xref: /openbmc/linux/arch/x86/include/asm/kvm_host.h (revision 26a9630c72ebac7c564db305a6aee54a8edde70e)
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 
93 #define CR0_RESERVED_BITS                                               \
94 	(~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \
95 			  | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \
96 			  | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG))
97 
98 #define CR4_RESERVED_BITS                                               \
99 	(~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\
100 			  | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE     \
101 			  | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR | X86_CR4_PCIDE \
102 			  | X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_FSGSBASE \
103 			  | X86_CR4_OSXMMEXCPT | X86_CR4_LA57 | X86_CR4_VMXE \
104 			  | X86_CR4_SMAP | X86_CR4_PKE | X86_CR4_UMIP))
105 
106 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR)
107 
108 
109 
110 #define INVALID_PAGE (~(hpa_t)0)
111 #define VALID_PAGE(x) ((x) != INVALID_PAGE)
112 
113 #define UNMAPPED_GVA (~(gpa_t)0)
114 
115 /* KVM Hugepage definitions for x86 */
116 #define KVM_MAX_HUGEPAGE_LEVEL	PG_LEVEL_1G
117 #define KVM_NR_PAGE_SIZES	(KVM_MAX_HUGEPAGE_LEVEL - PG_LEVEL_4K + 1)
118 #define KVM_HPAGE_GFN_SHIFT(x)	(((x) - 1) * 9)
119 #define KVM_HPAGE_SHIFT(x)	(PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
120 #define KVM_HPAGE_SIZE(x)	(1UL << KVM_HPAGE_SHIFT(x))
121 #define KVM_HPAGE_MASK(x)	(~(KVM_HPAGE_SIZE(x) - 1))
122 #define KVM_PAGES_PER_HPAGE(x)	(KVM_HPAGE_SIZE(x) / PAGE_SIZE)
123 
124 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
125 {
126 	/* KVM_HPAGE_GFN_SHIFT(PG_LEVEL_4K) must be 0. */
127 	return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
128 		(base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
129 }
130 
131 #define KVM_PERMILLE_MMU_PAGES 20
132 #define KVM_MIN_ALLOC_MMU_PAGES 64UL
133 #define KVM_MMU_HASH_SHIFT 12
134 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
135 #define KVM_MIN_FREE_MMU_PAGES 5
136 #define KVM_REFILL_PAGES 25
137 #define KVM_MAX_CPUID_ENTRIES 256
138 #define KVM_NR_FIXED_MTRR_REGION 88
139 #define KVM_NR_VAR_MTRR 8
140 
141 #define ASYNC_PF_PER_VCPU 64
142 
143 enum kvm_reg {
144 	VCPU_REGS_RAX = __VCPU_REGS_RAX,
145 	VCPU_REGS_RCX = __VCPU_REGS_RCX,
146 	VCPU_REGS_RDX = __VCPU_REGS_RDX,
147 	VCPU_REGS_RBX = __VCPU_REGS_RBX,
148 	VCPU_REGS_RSP = __VCPU_REGS_RSP,
149 	VCPU_REGS_RBP = __VCPU_REGS_RBP,
150 	VCPU_REGS_RSI = __VCPU_REGS_RSI,
151 	VCPU_REGS_RDI = __VCPU_REGS_RDI,
152 #ifdef CONFIG_X86_64
153 	VCPU_REGS_R8  = __VCPU_REGS_R8,
154 	VCPU_REGS_R9  = __VCPU_REGS_R9,
155 	VCPU_REGS_R10 = __VCPU_REGS_R10,
156 	VCPU_REGS_R11 = __VCPU_REGS_R11,
157 	VCPU_REGS_R12 = __VCPU_REGS_R12,
158 	VCPU_REGS_R13 = __VCPU_REGS_R13,
159 	VCPU_REGS_R14 = __VCPU_REGS_R14,
160 	VCPU_REGS_R15 = __VCPU_REGS_R15,
161 #endif
162 	VCPU_REGS_RIP,
163 	NR_VCPU_REGS,
164 
165 	VCPU_EXREG_PDPTR = NR_VCPU_REGS,
166 	VCPU_EXREG_CR0,
167 	VCPU_EXREG_CR3,
168 	VCPU_EXREG_CR4,
169 	VCPU_EXREG_RFLAGS,
170 	VCPU_EXREG_SEGMENTS,
171 	VCPU_EXREG_EXIT_INFO_1,
172 	VCPU_EXREG_EXIT_INFO_2,
173 };
174 
175 enum {
176 	VCPU_SREG_ES,
177 	VCPU_SREG_CS,
178 	VCPU_SREG_SS,
179 	VCPU_SREG_DS,
180 	VCPU_SREG_FS,
181 	VCPU_SREG_GS,
182 	VCPU_SREG_TR,
183 	VCPU_SREG_LDTR,
184 };
185 
186 enum exit_fastpath_completion {
187 	EXIT_FASTPATH_NONE,
188 	EXIT_FASTPATH_REENTER_GUEST,
189 	EXIT_FASTPATH_EXIT_HANDLED,
190 };
191 typedef enum exit_fastpath_completion fastpath_t;
192 
193 struct x86_emulate_ctxt;
194 struct x86_exception;
195 enum x86_intercept;
196 enum x86_intercept_stage;
197 
198 #define KVM_NR_DB_REGS	4
199 
200 #define DR6_BD		(1 << 13)
201 #define DR6_BS		(1 << 14)
202 #define DR6_BT		(1 << 15)
203 #define DR6_RTM		(1 << 16)
204 /*
205  * DR6_ACTIVE_LOW combines fixed-1 and active-low bits.
206  * We can regard all the bits in DR6_FIXED_1 as active_low bits;
207  * they will never be 0 for now, but when they are defined
208  * in the future it will require no code change.
209  *
210  * DR6_ACTIVE_LOW is also used as the init/reset value for DR6.
211  */
212 #define DR6_ACTIVE_LOW	0xffff0ff0
213 #define DR6_VOLATILE	0x0001e00f
214 #define DR6_FIXED_1	(DR6_ACTIVE_LOW & ~DR6_VOLATILE)
215 
216 #define DR7_BP_EN_MASK	0x000000ff
217 #define DR7_GE		(1 << 9)
218 #define DR7_GD		(1 << 13)
219 #define DR7_FIXED_1	0x00000400
220 #define DR7_VOLATILE	0xffff2bff
221 
222 #define PFERR_PRESENT_BIT 0
223 #define PFERR_WRITE_BIT 1
224 #define PFERR_USER_BIT 2
225 #define PFERR_RSVD_BIT 3
226 #define PFERR_FETCH_BIT 4
227 #define PFERR_PK_BIT 5
228 #define PFERR_GUEST_FINAL_BIT 32
229 #define PFERR_GUEST_PAGE_BIT 33
230 
231 #define PFERR_PRESENT_MASK (1U << PFERR_PRESENT_BIT)
232 #define PFERR_WRITE_MASK (1U << PFERR_WRITE_BIT)
233 #define PFERR_USER_MASK (1U << PFERR_USER_BIT)
234 #define PFERR_RSVD_MASK (1U << PFERR_RSVD_BIT)
235 #define PFERR_FETCH_MASK (1U << PFERR_FETCH_BIT)
236 #define PFERR_PK_MASK (1U << PFERR_PK_BIT)
237 #define PFERR_GUEST_FINAL_MASK (1ULL << PFERR_GUEST_FINAL_BIT)
238 #define PFERR_GUEST_PAGE_MASK (1ULL << PFERR_GUEST_PAGE_BIT)
239 
240 #define PFERR_NESTED_GUEST_PAGE (PFERR_GUEST_PAGE_MASK |	\
241 				 PFERR_WRITE_MASK |		\
242 				 PFERR_PRESENT_MASK)
243 
244 /* apic attention bits */
245 #define KVM_APIC_CHECK_VAPIC	0
246 /*
247  * The following bit is set with PV-EOI, unset on EOI.
248  * We detect PV-EOI changes by guest by comparing
249  * this bit with PV-EOI in guest memory.
250  * See the implementation in apic_update_pv_eoi.
251  */
252 #define KVM_APIC_PV_EOI_PENDING	1
253 
254 struct kvm_kernel_irq_routing_entry;
255 
256 /*
257  * the pages used as guest page table on soft mmu are tracked by
258  * kvm_memory_slot.arch.gfn_track which is 16 bits, so the role bits used
259  * by indirect shadow page can not be more than 15 bits.
260  *
261  * Currently, we used 14 bits that are @level, @gpte_is_8_bytes, @quadrant, @access,
262  * @nxe, @cr0_wp, @smep_andnot_wp and @smap_andnot_wp.
263  */
264 union kvm_mmu_page_role {
265 	u32 word;
266 	struct {
267 		unsigned level:4;
268 		unsigned gpte_is_8_bytes:1;
269 		unsigned quadrant:2;
270 		unsigned direct:1;
271 		unsigned access:3;
272 		unsigned invalid:1;
273 		unsigned nxe:1;
274 		unsigned cr0_wp:1;
275 		unsigned smep_andnot_wp:1;
276 		unsigned smap_andnot_wp:1;
277 		unsigned ad_disabled:1;
278 		unsigned guest_mode:1;
279 		unsigned :6;
280 
281 		/*
282 		 * This is left at the top of the word so that
283 		 * kvm_memslots_for_spte_role can extract it with a
284 		 * simple shift.  While there is room, give it a whole
285 		 * byte so it is also faster to load it from memory.
286 		 */
287 		unsigned smm:8;
288 	};
289 };
290 
291 union kvm_mmu_extended_role {
292 /*
293  * This structure complements kvm_mmu_page_role caching everything needed for
294  * MMU configuration. If nothing in both these structures changed, MMU
295  * re-configuration can be skipped. @valid bit is set on first usage so we don't
296  * treat all-zero structure as valid data.
297  */
298 	u32 word;
299 	struct {
300 		unsigned int valid:1;
301 		unsigned int execonly:1;
302 		unsigned int cr0_pg:1;
303 		unsigned int cr4_pae:1;
304 		unsigned int cr4_pse:1;
305 		unsigned int cr4_pke:1;
306 		unsigned int cr4_smap:1;
307 		unsigned int cr4_smep:1;
308 		unsigned int maxphyaddr:6;
309 	};
310 };
311 
312 union kvm_mmu_role {
313 	u64 as_u64;
314 	struct {
315 		union kvm_mmu_page_role base;
316 		union kvm_mmu_extended_role ext;
317 	};
318 };
319 
320 struct kvm_rmap_head {
321 	unsigned long val;
322 };
323 
324 struct kvm_pio_request {
325 	unsigned long linear_rip;
326 	unsigned long count;
327 	int in;
328 	int port;
329 	int size;
330 };
331 
332 #define PT64_ROOT_MAX_LEVEL 5
333 
334 struct rsvd_bits_validate {
335 	u64 rsvd_bits_mask[2][PT64_ROOT_MAX_LEVEL];
336 	u64 bad_mt_xwr;
337 };
338 
339 struct kvm_mmu_root_info {
340 	gpa_t pgd;
341 	hpa_t hpa;
342 };
343 
344 #define KVM_MMU_ROOT_INFO_INVALID \
345 	((struct kvm_mmu_root_info) { .pgd = INVALID_PAGE, .hpa = INVALID_PAGE })
346 
347 #define KVM_MMU_NUM_PREV_ROOTS 3
348 
349 #define KVM_HAVE_MMU_RWLOCK
350 
351 struct kvm_mmu_page;
352 
353 /*
354  * x86 supports 4 paging modes (5-level 64-bit, 4-level 64-bit, 3-level 32-bit,
355  * and 2-level 32-bit).  The kvm_mmu structure abstracts the details of the
356  * current mmu mode.
357  */
358 struct kvm_mmu {
359 	unsigned long (*get_guest_pgd)(struct kvm_vcpu *vcpu);
360 	u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
361 	int (*page_fault)(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u32 err,
362 			  bool prefault);
363 	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
364 				  struct x86_exception *fault);
365 	gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gpa_t gva_or_gpa,
366 			    u32 access, struct x86_exception *exception);
367 	gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
368 			       struct x86_exception *exception);
369 	int (*sync_page)(struct kvm_vcpu *vcpu,
370 			 struct kvm_mmu_page *sp);
371 	void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva, hpa_t root_hpa);
372 	hpa_t root_hpa;
373 	gpa_t root_pgd;
374 	union kvm_mmu_role mmu_role;
375 	u8 root_level;
376 	u8 shadow_root_level;
377 	u8 ept_ad;
378 	bool direct_map;
379 	struct kvm_mmu_root_info prev_roots[KVM_MMU_NUM_PREV_ROOTS];
380 
381 	/*
382 	 * Bitmap; bit set = permission fault
383 	 * Byte index: page fault error code [4:1]
384 	 * Bit index: pte permissions in ACC_* format
385 	 */
386 	u8 permissions[16];
387 
388 	/*
389 	* The pkru_mask indicates if protection key checks are needed.  It
390 	* consists of 16 domains indexed by page fault error code bits [4:1],
391 	* with PFEC.RSVD replaced by ACC_USER_MASK from the page tables.
392 	* Each domain has 2 bits which are ANDed with AD and WD from PKRU.
393 	*/
394 	u32 pkru_mask;
395 
396 	u64 *pae_root;
397 	u64 *lm_root;
398 
399 	/*
400 	 * check zero bits on shadow page table entries, these
401 	 * bits include not only hardware reserved bits but also
402 	 * the bits spte never used.
403 	 */
404 	struct rsvd_bits_validate shadow_zero_check;
405 
406 	struct rsvd_bits_validate guest_rsvd_check;
407 
408 	/* Can have large pages at levels 2..last_nonleaf_level-1. */
409 	u8 last_nonleaf_level;
410 
411 	bool nx;
412 
413 	u64 pdptrs[4]; /* pae */
414 };
415 
416 struct kvm_tlb_range {
417 	u64 start_gfn;
418 	u64 pages;
419 };
420 
421 enum pmc_type {
422 	KVM_PMC_GP = 0,
423 	KVM_PMC_FIXED,
424 };
425 
426 struct kvm_pmc {
427 	enum pmc_type type;
428 	u8 idx;
429 	u64 counter;
430 	u64 eventsel;
431 	struct perf_event *perf_event;
432 	struct kvm_vcpu *vcpu;
433 	/*
434 	 * eventsel value for general purpose counters,
435 	 * ctrl value for fixed counters.
436 	 */
437 	u64 current_config;
438 };
439 
440 struct kvm_pmu {
441 	unsigned nr_arch_gp_counters;
442 	unsigned nr_arch_fixed_counters;
443 	unsigned available_event_types;
444 	u64 fixed_ctr_ctrl;
445 	u64 global_ctrl;
446 	u64 global_status;
447 	u64 global_ovf_ctrl;
448 	u64 counter_bitmask[2];
449 	u64 global_ctrl_mask;
450 	u64 global_ovf_ctrl_mask;
451 	u64 reserved_bits;
452 	u8 version;
453 	struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
454 	struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED];
455 	struct irq_work irq_work;
456 	DECLARE_BITMAP(reprogram_pmi, X86_PMC_IDX_MAX);
457 	DECLARE_BITMAP(all_valid_pmc_idx, X86_PMC_IDX_MAX);
458 	DECLARE_BITMAP(pmc_in_use, X86_PMC_IDX_MAX);
459 
460 	/*
461 	 * The gate to release perf_events not marked in
462 	 * pmc_in_use only once in a vcpu time slice.
463 	 */
464 	bool need_cleanup;
465 
466 	/*
467 	 * The total number of programmed perf_events and it helps to avoid
468 	 * redundant check before cleanup if guest don't use vPMU at all.
469 	 */
470 	u8 event_count;
471 };
472 
473 struct kvm_pmu_ops;
474 
475 enum {
476 	KVM_DEBUGREG_BP_ENABLED = 1,
477 	KVM_DEBUGREG_WONT_EXIT = 2,
478 	KVM_DEBUGREG_RELOAD = 4,
479 };
480 
481 struct kvm_mtrr_range {
482 	u64 base;
483 	u64 mask;
484 	struct list_head node;
485 };
486 
487 struct kvm_mtrr {
488 	struct kvm_mtrr_range var_ranges[KVM_NR_VAR_MTRR];
489 	mtrr_type fixed_ranges[KVM_NR_FIXED_MTRR_REGION];
490 	u64 deftype;
491 
492 	struct list_head head;
493 };
494 
495 /* Hyper-V SynIC timer */
496 struct kvm_vcpu_hv_stimer {
497 	struct hrtimer timer;
498 	int index;
499 	union hv_stimer_config config;
500 	u64 count;
501 	u64 exp_time;
502 	struct hv_message msg;
503 	bool msg_pending;
504 };
505 
506 /* Hyper-V synthetic interrupt controller (SynIC)*/
507 struct kvm_vcpu_hv_synic {
508 	u64 version;
509 	u64 control;
510 	u64 msg_page;
511 	u64 evt_page;
512 	atomic64_t sint[HV_SYNIC_SINT_COUNT];
513 	atomic_t sint_to_gsi[HV_SYNIC_SINT_COUNT];
514 	DECLARE_BITMAP(auto_eoi_bitmap, 256);
515 	DECLARE_BITMAP(vec_bitmap, 256);
516 	bool active;
517 	bool dont_zero_synic_pages;
518 };
519 
520 /* Hyper-V per vcpu emulation context */
521 struct kvm_vcpu_hv {
522 	struct kvm_vcpu *vcpu;
523 	u32 vp_index;
524 	u64 hv_vapic;
525 	s64 runtime_offset;
526 	struct kvm_vcpu_hv_synic synic;
527 	struct kvm_hyperv_exit exit;
528 	struct kvm_vcpu_hv_stimer stimer[HV_SYNIC_STIMER_COUNT];
529 	DECLARE_BITMAP(stimer_pending_bitmap, HV_SYNIC_STIMER_COUNT);
530 	cpumask_t tlb_flush;
531 };
532 
533 /* Xen HVM per vcpu emulation context */
534 struct kvm_vcpu_xen {
535 	u64 hypercall_rip;
536 	bool vcpu_info_set;
537 	bool vcpu_time_info_set;
538 	struct gfn_to_hva_cache vcpu_info_cache;
539 	struct gfn_to_hva_cache vcpu_time_info_cache;
540 };
541 
542 struct kvm_vcpu_arch {
543 	/*
544 	 * rip and regs accesses must go through
545 	 * kvm_{register,rip}_{read,write} functions.
546 	 */
547 	unsigned long regs[NR_VCPU_REGS];
548 	u32 regs_avail;
549 	u32 regs_dirty;
550 
551 	unsigned long cr0;
552 	unsigned long cr0_guest_owned_bits;
553 	unsigned long cr2;
554 	unsigned long cr3;
555 	unsigned long cr4;
556 	unsigned long cr4_guest_owned_bits;
557 	unsigned long cr4_guest_rsvd_bits;
558 	unsigned long cr8;
559 	u32 host_pkru;
560 	u32 pkru;
561 	u32 hflags;
562 	u64 efer;
563 	u64 apic_base;
564 	struct kvm_lapic *apic;    /* kernel irqchip context */
565 	bool apicv_active;
566 	bool load_eoi_exitmap_pending;
567 	DECLARE_BITMAP(ioapic_handled_vectors, 256);
568 	unsigned long apic_attention;
569 	int32_t apic_arb_prio;
570 	int mp_state;
571 	u64 ia32_misc_enable_msr;
572 	u64 smbase;
573 	u64 smi_count;
574 	bool tpr_access_reporting;
575 	bool xsaves_enabled;
576 	u64 ia32_xss;
577 	u64 microcode_version;
578 	u64 arch_capabilities;
579 	u64 perf_capabilities;
580 
581 	/*
582 	 * Paging state of the vcpu
583 	 *
584 	 * If the vcpu runs in guest mode with two level paging this still saves
585 	 * the paging mode of the l1 guest. This context is always used to
586 	 * handle faults.
587 	 */
588 	struct kvm_mmu *mmu;
589 
590 	/* Non-nested MMU for L1 */
591 	struct kvm_mmu root_mmu;
592 
593 	/* L1 MMU when running nested */
594 	struct kvm_mmu guest_mmu;
595 
596 	/*
597 	 * Paging state of an L2 guest (used for nested npt)
598 	 *
599 	 * This context will save all necessary information to walk page tables
600 	 * of an L2 guest. This context is only initialized for page table
601 	 * walking and not for faulting since we never handle l2 page faults on
602 	 * the host.
603 	 */
604 	struct kvm_mmu nested_mmu;
605 
606 	/*
607 	 * Pointer to the mmu context currently used for
608 	 * gva_to_gpa translations.
609 	 */
610 	struct kvm_mmu *walk_mmu;
611 
612 	struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
613 	struct kvm_mmu_memory_cache mmu_shadow_page_cache;
614 	struct kvm_mmu_memory_cache mmu_gfn_array_cache;
615 	struct kvm_mmu_memory_cache mmu_page_header_cache;
616 
617 	/*
618 	 * QEMU userspace and the guest each have their own FPU state.
619 	 * In vcpu_run, we switch between the user and guest FPU contexts.
620 	 * While running a VCPU, the VCPU thread will have the guest FPU
621 	 * context.
622 	 *
623 	 * Note that while the PKRU state lives inside the fpu registers,
624 	 * it is switched out separately at VMENTER and VMEXIT time. The
625 	 * "guest_fpu" state here contains the guest FPU context, with the
626 	 * host PRKU bits.
627 	 */
628 	struct fpu *user_fpu;
629 	struct fpu *guest_fpu;
630 
631 	u64 xcr0;
632 	u64 guest_supported_xcr0;
633 
634 	struct kvm_pio_request pio;
635 	void *pio_data;
636 	void *guest_ins_data;
637 
638 	u8 event_exit_inst_len;
639 
640 	struct kvm_queued_exception {
641 		bool pending;
642 		bool injected;
643 		bool has_error_code;
644 		u8 nr;
645 		u32 error_code;
646 		unsigned long payload;
647 		bool has_payload;
648 		u8 nested_apf;
649 	} exception;
650 
651 	struct kvm_queued_interrupt {
652 		bool injected;
653 		bool soft;
654 		u8 nr;
655 	} interrupt;
656 
657 	int halt_request; /* real mode on Intel only */
658 
659 	int cpuid_nent;
660 	struct kvm_cpuid_entry2 *cpuid_entries;
661 
662 	u64 reserved_gpa_bits;
663 	int maxphyaddr;
664 	int max_tdp_level;
665 
666 	/* emulate context */
667 
668 	struct x86_emulate_ctxt *emulate_ctxt;
669 	bool emulate_regs_need_sync_to_vcpu;
670 	bool emulate_regs_need_sync_from_vcpu;
671 	int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
672 
673 	gpa_t time;
674 	struct pvclock_vcpu_time_info hv_clock;
675 	unsigned int hw_tsc_khz;
676 	struct gfn_to_hva_cache pv_time;
677 	bool pv_time_enabled;
678 	/* set guest stopped flag in pvclock flags field */
679 	bool pvclock_set_guest_stopped_request;
680 
681 	struct {
682 		u8 preempted;
683 		u64 msr_val;
684 		u64 last_steal;
685 		struct gfn_to_pfn_cache cache;
686 	} st;
687 
688 	u64 l1_tsc_offset;
689 	u64 tsc_offset;
690 	u64 last_guest_tsc;
691 	u64 last_host_tsc;
692 	u64 tsc_offset_adjustment;
693 	u64 this_tsc_nsec;
694 	u64 this_tsc_write;
695 	u64 this_tsc_generation;
696 	bool tsc_catchup;
697 	bool tsc_always_catchup;
698 	s8 virtual_tsc_shift;
699 	u32 virtual_tsc_mult;
700 	u32 virtual_tsc_khz;
701 	s64 ia32_tsc_adjust_msr;
702 	u64 msr_ia32_power_ctl;
703 	u64 tsc_scaling_ratio;
704 
705 	atomic_t nmi_queued;  /* unprocessed asynchronous NMIs */
706 	unsigned nmi_pending; /* NMI queued after currently running handler */
707 	bool nmi_injected;    /* Trying to inject an NMI this entry */
708 	bool smi_pending;    /* SMI queued after currently running handler */
709 
710 	struct kvm_mtrr mtrr_state;
711 	u64 pat;
712 
713 	unsigned switch_db_regs;
714 	unsigned long db[KVM_NR_DB_REGS];
715 	unsigned long dr6;
716 	unsigned long dr7;
717 	unsigned long eff_db[KVM_NR_DB_REGS];
718 	unsigned long guest_debug_dr7;
719 	u64 msr_platform_info;
720 	u64 msr_misc_features_enables;
721 
722 	u64 mcg_cap;
723 	u64 mcg_status;
724 	u64 mcg_ctl;
725 	u64 mcg_ext_ctl;
726 	u64 *mce_banks;
727 
728 	/* Cache MMIO info */
729 	u64 mmio_gva;
730 	unsigned mmio_access;
731 	gfn_t mmio_gfn;
732 	u64 mmio_gen;
733 
734 	struct kvm_pmu pmu;
735 
736 	/* used for guest single stepping over the given code position */
737 	unsigned long singlestep_rip;
738 
739 	bool hyperv_enabled;
740 	struct kvm_vcpu_hv *hyperv;
741 	struct kvm_vcpu_xen xen;
742 
743 	cpumask_var_t wbinvd_dirty_mask;
744 
745 	unsigned long last_retry_eip;
746 	unsigned long last_retry_addr;
747 
748 	struct {
749 		bool halted;
750 		gfn_t gfns[ASYNC_PF_PER_VCPU];
751 		struct gfn_to_hva_cache data;
752 		u64 msr_en_val; /* MSR_KVM_ASYNC_PF_EN */
753 		u64 msr_int_val; /* MSR_KVM_ASYNC_PF_INT */
754 		u16 vec;
755 		u32 id;
756 		bool send_user_only;
757 		u32 host_apf_flags;
758 		unsigned long nested_apf_token;
759 		bool delivery_as_pf_vmexit;
760 		bool pageready_pending;
761 	} apf;
762 
763 	/* OSVW MSRs (AMD only) */
764 	struct {
765 		u64 length;
766 		u64 status;
767 	} osvw;
768 
769 	struct {
770 		u64 msr_val;
771 		struct gfn_to_hva_cache data;
772 	} pv_eoi;
773 
774 	u64 msr_kvm_poll_control;
775 
776 	/*
777 	 * Indicates the guest is trying to write a gfn that contains one or
778 	 * more of the PTEs used to translate the write itself, i.e. the access
779 	 * is changing its own translation in the guest page tables.  KVM exits
780 	 * to userspace if emulation of the faulting instruction fails and this
781 	 * flag is set, as KVM cannot make forward progress.
782 	 *
783 	 * If emulation fails for a write to guest page tables, KVM unprotects
784 	 * (zaps) the shadow page for the target gfn and resumes the guest to
785 	 * retry the non-emulatable instruction (on hardware).  Unprotecting the
786 	 * gfn doesn't allow forward progress for a self-changing access because
787 	 * doing so also zaps the translation for the gfn, i.e. retrying the
788 	 * instruction will hit a !PRESENT fault, which results in a new shadow
789 	 * page and sends KVM back to square one.
790 	 */
791 	bool write_fault_to_shadow_pgtable;
792 
793 	/* set at EPT violation at this point */
794 	unsigned long exit_qualification;
795 
796 	/* pv related host specific info */
797 	struct {
798 		bool pv_unhalted;
799 	} pv;
800 
801 	int pending_ioapic_eoi;
802 	int pending_external_vector;
803 
804 	/* be preempted when it's in kernel-mode(cpl=0) */
805 	bool preempted_in_kernel;
806 
807 	/* Flush the L1 Data cache for L1TF mitigation on VMENTER */
808 	bool l1tf_flush_l1d;
809 
810 	/* Host CPU on which VM-entry was most recently attempted */
811 	unsigned int last_vmentry_cpu;
812 
813 	/* AMD MSRC001_0015 Hardware Configuration */
814 	u64 msr_hwcr;
815 
816 	/* pv related cpuid info */
817 	struct {
818 		/*
819 		 * value of the eax register in the KVM_CPUID_FEATURES CPUID
820 		 * leaf.
821 		 */
822 		u32 features;
823 
824 		/*
825 		 * indicates whether pv emulation should be disabled if features
826 		 * are not present in the guest's cpuid
827 		 */
828 		bool enforce;
829 	} pv_cpuid;
830 
831 	/* Protected Guests */
832 	bool guest_state_protected;
833 };
834 
835 struct kvm_lpage_info {
836 	int disallow_lpage;
837 };
838 
839 struct kvm_arch_memory_slot {
840 	struct kvm_rmap_head *rmap[KVM_NR_PAGE_SIZES];
841 	struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
842 	unsigned short *gfn_track[KVM_PAGE_TRACK_MAX];
843 };
844 
845 /*
846  * We use as the mode the number of bits allocated in the LDR for the
847  * logical processor ID.  It happens that these are all powers of two.
848  * This makes it is very easy to detect cases where the APICs are
849  * configured for multiple modes; in that case, we cannot use the map and
850  * hence cannot use kvm_irq_delivery_to_apic_fast either.
851  */
852 #define KVM_APIC_MODE_XAPIC_CLUSTER          4
853 #define KVM_APIC_MODE_XAPIC_FLAT             8
854 #define KVM_APIC_MODE_X2APIC                16
855 
856 struct kvm_apic_map {
857 	struct rcu_head rcu;
858 	u8 mode;
859 	u32 max_apic_id;
860 	union {
861 		struct kvm_lapic *xapic_flat_map[8];
862 		struct kvm_lapic *xapic_cluster_map[16][4];
863 	};
864 	struct kvm_lapic *phys_map[];
865 };
866 
867 /* Hyper-V synthetic debugger (SynDbg)*/
868 struct kvm_hv_syndbg {
869 	struct {
870 		u64 control;
871 		u64 status;
872 		u64 send_page;
873 		u64 recv_page;
874 		u64 pending_page;
875 	} control;
876 	u64 options;
877 };
878 
879 /* Hyper-V emulation context */
880 struct kvm_hv {
881 	struct mutex hv_lock;
882 	u64 hv_guest_os_id;
883 	u64 hv_hypercall;
884 	u64 hv_tsc_page;
885 
886 	/* Hyper-v based guest crash (NT kernel bugcheck) parameters */
887 	u64 hv_crash_param[HV_X64_MSR_CRASH_PARAMS];
888 	u64 hv_crash_ctl;
889 
890 	struct ms_hyperv_tsc_page tsc_ref;
891 
892 	struct idr conn_to_evt;
893 
894 	u64 hv_reenlightenment_control;
895 	u64 hv_tsc_emulation_control;
896 	u64 hv_tsc_emulation_status;
897 
898 	/* How many vCPUs have VP index != vCPU index */
899 	atomic_t num_mismatched_vp_indexes;
900 
901 	struct hv_partition_assist_pg *hv_pa_pg;
902 	struct kvm_hv_syndbg hv_syndbg;
903 };
904 
905 struct msr_bitmap_range {
906 	u32 flags;
907 	u32 nmsrs;
908 	u32 base;
909 	unsigned long *bitmap;
910 };
911 
912 /* Xen emulation context */
913 struct kvm_xen {
914 	bool long_mode;
915 	bool shinfo_set;
916 	u8 upcall_vector;
917 	struct gfn_to_hva_cache shinfo_cache;
918 };
919 
920 enum kvm_irqchip_mode {
921 	KVM_IRQCHIP_NONE,
922 	KVM_IRQCHIP_KERNEL,       /* created with KVM_CREATE_IRQCHIP */
923 	KVM_IRQCHIP_SPLIT,        /* created with KVM_CAP_SPLIT_IRQCHIP */
924 };
925 
926 #define APICV_INHIBIT_REASON_DISABLE    0
927 #define APICV_INHIBIT_REASON_HYPERV     1
928 #define APICV_INHIBIT_REASON_NESTED     2
929 #define APICV_INHIBIT_REASON_IRQWIN     3
930 #define APICV_INHIBIT_REASON_PIT_REINJ  4
931 #define APICV_INHIBIT_REASON_X2APIC	5
932 
933 struct kvm_arch {
934 	unsigned long n_used_mmu_pages;
935 	unsigned long n_requested_mmu_pages;
936 	unsigned long n_max_mmu_pages;
937 	unsigned int indirect_shadow_pages;
938 	u8 mmu_valid_gen;
939 	struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
940 	/*
941 	 * Hash table of struct kvm_mmu_page.
942 	 */
943 	struct list_head active_mmu_pages;
944 	struct list_head zapped_obsolete_pages;
945 	struct list_head lpage_disallowed_mmu_pages;
946 	struct kvm_page_track_notifier_node mmu_sp_tracker;
947 	struct kvm_page_track_notifier_head track_notifier_head;
948 
949 	struct list_head assigned_dev_head;
950 	struct iommu_domain *iommu_domain;
951 	bool iommu_noncoherent;
952 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA
953 	atomic_t noncoherent_dma_count;
954 #define __KVM_HAVE_ARCH_ASSIGNED_DEVICE
955 	atomic_t assigned_device_count;
956 	struct kvm_pic *vpic;
957 	struct kvm_ioapic *vioapic;
958 	struct kvm_pit *vpit;
959 	atomic_t vapics_in_nmi_mode;
960 	struct mutex apic_map_lock;
961 	struct kvm_apic_map *apic_map;
962 	atomic_t apic_map_dirty;
963 
964 	bool apic_access_page_done;
965 	unsigned long apicv_inhibit_reasons;
966 
967 	gpa_t wall_clock;
968 
969 	bool mwait_in_guest;
970 	bool hlt_in_guest;
971 	bool pause_in_guest;
972 	bool cstate_in_guest;
973 
974 	unsigned long irq_sources_bitmap;
975 	s64 kvmclock_offset;
976 	raw_spinlock_t tsc_write_lock;
977 	u64 last_tsc_nsec;
978 	u64 last_tsc_write;
979 	u32 last_tsc_khz;
980 	u64 cur_tsc_nsec;
981 	u64 cur_tsc_write;
982 	u64 cur_tsc_offset;
983 	u64 cur_tsc_generation;
984 	int nr_vcpus_matched_tsc;
985 
986 	spinlock_t pvclock_gtod_sync_lock;
987 	bool use_master_clock;
988 	u64 master_kernel_ns;
989 	u64 master_cycle_now;
990 	struct delayed_work kvmclock_update_work;
991 	struct delayed_work kvmclock_sync_work;
992 
993 	struct kvm_xen_hvm_config xen_hvm_config;
994 
995 	/* reads protected by irq_srcu, writes by irq_lock */
996 	struct hlist_head mask_notifier_list;
997 
998 	struct kvm_hv hyperv;
999 	struct kvm_xen xen;
1000 
1001 	#ifdef CONFIG_KVM_MMU_AUDIT
1002 	int audit_point;
1003 	#endif
1004 
1005 	bool backwards_tsc_observed;
1006 	bool boot_vcpu_runs_old_kvmclock;
1007 	u32 bsp_vcpu_id;
1008 
1009 	u64 disabled_quirks;
1010 
1011 	enum kvm_irqchip_mode irqchip_mode;
1012 	u8 nr_reserved_ioapic_pins;
1013 
1014 	bool disabled_lapic_found;
1015 
1016 	bool x2apic_format;
1017 	bool x2apic_broadcast_quirk_disabled;
1018 
1019 	bool guest_can_read_msr_platform_info;
1020 	bool exception_payload_enabled;
1021 
1022 	/* Deflect RDMSR and WRMSR to user space when they trigger a #GP */
1023 	u32 user_space_msr_mask;
1024 
1025 	struct {
1026 		u8 count;
1027 		bool default_allow:1;
1028 		struct msr_bitmap_range ranges[16];
1029 	} msr_filter;
1030 
1031 	bool bus_lock_detection_enabled;
1032 
1033 	struct kvm_pmu_event_filter *pmu_event_filter;
1034 	struct task_struct *nx_lpage_recovery_thread;
1035 
1036 #ifdef CONFIG_X86_64
1037 	/*
1038 	 * Whether the TDP MMU is enabled for this VM. This contains a
1039 	 * snapshot of the TDP MMU module parameter from when the VM was
1040 	 * created and remains unchanged for the life of the VM. If this is
1041 	 * true, TDP MMU handler functions will run for various MMU
1042 	 * operations.
1043 	 */
1044 	bool tdp_mmu_enabled;
1045 
1046 	/*
1047 	 * List of struct kvmp_mmu_pages being used as roots.
1048 	 * All struct kvm_mmu_pages in the list should have
1049 	 * tdp_mmu_page set.
1050 	 * All struct kvm_mmu_pages in the list should have a positive
1051 	 * root_count except when a thread holds the MMU lock and is removing
1052 	 * an entry from the list.
1053 	 */
1054 	struct list_head tdp_mmu_roots;
1055 
1056 	/*
1057 	 * List of struct kvmp_mmu_pages not being used as roots.
1058 	 * All struct kvm_mmu_pages in the list should have
1059 	 * tdp_mmu_page set and a root_count of 0.
1060 	 */
1061 	struct list_head tdp_mmu_pages;
1062 
1063 	/*
1064 	 * Protects accesses to the following fields when the MMU lock
1065 	 * is held in read mode:
1066 	 *  - tdp_mmu_pages (above)
1067 	 *  - the link field of struct kvm_mmu_pages used by the TDP MMU
1068 	 *  - lpage_disallowed_mmu_pages
1069 	 *  - the lpage_disallowed_link field of struct kvm_mmu_pages used
1070 	 *    by the TDP MMU
1071 	 * It is acceptable, but not necessary, to acquire this lock when
1072 	 * the thread holds the MMU lock in write mode.
1073 	 */
1074 	spinlock_t tdp_mmu_pages_lock;
1075 #endif /* CONFIG_X86_64 */
1076 };
1077 
1078 struct kvm_vm_stat {
1079 	ulong mmu_shadow_zapped;
1080 	ulong mmu_pte_write;
1081 	ulong mmu_pde_zapped;
1082 	ulong mmu_flooded;
1083 	ulong mmu_recycled;
1084 	ulong mmu_cache_miss;
1085 	ulong mmu_unsync;
1086 	ulong remote_tlb_flush;
1087 	ulong lpages;
1088 	ulong nx_lpage_splits;
1089 	ulong max_mmu_page_hash_collisions;
1090 };
1091 
1092 struct kvm_vcpu_stat {
1093 	u64 pf_fixed;
1094 	u64 pf_guest;
1095 	u64 tlb_flush;
1096 	u64 invlpg;
1097 
1098 	u64 exits;
1099 	u64 io_exits;
1100 	u64 mmio_exits;
1101 	u64 signal_exits;
1102 	u64 irq_window_exits;
1103 	u64 nmi_window_exits;
1104 	u64 l1d_flush;
1105 	u64 halt_exits;
1106 	u64 halt_successful_poll;
1107 	u64 halt_attempted_poll;
1108 	u64 halt_poll_invalid;
1109 	u64 halt_wakeup;
1110 	u64 request_irq_exits;
1111 	u64 irq_exits;
1112 	u64 host_state_reload;
1113 	u64 fpu_reload;
1114 	u64 insn_emulation;
1115 	u64 insn_emulation_fail;
1116 	u64 hypercalls;
1117 	u64 irq_injections;
1118 	u64 nmi_injections;
1119 	u64 req_event;
1120 	u64 halt_poll_success_ns;
1121 	u64 halt_poll_fail_ns;
1122 };
1123 
1124 struct x86_instruction_info;
1125 
1126 struct msr_data {
1127 	bool host_initiated;
1128 	u32 index;
1129 	u64 data;
1130 };
1131 
1132 struct kvm_lapic_irq {
1133 	u32 vector;
1134 	u16 delivery_mode;
1135 	u16 dest_mode;
1136 	bool level;
1137 	u16 trig_mode;
1138 	u32 shorthand;
1139 	u32 dest_id;
1140 	bool msi_redir_hint;
1141 };
1142 
1143 static inline u16 kvm_lapic_irq_dest_mode(bool dest_mode_logical)
1144 {
1145 	return dest_mode_logical ? APIC_DEST_LOGICAL : APIC_DEST_PHYSICAL;
1146 }
1147 
1148 struct kvm_x86_ops {
1149 	int (*hardware_enable)(void);
1150 	void (*hardware_disable)(void);
1151 	void (*hardware_unsetup)(void);
1152 	bool (*cpu_has_accelerated_tpr)(void);
1153 	bool (*has_emulated_msr)(struct kvm *kvm, u32 index);
1154 	void (*vcpu_after_set_cpuid)(struct kvm_vcpu *vcpu);
1155 
1156 	unsigned int vm_size;
1157 	int (*vm_init)(struct kvm *kvm);
1158 	void (*vm_destroy)(struct kvm *kvm);
1159 
1160 	/* Create, but do not attach this VCPU */
1161 	int (*vcpu_create)(struct kvm_vcpu *vcpu);
1162 	void (*vcpu_free)(struct kvm_vcpu *vcpu);
1163 	void (*vcpu_reset)(struct kvm_vcpu *vcpu, bool init_event);
1164 
1165 	void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
1166 	void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
1167 	void (*vcpu_put)(struct kvm_vcpu *vcpu);
1168 
1169 	void (*update_exception_bitmap)(struct kvm_vcpu *vcpu);
1170 	int (*get_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1171 	int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1172 	u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
1173 	void (*get_segment)(struct kvm_vcpu *vcpu,
1174 			    struct kvm_segment *var, int seg);
1175 	int (*get_cpl)(struct kvm_vcpu *vcpu);
1176 	void (*set_segment)(struct kvm_vcpu *vcpu,
1177 			    struct kvm_segment *var, int seg);
1178 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
1179 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
1180 	bool (*is_valid_cr4)(struct kvm_vcpu *vcpu, unsigned long cr0);
1181 	void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
1182 	int (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
1183 	void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1184 	void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1185 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1186 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1187 	void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu);
1188 	void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
1189 	void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
1190 	unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
1191 	void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
1192 
1193 	void (*tlb_flush_all)(struct kvm_vcpu *vcpu);
1194 	void (*tlb_flush_current)(struct kvm_vcpu *vcpu);
1195 	int  (*tlb_remote_flush)(struct kvm *kvm);
1196 	int  (*tlb_remote_flush_with_range)(struct kvm *kvm,
1197 			struct kvm_tlb_range *range);
1198 
1199 	/*
1200 	 * Flush any TLB entries associated with the given GVA.
1201 	 * Does not need to flush GPA->HPA mappings.
1202 	 * Can potentially get non-canonical addresses through INVLPGs, which
1203 	 * the implementation may choose to ignore if appropriate.
1204 	 */
1205 	void (*tlb_flush_gva)(struct kvm_vcpu *vcpu, gva_t addr);
1206 
1207 	/*
1208 	 * Flush any TLB entries created by the guest.  Like tlb_flush_gva(),
1209 	 * does not need to flush GPA->HPA mappings.
1210 	 */
1211 	void (*tlb_flush_guest)(struct kvm_vcpu *vcpu);
1212 
1213 	enum exit_fastpath_completion (*run)(struct kvm_vcpu *vcpu);
1214 	int (*handle_exit)(struct kvm_vcpu *vcpu,
1215 		enum exit_fastpath_completion exit_fastpath);
1216 	int (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
1217 	void (*update_emulated_instruction)(struct kvm_vcpu *vcpu);
1218 	void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
1219 	u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu);
1220 	void (*patch_hypercall)(struct kvm_vcpu *vcpu,
1221 				unsigned char *hypercall_addr);
1222 	void (*set_irq)(struct kvm_vcpu *vcpu);
1223 	void (*set_nmi)(struct kvm_vcpu *vcpu);
1224 	void (*queue_exception)(struct kvm_vcpu *vcpu);
1225 	void (*cancel_injection)(struct kvm_vcpu *vcpu);
1226 	int (*interrupt_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1227 	int (*nmi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1228 	bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
1229 	void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
1230 	void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
1231 	void (*enable_irq_window)(struct kvm_vcpu *vcpu);
1232 	void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
1233 	bool (*check_apicv_inhibit_reasons)(ulong bit);
1234 	void (*pre_update_apicv_exec_ctrl)(struct kvm *kvm, bool activate);
1235 	void (*refresh_apicv_exec_ctrl)(struct kvm_vcpu *vcpu);
1236 	void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
1237 	void (*hwapic_isr_update)(struct kvm_vcpu *vcpu, int isr);
1238 	bool (*guest_apic_has_interrupt)(struct kvm_vcpu *vcpu);
1239 	void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
1240 	void (*set_virtual_apic_mode)(struct kvm_vcpu *vcpu);
1241 	void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu);
1242 	int (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
1243 	int (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
1244 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
1245 	int (*set_identity_map_addr)(struct kvm *kvm, u64 ident_addr);
1246 	u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
1247 
1248 	void (*load_mmu_pgd)(struct kvm_vcpu *vcpu, unsigned long pgd,
1249 			     int pgd_level);
1250 
1251 	bool (*has_wbinvd_exit)(void);
1252 
1253 	/* Returns actual tsc_offset set in active VMCS */
1254 	u64 (*write_l1_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
1255 
1256 	/*
1257 	 * Retrieve somewhat arbitrary exit information.  Intended to be used
1258 	 * only from within tracepoints to avoid VMREADs when tracing is off.
1259 	 */
1260 	void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2,
1261 			      u32 *exit_int_info, u32 *exit_int_info_err_code);
1262 
1263 	int (*check_intercept)(struct kvm_vcpu *vcpu,
1264 			       struct x86_instruction_info *info,
1265 			       enum x86_intercept_stage stage,
1266 			       struct x86_exception *exception);
1267 	void (*handle_exit_irqoff)(struct kvm_vcpu *vcpu);
1268 
1269 	void (*request_immediate_exit)(struct kvm_vcpu *vcpu);
1270 
1271 	void (*sched_in)(struct kvm_vcpu *kvm, int cpu);
1272 
1273 	/*
1274 	 * Arch-specific dirty logging hooks. These hooks are only supposed to
1275 	 * be valid if the specific arch has hardware-accelerated dirty logging
1276 	 * mechanism. Currently only for PML on VMX.
1277 	 *
1278 	 *  - slot_enable_log_dirty:
1279 	 *	called when enabling log dirty mode for the slot.
1280 	 *  - slot_disable_log_dirty:
1281 	 *	called when disabling log dirty mode for the slot.
1282 	 *	also called when slot is created with log dirty disabled.
1283 	 *  - flush_log_dirty:
1284 	 *	called before reporting dirty_bitmap to userspace.
1285 	 *  - enable_log_dirty_pt_masked:
1286 	 *	called when reenabling log dirty for the GFNs in the mask after
1287 	 *	corresponding bits are cleared in slot->dirty_bitmap.
1288 	 */
1289 	void (*slot_enable_log_dirty)(struct kvm *kvm,
1290 				      struct kvm_memory_slot *slot);
1291 	void (*slot_disable_log_dirty)(struct kvm *kvm,
1292 				       struct kvm_memory_slot *slot);
1293 	void (*flush_log_dirty)(struct kvm *kvm);
1294 	void (*enable_log_dirty_pt_masked)(struct kvm *kvm,
1295 					   struct kvm_memory_slot *slot,
1296 					   gfn_t offset, unsigned long mask);
1297 	int (*cpu_dirty_log_size)(void);
1298 
1299 	/* pmu operations of sub-arch */
1300 	const struct kvm_pmu_ops *pmu_ops;
1301 	const struct kvm_x86_nested_ops *nested_ops;
1302 
1303 	/*
1304 	 * Architecture specific hooks for vCPU blocking due to
1305 	 * HLT instruction.
1306 	 * Returns for .pre_block():
1307 	 *    - 0 means continue to block the vCPU.
1308 	 *    - 1 means we cannot block the vCPU since some event
1309 	 *        happens during this period, such as, 'ON' bit in
1310 	 *        posted-interrupts descriptor is set.
1311 	 */
1312 	int (*pre_block)(struct kvm_vcpu *vcpu);
1313 	void (*post_block)(struct kvm_vcpu *vcpu);
1314 
1315 	void (*vcpu_blocking)(struct kvm_vcpu *vcpu);
1316 	void (*vcpu_unblocking)(struct kvm_vcpu *vcpu);
1317 
1318 	int (*update_pi_irte)(struct kvm *kvm, unsigned int host_irq,
1319 			      uint32_t guest_irq, bool set);
1320 	void (*apicv_post_state_restore)(struct kvm_vcpu *vcpu);
1321 	bool (*dy_apicv_has_pending_interrupt)(struct kvm_vcpu *vcpu);
1322 
1323 	int (*set_hv_timer)(struct kvm_vcpu *vcpu, u64 guest_deadline_tsc,
1324 			    bool *expired);
1325 	void (*cancel_hv_timer)(struct kvm_vcpu *vcpu);
1326 
1327 	void (*setup_mce)(struct kvm_vcpu *vcpu);
1328 
1329 	int (*smi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1330 	int (*pre_enter_smm)(struct kvm_vcpu *vcpu, char *smstate);
1331 	int (*pre_leave_smm)(struct kvm_vcpu *vcpu, const char *smstate);
1332 	void (*enable_smi_window)(struct kvm_vcpu *vcpu);
1333 
1334 	int (*mem_enc_op)(struct kvm *kvm, void __user *argp);
1335 	int (*mem_enc_reg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1336 	int (*mem_enc_unreg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1337 
1338 	int (*get_msr_feature)(struct kvm_msr_entry *entry);
1339 
1340 	bool (*can_emulate_instruction)(struct kvm_vcpu *vcpu, void *insn, int insn_len);
1341 
1342 	bool (*apic_init_signal_blocked)(struct kvm_vcpu *vcpu);
1343 	int (*enable_direct_tlbflush)(struct kvm_vcpu *vcpu);
1344 
1345 	void (*migrate_timers)(struct kvm_vcpu *vcpu);
1346 	void (*msr_filter_changed)(struct kvm_vcpu *vcpu);
1347 	int (*complete_emulated_msr)(struct kvm_vcpu *vcpu, int err);
1348 
1349 	void (*vcpu_deliver_sipi_vector)(struct kvm_vcpu *vcpu, u8 vector);
1350 };
1351 
1352 struct kvm_x86_nested_ops {
1353 	int (*check_events)(struct kvm_vcpu *vcpu);
1354 	bool (*hv_timer_pending)(struct kvm_vcpu *vcpu);
1355 	int (*get_state)(struct kvm_vcpu *vcpu,
1356 			 struct kvm_nested_state __user *user_kvm_nested_state,
1357 			 unsigned user_data_size);
1358 	int (*set_state)(struct kvm_vcpu *vcpu,
1359 			 struct kvm_nested_state __user *user_kvm_nested_state,
1360 			 struct kvm_nested_state *kvm_state);
1361 	bool (*get_nested_state_pages)(struct kvm_vcpu *vcpu);
1362 	int (*write_log_dirty)(struct kvm_vcpu *vcpu, gpa_t l2_gpa);
1363 
1364 	int (*enable_evmcs)(struct kvm_vcpu *vcpu,
1365 			    uint16_t *vmcs_version);
1366 	uint16_t (*get_evmcs_version)(struct kvm_vcpu *vcpu);
1367 };
1368 
1369 struct kvm_x86_init_ops {
1370 	int (*cpu_has_kvm_support)(void);
1371 	int (*disabled_by_bios)(void);
1372 	int (*check_processor_compatibility)(void);
1373 	int (*hardware_setup)(void);
1374 
1375 	struct kvm_x86_ops *runtime_ops;
1376 };
1377 
1378 struct kvm_arch_async_pf {
1379 	u32 token;
1380 	gfn_t gfn;
1381 	unsigned long cr3;
1382 	bool direct_map;
1383 };
1384 
1385 extern u64 __read_mostly host_efer;
1386 extern bool __read_mostly allow_smaller_maxphyaddr;
1387 extern struct kvm_x86_ops kvm_x86_ops;
1388 
1389 #define KVM_X86_OP(func) \
1390 	DECLARE_STATIC_CALL(kvm_x86_##func, *(((struct kvm_x86_ops *)0)->func));
1391 #define KVM_X86_OP_NULL KVM_X86_OP
1392 #include <asm/kvm-x86-ops.h>
1393 
1394 static inline void kvm_ops_static_call_update(void)
1395 {
1396 #define KVM_X86_OP(func) \
1397 	static_call_update(kvm_x86_##func, kvm_x86_ops.func);
1398 #define KVM_X86_OP_NULL KVM_X86_OP
1399 #include <asm/kvm-x86-ops.h>
1400 }
1401 
1402 #define __KVM_HAVE_ARCH_VM_ALLOC
1403 static inline struct kvm *kvm_arch_alloc_vm(void)
1404 {
1405 	return __vmalloc(kvm_x86_ops.vm_size, GFP_KERNEL_ACCOUNT | __GFP_ZERO);
1406 }
1407 void kvm_arch_free_vm(struct kvm *kvm);
1408 
1409 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
1410 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
1411 {
1412 	if (kvm_x86_ops.tlb_remote_flush &&
1413 	    !static_call(kvm_x86_tlb_remote_flush)(kvm))
1414 		return 0;
1415 	else
1416 		return -ENOTSUPP;
1417 }
1418 
1419 int kvm_mmu_module_init(void);
1420 void kvm_mmu_module_exit(void);
1421 
1422 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
1423 int kvm_mmu_create(struct kvm_vcpu *vcpu);
1424 void kvm_mmu_init_vm(struct kvm *kvm);
1425 void kvm_mmu_uninit_vm(struct kvm *kvm);
1426 void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
1427 		u64 dirty_mask, u64 nx_mask, u64 x_mask, u64 p_mask,
1428 		u64 acc_track_mask, u64 me_mask);
1429 
1430 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
1431 void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
1432 				      struct kvm_memory_slot *memslot,
1433 				      int start_level);
1434 void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
1435 				   const struct kvm_memory_slot *memslot);
1436 void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
1437 				   struct kvm_memory_slot *memslot);
1438 void kvm_mmu_slot_largepage_remove_write_access(struct kvm *kvm,
1439 					struct kvm_memory_slot *memslot);
1440 void kvm_mmu_slot_set_dirty(struct kvm *kvm,
1441 			    struct kvm_memory_slot *memslot);
1442 void kvm_mmu_clear_dirty_pt_masked(struct kvm *kvm,
1443 				   struct kvm_memory_slot *slot,
1444 				   gfn_t gfn_offset, unsigned long mask);
1445 void kvm_mmu_zap_all(struct kvm *kvm);
1446 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, u64 gen);
1447 unsigned long kvm_mmu_calculate_default_mmu_pages(struct kvm *kvm);
1448 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned long kvm_nr_mmu_pages);
1449 
1450 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
1451 bool pdptrs_changed(struct kvm_vcpu *vcpu);
1452 
1453 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
1454 			  const void *val, int bytes);
1455 
1456 struct kvm_irq_mask_notifier {
1457 	void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
1458 	int irq;
1459 	struct hlist_node link;
1460 };
1461 
1462 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
1463 				    struct kvm_irq_mask_notifier *kimn);
1464 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
1465 				      struct kvm_irq_mask_notifier *kimn);
1466 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
1467 			     bool mask);
1468 
1469 extern bool tdp_enabled;
1470 
1471 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
1472 
1473 /* control of guest tsc rate supported? */
1474 extern bool kvm_has_tsc_control;
1475 /* maximum supported tsc_khz for guests */
1476 extern u32  kvm_max_guest_tsc_khz;
1477 /* number of bits of the fractional part of the TSC scaling ratio */
1478 extern u8   kvm_tsc_scaling_ratio_frac_bits;
1479 /* maximum allowed value of TSC scaling ratio */
1480 extern u64  kvm_max_tsc_scaling_ratio;
1481 /* 1ull << kvm_tsc_scaling_ratio_frac_bits */
1482 extern u64  kvm_default_tsc_scaling_ratio;
1483 /* bus lock detection supported? */
1484 extern bool kvm_has_bus_lock_exit;
1485 
1486 extern u64 kvm_mce_cap_supported;
1487 
1488 /*
1489  * EMULTYPE_NO_DECODE - Set when re-emulating an instruction (after completing
1490  *			userspace I/O) to indicate that the emulation context
1491  *			should be resued as is, i.e. skip initialization of
1492  *			emulation context, instruction fetch and decode.
1493  *
1494  * EMULTYPE_TRAP_UD - Set when emulating an intercepted #UD from hardware.
1495  *		      Indicates that only select instructions (tagged with
1496  *		      EmulateOnUD) should be emulated (to minimize the emulator
1497  *		      attack surface).  See also EMULTYPE_TRAP_UD_FORCED.
1498  *
1499  * EMULTYPE_SKIP - Set when emulating solely to skip an instruction, i.e. to
1500  *		   decode the instruction length.  For use *only* by
1501  *		   kvm_x86_ops.skip_emulated_instruction() implementations.
1502  *
1503  * EMULTYPE_ALLOW_RETRY_PF - Set when the emulator should resume the guest to
1504  *			     retry native execution under certain conditions,
1505  *			     Can only be set in conjunction with EMULTYPE_PF.
1506  *
1507  * EMULTYPE_TRAP_UD_FORCED - Set when emulating an intercepted #UD that was
1508  *			     triggered by KVM's magic "force emulation" prefix,
1509  *			     which is opt in via module param (off by default).
1510  *			     Bypasses EmulateOnUD restriction despite emulating
1511  *			     due to an intercepted #UD (see EMULTYPE_TRAP_UD).
1512  *			     Used to test the full emulator from userspace.
1513  *
1514  * EMULTYPE_VMWARE_GP - Set when emulating an intercepted #GP for VMware
1515  *			backdoor emulation, which is opt in via module param.
1516  *			VMware backoor emulation handles select instructions
1517  *			and reinjects the #GP for all other cases.
1518  *
1519  * EMULTYPE_PF - Set when emulating MMIO by way of an intercepted #PF, in which
1520  *		 case the CR2/GPA value pass on the stack is valid.
1521  */
1522 #define EMULTYPE_NO_DECODE	    (1 << 0)
1523 #define EMULTYPE_TRAP_UD	    (1 << 1)
1524 #define EMULTYPE_SKIP		    (1 << 2)
1525 #define EMULTYPE_ALLOW_RETRY_PF	    (1 << 3)
1526 #define EMULTYPE_TRAP_UD_FORCED	    (1 << 4)
1527 #define EMULTYPE_VMWARE_GP	    (1 << 5)
1528 #define EMULTYPE_PF		    (1 << 6)
1529 
1530 int kvm_emulate_instruction(struct kvm_vcpu *vcpu, int emulation_type);
1531 int kvm_emulate_instruction_from_buffer(struct kvm_vcpu *vcpu,
1532 					void *insn, int insn_len);
1533 
1534 void kvm_enable_efer_bits(u64);
1535 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
1536 int __kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data, bool host_initiated);
1537 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data);
1538 int kvm_set_msr(struct kvm_vcpu *vcpu, u32 index, u64 data);
1539 int kvm_emulate_rdmsr(struct kvm_vcpu *vcpu);
1540 int kvm_emulate_wrmsr(struct kvm_vcpu *vcpu);
1541 
1542 int kvm_fast_pio(struct kvm_vcpu *vcpu, int size, unsigned short port, int in);
1543 int kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
1544 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
1545 int kvm_vcpu_halt(struct kvm_vcpu *vcpu);
1546 int kvm_emulate_ap_reset_hold(struct kvm_vcpu *vcpu);
1547 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
1548 
1549 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
1550 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
1551 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector);
1552 
1553 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
1554 		    int reason, bool has_error_code, u32 error_code);
1555 
1556 void kvm_free_guest_fpu(struct kvm_vcpu *vcpu);
1557 
1558 void kvm_post_set_cr0(struct kvm_vcpu *vcpu, unsigned long old_cr0, unsigned long cr0);
1559 void kvm_post_set_cr4(struct kvm_vcpu *vcpu, unsigned long old_cr4, unsigned long cr4);
1560 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
1561 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
1562 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
1563 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
1564 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
1565 void kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
1566 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
1567 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
1568 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
1569 int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr);
1570 
1571 int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1572 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1573 
1574 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
1575 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
1576 bool kvm_rdpmc(struct kvm_vcpu *vcpu);
1577 
1578 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1579 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1580 void kvm_queue_exception_p(struct kvm_vcpu *vcpu, unsigned nr, unsigned long payload);
1581 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1582 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1583 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
1584 bool kvm_inject_emulated_page_fault(struct kvm_vcpu *vcpu,
1585 				    struct x86_exception *fault);
1586 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1587 			    gfn_t gfn, void *data, int offset, int len,
1588 			    u32 access);
1589 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
1590 bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr);
1591 
1592 static inline int __kvm_irq_line_state(unsigned long *irq_state,
1593 				       int irq_source_id, int level)
1594 {
1595 	/* Logical OR for level trig interrupt */
1596 	if (level)
1597 		__set_bit(irq_source_id, irq_state);
1598 	else
1599 		__clear_bit(irq_source_id, irq_state);
1600 
1601 	return !!(*irq_state);
1602 }
1603 
1604 #define KVM_MMU_ROOT_CURRENT		BIT(0)
1605 #define KVM_MMU_ROOT_PREVIOUS(i)	BIT(1+i)
1606 #define KVM_MMU_ROOTS_ALL		(~0UL)
1607 
1608 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
1609 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
1610 
1611 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
1612 
1613 void kvm_update_dr7(struct kvm_vcpu *vcpu);
1614 
1615 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
1616 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
1617 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
1618 int kvm_mmu_load(struct kvm_vcpu *vcpu);
1619 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
1620 void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu);
1621 void kvm_mmu_free_roots(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1622 			ulong roots_to_free);
1623 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1624 			   struct x86_exception *exception);
1625 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
1626 			      struct x86_exception *exception);
1627 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
1628 			       struct x86_exception *exception);
1629 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
1630 			       struct x86_exception *exception);
1631 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
1632 				struct x86_exception *exception);
1633 
1634 bool kvm_apicv_activated(struct kvm *kvm);
1635 void kvm_apicv_init(struct kvm *kvm, bool enable);
1636 void kvm_vcpu_update_apicv(struct kvm_vcpu *vcpu);
1637 void kvm_request_apicv_update(struct kvm *kvm, bool activate,
1638 			      unsigned long bit);
1639 
1640 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
1641 
1642 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u64 error_code,
1643 		       void *insn, int insn_len);
1644 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
1645 void kvm_mmu_invalidate_gva(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1646 			    gva_t gva, hpa_t root_hpa);
1647 void kvm_mmu_invpcid_gva(struct kvm_vcpu *vcpu, gva_t gva, unsigned long pcid);
1648 void kvm_mmu_new_pgd(struct kvm_vcpu *vcpu, gpa_t new_pgd, bool skip_tlb_flush,
1649 		     bool skip_mmu_sync);
1650 
1651 void kvm_configure_mmu(bool enable_tdp, int tdp_max_root_level,
1652 		       int tdp_huge_page_level);
1653 
1654 static inline u16 kvm_read_ldt(void)
1655 {
1656 	u16 ldt;
1657 	asm("sldt %0" : "=g"(ldt));
1658 	return ldt;
1659 }
1660 
1661 static inline void kvm_load_ldt(u16 sel)
1662 {
1663 	asm("lldt %0" : : "rm"(sel));
1664 }
1665 
1666 #ifdef CONFIG_X86_64
1667 static inline unsigned long read_msr(unsigned long msr)
1668 {
1669 	u64 value;
1670 
1671 	rdmsrl(msr, value);
1672 	return value;
1673 }
1674 #endif
1675 
1676 static inline u32 get_rdx_init_val(void)
1677 {
1678 	return 0x600; /* P6 family */
1679 }
1680 
1681 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
1682 {
1683 	kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
1684 }
1685 
1686 #define TSS_IOPB_BASE_OFFSET 0x66
1687 #define TSS_BASE_SIZE 0x68
1688 #define TSS_IOPB_SIZE (65536 / 8)
1689 #define TSS_REDIRECTION_SIZE (256 / 8)
1690 #define RMODE_TSS_SIZE							\
1691 	(TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
1692 
1693 enum {
1694 	TASK_SWITCH_CALL = 0,
1695 	TASK_SWITCH_IRET = 1,
1696 	TASK_SWITCH_JMP = 2,
1697 	TASK_SWITCH_GATE = 3,
1698 };
1699 
1700 #define HF_GIF_MASK		(1 << 0)
1701 #define HF_NMI_MASK		(1 << 3)
1702 #define HF_IRET_MASK		(1 << 4)
1703 #define HF_GUEST_MASK		(1 << 5) /* VCPU is in guest-mode */
1704 #define HF_SMM_MASK		(1 << 6)
1705 #define HF_SMM_INSIDE_NMI_MASK	(1 << 7)
1706 
1707 #define __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
1708 #define KVM_ADDRESS_SPACE_NUM 2
1709 
1710 #define kvm_arch_vcpu_memslots_id(vcpu) ((vcpu)->arch.hflags & HF_SMM_MASK ? 1 : 0)
1711 #define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, (role).smm)
1712 
1713 asmlinkage void kvm_spurious_fault(void);
1714 
1715 /*
1716  * Hardware virtualization extension instructions may fault if a
1717  * reboot turns off virtualization while processes are running.
1718  * Usually after catching the fault we just panic; during reboot
1719  * instead the instruction is ignored.
1720  */
1721 #define __kvm_handle_fault_on_reboot(insn)				\
1722 	"666: \n\t"							\
1723 	insn "\n\t"							\
1724 	"jmp	668f \n\t"						\
1725 	"667: \n\t"							\
1726 	"1: \n\t"							\
1727 	".pushsection .discard.instr_begin \n\t"			\
1728 	".long 1b - . \n\t"						\
1729 	".popsection \n\t"						\
1730 	"call	kvm_spurious_fault \n\t"				\
1731 	"1: \n\t"							\
1732 	".pushsection .discard.instr_end \n\t"				\
1733 	".long 1b - . \n\t"						\
1734 	".popsection \n\t"						\
1735 	"668: \n\t"							\
1736 	_ASM_EXTABLE(666b, 667b)
1737 
1738 #define KVM_ARCH_WANT_MMU_NOTIFIER
1739 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end,
1740 			unsigned flags);
1741 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
1742 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
1743 int kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
1744 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1745 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1746 int kvm_cpu_has_extint(struct kvm_vcpu *v);
1747 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1748 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1749 void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event);
1750 void kvm_vcpu_reload_apic_access_page(struct kvm_vcpu *vcpu);
1751 
1752 int kvm_pv_send_ipi(struct kvm *kvm, unsigned long ipi_bitmap_low,
1753 		    unsigned long ipi_bitmap_high, u32 min,
1754 		    unsigned long icr, int op_64_bit);
1755 
1756 void kvm_define_user_return_msr(unsigned index, u32 msr);
1757 int kvm_set_user_return_msr(unsigned index, u64 val, u64 mask);
1758 
1759 u64 kvm_scale_tsc(struct kvm_vcpu *vcpu, u64 tsc);
1760 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc);
1761 
1762 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu);
1763 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1764 
1765 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
1766 void kvm_make_scan_ioapic_request(struct kvm *kvm);
1767 void kvm_make_scan_ioapic_request_mask(struct kvm *kvm,
1768 				       unsigned long *vcpu_bitmap);
1769 
1770 bool kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1771 				     struct kvm_async_pf *work);
1772 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1773 				 struct kvm_async_pf *work);
1774 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1775 			       struct kvm_async_pf *work);
1776 void kvm_arch_async_page_present_queued(struct kvm_vcpu *vcpu);
1777 bool kvm_arch_can_dequeue_async_page_present(struct kvm_vcpu *vcpu);
1778 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1779 
1780 int kvm_skip_emulated_instruction(struct kvm_vcpu *vcpu);
1781 int kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1782 void __kvm_request_immediate_exit(struct kvm_vcpu *vcpu);
1783 
1784 int kvm_is_in_guest(void);
1785 
1786 void __user *__x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa,
1787 				     u32 size);
1788 bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu);
1789 bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu);
1790 
1791 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
1792 			     struct kvm_vcpu **dest_vcpu);
1793 
1794 void kvm_set_msi_irq(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e,
1795 		     struct kvm_lapic_irq *irq);
1796 
1797 static inline bool kvm_irq_is_postable(struct kvm_lapic_irq *irq)
1798 {
1799 	/* We can only post Fixed and LowPrio IRQs */
1800 	return (irq->delivery_mode == APIC_DM_FIXED ||
1801 		irq->delivery_mode == APIC_DM_LOWEST);
1802 }
1803 
1804 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu)
1805 {
1806 	static_call_cond(kvm_x86_vcpu_blocking)(vcpu);
1807 }
1808 
1809 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu)
1810 {
1811 	static_call_cond(kvm_x86_vcpu_unblocking)(vcpu);
1812 }
1813 
1814 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
1815 
1816 static inline int kvm_cpu_get_apicid(int mps_cpu)
1817 {
1818 #ifdef CONFIG_X86_LOCAL_APIC
1819 	return default_cpu_present_to_apicid(mps_cpu);
1820 #else
1821 	WARN_ON_ONCE(1);
1822 	return BAD_APICID;
1823 #endif
1824 }
1825 
1826 #define put_smstate(type, buf, offset, val)                      \
1827 	*(type *)((buf) + (offset) - 0x7e00) = val
1828 
1829 #define GET_SMSTATE(type, buf, offset)		\
1830 	(*(type *)((buf) + (offset) - 0x7e00))
1831 
1832 int kvm_cpu_dirty_log_size(void);
1833 
1834 #endif /* _ASM_X86_KVM_HOST_H */
1835