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 PFERR_PRESENT_BIT 0
225 #define PFERR_WRITE_BIT 1
226 #define PFERR_USER_BIT 2
227 #define PFERR_RSVD_BIT 3
228 #define PFERR_FETCH_BIT 4
229 #define PFERR_PK_BIT 5
230 #define PFERR_GUEST_FINAL_BIT 32
231 #define PFERR_GUEST_PAGE_BIT 33
233 #define PFERR_PRESENT_MASK (1U << PFERR_PRESENT_BIT)
234 #define PFERR_WRITE_MASK (1U << PFERR_WRITE_BIT)
235 #define PFERR_USER_MASK (1U << PFERR_USER_BIT)
236 #define PFERR_RSVD_MASK (1U << PFERR_RSVD_BIT)
237 #define PFERR_FETCH_MASK (1U << PFERR_FETCH_BIT)
238 #define PFERR_PK_MASK (1U << PFERR_PK_BIT)
239 #define PFERR_GUEST_FINAL_MASK (1ULL << PFERR_GUEST_FINAL_BIT)
240 #define PFERR_GUEST_PAGE_MASK (1ULL << PFERR_GUEST_PAGE_BIT)
242 #define PFERR_NESTED_GUEST_PAGE (PFERR_GUEST_PAGE_MASK | \
246 /* apic attention bits */
247 #define KVM_APIC_CHECK_VAPIC 0
249 * The following bit is set with PV-EOI, unset on EOI.
250 * We detect PV-EOI changes by guest by comparing
251 * this bit with PV-EOI in guest memory.
252 * See the implementation in apic_update_pv_eoi.
254 #define KVM_APIC_PV_EOI_PENDING 1
256 struct kvm_kernel_irq_routing_entry;
259 * the pages used as guest page table on soft mmu are tracked by
260 * kvm_memory_slot.arch.gfn_track which is 16 bits, so the role bits used
261 * by indirect shadow page can not be more than 15 bits.
263 * Currently, we used 14 bits that are @level, @gpte_is_8_bytes, @quadrant, @access,
264 * @nxe, @cr0_wp, @smep_andnot_wp and @smap_andnot_wp.
266 union kvm_mmu_page_role {
270 unsigned gpte_is_8_bytes:1;
277 unsigned smep_andnot_wp:1;
278 unsigned smap_andnot_wp:1;
279 unsigned ad_disabled:1;
280 unsigned guest_mode:1;
284 * This is left at the top of the word so that
285 * kvm_memslots_for_spte_role can extract it with a
286 * simple shift. While there is room, give it a whole
287 * byte so it is also faster to load it from memory.
293 union kvm_mmu_extended_role {
295 * This structure complements kvm_mmu_page_role caching everything needed for
296 * MMU configuration. If nothing in both these structures changed, MMU
297 * re-configuration can be skipped. @valid bit is set on first usage so we don't
298 * treat all-zero structure as valid data.
302 unsigned int valid:1;
303 unsigned int execonly:1;
304 unsigned int cr0_pg:1;
305 unsigned int cr4_pae:1;
306 unsigned int cr4_pse:1;
307 unsigned int cr4_pke:1;
308 unsigned int cr4_smap:1;
309 unsigned int cr4_smep:1;
310 unsigned int maxphyaddr:6;
317 union kvm_mmu_page_role base;
318 union kvm_mmu_extended_role ext;
322 struct kvm_rmap_head {
326 struct kvm_pio_request {
327 unsigned long linear_rip;
334 #define PT64_ROOT_MAX_LEVEL 5
336 struct rsvd_bits_validate {
337 u64 rsvd_bits_mask[2][PT64_ROOT_MAX_LEVEL];
341 struct kvm_mmu_root_info {
346 #define KVM_MMU_ROOT_INFO_INVALID \
347 ((struct kvm_mmu_root_info) { .pgd = INVALID_PAGE, .hpa = INVALID_PAGE })
349 #define KVM_MMU_NUM_PREV_ROOTS 3
351 #define KVM_HAVE_MMU_RWLOCK
356 * x86 supports 4 paging modes (5-level 64-bit, 4-level 64-bit, 3-level 32-bit,
357 * and 2-level 32-bit). The kvm_mmu structure abstracts the details of the
361 unsigned long (*get_guest_pgd)(struct kvm_vcpu *vcpu);
362 u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
363 int (*page_fault)(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u32 err,
365 void (*inject_page_fault)(struct kvm_vcpu *vcpu,
366 struct x86_exception *fault);
367 gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gpa_t gva_or_gpa,
368 u32 access, struct x86_exception *exception);
369 gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
370 struct x86_exception *exception);
371 int (*sync_page)(struct kvm_vcpu *vcpu,
372 struct kvm_mmu_page *sp);
373 void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva, hpa_t root_hpa);
376 union kvm_mmu_role mmu_role;
378 u8 shadow_root_level;
381 struct kvm_mmu_root_info prev_roots[KVM_MMU_NUM_PREV_ROOTS];
384 * Bitmap; bit set = permission fault
385 * Byte index: page fault error code [4:1]
386 * Bit index: pte permissions in ACC_* format
391 * The pkru_mask indicates if protection key checks are needed. It
392 * consists of 16 domains indexed by page fault error code bits [4:1],
393 * with PFEC.RSVD replaced by ACC_USER_MASK from the page tables.
394 * Each domain has 2 bits which are ANDed with AD and WD from PKRU.
402 * check zero bits on shadow page table entries, these
403 * bits include not only hardware reserved bits but also
404 * the bits spte never used.
406 struct rsvd_bits_validate shadow_zero_check;
408 struct rsvd_bits_validate guest_rsvd_check;
410 /* Can have large pages at levels 2..last_nonleaf_level-1. */
411 u8 last_nonleaf_level;
415 u64 pdptrs[4]; /* pae */
418 struct kvm_tlb_range {
433 struct perf_event *perf_event;
434 struct kvm_vcpu *vcpu;
436 * eventsel value for general purpose counters,
437 * ctrl value for fixed counters.
443 unsigned nr_arch_gp_counters;
444 unsigned nr_arch_fixed_counters;
445 unsigned available_event_types;
450 u64 counter_bitmask[2];
451 u64 global_ctrl_mask;
452 u64 global_ovf_ctrl_mask;
455 struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
456 struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED];
457 struct irq_work irq_work;
458 DECLARE_BITMAP(reprogram_pmi, X86_PMC_IDX_MAX);
459 DECLARE_BITMAP(all_valid_pmc_idx, X86_PMC_IDX_MAX);
460 DECLARE_BITMAP(pmc_in_use, X86_PMC_IDX_MAX);
463 * The gate to release perf_events not marked in
464 * pmc_in_use only once in a vcpu time slice.
469 * The total number of programmed perf_events and it helps to avoid
470 * redundant check before cleanup if guest don't use vPMU at all.
478 KVM_DEBUGREG_BP_ENABLED = 1,
479 KVM_DEBUGREG_WONT_EXIT = 2,
480 KVM_DEBUGREG_RELOAD = 4,
483 struct kvm_mtrr_range {
486 struct list_head node;
490 struct kvm_mtrr_range var_ranges[KVM_NR_VAR_MTRR];
491 mtrr_type fixed_ranges[KVM_NR_FIXED_MTRR_REGION];
494 struct list_head head;
497 /* Hyper-V SynIC timer */
498 struct kvm_vcpu_hv_stimer {
499 struct hrtimer timer;
501 union hv_stimer_config config;
504 struct hv_message msg;
508 /* Hyper-V synthetic interrupt controller (SynIC)*/
509 struct kvm_vcpu_hv_synic {
514 atomic64_t sint[HV_SYNIC_SINT_COUNT];
515 atomic_t sint_to_gsi[HV_SYNIC_SINT_COUNT];
516 DECLARE_BITMAP(auto_eoi_bitmap, 256);
517 DECLARE_BITMAP(vec_bitmap, 256);
519 bool dont_zero_synic_pages;
522 /* Hyper-V per vcpu emulation context */
524 struct kvm_vcpu *vcpu;
528 struct kvm_vcpu_hv_synic synic;
529 struct kvm_hyperv_exit exit;
530 struct kvm_vcpu_hv_stimer stimer[HV_SYNIC_STIMER_COUNT];
531 DECLARE_BITMAP(stimer_pending_bitmap, HV_SYNIC_STIMER_COUNT);
535 /* Xen HVM per vcpu emulation context */
536 struct kvm_vcpu_xen {
538 u32 current_runstate;
540 bool vcpu_time_info_set;
542 struct gfn_to_hva_cache vcpu_info_cache;
543 struct gfn_to_hva_cache vcpu_time_info_cache;
544 struct gfn_to_hva_cache runstate_cache;
546 u64 runstate_entry_time;
547 u64 runstate_times[4];
550 struct kvm_vcpu_arch {
552 * rip and regs accesses must go through
553 * kvm_{register,rip}_{read,write} functions.
555 unsigned long regs[NR_VCPU_REGS];
560 unsigned long cr0_guest_owned_bits;
564 unsigned long cr4_guest_owned_bits;
565 unsigned long cr4_guest_rsvd_bits;
572 struct kvm_lapic *apic; /* kernel irqchip context */
574 bool load_eoi_exitmap_pending;
575 DECLARE_BITMAP(ioapic_handled_vectors, 256);
576 unsigned long apic_attention;
577 int32_t apic_arb_prio;
579 u64 ia32_misc_enable_msr;
582 bool tpr_access_reporting;
585 u64 microcode_version;
586 u64 arch_capabilities;
587 u64 perf_capabilities;
590 * Paging state of the vcpu
592 * If the vcpu runs in guest mode with two level paging this still saves
593 * the paging mode of the l1 guest. This context is always used to
598 /* Non-nested MMU for L1 */
599 struct kvm_mmu root_mmu;
601 /* L1 MMU when running nested */
602 struct kvm_mmu guest_mmu;
605 * Paging state of an L2 guest (used for nested npt)
607 * This context will save all necessary information to walk page tables
608 * of an L2 guest. This context is only initialized for page table
609 * walking and not for faulting since we never handle l2 page faults on
612 struct kvm_mmu nested_mmu;
615 * Pointer to the mmu context currently used for
616 * gva_to_gpa translations.
618 struct kvm_mmu *walk_mmu;
620 struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
621 struct kvm_mmu_memory_cache mmu_shadow_page_cache;
622 struct kvm_mmu_memory_cache mmu_gfn_array_cache;
623 struct kvm_mmu_memory_cache mmu_page_header_cache;
626 * QEMU userspace and the guest each have their own FPU state.
627 * In vcpu_run, we switch between the user and guest FPU contexts.
628 * While running a VCPU, the VCPU thread will have the guest FPU
631 * Note that while the PKRU state lives inside the fpu registers,
632 * it is switched out separately at VMENTER and VMEXIT time. The
633 * "guest_fpu" state here contains the guest FPU context, with the
636 struct fpu *user_fpu;
637 struct fpu *guest_fpu;
640 u64 guest_supported_xcr0;
642 struct kvm_pio_request pio;
644 void *guest_ins_data;
646 u8 event_exit_inst_len;
648 struct kvm_queued_exception {
654 unsigned long payload;
659 struct kvm_queued_interrupt {
665 int halt_request; /* real mode on Intel only */
668 struct kvm_cpuid_entry2 *cpuid_entries;
670 u64 reserved_gpa_bits;
674 /* emulate context */
676 struct x86_emulate_ctxt *emulate_ctxt;
677 bool emulate_regs_need_sync_to_vcpu;
678 bool emulate_regs_need_sync_from_vcpu;
679 int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
682 struct pvclock_vcpu_time_info hv_clock;
683 unsigned int hw_tsc_khz;
684 struct gfn_to_hva_cache pv_time;
685 bool pv_time_enabled;
686 /* set guest stopped flag in pvclock flags field */
687 bool pvclock_set_guest_stopped_request;
693 struct gfn_to_pfn_cache cache;
700 u64 tsc_offset_adjustment;
703 u64 this_tsc_generation;
705 bool tsc_always_catchup;
706 s8 virtual_tsc_shift;
707 u32 virtual_tsc_mult;
709 s64 ia32_tsc_adjust_msr;
710 u64 msr_ia32_power_ctl;
711 u64 tsc_scaling_ratio;
713 atomic_t nmi_queued; /* unprocessed asynchronous NMIs */
714 unsigned nmi_pending; /* NMI queued after currently running handler */
715 bool nmi_injected; /* Trying to inject an NMI this entry */
716 bool smi_pending; /* SMI queued after currently running handler */
718 struct kvm_mtrr mtrr_state;
721 unsigned switch_db_regs;
722 unsigned long db[KVM_NR_DB_REGS];
725 unsigned long eff_db[KVM_NR_DB_REGS];
726 unsigned long guest_debug_dr7;
727 u64 msr_platform_info;
728 u64 msr_misc_features_enables;
736 /* Cache MMIO info */
738 unsigned mmio_access;
744 /* used for guest single stepping over the given code position */
745 unsigned long singlestep_rip;
748 struct kvm_vcpu_hv *hyperv;
749 struct kvm_vcpu_xen xen;
751 cpumask_var_t wbinvd_dirty_mask;
753 unsigned long last_retry_eip;
754 unsigned long last_retry_addr;
758 gfn_t gfns[ASYNC_PF_PER_VCPU];
759 struct gfn_to_hva_cache data;
760 u64 msr_en_val; /* MSR_KVM_ASYNC_PF_EN */
761 u64 msr_int_val; /* MSR_KVM_ASYNC_PF_INT */
766 unsigned long nested_apf_token;
767 bool delivery_as_pf_vmexit;
768 bool pageready_pending;
771 /* OSVW MSRs (AMD only) */
779 struct gfn_to_hva_cache data;
782 u64 msr_kvm_poll_control;
785 * Indicates the guest is trying to write a gfn that contains one or
786 * more of the PTEs used to translate the write itself, i.e. the access
787 * is changing its own translation in the guest page tables. KVM exits
788 * to userspace if emulation of the faulting instruction fails and this
789 * flag is set, as KVM cannot make forward progress.
791 * If emulation fails for a write to guest page tables, KVM unprotects
792 * (zaps) the shadow page for the target gfn and resumes the guest to
793 * retry the non-emulatable instruction (on hardware). Unprotecting the
794 * gfn doesn't allow forward progress for a self-changing access because
795 * doing so also zaps the translation for the gfn, i.e. retrying the
796 * instruction will hit a !PRESENT fault, which results in a new shadow
797 * page and sends KVM back to square one.
799 bool write_fault_to_shadow_pgtable;
801 /* set at EPT violation at this point */
802 unsigned long exit_qualification;
804 /* pv related host specific info */
809 int pending_ioapic_eoi;
810 int pending_external_vector;
812 /* be preempted when it's in kernel-mode(cpl=0) */
813 bool preempted_in_kernel;
815 /* Flush the L1 Data cache for L1TF mitigation on VMENTER */
818 /* Host CPU on which VM-entry was most recently attempted */
819 unsigned int last_vmentry_cpu;
821 /* AMD MSRC001_0015 Hardware Configuration */
824 /* pv related cpuid info */
827 * value of the eax register in the KVM_CPUID_FEATURES CPUID
833 * indicates whether pv emulation should be disabled if features
834 * are not present in the guest's cpuid
839 /* Protected Guests */
840 bool guest_state_protected;
843 struct kvm_lpage_info {
847 struct kvm_arch_memory_slot {
848 struct kvm_rmap_head *rmap[KVM_NR_PAGE_SIZES];
849 struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
850 unsigned short *gfn_track[KVM_PAGE_TRACK_MAX];
854 * We use as the mode the number of bits allocated in the LDR for the
855 * logical processor ID. It happens that these are all powers of two.
856 * This makes it is very easy to detect cases where the APICs are
857 * configured for multiple modes; in that case, we cannot use the map and
858 * hence cannot use kvm_irq_delivery_to_apic_fast either.
860 #define KVM_APIC_MODE_XAPIC_CLUSTER 4
861 #define KVM_APIC_MODE_XAPIC_FLAT 8
862 #define KVM_APIC_MODE_X2APIC 16
864 struct kvm_apic_map {
869 struct kvm_lapic *xapic_flat_map[8];
870 struct kvm_lapic *xapic_cluster_map[16][4];
872 struct kvm_lapic *phys_map[];
875 /* Hyper-V synthetic debugger (SynDbg)*/
876 struct kvm_hv_syndbg {
887 /* Current state of Hyper-V TSC page clocksource */
888 enum hv_tsc_page_status {
889 /* TSC page was not set up or disabled */
890 HV_TSC_PAGE_UNSET = 0,
891 /* TSC page MSR was written by the guest, update pending */
892 HV_TSC_PAGE_GUEST_CHANGED,
893 /* TSC page MSR was written by KVM userspace, update pending */
894 HV_TSC_PAGE_HOST_CHANGED,
895 /* TSC page was properly set up and is currently active */
897 /* TSC page is currently being updated and therefore is inactive */
898 HV_TSC_PAGE_UPDATING,
899 /* TSC page was set up with an inaccessible GPA */
903 /* Hyper-V emulation context */
905 struct mutex hv_lock;
909 enum hv_tsc_page_status hv_tsc_page_status;
911 /* Hyper-v based guest crash (NT kernel bugcheck) parameters */
912 u64 hv_crash_param[HV_X64_MSR_CRASH_PARAMS];
915 struct ms_hyperv_tsc_page tsc_ref;
917 struct idr conn_to_evt;
919 u64 hv_reenlightenment_control;
920 u64 hv_tsc_emulation_control;
921 u64 hv_tsc_emulation_status;
923 /* How many vCPUs have VP index != vCPU index */
924 atomic_t num_mismatched_vp_indexes;
926 struct hv_partition_assist_pg *hv_pa_pg;
927 struct kvm_hv_syndbg hv_syndbg;
930 struct msr_bitmap_range {
934 unsigned long *bitmap;
937 /* Xen emulation context */
942 struct gfn_to_hva_cache shinfo_cache;
945 enum kvm_irqchip_mode {
947 KVM_IRQCHIP_KERNEL, /* created with KVM_CREATE_IRQCHIP */
948 KVM_IRQCHIP_SPLIT, /* created with KVM_CAP_SPLIT_IRQCHIP */
951 struct kvm_x86_msr_filter {
953 bool default_allow:1;
954 struct msr_bitmap_range ranges[16];
957 #define APICV_INHIBIT_REASON_DISABLE 0
958 #define APICV_INHIBIT_REASON_HYPERV 1
959 #define APICV_INHIBIT_REASON_NESTED 2
960 #define APICV_INHIBIT_REASON_IRQWIN 3
961 #define APICV_INHIBIT_REASON_PIT_REINJ 4
962 #define APICV_INHIBIT_REASON_X2APIC 5
965 unsigned long n_used_mmu_pages;
966 unsigned long n_requested_mmu_pages;
967 unsigned long n_max_mmu_pages;
968 unsigned int indirect_shadow_pages;
970 struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
971 struct list_head active_mmu_pages;
972 struct list_head zapped_obsolete_pages;
973 struct list_head lpage_disallowed_mmu_pages;
974 struct kvm_page_track_notifier_node mmu_sp_tracker;
975 struct kvm_page_track_notifier_head track_notifier_head;
977 struct list_head assigned_dev_head;
978 struct iommu_domain *iommu_domain;
979 bool iommu_noncoherent;
980 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA
981 atomic_t noncoherent_dma_count;
982 #define __KVM_HAVE_ARCH_ASSIGNED_DEVICE
983 atomic_t assigned_device_count;
984 struct kvm_pic *vpic;
985 struct kvm_ioapic *vioapic;
986 struct kvm_pit *vpit;
987 atomic_t vapics_in_nmi_mode;
988 struct mutex apic_map_lock;
989 struct kvm_apic_map __rcu *apic_map;
990 atomic_t apic_map_dirty;
992 bool apic_access_page_done;
993 unsigned long apicv_inhibit_reasons;
1000 bool cstate_in_guest;
1002 unsigned long irq_sources_bitmap;
1003 s64 kvmclock_offset;
1004 raw_spinlock_t tsc_write_lock;
1011 u64 cur_tsc_generation;
1012 int nr_vcpus_matched_tsc;
1014 spinlock_t pvclock_gtod_sync_lock;
1015 bool use_master_clock;
1016 u64 master_kernel_ns;
1017 u64 master_cycle_now;
1018 struct delayed_work kvmclock_update_work;
1019 struct delayed_work kvmclock_sync_work;
1021 struct kvm_xen_hvm_config xen_hvm_config;
1023 /* reads protected by irq_srcu, writes by irq_lock */
1024 struct hlist_head mask_notifier_list;
1026 struct kvm_hv hyperv;
1029 #ifdef CONFIG_KVM_MMU_AUDIT
1033 bool backwards_tsc_observed;
1034 bool boot_vcpu_runs_old_kvmclock;
1037 u64 disabled_quirks;
1038 int cpu_dirty_logging_count;
1040 enum kvm_irqchip_mode irqchip_mode;
1041 u8 nr_reserved_ioapic_pins;
1043 bool disabled_lapic_found;
1046 bool x2apic_broadcast_quirk_disabled;
1048 bool guest_can_read_msr_platform_info;
1049 bool exception_payload_enabled;
1051 bool bus_lock_detection_enabled;
1053 /* Deflect RDMSR and WRMSR to user space when they trigger a #GP */
1054 u32 user_space_msr_mask;
1055 struct kvm_x86_msr_filter __rcu *msr_filter;
1057 struct kvm_pmu_event_filter __rcu *pmu_event_filter;
1058 struct task_struct *nx_lpage_recovery_thread;
1060 #ifdef CONFIG_X86_64
1062 * Whether the TDP MMU is enabled for this VM. This contains a
1063 * snapshot of the TDP MMU module parameter from when the VM was
1064 * created and remains unchanged for the life of the VM. If this is
1065 * true, TDP MMU handler functions will run for various MMU
1068 bool tdp_mmu_enabled;
1071 * List of struct kvmp_mmu_pages being used as roots.
1072 * All struct kvm_mmu_pages in the list should have
1074 * All struct kvm_mmu_pages in the list should have a positive
1075 * root_count except when a thread holds the MMU lock and is removing
1076 * an entry from the list.
1078 struct list_head tdp_mmu_roots;
1081 * List of struct kvmp_mmu_pages not being used as roots.
1082 * All struct kvm_mmu_pages in the list should have
1083 * tdp_mmu_page set and a root_count of 0.
1085 struct list_head tdp_mmu_pages;
1088 * Protects accesses to the following fields when the MMU lock
1089 * is held in read mode:
1090 * - tdp_mmu_pages (above)
1091 * - the link field of struct kvm_mmu_pages used by the TDP MMU
1092 * - lpage_disallowed_mmu_pages
1093 * - the lpage_disallowed_link field of struct kvm_mmu_pages used
1095 * It is acceptable, but not necessary, to acquire this lock when
1096 * the thread holds the MMU lock in write mode.
1098 spinlock_t tdp_mmu_pages_lock;
1099 #endif /* CONFIG_X86_64 */
1102 struct kvm_vm_stat {
1103 ulong mmu_shadow_zapped;
1104 ulong mmu_pte_write;
1105 ulong mmu_pde_zapped;
1108 ulong mmu_cache_miss;
1110 ulong remote_tlb_flush;
1112 ulong nx_lpage_splits;
1113 ulong max_mmu_page_hash_collisions;
1116 struct kvm_vcpu_stat {
1126 u64 irq_window_exits;
1127 u64 nmi_window_exits;
1130 u64 halt_successful_poll;
1131 u64 halt_attempted_poll;
1132 u64 halt_poll_invalid;
1134 u64 request_irq_exits;
1136 u64 host_state_reload;
1139 u64 insn_emulation_fail;
1144 u64 halt_poll_success_ns;
1145 u64 halt_poll_fail_ns;
1148 struct x86_instruction_info;
1151 bool host_initiated;
1156 struct kvm_lapic_irq {
1164 bool msi_redir_hint;
1167 static inline u16 kvm_lapic_irq_dest_mode(bool dest_mode_logical)
1169 return dest_mode_logical ? APIC_DEST_LOGICAL : APIC_DEST_PHYSICAL;
1172 struct kvm_x86_ops {
1173 int (*hardware_enable)(void);
1174 void (*hardware_disable)(void);
1175 void (*hardware_unsetup)(void);
1176 bool (*cpu_has_accelerated_tpr)(void);
1177 bool (*has_emulated_msr)(struct kvm *kvm, u32 index);
1178 void (*vcpu_after_set_cpuid)(struct kvm_vcpu *vcpu);
1180 unsigned int vm_size;
1181 int (*vm_init)(struct kvm *kvm);
1182 void (*vm_destroy)(struct kvm *kvm);
1184 /* Create, but do not attach this VCPU */
1185 int (*vcpu_create)(struct kvm_vcpu *vcpu);
1186 void (*vcpu_free)(struct kvm_vcpu *vcpu);
1187 void (*vcpu_reset)(struct kvm_vcpu *vcpu, bool init_event);
1189 void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
1190 void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
1191 void (*vcpu_put)(struct kvm_vcpu *vcpu);
1193 void (*update_exception_bitmap)(struct kvm_vcpu *vcpu);
1194 int (*get_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1195 int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1196 u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
1197 void (*get_segment)(struct kvm_vcpu *vcpu,
1198 struct kvm_segment *var, int seg);
1199 int (*get_cpl)(struct kvm_vcpu *vcpu);
1200 void (*set_segment)(struct kvm_vcpu *vcpu,
1201 struct kvm_segment *var, int seg);
1202 void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
1203 void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
1204 bool (*is_valid_cr4)(struct kvm_vcpu *vcpu, unsigned long cr0);
1205 void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
1206 int (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
1207 void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1208 void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1209 void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1210 void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1211 void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu);
1212 void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
1213 void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
1214 unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
1215 void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
1217 void (*tlb_flush_all)(struct kvm_vcpu *vcpu);
1218 void (*tlb_flush_current)(struct kvm_vcpu *vcpu);
1219 int (*tlb_remote_flush)(struct kvm *kvm);
1220 int (*tlb_remote_flush_with_range)(struct kvm *kvm,
1221 struct kvm_tlb_range *range);
1224 * Flush any TLB entries associated with the given GVA.
1225 * Does not need to flush GPA->HPA mappings.
1226 * Can potentially get non-canonical addresses through INVLPGs, which
1227 * the implementation may choose to ignore if appropriate.
1229 void (*tlb_flush_gva)(struct kvm_vcpu *vcpu, gva_t addr);
1232 * Flush any TLB entries created by the guest. Like tlb_flush_gva(),
1233 * does not need to flush GPA->HPA mappings.
1235 void (*tlb_flush_guest)(struct kvm_vcpu *vcpu);
1237 enum exit_fastpath_completion (*run)(struct kvm_vcpu *vcpu);
1238 int (*handle_exit)(struct kvm_vcpu *vcpu,
1239 enum exit_fastpath_completion exit_fastpath);
1240 int (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
1241 void (*update_emulated_instruction)(struct kvm_vcpu *vcpu);
1242 void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
1243 u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu);
1244 void (*patch_hypercall)(struct kvm_vcpu *vcpu,
1245 unsigned char *hypercall_addr);
1246 void (*set_irq)(struct kvm_vcpu *vcpu);
1247 void (*set_nmi)(struct kvm_vcpu *vcpu);
1248 void (*queue_exception)(struct kvm_vcpu *vcpu);
1249 void (*cancel_injection)(struct kvm_vcpu *vcpu);
1250 int (*interrupt_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1251 int (*nmi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1252 bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
1253 void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
1254 void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
1255 void (*enable_irq_window)(struct kvm_vcpu *vcpu);
1256 void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
1257 bool (*check_apicv_inhibit_reasons)(ulong bit);
1258 void (*pre_update_apicv_exec_ctrl)(struct kvm *kvm, bool activate);
1259 void (*refresh_apicv_exec_ctrl)(struct kvm_vcpu *vcpu);
1260 void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
1261 void (*hwapic_isr_update)(struct kvm_vcpu *vcpu, int isr);
1262 bool (*guest_apic_has_interrupt)(struct kvm_vcpu *vcpu);
1263 void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
1264 void (*set_virtual_apic_mode)(struct kvm_vcpu *vcpu);
1265 void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu);
1266 int (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
1267 int (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
1268 int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
1269 int (*set_identity_map_addr)(struct kvm *kvm, u64 ident_addr);
1270 u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
1272 void (*load_mmu_pgd)(struct kvm_vcpu *vcpu, unsigned long pgd,
1275 bool (*has_wbinvd_exit)(void);
1277 /* Returns actual tsc_offset set in active VMCS */
1278 u64 (*write_l1_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
1281 * Retrieve somewhat arbitrary exit information. Intended to be used
1282 * only from within tracepoints to avoid VMREADs when tracing is off.
1284 void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2,
1285 u32 *exit_int_info, u32 *exit_int_info_err_code);
1287 int (*check_intercept)(struct kvm_vcpu *vcpu,
1288 struct x86_instruction_info *info,
1289 enum x86_intercept_stage stage,
1290 struct x86_exception *exception);
1291 void (*handle_exit_irqoff)(struct kvm_vcpu *vcpu);
1293 void (*request_immediate_exit)(struct kvm_vcpu *vcpu);
1295 void (*sched_in)(struct kvm_vcpu *kvm, int cpu);
1298 * Size of the CPU's dirty log buffer, i.e. VMX's PML buffer. A zero
1299 * value indicates CPU dirty logging is unsupported or disabled.
1301 int cpu_dirty_log_size;
1302 void (*update_cpu_dirty_logging)(struct kvm_vcpu *vcpu);
1304 /* pmu operations of sub-arch */
1305 const struct kvm_pmu_ops *pmu_ops;
1306 const struct kvm_x86_nested_ops *nested_ops;
1309 * Architecture specific hooks for vCPU blocking due to
1311 * Returns for .pre_block():
1312 * - 0 means continue to block the vCPU.
1313 * - 1 means we cannot block the vCPU since some event
1314 * happens during this period, such as, 'ON' bit in
1315 * posted-interrupts descriptor is set.
1317 int (*pre_block)(struct kvm_vcpu *vcpu);
1318 void (*post_block)(struct kvm_vcpu *vcpu);
1320 void (*vcpu_blocking)(struct kvm_vcpu *vcpu);
1321 void (*vcpu_unblocking)(struct kvm_vcpu *vcpu);
1323 int (*update_pi_irte)(struct kvm *kvm, unsigned int host_irq,
1324 uint32_t guest_irq, bool set);
1325 void (*apicv_post_state_restore)(struct kvm_vcpu *vcpu);
1326 bool (*dy_apicv_has_pending_interrupt)(struct kvm_vcpu *vcpu);
1328 int (*set_hv_timer)(struct kvm_vcpu *vcpu, u64 guest_deadline_tsc,
1330 void (*cancel_hv_timer)(struct kvm_vcpu *vcpu);
1332 void (*setup_mce)(struct kvm_vcpu *vcpu);
1334 int (*smi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1335 int (*pre_enter_smm)(struct kvm_vcpu *vcpu, char *smstate);
1336 int (*pre_leave_smm)(struct kvm_vcpu *vcpu, const char *smstate);
1337 void (*enable_smi_window)(struct kvm_vcpu *vcpu);
1339 int (*mem_enc_op)(struct kvm *kvm, void __user *argp);
1340 int (*mem_enc_reg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1341 int (*mem_enc_unreg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1343 int (*get_msr_feature)(struct kvm_msr_entry *entry);
1345 bool (*can_emulate_instruction)(struct kvm_vcpu *vcpu, void *insn, int insn_len);
1347 bool (*apic_init_signal_blocked)(struct kvm_vcpu *vcpu);
1348 int (*enable_direct_tlbflush)(struct kvm_vcpu *vcpu);
1350 void (*migrate_timers)(struct kvm_vcpu *vcpu);
1351 void (*msr_filter_changed)(struct kvm_vcpu *vcpu);
1352 int (*complete_emulated_msr)(struct kvm_vcpu *vcpu, int err);
1354 void (*vcpu_deliver_sipi_vector)(struct kvm_vcpu *vcpu, u8 vector);
1357 struct kvm_x86_nested_ops {
1358 int (*check_events)(struct kvm_vcpu *vcpu);
1359 bool (*hv_timer_pending)(struct kvm_vcpu *vcpu);
1360 int (*get_state)(struct kvm_vcpu *vcpu,
1361 struct kvm_nested_state __user *user_kvm_nested_state,
1362 unsigned user_data_size);
1363 int (*set_state)(struct kvm_vcpu *vcpu,
1364 struct kvm_nested_state __user *user_kvm_nested_state,
1365 struct kvm_nested_state *kvm_state);
1366 bool (*get_nested_state_pages)(struct kvm_vcpu *vcpu);
1367 int (*write_log_dirty)(struct kvm_vcpu *vcpu, gpa_t l2_gpa);
1369 int (*enable_evmcs)(struct kvm_vcpu *vcpu,
1370 uint16_t *vmcs_version);
1371 uint16_t (*get_evmcs_version)(struct kvm_vcpu *vcpu);
1374 struct kvm_x86_init_ops {
1375 int (*cpu_has_kvm_support)(void);
1376 int (*disabled_by_bios)(void);
1377 int (*check_processor_compatibility)(void);
1378 int (*hardware_setup)(void);
1380 struct kvm_x86_ops *runtime_ops;
1383 struct kvm_arch_async_pf {
1390 extern u64 __read_mostly host_efer;
1391 extern bool __read_mostly allow_smaller_maxphyaddr;
1392 extern struct kvm_x86_ops kvm_x86_ops;
1394 #define KVM_X86_OP(func) \
1395 DECLARE_STATIC_CALL(kvm_x86_##func, *(((struct kvm_x86_ops *)0)->func));
1396 #define KVM_X86_OP_NULL KVM_X86_OP
1397 #include <asm/kvm-x86-ops.h>
1399 static inline void kvm_ops_static_call_update(void)
1401 #define KVM_X86_OP(func) \
1402 static_call_update(kvm_x86_##func, kvm_x86_ops.func);
1403 #define KVM_X86_OP_NULL KVM_X86_OP
1404 #include <asm/kvm-x86-ops.h>
1407 #define __KVM_HAVE_ARCH_VM_ALLOC
1408 static inline struct kvm *kvm_arch_alloc_vm(void)
1410 return __vmalloc(kvm_x86_ops.vm_size, GFP_KERNEL_ACCOUNT | __GFP_ZERO);
1412 void kvm_arch_free_vm(struct kvm *kvm);
1414 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
1415 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
1417 if (kvm_x86_ops.tlb_remote_flush &&
1418 !static_call(kvm_x86_tlb_remote_flush)(kvm))
1424 int kvm_mmu_module_init(void);
1425 void kvm_mmu_module_exit(void);
1427 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
1428 int kvm_mmu_create(struct kvm_vcpu *vcpu);
1429 void kvm_mmu_init_vm(struct kvm *kvm);
1430 void kvm_mmu_uninit_vm(struct kvm *kvm);
1431 void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
1432 u64 dirty_mask, u64 nx_mask, u64 x_mask, u64 p_mask,
1433 u64 acc_track_mask, u64 me_mask);
1435 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
1436 void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
1437 struct kvm_memory_slot *memslot,
1439 void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
1440 const struct kvm_memory_slot *memslot);
1441 void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
1442 struct kvm_memory_slot *memslot);
1443 void kvm_mmu_slot_largepage_remove_write_access(struct kvm *kvm,
1444 struct kvm_memory_slot *memslot);
1445 void kvm_mmu_zap_all(struct kvm *kvm);
1446 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, u64 gen);
1447 unsigned long kvm_mmu_calculate_default_mmu_pages(struct kvm *kvm);
1448 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned long kvm_nr_mmu_pages);
1450 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
1451 bool pdptrs_changed(struct kvm_vcpu *vcpu);
1453 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
1454 const void *val, int bytes);
1456 struct kvm_irq_mask_notifier {
1457 void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
1459 struct hlist_node link;
1462 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
1463 struct kvm_irq_mask_notifier *kimn);
1464 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
1465 struct kvm_irq_mask_notifier *kimn);
1466 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
1469 extern bool tdp_enabled;
1471 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
1473 /* control of guest tsc rate supported? */
1474 extern bool kvm_has_tsc_control;
1475 /* maximum supported tsc_khz for guests */
1476 extern u32 kvm_max_guest_tsc_khz;
1477 /* number of bits of the fractional part of the TSC scaling ratio */
1478 extern u8 kvm_tsc_scaling_ratio_frac_bits;
1479 /* maximum allowed value of TSC scaling ratio */
1480 extern u64 kvm_max_tsc_scaling_ratio;
1481 /* 1ull << kvm_tsc_scaling_ratio_frac_bits */
1482 extern u64 kvm_default_tsc_scaling_ratio;
1483 /* bus lock detection supported? */
1484 extern bool kvm_has_bus_lock_exit;
1486 extern u64 kvm_mce_cap_supported;
1489 * EMULTYPE_NO_DECODE - Set when re-emulating an instruction (after completing
1490 * userspace I/O) to indicate that the emulation context
1491 * should be resued as is, i.e. skip initialization of
1492 * emulation context, instruction fetch and decode.
1494 * EMULTYPE_TRAP_UD - Set when emulating an intercepted #UD from hardware.
1495 * Indicates that only select instructions (tagged with
1496 * EmulateOnUD) should be emulated (to minimize the emulator
1497 * attack surface). See also EMULTYPE_TRAP_UD_FORCED.
1499 * EMULTYPE_SKIP - Set when emulating solely to skip an instruction, i.e. to
1500 * decode the instruction length. For use *only* by
1501 * kvm_x86_ops.skip_emulated_instruction() implementations.
1503 * EMULTYPE_ALLOW_RETRY_PF - Set when the emulator should resume the guest to
1504 * retry native execution under certain conditions,
1505 * Can only be set in conjunction with EMULTYPE_PF.
1507 * EMULTYPE_TRAP_UD_FORCED - Set when emulating an intercepted #UD that was
1508 * triggered by KVM's magic "force emulation" prefix,
1509 * which is opt in via module param (off by default).
1510 * Bypasses EmulateOnUD restriction despite emulating
1511 * due to an intercepted #UD (see EMULTYPE_TRAP_UD).
1512 * Used to test the full emulator from userspace.
1514 * EMULTYPE_VMWARE_GP - Set when emulating an intercepted #GP for VMware
1515 * backdoor emulation, which is opt in via module param.
1516 * VMware backoor emulation handles select instructions
1517 * and reinjects the #GP for all other cases.
1519 * EMULTYPE_PF - Set when emulating MMIO by way of an intercepted #PF, in which
1520 * case the CR2/GPA value pass on the stack is valid.
1522 #define EMULTYPE_NO_DECODE (1 << 0)
1523 #define EMULTYPE_TRAP_UD (1 << 1)
1524 #define EMULTYPE_SKIP (1 << 2)
1525 #define EMULTYPE_ALLOW_RETRY_PF (1 << 3)
1526 #define EMULTYPE_TRAP_UD_FORCED (1 << 4)
1527 #define EMULTYPE_VMWARE_GP (1 << 5)
1528 #define EMULTYPE_PF (1 << 6)
1530 int kvm_emulate_instruction(struct kvm_vcpu *vcpu, int emulation_type);
1531 int kvm_emulate_instruction_from_buffer(struct kvm_vcpu *vcpu,
1532 void *insn, int insn_len);
1534 void kvm_enable_efer_bits(u64);
1535 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
1536 int __kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data, bool host_initiated);
1537 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data);
1538 int kvm_set_msr(struct kvm_vcpu *vcpu, u32 index, u64 data);
1539 int kvm_emulate_rdmsr(struct kvm_vcpu *vcpu);
1540 int kvm_emulate_wrmsr(struct kvm_vcpu *vcpu);
1542 int kvm_fast_pio(struct kvm_vcpu *vcpu, int size, unsigned short port, int in);
1543 int kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
1544 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
1545 int kvm_vcpu_halt(struct kvm_vcpu *vcpu);
1546 int kvm_emulate_ap_reset_hold(struct kvm_vcpu *vcpu);
1547 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
1549 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
1550 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
1551 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector);
1553 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
1554 int reason, bool has_error_code, u32 error_code);
1556 void kvm_free_guest_fpu(struct kvm_vcpu *vcpu);
1558 void kvm_post_set_cr0(struct kvm_vcpu *vcpu, unsigned long old_cr0, unsigned long cr0);
1559 void kvm_post_set_cr4(struct kvm_vcpu *vcpu, unsigned long old_cr4, unsigned long cr4);
1560 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
1561 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
1562 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
1563 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
1564 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
1565 void kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
1566 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
1567 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
1568 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
1569 int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr);
1571 int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1572 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1574 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
1575 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
1576 bool kvm_rdpmc(struct kvm_vcpu *vcpu);
1578 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1579 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1580 void kvm_queue_exception_p(struct kvm_vcpu *vcpu, unsigned nr, unsigned long payload);
1581 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1582 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1583 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
1584 bool kvm_inject_emulated_page_fault(struct kvm_vcpu *vcpu,
1585 struct x86_exception *fault);
1586 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1587 gfn_t gfn, void *data, int offset, int len,
1589 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
1590 bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr);
1592 static inline int __kvm_irq_line_state(unsigned long *irq_state,
1593 int irq_source_id, int level)
1595 /* Logical OR for level trig interrupt */
1597 __set_bit(irq_source_id, irq_state);
1599 __clear_bit(irq_source_id, irq_state);
1601 return !!(*irq_state);
1604 #define KVM_MMU_ROOT_CURRENT BIT(0)
1605 #define KVM_MMU_ROOT_PREVIOUS(i) BIT(1+i)
1606 #define KVM_MMU_ROOTS_ALL (~0UL)
1608 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
1609 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
1611 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
1613 void kvm_update_dr7(struct kvm_vcpu *vcpu);
1615 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
1616 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
1617 int kvm_mmu_load(struct kvm_vcpu *vcpu);
1618 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
1619 void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu);
1620 void kvm_mmu_free_roots(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1621 ulong roots_to_free);
1622 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1623 struct x86_exception *exception);
1624 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
1625 struct x86_exception *exception);
1626 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
1627 struct x86_exception *exception);
1628 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
1629 struct x86_exception *exception);
1630 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
1631 struct x86_exception *exception);
1633 bool kvm_apicv_activated(struct kvm *kvm);
1634 void kvm_apicv_init(struct kvm *kvm, bool enable);
1635 void kvm_vcpu_update_apicv(struct kvm_vcpu *vcpu);
1636 void kvm_request_apicv_update(struct kvm *kvm, bool activate,
1639 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
1641 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u64 error_code,
1642 void *insn, int insn_len);
1643 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
1644 void kvm_mmu_invalidate_gva(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1645 gva_t gva, hpa_t root_hpa);
1646 void kvm_mmu_invpcid_gva(struct kvm_vcpu *vcpu, gva_t gva, unsigned long pcid);
1647 void kvm_mmu_new_pgd(struct kvm_vcpu *vcpu, gpa_t new_pgd, bool skip_tlb_flush,
1648 bool skip_mmu_sync);
1650 void kvm_configure_mmu(bool enable_tdp, int tdp_max_root_level,
1651 int tdp_huge_page_level);
1653 static inline u16 kvm_read_ldt(void)
1656 asm("sldt %0" : "=g"(ldt));
1660 static inline void kvm_load_ldt(u16 sel)
1662 asm("lldt %0" : : "rm"(sel));
1665 #ifdef CONFIG_X86_64
1666 static inline unsigned long read_msr(unsigned long msr)
1675 static inline u32 get_rdx_init_val(void)
1677 return 0x600; /* P6 family */
1680 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
1682 kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
1685 #define TSS_IOPB_BASE_OFFSET 0x66
1686 #define TSS_BASE_SIZE 0x68
1687 #define TSS_IOPB_SIZE (65536 / 8)
1688 #define TSS_REDIRECTION_SIZE (256 / 8)
1689 #define RMODE_TSS_SIZE \
1690 (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
1693 TASK_SWITCH_CALL = 0,
1694 TASK_SWITCH_IRET = 1,
1695 TASK_SWITCH_JMP = 2,
1696 TASK_SWITCH_GATE = 3,
1699 #define HF_GIF_MASK (1 << 0)
1700 #define HF_NMI_MASK (1 << 3)
1701 #define HF_IRET_MASK (1 << 4)
1702 #define HF_GUEST_MASK (1 << 5) /* VCPU is in guest-mode */
1703 #define HF_SMM_MASK (1 << 6)
1704 #define HF_SMM_INSIDE_NMI_MASK (1 << 7)
1706 #define __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
1707 #define KVM_ADDRESS_SPACE_NUM 2
1709 #define kvm_arch_vcpu_memslots_id(vcpu) ((vcpu)->arch.hflags & HF_SMM_MASK ? 1 : 0)
1710 #define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, (role).smm)
1712 asmlinkage void kvm_spurious_fault(void);
1715 * Hardware virtualization extension instructions may fault if a
1716 * reboot turns off virtualization while processes are running.
1717 * Usually after catching the fault we just panic; during reboot
1718 * instead the instruction is ignored.
1720 #define __kvm_handle_fault_on_reboot(insn) \
1726 ".pushsection .discard.instr_begin \n\t" \
1727 ".long 1b - . \n\t" \
1728 ".popsection \n\t" \
1729 "call kvm_spurious_fault \n\t" \
1731 ".pushsection .discard.instr_end \n\t" \
1732 ".long 1b - . \n\t" \
1733 ".popsection \n\t" \
1735 _ASM_EXTABLE(666b, 667b)
1737 #define KVM_ARCH_WANT_MMU_NOTIFIER
1738 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end,
1740 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
1741 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
1742 int kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
1743 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1744 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1745 int kvm_cpu_has_extint(struct kvm_vcpu *v);
1746 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1747 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1748 void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event);
1749 void kvm_vcpu_reload_apic_access_page(struct kvm_vcpu *vcpu);
1751 int kvm_pv_send_ipi(struct kvm *kvm, unsigned long ipi_bitmap_low,
1752 unsigned long ipi_bitmap_high, u32 min,
1753 unsigned long icr, int op_64_bit);
1755 void kvm_define_user_return_msr(unsigned index, u32 msr);
1756 int kvm_set_user_return_msr(unsigned index, u64 val, u64 mask);
1758 u64 kvm_scale_tsc(struct kvm_vcpu *vcpu, u64 tsc);
1759 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc);
1761 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu);
1762 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1764 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
1765 void kvm_make_scan_ioapic_request(struct kvm *kvm);
1766 void kvm_make_scan_ioapic_request_mask(struct kvm *kvm,
1767 unsigned long *vcpu_bitmap);
1769 bool kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1770 struct kvm_async_pf *work);
1771 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1772 struct kvm_async_pf *work);
1773 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1774 struct kvm_async_pf *work);
1775 void kvm_arch_async_page_present_queued(struct kvm_vcpu *vcpu);
1776 bool kvm_arch_can_dequeue_async_page_present(struct kvm_vcpu *vcpu);
1777 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1779 int kvm_skip_emulated_instruction(struct kvm_vcpu *vcpu);
1780 int kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1781 void __kvm_request_immediate_exit(struct kvm_vcpu *vcpu);
1783 int kvm_is_in_guest(void);
1785 void __user *__x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa,
1787 bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu);
1788 bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu);
1790 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
1791 struct kvm_vcpu **dest_vcpu);
1793 void kvm_set_msi_irq(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e,
1794 struct kvm_lapic_irq *irq);
1796 static inline bool kvm_irq_is_postable(struct kvm_lapic_irq *irq)
1798 /* We can only post Fixed and LowPrio IRQs */
1799 return (irq->delivery_mode == APIC_DM_FIXED ||
1800 irq->delivery_mode == APIC_DM_LOWEST);
1803 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu)
1805 static_call_cond(kvm_x86_vcpu_blocking)(vcpu);
1808 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu)
1810 static_call_cond(kvm_x86_vcpu_unblocking)(vcpu);
1813 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
1815 static inline int kvm_cpu_get_apicid(int mps_cpu)
1817 #ifdef CONFIG_X86_LOCAL_APIC
1818 return default_cpu_present_to_apicid(mps_cpu);
1825 #define put_smstate(type, buf, offset, val) \
1826 *(type *)((buf) + (offset) - 0x7e00) = val
1828 #define GET_SMSTATE(type, buf, offset) \
1829 (*(type *)((buf) + (offset) - 0x7e00))
1831 int kvm_cpu_dirty_log_size(void);
1833 #endif /* _ASM_X86_KVM_HOST_H */