xref: /openbmc/linux/arch/x86/include/asm/kvm_host.h (revision 930beb5a)
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 
28 #include <asm/pvclock-abi.h>
29 #include <asm/desc.h>
30 #include <asm/mtrr.h>
31 #include <asm/msr-index.h>
32 #include <asm/asm.h>
33 
34 #define KVM_MAX_VCPUS 255
35 #define KVM_SOFT_MAX_VCPUS 160
36 #define KVM_USER_MEM_SLOTS 125
37 /* memory slots that are not exposed to userspace */
38 #define KVM_PRIVATE_MEM_SLOTS 3
39 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
40 
41 #define KVM_MMIO_SIZE 16
42 
43 #define KVM_PIO_PAGE_OFFSET 1
44 #define KVM_COALESCED_MMIO_PAGE_OFFSET 2
45 
46 #define KVM_IRQCHIP_NUM_PINS  KVM_IOAPIC_NUM_PINS
47 
48 #define CR0_RESERVED_BITS                                               \
49 	(~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \
50 			  | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \
51 			  | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG))
52 
53 #define CR3_PAE_RESERVED_BITS ((X86_CR3_PWT | X86_CR3_PCD) - 1)
54 #define CR3_NONPAE_RESERVED_BITS ((PAGE_SIZE-1) & ~(X86_CR3_PWT | X86_CR3_PCD))
55 #define CR3_PCID_ENABLED_RESERVED_BITS 0xFFFFFF0000000000ULL
56 #define CR3_L_MODE_RESERVED_BITS (CR3_NONPAE_RESERVED_BITS |	\
57 				  0xFFFFFF0000000000ULL)
58 #define CR4_RESERVED_BITS                                               \
59 	(~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\
60 			  | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE     \
61 			  | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR | X86_CR4_PCIDE \
62 			  | X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_FSGSBASE \
63 			  | X86_CR4_OSXMMEXCPT | X86_CR4_VMXE))
64 
65 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR)
66 
67 
68 
69 #define INVALID_PAGE (~(hpa_t)0)
70 #define VALID_PAGE(x) ((x) != INVALID_PAGE)
71 
72 #define UNMAPPED_GVA (~(gpa_t)0)
73 
74 /* KVM Hugepage definitions for x86 */
75 #define KVM_NR_PAGE_SIZES	3
76 #define KVM_HPAGE_GFN_SHIFT(x)	(((x) - 1) * 9)
77 #define KVM_HPAGE_SHIFT(x)	(PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
78 #define KVM_HPAGE_SIZE(x)	(1UL << KVM_HPAGE_SHIFT(x))
79 #define KVM_HPAGE_MASK(x)	(~(KVM_HPAGE_SIZE(x) - 1))
80 #define KVM_PAGES_PER_HPAGE(x)	(KVM_HPAGE_SIZE(x) / PAGE_SIZE)
81 
82 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
83 {
84 	/* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */
85 	return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
86 		(base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
87 }
88 
89 #define SELECTOR_TI_MASK (1 << 2)
90 #define SELECTOR_RPL_MASK 0x03
91 
92 #define IOPL_SHIFT 12
93 
94 #define KVM_PERMILLE_MMU_PAGES 20
95 #define KVM_MIN_ALLOC_MMU_PAGES 64
96 #define KVM_MMU_HASH_SHIFT 10
97 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
98 #define KVM_MIN_FREE_MMU_PAGES 5
99 #define KVM_REFILL_PAGES 25
100 #define KVM_MAX_CPUID_ENTRIES 80
101 #define KVM_NR_FIXED_MTRR_REGION 88
102 #define KVM_NR_VAR_MTRR 8
103 
104 #define ASYNC_PF_PER_VCPU 64
105 
106 struct kvm_vcpu;
107 struct kvm;
108 struct kvm_async_pf;
109 
110 enum kvm_reg {
111 	VCPU_REGS_RAX = 0,
112 	VCPU_REGS_RCX = 1,
113 	VCPU_REGS_RDX = 2,
114 	VCPU_REGS_RBX = 3,
115 	VCPU_REGS_RSP = 4,
116 	VCPU_REGS_RBP = 5,
117 	VCPU_REGS_RSI = 6,
118 	VCPU_REGS_RDI = 7,
119 #ifdef CONFIG_X86_64
120 	VCPU_REGS_R8 = 8,
121 	VCPU_REGS_R9 = 9,
122 	VCPU_REGS_R10 = 10,
123 	VCPU_REGS_R11 = 11,
124 	VCPU_REGS_R12 = 12,
125 	VCPU_REGS_R13 = 13,
126 	VCPU_REGS_R14 = 14,
127 	VCPU_REGS_R15 = 15,
128 #endif
129 	VCPU_REGS_RIP,
130 	NR_VCPU_REGS
131 };
132 
133 enum kvm_reg_ex {
134 	VCPU_EXREG_PDPTR = NR_VCPU_REGS,
135 	VCPU_EXREG_CR3,
136 	VCPU_EXREG_RFLAGS,
137 	VCPU_EXREG_CPL,
138 	VCPU_EXREG_SEGMENTS,
139 };
140 
141 enum {
142 	VCPU_SREG_ES,
143 	VCPU_SREG_CS,
144 	VCPU_SREG_SS,
145 	VCPU_SREG_DS,
146 	VCPU_SREG_FS,
147 	VCPU_SREG_GS,
148 	VCPU_SREG_TR,
149 	VCPU_SREG_LDTR,
150 };
151 
152 #include <asm/kvm_emulate.h>
153 
154 #define KVM_NR_MEM_OBJS 40
155 
156 #define KVM_NR_DB_REGS	4
157 
158 #define DR6_BD		(1 << 13)
159 #define DR6_BS		(1 << 14)
160 #define DR6_FIXED_1	0xffff0ff0
161 #define DR6_VOLATILE	0x0000e00f
162 
163 #define DR7_BP_EN_MASK	0x000000ff
164 #define DR7_GE		(1 << 9)
165 #define DR7_GD		(1 << 13)
166 #define DR7_FIXED_1	0x00000400
167 #define DR7_VOLATILE	0xffff23ff
168 
169 /* apic attention bits */
170 #define KVM_APIC_CHECK_VAPIC	0
171 /*
172  * The following bit is set with PV-EOI, unset on EOI.
173  * We detect PV-EOI changes by guest by comparing
174  * this bit with PV-EOI in guest memory.
175  * See the implementation in apic_update_pv_eoi.
176  */
177 #define KVM_APIC_PV_EOI_PENDING	1
178 
179 /*
180  * We don't want allocation failures within the mmu code, so we preallocate
181  * enough memory for a single page fault in a cache.
182  */
183 struct kvm_mmu_memory_cache {
184 	int nobjs;
185 	void *objects[KVM_NR_MEM_OBJS];
186 };
187 
188 /*
189  * kvm_mmu_page_role, below, is defined as:
190  *
191  *   bits 0:3 - total guest paging levels (2-4, or zero for real mode)
192  *   bits 4:7 - page table level for this shadow (1-4)
193  *   bits 8:9 - page table quadrant for 2-level guests
194  *   bit   16 - direct mapping of virtual to physical mapping at gfn
195  *              used for real mode and two-dimensional paging
196  *   bits 17:19 - common access permissions for all ptes in this shadow page
197  */
198 union kvm_mmu_page_role {
199 	unsigned word;
200 	struct {
201 		unsigned level:4;
202 		unsigned cr4_pae:1;
203 		unsigned quadrant:2;
204 		unsigned pad_for_nice_hex_output:6;
205 		unsigned direct:1;
206 		unsigned access:3;
207 		unsigned invalid:1;
208 		unsigned nxe:1;
209 		unsigned cr0_wp:1;
210 		unsigned smep_andnot_wp:1;
211 	};
212 };
213 
214 struct kvm_mmu_page {
215 	struct list_head link;
216 	struct hlist_node hash_link;
217 
218 	/*
219 	 * The following two entries are used to key the shadow page in the
220 	 * hash table.
221 	 */
222 	gfn_t gfn;
223 	union kvm_mmu_page_role role;
224 
225 	u64 *spt;
226 	/* hold the gfn of each spte inside spt */
227 	gfn_t *gfns;
228 	bool unsync;
229 	int root_count;          /* Currently serving as active root */
230 	unsigned int unsync_children;
231 	unsigned long parent_ptes;	/* Reverse mapping for parent_pte */
232 
233 	/* The page is obsolete if mmu_valid_gen != kvm->arch.mmu_valid_gen.  */
234 	unsigned long mmu_valid_gen;
235 
236 	DECLARE_BITMAP(unsync_child_bitmap, 512);
237 
238 #ifdef CONFIG_X86_32
239 	/*
240 	 * Used out of the mmu-lock to avoid reading spte values while an
241 	 * update is in progress; see the comments in __get_spte_lockless().
242 	 */
243 	int clear_spte_count;
244 #endif
245 
246 	/* Number of writes since the last time traversal visited this page.  */
247 	int write_flooding_count;
248 };
249 
250 struct kvm_pio_request {
251 	unsigned long count;
252 	int in;
253 	int port;
254 	int size;
255 };
256 
257 /*
258  * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
259  * 32-bit).  The kvm_mmu structure abstracts the details of the current mmu
260  * mode.
261  */
262 struct kvm_mmu {
263 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long root);
264 	unsigned long (*get_cr3)(struct kvm_vcpu *vcpu);
265 	u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
266 	int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err,
267 			  bool prefault);
268 	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
269 				  struct x86_exception *fault);
270 	gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva, u32 access,
271 			    struct x86_exception *exception);
272 	gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access);
273 	int (*sync_page)(struct kvm_vcpu *vcpu,
274 			 struct kvm_mmu_page *sp);
275 	void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva);
276 	void (*update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
277 			   u64 *spte, const void *pte);
278 	hpa_t root_hpa;
279 	int root_level;
280 	int shadow_root_level;
281 	union kvm_mmu_page_role base_role;
282 	bool direct_map;
283 
284 	/*
285 	 * Bitmap; bit set = permission fault
286 	 * Byte index: page fault error code [4:1]
287 	 * Bit index: pte permissions in ACC_* format
288 	 */
289 	u8 permissions[16];
290 
291 	u64 *pae_root;
292 	u64 *lm_root;
293 	u64 rsvd_bits_mask[2][4];
294 	u64 bad_mt_xwr;
295 
296 	/*
297 	 * Bitmap: bit set = last pte in walk
298 	 * index[0:1]: level (zero-based)
299 	 * index[2]: pte.ps
300 	 */
301 	u8 last_pte_bitmap;
302 
303 	bool nx;
304 
305 	u64 pdptrs[4]; /* pae */
306 };
307 
308 enum pmc_type {
309 	KVM_PMC_GP = 0,
310 	KVM_PMC_FIXED,
311 };
312 
313 struct kvm_pmc {
314 	enum pmc_type type;
315 	u8 idx;
316 	u64 counter;
317 	u64 eventsel;
318 	struct perf_event *perf_event;
319 	struct kvm_vcpu *vcpu;
320 };
321 
322 struct kvm_pmu {
323 	unsigned nr_arch_gp_counters;
324 	unsigned nr_arch_fixed_counters;
325 	unsigned available_event_types;
326 	u64 fixed_ctr_ctrl;
327 	u64 global_ctrl;
328 	u64 global_status;
329 	u64 global_ovf_ctrl;
330 	u64 counter_bitmask[2];
331 	u64 global_ctrl_mask;
332 	u64 reserved_bits;
333 	u8 version;
334 	struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
335 	struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED];
336 	struct irq_work irq_work;
337 	u64 reprogram_pmi;
338 };
339 
340 struct kvm_vcpu_arch {
341 	/*
342 	 * rip and regs accesses must go through
343 	 * kvm_{register,rip}_{read,write} functions.
344 	 */
345 	unsigned long regs[NR_VCPU_REGS];
346 	u32 regs_avail;
347 	u32 regs_dirty;
348 
349 	unsigned long cr0;
350 	unsigned long cr0_guest_owned_bits;
351 	unsigned long cr2;
352 	unsigned long cr3;
353 	unsigned long cr4;
354 	unsigned long cr4_guest_owned_bits;
355 	unsigned long cr8;
356 	u32 hflags;
357 	u64 efer;
358 	u64 apic_base;
359 	struct kvm_lapic *apic;    /* kernel irqchip context */
360 	unsigned long apic_attention;
361 	int32_t apic_arb_prio;
362 	int mp_state;
363 	u64 ia32_misc_enable_msr;
364 	bool tpr_access_reporting;
365 
366 	/*
367 	 * Paging state of the vcpu
368 	 *
369 	 * If the vcpu runs in guest mode with two level paging this still saves
370 	 * the paging mode of the l1 guest. This context is always used to
371 	 * handle faults.
372 	 */
373 	struct kvm_mmu mmu;
374 
375 	/*
376 	 * Paging state of an L2 guest (used for nested npt)
377 	 *
378 	 * This context will save all necessary information to walk page tables
379 	 * of the an L2 guest. This context is only initialized for page table
380 	 * walking and not for faulting since we never handle l2 page faults on
381 	 * the host.
382 	 */
383 	struct kvm_mmu nested_mmu;
384 
385 	/*
386 	 * Pointer to the mmu context currently used for
387 	 * gva_to_gpa translations.
388 	 */
389 	struct kvm_mmu *walk_mmu;
390 
391 	struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
392 	struct kvm_mmu_memory_cache mmu_page_cache;
393 	struct kvm_mmu_memory_cache mmu_page_header_cache;
394 
395 	struct fpu guest_fpu;
396 	u64 xcr0;
397 	u64 guest_supported_xcr0;
398 	u32 guest_xstate_size;
399 
400 	struct kvm_pio_request pio;
401 	void *pio_data;
402 
403 	u8 event_exit_inst_len;
404 
405 	struct kvm_queued_exception {
406 		bool pending;
407 		bool has_error_code;
408 		bool reinject;
409 		u8 nr;
410 		u32 error_code;
411 	} exception;
412 
413 	struct kvm_queued_interrupt {
414 		bool pending;
415 		bool soft;
416 		u8 nr;
417 	} interrupt;
418 
419 	int halt_request; /* real mode on Intel only */
420 
421 	int cpuid_nent;
422 	struct kvm_cpuid_entry2 cpuid_entries[KVM_MAX_CPUID_ENTRIES];
423 	/* emulate context */
424 
425 	struct x86_emulate_ctxt emulate_ctxt;
426 	bool emulate_regs_need_sync_to_vcpu;
427 	bool emulate_regs_need_sync_from_vcpu;
428 	int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
429 
430 	gpa_t time;
431 	struct pvclock_vcpu_time_info hv_clock;
432 	unsigned int hw_tsc_khz;
433 	struct gfn_to_hva_cache pv_time;
434 	bool pv_time_enabled;
435 	/* set guest stopped flag in pvclock flags field */
436 	bool pvclock_set_guest_stopped_request;
437 
438 	struct {
439 		u64 msr_val;
440 		u64 last_steal;
441 		u64 accum_steal;
442 		struct gfn_to_hva_cache stime;
443 		struct kvm_steal_time steal;
444 	} st;
445 
446 	u64 last_guest_tsc;
447 	u64 last_kernel_ns;
448 	u64 last_host_tsc;
449 	u64 tsc_offset_adjustment;
450 	u64 this_tsc_nsec;
451 	u64 this_tsc_write;
452 	u8  this_tsc_generation;
453 	bool tsc_catchup;
454 	bool tsc_always_catchup;
455 	s8 virtual_tsc_shift;
456 	u32 virtual_tsc_mult;
457 	u32 virtual_tsc_khz;
458 	s64 ia32_tsc_adjust_msr;
459 
460 	atomic_t nmi_queued;  /* unprocessed asynchronous NMIs */
461 	unsigned nmi_pending; /* NMI queued after currently running handler */
462 	bool nmi_injected;    /* Trying to inject an NMI this entry */
463 
464 	struct mtrr_state_type mtrr_state;
465 	u32 pat;
466 
467 	int switch_db_regs;
468 	unsigned long db[KVM_NR_DB_REGS];
469 	unsigned long dr6;
470 	unsigned long dr7;
471 	unsigned long eff_db[KVM_NR_DB_REGS];
472 	unsigned long guest_debug_dr7;
473 
474 	u64 mcg_cap;
475 	u64 mcg_status;
476 	u64 mcg_ctl;
477 	u64 *mce_banks;
478 
479 	/* Cache MMIO info */
480 	u64 mmio_gva;
481 	unsigned access;
482 	gfn_t mmio_gfn;
483 
484 	struct kvm_pmu pmu;
485 
486 	/* used for guest single stepping over the given code position */
487 	unsigned long singlestep_rip;
488 
489 	/* fields used by HYPER-V emulation */
490 	u64 hv_vapic;
491 
492 	cpumask_var_t wbinvd_dirty_mask;
493 
494 	unsigned long last_retry_eip;
495 	unsigned long last_retry_addr;
496 
497 	struct {
498 		bool halted;
499 		gfn_t gfns[roundup_pow_of_two(ASYNC_PF_PER_VCPU)];
500 		struct gfn_to_hva_cache data;
501 		u64 msr_val;
502 		u32 id;
503 		bool send_user_only;
504 	} apf;
505 
506 	/* OSVW MSRs (AMD only) */
507 	struct {
508 		u64 length;
509 		u64 status;
510 	} osvw;
511 
512 	struct {
513 		u64 msr_val;
514 		struct gfn_to_hva_cache data;
515 	} pv_eoi;
516 
517 	/*
518 	 * Indicate whether the access faults on its page table in guest
519 	 * which is set when fix page fault and used to detect unhandeable
520 	 * instruction.
521 	 */
522 	bool write_fault_to_shadow_pgtable;
523 
524 	/* set at EPT violation at this point */
525 	unsigned long exit_qualification;
526 
527 	/* pv related host specific info */
528 	struct {
529 		bool pv_unhalted;
530 	} pv;
531 };
532 
533 struct kvm_lpage_info {
534 	int write_count;
535 };
536 
537 struct kvm_arch_memory_slot {
538 	unsigned long *rmap[KVM_NR_PAGE_SIZES];
539 	struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
540 };
541 
542 struct kvm_apic_map {
543 	struct rcu_head rcu;
544 	u8 ldr_bits;
545 	/* fields bellow are used to decode ldr values in different modes */
546 	u32 cid_shift, cid_mask, lid_mask;
547 	struct kvm_lapic *phys_map[256];
548 	/* first index is cluster id second is cpu id in a cluster */
549 	struct kvm_lapic *logical_map[16][16];
550 };
551 
552 struct kvm_arch {
553 	unsigned int n_used_mmu_pages;
554 	unsigned int n_requested_mmu_pages;
555 	unsigned int n_max_mmu_pages;
556 	unsigned int indirect_shadow_pages;
557 	unsigned long mmu_valid_gen;
558 	struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
559 	/*
560 	 * Hash table of struct kvm_mmu_page.
561 	 */
562 	struct list_head active_mmu_pages;
563 	struct list_head zapped_obsolete_pages;
564 
565 	struct list_head assigned_dev_head;
566 	struct iommu_domain *iommu_domain;
567 	bool iommu_noncoherent;
568 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA
569 	atomic_t noncoherent_dma_count;
570 	struct kvm_pic *vpic;
571 	struct kvm_ioapic *vioapic;
572 	struct kvm_pit *vpit;
573 	int vapics_in_nmi_mode;
574 	struct mutex apic_map_lock;
575 	struct kvm_apic_map *apic_map;
576 
577 	unsigned int tss_addr;
578 	struct page *apic_access_page;
579 
580 	gpa_t wall_clock;
581 
582 	struct page *ept_identity_pagetable;
583 	bool ept_identity_pagetable_done;
584 	gpa_t ept_identity_map_addr;
585 
586 	unsigned long irq_sources_bitmap;
587 	s64 kvmclock_offset;
588 	raw_spinlock_t tsc_write_lock;
589 	u64 last_tsc_nsec;
590 	u64 last_tsc_write;
591 	u32 last_tsc_khz;
592 	u64 cur_tsc_nsec;
593 	u64 cur_tsc_write;
594 	u64 cur_tsc_offset;
595 	u8  cur_tsc_generation;
596 	int nr_vcpus_matched_tsc;
597 
598 	spinlock_t pvclock_gtod_sync_lock;
599 	bool use_master_clock;
600 	u64 master_kernel_ns;
601 	cycle_t master_cycle_now;
602 
603 	struct kvm_xen_hvm_config xen_hvm_config;
604 
605 	/* fields used by HYPER-V emulation */
606 	u64 hv_guest_os_id;
607 	u64 hv_hypercall;
608 
609 	#ifdef CONFIG_KVM_MMU_AUDIT
610 	int audit_point;
611 	#endif
612 };
613 
614 struct kvm_vm_stat {
615 	u32 mmu_shadow_zapped;
616 	u32 mmu_pte_write;
617 	u32 mmu_pte_updated;
618 	u32 mmu_pde_zapped;
619 	u32 mmu_flooded;
620 	u32 mmu_recycled;
621 	u32 mmu_cache_miss;
622 	u32 mmu_unsync;
623 	u32 remote_tlb_flush;
624 	u32 lpages;
625 };
626 
627 struct kvm_vcpu_stat {
628 	u32 pf_fixed;
629 	u32 pf_guest;
630 	u32 tlb_flush;
631 	u32 invlpg;
632 
633 	u32 exits;
634 	u32 io_exits;
635 	u32 mmio_exits;
636 	u32 signal_exits;
637 	u32 irq_window_exits;
638 	u32 nmi_window_exits;
639 	u32 halt_exits;
640 	u32 halt_wakeup;
641 	u32 request_irq_exits;
642 	u32 irq_exits;
643 	u32 host_state_reload;
644 	u32 efer_reload;
645 	u32 fpu_reload;
646 	u32 insn_emulation;
647 	u32 insn_emulation_fail;
648 	u32 hypercalls;
649 	u32 irq_injections;
650 	u32 nmi_injections;
651 };
652 
653 struct x86_instruction_info;
654 
655 struct msr_data {
656 	bool host_initiated;
657 	u32 index;
658 	u64 data;
659 };
660 
661 struct kvm_x86_ops {
662 	int (*cpu_has_kvm_support)(void);          /* __init */
663 	int (*disabled_by_bios)(void);             /* __init */
664 	int (*hardware_enable)(void *dummy);
665 	void (*hardware_disable)(void *dummy);
666 	void (*check_processor_compatibility)(void *rtn);
667 	int (*hardware_setup)(void);               /* __init */
668 	void (*hardware_unsetup)(void);            /* __exit */
669 	bool (*cpu_has_accelerated_tpr)(void);
670 	void (*cpuid_update)(struct kvm_vcpu *vcpu);
671 
672 	/* Create, but do not attach this VCPU */
673 	struct kvm_vcpu *(*vcpu_create)(struct kvm *kvm, unsigned id);
674 	void (*vcpu_free)(struct kvm_vcpu *vcpu);
675 	void (*vcpu_reset)(struct kvm_vcpu *vcpu);
676 
677 	void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
678 	void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
679 	void (*vcpu_put)(struct kvm_vcpu *vcpu);
680 
681 	void (*update_db_bp_intercept)(struct kvm_vcpu *vcpu);
682 	int (*get_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata);
683 	int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
684 	u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
685 	void (*get_segment)(struct kvm_vcpu *vcpu,
686 			    struct kvm_segment *var, int seg);
687 	int (*get_cpl)(struct kvm_vcpu *vcpu);
688 	void (*set_segment)(struct kvm_vcpu *vcpu,
689 			    struct kvm_segment *var, int seg);
690 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
691 	void (*decache_cr0_guest_bits)(struct kvm_vcpu *vcpu);
692 	void (*decache_cr3)(struct kvm_vcpu *vcpu);
693 	void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu);
694 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
695 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
696 	int (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
697 	void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
698 	void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
699 	void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
700 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
701 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
702 	void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
703 	void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
704 	unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
705 	void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
706 	void (*fpu_activate)(struct kvm_vcpu *vcpu);
707 	void (*fpu_deactivate)(struct kvm_vcpu *vcpu);
708 
709 	void (*tlb_flush)(struct kvm_vcpu *vcpu);
710 
711 	void (*run)(struct kvm_vcpu *vcpu);
712 	int (*handle_exit)(struct kvm_vcpu *vcpu);
713 	void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
714 	void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
715 	u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
716 	void (*patch_hypercall)(struct kvm_vcpu *vcpu,
717 				unsigned char *hypercall_addr);
718 	void (*set_irq)(struct kvm_vcpu *vcpu);
719 	void (*set_nmi)(struct kvm_vcpu *vcpu);
720 	void (*queue_exception)(struct kvm_vcpu *vcpu, unsigned nr,
721 				bool has_error_code, u32 error_code,
722 				bool reinject);
723 	void (*cancel_injection)(struct kvm_vcpu *vcpu);
724 	int (*interrupt_allowed)(struct kvm_vcpu *vcpu);
725 	int (*nmi_allowed)(struct kvm_vcpu *vcpu);
726 	bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
727 	void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
728 	int (*enable_nmi_window)(struct kvm_vcpu *vcpu);
729 	int (*enable_irq_window)(struct kvm_vcpu *vcpu);
730 	void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
731 	int (*vm_has_apicv)(struct kvm *kvm);
732 	void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
733 	void (*hwapic_isr_update)(struct kvm *kvm, int isr);
734 	void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
735 	void (*set_virtual_x2apic_mode)(struct kvm_vcpu *vcpu, bool set);
736 	void (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
737 	void (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
738 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
739 	int (*get_tdp_level)(void);
740 	u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
741 	int (*get_lpage_level)(void);
742 	bool (*rdtscp_supported)(void);
743 	bool (*invpcid_supported)(void);
744 	void (*adjust_tsc_offset)(struct kvm_vcpu *vcpu, s64 adjustment, bool host);
745 
746 	void (*set_tdp_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
747 
748 	void (*set_supported_cpuid)(u32 func, struct kvm_cpuid_entry2 *entry);
749 
750 	bool (*has_wbinvd_exit)(void);
751 
752 	void (*set_tsc_khz)(struct kvm_vcpu *vcpu, u32 user_tsc_khz, bool scale);
753 	u64 (*read_tsc_offset)(struct kvm_vcpu *vcpu);
754 	void (*write_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
755 
756 	u64 (*compute_tsc_offset)(struct kvm_vcpu *vcpu, u64 target_tsc);
757 	u64 (*read_l1_tsc)(struct kvm_vcpu *vcpu, u64 host_tsc);
758 
759 	void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2);
760 
761 	int (*check_intercept)(struct kvm_vcpu *vcpu,
762 			       struct x86_instruction_info *info,
763 			       enum x86_intercept_stage stage);
764 	void (*handle_external_intr)(struct kvm_vcpu *vcpu);
765 };
766 
767 struct kvm_arch_async_pf {
768 	u32 token;
769 	gfn_t gfn;
770 	unsigned long cr3;
771 	bool direct_map;
772 };
773 
774 extern struct kvm_x86_ops *kvm_x86_ops;
775 
776 static inline void adjust_tsc_offset_guest(struct kvm_vcpu *vcpu,
777 					   s64 adjustment)
778 {
779 	kvm_x86_ops->adjust_tsc_offset(vcpu, adjustment, false);
780 }
781 
782 static inline void adjust_tsc_offset_host(struct kvm_vcpu *vcpu, s64 adjustment)
783 {
784 	kvm_x86_ops->adjust_tsc_offset(vcpu, adjustment, true);
785 }
786 
787 int kvm_mmu_module_init(void);
788 void kvm_mmu_module_exit(void);
789 
790 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
791 int kvm_mmu_create(struct kvm_vcpu *vcpu);
792 void kvm_mmu_setup(struct kvm_vcpu *vcpu);
793 void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
794 		u64 dirty_mask, u64 nx_mask, u64 x_mask);
795 
796 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
797 void kvm_mmu_slot_remove_write_access(struct kvm *kvm, int slot);
798 void kvm_mmu_write_protect_pt_masked(struct kvm *kvm,
799 				     struct kvm_memory_slot *slot,
800 				     gfn_t gfn_offset, unsigned long mask);
801 void kvm_mmu_zap_all(struct kvm *kvm);
802 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm);
803 unsigned int kvm_mmu_calculate_mmu_pages(struct kvm *kvm);
804 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int kvm_nr_mmu_pages);
805 
806 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
807 
808 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
809 			  const void *val, int bytes);
810 u8 kvm_get_guest_memory_type(struct kvm_vcpu *vcpu, gfn_t gfn);
811 
812 extern bool tdp_enabled;
813 
814 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
815 
816 /* control of guest tsc rate supported? */
817 extern bool kvm_has_tsc_control;
818 /* minimum supported tsc_khz for guests */
819 extern u32  kvm_min_guest_tsc_khz;
820 /* maximum supported tsc_khz for guests */
821 extern u32  kvm_max_guest_tsc_khz;
822 
823 enum emulation_result {
824 	EMULATE_DONE,         /* no further processing */
825 	EMULATE_USER_EXIT,    /* kvm_run ready for userspace exit */
826 	EMULATE_FAIL,         /* can't emulate this instruction */
827 };
828 
829 #define EMULTYPE_NO_DECODE	    (1 << 0)
830 #define EMULTYPE_TRAP_UD	    (1 << 1)
831 #define EMULTYPE_SKIP		    (1 << 2)
832 #define EMULTYPE_RETRY		    (1 << 3)
833 #define EMULTYPE_NO_REEXECUTE	    (1 << 4)
834 int x86_emulate_instruction(struct kvm_vcpu *vcpu, unsigned long cr2,
835 			    int emulation_type, void *insn, int insn_len);
836 
837 static inline int emulate_instruction(struct kvm_vcpu *vcpu,
838 			int emulation_type)
839 {
840 	return x86_emulate_instruction(vcpu, 0, emulation_type, NULL, 0);
841 }
842 
843 void kvm_enable_efer_bits(u64);
844 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
845 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 *data);
846 int kvm_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr);
847 
848 struct x86_emulate_ctxt;
849 
850 int kvm_fast_pio_out(struct kvm_vcpu *vcpu, int size, unsigned short port);
851 void kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
852 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
853 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
854 
855 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
856 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
857 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, unsigned int vector);
858 
859 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
860 		    int reason, bool has_error_code, u32 error_code);
861 
862 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
863 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
864 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
865 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
866 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
867 int kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
868 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
869 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
870 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
871 int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr);
872 
873 int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata);
874 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
875 
876 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
877 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
878 bool kvm_rdpmc(struct kvm_vcpu *vcpu);
879 
880 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
881 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
882 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
883 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
884 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
885 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
886 			    gfn_t gfn, void *data, int offset, int len,
887 			    u32 access);
888 void kvm_propagate_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
889 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
890 
891 static inline int __kvm_irq_line_state(unsigned long *irq_state,
892 				       int irq_source_id, int level)
893 {
894 	/* Logical OR for level trig interrupt */
895 	if (level)
896 		__set_bit(irq_source_id, irq_state);
897 	else
898 		__clear_bit(irq_source_id, irq_state);
899 
900 	return !!(*irq_state);
901 }
902 
903 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
904 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
905 
906 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
907 
908 int fx_init(struct kvm_vcpu *vcpu);
909 
910 void kvm_mmu_flush_tlb(struct kvm_vcpu *vcpu);
911 void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
912 		       const u8 *new, int bytes);
913 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
914 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
915 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
916 int kvm_mmu_load(struct kvm_vcpu *vcpu);
917 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
918 void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu);
919 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access);
920 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
921 			      struct x86_exception *exception);
922 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
923 			       struct x86_exception *exception);
924 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
925 			       struct x86_exception *exception);
926 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
927 				struct x86_exception *exception);
928 
929 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
930 
931 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva, u32 error_code,
932 		       void *insn, int insn_len);
933 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
934 void kvm_mmu_new_cr3(struct kvm_vcpu *vcpu);
935 
936 void kvm_enable_tdp(void);
937 void kvm_disable_tdp(void);
938 
939 static inline gpa_t translate_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access)
940 {
941 	return gpa;
942 }
943 
944 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
945 {
946 	struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
947 
948 	return (struct kvm_mmu_page *)page_private(page);
949 }
950 
951 static inline u16 kvm_read_ldt(void)
952 {
953 	u16 ldt;
954 	asm("sldt %0" : "=g"(ldt));
955 	return ldt;
956 }
957 
958 static inline void kvm_load_ldt(u16 sel)
959 {
960 	asm("lldt %0" : : "rm"(sel));
961 }
962 
963 #ifdef CONFIG_X86_64
964 static inline unsigned long read_msr(unsigned long msr)
965 {
966 	u64 value;
967 
968 	rdmsrl(msr, value);
969 	return value;
970 }
971 #endif
972 
973 static inline u32 get_rdx_init_val(void)
974 {
975 	return 0x600; /* P6 family */
976 }
977 
978 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
979 {
980 	kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
981 }
982 
983 #define TSS_IOPB_BASE_OFFSET 0x66
984 #define TSS_BASE_SIZE 0x68
985 #define TSS_IOPB_SIZE (65536 / 8)
986 #define TSS_REDIRECTION_SIZE (256 / 8)
987 #define RMODE_TSS_SIZE							\
988 	(TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
989 
990 enum {
991 	TASK_SWITCH_CALL = 0,
992 	TASK_SWITCH_IRET = 1,
993 	TASK_SWITCH_JMP = 2,
994 	TASK_SWITCH_GATE = 3,
995 };
996 
997 #define HF_GIF_MASK		(1 << 0)
998 #define HF_HIF_MASK		(1 << 1)
999 #define HF_VINTR_MASK		(1 << 2)
1000 #define HF_NMI_MASK		(1 << 3)
1001 #define HF_IRET_MASK		(1 << 4)
1002 #define HF_GUEST_MASK		(1 << 5) /* VCPU is in guest-mode */
1003 
1004 /*
1005  * Hardware virtualization extension instructions may fault if a
1006  * reboot turns off virtualization while processes are running.
1007  * Trap the fault and ignore the instruction if that happens.
1008  */
1009 asmlinkage void kvm_spurious_fault(void);
1010 
1011 #define ____kvm_handle_fault_on_reboot(insn, cleanup_insn)	\
1012 	"666: " insn "\n\t" \
1013 	"668: \n\t"                           \
1014 	".pushsection .fixup, \"ax\" \n" \
1015 	"667: \n\t" \
1016 	cleanup_insn "\n\t"		      \
1017 	"cmpb $0, kvm_rebooting \n\t"	      \
1018 	"jne 668b \n\t"      		      \
1019 	__ASM_SIZE(push) " $666b \n\t"	      \
1020 	"call kvm_spurious_fault \n\t"	      \
1021 	".popsection \n\t" \
1022 	_ASM_EXTABLE(666b, 667b)
1023 
1024 #define __kvm_handle_fault_on_reboot(insn)		\
1025 	____kvm_handle_fault_on_reboot(insn, "")
1026 
1027 #define KVM_ARCH_WANT_MMU_NOTIFIER
1028 int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
1029 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end);
1030 int kvm_age_hva(struct kvm *kvm, unsigned long hva);
1031 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
1032 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
1033 int cpuid_maxphyaddr(struct kvm_vcpu *vcpu);
1034 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1035 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1036 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1037 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1038 void kvm_vcpu_reset(struct kvm_vcpu *vcpu);
1039 
1040 void kvm_define_shared_msr(unsigned index, u32 msr);
1041 void kvm_set_shared_msr(unsigned index, u64 val, u64 mask);
1042 
1043 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1044 
1045 void kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1046 				     struct kvm_async_pf *work);
1047 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1048 				 struct kvm_async_pf *work);
1049 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1050 			       struct kvm_async_pf *work);
1051 bool kvm_arch_can_inject_async_page_present(struct kvm_vcpu *vcpu);
1052 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1053 
1054 void kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1055 
1056 int kvm_is_in_guest(void);
1057 
1058 void kvm_pmu_init(struct kvm_vcpu *vcpu);
1059 void kvm_pmu_destroy(struct kvm_vcpu *vcpu);
1060 void kvm_pmu_reset(struct kvm_vcpu *vcpu);
1061 void kvm_pmu_cpuid_update(struct kvm_vcpu *vcpu);
1062 bool kvm_pmu_msr(struct kvm_vcpu *vcpu, u32 msr);
1063 int kvm_pmu_get_msr(struct kvm_vcpu *vcpu, u32 msr, u64 *data);
1064 int kvm_pmu_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info);
1065 int kvm_pmu_read_pmc(struct kvm_vcpu *vcpu, unsigned pmc, u64 *data);
1066 void kvm_handle_pmu_event(struct kvm_vcpu *vcpu);
1067 void kvm_deliver_pmi(struct kvm_vcpu *vcpu);
1068 
1069 #endif /* _ASM_X86_KVM_HOST_H */
1070