xref: /openbmc/linux/arch/x86/include/asm/kvm_host.h (revision 4ed91d48259d9ddd378424d008f2e6559f7e78f8)
1 /*
2  * Kernel-based Virtual Machine driver for Linux
3  *
4  * This header defines architecture specific interfaces, x86 version
5  *
6  * This work is licensed under the terms of the GNU GPL, version 2.  See
7  * the COPYING file in the top-level directory.
8  *
9  */
10 
11 #ifndef _ASM_X86_KVM_HOST_H
12 #define _ASM_X86_KVM_HOST_H
13 
14 #include <linux/types.h>
15 #include <linux/mm.h>
16 #include <linux/mmu_notifier.h>
17 #include <linux/tracepoint.h>
18 #include <linux/cpumask.h>
19 #include <linux/irq_work.h>
20 
21 #include <linux/kvm.h>
22 #include <linux/kvm_para.h>
23 #include <linux/kvm_types.h>
24 #include <linux/perf_event.h>
25 #include <linux/pvclock_gtod.h>
26 #include <linux/clocksource.h>
27 #include <linux/irqbypass.h>
28 #include <linux/hyperv.h>
29 
30 #include <asm/apic.h>
31 #include <asm/pvclock-abi.h>
32 #include <asm/desc.h>
33 #include <asm/mtrr.h>
34 #include <asm/msr-index.h>
35 #include <asm/asm.h>
36 #include <asm/kvm_page_track.h>
37 
38 #define KVM_MAX_VCPUS 288
39 #define KVM_SOFT_MAX_VCPUS 240
40 #define KVM_MAX_VCPU_ID 1023
41 #define KVM_USER_MEM_SLOTS 509
42 /* memory slots that are not exposed to userspace */
43 #define KVM_PRIVATE_MEM_SLOTS 3
44 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
45 
46 #define KVM_PIO_PAGE_OFFSET 1
47 #define KVM_COALESCED_MMIO_PAGE_OFFSET 2
48 #define KVM_HALT_POLL_NS_DEFAULT 400000
49 
50 #define KVM_IRQCHIP_NUM_PINS  KVM_IOAPIC_NUM_PINS
51 
52 /* x86-specific vcpu->requests bit members */
53 #define KVM_REQ_MIGRATE_TIMER      8
54 #define KVM_REQ_REPORT_TPR_ACCESS  9
55 #define KVM_REQ_TRIPLE_FAULT      10
56 #define KVM_REQ_MMU_SYNC          11
57 #define KVM_REQ_CLOCK_UPDATE      12
58 #define KVM_REQ_EVENT             14
59 #define KVM_REQ_APF_HALT          15
60 #define KVM_REQ_STEAL_UPDATE      16
61 #define KVM_REQ_NMI               17
62 #define KVM_REQ_PMU               18
63 #define KVM_REQ_PMI               19
64 #define KVM_REQ_SMI               20
65 #define KVM_REQ_MASTERCLOCK_UPDATE 21
66 #define KVM_REQ_MCLOCK_INPROGRESS 22
67 #define KVM_REQ_SCAN_IOAPIC       23
68 #define KVM_REQ_GLOBAL_CLOCK_UPDATE 24
69 #define KVM_REQ_APIC_PAGE_RELOAD  25
70 #define KVM_REQ_HV_CRASH          26
71 #define KVM_REQ_IOAPIC_EOI_EXIT   27
72 #define KVM_REQ_HV_RESET          28
73 #define KVM_REQ_HV_EXIT           29
74 #define KVM_REQ_HV_STIMER         30
75 
76 #define CR0_RESERVED_BITS                                               \
77 	(~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \
78 			  | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \
79 			  | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG))
80 
81 #define CR3_L_MODE_RESERVED_BITS 0xFFFFFF0000000000ULL
82 #define CR3_PCID_INVD		 BIT_64(63)
83 #define CR4_RESERVED_BITS                                               \
84 	(~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\
85 			  | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE     \
86 			  | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR | X86_CR4_PCIDE \
87 			  | X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_FSGSBASE \
88 			  | X86_CR4_OSXMMEXCPT | X86_CR4_VMXE | X86_CR4_SMAP \
89 			  | X86_CR4_PKE))
90 
91 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR)
92 
93 
94 
95 #define INVALID_PAGE (~(hpa_t)0)
96 #define VALID_PAGE(x) ((x) != INVALID_PAGE)
97 
98 #define UNMAPPED_GVA (~(gpa_t)0)
99 
100 /* KVM Hugepage definitions for x86 */
101 #define KVM_NR_PAGE_SIZES	3
102 #define KVM_HPAGE_GFN_SHIFT(x)	(((x) - 1) * 9)
103 #define KVM_HPAGE_SHIFT(x)	(PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
104 #define KVM_HPAGE_SIZE(x)	(1UL << KVM_HPAGE_SHIFT(x))
105 #define KVM_HPAGE_MASK(x)	(~(KVM_HPAGE_SIZE(x) - 1))
106 #define KVM_PAGES_PER_HPAGE(x)	(KVM_HPAGE_SIZE(x) / PAGE_SIZE)
107 
108 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
109 {
110 	/* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */
111 	return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
112 		(base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
113 }
114 
115 #define KVM_PERMILLE_MMU_PAGES 20
116 #define KVM_MIN_ALLOC_MMU_PAGES 64
117 #define KVM_MMU_HASH_SHIFT 12
118 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
119 #define KVM_MIN_FREE_MMU_PAGES 5
120 #define KVM_REFILL_PAGES 25
121 #define KVM_MAX_CPUID_ENTRIES 80
122 #define KVM_NR_FIXED_MTRR_REGION 88
123 #define KVM_NR_VAR_MTRR 8
124 
125 #define ASYNC_PF_PER_VCPU 64
126 
127 enum kvm_reg {
128 	VCPU_REGS_RAX = 0,
129 	VCPU_REGS_RCX = 1,
130 	VCPU_REGS_RDX = 2,
131 	VCPU_REGS_RBX = 3,
132 	VCPU_REGS_RSP = 4,
133 	VCPU_REGS_RBP = 5,
134 	VCPU_REGS_RSI = 6,
135 	VCPU_REGS_RDI = 7,
136 #ifdef CONFIG_X86_64
137 	VCPU_REGS_R8 = 8,
138 	VCPU_REGS_R9 = 9,
139 	VCPU_REGS_R10 = 10,
140 	VCPU_REGS_R11 = 11,
141 	VCPU_REGS_R12 = 12,
142 	VCPU_REGS_R13 = 13,
143 	VCPU_REGS_R14 = 14,
144 	VCPU_REGS_R15 = 15,
145 #endif
146 	VCPU_REGS_RIP,
147 	NR_VCPU_REGS
148 };
149 
150 enum kvm_reg_ex {
151 	VCPU_EXREG_PDPTR = NR_VCPU_REGS,
152 	VCPU_EXREG_CR3,
153 	VCPU_EXREG_RFLAGS,
154 	VCPU_EXREG_SEGMENTS,
155 };
156 
157 enum {
158 	VCPU_SREG_ES,
159 	VCPU_SREG_CS,
160 	VCPU_SREG_SS,
161 	VCPU_SREG_DS,
162 	VCPU_SREG_FS,
163 	VCPU_SREG_GS,
164 	VCPU_SREG_TR,
165 	VCPU_SREG_LDTR,
166 };
167 
168 #include <asm/kvm_emulate.h>
169 
170 #define KVM_NR_MEM_OBJS 40
171 
172 #define KVM_NR_DB_REGS	4
173 
174 #define DR6_BD		(1 << 13)
175 #define DR6_BS		(1 << 14)
176 #define DR6_RTM		(1 << 16)
177 #define DR6_FIXED_1	0xfffe0ff0
178 #define DR6_INIT	0xffff0ff0
179 #define DR6_VOLATILE	0x0001e00f
180 
181 #define DR7_BP_EN_MASK	0x000000ff
182 #define DR7_GE		(1 << 9)
183 #define DR7_GD		(1 << 13)
184 #define DR7_FIXED_1	0x00000400
185 #define DR7_VOLATILE	0xffff2bff
186 
187 #define PFERR_PRESENT_BIT 0
188 #define PFERR_WRITE_BIT 1
189 #define PFERR_USER_BIT 2
190 #define PFERR_RSVD_BIT 3
191 #define PFERR_FETCH_BIT 4
192 #define PFERR_PK_BIT 5
193 #define PFERR_GUEST_FINAL_BIT 32
194 #define PFERR_GUEST_PAGE_BIT 33
195 
196 #define PFERR_PRESENT_MASK (1U << PFERR_PRESENT_BIT)
197 #define PFERR_WRITE_MASK (1U << PFERR_WRITE_BIT)
198 #define PFERR_USER_MASK (1U << PFERR_USER_BIT)
199 #define PFERR_RSVD_MASK (1U << PFERR_RSVD_BIT)
200 #define PFERR_FETCH_MASK (1U << PFERR_FETCH_BIT)
201 #define PFERR_PK_MASK (1U << PFERR_PK_BIT)
202 #define PFERR_GUEST_FINAL_MASK (1ULL << PFERR_GUEST_FINAL_BIT)
203 #define PFERR_GUEST_PAGE_MASK (1ULL << PFERR_GUEST_PAGE_BIT)
204 
205 #define PFERR_NESTED_GUEST_PAGE (PFERR_GUEST_PAGE_MASK |	\
206 				 PFERR_USER_MASK |		\
207 				 PFERR_WRITE_MASK |		\
208 				 PFERR_PRESENT_MASK)
209 
210 /*
211  * The mask used to denote special SPTEs, which can be either MMIO SPTEs or
212  * Access Tracking SPTEs. We use bit 62 instead of bit 63 to avoid conflicting
213  * with the SVE bit in EPT PTEs.
214  */
215 #define SPTE_SPECIAL_MASK (1ULL << 62)
216 
217 /* apic attention bits */
218 #define KVM_APIC_CHECK_VAPIC	0
219 /*
220  * The following bit is set with PV-EOI, unset on EOI.
221  * We detect PV-EOI changes by guest by comparing
222  * this bit with PV-EOI in guest memory.
223  * See the implementation in apic_update_pv_eoi.
224  */
225 #define KVM_APIC_PV_EOI_PENDING	1
226 
227 struct kvm_kernel_irq_routing_entry;
228 
229 /*
230  * We don't want allocation failures within the mmu code, so we preallocate
231  * enough memory for a single page fault in a cache.
232  */
233 struct kvm_mmu_memory_cache {
234 	int nobjs;
235 	void *objects[KVM_NR_MEM_OBJS];
236 };
237 
238 /*
239  * the pages used as guest page table on soft mmu are tracked by
240  * kvm_memory_slot.arch.gfn_track which is 16 bits, so the role bits used
241  * by indirect shadow page can not be more than 15 bits.
242  *
243  * Currently, we used 14 bits that are @level, @cr4_pae, @quadrant, @access,
244  * @nxe, @cr0_wp, @smep_andnot_wp and @smap_andnot_wp.
245  */
246 union kvm_mmu_page_role {
247 	unsigned word;
248 	struct {
249 		unsigned level:4;
250 		unsigned cr4_pae:1;
251 		unsigned quadrant:2;
252 		unsigned direct:1;
253 		unsigned access:3;
254 		unsigned invalid:1;
255 		unsigned nxe:1;
256 		unsigned cr0_wp:1;
257 		unsigned smep_andnot_wp:1;
258 		unsigned smap_andnot_wp:1;
259 		unsigned :8;
260 
261 		/*
262 		 * This is left at the top of the word so that
263 		 * kvm_memslots_for_spte_role can extract it with a
264 		 * simple shift.  While there is room, give it a whole
265 		 * byte so it is also faster to load it from memory.
266 		 */
267 		unsigned smm:8;
268 	};
269 };
270 
271 struct kvm_rmap_head {
272 	unsigned long val;
273 };
274 
275 struct kvm_mmu_page {
276 	struct list_head link;
277 	struct hlist_node hash_link;
278 
279 	/*
280 	 * The following two entries are used to key the shadow page in the
281 	 * hash table.
282 	 */
283 	gfn_t gfn;
284 	union kvm_mmu_page_role role;
285 
286 	u64 *spt;
287 	/* hold the gfn of each spte inside spt */
288 	gfn_t *gfns;
289 	bool unsync;
290 	int root_count;          /* Currently serving as active root */
291 	unsigned int unsync_children;
292 	struct kvm_rmap_head parent_ptes; /* rmap pointers to parent sptes */
293 
294 	/* The page is obsolete if mmu_valid_gen != kvm->arch.mmu_valid_gen.  */
295 	unsigned long mmu_valid_gen;
296 
297 	DECLARE_BITMAP(unsync_child_bitmap, 512);
298 
299 #ifdef CONFIG_X86_32
300 	/*
301 	 * Used out of the mmu-lock to avoid reading spte values while an
302 	 * update is in progress; see the comments in __get_spte_lockless().
303 	 */
304 	int clear_spte_count;
305 #endif
306 
307 	/* Number of writes since the last time traversal visited this page.  */
308 	atomic_t write_flooding_count;
309 };
310 
311 struct kvm_pio_request {
312 	unsigned long count;
313 	int in;
314 	int port;
315 	int size;
316 };
317 
318 struct rsvd_bits_validate {
319 	u64 rsvd_bits_mask[2][4];
320 	u64 bad_mt_xwr;
321 };
322 
323 /*
324  * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
325  * 32-bit).  The kvm_mmu structure abstracts the details of the current mmu
326  * mode.
327  */
328 struct kvm_mmu {
329 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long root);
330 	unsigned long (*get_cr3)(struct kvm_vcpu *vcpu);
331 	u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
332 	int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err,
333 			  bool prefault);
334 	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
335 				  struct x86_exception *fault);
336 	gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva, u32 access,
337 			    struct x86_exception *exception);
338 	gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
339 			       struct x86_exception *exception);
340 	int (*sync_page)(struct kvm_vcpu *vcpu,
341 			 struct kvm_mmu_page *sp);
342 	void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva);
343 	void (*update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
344 			   u64 *spte, const void *pte);
345 	hpa_t root_hpa;
346 	int root_level;
347 	int shadow_root_level;
348 	union kvm_mmu_page_role base_role;
349 	bool direct_map;
350 
351 	/*
352 	 * Bitmap; bit set = permission fault
353 	 * Byte index: page fault error code [4:1]
354 	 * Bit index: pte permissions in ACC_* format
355 	 */
356 	u8 permissions[16];
357 
358 	/*
359 	* The pkru_mask indicates if protection key checks are needed.  It
360 	* consists of 16 domains indexed by page fault error code bits [4:1],
361 	* with PFEC.RSVD replaced by ACC_USER_MASK from the page tables.
362 	* Each domain has 2 bits which are ANDed with AD and WD from PKRU.
363 	*/
364 	u32 pkru_mask;
365 
366 	u64 *pae_root;
367 	u64 *lm_root;
368 
369 	/*
370 	 * check zero bits on shadow page table entries, these
371 	 * bits include not only hardware reserved bits but also
372 	 * the bits spte never used.
373 	 */
374 	struct rsvd_bits_validate shadow_zero_check;
375 
376 	struct rsvd_bits_validate guest_rsvd_check;
377 
378 	/* Can have large pages at levels 2..last_nonleaf_level-1. */
379 	u8 last_nonleaf_level;
380 
381 	bool nx;
382 
383 	u64 pdptrs[4]; /* pae */
384 };
385 
386 enum pmc_type {
387 	KVM_PMC_GP = 0,
388 	KVM_PMC_FIXED,
389 };
390 
391 struct kvm_pmc {
392 	enum pmc_type type;
393 	u8 idx;
394 	u64 counter;
395 	u64 eventsel;
396 	struct perf_event *perf_event;
397 	struct kvm_vcpu *vcpu;
398 };
399 
400 struct kvm_pmu {
401 	unsigned nr_arch_gp_counters;
402 	unsigned nr_arch_fixed_counters;
403 	unsigned available_event_types;
404 	u64 fixed_ctr_ctrl;
405 	u64 global_ctrl;
406 	u64 global_status;
407 	u64 global_ovf_ctrl;
408 	u64 counter_bitmask[2];
409 	u64 global_ctrl_mask;
410 	u64 reserved_bits;
411 	u8 version;
412 	struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
413 	struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED];
414 	struct irq_work irq_work;
415 	u64 reprogram_pmi;
416 };
417 
418 struct kvm_pmu_ops;
419 
420 enum {
421 	KVM_DEBUGREG_BP_ENABLED = 1,
422 	KVM_DEBUGREG_WONT_EXIT = 2,
423 	KVM_DEBUGREG_RELOAD = 4,
424 };
425 
426 struct kvm_mtrr_range {
427 	u64 base;
428 	u64 mask;
429 	struct list_head node;
430 };
431 
432 struct kvm_mtrr {
433 	struct kvm_mtrr_range var_ranges[KVM_NR_VAR_MTRR];
434 	mtrr_type fixed_ranges[KVM_NR_FIXED_MTRR_REGION];
435 	u64 deftype;
436 
437 	struct list_head head;
438 };
439 
440 /* Hyper-V SynIC timer */
441 struct kvm_vcpu_hv_stimer {
442 	struct hrtimer timer;
443 	int index;
444 	u64 config;
445 	u64 count;
446 	u64 exp_time;
447 	struct hv_message msg;
448 	bool msg_pending;
449 };
450 
451 /* Hyper-V synthetic interrupt controller (SynIC)*/
452 struct kvm_vcpu_hv_synic {
453 	u64 version;
454 	u64 control;
455 	u64 msg_page;
456 	u64 evt_page;
457 	atomic64_t sint[HV_SYNIC_SINT_COUNT];
458 	atomic_t sint_to_gsi[HV_SYNIC_SINT_COUNT];
459 	DECLARE_BITMAP(auto_eoi_bitmap, 256);
460 	DECLARE_BITMAP(vec_bitmap, 256);
461 	bool active;
462 };
463 
464 /* Hyper-V per vcpu emulation context */
465 struct kvm_vcpu_hv {
466 	u64 hv_vapic;
467 	s64 runtime_offset;
468 	struct kvm_vcpu_hv_synic synic;
469 	struct kvm_hyperv_exit exit;
470 	struct kvm_vcpu_hv_stimer stimer[HV_SYNIC_STIMER_COUNT];
471 	DECLARE_BITMAP(stimer_pending_bitmap, HV_SYNIC_STIMER_COUNT);
472 };
473 
474 struct kvm_vcpu_arch {
475 	/*
476 	 * rip and regs accesses must go through
477 	 * kvm_{register,rip}_{read,write} functions.
478 	 */
479 	unsigned long regs[NR_VCPU_REGS];
480 	u32 regs_avail;
481 	u32 regs_dirty;
482 
483 	unsigned long cr0;
484 	unsigned long cr0_guest_owned_bits;
485 	unsigned long cr2;
486 	unsigned long cr3;
487 	unsigned long cr4;
488 	unsigned long cr4_guest_owned_bits;
489 	unsigned long cr8;
490 	u32 hflags;
491 	u64 efer;
492 	u64 apic_base;
493 	struct kvm_lapic *apic;    /* kernel irqchip context */
494 	bool apicv_active;
495 	DECLARE_BITMAP(ioapic_handled_vectors, 256);
496 	unsigned long apic_attention;
497 	int32_t apic_arb_prio;
498 	int mp_state;
499 	u64 ia32_misc_enable_msr;
500 	u64 smbase;
501 	bool tpr_access_reporting;
502 	u64 ia32_xss;
503 
504 	/*
505 	 * Paging state of the vcpu
506 	 *
507 	 * If the vcpu runs in guest mode with two level paging this still saves
508 	 * the paging mode of the l1 guest. This context is always used to
509 	 * handle faults.
510 	 */
511 	struct kvm_mmu mmu;
512 
513 	/*
514 	 * Paging state of an L2 guest (used for nested npt)
515 	 *
516 	 * This context will save all necessary information to walk page tables
517 	 * of the an L2 guest. This context is only initialized for page table
518 	 * walking and not for faulting since we never handle l2 page faults on
519 	 * the host.
520 	 */
521 	struct kvm_mmu nested_mmu;
522 
523 	/*
524 	 * Pointer to the mmu context currently used for
525 	 * gva_to_gpa translations.
526 	 */
527 	struct kvm_mmu *walk_mmu;
528 
529 	struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
530 	struct kvm_mmu_memory_cache mmu_page_cache;
531 	struct kvm_mmu_memory_cache mmu_page_header_cache;
532 
533 	struct fpu guest_fpu;
534 	u64 xcr0;
535 	u64 guest_supported_xcr0;
536 	u32 guest_xstate_size;
537 
538 	struct kvm_pio_request pio;
539 	void *pio_data;
540 
541 	u8 event_exit_inst_len;
542 
543 	struct kvm_queued_exception {
544 		bool pending;
545 		bool has_error_code;
546 		bool reinject;
547 		u8 nr;
548 		u32 error_code;
549 	} exception;
550 
551 	struct kvm_queued_interrupt {
552 		bool pending;
553 		bool soft;
554 		u8 nr;
555 	} interrupt;
556 
557 	int halt_request; /* real mode on Intel only */
558 
559 	int cpuid_nent;
560 	struct kvm_cpuid_entry2 cpuid_entries[KVM_MAX_CPUID_ENTRIES];
561 
562 	int maxphyaddr;
563 
564 	/* emulate context */
565 
566 	struct x86_emulate_ctxt emulate_ctxt;
567 	bool emulate_regs_need_sync_to_vcpu;
568 	bool emulate_regs_need_sync_from_vcpu;
569 	int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
570 
571 	gpa_t time;
572 	struct pvclock_vcpu_time_info hv_clock;
573 	unsigned int hw_tsc_khz;
574 	struct gfn_to_hva_cache pv_time;
575 	bool pv_time_enabled;
576 	/* set guest stopped flag in pvclock flags field */
577 	bool pvclock_set_guest_stopped_request;
578 
579 	struct {
580 		u64 msr_val;
581 		u64 last_steal;
582 		struct gfn_to_hva_cache stime;
583 		struct kvm_steal_time steal;
584 	} st;
585 
586 	u64 tsc_offset;
587 	u64 last_guest_tsc;
588 	u64 last_host_tsc;
589 	u64 tsc_offset_adjustment;
590 	u64 this_tsc_nsec;
591 	u64 this_tsc_write;
592 	u64 this_tsc_generation;
593 	bool tsc_catchup;
594 	bool tsc_always_catchup;
595 	s8 virtual_tsc_shift;
596 	u32 virtual_tsc_mult;
597 	u32 virtual_tsc_khz;
598 	s64 ia32_tsc_adjust_msr;
599 	u64 tsc_scaling_ratio;
600 
601 	atomic_t nmi_queued;  /* unprocessed asynchronous NMIs */
602 	unsigned nmi_pending; /* NMI queued after currently running handler */
603 	bool nmi_injected;    /* Trying to inject an NMI this entry */
604 	bool smi_pending;    /* SMI queued after currently running handler */
605 
606 	struct kvm_mtrr mtrr_state;
607 	u64 pat;
608 
609 	unsigned switch_db_regs;
610 	unsigned long db[KVM_NR_DB_REGS];
611 	unsigned long dr6;
612 	unsigned long dr7;
613 	unsigned long eff_db[KVM_NR_DB_REGS];
614 	unsigned long guest_debug_dr7;
615 
616 	u64 mcg_cap;
617 	u64 mcg_status;
618 	u64 mcg_ctl;
619 	u64 mcg_ext_ctl;
620 	u64 *mce_banks;
621 
622 	/* Cache MMIO info */
623 	u64 mmio_gva;
624 	unsigned access;
625 	gfn_t mmio_gfn;
626 	u64 mmio_gen;
627 
628 	struct kvm_pmu pmu;
629 
630 	/* used for guest single stepping over the given code position */
631 	unsigned long singlestep_rip;
632 
633 	struct kvm_vcpu_hv hyperv;
634 
635 	cpumask_var_t wbinvd_dirty_mask;
636 
637 	unsigned long last_retry_eip;
638 	unsigned long last_retry_addr;
639 
640 	struct {
641 		bool halted;
642 		gfn_t gfns[roundup_pow_of_two(ASYNC_PF_PER_VCPU)];
643 		struct gfn_to_hva_cache data;
644 		u64 msr_val;
645 		u32 id;
646 		bool send_user_only;
647 	} apf;
648 
649 	/* OSVW MSRs (AMD only) */
650 	struct {
651 		u64 length;
652 		u64 status;
653 	} osvw;
654 
655 	struct {
656 		u64 msr_val;
657 		struct gfn_to_hva_cache data;
658 	} pv_eoi;
659 
660 	/*
661 	 * Indicate whether the access faults on its page table in guest
662 	 * which is set when fix page fault and used to detect unhandeable
663 	 * instruction.
664 	 */
665 	bool write_fault_to_shadow_pgtable;
666 
667 	/* set at EPT violation at this point */
668 	unsigned long exit_qualification;
669 
670 	/* pv related host specific info */
671 	struct {
672 		bool pv_unhalted;
673 	} pv;
674 
675 	int pending_ioapic_eoi;
676 	int pending_external_vector;
677 
678 	/* GPA available (AMD only) */
679 	bool gpa_available;
680 };
681 
682 struct kvm_lpage_info {
683 	int disallow_lpage;
684 };
685 
686 struct kvm_arch_memory_slot {
687 	struct kvm_rmap_head *rmap[KVM_NR_PAGE_SIZES];
688 	struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
689 	unsigned short *gfn_track[KVM_PAGE_TRACK_MAX];
690 };
691 
692 /*
693  * We use as the mode the number of bits allocated in the LDR for the
694  * logical processor ID.  It happens that these are all powers of two.
695  * This makes it is very easy to detect cases where the APICs are
696  * configured for multiple modes; in that case, we cannot use the map and
697  * hence cannot use kvm_irq_delivery_to_apic_fast either.
698  */
699 #define KVM_APIC_MODE_XAPIC_CLUSTER          4
700 #define KVM_APIC_MODE_XAPIC_FLAT             8
701 #define KVM_APIC_MODE_X2APIC                16
702 
703 struct kvm_apic_map {
704 	struct rcu_head rcu;
705 	u8 mode;
706 	u32 max_apic_id;
707 	union {
708 		struct kvm_lapic *xapic_flat_map[8];
709 		struct kvm_lapic *xapic_cluster_map[16][4];
710 	};
711 	struct kvm_lapic *phys_map[];
712 };
713 
714 /* Hyper-V emulation context */
715 struct kvm_hv {
716 	struct mutex hv_lock;
717 	u64 hv_guest_os_id;
718 	u64 hv_hypercall;
719 	u64 hv_tsc_page;
720 
721 	/* Hyper-v based guest crash (NT kernel bugcheck) parameters */
722 	u64 hv_crash_param[HV_X64_MSR_CRASH_PARAMS];
723 	u64 hv_crash_ctl;
724 
725 	HV_REFERENCE_TSC_PAGE tsc_ref;
726 };
727 
728 enum kvm_irqchip_mode {
729 	KVM_IRQCHIP_NONE,
730 	KVM_IRQCHIP_KERNEL,       /* created with KVM_CREATE_IRQCHIP */
731 	KVM_IRQCHIP_SPLIT,        /* created with KVM_CAP_SPLIT_IRQCHIP */
732 };
733 
734 struct kvm_arch {
735 	unsigned int n_used_mmu_pages;
736 	unsigned int n_requested_mmu_pages;
737 	unsigned int n_max_mmu_pages;
738 	unsigned int indirect_shadow_pages;
739 	unsigned long mmu_valid_gen;
740 	struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
741 	/*
742 	 * Hash table of struct kvm_mmu_page.
743 	 */
744 	struct list_head active_mmu_pages;
745 	struct list_head zapped_obsolete_pages;
746 	struct kvm_page_track_notifier_node mmu_sp_tracker;
747 	struct kvm_page_track_notifier_head track_notifier_head;
748 
749 	struct list_head assigned_dev_head;
750 	struct iommu_domain *iommu_domain;
751 	bool iommu_noncoherent;
752 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA
753 	atomic_t noncoherent_dma_count;
754 #define __KVM_HAVE_ARCH_ASSIGNED_DEVICE
755 	atomic_t assigned_device_count;
756 	struct kvm_pic *vpic;
757 	struct kvm_ioapic *vioapic;
758 	struct kvm_pit *vpit;
759 	atomic_t vapics_in_nmi_mode;
760 	struct mutex apic_map_lock;
761 	struct kvm_apic_map *apic_map;
762 
763 	unsigned int tss_addr;
764 	bool apic_access_page_done;
765 
766 	gpa_t wall_clock;
767 
768 	bool ept_identity_pagetable_done;
769 	gpa_t ept_identity_map_addr;
770 
771 	unsigned long irq_sources_bitmap;
772 	s64 kvmclock_offset;
773 	raw_spinlock_t tsc_write_lock;
774 	u64 last_tsc_nsec;
775 	u64 last_tsc_write;
776 	u32 last_tsc_khz;
777 	u64 cur_tsc_nsec;
778 	u64 cur_tsc_write;
779 	u64 cur_tsc_offset;
780 	u64 cur_tsc_generation;
781 	int nr_vcpus_matched_tsc;
782 
783 	spinlock_t pvclock_gtod_sync_lock;
784 	bool use_master_clock;
785 	u64 master_kernel_ns;
786 	u64 master_cycle_now;
787 	struct delayed_work kvmclock_update_work;
788 	struct delayed_work kvmclock_sync_work;
789 
790 	struct kvm_xen_hvm_config xen_hvm_config;
791 
792 	/* reads protected by irq_srcu, writes by irq_lock */
793 	struct hlist_head mask_notifier_list;
794 
795 	struct kvm_hv hyperv;
796 
797 	#ifdef CONFIG_KVM_MMU_AUDIT
798 	int audit_point;
799 	#endif
800 
801 	bool boot_vcpu_runs_old_kvmclock;
802 	u32 bsp_vcpu_id;
803 
804 	u64 disabled_quirks;
805 
806 	enum kvm_irqchip_mode irqchip_mode;
807 	u8 nr_reserved_ioapic_pins;
808 
809 	bool disabled_lapic_found;
810 
811 	/* Struct members for AVIC */
812 	u32 avic_vm_id;
813 	u32 ldr_mode;
814 	struct page *avic_logical_id_table_page;
815 	struct page *avic_physical_id_table_page;
816 	struct hlist_node hnode;
817 
818 	bool x2apic_format;
819 	bool x2apic_broadcast_quirk_disabled;
820 };
821 
822 struct kvm_vm_stat {
823 	ulong mmu_shadow_zapped;
824 	ulong mmu_pte_write;
825 	ulong mmu_pte_updated;
826 	ulong mmu_pde_zapped;
827 	ulong mmu_flooded;
828 	ulong mmu_recycled;
829 	ulong mmu_cache_miss;
830 	ulong mmu_unsync;
831 	ulong remote_tlb_flush;
832 	ulong lpages;
833 	ulong max_mmu_page_hash_collisions;
834 };
835 
836 struct kvm_vcpu_stat {
837 	u64 pf_fixed;
838 	u64 pf_guest;
839 	u64 tlb_flush;
840 	u64 invlpg;
841 
842 	u64 exits;
843 	u64 io_exits;
844 	u64 mmio_exits;
845 	u64 signal_exits;
846 	u64 irq_window_exits;
847 	u64 nmi_window_exits;
848 	u64 halt_exits;
849 	u64 halt_successful_poll;
850 	u64 halt_attempted_poll;
851 	u64 halt_poll_invalid;
852 	u64 halt_wakeup;
853 	u64 request_irq_exits;
854 	u64 irq_exits;
855 	u64 host_state_reload;
856 	u64 efer_reload;
857 	u64 fpu_reload;
858 	u64 insn_emulation;
859 	u64 insn_emulation_fail;
860 	u64 hypercalls;
861 	u64 irq_injections;
862 	u64 nmi_injections;
863 	u64 req_event;
864 };
865 
866 struct x86_instruction_info;
867 
868 struct msr_data {
869 	bool host_initiated;
870 	u32 index;
871 	u64 data;
872 };
873 
874 struct kvm_lapic_irq {
875 	u32 vector;
876 	u16 delivery_mode;
877 	u16 dest_mode;
878 	bool level;
879 	u16 trig_mode;
880 	u32 shorthand;
881 	u32 dest_id;
882 	bool msi_redir_hint;
883 };
884 
885 struct kvm_x86_ops {
886 	int (*cpu_has_kvm_support)(void);          /* __init */
887 	int (*disabled_by_bios)(void);             /* __init */
888 	int (*hardware_enable)(void);
889 	void (*hardware_disable)(void);
890 	void (*check_processor_compatibility)(void *rtn);
891 	int (*hardware_setup)(void);               /* __init */
892 	void (*hardware_unsetup)(void);            /* __exit */
893 	bool (*cpu_has_accelerated_tpr)(void);
894 	bool (*cpu_has_high_real_mode_segbase)(void);
895 	void (*cpuid_update)(struct kvm_vcpu *vcpu);
896 
897 	int (*vm_init)(struct kvm *kvm);
898 	void (*vm_destroy)(struct kvm *kvm);
899 
900 	/* Create, but do not attach this VCPU */
901 	struct kvm_vcpu *(*vcpu_create)(struct kvm *kvm, unsigned id);
902 	void (*vcpu_free)(struct kvm_vcpu *vcpu);
903 	void (*vcpu_reset)(struct kvm_vcpu *vcpu, bool init_event);
904 
905 	void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
906 	void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
907 	void (*vcpu_put)(struct kvm_vcpu *vcpu);
908 
909 	void (*update_bp_intercept)(struct kvm_vcpu *vcpu);
910 	int (*get_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
911 	int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
912 	u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
913 	void (*get_segment)(struct kvm_vcpu *vcpu,
914 			    struct kvm_segment *var, int seg);
915 	int (*get_cpl)(struct kvm_vcpu *vcpu);
916 	void (*set_segment)(struct kvm_vcpu *vcpu,
917 			    struct kvm_segment *var, int seg);
918 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
919 	void (*decache_cr0_guest_bits)(struct kvm_vcpu *vcpu);
920 	void (*decache_cr3)(struct kvm_vcpu *vcpu);
921 	void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu);
922 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
923 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
924 	int (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
925 	void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
926 	void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
927 	void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
928 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
929 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
930 	u64 (*get_dr6)(struct kvm_vcpu *vcpu);
931 	void (*set_dr6)(struct kvm_vcpu *vcpu, unsigned long value);
932 	void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu);
933 	void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
934 	void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
935 	unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
936 	void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
937 	u32 (*get_pkru)(struct kvm_vcpu *vcpu);
938 
939 	void (*tlb_flush)(struct kvm_vcpu *vcpu);
940 
941 	void (*run)(struct kvm_vcpu *vcpu);
942 	int (*handle_exit)(struct kvm_vcpu *vcpu);
943 	void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
944 	void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
945 	u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu);
946 	void (*patch_hypercall)(struct kvm_vcpu *vcpu,
947 				unsigned char *hypercall_addr);
948 	void (*set_irq)(struct kvm_vcpu *vcpu);
949 	void (*set_nmi)(struct kvm_vcpu *vcpu);
950 	void (*queue_exception)(struct kvm_vcpu *vcpu, unsigned nr,
951 				bool has_error_code, u32 error_code,
952 				bool reinject);
953 	void (*cancel_injection)(struct kvm_vcpu *vcpu);
954 	int (*interrupt_allowed)(struct kvm_vcpu *vcpu);
955 	int (*nmi_allowed)(struct kvm_vcpu *vcpu);
956 	bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
957 	void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
958 	void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
959 	void (*enable_irq_window)(struct kvm_vcpu *vcpu);
960 	void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
961 	bool (*get_enable_apicv)(void);
962 	void (*refresh_apicv_exec_ctrl)(struct kvm_vcpu *vcpu);
963 	void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
964 	void (*hwapic_isr_update)(struct kvm_vcpu *vcpu, int isr);
965 	void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
966 	void (*set_virtual_x2apic_mode)(struct kvm_vcpu *vcpu, bool set);
967 	void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu, hpa_t hpa);
968 	void (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
969 	int (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
970 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
971 	int (*get_tdp_level)(void);
972 	u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
973 	int (*get_lpage_level)(void);
974 	bool (*rdtscp_supported)(void);
975 	bool (*invpcid_supported)(void);
976 
977 	void (*set_tdp_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
978 
979 	void (*set_supported_cpuid)(u32 func, struct kvm_cpuid_entry2 *entry);
980 
981 	bool (*has_wbinvd_exit)(void);
982 
983 	void (*write_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
984 
985 	void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2);
986 
987 	int (*check_intercept)(struct kvm_vcpu *vcpu,
988 			       struct x86_instruction_info *info,
989 			       enum x86_intercept_stage stage);
990 	void (*handle_external_intr)(struct kvm_vcpu *vcpu);
991 	bool (*mpx_supported)(void);
992 	bool (*xsaves_supported)(void);
993 
994 	int (*check_nested_events)(struct kvm_vcpu *vcpu, bool external_intr);
995 
996 	void (*sched_in)(struct kvm_vcpu *kvm, int cpu);
997 
998 	/*
999 	 * Arch-specific dirty logging hooks. These hooks are only supposed to
1000 	 * be valid if the specific arch has hardware-accelerated dirty logging
1001 	 * mechanism. Currently only for PML on VMX.
1002 	 *
1003 	 *  - slot_enable_log_dirty:
1004 	 *	called when enabling log dirty mode for the slot.
1005 	 *  - slot_disable_log_dirty:
1006 	 *	called when disabling log dirty mode for the slot.
1007 	 *	also called when slot is created with log dirty disabled.
1008 	 *  - flush_log_dirty:
1009 	 *	called before reporting dirty_bitmap to userspace.
1010 	 *  - enable_log_dirty_pt_masked:
1011 	 *	called when reenabling log dirty for the GFNs in the mask after
1012 	 *	corresponding bits are cleared in slot->dirty_bitmap.
1013 	 */
1014 	void (*slot_enable_log_dirty)(struct kvm *kvm,
1015 				      struct kvm_memory_slot *slot);
1016 	void (*slot_disable_log_dirty)(struct kvm *kvm,
1017 				       struct kvm_memory_slot *slot);
1018 	void (*flush_log_dirty)(struct kvm *kvm);
1019 	void (*enable_log_dirty_pt_masked)(struct kvm *kvm,
1020 					   struct kvm_memory_slot *slot,
1021 					   gfn_t offset, unsigned long mask);
1022 	/* pmu operations of sub-arch */
1023 	const struct kvm_pmu_ops *pmu_ops;
1024 
1025 	/*
1026 	 * Architecture specific hooks for vCPU blocking due to
1027 	 * HLT instruction.
1028 	 * Returns for .pre_block():
1029 	 *    - 0 means continue to block the vCPU.
1030 	 *    - 1 means we cannot block the vCPU since some event
1031 	 *        happens during this period, such as, 'ON' bit in
1032 	 *        posted-interrupts descriptor is set.
1033 	 */
1034 	int (*pre_block)(struct kvm_vcpu *vcpu);
1035 	void (*post_block)(struct kvm_vcpu *vcpu);
1036 
1037 	void (*vcpu_blocking)(struct kvm_vcpu *vcpu);
1038 	void (*vcpu_unblocking)(struct kvm_vcpu *vcpu);
1039 
1040 	int (*update_pi_irte)(struct kvm *kvm, unsigned int host_irq,
1041 			      uint32_t guest_irq, bool set);
1042 	void (*apicv_post_state_restore)(struct kvm_vcpu *vcpu);
1043 
1044 	int (*set_hv_timer)(struct kvm_vcpu *vcpu, u64 guest_deadline_tsc);
1045 	void (*cancel_hv_timer)(struct kvm_vcpu *vcpu);
1046 
1047 	void (*setup_mce)(struct kvm_vcpu *vcpu);
1048 };
1049 
1050 struct kvm_arch_async_pf {
1051 	u32 token;
1052 	gfn_t gfn;
1053 	unsigned long cr3;
1054 	bool direct_map;
1055 };
1056 
1057 extern struct kvm_x86_ops *kvm_x86_ops;
1058 
1059 int kvm_mmu_module_init(void);
1060 void kvm_mmu_module_exit(void);
1061 
1062 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
1063 int kvm_mmu_create(struct kvm_vcpu *vcpu);
1064 void kvm_mmu_setup(struct kvm_vcpu *vcpu);
1065 void kvm_mmu_init_vm(struct kvm *kvm);
1066 void kvm_mmu_uninit_vm(struct kvm *kvm);
1067 void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
1068 		u64 dirty_mask, u64 nx_mask, u64 x_mask, u64 p_mask,
1069 		u64 acc_track_mask);
1070 
1071 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
1072 void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
1073 				      struct kvm_memory_slot *memslot);
1074 void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
1075 				   const struct kvm_memory_slot *memslot);
1076 void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
1077 				   struct kvm_memory_slot *memslot);
1078 void kvm_mmu_slot_largepage_remove_write_access(struct kvm *kvm,
1079 					struct kvm_memory_slot *memslot);
1080 void kvm_mmu_slot_set_dirty(struct kvm *kvm,
1081 			    struct kvm_memory_slot *memslot);
1082 void kvm_mmu_clear_dirty_pt_masked(struct kvm *kvm,
1083 				   struct kvm_memory_slot *slot,
1084 				   gfn_t gfn_offset, unsigned long mask);
1085 void kvm_mmu_zap_all(struct kvm *kvm);
1086 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, struct kvm_memslots *slots);
1087 unsigned int kvm_mmu_calculate_mmu_pages(struct kvm *kvm);
1088 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int kvm_nr_mmu_pages);
1089 
1090 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
1091 bool pdptrs_changed(struct kvm_vcpu *vcpu);
1092 
1093 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
1094 			  const void *val, int bytes);
1095 
1096 struct kvm_irq_mask_notifier {
1097 	void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
1098 	int irq;
1099 	struct hlist_node link;
1100 };
1101 
1102 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
1103 				    struct kvm_irq_mask_notifier *kimn);
1104 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
1105 				      struct kvm_irq_mask_notifier *kimn);
1106 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
1107 			     bool mask);
1108 
1109 extern bool tdp_enabled;
1110 
1111 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
1112 
1113 /* control of guest tsc rate supported? */
1114 extern bool kvm_has_tsc_control;
1115 /* maximum supported tsc_khz for guests */
1116 extern u32  kvm_max_guest_tsc_khz;
1117 /* number of bits of the fractional part of the TSC scaling ratio */
1118 extern u8   kvm_tsc_scaling_ratio_frac_bits;
1119 /* maximum allowed value of TSC scaling ratio */
1120 extern u64  kvm_max_tsc_scaling_ratio;
1121 /* 1ull << kvm_tsc_scaling_ratio_frac_bits */
1122 extern u64  kvm_default_tsc_scaling_ratio;
1123 
1124 extern u64 kvm_mce_cap_supported;
1125 
1126 enum emulation_result {
1127 	EMULATE_DONE,         /* no further processing */
1128 	EMULATE_USER_EXIT,    /* kvm_run ready for userspace exit */
1129 	EMULATE_FAIL,         /* can't emulate this instruction */
1130 };
1131 
1132 #define EMULTYPE_NO_DECODE	    (1 << 0)
1133 #define EMULTYPE_TRAP_UD	    (1 << 1)
1134 #define EMULTYPE_SKIP		    (1 << 2)
1135 #define EMULTYPE_RETRY		    (1 << 3)
1136 #define EMULTYPE_NO_REEXECUTE	    (1 << 4)
1137 int x86_emulate_instruction(struct kvm_vcpu *vcpu, unsigned long cr2,
1138 			    int emulation_type, void *insn, int insn_len);
1139 
1140 static inline int emulate_instruction(struct kvm_vcpu *vcpu,
1141 			int emulation_type)
1142 {
1143 	return x86_emulate_instruction(vcpu, 0, emulation_type, NULL, 0);
1144 }
1145 
1146 void kvm_enable_efer_bits(u64);
1147 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
1148 int kvm_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr);
1149 int kvm_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr);
1150 
1151 struct x86_emulate_ctxt;
1152 
1153 int kvm_fast_pio_out(struct kvm_vcpu *vcpu, int size, unsigned short port);
1154 int kvm_fast_pio_in(struct kvm_vcpu *vcpu, int size, unsigned short port);
1155 int kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
1156 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
1157 int kvm_vcpu_halt(struct kvm_vcpu *vcpu);
1158 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
1159 
1160 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
1161 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
1162 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector);
1163 
1164 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
1165 		    int reason, bool has_error_code, u32 error_code);
1166 
1167 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
1168 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
1169 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
1170 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
1171 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
1172 int kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
1173 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
1174 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
1175 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
1176 int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr);
1177 
1178 int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1179 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1180 
1181 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
1182 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
1183 bool kvm_rdpmc(struct kvm_vcpu *vcpu);
1184 
1185 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1186 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1187 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1188 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1189 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
1190 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1191 			    gfn_t gfn, void *data, int offset, int len,
1192 			    u32 access);
1193 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
1194 bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr);
1195 
1196 static inline int __kvm_irq_line_state(unsigned long *irq_state,
1197 				       int irq_source_id, int level)
1198 {
1199 	/* Logical OR for level trig interrupt */
1200 	if (level)
1201 		__set_bit(irq_source_id, irq_state);
1202 	else
1203 		__clear_bit(irq_source_id, irq_state);
1204 
1205 	return !!(*irq_state);
1206 }
1207 
1208 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
1209 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
1210 
1211 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
1212 
1213 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
1214 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
1215 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
1216 int kvm_mmu_load(struct kvm_vcpu *vcpu);
1217 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
1218 void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu);
1219 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1220 			   struct x86_exception *exception);
1221 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
1222 			      struct x86_exception *exception);
1223 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
1224 			       struct x86_exception *exception);
1225 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
1226 			       struct x86_exception *exception);
1227 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
1228 				struct x86_exception *exception);
1229 
1230 void kvm_vcpu_deactivate_apicv(struct kvm_vcpu *vcpu);
1231 
1232 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
1233 
1234 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva, u64 error_code,
1235 		       void *insn, int insn_len);
1236 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
1237 void kvm_mmu_new_cr3(struct kvm_vcpu *vcpu);
1238 
1239 void kvm_enable_tdp(void);
1240 void kvm_disable_tdp(void);
1241 
1242 static inline gpa_t translate_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1243 				  struct x86_exception *exception)
1244 {
1245 	return gpa;
1246 }
1247 
1248 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
1249 {
1250 	struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
1251 
1252 	return (struct kvm_mmu_page *)page_private(page);
1253 }
1254 
1255 static inline u16 kvm_read_ldt(void)
1256 {
1257 	u16 ldt;
1258 	asm("sldt %0" : "=g"(ldt));
1259 	return ldt;
1260 }
1261 
1262 static inline void kvm_load_ldt(u16 sel)
1263 {
1264 	asm("lldt %0" : : "rm"(sel));
1265 }
1266 
1267 #ifdef CONFIG_X86_64
1268 static inline unsigned long read_msr(unsigned long msr)
1269 {
1270 	u64 value;
1271 
1272 	rdmsrl(msr, value);
1273 	return value;
1274 }
1275 #endif
1276 
1277 static inline u32 get_rdx_init_val(void)
1278 {
1279 	return 0x600; /* P6 family */
1280 }
1281 
1282 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
1283 {
1284 	kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
1285 }
1286 
1287 static inline u64 get_canonical(u64 la)
1288 {
1289 	return ((int64_t)la << 16) >> 16;
1290 }
1291 
1292 static inline bool is_noncanonical_address(u64 la)
1293 {
1294 #ifdef CONFIG_X86_64
1295 	return get_canonical(la) != la;
1296 #else
1297 	return false;
1298 #endif
1299 }
1300 
1301 #define TSS_IOPB_BASE_OFFSET 0x66
1302 #define TSS_BASE_SIZE 0x68
1303 #define TSS_IOPB_SIZE (65536 / 8)
1304 #define TSS_REDIRECTION_SIZE (256 / 8)
1305 #define RMODE_TSS_SIZE							\
1306 	(TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
1307 
1308 enum {
1309 	TASK_SWITCH_CALL = 0,
1310 	TASK_SWITCH_IRET = 1,
1311 	TASK_SWITCH_JMP = 2,
1312 	TASK_SWITCH_GATE = 3,
1313 };
1314 
1315 #define HF_GIF_MASK		(1 << 0)
1316 #define HF_HIF_MASK		(1 << 1)
1317 #define HF_VINTR_MASK		(1 << 2)
1318 #define HF_NMI_MASK		(1 << 3)
1319 #define HF_IRET_MASK		(1 << 4)
1320 #define HF_GUEST_MASK		(1 << 5) /* VCPU is in guest-mode */
1321 #define HF_SMM_MASK		(1 << 6)
1322 #define HF_SMM_INSIDE_NMI_MASK	(1 << 7)
1323 
1324 #define __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
1325 #define KVM_ADDRESS_SPACE_NUM 2
1326 
1327 #define kvm_arch_vcpu_memslots_id(vcpu) ((vcpu)->arch.hflags & HF_SMM_MASK ? 1 : 0)
1328 #define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, (role).smm)
1329 
1330 /*
1331  * Hardware virtualization extension instructions may fault if a
1332  * reboot turns off virtualization while processes are running.
1333  * Trap the fault and ignore the instruction if that happens.
1334  */
1335 asmlinkage void kvm_spurious_fault(void);
1336 
1337 #define ____kvm_handle_fault_on_reboot(insn, cleanup_insn)	\
1338 	"666: " insn "\n\t" \
1339 	"668: \n\t"                           \
1340 	".pushsection .fixup, \"ax\" \n" \
1341 	"667: \n\t" \
1342 	cleanup_insn "\n\t"		      \
1343 	"cmpb $0, kvm_rebooting \n\t"	      \
1344 	"jne 668b \n\t"      		      \
1345 	__ASM_SIZE(push) " $666b \n\t"	      \
1346 	"call kvm_spurious_fault \n\t"	      \
1347 	".popsection \n\t" \
1348 	_ASM_EXTABLE(666b, 667b)
1349 
1350 #define __kvm_handle_fault_on_reboot(insn)		\
1351 	____kvm_handle_fault_on_reboot(insn, "")
1352 
1353 #define KVM_ARCH_WANT_MMU_NOTIFIER
1354 int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
1355 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end);
1356 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
1357 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
1358 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
1359 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1360 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1361 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1362 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1363 void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event);
1364 void kvm_vcpu_reload_apic_access_page(struct kvm_vcpu *vcpu);
1365 void kvm_arch_mmu_notifier_invalidate_page(struct kvm *kvm,
1366 					   unsigned long address);
1367 
1368 void kvm_define_shared_msr(unsigned index, u32 msr);
1369 int kvm_set_shared_msr(unsigned index, u64 val, u64 mask);
1370 
1371 u64 kvm_scale_tsc(struct kvm_vcpu *vcpu, u64 tsc);
1372 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc);
1373 
1374 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu);
1375 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1376 
1377 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
1378 void kvm_make_scan_ioapic_request(struct kvm *kvm);
1379 
1380 void kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1381 				     struct kvm_async_pf *work);
1382 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1383 				 struct kvm_async_pf *work);
1384 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1385 			       struct kvm_async_pf *work);
1386 bool kvm_arch_can_inject_async_page_present(struct kvm_vcpu *vcpu);
1387 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1388 
1389 int kvm_skip_emulated_instruction(struct kvm_vcpu *vcpu);
1390 int kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1391 
1392 int kvm_is_in_guest(void);
1393 
1394 int __x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa, u32 size);
1395 int x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa, u32 size);
1396 bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu);
1397 bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu);
1398 
1399 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
1400 			     struct kvm_vcpu **dest_vcpu);
1401 
1402 void kvm_set_msi_irq(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e,
1403 		     struct kvm_lapic_irq *irq);
1404 
1405 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu)
1406 {
1407 	if (kvm_x86_ops->vcpu_blocking)
1408 		kvm_x86_ops->vcpu_blocking(vcpu);
1409 }
1410 
1411 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu)
1412 {
1413 	if (kvm_x86_ops->vcpu_unblocking)
1414 		kvm_x86_ops->vcpu_unblocking(vcpu);
1415 }
1416 
1417 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
1418 
1419 static inline int kvm_cpu_get_apicid(int mps_cpu)
1420 {
1421 #ifdef CONFIG_X86_LOCAL_APIC
1422 	return __default_cpu_present_to_apicid(mps_cpu);
1423 #else
1424 	WARN_ON_ONCE(1);
1425 	return BAD_APICID;
1426 #endif
1427 }
1428 
1429 #endif /* _ASM_X86_KVM_HOST_H */
1430