xref: /linux/arch/x86/include/asm/kvm_host.h (revision c6ed444fd6fffaaf2e3857d926ed18bf3df81e8e)
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_EVENT			KVM_ARCH_REQ(6)
59 #define KVM_REQ_APF_HALT		KVM_ARCH_REQ(7)
60 #define KVM_REQ_STEAL_UPDATE		KVM_ARCH_REQ(8)
61 #define KVM_REQ_NMI			KVM_ARCH_REQ(9)
62 #define KVM_REQ_PMU			KVM_ARCH_REQ(10)
63 #define KVM_REQ_PMI			KVM_ARCH_REQ(11)
64 #define KVM_REQ_SMI			KVM_ARCH_REQ(12)
65 #define KVM_REQ_MASTERCLOCK_UPDATE	KVM_ARCH_REQ(13)
66 #define KVM_REQ_MCLOCK_INPROGRESS \
67 	KVM_ARCH_REQ_FLAGS(14, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
68 #define KVM_REQ_SCAN_IOAPIC \
69 	KVM_ARCH_REQ_FLAGS(15, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
70 #define KVM_REQ_GLOBAL_CLOCK_UPDATE	KVM_ARCH_REQ(16)
71 #define KVM_REQ_APIC_PAGE_RELOAD \
72 	KVM_ARCH_REQ_FLAGS(17, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
73 #define KVM_REQ_HV_CRASH		KVM_ARCH_REQ(18)
74 #define KVM_REQ_IOAPIC_EOI_EXIT		KVM_ARCH_REQ(19)
75 #define KVM_REQ_HV_RESET		KVM_ARCH_REQ(20)
76 #define KVM_REQ_HV_EXIT			KVM_ARCH_REQ(21)
77 #define KVM_REQ_HV_STIMER		KVM_ARCH_REQ(22)
78 #define KVM_REQ_LOAD_EOI_EXITMAP	KVM_ARCH_REQ(23)
79 
80 #define CR0_RESERVED_BITS                                               \
81 	(~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \
82 			  | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \
83 			  | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG))
84 
85 #define CR3_PCID_INVD		 BIT_64(63)
86 #define CR4_RESERVED_BITS                                               \
87 	(~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\
88 			  | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE     \
89 			  | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR | X86_CR4_PCIDE \
90 			  | X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_FSGSBASE \
91 			  | X86_CR4_OSXMMEXCPT | X86_CR4_LA57 | X86_CR4_VMXE \
92 			  | X86_CR4_SMAP | X86_CR4_PKE | X86_CR4_UMIP))
93 
94 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR)
95 
96 
97 
98 #define INVALID_PAGE (~(hpa_t)0)
99 #define VALID_PAGE(x) ((x) != INVALID_PAGE)
100 
101 #define UNMAPPED_GVA (~(gpa_t)0)
102 
103 /* KVM Hugepage definitions for x86 */
104 #define KVM_NR_PAGE_SIZES	3
105 #define KVM_HPAGE_GFN_SHIFT(x)	(((x) - 1) * 9)
106 #define KVM_HPAGE_SHIFT(x)	(PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
107 #define KVM_HPAGE_SIZE(x)	(1UL << KVM_HPAGE_SHIFT(x))
108 #define KVM_HPAGE_MASK(x)	(~(KVM_HPAGE_SIZE(x) - 1))
109 #define KVM_PAGES_PER_HPAGE(x)	(KVM_HPAGE_SIZE(x) / PAGE_SIZE)
110 
111 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
112 {
113 	/* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */
114 	return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
115 		(base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
116 }
117 
118 #define KVM_PERMILLE_MMU_PAGES 20
119 #define KVM_MIN_ALLOC_MMU_PAGES 64
120 #define KVM_MMU_HASH_SHIFT 12
121 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
122 #define KVM_MIN_FREE_MMU_PAGES 5
123 #define KVM_REFILL_PAGES 25
124 #define KVM_MAX_CPUID_ENTRIES 80
125 #define KVM_NR_FIXED_MTRR_REGION 88
126 #define KVM_NR_VAR_MTRR 8
127 
128 #define ASYNC_PF_PER_VCPU 64
129 
130 enum kvm_reg {
131 	VCPU_REGS_RAX = 0,
132 	VCPU_REGS_RCX = 1,
133 	VCPU_REGS_RDX = 2,
134 	VCPU_REGS_RBX = 3,
135 	VCPU_REGS_RSP = 4,
136 	VCPU_REGS_RBP = 5,
137 	VCPU_REGS_RSI = 6,
138 	VCPU_REGS_RDI = 7,
139 #ifdef CONFIG_X86_64
140 	VCPU_REGS_R8 = 8,
141 	VCPU_REGS_R9 = 9,
142 	VCPU_REGS_R10 = 10,
143 	VCPU_REGS_R11 = 11,
144 	VCPU_REGS_R12 = 12,
145 	VCPU_REGS_R13 = 13,
146 	VCPU_REGS_R14 = 14,
147 	VCPU_REGS_R15 = 15,
148 #endif
149 	VCPU_REGS_RIP,
150 	NR_VCPU_REGS
151 };
152 
153 enum kvm_reg_ex {
154 	VCPU_EXREG_PDPTR = NR_VCPU_REGS,
155 	VCPU_EXREG_CR3,
156 	VCPU_EXREG_RFLAGS,
157 	VCPU_EXREG_SEGMENTS,
158 };
159 
160 enum {
161 	VCPU_SREG_ES,
162 	VCPU_SREG_CS,
163 	VCPU_SREG_SS,
164 	VCPU_SREG_DS,
165 	VCPU_SREG_FS,
166 	VCPU_SREG_GS,
167 	VCPU_SREG_TR,
168 	VCPU_SREG_LDTR,
169 };
170 
171 #include <asm/kvm_emulate.h>
172 
173 #define KVM_NR_MEM_OBJS 40
174 
175 #define KVM_NR_DB_REGS	4
176 
177 #define DR6_BD		(1 << 13)
178 #define DR6_BS		(1 << 14)
179 #define DR6_RTM		(1 << 16)
180 #define DR6_FIXED_1	0xfffe0ff0
181 #define DR6_INIT	0xffff0ff0
182 #define DR6_VOLATILE	0x0001e00f
183 
184 #define DR7_BP_EN_MASK	0x000000ff
185 #define DR7_GE		(1 << 9)
186 #define DR7_GD		(1 << 13)
187 #define DR7_FIXED_1	0x00000400
188 #define DR7_VOLATILE	0xffff2bff
189 
190 #define PFERR_PRESENT_BIT 0
191 #define PFERR_WRITE_BIT 1
192 #define PFERR_USER_BIT 2
193 #define PFERR_RSVD_BIT 3
194 #define PFERR_FETCH_BIT 4
195 #define PFERR_PK_BIT 5
196 #define PFERR_GUEST_FINAL_BIT 32
197 #define PFERR_GUEST_PAGE_BIT 33
198 
199 #define PFERR_PRESENT_MASK (1U << PFERR_PRESENT_BIT)
200 #define PFERR_WRITE_MASK (1U << PFERR_WRITE_BIT)
201 #define PFERR_USER_MASK (1U << PFERR_USER_BIT)
202 #define PFERR_RSVD_MASK (1U << PFERR_RSVD_BIT)
203 #define PFERR_FETCH_MASK (1U << PFERR_FETCH_BIT)
204 #define PFERR_PK_MASK (1U << PFERR_PK_BIT)
205 #define PFERR_GUEST_FINAL_MASK (1ULL << PFERR_GUEST_FINAL_BIT)
206 #define PFERR_GUEST_PAGE_MASK (1ULL << PFERR_GUEST_PAGE_BIT)
207 
208 #define PFERR_NESTED_GUEST_PAGE (PFERR_GUEST_PAGE_MASK |	\
209 				 PFERR_WRITE_MASK |		\
210 				 PFERR_PRESENT_MASK)
211 
212 /*
213  * The mask used to denote special SPTEs, which can be either MMIO SPTEs or
214  * Access Tracking SPTEs. We use bit 62 instead of bit 63 to avoid conflicting
215  * with the SVE bit in EPT PTEs.
216  */
217 #define SPTE_SPECIAL_MASK (1ULL << 62)
218 
219 /* apic attention bits */
220 #define KVM_APIC_CHECK_VAPIC	0
221 /*
222  * The following bit is set with PV-EOI, unset on EOI.
223  * We detect PV-EOI changes by guest by comparing
224  * this bit with PV-EOI in guest memory.
225  * See the implementation in apic_update_pv_eoi.
226  */
227 #define KVM_APIC_PV_EOI_PENDING	1
228 
229 struct kvm_kernel_irq_routing_entry;
230 
231 /*
232  * We don't want allocation failures within the mmu code, so we preallocate
233  * enough memory for a single page fault in a cache.
234  */
235 struct kvm_mmu_memory_cache {
236 	int nobjs;
237 	void *objects[KVM_NR_MEM_OBJS];
238 };
239 
240 /*
241  * the pages used as guest page table on soft mmu are tracked by
242  * kvm_memory_slot.arch.gfn_track which is 16 bits, so the role bits used
243  * by indirect shadow page can not be more than 15 bits.
244  *
245  * Currently, we used 14 bits that are @level, @cr4_pae, @quadrant, @access,
246  * @nxe, @cr0_wp, @smep_andnot_wp and @smap_andnot_wp.
247  */
248 union kvm_mmu_page_role {
249 	unsigned word;
250 	struct {
251 		unsigned level:4;
252 		unsigned cr4_pae:1;
253 		unsigned quadrant:2;
254 		unsigned direct:1;
255 		unsigned access:3;
256 		unsigned invalid:1;
257 		unsigned nxe:1;
258 		unsigned cr0_wp:1;
259 		unsigned smep_andnot_wp:1;
260 		unsigned smap_andnot_wp:1;
261 		unsigned ad_disabled:1;
262 		unsigned guest_mode:1;
263 		unsigned :6;
264 
265 		/*
266 		 * This is left at the top of the word so that
267 		 * kvm_memslots_for_spte_role can extract it with a
268 		 * simple shift.  While there is room, give it a whole
269 		 * byte so it is also faster to load it from memory.
270 		 */
271 		unsigned smm:8;
272 	};
273 };
274 
275 struct kvm_rmap_head {
276 	unsigned long val;
277 };
278 
279 struct kvm_mmu_page {
280 	struct list_head link;
281 	struct hlist_node hash_link;
282 
283 	/*
284 	 * The following two entries are used to key the shadow page in the
285 	 * hash table.
286 	 */
287 	gfn_t gfn;
288 	union kvm_mmu_page_role role;
289 
290 	u64 *spt;
291 	/* hold the gfn of each spte inside spt */
292 	gfn_t *gfns;
293 	bool unsync;
294 	int root_count;          /* Currently serving as active root */
295 	unsigned int unsync_children;
296 	struct kvm_rmap_head parent_ptes; /* rmap pointers to parent sptes */
297 
298 	/* The page is obsolete if mmu_valid_gen != kvm->arch.mmu_valid_gen.  */
299 	unsigned long mmu_valid_gen;
300 
301 	DECLARE_BITMAP(unsync_child_bitmap, 512);
302 
303 #ifdef CONFIG_X86_32
304 	/*
305 	 * Used out of the mmu-lock to avoid reading spte values while an
306 	 * update is in progress; see the comments in __get_spte_lockless().
307 	 */
308 	int clear_spte_count;
309 #endif
310 
311 	/* Number of writes since the last time traversal visited this page.  */
312 	atomic_t write_flooding_count;
313 };
314 
315 struct kvm_pio_request {
316 	unsigned long count;
317 	int in;
318 	int port;
319 	int size;
320 };
321 
322 #define PT64_ROOT_MAX_LEVEL 5
323 
324 struct rsvd_bits_validate {
325 	u64 rsvd_bits_mask[2][PT64_ROOT_MAX_LEVEL];
326 	u64 bad_mt_xwr;
327 };
328 
329 /*
330  * x86 supports 4 paging modes (5-level 64-bit, 4-level 64-bit, 3-level 32-bit,
331  * and 2-level 32-bit).  The kvm_mmu structure abstracts the details of the
332  * current mmu mode.
333  */
334 struct kvm_mmu {
335 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long root);
336 	unsigned long (*get_cr3)(struct kvm_vcpu *vcpu);
337 	u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
338 	int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err,
339 			  bool prefault);
340 	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
341 				  struct x86_exception *fault);
342 	gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva, u32 access,
343 			    struct x86_exception *exception);
344 	gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
345 			       struct x86_exception *exception);
346 	int (*sync_page)(struct kvm_vcpu *vcpu,
347 			 struct kvm_mmu_page *sp);
348 	void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva);
349 	void (*update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
350 			   u64 *spte, const void *pte);
351 	hpa_t root_hpa;
352 	union kvm_mmu_page_role base_role;
353 	u8 root_level;
354 	u8 shadow_root_level;
355 	u8 ept_ad;
356 	bool direct_map;
357 
358 	/*
359 	 * Bitmap; bit set = permission fault
360 	 * Byte index: page fault error code [4:1]
361 	 * Bit index: pte permissions in ACC_* format
362 	 */
363 	u8 permissions[16];
364 
365 	/*
366 	* The pkru_mask indicates if protection key checks are needed.  It
367 	* consists of 16 domains indexed by page fault error code bits [4:1],
368 	* with PFEC.RSVD replaced by ACC_USER_MASK from the page tables.
369 	* Each domain has 2 bits which are ANDed with AD and WD from PKRU.
370 	*/
371 	u32 pkru_mask;
372 
373 	u64 *pae_root;
374 	u64 *lm_root;
375 
376 	/*
377 	 * check zero bits on shadow page table entries, these
378 	 * bits include not only hardware reserved bits but also
379 	 * the bits spte never used.
380 	 */
381 	struct rsvd_bits_validate shadow_zero_check;
382 
383 	struct rsvd_bits_validate guest_rsvd_check;
384 
385 	/* Can have large pages at levels 2..last_nonleaf_level-1. */
386 	u8 last_nonleaf_level;
387 
388 	bool nx;
389 
390 	u64 pdptrs[4]; /* pae */
391 };
392 
393 enum pmc_type {
394 	KVM_PMC_GP = 0,
395 	KVM_PMC_FIXED,
396 };
397 
398 struct kvm_pmc {
399 	enum pmc_type type;
400 	u8 idx;
401 	u64 counter;
402 	u64 eventsel;
403 	struct perf_event *perf_event;
404 	struct kvm_vcpu *vcpu;
405 };
406 
407 struct kvm_pmu {
408 	unsigned nr_arch_gp_counters;
409 	unsigned nr_arch_fixed_counters;
410 	unsigned available_event_types;
411 	u64 fixed_ctr_ctrl;
412 	u64 global_ctrl;
413 	u64 global_status;
414 	u64 global_ovf_ctrl;
415 	u64 counter_bitmask[2];
416 	u64 global_ctrl_mask;
417 	u64 reserved_bits;
418 	u8 version;
419 	struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
420 	struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED];
421 	struct irq_work irq_work;
422 	u64 reprogram_pmi;
423 };
424 
425 struct kvm_pmu_ops;
426 
427 enum {
428 	KVM_DEBUGREG_BP_ENABLED = 1,
429 	KVM_DEBUGREG_WONT_EXIT = 2,
430 	KVM_DEBUGREG_RELOAD = 4,
431 };
432 
433 struct kvm_mtrr_range {
434 	u64 base;
435 	u64 mask;
436 	struct list_head node;
437 };
438 
439 struct kvm_mtrr {
440 	struct kvm_mtrr_range var_ranges[KVM_NR_VAR_MTRR];
441 	mtrr_type fixed_ranges[KVM_NR_FIXED_MTRR_REGION];
442 	u64 deftype;
443 
444 	struct list_head head;
445 };
446 
447 /* Hyper-V SynIC timer */
448 struct kvm_vcpu_hv_stimer {
449 	struct hrtimer timer;
450 	int index;
451 	u64 config;
452 	u64 count;
453 	u64 exp_time;
454 	struct hv_message msg;
455 	bool msg_pending;
456 };
457 
458 /* Hyper-V synthetic interrupt controller (SynIC)*/
459 struct kvm_vcpu_hv_synic {
460 	u64 version;
461 	u64 control;
462 	u64 msg_page;
463 	u64 evt_page;
464 	atomic64_t sint[HV_SYNIC_SINT_COUNT];
465 	atomic_t sint_to_gsi[HV_SYNIC_SINT_COUNT];
466 	DECLARE_BITMAP(auto_eoi_bitmap, 256);
467 	DECLARE_BITMAP(vec_bitmap, 256);
468 	bool active;
469 	bool dont_zero_synic_pages;
470 };
471 
472 /* Hyper-V per vcpu emulation context */
473 struct kvm_vcpu_hv {
474 	u32 vp_index;
475 	u64 hv_vapic;
476 	s64 runtime_offset;
477 	struct kvm_vcpu_hv_synic synic;
478 	struct kvm_hyperv_exit exit;
479 	struct kvm_vcpu_hv_stimer stimer[HV_SYNIC_STIMER_COUNT];
480 	DECLARE_BITMAP(stimer_pending_bitmap, HV_SYNIC_STIMER_COUNT);
481 	cpumask_t tlb_lush;
482 };
483 
484 struct kvm_vcpu_arch {
485 	/*
486 	 * rip and regs accesses must go through
487 	 * kvm_{register,rip}_{read,write} functions.
488 	 */
489 	unsigned long regs[NR_VCPU_REGS];
490 	u32 regs_avail;
491 	u32 regs_dirty;
492 
493 	unsigned long cr0;
494 	unsigned long cr0_guest_owned_bits;
495 	unsigned long cr2;
496 	unsigned long cr3;
497 	unsigned long cr4;
498 	unsigned long cr4_guest_owned_bits;
499 	unsigned long cr8;
500 	u32 pkru;
501 	u32 hflags;
502 	u64 efer;
503 	u64 apic_base;
504 	struct kvm_lapic *apic;    /* kernel irqchip context */
505 	bool apicv_active;
506 	bool load_eoi_exitmap_pending;
507 	DECLARE_BITMAP(ioapic_handled_vectors, 256);
508 	unsigned long apic_attention;
509 	int32_t apic_arb_prio;
510 	int mp_state;
511 	u64 ia32_misc_enable_msr;
512 	u64 smbase;
513 	u64 smi_count;
514 	bool tpr_access_reporting;
515 	u64 ia32_xss;
516 	u64 microcode_version;
517 
518 	/*
519 	 * Paging state of the vcpu
520 	 *
521 	 * If the vcpu runs in guest mode with two level paging this still saves
522 	 * the paging mode of the l1 guest. This context is always used to
523 	 * handle faults.
524 	 */
525 	struct kvm_mmu mmu;
526 
527 	/*
528 	 * Paging state of an L2 guest (used for nested npt)
529 	 *
530 	 * This context will save all necessary information to walk page tables
531 	 * of the an L2 guest. This context is only initialized for page table
532 	 * walking and not for faulting since we never handle l2 page faults on
533 	 * the host.
534 	 */
535 	struct kvm_mmu nested_mmu;
536 
537 	/*
538 	 * Pointer to the mmu context currently used for
539 	 * gva_to_gpa translations.
540 	 */
541 	struct kvm_mmu *walk_mmu;
542 
543 	struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
544 	struct kvm_mmu_memory_cache mmu_page_cache;
545 	struct kvm_mmu_memory_cache mmu_page_header_cache;
546 
547 	/*
548 	 * QEMU userspace and the guest each have their own FPU state.
549 	 * In vcpu_run, we switch between the user and guest FPU contexts.
550 	 * While running a VCPU, the VCPU thread will have the guest FPU
551 	 * context.
552 	 *
553 	 * Note that while the PKRU state lives inside the fpu registers,
554 	 * it is switched out separately at VMENTER and VMEXIT time. The
555 	 * "guest_fpu" state here contains the guest FPU context, with the
556 	 * host PRKU bits.
557 	 */
558 	struct fpu user_fpu;
559 	struct fpu guest_fpu;
560 
561 	u64 xcr0;
562 	u64 guest_supported_xcr0;
563 	u32 guest_xstate_size;
564 
565 	struct kvm_pio_request pio;
566 	void *pio_data;
567 
568 	u8 event_exit_inst_len;
569 
570 	struct kvm_queued_exception {
571 		bool pending;
572 		bool injected;
573 		bool has_error_code;
574 		u8 nr;
575 		u32 error_code;
576 		u8 nested_apf;
577 	} exception;
578 
579 	struct kvm_queued_interrupt {
580 		bool injected;
581 		bool soft;
582 		u8 nr;
583 	} interrupt;
584 
585 	int halt_request; /* real mode on Intel only */
586 
587 	int cpuid_nent;
588 	struct kvm_cpuid_entry2 cpuid_entries[KVM_MAX_CPUID_ENTRIES];
589 
590 	int maxphyaddr;
591 
592 	/* emulate context */
593 
594 	struct x86_emulate_ctxt emulate_ctxt;
595 	bool emulate_regs_need_sync_to_vcpu;
596 	bool emulate_regs_need_sync_from_vcpu;
597 	int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
598 
599 	gpa_t time;
600 	struct pvclock_vcpu_time_info hv_clock;
601 	unsigned int hw_tsc_khz;
602 	struct gfn_to_hva_cache pv_time;
603 	bool pv_time_enabled;
604 	/* set guest stopped flag in pvclock flags field */
605 	bool pvclock_set_guest_stopped_request;
606 
607 	struct {
608 		u64 msr_val;
609 		u64 last_steal;
610 		struct gfn_to_hva_cache stime;
611 		struct kvm_steal_time steal;
612 	} st;
613 
614 	u64 tsc_offset;
615 	u64 last_guest_tsc;
616 	u64 last_host_tsc;
617 	u64 tsc_offset_adjustment;
618 	u64 this_tsc_nsec;
619 	u64 this_tsc_write;
620 	u64 this_tsc_generation;
621 	bool tsc_catchup;
622 	bool tsc_always_catchup;
623 	s8 virtual_tsc_shift;
624 	u32 virtual_tsc_mult;
625 	u32 virtual_tsc_khz;
626 	s64 ia32_tsc_adjust_msr;
627 	u64 tsc_scaling_ratio;
628 
629 	atomic_t nmi_queued;  /* unprocessed asynchronous NMIs */
630 	unsigned nmi_pending; /* NMI queued after currently running handler */
631 	bool nmi_injected;    /* Trying to inject an NMI this entry */
632 	bool smi_pending;    /* SMI queued after currently running handler */
633 
634 	struct kvm_mtrr mtrr_state;
635 	u64 pat;
636 
637 	unsigned switch_db_regs;
638 	unsigned long db[KVM_NR_DB_REGS];
639 	unsigned long dr6;
640 	unsigned long dr7;
641 	unsigned long eff_db[KVM_NR_DB_REGS];
642 	unsigned long guest_debug_dr7;
643 	u64 msr_platform_info;
644 	u64 msr_misc_features_enables;
645 
646 	u64 mcg_cap;
647 	u64 mcg_status;
648 	u64 mcg_ctl;
649 	u64 mcg_ext_ctl;
650 	u64 *mce_banks;
651 
652 	/* Cache MMIO info */
653 	u64 mmio_gva;
654 	unsigned access;
655 	gfn_t mmio_gfn;
656 	u64 mmio_gen;
657 
658 	struct kvm_pmu pmu;
659 
660 	/* used for guest single stepping over the given code position */
661 	unsigned long singlestep_rip;
662 
663 	struct kvm_vcpu_hv hyperv;
664 
665 	cpumask_var_t wbinvd_dirty_mask;
666 
667 	unsigned long last_retry_eip;
668 	unsigned long last_retry_addr;
669 
670 	struct {
671 		bool halted;
672 		gfn_t gfns[roundup_pow_of_two(ASYNC_PF_PER_VCPU)];
673 		struct gfn_to_hva_cache data;
674 		u64 msr_val;
675 		u32 id;
676 		bool send_user_only;
677 		u32 host_apf_reason;
678 		unsigned long nested_apf_token;
679 		bool delivery_as_pf_vmexit;
680 	} apf;
681 
682 	/* OSVW MSRs (AMD only) */
683 	struct {
684 		u64 length;
685 		u64 status;
686 	} osvw;
687 
688 	struct {
689 		u64 msr_val;
690 		struct gfn_to_hva_cache data;
691 	} pv_eoi;
692 
693 	/*
694 	 * Indicate whether the access faults on its page table in guest
695 	 * which is set when fix page fault and used to detect unhandeable
696 	 * instruction.
697 	 */
698 	bool write_fault_to_shadow_pgtable;
699 
700 	/* set at EPT violation at this point */
701 	unsigned long exit_qualification;
702 
703 	/* pv related host specific info */
704 	struct {
705 		bool pv_unhalted;
706 	} pv;
707 
708 	int pending_ioapic_eoi;
709 	int pending_external_vector;
710 
711 	/* GPA available */
712 	bool gpa_available;
713 	gpa_t gpa_val;
714 
715 	/* be preempted when it's in kernel-mode(cpl=0) */
716 	bool preempted_in_kernel;
717 
718 	/* Flush the L1 Data cache for L1TF mitigation on VMENTER */
719 	bool l1tf_flush_l1d;
720 };
721 
722 struct kvm_lpage_info {
723 	int disallow_lpage;
724 };
725 
726 struct kvm_arch_memory_slot {
727 	struct kvm_rmap_head *rmap[KVM_NR_PAGE_SIZES];
728 	struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
729 	unsigned short *gfn_track[KVM_PAGE_TRACK_MAX];
730 };
731 
732 /*
733  * We use as the mode the number of bits allocated in the LDR for the
734  * logical processor ID.  It happens that these are all powers of two.
735  * This makes it is very easy to detect cases where the APICs are
736  * configured for multiple modes; in that case, we cannot use the map and
737  * hence cannot use kvm_irq_delivery_to_apic_fast either.
738  */
739 #define KVM_APIC_MODE_XAPIC_CLUSTER          4
740 #define KVM_APIC_MODE_XAPIC_FLAT             8
741 #define KVM_APIC_MODE_X2APIC                16
742 
743 struct kvm_apic_map {
744 	struct rcu_head rcu;
745 	u8 mode;
746 	u32 max_apic_id;
747 	union {
748 		struct kvm_lapic *xapic_flat_map[8];
749 		struct kvm_lapic *xapic_cluster_map[16][4];
750 	};
751 	struct kvm_lapic *phys_map[];
752 };
753 
754 /* Hyper-V emulation context */
755 struct kvm_hv {
756 	struct mutex hv_lock;
757 	u64 hv_guest_os_id;
758 	u64 hv_hypercall;
759 	u64 hv_tsc_page;
760 
761 	/* Hyper-v based guest crash (NT kernel bugcheck) parameters */
762 	u64 hv_crash_param[HV_X64_MSR_CRASH_PARAMS];
763 	u64 hv_crash_ctl;
764 
765 	HV_REFERENCE_TSC_PAGE tsc_ref;
766 
767 	struct idr conn_to_evt;
768 
769 	u64 hv_reenlightenment_control;
770 	u64 hv_tsc_emulation_control;
771 	u64 hv_tsc_emulation_status;
772 };
773 
774 enum kvm_irqchip_mode {
775 	KVM_IRQCHIP_NONE,
776 	KVM_IRQCHIP_KERNEL,       /* created with KVM_CREATE_IRQCHIP */
777 	KVM_IRQCHIP_SPLIT,        /* created with KVM_CAP_SPLIT_IRQCHIP */
778 };
779 
780 struct kvm_arch {
781 	unsigned int n_used_mmu_pages;
782 	unsigned int n_requested_mmu_pages;
783 	unsigned int n_max_mmu_pages;
784 	unsigned int indirect_shadow_pages;
785 	unsigned long mmu_valid_gen;
786 	struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
787 	/*
788 	 * Hash table of struct kvm_mmu_page.
789 	 */
790 	struct list_head active_mmu_pages;
791 	struct list_head zapped_obsolete_pages;
792 	struct kvm_page_track_notifier_node mmu_sp_tracker;
793 	struct kvm_page_track_notifier_head track_notifier_head;
794 
795 	struct list_head assigned_dev_head;
796 	struct iommu_domain *iommu_domain;
797 	bool iommu_noncoherent;
798 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA
799 	atomic_t noncoherent_dma_count;
800 #define __KVM_HAVE_ARCH_ASSIGNED_DEVICE
801 	atomic_t assigned_device_count;
802 	struct kvm_pic *vpic;
803 	struct kvm_ioapic *vioapic;
804 	struct kvm_pit *vpit;
805 	atomic_t vapics_in_nmi_mode;
806 	struct mutex apic_map_lock;
807 	struct kvm_apic_map *apic_map;
808 
809 	bool apic_access_page_done;
810 
811 	gpa_t wall_clock;
812 
813 	bool mwait_in_guest;
814 	bool hlt_in_guest;
815 	bool pause_in_guest;
816 
817 	unsigned long irq_sources_bitmap;
818 	s64 kvmclock_offset;
819 	raw_spinlock_t tsc_write_lock;
820 	u64 last_tsc_nsec;
821 	u64 last_tsc_write;
822 	u32 last_tsc_khz;
823 	u64 cur_tsc_nsec;
824 	u64 cur_tsc_write;
825 	u64 cur_tsc_offset;
826 	u64 cur_tsc_generation;
827 	int nr_vcpus_matched_tsc;
828 
829 	spinlock_t pvclock_gtod_sync_lock;
830 	bool use_master_clock;
831 	u64 master_kernel_ns;
832 	u64 master_cycle_now;
833 	struct delayed_work kvmclock_update_work;
834 	struct delayed_work kvmclock_sync_work;
835 
836 	struct kvm_xen_hvm_config xen_hvm_config;
837 
838 	/* reads protected by irq_srcu, writes by irq_lock */
839 	struct hlist_head mask_notifier_list;
840 
841 	struct kvm_hv hyperv;
842 
843 	#ifdef CONFIG_KVM_MMU_AUDIT
844 	int audit_point;
845 	#endif
846 
847 	bool backwards_tsc_observed;
848 	bool boot_vcpu_runs_old_kvmclock;
849 	u32 bsp_vcpu_id;
850 
851 	u64 disabled_quirks;
852 
853 	enum kvm_irqchip_mode irqchip_mode;
854 	u8 nr_reserved_ioapic_pins;
855 
856 	bool disabled_lapic_found;
857 
858 	bool x2apic_format;
859 	bool x2apic_broadcast_quirk_disabled;
860 };
861 
862 struct kvm_vm_stat {
863 	ulong mmu_shadow_zapped;
864 	ulong mmu_pte_write;
865 	ulong mmu_pte_updated;
866 	ulong mmu_pde_zapped;
867 	ulong mmu_flooded;
868 	ulong mmu_recycled;
869 	ulong mmu_cache_miss;
870 	ulong mmu_unsync;
871 	ulong remote_tlb_flush;
872 	ulong lpages;
873 	ulong max_mmu_page_hash_collisions;
874 };
875 
876 struct kvm_vcpu_stat {
877 	u64 pf_fixed;
878 	u64 pf_guest;
879 	u64 tlb_flush;
880 	u64 invlpg;
881 
882 	u64 exits;
883 	u64 io_exits;
884 	u64 mmio_exits;
885 	u64 signal_exits;
886 	u64 irq_window_exits;
887 	u64 nmi_window_exits;
888 	u64 l1d_flush;
889 	u64 halt_exits;
890 	u64 halt_successful_poll;
891 	u64 halt_attempted_poll;
892 	u64 halt_poll_invalid;
893 	u64 halt_wakeup;
894 	u64 request_irq_exits;
895 	u64 irq_exits;
896 	u64 host_state_reload;
897 	u64 fpu_reload;
898 	u64 insn_emulation;
899 	u64 insn_emulation_fail;
900 	u64 hypercalls;
901 	u64 irq_injections;
902 	u64 nmi_injections;
903 	u64 req_event;
904 };
905 
906 struct x86_instruction_info;
907 
908 struct msr_data {
909 	bool host_initiated;
910 	u32 index;
911 	u64 data;
912 };
913 
914 struct kvm_lapic_irq {
915 	u32 vector;
916 	u16 delivery_mode;
917 	u16 dest_mode;
918 	bool level;
919 	u16 trig_mode;
920 	u32 shorthand;
921 	u32 dest_id;
922 	bool msi_redir_hint;
923 };
924 
925 struct kvm_x86_ops {
926 	int (*cpu_has_kvm_support)(void);          /* __init */
927 	int (*disabled_by_bios)(void);             /* __init */
928 	int (*hardware_enable)(void);
929 	void (*hardware_disable)(void);
930 	void (*check_processor_compatibility)(void *rtn);
931 	int (*hardware_setup)(void);               /* __init */
932 	void (*hardware_unsetup)(void);            /* __exit */
933 	bool (*cpu_has_accelerated_tpr)(void);
934 	bool (*has_emulated_msr)(int index);
935 	void (*cpuid_update)(struct kvm_vcpu *vcpu);
936 
937 	struct kvm *(*vm_alloc)(void);
938 	void (*vm_free)(struct kvm *);
939 	int (*vm_init)(struct kvm *kvm);
940 	void (*vm_destroy)(struct kvm *kvm);
941 
942 	/* Create, but do not attach this VCPU */
943 	struct kvm_vcpu *(*vcpu_create)(struct kvm *kvm, unsigned id);
944 	void (*vcpu_free)(struct kvm_vcpu *vcpu);
945 	void (*vcpu_reset)(struct kvm_vcpu *vcpu, bool init_event);
946 
947 	void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
948 	void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
949 	void (*vcpu_put)(struct kvm_vcpu *vcpu);
950 
951 	void (*update_bp_intercept)(struct kvm_vcpu *vcpu);
952 	int (*get_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
953 	int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
954 	u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
955 	void (*get_segment)(struct kvm_vcpu *vcpu,
956 			    struct kvm_segment *var, int seg);
957 	int (*get_cpl)(struct kvm_vcpu *vcpu);
958 	void (*set_segment)(struct kvm_vcpu *vcpu,
959 			    struct kvm_segment *var, int seg);
960 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
961 	void (*decache_cr0_guest_bits)(struct kvm_vcpu *vcpu);
962 	void (*decache_cr3)(struct kvm_vcpu *vcpu);
963 	void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu);
964 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
965 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
966 	int (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
967 	void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
968 	void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
969 	void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
970 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
971 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
972 	u64 (*get_dr6)(struct kvm_vcpu *vcpu);
973 	void (*set_dr6)(struct kvm_vcpu *vcpu, unsigned long value);
974 	void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu);
975 	void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
976 	void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
977 	unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
978 	void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
979 
980 	void (*tlb_flush)(struct kvm_vcpu *vcpu, bool invalidate_gpa);
981 
982 	void (*run)(struct kvm_vcpu *vcpu);
983 	int (*handle_exit)(struct kvm_vcpu *vcpu);
984 	void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
985 	void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
986 	u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu);
987 	void (*patch_hypercall)(struct kvm_vcpu *vcpu,
988 				unsigned char *hypercall_addr);
989 	void (*set_irq)(struct kvm_vcpu *vcpu);
990 	void (*set_nmi)(struct kvm_vcpu *vcpu);
991 	void (*queue_exception)(struct kvm_vcpu *vcpu);
992 	void (*cancel_injection)(struct kvm_vcpu *vcpu);
993 	int (*interrupt_allowed)(struct kvm_vcpu *vcpu);
994 	int (*nmi_allowed)(struct kvm_vcpu *vcpu);
995 	bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
996 	void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
997 	void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
998 	void (*enable_irq_window)(struct kvm_vcpu *vcpu);
999 	void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
1000 	bool (*get_enable_apicv)(struct kvm_vcpu *vcpu);
1001 	void (*refresh_apicv_exec_ctrl)(struct kvm_vcpu *vcpu);
1002 	void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
1003 	void (*hwapic_isr_update)(struct kvm_vcpu *vcpu, int isr);
1004 	void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
1005 	void (*set_virtual_apic_mode)(struct kvm_vcpu *vcpu);
1006 	void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu, hpa_t hpa);
1007 	void (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
1008 	int (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
1009 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
1010 	int (*set_identity_map_addr)(struct kvm *kvm, u64 ident_addr);
1011 	int (*get_tdp_level)(struct kvm_vcpu *vcpu);
1012 	u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
1013 	int (*get_lpage_level)(void);
1014 	bool (*rdtscp_supported)(void);
1015 	bool (*invpcid_supported)(void);
1016 
1017 	void (*set_tdp_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
1018 
1019 	void (*set_supported_cpuid)(u32 func, struct kvm_cpuid_entry2 *entry);
1020 
1021 	bool (*has_wbinvd_exit)(void);
1022 
1023 	u64 (*read_l1_tsc_offset)(struct kvm_vcpu *vcpu);
1024 	void (*write_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
1025 
1026 	void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2);
1027 
1028 	int (*check_intercept)(struct kvm_vcpu *vcpu,
1029 			       struct x86_instruction_info *info,
1030 			       enum x86_intercept_stage stage);
1031 	void (*handle_external_intr)(struct kvm_vcpu *vcpu);
1032 	bool (*mpx_supported)(void);
1033 	bool (*xsaves_supported)(void);
1034 	bool (*umip_emulated)(void);
1035 
1036 	int (*check_nested_events)(struct kvm_vcpu *vcpu, bool external_intr);
1037 
1038 	void (*sched_in)(struct kvm_vcpu *kvm, int cpu);
1039 
1040 	/*
1041 	 * Arch-specific dirty logging hooks. These hooks are only supposed to
1042 	 * be valid if the specific arch has hardware-accelerated dirty logging
1043 	 * mechanism. Currently only for PML on VMX.
1044 	 *
1045 	 *  - slot_enable_log_dirty:
1046 	 *	called when enabling log dirty mode for the slot.
1047 	 *  - slot_disable_log_dirty:
1048 	 *	called when disabling log dirty mode for the slot.
1049 	 *	also called when slot is created with log dirty disabled.
1050 	 *  - flush_log_dirty:
1051 	 *	called before reporting dirty_bitmap to userspace.
1052 	 *  - enable_log_dirty_pt_masked:
1053 	 *	called when reenabling log dirty for the GFNs in the mask after
1054 	 *	corresponding bits are cleared in slot->dirty_bitmap.
1055 	 */
1056 	void (*slot_enable_log_dirty)(struct kvm *kvm,
1057 				      struct kvm_memory_slot *slot);
1058 	void (*slot_disable_log_dirty)(struct kvm *kvm,
1059 				       struct kvm_memory_slot *slot);
1060 	void (*flush_log_dirty)(struct kvm *kvm);
1061 	void (*enable_log_dirty_pt_masked)(struct kvm *kvm,
1062 					   struct kvm_memory_slot *slot,
1063 					   gfn_t offset, unsigned long mask);
1064 	int (*write_log_dirty)(struct kvm_vcpu *vcpu);
1065 
1066 	/* pmu operations of sub-arch */
1067 	const struct kvm_pmu_ops *pmu_ops;
1068 
1069 	/*
1070 	 * Architecture specific hooks for vCPU blocking due to
1071 	 * HLT instruction.
1072 	 * Returns for .pre_block():
1073 	 *    - 0 means continue to block the vCPU.
1074 	 *    - 1 means we cannot block the vCPU since some event
1075 	 *        happens during this period, such as, 'ON' bit in
1076 	 *        posted-interrupts descriptor is set.
1077 	 */
1078 	int (*pre_block)(struct kvm_vcpu *vcpu);
1079 	void (*post_block)(struct kvm_vcpu *vcpu);
1080 
1081 	void (*vcpu_blocking)(struct kvm_vcpu *vcpu);
1082 	void (*vcpu_unblocking)(struct kvm_vcpu *vcpu);
1083 
1084 	int (*update_pi_irte)(struct kvm *kvm, unsigned int host_irq,
1085 			      uint32_t guest_irq, bool set);
1086 	void (*apicv_post_state_restore)(struct kvm_vcpu *vcpu);
1087 
1088 	int (*set_hv_timer)(struct kvm_vcpu *vcpu, u64 guest_deadline_tsc);
1089 	void (*cancel_hv_timer)(struct kvm_vcpu *vcpu);
1090 
1091 	void (*setup_mce)(struct kvm_vcpu *vcpu);
1092 
1093 	int (*smi_allowed)(struct kvm_vcpu *vcpu);
1094 	int (*pre_enter_smm)(struct kvm_vcpu *vcpu, char *smstate);
1095 	int (*pre_leave_smm)(struct kvm_vcpu *vcpu, u64 smbase);
1096 	int (*enable_smi_window)(struct kvm_vcpu *vcpu);
1097 
1098 	int (*mem_enc_op)(struct kvm *kvm, void __user *argp);
1099 	int (*mem_enc_reg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1100 	int (*mem_enc_unreg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1101 
1102 	int (*get_msr_feature)(struct kvm_msr_entry *entry);
1103 };
1104 
1105 struct kvm_arch_async_pf {
1106 	u32 token;
1107 	gfn_t gfn;
1108 	unsigned long cr3;
1109 	bool direct_map;
1110 };
1111 
1112 extern struct kvm_x86_ops *kvm_x86_ops;
1113 
1114 #define __KVM_HAVE_ARCH_VM_ALLOC
1115 static inline struct kvm *kvm_arch_alloc_vm(void)
1116 {
1117 	return kvm_x86_ops->vm_alloc();
1118 }
1119 
1120 static inline void kvm_arch_free_vm(struct kvm *kvm)
1121 {
1122 	return kvm_x86_ops->vm_free(kvm);
1123 }
1124 
1125 int kvm_mmu_module_init(void);
1126 void kvm_mmu_module_exit(void);
1127 
1128 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
1129 int kvm_mmu_create(struct kvm_vcpu *vcpu);
1130 void kvm_mmu_setup(struct kvm_vcpu *vcpu);
1131 void kvm_mmu_init_vm(struct kvm *kvm);
1132 void kvm_mmu_uninit_vm(struct kvm *kvm);
1133 void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
1134 		u64 dirty_mask, u64 nx_mask, u64 x_mask, u64 p_mask,
1135 		u64 acc_track_mask, u64 me_mask);
1136 
1137 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
1138 void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
1139 				      struct kvm_memory_slot *memslot);
1140 void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
1141 				   const struct kvm_memory_slot *memslot);
1142 void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
1143 				   struct kvm_memory_slot *memslot);
1144 void kvm_mmu_slot_largepage_remove_write_access(struct kvm *kvm,
1145 					struct kvm_memory_slot *memslot);
1146 void kvm_mmu_slot_set_dirty(struct kvm *kvm,
1147 			    struct kvm_memory_slot *memslot);
1148 void kvm_mmu_clear_dirty_pt_masked(struct kvm *kvm,
1149 				   struct kvm_memory_slot *slot,
1150 				   gfn_t gfn_offset, unsigned long mask);
1151 void kvm_mmu_zap_all(struct kvm *kvm);
1152 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, struct kvm_memslots *slots);
1153 unsigned int kvm_mmu_calculate_mmu_pages(struct kvm *kvm);
1154 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int kvm_nr_mmu_pages);
1155 
1156 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
1157 bool pdptrs_changed(struct kvm_vcpu *vcpu);
1158 
1159 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
1160 			  const void *val, int bytes);
1161 
1162 struct kvm_irq_mask_notifier {
1163 	void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
1164 	int irq;
1165 	struct hlist_node link;
1166 };
1167 
1168 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
1169 				    struct kvm_irq_mask_notifier *kimn);
1170 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
1171 				      struct kvm_irq_mask_notifier *kimn);
1172 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
1173 			     bool mask);
1174 
1175 extern bool tdp_enabled;
1176 
1177 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
1178 
1179 /* control of guest tsc rate supported? */
1180 extern bool kvm_has_tsc_control;
1181 /* maximum supported tsc_khz for guests */
1182 extern u32  kvm_max_guest_tsc_khz;
1183 /* number of bits of the fractional part of the TSC scaling ratio */
1184 extern u8   kvm_tsc_scaling_ratio_frac_bits;
1185 /* maximum allowed value of TSC scaling ratio */
1186 extern u64  kvm_max_tsc_scaling_ratio;
1187 /* 1ull << kvm_tsc_scaling_ratio_frac_bits */
1188 extern u64  kvm_default_tsc_scaling_ratio;
1189 
1190 extern u64 kvm_mce_cap_supported;
1191 
1192 enum emulation_result {
1193 	EMULATE_DONE,         /* no further processing */
1194 	EMULATE_USER_EXIT,    /* kvm_run ready for userspace exit */
1195 	EMULATE_FAIL,         /* can't emulate this instruction */
1196 };
1197 
1198 #define EMULTYPE_NO_DECODE	    (1 << 0)
1199 #define EMULTYPE_TRAP_UD	    (1 << 1)
1200 #define EMULTYPE_SKIP		    (1 << 2)
1201 #define EMULTYPE_RETRY		    (1 << 3)
1202 #define EMULTYPE_NO_REEXECUTE	    (1 << 4)
1203 #define EMULTYPE_NO_UD_ON_FAIL	    (1 << 5)
1204 #define EMULTYPE_VMWARE		    (1 << 6)
1205 int x86_emulate_instruction(struct kvm_vcpu *vcpu, unsigned long cr2,
1206 			    int emulation_type, void *insn, int insn_len);
1207 
1208 static inline int emulate_instruction(struct kvm_vcpu *vcpu,
1209 			int emulation_type)
1210 {
1211 	return x86_emulate_instruction(vcpu, 0,
1212 			emulation_type | EMULTYPE_NO_REEXECUTE, NULL, 0);
1213 }
1214 
1215 void kvm_enable_efer_bits(u64);
1216 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
1217 int kvm_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr);
1218 int kvm_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr);
1219 
1220 struct x86_emulate_ctxt;
1221 
1222 int kvm_fast_pio(struct kvm_vcpu *vcpu, int size, unsigned short port, int in);
1223 int kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
1224 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
1225 int kvm_vcpu_halt(struct kvm_vcpu *vcpu);
1226 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
1227 
1228 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
1229 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
1230 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector);
1231 
1232 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
1233 		    int reason, bool has_error_code, u32 error_code);
1234 
1235 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
1236 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
1237 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
1238 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
1239 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
1240 int kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
1241 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
1242 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
1243 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
1244 int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr);
1245 
1246 int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1247 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1248 
1249 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
1250 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
1251 bool kvm_rdpmc(struct kvm_vcpu *vcpu);
1252 
1253 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1254 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1255 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1256 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1257 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
1258 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1259 			    gfn_t gfn, void *data, int offset, int len,
1260 			    u32 access);
1261 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
1262 bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr);
1263 
1264 static inline int __kvm_irq_line_state(unsigned long *irq_state,
1265 				       int irq_source_id, int level)
1266 {
1267 	/* Logical OR for level trig interrupt */
1268 	if (level)
1269 		__set_bit(irq_source_id, irq_state);
1270 	else
1271 		__clear_bit(irq_source_id, irq_state);
1272 
1273 	return !!(*irq_state);
1274 }
1275 
1276 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
1277 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
1278 
1279 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
1280 
1281 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
1282 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
1283 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
1284 int kvm_mmu_load(struct kvm_vcpu *vcpu);
1285 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
1286 void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu);
1287 void kvm_mmu_free_roots(struct kvm_vcpu *vcpu);
1288 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1289 			   struct x86_exception *exception);
1290 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
1291 			      struct x86_exception *exception);
1292 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
1293 			       struct x86_exception *exception);
1294 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
1295 			       struct x86_exception *exception);
1296 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
1297 				struct x86_exception *exception);
1298 
1299 void kvm_vcpu_deactivate_apicv(struct kvm_vcpu *vcpu);
1300 
1301 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
1302 
1303 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva, u64 error_code,
1304 		       void *insn, int insn_len);
1305 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
1306 void kvm_mmu_new_cr3(struct kvm_vcpu *vcpu);
1307 
1308 void kvm_enable_tdp(void);
1309 void kvm_disable_tdp(void);
1310 
1311 static inline gpa_t translate_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1312 				  struct x86_exception *exception)
1313 {
1314 	return gpa;
1315 }
1316 
1317 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
1318 {
1319 	struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
1320 
1321 	return (struct kvm_mmu_page *)page_private(page);
1322 }
1323 
1324 static inline u16 kvm_read_ldt(void)
1325 {
1326 	u16 ldt;
1327 	asm("sldt %0" : "=g"(ldt));
1328 	return ldt;
1329 }
1330 
1331 static inline void kvm_load_ldt(u16 sel)
1332 {
1333 	asm("lldt %0" : : "rm"(sel));
1334 }
1335 
1336 #ifdef CONFIG_X86_64
1337 static inline unsigned long read_msr(unsigned long msr)
1338 {
1339 	u64 value;
1340 
1341 	rdmsrl(msr, value);
1342 	return value;
1343 }
1344 #endif
1345 
1346 static inline u32 get_rdx_init_val(void)
1347 {
1348 	return 0x600; /* P6 family */
1349 }
1350 
1351 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
1352 {
1353 	kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
1354 }
1355 
1356 #define TSS_IOPB_BASE_OFFSET 0x66
1357 #define TSS_BASE_SIZE 0x68
1358 #define TSS_IOPB_SIZE (65536 / 8)
1359 #define TSS_REDIRECTION_SIZE (256 / 8)
1360 #define RMODE_TSS_SIZE							\
1361 	(TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
1362 
1363 enum {
1364 	TASK_SWITCH_CALL = 0,
1365 	TASK_SWITCH_IRET = 1,
1366 	TASK_SWITCH_JMP = 2,
1367 	TASK_SWITCH_GATE = 3,
1368 };
1369 
1370 #define HF_GIF_MASK		(1 << 0)
1371 #define HF_HIF_MASK		(1 << 1)
1372 #define HF_VINTR_MASK		(1 << 2)
1373 #define HF_NMI_MASK		(1 << 3)
1374 #define HF_IRET_MASK		(1 << 4)
1375 #define HF_GUEST_MASK		(1 << 5) /* VCPU is in guest-mode */
1376 #define HF_SMM_MASK		(1 << 6)
1377 #define HF_SMM_INSIDE_NMI_MASK	(1 << 7)
1378 
1379 #define __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
1380 #define KVM_ADDRESS_SPACE_NUM 2
1381 
1382 #define kvm_arch_vcpu_memslots_id(vcpu) ((vcpu)->arch.hflags & HF_SMM_MASK ? 1 : 0)
1383 #define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, (role).smm)
1384 
1385 /*
1386  * Hardware virtualization extension instructions may fault if a
1387  * reboot turns off virtualization while processes are running.
1388  * Trap the fault and ignore the instruction if that happens.
1389  */
1390 asmlinkage void kvm_spurious_fault(void);
1391 
1392 #define ____kvm_handle_fault_on_reboot(insn, cleanup_insn)	\
1393 	"666: " insn "\n\t" \
1394 	"668: \n\t"                           \
1395 	".pushsection .fixup, \"ax\" \n" \
1396 	"667: \n\t" \
1397 	cleanup_insn "\n\t"		      \
1398 	"cmpb $0, kvm_rebooting \n\t"	      \
1399 	"jne 668b \n\t"      		      \
1400 	__ASM_SIZE(push) " $666b \n\t"	      \
1401 	"call kvm_spurious_fault \n\t"	      \
1402 	".popsection \n\t" \
1403 	_ASM_EXTABLE(666b, 667b)
1404 
1405 #define __kvm_handle_fault_on_reboot(insn)		\
1406 	____kvm_handle_fault_on_reboot(insn, "")
1407 
1408 #define KVM_ARCH_WANT_MMU_NOTIFIER
1409 int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
1410 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end);
1411 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
1412 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
1413 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
1414 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1415 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1416 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1417 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1418 void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event);
1419 void kvm_vcpu_reload_apic_access_page(struct kvm_vcpu *vcpu);
1420 
1421 u64 kvm_get_arch_capabilities(void);
1422 void kvm_define_shared_msr(unsigned index, u32 msr);
1423 int kvm_set_shared_msr(unsigned index, u64 val, u64 mask);
1424 
1425 u64 kvm_scale_tsc(struct kvm_vcpu *vcpu, u64 tsc);
1426 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc);
1427 
1428 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu);
1429 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1430 
1431 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
1432 void kvm_make_scan_ioapic_request(struct kvm *kvm);
1433 
1434 void kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1435 				     struct kvm_async_pf *work);
1436 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1437 				 struct kvm_async_pf *work);
1438 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1439 			       struct kvm_async_pf *work);
1440 bool kvm_arch_can_inject_async_page_present(struct kvm_vcpu *vcpu);
1441 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1442 
1443 int kvm_skip_emulated_instruction(struct kvm_vcpu *vcpu);
1444 int kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1445 
1446 int kvm_is_in_guest(void);
1447 
1448 int __x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa, u32 size);
1449 int x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa, u32 size);
1450 bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu);
1451 bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu);
1452 
1453 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
1454 			     struct kvm_vcpu **dest_vcpu);
1455 
1456 void kvm_set_msi_irq(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e,
1457 		     struct kvm_lapic_irq *irq);
1458 
1459 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu)
1460 {
1461 	if (kvm_x86_ops->vcpu_blocking)
1462 		kvm_x86_ops->vcpu_blocking(vcpu);
1463 }
1464 
1465 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu)
1466 {
1467 	if (kvm_x86_ops->vcpu_unblocking)
1468 		kvm_x86_ops->vcpu_unblocking(vcpu);
1469 }
1470 
1471 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
1472 
1473 static inline int kvm_cpu_get_apicid(int mps_cpu)
1474 {
1475 #ifdef CONFIG_X86_LOCAL_APIC
1476 	return default_cpu_present_to_apicid(mps_cpu);
1477 #else
1478 	WARN_ON_ONCE(1);
1479 	return BAD_APICID;
1480 #endif
1481 }
1482 
1483 #define put_smstate(type, buf, offset, val)                      \
1484 	*(type *)((buf) + (offset) - 0x7e00) = val
1485 
1486 #endif /* _ASM_X86_KVM_HOST_H */
1487