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 #include <linux/kfifo.h>
29 
30 #include <asm/apic.h>
31 #include <asm/pvclock-abi.h>
32 #include <asm/desc.h>
33 #include <asm/mtrr.h>
34 #include <asm/msr-index.h>
35 #include <asm/asm.h>
36 #include <asm/kvm_page_track.h>
37 #include <asm/kvm_vcpu_regs.h>
38 #include <asm/hyperv-tlfs.h>
39 
40 #define __KVM_HAVE_ARCH_VCPU_DEBUGFS
41 
42 #define KVM_MAX_VCPUS 1024
43 
44 /*
45  * In x86, the VCPU ID corresponds to the APIC ID, and APIC IDs
46  * might be larger than the actual number of VCPUs because the
47  * APIC ID encodes CPU topology information.
48  *
49  * In the worst case, we'll need less than one extra bit for the
50  * Core ID, and less than one extra bit for the Package (Die) ID,
51  * so ratio of 4 should be enough.
52  */
53 #define KVM_VCPU_ID_RATIO 4
54 #define KVM_MAX_VCPU_IDS (KVM_MAX_VCPUS * KVM_VCPU_ID_RATIO)
55 
56 /* memory slots that are not exposed to userspace */
57 #define KVM_INTERNAL_MEM_SLOTS 3
58 
59 #define KVM_HALT_POLL_NS_DEFAULT 200000
60 
61 #define KVM_IRQCHIP_NUM_PINS  KVM_IOAPIC_NUM_PINS
62 
63 #define KVM_DIRTY_LOG_MANUAL_CAPS   (KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE | \
64 					KVM_DIRTY_LOG_INITIALLY_SET)
65 
66 #define KVM_BUS_LOCK_DETECTION_VALID_MODE	(KVM_BUS_LOCK_DETECTION_OFF | \
67 						 KVM_BUS_LOCK_DETECTION_EXIT)
68 
69 #define KVM_X86_NOTIFY_VMEXIT_VALID_BITS	(KVM_X86_NOTIFY_VMEXIT_ENABLED | \
70 						 KVM_X86_NOTIFY_VMEXIT_USER)
71 
72 /* x86-specific vcpu->requests bit members */
73 #define KVM_REQ_MIGRATE_TIMER		KVM_ARCH_REQ(0)
74 #define KVM_REQ_REPORT_TPR_ACCESS	KVM_ARCH_REQ(1)
75 #define KVM_REQ_TRIPLE_FAULT		KVM_ARCH_REQ(2)
76 #define KVM_REQ_MMU_SYNC		KVM_ARCH_REQ(3)
77 #define KVM_REQ_CLOCK_UPDATE		KVM_ARCH_REQ(4)
78 #define KVM_REQ_LOAD_MMU_PGD		KVM_ARCH_REQ(5)
79 #define KVM_REQ_EVENT			KVM_ARCH_REQ(6)
80 #define KVM_REQ_APF_HALT		KVM_ARCH_REQ(7)
81 #define KVM_REQ_STEAL_UPDATE		KVM_ARCH_REQ(8)
82 #define KVM_REQ_NMI			KVM_ARCH_REQ(9)
83 #define KVM_REQ_PMU			KVM_ARCH_REQ(10)
84 #define KVM_REQ_PMI			KVM_ARCH_REQ(11)
85 #ifdef CONFIG_KVM_SMM
86 #define KVM_REQ_SMI			KVM_ARCH_REQ(12)
87 #endif
88 #define KVM_REQ_MASTERCLOCK_UPDATE	KVM_ARCH_REQ(13)
89 #define KVM_REQ_MCLOCK_INPROGRESS \
90 	KVM_ARCH_REQ_FLAGS(14, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
91 #define KVM_REQ_SCAN_IOAPIC \
92 	KVM_ARCH_REQ_FLAGS(15, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
93 #define KVM_REQ_GLOBAL_CLOCK_UPDATE	KVM_ARCH_REQ(16)
94 #define KVM_REQ_APIC_PAGE_RELOAD \
95 	KVM_ARCH_REQ_FLAGS(17, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
96 #define KVM_REQ_HV_CRASH		KVM_ARCH_REQ(18)
97 #define KVM_REQ_IOAPIC_EOI_EXIT		KVM_ARCH_REQ(19)
98 #define KVM_REQ_HV_RESET		KVM_ARCH_REQ(20)
99 #define KVM_REQ_HV_EXIT			KVM_ARCH_REQ(21)
100 #define KVM_REQ_HV_STIMER		KVM_ARCH_REQ(22)
101 #define KVM_REQ_LOAD_EOI_EXITMAP	KVM_ARCH_REQ(23)
102 #define KVM_REQ_GET_NESTED_STATE_PAGES	KVM_ARCH_REQ(24)
103 #define KVM_REQ_APICV_UPDATE \
104 	KVM_ARCH_REQ_FLAGS(25, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
105 #define KVM_REQ_TLB_FLUSH_CURRENT	KVM_ARCH_REQ(26)
106 #define KVM_REQ_TLB_FLUSH_GUEST \
107 	KVM_ARCH_REQ_FLAGS(27, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
108 #define KVM_REQ_APF_READY		KVM_ARCH_REQ(28)
109 #define KVM_REQ_MSR_FILTER_CHANGED	KVM_ARCH_REQ(29)
110 #define KVM_REQ_UPDATE_CPU_DIRTY_LOGGING \
111 	KVM_ARCH_REQ_FLAGS(30, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
112 #define KVM_REQ_MMU_FREE_OBSOLETE_ROOTS \
113 	KVM_ARCH_REQ_FLAGS(31, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
114 #define KVM_REQ_HV_TLB_FLUSH \
115 	KVM_ARCH_REQ_FLAGS(32, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
116 
117 #define CR0_RESERVED_BITS                                               \
118 	(~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \
119 			  | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \
120 			  | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG))
121 
122 #define CR4_RESERVED_BITS                                               \
123 	(~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\
124 			  | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE     \
125 			  | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR | X86_CR4_PCIDE \
126 			  | X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_FSGSBASE \
127 			  | X86_CR4_OSXMMEXCPT | X86_CR4_LA57 | X86_CR4_VMXE \
128 			  | X86_CR4_SMAP | X86_CR4_PKE | X86_CR4_UMIP))
129 
130 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR)
131 
132 
133 
134 #define INVALID_PAGE (~(hpa_t)0)
135 #define VALID_PAGE(x) ((x) != INVALID_PAGE)
136 
137 /* KVM Hugepage definitions for x86 */
138 #define KVM_MAX_HUGEPAGE_LEVEL	PG_LEVEL_1G
139 #define KVM_NR_PAGE_SIZES	(KVM_MAX_HUGEPAGE_LEVEL - PG_LEVEL_4K + 1)
140 #define KVM_HPAGE_GFN_SHIFT(x)	(((x) - 1) * 9)
141 #define KVM_HPAGE_SHIFT(x)	(PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
142 #define KVM_HPAGE_SIZE(x)	(1UL << KVM_HPAGE_SHIFT(x))
143 #define KVM_HPAGE_MASK(x)	(~(KVM_HPAGE_SIZE(x) - 1))
144 #define KVM_PAGES_PER_HPAGE(x)	(KVM_HPAGE_SIZE(x) / PAGE_SIZE)
145 
146 #define KVM_MEMSLOT_PAGES_TO_MMU_PAGES_RATIO 50
147 #define KVM_MIN_ALLOC_MMU_PAGES 64UL
148 #define KVM_MMU_HASH_SHIFT 12
149 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
150 #define KVM_MIN_FREE_MMU_PAGES 5
151 #define KVM_REFILL_PAGES 25
152 #define KVM_MAX_CPUID_ENTRIES 256
153 #define KVM_NR_FIXED_MTRR_REGION 88
154 #define KVM_NR_VAR_MTRR 8
155 
156 #define ASYNC_PF_PER_VCPU 64
157 
158 enum kvm_reg {
159 	VCPU_REGS_RAX = __VCPU_REGS_RAX,
160 	VCPU_REGS_RCX = __VCPU_REGS_RCX,
161 	VCPU_REGS_RDX = __VCPU_REGS_RDX,
162 	VCPU_REGS_RBX = __VCPU_REGS_RBX,
163 	VCPU_REGS_RSP = __VCPU_REGS_RSP,
164 	VCPU_REGS_RBP = __VCPU_REGS_RBP,
165 	VCPU_REGS_RSI = __VCPU_REGS_RSI,
166 	VCPU_REGS_RDI = __VCPU_REGS_RDI,
167 #ifdef CONFIG_X86_64
168 	VCPU_REGS_R8  = __VCPU_REGS_R8,
169 	VCPU_REGS_R9  = __VCPU_REGS_R9,
170 	VCPU_REGS_R10 = __VCPU_REGS_R10,
171 	VCPU_REGS_R11 = __VCPU_REGS_R11,
172 	VCPU_REGS_R12 = __VCPU_REGS_R12,
173 	VCPU_REGS_R13 = __VCPU_REGS_R13,
174 	VCPU_REGS_R14 = __VCPU_REGS_R14,
175 	VCPU_REGS_R15 = __VCPU_REGS_R15,
176 #endif
177 	VCPU_REGS_RIP,
178 	NR_VCPU_REGS,
179 
180 	VCPU_EXREG_PDPTR = NR_VCPU_REGS,
181 	VCPU_EXREG_CR0,
182 	VCPU_EXREG_CR3,
183 	VCPU_EXREG_CR4,
184 	VCPU_EXREG_RFLAGS,
185 	VCPU_EXREG_SEGMENTS,
186 	VCPU_EXREG_EXIT_INFO_1,
187 	VCPU_EXREG_EXIT_INFO_2,
188 };
189 
190 enum {
191 	VCPU_SREG_ES,
192 	VCPU_SREG_CS,
193 	VCPU_SREG_SS,
194 	VCPU_SREG_DS,
195 	VCPU_SREG_FS,
196 	VCPU_SREG_GS,
197 	VCPU_SREG_TR,
198 	VCPU_SREG_LDTR,
199 };
200 
201 enum exit_fastpath_completion {
202 	EXIT_FASTPATH_NONE,
203 	EXIT_FASTPATH_REENTER_GUEST,
204 	EXIT_FASTPATH_EXIT_HANDLED,
205 };
206 typedef enum exit_fastpath_completion fastpath_t;
207 
208 struct x86_emulate_ctxt;
209 struct x86_exception;
210 union kvm_smram;
211 enum x86_intercept;
212 enum x86_intercept_stage;
213 
214 #define KVM_NR_DB_REGS	4
215 
216 #define DR6_BUS_LOCK   (1 << 11)
217 #define DR6_BD		(1 << 13)
218 #define DR6_BS		(1 << 14)
219 #define DR6_BT		(1 << 15)
220 #define DR6_RTM		(1 << 16)
221 /*
222  * DR6_ACTIVE_LOW combines fixed-1 and active-low bits.
223  * We can regard all the bits in DR6_FIXED_1 as active_low bits;
224  * they will never be 0 for now, but when they are defined
225  * in the future it will require no code change.
226  *
227  * DR6_ACTIVE_LOW is also used as the init/reset value for DR6.
228  */
229 #define DR6_ACTIVE_LOW	0xffff0ff0
230 #define DR6_VOLATILE	0x0001e80f
231 #define DR6_FIXED_1	(DR6_ACTIVE_LOW & ~DR6_VOLATILE)
232 
233 #define DR7_BP_EN_MASK	0x000000ff
234 #define DR7_GE		(1 << 9)
235 #define DR7_GD		(1 << 13)
236 #define DR7_FIXED_1	0x00000400
237 #define DR7_VOLATILE	0xffff2bff
238 
239 #define KVM_GUESTDBG_VALID_MASK \
240 	(KVM_GUESTDBG_ENABLE | \
241 	KVM_GUESTDBG_SINGLESTEP | \
242 	KVM_GUESTDBG_USE_HW_BP | \
243 	KVM_GUESTDBG_USE_SW_BP | \
244 	KVM_GUESTDBG_INJECT_BP | \
245 	KVM_GUESTDBG_INJECT_DB | \
246 	KVM_GUESTDBG_BLOCKIRQ)
247 
248 
249 #define PFERR_PRESENT_BIT 0
250 #define PFERR_WRITE_BIT 1
251 #define PFERR_USER_BIT 2
252 #define PFERR_RSVD_BIT 3
253 #define PFERR_FETCH_BIT 4
254 #define PFERR_PK_BIT 5
255 #define PFERR_SGX_BIT 15
256 #define PFERR_GUEST_FINAL_BIT 32
257 #define PFERR_GUEST_PAGE_BIT 33
258 #define PFERR_IMPLICIT_ACCESS_BIT 48
259 
260 #define PFERR_PRESENT_MASK	BIT(PFERR_PRESENT_BIT)
261 #define PFERR_WRITE_MASK	BIT(PFERR_WRITE_BIT)
262 #define PFERR_USER_MASK		BIT(PFERR_USER_BIT)
263 #define PFERR_RSVD_MASK		BIT(PFERR_RSVD_BIT)
264 #define PFERR_FETCH_MASK	BIT(PFERR_FETCH_BIT)
265 #define PFERR_PK_MASK		BIT(PFERR_PK_BIT)
266 #define PFERR_SGX_MASK		BIT(PFERR_SGX_BIT)
267 #define PFERR_GUEST_FINAL_MASK	BIT_ULL(PFERR_GUEST_FINAL_BIT)
268 #define PFERR_GUEST_PAGE_MASK	BIT_ULL(PFERR_GUEST_PAGE_BIT)
269 #define PFERR_IMPLICIT_ACCESS	BIT_ULL(PFERR_IMPLICIT_ACCESS_BIT)
270 
271 #define PFERR_NESTED_GUEST_PAGE (PFERR_GUEST_PAGE_MASK |	\
272 				 PFERR_WRITE_MASK |		\
273 				 PFERR_PRESENT_MASK)
274 
275 /* apic attention bits */
276 #define KVM_APIC_CHECK_VAPIC	0
277 /*
278  * The following bit is set with PV-EOI, unset on EOI.
279  * We detect PV-EOI changes by guest by comparing
280  * this bit with PV-EOI in guest memory.
281  * See the implementation in apic_update_pv_eoi.
282  */
283 #define KVM_APIC_PV_EOI_PENDING	1
284 
285 struct kvm_kernel_irq_routing_entry;
286 
287 /*
288  * kvm_mmu_page_role tracks the properties of a shadow page (where shadow page
289  * also includes TDP pages) to determine whether or not a page can be used in
290  * the given MMU context.  This is a subset of the overall kvm_cpu_role to
291  * minimize the size of kvm_memory_slot.arch.gfn_write_track, i.e. allows
292  * allocating 2 bytes per gfn instead of 4 bytes per gfn.
293  *
294  * Upper-level shadow pages having gptes are tracked for write-protection via
295  * gfn_write_track.  As above, gfn_write_track is a 16 bit counter, so KVM must
296  * not create more than 2^16-1 upper-level shadow pages at a single gfn,
297  * otherwise gfn_write_track will overflow and explosions will ensue.
298  *
299  * A unique shadow page (SP) for a gfn is created if and only if an existing SP
300  * cannot be reused.  The ability to reuse a SP is tracked by its role, which
301  * incorporates various mode bits and properties of the SP.  Roughly speaking,
302  * the number of unique SPs that can theoretically be created is 2^n, where n
303  * is the number of bits that are used to compute the role.
304  *
305  * But, even though there are 19 bits in the mask below, not all combinations
306  * of modes and flags are possible:
307  *
308  *   - invalid shadow pages are not accounted, so the bits are effectively 18
309  *
310  *   - quadrant will only be used if has_4_byte_gpte=1 (non-PAE paging);
311  *     execonly and ad_disabled are only used for nested EPT which has
312  *     has_4_byte_gpte=0.  Therefore, 2 bits are always unused.
313  *
314  *   - the 4 bits of level are effectively limited to the values 2/3/4/5,
315  *     as 4k SPs are not tracked (allowed to go unsync).  In addition non-PAE
316  *     paging has exactly one upper level, making level completely redundant
317  *     when has_4_byte_gpte=1.
318  *
319  *   - on top of this, smep_andnot_wp and smap_andnot_wp are only set if
320  *     cr0_wp=0, therefore these three bits only give rise to 5 possibilities.
321  *
322  * Therefore, the maximum number of possible upper-level shadow pages for a
323  * single gfn is a bit less than 2^13.
324  */
325 union kvm_mmu_page_role {
326 	u32 word;
327 	struct {
328 		unsigned level:4;
329 		unsigned has_4_byte_gpte:1;
330 		unsigned quadrant:2;
331 		unsigned direct:1;
332 		unsigned access:3;
333 		unsigned invalid:1;
334 		unsigned efer_nx:1;
335 		unsigned cr0_wp:1;
336 		unsigned smep_andnot_wp:1;
337 		unsigned smap_andnot_wp:1;
338 		unsigned ad_disabled:1;
339 		unsigned guest_mode:1;
340 		unsigned passthrough:1;
341 		unsigned :5;
342 
343 		/*
344 		 * This is left at the top of the word so that
345 		 * kvm_memslots_for_spte_role can extract it with a
346 		 * simple shift.  While there is room, give it a whole
347 		 * byte so it is also faster to load it from memory.
348 		 */
349 		unsigned smm:8;
350 	};
351 };
352 
353 /*
354  * kvm_mmu_extended_role complements kvm_mmu_page_role, tracking properties
355  * relevant to the current MMU configuration.   When loading CR0, CR4, or EFER,
356  * including on nested transitions, if nothing in the full role changes then
357  * MMU re-configuration can be skipped. @valid bit is set on first usage so we
358  * don't treat all-zero structure as valid data.
359  *
360  * The properties that are tracked in the extended role but not the page role
361  * are for things that either (a) do not affect the validity of the shadow page
362  * or (b) are indirectly reflected in the shadow page's role.  For example,
363  * CR4.PKE only affects permission checks for software walks of the guest page
364  * tables (because KVM doesn't support Protection Keys with shadow paging), and
365  * CR0.PG, CR4.PAE, and CR4.PSE are indirectly reflected in role.level.
366  *
367  * Note, SMEP and SMAP are not redundant with sm*p_andnot_wp in the page role.
368  * If CR0.WP=1, KVM can reuse shadow pages for the guest regardless of SMEP and
369  * SMAP, but the MMU's permission checks for software walks need to be SMEP and
370  * SMAP aware regardless of CR0.WP.
371  */
372 union kvm_mmu_extended_role {
373 	u32 word;
374 	struct {
375 		unsigned int valid:1;
376 		unsigned int execonly:1;
377 		unsigned int cr4_pse:1;
378 		unsigned int cr4_pke:1;
379 		unsigned int cr4_smap:1;
380 		unsigned int cr4_smep:1;
381 		unsigned int cr4_la57:1;
382 		unsigned int efer_lma:1;
383 	};
384 };
385 
386 union kvm_cpu_role {
387 	u64 as_u64;
388 	struct {
389 		union kvm_mmu_page_role base;
390 		union kvm_mmu_extended_role ext;
391 	};
392 };
393 
394 struct kvm_rmap_head {
395 	unsigned long val;
396 };
397 
398 struct kvm_pio_request {
399 	unsigned long linear_rip;
400 	unsigned long count;
401 	int in;
402 	int port;
403 	int size;
404 };
405 
406 #define PT64_ROOT_MAX_LEVEL 5
407 
408 struct rsvd_bits_validate {
409 	u64 rsvd_bits_mask[2][PT64_ROOT_MAX_LEVEL];
410 	u64 bad_mt_xwr;
411 };
412 
413 struct kvm_mmu_root_info {
414 	gpa_t pgd;
415 	hpa_t hpa;
416 };
417 
418 #define KVM_MMU_ROOT_INFO_INVALID \
419 	((struct kvm_mmu_root_info) { .pgd = INVALID_PAGE, .hpa = INVALID_PAGE })
420 
421 #define KVM_MMU_NUM_PREV_ROOTS 3
422 
423 #define KVM_MMU_ROOT_CURRENT		BIT(0)
424 #define KVM_MMU_ROOT_PREVIOUS(i)	BIT(1+i)
425 #define KVM_MMU_ROOTS_ALL		(BIT(1 + KVM_MMU_NUM_PREV_ROOTS) - 1)
426 
427 #define KVM_HAVE_MMU_RWLOCK
428 
429 struct kvm_mmu_page;
430 struct kvm_page_fault;
431 
432 /*
433  * x86 supports 4 paging modes (5-level 64-bit, 4-level 64-bit, 3-level 32-bit,
434  * and 2-level 32-bit).  The kvm_mmu structure abstracts the details of the
435  * current mmu mode.
436  */
437 struct kvm_mmu {
438 	unsigned long (*get_guest_pgd)(struct kvm_vcpu *vcpu);
439 	u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
440 	int (*page_fault)(struct kvm_vcpu *vcpu, struct kvm_page_fault *fault);
441 	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
442 				  struct x86_exception *fault);
443 	gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
444 			    gpa_t gva_or_gpa, u64 access,
445 			    struct x86_exception *exception);
446 	int (*sync_spte)(struct kvm_vcpu *vcpu,
447 			 struct kvm_mmu_page *sp, int i);
448 	struct kvm_mmu_root_info root;
449 	union kvm_cpu_role cpu_role;
450 	union kvm_mmu_page_role root_role;
451 
452 	/*
453 	* The pkru_mask indicates if protection key checks are needed.  It
454 	* consists of 16 domains indexed by page fault error code bits [4:1],
455 	* with PFEC.RSVD replaced by ACC_USER_MASK from the page tables.
456 	* Each domain has 2 bits which are ANDed with AD and WD from PKRU.
457 	*/
458 	u32 pkru_mask;
459 
460 	struct kvm_mmu_root_info prev_roots[KVM_MMU_NUM_PREV_ROOTS];
461 
462 	/*
463 	 * Bitmap; bit set = permission fault
464 	 * Byte index: page fault error code [4:1]
465 	 * Bit index: pte permissions in ACC_* format
466 	 */
467 	u8 permissions[16];
468 
469 	u64 *pae_root;
470 	u64 *pml4_root;
471 	u64 *pml5_root;
472 
473 	/*
474 	 * check zero bits on shadow page table entries, these
475 	 * bits include not only hardware reserved bits but also
476 	 * the bits spte never used.
477 	 */
478 	struct rsvd_bits_validate shadow_zero_check;
479 
480 	struct rsvd_bits_validate guest_rsvd_check;
481 
482 	u64 pdptrs[4]; /* pae */
483 };
484 
485 enum pmc_type {
486 	KVM_PMC_GP = 0,
487 	KVM_PMC_FIXED,
488 };
489 
490 struct kvm_pmc {
491 	enum pmc_type type;
492 	u8 idx;
493 	bool is_paused;
494 	bool intr;
495 	u64 counter;
496 	u64 prev_counter;
497 	u64 eventsel;
498 	struct perf_event *perf_event;
499 	struct kvm_vcpu *vcpu;
500 	/*
501 	 * only for creating or reusing perf_event,
502 	 * eventsel value for general purpose counters,
503 	 * ctrl value for fixed counters.
504 	 */
505 	u64 current_config;
506 };
507 
508 /* More counters may conflict with other existing Architectural MSRs */
509 #define KVM_INTEL_PMC_MAX_GENERIC	8
510 #define MSR_ARCH_PERFMON_PERFCTR_MAX	(MSR_ARCH_PERFMON_PERFCTR0 + KVM_INTEL_PMC_MAX_GENERIC - 1)
511 #define MSR_ARCH_PERFMON_EVENTSEL_MAX	(MSR_ARCH_PERFMON_EVENTSEL0 + KVM_INTEL_PMC_MAX_GENERIC - 1)
512 #define KVM_PMC_MAX_FIXED	3
513 #define MSR_ARCH_PERFMON_FIXED_CTR_MAX	(MSR_ARCH_PERFMON_FIXED_CTR0 + KVM_PMC_MAX_FIXED - 1)
514 #define KVM_AMD_PMC_MAX_GENERIC	6
515 struct kvm_pmu {
516 	u8 version;
517 	unsigned nr_arch_gp_counters;
518 	unsigned nr_arch_fixed_counters;
519 	unsigned available_event_types;
520 	u64 fixed_ctr_ctrl;
521 	u64 fixed_ctr_ctrl_mask;
522 	u64 global_ctrl;
523 	u64 global_status;
524 	u64 counter_bitmask[2];
525 	u64 global_ctrl_mask;
526 	u64 global_status_mask;
527 	u64 reserved_bits;
528 	u64 raw_event_mask;
529 	struct kvm_pmc gp_counters[KVM_INTEL_PMC_MAX_GENERIC];
530 	struct kvm_pmc fixed_counters[KVM_PMC_MAX_FIXED];
531 
532 	/*
533 	 * Overlay the bitmap with a 64-bit atomic so that all bits can be
534 	 * set in a single access, e.g. to reprogram all counters when the PMU
535 	 * filter changes.
536 	 */
537 	union {
538 		DECLARE_BITMAP(reprogram_pmi, X86_PMC_IDX_MAX);
539 		atomic64_t __reprogram_pmi;
540 	};
541 	DECLARE_BITMAP(all_valid_pmc_idx, X86_PMC_IDX_MAX);
542 	DECLARE_BITMAP(pmc_in_use, X86_PMC_IDX_MAX);
543 
544 	u64 ds_area;
545 	u64 pebs_enable;
546 	u64 pebs_enable_mask;
547 	u64 pebs_data_cfg;
548 	u64 pebs_data_cfg_mask;
549 
550 	/*
551 	 * If a guest counter is cross-mapped to host counter with different
552 	 * index, its PEBS capability will be temporarily disabled.
553 	 *
554 	 * The user should make sure that this mask is updated
555 	 * after disabling interrupts and before perf_guest_get_msrs();
556 	 */
557 	u64 host_cross_mapped_mask;
558 
559 	/*
560 	 * The gate to release perf_events not marked in
561 	 * pmc_in_use only once in a vcpu time slice.
562 	 */
563 	bool need_cleanup;
564 
565 	/*
566 	 * The total number of programmed perf_events and it helps to avoid
567 	 * redundant check before cleanup if guest don't use vPMU at all.
568 	 */
569 	u8 event_count;
570 };
571 
572 struct kvm_pmu_ops;
573 
574 enum {
575 	KVM_DEBUGREG_BP_ENABLED = 1,
576 	KVM_DEBUGREG_WONT_EXIT = 2,
577 };
578 
579 struct kvm_mtrr_range {
580 	u64 base;
581 	u64 mask;
582 	struct list_head node;
583 };
584 
585 struct kvm_mtrr {
586 	struct kvm_mtrr_range var_ranges[KVM_NR_VAR_MTRR];
587 	mtrr_type fixed_ranges[KVM_NR_FIXED_MTRR_REGION];
588 	u64 deftype;
589 
590 	struct list_head head;
591 };
592 
593 /* Hyper-V SynIC timer */
594 struct kvm_vcpu_hv_stimer {
595 	struct hrtimer timer;
596 	int index;
597 	union hv_stimer_config config;
598 	u64 count;
599 	u64 exp_time;
600 	struct hv_message msg;
601 	bool msg_pending;
602 };
603 
604 /* Hyper-V synthetic interrupt controller (SynIC)*/
605 struct kvm_vcpu_hv_synic {
606 	u64 version;
607 	u64 control;
608 	u64 msg_page;
609 	u64 evt_page;
610 	atomic64_t sint[HV_SYNIC_SINT_COUNT];
611 	atomic_t sint_to_gsi[HV_SYNIC_SINT_COUNT];
612 	DECLARE_BITMAP(auto_eoi_bitmap, 256);
613 	DECLARE_BITMAP(vec_bitmap, 256);
614 	bool active;
615 	bool dont_zero_synic_pages;
616 };
617 
618 /* The maximum number of entries on the TLB flush fifo. */
619 #define KVM_HV_TLB_FLUSH_FIFO_SIZE (16)
620 /*
621  * Note: the following 'magic' entry is made up by KVM to avoid putting
622  * anything besides GVA on the TLB flush fifo. It is theoretically possible
623  * to observe a request to flush 4095 PFNs starting from 0xfffffffffffff000
624  * which will look identical. KVM's action to 'flush everything' instead of
625  * flushing these particular addresses is, however, fully legitimate as
626  * flushing more than requested is always OK.
627  */
628 #define KVM_HV_TLB_FLUSHALL_ENTRY  ((u64)-1)
629 
630 enum hv_tlb_flush_fifos {
631 	HV_L1_TLB_FLUSH_FIFO,
632 	HV_L2_TLB_FLUSH_FIFO,
633 	HV_NR_TLB_FLUSH_FIFOS,
634 };
635 
636 struct kvm_vcpu_hv_tlb_flush_fifo {
637 	spinlock_t write_lock;
638 	DECLARE_KFIFO(entries, u64, KVM_HV_TLB_FLUSH_FIFO_SIZE);
639 };
640 
641 /* Hyper-V per vcpu emulation context */
642 struct kvm_vcpu_hv {
643 	struct kvm_vcpu *vcpu;
644 	u32 vp_index;
645 	u64 hv_vapic;
646 	s64 runtime_offset;
647 	struct kvm_vcpu_hv_synic synic;
648 	struct kvm_hyperv_exit exit;
649 	struct kvm_vcpu_hv_stimer stimer[HV_SYNIC_STIMER_COUNT];
650 	DECLARE_BITMAP(stimer_pending_bitmap, HV_SYNIC_STIMER_COUNT);
651 	bool enforce_cpuid;
652 	struct {
653 		u32 features_eax; /* HYPERV_CPUID_FEATURES.EAX */
654 		u32 features_ebx; /* HYPERV_CPUID_FEATURES.EBX */
655 		u32 features_edx; /* HYPERV_CPUID_FEATURES.EDX */
656 		u32 enlightenments_eax; /* HYPERV_CPUID_ENLIGHTMENT_INFO.EAX */
657 		u32 enlightenments_ebx; /* HYPERV_CPUID_ENLIGHTMENT_INFO.EBX */
658 		u32 syndbg_cap_eax; /* HYPERV_CPUID_SYNDBG_PLATFORM_CAPABILITIES.EAX */
659 		u32 nested_eax; /* HYPERV_CPUID_NESTED_FEATURES.EAX */
660 		u32 nested_ebx; /* HYPERV_CPUID_NESTED_FEATURES.EBX */
661 	} cpuid_cache;
662 
663 	struct kvm_vcpu_hv_tlb_flush_fifo tlb_flush_fifo[HV_NR_TLB_FLUSH_FIFOS];
664 
665 	/* Preallocated buffer for handling hypercalls passing sparse vCPU set */
666 	u64 sparse_banks[HV_MAX_SPARSE_VCPU_BANKS];
667 
668 	struct hv_vp_assist_page vp_assist_page;
669 
670 	struct {
671 		u64 pa_page_gpa;
672 		u64 vm_id;
673 		u32 vp_id;
674 	} nested;
675 };
676 
677 struct kvm_hypervisor_cpuid {
678 	u32 base;
679 	u32 limit;
680 };
681 
682 /* Xen HVM per vcpu emulation context */
683 struct kvm_vcpu_xen {
684 	u64 hypercall_rip;
685 	u32 current_runstate;
686 	u8 upcall_vector;
687 	struct gfn_to_pfn_cache vcpu_info_cache;
688 	struct gfn_to_pfn_cache vcpu_time_info_cache;
689 	struct gfn_to_pfn_cache runstate_cache;
690 	struct gfn_to_pfn_cache runstate2_cache;
691 	u64 last_steal;
692 	u64 runstate_entry_time;
693 	u64 runstate_times[4];
694 	unsigned long evtchn_pending_sel;
695 	u32 vcpu_id; /* The Xen / ACPI vCPU ID */
696 	u32 timer_virq;
697 	u64 timer_expires; /* In guest epoch */
698 	atomic_t timer_pending;
699 	struct hrtimer timer;
700 	int poll_evtchn;
701 	struct timer_list poll_timer;
702 	struct kvm_hypervisor_cpuid cpuid;
703 };
704 
705 struct kvm_queued_exception {
706 	bool pending;
707 	bool injected;
708 	bool has_error_code;
709 	u8 vector;
710 	u32 error_code;
711 	unsigned long payload;
712 	bool has_payload;
713 };
714 
715 struct kvm_vcpu_arch {
716 	/*
717 	 * rip and regs accesses must go through
718 	 * kvm_{register,rip}_{read,write} functions.
719 	 */
720 	unsigned long regs[NR_VCPU_REGS];
721 	u32 regs_avail;
722 	u32 regs_dirty;
723 
724 	unsigned long cr0;
725 	unsigned long cr0_guest_owned_bits;
726 	unsigned long cr2;
727 	unsigned long cr3;
728 	unsigned long cr4;
729 	unsigned long cr4_guest_owned_bits;
730 	unsigned long cr4_guest_rsvd_bits;
731 	unsigned long cr8;
732 	u32 host_pkru;
733 	u32 pkru;
734 	u32 hflags;
735 	u64 efer;
736 	u64 apic_base;
737 	struct kvm_lapic *apic;    /* kernel irqchip context */
738 	bool load_eoi_exitmap_pending;
739 	DECLARE_BITMAP(ioapic_handled_vectors, 256);
740 	unsigned long apic_attention;
741 	int32_t apic_arb_prio;
742 	int mp_state;
743 	u64 ia32_misc_enable_msr;
744 	u64 smbase;
745 	u64 smi_count;
746 	bool at_instruction_boundary;
747 	bool tpr_access_reporting;
748 	bool xfd_no_write_intercept;
749 	u64 ia32_xss;
750 	u64 microcode_version;
751 	u64 arch_capabilities;
752 	u64 perf_capabilities;
753 
754 	/*
755 	 * Paging state of the vcpu
756 	 *
757 	 * If the vcpu runs in guest mode with two level paging this still saves
758 	 * the paging mode of the l1 guest. This context is always used to
759 	 * handle faults.
760 	 */
761 	struct kvm_mmu *mmu;
762 
763 	/* Non-nested MMU for L1 */
764 	struct kvm_mmu root_mmu;
765 
766 	/* L1 MMU when running nested */
767 	struct kvm_mmu guest_mmu;
768 
769 	/*
770 	 * Paging state of an L2 guest (used for nested npt)
771 	 *
772 	 * This context will save all necessary information to walk page tables
773 	 * of an L2 guest. This context is only initialized for page table
774 	 * walking and not for faulting since we never handle l2 page faults on
775 	 * the host.
776 	 */
777 	struct kvm_mmu nested_mmu;
778 
779 	/*
780 	 * Pointer to the mmu context currently used for
781 	 * gva_to_gpa translations.
782 	 */
783 	struct kvm_mmu *walk_mmu;
784 
785 	struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
786 	struct kvm_mmu_memory_cache mmu_shadow_page_cache;
787 	struct kvm_mmu_memory_cache mmu_shadowed_info_cache;
788 	struct kvm_mmu_memory_cache mmu_page_header_cache;
789 
790 	/*
791 	 * QEMU userspace and the guest each have their own FPU state.
792 	 * In vcpu_run, we switch between the user and guest FPU contexts.
793 	 * While running a VCPU, the VCPU thread will have the guest FPU
794 	 * context.
795 	 *
796 	 * Note that while the PKRU state lives inside the fpu registers,
797 	 * it is switched out separately at VMENTER and VMEXIT time. The
798 	 * "guest_fpstate" state here contains the guest FPU context, with the
799 	 * host PRKU bits.
800 	 */
801 	struct fpu_guest guest_fpu;
802 
803 	u64 xcr0;
804 	u64 guest_supported_xcr0;
805 
806 	struct kvm_pio_request pio;
807 	void *pio_data;
808 	void *sev_pio_data;
809 	unsigned sev_pio_count;
810 
811 	u8 event_exit_inst_len;
812 
813 	bool exception_from_userspace;
814 
815 	/* Exceptions to be injected to the guest. */
816 	struct kvm_queued_exception exception;
817 	/* Exception VM-Exits to be synthesized to L1. */
818 	struct kvm_queued_exception exception_vmexit;
819 
820 	struct kvm_queued_interrupt {
821 		bool injected;
822 		bool soft;
823 		u8 nr;
824 	} interrupt;
825 
826 	int halt_request; /* real mode on Intel only */
827 
828 	int cpuid_nent;
829 	struct kvm_cpuid_entry2 *cpuid_entries;
830 	struct kvm_hypervisor_cpuid kvm_cpuid;
831 
832 	/*
833 	 * FIXME: Drop this macro and use KVM_NR_GOVERNED_FEATURES directly
834 	 * when "struct kvm_vcpu_arch" is no longer defined in an
835 	 * arch/x86/include/asm header.  The max is mostly arbitrary, i.e.
836 	 * can be increased as necessary.
837 	 */
838 #define KVM_MAX_NR_GOVERNED_FEATURES BITS_PER_LONG
839 
840 	/*
841 	 * Track whether or not the guest is allowed to use features that are
842 	 * governed by KVM, where "governed" means KVM needs to manage state
843 	 * and/or explicitly enable the feature in hardware.  Typically, but
844 	 * not always, governed features can be used by the guest if and only
845 	 * if both KVM and userspace want to expose the feature to the guest.
846 	 */
847 	struct {
848 		DECLARE_BITMAP(enabled, KVM_MAX_NR_GOVERNED_FEATURES);
849 	} governed_features;
850 
851 	u64 reserved_gpa_bits;
852 	int maxphyaddr;
853 
854 	/* emulate context */
855 
856 	struct x86_emulate_ctxt *emulate_ctxt;
857 	bool emulate_regs_need_sync_to_vcpu;
858 	bool emulate_regs_need_sync_from_vcpu;
859 	int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
860 
861 	gpa_t time;
862 	struct pvclock_vcpu_time_info hv_clock;
863 	unsigned int hw_tsc_khz;
864 	struct gfn_to_pfn_cache pv_time;
865 	/* set guest stopped flag in pvclock flags field */
866 	bool pvclock_set_guest_stopped_request;
867 
868 	struct {
869 		u8 preempted;
870 		u64 msr_val;
871 		u64 last_steal;
872 		struct gfn_to_hva_cache cache;
873 	} st;
874 
875 	u64 l1_tsc_offset;
876 	u64 tsc_offset; /* current tsc offset */
877 	u64 last_guest_tsc;
878 	u64 last_host_tsc;
879 	u64 tsc_offset_adjustment;
880 	u64 this_tsc_nsec;
881 	u64 this_tsc_write;
882 	u64 this_tsc_generation;
883 	bool tsc_catchup;
884 	bool tsc_always_catchup;
885 	s8 virtual_tsc_shift;
886 	u32 virtual_tsc_mult;
887 	u32 virtual_tsc_khz;
888 	s64 ia32_tsc_adjust_msr;
889 	u64 msr_ia32_power_ctl;
890 	u64 l1_tsc_scaling_ratio;
891 	u64 tsc_scaling_ratio; /* current scaling ratio */
892 
893 	atomic_t nmi_queued;  /* unprocessed asynchronous NMIs */
894 	/* Number of NMIs pending injection, not including hardware vNMIs. */
895 	unsigned int nmi_pending;
896 	bool nmi_injected;    /* Trying to inject an NMI this entry */
897 	bool smi_pending;    /* SMI queued after currently running handler */
898 	u8 handling_intr_from_guest;
899 
900 	struct kvm_mtrr mtrr_state;
901 	u64 pat;
902 
903 	unsigned switch_db_regs;
904 	unsigned long db[KVM_NR_DB_REGS];
905 	unsigned long dr6;
906 	unsigned long dr7;
907 	unsigned long eff_db[KVM_NR_DB_REGS];
908 	unsigned long guest_debug_dr7;
909 	u64 msr_platform_info;
910 	u64 msr_misc_features_enables;
911 
912 	u64 mcg_cap;
913 	u64 mcg_status;
914 	u64 mcg_ctl;
915 	u64 mcg_ext_ctl;
916 	u64 *mce_banks;
917 	u64 *mci_ctl2_banks;
918 
919 	/* Cache MMIO info */
920 	u64 mmio_gva;
921 	unsigned mmio_access;
922 	gfn_t mmio_gfn;
923 	u64 mmio_gen;
924 
925 	struct kvm_pmu pmu;
926 
927 	/* used for guest single stepping over the given code position */
928 	unsigned long singlestep_rip;
929 
930 	bool hyperv_enabled;
931 	struct kvm_vcpu_hv *hyperv;
932 	struct kvm_vcpu_xen xen;
933 
934 	cpumask_var_t wbinvd_dirty_mask;
935 
936 	unsigned long last_retry_eip;
937 	unsigned long last_retry_addr;
938 
939 	struct {
940 		bool halted;
941 		gfn_t gfns[ASYNC_PF_PER_VCPU];
942 		struct gfn_to_hva_cache data;
943 		u64 msr_en_val; /* MSR_KVM_ASYNC_PF_EN */
944 		u64 msr_int_val; /* MSR_KVM_ASYNC_PF_INT */
945 		u16 vec;
946 		u32 id;
947 		bool send_user_only;
948 		u32 host_apf_flags;
949 		bool delivery_as_pf_vmexit;
950 		bool pageready_pending;
951 	} apf;
952 
953 	/* OSVW MSRs (AMD only) */
954 	struct {
955 		u64 length;
956 		u64 status;
957 	} osvw;
958 
959 	struct {
960 		u64 msr_val;
961 		struct gfn_to_hva_cache data;
962 	} pv_eoi;
963 
964 	u64 msr_kvm_poll_control;
965 
966 	/* set at EPT violation at this point */
967 	unsigned long exit_qualification;
968 
969 	/* pv related host specific info */
970 	struct {
971 		bool pv_unhalted;
972 	} pv;
973 
974 	int pending_ioapic_eoi;
975 	int pending_external_vector;
976 
977 	/* be preempted when it's in kernel-mode(cpl=0) */
978 	bool preempted_in_kernel;
979 
980 	/* Flush the L1 Data cache for L1TF mitigation on VMENTER */
981 	bool l1tf_flush_l1d;
982 
983 	/* Host CPU on which VM-entry was most recently attempted */
984 	int last_vmentry_cpu;
985 
986 	/* AMD MSRC001_0015 Hardware Configuration */
987 	u64 msr_hwcr;
988 
989 	/* pv related cpuid info */
990 	struct {
991 		/*
992 		 * value of the eax register in the KVM_CPUID_FEATURES CPUID
993 		 * leaf.
994 		 */
995 		u32 features;
996 
997 		/*
998 		 * indicates whether pv emulation should be disabled if features
999 		 * are not present in the guest's cpuid
1000 		 */
1001 		bool enforce;
1002 	} pv_cpuid;
1003 
1004 	/* Protected Guests */
1005 	bool guest_state_protected;
1006 
1007 	/*
1008 	 * Set when PDPTS were loaded directly by the userspace without
1009 	 * reading the guest memory
1010 	 */
1011 	bool pdptrs_from_userspace;
1012 
1013 #if IS_ENABLED(CONFIG_HYPERV)
1014 	hpa_t hv_root_tdp;
1015 #endif
1016 };
1017 
1018 struct kvm_lpage_info {
1019 	int disallow_lpage;
1020 };
1021 
1022 struct kvm_arch_memory_slot {
1023 	struct kvm_rmap_head *rmap[KVM_NR_PAGE_SIZES];
1024 	struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
1025 	unsigned short *gfn_write_track;
1026 };
1027 
1028 /*
1029  * Track the mode of the optimized logical map, as the rules for decoding the
1030  * destination vary per mode.  Enabling the optimized logical map requires all
1031  * software-enabled local APIs to be in the same mode, each addressable APIC to
1032  * be mapped to only one MDA, and each MDA to map to at most one APIC.
1033  */
1034 enum kvm_apic_logical_mode {
1035 	/* All local APICs are software disabled. */
1036 	KVM_APIC_MODE_SW_DISABLED,
1037 	/* All software enabled local APICs in xAPIC cluster addressing mode. */
1038 	KVM_APIC_MODE_XAPIC_CLUSTER,
1039 	/* All software enabled local APICs in xAPIC flat addressing mode. */
1040 	KVM_APIC_MODE_XAPIC_FLAT,
1041 	/* All software enabled local APICs in x2APIC mode. */
1042 	KVM_APIC_MODE_X2APIC,
1043 	/*
1044 	 * Optimized map disabled, e.g. not all local APICs in the same logical
1045 	 * mode, same logical ID assigned to multiple APICs, etc.
1046 	 */
1047 	KVM_APIC_MODE_MAP_DISABLED,
1048 };
1049 
1050 struct kvm_apic_map {
1051 	struct rcu_head rcu;
1052 	enum kvm_apic_logical_mode logical_mode;
1053 	u32 max_apic_id;
1054 	union {
1055 		struct kvm_lapic *xapic_flat_map[8];
1056 		struct kvm_lapic *xapic_cluster_map[16][4];
1057 	};
1058 	struct kvm_lapic *phys_map[];
1059 };
1060 
1061 /* Hyper-V synthetic debugger (SynDbg)*/
1062 struct kvm_hv_syndbg {
1063 	struct {
1064 		u64 control;
1065 		u64 status;
1066 		u64 send_page;
1067 		u64 recv_page;
1068 		u64 pending_page;
1069 	} control;
1070 	u64 options;
1071 };
1072 
1073 /* Current state of Hyper-V TSC page clocksource */
1074 enum hv_tsc_page_status {
1075 	/* TSC page was not set up or disabled */
1076 	HV_TSC_PAGE_UNSET = 0,
1077 	/* TSC page MSR was written by the guest, update pending */
1078 	HV_TSC_PAGE_GUEST_CHANGED,
1079 	/* TSC page update was triggered from the host side */
1080 	HV_TSC_PAGE_HOST_CHANGED,
1081 	/* TSC page was properly set up and is currently active  */
1082 	HV_TSC_PAGE_SET,
1083 	/* TSC page was set up with an inaccessible GPA */
1084 	HV_TSC_PAGE_BROKEN,
1085 };
1086 
1087 /* Hyper-V emulation context */
1088 struct kvm_hv {
1089 	struct mutex hv_lock;
1090 	u64 hv_guest_os_id;
1091 	u64 hv_hypercall;
1092 	u64 hv_tsc_page;
1093 	enum hv_tsc_page_status hv_tsc_page_status;
1094 
1095 	/* Hyper-v based guest crash (NT kernel bugcheck) parameters */
1096 	u64 hv_crash_param[HV_X64_MSR_CRASH_PARAMS];
1097 	u64 hv_crash_ctl;
1098 
1099 	struct ms_hyperv_tsc_page tsc_ref;
1100 
1101 	struct idr conn_to_evt;
1102 
1103 	u64 hv_reenlightenment_control;
1104 	u64 hv_tsc_emulation_control;
1105 	u64 hv_tsc_emulation_status;
1106 	u64 hv_invtsc_control;
1107 
1108 	/* How many vCPUs have VP index != vCPU index */
1109 	atomic_t num_mismatched_vp_indexes;
1110 
1111 	/*
1112 	 * How many SynICs use 'AutoEOI' feature
1113 	 * (protected by arch.apicv_update_lock)
1114 	 */
1115 	unsigned int synic_auto_eoi_used;
1116 
1117 	struct hv_partition_assist_pg *hv_pa_pg;
1118 	struct kvm_hv_syndbg hv_syndbg;
1119 };
1120 
1121 struct msr_bitmap_range {
1122 	u32 flags;
1123 	u32 nmsrs;
1124 	u32 base;
1125 	unsigned long *bitmap;
1126 };
1127 
1128 /* Xen emulation context */
1129 struct kvm_xen {
1130 	struct mutex xen_lock;
1131 	u32 xen_version;
1132 	bool long_mode;
1133 	bool runstate_update_flag;
1134 	u8 upcall_vector;
1135 	struct gfn_to_pfn_cache shinfo_cache;
1136 	struct idr evtchn_ports;
1137 	unsigned long poll_mask[BITS_TO_LONGS(KVM_MAX_VCPUS)];
1138 };
1139 
1140 enum kvm_irqchip_mode {
1141 	KVM_IRQCHIP_NONE,
1142 	KVM_IRQCHIP_KERNEL,       /* created with KVM_CREATE_IRQCHIP */
1143 	KVM_IRQCHIP_SPLIT,        /* created with KVM_CAP_SPLIT_IRQCHIP */
1144 };
1145 
1146 struct kvm_x86_msr_filter {
1147 	u8 count;
1148 	bool default_allow:1;
1149 	struct msr_bitmap_range ranges[16];
1150 };
1151 
1152 struct kvm_x86_pmu_event_filter {
1153 	__u32 action;
1154 	__u32 nevents;
1155 	__u32 fixed_counter_bitmap;
1156 	__u32 flags;
1157 	__u32 nr_includes;
1158 	__u32 nr_excludes;
1159 	__u64 *includes;
1160 	__u64 *excludes;
1161 	__u64 events[];
1162 };
1163 
1164 enum kvm_apicv_inhibit {
1165 
1166 	/********************************************************************/
1167 	/* INHIBITs that are relevant to both Intel's APICv and AMD's AVIC. */
1168 	/********************************************************************/
1169 
1170 	/*
1171 	 * APIC acceleration is disabled by a module parameter
1172 	 * and/or not supported in hardware.
1173 	 */
1174 	APICV_INHIBIT_REASON_DISABLE,
1175 
1176 	/*
1177 	 * APIC acceleration is inhibited because AutoEOI feature is
1178 	 * being used by a HyperV guest.
1179 	 */
1180 	APICV_INHIBIT_REASON_HYPERV,
1181 
1182 	/*
1183 	 * APIC acceleration is inhibited because the userspace didn't yet
1184 	 * enable the kernel/split irqchip.
1185 	 */
1186 	APICV_INHIBIT_REASON_ABSENT,
1187 
1188 	/* APIC acceleration is inhibited because KVM_GUESTDBG_BLOCKIRQ
1189 	 * (out of band, debug measure of blocking all interrupts on this vCPU)
1190 	 * was enabled, to avoid AVIC/APICv bypassing it.
1191 	 */
1192 	APICV_INHIBIT_REASON_BLOCKIRQ,
1193 
1194 	/*
1195 	 * APICv is disabled because not all vCPUs have a 1:1 mapping between
1196 	 * APIC ID and vCPU, _and_ KVM is not applying its x2APIC hotplug hack.
1197 	 */
1198 	APICV_INHIBIT_REASON_PHYSICAL_ID_ALIASED,
1199 
1200 	/*
1201 	 * For simplicity, the APIC acceleration is inhibited
1202 	 * first time either APIC ID or APIC base are changed by the guest
1203 	 * from their reset values.
1204 	 */
1205 	APICV_INHIBIT_REASON_APIC_ID_MODIFIED,
1206 	APICV_INHIBIT_REASON_APIC_BASE_MODIFIED,
1207 
1208 	/******************************************************/
1209 	/* INHIBITs that are relevant only to the AMD's AVIC. */
1210 	/******************************************************/
1211 
1212 	/*
1213 	 * AVIC is inhibited on a vCPU because it runs a nested guest.
1214 	 *
1215 	 * This is needed because unlike APICv, the peers of this vCPU
1216 	 * cannot use the doorbell mechanism to signal interrupts via AVIC when
1217 	 * a vCPU runs nested.
1218 	 */
1219 	APICV_INHIBIT_REASON_NESTED,
1220 
1221 	/*
1222 	 * On SVM, the wait for the IRQ window is implemented with pending vIRQ,
1223 	 * which cannot be injected when the AVIC is enabled, thus AVIC
1224 	 * is inhibited while KVM waits for IRQ window.
1225 	 */
1226 	APICV_INHIBIT_REASON_IRQWIN,
1227 
1228 	/*
1229 	 * PIT (i8254) 're-inject' mode, relies on EOI intercept,
1230 	 * which AVIC doesn't support for edge triggered interrupts.
1231 	 */
1232 	APICV_INHIBIT_REASON_PIT_REINJ,
1233 
1234 	/*
1235 	 * AVIC is disabled because SEV doesn't support it.
1236 	 */
1237 	APICV_INHIBIT_REASON_SEV,
1238 
1239 	/*
1240 	 * AVIC is disabled because not all vCPUs with a valid LDR have a 1:1
1241 	 * mapping between logical ID and vCPU.
1242 	 */
1243 	APICV_INHIBIT_REASON_LOGICAL_ID_ALIASED,
1244 };
1245 
1246 struct kvm_arch {
1247 	unsigned long n_used_mmu_pages;
1248 	unsigned long n_requested_mmu_pages;
1249 	unsigned long n_max_mmu_pages;
1250 	unsigned int indirect_shadow_pages;
1251 	u8 mmu_valid_gen;
1252 	struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
1253 	struct list_head active_mmu_pages;
1254 	struct list_head zapped_obsolete_pages;
1255 	/*
1256 	 * A list of kvm_mmu_page structs that, if zapped, could possibly be
1257 	 * replaced by an NX huge page.  A shadow page is on this list if its
1258 	 * existence disallows an NX huge page (nx_huge_page_disallowed is set)
1259 	 * and there are no other conditions that prevent a huge page, e.g.
1260 	 * the backing host page is huge, dirtly logging is not enabled for its
1261 	 * memslot, etc...  Note, zapping shadow pages on this list doesn't
1262 	 * guarantee an NX huge page will be created in its stead, e.g. if the
1263 	 * guest attempts to execute from the region then KVM obviously can't
1264 	 * create an NX huge page (without hanging the guest).
1265 	 */
1266 	struct list_head possible_nx_huge_pages;
1267 #ifdef CONFIG_KVM_EXTERNAL_WRITE_TRACKING
1268 	struct kvm_page_track_notifier_head track_notifier_head;
1269 #endif
1270 	/*
1271 	 * Protects marking pages unsync during page faults, as TDP MMU page
1272 	 * faults only take mmu_lock for read.  For simplicity, the unsync
1273 	 * pages lock is always taken when marking pages unsync regardless of
1274 	 * whether mmu_lock is held for read or write.
1275 	 */
1276 	spinlock_t mmu_unsync_pages_lock;
1277 
1278 	struct list_head assigned_dev_head;
1279 	struct iommu_domain *iommu_domain;
1280 	bool iommu_noncoherent;
1281 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA
1282 	atomic_t noncoherent_dma_count;
1283 #define __KVM_HAVE_ARCH_ASSIGNED_DEVICE
1284 	atomic_t assigned_device_count;
1285 	struct kvm_pic *vpic;
1286 	struct kvm_ioapic *vioapic;
1287 	struct kvm_pit *vpit;
1288 	atomic_t vapics_in_nmi_mode;
1289 	struct mutex apic_map_lock;
1290 	struct kvm_apic_map __rcu *apic_map;
1291 	atomic_t apic_map_dirty;
1292 
1293 	bool apic_access_memslot_enabled;
1294 	bool apic_access_memslot_inhibited;
1295 
1296 	/* Protects apicv_inhibit_reasons */
1297 	struct rw_semaphore apicv_update_lock;
1298 	unsigned long apicv_inhibit_reasons;
1299 
1300 	gpa_t wall_clock;
1301 
1302 	bool mwait_in_guest;
1303 	bool hlt_in_guest;
1304 	bool pause_in_guest;
1305 	bool cstate_in_guest;
1306 
1307 	unsigned long irq_sources_bitmap;
1308 	s64 kvmclock_offset;
1309 
1310 	/*
1311 	 * This also protects nr_vcpus_matched_tsc which is read from a
1312 	 * preemption-disabled region, so it must be a raw spinlock.
1313 	 */
1314 	raw_spinlock_t tsc_write_lock;
1315 	u64 last_tsc_nsec;
1316 	u64 last_tsc_write;
1317 	u32 last_tsc_khz;
1318 	u64 last_tsc_offset;
1319 	u64 cur_tsc_nsec;
1320 	u64 cur_tsc_write;
1321 	u64 cur_tsc_offset;
1322 	u64 cur_tsc_generation;
1323 	int nr_vcpus_matched_tsc;
1324 
1325 	u32 default_tsc_khz;
1326 
1327 	seqcount_raw_spinlock_t pvclock_sc;
1328 	bool use_master_clock;
1329 	u64 master_kernel_ns;
1330 	u64 master_cycle_now;
1331 	struct delayed_work kvmclock_update_work;
1332 	struct delayed_work kvmclock_sync_work;
1333 
1334 	struct kvm_xen_hvm_config xen_hvm_config;
1335 
1336 	/* reads protected by irq_srcu, writes by irq_lock */
1337 	struct hlist_head mask_notifier_list;
1338 
1339 	struct kvm_hv hyperv;
1340 	struct kvm_xen xen;
1341 
1342 	bool backwards_tsc_observed;
1343 	bool boot_vcpu_runs_old_kvmclock;
1344 	u32 bsp_vcpu_id;
1345 
1346 	u64 disabled_quirks;
1347 
1348 	enum kvm_irqchip_mode irqchip_mode;
1349 	u8 nr_reserved_ioapic_pins;
1350 
1351 	bool disabled_lapic_found;
1352 
1353 	bool x2apic_format;
1354 	bool x2apic_broadcast_quirk_disabled;
1355 
1356 	bool guest_can_read_msr_platform_info;
1357 	bool exception_payload_enabled;
1358 
1359 	bool triple_fault_event;
1360 
1361 	bool bus_lock_detection_enabled;
1362 	bool enable_pmu;
1363 
1364 	u32 notify_window;
1365 	u32 notify_vmexit_flags;
1366 	/*
1367 	 * If exit_on_emulation_error is set, and the in-kernel instruction
1368 	 * emulator fails to emulate an instruction, allow userspace
1369 	 * the opportunity to look at it.
1370 	 */
1371 	bool exit_on_emulation_error;
1372 
1373 	/* Deflect RDMSR and WRMSR to user space when they trigger a #GP */
1374 	u32 user_space_msr_mask;
1375 	struct kvm_x86_msr_filter __rcu *msr_filter;
1376 
1377 	u32 hypercall_exit_enabled;
1378 
1379 	/* Guest can access the SGX PROVISIONKEY. */
1380 	bool sgx_provisioning_allowed;
1381 
1382 	struct kvm_x86_pmu_event_filter __rcu *pmu_event_filter;
1383 	struct task_struct *nx_huge_page_recovery_thread;
1384 
1385 #ifdef CONFIG_X86_64
1386 	/* The number of TDP MMU pages across all roots. */
1387 	atomic64_t tdp_mmu_pages;
1388 
1389 	/*
1390 	 * List of struct kvm_mmu_pages being used as roots.
1391 	 * All struct kvm_mmu_pages in the list should have
1392 	 * tdp_mmu_page set.
1393 	 *
1394 	 * For reads, this list is protected by:
1395 	 *	the MMU lock in read mode + RCU or
1396 	 *	the MMU lock in write mode
1397 	 *
1398 	 * For writes, this list is protected by:
1399 	 *	the MMU lock in read mode + the tdp_mmu_pages_lock or
1400 	 *	the MMU lock in write mode
1401 	 *
1402 	 * Roots will remain in the list until their tdp_mmu_root_count
1403 	 * drops to zero, at which point the thread that decremented the
1404 	 * count to zero should removed the root from the list and clean
1405 	 * it up, freeing the root after an RCU grace period.
1406 	 */
1407 	struct list_head tdp_mmu_roots;
1408 
1409 	/*
1410 	 * Protects accesses to the following fields when the MMU lock
1411 	 * is held in read mode:
1412 	 *  - tdp_mmu_roots (above)
1413 	 *  - the link field of kvm_mmu_page structs used by the TDP MMU
1414 	 *  - possible_nx_huge_pages;
1415 	 *  - the possible_nx_huge_page_link field of kvm_mmu_page structs used
1416 	 *    by the TDP MMU
1417 	 * It is acceptable, but not necessary, to acquire this lock when
1418 	 * the thread holds the MMU lock in write mode.
1419 	 */
1420 	spinlock_t tdp_mmu_pages_lock;
1421 #endif /* CONFIG_X86_64 */
1422 
1423 	/*
1424 	 * If set, at least one shadow root has been allocated. This flag
1425 	 * is used as one input when determining whether certain memslot
1426 	 * related allocations are necessary.
1427 	 */
1428 	bool shadow_root_allocated;
1429 
1430 #if IS_ENABLED(CONFIG_HYPERV)
1431 	hpa_t	hv_root_tdp;
1432 	spinlock_t hv_root_tdp_lock;
1433 #endif
1434 	/*
1435 	 * VM-scope maximum vCPU ID. Used to determine the size of structures
1436 	 * that increase along with the maximum vCPU ID, in which case, using
1437 	 * the global KVM_MAX_VCPU_IDS may lead to significant memory waste.
1438 	 */
1439 	u32 max_vcpu_ids;
1440 
1441 	bool disable_nx_huge_pages;
1442 
1443 	/*
1444 	 * Memory caches used to allocate shadow pages when performing eager
1445 	 * page splitting. No need for a shadowed_info_cache since eager page
1446 	 * splitting only allocates direct shadow pages.
1447 	 *
1448 	 * Protected by kvm->slots_lock.
1449 	 */
1450 	struct kvm_mmu_memory_cache split_shadow_page_cache;
1451 	struct kvm_mmu_memory_cache split_page_header_cache;
1452 
1453 	/*
1454 	 * Memory cache used to allocate pte_list_desc structs while splitting
1455 	 * huge pages. In the worst case, to split one huge page, 512
1456 	 * pte_list_desc structs are needed to add each lower level leaf sptep
1457 	 * to the rmap plus 1 to extend the parent_ptes rmap of the lower level
1458 	 * page table.
1459 	 *
1460 	 * Protected by kvm->slots_lock.
1461 	 */
1462 #define SPLIT_DESC_CACHE_MIN_NR_OBJECTS (SPTE_ENT_PER_PAGE + 1)
1463 	struct kvm_mmu_memory_cache split_desc_cache;
1464 };
1465 
1466 struct kvm_vm_stat {
1467 	struct kvm_vm_stat_generic generic;
1468 	u64 mmu_shadow_zapped;
1469 	u64 mmu_pte_write;
1470 	u64 mmu_pde_zapped;
1471 	u64 mmu_flooded;
1472 	u64 mmu_recycled;
1473 	u64 mmu_cache_miss;
1474 	u64 mmu_unsync;
1475 	union {
1476 		struct {
1477 			atomic64_t pages_4k;
1478 			atomic64_t pages_2m;
1479 			atomic64_t pages_1g;
1480 		};
1481 		atomic64_t pages[KVM_NR_PAGE_SIZES];
1482 	};
1483 	u64 nx_lpage_splits;
1484 	u64 max_mmu_page_hash_collisions;
1485 	u64 max_mmu_rmap_size;
1486 };
1487 
1488 struct kvm_vcpu_stat {
1489 	struct kvm_vcpu_stat_generic generic;
1490 	u64 pf_taken;
1491 	u64 pf_fixed;
1492 	u64 pf_emulate;
1493 	u64 pf_spurious;
1494 	u64 pf_fast;
1495 	u64 pf_mmio_spte_created;
1496 	u64 pf_guest;
1497 	u64 tlb_flush;
1498 	u64 invlpg;
1499 
1500 	u64 exits;
1501 	u64 io_exits;
1502 	u64 mmio_exits;
1503 	u64 signal_exits;
1504 	u64 irq_window_exits;
1505 	u64 nmi_window_exits;
1506 	u64 l1d_flush;
1507 	u64 halt_exits;
1508 	u64 request_irq_exits;
1509 	u64 irq_exits;
1510 	u64 host_state_reload;
1511 	u64 fpu_reload;
1512 	u64 insn_emulation;
1513 	u64 insn_emulation_fail;
1514 	u64 hypercalls;
1515 	u64 irq_injections;
1516 	u64 nmi_injections;
1517 	u64 req_event;
1518 	u64 nested_run;
1519 	u64 directed_yield_attempted;
1520 	u64 directed_yield_successful;
1521 	u64 preemption_reported;
1522 	u64 preemption_other;
1523 	u64 guest_mode;
1524 	u64 notify_window_exits;
1525 };
1526 
1527 struct x86_instruction_info;
1528 
1529 struct msr_data {
1530 	bool host_initiated;
1531 	u32 index;
1532 	u64 data;
1533 };
1534 
1535 struct kvm_lapic_irq {
1536 	u32 vector;
1537 	u16 delivery_mode;
1538 	u16 dest_mode;
1539 	bool level;
1540 	u16 trig_mode;
1541 	u32 shorthand;
1542 	u32 dest_id;
1543 	bool msi_redir_hint;
1544 };
1545 
kvm_lapic_irq_dest_mode(bool dest_mode_logical)1546 static inline u16 kvm_lapic_irq_dest_mode(bool dest_mode_logical)
1547 {
1548 	return dest_mode_logical ? APIC_DEST_LOGICAL : APIC_DEST_PHYSICAL;
1549 }
1550 
1551 struct kvm_x86_ops {
1552 	const char *name;
1553 
1554 	int (*check_processor_compatibility)(void);
1555 
1556 	int (*hardware_enable)(void);
1557 	void (*hardware_disable)(void);
1558 	void (*hardware_unsetup)(void);
1559 	bool (*has_emulated_msr)(struct kvm *kvm, u32 index);
1560 	void (*vcpu_after_set_cpuid)(struct kvm_vcpu *vcpu);
1561 
1562 	unsigned int vm_size;
1563 	int (*vm_init)(struct kvm *kvm);
1564 	void (*vm_destroy)(struct kvm *kvm);
1565 
1566 	/* Create, but do not attach this VCPU */
1567 	int (*vcpu_precreate)(struct kvm *kvm);
1568 	int (*vcpu_create)(struct kvm_vcpu *vcpu);
1569 	void (*vcpu_free)(struct kvm_vcpu *vcpu);
1570 	void (*vcpu_reset)(struct kvm_vcpu *vcpu, bool init_event);
1571 
1572 	void (*prepare_switch_to_guest)(struct kvm_vcpu *vcpu);
1573 	void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
1574 	void (*vcpu_put)(struct kvm_vcpu *vcpu);
1575 
1576 	void (*update_exception_bitmap)(struct kvm_vcpu *vcpu);
1577 	int (*get_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1578 	int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1579 	u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
1580 	void (*get_segment)(struct kvm_vcpu *vcpu,
1581 			    struct kvm_segment *var, int seg);
1582 	int (*get_cpl)(struct kvm_vcpu *vcpu);
1583 	void (*set_segment)(struct kvm_vcpu *vcpu,
1584 			    struct kvm_segment *var, int seg);
1585 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
1586 	bool (*is_valid_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
1587 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
1588 	void (*post_set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
1589 	bool (*is_valid_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
1590 	void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
1591 	int (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
1592 	void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1593 	void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1594 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1595 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1596 	void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu);
1597 	void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
1598 	void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
1599 	unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
1600 	void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
1601 	bool (*get_if_flag)(struct kvm_vcpu *vcpu);
1602 
1603 	void (*flush_tlb_all)(struct kvm_vcpu *vcpu);
1604 	void (*flush_tlb_current)(struct kvm_vcpu *vcpu);
1605 	int  (*flush_remote_tlbs)(struct kvm *kvm);
1606 	int  (*flush_remote_tlbs_range)(struct kvm *kvm, gfn_t gfn,
1607 					gfn_t nr_pages);
1608 
1609 	/*
1610 	 * Flush any TLB entries associated with the given GVA.
1611 	 * Does not need to flush GPA->HPA mappings.
1612 	 * Can potentially get non-canonical addresses through INVLPGs, which
1613 	 * the implementation may choose to ignore if appropriate.
1614 	 */
1615 	void (*flush_tlb_gva)(struct kvm_vcpu *vcpu, gva_t addr);
1616 
1617 	/*
1618 	 * Flush any TLB entries created by the guest.  Like tlb_flush_gva(),
1619 	 * does not need to flush GPA->HPA mappings.
1620 	 */
1621 	void (*flush_tlb_guest)(struct kvm_vcpu *vcpu);
1622 
1623 	int (*vcpu_pre_run)(struct kvm_vcpu *vcpu);
1624 	enum exit_fastpath_completion (*vcpu_run)(struct kvm_vcpu *vcpu);
1625 	int (*handle_exit)(struct kvm_vcpu *vcpu,
1626 		enum exit_fastpath_completion exit_fastpath);
1627 	int (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
1628 	void (*update_emulated_instruction)(struct kvm_vcpu *vcpu);
1629 	void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
1630 	u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu);
1631 	void (*patch_hypercall)(struct kvm_vcpu *vcpu,
1632 				unsigned char *hypercall_addr);
1633 	void (*inject_irq)(struct kvm_vcpu *vcpu, bool reinjected);
1634 	void (*inject_nmi)(struct kvm_vcpu *vcpu);
1635 	void (*inject_exception)(struct kvm_vcpu *vcpu);
1636 	void (*cancel_injection)(struct kvm_vcpu *vcpu);
1637 	int (*interrupt_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1638 	int (*nmi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1639 	bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
1640 	void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
1641 	/* Whether or not a virtual NMI is pending in hardware. */
1642 	bool (*is_vnmi_pending)(struct kvm_vcpu *vcpu);
1643 	/*
1644 	 * Attempt to pend a virtual NMI in harware.  Returns %true on success
1645 	 * to allow using static_call_ret0 as the fallback.
1646 	 */
1647 	bool (*set_vnmi_pending)(struct kvm_vcpu *vcpu);
1648 	void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
1649 	void (*enable_irq_window)(struct kvm_vcpu *vcpu);
1650 	void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
1651 	bool (*check_apicv_inhibit_reasons)(enum kvm_apicv_inhibit reason);
1652 	const unsigned long required_apicv_inhibits;
1653 	bool allow_apicv_in_x2apic_without_x2apic_virtualization;
1654 	void (*refresh_apicv_exec_ctrl)(struct kvm_vcpu *vcpu);
1655 	void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
1656 	void (*hwapic_isr_update)(int isr);
1657 	bool (*guest_apic_has_interrupt)(struct kvm_vcpu *vcpu);
1658 	void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
1659 	void (*set_virtual_apic_mode)(struct kvm_vcpu *vcpu);
1660 	void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu);
1661 	void (*deliver_interrupt)(struct kvm_lapic *apic, int delivery_mode,
1662 				  int trig_mode, int vector);
1663 	int (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
1664 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
1665 	int (*set_identity_map_addr)(struct kvm *kvm, u64 ident_addr);
1666 	u8 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
1667 
1668 	void (*load_mmu_pgd)(struct kvm_vcpu *vcpu, hpa_t root_hpa,
1669 			     int root_level);
1670 
1671 	bool (*has_wbinvd_exit)(void);
1672 
1673 	u64 (*get_l2_tsc_offset)(struct kvm_vcpu *vcpu);
1674 	u64 (*get_l2_tsc_multiplier)(struct kvm_vcpu *vcpu);
1675 	void (*write_tsc_offset)(struct kvm_vcpu *vcpu);
1676 	void (*write_tsc_multiplier)(struct kvm_vcpu *vcpu);
1677 
1678 	/*
1679 	 * Retrieve somewhat arbitrary exit information.  Intended to
1680 	 * be used only from within tracepoints or error paths.
1681 	 */
1682 	void (*get_exit_info)(struct kvm_vcpu *vcpu, u32 *reason,
1683 			      u64 *info1, u64 *info2,
1684 			      u32 *exit_int_info, u32 *exit_int_info_err_code);
1685 
1686 	int (*check_intercept)(struct kvm_vcpu *vcpu,
1687 			       struct x86_instruction_info *info,
1688 			       enum x86_intercept_stage stage,
1689 			       struct x86_exception *exception);
1690 	void (*handle_exit_irqoff)(struct kvm_vcpu *vcpu);
1691 
1692 	void (*request_immediate_exit)(struct kvm_vcpu *vcpu);
1693 
1694 	void (*sched_in)(struct kvm_vcpu *kvm, int cpu);
1695 
1696 	/*
1697 	 * Size of the CPU's dirty log buffer, i.e. VMX's PML buffer.  A zero
1698 	 * value indicates CPU dirty logging is unsupported or disabled.
1699 	 */
1700 	int cpu_dirty_log_size;
1701 	void (*update_cpu_dirty_logging)(struct kvm_vcpu *vcpu);
1702 
1703 	const struct kvm_x86_nested_ops *nested_ops;
1704 
1705 	void (*vcpu_blocking)(struct kvm_vcpu *vcpu);
1706 	void (*vcpu_unblocking)(struct kvm_vcpu *vcpu);
1707 
1708 	int (*pi_update_irte)(struct kvm *kvm, unsigned int host_irq,
1709 			      uint32_t guest_irq, bool set);
1710 	void (*pi_start_assignment)(struct kvm *kvm);
1711 	void (*apicv_pre_state_restore)(struct kvm_vcpu *vcpu);
1712 	void (*apicv_post_state_restore)(struct kvm_vcpu *vcpu);
1713 	bool (*dy_apicv_has_pending_interrupt)(struct kvm_vcpu *vcpu);
1714 
1715 	int (*set_hv_timer)(struct kvm_vcpu *vcpu, u64 guest_deadline_tsc,
1716 			    bool *expired);
1717 	void (*cancel_hv_timer)(struct kvm_vcpu *vcpu);
1718 
1719 	void (*setup_mce)(struct kvm_vcpu *vcpu);
1720 
1721 #ifdef CONFIG_KVM_SMM
1722 	int (*smi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1723 	int (*enter_smm)(struct kvm_vcpu *vcpu, union kvm_smram *smram);
1724 	int (*leave_smm)(struct kvm_vcpu *vcpu, const union kvm_smram *smram);
1725 	void (*enable_smi_window)(struct kvm_vcpu *vcpu);
1726 #endif
1727 
1728 	int (*mem_enc_ioctl)(struct kvm *kvm, void __user *argp);
1729 	int (*mem_enc_register_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1730 	int (*mem_enc_unregister_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1731 	int (*vm_copy_enc_context_from)(struct kvm *kvm, unsigned int source_fd);
1732 	int (*vm_move_enc_context_from)(struct kvm *kvm, unsigned int source_fd);
1733 	void (*guest_memory_reclaimed)(struct kvm *kvm);
1734 
1735 	int (*get_msr_feature)(struct kvm_msr_entry *entry);
1736 
1737 	bool (*can_emulate_instruction)(struct kvm_vcpu *vcpu, int emul_type,
1738 					void *insn, int insn_len);
1739 
1740 	bool (*apic_init_signal_blocked)(struct kvm_vcpu *vcpu);
1741 	int (*enable_l2_tlb_flush)(struct kvm_vcpu *vcpu);
1742 
1743 	void (*migrate_timers)(struct kvm_vcpu *vcpu);
1744 	void (*msr_filter_changed)(struct kvm_vcpu *vcpu);
1745 	int (*complete_emulated_msr)(struct kvm_vcpu *vcpu, int err);
1746 
1747 	void (*vcpu_deliver_sipi_vector)(struct kvm_vcpu *vcpu, u8 vector);
1748 
1749 	/*
1750 	 * Returns vCPU specific APICv inhibit reasons
1751 	 */
1752 	unsigned long (*vcpu_get_apicv_inhibit_reasons)(struct kvm_vcpu *vcpu);
1753 };
1754 
1755 struct kvm_x86_nested_ops {
1756 	void (*leave_nested)(struct kvm_vcpu *vcpu);
1757 	bool (*is_exception_vmexit)(struct kvm_vcpu *vcpu, u8 vector,
1758 				    u32 error_code);
1759 	int (*check_events)(struct kvm_vcpu *vcpu);
1760 	bool (*has_events)(struct kvm_vcpu *vcpu);
1761 	void (*triple_fault)(struct kvm_vcpu *vcpu);
1762 	int (*get_state)(struct kvm_vcpu *vcpu,
1763 			 struct kvm_nested_state __user *user_kvm_nested_state,
1764 			 unsigned user_data_size);
1765 	int (*set_state)(struct kvm_vcpu *vcpu,
1766 			 struct kvm_nested_state __user *user_kvm_nested_state,
1767 			 struct kvm_nested_state *kvm_state);
1768 	bool (*get_nested_state_pages)(struct kvm_vcpu *vcpu);
1769 	int (*write_log_dirty)(struct kvm_vcpu *vcpu, gpa_t l2_gpa);
1770 
1771 	int (*enable_evmcs)(struct kvm_vcpu *vcpu,
1772 			    uint16_t *vmcs_version);
1773 	uint16_t (*get_evmcs_version)(struct kvm_vcpu *vcpu);
1774 	void (*hv_inject_synthetic_vmexit_post_tlb_flush)(struct kvm_vcpu *vcpu);
1775 };
1776 
1777 struct kvm_x86_init_ops {
1778 	int (*hardware_setup)(void);
1779 	unsigned int (*handle_intel_pt_intr)(void);
1780 
1781 	struct kvm_x86_ops *runtime_ops;
1782 	struct kvm_pmu_ops *pmu_ops;
1783 };
1784 
1785 struct kvm_arch_async_pf {
1786 	u32 token;
1787 	gfn_t gfn;
1788 	unsigned long cr3;
1789 	bool direct_map;
1790 };
1791 
1792 extern u32 __read_mostly kvm_nr_uret_msrs;
1793 extern u64 __read_mostly host_efer;
1794 extern bool __read_mostly allow_smaller_maxphyaddr;
1795 extern bool __read_mostly enable_apicv;
1796 extern struct kvm_x86_ops kvm_x86_ops;
1797 
1798 #define KVM_X86_OP(func) \
1799 	DECLARE_STATIC_CALL(kvm_x86_##func, *(((struct kvm_x86_ops *)0)->func));
1800 #define KVM_X86_OP_OPTIONAL KVM_X86_OP
1801 #define KVM_X86_OP_OPTIONAL_RET0 KVM_X86_OP
1802 #include <asm/kvm-x86-ops.h>
1803 
1804 int kvm_x86_vendor_init(struct kvm_x86_init_ops *ops);
1805 void kvm_x86_vendor_exit(void);
1806 
1807 #define __KVM_HAVE_ARCH_VM_ALLOC
kvm_arch_alloc_vm(void)1808 static inline struct kvm *kvm_arch_alloc_vm(void)
1809 {
1810 	return __vmalloc(kvm_x86_ops.vm_size, GFP_KERNEL_ACCOUNT | __GFP_ZERO);
1811 }
1812 
1813 #define __KVM_HAVE_ARCH_VM_FREE
1814 void kvm_arch_free_vm(struct kvm *kvm);
1815 
1816 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS
kvm_arch_flush_remote_tlbs(struct kvm * kvm)1817 static inline int kvm_arch_flush_remote_tlbs(struct kvm *kvm)
1818 {
1819 	if (kvm_x86_ops.flush_remote_tlbs &&
1820 	    !static_call(kvm_x86_flush_remote_tlbs)(kvm))
1821 		return 0;
1822 	else
1823 		return -ENOTSUPP;
1824 }
1825 
1826 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS_RANGE
1827 
1828 #define kvm_arch_pmi_in_guest(vcpu) \
1829 	((vcpu) && (vcpu)->arch.handling_intr_from_guest)
1830 
1831 void __init kvm_mmu_x86_module_init(void);
1832 int kvm_mmu_vendor_module_init(void);
1833 void kvm_mmu_vendor_module_exit(void);
1834 
1835 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
1836 int kvm_mmu_create(struct kvm_vcpu *vcpu);
1837 void kvm_mmu_init_vm(struct kvm *kvm);
1838 void kvm_mmu_uninit_vm(struct kvm *kvm);
1839 
1840 void kvm_mmu_after_set_cpuid(struct kvm_vcpu *vcpu);
1841 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
1842 void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
1843 				      const struct kvm_memory_slot *memslot,
1844 				      int start_level);
1845 void kvm_mmu_slot_try_split_huge_pages(struct kvm *kvm,
1846 				       const struct kvm_memory_slot *memslot,
1847 				       int target_level);
1848 void kvm_mmu_try_split_huge_pages(struct kvm *kvm,
1849 				  const struct kvm_memory_slot *memslot,
1850 				  u64 start, u64 end,
1851 				  int target_level);
1852 void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
1853 				   const struct kvm_memory_slot *memslot);
1854 void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
1855 				   const struct kvm_memory_slot *memslot);
1856 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, u64 gen);
1857 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned long kvm_nr_mmu_pages);
1858 
1859 int load_pdptrs(struct kvm_vcpu *vcpu, unsigned long cr3);
1860 
1861 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
1862 			  const void *val, int bytes);
1863 
1864 struct kvm_irq_mask_notifier {
1865 	void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
1866 	int irq;
1867 	struct hlist_node link;
1868 };
1869 
1870 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
1871 				    struct kvm_irq_mask_notifier *kimn);
1872 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
1873 				      struct kvm_irq_mask_notifier *kimn);
1874 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
1875 			     bool mask);
1876 
1877 extern bool tdp_enabled;
1878 
1879 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
1880 
1881 /*
1882  * EMULTYPE_NO_DECODE - Set when re-emulating an instruction (after completing
1883  *			userspace I/O) to indicate that the emulation context
1884  *			should be reused as is, i.e. skip initialization of
1885  *			emulation context, instruction fetch and decode.
1886  *
1887  * EMULTYPE_TRAP_UD - Set when emulating an intercepted #UD from hardware.
1888  *		      Indicates that only select instructions (tagged with
1889  *		      EmulateOnUD) should be emulated (to minimize the emulator
1890  *		      attack surface).  See also EMULTYPE_TRAP_UD_FORCED.
1891  *
1892  * EMULTYPE_SKIP - Set when emulating solely to skip an instruction, i.e. to
1893  *		   decode the instruction length.  For use *only* by
1894  *		   kvm_x86_ops.skip_emulated_instruction() implementations if
1895  *		   EMULTYPE_COMPLETE_USER_EXIT is not set.
1896  *
1897  * EMULTYPE_ALLOW_RETRY_PF - Set when the emulator should resume the guest to
1898  *			     retry native execution under certain conditions,
1899  *			     Can only be set in conjunction with EMULTYPE_PF.
1900  *
1901  * EMULTYPE_TRAP_UD_FORCED - Set when emulating an intercepted #UD that was
1902  *			     triggered by KVM's magic "force emulation" prefix,
1903  *			     which is opt in via module param (off by default).
1904  *			     Bypasses EmulateOnUD restriction despite emulating
1905  *			     due to an intercepted #UD (see EMULTYPE_TRAP_UD).
1906  *			     Used to test the full emulator from userspace.
1907  *
1908  * EMULTYPE_VMWARE_GP - Set when emulating an intercepted #GP for VMware
1909  *			backdoor emulation, which is opt in via module param.
1910  *			VMware backdoor emulation handles select instructions
1911  *			and reinjects the #GP for all other cases.
1912  *
1913  * EMULTYPE_PF - Set when emulating MMIO by way of an intercepted #PF, in which
1914  *		 case the CR2/GPA value pass on the stack is valid.
1915  *
1916  * EMULTYPE_COMPLETE_USER_EXIT - Set when the emulator should update interruptibility
1917  *				 state and inject single-step #DBs after skipping
1918  *				 an instruction (after completing userspace I/O).
1919  *
1920  * EMULTYPE_WRITE_PF_TO_SP - Set when emulating an intercepted page fault that
1921  *			     is attempting to write a gfn that contains one or
1922  *			     more of the PTEs used to translate the write itself,
1923  *			     and the owning page table is being shadowed by KVM.
1924  *			     If emulation of the faulting instruction fails and
1925  *			     this flag is set, KVM will exit to userspace instead
1926  *			     of retrying emulation as KVM cannot make forward
1927  *			     progress.
1928  *
1929  *			     If emulation fails for a write to guest page tables,
1930  *			     KVM unprotects (zaps) the shadow page for the target
1931  *			     gfn and resumes the guest to retry the non-emulatable
1932  *			     instruction (on hardware).  Unprotecting the gfn
1933  *			     doesn't allow forward progress for a self-changing
1934  *			     access because doing so also zaps the translation for
1935  *			     the gfn, i.e. retrying the instruction will hit a
1936  *			     !PRESENT fault, which results in a new shadow page
1937  *			     and sends KVM back to square one.
1938  */
1939 #define EMULTYPE_NO_DECODE	    (1 << 0)
1940 #define EMULTYPE_TRAP_UD	    (1 << 1)
1941 #define EMULTYPE_SKIP		    (1 << 2)
1942 #define EMULTYPE_ALLOW_RETRY_PF	    (1 << 3)
1943 #define EMULTYPE_TRAP_UD_FORCED	    (1 << 4)
1944 #define EMULTYPE_VMWARE_GP	    (1 << 5)
1945 #define EMULTYPE_PF		    (1 << 6)
1946 #define EMULTYPE_COMPLETE_USER_EXIT (1 << 7)
1947 #define EMULTYPE_WRITE_PF_TO_SP	    (1 << 8)
1948 
1949 int kvm_emulate_instruction(struct kvm_vcpu *vcpu, int emulation_type);
1950 int kvm_emulate_instruction_from_buffer(struct kvm_vcpu *vcpu,
1951 					void *insn, int insn_len);
1952 void __kvm_prepare_emulation_failure_exit(struct kvm_vcpu *vcpu,
1953 					  u64 *data, u8 ndata);
1954 void kvm_prepare_emulation_failure_exit(struct kvm_vcpu *vcpu);
1955 
1956 void kvm_enable_efer_bits(u64);
1957 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
1958 int __kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data, bool host_initiated);
1959 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data);
1960 int kvm_set_msr(struct kvm_vcpu *vcpu, u32 index, u64 data);
1961 int kvm_emulate_rdmsr(struct kvm_vcpu *vcpu);
1962 int kvm_emulate_wrmsr(struct kvm_vcpu *vcpu);
1963 int kvm_emulate_as_nop(struct kvm_vcpu *vcpu);
1964 int kvm_emulate_invd(struct kvm_vcpu *vcpu);
1965 int kvm_emulate_mwait(struct kvm_vcpu *vcpu);
1966 int kvm_handle_invalid_op(struct kvm_vcpu *vcpu);
1967 int kvm_emulate_monitor(struct kvm_vcpu *vcpu);
1968 
1969 int kvm_fast_pio(struct kvm_vcpu *vcpu, int size, unsigned short port, int in);
1970 int kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
1971 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
1972 int kvm_emulate_halt_noskip(struct kvm_vcpu *vcpu);
1973 int kvm_emulate_ap_reset_hold(struct kvm_vcpu *vcpu);
1974 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
1975 
1976 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
1977 void kvm_set_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
1978 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
1979 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector);
1980 
1981 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
1982 		    int reason, bool has_error_code, u32 error_code);
1983 
1984 void kvm_post_set_cr0(struct kvm_vcpu *vcpu, unsigned long old_cr0, unsigned long cr0);
1985 void kvm_post_set_cr4(struct kvm_vcpu *vcpu, unsigned long old_cr4, unsigned long cr4);
1986 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
1987 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
1988 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
1989 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
1990 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
1991 void kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
1992 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
1993 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
1994 int kvm_emulate_xsetbv(struct kvm_vcpu *vcpu);
1995 
1996 int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1997 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1998 
1999 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
2000 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
2001 int kvm_emulate_rdpmc(struct kvm_vcpu *vcpu);
2002 
2003 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
2004 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
2005 void kvm_queue_exception_p(struct kvm_vcpu *vcpu, unsigned nr, unsigned long payload);
2006 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
2007 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
2008 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
2009 void kvm_inject_emulated_page_fault(struct kvm_vcpu *vcpu,
2010 				    struct x86_exception *fault);
2011 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
2012 bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr);
2013 
__kvm_irq_line_state(unsigned long * irq_state,int irq_source_id,int level)2014 static inline int __kvm_irq_line_state(unsigned long *irq_state,
2015 				       int irq_source_id, int level)
2016 {
2017 	/* Logical OR for level trig interrupt */
2018 	if (level)
2019 		__set_bit(irq_source_id, irq_state);
2020 	else
2021 		__clear_bit(irq_source_id, irq_state);
2022 
2023 	return !!(*irq_state);
2024 }
2025 
2026 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
2027 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
2028 
2029 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
2030 int kvm_get_nr_pending_nmis(struct kvm_vcpu *vcpu);
2031 
2032 void kvm_update_dr7(struct kvm_vcpu *vcpu);
2033 
2034 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
2035 void kvm_mmu_free_roots(struct kvm *kvm, struct kvm_mmu *mmu,
2036 			ulong roots_to_free);
2037 void kvm_mmu_free_guest_mode_roots(struct kvm *kvm, struct kvm_mmu *mmu);
2038 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
2039 			      struct x86_exception *exception);
2040 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
2041 			       struct x86_exception *exception);
2042 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
2043 				struct x86_exception *exception);
2044 
2045 bool kvm_apicv_activated(struct kvm *kvm);
2046 bool kvm_vcpu_apicv_activated(struct kvm_vcpu *vcpu);
2047 void __kvm_vcpu_update_apicv(struct kvm_vcpu *vcpu);
2048 void __kvm_set_or_clear_apicv_inhibit(struct kvm *kvm,
2049 				      enum kvm_apicv_inhibit reason, bool set);
2050 void kvm_set_or_clear_apicv_inhibit(struct kvm *kvm,
2051 				    enum kvm_apicv_inhibit reason, bool set);
2052 
kvm_set_apicv_inhibit(struct kvm * kvm,enum kvm_apicv_inhibit reason)2053 static inline void kvm_set_apicv_inhibit(struct kvm *kvm,
2054 					 enum kvm_apicv_inhibit reason)
2055 {
2056 	kvm_set_or_clear_apicv_inhibit(kvm, reason, true);
2057 }
2058 
kvm_clear_apicv_inhibit(struct kvm * kvm,enum kvm_apicv_inhibit reason)2059 static inline void kvm_clear_apicv_inhibit(struct kvm *kvm,
2060 					   enum kvm_apicv_inhibit reason)
2061 {
2062 	kvm_set_or_clear_apicv_inhibit(kvm, reason, false);
2063 }
2064 
2065 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
2066 
2067 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u64 error_code,
2068 		       void *insn, int insn_len);
2069 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
2070 void kvm_mmu_invalidate_addr(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
2071 			     u64 addr, unsigned long roots);
2072 void kvm_mmu_invpcid_gva(struct kvm_vcpu *vcpu, gva_t gva, unsigned long pcid);
2073 void kvm_mmu_new_pgd(struct kvm_vcpu *vcpu, gpa_t new_pgd);
2074 
2075 void kvm_configure_mmu(bool enable_tdp, int tdp_forced_root_level,
2076 		       int tdp_max_root_level, int tdp_huge_page_level);
2077 
kvm_read_ldt(void)2078 static inline u16 kvm_read_ldt(void)
2079 {
2080 	u16 ldt;
2081 	asm("sldt %0" : "=g"(ldt));
2082 	return ldt;
2083 }
2084 
kvm_load_ldt(u16 sel)2085 static inline void kvm_load_ldt(u16 sel)
2086 {
2087 	asm("lldt %0" : : "rm"(sel));
2088 }
2089 
2090 #ifdef CONFIG_X86_64
read_msr(unsigned long msr)2091 static inline unsigned long read_msr(unsigned long msr)
2092 {
2093 	u64 value;
2094 
2095 	rdmsrl(msr, value);
2096 	return value;
2097 }
2098 #endif
2099 
kvm_inject_gp(struct kvm_vcpu * vcpu,u32 error_code)2100 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
2101 {
2102 	kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
2103 }
2104 
2105 #define TSS_IOPB_BASE_OFFSET 0x66
2106 #define TSS_BASE_SIZE 0x68
2107 #define TSS_IOPB_SIZE (65536 / 8)
2108 #define TSS_REDIRECTION_SIZE (256 / 8)
2109 #define RMODE_TSS_SIZE							\
2110 	(TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
2111 
2112 enum {
2113 	TASK_SWITCH_CALL = 0,
2114 	TASK_SWITCH_IRET = 1,
2115 	TASK_SWITCH_JMP = 2,
2116 	TASK_SWITCH_GATE = 3,
2117 };
2118 
2119 #define HF_GUEST_MASK		(1 << 0) /* VCPU is in guest-mode */
2120 
2121 #ifdef CONFIG_KVM_SMM
2122 #define HF_SMM_MASK		(1 << 1)
2123 #define HF_SMM_INSIDE_NMI_MASK	(1 << 2)
2124 
2125 # define __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
2126 # define KVM_ADDRESS_SPACE_NUM 2
2127 # define kvm_arch_vcpu_memslots_id(vcpu) ((vcpu)->arch.hflags & HF_SMM_MASK ? 1 : 0)
2128 # define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, (role).smm)
2129 #else
2130 # define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, 0)
2131 #endif
2132 
2133 #define KVM_ARCH_WANT_MMU_NOTIFIER
2134 
2135 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
2136 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
2137 int kvm_cpu_has_extint(struct kvm_vcpu *v);
2138 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
2139 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
2140 void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event);
2141 
2142 int kvm_pv_send_ipi(struct kvm *kvm, unsigned long ipi_bitmap_low,
2143 		    unsigned long ipi_bitmap_high, u32 min,
2144 		    unsigned long icr, int op_64_bit);
2145 
2146 int kvm_add_user_return_msr(u32 msr);
2147 int kvm_find_user_return_msr(u32 msr);
2148 int kvm_set_user_return_msr(unsigned index, u64 val, u64 mask);
2149 
kvm_is_supported_user_return_msr(u32 msr)2150 static inline bool kvm_is_supported_user_return_msr(u32 msr)
2151 {
2152 	return kvm_find_user_return_msr(msr) >= 0;
2153 }
2154 
2155 u64 kvm_scale_tsc(u64 tsc, u64 ratio);
2156 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc);
2157 u64 kvm_calc_nested_tsc_offset(u64 l1_offset, u64 l2_offset, u64 l2_multiplier);
2158 u64 kvm_calc_nested_tsc_multiplier(u64 l1_multiplier, u64 l2_multiplier);
2159 
2160 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu);
2161 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
2162 
2163 void kvm_make_scan_ioapic_request(struct kvm *kvm);
2164 void kvm_make_scan_ioapic_request_mask(struct kvm *kvm,
2165 				       unsigned long *vcpu_bitmap);
2166 
2167 bool kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
2168 				     struct kvm_async_pf *work);
2169 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
2170 				 struct kvm_async_pf *work);
2171 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
2172 			       struct kvm_async_pf *work);
2173 void kvm_arch_async_page_present_queued(struct kvm_vcpu *vcpu);
2174 bool kvm_arch_can_dequeue_async_page_present(struct kvm_vcpu *vcpu);
2175 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
2176 
2177 int kvm_skip_emulated_instruction(struct kvm_vcpu *vcpu);
2178 int kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
2179 void __kvm_request_immediate_exit(struct kvm_vcpu *vcpu);
2180 
2181 void __user *__x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa,
2182 				     u32 size);
2183 bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu);
2184 bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu);
2185 
2186 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
2187 			     struct kvm_vcpu **dest_vcpu);
2188 
2189 void kvm_set_msi_irq(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e,
2190 		     struct kvm_lapic_irq *irq);
2191 
kvm_irq_is_postable(struct kvm_lapic_irq * irq)2192 static inline bool kvm_irq_is_postable(struct kvm_lapic_irq *irq)
2193 {
2194 	/* We can only post Fixed and LowPrio IRQs */
2195 	return (irq->delivery_mode == APIC_DM_FIXED ||
2196 		irq->delivery_mode == APIC_DM_LOWEST);
2197 }
2198 
kvm_arch_vcpu_blocking(struct kvm_vcpu * vcpu)2199 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu)
2200 {
2201 	static_call_cond(kvm_x86_vcpu_blocking)(vcpu);
2202 }
2203 
kvm_arch_vcpu_unblocking(struct kvm_vcpu * vcpu)2204 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu)
2205 {
2206 	static_call_cond(kvm_x86_vcpu_unblocking)(vcpu);
2207 }
2208 
kvm_cpu_get_apicid(int mps_cpu)2209 static inline int kvm_cpu_get_apicid(int mps_cpu)
2210 {
2211 #ifdef CONFIG_X86_LOCAL_APIC
2212 	return default_cpu_present_to_apicid(mps_cpu);
2213 #else
2214 	WARN_ON_ONCE(1);
2215 	return BAD_APICID;
2216 #endif
2217 }
2218 
2219 int memslot_rmap_alloc(struct kvm_memory_slot *slot, unsigned long npages);
2220 
2221 #define KVM_CLOCK_VALID_FLAGS						\
2222 	(KVM_CLOCK_TSC_STABLE | KVM_CLOCK_REALTIME | KVM_CLOCK_HOST_TSC)
2223 
2224 #define KVM_X86_VALID_QUIRKS			\
2225 	(KVM_X86_QUIRK_LINT0_REENABLED |	\
2226 	 KVM_X86_QUIRK_CD_NW_CLEARED |		\
2227 	 KVM_X86_QUIRK_LAPIC_MMIO_HOLE |	\
2228 	 KVM_X86_QUIRK_OUT_7E_INC_RIP |		\
2229 	 KVM_X86_QUIRK_MISC_ENABLE_NO_MWAIT |	\
2230 	 KVM_X86_QUIRK_FIX_HYPERCALL_INSN |	\
2231 	 KVM_X86_QUIRK_MWAIT_NEVER_UD_FAULTS)
2232 
2233 /*
2234  * KVM previously used a u32 field in kvm_run to indicate the hypercall was
2235  * initiated from long mode. KVM now sets bit 0 to indicate long mode, but the
2236  * remaining 31 lower bits must be 0 to preserve ABI.
2237  */
2238 #define KVM_EXIT_HYPERCALL_MBZ		GENMASK_ULL(31, 1)
2239 
2240 #endif /* _ASM_X86_KVM_HOST_H */
2241