xref: /openbmc/linux/arch/x86/include/asm/kvm_host.h (revision 44e2971f)
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 guest_mode:1;
262 		unsigned :6;
263 
264 		/*
265 		 * This is left at the top of the word so that
266 		 * kvm_memslots_for_spte_role can extract it with a
267 		 * simple shift.  While there is room, give it a whole
268 		 * byte so it is also faster to load it from memory.
269 		 */
270 		unsigned smm:8;
271 	};
272 };
273 
274 struct kvm_rmap_head {
275 	unsigned long val;
276 };
277 
278 struct kvm_mmu_page {
279 	struct list_head link;
280 	struct hlist_node hash_link;
281 
282 	/*
283 	 * The following two entries are used to key the shadow page in the
284 	 * hash table.
285 	 */
286 	gfn_t gfn;
287 	union kvm_mmu_page_role role;
288 
289 	u64 *spt;
290 	/* hold the gfn of each spte inside spt */
291 	gfn_t *gfns;
292 	bool unsync;
293 	int root_count;          /* Currently serving as active root */
294 	unsigned int unsync_children;
295 	struct kvm_rmap_head parent_ptes; /* rmap pointers to parent sptes */
296 
297 	/* The page is obsolete if mmu_valid_gen != kvm->arch.mmu_valid_gen.  */
298 	unsigned long mmu_valid_gen;
299 
300 	DECLARE_BITMAP(unsync_child_bitmap, 512);
301 
302 #ifdef CONFIG_X86_32
303 	/*
304 	 * Used out of the mmu-lock to avoid reading spte values while an
305 	 * update is in progress; see the comments in __get_spte_lockless().
306 	 */
307 	int clear_spte_count;
308 #endif
309 
310 	/* Number of writes since the last time traversal visited this page.  */
311 	atomic_t write_flooding_count;
312 };
313 
314 struct kvm_pio_request {
315 	unsigned long count;
316 	int in;
317 	int port;
318 	int size;
319 };
320 
321 #define PT64_ROOT_MAX_LEVEL 5
322 
323 struct rsvd_bits_validate {
324 	u64 rsvd_bits_mask[2][PT64_ROOT_MAX_LEVEL];
325 	u64 bad_mt_xwr;
326 };
327 
328 /*
329  * x86 supports 4 paging modes (5-level 64-bit, 4-level 64-bit, 3-level 32-bit,
330  * and 2-level 32-bit).  The kvm_mmu structure abstracts the details of the
331  * current mmu mode.
332  */
333 struct kvm_mmu {
334 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long root);
335 	unsigned long (*get_cr3)(struct kvm_vcpu *vcpu);
336 	u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
337 	int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err,
338 			  bool prefault);
339 	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
340 				  struct x86_exception *fault);
341 	gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva, u32 access,
342 			    struct x86_exception *exception);
343 	gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
344 			       struct x86_exception *exception);
345 	int (*sync_page)(struct kvm_vcpu *vcpu,
346 			 struct kvm_mmu_page *sp);
347 	void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva);
348 	void (*update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
349 			   u64 *spte, const void *pte);
350 	hpa_t root_hpa;
351 	union kvm_mmu_page_role base_role;
352 	u8 root_level;
353 	u8 shadow_root_level;
354 	u8 ept_ad;
355 	bool direct_map;
356 
357 	/*
358 	 * Bitmap; bit set = permission fault
359 	 * Byte index: page fault error code [4:1]
360 	 * Bit index: pte permissions in ACC_* format
361 	 */
362 	u8 permissions[16];
363 
364 	/*
365 	* The pkru_mask indicates if protection key checks are needed.  It
366 	* consists of 16 domains indexed by page fault error code bits [4:1],
367 	* with PFEC.RSVD replaced by ACC_USER_MASK from the page tables.
368 	* Each domain has 2 bits which are ANDed with AD and WD from PKRU.
369 	*/
370 	u32 pkru_mask;
371 
372 	u64 *pae_root;
373 	u64 *lm_root;
374 
375 	/*
376 	 * check zero bits on shadow page table entries, these
377 	 * bits include not only hardware reserved bits but also
378 	 * the bits spte never used.
379 	 */
380 	struct rsvd_bits_validate shadow_zero_check;
381 
382 	struct rsvd_bits_validate guest_rsvd_check;
383 
384 	/* Can have large pages at levels 2..last_nonleaf_level-1. */
385 	u8 last_nonleaf_level;
386 
387 	bool nx;
388 
389 	u64 pdptrs[4]; /* pae */
390 };
391 
392 enum pmc_type {
393 	KVM_PMC_GP = 0,
394 	KVM_PMC_FIXED,
395 };
396 
397 struct kvm_pmc {
398 	enum pmc_type type;
399 	u8 idx;
400 	u64 counter;
401 	u64 eventsel;
402 	struct perf_event *perf_event;
403 	struct kvm_vcpu *vcpu;
404 };
405 
406 struct kvm_pmu {
407 	unsigned nr_arch_gp_counters;
408 	unsigned nr_arch_fixed_counters;
409 	unsigned available_event_types;
410 	u64 fixed_ctr_ctrl;
411 	u64 global_ctrl;
412 	u64 global_status;
413 	u64 global_ovf_ctrl;
414 	u64 counter_bitmask[2];
415 	u64 global_ctrl_mask;
416 	u64 reserved_bits;
417 	u8 version;
418 	struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
419 	struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED];
420 	struct irq_work irq_work;
421 	u64 reprogram_pmi;
422 };
423 
424 struct kvm_pmu_ops;
425 
426 enum {
427 	KVM_DEBUGREG_BP_ENABLED = 1,
428 	KVM_DEBUGREG_WONT_EXIT = 2,
429 	KVM_DEBUGREG_RELOAD = 4,
430 };
431 
432 struct kvm_mtrr_range {
433 	u64 base;
434 	u64 mask;
435 	struct list_head node;
436 };
437 
438 struct kvm_mtrr {
439 	struct kvm_mtrr_range var_ranges[KVM_NR_VAR_MTRR];
440 	mtrr_type fixed_ranges[KVM_NR_FIXED_MTRR_REGION];
441 	u64 deftype;
442 
443 	struct list_head head;
444 };
445 
446 /* Hyper-V SynIC timer */
447 struct kvm_vcpu_hv_stimer {
448 	struct hrtimer timer;
449 	int index;
450 	u64 config;
451 	u64 count;
452 	u64 exp_time;
453 	struct hv_message msg;
454 	bool msg_pending;
455 };
456 
457 /* Hyper-V synthetic interrupt controller (SynIC)*/
458 struct kvm_vcpu_hv_synic {
459 	u64 version;
460 	u64 control;
461 	u64 msg_page;
462 	u64 evt_page;
463 	atomic64_t sint[HV_SYNIC_SINT_COUNT];
464 	atomic_t sint_to_gsi[HV_SYNIC_SINT_COUNT];
465 	DECLARE_BITMAP(auto_eoi_bitmap, 256);
466 	DECLARE_BITMAP(vec_bitmap, 256);
467 	bool active;
468 	bool dont_zero_synic_pages;
469 };
470 
471 /* Hyper-V per vcpu emulation context */
472 struct kvm_vcpu_hv {
473 	u32 vp_index;
474 	u64 hv_vapic;
475 	s64 runtime_offset;
476 	struct kvm_vcpu_hv_synic synic;
477 	struct kvm_hyperv_exit exit;
478 	struct kvm_vcpu_hv_stimer stimer[HV_SYNIC_STIMER_COUNT];
479 	DECLARE_BITMAP(stimer_pending_bitmap, HV_SYNIC_STIMER_COUNT);
480 	cpumask_t tlb_lush;
481 };
482 
483 struct kvm_vcpu_arch {
484 	/*
485 	 * rip and regs accesses must go through
486 	 * kvm_{register,rip}_{read,write} functions.
487 	 */
488 	unsigned long regs[NR_VCPU_REGS];
489 	u32 regs_avail;
490 	u32 regs_dirty;
491 
492 	unsigned long cr0;
493 	unsigned long cr0_guest_owned_bits;
494 	unsigned long cr2;
495 	unsigned long cr3;
496 	unsigned long cr4;
497 	unsigned long cr4_guest_owned_bits;
498 	unsigned long cr8;
499 	u32 pkru;
500 	u32 hflags;
501 	u64 efer;
502 	u64 apic_base;
503 	struct kvm_lapic *apic;    /* kernel irqchip context */
504 	bool apicv_active;
505 	bool load_eoi_exitmap_pending;
506 	DECLARE_BITMAP(ioapic_handled_vectors, 256);
507 	unsigned long apic_attention;
508 	int32_t apic_arb_prio;
509 	int mp_state;
510 	u64 ia32_misc_enable_msr;
511 	u64 smbase;
512 	u64 smi_count;
513 	bool tpr_access_reporting;
514 	u64 ia32_xss;
515 	u64 microcode_version;
516 
517 	/*
518 	 * Paging state of the vcpu
519 	 *
520 	 * If the vcpu runs in guest mode with two level paging this still saves
521 	 * the paging mode of the l1 guest. This context is always used to
522 	 * handle faults.
523 	 */
524 	struct kvm_mmu mmu;
525 
526 	/*
527 	 * Paging state of an L2 guest (used for nested npt)
528 	 *
529 	 * This context will save all necessary information to walk page tables
530 	 * of the an L2 guest. This context is only initialized for page table
531 	 * walking and not for faulting since we never handle l2 page faults on
532 	 * the host.
533 	 */
534 	struct kvm_mmu nested_mmu;
535 
536 	/*
537 	 * Pointer to the mmu context currently used for
538 	 * gva_to_gpa translations.
539 	 */
540 	struct kvm_mmu *walk_mmu;
541 
542 	struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
543 	struct kvm_mmu_memory_cache mmu_page_cache;
544 	struct kvm_mmu_memory_cache mmu_page_header_cache;
545 
546 	/*
547 	 * QEMU userspace and the guest each have their own FPU state.
548 	 * In vcpu_run, we switch between the user and guest FPU contexts.
549 	 * While running a VCPU, the VCPU thread will have the guest FPU
550 	 * context.
551 	 *
552 	 * Note that while the PKRU state lives inside the fpu registers,
553 	 * it is switched out separately at VMENTER and VMEXIT time. The
554 	 * "guest_fpu" state here contains the guest FPU context, with the
555 	 * host PRKU bits.
556 	 */
557 	struct fpu user_fpu;
558 	struct fpu guest_fpu;
559 
560 	u64 xcr0;
561 	u64 guest_supported_xcr0;
562 	u32 guest_xstate_size;
563 
564 	struct kvm_pio_request pio;
565 	void *pio_data;
566 
567 	u8 event_exit_inst_len;
568 
569 	struct kvm_queued_exception {
570 		bool pending;
571 		bool injected;
572 		bool has_error_code;
573 		u8 nr;
574 		u32 error_code;
575 		u8 nested_apf;
576 	} exception;
577 
578 	struct kvm_queued_interrupt {
579 		bool injected;
580 		bool soft;
581 		u8 nr;
582 	} interrupt;
583 
584 	int halt_request; /* real mode on Intel only */
585 
586 	int cpuid_nent;
587 	struct kvm_cpuid_entry2 cpuid_entries[KVM_MAX_CPUID_ENTRIES];
588 
589 	int maxphyaddr;
590 
591 	/* emulate context */
592 
593 	struct x86_emulate_ctxt emulate_ctxt;
594 	bool emulate_regs_need_sync_to_vcpu;
595 	bool emulate_regs_need_sync_from_vcpu;
596 	int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
597 
598 	gpa_t time;
599 	struct pvclock_vcpu_time_info hv_clock;
600 	unsigned int hw_tsc_khz;
601 	struct gfn_to_hva_cache pv_time;
602 	bool pv_time_enabled;
603 	/* set guest stopped flag in pvclock flags field */
604 	bool pvclock_set_guest_stopped_request;
605 
606 	struct {
607 		u64 msr_val;
608 		u64 last_steal;
609 		struct gfn_to_hva_cache stime;
610 		struct kvm_steal_time steal;
611 	} st;
612 
613 	u64 tsc_offset;
614 	u64 last_guest_tsc;
615 	u64 last_host_tsc;
616 	u64 tsc_offset_adjustment;
617 	u64 this_tsc_nsec;
618 	u64 this_tsc_write;
619 	u64 this_tsc_generation;
620 	bool tsc_catchup;
621 	bool tsc_always_catchup;
622 	s8 virtual_tsc_shift;
623 	u32 virtual_tsc_mult;
624 	u32 virtual_tsc_khz;
625 	s64 ia32_tsc_adjust_msr;
626 	u64 tsc_scaling_ratio;
627 
628 	atomic_t nmi_queued;  /* unprocessed asynchronous NMIs */
629 	unsigned nmi_pending; /* NMI queued after currently running handler */
630 	bool nmi_injected;    /* Trying to inject an NMI this entry */
631 	bool smi_pending;    /* SMI queued after currently running handler */
632 
633 	struct kvm_mtrr mtrr_state;
634 	u64 pat;
635 
636 	unsigned switch_db_regs;
637 	unsigned long db[KVM_NR_DB_REGS];
638 	unsigned long dr6;
639 	unsigned long dr7;
640 	unsigned long eff_db[KVM_NR_DB_REGS];
641 	unsigned long guest_debug_dr7;
642 	u64 msr_platform_info;
643 	u64 msr_misc_features_enables;
644 
645 	u64 mcg_cap;
646 	u64 mcg_status;
647 	u64 mcg_ctl;
648 	u64 mcg_ext_ctl;
649 	u64 *mce_banks;
650 
651 	/* Cache MMIO info */
652 	u64 mmio_gva;
653 	unsigned access;
654 	gfn_t mmio_gfn;
655 	u64 mmio_gen;
656 
657 	struct kvm_pmu pmu;
658 
659 	/* used for guest single stepping over the given code position */
660 	unsigned long singlestep_rip;
661 
662 	struct kvm_vcpu_hv hyperv;
663 
664 	cpumask_var_t wbinvd_dirty_mask;
665 
666 	unsigned long last_retry_eip;
667 	unsigned long last_retry_addr;
668 
669 	struct {
670 		bool halted;
671 		gfn_t gfns[roundup_pow_of_two(ASYNC_PF_PER_VCPU)];
672 		struct gfn_to_hva_cache data;
673 		u64 msr_val;
674 		u32 id;
675 		bool send_user_only;
676 		u32 host_apf_reason;
677 		unsigned long nested_apf_token;
678 		bool delivery_as_pf_vmexit;
679 	} apf;
680 
681 	/* OSVW MSRs (AMD only) */
682 	struct {
683 		u64 length;
684 		u64 status;
685 	} osvw;
686 
687 	struct {
688 		u64 msr_val;
689 		struct gfn_to_hva_cache data;
690 	} pv_eoi;
691 
692 	/*
693 	 * Indicate whether the access faults on its page table in guest
694 	 * which is set when fix page fault and used to detect unhandeable
695 	 * instruction.
696 	 */
697 	bool write_fault_to_shadow_pgtable;
698 
699 	/* set at EPT violation at this point */
700 	unsigned long exit_qualification;
701 
702 	/* pv related host specific info */
703 	struct {
704 		bool pv_unhalted;
705 	} pv;
706 
707 	int pending_ioapic_eoi;
708 	int pending_external_vector;
709 
710 	/* GPA available */
711 	bool gpa_available;
712 	gpa_t gpa_val;
713 
714 	/* be preempted when it's in kernel-mode(cpl=0) */
715 	bool preempted_in_kernel;
716 };
717 
718 struct kvm_lpage_info {
719 	int disallow_lpage;
720 };
721 
722 struct kvm_arch_memory_slot {
723 	struct kvm_rmap_head *rmap[KVM_NR_PAGE_SIZES];
724 	struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
725 	unsigned short *gfn_track[KVM_PAGE_TRACK_MAX];
726 };
727 
728 /*
729  * We use as the mode the number of bits allocated in the LDR for the
730  * logical processor ID.  It happens that these are all powers of two.
731  * This makes it is very easy to detect cases where the APICs are
732  * configured for multiple modes; in that case, we cannot use the map and
733  * hence cannot use kvm_irq_delivery_to_apic_fast either.
734  */
735 #define KVM_APIC_MODE_XAPIC_CLUSTER          4
736 #define KVM_APIC_MODE_XAPIC_FLAT             8
737 #define KVM_APIC_MODE_X2APIC                16
738 
739 struct kvm_apic_map {
740 	struct rcu_head rcu;
741 	u8 mode;
742 	u32 max_apic_id;
743 	union {
744 		struct kvm_lapic *xapic_flat_map[8];
745 		struct kvm_lapic *xapic_cluster_map[16][4];
746 	};
747 	struct kvm_lapic *phys_map[];
748 };
749 
750 /* Hyper-V emulation context */
751 struct kvm_hv {
752 	struct mutex hv_lock;
753 	u64 hv_guest_os_id;
754 	u64 hv_hypercall;
755 	u64 hv_tsc_page;
756 
757 	/* Hyper-v based guest crash (NT kernel bugcheck) parameters */
758 	u64 hv_crash_param[HV_X64_MSR_CRASH_PARAMS];
759 	u64 hv_crash_ctl;
760 
761 	HV_REFERENCE_TSC_PAGE tsc_ref;
762 
763 	struct idr conn_to_evt;
764 
765 	u64 hv_reenlightenment_control;
766 	u64 hv_tsc_emulation_control;
767 	u64 hv_tsc_emulation_status;
768 };
769 
770 enum kvm_irqchip_mode {
771 	KVM_IRQCHIP_NONE,
772 	KVM_IRQCHIP_KERNEL,       /* created with KVM_CREATE_IRQCHIP */
773 	KVM_IRQCHIP_SPLIT,        /* created with KVM_CAP_SPLIT_IRQCHIP */
774 };
775 
776 struct kvm_arch {
777 	unsigned int n_used_mmu_pages;
778 	unsigned int n_requested_mmu_pages;
779 	unsigned int n_max_mmu_pages;
780 	unsigned int indirect_shadow_pages;
781 	unsigned long mmu_valid_gen;
782 	struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
783 	/*
784 	 * Hash table of struct kvm_mmu_page.
785 	 */
786 	struct list_head active_mmu_pages;
787 	struct list_head zapped_obsolete_pages;
788 	struct kvm_page_track_notifier_node mmu_sp_tracker;
789 	struct kvm_page_track_notifier_head track_notifier_head;
790 
791 	struct list_head assigned_dev_head;
792 	struct iommu_domain *iommu_domain;
793 	bool iommu_noncoherent;
794 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA
795 	atomic_t noncoherent_dma_count;
796 #define __KVM_HAVE_ARCH_ASSIGNED_DEVICE
797 	atomic_t assigned_device_count;
798 	struct kvm_pic *vpic;
799 	struct kvm_ioapic *vioapic;
800 	struct kvm_pit *vpit;
801 	atomic_t vapics_in_nmi_mode;
802 	struct mutex apic_map_lock;
803 	struct kvm_apic_map *apic_map;
804 
805 	bool apic_access_page_done;
806 
807 	gpa_t wall_clock;
808 
809 	bool mwait_in_guest;
810 	bool hlt_in_guest;
811 	bool pause_in_guest;
812 
813 	unsigned long irq_sources_bitmap;
814 	s64 kvmclock_offset;
815 	raw_spinlock_t tsc_write_lock;
816 	u64 last_tsc_nsec;
817 	u64 last_tsc_write;
818 	u32 last_tsc_khz;
819 	u64 cur_tsc_nsec;
820 	u64 cur_tsc_write;
821 	u64 cur_tsc_offset;
822 	u64 cur_tsc_generation;
823 	int nr_vcpus_matched_tsc;
824 
825 	spinlock_t pvclock_gtod_sync_lock;
826 	bool use_master_clock;
827 	u64 master_kernel_ns;
828 	u64 master_cycle_now;
829 	struct delayed_work kvmclock_update_work;
830 	struct delayed_work kvmclock_sync_work;
831 
832 	struct kvm_xen_hvm_config xen_hvm_config;
833 
834 	/* reads protected by irq_srcu, writes by irq_lock */
835 	struct hlist_head mask_notifier_list;
836 
837 	struct kvm_hv hyperv;
838 
839 	#ifdef CONFIG_KVM_MMU_AUDIT
840 	int audit_point;
841 	#endif
842 
843 	bool backwards_tsc_observed;
844 	bool boot_vcpu_runs_old_kvmclock;
845 	u32 bsp_vcpu_id;
846 
847 	u64 disabled_quirks;
848 
849 	enum kvm_irqchip_mode irqchip_mode;
850 	u8 nr_reserved_ioapic_pins;
851 
852 	bool disabled_lapic_found;
853 
854 	bool x2apic_format;
855 	bool x2apic_broadcast_quirk_disabled;
856 };
857 
858 struct kvm_vm_stat {
859 	ulong mmu_shadow_zapped;
860 	ulong mmu_pte_write;
861 	ulong mmu_pte_updated;
862 	ulong mmu_pde_zapped;
863 	ulong mmu_flooded;
864 	ulong mmu_recycled;
865 	ulong mmu_cache_miss;
866 	ulong mmu_unsync;
867 	ulong remote_tlb_flush;
868 	ulong lpages;
869 	ulong max_mmu_page_hash_collisions;
870 };
871 
872 struct kvm_vcpu_stat {
873 	u64 pf_fixed;
874 	u64 pf_guest;
875 	u64 tlb_flush;
876 	u64 invlpg;
877 
878 	u64 exits;
879 	u64 io_exits;
880 	u64 mmio_exits;
881 	u64 signal_exits;
882 	u64 irq_window_exits;
883 	u64 nmi_window_exits;
884 	u64 halt_exits;
885 	u64 halt_successful_poll;
886 	u64 halt_attempted_poll;
887 	u64 halt_poll_invalid;
888 	u64 halt_wakeup;
889 	u64 request_irq_exits;
890 	u64 irq_exits;
891 	u64 host_state_reload;
892 	u64 fpu_reload;
893 	u64 insn_emulation;
894 	u64 insn_emulation_fail;
895 	u64 hypercalls;
896 	u64 irq_injections;
897 	u64 nmi_injections;
898 	u64 req_event;
899 };
900 
901 struct x86_instruction_info;
902 
903 struct msr_data {
904 	bool host_initiated;
905 	u32 index;
906 	u64 data;
907 };
908 
909 struct kvm_lapic_irq {
910 	u32 vector;
911 	u16 delivery_mode;
912 	u16 dest_mode;
913 	bool level;
914 	u16 trig_mode;
915 	u32 shorthand;
916 	u32 dest_id;
917 	bool msi_redir_hint;
918 };
919 
920 struct kvm_x86_ops {
921 	int (*cpu_has_kvm_support)(void);          /* __init */
922 	int (*disabled_by_bios)(void);             /* __init */
923 	int (*hardware_enable)(void);
924 	void (*hardware_disable)(void);
925 	void (*check_processor_compatibility)(void *rtn);
926 	int (*hardware_setup)(void);               /* __init */
927 	void (*hardware_unsetup)(void);            /* __exit */
928 	bool (*cpu_has_accelerated_tpr)(void);
929 	bool (*has_emulated_msr)(int index);
930 	void (*cpuid_update)(struct kvm_vcpu *vcpu);
931 
932 	struct kvm *(*vm_alloc)(void);
933 	void (*vm_free)(struct kvm *);
934 	int (*vm_init)(struct kvm *kvm);
935 	void (*vm_destroy)(struct kvm *kvm);
936 
937 	/* Create, but do not attach this VCPU */
938 	struct kvm_vcpu *(*vcpu_create)(struct kvm *kvm, unsigned id);
939 	void (*vcpu_free)(struct kvm_vcpu *vcpu);
940 	void (*vcpu_reset)(struct kvm_vcpu *vcpu, bool init_event);
941 
942 	void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
943 	void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
944 	void (*vcpu_put)(struct kvm_vcpu *vcpu);
945 
946 	void (*update_bp_intercept)(struct kvm_vcpu *vcpu);
947 	int (*get_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
948 	int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
949 	u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
950 	void (*get_segment)(struct kvm_vcpu *vcpu,
951 			    struct kvm_segment *var, int seg);
952 	int (*get_cpl)(struct kvm_vcpu *vcpu);
953 	void (*set_segment)(struct kvm_vcpu *vcpu,
954 			    struct kvm_segment *var, int seg);
955 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
956 	void (*decache_cr0_guest_bits)(struct kvm_vcpu *vcpu);
957 	void (*decache_cr3)(struct kvm_vcpu *vcpu);
958 	void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu);
959 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
960 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
961 	int (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
962 	void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
963 	void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
964 	void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
965 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
966 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
967 	u64 (*get_dr6)(struct kvm_vcpu *vcpu);
968 	void (*set_dr6)(struct kvm_vcpu *vcpu, unsigned long value);
969 	void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu);
970 	void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
971 	void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
972 	unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
973 	void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
974 
975 	void (*tlb_flush)(struct kvm_vcpu *vcpu, bool invalidate_gpa);
976 
977 	void (*run)(struct kvm_vcpu *vcpu);
978 	int (*handle_exit)(struct kvm_vcpu *vcpu);
979 	void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
980 	void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
981 	u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu);
982 	void (*patch_hypercall)(struct kvm_vcpu *vcpu,
983 				unsigned char *hypercall_addr);
984 	void (*set_irq)(struct kvm_vcpu *vcpu);
985 	void (*set_nmi)(struct kvm_vcpu *vcpu);
986 	void (*queue_exception)(struct kvm_vcpu *vcpu);
987 	void (*cancel_injection)(struct kvm_vcpu *vcpu);
988 	int (*interrupt_allowed)(struct kvm_vcpu *vcpu);
989 	int (*nmi_allowed)(struct kvm_vcpu *vcpu);
990 	bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
991 	void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
992 	void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
993 	void (*enable_irq_window)(struct kvm_vcpu *vcpu);
994 	void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
995 	bool (*get_enable_apicv)(struct kvm_vcpu *vcpu);
996 	void (*refresh_apicv_exec_ctrl)(struct kvm_vcpu *vcpu);
997 	void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
998 	void (*hwapic_isr_update)(struct kvm_vcpu *vcpu, int isr);
999 	void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
1000 	void (*set_virtual_apic_mode)(struct kvm_vcpu *vcpu);
1001 	void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu, hpa_t hpa);
1002 	void (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
1003 	int (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
1004 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
1005 	int (*set_identity_map_addr)(struct kvm *kvm, u64 ident_addr);
1006 	int (*get_tdp_level)(struct kvm_vcpu *vcpu);
1007 	u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
1008 	int (*get_lpage_level)(void);
1009 	bool (*rdtscp_supported)(void);
1010 	bool (*invpcid_supported)(void);
1011 
1012 	void (*set_tdp_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
1013 
1014 	void (*set_supported_cpuid)(u32 func, struct kvm_cpuid_entry2 *entry);
1015 
1016 	bool (*has_wbinvd_exit)(void);
1017 
1018 	u64 (*read_l1_tsc_offset)(struct kvm_vcpu *vcpu);
1019 	void (*write_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
1020 
1021 	void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2);
1022 
1023 	int (*check_intercept)(struct kvm_vcpu *vcpu,
1024 			       struct x86_instruction_info *info,
1025 			       enum x86_intercept_stage stage);
1026 	void (*handle_external_intr)(struct kvm_vcpu *vcpu);
1027 	bool (*mpx_supported)(void);
1028 	bool (*xsaves_supported)(void);
1029 	bool (*umip_emulated)(void);
1030 
1031 	int (*check_nested_events)(struct kvm_vcpu *vcpu, bool external_intr);
1032 
1033 	void (*sched_in)(struct kvm_vcpu *kvm, int cpu);
1034 
1035 	/*
1036 	 * Arch-specific dirty logging hooks. These hooks are only supposed to
1037 	 * be valid if the specific arch has hardware-accelerated dirty logging
1038 	 * mechanism. Currently only for PML on VMX.
1039 	 *
1040 	 *  - slot_enable_log_dirty:
1041 	 *	called when enabling log dirty mode for the slot.
1042 	 *  - slot_disable_log_dirty:
1043 	 *	called when disabling log dirty mode for the slot.
1044 	 *	also called when slot is created with log dirty disabled.
1045 	 *  - flush_log_dirty:
1046 	 *	called before reporting dirty_bitmap to userspace.
1047 	 *  - enable_log_dirty_pt_masked:
1048 	 *	called when reenabling log dirty for the GFNs in the mask after
1049 	 *	corresponding bits are cleared in slot->dirty_bitmap.
1050 	 */
1051 	void (*slot_enable_log_dirty)(struct kvm *kvm,
1052 				      struct kvm_memory_slot *slot);
1053 	void (*slot_disable_log_dirty)(struct kvm *kvm,
1054 				       struct kvm_memory_slot *slot);
1055 	void (*flush_log_dirty)(struct kvm *kvm);
1056 	void (*enable_log_dirty_pt_masked)(struct kvm *kvm,
1057 					   struct kvm_memory_slot *slot,
1058 					   gfn_t offset, unsigned long mask);
1059 	int (*write_log_dirty)(struct kvm_vcpu *vcpu);
1060 
1061 	/* pmu operations of sub-arch */
1062 	const struct kvm_pmu_ops *pmu_ops;
1063 
1064 	/*
1065 	 * Architecture specific hooks for vCPU blocking due to
1066 	 * HLT instruction.
1067 	 * Returns for .pre_block():
1068 	 *    - 0 means continue to block the vCPU.
1069 	 *    - 1 means we cannot block the vCPU since some event
1070 	 *        happens during this period, such as, 'ON' bit in
1071 	 *        posted-interrupts descriptor is set.
1072 	 */
1073 	int (*pre_block)(struct kvm_vcpu *vcpu);
1074 	void (*post_block)(struct kvm_vcpu *vcpu);
1075 
1076 	void (*vcpu_blocking)(struct kvm_vcpu *vcpu);
1077 	void (*vcpu_unblocking)(struct kvm_vcpu *vcpu);
1078 
1079 	int (*update_pi_irte)(struct kvm *kvm, unsigned int host_irq,
1080 			      uint32_t guest_irq, bool set);
1081 	void (*apicv_post_state_restore)(struct kvm_vcpu *vcpu);
1082 
1083 	int (*set_hv_timer)(struct kvm_vcpu *vcpu, u64 guest_deadline_tsc);
1084 	void (*cancel_hv_timer)(struct kvm_vcpu *vcpu);
1085 
1086 	void (*setup_mce)(struct kvm_vcpu *vcpu);
1087 
1088 	int (*smi_allowed)(struct kvm_vcpu *vcpu);
1089 	int (*pre_enter_smm)(struct kvm_vcpu *vcpu, char *smstate);
1090 	int (*pre_leave_smm)(struct kvm_vcpu *vcpu, u64 smbase);
1091 	int (*enable_smi_window)(struct kvm_vcpu *vcpu);
1092 
1093 	int (*mem_enc_op)(struct kvm *kvm, void __user *argp);
1094 	int (*mem_enc_reg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1095 	int (*mem_enc_unreg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1096 
1097 	int (*get_msr_feature)(struct kvm_msr_entry *entry);
1098 };
1099 
1100 struct kvm_arch_async_pf {
1101 	u32 token;
1102 	gfn_t gfn;
1103 	unsigned long cr3;
1104 	bool direct_map;
1105 };
1106 
1107 extern struct kvm_x86_ops *kvm_x86_ops;
1108 
1109 #define __KVM_HAVE_ARCH_VM_ALLOC
1110 static inline struct kvm *kvm_arch_alloc_vm(void)
1111 {
1112 	return kvm_x86_ops->vm_alloc();
1113 }
1114 
1115 static inline void kvm_arch_free_vm(struct kvm *kvm)
1116 {
1117 	return kvm_x86_ops->vm_free(kvm);
1118 }
1119 
1120 int kvm_mmu_module_init(void);
1121 void kvm_mmu_module_exit(void);
1122 
1123 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
1124 int kvm_mmu_create(struct kvm_vcpu *vcpu);
1125 void kvm_mmu_setup(struct kvm_vcpu *vcpu);
1126 void kvm_mmu_init_vm(struct kvm *kvm);
1127 void kvm_mmu_uninit_vm(struct kvm *kvm);
1128 void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
1129 		u64 dirty_mask, u64 nx_mask, u64 x_mask, u64 p_mask,
1130 		u64 acc_track_mask, u64 me_mask);
1131 
1132 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
1133 void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
1134 				      struct kvm_memory_slot *memslot);
1135 void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
1136 				   const struct kvm_memory_slot *memslot);
1137 void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
1138 				   struct kvm_memory_slot *memslot);
1139 void kvm_mmu_slot_largepage_remove_write_access(struct kvm *kvm,
1140 					struct kvm_memory_slot *memslot);
1141 void kvm_mmu_slot_set_dirty(struct kvm *kvm,
1142 			    struct kvm_memory_slot *memslot);
1143 void kvm_mmu_clear_dirty_pt_masked(struct kvm *kvm,
1144 				   struct kvm_memory_slot *slot,
1145 				   gfn_t gfn_offset, unsigned long mask);
1146 void kvm_mmu_zap_all(struct kvm *kvm);
1147 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, struct kvm_memslots *slots);
1148 unsigned int kvm_mmu_calculate_mmu_pages(struct kvm *kvm);
1149 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int kvm_nr_mmu_pages);
1150 
1151 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
1152 bool pdptrs_changed(struct kvm_vcpu *vcpu);
1153 
1154 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
1155 			  const void *val, int bytes);
1156 
1157 struct kvm_irq_mask_notifier {
1158 	void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
1159 	int irq;
1160 	struct hlist_node link;
1161 };
1162 
1163 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
1164 				    struct kvm_irq_mask_notifier *kimn);
1165 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
1166 				      struct kvm_irq_mask_notifier *kimn);
1167 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
1168 			     bool mask);
1169 
1170 extern bool tdp_enabled;
1171 
1172 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
1173 
1174 /* control of guest tsc rate supported? */
1175 extern bool kvm_has_tsc_control;
1176 /* maximum supported tsc_khz for guests */
1177 extern u32  kvm_max_guest_tsc_khz;
1178 /* number of bits of the fractional part of the TSC scaling ratio */
1179 extern u8   kvm_tsc_scaling_ratio_frac_bits;
1180 /* maximum allowed value of TSC scaling ratio */
1181 extern u64  kvm_max_tsc_scaling_ratio;
1182 /* 1ull << kvm_tsc_scaling_ratio_frac_bits */
1183 extern u64  kvm_default_tsc_scaling_ratio;
1184 
1185 extern u64 kvm_mce_cap_supported;
1186 
1187 enum emulation_result {
1188 	EMULATE_DONE,         /* no further processing */
1189 	EMULATE_USER_EXIT,    /* kvm_run ready for userspace exit */
1190 	EMULATE_FAIL,         /* can't emulate this instruction */
1191 };
1192 
1193 #define EMULTYPE_NO_DECODE	    (1 << 0)
1194 #define EMULTYPE_TRAP_UD	    (1 << 1)
1195 #define EMULTYPE_SKIP		    (1 << 2)
1196 #define EMULTYPE_RETRY		    (1 << 3)
1197 #define EMULTYPE_NO_REEXECUTE	    (1 << 4)
1198 #define EMULTYPE_NO_UD_ON_FAIL	    (1 << 5)
1199 #define EMULTYPE_VMWARE		    (1 << 6)
1200 int x86_emulate_instruction(struct kvm_vcpu *vcpu, unsigned long cr2,
1201 			    int emulation_type, void *insn, int insn_len);
1202 
1203 static inline int emulate_instruction(struct kvm_vcpu *vcpu,
1204 			int emulation_type)
1205 {
1206 	return x86_emulate_instruction(vcpu, 0,
1207 			emulation_type | EMULTYPE_NO_REEXECUTE, NULL, 0);
1208 }
1209 
1210 void kvm_enable_efer_bits(u64);
1211 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
1212 int kvm_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr);
1213 int kvm_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr);
1214 
1215 struct x86_emulate_ctxt;
1216 
1217 int kvm_fast_pio(struct kvm_vcpu *vcpu, int size, unsigned short port, int in);
1218 int kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
1219 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
1220 int kvm_vcpu_halt(struct kvm_vcpu *vcpu);
1221 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
1222 
1223 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
1224 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
1225 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector);
1226 
1227 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
1228 		    int reason, bool has_error_code, u32 error_code);
1229 
1230 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
1231 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
1232 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
1233 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
1234 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
1235 int kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
1236 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
1237 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
1238 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
1239 int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr);
1240 
1241 int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1242 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1243 
1244 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
1245 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
1246 bool kvm_rdpmc(struct kvm_vcpu *vcpu);
1247 
1248 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1249 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1250 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1251 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1252 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
1253 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1254 			    gfn_t gfn, void *data, int offset, int len,
1255 			    u32 access);
1256 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
1257 bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr);
1258 
1259 static inline int __kvm_irq_line_state(unsigned long *irq_state,
1260 				       int irq_source_id, int level)
1261 {
1262 	/* Logical OR for level trig interrupt */
1263 	if (level)
1264 		__set_bit(irq_source_id, irq_state);
1265 	else
1266 		__clear_bit(irq_source_id, irq_state);
1267 
1268 	return !!(*irq_state);
1269 }
1270 
1271 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
1272 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
1273 
1274 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
1275 
1276 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
1277 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
1278 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
1279 int kvm_mmu_load(struct kvm_vcpu *vcpu);
1280 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
1281 void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu);
1282 void kvm_mmu_free_roots(struct kvm_vcpu *vcpu);
1283 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1284 			   struct x86_exception *exception);
1285 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
1286 			      struct x86_exception *exception);
1287 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
1288 			       struct x86_exception *exception);
1289 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
1290 			       struct x86_exception *exception);
1291 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
1292 				struct x86_exception *exception);
1293 
1294 void kvm_vcpu_deactivate_apicv(struct kvm_vcpu *vcpu);
1295 
1296 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
1297 
1298 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva, u64 error_code,
1299 		       void *insn, int insn_len);
1300 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
1301 void kvm_mmu_new_cr3(struct kvm_vcpu *vcpu);
1302 
1303 void kvm_enable_tdp(void);
1304 void kvm_disable_tdp(void);
1305 
1306 static inline gpa_t translate_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1307 				  struct x86_exception *exception)
1308 {
1309 	return gpa;
1310 }
1311 
1312 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
1313 {
1314 	struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
1315 
1316 	return (struct kvm_mmu_page *)page_private(page);
1317 }
1318 
1319 static inline u16 kvm_read_ldt(void)
1320 {
1321 	u16 ldt;
1322 	asm("sldt %0" : "=g"(ldt));
1323 	return ldt;
1324 }
1325 
1326 static inline void kvm_load_ldt(u16 sel)
1327 {
1328 	asm("lldt %0" : : "rm"(sel));
1329 }
1330 
1331 #ifdef CONFIG_X86_64
1332 static inline unsigned long read_msr(unsigned long msr)
1333 {
1334 	u64 value;
1335 
1336 	rdmsrl(msr, value);
1337 	return value;
1338 }
1339 #endif
1340 
1341 static inline u32 get_rdx_init_val(void)
1342 {
1343 	return 0x600; /* P6 family */
1344 }
1345 
1346 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
1347 {
1348 	kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
1349 }
1350 
1351 #define TSS_IOPB_BASE_OFFSET 0x66
1352 #define TSS_BASE_SIZE 0x68
1353 #define TSS_IOPB_SIZE (65536 / 8)
1354 #define TSS_REDIRECTION_SIZE (256 / 8)
1355 #define RMODE_TSS_SIZE							\
1356 	(TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
1357 
1358 enum {
1359 	TASK_SWITCH_CALL = 0,
1360 	TASK_SWITCH_IRET = 1,
1361 	TASK_SWITCH_JMP = 2,
1362 	TASK_SWITCH_GATE = 3,
1363 };
1364 
1365 #define HF_GIF_MASK		(1 << 0)
1366 #define HF_HIF_MASK		(1 << 1)
1367 #define HF_VINTR_MASK		(1 << 2)
1368 #define HF_NMI_MASK		(1 << 3)
1369 #define HF_IRET_MASK		(1 << 4)
1370 #define HF_GUEST_MASK		(1 << 5) /* VCPU is in guest-mode */
1371 #define HF_SMM_MASK		(1 << 6)
1372 #define HF_SMM_INSIDE_NMI_MASK	(1 << 7)
1373 
1374 #define __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
1375 #define KVM_ADDRESS_SPACE_NUM 2
1376 
1377 #define kvm_arch_vcpu_memslots_id(vcpu) ((vcpu)->arch.hflags & HF_SMM_MASK ? 1 : 0)
1378 #define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, (role).smm)
1379 
1380 /*
1381  * Hardware virtualization extension instructions may fault if a
1382  * reboot turns off virtualization while processes are running.
1383  * Trap the fault and ignore the instruction if that happens.
1384  */
1385 asmlinkage void kvm_spurious_fault(void);
1386 
1387 #define ____kvm_handle_fault_on_reboot(insn, cleanup_insn)	\
1388 	"666: " insn "\n\t" \
1389 	"668: \n\t"                           \
1390 	".pushsection .fixup, \"ax\" \n" \
1391 	"667: \n\t" \
1392 	cleanup_insn "\n\t"		      \
1393 	"cmpb $0, kvm_rebooting \n\t"	      \
1394 	"jne 668b \n\t"      		      \
1395 	__ASM_SIZE(push) " $666b \n\t"	      \
1396 	"call kvm_spurious_fault \n\t"	      \
1397 	".popsection \n\t" \
1398 	_ASM_EXTABLE(666b, 667b)
1399 
1400 #define __kvm_handle_fault_on_reboot(insn)		\
1401 	____kvm_handle_fault_on_reboot(insn, "")
1402 
1403 #define KVM_ARCH_WANT_MMU_NOTIFIER
1404 int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
1405 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end);
1406 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
1407 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
1408 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
1409 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1410 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1411 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1412 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1413 void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event);
1414 void kvm_vcpu_reload_apic_access_page(struct kvm_vcpu *vcpu);
1415 
1416 void kvm_define_shared_msr(unsigned index, u32 msr);
1417 int kvm_set_shared_msr(unsigned index, u64 val, u64 mask);
1418 
1419 u64 kvm_scale_tsc(struct kvm_vcpu *vcpu, u64 tsc);
1420 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc);
1421 
1422 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu);
1423 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1424 
1425 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
1426 void kvm_make_scan_ioapic_request(struct kvm *kvm);
1427 
1428 void kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1429 				     struct kvm_async_pf *work);
1430 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1431 				 struct kvm_async_pf *work);
1432 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1433 			       struct kvm_async_pf *work);
1434 bool kvm_arch_can_inject_async_page_present(struct kvm_vcpu *vcpu);
1435 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1436 
1437 int kvm_skip_emulated_instruction(struct kvm_vcpu *vcpu);
1438 int kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1439 
1440 int kvm_is_in_guest(void);
1441 
1442 int __x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa, u32 size);
1443 int x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa, u32 size);
1444 bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu);
1445 bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu);
1446 
1447 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
1448 			     struct kvm_vcpu **dest_vcpu);
1449 
1450 void kvm_set_msi_irq(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e,
1451 		     struct kvm_lapic_irq *irq);
1452 
1453 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu)
1454 {
1455 	if (kvm_x86_ops->vcpu_blocking)
1456 		kvm_x86_ops->vcpu_blocking(vcpu);
1457 }
1458 
1459 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu)
1460 {
1461 	if (kvm_x86_ops->vcpu_unblocking)
1462 		kvm_x86_ops->vcpu_unblocking(vcpu);
1463 }
1464 
1465 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
1466 
1467 static inline int kvm_cpu_get_apicid(int mps_cpu)
1468 {
1469 #ifdef CONFIG_X86_LOCAL_APIC
1470 	return default_cpu_present_to_apicid(mps_cpu);
1471 #else
1472 	WARN_ON_ONCE(1);
1473 	return BAD_APICID;
1474 #endif
1475 }
1476 
1477 #define put_smstate(type, buf, offset, val)                      \
1478 	*(type *)((buf) + (offset) - 0x7e00) = val
1479 
1480 #endif /* _ASM_X86_KVM_HOST_H */
1481