xref: /openbmc/linux/arch/x86/include/asm/kvm_host.h (revision bcda5fd3)
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 288
41 #define KVM_SOFT_MAX_VCPUS 240
42 #define KVM_MAX_VCPU_ID 1023
43 /* memory slots that are not exposed to userspace */
44 #define KVM_PRIVATE_MEM_SLOTS 3
45 
46 #define KVM_HALT_POLL_NS_DEFAULT 200000
47 
48 #define KVM_IRQCHIP_NUM_PINS  KVM_IOAPIC_NUM_PINS
49 
50 #define KVM_DIRTY_LOG_MANUAL_CAPS   (KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE | \
51 					KVM_DIRTY_LOG_INITIALLY_SET)
52 
53 #define KVM_BUS_LOCK_DETECTION_VALID_MODE	(KVM_BUS_LOCK_DETECTION_OFF | \
54 						 KVM_BUS_LOCK_DETECTION_EXIT)
55 
56 /* x86-specific vcpu->requests bit members */
57 #define KVM_REQ_MIGRATE_TIMER		KVM_ARCH_REQ(0)
58 #define KVM_REQ_REPORT_TPR_ACCESS	KVM_ARCH_REQ(1)
59 #define KVM_REQ_TRIPLE_FAULT		KVM_ARCH_REQ(2)
60 #define KVM_REQ_MMU_SYNC		KVM_ARCH_REQ(3)
61 #define KVM_REQ_CLOCK_UPDATE		KVM_ARCH_REQ(4)
62 #define KVM_REQ_LOAD_MMU_PGD		KVM_ARCH_REQ(5)
63 #define KVM_REQ_EVENT			KVM_ARCH_REQ(6)
64 #define KVM_REQ_APF_HALT		KVM_ARCH_REQ(7)
65 #define KVM_REQ_STEAL_UPDATE		KVM_ARCH_REQ(8)
66 #define KVM_REQ_NMI			KVM_ARCH_REQ(9)
67 #define KVM_REQ_PMU			KVM_ARCH_REQ(10)
68 #define KVM_REQ_PMI			KVM_ARCH_REQ(11)
69 #define KVM_REQ_SMI			KVM_ARCH_REQ(12)
70 #define KVM_REQ_MASTERCLOCK_UPDATE	KVM_ARCH_REQ(13)
71 #define KVM_REQ_MCLOCK_INPROGRESS \
72 	KVM_ARCH_REQ_FLAGS(14, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
73 #define KVM_REQ_SCAN_IOAPIC \
74 	KVM_ARCH_REQ_FLAGS(15, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
75 #define KVM_REQ_GLOBAL_CLOCK_UPDATE	KVM_ARCH_REQ(16)
76 #define KVM_REQ_APIC_PAGE_RELOAD \
77 	KVM_ARCH_REQ_FLAGS(17, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
78 #define KVM_REQ_HV_CRASH		KVM_ARCH_REQ(18)
79 #define KVM_REQ_IOAPIC_EOI_EXIT		KVM_ARCH_REQ(19)
80 #define KVM_REQ_HV_RESET		KVM_ARCH_REQ(20)
81 #define KVM_REQ_HV_EXIT			KVM_ARCH_REQ(21)
82 #define KVM_REQ_HV_STIMER		KVM_ARCH_REQ(22)
83 #define KVM_REQ_LOAD_EOI_EXITMAP	KVM_ARCH_REQ(23)
84 #define KVM_REQ_GET_NESTED_STATE_PAGES	KVM_ARCH_REQ(24)
85 #define KVM_REQ_APICV_UPDATE \
86 	KVM_ARCH_REQ_FLAGS(25, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
87 #define KVM_REQ_TLB_FLUSH_CURRENT	KVM_ARCH_REQ(26)
88 #define KVM_REQ_TLB_FLUSH_GUEST \
89 	KVM_ARCH_REQ_FLAGS(27, KVM_REQUEST_NO_WAKEUP)
90 #define KVM_REQ_APF_READY		KVM_ARCH_REQ(28)
91 #define KVM_REQ_MSR_FILTER_CHANGED	KVM_ARCH_REQ(29)
92 #define KVM_REQ_UPDATE_CPU_DIRTY_LOGGING \
93 	KVM_ARCH_REQ_FLAGS(30, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
94 
95 #define CR0_RESERVED_BITS                                               \
96 	(~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \
97 			  | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \
98 			  | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG))
99 
100 #define CR4_RESERVED_BITS                                               \
101 	(~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\
102 			  | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE     \
103 			  | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR | X86_CR4_PCIDE \
104 			  | X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_FSGSBASE \
105 			  | X86_CR4_OSXMMEXCPT | X86_CR4_LA57 | X86_CR4_VMXE \
106 			  | X86_CR4_SMAP | X86_CR4_PKE | X86_CR4_UMIP))
107 
108 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR)
109 
110 
111 
112 #define INVALID_PAGE (~(hpa_t)0)
113 #define VALID_PAGE(x) ((x) != INVALID_PAGE)
114 
115 #define UNMAPPED_GVA (~(gpa_t)0)
116 #define INVALID_GPA (~(gpa_t)0)
117 
118 /* KVM Hugepage definitions for x86 */
119 #define KVM_MAX_HUGEPAGE_LEVEL	PG_LEVEL_1G
120 #define KVM_NR_PAGE_SIZES	(KVM_MAX_HUGEPAGE_LEVEL - PG_LEVEL_4K + 1)
121 #define KVM_HPAGE_GFN_SHIFT(x)	(((x) - 1) * 9)
122 #define KVM_HPAGE_SHIFT(x)	(PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
123 #define KVM_HPAGE_SIZE(x)	(1UL << KVM_HPAGE_SHIFT(x))
124 #define KVM_HPAGE_MASK(x)	(~(KVM_HPAGE_SIZE(x) - 1))
125 #define KVM_PAGES_PER_HPAGE(x)	(KVM_HPAGE_SIZE(x) / PAGE_SIZE)
126 
127 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
128 {
129 	/* KVM_HPAGE_GFN_SHIFT(PG_LEVEL_4K) must be 0. */
130 	return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
131 		(base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
132 }
133 
134 #define KVM_PERMILLE_MMU_PAGES 20
135 #define KVM_MIN_ALLOC_MMU_PAGES 64UL
136 #define KVM_MMU_HASH_SHIFT 12
137 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
138 #define KVM_MIN_FREE_MMU_PAGES 5
139 #define KVM_REFILL_PAGES 25
140 #define KVM_MAX_CPUID_ENTRIES 256
141 #define KVM_NR_FIXED_MTRR_REGION 88
142 #define KVM_NR_VAR_MTRR 8
143 
144 #define ASYNC_PF_PER_VCPU 64
145 
146 enum kvm_reg {
147 	VCPU_REGS_RAX = __VCPU_REGS_RAX,
148 	VCPU_REGS_RCX = __VCPU_REGS_RCX,
149 	VCPU_REGS_RDX = __VCPU_REGS_RDX,
150 	VCPU_REGS_RBX = __VCPU_REGS_RBX,
151 	VCPU_REGS_RSP = __VCPU_REGS_RSP,
152 	VCPU_REGS_RBP = __VCPU_REGS_RBP,
153 	VCPU_REGS_RSI = __VCPU_REGS_RSI,
154 	VCPU_REGS_RDI = __VCPU_REGS_RDI,
155 #ifdef CONFIG_X86_64
156 	VCPU_REGS_R8  = __VCPU_REGS_R8,
157 	VCPU_REGS_R9  = __VCPU_REGS_R9,
158 	VCPU_REGS_R10 = __VCPU_REGS_R10,
159 	VCPU_REGS_R11 = __VCPU_REGS_R11,
160 	VCPU_REGS_R12 = __VCPU_REGS_R12,
161 	VCPU_REGS_R13 = __VCPU_REGS_R13,
162 	VCPU_REGS_R14 = __VCPU_REGS_R14,
163 	VCPU_REGS_R15 = __VCPU_REGS_R15,
164 #endif
165 	VCPU_REGS_RIP,
166 	NR_VCPU_REGS,
167 
168 	VCPU_EXREG_PDPTR = NR_VCPU_REGS,
169 	VCPU_EXREG_CR0,
170 	VCPU_EXREG_CR3,
171 	VCPU_EXREG_CR4,
172 	VCPU_EXREG_RFLAGS,
173 	VCPU_EXREG_SEGMENTS,
174 	VCPU_EXREG_EXIT_INFO_1,
175 	VCPU_EXREG_EXIT_INFO_2,
176 };
177 
178 enum {
179 	VCPU_SREG_ES,
180 	VCPU_SREG_CS,
181 	VCPU_SREG_SS,
182 	VCPU_SREG_DS,
183 	VCPU_SREG_FS,
184 	VCPU_SREG_GS,
185 	VCPU_SREG_TR,
186 	VCPU_SREG_LDTR,
187 };
188 
189 enum exit_fastpath_completion {
190 	EXIT_FASTPATH_NONE,
191 	EXIT_FASTPATH_REENTER_GUEST,
192 	EXIT_FASTPATH_EXIT_HANDLED,
193 };
194 typedef enum exit_fastpath_completion fastpath_t;
195 
196 struct x86_emulate_ctxt;
197 struct x86_exception;
198 enum x86_intercept;
199 enum x86_intercept_stage;
200 
201 #define KVM_NR_DB_REGS	4
202 
203 #define DR6_BUS_LOCK   (1 << 11)
204 #define DR6_BD		(1 << 13)
205 #define DR6_BS		(1 << 14)
206 #define DR6_BT		(1 << 15)
207 #define DR6_RTM		(1 << 16)
208 /*
209  * DR6_ACTIVE_LOW combines fixed-1 and active-low bits.
210  * We can regard all the bits in DR6_FIXED_1 as active_low bits;
211  * they will never be 0 for now, but when they are defined
212  * in the future it will require no code change.
213  *
214  * DR6_ACTIVE_LOW is also used as the init/reset value for DR6.
215  */
216 #define DR6_ACTIVE_LOW	0xffff0ff0
217 #define DR6_VOLATILE	0x0001e80f
218 #define DR6_FIXED_1	(DR6_ACTIVE_LOW & ~DR6_VOLATILE)
219 
220 #define DR7_BP_EN_MASK	0x000000ff
221 #define DR7_GE		(1 << 9)
222 #define DR7_GD		(1 << 13)
223 #define DR7_FIXED_1	0x00000400
224 #define DR7_VOLATILE	0xffff2bff
225 
226 #define KVM_GUESTDBG_VALID_MASK \
227 	(KVM_GUESTDBG_ENABLE | \
228 	KVM_GUESTDBG_SINGLESTEP | \
229 	KVM_GUESTDBG_USE_HW_BP | \
230 	KVM_GUESTDBG_USE_SW_BP | \
231 	KVM_GUESTDBG_INJECT_BP | \
232 	KVM_GUESTDBG_INJECT_DB)
233 
234 
235 #define PFERR_PRESENT_BIT 0
236 #define PFERR_WRITE_BIT 1
237 #define PFERR_USER_BIT 2
238 #define PFERR_RSVD_BIT 3
239 #define PFERR_FETCH_BIT 4
240 #define PFERR_PK_BIT 5
241 #define PFERR_SGX_BIT 15
242 #define PFERR_GUEST_FINAL_BIT 32
243 #define PFERR_GUEST_PAGE_BIT 33
244 
245 #define PFERR_PRESENT_MASK (1U << PFERR_PRESENT_BIT)
246 #define PFERR_WRITE_MASK (1U << PFERR_WRITE_BIT)
247 #define PFERR_USER_MASK (1U << PFERR_USER_BIT)
248 #define PFERR_RSVD_MASK (1U << PFERR_RSVD_BIT)
249 #define PFERR_FETCH_MASK (1U << PFERR_FETCH_BIT)
250 #define PFERR_PK_MASK (1U << PFERR_PK_BIT)
251 #define PFERR_SGX_MASK (1U << PFERR_SGX_BIT)
252 #define PFERR_GUEST_FINAL_MASK (1ULL << PFERR_GUEST_FINAL_BIT)
253 #define PFERR_GUEST_PAGE_MASK (1ULL << PFERR_GUEST_PAGE_BIT)
254 
255 #define PFERR_NESTED_GUEST_PAGE (PFERR_GUEST_PAGE_MASK |	\
256 				 PFERR_WRITE_MASK |		\
257 				 PFERR_PRESENT_MASK)
258 
259 /* apic attention bits */
260 #define KVM_APIC_CHECK_VAPIC	0
261 /*
262  * The following bit is set with PV-EOI, unset on EOI.
263  * We detect PV-EOI changes by guest by comparing
264  * this bit with PV-EOI in guest memory.
265  * See the implementation in apic_update_pv_eoi.
266  */
267 #define KVM_APIC_PV_EOI_PENDING	1
268 
269 struct kvm_kernel_irq_routing_entry;
270 
271 /*
272  * kvm_mmu_page_role tracks the properties of a shadow page (where shadow page
273  * also includes TDP pages) to determine whether or not a page can be used in
274  * the given MMU context.  This is a subset of the overall kvm_mmu_role to
275  * minimize the size of kvm_memory_slot.arch.gfn_track, i.e. allows allocating
276  * 2 bytes per gfn instead of 4 bytes per gfn.
277  *
278  * Indirect upper-level shadow pages are tracked for write-protection via
279  * gfn_track.  As above, gfn_track is a 16 bit counter, so KVM must not create
280  * more than 2^16-1 upper-level shadow pages at a single gfn, otherwise
281  * gfn_track will overflow and explosions will ensure.
282  *
283  * A unique shadow page (SP) for a gfn is created if and only if an existing SP
284  * cannot be reused.  The ability to reuse a SP is tracked by its role, which
285  * incorporates various mode bits and properties of the SP.  Roughly speaking,
286  * the number of unique SPs that can theoretically be created is 2^n, where n
287  * is the number of bits that are used to compute the role.
288  *
289  * But, even though there are 18 bits in the mask below, not all combinations
290  * of modes and flags are possible.  The maximum number of possible upper-level
291  * shadow pages for a single gfn is in the neighborhood of 2^13.
292  *
293  *   - invalid shadow pages are not accounted.
294  *   - level is effectively limited to four combinations, not 16 as the number
295  *     bits would imply, as 4k SPs are not tracked (allowed to go unsync).
296  *   - level is effectively unused for non-PAE paging because there is exactly
297  *     one upper level (see 4k SP exception above).
298  *   - quadrant is used only for non-PAE paging and is exclusive with
299  *     gpte_is_8_bytes.
300  *   - execonly and ad_disabled are used only for nested EPT, which makes it
301  *     exclusive with quadrant.
302  */
303 union kvm_mmu_page_role {
304 	u32 word;
305 	struct {
306 		unsigned level:4;
307 		unsigned gpte_is_8_bytes:1;
308 		unsigned quadrant:2;
309 		unsigned direct:1;
310 		unsigned access:3;
311 		unsigned invalid:1;
312 		unsigned efer_nx:1;
313 		unsigned cr0_wp:1;
314 		unsigned smep_andnot_wp:1;
315 		unsigned smap_andnot_wp:1;
316 		unsigned ad_disabled:1;
317 		unsigned guest_mode:1;
318 		unsigned :6;
319 
320 		/*
321 		 * This is left at the top of the word so that
322 		 * kvm_memslots_for_spte_role can extract it with a
323 		 * simple shift.  While there is room, give it a whole
324 		 * byte so it is also faster to load it from memory.
325 		 */
326 		unsigned smm:8;
327 	};
328 };
329 
330 /*
331  * kvm_mmu_extended_role complements kvm_mmu_page_role, tracking properties
332  * relevant to the current MMU configuration.   When loading CR0, CR4, or EFER,
333  * including on nested transitions, if nothing in the full role changes then
334  * MMU re-configuration can be skipped. @valid bit is set on first usage so we
335  * don't treat all-zero structure as valid data.
336  *
337  * The properties that are tracked in the extended role but not the page role
338  * are for things that either (a) do not affect the validity of the shadow page
339  * or (b) are indirectly reflected in the shadow page's role.  For example,
340  * CR4.PKE only affects permission checks for software walks of the guest page
341  * tables (because KVM doesn't support Protection Keys with shadow paging), and
342  * CR0.PG, CR4.PAE, and CR4.PSE are indirectly reflected in role.level.
343  *
344  * Note, SMEP and SMAP are not redundant with sm*p_andnot_wp in the page role.
345  * If CR0.WP=1, KVM can reuse shadow pages for the guest regardless of SMEP and
346  * SMAP, but the MMU's permission checks for software walks need to be SMEP and
347  * SMAP aware regardless of CR0.WP.
348  */
349 union kvm_mmu_extended_role {
350 	u32 word;
351 	struct {
352 		unsigned int valid:1;
353 		unsigned int execonly:1;
354 		unsigned int cr0_pg:1;
355 		unsigned int cr4_pae:1;
356 		unsigned int cr4_pse:1;
357 		unsigned int cr4_pke:1;
358 		unsigned int cr4_smap:1;
359 		unsigned int cr4_smep:1;
360 		unsigned int cr4_la57:1;
361 	};
362 };
363 
364 union kvm_mmu_role {
365 	u64 as_u64;
366 	struct {
367 		union kvm_mmu_page_role base;
368 		union kvm_mmu_extended_role ext;
369 	};
370 };
371 
372 struct kvm_rmap_head {
373 	unsigned long val;
374 };
375 
376 struct kvm_pio_request {
377 	unsigned long linear_rip;
378 	unsigned long count;
379 	int in;
380 	int port;
381 	int size;
382 };
383 
384 #define PT64_ROOT_MAX_LEVEL 5
385 
386 struct rsvd_bits_validate {
387 	u64 rsvd_bits_mask[2][PT64_ROOT_MAX_LEVEL];
388 	u64 bad_mt_xwr;
389 };
390 
391 struct kvm_mmu_root_info {
392 	gpa_t pgd;
393 	hpa_t hpa;
394 };
395 
396 #define KVM_MMU_ROOT_INFO_INVALID \
397 	((struct kvm_mmu_root_info) { .pgd = INVALID_PAGE, .hpa = INVALID_PAGE })
398 
399 #define KVM_MMU_NUM_PREV_ROOTS 3
400 
401 #define KVM_HAVE_MMU_RWLOCK
402 
403 struct kvm_mmu_page;
404 
405 /*
406  * x86 supports 4 paging modes (5-level 64-bit, 4-level 64-bit, 3-level 32-bit,
407  * and 2-level 32-bit).  The kvm_mmu structure abstracts the details of the
408  * current mmu mode.
409  */
410 struct kvm_mmu {
411 	unsigned long (*get_guest_pgd)(struct kvm_vcpu *vcpu);
412 	u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
413 	int (*page_fault)(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u32 err,
414 			  bool prefault);
415 	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
416 				  struct x86_exception *fault);
417 	gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gpa_t gva_or_gpa,
418 			    u32 access, struct x86_exception *exception);
419 	gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
420 			       struct x86_exception *exception);
421 	int (*sync_page)(struct kvm_vcpu *vcpu,
422 			 struct kvm_mmu_page *sp);
423 	void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva, hpa_t root_hpa);
424 	hpa_t root_hpa;
425 	gpa_t root_pgd;
426 	union kvm_mmu_role mmu_role;
427 	u8 root_level;
428 	u8 shadow_root_level;
429 	u8 ept_ad;
430 	bool direct_map;
431 	struct kvm_mmu_root_info prev_roots[KVM_MMU_NUM_PREV_ROOTS];
432 
433 	/*
434 	 * Bitmap; bit set = permission fault
435 	 * Byte index: page fault error code [4:1]
436 	 * Bit index: pte permissions in ACC_* format
437 	 */
438 	u8 permissions[16];
439 
440 	/*
441 	* The pkru_mask indicates if protection key checks are needed.  It
442 	* consists of 16 domains indexed by page fault error code bits [4:1],
443 	* with PFEC.RSVD replaced by ACC_USER_MASK from the page tables.
444 	* Each domain has 2 bits which are ANDed with AD and WD from PKRU.
445 	*/
446 	u32 pkru_mask;
447 
448 	u64 *pae_root;
449 	u64 *pml4_root;
450 
451 	/*
452 	 * check zero bits on shadow page table entries, these
453 	 * bits include not only hardware reserved bits but also
454 	 * the bits spte never used.
455 	 */
456 	struct rsvd_bits_validate shadow_zero_check;
457 
458 	struct rsvd_bits_validate guest_rsvd_check;
459 
460 	u64 pdptrs[4]; /* pae */
461 };
462 
463 struct kvm_tlb_range {
464 	u64 start_gfn;
465 	u64 pages;
466 };
467 
468 enum pmc_type {
469 	KVM_PMC_GP = 0,
470 	KVM_PMC_FIXED,
471 };
472 
473 struct kvm_pmc {
474 	enum pmc_type type;
475 	u8 idx;
476 	u64 counter;
477 	u64 eventsel;
478 	struct perf_event *perf_event;
479 	struct kvm_vcpu *vcpu;
480 	/*
481 	 * eventsel value for general purpose counters,
482 	 * ctrl value for fixed counters.
483 	 */
484 	u64 current_config;
485 };
486 
487 struct kvm_pmu {
488 	unsigned nr_arch_gp_counters;
489 	unsigned nr_arch_fixed_counters;
490 	unsigned available_event_types;
491 	u64 fixed_ctr_ctrl;
492 	u64 global_ctrl;
493 	u64 global_status;
494 	u64 global_ovf_ctrl;
495 	u64 counter_bitmask[2];
496 	u64 global_ctrl_mask;
497 	u64 global_ovf_ctrl_mask;
498 	u64 reserved_bits;
499 	u8 version;
500 	struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
501 	struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED];
502 	struct irq_work irq_work;
503 	DECLARE_BITMAP(reprogram_pmi, X86_PMC_IDX_MAX);
504 	DECLARE_BITMAP(all_valid_pmc_idx, X86_PMC_IDX_MAX);
505 	DECLARE_BITMAP(pmc_in_use, X86_PMC_IDX_MAX);
506 
507 	/*
508 	 * The gate to release perf_events not marked in
509 	 * pmc_in_use only once in a vcpu time slice.
510 	 */
511 	bool need_cleanup;
512 
513 	/*
514 	 * The total number of programmed perf_events and it helps to avoid
515 	 * redundant check before cleanup if guest don't use vPMU at all.
516 	 */
517 	u8 event_count;
518 };
519 
520 struct kvm_pmu_ops;
521 
522 enum {
523 	KVM_DEBUGREG_BP_ENABLED = 1,
524 	KVM_DEBUGREG_WONT_EXIT = 2,
525 	KVM_DEBUGREG_RELOAD = 4,
526 };
527 
528 struct kvm_mtrr_range {
529 	u64 base;
530 	u64 mask;
531 	struct list_head node;
532 };
533 
534 struct kvm_mtrr {
535 	struct kvm_mtrr_range var_ranges[KVM_NR_VAR_MTRR];
536 	mtrr_type fixed_ranges[KVM_NR_FIXED_MTRR_REGION];
537 	u64 deftype;
538 
539 	struct list_head head;
540 };
541 
542 /* Hyper-V SynIC timer */
543 struct kvm_vcpu_hv_stimer {
544 	struct hrtimer timer;
545 	int index;
546 	union hv_stimer_config config;
547 	u64 count;
548 	u64 exp_time;
549 	struct hv_message msg;
550 	bool msg_pending;
551 };
552 
553 /* Hyper-V synthetic interrupt controller (SynIC)*/
554 struct kvm_vcpu_hv_synic {
555 	u64 version;
556 	u64 control;
557 	u64 msg_page;
558 	u64 evt_page;
559 	atomic64_t sint[HV_SYNIC_SINT_COUNT];
560 	atomic_t sint_to_gsi[HV_SYNIC_SINT_COUNT];
561 	DECLARE_BITMAP(auto_eoi_bitmap, 256);
562 	DECLARE_BITMAP(vec_bitmap, 256);
563 	bool active;
564 	bool dont_zero_synic_pages;
565 };
566 
567 /* Hyper-V per vcpu emulation context */
568 struct kvm_vcpu_hv {
569 	struct kvm_vcpu *vcpu;
570 	u32 vp_index;
571 	u64 hv_vapic;
572 	s64 runtime_offset;
573 	struct kvm_vcpu_hv_synic synic;
574 	struct kvm_hyperv_exit exit;
575 	struct kvm_vcpu_hv_stimer stimer[HV_SYNIC_STIMER_COUNT];
576 	DECLARE_BITMAP(stimer_pending_bitmap, HV_SYNIC_STIMER_COUNT);
577 	cpumask_t tlb_flush;
578 	bool enforce_cpuid;
579 	struct {
580 		u32 features_eax; /* HYPERV_CPUID_FEATURES.EAX */
581 		u32 features_ebx; /* HYPERV_CPUID_FEATURES.EBX */
582 		u32 features_edx; /* HYPERV_CPUID_FEATURES.EDX */
583 		u32 enlightenments_eax; /* HYPERV_CPUID_ENLIGHTMENT_INFO.EAX */
584 		u32 enlightenments_ebx; /* HYPERV_CPUID_ENLIGHTMENT_INFO.EBX */
585 		u32 syndbg_cap_eax; /* HYPERV_CPUID_SYNDBG_PLATFORM_CAPABILITIES.EAX */
586 	} cpuid_cache;
587 };
588 
589 /* Xen HVM per vcpu emulation context */
590 struct kvm_vcpu_xen {
591 	u64 hypercall_rip;
592 	u32 current_runstate;
593 	bool vcpu_info_set;
594 	bool vcpu_time_info_set;
595 	bool runstate_set;
596 	struct gfn_to_hva_cache vcpu_info_cache;
597 	struct gfn_to_hva_cache vcpu_time_info_cache;
598 	struct gfn_to_hva_cache runstate_cache;
599 	u64 last_steal;
600 	u64 runstate_entry_time;
601 	u64 runstate_times[4];
602 };
603 
604 struct kvm_vcpu_arch {
605 	/*
606 	 * rip and regs accesses must go through
607 	 * kvm_{register,rip}_{read,write} functions.
608 	 */
609 	unsigned long regs[NR_VCPU_REGS];
610 	u32 regs_avail;
611 	u32 regs_dirty;
612 
613 	unsigned long cr0;
614 	unsigned long cr0_guest_owned_bits;
615 	unsigned long cr2;
616 	unsigned long cr3;
617 	unsigned long cr4;
618 	unsigned long cr4_guest_owned_bits;
619 	unsigned long cr4_guest_rsvd_bits;
620 	unsigned long cr8;
621 	u32 host_pkru;
622 	u32 pkru;
623 	u32 hflags;
624 	u64 efer;
625 	u64 apic_base;
626 	struct kvm_lapic *apic;    /* kernel irqchip context */
627 	bool apicv_active;
628 	bool load_eoi_exitmap_pending;
629 	DECLARE_BITMAP(ioapic_handled_vectors, 256);
630 	unsigned long apic_attention;
631 	int32_t apic_arb_prio;
632 	int mp_state;
633 	u64 ia32_misc_enable_msr;
634 	u64 smbase;
635 	u64 smi_count;
636 	bool tpr_access_reporting;
637 	bool xsaves_enabled;
638 	u64 ia32_xss;
639 	u64 microcode_version;
640 	u64 arch_capabilities;
641 	u64 perf_capabilities;
642 
643 	/*
644 	 * Paging state of the vcpu
645 	 *
646 	 * If the vcpu runs in guest mode with two level paging this still saves
647 	 * the paging mode of the l1 guest. This context is always used to
648 	 * handle faults.
649 	 */
650 	struct kvm_mmu *mmu;
651 
652 	/* Non-nested MMU for L1 */
653 	struct kvm_mmu root_mmu;
654 
655 	/* L1 MMU when running nested */
656 	struct kvm_mmu guest_mmu;
657 
658 	/*
659 	 * Paging state of an L2 guest (used for nested npt)
660 	 *
661 	 * This context will save all necessary information to walk page tables
662 	 * of an L2 guest. This context is only initialized for page table
663 	 * walking and not for faulting since we never handle l2 page faults on
664 	 * the host.
665 	 */
666 	struct kvm_mmu nested_mmu;
667 
668 	/*
669 	 * Pointer to the mmu context currently used for
670 	 * gva_to_gpa translations.
671 	 */
672 	struct kvm_mmu *walk_mmu;
673 
674 	struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
675 	struct kvm_mmu_memory_cache mmu_shadow_page_cache;
676 	struct kvm_mmu_memory_cache mmu_gfn_array_cache;
677 	struct kvm_mmu_memory_cache mmu_page_header_cache;
678 
679 	/*
680 	 * QEMU userspace and the guest each have their own FPU state.
681 	 * In vcpu_run, we switch between the user and guest FPU contexts.
682 	 * While running a VCPU, the VCPU thread will have the guest FPU
683 	 * context.
684 	 *
685 	 * Note that while the PKRU state lives inside the fpu registers,
686 	 * it is switched out separately at VMENTER and VMEXIT time. The
687 	 * "guest_fpu" state here contains the guest FPU context, with the
688 	 * host PRKU bits.
689 	 */
690 	struct fpu *user_fpu;
691 	struct fpu *guest_fpu;
692 
693 	u64 xcr0;
694 	u64 guest_supported_xcr0;
695 
696 	struct kvm_pio_request pio;
697 	void *pio_data;
698 	void *guest_ins_data;
699 
700 	u8 event_exit_inst_len;
701 
702 	struct kvm_queued_exception {
703 		bool pending;
704 		bool injected;
705 		bool has_error_code;
706 		u8 nr;
707 		u32 error_code;
708 		unsigned long payload;
709 		bool has_payload;
710 		u8 nested_apf;
711 	} exception;
712 
713 	struct kvm_queued_interrupt {
714 		bool injected;
715 		bool soft;
716 		u8 nr;
717 	} interrupt;
718 
719 	int halt_request; /* real mode on Intel only */
720 
721 	int cpuid_nent;
722 	struct kvm_cpuid_entry2 *cpuid_entries;
723 
724 	u64 reserved_gpa_bits;
725 	int maxphyaddr;
726 	int max_tdp_level;
727 
728 	/* emulate context */
729 
730 	struct x86_emulate_ctxt *emulate_ctxt;
731 	bool emulate_regs_need_sync_to_vcpu;
732 	bool emulate_regs_need_sync_from_vcpu;
733 	int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
734 
735 	gpa_t time;
736 	struct pvclock_vcpu_time_info hv_clock;
737 	unsigned int hw_tsc_khz;
738 	struct gfn_to_hva_cache pv_time;
739 	bool pv_time_enabled;
740 	/* set guest stopped flag in pvclock flags field */
741 	bool pvclock_set_guest_stopped_request;
742 
743 	struct {
744 		u8 preempted;
745 		u64 msr_val;
746 		u64 last_steal;
747 		struct gfn_to_pfn_cache cache;
748 	} st;
749 
750 	u64 l1_tsc_offset;
751 	u64 tsc_offset; /* current tsc offset */
752 	u64 last_guest_tsc;
753 	u64 last_host_tsc;
754 	u64 tsc_offset_adjustment;
755 	u64 this_tsc_nsec;
756 	u64 this_tsc_write;
757 	u64 this_tsc_generation;
758 	bool tsc_catchup;
759 	bool tsc_always_catchup;
760 	s8 virtual_tsc_shift;
761 	u32 virtual_tsc_mult;
762 	u32 virtual_tsc_khz;
763 	s64 ia32_tsc_adjust_msr;
764 	u64 msr_ia32_power_ctl;
765 	u64 l1_tsc_scaling_ratio;
766 	u64 tsc_scaling_ratio; /* current scaling ratio */
767 
768 	atomic_t nmi_queued;  /* unprocessed asynchronous NMIs */
769 	unsigned nmi_pending; /* NMI queued after currently running handler */
770 	bool nmi_injected;    /* Trying to inject an NMI this entry */
771 	bool smi_pending;    /* SMI queued after currently running handler */
772 
773 	struct kvm_mtrr mtrr_state;
774 	u64 pat;
775 
776 	unsigned switch_db_regs;
777 	unsigned long db[KVM_NR_DB_REGS];
778 	unsigned long dr6;
779 	unsigned long dr7;
780 	unsigned long eff_db[KVM_NR_DB_REGS];
781 	unsigned long guest_debug_dr7;
782 	u64 msr_platform_info;
783 	u64 msr_misc_features_enables;
784 
785 	u64 mcg_cap;
786 	u64 mcg_status;
787 	u64 mcg_ctl;
788 	u64 mcg_ext_ctl;
789 	u64 *mce_banks;
790 
791 	/* Cache MMIO info */
792 	u64 mmio_gva;
793 	unsigned mmio_access;
794 	gfn_t mmio_gfn;
795 	u64 mmio_gen;
796 
797 	struct kvm_pmu pmu;
798 
799 	/* used for guest single stepping over the given code position */
800 	unsigned long singlestep_rip;
801 
802 	bool hyperv_enabled;
803 	struct kvm_vcpu_hv *hyperv;
804 	struct kvm_vcpu_xen xen;
805 
806 	cpumask_var_t wbinvd_dirty_mask;
807 
808 	unsigned long last_retry_eip;
809 	unsigned long last_retry_addr;
810 
811 	struct {
812 		bool halted;
813 		gfn_t gfns[ASYNC_PF_PER_VCPU];
814 		struct gfn_to_hva_cache data;
815 		u64 msr_en_val; /* MSR_KVM_ASYNC_PF_EN */
816 		u64 msr_int_val; /* MSR_KVM_ASYNC_PF_INT */
817 		u16 vec;
818 		u32 id;
819 		bool send_user_only;
820 		u32 host_apf_flags;
821 		unsigned long nested_apf_token;
822 		bool delivery_as_pf_vmexit;
823 		bool pageready_pending;
824 	} apf;
825 
826 	/* OSVW MSRs (AMD only) */
827 	struct {
828 		u64 length;
829 		u64 status;
830 	} osvw;
831 
832 	struct {
833 		u64 msr_val;
834 		struct gfn_to_hva_cache data;
835 	} pv_eoi;
836 
837 	u64 msr_kvm_poll_control;
838 
839 	/*
840 	 * Indicates the guest is trying to write a gfn that contains one or
841 	 * more of the PTEs used to translate the write itself, i.e. the access
842 	 * is changing its own translation in the guest page tables.  KVM exits
843 	 * to userspace if emulation of the faulting instruction fails and this
844 	 * flag is set, as KVM cannot make forward progress.
845 	 *
846 	 * If emulation fails for a write to guest page tables, KVM unprotects
847 	 * (zaps) the shadow page for the target gfn and resumes the guest to
848 	 * retry the non-emulatable instruction (on hardware).  Unprotecting the
849 	 * gfn doesn't allow forward progress for a self-changing access because
850 	 * doing so also zaps the translation for the gfn, i.e. retrying the
851 	 * instruction will hit a !PRESENT fault, which results in a new shadow
852 	 * page and sends KVM back to square one.
853 	 */
854 	bool write_fault_to_shadow_pgtable;
855 
856 	/* set at EPT violation at this point */
857 	unsigned long exit_qualification;
858 
859 	/* pv related host specific info */
860 	struct {
861 		bool pv_unhalted;
862 	} pv;
863 
864 	int pending_ioapic_eoi;
865 	int pending_external_vector;
866 
867 	/* be preempted when it's in kernel-mode(cpl=0) */
868 	bool preempted_in_kernel;
869 
870 	/* Flush the L1 Data cache for L1TF mitigation on VMENTER */
871 	bool l1tf_flush_l1d;
872 
873 	/* Host CPU on which VM-entry was most recently attempted */
874 	int last_vmentry_cpu;
875 
876 	/* AMD MSRC001_0015 Hardware Configuration */
877 	u64 msr_hwcr;
878 
879 	/* pv related cpuid info */
880 	struct {
881 		/*
882 		 * value of the eax register in the KVM_CPUID_FEATURES CPUID
883 		 * leaf.
884 		 */
885 		u32 features;
886 
887 		/*
888 		 * indicates whether pv emulation should be disabled if features
889 		 * are not present in the guest's cpuid
890 		 */
891 		bool enforce;
892 	} pv_cpuid;
893 
894 	/* Protected Guests */
895 	bool guest_state_protected;
896 
897 	/*
898 	 * Set when PDPTS were loaded directly by the userspace without
899 	 * reading the guest memory
900 	 */
901 	bool pdptrs_from_userspace;
902 
903 #if IS_ENABLED(CONFIG_HYPERV)
904 	hpa_t hv_root_tdp;
905 #endif
906 };
907 
908 struct kvm_lpage_info {
909 	int disallow_lpage;
910 };
911 
912 struct kvm_arch_memory_slot {
913 	struct kvm_rmap_head *rmap[KVM_NR_PAGE_SIZES];
914 	struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
915 	unsigned short *gfn_track[KVM_PAGE_TRACK_MAX];
916 };
917 
918 /*
919  * We use as the mode the number of bits allocated in the LDR for the
920  * logical processor ID.  It happens that these are all powers of two.
921  * This makes it is very easy to detect cases where the APICs are
922  * configured for multiple modes; in that case, we cannot use the map and
923  * hence cannot use kvm_irq_delivery_to_apic_fast either.
924  */
925 #define KVM_APIC_MODE_XAPIC_CLUSTER          4
926 #define KVM_APIC_MODE_XAPIC_FLAT             8
927 #define KVM_APIC_MODE_X2APIC                16
928 
929 struct kvm_apic_map {
930 	struct rcu_head rcu;
931 	u8 mode;
932 	u32 max_apic_id;
933 	union {
934 		struct kvm_lapic *xapic_flat_map[8];
935 		struct kvm_lapic *xapic_cluster_map[16][4];
936 	};
937 	struct kvm_lapic *phys_map[];
938 };
939 
940 /* Hyper-V synthetic debugger (SynDbg)*/
941 struct kvm_hv_syndbg {
942 	struct {
943 		u64 control;
944 		u64 status;
945 		u64 send_page;
946 		u64 recv_page;
947 		u64 pending_page;
948 	} control;
949 	u64 options;
950 };
951 
952 /* Current state of Hyper-V TSC page clocksource */
953 enum hv_tsc_page_status {
954 	/* TSC page was not set up or disabled */
955 	HV_TSC_PAGE_UNSET = 0,
956 	/* TSC page MSR was written by the guest, update pending */
957 	HV_TSC_PAGE_GUEST_CHANGED,
958 	/* TSC page MSR was written by KVM userspace, update pending */
959 	HV_TSC_PAGE_HOST_CHANGED,
960 	/* TSC page was properly set up and is currently active  */
961 	HV_TSC_PAGE_SET,
962 	/* TSC page is currently being updated and therefore is inactive */
963 	HV_TSC_PAGE_UPDATING,
964 	/* TSC page was set up with an inaccessible GPA */
965 	HV_TSC_PAGE_BROKEN,
966 };
967 
968 /* Hyper-V emulation context */
969 struct kvm_hv {
970 	struct mutex hv_lock;
971 	u64 hv_guest_os_id;
972 	u64 hv_hypercall;
973 	u64 hv_tsc_page;
974 	enum hv_tsc_page_status hv_tsc_page_status;
975 
976 	/* Hyper-v based guest crash (NT kernel bugcheck) parameters */
977 	u64 hv_crash_param[HV_X64_MSR_CRASH_PARAMS];
978 	u64 hv_crash_ctl;
979 
980 	struct ms_hyperv_tsc_page tsc_ref;
981 
982 	struct idr conn_to_evt;
983 
984 	u64 hv_reenlightenment_control;
985 	u64 hv_tsc_emulation_control;
986 	u64 hv_tsc_emulation_status;
987 
988 	/* How many vCPUs have VP index != vCPU index */
989 	atomic_t num_mismatched_vp_indexes;
990 
991 	struct hv_partition_assist_pg *hv_pa_pg;
992 	struct kvm_hv_syndbg hv_syndbg;
993 };
994 
995 struct msr_bitmap_range {
996 	u32 flags;
997 	u32 nmsrs;
998 	u32 base;
999 	unsigned long *bitmap;
1000 };
1001 
1002 /* Xen emulation context */
1003 struct kvm_xen {
1004 	bool long_mode;
1005 	bool shinfo_set;
1006 	u8 upcall_vector;
1007 	struct gfn_to_hva_cache shinfo_cache;
1008 };
1009 
1010 enum kvm_irqchip_mode {
1011 	KVM_IRQCHIP_NONE,
1012 	KVM_IRQCHIP_KERNEL,       /* created with KVM_CREATE_IRQCHIP */
1013 	KVM_IRQCHIP_SPLIT,        /* created with KVM_CAP_SPLIT_IRQCHIP */
1014 };
1015 
1016 struct kvm_x86_msr_filter {
1017 	u8 count;
1018 	bool default_allow:1;
1019 	struct msr_bitmap_range ranges[16];
1020 };
1021 
1022 #define APICV_INHIBIT_REASON_DISABLE    0
1023 #define APICV_INHIBIT_REASON_HYPERV     1
1024 #define APICV_INHIBIT_REASON_NESTED     2
1025 #define APICV_INHIBIT_REASON_IRQWIN     3
1026 #define APICV_INHIBIT_REASON_PIT_REINJ  4
1027 #define APICV_INHIBIT_REASON_X2APIC	5
1028 
1029 struct kvm_arch {
1030 	unsigned long n_used_mmu_pages;
1031 	unsigned long n_requested_mmu_pages;
1032 	unsigned long n_max_mmu_pages;
1033 	unsigned int indirect_shadow_pages;
1034 	u8 mmu_valid_gen;
1035 	struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
1036 	struct list_head active_mmu_pages;
1037 	struct list_head zapped_obsolete_pages;
1038 	struct list_head lpage_disallowed_mmu_pages;
1039 	struct kvm_page_track_notifier_node mmu_sp_tracker;
1040 	struct kvm_page_track_notifier_head track_notifier_head;
1041 	/*
1042 	 * Protects marking pages unsync during page faults, as TDP MMU page
1043 	 * faults only take mmu_lock for read.  For simplicity, the unsync
1044 	 * pages lock is always taken when marking pages unsync regardless of
1045 	 * whether mmu_lock is held for read or write.
1046 	 */
1047 	spinlock_t mmu_unsync_pages_lock;
1048 
1049 	struct list_head assigned_dev_head;
1050 	struct iommu_domain *iommu_domain;
1051 	bool iommu_noncoherent;
1052 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA
1053 	atomic_t noncoherent_dma_count;
1054 #define __KVM_HAVE_ARCH_ASSIGNED_DEVICE
1055 	atomic_t assigned_device_count;
1056 	struct kvm_pic *vpic;
1057 	struct kvm_ioapic *vioapic;
1058 	struct kvm_pit *vpit;
1059 	atomic_t vapics_in_nmi_mode;
1060 	struct mutex apic_map_lock;
1061 	struct kvm_apic_map __rcu *apic_map;
1062 	atomic_t apic_map_dirty;
1063 
1064 	bool apic_access_memslot_enabled;
1065 	unsigned long apicv_inhibit_reasons;
1066 
1067 	gpa_t wall_clock;
1068 
1069 	bool mwait_in_guest;
1070 	bool hlt_in_guest;
1071 	bool pause_in_guest;
1072 	bool cstate_in_guest;
1073 
1074 	unsigned long irq_sources_bitmap;
1075 	s64 kvmclock_offset;
1076 	raw_spinlock_t tsc_write_lock;
1077 	u64 last_tsc_nsec;
1078 	u64 last_tsc_write;
1079 	u32 last_tsc_khz;
1080 	u64 cur_tsc_nsec;
1081 	u64 cur_tsc_write;
1082 	u64 cur_tsc_offset;
1083 	u64 cur_tsc_generation;
1084 	int nr_vcpus_matched_tsc;
1085 
1086 	spinlock_t pvclock_gtod_sync_lock;
1087 	bool use_master_clock;
1088 	u64 master_kernel_ns;
1089 	u64 master_cycle_now;
1090 	struct delayed_work kvmclock_update_work;
1091 	struct delayed_work kvmclock_sync_work;
1092 
1093 	struct kvm_xen_hvm_config xen_hvm_config;
1094 
1095 	/* reads protected by irq_srcu, writes by irq_lock */
1096 	struct hlist_head mask_notifier_list;
1097 
1098 	struct kvm_hv hyperv;
1099 	struct kvm_xen xen;
1100 
1101 	#ifdef CONFIG_KVM_MMU_AUDIT
1102 	int audit_point;
1103 	#endif
1104 
1105 	bool backwards_tsc_observed;
1106 	bool boot_vcpu_runs_old_kvmclock;
1107 	u32 bsp_vcpu_id;
1108 
1109 	u64 disabled_quirks;
1110 	int cpu_dirty_logging_count;
1111 
1112 	enum kvm_irqchip_mode irqchip_mode;
1113 	u8 nr_reserved_ioapic_pins;
1114 
1115 	bool disabled_lapic_found;
1116 
1117 	bool x2apic_format;
1118 	bool x2apic_broadcast_quirk_disabled;
1119 
1120 	bool guest_can_read_msr_platform_info;
1121 	bool exception_payload_enabled;
1122 
1123 	bool bus_lock_detection_enabled;
1124 	/*
1125 	 * If exit_on_emulation_error is set, and the in-kernel instruction
1126 	 * emulator fails to emulate an instruction, allow userspace
1127 	 * the opportunity to look at it.
1128 	 */
1129 	bool exit_on_emulation_error;
1130 
1131 	/* Deflect RDMSR and WRMSR to user space when they trigger a #GP */
1132 	u32 user_space_msr_mask;
1133 	struct kvm_x86_msr_filter __rcu *msr_filter;
1134 
1135 	u32 hypercall_exit_enabled;
1136 
1137 	/* Guest can access the SGX PROVISIONKEY. */
1138 	bool sgx_provisioning_allowed;
1139 
1140 	struct kvm_pmu_event_filter __rcu *pmu_event_filter;
1141 	struct task_struct *nx_lpage_recovery_thread;
1142 
1143 #ifdef CONFIG_X86_64
1144 	/*
1145 	 * Whether the TDP MMU is enabled for this VM. This contains a
1146 	 * snapshot of the TDP MMU module parameter from when the VM was
1147 	 * created and remains unchanged for the life of the VM. If this is
1148 	 * true, TDP MMU handler functions will run for various MMU
1149 	 * operations.
1150 	 */
1151 	bool tdp_mmu_enabled;
1152 
1153 	/*
1154 	 * List of struct kvm_mmu_pages being used as roots.
1155 	 * All struct kvm_mmu_pages in the list should have
1156 	 * tdp_mmu_page set.
1157 	 *
1158 	 * For reads, this list is protected by:
1159 	 *	the MMU lock in read mode + RCU or
1160 	 *	the MMU lock in write mode
1161 	 *
1162 	 * For writes, this list is protected by:
1163 	 *	the MMU lock in read mode + the tdp_mmu_pages_lock or
1164 	 *	the MMU lock in write mode
1165 	 *
1166 	 * Roots will remain in the list until their tdp_mmu_root_count
1167 	 * drops to zero, at which point the thread that decremented the
1168 	 * count to zero should removed the root from the list and clean
1169 	 * it up, freeing the root after an RCU grace period.
1170 	 */
1171 	struct list_head tdp_mmu_roots;
1172 
1173 	/*
1174 	 * List of struct kvmp_mmu_pages not being used as roots.
1175 	 * All struct kvm_mmu_pages in the list should have
1176 	 * tdp_mmu_page set and a tdp_mmu_root_count of 0.
1177 	 */
1178 	struct list_head tdp_mmu_pages;
1179 
1180 	/*
1181 	 * Protects accesses to the following fields when the MMU lock
1182 	 * is held in read mode:
1183 	 *  - tdp_mmu_roots (above)
1184 	 *  - tdp_mmu_pages (above)
1185 	 *  - the link field of struct kvm_mmu_pages used by the TDP MMU
1186 	 *  - lpage_disallowed_mmu_pages
1187 	 *  - the lpage_disallowed_link field of struct kvm_mmu_pages used
1188 	 *    by the TDP MMU
1189 	 * It is acceptable, but not necessary, to acquire this lock when
1190 	 * the thread holds the MMU lock in write mode.
1191 	 */
1192 	spinlock_t tdp_mmu_pages_lock;
1193 #endif /* CONFIG_X86_64 */
1194 
1195 	/*
1196 	 * If set, rmaps have been allocated for all memslots and should be
1197 	 * allocated for any newly created or modified memslots.
1198 	 */
1199 	bool memslots_have_rmaps;
1200 
1201 #if IS_ENABLED(CONFIG_HYPERV)
1202 	hpa_t	hv_root_tdp;
1203 	spinlock_t hv_root_tdp_lock;
1204 #endif
1205 };
1206 
1207 struct kvm_vm_stat {
1208 	struct kvm_vm_stat_generic generic;
1209 	u64 mmu_shadow_zapped;
1210 	u64 mmu_pte_write;
1211 	u64 mmu_pde_zapped;
1212 	u64 mmu_flooded;
1213 	u64 mmu_recycled;
1214 	u64 mmu_cache_miss;
1215 	u64 mmu_unsync;
1216 	u64 lpages;
1217 	u64 nx_lpage_splits;
1218 	u64 max_mmu_page_hash_collisions;
1219 };
1220 
1221 struct kvm_vcpu_stat {
1222 	struct kvm_vcpu_stat_generic generic;
1223 	u64 pf_fixed;
1224 	u64 pf_guest;
1225 	u64 tlb_flush;
1226 	u64 invlpg;
1227 
1228 	u64 exits;
1229 	u64 io_exits;
1230 	u64 mmio_exits;
1231 	u64 signal_exits;
1232 	u64 irq_window_exits;
1233 	u64 nmi_window_exits;
1234 	u64 l1d_flush;
1235 	u64 halt_exits;
1236 	u64 request_irq_exits;
1237 	u64 irq_exits;
1238 	u64 host_state_reload;
1239 	u64 fpu_reload;
1240 	u64 insn_emulation;
1241 	u64 insn_emulation_fail;
1242 	u64 hypercalls;
1243 	u64 irq_injections;
1244 	u64 nmi_injections;
1245 	u64 req_event;
1246 	u64 nested_run;
1247 	u64 directed_yield_attempted;
1248 	u64 directed_yield_successful;
1249 	u64 guest_mode;
1250 };
1251 
1252 struct x86_instruction_info;
1253 
1254 struct msr_data {
1255 	bool host_initiated;
1256 	u32 index;
1257 	u64 data;
1258 };
1259 
1260 struct kvm_lapic_irq {
1261 	u32 vector;
1262 	u16 delivery_mode;
1263 	u16 dest_mode;
1264 	bool level;
1265 	u16 trig_mode;
1266 	u32 shorthand;
1267 	u32 dest_id;
1268 	bool msi_redir_hint;
1269 };
1270 
1271 static inline u16 kvm_lapic_irq_dest_mode(bool dest_mode_logical)
1272 {
1273 	return dest_mode_logical ? APIC_DEST_LOGICAL : APIC_DEST_PHYSICAL;
1274 }
1275 
1276 struct kvm_x86_ops {
1277 	int (*hardware_enable)(void);
1278 	void (*hardware_disable)(void);
1279 	void (*hardware_unsetup)(void);
1280 	bool (*cpu_has_accelerated_tpr)(void);
1281 	bool (*has_emulated_msr)(struct kvm *kvm, u32 index);
1282 	void (*vcpu_after_set_cpuid)(struct kvm_vcpu *vcpu);
1283 
1284 	unsigned int vm_size;
1285 	int (*vm_init)(struct kvm *kvm);
1286 	void (*vm_destroy)(struct kvm *kvm);
1287 
1288 	/* Create, but do not attach this VCPU */
1289 	int (*vcpu_create)(struct kvm_vcpu *vcpu);
1290 	void (*vcpu_free)(struct kvm_vcpu *vcpu);
1291 	void (*vcpu_reset)(struct kvm_vcpu *vcpu, bool init_event);
1292 
1293 	void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
1294 	void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
1295 	void (*vcpu_put)(struct kvm_vcpu *vcpu);
1296 
1297 	void (*update_exception_bitmap)(struct kvm_vcpu *vcpu);
1298 	int (*get_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1299 	int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1300 	u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
1301 	void (*get_segment)(struct kvm_vcpu *vcpu,
1302 			    struct kvm_segment *var, int seg);
1303 	int (*get_cpl)(struct kvm_vcpu *vcpu);
1304 	void (*set_segment)(struct kvm_vcpu *vcpu,
1305 			    struct kvm_segment *var, int seg);
1306 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
1307 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
1308 	bool (*is_valid_cr4)(struct kvm_vcpu *vcpu, unsigned long cr0);
1309 	void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
1310 	int (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
1311 	void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1312 	void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1313 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1314 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1315 	void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu);
1316 	void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
1317 	void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
1318 	unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
1319 	void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
1320 
1321 	void (*tlb_flush_all)(struct kvm_vcpu *vcpu);
1322 	void (*tlb_flush_current)(struct kvm_vcpu *vcpu);
1323 	int  (*tlb_remote_flush)(struct kvm *kvm);
1324 	int  (*tlb_remote_flush_with_range)(struct kvm *kvm,
1325 			struct kvm_tlb_range *range);
1326 
1327 	/*
1328 	 * Flush any TLB entries associated with the given GVA.
1329 	 * Does not need to flush GPA->HPA mappings.
1330 	 * Can potentially get non-canonical addresses through INVLPGs, which
1331 	 * the implementation may choose to ignore if appropriate.
1332 	 */
1333 	void (*tlb_flush_gva)(struct kvm_vcpu *vcpu, gva_t addr);
1334 
1335 	/*
1336 	 * Flush any TLB entries created by the guest.  Like tlb_flush_gva(),
1337 	 * does not need to flush GPA->HPA mappings.
1338 	 */
1339 	void (*tlb_flush_guest)(struct kvm_vcpu *vcpu);
1340 
1341 	enum exit_fastpath_completion (*run)(struct kvm_vcpu *vcpu);
1342 	int (*handle_exit)(struct kvm_vcpu *vcpu,
1343 		enum exit_fastpath_completion exit_fastpath);
1344 	int (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
1345 	void (*update_emulated_instruction)(struct kvm_vcpu *vcpu);
1346 	void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
1347 	u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu);
1348 	void (*patch_hypercall)(struct kvm_vcpu *vcpu,
1349 				unsigned char *hypercall_addr);
1350 	void (*set_irq)(struct kvm_vcpu *vcpu);
1351 	void (*set_nmi)(struct kvm_vcpu *vcpu);
1352 	void (*queue_exception)(struct kvm_vcpu *vcpu);
1353 	void (*cancel_injection)(struct kvm_vcpu *vcpu);
1354 	int (*interrupt_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1355 	int (*nmi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1356 	bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
1357 	void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
1358 	void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
1359 	void (*enable_irq_window)(struct kvm_vcpu *vcpu);
1360 	void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
1361 	bool (*check_apicv_inhibit_reasons)(ulong bit);
1362 	void (*pre_update_apicv_exec_ctrl)(struct kvm *kvm, bool activate);
1363 	void (*refresh_apicv_exec_ctrl)(struct kvm_vcpu *vcpu);
1364 	void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
1365 	void (*hwapic_isr_update)(struct kvm_vcpu *vcpu, int isr);
1366 	bool (*guest_apic_has_interrupt)(struct kvm_vcpu *vcpu);
1367 	void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
1368 	void (*set_virtual_apic_mode)(struct kvm_vcpu *vcpu);
1369 	void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu);
1370 	int (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
1371 	int (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
1372 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
1373 	int (*set_identity_map_addr)(struct kvm *kvm, u64 ident_addr);
1374 	u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
1375 
1376 	void (*load_mmu_pgd)(struct kvm_vcpu *vcpu, hpa_t root_hpa,
1377 			     int root_level);
1378 
1379 	bool (*has_wbinvd_exit)(void);
1380 
1381 	u64 (*get_l2_tsc_offset)(struct kvm_vcpu *vcpu);
1382 	u64 (*get_l2_tsc_multiplier)(struct kvm_vcpu *vcpu);
1383 	void (*write_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
1384 	void (*write_tsc_multiplier)(struct kvm_vcpu *vcpu, u64 multiplier);
1385 
1386 	/*
1387 	 * Retrieve somewhat arbitrary exit information.  Intended to be used
1388 	 * only from within tracepoints to avoid VMREADs when tracing is off.
1389 	 */
1390 	void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2,
1391 			      u32 *exit_int_info, u32 *exit_int_info_err_code);
1392 
1393 	int (*check_intercept)(struct kvm_vcpu *vcpu,
1394 			       struct x86_instruction_info *info,
1395 			       enum x86_intercept_stage stage,
1396 			       struct x86_exception *exception);
1397 	void (*handle_exit_irqoff)(struct kvm_vcpu *vcpu);
1398 
1399 	void (*request_immediate_exit)(struct kvm_vcpu *vcpu);
1400 
1401 	void (*sched_in)(struct kvm_vcpu *kvm, int cpu);
1402 
1403 	/*
1404 	 * Size of the CPU's dirty log buffer, i.e. VMX's PML buffer.  A zero
1405 	 * value indicates CPU dirty logging is unsupported or disabled.
1406 	 */
1407 	int cpu_dirty_log_size;
1408 	void (*update_cpu_dirty_logging)(struct kvm_vcpu *vcpu);
1409 
1410 	/* pmu operations of sub-arch */
1411 	const struct kvm_pmu_ops *pmu_ops;
1412 	const struct kvm_x86_nested_ops *nested_ops;
1413 
1414 	/*
1415 	 * Architecture specific hooks for vCPU blocking due to
1416 	 * HLT instruction.
1417 	 * Returns for .pre_block():
1418 	 *    - 0 means continue to block the vCPU.
1419 	 *    - 1 means we cannot block the vCPU since some event
1420 	 *        happens during this period, such as, 'ON' bit in
1421 	 *        posted-interrupts descriptor is set.
1422 	 */
1423 	int (*pre_block)(struct kvm_vcpu *vcpu);
1424 	void (*post_block)(struct kvm_vcpu *vcpu);
1425 
1426 	void (*vcpu_blocking)(struct kvm_vcpu *vcpu);
1427 	void (*vcpu_unblocking)(struct kvm_vcpu *vcpu);
1428 
1429 	int (*update_pi_irte)(struct kvm *kvm, unsigned int host_irq,
1430 			      uint32_t guest_irq, bool set);
1431 	void (*start_assignment)(struct kvm *kvm);
1432 	void (*apicv_post_state_restore)(struct kvm_vcpu *vcpu);
1433 	bool (*dy_apicv_has_pending_interrupt)(struct kvm_vcpu *vcpu);
1434 
1435 	int (*set_hv_timer)(struct kvm_vcpu *vcpu, u64 guest_deadline_tsc,
1436 			    bool *expired);
1437 	void (*cancel_hv_timer)(struct kvm_vcpu *vcpu);
1438 
1439 	void (*setup_mce)(struct kvm_vcpu *vcpu);
1440 
1441 	int (*smi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1442 	int (*enter_smm)(struct kvm_vcpu *vcpu, char *smstate);
1443 	int (*leave_smm)(struct kvm_vcpu *vcpu, const char *smstate);
1444 	void (*enable_smi_window)(struct kvm_vcpu *vcpu);
1445 
1446 	int (*mem_enc_op)(struct kvm *kvm, void __user *argp);
1447 	int (*mem_enc_reg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1448 	int (*mem_enc_unreg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1449 	int (*vm_copy_enc_context_from)(struct kvm *kvm, unsigned int source_fd);
1450 
1451 	int (*get_msr_feature)(struct kvm_msr_entry *entry);
1452 
1453 	bool (*can_emulate_instruction)(struct kvm_vcpu *vcpu, void *insn, int insn_len);
1454 
1455 	bool (*apic_init_signal_blocked)(struct kvm_vcpu *vcpu);
1456 	int (*enable_direct_tlbflush)(struct kvm_vcpu *vcpu);
1457 
1458 	void (*migrate_timers)(struct kvm_vcpu *vcpu);
1459 	void (*msr_filter_changed)(struct kvm_vcpu *vcpu);
1460 	int (*complete_emulated_msr)(struct kvm_vcpu *vcpu, int err);
1461 
1462 	void (*vcpu_deliver_sipi_vector)(struct kvm_vcpu *vcpu, u8 vector);
1463 };
1464 
1465 struct kvm_x86_nested_ops {
1466 	int (*check_events)(struct kvm_vcpu *vcpu);
1467 	bool (*hv_timer_pending)(struct kvm_vcpu *vcpu);
1468 	void (*triple_fault)(struct kvm_vcpu *vcpu);
1469 	int (*get_state)(struct kvm_vcpu *vcpu,
1470 			 struct kvm_nested_state __user *user_kvm_nested_state,
1471 			 unsigned user_data_size);
1472 	int (*set_state)(struct kvm_vcpu *vcpu,
1473 			 struct kvm_nested_state __user *user_kvm_nested_state,
1474 			 struct kvm_nested_state *kvm_state);
1475 	bool (*get_nested_state_pages)(struct kvm_vcpu *vcpu);
1476 	int (*write_log_dirty)(struct kvm_vcpu *vcpu, gpa_t l2_gpa);
1477 
1478 	int (*enable_evmcs)(struct kvm_vcpu *vcpu,
1479 			    uint16_t *vmcs_version);
1480 	uint16_t (*get_evmcs_version)(struct kvm_vcpu *vcpu);
1481 };
1482 
1483 struct kvm_x86_init_ops {
1484 	int (*cpu_has_kvm_support)(void);
1485 	int (*disabled_by_bios)(void);
1486 	int (*check_processor_compatibility)(void);
1487 	int (*hardware_setup)(void);
1488 
1489 	struct kvm_x86_ops *runtime_ops;
1490 };
1491 
1492 struct kvm_arch_async_pf {
1493 	u32 token;
1494 	gfn_t gfn;
1495 	unsigned long cr3;
1496 	bool direct_map;
1497 };
1498 
1499 extern u32 __read_mostly kvm_nr_uret_msrs;
1500 extern u64 __read_mostly host_efer;
1501 extern bool __read_mostly allow_smaller_maxphyaddr;
1502 extern bool __read_mostly enable_apicv;
1503 extern struct kvm_x86_ops kvm_x86_ops;
1504 
1505 #define KVM_X86_OP(func) \
1506 	DECLARE_STATIC_CALL(kvm_x86_##func, *(((struct kvm_x86_ops *)0)->func));
1507 #define KVM_X86_OP_NULL KVM_X86_OP
1508 #include <asm/kvm-x86-ops.h>
1509 
1510 static inline void kvm_ops_static_call_update(void)
1511 {
1512 #define KVM_X86_OP(func) \
1513 	static_call_update(kvm_x86_##func, kvm_x86_ops.func);
1514 #define KVM_X86_OP_NULL KVM_X86_OP
1515 #include <asm/kvm-x86-ops.h>
1516 }
1517 
1518 #define __KVM_HAVE_ARCH_VM_ALLOC
1519 static inline struct kvm *kvm_arch_alloc_vm(void)
1520 {
1521 	return __vmalloc(kvm_x86_ops.vm_size, GFP_KERNEL_ACCOUNT | __GFP_ZERO);
1522 }
1523 void kvm_arch_free_vm(struct kvm *kvm);
1524 
1525 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
1526 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
1527 {
1528 	if (kvm_x86_ops.tlb_remote_flush &&
1529 	    !static_call(kvm_x86_tlb_remote_flush)(kvm))
1530 		return 0;
1531 	else
1532 		return -ENOTSUPP;
1533 }
1534 
1535 int kvm_mmu_module_init(void);
1536 void kvm_mmu_module_exit(void);
1537 
1538 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
1539 int kvm_mmu_create(struct kvm_vcpu *vcpu);
1540 void kvm_mmu_init_vm(struct kvm *kvm);
1541 void kvm_mmu_uninit_vm(struct kvm *kvm);
1542 
1543 void kvm_mmu_after_set_cpuid(struct kvm_vcpu *vcpu);
1544 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
1545 void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
1546 				      struct kvm_memory_slot *memslot,
1547 				      int start_level);
1548 void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
1549 				   const struct kvm_memory_slot *memslot);
1550 void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
1551 				   struct kvm_memory_slot *memslot);
1552 void kvm_mmu_zap_all(struct kvm *kvm);
1553 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, u64 gen);
1554 unsigned long kvm_mmu_calculate_default_mmu_pages(struct kvm *kvm);
1555 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned long kvm_nr_mmu_pages);
1556 
1557 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
1558 
1559 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
1560 			  const void *val, int bytes);
1561 
1562 struct kvm_irq_mask_notifier {
1563 	void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
1564 	int irq;
1565 	struct hlist_node link;
1566 };
1567 
1568 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
1569 				    struct kvm_irq_mask_notifier *kimn);
1570 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
1571 				      struct kvm_irq_mask_notifier *kimn);
1572 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
1573 			     bool mask);
1574 
1575 extern bool tdp_enabled;
1576 
1577 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
1578 
1579 /* control of guest tsc rate supported? */
1580 extern bool kvm_has_tsc_control;
1581 /* maximum supported tsc_khz for guests */
1582 extern u32  kvm_max_guest_tsc_khz;
1583 /* number of bits of the fractional part of the TSC scaling ratio */
1584 extern u8   kvm_tsc_scaling_ratio_frac_bits;
1585 /* maximum allowed value of TSC scaling ratio */
1586 extern u64  kvm_max_tsc_scaling_ratio;
1587 /* 1ull << kvm_tsc_scaling_ratio_frac_bits */
1588 extern u64  kvm_default_tsc_scaling_ratio;
1589 /* bus lock detection supported? */
1590 extern bool kvm_has_bus_lock_exit;
1591 
1592 extern u64 kvm_mce_cap_supported;
1593 
1594 /*
1595  * EMULTYPE_NO_DECODE - Set when re-emulating an instruction (after completing
1596  *			userspace I/O) to indicate that the emulation context
1597  *			should be reused as is, i.e. skip initialization of
1598  *			emulation context, instruction fetch and decode.
1599  *
1600  * EMULTYPE_TRAP_UD - Set when emulating an intercepted #UD from hardware.
1601  *		      Indicates that only select instructions (tagged with
1602  *		      EmulateOnUD) should be emulated (to minimize the emulator
1603  *		      attack surface).  See also EMULTYPE_TRAP_UD_FORCED.
1604  *
1605  * EMULTYPE_SKIP - Set when emulating solely to skip an instruction, i.e. to
1606  *		   decode the instruction length.  For use *only* by
1607  *		   kvm_x86_ops.skip_emulated_instruction() implementations.
1608  *
1609  * EMULTYPE_ALLOW_RETRY_PF - Set when the emulator should resume the guest to
1610  *			     retry native execution under certain conditions,
1611  *			     Can only be set in conjunction with EMULTYPE_PF.
1612  *
1613  * EMULTYPE_TRAP_UD_FORCED - Set when emulating an intercepted #UD that was
1614  *			     triggered by KVM's magic "force emulation" prefix,
1615  *			     which is opt in via module param (off by default).
1616  *			     Bypasses EmulateOnUD restriction despite emulating
1617  *			     due to an intercepted #UD (see EMULTYPE_TRAP_UD).
1618  *			     Used to test the full emulator from userspace.
1619  *
1620  * EMULTYPE_VMWARE_GP - Set when emulating an intercepted #GP for VMware
1621  *			backdoor emulation, which is opt in via module param.
1622  *			VMware backdoor emulation handles select instructions
1623  *			and reinjects the #GP for all other cases.
1624  *
1625  * EMULTYPE_PF - Set when emulating MMIO by way of an intercepted #PF, in which
1626  *		 case the CR2/GPA value pass on the stack is valid.
1627  */
1628 #define EMULTYPE_NO_DECODE	    (1 << 0)
1629 #define EMULTYPE_TRAP_UD	    (1 << 1)
1630 #define EMULTYPE_SKIP		    (1 << 2)
1631 #define EMULTYPE_ALLOW_RETRY_PF	    (1 << 3)
1632 #define EMULTYPE_TRAP_UD_FORCED	    (1 << 4)
1633 #define EMULTYPE_VMWARE_GP	    (1 << 5)
1634 #define EMULTYPE_PF		    (1 << 6)
1635 
1636 int kvm_emulate_instruction(struct kvm_vcpu *vcpu, int emulation_type);
1637 int kvm_emulate_instruction_from_buffer(struct kvm_vcpu *vcpu,
1638 					void *insn, int insn_len);
1639 
1640 void kvm_enable_efer_bits(u64);
1641 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
1642 int __kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data, bool host_initiated);
1643 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data);
1644 int kvm_set_msr(struct kvm_vcpu *vcpu, u32 index, u64 data);
1645 int kvm_emulate_rdmsr(struct kvm_vcpu *vcpu);
1646 int kvm_emulate_wrmsr(struct kvm_vcpu *vcpu);
1647 int kvm_emulate_as_nop(struct kvm_vcpu *vcpu);
1648 int kvm_emulate_invd(struct kvm_vcpu *vcpu);
1649 int kvm_emulate_mwait(struct kvm_vcpu *vcpu);
1650 int kvm_handle_invalid_op(struct kvm_vcpu *vcpu);
1651 int kvm_emulate_monitor(struct kvm_vcpu *vcpu);
1652 
1653 int kvm_fast_pio(struct kvm_vcpu *vcpu, int size, unsigned short port, int in);
1654 int kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
1655 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
1656 int kvm_vcpu_halt(struct kvm_vcpu *vcpu);
1657 int kvm_emulate_ap_reset_hold(struct kvm_vcpu *vcpu);
1658 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
1659 
1660 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
1661 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
1662 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector);
1663 
1664 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
1665 		    int reason, bool has_error_code, u32 error_code);
1666 
1667 void kvm_free_guest_fpu(struct kvm_vcpu *vcpu);
1668 
1669 void kvm_post_set_cr0(struct kvm_vcpu *vcpu, unsigned long old_cr0, unsigned long cr0);
1670 void kvm_post_set_cr4(struct kvm_vcpu *vcpu, unsigned long old_cr4, unsigned long cr4);
1671 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
1672 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
1673 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
1674 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
1675 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
1676 void kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
1677 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
1678 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
1679 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
1680 int kvm_emulate_xsetbv(struct kvm_vcpu *vcpu);
1681 
1682 int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1683 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1684 
1685 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
1686 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
1687 int kvm_emulate_rdpmc(struct kvm_vcpu *vcpu);
1688 
1689 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1690 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1691 void kvm_queue_exception_p(struct kvm_vcpu *vcpu, unsigned nr, unsigned long payload);
1692 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1693 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1694 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
1695 bool kvm_inject_emulated_page_fault(struct kvm_vcpu *vcpu,
1696 				    struct x86_exception *fault);
1697 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1698 			    gfn_t gfn, void *data, int offset, int len,
1699 			    u32 access);
1700 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
1701 bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr);
1702 
1703 static inline int __kvm_irq_line_state(unsigned long *irq_state,
1704 				       int irq_source_id, int level)
1705 {
1706 	/* Logical OR for level trig interrupt */
1707 	if (level)
1708 		__set_bit(irq_source_id, irq_state);
1709 	else
1710 		__clear_bit(irq_source_id, irq_state);
1711 
1712 	return !!(*irq_state);
1713 }
1714 
1715 #define KVM_MMU_ROOT_CURRENT		BIT(0)
1716 #define KVM_MMU_ROOT_PREVIOUS(i)	BIT(1+i)
1717 #define KVM_MMU_ROOTS_ALL		(~0UL)
1718 
1719 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
1720 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
1721 
1722 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
1723 
1724 void kvm_update_dr7(struct kvm_vcpu *vcpu);
1725 
1726 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
1727 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
1728 void kvm_mmu_free_roots(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1729 			ulong roots_to_free);
1730 void kvm_mmu_free_guest_mode_roots(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu);
1731 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1732 			   struct x86_exception *exception);
1733 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
1734 			      struct x86_exception *exception);
1735 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
1736 			       struct x86_exception *exception);
1737 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
1738 			       struct x86_exception *exception);
1739 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
1740 				struct x86_exception *exception);
1741 
1742 bool kvm_apicv_activated(struct kvm *kvm);
1743 void kvm_vcpu_update_apicv(struct kvm_vcpu *vcpu);
1744 void kvm_request_apicv_update(struct kvm *kvm, bool activate,
1745 			      unsigned long bit);
1746 
1747 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
1748 
1749 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u64 error_code,
1750 		       void *insn, int insn_len);
1751 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
1752 void kvm_mmu_invalidate_gva(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1753 			    gva_t gva, hpa_t root_hpa);
1754 void kvm_mmu_invpcid_gva(struct kvm_vcpu *vcpu, gva_t gva, unsigned long pcid);
1755 void kvm_mmu_new_pgd(struct kvm_vcpu *vcpu, gpa_t new_pgd);
1756 
1757 void kvm_configure_mmu(bool enable_tdp, int tdp_max_root_level,
1758 		       int tdp_huge_page_level);
1759 
1760 static inline u16 kvm_read_ldt(void)
1761 {
1762 	u16 ldt;
1763 	asm("sldt %0" : "=g"(ldt));
1764 	return ldt;
1765 }
1766 
1767 static inline void kvm_load_ldt(u16 sel)
1768 {
1769 	asm("lldt %0" : : "rm"(sel));
1770 }
1771 
1772 #ifdef CONFIG_X86_64
1773 static inline unsigned long read_msr(unsigned long msr)
1774 {
1775 	u64 value;
1776 
1777 	rdmsrl(msr, value);
1778 	return value;
1779 }
1780 #endif
1781 
1782 static inline u32 get_rdx_init_val(void)
1783 {
1784 	return 0x600; /* P6 family */
1785 }
1786 
1787 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
1788 {
1789 	kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
1790 }
1791 
1792 #define TSS_IOPB_BASE_OFFSET 0x66
1793 #define TSS_BASE_SIZE 0x68
1794 #define TSS_IOPB_SIZE (65536 / 8)
1795 #define TSS_REDIRECTION_SIZE (256 / 8)
1796 #define RMODE_TSS_SIZE							\
1797 	(TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
1798 
1799 enum {
1800 	TASK_SWITCH_CALL = 0,
1801 	TASK_SWITCH_IRET = 1,
1802 	TASK_SWITCH_JMP = 2,
1803 	TASK_SWITCH_GATE = 3,
1804 };
1805 
1806 #define HF_GIF_MASK		(1 << 0)
1807 #define HF_NMI_MASK		(1 << 3)
1808 #define HF_IRET_MASK		(1 << 4)
1809 #define HF_GUEST_MASK		(1 << 5) /* VCPU is in guest-mode */
1810 #define HF_SMM_MASK		(1 << 6)
1811 #define HF_SMM_INSIDE_NMI_MASK	(1 << 7)
1812 
1813 #define __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
1814 #define KVM_ADDRESS_SPACE_NUM 2
1815 
1816 #define kvm_arch_vcpu_memslots_id(vcpu) ((vcpu)->arch.hflags & HF_SMM_MASK ? 1 : 0)
1817 #define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, (role).smm)
1818 
1819 asmlinkage void kvm_spurious_fault(void);
1820 
1821 /*
1822  * Hardware virtualization extension instructions may fault if a
1823  * reboot turns off virtualization while processes are running.
1824  * Usually after catching the fault we just panic; during reboot
1825  * instead the instruction is ignored.
1826  */
1827 #define __kvm_handle_fault_on_reboot(insn)				\
1828 	"666: \n\t"							\
1829 	insn "\n\t"							\
1830 	"jmp	668f \n\t"						\
1831 	"667: \n\t"							\
1832 	"1: \n\t"							\
1833 	".pushsection .discard.instr_begin \n\t"			\
1834 	".long 1b - . \n\t"						\
1835 	".popsection \n\t"						\
1836 	"call	kvm_spurious_fault \n\t"				\
1837 	"1: \n\t"							\
1838 	".pushsection .discard.instr_end \n\t"				\
1839 	".long 1b - . \n\t"						\
1840 	".popsection \n\t"						\
1841 	"668: \n\t"							\
1842 	_ASM_EXTABLE(666b, 667b)
1843 
1844 #define KVM_ARCH_WANT_MMU_NOTIFIER
1845 
1846 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1847 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1848 int kvm_cpu_has_extint(struct kvm_vcpu *v);
1849 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1850 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1851 void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event);
1852 void kvm_vcpu_reload_apic_access_page(struct kvm_vcpu *vcpu);
1853 
1854 int kvm_pv_send_ipi(struct kvm *kvm, unsigned long ipi_bitmap_low,
1855 		    unsigned long ipi_bitmap_high, u32 min,
1856 		    unsigned long icr, int op_64_bit);
1857 
1858 int kvm_add_user_return_msr(u32 msr);
1859 int kvm_find_user_return_msr(u32 msr);
1860 int kvm_set_user_return_msr(unsigned index, u64 val, u64 mask);
1861 
1862 static inline bool kvm_is_supported_user_return_msr(u32 msr)
1863 {
1864 	return kvm_find_user_return_msr(msr) >= 0;
1865 }
1866 
1867 u64 kvm_scale_tsc(struct kvm_vcpu *vcpu, u64 tsc, u64 ratio);
1868 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc);
1869 u64 kvm_calc_nested_tsc_offset(u64 l1_offset, u64 l2_offset, u64 l2_multiplier);
1870 u64 kvm_calc_nested_tsc_multiplier(u64 l1_multiplier, u64 l2_multiplier);
1871 
1872 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu);
1873 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1874 
1875 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
1876 void kvm_make_scan_ioapic_request(struct kvm *kvm);
1877 void kvm_make_scan_ioapic_request_mask(struct kvm *kvm,
1878 				       unsigned long *vcpu_bitmap);
1879 
1880 bool kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1881 				     struct kvm_async_pf *work);
1882 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1883 				 struct kvm_async_pf *work);
1884 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1885 			       struct kvm_async_pf *work);
1886 void kvm_arch_async_page_present_queued(struct kvm_vcpu *vcpu);
1887 bool kvm_arch_can_dequeue_async_page_present(struct kvm_vcpu *vcpu);
1888 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1889 
1890 int kvm_skip_emulated_instruction(struct kvm_vcpu *vcpu);
1891 int kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1892 void __kvm_request_immediate_exit(struct kvm_vcpu *vcpu);
1893 
1894 int kvm_is_in_guest(void);
1895 
1896 void __user *__x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa,
1897 				     u32 size);
1898 bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu);
1899 bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu);
1900 
1901 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
1902 			     struct kvm_vcpu **dest_vcpu);
1903 
1904 void kvm_set_msi_irq(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e,
1905 		     struct kvm_lapic_irq *irq);
1906 
1907 static inline bool kvm_irq_is_postable(struct kvm_lapic_irq *irq)
1908 {
1909 	/* We can only post Fixed and LowPrio IRQs */
1910 	return (irq->delivery_mode == APIC_DM_FIXED ||
1911 		irq->delivery_mode == APIC_DM_LOWEST);
1912 }
1913 
1914 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu)
1915 {
1916 	static_call_cond(kvm_x86_vcpu_blocking)(vcpu);
1917 }
1918 
1919 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu)
1920 {
1921 	static_call_cond(kvm_x86_vcpu_unblocking)(vcpu);
1922 }
1923 
1924 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
1925 
1926 static inline int kvm_cpu_get_apicid(int mps_cpu)
1927 {
1928 #ifdef CONFIG_X86_LOCAL_APIC
1929 	return default_cpu_present_to_apicid(mps_cpu);
1930 #else
1931 	WARN_ON_ONCE(1);
1932 	return BAD_APICID;
1933 #endif
1934 }
1935 
1936 #define put_smstate(type, buf, offset, val)                      \
1937 	*(type *)((buf) + (offset) - 0x7e00) = val
1938 
1939 #define GET_SMSTATE(type, buf, offset)		\
1940 	(*(type *)((buf) + (offset) - 0x7e00))
1941 
1942 int kvm_cpu_dirty_log_size(void);
1943 
1944 int alloc_all_memslots_rmaps(struct kvm *kvm);
1945 
1946 #endif /* _ASM_X86_KVM_HOST_H */
1947