xref: /linux/arch/x86/include/asm/kvm_host.h (revision aaa44952bbd1d4db14a4d676bf9595bb5db7e7b0)
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 
117 /* KVM Hugepage definitions for x86 */
118 #define KVM_MAX_HUGEPAGE_LEVEL	PG_LEVEL_1G
119 #define KVM_NR_PAGE_SIZES	(KVM_MAX_HUGEPAGE_LEVEL - PG_LEVEL_4K + 1)
120 #define KVM_HPAGE_GFN_SHIFT(x)	(((x) - 1) * 9)
121 #define KVM_HPAGE_SHIFT(x)	(PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
122 #define KVM_HPAGE_SIZE(x)	(1UL << KVM_HPAGE_SHIFT(x))
123 #define KVM_HPAGE_MASK(x)	(~(KVM_HPAGE_SIZE(x) - 1))
124 #define KVM_PAGES_PER_HPAGE(x)	(KVM_HPAGE_SIZE(x) / PAGE_SIZE)
125 
126 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
127 {
128 	/* KVM_HPAGE_GFN_SHIFT(PG_LEVEL_4K) must be 0. */
129 	return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
130 		(base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
131 }
132 
133 #define KVM_PERMILLE_MMU_PAGES 20
134 #define KVM_MIN_ALLOC_MMU_PAGES 64UL
135 #define KVM_MMU_HASH_SHIFT 12
136 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
137 #define KVM_MIN_FREE_MMU_PAGES 5
138 #define KVM_REFILL_PAGES 25
139 #define KVM_MAX_CPUID_ENTRIES 256
140 #define KVM_NR_FIXED_MTRR_REGION 88
141 #define KVM_NR_VAR_MTRR 8
142 
143 #define ASYNC_PF_PER_VCPU 64
144 
145 enum kvm_reg {
146 	VCPU_REGS_RAX = __VCPU_REGS_RAX,
147 	VCPU_REGS_RCX = __VCPU_REGS_RCX,
148 	VCPU_REGS_RDX = __VCPU_REGS_RDX,
149 	VCPU_REGS_RBX = __VCPU_REGS_RBX,
150 	VCPU_REGS_RSP = __VCPU_REGS_RSP,
151 	VCPU_REGS_RBP = __VCPU_REGS_RBP,
152 	VCPU_REGS_RSI = __VCPU_REGS_RSI,
153 	VCPU_REGS_RDI = __VCPU_REGS_RDI,
154 #ifdef CONFIG_X86_64
155 	VCPU_REGS_R8  = __VCPU_REGS_R8,
156 	VCPU_REGS_R9  = __VCPU_REGS_R9,
157 	VCPU_REGS_R10 = __VCPU_REGS_R10,
158 	VCPU_REGS_R11 = __VCPU_REGS_R11,
159 	VCPU_REGS_R12 = __VCPU_REGS_R12,
160 	VCPU_REGS_R13 = __VCPU_REGS_R13,
161 	VCPU_REGS_R14 = __VCPU_REGS_R14,
162 	VCPU_REGS_R15 = __VCPU_REGS_R15,
163 #endif
164 	VCPU_REGS_RIP,
165 	NR_VCPU_REGS,
166 
167 	VCPU_EXREG_PDPTR = NR_VCPU_REGS,
168 	VCPU_EXREG_CR0,
169 	VCPU_EXREG_CR3,
170 	VCPU_EXREG_CR4,
171 	VCPU_EXREG_RFLAGS,
172 	VCPU_EXREG_SEGMENTS,
173 	VCPU_EXREG_EXIT_INFO_1,
174 	VCPU_EXREG_EXIT_INFO_2,
175 };
176 
177 enum {
178 	VCPU_SREG_ES,
179 	VCPU_SREG_CS,
180 	VCPU_SREG_SS,
181 	VCPU_SREG_DS,
182 	VCPU_SREG_FS,
183 	VCPU_SREG_GS,
184 	VCPU_SREG_TR,
185 	VCPU_SREG_LDTR,
186 };
187 
188 enum exit_fastpath_completion {
189 	EXIT_FASTPATH_NONE,
190 	EXIT_FASTPATH_REENTER_GUEST,
191 	EXIT_FASTPATH_EXIT_HANDLED,
192 };
193 typedef enum exit_fastpath_completion fastpath_t;
194 
195 struct x86_emulate_ctxt;
196 struct x86_exception;
197 enum x86_intercept;
198 enum x86_intercept_stage;
199 
200 #define KVM_NR_DB_REGS	4
201 
202 #define DR6_BD		(1 << 13)
203 #define DR6_BS		(1 << 14)
204 #define DR6_BT		(1 << 15)
205 #define DR6_RTM		(1 << 16)
206 /*
207  * DR6_ACTIVE_LOW combines fixed-1 and active-low bits.
208  * We can regard all the bits in DR6_FIXED_1 as active_low bits;
209  * they will never be 0 for now, but when they are defined
210  * in the future it will require no code change.
211  *
212  * DR6_ACTIVE_LOW is also used as the init/reset value for DR6.
213  */
214 #define DR6_ACTIVE_LOW	0xffff0ff0
215 #define DR6_VOLATILE	0x0001e00f
216 #define DR6_FIXED_1	(DR6_ACTIVE_LOW & ~DR6_VOLATILE)
217 
218 #define DR7_BP_EN_MASK	0x000000ff
219 #define DR7_GE		(1 << 9)
220 #define DR7_GD		(1 << 13)
221 #define DR7_FIXED_1	0x00000400
222 #define DR7_VOLATILE	0xffff2bff
223 
224 #define KVM_GUESTDBG_VALID_MASK \
225 	(KVM_GUESTDBG_ENABLE | \
226 	KVM_GUESTDBG_SINGLESTEP | \
227 	KVM_GUESTDBG_USE_HW_BP | \
228 	KVM_GUESTDBG_USE_SW_BP | \
229 	KVM_GUESTDBG_INJECT_BP | \
230 	KVM_GUESTDBG_INJECT_DB)
231 
232 
233 #define PFERR_PRESENT_BIT 0
234 #define PFERR_WRITE_BIT 1
235 #define PFERR_USER_BIT 2
236 #define PFERR_RSVD_BIT 3
237 #define PFERR_FETCH_BIT 4
238 #define PFERR_PK_BIT 5
239 #define PFERR_SGX_BIT 15
240 #define PFERR_GUEST_FINAL_BIT 32
241 #define PFERR_GUEST_PAGE_BIT 33
242 
243 #define PFERR_PRESENT_MASK (1U << PFERR_PRESENT_BIT)
244 #define PFERR_WRITE_MASK (1U << PFERR_WRITE_BIT)
245 #define PFERR_USER_MASK (1U << PFERR_USER_BIT)
246 #define PFERR_RSVD_MASK (1U << PFERR_RSVD_BIT)
247 #define PFERR_FETCH_MASK (1U << PFERR_FETCH_BIT)
248 #define PFERR_PK_MASK (1U << PFERR_PK_BIT)
249 #define PFERR_SGX_MASK (1U << PFERR_SGX_BIT)
250 #define PFERR_GUEST_FINAL_MASK (1ULL << PFERR_GUEST_FINAL_BIT)
251 #define PFERR_GUEST_PAGE_MASK (1ULL << PFERR_GUEST_PAGE_BIT)
252 
253 #define PFERR_NESTED_GUEST_PAGE (PFERR_GUEST_PAGE_MASK |	\
254 				 PFERR_WRITE_MASK |		\
255 				 PFERR_PRESENT_MASK)
256 
257 /* apic attention bits */
258 #define KVM_APIC_CHECK_VAPIC	0
259 /*
260  * The following bit is set with PV-EOI, unset on EOI.
261  * We detect PV-EOI changes by guest by comparing
262  * this bit with PV-EOI in guest memory.
263  * See the implementation in apic_update_pv_eoi.
264  */
265 #define KVM_APIC_PV_EOI_PENDING	1
266 
267 struct kvm_kernel_irq_routing_entry;
268 
269 /*
270  * the pages used as guest page table on soft mmu are tracked by
271  * kvm_memory_slot.arch.gfn_track which is 16 bits, so the role bits used
272  * by indirect shadow page can not be more than 15 bits.
273  *
274  * Currently, we used 14 bits that are @level, @gpte_is_8_bytes, @quadrant, @access,
275  * @nxe, @cr0_wp, @smep_andnot_wp and @smap_andnot_wp.
276  */
277 union kvm_mmu_page_role {
278 	u32 word;
279 	struct {
280 		unsigned level:4;
281 		unsigned gpte_is_8_bytes:1;
282 		unsigned quadrant:2;
283 		unsigned direct:1;
284 		unsigned access:3;
285 		unsigned invalid:1;
286 		unsigned nxe:1;
287 		unsigned cr0_wp:1;
288 		unsigned smep_andnot_wp:1;
289 		unsigned smap_andnot_wp:1;
290 		unsigned ad_disabled:1;
291 		unsigned guest_mode:1;
292 		unsigned :6;
293 
294 		/*
295 		 * This is left at the top of the word so that
296 		 * kvm_memslots_for_spte_role can extract it with a
297 		 * simple shift.  While there is room, give it a whole
298 		 * byte so it is also faster to load it from memory.
299 		 */
300 		unsigned smm:8;
301 	};
302 };
303 
304 union kvm_mmu_extended_role {
305 /*
306  * This structure complements kvm_mmu_page_role caching everything needed for
307  * MMU configuration. If nothing in both these structures changed, MMU
308  * re-configuration can be skipped. @valid bit is set on first usage so we don't
309  * treat all-zero structure as valid data.
310  */
311 	u32 word;
312 	struct {
313 		unsigned int valid:1;
314 		unsigned int execonly:1;
315 		unsigned int cr0_pg:1;
316 		unsigned int cr4_pae:1;
317 		unsigned int cr4_pse:1;
318 		unsigned int cr4_pke:1;
319 		unsigned int cr4_smap:1;
320 		unsigned int cr4_smep:1;
321 		unsigned int maxphyaddr:6;
322 	};
323 };
324 
325 union kvm_mmu_role {
326 	u64 as_u64;
327 	struct {
328 		union kvm_mmu_page_role base;
329 		union kvm_mmu_extended_role ext;
330 	};
331 };
332 
333 struct kvm_rmap_head {
334 	unsigned long val;
335 };
336 
337 struct kvm_pio_request {
338 	unsigned long linear_rip;
339 	unsigned long count;
340 	int in;
341 	int port;
342 	int size;
343 };
344 
345 #define PT64_ROOT_MAX_LEVEL 5
346 
347 struct rsvd_bits_validate {
348 	u64 rsvd_bits_mask[2][PT64_ROOT_MAX_LEVEL];
349 	u64 bad_mt_xwr;
350 };
351 
352 struct kvm_mmu_root_info {
353 	gpa_t pgd;
354 	hpa_t hpa;
355 };
356 
357 #define KVM_MMU_ROOT_INFO_INVALID \
358 	((struct kvm_mmu_root_info) { .pgd = INVALID_PAGE, .hpa = INVALID_PAGE })
359 
360 #define KVM_MMU_NUM_PREV_ROOTS 3
361 
362 #define KVM_HAVE_MMU_RWLOCK
363 
364 struct kvm_mmu_page;
365 
366 /*
367  * x86 supports 4 paging modes (5-level 64-bit, 4-level 64-bit, 3-level 32-bit,
368  * and 2-level 32-bit).  The kvm_mmu structure abstracts the details of the
369  * current mmu mode.
370  */
371 struct kvm_mmu {
372 	unsigned long (*get_guest_pgd)(struct kvm_vcpu *vcpu);
373 	u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
374 	int (*page_fault)(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u32 err,
375 			  bool prefault);
376 	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
377 				  struct x86_exception *fault);
378 	gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gpa_t gva_or_gpa,
379 			    u32 access, struct x86_exception *exception);
380 	gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
381 			       struct x86_exception *exception);
382 	int (*sync_page)(struct kvm_vcpu *vcpu,
383 			 struct kvm_mmu_page *sp);
384 	void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva, hpa_t root_hpa);
385 	hpa_t root_hpa;
386 	gpa_t root_pgd;
387 	union kvm_mmu_role mmu_role;
388 	u8 root_level;
389 	u8 shadow_root_level;
390 	u8 ept_ad;
391 	bool direct_map;
392 	struct kvm_mmu_root_info prev_roots[KVM_MMU_NUM_PREV_ROOTS];
393 
394 	/*
395 	 * Bitmap; bit set = permission fault
396 	 * Byte index: page fault error code [4:1]
397 	 * Bit index: pte permissions in ACC_* format
398 	 */
399 	u8 permissions[16];
400 
401 	/*
402 	* The pkru_mask indicates if protection key checks are needed.  It
403 	* consists of 16 domains indexed by page fault error code bits [4:1],
404 	* with PFEC.RSVD replaced by ACC_USER_MASK from the page tables.
405 	* Each domain has 2 bits which are ANDed with AD and WD from PKRU.
406 	*/
407 	u32 pkru_mask;
408 
409 	u64 *pae_root;
410 	u64 *lm_root;
411 
412 	/*
413 	 * check zero bits on shadow page table entries, these
414 	 * bits include not only hardware reserved bits but also
415 	 * the bits spte never used.
416 	 */
417 	struct rsvd_bits_validate shadow_zero_check;
418 
419 	struct rsvd_bits_validate guest_rsvd_check;
420 
421 	/* Can have large pages at levels 2..last_nonleaf_level-1. */
422 	u8 last_nonleaf_level;
423 
424 	bool nx;
425 
426 	u64 pdptrs[4]; /* pae */
427 };
428 
429 struct kvm_tlb_range {
430 	u64 start_gfn;
431 	u64 pages;
432 };
433 
434 enum pmc_type {
435 	KVM_PMC_GP = 0,
436 	KVM_PMC_FIXED,
437 };
438 
439 struct kvm_pmc {
440 	enum pmc_type type;
441 	u8 idx;
442 	u64 counter;
443 	u64 eventsel;
444 	struct perf_event *perf_event;
445 	struct kvm_vcpu *vcpu;
446 	/*
447 	 * eventsel value for general purpose counters,
448 	 * ctrl value for fixed counters.
449 	 */
450 	u64 current_config;
451 };
452 
453 struct kvm_pmu {
454 	unsigned nr_arch_gp_counters;
455 	unsigned nr_arch_fixed_counters;
456 	unsigned available_event_types;
457 	u64 fixed_ctr_ctrl;
458 	u64 global_ctrl;
459 	u64 global_status;
460 	u64 global_ovf_ctrl;
461 	u64 counter_bitmask[2];
462 	u64 global_ctrl_mask;
463 	u64 global_ovf_ctrl_mask;
464 	u64 reserved_bits;
465 	u8 version;
466 	struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
467 	struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED];
468 	struct irq_work irq_work;
469 	DECLARE_BITMAP(reprogram_pmi, X86_PMC_IDX_MAX);
470 	DECLARE_BITMAP(all_valid_pmc_idx, X86_PMC_IDX_MAX);
471 	DECLARE_BITMAP(pmc_in_use, X86_PMC_IDX_MAX);
472 
473 	/*
474 	 * The gate to release perf_events not marked in
475 	 * pmc_in_use only once in a vcpu time slice.
476 	 */
477 	bool need_cleanup;
478 
479 	/*
480 	 * The total number of programmed perf_events and it helps to avoid
481 	 * redundant check before cleanup if guest don't use vPMU at all.
482 	 */
483 	u8 event_count;
484 };
485 
486 struct kvm_pmu_ops;
487 
488 enum {
489 	KVM_DEBUGREG_BP_ENABLED = 1,
490 	KVM_DEBUGREG_WONT_EXIT = 2,
491 	KVM_DEBUGREG_RELOAD = 4,
492 };
493 
494 struct kvm_mtrr_range {
495 	u64 base;
496 	u64 mask;
497 	struct list_head node;
498 };
499 
500 struct kvm_mtrr {
501 	struct kvm_mtrr_range var_ranges[KVM_NR_VAR_MTRR];
502 	mtrr_type fixed_ranges[KVM_NR_FIXED_MTRR_REGION];
503 	u64 deftype;
504 
505 	struct list_head head;
506 };
507 
508 /* Hyper-V SynIC timer */
509 struct kvm_vcpu_hv_stimer {
510 	struct hrtimer timer;
511 	int index;
512 	union hv_stimer_config config;
513 	u64 count;
514 	u64 exp_time;
515 	struct hv_message msg;
516 	bool msg_pending;
517 };
518 
519 /* Hyper-V synthetic interrupt controller (SynIC)*/
520 struct kvm_vcpu_hv_synic {
521 	u64 version;
522 	u64 control;
523 	u64 msg_page;
524 	u64 evt_page;
525 	atomic64_t sint[HV_SYNIC_SINT_COUNT];
526 	atomic_t sint_to_gsi[HV_SYNIC_SINT_COUNT];
527 	DECLARE_BITMAP(auto_eoi_bitmap, 256);
528 	DECLARE_BITMAP(vec_bitmap, 256);
529 	bool active;
530 	bool dont_zero_synic_pages;
531 };
532 
533 /* Hyper-V per vcpu emulation context */
534 struct kvm_vcpu_hv {
535 	struct kvm_vcpu *vcpu;
536 	u32 vp_index;
537 	u64 hv_vapic;
538 	s64 runtime_offset;
539 	struct kvm_vcpu_hv_synic synic;
540 	struct kvm_hyperv_exit exit;
541 	struct kvm_vcpu_hv_stimer stimer[HV_SYNIC_STIMER_COUNT];
542 	DECLARE_BITMAP(stimer_pending_bitmap, HV_SYNIC_STIMER_COUNT);
543 	cpumask_t tlb_flush;
544 };
545 
546 /* Xen HVM per vcpu emulation context */
547 struct kvm_vcpu_xen {
548 	u64 hypercall_rip;
549 	u32 current_runstate;
550 	bool vcpu_info_set;
551 	bool vcpu_time_info_set;
552 	bool runstate_set;
553 	struct gfn_to_hva_cache vcpu_info_cache;
554 	struct gfn_to_hva_cache vcpu_time_info_cache;
555 	struct gfn_to_hva_cache runstate_cache;
556 	u64 last_steal;
557 	u64 runstate_entry_time;
558 	u64 runstate_times[4];
559 };
560 
561 struct kvm_vcpu_arch {
562 	/*
563 	 * rip and regs accesses must go through
564 	 * kvm_{register,rip}_{read,write} functions.
565 	 */
566 	unsigned long regs[NR_VCPU_REGS];
567 	u32 regs_avail;
568 	u32 regs_dirty;
569 
570 	unsigned long cr0;
571 	unsigned long cr0_guest_owned_bits;
572 	unsigned long cr2;
573 	unsigned long cr3;
574 	unsigned long cr4;
575 	unsigned long cr4_guest_owned_bits;
576 	unsigned long cr4_guest_rsvd_bits;
577 	unsigned long cr8;
578 	u32 host_pkru;
579 	u32 pkru;
580 	u32 hflags;
581 	u64 efer;
582 	u64 apic_base;
583 	struct kvm_lapic *apic;    /* kernel irqchip context */
584 	bool apicv_active;
585 	bool load_eoi_exitmap_pending;
586 	DECLARE_BITMAP(ioapic_handled_vectors, 256);
587 	unsigned long apic_attention;
588 	int32_t apic_arb_prio;
589 	int mp_state;
590 	u64 ia32_misc_enable_msr;
591 	u64 smbase;
592 	u64 smi_count;
593 	bool tpr_access_reporting;
594 	bool xsaves_enabled;
595 	u64 ia32_xss;
596 	u64 microcode_version;
597 	u64 arch_capabilities;
598 	u64 perf_capabilities;
599 
600 	/*
601 	 * Paging state of the vcpu
602 	 *
603 	 * If the vcpu runs in guest mode with two level paging this still saves
604 	 * the paging mode of the l1 guest. This context is always used to
605 	 * handle faults.
606 	 */
607 	struct kvm_mmu *mmu;
608 
609 	/* Non-nested MMU for L1 */
610 	struct kvm_mmu root_mmu;
611 
612 	/* L1 MMU when running nested */
613 	struct kvm_mmu guest_mmu;
614 
615 	/*
616 	 * Paging state of an L2 guest (used for nested npt)
617 	 *
618 	 * This context will save all necessary information to walk page tables
619 	 * of an L2 guest. This context is only initialized for page table
620 	 * walking and not for faulting since we never handle l2 page faults on
621 	 * the host.
622 	 */
623 	struct kvm_mmu nested_mmu;
624 
625 	/*
626 	 * Pointer to the mmu context currently used for
627 	 * gva_to_gpa translations.
628 	 */
629 	struct kvm_mmu *walk_mmu;
630 
631 	struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
632 	struct kvm_mmu_memory_cache mmu_shadow_page_cache;
633 	struct kvm_mmu_memory_cache mmu_gfn_array_cache;
634 	struct kvm_mmu_memory_cache mmu_page_header_cache;
635 
636 	/*
637 	 * QEMU userspace and the guest each have their own FPU state.
638 	 * In vcpu_run, we switch between the user and guest FPU contexts.
639 	 * While running a VCPU, the VCPU thread will have the guest FPU
640 	 * context.
641 	 *
642 	 * Note that while the PKRU state lives inside the fpu registers,
643 	 * it is switched out separately at VMENTER and VMEXIT time. The
644 	 * "guest_fpu" state here contains the guest FPU context, with the
645 	 * host PRKU bits.
646 	 */
647 	struct fpu *user_fpu;
648 	struct fpu *guest_fpu;
649 
650 	u64 xcr0;
651 	u64 guest_supported_xcr0;
652 
653 	struct kvm_pio_request pio;
654 	void *pio_data;
655 	void *guest_ins_data;
656 
657 	u8 event_exit_inst_len;
658 
659 	struct kvm_queued_exception {
660 		bool pending;
661 		bool injected;
662 		bool has_error_code;
663 		u8 nr;
664 		u32 error_code;
665 		unsigned long payload;
666 		bool has_payload;
667 		u8 nested_apf;
668 	} exception;
669 
670 	struct kvm_queued_interrupt {
671 		bool injected;
672 		bool soft;
673 		u8 nr;
674 	} interrupt;
675 
676 	int halt_request; /* real mode on Intel only */
677 
678 	int cpuid_nent;
679 	struct kvm_cpuid_entry2 *cpuid_entries;
680 
681 	u64 reserved_gpa_bits;
682 	int maxphyaddr;
683 	int max_tdp_level;
684 
685 	/* emulate context */
686 
687 	struct x86_emulate_ctxt *emulate_ctxt;
688 	bool emulate_regs_need_sync_to_vcpu;
689 	bool emulate_regs_need_sync_from_vcpu;
690 	int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
691 
692 	gpa_t time;
693 	struct pvclock_vcpu_time_info hv_clock;
694 	unsigned int hw_tsc_khz;
695 	struct gfn_to_hva_cache pv_time;
696 	bool pv_time_enabled;
697 	/* set guest stopped flag in pvclock flags field */
698 	bool pvclock_set_guest_stopped_request;
699 
700 	struct {
701 		u8 preempted;
702 		u64 msr_val;
703 		u64 last_steal;
704 		struct gfn_to_pfn_cache cache;
705 	} st;
706 
707 	u64 l1_tsc_offset;
708 	u64 tsc_offset;
709 	u64 last_guest_tsc;
710 	u64 last_host_tsc;
711 	u64 tsc_offset_adjustment;
712 	u64 this_tsc_nsec;
713 	u64 this_tsc_write;
714 	u64 this_tsc_generation;
715 	bool tsc_catchup;
716 	bool tsc_always_catchup;
717 	s8 virtual_tsc_shift;
718 	u32 virtual_tsc_mult;
719 	u32 virtual_tsc_khz;
720 	s64 ia32_tsc_adjust_msr;
721 	u64 msr_ia32_power_ctl;
722 	u64 tsc_scaling_ratio;
723 
724 	atomic_t nmi_queued;  /* unprocessed asynchronous NMIs */
725 	unsigned nmi_pending; /* NMI queued after currently running handler */
726 	bool nmi_injected;    /* Trying to inject an NMI this entry */
727 	bool smi_pending;    /* SMI queued after currently running handler */
728 
729 	struct kvm_mtrr mtrr_state;
730 	u64 pat;
731 
732 	unsigned switch_db_regs;
733 	unsigned long db[KVM_NR_DB_REGS];
734 	unsigned long dr6;
735 	unsigned long dr7;
736 	unsigned long eff_db[KVM_NR_DB_REGS];
737 	unsigned long guest_debug_dr7;
738 	u64 msr_platform_info;
739 	u64 msr_misc_features_enables;
740 
741 	u64 mcg_cap;
742 	u64 mcg_status;
743 	u64 mcg_ctl;
744 	u64 mcg_ext_ctl;
745 	u64 *mce_banks;
746 
747 	/* Cache MMIO info */
748 	u64 mmio_gva;
749 	unsigned mmio_access;
750 	gfn_t mmio_gfn;
751 	u64 mmio_gen;
752 
753 	struct kvm_pmu pmu;
754 
755 	/* used for guest single stepping over the given code position */
756 	unsigned long singlestep_rip;
757 
758 	bool hyperv_enabled;
759 	struct kvm_vcpu_hv *hyperv;
760 	struct kvm_vcpu_xen xen;
761 
762 	cpumask_var_t wbinvd_dirty_mask;
763 
764 	unsigned long last_retry_eip;
765 	unsigned long last_retry_addr;
766 
767 	struct {
768 		bool halted;
769 		gfn_t gfns[ASYNC_PF_PER_VCPU];
770 		struct gfn_to_hva_cache data;
771 		u64 msr_en_val; /* MSR_KVM_ASYNC_PF_EN */
772 		u64 msr_int_val; /* MSR_KVM_ASYNC_PF_INT */
773 		u16 vec;
774 		u32 id;
775 		bool send_user_only;
776 		u32 host_apf_flags;
777 		unsigned long nested_apf_token;
778 		bool delivery_as_pf_vmexit;
779 		bool pageready_pending;
780 	} apf;
781 
782 	/* OSVW MSRs (AMD only) */
783 	struct {
784 		u64 length;
785 		u64 status;
786 	} osvw;
787 
788 	struct {
789 		u64 msr_val;
790 		struct gfn_to_hva_cache data;
791 	} pv_eoi;
792 
793 	u64 msr_kvm_poll_control;
794 
795 	/*
796 	 * Indicates the guest is trying to write a gfn that contains one or
797 	 * more of the PTEs used to translate the write itself, i.e. the access
798 	 * is changing its own translation in the guest page tables.  KVM exits
799 	 * to userspace if emulation of the faulting instruction fails and this
800 	 * flag is set, as KVM cannot make forward progress.
801 	 *
802 	 * If emulation fails for a write to guest page tables, KVM unprotects
803 	 * (zaps) the shadow page for the target gfn and resumes the guest to
804 	 * retry the non-emulatable instruction (on hardware).  Unprotecting the
805 	 * gfn doesn't allow forward progress for a self-changing access because
806 	 * doing so also zaps the translation for the gfn, i.e. retrying the
807 	 * instruction will hit a !PRESENT fault, which results in a new shadow
808 	 * page and sends KVM back to square one.
809 	 */
810 	bool write_fault_to_shadow_pgtable;
811 
812 	/* set at EPT violation at this point */
813 	unsigned long exit_qualification;
814 
815 	/* pv related host specific info */
816 	struct {
817 		bool pv_unhalted;
818 	} pv;
819 
820 	int pending_ioapic_eoi;
821 	int pending_external_vector;
822 
823 	/* be preempted when it's in kernel-mode(cpl=0) */
824 	bool preempted_in_kernel;
825 
826 	/* Flush the L1 Data cache for L1TF mitigation on VMENTER */
827 	bool l1tf_flush_l1d;
828 
829 	/* Host CPU on which VM-entry was most recently attempted */
830 	unsigned int last_vmentry_cpu;
831 
832 	/* AMD MSRC001_0015 Hardware Configuration */
833 	u64 msr_hwcr;
834 
835 	/* pv related cpuid info */
836 	struct {
837 		/*
838 		 * value of the eax register in the KVM_CPUID_FEATURES CPUID
839 		 * leaf.
840 		 */
841 		u32 features;
842 
843 		/*
844 		 * indicates whether pv emulation should be disabled if features
845 		 * are not present in the guest's cpuid
846 		 */
847 		bool enforce;
848 	} pv_cpuid;
849 
850 	/* Protected Guests */
851 	bool guest_state_protected;
852 };
853 
854 struct kvm_lpage_info {
855 	int disallow_lpage;
856 };
857 
858 struct kvm_arch_memory_slot {
859 	struct kvm_rmap_head *rmap[KVM_NR_PAGE_SIZES];
860 	struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
861 	unsigned short *gfn_track[KVM_PAGE_TRACK_MAX];
862 };
863 
864 /*
865  * We use as the mode the number of bits allocated in the LDR for the
866  * logical processor ID.  It happens that these are all powers of two.
867  * This makes it is very easy to detect cases where the APICs are
868  * configured for multiple modes; in that case, we cannot use the map and
869  * hence cannot use kvm_irq_delivery_to_apic_fast either.
870  */
871 #define KVM_APIC_MODE_XAPIC_CLUSTER          4
872 #define KVM_APIC_MODE_XAPIC_FLAT             8
873 #define KVM_APIC_MODE_X2APIC                16
874 
875 struct kvm_apic_map {
876 	struct rcu_head rcu;
877 	u8 mode;
878 	u32 max_apic_id;
879 	union {
880 		struct kvm_lapic *xapic_flat_map[8];
881 		struct kvm_lapic *xapic_cluster_map[16][4];
882 	};
883 	struct kvm_lapic *phys_map[];
884 };
885 
886 /* Hyper-V synthetic debugger (SynDbg)*/
887 struct kvm_hv_syndbg {
888 	struct {
889 		u64 control;
890 		u64 status;
891 		u64 send_page;
892 		u64 recv_page;
893 		u64 pending_page;
894 	} control;
895 	u64 options;
896 };
897 
898 /* Current state of Hyper-V TSC page clocksource */
899 enum hv_tsc_page_status {
900 	/* TSC page was not set up or disabled */
901 	HV_TSC_PAGE_UNSET = 0,
902 	/* TSC page MSR was written by the guest, update pending */
903 	HV_TSC_PAGE_GUEST_CHANGED,
904 	/* TSC page MSR was written by KVM userspace, update pending */
905 	HV_TSC_PAGE_HOST_CHANGED,
906 	/* TSC page was properly set up and is currently active  */
907 	HV_TSC_PAGE_SET,
908 	/* TSC page is currently being updated and therefore is inactive */
909 	HV_TSC_PAGE_UPDATING,
910 	/* TSC page was set up with an inaccessible GPA */
911 	HV_TSC_PAGE_BROKEN,
912 };
913 
914 /* Hyper-V emulation context */
915 struct kvm_hv {
916 	struct mutex hv_lock;
917 	u64 hv_guest_os_id;
918 	u64 hv_hypercall;
919 	u64 hv_tsc_page;
920 	enum hv_tsc_page_status hv_tsc_page_status;
921 
922 	/* Hyper-v based guest crash (NT kernel bugcheck) parameters */
923 	u64 hv_crash_param[HV_X64_MSR_CRASH_PARAMS];
924 	u64 hv_crash_ctl;
925 
926 	struct ms_hyperv_tsc_page tsc_ref;
927 
928 	struct idr conn_to_evt;
929 
930 	u64 hv_reenlightenment_control;
931 	u64 hv_tsc_emulation_control;
932 	u64 hv_tsc_emulation_status;
933 
934 	/* How many vCPUs have VP index != vCPU index */
935 	atomic_t num_mismatched_vp_indexes;
936 
937 	struct hv_partition_assist_pg *hv_pa_pg;
938 	struct kvm_hv_syndbg hv_syndbg;
939 };
940 
941 struct msr_bitmap_range {
942 	u32 flags;
943 	u32 nmsrs;
944 	u32 base;
945 	unsigned long *bitmap;
946 };
947 
948 /* Xen emulation context */
949 struct kvm_xen {
950 	bool long_mode;
951 	bool shinfo_set;
952 	u8 upcall_vector;
953 	struct gfn_to_hva_cache shinfo_cache;
954 };
955 
956 enum kvm_irqchip_mode {
957 	KVM_IRQCHIP_NONE,
958 	KVM_IRQCHIP_KERNEL,       /* created with KVM_CREATE_IRQCHIP */
959 	KVM_IRQCHIP_SPLIT,        /* created with KVM_CAP_SPLIT_IRQCHIP */
960 };
961 
962 struct kvm_x86_msr_filter {
963 	u8 count;
964 	bool default_allow:1;
965 	struct msr_bitmap_range ranges[16];
966 };
967 
968 #define APICV_INHIBIT_REASON_DISABLE    0
969 #define APICV_INHIBIT_REASON_HYPERV     1
970 #define APICV_INHIBIT_REASON_NESTED     2
971 #define APICV_INHIBIT_REASON_IRQWIN     3
972 #define APICV_INHIBIT_REASON_PIT_REINJ  4
973 #define APICV_INHIBIT_REASON_X2APIC	5
974 
975 struct kvm_arch {
976 	unsigned long n_used_mmu_pages;
977 	unsigned long n_requested_mmu_pages;
978 	unsigned long n_max_mmu_pages;
979 	unsigned int indirect_shadow_pages;
980 	u8 mmu_valid_gen;
981 	struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
982 	struct list_head active_mmu_pages;
983 	struct list_head zapped_obsolete_pages;
984 	struct list_head lpage_disallowed_mmu_pages;
985 	struct kvm_page_track_notifier_node mmu_sp_tracker;
986 	struct kvm_page_track_notifier_head track_notifier_head;
987 
988 	struct list_head assigned_dev_head;
989 	struct iommu_domain *iommu_domain;
990 	bool iommu_noncoherent;
991 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA
992 	atomic_t noncoherent_dma_count;
993 #define __KVM_HAVE_ARCH_ASSIGNED_DEVICE
994 	atomic_t assigned_device_count;
995 	struct kvm_pic *vpic;
996 	struct kvm_ioapic *vioapic;
997 	struct kvm_pit *vpit;
998 	atomic_t vapics_in_nmi_mode;
999 	struct mutex apic_map_lock;
1000 	struct kvm_apic_map __rcu *apic_map;
1001 	atomic_t apic_map_dirty;
1002 
1003 	bool apic_access_page_done;
1004 	unsigned long apicv_inhibit_reasons;
1005 
1006 	gpa_t wall_clock;
1007 
1008 	bool mwait_in_guest;
1009 	bool hlt_in_guest;
1010 	bool pause_in_guest;
1011 	bool cstate_in_guest;
1012 
1013 	unsigned long irq_sources_bitmap;
1014 	s64 kvmclock_offset;
1015 	raw_spinlock_t tsc_write_lock;
1016 	u64 last_tsc_nsec;
1017 	u64 last_tsc_write;
1018 	u32 last_tsc_khz;
1019 	u64 cur_tsc_nsec;
1020 	u64 cur_tsc_write;
1021 	u64 cur_tsc_offset;
1022 	u64 cur_tsc_generation;
1023 	int nr_vcpus_matched_tsc;
1024 
1025 	spinlock_t pvclock_gtod_sync_lock;
1026 	bool use_master_clock;
1027 	u64 master_kernel_ns;
1028 	u64 master_cycle_now;
1029 	struct delayed_work kvmclock_update_work;
1030 	struct delayed_work kvmclock_sync_work;
1031 
1032 	struct kvm_xen_hvm_config xen_hvm_config;
1033 
1034 	/* reads protected by irq_srcu, writes by irq_lock */
1035 	struct hlist_head mask_notifier_list;
1036 
1037 	struct kvm_hv hyperv;
1038 	struct kvm_xen xen;
1039 
1040 	#ifdef CONFIG_KVM_MMU_AUDIT
1041 	int audit_point;
1042 	#endif
1043 
1044 	bool backwards_tsc_observed;
1045 	bool boot_vcpu_runs_old_kvmclock;
1046 	u32 bsp_vcpu_id;
1047 
1048 	u64 disabled_quirks;
1049 	int cpu_dirty_logging_count;
1050 
1051 	enum kvm_irqchip_mode irqchip_mode;
1052 	u8 nr_reserved_ioapic_pins;
1053 
1054 	bool disabled_lapic_found;
1055 
1056 	bool x2apic_format;
1057 	bool x2apic_broadcast_quirk_disabled;
1058 
1059 	bool guest_can_read_msr_platform_info;
1060 	bool exception_payload_enabled;
1061 
1062 	bool bus_lock_detection_enabled;
1063 
1064 	/* Deflect RDMSR and WRMSR to user space when they trigger a #GP */
1065 	u32 user_space_msr_mask;
1066 	struct kvm_x86_msr_filter __rcu *msr_filter;
1067 
1068 	/* Guest can access the SGX PROVISIONKEY. */
1069 	bool sgx_provisioning_allowed;
1070 
1071 	struct kvm_pmu_event_filter __rcu *pmu_event_filter;
1072 	struct task_struct *nx_lpage_recovery_thread;
1073 
1074 #ifdef CONFIG_X86_64
1075 	/*
1076 	 * Whether the TDP MMU is enabled for this VM. This contains a
1077 	 * snapshot of the TDP MMU module parameter from when the VM was
1078 	 * created and remains unchanged for the life of the VM. If this is
1079 	 * true, TDP MMU handler functions will run for various MMU
1080 	 * operations.
1081 	 */
1082 	bool tdp_mmu_enabled;
1083 
1084 	/*
1085 	 * List of struct kvm_mmu_pages being used as roots.
1086 	 * All struct kvm_mmu_pages in the list should have
1087 	 * tdp_mmu_page set.
1088 	 *
1089 	 * For reads, this list is protected by:
1090 	 *	the MMU lock in read mode + RCU or
1091 	 *	the MMU lock in write mode
1092 	 *
1093 	 * For writes, this list is protected by:
1094 	 *	the MMU lock in read mode + the tdp_mmu_pages_lock or
1095 	 *	the MMU lock in write mode
1096 	 *
1097 	 * Roots will remain in the list until their tdp_mmu_root_count
1098 	 * drops to zero, at which point the thread that decremented the
1099 	 * count to zero should removed the root from the list and clean
1100 	 * it up, freeing the root after an RCU grace period.
1101 	 */
1102 	struct list_head tdp_mmu_roots;
1103 
1104 	/*
1105 	 * List of struct kvmp_mmu_pages not being used as roots.
1106 	 * All struct kvm_mmu_pages in the list should have
1107 	 * tdp_mmu_page set and a tdp_mmu_root_count of 0.
1108 	 */
1109 	struct list_head tdp_mmu_pages;
1110 
1111 	/*
1112 	 * Protects accesses to the following fields when the MMU lock
1113 	 * is held in read mode:
1114 	 *  - tdp_mmu_roots (above)
1115 	 *  - tdp_mmu_pages (above)
1116 	 *  - the link field of struct kvm_mmu_pages used by the TDP MMU
1117 	 *  - lpage_disallowed_mmu_pages
1118 	 *  - the lpage_disallowed_link field of struct kvm_mmu_pages used
1119 	 *    by the TDP MMU
1120 	 * It is acceptable, but not necessary, to acquire this lock when
1121 	 * the thread holds the MMU lock in write mode.
1122 	 */
1123 	spinlock_t tdp_mmu_pages_lock;
1124 #endif /* CONFIG_X86_64 */
1125 };
1126 
1127 struct kvm_vm_stat {
1128 	ulong mmu_shadow_zapped;
1129 	ulong mmu_pte_write;
1130 	ulong mmu_pde_zapped;
1131 	ulong mmu_flooded;
1132 	ulong mmu_recycled;
1133 	ulong mmu_cache_miss;
1134 	ulong mmu_unsync;
1135 	ulong remote_tlb_flush;
1136 	ulong lpages;
1137 	ulong nx_lpage_splits;
1138 	ulong max_mmu_page_hash_collisions;
1139 };
1140 
1141 struct kvm_vcpu_stat {
1142 	u64 pf_fixed;
1143 	u64 pf_guest;
1144 	u64 tlb_flush;
1145 	u64 invlpg;
1146 
1147 	u64 exits;
1148 	u64 io_exits;
1149 	u64 mmio_exits;
1150 	u64 signal_exits;
1151 	u64 irq_window_exits;
1152 	u64 nmi_window_exits;
1153 	u64 l1d_flush;
1154 	u64 halt_exits;
1155 	u64 halt_successful_poll;
1156 	u64 halt_attempted_poll;
1157 	u64 halt_poll_invalid;
1158 	u64 halt_wakeup;
1159 	u64 request_irq_exits;
1160 	u64 irq_exits;
1161 	u64 host_state_reload;
1162 	u64 fpu_reload;
1163 	u64 insn_emulation;
1164 	u64 insn_emulation_fail;
1165 	u64 hypercalls;
1166 	u64 irq_injections;
1167 	u64 nmi_injections;
1168 	u64 req_event;
1169 	u64 halt_poll_success_ns;
1170 	u64 halt_poll_fail_ns;
1171 	u64 nested_run;
1172 	u64 directed_yield_attempted;
1173 	u64 directed_yield_successful;
1174 };
1175 
1176 struct x86_instruction_info;
1177 
1178 struct msr_data {
1179 	bool host_initiated;
1180 	u32 index;
1181 	u64 data;
1182 };
1183 
1184 struct kvm_lapic_irq {
1185 	u32 vector;
1186 	u16 delivery_mode;
1187 	u16 dest_mode;
1188 	bool level;
1189 	u16 trig_mode;
1190 	u32 shorthand;
1191 	u32 dest_id;
1192 	bool msi_redir_hint;
1193 };
1194 
1195 static inline u16 kvm_lapic_irq_dest_mode(bool dest_mode_logical)
1196 {
1197 	return dest_mode_logical ? APIC_DEST_LOGICAL : APIC_DEST_PHYSICAL;
1198 }
1199 
1200 struct kvm_x86_ops {
1201 	int (*hardware_enable)(void);
1202 	void (*hardware_disable)(void);
1203 	void (*hardware_unsetup)(void);
1204 	bool (*cpu_has_accelerated_tpr)(void);
1205 	bool (*has_emulated_msr)(struct kvm *kvm, u32 index);
1206 	void (*vcpu_after_set_cpuid)(struct kvm_vcpu *vcpu);
1207 
1208 	unsigned int vm_size;
1209 	int (*vm_init)(struct kvm *kvm);
1210 	void (*vm_destroy)(struct kvm *kvm);
1211 
1212 	/* Create, but do not attach this VCPU */
1213 	int (*vcpu_create)(struct kvm_vcpu *vcpu);
1214 	void (*vcpu_free)(struct kvm_vcpu *vcpu);
1215 	void (*vcpu_reset)(struct kvm_vcpu *vcpu, bool init_event);
1216 
1217 	void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
1218 	void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
1219 	void (*vcpu_put)(struct kvm_vcpu *vcpu);
1220 
1221 	void (*update_exception_bitmap)(struct kvm_vcpu *vcpu);
1222 	int (*get_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1223 	int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1224 	u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
1225 	void (*get_segment)(struct kvm_vcpu *vcpu,
1226 			    struct kvm_segment *var, int seg);
1227 	int (*get_cpl)(struct kvm_vcpu *vcpu);
1228 	void (*set_segment)(struct kvm_vcpu *vcpu,
1229 			    struct kvm_segment *var, int seg);
1230 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
1231 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
1232 	bool (*is_valid_cr4)(struct kvm_vcpu *vcpu, unsigned long cr0);
1233 	void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
1234 	int (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
1235 	void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1236 	void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1237 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1238 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1239 	void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu);
1240 	void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
1241 	void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
1242 	unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
1243 	void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
1244 
1245 	void (*tlb_flush_all)(struct kvm_vcpu *vcpu);
1246 	void (*tlb_flush_current)(struct kvm_vcpu *vcpu);
1247 	int  (*tlb_remote_flush)(struct kvm *kvm);
1248 	int  (*tlb_remote_flush_with_range)(struct kvm *kvm,
1249 			struct kvm_tlb_range *range);
1250 
1251 	/*
1252 	 * Flush any TLB entries associated with the given GVA.
1253 	 * Does not need to flush GPA->HPA mappings.
1254 	 * Can potentially get non-canonical addresses through INVLPGs, which
1255 	 * the implementation may choose to ignore if appropriate.
1256 	 */
1257 	void (*tlb_flush_gva)(struct kvm_vcpu *vcpu, gva_t addr);
1258 
1259 	/*
1260 	 * Flush any TLB entries created by the guest.  Like tlb_flush_gva(),
1261 	 * does not need to flush GPA->HPA mappings.
1262 	 */
1263 	void (*tlb_flush_guest)(struct kvm_vcpu *vcpu);
1264 
1265 	enum exit_fastpath_completion (*run)(struct kvm_vcpu *vcpu);
1266 	int (*handle_exit)(struct kvm_vcpu *vcpu,
1267 		enum exit_fastpath_completion exit_fastpath);
1268 	int (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
1269 	void (*update_emulated_instruction)(struct kvm_vcpu *vcpu);
1270 	void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
1271 	u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu);
1272 	void (*patch_hypercall)(struct kvm_vcpu *vcpu,
1273 				unsigned char *hypercall_addr);
1274 	void (*set_irq)(struct kvm_vcpu *vcpu);
1275 	void (*set_nmi)(struct kvm_vcpu *vcpu);
1276 	void (*queue_exception)(struct kvm_vcpu *vcpu);
1277 	void (*cancel_injection)(struct kvm_vcpu *vcpu);
1278 	int (*interrupt_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1279 	int (*nmi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1280 	bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
1281 	void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
1282 	void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
1283 	void (*enable_irq_window)(struct kvm_vcpu *vcpu);
1284 	void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
1285 	bool (*check_apicv_inhibit_reasons)(ulong bit);
1286 	void (*pre_update_apicv_exec_ctrl)(struct kvm *kvm, bool activate);
1287 	void (*refresh_apicv_exec_ctrl)(struct kvm_vcpu *vcpu);
1288 	void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
1289 	void (*hwapic_isr_update)(struct kvm_vcpu *vcpu, int isr);
1290 	bool (*guest_apic_has_interrupt)(struct kvm_vcpu *vcpu);
1291 	void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
1292 	void (*set_virtual_apic_mode)(struct kvm_vcpu *vcpu);
1293 	void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu);
1294 	int (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
1295 	int (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
1296 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
1297 	int (*set_identity_map_addr)(struct kvm *kvm, u64 ident_addr);
1298 	u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
1299 
1300 	void (*load_mmu_pgd)(struct kvm_vcpu *vcpu, hpa_t root_hpa,
1301 			     int root_level);
1302 
1303 	bool (*has_wbinvd_exit)(void);
1304 
1305 	/* Returns actual tsc_offset set in active VMCS */
1306 	u64 (*write_l1_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
1307 
1308 	/*
1309 	 * Retrieve somewhat arbitrary exit information.  Intended to be used
1310 	 * only from within tracepoints to avoid VMREADs when tracing is off.
1311 	 */
1312 	void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2,
1313 			      u32 *exit_int_info, u32 *exit_int_info_err_code);
1314 
1315 	int (*check_intercept)(struct kvm_vcpu *vcpu,
1316 			       struct x86_instruction_info *info,
1317 			       enum x86_intercept_stage stage,
1318 			       struct x86_exception *exception);
1319 	void (*handle_exit_irqoff)(struct kvm_vcpu *vcpu);
1320 
1321 	void (*request_immediate_exit)(struct kvm_vcpu *vcpu);
1322 
1323 	void (*sched_in)(struct kvm_vcpu *kvm, int cpu);
1324 
1325 	/*
1326 	 * Size of the CPU's dirty log buffer, i.e. VMX's PML buffer.  A zero
1327 	 * value indicates CPU dirty logging is unsupported or disabled.
1328 	 */
1329 	int cpu_dirty_log_size;
1330 	void (*update_cpu_dirty_logging)(struct kvm_vcpu *vcpu);
1331 
1332 	/* pmu operations of sub-arch */
1333 	const struct kvm_pmu_ops *pmu_ops;
1334 	const struct kvm_x86_nested_ops *nested_ops;
1335 
1336 	/*
1337 	 * Architecture specific hooks for vCPU blocking due to
1338 	 * HLT instruction.
1339 	 * Returns for .pre_block():
1340 	 *    - 0 means continue to block the vCPU.
1341 	 *    - 1 means we cannot block the vCPU since some event
1342 	 *        happens during this period, such as, 'ON' bit in
1343 	 *        posted-interrupts descriptor is set.
1344 	 */
1345 	int (*pre_block)(struct kvm_vcpu *vcpu);
1346 	void (*post_block)(struct kvm_vcpu *vcpu);
1347 
1348 	void (*vcpu_blocking)(struct kvm_vcpu *vcpu);
1349 	void (*vcpu_unblocking)(struct kvm_vcpu *vcpu);
1350 
1351 	int (*update_pi_irte)(struct kvm *kvm, unsigned int host_irq,
1352 			      uint32_t guest_irq, bool set);
1353 	void (*apicv_post_state_restore)(struct kvm_vcpu *vcpu);
1354 	bool (*dy_apicv_has_pending_interrupt)(struct kvm_vcpu *vcpu);
1355 
1356 	int (*set_hv_timer)(struct kvm_vcpu *vcpu, u64 guest_deadline_tsc,
1357 			    bool *expired);
1358 	void (*cancel_hv_timer)(struct kvm_vcpu *vcpu);
1359 
1360 	void (*setup_mce)(struct kvm_vcpu *vcpu);
1361 
1362 	int (*smi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1363 	int (*pre_enter_smm)(struct kvm_vcpu *vcpu, char *smstate);
1364 	int (*pre_leave_smm)(struct kvm_vcpu *vcpu, const char *smstate);
1365 	void (*enable_smi_window)(struct kvm_vcpu *vcpu);
1366 
1367 	int (*mem_enc_op)(struct kvm *kvm, void __user *argp);
1368 	int (*mem_enc_reg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1369 	int (*mem_enc_unreg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1370 	int (*vm_copy_enc_context_from)(struct kvm *kvm, unsigned int source_fd);
1371 
1372 	int (*get_msr_feature)(struct kvm_msr_entry *entry);
1373 
1374 	bool (*can_emulate_instruction)(struct kvm_vcpu *vcpu, void *insn, int insn_len);
1375 
1376 	bool (*apic_init_signal_blocked)(struct kvm_vcpu *vcpu);
1377 	int (*enable_direct_tlbflush)(struct kvm_vcpu *vcpu);
1378 
1379 	void (*migrate_timers)(struct kvm_vcpu *vcpu);
1380 	void (*msr_filter_changed)(struct kvm_vcpu *vcpu);
1381 	int (*complete_emulated_msr)(struct kvm_vcpu *vcpu, int err);
1382 
1383 	void (*vcpu_deliver_sipi_vector)(struct kvm_vcpu *vcpu, u8 vector);
1384 };
1385 
1386 struct kvm_x86_nested_ops {
1387 	int (*check_events)(struct kvm_vcpu *vcpu);
1388 	bool (*hv_timer_pending)(struct kvm_vcpu *vcpu);
1389 	void (*triple_fault)(struct kvm_vcpu *vcpu);
1390 	int (*get_state)(struct kvm_vcpu *vcpu,
1391 			 struct kvm_nested_state __user *user_kvm_nested_state,
1392 			 unsigned user_data_size);
1393 	int (*set_state)(struct kvm_vcpu *vcpu,
1394 			 struct kvm_nested_state __user *user_kvm_nested_state,
1395 			 struct kvm_nested_state *kvm_state);
1396 	bool (*get_nested_state_pages)(struct kvm_vcpu *vcpu);
1397 	int (*write_log_dirty)(struct kvm_vcpu *vcpu, gpa_t l2_gpa);
1398 
1399 	int (*enable_evmcs)(struct kvm_vcpu *vcpu,
1400 			    uint16_t *vmcs_version);
1401 	uint16_t (*get_evmcs_version)(struct kvm_vcpu *vcpu);
1402 };
1403 
1404 struct kvm_x86_init_ops {
1405 	int (*cpu_has_kvm_support)(void);
1406 	int (*disabled_by_bios)(void);
1407 	int (*check_processor_compatibility)(void);
1408 	int (*hardware_setup)(void);
1409 
1410 	struct kvm_x86_ops *runtime_ops;
1411 };
1412 
1413 struct kvm_arch_async_pf {
1414 	u32 token;
1415 	gfn_t gfn;
1416 	unsigned long cr3;
1417 	bool direct_map;
1418 };
1419 
1420 extern u64 __read_mostly host_efer;
1421 extern bool __read_mostly allow_smaller_maxphyaddr;
1422 extern struct kvm_x86_ops kvm_x86_ops;
1423 
1424 #define KVM_X86_OP(func) \
1425 	DECLARE_STATIC_CALL(kvm_x86_##func, *(((struct kvm_x86_ops *)0)->func));
1426 #define KVM_X86_OP_NULL KVM_X86_OP
1427 #include <asm/kvm-x86-ops.h>
1428 
1429 static inline void kvm_ops_static_call_update(void)
1430 {
1431 #define KVM_X86_OP(func) \
1432 	static_call_update(kvm_x86_##func, kvm_x86_ops.func);
1433 #define KVM_X86_OP_NULL KVM_X86_OP
1434 #include <asm/kvm-x86-ops.h>
1435 }
1436 
1437 #define __KVM_HAVE_ARCH_VM_ALLOC
1438 static inline struct kvm *kvm_arch_alloc_vm(void)
1439 {
1440 	return __vmalloc(kvm_x86_ops.vm_size, GFP_KERNEL_ACCOUNT | __GFP_ZERO);
1441 }
1442 void kvm_arch_free_vm(struct kvm *kvm);
1443 
1444 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
1445 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
1446 {
1447 	if (kvm_x86_ops.tlb_remote_flush &&
1448 	    !static_call(kvm_x86_tlb_remote_flush)(kvm))
1449 		return 0;
1450 	else
1451 		return -ENOTSUPP;
1452 }
1453 
1454 int kvm_mmu_module_init(void);
1455 void kvm_mmu_module_exit(void);
1456 
1457 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
1458 int kvm_mmu_create(struct kvm_vcpu *vcpu);
1459 void kvm_mmu_init_vm(struct kvm *kvm);
1460 void kvm_mmu_uninit_vm(struct kvm *kvm);
1461 
1462 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
1463 void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
1464 				      struct kvm_memory_slot *memslot,
1465 				      int start_level);
1466 void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
1467 				   const struct kvm_memory_slot *memslot);
1468 void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
1469 				   struct kvm_memory_slot *memslot);
1470 void kvm_mmu_zap_all(struct kvm *kvm);
1471 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, u64 gen);
1472 unsigned long kvm_mmu_calculate_default_mmu_pages(struct kvm *kvm);
1473 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned long kvm_nr_mmu_pages);
1474 
1475 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
1476 bool pdptrs_changed(struct kvm_vcpu *vcpu);
1477 
1478 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
1479 			  const void *val, int bytes);
1480 
1481 struct kvm_irq_mask_notifier {
1482 	void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
1483 	int irq;
1484 	struct hlist_node link;
1485 };
1486 
1487 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
1488 				    struct kvm_irq_mask_notifier *kimn);
1489 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
1490 				      struct kvm_irq_mask_notifier *kimn);
1491 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
1492 			     bool mask);
1493 
1494 extern bool tdp_enabled;
1495 
1496 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
1497 
1498 /* control of guest tsc rate supported? */
1499 extern bool kvm_has_tsc_control;
1500 /* maximum supported tsc_khz for guests */
1501 extern u32  kvm_max_guest_tsc_khz;
1502 /* number of bits of the fractional part of the TSC scaling ratio */
1503 extern u8   kvm_tsc_scaling_ratio_frac_bits;
1504 /* maximum allowed value of TSC scaling ratio */
1505 extern u64  kvm_max_tsc_scaling_ratio;
1506 /* 1ull << kvm_tsc_scaling_ratio_frac_bits */
1507 extern u64  kvm_default_tsc_scaling_ratio;
1508 /* bus lock detection supported? */
1509 extern bool kvm_has_bus_lock_exit;
1510 
1511 extern u64 kvm_mce_cap_supported;
1512 
1513 /*
1514  * EMULTYPE_NO_DECODE - Set when re-emulating an instruction (after completing
1515  *			userspace I/O) to indicate that the emulation context
1516  *			should be reused as is, i.e. skip initialization of
1517  *			emulation context, instruction fetch and decode.
1518  *
1519  * EMULTYPE_TRAP_UD - Set when emulating an intercepted #UD from hardware.
1520  *		      Indicates that only select instructions (tagged with
1521  *		      EmulateOnUD) should be emulated (to minimize the emulator
1522  *		      attack surface).  See also EMULTYPE_TRAP_UD_FORCED.
1523  *
1524  * EMULTYPE_SKIP - Set when emulating solely to skip an instruction, i.e. to
1525  *		   decode the instruction length.  For use *only* by
1526  *		   kvm_x86_ops.skip_emulated_instruction() implementations.
1527  *
1528  * EMULTYPE_ALLOW_RETRY_PF - Set when the emulator should resume the guest to
1529  *			     retry native execution under certain conditions,
1530  *			     Can only be set in conjunction with EMULTYPE_PF.
1531  *
1532  * EMULTYPE_TRAP_UD_FORCED - Set when emulating an intercepted #UD that was
1533  *			     triggered by KVM's magic "force emulation" prefix,
1534  *			     which is opt in via module param (off by default).
1535  *			     Bypasses EmulateOnUD restriction despite emulating
1536  *			     due to an intercepted #UD (see EMULTYPE_TRAP_UD).
1537  *			     Used to test the full emulator from userspace.
1538  *
1539  * EMULTYPE_VMWARE_GP - Set when emulating an intercepted #GP for VMware
1540  *			backdoor emulation, which is opt in via module param.
1541  *			VMware backdoor emulation handles select instructions
1542  *			and reinjects the #GP for all other cases.
1543  *
1544  * EMULTYPE_PF - Set when emulating MMIO by way of an intercepted #PF, in which
1545  *		 case the CR2/GPA value pass on the stack is valid.
1546  */
1547 #define EMULTYPE_NO_DECODE	    (1 << 0)
1548 #define EMULTYPE_TRAP_UD	    (1 << 1)
1549 #define EMULTYPE_SKIP		    (1 << 2)
1550 #define EMULTYPE_ALLOW_RETRY_PF	    (1 << 3)
1551 #define EMULTYPE_TRAP_UD_FORCED	    (1 << 4)
1552 #define EMULTYPE_VMWARE_GP	    (1 << 5)
1553 #define EMULTYPE_PF		    (1 << 6)
1554 
1555 int kvm_emulate_instruction(struct kvm_vcpu *vcpu, int emulation_type);
1556 int kvm_emulate_instruction_from_buffer(struct kvm_vcpu *vcpu,
1557 					void *insn, int insn_len);
1558 
1559 void kvm_enable_efer_bits(u64);
1560 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
1561 int __kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data, bool host_initiated);
1562 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data);
1563 int kvm_set_msr(struct kvm_vcpu *vcpu, u32 index, u64 data);
1564 int kvm_emulate_rdmsr(struct kvm_vcpu *vcpu);
1565 int kvm_emulate_wrmsr(struct kvm_vcpu *vcpu);
1566 int kvm_emulate_as_nop(struct kvm_vcpu *vcpu);
1567 int kvm_emulate_invd(struct kvm_vcpu *vcpu);
1568 int kvm_emulate_mwait(struct kvm_vcpu *vcpu);
1569 int kvm_handle_invalid_op(struct kvm_vcpu *vcpu);
1570 int kvm_emulate_monitor(struct kvm_vcpu *vcpu);
1571 
1572 int kvm_fast_pio(struct kvm_vcpu *vcpu, int size, unsigned short port, int in);
1573 int kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
1574 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
1575 int kvm_vcpu_halt(struct kvm_vcpu *vcpu);
1576 int kvm_emulate_ap_reset_hold(struct kvm_vcpu *vcpu);
1577 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
1578 
1579 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
1580 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
1581 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector);
1582 
1583 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
1584 		    int reason, bool has_error_code, u32 error_code);
1585 
1586 void kvm_free_guest_fpu(struct kvm_vcpu *vcpu);
1587 
1588 void kvm_post_set_cr0(struct kvm_vcpu *vcpu, unsigned long old_cr0, unsigned long cr0);
1589 void kvm_post_set_cr4(struct kvm_vcpu *vcpu, unsigned long old_cr4, unsigned long cr4);
1590 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
1591 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
1592 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
1593 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
1594 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
1595 void kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
1596 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
1597 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
1598 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
1599 int kvm_emulate_xsetbv(struct kvm_vcpu *vcpu);
1600 
1601 int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1602 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1603 
1604 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
1605 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
1606 int kvm_emulate_rdpmc(struct kvm_vcpu *vcpu);
1607 
1608 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1609 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1610 void kvm_queue_exception_p(struct kvm_vcpu *vcpu, unsigned nr, unsigned long payload);
1611 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1612 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1613 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
1614 bool kvm_inject_emulated_page_fault(struct kvm_vcpu *vcpu,
1615 				    struct x86_exception *fault);
1616 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1617 			    gfn_t gfn, void *data, int offset, int len,
1618 			    u32 access);
1619 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
1620 bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr);
1621 
1622 static inline int __kvm_irq_line_state(unsigned long *irq_state,
1623 				       int irq_source_id, int level)
1624 {
1625 	/* Logical OR for level trig interrupt */
1626 	if (level)
1627 		__set_bit(irq_source_id, irq_state);
1628 	else
1629 		__clear_bit(irq_source_id, irq_state);
1630 
1631 	return !!(*irq_state);
1632 }
1633 
1634 #define KVM_MMU_ROOT_CURRENT		BIT(0)
1635 #define KVM_MMU_ROOT_PREVIOUS(i)	BIT(1+i)
1636 #define KVM_MMU_ROOTS_ALL		(~0UL)
1637 
1638 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
1639 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
1640 
1641 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
1642 
1643 void kvm_update_dr7(struct kvm_vcpu *vcpu);
1644 
1645 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
1646 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
1647 void kvm_mmu_free_roots(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1648 			ulong roots_to_free);
1649 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1650 			   struct x86_exception *exception);
1651 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
1652 			      struct x86_exception *exception);
1653 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
1654 			       struct x86_exception *exception);
1655 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
1656 			       struct x86_exception *exception);
1657 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
1658 				struct x86_exception *exception);
1659 
1660 bool kvm_apicv_activated(struct kvm *kvm);
1661 void kvm_apicv_init(struct kvm *kvm, bool enable);
1662 void kvm_vcpu_update_apicv(struct kvm_vcpu *vcpu);
1663 void kvm_request_apicv_update(struct kvm *kvm, bool activate,
1664 			      unsigned long bit);
1665 
1666 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
1667 
1668 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u64 error_code,
1669 		       void *insn, int insn_len);
1670 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
1671 void kvm_mmu_invalidate_gva(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1672 			    gva_t gva, hpa_t root_hpa);
1673 void kvm_mmu_invpcid_gva(struct kvm_vcpu *vcpu, gva_t gva, unsigned long pcid);
1674 void kvm_mmu_new_pgd(struct kvm_vcpu *vcpu, gpa_t new_pgd, bool skip_tlb_flush,
1675 		     bool skip_mmu_sync);
1676 
1677 void kvm_configure_mmu(bool enable_tdp, int tdp_max_root_level,
1678 		       int tdp_huge_page_level);
1679 
1680 static inline u16 kvm_read_ldt(void)
1681 {
1682 	u16 ldt;
1683 	asm("sldt %0" : "=g"(ldt));
1684 	return ldt;
1685 }
1686 
1687 static inline void kvm_load_ldt(u16 sel)
1688 {
1689 	asm("lldt %0" : : "rm"(sel));
1690 }
1691 
1692 #ifdef CONFIG_X86_64
1693 static inline unsigned long read_msr(unsigned long msr)
1694 {
1695 	u64 value;
1696 
1697 	rdmsrl(msr, value);
1698 	return value;
1699 }
1700 #endif
1701 
1702 static inline u32 get_rdx_init_val(void)
1703 {
1704 	return 0x600; /* P6 family */
1705 }
1706 
1707 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
1708 {
1709 	kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
1710 }
1711 
1712 #define TSS_IOPB_BASE_OFFSET 0x66
1713 #define TSS_BASE_SIZE 0x68
1714 #define TSS_IOPB_SIZE (65536 / 8)
1715 #define TSS_REDIRECTION_SIZE (256 / 8)
1716 #define RMODE_TSS_SIZE							\
1717 	(TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
1718 
1719 enum {
1720 	TASK_SWITCH_CALL = 0,
1721 	TASK_SWITCH_IRET = 1,
1722 	TASK_SWITCH_JMP = 2,
1723 	TASK_SWITCH_GATE = 3,
1724 };
1725 
1726 #define HF_GIF_MASK		(1 << 0)
1727 #define HF_NMI_MASK		(1 << 3)
1728 #define HF_IRET_MASK		(1 << 4)
1729 #define HF_GUEST_MASK		(1 << 5) /* VCPU is in guest-mode */
1730 #define HF_SMM_MASK		(1 << 6)
1731 #define HF_SMM_INSIDE_NMI_MASK	(1 << 7)
1732 
1733 #define __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
1734 #define KVM_ADDRESS_SPACE_NUM 2
1735 
1736 #define kvm_arch_vcpu_memslots_id(vcpu) ((vcpu)->arch.hflags & HF_SMM_MASK ? 1 : 0)
1737 #define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, (role).smm)
1738 
1739 asmlinkage void kvm_spurious_fault(void);
1740 
1741 /*
1742  * Hardware virtualization extension instructions may fault if a
1743  * reboot turns off virtualization while processes are running.
1744  * Usually after catching the fault we just panic; during reboot
1745  * instead the instruction is ignored.
1746  */
1747 #define __kvm_handle_fault_on_reboot(insn)				\
1748 	"666: \n\t"							\
1749 	insn "\n\t"							\
1750 	"jmp	668f \n\t"						\
1751 	"667: \n\t"							\
1752 	"1: \n\t"							\
1753 	".pushsection .discard.instr_begin \n\t"			\
1754 	".long 1b - . \n\t"						\
1755 	".popsection \n\t"						\
1756 	"call	kvm_spurious_fault \n\t"				\
1757 	"1: \n\t"							\
1758 	".pushsection .discard.instr_end \n\t"				\
1759 	".long 1b - . \n\t"						\
1760 	".popsection \n\t"						\
1761 	"668: \n\t"							\
1762 	_ASM_EXTABLE(666b, 667b)
1763 
1764 #define KVM_ARCH_WANT_MMU_NOTIFIER
1765 
1766 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1767 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1768 int kvm_cpu_has_extint(struct kvm_vcpu *v);
1769 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1770 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1771 void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event);
1772 void kvm_vcpu_reload_apic_access_page(struct kvm_vcpu *vcpu);
1773 
1774 int kvm_pv_send_ipi(struct kvm *kvm, unsigned long ipi_bitmap_low,
1775 		    unsigned long ipi_bitmap_high, u32 min,
1776 		    unsigned long icr, int op_64_bit);
1777 
1778 void kvm_define_user_return_msr(unsigned index, u32 msr);
1779 int kvm_set_user_return_msr(unsigned index, u64 val, u64 mask);
1780 
1781 u64 kvm_scale_tsc(struct kvm_vcpu *vcpu, u64 tsc);
1782 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc);
1783 
1784 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu);
1785 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1786 
1787 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
1788 void kvm_make_scan_ioapic_request(struct kvm *kvm);
1789 void kvm_make_scan_ioapic_request_mask(struct kvm *kvm,
1790 				       unsigned long *vcpu_bitmap);
1791 
1792 bool kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1793 				     struct kvm_async_pf *work);
1794 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1795 				 struct kvm_async_pf *work);
1796 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1797 			       struct kvm_async_pf *work);
1798 void kvm_arch_async_page_present_queued(struct kvm_vcpu *vcpu);
1799 bool kvm_arch_can_dequeue_async_page_present(struct kvm_vcpu *vcpu);
1800 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1801 
1802 int kvm_skip_emulated_instruction(struct kvm_vcpu *vcpu);
1803 int kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1804 void __kvm_request_immediate_exit(struct kvm_vcpu *vcpu);
1805 
1806 int kvm_is_in_guest(void);
1807 
1808 void __user *__x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa,
1809 				     u32 size);
1810 bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu);
1811 bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu);
1812 
1813 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
1814 			     struct kvm_vcpu **dest_vcpu);
1815 
1816 void kvm_set_msi_irq(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e,
1817 		     struct kvm_lapic_irq *irq);
1818 
1819 static inline bool kvm_irq_is_postable(struct kvm_lapic_irq *irq)
1820 {
1821 	/* We can only post Fixed and LowPrio IRQs */
1822 	return (irq->delivery_mode == APIC_DM_FIXED ||
1823 		irq->delivery_mode == APIC_DM_LOWEST);
1824 }
1825 
1826 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu)
1827 {
1828 	static_call_cond(kvm_x86_vcpu_blocking)(vcpu);
1829 }
1830 
1831 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu)
1832 {
1833 	static_call_cond(kvm_x86_vcpu_unblocking)(vcpu);
1834 }
1835 
1836 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
1837 
1838 static inline int kvm_cpu_get_apicid(int mps_cpu)
1839 {
1840 #ifdef CONFIG_X86_LOCAL_APIC
1841 	return default_cpu_present_to_apicid(mps_cpu);
1842 #else
1843 	WARN_ON_ONCE(1);
1844 	return BAD_APICID;
1845 #endif
1846 }
1847 
1848 #define put_smstate(type, buf, offset, val)                      \
1849 	*(type *)((buf) + (offset) - 0x7e00) = val
1850 
1851 #define GET_SMSTATE(type, buf, offset)		\
1852 	(*(type *)((buf) + (offset) - 0x7e00))
1853 
1854 int kvm_cpu_dirty_log_size(void);
1855 
1856 #endif /* _ASM_X86_KVM_HOST_H */
1857