xref: /openbmc/linux/arch/x86/include/asm/kvm_host.h (revision 06701297)
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 #define KVM_USER_MEM_SLOTS 509
44 /* memory slots that are not exposed to userspace */
45 #define KVM_PRIVATE_MEM_SLOTS 3
46 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
47 
48 #define KVM_HALT_POLL_NS_DEFAULT 200000
49 
50 #define KVM_IRQCHIP_NUM_PINS  KVM_IOAPIC_NUM_PINS
51 
52 #define KVM_DIRTY_LOG_MANUAL_CAPS   (KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE | \
53 					KVM_DIRTY_LOG_INITIALLY_SET)
54 
55 /* x86-specific vcpu->requests bit members */
56 #define KVM_REQ_MIGRATE_TIMER		KVM_ARCH_REQ(0)
57 #define KVM_REQ_REPORT_TPR_ACCESS	KVM_ARCH_REQ(1)
58 #define KVM_REQ_TRIPLE_FAULT		KVM_ARCH_REQ(2)
59 #define KVM_REQ_MMU_SYNC		KVM_ARCH_REQ(3)
60 #define KVM_REQ_CLOCK_UPDATE		KVM_ARCH_REQ(4)
61 #define KVM_REQ_LOAD_MMU_PGD		KVM_ARCH_REQ(5)
62 #define KVM_REQ_EVENT			KVM_ARCH_REQ(6)
63 #define KVM_REQ_APF_HALT		KVM_ARCH_REQ(7)
64 #define KVM_REQ_STEAL_UPDATE		KVM_ARCH_REQ(8)
65 #define KVM_REQ_NMI			KVM_ARCH_REQ(9)
66 #define KVM_REQ_PMU			KVM_ARCH_REQ(10)
67 #define KVM_REQ_PMI			KVM_ARCH_REQ(11)
68 #define KVM_REQ_SMI			KVM_ARCH_REQ(12)
69 #define KVM_REQ_MASTERCLOCK_UPDATE	KVM_ARCH_REQ(13)
70 #define KVM_REQ_MCLOCK_INPROGRESS \
71 	KVM_ARCH_REQ_FLAGS(14, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
72 #define KVM_REQ_SCAN_IOAPIC \
73 	KVM_ARCH_REQ_FLAGS(15, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
74 #define KVM_REQ_GLOBAL_CLOCK_UPDATE	KVM_ARCH_REQ(16)
75 #define KVM_REQ_APIC_PAGE_RELOAD \
76 	KVM_ARCH_REQ_FLAGS(17, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
77 #define KVM_REQ_HV_CRASH		KVM_ARCH_REQ(18)
78 #define KVM_REQ_IOAPIC_EOI_EXIT		KVM_ARCH_REQ(19)
79 #define KVM_REQ_HV_RESET		KVM_ARCH_REQ(20)
80 #define KVM_REQ_HV_EXIT			KVM_ARCH_REQ(21)
81 #define KVM_REQ_HV_STIMER		KVM_ARCH_REQ(22)
82 #define KVM_REQ_LOAD_EOI_EXITMAP	KVM_ARCH_REQ(23)
83 #define KVM_REQ_GET_NESTED_STATE_PAGES	KVM_ARCH_REQ(24)
84 #define KVM_REQ_APICV_UPDATE \
85 	KVM_ARCH_REQ_FLAGS(25, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
86 #define KVM_REQ_TLB_FLUSH_CURRENT	KVM_ARCH_REQ(26)
87 #define KVM_REQ_HV_TLB_FLUSH \
88 	KVM_ARCH_REQ_FLAGS(27, KVM_REQUEST_NO_WAKEUP)
89 #define KVM_REQ_APF_READY		KVM_ARCH_REQ(28)
90 #define KVM_REQ_MSR_FILTER_CHANGED	KVM_ARCH_REQ(29)
91 
92 #define CR0_RESERVED_BITS                                               \
93 	(~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \
94 			  | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \
95 			  | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG))
96 
97 #define CR4_RESERVED_BITS                                               \
98 	(~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\
99 			  | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE     \
100 			  | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR | X86_CR4_PCIDE \
101 			  | X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_FSGSBASE \
102 			  | X86_CR4_OSXMMEXCPT | X86_CR4_LA57 | X86_CR4_VMXE \
103 			  | X86_CR4_SMAP | X86_CR4_PKE | X86_CR4_UMIP))
104 
105 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR)
106 
107 
108 
109 #define INVALID_PAGE (~(hpa_t)0)
110 #define VALID_PAGE(x) ((x) != INVALID_PAGE)
111 
112 #define UNMAPPED_GVA (~(gpa_t)0)
113 
114 /* KVM Hugepage definitions for x86 */
115 #define KVM_MAX_HUGEPAGE_LEVEL	PG_LEVEL_1G
116 #define KVM_NR_PAGE_SIZES	(KVM_MAX_HUGEPAGE_LEVEL - PG_LEVEL_4K + 1)
117 #define KVM_HPAGE_GFN_SHIFT(x)	(((x) - 1) * 9)
118 #define KVM_HPAGE_SHIFT(x)	(PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
119 #define KVM_HPAGE_SIZE(x)	(1UL << KVM_HPAGE_SHIFT(x))
120 #define KVM_HPAGE_MASK(x)	(~(KVM_HPAGE_SIZE(x) - 1))
121 #define KVM_PAGES_PER_HPAGE(x)	(KVM_HPAGE_SIZE(x) / PAGE_SIZE)
122 
123 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
124 {
125 	/* KVM_HPAGE_GFN_SHIFT(PG_LEVEL_4K) must be 0. */
126 	return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
127 		(base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
128 }
129 
130 #define KVM_PERMILLE_MMU_PAGES 20
131 #define KVM_MIN_ALLOC_MMU_PAGES 64UL
132 #define KVM_MMU_HASH_SHIFT 12
133 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
134 #define KVM_MIN_FREE_MMU_PAGES 5
135 #define KVM_REFILL_PAGES 25
136 #define KVM_MAX_CPUID_ENTRIES 256
137 #define KVM_NR_FIXED_MTRR_REGION 88
138 #define KVM_NR_VAR_MTRR 8
139 
140 #define ASYNC_PF_PER_VCPU 64
141 
142 enum kvm_reg {
143 	VCPU_REGS_RAX = __VCPU_REGS_RAX,
144 	VCPU_REGS_RCX = __VCPU_REGS_RCX,
145 	VCPU_REGS_RDX = __VCPU_REGS_RDX,
146 	VCPU_REGS_RBX = __VCPU_REGS_RBX,
147 	VCPU_REGS_RSP = __VCPU_REGS_RSP,
148 	VCPU_REGS_RBP = __VCPU_REGS_RBP,
149 	VCPU_REGS_RSI = __VCPU_REGS_RSI,
150 	VCPU_REGS_RDI = __VCPU_REGS_RDI,
151 #ifdef CONFIG_X86_64
152 	VCPU_REGS_R8  = __VCPU_REGS_R8,
153 	VCPU_REGS_R9  = __VCPU_REGS_R9,
154 	VCPU_REGS_R10 = __VCPU_REGS_R10,
155 	VCPU_REGS_R11 = __VCPU_REGS_R11,
156 	VCPU_REGS_R12 = __VCPU_REGS_R12,
157 	VCPU_REGS_R13 = __VCPU_REGS_R13,
158 	VCPU_REGS_R14 = __VCPU_REGS_R14,
159 	VCPU_REGS_R15 = __VCPU_REGS_R15,
160 #endif
161 	VCPU_REGS_RIP,
162 	NR_VCPU_REGS,
163 
164 	VCPU_EXREG_PDPTR = NR_VCPU_REGS,
165 	VCPU_EXREG_CR0,
166 	VCPU_EXREG_CR3,
167 	VCPU_EXREG_CR4,
168 	VCPU_EXREG_RFLAGS,
169 	VCPU_EXREG_SEGMENTS,
170 	VCPU_EXREG_EXIT_INFO_1,
171 	VCPU_EXREG_EXIT_INFO_2,
172 };
173 
174 enum {
175 	VCPU_SREG_ES,
176 	VCPU_SREG_CS,
177 	VCPU_SREG_SS,
178 	VCPU_SREG_DS,
179 	VCPU_SREG_FS,
180 	VCPU_SREG_GS,
181 	VCPU_SREG_TR,
182 	VCPU_SREG_LDTR,
183 };
184 
185 enum exit_fastpath_completion {
186 	EXIT_FASTPATH_NONE,
187 	EXIT_FASTPATH_REENTER_GUEST,
188 	EXIT_FASTPATH_EXIT_HANDLED,
189 };
190 typedef enum exit_fastpath_completion fastpath_t;
191 
192 struct x86_emulate_ctxt;
193 struct x86_exception;
194 enum x86_intercept;
195 enum x86_intercept_stage;
196 
197 #define KVM_NR_DB_REGS	4
198 
199 #define DR6_BD		(1 << 13)
200 #define DR6_BS		(1 << 14)
201 #define DR6_BT		(1 << 15)
202 #define DR6_RTM		(1 << 16)
203 #define DR6_FIXED_1	0xfffe0ff0
204 #define DR6_INIT	0xffff0ff0
205 #define DR6_VOLATILE	0x0001e00f
206 
207 #define DR7_BP_EN_MASK	0x000000ff
208 #define DR7_GE		(1 << 9)
209 #define DR7_GD		(1 << 13)
210 #define DR7_FIXED_1	0x00000400
211 #define DR7_VOLATILE	0xffff2bff
212 
213 #define PFERR_PRESENT_BIT 0
214 #define PFERR_WRITE_BIT 1
215 #define PFERR_USER_BIT 2
216 #define PFERR_RSVD_BIT 3
217 #define PFERR_FETCH_BIT 4
218 #define PFERR_PK_BIT 5
219 #define PFERR_GUEST_FINAL_BIT 32
220 #define PFERR_GUEST_PAGE_BIT 33
221 
222 #define PFERR_PRESENT_MASK (1U << PFERR_PRESENT_BIT)
223 #define PFERR_WRITE_MASK (1U << PFERR_WRITE_BIT)
224 #define PFERR_USER_MASK (1U << PFERR_USER_BIT)
225 #define PFERR_RSVD_MASK (1U << PFERR_RSVD_BIT)
226 #define PFERR_FETCH_MASK (1U << PFERR_FETCH_BIT)
227 #define PFERR_PK_MASK (1U << PFERR_PK_BIT)
228 #define PFERR_GUEST_FINAL_MASK (1ULL << PFERR_GUEST_FINAL_BIT)
229 #define PFERR_GUEST_PAGE_MASK (1ULL << PFERR_GUEST_PAGE_BIT)
230 
231 #define PFERR_NESTED_GUEST_PAGE (PFERR_GUEST_PAGE_MASK |	\
232 				 PFERR_WRITE_MASK |		\
233 				 PFERR_PRESENT_MASK)
234 
235 /* apic attention bits */
236 #define KVM_APIC_CHECK_VAPIC	0
237 /*
238  * The following bit is set with PV-EOI, unset on EOI.
239  * We detect PV-EOI changes by guest by comparing
240  * this bit with PV-EOI in guest memory.
241  * See the implementation in apic_update_pv_eoi.
242  */
243 #define KVM_APIC_PV_EOI_PENDING	1
244 
245 struct kvm_kernel_irq_routing_entry;
246 
247 /*
248  * the pages used as guest page table on soft mmu are tracked by
249  * kvm_memory_slot.arch.gfn_track which is 16 bits, so the role bits used
250  * by indirect shadow page can not be more than 15 bits.
251  *
252  * Currently, we used 14 bits that are @level, @gpte_is_8_bytes, @quadrant, @access,
253  * @nxe, @cr0_wp, @smep_andnot_wp and @smap_andnot_wp.
254  */
255 union kvm_mmu_page_role {
256 	u32 word;
257 	struct {
258 		unsigned level:4;
259 		unsigned gpte_is_8_bytes:1;
260 		unsigned quadrant:2;
261 		unsigned direct:1;
262 		unsigned access:3;
263 		unsigned invalid:1;
264 		unsigned nxe:1;
265 		unsigned cr0_wp:1;
266 		unsigned smep_andnot_wp:1;
267 		unsigned smap_andnot_wp:1;
268 		unsigned ad_disabled:1;
269 		unsigned guest_mode:1;
270 		unsigned :6;
271 
272 		/*
273 		 * This is left at the top of the word so that
274 		 * kvm_memslots_for_spte_role can extract it with a
275 		 * simple shift.  While there is room, give it a whole
276 		 * byte so it is also faster to load it from memory.
277 		 */
278 		unsigned smm:8;
279 	};
280 };
281 
282 union kvm_mmu_extended_role {
283 /*
284  * This structure complements kvm_mmu_page_role caching everything needed for
285  * MMU configuration. If nothing in both these structures changed, MMU
286  * re-configuration can be skipped. @valid bit is set on first usage so we don't
287  * treat all-zero structure as valid data.
288  */
289 	u32 word;
290 	struct {
291 		unsigned int valid:1;
292 		unsigned int execonly:1;
293 		unsigned int cr0_pg:1;
294 		unsigned int cr4_pae:1;
295 		unsigned int cr4_pse:1;
296 		unsigned int cr4_pke:1;
297 		unsigned int cr4_smap:1;
298 		unsigned int cr4_smep:1;
299 		unsigned int maxphyaddr:6;
300 	};
301 };
302 
303 union kvm_mmu_role {
304 	u64 as_u64;
305 	struct {
306 		union kvm_mmu_page_role base;
307 		union kvm_mmu_extended_role ext;
308 	};
309 };
310 
311 struct kvm_rmap_head {
312 	unsigned long val;
313 };
314 
315 struct kvm_pio_request {
316 	unsigned long linear_rip;
317 	unsigned long count;
318 	int in;
319 	int port;
320 	int size;
321 };
322 
323 #define PT64_ROOT_MAX_LEVEL 5
324 
325 struct rsvd_bits_validate {
326 	u64 rsvd_bits_mask[2][PT64_ROOT_MAX_LEVEL];
327 	u64 bad_mt_xwr;
328 };
329 
330 struct kvm_mmu_root_info {
331 	gpa_t pgd;
332 	hpa_t hpa;
333 };
334 
335 #define KVM_MMU_ROOT_INFO_INVALID \
336 	((struct kvm_mmu_root_info) { .pgd = INVALID_PAGE, .hpa = INVALID_PAGE })
337 
338 #define KVM_MMU_NUM_PREV_ROOTS 3
339 
340 struct kvm_mmu_page;
341 
342 /*
343  * x86 supports 4 paging modes (5-level 64-bit, 4-level 64-bit, 3-level 32-bit,
344  * and 2-level 32-bit).  The kvm_mmu structure abstracts the details of the
345  * current mmu mode.
346  */
347 struct kvm_mmu {
348 	unsigned long (*get_guest_pgd)(struct kvm_vcpu *vcpu);
349 	u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
350 	int (*page_fault)(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u32 err,
351 			  bool prefault);
352 	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
353 				  struct x86_exception *fault);
354 	gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gpa_t gva_or_gpa,
355 			    u32 access, struct x86_exception *exception);
356 	gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
357 			       struct x86_exception *exception);
358 	int (*sync_page)(struct kvm_vcpu *vcpu,
359 			 struct kvm_mmu_page *sp);
360 	void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva, hpa_t root_hpa);
361 	void (*update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
362 			   u64 *spte, const void *pte);
363 	hpa_t root_hpa;
364 	gpa_t root_pgd;
365 	union kvm_mmu_role mmu_role;
366 	u8 root_level;
367 	u8 shadow_root_level;
368 	u8 ept_ad;
369 	bool direct_map;
370 	struct kvm_mmu_root_info prev_roots[KVM_MMU_NUM_PREV_ROOTS];
371 
372 	/*
373 	 * Bitmap; bit set = permission fault
374 	 * Byte index: page fault error code [4:1]
375 	 * Bit index: pte permissions in ACC_* format
376 	 */
377 	u8 permissions[16];
378 
379 	/*
380 	* The pkru_mask indicates if protection key checks are needed.  It
381 	* consists of 16 domains indexed by page fault error code bits [4:1],
382 	* with PFEC.RSVD replaced by ACC_USER_MASK from the page tables.
383 	* Each domain has 2 bits which are ANDed with AD and WD from PKRU.
384 	*/
385 	u32 pkru_mask;
386 
387 	u64 *pae_root;
388 	u64 *lm_root;
389 
390 	/*
391 	 * check zero bits on shadow page table entries, these
392 	 * bits include not only hardware reserved bits but also
393 	 * the bits spte never used.
394 	 */
395 	struct rsvd_bits_validate shadow_zero_check;
396 
397 	struct rsvd_bits_validate guest_rsvd_check;
398 
399 	/* Can have large pages at levels 2..last_nonleaf_level-1. */
400 	u8 last_nonleaf_level;
401 
402 	bool nx;
403 
404 	u64 pdptrs[4]; /* pae */
405 };
406 
407 struct kvm_tlb_range {
408 	u64 start_gfn;
409 	u64 pages;
410 };
411 
412 enum pmc_type {
413 	KVM_PMC_GP = 0,
414 	KVM_PMC_FIXED,
415 };
416 
417 struct kvm_pmc {
418 	enum pmc_type type;
419 	u8 idx;
420 	u64 counter;
421 	u64 eventsel;
422 	struct perf_event *perf_event;
423 	struct kvm_vcpu *vcpu;
424 	/*
425 	 * eventsel value for general purpose counters,
426 	 * ctrl value for fixed counters.
427 	 */
428 	u64 current_config;
429 };
430 
431 struct kvm_pmu {
432 	unsigned nr_arch_gp_counters;
433 	unsigned nr_arch_fixed_counters;
434 	unsigned available_event_types;
435 	u64 fixed_ctr_ctrl;
436 	u64 global_ctrl;
437 	u64 global_status;
438 	u64 global_ovf_ctrl;
439 	u64 counter_bitmask[2];
440 	u64 global_ctrl_mask;
441 	u64 global_ovf_ctrl_mask;
442 	u64 reserved_bits;
443 	u8 version;
444 	struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
445 	struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED];
446 	struct irq_work irq_work;
447 	DECLARE_BITMAP(reprogram_pmi, X86_PMC_IDX_MAX);
448 	DECLARE_BITMAP(all_valid_pmc_idx, X86_PMC_IDX_MAX);
449 	DECLARE_BITMAP(pmc_in_use, X86_PMC_IDX_MAX);
450 
451 	/*
452 	 * The gate to release perf_events not marked in
453 	 * pmc_in_use only once in a vcpu time slice.
454 	 */
455 	bool need_cleanup;
456 
457 	/*
458 	 * The total number of programmed perf_events and it helps to avoid
459 	 * redundant check before cleanup if guest don't use vPMU at all.
460 	 */
461 	u8 event_count;
462 };
463 
464 struct kvm_pmu_ops;
465 
466 enum {
467 	KVM_DEBUGREG_BP_ENABLED = 1,
468 	KVM_DEBUGREG_WONT_EXIT = 2,
469 	KVM_DEBUGREG_RELOAD = 4,
470 };
471 
472 struct kvm_mtrr_range {
473 	u64 base;
474 	u64 mask;
475 	struct list_head node;
476 };
477 
478 struct kvm_mtrr {
479 	struct kvm_mtrr_range var_ranges[KVM_NR_VAR_MTRR];
480 	mtrr_type fixed_ranges[KVM_NR_FIXED_MTRR_REGION];
481 	u64 deftype;
482 
483 	struct list_head head;
484 };
485 
486 /* Hyper-V SynIC timer */
487 struct kvm_vcpu_hv_stimer {
488 	struct hrtimer timer;
489 	int index;
490 	union hv_stimer_config config;
491 	u64 count;
492 	u64 exp_time;
493 	struct hv_message msg;
494 	bool msg_pending;
495 };
496 
497 /* Hyper-V synthetic interrupt controller (SynIC)*/
498 struct kvm_vcpu_hv_synic {
499 	u64 version;
500 	u64 control;
501 	u64 msg_page;
502 	u64 evt_page;
503 	atomic64_t sint[HV_SYNIC_SINT_COUNT];
504 	atomic_t sint_to_gsi[HV_SYNIC_SINT_COUNT];
505 	DECLARE_BITMAP(auto_eoi_bitmap, 256);
506 	DECLARE_BITMAP(vec_bitmap, 256);
507 	bool active;
508 	bool dont_zero_synic_pages;
509 };
510 
511 /* Hyper-V per vcpu emulation context */
512 struct kvm_vcpu_hv {
513 	u32 vp_index;
514 	u64 hv_vapic;
515 	s64 runtime_offset;
516 	struct kvm_vcpu_hv_synic synic;
517 	struct kvm_hyperv_exit exit;
518 	struct kvm_vcpu_hv_stimer stimer[HV_SYNIC_STIMER_COUNT];
519 	DECLARE_BITMAP(stimer_pending_bitmap, HV_SYNIC_STIMER_COUNT);
520 	cpumask_t tlb_flush;
521 };
522 
523 struct kvm_vcpu_arch {
524 	/*
525 	 * rip and regs accesses must go through
526 	 * kvm_{register,rip}_{read,write} functions.
527 	 */
528 	unsigned long regs[NR_VCPU_REGS];
529 	u32 regs_avail;
530 	u32 regs_dirty;
531 
532 	unsigned long cr0;
533 	unsigned long cr0_guest_owned_bits;
534 	unsigned long cr2;
535 	unsigned long cr3;
536 	unsigned long cr4;
537 	unsigned long cr4_guest_owned_bits;
538 	unsigned long cr4_guest_rsvd_bits;
539 	unsigned long cr8;
540 	u32 host_pkru;
541 	u32 pkru;
542 	u32 hflags;
543 	u64 efer;
544 	u64 apic_base;
545 	struct kvm_lapic *apic;    /* kernel irqchip context */
546 	bool apicv_active;
547 	bool load_eoi_exitmap_pending;
548 	DECLARE_BITMAP(ioapic_handled_vectors, 256);
549 	unsigned long apic_attention;
550 	int32_t apic_arb_prio;
551 	int mp_state;
552 	u64 ia32_misc_enable_msr;
553 	u64 smbase;
554 	u64 smi_count;
555 	bool tpr_access_reporting;
556 	bool xsaves_enabled;
557 	u64 ia32_xss;
558 	u64 microcode_version;
559 	u64 arch_capabilities;
560 	u64 perf_capabilities;
561 
562 	/*
563 	 * Paging state of the vcpu
564 	 *
565 	 * If the vcpu runs in guest mode with two level paging this still saves
566 	 * the paging mode of the l1 guest. This context is always used to
567 	 * handle faults.
568 	 */
569 	struct kvm_mmu *mmu;
570 
571 	/* Non-nested MMU for L1 */
572 	struct kvm_mmu root_mmu;
573 
574 	/* L1 MMU when running nested */
575 	struct kvm_mmu guest_mmu;
576 
577 	/*
578 	 * Paging state of an L2 guest (used for nested npt)
579 	 *
580 	 * This context will save all necessary information to walk page tables
581 	 * of an L2 guest. This context is only initialized for page table
582 	 * walking and not for faulting since we never handle l2 page faults on
583 	 * the host.
584 	 */
585 	struct kvm_mmu nested_mmu;
586 
587 	/*
588 	 * Pointer to the mmu context currently used for
589 	 * gva_to_gpa translations.
590 	 */
591 	struct kvm_mmu *walk_mmu;
592 
593 	struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
594 	struct kvm_mmu_memory_cache mmu_shadow_page_cache;
595 	struct kvm_mmu_memory_cache mmu_gfn_array_cache;
596 	struct kvm_mmu_memory_cache mmu_page_header_cache;
597 
598 	/*
599 	 * QEMU userspace and the guest each have their own FPU state.
600 	 * In vcpu_run, we switch between the user and guest FPU contexts.
601 	 * While running a VCPU, the VCPU thread will have the guest FPU
602 	 * context.
603 	 *
604 	 * Note that while the PKRU state lives inside the fpu registers,
605 	 * it is switched out separately at VMENTER and VMEXIT time. The
606 	 * "guest_fpu" state here contains the guest FPU context, with the
607 	 * host PRKU bits.
608 	 */
609 	struct fpu *user_fpu;
610 	struct fpu *guest_fpu;
611 
612 	u64 xcr0;
613 	u64 guest_supported_xcr0;
614 
615 	struct kvm_pio_request pio;
616 	void *pio_data;
617 	void *guest_ins_data;
618 
619 	u8 event_exit_inst_len;
620 
621 	struct kvm_queued_exception {
622 		bool pending;
623 		bool injected;
624 		bool has_error_code;
625 		u8 nr;
626 		u32 error_code;
627 		unsigned long payload;
628 		bool has_payload;
629 		u8 nested_apf;
630 	} exception;
631 
632 	struct kvm_queued_interrupt {
633 		bool injected;
634 		bool soft;
635 		u8 nr;
636 	} interrupt;
637 
638 	int halt_request; /* real mode on Intel only */
639 
640 	int cpuid_nent;
641 	struct kvm_cpuid_entry2 *cpuid_entries;
642 
643 	unsigned long cr3_lm_rsvd_bits;
644 	int maxphyaddr;
645 	int max_tdp_level;
646 
647 	/* emulate context */
648 
649 	struct x86_emulate_ctxt *emulate_ctxt;
650 	bool emulate_regs_need_sync_to_vcpu;
651 	bool emulate_regs_need_sync_from_vcpu;
652 	int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
653 
654 	gpa_t time;
655 	struct pvclock_vcpu_time_info hv_clock;
656 	unsigned int hw_tsc_khz;
657 	struct gfn_to_hva_cache pv_time;
658 	bool pv_time_enabled;
659 	/* set guest stopped flag in pvclock flags field */
660 	bool pvclock_set_guest_stopped_request;
661 
662 	struct {
663 		u8 preempted;
664 		u64 msr_val;
665 		u64 last_steal;
666 		struct gfn_to_pfn_cache cache;
667 	} st;
668 
669 	u64 l1_tsc_offset;
670 	u64 tsc_offset;
671 	u64 last_guest_tsc;
672 	u64 last_host_tsc;
673 	u64 tsc_offset_adjustment;
674 	u64 this_tsc_nsec;
675 	u64 this_tsc_write;
676 	u64 this_tsc_generation;
677 	bool tsc_catchup;
678 	bool tsc_always_catchup;
679 	s8 virtual_tsc_shift;
680 	u32 virtual_tsc_mult;
681 	u32 virtual_tsc_khz;
682 	s64 ia32_tsc_adjust_msr;
683 	u64 msr_ia32_power_ctl;
684 	u64 tsc_scaling_ratio;
685 
686 	atomic_t nmi_queued;  /* unprocessed asynchronous NMIs */
687 	unsigned nmi_pending; /* NMI queued after currently running handler */
688 	bool nmi_injected;    /* Trying to inject an NMI this entry */
689 	bool smi_pending;    /* SMI queued after currently running handler */
690 
691 	struct kvm_mtrr mtrr_state;
692 	u64 pat;
693 
694 	unsigned switch_db_regs;
695 	unsigned long db[KVM_NR_DB_REGS];
696 	unsigned long dr6;
697 	unsigned long dr7;
698 	unsigned long eff_db[KVM_NR_DB_REGS];
699 	unsigned long guest_debug_dr7;
700 	u64 msr_platform_info;
701 	u64 msr_misc_features_enables;
702 
703 	u64 mcg_cap;
704 	u64 mcg_status;
705 	u64 mcg_ctl;
706 	u64 mcg_ext_ctl;
707 	u64 *mce_banks;
708 
709 	/* Cache MMIO info */
710 	u64 mmio_gva;
711 	unsigned mmio_access;
712 	gfn_t mmio_gfn;
713 	u64 mmio_gen;
714 
715 	struct kvm_pmu pmu;
716 
717 	/* used for guest single stepping over the given code position */
718 	unsigned long singlestep_rip;
719 
720 	struct kvm_vcpu_hv hyperv;
721 
722 	cpumask_var_t wbinvd_dirty_mask;
723 
724 	unsigned long last_retry_eip;
725 	unsigned long last_retry_addr;
726 
727 	struct {
728 		bool halted;
729 		gfn_t gfns[ASYNC_PF_PER_VCPU];
730 		struct gfn_to_hva_cache data;
731 		u64 msr_en_val; /* MSR_KVM_ASYNC_PF_EN */
732 		u64 msr_int_val; /* MSR_KVM_ASYNC_PF_INT */
733 		u16 vec;
734 		u32 id;
735 		bool send_user_only;
736 		u32 host_apf_flags;
737 		unsigned long nested_apf_token;
738 		bool delivery_as_pf_vmexit;
739 		bool pageready_pending;
740 	} apf;
741 
742 	/* OSVW MSRs (AMD only) */
743 	struct {
744 		u64 length;
745 		u64 status;
746 	} osvw;
747 
748 	struct {
749 		u64 msr_val;
750 		struct gfn_to_hva_cache data;
751 	} pv_eoi;
752 
753 	u64 msr_kvm_poll_control;
754 
755 	/*
756 	 * Indicates the guest is trying to write a gfn that contains one or
757 	 * more of the PTEs used to translate the write itself, i.e. the access
758 	 * is changing its own translation in the guest page tables.  KVM exits
759 	 * to userspace if emulation of the faulting instruction fails and this
760 	 * flag is set, as KVM cannot make forward progress.
761 	 *
762 	 * If emulation fails for a write to guest page tables, KVM unprotects
763 	 * (zaps) the shadow page for the target gfn and resumes the guest to
764 	 * retry the non-emulatable instruction (on hardware).  Unprotecting the
765 	 * gfn doesn't allow forward progress for a self-changing access because
766 	 * doing so also zaps the translation for the gfn, i.e. retrying the
767 	 * instruction will hit a !PRESENT fault, which results in a new shadow
768 	 * page and sends KVM back to square one.
769 	 */
770 	bool write_fault_to_shadow_pgtable;
771 
772 	/* set at EPT violation at this point */
773 	unsigned long exit_qualification;
774 
775 	/* pv related host specific info */
776 	struct {
777 		bool pv_unhalted;
778 	} pv;
779 
780 	int pending_ioapic_eoi;
781 	int pending_external_vector;
782 
783 	/* be preempted when it's in kernel-mode(cpl=0) */
784 	bool preempted_in_kernel;
785 
786 	/* Flush the L1 Data cache for L1TF mitigation on VMENTER */
787 	bool l1tf_flush_l1d;
788 
789 	/* Host CPU on which VM-entry was most recently attempted */
790 	unsigned int last_vmentry_cpu;
791 
792 	/* AMD MSRC001_0015 Hardware Configuration */
793 	u64 msr_hwcr;
794 
795 	/* pv related cpuid info */
796 	struct {
797 		/*
798 		 * value of the eax register in the KVM_CPUID_FEATURES CPUID
799 		 * leaf.
800 		 */
801 		u32 features;
802 
803 		/*
804 		 * indicates whether pv emulation should be disabled if features
805 		 * are not present in the guest's cpuid
806 		 */
807 		bool enforce;
808 	} pv_cpuid;
809 
810 	/* Protected Guests */
811 	bool guest_state_protected;
812 };
813 
814 struct kvm_lpage_info {
815 	int disallow_lpage;
816 };
817 
818 struct kvm_arch_memory_slot {
819 	struct kvm_rmap_head *rmap[KVM_NR_PAGE_SIZES];
820 	struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
821 	unsigned short *gfn_track[KVM_PAGE_TRACK_MAX];
822 };
823 
824 /*
825  * We use as the mode the number of bits allocated in the LDR for the
826  * logical processor ID.  It happens that these are all powers of two.
827  * This makes it is very easy to detect cases where the APICs are
828  * configured for multiple modes; in that case, we cannot use the map and
829  * hence cannot use kvm_irq_delivery_to_apic_fast either.
830  */
831 #define KVM_APIC_MODE_XAPIC_CLUSTER          4
832 #define KVM_APIC_MODE_XAPIC_FLAT             8
833 #define KVM_APIC_MODE_X2APIC                16
834 
835 struct kvm_apic_map {
836 	struct rcu_head rcu;
837 	u8 mode;
838 	u32 max_apic_id;
839 	union {
840 		struct kvm_lapic *xapic_flat_map[8];
841 		struct kvm_lapic *xapic_cluster_map[16][4];
842 	};
843 	struct kvm_lapic *phys_map[];
844 };
845 
846 /* Hyper-V synthetic debugger (SynDbg)*/
847 struct kvm_hv_syndbg {
848 	struct {
849 		u64 control;
850 		u64 status;
851 		u64 send_page;
852 		u64 recv_page;
853 		u64 pending_page;
854 	} control;
855 	u64 options;
856 };
857 
858 /* Hyper-V emulation context */
859 struct kvm_hv {
860 	struct mutex hv_lock;
861 	u64 hv_guest_os_id;
862 	u64 hv_hypercall;
863 	u64 hv_tsc_page;
864 
865 	/* Hyper-v based guest crash (NT kernel bugcheck) parameters */
866 	u64 hv_crash_param[HV_X64_MSR_CRASH_PARAMS];
867 	u64 hv_crash_ctl;
868 
869 	struct ms_hyperv_tsc_page tsc_ref;
870 
871 	struct idr conn_to_evt;
872 
873 	u64 hv_reenlightenment_control;
874 	u64 hv_tsc_emulation_control;
875 	u64 hv_tsc_emulation_status;
876 
877 	/* How many vCPUs have VP index != vCPU index */
878 	atomic_t num_mismatched_vp_indexes;
879 
880 	struct hv_partition_assist_pg *hv_pa_pg;
881 	struct kvm_hv_syndbg hv_syndbg;
882 };
883 
884 struct msr_bitmap_range {
885 	u32 flags;
886 	u32 nmsrs;
887 	u32 base;
888 	unsigned long *bitmap;
889 };
890 
891 enum kvm_irqchip_mode {
892 	KVM_IRQCHIP_NONE,
893 	KVM_IRQCHIP_KERNEL,       /* created with KVM_CREATE_IRQCHIP */
894 	KVM_IRQCHIP_SPLIT,        /* created with KVM_CAP_SPLIT_IRQCHIP */
895 };
896 
897 #define APICV_INHIBIT_REASON_DISABLE    0
898 #define APICV_INHIBIT_REASON_HYPERV     1
899 #define APICV_INHIBIT_REASON_NESTED     2
900 #define APICV_INHIBIT_REASON_IRQWIN     3
901 #define APICV_INHIBIT_REASON_PIT_REINJ  4
902 #define APICV_INHIBIT_REASON_X2APIC	5
903 
904 struct kvm_arch {
905 	unsigned long n_used_mmu_pages;
906 	unsigned long n_requested_mmu_pages;
907 	unsigned long n_max_mmu_pages;
908 	unsigned int indirect_shadow_pages;
909 	u8 mmu_valid_gen;
910 	struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
911 	/*
912 	 * Hash table of struct kvm_mmu_page.
913 	 */
914 	struct list_head active_mmu_pages;
915 	struct list_head zapped_obsolete_pages;
916 	struct list_head lpage_disallowed_mmu_pages;
917 	struct kvm_page_track_notifier_node mmu_sp_tracker;
918 	struct kvm_page_track_notifier_head track_notifier_head;
919 
920 	struct list_head assigned_dev_head;
921 	struct iommu_domain *iommu_domain;
922 	bool iommu_noncoherent;
923 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA
924 	atomic_t noncoherent_dma_count;
925 #define __KVM_HAVE_ARCH_ASSIGNED_DEVICE
926 	atomic_t assigned_device_count;
927 	struct kvm_pic *vpic;
928 	struct kvm_ioapic *vioapic;
929 	struct kvm_pit *vpit;
930 	atomic_t vapics_in_nmi_mode;
931 	struct mutex apic_map_lock;
932 	struct kvm_apic_map *apic_map;
933 	atomic_t apic_map_dirty;
934 
935 	bool apic_access_page_done;
936 	unsigned long apicv_inhibit_reasons;
937 
938 	gpa_t wall_clock;
939 
940 	bool mwait_in_guest;
941 	bool hlt_in_guest;
942 	bool pause_in_guest;
943 	bool cstate_in_guest;
944 
945 	unsigned long irq_sources_bitmap;
946 	s64 kvmclock_offset;
947 	raw_spinlock_t tsc_write_lock;
948 	u64 last_tsc_nsec;
949 	u64 last_tsc_write;
950 	u32 last_tsc_khz;
951 	u64 cur_tsc_nsec;
952 	u64 cur_tsc_write;
953 	u64 cur_tsc_offset;
954 	u64 cur_tsc_generation;
955 	int nr_vcpus_matched_tsc;
956 
957 	spinlock_t pvclock_gtod_sync_lock;
958 	bool use_master_clock;
959 	u64 master_kernel_ns;
960 	u64 master_cycle_now;
961 	struct delayed_work kvmclock_update_work;
962 	struct delayed_work kvmclock_sync_work;
963 
964 	struct kvm_xen_hvm_config xen_hvm_config;
965 
966 	/* reads protected by irq_srcu, writes by irq_lock */
967 	struct hlist_head mask_notifier_list;
968 
969 	struct kvm_hv hyperv;
970 
971 	#ifdef CONFIG_KVM_MMU_AUDIT
972 	int audit_point;
973 	#endif
974 
975 	bool backwards_tsc_observed;
976 	bool boot_vcpu_runs_old_kvmclock;
977 	u32 bsp_vcpu_id;
978 
979 	u64 disabled_quirks;
980 
981 	enum kvm_irqchip_mode irqchip_mode;
982 	u8 nr_reserved_ioapic_pins;
983 
984 	bool disabled_lapic_found;
985 
986 	bool x2apic_format;
987 	bool x2apic_broadcast_quirk_disabled;
988 
989 	bool guest_can_read_msr_platform_info;
990 	bool exception_payload_enabled;
991 
992 	/* Deflect RDMSR and WRMSR to user space when they trigger a #GP */
993 	u32 user_space_msr_mask;
994 
995 	struct {
996 		u8 count;
997 		bool default_allow:1;
998 		struct msr_bitmap_range ranges[16];
999 	} msr_filter;
1000 
1001 	struct kvm_pmu_event_filter *pmu_event_filter;
1002 	struct task_struct *nx_lpage_recovery_thread;
1003 
1004 	/*
1005 	 * Whether the TDP MMU is enabled for this VM. This contains a
1006 	 * snapshot of the TDP MMU module parameter from when the VM was
1007 	 * created and remains unchanged for the life of the VM. If this is
1008 	 * true, TDP MMU handler functions will run for various MMU
1009 	 * operations.
1010 	 */
1011 	bool tdp_mmu_enabled;
1012 
1013 	/*
1014 	 * List of struct kvmp_mmu_pages being used as roots.
1015 	 * All struct kvm_mmu_pages in the list should have
1016 	 * tdp_mmu_page set.
1017 	 * All struct kvm_mmu_pages in the list should have a positive
1018 	 * root_count except when a thread holds the MMU lock and is removing
1019 	 * an entry from the list.
1020 	 */
1021 	struct list_head tdp_mmu_roots;
1022 
1023 	/*
1024 	 * List of struct kvmp_mmu_pages not being used as roots.
1025 	 * All struct kvm_mmu_pages in the list should have
1026 	 * tdp_mmu_page set and a root_count of 0.
1027 	 */
1028 	struct list_head tdp_mmu_pages;
1029 };
1030 
1031 struct kvm_vm_stat {
1032 	ulong mmu_shadow_zapped;
1033 	ulong mmu_pte_write;
1034 	ulong mmu_pte_updated;
1035 	ulong mmu_pde_zapped;
1036 	ulong mmu_flooded;
1037 	ulong mmu_recycled;
1038 	ulong mmu_cache_miss;
1039 	ulong mmu_unsync;
1040 	ulong remote_tlb_flush;
1041 	ulong lpages;
1042 	ulong nx_lpage_splits;
1043 	ulong max_mmu_page_hash_collisions;
1044 };
1045 
1046 struct kvm_vcpu_stat {
1047 	u64 pf_fixed;
1048 	u64 pf_guest;
1049 	u64 tlb_flush;
1050 	u64 invlpg;
1051 
1052 	u64 exits;
1053 	u64 io_exits;
1054 	u64 mmio_exits;
1055 	u64 signal_exits;
1056 	u64 irq_window_exits;
1057 	u64 nmi_window_exits;
1058 	u64 l1d_flush;
1059 	u64 halt_exits;
1060 	u64 halt_successful_poll;
1061 	u64 halt_attempted_poll;
1062 	u64 halt_poll_invalid;
1063 	u64 halt_wakeup;
1064 	u64 request_irq_exits;
1065 	u64 irq_exits;
1066 	u64 host_state_reload;
1067 	u64 fpu_reload;
1068 	u64 insn_emulation;
1069 	u64 insn_emulation_fail;
1070 	u64 hypercalls;
1071 	u64 irq_injections;
1072 	u64 nmi_injections;
1073 	u64 req_event;
1074 	u64 halt_poll_success_ns;
1075 	u64 halt_poll_fail_ns;
1076 };
1077 
1078 struct x86_instruction_info;
1079 
1080 struct msr_data {
1081 	bool host_initiated;
1082 	u32 index;
1083 	u64 data;
1084 };
1085 
1086 struct kvm_lapic_irq {
1087 	u32 vector;
1088 	u16 delivery_mode;
1089 	u16 dest_mode;
1090 	bool level;
1091 	u16 trig_mode;
1092 	u32 shorthand;
1093 	u32 dest_id;
1094 	bool msi_redir_hint;
1095 };
1096 
1097 static inline u16 kvm_lapic_irq_dest_mode(bool dest_mode_logical)
1098 {
1099 	return dest_mode_logical ? APIC_DEST_LOGICAL : APIC_DEST_PHYSICAL;
1100 }
1101 
1102 struct kvm_x86_ops {
1103 	int (*hardware_enable)(void);
1104 	void (*hardware_disable)(void);
1105 	void (*hardware_unsetup)(void);
1106 	bool (*cpu_has_accelerated_tpr)(void);
1107 	bool (*has_emulated_msr)(struct kvm *kvm, u32 index);
1108 	void (*vcpu_after_set_cpuid)(struct kvm_vcpu *vcpu);
1109 
1110 	unsigned int vm_size;
1111 	int (*vm_init)(struct kvm *kvm);
1112 	void (*vm_destroy)(struct kvm *kvm);
1113 
1114 	/* Create, but do not attach this VCPU */
1115 	int (*vcpu_create)(struct kvm_vcpu *vcpu);
1116 	void (*vcpu_free)(struct kvm_vcpu *vcpu);
1117 	void (*vcpu_reset)(struct kvm_vcpu *vcpu, bool init_event);
1118 
1119 	void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
1120 	void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
1121 	void (*vcpu_put)(struct kvm_vcpu *vcpu);
1122 
1123 	void (*update_exception_bitmap)(struct kvm_vcpu *vcpu);
1124 	int (*get_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1125 	int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1126 	u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
1127 	void (*get_segment)(struct kvm_vcpu *vcpu,
1128 			    struct kvm_segment *var, int seg);
1129 	int (*get_cpl)(struct kvm_vcpu *vcpu);
1130 	void (*set_segment)(struct kvm_vcpu *vcpu,
1131 			    struct kvm_segment *var, int seg);
1132 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
1133 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
1134 	bool (*is_valid_cr4)(struct kvm_vcpu *vcpu, unsigned long cr0);
1135 	void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
1136 	int (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
1137 	void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1138 	void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1139 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1140 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1141 	void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu);
1142 	void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
1143 	void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
1144 	unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
1145 	void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
1146 
1147 	void (*tlb_flush_all)(struct kvm_vcpu *vcpu);
1148 	void (*tlb_flush_current)(struct kvm_vcpu *vcpu);
1149 	int  (*tlb_remote_flush)(struct kvm *kvm);
1150 	int  (*tlb_remote_flush_with_range)(struct kvm *kvm,
1151 			struct kvm_tlb_range *range);
1152 
1153 	/*
1154 	 * Flush any TLB entries associated with the given GVA.
1155 	 * Does not need to flush GPA->HPA mappings.
1156 	 * Can potentially get non-canonical addresses through INVLPGs, which
1157 	 * the implementation may choose to ignore if appropriate.
1158 	 */
1159 	void (*tlb_flush_gva)(struct kvm_vcpu *vcpu, gva_t addr);
1160 
1161 	/*
1162 	 * Flush any TLB entries created by the guest.  Like tlb_flush_gva(),
1163 	 * does not need to flush GPA->HPA mappings.
1164 	 */
1165 	void (*tlb_flush_guest)(struct kvm_vcpu *vcpu);
1166 
1167 	enum exit_fastpath_completion (*run)(struct kvm_vcpu *vcpu);
1168 	int (*handle_exit)(struct kvm_vcpu *vcpu,
1169 		enum exit_fastpath_completion exit_fastpath);
1170 	int (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
1171 	void (*update_emulated_instruction)(struct kvm_vcpu *vcpu);
1172 	void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
1173 	u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu);
1174 	void (*patch_hypercall)(struct kvm_vcpu *vcpu,
1175 				unsigned char *hypercall_addr);
1176 	void (*set_irq)(struct kvm_vcpu *vcpu);
1177 	void (*set_nmi)(struct kvm_vcpu *vcpu);
1178 	void (*queue_exception)(struct kvm_vcpu *vcpu);
1179 	void (*cancel_injection)(struct kvm_vcpu *vcpu);
1180 	int (*interrupt_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1181 	int (*nmi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1182 	bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
1183 	void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
1184 	void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
1185 	void (*enable_irq_window)(struct kvm_vcpu *vcpu);
1186 	void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
1187 	bool (*check_apicv_inhibit_reasons)(ulong bit);
1188 	void (*pre_update_apicv_exec_ctrl)(struct kvm *kvm, bool activate);
1189 	void (*refresh_apicv_exec_ctrl)(struct kvm_vcpu *vcpu);
1190 	void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
1191 	void (*hwapic_isr_update)(struct kvm_vcpu *vcpu, int isr);
1192 	bool (*guest_apic_has_interrupt)(struct kvm_vcpu *vcpu);
1193 	void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
1194 	void (*set_virtual_apic_mode)(struct kvm_vcpu *vcpu);
1195 	void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu);
1196 	int (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
1197 	int (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
1198 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
1199 	int (*set_identity_map_addr)(struct kvm *kvm, u64 ident_addr);
1200 	u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
1201 
1202 	void (*load_mmu_pgd)(struct kvm_vcpu *vcpu, unsigned long pgd,
1203 			     int pgd_level);
1204 
1205 	bool (*has_wbinvd_exit)(void);
1206 
1207 	/* Returns actual tsc_offset set in active VMCS */
1208 	u64 (*write_l1_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
1209 
1210 	/*
1211 	 * Retrieve somewhat arbitrary exit information.  Intended to be used
1212 	 * only from within tracepoints to avoid VMREADs when tracing is off.
1213 	 */
1214 	void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2,
1215 			      u32 *exit_int_info, u32 *exit_int_info_err_code);
1216 
1217 	int (*check_intercept)(struct kvm_vcpu *vcpu,
1218 			       struct x86_instruction_info *info,
1219 			       enum x86_intercept_stage stage,
1220 			       struct x86_exception *exception);
1221 	void (*handle_exit_irqoff)(struct kvm_vcpu *vcpu);
1222 
1223 	void (*request_immediate_exit)(struct kvm_vcpu *vcpu);
1224 
1225 	void (*sched_in)(struct kvm_vcpu *kvm, int cpu);
1226 
1227 	/*
1228 	 * Arch-specific dirty logging hooks. These hooks are only supposed to
1229 	 * be valid if the specific arch has hardware-accelerated dirty logging
1230 	 * mechanism. Currently only for PML on VMX.
1231 	 *
1232 	 *  - slot_enable_log_dirty:
1233 	 *	called when enabling log dirty mode for the slot.
1234 	 *  - slot_disable_log_dirty:
1235 	 *	called when disabling log dirty mode for the slot.
1236 	 *	also called when slot is created with log dirty disabled.
1237 	 *  - flush_log_dirty:
1238 	 *	called before reporting dirty_bitmap to userspace.
1239 	 *  - enable_log_dirty_pt_masked:
1240 	 *	called when reenabling log dirty for the GFNs in the mask after
1241 	 *	corresponding bits are cleared in slot->dirty_bitmap.
1242 	 */
1243 	void (*slot_enable_log_dirty)(struct kvm *kvm,
1244 				      struct kvm_memory_slot *slot);
1245 	void (*slot_disable_log_dirty)(struct kvm *kvm,
1246 				       struct kvm_memory_slot *slot);
1247 	void (*flush_log_dirty)(struct kvm *kvm);
1248 	void (*enable_log_dirty_pt_masked)(struct kvm *kvm,
1249 					   struct kvm_memory_slot *slot,
1250 					   gfn_t offset, unsigned long mask);
1251 	int (*cpu_dirty_log_size)(void);
1252 
1253 	/* pmu operations of sub-arch */
1254 	const struct kvm_pmu_ops *pmu_ops;
1255 	const struct kvm_x86_nested_ops *nested_ops;
1256 
1257 	/*
1258 	 * Architecture specific hooks for vCPU blocking due to
1259 	 * HLT instruction.
1260 	 * Returns for .pre_block():
1261 	 *    - 0 means continue to block the vCPU.
1262 	 *    - 1 means we cannot block the vCPU since some event
1263 	 *        happens during this period, such as, 'ON' bit in
1264 	 *        posted-interrupts descriptor is set.
1265 	 */
1266 	int (*pre_block)(struct kvm_vcpu *vcpu);
1267 	void (*post_block)(struct kvm_vcpu *vcpu);
1268 
1269 	void (*vcpu_blocking)(struct kvm_vcpu *vcpu);
1270 	void (*vcpu_unblocking)(struct kvm_vcpu *vcpu);
1271 
1272 	int (*update_pi_irte)(struct kvm *kvm, unsigned int host_irq,
1273 			      uint32_t guest_irq, bool set);
1274 	void (*apicv_post_state_restore)(struct kvm_vcpu *vcpu);
1275 	bool (*dy_apicv_has_pending_interrupt)(struct kvm_vcpu *vcpu);
1276 
1277 	int (*set_hv_timer)(struct kvm_vcpu *vcpu, u64 guest_deadline_tsc,
1278 			    bool *expired);
1279 	void (*cancel_hv_timer)(struct kvm_vcpu *vcpu);
1280 
1281 	void (*setup_mce)(struct kvm_vcpu *vcpu);
1282 
1283 	int (*smi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1284 	int (*pre_enter_smm)(struct kvm_vcpu *vcpu, char *smstate);
1285 	int (*pre_leave_smm)(struct kvm_vcpu *vcpu, const char *smstate);
1286 	void (*enable_smi_window)(struct kvm_vcpu *vcpu);
1287 
1288 	int (*mem_enc_op)(struct kvm *kvm, void __user *argp);
1289 	int (*mem_enc_reg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1290 	int (*mem_enc_unreg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1291 
1292 	int (*get_msr_feature)(struct kvm_msr_entry *entry);
1293 
1294 	bool (*can_emulate_instruction)(struct kvm_vcpu *vcpu, void *insn, int insn_len);
1295 
1296 	bool (*apic_init_signal_blocked)(struct kvm_vcpu *vcpu);
1297 	int (*enable_direct_tlbflush)(struct kvm_vcpu *vcpu);
1298 
1299 	void (*migrate_timers)(struct kvm_vcpu *vcpu);
1300 	void (*msr_filter_changed)(struct kvm_vcpu *vcpu);
1301 	int (*complete_emulated_msr)(struct kvm_vcpu *vcpu, int err);
1302 
1303 	void (*vcpu_deliver_sipi_vector)(struct kvm_vcpu *vcpu, u8 vector);
1304 };
1305 
1306 struct kvm_x86_nested_ops {
1307 	int (*check_events)(struct kvm_vcpu *vcpu);
1308 	bool (*hv_timer_pending)(struct kvm_vcpu *vcpu);
1309 	int (*get_state)(struct kvm_vcpu *vcpu,
1310 			 struct kvm_nested_state __user *user_kvm_nested_state,
1311 			 unsigned user_data_size);
1312 	int (*set_state)(struct kvm_vcpu *vcpu,
1313 			 struct kvm_nested_state __user *user_kvm_nested_state,
1314 			 struct kvm_nested_state *kvm_state);
1315 	bool (*get_nested_state_pages)(struct kvm_vcpu *vcpu);
1316 	int (*write_log_dirty)(struct kvm_vcpu *vcpu, gpa_t l2_gpa);
1317 
1318 	int (*enable_evmcs)(struct kvm_vcpu *vcpu,
1319 			    uint16_t *vmcs_version);
1320 	uint16_t (*get_evmcs_version)(struct kvm_vcpu *vcpu);
1321 };
1322 
1323 struct kvm_x86_init_ops {
1324 	int (*cpu_has_kvm_support)(void);
1325 	int (*disabled_by_bios)(void);
1326 	int (*check_processor_compatibility)(void);
1327 	int (*hardware_setup)(void);
1328 
1329 	struct kvm_x86_ops *runtime_ops;
1330 };
1331 
1332 struct kvm_arch_async_pf {
1333 	u32 token;
1334 	gfn_t gfn;
1335 	unsigned long cr3;
1336 	bool direct_map;
1337 };
1338 
1339 extern u64 __read_mostly host_efer;
1340 extern bool __read_mostly allow_smaller_maxphyaddr;
1341 extern struct kvm_x86_ops kvm_x86_ops;
1342 
1343 #define __KVM_HAVE_ARCH_VM_ALLOC
1344 static inline struct kvm *kvm_arch_alloc_vm(void)
1345 {
1346 	return __vmalloc(kvm_x86_ops.vm_size, GFP_KERNEL_ACCOUNT | __GFP_ZERO);
1347 }
1348 void kvm_arch_free_vm(struct kvm *kvm);
1349 
1350 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
1351 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
1352 {
1353 	if (kvm_x86_ops.tlb_remote_flush &&
1354 	    !kvm_x86_ops.tlb_remote_flush(kvm))
1355 		return 0;
1356 	else
1357 		return -ENOTSUPP;
1358 }
1359 
1360 int kvm_mmu_module_init(void);
1361 void kvm_mmu_module_exit(void);
1362 
1363 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
1364 int kvm_mmu_create(struct kvm_vcpu *vcpu);
1365 void kvm_mmu_init_vm(struct kvm *kvm);
1366 void kvm_mmu_uninit_vm(struct kvm *kvm);
1367 void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
1368 		u64 dirty_mask, u64 nx_mask, u64 x_mask, u64 p_mask,
1369 		u64 acc_track_mask, u64 me_mask);
1370 
1371 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
1372 void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
1373 				      struct kvm_memory_slot *memslot,
1374 				      int start_level);
1375 void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
1376 				   const struct kvm_memory_slot *memslot);
1377 void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
1378 				   struct kvm_memory_slot *memslot);
1379 void kvm_mmu_slot_largepage_remove_write_access(struct kvm *kvm,
1380 					struct kvm_memory_slot *memslot);
1381 void kvm_mmu_slot_set_dirty(struct kvm *kvm,
1382 			    struct kvm_memory_slot *memslot);
1383 void kvm_mmu_clear_dirty_pt_masked(struct kvm *kvm,
1384 				   struct kvm_memory_slot *slot,
1385 				   gfn_t gfn_offset, unsigned long mask);
1386 void kvm_mmu_zap_all(struct kvm *kvm);
1387 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, u64 gen);
1388 unsigned long kvm_mmu_calculate_default_mmu_pages(struct kvm *kvm);
1389 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned long kvm_nr_mmu_pages);
1390 
1391 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
1392 bool pdptrs_changed(struct kvm_vcpu *vcpu);
1393 
1394 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
1395 			  const void *val, int bytes);
1396 
1397 struct kvm_irq_mask_notifier {
1398 	void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
1399 	int irq;
1400 	struct hlist_node link;
1401 };
1402 
1403 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
1404 				    struct kvm_irq_mask_notifier *kimn);
1405 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
1406 				      struct kvm_irq_mask_notifier *kimn);
1407 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
1408 			     bool mask);
1409 
1410 extern bool tdp_enabled;
1411 
1412 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
1413 
1414 /* control of guest tsc rate supported? */
1415 extern bool kvm_has_tsc_control;
1416 /* maximum supported tsc_khz for guests */
1417 extern u32  kvm_max_guest_tsc_khz;
1418 /* number of bits of the fractional part of the TSC scaling ratio */
1419 extern u8   kvm_tsc_scaling_ratio_frac_bits;
1420 /* maximum allowed value of TSC scaling ratio */
1421 extern u64  kvm_max_tsc_scaling_ratio;
1422 /* 1ull << kvm_tsc_scaling_ratio_frac_bits */
1423 extern u64  kvm_default_tsc_scaling_ratio;
1424 
1425 extern u64 kvm_mce_cap_supported;
1426 
1427 /*
1428  * EMULTYPE_NO_DECODE - Set when re-emulating an instruction (after completing
1429  *			userspace I/O) to indicate that the emulation context
1430  *			should be resued as is, i.e. skip initialization of
1431  *			emulation context, instruction fetch and decode.
1432  *
1433  * EMULTYPE_TRAP_UD - Set when emulating an intercepted #UD from hardware.
1434  *		      Indicates that only select instructions (tagged with
1435  *		      EmulateOnUD) should be emulated (to minimize the emulator
1436  *		      attack surface).  See also EMULTYPE_TRAP_UD_FORCED.
1437  *
1438  * EMULTYPE_SKIP - Set when emulating solely to skip an instruction, i.e. to
1439  *		   decode the instruction length.  For use *only* by
1440  *		   kvm_x86_ops.skip_emulated_instruction() implementations.
1441  *
1442  * EMULTYPE_ALLOW_RETRY_PF - Set when the emulator should resume the guest to
1443  *			     retry native execution under certain conditions,
1444  *			     Can only be set in conjunction with EMULTYPE_PF.
1445  *
1446  * EMULTYPE_TRAP_UD_FORCED - Set when emulating an intercepted #UD that was
1447  *			     triggered by KVM's magic "force emulation" prefix,
1448  *			     which is opt in via module param (off by default).
1449  *			     Bypasses EmulateOnUD restriction despite emulating
1450  *			     due to an intercepted #UD (see EMULTYPE_TRAP_UD).
1451  *			     Used to test the full emulator from userspace.
1452  *
1453  * EMULTYPE_VMWARE_GP - Set when emulating an intercepted #GP for VMware
1454  *			backdoor emulation, which is opt in via module param.
1455  *			VMware backoor emulation handles select instructions
1456  *			and reinjects the #GP for all other cases.
1457  *
1458  * EMULTYPE_PF - Set when emulating MMIO by way of an intercepted #PF, in which
1459  *		 case the CR2/GPA value pass on the stack is valid.
1460  */
1461 #define EMULTYPE_NO_DECODE	    (1 << 0)
1462 #define EMULTYPE_TRAP_UD	    (1 << 1)
1463 #define EMULTYPE_SKIP		    (1 << 2)
1464 #define EMULTYPE_ALLOW_RETRY_PF	    (1 << 3)
1465 #define EMULTYPE_TRAP_UD_FORCED	    (1 << 4)
1466 #define EMULTYPE_VMWARE_GP	    (1 << 5)
1467 #define EMULTYPE_PF		    (1 << 6)
1468 
1469 int kvm_emulate_instruction(struct kvm_vcpu *vcpu, int emulation_type);
1470 int kvm_emulate_instruction_from_buffer(struct kvm_vcpu *vcpu,
1471 					void *insn, int insn_len);
1472 
1473 void kvm_enable_efer_bits(u64);
1474 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
1475 int __kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data, bool host_initiated);
1476 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data);
1477 int kvm_set_msr(struct kvm_vcpu *vcpu, u32 index, u64 data);
1478 int kvm_emulate_rdmsr(struct kvm_vcpu *vcpu);
1479 int kvm_emulate_wrmsr(struct kvm_vcpu *vcpu);
1480 
1481 int kvm_fast_pio(struct kvm_vcpu *vcpu, int size, unsigned short port, int in);
1482 int kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
1483 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
1484 int kvm_vcpu_halt(struct kvm_vcpu *vcpu);
1485 int kvm_emulate_ap_reset_hold(struct kvm_vcpu *vcpu);
1486 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
1487 
1488 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
1489 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
1490 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector);
1491 
1492 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
1493 		    int reason, bool has_error_code, u32 error_code);
1494 
1495 void kvm_free_guest_fpu(struct kvm_vcpu *vcpu);
1496 
1497 void kvm_post_set_cr0(struct kvm_vcpu *vcpu, unsigned long old_cr0, unsigned long cr0);
1498 void kvm_post_set_cr4(struct kvm_vcpu *vcpu, unsigned long old_cr4, unsigned long cr4);
1499 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
1500 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
1501 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
1502 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
1503 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
1504 int kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
1505 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
1506 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
1507 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
1508 int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr);
1509 
1510 int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1511 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1512 
1513 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
1514 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
1515 bool kvm_rdpmc(struct kvm_vcpu *vcpu);
1516 
1517 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1518 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1519 void kvm_queue_exception_p(struct kvm_vcpu *vcpu, unsigned nr, unsigned long payload);
1520 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1521 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1522 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
1523 bool kvm_inject_emulated_page_fault(struct kvm_vcpu *vcpu,
1524 				    struct x86_exception *fault);
1525 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1526 			    gfn_t gfn, void *data, int offset, int len,
1527 			    u32 access);
1528 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
1529 bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr);
1530 
1531 static inline int __kvm_irq_line_state(unsigned long *irq_state,
1532 				       int irq_source_id, int level)
1533 {
1534 	/* Logical OR for level trig interrupt */
1535 	if (level)
1536 		__set_bit(irq_source_id, irq_state);
1537 	else
1538 		__clear_bit(irq_source_id, irq_state);
1539 
1540 	return !!(*irq_state);
1541 }
1542 
1543 #define KVM_MMU_ROOT_CURRENT		BIT(0)
1544 #define KVM_MMU_ROOT_PREVIOUS(i)	BIT(1+i)
1545 #define KVM_MMU_ROOTS_ALL		(~0UL)
1546 
1547 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
1548 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
1549 
1550 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
1551 
1552 void kvm_update_dr7(struct kvm_vcpu *vcpu);
1553 
1554 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
1555 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
1556 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
1557 int kvm_mmu_load(struct kvm_vcpu *vcpu);
1558 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
1559 void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu);
1560 void kvm_mmu_free_roots(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1561 			ulong roots_to_free);
1562 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1563 			   struct x86_exception *exception);
1564 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
1565 			      struct x86_exception *exception);
1566 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
1567 			       struct x86_exception *exception);
1568 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
1569 			       struct x86_exception *exception);
1570 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
1571 				struct x86_exception *exception);
1572 
1573 bool kvm_apicv_activated(struct kvm *kvm);
1574 void kvm_apicv_init(struct kvm *kvm, bool enable);
1575 void kvm_vcpu_update_apicv(struct kvm_vcpu *vcpu);
1576 void kvm_request_apicv_update(struct kvm *kvm, bool activate,
1577 			      unsigned long bit);
1578 
1579 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
1580 
1581 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u64 error_code,
1582 		       void *insn, int insn_len);
1583 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
1584 void kvm_mmu_invalidate_gva(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1585 			    gva_t gva, hpa_t root_hpa);
1586 void kvm_mmu_invpcid_gva(struct kvm_vcpu *vcpu, gva_t gva, unsigned long pcid);
1587 void kvm_mmu_new_pgd(struct kvm_vcpu *vcpu, gpa_t new_pgd, bool skip_tlb_flush,
1588 		     bool skip_mmu_sync);
1589 
1590 void kvm_configure_mmu(bool enable_tdp, int tdp_max_root_level,
1591 		       int tdp_huge_page_level);
1592 
1593 static inline u16 kvm_read_ldt(void)
1594 {
1595 	u16 ldt;
1596 	asm("sldt %0" : "=g"(ldt));
1597 	return ldt;
1598 }
1599 
1600 static inline void kvm_load_ldt(u16 sel)
1601 {
1602 	asm("lldt %0" : : "rm"(sel));
1603 }
1604 
1605 #ifdef CONFIG_X86_64
1606 static inline unsigned long read_msr(unsigned long msr)
1607 {
1608 	u64 value;
1609 
1610 	rdmsrl(msr, value);
1611 	return value;
1612 }
1613 #endif
1614 
1615 static inline u32 get_rdx_init_val(void)
1616 {
1617 	return 0x600; /* P6 family */
1618 }
1619 
1620 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
1621 {
1622 	kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
1623 }
1624 
1625 #define TSS_IOPB_BASE_OFFSET 0x66
1626 #define TSS_BASE_SIZE 0x68
1627 #define TSS_IOPB_SIZE (65536 / 8)
1628 #define TSS_REDIRECTION_SIZE (256 / 8)
1629 #define RMODE_TSS_SIZE							\
1630 	(TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
1631 
1632 enum {
1633 	TASK_SWITCH_CALL = 0,
1634 	TASK_SWITCH_IRET = 1,
1635 	TASK_SWITCH_JMP = 2,
1636 	TASK_SWITCH_GATE = 3,
1637 };
1638 
1639 #define HF_GIF_MASK		(1 << 0)
1640 #define HF_NMI_MASK		(1 << 3)
1641 #define HF_IRET_MASK		(1 << 4)
1642 #define HF_GUEST_MASK		(1 << 5) /* VCPU is in guest-mode */
1643 #define HF_SMM_MASK		(1 << 6)
1644 #define HF_SMM_INSIDE_NMI_MASK	(1 << 7)
1645 
1646 #define __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
1647 #define KVM_ADDRESS_SPACE_NUM 2
1648 
1649 #define kvm_arch_vcpu_memslots_id(vcpu) ((vcpu)->arch.hflags & HF_SMM_MASK ? 1 : 0)
1650 #define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, (role).smm)
1651 
1652 asmlinkage void kvm_spurious_fault(void);
1653 
1654 /*
1655  * Hardware virtualization extension instructions may fault if a
1656  * reboot turns off virtualization while processes are running.
1657  * Usually after catching the fault we just panic; during reboot
1658  * instead the instruction is ignored.
1659  */
1660 #define __kvm_handle_fault_on_reboot(insn)				\
1661 	"666: \n\t"							\
1662 	insn "\n\t"							\
1663 	"jmp	668f \n\t"						\
1664 	"667: \n\t"							\
1665 	"1: \n\t"							\
1666 	".pushsection .discard.instr_begin \n\t"			\
1667 	".long 1b - . \n\t"						\
1668 	".popsection \n\t"						\
1669 	"call	kvm_spurious_fault \n\t"				\
1670 	"1: \n\t"							\
1671 	".pushsection .discard.instr_end \n\t"				\
1672 	".long 1b - . \n\t"						\
1673 	".popsection \n\t"						\
1674 	"668: \n\t"							\
1675 	_ASM_EXTABLE(666b, 667b)
1676 
1677 #define KVM_ARCH_WANT_MMU_NOTIFIER
1678 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end,
1679 			unsigned flags);
1680 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
1681 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
1682 int kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
1683 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1684 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1685 int kvm_cpu_has_extint(struct kvm_vcpu *v);
1686 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1687 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1688 void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event);
1689 void kvm_vcpu_reload_apic_access_page(struct kvm_vcpu *vcpu);
1690 
1691 int kvm_pv_send_ipi(struct kvm *kvm, unsigned long ipi_bitmap_low,
1692 		    unsigned long ipi_bitmap_high, u32 min,
1693 		    unsigned long icr, int op_64_bit);
1694 
1695 void kvm_define_user_return_msr(unsigned index, u32 msr);
1696 int kvm_set_user_return_msr(unsigned index, u64 val, u64 mask);
1697 
1698 u64 kvm_scale_tsc(struct kvm_vcpu *vcpu, u64 tsc);
1699 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc);
1700 
1701 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu);
1702 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1703 
1704 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
1705 void kvm_make_scan_ioapic_request(struct kvm *kvm);
1706 void kvm_make_scan_ioapic_request_mask(struct kvm *kvm,
1707 				       unsigned long *vcpu_bitmap);
1708 
1709 bool kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1710 				     struct kvm_async_pf *work);
1711 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1712 				 struct kvm_async_pf *work);
1713 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1714 			       struct kvm_async_pf *work);
1715 void kvm_arch_async_page_present_queued(struct kvm_vcpu *vcpu);
1716 bool kvm_arch_can_dequeue_async_page_present(struct kvm_vcpu *vcpu);
1717 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1718 
1719 int kvm_skip_emulated_instruction(struct kvm_vcpu *vcpu);
1720 int kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1721 void __kvm_request_immediate_exit(struct kvm_vcpu *vcpu);
1722 
1723 int kvm_is_in_guest(void);
1724 
1725 void __user *__x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa,
1726 				     u32 size);
1727 bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu);
1728 bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu);
1729 
1730 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
1731 			     struct kvm_vcpu **dest_vcpu);
1732 
1733 void kvm_set_msi_irq(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e,
1734 		     struct kvm_lapic_irq *irq);
1735 
1736 static inline bool kvm_irq_is_postable(struct kvm_lapic_irq *irq)
1737 {
1738 	/* We can only post Fixed and LowPrio IRQs */
1739 	return (irq->delivery_mode == APIC_DM_FIXED ||
1740 		irq->delivery_mode == APIC_DM_LOWEST);
1741 }
1742 
1743 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu)
1744 {
1745 	if (kvm_x86_ops.vcpu_blocking)
1746 		kvm_x86_ops.vcpu_blocking(vcpu);
1747 }
1748 
1749 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu)
1750 {
1751 	if (kvm_x86_ops.vcpu_unblocking)
1752 		kvm_x86_ops.vcpu_unblocking(vcpu);
1753 }
1754 
1755 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
1756 
1757 static inline int kvm_cpu_get_apicid(int mps_cpu)
1758 {
1759 #ifdef CONFIG_X86_LOCAL_APIC
1760 	return default_cpu_present_to_apicid(mps_cpu);
1761 #else
1762 	WARN_ON_ONCE(1);
1763 	return BAD_APICID;
1764 #endif
1765 }
1766 
1767 #define put_smstate(type, buf, offset, val)                      \
1768 	*(type *)((buf) + (offset) - 0x7e00) = val
1769 
1770 #define GET_SMSTATE(type, buf, offset)		\
1771 	(*(type *)((buf) + (offset) - 0x7e00))
1772 
1773 int kvm_cpu_dirty_log_size(void);
1774 
1775 #endif /* _ASM_X86_KVM_HOST_H */
1776