xref: /openbmc/linux/arch/x86/include/asm/kvm_host.h (revision 15e3ae36)
1 /* SPDX-License-Identifier: GPL-2.0-only */
2 /*
3  * Kernel-based Virtual Machine driver for Linux
4  *
5  * This header defines architecture specific interfaces, x86 version
6  */
7 
8 #ifndef _ASM_X86_KVM_HOST_H
9 #define _ASM_X86_KVM_HOST_H
10 
11 #include <linux/types.h>
12 #include <linux/mm.h>
13 #include <linux/mmu_notifier.h>
14 #include <linux/tracepoint.h>
15 #include <linux/cpumask.h>
16 #include <linux/irq_work.h>
17 #include <linux/irq.h>
18 
19 #include <linux/kvm.h>
20 #include <linux/kvm_para.h>
21 #include <linux/kvm_types.h>
22 #include <linux/perf_event.h>
23 #include <linux/pvclock_gtod.h>
24 #include <linux/clocksource.h>
25 #include <linux/irqbypass.h>
26 #include <linux/hyperv.h>
27 
28 #include <asm/apic.h>
29 #include <asm/pvclock-abi.h>
30 #include <asm/desc.h>
31 #include <asm/mtrr.h>
32 #include <asm/msr-index.h>
33 #include <asm/asm.h>
34 #include <asm/kvm_page_track.h>
35 #include <asm/kvm_vcpu_regs.h>
36 #include <asm/hyperv-tlfs.h>
37 
38 #define __KVM_HAVE_ARCH_VCPU_DEBUGFS
39 
40 #define KVM_MAX_VCPUS 288
41 #define KVM_SOFT_MAX_VCPUS 240
42 #define KVM_MAX_VCPU_ID 1023
43 #define KVM_USER_MEM_SLOTS 509
44 /* memory slots that are not exposed to userspace */
45 #define KVM_PRIVATE_MEM_SLOTS 3
46 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
47 
48 #define KVM_HALT_POLL_NS_DEFAULT 200000
49 
50 #define KVM_IRQCHIP_NUM_PINS  KVM_IOAPIC_NUM_PINS
51 
52 #define KVM_DIRTY_LOG_MANUAL_CAPS   (KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE | \
53 					KVM_DIRTY_LOG_INITIALLY_SET)
54 
55 /* x86-specific vcpu->requests bit members */
56 #define KVM_REQ_MIGRATE_TIMER		KVM_ARCH_REQ(0)
57 #define KVM_REQ_REPORT_TPR_ACCESS	KVM_ARCH_REQ(1)
58 #define KVM_REQ_TRIPLE_FAULT		KVM_ARCH_REQ(2)
59 #define KVM_REQ_MMU_SYNC		KVM_ARCH_REQ(3)
60 #define KVM_REQ_CLOCK_UPDATE		KVM_ARCH_REQ(4)
61 #define KVM_REQ_LOAD_MMU_PGD		KVM_ARCH_REQ(5)
62 #define KVM_REQ_EVENT			KVM_ARCH_REQ(6)
63 #define KVM_REQ_APF_HALT		KVM_ARCH_REQ(7)
64 #define KVM_REQ_STEAL_UPDATE		KVM_ARCH_REQ(8)
65 #define KVM_REQ_NMI			KVM_ARCH_REQ(9)
66 #define KVM_REQ_PMU			KVM_ARCH_REQ(10)
67 #define KVM_REQ_PMI			KVM_ARCH_REQ(11)
68 #define KVM_REQ_SMI			KVM_ARCH_REQ(12)
69 #define KVM_REQ_MASTERCLOCK_UPDATE	KVM_ARCH_REQ(13)
70 #define KVM_REQ_MCLOCK_INPROGRESS \
71 	KVM_ARCH_REQ_FLAGS(14, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
72 #define KVM_REQ_SCAN_IOAPIC \
73 	KVM_ARCH_REQ_FLAGS(15, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
74 #define KVM_REQ_GLOBAL_CLOCK_UPDATE	KVM_ARCH_REQ(16)
75 #define KVM_REQ_APIC_PAGE_RELOAD \
76 	KVM_ARCH_REQ_FLAGS(17, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
77 #define KVM_REQ_HV_CRASH		KVM_ARCH_REQ(18)
78 #define KVM_REQ_IOAPIC_EOI_EXIT		KVM_ARCH_REQ(19)
79 #define KVM_REQ_HV_RESET		KVM_ARCH_REQ(20)
80 #define KVM_REQ_HV_EXIT			KVM_ARCH_REQ(21)
81 #define KVM_REQ_HV_STIMER		KVM_ARCH_REQ(22)
82 #define KVM_REQ_LOAD_EOI_EXITMAP	KVM_ARCH_REQ(23)
83 #define KVM_REQ_GET_VMCS12_PAGES	KVM_ARCH_REQ(24)
84 #define KVM_REQ_APICV_UPDATE \
85 	KVM_ARCH_REQ_FLAGS(25, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
86 
87 #define CR0_RESERVED_BITS                                               \
88 	(~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \
89 			  | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \
90 			  | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG))
91 
92 #define CR4_RESERVED_BITS                                               \
93 	(~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\
94 			  | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE     \
95 			  | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR | X86_CR4_PCIDE \
96 			  | X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_FSGSBASE \
97 			  | X86_CR4_OSXMMEXCPT | X86_CR4_LA57 | X86_CR4_VMXE \
98 			  | X86_CR4_SMAP | X86_CR4_PKE | X86_CR4_UMIP))
99 
100 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR)
101 
102 
103 
104 #define INVALID_PAGE (~(hpa_t)0)
105 #define VALID_PAGE(x) ((x) != INVALID_PAGE)
106 
107 #define UNMAPPED_GVA (~(gpa_t)0)
108 
109 /* KVM Hugepage definitions for x86 */
110 enum {
111 	PT_PAGE_TABLE_LEVEL   = 1,
112 	PT_DIRECTORY_LEVEL    = 2,
113 	PT_PDPE_LEVEL         = 3,
114 	/* set max level to the biggest one */
115 	PT_MAX_HUGEPAGE_LEVEL = PT_PDPE_LEVEL,
116 };
117 #define KVM_NR_PAGE_SIZES	(PT_MAX_HUGEPAGE_LEVEL - \
118 				 PT_PAGE_TABLE_LEVEL + 1)
119 #define KVM_HPAGE_GFN_SHIFT(x)	(((x) - 1) * 9)
120 #define KVM_HPAGE_SHIFT(x)	(PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
121 #define KVM_HPAGE_SIZE(x)	(1UL << KVM_HPAGE_SHIFT(x))
122 #define KVM_HPAGE_MASK(x)	(~(KVM_HPAGE_SIZE(x) - 1))
123 #define KVM_PAGES_PER_HPAGE(x)	(KVM_HPAGE_SIZE(x) / PAGE_SIZE)
124 
125 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
126 {
127 	/* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */
128 	return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
129 		(base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
130 }
131 
132 #define KVM_PERMILLE_MMU_PAGES 20
133 #define KVM_MIN_ALLOC_MMU_PAGES 64UL
134 #define KVM_MMU_HASH_SHIFT 12
135 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
136 #define KVM_MIN_FREE_MMU_PAGES 5
137 #define KVM_REFILL_PAGES 25
138 #define KVM_MAX_CPUID_ENTRIES 80
139 #define KVM_NR_FIXED_MTRR_REGION 88
140 #define KVM_NR_VAR_MTRR 8
141 
142 #define ASYNC_PF_PER_VCPU 64
143 
144 enum kvm_reg {
145 	VCPU_REGS_RAX = __VCPU_REGS_RAX,
146 	VCPU_REGS_RCX = __VCPU_REGS_RCX,
147 	VCPU_REGS_RDX = __VCPU_REGS_RDX,
148 	VCPU_REGS_RBX = __VCPU_REGS_RBX,
149 	VCPU_REGS_RSP = __VCPU_REGS_RSP,
150 	VCPU_REGS_RBP = __VCPU_REGS_RBP,
151 	VCPU_REGS_RSI = __VCPU_REGS_RSI,
152 	VCPU_REGS_RDI = __VCPU_REGS_RDI,
153 #ifdef CONFIG_X86_64
154 	VCPU_REGS_R8  = __VCPU_REGS_R8,
155 	VCPU_REGS_R9  = __VCPU_REGS_R9,
156 	VCPU_REGS_R10 = __VCPU_REGS_R10,
157 	VCPU_REGS_R11 = __VCPU_REGS_R11,
158 	VCPU_REGS_R12 = __VCPU_REGS_R12,
159 	VCPU_REGS_R13 = __VCPU_REGS_R13,
160 	VCPU_REGS_R14 = __VCPU_REGS_R14,
161 	VCPU_REGS_R15 = __VCPU_REGS_R15,
162 #endif
163 	VCPU_REGS_RIP,
164 	NR_VCPU_REGS,
165 
166 	VCPU_EXREG_PDPTR = NR_VCPU_REGS,
167 	VCPU_EXREG_CR3,
168 	VCPU_EXREG_RFLAGS,
169 	VCPU_EXREG_SEGMENTS,
170 };
171 
172 enum {
173 	VCPU_SREG_ES,
174 	VCPU_SREG_CS,
175 	VCPU_SREG_SS,
176 	VCPU_SREG_DS,
177 	VCPU_SREG_FS,
178 	VCPU_SREG_GS,
179 	VCPU_SREG_TR,
180 	VCPU_SREG_LDTR,
181 };
182 
183 enum exit_fastpath_completion {
184 	EXIT_FASTPATH_NONE,
185 	EXIT_FASTPATH_SKIP_EMUL_INS,
186 };
187 
188 struct x86_emulate_ctxt;
189 struct x86_exception;
190 enum x86_intercept;
191 enum x86_intercept_stage;
192 
193 #define KVM_NR_MEM_OBJS 40
194 
195 #define KVM_NR_DB_REGS	4
196 
197 #define DR6_BD		(1 << 13)
198 #define DR6_BS		(1 << 14)
199 #define DR6_BT		(1 << 15)
200 #define DR6_RTM		(1 << 16)
201 #define DR6_FIXED_1	0xfffe0ff0
202 #define DR6_INIT	0xffff0ff0
203 #define DR6_VOLATILE	0x0001e00f
204 
205 #define DR7_BP_EN_MASK	0x000000ff
206 #define DR7_GE		(1 << 9)
207 #define DR7_GD		(1 << 13)
208 #define DR7_FIXED_1	0x00000400
209 #define DR7_VOLATILE	0xffff2bff
210 
211 #define PFERR_PRESENT_BIT 0
212 #define PFERR_WRITE_BIT 1
213 #define PFERR_USER_BIT 2
214 #define PFERR_RSVD_BIT 3
215 #define PFERR_FETCH_BIT 4
216 #define PFERR_PK_BIT 5
217 #define PFERR_GUEST_FINAL_BIT 32
218 #define PFERR_GUEST_PAGE_BIT 33
219 
220 #define PFERR_PRESENT_MASK (1U << PFERR_PRESENT_BIT)
221 #define PFERR_WRITE_MASK (1U << PFERR_WRITE_BIT)
222 #define PFERR_USER_MASK (1U << PFERR_USER_BIT)
223 #define PFERR_RSVD_MASK (1U << PFERR_RSVD_BIT)
224 #define PFERR_FETCH_MASK (1U << PFERR_FETCH_BIT)
225 #define PFERR_PK_MASK (1U << PFERR_PK_BIT)
226 #define PFERR_GUEST_FINAL_MASK (1ULL << PFERR_GUEST_FINAL_BIT)
227 #define PFERR_GUEST_PAGE_MASK (1ULL << PFERR_GUEST_PAGE_BIT)
228 
229 #define PFERR_NESTED_GUEST_PAGE (PFERR_GUEST_PAGE_MASK |	\
230 				 PFERR_WRITE_MASK |		\
231 				 PFERR_PRESENT_MASK)
232 
233 /* apic attention bits */
234 #define KVM_APIC_CHECK_VAPIC	0
235 /*
236  * The following bit is set with PV-EOI, unset on EOI.
237  * We detect PV-EOI changes by guest by comparing
238  * this bit with PV-EOI in guest memory.
239  * See the implementation in apic_update_pv_eoi.
240  */
241 #define KVM_APIC_PV_EOI_PENDING	1
242 
243 struct kvm_kernel_irq_routing_entry;
244 
245 /*
246  * We don't want allocation failures within the mmu code, so we preallocate
247  * enough memory for a single page fault in a cache.
248  */
249 struct kvm_mmu_memory_cache {
250 	int nobjs;
251 	void *objects[KVM_NR_MEM_OBJS];
252 };
253 
254 /*
255  * the pages used as guest page table on soft mmu are tracked by
256  * kvm_memory_slot.arch.gfn_track which is 16 bits, so the role bits used
257  * by indirect shadow page can not be more than 15 bits.
258  *
259  * Currently, we used 14 bits that are @level, @gpte_is_8_bytes, @quadrant, @access,
260  * @nxe, @cr0_wp, @smep_andnot_wp and @smap_andnot_wp.
261  */
262 union kvm_mmu_page_role {
263 	u32 word;
264 	struct {
265 		unsigned level:4;
266 		unsigned gpte_is_8_bytes:1;
267 		unsigned quadrant:2;
268 		unsigned direct:1;
269 		unsigned access:3;
270 		unsigned invalid:1;
271 		unsigned nxe:1;
272 		unsigned cr0_wp:1;
273 		unsigned smep_andnot_wp:1;
274 		unsigned smap_andnot_wp:1;
275 		unsigned ad_disabled:1;
276 		unsigned guest_mode:1;
277 		unsigned :6;
278 
279 		/*
280 		 * This is left at the top of the word so that
281 		 * kvm_memslots_for_spte_role can extract it with a
282 		 * simple shift.  While there is room, give it a whole
283 		 * byte so it is also faster to load it from memory.
284 		 */
285 		unsigned smm:8;
286 	};
287 };
288 
289 union kvm_mmu_extended_role {
290 /*
291  * This structure complements kvm_mmu_page_role caching everything needed for
292  * MMU configuration. If nothing in both these structures changed, MMU
293  * re-configuration can be skipped. @valid bit is set on first usage so we don't
294  * treat all-zero structure as valid data.
295  */
296 	u32 word;
297 	struct {
298 		unsigned int valid:1;
299 		unsigned int execonly:1;
300 		unsigned int cr0_pg:1;
301 		unsigned int cr4_pae:1;
302 		unsigned int cr4_pse:1;
303 		unsigned int cr4_pke:1;
304 		unsigned int cr4_smap:1;
305 		unsigned int cr4_smep:1;
306 		unsigned int maxphyaddr:6;
307 	};
308 };
309 
310 union kvm_mmu_role {
311 	u64 as_u64;
312 	struct {
313 		union kvm_mmu_page_role base;
314 		union kvm_mmu_extended_role ext;
315 	};
316 };
317 
318 struct kvm_rmap_head {
319 	unsigned long val;
320 };
321 
322 struct kvm_mmu_page {
323 	struct list_head link;
324 	struct hlist_node hash_link;
325 	struct list_head lpage_disallowed_link;
326 
327 	bool unsync;
328 	u8 mmu_valid_gen;
329 	bool mmio_cached;
330 	bool lpage_disallowed; /* Can't be replaced by an equiv large page */
331 
332 	/*
333 	 * The following two entries are used to key the shadow page in the
334 	 * hash table.
335 	 */
336 	union kvm_mmu_page_role role;
337 	gfn_t gfn;
338 
339 	u64 *spt;
340 	/* hold the gfn of each spte inside spt */
341 	gfn_t *gfns;
342 	int root_count;          /* Currently serving as active root */
343 	unsigned int unsync_children;
344 	struct kvm_rmap_head parent_ptes; /* rmap pointers to parent sptes */
345 	DECLARE_BITMAP(unsync_child_bitmap, 512);
346 
347 #ifdef CONFIG_X86_32
348 	/*
349 	 * Used out of the mmu-lock to avoid reading spte values while an
350 	 * update is in progress; see the comments in __get_spte_lockless().
351 	 */
352 	int clear_spte_count;
353 #endif
354 
355 	/* Number of writes since the last time traversal visited this page.  */
356 	atomic_t write_flooding_count;
357 };
358 
359 struct kvm_pio_request {
360 	unsigned long linear_rip;
361 	unsigned long count;
362 	int in;
363 	int port;
364 	int size;
365 };
366 
367 #define PT64_ROOT_MAX_LEVEL 5
368 
369 struct rsvd_bits_validate {
370 	u64 rsvd_bits_mask[2][PT64_ROOT_MAX_LEVEL];
371 	u64 bad_mt_xwr;
372 };
373 
374 struct kvm_mmu_root_info {
375 	gpa_t cr3;
376 	hpa_t hpa;
377 };
378 
379 #define KVM_MMU_ROOT_INFO_INVALID \
380 	((struct kvm_mmu_root_info) { .cr3 = INVALID_PAGE, .hpa = INVALID_PAGE })
381 
382 #define KVM_MMU_NUM_PREV_ROOTS 3
383 
384 /*
385  * x86 supports 4 paging modes (5-level 64-bit, 4-level 64-bit, 3-level 32-bit,
386  * and 2-level 32-bit).  The kvm_mmu structure abstracts the details of the
387  * current mmu mode.
388  */
389 struct kvm_mmu {
390 	unsigned long (*get_guest_pgd)(struct kvm_vcpu *vcpu);
391 	u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
392 	int (*page_fault)(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u32 err,
393 			  bool prefault);
394 	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
395 				  struct x86_exception *fault);
396 	gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gpa_t gva_or_gpa,
397 			    u32 access, struct x86_exception *exception);
398 	gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
399 			       struct x86_exception *exception);
400 	int (*sync_page)(struct kvm_vcpu *vcpu,
401 			 struct kvm_mmu_page *sp);
402 	void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva, hpa_t root_hpa);
403 	void (*update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
404 			   u64 *spte, const void *pte);
405 	hpa_t root_hpa;
406 	gpa_t root_cr3;
407 	union kvm_mmu_role mmu_role;
408 	u8 root_level;
409 	u8 shadow_root_level;
410 	u8 ept_ad;
411 	bool direct_map;
412 	struct kvm_mmu_root_info prev_roots[KVM_MMU_NUM_PREV_ROOTS];
413 
414 	/*
415 	 * Bitmap; bit set = permission fault
416 	 * Byte index: page fault error code [4:1]
417 	 * Bit index: pte permissions in ACC_* format
418 	 */
419 	u8 permissions[16];
420 
421 	/*
422 	* The pkru_mask indicates if protection key checks are needed.  It
423 	* consists of 16 domains indexed by page fault error code bits [4:1],
424 	* with PFEC.RSVD replaced by ACC_USER_MASK from the page tables.
425 	* Each domain has 2 bits which are ANDed with AD and WD from PKRU.
426 	*/
427 	u32 pkru_mask;
428 
429 	u64 *pae_root;
430 	u64 *lm_root;
431 
432 	/*
433 	 * check zero bits on shadow page table entries, these
434 	 * bits include not only hardware reserved bits but also
435 	 * the bits spte never used.
436 	 */
437 	struct rsvd_bits_validate shadow_zero_check;
438 
439 	struct rsvd_bits_validate guest_rsvd_check;
440 
441 	/* Can have large pages at levels 2..last_nonleaf_level-1. */
442 	u8 last_nonleaf_level;
443 
444 	bool nx;
445 
446 	u64 pdptrs[4]; /* pae */
447 };
448 
449 struct kvm_tlb_range {
450 	u64 start_gfn;
451 	u64 pages;
452 };
453 
454 enum pmc_type {
455 	KVM_PMC_GP = 0,
456 	KVM_PMC_FIXED,
457 };
458 
459 struct kvm_pmc {
460 	enum pmc_type type;
461 	u8 idx;
462 	u64 counter;
463 	u64 eventsel;
464 	struct perf_event *perf_event;
465 	struct kvm_vcpu *vcpu;
466 	/*
467 	 * eventsel value for general purpose counters,
468 	 * ctrl value for fixed counters.
469 	 */
470 	u64 current_config;
471 };
472 
473 struct kvm_pmu {
474 	unsigned nr_arch_gp_counters;
475 	unsigned nr_arch_fixed_counters;
476 	unsigned available_event_types;
477 	u64 fixed_ctr_ctrl;
478 	u64 global_ctrl;
479 	u64 global_status;
480 	u64 global_ovf_ctrl;
481 	u64 counter_bitmask[2];
482 	u64 global_ctrl_mask;
483 	u64 global_ovf_ctrl_mask;
484 	u64 reserved_bits;
485 	u8 version;
486 	struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
487 	struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED];
488 	struct irq_work irq_work;
489 	DECLARE_BITMAP(reprogram_pmi, X86_PMC_IDX_MAX);
490 	DECLARE_BITMAP(all_valid_pmc_idx, X86_PMC_IDX_MAX);
491 	DECLARE_BITMAP(pmc_in_use, X86_PMC_IDX_MAX);
492 
493 	/*
494 	 * The gate to release perf_events not marked in
495 	 * pmc_in_use only once in a vcpu time slice.
496 	 */
497 	bool need_cleanup;
498 
499 	/*
500 	 * The total number of programmed perf_events and it helps to avoid
501 	 * redundant check before cleanup if guest don't use vPMU at all.
502 	 */
503 	u8 event_count;
504 };
505 
506 struct kvm_pmu_ops;
507 
508 enum {
509 	KVM_DEBUGREG_BP_ENABLED = 1,
510 	KVM_DEBUGREG_WONT_EXIT = 2,
511 	KVM_DEBUGREG_RELOAD = 4,
512 };
513 
514 struct kvm_mtrr_range {
515 	u64 base;
516 	u64 mask;
517 	struct list_head node;
518 };
519 
520 struct kvm_mtrr {
521 	struct kvm_mtrr_range var_ranges[KVM_NR_VAR_MTRR];
522 	mtrr_type fixed_ranges[KVM_NR_FIXED_MTRR_REGION];
523 	u64 deftype;
524 
525 	struct list_head head;
526 };
527 
528 /* Hyper-V SynIC timer */
529 struct kvm_vcpu_hv_stimer {
530 	struct hrtimer timer;
531 	int index;
532 	union hv_stimer_config config;
533 	u64 count;
534 	u64 exp_time;
535 	struct hv_message msg;
536 	bool msg_pending;
537 };
538 
539 /* Hyper-V synthetic interrupt controller (SynIC)*/
540 struct kvm_vcpu_hv_synic {
541 	u64 version;
542 	u64 control;
543 	u64 msg_page;
544 	u64 evt_page;
545 	atomic64_t sint[HV_SYNIC_SINT_COUNT];
546 	atomic_t sint_to_gsi[HV_SYNIC_SINT_COUNT];
547 	DECLARE_BITMAP(auto_eoi_bitmap, 256);
548 	DECLARE_BITMAP(vec_bitmap, 256);
549 	bool active;
550 	bool dont_zero_synic_pages;
551 };
552 
553 /* Hyper-V per vcpu emulation context */
554 struct kvm_vcpu_hv {
555 	u32 vp_index;
556 	u64 hv_vapic;
557 	s64 runtime_offset;
558 	struct kvm_vcpu_hv_synic synic;
559 	struct kvm_hyperv_exit exit;
560 	struct kvm_vcpu_hv_stimer stimer[HV_SYNIC_STIMER_COUNT];
561 	DECLARE_BITMAP(stimer_pending_bitmap, HV_SYNIC_STIMER_COUNT);
562 	cpumask_t tlb_flush;
563 };
564 
565 struct kvm_vcpu_arch {
566 	/*
567 	 * rip and regs accesses must go through
568 	 * kvm_{register,rip}_{read,write} functions.
569 	 */
570 	unsigned long regs[NR_VCPU_REGS];
571 	u32 regs_avail;
572 	u32 regs_dirty;
573 
574 	unsigned long cr0;
575 	unsigned long cr0_guest_owned_bits;
576 	unsigned long cr2;
577 	unsigned long cr3;
578 	unsigned long cr4;
579 	unsigned long cr4_guest_owned_bits;
580 	unsigned long cr8;
581 	u32 pkru;
582 	u32 hflags;
583 	u64 efer;
584 	u64 apic_base;
585 	struct kvm_lapic *apic;    /* kernel irqchip context */
586 	bool apicv_active;
587 	bool load_eoi_exitmap_pending;
588 	DECLARE_BITMAP(ioapic_handled_vectors, 256);
589 	unsigned long apic_attention;
590 	int32_t apic_arb_prio;
591 	int mp_state;
592 	u64 ia32_misc_enable_msr;
593 	u64 smbase;
594 	u64 smi_count;
595 	bool tpr_access_reporting;
596 	bool xsaves_enabled;
597 	u64 ia32_xss;
598 	u64 microcode_version;
599 	u64 arch_capabilities;
600 
601 	/*
602 	 * Paging state of the vcpu
603 	 *
604 	 * If the vcpu runs in guest mode with two level paging this still saves
605 	 * the paging mode of the l1 guest. This context is always used to
606 	 * handle faults.
607 	 */
608 	struct kvm_mmu *mmu;
609 
610 	/* Non-nested MMU for L1 */
611 	struct kvm_mmu root_mmu;
612 
613 	/* L1 MMU when running nested */
614 	struct kvm_mmu guest_mmu;
615 
616 	/*
617 	 * Paging state of an L2 guest (used for nested npt)
618 	 *
619 	 * This context will save all necessary information to walk page tables
620 	 * of an L2 guest. This context is only initialized for page table
621 	 * walking and not for faulting since we never handle l2 page faults on
622 	 * the host.
623 	 */
624 	struct kvm_mmu nested_mmu;
625 
626 	/*
627 	 * Pointer to the mmu context currently used for
628 	 * gva_to_gpa translations.
629 	 */
630 	struct kvm_mmu *walk_mmu;
631 
632 	struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
633 	struct kvm_mmu_memory_cache mmu_page_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 	u32 guest_xstate_size;
653 
654 	struct kvm_pio_request pio;
655 	void *pio_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[KVM_MAX_CPUID_ENTRIES];
680 
681 	int maxphyaddr;
682 
683 	/* emulate context */
684 
685 	struct x86_emulate_ctxt *emulate_ctxt;
686 	bool emulate_regs_need_sync_to_vcpu;
687 	bool emulate_regs_need_sync_from_vcpu;
688 	int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
689 
690 	gpa_t time;
691 	struct pvclock_vcpu_time_info hv_clock;
692 	unsigned int hw_tsc_khz;
693 	struct gfn_to_hva_cache pv_time;
694 	bool pv_time_enabled;
695 	/* set guest stopped flag in pvclock flags field */
696 	bool pvclock_set_guest_stopped_request;
697 
698 	struct {
699 		u8 preempted;
700 		u64 msr_val;
701 		u64 last_steal;
702 		struct gfn_to_pfn_cache cache;
703 	} st;
704 
705 	u64 tsc_offset;
706 	u64 last_guest_tsc;
707 	u64 last_host_tsc;
708 	u64 tsc_offset_adjustment;
709 	u64 this_tsc_nsec;
710 	u64 this_tsc_write;
711 	u64 this_tsc_generation;
712 	bool tsc_catchup;
713 	bool tsc_always_catchup;
714 	s8 virtual_tsc_shift;
715 	u32 virtual_tsc_mult;
716 	u32 virtual_tsc_khz;
717 	s64 ia32_tsc_adjust_msr;
718 	u64 msr_ia32_power_ctl;
719 	u64 tsc_scaling_ratio;
720 
721 	atomic_t nmi_queued;  /* unprocessed asynchronous NMIs */
722 	unsigned nmi_pending; /* NMI queued after currently running handler */
723 	bool nmi_injected;    /* Trying to inject an NMI this entry */
724 	bool smi_pending;    /* SMI queued after currently running handler */
725 
726 	struct kvm_mtrr mtrr_state;
727 	u64 pat;
728 
729 	unsigned switch_db_regs;
730 	unsigned long db[KVM_NR_DB_REGS];
731 	unsigned long dr6;
732 	unsigned long dr7;
733 	unsigned long eff_db[KVM_NR_DB_REGS];
734 	unsigned long guest_debug_dr7;
735 	u64 msr_platform_info;
736 	u64 msr_misc_features_enables;
737 
738 	u64 mcg_cap;
739 	u64 mcg_status;
740 	u64 mcg_ctl;
741 	u64 mcg_ext_ctl;
742 	u64 *mce_banks;
743 
744 	/* Cache MMIO info */
745 	u64 mmio_gva;
746 	unsigned mmio_access;
747 	gfn_t mmio_gfn;
748 	u64 mmio_gen;
749 
750 	struct kvm_pmu pmu;
751 
752 	/* used for guest single stepping over the given code position */
753 	unsigned long singlestep_rip;
754 
755 	struct kvm_vcpu_hv hyperv;
756 
757 	cpumask_var_t wbinvd_dirty_mask;
758 
759 	unsigned long last_retry_eip;
760 	unsigned long last_retry_addr;
761 
762 	struct {
763 		bool halted;
764 		gfn_t gfns[roundup_pow_of_two(ASYNC_PF_PER_VCPU)];
765 		struct gfn_to_hva_cache data;
766 		u64 msr_val;
767 		u32 id;
768 		bool send_user_only;
769 		u32 host_apf_reason;
770 		unsigned long nested_apf_token;
771 		bool delivery_as_pf_vmexit;
772 	} apf;
773 
774 	/* OSVW MSRs (AMD only) */
775 	struct {
776 		u64 length;
777 		u64 status;
778 	} osvw;
779 
780 	struct {
781 		u64 msr_val;
782 		struct gfn_to_hva_cache data;
783 	} pv_eoi;
784 
785 	u64 msr_kvm_poll_control;
786 
787 	/*
788 	 * Indicates the guest is trying to write a gfn that contains one or
789 	 * more of the PTEs used to translate the write itself, i.e. the access
790 	 * is changing its own translation in the guest page tables.  KVM exits
791 	 * to userspace if emulation of the faulting instruction fails and this
792 	 * flag is set, as KVM cannot make forward progress.
793 	 *
794 	 * If emulation fails for a write to guest page tables, KVM unprotects
795 	 * (zaps) the shadow page for the target gfn and resumes the guest to
796 	 * retry the non-emulatable instruction (on hardware).  Unprotecting the
797 	 * gfn doesn't allow forward progress for a self-changing access because
798 	 * doing so also zaps the translation for the gfn, i.e. retrying the
799 	 * instruction will hit a !PRESENT fault, which results in a new shadow
800 	 * page and sends KVM back to square one.
801 	 */
802 	bool write_fault_to_shadow_pgtable;
803 
804 	/* set at EPT violation at this point */
805 	unsigned long exit_qualification;
806 
807 	/* pv related host specific info */
808 	struct {
809 		bool pv_unhalted;
810 	} pv;
811 
812 	int pending_ioapic_eoi;
813 	int pending_external_vector;
814 
815 	/* be preempted when it's in kernel-mode(cpl=0) */
816 	bool preempted_in_kernel;
817 
818 	/* Flush the L1 Data cache for L1TF mitigation on VMENTER */
819 	bool l1tf_flush_l1d;
820 
821 	/* AMD MSRC001_0015 Hardware Configuration */
822 	u64 msr_hwcr;
823 };
824 
825 struct kvm_lpage_info {
826 	int disallow_lpage;
827 };
828 
829 struct kvm_arch_memory_slot {
830 	struct kvm_rmap_head *rmap[KVM_NR_PAGE_SIZES];
831 	struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
832 	unsigned short *gfn_track[KVM_PAGE_TRACK_MAX];
833 };
834 
835 /*
836  * We use as the mode the number of bits allocated in the LDR for the
837  * logical processor ID.  It happens that these are all powers of two.
838  * This makes it is very easy to detect cases where the APICs are
839  * configured for multiple modes; in that case, we cannot use the map and
840  * hence cannot use kvm_irq_delivery_to_apic_fast either.
841  */
842 #define KVM_APIC_MODE_XAPIC_CLUSTER          4
843 #define KVM_APIC_MODE_XAPIC_FLAT             8
844 #define KVM_APIC_MODE_X2APIC                16
845 
846 struct kvm_apic_map {
847 	struct rcu_head rcu;
848 	u8 mode;
849 	u32 max_apic_id;
850 	union {
851 		struct kvm_lapic *xapic_flat_map[8];
852 		struct kvm_lapic *xapic_cluster_map[16][4];
853 	};
854 	struct kvm_lapic *phys_map[];
855 };
856 
857 /* Hyper-V emulation context */
858 struct kvm_hv {
859 	struct mutex hv_lock;
860 	u64 hv_guest_os_id;
861 	u64 hv_hypercall;
862 	u64 hv_tsc_page;
863 
864 	/* Hyper-v based guest crash (NT kernel bugcheck) parameters */
865 	u64 hv_crash_param[HV_X64_MSR_CRASH_PARAMS];
866 	u64 hv_crash_ctl;
867 
868 	HV_REFERENCE_TSC_PAGE tsc_ref;
869 
870 	struct idr conn_to_evt;
871 
872 	u64 hv_reenlightenment_control;
873 	u64 hv_tsc_emulation_control;
874 	u64 hv_tsc_emulation_status;
875 
876 	/* How many vCPUs have VP index != vCPU index */
877 	atomic_t num_mismatched_vp_indexes;
878 
879 	struct hv_partition_assist_pg *hv_pa_pg;
880 };
881 
882 enum kvm_irqchip_mode {
883 	KVM_IRQCHIP_NONE,
884 	KVM_IRQCHIP_KERNEL,       /* created with KVM_CREATE_IRQCHIP */
885 	KVM_IRQCHIP_SPLIT,        /* created with KVM_CAP_SPLIT_IRQCHIP */
886 };
887 
888 #define APICV_INHIBIT_REASON_DISABLE    0
889 #define APICV_INHIBIT_REASON_HYPERV     1
890 #define APICV_INHIBIT_REASON_NESTED     2
891 #define APICV_INHIBIT_REASON_IRQWIN     3
892 #define APICV_INHIBIT_REASON_PIT_REINJ  4
893 #define APICV_INHIBIT_REASON_X2APIC	5
894 
895 struct kvm_arch {
896 	unsigned long n_used_mmu_pages;
897 	unsigned long n_requested_mmu_pages;
898 	unsigned long n_max_mmu_pages;
899 	unsigned int indirect_shadow_pages;
900 	u8 mmu_valid_gen;
901 	struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
902 	/*
903 	 * Hash table of struct kvm_mmu_page.
904 	 */
905 	struct list_head active_mmu_pages;
906 	struct list_head zapped_obsolete_pages;
907 	struct list_head lpage_disallowed_mmu_pages;
908 	struct kvm_page_track_notifier_node mmu_sp_tracker;
909 	struct kvm_page_track_notifier_head track_notifier_head;
910 
911 	struct list_head assigned_dev_head;
912 	struct iommu_domain *iommu_domain;
913 	bool iommu_noncoherent;
914 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA
915 	atomic_t noncoherent_dma_count;
916 #define __KVM_HAVE_ARCH_ASSIGNED_DEVICE
917 	atomic_t assigned_device_count;
918 	struct kvm_pic *vpic;
919 	struct kvm_ioapic *vioapic;
920 	struct kvm_pit *vpit;
921 	atomic_t vapics_in_nmi_mode;
922 	struct mutex apic_map_lock;
923 	struct kvm_apic_map *apic_map;
924 	bool apic_map_dirty;
925 
926 	bool apic_access_page_done;
927 	unsigned long apicv_inhibit_reasons;
928 
929 	gpa_t wall_clock;
930 
931 	bool mwait_in_guest;
932 	bool hlt_in_guest;
933 	bool pause_in_guest;
934 	bool cstate_in_guest;
935 
936 	unsigned long irq_sources_bitmap;
937 	s64 kvmclock_offset;
938 	raw_spinlock_t tsc_write_lock;
939 	u64 last_tsc_nsec;
940 	u64 last_tsc_write;
941 	u32 last_tsc_khz;
942 	u64 cur_tsc_nsec;
943 	u64 cur_tsc_write;
944 	u64 cur_tsc_offset;
945 	u64 cur_tsc_generation;
946 	int nr_vcpus_matched_tsc;
947 
948 	spinlock_t pvclock_gtod_sync_lock;
949 	bool use_master_clock;
950 	u64 master_kernel_ns;
951 	u64 master_cycle_now;
952 	struct delayed_work kvmclock_update_work;
953 	struct delayed_work kvmclock_sync_work;
954 
955 	struct kvm_xen_hvm_config xen_hvm_config;
956 
957 	/* reads protected by irq_srcu, writes by irq_lock */
958 	struct hlist_head mask_notifier_list;
959 
960 	struct kvm_hv hyperv;
961 
962 	#ifdef CONFIG_KVM_MMU_AUDIT
963 	int audit_point;
964 	#endif
965 
966 	bool backwards_tsc_observed;
967 	bool boot_vcpu_runs_old_kvmclock;
968 	u32 bsp_vcpu_id;
969 
970 	u64 disabled_quirks;
971 
972 	enum kvm_irqchip_mode irqchip_mode;
973 	u8 nr_reserved_ioapic_pins;
974 
975 	bool disabled_lapic_found;
976 
977 	bool x2apic_format;
978 	bool x2apic_broadcast_quirk_disabled;
979 
980 	bool guest_can_read_msr_platform_info;
981 	bool exception_payload_enabled;
982 
983 	struct kvm_pmu_event_filter *pmu_event_filter;
984 	struct task_struct *nx_lpage_recovery_thread;
985 };
986 
987 struct kvm_vm_stat {
988 	ulong mmu_shadow_zapped;
989 	ulong mmu_pte_write;
990 	ulong mmu_pte_updated;
991 	ulong mmu_pde_zapped;
992 	ulong mmu_flooded;
993 	ulong mmu_recycled;
994 	ulong mmu_cache_miss;
995 	ulong mmu_unsync;
996 	ulong remote_tlb_flush;
997 	ulong lpages;
998 	ulong nx_lpage_splits;
999 	ulong max_mmu_page_hash_collisions;
1000 };
1001 
1002 struct kvm_vcpu_stat {
1003 	u64 pf_fixed;
1004 	u64 pf_guest;
1005 	u64 tlb_flush;
1006 	u64 invlpg;
1007 
1008 	u64 exits;
1009 	u64 io_exits;
1010 	u64 mmio_exits;
1011 	u64 signal_exits;
1012 	u64 irq_window_exits;
1013 	u64 nmi_window_exits;
1014 	u64 l1d_flush;
1015 	u64 halt_exits;
1016 	u64 halt_successful_poll;
1017 	u64 halt_attempted_poll;
1018 	u64 halt_poll_invalid;
1019 	u64 halt_wakeup;
1020 	u64 request_irq_exits;
1021 	u64 irq_exits;
1022 	u64 host_state_reload;
1023 	u64 fpu_reload;
1024 	u64 insn_emulation;
1025 	u64 insn_emulation_fail;
1026 	u64 hypercalls;
1027 	u64 irq_injections;
1028 	u64 nmi_injections;
1029 	u64 req_event;
1030 };
1031 
1032 struct x86_instruction_info;
1033 
1034 struct msr_data {
1035 	bool host_initiated;
1036 	u32 index;
1037 	u64 data;
1038 };
1039 
1040 struct kvm_lapic_irq {
1041 	u32 vector;
1042 	u16 delivery_mode;
1043 	u16 dest_mode;
1044 	bool level;
1045 	u16 trig_mode;
1046 	u32 shorthand;
1047 	u32 dest_id;
1048 	bool msi_redir_hint;
1049 };
1050 
1051 static inline u16 kvm_lapic_irq_dest_mode(bool dest_mode_logical)
1052 {
1053 	return dest_mode_logical ? APIC_DEST_LOGICAL : APIC_DEST_PHYSICAL;
1054 }
1055 
1056 struct kvm_x86_ops {
1057 	int (*hardware_enable)(void);
1058 	void (*hardware_disable)(void);
1059 	void (*hardware_unsetup)(void);
1060 	bool (*cpu_has_accelerated_tpr)(void);
1061 	bool (*has_emulated_msr)(int index);
1062 	void (*cpuid_update)(struct kvm_vcpu *vcpu);
1063 
1064 	unsigned int vm_size;
1065 	int (*vm_init)(struct kvm *kvm);
1066 	void (*vm_destroy)(struct kvm *kvm);
1067 
1068 	/* Create, but do not attach this VCPU */
1069 	int (*vcpu_create)(struct kvm_vcpu *vcpu);
1070 	void (*vcpu_free)(struct kvm_vcpu *vcpu);
1071 	void (*vcpu_reset)(struct kvm_vcpu *vcpu, bool init_event);
1072 
1073 	void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
1074 	void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
1075 	void (*vcpu_put)(struct kvm_vcpu *vcpu);
1076 
1077 	void (*update_bp_intercept)(struct kvm_vcpu *vcpu);
1078 	int (*get_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1079 	int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1080 	u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
1081 	void (*get_segment)(struct kvm_vcpu *vcpu,
1082 			    struct kvm_segment *var, int seg);
1083 	int (*get_cpl)(struct kvm_vcpu *vcpu);
1084 	void (*set_segment)(struct kvm_vcpu *vcpu,
1085 			    struct kvm_segment *var, int seg);
1086 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
1087 	void (*decache_cr0_guest_bits)(struct kvm_vcpu *vcpu);
1088 	void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu);
1089 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
1090 	int (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
1091 	void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
1092 	void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1093 	void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1094 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1095 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1096 	u64 (*get_dr6)(struct kvm_vcpu *vcpu);
1097 	void (*set_dr6)(struct kvm_vcpu *vcpu, unsigned long value);
1098 	void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu);
1099 	void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
1100 	void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
1101 	unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
1102 	void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
1103 
1104 	void (*tlb_flush)(struct kvm_vcpu *vcpu, bool invalidate_gpa);
1105 	int  (*tlb_remote_flush)(struct kvm *kvm);
1106 	int  (*tlb_remote_flush_with_range)(struct kvm *kvm,
1107 			struct kvm_tlb_range *range);
1108 
1109 	/*
1110 	 * Flush any TLB entries associated with the given GVA.
1111 	 * Does not need to flush GPA->HPA mappings.
1112 	 * Can potentially get non-canonical addresses through INVLPGs, which
1113 	 * the implementation may choose to ignore if appropriate.
1114 	 */
1115 	void (*tlb_flush_gva)(struct kvm_vcpu *vcpu, gva_t addr);
1116 
1117 	void (*run)(struct kvm_vcpu *vcpu);
1118 	int (*handle_exit)(struct kvm_vcpu *vcpu,
1119 		enum exit_fastpath_completion exit_fastpath);
1120 	int (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
1121 	void (*update_emulated_instruction)(struct kvm_vcpu *vcpu);
1122 	void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
1123 	u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu);
1124 	void (*patch_hypercall)(struct kvm_vcpu *vcpu,
1125 				unsigned char *hypercall_addr);
1126 	void (*set_irq)(struct kvm_vcpu *vcpu);
1127 	void (*set_nmi)(struct kvm_vcpu *vcpu);
1128 	void (*queue_exception)(struct kvm_vcpu *vcpu);
1129 	void (*cancel_injection)(struct kvm_vcpu *vcpu);
1130 	int (*interrupt_allowed)(struct kvm_vcpu *vcpu);
1131 	int (*nmi_allowed)(struct kvm_vcpu *vcpu);
1132 	bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
1133 	void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
1134 	void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
1135 	void (*enable_irq_window)(struct kvm_vcpu *vcpu);
1136 	void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
1137 	bool (*check_apicv_inhibit_reasons)(ulong bit);
1138 	void (*pre_update_apicv_exec_ctrl)(struct kvm *kvm, bool activate);
1139 	void (*refresh_apicv_exec_ctrl)(struct kvm_vcpu *vcpu);
1140 	void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
1141 	void (*hwapic_isr_update)(struct kvm_vcpu *vcpu, int isr);
1142 	bool (*guest_apic_has_interrupt)(struct kvm_vcpu *vcpu);
1143 	void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
1144 	void (*set_virtual_apic_mode)(struct kvm_vcpu *vcpu);
1145 	void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu, hpa_t hpa);
1146 	int (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
1147 	int (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
1148 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
1149 	int (*set_identity_map_addr)(struct kvm *kvm, u64 ident_addr);
1150 	int (*get_tdp_level)(struct kvm_vcpu *vcpu);
1151 	u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
1152 
1153 	void (*load_mmu_pgd)(struct kvm_vcpu *vcpu, unsigned long cr3);
1154 
1155 	bool (*has_wbinvd_exit)(void);
1156 
1157 	u64 (*read_l1_tsc_offset)(struct kvm_vcpu *vcpu);
1158 	/* Returns actual tsc_offset set in active VMCS */
1159 	u64 (*write_l1_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
1160 
1161 	void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2);
1162 
1163 	int (*check_intercept)(struct kvm_vcpu *vcpu,
1164 			       struct x86_instruction_info *info,
1165 			       enum x86_intercept_stage stage,
1166 			       struct x86_exception *exception);
1167 	void (*handle_exit_irqoff)(struct kvm_vcpu *vcpu,
1168 		enum exit_fastpath_completion *exit_fastpath);
1169 
1170 	int (*check_nested_events)(struct kvm_vcpu *vcpu);
1171 	void (*request_immediate_exit)(struct kvm_vcpu *vcpu);
1172 
1173 	void (*sched_in)(struct kvm_vcpu *kvm, int cpu);
1174 
1175 	/*
1176 	 * Arch-specific dirty logging hooks. These hooks are only supposed to
1177 	 * be valid if the specific arch has hardware-accelerated dirty logging
1178 	 * mechanism. Currently only for PML on VMX.
1179 	 *
1180 	 *  - slot_enable_log_dirty:
1181 	 *	called when enabling log dirty mode for the slot.
1182 	 *  - slot_disable_log_dirty:
1183 	 *	called when disabling log dirty mode for the slot.
1184 	 *	also called when slot is created with log dirty disabled.
1185 	 *  - flush_log_dirty:
1186 	 *	called before reporting dirty_bitmap to userspace.
1187 	 *  - enable_log_dirty_pt_masked:
1188 	 *	called when reenabling log dirty for the GFNs in the mask after
1189 	 *	corresponding bits are cleared in slot->dirty_bitmap.
1190 	 */
1191 	void (*slot_enable_log_dirty)(struct kvm *kvm,
1192 				      struct kvm_memory_slot *slot);
1193 	void (*slot_disable_log_dirty)(struct kvm *kvm,
1194 				       struct kvm_memory_slot *slot);
1195 	void (*flush_log_dirty)(struct kvm *kvm);
1196 	void (*enable_log_dirty_pt_masked)(struct kvm *kvm,
1197 					   struct kvm_memory_slot *slot,
1198 					   gfn_t offset, unsigned long mask);
1199 	int (*write_log_dirty)(struct kvm_vcpu *vcpu);
1200 
1201 	/* pmu operations of sub-arch */
1202 	const struct kvm_pmu_ops *pmu_ops;
1203 
1204 	/*
1205 	 * Architecture specific hooks for vCPU blocking due to
1206 	 * HLT instruction.
1207 	 * Returns for .pre_block():
1208 	 *    - 0 means continue to block the vCPU.
1209 	 *    - 1 means we cannot block the vCPU since some event
1210 	 *        happens during this period, such as, 'ON' bit in
1211 	 *        posted-interrupts descriptor is set.
1212 	 */
1213 	int (*pre_block)(struct kvm_vcpu *vcpu);
1214 	void (*post_block)(struct kvm_vcpu *vcpu);
1215 
1216 	void (*vcpu_blocking)(struct kvm_vcpu *vcpu);
1217 	void (*vcpu_unblocking)(struct kvm_vcpu *vcpu);
1218 
1219 	int (*update_pi_irte)(struct kvm *kvm, unsigned int host_irq,
1220 			      uint32_t guest_irq, bool set);
1221 	void (*apicv_post_state_restore)(struct kvm_vcpu *vcpu);
1222 	bool (*dy_apicv_has_pending_interrupt)(struct kvm_vcpu *vcpu);
1223 
1224 	int (*set_hv_timer)(struct kvm_vcpu *vcpu, u64 guest_deadline_tsc,
1225 			    bool *expired);
1226 	void (*cancel_hv_timer)(struct kvm_vcpu *vcpu);
1227 
1228 	void (*setup_mce)(struct kvm_vcpu *vcpu);
1229 
1230 	int (*get_nested_state)(struct kvm_vcpu *vcpu,
1231 				struct kvm_nested_state __user *user_kvm_nested_state,
1232 				unsigned user_data_size);
1233 	int (*set_nested_state)(struct kvm_vcpu *vcpu,
1234 				struct kvm_nested_state __user *user_kvm_nested_state,
1235 				struct kvm_nested_state *kvm_state);
1236 	bool (*get_vmcs12_pages)(struct kvm_vcpu *vcpu);
1237 
1238 	int (*smi_allowed)(struct kvm_vcpu *vcpu);
1239 	int (*pre_enter_smm)(struct kvm_vcpu *vcpu, char *smstate);
1240 	int (*pre_leave_smm)(struct kvm_vcpu *vcpu, const char *smstate);
1241 	int (*enable_smi_window)(struct kvm_vcpu *vcpu);
1242 
1243 	int (*mem_enc_op)(struct kvm *kvm, void __user *argp);
1244 	int (*mem_enc_reg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1245 	int (*mem_enc_unreg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1246 
1247 	int (*get_msr_feature)(struct kvm_msr_entry *entry);
1248 
1249 	int (*nested_enable_evmcs)(struct kvm_vcpu *vcpu,
1250 				   uint16_t *vmcs_version);
1251 	uint16_t (*nested_get_evmcs_version)(struct kvm_vcpu *vcpu);
1252 
1253 	bool (*need_emulation_on_page_fault)(struct kvm_vcpu *vcpu);
1254 
1255 	bool (*apic_init_signal_blocked)(struct kvm_vcpu *vcpu);
1256 	int (*enable_direct_tlbflush)(struct kvm_vcpu *vcpu);
1257 };
1258 
1259 struct kvm_x86_init_ops {
1260 	int (*cpu_has_kvm_support)(void);
1261 	int (*disabled_by_bios)(void);
1262 	int (*check_processor_compatibility)(void);
1263 	int (*hardware_setup)(void);
1264 
1265 	struct kvm_x86_ops *runtime_ops;
1266 };
1267 
1268 struct kvm_arch_async_pf {
1269 	u32 token;
1270 	gfn_t gfn;
1271 	unsigned long cr3;
1272 	bool direct_map;
1273 };
1274 
1275 extern u64 __read_mostly host_efer;
1276 
1277 extern struct kvm_x86_ops kvm_x86_ops;
1278 extern struct kmem_cache *x86_fpu_cache;
1279 
1280 #define __KVM_HAVE_ARCH_VM_ALLOC
1281 static inline struct kvm *kvm_arch_alloc_vm(void)
1282 {
1283 	return __vmalloc(kvm_x86_ops.vm_size,
1284 			 GFP_KERNEL_ACCOUNT | __GFP_ZERO, PAGE_KERNEL);
1285 }
1286 void kvm_arch_free_vm(struct kvm *kvm);
1287 
1288 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
1289 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
1290 {
1291 	if (kvm_x86_ops.tlb_remote_flush &&
1292 	    !kvm_x86_ops.tlb_remote_flush(kvm))
1293 		return 0;
1294 	else
1295 		return -ENOTSUPP;
1296 }
1297 
1298 int kvm_mmu_module_init(void);
1299 void kvm_mmu_module_exit(void);
1300 
1301 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
1302 int kvm_mmu_create(struct kvm_vcpu *vcpu);
1303 void kvm_mmu_init_vm(struct kvm *kvm);
1304 void kvm_mmu_uninit_vm(struct kvm *kvm);
1305 void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
1306 		u64 dirty_mask, u64 nx_mask, u64 x_mask, u64 p_mask,
1307 		u64 acc_track_mask, u64 me_mask);
1308 
1309 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
1310 void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
1311 				      struct kvm_memory_slot *memslot,
1312 				      int start_level);
1313 void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
1314 				   const struct kvm_memory_slot *memslot);
1315 void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
1316 				   struct kvm_memory_slot *memslot);
1317 void kvm_mmu_slot_largepage_remove_write_access(struct kvm *kvm,
1318 					struct kvm_memory_slot *memslot);
1319 void kvm_mmu_slot_set_dirty(struct kvm *kvm,
1320 			    struct kvm_memory_slot *memslot);
1321 void kvm_mmu_clear_dirty_pt_masked(struct kvm *kvm,
1322 				   struct kvm_memory_slot *slot,
1323 				   gfn_t gfn_offset, unsigned long mask);
1324 void kvm_mmu_zap_all(struct kvm *kvm);
1325 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, u64 gen);
1326 unsigned long kvm_mmu_calculate_default_mmu_pages(struct kvm *kvm);
1327 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned long kvm_nr_mmu_pages);
1328 
1329 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
1330 bool pdptrs_changed(struct kvm_vcpu *vcpu);
1331 
1332 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
1333 			  const void *val, int bytes);
1334 
1335 struct kvm_irq_mask_notifier {
1336 	void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
1337 	int irq;
1338 	struct hlist_node link;
1339 };
1340 
1341 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
1342 				    struct kvm_irq_mask_notifier *kimn);
1343 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
1344 				      struct kvm_irq_mask_notifier *kimn);
1345 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
1346 			     bool mask);
1347 
1348 extern bool tdp_enabled;
1349 
1350 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
1351 
1352 /* control of guest tsc rate supported? */
1353 extern bool kvm_has_tsc_control;
1354 /* maximum supported tsc_khz for guests */
1355 extern u32  kvm_max_guest_tsc_khz;
1356 /* number of bits of the fractional part of the TSC scaling ratio */
1357 extern u8   kvm_tsc_scaling_ratio_frac_bits;
1358 /* maximum allowed value of TSC scaling ratio */
1359 extern u64  kvm_max_tsc_scaling_ratio;
1360 /* 1ull << kvm_tsc_scaling_ratio_frac_bits */
1361 extern u64  kvm_default_tsc_scaling_ratio;
1362 
1363 extern u64 kvm_mce_cap_supported;
1364 
1365 /*
1366  * EMULTYPE_NO_DECODE - Set when re-emulating an instruction (after completing
1367  *			userspace I/O) to indicate that the emulation context
1368  *			should be resued as is, i.e. skip initialization of
1369  *			emulation context, instruction fetch and decode.
1370  *
1371  * EMULTYPE_TRAP_UD - Set when emulating an intercepted #UD from hardware.
1372  *		      Indicates that only select instructions (tagged with
1373  *		      EmulateOnUD) should be emulated (to minimize the emulator
1374  *		      attack surface).  See also EMULTYPE_TRAP_UD_FORCED.
1375  *
1376  * EMULTYPE_SKIP - Set when emulating solely to skip an instruction, i.e. to
1377  *		   decode the instruction length.  For use *only* by
1378  *		   kvm_x86_ops.skip_emulated_instruction() implementations.
1379  *
1380  * EMULTYPE_ALLOW_RETRY_PF - Set when the emulator should resume the guest to
1381  *			     retry native execution under certain conditions,
1382  *			     Can only be set in conjunction with EMULTYPE_PF.
1383  *
1384  * EMULTYPE_TRAP_UD_FORCED - Set when emulating an intercepted #UD that was
1385  *			     triggered by KVM's magic "force emulation" prefix,
1386  *			     which is opt in via module param (off by default).
1387  *			     Bypasses EmulateOnUD restriction despite emulating
1388  *			     due to an intercepted #UD (see EMULTYPE_TRAP_UD).
1389  *			     Used to test the full emulator from userspace.
1390  *
1391  * EMULTYPE_VMWARE_GP - Set when emulating an intercepted #GP for VMware
1392  *			backdoor emulation, which is opt in via module param.
1393  *			VMware backoor emulation handles select instructions
1394  *			and reinjects the #GP for all other cases.
1395  *
1396  * EMULTYPE_PF - Set when emulating MMIO by way of an intercepted #PF, in which
1397  *		 case the CR2/GPA value pass on the stack is valid.
1398  */
1399 #define EMULTYPE_NO_DECODE	    (1 << 0)
1400 #define EMULTYPE_TRAP_UD	    (1 << 1)
1401 #define EMULTYPE_SKIP		    (1 << 2)
1402 #define EMULTYPE_ALLOW_RETRY_PF	    (1 << 3)
1403 #define EMULTYPE_TRAP_UD_FORCED	    (1 << 4)
1404 #define EMULTYPE_VMWARE_GP	    (1 << 5)
1405 #define EMULTYPE_PF		    (1 << 6)
1406 
1407 int kvm_emulate_instruction(struct kvm_vcpu *vcpu, int emulation_type);
1408 int kvm_emulate_instruction_from_buffer(struct kvm_vcpu *vcpu,
1409 					void *insn, int insn_len);
1410 
1411 void kvm_enable_efer_bits(u64);
1412 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
1413 int __kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data, bool host_initiated);
1414 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data);
1415 int kvm_set_msr(struct kvm_vcpu *vcpu, u32 index, u64 data);
1416 int kvm_emulate_rdmsr(struct kvm_vcpu *vcpu);
1417 int kvm_emulate_wrmsr(struct kvm_vcpu *vcpu);
1418 
1419 int kvm_fast_pio(struct kvm_vcpu *vcpu, int size, unsigned short port, int in);
1420 int kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
1421 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
1422 int kvm_vcpu_halt(struct kvm_vcpu *vcpu);
1423 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
1424 
1425 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
1426 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
1427 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector);
1428 
1429 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
1430 		    int reason, bool has_error_code, u32 error_code);
1431 
1432 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
1433 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
1434 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
1435 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
1436 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
1437 int kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
1438 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
1439 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
1440 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
1441 int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr);
1442 
1443 int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1444 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1445 
1446 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
1447 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
1448 bool kvm_rdpmc(struct kvm_vcpu *vcpu);
1449 
1450 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1451 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1452 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1453 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1454 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
1455 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1456 			    gfn_t gfn, void *data, int offset, int len,
1457 			    u32 access);
1458 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
1459 bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr);
1460 
1461 static inline int __kvm_irq_line_state(unsigned long *irq_state,
1462 				       int irq_source_id, int level)
1463 {
1464 	/* Logical OR for level trig interrupt */
1465 	if (level)
1466 		__set_bit(irq_source_id, irq_state);
1467 	else
1468 		__clear_bit(irq_source_id, irq_state);
1469 
1470 	return !!(*irq_state);
1471 }
1472 
1473 #define KVM_MMU_ROOT_CURRENT		BIT(0)
1474 #define KVM_MMU_ROOT_PREVIOUS(i)	BIT(1+i)
1475 #define KVM_MMU_ROOTS_ALL		(~0UL)
1476 
1477 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
1478 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
1479 
1480 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
1481 
1482 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
1483 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
1484 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
1485 int kvm_mmu_load(struct kvm_vcpu *vcpu);
1486 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
1487 void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu);
1488 void kvm_mmu_free_roots(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1489 			ulong roots_to_free);
1490 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1491 			   struct x86_exception *exception);
1492 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
1493 			      struct x86_exception *exception);
1494 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
1495 			       struct x86_exception *exception);
1496 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
1497 			       struct x86_exception *exception);
1498 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
1499 				struct x86_exception *exception);
1500 
1501 bool kvm_apicv_activated(struct kvm *kvm);
1502 void kvm_apicv_init(struct kvm *kvm, bool enable);
1503 void kvm_vcpu_update_apicv(struct kvm_vcpu *vcpu);
1504 void kvm_request_apicv_update(struct kvm *kvm, bool activate,
1505 			      unsigned long bit);
1506 
1507 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
1508 
1509 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u64 error_code,
1510 		       void *insn, int insn_len);
1511 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
1512 void kvm_mmu_invpcid_gva(struct kvm_vcpu *vcpu, gva_t gva, unsigned long pcid);
1513 void kvm_mmu_new_cr3(struct kvm_vcpu *vcpu, gpa_t new_cr3, bool skip_tlb_flush);
1514 
1515 void kvm_configure_mmu(bool enable_tdp, int tdp_page_level);
1516 
1517 static inline gpa_t translate_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1518 				  struct x86_exception *exception)
1519 {
1520 	return gpa;
1521 }
1522 
1523 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
1524 {
1525 	struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
1526 
1527 	return (struct kvm_mmu_page *)page_private(page);
1528 }
1529 
1530 static inline u16 kvm_read_ldt(void)
1531 {
1532 	u16 ldt;
1533 	asm("sldt %0" : "=g"(ldt));
1534 	return ldt;
1535 }
1536 
1537 static inline void kvm_load_ldt(u16 sel)
1538 {
1539 	asm("lldt %0" : : "rm"(sel));
1540 }
1541 
1542 #ifdef CONFIG_X86_64
1543 static inline unsigned long read_msr(unsigned long msr)
1544 {
1545 	u64 value;
1546 
1547 	rdmsrl(msr, value);
1548 	return value;
1549 }
1550 #endif
1551 
1552 static inline u32 get_rdx_init_val(void)
1553 {
1554 	return 0x600; /* P6 family */
1555 }
1556 
1557 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
1558 {
1559 	kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
1560 }
1561 
1562 #define TSS_IOPB_BASE_OFFSET 0x66
1563 #define TSS_BASE_SIZE 0x68
1564 #define TSS_IOPB_SIZE (65536 / 8)
1565 #define TSS_REDIRECTION_SIZE (256 / 8)
1566 #define RMODE_TSS_SIZE							\
1567 	(TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
1568 
1569 enum {
1570 	TASK_SWITCH_CALL = 0,
1571 	TASK_SWITCH_IRET = 1,
1572 	TASK_SWITCH_JMP = 2,
1573 	TASK_SWITCH_GATE = 3,
1574 };
1575 
1576 #define HF_GIF_MASK		(1 << 0)
1577 #define HF_HIF_MASK		(1 << 1)
1578 #define HF_VINTR_MASK		(1 << 2)
1579 #define HF_NMI_MASK		(1 << 3)
1580 #define HF_IRET_MASK		(1 << 4)
1581 #define HF_GUEST_MASK		(1 << 5) /* VCPU is in guest-mode */
1582 #define HF_SMM_MASK		(1 << 6)
1583 #define HF_SMM_INSIDE_NMI_MASK	(1 << 7)
1584 
1585 #define __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
1586 #define KVM_ADDRESS_SPACE_NUM 2
1587 
1588 #define kvm_arch_vcpu_memslots_id(vcpu) ((vcpu)->arch.hflags & HF_SMM_MASK ? 1 : 0)
1589 #define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, (role).smm)
1590 
1591 asmlinkage void kvm_spurious_fault(void);
1592 
1593 /*
1594  * Hardware virtualization extension instructions may fault if a
1595  * reboot turns off virtualization while processes are running.
1596  * Usually after catching the fault we just panic; during reboot
1597  * instead the instruction is ignored.
1598  */
1599 #define __kvm_handle_fault_on_reboot(insn)				\
1600 	"666: \n\t"							\
1601 	insn "\n\t"							\
1602 	"jmp	668f \n\t"						\
1603 	"667: \n\t"							\
1604 	"call	kvm_spurious_fault \n\t"				\
1605 	"668: \n\t"							\
1606 	_ASM_EXTABLE(666b, 667b)
1607 
1608 #define KVM_ARCH_WANT_MMU_NOTIFIER
1609 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end);
1610 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
1611 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
1612 int kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
1613 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1614 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1615 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1616 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1617 void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event);
1618 void kvm_vcpu_reload_apic_access_page(struct kvm_vcpu *vcpu);
1619 
1620 int kvm_pv_send_ipi(struct kvm *kvm, unsigned long ipi_bitmap_low,
1621 		    unsigned long ipi_bitmap_high, u32 min,
1622 		    unsigned long icr, int op_64_bit);
1623 
1624 void kvm_define_shared_msr(unsigned index, u32 msr);
1625 int kvm_set_shared_msr(unsigned index, u64 val, u64 mask);
1626 
1627 u64 kvm_scale_tsc(struct kvm_vcpu *vcpu, u64 tsc);
1628 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc);
1629 
1630 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu);
1631 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1632 
1633 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
1634 void kvm_make_scan_ioapic_request(struct kvm *kvm);
1635 void kvm_make_scan_ioapic_request_mask(struct kvm *kvm,
1636 				       unsigned long *vcpu_bitmap);
1637 
1638 void kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1639 				     struct kvm_async_pf *work);
1640 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1641 				 struct kvm_async_pf *work);
1642 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1643 			       struct kvm_async_pf *work);
1644 bool kvm_arch_can_inject_async_page_present(struct kvm_vcpu *vcpu);
1645 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1646 
1647 int kvm_skip_emulated_instruction(struct kvm_vcpu *vcpu);
1648 int kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1649 void __kvm_request_immediate_exit(struct kvm_vcpu *vcpu);
1650 
1651 int kvm_is_in_guest(void);
1652 
1653 int __x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa, u32 size);
1654 bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu);
1655 bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu);
1656 
1657 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
1658 			     struct kvm_vcpu **dest_vcpu);
1659 
1660 void kvm_set_msi_irq(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e,
1661 		     struct kvm_lapic_irq *irq);
1662 
1663 static inline bool kvm_irq_is_postable(struct kvm_lapic_irq *irq)
1664 {
1665 	/* We can only post Fixed and LowPrio IRQs */
1666 	return (irq->delivery_mode == dest_Fixed ||
1667 		irq->delivery_mode == dest_LowestPrio);
1668 }
1669 
1670 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu)
1671 {
1672 	if (kvm_x86_ops.vcpu_blocking)
1673 		kvm_x86_ops.vcpu_blocking(vcpu);
1674 }
1675 
1676 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu)
1677 {
1678 	if (kvm_x86_ops.vcpu_unblocking)
1679 		kvm_x86_ops.vcpu_unblocking(vcpu);
1680 }
1681 
1682 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
1683 
1684 static inline int kvm_cpu_get_apicid(int mps_cpu)
1685 {
1686 #ifdef CONFIG_X86_LOCAL_APIC
1687 	return default_cpu_present_to_apicid(mps_cpu);
1688 #else
1689 	WARN_ON_ONCE(1);
1690 	return BAD_APICID;
1691 #endif
1692 }
1693 
1694 #define put_smstate(type, buf, offset, val)                      \
1695 	*(type *)((buf) + (offset) - 0x7e00) = val
1696 
1697 #define GET_SMSTATE(type, buf, offset)		\
1698 	(*(type *)((buf) + (offset) - 0x7e00))
1699 
1700 #endif /* _ASM_X86_KVM_HOST_H */
1701