xref: /linux/arch/x86/include/asm/kvm_host.h (revision ff5599816711d2e67da2d7561fd36ac48debd433)
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 #define SELECTOR_TI_MASK (1 << 2)
83 #define SELECTOR_RPL_MASK 0x03
84 
85 #define IOPL_SHIFT 12
86 
87 #define KVM_PERMILLE_MMU_PAGES 20
88 #define KVM_MIN_ALLOC_MMU_PAGES 64
89 #define KVM_MMU_HASH_SHIFT 10
90 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
91 #define KVM_MIN_FREE_MMU_PAGES 5
92 #define KVM_REFILL_PAGES 25
93 #define KVM_MAX_CPUID_ENTRIES 80
94 #define KVM_NR_FIXED_MTRR_REGION 88
95 #define KVM_NR_VAR_MTRR 8
96 
97 #define ASYNC_PF_PER_VCPU 64
98 
99 struct kvm_vcpu;
100 struct kvm;
101 struct kvm_async_pf;
102 
103 enum kvm_reg {
104 	VCPU_REGS_RAX = 0,
105 	VCPU_REGS_RCX = 1,
106 	VCPU_REGS_RDX = 2,
107 	VCPU_REGS_RBX = 3,
108 	VCPU_REGS_RSP = 4,
109 	VCPU_REGS_RBP = 5,
110 	VCPU_REGS_RSI = 6,
111 	VCPU_REGS_RDI = 7,
112 #ifdef CONFIG_X86_64
113 	VCPU_REGS_R8 = 8,
114 	VCPU_REGS_R9 = 9,
115 	VCPU_REGS_R10 = 10,
116 	VCPU_REGS_R11 = 11,
117 	VCPU_REGS_R12 = 12,
118 	VCPU_REGS_R13 = 13,
119 	VCPU_REGS_R14 = 14,
120 	VCPU_REGS_R15 = 15,
121 #endif
122 	VCPU_REGS_RIP,
123 	NR_VCPU_REGS
124 };
125 
126 enum kvm_reg_ex {
127 	VCPU_EXREG_PDPTR = NR_VCPU_REGS,
128 	VCPU_EXREG_CR3,
129 	VCPU_EXREG_RFLAGS,
130 	VCPU_EXREG_CPL,
131 	VCPU_EXREG_SEGMENTS,
132 };
133 
134 enum {
135 	VCPU_SREG_ES,
136 	VCPU_SREG_CS,
137 	VCPU_SREG_SS,
138 	VCPU_SREG_DS,
139 	VCPU_SREG_FS,
140 	VCPU_SREG_GS,
141 	VCPU_SREG_TR,
142 	VCPU_SREG_LDTR,
143 };
144 
145 #include <asm/kvm_emulate.h>
146 
147 #define KVM_NR_MEM_OBJS 40
148 
149 #define KVM_NR_DB_REGS	4
150 
151 #define DR6_BD		(1 << 13)
152 #define DR6_BS		(1 << 14)
153 #define DR6_FIXED_1	0xffff0ff0
154 #define DR6_VOLATILE	0x0000e00f
155 
156 #define DR7_BP_EN_MASK	0x000000ff
157 #define DR7_GE		(1 << 9)
158 #define DR7_GD		(1 << 13)
159 #define DR7_FIXED_1	0x00000400
160 #define DR7_VOLATILE	0xffff23ff
161 
162 /* apic attention bits */
163 #define KVM_APIC_CHECK_VAPIC	0
164 /*
165  * The following bit is set with PV-EOI, unset on EOI.
166  * We detect PV-EOI changes by guest by comparing
167  * this bit with PV-EOI in guest memory.
168  * See the implementation in apic_update_pv_eoi.
169  */
170 #define KVM_APIC_PV_EOI_PENDING	1
171 
172 /*
173  * We don't want allocation failures within the mmu code, so we preallocate
174  * enough memory for a single page fault in a cache.
175  */
176 struct kvm_mmu_memory_cache {
177 	int nobjs;
178 	void *objects[KVM_NR_MEM_OBJS];
179 };
180 
181 /*
182  * kvm_mmu_page_role, below, is defined as:
183  *
184  *   bits 0:3 - total guest paging levels (2-4, or zero for real mode)
185  *   bits 4:7 - page table level for this shadow (1-4)
186  *   bits 8:9 - page table quadrant for 2-level guests
187  *   bit   16 - direct mapping of virtual to physical mapping at gfn
188  *              used for real mode and two-dimensional paging
189  *   bits 17:19 - common access permissions for all ptes in this shadow page
190  */
191 union kvm_mmu_page_role {
192 	unsigned word;
193 	struct {
194 		unsigned level:4;
195 		unsigned cr4_pae:1;
196 		unsigned quadrant:2;
197 		unsigned pad_for_nice_hex_output:6;
198 		unsigned direct:1;
199 		unsigned access:3;
200 		unsigned invalid:1;
201 		unsigned nxe:1;
202 		unsigned cr0_wp:1;
203 		unsigned smep_andnot_wp:1;
204 	};
205 };
206 
207 struct kvm_mmu_page {
208 	struct list_head link;
209 	struct hlist_node hash_link;
210 
211 	/*
212 	 * The following two entries are used to key the shadow page in the
213 	 * hash table.
214 	 */
215 	gfn_t gfn;
216 	union kvm_mmu_page_role role;
217 
218 	u64 *spt;
219 	/* hold the gfn of each spte inside spt */
220 	gfn_t *gfns;
221 	bool unsync;
222 	int root_count;          /* Currently serving as active root */
223 	unsigned int unsync_children;
224 	unsigned long parent_ptes;	/* Reverse mapping for parent_pte */
225 
226 	/* The page is obsolete if mmu_valid_gen != kvm->arch.mmu_valid_gen.  */
227 	unsigned long mmu_valid_gen;
228 
229 	DECLARE_BITMAP(unsync_child_bitmap, 512);
230 
231 #ifdef CONFIG_X86_32
232 	/*
233 	 * Used out of the mmu-lock to avoid reading spte values while an
234 	 * update is in progress; see the comments in __get_spte_lockless().
235 	 */
236 	int clear_spte_count;
237 #endif
238 
239 	/* Number of writes since the last time traversal visited this page.  */
240 	int write_flooding_count;
241 };
242 
243 struct kvm_pio_request {
244 	unsigned long count;
245 	int in;
246 	int port;
247 	int size;
248 };
249 
250 /*
251  * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
252  * 32-bit).  The kvm_mmu structure abstracts the details of the current mmu
253  * mode.
254  */
255 struct kvm_mmu {
256 	void (*new_cr3)(struct kvm_vcpu *vcpu);
257 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long root);
258 	unsigned long (*get_cr3)(struct kvm_vcpu *vcpu);
259 	u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
260 	int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err,
261 			  bool prefault);
262 	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
263 				  struct x86_exception *fault);
264 	void (*free)(struct kvm_vcpu *vcpu);
265 	gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva, u32 access,
266 			    struct x86_exception *exception);
267 	gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access);
268 	int (*sync_page)(struct kvm_vcpu *vcpu,
269 			 struct kvm_mmu_page *sp);
270 	void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva);
271 	void (*update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
272 			   u64 *spte, const void *pte);
273 	hpa_t root_hpa;
274 	int root_level;
275 	int shadow_root_level;
276 	union kvm_mmu_page_role base_role;
277 	bool direct_map;
278 
279 	/*
280 	 * Bitmap; bit set = permission fault
281 	 * Byte index: page fault error code [4:1]
282 	 * Bit index: pte permissions in ACC_* format
283 	 */
284 	u8 permissions[16];
285 
286 	u64 *pae_root;
287 	u64 *lm_root;
288 	u64 rsvd_bits_mask[2][4];
289 
290 	/*
291 	 * Bitmap: bit set = last pte in walk
292 	 * index[0:1]: level (zero-based)
293 	 * index[2]: pte.ps
294 	 */
295 	u8 last_pte_bitmap;
296 
297 	bool nx;
298 
299 	u64 pdptrs[4]; /* pae */
300 };
301 
302 enum pmc_type {
303 	KVM_PMC_GP = 0,
304 	KVM_PMC_FIXED,
305 };
306 
307 struct kvm_pmc {
308 	enum pmc_type type;
309 	u8 idx;
310 	u64 counter;
311 	u64 eventsel;
312 	struct perf_event *perf_event;
313 	struct kvm_vcpu *vcpu;
314 };
315 
316 struct kvm_pmu {
317 	unsigned nr_arch_gp_counters;
318 	unsigned nr_arch_fixed_counters;
319 	unsigned available_event_types;
320 	u64 fixed_ctr_ctrl;
321 	u64 global_ctrl;
322 	u64 global_status;
323 	u64 global_ovf_ctrl;
324 	u64 counter_bitmask[2];
325 	u64 global_ctrl_mask;
326 	u8 version;
327 	struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
328 	struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED];
329 	struct irq_work irq_work;
330 	u64 reprogram_pmi;
331 };
332 
333 struct kvm_vcpu_arch {
334 	/*
335 	 * rip and regs accesses must go through
336 	 * kvm_{register,rip}_{read,write} functions.
337 	 */
338 	unsigned long regs[NR_VCPU_REGS];
339 	u32 regs_avail;
340 	u32 regs_dirty;
341 
342 	unsigned long cr0;
343 	unsigned long cr0_guest_owned_bits;
344 	unsigned long cr2;
345 	unsigned long cr3;
346 	unsigned long cr4;
347 	unsigned long cr4_guest_owned_bits;
348 	unsigned long cr8;
349 	u32 hflags;
350 	u64 efer;
351 	u64 apic_base;
352 	struct kvm_lapic *apic;    /* kernel irqchip context */
353 	unsigned long apic_attention;
354 	int32_t apic_arb_prio;
355 	int mp_state;
356 	u64 ia32_misc_enable_msr;
357 	bool tpr_access_reporting;
358 
359 	/*
360 	 * Paging state of the vcpu
361 	 *
362 	 * If the vcpu runs in guest mode with two level paging this still saves
363 	 * the paging mode of the l1 guest. This context is always used to
364 	 * handle faults.
365 	 */
366 	struct kvm_mmu mmu;
367 
368 	/*
369 	 * Paging state of an L2 guest (used for nested npt)
370 	 *
371 	 * This context will save all necessary information to walk page tables
372 	 * of the an L2 guest. This context is only initialized for page table
373 	 * walking and not for faulting since we never handle l2 page faults on
374 	 * the host.
375 	 */
376 	struct kvm_mmu nested_mmu;
377 
378 	/*
379 	 * Pointer to the mmu context currently used for
380 	 * gva_to_gpa translations.
381 	 */
382 	struct kvm_mmu *walk_mmu;
383 
384 	struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
385 	struct kvm_mmu_memory_cache mmu_page_cache;
386 	struct kvm_mmu_memory_cache mmu_page_header_cache;
387 
388 	struct fpu guest_fpu;
389 	u64 xcr0;
390 
391 	struct kvm_pio_request pio;
392 	void *pio_data;
393 
394 	u8 event_exit_inst_len;
395 
396 	struct kvm_queued_exception {
397 		bool pending;
398 		bool has_error_code;
399 		bool reinject;
400 		u8 nr;
401 		u32 error_code;
402 	} exception;
403 
404 	struct kvm_queued_interrupt {
405 		bool pending;
406 		bool soft;
407 		u8 nr;
408 	} interrupt;
409 
410 	int halt_request; /* real mode on Intel only */
411 
412 	int cpuid_nent;
413 	struct kvm_cpuid_entry2 cpuid_entries[KVM_MAX_CPUID_ENTRIES];
414 	/* emulate context */
415 
416 	struct x86_emulate_ctxt emulate_ctxt;
417 	bool emulate_regs_need_sync_to_vcpu;
418 	bool emulate_regs_need_sync_from_vcpu;
419 	int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
420 
421 	gpa_t time;
422 	struct pvclock_vcpu_time_info hv_clock;
423 	unsigned int hw_tsc_khz;
424 	struct gfn_to_hva_cache pv_time;
425 	bool pv_time_enabled;
426 	/* set guest stopped flag in pvclock flags field */
427 	bool pvclock_set_guest_stopped_request;
428 
429 	struct {
430 		u64 msr_val;
431 		u64 last_steal;
432 		u64 accum_steal;
433 		struct gfn_to_hva_cache stime;
434 		struct kvm_steal_time steal;
435 	} st;
436 
437 	u64 last_guest_tsc;
438 	u64 last_kernel_ns;
439 	u64 last_host_tsc;
440 	u64 tsc_offset_adjustment;
441 	u64 this_tsc_nsec;
442 	u64 this_tsc_write;
443 	u8  this_tsc_generation;
444 	bool tsc_catchup;
445 	bool tsc_always_catchup;
446 	s8 virtual_tsc_shift;
447 	u32 virtual_tsc_mult;
448 	u32 virtual_tsc_khz;
449 	s64 ia32_tsc_adjust_msr;
450 
451 	atomic_t nmi_queued;  /* unprocessed asynchronous NMIs */
452 	unsigned nmi_pending; /* NMI queued after currently running handler */
453 	bool nmi_injected;    /* Trying to inject an NMI this entry */
454 
455 	struct mtrr_state_type mtrr_state;
456 	u32 pat;
457 
458 	int switch_db_regs;
459 	unsigned long db[KVM_NR_DB_REGS];
460 	unsigned long dr6;
461 	unsigned long dr7;
462 	unsigned long eff_db[KVM_NR_DB_REGS];
463 	unsigned long guest_debug_dr7;
464 
465 	u64 mcg_cap;
466 	u64 mcg_status;
467 	u64 mcg_ctl;
468 	u64 *mce_banks;
469 
470 	/* Cache MMIO info */
471 	u64 mmio_gva;
472 	unsigned access;
473 	gfn_t mmio_gfn;
474 
475 	struct kvm_pmu pmu;
476 
477 	/* used for guest single stepping over the given code position */
478 	unsigned long singlestep_rip;
479 
480 	/* fields used by HYPER-V emulation */
481 	u64 hv_vapic;
482 
483 	cpumask_var_t wbinvd_dirty_mask;
484 
485 	unsigned long last_retry_eip;
486 	unsigned long last_retry_addr;
487 
488 	struct {
489 		bool halted;
490 		gfn_t gfns[roundup_pow_of_two(ASYNC_PF_PER_VCPU)];
491 		struct gfn_to_hva_cache data;
492 		u64 msr_val;
493 		u32 id;
494 		bool send_user_only;
495 	} apf;
496 
497 	/* OSVW MSRs (AMD only) */
498 	struct {
499 		u64 length;
500 		u64 status;
501 	} osvw;
502 
503 	struct {
504 		u64 msr_val;
505 		struct gfn_to_hva_cache data;
506 	} pv_eoi;
507 
508 	/*
509 	 * Indicate whether the access faults on its page table in guest
510 	 * which is set when fix page fault and used to detect unhandeable
511 	 * instruction.
512 	 */
513 	bool write_fault_to_shadow_pgtable;
514 };
515 
516 struct kvm_lpage_info {
517 	int write_count;
518 };
519 
520 struct kvm_arch_memory_slot {
521 	unsigned long *rmap[KVM_NR_PAGE_SIZES];
522 	struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
523 };
524 
525 struct kvm_apic_map {
526 	struct rcu_head rcu;
527 	u8 ldr_bits;
528 	/* fields bellow are used to decode ldr values in different modes */
529 	u32 cid_shift, cid_mask, lid_mask;
530 	struct kvm_lapic *phys_map[256];
531 	/* first index is cluster id second is cpu id in a cluster */
532 	struct kvm_lapic *logical_map[16][16];
533 };
534 
535 struct kvm_arch {
536 	unsigned int n_used_mmu_pages;
537 	unsigned int n_requested_mmu_pages;
538 	unsigned int n_max_mmu_pages;
539 	unsigned int indirect_shadow_pages;
540 	unsigned long mmu_valid_gen;
541 	struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
542 	/*
543 	 * Hash table of struct kvm_mmu_page.
544 	 */
545 	struct list_head active_mmu_pages;
546 	struct list_head zapped_obsolete_pages;
547 
548 	struct list_head assigned_dev_head;
549 	struct iommu_domain *iommu_domain;
550 	int iommu_flags;
551 	struct kvm_pic *vpic;
552 	struct kvm_ioapic *vioapic;
553 	struct kvm_pit *vpit;
554 	int vapics_in_nmi_mode;
555 	struct mutex apic_map_lock;
556 	struct kvm_apic_map *apic_map;
557 
558 	unsigned int tss_addr;
559 	struct page *apic_access_page;
560 
561 	gpa_t wall_clock;
562 
563 	struct page *ept_identity_pagetable;
564 	bool ept_identity_pagetable_done;
565 	gpa_t ept_identity_map_addr;
566 
567 	unsigned long irq_sources_bitmap;
568 	s64 kvmclock_offset;
569 	raw_spinlock_t tsc_write_lock;
570 	u64 last_tsc_nsec;
571 	u64 last_tsc_write;
572 	u32 last_tsc_khz;
573 	u64 cur_tsc_nsec;
574 	u64 cur_tsc_write;
575 	u64 cur_tsc_offset;
576 	u8  cur_tsc_generation;
577 	int nr_vcpus_matched_tsc;
578 
579 	spinlock_t pvclock_gtod_sync_lock;
580 	bool use_master_clock;
581 	u64 master_kernel_ns;
582 	cycle_t master_cycle_now;
583 
584 	struct kvm_xen_hvm_config xen_hvm_config;
585 
586 	/* fields used by HYPER-V emulation */
587 	u64 hv_guest_os_id;
588 	u64 hv_hypercall;
589 
590 	#ifdef CONFIG_KVM_MMU_AUDIT
591 	int audit_point;
592 	#endif
593 };
594 
595 struct kvm_vm_stat {
596 	u32 mmu_shadow_zapped;
597 	u32 mmu_pte_write;
598 	u32 mmu_pte_updated;
599 	u32 mmu_pde_zapped;
600 	u32 mmu_flooded;
601 	u32 mmu_recycled;
602 	u32 mmu_cache_miss;
603 	u32 mmu_unsync;
604 	u32 remote_tlb_flush;
605 	u32 lpages;
606 };
607 
608 struct kvm_vcpu_stat {
609 	u32 pf_fixed;
610 	u32 pf_guest;
611 	u32 tlb_flush;
612 	u32 invlpg;
613 
614 	u32 exits;
615 	u32 io_exits;
616 	u32 mmio_exits;
617 	u32 signal_exits;
618 	u32 irq_window_exits;
619 	u32 nmi_window_exits;
620 	u32 halt_exits;
621 	u32 halt_wakeup;
622 	u32 request_irq_exits;
623 	u32 irq_exits;
624 	u32 host_state_reload;
625 	u32 efer_reload;
626 	u32 fpu_reload;
627 	u32 insn_emulation;
628 	u32 insn_emulation_fail;
629 	u32 hypercalls;
630 	u32 irq_injections;
631 	u32 nmi_injections;
632 };
633 
634 struct x86_instruction_info;
635 
636 struct msr_data {
637 	bool host_initiated;
638 	u32 index;
639 	u64 data;
640 };
641 
642 struct kvm_x86_ops {
643 	int (*cpu_has_kvm_support)(void);          /* __init */
644 	int (*disabled_by_bios)(void);             /* __init */
645 	int (*hardware_enable)(void *dummy);
646 	void (*hardware_disable)(void *dummy);
647 	void (*check_processor_compatibility)(void *rtn);
648 	int (*hardware_setup)(void);               /* __init */
649 	void (*hardware_unsetup)(void);            /* __exit */
650 	bool (*cpu_has_accelerated_tpr)(void);
651 	void (*cpuid_update)(struct kvm_vcpu *vcpu);
652 
653 	/* Create, but do not attach this VCPU */
654 	struct kvm_vcpu *(*vcpu_create)(struct kvm *kvm, unsigned id);
655 	void (*vcpu_free)(struct kvm_vcpu *vcpu);
656 	void (*vcpu_reset)(struct kvm_vcpu *vcpu);
657 
658 	void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
659 	void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
660 	void (*vcpu_put)(struct kvm_vcpu *vcpu);
661 
662 	void (*update_db_bp_intercept)(struct kvm_vcpu *vcpu);
663 	int (*get_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata);
664 	int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
665 	u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
666 	void (*get_segment)(struct kvm_vcpu *vcpu,
667 			    struct kvm_segment *var, int seg);
668 	int (*get_cpl)(struct kvm_vcpu *vcpu);
669 	void (*set_segment)(struct kvm_vcpu *vcpu,
670 			    struct kvm_segment *var, int seg);
671 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
672 	void (*decache_cr0_guest_bits)(struct kvm_vcpu *vcpu);
673 	void (*decache_cr3)(struct kvm_vcpu *vcpu);
674 	void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu);
675 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
676 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
677 	int (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
678 	void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
679 	void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
680 	void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
681 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
682 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
683 	void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
684 	void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
685 	unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
686 	void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
687 	void (*fpu_activate)(struct kvm_vcpu *vcpu);
688 	void (*fpu_deactivate)(struct kvm_vcpu *vcpu);
689 
690 	void (*tlb_flush)(struct kvm_vcpu *vcpu);
691 
692 	void (*run)(struct kvm_vcpu *vcpu);
693 	int (*handle_exit)(struct kvm_vcpu *vcpu);
694 	void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
695 	void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
696 	u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
697 	void (*patch_hypercall)(struct kvm_vcpu *vcpu,
698 				unsigned char *hypercall_addr);
699 	void (*set_irq)(struct kvm_vcpu *vcpu);
700 	void (*set_nmi)(struct kvm_vcpu *vcpu);
701 	void (*queue_exception)(struct kvm_vcpu *vcpu, unsigned nr,
702 				bool has_error_code, u32 error_code,
703 				bool reinject);
704 	void (*cancel_injection)(struct kvm_vcpu *vcpu);
705 	int (*interrupt_allowed)(struct kvm_vcpu *vcpu);
706 	int (*nmi_allowed)(struct kvm_vcpu *vcpu);
707 	bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
708 	void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
709 	int (*enable_nmi_window)(struct kvm_vcpu *vcpu);
710 	int (*enable_irq_window)(struct kvm_vcpu *vcpu);
711 	void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
712 	int (*vm_has_apicv)(struct kvm *kvm);
713 	void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
714 	void (*hwapic_isr_update)(struct kvm *kvm, int isr);
715 	void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
716 	void (*set_virtual_x2apic_mode)(struct kvm_vcpu *vcpu, bool set);
717 	void (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
718 	void (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
719 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
720 	int (*get_tdp_level)(void);
721 	u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
722 	int (*get_lpage_level)(void);
723 	bool (*rdtscp_supported)(void);
724 	bool (*invpcid_supported)(void);
725 	void (*adjust_tsc_offset)(struct kvm_vcpu *vcpu, s64 adjustment, bool host);
726 
727 	void (*set_tdp_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
728 
729 	void (*set_supported_cpuid)(u32 func, struct kvm_cpuid_entry2 *entry);
730 
731 	bool (*has_wbinvd_exit)(void);
732 
733 	void (*set_tsc_khz)(struct kvm_vcpu *vcpu, u32 user_tsc_khz, bool scale);
734 	u64 (*read_tsc_offset)(struct kvm_vcpu *vcpu);
735 	void (*write_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
736 
737 	u64 (*compute_tsc_offset)(struct kvm_vcpu *vcpu, u64 target_tsc);
738 	u64 (*read_l1_tsc)(struct kvm_vcpu *vcpu, u64 host_tsc);
739 
740 	void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2);
741 
742 	int (*check_intercept)(struct kvm_vcpu *vcpu,
743 			       struct x86_instruction_info *info,
744 			       enum x86_intercept_stage stage);
745 	void (*handle_external_intr)(struct kvm_vcpu *vcpu);
746 };
747 
748 struct kvm_arch_async_pf {
749 	u32 token;
750 	gfn_t gfn;
751 	unsigned long cr3;
752 	bool direct_map;
753 };
754 
755 extern struct kvm_x86_ops *kvm_x86_ops;
756 
757 static inline void adjust_tsc_offset_guest(struct kvm_vcpu *vcpu,
758 					   s64 adjustment)
759 {
760 	kvm_x86_ops->adjust_tsc_offset(vcpu, adjustment, false);
761 }
762 
763 static inline void adjust_tsc_offset_host(struct kvm_vcpu *vcpu, s64 adjustment)
764 {
765 	kvm_x86_ops->adjust_tsc_offset(vcpu, adjustment, true);
766 }
767 
768 int kvm_mmu_module_init(void);
769 void kvm_mmu_module_exit(void);
770 
771 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
772 int kvm_mmu_create(struct kvm_vcpu *vcpu);
773 int kvm_mmu_setup(struct kvm_vcpu *vcpu);
774 void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
775 		u64 dirty_mask, u64 nx_mask, u64 x_mask);
776 
777 int kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
778 void kvm_mmu_slot_remove_write_access(struct kvm *kvm, int slot);
779 void kvm_mmu_write_protect_pt_masked(struct kvm *kvm,
780 				     struct kvm_memory_slot *slot,
781 				     gfn_t gfn_offset, unsigned long mask);
782 void kvm_mmu_zap_all(struct kvm *kvm);
783 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm);
784 unsigned int kvm_mmu_calculate_mmu_pages(struct kvm *kvm);
785 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int kvm_nr_mmu_pages);
786 
787 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
788 
789 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
790 			  const void *val, int bytes);
791 u8 kvm_get_guest_memory_type(struct kvm_vcpu *vcpu, gfn_t gfn);
792 
793 extern bool tdp_enabled;
794 
795 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
796 
797 /* control of guest tsc rate supported? */
798 extern bool kvm_has_tsc_control;
799 /* minimum supported tsc_khz for guests */
800 extern u32  kvm_min_guest_tsc_khz;
801 /* maximum supported tsc_khz for guests */
802 extern u32  kvm_max_guest_tsc_khz;
803 
804 enum emulation_result {
805 	EMULATE_DONE,       /* no further processing */
806 	EMULATE_DO_MMIO,      /* kvm_run filled with mmio request */
807 	EMULATE_FAIL,         /* can't emulate this instruction */
808 };
809 
810 #define EMULTYPE_NO_DECODE	    (1 << 0)
811 #define EMULTYPE_TRAP_UD	    (1 << 1)
812 #define EMULTYPE_SKIP		    (1 << 2)
813 #define EMULTYPE_RETRY		    (1 << 3)
814 #define EMULTYPE_NO_REEXECUTE	    (1 << 4)
815 int x86_emulate_instruction(struct kvm_vcpu *vcpu, unsigned long cr2,
816 			    int emulation_type, void *insn, int insn_len);
817 
818 static inline int emulate_instruction(struct kvm_vcpu *vcpu,
819 			int emulation_type)
820 {
821 	return x86_emulate_instruction(vcpu, 0, emulation_type, NULL, 0);
822 }
823 
824 void kvm_enable_efer_bits(u64);
825 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
826 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 *data);
827 int kvm_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr);
828 
829 struct x86_emulate_ctxt;
830 
831 int kvm_fast_pio_out(struct kvm_vcpu *vcpu, int size, unsigned short port);
832 void kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
833 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
834 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
835 
836 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
837 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
838 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, unsigned int vector);
839 
840 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
841 		    int reason, bool has_error_code, u32 error_code);
842 
843 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
844 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
845 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
846 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
847 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
848 int kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
849 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
850 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
851 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
852 int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr);
853 
854 int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata);
855 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
856 
857 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
858 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
859 bool kvm_rdpmc(struct kvm_vcpu *vcpu);
860 
861 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
862 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
863 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
864 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
865 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
866 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
867 			    gfn_t gfn, void *data, int offset, int len,
868 			    u32 access);
869 void kvm_propagate_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
870 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
871 
872 static inline int __kvm_irq_line_state(unsigned long *irq_state,
873 				       int irq_source_id, int level)
874 {
875 	/* Logical OR for level trig interrupt */
876 	if (level)
877 		__set_bit(irq_source_id, irq_state);
878 	else
879 		__clear_bit(irq_source_id, irq_state);
880 
881 	return !!(*irq_state);
882 }
883 
884 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
885 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
886 
887 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
888 
889 int fx_init(struct kvm_vcpu *vcpu);
890 
891 void kvm_mmu_flush_tlb(struct kvm_vcpu *vcpu);
892 void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
893 		       const u8 *new, int bytes);
894 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
895 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
896 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
897 int kvm_mmu_load(struct kvm_vcpu *vcpu);
898 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
899 void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu);
900 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access);
901 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
902 			      struct x86_exception *exception);
903 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
904 			       struct x86_exception *exception);
905 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
906 			       struct x86_exception *exception);
907 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
908 				struct x86_exception *exception);
909 
910 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
911 
912 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva, u32 error_code,
913 		       void *insn, int insn_len);
914 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
915 
916 void kvm_enable_tdp(void);
917 void kvm_disable_tdp(void);
918 
919 int complete_pio(struct kvm_vcpu *vcpu);
920 bool kvm_check_iopl(struct kvm_vcpu *vcpu);
921 
922 static inline gpa_t translate_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access)
923 {
924 	return gpa;
925 }
926 
927 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
928 {
929 	struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
930 
931 	return (struct kvm_mmu_page *)page_private(page);
932 }
933 
934 static inline u16 kvm_read_ldt(void)
935 {
936 	u16 ldt;
937 	asm("sldt %0" : "=g"(ldt));
938 	return ldt;
939 }
940 
941 static inline void kvm_load_ldt(u16 sel)
942 {
943 	asm("lldt %0" : : "rm"(sel));
944 }
945 
946 #ifdef CONFIG_X86_64
947 static inline unsigned long read_msr(unsigned long msr)
948 {
949 	u64 value;
950 
951 	rdmsrl(msr, value);
952 	return value;
953 }
954 #endif
955 
956 static inline u32 get_rdx_init_val(void)
957 {
958 	return 0x600; /* P6 family */
959 }
960 
961 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
962 {
963 	kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
964 }
965 
966 #define TSS_IOPB_BASE_OFFSET 0x66
967 #define TSS_BASE_SIZE 0x68
968 #define TSS_IOPB_SIZE (65536 / 8)
969 #define TSS_REDIRECTION_SIZE (256 / 8)
970 #define RMODE_TSS_SIZE							\
971 	(TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
972 
973 enum {
974 	TASK_SWITCH_CALL = 0,
975 	TASK_SWITCH_IRET = 1,
976 	TASK_SWITCH_JMP = 2,
977 	TASK_SWITCH_GATE = 3,
978 };
979 
980 #define HF_GIF_MASK		(1 << 0)
981 #define HF_HIF_MASK		(1 << 1)
982 #define HF_VINTR_MASK		(1 << 2)
983 #define HF_NMI_MASK		(1 << 3)
984 #define HF_IRET_MASK		(1 << 4)
985 #define HF_GUEST_MASK		(1 << 5) /* VCPU is in guest-mode */
986 
987 /*
988  * Hardware virtualization extension instructions may fault if a
989  * reboot turns off virtualization while processes are running.
990  * Trap the fault and ignore the instruction if that happens.
991  */
992 asmlinkage void kvm_spurious_fault(void);
993 
994 #define ____kvm_handle_fault_on_reboot(insn, cleanup_insn)	\
995 	"666: " insn "\n\t" \
996 	"668: \n\t"                           \
997 	".pushsection .fixup, \"ax\" \n" \
998 	"667: \n\t" \
999 	cleanup_insn "\n\t"		      \
1000 	"cmpb $0, kvm_rebooting \n\t"	      \
1001 	"jne 668b \n\t"      		      \
1002 	__ASM_SIZE(push) " $666b \n\t"	      \
1003 	"call kvm_spurious_fault \n\t"	      \
1004 	".popsection \n\t" \
1005 	_ASM_EXTABLE(666b, 667b)
1006 
1007 #define __kvm_handle_fault_on_reboot(insn)		\
1008 	____kvm_handle_fault_on_reboot(insn, "")
1009 
1010 #define KVM_ARCH_WANT_MMU_NOTIFIER
1011 int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
1012 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end);
1013 int kvm_age_hva(struct kvm *kvm, unsigned long hva);
1014 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
1015 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
1016 int cpuid_maxphyaddr(struct kvm_vcpu *vcpu);
1017 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1018 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1019 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1020 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1021 void kvm_vcpu_reset(struct kvm_vcpu *vcpu);
1022 
1023 void kvm_define_shared_msr(unsigned index, u32 msr);
1024 void kvm_set_shared_msr(unsigned index, u64 val, u64 mask);
1025 
1026 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1027 
1028 void kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1029 				     struct kvm_async_pf *work);
1030 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1031 				 struct kvm_async_pf *work);
1032 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1033 			       struct kvm_async_pf *work);
1034 bool kvm_arch_can_inject_async_page_present(struct kvm_vcpu *vcpu);
1035 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1036 
1037 void kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1038 
1039 int kvm_is_in_guest(void);
1040 
1041 void kvm_pmu_init(struct kvm_vcpu *vcpu);
1042 void kvm_pmu_destroy(struct kvm_vcpu *vcpu);
1043 void kvm_pmu_reset(struct kvm_vcpu *vcpu);
1044 void kvm_pmu_cpuid_update(struct kvm_vcpu *vcpu);
1045 bool kvm_pmu_msr(struct kvm_vcpu *vcpu, u32 msr);
1046 int kvm_pmu_get_msr(struct kvm_vcpu *vcpu, u32 msr, u64 *data);
1047 int kvm_pmu_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info);
1048 int kvm_pmu_read_pmc(struct kvm_vcpu *vcpu, unsigned pmc, u64 *data);
1049 void kvm_handle_pmu_event(struct kvm_vcpu *vcpu);
1050 void kvm_deliver_pmi(struct kvm_vcpu *vcpu);
1051 
1052 #endif /* _ASM_X86_KVM_HOST_H */
1053