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)
117 /* KVM Hugepage definitions for x86 */
118 #define KVM_MAX_HUGEPAGE_LEVEL PG_LEVEL_1G
119 #define KVM_NR_PAGE_SIZES (KVM_MAX_HUGEPAGE_LEVEL - PG_LEVEL_4K + 1)
120 #define KVM_HPAGE_GFN_SHIFT(x) (((x) - 1) * 9)
121 #define KVM_HPAGE_SHIFT(x) (PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
122 #define KVM_HPAGE_SIZE(x) (1UL << KVM_HPAGE_SHIFT(x))
123 #define KVM_HPAGE_MASK(x) (~(KVM_HPAGE_SIZE(x) - 1))
124 #define KVM_PAGES_PER_HPAGE(x) (KVM_HPAGE_SIZE(x) / PAGE_SIZE)
126 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
128 /* KVM_HPAGE_GFN_SHIFT(PG_LEVEL_4K) must be 0. */
129 return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
130 (base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
133 #define KVM_PERMILLE_MMU_PAGES 20
134 #define KVM_MIN_ALLOC_MMU_PAGES 64UL
135 #define KVM_MMU_HASH_SHIFT 12
136 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
137 #define KVM_MIN_FREE_MMU_PAGES 5
138 #define KVM_REFILL_PAGES 25
139 #define KVM_MAX_CPUID_ENTRIES 256
140 #define KVM_NR_FIXED_MTRR_REGION 88
141 #define KVM_NR_VAR_MTRR 8
143 #define ASYNC_PF_PER_VCPU 64
146 VCPU_REGS_RAX = __VCPU_REGS_RAX,
147 VCPU_REGS_RCX = __VCPU_REGS_RCX,
148 VCPU_REGS_RDX = __VCPU_REGS_RDX,
149 VCPU_REGS_RBX = __VCPU_REGS_RBX,
150 VCPU_REGS_RSP = __VCPU_REGS_RSP,
151 VCPU_REGS_RBP = __VCPU_REGS_RBP,
152 VCPU_REGS_RSI = __VCPU_REGS_RSI,
153 VCPU_REGS_RDI = __VCPU_REGS_RDI,
155 VCPU_REGS_R8 = __VCPU_REGS_R8,
156 VCPU_REGS_R9 = __VCPU_REGS_R9,
157 VCPU_REGS_R10 = __VCPU_REGS_R10,
158 VCPU_REGS_R11 = __VCPU_REGS_R11,
159 VCPU_REGS_R12 = __VCPU_REGS_R12,
160 VCPU_REGS_R13 = __VCPU_REGS_R13,
161 VCPU_REGS_R14 = __VCPU_REGS_R14,
162 VCPU_REGS_R15 = __VCPU_REGS_R15,
167 VCPU_EXREG_PDPTR = NR_VCPU_REGS,
173 VCPU_EXREG_EXIT_INFO_1,
174 VCPU_EXREG_EXIT_INFO_2,
188 enum exit_fastpath_completion {
190 EXIT_FASTPATH_REENTER_GUEST,
191 EXIT_FASTPATH_EXIT_HANDLED,
193 typedef enum exit_fastpath_completion fastpath_t;
195 struct x86_emulate_ctxt;
196 struct x86_exception;
198 enum x86_intercept_stage;
200 #define KVM_NR_DB_REGS 4
202 #define DR6_BD (1 << 13)
203 #define DR6_BS (1 << 14)
204 #define DR6_BT (1 << 15)
205 #define DR6_RTM (1 << 16)
207 * DR6_ACTIVE_LOW combines fixed-1 and active-low bits.
208 * We can regard all the bits in DR6_FIXED_1 as active_low bits;
209 * they will never be 0 for now, but when they are defined
210 * in the future it will require no code change.
212 * DR6_ACTIVE_LOW is also used as the init/reset value for DR6.
214 #define DR6_ACTIVE_LOW 0xffff0ff0
215 #define DR6_VOLATILE 0x0001e00f
216 #define DR6_FIXED_1 (DR6_ACTIVE_LOW & ~DR6_VOLATILE)
218 #define DR7_BP_EN_MASK 0x000000ff
219 #define DR7_GE (1 << 9)
220 #define DR7_GD (1 << 13)
221 #define DR7_FIXED_1 0x00000400
222 #define DR7_VOLATILE 0xffff2bff
224 #define KVM_GUESTDBG_VALID_MASK \
225 (KVM_GUESTDBG_ENABLE | \
226 KVM_GUESTDBG_SINGLESTEP | \
227 KVM_GUESTDBG_USE_HW_BP | \
228 KVM_GUESTDBG_USE_SW_BP | \
229 KVM_GUESTDBG_INJECT_BP | \
230 KVM_GUESTDBG_INJECT_DB)
233 #define PFERR_PRESENT_BIT 0
234 #define PFERR_WRITE_BIT 1
235 #define PFERR_USER_BIT 2
236 #define PFERR_RSVD_BIT 3
237 #define PFERR_FETCH_BIT 4
238 #define PFERR_PK_BIT 5
239 #define PFERR_GUEST_FINAL_BIT 32
240 #define PFERR_GUEST_PAGE_BIT 33
242 #define PFERR_PRESENT_MASK (1U << PFERR_PRESENT_BIT)
243 #define PFERR_WRITE_MASK (1U << PFERR_WRITE_BIT)
244 #define PFERR_USER_MASK (1U << PFERR_USER_BIT)
245 #define PFERR_RSVD_MASK (1U << PFERR_RSVD_BIT)
246 #define PFERR_FETCH_MASK (1U << PFERR_FETCH_BIT)
247 #define PFERR_PK_MASK (1U << PFERR_PK_BIT)
248 #define PFERR_GUEST_FINAL_MASK (1ULL << PFERR_GUEST_FINAL_BIT)
249 #define PFERR_GUEST_PAGE_MASK (1ULL << PFERR_GUEST_PAGE_BIT)
251 #define PFERR_NESTED_GUEST_PAGE (PFERR_GUEST_PAGE_MASK | \
255 /* apic attention bits */
256 #define KVM_APIC_CHECK_VAPIC 0
258 * The following bit is set with PV-EOI, unset on EOI.
259 * We detect PV-EOI changes by guest by comparing
260 * this bit with PV-EOI in guest memory.
261 * See the implementation in apic_update_pv_eoi.
263 #define KVM_APIC_PV_EOI_PENDING 1
265 struct kvm_kernel_irq_routing_entry;
268 * the pages used as guest page table on soft mmu are tracked by
269 * kvm_memory_slot.arch.gfn_track which is 16 bits, so the role bits used
270 * by indirect shadow page can not be more than 15 bits.
272 * Currently, we used 14 bits that are @level, @gpte_is_8_bytes, @quadrant, @access,
273 * @nxe, @cr0_wp, @smep_andnot_wp and @smap_andnot_wp.
275 union kvm_mmu_page_role {
279 unsigned gpte_is_8_bytes:1;
286 unsigned smep_andnot_wp:1;
287 unsigned smap_andnot_wp:1;
288 unsigned ad_disabled:1;
289 unsigned guest_mode:1;
293 * This is left at the top of the word so that
294 * kvm_memslots_for_spte_role can extract it with a
295 * simple shift. While there is room, give it a whole
296 * byte so it is also faster to load it from memory.
302 union kvm_mmu_extended_role {
304 * This structure complements kvm_mmu_page_role caching everything needed for
305 * MMU configuration. If nothing in both these structures changed, MMU
306 * re-configuration can be skipped. @valid bit is set on first usage so we don't
307 * treat all-zero structure as valid data.
311 unsigned int valid:1;
312 unsigned int execonly:1;
313 unsigned int cr0_pg:1;
314 unsigned int cr4_pae:1;
315 unsigned int cr4_pse:1;
316 unsigned int cr4_pke:1;
317 unsigned int cr4_smap:1;
318 unsigned int cr4_smep:1;
319 unsigned int maxphyaddr:6;
326 union kvm_mmu_page_role base;
327 union kvm_mmu_extended_role ext;
331 struct kvm_rmap_head {
335 struct kvm_pio_request {
336 unsigned long linear_rip;
343 #define PT64_ROOT_MAX_LEVEL 5
345 struct rsvd_bits_validate {
346 u64 rsvd_bits_mask[2][PT64_ROOT_MAX_LEVEL];
350 struct kvm_mmu_root_info {
355 #define KVM_MMU_ROOT_INFO_INVALID \
356 ((struct kvm_mmu_root_info) { .pgd = INVALID_PAGE, .hpa = INVALID_PAGE })
358 #define KVM_MMU_NUM_PREV_ROOTS 3
360 #define KVM_HAVE_MMU_RWLOCK
365 * x86 supports 4 paging modes (5-level 64-bit, 4-level 64-bit, 3-level 32-bit,
366 * and 2-level 32-bit). The kvm_mmu structure abstracts the details of the
370 unsigned long (*get_guest_pgd)(struct kvm_vcpu *vcpu);
371 u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
372 int (*page_fault)(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u32 err,
374 void (*inject_page_fault)(struct kvm_vcpu *vcpu,
375 struct x86_exception *fault);
376 gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gpa_t gva_or_gpa,
377 u32 access, struct x86_exception *exception);
378 gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
379 struct x86_exception *exception);
380 int (*sync_page)(struct kvm_vcpu *vcpu,
381 struct kvm_mmu_page *sp);
382 void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva, hpa_t root_hpa);
385 union kvm_mmu_role mmu_role;
387 u8 shadow_root_level;
390 struct kvm_mmu_root_info prev_roots[KVM_MMU_NUM_PREV_ROOTS];
393 * Bitmap; bit set = permission fault
394 * Byte index: page fault error code [4:1]
395 * Bit index: pte permissions in ACC_* format
400 * The pkru_mask indicates if protection key checks are needed. It
401 * consists of 16 domains indexed by page fault error code bits [4:1],
402 * with PFEC.RSVD replaced by ACC_USER_MASK from the page tables.
403 * Each domain has 2 bits which are ANDed with AD and WD from PKRU.
411 * check zero bits on shadow page table entries, these
412 * bits include not only hardware reserved bits but also
413 * the bits spte never used.
415 struct rsvd_bits_validate shadow_zero_check;
417 struct rsvd_bits_validate guest_rsvd_check;
419 /* Can have large pages at levels 2..last_nonleaf_level-1. */
420 u8 last_nonleaf_level;
424 u64 pdptrs[4]; /* pae */
427 struct kvm_tlb_range {
442 struct perf_event *perf_event;
443 struct kvm_vcpu *vcpu;
445 * eventsel value for general purpose counters,
446 * ctrl value for fixed counters.
452 unsigned nr_arch_gp_counters;
453 unsigned nr_arch_fixed_counters;
454 unsigned available_event_types;
459 u64 counter_bitmask[2];
460 u64 global_ctrl_mask;
461 u64 global_ovf_ctrl_mask;
464 struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
465 struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED];
466 struct irq_work irq_work;
467 DECLARE_BITMAP(reprogram_pmi, X86_PMC_IDX_MAX);
468 DECLARE_BITMAP(all_valid_pmc_idx, X86_PMC_IDX_MAX);
469 DECLARE_BITMAP(pmc_in_use, X86_PMC_IDX_MAX);
472 * The gate to release perf_events not marked in
473 * pmc_in_use only once in a vcpu time slice.
478 * The total number of programmed perf_events and it helps to avoid
479 * redundant check before cleanup if guest don't use vPMU at all.
487 KVM_DEBUGREG_BP_ENABLED = 1,
488 KVM_DEBUGREG_WONT_EXIT = 2,
489 KVM_DEBUGREG_RELOAD = 4,
492 struct kvm_mtrr_range {
495 struct list_head node;
499 struct kvm_mtrr_range var_ranges[KVM_NR_VAR_MTRR];
500 mtrr_type fixed_ranges[KVM_NR_FIXED_MTRR_REGION];
503 struct list_head head;
506 /* Hyper-V SynIC timer */
507 struct kvm_vcpu_hv_stimer {
508 struct hrtimer timer;
510 union hv_stimer_config config;
513 struct hv_message msg;
517 /* Hyper-V synthetic interrupt controller (SynIC)*/
518 struct kvm_vcpu_hv_synic {
523 atomic64_t sint[HV_SYNIC_SINT_COUNT];
524 atomic_t sint_to_gsi[HV_SYNIC_SINT_COUNT];
525 DECLARE_BITMAP(auto_eoi_bitmap, 256);
526 DECLARE_BITMAP(vec_bitmap, 256);
528 bool dont_zero_synic_pages;
531 /* Hyper-V per vcpu emulation context */
533 struct kvm_vcpu *vcpu;
537 struct kvm_vcpu_hv_synic synic;
538 struct kvm_hyperv_exit exit;
539 struct kvm_vcpu_hv_stimer stimer[HV_SYNIC_STIMER_COUNT];
540 DECLARE_BITMAP(stimer_pending_bitmap, HV_SYNIC_STIMER_COUNT);
544 /* Xen HVM per vcpu emulation context */
545 struct kvm_vcpu_xen {
547 u32 current_runstate;
549 bool vcpu_time_info_set;
551 struct gfn_to_hva_cache vcpu_info_cache;
552 struct gfn_to_hva_cache vcpu_time_info_cache;
553 struct gfn_to_hva_cache runstate_cache;
555 u64 runstate_entry_time;
556 u64 runstate_times[4];
559 struct kvm_vcpu_arch {
561 * rip and regs accesses must go through
562 * kvm_{register,rip}_{read,write} functions.
564 unsigned long regs[NR_VCPU_REGS];
569 unsigned long cr0_guest_owned_bits;
573 unsigned long cr4_guest_owned_bits;
574 unsigned long cr4_guest_rsvd_bits;
581 struct kvm_lapic *apic; /* kernel irqchip context */
583 bool load_eoi_exitmap_pending;
584 DECLARE_BITMAP(ioapic_handled_vectors, 256);
585 unsigned long apic_attention;
586 int32_t apic_arb_prio;
588 u64 ia32_misc_enable_msr;
591 bool tpr_access_reporting;
594 u64 microcode_version;
595 u64 arch_capabilities;
596 u64 perf_capabilities;
599 * Paging state of the vcpu
601 * If the vcpu runs in guest mode with two level paging this still saves
602 * the paging mode of the l1 guest. This context is always used to
607 /* Non-nested MMU for L1 */
608 struct kvm_mmu root_mmu;
610 /* L1 MMU when running nested */
611 struct kvm_mmu guest_mmu;
614 * Paging state of an L2 guest (used for nested npt)
616 * This context will save all necessary information to walk page tables
617 * of an L2 guest. This context is only initialized for page table
618 * walking and not for faulting since we never handle l2 page faults on
621 struct kvm_mmu nested_mmu;
624 * Pointer to the mmu context currently used for
625 * gva_to_gpa translations.
627 struct kvm_mmu *walk_mmu;
629 struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
630 struct kvm_mmu_memory_cache mmu_shadow_page_cache;
631 struct kvm_mmu_memory_cache mmu_gfn_array_cache;
632 struct kvm_mmu_memory_cache mmu_page_header_cache;
635 * QEMU userspace and the guest each have their own FPU state.
636 * In vcpu_run, we switch between the user and guest FPU contexts.
637 * While running a VCPU, the VCPU thread will have the guest FPU
640 * Note that while the PKRU state lives inside the fpu registers,
641 * it is switched out separately at VMENTER and VMEXIT time. The
642 * "guest_fpu" state here contains the guest FPU context, with the
645 struct fpu *user_fpu;
646 struct fpu *guest_fpu;
649 u64 guest_supported_xcr0;
651 struct kvm_pio_request pio;
653 void *guest_ins_data;
655 u8 event_exit_inst_len;
657 struct kvm_queued_exception {
663 unsigned long payload;
668 struct kvm_queued_interrupt {
674 int halt_request; /* real mode on Intel only */
677 struct kvm_cpuid_entry2 *cpuid_entries;
679 u64 reserved_gpa_bits;
683 /* emulate context */
685 struct x86_emulate_ctxt *emulate_ctxt;
686 bool emulate_regs_need_sync_to_vcpu;
687 bool emulate_regs_need_sync_from_vcpu;
688 int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
691 struct pvclock_vcpu_time_info hv_clock;
692 unsigned int hw_tsc_khz;
693 struct gfn_to_hva_cache pv_time;
694 bool pv_time_enabled;
695 /* set guest stopped flag in pvclock flags field */
696 bool pvclock_set_guest_stopped_request;
702 struct gfn_to_pfn_cache cache;
709 u64 tsc_offset_adjustment;
712 u64 this_tsc_generation;
714 bool tsc_always_catchup;
715 s8 virtual_tsc_shift;
716 u32 virtual_tsc_mult;
718 s64 ia32_tsc_adjust_msr;
719 u64 msr_ia32_power_ctl;
720 u64 tsc_scaling_ratio;
722 atomic_t nmi_queued; /* unprocessed asynchronous NMIs */
723 unsigned nmi_pending; /* NMI queued after currently running handler */
724 bool nmi_injected; /* Trying to inject an NMI this entry */
725 bool smi_pending; /* SMI queued after currently running handler */
727 struct kvm_mtrr mtrr_state;
730 unsigned switch_db_regs;
731 unsigned long db[KVM_NR_DB_REGS];
734 unsigned long eff_db[KVM_NR_DB_REGS];
735 unsigned long guest_debug_dr7;
736 u64 msr_platform_info;
737 u64 msr_misc_features_enables;
745 /* Cache MMIO info */
747 unsigned mmio_access;
753 /* used for guest single stepping over the given code position */
754 unsigned long singlestep_rip;
757 struct kvm_vcpu_hv *hyperv;
758 struct kvm_vcpu_xen xen;
760 cpumask_var_t wbinvd_dirty_mask;
762 unsigned long last_retry_eip;
763 unsigned long last_retry_addr;
767 gfn_t gfns[ASYNC_PF_PER_VCPU];
768 struct gfn_to_hva_cache data;
769 u64 msr_en_val; /* MSR_KVM_ASYNC_PF_EN */
770 u64 msr_int_val; /* MSR_KVM_ASYNC_PF_INT */
775 unsigned long nested_apf_token;
776 bool delivery_as_pf_vmexit;
777 bool pageready_pending;
780 /* OSVW MSRs (AMD only) */
788 struct gfn_to_hva_cache data;
791 u64 msr_kvm_poll_control;
794 * Indicates the guest is trying to write a gfn that contains one or
795 * more of the PTEs used to translate the write itself, i.e. the access
796 * is changing its own translation in the guest page tables. KVM exits
797 * to userspace if emulation of the faulting instruction fails and this
798 * flag is set, as KVM cannot make forward progress.
800 * If emulation fails for a write to guest page tables, KVM unprotects
801 * (zaps) the shadow page for the target gfn and resumes the guest to
802 * retry the non-emulatable instruction (on hardware). Unprotecting the
803 * gfn doesn't allow forward progress for a self-changing access because
804 * doing so also zaps the translation for the gfn, i.e. retrying the
805 * instruction will hit a !PRESENT fault, which results in a new shadow
806 * page and sends KVM back to square one.
808 bool write_fault_to_shadow_pgtable;
810 /* set at EPT violation at this point */
811 unsigned long exit_qualification;
813 /* pv related host specific info */
818 int pending_ioapic_eoi;
819 int pending_external_vector;
821 /* be preempted when it's in kernel-mode(cpl=0) */
822 bool preempted_in_kernel;
824 /* Flush the L1 Data cache for L1TF mitigation on VMENTER */
827 /* Host CPU on which VM-entry was most recently attempted */
828 unsigned int last_vmentry_cpu;
830 /* AMD MSRC001_0015 Hardware Configuration */
833 /* pv related cpuid info */
836 * value of the eax register in the KVM_CPUID_FEATURES CPUID
842 * indicates whether pv emulation should be disabled if features
843 * are not present in the guest's cpuid
848 /* Protected Guests */
849 bool guest_state_protected;
852 struct kvm_lpage_info {
856 struct kvm_arch_memory_slot {
857 struct kvm_rmap_head *rmap[KVM_NR_PAGE_SIZES];
858 struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
859 unsigned short *gfn_track[KVM_PAGE_TRACK_MAX];
863 * We use as the mode the number of bits allocated in the LDR for the
864 * logical processor ID. It happens that these are all powers of two.
865 * This makes it is very easy to detect cases where the APICs are
866 * configured for multiple modes; in that case, we cannot use the map and
867 * hence cannot use kvm_irq_delivery_to_apic_fast either.
869 #define KVM_APIC_MODE_XAPIC_CLUSTER 4
870 #define KVM_APIC_MODE_XAPIC_FLAT 8
871 #define KVM_APIC_MODE_X2APIC 16
873 struct kvm_apic_map {
878 struct kvm_lapic *xapic_flat_map[8];
879 struct kvm_lapic *xapic_cluster_map[16][4];
881 struct kvm_lapic *phys_map[];
884 /* Hyper-V synthetic debugger (SynDbg)*/
885 struct kvm_hv_syndbg {
896 /* Hyper-V emulation context */
898 struct mutex hv_lock;
903 /* Hyper-v based guest crash (NT kernel bugcheck) parameters */
904 u64 hv_crash_param[HV_X64_MSR_CRASH_PARAMS];
907 struct ms_hyperv_tsc_page tsc_ref;
909 struct idr conn_to_evt;
911 u64 hv_reenlightenment_control;
912 u64 hv_tsc_emulation_control;
913 u64 hv_tsc_emulation_status;
915 /* How many vCPUs have VP index != vCPU index */
916 atomic_t num_mismatched_vp_indexes;
918 struct hv_partition_assist_pg *hv_pa_pg;
919 struct kvm_hv_syndbg hv_syndbg;
922 struct msr_bitmap_range {
926 unsigned long *bitmap;
929 /* Xen emulation context */
934 struct gfn_to_hva_cache shinfo_cache;
937 enum kvm_irqchip_mode {
939 KVM_IRQCHIP_KERNEL, /* created with KVM_CREATE_IRQCHIP */
940 KVM_IRQCHIP_SPLIT, /* created with KVM_CAP_SPLIT_IRQCHIP */
943 #define APICV_INHIBIT_REASON_DISABLE 0
944 #define APICV_INHIBIT_REASON_HYPERV 1
945 #define APICV_INHIBIT_REASON_NESTED 2
946 #define APICV_INHIBIT_REASON_IRQWIN 3
947 #define APICV_INHIBIT_REASON_PIT_REINJ 4
948 #define APICV_INHIBIT_REASON_X2APIC 5
951 unsigned long n_used_mmu_pages;
952 unsigned long n_requested_mmu_pages;
953 unsigned long n_max_mmu_pages;
954 unsigned int indirect_shadow_pages;
956 struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
957 struct list_head active_mmu_pages;
958 struct list_head zapped_obsolete_pages;
959 struct list_head lpage_disallowed_mmu_pages;
960 struct kvm_page_track_notifier_node mmu_sp_tracker;
961 struct kvm_page_track_notifier_head track_notifier_head;
963 struct list_head assigned_dev_head;
964 struct iommu_domain *iommu_domain;
965 bool iommu_noncoherent;
966 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA
967 atomic_t noncoherent_dma_count;
968 #define __KVM_HAVE_ARCH_ASSIGNED_DEVICE
969 atomic_t assigned_device_count;
970 struct kvm_pic *vpic;
971 struct kvm_ioapic *vioapic;
972 struct kvm_pit *vpit;
973 atomic_t vapics_in_nmi_mode;
974 struct mutex apic_map_lock;
975 struct kvm_apic_map __rcu *apic_map;
976 atomic_t apic_map_dirty;
978 bool apic_access_page_done;
979 unsigned long apicv_inhibit_reasons;
986 bool cstate_in_guest;
988 unsigned long irq_sources_bitmap;
990 raw_spinlock_t tsc_write_lock;
997 u64 cur_tsc_generation;
998 int nr_vcpus_matched_tsc;
1000 spinlock_t pvclock_gtod_sync_lock;
1001 bool use_master_clock;
1002 u64 master_kernel_ns;
1003 u64 master_cycle_now;
1004 struct delayed_work kvmclock_update_work;
1005 struct delayed_work kvmclock_sync_work;
1007 struct kvm_xen_hvm_config xen_hvm_config;
1009 /* reads protected by irq_srcu, writes by irq_lock */
1010 struct hlist_head mask_notifier_list;
1012 struct kvm_hv hyperv;
1015 #ifdef CONFIG_KVM_MMU_AUDIT
1019 bool backwards_tsc_observed;
1020 bool boot_vcpu_runs_old_kvmclock;
1023 u64 disabled_quirks;
1024 int cpu_dirty_logging_count;
1026 enum kvm_irqchip_mode irqchip_mode;
1027 u8 nr_reserved_ioapic_pins;
1029 bool disabled_lapic_found;
1032 bool x2apic_broadcast_quirk_disabled;
1034 bool guest_can_read_msr_platform_info;
1035 bool exception_payload_enabled;
1037 /* Deflect RDMSR and WRMSR to user space when they trigger a #GP */
1038 u32 user_space_msr_mask;
1042 bool default_allow:1;
1043 struct msr_bitmap_range ranges[16];
1046 bool bus_lock_detection_enabled;
1048 struct kvm_pmu_event_filter __rcu *pmu_event_filter;
1049 struct task_struct *nx_lpage_recovery_thread;
1051 #ifdef CONFIG_X86_64
1053 * Whether the TDP MMU is enabled for this VM. This contains a
1054 * snapshot of the TDP MMU module parameter from when the VM was
1055 * created and remains unchanged for the life of the VM. If this is
1056 * true, TDP MMU handler functions will run for various MMU
1059 bool tdp_mmu_enabled;
1062 * List of struct kvmp_mmu_pages being used as roots.
1063 * All struct kvm_mmu_pages in the list should have
1065 * All struct kvm_mmu_pages in the list should have a positive
1066 * root_count except when a thread holds the MMU lock and is removing
1067 * an entry from the list.
1069 struct list_head tdp_mmu_roots;
1072 * List of struct kvmp_mmu_pages not being used as roots.
1073 * All struct kvm_mmu_pages in the list should have
1074 * tdp_mmu_page set and a root_count of 0.
1076 struct list_head tdp_mmu_pages;
1079 * Protects accesses to the following fields when the MMU lock
1080 * is held in read mode:
1081 * - tdp_mmu_pages (above)
1082 * - the link field of struct kvm_mmu_pages used by the TDP MMU
1083 * - lpage_disallowed_mmu_pages
1084 * - the lpage_disallowed_link field of struct kvm_mmu_pages used
1086 * It is acceptable, but not necessary, to acquire this lock when
1087 * the thread holds the MMU lock in write mode.
1089 spinlock_t tdp_mmu_pages_lock;
1090 #endif /* CONFIG_X86_64 */
1093 struct kvm_vm_stat {
1094 ulong mmu_shadow_zapped;
1095 ulong mmu_pte_write;
1096 ulong mmu_pde_zapped;
1099 ulong mmu_cache_miss;
1101 ulong remote_tlb_flush;
1103 ulong nx_lpage_splits;
1104 ulong max_mmu_page_hash_collisions;
1107 struct kvm_vcpu_stat {
1117 u64 irq_window_exits;
1118 u64 nmi_window_exits;
1121 u64 halt_successful_poll;
1122 u64 halt_attempted_poll;
1123 u64 halt_poll_invalid;
1125 u64 request_irq_exits;
1127 u64 host_state_reload;
1130 u64 insn_emulation_fail;
1135 u64 halt_poll_success_ns;
1136 u64 halt_poll_fail_ns;
1140 struct x86_instruction_info;
1143 bool host_initiated;
1148 struct kvm_lapic_irq {
1156 bool msi_redir_hint;
1159 static inline u16 kvm_lapic_irq_dest_mode(bool dest_mode_logical)
1161 return dest_mode_logical ? APIC_DEST_LOGICAL : APIC_DEST_PHYSICAL;
1164 struct kvm_x86_ops {
1165 int (*hardware_enable)(void);
1166 void (*hardware_disable)(void);
1167 void (*hardware_unsetup)(void);
1168 bool (*cpu_has_accelerated_tpr)(void);
1169 bool (*has_emulated_msr)(struct kvm *kvm, u32 index);
1170 void (*vcpu_after_set_cpuid)(struct kvm_vcpu *vcpu);
1172 unsigned int vm_size;
1173 int (*vm_init)(struct kvm *kvm);
1174 void (*vm_destroy)(struct kvm *kvm);
1176 /* Create, but do not attach this VCPU */
1177 int (*vcpu_create)(struct kvm_vcpu *vcpu);
1178 void (*vcpu_free)(struct kvm_vcpu *vcpu);
1179 void (*vcpu_reset)(struct kvm_vcpu *vcpu, bool init_event);
1181 void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
1182 void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
1183 void (*vcpu_put)(struct kvm_vcpu *vcpu);
1185 void (*update_exception_bitmap)(struct kvm_vcpu *vcpu);
1186 int (*get_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1187 int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1188 u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
1189 void (*get_segment)(struct kvm_vcpu *vcpu,
1190 struct kvm_segment *var, int seg);
1191 int (*get_cpl)(struct kvm_vcpu *vcpu);
1192 void (*set_segment)(struct kvm_vcpu *vcpu,
1193 struct kvm_segment *var, int seg);
1194 void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
1195 void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
1196 bool (*is_valid_cr4)(struct kvm_vcpu *vcpu, unsigned long cr0);
1197 void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
1198 int (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
1199 void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1200 void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1201 void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1202 void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1203 void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu);
1204 void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
1205 void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
1206 unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
1207 void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
1209 void (*tlb_flush_all)(struct kvm_vcpu *vcpu);
1210 void (*tlb_flush_current)(struct kvm_vcpu *vcpu);
1211 int (*tlb_remote_flush)(struct kvm *kvm);
1212 int (*tlb_remote_flush_with_range)(struct kvm *kvm,
1213 struct kvm_tlb_range *range);
1216 * Flush any TLB entries associated with the given GVA.
1217 * Does not need to flush GPA->HPA mappings.
1218 * Can potentially get non-canonical addresses through INVLPGs, which
1219 * the implementation may choose to ignore if appropriate.
1221 void (*tlb_flush_gva)(struct kvm_vcpu *vcpu, gva_t addr);
1224 * Flush any TLB entries created by the guest. Like tlb_flush_gva(),
1225 * does not need to flush GPA->HPA mappings.
1227 void (*tlb_flush_guest)(struct kvm_vcpu *vcpu);
1229 enum exit_fastpath_completion (*run)(struct kvm_vcpu *vcpu);
1230 int (*handle_exit)(struct kvm_vcpu *vcpu,
1231 enum exit_fastpath_completion exit_fastpath);
1232 int (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
1233 void (*update_emulated_instruction)(struct kvm_vcpu *vcpu);
1234 void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
1235 u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu);
1236 void (*patch_hypercall)(struct kvm_vcpu *vcpu,
1237 unsigned char *hypercall_addr);
1238 void (*set_irq)(struct kvm_vcpu *vcpu);
1239 void (*set_nmi)(struct kvm_vcpu *vcpu);
1240 void (*queue_exception)(struct kvm_vcpu *vcpu);
1241 void (*cancel_injection)(struct kvm_vcpu *vcpu);
1242 int (*interrupt_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1243 int (*nmi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1244 bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
1245 void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
1246 void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
1247 void (*enable_irq_window)(struct kvm_vcpu *vcpu);
1248 void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
1249 bool (*check_apicv_inhibit_reasons)(ulong bit);
1250 void (*pre_update_apicv_exec_ctrl)(struct kvm *kvm, bool activate);
1251 void (*refresh_apicv_exec_ctrl)(struct kvm_vcpu *vcpu);
1252 void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
1253 void (*hwapic_isr_update)(struct kvm_vcpu *vcpu, int isr);
1254 bool (*guest_apic_has_interrupt)(struct kvm_vcpu *vcpu);
1255 void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
1256 void (*set_virtual_apic_mode)(struct kvm_vcpu *vcpu);
1257 void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu);
1258 int (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
1259 int (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
1260 int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
1261 int (*set_identity_map_addr)(struct kvm *kvm, u64 ident_addr);
1262 u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
1264 void (*load_mmu_pgd)(struct kvm_vcpu *vcpu, hpa_t root_hpa,
1267 bool (*has_wbinvd_exit)(void);
1269 /* Returns actual tsc_offset set in active VMCS */
1270 u64 (*write_l1_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
1273 * Retrieve somewhat arbitrary exit information. Intended to be used
1274 * only from within tracepoints to avoid VMREADs when tracing is off.
1276 void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2,
1277 u32 *exit_int_info, u32 *exit_int_info_err_code);
1279 int (*check_intercept)(struct kvm_vcpu *vcpu,
1280 struct x86_instruction_info *info,
1281 enum x86_intercept_stage stage,
1282 struct x86_exception *exception);
1283 void (*handle_exit_irqoff)(struct kvm_vcpu *vcpu);
1285 void (*request_immediate_exit)(struct kvm_vcpu *vcpu);
1287 void (*sched_in)(struct kvm_vcpu *kvm, int cpu);
1290 * Size of the CPU's dirty log buffer, i.e. VMX's PML buffer. A zero
1291 * value indicates CPU dirty logging is unsupported or disabled.
1293 int cpu_dirty_log_size;
1294 void (*update_cpu_dirty_logging)(struct kvm_vcpu *vcpu);
1296 /* pmu operations of sub-arch */
1297 const struct kvm_pmu_ops *pmu_ops;
1298 const struct kvm_x86_nested_ops *nested_ops;
1301 * Architecture specific hooks for vCPU blocking due to
1303 * Returns for .pre_block():
1304 * - 0 means continue to block the vCPU.
1305 * - 1 means we cannot block the vCPU since some event
1306 * happens during this period, such as, 'ON' bit in
1307 * posted-interrupts descriptor is set.
1309 int (*pre_block)(struct kvm_vcpu *vcpu);
1310 void (*post_block)(struct kvm_vcpu *vcpu);
1312 void (*vcpu_blocking)(struct kvm_vcpu *vcpu);
1313 void (*vcpu_unblocking)(struct kvm_vcpu *vcpu);
1315 int (*update_pi_irte)(struct kvm *kvm, unsigned int host_irq,
1316 uint32_t guest_irq, bool set);
1317 void (*apicv_post_state_restore)(struct kvm_vcpu *vcpu);
1318 bool (*dy_apicv_has_pending_interrupt)(struct kvm_vcpu *vcpu);
1320 int (*set_hv_timer)(struct kvm_vcpu *vcpu, u64 guest_deadline_tsc,
1322 void (*cancel_hv_timer)(struct kvm_vcpu *vcpu);
1324 void (*setup_mce)(struct kvm_vcpu *vcpu);
1326 int (*smi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1327 int (*pre_enter_smm)(struct kvm_vcpu *vcpu, char *smstate);
1328 int (*pre_leave_smm)(struct kvm_vcpu *vcpu, const char *smstate);
1329 void (*enable_smi_window)(struct kvm_vcpu *vcpu);
1331 int (*mem_enc_op)(struct kvm *kvm, void __user *argp);
1332 int (*mem_enc_reg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1333 int (*mem_enc_unreg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1335 int (*get_msr_feature)(struct kvm_msr_entry *entry);
1337 bool (*can_emulate_instruction)(struct kvm_vcpu *vcpu, void *insn, int insn_len);
1339 bool (*apic_init_signal_blocked)(struct kvm_vcpu *vcpu);
1340 int (*enable_direct_tlbflush)(struct kvm_vcpu *vcpu);
1342 void (*migrate_timers)(struct kvm_vcpu *vcpu);
1343 void (*msr_filter_changed)(struct kvm_vcpu *vcpu);
1344 int (*complete_emulated_msr)(struct kvm_vcpu *vcpu, int err);
1346 void (*vcpu_deliver_sipi_vector)(struct kvm_vcpu *vcpu, u8 vector);
1349 struct kvm_x86_nested_ops {
1350 int (*check_events)(struct kvm_vcpu *vcpu);
1351 bool (*hv_timer_pending)(struct kvm_vcpu *vcpu);
1352 void (*triple_fault)(struct kvm_vcpu *vcpu);
1353 int (*get_state)(struct kvm_vcpu *vcpu,
1354 struct kvm_nested_state __user *user_kvm_nested_state,
1355 unsigned user_data_size);
1356 int (*set_state)(struct kvm_vcpu *vcpu,
1357 struct kvm_nested_state __user *user_kvm_nested_state,
1358 struct kvm_nested_state *kvm_state);
1359 bool (*get_nested_state_pages)(struct kvm_vcpu *vcpu);
1360 int (*write_log_dirty)(struct kvm_vcpu *vcpu, gpa_t l2_gpa);
1362 int (*enable_evmcs)(struct kvm_vcpu *vcpu,
1363 uint16_t *vmcs_version);
1364 uint16_t (*get_evmcs_version)(struct kvm_vcpu *vcpu);
1367 struct kvm_x86_init_ops {
1368 int (*cpu_has_kvm_support)(void);
1369 int (*disabled_by_bios)(void);
1370 int (*check_processor_compatibility)(void);
1371 int (*hardware_setup)(void);
1373 struct kvm_x86_ops *runtime_ops;
1376 struct kvm_arch_async_pf {
1383 extern u64 __read_mostly host_efer;
1384 extern bool __read_mostly allow_smaller_maxphyaddr;
1385 extern struct kvm_x86_ops kvm_x86_ops;
1387 #define KVM_X86_OP(func) \
1388 DECLARE_STATIC_CALL(kvm_x86_##func, *(((struct kvm_x86_ops *)0)->func));
1389 #define KVM_X86_OP_NULL KVM_X86_OP
1390 #include <asm/kvm-x86-ops.h>
1392 static inline void kvm_ops_static_call_update(void)
1394 #define KVM_X86_OP(func) \
1395 static_call_update(kvm_x86_##func, kvm_x86_ops.func);
1396 #define KVM_X86_OP_NULL KVM_X86_OP
1397 #include <asm/kvm-x86-ops.h>
1400 #define __KVM_HAVE_ARCH_VM_ALLOC
1401 static inline struct kvm *kvm_arch_alloc_vm(void)
1403 return __vmalloc(kvm_x86_ops.vm_size, GFP_KERNEL_ACCOUNT | __GFP_ZERO);
1405 void kvm_arch_free_vm(struct kvm *kvm);
1407 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
1408 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
1410 if (kvm_x86_ops.tlb_remote_flush &&
1411 !static_call(kvm_x86_tlb_remote_flush)(kvm))
1417 int kvm_mmu_module_init(void);
1418 void kvm_mmu_module_exit(void);
1420 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
1421 int kvm_mmu_create(struct kvm_vcpu *vcpu);
1422 void kvm_mmu_init_vm(struct kvm *kvm);
1423 void kvm_mmu_uninit_vm(struct kvm *kvm);
1425 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
1426 void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
1427 struct kvm_memory_slot *memslot,
1429 void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
1430 const struct kvm_memory_slot *memslot);
1431 void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
1432 struct kvm_memory_slot *memslot);
1433 void kvm_mmu_slot_largepage_remove_write_access(struct kvm *kvm,
1434 struct kvm_memory_slot *memslot);
1435 void kvm_mmu_zap_all(struct kvm *kvm);
1436 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, u64 gen);
1437 unsigned long kvm_mmu_calculate_default_mmu_pages(struct kvm *kvm);
1438 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned long kvm_nr_mmu_pages);
1440 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
1441 bool pdptrs_changed(struct kvm_vcpu *vcpu);
1443 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
1444 const void *val, int bytes);
1446 struct kvm_irq_mask_notifier {
1447 void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
1449 struct hlist_node link;
1452 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
1453 struct kvm_irq_mask_notifier *kimn);
1454 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
1455 struct kvm_irq_mask_notifier *kimn);
1456 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
1459 extern bool tdp_enabled;
1461 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
1463 /* control of guest tsc rate supported? */
1464 extern bool kvm_has_tsc_control;
1465 /* maximum supported tsc_khz for guests */
1466 extern u32 kvm_max_guest_tsc_khz;
1467 /* number of bits of the fractional part of the TSC scaling ratio */
1468 extern u8 kvm_tsc_scaling_ratio_frac_bits;
1469 /* maximum allowed value of TSC scaling ratio */
1470 extern u64 kvm_max_tsc_scaling_ratio;
1471 /* 1ull << kvm_tsc_scaling_ratio_frac_bits */
1472 extern u64 kvm_default_tsc_scaling_ratio;
1473 /* bus lock detection supported? */
1474 extern bool kvm_has_bus_lock_exit;
1476 extern u64 kvm_mce_cap_supported;
1479 * EMULTYPE_NO_DECODE - Set when re-emulating an instruction (after completing
1480 * userspace I/O) to indicate that the emulation context
1481 * should be resued as is, i.e. skip initialization of
1482 * emulation context, instruction fetch and decode.
1484 * EMULTYPE_TRAP_UD - Set when emulating an intercepted #UD from hardware.
1485 * Indicates that only select instructions (tagged with
1486 * EmulateOnUD) should be emulated (to minimize the emulator
1487 * attack surface). See also EMULTYPE_TRAP_UD_FORCED.
1489 * EMULTYPE_SKIP - Set when emulating solely to skip an instruction, i.e. to
1490 * decode the instruction length. For use *only* by
1491 * kvm_x86_ops.skip_emulated_instruction() implementations.
1493 * EMULTYPE_ALLOW_RETRY_PF - Set when the emulator should resume the guest to
1494 * retry native execution under certain conditions,
1495 * Can only be set in conjunction with EMULTYPE_PF.
1497 * EMULTYPE_TRAP_UD_FORCED - Set when emulating an intercepted #UD that was
1498 * triggered by KVM's magic "force emulation" prefix,
1499 * which is opt in via module param (off by default).
1500 * Bypasses EmulateOnUD restriction despite emulating
1501 * due to an intercepted #UD (see EMULTYPE_TRAP_UD).
1502 * Used to test the full emulator from userspace.
1504 * EMULTYPE_VMWARE_GP - Set when emulating an intercepted #GP for VMware
1505 * backdoor emulation, which is opt in via module param.
1506 * VMware backoor emulation handles select instructions
1507 * and reinjects the #GP for all other cases.
1509 * EMULTYPE_PF - Set when emulating MMIO by way of an intercepted #PF, in which
1510 * case the CR2/GPA value pass on the stack is valid.
1512 #define EMULTYPE_NO_DECODE (1 << 0)
1513 #define EMULTYPE_TRAP_UD (1 << 1)
1514 #define EMULTYPE_SKIP (1 << 2)
1515 #define EMULTYPE_ALLOW_RETRY_PF (1 << 3)
1516 #define EMULTYPE_TRAP_UD_FORCED (1 << 4)
1517 #define EMULTYPE_VMWARE_GP (1 << 5)
1518 #define EMULTYPE_PF (1 << 6)
1520 int kvm_emulate_instruction(struct kvm_vcpu *vcpu, int emulation_type);
1521 int kvm_emulate_instruction_from_buffer(struct kvm_vcpu *vcpu,
1522 void *insn, int insn_len);
1524 void kvm_enable_efer_bits(u64);
1525 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
1526 int __kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data, bool host_initiated);
1527 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data);
1528 int kvm_set_msr(struct kvm_vcpu *vcpu, u32 index, u64 data);
1529 int kvm_emulate_rdmsr(struct kvm_vcpu *vcpu);
1530 int kvm_emulate_wrmsr(struct kvm_vcpu *vcpu);
1531 int kvm_emulate_as_nop(struct kvm_vcpu *vcpu);
1532 int kvm_emulate_invd(struct kvm_vcpu *vcpu);
1533 int kvm_emulate_mwait(struct kvm_vcpu *vcpu);
1534 int kvm_handle_invalid_op(struct kvm_vcpu *vcpu);
1535 int kvm_emulate_monitor(struct kvm_vcpu *vcpu);
1537 int kvm_fast_pio(struct kvm_vcpu *vcpu, int size, unsigned short port, int in);
1538 int kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
1539 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
1540 int kvm_vcpu_halt(struct kvm_vcpu *vcpu);
1541 int kvm_emulate_ap_reset_hold(struct kvm_vcpu *vcpu);
1542 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
1544 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
1545 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
1546 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector);
1548 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
1549 int reason, bool has_error_code, u32 error_code);
1551 void kvm_free_guest_fpu(struct kvm_vcpu *vcpu);
1553 void kvm_post_set_cr0(struct kvm_vcpu *vcpu, unsigned long old_cr0, unsigned long cr0);
1554 void kvm_post_set_cr4(struct kvm_vcpu *vcpu, unsigned long old_cr4, unsigned long cr4);
1555 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
1556 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
1557 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
1558 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
1559 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
1560 void kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
1561 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
1562 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
1563 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
1564 int kvm_emulate_xsetbv(struct kvm_vcpu *vcpu);
1566 int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1567 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1569 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
1570 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
1571 int kvm_emulate_rdpmc(struct kvm_vcpu *vcpu);
1573 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1574 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1575 void kvm_queue_exception_p(struct kvm_vcpu *vcpu, unsigned nr, unsigned long payload);
1576 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1577 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1578 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
1579 bool kvm_inject_emulated_page_fault(struct kvm_vcpu *vcpu,
1580 struct x86_exception *fault);
1581 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1582 gfn_t gfn, void *data, int offset, int len,
1584 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
1585 bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr);
1587 static inline int __kvm_irq_line_state(unsigned long *irq_state,
1588 int irq_source_id, int level)
1590 /* Logical OR for level trig interrupt */
1592 __set_bit(irq_source_id, irq_state);
1594 __clear_bit(irq_source_id, irq_state);
1596 return !!(*irq_state);
1599 #define KVM_MMU_ROOT_CURRENT BIT(0)
1600 #define KVM_MMU_ROOT_PREVIOUS(i) BIT(1+i)
1601 #define KVM_MMU_ROOTS_ALL (~0UL)
1603 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
1604 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
1606 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
1608 void kvm_update_dr7(struct kvm_vcpu *vcpu);
1610 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
1611 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
1612 void kvm_mmu_free_roots(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1613 ulong roots_to_free);
1614 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1615 struct x86_exception *exception);
1616 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
1617 struct x86_exception *exception);
1618 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
1619 struct x86_exception *exception);
1620 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
1621 struct x86_exception *exception);
1622 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
1623 struct x86_exception *exception);
1625 bool kvm_apicv_activated(struct kvm *kvm);
1626 void kvm_apicv_init(struct kvm *kvm, bool enable);
1627 void kvm_vcpu_update_apicv(struct kvm_vcpu *vcpu);
1628 void kvm_request_apicv_update(struct kvm *kvm, bool activate,
1631 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
1633 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u64 error_code,
1634 void *insn, int insn_len);
1635 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
1636 void kvm_mmu_invalidate_gva(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1637 gva_t gva, hpa_t root_hpa);
1638 void kvm_mmu_invpcid_gva(struct kvm_vcpu *vcpu, gva_t gva, unsigned long pcid);
1639 void kvm_mmu_new_pgd(struct kvm_vcpu *vcpu, gpa_t new_pgd, bool skip_tlb_flush,
1640 bool skip_mmu_sync);
1642 void kvm_configure_mmu(bool enable_tdp, int tdp_max_root_level,
1643 int tdp_huge_page_level);
1645 static inline u16 kvm_read_ldt(void)
1648 asm("sldt %0" : "=g"(ldt));
1652 static inline void kvm_load_ldt(u16 sel)
1654 asm("lldt %0" : : "rm"(sel));
1657 #ifdef CONFIG_X86_64
1658 static inline unsigned long read_msr(unsigned long msr)
1667 static inline u32 get_rdx_init_val(void)
1669 return 0x600; /* P6 family */
1672 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
1674 kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
1677 #define TSS_IOPB_BASE_OFFSET 0x66
1678 #define TSS_BASE_SIZE 0x68
1679 #define TSS_IOPB_SIZE (65536 / 8)
1680 #define TSS_REDIRECTION_SIZE (256 / 8)
1681 #define RMODE_TSS_SIZE \
1682 (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
1685 TASK_SWITCH_CALL = 0,
1686 TASK_SWITCH_IRET = 1,
1687 TASK_SWITCH_JMP = 2,
1688 TASK_SWITCH_GATE = 3,
1691 #define HF_GIF_MASK (1 << 0)
1692 #define HF_NMI_MASK (1 << 3)
1693 #define HF_IRET_MASK (1 << 4)
1694 #define HF_GUEST_MASK (1 << 5) /* VCPU is in guest-mode */
1695 #define HF_SMM_MASK (1 << 6)
1696 #define HF_SMM_INSIDE_NMI_MASK (1 << 7)
1698 #define __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
1699 #define KVM_ADDRESS_SPACE_NUM 2
1701 #define kvm_arch_vcpu_memslots_id(vcpu) ((vcpu)->arch.hflags & HF_SMM_MASK ? 1 : 0)
1702 #define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, (role).smm)
1704 asmlinkage void kvm_spurious_fault(void);
1707 * Hardware virtualization extension instructions may fault if a
1708 * reboot turns off virtualization while processes are running.
1709 * Usually after catching the fault we just panic; during reboot
1710 * instead the instruction is ignored.
1712 #define __kvm_handle_fault_on_reboot(insn) \
1718 ".pushsection .discard.instr_begin \n\t" \
1719 ".long 1b - . \n\t" \
1720 ".popsection \n\t" \
1721 "call kvm_spurious_fault \n\t" \
1723 ".pushsection .discard.instr_end \n\t" \
1724 ".long 1b - . \n\t" \
1725 ".popsection \n\t" \
1727 _ASM_EXTABLE(666b, 667b)
1729 #define KVM_ARCH_WANT_MMU_NOTIFIER
1731 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1732 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1733 int kvm_cpu_has_extint(struct kvm_vcpu *v);
1734 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1735 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1736 void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event);
1737 void kvm_vcpu_reload_apic_access_page(struct kvm_vcpu *vcpu);
1739 int kvm_pv_send_ipi(struct kvm *kvm, unsigned long ipi_bitmap_low,
1740 unsigned long ipi_bitmap_high, u32 min,
1741 unsigned long icr, int op_64_bit);
1743 void kvm_define_user_return_msr(unsigned index, u32 msr);
1744 int kvm_set_user_return_msr(unsigned index, u64 val, u64 mask);
1746 u64 kvm_scale_tsc(struct kvm_vcpu *vcpu, u64 tsc);
1747 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc);
1749 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu);
1750 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1752 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
1753 void kvm_make_scan_ioapic_request(struct kvm *kvm);
1754 void kvm_make_scan_ioapic_request_mask(struct kvm *kvm,
1755 unsigned long *vcpu_bitmap);
1757 bool kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1758 struct kvm_async_pf *work);
1759 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1760 struct kvm_async_pf *work);
1761 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1762 struct kvm_async_pf *work);
1763 void kvm_arch_async_page_present_queued(struct kvm_vcpu *vcpu);
1764 bool kvm_arch_can_dequeue_async_page_present(struct kvm_vcpu *vcpu);
1765 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1767 int kvm_skip_emulated_instruction(struct kvm_vcpu *vcpu);
1768 int kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1769 void __kvm_request_immediate_exit(struct kvm_vcpu *vcpu);
1771 int kvm_is_in_guest(void);
1773 void __user *__x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa,
1775 bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu);
1776 bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu);
1778 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
1779 struct kvm_vcpu **dest_vcpu);
1781 void kvm_set_msi_irq(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e,
1782 struct kvm_lapic_irq *irq);
1784 static inline bool kvm_irq_is_postable(struct kvm_lapic_irq *irq)
1786 /* We can only post Fixed and LowPrio IRQs */
1787 return (irq->delivery_mode == APIC_DM_FIXED ||
1788 irq->delivery_mode == APIC_DM_LOWEST);
1791 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu)
1793 static_call_cond(kvm_x86_vcpu_blocking)(vcpu);
1796 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu)
1798 static_call_cond(kvm_x86_vcpu_unblocking)(vcpu);
1801 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
1803 static inline int kvm_cpu_get_apicid(int mps_cpu)
1805 #ifdef CONFIG_X86_LOCAL_APIC
1806 return default_cpu_present_to_apicid(mps_cpu);
1813 #define put_smstate(type, buf, offset, val) \
1814 *(type *)((buf) + (offset) - 0x7e00) = val
1816 #define GET_SMSTATE(type, buf, offset) \
1817 (*(type *)((buf) + (offset) - 0x7e00))
1819 int kvm_cpu_dirty_log_size(void);
1821 #endif /* _ASM_X86_KVM_HOST_H */