1 /* SPDX-License-Identifier: GPL-2.0-only */
3 * Kernel-based Virtual Machine driver for Linux
5 * This header defines architecture specific interfaces, x86 version
8 #ifndef _ASM_X86_KVM_HOST_H
9 #define _ASM_X86_KVM_HOST_H
11 #include <linux/types.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>
19 #include <linux/kvm.h>
20 #include <linux/kvm_para.h>
21 #include <linux/kvm_types.h>
22 #include <linux/perf_event.h>
23 #include <linux/pvclock_gtod.h>
24 #include <linux/clocksource.h>
25 #include <linux/irqbypass.h>
26 #include <linux/hyperv.h>
29 #include <asm/pvclock-abi.h>
32 #include <asm/msr-index.h>
34 #include <asm/kvm_page_track.h>
35 #include <asm/kvm_vcpu_regs.h>
36 #include <asm/hyperv-tlfs.h>
38 #define __KVM_HAVE_ARCH_VCPU_DEBUGFS
40 #define KVM_MAX_VCPUS 288
41 #define KVM_SOFT_MAX_VCPUS 240
42 #define KVM_MAX_VCPU_ID 1023
43 /* memory slots that are not exposed to userspace */
44 #define KVM_PRIVATE_MEM_SLOTS 3
46 #define KVM_HALT_POLL_NS_DEFAULT 200000
48 #define KVM_IRQCHIP_NUM_PINS KVM_IOAPIC_NUM_PINS
50 #define KVM_DIRTY_LOG_MANUAL_CAPS (KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE | \
51 KVM_DIRTY_LOG_INITIALLY_SET)
53 #define KVM_BUS_LOCK_DETECTION_VALID_MODE (KVM_BUS_LOCK_DETECTION_OFF | \
54 KVM_BUS_LOCK_DETECTION_EXIT)
56 /* x86-specific vcpu->requests bit members */
57 #define KVM_REQ_MIGRATE_TIMER KVM_ARCH_REQ(0)
58 #define KVM_REQ_REPORT_TPR_ACCESS KVM_ARCH_REQ(1)
59 #define KVM_REQ_TRIPLE_FAULT KVM_ARCH_REQ(2)
60 #define KVM_REQ_MMU_SYNC KVM_ARCH_REQ(3)
61 #define KVM_REQ_CLOCK_UPDATE KVM_ARCH_REQ(4)
62 #define KVM_REQ_LOAD_MMU_PGD KVM_ARCH_REQ(5)
63 #define KVM_REQ_EVENT KVM_ARCH_REQ(6)
64 #define KVM_REQ_APF_HALT KVM_ARCH_REQ(7)
65 #define KVM_REQ_STEAL_UPDATE KVM_ARCH_REQ(8)
66 #define KVM_REQ_NMI KVM_ARCH_REQ(9)
67 #define KVM_REQ_PMU KVM_ARCH_REQ(10)
68 #define KVM_REQ_PMI KVM_ARCH_REQ(11)
69 #define KVM_REQ_SMI KVM_ARCH_REQ(12)
70 #define KVM_REQ_MASTERCLOCK_UPDATE KVM_ARCH_REQ(13)
71 #define KVM_REQ_MCLOCK_INPROGRESS \
72 KVM_ARCH_REQ_FLAGS(14, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
73 #define KVM_REQ_SCAN_IOAPIC \
74 KVM_ARCH_REQ_FLAGS(15, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
75 #define KVM_REQ_GLOBAL_CLOCK_UPDATE KVM_ARCH_REQ(16)
76 #define KVM_REQ_APIC_PAGE_RELOAD \
77 KVM_ARCH_REQ_FLAGS(17, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
78 #define KVM_REQ_HV_CRASH KVM_ARCH_REQ(18)
79 #define KVM_REQ_IOAPIC_EOI_EXIT KVM_ARCH_REQ(19)
80 #define KVM_REQ_HV_RESET KVM_ARCH_REQ(20)
81 #define KVM_REQ_HV_EXIT KVM_ARCH_REQ(21)
82 #define KVM_REQ_HV_STIMER KVM_ARCH_REQ(22)
83 #define KVM_REQ_LOAD_EOI_EXITMAP KVM_ARCH_REQ(23)
84 #define KVM_REQ_GET_NESTED_STATE_PAGES KVM_ARCH_REQ(24)
85 #define KVM_REQ_APICV_UPDATE \
86 KVM_ARCH_REQ_FLAGS(25, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
87 #define KVM_REQ_TLB_FLUSH_CURRENT KVM_ARCH_REQ(26)
88 #define KVM_REQ_HV_TLB_FLUSH \
89 KVM_ARCH_REQ_FLAGS(27, KVM_REQUEST_NO_WAKEUP)
90 #define KVM_REQ_APF_READY KVM_ARCH_REQ(28)
91 #define KVM_REQ_MSR_FILTER_CHANGED KVM_ARCH_REQ(29)
92 #define KVM_REQ_UPDATE_CPU_DIRTY_LOGGING \
93 KVM_ARCH_REQ_FLAGS(30, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
95 #define CR0_RESERVED_BITS \
96 (~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \
97 | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \
98 | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG))
100 #define CR4_RESERVED_BITS \
101 (~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\
102 | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE \
103 | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR | X86_CR4_PCIDE \
104 | X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_FSGSBASE \
105 | X86_CR4_OSXMMEXCPT | X86_CR4_LA57 | X86_CR4_VMXE \
106 | X86_CR4_SMAP | X86_CR4_PKE | X86_CR4_UMIP))
108 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR)
112 #define INVALID_PAGE (~(hpa_t)0)
113 #define VALID_PAGE(x) ((x) != INVALID_PAGE)
115 #define UNMAPPED_GVA (~(gpa_t)0)
116 #define INVALID_GPA (~(gpa_t)0)
118 /* KVM Hugepage definitions for x86 */
119 #define KVM_MAX_HUGEPAGE_LEVEL PG_LEVEL_1G
120 #define KVM_NR_PAGE_SIZES (KVM_MAX_HUGEPAGE_LEVEL - PG_LEVEL_4K + 1)
121 #define KVM_HPAGE_GFN_SHIFT(x) (((x) - 1) * 9)
122 #define KVM_HPAGE_SHIFT(x) (PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
123 #define KVM_HPAGE_SIZE(x) (1UL << KVM_HPAGE_SHIFT(x))
124 #define KVM_HPAGE_MASK(x) (~(KVM_HPAGE_SIZE(x) - 1))
125 #define KVM_PAGES_PER_HPAGE(x) (KVM_HPAGE_SIZE(x) / PAGE_SIZE)
127 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
129 /* KVM_HPAGE_GFN_SHIFT(PG_LEVEL_4K) must be 0. */
130 return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
131 (base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
134 #define KVM_PERMILLE_MMU_PAGES 20
135 #define KVM_MIN_ALLOC_MMU_PAGES 64UL
136 #define KVM_MMU_HASH_SHIFT 12
137 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
138 #define KVM_MIN_FREE_MMU_PAGES 5
139 #define KVM_REFILL_PAGES 25
140 #define KVM_MAX_CPUID_ENTRIES 256
141 #define KVM_NR_FIXED_MTRR_REGION 88
142 #define KVM_NR_VAR_MTRR 8
144 #define ASYNC_PF_PER_VCPU 64
147 VCPU_REGS_RAX = __VCPU_REGS_RAX,
148 VCPU_REGS_RCX = __VCPU_REGS_RCX,
149 VCPU_REGS_RDX = __VCPU_REGS_RDX,
150 VCPU_REGS_RBX = __VCPU_REGS_RBX,
151 VCPU_REGS_RSP = __VCPU_REGS_RSP,
152 VCPU_REGS_RBP = __VCPU_REGS_RBP,
153 VCPU_REGS_RSI = __VCPU_REGS_RSI,
154 VCPU_REGS_RDI = __VCPU_REGS_RDI,
156 VCPU_REGS_R8 = __VCPU_REGS_R8,
157 VCPU_REGS_R9 = __VCPU_REGS_R9,
158 VCPU_REGS_R10 = __VCPU_REGS_R10,
159 VCPU_REGS_R11 = __VCPU_REGS_R11,
160 VCPU_REGS_R12 = __VCPU_REGS_R12,
161 VCPU_REGS_R13 = __VCPU_REGS_R13,
162 VCPU_REGS_R14 = __VCPU_REGS_R14,
163 VCPU_REGS_R15 = __VCPU_REGS_R15,
168 VCPU_EXREG_PDPTR = NR_VCPU_REGS,
174 VCPU_EXREG_EXIT_INFO_1,
175 VCPU_EXREG_EXIT_INFO_2,
189 enum exit_fastpath_completion {
191 EXIT_FASTPATH_REENTER_GUEST,
192 EXIT_FASTPATH_EXIT_HANDLED,
194 typedef enum exit_fastpath_completion fastpath_t;
196 struct x86_emulate_ctxt;
197 struct x86_exception;
199 enum x86_intercept_stage;
201 #define KVM_NR_DB_REGS 4
203 #define DR6_BD (1 << 13)
204 #define DR6_BS (1 << 14)
205 #define DR6_BT (1 << 15)
206 #define DR6_RTM (1 << 16)
208 * DR6_ACTIVE_LOW combines fixed-1 and active-low bits.
209 * We can regard all the bits in DR6_FIXED_1 as active_low bits;
210 * they will never be 0 for now, but when they are defined
211 * in the future it will require no code change.
213 * DR6_ACTIVE_LOW is also used as the init/reset value for DR6.
215 #define DR6_ACTIVE_LOW 0xffff0ff0
216 #define DR6_VOLATILE 0x0001e00f
217 #define DR6_FIXED_1 (DR6_ACTIVE_LOW & ~DR6_VOLATILE)
219 #define DR7_BP_EN_MASK 0x000000ff
220 #define DR7_GE (1 << 9)
221 #define DR7_GD (1 << 13)
222 #define DR7_FIXED_1 0x00000400
223 #define DR7_VOLATILE 0xffff2bff
225 #define KVM_GUESTDBG_VALID_MASK \
226 (KVM_GUESTDBG_ENABLE | \
227 KVM_GUESTDBG_SINGLESTEP | \
228 KVM_GUESTDBG_USE_HW_BP | \
229 KVM_GUESTDBG_USE_SW_BP | \
230 KVM_GUESTDBG_INJECT_BP | \
231 KVM_GUESTDBG_INJECT_DB)
234 #define PFERR_PRESENT_BIT 0
235 #define PFERR_WRITE_BIT 1
236 #define PFERR_USER_BIT 2
237 #define PFERR_RSVD_BIT 3
238 #define PFERR_FETCH_BIT 4
239 #define PFERR_PK_BIT 5
240 #define PFERR_SGX_BIT 15
241 #define PFERR_GUEST_FINAL_BIT 32
242 #define PFERR_GUEST_PAGE_BIT 33
244 #define PFERR_PRESENT_MASK (1U << PFERR_PRESENT_BIT)
245 #define PFERR_WRITE_MASK (1U << PFERR_WRITE_BIT)
246 #define PFERR_USER_MASK (1U << PFERR_USER_BIT)
247 #define PFERR_RSVD_MASK (1U << PFERR_RSVD_BIT)
248 #define PFERR_FETCH_MASK (1U << PFERR_FETCH_BIT)
249 #define PFERR_PK_MASK (1U << PFERR_PK_BIT)
250 #define PFERR_SGX_MASK (1U << PFERR_SGX_BIT)
251 #define PFERR_GUEST_FINAL_MASK (1ULL << PFERR_GUEST_FINAL_BIT)
252 #define PFERR_GUEST_PAGE_MASK (1ULL << PFERR_GUEST_PAGE_BIT)
254 #define PFERR_NESTED_GUEST_PAGE (PFERR_GUEST_PAGE_MASK | \
258 /* apic attention bits */
259 #define KVM_APIC_CHECK_VAPIC 0
261 * The following bit is set with PV-EOI, unset on EOI.
262 * We detect PV-EOI changes by guest by comparing
263 * this bit with PV-EOI in guest memory.
264 * See the implementation in apic_update_pv_eoi.
266 #define KVM_APIC_PV_EOI_PENDING 1
268 struct kvm_kernel_irq_routing_entry;
271 * the pages used as guest page table on soft mmu are tracked by
272 * kvm_memory_slot.arch.gfn_track which is 16 bits, so the role bits used
273 * by indirect shadow page can not be more than 15 bits.
275 * Currently, we used 14 bits that are @level, @gpte_is_8_bytes, @quadrant, @access,
276 * @nxe, @cr0_wp, @smep_andnot_wp and @smap_andnot_wp.
278 union kvm_mmu_page_role {
282 unsigned gpte_is_8_bytes:1;
289 unsigned smep_andnot_wp:1;
290 unsigned smap_andnot_wp:1;
291 unsigned ad_disabled:1;
292 unsigned guest_mode:1;
296 * This is left at the top of the word so that
297 * kvm_memslots_for_spte_role can extract it with a
298 * simple shift. While there is room, give it a whole
299 * byte so it is also faster to load it from memory.
305 union kvm_mmu_extended_role {
307 * This structure complements kvm_mmu_page_role caching everything needed for
308 * MMU configuration. If nothing in both these structures changed, MMU
309 * re-configuration can be skipped. @valid bit is set on first usage so we don't
310 * treat all-zero structure as valid data.
314 unsigned int valid:1;
315 unsigned int execonly:1;
316 unsigned int cr0_pg:1;
317 unsigned int cr4_pae:1;
318 unsigned int cr4_pse:1;
319 unsigned int cr4_pke:1;
320 unsigned int cr4_smap:1;
321 unsigned int cr4_smep:1;
322 unsigned int maxphyaddr:6;
329 union kvm_mmu_page_role base;
330 union kvm_mmu_extended_role ext;
334 struct kvm_rmap_head {
338 struct kvm_pio_request {
339 unsigned long linear_rip;
346 #define PT64_ROOT_MAX_LEVEL 5
348 struct rsvd_bits_validate {
349 u64 rsvd_bits_mask[2][PT64_ROOT_MAX_LEVEL];
353 struct kvm_mmu_root_info {
358 #define KVM_MMU_ROOT_INFO_INVALID \
359 ((struct kvm_mmu_root_info) { .pgd = INVALID_PAGE, .hpa = INVALID_PAGE })
361 #define KVM_MMU_NUM_PREV_ROOTS 3
363 #define KVM_HAVE_MMU_RWLOCK
368 * x86 supports 4 paging modes (5-level 64-bit, 4-level 64-bit, 3-level 32-bit,
369 * and 2-level 32-bit). The kvm_mmu structure abstracts the details of the
373 unsigned long (*get_guest_pgd)(struct kvm_vcpu *vcpu);
374 u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
375 int (*page_fault)(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u32 err,
377 void (*inject_page_fault)(struct kvm_vcpu *vcpu,
378 struct x86_exception *fault);
379 gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gpa_t gva_or_gpa,
380 u32 access, struct x86_exception *exception);
381 gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
382 struct x86_exception *exception);
383 int (*sync_page)(struct kvm_vcpu *vcpu,
384 struct kvm_mmu_page *sp);
385 void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva, hpa_t root_hpa);
388 union kvm_mmu_role mmu_role;
390 u8 shadow_root_level;
393 struct kvm_mmu_root_info prev_roots[KVM_MMU_NUM_PREV_ROOTS];
396 * Bitmap; bit set = permission fault
397 * Byte index: page fault error code [4:1]
398 * Bit index: pte permissions in ACC_* format
403 * The pkru_mask indicates if protection key checks are needed. It
404 * consists of 16 domains indexed by page fault error code bits [4:1],
405 * with PFEC.RSVD replaced by ACC_USER_MASK from the page tables.
406 * Each domain has 2 bits which are ANDed with AD and WD from PKRU.
414 * check zero bits on shadow page table entries, these
415 * bits include not only hardware reserved bits but also
416 * the bits spte never used.
418 struct rsvd_bits_validate shadow_zero_check;
420 struct rsvd_bits_validate guest_rsvd_check;
422 /* Can have large pages at levels 2..last_nonleaf_level-1. */
423 u8 last_nonleaf_level;
427 u64 pdptrs[4]; /* pae */
430 struct kvm_tlb_range {
445 struct perf_event *perf_event;
446 struct kvm_vcpu *vcpu;
448 * eventsel value for general purpose counters,
449 * ctrl value for fixed counters.
455 unsigned nr_arch_gp_counters;
456 unsigned nr_arch_fixed_counters;
457 unsigned available_event_types;
462 u64 counter_bitmask[2];
463 u64 global_ctrl_mask;
464 u64 global_ovf_ctrl_mask;
467 struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
468 struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED];
469 struct irq_work irq_work;
470 DECLARE_BITMAP(reprogram_pmi, X86_PMC_IDX_MAX);
471 DECLARE_BITMAP(all_valid_pmc_idx, X86_PMC_IDX_MAX);
472 DECLARE_BITMAP(pmc_in_use, X86_PMC_IDX_MAX);
475 * The gate to release perf_events not marked in
476 * pmc_in_use only once in a vcpu time slice.
481 * The total number of programmed perf_events and it helps to avoid
482 * redundant check before cleanup if guest don't use vPMU at all.
490 KVM_DEBUGREG_BP_ENABLED = 1,
491 KVM_DEBUGREG_WONT_EXIT = 2,
492 KVM_DEBUGREG_RELOAD = 4,
495 struct kvm_mtrr_range {
498 struct list_head node;
502 struct kvm_mtrr_range var_ranges[KVM_NR_VAR_MTRR];
503 mtrr_type fixed_ranges[KVM_NR_FIXED_MTRR_REGION];
506 struct list_head head;
509 /* Hyper-V SynIC timer */
510 struct kvm_vcpu_hv_stimer {
511 struct hrtimer timer;
513 union hv_stimer_config config;
516 struct hv_message msg;
520 /* Hyper-V synthetic interrupt controller (SynIC)*/
521 struct kvm_vcpu_hv_synic {
526 atomic64_t sint[HV_SYNIC_SINT_COUNT];
527 atomic_t sint_to_gsi[HV_SYNIC_SINT_COUNT];
528 DECLARE_BITMAP(auto_eoi_bitmap, 256);
529 DECLARE_BITMAP(vec_bitmap, 256);
531 bool dont_zero_synic_pages;
534 /* Hyper-V per vcpu emulation context */
536 struct kvm_vcpu *vcpu;
540 struct kvm_vcpu_hv_synic synic;
541 struct kvm_hyperv_exit exit;
542 struct kvm_vcpu_hv_stimer stimer[HV_SYNIC_STIMER_COUNT];
543 DECLARE_BITMAP(stimer_pending_bitmap, HV_SYNIC_STIMER_COUNT);
547 /* Xen HVM per vcpu emulation context */
548 struct kvm_vcpu_xen {
550 u32 current_runstate;
552 bool vcpu_time_info_set;
554 struct gfn_to_hva_cache vcpu_info_cache;
555 struct gfn_to_hva_cache vcpu_time_info_cache;
556 struct gfn_to_hva_cache runstate_cache;
558 u64 runstate_entry_time;
559 u64 runstate_times[4];
562 struct kvm_vcpu_arch {
564 * rip and regs accesses must go through
565 * kvm_{register,rip}_{read,write} functions.
567 unsigned long regs[NR_VCPU_REGS];
572 unsigned long cr0_guest_owned_bits;
576 unsigned long cr4_guest_owned_bits;
577 unsigned long cr4_guest_rsvd_bits;
584 struct kvm_lapic *apic; /* kernel irqchip context */
586 bool load_eoi_exitmap_pending;
587 DECLARE_BITMAP(ioapic_handled_vectors, 256);
588 unsigned long apic_attention;
589 int32_t apic_arb_prio;
591 u64 ia32_misc_enable_msr;
594 bool tpr_access_reporting;
597 u64 microcode_version;
598 u64 arch_capabilities;
599 u64 perf_capabilities;
602 * Paging state of the vcpu
604 * If the vcpu runs in guest mode with two level paging this still saves
605 * the paging mode of the l1 guest. This context is always used to
610 /* Non-nested MMU for L1 */
611 struct kvm_mmu root_mmu;
613 /* L1 MMU when running nested */
614 struct kvm_mmu guest_mmu;
617 * Paging state of an L2 guest (used for nested npt)
619 * This context will save all necessary information to walk page tables
620 * of an L2 guest. This context is only initialized for page table
621 * walking and not for faulting since we never handle l2 page faults on
624 struct kvm_mmu nested_mmu;
627 * Pointer to the mmu context currently used for
628 * gva_to_gpa translations.
630 struct kvm_mmu *walk_mmu;
632 struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
633 struct kvm_mmu_memory_cache mmu_shadow_page_cache;
634 struct kvm_mmu_memory_cache mmu_gfn_array_cache;
635 struct kvm_mmu_memory_cache mmu_page_header_cache;
638 * QEMU userspace and the guest each have their own FPU state.
639 * In vcpu_run, we switch between the user and guest FPU contexts.
640 * While running a VCPU, the VCPU thread will have the guest FPU
643 * Note that while the PKRU state lives inside the fpu registers,
644 * it is switched out separately at VMENTER and VMEXIT time. The
645 * "guest_fpu" state here contains the guest FPU context, with the
648 struct fpu *user_fpu;
649 struct fpu *guest_fpu;
652 u64 guest_supported_xcr0;
654 struct kvm_pio_request pio;
656 void *guest_ins_data;
658 u8 event_exit_inst_len;
660 struct kvm_queued_exception {
666 unsigned long payload;
671 struct kvm_queued_interrupt {
677 int halt_request; /* real mode on Intel only */
680 struct kvm_cpuid_entry2 *cpuid_entries;
682 u64 reserved_gpa_bits;
686 /* emulate context */
688 struct x86_emulate_ctxt *emulate_ctxt;
689 bool emulate_regs_need_sync_to_vcpu;
690 bool emulate_regs_need_sync_from_vcpu;
691 int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
694 struct pvclock_vcpu_time_info hv_clock;
695 unsigned int hw_tsc_khz;
696 struct gfn_to_hva_cache pv_time;
697 bool pv_time_enabled;
698 /* set guest stopped flag in pvclock flags field */
699 bool pvclock_set_guest_stopped_request;
705 struct gfn_to_pfn_cache cache;
712 u64 tsc_offset_adjustment;
715 u64 this_tsc_generation;
717 bool tsc_always_catchup;
718 s8 virtual_tsc_shift;
719 u32 virtual_tsc_mult;
721 s64 ia32_tsc_adjust_msr;
722 u64 msr_ia32_power_ctl;
723 u64 tsc_scaling_ratio;
725 atomic_t nmi_queued; /* unprocessed asynchronous NMIs */
726 unsigned nmi_pending; /* NMI queued after currently running handler */
727 bool nmi_injected; /* Trying to inject an NMI this entry */
728 bool smi_pending; /* SMI queued after currently running handler */
730 struct kvm_mtrr mtrr_state;
733 unsigned switch_db_regs;
734 unsigned long db[KVM_NR_DB_REGS];
737 unsigned long eff_db[KVM_NR_DB_REGS];
738 unsigned long guest_debug_dr7;
739 u64 msr_platform_info;
740 u64 msr_misc_features_enables;
748 /* Cache MMIO info */
750 unsigned mmio_access;
756 /* used for guest single stepping over the given code position */
757 unsigned long singlestep_rip;
760 struct kvm_vcpu_hv *hyperv;
761 struct kvm_vcpu_xen xen;
763 cpumask_var_t wbinvd_dirty_mask;
765 unsigned long last_retry_eip;
766 unsigned long last_retry_addr;
770 gfn_t gfns[ASYNC_PF_PER_VCPU];
771 struct gfn_to_hva_cache data;
772 u64 msr_en_val; /* MSR_KVM_ASYNC_PF_EN */
773 u64 msr_int_val; /* MSR_KVM_ASYNC_PF_INT */
778 unsigned long nested_apf_token;
779 bool delivery_as_pf_vmexit;
780 bool pageready_pending;
783 /* OSVW MSRs (AMD only) */
791 struct gfn_to_hva_cache data;
794 u64 msr_kvm_poll_control;
797 * Indicates the guest is trying to write a gfn that contains one or
798 * more of the PTEs used to translate the write itself, i.e. the access
799 * is changing its own translation in the guest page tables. KVM exits
800 * to userspace if emulation of the faulting instruction fails and this
801 * flag is set, as KVM cannot make forward progress.
803 * If emulation fails for a write to guest page tables, KVM unprotects
804 * (zaps) the shadow page for the target gfn and resumes the guest to
805 * retry the non-emulatable instruction (on hardware). Unprotecting the
806 * gfn doesn't allow forward progress for a self-changing access because
807 * doing so also zaps the translation for the gfn, i.e. retrying the
808 * instruction will hit a !PRESENT fault, which results in a new shadow
809 * page and sends KVM back to square one.
811 bool write_fault_to_shadow_pgtable;
813 /* set at EPT violation at this point */
814 unsigned long exit_qualification;
816 /* pv related host specific info */
821 int pending_ioapic_eoi;
822 int pending_external_vector;
824 /* be preempted when it's in kernel-mode(cpl=0) */
825 bool preempted_in_kernel;
827 /* Flush the L1 Data cache for L1TF mitigation on VMENTER */
830 /* Host CPU on which VM-entry was most recently attempted */
831 unsigned int last_vmentry_cpu;
833 /* AMD MSRC001_0015 Hardware Configuration */
836 /* pv related cpuid info */
839 * value of the eax register in the KVM_CPUID_FEATURES CPUID
845 * indicates whether pv emulation should be disabled if features
846 * are not present in the guest's cpuid
851 /* Protected Guests */
852 bool guest_state_protected;
855 struct kvm_lpage_info {
859 struct kvm_arch_memory_slot {
860 struct kvm_rmap_head *rmap[KVM_NR_PAGE_SIZES];
861 struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
862 unsigned short *gfn_track[KVM_PAGE_TRACK_MAX];
866 * We use as the mode the number of bits allocated in the LDR for the
867 * logical processor ID. It happens that these are all powers of two.
868 * This makes it is very easy to detect cases where the APICs are
869 * configured for multiple modes; in that case, we cannot use the map and
870 * hence cannot use kvm_irq_delivery_to_apic_fast either.
872 #define KVM_APIC_MODE_XAPIC_CLUSTER 4
873 #define KVM_APIC_MODE_XAPIC_FLAT 8
874 #define KVM_APIC_MODE_X2APIC 16
876 struct kvm_apic_map {
881 struct kvm_lapic *xapic_flat_map[8];
882 struct kvm_lapic *xapic_cluster_map[16][4];
884 struct kvm_lapic *phys_map[];
887 /* Hyper-V synthetic debugger (SynDbg)*/
888 struct kvm_hv_syndbg {
899 /* Current state of Hyper-V TSC page clocksource */
900 enum hv_tsc_page_status {
901 /* TSC page was not set up or disabled */
902 HV_TSC_PAGE_UNSET = 0,
903 /* TSC page MSR was written by the guest, update pending */
904 HV_TSC_PAGE_GUEST_CHANGED,
905 /* TSC page MSR was written by KVM userspace, update pending */
906 HV_TSC_PAGE_HOST_CHANGED,
907 /* TSC page was properly set up and is currently active */
909 /* TSC page is currently being updated and therefore is inactive */
910 HV_TSC_PAGE_UPDATING,
911 /* TSC page was set up with an inaccessible GPA */
915 /* Hyper-V emulation context */
917 struct mutex hv_lock;
921 enum hv_tsc_page_status hv_tsc_page_status;
923 /* Hyper-v based guest crash (NT kernel bugcheck) parameters */
924 u64 hv_crash_param[HV_X64_MSR_CRASH_PARAMS];
927 struct ms_hyperv_tsc_page tsc_ref;
929 struct idr conn_to_evt;
931 u64 hv_reenlightenment_control;
932 u64 hv_tsc_emulation_control;
933 u64 hv_tsc_emulation_status;
935 /* How many vCPUs have VP index != vCPU index */
936 atomic_t num_mismatched_vp_indexes;
938 struct hv_partition_assist_pg *hv_pa_pg;
939 struct kvm_hv_syndbg hv_syndbg;
942 struct msr_bitmap_range {
946 unsigned long *bitmap;
949 /* Xen emulation context */
954 struct gfn_to_hva_cache shinfo_cache;
957 enum kvm_irqchip_mode {
959 KVM_IRQCHIP_KERNEL, /* created with KVM_CREATE_IRQCHIP */
960 KVM_IRQCHIP_SPLIT, /* created with KVM_CAP_SPLIT_IRQCHIP */
963 struct kvm_x86_msr_filter {
965 bool default_allow:1;
966 struct msr_bitmap_range ranges[16];
969 #define APICV_INHIBIT_REASON_DISABLE 0
970 #define APICV_INHIBIT_REASON_HYPERV 1
971 #define APICV_INHIBIT_REASON_NESTED 2
972 #define APICV_INHIBIT_REASON_IRQWIN 3
973 #define APICV_INHIBIT_REASON_PIT_REINJ 4
974 #define APICV_INHIBIT_REASON_X2APIC 5
977 unsigned long n_used_mmu_pages;
978 unsigned long n_requested_mmu_pages;
979 unsigned long n_max_mmu_pages;
980 unsigned int indirect_shadow_pages;
982 struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
983 struct list_head active_mmu_pages;
984 struct list_head zapped_obsolete_pages;
985 struct list_head lpage_disallowed_mmu_pages;
986 struct kvm_page_track_notifier_node mmu_sp_tracker;
987 struct kvm_page_track_notifier_head track_notifier_head;
989 struct list_head assigned_dev_head;
990 struct iommu_domain *iommu_domain;
991 bool iommu_noncoherent;
992 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA
993 atomic_t noncoherent_dma_count;
994 #define __KVM_HAVE_ARCH_ASSIGNED_DEVICE
995 atomic_t assigned_device_count;
996 struct kvm_pic *vpic;
997 struct kvm_ioapic *vioapic;
998 struct kvm_pit *vpit;
999 atomic_t vapics_in_nmi_mode;
1000 struct mutex apic_map_lock;
1001 struct kvm_apic_map __rcu *apic_map;
1002 atomic_t apic_map_dirty;
1004 bool apic_access_page_done;
1005 unsigned long apicv_inhibit_reasons;
1009 bool mwait_in_guest;
1011 bool pause_in_guest;
1012 bool cstate_in_guest;
1014 unsigned long irq_sources_bitmap;
1015 s64 kvmclock_offset;
1016 raw_spinlock_t tsc_write_lock;
1023 u64 cur_tsc_generation;
1024 int nr_vcpus_matched_tsc;
1026 spinlock_t pvclock_gtod_sync_lock;
1027 bool use_master_clock;
1028 u64 master_kernel_ns;
1029 u64 master_cycle_now;
1030 struct delayed_work kvmclock_update_work;
1031 struct delayed_work kvmclock_sync_work;
1033 struct kvm_xen_hvm_config xen_hvm_config;
1035 /* reads protected by irq_srcu, writes by irq_lock */
1036 struct hlist_head mask_notifier_list;
1038 struct kvm_hv hyperv;
1041 #ifdef CONFIG_KVM_MMU_AUDIT
1045 bool backwards_tsc_observed;
1046 bool boot_vcpu_runs_old_kvmclock;
1049 u64 disabled_quirks;
1050 int cpu_dirty_logging_count;
1052 enum kvm_irqchip_mode irqchip_mode;
1053 u8 nr_reserved_ioapic_pins;
1055 bool disabled_lapic_found;
1058 bool x2apic_broadcast_quirk_disabled;
1060 bool guest_can_read_msr_platform_info;
1061 bool exception_payload_enabled;
1063 bool bus_lock_detection_enabled;
1065 /* Deflect RDMSR and WRMSR to user space when they trigger a #GP */
1066 u32 user_space_msr_mask;
1067 struct kvm_x86_msr_filter __rcu *msr_filter;
1069 /* Guest can access the SGX PROVISIONKEY. */
1070 bool sgx_provisioning_allowed;
1072 struct kvm_pmu_event_filter __rcu *pmu_event_filter;
1073 struct task_struct *nx_lpage_recovery_thread;
1075 #ifdef CONFIG_X86_64
1077 * Whether the TDP MMU is enabled for this VM. This contains a
1078 * snapshot of the TDP MMU module parameter from when the VM was
1079 * created and remains unchanged for the life of the VM. If this is
1080 * true, TDP MMU handler functions will run for various MMU
1083 bool tdp_mmu_enabled;
1086 * List of struct kvm_mmu_pages being used as roots.
1087 * All struct kvm_mmu_pages in the list should have
1090 * For reads, this list is protected by:
1091 * the MMU lock in read mode + RCU or
1092 * the MMU lock in write mode
1094 * For writes, this list is protected by:
1095 * the MMU lock in read mode + the tdp_mmu_pages_lock or
1096 * the MMU lock in write mode
1098 * Roots will remain in the list until their tdp_mmu_root_count
1099 * drops to zero, at which point the thread that decremented the
1100 * count to zero should removed the root from the list and clean
1101 * it up, freeing the root after an RCU grace period.
1103 struct list_head tdp_mmu_roots;
1106 * List of struct kvmp_mmu_pages not being used as roots.
1107 * All struct kvm_mmu_pages in the list should have
1108 * tdp_mmu_page set and a tdp_mmu_root_count of 0.
1110 struct list_head tdp_mmu_pages;
1113 * Protects accesses to the following fields when the MMU lock
1114 * is held in read mode:
1115 * - tdp_mmu_roots (above)
1116 * - tdp_mmu_pages (above)
1117 * - the link field of struct kvm_mmu_pages used by the TDP MMU
1118 * - lpage_disallowed_mmu_pages
1119 * - the lpage_disallowed_link field of struct kvm_mmu_pages used
1121 * It is acceptable, but not necessary, to acquire this lock when
1122 * the thread holds the MMU lock in write mode.
1124 spinlock_t tdp_mmu_pages_lock;
1125 #endif /* CONFIG_X86_64 */
1128 struct kvm_vm_stat {
1129 ulong mmu_shadow_zapped;
1130 ulong mmu_pte_write;
1131 ulong mmu_pde_zapped;
1134 ulong mmu_cache_miss;
1136 ulong remote_tlb_flush;
1138 ulong nx_lpage_splits;
1139 ulong max_mmu_page_hash_collisions;
1142 struct kvm_vcpu_stat {
1152 u64 irq_window_exits;
1153 u64 nmi_window_exits;
1156 u64 halt_successful_poll;
1157 u64 halt_attempted_poll;
1158 u64 halt_poll_invalid;
1160 u64 request_irq_exits;
1162 u64 host_state_reload;
1165 u64 insn_emulation_fail;
1170 u64 halt_poll_success_ns;
1171 u64 halt_poll_fail_ns;
1173 u64 directed_yield_attempted;
1174 u64 directed_yield_successful;
1177 struct x86_instruction_info;
1180 bool host_initiated;
1185 struct kvm_lapic_irq {
1193 bool msi_redir_hint;
1196 static inline u16 kvm_lapic_irq_dest_mode(bool dest_mode_logical)
1198 return dest_mode_logical ? APIC_DEST_LOGICAL : APIC_DEST_PHYSICAL;
1201 struct kvm_x86_ops {
1202 int (*hardware_enable)(void);
1203 void (*hardware_disable)(void);
1204 void (*hardware_unsetup)(void);
1205 bool (*cpu_has_accelerated_tpr)(void);
1206 bool (*has_emulated_msr)(struct kvm *kvm, u32 index);
1207 void (*vcpu_after_set_cpuid)(struct kvm_vcpu *vcpu);
1209 unsigned int vm_size;
1210 int (*vm_init)(struct kvm *kvm);
1211 void (*vm_destroy)(struct kvm *kvm);
1213 /* Create, but do not attach this VCPU */
1214 int (*vcpu_create)(struct kvm_vcpu *vcpu);
1215 void (*vcpu_free)(struct kvm_vcpu *vcpu);
1216 void (*vcpu_reset)(struct kvm_vcpu *vcpu, bool init_event);
1218 void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
1219 void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
1220 void (*vcpu_put)(struct kvm_vcpu *vcpu);
1222 void (*update_exception_bitmap)(struct kvm_vcpu *vcpu);
1223 int (*get_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1224 int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1225 u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
1226 void (*get_segment)(struct kvm_vcpu *vcpu,
1227 struct kvm_segment *var, int seg);
1228 int (*get_cpl)(struct kvm_vcpu *vcpu);
1229 void (*set_segment)(struct kvm_vcpu *vcpu,
1230 struct kvm_segment *var, int seg);
1231 void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
1232 void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
1233 bool (*is_valid_cr4)(struct kvm_vcpu *vcpu, unsigned long cr0);
1234 void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
1235 int (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
1236 void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1237 void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1238 void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1239 void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1240 void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu);
1241 void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
1242 void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
1243 unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
1244 void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
1246 void (*tlb_flush_all)(struct kvm_vcpu *vcpu);
1247 void (*tlb_flush_current)(struct kvm_vcpu *vcpu);
1248 int (*tlb_remote_flush)(struct kvm *kvm);
1249 int (*tlb_remote_flush_with_range)(struct kvm *kvm,
1250 struct kvm_tlb_range *range);
1253 * Flush any TLB entries associated with the given GVA.
1254 * Does not need to flush GPA->HPA mappings.
1255 * Can potentially get non-canonical addresses through INVLPGs, which
1256 * the implementation may choose to ignore if appropriate.
1258 void (*tlb_flush_gva)(struct kvm_vcpu *vcpu, gva_t addr);
1261 * Flush any TLB entries created by the guest. Like tlb_flush_gva(),
1262 * does not need to flush GPA->HPA mappings.
1264 void (*tlb_flush_guest)(struct kvm_vcpu *vcpu);
1266 enum exit_fastpath_completion (*run)(struct kvm_vcpu *vcpu);
1267 int (*handle_exit)(struct kvm_vcpu *vcpu,
1268 enum exit_fastpath_completion exit_fastpath);
1269 int (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
1270 void (*update_emulated_instruction)(struct kvm_vcpu *vcpu);
1271 void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
1272 u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu);
1273 void (*patch_hypercall)(struct kvm_vcpu *vcpu,
1274 unsigned char *hypercall_addr);
1275 void (*set_irq)(struct kvm_vcpu *vcpu);
1276 void (*set_nmi)(struct kvm_vcpu *vcpu);
1277 void (*queue_exception)(struct kvm_vcpu *vcpu);
1278 void (*cancel_injection)(struct kvm_vcpu *vcpu);
1279 int (*interrupt_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1280 int (*nmi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1281 bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
1282 void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
1283 void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
1284 void (*enable_irq_window)(struct kvm_vcpu *vcpu);
1285 void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
1286 bool (*check_apicv_inhibit_reasons)(ulong bit);
1287 void (*pre_update_apicv_exec_ctrl)(struct kvm *kvm, bool activate);
1288 void (*refresh_apicv_exec_ctrl)(struct kvm_vcpu *vcpu);
1289 void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
1290 void (*hwapic_isr_update)(struct kvm_vcpu *vcpu, int isr);
1291 bool (*guest_apic_has_interrupt)(struct kvm_vcpu *vcpu);
1292 void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
1293 void (*set_virtual_apic_mode)(struct kvm_vcpu *vcpu);
1294 void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu);
1295 int (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
1296 int (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
1297 int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
1298 int (*set_identity_map_addr)(struct kvm *kvm, u64 ident_addr);
1299 u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
1301 void (*load_mmu_pgd)(struct kvm_vcpu *vcpu, hpa_t root_hpa,
1304 bool (*has_wbinvd_exit)(void);
1306 /* Returns actual tsc_offset set in active VMCS */
1307 u64 (*write_l1_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
1310 * Retrieve somewhat arbitrary exit information. Intended to be used
1311 * only from within tracepoints to avoid VMREADs when tracing is off.
1313 void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2,
1314 u32 *exit_int_info, u32 *exit_int_info_err_code);
1316 int (*check_intercept)(struct kvm_vcpu *vcpu,
1317 struct x86_instruction_info *info,
1318 enum x86_intercept_stage stage,
1319 struct x86_exception *exception);
1320 void (*handle_exit_irqoff)(struct kvm_vcpu *vcpu);
1322 void (*request_immediate_exit)(struct kvm_vcpu *vcpu);
1324 void (*sched_in)(struct kvm_vcpu *kvm, int cpu);
1327 * Size of the CPU's dirty log buffer, i.e. VMX's PML buffer. A zero
1328 * value indicates CPU dirty logging is unsupported or disabled.
1330 int cpu_dirty_log_size;
1331 void (*update_cpu_dirty_logging)(struct kvm_vcpu *vcpu);
1333 /* pmu operations of sub-arch */
1334 const struct kvm_pmu_ops *pmu_ops;
1335 const struct kvm_x86_nested_ops *nested_ops;
1338 * Architecture specific hooks for vCPU blocking due to
1340 * Returns for .pre_block():
1341 * - 0 means continue to block the vCPU.
1342 * - 1 means we cannot block the vCPU since some event
1343 * happens during this period, such as, 'ON' bit in
1344 * posted-interrupts descriptor is set.
1346 int (*pre_block)(struct kvm_vcpu *vcpu);
1347 void (*post_block)(struct kvm_vcpu *vcpu);
1349 void (*vcpu_blocking)(struct kvm_vcpu *vcpu);
1350 void (*vcpu_unblocking)(struct kvm_vcpu *vcpu);
1352 int (*update_pi_irte)(struct kvm *kvm, unsigned int host_irq,
1353 uint32_t guest_irq, bool set);
1354 void (*apicv_post_state_restore)(struct kvm_vcpu *vcpu);
1355 bool (*dy_apicv_has_pending_interrupt)(struct kvm_vcpu *vcpu);
1357 int (*set_hv_timer)(struct kvm_vcpu *vcpu, u64 guest_deadline_tsc,
1359 void (*cancel_hv_timer)(struct kvm_vcpu *vcpu);
1361 void (*setup_mce)(struct kvm_vcpu *vcpu);
1363 int (*smi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1364 int (*pre_enter_smm)(struct kvm_vcpu *vcpu, char *smstate);
1365 int (*pre_leave_smm)(struct kvm_vcpu *vcpu, const char *smstate);
1366 void (*enable_smi_window)(struct kvm_vcpu *vcpu);
1368 int (*mem_enc_op)(struct kvm *kvm, void __user *argp);
1369 int (*mem_enc_reg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1370 int (*mem_enc_unreg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1371 int (*vm_copy_enc_context_from)(struct kvm *kvm, unsigned int source_fd);
1373 int (*get_msr_feature)(struct kvm_msr_entry *entry);
1375 bool (*can_emulate_instruction)(struct kvm_vcpu *vcpu, void *insn, int insn_len);
1377 bool (*apic_init_signal_blocked)(struct kvm_vcpu *vcpu);
1378 int (*enable_direct_tlbflush)(struct kvm_vcpu *vcpu);
1380 void (*migrate_timers)(struct kvm_vcpu *vcpu);
1381 void (*msr_filter_changed)(struct kvm_vcpu *vcpu);
1382 int (*complete_emulated_msr)(struct kvm_vcpu *vcpu, int err);
1384 void (*vcpu_deliver_sipi_vector)(struct kvm_vcpu *vcpu, u8 vector);
1387 struct kvm_x86_nested_ops {
1388 int (*check_events)(struct kvm_vcpu *vcpu);
1389 bool (*hv_timer_pending)(struct kvm_vcpu *vcpu);
1390 void (*triple_fault)(struct kvm_vcpu *vcpu);
1391 int (*get_state)(struct kvm_vcpu *vcpu,
1392 struct kvm_nested_state __user *user_kvm_nested_state,
1393 unsigned user_data_size);
1394 int (*set_state)(struct kvm_vcpu *vcpu,
1395 struct kvm_nested_state __user *user_kvm_nested_state,
1396 struct kvm_nested_state *kvm_state);
1397 bool (*get_nested_state_pages)(struct kvm_vcpu *vcpu);
1398 int (*write_log_dirty)(struct kvm_vcpu *vcpu, gpa_t l2_gpa);
1400 int (*enable_evmcs)(struct kvm_vcpu *vcpu,
1401 uint16_t *vmcs_version);
1402 uint16_t (*get_evmcs_version)(struct kvm_vcpu *vcpu);
1405 struct kvm_x86_init_ops {
1406 int (*cpu_has_kvm_support)(void);
1407 int (*disabled_by_bios)(void);
1408 int (*check_processor_compatibility)(void);
1409 int (*hardware_setup)(void);
1411 struct kvm_x86_ops *runtime_ops;
1414 struct kvm_arch_async_pf {
1421 extern u64 __read_mostly host_efer;
1422 extern bool __read_mostly allow_smaller_maxphyaddr;
1423 extern struct kvm_x86_ops kvm_x86_ops;
1425 #define KVM_X86_OP(func) \
1426 DECLARE_STATIC_CALL(kvm_x86_##func, *(((struct kvm_x86_ops *)0)->func));
1427 #define KVM_X86_OP_NULL KVM_X86_OP
1428 #include <asm/kvm-x86-ops.h>
1430 static inline void kvm_ops_static_call_update(void)
1432 #define KVM_X86_OP(func) \
1433 static_call_update(kvm_x86_##func, kvm_x86_ops.func);
1434 #define KVM_X86_OP_NULL KVM_X86_OP
1435 #include <asm/kvm-x86-ops.h>
1438 #define __KVM_HAVE_ARCH_VM_ALLOC
1439 static inline struct kvm *kvm_arch_alloc_vm(void)
1441 return __vmalloc(kvm_x86_ops.vm_size, GFP_KERNEL_ACCOUNT | __GFP_ZERO);
1443 void kvm_arch_free_vm(struct kvm *kvm);
1445 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
1446 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
1448 if (kvm_x86_ops.tlb_remote_flush &&
1449 !static_call(kvm_x86_tlb_remote_flush)(kvm))
1455 int kvm_mmu_module_init(void);
1456 void kvm_mmu_module_exit(void);
1458 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
1459 int kvm_mmu_create(struct kvm_vcpu *vcpu);
1460 void kvm_mmu_init_vm(struct kvm *kvm);
1461 void kvm_mmu_uninit_vm(struct kvm *kvm);
1463 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
1464 void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
1465 struct kvm_memory_slot *memslot,
1467 void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
1468 const struct kvm_memory_slot *memslot);
1469 void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
1470 struct kvm_memory_slot *memslot);
1471 void kvm_mmu_zap_all(struct kvm *kvm);
1472 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, u64 gen);
1473 unsigned long kvm_mmu_calculate_default_mmu_pages(struct kvm *kvm);
1474 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned long kvm_nr_mmu_pages);
1476 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
1477 bool pdptrs_changed(struct kvm_vcpu *vcpu);
1479 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
1480 const void *val, int bytes);
1482 struct kvm_irq_mask_notifier {
1483 void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
1485 struct hlist_node link;
1488 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
1489 struct kvm_irq_mask_notifier *kimn);
1490 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
1491 struct kvm_irq_mask_notifier *kimn);
1492 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
1495 extern bool tdp_enabled;
1497 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
1499 /* control of guest tsc rate supported? */
1500 extern bool kvm_has_tsc_control;
1501 /* maximum supported tsc_khz for guests */
1502 extern u32 kvm_max_guest_tsc_khz;
1503 /* number of bits of the fractional part of the TSC scaling ratio */
1504 extern u8 kvm_tsc_scaling_ratio_frac_bits;
1505 /* maximum allowed value of TSC scaling ratio */
1506 extern u64 kvm_max_tsc_scaling_ratio;
1507 /* 1ull << kvm_tsc_scaling_ratio_frac_bits */
1508 extern u64 kvm_default_tsc_scaling_ratio;
1509 /* bus lock detection supported? */
1510 extern bool kvm_has_bus_lock_exit;
1512 extern u64 kvm_mce_cap_supported;
1515 * EMULTYPE_NO_DECODE - Set when re-emulating an instruction (after completing
1516 * userspace I/O) to indicate that the emulation context
1517 * should be reused as is, i.e. skip initialization of
1518 * emulation context, instruction fetch and decode.
1520 * EMULTYPE_TRAP_UD - Set when emulating an intercepted #UD from hardware.
1521 * Indicates that only select instructions (tagged with
1522 * EmulateOnUD) should be emulated (to minimize the emulator
1523 * attack surface). See also EMULTYPE_TRAP_UD_FORCED.
1525 * EMULTYPE_SKIP - Set when emulating solely to skip an instruction, i.e. to
1526 * decode the instruction length. For use *only* by
1527 * kvm_x86_ops.skip_emulated_instruction() implementations.
1529 * EMULTYPE_ALLOW_RETRY_PF - Set when the emulator should resume the guest to
1530 * retry native execution under certain conditions,
1531 * Can only be set in conjunction with EMULTYPE_PF.
1533 * EMULTYPE_TRAP_UD_FORCED - Set when emulating an intercepted #UD that was
1534 * triggered by KVM's magic "force emulation" prefix,
1535 * which is opt in via module param (off by default).
1536 * Bypasses EmulateOnUD restriction despite emulating
1537 * due to an intercepted #UD (see EMULTYPE_TRAP_UD).
1538 * Used to test the full emulator from userspace.
1540 * EMULTYPE_VMWARE_GP - Set when emulating an intercepted #GP for VMware
1541 * backdoor emulation, which is opt in via module param.
1542 * VMware backdoor emulation handles select instructions
1543 * and reinjects the #GP for all other cases.
1545 * EMULTYPE_PF - Set when emulating MMIO by way of an intercepted #PF, in which
1546 * case the CR2/GPA value pass on the stack is valid.
1548 #define EMULTYPE_NO_DECODE (1 << 0)
1549 #define EMULTYPE_TRAP_UD (1 << 1)
1550 #define EMULTYPE_SKIP (1 << 2)
1551 #define EMULTYPE_ALLOW_RETRY_PF (1 << 3)
1552 #define EMULTYPE_TRAP_UD_FORCED (1 << 4)
1553 #define EMULTYPE_VMWARE_GP (1 << 5)
1554 #define EMULTYPE_PF (1 << 6)
1556 int kvm_emulate_instruction(struct kvm_vcpu *vcpu, int emulation_type);
1557 int kvm_emulate_instruction_from_buffer(struct kvm_vcpu *vcpu,
1558 void *insn, int insn_len);
1560 void kvm_enable_efer_bits(u64);
1561 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
1562 int __kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data, bool host_initiated);
1563 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data);
1564 int kvm_set_msr(struct kvm_vcpu *vcpu, u32 index, u64 data);
1565 int kvm_emulate_rdmsr(struct kvm_vcpu *vcpu);
1566 int kvm_emulate_wrmsr(struct kvm_vcpu *vcpu);
1567 int kvm_emulate_as_nop(struct kvm_vcpu *vcpu);
1568 int kvm_emulate_invd(struct kvm_vcpu *vcpu);
1569 int kvm_emulate_mwait(struct kvm_vcpu *vcpu);
1570 int kvm_handle_invalid_op(struct kvm_vcpu *vcpu);
1571 int kvm_emulate_monitor(struct kvm_vcpu *vcpu);
1573 int kvm_fast_pio(struct kvm_vcpu *vcpu, int size, unsigned short port, int in);
1574 int kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
1575 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
1576 int kvm_vcpu_halt(struct kvm_vcpu *vcpu);
1577 int kvm_emulate_ap_reset_hold(struct kvm_vcpu *vcpu);
1578 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
1580 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
1581 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
1582 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector);
1584 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
1585 int reason, bool has_error_code, u32 error_code);
1587 void kvm_free_guest_fpu(struct kvm_vcpu *vcpu);
1589 void kvm_post_set_cr0(struct kvm_vcpu *vcpu, unsigned long old_cr0, unsigned long cr0);
1590 void kvm_post_set_cr4(struct kvm_vcpu *vcpu, unsigned long old_cr4, unsigned long cr4);
1591 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
1592 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
1593 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
1594 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
1595 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
1596 void kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
1597 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
1598 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
1599 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
1600 int kvm_emulate_xsetbv(struct kvm_vcpu *vcpu);
1602 int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1603 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1605 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
1606 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
1607 int kvm_emulate_rdpmc(struct kvm_vcpu *vcpu);
1609 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1610 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1611 void kvm_queue_exception_p(struct kvm_vcpu *vcpu, unsigned nr, unsigned long payload);
1612 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1613 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1614 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
1615 bool kvm_inject_emulated_page_fault(struct kvm_vcpu *vcpu,
1616 struct x86_exception *fault);
1617 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1618 gfn_t gfn, void *data, int offset, int len,
1620 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
1621 bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr);
1623 static inline int __kvm_irq_line_state(unsigned long *irq_state,
1624 int irq_source_id, int level)
1626 /* Logical OR for level trig interrupt */
1628 __set_bit(irq_source_id, irq_state);
1630 __clear_bit(irq_source_id, irq_state);
1632 return !!(*irq_state);
1635 #define KVM_MMU_ROOT_CURRENT BIT(0)
1636 #define KVM_MMU_ROOT_PREVIOUS(i) BIT(1+i)
1637 #define KVM_MMU_ROOTS_ALL (~0UL)
1639 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
1640 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
1642 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
1644 void kvm_update_dr7(struct kvm_vcpu *vcpu);
1646 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
1647 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
1648 void kvm_mmu_free_roots(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1649 ulong roots_to_free);
1650 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1651 struct x86_exception *exception);
1652 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
1653 struct x86_exception *exception);
1654 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
1655 struct x86_exception *exception);
1656 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
1657 struct x86_exception *exception);
1658 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
1659 struct x86_exception *exception);
1661 bool kvm_apicv_activated(struct kvm *kvm);
1662 void kvm_apicv_init(struct kvm *kvm, bool enable);
1663 void kvm_vcpu_update_apicv(struct kvm_vcpu *vcpu);
1664 void kvm_request_apicv_update(struct kvm *kvm, bool activate,
1667 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
1669 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u64 error_code,
1670 void *insn, int insn_len);
1671 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
1672 void kvm_mmu_invalidate_gva(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1673 gva_t gva, hpa_t root_hpa);
1674 void kvm_mmu_invpcid_gva(struct kvm_vcpu *vcpu, gva_t gva, unsigned long pcid);
1675 void kvm_mmu_new_pgd(struct kvm_vcpu *vcpu, gpa_t new_pgd, bool skip_tlb_flush,
1676 bool skip_mmu_sync);
1678 void kvm_configure_mmu(bool enable_tdp, int tdp_max_root_level,
1679 int tdp_huge_page_level);
1681 static inline u16 kvm_read_ldt(void)
1684 asm("sldt %0" : "=g"(ldt));
1688 static inline void kvm_load_ldt(u16 sel)
1690 asm("lldt %0" : : "rm"(sel));
1693 #ifdef CONFIG_X86_64
1694 static inline unsigned long read_msr(unsigned long msr)
1703 static inline u32 get_rdx_init_val(void)
1705 return 0x600; /* P6 family */
1708 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
1710 kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
1713 #define TSS_IOPB_BASE_OFFSET 0x66
1714 #define TSS_BASE_SIZE 0x68
1715 #define TSS_IOPB_SIZE (65536 / 8)
1716 #define TSS_REDIRECTION_SIZE (256 / 8)
1717 #define RMODE_TSS_SIZE \
1718 (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
1721 TASK_SWITCH_CALL = 0,
1722 TASK_SWITCH_IRET = 1,
1723 TASK_SWITCH_JMP = 2,
1724 TASK_SWITCH_GATE = 3,
1727 #define HF_GIF_MASK (1 << 0)
1728 #define HF_NMI_MASK (1 << 3)
1729 #define HF_IRET_MASK (1 << 4)
1730 #define HF_GUEST_MASK (1 << 5) /* VCPU is in guest-mode */
1731 #define HF_SMM_MASK (1 << 6)
1732 #define HF_SMM_INSIDE_NMI_MASK (1 << 7)
1734 #define __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
1735 #define KVM_ADDRESS_SPACE_NUM 2
1737 #define kvm_arch_vcpu_memslots_id(vcpu) ((vcpu)->arch.hflags & HF_SMM_MASK ? 1 : 0)
1738 #define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, (role).smm)
1740 asmlinkage void kvm_spurious_fault(void);
1743 * Hardware virtualization extension instructions may fault if a
1744 * reboot turns off virtualization while processes are running.
1745 * Usually after catching the fault we just panic; during reboot
1746 * instead the instruction is ignored.
1748 #define __kvm_handle_fault_on_reboot(insn) \
1754 ".pushsection .discard.instr_begin \n\t" \
1755 ".long 1b - . \n\t" \
1756 ".popsection \n\t" \
1757 "call kvm_spurious_fault \n\t" \
1759 ".pushsection .discard.instr_end \n\t" \
1760 ".long 1b - . \n\t" \
1761 ".popsection \n\t" \
1763 _ASM_EXTABLE(666b, 667b)
1765 #define KVM_ARCH_WANT_MMU_NOTIFIER
1767 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1768 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1769 int kvm_cpu_has_extint(struct kvm_vcpu *v);
1770 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1771 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1772 void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event);
1773 void kvm_vcpu_reload_apic_access_page(struct kvm_vcpu *vcpu);
1775 int kvm_pv_send_ipi(struct kvm *kvm, unsigned long ipi_bitmap_low,
1776 unsigned long ipi_bitmap_high, u32 min,
1777 unsigned long icr, int op_64_bit);
1779 void kvm_define_user_return_msr(unsigned index, u32 msr);
1780 int kvm_set_user_return_msr(unsigned index, u64 val, u64 mask);
1782 u64 kvm_scale_tsc(struct kvm_vcpu *vcpu, u64 tsc);
1783 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc);
1785 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu);
1786 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1788 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
1789 void kvm_make_scan_ioapic_request(struct kvm *kvm);
1790 void kvm_make_scan_ioapic_request_mask(struct kvm *kvm,
1791 unsigned long *vcpu_bitmap);
1793 bool kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1794 struct kvm_async_pf *work);
1795 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1796 struct kvm_async_pf *work);
1797 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1798 struct kvm_async_pf *work);
1799 void kvm_arch_async_page_present_queued(struct kvm_vcpu *vcpu);
1800 bool kvm_arch_can_dequeue_async_page_present(struct kvm_vcpu *vcpu);
1801 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1803 int kvm_skip_emulated_instruction(struct kvm_vcpu *vcpu);
1804 int kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1805 void __kvm_request_immediate_exit(struct kvm_vcpu *vcpu);
1807 int kvm_is_in_guest(void);
1809 void __user *__x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa,
1811 bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu);
1812 bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu);
1814 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
1815 struct kvm_vcpu **dest_vcpu);
1817 void kvm_set_msi_irq(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e,
1818 struct kvm_lapic_irq *irq);
1820 static inline bool kvm_irq_is_postable(struct kvm_lapic_irq *irq)
1822 /* We can only post Fixed and LowPrio IRQs */
1823 return (irq->delivery_mode == APIC_DM_FIXED ||
1824 irq->delivery_mode == APIC_DM_LOWEST);
1827 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu)
1829 static_call_cond(kvm_x86_vcpu_blocking)(vcpu);
1832 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu)
1834 static_call_cond(kvm_x86_vcpu_unblocking)(vcpu);
1837 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
1839 static inline int kvm_cpu_get_apicid(int mps_cpu)
1841 #ifdef CONFIG_X86_LOCAL_APIC
1842 return default_cpu_present_to_apicid(mps_cpu);
1849 #define put_smstate(type, buf, offset, val) \
1850 *(type *)((buf) + (offset) - 0x7e00) = val
1852 #define GET_SMSTATE(type, buf, offset) \
1853 (*(type *)((buf) + (offset) - 0x7e00))
1855 int kvm_cpu_dirty_log_size(void);
1857 #endif /* _ASM_X86_KVM_HOST_H */