xref: /openbmc/linux/arch/x86/include/asm/kvm_host.h (revision 2359ccdd)
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 #include <asm/hyperv-tlfs.h>
38 
39 #define KVM_MAX_VCPUS 288
40 #define KVM_SOFT_MAX_VCPUS 240
41 #define KVM_MAX_VCPU_ID 1023
42 #define KVM_USER_MEM_SLOTS 509
43 /* memory slots that are not exposed to userspace */
44 #define KVM_PRIVATE_MEM_SLOTS 3
45 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
46 
47 #define KVM_HALT_POLL_NS_DEFAULT 200000
48 
49 #define KVM_IRQCHIP_NUM_PINS  KVM_IOAPIC_NUM_PINS
50 
51 /* x86-specific vcpu->requests bit members */
52 #define KVM_REQ_MIGRATE_TIMER		KVM_ARCH_REQ(0)
53 #define KVM_REQ_REPORT_TPR_ACCESS	KVM_ARCH_REQ(1)
54 #define KVM_REQ_TRIPLE_FAULT		KVM_ARCH_REQ(2)
55 #define KVM_REQ_MMU_SYNC		KVM_ARCH_REQ(3)
56 #define KVM_REQ_CLOCK_UPDATE		KVM_ARCH_REQ(4)
57 #define KVM_REQ_EVENT			KVM_ARCH_REQ(6)
58 #define KVM_REQ_APF_HALT		KVM_ARCH_REQ(7)
59 #define KVM_REQ_STEAL_UPDATE		KVM_ARCH_REQ(8)
60 #define KVM_REQ_NMI			KVM_ARCH_REQ(9)
61 #define KVM_REQ_PMU			KVM_ARCH_REQ(10)
62 #define KVM_REQ_PMI			KVM_ARCH_REQ(11)
63 #define KVM_REQ_SMI			KVM_ARCH_REQ(12)
64 #define KVM_REQ_MASTERCLOCK_UPDATE	KVM_ARCH_REQ(13)
65 #define KVM_REQ_MCLOCK_INPROGRESS \
66 	KVM_ARCH_REQ_FLAGS(14, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
67 #define KVM_REQ_SCAN_IOAPIC \
68 	KVM_ARCH_REQ_FLAGS(15, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
69 #define KVM_REQ_GLOBAL_CLOCK_UPDATE	KVM_ARCH_REQ(16)
70 #define KVM_REQ_APIC_PAGE_RELOAD \
71 	KVM_ARCH_REQ_FLAGS(17, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
72 #define KVM_REQ_HV_CRASH		KVM_ARCH_REQ(18)
73 #define KVM_REQ_IOAPIC_EOI_EXIT		KVM_ARCH_REQ(19)
74 #define KVM_REQ_HV_RESET		KVM_ARCH_REQ(20)
75 #define KVM_REQ_HV_EXIT			KVM_ARCH_REQ(21)
76 #define KVM_REQ_HV_STIMER		KVM_ARCH_REQ(22)
77 #define KVM_REQ_LOAD_EOI_EXITMAP	KVM_ARCH_REQ(23)
78 
79 #define CR0_RESERVED_BITS                                               \
80 	(~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \
81 			  | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \
82 			  | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG))
83 
84 #define CR3_PCID_INVD		 BIT_64(63)
85 #define CR4_RESERVED_BITS                                               \
86 	(~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\
87 			  | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE     \
88 			  | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR | X86_CR4_PCIDE \
89 			  | X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_FSGSBASE \
90 			  | X86_CR4_OSXMMEXCPT | X86_CR4_LA57 | X86_CR4_VMXE \
91 			  | X86_CR4_SMAP | X86_CR4_PKE | X86_CR4_UMIP))
92 
93 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR)
94 
95 
96 
97 #define INVALID_PAGE (~(hpa_t)0)
98 #define VALID_PAGE(x) ((x) != INVALID_PAGE)
99 
100 #define UNMAPPED_GVA (~(gpa_t)0)
101 
102 /* KVM Hugepage definitions for x86 */
103 #define KVM_NR_PAGE_SIZES	3
104 #define KVM_HPAGE_GFN_SHIFT(x)	(((x) - 1) * 9)
105 #define KVM_HPAGE_SHIFT(x)	(PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
106 #define KVM_HPAGE_SIZE(x)	(1UL << KVM_HPAGE_SHIFT(x))
107 #define KVM_HPAGE_MASK(x)	(~(KVM_HPAGE_SIZE(x) - 1))
108 #define KVM_PAGES_PER_HPAGE(x)	(KVM_HPAGE_SIZE(x) / PAGE_SIZE)
109 
110 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
111 {
112 	/* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */
113 	return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
114 		(base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
115 }
116 
117 #define KVM_PERMILLE_MMU_PAGES 20
118 #define KVM_MIN_ALLOC_MMU_PAGES 64
119 #define KVM_MMU_HASH_SHIFT 12
120 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
121 #define KVM_MIN_FREE_MMU_PAGES 5
122 #define KVM_REFILL_PAGES 25
123 #define KVM_MAX_CPUID_ENTRIES 80
124 #define KVM_NR_FIXED_MTRR_REGION 88
125 #define KVM_NR_VAR_MTRR 8
126 
127 #define ASYNC_PF_PER_VCPU 64
128 
129 enum kvm_reg {
130 	VCPU_REGS_RAX = 0,
131 	VCPU_REGS_RCX = 1,
132 	VCPU_REGS_RDX = 2,
133 	VCPU_REGS_RBX = 3,
134 	VCPU_REGS_RSP = 4,
135 	VCPU_REGS_RBP = 5,
136 	VCPU_REGS_RSI = 6,
137 	VCPU_REGS_RDI = 7,
138 #ifdef CONFIG_X86_64
139 	VCPU_REGS_R8 = 8,
140 	VCPU_REGS_R9 = 9,
141 	VCPU_REGS_R10 = 10,
142 	VCPU_REGS_R11 = 11,
143 	VCPU_REGS_R12 = 12,
144 	VCPU_REGS_R13 = 13,
145 	VCPU_REGS_R14 = 14,
146 	VCPU_REGS_R15 = 15,
147 #endif
148 	VCPU_REGS_RIP,
149 	NR_VCPU_REGS
150 };
151 
152 enum kvm_reg_ex {
153 	VCPU_EXREG_PDPTR = NR_VCPU_REGS,
154 	VCPU_EXREG_CR3,
155 	VCPU_EXREG_RFLAGS,
156 	VCPU_EXREG_SEGMENTS,
157 };
158 
159 enum {
160 	VCPU_SREG_ES,
161 	VCPU_SREG_CS,
162 	VCPU_SREG_SS,
163 	VCPU_SREG_DS,
164 	VCPU_SREG_FS,
165 	VCPU_SREG_GS,
166 	VCPU_SREG_TR,
167 	VCPU_SREG_LDTR,
168 };
169 
170 #include <asm/kvm_emulate.h>
171 
172 #define KVM_NR_MEM_OBJS 40
173 
174 #define KVM_NR_DB_REGS	4
175 
176 #define DR6_BD		(1 << 13)
177 #define DR6_BS		(1 << 14)
178 #define DR6_RTM		(1 << 16)
179 #define DR6_FIXED_1	0xfffe0ff0
180 #define DR6_INIT	0xffff0ff0
181 #define DR6_VOLATILE	0x0001e00f
182 
183 #define DR7_BP_EN_MASK	0x000000ff
184 #define DR7_GE		(1 << 9)
185 #define DR7_GD		(1 << 13)
186 #define DR7_FIXED_1	0x00000400
187 #define DR7_VOLATILE	0xffff2bff
188 
189 #define PFERR_PRESENT_BIT 0
190 #define PFERR_WRITE_BIT 1
191 #define PFERR_USER_BIT 2
192 #define PFERR_RSVD_BIT 3
193 #define PFERR_FETCH_BIT 4
194 #define PFERR_PK_BIT 5
195 #define PFERR_GUEST_FINAL_BIT 32
196 #define PFERR_GUEST_PAGE_BIT 33
197 
198 #define PFERR_PRESENT_MASK (1U << PFERR_PRESENT_BIT)
199 #define PFERR_WRITE_MASK (1U << PFERR_WRITE_BIT)
200 #define PFERR_USER_MASK (1U << PFERR_USER_BIT)
201 #define PFERR_RSVD_MASK (1U << PFERR_RSVD_BIT)
202 #define PFERR_FETCH_MASK (1U << PFERR_FETCH_BIT)
203 #define PFERR_PK_MASK (1U << PFERR_PK_BIT)
204 #define PFERR_GUEST_FINAL_MASK (1ULL << PFERR_GUEST_FINAL_BIT)
205 #define PFERR_GUEST_PAGE_MASK (1ULL << PFERR_GUEST_PAGE_BIT)
206 
207 #define PFERR_NESTED_GUEST_PAGE (PFERR_GUEST_PAGE_MASK |	\
208 				 PFERR_WRITE_MASK |		\
209 				 PFERR_PRESENT_MASK)
210 
211 /*
212  * The mask used to denote special SPTEs, which can be either MMIO SPTEs or
213  * Access Tracking SPTEs. We use bit 62 instead of bit 63 to avoid conflicting
214  * with the SVE bit in EPT PTEs.
215  */
216 #define SPTE_SPECIAL_MASK (1ULL << 62)
217 
218 /* apic attention bits */
219 #define KVM_APIC_CHECK_VAPIC	0
220 /*
221  * The following bit is set with PV-EOI, unset on EOI.
222  * We detect PV-EOI changes by guest by comparing
223  * this bit with PV-EOI in guest memory.
224  * See the implementation in apic_update_pv_eoi.
225  */
226 #define KVM_APIC_PV_EOI_PENDING	1
227 
228 struct kvm_kernel_irq_routing_entry;
229 
230 /*
231  * We don't want allocation failures within the mmu code, so we preallocate
232  * enough memory for a single page fault in a cache.
233  */
234 struct kvm_mmu_memory_cache {
235 	int nobjs;
236 	void *objects[KVM_NR_MEM_OBJS];
237 };
238 
239 /*
240  * the pages used as guest page table on soft mmu are tracked by
241  * kvm_memory_slot.arch.gfn_track which is 16 bits, so the role bits used
242  * by indirect shadow page can not be more than 15 bits.
243  *
244  * Currently, we used 14 bits that are @level, @cr4_pae, @quadrant, @access,
245  * @nxe, @cr0_wp, @smep_andnot_wp and @smap_andnot_wp.
246  */
247 union kvm_mmu_page_role {
248 	unsigned word;
249 	struct {
250 		unsigned level:4;
251 		unsigned cr4_pae:1;
252 		unsigned quadrant:2;
253 		unsigned direct:1;
254 		unsigned access:3;
255 		unsigned invalid:1;
256 		unsigned nxe:1;
257 		unsigned cr0_wp:1;
258 		unsigned smep_andnot_wp:1;
259 		unsigned smap_andnot_wp:1;
260 		unsigned ad_disabled:1;
261 		unsigned :7;
262 
263 		/*
264 		 * This is left at the top of the word so that
265 		 * kvm_memslots_for_spte_role can extract it with a
266 		 * simple shift.  While there is room, give it a whole
267 		 * byte so it is also faster to load it from memory.
268 		 */
269 		unsigned smm:8;
270 	};
271 };
272 
273 struct kvm_rmap_head {
274 	unsigned long val;
275 };
276 
277 struct kvm_mmu_page {
278 	struct list_head link;
279 	struct hlist_node hash_link;
280 
281 	/*
282 	 * The following two entries are used to key the shadow page in the
283 	 * hash table.
284 	 */
285 	gfn_t gfn;
286 	union kvm_mmu_page_role role;
287 
288 	u64 *spt;
289 	/* hold the gfn of each spte inside spt */
290 	gfn_t *gfns;
291 	bool unsync;
292 	int root_count;          /* Currently serving as active root */
293 	unsigned int unsync_children;
294 	struct kvm_rmap_head parent_ptes; /* rmap pointers to parent sptes */
295 
296 	/* The page is obsolete if mmu_valid_gen != kvm->arch.mmu_valid_gen.  */
297 	unsigned long mmu_valid_gen;
298 
299 	DECLARE_BITMAP(unsync_child_bitmap, 512);
300 
301 #ifdef CONFIG_X86_32
302 	/*
303 	 * Used out of the mmu-lock to avoid reading spte values while an
304 	 * update is in progress; see the comments in __get_spte_lockless().
305 	 */
306 	int clear_spte_count;
307 #endif
308 
309 	/* Number of writes since the last time traversal visited this page.  */
310 	atomic_t write_flooding_count;
311 };
312 
313 struct kvm_pio_request {
314 	unsigned long count;
315 	int in;
316 	int port;
317 	int size;
318 };
319 
320 #define PT64_ROOT_MAX_LEVEL 5
321 
322 struct rsvd_bits_validate {
323 	u64 rsvd_bits_mask[2][PT64_ROOT_MAX_LEVEL];
324 	u64 bad_mt_xwr;
325 };
326 
327 /*
328  * x86 supports 4 paging modes (5-level 64-bit, 4-level 64-bit, 3-level 32-bit,
329  * and 2-level 32-bit).  The kvm_mmu structure abstracts the details of the
330  * current mmu mode.
331  */
332 struct kvm_mmu {
333 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long root);
334 	unsigned long (*get_cr3)(struct kvm_vcpu *vcpu);
335 	u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
336 	int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err,
337 			  bool prefault);
338 	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
339 				  struct x86_exception *fault);
340 	gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva, u32 access,
341 			    struct x86_exception *exception);
342 	gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
343 			       struct x86_exception *exception);
344 	int (*sync_page)(struct kvm_vcpu *vcpu,
345 			 struct kvm_mmu_page *sp);
346 	void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva);
347 	void (*update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
348 			   u64 *spte, const void *pte);
349 	hpa_t root_hpa;
350 	union kvm_mmu_page_role base_role;
351 	u8 root_level;
352 	u8 shadow_root_level;
353 	u8 ept_ad;
354 	bool direct_map;
355 
356 	/*
357 	 * Bitmap; bit set = permission fault
358 	 * Byte index: page fault error code [4:1]
359 	 * Bit index: pte permissions in ACC_* format
360 	 */
361 	u8 permissions[16];
362 
363 	/*
364 	* The pkru_mask indicates if protection key checks are needed.  It
365 	* consists of 16 domains indexed by page fault error code bits [4:1],
366 	* with PFEC.RSVD replaced by ACC_USER_MASK from the page tables.
367 	* Each domain has 2 bits which are ANDed with AD and WD from PKRU.
368 	*/
369 	u32 pkru_mask;
370 
371 	u64 *pae_root;
372 	u64 *lm_root;
373 
374 	/*
375 	 * check zero bits on shadow page table entries, these
376 	 * bits include not only hardware reserved bits but also
377 	 * the bits spte never used.
378 	 */
379 	struct rsvd_bits_validate shadow_zero_check;
380 
381 	struct rsvd_bits_validate guest_rsvd_check;
382 
383 	/* Can have large pages at levels 2..last_nonleaf_level-1. */
384 	u8 last_nonleaf_level;
385 
386 	bool nx;
387 
388 	u64 pdptrs[4]; /* pae */
389 };
390 
391 enum pmc_type {
392 	KVM_PMC_GP = 0,
393 	KVM_PMC_FIXED,
394 };
395 
396 struct kvm_pmc {
397 	enum pmc_type type;
398 	u8 idx;
399 	u64 counter;
400 	u64 eventsel;
401 	struct perf_event *perf_event;
402 	struct kvm_vcpu *vcpu;
403 };
404 
405 struct kvm_pmu {
406 	unsigned nr_arch_gp_counters;
407 	unsigned nr_arch_fixed_counters;
408 	unsigned available_event_types;
409 	u64 fixed_ctr_ctrl;
410 	u64 global_ctrl;
411 	u64 global_status;
412 	u64 global_ovf_ctrl;
413 	u64 counter_bitmask[2];
414 	u64 global_ctrl_mask;
415 	u64 reserved_bits;
416 	u8 version;
417 	struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
418 	struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED];
419 	struct irq_work irq_work;
420 	u64 reprogram_pmi;
421 };
422 
423 struct kvm_pmu_ops;
424 
425 enum {
426 	KVM_DEBUGREG_BP_ENABLED = 1,
427 	KVM_DEBUGREG_WONT_EXIT = 2,
428 	KVM_DEBUGREG_RELOAD = 4,
429 };
430 
431 struct kvm_mtrr_range {
432 	u64 base;
433 	u64 mask;
434 	struct list_head node;
435 };
436 
437 struct kvm_mtrr {
438 	struct kvm_mtrr_range var_ranges[KVM_NR_VAR_MTRR];
439 	mtrr_type fixed_ranges[KVM_NR_FIXED_MTRR_REGION];
440 	u64 deftype;
441 
442 	struct list_head head;
443 };
444 
445 /* Hyper-V SynIC timer */
446 struct kvm_vcpu_hv_stimer {
447 	struct hrtimer timer;
448 	int index;
449 	u64 config;
450 	u64 count;
451 	u64 exp_time;
452 	struct hv_message msg;
453 	bool msg_pending;
454 };
455 
456 /* Hyper-V synthetic interrupt controller (SynIC)*/
457 struct kvm_vcpu_hv_synic {
458 	u64 version;
459 	u64 control;
460 	u64 msg_page;
461 	u64 evt_page;
462 	atomic64_t sint[HV_SYNIC_SINT_COUNT];
463 	atomic_t sint_to_gsi[HV_SYNIC_SINT_COUNT];
464 	DECLARE_BITMAP(auto_eoi_bitmap, 256);
465 	DECLARE_BITMAP(vec_bitmap, 256);
466 	bool active;
467 	bool dont_zero_synic_pages;
468 };
469 
470 /* Hyper-V per vcpu emulation context */
471 struct kvm_vcpu_hv {
472 	u32 vp_index;
473 	u64 hv_vapic;
474 	s64 runtime_offset;
475 	struct kvm_vcpu_hv_synic synic;
476 	struct kvm_hyperv_exit exit;
477 	struct kvm_vcpu_hv_stimer stimer[HV_SYNIC_STIMER_COUNT];
478 	DECLARE_BITMAP(stimer_pending_bitmap, HV_SYNIC_STIMER_COUNT);
479 };
480 
481 struct kvm_vcpu_arch {
482 	/*
483 	 * rip and regs accesses must go through
484 	 * kvm_{register,rip}_{read,write} functions.
485 	 */
486 	unsigned long regs[NR_VCPU_REGS];
487 	u32 regs_avail;
488 	u32 regs_dirty;
489 
490 	unsigned long cr0;
491 	unsigned long cr0_guest_owned_bits;
492 	unsigned long cr2;
493 	unsigned long cr3;
494 	unsigned long cr4;
495 	unsigned long cr4_guest_owned_bits;
496 	unsigned long cr8;
497 	u32 pkru;
498 	u32 hflags;
499 	u64 efer;
500 	u64 apic_base;
501 	struct kvm_lapic *apic;    /* kernel irqchip context */
502 	bool apicv_active;
503 	bool load_eoi_exitmap_pending;
504 	DECLARE_BITMAP(ioapic_handled_vectors, 256);
505 	unsigned long apic_attention;
506 	int32_t apic_arb_prio;
507 	int mp_state;
508 	u64 ia32_misc_enable_msr;
509 	u64 smbase;
510 	u64 smi_count;
511 	bool tpr_access_reporting;
512 	u64 ia32_xss;
513 	u64 microcode_version;
514 
515 	/*
516 	 * Paging state of the vcpu
517 	 *
518 	 * If the vcpu runs in guest mode with two level paging this still saves
519 	 * the paging mode of the l1 guest. This context is always used to
520 	 * handle faults.
521 	 */
522 	struct kvm_mmu mmu;
523 
524 	/*
525 	 * Paging state of an L2 guest (used for nested npt)
526 	 *
527 	 * This context will save all necessary information to walk page tables
528 	 * of the an L2 guest. This context is only initialized for page table
529 	 * walking and not for faulting since we never handle l2 page faults on
530 	 * the host.
531 	 */
532 	struct kvm_mmu nested_mmu;
533 
534 	/*
535 	 * Pointer to the mmu context currently used for
536 	 * gva_to_gpa translations.
537 	 */
538 	struct kvm_mmu *walk_mmu;
539 
540 	struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
541 	struct kvm_mmu_memory_cache mmu_page_cache;
542 	struct kvm_mmu_memory_cache mmu_page_header_cache;
543 
544 	/*
545 	 * QEMU userspace and the guest each have their own FPU state.
546 	 * In vcpu_run, we switch between the user and guest FPU contexts.
547 	 * While running a VCPU, the VCPU thread will have the guest FPU
548 	 * context.
549 	 *
550 	 * Note that while the PKRU state lives inside the fpu registers,
551 	 * it is switched out separately at VMENTER and VMEXIT time. The
552 	 * "guest_fpu" state here contains the guest FPU context, with the
553 	 * host PRKU bits.
554 	 */
555 	struct fpu user_fpu;
556 	struct fpu guest_fpu;
557 
558 	u64 xcr0;
559 	u64 guest_supported_xcr0;
560 	u32 guest_xstate_size;
561 
562 	struct kvm_pio_request pio;
563 	void *pio_data;
564 
565 	u8 event_exit_inst_len;
566 
567 	struct kvm_queued_exception {
568 		bool pending;
569 		bool injected;
570 		bool has_error_code;
571 		u8 nr;
572 		u32 error_code;
573 		u8 nested_apf;
574 	} exception;
575 
576 	struct kvm_queued_interrupt {
577 		bool injected;
578 		bool soft;
579 		u8 nr;
580 	} interrupt;
581 
582 	int halt_request; /* real mode on Intel only */
583 
584 	int cpuid_nent;
585 	struct kvm_cpuid_entry2 cpuid_entries[KVM_MAX_CPUID_ENTRIES];
586 
587 	int maxphyaddr;
588 
589 	/* emulate context */
590 
591 	struct x86_emulate_ctxt emulate_ctxt;
592 	bool emulate_regs_need_sync_to_vcpu;
593 	bool emulate_regs_need_sync_from_vcpu;
594 	int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
595 
596 	gpa_t time;
597 	struct pvclock_vcpu_time_info hv_clock;
598 	unsigned int hw_tsc_khz;
599 	struct gfn_to_hva_cache pv_time;
600 	bool pv_time_enabled;
601 	/* set guest stopped flag in pvclock flags field */
602 	bool pvclock_set_guest_stopped_request;
603 
604 	struct {
605 		u64 msr_val;
606 		u64 last_steal;
607 		struct gfn_to_hva_cache stime;
608 		struct kvm_steal_time steal;
609 	} st;
610 
611 	u64 tsc_offset;
612 	u64 last_guest_tsc;
613 	u64 last_host_tsc;
614 	u64 tsc_offset_adjustment;
615 	u64 this_tsc_nsec;
616 	u64 this_tsc_write;
617 	u64 this_tsc_generation;
618 	bool tsc_catchup;
619 	bool tsc_always_catchup;
620 	s8 virtual_tsc_shift;
621 	u32 virtual_tsc_mult;
622 	u32 virtual_tsc_khz;
623 	s64 ia32_tsc_adjust_msr;
624 	u64 tsc_scaling_ratio;
625 
626 	atomic_t nmi_queued;  /* unprocessed asynchronous NMIs */
627 	unsigned nmi_pending; /* NMI queued after currently running handler */
628 	bool nmi_injected;    /* Trying to inject an NMI this entry */
629 	bool smi_pending;    /* SMI queued after currently running handler */
630 
631 	struct kvm_mtrr mtrr_state;
632 	u64 pat;
633 
634 	unsigned switch_db_regs;
635 	unsigned long db[KVM_NR_DB_REGS];
636 	unsigned long dr6;
637 	unsigned long dr7;
638 	unsigned long eff_db[KVM_NR_DB_REGS];
639 	unsigned long guest_debug_dr7;
640 	u64 msr_platform_info;
641 	u64 msr_misc_features_enables;
642 
643 	u64 mcg_cap;
644 	u64 mcg_status;
645 	u64 mcg_ctl;
646 	u64 mcg_ext_ctl;
647 	u64 *mce_banks;
648 
649 	/* Cache MMIO info */
650 	u64 mmio_gva;
651 	unsigned access;
652 	gfn_t mmio_gfn;
653 	u64 mmio_gen;
654 
655 	struct kvm_pmu pmu;
656 
657 	/* used for guest single stepping over the given code position */
658 	unsigned long singlestep_rip;
659 
660 	struct kvm_vcpu_hv hyperv;
661 
662 	cpumask_var_t wbinvd_dirty_mask;
663 
664 	unsigned long last_retry_eip;
665 	unsigned long last_retry_addr;
666 
667 	struct {
668 		bool halted;
669 		gfn_t gfns[roundup_pow_of_two(ASYNC_PF_PER_VCPU)];
670 		struct gfn_to_hva_cache data;
671 		u64 msr_val;
672 		u32 id;
673 		bool send_user_only;
674 		u32 host_apf_reason;
675 		unsigned long nested_apf_token;
676 		bool delivery_as_pf_vmexit;
677 	} apf;
678 
679 	/* OSVW MSRs (AMD only) */
680 	struct {
681 		u64 length;
682 		u64 status;
683 	} osvw;
684 
685 	struct {
686 		u64 msr_val;
687 		struct gfn_to_hva_cache data;
688 	} pv_eoi;
689 
690 	/*
691 	 * Indicate whether the access faults on its page table in guest
692 	 * which is set when fix page fault and used to detect unhandeable
693 	 * instruction.
694 	 */
695 	bool write_fault_to_shadow_pgtable;
696 
697 	/* set at EPT violation at this point */
698 	unsigned long exit_qualification;
699 
700 	/* pv related host specific info */
701 	struct {
702 		bool pv_unhalted;
703 	} pv;
704 
705 	int pending_ioapic_eoi;
706 	int pending_external_vector;
707 
708 	/* GPA available */
709 	bool gpa_available;
710 	gpa_t gpa_val;
711 
712 	/* be preempted when it's in kernel-mode(cpl=0) */
713 	bool preempted_in_kernel;
714 };
715 
716 struct kvm_lpage_info {
717 	int disallow_lpage;
718 };
719 
720 struct kvm_arch_memory_slot {
721 	struct kvm_rmap_head *rmap[KVM_NR_PAGE_SIZES];
722 	struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
723 	unsigned short *gfn_track[KVM_PAGE_TRACK_MAX];
724 };
725 
726 /*
727  * We use as the mode the number of bits allocated in the LDR for the
728  * logical processor ID.  It happens that these are all powers of two.
729  * This makes it is very easy to detect cases where the APICs are
730  * configured for multiple modes; in that case, we cannot use the map and
731  * hence cannot use kvm_irq_delivery_to_apic_fast either.
732  */
733 #define KVM_APIC_MODE_XAPIC_CLUSTER          4
734 #define KVM_APIC_MODE_XAPIC_FLAT             8
735 #define KVM_APIC_MODE_X2APIC                16
736 
737 struct kvm_apic_map {
738 	struct rcu_head rcu;
739 	u8 mode;
740 	u32 max_apic_id;
741 	union {
742 		struct kvm_lapic *xapic_flat_map[8];
743 		struct kvm_lapic *xapic_cluster_map[16][4];
744 	};
745 	struct kvm_lapic *phys_map[];
746 };
747 
748 /* Hyper-V emulation context */
749 struct kvm_hv {
750 	struct mutex hv_lock;
751 	u64 hv_guest_os_id;
752 	u64 hv_hypercall;
753 	u64 hv_tsc_page;
754 
755 	/* Hyper-v based guest crash (NT kernel bugcheck) parameters */
756 	u64 hv_crash_param[HV_X64_MSR_CRASH_PARAMS];
757 	u64 hv_crash_ctl;
758 
759 	HV_REFERENCE_TSC_PAGE tsc_ref;
760 
761 	struct idr conn_to_evt;
762 
763 	u64 hv_reenlightenment_control;
764 	u64 hv_tsc_emulation_control;
765 	u64 hv_tsc_emulation_status;
766 };
767 
768 enum kvm_irqchip_mode {
769 	KVM_IRQCHIP_NONE,
770 	KVM_IRQCHIP_KERNEL,       /* created with KVM_CREATE_IRQCHIP */
771 	KVM_IRQCHIP_SPLIT,        /* created with KVM_CAP_SPLIT_IRQCHIP */
772 };
773 
774 struct kvm_arch {
775 	unsigned int n_used_mmu_pages;
776 	unsigned int n_requested_mmu_pages;
777 	unsigned int n_max_mmu_pages;
778 	unsigned int indirect_shadow_pages;
779 	unsigned long mmu_valid_gen;
780 	struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
781 	/*
782 	 * Hash table of struct kvm_mmu_page.
783 	 */
784 	struct list_head active_mmu_pages;
785 	struct list_head zapped_obsolete_pages;
786 	struct kvm_page_track_notifier_node mmu_sp_tracker;
787 	struct kvm_page_track_notifier_head track_notifier_head;
788 
789 	struct list_head assigned_dev_head;
790 	struct iommu_domain *iommu_domain;
791 	bool iommu_noncoherent;
792 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA
793 	atomic_t noncoherent_dma_count;
794 #define __KVM_HAVE_ARCH_ASSIGNED_DEVICE
795 	atomic_t assigned_device_count;
796 	struct kvm_pic *vpic;
797 	struct kvm_ioapic *vioapic;
798 	struct kvm_pit *vpit;
799 	atomic_t vapics_in_nmi_mode;
800 	struct mutex apic_map_lock;
801 	struct kvm_apic_map *apic_map;
802 
803 	bool apic_access_page_done;
804 
805 	gpa_t wall_clock;
806 
807 	bool mwait_in_guest;
808 	bool hlt_in_guest;
809 	bool pause_in_guest;
810 
811 	unsigned long irq_sources_bitmap;
812 	s64 kvmclock_offset;
813 	raw_spinlock_t tsc_write_lock;
814 	u64 last_tsc_nsec;
815 	u64 last_tsc_write;
816 	u32 last_tsc_khz;
817 	u64 cur_tsc_nsec;
818 	u64 cur_tsc_write;
819 	u64 cur_tsc_offset;
820 	u64 cur_tsc_generation;
821 	int nr_vcpus_matched_tsc;
822 
823 	spinlock_t pvclock_gtod_sync_lock;
824 	bool use_master_clock;
825 	u64 master_kernel_ns;
826 	u64 master_cycle_now;
827 	struct delayed_work kvmclock_update_work;
828 	struct delayed_work kvmclock_sync_work;
829 
830 	struct kvm_xen_hvm_config xen_hvm_config;
831 
832 	/* reads protected by irq_srcu, writes by irq_lock */
833 	struct hlist_head mask_notifier_list;
834 
835 	struct kvm_hv hyperv;
836 
837 	#ifdef CONFIG_KVM_MMU_AUDIT
838 	int audit_point;
839 	#endif
840 
841 	bool backwards_tsc_observed;
842 	bool boot_vcpu_runs_old_kvmclock;
843 	u32 bsp_vcpu_id;
844 
845 	u64 disabled_quirks;
846 
847 	enum kvm_irqchip_mode irqchip_mode;
848 	u8 nr_reserved_ioapic_pins;
849 
850 	bool disabled_lapic_found;
851 
852 	bool x2apic_format;
853 	bool x2apic_broadcast_quirk_disabled;
854 };
855 
856 struct kvm_vm_stat {
857 	ulong mmu_shadow_zapped;
858 	ulong mmu_pte_write;
859 	ulong mmu_pte_updated;
860 	ulong mmu_pde_zapped;
861 	ulong mmu_flooded;
862 	ulong mmu_recycled;
863 	ulong mmu_cache_miss;
864 	ulong mmu_unsync;
865 	ulong remote_tlb_flush;
866 	ulong lpages;
867 	ulong max_mmu_page_hash_collisions;
868 };
869 
870 struct kvm_vcpu_stat {
871 	u64 pf_fixed;
872 	u64 pf_guest;
873 	u64 tlb_flush;
874 	u64 invlpg;
875 
876 	u64 exits;
877 	u64 io_exits;
878 	u64 mmio_exits;
879 	u64 signal_exits;
880 	u64 irq_window_exits;
881 	u64 nmi_window_exits;
882 	u64 halt_exits;
883 	u64 halt_successful_poll;
884 	u64 halt_attempted_poll;
885 	u64 halt_poll_invalid;
886 	u64 halt_wakeup;
887 	u64 request_irq_exits;
888 	u64 irq_exits;
889 	u64 host_state_reload;
890 	u64 fpu_reload;
891 	u64 insn_emulation;
892 	u64 insn_emulation_fail;
893 	u64 hypercalls;
894 	u64 irq_injections;
895 	u64 nmi_injections;
896 	u64 req_event;
897 };
898 
899 struct x86_instruction_info;
900 
901 struct msr_data {
902 	bool host_initiated;
903 	u32 index;
904 	u64 data;
905 };
906 
907 struct kvm_lapic_irq {
908 	u32 vector;
909 	u16 delivery_mode;
910 	u16 dest_mode;
911 	bool level;
912 	u16 trig_mode;
913 	u32 shorthand;
914 	u32 dest_id;
915 	bool msi_redir_hint;
916 };
917 
918 struct kvm_x86_ops {
919 	int (*cpu_has_kvm_support)(void);          /* __init */
920 	int (*disabled_by_bios)(void);             /* __init */
921 	int (*hardware_enable)(void);
922 	void (*hardware_disable)(void);
923 	void (*check_processor_compatibility)(void *rtn);
924 	int (*hardware_setup)(void);               /* __init */
925 	void (*hardware_unsetup)(void);            /* __exit */
926 	bool (*cpu_has_accelerated_tpr)(void);
927 	bool (*cpu_has_high_real_mode_segbase)(void);
928 	void (*cpuid_update)(struct kvm_vcpu *vcpu);
929 
930 	struct kvm *(*vm_alloc)(void);
931 	void (*vm_free)(struct kvm *);
932 	int (*vm_init)(struct kvm *kvm);
933 	void (*vm_destroy)(struct kvm *kvm);
934 
935 	/* Create, but do not attach this VCPU */
936 	struct kvm_vcpu *(*vcpu_create)(struct kvm *kvm, unsigned id);
937 	void (*vcpu_free)(struct kvm_vcpu *vcpu);
938 	void (*vcpu_reset)(struct kvm_vcpu *vcpu, bool init_event);
939 
940 	void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
941 	void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
942 	void (*vcpu_put)(struct kvm_vcpu *vcpu);
943 
944 	void (*update_bp_intercept)(struct kvm_vcpu *vcpu);
945 	int (*get_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
946 	int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
947 	u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
948 	void (*get_segment)(struct kvm_vcpu *vcpu,
949 			    struct kvm_segment *var, int seg);
950 	int (*get_cpl)(struct kvm_vcpu *vcpu);
951 	void (*set_segment)(struct kvm_vcpu *vcpu,
952 			    struct kvm_segment *var, int seg);
953 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
954 	void (*decache_cr0_guest_bits)(struct kvm_vcpu *vcpu);
955 	void (*decache_cr3)(struct kvm_vcpu *vcpu);
956 	void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu);
957 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
958 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
959 	int (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
960 	void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
961 	void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
962 	void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
963 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
964 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
965 	u64 (*get_dr6)(struct kvm_vcpu *vcpu);
966 	void (*set_dr6)(struct kvm_vcpu *vcpu, unsigned long value);
967 	void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu);
968 	void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
969 	void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
970 	unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
971 	void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
972 
973 	void (*tlb_flush)(struct kvm_vcpu *vcpu, bool invalidate_gpa);
974 
975 	void (*run)(struct kvm_vcpu *vcpu);
976 	int (*handle_exit)(struct kvm_vcpu *vcpu);
977 	void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
978 	void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
979 	u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu);
980 	void (*patch_hypercall)(struct kvm_vcpu *vcpu,
981 				unsigned char *hypercall_addr);
982 	void (*set_irq)(struct kvm_vcpu *vcpu);
983 	void (*set_nmi)(struct kvm_vcpu *vcpu);
984 	void (*queue_exception)(struct kvm_vcpu *vcpu);
985 	void (*cancel_injection)(struct kvm_vcpu *vcpu);
986 	int (*interrupt_allowed)(struct kvm_vcpu *vcpu);
987 	int (*nmi_allowed)(struct kvm_vcpu *vcpu);
988 	bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
989 	void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
990 	void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
991 	void (*enable_irq_window)(struct kvm_vcpu *vcpu);
992 	void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
993 	bool (*get_enable_apicv)(struct kvm_vcpu *vcpu);
994 	void (*refresh_apicv_exec_ctrl)(struct kvm_vcpu *vcpu);
995 	void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
996 	void (*hwapic_isr_update)(struct kvm_vcpu *vcpu, int isr);
997 	void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
998 	void (*set_virtual_x2apic_mode)(struct kvm_vcpu *vcpu, bool set);
999 	void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu, hpa_t hpa);
1000 	void (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
1001 	int (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
1002 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
1003 	int (*set_identity_map_addr)(struct kvm *kvm, u64 ident_addr);
1004 	int (*get_tdp_level)(struct kvm_vcpu *vcpu);
1005 	u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
1006 	int (*get_lpage_level)(void);
1007 	bool (*rdtscp_supported)(void);
1008 	bool (*invpcid_supported)(void);
1009 
1010 	void (*set_tdp_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
1011 
1012 	void (*set_supported_cpuid)(u32 func, struct kvm_cpuid_entry2 *entry);
1013 
1014 	bool (*has_wbinvd_exit)(void);
1015 
1016 	void (*write_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
1017 
1018 	void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2);
1019 
1020 	int (*check_intercept)(struct kvm_vcpu *vcpu,
1021 			       struct x86_instruction_info *info,
1022 			       enum x86_intercept_stage stage);
1023 	void (*handle_external_intr)(struct kvm_vcpu *vcpu);
1024 	bool (*mpx_supported)(void);
1025 	bool (*xsaves_supported)(void);
1026 	bool (*umip_emulated)(void);
1027 
1028 	int (*check_nested_events)(struct kvm_vcpu *vcpu, bool external_intr);
1029 
1030 	void (*sched_in)(struct kvm_vcpu *kvm, int cpu);
1031 
1032 	/*
1033 	 * Arch-specific dirty logging hooks. These hooks are only supposed to
1034 	 * be valid if the specific arch has hardware-accelerated dirty logging
1035 	 * mechanism. Currently only for PML on VMX.
1036 	 *
1037 	 *  - slot_enable_log_dirty:
1038 	 *	called when enabling log dirty mode for the slot.
1039 	 *  - slot_disable_log_dirty:
1040 	 *	called when disabling log dirty mode for the slot.
1041 	 *	also called when slot is created with log dirty disabled.
1042 	 *  - flush_log_dirty:
1043 	 *	called before reporting dirty_bitmap to userspace.
1044 	 *  - enable_log_dirty_pt_masked:
1045 	 *	called when reenabling log dirty for the GFNs in the mask after
1046 	 *	corresponding bits are cleared in slot->dirty_bitmap.
1047 	 */
1048 	void (*slot_enable_log_dirty)(struct kvm *kvm,
1049 				      struct kvm_memory_slot *slot);
1050 	void (*slot_disable_log_dirty)(struct kvm *kvm,
1051 				       struct kvm_memory_slot *slot);
1052 	void (*flush_log_dirty)(struct kvm *kvm);
1053 	void (*enable_log_dirty_pt_masked)(struct kvm *kvm,
1054 					   struct kvm_memory_slot *slot,
1055 					   gfn_t offset, unsigned long mask);
1056 	int (*write_log_dirty)(struct kvm_vcpu *vcpu);
1057 
1058 	/* pmu operations of sub-arch */
1059 	const struct kvm_pmu_ops *pmu_ops;
1060 
1061 	/*
1062 	 * Architecture specific hooks for vCPU blocking due to
1063 	 * HLT instruction.
1064 	 * Returns for .pre_block():
1065 	 *    - 0 means continue to block the vCPU.
1066 	 *    - 1 means we cannot block the vCPU since some event
1067 	 *        happens during this period, such as, 'ON' bit in
1068 	 *        posted-interrupts descriptor is set.
1069 	 */
1070 	int (*pre_block)(struct kvm_vcpu *vcpu);
1071 	void (*post_block)(struct kvm_vcpu *vcpu);
1072 
1073 	void (*vcpu_blocking)(struct kvm_vcpu *vcpu);
1074 	void (*vcpu_unblocking)(struct kvm_vcpu *vcpu);
1075 
1076 	int (*update_pi_irte)(struct kvm *kvm, unsigned int host_irq,
1077 			      uint32_t guest_irq, bool set);
1078 	void (*apicv_post_state_restore)(struct kvm_vcpu *vcpu);
1079 
1080 	int (*set_hv_timer)(struct kvm_vcpu *vcpu, u64 guest_deadline_tsc);
1081 	void (*cancel_hv_timer)(struct kvm_vcpu *vcpu);
1082 
1083 	void (*setup_mce)(struct kvm_vcpu *vcpu);
1084 
1085 	int (*smi_allowed)(struct kvm_vcpu *vcpu);
1086 	int (*pre_enter_smm)(struct kvm_vcpu *vcpu, char *smstate);
1087 	int (*pre_leave_smm)(struct kvm_vcpu *vcpu, u64 smbase);
1088 	int (*enable_smi_window)(struct kvm_vcpu *vcpu);
1089 
1090 	int (*mem_enc_op)(struct kvm *kvm, void __user *argp);
1091 	int (*mem_enc_reg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1092 	int (*mem_enc_unreg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1093 
1094 	int (*get_msr_feature)(struct kvm_msr_entry *entry);
1095 };
1096 
1097 struct kvm_arch_async_pf {
1098 	u32 token;
1099 	gfn_t gfn;
1100 	unsigned long cr3;
1101 	bool direct_map;
1102 };
1103 
1104 extern struct kvm_x86_ops *kvm_x86_ops;
1105 
1106 #define __KVM_HAVE_ARCH_VM_ALLOC
1107 static inline struct kvm *kvm_arch_alloc_vm(void)
1108 {
1109 	return kvm_x86_ops->vm_alloc();
1110 }
1111 
1112 static inline void kvm_arch_free_vm(struct kvm *kvm)
1113 {
1114 	return kvm_x86_ops->vm_free(kvm);
1115 }
1116 
1117 int kvm_mmu_module_init(void);
1118 void kvm_mmu_module_exit(void);
1119 
1120 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
1121 int kvm_mmu_create(struct kvm_vcpu *vcpu);
1122 void kvm_mmu_setup(struct kvm_vcpu *vcpu);
1123 void kvm_mmu_init_vm(struct kvm *kvm);
1124 void kvm_mmu_uninit_vm(struct kvm *kvm);
1125 void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
1126 		u64 dirty_mask, u64 nx_mask, u64 x_mask, u64 p_mask,
1127 		u64 acc_track_mask, u64 me_mask);
1128 
1129 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
1130 void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
1131 				      struct kvm_memory_slot *memslot);
1132 void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
1133 				   const struct kvm_memory_slot *memslot);
1134 void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
1135 				   struct kvm_memory_slot *memslot);
1136 void kvm_mmu_slot_largepage_remove_write_access(struct kvm *kvm,
1137 					struct kvm_memory_slot *memslot);
1138 void kvm_mmu_slot_set_dirty(struct kvm *kvm,
1139 			    struct kvm_memory_slot *memslot);
1140 void kvm_mmu_clear_dirty_pt_masked(struct kvm *kvm,
1141 				   struct kvm_memory_slot *slot,
1142 				   gfn_t gfn_offset, unsigned long mask);
1143 void kvm_mmu_zap_all(struct kvm *kvm);
1144 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, struct kvm_memslots *slots);
1145 unsigned int kvm_mmu_calculate_mmu_pages(struct kvm *kvm);
1146 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int kvm_nr_mmu_pages);
1147 
1148 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
1149 bool pdptrs_changed(struct kvm_vcpu *vcpu);
1150 
1151 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
1152 			  const void *val, int bytes);
1153 
1154 struct kvm_irq_mask_notifier {
1155 	void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
1156 	int irq;
1157 	struct hlist_node link;
1158 };
1159 
1160 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
1161 				    struct kvm_irq_mask_notifier *kimn);
1162 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
1163 				      struct kvm_irq_mask_notifier *kimn);
1164 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
1165 			     bool mask);
1166 
1167 extern bool tdp_enabled;
1168 
1169 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
1170 
1171 /* control of guest tsc rate supported? */
1172 extern bool kvm_has_tsc_control;
1173 /* maximum supported tsc_khz for guests */
1174 extern u32  kvm_max_guest_tsc_khz;
1175 /* number of bits of the fractional part of the TSC scaling ratio */
1176 extern u8   kvm_tsc_scaling_ratio_frac_bits;
1177 /* maximum allowed value of TSC scaling ratio */
1178 extern u64  kvm_max_tsc_scaling_ratio;
1179 /* 1ull << kvm_tsc_scaling_ratio_frac_bits */
1180 extern u64  kvm_default_tsc_scaling_ratio;
1181 
1182 extern u64 kvm_mce_cap_supported;
1183 
1184 enum emulation_result {
1185 	EMULATE_DONE,         /* no further processing */
1186 	EMULATE_USER_EXIT,    /* kvm_run ready for userspace exit */
1187 	EMULATE_FAIL,         /* can't emulate this instruction */
1188 };
1189 
1190 #define EMULTYPE_NO_DECODE	    (1 << 0)
1191 #define EMULTYPE_TRAP_UD	    (1 << 1)
1192 #define EMULTYPE_SKIP		    (1 << 2)
1193 #define EMULTYPE_RETRY		    (1 << 3)
1194 #define EMULTYPE_NO_REEXECUTE	    (1 << 4)
1195 #define EMULTYPE_NO_UD_ON_FAIL	    (1 << 5)
1196 #define EMULTYPE_VMWARE		    (1 << 6)
1197 int x86_emulate_instruction(struct kvm_vcpu *vcpu, unsigned long cr2,
1198 			    int emulation_type, void *insn, int insn_len);
1199 
1200 static inline int emulate_instruction(struct kvm_vcpu *vcpu,
1201 			int emulation_type)
1202 {
1203 	return x86_emulate_instruction(vcpu, 0,
1204 			emulation_type | EMULTYPE_NO_REEXECUTE, NULL, 0);
1205 }
1206 
1207 void kvm_enable_efer_bits(u64);
1208 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
1209 int kvm_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr);
1210 int kvm_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr);
1211 
1212 struct x86_emulate_ctxt;
1213 
1214 int kvm_fast_pio(struct kvm_vcpu *vcpu, int size, unsigned short port, int in);
1215 int kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
1216 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
1217 int kvm_vcpu_halt(struct kvm_vcpu *vcpu);
1218 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
1219 
1220 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
1221 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
1222 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector);
1223 
1224 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
1225 		    int reason, bool has_error_code, u32 error_code);
1226 
1227 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
1228 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
1229 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
1230 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
1231 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
1232 int kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
1233 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
1234 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
1235 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
1236 int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr);
1237 
1238 int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1239 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1240 
1241 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
1242 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
1243 bool kvm_rdpmc(struct kvm_vcpu *vcpu);
1244 
1245 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1246 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1247 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1248 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1249 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
1250 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1251 			    gfn_t gfn, void *data, int offset, int len,
1252 			    u32 access);
1253 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
1254 bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr);
1255 
1256 static inline int __kvm_irq_line_state(unsigned long *irq_state,
1257 				       int irq_source_id, int level)
1258 {
1259 	/* Logical OR for level trig interrupt */
1260 	if (level)
1261 		__set_bit(irq_source_id, irq_state);
1262 	else
1263 		__clear_bit(irq_source_id, irq_state);
1264 
1265 	return !!(*irq_state);
1266 }
1267 
1268 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
1269 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
1270 
1271 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
1272 
1273 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
1274 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
1275 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
1276 int kvm_mmu_load(struct kvm_vcpu *vcpu);
1277 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
1278 void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu);
1279 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1280 			   struct x86_exception *exception);
1281 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
1282 			      struct x86_exception *exception);
1283 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
1284 			       struct x86_exception *exception);
1285 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
1286 			       struct x86_exception *exception);
1287 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
1288 				struct x86_exception *exception);
1289 
1290 void kvm_vcpu_deactivate_apicv(struct kvm_vcpu *vcpu);
1291 
1292 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
1293 
1294 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva, u64 error_code,
1295 		       void *insn, int insn_len);
1296 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
1297 void kvm_mmu_new_cr3(struct kvm_vcpu *vcpu);
1298 
1299 void kvm_enable_tdp(void);
1300 void kvm_disable_tdp(void);
1301 
1302 static inline gpa_t translate_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1303 				  struct x86_exception *exception)
1304 {
1305 	return gpa;
1306 }
1307 
1308 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
1309 {
1310 	struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
1311 
1312 	return (struct kvm_mmu_page *)page_private(page);
1313 }
1314 
1315 static inline u16 kvm_read_ldt(void)
1316 {
1317 	u16 ldt;
1318 	asm("sldt %0" : "=g"(ldt));
1319 	return ldt;
1320 }
1321 
1322 static inline void kvm_load_ldt(u16 sel)
1323 {
1324 	asm("lldt %0" : : "rm"(sel));
1325 }
1326 
1327 #ifdef CONFIG_X86_64
1328 static inline unsigned long read_msr(unsigned long msr)
1329 {
1330 	u64 value;
1331 
1332 	rdmsrl(msr, value);
1333 	return value;
1334 }
1335 #endif
1336 
1337 static inline u32 get_rdx_init_val(void)
1338 {
1339 	return 0x600; /* P6 family */
1340 }
1341 
1342 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
1343 {
1344 	kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
1345 }
1346 
1347 #define TSS_IOPB_BASE_OFFSET 0x66
1348 #define TSS_BASE_SIZE 0x68
1349 #define TSS_IOPB_SIZE (65536 / 8)
1350 #define TSS_REDIRECTION_SIZE (256 / 8)
1351 #define RMODE_TSS_SIZE							\
1352 	(TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
1353 
1354 enum {
1355 	TASK_SWITCH_CALL = 0,
1356 	TASK_SWITCH_IRET = 1,
1357 	TASK_SWITCH_JMP = 2,
1358 	TASK_SWITCH_GATE = 3,
1359 };
1360 
1361 #define HF_GIF_MASK		(1 << 0)
1362 #define HF_HIF_MASK		(1 << 1)
1363 #define HF_VINTR_MASK		(1 << 2)
1364 #define HF_NMI_MASK		(1 << 3)
1365 #define HF_IRET_MASK		(1 << 4)
1366 #define HF_GUEST_MASK		(1 << 5) /* VCPU is in guest-mode */
1367 #define HF_SMM_MASK		(1 << 6)
1368 #define HF_SMM_INSIDE_NMI_MASK	(1 << 7)
1369 
1370 #define __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
1371 #define KVM_ADDRESS_SPACE_NUM 2
1372 
1373 #define kvm_arch_vcpu_memslots_id(vcpu) ((vcpu)->arch.hflags & HF_SMM_MASK ? 1 : 0)
1374 #define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, (role).smm)
1375 
1376 /*
1377  * Hardware virtualization extension instructions may fault if a
1378  * reboot turns off virtualization while processes are running.
1379  * Trap the fault and ignore the instruction if that happens.
1380  */
1381 asmlinkage void kvm_spurious_fault(void);
1382 
1383 #define ____kvm_handle_fault_on_reboot(insn, cleanup_insn)	\
1384 	"666: " insn "\n\t" \
1385 	"668: \n\t"                           \
1386 	".pushsection .fixup, \"ax\" \n" \
1387 	"667: \n\t" \
1388 	cleanup_insn "\n\t"		      \
1389 	"cmpb $0, kvm_rebooting \n\t"	      \
1390 	"jne 668b \n\t"      		      \
1391 	__ASM_SIZE(push) " $666b \n\t"	      \
1392 	"call kvm_spurious_fault \n\t"	      \
1393 	".popsection \n\t" \
1394 	_ASM_EXTABLE(666b, 667b)
1395 
1396 #define __kvm_handle_fault_on_reboot(insn)		\
1397 	____kvm_handle_fault_on_reboot(insn, "")
1398 
1399 #define KVM_ARCH_WANT_MMU_NOTIFIER
1400 int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
1401 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end);
1402 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
1403 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
1404 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
1405 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1406 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1407 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1408 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1409 void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event);
1410 void kvm_vcpu_reload_apic_access_page(struct kvm_vcpu *vcpu);
1411 
1412 void kvm_define_shared_msr(unsigned index, u32 msr);
1413 int kvm_set_shared_msr(unsigned index, u64 val, u64 mask);
1414 
1415 u64 kvm_scale_tsc(struct kvm_vcpu *vcpu, u64 tsc);
1416 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc);
1417 
1418 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu);
1419 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1420 
1421 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
1422 void kvm_make_scan_ioapic_request(struct kvm *kvm);
1423 
1424 void kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1425 				     struct kvm_async_pf *work);
1426 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1427 				 struct kvm_async_pf *work);
1428 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1429 			       struct kvm_async_pf *work);
1430 bool kvm_arch_can_inject_async_page_present(struct kvm_vcpu *vcpu);
1431 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1432 
1433 int kvm_skip_emulated_instruction(struct kvm_vcpu *vcpu);
1434 int kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1435 
1436 int kvm_is_in_guest(void);
1437 
1438 int __x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa, u32 size);
1439 int x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa, u32 size);
1440 bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu);
1441 bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu);
1442 
1443 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
1444 			     struct kvm_vcpu **dest_vcpu);
1445 
1446 void kvm_set_msi_irq(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e,
1447 		     struct kvm_lapic_irq *irq);
1448 
1449 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu)
1450 {
1451 	if (kvm_x86_ops->vcpu_blocking)
1452 		kvm_x86_ops->vcpu_blocking(vcpu);
1453 }
1454 
1455 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu)
1456 {
1457 	if (kvm_x86_ops->vcpu_unblocking)
1458 		kvm_x86_ops->vcpu_unblocking(vcpu);
1459 }
1460 
1461 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
1462 
1463 static inline int kvm_cpu_get_apicid(int mps_cpu)
1464 {
1465 #ifdef CONFIG_X86_LOCAL_APIC
1466 	return default_cpu_present_to_apicid(mps_cpu);
1467 #else
1468 	WARN_ON_ONCE(1);
1469 	return BAD_APICID;
1470 #endif
1471 }
1472 
1473 #define put_smstate(type, buf, offset, val)                      \
1474 	*(type *)((buf) + (offset) - 0x7e00) = val
1475 
1476 #endif /* _ASM_X86_KVM_HOST_H */
1477