xref: /openbmc/linux/arch/x86/include/asm/kvm_host.h (revision 9977a8c3497a8f7f7f951994f298a8e4d961234f)
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 #include <linux/irqbypass.h>
28 #include <linux/hyperv.h>
29 
30 #include <asm/apic.h>
31 #include <asm/pvclock-abi.h>
32 #include <asm/desc.h>
33 #include <asm/mtrr.h>
34 #include <asm/msr-index.h>
35 #include <asm/asm.h>
36 #include <asm/kvm_page_track.h>
37 
38 #define KVM_MAX_VCPUS 288
39 #define KVM_SOFT_MAX_VCPUS 240
40 #define KVM_MAX_VCPU_ID 1023
41 #define KVM_USER_MEM_SLOTS 509
42 /* memory slots that are not exposed to userspace */
43 #define KVM_PRIVATE_MEM_SLOTS 3
44 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
45 
46 #define KVM_HALT_POLL_NS_DEFAULT 200000
47 
48 #define KVM_IRQCHIP_NUM_PINS  KVM_IOAPIC_NUM_PINS
49 
50 /* x86-specific vcpu->requests bit members */
51 #define KVM_REQ_MIGRATE_TIMER		KVM_ARCH_REQ(0)
52 #define KVM_REQ_REPORT_TPR_ACCESS	KVM_ARCH_REQ(1)
53 #define KVM_REQ_TRIPLE_FAULT		KVM_ARCH_REQ(2)
54 #define KVM_REQ_MMU_SYNC		KVM_ARCH_REQ(3)
55 #define KVM_REQ_CLOCK_UPDATE		KVM_ARCH_REQ(4)
56 #define KVM_REQ_EVENT			KVM_ARCH_REQ(6)
57 #define KVM_REQ_APF_HALT		KVM_ARCH_REQ(7)
58 #define KVM_REQ_STEAL_UPDATE		KVM_ARCH_REQ(8)
59 #define KVM_REQ_NMI			KVM_ARCH_REQ(9)
60 #define KVM_REQ_PMU			KVM_ARCH_REQ(10)
61 #define KVM_REQ_PMI			KVM_ARCH_REQ(11)
62 #define KVM_REQ_SMI			KVM_ARCH_REQ(12)
63 #define KVM_REQ_MASTERCLOCK_UPDATE	KVM_ARCH_REQ(13)
64 #define KVM_REQ_MCLOCK_INPROGRESS \
65 	KVM_ARCH_REQ_FLAGS(14, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
66 #define KVM_REQ_SCAN_IOAPIC \
67 	KVM_ARCH_REQ_FLAGS(15, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
68 #define KVM_REQ_GLOBAL_CLOCK_UPDATE	KVM_ARCH_REQ(16)
69 #define KVM_REQ_APIC_PAGE_RELOAD \
70 	KVM_ARCH_REQ_FLAGS(17, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
71 #define KVM_REQ_HV_CRASH		KVM_ARCH_REQ(18)
72 #define KVM_REQ_IOAPIC_EOI_EXIT		KVM_ARCH_REQ(19)
73 #define KVM_REQ_HV_RESET		KVM_ARCH_REQ(20)
74 #define KVM_REQ_HV_EXIT			KVM_ARCH_REQ(21)
75 #define KVM_REQ_HV_STIMER		KVM_ARCH_REQ(22)
76 
77 #define CR0_RESERVED_BITS                                               \
78 	(~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \
79 			  | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \
80 			  | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG))
81 
82 #define CR3_PCID_INVD		 BIT_64(63)
83 #define CR4_RESERVED_BITS                                               \
84 	(~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\
85 			  | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE     \
86 			  | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR | X86_CR4_PCIDE \
87 			  | X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_FSGSBASE \
88 			  | X86_CR4_OSXMMEXCPT | X86_CR4_LA57 | X86_CR4_VMXE \
89 			  | X86_CR4_SMAP | X86_CR4_PKE | X86_CR4_UMIP))
90 
91 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR)
92 
93 
94 
95 #define INVALID_PAGE (~(hpa_t)0)
96 #define VALID_PAGE(x) ((x) != INVALID_PAGE)
97 
98 #define UNMAPPED_GVA (~(gpa_t)0)
99 
100 /* KVM Hugepage definitions for x86 */
101 #define KVM_NR_PAGE_SIZES	3
102 #define KVM_HPAGE_GFN_SHIFT(x)	(((x) - 1) * 9)
103 #define KVM_HPAGE_SHIFT(x)	(PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
104 #define KVM_HPAGE_SIZE(x)	(1UL << KVM_HPAGE_SHIFT(x))
105 #define KVM_HPAGE_MASK(x)	(~(KVM_HPAGE_SIZE(x) - 1))
106 #define KVM_PAGES_PER_HPAGE(x)	(KVM_HPAGE_SIZE(x) / PAGE_SIZE)
107 
108 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
109 {
110 	/* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */
111 	return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
112 		(base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
113 }
114 
115 #define KVM_PERMILLE_MMU_PAGES 20
116 #define KVM_MIN_ALLOC_MMU_PAGES 64
117 #define KVM_MMU_HASH_SHIFT 12
118 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
119 #define KVM_MIN_FREE_MMU_PAGES 5
120 #define KVM_REFILL_PAGES 25
121 #define KVM_MAX_CPUID_ENTRIES 80
122 #define KVM_NR_FIXED_MTRR_REGION 88
123 #define KVM_NR_VAR_MTRR 8
124 
125 #define ASYNC_PF_PER_VCPU 64
126 
127 enum kvm_reg {
128 	VCPU_REGS_RAX = 0,
129 	VCPU_REGS_RCX = 1,
130 	VCPU_REGS_RDX = 2,
131 	VCPU_REGS_RBX = 3,
132 	VCPU_REGS_RSP = 4,
133 	VCPU_REGS_RBP = 5,
134 	VCPU_REGS_RSI = 6,
135 	VCPU_REGS_RDI = 7,
136 #ifdef CONFIG_X86_64
137 	VCPU_REGS_R8 = 8,
138 	VCPU_REGS_R9 = 9,
139 	VCPU_REGS_R10 = 10,
140 	VCPU_REGS_R11 = 11,
141 	VCPU_REGS_R12 = 12,
142 	VCPU_REGS_R13 = 13,
143 	VCPU_REGS_R14 = 14,
144 	VCPU_REGS_R15 = 15,
145 #endif
146 	VCPU_REGS_RIP,
147 	NR_VCPU_REGS
148 };
149 
150 enum kvm_reg_ex {
151 	VCPU_EXREG_PDPTR = NR_VCPU_REGS,
152 	VCPU_EXREG_CR3,
153 	VCPU_EXREG_RFLAGS,
154 	VCPU_EXREG_SEGMENTS,
155 };
156 
157 enum {
158 	VCPU_SREG_ES,
159 	VCPU_SREG_CS,
160 	VCPU_SREG_SS,
161 	VCPU_SREG_DS,
162 	VCPU_SREG_FS,
163 	VCPU_SREG_GS,
164 	VCPU_SREG_TR,
165 	VCPU_SREG_LDTR,
166 };
167 
168 #include <asm/kvm_emulate.h>
169 
170 #define KVM_NR_MEM_OBJS 40
171 
172 #define KVM_NR_DB_REGS	4
173 
174 #define DR6_BD		(1 << 13)
175 #define DR6_BS		(1 << 14)
176 #define DR6_RTM		(1 << 16)
177 #define DR6_FIXED_1	0xfffe0ff0
178 #define DR6_INIT	0xffff0ff0
179 #define DR6_VOLATILE	0x0001e00f
180 
181 #define DR7_BP_EN_MASK	0x000000ff
182 #define DR7_GE		(1 << 9)
183 #define DR7_GD		(1 << 13)
184 #define DR7_FIXED_1	0x00000400
185 #define DR7_VOLATILE	0xffff2bff
186 
187 #define PFERR_PRESENT_BIT 0
188 #define PFERR_WRITE_BIT 1
189 #define PFERR_USER_BIT 2
190 #define PFERR_RSVD_BIT 3
191 #define PFERR_FETCH_BIT 4
192 #define PFERR_PK_BIT 5
193 #define PFERR_GUEST_FINAL_BIT 32
194 #define PFERR_GUEST_PAGE_BIT 33
195 
196 #define PFERR_PRESENT_MASK (1U << PFERR_PRESENT_BIT)
197 #define PFERR_WRITE_MASK (1U << PFERR_WRITE_BIT)
198 #define PFERR_USER_MASK (1U << PFERR_USER_BIT)
199 #define PFERR_RSVD_MASK (1U << PFERR_RSVD_BIT)
200 #define PFERR_FETCH_MASK (1U << PFERR_FETCH_BIT)
201 #define PFERR_PK_MASK (1U << PFERR_PK_BIT)
202 #define PFERR_GUEST_FINAL_MASK (1ULL << PFERR_GUEST_FINAL_BIT)
203 #define PFERR_GUEST_PAGE_MASK (1ULL << PFERR_GUEST_PAGE_BIT)
204 
205 #define PFERR_NESTED_GUEST_PAGE (PFERR_GUEST_PAGE_MASK |	\
206 				 PFERR_WRITE_MASK |		\
207 				 PFERR_PRESENT_MASK)
208 
209 /*
210  * The mask used to denote special SPTEs, which can be either MMIO SPTEs or
211  * Access Tracking SPTEs. We use bit 62 instead of bit 63 to avoid conflicting
212  * with the SVE bit in EPT PTEs.
213  */
214 #define SPTE_SPECIAL_MASK (1ULL << 62)
215 
216 /* apic attention bits */
217 #define KVM_APIC_CHECK_VAPIC	0
218 /*
219  * The following bit is set with PV-EOI, unset on EOI.
220  * We detect PV-EOI changes by guest by comparing
221  * this bit with PV-EOI in guest memory.
222  * See the implementation in apic_update_pv_eoi.
223  */
224 #define KVM_APIC_PV_EOI_PENDING	1
225 
226 struct kvm_kernel_irq_routing_entry;
227 
228 /*
229  * We don't want allocation failures within the mmu code, so we preallocate
230  * enough memory for a single page fault in a cache.
231  */
232 struct kvm_mmu_memory_cache {
233 	int nobjs;
234 	void *objects[KVM_NR_MEM_OBJS];
235 };
236 
237 /*
238  * the pages used as guest page table on soft mmu are tracked by
239  * kvm_memory_slot.arch.gfn_track which is 16 bits, so the role bits used
240  * by indirect shadow page can not be more than 15 bits.
241  *
242  * Currently, we used 14 bits that are @level, @cr4_pae, @quadrant, @access,
243  * @nxe, @cr0_wp, @smep_andnot_wp and @smap_andnot_wp.
244  */
245 union kvm_mmu_page_role {
246 	unsigned word;
247 	struct {
248 		unsigned level:4;
249 		unsigned cr4_pae:1;
250 		unsigned quadrant:2;
251 		unsigned direct:1;
252 		unsigned access:3;
253 		unsigned invalid:1;
254 		unsigned nxe:1;
255 		unsigned cr0_wp:1;
256 		unsigned smep_andnot_wp:1;
257 		unsigned smap_andnot_wp:1;
258 		unsigned ad_disabled:1;
259 		unsigned :7;
260 
261 		/*
262 		 * This is left at the top of the word so that
263 		 * kvm_memslots_for_spte_role can extract it with a
264 		 * simple shift.  While there is room, give it a whole
265 		 * byte so it is also faster to load it from memory.
266 		 */
267 		unsigned smm:8;
268 	};
269 };
270 
271 struct kvm_rmap_head {
272 	unsigned long val;
273 };
274 
275 struct kvm_mmu_page {
276 	struct list_head link;
277 	struct hlist_node hash_link;
278 
279 	/*
280 	 * The following two entries are used to key the shadow page in the
281 	 * hash table.
282 	 */
283 	gfn_t gfn;
284 	union kvm_mmu_page_role role;
285 
286 	u64 *spt;
287 	/* hold the gfn of each spte inside spt */
288 	gfn_t *gfns;
289 	bool unsync;
290 	int root_count;          /* Currently serving as active root */
291 	unsigned int unsync_children;
292 	struct kvm_rmap_head parent_ptes; /* rmap pointers to parent sptes */
293 
294 	/* The page is obsolete if mmu_valid_gen != kvm->arch.mmu_valid_gen.  */
295 	unsigned long mmu_valid_gen;
296 
297 	DECLARE_BITMAP(unsync_child_bitmap, 512);
298 
299 #ifdef CONFIG_X86_32
300 	/*
301 	 * Used out of the mmu-lock to avoid reading spte values while an
302 	 * update is in progress; see the comments in __get_spte_lockless().
303 	 */
304 	int clear_spte_count;
305 #endif
306 
307 	/* Number of writes since the last time traversal visited this page.  */
308 	atomic_t write_flooding_count;
309 };
310 
311 struct kvm_pio_request {
312 	unsigned long count;
313 	int in;
314 	int port;
315 	int size;
316 };
317 
318 #define PT64_ROOT_MAX_LEVEL 5
319 
320 struct rsvd_bits_validate {
321 	u64 rsvd_bits_mask[2][PT64_ROOT_MAX_LEVEL];
322 	u64 bad_mt_xwr;
323 };
324 
325 /*
326  * x86 supports 4 paging modes (5-level 64-bit, 4-level 64-bit, 3-level 32-bit,
327  * and 2-level 32-bit).  The kvm_mmu structure abstracts the details of the
328  * current mmu mode.
329  */
330 struct kvm_mmu {
331 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long root);
332 	unsigned long (*get_cr3)(struct kvm_vcpu *vcpu);
333 	u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
334 	int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err,
335 			  bool prefault);
336 	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
337 				  struct x86_exception *fault);
338 	gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva, u32 access,
339 			    struct x86_exception *exception);
340 	gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
341 			       struct x86_exception *exception);
342 	int (*sync_page)(struct kvm_vcpu *vcpu,
343 			 struct kvm_mmu_page *sp);
344 	void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva);
345 	void (*update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
346 			   u64 *spte, const void *pte);
347 	hpa_t root_hpa;
348 	union kvm_mmu_page_role base_role;
349 	u8 root_level;
350 	u8 shadow_root_level;
351 	u8 ept_ad;
352 	bool direct_map;
353 
354 	/*
355 	 * Bitmap; bit set = permission fault
356 	 * Byte index: page fault error code [4:1]
357 	 * Bit index: pte permissions in ACC_* format
358 	 */
359 	u8 permissions[16];
360 
361 	/*
362 	* The pkru_mask indicates if protection key checks are needed.  It
363 	* consists of 16 domains indexed by page fault error code bits [4:1],
364 	* with PFEC.RSVD replaced by ACC_USER_MASK from the page tables.
365 	* Each domain has 2 bits which are ANDed with AD and WD from PKRU.
366 	*/
367 	u32 pkru_mask;
368 
369 	u64 *pae_root;
370 	u64 *lm_root;
371 
372 	/*
373 	 * check zero bits on shadow page table entries, these
374 	 * bits include not only hardware reserved bits but also
375 	 * the bits spte never used.
376 	 */
377 	struct rsvd_bits_validate shadow_zero_check;
378 
379 	struct rsvd_bits_validate guest_rsvd_check;
380 
381 	/* Can have large pages at levels 2..last_nonleaf_level-1. */
382 	u8 last_nonleaf_level;
383 
384 	bool nx;
385 
386 	u64 pdptrs[4]; /* pae */
387 };
388 
389 enum pmc_type {
390 	KVM_PMC_GP = 0,
391 	KVM_PMC_FIXED,
392 };
393 
394 struct kvm_pmc {
395 	enum pmc_type type;
396 	u8 idx;
397 	u64 counter;
398 	u64 eventsel;
399 	struct perf_event *perf_event;
400 	struct kvm_vcpu *vcpu;
401 };
402 
403 struct kvm_pmu {
404 	unsigned nr_arch_gp_counters;
405 	unsigned nr_arch_fixed_counters;
406 	unsigned available_event_types;
407 	u64 fixed_ctr_ctrl;
408 	u64 global_ctrl;
409 	u64 global_status;
410 	u64 global_ovf_ctrl;
411 	u64 counter_bitmask[2];
412 	u64 global_ctrl_mask;
413 	u64 reserved_bits;
414 	u8 version;
415 	struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
416 	struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED];
417 	struct irq_work irq_work;
418 	u64 reprogram_pmi;
419 };
420 
421 struct kvm_pmu_ops;
422 
423 enum {
424 	KVM_DEBUGREG_BP_ENABLED = 1,
425 	KVM_DEBUGREG_WONT_EXIT = 2,
426 	KVM_DEBUGREG_RELOAD = 4,
427 };
428 
429 struct kvm_mtrr_range {
430 	u64 base;
431 	u64 mask;
432 	struct list_head node;
433 };
434 
435 struct kvm_mtrr {
436 	struct kvm_mtrr_range var_ranges[KVM_NR_VAR_MTRR];
437 	mtrr_type fixed_ranges[KVM_NR_FIXED_MTRR_REGION];
438 	u64 deftype;
439 
440 	struct list_head head;
441 };
442 
443 /* Hyper-V SynIC timer */
444 struct kvm_vcpu_hv_stimer {
445 	struct hrtimer timer;
446 	int index;
447 	u64 config;
448 	u64 count;
449 	u64 exp_time;
450 	struct hv_message msg;
451 	bool msg_pending;
452 };
453 
454 /* Hyper-V synthetic interrupt controller (SynIC)*/
455 struct kvm_vcpu_hv_synic {
456 	u64 version;
457 	u64 control;
458 	u64 msg_page;
459 	u64 evt_page;
460 	atomic64_t sint[HV_SYNIC_SINT_COUNT];
461 	atomic_t sint_to_gsi[HV_SYNIC_SINT_COUNT];
462 	DECLARE_BITMAP(auto_eoi_bitmap, 256);
463 	DECLARE_BITMAP(vec_bitmap, 256);
464 	bool active;
465 	bool dont_zero_synic_pages;
466 };
467 
468 /* Hyper-V per vcpu emulation context */
469 struct kvm_vcpu_hv {
470 	u32 vp_index;
471 	u64 hv_vapic;
472 	s64 runtime_offset;
473 	struct kvm_vcpu_hv_synic synic;
474 	struct kvm_hyperv_exit exit;
475 	struct kvm_vcpu_hv_stimer stimer[HV_SYNIC_STIMER_COUNT];
476 	DECLARE_BITMAP(stimer_pending_bitmap, HV_SYNIC_STIMER_COUNT);
477 };
478 
479 struct kvm_vcpu_arch {
480 	/*
481 	 * rip and regs accesses must go through
482 	 * kvm_{register,rip}_{read,write} functions.
483 	 */
484 	unsigned long regs[NR_VCPU_REGS];
485 	u32 regs_avail;
486 	u32 regs_dirty;
487 
488 	unsigned long cr0;
489 	unsigned long cr0_guest_owned_bits;
490 	unsigned long cr2;
491 	unsigned long cr3;
492 	unsigned long cr4;
493 	unsigned long cr4_guest_owned_bits;
494 	unsigned long cr8;
495 	u32 pkru;
496 	u32 hflags;
497 	u64 efer;
498 	u64 apic_base;
499 	struct kvm_lapic *apic;    /* kernel irqchip context */
500 	bool apicv_active;
501 	DECLARE_BITMAP(ioapic_handled_vectors, 256);
502 	unsigned long apic_attention;
503 	int32_t apic_arb_prio;
504 	int mp_state;
505 	u64 ia32_misc_enable_msr;
506 	u64 smbase;
507 	u64 smi_count;
508 	bool tpr_access_reporting;
509 	u64 ia32_xss;
510 
511 	/*
512 	 * Paging state of the vcpu
513 	 *
514 	 * If the vcpu runs in guest mode with two level paging this still saves
515 	 * the paging mode of the l1 guest. This context is always used to
516 	 * handle faults.
517 	 */
518 	struct kvm_mmu mmu;
519 
520 	/*
521 	 * Paging state of an L2 guest (used for nested npt)
522 	 *
523 	 * This context will save all necessary information to walk page tables
524 	 * of the an L2 guest. This context is only initialized for page table
525 	 * walking and not for faulting since we never handle l2 page faults on
526 	 * the host.
527 	 */
528 	struct kvm_mmu nested_mmu;
529 
530 	/*
531 	 * Pointer to the mmu context currently used for
532 	 * gva_to_gpa translations.
533 	 */
534 	struct kvm_mmu *walk_mmu;
535 
536 	struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
537 	struct kvm_mmu_memory_cache mmu_page_cache;
538 	struct kvm_mmu_memory_cache mmu_page_header_cache;
539 
540 	/*
541 	 * QEMU userspace and the guest each have their own FPU state.
542 	 * In vcpu_run, we switch between the user and guest FPU contexts.
543 	 * While running a VCPU, the VCPU thread will have the guest FPU
544 	 * context.
545 	 *
546 	 * Note that while the PKRU state lives inside the fpu registers,
547 	 * it is switched out separately at VMENTER and VMEXIT time. The
548 	 * "guest_fpu" state here contains the guest FPU context, with the
549 	 * host PRKU bits.
550 	 */
551 	struct fpu user_fpu;
552 	struct fpu guest_fpu;
553 
554 	u64 xcr0;
555 	u64 guest_supported_xcr0;
556 	u32 guest_xstate_size;
557 
558 	struct kvm_pio_request pio;
559 	void *pio_data;
560 
561 	u8 event_exit_inst_len;
562 
563 	struct kvm_queued_exception {
564 		bool pending;
565 		bool injected;
566 		bool has_error_code;
567 		u8 nr;
568 		u32 error_code;
569 		u8 nested_apf;
570 	} exception;
571 
572 	struct kvm_queued_interrupt {
573 		bool pending;
574 		bool soft;
575 		u8 nr;
576 	} interrupt;
577 
578 	int halt_request; /* real mode on Intel only */
579 
580 	int cpuid_nent;
581 	struct kvm_cpuid_entry2 cpuid_entries[KVM_MAX_CPUID_ENTRIES];
582 
583 	int maxphyaddr;
584 
585 	/* emulate context */
586 
587 	struct x86_emulate_ctxt emulate_ctxt;
588 	bool emulate_regs_need_sync_to_vcpu;
589 	bool emulate_regs_need_sync_from_vcpu;
590 	int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
591 
592 	gpa_t time;
593 	struct pvclock_vcpu_time_info hv_clock;
594 	unsigned int hw_tsc_khz;
595 	struct gfn_to_hva_cache pv_time;
596 	bool pv_time_enabled;
597 	/* set guest stopped flag in pvclock flags field */
598 	bool pvclock_set_guest_stopped_request;
599 
600 	struct {
601 		u64 msr_val;
602 		u64 last_steal;
603 		struct gfn_to_hva_cache stime;
604 		struct kvm_steal_time steal;
605 	} st;
606 
607 	u64 tsc_offset;
608 	u64 last_guest_tsc;
609 	u64 last_host_tsc;
610 	u64 tsc_offset_adjustment;
611 	u64 this_tsc_nsec;
612 	u64 this_tsc_write;
613 	u64 this_tsc_generation;
614 	bool tsc_catchup;
615 	bool tsc_always_catchup;
616 	s8 virtual_tsc_shift;
617 	u32 virtual_tsc_mult;
618 	u32 virtual_tsc_khz;
619 	s64 ia32_tsc_adjust_msr;
620 	u64 tsc_scaling_ratio;
621 
622 	atomic_t nmi_queued;  /* unprocessed asynchronous NMIs */
623 	unsigned nmi_pending; /* NMI queued after currently running handler */
624 	bool nmi_injected;    /* Trying to inject an NMI this entry */
625 	bool smi_pending;    /* SMI queued after currently running handler */
626 
627 	struct kvm_mtrr mtrr_state;
628 	u64 pat;
629 
630 	unsigned switch_db_regs;
631 	unsigned long db[KVM_NR_DB_REGS];
632 	unsigned long dr6;
633 	unsigned long dr7;
634 	unsigned long eff_db[KVM_NR_DB_REGS];
635 	unsigned long guest_debug_dr7;
636 	u64 msr_platform_info;
637 	u64 msr_misc_features_enables;
638 
639 	u64 mcg_cap;
640 	u64 mcg_status;
641 	u64 mcg_ctl;
642 	u64 mcg_ext_ctl;
643 	u64 *mce_banks;
644 
645 	/* Cache MMIO info */
646 	u64 mmio_gva;
647 	unsigned access;
648 	gfn_t mmio_gfn;
649 	u64 mmio_gen;
650 
651 	struct kvm_pmu pmu;
652 
653 	/* used for guest single stepping over the given code position */
654 	unsigned long singlestep_rip;
655 
656 	struct kvm_vcpu_hv hyperv;
657 
658 	cpumask_var_t wbinvd_dirty_mask;
659 
660 	unsigned long last_retry_eip;
661 	unsigned long last_retry_addr;
662 
663 	struct {
664 		bool halted;
665 		gfn_t gfns[roundup_pow_of_two(ASYNC_PF_PER_VCPU)];
666 		struct gfn_to_hva_cache data;
667 		u64 msr_val;
668 		u32 id;
669 		bool send_user_only;
670 		u32 host_apf_reason;
671 		unsigned long nested_apf_token;
672 		bool delivery_as_pf_vmexit;
673 	} apf;
674 
675 	/* OSVW MSRs (AMD only) */
676 	struct {
677 		u64 length;
678 		u64 status;
679 	} osvw;
680 
681 	struct {
682 		u64 msr_val;
683 		struct gfn_to_hva_cache data;
684 	} pv_eoi;
685 
686 	/*
687 	 * Indicate whether the access faults on its page table in guest
688 	 * which is set when fix page fault and used to detect unhandeable
689 	 * instruction.
690 	 */
691 	bool write_fault_to_shadow_pgtable;
692 
693 	/* set at EPT violation at this point */
694 	unsigned long exit_qualification;
695 
696 	/* pv related host specific info */
697 	struct {
698 		bool pv_unhalted;
699 	} pv;
700 
701 	int pending_ioapic_eoi;
702 	int pending_external_vector;
703 
704 	/* GPA available */
705 	bool gpa_available;
706 	gpa_t gpa_val;
707 
708 	/* be preempted when it's in kernel-mode(cpl=0) */
709 	bool preempted_in_kernel;
710 };
711 
712 struct kvm_lpage_info {
713 	int disallow_lpage;
714 };
715 
716 struct kvm_arch_memory_slot {
717 	struct kvm_rmap_head *rmap[KVM_NR_PAGE_SIZES];
718 	struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
719 	unsigned short *gfn_track[KVM_PAGE_TRACK_MAX];
720 };
721 
722 /*
723  * We use as the mode the number of bits allocated in the LDR for the
724  * logical processor ID.  It happens that these are all powers of two.
725  * This makes it is very easy to detect cases where the APICs are
726  * configured for multiple modes; in that case, we cannot use the map and
727  * hence cannot use kvm_irq_delivery_to_apic_fast either.
728  */
729 #define KVM_APIC_MODE_XAPIC_CLUSTER          4
730 #define KVM_APIC_MODE_XAPIC_FLAT             8
731 #define KVM_APIC_MODE_X2APIC                16
732 
733 struct kvm_apic_map {
734 	struct rcu_head rcu;
735 	u8 mode;
736 	u32 max_apic_id;
737 	union {
738 		struct kvm_lapic *xapic_flat_map[8];
739 		struct kvm_lapic *xapic_cluster_map[16][4];
740 	};
741 	struct kvm_lapic *phys_map[];
742 };
743 
744 /* Hyper-V emulation context */
745 struct kvm_hv {
746 	struct mutex hv_lock;
747 	u64 hv_guest_os_id;
748 	u64 hv_hypercall;
749 	u64 hv_tsc_page;
750 
751 	/* Hyper-v based guest crash (NT kernel bugcheck) parameters */
752 	u64 hv_crash_param[HV_X64_MSR_CRASH_PARAMS];
753 	u64 hv_crash_ctl;
754 
755 	HV_REFERENCE_TSC_PAGE tsc_ref;
756 };
757 
758 enum kvm_irqchip_mode {
759 	KVM_IRQCHIP_NONE,
760 	KVM_IRQCHIP_KERNEL,       /* created with KVM_CREATE_IRQCHIP */
761 	KVM_IRQCHIP_SPLIT,        /* created with KVM_CAP_SPLIT_IRQCHIP */
762 };
763 
764 struct kvm_sev_info {
765 	bool active;		/* SEV enabled guest */
766 	unsigned int asid;	/* ASID used for this guest */
767 	unsigned int handle;	/* SEV firmware handle */
768 	int fd;			/* SEV device fd */
769 	unsigned long pages_locked; /* Number of pages locked */
770 	struct list_head regions_list;  /* List of registered regions */
771 };
772 
773 struct kvm_arch {
774 	unsigned int n_used_mmu_pages;
775 	unsigned int n_requested_mmu_pages;
776 	unsigned int n_max_mmu_pages;
777 	unsigned int indirect_shadow_pages;
778 	unsigned long mmu_valid_gen;
779 	struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
780 	/*
781 	 * Hash table of struct kvm_mmu_page.
782 	 */
783 	struct list_head active_mmu_pages;
784 	struct list_head zapped_obsolete_pages;
785 	struct kvm_page_track_notifier_node mmu_sp_tracker;
786 	struct kvm_page_track_notifier_head track_notifier_head;
787 
788 	struct list_head assigned_dev_head;
789 	struct iommu_domain *iommu_domain;
790 	bool iommu_noncoherent;
791 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA
792 	atomic_t noncoherent_dma_count;
793 #define __KVM_HAVE_ARCH_ASSIGNED_DEVICE
794 	atomic_t assigned_device_count;
795 	struct kvm_pic *vpic;
796 	struct kvm_ioapic *vioapic;
797 	struct kvm_pit *vpit;
798 	atomic_t vapics_in_nmi_mode;
799 	struct mutex apic_map_lock;
800 	struct kvm_apic_map *apic_map;
801 
802 	unsigned int tss_addr;
803 	bool apic_access_page_done;
804 
805 	gpa_t wall_clock;
806 
807 	bool ept_identity_pagetable_done;
808 	gpa_t ept_identity_map_addr;
809 
810 	unsigned long irq_sources_bitmap;
811 	s64 kvmclock_offset;
812 	raw_spinlock_t tsc_write_lock;
813 	u64 last_tsc_nsec;
814 	u64 last_tsc_write;
815 	u32 last_tsc_khz;
816 	u64 cur_tsc_nsec;
817 	u64 cur_tsc_write;
818 	u64 cur_tsc_offset;
819 	u64 cur_tsc_generation;
820 	int nr_vcpus_matched_tsc;
821 
822 	spinlock_t pvclock_gtod_sync_lock;
823 	bool use_master_clock;
824 	u64 master_kernel_ns;
825 	u64 master_cycle_now;
826 	struct delayed_work kvmclock_update_work;
827 	struct delayed_work kvmclock_sync_work;
828 
829 	struct kvm_xen_hvm_config xen_hvm_config;
830 
831 	/* reads protected by irq_srcu, writes by irq_lock */
832 	struct hlist_head mask_notifier_list;
833 
834 	struct kvm_hv hyperv;
835 
836 	#ifdef CONFIG_KVM_MMU_AUDIT
837 	int audit_point;
838 	#endif
839 
840 	bool backwards_tsc_observed;
841 	bool boot_vcpu_runs_old_kvmclock;
842 	u32 bsp_vcpu_id;
843 
844 	u64 disabled_quirks;
845 
846 	enum kvm_irqchip_mode irqchip_mode;
847 	u8 nr_reserved_ioapic_pins;
848 
849 	bool disabled_lapic_found;
850 
851 	/* Struct members for AVIC */
852 	u32 avic_vm_id;
853 	u32 ldr_mode;
854 	struct page *avic_logical_id_table_page;
855 	struct page *avic_physical_id_table_page;
856 	struct hlist_node hnode;
857 
858 	bool x2apic_format;
859 	bool x2apic_broadcast_quirk_disabled;
860 
861 	struct kvm_sev_info sev_info;
862 };
863 
864 struct kvm_vm_stat {
865 	ulong mmu_shadow_zapped;
866 	ulong mmu_pte_write;
867 	ulong mmu_pte_updated;
868 	ulong mmu_pde_zapped;
869 	ulong mmu_flooded;
870 	ulong mmu_recycled;
871 	ulong mmu_cache_miss;
872 	ulong mmu_unsync;
873 	ulong remote_tlb_flush;
874 	ulong lpages;
875 	ulong max_mmu_page_hash_collisions;
876 };
877 
878 struct kvm_vcpu_stat {
879 	u64 pf_fixed;
880 	u64 pf_guest;
881 	u64 tlb_flush;
882 	u64 invlpg;
883 
884 	u64 exits;
885 	u64 io_exits;
886 	u64 mmio_exits;
887 	u64 signal_exits;
888 	u64 irq_window_exits;
889 	u64 nmi_window_exits;
890 	u64 halt_exits;
891 	u64 halt_successful_poll;
892 	u64 halt_attempted_poll;
893 	u64 halt_poll_invalid;
894 	u64 halt_wakeup;
895 	u64 request_irq_exits;
896 	u64 irq_exits;
897 	u64 host_state_reload;
898 	u64 fpu_reload;
899 	u64 insn_emulation;
900 	u64 insn_emulation_fail;
901 	u64 hypercalls;
902 	u64 irq_injections;
903 	u64 nmi_injections;
904 	u64 req_event;
905 };
906 
907 struct x86_instruction_info;
908 
909 struct msr_data {
910 	bool host_initiated;
911 	u32 index;
912 	u64 data;
913 };
914 
915 struct kvm_lapic_irq {
916 	u32 vector;
917 	u16 delivery_mode;
918 	u16 dest_mode;
919 	bool level;
920 	u16 trig_mode;
921 	u32 shorthand;
922 	u32 dest_id;
923 	bool msi_redir_hint;
924 };
925 
926 struct kvm_x86_ops {
927 	int (*cpu_has_kvm_support)(void);          /* __init */
928 	int (*disabled_by_bios)(void);             /* __init */
929 	int (*hardware_enable)(void);
930 	void (*hardware_disable)(void);
931 	void (*check_processor_compatibility)(void *rtn);
932 	int (*hardware_setup)(void);               /* __init */
933 	void (*hardware_unsetup)(void);            /* __exit */
934 	bool (*cpu_has_accelerated_tpr)(void);
935 	bool (*cpu_has_high_real_mode_segbase)(void);
936 	void (*cpuid_update)(struct kvm_vcpu *vcpu);
937 
938 	int (*vm_init)(struct kvm *kvm);
939 	void (*vm_destroy)(struct kvm *kvm);
940 
941 	/* Create, but do not attach this VCPU */
942 	struct kvm_vcpu *(*vcpu_create)(struct kvm *kvm, unsigned id);
943 	void (*vcpu_free)(struct kvm_vcpu *vcpu);
944 	void (*vcpu_reset)(struct kvm_vcpu *vcpu, bool init_event);
945 
946 	void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
947 	void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
948 	void (*vcpu_put)(struct kvm_vcpu *vcpu);
949 
950 	void (*update_bp_intercept)(struct kvm_vcpu *vcpu);
951 	int (*get_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
952 	int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
953 	u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
954 	void (*get_segment)(struct kvm_vcpu *vcpu,
955 			    struct kvm_segment *var, int seg);
956 	int (*get_cpl)(struct kvm_vcpu *vcpu);
957 	void (*set_segment)(struct kvm_vcpu *vcpu,
958 			    struct kvm_segment *var, int seg);
959 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
960 	void (*decache_cr0_guest_bits)(struct kvm_vcpu *vcpu);
961 	void (*decache_cr3)(struct kvm_vcpu *vcpu);
962 	void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu);
963 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
964 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
965 	int (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
966 	void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
967 	void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
968 	void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
969 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
970 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
971 	u64 (*get_dr6)(struct kvm_vcpu *vcpu);
972 	void (*set_dr6)(struct kvm_vcpu *vcpu, unsigned long value);
973 	void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu);
974 	void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
975 	void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
976 	unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
977 	void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
978 
979 	void (*tlb_flush)(struct kvm_vcpu *vcpu, bool invalidate_gpa);
980 
981 	void (*run)(struct kvm_vcpu *vcpu);
982 	int (*handle_exit)(struct kvm_vcpu *vcpu);
983 	void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
984 	void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
985 	u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu);
986 	void (*patch_hypercall)(struct kvm_vcpu *vcpu,
987 				unsigned char *hypercall_addr);
988 	void (*set_irq)(struct kvm_vcpu *vcpu);
989 	void (*set_nmi)(struct kvm_vcpu *vcpu);
990 	void (*queue_exception)(struct kvm_vcpu *vcpu);
991 	void (*cancel_injection)(struct kvm_vcpu *vcpu);
992 	int (*interrupt_allowed)(struct kvm_vcpu *vcpu);
993 	int (*nmi_allowed)(struct kvm_vcpu *vcpu);
994 	bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
995 	void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
996 	void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
997 	void (*enable_irq_window)(struct kvm_vcpu *vcpu);
998 	void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
999 	bool (*get_enable_apicv)(struct kvm_vcpu *vcpu);
1000 	void (*refresh_apicv_exec_ctrl)(struct kvm_vcpu *vcpu);
1001 	void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
1002 	void (*hwapic_isr_update)(struct kvm_vcpu *vcpu, int isr);
1003 	void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
1004 	void (*set_virtual_x2apic_mode)(struct kvm_vcpu *vcpu, bool set);
1005 	void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu, hpa_t hpa);
1006 	void (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
1007 	int (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
1008 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
1009 	int (*get_tdp_level)(struct kvm_vcpu *vcpu);
1010 	u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
1011 	int (*get_lpage_level)(void);
1012 	bool (*rdtscp_supported)(void);
1013 	bool (*invpcid_supported)(void);
1014 
1015 	void (*set_tdp_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
1016 
1017 	void (*set_supported_cpuid)(u32 func, struct kvm_cpuid_entry2 *entry);
1018 
1019 	bool (*has_wbinvd_exit)(void);
1020 
1021 	void (*write_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
1022 
1023 	void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2);
1024 
1025 	int (*check_intercept)(struct kvm_vcpu *vcpu,
1026 			       struct x86_instruction_info *info,
1027 			       enum x86_intercept_stage stage);
1028 	void (*handle_external_intr)(struct kvm_vcpu *vcpu);
1029 	bool (*mpx_supported)(void);
1030 	bool (*xsaves_supported)(void);
1031 	bool (*umip_emulated)(void);
1032 
1033 	int (*check_nested_events)(struct kvm_vcpu *vcpu, bool external_intr);
1034 
1035 	void (*sched_in)(struct kvm_vcpu *kvm, int cpu);
1036 
1037 	/*
1038 	 * Arch-specific dirty logging hooks. These hooks are only supposed to
1039 	 * be valid if the specific arch has hardware-accelerated dirty logging
1040 	 * mechanism. Currently only for PML on VMX.
1041 	 *
1042 	 *  - slot_enable_log_dirty:
1043 	 *	called when enabling log dirty mode for the slot.
1044 	 *  - slot_disable_log_dirty:
1045 	 *	called when disabling log dirty mode for the slot.
1046 	 *	also called when slot is created with log dirty disabled.
1047 	 *  - flush_log_dirty:
1048 	 *	called before reporting dirty_bitmap to userspace.
1049 	 *  - enable_log_dirty_pt_masked:
1050 	 *	called when reenabling log dirty for the GFNs in the mask after
1051 	 *	corresponding bits are cleared in slot->dirty_bitmap.
1052 	 */
1053 	void (*slot_enable_log_dirty)(struct kvm *kvm,
1054 				      struct kvm_memory_slot *slot);
1055 	void (*slot_disable_log_dirty)(struct kvm *kvm,
1056 				       struct kvm_memory_slot *slot);
1057 	void (*flush_log_dirty)(struct kvm *kvm);
1058 	void (*enable_log_dirty_pt_masked)(struct kvm *kvm,
1059 					   struct kvm_memory_slot *slot,
1060 					   gfn_t offset, unsigned long mask);
1061 	int (*write_log_dirty)(struct kvm_vcpu *vcpu);
1062 
1063 	/* pmu operations of sub-arch */
1064 	const struct kvm_pmu_ops *pmu_ops;
1065 
1066 	/*
1067 	 * Architecture specific hooks for vCPU blocking due to
1068 	 * HLT instruction.
1069 	 * Returns for .pre_block():
1070 	 *    - 0 means continue to block the vCPU.
1071 	 *    - 1 means we cannot block the vCPU since some event
1072 	 *        happens during this period, such as, 'ON' bit in
1073 	 *        posted-interrupts descriptor is set.
1074 	 */
1075 	int (*pre_block)(struct kvm_vcpu *vcpu);
1076 	void (*post_block)(struct kvm_vcpu *vcpu);
1077 
1078 	void (*vcpu_blocking)(struct kvm_vcpu *vcpu);
1079 	void (*vcpu_unblocking)(struct kvm_vcpu *vcpu);
1080 
1081 	int (*update_pi_irte)(struct kvm *kvm, unsigned int host_irq,
1082 			      uint32_t guest_irq, bool set);
1083 	void (*apicv_post_state_restore)(struct kvm_vcpu *vcpu);
1084 
1085 	int (*set_hv_timer)(struct kvm_vcpu *vcpu, u64 guest_deadline_tsc);
1086 	void (*cancel_hv_timer)(struct kvm_vcpu *vcpu);
1087 
1088 	void (*setup_mce)(struct kvm_vcpu *vcpu);
1089 
1090 	int (*smi_allowed)(struct kvm_vcpu *vcpu);
1091 	int (*pre_enter_smm)(struct kvm_vcpu *vcpu, char *smstate);
1092 	int (*pre_leave_smm)(struct kvm_vcpu *vcpu, u64 smbase);
1093 	int (*enable_smi_window)(struct kvm_vcpu *vcpu);
1094 
1095 	int (*mem_enc_op)(struct kvm *kvm, void __user *argp);
1096 	int (*mem_enc_reg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1097 	int (*mem_enc_unreg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1098 };
1099 
1100 struct kvm_arch_async_pf {
1101 	u32 token;
1102 	gfn_t gfn;
1103 	unsigned long cr3;
1104 	bool direct_map;
1105 };
1106 
1107 extern struct kvm_x86_ops *kvm_x86_ops;
1108 
1109 int kvm_mmu_module_init(void);
1110 void kvm_mmu_module_exit(void);
1111 
1112 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
1113 int kvm_mmu_create(struct kvm_vcpu *vcpu);
1114 void kvm_mmu_setup(struct kvm_vcpu *vcpu);
1115 void kvm_mmu_init_vm(struct kvm *kvm);
1116 void kvm_mmu_uninit_vm(struct kvm *kvm);
1117 void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
1118 		u64 dirty_mask, u64 nx_mask, u64 x_mask, u64 p_mask,
1119 		u64 acc_track_mask, u64 me_mask);
1120 
1121 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
1122 void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
1123 				      struct kvm_memory_slot *memslot);
1124 void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
1125 				   const struct kvm_memory_slot *memslot);
1126 void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
1127 				   struct kvm_memory_slot *memslot);
1128 void kvm_mmu_slot_largepage_remove_write_access(struct kvm *kvm,
1129 					struct kvm_memory_slot *memslot);
1130 void kvm_mmu_slot_set_dirty(struct kvm *kvm,
1131 			    struct kvm_memory_slot *memslot);
1132 void kvm_mmu_clear_dirty_pt_masked(struct kvm *kvm,
1133 				   struct kvm_memory_slot *slot,
1134 				   gfn_t gfn_offset, unsigned long mask);
1135 void kvm_mmu_zap_all(struct kvm *kvm);
1136 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, struct kvm_memslots *slots);
1137 unsigned int kvm_mmu_calculate_mmu_pages(struct kvm *kvm);
1138 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int kvm_nr_mmu_pages);
1139 
1140 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
1141 bool pdptrs_changed(struct kvm_vcpu *vcpu);
1142 
1143 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
1144 			  const void *val, int bytes);
1145 
1146 struct kvm_irq_mask_notifier {
1147 	void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
1148 	int irq;
1149 	struct hlist_node link;
1150 };
1151 
1152 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
1153 				    struct kvm_irq_mask_notifier *kimn);
1154 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
1155 				      struct kvm_irq_mask_notifier *kimn);
1156 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
1157 			     bool mask);
1158 
1159 extern bool tdp_enabled;
1160 
1161 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
1162 
1163 /* control of guest tsc rate supported? */
1164 extern bool kvm_has_tsc_control;
1165 /* maximum supported tsc_khz for guests */
1166 extern u32  kvm_max_guest_tsc_khz;
1167 /* number of bits of the fractional part of the TSC scaling ratio */
1168 extern u8   kvm_tsc_scaling_ratio_frac_bits;
1169 /* maximum allowed value of TSC scaling ratio */
1170 extern u64  kvm_max_tsc_scaling_ratio;
1171 /* 1ull << kvm_tsc_scaling_ratio_frac_bits */
1172 extern u64  kvm_default_tsc_scaling_ratio;
1173 
1174 extern u64 kvm_mce_cap_supported;
1175 
1176 enum emulation_result {
1177 	EMULATE_DONE,         /* no further processing */
1178 	EMULATE_USER_EXIT,    /* kvm_run ready for userspace exit */
1179 	EMULATE_FAIL,         /* can't emulate this instruction */
1180 };
1181 
1182 #define EMULTYPE_NO_DECODE	    (1 << 0)
1183 #define EMULTYPE_TRAP_UD	    (1 << 1)
1184 #define EMULTYPE_SKIP		    (1 << 2)
1185 #define EMULTYPE_RETRY		    (1 << 3)
1186 #define EMULTYPE_NO_REEXECUTE	    (1 << 4)
1187 int x86_emulate_instruction(struct kvm_vcpu *vcpu, unsigned long cr2,
1188 			    int emulation_type, void *insn, int insn_len);
1189 
1190 static inline int emulate_instruction(struct kvm_vcpu *vcpu,
1191 			int emulation_type)
1192 {
1193 	return x86_emulate_instruction(vcpu, 0,
1194 			emulation_type | EMULTYPE_NO_REEXECUTE, NULL, 0);
1195 }
1196 
1197 void kvm_enable_efer_bits(u64);
1198 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
1199 int kvm_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr);
1200 int kvm_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr);
1201 
1202 struct x86_emulate_ctxt;
1203 
1204 int kvm_fast_pio_out(struct kvm_vcpu *vcpu, int size, unsigned short port);
1205 int kvm_fast_pio_in(struct kvm_vcpu *vcpu, int size, unsigned short port);
1206 int kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
1207 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
1208 int kvm_vcpu_halt(struct kvm_vcpu *vcpu);
1209 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
1210 
1211 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
1212 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
1213 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector);
1214 
1215 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
1216 		    int reason, bool has_error_code, u32 error_code);
1217 
1218 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
1219 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
1220 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
1221 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
1222 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
1223 int kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
1224 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
1225 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
1226 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
1227 int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr);
1228 
1229 int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1230 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1231 
1232 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
1233 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
1234 bool kvm_rdpmc(struct kvm_vcpu *vcpu);
1235 
1236 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1237 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1238 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1239 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1240 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
1241 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1242 			    gfn_t gfn, void *data, int offset, int len,
1243 			    u32 access);
1244 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
1245 bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr);
1246 
1247 static inline int __kvm_irq_line_state(unsigned long *irq_state,
1248 				       int irq_source_id, int level)
1249 {
1250 	/* Logical OR for level trig interrupt */
1251 	if (level)
1252 		__set_bit(irq_source_id, irq_state);
1253 	else
1254 		__clear_bit(irq_source_id, irq_state);
1255 
1256 	return !!(*irq_state);
1257 }
1258 
1259 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
1260 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
1261 
1262 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
1263 
1264 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
1265 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
1266 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
1267 int kvm_mmu_load(struct kvm_vcpu *vcpu);
1268 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
1269 void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu);
1270 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1271 			   struct x86_exception *exception);
1272 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
1273 			      struct x86_exception *exception);
1274 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
1275 			       struct x86_exception *exception);
1276 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
1277 			       struct x86_exception *exception);
1278 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
1279 				struct x86_exception *exception);
1280 
1281 void kvm_vcpu_deactivate_apicv(struct kvm_vcpu *vcpu);
1282 
1283 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
1284 
1285 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva, u64 error_code,
1286 		       void *insn, int insn_len);
1287 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
1288 void kvm_mmu_new_cr3(struct kvm_vcpu *vcpu);
1289 
1290 void kvm_enable_tdp(void);
1291 void kvm_disable_tdp(void);
1292 
1293 static inline gpa_t translate_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1294 				  struct x86_exception *exception)
1295 {
1296 	return gpa;
1297 }
1298 
1299 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
1300 {
1301 	struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
1302 
1303 	return (struct kvm_mmu_page *)page_private(page);
1304 }
1305 
1306 static inline u16 kvm_read_ldt(void)
1307 {
1308 	u16 ldt;
1309 	asm("sldt %0" : "=g"(ldt));
1310 	return ldt;
1311 }
1312 
1313 static inline void kvm_load_ldt(u16 sel)
1314 {
1315 	asm("lldt %0" : : "rm"(sel));
1316 }
1317 
1318 #ifdef CONFIG_X86_64
1319 static inline unsigned long read_msr(unsigned long msr)
1320 {
1321 	u64 value;
1322 
1323 	rdmsrl(msr, value);
1324 	return value;
1325 }
1326 #endif
1327 
1328 static inline u32 get_rdx_init_val(void)
1329 {
1330 	return 0x600; /* P6 family */
1331 }
1332 
1333 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
1334 {
1335 	kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
1336 }
1337 
1338 #define TSS_IOPB_BASE_OFFSET 0x66
1339 #define TSS_BASE_SIZE 0x68
1340 #define TSS_IOPB_SIZE (65536 / 8)
1341 #define TSS_REDIRECTION_SIZE (256 / 8)
1342 #define RMODE_TSS_SIZE							\
1343 	(TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
1344 
1345 enum {
1346 	TASK_SWITCH_CALL = 0,
1347 	TASK_SWITCH_IRET = 1,
1348 	TASK_SWITCH_JMP = 2,
1349 	TASK_SWITCH_GATE = 3,
1350 };
1351 
1352 #define HF_GIF_MASK		(1 << 0)
1353 #define HF_HIF_MASK		(1 << 1)
1354 #define HF_VINTR_MASK		(1 << 2)
1355 #define HF_NMI_MASK		(1 << 3)
1356 #define HF_IRET_MASK		(1 << 4)
1357 #define HF_GUEST_MASK		(1 << 5) /* VCPU is in guest-mode */
1358 #define HF_SMM_MASK		(1 << 6)
1359 #define HF_SMM_INSIDE_NMI_MASK	(1 << 7)
1360 
1361 #define __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
1362 #define KVM_ADDRESS_SPACE_NUM 2
1363 
1364 #define kvm_arch_vcpu_memslots_id(vcpu) ((vcpu)->arch.hflags & HF_SMM_MASK ? 1 : 0)
1365 #define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, (role).smm)
1366 
1367 /*
1368  * Hardware virtualization extension instructions may fault if a
1369  * reboot turns off virtualization while processes are running.
1370  * Trap the fault and ignore the instruction if that happens.
1371  */
1372 asmlinkage void kvm_spurious_fault(void);
1373 
1374 #define ____kvm_handle_fault_on_reboot(insn, cleanup_insn)	\
1375 	"666: " insn "\n\t" \
1376 	"668: \n\t"                           \
1377 	".pushsection .fixup, \"ax\" \n" \
1378 	"667: \n\t" \
1379 	cleanup_insn "\n\t"		      \
1380 	"cmpb $0, kvm_rebooting \n\t"	      \
1381 	"jne 668b \n\t"      		      \
1382 	__ASM_SIZE(push) " $666b \n\t"	      \
1383 	"call kvm_spurious_fault \n\t"	      \
1384 	".popsection \n\t" \
1385 	_ASM_EXTABLE(666b, 667b)
1386 
1387 #define __kvm_handle_fault_on_reboot(insn)		\
1388 	____kvm_handle_fault_on_reboot(insn, "")
1389 
1390 #define KVM_ARCH_WANT_MMU_NOTIFIER
1391 int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
1392 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end);
1393 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
1394 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
1395 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
1396 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1397 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1398 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1399 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1400 void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event);
1401 void kvm_vcpu_reload_apic_access_page(struct kvm_vcpu *vcpu);
1402 
1403 void kvm_define_shared_msr(unsigned index, u32 msr);
1404 int kvm_set_shared_msr(unsigned index, u64 val, u64 mask);
1405 
1406 u64 kvm_scale_tsc(struct kvm_vcpu *vcpu, u64 tsc);
1407 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc);
1408 
1409 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu);
1410 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1411 
1412 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
1413 void kvm_make_scan_ioapic_request(struct kvm *kvm);
1414 
1415 void kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1416 				     struct kvm_async_pf *work);
1417 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1418 				 struct kvm_async_pf *work);
1419 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1420 			       struct kvm_async_pf *work);
1421 bool kvm_arch_can_inject_async_page_present(struct kvm_vcpu *vcpu);
1422 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1423 
1424 int kvm_skip_emulated_instruction(struct kvm_vcpu *vcpu);
1425 int kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1426 
1427 int kvm_is_in_guest(void);
1428 
1429 int __x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa, u32 size);
1430 int x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa, u32 size);
1431 bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu);
1432 bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu);
1433 
1434 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
1435 			     struct kvm_vcpu **dest_vcpu);
1436 
1437 void kvm_set_msi_irq(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e,
1438 		     struct kvm_lapic_irq *irq);
1439 
1440 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu)
1441 {
1442 	if (kvm_x86_ops->vcpu_blocking)
1443 		kvm_x86_ops->vcpu_blocking(vcpu);
1444 }
1445 
1446 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu)
1447 {
1448 	if (kvm_x86_ops->vcpu_unblocking)
1449 		kvm_x86_ops->vcpu_unblocking(vcpu);
1450 }
1451 
1452 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
1453 
1454 static inline int kvm_cpu_get_apicid(int mps_cpu)
1455 {
1456 #ifdef CONFIG_X86_LOCAL_APIC
1457 	return default_cpu_present_to_apicid(mps_cpu);
1458 #else
1459 	WARN_ON_ONCE(1);
1460 	return BAD_APICID;
1461 #endif
1462 }
1463 
1464 #define put_smstate(type, buf, offset, val)                      \
1465 	*(type *)((buf) + (offset) - 0x7e00) = val
1466 
1467 void kvm_arch_mmu_notifier_invalidate_range(struct kvm *kvm,
1468 		unsigned long start, unsigned long end);
1469 
1470 #endif /* _ASM_X86_KVM_HOST_H */
1471