xref: /linux/arch/x86/include/asm/kvm_host.h (revision b43ab901d671e3e3cad425ea5e9a3c74e266dcdd)
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 
26 #include <asm/pvclock-abi.h>
27 #include <asm/desc.h>
28 #include <asm/mtrr.h>
29 #include <asm/msr-index.h>
30 
31 #define KVM_MAX_VCPUS 254
32 #define KVM_SOFT_MAX_VCPUS 64
33 #define KVM_MEMORY_SLOTS 32
34 /* memory slots that does not exposed to userspace */
35 #define KVM_PRIVATE_MEM_SLOTS 4
36 #define KVM_MEM_SLOTS_NUM (KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS)
37 
38 #define KVM_MMIO_SIZE 16
39 
40 #define KVM_PIO_PAGE_OFFSET 1
41 #define KVM_COALESCED_MMIO_PAGE_OFFSET 2
42 
43 #define CR0_RESERVED_BITS                                               \
44 	(~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \
45 			  | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \
46 			  | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG))
47 
48 #define CR3_PAE_RESERVED_BITS ((X86_CR3_PWT | X86_CR3_PCD) - 1)
49 #define CR3_NONPAE_RESERVED_BITS ((PAGE_SIZE-1) & ~(X86_CR3_PWT | X86_CR3_PCD))
50 #define CR3_L_MODE_RESERVED_BITS (CR3_NONPAE_RESERVED_BITS |	\
51 				  0xFFFFFF0000000000ULL)
52 #define CR4_RESERVED_BITS                                               \
53 	(~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\
54 			  | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE     \
55 			  | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR  \
56 			  | X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_RDWRGSFS \
57 			  | X86_CR4_OSXMMEXCPT | X86_CR4_VMXE))
58 
59 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR)
60 
61 
62 
63 #define INVALID_PAGE (~(hpa_t)0)
64 #define VALID_PAGE(x) ((x) != INVALID_PAGE)
65 
66 #define UNMAPPED_GVA (~(gpa_t)0)
67 
68 /* KVM Hugepage definitions for x86 */
69 #define KVM_NR_PAGE_SIZES	3
70 #define KVM_HPAGE_GFN_SHIFT(x)	(((x) - 1) * 9)
71 #define KVM_HPAGE_SHIFT(x)	(PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
72 #define KVM_HPAGE_SIZE(x)	(1UL << KVM_HPAGE_SHIFT(x))
73 #define KVM_HPAGE_MASK(x)	(~(KVM_HPAGE_SIZE(x) - 1))
74 #define KVM_PAGES_PER_HPAGE(x)	(KVM_HPAGE_SIZE(x) / PAGE_SIZE)
75 
76 #define DE_VECTOR 0
77 #define DB_VECTOR 1
78 #define BP_VECTOR 3
79 #define OF_VECTOR 4
80 #define BR_VECTOR 5
81 #define UD_VECTOR 6
82 #define NM_VECTOR 7
83 #define DF_VECTOR 8
84 #define TS_VECTOR 10
85 #define NP_VECTOR 11
86 #define SS_VECTOR 12
87 #define GP_VECTOR 13
88 #define PF_VECTOR 14
89 #define MF_VECTOR 16
90 #define MC_VECTOR 18
91 
92 #define SELECTOR_TI_MASK (1 << 2)
93 #define SELECTOR_RPL_MASK 0x03
94 
95 #define IOPL_SHIFT 12
96 
97 #define KVM_PERMILLE_MMU_PAGES 20
98 #define KVM_MIN_ALLOC_MMU_PAGES 64
99 #define KVM_MMU_HASH_SHIFT 10
100 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
101 #define KVM_MIN_FREE_MMU_PAGES 5
102 #define KVM_REFILL_PAGES 25
103 #define KVM_MAX_CPUID_ENTRIES 80
104 #define KVM_NR_FIXED_MTRR_REGION 88
105 #define KVM_NR_VAR_MTRR 8
106 
107 #define ASYNC_PF_PER_VCPU 64
108 
109 extern raw_spinlock_t kvm_lock;
110 extern struct list_head vm_list;
111 
112 struct kvm_vcpu;
113 struct kvm;
114 struct kvm_async_pf;
115 
116 enum kvm_reg {
117 	VCPU_REGS_RAX = 0,
118 	VCPU_REGS_RCX = 1,
119 	VCPU_REGS_RDX = 2,
120 	VCPU_REGS_RBX = 3,
121 	VCPU_REGS_RSP = 4,
122 	VCPU_REGS_RBP = 5,
123 	VCPU_REGS_RSI = 6,
124 	VCPU_REGS_RDI = 7,
125 #ifdef CONFIG_X86_64
126 	VCPU_REGS_R8 = 8,
127 	VCPU_REGS_R9 = 9,
128 	VCPU_REGS_R10 = 10,
129 	VCPU_REGS_R11 = 11,
130 	VCPU_REGS_R12 = 12,
131 	VCPU_REGS_R13 = 13,
132 	VCPU_REGS_R14 = 14,
133 	VCPU_REGS_R15 = 15,
134 #endif
135 	VCPU_REGS_RIP,
136 	NR_VCPU_REGS
137 };
138 
139 enum kvm_reg_ex {
140 	VCPU_EXREG_PDPTR = NR_VCPU_REGS,
141 	VCPU_EXREG_CR3,
142 	VCPU_EXREG_RFLAGS,
143 	VCPU_EXREG_CPL,
144 	VCPU_EXREG_SEGMENTS,
145 };
146 
147 enum {
148 	VCPU_SREG_ES,
149 	VCPU_SREG_CS,
150 	VCPU_SREG_SS,
151 	VCPU_SREG_DS,
152 	VCPU_SREG_FS,
153 	VCPU_SREG_GS,
154 	VCPU_SREG_TR,
155 	VCPU_SREG_LDTR,
156 };
157 
158 #include <asm/kvm_emulate.h>
159 
160 #define KVM_NR_MEM_OBJS 40
161 
162 #define KVM_NR_DB_REGS	4
163 
164 #define DR6_BD		(1 << 13)
165 #define DR6_BS		(1 << 14)
166 #define DR6_FIXED_1	0xffff0ff0
167 #define DR6_VOLATILE	0x0000e00f
168 
169 #define DR7_BP_EN_MASK	0x000000ff
170 #define DR7_GE		(1 << 9)
171 #define DR7_GD		(1 << 13)
172 #define DR7_FIXED_1	0x00000400
173 #define DR7_VOLATILE	0xffff23ff
174 
175 /*
176  * We don't want allocation failures within the mmu code, so we preallocate
177  * enough memory for a single page fault in a cache.
178  */
179 struct kvm_mmu_memory_cache {
180 	int nobjs;
181 	void *objects[KVM_NR_MEM_OBJS];
182 };
183 
184 #define NR_PTE_CHAIN_ENTRIES 5
185 
186 struct kvm_pte_chain {
187 	u64 *parent_ptes[NR_PTE_CHAIN_ENTRIES];
188 	struct hlist_node link;
189 };
190 
191 /*
192  * kvm_mmu_page_role, below, is defined as:
193  *
194  *   bits 0:3 - total guest paging levels (2-4, or zero for real mode)
195  *   bits 4:7 - page table level for this shadow (1-4)
196  *   bits 8:9 - page table quadrant for 2-level guests
197  *   bit   16 - direct mapping of virtual to physical mapping at gfn
198  *              used for real mode and two-dimensional paging
199  *   bits 17:19 - common access permissions for all ptes in this shadow page
200  */
201 union kvm_mmu_page_role {
202 	unsigned word;
203 	struct {
204 		unsigned level:4;
205 		unsigned cr4_pae:1;
206 		unsigned quadrant:2;
207 		unsigned pad_for_nice_hex_output:6;
208 		unsigned direct:1;
209 		unsigned access:3;
210 		unsigned invalid:1;
211 		unsigned nxe:1;
212 		unsigned cr0_wp:1;
213 		unsigned smep_andnot_wp:1;
214 	};
215 };
216 
217 struct kvm_mmu_page {
218 	struct list_head link;
219 	struct hlist_node hash_link;
220 
221 	/*
222 	 * The following two entries are used to key the shadow page in the
223 	 * hash table.
224 	 */
225 	gfn_t gfn;
226 	union kvm_mmu_page_role role;
227 
228 	u64 *spt;
229 	/* hold the gfn of each spte inside spt */
230 	gfn_t *gfns;
231 	/*
232 	 * One bit set per slot which has memory
233 	 * in this shadow page.
234 	 */
235 	DECLARE_BITMAP(slot_bitmap, KVM_MEM_SLOTS_NUM);
236 	bool unsync;
237 	int root_count;          /* Currently serving as active root */
238 	unsigned int unsync_children;
239 	unsigned long parent_ptes;	/* Reverse mapping for parent_pte */
240 	DECLARE_BITMAP(unsync_child_bitmap, 512);
241 
242 #ifdef CONFIG_X86_32
243 	int clear_spte_count;
244 #endif
245 
246 	int write_flooding_count;
247 
248 	struct rcu_head rcu;
249 };
250 
251 struct kvm_pio_request {
252 	unsigned long count;
253 	int in;
254 	int port;
255 	int size;
256 };
257 
258 /*
259  * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
260  * 32-bit).  The kvm_mmu structure abstracts the details of the current mmu
261  * mode.
262  */
263 struct kvm_mmu {
264 	void (*new_cr3)(struct kvm_vcpu *vcpu);
265 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long root);
266 	unsigned long (*get_cr3)(struct kvm_vcpu *vcpu);
267 	u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
268 	int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err,
269 			  bool prefault);
270 	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
271 				  struct x86_exception *fault);
272 	void (*free)(struct kvm_vcpu *vcpu);
273 	gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva, u32 access,
274 			    struct x86_exception *exception);
275 	gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access);
276 	int (*sync_page)(struct kvm_vcpu *vcpu,
277 			 struct kvm_mmu_page *sp);
278 	void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva);
279 	void (*update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
280 			   u64 *spte, const void *pte);
281 	hpa_t root_hpa;
282 	int root_level;
283 	int shadow_root_level;
284 	union kvm_mmu_page_role base_role;
285 	bool direct_map;
286 
287 	u64 *pae_root;
288 	u64 *lm_root;
289 	u64 rsvd_bits_mask[2][4];
290 
291 	bool nx;
292 
293 	u64 pdptrs[4]; /* pae */
294 };
295 
296 enum pmc_type {
297 	KVM_PMC_GP = 0,
298 	KVM_PMC_FIXED,
299 };
300 
301 struct kvm_pmc {
302 	enum pmc_type type;
303 	u8 idx;
304 	u64 counter;
305 	u64 eventsel;
306 	struct perf_event *perf_event;
307 	struct kvm_vcpu *vcpu;
308 };
309 
310 struct kvm_pmu {
311 	unsigned nr_arch_gp_counters;
312 	unsigned nr_arch_fixed_counters;
313 	unsigned available_event_types;
314 	u64 fixed_ctr_ctrl;
315 	u64 global_ctrl;
316 	u64 global_status;
317 	u64 global_ovf_ctrl;
318 	u64 counter_bitmask[2];
319 	u64 global_ctrl_mask;
320 	u8 version;
321 	struct kvm_pmc gp_counters[X86_PMC_MAX_GENERIC];
322 	struct kvm_pmc fixed_counters[X86_PMC_MAX_FIXED];
323 	struct irq_work irq_work;
324 	u64 reprogram_pmi;
325 };
326 
327 struct kvm_vcpu_arch {
328 	/*
329 	 * rip and regs accesses must go through
330 	 * kvm_{register,rip}_{read,write} functions.
331 	 */
332 	unsigned long regs[NR_VCPU_REGS];
333 	u32 regs_avail;
334 	u32 regs_dirty;
335 
336 	unsigned long cr0;
337 	unsigned long cr0_guest_owned_bits;
338 	unsigned long cr2;
339 	unsigned long cr3;
340 	unsigned long cr4;
341 	unsigned long cr4_guest_owned_bits;
342 	unsigned long cr8;
343 	u32 hflags;
344 	u64 efer;
345 	u64 apic_base;
346 	struct kvm_lapic *apic;    /* kernel irqchip context */
347 	int32_t apic_arb_prio;
348 	int mp_state;
349 	int sipi_vector;
350 	u64 ia32_misc_enable_msr;
351 	bool tpr_access_reporting;
352 
353 	/*
354 	 * Paging state of the vcpu
355 	 *
356 	 * If the vcpu runs in guest mode with two level paging this still saves
357 	 * the paging mode of the l1 guest. This context is always used to
358 	 * handle faults.
359 	 */
360 	struct kvm_mmu mmu;
361 
362 	/*
363 	 * Paging state of an L2 guest (used for nested npt)
364 	 *
365 	 * This context will save all necessary information to walk page tables
366 	 * of the an L2 guest. This context is only initialized for page table
367 	 * walking and not for faulting since we never handle l2 page faults on
368 	 * the host.
369 	 */
370 	struct kvm_mmu nested_mmu;
371 
372 	/*
373 	 * Pointer to the mmu context currently used for
374 	 * gva_to_gpa translations.
375 	 */
376 	struct kvm_mmu *walk_mmu;
377 
378 	struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
379 	struct kvm_mmu_memory_cache mmu_page_cache;
380 	struct kvm_mmu_memory_cache mmu_page_header_cache;
381 
382 	struct fpu guest_fpu;
383 	u64 xcr0;
384 
385 	struct kvm_pio_request pio;
386 	void *pio_data;
387 
388 	u8 event_exit_inst_len;
389 
390 	struct kvm_queued_exception {
391 		bool pending;
392 		bool has_error_code;
393 		bool reinject;
394 		u8 nr;
395 		u32 error_code;
396 	} exception;
397 
398 	struct kvm_queued_interrupt {
399 		bool pending;
400 		bool soft;
401 		u8 nr;
402 	} interrupt;
403 
404 	int halt_request; /* real mode on Intel only */
405 
406 	int cpuid_nent;
407 	struct kvm_cpuid_entry2 cpuid_entries[KVM_MAX_CPUID_ENTRIES];
408 	/* emulate context */
409 
410 	struct x86_emulate_ctxt emulate_ctxt;
411 	bool emulate_regs_need_sync_to_vcpu;
412 	bool emulate_regs_need_sync_from_vcpu;
413 
414 	gpa_t time;
415 	struct pvclock_vcpu_time_info hv_clock;
416 	unsigned int hw_tsc_khz;
417 	unsigned int time_offset;
418 	struct page *time_page;
419 
420 	struct {
421 		u64 msr_val;
422 		u64 last_steal;
423 		u64 accum_steal;
424 		struct gfn_to_hva_cache stime;
425 		struct kvm_steal_time steal;
426 	} st;
427 
428 	u64 last_guest_tsc;
429 	u64 last_kernel_ns;
430 	u64 last_tsc_nsec;
431 	u64 last_tsc_write;
432 	u32 virtual_tsc_khz;
433 	bool tsc_catchup;
434 	u32  tsc_catchup_mult;
435 	s8   tsc_catchup_shift;
436 
437 	atomic_t nmi_queued;  /* unprocessed asynchronous NMIs */
438 	unsigned nmi_pending; /* NMI queued after currently running handler */
439 	bool nmi_injected;    /* Trying to inject an NMI this entry */
440 
441 	struct mtrr_state_type mtrr_state;
442 	u32 pat;
443 
444 	int switch_db_regs;
445 	unsigned long db[KVM_NR_DB_REGS];
446 	unsigned long dr6;
447 	unsigned long dr7;
448 	unsigned long eff_db[KVM_NR_DB_REGS];
449 
450 	u64 mcg_cap;
451 	u64 mcg_status;
452 	u64 mcg_ctl;
453 	u64 *mce_banks;
454 
455 	/* Cache MMIO info */
456 	u64 mmio_gva;
457 	unsigned access;
458 	gfn_t mmio_gfn;
459 
460 	struct kvm_pmu pmu;
461 
462 	/* used for guest single stepping over the given code position */
463 	unsigned long singlestep_rip;
464 
465 	/* fields used by HYPER-V emulation */
466 	u64 hv_vapic;
467 
468 	cpumask_var_t wbinvd_dirty_mask;
469 
470 	unsigned long last_retry_eip;
471 	unsigned long last_retry_addr;
472 
473 	struct {
474 		bool halted;
475 		gfn_t gfns[roundup_pow_of_two(ASYNC_PF_PER_VCPU)];
476 		struct gfn_to_hva_cache data;
477 		u64 msr_val;
478 		u32 id;
479 		bool send_user_only;
480 	} apf;
481 };
482 
483 struct kvm_arch {
484 	unsigned int n_used_mmu_pages;
485 	unsigned int n_requested_mmu_pages;
486 	unsigned int n_max_mmu_pages;
487 	unsigned int indirect_shadow_pages;
488 	struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
489 	/*
490 	 * Hash table of struct kvm_mmu_page.
491 	 */
492 	struct list_head active_mmu_pages;
493 	struct list_head assigned_dev_head;
494 	struct iommu_domain *iommu_domain;
495 	int iommu_flags;
496 	struct kvm_pic *vpic;
497 	struct kvm_ioapic *vioapic;
498 	struct kvm_pit *vpit;
499 	int vapics_in_nmi_mode;
500 
501 	unsigned int tss_addr;
502 	struct page *apic_access_page;
503 
504 	gpa_t wall_clock;
505 
506 	struct page *ept_identity_pagetable;
507 	bool ept_identity_pagetable_done;
508 	gpa_t ept_identity_map_addr;
509 
510 	unsigned long irq_sources_bitmap;
511 	s64 kvmclock_offset;
512 	raw_spinlock_t tsc_write_lock;
513 	u64 last_tsc_nsec;
514 	u64 last_tsc_offset;
515 	u64 last_tsc_write;
516 
517 	struct kvm_xen_hvm_config xen_hvm_config;
518 
519 	/* fields used by HYPER-V emulation */
520 	u64 hv_guest_os_id;
521 	u64 hv_hypercall;
522 
523 	atomic_t reader_counter;
524 
525 	#ifdef CONFIG_KVM_MMU_AUDIT
526 	int audit_point;
527 	#endif
528 };
529 
530 struct kvm_vm_stat {
531 	u32 mmu_shadow_zapped;
532 	u32 mmu_pte_write;
533 	u32 mmu_pte_updated;
534 	u32 mmu_pde_zapped;
535 	u32 mmu_flooded;
536 	u32 mmu_recycled;
537 	u32 mmu_cache_miss;
538 	u32 mmu_unsync;
539 	u32 remote_tlb_flush;
540 	u32 lpages;
541 };
542 
543 struct kvm_vcpu_stat {
544 	u32 pf_fixed;
545 	u32 pf_guest;
546 	u32 tlb_flush;
547 	u32 invlpg;
548 
549 	u32 exits;
550 	u32 io_exits;
551 	u32 mmio_exits;
552 	u32 signal_exits;
553 	u32 irq_window_exits;
554 	u32 nmi_window_exits;
555 	u32 halt_exits;
556 	u32 halt_wakeup;
557 	u32 request_irq_exits;
558 	u32 irq_exits;
559 	u32 host_state_reload;
560 	u32 efer_reload;
561 	u32 fpu_reload;
562 	u32 insn_emulation;
563 	u32 insn_emulation_fail;
564 	u32 hypercalls;
565 	u32 irq_injections;
566 	u32 nmi_injections;
567 };
568 
569 struct x86_instruction_info;
570 
571 struct kvm_x86_ops {
572 	int (*cpu_has_kvm_support)(void);          /* __init */
573 	int (*disabled_by_bios)(void);             /* __init */
574 	int (*hardware_enable)(void *dummy);
575 	void (*hardware_disable)(void *dummy);
576 	void (*check_processor_compatibility)(void *rtn);
577 	int (*hardware_setup)(void);               /* __init */
578 	void (*hardware_unsetup)(void);            /* __exit */
579 	bool (*cpu_has_accelerated_tpr)(void);
580 	void (*cpuid_update)(struct kvm_vcpu *vcpu);
581 
582 	/* Create, but do not attach this VCPU */
583 	struct kvm_vcpu *(*vcpu_create)(struct kvm *kvm, unsigned id);
584 	void (*vcpu_free)(struct kvm_vcpu *vcpu);
585 	int (*vcpu_reset)(struct kvm_vcpu *vcpu);
586 
587 	void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
588 	void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
589 	void (*vcpu_put)(struct kvm_vcpu *vcpu);
590 
591 	void (*set_guest_debug)(struct kvm_vcpu *vcpu,
592 				struct kvm_guest_debug *dbg);
593 	int (*get_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata);
594 	int (*set_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
595 	u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
596 	void (*get_segment)(struct kvm_vcpu *vcpu,
597 			    struct kvm_segment *var, int seg);
598 	int (*get_cpl)(struct kvm_vcpu *vcpu);
599 	void (*set_segment)(struct kvm_vcpu *vcpu,
600 			    struct kvm_segment *var, int seg);
601 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
602 	void (*decache_cr0_guest_bits)(struct kvm_vcpu *vcpu);
603 	void (*decache_cr3)(struct kvm_vcpu *vcpu);
604 	void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu);
605 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
606 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
607 	int (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
608 	void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
609 	void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
610 	void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
611 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
612 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
613 	void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
614 	void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
615 	unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
616 	void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
617 	void (*fpu_activate)(struct kvm_vcpu *vcpu);
618 	void (*fpu_deactivate)(struct kvm_vcpu *vcpu);
619 
620 	void (*tlb_flush)(struct kvm_vcpu *vcpu);
621 
622 	void (*run)(struct kvm_vcpu *vcpu);
623 	int (*handle_exit)(struct kvm_vcpu *vcpu);
624 	void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
625 	void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
626 	u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
627 	void (*patch_hypercall)(struct kvm_vcpu *vcpu,
628 				unsigned char *hypercall_addr);
629 	void (*set_irq)(struct kvm_vcpu *vcpu);
630 	void (*set_nmi)(struct kvm_vcpu *vcpu);
631 	void (*queue_exception)(struct kvm_vcpu *vcpu, unsigned nr,
632 				bool has_error_code, u32 error_code,
633 				bool reinject);
634 	void (*cancel_injection)(struct kvm_vcpu *vcpu);
635 	int (*interrupt_allowed)(struct kvm_vcpu *vcpu);
636 	int (*nmi_allowed)(struct kvm_vcpu *vcpu);
637 	bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
638 	void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
639 	void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
640 	void (*enable_irq_window)(struct kvm_vcpu *vcpu);
641 	void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
642 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
643 	int (*get_tdp_level)(void);
644 	u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
645 	int (*get_lpage_level)(void);
646 	bool (*rdtscp_supported)(void);
647 	void (*adjust_tsc_offset)(struct kvm_vcpu *vcpu, s64 adjustment);
648 
649 	void (*set_tdp_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
650 
651 	void (*set_supported_cpuid)(u32 func, struct kvm_cpuid_entry2 *entry);
652 
653 	bool (*has_wbinvd_exit)(void);
654 
655 	void (*set_tsc_khz)(struct kvm_vcpu *vcpu, u32 user_tsc_khz);
656 	void (*write_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
657 
658 	u64 (*compute_tsc_offset)(struct kvm_vcpu *vcpu, u64 target_tsc);
659 	u64 (*read_l1_tsc)(struct kvm_vcpu *vcpu);
660 
661 	void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2);
662 
663 	int (*check_intercept)(struct kvm_vcpu *vcpu,
664 			       struct x86_instruction_info *info,
665 			       enum x86_intercept_stage stage);
666 };
667 
668 struct kvm_arch_async_pf {
669 	u32 token;
670 	gfn_t gfn;
671 	unsigned long cr3;
672 	bool direct_map;
673 };
674 
675 extern struct kvm_x86_ops *kvm_x86_ops;
676 
677 int kvm_mmu_module_init(void);
678 void kvm_mmu_module_exit(void);
679 
680 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
681 int kvm_mmu_create(struct kvm_vcpu *vcpu);
682 int kvm_mmu_setup(struct kvm_vcpu *vcpu);
683 void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
684 		u64 dirty_mask, u64 nx_mask, u64 x_mask);
685 
686 int kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
687 void kvm_mmu_slot_remove_write_access(struct kvm *kvm, int slot);
688 int kvm_mmu_rmap_write_protect(struct kvm *kvm, u64 gfn,
689 			       struct kvm_memory_slot *slot);
690 void kvm_mmu_zap_all(struct kvm *kvm);
691 unsigned int kvm_mmu_calculate_mmu_pages(struct kvm *kvm);
692 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int kvm_nr_mmu_pages);
693 
694 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
695 
696 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
697 			  const void *val, int bytes);
698 u8 kvm_get_guest_memory_type(struct kvm_vcpu *vcpu, gfn_t gfn);
699 
700 extern bool tdp_enabled;
701 
702 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
703 
704 /* control of guest tsc rate supported? */
705 extern bool kvm_has_tsc_control;
706 /* minimum supported tsc_khz for guests */
707 extern u32  kvm_min_guest_tsc_khz;
708 /* maximum supported tsc_khz for guests */
709 extern u32  kvm_max_guest_tsc_khz;
710 
711 enum emulation_result {
712 	EMULATE_DONE,       /* no further processing */
713 	EMULATE_DO_MMIO,      /* kvm_run filled with mmio request */
714 	EMULATE_FAIL,         /* can't emulate this instruction */
715 };
716 
717 #define EMULTYPE_NO_DECODE	    (1 << 0)
718 #define EMULTYPE_TRAP_UD	    (1 << 1)
719 #define EMULTYPE_SKIP		    (1 << 2)
720 #define EMULTYPE_RETRY		    (1 << 3)
721 int x86_emulate_instruction(struct kvm_vcpu *vcpu, unsigned long cr2,
722 			    int emulation_type, void *insn, int insn_len);
723 
724 static inline int emulate_instruction(struct kvm_vcpu *vcpu,
725 			int emulation_type)
726 {
727 	return x86_emulate_instruction(vcpu, 0, emulation_type, NULL, 0);
728 }
729 
730 void kvm_enable_efer_bits(u64);
731 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 *data);
732 int kvm_set_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
733 
734 struct x86_emulate_ctxt;
735 
736 int kvm_fast_pio_out(struct kvm_vcpu *vcpu, int size, unsigned short port);
737 void kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
738 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
739 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
740 
741 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
742 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
743 
744 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int reason,
745 		    bool has_error_code, u32 error_code);
746 
747 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
748 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
749 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
750 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
751 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
752 int kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
753 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
754 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
755 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
756 int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr);
757 
758 int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata);
759 int kvm_set_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 data);
760 
761 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
762 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
763 bool kvm_rdpmc(struct kvm_vcpu *vcpu);
764 
765 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
766 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
767 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
768 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
769 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
770 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
771 			    gfn_t gfn, void *data, int offset, int len,
772 			    u32 access);
773 void kvm_propagate_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
774 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
775 
776 int kvm_pic_set_irq(void *opaque, int irq, int level);
777 
778 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
779 
780 int fx_init(struct kvm_vcpu *vcpu);
781 
782 void kvm_mmu_flush_tlb(struct kvm_vcpu *vcpu);
783 void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
784 		       const u8 *new, int bytes);
785 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
786 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
787 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
788 int kvm_mmu_load(struct kvm_vcpu *vcpu);
789 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
790 void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu);
791 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access);
792 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
793 			      struct x86_exception *exception);
794 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
795 			       struct x86_exception *exception);
796 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
797 			       struct x86_exception *exception);
798 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
799 				struct x86_exception *exception);
800 
801 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
802 
803 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva, u32 error_code,
804 		       void *insn, int insn_len);
805 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
806 
807 void kvm_enable_tdp(void);
808 void kvm_disable_tdp(void);
809 
810 int complete_pio(struct kvm_vcpu *vcpu);
811 bool kvm_check_iopl(struct kvm_vcpu *vcpu);
812 
813 static inline gpa_t translate_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access)
814 {
815 	return gpa;
816 }
817 
818 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
819 {
820 	struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
821 
822 	return (struct kvm_mmu_page *)page_private(page);
823 }
824 
825 static inline u16 kvm_read_ldt(void)
826 {
827 	u16 ldt;
828 	asm("sldt %0" : "=g"(ldt));
829 	return ldt;
830 }
831 
832 static inline void kvm_load_ldt(u16 sel)
833 {
834 	asm("lldt %0" : : "rm"(sel));
835 }
836 
837 #ifdef CONFIG_X86_64
838 static inline unsigned long read_msr(unsigned long msr)
839 {
840 	u64 value;
841 
842 	rdmsrl(msr, value);
843 	return value;
844 }
845 #endif
846 
847 static inline u32 get_rdx_init_val(void)
848 {
849 	return 0x600; /* P6 family */
850 }
851 
852 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
853 {
854 	kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
855 }
856 
857 #define TSS_IOPB_BASE_OFFSET 0x66
858 #define TSS_BASE_SIZE 0x68
859 #define TSS_IOPB_SIZE (65536 / 8)
860 #define TSS_REDIRECTION_SIZE (256 / 8)
861 #define RMODE_TSS_SIZE							\
862 	(TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
863 
864 enum {
865 	TASK_SWITCH_CALL = 0,
866 	TASK_SWITCH_IRET = 1,
867 	TASK_SWITCH_JMP = 2,
868 	TASK_SWITCH_GATE = 3,
869 };
870 
871 #define HF_GIF_MASK		(1 << 0)
872 #define HF_HIF_MASK		(1 << 1)
873 #define HF_VINTR_MASK		(1 << 2)
874 #define HF_NMI_MASK		(1 << 3)
875 #define HF_IRET_MASK		(1 << 4)
876 #define HF_GUEST_MASK		(1 << 5) /* VCPU is in guest-mode */
877 
878 /*
879  * Hardware virtualization extension instructions may fault if a
880  * reboot turns off virtualization while processes are running.
881  * Trap the fault and ignore the instruction if that happens.
882  */
883 asmlinkage void kvm_spurious_fault(void);
884 extern bool kvm_rebooting;
885 
886 #define ____kvm_handle_fault_on_reboot(insn, cleanup_insn)	\
887 	"666: " insn "\n\t" \
888 	"668: \n\t"                           \
889 	".pushsection .fixup, \"ax\" \n" \
890 	"667: \n\t" \
891 	cleanup_insn "\n\t"		      \
892 	"cmpb $0, kvm_rebooting \n\t"	      \
893 	"jne 668b \n\t"      		      \
894 	__ASM_SIZE(push) " $666b \n\t"	      \
895 	"call kvm_spurious_fault \n\t"	      \
896 	".popsection \n\t" \
897 	".pushsection __ex_table, \"a\" \n\t" \
898 	_ASM_PTR " 666b, 667b \n\t" \
899 	".popsection"
900 
901 #define __kvm_handle_fault_on_reboot(insn)		\
902 	____kvm_handle_fault_on_reboot(insn, "")
903 
904 #define KVM_ARCH_WANT_MMU_NOTIFIER
905 int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
906 int kvm_age_hva(struct kvm *kvm, unsigned long hva);
907 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
908 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
909 int cpuid_maxphyaddr(struct kvm_vcpu *vcpu);
910 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
911 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
912 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
913 
914 void kvm_define_shared_msr(unsigned index, u32 msr);
915 void kvm_set_shared_msr(unsigned index, u64 val, u64 mask);
916 
917 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
918 
919 void kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
920 				     struct kvm_async_pf *work);
921 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
922 				 struct kvm_async_pf *work);
923 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
924 			       struct kvm_async_pf *work);
925 bool kvm_arch_can_inject_async_page_present(struct kvm_vcpu *vcpu);
926 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
927 
928 void kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
929 
930 int kvm_is_in_guest(void);
931 
932 void kvm_pmu_init(struct kvm_vcpu *vcpu);
933 void kvm_pmu_destroy(struct kvm_vcpu *vcpu);
934 void kvm_pmu_reset(struct kvm_vcpu *vcpu);
935 void kvm_pmu_cpuid_update(struct kvm_vcpu *vcpu);
936 bool kvm_pmu_msr(struct kvm_vcpu *vcpu, u32 msr);
937 int kvm_pmu_get_msr(struct kvm_vcpu *vcpu, u32 msr, u64 *data);
938 int kvm_pmu_set_msr(struct kvm_vcpu *vcpu, u32 msr, u64 data);
939 int kvm_pmu_read_pmc(struct kvm_vcpu *vcpu, unsigned pmc, u64 *data);
940 void kvm_handle_pmu_event(struct kvm_vcpu *vcpu);
941 void kvm_deliver_pmi(struct kvm_vcpu *vcpu);
942 
943 #endif /* _ASM_X86_KVM_HOST_H */
944