xref: /openbmc/linux/arch/x86/include/asm/kvm_host.h (revision 2bc7d3e0)
1 /* SPDX-License-Identifier: GPL-2.0-only */
2 /*
3  * Kernel-based Virtual Machine driver for Linux
4  *
5  * This header defines architecture specific interfaces, x86 version
6  */
7 
8 #ifndef _ASM_X86_KVM_HOST_H
9 #define _ASM_X86_KVM_HOST_H
10 
11 #include <linux/types.h>
12 #include <linux/mm.h>
13 #include <linux/mmu_notifier.h>
14 #include <linux/tracepoint.h>
15 #include <linux/cpumask.h>
16 #include <linux/irq_work.h>
17 #include <linux/irq.h>
18 #include <linux/workqueue.h>
19 
20 #include <linux/kvm.h>
21 #include <linux/kvm_para.h>
22 #include <linux/kvm_types.h>
23 #include <linux/perf_event.h>
24 #include <linux/pvclock_gtod.h>
25 #include <linux/clocksource.h>
26 #include <linux/irqbypass.h>
27 #include <linux/hyperv.h>
28 
29 #include <asm/apic.h>
30 #include <asm/pvclock-abi.h>
31 #include <asm/desc.h>
32 #include <asm/mtrr.h>
33 #include <asm/msr-index.h>
34 #include <asm/asm.h>
35 #include <asm/kvm_page_track.h>
36 #include <asm/kvm_vcpu_regs.h>
37 #include <asm/hyperv-tlfs.h>
38 
39 #define __KVM_HAVE_ARCH_VCPU_DEBUGFS
40 
41 #define KVM_MAX_VCPUS 1024
42 
43 /*
44  * In x86, the VCPU ID corresponds to the APIC ID, and APIC IDs
45  * might be larger than the actual number of VCPUs because the
46  * APIC ID encodes CPU topology information.
47  *
48  * In the worst case, we'll need less than one extra bit for the
49  * Core ID, and less than one extra bit for the Package (Die) ID,
50  * so ratio of 4 should be enough.
51  */
52 #define KVM_VCPU_ID_RATIO 4
53 #define KVM_MAX_VCPU_IDS (KVM_MAX_VCPUS * KVM_VCPU_ID_RATIO)
54 
55 /* memory slots that are not exposed to userspace */
56 #define KVM_PRIVATE_MEM_SLOTS 3
57 
58 #define KVM_HALT_POLL_NS_DEFAULT 200000
59 
60 #define KVM_IRQCHIP_NUM_PINS  KVM_IOAPIC_NUM_PINS
61 
62 #define KVM_DIRTY_LOG_MANUAL_CAPS   (KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE | \
63 					KVM_DIRTY_LOG_INITIALLY_SET)
64 
65 #define KVM_BUS_LOCK_DETECTION_VALID_MODE	(KVM_BUS_LOCK_DETECTION_OFF | \
66 						 KVM_BUS_LOCK_DETECTION_EXIT)
67 
68 /* x86-specific vcpu->requests bit members */
69 #define KVM_REQ_MIGRATE_TIMER		KVM_ARCH_REQ(0)
70 #define KVM_REQ_REPORT_TPR_ACCESS	KVM_ARCH_REQ(1)
71 #define KVM_REQ_TRIPLE_FAULT		KVM_ARCH_REQ(2)
72 #define KVM_REQ_MMU_SYNC		KVM_ARCH_REQ(3)
73 #define KVM_REQ_CLOCK_UPDATE		KVM_ARCH_REQ(4)
74 #define KVM_REQ_LOAD_MMU_PGD		KVM_ARCH_REQ(5)
75 #define KVM_REQ_EVENT			KVM_ARCH_REQ(6)
76 #define KVM_REQ_APF_HALT		KVM_ARCH_REQ(7)
77 #define KVM_REQ_STEAL_UPDATE		KVM_ARCH_REQ(8)
78 #define KVM_REQ_NMI			KVM_ARCH_REQ(9)
79 #define KVM_REQ_PMU			KVM_ARCH_REQ(10)
80 #define KVM_REQ_PMI			KVM_ARCH_REQ(11)
81 #define KVM_REQ_SMI			KVM_ARCH_REQ(12)
82 #define KVM_REQ_MASTERCLOCK_UPDATE	KVM_ARCH_REQ(13)
83 #define KVM_REQ_MCLOCK_INPROGRESS \
84 	KVM_ARCH_REQ_FLAGS(14, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
85 #define KVM_REQ_SCAN_IOAPIC \
86 	KVM_ARCH_REQ_FLAGS(15, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
87 #define KVM_REQ_GLOBAL_CLOCK_UPDATE	KVM_ARCH_REQ(16)
88 #define KVM_REQ_APIC_PAGE_RELOAD \
89 	KVM_ARCH_REQ_FLAGS(17, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
90 #define KVM_REQ_HV_CRASH		KVM_ARCH_REQ(18)
91 #define KVM_REQ_IOAPIC_EOI_EXIT		KVM_ARCH_REQ(19)
92 #define KVM_REQ_HV_RESET		KVM_ARCH_REQ(20)
93 #define KVM_REQ_HV_EXIT			KVM_ARCH_REQ(21)
94 #define KVM_REQ_HV_STIMER		KVM_ARCH_REQ(22)
95 #define KVM_REQ_LOAD_EOI_EXITMAP	KVM_ARCH_REQ(23)
96 #define KVM_REQ_GET_NESTED_STATE_PAGES	KVM_ARCH_REQ(24)
97 #define KVM_REQ_APICV_UPDATE \
98 	KVM_ARCH_REQ_FLAGS(25, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
99 #define KVM_REQ_TLB_FLUSH_CURRENT	KVM_ARCH_REQ(26)
100 #define KVM_REQ_TLB_FLUSH_GUEST \
101 	KVM_ARCH_REQ_FLAGS(27, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
102 #define KVM_REQ_APF_READY		KVM_ARCH_REQ(28)
103 #define KVM_REQ_MSR_FILTER_CHANGED	KVM_ARCH_REQ(29)
104 #define KVM_REQ_UPDATE_CPU_DIRTY_LOGGING \
105 	KVM_ARCH_REQ_FLAGS(30, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
106 #define KVM_REQ_MMU_FREE_OBSOLETE_ROOTS \
107 	KVM_ARCH_REQ_FLAGS(31, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
108 
109 #define CR0_RESERVED_BITS                                               \
110 	(~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \
111 			  | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \
112 			  | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG))
113 
114 #define CR4_RESERVED_BITS                                               \
115 	(~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\
116 			  | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE     \
117 			  | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR | X86_CR4_PCIDE \
118 			  | X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_FSGSBASE \
119 			  | X86_CR4_OSXMMEXCPT | X86_CR4_LA57 | X86_CR4_VMXE \
120 			  | X86_CR4_SMAP | X86_CR4_PKE | X86_CR4_UMIP))
121 
122 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR)
123 
124 
125 
126 #define INVALID_PAGE (~(hpa_t)0)
127 #define VALID_PAGE(x) ((x) != INVALID_PAGE)
128 
129 #define UNMAPPED_GVA (~(gpa_t)0)
130 #define INVALID_GPA (~(gpa_t)0)
131 
132 /* KVM Hugepage definitions for x86 */
133 #define KVM_MAX_HUGEPAGE_LEVEL	PG_LEVEL_1G
134 #define KVM_NR_PAGE_SIZES	(KVM_MAX_HUGEPAGE_LEVEL - PG_LEVEL_4K + 1)
135 #define KVM_HPAGE_GFN_SHIFT(x)	(((x) - 1) * 9)
136 #define KVM_HPAGE_SHIFT(x)	(PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
137 #define KVM_HPAGE_SIZE(x)	(1UL << KVM_HPAGE_SHIFT(x))
138 #define KVM_HPAGE_MASK(x)	(~(KVM_HPAGE_SIZE(x) - 1))
139 #define KVM_PAGES_PER_HPAGE(x)	(KVM_HPAGE_SIZE(x) / PAGE_SIZE)
140 
141 #define KVM_MEMSLOT_PAGES_TO_MMU_PAGES_RATIO 50
142 #define KVM_MIN_ALLOC_MMU_PAGES 64UL
143 #define KVM_MMU_HASH_SHIFT 12
144 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
145 #define KVM_MIN_FREE_MMU_PAGES 5
146 #define KVM_REFILL_PAGES 25
147 #define KVM_MAX_CPUID_ENTRIES 256
148 #define KVM_NR_FIXED_MTRR_REGION 88
149 #define KVM_NR_VAR_MTRR 8
150 
151 #define ASYNC_PF_PER_VCPU 64
152 
153 enum kvm_reg {
154 	VCPU_REGS_RAX = __VCPU_REGS_RAX,
155 	VCPU_REGS_RCX = __VCPU_REGS_RCX,
156 	VCPU_REGS_RDX = __VCPU_REGS_RDX,
157 	VCPU_REGS_RBX = __VCPU_REGS_RBX,
158 	VCPU_REGS_RSP = __VCPU_REGS_RSP,
159 	VCPU_REGS_RBP = __VCPU_REGS_RBP,
160 	VCPU_REGS_RSI = __VCPU_REGS_RSI,
161 	VCPU_REGS_RDI = __VCPU_REGS_RDI,
162 #ifdef CONFIG_X86_64
163 	VCPU_REGS_R8  = __VCPU_REGS_R8,
164 	VCPU_REGS_R9  = __VCPU_REGS_R9,
165 	VCPU_REGS_R10 = __VCPU_REGS_R10,
166 	VCPU_REGS_R11 = __VCPU_REGS_R11,
167 	VCPU_REGS_R12 = __VCPU_REGS_R12,
168 	VCPU_REGS_R13 = __VCPU_REGS_R13,
169 	VCPU_REGS_R14 = __VCPU_REGS_R14,
170 	VCPU_REGS_R15 = __VCPU_REGS_R15,
171 #endif
172 	VCPU_REGS_RIP,
173 	NR_VCPU_REGS,
174 
175 	VCPU_EXREG_PDPTR = NR_VCPU_REGS,
176 	VCPU_EXREG_CR0,
177 	VCPU_EXREG_CR3,
178 	VCPU_EXREG_CR4,
179 	VCPU_EXREG_RFLAGS,
180 	VCPU_EXREG_SEGMENTS,
181 	VCPU_EXREG_EXIT_INFO_1,
182 	VCPU_EXREG_EXIT_INFO_2,
183 };
184 
185 enum {
186 	VCPU_SREG_ES,
187 	VCPU_SREG_CS,
188 	VCPU_SREG_SS,
189 	VCPU_SREG_DS,
190 	VCPU_SREG_FS,
191 	VCPU_SREG_GS,
192 	VCPU_SREG_TR,
193 	VCPU_SREG_LDTR,
194 };
195 
196 enum exit_fastpath_completion {
197 	EXIT_FASTPATH_NONE,
198 	EXIT_FASTPATH_REENTER_GUEST,
199 	EXIT_FASTPATH_EXIT_HANDLED,
200 };
201 typedef enum exit_fastpath_completion fastpath_t;
202 
203 struct x86_emulate_ctxt;
204 struct x86_exception;
205 enum x86_intercept;
206 enum x86_intercept_stage;
207 
208 #define KVM_NR_DB_REGS	4
209 
210 #define DR6_BUS_LOCK   (1 << 11)
211 #define DR6_BD		(1 << 13)
212 #define DR6_BS		(1 << 14)
213 #define DR6_BT		(1 << 15)
214 #define DR6_RTM		(1 << 16)
215 /*
216  * DR6_ACTIVE_LOW combines fixed-1 and active-low bits.
217  * We can regard all the bits in DR6_FIXED_1 as active_low bits;
218  * they will never be 0 for now, but when they are defined
219  * in the future it will require no code change.
220  *
221  * DR6_ACTIVE_LOW is also used as the init/reset value for DR6.
222  */
223 #define DR6_ACTIVE_LOW	0xffff0ff0
224 #define DR6_VOLATILE	0x0001e80f
225 #define DR6_FIXED_1	(DR6_ACTIVE_LOW & ~DR6_VOLATILE)
226 
227 #define DR7_BP_EN_MASK	0x000000ff
228 #define DR7_GE		(1 << 9)
229 #define DR7_GD		(1 << 13)
230 #define DR7_FIXED_1	0x00000400
231 #define DR7_VOLATILE	0xffff2bff
232 
233 #define KVM_GUESTDBG_VALID_MASK \
234 	(KVM_GUESTDBG_ENABLE | \
235 	KVM_GUESTDBG_SINGLESTEP | \
236 	KVM_GUESTDBG_USE_HW_BP | \
237 	KVM_GUESTDBG_USE_SW_BP | \
238 	KVM_GUESTDBG_INJECT_BP | \
239 	KVM_GUESTDBG_INJECT_DB | \
240 	KVM_GUESTDBG_BLOCKIRQ)
241 
242 
243 #define PFERR_PRESENT_BIT 0
244 #define PFERR_WRITE_BIT 1
245 #define PFERR_USER_BIT 2
246 #define PFERR_RSVD_BIT 3
247 #define PFERR_FETCH_BIT 4
248 #define PFERR_PK_BIT 5
249 #define PFERR_SGX_BIT 15
250 #define PFERR_GUEST_FINAL_BIT 32
251 #define PFERR_GUEST_PAGE_BIT 33
252 #define PFERR_IMPLICIT_ACCESS_BIT 48
253 
254 #define PFERR_PRESENT_MASK (1U << PFERR_PRESENT_BIT)
255 #define PFERR_WRITE_MASK (1U << PFERR_WRITE_BIT)
256 #define PFERR_USER_MASK (1U << PFERR_USER_BIT)
257 #define PFERR_RSVD_MASK (1U << PFERR_RSVD_BIT)
258 #define PFERR_FETCH_MASK (1U << PFERR_FETCH_BIT)
259 #define PFERR_PK_MASK (1U << PFERR_PK_BIT)
260 #define PFERR_SGX_MASK (1U << PFERR_SGX_BIT)
261 #define PFERR_GUEST_FINAL_MASK (1ULL << PFERR_GUEST_FINAL_BIT)
262 #define PFERR_GUEST_PAGE_MASK (1ULL << PFERR_GUEST_PAGE_BIT)
263 #define PFERR_IMPLICIT_ACCESS (1ULL << PFERR_IMPLICIT_ACCESS_BIT)
264 
265 #define PFERR_NESTED_GUEST_PAGE (PFERR_GUEST_PAGE_MASK |	\
266 				 PFERR_WRITE_MASK |		\
267 				 PFERR_PRESENT_MASK)
268 
269 /* apic attention bits */
270 #define KVM_APIC_CHECK_VAPIC	0
271 /*
272  * The following bit is set with PV-EOI, unset on EOI.
273  * We detect PV-EOI changes by guest by comparing
274  * this bit with PV-EOI in guest memory.
275  * See the implementation in apic_update_pv_eoi.
276  */
277 #define KVM_APIC_PV_EOI_PENDING	1
278 
279 struct kvm_kernel_irq_routing_entry;
280 
281 /*
282  * kvm_mmu_page_role tracks the properties of a shadow page (where shadow page
283  * also includes TDP pages) to determine whether or not a page can be used in
284  * the given MMU context.  This is a subset of the overall kvm_cpu_role to
285  * minimize the size of kvm_memory_slot.arch.gfn_track, i.e. allows allocating
286  * 2 bytes per gfn instead of 4 bytes per gfn.
287  *
288  * Upper-level shadow pages having gptes are tracked for write-protection via
289  * gfn_track.  As above, gfn_track is a 16 bit counter, so KVM must not create
290  * more than 2^16-1 upper-level shadow pages at a single gfn, otherwise
291  * gfn_track will overflow and explosions will ensure.
292  *
293  * A unique shadow page (SP) for a gfn is created if and only if an existing SP
294  * cannot be reused.  The ability to reuse a SP is tracked by its role, which
295  * incorporates various mode bits and properties of the SP.  Roughly speaking,
296  * the number of unique SPs that can theoretically be created is 2^n, where n
297  * is the number of bits that are used to compute the role.
298  *
299  * But, even though there are 19 bits in the mask below, not all combinations
300  * of modes and flags are possible:
301  *
302  *   - invalid shadow pages are not accounted, so the bits are effectively 18
303  *
304  *   - quadrant will only be used if has_4_byte_gpte=1 (non-PAE paging);
305  *     execonly and ad_disabled are only used for nested EPT which has
306  *     has_4_byte_gpte=0.  Therefore, 2 bits are always unused.
307  *
308  *   - the 4 bits of level are effectively limited to the values 2/3/4/5,
309  *     as 4k SPs are not tracked (allowed to go unsync).  In addition non-PAE
310  *     paging has exactly one upper level, making level completely redundant
311  *     when has_4_byte_gpte=1.
312  *
313  *   - on top of this, smep_andnot_wp and smap_andnot_wp are only set if
314  *     cr0_wp=0, therefore these three bits only give rise to 5 possibilities.
315  *
316  * Therefore, the maximum number of possible upper-level shadow pages for a
317  * single gfn is a bit less than 2^13.
318  */
319 union kvm_mmu_page_role {
320 	u32 word;
321 	struct {
322 		unsigned level:4;
323 		unsigned has_4_byte_gpte:1;
324 		unsigned quadrant:2;
325 		unsigned direct:1;
326 		unsigned access:3;
327 		unsigned invalid:1;
328 		unsigned efer_nx:1;
329 		unsigned cr0_wp:1;
330 		unsigned smep_andnot_wp:1;
331 		unsigned smap_andnot_wp:1;
332 		unsigned ad_disabled:1;
333 		unsigned guest_mode:1;
334 		unsigned passthrough:1;
335 		unsigned :5;
336 
337 		/*
338 		 * This is left at the top of the word so that
339 		 * kvm_memslots_for_spte_role can extract it with a
340 		 * simple shift.  While there is room, give it a whole
341 		 * byte so it is also faster to load it from memory.
342 		 */
343 		unsigned smm:8;
344 	};
345 };
346 
347 /*
348  * kvm_mmu_extended_role complements kvm_mmu_page_role, tracking properties
349  * relevant to the current MMU configuration.   When loading CR0, CR4, or EFER,
350  * including on nested transitions, if nothing in the full role changes then
351  * MMU re-configuration can be skipped. @valid bit is set on first usage so we
352  * don't treat all-zero structure as valid data.
353  *
354  * The properties that are tracked in the extended role but not the page role
355  * are for things that either (a) do not affect the validity of the shadow page
356  * or (b) are indirectly reflected in the shadow page's role.  For example,
357  * CR4.PKE only affects permission checks for software walks of the guest page
358  * tables (because KVM doesn't support Protection Keys with shadow paging), and
359  * CR0.PG, CR4.PAE, and CR4.PSE are indirectly reflected in role.level.
360  *
361  * Note, SMEP and SMAP are not redundant with sm*p_andnot_wp in the page role.
362  * If CR0.WP=1, KVM can reuse shadow pages for the guest regardless of SMEP and
363  * SMAP, but the MMU's permission checks for software walks need to be SMEP and
364  * SMAP aware regardless of CR0.WP.
365  */
366 union kvm_mmu_extended_role {
367 	u32 word;
368 	struct {
369 		unsigned int valid:1;
370 		unsigned int execonly:1;
371 		unsigned int cr4_pse:1;
372 		unsigned int cr4_pke:1;
373 		unsigned int cr4_smap:1;
374 		unsigned int cr4_smep:1;
375 		unsigned int cr4_la57:1;
376 		unsigned int efer_lma:1;
377 	};
378 };
379 
380 union kvm_cpu_role {
381 	u64 as_u64;
382 	struct {
383 		union kvm_mmu_page_role base;
384 		union kvm_mmu_extended_role ext;
385 	};
386 };
387 
388 struct kvm_rmap_head {
389 	unsigned long val;
390 };
391 
392 struct kvm_pio_request {
393 	unsigned long linear_rip;
394 	unsigned long count;
395 	int in;
396 	int port;
397 	int size;
398 };
399 
400 #define PT64_ROOT_MAX_LEVEL 5
401 
402 struct rsvd_bits_validate {
403 	u64 rsvd_bits_mask[2][PT64_ROOT_MAX_LEVEL];
404 	u64 bad_mt_xwr;
405 };
406 
407 struct kvm_mmu_root_info {
408 	gpa_t pgd;
409 	hpa_t hpa;
410 };
411 
412 #define KVM_MMU_ROOT_INFO_INVALID \
413 	((struct kvm_mmu_root_info) { .pgd = INVALID_PAGE, .hpa = INVALID_PAGE })
414 
415 #define KVM_MMU_NUM_PREV_ROOTS 3
416 
417 #define KVM_HAVE_MMU_RWLOCK
418 
419 struct kvm_mmu_page;
420 struct kvm_page_fault;
421 
422 /*
423  * x86 supports 4 paging modes (5-level 64-bit, 4-level 64-bit, 3-level 32-bit,
424  * and 2-level 32-bit).  The kvm_mmu structure abstracts the details of the
425  * current mmu mode.
426  */
427 struct kvm_mmu {
428 	unsigned long (*get_guest_pgd)(struct kvm_vcpu *vcpu);
429 	u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
430 	int (*page_fault)(struct kvm_vcpu *vcpu, struct kvm_page_fault *fault);
431 	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
432 				  struct x86_exception *fault);
433 	gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
434 			    gpa_t gva_or_gpa, u64 access,
435 			    struct x86_exception *exception);
436 	int (*sync_page)(struct kvm_vcpu *vcpu,
437 			 struct kvm_mmu_page *sp);
438 	void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva, hpa_t root_hpa);
439 	struct kvm_mmu_root_info root;
440 	union kvm_cpu_role cpu_role;
441 	union kvm_mmu_page_role root_role;
442 
443 	/*
444 	* The pkru_mask indicates if protection key checks are needed.  It
445 	* consists of 16 domains indexed by page fault error code bits [4:1],
446 	* with PFEC.RSVD replaced by ACC_USER_MASK from the page tables.
447 	* Each domain has 2 bits which are ANDed with AD and WD from PKRU.
448 	*/
449 	u32 pkru_mask;
450 
451 	struct kvm_mmu_root_info prev_roots[KVM_MMU_NUM_PREV_ROOTS];
452 
453 	/*
454 	 * Bitmap; bit set = permission fault
455 	 * Byte index: page fault error code [4:1]
456 	 * Bit index: pte permissions in ACC_* format
457 	 */
458 	u8 permissions[16];
459 
460 	u64 *pae_root;
461 	u64 *pml4_root;
462 	u64 *pml5_root;
463 
464 	/*
465 	 * check zero bits on shadow page table entries, these
466 	 * bits include not only hardware reserved bits but also
467 	 * the bits spte never used.
468 	 */
469 	struct rsvd_bits_validate shadow_zero_check;
470 
471 	struct rsvd_bits_validate guest_rsvd_check;
472 
473 	u64 pdptrs[4]; /* pae */
474 };
475 
476 struct kvm_tlb_range {
477 	u64 start_gfn;
478 	u64 pages;
479 };
480 
481 enum pmc_type {
482 	KVM_PMC_GP = 0,
483 	KVM_PMC_FIXED,
484 };
485 
486 struct kvm_pmc {
487 	enum pmc_type type;
488 	u8 idx;
489 	u64 counter;
490 	u64 eventsel;
491 	struct perf_event *perf_event;
492 	struct kvm_vcpu *vcpu;
493 	/*
494 	 * eventsel value for general purpose counters,
495 	 * ctrl value for fixed counters.
496 	 */
497 	u64 current_config;
498 	bool is_paused;
499 	bool intr;
500 };
501 
502 #define KVM_PMC_MAX_FIXED	3
503 struct kvm_pmu {
504 	unsigned nr_arch_gp_counters;
505 	unsigned nr_arch_fixed_counters;
506 	unsigned available_event_types;
507 	u64 fixed_ctr_ctrl;
508 	u64 global_ctrl;
509 	u64 global_status;
510 	u64 counter_bitmask[2];
511 	u64 global_ctrl_mask;
512 	u64 global_ovf_ctrl_mask;
513 	u64 reserved_bits;
514 	u64 raw_event_mask;
515 	u8 version;
516 	struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
517 	struct kvm_pmc fixed_counters[KVM_PMC_MAX_FIXED];
518 	struct irq_work irq_work;
519 	DECLARE_BITMAP(reprogram_pmi, X86_PMC_IDX_MAX);
520 	DECLARE_BITMAP(all_valid_pmc_idx, X86_PMC_IDX_MAX);
521 	DECLARE_BITMAP(pmc_in_use, X86_PMC_IDX_MAX);
522 
523 	/*
524 	 * The gate to release perf_events not marked in
525 	 * pmc_in_use only once in a vcpu time slice.
526 	 */
527 	bool need_cleanup;
528 
529 	/*
530 	 * The total number of programmed perf_events and it helps to avoid
531 	 * redundant check before cleanup if guest don't use vPMU at all.
532 	 */
533 	u8 event_count;
534 };
535 
536 struct kvm_pmu_ops;
537 
538 enum {
539 	KVM_DEBUGREG_BP_ENABLED = 1,
540 	KVM_DEBUGREG_WONT_EXIT = 2,
541 };
542 
543 struct kvm_mtrr_range {
544 	u64 base;
545 	u64 mask;
546 	struct list_head node;
547 };
548 
549 struct kvm_mtrr {
550 	struct kvm_mtrr_range var_ranges[KVM_NR_VAR_MTRR];
551 	mtrr_type fixed_ranges[KVM_NR_FIXED_MTRR_REGION];
552 	u64 deftype;
553 
554 	struct list_head head;
555 };
556 
557 /* Hyper-V SynIC timer */
558 struct kvm_vcpu_hv_stimer {
559 	struct hrtimer timer;
560 	int index;
561 	union hv_stimer_config config;
562 	u64 count;
563 	u64 exp_time;
564 	struct hv_message msg;
565 	bool msg_pending;
566 };
567 
568 /* Hyper-V synthetic interrupt controller (SynIC)*/
569 struct kvm_vcpu_hv_synic {
570 	u64 version;
571 	u64 control;
572 	u64 msg_page;
573 	u64 evt_page;
574 	atomic64_t sint[HV_SYNIC_SINT_COUNT];
575 	atomic_t sint_to_gsi[HV_SYNIC_SINT_COUNT];
576 	DECLARE_BITMAP(auto_eoi_bitmap, 256);
577 	DECLARE_BITMAP(vec_bitmap, 256);
578 	bool active;
579 	bool dont_zero_synic_pages;
580 };
581 
582 /* Hyper-V per vcpu emulation context */
583 struct kvm_vcpu_hv {
584 	struct kvm_vcpu *vcpu;
585 	u32 vp_index;
586 	u64 hv_vapic;
587 	s64 runtime_offset;
588 	struct kvm_vcpu_hv_synic synic;
589 	struct kvm_hyperv_exit exit;
590 	struct kvm_vcpu_hv_stimer stimer[HV_SYNIC_STIMER_COUNT];
591 	DECLARE_BITMAP(stimer_pending_bitmap, HV_SYNIC_STIMER_COUNT);
592 	bool enforce_cpuid;
593 	struct {
594 		u32 features_eax; /* HYPERV_CPUID_FEATURES.EAX */
595 		u32 features_ebx; /* HYPERV_CPUID_FEATURES.EBX */
596 		u32 features_edx; /* HYPERV_CPUID_FEATURES.EDX */
597 		u32 enlightenments_eax; /* HYPERV_CPUID_ENLIGHTMENT_INFO.EAX */
598 		u32 enlightenments_ebx; /* HYPERV_CPUID_ENLIGHTMENT_INFO.EBX */
599 		u32 syndbg_cap_eax; /* HYPERV_CPUID_SYNDBG_PLATFORM_CAPABILITIES.EAX */
600 	} cpuid_cache;
601 };
602 
603 /* Xen HVM per vcpu emulation context */
604 struct kvm_vcpu_xen {
605 	u64 hypercall_rip;
606 	u32 current_runstate;
607 	u8 upcall_vector;
608 	struct gfn_to_pfn_cache vcpu_info_cache;
609 	struct gfn_to_pfn_cache vcpu_time_info_cache;
610 	struct gfn_to_pfn_cache runstate_cache;
611 	u64 last_steal;
612 	u64 runstate_entry_time;
613 	u64 runstate_times[4];
614 	unsigned long evtchn_pending_sel;
615 	u32 vcpu_id; /* The Xen / ACPI vCPU ID */
616 	u32 timer_virq;
617 	u64 timer_expires; /* In guest epoch */
618 	atomic_t timer_pending;
619 	struct hrtimer timer;
620 	int poll_evtchn;
621 	struct timer_list poll_timer;
622 };
623 
624 struct kvm_vcpu_arch {
625 	/*
626 	 * rip and regs accesses must go through
627 	 * kvm_{register,rip}_{read,write} functions.
628 	 */
629 	unsigned long regs[NR_VCPU_REGS];
630 	u32 regs_avail;
631 	u32 regs_dirty;
632 
633 	unsigned long cr0;
634 	unsigned long cr0_guest_owned_bits;
635 	unsigned long cr2;
636 	unsigned long cr3;
637 	unsigned long cr4;
638 	unsigned long cr4_guest_owned_bits;
639 	unsigned long cr4_guest_rsvd_bits;
640 	unsigned long cr8;
641 	u32 host_pkru;
642 	u32 pkru;
643 	u32 hflags;
644 	u64 efer;
645 	u64 apic_base;
646 	struct kvm_lapic *apic;    /* kernel irqchip context */
647 	bool apicv_active;
648 	bool load_eoi_exitmap_pending;
649 	DECLARE_BITMAP(ioapic_handled_vectors, 256);
650 	unsigned long apic_attention;
651 	int32_t apic_arb_prio;
652 	int mp_state;
653 	u64 ia32_misc_enable_msr;
654 	u64 smbase;
655 	u64 smi_count;
656 	bool tpr_access_reporting;
657 	bool xsaves_enabled;
658 	bool xfd_no_write_intercept;
659 	u64 ia32_xss;
660 	u64 microcode_version;
661 	u64 arch_capabilities;
662 	u64 perf_capabilities;
663 
664 	/*
665 	 * Paging state of the vcpu
666 	 *
667 	 * If the vcpu runs in guest mode with two level paging this still saves
668 	 * the paging mode of the l1 guest. This context is always used to
669 	 * handle faults.
670 	 */
671 	struct kvm_mmu *mmu;
672 
673 	/* Non-nested MMU for L1 */
674 	struct kvm_mmu root_mmu;
675 
676 	/* L1 MMU when running nested */
677 	struct kvm_mmu guest_mmu;
678 
679 	/*
680 	 * Paging state of an L2 guest (used for nested npt)
681 	 *
682 	 * This context will save all necessary information to walk page tables
683 	 * of an L2 guest. This context is only initialized for page table
684 	 * walking and not for faulting since we never handle l2 page faults on
685 	 * the host.
686 	 */
687 	struct kvm_mmu nested_mmu;
688 
689 	/*
690 	 * Pointer to the mmu context currently used for
691 	 * gva_to_gpa translations.
692 	 */
693 	struct kvm_mmu *walk_mmu;
694 
695 	struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
696 	struct kvm_mmu_memory_cache mmu_shadow_page_cache;
697 	struct kvm_mmu_memory_cache mmu_gfn_array_cache;
698 	struct kvm_mmu_memory_cache mmu_page_header_cache;
699 
700 	/*
701 	 * QEMU userspace and the guest each have their own FPU state.
702 	 * In vcpu_run, we switch between the user and guest FPU contexts.
703 	 * While running a VCPU, the VCPU thread will have the guest FPU
704 	 * context.
705 	 *
706 	 * Note that while the PKRU state lives inside the fpu registers,
707 	 * it is switched out separately at VMENTER and VMEXIT time. The
708 	 * "guest_fpstate" state here contains the guest FPU context, with the
709 	 * host PRKU bits.
710 	 */
711 	struct fpu_guest guest_fpu;
712 
713 	u64 xcr0;
714 
715 	struct kvm_pio_request pio;
716 	void *pio_data;
717 	void *sev_pio_data;
718 	unsigned sev_pio_count;
719 
720 	u8 event_exit_inst_len;
721 
722 	struct kvm_queued_exception {
723 		bool pending;
724 		bool injected;
725 		bool has_error_code;
726 		u8 nr;
727 		u32 error_code;
728 		unsigned long payload;
729 		bool has_payload;
730 		u8 nested_apf;
731 	} exception;
732 
733 	struct kvm_queued_interrupt {
734 		bool injected;
735 		bool soft;
736 		u8 nr;
737 	} interrupt;
738 
739 	int halt_request; /* real mode on Intel only */
740 
741 	int cpuid_nent;
742 	struct kvm_cpuid_entry2 *cpuid_entries;
743 	u32 kvm_cpuid_base;
744 
745 	u64 reserved_gpa_bits;
746 	int maxphyaddr;
747 
748 	/* emulate context */
749 
750 	struct x86_emulate_ctxt *emulate_ctxt;
751 	bool emulate_regs_need_sync_to_vcpu;
752 	bool emulate_regs_need_sync_from_vcpu;
753 	int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
754 
755 	gpa_t time;
756 	struct pvclock_vcpu_time_info hv_clock;
757 	unsigned int hw_tsc_khz;
758 	struct gfn_to_pfn_cache pv_time;
759 	/* set guest stopped flag in pvclock flags field */
760 	bool pvclock_set_guest_stopped_request;
761 
762 	struct {
763 		u8 preempted;
764 		u64 msr_val;
765 		u64 last_steal;
766 		struct gfn_to_hva_cache cache;
767 	} st;
768 
769 	u64 l1_tsc_offset;
770 	u64 tsc_offset; /* current tsc offset */
771 	u64 last_guest_tsc;
772 	u64 last_host_tsc;
773 	u64 tsc_offset_adjustment;
774 	u64 this_tsc_nsec;
775 	u64 this_tsc_write;
776 	u64 this_tsc_generation;
777 	bool tsc_catchup;
778 	bool tsc_always_catchup;
779 	s8 virtual_tsc_shift;
780 	u32 virtual_tsc_mult;
781 	u32 virtual_tsc_khz;
782 	s64 ia32_tsc_adjust_msr;
783 	u64 msr_ia32_power_ctl;
784 	u64 l1_tsc_scaling_ratio;
785 	u64 tsc_scaling_ratio; /* current scaling ratio */
786 
787 	atomic_t nmi_queued;  /* unprocessed asynchronous NMIs */
788 	unsigned nmi_pending; /* NMI queued after currently running handler */
789 	bool nmi_injected;    /* Trying to inject an NMI this entry */
790 	bool smi_pending;    /* SMI queued after currently running handler */
791 	u8 handling_intr_from_guest;
792 
793 	struct kvm_mtrr mtrr_state;
794 	u64 pat;
795 
796 	unsigned switch_db_regs;
797 	unsigned long db[KVM_NR_DB_REGS];
798 	unsigned long dr6;
799 	unsigned long dr7;
800 	unsigned long eff_db[KVM_NR_DB_REGS];
801 	unsigned long guest_debug_dr7;
802 	u64 msr_platform_info;
803 	u64 msr_misc_features_enables;
804 
805 	u64 mcg_cap;
806 	u64 mcg_status;
807 	u64 mcg_ctl;
808 	u64 mcg_ext_ctl;
809 	u64 *mce_banks;
810 
811 	/* Cache MMIO info */
812 	u64 mmio_gva;
813 	unsigned mmio_access;
814 	gfn_t mmio_gfn;
815 	u64 mmio_gen;
816 
817 	struct kvm_pmu pmu;
818 
819 	/* used for guest single stepping over the given code position */
820 	unsigned long singlestep_rip;
821 
822 	bool hyperv_enabled;
823 	struct kvm_vcpu_hv *hyperv;
824 	struct kvm_vcpu_xen xen;
825 
826 	cpumask_var_t wbinvd_dirty_mask;
827 
828 	unsigned long last_retry_eip;
829 	unsigned long last_retry_addr;
830 
831 	struct {
832 		bool halted;
833 		gfn_t gfns[ASYNC_PF_PER_VCPU];
834 		struct gfn_to_hva_cache data;
835 		u64 msr_en_val; /* MSR_KVM_ASYNC_PF_EN */
836 		u64 msr_int_val; /* MSR_KVM_ASYNC_PF_INT */
837 		u16 vec;
838 		u32 id;
839 		bool send_user_only;
840 		u32 host_apf_flags;
841 		unsigned long nested_apf_token;
842 		bool delivery_as_pf_vmexit;
843 		bool pageready_pending;
844 	} apf;
845 
846 	/* OSVW MSRs (AMD only) */
847 	struct {
848 		u64 length;
849 		u64 status;
850 	} osvw;
851 
852 	struct {
853 		u64 msr_val;
854 		struct gfn_to_hva_cache data;
855 	} pv_eoi;
856 
857 	u64 msr_kvm_poll_control;
858 
859 	/*
860 	 * Indicates the guest is trying to write a gfn that contains one or
861 	 * more of the PTEs used to translate the write itself, i.e. the access
862 	 * is changing its own translation in the guest page tables.  KVM exits
863 	 * to userspace if emulation of the faulting instruction fails and this
864 	 * flag is set, as KVM cannot make forward progress.
865 	 *
866 	 * If emulation fails for a write to guest page tables, KVM unprotects
867 	 * (zaps) the shadow page for the target gfn and resumes the guest to
868 	 * retry the non-emulatable instruction (on hardware).  Unprotecting the
869 	 * gfn doesn't allow forward progress for a self-changing access because
870 	 * doing so also zaps the translation for the gfn, i.e. retrying the
871 	 * instruction will hit a !PRESENT fault, which results in a new shadow
872 	 * page and sends KVM back to square one.
873 	 */
874 	bool write_fault_to_shadow_pgtable;
875 
876 	/* set at EPT violation at this point */
877 	unsigned long exit_qualification;
878 
879 	/* pv related host specific info */
880 	struct {
881 		bool pv_unhalted;
882 	} pv;
883 
884 	int pending_ioapic_eoi;
885 	int pending_external_vector;
886 
887 	/* be preempted when it's in kernel-mode(cpl=0) */
888 	bool preempted_in_kernel;
889 
890 	/* Flush the L1 Data cache for L1TF mitigation on VMENTER */
891 	bool l1tf_flush_l1d;
892 
893 	/* Host CPU on which VM-entry was most recently attempted */
894 	int last_vmentry_cpu;
895 
896 	/* AMD MSRC001_0015 Hardware Configuration */
897 	u64 msr_hwcr;
898 
899 	/* pv related cpuid info */
900 	struct {
901 		/*
902 		 * value of the eax register in the KVM_CPUID_FEATURES CPUID
903 		 * leaf.
904 		 */
905 		u32 features;
906 
907 		/*
908 		 * indicates whether pv emulation should be disabled if features
909 		 * are not present in the guest's cpuid
910 		 */
911 		bool enforce;
912 	} pv_cpuid;
913 
914 	/* Protected Guests */
915 	bool guest_state_protected;
916 
917 	/*
918 	 * Set when PDPTS were loaded directly by the userspace without
919 	 * reading the guest memory
920 	 */
921 	bool pdptrs_from_userspace;
922 
923 #if IS_ENABLED(CONFIG_HYPERV)
924 	hpa_t hv_root_tdp;
925 #endif
926 };
927 
928 struct kvm_lpage_info {
929 	int disallow_lpage;
930 };
931 
932 struct kvm_arch_memory_slot {
933 	struct kvm_rmap_head *rmap[KVM_NR_PAGE_SIZES];
934 	struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
935 	unsigned short *gfn_track[KVM_PAGE_TRACK_MAX];
936 };
937 
938 /*
939  * We use as the mode the number of bits allocated in the LDR for the
940  * logical processor ID.  It happens that these are all powers of two.
941  * This makes it is very easy to detect cases where the APICs are
942  * configured for multiple modes; in that case, we cannot use the map and
943  * hence cannot use kvm_irq_delivery_to_apic_fast either.
944  */
945 #define KVM_APIC_MODE_XAPIC_CLUSTER          4
946 #define KVM_APIC_MODE_XAPIC_FLAT             8
947 #define KVM_APIC_MODE_X2APIC                16
948 
949 struct kvm_apic_map {
950 	struct rcu_head rcu;
951 	u8 mode;
952 	u32 max_apic_id;
953 	union {
954 		struct kvm_lapic *xapic_flat_map[8];
955 		struct kvm_lapic *xapic_cluster_map[16][4];
956 	};
957 	struct kvm_lapic *phys_map[];
958 };
959 
960 /* Hyper-V synthetic debugger (SynDbg)*/
961 struct kvm_hv_syndbg {
962 	struct {
963 		u64 control;
964 		u64 status;
965 		u64 send_page;
966 		u64 recv_page;
967 		u64 pending_page;
968 	} control;
969 	u64 options;
970 };
971 
972 /* Current state of Hyper-V TSC page clocksource */
973 enum hv_tsc_page_status {
974 	/* TSC page was not set up or disabled */
975 	HV_TSC_PAGE_UNSET = 0,
976 	/* TSC page MSR was written by the guest, update pending */
977 	HV_TSC_PAGE_GUEST_CHANGED,
978 	/* TSC page update was triggered from the host side */
979 	HV_TSC_PAGE_HOST_CHANGED,
980 	/* TSC page was properly set up and is currently active  */
981 	HV_TSC_PAGE_SET,
982 	/* TSC page was set up with an inaccessible GPA */
983 	HV_TSC_PAGE_BROKEN,
984 };
985 
986 /* Hyper-V emulation context */
987 struct kvm_hv {
988 	struct mutex hv_lock;
989 	u64 hv_guest_os_id;
990 	u64 hv_hypercall;
991 	u64 hv_tsc_page;
992 	enum hv_tsc_page_status hv_tsc_page_status;
993 
994 	/* Hyper-v based guest crash (NT kernel bugcheck) parameters */
995 	u64 hv_crash_param[HV_X64_MSR_CRASH_PARAMS];
996 	u64 hv_crash_ctl;
997 
998 	struct ms_hyperv_tsc_page tsc_ref;
999 
1000 	struct idr conn_to_evt;
1001 
1002 	u64 hv_reenlightenment_control;
1003 	u64 hv_tsc_emulation_control;
1004 	u64 hv_tsc_emulation_status;
1005 
1006 	/* How many vCPUs have VP index != vCPU index */
1007 	atomic_t num_mismatched_vp_indexes;
1008 
1009 	/*
1010 	 * How many SynICs use 'AutoEOI' feature
1011 	 * (protected by arch.apicv_update_lock)
1012 	 */
1013 	unsigned int synic_auto_eoi_used;
1014 
1015 	struct hv_partition_assist_pg *hv_pa_pg;
1016 	struct kvm_hv_syndbg hv_syndbg;
1017 };
1018 
1019 struct msr_bitmap_range {
1020 	u32 flags;
1021 	u32 nmsrs;
1022 	u32 base;
1023 	unsigned long *bitmap;
1024 };
1025 
1026 /* Xen emulation context */
1027 struct kvm_xen {
1028 	u32 xen_version;
1029 	bool long_mode;
1030 	u8 upcall_vector;
1031 	struct gfn_to_pfn_cache shinfo_cache;
1032 	struct idr evtchn_ports;
1033 	unsigned long poll_mask[BITS_TO_LONGS(KVM_MAX_VCPUS)];
1034 };
1035 
1036 enum kvm_irqchip_mode {
1037 	KVM_IRQCHIP_NONE,
1038 	KVM_IRQCHIP_KERNEL,       /* created with KVM_CREATE_IRQCHIP */
1039 	KVM_IRQCHIP_SPLIT,        /* created with KVM_CAP_SPLIT_IRQCHIP */
1040 };
1041 
1042 struct kvm_x86_msr_filter {
1043 	u8 count;
1044 	bool default_allow:1;
1045 	struct msr_bitmap_range ranges[16];
1046 };
1047 
1048 enum kvm_apicv_inhibit {
1049 	APICV_INHIBIT_REASON_DISABLE,
1050 	APICV_INHIBIT_REASON_HYPERV,
1051 	APICV_INHIBIT_REASON_NESTED,
1052 	APICV_INHIBIT_REASON_IRQWIN,
1053 	APICV_INHIBIT_REASON_PIT_REINJ,
1054 	APICV_INHIBIT_REASON_X2APIC,
1055 	APICV_INHIBIT_REASON_BLOCKIRQ,
1056 	APICV_INHIBIT_REASON_ABSENT,
1057 	APICV_INHIBIT_REASON_SEV,
1058 };
1059 
1060 struct kvm_arch {
1061 	unsigned long n_used_mmu_pages;
1062 	unsigned long n_requested_mmu_pages;
1063 	unsigned long n_max_mmu_pages;
1064 	unsigned int indirect_shadow_pages;
1065 	u8 mmu_valid_gen;
1066 	struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
1067 	struct list_head active_mmu_pages;
1068 	struct list_head zapped_obsolete_pages;
1069 	struct list_head lpage_disallowed_mmu_pages;
1070 	struct kvm_page_track_notifier_node mmu_sp_tracker;
1071 	struct kvm_page_track_notifier_head track_notifier_head;
1072 	/*
1073 	 * Protects marking pages unsync during page faults, as TDP MMU page
1074 	 * faults only take mmu_lock for read.  For simplicity, the unsync
1075 	 * pages lock is always taken when marking pages unsync regardless of
1076 	 * whether mmu_lock is held for read or write.
1077 	 */
1078 	spinlock_t mmu_unsync_pages_lock;
1079 
1080 	struct list_head assigned_dev_head;
1081 	struct iommu_domain *iommu_domain;
1082 	bool iommu_noncoherent;
1083 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA
1084 	atomic_t noncoherent_dma_count;
1085 #define __KVM_HAVE_ARCH_ASSIGNED_DEVICE
1086 	atomic_t assigned_device_count;
1087 	struct kvm_pic *vpic;
1088 	struct kvm_ioapic *vioapic;
1089 	struct kvm_pit *vpit;
1090 	atomic_t vapics_in_nmi_mode;
1091 	struct mutex apic_map_lock;
1092 	struct kvm_apic_map __rcu *apic_map;
1093 	atomic_t apic_map_dirty;
1094 
1095 	/* Protects apic_access_memslot_enabled and apicv_inhibit_reasons */
1096 	struct rw_semaphore apicv_update_lock;
1097 
1098 	bool apic_access_memslot_enabled;
1099 	unsigned long apicv_inhibit_reasons;
1100 
1101 	gpa_t wall_clock;
1102 
1103 	bool mwait_in_guest;
1104 	bool hlt_in_guest;
1105 	bool pause_in_guest;
1106 	bool cstate_in_guest;
1107 
1108 	unsigned long irq_sources_bitmap;
1109 	s64 kvmclock_offset;
1110 
1111 	/*
1112 	 * This also protects nr_vcpus_matched_tsc which is read from a
1113 	 * preemption-disabled region, so it must be a raw spinlock.
1114 	 */
1115 	raw_spinlock_t tsc_write_lock;
1116 	u64 last_tsc_nsec;
1117 	u64 last_tsc_write;
1118 	u32 last_tsc_khz;
1119 	u64 last_tsc_offset;
1120 	u64 cur_tsc_nsec;
1121 	u64 cur_tsc_write;
1122 	u64 cur_tsc_offset;
1123 	u64 cur_tsc_generation;
1124 	int nr_vcpus_matched_tsc;
1125 
1126 	u32 default_tsc_khz;
1127 
1128 	seqcount_raw_spinlock_t pvclock_sc;
1129 	bool use_master_clock;
1130 	u64 master_kernel_ns;
1131 	u64 master_cycle_now;
1132 	struct delayed_work kvmclock_update_work;
1133 	struct delayed_work kvmclock_sync_work;
1134 
1135 	struct kvm_xen_hvm_config xen_hvm_config;
1136 
1137 	/* reads protected by irq_srcu, writes by irq_lock */
1138 	struct hlist_head mask_notifier_list;
1139 
1140 	struct kvm_hv hyperv;
1141 	struct kvm_xen xen;
1142 
1143 	bool backwards_tsc_observed;
1144 	bool boot_vcpu_runs_old_kvmclock;
1145 	u32 bsp_vcpu_id;
1146 
1147 	u64 disabled_quirks;
1148 	int cpu_dirty_logging_count;
1149 
1150 	enum kvm_irqchip_mode irqchip_mode;
1151 	u8 nr_reserved_ioapic_pins;
1152 
1153 	bool disabled_lapic_found;
1154 
1155 	bool x2apic_format;
1156 	bool x2apic_broadcast_quirk_disabled;
1157 
1158 	bool guest_can_read_msr_platform_info;
1159 	bool exception_payload_enabled;
1160 
1161 	bool bus_lock_detection_enabled;
1162 	bool enable_pmu;
1163 	/*
1164 	 * If exit_on_emulation_error is set, and the in-kernel instruction
1165 	 * emulator fails to emulate an instruction, allow userspace
1166 	 * the opportunity to look at it.
1167 	 */
1168 	bool exit_on_emulation_error;
1169 
1170 	/* Deflect RDMSR and WRMSR to user space when they trigger a #GP */
1171 	u32 user_space_msr_mask;
1172 	struct kvm_x86_msr_filter __rcu *msr_filter;
1173 
1174 	u32 hypercall_exit_enabled;
1175 
1176 	/* Guest can access the SGX PROVISIONKEY. */
1177 	bool sgx_provisioning_allowed;
1178 
1179 	struct kvm_pmu_event_filter __rcu *pmu_event_filter;
1180 	struct task_struct *nx_lpage_recovery_thread;
1181 
1182 #ifdef CONFIG_X86_64
1183 	/*
1184 	 * Whether the TDP MMU is enabled for this VM. This contains a
1185 	 * snapshot of the TDP MMU module parameter from when the VM was
1186 	 * created and remains unchanged for the life of the VM. If this is
1187 	 * true, TDP MMU handler functions will run for various MMU
1188 	 * operations.
1189 	 */
1190 	bool tdp_mmu_enabled;
1191 
1192 	/*
1193 	 * List of struct kvm_mmu_pages being used as roots.
1194 	 * All struct kvm_mmu_pages in the list should have
1195 	 * tdp_mmu_page set.
1196 	 *
1197 	 * For reads, this list is protected by:
1198 	 *	the MMU lock in read mode + RCU or
1199 	 *	the MMU lock in write mode
1200 	 *
1201 	 * For writes, this list is protected by:
1202 	 *	the MMU lock in read mode + the tdp_mmu_pages_lock or
1203 	 *	the MMU lock in write mode
1204 	 *
1205 	 * Roots will remain in the list until their tdp_mmu_root_count
1206 	 * drops to zero, at which point the thread that decremented the
1207 	 * count to zero should removed the root from the list and clean
1208 	 * it up, freeing the root after an RCU grace period.
1209 	 */
1210 	struct list_head tdp_mmu_roots;
1211 
1212 	/*
1213 	 * List of struct kvmp_mmu_pages not being used as roots.
1214 	 * All struct kvm_mmu_pages in the list should have
1215 	 * tdp_mmu_page set and a tdp_mmu_root_count of 0.
1216 	 */
1217 	struct list_head tdp_mmu_pages;
1218 
1219 	/*
1220 	 * Protects accesses to the following fields when the MMU lock
1221 	 * is held in read mode:
1222 	 *  - tdp_mmu_roots (above)
1223 	 *  - tdp_mmu_pages (above)
1224 	 *  - the link field of struct kvm_mmu_pages used by the TDP MMU
1225 	 *  - lpage_disallowed_mmu_pages
1226 	 *  - the lpage_disallowed_link field of struct kvm_mmu_pages used
1227 	 *    by the TDP MMU
1228 	 * It is acceptable, but not necessary, to acquire this lock when
1229 	 * the thread holds the MMU lock in write mode.
1230 	 */
1231 	spinlock_t tdp_mmu_pages_lock;
1232 	struct workqueue_struct *tdp_mmu_zap_wq;
1233 #endif /* CONFIG_X86_64 */
1234 
1235 	/*
1236 	 * If set, at least one shadow root has been allocated. This flag
1237 	 * is used as one input when determining whether certain memslot
1238 	 * related allocations are necessary.
1239 	 */
1240 	bool shadow_root_allocated;
1241 
1242 #if IS_ENABLED(CONFIG_HYPERV)
1243 	hpa_t	hv_root_tdp;
1244 	spinlock_t hv_root_tdp_lock;
1245 #endif
1246 };
1247 
1248 struct kvm_vm_stat {
1249 	struct kvm_vm_stat_generic generic;
1250 	u64 mmu_shadow_zapped;
1251 	u64 mmu_pte_write;
1252 	u64 mmu_pde_zapped;
1253 	u64 mmu_flooded;
1254 	u64 mmu_recycled;
1255 	u64 mmu_cache_miss;
1256 	u64 mmu_unsync;
1257 	union {
1258 		struct {
1259 			atomic64_t pages_4k;
1260 			atomic64_t pages_2m;
1261 			atomic64_t pages_1g;
1262 		};
1263 		atomic64_t pages[KVM_NR_PAGE_SIZES];
1264 	};
1265 	u64 nx_lpage_splits;
1266 	u64 max_mmu_page_hash_collisions;
1267 	u64 max_mmu_rmap_size;
1268 };
1269 
1270 struct kvm_vcpu_stat {
1271 	struct kvm_vcpu_stat_generic generic;
1272 	u64 pf_taken;
1273 	u64 pf_fixed;
1274 	u64 pf_emulate;
1275 	u64 pf_spurious;
1276 	u64 pf_fast;
1277 	u64 pf_mmio_spte_created;
1278 	u64 pf_guest;
1279 	u64 tlb_flush;
1280 	u64 invlpg;
1281 
1282 	u64 exits;
1283 	u64 io_exits;
1284 	u64 mmio_exits;
1285 	u64 signal_exits;
1286 	u64 irq_window_exits;
1287 	u64 nmi_window_exits;
1288 	u64 l1d_flush;
1289 	u64 halt_exits;
1290 	u64 request_irq_exits;
1291 	u64 irq_exits;
1292 	u64 host_state_reload;
1293 	u64 fpu_reload;
1294 	u64 insn_emulation;
1295 	u64 insn_emulation_fail;
1296 	u64 hypercalls;
1297 	u64 irq_injections;
1298 	u64 nmi_injections;
1299 	u64 req_event;
1300 	u64 nested_run;
1301 	u64 directed_yield_attempted;
1302 	u64 directed_yield_successful;
1303 	u64 guest_mode;
1304 };
1305 
1306 struct x86_instruction_info;
1307 
1308 struct msr_data {
1309 	bool host_initiated;
1310 	u32 index;
1311 	u64 data;
1312 };
1313 
1314 struct kvm_lapic_irq {
1315 	u32 vector;
1316 	u16 delivery_mode;
1317 	u16 dest_mode;
1318 	bool level;
1319 	u16 trig_mode;
1320 	u32 shorthand;
1321 	u32 dest_id;
1322 	bool msi_redir_hint;
1323 };
1324 
1325 static inline u16 kvm_lapic_irq_dest_mode(bool dest_mode_logical)
1326 {
1327 	return dest_mode_logical ? APIC_DEST_LOGICAL : APIC_DEST_PHYSICAL;
1328 }
1329 
1330 struct kvm_x86_ops {
1331 	const char *name;
1332 
1333 	int (*hardware_enable)(void);
1334 	void (*hardware_disable)(void);
1335 	void (*hardware_unsetup)(void);
1336 	bool (*has_emulated_msr)(struct kvm *kvm, u32 index);
1337 	void (*vcpu_after_set_cpuid)(struct kvm_vcpu *vcpu);
1338 
1339 	unsigned int vm_size;
1340 	int (*vm_init)(struct kvm *kvm);
1341 	void (*vm_destroy)(struct kvm *kvm);
1342 
1343 	/* Create, but do not attach this VCPU */
1344 	int (*vcpu_create)(struct kvm_vcpu *vcpu);
1345 	void (*vcpu_free)(struct kvm_vcpu *vcpu);
1346 	void (*vcpu_reset)(struct kvm_vcpu *vcpu, bool init_event);
1347 
1348 	void (*prepare_switch_to_guest)(struct kvm_vcpu *vcpu);
1349 	void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
1350 	void (*vcpu_put)(struct kvm_vcpu *vcpu);
1351 
1352 	void (*update_exception_bitmap)(struct kvm_vcpu *vcpu);
1353 	int (*get_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1354 	int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1355 	u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
1356 	void (*get_segment)(struct kvm_vcpu *vcpu,
1357 			    struct kvm_segment *var, int seg);
1358 	int (*get_cpl)(struct kvm_vcpu *vcpu);
1359 	void (*set_segment)(struct kvm_vcpu *vcpu,
1360 			    struct kvm_segment *var, int seg);
1361 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
1362 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
1363 	void (*post_set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
1364 	bool (*is_valid_cr4)(struct kvm_vcpu *vcpu, unsigned long cr0);
1365 	void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
1366 	int (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
1367 	void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1368 	void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1369 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1370 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1371 	void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu);
1372 	void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
1373 	void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
1374 	unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
1375 	void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
1376 	bool (*get_if_flag)(struct kvm_vcpu *vcpu);
1377 
1378 	void (*flush_tlb_all)(struct kvm_vcpu *vcpu);
1379 	void (*flush_tlb_current)(struct kvm_vcpu *vcpu);
1380 	int  (*tlb_remote_flush)(struct kvm *kvm);
1381 	int  (*tlb_remote_flush_with_range)(struct kvm *kvm,
1382 			struct kvm_tlb_range *range);
1383 
1384 	/*
1385 	 * Flush any TLB entries associated with the given GVA.
1386 	 * Does not need to flush GPA->HPA mappings.
1387 	 * Can potentially get non-canonical addresses through INVLPGs, which
1388 	 * the implementation may choose to ignore if appropriate.
1389 	 */
1390 	void (*flush_tlb_gva)(struct kvm_vcpu *vcpu, gva_t addr);
1391 
1392 	/*
1393 	 * Flush any TLB entries created by the guest.  Like tlb_flush_gva(),
1394 	 * does not need to flush GPA->HPA mappings.
1395 	 */
1396 	void (*flush_tlb_guest)(struct kvm_vcpu *vcpu);
1397 
1398 	int (*vcpu_pre_run)(struct kvm_vcpu *vcpu);
1399 	enum exit_fastpath_completion (*vcpu_run)(struct kvm_vcpu *vcpu);
1400 	int (*handle_exit)(struct kvm_vcpu *vcpu,
1401 		enum exit_fastpath_completion exit_fastpath);
1402 	int (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
1403 	void (*update_emulated_instruction)(struct kvm_vcpu *vcpu);
1404 	void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
1405 	u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu);
1406 	void (*patch_hypercall)(struct kvm_vcpu *vcpu,
1407 				unsigned char *hypercall_addr);
1408 	void (*inject_irq)(struct kvm_vcpu *vcpu);
1409 	void (*inject_nmi)(struct kvm_vcpu *vcpu);
1410 	void (*queue_exception)(struct kvm_vcpu *vcpu);
1411 	void (*cancel_injection)(struct kvm_vcpu *vcpu);
1412 	int (*interrupt_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1413 	int (*nmi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1414 	bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
1415 	void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
1416 	void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
1417 	void (*enable_irq_window)(struct kvm_vcpu *vcpu);
1418 	void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
1419 	bool (*check_apicv_inhibit_reasons)(enum kvm_apicv_inhibit reason);
1420 	void (*refresh_apicv_exec_ctrl)(struct kvm_vcpu *vcpu);
1421 	void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
1422 	void (*hwapic_isr_update)(struct kvm_vcpu *vcpu, int isr);
1423 	bool (*guest_apic_has_interrupt)(struct kvm_vcpu *vcpu);
1424 	void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
1425 	void (*set_virtual_apic_mode)(struct kvm_vcpu *vcpu);
1426 	void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu);
1427 	void (*deliver_interrupt)(struct kvm_lapic *apic, int delivery_mode,
1428 				  int trig_mode, int vector);
1429 	int (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
1430 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
1431 	int (*set_identity_map_addr)(struct kvm *kvm, u64 ident_addr);
1432 	u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
1433 
1434 	void (*load_mmu_pgd)(struct kvm_vcpu *vcpu, hpa_t root_hpa,
1435 			     int root_level);
1436 
1437 	bool (*has_wbinvd_exit)(void);
1438 
1439 	u64 (*get_l2_tsc_offset)(struct kvm_vcpu *vcpu);
1440 	u64 (*get_l2_tsc_multiplier)(struct kvm_vcpu *vcpu);
1441 	void (*write_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
1442 	void (*write_tsc_multiplier)(struct kvm_vcpu *vcpu, u64 multiplier);
1443 
1444 	/*
1445 	 * Retrieve somewhat arbitrary exit information.  Intended to
1446 	 * be used only from within tracepoints or error paths.
1447 	 */
1448 	void (*get_exit_info)(struct kvm_vcpu *vcpu, u32 *reason,
1449 			      u64 *info1, u64 *info2,
1450 			      u32 *exit_int_info, u32 *exit_int_info_err_code);
1451 
1452 	int (*check_intercept)(struct kvm_vcpu *vcpu,
1453 			       struct x86_instruction_info *info,
1454 			       enum x86_intercept_stage stage,
1455 			       struct x86_exception *exception);
1456 	void (*handle_exit_irqoff)(struct kvm_vcpu *vcpu);
1457 
1458 	void (*request_immediate_exit)(struct kvm_vcpu *vcpu);
1459 
1460 	void (*sched_in)(struct kvm_vcpu *kvm, int cpu);
1461 
1462 	/*
1463 	 * Size of the CPU's dirty log buffer, i.e. VMX's PML buffer.  A zero
1464 	 * value indicates CPU dirty logging is unsupported or disabled.
1465 	 */
1466 	int cpu_dirty_log_size;
1467 	void (*update_cpu_dirty_logging)(struct kvm_vcpu *vcpu);
1468 
1469 	const struct kvm_x86_nested_ops *nested_ops;
1470 
1471 	void (*vcpu_blocking)(struct kvm_vcpu *vcpu);
1472 	void (*vcpu_unblocking)(struct kvm_vcpu *vcpu);
1473 
1474 	int (*pi_update_irte)(struct kvm *kvm, unsigned int host_irq,
1475 			      uint32_t guest_irq, bool set);
1476 	void (*pi_start_assignment)(struct kvm *kvm);
1477 	void (*apicv_post_state_restore)(struct kvm_vcpu *vcpu);
1478 	bool (*dy_apicv_has_pending_interrupt)(struct kvm_vcpu *vcpu);
1479 
1480 	int (*set_hv_timer)(struct kvm_vcpu *vcpu, u64 guest_deadline_tsc,
1481 			    bool *expired);
1482 	void (*cancel_hv_timer)(struct kvm_vcpu *vcpu);
1483 
1484 	void (*setup_mce)(struct kvm_vcpu *vcpu);
1485 
1486 	int (*smi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1487 	int (*enter_smm)(struct kvm_vcpu *vcpu, char *smstate);
1488 	int (*leave_smm)(struct kvm_vcpu *vcpu, const char *smstate);
1489 	void (*enable_smi_window)(struct kvm_vcpu *vcpu);
1490 
1491 	int (*mem_enc_ioctl)(struct kvm *kvm, void __user *argp);
1492 	int (*mem_enc_register_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1493 	int (*mem_enc_unregister_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1494 	int (*vm_copy_enc_context_from)(struct kvm *kvm, unsigned int source_fd);
1495 	int (*vm_move_enc_context_from)(struct kvm *kvm, unsigned int source_fd);
1496 	void (*guest_memory_reclaimed)(struct kvm *kvm);
1497 
1498 	int (*get_msr_feature)(struct kvm_msr_entry *entry);
1499 
1500 	bool (*can_emulate_instruction)(struct kvm_vcpu *vcpu, int emul_type,
1501 					void *insn, int insn_len);
1502 
1503 	bool (*apic_init_signal_blocked)(struct kvm_vcpu *vcpu);
1504 	int (*enable_direct_tlbflush)(struct kvm_vcpu *vcpu);
1505 
1506 	void (*migrate_timers)(struct kvm_vcpu *vcpu);
1507 	void (*msr_filter_changed)(struct kvm_vcpu *vcpu);
1508 	int (*complete_emulated_msr)(struct kvm_vcpu *vcpu, int err);
1509 
1510 	void (*vcpu_deliver_sipi_vector)(struct kvm_vcpu *vcpu, u8 vector);
1511 
1512 	/*
1513 	 * Returns vCPU specific APICv inhibit reasons
1514 	 */
1515 	unsigned long (*vcpu_get_apicv_inhibit_reasons)(struct kvm_vcpu *vcpu);
1516 };
1517 
1518 struct kvm_x86_nested_ops {
1519 	void (*leave_nested)(struct kvm_vcpu *vcpu);
1520 	int (*check_events)(struct kvm_vcpu *vcpu);
1521 	bool (*handle_page_fault_workaround)(struct kvm_vcpu *vcpu,
1522 					     struct x86_exception *fault);
1523 	bool (*hv_timer_pending)(struct kvm_vcpu *vcpu);
1524 	void (*triple_fault)(struct kvm_vcpu *vcpu);
1525 	int (*get_state)(struct kvm_vcpu *vcpu,
1526 			 struct kvm_nested_state __user *user_kvm_nested_state,
1527 			 unsigned user_data_size);
1528 	int (*set_state)(struct kvm_vcpu *vcpu,
1529 			 struct kvm_nested_state __user *user_kvm_nested_state,
1530 			 struct kvm_nested_state *kvm_state);
1531 	bool (*get_nested_state_pages)(struct kvm_vcpu *vcpu);
1532 	int (*write_log_dirty)(struct kvm_vcpu *vcpu, gpa_t l2_gpa);
1533 
1534 	int (*enable_evmcs)(struct kvm_vcpu *vcpu,
1535 			    uint16_t *vmcs_version);
1536 	uint16_t (*get_evmcs_version)(struct kvm_vcpu *vcpu);
1537 };
1538 
1539 struct kvm_x86_init_ops {
1540 	int (*cpu_has_kvm_support)(void);
1541 	int (*disabled_by_bios)(void);
1542 	int (*check_processor_compatibility)(void);
1543 	int (*hardware_setup)(void);
1544 	unsigned int (*handle_intel_pt_intr)(void);
1545 
1546 	struct kvm_x86_ops *runtime_ops;
1547 	struct kvm_pmu_ops *pmu_ops;
1548 };
1549 
1550 struct kvm_arch_async_pf {
1551 	u32 token;
1552 	gfn_t gfn;
1553 	unsigned long cr3;
1554 	bool direct_map;
1555 };
1556 
1557 extern u32 __read_mostly kvm_nr_uret_msrs;
1558 extern u64 __read_mostly host_efer;
1559 extern bool __read_mostly allow_smaller_maxphyaddr;
1560 extern bool __read_mostly enable_apicv;
1561 extern struct kvm_x86_ops kvm_x86_ops;
1562 
1563 #define KVM_X86_OP(func) \
1564 	DECLARE_STATIC_CALL(kvm_x86_##func, *(((struct kvm_x86_ops *)0)->func));
1565 #define KVM_X86_OP_OPTIONAL KVM_X86_OP
1566 #define KVM_X86_OP_OPTIONAL_RET0 KVM_X86_OP
1567 #include <asm/kvm-x86-ops.h>
1568 
1569 #define __KVM_HAVE_ARCH_VM_ALLOC
1570 static inline struct kvm *kvm_arch_alloc_vm(void)
1571 {
1572 	return __vmalloc(kvm_x86_ops.vm_size, GFP_KERNEL_ACCOUNT | __GFP_ZERO);
1573 }
1574 
1575 #define __KVM_HAVE_ARCH_VM_FREE
1576 void kvm_arch_free_vm(struct kvm *kvm);
1577 
1578 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
1579 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
1580 {
1581 	if (kvm_x86_ops.tlb_remote_flush &&
1582 	    !static_call(kvm_x86_tlb_remote_flush)(kvm))
1583 		return 0;
1584 	else
1585 		return -ENOTSUPP;
1586 }
1587 
1588 #define kvm_arch_pmi_in_guest(vcpu) \
1589 	((vcpu) && (vcpu)->arch.handling_intr_from_guest)
1590 
1591 void kvm_mmu_x86_module_init(void);
1592 int kvm_mmu_vendor_module_init(void);
1593 void kvm_mmu_vendor_module_exit(void);
1594 
1595 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
1596 int kvm_mmu_create(struct kvm_vcpu *vcpu);
1597 int kvm_mmu_init_vm(struct kvm *kvm);
1598 void kvm_mmu_uninit_vm(struct kvm *kvm);
1599 
1600 void kvm_mmu_after_set_cpuid(struct kvm_vcpu *vcpu);
1601 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
1602 void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
1603 				      const struct kvm_memory_slot *memslot,
1604 				      int start_level);
1605 void kvm_mmu_slot_try_split_huge_pages(struct kvm *kvm,
1606 				       const struct kvm_memory_slot *memslot,
1607 				       int target_level);
1608 void kvm_mmu_try_split_huge_pages(struct kvm *kvm,
1609 				  const struct kvm_memory_slot *memslot,
1610 				  u64 start, u64 end,
1611 				  int target_level);
1612 void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
1613 				   const struct kvm_memory_slot *memslot);
1614 void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
1615 				   const struct kvm_memory_slot *memslot);
1616 void kvm_mmu_zap_all(struct kvm *kvm);
1617 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, u64 gen);
1618 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned long kvm_nr_mmu_pages);
1619 
1620 int load_pdptrs(struct kvm_vcpu *vcpu, unsigned long cr3);
1621 
1622 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
1623 			  const void *val, int bytes);
1624 
1625 struct kvm_irq_mask_notifier {
1626 	void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
1627 	int irq;
1628 	struct hlist_node link;
1629 };
1630 
1631 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
1632 				    struct kvm_irq_mask_notifier *kimn);
1633 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
1634 				      struct kvm_irq_mask_notifier *kimn);
1635 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
1636 			     bool mask);
1637 
1638 extern bool tdp_enabled;
1639 
1640 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
1641 
1642 /* control of guest tsc rate supported? */
1643 extern bool kvm_has_tsc_control;
1644 /* maximum supported tsc_khz for guests */
1645 extern u32  kvm_max_guest_tsc_khz;
1646 /* number of bits of the fractional part of the TSC scaling ratio */
1647 extern u8   kvm_tsc_scaling_ratio_frac_bits;
1648 /* maximum allowed value of TSC scaling ratio */
1649 extern u64  kvm_max_tsc_scaling_ratio;
1650 /* 1ull << kvm_tsc_scaling_ratio_frac_bits */
1651 extern u64  kvm_default_tsc_scaling_ratio;
1652 /* bus lock detection supported? */
1653 extern bool kvm_has_bus_lock_exit;
1654 
1655 extern u64 kvm_mce_cap_supported;
1656 
1657 /*
1658  * EMULTYPE_NO_DECODE - Set when re-emulating an instruction (after completing
1659  *			userspace I/O) to indicate that the emulation context
1660  *			should be reused as is, i.e. skip initialization of
1661  *			emulation context, instruction fetch and decode.
1662  *
1663  * EMULTYPE_TRAP_UD - Set when emulating an intercepted #UD from hardware.
1664  *		      Indicates that only select instructions (tagged with
1665  *		      EmulateOnUD) should be emulated (to minimize the emulator
1666  *		      attack surface).  See also EMULTYPE_TRAP_UD_FORCED.
1667  *
1668  * EMULTYPE_SKIP - Set when emulating solely to skip an instruction, i.e. to
1669  *		   decode the instruction length.  For use *only* by
1670  *		   kvm_x86_ops.skip_emulated_instruction() implementations if
1671  *		   EMULTYPE_COMPLETE_USER_EXIT is not set.
1672  *
1673  * EMULTYPE_ALLOW_RETRY_PF - Set when the emulator should resume the guest to
1674  *			     retry native execution under certain conditions,
1675  *			     Can only be set in conjunction with EMULTYPE_PF.
1676  *
1677  * EMULTYPE_TRAP_UD_FORCED - Set when emulating an intercepted #UD that was
1678  *			     triggered by KVM's magic "force emulation" prefix,
1679  *			     which is opt in via module param (off by default).
1680  *			     Bypasses EmulateOnUD restriction despite emulating
1681  *			     due to an intercepted #UD (see EMULTYPE_TRAP_UD).
1682  *			     Used to test the full emulator from userspace.
1683  *
1684  * EMULTYPE_VMWARE_GP - Set when emulating an intercepted #GP for VMware
1685  *			backdoor emulation, which is opt in via module param.
1686  *			VMware backdoor emulation handles select instructions
1687  *			and reinjects the #GP for all other cases.
1688  *
1689  * EMULTYPE_PF - Set when emulating MMIO by way of an intercepted #PF, in which
1690  *		 case the CR2/GPA value pass on the stack is valid.
1691  *
1692  * EMULTYPE_COMPLETE_USER_EXIT - Set when the emulator should update interruptibility
1693  *				 state and inject single-step #DBs after skipping
1694  *				 an instruction (after completing userspace I/O).
1695  */
1696 #define EMULTYPE_NO_DECODE	    (1 << 0)
1697 #define EMULTYPE_TRAP_UD	    (1 << 1)
1698 #define EMULTYPE_SKIP		    (1 << 2)
1699 #define EMULTYPE_ALLOW_RETRY_PF	    (1 << 3)
1700 #define EMULTYPE_TRAP_UD_FORCED	    (1 << 4)
1701 #define EMULTYPE_VMWARE_GP	    (1 << 5)
1702 #define EMULTYPE_PF		    (1 << 6)
1703 #define EMULTYPE_COMPLETE_USER_EXIT (1 << 7)
1704 
1705 int kvm_emulate_instruction(struct kvm_vcpu *vcpu, int emulation_type);
1706 int kvm_emulate_instruction_from_buffer(struct kvm_vcpu *vcpu,
1707 					void *insn, int insn_len);
1708 void __kvm_prepare_emulation_failure_exit(struct kvm_vcpu *vcpu,
1709 					  u64 *data, u8 ndata);
1710 void kvm_prepare_emulation_failure_exit(struct kvm_vcpu *vcpu);
1711 
1712 void kvm_enable_efer_bits(u64);
1713 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
1714 int __kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data, bool host_initiated);
1715 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data);
1716 int kvm_set_msr(struct kvm_vcpu *vcpu, u32 index, u64 data);
1717 int kvm_emulate_rdmsr(struct kvm_vcpu *vcpu);
1718 int kvm_emulate_wrmsr(struct kvm_vcpu *vcpu);
1719 int kvm_emulate_as_nop(struct kvm_vcpu *vcpu);
1720 int kvm_emulate_invd(struct kvm_vcpu *vcpu);
1721 int kvm_emulate_mwait(struct kvm_vcpu *vcpu);
1722 int kvm_handle_invalid_op(struct kvm_vcpu *vcpu);
1723 int kvm_emulate_monitor(struct kvm_vcpu *vcpu);
1724 
1725 int kvm_fast_pio(struct kvm_vcpu *vcpu, int size, unsigned short port, int in);
1726 int kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
1727 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
1728 int kvm_emulate_halt_noskip(struct kvm_vcpu *vcpu);
1729 int kvm_emulate_ap_reset_hold(struct kvm_vcpu *vcpu);
1730 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
1731 
1732 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
1733 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
1734 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector);
1735 
1736 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
1737 		    int reason, bool has_error_code, u32 error_code);
1738 
1739 void kvm_post_set_cr0(struct kvm_vcpu *vcpu, unsigned long old_cr0, unsigned long cr0);
1740 void kvm_post_set_cr4(struct kvm_vcpu *vcpu, unsigned long old_cr4, unsigned long cr4);
1741 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
1742 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
1743 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
1744 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
1745 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
1746 void kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
1747 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
1748 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
1749 int kvm_emulate_xsetbv(struct kvm_vcpu *vcpu);
1750 
1751 int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1752 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1753 
1754 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
1755 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
1756 int kvm_emulate_rdpmc(struct kvm_vcpu *vcpu);
1757 
1758 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1759 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1760 void kvm_queue_exception_p(struct kvm_vcpu *vcpu, unsigned nr, unsigned long payload);
1761 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1762 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1763 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
1764 bool kvm_inject_emulated_page_fault(struct kvm_vcpu *vcpu,
1765 				    struct x86_exception *fault);
1766 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
1767 bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr);
1768 
1769 static inline int __kvm_irq_line_state(unsigned long *irq_state,
1770 				       int irq_source_id, int level)
1771 {
1772 	/* Logical OR for level trig interrupt */
1773 	if (level)
1774 		__set_bit(irq_source_id, irq_state);
1775 	else
1776 		__clear_bit(irq_source_id, irq_state);
1777 
1778 	return !!(*irq_state);
1779 }
1780 
1781 #define KVM_MMU_ROOT_CURRENT		BIT(0)
1782 #define KVM_MMU_ROOT_PREVIOUS(i)	BIT(1+i)
1783 #define KVM_MMU_ROOTS_ALL		(~0UL)
1784 
1785 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
1786 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
1787 
1788 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
1789 
1790 void kvm_update_dr7(struct kvm_vcpu *vcpu);
1791 
1792 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
1793 void kvm_mmu_free_roots(struct kvm *kvm, struct kvm_mmu *mmu,
1794 			ulong roots_to_free);
1795 void kvm_mmu_free_guest_mode_roots(struct kvm *kvm, struct kvm_mmu *mmu);
1796 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
1797 			      struct x86_exception *exception);
1798 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
1799 			       struct x86_exception *exception);
1800 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
1801 			       struct x86_exception *exception);
1802 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
1803 				struct x86_exception *exception);
1804 
1805 bool kvm_apicv_activated(struct kvm *kvm);
1806 bool kvm_vcpu_apicv_activated(struct kvm_vcpu *vcpu);
1807 void kvm_vcpu_update_apicv(struct kvm_vcpu *vcpu);
1808 void __kvm_set_or_clear_apicv_inhibit(struct kvm *kvm,
1809 				      enum kvm_apicv_inhibit reason, bool set);
1810 void kvm_set_or_clear_apicv_inhibit(struct kvm *kvm,
1811 				    enum kvm_apicv_inhibit reason, bool set);
1812 
1813 static inline void kvm_set_apicv_inhibit(struct kvm *kvm,
1814 					 enum kvm_apicv_inhibit reason)
1815 {
1816 	kvm_set_or_clear_apicv_inhibit(kvm, reason, true);
1817 }
1818 
1819 static inline void kvm_clear_apicv_inhibit(struct kvm *kvm,
1820 					   enum kvm_apicv_inhibit reason)
1821 {
1822 	kvm_set_or_clear_apicv_inhibit(kvm, reason, false);
1823 }
1824 
1825 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
1826 
1827 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u64 error_code,
1828 		       void *insn, int insn_len);
1829 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
1830 void kvm_mmu_invalidate_gva(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1831 			    gva_t gva, hpa_t root_hpa);
1832 void kvm_mmu_invpcid_gva(struct kvm_vcpu *vcpu, gva_t gva, unsigned long pcid);
1833 void kvm_mmu_new_pgd(struct kvm_vcpu *vcpu, gpa_t new_pgd);
1834 
1835 void kvm_configure_mmu(bool enable_tdp, int tdp_forced_root_level,
1836 		       int tdp_max_root_level, int tdp_huge_page_level);
1837 
1838 static inline u16 kvm_read_ldt(void)
1839 {
1840 	u16 ldt;
1841 	asm("sldt %0" : "=g"(ldt));
1842 	return ldt;
1843 }
1844 
1845 static inline void kvm_load_ldt(u16 sel)
1846 {
1847 	asm("lldt %0" : : "rm"(sel));
1848 }
1849 
1850 #ifdef CONFIG_X86_64
1851 static inline unsigned long read_msr(unsigned long msr)
1852 {
1853 	u64 value;
1854 
1855 	rdmsrl(msr, value);
1856 	return value;
1857 }
1858 #endif
1859 
1860 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
1861 {
1862 	kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
1863 }
1864 
1865 #define TSS_IOPB_BASE_OFFSET 0x66
1866 #define TSS_BASE_SIZE 0x68
1867 #define TSS_IOPB_SIZE (65536 / 8)
1868 #define TSS_REDIRECTION_SIZE (256 / 8)
1869 #define RMODE_TSS_SIZE							\
1870 	(TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
1871 
1872 enum {
1873 	TASK_SWITCH_CALL = 0,
1874 	TASK_SWITCH_IRET = 1,
1875 	TASK_SWITCH_JMP = 2,
1876 	TASK_SWITCH_GATE = 3,
1877 };
1878 
1879 #define HF_GIF_MASK		(1 << 0)
1880 #define HF_NMI_MASK		(1 << 3)
1881 #define HF_IRET_MASK		(1 << 4)
1882 #define HF_GUEST_MASK		(1 << 5) /* VCPU is in guest-mode */
1883 #define HF_SMM_MASK		(1 << 6)
1884 #define HF_SMM_INSIDE_NMI_MASK	(1 << 7)
1885 
1886 #define __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
1887 #define KVM_ADDRESS_SPACE_NUM 2
1888 
1889 #define kvm_arch_vcpu_memslots_id(vcpu) ((vcpu)->arch.hflags & HF_SMM_MASK ? 1 : 0)
1890 #define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, (role).smm)
1891 
1892 #define KVM_ARCH_WANT_MMU_NOTIFIER
1893 
1894 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1895 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1896 int kvm_cpu_has_extint(struct kvm_vcpu *v);
1897 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1898 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1899 void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event);
1900 
1901 int kvm_pv_send_ipi(struct kvm *kvm, unsigned long ipi_bitmap_low,
1902 		    unsigned long ipi_bitmap_high, u32 min,
1903 		    unsigned long icr, int op_64_bit);
1904 
1905 int kvm_add_user_return_msr(u32 msr);
1906 int kvm_find_user_return_msr(u32 msr);
1907 int kvm_set_user_return_msr(unsigned index, u64 val, u64 mask);
1908 
1909 static inline bool kvm_is_supported_user_return_msr(u32 msr)
1910 {
1911 	return kvm_find_user_return_msr(msr) >= 0;
1912 }
1913 
1914 u64 kvm_scale_tsc(u64 tsc, u64 ratio);
1915 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc);
1916 u64 kvm_calc_nested_tsc_offset(u64 l1_offset, u64 l2_offset, u64 l2_multiplier);
1917 u64 kvm_calc_nested_tsc_multiplier(u64 l1_multiplier, u64 l2_multiplier);
1918 
1919 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu);
1920 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1921 
1922 void kvm_make_scan_ioapic_request(struct kvm *kvm);
1923 void kvm_make_scan_ioapic_request_mask(struct kvm *kvm,
1924 				       unsigned long *vcpu_bitmap);
1925 
1926 bool kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1927 				     struct kvm_async_pf *work);
1928 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1929 				 struct kvm_async_pf *work);
1930 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1931 			       struct kvm_async_pf *work);
1932 void kvm_arch_async_page_present_queued(struct kvm_vcpu *vcpu);
1933 bool kvm_arch_can_dequeue_async_page_present(struct kvm_vcpu *vcpu);
1934 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1935 
1936 int kvm_skip_emulated_instruction(struct kvm_vcpu *vcpu);
1937 int kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1938 void __kvm_request_immediate_exit(struct kvm_vcpu *vcpu);
1939 
1940 void __user *__x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa,
1941 				     u32 size);
1942 bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu);
1943 bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu);
1944 
1945 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
1946 			     struct kvm_vcpu **dest_vcpu);
1947 
1948 void kvm_set_msi_irq(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e,
1949 		     struct kvm_lapic_irq *irq);
1950 
1951 static inline bool kvm_irq_is_postable(struct kvm_lapic_irq *irq)
1952 {
1953 	/* We can only post Fixed and LowPrio IRQs */
1954 	return (irq->delivery_mode == APIC_DM_FIXED ||
1955 		irq->delivery_mode == APIC_DM_LOWEST);
1956 }
1957 
1958 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu)
1959 {
1960 	static_call_cond(kvm_x86_vcpu_blocking)(vcpu);
1961 }
1962 
1963 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu)
1964 {
1965 	static_call_cond(kvm_x86_vcpu_unblocking)(vcpu);
1966 }
1967 
1968 static inline int kvm_cpu_get_apicid(int mps_cpu)
1969 {
1970 #ifdef CONFIG_X86_LOCAL_APIC
1971 	return default_cpu_present_to_apicid(mps_cpu);
1972 #else
1973 	WARN_ON_ONCE(1);
1974 	return BAD_APICID;
1975 #endif
1976 }
1977 
1978 #define put_smstate(type, buf, offset, val)                      \
1979 	*(type *)((buf) + (offset) - 0x7e00) = val
1980 
1981 #define GET_SMSTATE(type, buf, offset)		\
1982 	(*(type *)((buf) + (offset) - 0x7e00))
1983 
1984 int kvm_cpu_dirty_log_size(void);
1985 
1986 int memslot_rmap_alloc(struct kvm_memory_slot *slot, unsigned long npages);
1987 
1988 #define KVM_CLOCK_VALID_FLAGS						\
1989 	(KVM_CLOCK_TSC_STABLE | KVM_CLOCK_REALTIME | KVM_CLOCK_HOST_TSC)
1990 
1991 #define KVM_X86_VALID_QUIRKS			\
1992 	(KVM_X86_QUIRK_LINT0_REENABLED |	\
1993 	 KVM_X86_QUIRK_CD_NW_CLEARED |		\
1994 	 KVM_X86_QUIRK_LAPIC_MMIO_HOLE |	\
1995 	 KVM_X86_QUIRK_OUT_7E_INC_RIP |		\
1996 	 KVM_X86_QUIRK_MISC_ENABLE_NO_MWAIT |	\
1997 	 KVM_X86_QUIRK_FIX_HYPERCALL_INSN)
1998 
1999 #endif /* _ASM_X86_KVM_HOST_H */
2000