xref: /linux/arch/x86/include/asm/kvm_host.h (revision e58e871becec2d3b04ed91c0c16fe8deac9c9dfa)
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_HALT_POLL_NS_DEFAULT 200000
47 
48 #define KVM_IRQCHIP_NUM_PINS  KVM_IOAPIC_NUM_PINS
49 
50 /* x86-specific vcpu->requests bit members */
51 #define KVM_REQ_MIGRATE_TIMER      8
52 #define KVM_REQ_REPORT_TPR_ACCESS  9
53 #define KVM_REQ_TRIPLE_FAULT      10
54 #define KVM_REQ_MMU_SYNC          11
55 #define KVM_REQ_CLOCK_UPDATE      12
56 #define KVM_REQ_EVENT             14
57 #define KVM_REQ_APF_HALT          15
58 #define KVM_REQ_STEAL_UPDATE      16
59 #define KVM_REQ_NMI               17
60 #define KVM_REQ_PMU               18
61 #define KVM_REQ_PMI               19
62 #define KVM_REQ_SMI               20
63 #define KVM_REQ_MASTERCLOCK_UPDATE 21
64 #define KVM_REQ_MCLOCK_INPROGRESS (22 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
65 #define KVM_REQ_SCAN_IOAPIC       (23 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
66 #define KVM_REQ_GLOBAL_CLOCK_UPDATE 24
67 #define KVM_REQ_APIC_PAGE_RELOAD  (25 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
68 #define KVM_REQ_HV_CRASH          26
69 #define KVM_REQ_IOAPIC_EOI_EXIT   27
70 #define KVM_REQ_HV_RESET          28
71 #define KVM_REQ_HV_EXIT           29
72 #define KVM_REQ_HV_STIMER         30
73 
74 #define CR0_RESERVED_BITS                                               \
75 	(~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \
76 			  | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \
77 			  | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG))
78 
79 #define CR3_L_MODE_RESERVED_BITS 0xFFFFFF0000000000ULL
80 #define CR3_PCID_INVD		 BIT_64(63)
81 #define CR4_RESERVED_BITS                                               \
82 	(~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\
83 			  | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE     \
84 			  | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR | X86_CR4_PCIDE \
85 			  | X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_FSGSBASE \
86 			  | X86_CR4_OSXMMEXCPT | X86_CR4_VMXE | X86_CR4_SMAP \
87 			  | X86_CR4_PKE))
88 
89 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR)
90 
91 
92 
93 #define INVALID_PAGE (~(hpa_t)0)
94 #define VALID_PAGE(x) ((x) != INVALID_PAGE)
95 
96 #define UNMAPPED_GVA (~(gpa_t)0)
97 
98 /* KVM Hugepage definitions for x86 */
99 #define KVM_NR_PAGE_SIZES	3
100 #define KVM_HPAGE_GFN_SHIFT(x)	(((x) - 1) * 9)
101 #define KVM_HPAGE_SHIFT(x)	(PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
102 #define KVM_HPAGE_SIZE(x)	(1UL << KVM_HPAGE_SHIFT(x))
103 #define KVM_HPAGE_MASK(x)	(~(KVM_HPAGE_SIZE(x) - 1))
104 #define KVM_PAGES_PER_HPAGE(x)	(KVM_HPAGE_SIZE(x) / PAGE_SIZE)
105 
106 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
107 {
108 	/* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */
109 	return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
110 		(base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
111 }
112 
113 #define KVM_PERMILLE_MMU_PAGES 20
114 #define KVM_MIN_ALLOC_MMU_PAGES 64
115 #define KVM_MMU_HASH_SHIFT 12
116 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
117 #define KVM_MIN_FREE_MMU_PAGES 5
118 #define KVM_REFILL_PAGES 25
119 #define KVM_MAX_CPUID_ENTRIES 80
120 #define KVM_NR_FIXED_MTRR_REGION 88
121 #define KVM_NR_VAR_MTRR 8
122 
123 #define ASYNC_PF_PER_VCPU 64
124 
125 enum kvm_reg {
126 	VCPU_REGS_RAX = 0,
127 	VCPU_REGS_RCX = 1,
128 	VCPU_REGS_RDX = 2,
129 	VCPU_REGS_RBX = 3,
130 	VCPU_REGS_RSP = 4,
131 	VCPU_REGS_RBP = 5,
132 	VCPU_REGS_RSI = 6,
133 	VCPU_REGS_RDI = 7,
134 #ifdef CONFIG_X86_64
135 	VCPU_REGS_R8 = 8,
136 	VCPU_REGS_R9 = 9,
137 	VCPU_REGS_R10 = 10,
138 	VCPU_REGS_R11 = 11,
139 	VCPU_REGS_R12 = 12,
140 	VCPU_REGS_R13 = 13,
141 	VCPU_REGS_R14 = 14,
142 	VCPU_REGS_R15 = 15,
143 #endif
144 	VCPU_REGS_RIP,
145 	NR_VCPU_REGS
146 };
147 
148 enum kvm_reg_ex {
149 	VCPU_EXREG_PDPTR = NR_VCPU_REGS,
150 	VCPU_EXREG_CR3,
151 	VCPU_EXREG_RFLAGS,
152 	VCPU_EXREG_SEGMENTS,
153 };
154 
155 enum {
156 	VCPU_SREG_ES,
157 	VCPU_SREG_CS,
158 	VCPU_SREG_SS,
159 	VCPU_SREG_DS,
160 	VCPU_SREG_FS,
161 	VCPU_SREG_GS,
162 	VCPU_SREG_TR,
163 	VCPU_SREG_LDTR,
164 };
165 
166 #include <asm/kvm_emulate.h>
167 
168 #define KVM_NR_MEM_OBJS 40
169 
170 #define KVM_NR_DB_REGS	4
171 
172 #define DR6_BD		(1 << 13)
173 #define DR6_BS		(1 << 14)
174 #define DR6_RTM		(1 << 16)
175 #define DR6_FIXED_1	0xfffe0ff0
176 #define DR6_INIT	0xffff0ff0
177 #define DR6_VOLATILE	0x0001e00f
178 
179 #define DR7_BP_EN_MASK	0x000000ff
180 #define DR7_GE		(1 << 9)
181 #define DR7_GD		(1 << 13)
182 #define DR7_FIXED_1	0x00000400
183 #define DR7_VOLATILE	0xffff2bff
184 
185 #define PFERR_PRESENT_BIT 0
186 #define PFERR_WRITE_BIT 1
187 #define PFERR_USER_BIT 2
188 #define PFERR_RSVD_BIT 3
189 #define PFERR_FETCH_BIT 4
190 #define PFERR_PK_BIT 5
191 #define PFERR_GUEST_FINAL_BIT 32
192 #define PFERR_GUEST_PAGE_BIT 33
193 
194 #define PFERR_PRESENT_MASK (1U << PFERR_PRESENT_BIT)
195 #define PFERR_WRITE_MASK (1U << PFERR_WRITE_BIT)
196 #define PFERR_USER_MASK (1U << PFERR_USER_BIT)
197 #define PFERR_RSVD_MASK (1U << PFERR_RSVD_BIT)
198 #define PFERR_FETCH_MASK (1U << PFERR_FETCH_BIT)
199 #define PFERR_PK_MASK (1U << PFERR_PK_BIT)
200 #define PFERR_GUEST_FINAL_MASK (1ULL << PFERR_GUEST_FINAL_BIT)
201 #define PFERR_GUEST_PAGE_MASK (1ULL << PFERR_GUEST_PAGE_BIT)
202 
203 #define PFERR_NESTED_GUEST_PAGE (PFERR_GUEST_PAGE_MASK |	\
204 				 PFERR_USER_MASK |		\
205 				 PFERR_WRITE_MASK |		\
206 				 PFERR_PRESENT_MASK)
207 
208 /*
209  * The mask used to denote special SPTEs, which can be either MMIO SPTEs or
210  * Access Tracking SPTEs. We use bit 62 instead of bit 63 to avoid conflicting
211  * with the SVE bit in EPT PTEs.
212  */
213 #define SPTE_SPECIAL_MASK (1ULL << 62)
214 
215 /* apic attention bits */
216 #define KVM_APIC_CHECK_VAPIC	0
217 /*
218  * The following bit is set with PV-EOI, unset on EOI.
219  * We detect PV-EOI changes by guest by comparing
220  * this bit with PV-EOI in guest memory.
221  * See the implementation in apic_update_pv_eoi.
222  */
223 #define KVM_APIC_PV_EOI_PENDING	1
224 
225 struct kvm_kernel_irq_routing_entry;
226 
227 /*
228  * We don't want allocation failures within the mmu code, so we preallocate
229  * enough memory for a single page fault in a cache.
230  */
231 struct kvm_mmu_memory_cache {
232 	int nobjs;
233 	void *objects[KVM_NR_MEM_OBJS];
234 };
235 
236 /*
237  * the pages used as guest page table on soft mmu are tracked by
238  * kvm_memory_slot.arch.gfn_track which is 16 bits, so the role bits used
239  * by indirect shadow page can not be more than 15 bits.
240  *
241  * Currently, we used 14 bits that are @level, @cr4_pae, @quadrant, @access,
242  * @nxe, @cr0_wp, @smep_andnot_wp and @smap_andnot_wp.
243  */
244 union kvm_mmu_page_role {
245 	unsigned word;
246 	struct {
247 		unsigned level:4;
248 		unsigned cr4_pae:1;
249 		unsigned quadrant:2;
250 		unsigned direct:1;
251 		unsigned access:3;
252 		unsigned invalid:1;
253 		unsigned nxe:1;
254 		unsigned cr0_wp:1;
255 		unsigned smep_andnot_wp:1;
256 		unsigned smap_andnot_wp:1;
257 		unsigned :8;
258 
259 		/*
260 		 * This is left at the top of the word so that
261 		 * kvm_memslots_for_spte_role can extract it with a
262 		 * simple shift.  While there is room, give it a whole
263 		 * byte so it is also faster to load it from memory.
264 		 */
265 		unsigned smm:8;
266 	};
267 };
268 
269 struct kvm_rmap_head {
270 	unsigned long val;
271 };
272 
273 struct kvm_mmu_page {
274 	struct list_head link;
275 	struct hlist_node hash_link;
276 
277 	/*
278 	 * The following two entries are used to key the shadow page in the
279 	 * hash table.
280 	 */
281 	gfn_t gfn;
282 	union kvm_mmu_page_role role;
283 
284 	u64 *spt;
285 	/* hold the gfn of each spte inside spt */
286 	gfn_t *gfns;
287 	bool unsync;
288 	int root_count;          /* Currently serving as active root */
289 	unsigned int unsync_children;
290 	struct kvm_rmap_head parent_ptes; /* rmap pointers to parent sptes */
291 
292 	/* The page is obsolete if mmu_valid_gen != kvm->arch.mmu_valid_gen.  */
293 	unsigned long mmu_valid_gen;
294 
295 	DECLARE_BITMAP(unsync_child_bitmap, 512);
296 
297 #ifdef CONFIG_X86_32
298 	/*
299 	 * Used out of the mmu-lock to avoid reading spte values while an
300 	 * update is in progress; see the comments in __get_spte_lockless().
301 	 */
302 	int clear_spte_count;
303 #endif
304 
305 	/* Number of writes since the last time traversal visited this page.  */
306 	atomic_t write_flooding_count;
307 };
308 
309 struct kvm_pio_request {
310 	unsigned long count;
311 	int in;
312 	int port;
313 	int size;
314 };
315 
316 struct rsvd_bits_validate {
317 	u64 rsvd_bits_mask[2][4];
318 	u64 bad_mt_xwr;
319 };
320 
321 /*
322  * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
323  * 32-bit).  The kvm_mmu structure abstracts the details of the current mmu
324  * mode.
325  */
326 struct kvm_mmu {
327 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long root);
328 	unsigned long (*get_cr3)(struct kvm_vcpu *vcpu);
329 	u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
330 	int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err,
331 			  bool prefault);
332 	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
333 				  struct x86_exception *fault);
334 	gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva, u32 access,
335 			    struct x86_exception *exception);
336 	gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
337 			       struct x86_exception *exception);
338 	int (*sync_page)(struct kvm_vcpu *vcpu,
339 			 struct kvm_mmu_page *sp);
340 	void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva);
341 	void (*update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
342 			   u64 *spte, const void *pte);
343 	hpa_t root_hpa;
344 	union kvm_mmu_page_role base_role;
345 	u8 root_level;
346 	u8 shadow_root_level;
347 	u8 ept_ad;
348 	bool direct_map;
349 
350 	/*
351 	 * Bitmap; bit set = permission fault
352 	 * Byte index: page fault error code [4:1]
353 	 * Bit index: pte permissions in ACC_* format
354 	 */
355 	u8 permissions[16];
356 
357 	/*
358 	* The pkru_mask indicates if protection key checks are needed.  It
359 	* consists of 16 domains indexed by page fault error code bits [4:1],
360 	* with PFEC.RSVD replaced by ACC_USER_MASK from the page tables.
361 	* Each domain has 2 bits which are ANDed with AD and WD from PKRU.
362 	*/
363 	u32 pkru_mask;
364 
365 	u64 *pae_root;
366 	u64 *lm_root;
367 
368 	/*
369 	 * check zero bits on shadow page table entries, these
370 	 * bits include not only hardware reserved bits but also
371 	 * the bits spte never used.
372 	 */
373 	struct rsvd_bits_validate shadow_zero_check;
374 
375 	struct rsvd_bits_validate guest_rsvd_check;
376 
377 	/* Can have large pages at levels 2..last_nonleaf_level-1. */
378 	u8 last_nonleaf_level;
379 
380 	bool nx;
381 
382 	u64 pdptrs[4]; /* pae */
383 };
384 
385 enum pmc_type {
386 	KVM_PMC_GP = 0,
387 	KVM_PMC_FIXED,
388 };
389 
390 struct kvm_pmc {
391 	enum pmc_type type;
392 	u8 idx;
393 	u64 counter;
394 	u64 eventsel;
395 	struct perf_event *perf_event;
396 	struct kvm_vcpu *vcpu;
397 };
398 
399 struct kvm_pmu {
400 	unsigned nr_arch_gp_counters;
401 	unsigned nr_arch_fixed_counters;
402 	unsigned available_event_types;
403 	u64 fixed_ctr_ctrl;
404 	u64 global_ctrl;
405 	u64 global_status;
406 	u64 global_ovf_ctrl;
407 	u64 counter_bitmask[2];
408 	u64 global_ctrl_mask;
409 	u64 reserved_bits;
410 	u8 version;
411 	struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
412 	struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED];
413 	struct irq_work irq_work;
414 	u64 reprogram_pmi;
415 };
416 
417 struct kvm_pmu_ops;
418 
419 enum {
420 	KVM_DEBUGREG_BP_ENABLED = 1,
421 	KVM_DEBUGREG_WONT_EXIT = 2,
422 	KVM_DEBUGREG_RELOAD = 4,
423 };
424 
425 struct kvm_mtrr_range {
426 	u64 base;
427 	u64 mask;
428 	struct list_head node;
429 };
430 
431 struct kvm_mtrr {
432 	struct kvm_mtrr_range var_ranges[KVM_NR_VAR_MTRR];
433 	mtrr_type fixed_ranges[KVM_NR_FIXED_MTRR_REGION];
434 	u64 deftype;
435 
436 	struct list_head head;
437 };
438 
439 /* Hyper-V SynIC timer */
440 struct kvm_vcpu_hv_stimer {
441 	struct hrtimer timer;
442 	int index;
443 	u64 config;
444 	u64 count;
445 	u64 exp_time;
446 	struct hv_message msg;
447 	bool msg_pending;
448 };
449 
450 /* Hyper-V synthetic interrupt controller (SynIC)*/
451 struct kvm_vcpu_hv_synic {
452 	u64 version;
453 	u64 control;
454 	u64 msg_page;
455 	u64 evt_page;
456 	atomic64_t sint[HV_SYNIC_SINT_COUNT];
457 	atomic_t sint_to_gsi[HV_SYNIC_SINT_COUNT];
458 	DECLARE_BITMAP(auto_eoi_bitmap, 256);
459 	DECLARE_BITMAP(vec_bitmap, 256);
460 	bool active;
461 };
462 
463 /* Hyper-V per vcpu emulation context */
464 struct kvm_vcpu_hv {
465 	u64 hv_vapic;
466 	s64 runtime_offset;
467 	struct kvm_vcpu_hv_synic synic;
468 	struct kvm_hyperv_exit exit;
469 	struct kvm_vcpu_hv_stimer stimer[HV_SYNIC_STIMER_COUNT];
470 	DECLARE_BITMAP(stimer_pending_bitmap, HV_SYNIC_STIMER_COUNT);
471 };
472 
473 struct kvm_vcpu_arch {
474 	/*
475 	 * rip and regs accesses must go through
476 	 * kvm_{register,rip}_{read,write} functions.
477 	 */
478 	unsigned long regs[NR_VCPU_REGS];
479 	u32 regs_avail;
480 	u32 regs_dirty;
481 
482 	unsigned long cr0;
483 	unsigned long cr0_guest_owned_bits;
484 	unsigned long cr2;
485 	unsigned long cr3;
486 	unsigned long cr4;
487 	unsigned long cr4_guest_owned_bits;
488 	unsigned long cr8;
489 	u32 hflags;
490 	u64 efer;
491 	u64 apic_base;
492 	struct kvm_lapic *apic;    /* kernel irqchip context */
493 	bool apicv_active;
494 	DECLARE_BITMAP(ioapic_handled_vectors, 256);
495 	unsigned long apic_attention;
496 	int32_t apic_arb_prio;
497 	int mp_state;
498 	u64 ia32_misc_enable_msr;
499 	u64 smbase;
500 	bool tpr_access_reporting;
501 	u64 ia32_xss;
502 
503 	/*
504 	 * Paging state of the vcpu
505 	 *
506 	 * If the vcpu runs in guest mode with two level paging this still saves
507 	 * the paging mode of the l1 guest. This context is always used to
508 	 * handle faults.
509 	 */
510 	struct kvm_mmu mmu;
511 
512 	/*
513 	 * Paging state of an L2 guest (used for nested npt)
514 	 *
515 	 * This context will save all necessary information to walk page tables
516 	 * of the an L2 guest. This context is only initialized for page table
517 	 * walking and not for faulting since we never handle l2 page faults on
518 	 * the host.
519 	 */
520 	struct kvm_mmu nested_mmu;
521 
522 	/*
523 	 * Pointer to the mmu context currently used for
524 	 * gva_to_gpa translations.
525 	 */
526 	struct kvm_mmu *walk_mmu;
527 
528 	struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
529 	struct kvm_mmu_memory_cache mmu_page_cache;
530 	struct kvm_mmu_memory_cache mmu_page_header_cache;
531 
532 	struct fpu guest_fpu;
533 	u64 xcr0;
534 	u64 guest_supported_xcr0;
535 	u32 guest_xstate_size;
536 
537 	struct kvm_pio_request pio;
538 	void *pio_data;
539 
540 	u8 event_exit_inst_len;
541 
542 	struct kvm_queued_exception {
543 		bool pending;
544 		bool has_error_code;
545 		bool reinject;
546 		u8 nr;
547 		u32 error_code;
548 	} exception;
549 
550 	struct kvm_queued_interrupt {
551 		bool pending;
552 		bool soft;
553 		u8 nr;
554 	} interrupt;
555 
556 	int halt_request; /* real mode on Intel only */
557 
558 	int cpuid_nent;
559 	struct kvm_cpuid_entry2 cpuid_entries[KVM_MAX_CPUID_ENTRIES];
560 
561 	int maxphyaddr;
562 
563 	/* emulate context */
564 
565 	struct x86_emulate_ctxt emulate_ctxt;
566 	bool emulate_regs_need_sync_to_vcpu;
567 	bool emulate_regs_need_sync_from_vcpu;
568 	int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
569 
570 	gpa_t time;
571 	struct pvclock_vcpu_time_info hv_clock;
572 	unsigned int hw_tsc_khz;
573 	struct gfn_to_hva_cache pv_time;
574 	bool pv_time_enabled;
575 	/* set guest stopped flag in pvclock flags field */
576 	bool pvclock_set_guest_stopped_request;
577 
578 	struct {
579 		u64 msr_val;
580 		u64 last_steal;
581 		struct gfn_to_hva_cache stime;
582 		struct kvm_steal_time steal;
583 	} st;
584 
585 	u64 tsc_offset;
586 	u64 last_guest_tsc;
587 	u64 last_host_tsc;
588 	u64 tsc_offset_adjustment;
589 	u64 this_tsc_nsec;
590 	u64 this_tsc_write;
591 	u64 this_tsc_generation;
592 	bool tsc_catchup;
593 	bool tsc_always_catchup;
594 	s8 virtual_tsc_shift;
595 	u32 virtual_tsc_mult;
596 	u32 virtual_tsc_khz;
597 	s64 ia32_tsc_adjust_msr;
598 	u64 tsc_scaling_ratio;
599 
600 	atomic_t nmi_queued;  /* unprocessed asynchronous NMIs */
601 	unsigned nmi_pending; /* NMI queued after currently running handler */
602 	bool nmi_injected;    /* Trying to inject an NMI this entry */
603 	bool smi_pending;    /* SMI queued after currently running handler */
604 
605 	struct kvm_mtrr mtrr_state;
606 	u64 pat;
607 
608 	unsigned switch_db_regs;
609 	unsigned long db[KVM_NR_DB_REGS];
610 	unsigned long dr6;
611 	unsigned long dr7;
612 	unsigned long eff_db[KVM_NR_DB_REGS];
613 	unsigned long guest_debug_dr7;
614 	u64 msr_platform_info;
615 	u64 msr_misc_features_enables;
616 
617 	u64 mcg_cap;
618 	u64 mcg_status;
619 	u64 mcg_ctl;
620 	u64 mcg_ext_ctl;
621 	u64 *mce_banks;
622 
623 	/* Cache MMIO info */
624 	u64 mmio_gva;
625 	unsigned access;
626 	gfn_t mmio_gfn;
627 	u64 mmio_gen;
628 
629 	struct kvm_pmu pmu;
630 
631 	/* used for guest single stepping over the given code position */
632 	unsigned long singlestep_rip;
633 
634 	struct kvm_vcpu_hv hyperv;
635 
636 	cpumask_var_t wbinvd_dirty_mask;
637 
638 	unsigned long last_retry_eip;
639 	unsigned long last_retry_addr;
640 
641 	struct {
642 		bool halted;
643 		gfn_t gfns[roundup_pow_of_two(ASYNC_PF_PER_VCPU)];
644 		struct gfn_to_hva_cache data;
645 		u64 msr_val;
646 		u32 id;
647 		bool send_user_only;
648 	} apf;
649 
650 	/* OSVW MSRs (AMD only) */
651 	struct {
652 		u64 length;
653 		u64 status;
654 	} osvw;
655 
656 	struct {
657 		u64 msr_val;
658 		struct gfn_to_hva_cache data;
659 	} pv_eoi;
660 
661 	/*
662 	 * Indicate whether the access faults on its page table in guest
663 	 * which is set when fix page fault and used to detect unhandeable
664 	 * instruction.
665 	 */
666 	bool write_fault_to_shadow_pgtable;
667 
668 	/* set at EPT violation at this point */
669 	unsigned long exit_qualification;
670 
671 	/* pv related host specific info */
672 	struct {
673 		bool pv_unhalted;
674 	} pv;
675 
676 	int pending_ioapic_eoi;
677 	int pending_external_vector;
678 
679 	/* GPA available (AMD only) */
680 	bool gpa_available;
681 };
682 
683 struct kvm_lpage_info {
684 	int disallow_lpage;
685 };
686 
687 struct kvm_arch_memory_slot {
688 	struct kvm_rmap_head *rmap[KVM_NR_PAGE_SIZES];
689 	struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
690 	unsigned short *gfn_track[KVM_PAGE_TRACK_MAX];
691 };
692 
693 /*
694  * We use as the mode the number of bits allocated in the LDR for the
695  * logical processor ID.  It happens that these are all powers of two.
696  * This makes it is very easy to detect cases where the APICs are
697  * configured for multiple modes; in that case, we cannot use the map and
698  * hence cannot use kvm_irq_delivery_to_apic_fast either.
699  */
700 #define KVM_APIC_MODE_XAPIC_CLUSTER          4
701 #define KVM_APIC_MODE_XAPIC_FLAT             8
702 #define KVM_APIC_MODE_X2APIC                16
703 
704 struct kvm_apic_map {
705 	struct rcu_head rcu;
706 	u8 mode;
707 	u32 max_apic_id;
708 	union {
709 		struct kvm_lapic *xapic_flat_map[8];
710 		struct kvm_lapic *xapic_cluster_map[16][4];
711 	};
712 	struct kvm_lapic *phys_map[];
713 };
714 
715 /* Hyper-V emulation context */
716 struct kvm_hv {
717 	struct mutex hv_lock;
718 	u64 hv_guest_os_id;
719 	u64 hv_hypercall;
720 	u64 hv_tsc_page;
721 
722 	/* Hyper-v based guest crash (NT kernel bugcheck) parameters */
723 	u64 hv_crash_param[HV_X64_MSR_CRASH_PARAMS];
724 	u64 hv_crash_ctl;
725 
726 	HV_REFERENCE_TSC_PAGE tsc_ref;
727 };
728 
729 enum kvm_irqchip_mode {
730 	KVM_IRQCHIP_NONE,
731 	KVM_IRQCHIP_KERNEL,       /* created with KVM_CREATE_IRQCHIP */
732 	KVM_IRQCHIP_SPLIT,        /* created with KVM_CAP_SPLIT_IRQCHIP */
733 };
734 
735 struct kvm_arch {
736 	unsigned int n_used_mmu_pages;
737 	unsigned int n_requested_mmu_pages;
738 	unsigned int n_max_mmu_pages;
739 	unsigned int indirect_shadow_pages;
740 	unsigned long mmu_valid_gen;
741 	struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
742 	/*
743 	 * Hash table of struct kvm_mmu_page.
744 	 */
745 	struct list_head active_mmu_pages;
746 	struct list_head zapped_obsolete_pages;
747 	struct kvm_page_track_notifier_node mmu_sp_tracker;
748 	struct kvm_page_track_notifier_head track_notifier_head;
749 
750 	struct list_head assigned_dev_head;
751 	struct iommu_domain *iommu_domain;
752 	bool iommu_noncoherent;
753 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA
754 	atomic_t noncoherent_dma_count;
755 #define __KVM_HAVE_ARCH_ASSIGNED_DEVICE
756 	atomic_t assigned_device_count;
757 	struct kvm_pic *vpic;
758 	struct kvm_ioapic *vioapic;
759 	struct kvm_pit *vpit;
760 	atomic_t vapics_in_nmi_mode;
761 	struct mutex apic_map_lock;
762 	struct kvm_apic_map *apic_map;
763 
764 	unsigned int tss_addr;
765 	bool apic_access_page_done;
766 
767 	gpa_t wall_clock;
768 
769 	bool ept_identity_pagetable_done;
770 	gpa_t ept_identity_map_addr;
771 
772 	unsigned long irq_sources_bitmap;
773 	s64 kvmclock_offset;
774 	raw_spinlock_t tsc_write_lock;
775 	u64 last_tsc_nsec;
776 	u64 last_tsc_write;
777 	u32 last_tsc_khz;
778 	u64 cur_tsc_nsec;
779 	u64 cur_tsc_write;
780 	u64 cur_tsc_offset;
781 	u64 cur_tsc_generation;
782 	int nr_vcpus_matched_tsc;
783 
784 	spinlock_t pvclock_gtod_sync_lock;
785 	bool use_master_clock;
786 	u64 master_kernel_ns;
787 	u64 master_cycle_now;
788 	struct delayed_work kvmclock_update_work;
789 	struct delayed_work kvmclock_sync_work;
790 
791 	struct kvm_xen_hvm_config xen_hvm_config;
792 
793 	/* reads protected by irq_srcu, writes by irq_lock */
794 	struct hlist_head mask_notifier_list;
795 
796 	struct kvm_hv hyperv;
797 
798 	#ifdef CONFIG_KVM_MMU_AUDIT
799 	int audit_point;
800 	#endif
801 
802 	bool boot_vcpu_runs_old_kvmclock;
803 	u32 bsp_vcpu_id;
804 
805 	u64 disabled_quirks;
806 
807 	enum kvm_irqchip_mode irqchip_mode;
808 	u8 nr_reserved_ioapic_pins;
809 
810 	bool disabled_lapic_found;
811 
812 	/* Struct members for AVIC */
813 	u32 avic_vm_id;
814 	u32 ldr_mode;
815 	struct page *avic_logical_id_table_page;
816 	struct page *avic_physical_id_table_page;
817 	struct hlist_node hnode;
818 
819 	bool x2apic_format;
820 	bool x2apic_broadcast_quirk_disabled;
821 };
822 
823 struct kvm_vm_stat {
824 	ulong mmu_shadow_zapped;
825 	ulong mmu_pte_write;
826 	ulong mmu_pte_updated;
827 	ulong mmu_pde_zapped;
828 	ulong mmu_flooded;
829 	ulong mmu_recycled;
830 	ulong mmu_cache_miss;
831 	ulong mmu_unsync;
832 	ulong remote_tlb_flush;
833 	ulong lpages;
834 	ulong max_mmu_page_hash_collisions;
835 };
836 
837 struct kvm_vcpu_stat {
838 	u64 pf_fixed;
839 	u64 pf_guest;
840 	u64 tlb_flush;
841 	u64 invlpg;
842 
843 	u64 exits;
844 	u64 io_exits;
845 	u64 mmio_exits;
846 	u64 signal_exits;
847 	u64 irq_window_exits;
848 	u64 nmi_window_exits;
849 	u64 halt_exits;
850 	u64 halt_successful_poll;
851 	u64 halt_attempted_poll;
852 	u64 halt_poll_invalid;
853 	u64 halt_wakeup;
854 	u64 request_irq_exits;
855 	u64 irq_exits;
856 	u64 host_state_reload;
857 	u64 efer_reload;
858 	u64 fpu_reload;
859 	u64 insn_emulation;
860 	u64 insn_emulation_fail;
861 	u64 hypercalls;
862 	u64 irq_injections;
863 	u64 nmi_injections;
864 	u64 req_event;
865 };
866 
867 struct x86_instruction_info;
868 
869 struct msr_data {
870 	bool host_initiated;
871 	u32 index;
872 	u64 data;
873 };
874 
875 struct kvm_lapic_irq {
876 	u32 vector;
877 	u16 delivery_mode;
878 	u16 dest_mode;
879 	bool level;
880 	u16 trig_mode;
881 	u32 shorthand;
882 	u32 dest_id;
883 	bool msi_redir_hint;
884 };
885 
886 struct kvm_x86_ops {
887 	int (*cpu_has_kvm_support)(void);          /* __init */
888 	int (*disabled_by_bios)(void);             /* __init */
889 	int (*hardware_enable)(void);
890 	void (*hardware_disable)(void);
891 	void (*check_processor_compatibility)(void *rtn);
892 	int (*hardware_setup)(void);               /* __init */
893 	void (*hardware_unsetup)(void);            /* __exit */
894 	bool (*cpu_has_accelerated_tpr)(void);
895 	bool (*cpu_has_high_real_mode_segbase)(void);
896 	void (*cpuid_update)(struct kvm_vcpu *vcpu);
897 
898 	int (*vm_init)(struct kvm *kvm);
899 	void (*vm_destroy)(struct kvm *kvm);
900 
901 	/* Create, but do not attach this VCPU */
902 	struct kvm_vcpu *(*vcpu_create)(struct kvm *kvm, unsigned id);
903 	void (*vcpu_free)(struct kvm_vcpu *vcpu);
904 	void (*vcpu_reset)(struct kvm_vcpu *vcpu, bool init_event);
905 
906 	void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
907 	void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
908 	void (*vcpu_put)(struct kvm_vcpu *vcpu);
909 
910 	void (*update_bp_intercept)(struct kvm_vcpu *vcpu);
911 	int (*get_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
912 	int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
913 	u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
914 	void (*get_segment)(struct kvm_vcpu *vcpu,
915 			    struct kvm_segment *var, int seg);
916 	int (*get_cpl)(struct kvm_vcpu *vcpu);
917 	void (*set_segment)(struct kvm_vcpu *vcpu,
918 			    struct kvm_segment *var, int seg);
919 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
920 	void (*decache_cr0_guest_bits)(struct kvm_vcpu *vcpu);
921 	void (*decache_cr3)(struct kvm_vcpu *vcpu);
922 	void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu);
923 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
924 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
925 	int (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
926 	void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
927 	void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
928 	void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
929 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
930 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
931 	u64 (*get_dr6)(struct kvm_vcpu *vcpu);
932 	void (*set_dr6)(struct kvm_vcpu *vcpu, unsigned long value);
933 	void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu);
934 	void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
935 	void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
936 	unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
937 	void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
938 	u32 (*get_pkru)(struct kvm_vcpu *vcpu);
939 
940 	void (*tlb_flush)(struct kvm_vcpu *vcpu);
941 
942 	void (*run)(struct kvm_vcpu *vcpu);
943 	int (*handle_exit)(struct kvm_vcpu *vcpu);
944 	void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
945 	void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
946 	u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu);
947 	void (*patch_hypercall)(struct kvm_vcpu *vcpu,
948 				unsigned char *hypercall_addr);
949 	void (*set_irq)(struct kvm_vcpu *vcpu);
950 	void (*set_nmi)(struct kvm_vcpu *vcpu);
951 	void (*queue_exception)(struct kvm_vcpu *vcpu, unsigned nr,
952 				bool has_error_code, u32 error_code,
953 				bool reinject);
954 	void (*cancel_injection)(struct kvm_vcpu *vcpu);
955 	int (*interrupt_allowed)(struct kvm_vcpu *vcpu);
956 	int (*nmi_allowed)(struct kvm_vcpu *vcpu);
957 	bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
958 	void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
959 	void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
960 	void (*enable_irq_window)(struct kvm_vcpu *vcpu);
961 	void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
962 	bool (*get_enable_apicv)(void);
963 	void (*refresh_apicv_exec_ctrl)(struct kvm_vcpu *vcpu);
964 	void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
965 	void (*hwapic_isr_update)(struct kvm_vcpu *vcpu, int isr);
966 	void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
967 	void (*set_virtual_x2apic_mode)(struct kvm_vcpu *vcpu, bool set);
968 	void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu, hpa_t hpa);
969 	void (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
970 	int (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
971 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
972 	int (*get_tdp_level)(void);
973 	u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
974 	int (*get_lpage_level)(void);
975 	bool (*rdtscp_supported)(void);
976 	bool (*invpcid_supported)(void);
977 
978 	void (*set_tdp_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
979 
980 	void (*set_supported_cpuid)(u32 func, struct kvm_cpuid_entry2 *entry);
981 
982 	bool (*has_wbinvd_exit)(void);
983 
984 	void (*write_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
985 
986 	void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2);
987 
988 	int (*check_intercept)(struct kvm_vcpu *vcpu,
989 			       struct x86_instruction_info *info,
990 			       enum x86_intercept_stage stage);
991 	void (*handle_external_intr)(struct kvm_vcpu *vcpu);
992 	bool (*mpx_supported)(void);
993 	bool (*xsaves_supported)(void);
994 
995 	int (*check_nested_events)(struct kvm_vcpu *vcpu, bool external_intr);
996 
997 	void (*sched_in)(struct kvm_vcpu *kvm, int cpu);
998 
999 	/*
1000 	 * Arch-specific dirty logging hooks. These hooks are only supposed to
1001 	 * be valid if the specific arch has hardware-accelerated dirty logging
1002 	 * mechanism. Currently only for PML on VMX.
1003 	 *
1004 	 *  - slot_enable_log_dirty:
1005 	 *	called when enabling log dirty mode for the slot.
1006 	 *  - slot_disable_log_dirty:
1007 	 *	called when disabling log dirty mode for the slot.
1008 	 *	also called when slot is created with log dirty disabled.
1009 	 *  - flush_log_dirty:
1010 	 *	called before reporting dirty_bitmap to userspace.
1011 	 *  - enable_log_dirty_pt_masked:
1012 	 *	called when reenabling log dirty for the GFNs in the mask after
1013 	 *	corresponding bits are cleared in slot->dirty_bitmap.
1014 	 */
1015 	void (*slot_enable_log_dirty)(struct kvm *kvm,
1016 				      struct kvm_memory_slot *slot);
1017 	void (*slot_disable_log_dirty)(struct kvm *kvm,
1018 				       struct kvm_memory_slot *slot);
1019 	void (*flush_log_dirty)(struct kvm *kvm);
1020 	void (*enable_log_dirty_pt_masked)(struct kvm *kvm,
1021 					   struct kvm_memory_slot *slot,
1022 					   gfn_t offset, unsigned long mask);
1023 	int (*write_log_dirty)(struct kvm_vcpu *vcpu);
1024 
1025 	/* pmu operations of sub-arch */
1026 	const struct kvm_pmu_ops *pmu_ops;
1027 
1028 	/*
1029 	 * Architecture specific hooks for vCPU blocking due to
1030 	 * HLT instruction.
1031 	 * Returns for .pre_block():
1032 	 *    - 0 means continue to block the vCPU.
1033 	 *    - 1 means we cannot block the vCPU since some event
1034 	 *        happens during this period, such as, 'ON' bit in
1035 	 *        posted-interrupts descriptor is set.
1036 	 */
1037 	int (*pre_block)(struct kvm_vcpu *vcpu);
1038 	void (*post_block)(struct kvm_vcpu *vcpu);
1039 
1040 	void (*vcpu_blocking)(struct kvm_vcpu *vcpu);
1041 	void (*vcpu_unblocking)(struct kvm_vcpu *vcpu);
1042 
1043 	int (*update_pi_irte)(struct kvm *kvm, unsigned int host_irq,
1044 			      uint32_t guest_irq, bool set);
1045 	void (*apicv_post_state_restore)(struct kvm_vcpu *vcpu);
1046 
1047 	int (*set_hv_timer)(struct kvm_vcpu *vcpu, u64 guest_deadline_tsc);
1048 	void (*cancel_hv_timer)(struct kvm_vcpu *vcpu);
1049 
1050 	void (*setup_mce)(struct kvm_vcpu *vcpu);
1051 };
1052 
1053 struct kvm_arch_async_pf {
1054 	u32 token;
1055 	gfn_t gfn;
1056 	unsigned long cr3;
1057 	bool direct_map;
1058 };
1059 
1060 extern struct kvm_x86_ops *kvm_x86_ops;
1061 
1062 int kvm_mmu_module_init(void);
1063 void kvm_mmu_module_exit(void);
1064 
1065 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
1066 int kvm_mmu_create(struct kvm_vcpu *vcpu);
1067 void kvm_mmu_setup(struct kvm_vcpu *vcpu);
1068 void kvm_mmu_init_vm(struct kvm *kvm);
1069 void kvm_mmu_uninit_vm(struct kvm *kvm);
1070 void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
1071 		u64 dirty_mask, u64 nx_mask, u64 x_mask, u64 p_mask,
1072 		u64 acc_track_mask);
1073 
1074 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
1075 void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
1076 				      struct kvm_memory_slot *memslot);
1077 void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
1078 				   const struct kvm_memory_slot *memslot);
1079 void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
1080 				   struct kvm_memory_slot *memslot);
1081 void kvm_mmu_slot_largepage_remove_write_access(struct kvm *kvm,
1082 					struct kvm_memory_slot *memslot);
1083 void kvm_mmu_slot_set_dirty(struct kvm *kvm,
1084 			    struct kvm_memory_slot *memslot);
1085 void kvm_mmu_clear_dirty_pt_masked(struct kvm *kvm,
1086 				   struct kvm_memory_slot *slot,
1087 				   gfn_t gfn_offset, unsigned long mask);
1088 void kvm_mmu_zap_all(struct kvm *kvm);
1089 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, struct kvm_memslots *slots);
1090 unsigned int kvm_mmu_calculate_mmu_pages(struct kvm *kvm);
1091 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int kvm_nr_mmu_pages);
1092 
1093 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
1094 bool pdptrs_changed(struct kvm_vcpu *vcpu);
1095 
1096 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
1097 			  const void *val, int bytes);
1098 
1099 struct kvm_irq_mask_notifier {
1100 	void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
1101 	int irq;
1102 	struct hlist_node link;
1103 };
1104 
1105 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
1106 				    struct kvm_irq_mask_notifier *kimn);
1107 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
1108 				      struct kvm_irq_mask_notifier *kimn);
1109 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
1110 			     bool mask);
1111 
1112 extern bool tdp_enabled;
1113 
1114 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
1115 
1116 /* control of guest tsc rate supported? */
1117 extern bool kvm_has_tsc_control;
1118 /* maximum supported tsc_khz for guests */
1119 extern u32  kvm_max_guest_tsc_khz;
1120 /* number of bits of the fractional part of the TSC scaling ratio */
1121 extern u8   kvm_tsc_scaling_ratio_frac_bits;
1122 /* maximum allowed value of TSC scaling ratio */
1123 extern u64  kvm_max_tsc_scaling_ratio;
1124 /* 1ull << kvm_tsc_scaling_ratio_frac_bits */
1125 extern u64  kvm_default_tsc_scaling_ratio;
1126 
1127 extern u64 kvm_mce_cap_supported;
1128 
1129 enum emulation_result {
1130 	EMULATE_DONE,         /* no further processing */
1131 	EMULATE_USER_EXIT,    /* kvm_run ready for userspace exit */
1132 	EMULATE_FAIL,         /* can't emulate this instruction */
1133 };
1134 
1135 #define EMULTYPE_NO_DECODE	    (1 << 0)
1136 #define EMULTYPE_TRAP_UD	    (1 << 1)
1137 #define EMULTYPE_SKIP		    (1 << 2)
1138 #define EMULTYPE_RETRY		    (1 << 3)
1139 #define EMULTYPE_NO_REEXECUTE	    (1 << 4)
1140 int x86_emulate_instruction(struct kvm_vcpu *vcpu, unsigned long cr2,
1141 			    int emulation_type, void *insn, int insn_len);
1142 
1143 static inline int emulate_instruction(struct kvm_vcpu *vcpu,
1144 			int emulation_type)
1145 {
1146 	return x86_emulate_instruction(vcpu, 0, emulation_type, NULL, 0);
1147 }
1148 
1149 void kvm_enable_efer_bits(u64);
1150 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
1151 int kvm_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr);
1152 int kvm_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr);
1153 
1154 struct x86_emulate_ctxt;
1155 
1156 int kvm_fast_pio_out(struct kvm_vcpu *vcpu, int size, unsigned short port);
1157 int kvm_fast_pio_in(struct kvm_vcpu *vcpu, int size, unsigned short port);
1158 int kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
1159 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
1160 int kvm_vcpu_halt(struct kvm_vcpu *vcpu);
1161 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
1162 
1163 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
1164 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
1165 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector);
1166 
1167 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
1168 		    int reason, bool has_error_code, u32 error_code);
1169 
1170 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
1171 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
1172 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
1173 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
1174 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
1175 int kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
1176 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
1177 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
1178 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
1179 int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr);
1180 
1181 int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1182 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1183 
1184 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
1185 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
1186 bool kvm_rdpmc(struct kvm_vcpu *vcpu);
1187 
1188 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1189 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1190 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1191 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1192 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
1193 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1194 			    gfn_t gfn, void *data, int offset, int len,
1195 			    u32 access);
1196 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
1197 bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr);
1198 
1199 static inline int __kvm_irq_line_state(unsigned long *irq_state,
1200 				       int irq_source_id, int level)
1201 {
1202 	/* Logical OR for level trig interrupt */
1203 	if (level)
1204 		__set_bit(irq_source_id, irq_state);
1205 	else
1206 		__clear_bit(irq_source_id, irq_state);
1207 
1208 	return !!(*irq_state);
1209 }
1210 
1211 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
1212 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
1213 
1214 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
1215 
1216 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
1217 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
1218 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
1219 int kvm_mmu_load(struct kvm_vcpu *vcpu);
1220 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
1221 void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu);
1222 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1223 			   struct x86_exception *exception);
1224 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
1225 			      struct x86_exception *exception);
1226 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
1227 			       struct x86_exception *exception);
1228 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
1229 			       struct x86_exception *exception);
1230 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
1231 				struct x86_exception *exception);
1232 
1233 void kvm_vcpu_deactivate_apicv(struct kvm_vcpu *vcpu);
1234 
1235 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
1236 
1237 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva, u64 error_code,
1238 		       void *insn, int insn_len);
1239 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
1240 void kvm_mmu_new_cr3(struct kvm_vcpu *vcpu);
1241 
1242 void kvm_enable_tdp(void);
1243 void kvm_disable_tdp(void);
1244 
1245 static inline gpa_t translate_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1246 				  struct x86_exception *exception)
1247 {
1248 	return gpa;
1249 }
1250 
1251 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
1252 {
1253 	struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
1254 
1255 	return (struct kvm_mmu_page *)page_private(page);
1256 }
1257 
1258 static inline u16 kvm_read_ldt(void)
1259 {
1260 	u16 ldt;
1261 	asm("sldt %0" : "=g"(ldt));
1262 	return ldt;
1263 }
1264 
1265 static inline void kvm_load_ldt(u16 sel)
1266 {
1267 	asm("lldt %0" : : "rm"(sel));
1268 }
1269 
1270 #ifdef CONFIG_X86_64
1271 static inline unsigned long read_msr(unsigned long msr)
1272 {
1273 	u64 value;
1274 
1275 	rdmsrl(msr, value);
1276 	return value;
1277 }
1278 #endif
1279 
1280 static inline u32 get_rdx_init_val(void)
1281 {
1282 	return 0x600; /* P6 family */
1283 }
1284 
1285 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
1286 {
1287 	kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
1288 }
1289 
1290 static inline u64 get_canonical(u64 la)
1291 {
1292 	return ((int64_t)la << 16) >> 16;
1293 }
1294 
1295 static inline bool is_noncanonical_address(u64 la)
1296 {
1297 #ifdef CONFIG_X86_64
1298 	return get_canonical(la) != la;
1299 #else
1300 	return false;
1301 #endif
1302 }
1303 
1304 #define TSS_IOPB_BASE_OFFSET 0x66
1305 #define TSS_BASE_SIZE 0x68
1306 #define TSS_IOPB_SIZE (65536 / 8)
1307 #define TSS_REDIRECTION_SIZE (256 / 8)
1308 #define RMODE_TSS_SIZE							\
1309 	(TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
1310 
1311 enum {
1312 	TASK_SWITCH_CALL = 0,
1313 	TASK_SWITCH_IRET = 1,
1314 	TASK_SWITCH_JMP = 2,
1315 	TASK_SWITCH_GATE = 3,
1316 };
1317 
1318 #define HF_GIF_MASK		(1 << 0)
1319 #define HF_HIF_MASK		(1 << 1)
1320 #define HF_VINTR_MASK		(1 << 2)
1321 #define HF_NMI_MASK		(1 << 3)
1322 #define HF_IRET_MASK		(1 << 4)
1323 #define HF_GUEST_MASK		(1 << 5) /* VCPU is in guest-mode */
1324 #define HF_SMM_MASK		(1 << 6)
1325 #define HF_SMM_INSIDE_NMI_MASK	(1 << 7)
1326 
1327 #define __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
1328 #define KVM_ADDRESS_SPACE_NUM 2
1329 
1330 #define kvm_arch_vcpu_memslots_id(vcpu) ((vcpu)->arch.hflags & HF_SMM_MASK ? 1 : 0)
1331 #define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, (role).smm)
1332 
1333 /*
1334  * Hardware virtualization extension instructions may fault if a
1335  * reboot turns off virtualization while processes are running.
1336  * Trap the fault and ignore the instruction if that happens.
1337  */
1338 asmlinkage void kvm_spurious_fault(void);
1339 
1340 #define ____kvm_handle_fault_on_reboot(insn, cleanup_insn)	\
1341 	"666: " insn "\n\t" \
1342 	"668: \n\t"                           \
1343 	".pushsection .fixup, \"ax\" \n" \
1344 	"667: \n\t" \
1345 	cleanup_insn "\n\t"		      \
1346 	"cmpb $0, kvm_rebooting \n\t"	      \
1347 	"jne 668b \n\t"      		      \
1348 	__ASM_SIZE(push) " $666b \n\t"	      \
1349 	"call kvm_spurious_fault \n\t"	      \
1350 	".popsection \n\t" \
1351 	_ASM_EXTABLE(666b, 667b)
1352 
1353 #define __kvm_handle_fault_on_reboot(insn)		\
1354 	____kvm_handle_fault_on_reboot(insn, "")
1355 
1356 #define KVM_ARCH_WANT_MMU_NOTIFIER
1357 int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
1358 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end);
1359 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
1360 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
1361 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
1362 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1363 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1364 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1365 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1366 void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event);
1367 void kvm_vcpu_reload_apic_access_page(struct kvm_vcpu *vcpu);
1368 void kvm_arch_mmu_notifier_invalidate_page(struct kvm *kvm,
1369 					   unsigned long address);
1370 
1371 void kvm_define_shared_msr(unsigned index, u32 msr);
1372 int kvm_set_shared_msr(unsigned index, u64 val, u64 mask);
1373 
1374 u64 kvm_scale_tsc(struct kvm_vcpu *vcpu, u64 tsc);
1375 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc);
1376 
1377 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu);
1378 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1379 
1380 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
1381 void kvm_make_scan_ioapic_request(struct kvm *kvm);
1382 
1383 void kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1384 				     struct kvm_async_pf *work);
1385 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1386 				 struct kvm_async_pf *work);
1387 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1388 			       struct kvm_async_pf *work);
1389 bool kvm_arch_can_inject_async_page_present(struct kvm_vcpu *vcpu);
1390 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1391 
1392 int kvm_skip_emulated_instruction(struct kvm_vcpu *vcpu);
1393 int kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1394 
1395 int kvm_is_in_guest(void);
1396 
1397 int __x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa, u32 size);
1398 int x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa, u32 size);
1399 bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu);
1400 bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu);
1401 
1402 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
1403 			     struct kvm_vcpu **dest_vcpu);
1404 
1405 void kvm_set_msi_irq(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e,
1406 		     struct kvm_lapic_irq *irq);
1407 
1408 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu)
1409 {
1410 	if (kvm_x86_ops->vcpu_blocking)
1411 		kvm_x86_ops->vcpu_blocking(vcpu);
1412 }
1413 
1414 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu)
1415 {
1416 	if (kvm_x86_ops->vcpu_unblocking)
1417 		kvm_x86_ops->vcpu_unblocking(vcpu);
1418 }
1419 
1420 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
1421 
1422 static inline int kvm_cpu_get_apicid(int mps_cpu)
1423 {
1424 #ifdef CONFIG_X86_LOCAL_APIC
1425 	return __default_cpu_present_to_apicid(mps_cpu);
1426 #else
1427 	WARN_ON_ONCE(1);
1428 	return BAD_APICID;
1429 #endif
1430 }
1431 
1432 #endif /* _ASM_X86_KVM_HOST_H */
1433