
922 	enum hv_tsc_page_status hv_tsc_page_status;
923 
924 	/* Hyper-v based guest crash (NT kernel bugcheck) parameters */
925 	u64 hv_crash_param[HV_X64_MSR_CRASH_PARAMS];
926 	u64 hv_crash_ctl;
927 
928 	struct ms_hyperv_tsc_page tsc_ref;
929 
930 	struct idr conn_to_evt;
931 
932 	u64 hv_reenlightenment_control;
933 	u64 hv_tsc_emulation_control;
934 	u64 hv_tsc_emulation_status;
935 
936 	/* How many vCPUs have VP index != vCPU index */
937 	atomic_t num_mismatched_vp_indexes;
938 
939 	struct hv_partition_assist_pg *hv_pa_pg;
940 	struct kvm_hv_syndbg hv_syndbg;
941 };
942 
943 struct msr_bitmap_range {
944 	u32 flags;
945 	u32 nmsrs;
946 	u32 base;
947 	unsigned long *bitmap;
948 };
949 
950 /* Xen emulation context */
951 struct kvm_xen {
952 	bool long_mode;
953 	bool shinfo_set;
954 	u8 upcall_vector;
955 	struct gfn_to_hva_cache shinfo_cache;
956 };
957 
958 enum kvm_irqchip_mode {
959 	KVM_IRQCHIP_NONE,
960 	KVM_IRQCHIP_KERNEL,       /* created with KVM_CREATE_IRQCHIP */
961 	KVM_IRQCHIP_SPLIT,        /* created with KVM_CAP_SPLIT_IRQCHIP */
962 };
963 
964 struct kvm_x86_msr_filter {
965 	u8 count;
966 	bool default_allow:1;
967 	struct msr_bitmap_range ranges[16];
968 };
969 
970 #define APICV_INHIBIT_REASON_DISABLE    0
971 #define APICV_INHIBIT_REASON_HYPERV     1
972 #define APICV_INHIBIT_REASON_NESTED     2
973 #define APICV_INHIBIT_REASON_IRQWIN     3
974 #define APICV_INHIBIT_REASON_PIT_REINJ  4
975 #define APICV_INHIBIT_REASON_X2APIC	5
976 
977 struct kvm_arch {
978 	unsigned long n_used_mmu_pages;
979 	unsigned long n_requested_mmu_pages;
980 	unsigned long n_max_mmu_pages;
981 	unsigned int indirect_shadow_pages;
982 	u8 mmu_valid_gen;
983 	struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
984 	struct list_head active_mmu_pages;
985 	struct list_head zapped_obsolete_pages;
986 	struct list_head lpage_disallowed_mmu_pages;
987 	struct kvm_page_track_notifier_node mmu_sp_tracker;
988 	struct kvm_page_track_notifier_head track_notifier_head;
989 
990 	struct list_head assigned_dev_head;
991 	struct iommu_domain *iommu_domain;
992 	bool iommu_noncoherent;
993 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA
994 	atomic_t noncoherent_dma_count;
995 #define __KVM_HAVE_ARCH_ASSIGNED_DEVICE
996 	atomic_t assigned_device_count;
997 	struct kvm_pic *vpic;
998 	struct kvm_ioapic *vioapic;
999 	struct kvm_pit *vpit;
1000 	atomic_t vapics_in_nmi_mode;
1001 	struct mutex apic_map_lock;
1002 	struct kvm_apic_map __rcu *apic_map;
1003 	atomic_t apic_map_dirty;
1004 
1005 	bool apic_access_page_done;
1006 	unsigned long apicv_inhibit_reasons;
1007 
1008 	gpa_t wall_clock;
1009 
1010 	bool mwait_in_guest;
1011 	bool hlt_in_guest;
1012 	bool pause_in_guest;
1013 	bool cstate_in_guest;
1014 
1015 	unsigned long irq_sources_bitmap;
1016 	s64 kvmclock_offset;
1017 	raw_spinlock_t tsc_write_lock;
1018 	u64 last_tsc_nsec;
1019 	u64 last_tsc_write;
1020 	u32 last_tsc_khz;
1021 	u64 cur_tsc_nsec;
1022 	u64 cur_tsc_write;
1023 	u64 cur_tsc_offset;
1024 	u64 cur_tsc_generation;
1025 	int nr_vcpus_matched_tsc;
1026 
1027 	spinlock_t pvclock_gtod_sync_lock;
1028 	bool use_master_clock;
1029 	u64 master_kernel_ns;
1030 	u64 master_cycle_now;
1031 	struct delayed_work kvmclock_update_work;
1032 	struct delayed_work kvmclock_sync_work;
1033 
1034 	struct kvm_xen_hvm_config xen_hvm_config;
1035 
1036 	/* reads protected by irq_srcu, writes by irq_lock */
1037 	struct hlist_head mask_notifier_list;
1038 
1039 	struct kvm_hv hyperv;
1040 	struct kvm_xen xen;
1041 
1042 	#ifdef CONFIG_KVM_MMU_AUDIT
1043 	int audit_point;
1044 	#endif
1045 
1046 	bool backwards_tsc_observed;
1047 	bool boot_vcpu_runs_old_kvmclock;
1048 	u32 bsp_vcpu_id;
1049 
1050 	u64 disabled_quirks;
1051 	int cpu_dirty_logging_count;
1052 
1053 	enum kvm_irqchip_mode irqchip_mode;
1054 	u8 nr_reserved_ioapic_pins;
1055 
1056 	bool disabled_lapic_found;
1057 
1058 	bool x2apic_format;
1059 	bool x2apic_broadcast_quirk_disabled;
1060 
1061 	bool guest_can_read_msr_platform_info;
1062 	bool exception_payload_enabled;
1063 
1064 	bool bus_lock_detection_enabled;
1065 
1066 	/* Deflect RDMSR and WRMSR to user space when they trigger a #GP */
1067 	u32 user_space_msr_mask;
1068 	struct kvm_x86_msr_filter __rcu *msr_filter;
1069 
1070 	/* Guest can access the SGX PROVISIONKEY. */
1071 	bool sgx_provisioning_allowed;
1072 
1073 	struct kvm_pmu_event_filter __rcu *pmu_event_filter;
1074 	struct task_struct *nx_lpage_recovery_thread;
1075 
1076 #ifdef CONFIG_X86_64
1077 	/*
1078 	 * Whether the TDP MMU is enabled for this VM. This contains a
1079 	 * snapshot of the TDP MMU module parameter from when the VM was
1080 	 * created and remains unchanged for the life of the VM. If this is
1081 	 * true, TDP MMU handler functions will run for various MMU
1082 	 * operations.
1083 	 */
1084 	bool tdp_mmu_enabled;
1085 
1086 	/*
1087 	 * List of struct kvm_mmu_pages being used as roots.
1088 	 * All struct kvm_mmu_pages in the list should have
1089 	 * tdp_mmu_page set.
1090 	 *
1091 	 * For reads, this list is protected by:
1092 	 *	the MMU lock in read mode + RCU or
1093 	 *	the MMU lock in write mode
1094 	 *
1095 	 * For writes, this list is protected by:
1096 	 *	the MMU lock in read mode + the tdp_mmu_pages_lock or
1097 	 *	the MMU lock in write mode
1098 	 *
1099 	 * Roots will remain in the list until their tdp_mmu_root_count
1100 	 * drops to zero, at which point the thread that decremented the
1101 	 * count to zero should removed the root from the list and clean
1102 	 * it up, freeing the root after an RCU grace period.
1103 	 */
1104 	struct list_head tdp_mmu_roots;
1105 
1106 	/*
1107 	 * List of struct kvmp_mmu_pages not being used as roots.
1108 	 * All struct kvm_mmu_pages in the list should have
1109 	 * tdp_mmu_page set and a tdp_mmu_root_count of 0.
1110 	 */
1111 	struct list_head tdp_mmu_pages;
1112 
1113 	/*
1114 	 * Protects accesses to the following fields when the MMU lock
1115 	 * is held in read mode:
1116 	 *  - tdp_mmu_roots (above)
1117 	 *  - tdp_mmu_pages (above)
1118 	 *  - the link field of struct kvm_mmu_pages used by the TDP MMU
1119 	 *  - lpage_disallowed_mmu_pages
1120 	 *  - the lpage_disallowed_link field of struct kvm_mmu_pages used
1121 	 *    by the TDP MMU
1122 	 * It is acceptable, but not necessary, to acquire this lock when
1123 	 * the thread holds the MMU lock in write mode.
1124 	 */
1125 	spinlock_t tdp_mmu_pages_lock;
1126 #endif /* CONFIG_X86_64 */
1127 };
1128 
1129 struct kvm_vm_stat {
1130 	ulong mmu_shadow_zapped;
1131 	ulong mmu_pte_write;
1132 	ulong mmu_pde_zapped;
1133 	ulong mmu_flooded;
1134 	ulong mmu_recycled;
1135 	ulong mmu_cache_miss;
1136 	ulong mmu_unsync;
1137 	ulong remote_tlb_flush;
1138 	ulong lpages;
1139 	ulong nx_lpage_splits;
1140 	ulong max_mmu_page_hash_collisions;
1141 };
1142 
1143 struct kvm_vcpu_stat {
1144 	u64 pf_fixed;
1145 	u64 pf_guest;
1146 	u64 tlb_flush;
1147 	u64 invlpg;
1148 
1149 	u64 exits;
1150 	u64 io_exits;
1151 	u64 mmio_exits;
1152 	u64 signal_exits;
1153 	u64 irq_window_exits;
1154 	u64 nmi_window_exits;
1155 	u64 l1d_flush;
1156 	u64 halt_exits;
1157 	u64 halt_successful_poll;
1158 	u64 halt_attempted_poll;
1159 	u64 halt_poll_invalid;
1160 	u64 halt_wakeup;
1161 	u64 request_irq_exits;
1162 	u64 irq_exits;
1163 	u64 host_state_reload;
1164 	u64 fpu_reload;
1165 	u64 insn_emulation;
1166 	u64 insn_emulation_fail;
1167 	u64 hypercalls;
1168 	u64 irq_injections;
1169 	u64 nmi_injections;
1170 	u64 req_event;
1171 	u64 halt_poll_success_ns;
1172 	u64 halt_poll_fail_ns;
1173 	u64 nested_run;
1174 	u64 directed_yield_attempted;
1175 	u64 directed_yield_successful;
1176 };
1177 
1178 struct x86_instruction_info;
1179 
1180 struct msr_data {
1181 	bool host_initiated;
1182 	u32 index;
1183 	u64 data;
1184 };
1185 
1186 struct kvm_lapic_irq {
1187 	u32 vector;
1188 	u16 delivery_mode;
1189 	u16 dest_mode;
1190 	bool level;
1191 	u16 trig_mode;
1192 	u32 shorthand;
1193 	u32 dest_id;
1194 	bool msi_redir_hint;
1195 };
1196 
1197 static inline u16 kvm_lapic_irq_dest_mode(bool dest_mode_logical)
1198 {
1199 	return dest_mode_logical ? APIC_DEST_LOGICAL : APIC_DEST_PHYSICAL;
1200 }
1201 
1202 struct kvm_x86_ops {
1203 	int (*hardware_enable)(void);
1204 	void (*hardware_disable)(void);
1205 	void (*hardware_unsetup)(void);
1206 	bool (*cpu_has_accelerated_tpr)(void);
1207 	bool (*has_emulated_msr)(struct kvm *kvm, u32 index);
1208 	void (*vcpu_after_set_cpuid)(struct kvm_vcpu *vcpu);
1209 
1210 	unsigned int vm_size;
1211 	int (*vm_init)(struct kvm *kvm);
1212 	void (*vm_destroy)(struct kvm *kvm);
1213 
1214 	/* Create, but do not attach this VCPU */
1215 	int (*vcpu_create)(struct kvm_vcpu *vcpu);
1216 	void (*vcpu_free)(struct kvm_vcpu *vcpu);
1217 	void (*vcpu_reset)(struct kvm_vcpu *vcpu, bool init_event);
1218 
1219 	void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
1220 	void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
1221 	void (*vcpu_put)(struct kvm_vcpu *vcpu);
1222 
1223 	void (*update_exception_bitmap)(struct kvm_vcpu *vcpu);
1224 	int (*get_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1225 	int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1226 	u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
1227 	void (*get_segment)(struct kvm_vcpu *vcpu,
1228 			    struct kvm_segment *var, int seg);
1229 	int (*get_cpl)(struct kvm_vcpu *vcpu);
1230 	void (*set_segment)(struct kvm_vcpu *vcpu,
1231 			    struct kvm_segment *var, int seg);
1232 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
1233 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
1234 	bool (*is_valid_cr4)(struct kvm_vcpu *vcpu, unsigned long cr0);
1235 	void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
1236 	int (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
1237 	void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1238 	void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1239 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1240 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1241 	void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu);
1242 	void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
1243 	void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
1244 	unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
1245 	void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
1246 
1247 	void (*tlb_flush_all)(struct kvm_vcpu *vcpu);
1248 	void (*tlb_flush_current)(struct kvm_vcpu *vcpu);
1249 	int  (*tlb_remote_flush)(struct kvm *kvm);
1250 	int  (*tlb_remote_flush_with_range)(struct kvm *kvm,
1251 			struct kvm_tlb_range *range);
1252 
1253 	/*
1254 	 * Flush any TLB entries associated with the given GVA.
1255 	 * Does not need to flush GPA->HPA mappings.
1256 	 * Can potentially get non-canonical addresses through INVLPGs, which
1257 	 * the implementation may choose to ignore if appropriate.
1258 	 */
1259 	void (*tlb_flush_gva)(struct kvm_vcpu *vcpu, gva_t addr);
1260 
1261 	/*
1262 	 * Flush any TLB entries created by the guest.  Like tlb_flush_gva(),
1263 	 * does not need to flush GPA->HPA mappings.
1264 	 */
1265 	void (*tlb_flush_guest)(struct kvm_vcpu *vcpu);
1266 
1267 	enum exit_fastpath_completion (*run)(struct kvm_vcpu *vcpu);
1268 	int (*handle_exit)(struct kvm_vcpu *vcpu,
1269 		enum exit_fastpath_completion exit_fastpath);
1270 	int (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
1271 	void (*update_emulated_instruction)(struct kvm_vcpu *vcpu);
1272 	void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
1273 	u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu);
1274 	void (*patch_hypercall)(struct kvm_vcpu *vcpu,
1275 				unsigned char *hypercall_addr);
1276 	void (*set_irq)(struct kvm_vcpu *vcpu);
1277 	void (*set_nmi)(struct kvm_vcpu *vcpu);
1278 	void (*queue_exception)(struct kvm_vcpu *vcpu);
1279 	void (*cancel_injection)(struct kvm_vcpu *vcpu);
1280 	int (*interrupt_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1281 	int (*nmi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1282 	bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
1283 	void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
1284 	void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
1285 	void (*enable_irq_window)(struct kvm_vcpu *vcpu);
1286 	void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
1287 	bool (*check_apicv_inhibit_reasons)(ulong bit);
1288 	void (*pre_update_apicv_exec_ctrl)(struct kvm *kvm, bool activate);
1289 	void (*refresh_apicv_exec_ctrl)(struct kvm_vcpu *vcpu);
1290 	void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
1291 	void (*hwapic_isr_update)(struct kvm_vcpu *vcpu, int isr);
1292 	bool (*guest_apic_has_interrupt)(struct kvm_vcpu *vcpu);
1293 	void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
1294 	void (*set_virtual_apic_mode)(struct kvm_vcpu *vcpu);
1295 	void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu);
1296 	int (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
1297 	int (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
1298 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
1299 	int (*set_identity_map_addr)(struct kvm *kvm, u64 ident_addr);
1300 	u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
1301 
1302 	void (*load_mmu_pgd)(struct kvm_vcpu *vcpu, hpa_t root_hpa,
1303 			     int root_level);
1304 
1305 	bool (*has_wbinvd_exit)(void);
1306 
1307 	/* Returns actual tsc_offset set in active VMCS */
1308 	u64 (*write_l1_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
1309 
1310 	/*
1311 	 * Retrieve somewhat arbitrary exit information.  Intended to be used
1312 	 * only from within tracepoints to avoid VMREADs when tracing is off.
1313 	 */
1314 	void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2,
1315 			      u32 *exit_int_info, u32 *exit_int_info_err_code);
1316 
1317 	int (*check_intercept)(struct kvm_vcpu *vcpu,
1318 			       struct x86_instruction_info *info,
1319 			       enum x86_intercept_stage stage,
1320 			       struct x86_exception *exception);
1321 	void (*handle_exit_irqoff)(struct kvm_vcpu *vcpu);
1322 
1323 	void (*request_immediate_exit)(struct kvm_vcpu *vcpu);
1324 
1325 	void (*sched_in)(struct kvm_vcpu *kvm, int cpu);
1326 
1327 	/*
1328 	 * Size of the CPU's dirty log buffer, i.e. VMX's PML buffer.  A zero
1329 	 * value indicates CPU dirty logging is unsupported or disabled.
1330 	 */
1331 	int cpu_dirty_log_size;
1332 	void (*update_cpu_dirty_logging)(struct kvm_vcpu *vcpu);
1333 
1334 	/* pmu operations of sub-arch */
1335 	const struct kvm_pmu_ops *pmu_ops;
1336 	const struct kvm_x86_nested_ops *nested_ops;
1337 
1338 	/*
1339 	 * Architecture specific hooks for vCPU blocking due to
1340 	 * HLT instruction.
1341 	 * Returns for .pre_block():
1342 	 *    - 0 means continue to block the vCPU.
1343 	 *    - 1 means we cannot block the vCPU since some event
1344 	 *        happens during this period, such as, 'ON' bit in
1345 	 *        posted-interrupts descriptor is set.
1346 	 */
1347 	int (*pre_block)(struct kvm_vcpu *vcpu);
1348 	void (*post_block)(struct kvm_vcpu *vcpu);
1349 
1350 	void (*vcpu_blocking)(struct kvm_vcpu *vcpu);
1351 	void (*vcpu_unblocking)(struct kvm_vcpu *vcpu);
1352 
1353 	int (*update_pi_irte)(struct kvm *kvm, unsigned int host_irq,
1354 			      uint32_t guest_irq, bool set);
1355 	void (*apicv_post_state_restore)(struct kvm_vcpu *vcpu);
1356 	bool (*dy_apicv_has_pending_interrupt)(struct kvm_vcpu *vcpu);
1357 
1358 	int (*set_hv_timer)(struct kvm_vcpu *vcpu, u64 guest_deadline_tsc,
1359 			    bool *expired);
1360 	void (*cancel_hv_timer)(struct kvm_vcpu *vcpu);
1361 
1362 	void (*setup_mce)(struct kvm_vcpu *vcpu);
1363 
1364 	int (*smi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1365 	int (*pre_enter_smm)(struct kvm_vcpu *vcpu, char *smstate);
1366 	int (*pre_leave_smm)(struct kvm_vcpu *vcpu, const char *smstate);
1367 	void (*enable_smi_window)(struct kvm_vcpu *vcpu);
1368 
1369 	int (*mem_enc_op)(struct kvm *kvm, void __user *argp);
1370 	int (*mem_enc_reg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1371 	int (*mem_enc_unreg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1372 	int (*vm_copy_enc_context_from)(struct kvm *kvm, unsigned int source_fd);
1373 
1374 	int (*get_msr_feature)(struct kvm_msr_entry *entry);
1375 
1376 	bool (*can_emulate_instruction)(struct kvm_vcpu *vcpu, void *insn, int insn_len);
1377 
1378 	bool (*apic_init_signal_blocked)(struct kvm_vcpu *vcpu);
1379 	int (*enable_direct_tlbflush)(struct kvm_vcpu *vcpu);
1380 
1381 	void (*migrate_timers)(struct kvm_vcpu *vcpu);
1382 	void (*msr_filter_changed)(struct kvm_vcpu *vcpu);
1383 	int (*complete_emulated_msr)(struct kvm_vcpu *vcpu, int err);
1384 
1385 	void (*vcpu_deliver_sipi_vector)(struct kvm_vcpu *vcpu, u8 vector);
1386 };
1387 
1388 struct kvm_x86_nested_ops {
1389 	int (*check_events)(struct kvm_vcpu *vcpu);
1390 	bool (*hv_timer_pending)(struct kvm_vcpu *vcpu);
1391 	void (*triple_fault)(struct kvm_vcpu *vcpu);
1392 	int (*get_state)(struct kvm_vcpu *vcpu,
1393 			 struct kvm_nested_state __user *user_kvm_nested_state,
1394 			 unsigned user_data_size);
1395 	int (*set_state)(struct kvm_vcpu *vcpu,
1396 			 struct kvm_nested_state __user *user_kvm_nested_state,
1397 			 struct kvm_nested_state *kvm_state);
1398 	bool (*get_nested_state_pages)(struct kvm_vcpu *vcpu);
1399 	int (*write_log_dirty)(struct kvm_vcpu *vcpu, gpa_t l2_gpa);
1400 
1401 	int (*enable_evmcs)(struct kvm_vcpu *vcpu,
1402 			    uint16_t *vmcs_version);
1403 	uint16_t (*get_evmcs_version)(struct kvm_vcpu *vcpu);
1404 };
1405 
1406 struct kvm_x86_init_ops {
1407 	int (*cpu_has_kvm_support)(void);
1408 	int (*disabled_by_bios)(void);
1409 	int (*check_processor_compatibility)(void);
1410 	int (*hardware_setup)(void);
1411 
1412 	struct kvm_x86_ops *runtime_ops;
1413 };
1414 
1415 struct kvm_arch_async_pf {
1416 	u32 token;
1417 	gfn_t gfn;
1418 	unsigned long cr3;
1419 	bool direct_map;
1420 };
1421 
1422 extern u32 __read_mostly kvm_nr_uret_msrs;
1423 extern u64 __read_mostly host_efer;
1424 extern bool __read_mostly allow_smaller_maxphyaddr;
1425 extern struct kvm_x86_ops kvm_x86_ops;
1426 
1427 #define KVM_X86_OP(func) \
1428 	DECLARE_STATIC_CALL(kvm_x86_##func, *(((struct kvm_x86_ops *)0)->func));
1429 #define KVM_X86_OP_NULL KVM_X86_OP
1430 #include <asm/kvm-x86-ops.h>
1431 
1432 static inline void kvm_ops_static_call_update(void)
1433 {
1434 #define KVM_X86_OP(func) \
1435 	static_call_update(kvm_x86_##func, kvm_x86_ops.func);
1436 #define KVM_X86_OP_NULL KVM_X86_OP
1437 #include <asm/kvm-x86-ops.h>
1438 }
1439 
1440 #define __KVM_HAVE_ARCH_VM_ALLOC
1441 static inline struct kvm *kvm_arch_alloc_vm(void)
1442 {
1443 	return __vmalloc(kvm_x86_ops.vm_size, GFP_KERNEL_ACCOUNT | __GFP_ZERO);
1444 }
1445 void kvm_arch_free_vm(struct kvm *kvm);
1446 
1447 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
1448 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
1449 {
1450 	if (kvm_x86_ops.tlb_remote_flush &&
1451 	    !static_call(kvm_x86_tlb_remote_flush)(kvm))
1452 		return 0;
1453 	else
1454 		return -ENOTSUPP;
1455 }
1456 
1457 int kvm_mmu_module_init(void);
1458 void kvm_mmu_module_exit(void);
1459 
1460 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
1461 int kvm_mmu_create(struct kvm_vcpu *vcpu);
1462 void kvm_mmu_init_vm(struct kvm *kvm);
1463 void kvm_mmu_uninit_vm(struct kvm *kvm);
1464 
1465 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
1466 void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
1467 				      struct kvm_memory_slot *memslot,
1468 				      int start_level);
1469 void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
1470 				   const struct kvm_memory_slot *memslot);
1471 void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
1472 				   struct kvm_memory_slot *memslot);
1473 void kvm_mmu_zap_all(struct kvm *kvm);
1474 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, u64 gen);
1475 unsigned long kvm_mmu_calculate_default_mmu_pages(struct kvm *kvm);
1476 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned long kvm_nr_mmu_pages);
1477 
1478 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
1479 bool pdptrs_changed(struct kvm_vcpu *vcpu);
1480 
1481 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
1482 			  const void *val, int bytes);
1483 
1484 struct kvm_irq_mask_notifier {
1485 	void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
1486 	int irq;
1487 	struct hlist_node link;
1488 };
1489 
1490 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
1491 				    struct kvm_irq_mask_notifier *kimn);
1492 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
1493 				      struct kvm_irq_mask_notifier *kimn);
1494 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
1495 			     bool mask);
1496 
1497 extern bool tdp_enabled;
1498 
1499 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
1500 
1501 /* control of guest tsc rate supported? */
1502 extern bool kvm_has_tsc_control;
1503 /* maximum supported tsc_khz for guests */
1504 extern u32  kvm_max_guest_tsc_khz;
1505 /* number of bits of the fractional part of the TSC scaling ratio */
1506 extern u8   kvm_tsc_scaling_ratio_frac_bits;
1507 /* maximum allowed value of TSC scaling ratio */
1508 extern u64  kvm_max_tsc_scaling_ratio;
1509 /* 1ull << kvm_tsc_scaling_ratio_frac_bits */
1510 extern u64  kvm_default_tsc_scaling_ratio;
1511 /* bus lock detection supported? */
1512 extern bool kvm_has_bus_lock_exit;
1513 
1514 extern u64 kvm_mce_cap_supported;
1515 
1516 /*
1517  * EMULTYPE_NO_DECODE - Set when re-emulating an instruction (after completing
1518  *			userspace I/O) to indicate that the emulation context
1519  *			should be reused as is, i.e. skip initialization of
1520  *			emulation context, instruction fetch and decode.
1521  *
1522  * EMULTYPE_TRAP_UD - Set when emulating an intercepted #UD from hardware.
1523  *		      Indicates that only select instructions (tagged with
1524  *		      EmulateOnUD) should be emulated (to minimize the emulator
1525  *		      attack surface).  See also EMULTYPE_TRAP_UD_FORCED.
1526  *
1527  * EMULTYPE_SKIP - Set when emulating solely to skip an instruction, i.e. to
1528  *		   decode the instruction length.  For use *only* by
1529  *		   kvm_x86_ops.skip_emulated_instruction() implementations.
1530  *
1531  * EMULTYPE_ALLOW_RETRY_PF - Set when the emulator should resume the guest to
1532  *			     retry native execution under certain conditions,
1533  *			     Can only be set in conjunction with EMULTYPE_PF.
1534  *
1535  * EMULTYPE_TRAP_UD_FORCED - Set when emulating an intercepted #UD that was
1536  *			     triggered by KVM's magic "force emulation" prefix,
1537  *			     which is opt in via module param (off by default).
1538  *			     Bypasses EmulateOnUD restriction despite emulating
1539  *			     due to an intercepted #UD (see EMULTYPE_TRAP_UD).
1540  *			     Used to test the full emulator from userspace.
1541  *
1542  * EMULTYPE_VMWARE_GP - Set when emulating an intercepted #GP for VMware
1543  *			backdoor emulation, which is opt in via module param.
1544  *			VMware backdoor emulation handles select instructions
1545  *			and reinjects the #GP for all other cases.
1546  *
1547  * EMULTYPE_PF - Set when emulating MMIO by way of an intercepted #PF, in which
1548  *		 case the CR2/GPA value pass on the stack is valid.
1549  */
1550 #define EMULTYPE_NO_DECODE	    (1 << 0)
1551 #define EMULTYPE_TRAP_UD	    (1 << 1)
1552 #define EMULTYPE_SKIP		    (1 << 2)
1553 #define EMULTYPE_ALLOW_RETRY_PF	    (1 << 3)
1554 #define EMULTYPE_TRAP_UD_FORCED	    (1 << 4)
1555 #define EMULTYPE_VMWARE_GP	    (1 << 5)
1556 #define EMULTYPE_PF		    (1 << 6)
1557 
1558 int kvm_emulate_instruction(struct kvm_vcpu *vcpu, int emulation_type);
1559 int kvm_emulate_instruction_from_buffer(struct kvm_vcpu *vcpu,
1560 					void *insn, int insn_len);
1561 
1562 void kvm_enable_efer_bits(u64);
1563 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
1564 int __kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data, bool host_initiated);
1565 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data);
1566 int kvm_set_msr(struct kvm_vcpu *vcpu, u32 index, u64 data);
1567 int kvm_emulate_rdmsr(struct kvm_vcpu *vcpu);
1568 int kvm_emulate_wrmsr(struct kvm_vcpu *vcpu);
1569 int kvm_emulate_as_nop(struct kvm_vcpu *vcpu);
1570 int kvm_emulate_invd(struct kvm_vcpu *vcpu);
1571 int kvm_emulate_mwait(struct kvm_vcpu *vcpu);
1572 int kvm_handle_invalid_op(struct kvm_vcpu *vcpu);
1573 int kvm_emulate_monitor(struct kvm_vcpu *vcpu);
1574 
1575 int kvm_fast_pio(struct kvm_vcpu *vcpu, int size, unsigned short port, int in);
1576 int kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
1577 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
1578 int kvm_vcpu_halt(struct kvm_vcpu *vcpu);
1579 int kvm_emulate_ap_reset_hold(struct kvm_vcpu *vcpu);
1580 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
1581 
1582 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
1583 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
1584 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector);
1585 
1586 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
1587 		    int reason, bool has_error_code, u32 error_code);
1588 
1589 void kvm_free_guest_fpu(struct kvm_vcpu *vcpu);
1590 
1591 void kvm_post_set_cr0(struct kvm_vcpu *vcpu, unsigned long old_cr0, unsigned long cr0);
1592 void kvm_post_set_cr4(struct kvm_vcpu *vcpu, unsigned long old_cr4, unsigned long cr4);
1593 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
1594 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
1595 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
1596 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
1597 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
1598 void kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
1599 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
1600 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
1601 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
1602 int kvm_emulate_xsetbv(struct kvm_vcpu *vcpu);
1603 
1604 int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1605 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1606 
1607 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
1608 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
1609 int kvm_emulate_rdpmc(struct kvm_vcpu *vcpu);
1610 
1611 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1612 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1613 void kvm_queue_exception_p(struct kvm_vcpu *vcpu, unsigned nr, unsigned long payload);
1614 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1615 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1616 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
1617 bool kvm_inject_emulated_page_fault(struct kvm_vcpu *vcpu,
1618 				    struct x86_exception *fault);
1619 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1620 			    gfn_t gfn, void *data, int offset, int len,
1621 			    u32 access);
1622 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
1623 bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr);
1624 
1625 static inline int __kvm_irq_line_state(unsigned long *irq_state,
1626 				       int irq_source_id, int level)
1627 {
1628 	/* Logical OR for level trig interrupt */
1629 	if (level)
1630 		__set_bit(irq_source_id, irq_state);
1631 	else
1632 		__clear_bit(irq_source_id, irq_state);
1633 
1634 	return !!(*irq_state);
1635 }
1636 
1637 #define KVM_MMU_ROOT_CURRENT		BIT(0)
1638 #define KVM_MMU_ROOT_PREVIOUS(i)	BIT(1+i)
1639 #define KVM_MMU_ROOTS_ALL		(~0UL)
1640 
1641 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
1642 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
1643 
1644 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
1645 
1646 void kvm_update_dr7(struct kvm_vcpu *vcpu);
1647 
1648 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
1649 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
1650 void kvm_mmu_free_roots(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1651 			ulong roots_to_free);
1652 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1653 			   struct x86_exception *exception);
1654 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
1655 			      struct x86_exception *exception);
1656 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
1657 			       struct x86_exception *exception);
1658 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
1659 			       struct x86_exception *exception);
1660 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
1661 				struct x86_exception *exception);
1662 
1663 bool kvm_apicv_activated(struct kvm *kvm);
1664 void kvm_apicv_init(struct kvm *kvm, bool enable);
1665 void kvm_vcpu_update_apicv(struct kvm_vcpu *vcpu);
1666 void kvm_request_apicv_update(struct kvm *kvm, bool activate,
1667 			      unsigned long bit);
1668 
1669 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
1670 
1671 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u64 error_code,
1672 		       void *insn, int insn_len);
1673 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
1674 void kvm_mmu_invalidate_gva(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1675 			    gva_t gva, hpa_t root_hpa);
1676 void kvm_mmu_invpcid_gva(struct kvm_vcpu *vcpu, gva_t gva, unsigned long pcid);
1677 void kvm_mmu_new_pgd(struct kvm_vcpu *vcpu, gpa_t new_pgd, bool skip_tlb_flush,
1678 		     bool skip_mmu_sync);
1679 
1680 void kvm_configure_mmu(bool enable_tdp, int tdp_max_root_level,
1681 		       int tdp_huge_page_level);
1682 
1683 static inline u16 kvm_read_ldt(void)
1684 {
1685 	u16 ldt;
1686 	asm("sldt %0" : "=g"(ldt));
1687 	return ldt;
1688 }
1689 
1690 static inline void kvm_load_ldt(u16 sel)
1691 {
1692 	asm("lldt %0" : : "rm"(sel));
1693 }
1694 
1695 #ifdef CONFIG_X86_64
1696 static inline unsigned long read_msr(unsigned long msr)
1697 {
1698 	u64 value;
1699 
1700 	rdmsrl(msr, value);
1701 	return value;
1702 }
1703 #endif
1704 
1705 static inline u32 get_rdx_init_val(void)
1706 {
1707 	return 0x600; /* P6 family */
1708 }
1709 
1710 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
1711 {
1712 	kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
1713 }
1714 
1715 #define TSS_IOPB_BASE_OFFSET 0x66
1716 #define TSS_BASE_SIZE 0x68
1717 #define TSS_IOPB_SIZE (65536 / 8)
1718 #define TSS_REDIRECTION_SIZE (256 / 8)
1719 #define RMODE_TSS_SIZE							\
1720 	(TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
1721 
1722 enum {
1723 	TASK_SWITCH_CALL = 0,
1724 	TASK_SWITCH_IRET = 1,
1725 	TASK_SWITCH_JMP = 2,
1726 	TASK_SWITCH_GATE = 3,
1727 };
1728 
1729 #define HF_GIF_MASK		(1 << 0)
1730 #define HF_NMI_MASK		(1 << 3)
1731 #define HF_IRET_MASK		(1 << 4)
1732 #define HF_GUEST_MASK		(1 << 5) /* VCPU is in guest-mode */
1733 #define HF_SMM_MASK		(1 << 6)
1734 #define HF_SMM_INSIDE_NMI_MASK	(1 << 7)
1735 
1736 #define __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
1737 #define KVM_ADDRESS_SPACE_NUM 2
1738 
1739 #define kvm_arch_vcpu_memslots_id(vcpu) ((vcpu)->arch.hflags & HF_SMM_MASK ? 1 : 0)
1740 #define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, (role).smm)
1741 
1742 asmlinkage void kvm_spurious_fault(void);
1743 
1744 /*
1745  * Hardware virtualization extension instructions may fault if a
1746  * reboot turns off virtualization while processes are running.
1747  * Usually after catching the fault we just panic; during reboot
1748  * instead the instruction is ignored.
1749  */
1750 #define __kvm_handle_fault_on_reboot(insn)				\
1751 	"666: \n\t"							\
1752 	insn "\n\t"							\
1753 	"jmp	668f \n\t"						\
1754 	"667: \n\t"							\
1755 	"1: \n\t"							\
1756 	".pushsection .discard.instr_begin \n\t"			\
1757 	".long 1b - . \n\t"						\
1758 	".popsection \n\t"						\
1759 	"call	kvm_spurious_fault \n\t"				\
1760 	"1: \n\t"							\
1761 	".pushsection .discard.instr_end \n\t"				\
1762 	".long 1b - . \n\t"						\
1763 	".popsection \n\t"						\
1764 	"668: \n\t"							\
1765 	_ASM_EXTABLE(666b, 667b)
1766 
1767 #define KVM_ARCH_WANT_MMU_NOTIFIER
1768 
1769 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1770 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1771 int kvm_cpu_has_extint(struct kvm_vcpu *v);
1772 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1773 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1774 void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event);
1775 void kvm_vcpu_reload_apic_access_page(struct kvm_vcpu *vcpu);
1776 
1777 int kvm_pv_send_ipi(struct kvm *kvm, unsigned long ipi_bitmap_low,
1778 		    unsigned long ipi_bitmap_high, u32 min,
1779 		    unsigned long icr, int op_64_bit);
1780 
1781 int kvm_add_user_return_msr(u32 msr);
1782 int kvm_find_user_return_msr(u32 msr);
1783 int kvm_set_user_return_msr(unsigned index, u64 val, u64 mask);
1784 
1785 static inline bool kvm_is_supported_user_return_msr(u32 msr)
1786 {
1787 	return kvm_find_user_return_msr(msr) >= 0;
1788 }
1789 
1790 u64 kvm_scale_tsc(struct kvm_vcpu *vcpu, u64 tsc);
1791 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc);
1792 
1793 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu);
1794 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1795 
1796 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
1797 void kvm_make_scan_ioapic_request(struct kvm *kvm);
1798 void kvm_make_scan_ioapic_request_mask(struct kvm *kvm,
1799 				       unsigned long *vcpu_bitmap);
1800 
1801 bool kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1802 				     struct kvm_async_pf *work);
1803 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1804 				 struct kvm_async_pf *work);
1805 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1806 			       struct kvm_async_pf *work);
1807 void kvm_arch_async_page_present_queued(struct kvm_vcpu *vcpu);
1808 bool kvm_arch_can_dequeue_async_page_present(struct kvm_vcpu *vcpu);
1809 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1810 
1811 int kvm_skip_emulated_instruction(struct kvm_vcpu *vcpu);
1812 int kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1813 void __kvm_request_immediate_exit(struct kvm_vcpu *vcpu);
1814 
1815 int kvm_is_in_guest(void);
1816 
1817 void __user *__x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa,
1818 				     u32 size);
1819 bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu);
1820 bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu);
1821 
1822 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
1823 			     struct kvm_vcpu **dest_vcpu);
1824 
1825 void kvm_set_msi_irq(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e,
1826 		     struct kvm_lapic_irq *irq);
1827 
1828 static inline bool kvm_irq_is_postable(struct kvm_lapic_irq *irq)
1829 {
1830 	/* We can only post Fixed and LowPrio IRQs */
1831 	return (irq->delivery_mode == APIC_DM_FIXED ||
1832 		irq->delivery_mode == APIC_DM_LOWEST);
1833 }
1834 
1835 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu)
1836 {
1837 	static_call_cond(kvm_x86_vcpu_blocking)(vcpu);
1838 }
1839 
1840 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu)
1841 {
1842 	static_call_cond(kvm_x86_vcpu_unblocking)(vcpu);
1843 }
1844 
1845 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
1846 
1847 static inline int kvm_cpu_get_apicid(int mps_cpu)
1848 {
1849 #ifdef CONFIG_X86_LOCAL_APIC
1850 	return default_cpu_present_to_apicid(mps_cpu);
1851 #else
1852 	WARN_ON_ONCE(1);
1853 	return BAD_APICID;
1854 #endif
1855 }
1856 
1857 #define put_smstate(type, buf, offset, val)                      \
1858 	*(type *)((buf) + (offset) - 0x7e00) = val
1859 
1860 #define GET_SMSTATE(type, buf, offset)		\
1861 	(*(type *)((buf) + (offset) - 0x7e00))
1862 
1863 int kvm_cpu_dirty_log_size(void);
1864 
1865 #endif /* _ASM_X86_KVM_HOST_H */
1866