xref: /openbmc/linux/arch/x86/include/asm/kvm_host.h (revision ba936421)
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 #include <linux/workqueue.h>
19 
20 #include <linux/kvm.h>
21 #include <linux/kvm_para.h>
22 #include <linux/kvm_types.h>
23 #include <linux/perf_event.h>
24 #include <linux/pvclock_gtod.h>
25 #include <linux/clocksource.h>
26 #include <linux/irqbypass.h>
27 #include <linux/hyperv.h>
28 
29 #include <asm/apic.h>
30 #include <asm/pvclock-abi.h>
31 #include <asm/desc.h>
32 #include <asm/mtrr.h>
33 #include <asm/msr-index.h>
34 #include <asm/asm.h>
35 #include <asm/kvm_page_track.h>
36 #include <asm/kvm_vcpu_regs.h>
37 #include <asm/hyperv-tlfs.h>
38 
39 #define __KVM_HAVE_ARCH_VCPU_DEBUGFS
40 
41 #define KVM_MAX_VCPUS 1024
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_IDS (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_WAIT | 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 #define KVM_REQ_MMU_FREE_OBSOLETE_ROOTS \
107 	KVM_ARCH_REQ_FLAGS(31, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
108 
109 #define CR0_RESERVED_BITS                                               \
110 	(~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \
111 			  | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \
112 			  | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG))
113 
114 #define CR4_RESERVED_BITS                                               \
115 	(~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\
116 			  | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE     \
117 			  | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR | X86_CR4_PCIDE \
118 			  | X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_FSGSBASE \
119 			  | X86_CR4_OSXMMEXCPT | X86_CR4_LA57 | X86_CR4_VMXE \
120 			  | X86_CR4_SMAP | X86_CR4_PKE | X86_CR4_UMIP))
121 
122 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR)
123 
124 
125 
126 #define INVALID_PAGE (~(hpa_t)0)
127 #define VALID_PAGE(x) ((x) != INVALID_PAGE)
128 
129 #define UNMAPPED_GVA (~(gpa_t)0)
130 #define INVALID_GPA (~(gpa_t)0)
131 
132 /* KVM Hugepage definitions for x86 */
133 #define KVM_MAX_HUGEPAGE_LEVEL	PG_LEVEL_1G
134 #define KVM_NR_PAGE_SIZES	(KVM_MAX_HUGEPAGE_LEVEL - PG_LEVEL_4K + 1)
135 #define KVM_HPAGE_GFN_SHIFT(x)	(((x) - 1) * 9)
136 #define KVM_HPAGE_SHIFT(x)	(PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
137 #define KVM_HPAGE_SIZE(x)	(1UL << KVM_HPAGE_SHIFT(x))
138 #define KVM_HPAGE_MASK(x)	(~(KVM_HPAGE_SIZE(x) - 1))
139 #define KVM_PAGES_PER_HPAGE(x)	(KVM_HPAGE_SIZE(x) / PAGE_SIZE)
140 
141 #define KVM_MEMSLOT_PAGES_TO_MMU_PAGES_RATIO 50
142 #define KVM_MIN_ALLOC_MMU_PAGES 64UL
143 #define KVM_MMU_HASH_SHIFT 12
144 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
145 #define KVM_MIN_FREE_MMU_PAGES 5
146 #define KVM_REFILL_PAGES 25
147 #define KVM_MAX_CPUID_ENTRIES 256
148 #define KVM_NR_FIXED_MTRR_REGION 88
149 #define KVM_NR_VAR_MTRR 8
150 
151 #define ASYNC_PF_PER_VCPU 64
152 
153 enum kvm_reg {
154 	VCPU_REGS_RAX = __VCPU_REGS_RAX,
155 	VCPU_REGS_RCX = __VCPU_REGS_RCX,
156 	VCPU_REGS_RDX = __VCPU_REGS_RDX,
157 	VCPU_REGS_RBX = __VCPU_REGS_RBX,
158 	VCPU_REGS_RSP = __VCPU_REGS_RSP,
159 	VCPU_REGS_RBP = __VCPU_REGS_RBP,
160 	VCPU_REGS_RSI = __VCPU_REGS_RSI,
161 	VCPU_REGS_RDI = __VCPU_REGS_RDI,
162 #ifdef CONFIG_X86_64
163 	VCPU_REGS_R8  = __VCPU_REGS_R8,
164 	VCPU_REGS_R9  = __VCPU_REGS_R9,
165 	VCPU_REGS_R10 = __VCPU_REGS_R10,
166 	VCPU_REGS_R11 = __VCPU_REGS_R11,
167 	VCPU_REGS_R12 = __VCPU_REGS_R12,
168 	VCPU_REGS_R13 = __VCPU_REGS_R13,
169 	VCPU_REGS_R14 = __VCPU_REGS_R14,
170 	VCPU_REGS_R15 = __VCPU_REGS_R15,
171 #endif
172 	VCPU_REGS_RIP,
173 	NR_VCPU_REGS,
174 
175 	VCPU_EXREG_PDPTR = NR_VCPU_REGS,
176 	VCPU_EXREG_CR0,
177 	VCPU_EXREG_CR3,
178 	VCPU_EXREG_CR4,
179 	VCPU_EXREG_RFLAGS,
180 	VCPU_EXREG_SEGMENTS,
181 	VCPU_EXREG_EXIT_INFO_1,
182 	VCPU_EXREG_EXIT_INFO_2,
183 };
184 
185 enum {
186 	VCPU_SREG_ES,
187 	VCPU_SREG_CS,
188 	VCPU_SREG_SS,
189 	VCPU_SREG_DS,
190 	VCPU_SREG_FS,
191 	VCPU_SREG_GS,
192 	VCPU_SREG_TR,
193 	VCPU_SREG_LDTR,
194 };
195 
196 enum exit_fastpath_completion {
197 	EXIT_FASTPATH_NONE,
198 	EXIT_FASTPATH_REENTER_GUEST,
199 	EXIT_FASTPATH_EXIT_HANDLED,
200 };
201 typedef enum exit_fastpath_completion fastpath_t;
202 
203 struct x86_emulate_ctxt;
204 struct x86_exception;
205 enum x86_intercept;
206 enum x86_intercept_stage;
207 
208 #define KVM_NR_DB_REGS	4
209 
210 #define DR6_BUS_LOCK   (1 << 11)
211 #define DR6_BD		(1 << 13)
212 #define DR6_BS		(1 << 14)
213 #define DR6_BT		(1 << 15)
214 #define DR6_RTM		(1 << 16)
215 /*
216  * DR6_ACTIVE_LOW combines fixed-1 and active-low bits.
217  * We can regard all the bits in DR6_FIXED_1 as active_low bits;
218  * they will never be 0 for now, but when they are defined
219  * in the future it will require no code change.
220  *
221  * DR6_ACTIVE_LOW is also used as the init/reset value for DR6.
222  */
223 #define DR6_ACTIVE_LOW	0xffff0ff0
224 #define DR6_VOLATILE	0x0001e80f
225 #define DR6_FIXED_1	(DR6_ACTIVE_LOW & ~DR6_VOLATILE)
226 
227 #define DR7_BP_EN_MASK	0x000000ff
228 #define DR7_GE		(1 << 9)
229 #define DR7_GD		(1 << 13)
230 #define DR7_FIXED_1	0x00000400
231 #define DR7_VOLATILE	0xffff2bff
232 
233 #define KVM_GUESTDBG_VALID_MASK \
234 	(KVM_GUESTDBG_ENABLE | \
235 	KVM_GUESTDBG_SINGLESTEP | \
236 	KVM_GUESTDBG_USE_HW_BP | \
237 	KVM_GUESTDBG_USE_SW_BP | \
238 	KVM_GUESTDBG_INJECT_BP | \
239 	KVM_GUESTDBG_INJECT_DB | \
240 	KVM_GUESTDBG_BLOCKIRQ)
241 
242 
243 #define PFERR_PRESENT_BIT 0
244 #define PFERR_WRITE_BIT 1
245 #define PFERR_USER_BIT 2
246 #define PFERR_RSVD_BIT 3
247 #define PFERR_FETCH_BIT 4
248 #define PFERR_PK_BIT 5
249 #define PFERR_SGX_BIT 15
250 #define PFERR_GUEST_FINAL_BIT 32
251 #define PFERR_GUEST_PAGE_BIT 33
252 #define PFERR_IMPLICIT_ACCESS_BIT 48
253 
254 #define PFERR_PRESENT_MASK (1U << PFERR_PRESENT_BIT)
255 #define PFERR_WRITE_MASK (1U << PFERR_WRITE_BIT)
256 #define PFERR_USER_MASK (1U << PFERR_USER_BIT)
257 #define PFERR_RSVD_MASK (1U << PFERR_RSVD_BIT)
258 #define PFERR_FETCH_MASK (1U << PFERR_FETCH_BIT)
259 #define PFERR_PK_MASK (1U << PFERR_PK_BIT)
260 #define PFERR_SGX_MASK (1U << PFERR_SGX_BIT)
261 #define PFERR_GUEST_FINAL_MASK (1ULL << PFERR_GUEST_FINAL_BIT)
262 #define PFERR_GUEST_PAGE_MASK (1ULL << PFERR_GUEST_PAGE_BIT)
263 #define PFERR_IMPLICIT_ACCESS (1ULL << PFERR_IMPLICIT_ACCESS_BIT)
264 
265 #define PFERR_NESTED_GUEST_PAGE (PFERR_GUEST_PAGE_MASK |	\
266 				 PFERR_WRITE_MASK |		\
267 				 PFERR_PRESENT_MASK)
268 
269 /* apic attention bits */
270 #define KVM_APIC_CHECK_VAPIC	0
271 /*
272  * The following bit is set with PV-EOI, unset on EOI.
273  * We detect PV-EOI changes by guest by comparing
274  * this bit with PV-EOI in guest memory.
275  * See the implementation in apic_update_pv_eoi.
276  */
277 #define KVM_APIC_PV_EOI_PENDING	1
278 
279 struct kvm_kernel_irq_routing_entry;
280 
281 /*
282  * kvm_mmu_page_role tracks the properties of a shadow page (where shadow page
283  * also includes TDP pages) to determine whether or not a page can be used in
284  * the given MMU context.  This is a subset of the overall kvm_mmu_role to
285  * minimize the size of kvm_memory_slot.arch.gfn_track, i.e. allows allocating
286  * 2 bytes per gfn instead of 4 bytes per gfn.
287  *
288  * Indirect upper-level shadow pages are tracked for write-protection via
289  * gfn_track.  As above, gfn_track is a 16 bit counter, so KVM must not create
290  * more than 2^16-1 upper-level shadow pages at a single gfn, otherwise
291  * gfn_track will overflow and explosions will ensure.
292  *
293  * A unique shadow page (SP) for a gfn is created if and only if an existing SP
294  * cannot be reused.  The ability to reuse a SP is tracked by its role, which
295  * incorporates various mode bits and properties of the SP.  Roughly speaking,
296  * the number of unique SPs that can theoretically be created is 2^n, where n
297  * is the number of bits that are used to compute the role.
298  *
299  * But, even though there are 19 bits in the mask below, not all combinations
300  * of modes and flags are possible:
301  *
302  *   - invalid shadow pages are not accounted, so the bits are effectively 18
303  *
304  *   - quadrant will only be used if has_4_byte_gpte=1 (non-PAE paging);
305  *     execonly and ad_disabled are only used for nested EPT which has
306  *     has_4_byte_gpte=0.  Therefore, 2 bits are always unused.
307  *
308  *   - the 4 bits of level are effectively limited to the values 2/3/4/5,
309  *     as 4k SPs are not tracked (allowed to go unsync).  In addition non-PAE
310  *     paging has exactly one upper level, making level completely redundant
311  *     when has_4_byte_gpte=1.
312  *
313  *   - on top of this, smep_andnot_wp and smap_andnot_wp are only set if
314  *     cr0_wp=0, therefore these three bits only give rise to 5 possibilities.
315  *
316  * Therefore, the maximum number of possible upper-level shadow pages for a
317  * single gfn is a bit less than 2^13.
318  */
319 union kvm_mmu_page_role {
320 	u32 word;
321 	struct {
322 		unsigned level:4;
323 		unsigned has_4_byte_gpte:1;
324 		unsigned quadrant:2;
325 		unsigned direct:1;
326 		unsigned access:3;
327 		unsigned invalid:1;
328 		unsigned efer_nx:1;
329 		unsigned cr0_wp:1;
330 		unsigned smep_andnot_wp:1;
331 		unsigned smap_andnot_wp:1;
332 		unsigned ad_disabled:1;
333 		unsigned guest_mode:1;
334 		unsigned :6;
335 
336 		/*
337 		 * This is left at the top of the word so that
338 		 * kvm_memslots_for_spte_role can extract it with a
339 		 * simple shift.  While there is room, give it a whole
340 		 * byte so it is also faster to load it from memory.
341 		 */
342 		unsigned smm:8;
343 	};
344 };
345 
346 /*
347  * kvm_mmu_extended_role complements kvm_mmu_page_role, tracking properties
348  * relevant to the current MMU configuration.   When loading CR0, CR4, or EFER,
349  * including on nested transitions, if nothing in the full role changes then
350  * MMU re-configuration can be skipped. @valid bit is set on first usage so we
351  * don't treat all-zero structure as valid data.
352  *
353  * The properties that are tracked in the extended role but not the page role
354  * are for things that either (a) do not affect the validity of the shadow page
355  * or (b) are indirectly reflected in the shadow page's role.  For example,
356  * CR4.PKE only affects permission checks for software walks of the guest page
357  * tables (because KVM doesn't support Protection Keys with shadow paging), and
358  * CR0.PG, CR4.PAE, and CR4.PSE are indirectly reflected in role.level.
359  *
360  * Note, SMEP and SMAP are not redundant with sm*p_andnot_wp in the page role.
361  * If CR0.WP=1, KVM can reuse shadow pages for the guest regardless of SMEP and
362  * SMAP, but the MMU's permission checks for software walks need to be SMEP and
363  * SMAP aware regardless of CR0.WP.
364  */
365 union kvm_mmu_extended_role {
366 	u32 word;
367 	struct {
368 		unsigned int valid:1;
369 		unsigned int execonly:1;
370 		unsigned int cr0_pg:1;
371 		unsigned int cr4_pae:1;
372 		unsigned int cr4_pse:1;
373 		unsigned int cr4_pke:1;
374 		unsigned int cr4_smap:1;
375 		unsigned int cr4_smep:1;
376 		unsigned int cr4_la57:1;
377 		unsigned int efer_lma:1;
378 	};
379 };
380 
381 union kvm_mmu_role {
382 	u64 as_u64;
383 	struct {
384 		union kvm_mmu_page_role base;
385 		union kvm_mmu_extended_role ext;
386 	};
387 };
388 
389 struct kvm_rmap_head {
390 	unsigned long val;
391 };
392 
393 struct kvm_pio_request {
394 	unsigned long linear_rip;
395 	unsigned long count;
396 	int in;
397 	int port;
398 	int size;
399 };
400 
401 #define PT64_ROOT_MAX_LEVEL 5
402 
403 struct rsvd_bits_validate {
404 	u64 rsvd_bits_mask[2][PT64_ROOT_MAX_LEVEL];
405 	u64 bad_mt_xwr;
406 };
407 
408 struct kvm_mmu_root_info {
409 	gpa_t pgd;
410 	hpa_t hpa;
411 };
412 
413 #define KVM_MMU_ROOT_INFO_INVALID \
414 	((struct kvm_mmu_root_info) { .pgd = INVALID_PAGE, .hpa = INVALID_PAGE })
415 
416 #define KVM_MMU_NUM_PREV_ROOTS 3
417 
418 #define KVM_HAVE_MMU_RWLOCK
419 
420 struct kvm_mmu_page;
421 struct kvm_page_fault;
422 
423 /*
424  * x86 supports 4 paging modes (5-level 64-bit, 4-level 64-bit, 3-level 32-bit,
425  * and 2-level 32-bit).  The kvm_mmu structure abstracts the details of the
426  * current mmu mode.
427  */
428 struct kvm_mmu {
429 	unsigned long (*get_guest_pgd)(struct kvm_vcpu *vcpu);
430 	u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
431 	int (*page_fault)(struct kvm_vcpu *vcpu, struct kvm_page_fault *fault);
432 	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
433 				  struct x86_exception *fault);
434 	gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
435 			    gpa_t gva_or_gpa, u64 access,
436 			    struct x86_exception *exception);
437 	int (*sync_page)(struct kvm_vcpu *vcpu,
438 			 struct kvm_mmu_page *sp);
439 	void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva, hpa_t root_hpa);
440 	struct kvm_mmu_root_info root;
441 	union kvm_mmu_role mmu_role;
442 	u8 root_level;
443 	u8 shadow_root_level;
444 	u8 ept_ad;
445 	bool direct_map;
446 	struct kvm_mmu_root_info prev_roots[KVM_MMU_NUM_PREV_ROOTS];
447 
448 	/*
449 	 * Bitmap; bit set = permission fault
450 	 * Byte index: page fault error code [4:1]
451 	 * Bit index: pte permissions in ACC_* format
452 	 */
453 	u8 permissions[16];
454 
455 	/*
456 	* The pkru_mask indicates if protection key checks are needed.  It
457 	* consists of 16 domains indexed by page fault error code bits [4:1],
458 	* with PFEC.RSVD replaced by ACC_USER_MASK from the page tables.
459 	* Each domain has 2 bits which are ANDed with AD and WD from PKRU.
460 	*/
461 	u32 pkru_mask;
462 
463 	u64 *pae_root;
464 	u64 *pml4_root;
465 	u64 *pml5_root;
466 
467 	/*
468 	 * check zero bits on shadow page table entries, these
469 	 * bits include not only hardware reserved bits but also
470 	 * the bits spte never used.
471 	 */
472 	struct rsvd_bits_validate shadow_zero_check;
473 
474 	struct rsvd_bits_validate guest_rsvd_check;
475 
476 	u64 pdptrs[4]; /* pae */
477 };
478 
479 struct kvm_tlb_range {
480 	u64 start_gfn;
481 	u64 pages;
482 };
483 
484 enum pmc_type {
485 	KVM_PMC_GP = 0,
486 	KVM_PMC_FIXED,
487 };
488 
489 struct kvm_pmc {
490 	enum pmc_type type;
491 	u8 idx;
492 	u64 counter;
493 	u64 eventsel;
494 	struct perf_event *perf_event;
495 	struct kvm_vcpu *vcpu;
496 	/*
497 	 * eventsel value for general purpose counters,
498 	 * ctrl value for fixed counters.
499 	 */
500 	u64 current_config;
501 	bool is_paused;
502 	bool intr;
503 };
504 
505 #define KVM_PMC_MAX_FIXED	3
506 struct kvm_pmu {
507 	unsigned nr_arch_gp_counters;
508 	unsigned nr_arch_fixed_counters;
509 	unsigned available_event_types;
510 	u64 fixed_ctr_ctrl;
511 	u64 global_ctrl;
512 	u64 global_status;
513 	u64 counter_bitmask[2];
514 	u64 global_ctrl_mask;
515 	u64 global_ovf_ctrl_mask;
516 	u64 reserved_bits;
517 	u64 raw_event_mask;
518 	u8 version;
519 	struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
520 	struct kvm_pmc fixed_counters[KVM_PMC_MAX_FIXED];
521 	struct irq_work irq_work;
522 	DECLARE_BITMAP(reprogram_pmi, X86_PMC_IDX_MAX);
523 	DECLARE_BITMAP(all_valid_pmc_idx, X86_PMC_IDX_MAX);
524 	DECLARE_BITMAP(pmc_in_use, X86_PMC_IDX_MAX);
525 
526 	/*
527 	 * The gate to release perf_events not marked in
528 	 * pmc_in_use only once in a vcpu time slice.
529 	 */
530 	bool need_cleanup;
531 
532 	/*
533 	 * The total number of programmed perf_events and it helps to avoid
534 	 * redundant check before cleanup if guest don't use vPMU at all.
535 	 */
536 	u8 event_count;
537 };
538 
539 struct kvm_pmu_ops;
540 
541 enum {
542 	KVM_DEBUGREG_BP_ENABLED = 1,
543 	KVM_DEBUGREG_WONT_EXIT = 2,
544 };
545 
546 struct kvm_mtrr_range {
547 	u64 base;
548 	u64 mask;
549 	struct list_head node;
550 };
551 
552 struct kvm_mtrr {
553 	struct kvm_mtrr_range var_ranges[KVM_NR_VAR_MTRR];
554 	mtrr_type fixed_ranges[KVM_NR_FIXED_MTRR_REGION];
555 	u64 deftype;
556 
557 	struct list_head head;
558 };
559 
560 /* Hyper-V SynIC timer */
561 struct kvm_vcpu_hv_stimer {
562 	struct hrtimer timer;
563 	int index;
564 	union hv_stimer_config config;
565 	u64 count;
566 	u64 exp_time;
567 	struct hv_message msg;
568 	bool msg_pending;
569 };
570 
571 /* Hyper-V synthetic interrupt controller (SynIC)*/
572 struct kvm_vcpu_hv_synic {
573 	u64 version;
574 	u64 control;
575 	u64 msg_page;
576 	u64 evt_page;
577 	atomic64_t sint[HV_SYNIC_SINT_COUNT];
578 	atomic_t sint_to_gsi[HV_SYNIC_SINT_COUNT];
579 	DECLARE_BITMAP(auto_eoi_bitmap, 256);
580 	DECLARE_BITMAP(vec_bitmap, 256);
581 	bool active;
582 	bool dont_zero_synic_pages;
583 };
584 
585 /* Hyper-V per vcpu emulation context */
586 struct kvm_vcpu_hv {
587 	struct kvm_vcpu *vcpu;
588 	u32 vp_index;
589 	u64 hv_vapic;
590 	s64 runtime_offset;
591 	struct kvm_vcpu_hv_synic synic;
592 	struct kvm_hyperv_exit exit;
593 	struct kvm_vcpu_hv_stimer stimer[HV_SYNIC_STIMER_COUNT];
594 	DECLARE_BITMAP(stimer_pending_bitmap, HV_SYNIC_STIMER_COUNT);
595 	bool enforce_cpuid;
596 	struct {
597 		u32 features_eax; /* HYPERV_CPUID_FEATURES.EAX */
598 		u32 features_ebx; /* HYPERV_CPUID_FEATURES.EBX */
599 		u32 features_edx; /* HYPERV_CPUID_FEATURES.EDX */
600 		u32 enlightenments_eax; /* HYPERV_CPUID_ENLIGHTMENT_INFO.EAX */
601 		u32 enlightenments_ebx; /* HYPERV_CPUID_ENLIGHTMENT_INFO.EBX */
602 		u32 syndbg_cap_eax; /* HYPERV_CPUID_SYNDBG_PLATFORM_CAPABILITIES.EAX */
603 	} cpuid_cache;
604 };
605 
606 /* Xen HVM per vcpu emulation context */
607 struct kvm_vcpu_xen {
608 	u64 hypercall_rip;
609 	u32 current_runstate;
610 	bool vcpu_info_set;
611 	bool vcpu_time_info_set;
612 	bool runstate_set;
613 	struct gfn_to_hva_cache vcpu_info_cache;
614 	struct gfn_to_hva_cache vcpu_time_info_cache;
615 	struct gfn_to_hva_cache runstate_cache;
616 	u64 last_steal;
617 	u64 runstate_entry_time;
618 	u64 runstate_times[4];
619 	unsigned long evtchn_pending_sel;
620 };
621 
622 struct kvm_vcpu_arch {
623 	/*
624 	 * rip and regs accesses must go through
625 	 * kvm_{register,rip}_{read,write} functions.
626 	 */
627 	unsigned long regs[NR_VCPU_REGS];
628 	u32 regs_avail;
629 	u32 regs_dirty;
630 
631 	unsigned long cr0;
632 	unsigned long cr0_guest_owned_bits;
633 	unsigned long cr2;
634 	unsigned long cr3;
635 	unsigned long cr4;
636 	unsigned long cr4_guest_owned_bits;
637 	unsigned long cr4_guest_rsvd_bits;
638 	unsigned long cr8;
639 	u32 host_pkru;
640 	u32 pkru;
641 	u32 hflags;
642 	u64 efer;
643 	u64 apic_base;
644 	struct kvm_lapic *apic;    /* kernel irqchip context */
645 	bool apicv_active;
646 	bool load_eoi_exitmap_pending;
647 	DECLARE_BITMAP(ioapic_handled_vectors, 256);
648 	unsigned long apic_attention;
649 	int32_t apic_arb_prio;
650 	int mp_state;
651 	u64 ia32_misc_enable_msr;
652 	u64 smbase;
653 	u64 smi_count;
654 	bool tpr_access_reporting;
655 	bool xsaves_enabled;
656 	bool xfd_no_write_intercept;
657 	u64 ia32_xss;
658 	u64 microcode_version;
659 	u64 arch_capabilities;
660 	u64 perf_capabilities;
661 
662 	/*
663 	 * Paging state of the vcpu
664 	 *
665 	 * If the vcpu runs in guest mode with two level paging this still saves
666 	 * the paging mode of the l1 guest. This context is always used to
667 	 * handle faults.
668 	 */
669 	struct kvm_mmu *mmu;
670 
671 	/* Non-nested MMU for L1 */
672 	struct kvm_mmu root_mmu;
673 
674 	/* L1 MMU when running nested */
675 	struct kvm_mmu guest_mmu;
676 
677 	/*
678 	 * Paging state of an L2 guest (used for nested npt)
679 	 *
680 	 * This context will save all necessary information to walk page tables
681 	 * of an L2 guest. This context is only initialized for page table
682 	 * walking and not for faulting since we never handle l2 page faults on
683 	 * the host.
684 	 */
685 	struct kvm_mmu nested_mmu;
686 
687 	/*
688 	 * Pointer to the mmu context currently used for
689 	 * gva_to_gpa translations.
690 	 */
691 	struct kvm_mmu *walk_mmu;
692 
693 	struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
694 	struct kvm_mmu_memory_cache mmu_shadow_page_cache;
695 	struct kvm_mmu_memory_cache mmu_gfn_array_cache;
696 	struct kvm_mmu_memory_cache mmu_page_header_cache;
697 
698 	/*
699 	 * QEMU userspace and the guest each have their own FPU state.
700 	 * In vcpu_run, we switch between the user and guest FPU contexts.
701 	 * While running a VCPU, the VCPU thread will have the guest FPU
702 	 * context.
703 	 *
704 	 * Note that while the PKRU state lives inside the fpu registers,
705 	 * it is switched out separately at VMENTER and VMEXIT time. The
706 	 * "guest_fpstate" state here contains the guest FPU context, with the
707 	 * host PRKU bits.
708 	 */
709 	struct fpu_guest guest_fpu;
710 
711 	u64 xcr0;
712 
713 	struct kvm_pio_request pio;
714 	void *pio_data;
715 	void *sev_pio_data;
716 	unsigned sev_pio_count;
717 
718 	u8 event_exit_inst_len;
719 
720 	struct kvm_queued_exception {
721 		bool pending;
722 		bool injected;
723 		bool has_error_code;
724 		u8 nr;
725 		u32 error_code;
726 		unsigned long payload;
727 		bool has_payload;
728 		u8 nested_apf;
729 	} exception;
730 
731 	struct kvm_queued_interrupt {
732 		bool injected;
733 		bool soft;
734 		u8 nr;
735 	} interrupt;
736 
737 	int halt_request; /* real mode on Intel only */
738 
739 	int cpuid_nent;
740 	struct kvm_cpuid_entry2 *cpuid_entries;
741 	u32 kvm_cpuid_base;
742 
743 	u64 reserved_gpa_bits;
744 	int maxphyaddr;
745 
746 	/* emulate context */
747 
748 	struct x86_emulate_ctxt *emulate_ctxt;
749 	bool emulate_regs_need_sync_to_vcpu;
750 	bool emulate_regs_need_sync_from_vcpu;
751 	int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
752 
753 	gpa_t time;
754 	struct pvclock_vcpu_time_info hv_clock;
755 	unsigned int hw_tsc_khz;
756 	struct gfn_to_hva_cache pv_time;
757 	bool pv_time_enabled;
758 	/* set guest stopped flag in pvclock flags field */
759 	bool pvclock_set_guest_stopped_request;
760 
761 	struct {
762 		u8 preempted;
763 		u64 msr_val;
764 		u64 last_steal;
765 		struct gfn_to_hva_cache cache;
766 	} st;
767 
768 	u64 l1_tsc_offset;
769 	u64 tsc_offset; /* current tsc offset */
770 	u64 last_guest_tsc;
771 	u64 last_host_tsc;
772 	u64 tsc_offset_adjustment;
773 	u64 this_tsc_nsec;
774 	u64 this_tsc_write;
775 	u64 this_tsc_generation;
776 	bool tsc_catchup;
777 	bool tsc_always_catchup;
778 	s8 virtual_tsc_shift;
779 	u32 virtual_tsc_mult;
780 	u32 virtual_tsc_khz;
781 	s64 ia32_tsc_adjust_msr;
782 	u64 msr_ia32_power_ctl;
783 	u64 l1_tsc_scaling_ratio;
784 	u64 tsc_scaling_ratio; /* current scaling ratio */
785 
786 	atomic_t nmi_queued;  /* unprocessed asynchronous NMIs */
787 	unsigned nmi_pending; /* NMI queued after currently running handler */
788 	bool nmi_injected;    /* Trying to inject an NMI this entry */
789 	bool smi_pending;    /* SMI queued after currently running handler */
790 	u8 handling_intr_from_guest;
791 
792 	struct kvm_mtrr mtrr_state;
793 	u64 pat;
794 
795 	unsigned switch_db_regs;
796 	unsigned long db[KVM_NR_DB_REGS];
797 	unsigned long dr6;
798 	unsigned long dr7;
799 	unsigned long eff_db[KVM_NR_DB_REGS];
800 	unsigned long guest_debug_dr7;
801 	u64 msr_platform_info;
802 	u64 msr_misc_features_enables;
803 
804 	u64 mcg_cap;
805 	u64 mcg_status;
806 	u64 mcg_ctl;
807 	u64 mcg_ext_ctl;
808 	u64 *mce_banks;
809 
810 	/* Cache MMIO info */
811 	u64 mmio_gva;
812 	unsigned mmio_access;
813 	gfn_t mmio_gfn;
814 	u64 mmio_gen;
815 
816 	struct kvm_pmu pmu;
817 
818 	/* used for guest single stepping over the given code position */
819 	unsigned long singlestep_rip;
820 
821 	bool hyperv_enabled;
822 	struct kvm_vcpu_hv *hyperv;
823 	struct kvm_vcpu_xen xen;
824 
825 	cpumask_var_t wbinvd_dirty_mask;
826 
827 	unsigned long last_retry_eip;
828 	unsigned long last_retry_addr;
829 
830 	struct {
831 		bool halted;
832 		gfn_t gfns[ASYNC_PF_PER_VCPU];
833 		struct gfn_to_hva_cache data;
834 		u64 msr_en_val; /* MSR_KVM_ASYNC_PF_EN */
835 		u64 msr_int_val; /* MSR_KVM_ASYNC_PF_INT */
836 		u16 vec;
837 		u32 id;
838 		bool send_user_only;
839 		u32 host_apf_flags;
840 		unsigned long nested_apf_token;
841 		bool delivery_as_pf_vmexit;
842 		bool pageready_pending;
843 	} apf;
844 
845 	/* OSVW MSRs (AMD only) */
846 	struct {
847 		u64 length;
848 		u64 status;
849 	} osvw;
850 
851 	struct {
852 		u64 msr_val;
853 		struct gfn_to_hva_cache data;
854 	} pv_eoi;
855 
856 	u64 msr_kvm_poll_control;
857 
858 	/*
859 	 * Indicates the guest is trying to write a gfn that contains one or
860 	 * more of the PTEs used to translate the write itself, i.e. the access
861 	 * is changing its own translation in the guest page tables.  KVM exits
862 	 * to userspace if emulation of the faulting instruction fails and this
863 	 * flag is set, as KVM cannot make forward progress.
864 	 *
865 	 * If emulation fails for a write to guest page tables, KVM unprotects
866 	 * (zaps) the shadow page for the target gfn and resumes the guest to
867 	 * retry the non-emulatable instruction (on hardware).  Unprotecting the
868 	 * gfn doesn't allow forward progress for a self-changing access because
869 	 * doing so also zaps the translation for the gfn, i.e. retrying the
870 	 * instruction will hit a !PRESENT fault, which results in a new shadow
871 	 * page and sends KVM back to square one.
872 	 */
873 	bool write_fault_to_shadow_pgtable;
874 
875 	/* set at EPT violation at this point */
876 	unsigned long exit_qualification;
877 
878 	/* pv related host specific info */
879 	struct {
880 		bool pv_unhalted;
881 	} pv;
882 
883 	int pending_ioapic_eoi;
884 	int pending_external_vector;
885 
886 	/* be preempted when it's in kernel-mode(cpl=0) */
887 	bool preempted_in_kernel;
888 
889 	/* Flush the L1 Data cache for L1TF mitigation on VMENTER */
890 	bool l1tf_flush_l1d;
891 
892 	/* Host CPU on which VM-entry was most recently attempted */
893 	int last_vmentry_cpu;
894 
895 	/* AMD MSRC001_0015 Hardware Configuration */
896 	u64 msr_hwcr;
897 
898 	/* pv related cpuid info */
899 	struct {
900 		/*
901 		 * value of the eax register in the KVM_CPUID_FEATURES CPUID
902 		 * leaf.
903 		 */
904 		u32 features;
905 
906 		/*
907 		 * indicates whether pv emulation should be disabled if features
908 		 * are not present in the guest's cpuid
909 		 */
910 		bool enforce;
911 	} pv_cpuid;
912 
913 	/* Protected Guests */
914 	bool guest_state_protected;
915 
916 	/*
917 	 * Set when PDPTS were loaded directly by the userspace without
918 	 * reading the guest memory
919 	 */
920 	bool pdptrs_from_userspace;
921 
922 #if IS_ENABLED(CONFIG_HYPERV)
923 	hpa_t hv_root_tdp;
924 #endif
925 };
926 
927 struct kvm_lpage_info {
928 	int disallow_lpage;
929 };
930 
931 struct kvm_arch_memory_slot {
932 	struct kvm_rmap_head *rmap[KVM_NR_PAGE_SIZES];
933 	struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
934 	unsigned short *gfn_track[KVM_PAGE_TRACK_MAX];
935 };
936 
937 /*
938  * We use as the mode the number of bits allocated in the LDR for the
939  * logical processor ID.  It happens that these are all powers of two.
940  * This makes it is very easy to detect cases where the APICs are
941  * configured for multiple modes; in that case, we cannot use the map and
942  * hence cannot use kvm_irq_delivery_to_apic_fast either.
943  */
944 #define KVM_APIC_MODE_XAPIC_CLUSTER          4
945 #define KVM_APIC_MODE_XAPIC_FLAT             8
946 #define KVM_APIC_MODE_X2APIC                16
947 
948 struct kvm_apic_map {
949 	struct rcu_head rcu;
950 	u8 mode;
951 	u32 max_apic_id;
952 	union {
953 		struct kvm_lapic *xapic_flat_map[8];
954 		struct kvm_lapic *xapic_cluster_map[16][4];
955 	};
956 	struct kvm_lapic *phys_map[];
957 };
958 
959 /* Hyper-V synthetic debugger (SynDbg)*/
960 struct kvm_hv_syndbg {
961 	struct {
962 		u64 control;
963 		u64 status;
964 		u64 send_page;
965 		u64 recv_page;
966 		u64 pending_page;
967 	} control;
968 	u64 options;
969 };
970 
971 /* Current state of Hyper-V TSC page clocksource */
972 enum hv_tsc_page_status {
973 	/* TSC page was not set up or disabled */
974 	HV_TSC_PAGE_UNSET = 0,
975 	/* TSC page MSR was written by the guest, update pending */
976 	HV_TSC_PAGE_GUEST_CHANGED,
977 	/* TSC page MSR was written by KVM userspace, update pending */
978 	HV_TSC_PAGE_HOST_CHANGED,
979 	/* TSC page was properly set up and is currently active  */
980 	HV_TSC_PAGE_SET,
981 	/* TSC page is currently being updated and therefore is inactive */
982 	HV_TSC_PAGE_UPDATING,
983 	/* TSC page was set up with an inaccessible GPA */
984 	HV_TSC_PAGE_BROKEN,
985 };
986 
987 /* Hyper-V emulation context */
988 struct kvm_hv {
989 	struct mutex hv_lock;
990 	u64 hv_guest_os_id;
991 	u64 hv_hypercall;
992 	u64 hv_tsc_page;
993 	enum hv_tsc_page_status hv_tsc_page_status;
994 
995 	/* Hyper-v based guest crash (NT kernel bugcheck) parameters */
996 	u64 hv_crash_param[HV_X64_MSR_CRASH_PARAMS];
997 	u64 hv_crash_ctl;
998 
999 	struct ms_hyperv_tsc_page tsc_ref;
1000 
1001 	struct idr conn_to_evt;
1002 
1003 	u64 hv_reenlightenment_control;
1004 	u64 hv_tsc_emulation_control;
1005 	u64 hv_tsc_emulation_status;
1006 
1007 	/* How many vCPUs have VP index != vCPU index */
1008 	atomic_t num_mismatched_vp_indexes;
1009 
1010 	/*
1011 	 * How many SynICs use 'AutoEOI' feature
1012 	 * (protected by arch.apicv_update_lock)
1013 	 */
1014 	unsigned int synic_auto_eoi_used;
1015 
1016 	struct hv_partition_assist_pg *hv_pa_pg;
1017 	struct kvm_hv_syndbg hv_syndbg;
1018 };
1019 
1020 struct msr_bitmap_range {
1021 	u32 flags;
1022 	u32 nmsrs;
1023 	u32 base;
1024 	unsigned long *bitmap;
1025 };
1026 
1027 /* Xen emulation context */
1028 struct kvm_xen {
1029 	bool long_mode;
1030 	u8 upcall_vector;
1031 	struct gfn_to_pfn_cache shinfo_cache;
1032 };
1033 
1034 enum kvm_irqchip_mode {
1035 	KVM_IRQCHIP_NONE,
1036 	KVM_IRQCHIP_KERNEL,       /* created with KVM_CREATE_IRQCHIP */
1037 	KVM_IRQCHIP_SPLIT,        /* created with KVM_CAP_SPLIT_IRQCHIP */
1038 };
1039 
1040 struct kvm_x86_msr_filter {
1041 	u8 count;
1042 	bool default_allow:1;
1043 	struct msr_bitmap_range ranges[16];
1044 };
1045 
1046 enum kvm_apicv_inhibit {
1047 	APICV_INHIBIT_REASON_DISABLE,
1048 	APICV_INHIBIT_REASON_HYPERV,
1049 	APICV_INHIBIT_REASON_NESTED,
1050 	APICV_INHIBIT_REASON_IRQWIN,
1051 	APICV_INHIBIT_REASON_PIT_REINJ,
1052 	APICV_INHIBIT_REASON_X2APIC,
1053 	APICV_INHIBIT_REASON_BLOCKIRQ,
1054 	APICV_INHIBIT_REASON_ABSENT,
1055 };
1056 
1057 struct kvm_arch {
1058 	unsigned long n_used_mmu_pages;
1059 	unsigned long n_requested_mmu_pages;
1060 	unsigned long n_max_mmu_pages;
1061 	unsigned int indirect_shadow_pages;
1062 	u8 mmu_valid_gen;
1063 	struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
1064 	struct list_head active_mmu_pages;
1065 	struct list_head zapped_obsolete_pages;
1066 	struct list_head lpage_disallowed_mmu_pages;
1067 	struct kvm_page_track_notifier_node mmu_sp_tracker;
1068 	struct kvm_page_track_notifier_head track_notifier_head;
1069 	/*
1070 	 * Protects marking pages unsync during page faults, as TDP MMU page
1071 	 * faults only take mmu_lock for read.  For simplicity, the unsync
1072 	 * pages lock is always taken when marking pages unsync regardless of
1073 	 * whether mmu_lock is held for read or write.
1074 	 */
1075 	spinlock_t mmu_unsync_pages_lock;
1076 
1077 	struct list_head assigned_dev_head;
1078 	struct iommu_domain *iommu_domain;
1079 	bool iommu_noncoherent;
1080 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA
1081 	atomic_t noncoherent_dma_count;
1082 #define __KVM_HAVE_ARCH_ASSIGNED_DEVICE
1083 	atomic_t assigned_device_count;
1084 	struct kvm_pic *vpic;
1085 	struct kvm_ioapic *vioapic;
1086 	struct kvm_pit *vpit;
1087 	atomic_t vapics_in_nmi_mode;
1088 	struct mutex apic_map_lock;
1089 	struct kvm_apic_map __rcu *apic_map;
1090 	atomic_t apic_map_dirty;
1091 
1092 	/* Protects apic_access_memslot_enabled and apicv_inhibit_reasons */
1093 	struct rw_semaphore apicv_update_lock;
1094 
1095 	bool apic_access_memslot_enabled;
1096 	unsigned long apicv_inhibit_reasons;
1097 
1098 	gpa_t wall_clock;
1099 
1100 	bool mwait_in_guest;
1101 	bool hlt_in_guest;
1102 	bool pause_in_guest;
1103 	bool cstate_in_guest;
1104 
1105 	unsigned long irq_sources_bitmap;
1106 	s64 kvmclock_offset;
1107 
1108 	/*
1109 	 * This also protects nr_vcpus_matched_tsc which is read from a
1110 	 * preemption-disabled region, so it must be a raw spinlock.
1111 	 */
1112 	raw_spinlock_t tsc_write_lock;
1113 	u64 last_tsc_nsec;
1114 	u64 last_tsc_write;
1115 	u32 last_tsc_khz;
1116 	u64 last_tsc_offset;
1117 	u64 cur_tsc_nsec;
1118 	u64 cur_tsc_write;
1119 	u64 cur_tsc_offset;
1120 	u64 cur_tsc_generation;
1121 	int nr_vcpus_matched_tsc;
1122 
1123 	seqcount_raw_spinlock_t pvclock_sc;
1124 	bool use_master_clock;
1125 	u64 master_kernel_ns;
1126 	u64 master_cycle_now;
1127 	struct delayed_work kvmclock_update_work;
1128 	struct delayed_work kvmclock_sync_work;
1129 
1130 	struct kvm_xen_hvm_config xen_hvm_config;
1131 
1132 	/* reads protected by irq_srcu, writes by irq_lock */
1133 	struct hlist_head mask_notifier_list;
1134 
1135 	struct kvm_hv hyperv;
1136 	struct kvm_xen xen;
1137 
1138 	bool backwards_tsc_observed;
1139 	bool boot_vcpu_runs_old_kvmclock;
1140 	u32 bsp_vcpu_id;
1141 
1142 	u64 disabled_quirks;
1143 	int cpu_dirty_logging_count;
1144 
1145 	enum kvm_irqchip_mode irqchip_mode;
1146 	u8 nr_reserved_ioapic_pins;
1147 
1148 	bool disabled_lapic_found;
1149 
1150 	bool x2apic_format;
1151 	bool x2apic_broadcast_quirk_disabled;
1152 
1153 	bool guest_can_read_msr_platform_info;
1154 	bool exception_payload_enabled;
1155 
1156 	bool bus_lock_detection_enabled;
1157 	bool enable_pmu;
1158 	/*
1159 	 * If exit_on_emulation_error is set, and the in-kernel instruction
1160 	 * emulator fails to emulate an instruction, allow userspace
1161 	 * the opportunity to look at it.
1162 	 */
1163 	bool exit_on_emulation_error;
1164 
1165 	/* Deflect RDMSR and WRMSR to user space when they trigger a #GP */
1166 	u32 user_space_msr_mask;
1167 	struct kvm_x86_msr_filter __rcu *msr_filter;
1168 
1169 	u32 hypercall_exit_enabled;
1170 
1171 	/* Guest can access the SGX PROVISIONKEY. */
1172 	bool sgx_provisioning_allowed;
1173 
1174 	struct kvm_pmu_event_filter __rcu *pmu_event_filter;
1175 	struct task_struct *nx_lpage_recovery_thread;
1176 
1177 #ifdef CONFIG_X86_64
1178 	/*
1179 	 * Whether the TDP MMU is enabled for this VM. This contains a
1180 	 * snapshot of the TDP MMU module parameter from when the VM was
1181 	 * created and remains unchanged for the life of the VM. If this is
1182 	 * true, TDP MMU handler functions will run for various MMU
1183 	 * operations.
1184 	 */
1185 	bool tdp_mmu_enabled;
1186 
1187 	/*
1188 	 * List of struct kvm_mmu_pages being used as roots.
1189 	 * All struct kvm_mmu_pages in the list should have
1190 	 * tdp_mmu_page set.
1191 	 *
1192 	 * For reads, this list is protected by:
1193 	 *	the MMU lock in read mode + RCU or
1194 	 *	the MMU lock in write mode
1195 	 *
1196 	 * For writes, this list is protected by:
1197 	 *	the MMU lock in read mode + the tdp_mmu_pages_lock or
1198 	 *	the MMU lock in write mode
1199 	 *
1200 	 * Roots will remain in the list until their tdp_mmu_root_count
1201 	 * drops to zero, at which point the thread that decremented the
1202 	 * count to zero should removed the root from the list and clean
1203 	 * it up, freeing the root after an RCU grace period.
1204 	 */
1205 	struct list_head tdp_mmu_roots;
1206 
1207 	/*
1208 	 * List of struct kvmp_mmu_pages not being used as roots.
1209 	 * All struct kvm_mmu_pages in the list should have
1210 	 * tdp_mmu_page set and a tdp_mmu_root_count of 0.
1211 	 */
1212 	struct list_head tdp_mmu_pages;
1213 
1214 	/*
1215 	 * Protects accesses to the following fields when the MMU lock
1216 	 * is held in read mode:
1217 	 *  - tdp_mmu_roots (above)
1218 	 *  - tdp_mmu_pages (above)
1219 	 *  - the link field of struct kvm_mmu_pages used by the TDP MMU
1220 	 *  - lpage_disallowed_mmu_pages
1221 	 *  - the lpage_disallowed_link field of struct kvm_mmu_pages used
1222 	 *    by the TDP MMU
1223 	 * It is acceptable, but not necessary, to acquire this lock when
1224 	 * the thread holds the MMU lock in write mode.
1225 	 */
1226 	spinlock_t tdp_mmu_pages_lock;
1227 	struct workqueue_struct *tdp_mmu_zap_wq;
1228 #endif /* CONFIG_X86_64 */
1229 
1230 	/*
1231 	 * If set, at least one shadow root has been allocated. This flag
1232 	 * is used as one input when determining whether certain memslot
1233 	 * related allocations are necessary.
1234 	 */
1235 	bool shadow_root_allocated;
1236 
1237 #if IS_ENABLED(CONFIG_HYPERV)
1238 	hpa_t	hv_root_tdp;
1239 	spinlock_t hv_root_tdp_lock;
1240 #endif
1241 };
1242 
1243 struct kvm_vm_stat {
1244 	struct kvm_vm_stat_generic generic;
1245 	u64 mmu_shadow_zapped;
1246 	u64 mmu_pte_write;
1247 	u64 mmu_pde_zapped;
1248 	u64 mmu_flooded;
1249 	u64 mmu_recycled;
1250 	u64 mmu_cache_miss;
1251 	u64 mmu_unsync;
1252 	union {
1253 		struct {
1254 			atomic64_t pages_4k;
1255 			atomic64_t pages_2m;
1256 			atomic64_t pages_1g;
1257 		};
1258 		atomic64_t pages[KVM_NR_PAGE_SIZES];
1259 	};
1260 	u64 nx_lpage_splits;
1261 	u64 max_mmu_page_hash_collisions;
1262 	u64 max_mmu_rmap_size;
1263 };
1264 
1265 struct kvm_vcpu_stat {
1266 	struct kvm_vcpu_stat_generic generic;
1267 	u64 pf_fixed;
1268 	u64 pf_guest;
1269 	u64 tlb_flush;
1270 	u64 invlpg;
1271 
1272 	u64 exits;
1273 	u64 io_exits;
1274 	u64 mmio_exits;
1275 	u64 signal_exits;
1276 	u64 irq_window_exits;
1277 	u64 nmi_window_exits;
1278 	u64 l1d_flush;
1279 	u64 halt_exits;
1280 	u64 request_irq_exits;
1281 	u64 irq_exits;
1282 	u64 host_state_reload;
1283 	u64 fpu_reload;
1284 	u64 insn_emulation;
1285 	u64 insn_emulation_fail;
1286 	u64 hypercalls;
1287 	u64 irq_injections;
1288 	u64 nmi_injections;
1289 	u64 req_event;
1290 	u64 nested_run;
1291 	u64 directed_yield_attempted;
1292 	u64 directed_yield_successful;
1293 	u64 guest_mode;
1294 };
1295 
1296 struct x86_instruction_info;
1297 
1298 struct msr_data {
1299 	bool host_initiated;
1300 	u32 index;
1301 	u64 data;
1302 };
1303 
1304 struct kvm_lapic_irq {
1305 	u32 vector;
1306 	u16 delivery_mode;
1307 	u16 dest_mode;
1308 	bool level;
1309 	u16 trig_mode;
1310 	u32 shorthand;
1311 	u32 dest_id;
1312 	bool msi_redir_hint;
1313 };
1314 
1315 static inline u16 kvm_lapic_irq_dest_mode(bool dest_mode_logical)
1316 {
1317 	return dest_mode_logical ? APIC_DEST_LOGICAL : APIC_DEST_PHYSICAL;
1318 }
1319 
1320 struct kvm_x86_ops {
1321 	const char *name;
1322 
1323 	int (*hardware_enable)(void);
1324 	void (*hardware_disable)(void);
1325 	void (*hardware_unsetup)(void);
1326 	bool (*has_emulated_msr)(struct kvm *kvm, u32 index);
1327 	void (*vcpu_after_set_cpuid)(struct kvm_vcpu *vcpu);
1328 
1329 	unsigned int vm_size;
1330 	int (*vm_init)(struct kvm *kvm);
1331 	void (*vm_destroy)(struct kvm *kvm);
1332 
1333 	/* Create, but do not attach this VCPU */
1334 	int (*vcpu_create)(struct kvm_vcpu *vcpu);
1335 	void (*vcpu_free)(struct kvm_vcpu *vcpu);
1336 	void (*vcpu_reset)(struct kvm_vcpu *vcpu, bool init_event);
1337 
1338 	void (*prepare_switch_to_guest)(struct kvm_vcpu *vcpu);
1339 	void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
1340 	void (*vcpu_put)(struct kvm_vcpu *vcpu);
1341 
1342 	void (*update_exception_bitmap)(struct kvm_vcpu *vcpu);
1343 	int (*get_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1344 	int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1345 	u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
1346 	void (*get_segment)(struct kvm_vcpu *vcpu,
1347 			    struct kvm_segment *var, int seg);
1348 	int (*get_cpl)(struct kvm_vcpu *vcpu);
1349 	void (*set_segment)(struct kvm_vcpu *vcpu,
1350 			    struct kvm_segment *var, int seg);
1351 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
1352 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
1353 	void (*post_set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
1354 	bool (*is_valid_cr4)(struct kvm_vcpu *vcpu, unsigned long cr0);
1355 	void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
1356 	int (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
1357 	void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1358 	void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1359 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1360 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1361 	void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu);
1362 	void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
1363 	void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
1364 	unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
1365 	void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
1366 	bool (*get_if_flag)(struct kvm_vcpu *vcpu);
1367 
1368 	void (*flush_tlb_all)(struct kvm_vcpu *vcpu);
1369 	void (*flush_tlb_current)(struct kvm_vcpu *vcpu);
1370 	int  (*tlb_remote_flush)(struct kvm *kvm);
1371 	int  (*tlb_remote_flush_with_range)(struct kvm *kvm,
1372 			struct kvm_tlb_range *range);
1373 
1374 	/*
1375 	 * Flush any TLB entries associated with the given GVA.
1376 	 * Does not need to flush GPA->HPA mappings.
1377 	 * Can potentially get non-canonical addresses through INVLPGs, which
1378 	 * the implementation may choose to ignore if appropriate.
1379 	 */
1380 	void (*flush_tlb_gva)(struct kvm_vcpu *vcpu, gva_t addr);
1381 
1382 	/*
1383 	 * Flush any TLB entries created by the guest.  Like tlb_flush_gva(),
1384 	 * does not need to flush GPA->HPA mappings.
1385 	 */
1386 	void (*flush_tlb_guest)(struct kvm_vcpu *vcpu);
1387 
1388 	int (*vcpu_pre_run)(struct kvm_vcpu *vcpu);
1389 	enum exit_fastpath_completion (*vcpu_run)(struct kvm_vcpu *vcpu);
1390 	int (*handle_exit)(struct kvm_vcpu *vcpu,
1391 		enum exit_fastpath_completion exit_fastpath);
1392 	int (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
1393 	void (*update_emulated_instruction)(struct kvm_vcpu *vcpu);
1394 	void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
1395 	u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu);
1396 	void (*patch_hypercall)(struct kvm_vcpu *vcpu,
1397 				unsigned char *hypercall_addr);
1398 	void (*inject_irq)(struct kvm_vcpu *vcpu);
1399 	void (*inject_nmi)(struct kvm_vcpu *vcpu);
1400 	void (*queue_exception)(struct kvm_vcpu *vcpu);
1401 	void (*cancel_injection)(struct kvm_vcpu *vcpu);
1402 	int (*interrupt_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1403 	int (*nmi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1404 	bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
1405 	void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
1406 	void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
1407 	void (*enable_irq_window)(struct kvm_vcpu *vcpu);
1408 	void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
1409 	bool (*check_apicv_inhibit_reasons)(enum kvm_apicv_inhibit reason);
1410 	void (*refresh_apicv_exec_ctrl)(struct kvm_vcpu *vcpu);
1411 	void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
1412 	void (*hwapic_isr_update)(struct kvm_vcpu *vcpu, int isr);
1413 	bool (*guest_apic_has_interrupt)(struct kvm_vcpu *vcpu);
1414 	void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
1415 	void (*set_virtual_apic_mode)(struct kvm_vcpu *vcpu);
1416 	void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu);
1417 	void (*deliver_interrupt)(struct kvm_lapic *apic, int delivery_mode,
1418 				  int trig_mode, int vector);
1419 	int (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
1420 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
1421 	int (*set_identity_map_addr)(struct kvm *kvm, u64 ident_addr);
1422 	u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
1423 
1424 	void (*load_mmu_pgd)(struct kvm_vcpu *vcpu, hpa_t root_hpa,
1425 			     int root_level);
1426 
1427 	bool (*has_wbinvd_exit)(void);
1428 
1429 	u64 (*get_l2_tsc_offset)(struct kvm_vcpu *vcpu);
1430 	u64 (*get_l2_tsc_multiplier)(struct kvm_vcpu *vcpu);
1431 	void (*write_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
1432 	void (*write_tsc_multiplier)(struct kvm_vcpu *vcpu, u64 multiplier);
1433 
1434 	/*
1435 	 * Retrieve somewhat arbitrary exit information.  Intended to
1436 	 * be used only from within tracepoints or error paths.
1437 	 */
1438 	void (*get_exit_info)(struct kvm_vcpu *vcpu, u32 *reason,
1439 			      u64 *info1, u64 *info2,
1440 			      u32 *exit_int_info, u32 *exit_int_info_err_code);
1441 
1442 	int (*check_intercept)(struct kvm_vcpu *vcpu,
1443 			       struct x86_instruction_info *info,
1444 			       enum x86_intercept_stage stage,
1445 			       struct x86_exception *exception);
1446 	void (*handle_exit_irqoff)(struct kvm_vcpu *vcpu);
1447 
1448 	void (*request_immediate_exit)(struct kvm_vcpu *vcpu);
1449 
1450 	void (*sched_in)(struct kvm_vcpu *kvm, int cpu);
1451 
1452 	/*
1453 	 * Size of the CPU's dirty log buffer, i.e. VMX's PML buffer.  A zero
1454 	 * value indicates CPU dirty logging is unsupported or disabled.
1455 	 */
1456 	int cpu_dirty_log_size;
1457 	void (*update_cpu_dirty_logging)(struct kvm_vcpu *vcpu);
1458 
1459 	/* pmu operations of sub-arch */
1460 	const struct kvm_pmu_ops *pmu_ops;
1461 	const struct kvm_x86_nested_ops *nested_ops;
1462 
1463 	void (*vcpu_blocking)(struct kvm_vcpu *vcpu);
1464 	void (*vcpu_unblocking)(struct kvm_vcpu *vcpu);
1465 
1466 	int (*pi_update_irte)(struct kvm *kvm, unsigned int host_irq,
1467 			      uint32_t guest_irq, bool set);
1468 	void (*pi_start_assignment)(struct kvm *kvm);
1469 	void (*apicv_post_state_restore)(struct kvm_vcpu *vcpu);
1470 	bool (*dy_apicv_has_pending_interrupt)(struct kvm_vcpu *vcpu);
1471 
1472 	int (*set_hv_timer)(struct kvm_vcpu *vcpu, u64 guest_deadline_tsc,
1473 			    bool *expired);
1474 	void (*cancel_hv_timer)(struct kvm_vcpu *vcpu);
1475 
1476 	void (*setup_mce)(struct kvm_vcpu *vcpu);
1477 
1478 	int (*smi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1479 	int (*enter_smm)(struct kvm_vcpu *vcpu, char *smstate);
1480 	int (*leave_smm)(struct kvm_vcpu *vcpu, const char *smstate);
1481 	void (*enable_smi_window)(struct kvm_vcpu *vcpu);
1482 
1483 	int (*mem_enc_ioctl)(struct kvm *kvm, void __user *argp);
1484 	int (*mem_enc_register_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1485 	int (*mem_enc_unregister_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1486 	int (*vm_copy_enc_context_from)(struct kvm *kvm, unsigned int source_fd);
1487 	int (*vm_move_enc_context_from)(struct kvm *kvm, unsigned int source_fd);
1488 
1489 	int (*get_msr_feature)(struct kvm_msr_entry *entry);
1490 
1491 	bool (*can_emulate_instruction)(struct kvm_vcpu *vcpu, int emul_type,
1492 					void *insn, int insn_len);
1493 
1494 	bool (*apic_init_signal_blocked)(struct kvm_vcpu *vcpu);
1495 	int (*enable_direct_tlbflush)(struct kvm_vcpu *vcpu);
1496 
1497 	void (*migrate_timers)(struct kvm_vcpu *vcpu);
1498 	void (*msr_filter_changed)(struct kvm_vcpu *vcpu);
1499 	int (*complete_emulated_msr)(struct kvm_vcpu *vcpu, int err);
1500 
1501 	void (*vcpu_deliver_sipi_vector)(struct kvm_vcpu *vcpu, u8 vector);
1502 };
1503 
1504 struct kvm_x86_nested_ops {
1505 	void (*leave_nested)(struct kvm_vcpu *vcpu);
1506 	int (*check_events)(struct kvm_vcpu *vcpu);
1507 	bool (*hv_timer_pending)(struct kvm_vcpu *vcpu);
1508 	void (*triple_fault)(struct kvm_vcpu *vcpu);
1509 	int (*get_state)(struct kvm_vcpu *vcpu,
1510 			 struct kvm_nested_state __user *user_kvm_nested_state,
1511 			 unsigned user_data_size);
1512 	int (*set_state)(struct kvm_vcpu *vcpu,
1513 			 struct kvm_nested_state __user *user_kvm_nested_state,
1514 			 struct kvm_nested_state *kvm_state);
1515 	bool (*get_nested_state_pages)(struct kvm_vcpu *vcpu);
1516 	int (*write_log_dirty)(struct kvm_vcpu *vcpu, gpa_t l2_gpa);
1517 
1518 	int (*enable_evmcs)(struct kvm_vcpu *vcpu,
1519 			    uint16_t *vmcs_version);
1520 	uint16_t (*get_evmcs_version)(struct kvm_vcpu *vcpu);
1521 };
1522 
1523 struct kvm_x86_init_ops {
1524 	int (*cpu_has_kvm_support)(void);
1525 	int (*disabled_by_bios)(void);
1526 	int (*check_processor_compatibility)(void);
1527 	int (*hardware_setup)(void);
1528 	unsigned int (*handle_intel_pt_intr)(void);
1529 
1530 	struct kvm_x86_ops *runtime_ops;
1531 };
1532 
1533 struct kvm_arch_async_pf {
1534 	u32 token;
1535 	gfn_t gfn;
1536 	unsigned long cr3;
1537 	bool direct_map;
1538 };
1539 
1540 extern u32 __read_mostly kvm_nr_uret_msrs;
1541 extern u64 __read_mostly host_efer;
1542 extern bool __read_mostly allow_smaller_maxphyaddr;
1543 extern bool __read_mostly enable_apicv;
1544 extern struct kvm_x86_ops kvm_x86_ops;
1545 
1546 #define KVM_X86_OP(func) \
1547 	DECLARE_STATIC_CALL(kvm_x86_##func, *(((struct kvm_x86_ops *)0)->func));
1548 #define KVM_X86_OP_OPTIONAL KVM_X86_OP
1549 #define KVM_X86_OP_OPTIONAL_RET0 KVM_X86_OP
1550 #include <asm/kvm-x86-ops.h>
1551 
1552 static inline void kvm_ops_static_call_update(void)
1553 {
1554 #define __KVM_X86_OP(func) \
1555 	static_call_update(kvm_x86_##func, kvm_x86_ops.func);
1556 #define KVM_X86_OP(func) \
1557 	WARN_ON(!kvm_x86_ops.func); __KVM_X86_OP(func)
1558 #define KVM_X86_OP_OPTIONAL __KVM_X86_OP
1559 #define KVM_X86_OP_OPTIONAL_RET0(func) \
1560 	static_call_update(kvm_x86_##func, (void *)kvm_x86_ops.func ? : \
1561 					   (void *)__static_call_return0);
1562 #include <asm/kvm-x86-ops.h>
1563 #undef __KVM_X86_OP
1564 }
1565 
1566 #define __KVM_HAVE_ARCH_VM_ALLOC
1567 static inline struct kvm *kvm_arch_alloc_vm(void)
1568 {
1569 	return __vmalloc(kvm_x86_ops.vm_size, GFP_KERNEL_ACCOUNT | __GFP_ZERO);
1570 }
1571 
1572 #define __KVM_HAVE_ARCH_VM_FREE
1573 void kvm_arch_free_vm(struct kvm *kvm);
1574 
1575 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
1576 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
1577 {
1578 	if (kvm_x86_ops.tlb_remote_flush &&
1579 	    !static_call(kvm_x86_tlb_remote_flush)(kvm))
1580 		return 0;
1581 	else
1582 		return -ENOTSUPP;
1583 }
1584 
1585 #define kvm_arch_pmi_in_guest(vcpu) \
1586 	((vcpu) && (vcpu)->arch.handling_intr_from_guest)
1587 
1588 int kvm_mmu_module_init(void);
1589 void kvm_mmu_module_exit(void);
1590 
1591 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
1592 int kvm_mmu_create(struct kvm_vcpu *vcpu);
1593 int kvm_mmu_init_vm(struct kvm *kvm);
1594 void kvm_mmu_uninit_vm(struct kvm *kvm);
1595 
1596 void kvm_mmu_after_set_cpuid(struct kvm_vcpu *vcpu);
1597 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
1598 void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
1599 				      const struct kvm_memory_slot *memslot,
1600 				      int start_level);
1601 void kvm_mmu_slot_try_split_huge_pages(struct kvm *kvm,
1602 				       const struct kvm_memory_slot *memslot,
1603 				       int target_level);
1604 void kvm_mmu_try_split_huge_pages(struct kvm *kvm,
1605 				  const struct kvm_memory_slot *memslot,
1606 				  u64 start, u64 end,
1607 				  int target_level);
1608 void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
1609 				   const struct kvm_memory_slot *memslot);
1610 void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
1611 				   const struct kvm_memory_slot *memslot);
1612 void kvm_mmu_zap_all(struct kvm *kvm);
1613 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, u64 gen);
1614 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned long kvm_nr_mmu_pages);
1615 
1616 int load_pdptrs(struct kvm_vcpu *vcpu, unsigned long cr3);
1617 
1618 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
1619 			  const void *val, int bytes);
1620 
1621 struct kvm_irq_mask_notifier {
1622 	void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
1623 	int irq;
1624 	struct hlist_node link;
1625 };
1626 
1627 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
1628 				    struct kvm_irq_mask_notifier *kimn);
1629 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
1630 				      struct kvm_irq_mask_notifier *kimn);
1631 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
1632 			     bool mask);
1633 
1634 extern bool tdp_enabled;
1635 
1636 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
1637 
1638 /* control of guest tsc rate supported? */
1639 extern bool kvm_has_tsc_control;
1640 /* maximum supported tsc_khz for guests */
1641 extern u32  kvm_max_guest_tsc_khz;
1642 /* number of bits of the fractional part of the TSC scaling ratio */
1643 extern u8   kvm_tsc_scaling_ratio_frac_bits;
1644 /* maximum allowed value of TSC scaling ratio */
1645 extern u64  kvm_max_tsc_scaling_ratio;
1646 /* 1ull << kvm_tsc_scaling_ratio_frac_bits */
1647 extern u64  kvm_default_tsc_scaling_ratio;
1648 /* bus lock detection supported? */
1649 extern bool kvm_has_bus_lock_exit;
1650 
1651 extern u64 kvm_mce_cap_supported;
1652 
1653 /*
1654  * EMULTYPE_NO_DECODE - Set when re-emulating an instruction (after completing
1655  *			userspace I/O) to indicate that the emulation context
1656  *			should be reused as is, i.e. skip initialization of
1657  *			emulation context, instruction fetch and decode.
1658  *
1659  * EMULTYPE_TRAP_UD - Set when emulating an intercepted #UD from hardware.
1660  *		      Indicates that only select instructions (tagged with
1661  *		      EmulateOnUD) should be emulated (to minimize the emulator
1662  *		      attack surface).  See also EMULTYPE_TRAP_UD_FORCED.
1663  *
1664  * EMULTYPE_SKIP - Set when emulating solely to skip an instruction, i.e. to
1665  *		   decode the instruction length.  For use *only* by
1666  *		   kvm_x86_ops.skip_emulated_instruction() implementations if
1667  *		   EMULTYPE_COMPLETE_USER_EXIT is not set.
1668  *
1669  * EMULTYPE_ALLOW_RETRY_PF - Set when the emulator should resume the guest to
1670  *			     retry native execution under certain conditions,
1671  *			     Can only be set in conjunction with EMULTYPE_PF.
1672  *
1673  * EMULTYPE_TRAP_UD_FORCED - Set when emulating an intercepted #UD that was
1674  *			     triggered by KVM's magic "force emulation" prefix,
1675  *			     which is opt in via module param (off by default).
1676  *			     Bypasses EmulateOnUD restriction despite emulating
1677  *			     due to an intercepted #UD (see EMULTYPE_TRAP_UD).
1678  *			     Used to test the full emulator from userspace.
1679  *
1680  * EMULTYPE_VMWARE_GP - Set when emulating an intercepted #GP for VMware
1681  *			backdoor emulation, which is opt in via module param.
1682  *			VMware backdoor emulation handles select instructions
1683  *			and reinjects the #GP for all other cases.
1684  *
1685  * EMULTYPE_PF - Set when emulating MMIO by way of an intercepted #PF, in which
1686  *		 case the CR2/GPA value pass on the stack is valid.
1687  *
1688  * EMULTYPE_COMPLETE_USER_EXIT - Set when the emulator should update interruptibility
1689  *				 state and inject single-step #DBs after skipping
1690  *				 an instruction (after completing userspace I/O).
1691  */
1692 #define EMULTYPE_NO_DECODE	    (1 << 0)
1693 #define EMULTYPE_TRAP_UD	    (1 << 1)
1694 #define EMULTYPE_SKIP		    (1 << 2)
1695 #define EMULTYPE_ALLOW_RETRY_PF	    (1 << 3)
1696 #define EMULTYPE_TRAP_UD_FORCED	    (1 << 4)
1697 #define EMULTYPE_VMWARE_GP	    (1 << 5)
1698 #define EMULTYPE_PF		    (1 << 6)
1699 #define EMULTYPE_COMPLETE_USER_EXIT (1 << 7)
1700 
1701 int kvm_emulate_instruction(struct kvm_vcpu *vcpu, int emulation_type);
1702 int kvm_emulate_instruction_from_buffer(struct kvm_vcpu *vcpu,
1703 					void *insn, int insn_len);
1704 void __kvm_prepare_emulation_failure_exit(struct kvm_vcpu *vcpu,
1705 					  u64 *data, u8 ndata);
1706 void kvm_prepare_emulation_failure_exit(struct kvm_vcpu *vcpu);
1707 
1708 void kvm_enable_efer_bits(u64);
1709 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
1710 int __kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data, bool host_initiated);
1711 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data);
1712 int kvm_set_msr(struct kvm_vcpu *vcpu, u32 index, u64 data);
1713 int kvm_emulate_rdmsr(struct kvm_vcpu *vcpu);
1714 int kvm_emulate_wrmsr(struct kvm_vcpu *vcpu);
1715 int kvm_emulate_as_nop(struct kvm_vcpu *vcpu);
1716 int kvm_emulate_invd(struct kvm_vcpu *vcpu);
1717 int kvm_emulate_mwait(struct kvm_vcpu *vcpu);
1718 int kvm_handle_invalid_op(struct kvm_vcpu *vcpu);
1719 int kvm_emulate_monitor(struct kvm_vcpu *vcpu);
1720 
1721 int kvm_fast_pio(struct kvm_vcpu *vcpu, int size, unsigned short port, int in);
1722 int kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
1723 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
1724 int kvm_emulate_halt_noskip(struct kvm_vcpu *vcpu);
1725 int kvm_emulate_ap_reset_hold(struct kvm_vcpu *vcpu);
1726 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
1727 
1728 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
1729 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
1730 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector);
1731 
1732 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
1733 		    int reason, bool has_error_code, u32 error_code);
1734 
1735 void kvm_post_set_cr0(struct kvm_vcpu *vcpu, unsigned long old_cr0, unsigned long cr0);
1736 void kvm_post_set_cr4(struct kvm_vcpu *vcpu, unsigned long old_cr4, unsigned long cr4);
1737 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
1738 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
1739 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
1740 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
1741 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
1742 void kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
1743 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
1744 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
1745 int kvm_emulate_xsetbv(struct kvm_vcpu *vcpu);
1746 
1747 int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1748 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1749 
1750 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
1751 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
1752 int kvm_emulate_rdpmc(struct kvm_vcpu *vcpu);
1753 
1754 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1755 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1756 void kvm_queue_exception_p(struct kvm_vcpu *vcpu, unsigned nr, unsigned long payload);
1757 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1758 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1759 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
1760 bool kvm_inject_emulated_page_fault(struct kvm_vcpu *vcpu,
1761 				    struct x86_exception *fault);
1762 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
1763 bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr);
1764 
1765 static inline int __kvm_irq_line_state(unsigned long *irq_state,
1766 				       int irq_source_id, int level)
1767 {
1768 	/* Logical OR for level trig interrupt */
1769 	if (level)
1770 		__set_bit(irq_source_id, irq_state);
1771 	else
1772 		__clear_bit(irq_source_id, irq_state);
1773 
1774 	return !!(*irq_state);
1775 }
1776 
1777 #define KVM_MMU_ROOT_CURRENT		BIT(0)
1778 #define KVM_MMU_ROOT_PREVIOUS(i)	BIT(1+i)
1779 #define KVM_MMU_ROOTS_ALL		(~0UL)
1780 
1781 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
1782 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
1783 
1784 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
1785 
1786 void kvm_update_dr7(struct kvm_vcpu *vcpu);
1787 
1788 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
1789 void kvm_mmu_free_roots(struct kvm *kvm, struct kvm_mmu *mmu,
1790 			ulong roots_to_free);
1791 void kvm_mmu_free_guest_mode_roots(struct kvm *kvm, struct kvm_mmu *mmu);
1792 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
1793 			      struct x86_exception *exception);
1794 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
1795 			       struct x86_exception *exception);
1796 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
1797 			       struct x86_exception *exception);
1798 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
1799 				struct x86_exception *exception);
1800 
1801 bool kvm_apicv_activated(struct kvm *kvm);
1802 void kvm_vcpu_update_apicv(struct kvm_vcpu *vcpu);
1803 void __kvm_set_or_clear_apicv_inhibit(struct kvm *kvm,
1804 				      enum kvm_apicv_inhibit reason, bool set);
1805 void kvm_set_or_clear_apicv_inhibit(struct kvm *kvm,
1806 				    enum kvm_apicv_inhibit reason, bool set);
1807 
1808 static inline void kvm_set_apicv_inhibit(struct kvm *kvm,
1809 					 enum kvm_apicv_inhibit reason)
1810 {
1811 	kvm_set_or_clear_apicv_inhibit(kvm, reason, true);
1812 }
1813 
1814 static inline void kvm_clear_apicv_inhibit(struct kvm *kvm,
1815 					   enum kvm_apicv_inhibit reason)
1816 {
1817 	kvm_set_or_clear_apicv_inhibit(kvm, reason, false);
1818 }
1819 
1820 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
1821 
1822 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u64 error_code,
1823 		       void *insn, int insn_len);
1824 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
1825 void kvm_mmu_invalidate_gva(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1826 			    gva_t gva, hpa_t root_hpa);
1827 void kvm_mmu_invpcid_gva(struct kvm_vcpu *vcpu, gva_t gva, unsigned long pcid);
1828 void kvm_mmu_new_pgd(struct kvm_vcpu *vcpu, gpa_t new_pgd);
1829 
1830 void kvm_configure_mmu(bool enable_tdp, int tdp_forced_root_level,
1831 		       int tdp_max_root_level, int tdp_huge_page_level);
1832 
1833 static inline u16 kvm_read_ldt(void)
1834 {
1835 	u16 ldt;
1836 	asm("sldt %0" : "=g"(ldt));
1837 	return ldt;
1838 }
1839 
1840 static inline void kvm_load_ldt(u16 sel)
1841 {
1842 	asm("lldt %0" : : "rm"(sel));
1843 }
1844 
1845 #ifdef CONFIG_X86_64
1846 static inline unsigned long read_msr(unsigned long msr)
1847 {
1848 	u64 value;
1849 
1850 	rdmsrl(msr, value);
1851 	return value;
1852 }
1853 #endif
1854 
1855 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
1856 {
1857 	kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
1858 }
1859 
1860 #define TSS_IOPB_BASE_OFFSET 0x66
1861 #define TSS_BASE_SIZE 0x68
1862 #define TSS_IOPB_SIZE (65536 / 8)
1863 #define TSS_REDIRECTION_SIZE (256 / 8)
1864 #define RMODE_TSS_SIZE							\
1865 	(TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
1866 
1867 enum {
1868 	TASK_SWITCH_CALL = 0,
1869 	TASK_SWITCH_IRET = 1,
1870 	TASK_SWITCH_JMP = 2,
1871 	TASK_SWITCH_GATE = 3,
1872 };
1873 
1874 #define HF_GIF_MASK		(1 << 0)
1875 #define HF_NMI_MASK		(1 << 3)
1876 #define HF_IRET_MASK		(1 << 4)
1877 #define HF_GUEST_MASK		(1 << 5) /* VCPU is in guest-mode */
1878 #define HF_SMM_MASK		(1 << 6)
1879 #define HF_SMM_INSIDE_NMI_MASK	(1 << 7)
1880 
1881 #define __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
1882 #define KVM_ADDRESS_SPACE_NUM 2
1883 
1884 #define kvm_arch_vcpu_memslots_id(vcpu) ((vcpu)->arch.hflags & HF_SMM_MASK ? 1 : 0)
1885 #define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, (role).smm)
1886 
1887 #define KVM_ARCH_WANT_MMU_NOTIFIER
1888 
1889 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1890 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1891 int kvm_cpu_has_extint(struct kvm_vcpu *v);
1892 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1893 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1894 void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event);
1895 
1896 int kvm_pv_send_ipi(struct kvm *kvm, unsigned long ipi_bitmap_low,
1897 		    unsigned long ipi_bitmap_high, u32 min,
1898 		    unsigned long icr, int op_64_bit);
1899 
1900 int kvm_add_user_return_msr(u32 msr);
1901 int kvm_find_user_return_msr(u32 msr);
1902 int kvm_set_user_return_msr(unsigned index, u64 val, u64 mask);
1903 
1904 static inline bool kvm_is_supported_user_return_msr(u32 msr)
1905 {
1906 	return kvm_find_user_return_msr(msr) >= 0;
1907 }
1908 
1909 u64 kvm_scale_tsc(u64 tsc, u64 ratio);
1910 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc);
1911 u64 kvm_calc_nested_tsc_offset(u64 l1_offset, u64 l2_offset, u64 l2_multiplier);
1912 u64 kvm_calc_nested_tsc_multiplier(u64 l1_multiplier, u64 l2_multiplier);
1913 
1914 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu);
1915 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1916 
1917 void kvm_make_scan_ioapic_request(struct kvm *kvm);
1918 void kvm_make_scan_ioapic_request_mask(struct kvm *kvm,
1919 				       unsigned long *vcpu_bitmap);
1920 
1921 bool kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1922 				     struct kvm_async_pf *work);
1923 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1924 				 struct kvm_async_pf *work);
1925 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1926 			       struct kvm_async_pf *work);
1927 void kvm_arch_async_page_present_queued(struct kvm_vcpu *vcpu);
1928 bool kvm_arch_can_dequeue_async_page_present(struct kvm_vcpu *vcpu);
1929 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1930 
1931 int kvm_skip_emulated_instruction(struct kvm_vcpu *vcpu);
1932 int kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1933 void __kvm_request_immediate_exit(struct kvm_vcpu *vcpu);
1934 
1935 void __user *__x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa,
1936 				     u32 size);
1937 bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu);
1938 bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu);
1939 
1940 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
1941 			     struct kvm_vcpu **dest_vcpu);
1942 
1943 void kvm_set_msi_irq(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e,
1944 		     struct kvm_lapic_irq *irq);
1945 
1946 static inline bool kvm_irq_is_postable(struct kvm_lapic_irq *irq)
1947 {
1948 	/* We can only post Fixed and LowPrio IRQs */
1949 	return (irq->delivery_mode == APIC_DM_FIXED ||
1950 		irq->delivery_mode == APIC_DM_LOWEST);
1951 }
1952 
1953 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu)
1954 {
1955 	static_call_cond(kvm_x86_vcpu_blocking)(vcpu);
1956 }
1957 
1958 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu)
1959 {
1960 	static_call_cond(kvm_x86_vcpu_unblocking)(vcpu);
1961 }
1962 
1963 static inline int kvm_cpu_get_apicid(int mps_cpu)
1964 {
1965 #ifdef CONFIG_X86_LOCAL_APIC
1966 	return default_cpu_present_to_apicid(mps_cpu);
1967 #else
1968 	WARN_ON_ONCE(1);
1969 	return BAD_APICID;
1970 #endif
1971 }
1972 
1973 #define put_smstate(type, buf, offset, val)                      \
1974 	*(type *)((buf) + (offset) - 0x7e00) = val
1975 
1976 #define GET_SMSTATE(type, buf, offset)		\
1977 	(*(type *)((buf) + (offset) - 0x7e00))
1978 
1979 int kvm_cpu_dirty_log_size(void);
1980 
1981 int memslot_rmap_alloc(struct kvm_memory_slot *slot, unsigned long npages);
1982 
1983 #define KVM_CLOCK_VALID_FLAGS						\
1984 	(KVM_CLOCK_TSC_STABLE | KVM_CLOCK_REALTIME | KVM_CLOCK_HOST_TSC)
1985 
1986 #define KVM_X86_VALID_QUIRKS			\
1987 	(KVM_X86_QUIRK_LINT0_REENABLED |	\
1988 	 KVM_X86_QUIRK_CD_NW_CLEARED |		\
1989 	 KVM_X86_QUIRK_LAPIC_MMIO_HOLE |	\
1990 	 KVM_X86_QUIRK_OUT_7E_INC_RIP |		\
1991 	 KVM_X86_QUIRK_MISC_ENABLE_NO_MWAIT)
1992 
1993 #endif /* _ASM_X86_KVM_HOST_H */
1994