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 #define KVM_USER_MEM_SLOTS 509
44 /* memory slots that are not exposed to userspace */
45 #define KVM_PRIVATE_MEM_SLOTS 3
46 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
48 #define KVM_HALT_POLL_NS_DEFAULT 200000
50 #define KVM_IRQCHIP_NUM_PINS KVM_IOAPIC_NUM_PINS
52 #define KVM_DIRTY_LOG_MANUAL_CAPS (KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE | \
53 KVM_DIRTY_LOG_INITIALLY_SET)
55 #define KVM_BUS_LOCK_DETECTION_VALID_MODE (KVM_BUS_LOCK_DETECTION_OFF | \
56 KVM_BUS_LOCK_DETECTION_EXIT)
58 /* x86-specific vcpu->requests bit members */
59 #define KVM_REQ_MIGRATE_TIMER KVM_ARCH_REQ(0)
60 #define KVM_REQ_REPORT_TPR_ACCESS KVM_ARCH_REQ(1)
61 #define KVM_REQ_TRIPLE_FAULT KVM_ARCH_REQ(2)
62 #define KVM_REQ_MMU_SYNC KVM_ARCH_REQ(3)
63 #define KVM_REQ_CLOCK_UPDATE KVM_ARCH_REQ(4)
64 #define KVM_REQ_LOAD_MMU_PGD KVM_ARCH_REQ(5)
65 #define KVM_REQ_EVENT KVM_ARCH_REQ(6)
66 #define KVM_REQ_APF_HALT KVM_ARCH_REQ(7)
67 #define KVM_REQ_STEAL_UPDATE KVM_ARCH_REQ(8)
68 #define KVM_REQ_NMI KVM_ARCH_REQ(9)
69 #define KVM_REQ_PMU KVM_ARCH_REQ(10)
70 #define KVM_REQ_PMI KVM_ARCH_REQ(11)
71 #define KVM_REQ_SMI KVM_ARCH_REQ(12)
72 #define KVM_REQ_MASTERCLOCK_UPDATE KVM_ARCH_REQ(13)
73 #define KVM_REQ_MCLOCK_INPROGRESS \
74 KVM_ARCH_REQ_FLAGS(14, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
75 #define KVM_REQ_SCAN_IOAPIC \
76 KVM_ARCH_REQ_FLAGS(15, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
77 #define KVM_REQ_GLOBAL_CLOCK_UPDATE KVM_ARCH_REQ(16)
78 #define KVM_REQ_APIC_PAGE_RELOAD \
79 KVM_ARCH_REQ_FLAGS(17, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
80 #define KVM_REQ_HV_CRASH KVM_ARCH_REQ(18)
81 #define KVM_REQ_IOAPIC_EOI_EXIT KVM_ARCH_REQ(19)
82 #define KVM_REQ_HV_RESET KVM_ARCH_REQ(20)
83 #define KVM_REQ_HV_EXIT KVM_ARCH_REQ(21)
84 #define KVM_REQ_HV_STIMER KVM_ARCH_REQ(22)
85 #define KVM_REQ_LOAD_EOI_EXITMAP KVM_ARCH_REQ(23)
86 #define KVM_REQ_GET_NESTED_STATE_PAGES KVM_ARCH_REQ(24)
87 #define KVM_REQ_APICV_UPDATE \
88 KVM_ARCH_REQ_FLAGS(25, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
89 #define KVM_REQ_TLB_FLUSH_CURRENT KVM_ARCH_REQ(26)
90 #define KVM_REQ_HV_TLB_FLUSH \
91 KVM_ARCH_REQ_FLAGS(27, KVM_REQUEST_NO_WAKEUP)
92 #define KVM_REQ_APF_READY KVM_ARCH_REQ(28)
93 #define KVM_REQ_MSR_FILTER_CHANGED KVM_ARCH_REQ(29)
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
354 * x86 supports 4 paging modes (5-level 64-bit, 4-level 64-bit, 3-level 32-bit,
355 * and 2-level 32-bit). The kvm_mmu structure abstracts the details of the
359 unsigned long (*get_guest_pgd)(struct kvm_vcpu *vcpu);
360 u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
361 int (*page_fault)(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u32 err,
363 void (*inject_page_fault)(struct kvm_vcpu *vcpu,
364 struct x86_exception *fault);
365 gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gpa_t gva_or_gpa,
366 u32 access, struct x86_exception *exception);
367 gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
368 struct x86_exception *exception);
369 int (*sync_page)(struct kvm_vcpu *vcpu,
370 struct kvm_mmu_page *sp);
371 void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva, hpa_t root_hpa);
374 union kvm_mmu_role mmu_role;
376 u8 shadow_root_level;
379 struct kvm_mmu_root_info prev_roots[KVM_MMU_NUM_PREV_ROOTS];
382 * Bitmap; bit set = permission fault
383 * Byte index: page fault error code [4:1]
384 * Bit index: pte permissions in ACC_* format
389 * The pkru_mask indicates if protection key checks are needed. It
390 * consists of 16 domains indexed by page fault error code bits [4:1],
391 * with PFEC.RSVD replaced by ACC_USER_MASK from the page tables.
392 * Each domain has 2 bits which are ANDed with AD and WD from PKRU.
400 * check zero bits on shadow page table entries, these
401 * bits include not only hardware reserved bits but also
402 * the bits spte never used.
404 struct rsvd_bits_validate shadow_zero_check;
406 struct rsvd_bits_validate guest_rsvd_check;
408 /* Can have large pages at levels 2..last_nonleaf_level-1. */
409 u8 last_nonleaf_level;
413 u64 pdptrs[4]; /* pae */
416 struct kvm_tlb_range {
431 struct perf_event *perf_event;
432 struct kvm_vcpu *vcpu;
434 * eventsel value for general purpose counters,
435 * ctrl value for fixed counters.
441 unsigned nr_arch_gp_counters;
442 unsigned nr_arch_fixed_counters;
443 unsigned available_event_types;
448 u64 counter_bitmask[2];
449 u64 global_ctrl_mask;
450 u64 global_ovf_ctrl_mask;
453 struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
454 struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED];
455 struct irq_work irq_work;
456 DECLARE_BITMAP(reprogram_pmi, X86_PMC_IDX_MAX);
457 DECLARE_BITMAP(all_valid_pmc_idx, X86_PMC_IDX_MAX);
458 DECLARE_BITMAP(pmc_in_use, X86_PMC_IDX_MAX);
461 * The gate to release perf_events not marked in
462 * pmc_in_use only once in a vcpu time slice.
467 * The total number of programmed perf_events and it helps to avoid
468 * redundant check before cleanup if guest don't use vPMU at all.
476 KVM_DEBUGREG_BP_ENABLED = 1,
477 KVM_DEBUGREG_WONT_EXIT = 2,
478 KVM_DEBUGREG_RELOAD = 4,
481 struct kvm_mtrr_range {
484 struct list_head node;
488 struct kvm_mtrr_range var_ranges[KVM_NR_VAR_MTRR];
489 mtrr_type fixed_ranges[KVM_NR_FIXED_MTRR_REGION];
492 struct list_head head;
495 /* Hyper-V SynIC timer */
496 struct kvm_vcpu_hv_stimer {
497 struct hrtimer timer;
499 union hv_stimer_config config;
502 struct hv_message msg;
506 /* Hyper-V synthetic interrupt controller (SynIC)*/
507 struct kvm_vcpu_hv_synic {
512 atomic64_t sint[HV_SYNIC_SINT_COUNT];
513 atomic_t sint_to_gsi[HV_SYNIC_SINT_COUNT];
514 DECLARE_BITMAP(auto_eoi_bitmap, 256);
515 DECLARE_BITMAP(vec_bitmap, 256);
517 bool dont_zero_synic_pages;
520 /* Hyper-V per vcpu emulation context */
525 struct kvm_vcpu_hv_synic synic;
526 struct kvm_hyperv_exit exit;
527 struct kvm_vcpu_hv_stimer stimer[HV_SYNIC_STIMER_COUNT];
528 DECLARE_BITMAP(stimer_pending_bitmap, HV_SYNIC_STIMER_COUNT);
532 struct kvm_vcpu_arch {
534 * rip and regs accesses must go through
535 * kvm_{register,rip}_{read,write} functions.
537 unsigned long regs[NR_VCPU_REGS];
542 unsigned long cr0_guest_owned_bits;
546 unsigned long cr4_guest_owned_bits;
547 unsigned long cr4_guest_rsvd_bits;
554 struct kvm_lapic *apic; /* kernel irqchip context */
556 bool load_eoi_exitmap_pending;
557 DECLARE_BITMAP(ioapic_handled_vectors, 256);
558 unsigned long apic_attention;
559 int32_t apic_arb_prio;
561 u64 ia32_misc_enable_msr;
564 bool tpr_access_reporting;
567 u64 microcode_version;
568 u64 arch_capabilities;
569 u64 perf_capabilities;
572 * Paging state of the vcpu
574 * If the vcpu runs in guest mode with two level paging this still saves
575 * the paging mode of the l1 guest. This context is always used to
580 /* Non-nested MMU for L1 */
581 struct kvm_mmu root_mmu;
583 /* L1 MMU when running nested */
584 struct kvm_mmu guest_mmu;
587 * Paging state of an L2 guest (used for nested npt)
589 * This context will save all necessary information to walk page tables
590 * of an L2 guest. This context is only initialized for page table
591 * walking and not for faulting since we never handle l2 page faults on
594 struct kvm_mmu nested_mmu;
597 * Pointer to the mmu context currently used for
598 * gva_to_gpa translations.
600 struct kvm_mmu *walk_mmu;
602 struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
603 struct kvm_mmu_memory_cache mmu_shadow_page_cache;
604 struct kvm_mmu_memory_cache mmu_gfn_array_cache;
605 struct kvm_mmu_memory_cache mmu_page_header_cache;
608 * QEMU userspace and the guest each have their own FPU state.
609 * In vcpu_run, we switch between the user and guest FPU contexts.
610 * While running a VCPU, the VCPU thread will have the guest FPU
613 * Note that while the PKRU state lives inside the fpu registers,
614 * it is switched out separately at VMENTER and VMEXIT time. The
615 * "guest_fpu" state here contains the guest FPU context, with the
618 struct fpu *user_fpu;
619 struct fpu *guest_fpu;
622 u64 guest_supported_xcr0;
624 struct kvm_pio_request pio;
626 void *guest_ins_data;
628 u8 event_exit_inst_len;
630 struct kvm_queued_exception {
636 unsigned long payload;
641 struct kvm_queued_interrupt {
647 int halt_request; /* real mode on Intel only */
650 struct kvm_cpuid_entry2 *cpuid_entries;
652 unsigned long cr3_lm_rsvd_bits;
656 /* emulate context */
658 struct x86_emulate_ctxt *emulate_ctxt;
659 bool emulate_regs_need_sync_to_vcpu;
660 bool emulate_regs_need_sync_from_vcpu;
661 int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
664 struct pvclock_vcpu_time_info hv_clock;
665 unsigned int hw_tsc_khz;
666 struct gfn_to_hva_cache pv_time;
667 bool pv_time_enabled;
668 /* set guest stopped flag in pvclock flags field */
669 bool pvclock_set_guest_stopped_request;
675 struct gfn_to_pfn_cache cache;
682 u64 tsc_offset_adjustment;
685 u64 this_tsc_generation;
687 bool tsc_always_catchup;
688 s8 virtual_tsc_shift;
689 u32 virtual_tsc_mult;
691 s64 ia32_tsc_adjust_msr;
692 u64 msr_ia32_power_ctl;
693 u64 tsc_scaling_ratio;
695 atomic_t nmi_queued; /* unprocessed asynchronous NMIs */
696 unsigned nmi_pending; /* NMI queued after currently running handler */
697 bool nmi_injected; /* Trying to inject an NMI this entry */
698 bool smi_pending; /* SMI queued after currently running handler */
700 struct kvm_mtrr mtrr_state;
703 unsigned switch_db_regs;
704 unsigned long db[KVM_NR_DB_REGS];
707 unsigned long eff_db[KVM_NR_DB_REGS];
708 unsigned long guest_debug_dr7;
709 u64 msr_platform_info;
710 u64 msr_misc_features_enables;
718 /* Cache MMIO info */
720 unsigned mmio_access;
726 /* used for guest single stepping over the given code position */
727 unsigned long singlestep_rip;
729 struct kvm_vcpu_hv hyperv;
731 cpumask_var_t wbinvd_dirty_mask;
733 unsigned long last_retry_eip;
734 unsigned long last_retry_addr;
738 gfn_t gfns[ASYNC_PF_PER_VCPU];
739 struct gfn_to_hva_cache data;
740 u64 msr_en_val; /* MSR_KVM_ASYNC_PF_EN */
741 u64 msr_int_val; /* MSR_KVM_ASYNC_PF_INT */
746 unsigned long nested_apf_token;
747 bool delivery_as_pf_vmexit;
748 bool pageready_pending;
751 /* OSVW MSRs (AMD only) */
759 struct gfn_to_hva_cache data;
762 u64 msr_kvm_poll_control;
765 * Indicates the guest is trying to write a gfn that contains one or
766 * more of the PTEs used to translate the write itself, i.e. the access
767 * is changing its own translation in the guest page tables. KVM exits
768 * to userspace if emulation of the faulting instruction fails and this
769 * flag is set, as KVM cannot make forward progress.
771 * If emulation fails for a write to guest page tables, KVM unprotects
772 * (zaps) the shadow page for the target gfn and resumes the guest to
773 * retry the non-emulatable instruction (on hardware). Unprotecting the
774 * gfn doesn't allow forward progress for a self-changing access because
775 * doing so also zaps the translation for the gfn, i.e. retrying the
776 * instruction will hit a !PRESENT fault, which results in a new shadow
777 * page and sends KVM back to square one.
779 bool write_fault_to_shadow_pgtable;
781 /* set at EPT violation at this point */
782 unsigned long exit_qualification;
784 /* pv related host specific info */
789 int pending_ioapic_eoi;
790 int pending_external_vector;
792 /* be preempted when it's in kernel-mode(cpl=0) */
793 bool preempted_in_kernel;
795 /* Flush the L1 Data cache for L1TF mitigation on VMENTER */
798 /* Host CPU on which VM-entry was most recently attempted */
799 unsigned int last_vmentry_cpu;
801 /* AMD MSRC001_0015 Hardware Configuration */
804 /* pv related cpuid info */
807 * value of the eax register in the KVM_CPUID_FEATURES CPUID
813 * indicates whether pv emulation should be disabled if features
814 * are not present in the guest's cpuid
819 /* Protected Guests */
820 bool guest_state_protected;
823 struct kvm_lpage_info {
827 struct kvm_arch_memory_slot {
828 struct kvm_rmap_head *rmap[KVM_NR_PAGE_SIZES];
829 struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
830 unsigned short *gfn_track[KVM_PAGE_TRACK_MAX];
834 * We use as the mode the number of bits allocated in the LDR for the
835 * logical processor ID. It happens that these are all powers of two.
836 * This makes it is very easy to detect cases where the APICs are
837 * configured for multiple modes; in that case, we cannot use the map and
838 * hence cannot use kvm_irq_delivery_to_apic_fast either.
840 #define KVM_APIC_MODE_XAPIC_CLUSTER 4
841 #define KVM_APIC_MODE_XAPIC_FLAT 8
842 #define KVM_APIC_MODE_X2APIC 16
844 struct kvm_apic_map {
849 struct kvm_lapic *xapic_flat_map[8];
850 struct kvm_lapic *xapic_cluster_map[16][4];
852 struct kvm_lapic *phys_map[];
855 /* Hyper-V synthetic debugger (SynDbg)*/
856 struct kvm_hv_syndbg {
867 /* Hyper-V emulation context */
869 struct mutex hv_lock;
874 /* Hyper-v based guest crash (NT kernel bugcheck) parameters */
875 u64 hv_crash_param[HV_X64_MSR_CRASH_PARAMS];
878 struct ms_hyperv_tsc_page tsc_ref;
880 struct idr conn_to_evt;
882 u64 hv_reenlightenment_control;
883 u64 hv_tsc_emulation_control;
884 u64 hv_tsc_emulation_status;
886 /* How many vCPUs have VP index != vCPU index */
887 atomic_t num_mismatched_vp_indexes;
889 struct hv_partition_assist_pg *hv_pa_pg;
890 struct kvm_hv_syndbg hv_syndbg;
893 struct msr_bitmap_range {
897 unsigned long *bitmap;
900 enum kvm_irqchip_mode {
902 KVM_IRQCHIP_KERNEL, /* created with KVM_CREATE_IRQCHIP */
903 KVM_IRQCHIP_SPLIT, /* created with KVM_CAP_SPLIT_IRQCHIP */
906 #define APICV_INHIBIT_REASON_DISABLE 0
907 #define APICV_INHIBIT_REASON_HYPERV 1
908 #define APICV_INHIBIT_REASON_NESTED 2
909 #define APICV_INHIBIT_REASON_IRQWIN 3
910 #define APICV_INHIBIT_REASON_PIT_REINJ 4
911 #define APICV_INHIBIT_REASON_X2APIC 5
914 unsigned long n_used_mmu_pages;
915 unsigned long n_requested_mmu_pages;
916 unsigned long n_max_mmu_pages;
917 unsigned int indirect_shadow_pages;
919 struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
921 * Hash table of struct kvm_mmu_page.
923 struct list_head active_mmu_pages;
924 struct list_head zapped_obsolete_pages;
925 struct list_head lpage_disallowed_mmu_pages;
926 struct kvm_page_track_notifier_node mmu_sp_tracker;
927 struct kvm_page_track_notifier_head track_notifier_head;
929 struct list_head assigned_dev_head;
930 struct iommu_domain *iommu_domain;
931 bool iommu_noncoherent;
932 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA
933 atomic_t noncoherent_dma_count;
934 #define __KVM_HAVE_ARCH_ASSIGNED_DEVICE
935 atomic_t assigned_device_count;
936 struct kvm_pic *vpic;
937 struct kvm_ioapic *vioapic;
938 struct kvm_pit *vpit;
939 atomic_t vapics_in_nmi_mode;
940 struct mutex apic_map_lock;
941 struct kvm_apic_map *apic_map;
942 atomic_t apic_map_dirty;
944 bool apic_access_page_done;
945 unsigned long apicv_inhibit_reasons;
952 bool cstate_in_guest;
954 unsigned long irq_sources_bitmap;
956 raw_spinlock_t tsc_write_lock;
963 u64 cur_tsc_generation;
964 int nr_vcpus_matched_tsc;
966 spinlock_t pvclock_gtod_sync_lock;
967 bool use_master_clock;
968 u64 master_kernel_ns;
969 u64 master_cycle_now;
970 struct delayed_work kvmclock_update_work;
971 struct delayed_work kvmclock_sync_work;
973 struct kvm_xen_hvm_config xen_hvm_config;
975 /* reads protected by irq_srcu, writes by irq_lock */
976 struct hlist_head mask_notifier_list;
978 struct kvm_hv hyperv;
980 #ifdef CONFIG_KVM_MMU_AUDIT
984 bool backwards_tsc_observed;
985 bool boot_vcpu_runs_old_kvmclock;
990 enum kvm_irqchip_mode irqchip_mode;
991 u8 nr_reserved_ioapic_pins;
993 bool disabled_lapic_found;
996 bool x2apic_broadcast_quirk_disabled;
998 bool guest_can_read_msr_platform_info;
999 bool exception_payload_enabled;
1001 /* Deflect RDMSR and WRMSR to user space when they trigger a #GP */
1002 u32 user_space_msr_mask;
1006 bool default_allow:1;
1007 struct msr_bitmap_range ranges[16];
1010 bool bus_lock_detection_enabled;
1012 struct kvm_pmu_event_filter *pmu_event_filter;
1013 struct task_struct *nx_lpage_recovery_thread;
1016 * Whether the TDP MMU is enabled for this VM. This contains a
1017 * snapshot of the TDP MMU module parameter from when the VM was
1018 * created and remains unchanged for the life of the VM. If this is
1019 * true, TDP MMU handler functions will run for various MMU
1022 bool tdp_mmu_enabled;
1025 * List of struct kvmp_mmu_pages being used as roots.
1026 * All struct kvm_mmu_pages in the list should have
1028 * All struct kvm_mmu_pages in the list should have a positive
1029 * root_count except when a thread holds the MMU lock and is removing
1030 * an entry from the list.
1032 struct list_head tdp_mmu_roots;
1035 * List of struct kvmp_mmu_pages not being used as roots.
1036 * All struct kvm_mmu_pages in the list should have
1037 * tdp_mmu_page set and a root_count of 0.
1039 struct list_head tdp_mmu_pages;
1042 struct kvm_vm_stat {
1043 ulong mmu_shadow_zapped;
1044 ulong mmu_pte_write;
1045 ulong mmu_pde_zapped;
1048 ulong mmu_cache_miss;
1050 ulong remote_tlb_flush;
1052 ulong nx_lpage_splits;
1053 ulong max_mmu_page_hash_collisions;
1056 struct kvm_vcpu_stat {
1066 u64 irq_window_exits;
1067 u64 nmi_window_exits;
1070 u64 halt_successful_poll;
1071 u64 halt_attempted_poll;
1072 u64 halt_poll_invalid;
1074 u64 request_irq_exits;
1076 u64 host_state_reload;
1079 u64 insn_emulation_fail;
1084 u64 halt_poll_success_ns;
1085 u64 halt_poll_fail_ns;
1088 struct x86_instruction_info;
1091 bool host_initiated;
1096 struct kvm_lapic_irq {
1104 bool msi_redir_hint;
1107 static inline u16 kvm_lapic_irq_dest_mode(bool dest_mode_logical)
1109 return dest_mode_logical ? APIC_DEST_LOGICAL : APIC_DEST_PHYSICAL;
1112 struct kvm_x86_ops {
1113 int (*hardware_enable)(void);
1114 void (*hardware_disable)(void);
1115 void (*hardware_unsetup)(void);
1116 bool (*cpu_has_accelerated_tpr)(void);
1117 bool (*has_emulated_msr)(struct kvm *kvm, u32 index);
1118 void (*vcpu_after_set_cpuid)(struct kvm_vcpu *vcpu);
1120 unsigned int vm_size;
1121 int (*vm_init)(struct kvm *kvm);
1122 void (*vm_destroy)(struct kvm *kvm);
1124 /* Create, but do not attach this VCPU */
1125 int (*vcpu_create)(struct kvm_vcpu *vcpu);
1126 void (*vcpu_free)(struct kvm_vcpu *vcpu);
1127 void (*vcpu_reset)(struct kvm_vcpu *vcpu, bool init_event);
1129 void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
1130 void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
1131 void (*vcpu_put)(struct kvm_vcpu *vcpu);
1133 void (*update_exception_bitmap)(struct kvm_vcpu *vcpu);
1134 int (*get_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1135 int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1136 u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
1137 void (*get_segment)(struct kvm_vcpu *vcpu,
1138 struct kvm_segment *var, int seg);
1139 int (*get_cpl)(struct kvm_vcpu *vcpu);
1140 void (*set_segment)(struct kvm_vcpu *vcpu,
1141 struct kvm_segment *var, int seg);
1142 void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
1143 void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
1144 bool (*is_valid_cr4)(struct kvm_vcpu *vcpu, unsigned long cr0);
1145 void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
1146 int (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
1147 void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1148 void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1149 void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1150 void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1151 void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu);
1152 void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
1153 void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
1154 unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
1155 void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
1157 void (*tlb_flush_all)(struct kvm_vcpu *vcpu);
1158 void (*tlb_flush_current)(struct kvm_vcpu *vcpu);
1159 int (*tlb_remote_flush)(struct kvm *kvm);
1160 int (*tlb_remote_flush_with_range)(struct kvm *kvm,
1161 struct kvm_tlb_range *range);
1164 * Flush any TLB entries associated with the given GVA.
1165 * Does not need to flush GPA->HPA mappings.
1166 * Can potentially get non-canonical addresses through INVLPGs, which
1167 * the implementation may choose to ignore if appropriate.
1169 void (*tlb_flush_gva)(struct kvm_vcpu *vcpu, gva_t addr);
1172 * Flush any TLB entries created by the guest. Like tlb_flush_gva(),
1173 * does not need to flush GPA->HPA mappings.
1175 void (*tlb_flush_guest)(struct kvm_vcpu *vcpu);
1177 enum exit_fastpath_completion (*run)(struct kvm_vcpu *vcpu);
1178 int (*handle_exit)(struct kvm_vcpu *vcpu,
1179 enum exit_fastpath_completion exit_fastpath);
1180 int (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
1181 void (*update_emulated_instruction)(struct kvm_vcpu *vcpu);
1182 void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
1183 u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu);
1184 void (*patch_hypercall)(struct kvm_vcpu *vcpu,
1185 unsigned char *hypercall_addr);
1186 void (*set_irq)(struct kvm_vcpu *vcpu);
1187 void (*set_nmi)(struct kvm_vcpu *vcpu);
1188 void (*queue_exception)(struct kvm_vcpu *vcpu);
1189 void (*cancel_injection)(struct kvm_vcpu *vcpu);
1190 int (*interrupt_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1191 int (*nmi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1192 bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
1193 void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
1194 void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
1195 void (*enable_irq_window)(struct kvm_vcpu *vcpu);
1196 void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
1197 bool (*check_apicv_inhibit_reasons)(ulong bit);
1198 void (*pre_update_apicv_exec_ctrl)(struct kvm *kvm, bool activate);
1199 void (*refresh_apicv_exec_ctrl)(struct kvm_vcpu *vcpu);
1200 void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
1201 void (*hwapic_isr_update)(struct kvm_vcpu *vcpu, int isr);
1202 bool (*guest_apic_has_interrupt)(struct kvm_vcpu *vcpu);
1203 void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
1204 void (*set_virtual_apic_mode)(struct kvm_vcpu *vcpu);
1205 void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu);
1206 int (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
1207 int (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
1208 int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
1209 int (*set_identity_map_addr)(struct kvm *kvm, u64 ident_addr);
1210 u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
1212 void (*load_mmu_pgd)(struct kvm_vcpu *vcpu, unsigned long pgd,
1215 bool (*has_wbinvd_exit)(void);
1217 /* Returns actual tsc_offset set in active VMCS */
1218 u64 (*write_l1_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
1221 * Retrieve somewhat arbitrary exit information. Intended to be used
1222 * only from within tracepoints to avoid VMREADs when tracing is off.
1224 void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2,
1225 u32 *exit_int_info, u32 *exit_int_info_err_code);
1227 int (*check_intercept)(struct kvm_vcpu *vcpu,
1228 struct x86_instruction_info *info,
1229 enum x86_intercept_stage stage,
1230 struct x86_exception *exception);
1231 void (*handle_exit_irqoff)(struct kvm_vcpu *vcpu);
1233 void (*request_immediate_exit)(struct kvm_vcpu *vcpu);
1235 void (*sched_in)(struct kvm_vcpu *kvm, int cpu);
1238 * Arch-specific dirty logging hooks. These hooks are only supposed to
1239 * be valid if the specific arch has hardware-accelerated dirty logging
1240 * mechanism. Currently only for PML on VMX.
1242 * - slot_enable_log_dirty:
1243 * called when enabling log dirty mode for the slot.
1244 * - slot_disable_log_dirty:
1245 * called when disabling log dirty mode for the slot.
1246 * also called when slot is created with log dirty disabled.
1247 * - flush_log_dirty:
1248 * called before reporting dirty_bitmap to userspace.
1249 * - enable_log_dirty_pt_masked:
1250 * called when reenabling log dirty for the GFNs in the mask after
1251 * corresponding bits are cleared in slot->dirty_bitmap.
1253 void (*slot_enable_log_dirty)(struct kvm *kvm,
1254 struct kvm_memory_slot *slot);
1255 void (*slot_disable_log_dirty)(struct kvm *kvm,
1256 struct kvm_memory_slot *slot);
1257 void (*flush_log_dirty)(struct kvm *kvm);
1258 void (*enable_log_dirty_pt_masked)(struct kvm *kvm,
1259 struct kvm_memory_slot *slot,
1260 gfn_t offset, unsigned long mask);
1261 int (*cpu_dirty_log_size)(void);
1263 /* pmu operations of sub-arch */
1264 const struct kvm_pmu_ops *pmu_ops;
1265 const struct kvm_x86_nested_ops *nested_ops;
1268 * Architecture specific hooks for vCPU blocking due to
1270 * Returns for .pre_block():
1271 * - 0 means continue to block the vCPU.
1272 * - 1 means we cannot block the vCPU since some event
1273 * happens during this period, such as, 'ON' bit in
1274 * posted-interrupts descriptor is set.
1276 int (*pre_block)(struct kvm_vcpu *vcpu);
1277 void (*post_block)(struct kvm_vcpu *vcpu);
1279 void (*vcpu_blocking)(struct kvm_vcpu *vcpu);
1280 void (*vcpu_unblocking)(struct kvm_vcpu *vcpu);
1282 int (*update_pi_irte)(struct kvm *kvm, unsigned int host_irq,
1283 uint32_t guest_irq, bool set);
1284 void (*apicv_post_state_restore)(struct kvm_vcpu *vcpu);
1285 bool (*dy_apicv_has_pending_interrupt)(struct kvm_vcpu *vcpu);
1287 int (*set_hv_timer)(struct kvm_vcpu *vcpu, u64 guest_deadline_tsc,
1289 void (*cancel_hv_timer)(struct kvm_vcpu *vcpu);
1291 void (*setup_mce)(struct kvm_vcpu *vcpu);
1293 int (*smi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1294 int (*pre_enter_smm)(struct kvm_vcpu *vcpu, char *smstate);
1295 int (*pre_leave_smm)(struct kvm_vcpu *vcpu, const char *smstate);
1296 void (*enable_smi_window)(struct kvm_vcpu *vcpu);
1298 int (*mem_enc_op)(struct kvm *kvm, void __user *argp);
1299 int (*mem_enc_reg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1300 int (*mem_enc_unreg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1302 int (*get_msr_feature)(struct kvm_msr_entry *entry);
1304 bool (*can_emulate_instruction)(struct kvm_vcpu *vcpu, void *insn, int insn_len);
1306 bool (*apic_init_signal_blocked)(struct kvm_vcpu *vcpu);
1307 int (*enable_direct_tlbflush)(struct kvm_vcpu *vcpu);
1309 void (*migrate_timers)(struct kvm_vcpu *vcpu);
1310 void (*msr_filter_changed)(struct kvm_vcpu *vcpu);
1311 int (*complete_emulated_msr)(struct kvm_vcpu *vcpu, int err);
1313 void (*vcpu_deliver_sipi_vector)(struct kvm_vcpu *vcpu, u8 vector);
1316 struct kvm_x86_nested_ops {
1317 int (*check_events)(struct kvm_vcpu *vcpu);
1318 bool (*hv_timer_pending)(struct kvm_vcpu *vcpu);
1319 int (*get_state)(struct kvm_vcpu *vcpu,
1320 struct kvm_nested_state __user *user_kvm_nested_state,
1321 unsigned user_data_size);
1322 int (*set_state)(struct kvm_vcpu *vcpu,
1323 struct kvm_nested_state __user *user_kvm_nested_state,
1324 struct kvm_nested_state *kvm_state);
1325 bool (*get_nested_state_pages)(struct kvm_vcpu *vcpu);
1326 int (*write_log_dirty)(struct kvm_vcpu *vcpu, gpa_t l2_gpa);
1328 int (*enable_evmcs)(struct kvm_vcpu *vcpu,
1329 uint16_t *vmcs_version);
1330 uint16_t (*get_evmcs_version)(struct kvm_vcpu *vcpu);
1333 struct kvm_x86_init_ops {
1334 int (*cpu_has_kvm_support)(void);
1335 int (*disabled_by_bios)(void);
1336 int (*check_processor_compatibility)(void);
1337 int (*hardware_setup)(void);
1339 struct kvm_x86_ops *runtime_ops;
1342 struct kvm_arch_async_pf {
1349 extern u64 __read_mostly host_efer;
1350 extern bool __read_mostly allow_smaller_maxphyaddr;
1351 extern struct kvm_x86_ops kvm_x86_ops;
1353 #define __KVM_HAVE_ARCH_VM_ALLOC
1354 static inline struct kvm *kvm_arch_alloc_vm(void)
1356 return __vmalloc(kvm_x86_ops.vm_size, GFP_KERNEL_ACCOUNT | __GFP_ZERO);
1358 void kvm_arch_free_vm(struct kvm *kvm);
1360 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
1361 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
1363 if (kvm_x86_ops.tlb_remote_flush &&
1364 !kvm_x86_ops.tlb_remote_flush(kvm))
1370 int kvm_mmu_module_init(void);
1371 void kvm_mmu_module_exit(void);
1373 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
1374 int kvm_mmu_create(struct kvm_vcpu *vcpu);
1375 void kvm_mmu_init_vm(struct kvm *kvm);
1376 void kvm_mmu_uninit_vm(struct kvm *kvm);
1377 void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
1378 u64 dirty_mask, u64 nx_mask, u64 x_mask, u64 p_mask,
1379 u64 acc_track_mask, u64 me_mask);
1381 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
1382 void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
1383 struct kvm_memory_slot *memslot,
1385 void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
1386 const struct kvm_memory_slot *memslot);
1387 void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
1388 struct kvm_memory_slot *memslot);
1389 void kvm_mmu_slot_largepage_remove_write_access(struct kvm *kvm,
1390 struct kvm_memory_slot *memslot);
1391 void kvm_mmu_slot_set_dirty(struct kvm *kvm,
1392 struct kvm_memory_slot *memslot);
1393 void kvm_mmu_clear_dirty_pt_masked(struct kvm *kvm,
1394 struct kvm_memory_slot *slot,
1395 gfn_t gfn_offset, unsigned long mask);
1396 void kvm_mmu_zap_all(struct kvm *kvm);
1397 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, u64 gen);
1398 unsigned long kvm_mmu_calculate_default_mmu_pages(struct kvm *kvm);
1399 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned long kvm_nr_mmu_pages);
1401 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
1402 bool pdptrs_changed(struct kvm_vcpu *vcpu);
1404 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
1405 const void *val, int bytes);
1407 struct kvm_irq_mask_notifier {
1408 void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
1410 struct hlist_node link;
1413 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
1414 struct kvm_irq_mask_notifier *kimn);
1415 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
1416 struct kvm_irq_mask_notifier *kimn);
1417 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
1420 extern bool tdp_enabled;
1422 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
1424 /* control of guest tsc rate supported? */
1425 extern bool kvm_has_tsc_control;
1426 /* maximum supported tsc_khz for guests */
1427 extern u32 kvm_max_guest_tsc_khz;
1428 /* number of bits of the fractional part of the TSC scaling ratio */
1429 extern u8 kvm_tsc_scaling_ratio_frac_bits;
1430 /* maximum allowed value of TSC scaling ratio */
1431 extern u64 kvm_max_tsc_scaling_ratio;
1432 /* 1ull << kvm_tsc_scaling_ratio_frac_bits */
1433 extern u64 kvm_default_tsc_scaling_ratio;
1434 /* bus lock detection supported? */
1435 extern bool kvm_has_bus_lock_exit;
1437 extern u64 kvm_mce_cap_supported;
1440 * EMULTYPE_NO_DECODE - Set when re-emulating an instruction (after completing
1441 * userspace I/O) to indicate that the emulation context
1442 * should be resued as is, i.e. skip initialization of
1443 * emulation context, instruction fetch and decode.
1445 * EMULTYPE_TRAP_UD - Set when emulating an intercepted #UD from hardware.
1446 * Indicates that only select instructions (tagged with
1447 * EmulateOnUD) should be emulated (to minimize the emulator
1448 * attack surface). See also EMULTYPE_TRAP_UD_FORCED.
1450 * EMULTYPE_SKIP - Set when emulating solely to skip an instruction, i.e. to
1451 * decode the instruction length. For use *only* by
1452 * kvm_x86_ops.skip_emulated_instruction() implementations.
1454 * EMULTYPE_ALLOW_RETRY_PF - Set when the emulator should resume the guest to
1455 * retry native execution under certain conditions,
1456 * Can only be set in conjunction with EMULTYPE_PF.
1458 * EMULTYPE_TRAP_UD_FORCED - Set when emulating an intercepted #UD that was
1459 * triggered by KVM's magic "force emulation" prefix,
1460 * which is opt in via module param (off by default).
1461 * Bypasses EmulateOnUD restriction despite emulating
1462 * due to an intercepted #UD (see EMULTYPE_TRAP_UD).
1463 * Used to test the full emulator from userspace.
1465 * EMULTYPE_VMWARE_GP - Set when emulating an intercepted #GP for VMware
1466 * backdoor emulation, which is opt in via module param.
1467 * VMware backoor emulation handles select instructions
1468 * and reinjects the #GP for all other cases.
1470 * EMULTYPE_PF - Set when emulating MMIO by way of an intercepted #PF, in which
1471 * case the CR2/GPA value pass on the stack is valid.
1473 #define EMULTYPE_NO_DECODE (1 << 0)
1474 #define EMULTYPE_TRAP_UD (1 << 1)
1475 #define EMULTYPE_SKIP (1 << 2)
1476 #define EMULTYPE_ALLOW_RETRY_PF (1 << 3)
1477 #define EMULTYPE_TRAP_UD_FORCED (1 << 4)
1478 #define EMULTYPE_VMWARE_GP (1 << 5)
1479 #define EMULTYPE_PF (1 << 6)
1481 int kvm_emulate_instruction(struct kvm_vcpu *vcpu, int emulation_type);
1482 int kvm_emulate_instruction_from_buffer(struct kvm_vcpu *vcpu,
1483 void *insn, int insn_len);
1485 void kvm_enable_efer_bits(u64);
1486 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
1487 int __kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data, bool host_initiated);
1488 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data);
1489 int kvm_set_msr(struct kvm_vcpu *vcpu, u32 index, u64 data);
1490 int kvm_emulate_rdmsr(struct kvm_vcpu *vcpu);
1491 int kvm_emulate_wrmsr(struct kvm_vcpu *vcpu);
1493 int kvm_fast_pio(struct kvm_vcpu *vcpu, int size, unsigned short port, int in);
1494 int kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
1495 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
1496 int kvm_vcpu_halt(struct kvm_vcpu *vcpu);
1497 int kvm_emulate_ap_reset_hold(struct kvm_vcpu *vcpu);
1498 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
1500 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
1501 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
1502 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector);
1504 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
1505 int reason, bool has_error_code, u32 error_code);
1507 void kvm_free_guest_fpu(struct kvm_vcpu *vcpu);
1509 void kvm_post_set_cr0(struct kvm_vcpu *vcpu, unsigned long old_cr0, unsigned long cr0);
1510 void kvm_post_set_cr4(struct kvm_vcpu *vcpu, unsigned long old_cr4, unsigned long cr4);
1511 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
1512 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
1513 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
1514 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
1515 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
1516 int kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
1517 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
1518 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
1519 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
1520 int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr);
1522 int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1523 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1525 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
1526 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
1527 bool kvm_rdpmc(struct kvm_vcpu *vcpu);
1529 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1530 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1531 void kvm_queue_exception_p(struct kvm_vcpu *vcpu, unsigned nr, unsigned long payload);
1532 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1533 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1534 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
1535 bool kvm_inject_emulated_page_fault(struct kvm_vcpu *vcpu,
1536 struct x86_exception *fault);
1537 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1538 gfn_t gfn, void *data, int offset, int len,
1540 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
1541 bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr);
1543 static inline int __kvm_irq_line_state(unsigned long *irq_state,
1544 int irq_source_id, int level)
1546 /* Logical OR for level trig interrupt */
1548 __set_bit(irq_source_id, irq_state);
1550 __clear_bit(irq_source_id, irq_state);
1552 return !!(*irq_state);
1555 #define KVM_MMU_ROOT_CURRENT BIT(0)
1556 #define KVM_MMU_ROOT_PREVIOUS(i) BIT(1+i)
1557 #define KVM_MMU_ROOTS_ALL (~0UL)
1559 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
1560 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
1562 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
1564 void kvm_update_dr7(struct kvm_vcpu *vcpu);
1566 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
1567 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
1568 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
1569 int kvm_mmu_load(struct kvm_vcpu *vcpu);
1570 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
1571 void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu);
1572 void kvm_mmu_free_roots(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1573 ulong roots_to_free);
1574 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1575 struct x86_exception *exception);
1576 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
1577 struct x86_exception *exception);
1578 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
1579 struct x86_exception *exception);
1580 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
1581 struct x86_exception *exception);
1582 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
1583 struct x86_exception *exception);
1585 bool kvm_apicv_activated(struct kvm *kvm);
1586 void kvm_apicv_init(struct kvm *kvm, bool enable);
1587 void kvm_vcpu_update_apicv(struct kvm_vcpu *vcpu);
1588 void kvm_request_apicv_update(struct kvm *kvm, bool activate,
1591 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
1593 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u64 error_code,
1594 void *insn, int insn_len);
1595 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
1596 void kvm_mmu_invalidate_gva(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1597 gva_t gva, hpa_t root_hpa);
1598 void kvm_mmu_invpcid_gva(struct kvm_vcpu *vcpu, gva_t gva, unsigned long pcid);
1599 void kvm_mmu_new_pgd(struct kvm_vcpu *vcpu, gpa_t new_pgd, bool skip_tlb_flush,
1600 bool skip_mmu_sync);
1602 void kvm_configure_mmu(bool enable_tdp, int tdp_max_root_level,
1603 int tdp_huge_page_level);
1605 static inline u16 kvm_read_ldt(void)
1608 asm("sldt %0" : "=g"(ldt));
1612 static inline void kvm_load_ldt(u16 sel)
1614 asm("lldt %0" : : "rm"(sel));
1617 #ifdef CONFIG_X86_64
1618 static inline unsigned long read_msr(unsigned long msr)
1627 static inline u32 get_rdx_init_val(void)
1629 return 0x600; /* P6 family */
1632 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
1634 kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
1637 #define TSS_IOPB_BASE_OFFSET 0x66
1638 #define TSS_BASE_SIZE 0x68
1639 #define TSS_IOPB_SIZE (65536 / 8)
1640 #define TSS_REDIRECTION_SIZE (256 / 8)
1641 #define RMODE_TSS_SIZE \
1642 (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
1645 TASK_SWITCH_CALL = 0,
1646 TASK_SWITCH_IRET = 1,
1647 TASK_SWITCH_JMP = 2,
1648 TASK_SWITCH_GATE = 3,
1651 #define HF_GIF_MASK (1 << 0)
1652 #define HF_NMI_MASK (1 << 3)
1653 #define HF_IRET_MASK (1 << 4)
1654 #define HF_GUEST_MASK (1 << 5) /* VCPU is in guest-mode */
1655 #define HF_SMM_MASK (1 << 6)
1656 #define HF_SMM_INSIDE_NMI_MASK (1 << 7)
1658 #define __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
1659 #define KVM_ADDRESS_SPACE_NUM 2
1661 #define kvm_arch_vcpu_memslots_id(vcpu) ((vcpu)->arch.hflags & HF_SMM_MASK ? 1 : 0)
1662 #define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, (role).smm)
1664 asmlinkage void kvm_spurious_fault(void);
1667 * Hardware virtualization extension instructions may fault if a
1668 * reboot turns off virtualization while processes are running.
1669 * Usually after catching the fault we just panic; during reboot
1670 * instead the instruction is ignored.
1672 #define __kvm_handle_fault_on_reboot(insn) \
1678 ".pushsection .discard.instr_begin \n\t" \
1679 ".long 1b - . \n\t" \
1680 ".popsection \n\t" \
1681 "call kvm_spurious_fault \n\t" \
1683 ".pushsection .discard.instr_end \n\t" \
1684 ".long 1b - . \n\t" \
1685 ".popsection \n\t" \
1687 _ASM_EXTABLE(666b, 667b)
1689 #define KVM_ARCH_WANT_MMU_NOTIFIER
1690 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end,
1692 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
1693 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
1694 int kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
1695 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1696 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1697 int kvm_cpu_has_extint(struct kvm_vcpu *v);
1698 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1699 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1700 void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event);
1701 void kvm_vcpu_reload_apic_access_page(struct kvm_vcpu *vcpu);
1703 int kvm_pv_send_ipi(struct kvm *kvm, unsigned long ipi_bitmap_low,
1704 unsigned long ipi_bitmap_high, u32 min,
1705 unsigned long icr, int op_64_bit);
1707 void kvm_define_user_return_msr(unsigned index, u32 msr);
1708 int kvm_set_user_return_msr(unsigned index, u64 val, u64 mask);
1710 u64 kvm_scale_tsc(struct kvm_vcpu *vcpu, u64 tsc);
1711 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc);
1713 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu);
1714 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1716 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
1717 void kvm_make_scan_ioapic_request(struct kvm *kvm);
1718 void kvm_make_scan_ioapic_request_mask(struct kvm *kvm,
1719 unsigned long *vcpu_bitmap);
1721 bool kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1722 struct kvm_async_pf *work);
1723 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1724 struct kvm_async_pf *work);
1725 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1726 struct kvm_async_pf *work);
1727 void kvm_arch_async_page_present_queued(struct kvm_vcpu *vcpu);
1728 bool kvm_arch_can_dequeue_async_page_present(struct kvm_vcpu *vcpu);
1729 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1731 int kvm_skip_emulated_instruction(struct kvm_vcpu *vcpu);
1732 int kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1733 void __kvm_request_immediate_exit(struct kvm_vcpu *vcpu);
1735 int kvm_is_in_guest(void);
1737 void __user *__x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa,
1739 bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu);
1740 bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu);
1742 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
1743 struct kvm_vcpu **dest_vcpu);
1745 void kvm_set_msi_irq(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e,
1746 struct kvm_lapic_irq *irq);
1748 static inline bool kvm_irq_is_postable(struct kvm_lapic_irq *irq)
1750 /* We can only post Fixed and LowPrio IRQs */
1751 return (irq->delivery_mode == APIC_DM_FIXED ||
1752 irq->delivery_mode == APIC_DM_LOWEST);
1755 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu)
1757 if (kvm_x86_ops.vcpu_blocking)
1758 kvm_x86_ops.vcpu_blocking(vcpu);
1761 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu)
1763 if (kvm_x86_ops.vcpu_unblocking)
1764 kvm_x86_ops.vcpu_unblocking(vcpu);
1767 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
1769 static inline int kvm_cpu_get_apicid(int mps_cpu)
1771 #ifdef CONFIG_X86_LOCAL_APIC
1772 return default_cpu_present_to_apicid(mps_cpu);
1779 #define put_smstate(type, buf, offset, val) \
1780 *(type *)((buf) + (offset) - 0x7e00) = val
1782 #define GET_SMSTATE(type, buf, offset) \
1783 (*(type *)((buf) + (offset) - 0x7e00))
1785 int kvm_cpu_dirty_log_size(void);
1787 #endif /* _ASM_X86_KVM_HOST_H */