xref: /openbmc/linux/arch/x86/include/asm/kvm_host.h (revision 781095f903f398148cd0b646d3984234a715f29e)
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/pvclock-abi.h>
31 #include <asm/desc.h>
32 #include <asm/mtrr.h>
33 #include <asm/msr-index.h>
34 #include <asm/asm.h>
35 
36 #define KVM_MAX_VCPUS 255
37 #define KVM_SOFT_MAX_VCPUS 160
38 #define KVM_USER_MEM_SLOTS 509
39 /* memory slots that are not exposed to userspace */
40 #define KVM_PRIVATE_MEM_SLOTS 3
41 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
42 
43 #define KVM_PIO_PAGE_OFFSET 1
44 #define KVM_COALESCED_MMIO_PAGE_OFFSET 2
45 #define KVM_HALT_POLL_NS_DEFAULT 500000
46 
47 #define KVM_IRQCHIP_NUM_PINS  KVM_IOAPIC_NUM_PINS
48 
49 /* x86-specific vcpu->requests bit members */
50 #define KVM_REQ_MIGRATE_TIMER      8
51 #define KVM_REQ_REPORT_TPR_ACCESS  9
52 #define KVM_REQ_TRIPLE_FAULT      10
53 #define KVM_REQ_MMU_SYNC          11
54 #define KVM_REQ_CLOCK_UPDATE      12
55 #define KVM_REQ_DEACTIVATE_FPU    13
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
65 #define KVM_REQ_SCAN_IOAPIC       23
66 #define KVM_REQ_GLOBAL_CLOCK_UPDATE 24
67 #define KVM_REQ_APIC_PAGE_RELOAD  25
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 
88 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR)
89 
90 
91 
92 #define INVALID_PAGE (~(hpa_t)0)
93 #define VALID_PAGE(x) ((x) != INVALID_PAGE)
94 
95 #define UNMAPPED_GVA (~(gpa_t)0)
96 
97 /* KVM Hugepage definitions for x86 */
98 #define KVM_NR_PAGE_SIZES	3
99 #define KVM_HPAGE_GFN_SHIFT(x)	(((x) - 1) * 9)
100 #define KVM_HPAGE_SHIFT(x)	(PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
101 #define KVM_HPAGE_SIZE(x)	(1UL << KVM_HPAGE_SHIFT(x))
102 #define KVM_HPAGE_MASK(x)	(~(KVM_HPAGE_SIZE(x) - 1))
103 #define KVM_PAGES_PER_HPAGE(x)	(KVM_HPAGE_SIZE(x) / PAGE_SIZE)
104 
105 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
106 {
107 	/* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */
108 	return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
109 		(base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
110 }
111 
112 #define KVM_PERMILLE_MMU_PAGES 20
113 #define KVM_MIN_ALLOC_MMU_PAGES 64
114 #define KVM_MMU_HASH_SHIFT 10
115 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
116 #define KVM_MIN_FREE_MMU_PAGES 5
117 #define KVM_REFILL_PAGES 25
118 #define KVM_MAX_CPUID_ENTRIES 80
119 #define KVM_NR_FIXED_MTRR_REGION 88
120 #define KVM_NR_VAR_MTRR 8
121 
122 #define ASYNC_PF_PER_VCPU 64
123 
124 enum kvm_reg {
125 	VCPU_REGS_RAX = 0,
126 	VCPU_REGS_RCX = 1,
127 	VCPU_REGS_RDX = 2,
128 	VCPU_REGS_RBX = 3,
129 	VCPU_REGS_RSP = 4,
130 	VCPU_REGS_RBP = 5,
131 	VCPU_REGS_RSI = 6,
132 	VCPU_REGS_RDI = 7,
133 #ifdef CONFIG_X86_64
134 	VCPU_REGS_R8 = 8,
135 	VCPU_REGS_R9 = 9,
136 	VCPU_REGS_R10 = 10,
137 	VCPU_REGS_R11 = 11,
138 	VCPU_REGS_R12 = 12,
139 	VCPU_REGS_R13 = 13,
140 	VCPU_REGS_R14 = 14,
141 	VCPU_REGS_R15 = 15,
142 #endif
143 	VCPU_REGS_RIP,
144 	NR_VCPU_REGS
145 };
146 
147 enum kvm_reg_ex {
148 	VCPU_EXREG_PDPTR = NR_VCPU_REGS,
149 	VCPU_EXREG_CR3,
150 	VCPU_EXREG_RFLAGS,
151 	VCPU_EXREG_SEGMENTS,
152 };
153 
154 enum {
155 	VCPU_SREG_ES,
156 	VCPU_SREG_CS,
157 	VCPU_SREG_SS,
158 	VCPU_SREG_DS,
159 	VCPU_SREG_FS,
160 	VCPU_SREG_GS,
161 	VCPU_SREG_TR,
162 	VCPU_SREG_LDTR,
163 };
164 
165 #include <asm/kvm_emulate.h>
166 
167 #define KVM_NR_MEM_OBJS 40
168 
169 #define KVM_NR_DB_REGS	4
170 
171 #define DR6_BD		(1 << 13)
172 #define DR6_BS		(1 << 14)
173 #define DR6_RTM		(1 << 16)
174 #define DR6_FIXED_1	0xfffe0ff0
175 #define DR6_INIT	0xffff0ff0
176 #define DR6_VOLATILE	0x0001e00f
177 
178 #define DR7_BP_EN_MASK	0x000000ff
179 #define DR7_GE		(1 << 9)
180 #define DR7_GD		(1 << 13)
181 #define DR7_FIXED_1	0x00000400
182 #define DR7_VOLATILE	0xffff2bff
183 
184 #define PFERR_PRESENT_BIT 0
185 #define PFERR_WRITE_BIT 1
186 #define PFERR_USER_BIT 2
187 #define PFERR_RSVD_BIT 3
188 #define PFERR_FETCH_BIT 4
189 
190 #define PFERR_PRESENT_MASK (1U << PFERR_PRESENT_BIT)
191 #define PFERR_WRITE_MASK (1U << PFERR_WRITE_BIT)
192 #define PFERR_USER_MASK (1U << PFERR_USER_BIT)
193 #define PFERR_RSVD_MASK (1U << PFERR_RSVD_BIT)
194 #define PFERR_FETCH_MASK (1U << PFERR_FETCH_BIT)
195 
196 /* apic attention bits */
197 #define KVM_APIC_CHECK_VAPIC	0
198 /*
199  * The following bit is set with PV-EOI, unset on EOI.
200  * We detect PV-EOI changes by guest by comparing
201  * this bit with PV-EOI in guest memory.
202  * See the implementation in apic_update_pv_eoi.
203  */
204 #define KVM_APIC_PV_EOI_PENDING	1
205 
206 struct kvm_kernel_irq_routing_entry;
207 
208 /*
209  * We don't want allocation failures within the mmu code, so we preallocate
210  * enough memory for a single page fault in a cache.
211  */
212 struct kvm_mmu_memory_cache {
213 	int nobjs;
214 	void *objects[KVM_NR_MEM_OBJS];
215 };
216 
217 union kvm_mmu_page_role {
218 	unsigned word;
219 	struct {
220 		unsigned level:4;
221 		unsigned cr4_pae:1;
222 		unsigned quadrant:2;
223 		unsigned direct:1;
224 		unsigned access:3;
225 		unsigned invalid:1;
226 		unsigned nxe:1;
227 		unsigned cr0_wp:1;
228 		unsigned smep_andnot_wp:1;
229 		unsigned smap_andnot_wp:1;
230 		unsigned :8;
231 
232 		/*
233 		 * This is left at the top of the word so that
234 		 * kvm_memslots_for_spte_role can extract it with a
235 		 * simple shift.  While there is room, give it a whole
236 		 * byte so it is also faster to load it from memory.
237 		 */
238 		unsigned smm:8;
239 	};
240 };
241 
242 struct kvm_rmap_head {
243 	unsigned long val;
244 };
245 
246 struct kvm_mmu_page {
247 	struct list_head link;
248 	struct hlist_node hash_link;
249 
250 	/*
251 	 * The following two entries are used to key the shadow page in the
252 	 * hash table.
253 	 */
254 	gfn_t gfn;
255 	union kvm_mmu_page_role role;
256 
257 	u64 *spt;
258 	/* hold the gfn of each spte inside spt */
259 	gfn_t *gfns;
260 	bool unsync;
261 	int root_count;          /* Currently serving as active root */
262 	unsigned int unsync_children;
263 	struct kvm_rmap_head parent_ptes; /* rmap pointers to parent sptes */
264 
265 	/* The page is obsolete if mmu_valid_gen != kvm->arch.mmu_valid_gen.  */
266 	unsigned long mmu_valid_gen;
267 
268 	DECLARE_BITMAP(unsync_child_bitmap, 512);
269 
270 #ifdef CONFIG_X86_32
271 	/*
272 	 * Used out of the mmu-lock to avoid reading spte values while an
273 	 * update is in progress; see the comments in __get_spte_lockless().
274 	 */
275 	int clear_spte_count;
276 #endif
277 
278 	/* Number of writes since the last time traversal visited this page.  */
279 	int write_flooding_count;
280 };
281 
282 struct kvm_pio_request {
283 	unsigned long count;
284 	int in;
285 	int port;
286 	int size;
287 };
288 
289 struct rsvd_bits_validate {
290 	u64 rsvd_bits_mask[2][4];
291 	u64 bad_mt_xwr;
292 };
293 
294 /*
295  * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
296  * 32-bit).  The kvm_mmu structure abstracts the details of the current mmu
297  * mode.
298  */
299 struct kvm_mmu {
300 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long root);
301 	unsigned long (*get_cr3)(struct kvm_vcpu *vcpu);
302 	u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
303 	int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err,
304 			  bool prefault);
305 	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
306 				  struct x86_exception *fault);
307 	gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva, u32 access,
308 			    struct x86_exception *exception);
309 	gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
310 			       struct x86_exception *exception);
311 	int (*sync_page)(struct kvm_vcpu *vcpu,
312 			 struct kvm_mmu_page *sp);
313 	void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva);
314 	void (*update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
315 			   u64 *spte, const void *pte);
316 	hpa_t root_hpa;
317 	int root_level;
318 	int shadow_root_level;
319 	union kvm_mmu_page_role base_role;
320 	bool direct_map;
321 
322 	/*
323 	 * Bitmap; bit set = permission fault
324 	 * Byte index: page fault error code [4:1]
325 	 * Bit index: pte permissions in ACC_* format
326 	 */
327 	u8 permissions[16];
328 
329 	u64 *pae_root;
330 	u64 *lm_root;
331 
332 	/*
333 	 * check zero bits on shadow page table entries, these
334 	 * bits include not only hardware reserved bits but also
335 	 * the bits spte never used.
336 	 */
337 	struct rsvd_bits_validate shadow_zero_check;
338 
339 	struct rsvd_bits_validate guest_rsvd_check;
340 
341 	/*
342 	 * Bitmap: bit set = last pte in walk
343 	 * index[0:1]: level (zero-based)
344 	 * index[2]: pte.ps
345 	 */
346 	u8 last_pte_bitmap;
347 
348 	bool nx;
349 
350 	u64 pdptrs[4]; /* pae */
351 };
352 
353 enum pmc_type {
354 	KVM_PMC_GP = 0,
355 	KVM_PMC_FIXED,
356 };
357 
358 struct kvm_pmc {
359 	enum pmc_type type;
360 	u8 idx;
361 	u64 counter;
362 	u64 eventsel;
363 	struct perf_event *perf_event;
364 	struct kvm_vcpu *vcpu;
365 };
366 
367 struct kvm_pmu {
368 	unsigned nr_arch_gp_counters;
369 	unsigned nr_arch_fixed_counters;
370 	unsigned available_event_types;
371 	u64 fixed_ctr_ctrl;
372 	u64 global_ctrl;
373 	u64 global_status;
374 	u64 global_ovf_ctrl;
375 	u64 counter_bitmask[2];
376 	u64 global_ctrl_mask;
377 	u64 reserved_bits;
378 	u8 version;
379 	struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
380 	struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED];
381 	struct irq_work irq_work;
382 	u64 reprogram_pmi;
383 };
384 
385 struct kvm_pmu_ops;
386 
387 enum {
388 	KVM_DEBUGREG_BP_ENABLED = 1,
389 	KVM_DEBUGREG_WONT_EXIT = 2,
390 	KVM_DEBUGREG_RELOAD = 4,
391 };
392 
393 struct kvm_mtrr_range {
394 	u64 base;
395 	u64 mask;
396 	struct list_head node;
397 };
398 
399 struct kvm_mtrr {
400 	struct kvm_mtrr_range var_ranges[KVM_NR_VAR_MTRR];
401 	mtrr_type fixed_ranges[KVM_NR_FIXED_MTRR_REGION];
402 	u64 deftype;
403 
404 	struct list_head head;
405 };
406 
407 /* Hyper-V SynIC timer */
408 struct kvm_vcpu_hv_stimer {
409 	struct hrtimer timer;
410 	int index;
411 	u64 config;
412 	u64 count;
413 	u64 exp_time;
414 	struct hv_message msg;
415 	bool msg_pending;
416 };
417 
418 /* Hyper-V synthetic interrupt controller (SynIC)*/
419 struct kvm_vcpu_hv_synic {
420 	u64 version;
421 	u64 control;
422 	u64 msg_page;
423 	u64 evt_page;
424 	atomic64_t sint[HV_SYNIC_SINT_COUNT];
425 	atomic_t sint_to_gsi[HV_SYNIC_SINT_COUNT];
426 	DECLARE_BITMAP(auto_eoi_bitmap, 256);
427 	DECLARE_BITMAP(vec_bitmap, 256);
428 	bool active;
429 };
430 
431 /* Hyper-V per vcpu emulation context */
432 struct kvm_vcpu_hv {
433 	u64 hv_vapic;
434 	s64 runtime_offset;
435 	struct kvm_vcpu_hv_synic synic;
436 	struct kvm_hyperv_exit exit;
437 	struct kvm_vcpu_hv_stimer stimer[HV_SYNIC_STIMER_COUNT];
438 	DECLARE_BITMAP(stimer_pending_bitmap, HV_SYNIC_STIMER_COUNT);
439 };
440 
441 struct kvm_vcpu_arch {
442 	/*
443 	 * rip and regs accesses must go through
444 	 * kvm_{register,rip}_{read,write} functions.
445 	 */
446 	unsigned long regs[NR_VCPU_REGS];
447 	u32 regs_avail;
448 	u32 regs_dirty;
449 
450 	unsigned long cr0;
451 	unsigned long cr0_guest_owned_bits;
452 	unsigned long cr2;
453 	unsigned long cr3;
454 	unsigned long cr4;
455 	unsigned long cr4_guest_owned_bits;
456 	unsigned long cr8;
457 	u32 hflags;
458 	u64 efer;
459 	u64 apic_base;
460 	struct kvm_lapic *apic;    /* kernel irqchip context */
461 	bool apicv_active;
462 	DECLARE_BITMAP(ioapic_handled_vectors, 256);
463 	unsigned long apic_attention;
464 	int32_t apic_arb_prio;
465 	int mp_state;
466 	u64 ia32_misc_enable_msr;
467 	u64 smbase;
468 	bool tpr_access_reporting;
469 	u64 ia32_xss;
470 
471 	/*
472 	 * Paging state of the vcpu
473 	 *
474 	 * If the vcpu runs in guest mode with two level paging this still saves
475 	 * the paging mode of the l1 guest. This context is always used to
476 	 * handle faults.
477 	 */
478 	struct kvm_mmu mmu;
479 
480 	/*
481 	 * Paging state of an L2 guest (used for nested npt)
482 	 *
483 	 * This context will save all necessary information to walk page tables
484 	 * of the an L2 guest. This context is only initialized for page table
485 	 * walking and not for faulting since we never handle l2 page faults on
486 	 * the host.
487 	 */
488 	struct kvm_mmu nested_mmu;
489 
490 	/*
491 	 * Pointer to the mmu context currently used for
492 	 * gva_to_gpa translations.
493 	 */
494 	struct kvm_mmu *walk_mmu;
495 
496 	struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
497 	struct kvm_mmu_memory_cache mmu_page_cache;
498 	struct kvm_mmu_memory_cache mmu_page_header_cache;
499 
500 	struct fpu guest_fpu;
501 	bool eager_fpu;
502 	u64 xcr0;
503 	u64 guest_supported_xcr0;
504 	u32 guest_xstate_size;
505 
506 	struct kvm_pio_request pio;
507 	void *pio_data;
508 
509 	u8 event_exit_inst_len;
510 
511 	struct kvm_queued_exception {
512 		bool pending;
513 		bool has_error_code;
514 		bool reinject;
515 		u8 nr;
516 		u32 error_code;
517 	} exception;
518 
519 	struct kvm_queued_interrupt {
520 		bool pending;
521 		bool soft;
522 		u8 nr;
523 	} interrupt;
524 
525 	int halt_request; /* real mode on Intel only */
526 
527 	int cpuid_nent;
528 	struct kvm_cpuid_entry2 cpuid_entries[KVM_MAX_CPUID_ENTRIES];
529 
530 	int maxphyaddr;
531 
532 	/* emulate context */
533 
534 	struct x86_emulate_ctxt emulate_ctxt;
535 	bool emulate_regs_need_sync_to_vcpu;
536 	bool emulate_regs_need_sync_from_vcpu;
537 	int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
538 
539 	gpa_t time;
540 	struct pvclock_vcpu_time_info hv_clock;
541 	unsigned int hw_tsc_khz;
542 	struct gfn_to_hva_cache pv_time;
543 	bool pv_time_enabled;
544 	/* set guest stopped flag in pvclock flags field */
545 	bool pvclock_set_guest_stopped_request;
546 
547 	struct {
548 		u64 msr_val;
549 		u64 last_steal;
550 		u64 accum_steal;
551 		struct gfn_to_hva_cache stime;
552 		struct kvm_steal_time steal;
553 	} st;
554 
555 	u64 last_guest_tsc;
556 	u64 last_host_tsc;
557 	u64 tsc_offset_adjustment;
558 	u64 this_tsc_nsec;
559 	u64 this_tsc_write;
560 	u64 this_tsc_generation;
561 	bool tsc_catchup;
562 	bool tsc_always_catchup;
563 	s8 virtual_tsc_shift;
564 	u32 virtual_tsc_mult;
565 	u32 virtual_tsc_khz;
566 	s64 ia32_tsc_adjust_msr;
567 	u64 tsc_scaling_ratio;
568 
569 	atomic_t nmi_queued;  /* unprocessed asynchronous NMIs */
570 	unsigned nmi_pending; /* NMI queued after currently running handler */
571 	bool nmi_injected;    /* Trying to inject an NMI this entry */
572 	bool smi_pending;    /* SMI queued after currently running handler */
573 
574 	struct kvm_mtrr mtrr_state;
575 	u64 pat;
576 
577 	unsigned switch_db_regs;
578 	unsigned long db[KVM_NR_DB_REGS];
579 	unsigned long dr6;
580 	unsigned long dr7;
581 	unsigned long eff_db[KVM_NR_DB_REGS];
582 	unsigned long guest_debug_dr7;
583 
584 	u64 mcg_cap;
585 	u64 mcg_status;
586 	u64 mcg_ctl;
587 	u64 *mce_banks;
588 
589 	/* Cache MMIO info */
590 	u64 mmio_gva;
591 	unsigned access;
592 	gfn_t mmio_gfn;
593 	u64 mmio_gen;
594 
595 	struct kvm_pmu pmu;
596 
597 	/* used for guest single stepping over the given code position */
598 	unsigned long singlestep_rip;
599 
600 	struct kvm_vcpu_hv hyperv;
601 
602 	cpumask_var_t wbinvd_dirty_mask;
603 
604 	unsigned long last_retry_eip;
605 	unsigned long last_retry_addr;
606 
607 	struct {
608 		bool halted;
609 		gfn_t gfns[roundup_pow_of_two(ASYNC_PF_PER_VCPU)];
610 		struct gfn_to_hva_cache data;
611 		u64 msr_val;
612 		u32 id;
613 		bool send_user_only;
614 	} apf;
615 
616 	/* OSVW MSRs (AMD only) */
617 	struct {
618 		u64 length;
619 		u64 status;
620 	} osvw;
621 
622 	struct {
623 		u64 msr_val;
624 		struct gfn_to_hva_cache data;
625 	} pv_eoi;
626 
627 	/*
628 	 * Indicate whether the access faults on its page table in guest
629 	 * which is set when fix page fault and used to detect unhandeable
630 	 * instruction.
631 	 */
632 	bool write_fault_to_shadow_pgtable;
633 
634 	/* set at EPT violation at this point */
635 	unsigned long exit_qualification;
636 
637 	/* pv related host specific info */
638 	struct {
639 		bool pv_unhalted;
640 	} pv;
641 
642 	int pending_ioapic_eoi;
643 	int pending_external_vector;
644 };
645 
646 struct kvm_lpage_info {
647 	int write_count;
648 };
649 
650 struct kvm_arch_memory_slot {
651 	struct kvm_rmap_head *rmap[KVM_NR_PAGE_SIZES];
652 	struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
653 };
654 
655 /*
656  * We use as the mode the number of bits allocated in the LDR for the
657  * logical processor ID.  It happens that these are all powers of two.
658  * This makes it is very easy to detect cases where the APICs are
659  * configured for multiple modes; in that case, we cannot use the map and
660  * hence cannot use kvm_irq_delivery_to_apic_fast either.
661  */
662 #define KVM_APIC_MODE_XAPIC_CLUSTER          4
663 #define KVM_APIC_MODE_XAPIC_FLAT             8
664 #define KVM_APIC_MODE_X2APIC                16
665 
666 struct kvm_apic_map {
667 	struct rcu_head rcu;
668 	u8 mode;
669 	struct kvm_lapic *phys_map[256];
670 	/* first index is cluster id second is cpu id in a cluster */
671 	struct kvm_lapic *logical_map[16][16];
672 };
673 
674 /* Hyper-V emulation context */
675 struct kvm_hv {
676 	u64 hv_guest_os_id;
677 	u64 hv_hypercall;
678 	u64 hv_tsc_page;
679 
680 	/* Hyper-v based guest crash (NT kernel bugcheck) parameters */
681 	u64 hv_crash_param[HV_X64_MSR_CRASH_PARAMS];
682 	u64 hv_crash_ctl;
683 };
684 
685 struct kvm_arch {
686 	unsigned int n_used_mmu_pages;
687 	unsigned int n_requested_mmu_pages;
688 	unsigned int n_max_mmu_pages;
689 	unsigned int indirect_shadow_pages;
690 	unsigned long mmu_valid_gen;
691 	struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
692 	/*
693 	 * Hash table of struct kvm_mmu_page.
694 	 */
695 	struct list_head active_mmu_pages;
696 	struct list_head zapped_obsolete_pages;
697 
698 	struct list_head assigned_dev_head;
699 	struct iommu_domain *iommu_domain;
700 	bool iommu_noncoherent;
701 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA
702 	atomic_t noncoherent_dma_count;
703 #define __KVM_HAVE_ARCH_ASSIGNED_DEVICE
704 	atomic_t assigned_device_count;
705 	struct kvm_pic *vpic;
706 	struct kvm_ioapic *vioapic;
707 	struct kvm_pit *vpit;
708 	atomic_t vapics_in_nmi_mode;
709 	struct mutex apic_map_lock;
710 	struct kvm_apic_map *apic_map;
711 
712 	unsigned int tss_addr;
713 	bool apic_access_page_done;
714 
715 	gpa_t wall_clock;
716 
717 	bool ept_identity_pagetable_done;
718 	gpa_t ept_identity_map_addr;
719 
720 	unsigned long irq_sources_bitmap;
721 	s64 kvmclock_offset;
722 	raw_spinlock_t tsc_write_lock;
723 	u64 last_tsc_nsec;
724 	u64 last_tsc_write;
725 	u32 last_tsc_khz;
726 	u64 cur_tsc_nsec;
727 	u64 cur_tsc_write;
728 	u64 cur_tsc_offset;
729 	u64 cur_tsc_generation;
730 	int nr_vcpus_matched_tsc;
731 
732 	spinlock_t pvclock_gtod_sync_lock;
733 	bool use_master_clock;
734 	u64 master_kernel_ns;
735 	cycle_t master_cycle_now;
736 	struct delayed_work kvmclock_update_work;
737 	struct delayed_work kvmclock_sync_work;
738 
739 	struct kvm_xen_hvm_config xen_hvm_config;
740 
741 	/* reads protected by irq_srcu, writes by irq_lock */
742 	struct hlist_head mask_notifier_list;
743 
744 	struct kvm_hv hyperv;
745 
746 	#ifdef CONFIG_KVM_MMU_AUDIT
747 	int audit_point;
748 	#endif
749 
750 	bool boot_vcpu_runs_old_kvmclock;
751 	u32 bsp_vcpu_id;
752 
753 	u64 disabled_quirks;
754 
755 	bool irqchip_split;
756 	u8 nr_reserved_ioapic_pins;
757 };
758 
759 struct kvm_vm_stat {
760 	u32 mmu_shadow_zapped;
761 	u32 mmu_pte_write;
762 	u32 mmu_pte_updated;
763 	u32 mmu_pde_zapped;
764 	u32 mmu_flooded;
765 	u32 mmu_recycled;
766 	u32 mmu_cache_miss;
767 	u32 mmu_unsync;
768 	u32 remote_tlb_flush;
769 	u32 lpages;
770 };
771 
772 struct kvm_vcpu_stat {
773 	u32 pf_fixed;
774 	u32 pf_guest;
775 	u32 tlb_flush;
776 	u32 invlpg;
777 
778 	u32 exits;
779 	u32 io_exits;
780 	u32 mmio_exits;
781 	u32 signal_exits;
782 	u32 irq_window_exits;
783 	u32 nmi_window_exits;
784 	u32 halt_exits;
785 	u32 halt_successful_poll;
786 	u32 halt_attempted_poll;
787 	u32 halt_wakeup;
788 	u32 request_irq_exits;
789 	u32 irq_exits;
790 	u32 host_state_reload;
791 	u32 efer_reload;
792 	u32 fpu_reload;
793 	u32 insn_emulation;
794 	u32 insn_emulation_fail;
795 	u32 hypercalls;
796 	u32 irq_injections;
797 	u32 nmi_injections;
798 };
799 
800 struct x86_instruction_info;
801 
802 struct msr_data {
803 	bool host_initiated;
804 	u32 index;
805 	u64 data;
806 };
807 
808 struct kvm_lapic_irq {
809 	u32 vector;
810 	u16 delivery_mode;
811 	u16 dest_mode;
812 	bool level;
813 	u16 trig_mode;
814 	u32 shorthand;
815 	u32 dest_id;
816 	bool msi_redir_hint;
817 };
818 
819 struct kvm_x86_ops {
820 	int (*cpu_has_kvm_support)(void);          /* __init */
821 	int (*disabled_by_bios)(void);             /* __init */
822 	int (*hardware_enable)(void);
823 	void (*hardware_disable)(void);
824 	void (*check_processor_compatibility)(void *rtn);
825 	int (*hardware_setup)(void);               /* __init */
826 	void (*hardware_unsetup)(void);            /* __exit */
827 	bool (*cpu_has_accelerated_tpr)(void);
828 	bool (*cpu_has_high_real_mode_segbase)(void);
829 	void (*cpuid_update)(struct kvm_vcpu *vcpu);
830 
831 	/* Create, but do not attach this VCPU */
832 	struct kvm_vcpu *(*vcpu_create)(struct kvm *kvm, unsigned id);
833 	void (*vcpu_free)(struct kvm_vcpu *vcpu);
834 	void (*vcpu_reset)(struct kvm_vcpu *vcpu, bool init_event);
835 
836 	void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
837 	void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
838 	void (*vcpu_put)(struct kvm_vcpu *vcpu);
839 
840 	void (*update_bp_intercept)(struct kvm_vcpu *vcpu);
841 	int (*get_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
842 	int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
843 	u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
844 	void (*get_segment)(struct kvm_vcpu *vcpu,
845 			    struct kvm_segment *var, int seg);
846 	int (*get_cpl)(struct kvm_vcpu *vcpu);
847 	void (*set_segment)(struct kvm_vcpu *vcpu,
848 			    struct kvm_segment *var, int seg);
849 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
850 	void (*decache_cr0_guest_bits)(struct kvm_vcpu *vcpu);
851 	void (*decache_cr3)(struct kvm_vcpu *vcpu);
852 	void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu);
853 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
854 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
855 	int (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
856 	void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
857 	void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
858 	void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
859 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
860 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
861 	u64 (*get_dr6)(struct kvm_vcpu *vcpu);
862 	void (*set_dr6)(struct kvm_vcpu *vcpu, unsigned long value);
863 	void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu);
864 	void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
865 	void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
866 	unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
867 	void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
868 	void (*fpu_activate)(struct kvm_vcpu *vcpu);
869 	void (*fpu_deactivate)(struct kvm_vcpu *vcpu);
870 
871 	void (*tlb_flush)(struct kvm_vcpu *vcpu);
872 
873 	void (*run)(struct kvm_vcpu *vcpu);
874 	int (*handle_exit)(struct kvm_vcpu *vcpu);
875 	void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
876 	void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
877 	u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu);
878 	void (*patch_hypercall)(struct kvm_vcpu *vcpu,
879 				unsigned char *hypercall_addr);
880 	void (*set_irq)(struct kvm_vcpu *vcpu);
881 	void (*set_nmi)(struct kvm_vcpu *vcpu);
882 	void (*queue_exception)(struct kvm_vcpu *vcpu, unsigned nr,
883 				bool has_error_code, u32 error_code,
884 				bool reinject);
885 	void (*cancel_injection)(struct kvm_vcpu *vcpu);
886 	int (*interrupt_allowed)(struct kvm_vcpu *vcpu);
887 	int (*nmi_allowed)(struct kvm_vcpu *vcpu);
888 	bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
889 	void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
890 	void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
891 	void (*enable_irq_window)(struct kvm_vcpu *vcpu);
892 	void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
893 	bool (*get_enable_apicv)(void);
894 	void (*refresh_apicv_exec_ctrl)(struct kvm_vcpu *vcpu);
895 	void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
896 	void (*hwapic_isr_update)(struct kvm *kvm, int isr);
897 	void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
898 	void (*set_virtual_x2apic_mode)(struct kvm_vcpu *vcpu, bool set);
899 	void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu, hpa_t hpa);
900 	void (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
901 	void (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
902 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
903 	int (*get_tdp_level)(void);
904 	u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
905 	int (*get_lpage_level)(void);
906 	bool (*rdtscp_supported)(void);
907 	bool (*invpcid_supported)(void);
908 	void (*adjust_tsc_offset_guest)(struct kvm_vcpu *vcpu, s64 adjustment);
909 
910 	void (*set_tdp_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
911 
912 	void (*set_supported_cpuid)(u32 func, struct kvm_cpuid_entry2 *entry);
913 
914 	bool (*has_wbinvd_exit)(void);
915 
916 	u64 (*read_tsc_offset)(struct kvm_vcpu *vcpu);
917 	void (*write_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
918 
919 	u64 (*read_l1_tsc)(struct kvm_vcpu *vcpu, u64 host_tsc);
920 
921 	void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2);
922 
923 	int (*check_intercept)(struct kvm_vcpu *vcpu,
924 			       struct x86_instruction_info *info,
925 			       enum x86_intercept_stage stage);
926 	void (*handle_external_intr)(struct kvm_vcpu *vcpu);
927 	bool (*mpx_supported)(void);
928 	bool (*xsaves_supported)(void);
929 
930 	int (*check_nested_events)(struct kvm_vcpu *vcpu, bool external_intr);
931 
932 	void (*sched_in)(struct kvm_vcpu *kvm, int cpu);
933 
934 	/*
935 	 * Arch-specific dirty logging hooks. These hooks are only supposed to
936 	 * be valid if the specific arch has hardware-accelerated dirty logging
937 	 * mechanism. Currently only for PML on VMX.
938 	 *
939 	 *  - slot_enable_log_dirty:
940 	 *	called when enabling log dirty mode for the slot.
941 	 *  - slot_disable_log_dirty:
942 	 *	called when disabling log dirty mode for the slot.
943 	 *	also called when slot is created with log dirty disabled.
944 	 *  - flush_log_dirty:
945 	 *	called before reporting dirty_bitmap to userspace.
946 	 *  - enable_log_dirty_pt_masked:
947 	 *	called when reenabling log dirty for the GFNs in the mask after
948 	 *	corresponding bits are cleared in slot->dirty_bitmap.
949 	 */
950 	void (*slot_enable_log_dirty)(struct kvm *kvm,
951 				      struct kvm_memory_slot *slot);
952 	void (*slot_disable_log_dirty)(struct kvm *kvm,
953 				       struct kvm_memory_slot *slot);
954 	void (*flush_log_dirty)(struct kvm *kvm);
955 	void (*enable_log_dirty_pt_masked)(struct kvm *kvm,
956 					   struct kvm_memory_slot *slot,
957 					   gfn_t offset, unsigned long mask);
958 	/* pmu operations of sub-arch */
959 	const struct kvm_pmu_ops *pmu_ops;
960 
961 	/*
962 	 * Architecture specific hooks for vCPU blocking due to
963 	 * HLT instruction.
964 	 * Returns for .pre_block():
965 	 *    - 0 means continue to block the vCPU.
966 	 *    - 1 means we cannot block the vCPU since some event
967 	 *        happens during this period, such as, 'ON' bit in
968 	 *        posted-interrupts descriptor is set.
969 	 */
970 	int (*pre_block)(struct kvm_vcpu *vcpu);
971 	void (*post_block)(struct kvm_vcpu *vcpu);
972 	int (*update_pi_irte)(struct kvm *kvm, unsigned int host_irq,
973 			      uint32_t guest_irq, bool set);
974 };
975 
976 struct kvm_arch_async_pf {
977 	u32 token;
978 	gfn_t gfn;
979 	unsigned long cr3;
980 	bool direct_map;
981 };
982 
983 extern struct kvm_x86_ops *kvm_x86_ops;
984 
985 int kvm_mmu_module_init(void);
986 void kvm_mmu_module_exit(void);
987 
988 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
989 int kvm_mmu_create(struct kvm_vcpu *vcpu);
990 void kvm_mmu_setup(struct kvm_vcpu *vcpu);
991 void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
992 		u64 dirty_mask, u64 nx_mask, u64 x_mask);
993 
994 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
995 void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
996 				      struct kvm_memory_slot *memslot);
997 void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
998 				   const struct kvm_memory_slot *memslot);
999 void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
1000 				   struct kvm_memory_slot *memslot);
1001 void kvm_mmu_slot_largepage_remove_write_access(struct kvm *kvm,
1002 					struct kvm_memory_slot *memslot);
1003 void kvm_mmu_slot_set_dirty(struct kvm *kvm,
1004 			    struct kvm_memory_slot *memslot);
1005 void kvm_mmu_clear_dirty_pt_masked(struct kvm *kvm,
1006 				   struct kvm_memory_slot *slot,
1007 				   gfn_t gfn_offset, unsigned long mask);
1008 void kvm_mmu_zap_all(struct kvm *kvm);
1009 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, struct kvm_memslots *slots);
1010 unsigned int kvm_mmu_calculate_mmu_pages(struct kvm *kvm);
1011 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int kvm_nr_mmu_pages);
1012 
1013 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
1014 
1015 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
1016 			  const void *val, int bytes);
1017 
1018 struct kvm_irq_mask_notifier {
1019 	void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
1020 	int irq;
1021 	struct hlist_node link;
1022 };
1023 
1024 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
1025 				    struct kvm_irq_mask_notifier *kimn);
1026 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
1027 				      struct kvm_irq_mask_notifier *kimn);
1028 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
1029 			     bool mask);
1030 
1031 extern bool tdp_enabled;
1032 
1033 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
1034 
1035 /* control of guest tsc rate supported? */
1036 extern bool kvm_has_tsc_control;
1037 /* maximum supported tsc_khz for guests */
1038 extern u32  kvm_max_guest_tsc_khz;
1039 /* number of bits of the fractional part of the TSC scaling ratio */
1040 extern u8   kvm_tsc_scaling_ratio_frac_bits;
1041 /* maximum allowed value of TSC scaling ratio */
1042 extern u64  kvm_max_tsc_scaling_ratio;
1043 
1044 enum emulation_result {
1045 	EMULATE_DONE,         /* no further processing */
1046 	EMULATE_USER_EXIT,    /* kvm_run ready for userspace exit */
1047 	EMULATE_FAIL,         /* can't emulate this instruction */
1048 };
1049 
1050 #define EMULTYPE_NO_DECODE	    (1 << 0)
1051 #define EMULTYPE_TRAP_UD	    (1 << 1)
1052 #define EMULTYPE_SKIP		    (1 << 2)
1053 #define EMULTYPE_RETRY		    (1 << 3)
1054 #define EMULTYPE_NO_REEXECUTE	    (1 << 4)
1055 int x86_emulate_instruction(struct kvm_vcpu *vcpu, unsigned long cr2,
1056 			    int emulation_type, void *insn, int insn_len);
1057 
1058 static inline int emulate_instruction(struct kvm_vcpu *vcpu,
1059 			int emulation_type)
1060 {
1061 	return x86_emulate_instruction(vcpu, 0, emulation_type, NULL, 0);
1062 }
1063 
1064 void kvm_enable_efer_bits(u64);
1065 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
1066 int kvm_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr);
1067 int kvm_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr);
1068 
1069 struct x86_emulate_ctxt;
1070 
1071 int kvm_fast_pio_out(struct kvm_vcpu *vcpu, int size, unsigned short port);
1072 void kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
1073 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
1074 int kvm_vcpu_halt(struct kvm_vcpu *vcpu);
1075 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
1076 
1077 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
1078 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
1079 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector);
1080 
1081 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
1082 		    int reason, bool has_error_code, u32 error_code);
1083 
1084 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
1085 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
1086 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
1087 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
1088 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
1089 int kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
1090 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
1091 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
1092 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
1093 int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr);
1094 
1095 int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1096 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1097 
1098 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
1099 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
1100 bool kvm_rdpmc(struct kvm_vcpu *vcpu);
1101 
1102 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1103 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1104 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1105 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1106 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
1107 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1108 			    gfn_t gfn, void *data, int offset, int len,
1109 			    u32 access);
1110 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
1111 bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr);
1112 
1113 static inline int __kvm_irq_line_state(unsigned long *irq_state,
1114 				       int irq_source_id, int level)
1115 {
1116 	/* Logical OR for level trig interrupt */
1117 	if (level)
1118 		__set_bit(irq_source_id, irq_state);
1119 	else
1120 		__clear_bit(irq_source_id, irq_state);
1121 
1122 	return !!(*irq_state);
1123 }
1124 
1125 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
1126 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
1127 
1128 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
1129 
1130 void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
1131 		       const u8 *new, int bytes);
1132 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
1133 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
1134 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
1135 int kvm_mmu_load(struct kvm_vcpu *vcpu);
1136 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
1137 void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu);
1138 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1139 			   struct x86_exception *exception);
1140 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
1141 			      struct x86_exception *exception);
1142 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
1143 			       struct x86_exception *exception);
1144 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
1145 			       struct x86_exception *exception);
1146 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
1147 				struct x86_exception *exception);
1148 
1149 void kvm_vcpu_deactivate_apicv(struct kvm_vcpu *vcpu);
1150 
1151 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
1152 
1153 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva, u32 error_code,
1154 		       void *insn, int insn_len);
1155 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
1156 void kvm_mmu_new_cr3(struct kvm_vcpu *vcpu);
1157 
1158 void kvm_enable_tdp(void);
1159 void kvm_disable_tdp(void);
1160 
1161 static inline gpa_t translate_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1162 				  struct x86_exception *exception)
1163 {
1164 	return gpa;
1165 }
1166 
1167 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
1168 {
1169 	struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
1170 
1171 	return (struct kvm_mmu_page *)page_private(page);
1172 }
1173 
1174 static inline u16 kvm_read_ldt(void)
1175 {
1176 	u16 ldt;
1177 	asm("sldt %0" : "=g"(ldt));
1178 	return ldt;
1179 }
1180 
1181 static inline void kvm_load_ldt(u16 sel)
1182 {
1183 	asm("lldt %0" : : "rm"(sel));
1184 }
1185 
1186 #ifdef CONFIG_X86_64
1187 static inline unsigned long read_msr(unsigned long msr)
1188 {
1189 	u64 value;
1190 
1191 	rdmsrl(msr, value);
1192 	return value;
1193 }
1194 #endif
1195 
1196 static inline u32 get_rdx_init_val(void)
1197 {
1198 	return 0x600; /* P6 family */
1199 }
1200 
1201 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
1202 {
1203 	kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
1204 }
1205 
1206 static inline u64 get_canonical(u64 la)
1207 {
1208 	return ((int64_t)la << 16) >> 16;
1209 }
1210 
1211 static inline bool is_noncanonical_address(u64 la)
1212 {
1213 #ifdef CONFIG_X86_64
1214 	return get_canonical(la) != la;
1215 #else
1216 	return false;
1217 #endif
1218 }
1219 
1220 #define TSS_IOPB_BASE_OFFSET 0x66
1221 #define TSS_BASE_SIZE 0x68
1222 #define TSS_IOPB_SIZE (65536 / 8)
1223 #define TSS_REDIRECTION_SIZE (256 / 8)
1224 #define RMODE_TSS_SIZE							\
1225 	(TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
1226 
1227 enum {
1228 	TASK_SWITCH_CALL = 0,
1229 	TASK_SWITCH_IRET = 1,
1230 	TASK_SWITCH_JMP = 2,
1231 	TASK_SWITCH_GATE = 3,
1232 };
1233 
1234 #define HF_GIF_MASK		(1 << 0)
1235 #define HF_HIF_MASK		(1 << 1)
1236 #define HF_VINTR_MASK		(1 << 2)
1237 #define HF_NMI_MASK		(1 << 3)
1238 #define HF_IRET_MASK		(1 << 4)
1239 #define HF_GUEST_MASK		(1 << 5) /* VCPU is in guest-mode */
1240 #define HF_SMM_MASK		(1 << 6)
1241 #define HF_SMM_INSIDE_NMI_MASK	(1 << 7)
1242 
1243 #define __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
1244 #define KVM_ADDRESS_SPACE_NUM 2
1245 
1246 #define kvm_arch_vcpu_memslots_id(vcpu) ((vcpu)->arch.hflags & HF_SMM_MASK ? 1 : 0)
1247 #define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, (role).smm)
1248 
1249 /*
1250  * Hardware virtualization extension instructions may fault if a
1251  * reboot turns off virtualization while processes are running.
1252  * Trap the fault and ignore the instruction if that happens.
1253  */
1254 asmlinkage void kvm_spurious_fault(void);
1255 
1256 #define ____kvm_handle_fault_on_reboot(insn, cleanup_insn)	\
1257 	"666: " insn "\n\t" \
1258 	"668: \n\t"                           \
1259 	".pushsection .fixup, \"ax\" \n" \
1260 	"667: \n\t" \
1261 	cleanup_insn "\n\t"		      \
1262 	"cmpb $0, kvm_rebooting \n\t"	      \
1263 	"jne 668b \n\t"      		      \
1264 	__ASM_SIZE(push) " $666b \n\t"	      \
1265 	"call kvm_spurious_fault \n\t"	      \
1266 	".popsection \n\t" \
1267 	_ASM_EXTABLE(666b, 667b)
1268 
1269 #define __kvm_handle_fault_on_reboot(insn)		\
1270 	____kvm_handle_fault_on_reboot(insn, "")
1271 
1272 #define KVM_ARCH_WANT_MMU_NOTIFIER
1273 int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
1274 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end);
1275 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
1276 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
1277 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
1278 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1279 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1280 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1281 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1282 void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event);
1283 void kvm_vcpu_reload_apic_access_page(struct kvm_vcpu *vcpu);
1284 void kvm_arch_mmu_notifier_invalidate_page(struct kvm *kvm,
1285 					   unsigned long address);
1286 
1287 void kvm_define_shared_msr(unsigned index, u32 msr);
1288 int kvm_set_shared_msr(unsigned index, u64 val, u64 mask);
1289 
1290 u64 kvm_scale_tsc(struct kvm_vcpu *vcpu, u64 tsc);
1291 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc);
1292 
1293 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu);
1294 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1295 
1296 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
1297 void kvm_make_scan_ioapic_request(struct kvm *kvm);
1298 
1299 void kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1300 				     struct kvm_async_pf *work);
1301 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1302 				 struct kvm_async_pf *work);
1303 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1304 			       struct kvm_async_pf *work);
1305 bool kvm_arch_can_inject_async_page_present(struct kvm_vcpu *vcpu);
1306 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1307 
1308 void kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1309 
1310 int kvm_is_in_guest(void);
1311 
1312 int __x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa, u32 size);
1313 int x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa, u32 size);
1314 bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu);
1315 bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu);
1316 
1317 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
1318 			     struct kvm_vcpu **dest_vcpu);
1319 
1320 void kvm_set_msi_irq(struct kvm_kernel_irq_routing_entry *e,
1321 		     struct kvm_lapic_irq *irq);
1322 
1323 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu) {}
1324 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu) {}
1325 
1326 #endif /* _ASM_X86_KVM_HOST_H */
1327