xref: /openbmc/linux/arch/x86/include/asm/kvm_host.h (revision 22a41e9a5044bf3519f05b4a00e99af34bfeb40c)
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 
253 #define PFERR_PRESENT_MASK (1U << PFERR_PRESENT_BIT)
254 #define PFERR_WRITE_MASK (1U << PFERR_WRITE_BIT)
255 #define PFERR_USER_MASK (1U << PFERR_USER_BIT)
256 #define PFERR_RSVD_MASK (1U << PFERR_RSVD_BIT)
257 #define PFERR_FETCH_MASK (1U << PFERR_FETCH_BIT)
258 #define PFERR_PK_MASK (1U << PFERR_PK_BIT)
259 #define PFERR_SGX_MASK (1U << PFERR_SGX_BIT)
260 #define PFERR_GUEST_FINAL_MASK (1ULL << PFERR_GUEST_FINAL_BIT)
261 #define PFERR_GUEST_PAGE_MASK (1ULL << PFERR_GUEST_PAGE_BIT)
262 
263 #define PFERR_NESTED_GUEST_PAGE (PFERR_GUEST_PAGE_MASK |	\
264 				 PFERR_WRITE_MASK |		\
265 				 PFERR_PRESENT_MASK)
266 
267 /* apic attention bits */
268 #define KVM_APIC_CHECK_VAPIC	0
269 /*
270  * The following bit is set with PV-EOI, unset on EOI.
271  * We detect PV-EOI changes by guest by comparing
272  * this bit with PV-EOI in guest memory.
273  * See the implementation in apic_update_pv_eoi.
274  */
275 #define KVM_APIC_PV_EOI_PENDING	1
276 
277 struct kvm_kernel_irq_routing_entry;
278 
279 /*
280  * kvm_mmu_page_role tracks the properties of a shadow page (where shadow page
281  * also includes TDP pages) to determine whether or not a page can be used in
282  * the given MMU context.  This is a subset of the overall kvm_mmu_role to
283  * minimize the size of kvm_memory_slot.arch.gfn_track, i.e. allows allocating
284  * 2 bytes per gfn instead of 4 bytes per gfn.
285  *
286  * Indirect upper-level shadow pages are tracked for write-protection via
287  * gfn_track.  As above, gfn_track is a 16 bit counter, so KVM must not create
288  * more than 2^16-1 upper-level shadow pages at a single gfn, otherwise
289  * gfn_track will overflow and explosions will ensure.
290  *
291  * A unique shadow page (SP) for a gfn is created if and only if an existing SP
292  * cannot be reused.  The ability to reuse a SP is tracked by its role, which
293  * incorporates various mode bits and properties of the SP.  Roughly speaking,
294  * the number of unique SPs that can theoretically be created is 2^n, where n
295  * is the number of bits that are used to compute the role.
296  *
297  * But, even though there are 19 bits in the mask below, not all combinations
298  * of modes and flags are possible:
299  *
300  *   - invalid shadow pages are not accounted, so the bits are effectively 18
301  *
302  *   - quadrant will only be used if has_4_byte_gpte=1 (non-PAE paging);
303  *     execonly and ad_disabled are only used for nested EPT which has
304  *     has_4_byte_gpte=0.  Therefore, 2 bits are always unused.
305  *
306  *   - the 4 bits of level are effectively limited to the values 2/3/4/5,
307  *     as 4k SPs are not tracked (allowed to go unsync).  In addition non-PAE
308  *     paging has exactly one upper level, making level completely redundant
309  *     when has_4_byte_gpte=1.
310  *
311  *   - on top of this, smep_andnot_wp and smap_andnot_wp are only set if
312  *     cr0_wp=0, therefore these three bits only give rise to 5 possibilities.
313  *
314  * Therefore, the maximum number of possible upper-level shadow pages for a
315  * single gfn is a bit less than 2^13.
316  */
317 union kvm_mmu_page_role {
318 	u32 word;
319 	struct {
320 		unsigned level:4;
321 		unsigned has_4_byte_gpte:1;
322 		unsigned quadrant:2;
323 		unsigned direct:1;
324 		unsigned access:3;
325 		unsigned invalid:1;
326 		unsigned efer_nx:1;
327 		unsigned cr0_wp:1;
328 		unsigned smep_andnot_wp:1;
329 		unsigned smap_andnot_wp:1;
330 		unsigned ad_disabled:1;
331 		unsigned guest_mode:1;
332 		unsigned :6;
333 
334 		/*
335 		 * This is left at the top of the word so that
336 		 * kvm_memslots_for_spte_role can extract it with a
337 		 * simple shift.  While there is room, give it a whole
338 		 * byte so it is also faster to load it from memory.
339 		 */
340 		unsigned smm:8;
341 	};
342 };
343 
344 /*
345  * kvm_mmu_extended_role complements kvm_mmu_page_role, tracking properties
346  * relevant to the current MMU configuration.   When loading CR0, CR4, or EFER,
347  * including on nested transitions, if nothing in the full role changes then
348  * MMU re-configuration can be skipped. @valid bit is set on first usage so we
349  * don't treat all-zero structure as valid data.
350  *
351  * The properties that are tracked in the extended role but not the page role
352  * are for things that either (a) do not affect the validity of the shadow page
353  * or (b) are indirectly reflected in the shadow page's role.  For example,
354  * CR4.PKE only affects permission checks for software walks of the guest page
355  * tables (because KVM doesn't support Protection Keys with shadow paging), and
356  * CR0.PG, CR4.PAE, and CR4.PSE are indirectly reflected in role.level.
357  *
358  * Note, SMEP and SMAP are not redundant with sm*p_andnot_wp in the page role.
359  * If CR0.WP=1, KVM can reuse shadow pages for the guest regardless of SMEP and
360  * SMAP, but the MMU's permission checks for software walks need to be SMEP and
361  * SMAP aware regardless of CR0.WP.
362  */
363 union kvm_mmu_extended_role {
364 	u32 word;
365 	struct {
366 		unsigned int valid:1;
367 		unsigned int execonly:1;
368 		unsigned int cr0_pg:1;
369 		unsigned int cr4_pae:1;
370 		unsigned int cr4_pse:1;
371 		unsigned int cr4_pke:1;
372 		unsigned int cr4_smap:1;
373 		unsigned int cr4_smep:1;
374 		unsigned int cr4_la57:1;
375 		unsigned int efer_lma:1;
376 	};
377 };
378 
379 union kvm_mmu_role {
380 	u64 as_u64;
381 	struct {
382 		union kvm_mmu_page_role base;
383 		union kvm_mmu_extended_role ext;
384 	};
385 };
386 
387 struct kvm_rmap_head {
388 	unsigned long val;
389 };
390 
391 struct kvm_pio_request {
392 	unsigned long linear_rip;
393 	unsigned long count;
394 	int in;
395 	int port;
396 	int size;
397 };
398 
399 #define PT64_ROOT_MAX_LEVEL 5
400 
401 struct rsvd_bits_validate {
402 	u64 rsvd_bits_mask[2][PT64_ROOT_MAX_LEVEL];
403 	u64 bad_mt_xwr;
404 };
405 
406 struct kvm_mmu_root_info {
407 	gpa_t pgd;
408 	hpa_t hpa;
409 };
410 
411 #define KVM_MMU_ROOT_INFO_INVALID \
412 	((struct kvm_mmu_root_info) { .pgd = INVALID_PAGE, .hpa = INVALID_PAGE })
413 
414 #define KVM_MMU_NUM_PREV_ROOTS 3
415 
416 #define KVM_HAVE_MMU_RWLOCK
417 
418 struct kvm_mmu_page;
419 struct kvm_page_fault;
420 
421 /*
422  * x86 supports 4 paging modes (5-level 64-bit, 4-level 64-bit, 3-level 32-bit,
423  * and 2-level 32-bit).  The kvm_mmu structure abstracts the details of the
424  * current mmu mode.
425  */
426 struct kvm_mmu {
427 	unsigned long (*get_guest_pgd)(struct kvm_vcpu *vcpu);
428 	u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
429 	int (*page_fault)(struct kvm_vcpu *vcpu, struct kvm_page_fault *fault);
430 	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
431 				  struct x86_exception *fault);
432 	gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
433 			    gpa_t gva_or_gpa, u32 access,
434 			    struct x86_exception *exception);
435 	int (*sync_page)(struct kvm_vcpu *vcpu,
436 			 struct kvm_mmu_page *sp);
437 	void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva, hpa_t root_hpa);
438 	struct kvm_mmu_root_info root;
439 	union kvm_mmu_role mmu_role;
440 	u8 root_level;
441 	u8 shadow_root_level;
442 	u8 ept_ad;
443 	bool direct_map;
444 	struct kvm_mmu_root_info prev_roots[KVM_MMU_NUM_PREV_ROOTS];
445 
446 	/*
447 	 * Bitmap; bit set = permission fault
448 	 * Byte index: page fault error code [4:1]
449 	 * Bit index: pte permissions in ACC_* format
450 	 */
451 	u8 permissions[16];
452 
453 	/*
454 	* The pkru_mask indicates if protection key checks are needed.  It
455 	* consists of 16 domains indexed by page fault error code bits [4:1],
456 	* with PFEC.RSVD replaced by ACC_USER_MASK from the page tables.
457 	* Each domain has 2 bits which are ANDed with AD and WD from PKRU.
458 	*/
459 	u32 pkru_mask;
460 
461 	u64 *pae_root;
462 	u64 *pml4_root;
463 	u64 *pml5_root;
464 
465 	/*
466 	 * check zero bits on shadow page table entries, these
467 	 * bits include not only hardware reserved bits but also
468 	 * the bits spte never used.
469 	 */
470 	struct rsvd_bits_validate shadow_zero_check;
471 
472 	struct rsvd_bits_validate guest_rsvd_check;
473 
474 	u64 pdptrs[4]; /* pae */
475 };
476 
477 struct kvm_tlb_range {
478 	u64 start_gfn;
479 	u64 pages;
480 };
481 
482 enum pmc_type {
483 	KVM_PMC_GP = 0,
484 	KVM_PMC_FIXED,
485 };
486 
487 struct kvm_pmc {
488 	enum pmc_type type;
489 	u8 idx;
490 	u64 counter;
491 	u64 eventsel;
492 	struct perf_event *perf_event;
493 	struct kvm_vcpu *vcpu;
494 	/*
495 	 * eventsel value for general purpose counters,
496 	 * ctrl value for fixed counters.
497 	 */
498 	u64 current_config;
499 	bool is_paused;
500 	bool intr;
501 };
502 
503 #define KVM_PMC_MAX_FIXED	3
504 struct kvm_pmu {
505 	unsigned nr_arch_gp_counters;
506 	unsigned nr_arch_fixed_counters;
507 	unsigned available_event_types;
508 	u64 fixed_ctr_ctrl;
509 	u64 global_ctrl;
510 	u64 global_status;
511 	u64 counter_bitmask[2];
512 	u64 global_ctrl_mask;
513 	u64 global_ovf_ctrl_mask;
514 	u64 reserved_bits;
515 	u8 version;
516 	struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
517 	struct kvm_pmc fixed_counters[KVM_PMC_MAX_FIXED];
518 	struct irq_work irq_work;
519 	DECLARE_BITMAP(reprogram_pmi, X86_PMC_IDX_MAX);
520 	DECLARE_BITMAP(all_valid_pmc_idx, X86_PMC_IDX_MAX);
521 	DECLARE_BITMAP(pmc_in_use, X86_PMC_IDX_MAX);
522 
523 	/*
524 	 * The gate to release perf_events not marked in
525 	 * pmc_in_use only once in a vcpu time slice.
526 	 */
527 	bool need_cleanup;
528 
529 	/*
530 	 * The total number of programmed perf_events and it helps to avoid
531 	 * redundant check before cleanup if guest don't use vPMU at all.
532 	 */
533 	u8 event_count;
534 };
535 
536 struct kvm_pmu_ops;
537 
538 enum {
539 	KVM_DEBUGREG_BP_ENABLED = 1,
540 	KVM_DEBUGREG_WONT_EXIT = 2,
541 };
542 
543 struct kvm_mtrr_range {
544 	u64 base;
545 	u64 mask;
546 	struct list_head node;
547 };
548 
549 struct kvm_mtrr {
550 	struct kvm_mtrr_range var_ranges[KVM_NR_VAR_MTRR];
551 	mtrr_type fixed_ranges[KVM_NR_FIXED_MTRR_REGION];
552 	u64 deftype;
553 
554 	struct list_head head;
555 };
556 
557 /* Hyper-V SynIC timer */
558 struct kvm_vcpu_hv_stimer {
559 	struct hrtimer timer;
560 	int index;
561 	union hv_stimer_config config;
562 	u64 count;
563 	u64 exp_time;
564 	struct hv_message msg;
565 	bool msg_pending;
566 };
567 
568 /* Hyper-V synthetic interrupt controller (SynIC)*/
569 struct kvm_vcpu_hv_synic {
570 	u64 version;
571 	u64 control;
572 	u64 msg_page;
573 	u64 evt_page;
574 	atomic64_t sint[HV_SYNIC_SINT_COUNT];
575 	atomic_t sint_to_gsi[HV_SYNIC_SINT_COUNT];
576 	DECLARE_BITMAP(auto_eoi_bitmap, 256);
577 	DECLARE_BITMAP(vec_bitmap, 256);
578 	bool active;
579 	bool dont_zero_synic_pages;
580 };
581 
582 /* Hyper-V per vcpu emulation context */
583 struct kvm_vcpu_hv {
584 	struct kvm_vcpu *vcpu;
585 	u32 vp_index;
586 	u64 hv_vapic;
587 	s64 runtime_offset;
588 	struct kvm_vcpu_hv_synic synic;
589 	struct kvm_hyperv_exit exit;
590 	struct kvm_vcpu_hv_stimer stimer[HV_SYNIC_STIMER_COUNT];
591 	DECLARE_BITMAP(stimer_pending_bitmap, HV_SYNIC_STIMER_COUNT);
592 	bool enforce_cpuid;
593 	struct {
594 		u32 features_eax; /* HYPERV_CPUID_FEATURES.EAX */
595 		u32 features_ebx; /* HYPERV_CPUID_FEATURES.EBX */
596 		u32 features_edx; /* HYPERV_CPUID_FEATURES.EDX */
597 		u32 enlightenments_eax; /* HYPERV_CPUID_ENLIGHTMENT_INFO.EAX */
598 		u32 enlightenments_ebx; /* HYPERV_CPUID_ENLIGHTMENT_INFO.EBX */
599 		u32 syndbg_cap_eax; /* HYPERV_CPUID_SYNDBG_PLATFORM_CAPABILITIES.EAX */
600 	} cpuid_cache;
601 };
602 
603 /* Xen HVM per vcpu emulation context */
604 struct kvm_vcpu_xen {
605 	u64 hypercall_rip;
606 	u32 current_runstate;
607 	bool vcpu_info_set;
608 	bool vcpu_time_info_set;
609 	bool runstate_set;
610 	struct gfn_to_hva_cache vcpu_info_cache;
611 	struct gfn_to_hva_cache vcpu_time_info_cache;
612 	struct gfn_to_hva_cache runstate_cache;
613 	u64 last_steal;
614 	u64 runstate_entry_time;
615 	u64 runstate_times[4];
616 	unsigned long evtchn_pending_sel;
617 };
618 
619 struct kvm_vcpu_arch {
620 	/*
621 	 * rip and regs accesses must go through
622 	 * kvm_{register,rip}_{read,write} functions.
623 	 */
624 	unsigned long regs[NR_VCPU_REGS];
625 	u32 regs_avail;
626 	u32 regs_dirty;
627 
628 	unsigned long cr0;
629 	unsigned long cr0_guest_owned_bits;
630 	unsigned long cr2;
631 	unsigned long cr3;
632 	unsigned long cr4;
633 	unsigned long cr4_guest_owned_bits;
634 	unsigned long cr4_guest_rsvd_bits;
635 	unsigned long cr8;
636 	u32 host_pkru;
637 	u32 pkru;
638 	u32 hflags;
639 	u64 efer;
640 	u64 apic_base;
641 	struct kvm_lapic *apic;    /* kernel irqchip context */
642 	bool apicv_active;
643 	bool load_eoi_exitmap_pending;
644 	DECLARE_BITMAP(ioapic_handled_vectors, 256);
645 	unsigned long apic_attention;
646 	int32_t apic_arb_prio;
647 	int mp_state;
648 	u64 ia32_misc_enable_msr;
649 	u64 smbase;
650 	u64 smi_count;
651 	bool tpr_access_reporting;
652 	bool xsaves_enabled;
653 	bool xfd_no_write_intercept;
654 	u64 ia32_xss;
655 	u64 microcode_version;
656 	u64 arch_capabilities;
657 	u64 perf_capabilities;
658 
659 	/*
660 	 * Paging state of the vcpu
661 	 *
662 	 * If the vcpu runs in guest mode with two level paging this still saves
663 	 * the paging mode of the l1 guest. This context is always used to
664 	 * handle faults.
665 	 */
666 	struct kvm_mmu *mmu;
667 
668 	/* Non-nested MMU for L1 */
669 	struct kvm_mmu root_mmu;
670 
671 	/* L1 MMU when running nested */
672 	struct kvm_mmu guest_mmu;
673 
674 	/*
675 	 * Paging state of an L2 guest (used for nested npt)
676 	 *
677 	 * This context will save all necessary information to walk page tables
678 	 * of an L2 guest. This context is only initialized for page table
679 	 * walking and not for faulting since we never handle l2 page faults on
680 	 * the host.
681 	 */
682 	struct kvm_mmu nested_mmu;
683 
684 	/*
685 	 * Pointer to the mmu context currently used for
686 	 * gva_to_gpa translations.
687 	 */
688 	struct kvm_mmu *walk_mmu;
689 
690 	struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
691 	struct kvm_mmu_memory_cache mmu_shadow_page_cache;
692 	struct kvm_mmu_memory_cache mmu_gfn_array_cache;
693 	struct kvm_mmu_memory_cache mmu_page_header_cache;
694 
695 	/*
696 	 * QEMU userspace and the guest each have their own FPU state.
697 	 * In vcpu_run, we switch between the user and guest FPU contexts.
698 	 * While running a VCPU, the VCPU thread will have the guest FPU
699 	 * context.
700 	 *
701 	 * Note that while the PKRU state lives inside the fpu registers,
702 	 * it is switched out separately at VMENTER and VMEXIT time. The
703 	 * "guest_fpstate" state here contains the guest FPU context, with the
704 	 * host PRKU bits.
705 	 */
706 	struct fpu_guest guest_fpu;
707 
708 	u64 xcr0;
709 
710 	struct kvm_pio_request pio;
711 	void *pio_data;
712 	void *sev_pio_data;
713 	unsigned sev_pio_count;
714 
715 	u8 event_exit_inst_len;
716 
717 	struct kvm_queued_exception {
718 		bool pending;
719 		bool injected;
720 		bool has_error_code;
721 		u8 nr;
722 		u32 error_code;
723 		unsigned long payload;
724 		bool has_payload;
725 		u8 nested_apf;
726 	} exception;
727 
728 	struct kvm_queued_interrupt {
729 		bool injected;
730 		bool soft;
731 		u8 nr;
732 	} interrupt;
733 
734 	int halt_request; /* real mode on Intel only */
735 
736 	int cpuid_nent;
737 	struct kvm_cpuid_entry2 *cpuid_entries;
738 	u32 kvm_cpuid_base;
739 
740 	u64 reserved_gpa_bits;
741 	int maxphyaddr;
742 
743 	/* emulate context */
744 
745 	struct x86_emulate_ctxt *emulate_ctxt;
746 	bool emulate_regs_need_sync_to_vcpu;
747 	bool emulate_regs_need_sync_from_vcpu;
748 	int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
749 
750 	gpa_t time;
751 	struct pvclock_vcpu_time_info hv_clock;
752 	unsigned int hw_tsc_khz;
753 	struct gfn_to_hva_cache pv_time;
754 	bool pv_time_enabled;
755 	/* set guest stopped flag in pvclock flags field */
756 	bool pvclock_set_guest_stopped_request;
757 
758 	struct {
759 		u8 preempted;
760 		u64 msr_val;
761 		u64 last_steal;
762 		struct gfn_to_hva_cache cache;
763 	} st;
764 
765 	u64 l1_tsc_offset;
766 	u64 tsc_offset; /* current tsc offset */
767 	u64 last_guest_tsc;
768 	u64 last_host_tsc;
769 	u64 tsc_offset_adjustment;
770 	u64 this_tsc_nsec;
771 	u64 this_tsc_write;
772 	u64 this_tsc_generation;
773 	bool tsc_catchup;
774 	bool tsc_always_catchup;
775 	s8 virtual_tsc_shift;
776 	u32 virtual_tsc_mult;
777 	u32 virtual_tsc_khz;
778 	s64 ia32_tsc_adjust_msr;
779 	u64 msr_ia32_power_ctl;
780 	u64 l1_tsc_scaling_ratio;
781 	u64 tsc_scaling_ratio; /* current scaling ratio */
782 
783 	atomic_t nmi_queued;  /* unprocessed asynchronous NMIs */
784 	unsigned nmi_pending; /* NMI queued after currently running handler */
785 	bool nmi_injected;    /* Trying to inject an NMI this entry */
786 	bool smi_pending;    /* SMI queued after currently running handler */
787 	u8 handling_intr_from_guest;
788 
789 	struct kvm_mtrr mtrr_state;
790 	u64 pat;
791 
792 	unsigned switch_db_regs;
793 	unsigned long db[KVM_NR_DB_REGS];
794 	unsigned long dr6;
795 	unsigned long dr7;
796 	unsigned long eff_db[KVM_NR_DB_REGS];
797 	unsigned long guest_debug_dr7;
798 	u64 msr_platform_info;
799 	u64 msr_misc_features_enables;
800 
801 	u64 mcg_cap;
802 	u64 mcg_status;
803 	u64 mcg_ctl;
804 	u64 mcg_ext_ctl;
805 	u64 *mce_banks;
806 
807 	/* Cache MMIO info */
808 	u64 mmio_gva;
809 	unsigned mmio_access;
810 	gfn_t mmio_gfn;
811 	u64 mmio_gen;
812 
813 	struct kvm_pmu pmu;
814 
815 	/* used for guest single stepping over the given code position */
816 	unsigned long singlestep_rip;
817 
818 	bool hyperv_enabled;
819 	struct kvm_vcpu_hv *hyperv;
820 	struct kvm_vcpu_xen xen;
821 
822 	cpumask_var_t wbinvd_dirty_mask;
823 
824 	unsigned long last_retry_eip;
825 	unsigned long last_retry_addr;
826 
827 	struct {
828 		bool halted;
829 		gfn_t gfns[ASYNC_PF_PER_VCPU];
830 		struct gfn_to_hva_cache data;
831 		u64 msr_en_val; /* MSR_KVM_ASYNC_PF_EN */
832 		u64 msr_int_val; /* MSR_KVM_ASYNC_PF_INT */
833 		u16 vec;
834 		u32 id;
835 		bool send_user_only;
836 		u32 host_apf_flags;
837 		unsigned long nested_apf_token;
838 		bool delivery_as_pf_vmexit;
839 		bool pageready_pending;
840 	} apf;
841 
842 	/* OSVW MSRs (AMD only) */
843 	struct {
844 		u64 length;
845 		u64 status;
846 	} osvw;
847 
848 	struct {
849 		u64 msr_val;
850 		struct gfn_to_hva_cache data;
851 	} pv_eoi;
852 
853 	u64 msr_kvm_poll_control;
854 
855 	/*
856 	 * Indicates the guest is trying to write a gfn that contains one or
857 	 * more of the PTEs used to translate the write itself, i.e. the access
858 	 * is changing its own translation in the guest page tables.  KVM exits
859 	 * to userspace if emulation of the faulting instruction fails and this
860 	 * flag is set, as KVM cannot make forward progress.
861 	 *
862 	 * If emulation fails for a write to guest page tables, KVM unprotects
863 	 * (zaps) the shadow page for the target gfn and resumes the guest to
864 	 * retry the non-emulatable instruction (on hardware).  Unprotecting the
865 	 * gfn doesn't allow forward progress for a self-changing access because
866 	 * doing so also zaps the translation for the gfn, i.e. retrying the
867 	 * instruction will hit a !PRESENT fault, which results in a new shadow
868 	 * page and sends KVM back to square one.
869 	 */
870 	bool write_fault_to_shadow_pgtable;
871 
872 	/* set at EPT violation at this point */
873 	unsigned long exit_qualification;
874 
875 	/* pv related host specific info */
876 	struct {
877 		bool pv_unhalted;
878 	} pv;
879 
880 	int pending_ioapic_eoi;
881 	int pending_external_vector;
882 
883 	/* be preempted when it's in kernel-mode(cpl=0) */
884 	bool preempted_in_kernel;
885 
886 	/* Flush the L1 Data cache for L1TF mitigation on VMENTER */
887 	bool l1tf_flush_l1d;
888 
889 	/* Host CPU on which VM-entry was most recently attempted */
890 	int last_vmentry_cpu;
891 
892 	/* AMD MSRC001_0015 Hardware Configuration */
893 	u64 msr_hwcr;
894 
895 	/* pv related cpuid info */
896 	struct {
897 		/*
898 		 * value of the eax register in the KVM_CPUID_FEATURES CPUID
899 		 * leaf.
900 		 */
901 		u32 features;
902 
903 		/*
904 		 * indicates whether pv emulation should be disabled if features
905 		 * are not present in the guest's cpuid
906 		 */
907 		bool enforce;
908 	} pv_cpuid;
909 
910 	/* Protected Guests */
911 	bool guest_state_protected;
912 
913 	/*
914 	 * Set when PDPTS were loaded directly by the userspace without
915 	 * reading the guest memory
916 	 */
917 	bool pdptrs_from_userspace;
918 
919 #if IS_ENABLED(CONFIG_HYPERV)
920 	hpa_t hv_root_tdp;
921 #endif
922 };
923 
924 struct kvm_lpage_info {
925 	int disallow_lpage;
926 };
927 
928 struct kvm_arch_memory_slot {
929 	struct kvm_rmap_head *rmap[KVM_NR_PAGE_SIZES];
930 	struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
931 	unsigned short *gfn_track[KVM_PAGE_TRACK_MAX];
932 };
933 
934 /*
935  * We use as the mode the number of bits allocated in the LDR for the
936  * logical processor ID.  It happens that these are all powers of two.
937  * This makes it is very easy to detect cases where the APICs are
938  * configured for multiple modes; in that case, we cannot use the map and
939  * hence cannot use kvm_irq_delivery_to_apic_fast either.
940  */
941 #define KVM_APIC_MODE_XAPIC_CLUSTER          4
942 #define KVM_APIC_MODE_XAPIC_FLAT             8
943 #define KVM_APIC_MODE_X2APIC                16
944 
945 struct kvm_apic_map {
946 	struct rcu_head rcu;
947 	u8 mode;
948 	u32 max_apic_id;
949 	union {
950 		struct kvm_lapic *xapic_flat_map[8];
951 		struct kvm_lapic *xapic_cluster_map[16][4];
952 	};
953 	struct kvm_lapic *phys_map[];
954 };
955 
956 /* Hyper-V synthetic debugger (SynDbg)*/
957 struct kvm_hv_syndbg {
958 	struct {
959 		u64 control;
960 		u64 status;
961 		u64 send_page;
962 		u64 recv_page;
963 		u64 pending_page;
964 	} control;
965 	u64 options;
966 };
967 
968 /* Current state of Hyper-V TSC page clocksource */
969 enum hv_tsc_page_status {
970 	/* TSC page was not set up or disabled */
971 	HV_TSC_PAGE_UNSET = 0,
972 	/* TSC page MSR was written by the guest, update pending */
973 	HV_TSC_PAGE_GUEST_CHANGED,
974 	/* TSC page MSR was written by KVM userspace, update pending */
975 	HV_TSC_PAGE_HOST_CHANGED,
976 	/* TSC page was properly set up and is currently active  */
977 	HV_TSC_PAGE_SET,
978 	/* TSC page is currently being updated and therefore is inactive */
979 	HV_TSC_PAGE_UPDATING,
980 	/* TSC page was set up with an inaccessible GPA */
981 	HV_TSC_PAGE_BROKEN,
982 };
983 
984 /* Hyper-V emulation context */
985 struct kvm_hv {
986 	struct mutex hv_lock;
987 	u64 hv_guest_os_id;
988 	u64 hv_hypercall;
989 	u64 hv_tsc_page;
990 	enum hv_tsc_page_status hv_tsc_page_status;
991 
992 	/* Hyper-v based guest crash (NT kernel bugcheck) parameters */
993 	u64 hv_crash_param[HV_X64_MSR_CRASH_PARAMS];
994 	u64 hv_crash_ctl;
995 
996 	struct ms_hyperv_tsc_page tsc_ref;
997 
998 	struct idr conn_to_evt;
999 
1000 	u64 hv_reenlightenment_control;
1001 	u64 hv_tsc_emulation_control;
1002 	u64 hv_tsc_emulation_status;
1003 
1004 	/* How many vCPUs have VP index != vCPU index */
1005 	atomic_t num_mismatched_vp_indexes;
1006 
1007 	/*
1008 	 * How many SynICs use 'AutoEOI' feature
1009 	 * (protected by arch.apicv_update_lock)
1010 	 */
1011 	unsigned int synic_auto_eoi_used;
1012 
1013 	struct hv_partition_assist_pg *hv_pa_pg;
1014 	struct kvm_hv_syndbg hv_syndbg;
1015 };
1016 
1017 struct msr_bitmap_range {
1018 	u32 flags;
1019 	u32 nmsrs;
1020 	u32 base;
1021 	unsigned long *bitmap;
1022 };
1023 
1024 /* Xen emulation context */
1025 struct kvm_xen {
1026 	bool long_mode;
1027 	u8 upcall_vector;
1028 	struct gfn_to_pfn_cache shinfo_cache;
1029 };
1030 
1031 enum kvm_irqchip_mode {
1032 	KVM_IRQCHIP_NONE,
1033 	KVM_IRQCHIP_KERNEL,       /* created with KVM_CREATE_IRQCHIP */
1034 	KVM_IRQCHIP_SPLIT,        /* created with KVM_CAP_SPLIT_IRQCHIP */
1035 };
1036 
1037 struct kvm_x86_msr_filter {
1038 	u8 count;
1039 	bool default_allow:1;
1040 	struct msr_bitmap_range ranges[16];
1041 };
1042 
1043 #define APICV_INHIBIT_REASON_DISABLE    0
1044 #define APICV_INHIBIT_REASON_HYPERV     1
1045 #define APICV_INHIBIT_REASON_NESTED     2
1046 #define APICV_INHIBIT_REASON_IRQWIN     3
1047 #define APICV_INHIBIT_REASON_PIT_REINJ  4
1048 #define APICV_INHIBIT_REASON_X2APIC	5
1049 #define APICV_INHIBIT_REASON_BLOCKIRQ	6
1050 #define APICV_INHIBIT_REASON_ABSENT	7
1051 
1052 struct kvm_arch {
1053 	unsigned long n_used_mmu_pages;
1054 	unsigned long n_requested_mmu_pages;
1055 	unsigned long n_max_mmu_pages;
1056 	unsigned int indirect_shadow_pages;
1057 	u8 mmu_valid_gen;
1058 	struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
1059 	struct list_head active_mmu_pages;
1060 	struct list_head zapped_obsolete_pages;
1061 	struct list_head lpage_disallowed_mmu_pages;
1062 	struct kvm_page_track_notifier_node mmu_sp_tracker;
1063 	struct kvm_page_track_notifier_head track_notifier_head;
1064 	/*
1065 	 * Protects marking pages unsync during page faults, as TDP MMU page
1066 	 * faults only take mmu_lock for read.  For simplicity, the unsync
1067 	 * pages lock is always taken when marking pages unsync regardless of
1068 	 * whether mmu_lock is held for read or write.
1069 	 */
1070 	spinlock_t mmu_unsync_pages_lock;
1071 
1072 	struct list_head assigned_dev_head;
1073 	struct iommu_domain *iommu_domain;
1074 	bool iommu_noncoherent;
1075 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA
1076 	atomic_t noncoherent_dma_count;
1077 #define __KVM_HAVE_ARCH_ASSIGNED_DEVICE
1078 	atomic_t assigned_device_count;
1079 	struct kvm_pic *vpic;
1080 	struct kvm_ioapic *vioapic;
1081 	struct kvm_pit *vpit;
1082 	atomic_t vapics_in_nmi_mode;
1083 	struct mutex apic_map_lock;
1084 	struct kvm_apic_map __rcu *apic_map;
1085 	atomic_t apic_map_dirty;
1086 
1087 	/* Protects apic_access_memslot_enabled and apicv_inhibit_reasons */
1088 	struct rw_semaphore apicv_update_lock;
1089 
1090 	bool apic_access_memslot_enabled;
1091 	unsigned long apicv_inhibit_reasons;
1092 
1093 	gpa_t wall_clock;
1094 
1095 	bool mwait_in_guest;
1096 	bool hlt_in_guest;
1097 	bool pause_in_guest;
1098 	bool cstate_in_guest;
1099 
1100 	unsigned long irq_sources_bitmap;
1101 	s64 kvmclock_offset;
1102 
1103 	/*
1104 	 * This also protects nr_vcpus_matched_tsc which is read from a
1105 	 * preemption-disabled region, so it must be a raw spinlock.
1106 	 */
1107 	raw_spinlock_t tsc_write_lock;
1108 	u64 last_tsc_nsec;
1109 	u64 last_tsc_write;
1110 	u32 last_tsc_khz;
1111 	u64 last_tsc_offset;
1112 	u64 cur_tsc_nsec;
1113 	u64 cur_tsc_write;
1114 	u64 cur_tsc_offset;
1115 	u64 cur_tsc_generation;
1116 	int nr_vcpus_matched_tsc;
1117 
1118 	seqcount_raw_spinlock_t pvclock_sc;
1119 	bool use_master_clock;
1120 	u64 master_kernel_ns;
1121 	u64 master_cycle_now;
1122 	struct delayed_work kvmclock_update_work;
1123 	struct delayed_work kvmclock_sync_work;
1124 
1125 	struct kvm_xen_hvm_config xen_hvm_config;
1126 
1127 	/* reads protected by irq_srcu, writes by irq_lock */
1128 	struct hlist_head mask_notifier_list;
1129 
1130 	struct kvm_hv hyperv;
1131 	struct kvm_xen xen;
1132 
1133 	bool backwards_tsc_observed;
1134 	bool boot_vcpu_runs_old_kvmclock;
1135 	u32 bsp_vcpu_id;
1136 
1137 	u64 disabled_quirks;
1138 	int cpu_dirty_logging_count;
1139 
1140 	enum kvm_irqchip_mode irqchip_mode;
1141 	u8 nr_reserved_ioapic_pins;
1142 
1143 	bool disabled_lapic_found;
1144 
1145 	bool x2apic_format;
1146 	bool x2apic_broadcast_quirk_disabled;
1147 
1148 	bool guest_can_read_msr_platform_info;
1149 	bool exception_payload_enabled;
1150 
1151 	bool bus_lock_detection_enabled;
1152 	bool enable_pmu;
1153 	/*
1154 	 * If exit_on_emulation_error is set, and the in-kernel instruction
1155 	 * emulator fails to emulate an instruction, allow userspace
1156 	 * the opportunity to look at it.
1157 	 */
1158 	bool exit_on_emulation_error;
1159 
1160 	/* Deflect RDMSR and WRMSR to user space when they trigger a #GP */
1161 	u32 user_space_msr_mask;
1162 	struct kvm_x86_msr_filter __rcu *msr_filter;
1163 
1164 	u32 hypercall_exit_enabled;
1165 
1166 	/* Guest can access the SGX PROVISIONKEY. */
1167 	bool sgx_provisioning_allowed;
1168 
1169 	struct kvm_pmu_event_filter __rcu *pmu_event_filter;
1170 	struct task_struct *nx_lpage_recovery_thread;
1171 
1172 #ifdef CONFIG_X86_64
1173 	/*
1174 	 * Whether the TDP MMU is enabled for this VM. This contains a
1175 	 * snapshot of the TDP MMU module parameter from when the VM was
1176 	 * created and remains unchanged for the life of the VM. If this is
1177 	 * true, TDP MMU handler functions will run for various MMU
1178 	 * operations.
1179 	 */
1180 	bool tdp_mmu_enabled;
1181 
1182 	/*
1183 	 * List of struct kvm_mmu_pages being used as roots.
1184 	 * All struct kvm_mmu_pages in the list should have
1185 	 * tdp_mmu_page set.
1186 	 *
1187 	 * For reads, this list is protected by:
1188 	 *	the MMU lock in read mode + RCU or
1189 	 *	the MMU lock in write mode
1190 	 *
1191 	 * For writes, this list is protected by:
1192 	 *	the MMU lock in read mode + the tdp_mmu_pages_lock or
1193 	 *	the MMU lock in write mode
1194 	 *
1195 	 * Roots will remain in the list until their tdp_mmu_root_count
1196 	 * drops to zero, at which point the thread that decremented the
1197 	 * count to zero should removed the root from the list and clean
1198 	 * it up, freeing the root after an RCU grace period.
1199 	 */
1200 	struct list_head tdp_mmu_roots;
1201 
1202 	/*
1203 	 * List of struct kvmp_mmu_pages not being used as roots.
1204 	 * All struct kvm_mmu_pages in the list should have
1205 	 * tdp_mmu_page set and a tdp_mmu_root_count of 0.
1206 	 */
1207 	struct list_head tdp_mmu_pages;
1208 
1209 	/*
1210 	 * Protects accesses to the following fields when the MMU lock
1211 	 * is held in read mode:
1212 	 *  - tdp_mmu_roots (above)
1213 	 *  - tdp_mmu_pages (above)
1214 	 *  - the link field of struct kvm_mmu_pages used by the TDP MMU
1215 	 *  - lpage_disallowed_mmu_pages
1216 	 *  - the lpage_disallowed_link field of struct kvm_mmu_pages used
1217 	 *    by the TDP MMU
1218 	 * It is acceptable, but not necessary, to acquire this lock when
1219 	 * the thread holds the MMU lock in write mode.
1220 	 */
1221 	spinlock_t tdp_mmu_pages_lock;
1222 	struct workqueue_struct *tdp_mmu_zap_wq;
1223 #endif /* CONFIG_X86_64 */
1224 
1225 	/*
1226 	 * If set, at least one shadow root has been allocated. This flag
1227 	 * is used as one input when determining whether certain memslot
1228 	 * related allocations are necessary.
1229 	 */
1230 	bool shadow_root_allocated;
1231 
1232 #if IS_ENABLED(CONFIG_HYPERV)
1233 	hpa_t	hv_root_tdp;
1234 	spinlock_t hv_root_tdp_lock;
1235 #endif
1236 };
1237 
1238 struct kvm_vm_stat {
1239 	struct kvm_vm_stat_generic generic;
1240 	u64 mmu_shadow_zapped;
1241 	u64 mmu_pte_write;
1242 	u64 mmu_pde_zapped;
1243 	u64 mmu_flooded;
1244 	u64 mmu_recycled;
1245 	u64 mmu_cache_miss;
1246 	u64 mmu_unsync;
1247 	union {
1248 		struct {
1249 			atomic64_t pages_4k;
1250 			atomic64_t pages_2m;
1251 			atomic64_t pages_1g;
1252 		};
1253 		atomic64_t pages[KVM_NR_PAGE_SIZES];
1254 	};
1255 	u64 nx_lpage_splits;
1256 	u64 max_mmu_page_hash_collisions;
1257 	u64 max_mmu_rmap_size;
1258 };
1259 
1260 struct kvm_vcpu_stat {
1261 	struct kvm_vcpu_stat_generic generic;
1262 	u64 pf_fixed;
1263 	u64 pf_guest;
1264 	u64 tlb_flush;
1265 	u64 invlpg;
1266 
1267 	u64 exits;
1268 	u64 io_exits;
1269 	u64 mmio_exits;
1270 	u64 signal_exits;
1271 	u64 irq_window_exits;
1272 	u64 nmi_window_exits;
1273 	u64 l1d_flush;
1274 	u64 halt_exits;
1275 	u64 request_irq_exits;
1276 	u64 irq_exits;
1277 	u64 host_state_reload;
1278 	u64 fpu_reload;
1279 	u64 insn_emulation;
1280 	u64 insn_emulation_fail;
1281 	u64 hypercalls;
1282 	u64 irq_injections;
1283 	u64 nmi_injections;
1284 	u64 req_event;
1285 	u64 nested_run;
1286 	u64 directed_yield_attempted;
1287 	u64 directed_yield_successful;
1288 	u64 guest_mode;
1289 };
1290 
1291 struct x86_instruction_info;
1292 
1293 struct msr_data {
1294 	bool host_initiated;
1295 	u32 index;
1296 	u64 data;
1297 };
1298 
1299 struct kvm_lapic_irq {
1300 	u32 vector;
1301 	u16 delivery_mode;
1302 	u16 dest_mode;
1303 	bool level;
1304 	u16 trig_mode;
1305 	u32 shorthand;
1306 	u32 dest_id;
1307 	bool msi_redir_hint;
1308 };
1309 
1310 static inline u16 kvm_lapic_irq_dest_mode(bool dest_mode_logical)
1311 {
1312 	return dest_mode_logical ? APIC_DEST_LOGICAL : APIC_DEST_PHYSICAL;
1313 }
1314 
1315 struct kvm_x86_ops {
1316 	const char *name;
1317 
1318 	int (*hardware_enable)(void);
1319 	void (*hardware_disable)(void);
1320 	void (*hardware_unsetup)(void);
1321 	bool (*has_emulated_msr)(struct kvm *kvm, u32 index);
1322 	void (*vcpu_after_set_cpuid)(struct kvm_vcpu *vcpu);
1323 
1324 	unsigned int vm_size;
1325 	int (*vm_init)(struct kvm *kvm);
1326 	void (*vm_destroy)(struct kvm *kvm);
1327 
1328 	/* Create, but do not attach this VCPU */
1329 	int (*vcpu_create)(struct kvm_vcpu *vcpu);
1330 	void (*vcpu_free)(struct kvm_vcpu *vcpu);
1331 	void (*vcpu_reset)(struct kvm_vcpu *vcpu, bool init_event);
1332 
1333 	void (*prepare_switch_to_guest)(struct kvm_vcpu *vcpu);
1334 	void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
1335 	void (*vcpu_put)(struct kvm_vcpu *vcpu);
1336 
1337 	void (*update_exception_bitmap)(struct kvm_vcpu *vcpu);
1338 	int (*get_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1339 	int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1340 	u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
1341 	void (*get_segment)(struct kvm_vcpu *vcpu,
1342 			    struct kvm_segment *var, int seg);
1343 	int (*get_cpl)(struct kvm_vcpu *vcpu);
1344 	void (*set_segment)(struct kvm_vcpu *vcpu,
1345 			    struct kvm_segment *var, int seg);
1346 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
1347 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
1348 	void (*post_set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
1349 	bool (*is_valid_cr4)(struct kvm_vcpu *vcpu, unsigned long cr0);
1350 	void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
1351 	int (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
1352 	void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1353 	void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1354 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1355 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1356 	void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu);
1357 	void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
1358 	void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
1359 	unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
1360 	void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
1361 	bool (*get_if_flag)(struct kvm_vcpu *vcpu);
1362 
1363 	void (*flush_tlb_all)(struct kvm_vcpu *vcpu);
1364 	void (*flush_tlb_current)(struct kvm_vcpu *vcpu);
1365 	int  (*tlb_remote_flush)(struct kvm *kvm);
1366 	int  (*tlb_remote_flush_with_range)(struct kvm *kvm,
1367 			struct kvm_tlb_range *range);
1368 
1369 	/*
1370 	 * Flush any TLB entries associated with the given GVA.
1371 	 * Does not need to flush GPA->HPA mappings.
1372 	 * Can potentially get non-canonical addresses through INVLPGs, which
1373 	 * the implementation may choose to ignore if appropriate.
1374 	 */
1375 	void (*flush_tlb_gva)(struct kvm_vcpu *vcpu, gva_t addr);
1376 
1377 	/*
1378 	 * Flush any TLB entries created by the guest.  Like tlb_flush_gva(),
1379 	 * does not need to flush GPA->HPA mappings.
1380 	 */
1381 	void (*flush_tlb_guest)(struct kvm_vcpu *vcpu);
1382 
1383 	int (*vcpu_pre_run)(struct kvm_vcpu *vcpu);
1384 	enum exit_fastpath_completion (*vcpu_run)(struct kvm_vcpu *vcpu);
1385 	int (*handle_exit)(struct kvm_vcpu *vcpu,
1386 		enum exit_fastpath_completion exit_fastpath);
1387 	int (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
1388 	void (*update_emulated_instruction)(struct kvm_vcpu *vcpu);
1389 	void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
1390 	u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu);
1391 	void (*patch_hypercall)(struct kvm_vcpu *vcpu,
1392 				unsigned char *hypercall_addr);
1393 	void (*inject_irq)(struct kvm_vcpu *vcpu);
1394 	void (*inject_nmi)(struct kvm_vcpu *vcpu);
1395 	void (*queue_exception)(struct kvm_vcpu *vcpu);
1396 	void (*cancel_injection)(struct kvm_vcpu *vcpu);
1397 	int (*interrupt_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1398 	int (*nmi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1399 	bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
1400 	void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
1401 	void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
1402 	void (*enable_irq_window)(struct kvm_vcpu *vcpu);
1403 	void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
1404 	bool (*check_apicv_inhibit_reasons)(ulong bit);
1405 	void (*refresh_apicv_exec_ctrl)(struct kvm_vcpu *vcpu);
1406 	void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
1407 	void (*hwapic_isr_update)(struct kvm_vcpu *vcpu, int isr);
1408 	bool (*guest_apic_has_interrupt)(struct kvm_vcpu *vcpu);
1409 	void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
1410 	void (*set_virtual_apic_mode)(struct kvm_vcpu *vcpu);
1411 	void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu);
1412 	void (*deliver_interrupt)(struct kvm_lapic *apic, int delivery_mode,
1413 				  int trig_mode, int vector);
1414 	int (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
1415 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
1416 	int (*set_identity_map_addr)(struct kvm *kvm, u64 ident_addr);
1417 	u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
1418 
1419 	void (*load_mmu_pgd)(struct kvm_vcpu *vcpu, hpa_t root_hpa,
1420 			     int root_level);
1421 
1422 	bool (*has_wbinvd_exit)(void);
1423 
1424 	u64 (*get_l2_tsc_offset)(struct kvm_vcpu *vcpu);
1425 	u64 (*get_l2_tsc_multiplier)(struct kvm_vcpu *vcpu);
1426 	void (*write_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
1427 	void (*write_tsc_multiplier)(struct kvm_vcpu *vcpu, u64 multiplier);
1428 
1429 	/*
1430 	 * Retrieve somewhat arbitrary exit information.  Intended to
1431 	 * be used only from within tracepoints or error paths.
1432 	 */
1433 	void (*get_exit_info)(struct kvm_vcpu *vcpu, u32 *reason,
1434 			      u64 *info1, u64 *info2,
1435 			      u32 *exit_int_info, u32 *exit_int_info_err_code);
1436 
1437 	int (*check_intercept)(struct kvm_vcpu *vcpu,
1438 			       struct x86_instruction_info *info,
1439 			       enum x86_intercept_stage stage,
1440 			       struct x86_exception *exception);
1441 	void (*handle_exit_irqoff)(struct kvm_vcpu *vcpu);
1442 
1443 	void (*request_immediate_exit)(struct kvm_vcpu *vcpu);
1444 
1445 	void (*sched_in)(struct kvm_vcpu *kvm, int cpu);
1446 
1447 	/*
1448 	 * Size of the CPU's dirty log buffer, i.e. VMX's PML buffer.  A zero
1449 	 * value indicates CPU dirty logging is unsupported or disabled.
1450 	 */
1451 	int cpu_dirty_log_size;
1452 	void (*update_cpu_dirty_logging)(struct kvm_vcpu *vcpu);
1453 
1454 	/* pmu operations of sub-arch */
1455 	const struct kvm_pmu_ops *pmu_ops;
1456 	const struct kvm_x86_nested_ops *nested_ops;
1457 
1458 	void (*vcpu_blocking)(struct kvm_vcpu *vcpu);
1459 	void (*vcpu_unblocking)(struct kvm_vcpu *vcpu);
1460 
1461 	int (*pi_update_irte)(struct kvm *kvm, unsigned int host_irq,
1462 			      uint32_t guest_irq, bool set);
1463 	void (*pi_start_assignment)(struct kvm *kvm);
1464 	void (*apicv_post_state_restore)(struct kvm_vcpu *vcpu);
1465 	bool (*dy_apicv_has_pending_interrupt)(struct kvm_vcpu *vcpu);
1466 
1467 	int (*set_hv_timer)(struct kvm_vcpu *vcpu, u64 guest_deadline_tsc,
1468 			    bool *expired);
1469 	void (*cancel_hv_timer)(struct kvm_vcpu *vcpu);
1470 
1471 	void (*setup_mce)(struct kvm_vcpu *vcpu);
1472 
1473 	int (*smi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1474 	int (*enter_smm)(struct kvm_vcpu *vcpu, char *smstate);
1475 	int (*leave_smm)(struct kvm_vcpu *vcpu, const char *smstate);
1476 	void (*enable_smi_window)(struct kvm_vcpu *vcpu);
1477 
1478 	int (*mem_enc_ioctl)(struct kvm *kvm, void __user *argp);
1479 	int (*mem_enc_register_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1480 	int (*mem_enc_unregister_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1481 	int (*vm_copy_enc_context_from)(struct kvm *kvm, unsigned int source_fd);
1482 	int (*vm_move_enc_context_from)(struct kvm *kvm, unsigned int source_fd);
1483 
1484 	int (*get_msr_feature)(struct kvm_msr_entry *entry);
1485 
1486 	bool (*can_emulate_instruction)(struct kvm_vcpu *vcpu, int emul_type,
1487 					void *insn, int insn_len);
1488 
1489 	bool (*apic_init_signal_blocked)(struct kvm_vcpu *vcpu);
1490 	int (*enable_direct_tlbflush)(struct kvm_vcpu *vcpu);
1491 
1492 	void (*migrate_timers)(struct kvm_vcpu *vcpu);
1493 	void (*msr_filter_changed)(struct kvm_vcpu *vcpu);
1494 	int (*complete_emulated_msr)(struct kvm_vcpu *vcpu, int err);
1495 
1496 	void (*vcpu_deliver_sipi_vector)(struct kvm_vcpu *vcpu, u8 vector);
1497 };
1498 
1499 struct kvm_x86_nested_ops {
1500 	void (*leave_nested)(struct kvm_vcpu *vcpu);
1501 	int (*check_events)(struct kvm_vcpu *vcpu);
1502 	bool (*hv_timer_pending)(struct kvm_vcpu *vcpu);
1503 	void (*triple_fault)(struct kvm_vcpu *vcpu);
1504 	int (*get_state)(struct kvm_vcpu *vcpu,
1505 			 struct kvm_nested_state __user *user_kvm_nested_state,
1506 			 unsigned user_data_size);
1507 	int (*set_state)(struct kvm_vcpu *vcpu,
1508 			 struct kvm_nested_state __user *user_kvm_nested_state,
1509 			 struct kvm_nested_state *kvm_state);
1510 	bool (*get_nested_state_pages)(struct kvm_vcpu *vcpu);
1511 	int (*write_log_dirty)(struct kvm_vcpu *vcpu, gpa_t l2_gpa);
1512 
1513 	int (*enable_evmcs)(struct kvm_vcpu *vcpu,
1514 			    uint16_t *vmcs_version);
1515 	uint16_t (*get_evmcs_version)(struct kvm_vcpu *vcpu);
1516 };
1517 
1518 struct kvm_x86_init_ops {
1519 	int (*cpu_has_kvm_support)(void);
1520 	int (*disabled_by_bios)(void);
1521 	int (*check_processor_compatibility)(void);
1522 	int (*hardware_setup)(void);
1523 	unsigned int (*handle_intel_pt_intr)(void);
1524 
1525 	struct kvm_x86_ops *runtime_ops;
1526 };
1527 
1528 struct kvm_arch_async_pf {
1529 	u32 token;
1530 	gfn_t gfn;
1531 	unsigned long cr3;
1532 	bool direct_map;
1533 };
1534 
1535 extern u32 __read_mostly kvm_nr_uret_msrs;
1536 extern u64 __read_mostly host_efer;
1537 extern bool __read_mostly allow_smaller_maxphyaddr;
1538 extern bool __read_mostly enable_apicv;
1539 extern struct kvm_x86_ops kvm_x86_ops;
1540 
1541 #define KVM_X86_OP(func) \
1542 	DECLARE_STATIC_CALL(kvm_x86_##func, *(((struct kvm_x86_ops *)0)->func));
1543 #define KVM_X86_OP_OPTIONAL KVM_X86_OP
1544 #define KVM_X86_OP_OPTIONAL_RET0 KVM_X86_OP
1545 #include <asm/kvm-x86-ops.h>
1546 
1547 static inline void kvm_ops_static_call_update(void)
1548 {
1549 #define __KVM_X86_OP(func) \
1550 	static_call_update(kvm_x86_##func, kvm_x86_ops.func);
1551 #define KVM_X86_OP(func) \
1552 	WARN_ON(!kvm_x86_ops.func); __KVM_X86_OP(func)
1553 #define KVM_X86_OP_OPTIONAL __KVM_X86_OP
1554 #define KVM_X86_OP_OPTIONAL_RET0(func) \
1555 	static_call_update(kvm_x86_##func, (void *)kvm_x86_ops.func ? : \
1556 					   (void *)__static_call_return0);
1557 #include <asm/kvm-x86-ops.h>
1558 #undef __KVM_X86_OP
1559 }
1560 
1561 #define __KVM_HAVE_ARCH_VM_ALLOC
1562 static inline struct kvm *kvm_arch_alloc_vm(void)
1563 {
1564 	return __vmalloc(kvm_x86_ops.vm_size, GFP_KERNEL_ACCOUNT | __GFP_ZERO);
1565 }
1566 
1567 #define __KVM_HAVE_ARCH_VM_FREE
1568 void kvm_arch_free_vm(struct kvm *kvm);
1569 
1570 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
1571 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
1572 {
1573 	if (kvm_x86_ops.tlb_remote_flush &&
1574 	    !static_call(kvm_x86_tlb_remote_flush)(kvm))
1575 		return 0;
1576 	else
1577 		return -ENOTSUPP;
1578 }
1579 
1580 #define kvm_arch_pmi_in_guest(vcpu) \
1581 	((vcpu) && (vcpu)->arch.handling_intr_from_guest)
1582 
1583 int kvm_mmu_module_init(void);
1584 void kvm_mmu_module_exit(void);
1585 
1586 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
1587 int kvm_mmu_create(struct kvm_vcpu *vcpu);
1588 void kvm_mmu_init_vm(struct kvm *kvm);
1589 void kvm_mmu_uninit_vm(struct kvm *kvm);
1590 
1591 void kvm_mmu_after_set_cpuid(struct kvm_vcpu *vcpu);
1592 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
1593 void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
1594 				      const struct kvm_memory_slot *memslot,
1595 				      int start_level);
1596 void kvm_mmu_slot_try_split_huge_pages(struct kvm *kvm,
1597 				       const struct kvm_memory_slot *memslot,
1598 				       int target_level);
1599 void kvm_mmu_try_split_huge_pages(struct kvm *kvm,
1600 				  const struct kvm_memory_slot *memslot,
1601 				  u64 start, u64 end,
1602 				  int target_level);
1603 void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
1604 				   const struct kvm_memory_slot *memslot);
1605 void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
1606 				   const struct kvm_memory_slot *memslot);
1607 void kvm_mmu_zap_all(struct kvm *kvm);
1608 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, u64 gen);
1609 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned long kvm_nr_mmu_pages);
1610 
1611 int load_pdptrs(struct kvm_vcpu *vcpu, unsigned long cr3);
1612 
1613 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
1614 			  const void *val, int bytes);
1615 
1616 struct kvm_irq_mask_notifier {
1617 	void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
1618 	int irq;
1619 	struct hlist_node link;
1620 };
1621 
1622 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
1623 				    struct kvm_irq_mask_notifier *kimn);
1624 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
1625 				      struct kvm_irq_mask_notifier *kimn);
1626 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
1627 			     bool mask);
1628 
1629 extern bool tdp_enabled;
1630 
1631 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
1632 
1633 /* control of guest tsc rate supported? */
1634 extern bool kvm_has_tsc_control;
1635 /* maximum supported tsc_khz for guests */
1636 extern u32  kvm_max_guest_tsc_khz;
1637 /* number of bits of the fractional part of the TSC scaling ratio */
1638 extern u8   kvm_tsc_scaling_ratio_frac_bits;
1639 /* maximum allowed value of TSC scaling ratio */
1640 extern u64  kvm_max_tsc_scaling_ratio;
1641 /* 1ull << kvm_tsc_scaling_ratio_frac_bits */
1642 extern u64  kvm_default_tsc_scaling_ratio;
1643 /* bus lock detection supported? */
1644 extern bool kvm_has_bus_lock_exit;
1645 
1646 extern u64 kvm_mce_cap_supported;
1647 
1648 /*
1649  * EMULTYPE_NO_DECODE - Set when re-emulating an instruction (after completing
1650  *			userspace I/O) to indicate that the emulation context
1651  *			should be reused as is, i.e. skip initialization of
1652  *			emulation context, instruction fetch and decode.
1653  *
1654  * EMULTYPE_TRAP_UD - Set when emulating an intercepted #UD from hardware.
1655  *		      Indicates that only select instructions (tagged with
1656  *		      EmulateOnUD) should be emulated (to minimize the emulator
1657  *		      attack surface).  See also EMULTYPE_TRAP_UD_FORCED.
1658  *
1659  * EMULTYPE_SKIP - Set when emulating solely to skip an instruction, i.e. to
1660  *		   decode the instruction length.  For use *only* by
1661  *		   kvm_x86_ops.skip_emulated_instruction() implementations if
1662  *		   EMULTYPE_COMPLETE_USER_EXIT is not set.
1663  *
1664  * EMULTYPE_ALLOW_RETRY_PF - Set when the emulator should resume the guest to
1665  *			     retry native execution under certain conditions,
1666  *			     Can only be set in conjunction with EMULTYPE_PF.
1667  *
1668  * EMULTYPE_TRAP_UD_FORCED - Set when emulating an intercepted #UD that was
1669  *			     triggered by KVM's magic "force emulation" prefix,
1670  *			     which is opt in via module param (off by default).
1671  *			     Bypasses EmulateOnUD restriction despite emulating
1672  *			     due to an intercepted #UD (see EMULTYPE_TRAP_UD).
1673  *			     Used to test the full emulator from userspace.
1674  *
1675  * EMULTYPE_VMWARE_GP - Set when emulating an intercepted #GP for VMware
1676  *			backdoor emulation, which is opt in via module param.
1677  *			VMware backdoor emulation handles select instructions
1678  *			and reinjects the #GP for all other cases.
1679  *
1680  * EMULTYPE_PF - Set when emulating MMIO by way of an intercepted #PF, in which
1681  *		 case the CR2/GPA value pass on the stack is valid.
1682  *
1683  * EMULTYPE_COMPLETE_USER_EXIT - Set when the emulator should update interruptibility
1684  *				 state and inject single-step #DBs after skipping
1685  *				 an instruction (after completing userspace I/O).
1686  */
1687 #define EMULTYPE_NO_DECODE	    (1 << 0)
1688 #define EMULTYPE_TRAP_UD	    (1 << 1)
1689 #define EMULTYPE_SKIP		    (1 << 2)
1690 #define EMULTYPE_ALLOW_RETRY_PF	    (1 << 3)
1691 #define EMULTYPE_TRAP_UD_FORCED	    (1 << 4)
1692 #define EMULTYPE_VMWARE_GP	    (1 << 5)
1693 #define EMULTYPE_PF		    (1 << 6)
1694 #define EMULTYPE_COMPLETE_USER_EXIT (1 << 7)
1695 
1696 int kvm_emulate_instruction(struct kvm_vcpu *vcpu, int emulation_type);
1697 int kvm_emulate_instruction_from_buffer(struct kvm_vcpu *vcpu,
1698 					void *insn, int insn_len);
1699 void __kvm_prepare_emulation_failure_exit(struct kvm_vcpu *vcpu,
1700 					  u64 *data, u8 ndata);
1701 void kvm_prepare_emulation_failure_exit(struct kvm_vcpu *vcpu);
1702 
1703 void kvm_enable_efer_bits(u64);
1704 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
1705 int __kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data, bool host_initiated);
1706 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data);
1707 int kvm_set_msr(struct kvm_vcpu *vcpu, u32 index, u64 data);
1708 int kvm_emulate_rdmsr(struct kvm_vcpu *vcpu);
1709 int kvm_emulate_wrmsr(struct kvm_vcpu *vcpu);
1710 int kvm_emulate_as_nop(struct kvm_vcpu *vcpu);
1711 int kvm_emulate_invd(struct kvm_vcpu *vcpu);
1712 int kvm_emulate_mwait(struct kvm_vcpu *vcpu);
1713 int kvm_handle_invalid_op(struct kvm_vcpu *vcpu);
1714 int kvm_emulate_monitor(struct kvm_vcpu *vcpu);
1715 
1716 int kvm_fast_pio(struct kvm_vcpu *vcpu, int size, unsigned short port, int in);
1717 int kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
1718 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
1719 int kvm_emulate_halt_noskip(struct kvm_vcpu *vcpu);
1720 int kvm_emulate_ap_reset_hold(struct kvm_vcpu *vcpu);
1721 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
1722 
1723 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
1724 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
1725 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector);
1726 
1727 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
1728 		    int reason, bool has_error_code, u32 error_code);
1729 
1730 void kvm_post_set_cr0(struct kvm_vcpu *vcpu, unsigned long old_cr0, unsigned long cr0);
1731 void kvm_post_set_cr4(struct kvm_vcpu *vcpu, unsigned long old_cr4, unsigned long cr4);
1732 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
1733 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
1734 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
1735 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
1736 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
1737 void kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
1738 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
1739 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
1740 int kvm_emulate_xsetbv(struct kvm_vcpu *vcpu);
1741 
1742 int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1743 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1744 
1745 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
1746 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
1747 int kvm_emulate_rdpmc(struct kvm_vcpu *vcpu);
1748 
1749 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1750 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1751 void kvm_queue_exception_p(struct kvm_vcpu *vcpu, unsigned nr, unsigned long payload);
1752 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1753 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1754 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
1755 bool kvm_inject_emulated_page_fault(struct kvm_vcpu *vcpu,
1756 				    struct x86_exception *fault);
1757 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
1758 bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr);
1759 
1760 static inline int __kvm_irq_line_state(unsigned long *irq_state,
1761 				       int irq_source_id, int level)
1762 {
1763 	/* Logical OR for level trig interrupt */
1764 	if (level)
1765 		__set_bit(irq_source_id, irq_state);
1766 	else
1767 		__clear_bit(irq_source_id, irq_state);
1768 
1769 	return !!(*irq_state);
1770 }
1771 
1772 #define KVM_MMU_ROOT_CURRENT		BIT(0)
1773 #define KVM_MMU_ROOT_PREVIOUS(i)	BIT(1+i)
1774 #define KVM_MMU_ROOTS_ALL		(~0UL)
1775 
1776 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
1777 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
1778 
1779 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
1780 
1781 void kvm_update_dr7(struct kvm_vcpu *vcpu);
1782 
1783 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
1784 void kvm_mmu_free_roots(struct kvm *kvm, struct kvm_mmu *mmu,
1785 			ulong roots_to_free);
1786 void kvm_mmu_free_guest_mode_roots(struct kvm *kvm, struct kvm_mmu *mmu);
1787 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
1788 			      struct x86_exception *exception);
1789 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
1790 			       struct x86_exception *exception);
1791 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
1792 			       struct x86_exception *exception);
1793 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
1794 				struct x86_exception *exception);
1795 
1796 bool kvm_apicv_activated(struct kvm *kvm);
1797 void kvm_vcpu_update_apicv(struct kvm_vcpu *vcpu);
1798 void kvm_request_apicv_update(struct kvm *kvm, bool activate,
1799 			      unsigned long bit);
1800 
1801 void __kvm_request_apicv_update(struct kvm *kvm, bool activate,
1802 				unsigned long bit);
1803 
1804 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
1805 
1806 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u64 error_code,
1807 		       void *insn, int insn_len);
1808 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
1809 void kvm_mmu_invalidate_gva(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1810 			    gva_t gva, hpa_t root_hpa);
1811 void kvm_mmu_invpcid_gva(struct kvm_vcpu *vcpu, gva_t gva, unsigned long pcid);
1812 void kvm_mmu_new_pgd(struct kvm_vcpu *vcpu, gpa_t new_pgd);
1813 
1814 void kvm_configure_mmu(bool enable_tdp, int tdp_forced_root_level,
1815 		       int tdp_max_root_level, int tdp_huge_page_level);
1816 
1817 static inline u16 kvm_read_ldt(void)
1818 {
1819 	u16 ldt;
1820 	asm("sldt %0" : "=g"(ldt));
1821 	return ldt;
1822 }
1823 
1824 static inline void kvm_load_ldt(u16 sel)
1825 {
1826 	asm("lldt %0" : : "rm"(sel));
1827 }
1828 
1829 #ifdef CONFIG_X86_64
1830 static inline unsigned long read_msr(unsigned long msr)
1831 {
1832 	u64 value;
1833 
1834 	rdmsrl(msr, value);
1835 	return value;
1836 }
1837 #endif
1838 
1839 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
1840 {
1841 	kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
1842 }
1843 
1844 #define TSS_IOPB_BASE_OFFSET 0x66
1845 #define TSS_BASE_SIZE 0x68
1846 #define TSS_IOPB_SIZE (65536 / 8)
1847 #define TSS_REDIRECTION_SIZE (256 / 8)
1848 #define RMODE_TSS_SIZE							\
1849 	(TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
1850 
1851 enum {
1852 	TASK_SWITCH_CALL = 0,
1853 	TASK_SWITCH_IRET = 1,
1854 	TASK_SWITCH_JMP = 2,
1855 	TASK_SWITCH_GATE = 3,
1856 };
1857 
1858 #define HF_GIF_MASK		(1 << 0)
1859 #define HF_NMI_MASK		(1 << 3)
1860 #define HF_IRET_MASK		(1 << 4)
1861 #define HF_GUEST_MASK		(1 << 5) /* VCPU is in guest-mode */
1862 #define HF_SMM_MASK		(1 << 6)
1863 #define HF_SMM_INSIDE_NMI_MASK	(1 << 7)
1864 
1865 #define __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
1866 #define KVM_ADDRESS_SPACE_NUM 2
1867 
1868 #define kvm_arch_vcpu_memslots_id(vcpu) ((vcpu)->arch.hflags & HF_SMM_MASK ? 1 : 0)
1869 #define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, (role).smm)
1870 
1871 #define KVM_ARCH_WANT_MMU_NOTIFIER
1872 
1873 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1874 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1875 int kvm_cpu_has_extint(struct kvm_vcpu *v);
1876 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1877 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1878 void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event);
1879 
1880 int kvm_pv_send_ipi(struct kvm *kvm, unsigned long ipi_bitmap_low,
1881 		    unsigned long ipi_bitmap_high, u32 min,
1882 		    unsigned long icr, int op_64_bit);
1883 
1884 int kvm_add_user_return_msr(u32 msr);
1885 int kvm_find_user_return_msr(u32 msr);
1886 int kvm_set_user_return_msr(unsigned index, u64 val, u64 mask);
1887 
1888 static inline bool kvm_is_supported_user_return_msr(u32 msr)
1889 {
1890 	return kvm_find_user_return_msr(msr) >= 0;
1891 }
1892 
1893 u64 kvm_scale_tsc(u64 tsc, u64 ratio);
1894 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc);
1895 u64 kvm_calc_nested_tsc_offset(u64 l1_offset, u64 l2_offset, u64 l2_multiplier);
1896 u64 kvm_calc_nested_tsc_multiplier(u64 l1_multiplier, u64 l2_multiplier);
1897 
1898 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu);
1899 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1900 
1901 void kvm_make_scan_ioapic_request(struct kvm *kvm);
1902 void kvm_make_scan_ioapic_request_mask(struct kvm *kvm,
1903 				       unsigned long *vcpu_bitmap);
1904 
1905 bool kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1906 				     struct kvm_async_pf *work);
1907 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1908 				 struct kvm_async_pf *work);
1909 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1910 			       struct kvm_async_pf *work);
1911 void kvm_arch_async_page_present_queued(struct kvm_vcpu *vcpu);
1912 bool kvm_arch_can_dequeue_async_page_present(struct kvm_vcpu *vcpu);
1913 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1914 
1915 int kvm_skip_emulated_instruction(struct kvm_vcpu *vcpu);
1916 int kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1917 void __kvm_request_immediate_exit(struct kvm_vcpu *vcpu);
1918 
1919 void __user *__x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa,
1920 				     u32 size);
1921 bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu);
1922 bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu);
1923 
1924 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
1925 			     struct kvm_vcpu **dest_vcpu);
1926 
1927 void kvm_set_msi_irq(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e,
1928 		     struct kvm_lapic_irq *irq);
1929 
1930 static inline bool kvm_irq_is_postable(struct kvm_lapic_irq *irq)
1931 {
1932 	/* We can only post Fixed and LowPrio IRQs */
1933 	return (irq->delivery_mode == APIC_DM_FIXED ||
1934 		irq->delivery_mode == APIC_DM_LOWEST);
1935 }
1936 
1937 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu)
1938 {
1939 	static_call_cond(kvm_x86_vcpu_blocking)(vcpu);
1940 }
1941 
1942 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu)
1943 {
1944 	static_call_cond(kvm_x86_vcpu_unblocking)(vcpu);
1945 }
1946 
1947 static inline int kvm_cpu_get_apicid(int mps_cpu)
1948 {
1949 #ifdef CONFIG_X86_LOCAL_APIC
1950 	return default_cpu_present_to_apicid(mps_cpu);
1951 #else
1952 	WARN_ON_ONCE(1);
1953 	return BAD_APICID;
1954 #endif
1955 }
1956 
1957 #define put_smstate(type, buf, offset, val)                      \
1958 	*(type *)((buf) + (offset) - 0x7e00) = val
1959 
1960 #define GET_SMSTATE(type, buf, offset)		\
1961 	(*(type *)((buf) + (offset) - 0x7e00))
1962 
1963 int kvm_cpu_dirty_log_size(void);
1964 
1965 int memslot_rmap_alloc(struct kvm_memory_slot *slot, unsigned long npages);
1966 
1967 #define KVM_CLOCK_VALID_FLAGS						\
1968 	(KVM_CLOCK_TSC_STABLE | KVM_CLOCK_REALTIME | KVM_CLOCK_HOST_TSC)
1969 
1970 #define KVM_X86_VALID_QUIRKS			\
1971 	(KVM_X86_QUIRK_LINT0_REENABLED |	\
1972 	 KVM_X86_QUIRK_CD_NW_CLEARED |		\
1973 	 KVM_X86_QUIRK_LAPIC_MMIO_HOLE |	\
1974 	 KVM_X86_QUIRK_OUT_7E_INC_RIP |		\
1975 	 KVM_X86_QUIRK_MISC_ENABLE_NO_MWAIT)
1976 
1977 #endif /* _ASM_X86_KVM_HOST_H */
1978