xref: /openbmc/linux/arch/x86/include/asm/kvm_host.h (revision 5b394b2d)
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 #include <linux/irq.h>
21 
22 #include <linux/kvm.h>
23 #include <linux/kvm_para.h>
24 #include <linux/kvm_types.h>
25 #include <linux/perf_event.h>
26 #include <linux/pvclock_gtod.h>
27 #include <linux/clocksource.h>
28 #include <linux/irqbypass.h>
29 #include <linux/hyperv.h>
30 
31 #include <asm/apic.h>
32 #include <asm/pvclock-abi.h>
33 #include <asm/desc.h>
34 #include <asm/mtrr.h>
35 #include <asm/msr-index.h>
36 #include <asm/asm.h>
37 #include <asm/kvm_page_track.h>
38 #include <asm/hyperv-tlfs.h>
39 
40 #define KVM_MAX_VCPUS 288
41 #define KVM_SOFT_MAX_VCPUS 240
42 #define KVM_MAX_VCPU_ID 1023
43 #define KVM_USER_MEM_SLOTS 509
44 /* memory slots that are not exposed to userspace */
45 #define KVM_PRIVATE_MEM_SLOTS 3
46 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
47 
48 #define KVM_HALT_POLL_NS_DEFAULT 200000
49 
50 #define KVM_IRQCHIP_NUM_PINS  KVM_IOAPIC_NUM_PINS
51 
52 /* x86-specific vcpu->requests bit members */
53 #define KVM_REQ_MIGRATE_TIMER		KVM_ARCH_REQ(0)
54 #define KVM_REQ_REPORT_TPR_ACCESS	KVM_ARCH_REQ(1)
55 #define KVM_REQ_TRIPLE_FAULT		KVM_ARCH_REQ(2)
56 #define KVM_REQ_MMU_SYNC		KVM_ARCH_REQ(3)
57 #define KVM_REQ_CLOCK_UPDATE		KVM_ARCH_REQ(4)
58 #define KVM_REQ_LOAD_CR3		KVM_ARCH_REQ(5)
59 #define KVM_REQ_EVENT			KVM_ARCH_REQ(6)
60 #define KVM_REQ_APF_HALT		KVM_ARCH_REQ(7)
61 #define KVM_REQ_STEAL_UPDATE		KVM_ARCH_REQ(8)
62 #define KVM_REQ_NMI			KVM_ARCH_REQ(9)
63 #define KVM_REQ_PMU			KVM_ARCH_REQ(10)
64 #define KVM_REQ_PMI			KVM_ARCH_REQ(11)
65 #define KVM_REQ_SMI			KVM_ARCH_REQ(12)
66 #define KVM_REQ_MASTERCLOCK_UPDATE	KVM_ARCH_REQ(13)
67 #define KVM_REQ_MCLOCK_INPROGRESS \
68 	KVM_ARCH_REQ_FLAGS(14, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
69 #define KVM_REQ_SCAN_IOAPIC \
70 	KVM_ARCH_REQ_FLAGS(15, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
71 #define KVM_REQ_GLOBAL_CLOCK_UPDATE	KVM_ARCH_REQ(16)
72 #define KVM_REQ_APIC_PAGE_RELOAD \
73 	KVM_ARCH_REQ_FLAGS(17, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
74 #define KVM_REQ_HV_CRASH		KVM_ARCH_REQ(18)
75 #define KVM_REQ_IOAPIC_EOI_EXIT		KVM_ARCH_REQ(19)
76 #define KVM_REQ_HV_RESET		KVM_ARCH_REQ(20)
77 #define KVM_REQ_HV_EXIT			KVM_ARCH_REQ(21)
78 #define KVM_REQ_HV_STIMER		KVM_ARCH_REQ(22)
79 #define KVM_REQ_LOAD_EOI_EXITMAP	KVM_ARCH_REQ(23)
80 #define KVM_REQ_GET_VMCS12_PAGES	KVM_ARCH_REQ(24)
81 
82 #define CR0_RESERVED_BITS                                               \
83 	(~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \
84 			  | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \
85 			  | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG))
86 
87 #define CR4_RESERVED_BITS                                               \
88 	(~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\
89 			  | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE     \
90 			  | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR | X86_CR4_PCIDE \
91 			  | X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_FSGSBASE \
92 			  | X86_CR4_OSXMMEXCPT | X86_CR4_LA57 | X86_CR4_VMXE \
93 			  | X86_CR4_SMAP | X86_CR4_PKE | X86_CR4_UMIP))
94 
95 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR)
96 
97 
98 
99 #define INVALID_PAGE (~(hpa_t)0)
100 #define VALID_PAGE(x) ((x) != INVALID_PAGE)
101 
102 #define UNMAPPED_GVA (~(gpa_t)0)
103 
104 /* KVM Hugepage definitions for x86 */
105 #define KVM_NR_PAGE_SIZES	3
106 #define KVM_HPAGE_GFN_SHIFT(x)	(((x) - 1) * 9)
107 #define KVM_HPAGE_SHIFT(x)	(PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
108 #define KVM_HPAGE_SIZE(x)	(1UL << KVM_HPAGE_SHIFT(x))
109 #define KVM_HPAGE_MASK(x)	(~(KVM_HPAGE_SIZE(x) - 1))
110 #define KVM_PAGES_PER_HPAGE(x)	(KVM_HPAGE_SIZE(x) / PAGE_SIZE)
111 
112 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
113 {
114 	/* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */
115 	return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
116 		(base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
117 }
118 
119 #define KVM_PERMILLE_MMU_PAGES 20
120 #define KVM_MIN_ALLOC_MMU_PAGES 64
121 #define KVM_MMU_HASH_SHIFT 12
122 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
123 #define KVM_MIN_FREE_MMU_PAGES 5
124 #define KVM_REFILL_PAGES 25
125 #define KVM_MAX_CPUID_ENTRIES 80
126 #define KVM_NR_FIXED_MTRR_REGION 88
127 #define KVM_NR_VAR_MTRR 8
128 
129 #define ASYNC_PF_PER_VCPU 64
130 
131 enum kvm_reg {
132 	VCPU_REGS_RAX = 0,
133 	VCPU_REGS_RCX = 1,
134 	VCPU_REGS_RDX = 2,
135 	VCPU_REGS_RBX = 3,
136 	VCPU_REGS_RSP = 4,
137 	VCPU_REGS_RBP = 5,
138 	VCPU_REGS_RSI = 6,
139 	VCPU_REGS_RDI = 7,
140 #ifdef CONFIG_X86_64
141 	VCPU_REGS_R8 = 8,
142 	VCPU_REGS_R9 = 9,
143 	VCPU_REGS_R10 = 10,
144 	VCPU_REGS_R11 = 11,
145 	VCPU_REGS_R12 = 12,
146 	VCPU_REGS_R13 = 13,
147 	VCPU_REGS_R14 = 14,
148 	VCPU_REGS_R15 = 15,
149 #endif
150 	VCPU_REGS_RIP,
151 	NR_VCPU_REGS
152 };
153 
154 enum kvm_reg_ex {
155 	VCPU_EXREG_PDPTR = NR_VCPU_REGS,
156 	VCPU_EXREG_CR3,
157 	VCPU_EXREG_RFLAGS,
158 	VCPU_EXREG_SEGMENTS,
159 };
160 
161 enum {
162 	VCPU_SREG_ES,
163 	VCPU_SREG_CS,
164 	VCPU_SREG_SS,
165 	VCPU_SREG_DS,
166 	VCPU_SREG_FS,
167 	VCPU_SREG_GS,
168 	VCPU_SREG_TR,
169 	VCPU_SREG_LDTR,
170 };
171 
172 #include <asm/kvm_emulate.h>
173 
174 #define KVM_NR_MEM_OBJS 40
175 
176 #define KVM_NR_DB_REGS	4
177 
178 #define DR6_BD		(1 << 13)
179 #define DR6_BS		(1 << 14)
180 #define DR6_RTM		(1 << 16)
181 #define DR6_FIXED_1	0xfffe0ff0
182 #define DR6_INIT	0xffff0ff0
183 #define DR6_VOLATILE	0x0001e00f
184 
185 #define DR7_BP_EN_MASK	0x000000ff
186 #define DR7_GE		(1 << 9)
187 #define DR7_GD		(1 << 13)
188 #define DR7_FIXED_1	0x00000400
189 #define DR7_VOLATILE	0xffff2bff
190 
191 #define PFERR_PRESENT_BIT 0
192 #define PFERR_WRITE_BIT 1
193 #define PFERR_USER_BIT 2
194 #define PFERR_RSVD_BIT 3
195 #define PFERR_FETCH_BIT 4
196 #define PFERR_PK_BIT 5
197 #define PFERR_GUEST_FINAL_BIT 32
198 #define PFERR_GUEST_PAGE_BIT 33
199 
200 #define PFERR_PRESENT_MASK (1U << PFERR_PRESENT_BIT)
201 #define PFERR_WRITE_MASK (1U << PFERR_WRITE_BIT)
202 #define PFERR_USER_MASK (1U << PFERR_USER_BIT)
203 #define PFERR_RSVD_MASK (1U << PFERR_RSVD_BIT)
204 #define PFERR_FETCH_MASK (1U << PFERR_FETCH_BIT)
205 #define PFERR_PK_MASK (1U << PFERR_PK_BIT)
206 #define PFERR_GUEST_FINAL_MASK (1ULL << PFERR_GUEST_FINAL_BIT)
207 #define PFERR_GUEST_PAGE_MASK (1ULL << PFERR_GUEST_PAGE_BIT)
208 
209 #define PFERR_NESTED_GUEST_PAGE (PFERR_GUEST_PAGE_MASK |	\
210 				 PFERR_WRITE_MASK |		\
211 				 PFERR_PRESENT_MASK)
212 
213 /*
214  * The mask used to denote special SPTEs, which can be either MMIO SPTEs or
215  * Access Tracking SPTEs. We use bit 62 instead of bit 63 to avoid conflicting
216  * with the SVE bit in EPT PTEs.
217  */
218 #define SPTE_SPECIAL_MASK (1ULL << 62)
219 
220 /* apic attention bits */
221 #define KVM_APIC_CHECK_VAPIC	0
222 /*
223  * The following bit is set with PV-EOI, unset on EOI.
224  * We detect PV-EOI changes by guest by comparing
225  * this bit with PV-EOI in guest memory.
226  * See the implementation in apic_update_pv_eoi.
227  */
228 #define KVM_APIC_PV_EOI_PENDING	1
229 
230 struct kvm_kernel_irq_routing_entry;
231 
232 /*
233  * We don't want allocation failures within the mmu code, so we preallocate
234  * enough memory for a single page fault in a cache.
235  */
236 struct kvm_mmu_memory_cache {
237 	int nobjs;
238 	void *objects[KVM_NR_MEM_OBJS];
239 };
240 
241 /*
242  * the pages used as guest page table on soft mmu are tracked by
243  * kvm_memory_slot.arch.gfn_track which is 16 bits, so the role bits used
244  * by indirect shadow page can not be more than 15 bits.
245  *
246  * Currently, we used 14 bits that are @level, @cr4_pae, @quadrant, @access,
247  * @nxe, @cr0_wp, @smep_andnot_wp and @smap_andnot_wp.
248  */
249 union kvm_mmu_page_role {
250 	unsigned word;
251 	struct {
252 		unsigned level:4;
253 		unsigned cr4_pae:1;
254 		unsigned quadrant:2;
255 		unsigned direct:1;
256 		unsigned access:3;
257 		unsigned invalid:1;
258 		unsigned nxe:1;
259 		unsigned cr0_wp:1;
260 		unsigned smep_andnot_wp:1;
261 		unsigned smap_andnot_wp:1;
262 		unsigned ad_disabled:1;
263 		unsigned guest_mode:1;
264 		unsigned :6;
265 
266 		/*
267 		 * This is left at the top of the word so that
268 		 * kvm_memslots_for_spte_role can extract it with a
269 		 * simple shift.  While there is room, give it a whole
270 		 * byte so it is also faster to load it from memory.
271 		 */
272 		unsigned smm:8;
273 	};
274 };
275 
276 struct kvm_rmap_head {
277 	unsigned long val;
278 };
279 
280 struct kvm_mmu_page {
281 	struct list_head link;
282 	struct hlist_node hash_link;
283 
284 	/*
285 	 * The following two entries are used to key the shadow page in the
286 	 * hash table.
287 	 */
288 	gfn_t gfn;
289 	union kvm_mmu_page_role role;
290 
291 	u64 *spt;
292 	/* hold the gfn of each spte inside spt */
293 	gfn_t *gfns;
294 	bool unsync;
295 	int root_count;          /* Currently serving as active root */
296 	unsigned int unsync_children;
297 	struct kvm_rmap_head parent_ptes; /* rmap pointers to parent sptes */
298 
299 	/* The page is obsolete if mmu_valid_gen != kvm->arch.mmu_valid_gen.  */
300 	unsigned long mmu_valid_gen;
301 
302 	DECLARE_BITMAP(unsync_child_bitmap, 512);
303 
304 #ifdef CONFIG_X86_32
305 	/*
306 	 * Used out of the mmu-lock to avoid reading spte values while an
307 	 * update is in progress; see the comments in __get_spte_lockless().
308 	 */
309 	int clear_spte_count;
310 #endif
311 
312 	/* Number of writes since the last time traversal visited this page.  */
313 	atomic_t write_flooding_count;
314 };
315 
316 struct kvm_pio_request {
317 	unsigned long count;
318 	int in;
319 	int port;
320 	int size;
321 };
322 
323 #define PT64_ROOT_MAX_LEVEL 5
324 
325 struct rsvd_bits_validate {
326 	u64 rsvd_bits_mask[2][PT64_ROOT_MAX_LEVEL];
327 	u64 bad_mt_xwr;
328 };
329 
330 struct kvm_mmu_root_info {
331 	gpa_t cr3;
332 	hpa_t hpa;
333 };
334 
335 #define KVM_MMU_ROOT_INFO_INVALID \
336 	((struct kvm_mmu_root_info) { .cr3 = INVALID_PAGE, .hpa = INVALID_PAGE })
337 
338 #define KVM_MMU_NUM_PREV_ROOTS 3
339 
340 /*
341  * x86 supports 4 paging modes (5-level 64-bit, 4-level 64-bit, 3-level 32-bit,
342  * and 2-level 32-bit).  The kvm_mmu structure abstracts the details of the
343  * current mmu mode.
344  */
345 struct kvm_mmu {
346 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long root);
347 	unsigned long (*get_cr3)(struct kvm_vcpu *vcpu);
348 	u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
349 	int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err,
350 			  bool prefault);
351 	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
352 				  struct x86_exception *fault);
353 	gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva, u32 access,
354 			    struct x86_exception *exception);
355 	gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
356 			       struct x86_exception *exception);
357 	int (*sync_page)(struct kvm_vcpu *vcpu,
358 			 struct kvm_mmu_page *sp);
359 	void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva, hpa_t root_hpa);
360 	void (*update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
361 			   u64 *spte, const void *pte);
362 	hpa_t root_hpa;
363 	union kvm_mmu_page_role base_role;
364 	u8 root_level;
365 	u8 shadow_root_level;
366 	u8 ept_ad;
367 	bool direct_map;
368 	struct kvm_mmu_root_info prev_roots[KVM_MMU_NUM_PREV_ROOTS];
369 
370 	/*
371 	 * Bitmap; bit set = permission fault
372 	 * Byte index: page fault error code [4:1]
373 	 * Bit index: pte permissions in ACC_* format
374 	 */
375 	u8 permissions[16];
376 
377 	/*
378 	* The pkru_mask indicates if protection key checks are needed.  It
379 	* consists of 16 domains indexed by page fault error code bits [4:1],
380 	* with PFEC.RSVD replaced by ACC_USER_MASK from the page tables.
381 	* Each domain has 2 bits which are ANDed with AD and WD from PKRU.
382 	*/
383 	u32 pkru_mask;
384 
385 	u64 *pae_root;
386 	u64 *lm_root;
387 
388 	/*
389 	 * check zero bits on shadow page table entries, these
390 	 * bits include not only hardware reserved bits but also
391 	 * the bits spte never used.
392 	 */
393 	struct rsvd_bits_validate shadow_zero_check;
394 
395 	struct rsvd_bits_validate guest_rsvd_check;
396 
397 	/* Can have large pages at levels 2..last_nonleaf_level-1. */
398 	u8 last_nonleaf_level;
399 
400 	bool nx;
401 
402 	u64 pdptrs[4]; /* pae */
403 };
404 
405 enum pmc_type {
406 	KVM_PMC_GP = 0,
407 	KVM_PMC_FIXED,
408 };
409 
410 struct kvm_pmc {
411 	enum pmc_type type;
412 	u8 idx;
413 	u64 counter;
414 	u64 eventsel;
415 	struct perf_event *perf_event;
416 	struct kvm_vcpu *vcpu;
417 };
418 
419 struct kvm_pmu {
420 	unsigned nr_arch_gp_counters;
421 	unsigned nr_arch_fixed_counters;
422 	unsigned available_event_types;
423 	u64 fixed_ctr_ctrl;
424 	u64 global_ctrl;
425 	u64 global_status;
426 	u64 global_ovf_ctrl;
427 	u64 counter_bitmask[2];
428 	u64 global_ctrl_mask;
429 	u64 reserved_bits;
430 	u8 version;
431 	struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
432 	struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED];
433 	struct irq_work irq_work;
434 	u64 reprogram_pmi;
435 };
436 
437 struct kvm_pmu_ops;
438 
439 enum {
440 	KVM_DEBUGREG_BP_ENABLED = 1,
441 	KVM_DEBUGREG_WONT_EXIT = 2,
442 	KVM_DEBUGREG_RELOAD = 4,
443 };
444 
445 struct kvm_mtrr_range {
446 	u64 base;
447 	u64 mask;
448 	struct list_head node;
449 };
450 
451 struct kvm_mtrr {
452 	struct kvm_mtrr_range var_ranges[KVM_NR_VAR_MTRR];
453 	mtrr_type fixed_ranges[KVM_NR_FIXED_MTRR_REGION];
454 	u64 deftype;
455 
456 	struct list_head head;
457 };
458 
459 /* Hyper-V SynIC timer */
460 struct kvm_vcpu_hv_stimer {
461 	struct hrtimer timer;
462 	int index;
463 	u64 config;
464 	u64 count;
465 	u64 exp_time;
466 	struct hv_message msg;
467 	bool msg_pending;
468 };
469 
470 /* Hyper-V synthetic interrupt controller (SynIC)*/
471 struct kvm_vcpu_hv_synic {
472 	u64 version;
473 	u64 control;
474 	u64 msg_page;
475 	u64 evt_page;
476 	atomic64_t sint[HV_SYNIC_SINT_COUNT];
477 	atomic_t sint_to_gsi[HV_SYNIC_SINT_COUNT];
478 	DECLARE_BITMAP(auto_eoi_bitmap, 256);
479 	DECLARE_BITMAP(vec_bitmap, 256);
480 	bool active;
481 	bool dont_zero_synic_pages;
482 };
483 
484 /* Hyper-V per vcpu emulation context */
485 struct kvm_vcpu_hv {
486 	u32 vp_index;
487 	u64 hv_vapic;
488 	s64 runtime_offset;
489 	struct kvm_vcpu_hv_synic synic;
490 	struct kvm_hyperv_exit exit;
491 	struct kvm_vcpu_hv_stimer stimer[HV_SYNIC_STIMER_COUNT];
492 	DECLARE_BITMAP(stimer_pending_bitmap, HV_SYNIC_STIMER_COUNT);
493 	cpumask_t tlb_lush;
494 };
495 
496 struct kvm_vcpu_arch {
497 	/*
498 	 * rip and regs accesses must go through
499 	 * kvm_{register,rip}_{read,write} functions.
500 	 */
501 	unsigned long regs[NR_VCPU_REGS];
502 	u32 regs_avail;
503 	u32 regs_dirty;
504 
505 	unsigned long cr0;
506 	unsigned long cr0_guest_owned_bits;
507 	unsigned long cr2;
508 	unsigned long cr3;
509 	unsigned long cr4;
510 	unsigned long cr4_guest_owned_bits;
511 	unsigned long cr8;
512 	u32 pkru;
513 	u32 hflags;
514 	u64 efer;
515 	u64 apic_base;
516 	struct kvm_lapic *apic;    /* kernel irqchip context */
517 	bool apicv_active;
518 	bool load_eoi_exitmap_pending;
519 	DECLARE_BITMAP(ioapic_handled_vectors, 256);
520 	unsigned long apic_attention;
521 	int32_t apic_arb_prio;
522 	int mp_state;
523 	u64 ia32_misc_enable_msr;
524 	u64 smbase;
525 	u64 smi_count;
526 	bool tpr_access_reporting;
527 	u64 ia32_xss;
528 	u64 microcode_version;
529 
530 	/*
531 	 * Paging state of the vcpu
532 	 *
533 	 * If the vcpu runs in guest mode with two level paging this still saves
534 	 * the paging mode of the l1 guest. This context is always used to
535 	 * handle faults.
536 	 */
537 	struct kvm_mmu mmu;
538 
539 	/*
540 	 * Paging state of an L2 guest (used for nested npt)
541 	 *
542 	 * This context will save all necessary information to walk page tables
543 	 * of the an L2 guest. This context is only initialized for page table
544 	 * walking and not for faulting since we never handle l2 page faults on
545 	 * the host.
546 	 */
547 	struct kvm_mmu nested_mmu;
548 
549 	/*
550 	 * Pointer to the mmu context currently used for
551 	 * gva_to_gpa translations.
552 	 */
553 	struct kvm_mmu *walk_mmu;
554 
555 	struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
556 	struct kvm_mmu_memory_cache mmu_page_cache;
557 	struct kvm_mmu_memory_cache mmu_page_header_cache;
558 
559 	/*
560 	 * QEMU userspace and the guest each have their own FPU state.
561 	 * In vcpu_run, we switch between the user and guest FPU contexts.
562 	 * While running a VCPU, the VCPU thread will have the guest FPU
563 	 * context.
564 	 *
565 	 * Note that while the PKRU state lives inside the fpu registers,
566 	 * it is switched out separately at VMENTER and VMEXIT time. The
567 	 * "guest_fpu" state here contains the guest FPU context, with the
568 	 * host PRKU bits.
569 	 */
570 	struct fpu user_fpu;
571 	struct fpu guest_fpu;
572 
573 	u64 xcr0;
574 	u64 guest_supported_xcr0;
575 	u32 guest_xstate_size;
576 
577 	struct kvm_pio_request pio;
578 	void *pio_data;
579 
580 	u8 event_exit_inst_len;
581 
582 	struct kvm_queued_exception {
583 		bool pending;
584 		bool injected;
585 		bool has_error_code;
586 		u8 nr;
587 		u32 error_code;
588 		u8 nested_apf;
589 	} exception;
590 
591 	struct kvm_queued_interrupt {
592 		bool injected;
593 		bool soft;
594 		u8 nr;
595 	} interrupt;
596 
597 	int halt_request; /* real mode on Intel only */
598 
599 	int cpuid_nent;
600 	struct kvm_cpuid_entry2 cpuid_entries[KVM_MAX_CPUID_ENTRIES];
601 
602 	int maxphyaddr;
603 
604 	/* emulate context */
605 
606 	struct x86_emulate_ctxt emulate_ctxt;
607 	bool emulate_regs_need_sync_to_vcpu;
608 	bool emulate_regs_need_sync_from_vcpu;
609 	int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
610 
611 	gpa_t time;
612 	struct pvclock_vcpu_time_info hv_clock;
613 	unsigned int hw_tsc_khz;
614 	struct gfn_to_hva_cache pv_time;
615 	bool pv_time_enabled;
616 	/* set guest stopped flag in pvclock flags field */
617 	bool pvclock_set_guest_stopped_request;
618 
619 	struct {
620 		u64 msr_val;
621 		u64 last_steal;
622 		struct gfn_to_hva_cache stime;
623 		struct kvm_steal_time steal;
624 	} st;
625 
626 	u64 tsc_offset;
627 	u64 last_guest_tsc;
628 	u64 last_host_tsc;
629 	u64 tsc_offset_adjustment;
630 	u64 this_tsc_nsec;
631 	u64 this_tsc_write;
632 	u64 this_tsc_generation;
633 	bool tsc_catchup;
634 	bool tsc_always_catchup;
635 	s8 virtual_tsc_shift;
636 	u32 virtual_tsc_mult;
637 	u32 virtual_tsc_khz;
638 	s64 ia32_tsc_adjust_msr;
639 	u64 tsc_scaling_ratio;
640 
641 	atomic_t nmi_queued;  /* unprocessed asynchronous NMIs */
642 	unsigned nmi_pending; /* NMI queued after currently running handler */
643 	bool nmi_injected;    /* Trying to inject an NMI this entry */
644 	bool smi_pending;    /* SMI queued after currently running handler */
645 
646 	struct kvm_mtrr mtrr_state;
647 	u64 pat;
648 
649 	unsigned switch_db_regs;
650 	unsigned long db[KVM_NR_DB_REGS];
651 	unsigned long dr6;
652 	unsigned long dr7;
653 	unsigned long eff_db[KVM_NR_DB_REGS];
654 	unsigned long guest_debug_dr7;
655 	u64 msr_platform_info;
656 	u64 msr_misc_features_enables;
657 
658 	u64 mcg_cap;
659 	u64 mcg_status;
660 	u64 mcg_ctl;
661 	u64 mcg_ext_ctl;
662 	u64 *mce_banks;
663 
664 	/* Cache MMIO info */
665 	u64 mmio_gva;
666 	unsigned access;
667 	gfn_t mmio_gfn;
668 	u64 mmio_gen;
669 
670 	struct kvm_pmu pmu;
671 
672 	/* used for guest single stepping over the given code position */
673 	unsigned long singlestep_rip;
674 
675 	struct kvm_vcpu_hv hyperv;
676 
677 	cpumask_var_t wbinvd_dirty_mask;
678 
679 	unsigned long last_retry_eip;
680 	unsigned long last_retry_addr;
681 
682 	struct {
683 		bool halted;
684 		gfn_t gfns[roundup_pow_of_two(ASYNC_PF_PER_VCPU)];
685 		struct gfn_to_hva_cache data;
686 		u64 msr_val;
687 		u32 id;
688 		bool send_user_only;
689 		u32 host_apf_reason;
690 		unsigned long nested_apf_token;
691 		bool delivery_as_pf_vmexit;
692 	} apf;
693 
694 	/* OSVW MSRs (AMD only) */
695 	struct {
696 		u64 length;
697 		u64 status;
698 	} osvw;
699 
700 	struct {
701 		u64 msr_val;
702 		struct gfn_to_hva_cache data;
703 	} pv_eoi;
704 
705 	/*
706 	 * Indicate whether the access faults on its page table in guest
707 	 * which is set when fix page fault and used to detect unhandeable
708 	 * instruction.
709 	 */
710 	bool write_fault_to_shadow_pgtable;
711 
712 	/* set at EPT violation at this point */
713 	unsigned long exit_qualification;
714 
715 	/* pv related host specific info */
716 	struct {
717 		bool pv_unhalted;
718 	} pv;
719 
720 	int pending_ioapic_eoi;
721 	int pending_external_vector;
722 
723 	/* GPA available */
724 	bool gpa_available;
725 	gpa_t gpa_val;
726 
727 	/* be preempted when it's in kernel-mode(cpl=0) */
728 	bool preempted_in_kernel;
729 
730 	/* Flush the L1 Data cache for L1TF mitigation on VMENTER */
731 	bool l1tf_flush_l1d;
732 };
733 
734 struct kvm_lpage_info {
735 	int disallow_lpage;
736 };
737 
738 struct kvm_arch_memory_slot {
739 	struct kvm_rmap_head *rmap[KVM_NR_PAGE_SIZES];
740 	struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
741 	unsigned short *gfn_track[KVM_PAGE_TRACK_MAX];
742 };
743 
744 /*
745  * We use as the mode the number of bits allocated in the LDR for the
746  * logical processor ID.  It happens that these are all powers of two.
747  * This makes it is very easy to detect cases where the APICs are
748  * configured for multiple modes; in that case, we cannot use the map and
749  * hence cannot use kvm_irq_delivery_to_apic_fast either.
750  */
751 #define KVM_APIC_MODE_XAPIC_CLUSTER          4
752 #define KVM_APIC_MODE_XAPIC_FLAT             8
753 #define KVM_APIC_MODE_X2APIC                16
754 
755 struct kvm_apic_map {
756 	struct rcu_head rcu;
757 	u8 mode;
758 	u32 max_apic_id;
759 	union {
760 		struct kvm_lapic *xapic_flat_map[8];
761 		struct kvm_lapic *xapic_cluster_map[16][4];
762 	};
763 	struct kvm_lapic *phys_map[];
764 };
765 
766 /* Hyper-V emulation context */
767 struct kvm_hv {
768 	struct mutex hv_lock;
769 	u64 hv_guest_os_id;
770 	u64 hv_hypercall;
771 	u64 hv_tsc_page;
772 
773 	/* Hyper-v based guest crash (NT kernel bugcheck) parameters */
774 	u64 hv_crash_param[HV_X64_MSR_CRASH_PARAMS];
775 	u64 hv_crash_ctl;
776 
777 	HV_REFERENCE_TSC_PAGE tsc_ref;
778 
779 	struct idr conn_to_evt;
780 
781 	u64 hv_reenlightenment_control;
782 	u64 hv_tsc_emulation_control;
783 	u64 hv_tsc_emulation_status;
784 };
785 
786 enum kvm_irqchip_mode {
787 	KVM_IRQCHIP_NONE,
788 	KVM_IRQCHIP_KERNEL,       /* created with KVM_CREATE_IRQCHIP */
789 	KVM_IRQCHIP_SPLIT,        /* created with KVM_CAP_SPLIT_IRQCHIP */
790 };
791 
792 struct kvm_arch {
793 	unsigned int n_used_mmu_pages;
794 	unsigned int n_requested_mmu_pages;
795 	unsigned int n_max_mmu_pages;
796 	unsigned int indirect_shadow_pages;
797 	unsigned long mmu_valid_gen;
798 	struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
799 	/*
800 	 * Hash table of struct kvm_mmu_page.
801 	 */
802 	struct list_head active_mmu_pages;
803 	struct list_head zapped_obsolete_pages;
804 	struct kvm_page_track_notifier_node mmu_sp_tracker;
805 	struct kvm_page_track_notifier_head track_notifier_head;
806 
807 	struct list_head assigned_dev_head;
808 	struct iommu_domain *iommu_domain;
809 	bool iommu_noncoherent;
810 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA
811 	atomic_t noncoherent_dma_count;
812 #define __KVM_HAVE_ARCH_ASSIGNED_DEVICE
813 	atomic_t assigned_device_count;
814 	struct kvm_pic *vpic;
815 	struct kvm_ioapic *vioapic;
816 	struct kvm_pit *vpit;
817 	atomic_t vapics_in_nmi_mode;
818 	struct mutex apic_map_lock;
819 	struct kvm_apic_map *apic_map;
820 
821 	bool apic_access_page_done;
822 
823 	gpa_t wall_clock;
824 
825 	bool mwait_in_guest;
826 	bool hlt_in_guest;
827 	bool pause_in_guest;
828 
829 	unsigned long irq_sources_bitmap;
830 	s64 kvmclock_offset;
831 	raw_spinlock_t tsc_write_lock;
832 	u64 last_tsc_nsec;
833 	u64 last_tsc_write;
834 	u32 last_tsc_khz;
835 	u64 cur_tsc_nsec;
836 	u64 cur_tsc_write;
837 	u64 cur_tsc_offset;
838 	u64 cur_tsc_generation;
839 	int nr_vcpus_matched_tsc;
840 
841 	spinlock_t pvclock_gtod_sync_lock;
842 	bool use_master_clock;
843 	u64 master_kernel_ns;
844 	u64 master_cycle_now;
845 	struct delayed_work kvmclock_update_work;
846 	struct delayed_work kvmclock_sync_work;
847 
848 	struct kvm_xen_hvm_config xen_hvm_config;
849 
850 	/* reads protected by irq_srcu, writes by irq_lock */
851 	struct hlist_head mask_notifier_list;
852 
853 	struct kvm_hv hyperv;
854 
855 	#ifdef CONFIG_KVM_MMU_AUDIT
856 	int audit_point;
857 	#endif
858 
859 	bool backwards_tsc_observed;
860 	bool boot_vcpu_runs_old_kvmclock;
861 	u32 bsp_vcpu_id;
862 
863 	u64 disabled_quirks;
864 
865 	enum kvm_irqchip_mode irqchip_mode;
866 	u8 nr_reserved_ioapic_pins;
867 
868 	bool disabled_lapic_found;
869 
870 	bool x2apic_format;
871 	bool x2apic_broadcast_quirk_disabled;
872 };
873 
874 struct kvm_vm_stat {
875 	ulong mmu_shadow_zapped;
876 	ulong mmu_pte_write;
877 	ulong mmu_pte_updated;
878 	ulong mmu_pde_zapped;
879 	ulong mmu_flooded;
880 	ulong mmu_recycled;
881 	ulong mmu_cache_miss;
882 	ulong mmu_unsync;
883 	ulong remote_tlb_flush;
884 	ulong lpages;
885 	ulong max_mmu_page_hash_collisions;
886 };
887 
888 struct kvm_vcpu_stat {
889 	u64 pf_fixed;
890 	u64 pf_guest;
891 	u64 tlb_flush;
892 	u64 invlpg;
893 
894 	u64 exits;
895 	u64 io_exits;
896 	u64 mmio_exits;
897 	u64 signal_exits;
898 	u64 irq_window_exits;
899 	u64 nmi_window_exits;
900 	u64 l1d_flush;
901 	u64 halt_exits;
902 	u64 halt_successful_poll;
903 	u64 halt_attempted_poll;
904 	u64 halt_poll_invalid;
905 	u64 halt_wakeup;
906 	u64 request_irq_exits;
907 	u64 irq_exits;
908 	u64 host_state_reload;
909 	u64 fpu_reload;
910 	u64 insn_emulation;
911 	u64 insn_emulation_fail;
912 	u64 hypercalls;
913 	u64 irq_injections;
914 	u64 nmi_injections;
915 	u64 req_event;
916 };
917 
918 struct x86_instruction_info;
919 
920 struct msr_data {
921 	bool host_initiated;
922 	u32 index;
923 	u64 data;
924 };
925 
926 struct kvm_lapic_irq {
927 	u32 vector;
928 	u16 delivery_mode;
929 	u16 dest_mode;
930 	bool level;
931 	u16 trig_mode;
932 	u32 shorthand;
933 	u32 dest_id;
934 	bool msi_redir_hint;
935 };
936 
937 struct kvm_x86_ops {
938 	int (*cpu_has_kvm_support)(void);          /* __init */
939 	int (*disabled_by_bios)(void);             /* __init */
940 	int (*hardware_enable)(void);
941 	void (*hardware_disable)(void);
942 	void (*check_processor_compatibility)(void *rtn);
943 	int (*hardware_setup)(void);               /* __init */
944 	void (*hardware_unsetup)(void);            /* __exit */
945 	bool (*cpu_has_accelerated_tpr)(void);
946 	bool (*has_emulated_msr)(int index);
947 	void (*cpuid_update)(struct kvm_vcpu *vcpu);
948 
949 	struct kvm *(*vm_alloc)(void);
950 	void (*vm_free)(struct kvm *);
951 	int (*vm_init)(struct kvm *kvm);
952 	void (*vm_destroy)(struct kvm *kvm);
953 
954 	/* Create, but do not attach this VCPU */
955 	struct kvm_vcpu *(*vcpu_create)(struct kvm *kvm, unsigned id);
956 	void (*vcpu_free)(struct kvm_vcpu *vcpu);
957 	void (*vcpu_reset)(struct kvm_vcpu *vcpu, bool init_event);
958 
959 	void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
960 	void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
961 	void (*vcpu_put)(struct kvm_vcpu *vcpu);
962 
963 	void (*update_bp_intercept)(struct kvm_vcpu *vcpu);
964 	int (*get_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
965 	int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
966 	u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
967 	void (*get_segment)(struct kvm_vcpu *vcpu,
968 			    struct kvm_segment *var, int seg);
969 	int (*get_cpl)(struct kvm_vcpu *vcpu);
970 	void (*set_segment)(struct kvm_vcpu *vcpu,
971 			    struct kvm_segment *var, int seg);
972 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
973 	void (*decache_cr0_guest_bits)(struct kvm_vcpu *vcpu);
974 	void (*decache_cr3)(struct kvm_vcpu *vcpu);
975 	void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu);
976 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
977 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
978 	int (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
979 	void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
980 	void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
981 	void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
982 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
983 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
984 	u64 (*get_dr6)(struct kvm_vcpu *vcpu);
985 	void (*set_dr6)(struct kvm_vcpu *vcpu, unsigned long value);
986 	void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu);
987 	void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
988 	void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
989 	unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
990 	void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
991 
992 	void (*tlb_flush)(struct kvm_vcpu *vcpu, bool invalidate_gpa);
993 	int  (*tlb_remote_flush)(struct kvm *kvm);
994 
995 	/*
996 	 * Flush any TLB entries associated with the given GVA.
997 	 * Does not need to flush GPA->HPA mappings.
998 	 * Can potentially get non-canonical addresses through INVLPGs, which
999 	 * the implementation may choose to ignore if appropriate.
1000 	 */
1001 	void (*tlb_flush_gva)(struct kvm_vcpu *vcpu, gva_t addr);
1002 
1003 	void (*run)(struct kvm_vcpu *vcpu);
1004 	int (*handle_exit)(struct kvm_vcpu *vcpu);
1005 	void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
1006 	void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
1007 	u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu);
1008 	void (*patch_hypercall)(struct kvm_vcpu *vcpu,
1009 				unsigned char *hypercall_addr);
1010 	void (*set_irq)(struct kvm_vcpu *vcpu);
1011 	void (*set_nmi)(struct kvm_vcpu *vcpu);
1012 	void (*queue_exception)(struct kvm_vcpu *vcpu);
1013 	void (*cancel_injection)(struct kvm_vcpu *vcpu);
1014 	int (*interrupt_allowed)(struct kvm_vcpu *vcpu);
1015 	int (*nmi_allowed)(struct kvm_vcpu *vcpu);
1016 	bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
1017 	void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
1018 	void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
1019 	void (*enable_irq_window)(struct kvm_vcpu *vcpu);
1020 	void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
1021 	bool (*get_enable_apicv)(struct kvm_vcpu *vcpu);
1022 	void (*refresh_apicv_exec_ctrl)(struct kvm_vcpu *vcpu);
1023 	void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
1024 	void (*hwapic_isr_update)(struct kvm_vcpu *vcpu, int isr);
1025 	void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
1026 	void (*set_virtual_apic_mode)(struct kvm_vcpu *vcpu);
1027 	void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu, hpa_t hpa);
1028 	void (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
1029 	int (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
1030 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
1031 	int (*set_identity_map_addr)(struct kvm *kvm, u64 ident_addr);
1032 	int (*get_tdp_level)(struct kvm_vcpu *vcpu);
1033 	u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
1034 	int (*get_lpage_level)(void);
1035 	bool (*rdtscp_supported)(void);
1036 	bool (*invpcid_supported)(void);
1037 
1038 	void (*set_tdp_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
1039 
1040 	void (*set_supported_cpuid)(u32 func, struct kvm_cpuid_entry2 *entry);
1041 
1042 	bool (*has_wbinvd_exit)(void);
1043 
1044 	u64 (*read_l1_tsc_offset)(struct kvm_vcpu *vcpu);
1045 	void (*write_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
1046 
1047 	void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2);
1048 
1049 	int (*check_intercept)(struct kvm_vcpu *vcpu,
1050 			       struct x86_instruction_info *info,
1051 			       enum x86_intercept_stage stage);
1052 	void (*handle_external_intr)(struct kvm_vcpu *vcpu);
1053 	bool (*mpx_supported)(void);
1054 	bool (*xsaves_supported)(void);
1055 	bool (*umip_emulated)(void);
1056 
1057 	int (*check_nested_events)(struct kvm_vcpu *vcpu, bool external_intr);
1058 
1059 	void (*sched_in)(struct kvm_vcpu *kvm, int cpu);
1060 
1061 	/*
1062 	 * Arch-specific dirty logging hooks. These hooks are only supposed to
1063 	 * be valid if the specific arch has hardware-accelerated dirty logging
1064 	 * mechanism. Currently only for PML on VMX.
1065 	 *
1066 	 *  - slot_enable_log_dirty:
1067 	 *	called when enabling log dirty mode for the slot.
1068 	 *  - slot_disable_log_dirty:
1069 	 *	called when disabling log dirty mode for the slot.
1070 	 *	also called when slot is created with log dirty disabled.
1071 	 *  - flush_log_dirty:
1072 	 *	called before reporting dirty_bitmap to userspace.
1073 	 *  - enable_log_dirty_pt_masked:
1074 	 *	called when reenabling log dirty for the GFNs in the mask after
1075 	 *	corresponding bits are cleared in slot->dirty_bitmap.
1076 	 */
1077 	void (*slot_enable_log_dirty)(struct kvm *kvm,
1078 				      struct kvm_memory_slot *slot);
1079 	void (*slot_disable_log_dirty)(struct kvm *kvm,
1080 				       struct kvm_memory_slot *slot);
1081 	void (*flush_log_dirty)(struct kvm *kvm);
1082 	void (*enable_log_dirty_pt_masked)(struct kvm *kvm,
1083 					   struct kvm_memory_slot *slot,
1084 					   gfn_t offset, unsigned long mask);
1085 	int (*write_log_dirty)(struct kvm_vcpu *vcpu);
1086 
1087 	/* pmu operations of sub-arch */
1088 	const struct kvm_pmu_ops *pmu_ops;
1089 
1090 	/*
1091 	 * Architecture specific hooks for vCPU blocking due to
1092 	 * HLT instruction.
1093 	 * Returns for .pre_block():
1094 	 *    - 0 means continue to block the vCPU.
1095 	 *    - 1 means we cannot block the vCPU since some event
1096 	 *        happens during this period, such as, 'ON' bit in
1097 	 *        posted-interrupts descriptor is set.
1098 	 */
1099 	int (*pre_block)(struct kvm_vcpu *vcpu);
1100 	void (*post_block)(struct kvm_vcpu *vcpu);
1101 
1102 	void (*vcpu_blocking)(struct kvm_vcpu *vcpu);
1103 	void (*vcpu_unblocking)(struct kvm_vcpu *vcpu);
1104 
1105 	int (*update_pi_irte)(struct kvm *kvm, unsigned int host_irq,
1106 			      uint32_t guest_irq, bool set);
1107 	void (*apicv_post_state_restore)(struct kvm_vcpu *vcpu);
1108 
1109 	int (*set_hv_timer)(struct kvm_vcpu *vcpu, u64 guest_deadline_tsc);
1110 	void (*cancel_hv_timer)(struct kvm_vcpu *vcpu);
1111 
1112 	void (*setup_mce)(struct kvm_vcpu *vcpu);
1113 
1114 	int (*get_nested_state)(struct kvm_vcpu *vcpu,
1115 				struct kvm_nested_state __user *user_kvm_nested_state,
1116 				unsigned user_data_size);
1117 	int (*set_nested_state)(struct kvm_vcpu *vcpu,
1118 				struct kvm_nested_state __user *user_kvm_nested_state,
1119 				struct kvm_nested_state *kvm_state);
1120 	void (*get_vmcs12_pages)(struct kvm_vcpu *vcpu);
1121 
1122 	int (*smi_allowed)(struct kvm_vcpu *vcpu);
1123 	int (*pre_enter_smm)(struct kvm_vcpu *vcpu, char *smstate);
1124 	int (*pre_leave_smm)(struct kvm_vcpu *vcpu, u64 smbase);
1125 	int (*enable_smi_window)(struct kvm_vcpu *vcpu);
1126 
1127 	int (*mem_enc_op)(struct kvm *kvm, void __user *argp);
1128 	int (*mem_enc_reg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1129 	int (*mem_enc_unreg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1130 
1131 	int (*get_msr_feature)(struct kvm_msr_entry *entry);
1132 };
1133 
1134 struct kvm_arch_async_pf {
1135 	u32 token;
1136 	gfn_t gfn;
1137 	unsigned long cr3;
1138 	bool direct_map;
1139 };
1140 
1141 extern struct kvm_x86_ops *kvm_x86_ops;
1142 
1143 #define __KVM_HAVE_ARCH_VM_ALLOC
1144 static inline struct kvm *kvm_arch_alloc_vm(void)
1145 {
1146 	return kvm_x86_ops->vm_alloc();
1147 }
1148 
1149 static inline void kvm_arch_free_vm(struct kvm *kvm)
1150 {
1151 	return kvm_x86_ops->vm_free(kvm);
1152 }
1153 
1154 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
1155 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
1156 {
1157 	if (kvm_x86_ops->tlb_remote_flush &&
1158 	    !kvm_x86_ops->tlb_remote_flush(kvm))
1159 		return 0;
1160 	else
1161 		return -ENOTSUPP;
1162 }
1163 
1164 int kvm_mmu_module_init(void);
1165 void kvm_mmu_module_exit(void);
1166 
1167 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
1168 int kvm_mmu_create(struct kvm_vcpu *vcpu);
1169 void kvm_mmu_setup(struct kvm_vcpu *vcpu);
1170 void kvm_mmu_init_vm(struct kvm *kvm);
1171 void kvm_mmu_uninit_vm(struct kvm *kvm);
1172 void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
1173 		u64 dirty_mask, u64 nx_mask, u64 x_mask, u64 p_mask,
1174 		u64 acc_track_mask, u64 me_mask);
1175 
1176 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
1177 void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
1178 				      struct kvm_memory_slot *memslot);
1179 void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
1180 				   const struct kvm_memory_slot *memslot);
1181 void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
1182 				   struct kvm_memory_slot *memslot);
1183 void kvm_mmu_slot_largepage_remove_write_access(struct kvm *kvm,
1184 					struct kvm_memory_slot *memslot);
1185 void kvm_mmu_slot_set_dirty(struct kvm *kvm,
1186 			    struct kvm_memory_slot *memslot);
1187 void kvm_mmu_clear_dirty_pt_masked(struct kvm *kvm,
1188 				   struct kvm_memory_slot *slot,
1189 				   gfn_t gfn_offset, unsigned long mask);
1190 void kvm_mmu_zap_all(struct kvm *kvm);
1191 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, struct kvm_memslots *slots);
1192 unsigned int kvm_mmu_calculate_mmu_pages(struct kvm *kvm);
1193 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int kvm_nr_mmu_pages);
1194 
1195 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
1196 bool pdptrs_changed(struct kvm_vcpu *vcpu);
1197 
1198 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
1199 			  const void *val, int bytes);
1200 
1201 struct kvm_irq_mask_notifier {
1202 	void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
1203 	int irq;
1204 	struct hlist_node link;
1205 };
1206 
1207 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
1208 				    struct kvm_irq_mask_notifier *kimn);
1209 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
1210 				      struct kvm_irq_mask_notifier *kimn);
1211 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
1212 			     bool mask);
1213 
1214 extern bool tdp_enabled;
1215 
1216 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
1217 
1218 /* control of guest tsc rate supported? */
1219 extern bool kvm_has_tsc_control;
1220 /* maximum supported tsc_khz for guests */
1221 extern u32  kvm_max_guest_tsc_khz;
1222 /* number of bits of the fractional part of the TSC scaling ratio */
1223 extern u8   kvm_tsc_scaling_ratio_frac_bits;
1224 /* maximum allowed value of TSC scaling ratio */
1225 extern u64  kvm_max_tsc_scaling_ratio;
1226 /* 1ull << kvm_tsc_scaling_ratio_frac_bits */
1227 extern u64  kvm_default_tsc_scaling_ratio;
1228 
1229 extern u64 kvm_mce_cap_supported;
1230 
1231 enum emulation_result {
1232 	EMULATE_DONE,         /* no further processing */
1233 	EMULATE_USER_EXIT,    /* kvm_run ready for userspace exit */
1234 	EMULATE_FAIL,         /* can't emulate this instruction */
1235 };
1236 
1237 #define EMULTYPE_NO_DECODE	    (1 << 0)
1238 #define EMULTYPE_TRAP_UD	    (1 << 1)
1239 #define EMULTYPE_SKIP		    (1 << 2)
1240 #define EMULTYPE_RETRY		    (1 << 3)
1241 #define EMULTYPE_NO_REEXECUTE	    (1 << 4)
1242 #define EMULTYPE_NO_UD_ON_FAIL	    (1 << 5)
1243 #define EMULTYPE_VMWARE		    (1 << 6)
1244 int x86_emulate_instruction(struct kvm_vcpu *vcpu, unsigned long cr2,
1245 			    int emulation_type, void *insn, int insn_len);
1246 
1247 static inline int emulate_instruction(struct kvm_vcpu *vcpu,
1248 			int emulation_type)
1249 {
1250 	return x86_emulate_instruction(vcpu, 0,
1251 			emulation_type | EMULTYPE_NO_REEXECUTE, NULL, 0);
1252 }
1253 
1254 void kvm_enable_efer_bits(u64);
1255 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
1256 int kvm_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr);
1257 int kvm_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr);
1258 
1259 struct x86_emulate_ctxt;
1260 
1261 int kvm_fast_pio(struct kvm_vcpu *vcpu, int size, unsigned short port, int in);
1262 int kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
1263 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
1264 int kvm_vcpu_halt(struct kvm_vcpu *vcpu);
1265 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
1266 
1267 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
1268 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
1269 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector);
1270 
1271 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
1272 		    int reason, bool has_error_code, u32 error_code);
1273 
1274 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
1275 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
1276 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
1277 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
1278 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
1279 int kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
1280 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
1281 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
1282 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
1283 int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr);
1284 
1285 int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1286 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1287 
1288 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
1289 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
1290 bool kvm_rdpmc(struct kvm_vcpu *vcpu);
1291 
1292 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1293 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1294 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1295 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1296 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
1297 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1298 			    gfn_t gfn, void *data, int offset, int len,
1299 			    u32 access);
1300 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
1301 bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr);
1302 
1303 static inline int __kvm_irq_line_state(unsigned long *irq_state,
1304 				       int irq_source_id, int level)
1305 {
1306 	/* Logical OR for level trig interrupt */
1307 	if (level)
1308 		__set_bit(irq_source_id, irq_state);
1309 	else
1310 		__clear_bit(irq_source_id, irq_state);
1311 
1312 	return !!(*irq_state);
1313 }
1314 
1315 #define KVM_MMU_ROOT_CURRENT		BIT(0)
1316 #define KVM_MMU_ROOT_PREVIOUS(i)	BIT(1+i)
1317 #define KVM_MMU_ROOTS_ALL		(~0UL)
1318 
1319 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
1320 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
1321 
1322 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
1323 
1324 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
1325 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
1326 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
1327 int kvm_mmu_load(struct kvm_vcpu *vcpu);
1328 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
1329 void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu);
1330 void kvm_mmu_free_roots(struct kvm_vcpu *vcpu, ulong roots_to_free);
1331 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1332 			   struct x86_exception *exception);
1333 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
1334 			      struct x86_exception *exception);
1335 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
1336 			       struct x86_exception *exception);
1337 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
1338 			       struct x86_exception *exception);
1339 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
1340 				struct x86_exception *exception);
1341 
1342 void kvm_vcpu_deactivate_apicv(struct kvm_vcpu *vcpu);
1343 
1344 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
1345 
1346 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva, u64 error_code,
1347 		       void *insn, int insn_len);
1348 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
1349 void kvm_mmu_invpcid_gva(struct kvm_vcpu *vcpu, gva_t gva, unsigned long pcid);
1350 void kvm_mmu_new_cr3(struct kvm_vcpu *vcpu, gpa_t new_cr3, bool skip_tlb_flush);
1351 
1352 void kvm_enable_tdp(void);
1353 void kvm_disable_tdp(void);
1354 
1355 static inline gpa_t translate_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1356 				  struct x86_exception *exception)
1357 {
1358 	return gpa;
1359 }
1360 
1361 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
1362 {
1363 	struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
1364 
1365 	return (struct kvm_mmu_page *)page_private(page);
1366 }
1367 
1368 static inline u16 kvm_read_ldt(void)
1369 {
1370 	u16 ldt;
1371 	asm("sldt %0" : "=g"(ldt));
1372 	return ldt;
1373 }
1374 
1375 static inline void kvm_load_ldt(u16 sel)
1376 {
1377 	asm("lldt %0" : : "rm"(sel));
1378 }
1379 
1380 #ifdef CONFIG_X86_64
1381 static inline unsigned long read_msr(unsigned long msr)
1382 {
1383 	u64 value;
1384 
1385 	rdmsrl(msr, value);
1386 	return value;
1387 }
1388 #endif
1389 
1390 static inline u32 get_rdx_init_val(void)
1391 {
1392 	return 0x600; /* P6 family */
1393 }
1394 
1395 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
1396 {
1397 	kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
1398 }
1399 
1400 #define TSS_IOPB_BASE_OFFSET 0x66
1401 #define TSS_BASE_SIZE 0x68
1402 #define TSS_IOPB_SIZE (65536 / 8)
1403 #define TSS_REDIRECTION_SIZE (256 / 8)
1404 #define RMODE_TSS_SIZE							\
1405 	(TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
1406 
1407 enum {
1408 	TASK_SWITCH_CALL = 0,
1409 	TASK_SWITCH_IRET = 1,
1410 	TASK_SWITCH_JMP = 2,
1411 	TASK_SWITCH_GATE = 3,
1412 };
1413 
1414 #define HF_GIF_MASK		(1 << 0)
1415 #define HF_HIF_MASK		(1 << 1)
1416 #define HF_VINTR_MASK		(1 << 2)
1417 #define HF_NMI_MASK		(1 << 3)
1418 #define HF_IRET_MASK		(1 << 4)
1419 #define HF_GUEST_MASK		(1 << 5) /* VCPU is in guest-mode */
1420 #define HF_SMM_MASK		(1 << 6)
1421 #define HF_SMM_INSIDE_NMI_MASK	(1 << 7)
1422 
1423 #define __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
1424 #define KVM_ADDRESS_SPACE_NUM 2
1425 
1426 #define kvm_arch_vcpu_memslots_id(vcpu) ((vcpu)->arch.hflags & HF_SMM_MASK ? 1 : 0)
1427 #define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, (role).smm)
1428 
1429 /*
1430  * Hardware virtualization extension instructions may fault if a
1431  * reboot turns off virtualization while processes are running.
1432  * Trap the fault and ignore the instruction if that happens.
1433  */
1434 asmlinkage void kvm_spurious_fault(void);
1435 
1436 #define ____kvm_handle_fault_on_reboot(insn, cleanup_insn)	\
1437 	"666: " insn "\n\t" \
1438 	"668: \n\t"                           \
1439 	".pushsection .fixup, \"ax\" \n" \
1440 	"667: \n\t" \
1441 	cleanup_insn "\n\t"		      \
1442 	"cmpb $0, kvm_rebooting \n\t"	      \
1443 	"jne 668b \n\t"      		      \
1444 	__ASM_SIZE(push) " $666b \n\t"	      \
1445 	"call kvm_spurious_fault \n\t"	      \
1446 	".popsection \n\t" \
1447 	_ASM_EXTABLE(666b, 667b)
1448 
1449 #define __kvm_handle_fault_on_reboot(insn)		\
1450 	____kvm_handle_fault_on_reboot(insn, "")
1451 
1452 #define KVM_ARCH_WANT_MMU_NOTIFIER
1453 int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
1454 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end);
1455 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
1456 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
1457 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
1458 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1459 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1460 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1461 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1462 void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event);
1463 void kvm_vcpu_reload_apic_access_page(struct kvm_vcpu *vcpu);
1464 
1465 int kvm_pv_send_ipi(struct kvm *kvm, unsigned long ipi_bitmap_low,
1466     		    unsigned long ipi_bitmap_high, int min,
1467 		    unsigned long icr, int op_64_bit);
1468 
1469 u64 kvm_get_arch_capabilities(void);
1470 void kvm_define_shared_msr(unsigned index, u32 msr);
1471 int kvm_set_shared_msr(unsigned index, u64 val, u64 mask);
1472 
1473 u64 kvm_scale_tsc(struct kvm_vcpu *vcpu, u64 tsc);
1474 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc);
1475 
1476 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu);
1477 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1478 
1479 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
1480 void kvm_make_scan_ioapic_request(struct kvm *kvm);
1481 
1482 void kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1483 				     struct kvm_async_pf *work);
1484 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1485 				 struct kvm_async_pf *work);
1486 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1487 			       struct kvm_async_pf *work);
1488 bool kvm_arch_can_inject_async_page_present(struct kvm_vcpu *vcpu);
1489 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1490 
1491 int kvm_skip_emulated_instruction(struct kvm_vcpu *vcpu);
1492 int kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1493 
1494 int kvm_is_in_guest(void);
1495 
1496 int __x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa, u32 size);
1497 int x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa, u32 size);
1498 bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu);
1499 bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu);
1500 
1501 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
1502 			     struct kvm_vcpu **dest_vcpu);
1503 
1504 void kvm_set_msi_irq(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e,
1505 		     struct kvm_lapic_irq *irq);
1506 
1507 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu)
1508 {
1509 	if (kvm_x86_ops->vcpu_blocking)
1510 		kvm_x86_ops->vcpu_blocking(vcpu);
1511 }
1512 
1513 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu)
1514 {
1515 	if (kvm_x86_ops->vcpu_unblocking)
1516 		kvm_x86_ops->vcpu_unblocking(vcpu);
1517 }
1518 
1519 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
1520 
1521 static inline int kvm_cpu_get_apicid(int mps_cpu)
1522 {
1523 #ifdef CONFIG_X86_LOCAL_APIC
1524 	return default_cpu_present_to_apicid(mps_cpu);
1525 #else
1526 	WARN_ON_ONCE(1);
1527 	return BAD_APICID;
1528 #endif
1529 }
1530 
1531 #define put_smstate(type, buf, offset, val)                      \
1532 	*(type *)((buf) + (offset) - 0x7e00) = val
1533 
1534 #endif /* _ASM_X86_KVM_HOST_H */
1535