xref: /openbmc/linux/arch/x86/include/asm/kvm_host.h (revision a8da474e)
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 
29 #include <asm/pvclock-abi.h>
30 #include <asm/desc.h>
31 #include <asm/mtrr.h>
32 #include <asm/msr-index.h>
33 #include <asm/asm.h>
34 
35 #define KVM_MAX_VCPUS 255
36 #define KVM_SOFT_MAX_VCPUS 160
37 #define KVM_USER_MEM_SLOTS 509
38 /* memory slots that are not exposed to userspace */
39 #define KVM_PRIVATE_MEM_SLOTS 3
40 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
41 
42 #define KVM_PIO_PAGE_OFFSET 1
43 #define KVM_COALESCED_MMIO_PAGE_OFFSET 2
44 #define KVM_HALT_POLL_NS_DEFAULT 500000
45 
46 #define KVM_IRQCHIP_NUM_PINS  KVM_IOAPIC_NUM_PINS
47 
48 #define CR0_RESERVED_BITS                                               \
49 	(~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \
50 			  | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \
51 			  | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG))
52 
53 #define CR3_L_MODE_RESERVED_BITS 0xFFFFFF0000000000ULL
54 #define CR3_PCID_INVD		 BIT_64(63)
55 #define CR4_RESERVED_BITS                                               \
56 	(~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\
57 			  | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE     \
58 			  | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR | X86_CR4_PCIDE \
59 			  | X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_FSGSBASE \
60 			  | X86_CR4_OSXMMEXCPT | X86_CR4_VMXE | X86_CR4_SMAP))
61 
62 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR)
63 
64 
65 
66 #define INVALID_PAGE (~(hpa_t)0)
67 #define VALID_PAGE(x) ((x) != INVALID_PAGE)
68 
69 #define UNMAPPED_GVA (~(gpa_t)0)
70 
71 /* KVM Hugepage definitions for x86 */
72 #define KVM_NR_PAGE_SIZES	3
73 #define KVM_HPAGE_GFN_SHIFT(x)	(((x) - 1) * 9)
74 #define KVM_HPAGE_SHIFT(x)	(PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
75 #define KVM_HPAGE_SIZE(x)	(1UL << KVM_HPAGE_SHIFT(x))
76 #define KVM_HPAGE_MASK(x)	(~(KVM_HPAGE_SIZE(x) - 1))
77 #define KVM_PAGES_PER_HPAGE(x)	(KVM_HPAGE_SIZE(x) / PAGE_SIZE)
78 
79 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
80 {
81 	/* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */
82 	return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
83 		(base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
84 }
85 
86 #define KVM_PERMILLE_MMU_PAGES 20
87 #define KVM_MIN_ALLOC_MMU_PAGES 64
88 #define KVM_MMU_HASH_SHIFT 10
89 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
90 #define KVM_MIN_FREE_MMU_PAGES 5
91 #define KVM_REFILL_PAGES 25
92 #define KVM_MAX_CPUID_ENTRIES 80
93 #define KVM_NR_FIXED_MTRR_REGION 88
94 #define KVM_NR_VAR_MTRR 8
95 
96 #define ASYNC_PF_PER_VCPU 64
97 
98 enum kvm_reg {
99 	VCPU_REGS_RAX = 0,
100 	VCPU_REGS_RCX = 1,
101 	VCPU_REGS_RDX = 2,
102 	VCPU_REGS_RBX = 3,
103 	VCPU_REGS_RSP = 4,
104 	VCPU_REGS_RBP = 5,
105 	VCPU_REGS_RSI = 6,
106 	VCPU_REGS_RDI = 7,
107 #ifdef CONFIG_X86_64
108 	VCPU_REGS_R8 = 8,
109 	VCPU_REGS_R9 = 9,
110 	VCPU_REGS_R10 = 10,
111 	VCPU_REGS_R11 = 11,
112 	VCPU_REGS_R12 = 12,
113 	VCPU_REGS_R13 = 13,
114 	VCPU_REGS_R14 = 14,
115 	VCPU_REGS_R15 = 15,
116 #endif
117 	VCPU_REGS_RIP,
118 	NR_VCPU_REGS
119 };
120 
121 enum kvm_reg_ex {
122 	VCPU_EXREG_PDPTR = NR_VCPU_REGS,
123 	VCPU_EXREG_CR3,
124 	VCPU_EXREG_RFLAGS,
125 	VCPU_EXREG_SEGMENTS,
126 };
127 
128 enum {
129 	VCPU_SREG_ES,
130 	VCPU_SREG_CS,
131 	VCPU_SREG_SS,
132 	VCPU_SREG_DS,
133 	VCPU_SREG_FS,
134 	VCPU_SREG_GS,
135 	VCPU_SREG_TR,
136 	VCPU_SREG_LDTR,
137 };
138 
139 #include <asm/kvm_emulate.h>
140 
141 #define KVM_NR_MEM_OBJS 40
142 
143 #define KVM_NR_DB_REGS	4
144 
145 #define DR6_BD		(1 << 13)
146 #define DR6_BS		(1 << 14)
147 #define DR6_RTM		(1 << 16)
148 #define DR6_FIXED_1	0xfffe0ff0
149 #define DR6_INIT	0xffff0ff0
150 #define DR6_VOLATILE	0x0001e00f
151 
152 #define DR7_BP_EN_MASK	0x000000ff
153 #define DR7_GE		(1 << 9)
154 #define DR7_GD		(1 << 13)
155 #define DR7_FIXED_1	0x00000400
156 #define DR7_VOLATILE	0xffff2bff
157 
158 #define PFERR_PRESENT_BIT 0
159 #define PFERR_WRITE_BIT 1
160 #define PFERR_USER_BIT 2
161 #define PFERR_RSVD_BIT 3
162 #define PFERR_FETCH_BIT 4
163 
164 #define PFERR_PRESENT_MASK (1U << PFERR_PRESENT_BIT)
165 #define PFERR_WRITE_MASK (1U << PFERR_WRITE_BIT)
166 #define PFERR_USER_MASK (1U << PFERR_USER_BIT)
167 #define PFERR_RSVD_MASK (1U << PFERR_RSVD_BIT)
168 #define PFERR_FETCH_MASK (1U << PFERR_FETCH_BIT)
169 
170 /* apic attention bits */
171 #define KVM_APIC_CHECK_VAPIC	0
172 /*
173  * The following bit is set with PV-EOI, unset on EOI.
174  * We detect PV-EOI changes by guest by comparing
175  * this bit with PV-EOI in guest memory.
176  * See the implementation in apic_update_pv_eoi.
177  */
178 #define KVM_APIC_PV_EOI_PENDING	1
179 
180 struct kvm_kernel_irq_routing_entry;
181 
182 /*
183  * We don't want allocation failures within the mmu code, so we preallocate
184  * enough memory for a single page fault in a cache.
185  */
186 struct kvm_mmu_memory_cache {
187 	int nobjs;
188 	void *objects[KVM_NR_MEM_OBJS];
189 };
190 
191 union kvm_mmu_page_role {
192 	unsigned word;
193 	struct {
194 		unsigned level:4;
195 		unsigned cr4_pae:1;
196 		unsigned quadrant:2;
197 		unsigned direct:1;
198 		unsigned access:3;
199 		unsigned invalid:1;
200 		unsigned nxe:1;
201 		unsigned cr0_wp:1;
202 		unsigned smep_andnot_wp:1;
203 		unsigned smap_andnot_wp:1;
204 		unsigned :8;
205 
206 		/*
207 		 * This is left at the top of the word so that
208 		 * kvm_memslots_for_spte_role can extract it with a
209 		 * simple shift.  While there is room, give it a whole
210 		 * byte so it is also faster to load it from memory.
211 		 */
212 		unsigned smm:8;
213 	};
214 };
215 
216 struct kvm_mmu_page {
217 	struct list_head link;
218 	struct hlist_node hash_link;
219 
220 	/*
221 	 * The following two entries are used to key the shadow page in the
222 	 * hash table.
223 	 */
224 	gfn_t gfn;
225 	union kvm_mmu_page_role role;
226 
227 	u64 *spt;
228 	/* hold the gfn of each spte inside spt */
229 	gfn_t *gfns;
230 	bool unsync;
231 	int root_count;          /* Currently serving as active root */
232 	unsigned int unsync_children;
233 	unsigned long parent_ptes;	/* Reverse mapping for parent_pte */
234 
235 	/* The page is obsolete if mmu_valid_gen != kvm->arch.mmu_valid_gen.  */
236 	unsigned long mmu_valid_gen;
237 
238 	DECLARE_BITMAP(unsync_child_bitmap, 512);
239 
240 #ifdef CONFIG_X86_32
241 	/*
242 	 * Used out of the mmu-lock to avoid reading spte values while an
243 	 * update is in progress; see the comments in __get_spte_lockless().
244 	 */
245 	int clear_spte_count;
246 #endif
247 
248 	/* Number of writes since the last time traversal visited this page.  */
249 	int write_flooding_count;
250 };
251 
252 struct kvm_pio_request {
253 	unsigned long count;
254 	int in;
255 	int port;
256 	int size;
257 };
258 
259 struct rsvd_bits_validate {
260 	u64 rsvd_bits_mask[2][4];
261 	u64 bad_mt_xwr;
262 };
263 
264 /*
265  * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
266  * 32-bit).  The kvm_mmu structure abstracts the details of the current mmu
267  * mode.
268  */
269 struct kvm_mmu {
270 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long root);
271 	unsigned long (*get_cr3)(struct kvm_vcpu *vcpu);
272 	u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
273 	int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err,
274 			  bool prefault);
275 	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
276 				  struct x86_exception *fault);
277 	gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva, u32 access,
278 			    struct x86_exception *exception);
279 	gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
280 			       struct x86_exception *exception);
281 	int (*sync_page)(struct kvm_vcpu *vcpu,
282 			 struct kvm_mmu_page *sp);
283 	void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva);
284 	void (*update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
285 			   u64 *spte, const void *pte);
286 	hpa_t root_hpa;
287 	int root_level;
288 	int shadow_root_level;
289 	union kvm_mmu_page_role base_role;
290 	bool direct_map;
291 
292 	/*
293 	 * Bitmap; bit set = permission fault
294 	 * Byte index: page fault error code [4:1]
295 	 * Bit index: pte permissions in ACC_* format
296 	 */
297 	u8 permissions[16];
298 
299 	u64 *pae_root;
300 	u64 *lm_root;
301 
302 	/*
303 	 * check zero bits on shadow page table entries, these
304 	 * bits include not only hardware reserved bits but also
305 	 * the bits spte never used.
306 	 */
307 	struct rsvd_bits_validate shadow_zero_check;
308 
309 	struct rsvd_bits_validate guest_rsvd_check;
310 
311 	/*
312 	 * Bitmap: bit set = last pte in walk
313 	 * index[0:1]: level (zero-based)
314 	 * index[2]: pte.ps
315 	 */
316 	u8 last_pte_bitmap;
317 
318 	bool nx;
319 
320 	u64 pdptrs[4]; /* pae */
321 };
322 
323 enum pmc_type {
324 	KVM_PMC_GP = 0,
325 	KVM_PMC_FIXED,
326 };
327 
328 struct kvm_pmc {
329 	enum pmc_type type;
330 	u8 idx;
331 	u64 counter;
332 	u64 eventsel;
333 	struct perf_event *perf_event;
334 	struct kvm_vcpu *vcpu;
335 };
336 
337 struct kvm_pmu {
338 	unsigned nr_arch_gp_counters;
339 	unsigned nr_arch_fixed_counters;
340 	unsigned available_event_types;
341 	u64 fixed_ctr_ctrl;
342 	u64 global_ctrl;
343 	u64 global_status;
344 	u64 global_ovf_ctrl;
345 	u64 counter_bitmask[2];
346 	u64 global_ctrl_mask;
347 	u64 reserved_bits;
348 	u8 version;
349 	struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
350 	struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED];
351 	struct irq_work irq_work;
352 	u64 reprogram_pmi;
353 };
354 
355 struct kvm_pmu_ops;
356 
357 enum {
358 	KVM_DEBUGREG_BP_ENABLED = 1,
359 	KVM_DEBUGREG_WONT_EXIT = 2,
360 	KVM_DEBUGREG_RELOAD = 4,
361 };
362 
363 struct kvm_mtrr_range {
364 	u64 base;
365 	u64 mask;
366 	struct list_head node;
367 };
368 
369 struct kvm_mtrr {
370 	struct kvm_mtrr_range var_ranges[KVM_NR_VAR_MTRR];
371 	mtrr_type fixed_ranges[KVM_NR_FIXED_MTRR_REGION];
372 	u64 deftype;
373 
374 	struct list_head head;
375 };
376 
377 /* Hyper-V per vcpu emulation context */
378 struct kvm_vcpu_hv {
379 	u64 hv_vapic;
380 	s64 runtime_offset;
381 };
382 
383 struct kvm_vcpu_arch {
384 	/*
385 	 * rip and regs accesses must go through
386 	 * kvm_{register,rip}_{read,write} functions.
387 	 */
388 	unsigned long regs[NR_VCPU_REGS];
389 	u32 regs_avail;
390 	u32 regs_dirty;
391 
392 	unsigned long cr0;
393 	unsigned long cr0_guest_owned_bits;
394 	unsigned long cr2;
395 	unsigned long cr3;
396 	unsigned long cr4;
397 	unsigned long cr4_guest_owned_bits;
398 	unsigned long cr8;
399 	u32 hflags;
400 	u64 efer;
401 	u64 apic_base;
402 	struct kvm_lapic *apic;    /* kernel irqchip context */
403 	u64 eoi_exit_bitmap[4];
404 	unsigned long apic_attention;
405 	int32_t apic_arb_prio;
406 	int mp_state;
407 	u64 ia32_misc_enable_msr;
408 	u64 smbase;
409 	bool tpr_access_reporting;
410 	u64 ia32_xss;
411 
412 	/*
413 	 * Paging state of the vcpu
414 	 *
415 	 * If the vcpu runs in guest mode with two level paging this still saves
416 	 * the paging mode of the l1 guest. This context is always used to
417 	 * handle faults.
418 	 */
419 	struct kvm_mmu mmu;
420 
421 	/*
422 	 * Paging state of an L2 guest (used for nested npt)
423 	 *
424 	 * This context will save all necessary information to walk page tables
425 	 * of the an L2 guest. This context is only initialized for page table
426 	 * walking and not for faulting since we never handle l2 page faults on
427 	 * the host.
428 	 */
429 	struct kvm_mmu nested_mmu;
430 
431 	/*
432 	 * Pointer to the mmu context currently used for
433 	 * gva_to_gpa translations.
434 	 */
435 	struct kvm_mmu *walk_mmu;
436 
437 	struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
438 	struct kvm_mmu_memory_cache mmu_page_cache;
439 	struct kvm_mmu_memory_cache mmu_page_header_cache;
440 
441 	struct fpu guest_fpu;
442 	bool eager_fpu;
443 	u64 xcr0;
444 	u64 guest_supported_xcr0;
445 	u32 guest_xstate_size;
446 
447 	struct kvm_pio_request pio;
448 	void *pio_data;
449 
450 	u8 event_exit_inst_len;
451 
452 	struct kvm_queued_exception {
453 		bool pending;
454 		bool has_error_code;
455 		bool reinject;
456 		u8 nr;
457 		u32 error_code;
458 	} exception;
459 
460 	struct kvm_queued_interrupt {
461 		bool pending;
462 		bool soft;
463 		u8 nr;
464 	} interrupt;
465 
466 	int halt_request; /* real mode on Intel only */
467 
468 	int cpuid_nent;
469 	struct kvm_cpuid_entry2 cpuid_entries[KVM_MAX_CPUID_ENTRIES];
470 
471 	int maxphyaddr;
472 
473 	/* emulate context */
474 
475 	struct x86_emulate_ctxt emulate_ctxt;
476 	bool emulate_regs_need_sync_to_vcpu;
477 	bool emulate_regs_need_sync_from_vcpu;
478 	int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
479 
480 	gpa_t time;
481 	struct pvclock_vcpu_time_info hv_clock;
482 	unsigned int hw_tsc_khz;
483 	struct gfn_to_hva_cache pv_time;
484 	bool pv_time_enabled;
485 	/* set guest stopped flag in pvclock flags field */
486 	bool pvclock_set_guest_stopped_request;
487 
488 	struct {
489 		u64 msr_val;
490 		u64 last_steal;
491 		u64 accum_steal;
492 		struct gfn_to_hva_cache stime;
493 		struct kvm_steal_time steal;
494 	} st;
495 
496 	u64 last_guest_tsc;
497 	u64 last_host_tsc;
498 	u64 tsc_offset_adjustment;
499 	u64 this_tsc_nsec;
500 	u64 this_tsc_write;
501 	u64 this_tsc_generation;
502 	bool tsc_catchup;
503 	bool tsc_always_catchup;
504 	s8 virtual_tsc_shift;
505 	u32 virtual_tsc_mult;
506 	u32 virtual_tsc_khz;
507 	s64 ia32_tsc_adjust_msr;
508 	u64 tsc_scaling_ratio;
509 
510 	atomic_t nmi_queued;  /* unprocessed asynchronous NMIs */
511 	unsigned nmi_pending; /* NMI queued after currently running handler */
512 	bool nmi_injected;    /* Trying to inject an NMI this entry */
513 	bool smi_pending;    /* SMI queued after currently running handler */
514 
515 	struct kvm_mtrr mtrr_state;
516 	u64 pat;
517 
518 	unsigned switch_db_regs;
519 	unsigned long db[KVM_NR_DB_REGS];
520 	unsigned long dr6;
521 	unsigned long dr7;
522 	unsigned long eff_db[KVM_NR_DB_REGS];
523 	unsigned long guest_debug_dr7;
524 
525 	u64 mcg_cap;
526 	u64 mcg_status;
527 	u64 mcg_ctl;
528 	u64 *mce_banks;
529 
530 	/* Cache MMIO info */
531 	u64 mmio_gva;
532 	unsigned access;
533 	gfn_t mmio_gfn;
534 	u64 mmio_gen;
535 
536 	struct kvm_pmu pmu;
537 
538 	/* used for guest single stepping over the given code position */
539 	unsigned long singlestep_rip;
540 
541 	struct kvm_vcpu_hv hyperv;
542 
543 	cpumask_var_t wbinvd_dirty_mask;
544 
545 	unsigned long last_retry_eip;
546 	unsigned long last_retry_addr;
547 
548 	struct {
549 		bool halted;
550 		gfn_t gfns[roundup_pow_of_two(ASYNC_PF_PER_VCPU)];
551 		struct gfn_to_hva_cache data;
552 		u64 msr_val;
553 		u32 id;
554 		bool send_user_only;
555 	} apf;
556 
557 	/* OSVW MSRs (AMD only) */
558 	struct {
559 		u64 length;
560 		u64 status;
561 	} osvw;
562 
563 	struct {
564 		u64 msr_val;
565 		struct gfn_to_hva_cache data;
566 	} pv_eoi;
567 
568 	/*
569 	 * Indicate whether the access faults on its page table in guest
570 	 * which is set when fix page fault and used to detect unhandeable
571 	 * instruction.
572 	 */
573 	bool write_fault_to_shadow_pgtable;
574 
575 	/* set at EPT violation at this point */
576 	unsigned long exit_qualification;
577 
578 	/* pv related host specific info */
579 	struct {
580 		bool pv_unhalted;
581 	} pv;
582 
583 	int pending_ioapic_eoi;
584 	int pending_external_vector;
585 };
586 
587 struct kvm_lpage_info {
588 	int write_count;
589 };
590 
591 struct kvm_arch_memory_slot {
592 	unsigned long *rmap[KVM_NR_PAGE_SIZES];
593 	struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
594 };
595 
596 /*
597  * We use as the mode the number of bits allocated in the LDR for the
598  * logical processor ID.  It happens that these are all powers of two.
599  * This makes it is very easy to detect cases where the APICs are
600  * configured for multiple modes; in that case, we cannot use the map and
601  * hence cannot use kvm_irq_delivery_to_apic_fast either.
602  */
603 #define KVM_APIC_MODE_XAPIC_CLUSTER          4
604 #define KVM_APIC_MODE_XAPIC_FLAT             8
605 #define KVM_APIC_MODE_X2APIC                16
606 
607 struct kvm_apic_map {
608 	struct rcu_head rcu;
609 	u8 mode;
610 	struct kvm_lapic *phys_map[256];
611 	/* first index is cluster id second is cpu id in a cluster */
612 	struct kvm_lapic *logical_map[16][16];
613 };
614 
615 /* Hyper-V emulation context */
616 struct kvm_hv {
617 	u64 hv_guest_os_id;
618 	u64 hv_hypercall;
619 	u64 hv_tsc_page;
620 
621 	/* Hyper-v based guest crash (NT kernel bugcheck) parameters */
622 	u64 hv_crash_param[HV_X64_MSR_CRASH_PARAMS];
623 	u64 hv_crash_ctl;
624 };
625 
626 struct kvm_arch {
627 	unsigned int n_used_mmu_pages;
628 	unsigned int n_requested_mmu_pages;
629 	unsigned int n_max_mmu_pages;
630 	unsigned int indirect_shadow_pages;
631 	unsigned long mmu_valid_gen;
632 	struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
633 	/*
634 	 * Hash table of struct kvm_mmu_page.
635 	 */
636 	struct list_head active_mmu_pages;
637 	struct list_head zapped_obsolete_pages;
638 
639 	struct list_head assigned_dev_head;
640 	struct iommu_domain *iommu_domain;
641 	bool iommu_noncoherent;
642 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA
643 	atomic_t noncoherent_dma_count;
644 #define __KVM_HAVE_ARCH_ASSIGNED_DEVICE
645 	atomic_t assigned_device_count;
646 	struct kvm_pic *vpic;
647 	struct kvm_ioapic *vioapic;
648 	struct kvm_pit *vpit;
649 	atomic_t vapics_in_nmi_mode;
650 	struct mutex apic_map_lock;
651 	struct kvm_apic_map *apic_map;
652 
653 	unsigned int tss_addr;
654 	bool apic_access_page_done;
655 
656 	gpa_t wall_clock;
657 
658 	bool ept_identity_pagetable_done;
659 	gpa_t ept_identity_map_addr;
660 
661 	unsigned long irq_sources_bitmap;
662 	s64 kvmclock_offset;
663 	raw_spinlock_t tsc_write_lock;
664 	u64 last_tsc_nsec;
665 	u64 last_tsc_write;
666 	u32 last_tsc_khz;
667 	u64 cur_tsc_nsec;
668 	u64 cur_tsc_write;
669 	u64 cur_tsc_offset;
670 	u64 cur_tsc_generation;
671 	int nr_vcpus_matched_tsc;
672 
673 	spinlock_t pvclock_gtod_sync_lock;
674 	bool use_master_clock;
675 	u64 master_kernel_ns;
676 	cycle_t master_cycle_now;
677 	struct delayed_work kvmclock_update_work;
678 	struct delayed_work kvmclock_sync_work;
679 
680 	struct kvm_xen_hvm_config xen_hvm_config;
681 
682 	/* reads protected by irq_srcu, writes by irq_lock */
683 	struct hlist_head mask_notifier_list;
684 
685 	struct kvm_hv hyperv;
686 
687 	#ifdef CONFIG_KVM_MMU_AUDIT
688 	int audit_point;
689 	#endif
690 
691 	bool boot_vcpu_runs_old_kvmclock;
692 	u32 bsp_vcpu_id;
693 
694 	u64 disabled_quirks;
695 
696 	bool irqchip_split;
697 	u8 nr_reserved_ioapic_pins;
698 };
699 
700 struct kvm_vm_stat {
701 	u32 mmu_shadow_zapped;
702 	u32 mmu_pte_write;
703 	u32 mmu_pte_updated;
704 	u32 mmu_pde_zapped;
705 	u32 mmu_flooded;
706 	u32 mmu_recycled;
707 	u32 mmu_cache_miss;
708 	u32 mmu_unsync;
709 	u32 remote_tlb_flush;
710 	u32 lpages;
711 };
712 
713 struct kvm_vcpu_stat {
714 	u32 pf_fixed;
715 	u32 pf_guest;
716 	u32 tlb_flush;
717 	u32 invlpg;
718 
719 	u32 exits;
720 	u32 io_exits;
721 	u32 mmio_exits;
722 	u32 signal_exits;
723 	u32 irq_window_exits;
724 	u32 nmi_window_exits;
725 	u32 halt_exits;
726 	u32 halt_successful_poll;
727 	u32 halt_attempted_poll;
728 	u32 halt_wakeup;
729 	u32 request_irq_exits;
730 	u32 irq_exits;
731 	u32 host_state_reload;
732 	u32 efer_reload;
733 	u32 fpu_reload;
734 	u32 insn_emulation;
735 	u32 insn_emulation_fail;
736 	u32 hypercalls;
737 	u32 irq_injections;
738 	u32 nmi_injections;
739 };
740 
741 struct x86_instruction_info;
742 
743 struct msr_data {
744 	bool host_initiated;
745 	u32 index;
746 	u64 data;
747 };
748 
749 struct kvm_lapic_irq {
750 	u32 vector;
751 	u16 delivery_mode;
752 	u16 dest_mode;
753 	bool level;
754 	u16 trig_mode;
755 	u32 shorthand;
756 	u32 dest_id;
757 	bool msi_redir_hint;
758 };
759 
760 struct kvm_x86_ops {
761 	int (*cpu_has_kvm_support)(void);          /* __init */
762 	int (*disabled_by_bios)(void);             /* __init */
763 	int (*hardware_enable)(void);
764 	void (*hardware_disable)(void);
765 	void (*check_processor_compatibility)(void *rtn);
766 	int (*hardware_setup)(void);               /* __init */
767 	void (*hardware_unsetup)(void);            /* __exit */
768 	bool (*cpu_has_accelerated_tpr)(void);
769 	bool (*cpu_has_high_real_mode_segbase)(void);
770 	void (*cpuid_update)(struct kvm_vcpu *vcpu);
771 
772 	/* Create, but do not attach this VCPU */
773 	struct kvm_vcpu *(*vcpu_create)(struct kvm *kvm, unsigned id);
774 	void (*vcpu_free)(struct kvm_vcpu *vcpu);
775 	void (*vcpu_reset)(struct kvm_vcpu *vcpu, bool init_event);
776 
777 	void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
778 	void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
779 	void (*vcpu_put)(struct kvm_vcpu *vcpu);
780 
781 	void (*update_bp_intercept)(struct kvm_vcpu *vcpu);
782 	int (*get_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
783 	int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
784 	u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
785 	void (*get_segment)(struct kvm_vcpu *vcpu,
786 			    struct kvm_segment *var, int seg);
787 	int (*get_cpl)(struct kvm_vcpu *vcpu);
788 	void (*set_segment)(struct kvm_vcpu *vcpu,
789 			    struct kvm_segment *var, int seg);
790 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
791 	void (*decache_cr0_guest_bits)(struct kvm_vcpu *vcpu);
792 	void (*decache_cr3)(struct kvm_vcpu *vcpu);
793 	void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu);
794 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
795 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
796 	int (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
797 	void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
798 	void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
799 	void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
800 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
801 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
802 	u64 (*get_dr6)(struct kvm_vcpu *vcpu);
803 	void (*set_dr6)(struct kvm_vcpu *vcpu, unsigned long value);
804 	void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu);
805 	void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
806 	void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
807 	unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
808 	void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
809 	void (*fpu_activate)(struct kvm_vcpu *vcpu);
810 	void (*fpu_deactivate)(struct kvm_vcpu *vcpu);
811 
812 	void (*tlb_flush)(struct kvm_vcpu *vcpu);
813 
814 	void (*run)(struct kvm_vcpu *vcpu);
815 	int (*handle_exit)(struct kvm_vcpu *vcpu);
816 	void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
817 	void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
818 	u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu);
819 	void (*patch_hypercall)(struct kvm_vcpu *vcpu,
820 				unsigned char *hypercall_addr);
821 	void (*set_irq)(struct kvm_vcpu *vcpu);
822 	void (*set_nmi)(struct kvm_vcpu *vcpu);
823 	void (*queue_exception)(struct kvm_vcpu *vcpu, unsigned nr,
824 				bool has_error_code, u32 error_code,
825 				bool reinject);
826 	void (*cancel_injection)(struct kvm_vcpu *vcpu);
827 	int (*interrupt_allowed)(struct kvm_vcpu *vcpu);
828 	int (*nmi_allowed)(struct kvm_vcpu *vcpu);
829 	bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
830 	void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
831 	void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
832 	void (*enable_irq_window)(struct kvm_vcpu *vcpu);
833 	void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
834 	int (*cpu_uses_apicv)(struct kvm_vcpu *vcpu);
835 	void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
836 	void (*hwapic_isr_update)(struct kvm *kvm, int isr);
837 	void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu);
838 	void (*set_virtual_x2apic_mode)(struct kvm_vcpu *vcpu, bool set);
839 	void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu, hpa_t hpa);
840 	void (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
841 	void (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
842 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
843 	int (*get_tdp_level)(void);
844 	u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
845 	int (*get_lpage_level)(void);
846 	bool (*rdtscp_supported)(void);
847 	bool (*invpcid_supported)(void);
848 	void (*adjust_tsc_offset_guest)(struct kvm_vcpu *vcpu, s64 adjustment);
849 
850 	void (*set_tdp_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
851 
852 	void (*set_supported_cpuid)(u32 func, struct kvm_cpuid_entry2 *entry);
853 
854 	bool (*has_wbinvd_exit)(void);
855 
856 	u64 (*read_tsc_offset)(struct kvm_vcpu *vcpu);
857 	void (*write_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
858 
859 	u64 (*read_l1_tsc)(struct kvm_vcpu *vcpu, u64 host_tsc);
860 
861 	void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2);
862 
863 	int (*check_intercept)(struct kvm_vcpu *vcpu,
864 			       struct x86_instruction_info *info,
865 			       enum x86_intercept_stage stage);
866 	void (*handle_external_intr)(struct kvm_vcpu *vcpu);
867 	bool (*mpx_supported)(void);
868 	bool (*xsaves_supported)(void);
869 
870 	int (*check_nested_events)(struct kvm_vcpu *vcpu, bool external_intr);
871 
872 	void (*sched_in)(struct kvm_vcpu *kvm, int cpu);
873 
874 	/*
875 	 * Arch-specific dirty logging hooks. These hooks are only supposed to
876 	 * be valid if the specific arch has hardware-accelerated dirty logging
877 	 * mechanism. Currently only for PML on VMX.
878 	 *
879 	 *  - slot_enable_log_dirty:
880 	 *	called when enabling log dirty mode for the slot.
881 	 *  - slot_disable_log_dirty:
882 	 *	called when disabling log dirty mode for the slot.
883 	 *	also called when slot is created with log dirty disabled.
884 	 *  - flush_log_dirty:
885 	 *	called before reporting dirty_bitmap to userspace.
886 	 *  - enable_log_dirty_pt_masked:
887 	 *	called when reenabling log dirty for the GFNs in the mask after
888 	 *	corresponding bits are cleared in slot->dirty_bitmap.
889 	 */
890 	void (*slot_enable_log_dirty)(struct kvm *kvm,
891 				      struct kvm_memory_slot *slot);
892 	void (*slot_disable_log_dirty)(struct kvm *kvm,
893 				       struct kvm_memory_slot *slot);
894 	void (*flush_log_dirty)(struct kvm *kvm);
895 	void (*enable_log_dirty_pt_masked)(struct kvm *kvm,
896 					   struct kvm_memory_slot *slot,
897 					   gfn_t offset, unsigned long mask);
898 	/* pmu operations of sub-arch */
899 	const struct kvm_pmu_ops *pmu_ops;
900 
901 	/*
902 	 * Architecture specific hooks for vCPU blocking due to
903 	 * HLT instruction.
904 	 * Returns for .pre_block():
905 	 *    - 0 means continue to block the vCPU.
906 	 *    - 1 means we cannot block the vCPU since some event
907 	 *        happens during this period, such as, 'ON' bit in
908 	 *        posted-interrupts descriptor is set.
909 	 */
910 	int (*pre_block)(struct kvm_vcpu *vcpu);
911 	void (*post_block)(struct kvm_vcpu *vcpu);
912 	int (*update_pi_irte)(struct kvm *kvm, unsigned int host_irq,
913 			      uint32_t guest_irq, bool set);
914 };
915 
916 struct kvm_arch_async_pf {
917 	u32 token;
918 	gfn_t gfn;
919 	unsigned long cr3;
920 	bool direct_map;
921 };
922 
923 extern struct kvm_x86_ops *kvm_x86_ops;
924 
925 int kvm_mmu_module_init(void);
926 void kvm_mmu_module_exit(void);
927 
928 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
929 int kvm_mmu_create(struct kvm_vcpu *vcpu);
930 void kvm_mmu_setup(struct kvm_vcpu *vcpu);
931 void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
932 		u64 dirty_mask, u64 nx_mask, u64 x_mask);
933 
934 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
935 void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
936 				      struct kvm_memory_slot *memslot);
937 void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
938 				   const struct kvm_memory_slot *memslot);
939 void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
940 				   struct kvm_memory_slot *memslot);
941 void kvm_mmu_slot_largepage_remove_write_access(struct kvm *kvm,
942 					struct kvm_memory_slot *memslot);
943 void kvm_mmu_slot_set_dirty(struct kvm *kvm,
944 			    struct kvm_memory_slot *memslot);
945 void kvm_mmu_clear_dirty_pt_masked(struct kvm *kvm,
946 				   struct kvm_memory_slot *slot,
947 				   gfn_t gfn_offset, unsigned long mask);
948 void kvm_mmu_zap_all(struct kvm *kvm);
949 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, struct kvm_memslots *slots);
950 unsigned int kvm_mmu_calculate_mmu_pages(struct kvm *kvm);
951 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int kvm_nr_mmu_pages);
952 
953 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
954 
955 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
956 			  const void *val, int bytes);
957 
958 struct kvm_irq_mask_notifier {
959 	void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
960 	int irq;
961 	struct hlist_node link;
962 };
963 
964 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
965 				    struct kvm_irq_mask_notifier *kimn);
966 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
967 				      struct kvm_irq_mask_notifier *kimn);
968 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
969 			     bool mask);
970 
971 extern bool tdp_enabled;
972 
973 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
974 
975 /* control of guest tsc rate supported? */
976 extern bool kvm_has_tsc_control;
977 /* maximum supported tsc_khz for guests */
978 extern u32  kvm_max_guest_tsc_khz;
979 /* number of bits of the fractional part of the TSC scaling ratio */
980 extern u8   kvm_tsc_scaling_ratio_frac_bits;
981 /* maximum allowed value of TSC scaling ratio */
982 extern u64  kvm_max_tsc_scaling_ratio;
983 
984 enum emulation_result {
985 	EMULATE_DONE,         /* no further processing */
986 	EMULATE_USER_EXIT,    /* kvm_run ready for userspace exit */
987 	EMULATE_FAIL,         /* can't emulate this instruction */
988 };
989 
990 #define EMULTYPE_NO_DECODE	    (1 << 0)
991 #define EMULTYPE_TRAP_UD	    (1 << 1)
992 #define EMULTYPE_SKIP		    (1 << 2)
993 #define EMULTYPE_RETRY		    (1 << 3)
994 #define EMULTYPE_NO_REEXECUTE	    (1 << 4)
995 int x86_emulate_instruction(struct kvm_vcpu *vcpu, unsigned long cr2,
996 			    int emulation_type, void *insn, int insn_len);
997 
998 static inline int emulate_instruction(struct kvm_vcpu *vcpu,
999 			int emulation_type)
1000 {
1001 	return x86_emulate_instruction(vcpu, 0, emulation_type, NULL, 0);
1002 }
1003 
1004 void kvm_enable_efer_bits(u64);
1005 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
1006 int kvm_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr);
1007 int kvm_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr);
1008 
1009 struct x86_emulate_ctxt;
1010 
1011 int kvm_fast_pio_out(struct kvm_vcpu *vcpu, int size, unsigned short port);
1012 void kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
1013 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
1014 int kvm_vcpu_halt(struct kvm_vcpu *vcpu);
1015 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
1016 
1017 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
1018 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
1019 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector);
1020 
1021 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
1022 		    int reason, bool has_error_code, u32 error_code);
1023 
1024 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
1025 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
1026 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
1027 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
1028 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
1029 int kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
1030 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
1031 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
1032 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
1033 int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr);
1034 
1035 int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1036 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1037 
1038 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
1039 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
1040 bool kvm_rdpmc(struct kvm_vcpu *vcpu);
1041 
1042 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1043 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1044 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1045 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1046 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
1047 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1048 			    gfn_t gfn, void *data, int offset, int len,
1049 			    u32 access);
1050 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
1051 bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr);
1052 
1053 static inline int __kvm_irq_line_state(unsigned long *irq_state,
1054 				       int irq_source_id, int level)
1055 {
1056 	/* Logical OR for level trig interrupt */
1057 	if (level)
1058 		__set_bit(irq_source_id, irq_state);
1059 	else
1060 		__clear_bit(irq_source_id, irq_state);
1061 
1062 	return !!(*irq_state);
1063 }
1064 
1065 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
1066 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
1067 
1068 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
1069 
1070 void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
1071 		       const u8 *new, int bytes);
1072 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
1073 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
1074 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
1075 int kvm_mmu_load(struct kvm_vcpu *vcpu);
1076 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
1077 void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu);
1078 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1079 			   struct x86_exception *exception);
1080 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
1081 			      struct x86_exception *exception);
1082 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
1083 			       struct x86_exception *exception);
1084 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
1085 			       struct x86_exception *exception);
1086 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
1087 				struct x86_exception *exception);
1088 
1089 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
1090 
1091 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva, u32 error_code,
1092 		       void *insn, int insn_len);
1093 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
1094 void kvm_mmu_new_cr3(struct kvm_vcpu *vcpu);
1095 
1096 void kvm_enable_tdp(void);
1097 void kvm_disable_tdp(void);
1098 
1099 static inline gpa_t translate_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1100 				  struct x86_exception *exception)
1101 {
1102 	return gpa;
1103 }
1104 
1105 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
1106 {
1107 	struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
1108 
1109 	return (struct kvm_mmu_page *)page_private(page);
1110 }
1111 
1112 static inline u16 kvm_read_ldt(void)
1113 {
1114 	u16 ldt;
1115 	asm("sldt %0" : "=g"(ldt));
1116 	return ldt;
1117 }
1118 
1119 static inline void kvm_load_ldt(u16 sel)
1120 {
1121 	asm("lldt %0" : : "rm"(sel));
1122 }
1123 
1124 #ifdef CONFIG_X86_64
1125 static inline unsigned long read_msr(unsigned long msr)
1126 {
1127 	u64 value;
1128 
1129 	rdmsrl(msr, value);
1130 	return value;
1131 }
1132 #endif
1133 
1134 static inline u32 get_rdx_init_val(void)
1135 {
1136 	return 0x600; /* P6 family */
1137 }
1138 
1139 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
1140 {
1141 	kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
1142 }
1143 
1144 static inline u64 get_canonical(u64 la)
1145 {
1146 	return ((int64_t)la << 16) >> 16;
1147 }
1148 
1149 static inline bool is_noncanonical_address(u64 la)
1150 {
1151 #ifdef CONFIG_X86_64
1152 	return get_canonical(la) != la;
1153 #else
1154 	return false;
1155 #endif
1156 }
1157 
1158 #define TSS_IOPB_BASE_OFFSET 0x66
1159 #define TSS_BASE_SIZE 0x68
1160 #define TSS_IOPB_SIZE (65536 / 8)
1161 #define TSS_REDIRECTION_SIZE (256 / 8)
1162 #define RMODE_TSS_SIZE							\
1163 	(TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
1164 
1165 enum {
1166 	TASK_SWITCH_CALL = 0,
1167 	TASK_SWITCH_IRET = 1,
1168 	TASK_SWITCH_JMP = 2,
1169 	TASK_SWITCH_GATE = 3,
1170 };
1171 
1172 #define HF_GIF_MASK		(1 << 0)
1173 #define HF_HIF_MASK		(1 << 1)
1174 #define HF_VINTR_MASK		(1 << 2)
1175 #define HF_NMI_MASK		(1 << 3)
1176 #define HF_IRET_MASK		(1 << 4)
1177 #define HF_GUEST_MASK		(1 << 5) /* VCPU is in guest-mode */
1178 #define HF_SMM_MASK		(1 << 6)
1179 #define HF_SMM_INSIDE_NMI_MASK	(1 << 7)
1180 
1181 #define __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
1182 #define KVM_ADDRESS_SPACE_NUM 2
1183 
1184 #define kvm_arch_vcpu_memslots_id(vcpu) ((vcpu)->arch.hflags & HF_SMM_MASK ? 1 : 0)
1185 #define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, (role).smm)
1186 
1187 /*
1188  * Hardware virtualization extension instructions may fault if a
1189  * reboot turns off virtualization while processes are running.
1190  * Trap the fault and ignore the instruction if that happens.
1191  */
1192 asmlinkage void kvm_spurious_fault(void);
1193 
1194 #define ____kvm_handle_fault_on_reboot(insn, cleanup_insn)	\
1195 	"666: " insn "\n\t" \
1196 	"668: \n\t"                           \
1197 	".pushsection .fixup, \"ax\" \n" \
1198 	"667: \n\t" \
1199 	cleanup_insn "\n\t"		      \
1200 	"cmpb $0, kvm_rebooting \n\t"	      \
1201 	"jne 668b \n\t"      		      \
1202 	__ASM_SIZE(push) " $666b \n\t"	      \
1203 	"call kvm_spurious_fault \n\t"	      \
1204 	".popsection \n\t" \
1205 	_ASM_EXTABLE(666b, 667b)
1206 
1207 #define __kvm_handle_fault_on_reboot(insn)		\
1208 	____kvm_handle_fault_on_reboot(insn, "")
1209 
1210 #define KVM_ARCH_WANT_MMU_NOTIFIER
1211 int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
1212 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end);
1213 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
1214 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
1215 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
1216 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1217 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1218 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1219 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1220 void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event);
1221 void kvm_vcpu_reload_apic_access_page(struct kvm_vcpu *vcpu);
1222 void kvm_arch_mmu_notifier_invalidate_page(struct kvm *kvm,
1223 					   unsigned long address);
1224 
1225 void kvm_define_shared_msr(unsigned index, u32 msr);
1226 int kvm_set_shared_msr(unsigned index, u64 val, u64 mask);
1227 
1228 u64 kvm_scale_tsc(struct kvm_vcpu *vcpu, u64 tsc);
1229 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc);
1230 
1231 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu);
1232 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1233 
1234 void kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1235 				     struct kvm_async_pf *work);
1236 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1237 				 struct kvm_async_pf *work);
1238 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1239 			       struct kvm_async_pf *work);
1240 bool kvm_arch_can_inject_async_page_present(struct kvm_vcpu *vcpu);
1241 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1242 
1243 void kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1244 
1245 int kvm_is_in_guest(void);
1246 
1247 int __x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa, u32 size);
1248 int x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa, u32 size);
1249 bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu);
1250 bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu);
1251 
1252 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
1253 			     struct kvm_vcpu **dest_vcpu);
1254 
1255 void kvm_set_msi_irq(struct kvm_kernel_irq_routing_entry *e,
1256 		     struct kvm_lapic_irq *irq);
1257 
1258 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu) {}
1259 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu) {}
1260 
1261 #endif /* _ASM_X86_KVM_HOST_H */
1262