xref: /openbmc/linux/arch/x86/include/asm/kvm_host.h (revision ee8ec048)
1 /* SPDX-License-Identifier: GPL-2.0-only */
2 /*
3  * Kernel-based Virtual Machine driver for Linux
4  *
5  * This header defines architecture specific interfaces, x86 version
6  */
7 
8 #ifndef _ASM_X86_KVM_HOST_H
9 #define _ASM_X86_KVM_HOST_H
10 
11 #include <linux/types.h>
12 #include <linux/mm.h>
13 #include <linux/mmu_notifier.h>
14 #include <linux/tracepoint.h>
15 #include <linux/cpumask.h>
16 #include <linux/irq_work.h>
17 #include <linux/irq.h>
18 
19 #include <linux/kvm.h>
20 #include <linux/kvm_para.h>
21 #include <linux/kvm_types.h>
22 #include <linux/perf_event.h>
23 #include <linux/pvclock_gtod.h>
24 #include <linux/clocksource.h>
25 #include <linux/irqbypass.h>
26 #include <linux/hyperv.h>
27 
28 #include <asm/apic.h>
29 #include <asm/pvclock-abi.h>
30 #include <asm/desc.h>
31 #include <asm/mtrr.h>
32 #include <asm/msr-index.h>
33 #include <asm/asm.h>
34 #include <asm/kvm_page_track.h>
35 #include <asm/kvm_vcpu_regs.h>
36 #include <asm/hyperv-tlfs.h>
37 
38 #define __KVM_HAVE_ARCH_VCPU_DEBUGFS
39 
40 #define KVM_MAX_VCPUS 1024
41 
42 /*
43  * In x86, the VCPU ID corresponds to the APIC ID, and APIC IDs
44  * might be larger than the actual number of VCPUs because the
45  * APIC ID encodes CPU topology information.
46  *
47  * In the worst case, we'll need less than one extra bit for the
48  * Core ID, and less than one extra bit for the Package (Die) ID,
49  * so ratio of 4 should be enough.
50  */
51 #define KVM_VCPU_ID_RATIO 4
52 #define KVM_MAX_VCPU_IDS (KVM_MAX_VCPUS * KVM_VCPU_ID_RATIO)
53 
54 /* memory slots that are not exposed to userspace */
55 #define KVM_PRIVATE_MEM_SLOTS 3
56 
57 #define KVM_HALT_POLL_NS_DEFAULT 200000
58 
59 #define KVM_IRQCHIP_NUM_PINS  KVM_IOAPIC_NUM_PINS
60 
61 #define KVM_DIRTY_LOG_MANUAL_CAPS   (KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE | \
62 					KVM_DIRTY_LOG_INITIALLY_SET)
63 
64 #define KVM_BUS_LOCK_DETECTION_VALID_MODE	(KVM_BUS_LOCK_DETECTION_OFF | \
65 						 KVM_BUS_LOCK_DETECTION_EXIT)
66 
67 /* x86-specific vcpu->requests bit members */
68 #define KVM_REQ_MIGRATE_TIMER		KVM_ARCH_REQ(0)
69 #define KVM_REQ_REPORT_TPR_ACCESS	KVM_ARCH_REQ(1)
70 #define KVM_REQ_TRIPLE_FAULT		KVM_ARCH_REQ(2)
71 #define KVM_REQ_MMU_SYNC		KVM_ARCH_REQ(3)
72 #define KVM_REQ_CLOCK_UPDATE		KVM_ARCH_REQ(4)
73 #define KVM_REQ_LOAD_MMU_PGD		KVM_ARCH_REQ(5)
74 #define KVM_REQ_EVENT			KVM_ARCH_REQ(6)
75 #define KVM_REQ_APF_HALT		KVM_ARCH_REQ(7)
76 #define KVM_REQ_STEAL_UPDATE		KVM_ARCH_REQ(8)
77 #define KVM_REQ_NMI			KVM_ARCH_REQ(9)
78 #define KVM_REQ_PMU			KVM_ARCH_REQ(10)
79 #define KVM_REQ_PMI			KVM_ARCH_REQ(11)
80 #define KVM_REQ_SMI			KVM_ARCH_REQ(12)
81 #define KVM_REQ_MASTERCLOCK_UPDATE	KVM_ARCH_REQ(13)
82 #define KVM_REQ_MCLOCK_INPROGRESS \
83 	KVM_ARCH_REQ_FLAGS(14, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
84 #define KVM_REQ_SCAN_IOAPIC \
85 	KVM_ARCH_REQ_FLAGS(15, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
86 #define KVM_REQ_GLOBAL_CLOCK_UPDATE	KVM_ARCH_REQ(16)
87 #define KVM_REQ_APIC_PAGE_RELOAD \
88 	KVM_ARCH_REQ_FLAGS(17, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
89 #define KVM_REQ_HV_CRASH		KVM_ARCH_REQ(18)
90 #define KVM_REQ_IOAPIC_EOI_EXIT		KVM_ARCH_REQ(19)
91 #define KVM_REQ_HV_RESET		KVM_ARCH_REQ(20)
92 #define KVM_REQ_HV_EXIT			KVM_ARCH_REQ(21)
93 #define KVM_REQ_HV_STIMER		KVM_ARCH_REQ(22)
94 #define KVM_REQ_LOAD_EOI_EXITMAP	KVM_ARCH_REQ(23)
95 #define KVM_REQ_GET_NESTED_STATE_PAGES	KVM_ARCH_REQ(24)
96 #define KVM_REQ_APICV_UPDATE \
97 	KVM_ARCH_REQ_FLAGS(25, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
98 #define KVM_REQ_TLB_FLUSH_CURRENT	KVM_ARCH_REQ(26)
99 #define KVM_REQ_TLB_FLUSH_GUEST \
100 	KVM_ARCH_REQ_FLAGS(27, KVM_REQUEST_NO_WAKEUP)
101 #define KVM_REQ_APF_READY		KVM_ARCH_REQ(28)
102 #define KVM_REQ_MSR_FILTER_CHANGED	KVM_ARCH_REQ(29)
103 #define KVM_REQ_UPDATE_CPU_DIRTY_LOGGING \
104 	KVM_ARCH_REQ_FLAGS(30, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
105 
106 #define CR0_RESERVED_BITS                                               \
107 	(~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \
108 			  | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \
109 			  | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG))
110 
111 #define CR4_RESERVED_BITS                                               \
112 	(~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\
113 			  | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE     \
114 			  | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR | X86_CR4_PCIDE \
115 			  | X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_FSGSBASE \
116 			  | X86_CR4_OSXMMEXCPT | X86_CR4_LA57 | X86_CR4_VMXE \
117 			  | X86_CR4_SMAP | X86_CR4_PKE | X86_CR4_UMIP))
118 
119 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR)
120 
121 
122 
123 #define INVALID_PAGE (~(hpa_t)0)
124 #define VALID_PAGE(x) ((x) != INVALID_PAGE)
125 
126 #define UNMAPPED_GVA (~(gpa_t)0)
127 #define INVALID_GPA (~(gpa_t)0)
128 
129 /* KVM Hugepage definitions for x86 */
130 #define KVM_MAX_HUGEPAGE_LEVEL	PG_LEVEL_1G
131 #define KVM_NR_PAGE_SIZES	(KVM_MAX_HUGEPAGE_LEVEL - PG_LEVEL_4K + 1)
132 #define KVM_HPAGE_GFN_SHIFT(x)	(((x) - 1) * 9)
133 #define KVM_HPAGE_SHIFT(x)	(PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
134 #define KVM_HPAGE_SIZE(x)	(1UL << KVM_HPAGE_SHIFT(x))
135 #define KVM_HPAGE_MASK(x)	(~(KVM_HPAGE_SIZE(x) - 1))
136 #define KVM_PAGES_PER_HPAGE(x)	(KVM_HPAGE_SIZE(x) / PAGE_SIZE)
137 
138 #define KVM_PERMILLE_MMU_PAGES 20
139 #define KVM_MIN_ALLOC_MMU_PAGES 64UL
140 #define KVM_MMU_HASH_SHIFT 12
141 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
142 #define KVM_MIN_FREE_MMU_PAGES 5
143 #define KVM_REFILL_PAGES 25
144 #define KVM_MAX_CPUID_ENTRIES 256
145 #define KVM_NR_FIXED_MTRR_REGION 88
146 #define KVM_NR_VAR_MTRR 8
147 
148 #define ASYNC_PF_PER_VCPU 64
149 
150 enum kvm_reg {
151 	VCPU_REGS_RAX = __VCPU_REGS_RAX,
152 	VCPU_REGS_RCX = __VCPU_REGS_RCX,
153 	VCPU_REGS_RDX = __VCPU_REGS_RDX,
154 	VCPU_REGS_RBX = __VCPU_REGS_RBX,
155 	VCPU_REGS_RSP = __VCPU_REGS_RSP,
156 	VCPU_REGS_RBP = __VCPU_REGS_RBP,
157 	VCPU_REGS_RSI = __VCPU_REGS_RSI,
158 	VCPU_REGS_RDI = __VCPU_REGS_RDI,
159 #ifdef CONFIG_X86_64
160 	VCPU_REGS_R8  = __VCPU_REGS_R8,
161 	VCPU_REGS_R9  = __VCPU_REGS_R9,
162 	VCPU_REGS_R10 = __VCPU_REGS_R10,
163 	VCPU_REGS_R11 = __VCPU_REGS_R11,
164 	VCPU_REGS_R12 = __VCPU_REGS_R12,
165 	VCPU_REGS_R13 = __VCPU_REGS_R13,
166 	VCPU_REGS_R14 = __VCPU_REGS_R14,
167 	VCPU_REGS_R15 = __VCPU_REGS_R15,
168 #endif
169 	VCPU_REGS_RIP,
170 	NR_VCPU_REGS,
171 
172 	VCPU_EXREG_PDPTR = NR_VCPU_REGS,
173 	VCPU_EXREG_CR0,
174 	VCPU_EXREG_CR3,
175 	VCPU_EXREG_CR4,
176 	VCPU_EXREG_RFLAGS,
177 	VCPU_EXREG_SEGMENTS,
178 	VCPU_EXREG_EXIT_INFO_1,
179 	VCPU_EXREG_EXIT_INFO_2,
180 };
181 
182 enum {
183 	VCPU_SREG_ES,
184 	VCPU_SREG_CS,
185 	VCPU_SREG_SS,
186 	VCPU_SREG_DS,
187 	VCPU_SREG_FS,
188 	VCPU_SREG_GS,
189 	VCPU_SREG_TR,
190 	VCPU_SREG_LDTR,
191 };
192 
193 enum exit_fastpath_completion {
194 	EXIT_FASTPATH_NONE,
195 	EXIT_FASTPATH_REENTER_GUEST,
196 	EXIT_FASTPATH_EXIT_HANDLED,
197 };
198 typedef enum exit_fastpath_completion fastpath_t;
199 
200 struct x86_emulate_ctxt;
201 struct x86_exception;
202 enum x86_intercept;
203 enum x86_intercept_stage;
204 
205 #define KVM_NR_DB_REGS	4
206 
207 #define DR6_BUS_LOCK   (1 << 11)
208 #define DR6_BD		(1 << 13)
209 #define DR6_BS		(1 << 14)
210 #define DR6_BT		(1 << 15)
211 #define DR6_RTM		(1 << 16)
212 /*
213  * DR6_ACTIVE_LOW combines fixed-1 and active-low bits.
214  * We can regard all the bits in DR6_FIXED_1 as active_low bits;
215  * they will never be 0 for now, but when they are defined
216  * in the future it will require no code change.
217  *
218  * DR6_ACTIVE_LOW is also used as the init/reset value for DR6.
219  */
220 #define DR6_ACTIVE_LOW	0xffff0ff0
221 #define DR6_VOLATILE	0x0001e80f
222 #define DR6_FIXED_1	(DR6_ACTIVE_LOW & ~DR6_VOLATILE)
223 
224 #define DR7_BP_EN_MASK	0x000000ff
225 #define DR7_GE		(1 << 9)
226 #define DR7_GD		(1 << 13)
227 #define DR7_FIXED_1	0x00000400
228 #define DR7_VOLATILE	0xffff2bff
229 
230 #define KVM_GUESTDBG_VALID_MASK \
231 	(KVM_GUESTDBG_ENABLE | \
232 	KVM_GUESTDBG_SINGLESTEP | \
233 	KVM_GUESTDBG_USE_HW_BP | \
234 	KVM_GUESTDBG_USE_SW_BP | \
235 	KVM_GUESTDBG_INJECT_BP | \
236 	KVM_GUESTDBG_INJECT_DB | \
237 	KVM_GUESTDBG_BLOCKIRQ)
238 
239 
240 #define PFERR_PRESENT_BIT 0
241 #define PFERR_WRITE_BIT 1
242 #define PFERR_USER_BIT 2
243 #define PFERR_RSVD_BIT 3
244 #define PFERR_FETCH_BIT 4
245 #define PFERR_PK_BIT 5
246 #define PFERR_SGX_BIT 15
247 #define PFERR_GUEST_FINAL_BIT 32
248 #define PFERR_GUEST_PAGE_BIT 33
249 
250 #define PFERR_PRESENT_MASK (1U << PFERR_PRESENT_BIT)
251 #define PFERR_WRITE_MASK (1U << PFERR_WRITE_BIT)
252 #define PFERR_USER_MASK (1U << PFERR_USER_BIT)
253 #define PFERR_RSVD_MASK (1U << PFERR_RSVD_BIT)
254 #define PFERR_FETCH_MASK (1U << PFERR_FETCH_BIT)
255 #define PFERR_PK_MASK (1U << PFERR_PK_BIT)
256 #define PFERR_SGX_MASK (1U << PFERR_SGX_BIT)
257 #define PFERR_GUEST_FINAL_MASK (1ULL << PFERR_GUEST_FINAL_BIT)
258 #define PFERR_GUEST_PAGE_MASK (1ULL << PFERR_GUEST_PAGE_BIT)
259 
260 #define PFERR_NESTED_GUEST_PAGE (PFERR_GUEST_PAGE_MASK |	\
261 				 PFERR_WRITE_MASK |		\
262 				 PFERR_PRESENT_MASK)
263 
264 /* apic attention bits */
265 #define KVM_APIC_CHECK_VAPIC	0
266 /*
267  * The following bit is set with PV-EOI, unset on EOI.
268  * We detect PV-EOI changes by guest by comparing
269  * this bit with PV-EOI in guest memory.
270  * See the implementation in apic_update_pv_eoi.
271  */
272 #define KVM_APIC_PV_EOI_PENDING	1
273 
274 struct kvm_kernel_irq_routing_entry;
275 
276 /*
277  * kvm_mmu_page_role tracks the properties of a shadow page (where shadow page
278  * also includes TDP pages) to determine whether or not a page can be used in
279  * the given MMU context.  This is a subset of the overall kvm_mmu_role to
280  * minimize the size of kvm_memory_slot.arch.gfn_track, i.e. allows allocating
281  * 2 bytes per gfn instead of 4 bytes per gfn.
282  *
283  * Indirect upper-level shadow pages are tracked for write-protection via
284  * gfn_track.  As above, gfn_track is a 16 bit counter, so KVM must not create
285  * more than 2^16-1 upper-level shadow pages at a single gfn, otherwise
286  * gfn_track will overflow and explosions will ensure.
287  *
288  * A unique shadow page (SP) for a gfn is created if and only if an existing SP
289  * cannot be reused.  The ability to reuse a SP is tracked by its role, which
290  * incorporates various mode bits and properties of the SP.  Roughly speaking,
291  * the number of unique SPs that can theoretically be created is 2^n, where n
292  * is the number of bits that are used to compute the role.
293  *
294  * But, even though there are 18 bits in the mask below, not all combinations
295  * of modes and flags are possible.  The maximum number of possible upper-level
296  * shadow pages for a single gfn is in the neighborhood of 2^13.
297  *
298  *   - invalid shadow pages are not accounted.
299  *   - level is effectively limited to four combinations, not 16 as the number
300  *     bits would imply, as 4k SPs are not tracked (allowed to go unsync).
301  *   - level is effectively unused for non-PAE paging because there is exactly
302  *     one upper level (see 4k SP exception above).
303  *   - quadrant is used only for non-PAE paging and is exclusive with
304  *     gpte_is_8_bytes.
305  *   - execonly and ad_disabled are used only for nested EPT, which makes it
306  *     exclusive with quadrant.
307  */
308 union kvm_mmu_page_role {
309 	u32 word;
310 	struct {
311 		unsigned level:4;
312 		unsigned gpte_is_8_bytes:1;
313 		unsigned quadrant:2;
314 		unsigned direct:1;
315 		unsigned access:3;
316 		unsigned invalid:1;
317 		unsigned efer_nx:1;
318 		unsigned cr0_wp:1;
319 		unsigned smep_andnot_wp:1;
320 		unsigned smap_andnot_wp:1;
321 		unsigned ad_disabled:1;
322 		unsigned guest_mode:1;
323 		unsigned :6;
324 
325 		/*
326 		 * This is left at the top of the word so that
327 		 * kvm_memslots_for_spte_role can extract it with a
328 		 * simple shift.  While there is room, give it a whole
329 		 * byte so it is also faster to load it from memory.
330 		 */
331 		unsigned smm:8;
332 	};
333 };
334 
335 /*
336  * kvm_mmu_extended_role complements kvm_mmu_page_role, tracking properties
337  * relevant to the current MMU configuration.   When loading CR0, CR4, or EFER,
338  * including on nested transitions, if nothing in the full role changes then
339  * MMU re-configuration can be skipped. @valid bit is set on first usage so we
340  * don't treat all-zero structure as valid data.
341  *
342  * The properties that are tracked in the extended role but not the page role
343  * are for things that either (a) do not affect the validity of the shadow page
344  * or (b) are indirectly reflected in the shadow page's role.  For example,
345  * CR4.PKE only affects permission checks for software walks of the guest page
346  * tables (because KVM doesn't support Protection Keys with shadow paging), and
347  * CR0.PG, CR4.PAE, and CR4.PSE are indirectly reflected in role.level.
348  *
349  * Note, SMEP and SMAP are not redundant with sm*p_andnot_wp in the page role.
350  * If CR0.WP=1, KVM can reuse shadow pages for the guest regardless of SMEP and
351  * SMAP, but the MMU's permission checks for software walks need to be SMEP and
352  * SMAP aware regardless of CR0.WP.
353  */
354 union kvm_mmu_extended_role {
355 	u32 word;
356 	struct {
357 		unsigned int valid:1;
358 		unsigned int execonly:1;
359 		unsigned int cr0_pg:1;
360 		unsigned int cr4_pae:1;
361 		unsigned int cr4_pse:1;
362 		unsigned int cr4_pke:1;
363 		unsigned int cr4_smap:1;
364 		unsigned int cr4_smep:1;
365 		unsigned int cr4_la57:1;
366 	};
367 };
368 
369 union kvm_mmu_role {
370 	u64 as_u64;
371 	struct {
372 		union kvm_mmu_page_role base;
373 		union kvm_mmu_extended_role ext;
374 	};
375 };
376 
377 struct kvm_rmap_head {
378 	unsigned long val;
379 };
380 
381 struct kvm_pio_request {
382 	unsigned long linear_rip;
383 	unsigned long count;
384 	int in;
385 	int port;
386 	int size;
387 };
388 
389 #define PT64_ROOT_MAX_LEVEL 5
390 
391 struct rsvd_bits_validate {
392 	u64 rsvd_bits_mask[2][PT64_ROOT_MAX_LEVEL];
393 	u64 bad_mt_xwr;
394 };
395 
396 struct kvm_mmu_root_info {
397 	gpa_t pgd;
398 	hpa_t hpa;
399 };
400 
401 #define KVM_MMU_ROOT_INFO_INVALID \
402 	((struct kvm_mmu_root_info) { .pgd = INVALID_PAGE, .hpa = INVALID_PAGE })
403 
404 #define KVM_MMU_NUM_PREV_ROOTS 3
405 
406 #define KVM_HAVE_MMU_RWLOCK
407 
408 struct kvm_mmu_page;
409 struct kvm_page_fault;
410 
411 /*
412  * x86 supports 4 paging modes (5-level 64-bit, 4-level 64-bit, 3-level 32-bit,
413  * and 2-level 32-bit).  The kvm_mmu structure abstracts the details of the
414  * current mmu mode.
415  */
416 struct kvm_mmu {
417 	unsigned long (*get_guest_pgd)(struct kvm_vcpu *vcpu);
418 	u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
419 	int (*page_fault)(struct kvm_vcpu *vcpu, struct kvm_page_fault *fault);
420 	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
421 				  struct x86_exception *fault);
422 	gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gpa_t gva_or_gpa,
423 			    u32 access, struct x86_exception *exception);
424 	gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
425 			       struct x86_exception *exception);
426 	int (*sync_page)(struct kvm_vcpu *vcpu,
427 			 struct kvm_mmu_page *sp);
428 	void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva, hpa_t root_hpa);
429 	hpa_t root_hpa;
430 	gpa_t root_pgd;
431 	union kvm_mmu_role mmu_role;
432 	u8 root_level;
433 	u8 shadow_root_level;
434 	u8 ept_ad;
435 	bool direct_map;
436 	struct kvm_mmu_root_info prev_roots[KVM_MMU_NUM_PREV_ROOTS];
437 
438 	/*
439 	 * Bitmap; bit set = permission fault
440 	 * Byte index: page fault error code [4:1]
441 	 * Bit index: pte permissions in ACC_* format
442 	 */
443 	u8 permissions[16];
444 
445 	/*
446 	* The pkru_mask indicates if protection key checks are needed.  It
447 	* consists of 16 domains indexed by page fault error code bits [4:1],
448 	* with PFEC.RSVD replaced by ACC_USER_MASK from the page tables.
449 	* Each domain has 2 bits which are ANDed with AD and WD from PKRU.
450 	*/
451 	u32 pkru_mask;
452 
453 	u64 *pae_root;
454 	u64 *pml4_root;
455 	u64 *pml5_root;
456 
457 	/*
458 	 * check zero bits on shadow page table entries, these
459 	 * bits include not only hardware reserved bits but also
460 	 * the bits spte never used.
461 	 */
462 	struct rsvd_bits_validate shadow_zero_check;
463 
464 	struct rsvd_bits_validate guest_rsvd_check;
465 
466 	u64 pdptrs[4]; /* pae */
467 };
468 
469 struct kvm_tlb_range {
470 	u64 start_gfn;
471 	u64 pages;
472 };
473 
474 enum pmc_type {
475 	KVM_PMC_GP = 0,
476 	KVM_PMC_FIXED,
477 };
478 
479 struct kvm_pmc {
480 	enum pmc_type type;
481 	u8 idx;
482 	u64 counter;
483 	u64 eventsel;
484 	struct perf_event *perf_event;
485 	struct kvm_vcpu *vcpu;
486 	/*
487 	 * eventsel value for general purpose counters,
488 	 * ctrl value for fixed counters.
489 	 */
490 	u64 current_config;
491 	bool is_paused;
492 };
493 
494 struct kvm_pmu {
495 	unsigned nr_arch_gp_counters;
496 	unsigned nr_arch_fixed_counters;
497 	unsigned available_event_types;
498 	u64 fixed_ctr_ctrl;
499 	u64 global_ctrl;
500 	u64 global_status;
501 	u64 counter_bitmask[2];
502 	u64 global_ctrl_mask;
503 	u64 global_ovf_ctrl_mask;
504 	u64 reserved_bits;
505 	u8 version;
506 	struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
507 	struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED];
508 	struct irq_work irq_work;
509 	DECLARE_BITMAP(reprogram_pmi, X86_PMC_IDX_MAX);
510 	DECLARE_BITMAP(all_valid_pmc_idx, X86_PMC_IDX_MAX);
511 	DECLARE_BITMAP(pmc_in_use, X86_PMC_IDX_MAX);
512 
513 	/*
514 	 * The gate to release perf_events not marked in
515 	 * pmc_in_use only once in a vcpu time slice.
516 	 */
517 	bool need_cleanup;
518 
519 	/*
520 	 * The total number of programmed perf_events and it helps to avoid
521 	 * redundant check before cleanup if guest don't use vPMU at all.
522 	 */
523 	u8 event_count;
524 };
525 
526 struct kvm_pmu_ops;
527 
528 enum {
529 	KVM_DEBUGREG_BP_ENABLED = 1,
530 	KVM_DEBUGREG_WONT_EXIT = 2,
531 };
532 
533 struct kvm_mtrr_range {
534 	u64 base;
535 	u64 mask;
536 	struct list_head node;
537 };
538 
539 struct kvm_mtrr {
540 	struct kvm_mtrr_range var_ranges[KVM_NR_VAR_MTRR];
541 	mtrr_type fixed_ranges[KVM_NR_FIXED_MTRR_REGION];
542 	u64 deftype;
543 
544 	struct list_head head;
545 };
546 
547 /* Hyper-V SynIC timer */
548 struct kvm_vcpu_hv_stimer {
549 	struct hrtimer timer;
550 	int index;
551 	union hv_stimer_config config;
552 	u64 count;
553 	u64 exp_time;
554 	struct hv_message msg;
555 	bool msg_pending;
556 };
557 
558 /* Hyper-V synthetic interrupt controller (SynIC)*/
559 struct kvm_vcpu_hv_synic {
560 	u64 version;
561 	u64 control;
562 	u64 msg_page;
563 	u64 evt_page;
564 	atomic64_t sint[HV_SYNIC_SINT_COUNT];
565 	atomic_t sint_to_gsi[HV_SYNIC_SINT_COUNT];
566 	DECLARE_BITMAP(auto_eoi_bitmap, 256);
567 	DECLARE_BITMAP(vec_bitmap, 256);
568 	bool active;
569 	bool dont_zero_synic_pages;
570 };
571 
572 /* Hyper-V per vcpu emulation context */
573 struct kvm_vcpu_hv {
574 	struct kvm_vcpu *vcpu;
575 	u32 vp_index;
576 	u64 hv_vapic;
577 	s64 runtime_offset;
578 	struct kvm_vcpu_hv_synic synic;
579 	struct kvm_hyperv_exit exit;
580 	struct kvm_vcpu_hv_stimer stimer[HV_SYNIC_STIMER_COUNT];
581 	DECLARE_BITMAP(stimer_pending_bitmap, HV_SYNIC_STIMER_COUNT);
582 	bool enforce_cpuid;
583 	struct {
584 		u32 features_eax; /* HYPERV_CPUID_FEATURES.EAX */
585 		u32 features_ebx; /* HYPERV_CPUID_FEATURES.EBX */
586 		u32 features_edx; /* HYPERV_CPUID_FEATURES.EDX */
587 		u32 enlightenments_eax; /* HYPERV_CPUID_ENLIGHTMENT_INFO.EAX */
588 		u32 enlightenments_ebx; /* HYPERV_CPUID_ENLIGHTMENT_INFO.EBX */
589 		u32 syndbg_cap_eax; /* HYPERV_CPUID_SYNDBG_PLATFORM_CAPABILITIES.EAX */
590 	} cpuid_cache;
591 };
592 
593 /* Xen HVM per vcpu emulation context */
594 struct kvm_vcpu_xen {
595 	u64 hypercall_rip;
596 	u32 current_runstate;
597 	bool vcpu_info_set;
598 	bool vcpu_time_info_set;
599 	bool runstate_set;
600 	struct gfn_to_hva_cache vcpu_info_cache;
601 	struct gfn_to_hva_cache vcpu_time_info_cache;
602 	struct gfn_to_hva_cache runstate_cache;
603 	u64 last_steal;
604 	u64 runstate_entry_time;
605 	u64 runstate_times[4];
606 };
607 
608 struct kvm_vcpu_arch {
609 	/*
610 	 * rip and regs accesses must go through
611 	 * kvm_{register,rip}_{read,write} functions.
612 	 */
613 	unsigned long regs[NR_VCPU_REGS];
614 	u32 regs_avail;
615 	u32 regs_dirty;
616 
617 	unsigned long cr0;
618 	unsigned long cr0_guest_owned_bits;
619 	unsigned long cr2;
620 	unsigned long cr3;
621 	unsigned long cr4;
622 	unsigned long cr4_guest_owned_bits;
623 	unsigned long cr4_guest_rsvd_bits;
624 	unsigned long cr8;
625 	u32 host_pkru;
626 	u32 pkru;
627 	u32 hflags;
628 	u64 efer;
629 	u64 apic_base;
630 	struct kvm_lapic *apic;    /* kernel irqchip context */
631 	bool apicv_active;
632 	bool load_eoi_exitmap_pending;
633 	DECLARE_BITMAP(ioapic_handled_vectors, 256);
634 	unsigned long apic_attention;
635 	int32_t apic_arb_prio;
636 	int mp_state;
637 	u64 ia32_misc_enable_msr;
638 	u64 smbase;
639 	u64 smi_count;
640 	bool tpr_access_reporting;
641 	bool xsaves_enabled;
642 	u64 ia32_xss;
643 	u64 microcode_version;
644 	u64 arch_capabilities;
645 	u64 perf_capabilities;
646 
647 	/*
648 	 * Paging state of the vcpu
649 	 *
650 	 * If the vcpu runs in guest mode with two level paging this still saves
651 	 * the paging mode of the l1 guest. This context is always used to
652 	 * handle faults.
653 	 */
654 	struct kvm_mmu *mmu;
655 
656 	/* Non-nested MMU for L1 */
657 	struct kvm_mmu root_mmu;
658 
659 	/* L1 MMU when running nested */
660 	struct kvm_mmu guest_mmu;
661 
662 	/*
663 	 * Paging state of an L2 guest (used for nested npt)
664 	 *
665 	 * This context will save all necessary information to walk page tables
666 	 * of an L2 guest. This context is only initialized for page table
667 	 * walking and not for faulting since we never handle l2 page faults on
668 	 * the host.
669 	 */
670 	struct kvm_mmu nested_mmu;
671 
672 	/*
673 	 * Pointer to the mmu context currently used for
674 	 * gva_to_gpa translations.
675 	 */
676 	struct kvm_mmu *walk_mmu;
677 
678 	struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
679 	struct kvm_mmu_memory_cache mmu_shadow_page_cache;
680 	struct kvm_mmu_memory_cache mmu_gfn_array_cache;
681 	struct kvm_mmu_memory_cache mmu_page_header_cache;
682 
683 	/*
684 	 * QEMU userspace and the guest each have their own FPU state.
685 	 * In vcpu_run, we switch between the user and guest FPU contexts.
686 	 * While running a VCPU, the VCPU thread will have the guest FPU
687 	 * context.
688 	 *
689 	 * Note that while the PKRU state lives inside the fpu registers,
690 	 * it is switched out separately at VMENTER and VMEXIT time. The
691 	 * "guest_fpstate" state here contains the guest FPU context, with the
692 	 * host PRKU bits.
693 	 */
694 	struct fpu_guest guest_fpu;
695 
696 	u64 xcr0;
697 	u64 guest_supported_xcr0;
698 
699 	struct kvm_pio_request pio;
700 	void *pio_data;
701 	void *sev_pio_data;
702 	unsigned sev_pio_count;
703 
704 	u8 event_exit_inst_len;
705 
706 	struct kvm_queued_exception {
707 		bool pending;
708 		bool injected;
709 		bool has_error_code;
710 		u8 nr;
711 		u32 error_code;
712 		unsigned long payload;
713 		bool has_payload;
714 		u8 nested_apf;
715 	} exception;
716 
717 	struct kvm_queued_interrupt {
718 		bool injected;
719 		bool soft;
720 		u8 nr;
721 	} interrupt;
722 
723 	int halt_request; /* real mode on Intel only */
724 
725 	int cpuid_nent;
726 	struct kvm_cpuid_entry2 *cpuid_entries;
727 	u32 kvm_cpuid_base;
728 
729 	u64 reserved_gpa_bits;
730 	int maxphyaddr;
731 
732 	/* emulate context */
733 
734 	struct x86_emulate_ctxt *emulate_ctxt;
735 	bool emulate_regs_need_sync_to_vcpu;
736 	bool emulate_regs_need_sync_from_vcpu;
737 	int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
738 
739 	gpa_t time;
740 	struct pvclock_vcpu_time_info hv_clock;
741 	unsigned int hw_tsc_khz;
742 	struct gfn_to_hva_cache pv_time;
743 	bool pv_time_enabled;
744 	/* set guest stopped flag in pvclock flags field */
745 	bool pvclock_set_guest_stopped_request;
746 
747 	struct {
748 		u8 preempted;
749 		u64 msr_val;
750 		u64 last_steal;
751 		struct gfn_to_hva_cache cache;
752 	} st;
753 
754 	u64 l1_tsc_offset;
755 	u64 tsc_offset; /* current tsc offset */
756 	u64 last_guest_tsc;
757 	u64 last_host_tsc;
758 	u64 tsc_offset_adjustment;
759 	u64 this_tsc_nsec;
760 	u64 this_tsc_write;
761 	u64 this_tsc_generation;
762 	bool tsc_catchup;
763 	bool tsc_always_catchup;
764 	s8 virtual_tsc_shift;
765 	u32 virtual_tsc_mult;
766 	u32 virtual_tsc_khz;
767 	s64 ia32_tsc_adjust_msr;
768 	u64 msr_ia32_power_ctl;
769 	u64 l1_tsc_scaling_ratio;
770 	u64 tsc_scaling_ratio; /* current scaling ratio */
771 
772 	atomic_t nmi_queued;  /* unprocessed asynchronous NMIs */
773 	unsigned nmi_pending; /* NMI queued after currently running handler */
774 	bool nmi_injected;    /* Trying to inject an NMI this entry */
775 	bool smi_pending;    /* SMI queued after currently running handler */
776 
777 	struct kvm_mtrr mtrr_state;
778 	u64 pat;
779 
780 	unsigned switch_db_regs;
781 	unsigned long db[KVM_NR_DB_REGS];
782 	unsigned long dr6;
783 	unsigned long dr7;
784 	unsigned long eff_db[KVM_NR_DB_REGS];
785 	unsigned long guest_debug_dr7;
786 	u64 msr_platform_info;
787 	u64 msr_misc_features_enables;
788 
789 	u64 mcg_cap;
790 	u64 mcg_status;
791 	u64 mcg_ctl;
792 	u64 mcg_ext_ctl;
793 	u64 *mce_banks;
794 
795 	/* Cache MMIO info */
796 	u64 mmio_gva;
797 	unsigned mmio_access;
798 	gfn_t mmio_gfn;
799 	u64 mmio_gen;
800 
801 	struct kvm_pmu pmu;
802 
803 	/* used for guest single stepping over the given code position */
804 	unsigned long singlestep_rip;
805 
806 	bool hyperv_enabled;
807 	struct kvm_vcpu_hv *hyperv;
808 	struct kvm_vcpu_xen xen;
809 
810 	cpumask_var_t wbinvd_dirty_mask;
811 
812 	unsigned long last_retry_eip;
813 	unsigned long last_retry_addr;
814 
815 	struct {
816 		bool halted;
817 		gfn_t gfns[ASYNC_PF_PER_VCPU];
818 		struct gfn_to_hva_cache data;
819 		u64 msr_en_val; /* MSR_KVM_ASYNC_PF_EN */
820 		u64 msr_int_val; /* MSR_KVM_ASYNC_PF_INT */
821 		u16 vec;
822 		u32 id;
823 		bool send_user_only;
824 		u32 host_apf_flags;
825 		unsigned long nested_apf_token;
826 		bool delivery_as_pf_vmexit;
827 		bool pageready_pending;
828 	} apf;
829 
830 	/* OSVW MSRs (AMD only) */
831 	struct {
832 		u64 length;
833 		u64 status;
834 	} osvw;
835 
836 	struct {
837 		u64 msr_val;
838 		struct gfn_to_hva_cache data;
839 	} pv_eoi;
840 
841 	u64 msr_kvm_poll_control;
842 
843 	/*
844 	 * Indicates the guest is trying to write a gfn that contains one or
845 	 * more of the PTEs used to translate the write itself, i.e. the access
846 	 * is changing its own translation in the guest page tables.  KVM exits
847 	 * to userspace if emulation of the faulting instruction fails and this
848 	 * flag is set, as KVM cannot make forward progress.
849 	 *
850 	 * If emulation fails for a write to guest page tables, KVM unprotects
851 	 * (zaps) the shadow page for the target gfn and resumes the guest to
852 	 * retry the non-emulatable instruction (on hardware).  Unprotecting the
853 	 * gfn doesn't allow forward progress for a self-changing access because
854 	 * doing so also zaps the translation for the gfn, i.e. retrying the
855 	 * instruction will hit a !PRESENT fault, which results in a new shadow
856 	 * page and sends KVM back to square one.
857 	 */
858 	bool write_fault_to_shadow_pgtable;
859 
860 	/* set at EPT violation at this point */
861 	unsigned long exit_qualification;
862 
863 	/* pv related host specific info */
864 	struct {
865 		bool pv_unhalted;
866 	} pv;
867 
868 	int pending_ioapic_eoi;
869 	int pending_external_vector;
870 
871 	/* be preempted when it's in kernel-mode(cpl=0) */
872 	bool preempted_in_kernel;
873 
874 	/* Flush the L1 Data cache for L1TF mitigation on VMENTER */
875 	bool l1tf_flush_l1d;
876 
877 	/* Host CPU on which VM-entry was most recently attempted */
878 	int last_vmentry_cpu;
879 
880 	/* AMD MSRC001_0015 Hardware Configuration */
881 	u64 msr_hwcr;
882 
883 	/* pv related cpuid info */
884 	struct {
885 		/*
886 		 * value of the eax register in the KVM_CPUID_FEATURES CPUID
887 		 * leaf.
888 		 */
889 		u32 features;
890 
891 		/*
892 		 * indicates whether pv emulation should be disabled if features
893 		 * are not present in the guest's cpuid
894 		 */
895 		bool enforce;
896 	} pv_cpuid;
897 
898 	/* Protected Guests */
899 	bool guest_state_protected;
900 
901 	/*
902 	 * Set when PDPTS were loaded directly by the userspace without
903 	 * reading the guest memory
904 	 */
905 	bool pdptrs_from_userspace;
906 
907 #if IS_ENABLED(CONFIG_HYPERV)
908 	hpa_t hv_root_tdp;
909 #endif
910 };
911 
912 struct kvm_lpage_info {
913 	int disallow_lpage;
914 };
915 
916 struct kvm_arch_memory_slot {
917 	struct kvm_rmap_head *rmap[KVM_NR_PAGE_SIZES];
918 	struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
919 	unsigned short *gfn_track[KVM_PAGE_TRACK_MAX];
920 };
921 
922 /*
923  * We use as the mode the number of bits allocated in the LDR for the
924  * logical processor ID.  It happens that these are all powers of two.
925  * This makes it is very easy to detect cases where the APICs are
926  * configured for multiple modes; in that case, we cannot use the map and
927  * hence cannot use kvm_irq_delivery_to_apic_fast either.
928  */
929 #define KVM_APIC_MODE_XAPIC_CLUSTER          4
930 #define KVM_APIC_MODE_XAPIC_FLAT             8
931 #define KVM_APIC_MODE_X2APIC                16
932 
933 struct kvm_apic_map {
934 	struct rcu_head rcu;
935 	u8 mode;
936 	u32 max_apic_id;
937 	union {
938 		struct kvm_lapic *xapic_flat_map[8];
939 		struct kvm_lapic *xapic_cluster_map[16][4];
940 	};
941 	struct kvm_lapic *phys_map[];
942 };
943 
944 /* Hyper-V synthetic debugger (SynDbg)*/
945 struct kvm_hv_syndbg {
946 	struct {
947 		u64 control;
948 		u64 status;
949 		u64 send_page;
950 		u64 recv_page;
951 		u64 pending_page;
952 	} control;
953 	u64 options;
954 };
955 
956 /* Current state of Hyper-V TSC page clocksource */
957 enum hv_tsc_page_status {
958 	/* TSC page was not set up or disabled */
959 	HV_TSC_PAGE_UNSET = 0,
960 	/* TSC page MSR was written by the guest, update pending */
961 	HV_TSC_PAGE_GUEST_CHANGED,
962 	/* TSC page MSR was written by KVM userspace, update pending */
963 	HV_TSC_PAGE_HOST_CHANGED,
964 	/* TSC page was properly set up and is currently active  */
965 	HV_TSC_PAGE_SET,
966 	/* TSC page is currently being updated and therefore is inactive */
967 	HV_TSC_PAGE_UPDATING,
968 	/* TSC page was set up with an inaccessible GPA */
969 	HV_TSC_PAGE_BROKEN,
970 };
971 
972 /* Hyper-V emulation context */
973 struct kvm_hv {
974 	struct mutex hv_lock;
975 	u64 hv_guest_os_id;
976 	u64 hv_hypercall;
977 	u64 hv_tsc_page;
978 	enum hv_tsc_page_status hv_tsc_page_status;
979 
980 	/* Hyper-v based guest crash (NT kernel bugcheck) parameters */
981 	u64 hv_crash_param[HV_X64_MSR_CRASH_PARAMS];
982 	u64 hv_crash_ctl;
983 
984 	struct ms_hyperv_tsc_page tsc_ref;
985 
986 	struct idr conn_to_evt;
987 
988 	u64 hv_reenlightenment_control;
989 	u64 hv_tsc_emulation_control;
990 	u64 hv_tsc_emulation_status;
991 
992 	/* How many vCPUs have VP index != vCPU index */
993 	atomic_t num_mismatched_vp_indexes;
994 
995 	/*
996 	 * How many SynICs use 'AutoEOI' feature
997 	 * (protected by arch.apicv_update_lock)
998 	 */
999 	unsigned int synic_auto_eoi_used;
1000 
1001 	struct hv_partition_assist_pg *hv_pa_pg;
1002 	struct kvm_hv_syndbg hv_syndbg;
1003 };
1004 
1005 struct msr_bitmap_range {
1006 	u32 flags;
1007 	u32 nmsrs;
1008 	u32 base;
1009 	unsigned long *bitmap;
1010 };
1011 
1012 /* Xen emulation context */
1013 struct kvm_xen {
1014 	bool long_mode;
1015 	u8 upcall_vector;
1016 	gfn_t shinfo_gfn;
1017 };
1018 
1019 enum kvm_irqchip_mode {
1020 	KVM_IRQCHIP_NONE,
1021 	KVM_IRQCHIP_KERNEL,       /* created with KVM_CREATE_IRQCHIP */
1022 	KVM_IRQCHIP_SPLIT,        /* created with KVM_CAP_SPLIT_IRQCHIP */
1023 };
1024 
1025 struct kvm_x86_msr_filter {
1026 	u8 count;
1027 	bool default_allow:1;
1028 	struct msr_bitmap_range ranges[16];
1029 };
1030 
1031 #define APICV_INHIBIT_REASON_DISABLE    0
1032 #define APICV_INHIBIT_REASON_HYPERV     1
1033 #define APICV_INHIBIT_REASON_NESTED     2
1034 #define APICV_INHIBIT_REASON_IRQWIN     3
1035 #define APICV_INHIBIT_REASON_PIT_REINJ  4
1036 #define APICV_INHIBIT_REASON_X2APIC	5
1037 #define APICV_INHIBIT_REASON_BLOCKIRQ	6
1038 
1039 struct kvm_arch {
1040 	unsigned long n_used_mmu_pages;
1041 	unsigned long n_requested_mmu_pages;
1042 	unsigned long n_max_mmu_pages;
1043 	unsigned int indirect_shadow_pages;
1044 	u8 mmu_valid_gen;
1045 	struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
1046 	struct list_head active_mmu_pages;
1047 	struct list_head zapped_obsolete_pages;
1048 	struct list_head lpage_disallowed_mmu_pages;
1049 	struct kvm_page_track_notifier_node mmu_sp_tracker;
1050 	struct kvm_page_track_notifier_head track_notifier_head;
1051 	/*
1052 	 * Protects marking pages unsync during page faults, as TDP MMU page
1053 	 * faults only take mmu_lock for read.  For simplicity, the unsync
1054 	 * pages lock is always taken when marking pages unsync regardless of
1055 	 * whether mmu_lock is held for read or write.
1056 	 */
1057 	spinlock_t mmu_unsync_pages_lock;
1058 
1059 	struct list_head assigned_dev_head;
1060 	struct iommu_domain *iommu_domain;
1061 	bool iommu_noncoherent;
1062 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA
1063 	atomic_t noncoherent_dma_count;
1064 #define __KVM_HAVE_ARCH_ASSIGNED_DEVICE
1065 	atomic_t assigned_device_count;
1066 	struct kvm_pic *vpic;
1067 	struct kvm_ioapic *vioapic;
1068 	struct kvm_pit *vpit;
1069 	atomic_t vapics_in_nmi_mode;
1070 	struct mutex apic_map_lock;
1071 	struct kvm_apic_map __rcu *apic_map;
1072 	atomic_t apic_map_dirty;
1073 
1074 	/* Protects apic_access_memslot_enabled and apicv_inhibit_reasons */
1075 	struct rw_semaphore apicv_update_lock;
1076 
1077 	bool apic_access_memslot_enabled;
1078 	unsigned long apicv_inhibit_reasons;
1079 
1080 	gpa_t wall_clock;
1081 
1082 	bool mwait_in_guest;
1083 	bool hlt_in_guest;
1084 	bool pause_in_guest;
1085 	bool cstate_in_guest;
1086 
1087 	unsigned long irq_sources_bitmap;
1088 	s64 kvmclock_offset;
1089 
1090 	/*
1091 	 * This also protects nr_vcpus_matched_tsc which is read from a
1092 	 * preemption-disabled region, so it must be a raw spinlock.
1093 	 */
1094 	raw_spinlock_t tsc_write_lock;
1095 	u64 last_tsc_nsec;
1096 	u64 last_tsc_write;
1097 	u32 last_tsc_khz;
1098 	u64 last_tsc_offset;
1099 	u64 cur_tsc_nsec;
1100 	u64 cur_tsc_write;
1101 	u64 cur_tsc_offset;
1102 	u64 cur_tsc_generation;
1103 	int nr_vcpus_matched_tsc;
1104 
1105 	seqcount_raw_spinlock_t pvclock_sc;
1106 	bool use_master_clock;
1107 	u64 master_kernel_ns;
1108 	u64 master_cycle_now;
1109 	struct delayed_work kvmclock_update_work;
1110 	struct delayed_work kvmclock_sync_work;
1111 
1112 	struct kvm_xen_hvm_config xen_hvm_config;
1113 
1114 	/* reads protected by irq_srcu, writes by irq_lock */
1115 	struct hlist_head mask_notifier_list;
1116 
1117 	struct kvm_hv hyperv;
1118 	struct kvm_xen xen;
1119 
1120 	#ifdef CONFIG_KVM_MMU_AUDIT
1121 	int audit_point;
1122 	#endif
1123 
1124 	bool backwards_tsc_observed;
1125 	bool boot_vcpu_runs_old_kvmclock;
1126 	u32 bsp_vcpu_id;
1127 
1128 	u64 disabled_quirks;
1129 	int cpu_dirty_logging_count;
1130 
1131 	enum kvm_irqchip_mode irqchip_mode;
1132 	u8 nr_reserved_ioapic_pins;
1133 
1134 	bool disabled_lapic_found;
1135 
1136 	bool x2apic_format;
1137 	bool x2apic_broadcast_quirk_disabled;
1138 
1139 	bool guest_can_read_msr_platform_info;
1140 	bool exception_payload_enabled;
1141 
1142 	bool bus_lock_detection_enabled;
1143 	/*
1144 	 * If exit_on_emulation_error is set, and the in-kernel instruction
1145 	 * emulator fails to emulate an instruction, allow userspace
1146 	 * the opportunity to look at it.
1147 	 */
1148 	bool exit_on_emulation_error;
1149 
1150 	/* Deflect RDMSR and WRMSR to user space when they trigger a #GP */
1151 	u32 user_space_msr_mask;
1152 	struct kvm_x86_msr_filter __rcu *msr_filter;
1153 
1154 	u32 hypercall_exit_enabled;
1155 
1156 	/* Guest can access the SGX PROVISIONKEY. */
1157 	bool sgx_provisioning_allowed;
1158 
1159 	struct kvm_pmu_event_filter __rcu *pmu_event_filter;
1160 	struct task_struct *nx_lpage_recovery_thread;
1161 
1162 #ifdef CONFIG_X86_64
1163 	/*
1164 	 * Whether the TDP MMU is enabled for this VM. This contains a
1165 	 * snapshot of the TDP MMU module parameter from when the VM was
1166 	 * created and remains unchanged for the life of the VM. If this is
1167 	 * true, TDP MMU handler functions will run for various MMU
1168 	 * operations.
1169 	 */
1170 	bool tdp_mmu_enabled;
1171 
1172 	/*
1173 	 * List of struct kvm_mmu_pages being used as roots.
1174 	 * All struct kvm_mmu_pages in the list should have
1175 	 * tdp_mmu_page set.
1176 	 *
1177 	 * For reads, this list is protected by:
1178 	 *	the MMU lock in read mode + RCU or
1179 	 *	the MMU lock in write mode
1180 	 *
1181 	 * For writes, this list is protected by:
1182 	 *	the MMU lock in read mode + the tdp_mmu_pages_lock or
1183 	 *	the MMU lock in write mode
1184 	 *
1185 	 * Roots will remain in the list until their tdp_mmu_root_count
1186 	 * drops to zero, at which point the thread that decremented the
1187 	 * count to zero should removed the root from the list and clean
1188 	 * it up, freeing the root after an RCU grace period.
1189 	 */
1190 	struct list_head tdp_mmu_roots;
1191 
1192 	/*
1193 	 * List of struct kvmp_mmu_pages not being used as roots.
1194 	 * All struct kvm_mmu_pages in the list should have
1195 	 * tdp_mmu_page set and a tdp_mmu_root_count of 0.
1196 	 */
1197 	struct list_head tdp_mmu_pages;
1198 
1199 	/*
1200 	 * Protects accesses to the following fields when the MMU lock
1201 	 * is held in read mode:
1202 	 *  - tdp_mmu_roots (above)
1203 	 *  - tdp_mmu_pages (above)
1204 	 *  - the link field of struct kvm_mmu_pages used by the TDP MMU
1205 	 *  - lpage_disallowed_mmu_pages
1206 	 *  - the lpage_disallowed_link field of struct kvm_mmu_pages used
1207 	 *    by the TDP MMU
1208 	 * It is acceptable, but not necessary, to acquire this lock when
1209 	 * the thread holds the MMU lock in write mode.
1210 	 */
1211 	spinlock_t tdp_mmu_pages_lock;
1212 #endif /* CONFIG_X86_64 */
1213 
1214 	/*
1215 	 * If set, at least one shadow root has been allocated. This flag
1216 	 * is used as one input when determining whether certain memslot
1217 	 * related allocations are necessary.
1218 	 */
1219 	bool shadow_root_allocated;
1220 
1221 #if IS_ENABLED(CONFIG_HYPERV)
1222 	hpa_t	hv_root_tdp;
1223 	spinlock_t hv_root_tdp_lock;
1224 #endif
1225 };
1226 
1227 struct kvm_vm_stat {
1228 	struct kvm_vm_stat_generic generic;
1229 	u64 mmu_shadow_zapped;
1230 	u64 mmu_pte_write;
1231 	u64 mmu_pde_zapped;
1232 	u64 mmu_flooded;
1233 	u64 mmu_recycled;
1234 	u64 mmu_cache_miss;
1235 	u64 mmu_unsync;
1236 	union {
1237 		struct {
1238 			atomic64_t pages_4k;
1239 			atomic64_t pages_2m;
1240 			atomic64_t pages_1g;
1241 		};
1242 		atomic64_t pages[KVM_NR_PAGE_SIZES];
1243 	};
1244 	u64 nx_lpage_splits;
1245 	u64 max_mmu_page_hash_collisions;
1246 	u64 max_mmu_rmap_size;
1247 };
1248 
1249 struct kvm_vcpu_stat {
1250 	struct kvm_vcpu_stat_generic generic;
1251 	u64 pf_fixed;
1252 	u64 pf_guest;
1253 	u64 tlb_flush;
1254 	u64 invlpg;
1255 
1256 	u64 exits;
1257 	u64 io_exits;
1258 	u64 mmio_exits;
1259 	u64 signal_exits;
1260 	u64 irq_window_exits;
1261 	u64 nmi_window_exits;
1262 	u64 l1d_flush;
1263 	u64 halt_exits;
1264 	u64 request_irq_exits;
1265 	u64 irq_exits;
1266 	u64 host_state_reload;
1267 	u64 fpu_reload;
1268 	u64 insn_emulation;
1269 	u64 insn_emulation_fail;
1270 	u64 hypercalls;
1271 	u64 irq_injections;
1272 	u64 nmi_injections;
1273 	u64 req_event;
1274 	u64 nested_run;
1275 	u64 directed_yield_attempted;
1276 	u64 directed_yield_successful;
1277 	u64 guest_mode;
1278 };
1279 
1280 struct x86_instruction_info;
1281 
1282 struct msr_data {
1283 	bool host_initiated;
1284 	u32 index;
1285 	u64 data;
1286 };
1287 
1288 struct kvm_lapic_irq {
1289 	u32 vector;
1290 	u16 delivery_mode;
1291 	u16 dest_mode;
1292 	bool level;
1293 	u16 trig_mode;
1294 	u32 shorthand;
1295 	u32 dest_id;
1296 	bool msi_redir_hint;
1297 };
1298 
1299 static inline u16 kvm_lapic_irq_dest_mode(bool dest_mode_logical)
1300 {
1301 	return dest_mode_logical ? APIC_DEST_LOGICAL : APIC_DEST_PHYSICAL;
1302 }
1303 
1304 struct kvm_x86_ops {
1305 	const char *name;
1306 
1307 	int (*hardware_enable)(void);
1308 	void (*hardware_disable)(void);
1309 	void (*hardware_unsetup)(void);
1310 	bool (*cpu_has_accelerated_tpr)(void);
1311 	bool (*has_emulated_msr)(struct kvm *kvm, u32 index);
1312 	void (*vcpu_after_set_cpuid)(struct kvm_vcpu *vcpu);
1313 
1314 	unsigned int vm_size;
1315 	int (*vm_init)(struct kvm *kvm);
1316 	void (*vm_destroy)(struct kvm *kvm);
1317 
1318 	/* Create, but do not attach this VCPU */
1319 	int (*vcpu_create)(struct kvm_vcpu *vcpu);
1320 	void (*vcpu_free)(struct kvm_vcpu *vcpu);
1321 	void (*vcpu_reset)(struct kvm_vcpu *vcpu, bool init_event);
1322 
1323 	void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
1324 	void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
1325 	void (*vcpu_put)(struct kvm_vcpu *vcpu);
1326 
1327 	void (*update_exception_bitmap)(struct kvm_vcpu *vcpu);
1328 	int (*get_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1329 	int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1330 	u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
1331 	void (*get_segment)(struct kvm_vcpu *vcpu,
1332 			    struct kvm_segment *var, int seg);
1333 	int (*get_cpl)(struct kvm_vcpu *vcpu);
1334 	void (*set_segment)(struct kvm_vcpu *vcpu,
1335 			    struct kvm_segment *var, int seg);
1336 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
1337 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
1338 	bool (*is_valid_cr4)(struct kvm_vcpu *vcpu, unsigned long cr0);
1339 	void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
1340 	int (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
1341 	void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1342 	void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1343 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1344 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1345 	void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu);
1346 	void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
1347 	void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
1348 	unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
1349 	void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
1350 
1351 	void (*tlb_flush_all)(struct kvm_vcpu *vcpu);
1352 	void (*tlb_flush_current)(struct kvm_vcpu *vcpu);
1353 	int  (*tlb_remote_flush)(struct kvm *kvm);
1354 	int  (*tlb_remote_flush_with_range)(struct kvm *kvm,
1355 			struct kvm_tlb_range *range);
1356 
1357 	/*
1358 	 * Flush any TLB entries associated with the given GVA.
1359 	 * Does not need to flush GPA->HPA mappings.
1360 	 * Can potentially get non-canonical addresses through INVLPGs, which
1361 	 * the implementation may choose to ignore if appropriate.
1362 	 */
1363 	void (*tlb_flush_gva)(struct kvm_vcpu *vcpu, gva_t addr);
1364 
1365 	/*
1366 	 * Flush any TLB entries created by the guest.  Like tlb_flush_gva(),
1367 	 * does not need to flush GPA->HPA mappings.
1368 	 */
1369 	void (*tlb_flush_guest)(struct kvm_vcpu *vcpu);
1370 
1371 	enum exit_fastpath_completion (*run)(struct kvm_vcpu *vcpu);
1372 	int (*handle_exit)(struct kvm_vcpu *vcpu,
1373 		enum exit_fastpath_completion exit_fastpath);
1374 	int (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
1375 	void (*update_emulated_instruction)(struct kvm_vcpu *vcpu);
1376 	void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
1377 	u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu);
1378 	void (*patch_hypercall)(struct kvm_vcpu *vcpu,
1379 				unsigned char *hypercall_addr);
1380 	void (*set_irq)(struct kvm_vcpu *vcpu);
1381 	void (*set_nmi)(struct kvm_vcpu *vcpu);
1382 	void (*queue_exception)(struct kvm_vcpu *vcpu);
1383 	void (*cancel_injection)(struct kvm_vcpu *vcpu);
1384 	int (*interrupt_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1385 	int (*nmi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1386 	bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
1387 	void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
1388 	void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
1389 	void (*enable_irq_window)(struct kvm_vcpu *vcpu);
1390 	void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
1391 	bool (*check_apicv_inhibit_reasons)(ulong bit);
1392 	void (*refresh_apicv_exec_ctrl)(struct kvm_vcpu *vcpu);
1393 	void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
1394 	void (*hwapic_isr_update)(struct kvm_vcpu *vcpu, int isr);
1395 	bool (*guest_apic_has_interrupt)(struct kvm_vcpu *vcpu);
1396 	void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
1397 	void (*set_virtual_apic_mode)(struct kvm_vcpu *vcpu);
1398 	void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu);
1399 	int (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
1400 	int (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
1401 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
1402 	int (*set_identity_map_addr)(struct kvm *kvm, u64 ident_addr);
1403 	u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
1404 
1405 	void (*load_mmu_pgd)(struct kvm_vcpu *vcpu, hpa_t root_hpa,
1406 			     int root_level);
1407 
1408 	bool (*has_wbinvd_exit)(void);
1409 
1410 	u64 (*get_l2_tsc_offset)(struct kvm_vcpu *vcpu);
1411 	u64 (*get_l2_tsc_multiplier)(struct kvm_vcpu *vcpu);
1412 	void (*write_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
1413 	void (*write_tsc_multiplier)(struct kvm_vcpu *vcpu, u64 multiplier);
1414 
1415 	/*
1416 	 * Retrieve somewhat arbitrary exit information.  Intended to
1417 	 * be used only from within tracepoints or error paths.
1418 	 */
1419 	void (*get_exit_info)(struct kvm_vcpu *vcpu, u32 *reason,
1420 			      u64 *info1, u64 *info2,
1421 			      u32 *exit_int_info, u32 *exit_int_info_err_code);
1422 
1423 	int (*check_intercept)(struct kvm_vcpu *vcpu,
1424 			       struct x86_instruction_info *info,
1425 			       enum x86_intercept_stage stage,
1426 			       struct x86_exception *exception);
1427 	void (*handle_exit_irqoff)(struct kvm_vcpu *vcpu);
1428 
1429 	void (*request_immediate_exit)(struct kvm_vcpu *vcpu);
1430 
1431 	void (*sched_in)(struct kvm_vcpu *kvm, int cpu);
1432 
1433 	/*
1434 	 * Size of the CPU's dirty log buffer, i.e. VMX's PML buffer.  A zero
1435 	 * value indicates CPU dirty logging is unsupported or disabled.
1436 	 */
1437 	int cpu_dirty_log_size;
1438 	void (*update_cpu_dirty_logging)(struct kvm_vcpu *vcpu);
1439 
1440 	/* pmu operations of sub-arch */
1441 	const struct kvm_pmu_ops *pmu_ops;
1442 	const struct kvm_x86_nested_ops *nested_ops;
1443 
1444 	/*
1445 	 * Architecture specific hooks for vCPU blocking due to
1446 	 * HLT instruction.
1447 	 * Returns for .pre_block():
1448 	 *    - 0 means continue to block the vCPU.
1449 	 *    - 1 means we cannot block the vCPU since some event
1450 	 *        happens during this period, such as, 'ON' bit in
1451 	 *        posted-interrupts descriptor is set.
1452 	 */
1453 	int (*pre_block)(struct kvm_vcpu *vcpu);
1454 	void (*post_block)(struct kvm_vcpu *vcpu);
1455 
1456 	void (*vcpu_blocking)(struct kvm_vcpu *vcpu);
1457 	void (*vcpu_unblocking)(struct kvm_vcpu *vcpu);
1458 
1459 	int (*update_pi_irte)(struct kvm *kvm, unsigned int host_irq,
1460 			      uint32_t guest_irq, bool set);
1461 	void (*start_assignment)(struct kvm *kvm);
1462 	void (*apicv_post_state_restore)(struct kvm_vcpu *vcpu);
1463 	bool (*dy_apicv_has_pending_interrupt)(struct kvm_vcpu *vcpu);
1464 
1465 	int (*set_hv_timer)(struct kvm_vcpu *vcpu, u64 guest_deadline_tsc,
1466 			    bool *expired);
1467 	void (*cancel_hv_timer)(struct kvm_vcpu *vcpu);
1468 
1469 	void (*setup_mce)(struct kvm_vcpu *vcpu);
1470 
1471 	int (*smi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1472 	int (*enter_smm)(struct kvm_vcpu *vcpu, char *smstate);
1473 	int (*leave_smm)(struct kvm_vcpu *vcpu, const char *smstate);
1474 	void (*enable_smi_window)(struct kvm_vcpu *vcpu);
1475 
1476 	int (*mem_enc_op)(struct kvm *kvm, void __user *argp);
1477 	int (*mem_enc_reg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1478 	int (*mem_enc_unreg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1479 	int (*vm_copy_enc_context_from)(struct kvm *kvm, unsigned int source_fd);
1480 	int (*vm_move_enc_context_from)(struct kvm *kvm, unsigned int source_fd);
1481 
1482 	int (*get_msr_feature)(struct kvm_msr_entry *entry);
1483 
1484 	bool (*can_emulate_instruction)(struct kvm_vcpu *vcpu, void *insn, int insn_len);
1485 
1486 	bool (*apic_init_signal_blocked)(struct kvm_vcpu *vcpu);
1487 	int (*enable_direct_tlbflush)(struct kvm_vcpu *vcpu);
1488 
1489 	void (*migrate_timers)(struct kvm_vcpu *vcpu);
1490 	void (*msr_filter_changed)(struct kvm_vcpu *vcpu);
1491 	int (*complete_emulated_msr)(struct kvm_vcpu *vcpu, int err);
1492 
1493 	void (*vcpu_deliver_sipi_vector)(struct kvm_vcpu *vcpu, u8 vector);
1494 };
1495 
1496 struct kvm_x86_nested_ops {
1497 	int (*check_events)(struct kvm_vcpu *vcpu);
1498 	bool (*hv_timer_pending)(struct kvm_vcpu *vcpu);
1499 	void (*triple_fault)(struct kvm_vcpu *vcpu);
1500 	int (*get_state)(struct kvm_vcpu *vcpu,
1501 			 struct kvm_nested_state __user *user_kvm_nested_state,
1502 			 unsigned user_data_size);
1503 	int (*set_state)(struct kvm_vcpu *vcpu,
1504 			 struct kvm_nested_state __user *user_kvm_nested_state,
1505 			 struct kvm_nested_state *kvm_state);
1506 	bool (*get_nested_state_pages)(struct kvm_vcpu *vcpu);
1507 	int (*write_log_dirty)(struct kvm_vcpu *vcpu, gpa_t l2_gpa);
1508 
1509 	int (*enable_evmcs)(struct kvm_vcpu *vcpu,
1510 			    uint16_t *vmcs_version);
1511 	uint16_t (*get_evmcs_version)(struct kvm_vcpu *vcpu);
1512 };
1513 
1514 struct kvm_x86_init_ops {
1515 	int (*cpu_has_kvm_support)(void);
1516 	int (*disabled_by_bios)(void);
1517 	int (*check_processor_compatibility)(void);
1518 	int (*hardware_setup)(void);
1519 
1520 	struct kvm_x86_ops *runtime_ops;
1521 };
1522 
1523 struct kvm_arch_async_pf {
1524 	u32 token;
1525 	gfn_t gfn;
1526 	unsigned long cr3;
1527 	bool direct_map;
1528 };
1529 
1530 extern u32 __read_mostly kvm_nr_uret_msrs;
1531 extern u64 __read_mostly host_efer;
1532 extern bool __read_mostly allow_smaller_maxphyaddr;
1533 extern bool __read_mostly enable_apicv;
1534 extern struct kvm_x86_ops kvm_x86_ops;
1535 
1536 #define KVM_X86_OP(func) \
1537 	DECLARE_STATIC_CALL(kvm_x86_##func, *(((struct kvm_x86_ops *)0)->func));
1538 #define KVM_X86_OP_NULL KVM_X86_OP
1539 #include <asm/kvm-x86-ops.h>
1540 
1541 static inline void kvm_ops_static_call_update(void)
1542 {
1543 #define KVM_X86_OP(func) \
1544 	static_call_update(kvm_x86_##func, kvm_x86_ops.func);
1545 #define KVM_X86_OP_NULL KVM_X86_OP
1546 #include <asm/kvm-x86-ops.h>
1547 }
1548 
1549 #define __KVM_HAVE_ARCH_VM_ALLOC
1550 static inline struct kvm *kvm_arch_alloc_vm(void)
1551 {
1552 	return __vmalloc(kvm_x86_ops.vm_size, GFP_KERNEL_ACCOUNT | __GFP_ZERO);
1553 }
1554 
1555 #define __KVM_HAVE_ARCH_VM_FREE
1556 void kvm_arch_free_vm(struct kvm *kvm);
1557 
1558 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
1559 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
1560 {
1561 	if (kvm_x86_ops.tlb_remote_flush &&
1562 	    !static_call(kvm_x86_tlb_remote_flush)(kvm))
1563 		return 0;
1564 	else
1565 		return -ENOTSUPP;
1566 }
1567 
1568 int kvm_mmu_module_init(void);
1569 void kvm_mmu_module_exit(void);
1570 
1571 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
1572 int kvm_mmu_create(struct kvm_vcpu *vcpu);
1573 void kvm_mmu_init_vm(struct kvm *kvm);
1574 void kvm_mmu_uninit_vm(struct kvm *kvm);
1575 
1576 void kvm_mmu_after_set_cpuid(struct kvm_vcpu *vcpu);
1577 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
1578 void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
1579 				      const struct kvm_memory_slot *memslot,
1580 				      int start_level);
1581 void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
1582 				   const struct kvm_memory_slot *memslot);
1583 void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
1584 				   const struct kvm_memory_slot *memslot);
1585 void kvm_mmu_zap_all(struct kvm *kvm);
1586 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, u64 gen);
1587 unsigned long kvm_mmu_calculate_default_mmu_pages(struct kvm *kvm);
1588 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned long kvm_nr_mmu_pages);
1589 
1590 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
1591 
1592 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
1593 			  const void *val, int bytes);
1594 
1595 struct kvm_irq_mask_notifier {
1596 	void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
1597 	int irq;
1598 	struct hlist_node link;
1599 };
1600 
1601 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
1602 				    struct kvm_irq_mask_notifier *kimn);
1603 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
1604 				      struct kvm_irq_mask_notifier *kimn);
1605 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
1606 			     bool mask);
1607 
1608 extern bool tdp_enabled;
1609 
1610 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
1611 
1612 /* control of guest tsc rate supported? */
1613 extern bool kvm_has_tsc_control;
1614 /* maximum supported tsc_khz for guests */
1615 extern u32  kvm_max_guest_tsc_khz;
1616 /* number of bits of the fractional part of the TSC scaling ratio */
1617 extern u8   kvm_tsc_scaling_ratio_frac_bits;
1618 /* maximum allowed value of TSC scaling ratio */
1619 extern u64  kvm_max_tsc_scaling_ratio;
1620 /* 1ull << kvm_tsc_scaling_ratio_frac_bits */
1621 extern u64  kvm_default_tsc_scaling_ratio;
1622 /* bus lock detection supported? */
1623 extern bool kvm_has_bus_lock_exit;
1624 
1625 extern u64 kvm_mce_cap_supported;
1626 
1627 /*
1628  * EMULTYPE_NO_DECODE - Set when re-emulating an instruction (after completing
1629  *			userspace I/O) to indicate that the emulation context
1630  *			should be reused as is, i.e. skip initialization of
1631  *			emulation context, instruction fetch and decode.
1632  *
1633  * EMULTYPE_TRAP_UD - Set when emulating an intercepted #UD from hardware.
1634  *		      Indicates that only select instructions (tagged with
1635  *		      EmulateOnUD) should be emulated (to minimize the emulator
1636  *		      attack surface).  See also EMULTYPE_TRAP_UD_FORCED.
1637  *
1638  * EMULTYPE_SKIP - Set when emulating solely to skip an instruction, i.e. to
1639  *		   decode the instruction length.  For use *only* by
1640  *		   kvm_x86_ops.skip_emulated_instruction() implementations.
1641  *
1642  * EMULTYPE_ALLOW_RETRY_PF - Set when the emulator should resume the guest to
1643  *			     retry native execution under certain conditions,
1644  *			     Can only be set in conjunction with EMULTYPE_PF.
1645  *
1646  * EMULTYPE_TRAP_UD_FORCED - Set when emulating an intercepted #UD that was
1647  *			     triggered by KVM's magic "force emulation" prefix,
1648  *			     which is opt in via module param (off by default).
1649  *			     Bypasses EmulateOnUD restriction despite emulating
1650  *			     due to an intercepted #UD (see EMULTYPE_TRAP_UD).
1651  *			     Used to test the full emulator from userspace.
1652  *
1653  * EMULTYPE_VMWARE_GP - Set when emulating an intercepted #GP for VMware
1654  *			backdoor emulation, which is opt in via module param.
1655  *			VMware backdoor emulation handles select instructions
1656  *			and reinjects the #GP for all other cases.
1657  *
1658  * EMULTYPE_PF - Set when emulating MMIO by way of an intercepted #PF, in which
1659  *		 case the CR2/GPA value pass on the stack is valid.
1660  */
1661 #define EMULTYPE_NO_DECODE	    (1 << 0)
1662 #define EMULTYPE_TRAP_UD	    (1 << 1)
1663 #define EMULTYPE_SKIP		    (1 << 2)
1664 #define EMULTYPE_ALLOW_RETRY_PF	    (1 << 3)
1665 #define EMULTYPE_TRAP_UD_FORCED	    (1 << 4)
1666 #define EMULTYPE_VMWARE_GP	    (1 << 5)
1667 #define EMULTYPE_PF		    (1 << 6)
1668 
1669 int kvm_emulate_instruction(struct kvm_vcpu *vcpu, int emulation_type);
1670 int kvm_emulate_instruction_from_buffer(struct kvm_vcpu *vcpu,
1671 					void *insn, int insn_len);
1672 void __kvm_prepare_emulation_failure_exit(struct kvm_vcpu *vcpu,
1673 					  u64 *data, u8 ndata);
1674 void kvm_prepare_emulation_failure_exit(struct kvm_vcpu *vcpu);
1675 
1676 void kvm_enable_efer_bits(u64);
1677 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
1678 int __kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data, bool host_initiated);
1679 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data);
1680 int kvm_set_msr(struct kvm_vcpu *vcpu, u32 index, u64 data);
1681 int kvm_emulate_rdmsr(struct kvm_vcpu *vcpu);
1682 int kvm_emulate_wrmsr(struct kvm_vcpu *vcpu);
1683 int kvm_emulate_as_nop(struct kvm_vcpu *vcpu);
1684 int kvm_emulate_invd(struct kvm_vcpu *vcpu);
1685 int kvm_emulate_mwait(struct kvm_vcpu *vcpu);
1686 int kvm_handle_invalid_op(struct kvm_vcpu *vcpu);
1687 int kvm_emulate_monitor(struct kvm_vcpu *vcpu);
1688 
1689 int kvm_fast_pio(struct kvm_vcpu *vcpu, int size, unsigned short port, int in);
1690 int kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
1691 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
1692 int kvm_vcpu_halt(struct kvm_vcpu *vcpu);
1693 int kvm_emulate_ap_reset_hold(struct kvm_vcpu *vcpu);
1694 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
1695 
1696 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
1697 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
1698 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector);
1699 
1700 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
1701 		    int reason, bool has_error_code, u32 error_code);
1702 
1703 void kvm_post_set_cr0(struct kvm_vcpu *vcpu, unsigned long old_cr0, unsigned long cr0);
1704 void kvm_post_set_cr4(struct kvm_vcpu *vcpu, unsigned long old_cr4, unsigned long cr4);
1705 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
1706 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
1707 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
1708 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
1709 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
1710 void kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
1711 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
1712 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
1713 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
1714 int kvm_emulate_xsetbv(struct kvm_vcpu *vcpu);
1715 
1716 int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1717 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1718 
1719 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
1720 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
1721 int kvm_emulate_rdpmc(struct kvm_vcpu *vcpu);
1722 
1723 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1724 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1725 void kvm_queue_exception_p(struct kvm_vcpu *vcpu, unsigned nr, unsigned long payload);
1726 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1727 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1728 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
1729 bool kvm_inject_emulated_page_fault(struct kvm_vcpu *vcpu,
1730 				    struct x86_exception *fault);
1731 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
1732 bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr);
1733 
1734 static inline int __kvm_irq_line_state(unsigned long *irq_state,
1735 				       int irq_source_id, int level)
1736 {
1737 	/* Logical OR for level trig interrupt */
1738 	if (level)
1739 		__set_bit(irq_source_id, irq_state);
1740 	else
1741 		__clear_bit(irq_source_id, irq_state);
1742 
1743 	return !!(*irq_state);
1744 }
1745 
1746 #define KVM_MMU_ROOT_CURRENT		BIT(0)
1747 #define KVM_MMU_ROOT_PREVIOUS(i)	BIT(1+i)
1748 #define KVM_MMU_ROOTS_ALL		(~0UL)
1749 
1750 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
1751 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
1752 
1753 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
1754 
1755 void kvm_update_dr7(struct kvm_vcpu *vcpu);
1756 
1757 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
1758 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
1759 void kvm_mmu_free_roots(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1760 			ulong roots_to_free);
1761 void kvm_mmu_free_guest_mode_roots(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu);
1762 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1763 			   struct x86_exception *exception);
1764 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
1765 			      struct x86_exception *exception);
1766 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
1767 			       struct x86_exception *exception);
1768 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
1769 			       struct x86_exception *exception);
1770 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
1771 				struct x86_exception *exception);
1772 
1773 bool kvm_apicv_activated(struct kvm *kvm);
1774 void kvm_vcpu_update_apicv(struct kvm_vcpu *vcpu);
1775 void kvm_request_apicv_update(struct kvm *kvm, bool activate,
1776 			      unsigned long bit);
1777 
1778 void __kvm_request_apicv_update(struct kvm *kvm, bool activate,
1779 				unsigned long bit);
1780 
1781 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
1782 
1783 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u64 error_code,
1784 		       void *insn, int insn_len);
1785 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
1786 void kvm_mmu_invalidate_gva(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1787 			    gva_t gva, hpa_t root_hpa);
1788 void kvm_mmu_invpcid_gva(struct kvm_vcpu *vcpu, gva_t gva, unsigned long pcid);
1789 void kvm_mmu_new_pgd(struct kvm_vcpu *vcpu, gpa_t new_pgd);
1790 
1791 void kvm_configure_mmu(bool enable_tdp, int tdp_forced_root_level,
1792 		       int tdp_max_root_level, int tdp_huge_page_level);
1793 
1794 static inline u16 kvm_read_ldt(void)
1795 {
1796 	u16 ldt;
1797 	asm("sldt %0" : "=g"(ldt));
1798 	return ldt;
1799 }
1800 
1801 static inline void kvm_load_ldt(u16 sel)
1802 {
1803 	asm("lldt %0" : : "rm"(sel));
1804 }
1805 
1806 #ifdef CONFIG_X86_64
1807 static inline unsigned long read_msr(unsigned long msr)
1808 {
1809 	u64 value;
1810 
1811 	rdmsrl(msr, value);
1812 	return value;
1813 }
1814 #endif
1815 
1816 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
1817 {
1818 	kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
1819 }
1820 
1821 #define TSS_IOPB_BASE_OFFSET 0x66
1822 #define TSS_BASE_SIZE 0x68
1823 #define TSS_IOPB_SIZE (65536 / 8)
1824 #define TSS_REDIRECTION_SIZE (256 / 8)
1825 #define RMODE_TSS_SIZE							\
1826 	(TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
1827 
1828 enum {
1829 	TASK_SWITCH_CALL = 0,
1830 	TASK_SWITCH_IRET = 1,
1831 	TASK_SWITCH_JMP = 2,
1832 	TASK_SWITCH_GATE = 3,
1833 };
1834 
1835 #define HF_GIF_MASK		(1 << 0)
1836 #define HF_NMI_MASK		(1 << 3)
1837 #define HF_IRET_MASK		(1 << 4)
1838 #define HF_GUEST_MASK		(1 << 5) /* VCPU is in guest-mode */
1839 #define HF_SMM_MASK		(1 << 6)
1840 #define HF_SMM_INSIDE_NMI_MASK	(1 << 7)
1841 
1842 #define __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
1843 #define KVM_ADDRESS_SPACE_NUM 2
1844 
1845 #define kvm_arch_vcpu_memslots_id(vcpu) ((vcpu)->arch.hflags & HF_SMM_MASK ? 1 : 0)
1846 #define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, (role).smm)
1847 
1848 #define KVM_ARCH_WANT_MMU_NOTIFIER
1849 
1850 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1851 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1852 int kvm_cpu_has_extint(struct kvm_vcpu *v);
1853 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1854 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1855 void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event);
1856 void kvm_vcpu_reload_apic_access_page(struct kvm_vcpu *vcpu);
1857 
1858 int kvm_pv_send_ipi(struct kvm *kvm, unsigned long ipi_bitmap_low,
1859 		    unsigned long ipi_bitmap_high, u32 min,
1860 		    unsigned long icr, int op_64_bit);
1861 
1862 int kvm_add_user_return_msr(u32 msr);
1863 int kvm_find_user_return_msr(u32 msr);
1864 int kvm_set_user_return_msr(unsigned index, u64 val, u64 mask);
1865 
1866 static inline bool kvm_is_supported_user_return_msr(u32 msr)
1867 {
1868 	return kvm_find_user_return_msr(msr) >= 0;
1869 }
1870 
1871 u64 kvm_scale_tsc(struct kvm_vcpu *vcpu, u64 tsc, u64 ratio);
1872 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc);
1873 u64 kvm_calc_nested_tsc_offset(u64 l1_offset, u64 l2_offset, u64 l2_multiplier);
1874 u64 kvm_calc_nested_tsc_multiplier(u64 l1_multiplier, u64 l2_multiplier);
1875 
1876 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu);
1877 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1878 
1879 void kvm_make_scan_ioapic_request(struct kvm *kvm);
1880 void kvm_make_scan_ioapic_request_mask(struct kvm *kvm,
1881 				       unsigned long *vcpu_bitmap);
1882 
1883 bool kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1884 				     struct kvm_async_pf *work);
1885 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1886 				 struct kvm_async_pf *work);
1887 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1888 			       struct kvm_async_pf *work);
1889 void kvm_arch_async_page_present_queued(struct kvm_vcpu *vcpu);
1890 bool kvm_arch_can_dequeue_async_page_present(struct kvm_vcpu *vcpu);
1891 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1892 
1893 int kvm_skip_emulated_instruction(struct kvm_vcpu *vcpu);
1894 int kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1895 void __kvm_request_immediate_exit(struct kvm_vcpu *vcpu);
1896 
1897 int kvm_is_in_guest(void);
1898 
1899 void __user *__x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa,
1900 				     u32 size);
1901 bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu);
1902 bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu);
1903 
1904 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
1905 			     struct kvm_vcpu **dest_vcpu);
1906 
1907 void kvm_set_msi_irq(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e,
1908 		     struct kvm_lapic_irq *irq);
1909 
1910 static inline bool kvm_irq_is_postable(struct kvm_lapic_irq *irq)
1911 {
1912 	/* We can only post Fixed and LowPrio IRQs */
1913 	return (irq->delivery_mode == APIC_DM_FIXED ||
1914 		irq->delivery_mode == APIC_DM_LOWEST);
1915 }
1916 
1917 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu)
1918 {
1919 	static_call_cond(kvm_x86_vcpu_blocking)(vcpu);
1920 }
1921 
1922 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu)
1923 {
1924 	static_call_cond(kvm_x86_vcpu_unblocking)(vcpu);
1925 }
1926 
1927 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
1928 
1929 static inline int kvm_cpu_get_apicid(int mps_cpu)
1930 {
1931 #ifdef CONFIG_X86_LOCAL_APIC
1932 	return default_cpu_present_to_apicid(mps_cpu);
1933 #else
1934 	WARN_ON_ONCE(1);
1935 	return BAD_APICID;
1936 #endif
1937 }
1938 
1939 #define put_smstate(type, buf, offset, val)                      \
1940 	*(type *)((buf) + (offset) - 0x7e00) = val
1941 
1942 #define GET_SMSTATE(type, buf, offset)		\
1943 	(*(type *)((buf) + (offset) - 0x7e00))
1944 
1945 int kvm_cpu_dirty_log_size(void);
1946 
1947 int memslot_rmap_alloc(struct kvm_memory_slot *slot, unsigned long npages);
1948 
1949 #define KVM_CLOCK_VALID_FLAGS						\
1950 	(KVM_CLOCK_TSC_STABLE | KVM_CLOCK_REALTIME | KVM_CLOCK_HOST_TSC)
1951 
1952 #endif /* _ASM_X86_KVM_HOST_H */
1953