xref: /openbmc/linux/arch/x86/include/asm/kvm_host.h (revision ab73b751)
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 #include <asm/asm.h>
31 
32 #define KVM_MAX_VCPUS 254
33 #define KVM_SOFT_MAX_VCPUS 160
34 #define KVM_MEMORY_SLOTS 32
35 /* memory slots that does not exposed to userspace */
36 #define KVM_PRIVATE_MEM_SLOTS 4
37 #define KVM_MEM_SLOTS_NUM (KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS)
38 
39 #define KVM_MMIO_SIZE 16
40 
41 #define KVM_PIO_PAGE_OFFSET 1
42 #define KVM_COALESCED_MMIO_PAGE_OFFSET 2
43 
44 #define CR0_RESERVED_BITS                                               \
45 	(~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \
46 			  | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \
47 			  | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG))
48 
49 #define CR3_PAE_RESERVED_BITS ((X86_CR3_PWT | X86_CR3_PCD) - 1)
50 #define CR3_NONPAE_RESERVED_BITS ((PAGE_SIZE-1) & ~(X86_CR3_PWT | X86_CR3_PCD))
51 #define CR3_L_MODE_RESERVED_BITS (CR3_NONPAE_RESERVED_BITS |	\
52 				  0xFFFFFF0000000000ULL)
53 #define CR4_RESERVED_BITS                                               \
54 	(~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\
55 			  | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE     \
56 			  | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR  \
57 			  | X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_RDWRGSFS \
58 			  | X86_CR4_OSXMMEXCPT | X86_CR4_VMXE))
59 
60 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR)
61 
62 
63 
64 #define INVALID_PAGE (~(hpa_t)0)
65 #define VALID_PAGE(x) ((x) != INVALID_PAGE)
66 
67 #define UNMAPPED_GVA (~(gpa_t)0)
68 
69 /* KVM Hugepage definitions for x86 */
70 #define KVM_NR_PAGE_SIZES	3
71 #define KVM_HPAGE_GFN_SHIFT(x)	(((x) - 1) * 9)
72 #define KVM_HPAGE_SHIFT(x)	(PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
73 #define KVM_HPAGE_SIZE(x)	(1UL << KVM_HPAGE_SHIFT(x))
74 #define KVM_HPAGE_MASK(x)	(~(KVM_HPAGE_SIZE(x) - 1))
75 #define KVM_PAGES_PER_HPAGE(x)	(KVM_HPAGE_SIZE(x) / PAGE_SIZE)
76 
77 #define DE_VECTOR 0
78 #define DB_VECTOR 1
79 #define BP_VECTOR 3
80 #define OF_VECTOR 4
81 #define BR_VECTOR 5
82 #define UD_VECTOR 6
83 #define NM_VECTOR 7
84 #define DF_VECTOR 8
85 #define TS_VECTOR 10
86 #define NP_VECTOR 11
87 #define SS_VECTOR 12
88 #define GP_VECTOR 13
89 #define PF_VECTOR 14
90 #define MF_VECTOR 16
91 #define MC_VECTOR 18
92 
93 #define SELECTOR_TI_MASK (1 << 2)
94 #define SELECTOR_RPL_MASK 0x03
95 
96 #define IOPL_SHIFT 12
97 
98 #define KVM_PERMILLE_MMU_PAGES 20
99 #define KVM_MIN_ALLOC_MMU_PAGES 64
100 #define KVM_MMU_HASH_SHIFT 10
101 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
102 #define KVM_MIN_FREE_MMU_PAGES 5
103 #define KVM_REFILL_PAGES 25
104 #define KVM_MAX_CPUID_ENTRIES 80
105 #define KVM_NR_FIXED_MTRR_REGION 88
106 #define KVM_NR_VAR_MTRR 8
107 
108 #define ASYNC_PF_PER_VCPU 64
109 
110 extern raw_spinlock_t kvm_lock;
111 extern struct list_head vm_list;
112 
113 struct kvm_vcpu;
114 struct kvm;
115 struct kvm_async_pf;
116 
117 enum kvm_reg {
118 	VCPU_REGS_RAX = 0,
119 	VCPU_REGS_RCX = 1,
120 	VCPU_REGS_RDX = 2,
121 	VCPU_REGS_RBX = 3,
122 	VCPU_REGS_RSP = 4,
123 	VCPU_REGS_RBP = 5,
124 	VCPU_REGS_RSI = 6,
125 	VCPU_REGS_RDI = 7,
126 #ifdef CONFIG_X86_64
127 	VCPU_REGS_R8 = 8,
128 	VCPU_REGS_R9 = 9,
129 	VCPU_REGS_R10 = 10,
130 	VCPU_REGS_R11 = 11,
131 	VCPU_REGS_R12 = 12,
132 	VCPU_REGS_R13 = 13,
133 	VCPU_REGS_R14 = 14,
134 	VCPU_REGS_R15 = 15,
135 #endif
136 	VCPU_REGS_RIP,
137 	NR_VCPU_REGS
138 };
139 
140 enum kvm_reg_ex {
141 	VCPU_EXREG_PDPTR = NR_VCPU_REGS,
142 	VCPU_EXREG_CR3,
143 	VCPU_EXREG_RFLAGS,
144 	VCPU_EXREG_CPL,
145 	VCPU_EXREG_SEGMENTS,
146 };
147 
148 enum {
149 	VCPU_SREG_ES,
150 	VCPU_SREG_CS,
151 	VCPU_SREG_SS,
152 	VCPU_SREG_DS,
153 	VCPU_SREG_FS,
154 	VCPU_SREG_GS,
155 	VCPU_SREG_TR,
156 	VCPU_SREG_LDTR,
157 };
158 
159 #include <asm/kvm_emulate.h>
160 
161 #define KVM_NR_MEM_OBJS 40
162 
163 #define KVM_NR_DB_REGS	4
164 
165 #define DR6_BD		(1 << 13)
166 #define DR6_BS		(1 << 14)
167 #define DR6_FIXED_1	0xffff0ff0
168 #define DR6_VOLATILE	0x0000e00f
169 
170 #define DR7_BP_EN_MASK	0x000000ff
171 #define DR7_GE		(1 << 9)
172 #define DR7_GD		(1 << 13)
173 #define DR7_FIXED_1	0x00000400
174 #define DR7_VOLATILE	0xffff23ff
175 
176 /* apic attention bits */
177 #define KVM_APIC_CHECK_VAPIC	0
178 
179 /*
180  * We don't want allocation failures within the mmu code, so we preallocate
181  * enough memory for a single page fault in a cache.
182  */
183 struct kvm_mmu_memory_cache {
184 	int nobjs;
185 	void *objects[KVM_NR_MEM_OBJS];
186 };
187 
188 /*
189  * kvm_mmu_page_role, below, is defined as:
190  *
191  *   bits 0:3 - total guest paging levels (2-4, or zero for real mode)
192  *   bits 4:7 - page table level for this shadow (1-4)
193  *   bits 8:9 - page table quadrant for 2-level guests
194  *   bit   16 - direct mapping of virtual to physical mapping at gfn
195  *              used for real mode and two-dimensional paging
196  *   bits 17:19 - common access permissions for all ptes in this shadow page
197  */
198 union kvm_mmu_page_role {
199 	unsigned word;
200 	struct {
201 		unsigned level:4;
202 		unsigned cr4_pae:1;
203 		unsigned quadrant:2;
204 		unsigned pad_for_nice_hex_output:6;
205 		unsigned direct:1;
206 		unsigned access:3;
207 		unsigned invalid:1;
208 		unsigned nxe:1;
209 		unsigned cr0_wp:1;
210 		unsigned smep_andnot_wp:1;
211 	};
212 };
213 
214 struct kvm_mmu_page {
215 	struct list_head link;
216 	struct hlist_node hash_link;
217 
218 	/*
219 	 * The following two entries are used to key the shadow page in the
220 	 * hash table.
221 	 */
222 	gfn_t gfn;
223 	union kvm_mmu_page_role role;
224 
225 	u64 *spt;
226 	/* hold the gfn of each spte inside spt */
227 	gfn_t *gfns;
228 	/*
229 	 * One bit set per slot which has memory
230 	 * in this shadow page.
231 	 */
232 	DECLARE_BITMAP(slot_bitmap, KVM_MEM_SLOTS_NUM);
233 	bool unsync;
234 	int root_count;          /* Currently serving as active root */
235 	unsigned int unsync_children;
236 	unsigned long parent_ptes;	/* Reverse mapping for parent_pte */
237 	DECLARE_BITMAP(unsync_child_bitmap, 512);
238 
239 #ifdef CONFIG_X86_32
240 	int clear_spte_count;
241 #endif
242 
243 	int write_flooding_count;
244 };
245 
246 struct kvm_pio_request {
247 	unsigned long count;
248 	int in;
249 	int port;
250 	int size;
251 };
252 
253 /*
254  * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
255  * 32-bit).  The kvm_mmu structure abstracts the details of the current mmu
256  * mode.
257  */
258 struct kvm_mmu {
259 	void (*new_cr3)(struct kvm_vcpu *vcpu);
260 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long root);
261 	unsigned long (*get_cr3)(struct kvm_vcpu *vcpu);
262 	u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
263 	int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err,
264 			  bool prefault);
265 	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
266 				  struct x86_exception *fault);
267 	void (*free)(struct kvm_vcpu *vcpu);
268 	gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva, u32 access,
269 			    struct x86_exception *exception);
270 	gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access);
271 	int (*sync_page)(struct kvm_vcpu *vcpu,
272 			 struct kvm_mmu_page *sp);
273 	void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva);
274 	void (*update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
275 			   u64 *spte, const void *pte);
276 	hpa_t root_hpa;
277 	int root_level;
278 	int shadow_root_level;
279 	union kvm_mmu_page_role base_role;
280 	bool direct_map;
281 
282 	u64 *pae_root;
283 	u64 *lm_root;
284 	u64 rsvd_bits_mask[2][4];
285 
286 	bool nx;
287 
288 	u64 pdptrs[4]; /* pae */
289 };
290 
291 enum pmc_type {
292 	KVM_PMC_GP = 0,
293 	KVM_PMC_FIXED,
294 };
295 
296 struct kvm_pmc {
297 	enum pmc_type type;
298 	u8 idx;
299 	u64 counter;
300 	u64 eventsel;
301 	struct perf_event *perf_event;
302 	struct kvm_vcpu *vcpu;
303 };
304 
305 struct kvm_pmu {
306 	unsigned nr_arch_gp_counters;
307 	unsigned nr_arch_fixed_counters;
308 	unsigned available_event_types;
309 	u64 fixed_ctr_ctrl;
310 	u64 global_ctrl;
311 	u64 global_status;
312 	u64 global_ovf_ctrl;
313 	u64 counter_bitmask[2];
314 	u64 global_ctrl_mask;
315 	u8 version;
316 	struct kvm_pmc gp_counters[X86_PMC_MAX_GENERIC];
317 	struct kvm_pmc fixed_counters[X86_PMC_MAX_FIXED];
318 	struct irq_work irq_work;
319 	u64 reprogram_pmi;
320 };
321 
322 struct kvm_vcpu_arch {
323 	/*
324 	 * rip and regs accesses must go through
325 	 * kvm_{register,rip}_{read,write} functions.
326 	 */
327 	unsigned long regs[NR_VCPU_REGS];
328 	u32 regs_avail;
329 	u32 regs_dirty;
330 
331 	unsigned long cr0;
332 	unsigned long cr0_guest_owned_bits;
333 	unsigned long cr2;
334 	unsigned long cr3;
335 	unsigned long cr4;
336 	unsigned long cr4_guest_owned_bits;
337 	unsigned long cr8;
338 	u32 hflags;
339 	u64 efer;
340 	u64 apic_base;
341 	struct kvm_lapic *apic;    /* kernel irqchip context */
342 	unsigned long apic_attention;
343 	int32_t apic_arb_prio;
344 	int mp_state;
345 	int sipi_vector;
346 	u64 ia32_misc_enable_msr;
347 	bool tpr_access_reporting;
348 
349 	/*
350 	 * Paging state of the vcpu
351 	 *
352 	 * If the vcpu runs in guest mode with two level paging this still saves
353 	 * the paging mode of the l1 guest. This context is always used to
354 	 * handle faults.
355 	 */
356 	struct kvm_mmu mmu;
357 
358 	/*
359 	 * Paging state of an L2 guest (used for nested npt)
360 	 *
361 	 * This context will save all necessary information to walk page tables
362 	 * of the an L2 guest. This context is only initialized for page table
363 	 * walking and not for faulting since we never handle l2 page faults on
364 	 * the host.
365 	 */
366 	struct kvm_mmu nested_mmu;
367 
368 	/*
369 	 * Pointer to the mmu context currently used for
370 	 * gva_to_gpa translations.
371 	 */
372 	struct kvm_mmu *walk_mmu;
373 
374 	struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
375 	struct kvm_mmu_memory_cache mmu_page_cache;
376 	struct kvm_mmu_memory_cache mmu_page_header_cache;
377 
378 	struct fpu guest_fpu;
379 	u64 xcr0;
380 
381 	struct kvm_pio_request pio;
382 	void *pio_data;
383 
384 	u8 event_exit_inst_len;
385 
386 	struct kvm_queued_exception {
387 		bool pending;
388 		bool has_error_code;
389 		bool reinject;
390 		u8 nr;
391 		u32 error_code;
392 	} exception;
393 
394 	struct kvm_queued_interrupt {
395 		bool pending;
396 		bool soft;
397 		u8 nr;
398 	} interrupt;
399 
400 	int halt_request; /* real mode on Intel only */
401 
402 	int cpuid_nent;
403 	struct kvm_cpuid_entry2 cpuid_entries[KVM_MAX_CPUID_ENTRIES];
404 	/* emulate context */
405 
406 	struct x86_emulate_ctxt emulate_ctxt;
407 	bool emulate_regs_need_sync_to_vcpu;
408 	bool emulate_regs_need_sync_from_vcpu;
409 
410 	gpa_t time;
411 	struct pvclock_vcpu_time_info hv_clock;
412 	unsigned int hw_tsc_khz;
413 	unsigned int time_offset;
414 	struct page *time_page;
415 
416 	struct {
417 		u64 msr_val;
418 		u64 last_steal;
419 		u64 accum_steal;
420 		struct gfn_to_hva_cache stime;
421 		struct kvm_steal_time steal;
422 	} st;
423 
424 	u64 last_guest_tsc;
425 	u64 last_kernel_ns;
426 	u64 last_host_tsc;
427 	u64 tsc_offset_adjustment;
428 	u64 this_tsc_nsec;
429 	u64 this_tsc_write;
430 	u8  this_tsc_generation;
431 	bool tsc_catchup;
432 	bool tsc_always_catchup;
433 	s8 virtual_tsc_shift;
434 	u32 virtual_tsc_mult;
435 	u32 virtual_tsc_khz;
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 	/* OSVW MSRs (AMD only) */
483 	struct {
484 		u64 length;
485 		u64 status;
486 	} osvw;
487 };
488 
489 struct kvm_lpage_info {
490 	unsigned long rmap_pde;
491 	int write_count;
492 };
493 
494 struct kvm_arch_memory_slot {
495 	struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
496 };
497 
498 struct kvm_arch {
499 	unsigned int n_used_mmu_pages;
500 	unsigned int n_requested_mmu_pages;
501 	unsigned int n_max_mmu_pages;
502 	unsigned int indirect_shadow_pages;
503 	struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
504 	/*
505 	 * Hash table of struct kvm_mmu_page.
506 	 */
507 	struct list_head active_mmu_pages;
508 	struct list_head assigned_dev_head;
509 	struct iommu_domain *iommu_domain;
510 	int iommu_flags;
511 	struct kvm_pic *vpic;
512 	struct kvm_ioapic *vioapic;
513 	struct kvm_pit *vpit;
514 	int vapics_in_nmi_mode;
515 
516 	unsigned int tss_addr;
517 	struct page *apic_access_page;
518 
519 	gpa_t wall_clock;
520 
521 	struct page *ept_identity_pagetable;
522 	bool ept_identity_pagetable_done;
523 	gpa_t ept_identity_map_addr;
524 
525 	unsigned long irq_sources_bitmap;
526 	s64 kvmclock_offset;
527 	raw_spinlock_t tsc_write_lock;
528 	u64 last_tsc_nsec;
529 	u64 last_tsc_write;
530 	u32 last_tsc_khz;
531 	u64 cur_tsc_nsec;
532 	u64 cur_tsc_write;
533 	u64 cur_tsc_offset;
534 	u8  cur_tsc_generation;
535 
536 	struct kvm_xen_hvm_config xen_hvm_config;
537 
538 	/* fields used by HYPER-V emulation */
539 	u64 hv_guest_os_id;
540 	u64 hv_hypercall;
541 
542 	#ifdef CONFIG_KVM_MMU_AUDIT
543 	int audit_point;
544 	#endif
545 };
546 
547 struct kvm_vm_stat {
548 	u32 mmu_shadow_zapped;
549 	u32 mmu_pte_write;
550 	u32 mmu_pte_updated;
551 	u32 mmu_pde_zapped;
552 	u32 mmu_flooded;
553 	u32 mmu_recycled;
554 	u32 mmu_cache_miss;
555 	u32 mmu_unsync;
556 	u32 remote_tlb_flush;
557 	u32 lpages;
558 };
559 
560 struct kvm_vcpu_stat {
561 	u32 pf_fixed;
562 	u32 pf_guest;
563 	u32 tlb_flush;
564 	u32 invlpg;
565 
566 	u32 exits;
567 	u32 io_exits;
568 	u32 mmio_exits;
569 	u32 signal_exits;
570 	u32 irq_window_exits;
571 	u32 nmi_window_exits;
572 	u32 halt_exits;
573 	u32 halt_wakeup;
574 	u32 request_irq_exits;
575 	u32 irq_exits;
576 	u32 host_state_reload;
577 	u32 efer_reload;
578 	u32 fpu_reload;
579 	u32 insn_emulation;
580 	u32 insn_emulation_fail;
581 	u32 hypercalls;
582 	u32 irq_injections;
583 	u32 nmi_injections;
584 };
585 
586 struct x86_instruction_info;
587 
588 struct kvm_x86_ops {
589 	int (*cpu_has_kvm_support)(void);          /* __init */
590 	int (*disabled_by_bios)(void);             /* __init */
591 	int (*hardware_enable)(void *dummy);
592 	void (*hardware_disable)(void *dummy);
593 	void (*check_processor_compatibility)(void *rtn);
594 	int (*hardware_setup)(void);               /* __init */
595 	void (*hardware_unsetup)(void);            /* __exit */
596 	bool (*cpu_has_accelerated_tpr)(void);
597 	void (*cpuid_update)(struct kvm_vcpu *vcpu);
598 
599 	/* Create, but do not attach this VCPU */
600 	struct kvm_vcpu *(*vcpu_create)(struct kvm *kvm, unsigned id);
601 	void (*vcpu_free)(struct kvm_vcpu *vcpu);
602 	int (*vcpu_reset)(struct kvm_vcpu *vcpu);
603 
604 	void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
605 	void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
606 	void (*vcpu_put)(struct kvm_vcpu *vcpu);
607 
608 	void (*set_guest_debug)(struct kvm_vcpu *vcpu,
609 				struct kvm_guest_debug *dbg);
610 	int (*get_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata);
611 	int (*set_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
612 	u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
613 	void (*get_segment)(struct kvm_vcpu *vcpu,
614 			    struct kvm_segment *var, int seg);
615 	int (*get_cpl)(struct kvm_vcpu *vcpu);
616 	void (*set_segment)(struct kvm_vcpu *vcpu,
617 			    struct kvm_segment *var, int seg);
618 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
619 	void (*decache_cr0_guest_bits)(struct kvm_vcpu *vcpu);
620 	void (*decache_cr3)(struct kvm_vcpu *vcpu);
621 	void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu);
622 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
623 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
624 	int (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
625 	void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
626 	void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
627 	void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
628 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
629 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
630 	void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
631 	void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
632 	unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
633 	void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
634 	void (*fpu_activate)(struct kvm_vcpu *vcpu);
635 	void (*fpu_deactivate)(struct kvm_vcpu *vcpu);
636 
637 	void (*tlb_flush)(struct kvm_vcpu *vcpu);
638 
639 	void (*run)(struct kvm_vcpu *vcpu);
640 	int (*handle_exit)(struct kvm_vcpu *vcpu);
641 	void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
642 	void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
643 	u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
644 	void (*patch_hypercall)(struct kvm_vcpu *vcpu,
645 				unsigned char *hypercall_addr);
646 	void (*set_irq)(struct kvm_vcpu *vcpu);
647 	void (*set_nmi)(struct kvm_vcpu *vcpu);
648 	void (*queue_exception)(struct kvm_vcpu *vcpu, unsigned nr,
649 				bool has_error_code, u32 error_code,
650 				bool reinject);
651 	void (*cancel_injection)(struct kvm_vcpu *vcpu);
652 	int (*interrupt_allowed)(struct kvm_vcpu *vcpu);
653 	int (*nmi_allowed)(struct kvm_vcpu *vcpu);
654 	bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
655 	void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
656 	void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
657 	void (*enable_irq_window)(struct kvm_vcpu *vcpu);
658 	void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
659 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
660 	int (*get_tdp_level)(void);
661 	u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
662 	int (*get_lpage_level)(void);
663 	bool (*rdtscp_supported)(void);
664 	void (*adjust_tsc_offset)(struct kvm_vcpu *vcpu, s64 adjustment, bool host);
665 
666 	void (*set_tdp_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
667 
668 	void (*set_supported_cpuid)(u32 func, struct kvm_cpuid_entry2 *entry);
669 
670 	bool (*has_wbinvd_exit)(void);
671 
672 	void (*set_tsc_khz)(struct kvm_vcpu *vcpu, u32 user_tsc_khz, bool scale);
673 	void (*write_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
674 
675 	u64 (*compute_tsc_offset)(struct kvm_vcpu *vcpu, u64 target_tsc);
676 	u64 (*read_l1_tsc)(struct kvm_vcpu *vcpu);
677 
678 	void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2);
679 
680 	int (*check_intercept)(struct kvm_vcpu *vcpu,
681 			       struct x86_instruction_info *info,
682 			       enum x86_intercept_stage stage);
683 };
684 
685 struct kvm_arch_async_pf {
686 	u32 token;
687 	gfn_t gfn;
688 	unsigned long cr3;
689 	bool direct_map;
690 };
691 
692 extern struct kvm_x86_ops *kvm_x86_ops;
693 
694 static inline void adjust_tsc_offset_guest(struct kvm_vcpu *vcpu,
695 					   s64 adjustment)
696 {
697 	kvm_x86_ops->adjust_tsc_offset(vcpu, adjustment, false);
698 }
699 
700 static inline void adjust_tsc_offset_host(struct kvm_vcpu *vcpu, s64 adjustment)
701 {
702 	kvm_x86_ops->adjust_tsc_offset(vcpu, adjustment, true);
703 }
704 
705 int kvm_mmu_module_init(void);
706 void kvm_mmu_module_exit(void);
707 
708 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
709 int kvm_mmu_create(struct kvm_vcpu *vcpu);
710 int kvm_mmu_setup(struct kvm_vcpu *vcpu);
711 void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
712 		u64 dirty_mask, u64 nx_mask, u64 x_mask);
713 
714 int kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
715 void kvm_mmu_slot_remove_write_access(struct kvm *kvm, int slot);
716 void kvm_mmu_write_protect_pt_masked(struct kvm *kvm,
717 				     struct kvm_memory_slot *slot,
718 				     gfn_t gfn_offset, unsigned long mask);
719 void kvm_mmu_zap_all(struct kvm *kvm);
720 unsigned int kvm_mmu_calculate_mmu_pages(struct kvm *kvm);
721 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int kvm_nr_mmu_pages);
722 
723 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
724 
725 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
726 			  const void *val, int bytes);
727 u8 kvm_get_guest_memory_type(struct kvm_vcpu *vcpu, gfn_t gfn);
728 
729 extern bool tdp_enabled;
730 
731 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
732 
733 /* control of guest tsc rate supported? */
734 extern bool kvm_has_tsc_control;
735 /* minimum supported tsc_khz for guests */
736 extern u32  kvm_min_guest_tsc_khz;
737 /* maximum supported tsc_khz for guests */
738 extern u32  kvm_max_guest_tsc_khz;
739 
740 enum emulation_result {
741 	EMULATE_DONE,       /* no further processing */
742 	EMULATE_DO_MMIO,      /* kvm_run filled with mmio request */
743 	EMULATE_FAIL,         /* can't emulate this instruction */
744 };
745 
746 #define EMULTYPE_NO_DECODE	    (1 << 0)
747 #define EMULTYPE_TRAP_UD	    (1 << 1)
748 #define EMULTYPE_SKIP		    (1 << 2)
749 #define EMULTYPE_RETRY		    (1 << 3)
750 int x86_emulate_instruction(struct kvm_vcpu *vcpu, unsigned long cr2,
751 			    int emulation_type, void *insn, int insn_len);
752 
753 static inline int emulate_instruction(struct kvm_vcpu *vcpu,
754 			int emulation_type)
755 {
756 	return x86_emulate_instruction(vcpu, 0, emulation_type, NULL, 0);
757 }
758 
759 void kvm_enable_efer_bits(u64);
760 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 *data);
761 int kvm_set_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
762 
763 struct x86_emulate_ctxt;
764 
765 int kvm_fast_pio_out(struct kvm_vcpu *vcpu, int size, unsigned short port);
766 void kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
767 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
768 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
769 
770 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
771 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
772 
773 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
774 		    int reason, bool has_error_code, u32 error_code);
775 
776 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
777 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
778 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
779 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
780 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
781 int kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
782 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
783 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
784 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
785 int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr);
786 
787 int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata);
788 int kvm_set_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 data);
789 
790 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
791 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
792 bool kvm_rdpmc(struct kvm_vcpu *vcpu);
793 
794 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
795 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
796 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
797 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
798 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
799 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
800 			    gfn_t gfn, void *data, int offset, int len,
801 			    u32 access);
802 void kvm_propagate_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
803 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
804 
805 int kvm_pic_set_irq(void *opaque, int irq, int level);
806 
807 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
808 
809 int fx_init(struct kvm_vcpu *vcpu);
810 
811 void kvm_mmu_flush_tlb(struct kvm_vcpu *vcpu);
812 void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
813 		       const u8 *new, int bytes);
814 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
815 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
816 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
817 int kvm_mmu_load(struct kvm_vcpu *vcpu);
818 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
819 void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu);
820 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access);
821 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
822 			      struct x86_exception *exception);
823 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
824 			       struct x86_exception *exception);
825 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
826 			       struct x86_exception *exception);
827 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
828 				struct x86_exception *exception);
829 
830 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
831 
832 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva, u32 error_code,
833 		       void *insn, int insn_len);
834 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
835 
836 void kvm_enable_tdp(void);
837 void kvm_disable_tdp(void);
838 
839 int complete_pio(struct kvm_vcpu *vcpu);
840 bool kvm_check_iopl(struct kvm_vcpu *vcpu);
841 
842 static inline gpa_t translate_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access)
843 {
844 	return gpa;
845 }
846 
847 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
848 {
849 	struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
850 
851 	return (struct kvm_mmu_page *)page_private(page);
852 }
853 
854 static inline u16 kvm_read_ldt(void)
855 {
856 	u16 ldt;
857 	asm("sldt %0" : "=g"(ldt));
858 	return ldt;
859 }
860 
861 static inline void kvm_load_ldt(u16 sel)
862 {
863 	asm("lldt %0" : : "rm"(sel));
864 }
865 
866 #ifdef CONFIG_X86_64
867 static inline unsigned long read_msr(unsigned long msr)
868 {
869 	u64 value;
870 
871 	rdmsrl(msr, value);
872 	return value;
873 }
874 #endif
875 
876 static inline u32 get_rdx_init_val(void)
877 {
878 	return 0x600; /* P6 family */
879 }
880 
881 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
882 {
883 	kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
884 }
885 
886 #define TSS_IOPB_BASE_OFFSET 0x66
887 #define TSS_BASE_SIZE 0x68
888 #define TSS_IOPB_SIZE (65536 / 8)
889 #define TSS_REDIRECTION_SIZE (256 / 8)
890 #define RMODE_TSS_SIZE							\
891 	(TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
892 
893 enum {
894 	TASK_SWITCH_CALL = 0,
895 	TASK_SWITCH_IRET = 1,
896 	TASK_SWITCH_JMP = 2,
897 	TASK_SWITCH_GATE = 3,
898 };
899 
900 #define HF_GIF_MASK		(1 << 0)
901 #define HF_HIF_MASK		(1 << 1)
902 #define HF_VINTR_MASK		(1 << 2)
903 #define HF_NMI_MASK		(1 << 3)
904 #define HF_IRET_MASK		(1 << 4)
905 #define HF_GUEST_MASK		(1 << 5) /* VCPU is in guest-mode */
906 
907 /*
908  * Hardware virtualization extension instructions may fault if a
909  * reboot turns off virtualization while processes are running.
910  * Trap the fault and ignore the instruction if that happens.
911  */
912 asmlinkage void kvm_spurious_fault(void);
913 extern bool kvm_rebooting;
914 
915 #define ____kvm_handle_fault_on_reboot(insn, cleanup_insn)	\
916 	"666: " insn "\n\t" \
917 	"668: \n\t"                           \
918 	".pushsection .fixup, \"ax\" \n" \
919 	"667: \n\t" \
920 	cleanup_insn "\n\t"		      \
921 	"cmpb $0, kvm_rebooting \n\t"	      \
922 	"jne 668b \n\t"      		      \
923 	__ASM_SIZE(push) " $666b \n\t"	      \
924 	"call kvm_spurious_fault \n\t"	      \
925 	".popsection \n\t" \
926 	_ASM_EXTABLE(666b, 667b)
927 
928 #define __kvm_handle_fault_on_reboot(insn)		\
929 	____kvm_handle_fault_on_reboot(insn, "")
930 
931 #define KVM_ARCH_WANT_MMU_NOTIFIER
932 int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
933 int kvm_age_hva(struct kvm *kvm, unsigned long hva);
934 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
935 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
936 int cpuid_maxphyaddr(struct kvm_vcpu *vcpu);
937 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
938 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
939 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
940 
941 void kvm_define_shared_msr(unsigned index, u32 msr);
942 void kvm_set_shared_msr(unsigned index, u64 val, u64 mask);
943 
944 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
945 
946 void kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
947 				     struct kvm_async_pf *work);
948 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
949 				 struct kvm_async_pf *work);
950 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
951 			       struct kvm_async_pf *work);
952 bool kvm_arch_can_inject_async_page_present(struct kvm_vcpu *vcpu);
953 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
954 
955 void kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
956 
957 int kvm_is_in_guest(void);
958 
959 void kvm_pmu_init(struct kvm_vcpu *vcpu);
960 void kvm_pmu_destroy(struct kvm_vcpu *vcpu);
961 void kvm_pmu_reset(struct kvm_vcpu *vcpu);
962 void kvm_pmu_cpuid_update(struct kvm_vcpu *vcpu);
963 bool kvm_pmu_msr(struct kvm_vcpu *vcpu, u32 msr);
964 int kvm_pmu_get_msr(struct kvm_vcpu *vcpu, u32 msr, u64 *data);
965 int kvm_pmu_set_msr(struct kvm_vcpu *vcpu, u32 msr, u64 data);
966 int kvm_pmu_read_pmc(struct kvm_vcpu *vcpu, unsigned pmc, u64 *data);
967 void kvm_handle_pmu_event(struct kvm_vcpu *vcpu);
968 void kvm_deliver_pmi(struct kvm_vcpu *vcpu);
969 
970 #endif /* _ASM_X86_KVM_HOST_H */
971