xref: /linux/arch/x86/include/asm/kvm_host.h (revision 424f0750edd5af866f80f5e65998e0610503cb5c)
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 
20 #include <linux/kvm.h>
21 #include <linux/kvm_para.h>
22 #include <linux/kvm_types.h>
23 
24 #include <asm/pvclock-abi.h>
25 #include <asm/desc.h>
26 #include <asm/mtrr.h>
27 #include <asm/msr-index.h>
28 
29 #define KVM_MAX_VCPUS 254
30 #define KVM_SOFT_MAX_VCPUS 64
31 #define KVM_MEMORY_SLOTS 32
32 /* memory slots that does not exposed to userspace */
33 #define KVM_PRIVATE_MEM_SLOTS 4
34 #define KVM_MMIO_SIZE 16
35 
36 #define KVM_PIO_PAGE_OFFSET 1
37 #define KVM_COALESCED_MMIO_PAGE_OFFSET 2
38 
39 #define CR0_RESERVED_BITS                                               \
40 	(~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \
41 			  | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \
42 			  | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG))
43 
44 #define CR3_PAE_RESERVED_BITS ((X86_CR3_PWT | X86_CR3_PCD) - 1)
45 #define CR3_NONPAE_RESERVED_BITS ((PAGE_SIZE-1) & ~(X86_CR3_PWT | X86_CR3_PCD))
46 #define CR3_L_MODE_RESERVED_BITS (CR3_NONPAE_RESERVED_BITS |	\
47 				  0xFFFFFF0000000000ULL)
48 #define CR4_RESERVED_BITS                                               \
49 	(~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\
50 			  | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE     \
51 			  | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR  \
52 			  | X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_RDWRGSFS \
53 			  | X86_CR4_OSXMMEXCPT | X86_CR4_VMXE))
54 
55 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR)
56 
57 
58 
59 #define INVALID_PAGE (~(hpa_t)0)
60 #define VALID_PAGE(x) ((x) != INVALID_PAGE)
61 
62 #define UNMAPPED_GVA (~(gpa_t)0)
63 
64 /* KVM Hugepage definitions for x86 */
65 #define KVM_NR_PAGE_SIZES	3
66 #define KVM_HPAGE_GFN_SHIFT(x)	(((x) - 1) * 9)
67 #define KVM_HPAGE_SHIFT(x)	(PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
68 #define KVM_HPAGE_SIZE(x)	(1UL << KVM_HPAGE_SHIFT(x))
69 #define KVM_HPAGE_MASK(x)	(~(KVM_HPAGE_SIZE(x) - 1))
70 #define KVM_PAGES_PER_HPAGE(x)	(KVM_HPAGE_SIZE(x) / PAGE_SIZE)
71 
72 #define DE_VECTOR 0
73 #define DB_VECTOR 1
74 #define BP_VECTOR 3
75 #define OF_VECTOR 4
76 #define BR_VECTOR 5
77 #define UD_VECTOR 6
78 #define NM_VECTOR 7
79 #define DF_VECTOR 8
80 #define TS_VECTOR 10
81 #define NP_VECTOR 11
82 #define SS_VECTOR 12
83 #define GP_VECTOR 13
84 #define PF_VECTOR 14
85 #define MF_VECTOR 16
86 #define MC_VECTOR 18
87 
88 #define SELECTOR_TI_MASK (1 << 2)
89 #define SELECTOR_RPL_MASK 0x03
90 
91 #define IOPL_SHIFT 12
92 
93 #define KVM_PERMILLE_MMU_PAGES 20
94 #define KVM_MIN_ALLOC_MMU_PAGES 64
95 #define KVM_MMU_HASH_SHIFT 10
96 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
97 #define KVM_MIN_FREE_MMU_PAGES 5
98 #define KVM_REFILL_PAGES 25
99 #define KVM_MAX_CPUID_ENTRIES 80
100 #define KVM_NR_FIXED_MTRR_REGION 88
101 #define KVM_NR_VAR_MTRR 8
102 
103 #define ASYNC_PF_PER_VCPU 64
104 
105 extern raw_spinlock_t kvm_lock;
106 extern struct list_head vm_list;
107 
108 struct kvm_vcpu;
109 struct kvm;
110 struct kvm_async_pf;
111 
112 enum kvm_reg {
113 	VCPU_REGS_RAX = 0,
114 	VCPU_REGS_RCX = 1,
115 	VCPU_REGS_RDX = 2,
116 	VCPU_REGS_RBX = 3,
117 	VCPU_REGS_RSP = 4,
118 	VCPU_REGS_RBP = 5,
119 	VCPU_REGS_RSI = 6,
120 	VCPU_REGS_RDI = 7,
121 #ifdef CONFIG_X86_64
122 	VCPU_REGS_R8 = 8,
123 	VCPU_REGS_R9 = 9,
124 	VCPU_REGS_R10 = 10,
125 	VCPU_REGS_R11 = 11,
126 	VCPU_REGS_R12 = 12,
127 	VCPU_REGS_R13 = 13,
128 	VCPU_REGS_R14 = 14,
129 	VCPU_REGS_R15 = 15,
130 #endif
131 	VCPU_REGS_RIP,
132 	NR_VCPU_REGS
133 };
134 
135 enum kvm_reg_ex {
136 	VCPU_EXREG_PDPTR = NR_VCPU_REGS,
137 	VCPU_EXREG_CR3,
138 	VCPU_EXREG_RFLAGS,
139 	VCPU_EXREG_CPL,
140 	VCPU_EXREG_SEGMENTS,
141 };
142 
143 enum {
144 	VCPU_SREG_ES,
145 	VCPU_SREG_CS,
146 	VCPU_SREG_SS,
147 	VCPU_SREG_DS,
148 	VCPU_SREG_FS,
149 	VCPU_SREG_GS,
150 	VCPU_SREG_TR,
151 	VCPU_SREG_LDTR,
152 };
153 
154 #include <asm/kvm_emulate.h>
155 
156 #define KVM_NR_MEM_OBJS 40
157 
158 #define KVM_NR_DB_REGS	4
159 
160 #define DR6_BD		(1 << 13)
161 #define DR6_BS		(1 << 14)
162 #define DR6_FIXED_1	0xffff0ff0
163 #define DR6_VOLATILE	0x0000e00f
164 
165 #define DR7_BP_EN_MASK	0x000000ff
166 #define DR7_GE		(1 << 9)
167 #define DR7_GD		(1 << 13)
168 #define DR7_FIXED_1	0x00000400
169 #define DR7_VOLATILE	0xffff23ff
170 
171 /*
172  * We don't want allocation failures within the mmu code, so we preallocate
173  * enough memory for a single page fault in a cache.
174  */
175 struct kvm_mmu_memory_cache {
176 	int nobjs;
177 	void *objects[KVM_NR_MEM_OBJS];
178 };
179 
180 #define NR_PTE_CHAIN_ENTRIES 5
181 
182 struct kvm_pte_chain {
183 	u64 *parent_ptes[NR_PTE_CHAIN_ENTRIES];
184 	struct hlist_node link;
185 };
186 
187 /*
188  * kvm_mmu_page_role, below, is defined as:
189  *
190  *   bits 0:3 - total guest paging levels (2-4, or zero for real mode)
191  *   bits 4:7 - page table level for this shadow (1-4)
192  *   bits 8:9 - page table quadrant for 2-level guests
193  *   bit   16 - direct mapping of virtual to physical mapping at gfn
194  *              used for real mode and two-dimensional paging
195  *   bits 17:19 - common access permissions for all ptes in this shadow page
196  */
197 union kvm_mmu_page_role {
198 	unsigned word;
199 	struct {
200 		unsigned level:4;
201 		unsigned cr4_pae:1;
202 		unsigned quadrant:2;
203 		unsigned pad_for_nice_hex_output:6;
204 		unsigned direct:1;
205 		unsigned access:3;
206 		unsigned invalid:1;
207 		unsigned nxe:1;
208 		unsigned cr0_wp:1;
209 		unsigned smep_andnot_wp:1;
210 	};
211 };
212 
213 struct kvm_mmu_page {
214 	struct list_head link;
215 	struct hlist_node hash_link;
216 
217 	/*
218 	 * The following two entries are used to key the shadow page in the
219 	 * hash table.
220 	 */
221 	gfn_t gfn;
222 	union kvm_mmu_page_role role;
223 
224 	u64 *spt;
225 	/* hold the gfn of each spte inside spt */
226 	gfn_t *gfns;
227 	/*
228 	 * One bit set per slot which has memory
229 	 * in this shadow page.
230 	 */
231 	DECLARE_BITMAP(slot_bitmap, KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS);
232 	bool unsync;
233 	int root_count;          /* Currently serving as active root */
234 	unsigned int unsync_children;
235 	unsigned long parent_ptes;	/* Reverse mapping for parent_pte */
236 	DECLARE_BITMAP(unsync_child_bitmap, 512);
237 
238 #ifdef CONFIG_X86_32
239 	int clear_spte_count;
240 #endif
241 
242 	struct rcu_head rcu;
243 };
244 
245 struct kvm_pv_mmu_op_buffer {
246 	void *ptr;
247 	unsigned len;
248 	unsigned processed;
249 	char buf[512] __aligned(sizeof(long));
250 };
251 
252 struct kvm_pio_request {
253 	unsigned long count;
254 	int in;
255 	int port;
256 	int size;
257 };
258 
259 /*
260  * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
261  * 32-bit).  The kvm_mmu structure abstracts the details of the current mmu
262  * mode.
263  */
264 struct kvm_mmu {
265 	void (*new_cr3)(struct kvm_vcpu *vcpu);
266 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long root);
267 	unsigned long (*get_cr3)(struct kvm_vcpu *vcpu);
268 	u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
269 	int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err,
270 			  bool prefault);
271 	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
272 				  struct x86_exception *fault);
273 	void (*free)(struct kvm_vcpu *vcpu);
274 	gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva, u32 access,
275 			    struct x86_exception *exception);
276 	gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access);
277 	int (*sync_page)(struct kvm_vcpu *vcpu,
278 			 struct kvm_mmu_page *sp);
279 	void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva);
280 	void (*update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
281 			   u64 *spte, const void *pte);
282 	hpa_t root_hpa;
283 	int root_level;
284 	int shadow_root_level;
285 	union kvm_mmu_page_role base_role;
286 	bool direct_map;
287 
288 	u64 *pae_root;
289 	u64 *lm_root;
290 	u64 rsvd_bits_mask[2][4];
291 
292 	bool nx;
293 
294 	u64 pdptrs[4]; /* pae */
295 };
296 
297 struct kvm_vcpu_arch {
298 	/*
299 	 * rip and regs accesses must go through
300 	 * kvm_{register,rip}_{read,write} functions.
301 	 */
302 	unsigned long regs[NR_VCPU_REGS];
303 	u32 regs_avail;
304 	u32 regs_dirty;
305 
306 	unsigned long cr0;
307 	unsigned long cr0_guest_owned_bits;
308 	unsigned long cr2;
309 	unsigned long cr3;
310 	unsigned long cr4;
311 	unsigned long cr4_guest_owned_bits;
312 	unsigned long cr8;
313 	u32 hflags;
314 	u64 efer;
315 	u64 apic_base;
316 	struct kvm_lapic *apic;    /* kernel irqchip context */
317 	int32_t apic_arb_prio;
318 	int mp_state;
319 	int sipi_vector;
320 	u64 ia32_misc_enable_msr;
321 	bool tpr_access_reporting;
322 
323 	/*
324 	 * Paging state of the vcpu
325 	 *
326 	 * If the vcpu runs in guest mode with two level paging this still saves
327 	 * the paging mode of the l1 guest. This context is always used to
328 	 * handle faults.
329 	 */
330 	struct kvm_mmu mmu;
331 
332 	/*
333 	 * Paging state of an L2 guest (used for nested npt)
334 	 *
335 	 * This context will save all necessary information to walk page tables
336 	 * of the an L2 guest. This context is only initialized for page table
337 	 * walking and not for faulting since we never handle l2 page faults on
338 	 * the host.
339 	 */
340 	struct kvm_mmu nested_mmu;
341 
342 	/*
343 	 * Pointer to the mmu context currently used for
344 	 * gva_to_gpa translations.
345 	 */
346 	struct kvm_mmu *walk_mmu;
347 
348 	/* only needed in kvm_pv_mmu_op() path, but it's hot so
349 	 * put it here to avoid allocation */
350 	struct kvm_pv_mmu_op_buffer mmu_op_buffer;
351 
352 	struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
353 	struct kvm_mmu_memory_cache mmu_page_cache;
354 	struct kvm_mmu_memory_cache mmu_page_header_cache;
355 
356 	gfn_t last_pt_write_gfn;
357 	int   last_pt_write_count;
358 	u64  *last_pte_updated;
359 	gfn_t last_pte_gfn;
360 
361 	struct fpu guest_fpu;
362 	u64 xcr0;
363 
364 	struct kvm_pio_request pio;
365 	void *pio_data;
366 
367 	u8 event_exit_inst_len;
368 
369 	struct kvm_queued_exception {
370 		bool pending;
371 		bool has_error_code;
372 		bool reinject;
373 		u8 nr;
374 		u32 error_code;
375 	} exception;
376 
377 	struct kvm_queued_interrupt {
378 		bool pending;
379 		bool soft;
380 		u8 nr;
381 	} interrupt;
382 
383 	int halt_request; /* real mode on Intel only */
384 
385 	int cpuid_nent;
386 	struct kvm_cpuid_entry2 cpuid_entries[KVM_MAX_CPUID_ENTRIES];
387 	/* emulate context */
388 
389 	struct x86_emulate_ctxt emulate_ctxt;
390 	bool emulate_regs_need_sync_to_vcpu;
391 	bool emulate_regs_need_sync_from_vcpu;
392 
393 	gpa_t time;
394 	struct pvclock_vcpu_time_info hv_clock;
395 	unsigned int hw_tsc_khz;
396 	unsigned int time_offset;
397 	struct page *time_page;
398 
399 	struct {
400 		u64 msr_val;
401 		u64 last_steal;
402 		u64 accum_steal;
403 		struct gfn_to_hva_cache stime;
404 		struct kvm_steal_time steal;
405 	} st;
406 
407 	u64 last_guest_tsc;
408 	u64 last_kernel_ns;
409 	u64 last_tsc_nsec;
410 	u64 last_tsc_write;
411 	u32 virtual_tsc_khz;
412 	bool tsc_catchup;
413 	u32  tsc_catchup_mult;
414 	s8   tsc_catchup_shift;
415 
416 	atomic_t nmi_queued;  /* unprocessed asynchronous NMIs */
417 	unsigned nmi_pending; /* NMI queued after currently running handler */
418 	bool nmi_injected;    /* Trying to inject an NMI this entry */
419 
420 	struct mtrr_state_type mtrr_state;
421 	u32 pat;
422 
423 	int switch_db_regs;
424 	unsigned long db[KVM_NR_DB_REGS];
425 	unsigned long dr6;
426 	unsigned long dr7;
427 	unsigned long eff_db[KVM_NR_DB_REGS];
428 
429 	u64 mcg_cap;
430 	u64 mcg_status;
431 	u64 mcg_ctl;
432 	u64 *mce_banks;
433 
434 	/* Cache MMIO info */
435 	u64 mmio_gva;
436 	unsigned access;
437 	gfn_t mmio_gfn;
438 
439 	/* used for guest single stepping over the given code position */
440 	unsigned long singlestep_rip;
441 
442 	/* fields used by HYPER-V emulation */
443 	u64 hv_vapic;
444 
445 	cpumask_var_t wbinvd_dirty_mask;
446 
447 	struct {
448 		bool halted;
449 		gfn_t gfns[roundup_pow_of_two(ASYNC_PF_PER_VCPU)];
450 		struct gfn_to_hva_cache data;
451 		u64 msr_val;
452 		u32 id;
453 		bool send_user_only;
454 	} apf;
455 };
456 
457 struct kvm_arch {
458 	unsigned int n_used_mmu_pages;
459 	unsigned int n_requested_mmu_pages;
460 	unsigned int n_max_mmu_pages;
461 	unsigned int indirect_shadow_pages;
462 	atomic_t invlpg_counter;
463 	struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
464 	/*
465 	 * Hash table of struct kvm_mmu_page.
466 	 */
467 	struct list_head active_mmu_pages;
468 	struct list_head assigned_dev_head;
469 	struct iommu_domain *iommu_domain;
470 	int iommu_flags;
471 	struct kvm_pic *vpic;
472 	struct kvm_ioapic *vioapic;
473 	struct kvm_pit *vpit;
474 	int vapics_in_nmi_mode;
475 
476 	unsigned int tss_addr;
477 	struct page *apic_access_page;
478 
479 	gpa_t wall_clock;
480 
481 	struct page *ept_identity_pagetable;
482 	bool ept_identity_pagetable_done;
483 	gpa_t ept_identity_map_addr;
484 
485 	unsigned long irq_sources_bitmap;
486 	s64 kvmclock_offset;
487 	raw_spinlock_t tsc_write_lock;
488 	u64 last_tsc_nsec;
489 	u64 last_tsc_offset;
490 	u64 last_tsc_write;
491 
492 	struct kvm_xen_hvm_config xen_hvm_config;
493 
494 	/* fields used by HYPER-V emulation */
495 	u64 hv_guest_os_id;
496 	u64 hv_hypercall;
497 
498 	atomic_t reader_counter;
499 
500 	#ifdef CONFIG_KVM_MMU_AUDIT
501 	int audit_point;
502 	#endif
503 };
504 
505 struct kvm_vm_stat {
506 	u32 mmu_shadow_zapped;
507 	u32 mmu_pte_write;
508 	u32 mmu_pte_updated;
509 	u32 mmu_pde_zapped;
510 	u32 mmu_flooded;
511 	u32 mmu_recycled;
512 	u32 mmu_cache_miss;
513 	u32 mmu_unsync;
514 	u32 remote_tlb_flush;
515 	u32 lpages;
516 };
517 
518 struct kvm_vcpu_stat {
519 	u32 pf_fixed;
520 	u32 pf_guest;
521 	u32 tlb_flush;
522 	u32 invlpg;
523 
524 	u32 exits;
525 	u32 io_exits;
526 	u32 mmio_exits;
527 	u32 signal_exits;
528 	u32 irq_window_exits;
529 	u32 nmi_window_exits;
530 	u32 halt_exits;
531 	u32 halt_wakeup;
532 	u32 request_irq_exits;
533 	u32 irq_exits;
534 	u32 host_state_reload;
535 	u32 efer_reload;
536 	u32 fpu_reload;
537 	u32 insn_emulation;
538 	u32 insn_emulation_fail;
539 	u32 hypercalls;
540 	u32 irq_injections;
541 	u32 nmi_injections;
542 };
543 
544 struct x86_instruction_info;
545 
546 struct kvm_x86_ops {
547 	int (*cpu_has_kvm_support)(void);          /* __init */
548 	int (*disabled_by_bios)(void);             /* __init */
549 	int (*hardware_enable)(void *dummy);
550 	void (*hardware_disable)(void *dummy);
551 	void (*check_processor_compatibility)(void *rtn);
552 	int (*hardware_setup)(void);               /* __init */
553 	void (*hardware_unsetup)(void);            /* __exit */
554 	bool (*cpu_has_accelerated_tpr)(void);
555 	void (*cpuid_update)(struct kvm_vcpu *vcpu);
556 
557 	/* Create, but do not attach this VCPU */
558 	struct kvm_vcpu *(*vcpu_create)(struct kvm *kvm, unsigned id);
559 	void (*vcpu_free)(struct kvm_vcpu *vcpu);
560 	int (*vcpu_reset)(struct kvm_vcpu *vcpu);
561 
562 	void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
563 	void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
564 	void (*vcpu_put)(struct kvm_vcpu *vcpu);
565 
566 	void (*set_guest_debug)(struct kvm_vcpu *vcpu,
567 				struct kvm_guest_debug *dbg);
568 	int (*get_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata);
569 	int (*set_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
570 	u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
571 	void (*get_segment)(struct kvm_vcpu *vcpu,
572 			    struct kvm_segment *var, int seg);
573 	int (*get_cpl)(struct kvm_vcpu *vcpu);
574 	void (*set_segment)(struct kvm_vcpu *vcpu,
575 			    struct kvm_segment *var, int seg);
576 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
577 	void (*decache_cr0_guest_bits)(struct kvm_vcpu *vcpu);
578 	void (*decache_cr3)(struct kvm_vcpu *vcpu);
579 	void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu);
580 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
581 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
582 	int (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
583 	void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
584 	void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
585 	void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
586 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
587 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
588 	void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
589 	void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
590 	unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
591 	void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
592 	void (*fpu_activate)(struct kvm_vcpu *vcpu);
593 	void (*fpu_deactivate)(struct kvm_vcpu *vcpu);
594 
595 	void (*tlb_flush)(struct kvm_vcpu *vcpu);
596 
597 	void (*run)(struct kvm_vcpu *vcpu);
598 	int (*handle_exit)(struct kvm_vcpu *vcpu);
599 	void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
600 	void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
601 	u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
602 	void (*patch_hypercall)(struct kvm_vcpu *vcpu,
603 				unsigned char *hypercall_addr);
604 	void (*set_irq)(struct kvm_vcpu *vcpu);
605 	void (*set_nmi)(struct kvm_vcpu *vcpu);
606 	void (*queue_exception)(struct kvm_vcpu *vcpu, unsigned nr,
607 				bool has_error_code, u32 error_code,
608 				bool reinject);
609 	void (*cancel_injection)(struct kvm_vcpu *vcpu);
610 	int (*interrupt_allowed)(struct kvm_vcpu *vcpu);
611 	int (*nmi_allowed)(struct kvm_vcpu *vcpu);
612 	bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
613 	void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
614 	void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
615 	void (*enable_irq_window)(struct kvm_vcpu *vcpu);
616 	void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
617 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
618 	int (*get_tdp_level)(void);
619 	u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
620 	int (*get_lpage_level)(void);
621 	bool (*rdtscp_supported)(void);
622 	void (*adjust_tsc_offset)(struct kvm_vcpu *vcpu, s64 adjustment);
623 
624 	void (*set_tdp_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
625 
626 	void (*set_supported_cpuid)(u32 func, struct kvm_cpuid_entry2 *entry);
627 
628 	bool (*has_wbinvd_exit)(void);
629 
630 	void (*set_tsc_khz)(struct kvm_vcpu *vcpu, u32 user_tsc_khz);
631 	void (*write_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
632 
633 	u64 (*compute_tsc_offset)(struct kvm_vcpu *vcpu, u64 target_tsc);
634 	u64 (*read_l1_tsc)(struct kvm_vcpu *vcpu);
635 
636 	void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2);
637 
638 	int (*check_intercept)(struct kvm_vcpu *vcpu,
639 			       struct x86_instruction_info *info,
640 			       enum x86_intercept_stage stage);
641 };
642 
643 struct kvm_arch_async_pf {
644 	u32 token;
645 	gfn_t gfn;
646 	unsigned long cr3;
647 	bool direct_map;
648 };
649 
650 extern struct kvm_x86_ops *kvm_x86_ops;
651 
652 int kvm_mmu_module_init(void);
653 void kvm_mmu_module_exit(void);
654 
655 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
656 int kvm_mmu_create(struct kvm_vcpu *vcpu);
657 int kvm_mmu_setup(struct kvm_vcpu *vcpu);
658 void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
659 		u64 dirty_mask, u64 nx_mask, u64 x_mask);
660 
661 int kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
662 void kvm_mmu_slot_remove_write_access(struct kvm *kvm, int slot);
663 void kvm_mmu_zap_all(struct kvm *kvm);
664 unsigned int kvm_mmu_calculate_mmu_pages(struct kvm *kvm);
665 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int kvm_nr_mmu_pages);
666 
667 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
668 
669 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
670 			  const void *val, int bytes);
671 int kvm_pv_mmu_op(struct kvm_vcpu *vcpu, unsigned long bytes,
672 		  gpa_t addr, unsigned long *ret);
673 u8 kvm_get_guest_memory_type(struct kvm_vcpu *vcpu, gfn_t gfn);
674 
675 extern bool tdp_enabled;
676 
677 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
678 
679 /* control of guest tsc rate supported? */
680 extern bool kvm_has_tsc_control;
681 /* minimum supported tsc_khz for guests */
682 extern u32  kvm_min_guest_tsc_khz;
683 /* maximum supported tsc_khz for guests */
684 extern u32  kvm_max_guest_tsc_khz;
685 
686 enum emulation_result {
687 	EMULATE_DONE,       /* no further processing */
688 	EMULATE_DO_MMIO,      /* kvm_run filled with mmio request */
689 	EMULATE_FAIL,         /* can't emulate this instruction */
690 };
691 
692 #define EMULTYPE_NO_DECODE	    (1 << 0)
693 #define EMULTYPE_TRAP_UD	    (1 << 1)
694 #define EMULTYPE_SKIP		    (1 << 2)
695 int x86_emulate_instruction(struct kvm_vcpu *vcpu, unsigned long cr2,
696 			    int emulation_type, void *insn, int insn_len);
697 
698 static inline int emulate_instruction(struct kvm_vcpu *vcpu,
699 			int emulation_type)
700 {
701 	return x86_emulate_instruction(vcpu, 0, emulation_type, NULL, 0);
702 }
703 
704 void kvm_enable_efer_bits(u64);
705 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 *data);
706 int kvm_set_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
707 
708 struct x86_emulate_ctxt;
709 
710 int kvm_fast_pio_out(struct kvm_vcpu *vcpu, int size, unsigned short port);
711 void kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
712 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
713 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
714 
715 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
716 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
717 
718 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int reason,
719 		    bool has_error_code, u32 error_code);
720 
721 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
722 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
723 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
724 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
725 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
726 int kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
727 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
728 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
729 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
730 int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr);
731 
732 int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata);
733 int kvm_set_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 data);
734 
735 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
736 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
737 
738 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
739 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
740 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
741 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
742 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
743 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
744 			    gfn_t gfn, void *data, int offset, int len,
745 			    u32 access);
746 void kvm_propagate_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
747 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
748 
749 int kvm_pic_set_irq(void *opaque, int irq, int level);
750 
751 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
752 
753 int fx_init(struct kvm_vcpu *vcpu);
754 
755 void kvm_mmu_flush_tlb(struct kvm_vcpu *vcpu);
756 void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
757 		       const u8 *new, int bytes,
758 		       bool guest_initiated);
759 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
760 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
761 int kvm_mmu_load(struct kvm_vcpu *vcpu);
762 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
763 void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu);
764 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
765 			      struct x86_exception *exception);
766 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
767 			       struct x86_exception *exception);
768 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
769 			       struct x86_exception *exception);
770 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
771 				struct x86_exception *exception);
772 
773 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
774 
775 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva, u32 error_code,
776 		       void *insn, int insn_len);
777 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
778 
779 void kvm_enable_tdp(void);
780 void kvm_disable_tdp(void);
781 
782 int complete_pio(struct kvm_vcpu *vcpu);
783 bool kvm_check_iopl(struct kvm_vcpu *vcpu);
784 
785 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
786 {
787 	struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
788 
789 	return (struct kvm_mmu_page *)page_private(page);
790 }
791 
792 static inline u16 kvm_read_ldt(void)
793 {
794 	u16 ldt;
795 	asm("sldt %0" : "=g"(ldt));
796 	return ldt;
797 }
798 
799 static inline void kvm_load_ldt(u16 sel)
800 {
801 	asm("lldt %0" : : "rm"(sel));
802 }
803 
804 #ifdef CONFIG_X86_64
805 static inline unsigned long read_msr(unsigned long msr)
806 {
807 	u64 value;
808 
809 	rdmsrl(msr, value);
810 	return value;
811 }
812 #endif
813 
814 static inline u32 get_rdx_init_val(void)
815 {
816 	return 0x600; /* P6 family */
817 }
818 
819 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
820 {
821 	kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
822 }
823 
824 #define TSS_IOPB_BASE_OFFSET 0x66
825 #define TSS_BASE_SIZE 0x68
826 #define TSS_IOPB_SIZE (65536 / 8)
827 #define TSS_REDIRECTION_SIZE (256 / 8)
828 #define RMODE_TSS_SIZE							\
829 	(TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
830 
831 enum {
832 	TASK_SWITCH_CALL = 0,
833 	TASK_SWITCH_IRET = 1,
834 	TASK_SWITCH_JMP = 2,
835 	TASK_SWITCH_GATE = 3,
836 };
837 
838 #define HF_GIF_MASK		(1 << 0)
839 #define HF_HIF_MASK		(1 << 1)
840 #define HF_VINTR_MASK		(1 << 2)
841 #define HF_NMI_MASK		(1 << 3)
842 #define HF_IRET_MASK		(1 << 4)
843 #define HF_GUEST_MASK		(1 << 5) /* VCPU is in guest-mode */
844 
845 /*
846  * Hardware virtualization extension instructions may fault if a
847  * reboot turns off virtualization while processes are running.
848  * Trap the fault and ignore the instruction if that happens.
849  */
850 asmlinkage void kvm_spurious_fault(void);
851 extern bool kvm_rebooting;
852 
853 #define ____kvm_handle_fault_on_reboot(insn, cleanup_insn)	\
854 	"666: " insn "\n\t" \
855 	"668: \n\t"                           \
856 	".pushsection .fixup, \"ax\" \n" \
857 	"667: \n\t" \
858 	cleanup_insn "\n\t"		      \
859 	"cmpb $0, kvm_rebooting \n\t"	      \
860 	"jne 668b \n\t"      		      \
861 	__ASM_SIZE(push) " $666b \n\t"	      \
862 	"call kvm_spurious_fault \n\t"	      \
863 	".popsection \n\t" \
864 	".pushsection __ex_table, \"a\" \n\t" \
865 	_ASM_PTR " 666b, 667b \n\t" \
866 	".popsection"
867 
868 #define __kvm_handle_fault_on_reboot(insn)		\
869 	____kvm_handle_fault_on_reboot(insn, "")
870 
871 #define KVM_ARCH_WANT_MMU_NOTIFIER
872 int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
873 int kvm_age_hva(struct kvm *kvm, unsigned long hva);
874 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
875 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
876 int cpuid_maxphyaddr(struct kvm_vcpu *vcpu);
877 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
878 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
879 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
880 
881 void kvm_define_shared_msr(unsigned index, u32 msr);
882 void kvm_set_shared_msr(unsigned index, u64 val, u64 mask);
883 
884 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
885 
886 void kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
887 				     struct kvm_async_pf *work);
888 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
889 				 struct kvm_async_pf *work);
890 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
891 			       struct kvm_async_pf *work);
892 bool kvm_arch_can_inject_async_page_present(struct kvm_vcpu *vcpu);
893 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
894 
895 void kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
896 
897 #endif /* _ASM_X86_KVM_HOST_H */
898