xref: /openbmc/linux/arch/x86/include/asm/kvm_host.h (revision 8fdff1dc)
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 254
35 #define KVM_SOFT_MAX_VCPUS 160
36 #define KVM_MEMORY_SLOTS 32
37 /* memory slots that does not exposed to userspace */
38 #define KVM_PRIVATE_MEM_SLOTS 4
39 #define KVM_MEM_SLOTS_NUM (KVM_MEMORY_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 CR0_RESERVED_BITS                                               \
47 	(~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \
48 			  | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \
49 			  | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG))
50 
51 #define CR3_PAE_RESERVED_BITS ((X86_CR3_PWT | X86_CR3_PCD) - 1)
52 #define CR3_NONPAE_RESERVED_BITS ((PAGE_SIZE-1) & ~(X86_CR3_PWT | X86_CR3_PCD))
53 #define CR3_PCID_ENABLED_RESERVED_BITS 0xFFFFFF0000000000ULL
54 #define CR3_L_MODE_RESERVED_BITS (CR3_NONPAE_RESERVED_BITS |	\
55 				  0xFFFFFF0000000000ULL)
56 #define CR4_RESERVED_BITS                                               \
57 	(~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\
58 			  | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE     \
59 			  | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR | X86_CR4_PCIDE \
60 			  | X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_RDWRGSFS \
61 			  | X86_CR4_OSXMMEXCPT | X86_CR4_VMXE))
62 
63 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR)
64 
65 
66 
67 #define INVALID_PAGE (~(hpa_t)0)
68 #define VALID_PAGE(x) ((x) != INVALID_PAGE)
69 
70 #define UNMAPPED_GVA (~(gpa_t)0)
71 
72 /* KVM Hugepage definitions for x86 */
73 #define KVM_NR_PAGE_SIZES	3
74 #define KVM_HPAGE_GFN_SHIFT(x)	(((x) - 1) * 9)
75 #define KVM_HPAGE_SHIFT(x)	(PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
76 #define KVM_HPAGE_SIZE(x)	(1UL << KVM_HPAGE_SHIFT(x))
77 #define KVM_HPAGE_MASK(x)	(~(KVM_HPAGE_SIZE(x) - 1))
78 #define KVM_PAGES_PER_HPAGE(x)	(KVM_HPAGE_SIZE(x) / PAGE_SIZE)
79 
80 #define SELECTOR_TI_MASK (1 << 2)
81 #define SELECTOR_RPL_MASK 0x03
82 
83 #define IOPL_SHIFT 12
84 
85 #define KVM_PERMILLE_MMU_PAGES 20
86 #define KVM_MIN_ALLOC_MMU_PAGES 64
87 #define KVM_MMU_HASH_SHIFT 10
88 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
89 #define KVM_MIN_FREE_MMU_PAGES 5
90 #define KVM_REFILL_PAGES 25
91 #define KVM_MAX_CPUID_ENTRIES 80
92 #define KVM_NR_FIXED_MTRR_REGION 88
93 #define KVM_NR_VAR_MTRR 8
94 
95 #define ASYNC_PF_PER_VCPU 64
96 
97 extern raw_spinlock_t kvm_lock;
98 extern struct list_head vm_list;
99 
100 struct kvm_vcpu;
101 struct kvm;
102 struct kvm_async_pf;
103 
104 enum kvm_reg {
105 	VCPU_REGS_RAX = 0,
106 	VCPU_REGS_RCX = 1,
107 	VCPU_REGS_RDX = 2,
108 	VCPU_REGS_RBX = 3,
109 	VCPU_REGS_RSP = 4,
110 	VCPU_REGS_RBP = 5,
111 	VCPU_REGS_RSI = 6,
112 	VCPU_REGS_RDI = 7,
113 #ifdef CONFIG_X86_64
114 	VCPU_REGS_R8 = 8,
115 	VCPU_REGS_R9 = 9,
116 	VCPU_REGS_R10 = 10,
117 	VCPU_REGS_R11 = 11,
118 	VCPU_REGS_R12 = 12,
119 	VCPU_REGS_R13 = 13,
120 	VCPU_REGS_R14 = 14,
121 	VCPU_REGS_R15 = 15,
122 #endif
123 	VCPU_REGS_RIP,
124 	NR_VCPU_REGS
125 };
126 
127 enum kvm_reg_ex {
128 	VCPU_EXREG_PDPTR = NR_VCPU_REGS,
129 	VCPU_EXREG_CR3,
130 	VCPU_EXREG_RFLAGS,
131 	VCPU_EXREG_CPL,
132 	VCPU_EXREG_SEGMENTS,
133 };
134 
135 enum {
136 	VCPU_SREG_ES,
137 	VCPU_SREG_CS,
138 	VCPU_SREG_SS,
139 	VCPU_SREG_DS,
140 	VCPU_SREG_FS,
141 	VCPU_SREG_GS,
142 	VCPU_SREG_TR,
143 	VCPU_SREG_LDTR,
144 };
145 
146 #include <asm/kvm_emulate.h>
147 
148 #define KVM_NR_MEM_OBJS 40
149 
150 #define KVM_NR_DB_REGS	4
151 
152 #define DR6_BD		(1 << 13)
153 #define DR6_BS		(1 << 14)
154 #define DR6_FIXED_1	0xffff0ff0
155 #define DR6_VOLATILE	0x0000e00f
156 
157 #define DR7_BP_EN_MASK	0x000000ff
158 #define DR7_GE		(1 << 9)
159 #define DR7_GD		(1 << 13)
160 #define DR7_FIXED_1	0x00000400
161 #define DR7_VOLATILE	0xffff23ff
162 
163 /* apic attention bits */
164 #define KVM_APIC_CHECK_VAPIC	0
165 /*
166  * The following bit is set with PV-EOI, unset on EOI.
167  * We detect PV-EOI changes by guest by comparing
168  * this bit with PV-EOI in guest memory.
169  * See the implementation in apic_update_pv_eoi.
170  */
171 #define KVM_APIC_PV_EOI_PENDING	1
172 
173 /*
174  * We don't want allocation failures within the mmu code, so we preallocate
175  * enough memory for a single page fault in a cache.
176  */
177 struct kvm_mmu_memory_cache {
178 	int nobjs;
179 	void *objects[KVM_NR_MEM_OBJS];
180 };
181 
182 /*
183  * kvm_mmu_page_role, below, is defined as:
184  *
185  *   bits 0:3 - total guest paging levels (2-4, or zero for real mode)
186  *   bits 4:7 - page table level for this shadow (1-4)
187  *   bits 8:9 - page table quadrant for 2-level guests
188  *   bit   16 - direct mapping of virtual to physical mapping at gfn
189  *              used for real mode and two-dimensional paging
190  *   bits 17:19 - common access permissions for all ptes in this shadow page
191  */
192 union kvm_mmu_page_role {
193 	unsigned word;
194 	struct {
195 		unsigned level:4;
196 		unsigned cr4_pae:1;
197 		unsigned quadrant:2;
198 		unsigned pad_for_nice_hex_output:6;
199 		unsigned direct:1;
200 		unsigned access:3;
201 		unsigned invalid:1;
202 		unsigned nxe:1;
203 		unsigned cr0_wp:1;
204 		unsigned smep_andnot_wp:1;
205 	};
206 };
207 
208 struct kvm_mmu_page {
209 	struct list_head link;
210 	struct hlist_node hash_link;
211 
212 	/*
213 	 * The following two entries are used to key the shadow page in the
214 	 * hash table.
215 	 */
216 	gfn_t gfn;
217 	union kvm_mmu_page_role role;
218 
219 	u64 *spt;
220 	/* hold the gfn of each spte inside spt */
221 	gfn_t *gfns;
222 	/*
223 	 * One bit set per slot which has memory
224 	 * in this shadow page.
225 	 */
226 	DECLARE_BITMAP(slot_bitmap, KVM_MEM_SLOTS_NUM);
227 	bool unsync;
228 	int root_count;          /* Currently serving as active root */
229 	unsigned int unsync_children;
230 	unsigned long parent_ptes;	/* Reverse mapping for parent_pte */
231 	DECLARE_BITMAP(unsync_child_bitmap, 512);
232 
233 #ifdef CONFIG_X86_32
234 	int clear_spte_count;
235 #endif
236 
237 	int write_flooding_count;
238 };
239 
240 struct kvm_pio_request {
241 	unsigned long count;
242 	int in;
243 	int port;
244 	int size;
245 };
246 
247 /*
248  * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
249  * 32-bit).  The kvm_mmu structure abstracts the details of the current mmu
250  * mode.
251  */
252 struct kvm_mmu {
253 	void (*new_cr3)(struct kvm_vcpu *vcpu);
254 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long root);
255 	unsigned long (*get_cr3)(struct kvm_vcpu *vcpu);
256 	u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
257 	int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err,
258 			  bool prefault);
259 	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
260 				  struct x86_exception *fault);
261 	void (*free)(struct kvm_vcpu *vcpu);
262 	gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva, u32 access,
263 			    struct x86_exception *exception);
264 	gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access);
265 	int (*sync_page)(struct kvm_vcpu *vcpu,
266 			 struct kvm_mmu_page *sp);
267 	void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva);
268 	void (*update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
269 			   u64 *spte, const void *pte);
270 	hpa_t root_hpa;
271 	int root_level;
272 	int shadow_root_level;
273 	union kvm_mmu_page_role base_role;
274 	bool direct_map;
275 
276 	/*
277 	 * Bitmap; bit set = permission fault
278 	 * Byte index: page fault error code [4:1]
279 	 * Bit index: pte permissions in ACC_* format
280 	 */
281 	u8 permissions[16];
282 
283 	u64 *pae_root;
284 	u64 *lm_root;
285 	u64 rsvd_bits_mask[2][4];
286 
287 	/*
288 	 * Bitmap: bit set = last pte in walk
289 	 * index[0:1]: level (zero-based)
290 	 * index[2]: pte.ps
291 	 */
292 	u8 last_pte_bitmap;
293 
294 	bool nx;
295 
296 	u64 pdptrs[4]; /* pae */
297 };
298 
299 enum pmc_type {
300 	KVM_PMC_GP = 0,
301 	KVM_PMC_FIXED,
302 };
303 
304 struct kvm_pmc {
305 	enum pmc_type type;
306 	u8 idx;
307 	u64 counter;
308 	u64 eventsel;
309 	struct perf_event *perf_event;
310 	struct kvm_vcpu *vcpu;
311 };
312 
313 struct kvm_pmu {
314 	unsigned nr_arch_gp_counters;
315 	unsigned nr_arch_fixed_counters;
316 	unsigned available_event_types;
317 	u64 fixed_ctr_ctrl;
318 	u64 global_ctrl;
319 	u64 global_status;
320 	u64 global_ovf_ctrl;
321 	u64 counter_bitmask[2];
322 	u64 global_ctrl_mask;
323 	u8 version;
324 	struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
325 	struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED];
326 	struct irq_work irq_work;
327 	u64 reprogram_pmi;
328 };
329 
330 struct kvm_vcpu_arch {
331 	/*
332 	 * rip and regs accesses must go through
333 	 * kvm_{register,rip}_{read,write} functions.
334 	 */
335 	unsigned long regs[NR_VCPU_REGS];
336 	u32 regs_avail;
337 	u32 regs_dirty;
338 
339 	unsigned long cr0;
340 	unsigned long cr0_guest_owned_bits;
341 	unsigned long cr2;
342 	unsigned long cr3;
343 	unsigned long cr4;
344 	unsigned long cr4_guest_owned_bits;
345 	unsigned long cr8;
346 	u32 hflags;
347 	u64 efer;
348 	u64 apic_base;
349 	struct kvm_lapic *apic;    /* kernel irqchip context */
350 	unsigned long apic_attention;
351 	int32_t apic_arb_prio;
352 	int mp_state;
353 	int sipi_vector;
354 	u64 ia32_misc_enable_msr;
355 	bool tpr_access_reporting;
356 
357 	/*
358 	 * Paging state of the vcpu
359 	 *
360 	 * If the vcpu runs in guest mode with two level paging this still saves
361 	 * the paging mode of the l1 guest. This context is always used to
362 	 * handle faults.
363 	 */
364 	struct kvm_mmu mmu;
365 
366 	/*
367 	 * Paging state of an L2 guest (used for nested npt)
368 	 *
369 	 * This context will save all necessary information to walk page tables
370 	 * of the an L2 guest. This context is only initialized for page table
371 	 * walking and not for faulting since we never handle l2 page faults on
372 	 * the host.
373 	 */
374 	struct kvm_mmu nested_mmu;
375 
376 	/*
377 	 * Pointer to the mmu context currently used for
378 	 * gva_to_gpa translations.
379 	 */
380 	struct kvm_mmu *walk_mmu;
381 
382 	struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
383 	struct kvm_mmu_memory_cache mmu_page_cache;
384 	struct kvm_mmu_memory_cache mmu_page_header_cache;
385 
386 	struct fpu guest_fpu;
387 	u64 xcr0;
388 
389 	struct kvm_pio_request pio;
390 	void *pio_data;
391 
392 	u8 event_exit_inst_len;
393 
394 	struct kvm_queued_exception {
395 		bool pending;
396 		bool has_error_code;
397 		bool reinject;
398 		u8 nr;
399 		u32 error_code;
400 	} exception;
401 
402 	struct kvm_queued_interrupt {
403 		bool pending;
404 		bool soft;
405 		u8 nr;
406 	} interrupt;
407 
408 	int halt_request; /* real mode on Intel only */
409 
410 	int cpuid_nent;
411 	struct kvm_cpuid_entry2 cpuid_entries[KVM_MAX_CPUID_ENTRIES];
412 	/* emulate context */
413 
414 	struct x86_emulate_ctxt emulate_ctxt;
415 	bool emulate_regs_need_sync_to_vcpu;
416 	bool emulate_regs_need_sync_from_vcpu;
417 	int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
418 
419 	gpa_t time;
420 	struct pvclock_vcpu_time_info hv_clock;
421 	unsigned int hw_tsc_khz;
422 	unsigned int time_offset;
423 	struct page *time_page;
424 	/* set guest stopped flag in pvclock flags field */
425 	bool pvclock_set_guest_stopped_request;
426 
427 	struct {
428 		u64 msr_val;
429 		u64 last_steal;
430 		u64 accum_steal;
431 		struct gfn_to_hva_cache stime;
432 		struct kvm_steal_time steal;
433 	} st;
434 
435 	u64 last_guest_tsc;
436 	u64 last_kernel_ns;
437 	u64 last_host_tsc;
438 	u64 tsc_offset_adjustment;
439 	u64 this_tsc_nsec;
440 	u64 this_tsc_write;
441 	u8  this_tsc_generation;
442 	bool tsc_catchup;
443 	bool tsc_always_catchup;
444 	s8 virtual_tsc_shift;
445 	u32 virtual_tsc_mult;
446 	u32 virtual_tsc_khz;
447 	s64 ia32_tsc_adjust_msr;
448 
449 	atomic_t nmi_queued;  /* unprocessed asynchronous NMIs */
450 	unsigned nmi_pending; /* NMI queued after currently running handler */
451 	bool nmi_injected;    /* Trying to inject an NMI this entry */
452 
453 	struct mtrr_state_type mtrr_state;
454 	u32 pat;
455 
456 	int switch_db_regs;
457 	unsigned long db[KVM_NR_DB_REGS];
458 	unsigned long dr6;
459 	unsigned long dr7;
460 	unsigned long eff_db[KVM_NR_DB_REGS];
461 	unsigned long guest_debug_dr7;
462 
463 	u64 mcg_cap;
464 	u64 mcg_status;
465 	u64 mcg_ctl;
466 	u64 *mce_banks;
467 
468 	/* Cache MMIO info */
469 	u64 mmio_gva;
470 	unsigned access;
471 	gfn_t mmio_gfn;
472 
473 	struct kvm_pmu pmu;
474 
475 	/* used for guest single stepping over the given code position */
476 	unsigned long singlestep_rip;
477 
478 	/* fields used by HYPER-V emulation */
479 	u64 hv_vapic;
480 
481 	cpumask_var_t wbinvd_dirty_mask;
482 
483 	unsigned long last_retry_eip;
484 	unsigned long last_retry_addr;
485 
486 	struct {
487 		bool halted;
488 		gfn_t gfns[roundup_pow_of_two(ASYNC_PF_PER_VCPU)];
489 		struct gfn_to_hva_cache data;
490 		u64 msr_val;
491 		u32 id;
492 		bool send_user_only;
493 	} apf;
494 
495 	/* OSVW MSRs (AMD only) */
496 	struct {
497 		u64 length;
498 		u64 status;
499 	} osvw;
500 
501 	struct {
502 		u64 msr_val;
503 		struct gfn_to_hva_cache data;
504 	} pv_eoi;
505 };
506 
507 struct kvm_lpage_info {
508 	int write_count;
509 };
510 
511 struct kvm_arch_memory_slot {
512 	unsigned long *rmap[KVM_NR_PAGE_SIZES];
513 	struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
514 };
515 
516 struct kvm_apic_map {
517 	struct rcu_head rcu;
518 	u8 ldr_bits;
519 	/* fields bellow are used to decode ldr values in different modes */
520 	u32 cid_shift, cid_mask, lid_mask;
521 	struct kvm_lapic *phys_map[256];
522 	/* first index is cluster id second is cpu id in a cluster */
523 	struct kvm_lapic *logical_map[16][16];
524 };
525 
526 struct kvm_arch {
527 	unsigned int n_used_mmu_pages;
528 	unsigned int n_requested_mmu_pages;
529 	unsigned int n_max_mmu_pages;
530 	unsigned int indirect_shadow_pages;
531 	struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
532 	/*
533 	 * Hash table of struct kvm_mmu_page.
534 	 */
535 	struct list_head active_mmu_pages;
536 	struct list_head assigned_dev_head;
537 	struct iommu_domain *iommu_domain;
538 	int iommu_flags;
539 	struct kvm_pic *vpic;
540 	struct kvm_ioapic *vioapic;
541 	struct kvm_pit *vpit;
542 	int vapics_in_nmi_mode;
543 	struct mutex apic_map_lock;
544 	struct kvm_apic_map *apic_map;
545 
546 	unsigned int tss_addr;
547 	struct page *apic_access_page;
548 
549 	gpa_t wall_clock;
550 
551 	struct page *ept_identity_pagetable;
552 	bool ept_identity_pagetable_done;
553 	gpa_t ept_identity_map_addr;
554 
555 	unsigned long irq_sources_bitmap;
556 	s64 kvmclock_offset;
557 	raw_spinlock_t tsc_write_lock;
558 	u64 last_tsc_nsec;
559 	u64 last_tsc_write;
560 	u32 last_tsc_khz;
561 	u64 cur_tsc_nsec;
562 	u64 cur_tsc_write;
563 	u64 cur_tsc_offset;
564 	u8  cur_tsc_generation;
565 	int nr_vcpus_matched_tsc;
566 
567 	spinlock_t pvclock_gtod_sync_lock;
568 	bool use_master_clock;
569 	u64 master_kernel_ns;
570 	cycle_t master_cycle_now;
571 
572 	struct kvm_xen_hvm_config xen_hvm_config;
573 
574 	/* fields used by HYPER-V emulation */
575 	u64 hv_guest_os_id;
576 	u64 hv_hypercall;
577 
578 	#ifdef CONFIG_KVM_MMU_AUDIT
579 	int audit_point;
580 	#endif
581 };
582 
583 struct kvm_vm_stat {
584 	u32 mmu_shadow_zapped;
585 	u32 mmu_pte_write;
586 	u32 mmu_pte_updated;
587 	u32 mmu_pde_zapped;
588 	u32 mmu_flooded;
589 	u32 mmu_recycled;
590 	u32 mmu_cache_miss;
591 	u32 mmu_unsync;
592 	u32 remote_tlb_flush;
593 	u32 lpages;
594 };
595 
596 struct kvm_vcpu_stat {
597 	u32 pf_fixed;
598 	u32 pf_guest;
599 	u32 tlb_flush;
600 	u32 invlpg;
601 
602 	u32 exits;
603 	u32 io_exits;
604 	u32 mmio_exits;
605 	u32 signal_exits;
606 	u32 irq_window_exits;
607 	u32 nmi_window_exits;
608 	u32 halt_exits;
609 	u32 halt_wakeup;
610 	u32 request_irq_exits;
611 	u32 irq_exits;
612 	u32 host_state_reload;
613 	u32 efer_reload;
614 	u32 fpu_reload;
615 	u32 insn_emulation;
616 	u32 insn_emulation_fail;
617 	u32 hypercalls;
618 	u32 irq_injections;
619 	u32 nmi_injections;
620 };
621 
622 struct x86_instruction_info;
623 
624 struct msr_data {
625 	bool host_initiated;
626 	u32 index;
627 	u64 data;
628 };
629 
630 struct kvm_x86_ops {
631 	int (*cpu_has_kvm_support)(void);          /* __init */
632 	int (*disabled_by_bios)(void);             /* __init */
633 	int (*hardware_enable)(void *dummy);
634 	void (*hardware_disable)(void *dummy);
635 	void (*check_processor_compatibility)(void *rtn);
636 	int (*hardware_setup)(void);               /* __init */
637 	void (*hardware_unsetup)(void);            /* __exit */
638 	bool (*cpu_has_accelerated_tpr)(void);
639 	void (*cpuid_update)(struct kvm_vcpu *vcpu);
640 
641 	/* Create, but do not attach this VCPU */
642 	struct kvm_vcpu *(*vcpu_create)(struct kvm *kvm, unsigned id);
643 	void (*vcpu_free)(struct kvm_vcpu *vcpu);
644 	int (*vcpu_reset)(struct kvm_vcpu *vcpu);
645 
646 	void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
647 	void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
648 	void (*vcpu_put)(struct kvm_vcpu *vcpu);
649 
650 	void (*update_db_bp_intercept)(struct kvm_vcpu *vcpu);
651 	int (*get_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata);
652 	int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
653 	u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
654 	void (*get_segment)(struct kvm_vcpu *vcpu,
655 			    struct kvm_segment *var, int seg);
656 	int (*get_cpl)(struct kvm_vcpu *vcpu);
657 	void (*set_segment)(struct kvm_vcpu *vcpu,
658 			    struct kvm_segment *var, int seg);
659 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
660 	void (*decache_cr0_guest_bits)(struct kvm_vcpu *vcpu);
661 	void (*decache_cr3)(struct kvm_vcpu *vcpu);
662 	void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu);
663 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
664 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
665 	int (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
666 	void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
667 	void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
668 	void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
669 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
670 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
671 	void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
672 	void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
673 	unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
674 	void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
675 	void (*fpu_activate)(struct kvm_vcpu *vcpu);
676 	void (*fpu_deactivate)(struct kvm_vcpu *vcpu);
677 
678 	void (*tlb_flush)(struct kvm_vcpu *vcpu);
679 
680 	void (*run)(struct kvm_vcpu *vcpu);
681 	int (*handle_exit)(struct kvm_vcpu *vcpu);
682 	void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
683 	void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
684 	u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
685 	void (*patch_hypercall)(struct kvm_vcpu *vcpu,
686 				unsigned char *hypercall_addr);
687 	void (*set_irq)(struct kvm_vcpu *vcpu);
688 	void (*set_nmi)(struct kvm_vcpu *vcpu);
689 	void (*queue_exception)(struct kvm_vcpu *vcpu, unsigned nr,
690 				bool has_error_code, u32 error_code,
691 				bool reinject);
692 	void (*cancel_injection)(struct kvm_vcpu *vcpu);
693 	int (*interrupt_allowed)(struct kvm_vcpu *vcpu);
694 	int (*nmi_allowed)(struct kvm_vcpu *vcpu);
695 	bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
696 	void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
697 	void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
698 	void (*enable_irq_window)(struct kvm_vcpu *vcpu);
699 	void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
700 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
701 	int (*get_tdp_level)(void);
702 	u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
703 	int (*get_lpage_level)(void);
704 	bool (*rdtscp_supported)(void);
705 	bool (*invpcid_supported)(void);
706 	void (*adjust_tsc_offset)(struct kvm_vcpu *vcpu, s64 adjustment, bool host);
707 
708 	void (*set_tdp_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
709 
710 	void (*set_supported_cpuid)(u32 func, struct kvm_cpuid_entry2 *entry);
711 
712 	bool (*has_wbinvd_exit)(void);
713 
714 	void (*set_tsc_khz)(struct kvm_vcpu *vcpu, u32 user_tsc_khz, bool scale);
715 	u64 (*read_tsc_offset)(struct kvm_vcpu *vcpu);
716 	void (*write_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
717 
718 	u64 (*compute_tsc_offset)(struct kvm_vcpu *vcpu, u64 target_tsc);
719 	u64 (*read_l1_tsc)(struct kvm_vcpu *vcpu, u64 host_tsc);
720 
721 	void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2);
722 
723 	int (*check_intercept)(struct kvm_vcpu *vcpu,
724 			       struct x86_instruction_info *info,
725 			       enum x86_intercept_stage stage);
726 };
727 
728 struct kvm_arch_async_pf {
729 	u32 token;
730 	gfn_t gfn;
731 	unsigned long cr3;
732 	bool direct_map;
733 };
734 
735 extern struct kvm_x86_ops *kvm_x86_ops;
736 
737 static inline void adjust_tsc_offset_guest(struct kvm_vcpu *vcpu,
738 					   s64 adjustment)
739 {
740 	kvm_x86_ops->adjust_tsc_offset(vcpu, adjustment, false);
741 }
742 
743 static inline void adjust_tsc_offset_host(struct kvm_vcpu *vcpu, s64 adjustment)
744 {
745 	kvm_x86_ops->adjust_tsc_offset(vcpu, adjustment, true);
746 }
747 
748 int kvm_mmu_module_init(void);
749 void kvm_mmu_module_exit(void);
750 
751 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
752 int kvm_mmu_create(struct kvm_vcpu *vcpu);
753 int kvm_mmu_setup(struct kvm_vcpu *vcpu);
754 void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
755 		u64 dirty_mask, u64 nx_mask, u64 x_mask);
756 
757 int kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
758 void kvm_mmu_slot_remove_write_access(struct kvm *kvm, int slot);
759 void kvm_mmu_write_protect_pt_masked(struct kvm *kvm,
760 				     struct kvm_memory_slot *slot,
761 				     gfn_t gfn_offset, unsigned long mask);
762 void kvm_mmu_zap_all(struct kvm *kvm);
763 unsigned int kvm_mmu_calculate_mmu_pages(struct kvm *kvm);
764 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int kvm_nr_mmu_pages);
765 
766 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
767 
768 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
769 			  const void *val, int bytes);
770 u8 kvm_get_guest_memory_type(struct kvm_vcpu *vcpu, gfn_t gfn);
771 
772 extern bool tdp_enabled;
773 
774 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
775 
776 /* control of guest tsc rate supported? */
777 extern bool kvm_has_tsc_control;
778 /* minimum supported tsc_khz for guests */
779 extern u32  kvm_min_guest_tsc_khz;
780 /* maximum supported tsc_khz for guests */
781 extern u32  kvm_max_guest_tsc_khz;
782 
783 enum emulation_result {
784 	EMULATE_DONE,       /* no further processing */
785 	EMULATE_DO_MMIO,      /* kvm_run filled with mmio request */
786 	EMULATE_FAIL,         /* can't emulate this instruction */
787 };
788 
789 #define EMULTYPE_NO_DECODE	    (1 << 0)
790 #define EMULTYPE_TRAP_UD	    (1 << 1)
791 #define EMULTYPE_SKIP		    (1 << 2)
792 #define EMULTYPE_RETRY		    (1 << 3)
793 int x86_emulate_instruction(struct kvm_vcpu *vcpu, unsigned long cr2,
794 			    int emulation_type, void *insn, int insn_len);
795 
796 static inline int emulate_instruction(struct kvm_vcpu *vcpu,
797 			int emulation_type)
798 {
799 	return x86_emulate_instruction(vcpu, 0, emulation_type, NULL, 0);
800 }
801 
802 void kvm_enable_efer_bits(u64);
803 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 *data);
804 int kvm_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr);
805 
806 struct x86_emulate_ctxt;
807 
808 int kvm_fast_pio_out(struct kvm_vcpu *vcpu, int size, unsigned short port);
809 void kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
810 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
811 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
812 
813 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
814 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
815 
816 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
817 		    int reason, bool has_error_code, u32 error_code);
818 
819 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
820 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
821 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
822 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
823 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
824 int kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
825 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
826 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
827 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
828 int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr);
829 
830 int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata);
831 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
832 
833 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
834 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
835 bool kvm_rdpmc(struct kvm_vcpu *vcpu);
836 
837 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
838 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
839 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
840 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
841 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
842 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
843 			    gfn_t gfn, void *data, int offset, int len,
844 			    u32 access);
845 void kvm_propagate_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
846 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
847 
848 static inline int __kvm_irq_line_state(unsigned long *irq_state,
849 				       int irq_source_id, int level)
850 {
851 	/* Logical OR for level trig interrupt */
852 	if (level)
853 		__set_bit(irq_source_id, irq_state);
854 	else
855 		__clear_bit(irq_source_id, irq_state);
856 
857 	return !!(*irq_state);
858 }
859 
860 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
861 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
862 
863 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
864 
865 int fx_init(struct kvm_vcpu *vcpu);
866 
867 void kvm_mmu_flush_tlb(struct kvm_vcpu *vcpu);
868 void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
869 		       const u8 *new, int bytes);
870 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
871 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
872 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
873 int kvm_mmu_load(struct kvm_vcpu *vcpu);
874 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
875 void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu);
876 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access);
877 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
878 			      struct x86_exception *exception);
879 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
880 			       struct x86_exception *exception);
881 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
882 			       struct x86_exception *exception);
883 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
884 				struct x86_exception *exception);
885 
886 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
887 
888 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva, u32 error_code,
889 		       void *insn, int insn_len);
890 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
891 
892 void kvm_enable_tdp(void);
893 void kvm_disable_tdp(void);
894 
895 int complete_pio(struct kvm_vcpu *vcpu);
896 bool kvm_check_iopl(struct kvm_vcpu *vcpu);
897 
898 static inline gpa_t translate_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access)
899 {
900 	return gpa;
901 }
902 
903 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
904 {
905 	struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
906 
907 	return (struct kvm_mmu_page *)page_private(page);
908 }
909 
910 static inline u16 kvm_read_ldt(void)
911 {
912 	u16 ldt;
913 	asm("sldt %0" : "=g"(ldt));
914 	return ldt;
915 }
916 
917 static inline void kvm_load_ldt(u16 sel)
918 {
919 	asm("lldt %0" : : "rm"(sel));
920 }
921 
922 #ifdef CONFIG_X86_64
923 static inline unsigned long read_msr(unsigned long msr)
924 {
925 	u64 value;
926 
927 	rdmsrl(msr, value);
928 	return value;
929 }
930 #endif
931 
932 static inline u32 get_rdx_init_val(void)
933 {
934 	return 0x600; /* P6 family */
935 }
936 
937 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
938 {
939 	kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
940 }
941 
942 #define TSS_IOPB_BASE_OFFSET 0x66
943 #define TSS_BASE_SIZE 0x68
944 #define TSS_IOPB_SIZE (65536 / 8)
945 #define TSS_REDIRECTION_SIZE (256 / 8)
946 #define RMODE_TSS_SIZE							\
947 	(TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
948 
949 enum {
950 	TASK_SWITCH_CALL = 0,
951 	TASK_SWITCH_IRET = 1,
952 	TASK_SWITCH_JMP = 2,
953 	TASK_SWITCH_GATE = 3,
954 };
955 
956 #define HF_GIF_MASK		(1 << 0)
957 #define HF_HIF_MASK		(1 << 1)
958 #define HF_VINTR_MASK		(1 << 2)
959 #define HF_NMI_MASK		(1 << 3)
960 #define HF_IRET_MASK		(1 << 4)
961 #define HF_GUEST_MASK		(1 << 5) /* VCPU is in guest-mode */
962 
963 /*
964  * Hardware virtualization extension instructions may fault if a
965  * reboot turns off virtualization while processes are running.
966  * Trap the fault and ignore the instruction if that happens.
967  */
968 asmlinkage void kvm_spurious_fault(void);
969 extern bool kvm_rebooting;
970 
971 #define ____kvm_handle_fault_on_reboot(insn, cleanup_insn)	\
972 	"666: " insn "\n\t" \
973 	"668: \n\t"                           \
974 	".pushsection .fixup, \"ax\" \n" \
975 	"667: \n\t" \
976 	cleanup_insn "\n\t"		      \
977 	"cmpb $0, kvm_rebooting \n\t"	      \
978 	"jne 668b \n\t"      		      \
979 	__ASM_SIZE(push) " $666b \n\t"	      \
980 	"call kvm_spurious_fault \n\t"	      \
981 	".popsection \n\t" \
982 	_ASM_EXTABLE(666b, 667b)
983 
984 #define __kvm_handle_fault_on_reboot(insn)		\
985 	____kvm_handle_fault_on_reboot(insn, "")
986 
987 #define KVM_ARCH_WANT_MMU_NOTIFIER
988 int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
989 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end);
990 int kvm_age_hva(struct kvm *kvm, unsigned long hva);
991 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
992 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
993 int cpuid_maxphyaddr(struct kvm_vcpu *vcpu);
994 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
995 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
996 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
997 
998 void kvm_define_shared_msr(unsigned index, u32 msr);
999 void kvm_set_shared_msr(unsigned index, u64 val, u64 mask);
1000 
1001 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1002 
1003 void kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1004 				     struct kvm_async_pf *work);
1005 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1006 				 struct kvm_async_pf *work);
1007 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1008 			       struct kvm_async_pf *work);
1009 bool kvm_arch_can_inject_async_page_present(struct kvm_vcpu *vcpu);
1010 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1011 
1012 void kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1013 
1014 int kvm_is_in_guest(void);
1015 
1016 void kvm_pmu_init(struct kvm_vcpu *vcpu);
1017 void kvm_pmu_destroy(struct kvm_vcpu *vcpu);
1018 void kvm_pmu_reset(struct kvm_vcpu *vcpu);
1019 void kvm_pmu_cpuid_update(struct kvm_vcpu *vcpu);
1020 bool kvm_pmu_msr(struct kvm_vcpu *vcpu, u32 msr);
1021 int kvm_pmu_get_msr(struct kvm_vcpu *vcpu, u32 msr, u64 *data);
1022 int kvm_pmu_set_msr(struct kvm_vcpu *vcpu, u32 msr, u64 data);
1023 int kvm_pmu_read_pmc(struct kvm_vcpu *vcpu, unsigned pmc, u64 *data);
1024 void kvm_handle_pmu_event(struct kvm_vcpu *vcpu);
1025 void kvm_deliver_pmi(struct kvm_vcpu *vcpu);
1026 
1027 #endif /* _ASM_X86_KVM_HOST_H */
1028