xref: /openbmc/linux/arch/x86/include/asm/kvm_host.h (revision 76a4f7cc)
1 /* SPDX-License-Identifier: GPL-2.0-only */
2 /*
3  * Kernel-based Virtual Machine driver for Linux
4  *
5  * This header defines architecture specific interfaces, x86 version
6  */
7 
8 #ifndef _ASM_X86_KVM_HOST_H
9 #define _ASM_X86_KVM_HOST_H
10 
11 #include <linux/types.h>
12 #include <linux/mm.h>
13 #include <linux/mmu_notifier.h>
14 #include <linux/tracepoint.h>
15 #include <linux/cpumask.h>
16 #include <linux/irq_work.h>
17 #include <linux/irq.h>
18 
19 #include <linux/kvm.h>
20 #include <linux/kvm_para.h>
21 #include <linux/kvm_types.h>
22 #include <linux/perf_event.h>
23 #include <linux/pvclock_gtod.h>
24 #include <linux/clocksource.h>
25 #include <linux/irqbypass.h>
26 #include <linux/hyperv.h>
27 
28 #include <asm/apic.h>
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 #include <asm/kvm_page_track.h>
35 #include <asm/kvm_vcpu_regs.h>
36 #include <asm/hyperv-tlfs.h>
37 
38 #define __KVM_HAVE_ARCH_VCPU_DEBUGFS
39 
40 #define KVM_MAX_VCPUS 1024
41 #define KVM_SOFT_MAX_VCPUS 710
42 
43 /*
44  * In x86, the VCPU ID corresponds to the APIC ID, and APIC IDs
45  * might be larger than the actual number of VCPUs because the
46  * APIC ID encodes CPU topology information.
47  *
48  * In the worst case, we'll need less than one extra bit for the
49  * Core ID, and less than one extra bit for the Package (Die) ID,
50  * so ratio of 4 should be enough.
51  */
52 #define KVM_VCPU_ID_RATIO 4
53 #define KVM_MAX_VCPU_ID (KVM_MAX_VCPUS * KVM_VCPU_ID_RATIO)
54 
55 /* memory slots that are not exposed to userspace */
56 #define KVM_PRIVATE_MEM_SLOTS 3
57 
58 #define KVM_HALT_POLL_NS_DEFAULT 200000
59 
60 #define KVM_IRQCHIP_NUM_PINS  KVM_IOAPIC_NUM_PINS
61 
62 #define KVM_DIRTY_LOG_MANUAL_CAPS   (KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE | \
63 					KVM_DIRTY_LOG_INITIALLY_SET)
64 
65 #define KVM_BUS_LOCK_DETECTION_VALID_MODE	(KVM_BUS_LOCK_DETECTION_OFF | \
66 						 KVM_BUS_LOCK_DETECTION_EXIT)
67 
68 /* x86-specific vcpu->requests bit members */
69 #define KVM_REQ_MIGRATE_TIMER		KVM_ARCH_REQ(0)
70 #define KVM_REQ_REPORT_TPR_ACCESS	KVM_ARCH_REQ(1)
71 #define KVM_REQ_TRIPLE_FAULT		KVM_ARCH_REQ(2)
72 #define KVM_REQ_MMU_SYNC		KVM_ARCH_REQ(3)
73 #define KVM_REQ_CLOCK_UPDATE		KVM_ARCH_REQ(4)
74 #define KVM_REQ_LOAD_MMU_PGD		KVM_ARCH_REQ(5)
75 #define KVM_REQ_EVENT			KVM_ARCH_REQ(6)
76 #define KVM_REQ_APF_HALT		KVM_ARCH_REQ(7)
77 #define KVM_REQ_STEAL_UPDATE		KVM_ARCH_REQ(8)
78 #define KVM_REQ_NMI			KVM_ARCH_REQ(9)
79 #define KVM_REQ_PMU			KVM_ARCH_REQ(10)
80 #define KVM_REQ_PMI			KVM_ARCH_REQ(11)
81 #define KVM_REQ_SMI			KVM_ARCH_REQ(12)
82 #define KVM_REQ_MASTERCLOCK_UPDATE	KVM_ARCH_REQ(13)
83 #define KVM_REQ_MCLOCK_INPROGRESS \
84 	KVM_ARCH_REQ_FLAGS(14, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
85 #define KVM_REQ_SCAN_IOAPIC \
86 	KVM_ARCH_REQ_FLAGS(15, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
87 #define KVM_REQ_GLOBAL_CLOCK_UPDATE	KVM_ARCH_REQ(16)
88 #define KVM_REQ_APIC_PAGE_RELOAD \
89 	KVM_ARCH_REQ_FLAGS(17, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
90 #define KVM_REQ_HV_CRASH		KVM_ARCH_REQ(18)
91 #define KVM_REQ_IOAPIC_EOI_EXIT		KVM_ARCH_REQ(19)
92 #define KVM_REQ_HV_RESET		KVM_ARCH_REQ(20)
93 #define KVM_REQ_HV_EXIT			KVM_ARCH_REQ(21)
94 #define KVM_REQ_HV_STIMER		KVM_ARCH_REQ(22)
95 #define KVM_REQ_LOAD_EOI_EXITMAP	KVM_ARCH_REQ(23)
96 #define KVM_REQ_GET_NESTED_STATE_PAGES	KVM_ARCH_REQ(24)
97 #define KVM_REQ_APICV_UPDATE \
98 	KVM_ARCH_REQ_FLAGS(25, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
99 #define KVM_REQ_TLB_FLUSH_CURRENT	KVM_ARCH_REQ(26)
100 #define KVM_REQ_TLB_FLUSH_GUEST \
101 	KVM_ARCH_REQ_FLAGS(27, KVM_REQUEST_NO_WAKEUP)
102 #define KVM_REQ_APF_READY		KVM_ARCH_REQ(28)
103 #define KVM_REQ_MSR_FILTER_CHANGED	KVM_ARCH_REQ(29)
104 #define KVM_REQ_UPDATE_CPU_DIRTY_LOGGING \
105 	KVM_ARCH_REQ_FLAGS(30, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
106 
107 #define CR0_RESERVED_BITS                                               \
108 	(~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \
109 			  | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \
110 			  | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG))
111 
112 #define CR4_RESERVED_BITS                                               \
113 	(~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\
114 			  | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE     \
115 			  | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR | X86_CR4_PCIDE \
116 			  | X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_FSGSBASE \
117 			  | X86_CR4_OSXMMEXCPT | X86_CR4_LA57 | X86_CR4_VMXE \
118 			  | X86_CR4_SMAP | X86_CR4_PKE | X86_CR4_UMIP))
119 
120 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR)
121 
122 
123 
124 #define INVALID_PAGE (~(hpa_t)0)
125 #define VALID_PAGE(x) ((x) != INVALID_PAGE)
126 
127 #define UNMAPPED_GVA (~(gpa_t)0)
128 #define INVALID_GPA (~(gpa_t)0)
129 
130 /* KVM Hugepage definitions for x86 */
131 #define KVM_MAX_HUGEPAGE_LEVEL	PG_LEVEL_1G
132 #define KVM_NR_PAGE_SIZES	(KVM_MAX_HUGEPAGE_LEVEL - PG_LEVEL_4K + 1)
133 #define KVM_HPAGE_GFN_SHIFT(x)	(((x) - 1) * 9)
134 #define KVM_HPAGE_SHIFT(x)	(PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
135 #define KVM_HPAGE_SIZE(x)	(1UL << KVM_HPAGE_SHIFT(x))
136 #define KVM_HPAGE_MASK(x)	(~(KVM_HPAGE_SIZE(x) - 1))
137 #define KVM_PAGES_PER_HPAGE(x)	(KVM_HPAGE_SIZE(x) / PAGE_SIZE)
138 
139 #define KVM_PERMILLE_MMU_PAGES 20
140 #define KVM_MIN_ALLOC_MMU_PAGES 64UL
141 #define KVM_MMU_HASH_SHIFT 12
142 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
143 #define KVM_MIN_FREE_MMU_PAGES 5
144 #define KVM_REFILL_PAGES 25
145 #define KVM_MAX_CPUID_ENTRIES 256
146 #define KVM_NR_FIXED_MTRR_REGION 88
147 #define KVM_NR_VAR_MTRR 8
148 
149 #define ASYNC_PF_PER_VCPU 64
150 
151 enum kvm_reg {
152 	VCPU_REGS_RAX = __VCPU_REGS_RAX,
153 	VCPU_REGS_RCX = __VCPU_REGS_RCX,
154 	VCPU_REGS_RDX = __VCPU_REGS_RDX,
155 	VCPU_REGS_RBX = __VCPU_REGS_RBX,
156 	VCPU_REGS_RSP = __VCPU_REGS_RSP,
157 	VCPU_REGS_RBP = __VCPU_REGS_RBP,
158 	VCPU_REGS_RSI = __VCPU_REGS_RSI,
159 	VCPU_REGS_RDI = __VCPU_REGS_RDI,
160 #ifdef CONFIG_X86_64
161 	VCPU_REGS_R8  = __VCPU_REGS_R8,
162 	VCPU_REGS_R9  = __VCPU_REGS_R9,
163 	VCPU_REGS_R10 = __VCPU_REGS_R10,
164 	VCPU_REGS_R11 = __VCPU_REGS_R11,
165 	VCPU_REGS_R12 = __VCPU_REGS_R12,
166 	VCPU_REGS_R13 = __VCPU_REGS_R13,
167 	VCPU_REGS_R14 = __VCPU_REGS_R14,
168 	VCPU_REGS_R15 = __VCPU_REGS_R15,
169 #endif
170 	VCPU_REGS_RIP,
171 	NR_VCPU_REGS,
172 
173 	VCPU_EXREG_PDPTR = NR_VCPU_REGS,
174 	VCPU_EXREG_CR0,
175 	VCPU_EXREG_CR3,
176 	VCPU_EXREG_CR4,
177 	VCPU_EXREG_RFLAGS,
178 	VCPU_EXREG_SEGMENTS,
179 	VCPU_EXREG_EXIT_INFO_1,
180 	VCPU_EXREG_EXIT_INFO_2,
181 };
182 
183 enum {
184 	VCPU_SREG_ES,
185 	VCPU_SREG_CS,
186 	VCPU_SREG_SS,
187 	VCPU_SREG_DS,
188 	VCPU_SREG_FS,
189 	VCPU_SREG_GS,
190 	VCPU_SREG_TR,
191 	VCPU_SREG_LDTR,
192 };
193 
194 enum exit_fastpath_completion {
195 	EXIT_FASTPATH_NONE,
196 	EXIT_FASTPATH_REENTER_GUEST,
197 	EXIT_FASTPATH_EXIT_HANDLED,
198 };
199 typedef enum exit_fastpath_completion fastpath_t;
200 
201 struct x86_emulate_ctxt;
202 struct x86_exception;
203 enum x86_intercept;
204 enum x86_intercept_stage;
205 
206 #define KVM_NR_DB_REGS	4
207 
208 #define DR6_BUS_LOCK   (1 << 11)
209 #define DR6_BD		(1 << 13)
210 #define DR6_BS		(1 << 14)
211 #define DR6_BT		(1 << 15)
212 #define DR6_RTM		(1 << 16)
213 /*
214  * DR6_ACTIVE_LOW combines fixed-1 and active-low bits.
215  * We can regard all the bits in DR6_FIXED_1 as active_low bits;
216  * they will never be 0 for now, but when they are defined
217  * in the future it will require no code change.
218  *
219  * DR6_ACTIVE_LOW is also used as the init/reset value for DR6.
220  */
221 #define DR6_ACTIVE_LOW	0xffff0ff0
222 #define DR6_VOLATILE	0x0001e80f
223 #define DR6_FIXED_1	(DR6_ACTIVE_LOW & ~DR6_VOLATILE)
224 
225 #define DR7_BP_EN_MASK	0x000000ff
226 #define DR7_GE		(1 << 9)
227 #define DR7_GD		(1 << 13)
228 #define DR7_FIXED_1	0x00000400
229 #define DR7_VOLATILE	0xffff2bff
230 
231 #define KVM_GUESTDBG_VALID_MASK \
232 	(KVM_GUESTDBG_ENABLE | \
233 	KVM_GUESTDBG_SINGLESTEP | \
234 	KVM_GUESTDBG_USE_HW_BP | \
235 	KVM_GUESTDBG_USE_SW_BP | \
236 	KVM_GUESTDBG_INJECT_BP | \
237 	KVM_GUESTDBG_INJECT_DB | \
238 	KVM_GUESTDBG_BLOCKIRQ)
239 
240 
241 #define PFERR_PRESENT_BIT 0
242 #define PFERR_WRITE_BIT 1
243 #define PFERR_USER_BIT 2
244 #define PFERR_RSVD_BIT 3
245 #define PFERR_FETCH_BIT 4
246 #define PFERR_PK_BIT 5
247 #define PFERR_SGX_BIT 15
248 #define PFERR_GUEST_FINAL_BIT 32
249 #define PFERR_GUEST_PAGE_BIT 33
250 
251 #define PFERR_PRESENT_MASK (1U << PFERR_PRESENT_BIT)
252 #define PFERR_WRITE_MASK (1U << PFERR_WRITE_BIT)
253 #define PFERR_USER_MASK (1U << PFERR_USER_BIT)
254 #define PFERR_RSVD_MASK (1U << PFERR_RSVD_BIT)
255 #define PFERR_FETCH_MASK (1U << PFERR_FETCH_BIT)
256 #define PFERR_PK_MASK (1U << PFERR_PK_BIT)
257 #define PFERR_SGX_MASK (1U << PFERR_SGX_BIT)
258 #define PFERR_GUEST_FINAL_MASK (1ULL << PFERR_GUEST_FINAL_BIT)
259 #define PFERR_GUEST_PAGE_MASK (1ULL << PFERR_GUEST_PAGE_BIT)
260 
261 #define PFERR_NESTED_GUEST_PAGE (PFERR_GUEST_PAGE_MASK |	\
262 				 PFERR_WRITE_MASK |		\
263 				 PFERR_PRESENT_MASK)
264 
265 /* apic attention bits */
266 #define KVM_APIC_CHECK_VAPIC	0
267 /*
268  * The following bit is set with PV-EOI, unset on EOI.
269  * We detect PV-EOI changes by guest by comparing
270  * this bit with PV-EOI in guest memory.
271  * See the implementation in apic_update_pv_eoi.
272  */
273 #define KVM_APIC_PV_EOI_PENDING	1
274 
275 struct kvm_kernel_irq_routing_entry;
276 
277 /*
278  * kvm_mmu_page_role tracks the properties of a shadow page (where shadow page
279  * also includes TDP pages) to determine whether or not a page can be used in
280  * the given MMU context.  This is a subset of the overall kvm_mmu_role to
281  * minimize the size of kvm_memory_slot.arch.gfn_track, i.e. allows allocating
282  * 2 bytes per gfn instead of 4 bytes per gfn.
283  *
284  * Indirect upper-level shadow pages are tracked for write-protection via
285  * gfn_track.  As above, gfn_track is a 16 bit counter, so KVM must not create
286  * more than 2^16-1 upper-level shadow pages at a single gfn, otherwise
287  * gfn_track will overflow and explosions will ensure.
288  *
289  * A unique shadow page (SP) for a gfn is created if and only if an existing SP
290  * cannot be reused.  The ability to reuse a SP is tracked by its role, which
291  * incorporates various mode bits and properties of the SP.  Roughly speaking,
292  * the number of unique SPs that can theoretically be created is 2^n, where n
293  * is the number of bits that are used to compute the role.
294  *
295  * But, even though there are 18 bits in the mask below, not all combinations
296  * of modes and flags are possible.  The maximum number of possible upper-level
297  * shadow pages for a single gfn is in the neighborhood of 2^13.
298  *
299  *   - invalid shadow pages are not accounted.
300  *   - level is effectively limited to four combinations, not 16 as the number
301  *     bits would imply, as 4k SPs are not tracked (allowed to go unsync).
302  *   - level is effectively unused for non-PAE paging because there is exactly
303  *     one upper level (see 4k SP exception above).
304  *   - quadrant is used only for non-PAE paging and is exclusive with
305  *     gpte_is_8_bytes.
306  *   - execonly and ad_disabled are used only for nested EPT, which makes it
307  *     exclusive with quadrant.
308  */
309 union kvm_mmu_page_role {
310 	u32 word;
311 	struct {
312 		unsigned level:4;
313 		unsigned gpte_is_8_bytes:1;
314 		unsigned quadrant:2;
315 		unsigned direct:1;
316 		unsigned access:3;
317 		unsigned invalid:1;
318 		unsigned efer_nx:1;
319 		unsigned cr0_wp:1;
320 		unsigned smep_andnot_wp:1;
321 		unsigned smap_andnot_wp:1;
322 		unsigned ad_disabled:1;
323 		unsigned guest_mode:1;
324 		unsigned :6;
325 
326 		/*
327 		 * This is left at the top of the word so that
328 		 * kvm_memslots_for_spte_role can extract it with a
329 		 * simple shift.  While there is room, give it a whole
330 		 * byte so it is also faster to load it from memory.
331 		 */
332 		unsigned smm:8;
333 	};
334 };
335 
336 /*
337  * kvm_mmu_extended_role complements kvm_mmu_page_role, tracking properties
338  * relevant to the current MMU configuration.   When loading CR0, CR4, or EFER,
339  * including on nested transitions, if nothing in the full role changes then
340  * MMU re-configuration can be skipped. @valid bit is set on first usage so we
341  * don't treat all-zero structure as valid data.
342  *
343  * The properties that are tracked in the extended role but not the page role
344  * are for things that either (a) do not affect the validity of the shadow page
345  * or (b) are indirectly reflected in the shadow page's role.  For example,
346  * CR4.PKE only affects permission checks for software walks of the guest page
347  * tables (because KVM doesn't support Protection Keys with shadow paging), and
348  * CR0.PG, CR4.PAE, and CR4.PSE are indirectly reflected in role.level.
349  *
350  * Note, SMEP and SMAP are not redundant with sm*p_andnot_wp in the page role.
351  * If CR0.WP=1, KVM can reuse shadow pages for the guest regardless of SMEP and
352  * SMAP, but the MMU's permission checks for software walks need to be SMEP and
353  * SMAP aware regardless of CR0.WP.
354  */
355 union kvm_mmu_extended_role {
356 	u32 word;
357 	struct {
358 		unsigned int valid:1;
359 		unsigned int execonly:1;
360 		unsigned int cr0_pg:1;
361 		unsigned int cr4_pae:1;
362 		unsigned int cr4_pse:1;
363 		unsigned int cr4_pke:1;
364 		unsigned int cr4_smap:1;
365 		unsigned int cr4_smep:1;
366 		unsigned int cr4_la57:1;
367 	};
368 };
369 
370 union kvm_mmu_role {
371 	u64 as_u64;
372 	struct {
373 		union kvm_mmu_page_role base;
374 		union kvm_mmu_extended_role ext;
375 	};
376 };
377 
378 struct kvm_rmap_head {
379 	unsigned long val;
380 };
381 
382 struct kvm_pio_request {
383 	unsigned long linear_rip;
384 	unsigned long count;
385 	int in;
386 	int port;
387 	int size;
388 };
389 
390 #define PT64_ROOT_MAX_LEVEL 5
391 
392 struct rsvd_bits_validate {
393 	u64 rsvd_bits_mask[2][PT64_ROOT_MAX_LEVEL];
394 	u64 bad_mt_xwr;
395 };
396 
397 struct kvm_mmu_root_info {
398 	gpa_t pgd;
399 	hpa_t hpa;
400 };
401 
402 #define KVM_MMU_ROOT_INFO_INVALID \
403 	((struct kvm_mmu_root_info) { .pgd = INVALID_PAGE, .hpa = INVALID_PAGE })
404 
405 #define KVM_MMU_NUM_PREV_ROOTS 3
406 
407 #define KVM_HAVE_MMU_RWLOCK
408 
409 struct kvm_mmu_page;
410 
411 /*
412  * x86 supports 4 paging modes (5-level 64-bit, 4-level 64-bit, 3-level 32-bit,
413  * and 2-level 32-bit).  The kvm_mmu structure abstracts the details of the
414  * current mmu mode.
415  */
416 struct kvm_mmu {
417 	unsigned long (*get_guest_pgd)(struct kvm_vcpu *vcpu);
418 	u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
419 	int (*page_fault)(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u32 err,
420 			  bool prefault);
421 	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
422 				  struct x86_exception *fault);
423 	gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gpa_t gva_or_gpa,
424 			    u32 access, struct x86_exception *exception);
425 	gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
426 			       struct x86_exception *exception);
427 	int (*sync_page)(struct kvm_vcpu *vcpu,
428 			 struct kvm_mmu_page *sp);
429 	void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva, hpa_t root_hpa);
430 	hpa_t root_hpa;
431 	gpa_t root_pgd;
432 	union kvm_mmu_role mmu_role;
433 	u8 root_level;
434 	u8 shadow_root_level;
435 	u8 ept_ad;
436 	bool direct_map;
437 	struct kvm_mmu_root_info prev_roots[KVM_MMU_NUM_PREV_ROOTS];
438 
439 	/*
440 	 * Bitmap; bit set = permission fault
441 	 * Byte index: page fault error code [4:1]
442 	 * Bit index: pte permissions in ACC_* format
443 	 */
444 	u8 permissions[16];
445 
446 	/*
447 	* The pkru_mask indicates if protection key checks are needed.  It
448 	* consists of 16 domains indexed by page fault error code bits [4:1],
449 	* with PFEC.RSVD replaced by ACC_USER_MASK from the page tables.
450 	* Each domain has 2 bits which are ANDed with AD and WD from PKRU.
451 	*/
452 	u32 pkru_mask;
453 
454 	u64 *pae_root;
455 	u64 *pml4_root;
456 	u64 *pml5_root;
457 
458 	/*
459 	 * check zero bits on shadow page table entries, these
460 	 * bits include not only hardware reserved bits but also
461 	 * the bits spte never used.
462 	 */
463 	struct rsvd_bits_validate shadow_zero_check;
464 
465 	struct rsvd_bits_validate guest_rsvd_check;
466 
467 	u64 pdptrs[4]; /* pae */
468 };
469 
470 struct kvm_tlb_range {
471 	u64 start_gfn;
472 	u64 pages;
473 };
474 
475 enum pmc_type {
476 	KVM_PMC_GP = 0,
477 	KVM_PMC_FIXED,
478 };
479 
480 struct kvm_pmc {
481 	enum pmc_type type;
482 	u8 idx;
483 	u64 counter;
484 	u64 eventsel;
485 	struct perf_event *perf_event;
486 	struct kvm_vcpu *vcpu;
487 	/*
488 	 * eventsel value for general purpose counters,
489 	 * ctrl value for fixed counters.
490 	 */
491 	u64 current_config;
492 	bool is_paused;
493 };
494 
495 struct kvm_pmu {
496 	unsigned nr_arch_gp_counters;
497 	unsigned nr_arch_fixed_counters;
498 	unsigned available_event_types;
499 	u64 fixed_ctr_ctrl;
500 	u64 global_ctrl;
501 	u64 global_status;
502 	u64 global_ovf_ctrl;
503 	u64 counter_bitmask[2];
504 	u64 global_ctrl_mask;
505 	u64 global_ovf_ctrl_mask;
506 	u64 reserved_bits;
507 	u8 version;
508 	struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
509 	struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED];
510 	struct irq_work irq_work;
511 	DECLARE_BITMAP(reprogram_pmi, X86_PMC_IDX_MAX);
512 	DECLARE_BITMAP(all_valid_pmc_idx, X86_PMC_IDX_MAX);
513 	DECLARE_BITMAP(pmc_in_use, X86_PMC_IDX_MAX);
514 
515 	/*
516 	 * The gate to release perf_events not marked in
517 	 * pmc_in_use only once in a vcpu time slice.
518 	 */
519 	bool need_cleanup;
520 
521 	/*
522 	 * The total number of programmed perf_events and it helps to avoid
523 	 * redundant check before cleanup if guest don't use vPMU at all.
524 	 */
525 	u8 event_count;
526 };
527 
528 struct kvm_pmu_ops;
529 
530 enum {
531 	KVM_DEBUGREG_BP_ENABLED = 1,
532 	KVM_DEBUGREG_WONT_EXIT = 2,
533 };
534 
535 struct kvm_mtrr_range {
536 	u64 base;
537 	u64 mask;
538 	struct list_head node;
539 };
540 
541 struct kvm_mtrr {
542 	struct kvm_mtrr_range var_ranges[KVM_NR_VAR_MTRR];
543 	mtrr_type fixed_ranges[KVM_NR_FIXED_MTRR_REGION];
544 	u64 deftype;
545 
546 	struct list_head head;
547 };
548 
549 /* Hyper-V SynIC timer */
550 struct kvm_vcpu_hv_stimer {
551 	struct hrtimer timer;
552 	int index;
553 	union hv_stimer_config config;
554 	u64 count;
555 	u64 exp_time;
556 	struct hv_message msg;
557 	bool msg_pending;
558 };
559 
560 /* Hyper-V synthetic interrupt controller (SynIC)*/
561 struct kvm_vcpu_hv_synic {
562 	u64 version;
563 	u64 control;
564 	u64 msg_page;
565 	u64 evt_page;
566 	atomic64_t sint[HV_SYNIC_SINT_COUNT];
567 	atomic_t sint_to_gsi[HV_SYNIC_SINT_COUNT];
568 	DECLARE_BITMAP(auto_eoi_bitmap, 256);
569 	DECLARE_BITMAP(vec_bitmap, 256);
570 	bool active;
571 	bool dont_zero_synic_pages;
572 };
573 
574 /* Hyper-V per vcpu emulation context */
575 struct kvm_vcpu_hv {
576 	struct kvm_vcpu *vcpu;
577 	u32 vp_index;
578 	u64 hv_vapic;
579 	s64 runtime_offset;
580 	struct kvm_vcpu_hv_synic synic;
581 	struct kvm_hyperv_exit exit;
582 	struct kvm_vcpu_hv_stimer stimer[HV_SYNIC_STIMER_COUNT];
583 	DECLARE_BITMAP(stimer_pending_bitmap, HV_SYNIC_STIMER_COUNT);
584 	cpumask_t tlb_flush;
585 	bool enforce_cpuid;
586 	struct {
587 		u32 features_eax; /* HYPERV_CPUID_FEATURES.EAX */
588 		u32 features_ebx; /* HYPERV_CPUID_FEATURES.EBX */
589 		u32 features_edx; /* HYPERV_CPUID_FEATURES.EDX */
590 		u32 enlightenments_eax; /* HYPERV_CPUID_ENLIGHTMENT_INFO.EAX */
591 		u32 enlightenments_ebx; /* HYPERV_CPUID_ENLIGHTMENT_INFO.EBX */
592 		u32 syndbg_cap_eax; /* HYPERV_CPUID_SYNDBG_PLATFORM_CAPABILITIES.EAX */
593 	} cpuid_cache;
594 };
595 
596 /* Xen HVM per vcpu emulation context */
597 struct kvm_vcpu_xen {
598 	u64 hypercall_rip;
599 	u32 current_runstate;
600 	bool vcpu_info_set;
601 	bool vcpu_time_info_set;
602 	bool runstate_set;
603 	struct gfn_to_hva_cache vcpu_info_cache;
604 	struct gfn_to_hva_cache vcpu_time_info_cache;
605 	struct gfn_to_hva_cache runstate_cache;
606 	u64 last_steal;
607 	u64 runstate_entry_time;
608 	u64 runstate_times[4];
609 };
610 
611 struct kvm_vcpu_arch {
612 	/*
613 	 * rip and regs accesses must go through
614 	 * kvm_{register,rip}_{read,write} functions.
615 	 */
616 	unsigned long regs[NR_VCPU_REGS];
617 	u32 regs_avail;
618 	u32 regs_dirty;
619 
620 	unsigned long cr0;
621 	unsigned long cr0_guest_owned_bits;
622 	unsigned long cr2;
623 	unsigned long cr3;
624 	unsigned long cr4;
625 	unsigned long cr4_guest_owned_bits;
626 	unsigned long cr4_guest_rsvd_bits;
627 	unsigned long cr8;
628 	u32 host_pkru;
629 	u32 pkru;
630 	u32 hflags;
631 	u64 efer;
632 	u64 apic_base;
633 	struct kvm_lapic *apic;    /* kernel irqchip context */
634 	bool apicv_active;
635 	bool load_eoi_exitmap_pending;
636 	DECLARE_BITMAP(ioapic_handled_vectors, 256);
637 	unsigned long apic_attention;
638 	int32_t apic_arb_prio;
639 	int mp_state;
640 	u64 ia32_misc_enable_msr;
641 	u64 smbase;
642 	u64 smi_count;
643 	bool tpr_access_reporting;
644 	bool xsaves_enabled;
645 	u64 ia32_xss;
646 	u64 microcode_version;
647 	u64 arch_capabilities;
648 	u64 perf_capabilities;
649 
650 	/*
651 	 * Paging state of the vcpu
652 	 *
653 	 * If the vcpu runs in guest mode with two level paging this still saves
654 	 * the paging mode of the l1 guest. This context is always used to
655 	 * handle faults.
656 	 */
657 	struct kvm_mmu *mmu;
658 
659 	/* Non-nested MMU for L1 */
660 	struct kvm_mmu root_mmu;
661 
662 	/* L1 MMU when running nested */
663 	struct kvm_mmu guest_mmu;
664 
665 	/*
666 	 * Paging state of an L2 guest (used for nested npt)
667 	 *
668 	 * This context will save all necessary information to walk page tables
669 	 * of an L2 guest. This context is only initialized for page table
670 	 * walking and not for faulting since we never handle l2 page faults on
671 	 * the host.
672 	 */
673 	struct kvm_mmu nested_mmu;
674 
675 	/*
676 	 * Pointer to the mmu context currently used for
677 	 * gva_to_gpa translations.
678 	 */
679 	struct kvm_mmu *walk_mmu;
680 
681 	struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
682 	struct kvm_mmu_memory_cache mmu_shadow_page_cache;
683 	struct kvm_mmu_memory_cache mmu_gfn_array_cache;
684 	struct kvm_mmu_memory_cache mmu_page_header_cache;
685 
686 	/*
687 	 * QEMU userspace and the guest each have their own FPU state.
688 	 * In vcpu_run, we switch between the user and guest FPU contexts.
689 	 * While running a VCPU, the VCPU thread will have the guest FPU
690 	 * context.
691 	 *
692 	 * Note that while the PKRU state lives inside the fpu registers,
693 	 * it is switched out separately at VMENTER and VMEXIT time. The
694 	 * "guest_fpu" state here contains the guest FPU context, with the
695 	 * host PRKU bits.
696 	 */
697 	struct fpu *user_fpu;
698 	struct fpu *guest_fpu;
699 
700 	u64 xcr0;
701 	u64 guest_supported_xcr0;
702 
703 	struct kvm_pio_request pio;
704 	void *pio_data;
705 	void *guest_ins_data;
706 
707 	u8 event_exit_inst_len;
708 
709 	struct kvm_queued_exception {
710 		bool pending;
711 		bool injected;
712 		bool has_error_code;
713 		u8 nr;
714 		u32 error_code;
715 		unsigned long payload;
716 		bool has_payload;
717 		u8 nested_apf;
718 	} exception;
719 
720 	struct kvm_queued_interrupt {
721 		bool injected;
722 		bool soft;
723 		u8 nr;
724 	} interrupt;
725 
726 	int halt_request; /* real mode on Intel only */
727 
728 	int cpuid_nent;
729 	struct kvm_cpuid_entry2 *cpuid_entries;
730 
731 	u64 reserved_gpa_bits;
732 	int maxphyaddr;
733 
734 	/* emulate context */
735 
736 	struct x86_emulate_ctxt *emulate_ctxt;
737 	bool emulate_regs_need_sync_to_vcpu;
738 	bool emulate_regs_need_sync_from_vcpu;
739 	int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
740 
741 	gpa_t time;
742 	struct pvclock_vcpu_time_info hv_clock;
743 	unsigned int hw_tsc_khz;
744 	struct gfn_to_hva_cache pv_time;
745 	bool pv_time_enabled;
746 	/* set guest stopped flag in pvclock flags field */
747 	bool pvclock_set_guest_stopped_request;
748 
749 	struct {
750 		u8 preempted;
751 		u64 msr_val;
752 		u64 last_steal;
753 		struct gfn_to_pfn_cache cache;
754 	} st;
755 
756 	u64 l1_tsc_offset;
757 	u64 tsc_offset; /* current tsc offset */
758 	u64 last_guest_tsc;
759 	u64 last_host_tsc;
760 	u64 tsc_offset_adjustment;
761 	u64 this_tsc_nsec;
762 	u64 this_tsc_write;
763 	u64 this_tsc_generation;
764 	bool tsc_catchup;
765 	bool tsc_always_catchup;
766 	s8 virtual_tsc_shift;
767 	u32 virtual_tsc_mult;
768 	u32 virtual_tsc_khz;
769 	s64 ia32_tsc_adjust_msr;
770 	u64 msr_ia32_power_ctl;
771 	u64 l1_tsc_scaling_ratio;
772 	u64 tsc_scaling_ratio; /* current scaling ratio */
773 
774 	atomic_t nmi_queued;  /* unprocessed asynchronous NMIs */
775 	unsigned nmi_pending; /* NMI queued after currently running handler */
776 	bool nmi_injected;    /* Trying to inject an NMI this entry */
777 	bool smi_pending;    /* SMI queued after currently running handler */
778 
779 	struct kvm_mtrr mtrr_state;
780 	u64 pat;
781 
782 	unsigned switch_db_regs;
783 	unsigned long db[KVM_NR_DB_REGS];
784 	unsigned long dr6;
785 	unsigned long dr7;
786 	unsigned long eff_db[KVM_NR_DB_REGS];
787 	unsigned long guest_debug_dr7;
788 	u64 msr_platform_info;
789 	u64 msr_misc_features_enables;
790 
791 	u64 mcg_cap;
792 	u64 mcg_status;
793 	u64 mcg_ctl;
794 	u64 mcg_ext_ctl;
795 	u64 *mce_banks;
796 
797 	/* Cache MMIO info */
798 	u64 mmio_gva;
799 	unsigned mmio_access;
800 	gfn_t mmio_gfn;
801 	u64 mmio_gen;
802 
803 	struct kvm_pmu pmu;
804 
805 	/* used for guest single stepping over the given code position */
806 	unsigned long singlestep_rip;
807 
808 	bool hyperv_enabled;
809 	struct kvm_vcpu_hv *hyperv;
810 	struct kvm_vcpu_xen xen;
811 
812 	cpumask_var_t wbinvd_dirty_mask;
813 
814 	unsigned long last_retry_eip;
815 	unsigned long last_retry_addr;
816 
817 	struct {
818 		bool halted;
819 		gfn_t gfns[ASYNC_PF_PER_VCPU];
820 		struct gfn_to_hva_cache data;
821 		u64 msr_en_val; /* MSR_KVM_ASYNC_PF_EN */
822 		u64 msr_int_val; /* MSR_KVM_ASYNC_PF_INT */
823 		u16 vec;
824 		u32 id;
825 		bool send_user_only;
826 		u32 host_apf_flags;
827 		unsigned long nested_apf_token;
828 		bool delivery_as_pf_vmexit;
829 		bool pageready_pending;
830 	} apf;
831 
832 	/* OSVW MSRs (AMD only) */
833 	struct {
834 		u64 length;
835 		u64 status;
836 	} osvw;
837 
838 	struct {
839 		u64 msr_val;
840 		struct gfn_to_hva_cache data;
841 	} pv_eoi;
842 
843 	u64 msr_kvm_poll_control;
844 
845 	/*
846 	 * Indicates the guest is trying to write a gfn that contains one or
847 	 * more of the PTEs used to translate the write itself, i.e. the access
848 	 * is changing its own translation in the guest page tables.  KVM exits
849 	 * to userspace if emulation of the faulting instruction fails and this
850 	 * flag is set, as KVM cannot make forward progress.
851 	 *
852 	 * If emulation fails for a write to guest page tables, KVM unprotects
853 	 * (zaps) the shadow page for the target gfn and resumes the guest to
854 	 * retry the non-emulatable instruction (on hardware).  Unprotecting the
855 	 * gfn doesn't allow forward progress for a self-changing access because
856 	 * doing so also zaps the translation for the gfn, i.e. retrying the
857 	 * instruction will hit a !PRESENT fault, which results in a new shadow
858 	 * page and sends KVM back to square one.
859 	 */
860 	bool write_fault_to_shadow_pgtable;
861 
862 	/* set at EPT violation at this point */
863 	unsigned long exit_qualification;
864 
865 	/* pv related host specific info */
866 	struct {
867 		bool pv_unhalted;
868 	} pv;
869 
870 	int pending_ioapic_eoi;
871 	int pending_external_vector;
872 
873 	/* be preempted when it's in kernel-mode(cpl=0) */
874 	bool preempted_in_kernel;
875 
876 	/* Flush the L1 Data cache for L1TF mitigation on VMENTER */
877 	bool l1tf_flush_l1d;
878 
879 	/* Host CPU on which VM-entry was most recently attempted */
880 	int last_vmentry_cpu;
881 
882 	/* AMD MSRC001_0015 Hardware Configuration */
883 	u64 msr_hwcr;
884 
885 	/* pv related cpuid info */
886 	struct {
887 		/*
888 		 * value of the eax register in the KVM_CPUID_FEATURES CPUID
889 		 * leaf.
890 		 */
891 		u32 features;
892 
893 		/*
894 		 * indicates whether pv emulation should be disabled if features
895 		 * are not present in the guest's cpuid
896 		 */
897 		bool enforce;
898 	} pv_cpuid;
899 
900 	/* Protected Guests */
901 	bool guest_state_protected;
902 
903 	/*
904 	 * Set when PDPTS were loaded directly by the userspace without
905 	 * reading the guest memory
906 	 */
907 	bool pdptrs_from_userspace;
908 
909 #if IS_ENABLED(CONFIG_HYPERV)
910 	hpa_t hv_root_tdp;
911 #endif
912 };
913 
914 struct kvm_lpage_info {
915 	int disallow_lpage;
916 };
917 
918 struct kvm_arch_memory_slot {
919 	struct kvm_rmap_head *rmap[KVM_NR_PAGE_SIZES];
920 	struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
921 	unsigned short *gfn_track[KVM_PAGE_TRACK_MAX];
922 };
923 
924 /*
925  * We use as the mode the number of bits allocated in the LDR for the
926  * logical processor ID.  It happens that these are all powers of two.
927  * This makes it is very easy to detect cases where the APICs are
928  * configured for multiple modes; in that case, we cannot use the map and
929  * hence cannot use kvm_irq_delivery_to_apic_fast either.
930  */
931 #define KVM_APIC_MODE_XAPIC_CLUSTER          4
932 #define KVM_APIC_MODE_XAPIC_FLAT             8
933 #define KVM_APIC_MODE_X2APIC                16
934 
935 struct kvm_apic_map {
936 	struct rcu_head rcu;
937 	u8 mode;
938 	u32 max_apic_id;
939 	union {
940 		struct kvm_lapic *xapic_flat_map[8];
941 		struct kvm_lapic *xapic_cluster_map[16][4];
942 	};
943 	struct kvm_lapic *phys_map[];
944 };
945 
946 /* Hyper-V synthetic debugger (SynDbg)*/
947 struct kvm_hv_syndbg {
948 	struct {
949 		u64 control;
950 		u64 status;
951 		u64 send_page;
952 		u64 recv_page;
953 		u64 pending_page;
954 	} control;
955 	u64 options;
956 };
957 
958 /* Current state of Hyper-V TSC page clocksource */
959 enum hv_tsc_page_status {
960 	/* TSC page was not set up or disabled */
961 	HV_TSC_PAGE_UNSET = 0,
962 	/* TSC page MSR was written by the guest, update pending */
963 	HV_TSC_PAGE_GUEST_CHANGED,
964 	/* TSC page MSR was written by KVM userspace, update pending */
965 	HV_TSC_PAGE_HOST_CHANGED,
966 	/* TSC page was properly set up and is currently active  */
967 	HV_TSC_PAGE_SET,
968 	/* TSC page is currently being updated and therefore is inactive */
969 	HV_TSC_PAGE_UPDATING,
970 	/* TSC page was set up with an inaccessible GPA */
971 	HV_TSC_PAGE_BROKEN,
972 };
973 
974 /* Hyper-V emulation context */
975 struct kvm_hv {
976 	struct mutex hv_lock;
977 	u64 hv_guest_os_id;
978 	u64 hv_hypercall;
979 	u64 hv_tsc_page;
980 	enum hv_tsc_page_status hv_tsc_page_status;
981 
982 	/* Hyper-v based guest crash (NT kernel bugcheck) parameters */
983 	u64 hv_crash_param[HV_X64_MSR_CRASH_PARAMS];
984 	u64 hv_crash_ctl;
985 
986 	struct ms_hyperv_tsc_page tsc_ref;
987 
988 	struct idr conn_to_evt;
989 
990 	u64 hv_reenlightenment_control;
991 	u64 hv_tsc_emulation_control;
992 	u64 hv_tsc_emulation_status;
993 
994 	/* How many vCPUs have VP index != vCPU index */
995 	atomic_t num_mismatched_vp_indexes;
996 
997 	/*
998 	 * How many SynICs use 'AutoEOI' feature
999 	 * (protected by arch.apicv_update_lock)
1000 	 */
1001 	unsigned int synic_auto_eoi_used;
1002 
1003 	struct hv_partition_assist_pg *hv_pa_pg;
1004 	struct kvm_hv_syndbg hv_syndbg;
1005 };
1006 
1007 struct msr_bitmap_range {
1008 	u32 flags;
1009 	u32 nmsrs;
1010 	u32 base;
1011 	unsigned long *bitmap;
1012 };
1013 
1014 /* Xen emulation context */
1015 struct kvm_xen {
1016 	bool long_mode;
1017 	u8 upcall_vector;
1018 	gfn_t shinfo_gfn;
1019 };
1020 
1021 enum kvm_irqchip_mode {
1022 	KVM_IRQCHIP_NONE,
1023 	KVM_IRQCHIP_KERNEL,       /* created with KVM_CREATE_IRQCHIP */
1024 	KVM_IRQCHIP_SPLIT,        /* created with KVM_CAP_SPLIT_IRQCHIP */
1025 };
1026 
1027 struct kvm_x86_msr_filter {
1028 	u8 count;
1029 	bool default_allow:1;
1030 	struct msr_bitmap_range ranges[16];
1031 };
1032 
1033 #define APICV_INHIBIT_REASON_DISABLE    0
1034 #define APICV_INHIBIT_REASON_HYPERV     1
1035 #define APICV_INHIBIT_REASON_NESTED     2
1036 #define APICV_INHIBIT_REASON_IRQWIN     3
1037 #define APICV_INHIBIT_REASON_PIT_REINJ  4
1038 #define APICV_INHIBIT_REASON_X2APIC	5
1039 
1040 struct kvm_arch {
1041 	unsigned long n_used_mmu_pages;
1042 	unsigned long n_requested_mmu_pages;
1043 	unsigned long n_max_mmu_pages;
1044 	unsigned int indirect_shadow_pages;
1045 	u8 mmu_valid_gen;
1046 	struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
1047 	struct list_head active_mmu_pages;
1048 	struct list_head zapped_obsolete_pages;
1049 	struct list_head lpage_disallowed_mmu_pages;
1050 	struct kvm_page_track_notifier_node mmu_sp_tracker;
1051 	struct kvm_page_track_notifier_head track_notifier_head;
1052 	/*
1053 	 * Protects marking pages unsync during page faults, as TDP MMU page
1054 	 * faults only take mmu_lock for read.  For simplicity, the unsync
1055 	 * pages lock is always taken when marking pages unsync regardless of
1056 	 * whether mmu_lock is held for read or write.
1057 	 */
1058 	spinlock_t mmu_unsync_pages_lock;
1059 
1060 	struct list_head assigned_dev_head;
1061 	struct iommu_domain *iommu_domain;
1062 	bool iommu_noncoherent;
1063 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA
1064 	atomic_t noncoherent_dma_count;
1065 #define __KVM_HAVE_ARCH_ASSIGNED_DEVICE
1066 	atomic_t assigned_device_count;
1067 	struct kvm_pic *vpic;
1068 	struct kvm_ioapic *vioapic;
1069 	struct kvm_pit *vpit;
1070 	atomic_t vapics_in_nmi_mode;
1071 	struct mutex apic_map_lock;
1072 	struct kvm_apic_map __rcu *apic_map;
1073 	atomic_t apic_map_dirty;
1074 
1075 	/* Protects apic_access_memslot_enabled and apicv_inhibit_reasons */
1076 	struct mutex apicv_update_lock;
1077 
1078 	bool apic_access_memslot_enabled;
1079 	unsigned long apicv_inhibit_reasons;
1080 
1081 	gpa_t wall_clock;
1082 
1083 	bool mwait_in_guest;
1084 	bool hlt_in_guest;
1085 	bool pause_in_guest;
1086 	bool cstate_in_guest;
1087 
1088 	unsigned long irq_sources_bitmap;
1089 	s64 kvmclock_offset;
1090 	raw_spinlock_t tsc_write_lock;
1091 	u64 last_tsc_nsec;
1092 	u64 last_tsc_write;
1093 	u32 last_tsc_khz;
1094 	u64 cur_tsc_nsec;
1095 	u64 cur_tsc_write;
1096 	u64 cur_tsc_offset;
1097 	u64 cur_tsc_generation;
1098 	int nr_vcpus_matched_tsc;
1099 
1100 	spinlock_t pvclock_gtod_sync_lock;
1101 	bool use_master_clock;
1102 	u64 master_kernel_ns;
1103 	u64 master_cycle_now;
1104 	struct delayed_work kvmclock_update_work;
1105 	struct delayed_work kvmclock_sync_work;
1106 
1107 	struct kvm_xen_hvm_config xen_hvm_config;
1108 
1109 	/* reads protected by irq_srcu, writes by irq_lock */
1110 	struct hlist_head mask_notifier_list;
1111 
1112 	struct kvm_hv hyperv;
1113 	struct kvm_xen xen;
1114 
1115 	#ifdef CONFIG_KVM_MMU_AUDIT
1116 	int audit_point;
1117 	#endif
1118 
1119 	bool backwards_tsc_observed;
1120 	bool boot_vcpu_runs_old_kvmclock;
1121 	u32 bsp_vcpu_id;
1122 
1123 	u64 disabled_quirks;
1124 	int cpu_dirty_logging_count;
1125 
1126 	enum kvm_irqchip_mode irqchip_mode;
1127 	u8 nr_reserved_ioapic_pins;
1128 
1129 	bool disabled_lapic_found;
1130 
1131 	bool x2apic_format;
1132 	bool x2apic_broadcast_quirk_disabled;
1133 
1134 	bool guest_can_read_msr_platform_info;
1135 	bool exception_payload_enabled;
1136 
1137 	bool bus_lock_detection_enabled;
1138 	/*
1139 	 * If exit_on_emulation_error is set, and the in-kernel instruction
1140 	 * emulator fails to emulate an instruction, allow userspace
1141 	 * the opportunity to look at it.
1142 	 */
1143 	bool exit_on_emulation_error;
1144 
1145 	/* Deflect RDMSR and WRMSR to user space when they trigger a #GP */
1146 	u32 user_space_msr_mask;
1147 	struct kvm_x86_msr_filter __rcu *msr_filter;
1148 
1149 	u32 hypercall_exit_enabled;
1150 
1151 	/* Guest can access the SGX PROVISIONKEY. */
1152 	bool sgx_provisioning_allowed;
1153 
1154 	struct kvm_pmu_event_filter __rcu *pmu_event_filter;
1155 	struct task_struct *nx_lpage_recovery_thread;
1156 
1157 #ifdef CONFIG_X86_64
1158 	/*
1159 	 * Whether the TDP MMU is enabled for this VM. This contains a
1160 	 * snapshot of the TDP MMU module parameter from when the VM was
1161 	 * created and remains unchanged for the life of the VM. If this is
1162 	 * true, TDP MMU handler functions will run for various MMU
1163 	 * operations.
1164 	 */
1165 	bool tdp_mmu_enabled;
1166 
1167 	/*
1168 	 * List of struct kvm_mmu_pages being used as roots.
1169 	 * All struct kvm_mmu_pages in the list should have
1170 	 * tdp_mmu_page set.
1171 	 *
1172 	 * For reads, this list is protected by:
1173 	 *	the MMU lock in read mode + RCU or
1174 	 *	the MMU lock in write mode
1175 	 *
1176 	 * For writes, this list is protected by:
1177 	 *	the MMU lock in read mode + the tdp_mmu_pages_lock or
1178 	 *	the MMU lock in write mode
1179 	 *
1180 	 * Roots will remain in the list until their tdp_mmu_root_count
1181 	 * drops to zero, at which point the thread that decremented the
1182 	 * count to zero should removed the root from the list and clean
1183 	 * it up, freeing the root after an RCU grace period.
1184 	 */
1185 	struct list_head tdp_mmu_roots;
1186 
1187 	/*
1188 	 * List of struct kvmp_mmu_pages not being used as roots.
1189 	 * All struct kvm_mmu_pages in the list should have
1190 	 * tdp_mmu_page set and a tdp_mmu_root_count of 0.
1191 	 */
1192 	struct list_head tdp_mmu_pages;
1193 
1194 	/*
1195 	 * Protects accesses to the following fields when the MMU lock
1196 	 * is held in read mode:
1197 	 *  - tdp_mmu_roots (above)
1198 	 *  - tdp_mmu_pages (above)
1199 	 *  - the link field of struct kvm_mmu_pages used by the TDP MMU
1200 	 *  - lpage_disallowed_mmu_pages
1201 	 *  - the lpage_disallowed_link field of struct kvm_mmu_pages used
1202 	 *    by the TDP MMU
1203 	 * It is acceptable, but not necessary, to acquire this lock when
1204 	 * the thread holds the MMU lock in write mode.
1205 	 */
1206 	spinlock_t tdp_mmu_pages_lock;
1207 #endif /* CONFIG_X86_64 */
1208 
1209 	/*
1210 	 * If set, rmaps have been allocated for all memslots and should be
1211 	 * allocated for any newly created or modified memslots.
1212 	 */
1213 	bool memslots_have_rmaps;
1214 
1215 #if IS_ENABLED(CONFIG_HYPERV)
1216 	hpa_t	hv_root_tdp;
1217 	spinlock_t hv_root_tdp_lock;
1218 #endif
1219 };
1220 
1221 struct kvm_vm_stat {
1222 	struct kvm_vm_stat_generic generic;
1223 	u64 mmu_shadow_zapped;
1224 	u64 mmu_pte_write;
1225 	u64 mmu_pde_zapped;
1226 	u64 mmu_flooded;
1227 	u64 mmu_recycled;
1228 	u64 mmu_cache_miss;
1229 	u64 mmu_unsync;
1230 	union {
1231 		struct {
1232 			atomic64_t pages_4k;
1233 			atomic64_t pages_2m;
1234 			atomic64_t pages_1g;
1235 		};
1236 		atomic64_t pages[KVM_NR_PAGE_SIZES];
1237 	};
1238 	u64 nx_lpage_splits;
1239 	u64 max_mmu_page_hash_collisions;
1240 	u64 max_mmu_rmap_size;
1241 };
1242 
1243 struct kvm_vcpu_stat {
1244 	struct kvm_vcpu_stat_generic generic;
1245 	u64 pf_fixed;
1246 	u64 pf_guest;
1247 	u64 tlb_flush;
1248 	u64 invlpg;
1249 
1250 	u64 exits;
1251 	u64 io_exits;
1252 	u64 mmio_exits;
1253 	u64 signal_exits;
1254 	u64 irq_window_exits;
1255 	u64 nmi_window_exits;
1256 	u64 l1d_flush;
1257 	u64 halt_exits;
1258 	u64 request_irq_exits;
1259 	u64 irq_exits;
1260 	u64 host_state_reload;
1261 	u64 fpu_reload;
1262 	u64 insn_emulation;
1263 	u64 insn_emulation_fail;
1264 	u64 hypercalls;
1265 	u64 irq_injections;
1266 	u64 nmi_injections;
1267 	u64 req_event;
1268 	u64 nested_run;
1269 	u64 directed_yield_attempted;
1270 	u64 directed_yield_successful;
1271 	u64 guest_mode;
1272 };
1273 
1274 struct x86_instruction_info;
1275 
1276 struct msr_data {
1277 	bool host_initiated;
1278 	u32 index;
1279 	u64 data;
1280 };
1281 
1282 struct kvm_lapic_irq {
1283 	u32 vector;
1284 	u16 delivery_mode;
1285 	u16 dest_mode;
1286 	bool level;
1287 	u16 trig_mode;
1288 	u32 shorthand;
1289 	u32 dest_id;
1290 	bool msi_redir_hint;
1291 };
1292 
1293 static inline u16 kvm_lapic_irq_dest_mode(bool dest_mode_logical)
1294 {
1295 	return dest_mode_logical ? APIC_DEST_LOGICAL : APIC_DEST_PHYSICAL;
1296 }
1297 
1298 struct kvm_x86_ops {
1299 	int (*hardware_enable)(void);
1300 	void (*hardware_disable)(void);
1301 	void (*hardware_unsetup)(void);
1302 	bool (*cpu_has_accelerated_tpr)(void);
1303 	bool (*has_emulated_msr)(struct kvm *kvm, u32 index);
1304 	void (*vcpu_after_set_cpuid)(struct kvm_vcpu *vcpu);
1305 
1306 	unsigned int vm_size;
1307 	int (*vm_init)(struct kvm *kvm);
1308 	void (*vm_destroy)(struct kvm *kvm);
1309 
1310 	/* Create, but do not attach this VCPU */
1311 	int (*vcpu_create)(struct kvm_vcpu *vcpu);
1312 	void (*vcpu_free)(struct kvm_vcpu *vcpu);
1313 	void (*vcpu_reset)(struct kvm_vcpu *vcpu, bool init_event);
1314 
1315 	void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
1316 	void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
1317 	void (*vcpu_put)(struct kvm_vcpu *vcpu);
1318 
1319 	void (*update_exception_bitmap)(struct kvm_vcpu *vcpu);
1320 	int (*get_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1321 	int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1322 	u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
1323 	void (*get_segment)(struct kvm_vcpu *vcpu,
1324 			    struct kvm_segment *var, int seg);
1325 	int (*get_cpl)(struct kvm_vcpu *vcpu);
1326 	void (*set_segment)(struct kvm_vcpu *vcpu,
1327 			    struct kvm_segment *var, int seg);
1328 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
1329 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
1330 	bool (*is_valid_cr4)(struct kvm_vcpu *vcpu, unsigned long cr0);
1331 	void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
1332 	int (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
1333 	void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1334 	void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1335 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1336 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1337 	void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu);
1338 	void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
1339 	void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
1340 	unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
1341 	void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
1342 
1343 	void (*tlb_flush_all)(struct kvm_vcpu *vcpu);
1344 	void (*tlb_flush_current)(struct kvm_vcpu *vcpu);
1345 	int  (*tlb_remote_flush)(struct kvm *kvm);
1346 	int  (*tlb_remote_flush_with_range)(struct kvm *kvm,
1347 			struct kvm_tlb_range *range);
1348 
1349 	/*
1350 	 * Flush any TLB entries associated with the given GVA.
1351 	 * Does not need to flush GPA->HPA mappings.
1352 	 * Can potentially get non-canonical addresses through INVLPGs, which
1353 	 * the implementation may choose to ignore if appropriate.
1354 	 */
1355 	void (*tlb_flush_gva)(struct kvm_vcpu *vcpu, gva_t addr);
1356 
1357 	/*
1358 	 * Flush any TLB entries created by the guest.  Like tlb_flush_gva(),
1359 	 * does not need to flush GPA->HPA mappings.
1360 	 */
1361 	void (*tlb_flush_guest)(struct kvm_vcpu *vcpu);
1362 
1363 	enum exit_fastpath_completion (*run)(struct kvm_vcpu *vcpu);
1364 	int (*handle_exit)(struct kvm_vcpu *vcpu,
1365 		enum exit_fastpath_completion exit_fastpath);
1366 	int (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
1367 	void (*update_emulated_instruction)(struct kvm_vcpu *vcpu);
1368 	void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
1369 	u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu);
1370 	void (*patch_hypercall)(struct kvm_vcpu *vcpu,
1371 				unsigned char *hypercall_addr);
1372 	void (*set_irq)(struct kvm_vcpu *vcpu);
1373 	void (*set_nmi)(struct kvm_vcpu *vcpu);
1374 	void (*queue_exception)(struct kvm_vcpu *vcpu);
1375 	void (*cancel_injection)(struct kvm_vcpu *vcpu);
1376 	int (*interrupt_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1377 	int (*nmi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1378 	bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
1379 	void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
1380 	void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
1381 	void (*enable_irq_window)(struct kvm_vcpu *vcpu);
1382 	void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
1383 	bool (*check_apicv_inhibit_reasons)(ulong bit);
1384 	void (*refresh_apicv_exec_ctrl)(struct kvm_vcpu *vcpu);
1385 	void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
1386 	void (*hwapic_isr_update)(struct kvm_vcpu *vcpu, int isr);
1387 	bool (*guest_apic_has_interrupt)(struct kvm_vcpu *vcpu);
1388 	void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
1389 	void (*set_virtual_apic_mode)(struct kvm_vcpu *vcpu);
1390 	void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu);
1391 	int (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
1392 	int (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
1393 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
1394 	int (*set_identity_map_addr)(struct kvm *kvm, u64 ident_addr);
1395 	u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
1396 
1397 	void (*load_mmu_pgd)(struct kvm_vcpu *vcpu, hpa_t root_hpa,
1398 			     int root_level);
1399 
1400 	bool (*has_wbinvd_exit)(void);
1401 
1402 	u64 (*get_l2_tsc_offset)(struct kvm_vcpu *vcpu);
1403 	u64 (*get_l2_tsc_multiplier)(struct kvm_vcpu *vcpu);
1404 	void (*write_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
1405 	void (*write_tsc_multiplier)(struct kvm_vcpu *vcpu, u64 multiplier);
1406 
1407 	/*
1408 	 * Retrieve somewhat arbitrary exit information.  Intended to be used
1409 	 * only from within tracepoints to avoid VMREADs when tracing is off.
1410 	 */
1411 	void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2,
1412 			      u32 *exit_int_info, u32 *exit_int_info_err_code);
1413 
1414 	int (*check_intercept)(struct kvm_vcpu *vcpu,
1415 			       struct x86_instruction_info *info,
1416 			       enum x86_intercept_stage stage,
1417 			       struct x86_exception *exception);
1418 	void (*handle_exit_irqoff)(struct kvm_vcpu *vcpu);
1419 
1420 	void (*request_immediate_exit)(struct kvm_vcpu *vcpu);
1421 
1422 	void (*sched_in)(struct kvm_vcpu *kvm, int cpu);
1423 
1424 	/*
1425 	 * Size of the CPU's dirty log buffer, i.e. VMX's PML buffer.  A zero
1426 	 * value indicates CPU dirty logging is unsupported or disabled.
1427 	 */
1428 	int cpu_dirty_log_size;
1429 	void (*update_cpu_dirty_logging)(struct kvm_vcpu *vcpu);
1430 
1431 	/* pmu operations of sub-arch */
1432 	const struct kvm_pmu_ops *pmu_ops;
1433 	const struct kvm_x86_nested_ops *nested_ops;
1434 
1435 	/*
1436 	 * Architecture specific hooks for vCPU blocking due to
1437 	 * HLT instruction.
1438 	 * Returns for .pre_block():
1439 	 *    - 0 means continue to block the vCPU.
1440 	 *    - 1 means we cannot block the vCPU since some event
1441 	 *        happens during this period, such as, 'ON' bit in
1442 	 *        posted-interrupts descriptor is set.
1443 	 */
1444 	int (*pre_block)(struct kvm_vcpu *vcpu);
1445 	void (*post_block)(struct kvm_vcpu *vcpu);
1446 
1447 	void (*vcpu_blocking)(struct kvm_vcpu *vcpu);
1448 	void (*vcpu_unblocking)(struct kvm_vcpu *vcpu);
1449 
1450 	int (*update_pi_irte)(struct kvm *kvm, unsigned int host_irq,
1451 			      uint32_t guest_irq, bool set);
1452 	void (*start_assignment)(struct kvm *kvm);
1453 	void (*apicv_post_state_restore)(struct kvm_vcpu *vcpu);
1454 	bool (*dy_apicv_has_pending_interrupt)(struct kvm_vcpu *vcpu);
1455 
1456 	int (*set_hv_timer)(struct kvm_vcpu *vcpu, u64 guest_deadline_tsc,
1457 			    bool *expired);
1458 	void (*cancel_hv_timer)(struct kvm_vcpu *vcpu);
1459 
1460 	void (*setup_mce)(struct kvm_vcpu *vcpu);
1461 
1462 	int (*smi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1463 	int (*enter_smm)(struct kvm_vcpu *vcpu, char *smstate);
1464 	int (*leave_smm)(struct kvm_vcpu *vcpu, const char *smstate);
1465 	void (*enable_smi_window)(struct kvm_vcpu *vcpu);
1466 
1467 	int (*mem_enc_op)(struct kvm *kvm, void __user *argp);
1468 	int (*mem_enc_reg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1469 	int (*mem_enc_unreg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1470 	int (*vm_copy_enc_context_from)(struct kvm *kvm, unsigned int source_fd);
1471 
1472 	int (*get_msr_feature)(struct kvm_msr_entry *entry);
1473 
1474 	bool (*can_emulate_instruction)(struct kvm_vcpu *vcpu, void *insn, int insn_len);
1475 
1476 	bool (*apic_init_signal_blocked)(struct kvm_vcpu *vcpu);
1477 	int (*enable_direct_tlbflush)(struct kvm_vcpu *vcpu);
1478 
1479 	void (*migrate_timers)(struct kvm_vcpu *vcpu);
1480 	void (*msr_filter_changed)(struct kvm_vcpu *vcpu);
1481 	int (*complete_emulated_msr)(struct kvm_vcpu *vcpu, int err);
1482 
1483 	void (*vcpu_deliver_sipi_vector)(struct kvm_vcpu *vcpu, u8 vector);
1484 };
1485 
1486 struct kvm_x86_nested_ops {
1487 	int (*check_events)(struct kvm_vcpu *vcpu);
1488 	bool (*hv_timer_pending)(struct kvm_vcpu *vcpu);
1489 	void (*triple_fault)(struct kvm_vcpu *vcpu);
1490 	int (*get_state)(struct kvm_vcpu *vcpu,
1491 			 struct kvm_nested_state __user *user_kvm_nested_state,
1492 			 unsigned user_data_size);
1493 	int (*set_state)(struct kvm_vcpu *vcpu,
1494 			 struct kvm_nested_state __user *user_kvm_nested_state,
1495 			 struct kvm_nested_state *kvm_state);
1496 	bool (*get_nested_state_pages)(struct kvm_vcpu *vcpu);
1497 	int (*write_log_dirty)(struct kvm_vcpu *vcpu, gpa_t l2_gpa);
1498 
1499 	int (*enable_evmcs)(struct kvm_vcpu *vcpu,
1500 			    uint16_t *vmcs_version);
1501 	uint16_t (*get_evmcs_version)(struct kvm_vcpu *vcpu);
1502 };
1503 
1504 struct kvm_x86_init_ops {
1505 	int (*cpu_has_kvm_support)(void);
1506 	int (*disabled_by_bios)(void);
1507 	int (*check_processor_compatibility)(void);
1508 	int (*hardware_setup)(void);
1509 
1510 	struct kvm_x86_ops *runtime_ops;
1511 };
1512 
1513 struct kvm_arch_async_pf {
1514 	u32 token;
1515 	gfn_t gfn;
1516 	unsigned long cr3;
1517 	bool direct_map;
1518 };
1519 
1520 extern u32 __read_mostly kvm_nr_uret_msrs;
1521 extern u64 __read_mostly host_efer;
1522 extern bool __read_mostly allow_smaller_maxphyaddr;
1523 extern bool __read_mostly enable_apicv;
1524 extern struct kvm_x86_ops kvm_x86_ops;
1525 
1526 #define KVM_X86_OP(func) \
1527 	DECLARE_STATIC_CALL(kvm_x86_##func, *(((struct kvm_x86_ops *)0)->func));
1528 #define KVM_X86_OP_NULL KVM_X86_OP
1529 #include <asm/kvm-x86-ops.h>
1530 
1531 static inline void kvm_ops_static_call_update(void)
1532 {
1533 #define KVM_X86_OP(func) \
1534 	static_call_update(kvm_x86_##func, kvm_x86_ops.func);
1535 #define KVM_X86_OP_NULL KVM_X86_OP
1536 #include <asm/kvm-x86-ops.h>
1537 }
1538 
1539 #define __KVM_HAVE_ARCH_VM_ALLOC
1540 static inline struct kvm *kvm_arch_alloc_vm(void)
1541 {
1542 	return __vmalloc(kvm_x86_ops.vm_size, GFP_KERNEL_ACCOUNT | __GFP_ZERO);
1543 }
1544 void kvm_arch_free_vm(struct kvm *kvm);
1545 
1546 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
1547 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
1548 {
1549 	if (kvm_x86_ops.tlb_remote_flush &&
1550 	    !static_call(kvm_x86_tlb_remote_flush)(kvm))
1551 		return 0;
1552 	else
1553 		return -ENOTSUPP;
1554 }
1555 
1556 int kvm_mmu_module_init(void);
1557 void kvm_mmu_module_exit(void);
1558 
1559 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
1560 int kvm_mmu_create(struct kvm_vcpu *vcpu);
1561 void kvm_mmu_init_vm(struct kvm *kvm);
1562 void kvm_mmu_uninit_vm(struct kvm *kvm);
1563 
1564 void kvm_mmu_after_set_cpuid(struct kvm_vcpu *vcpu);
1565 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
1566 void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
1567 				      const struct kvm_memory_slot *memslot,
1568 				      int start_level);
1569 void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
1570 				   const struct kvm_memory_slot *memslot);
1571 void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
1572 				   const struct kvm_memory_slot *memslot);
1573 void kvm_mmu_zap_all(struct kvm *kvm);
1574 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, u64 gen);
1575 unsigned long kvm_mmu_calculate_default_mmu_pages(struct kvm *kvm);
1576 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned long kvm_nr_mmu_pages);
1577 
1578 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
1579 
1580 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
1581 			  const void *val, int bytes);
1582 
1583 struct kvm_irq_mask_notifier {
1584 	void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
1585 	int irq;
1586 	struct hlist_node link;
1587 };
1588 
1589 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
1590 				    struct kvm_irq_mask_notifier *kimn);
1591 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
1592 				      struct kvm_irq_mask_notifier *kimn);
1593 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
1594 			     bool mask);
1595 
1596 extern bool tdp_enabled;
1597 
1598 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
1599 
1600 /* control of guest tsc rate supported? */
1601 extern bool kvm_has_tsc_control;
1602 /* maximum supported tsc_khz for guests */
1603 extern u32  kvm_max_guest_tsc_khz;
1604 /* number of bits of the fractional part of the TSC scaling ratio */
1605 extern u8   kvm_tsc_scaling_ratio_frac_bits;
1606 /* maximum allowed value of TSC scaling ratio */
1607 extern u64  kvm_max_tsc_scaling_ratio;
1608 /* 1ull << kvm_tsc_scaling_ratio_frac_bits */
1609 extern u64  kvm_default_tsc_scaling_ratio;
1610 /* bus lock detection supported? */
1611 extern bool kvm_has_bus_lock_exit;
1612 
1613 extern u64 kvm_mce_cap_supported;
1614 
1615 /*
1616  * EMULTYPE_NO_DECODE - Set when re-emulating an instruction (after completing
1617  *			userspace I/O) to indicate that the emulation context
1618  *			should be reused as is, i.e. skip initialization of
1619  *			emulation context, instruction fetch and decode.
1620  *
1621  * EMULTYPE_TRAP_UD - Set when emulating an intercepted #UD from hardware.
1622  *		      Indicates that only select instructions (tagged with
1623  *		      EmulateOnUD) should be emulated (to minimize the emulator
1624  *		      attack surface).  See also EMULTYPE_TRAP_UD_FORCED.
1625  *
1626  * EMULTYPE_SKIP - Set when emulating solely to skip an instruction, i.e. to
1627  *		   decode the instruction length.  For use *only* by
1628  *		   kvm_x86_ops.skip_emulated_instruction() implementations.
1629  *
1630  * EMULTYPE_ALLOW_RETRY_PF - Set when the emulator should resume the guest to
1631  *			     retry native execution under certain conditions,
1632  *			     Can only be set in conjunction with EMULTYPE_PF.
1633  *
1634  * EMULTYPE_TRAP_UD_FORCED - Set when emulating an intercepted #UD that was
1635  *			     triggered by KVM's magic "force emulation" prefix,
1636  *			     which is opt in via module param (off by default).
1637  *			     Bypasses EmulateOnUD restriction despite emulating
1638  *			     due to an intercepted #UD (see EMULTYPE_TRAP_UD).
1639  *			     Used to test the full emulator from userspace.
1640  *
1641  * EMULTYPE_VMWARE_GP - Set when emulating an intercepted #GP for VMware
1642  *			backdoor emulation, which is opt in via module param.
1643  *			VMware backdoor emulation handles select instructions
1644  *			and reinjects the #GP for all other cases.
1645  *
1646  * EMULTYPE_PF - Set when emulating MMIO by way of an intercepted #PF, in which
1647  *		 case the CR2/GPA value pass on the stack is valid.
1648  */
1649 #define EMULTYPE_NO_DECODE	    (1 << 0)
1650 #define EMULTYPE_TRAP_UD	    (1 << 1)
1651 #define EMULTYPE_SKIP		    (1 << 2)
1652 #define EMULTYPE_ALLOW_RETRY_PF	    (1 << 3)
1653 #define EMULTYPE_TRAP_UD_FORCED	    (1 << 4)
1654 #define EMULTYPE_VMWARE_GP	    (1 << 5)
1655 #define EMULTYPE_PF		    (1 << 6)
1656 
1657 int kvm_emulate_instruction(struct kvm_vcpu *vcpu, int emulation_type);
1658 int kvm_emulate_instruction_from_buffer(struct kvm_vcpu *vcpu,
1659 					void *insn, int insn_len);
1660 
1661 void kvm_enable_efer_bits(u64);
1662 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
1663 int __kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data, bool host_initiated);
1664 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data);
1665 int kvm_set_msr(struct kvm_vcpu *vcpu, u32 index, u64 data);
1666 int kvm_emulate_rdmsr(struct kvm_vcpu *vcpu);
1667 int kvm_emulate_wrmsr(struct kvm_vcpu *vcpu);
1668 int kvm_emulate_as_nop(struct kvm_vcpu *vcpu);
1669 int kvm_emulate_invd(struct kvm_vcpu *vcpu);
1670 int kvm_emulate_mwait(struct kvm_vcpu *vcpu);
1671 int kvm_handle_invalid_op(struct kvm_vcpu *vcpu);
1672 int kvm_emulate_monitor(struct kvm_vcpu *vcpu);
1673 
1674 int kvm_fast_pio(struct kvm_vcpu *vcpu, int size, unsigned short port, int in);
1675 int kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
1676 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
1677 int kvm_vcpu_halt(struct kvm_vcpu *vcpu);
1678 int kvm_emulate_ap_reset_hold(struct kvm_vcpu *vcpu);
1679 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
1680 
1681 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
1682 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
1683 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector);
1684 
1685 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
1686 		    int reason, bool has_error_code, u32 error_code);
1687 
1688 void kvm_free_guest_fpu(struct kvm_vcpu *vcpu);
1689 
1690 void kvm_post_set_cr0(struct kvm_vcpu *vcpu, unsigned long old_cr0, unsigned long cr0);
1691 void kvm_post_set_cr4(struct kvm_vcpu *vcpu, unsigned long old_cr4, unsigned long cr4);
1692 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
1693 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
1694 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
1695 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
1696 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
1697 void kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
1698 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
1699 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
1700 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
1701 int kvm_emulate_xsetbv(struct kvm_vcpu *vcpu);
1702 
1703 int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1704 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1705 
1706 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
1707 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
1708 int kvm_emulate_rdpmc(struct kvm_vcpu *vcpu);
1709 
1710 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1711 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1712 void kvm_queue_exception_p(struct kvm_vcpu *vcpu, unsigned nr, unsigned long payload);
1713 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1714 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1715 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
1716 bool kvm_inject_emulated_page_fault(struct kvm_vcpu *vcpu,
1717 				    struct x86_exception *fault);
1718 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1719 			    gfn_t gfn, void *data, int offset, int len,
1720 			    u32 access);
1721 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
1722 bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr);
1723 
1724 static inline int __kvm_irq_line_state(unsigned long *irq_state,
1725 				       int irq_source_id, int level)
1726 {
1727 	/* Logical OR for level trig interrupt */
1728 	if (level)
1729 		__set_bit(irq_source_id, irq_state);
1730 	else
1731 		__clear_bit(irq_source_id, irq_state);
1732 
1733 	return !!(*irq_state);
1734 }
1735 
1736 #define KVM_MMU_ROOT_CURRENT		BIT(0)
1737 #define KVM_MMU_ROOT_PREVIOUS(i)	BIT(1+i)
1738 #define KVM_MMU_ROOTS_ALL		(~0UL)
1739 
1740 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
1741 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
1742 
1743 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
1744 
1745 void kvm_update_dr7(struct kvm_vcpu *vcpu);
1746 
1747 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
1748 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
1749 void kvm_mmu_free_roots(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1750 			ulong roots_to_free);
1751 void kvm_mmu_free_guest_mode_roots(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu);
1752 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1753 			   struct x86_exception *exception);
1754 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
1755 			      struct x86_exception *exception);
1756 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
1757 			       struct x86_exception *exception);
1758 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
1759 			       struct x86_exception *exception);
1760 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
1761 				struct x86_exception *exception);
1762 
1763 bool kvm_apicv_activated(struct kvm *kvm);
1764 void kvm_vcpu_update_apicv(struct kvm_vcpu *vcpu);
1765 void kvm_request_apicv_update(struct kvm *kvm, bool activate,
1766 			      unsigned long bit);
1767 
1768 void __kvm_request_apicv_update(struct kvm *kvm, bool activate,
1769 				unsigned long bit);
1770 
1771 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
1772 
1773 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u64 error_code,
1774 		       void *insn, int insn_len);
1775 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
1776 void kvm_mmu_invalidate_gva(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1777 			    gva_t gva, hpa_t root_hpa);
1778 void kvm_mmu_invpcid_gva(struct kvm_vcpu *vcpu, gva_t gva, unsigned long pcid);
1779 void kvm_mmu_new_pgd(struct kvm_vcpu *vcpu, gpa_t new_pgd);
1780 
1781 void kvm_configure_mmu(bool enable_tdp, int tdp_forced_root_level,
1782 		       int tdp_max_root_level, int tdp_huge_page_level);
1783 
1784 static inline u16 kvm_read_ldt(void)
1785 {
1786 	u16 ldt;
1787 	asm("sldt %0" : "=g"(ldt));
1788 	return ldt;
1789 }
1790 
1791 static inline void kvm_load_ldt(u16 sel)
1792 {
1793 	asm("lldt %0" : : "rm"(sel));
1794 }
1795 
1796 #ifdef CONFIG_X86_64
1797 static inline unsigned long read_msr(unsigned long msr)
1798 {
1799 	u64 value;
1800 
1801 	rdmsrl(msr, value);
1802 	return value;
1803 }
1804 #endif
1805 
1806 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
1807 {
1808 	kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
1809 }
1810 
1811 #define TSS_IOPB_BASE_OFFSET 0x66
1812 #define TSS_BASE_SIZE 0x68
1813 #define TSS_IOPB_SIZE (65536 / 8)
1814 #define TSS_REDIRECTION_SIZE (256 / 8)
1815 #define RMODE_TSS_SIZE							\
1816 	(TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
1817 
1818 enum {
1819 	TASK_SWITCH_CALL = 0,
1820 	TASK_SWITCH_IRET = 1,
1821 	TASK_SWITCH_JMP = 2,
1822 	TASK_SWITCH_GATE = 3,
1823 };
1824 
1825 #define HF_GIF_MASK		(1 << 0)
1826 #define HF_NMI_MASK		(1 << 3)
1827 #define HF_IRET_MASK		(1 << 4)
1828 #define HF_GUEST_MASK		(1 << 5) /* VCPU is in guest-mode */
1829 #define HF_SMM_MASK		(1 << 6)
1830 #define HF_SMM_INSIDE_NMI_MASK	(1 << 7)
1831 
1832 #define __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
1833 #define KVM_ADDRESS_SPACE_NUM 2
1834 
1835 #define kvm_arch_vcpu_memslots_id(vcpu) ((vcpu)->arch.hflags & HF_SMM_MASK ? 1 : 0)
1836 #define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, (role).smm)
1837 
1838 #define KVM_ARCH_WANT_MMU_NOTIFIER
1839 
1840 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1841 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1842 int kvm_cpu_has_extint(struct kvm_vcpu *v);
1843 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1844 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1845 void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event);
1846 void kvm_vcpu_reload_apic_access_page(struct kvm_vcpu *vcpu);
1847 
1848 int kvm_pv_send_ipi(struct kvm *kvm, unsigned long ipi_bitmap_low,
1849 		    unsigned long ipi_bitmap_high, u32 min,
1850 		    unsigned long icr, int op_64_bit);
1851 
1852 int kvm_add_user_return_msr(u32 msr);
1853 int kvm_find_user_return_msr(u32 msr);
1854 int kvm_set_user_return_msr(unsigned index, u64 val, u64 mask);
1855 
1856 static inline bool kvm_is_supported_user_return_msr(u32 msr)
1857 {
1858 	return kvm_find_user_return_msr(msr) >= 0;
1859 }
1860 
1861 u64 kvm_scale_tsc(struct kvm_vcpu *vcpu, u64 tsc, u64 ratio);
1862 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc);
1863 u64 kvm_calc_nested_tsc_offset(u64 l1_offset, u64 l2_offset, u64 l2_multiplier);
1864 u64 kvm_calc_nested_tsc_multiplier(u64 l1_multiplier, u64 l2_multiplier);
1865 
1866 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu);
1867 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1868 
1869 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
1870 void kvm_make_scan_ioapic_request(struct kvm *kvm);
1871 void kvm_make_scan_ioapic_request_mask(struct kvm *kvm,
1872 				       unsigned long *vcpu_bitmap);
1873 
1874 bool kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1875 				     struct kvm_async_pf *work);
1876 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1877 				 struct kvm_async_pf *work);
1878 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1879 			       struct kvm_async_pf *work);
1880 void kvm_arch_async_page_present_queued(struct kvm_vcpu *vcpu);
1881 bool kvm_arch_can_dequeue_async_page_present(struct kvm_vcpu *vcpu);
1882 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1883 
1884 int kvm_skip_emulated_instruction(struct kvm_vcpu *vcpu);
1885 int kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1886 void __kvm_request_immediate_exit(struct kvm_vcpu *vcpu);
1887 
1888 int kvm_is_in_guest(void);
1889 
1890 void __user *__x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa,
1891 				     u32 size);
1892 bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu);
1893 bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu);
1894 
1895 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
1896 			     struct kvm_vcpu **dest_vcpu);
1897 
1898 void kvm_set_msi_irq(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e,
1899 		     struct kvm_lapic_irq *irq);
1900 
1901 static inline bool kvm_irq_is_postable(struct kvm_lapic_irq *irq)
1902 {
1903 	/* We can only post Fixed and LowPrio IRQs */
1904 	return (irq->delivery_mode == APIC_DM_FIXED ||
1905 		irq->delivery_mode == APIC_DM_LOWEST);
1906 }
1907 
1908 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu)
1909 {
1910 	static_call_cond(kvm_x86_vcpu_blocking)(vcpu);
1911 }
1912 
1913 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu)
1914 {
1915 	static_call_cond(kvm_x86_vcpu_unblocking)(vcpu);
1916 }
1917 
1918 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
1919 
1920 static inline int kvm_cpu_get_apicid(int mps_cpu)
1921 {
1922 #ifdef CONFIG_X86_LOCAL_APIC
1923 	return default_cpu_present_to_apicid(mps_cpu);
1924 #else
1925 	WARN_ON_ONCE(1);
1926 	return BAD_APICID;
1927 #endif
1928 }
1929 
1930 #define put_smstate(type, buf, offset, val)                      \
1931 	*(type *)((buf) + (offset) - 0x7e00) = val
1932 
1933 #define GET_SMSTATE(type, buf, offset)		\
1934 	(*(type *)((buf) + (offset) - 0x7e00))
1935 
1936 int kvm_cpu_dirty_log_size(void);
1937 
1938 int alloc_all_memslots_rmaps(struct kvm *kvm);
1939 
1940 #endif /* _ASM_X86_KVM_HOST_H */
1941