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 /* x86-specific vcpu->requests bit members */
56 #define KVM_REQ_MIGRATE_TIMER KVM_ARCH_REQ(0)
57 #define KVM_REQ_REPORT_TPR_ACCESS KVM_ARCH_REQ(1)
58 #define KVM_REQ_TRIPLE_FAULT KVM_ARCH_REQ(2)
59 #define KVM_REQ_MMU_SYNC KVM_ARCH_REQ(3)
60 #define KVM_REQ_CLOCK_UPDATE KVM_ARCH_REQ(4)
61 #define KVM_REQ_LOAD_MMU_PGD KVM_ARCH_REQ(5)
62 #define KVM_REQ_EVENT KVM_ARCH_REQ(6)
63 #define KVM_REQ_APF_HALT KVM_ARCH_REQ(7)
64 #define KVM_REQ_STEAL_UPDATE KVM_ARCH_REQ(8)
65 #define KVM_REQ_NMI KVM_ARCH_REQ(9)
66 #define KVM_REQ_PMU KVM_ARCH_REQ(10)
67 #define KVM_REQ_PMI KVM_ARCH_REQ(11)
68 #define KVM_REQ_SMI KVM_ARCH_REQ(12)
69 #define KVM_REQ_MASTERCLOCK_UPDATE KVM_ARCH_REQ(13)
70 #define KVM_REQ_MCLOCK_INPROGRESS \
71 KVM_ARCH_REQ_FLAGS(14, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
72 #define KVM_REQ_SCAN_IOAPIC \
73 KVM_ARCH_REQ_FLAGS(15, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
74 #define KVM_REQ_GLOBAL_CLOCK_UPDATE KVM_ARCH_REQ(16)
75 #define KVM_REQ_APIC_PAGE_RELOAD \
76 KVM_ARCH_REQ_FLAGS(17, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
77 #define KVM_REQ_HV_CRASH KVM_ARCH_REQ(18)
78 #define KVM_REQ_IOAPIC_EOI_EXIT KVM_ARCH_REQ(19)
79 #define KVM_REQ_HV_RESET KVM_ARCH_REQ(20)
80 #define KVM_REQ_HV_EXIT KVM_ARCH_REQ(21)
81 #define KVM_REQ_HV_STIMER KVM_ARCH_REQ(22)
82 #define KVM_REQ_LOAD_EOI_EXITMAP KVM_ARCH_REQ(23)
83 #define KVM_REQ_GET_NESTED_STATE_PAGES KVM_ARCH_REQ(24)
84 #define KVM_REQ_APICV_UPDATE \
85 KVM_ARCH_REQ_FLAGS(25, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
86 #define KVM_REQ_TLB_FLUSH_CURRENT KVM_ARCH_REQ(26)
87 #define KVM_REQ_HV_TLB_FLUSH \
88 KVM_ARCH_REQ_FLAGS(27, KVM_REQUEST_NO_WAKEUP)
89 #define KVM_REQ_APF_READY KVM_ARCH_REQ(28)
90 #define KVM_REQ_MSR_FILTER_CHANGED KVM_ARCH_REQ(29)
92 #define CR0_RESERVED_BITS \
93 (~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \
94 | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \
95 | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG))
97 #define CR4_RESERVED_BITS \
98 (~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\
99 | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE \
100 | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR | X86_CR4_PCIDE \
101 | X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_FSGSBASE \
102 | X86_CR4_OSXMMEXCPT | X86_CR4_LA57 | X86_CR4_VMXE \
103 | X86_CR4_SMAP | X86_CR4_PKE | X86_CR4_UMIP))
105 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR)
109 #define INVALID_PAGE (~(hpa_t)0)
110 #define VALID_PAGE(x) ((x) != INVALID_PAGE)
112 #define UNMAPPED_GVA (~(gpa_t)0)
114 /* KVM Hugepage definitions for x86 */
115 #define KVM_MAX_HUGEPAGE_LEVEL PG_LEVEL_1G
116 #define KVM_NR_PAGE_SIZES (KVM_MAX_HUGEPAGE_LEVEL - PG_LEVEL_4K + 1)
117 #define KVM_HPAGE_GFN_SHIFT(x) (((x) - 1) * 9)
118 #define KVM_HPAGE_SHIFT(x) (PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
119 #define KVM_HPAGE_SIZE(x) (1UL << KVM_HPAGE_SHIFT(x))
120 #define KVM_HPAGE_MASK(x) (~(KVM_HPAGE_SIZE(x) - 1))
121 #define KVM_PAGES_PER_HPAGE(x) (KVM_HPAGE_SIZE(x) / PAGE_SIZE)
123 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
125 /* KVM_HPAGE_GFN_SHIFT(PG_LEVEL_4K) must be 0. */
126 return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
127 (base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
130 #define KVM_PERMILLE_MMU_PAGES 20
131 #define KVM_MIN_ALLOC_MMU_PAGES 64UL
132 #define KVM_MMU_HASH_SHIFT 12
133 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
134 #define KVM_MIN_FREE_MMU_PAGES 5
135 #define KVM_REFILL_PAGES 25
136 #define KVM_MAX_CPUID_ENTRIES 80
137 #define KVM_NR_FIXED_MTRR_REGION 88
138 #define KVM_NR_VAR_MTRR 8
140 #define ASYNC_PF_PER_VCPU 64
143 VCPU_REGS_RAX = __VCPU_REGS_RAX,
144 VCPU_REGS_RCX = __VCPU_REGS_RCX,
145 VCPU_REGS_RDX = __VCPU_REGS_RDX,
146 VCPU_REGS_RBX = __VCPU_REGS_RBX,
147 VCPU_REGS_RSP = __VCPU_REGS_RSP,
148 VCPU_REGS_RBP = __VCPU_REGS_RBP,
149 VCPU_REGS_RSI = __VCPU_REGS_RSI,
150 VCPU_REGS_RDI = __VCPU_REGS_RDI,
152 VCPU_REGS_R8 = __VCPU_REGS_R8,
153 VCPU_REGS_R9 = __VCPU_REGS_R9,
154 VCPU_REGS_R10 = __VCPU_REGS_R10,
155 VCPU_REGS_R11 = __VCPU_REGS_R11,
156 VCPU_REGS_R12 = __VCPU_REGS_R12,
157 VCPU_REGS_R13 = __VCPU_REGS_R13,
158 VCPU_REGS_R14 = __VCPU_REGS_R14,
159 VCPU_REGS_R15 = __VCPU_REGS_R15,
164 VCPU_EXREG_PDPTR = NR_VCPU_REGS,
170 VCPU_EXREG_EXIT_INFO_1,
171 VCPU_EXREG_EXIT_INFO_2,
185 enum exit_fastpath_completion {
187 EXIT_FASTPATH_REENTER_GUEST,
188 EXIT_FASTPATH_EXIT_HANDLED,
190 typedef enum exit_fastpath_completion fastpath_t;
192 struct x86_emulate_ctxt;
193 struct x86_exception;
195 enum x86_intercept_stage;
197 #define KVM_NR_DB_REGS 4
199 #define DR6_BD (1 << 13)
200 #define DR6_BS (1 << 14)
201 #define DR6_BT (1 << 15)
202 #define DR6_RTM (1 << 16)
203 #define DR6_FIXED_1 0xfffe0ff0
204 #define DR6_INIT 0xffff0ff0
205 #define DR6_VOLATILE 0x0001e00f
207 #define DR7_BP_EN_MASK 0x000000ff
208 #define DR7_GE (1 << 9)
209 #define DR7_GD (1 << 13)
210 #define DR7_FIXED_1 0x00000400
211 #define DR7_VOLATILE 0xffff2bff
213 #define PFERR_PRESENT_BIT 0
214 #define PFERR_WRITE_BIT 1
215 #define PFERR_USER_BIT 2
216 #define PFERR_RSVD_BIT 3
217 #define PFERR_FETCH_BIT 4
218 #define PFERR_PK_BIT 5
219 #define PFERR_GUEST_FINAL_BIT 32
220 #define PFERR_GUEST_PAGE_BIT 33
222 #define PFERR_PRESENT_MASK (1U << PFERR_PRESENT_BIT)
223 #define PFERR_WRITE_MASK (1U << PFERR_WRITE_BIT)
224 #define PFERR_USER_MASK (1U << PFERR_USER_BIT)
225 #define PFERR_RSVD_MASK (1U << PFERR_RSVD_BIT)
226 #define PFERR_FETCH_MASK (1U << PFERR_FETCH_BIT)
227 #define PFERR_PK_MASK (1U << PFERR_PK_BIT)
228 #define PFERR_GUEST_FINAL_MASK (1ULL << PFERR_GUEST_FINAL_BIT)
229 #define PFERR_GUEST_PAGE_MASK (1ULL << PFERR_GUEST_PAGE_BIT)
231 #define PFERR_NESTED_GUEST_PAGE (PFERR_GUEST_PAGE_MASK | \
235 /* apic attention bits */
236 #define KVM_APIC_CHECK_VAPIC 0
238 * The following bit is set with PV-EOI, unset on EOI.
239 * We detect PV-EOI changes by guest by comparing
240 * this bit with PV-EOI in guest memory.
241 * See the implementation in apic_update_pv_eoi.
243 #define KVM_APIC_PV_EOI_PENDING 1
245 struct kvm_kernel_irq_routing_entry;
248 * the pages used as guest page table on soft mmu are tracked by
249 * kvm_memory_slot.arch.gfn_track which is 16 bits, so the role bits used
250 * by indirect shadow page can not be more than 15 bits.
252 * Currently, we used 14 bits that are @level, @gpte_is_8_bytes, @quadrant, @access,
253 * @nxe, @cr0_wp, @smep_andnot_wp and @smap_andnot_wp.
255 union kvm_mmu_page_role {
259 unsigned gpte_is_8_bytes:1;
266 unsigned smep_andnot_wp:1;
267 unsigned smap_andnot_wp:1;
268 unsigned ad_disabled:1;
269 unsigned guest_mode:1;
273 * This is left at the top of the word so that
274 * kvm_memslots_for_spte_role can extract it with a
275 * simple shift. While there is room, give it a whole
276 * byte so it is also faster to load it from memory.
282 union kvm_mmu_extended_role {
284 * This structure complements kvm_mmu_page_role caching everything needed for
285 * MMU configuration. If nothing in both these structures changed, MMU
286 * re-configuration can be skipped. @valid bit is set on first usage so we don't
287 * treat all-zero structure as valid data.
291 unsigned int valid:1;
292 unsigned int execonly:1;
293 unsigned int cr0_pg:1;
294 unsigned int cr4_pae:1;
295 unsigned int cr4_pse:1;
296 unsigned int cr4_pke:1;
297 unsigned int cr4_smap:1;
298 unsigned int cr4_smep:1;
299 unsigned int maxphyaddr:6;
306 union kvm_mmu_page_role base;
307 union kvm_mmu_extended_role ext;
311 struct kvm_rmap_head {
315 struct kvm_pio_request {
316 unsigned long linear_rip;
323 #define PT64_ROOT_MAX_LEVEL 5
325 struct rsvd_bits_validate {
326 u64 rsvd_bits_mask[2][PT64_ROOT_MAX_LEVEL];
330 struct kvm_mmu_root_info {
335 #define KVM_MMU_ROOT_INFO_INVALID \
336 ((struct kvm_mmu_root_info) { .pgd = INVALID_PAGE, .hpa = INVALID_PAGE })
338 #define KVM_MMU_NUM_PREV_ROOTS 3
343 * x86 supports 4 paging modes (5-level 64-bit, 4-level 64-bit, 3-level 32-bit,
344 * and 2-level 32-bit). The kvm_mmu structure abstracts the details of the
348 unsigned long (*get_guest_pgd)(struct kvm_vcpu *vcpu);
349 u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
350 int (*page_fault)(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u32 err,
352 void (*inject_page_fault)(struct kvm_vcpu *vcpu,
353 struct x86_exception *fault);
354 gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gpa_t gva_or_gpa,
355 u32 access, struct x86_exception *exception);
356 gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
357 struct x86_exception *exception);
358 int (*sync_page)(struct kvm_vcpu *vcpu,
359 struct kvm_mmu_page *sp);
360 void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva, hpa_t root_hpa);
361 void (*update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
362 u64 *spte, const void *pte);
365 union kvm_mmu_role mmu_role;
367 u8 shadow_root_level;
370 struct kvm_mmu_root_info prev_roots[KVM_MMU_NUM_PREV_ROOTS];
373 * Bitmap; bit set = permission fault
374 * Byte index: page fault error code [4:1]
375 * Bit index: pte permissions in ACC_* format
380 * The pkru_mask indicates if protection key checks are needed. It
381 * consists of 16 domains indexed by page fault error code bits [4:1],
382 * with PFEC.RSVD replaced by ACC_USER_MASK from the page tables.
383 * Each domain has 2 bits which are ANDed with AD and WD from PKRU.
391 * check zero bits on shadow page table entries, these
392 * bits include not only hardware reserved bits but also
393 * the bits spte never used.
395 struct rsvd_bits_validate shadow_zero_check;
397 struct rsvd_bits_validate guest_rsvd_check;
399 /* Can have large pages at levels 2..last_nonleaf_level-1. */
400 u8 last_nonleaf_level;
404 u64 pdptrs[4]; /* pae */
407 struct kvm_tlb_range {
422 struct perf_event *perf_event;
423 struct kvm_vcpu *vcpu;
425 * eventsel value for general purpose counters,
426 * ctrl value for fixed counters.
432 unsigned nr_arch_gp_counters;
433 unsigned nr_arch_fixed_counters;
434 unsigned available_event_types;
439 u64 counter_bitmask[2];
440 u64 global_ctrl_mask;
441 u64 global_ovf_ctrl_mask;
444 struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
445 struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED];
446 struct irq_work irq_work;
447 DECLARE_BITMAP(reprogram_pmi, X86_PMC_IDX_MAX);
448 DECLARE_BITMAP(all_valid_pmc_idx, X86_PMC_IDX_MAX);
449 DECLARE_BITMAP(pmc_in_use, X86_PMC_IDX_MAX);
452 * The gate to release perf_events not marked in
453 * pmc_in_use only once in a vcpu time slice.
458 * The total number of programmed perf_events and it helps to avoid
459 * redundant check before cleanup if guest don't use vPMU at all.
467 KVM_DEBUGREG_BP_ENABLED = 1,
468 KVM_DEBUGREG_WONT_EXIT = 2,
469 KVM_DEBUGREG_RELOAD = 4,
472 struct kvm_mtrr_range {
475 struct list_head node;
479 struct kvm_mtrr_range var_ranges[KVM_NR_VAR_MTRR];
480 mtrr_type fixed_ranges[KVM_NR_FIXED_MTRR_REGION];
483 struct list_head head;
486 /* Hyper-V SynIC timer */
487 struct kvm_vcpu_hv_stimer {
488 struct hrtimer timer;
490 union hv_stimer_config config;
493 struct hv_message msg;
497 /* Hyper-V synthetic interrupt controller (SynIC)*/
498 struct kvm_vcpu_hv_synic {
503 atomic64_t sint[HV_SYNIC_SINT_COUNT];
504 atomic_t sint_to_gsi[HV_SYNIC_SINT_COUNT];
505 DECLARE_BITMAP(auto_eoi_bitmap, 256);
506 DECLARE_BITMAP(vec_bitmap, 256);
508 bool dont_zero_synic_pages;
511 /* Hyper-V per vcpu emulation context */
516 struct kvm_vcpu_hv_synic synic;
517 struct kvm_hyperv_exit exit;
518 struct kvm_vcpu_hv_stimer stimer[HV_SYNIC_STIMER_COUNT];
519 DECLARE_BITMAP(stimer_pending_bitmap, HV_SYNIC_STIMER_COUNT);
523 struct kvm_vcpu_arch {
525 * rip and regs accesses must go through
526 * kvm_{register,rip}_{read,write} functions.
528 unsigned long regs[NR_VCPU_REGS];
533 unsigned long cr0_guest_owned_bits;
537 unsigned long cr4_guest_owned_bits;
538 unsigned long cr4_guest_rsvd_bits;
545 struct kvm_lapic *apic; /* kernel irqchip context */
547 bool load_eoi_exitmap_pending;
548 DECLARE_BITMAP(ioapic_handled_vectors, 256);
549 unsigned long apic_attention;
550 int32_t apic_arb_prio;
552 u64 ia32_misc_enable_msr;
555 bool tpr_access_reporting;
558 u64 microcode_version;
559 u64 arch_capabilities;
560 u64 perf_capabilities;
563 * Paging state of the vcpu
565 * If the vcpu runs in guest mode with two level paging this still saves
566 * the paging mode of the l1 guest. This context is always used to
571 /* Non-nested MMU for L1 */
572 struct kvm_mmu root_mmu;
574 /* L1 MMU when running nested */
575 struct kvm_mmu guest_mmu;
578 * Paging state of an L2 guest (used for nested npt)
580 * This context will save all necessary information to walk page tables
581 * of an L2 guest. This context is only initialized for page table
582 * walking and not for faulting since we never handle l2 page faults on
585 struct kvm_mmu nested_mmu;
588 * Pointer to the mmu context currently used for
589 * gva_to_gpa translations.
591 struct kvm_mmu *walk_mmu;
593 struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
594 struct kvm_mmu_memory_cache mmu_shadow_page_cache;
595 struct kvm_mmu_memory_cache mmu_gfn_array_cache;
596 struct kvm_mmu_memory_cache mmu_page_header_cache;
599 * QEMU userspace and the guest each have their own FPU state.
600 * In vcpu_run, we switch between the user and guest FPU contexts.
601 * While running a VCPU, the VCPU thread will have the guest FPU
604 * Note that while the PKRU state lives inside the fpu registers,
605 * it is switched out separately at VMENTER and VMEXIT time. The
606 * "guest_fpu" state here contains the guest FPU context, with the
609 struct fpu *user_fpu;
610 struct fpu *guest_fpu;
613 u64 guest_supported_xcr0;
615 struct kvm_pio_request pio;
618 u8 event_exit_inst_len;
620 struct kvm_queued_exception {
626 unsigned long payload;
631 struct kvm_queued_interrupt {
637 int halt_request; /* real mode on Intel only */
640 struct kvm_cpuid_entry2 cpuid_entries[KVM_MAX_CPUID_ENTRIES];
645 /* emulate context */
647 struct x86_emulate_ctxt *emulate_ctxt;
648 bool emulate_regs_need_sync_to_vcpu;
649 bool emulate_regs_need_sync_from_vcpu;
650 int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
653 struct pvclock_vcpu_time_info hv_clock;
654 unsigned int hw_tsc_khz;
655 struct gfn_to_hva_cache pv_time;
656 bool pv_time_enabled;
657 /* set guest stopped flag in pvclock flags field */
658 bool pvclock_set_guest_stopped_request;
664 struct gfn_to_pfn_cache cache;
671 u64 tsc_offset_adjustment;
674 u64 this_tsc_generation;
676 bool tsc_always_catchup;
677 s8 virtual_tsc_shift;
678 u32 virtual_tsc_mult;
680 s64 ia32_tsc_adjust_msr;
681 u64 msr_ia32_power_ctl;
682 u64 tsc_scaling_ratio;
684 atomic_t nmi_queued; /* unprocessed asynchronous NMIs */
685 unsigned nmi_pending; /* NMI queued after currently running handler */
686 bool nmi_injected; /* Trying to inject an NMI this entry */
687 bool smi_pending; /* SMI queued after currently running handler */
689 struct kvm_mtrr mtrr_state;
692 unsigned switch_db_regs;
693 unsigned long db[KVM_NR_DB_REGS];
696 unsigned long eff_db[KVM_NR_DB_REGS];
697 unsigned long guest_debug_dr7;
698 u64 msr_platform_info;
699 u64 msr_misc_features_enables;
707 /* Cache MMIO info */
709 unsigned mmio_access;
715 /* used for guest single stepping over the given code position */
716 unsigned long singlestep_rip;
718 struct kvm_vcpu_hv hyperv;
720 cpumask_var_t wbinvd_dirty_mask;
722 unsigned long last_retry_eip;
723 unsigned long last_retry_addr;
727 gfn_t gfns[ASYNC_PF_PER_VCPU];
728 struct gfn_to_hva_cache data;
729 u64 msr_en_val; /* MSR_KVM_ASYNC_PF_EN */
730 u64 msr_int_val; /* MSR_KVM_ASYNC_PF_INT */
735 unsigned long nested_apf_token;
736 bool delivery_as_pf_vmexit;
737 bool pageready_pending;
740 /* OSVW MSRs (AMD only) */
748 struct gfn_to_hva_cache data;
751 u64 msr_kvm_poll_control;
754 * Indicates the guest is trying to write a gfn that contains one or
755 * more of the PTEs used to translate the write itself, i.e. the access
756 * is changing its own translation in the guest page tables. KVM exits
757 * to userspace if emulation of the faulting instruction fails and this
758 * flag is set, as KVM cannot make forward progress.
760 * If emulation fails for a write to guest page tables, KVM unprotects
761 * (zaps) the shadow page for the target gfn and resumes the guest to
762 * retry the non-emulatable instruction (on hardware). Unprotecting the
763 * gfn doesn't allow forward progress for a self-changing access because
764 * doing so also zaps the translation for the gfn, i.e. retrying the
765 * instruction will hit a !PRESENT fault, which results in a new shadow
766 * page and sends KVM back to square one.
768 bool write_fault_to_shadow_pgtable;
770 /* set at EPT violation at this point */
771 unsigned long exit_qualification;
773 /* pv related host specific info */
778 int pending_ioapic_eoi;
779 int pending_external_vector;
781 /* be preempted when it's in kernel-mode(cpl=0) */
782 bool preempted_in_kernel;
784 /* Flush the L1 Data cache for L1TF mitigation on VMENTER */
787 /* Host CPU on which VM-entry was most recently attempted */
788 unsigned int last_vmentry_cpu;
790 /* AMD MSRC001_0015 Hardware Configuration */
794 struct kvm_lpage_info {
798 struct kvm_arch_memory_slot {
799 struct kvm_rmap_head *rmap[KVM_NR_PAGE_SIZES];
800 struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
801 unsigned short *gfn_track[KVM_PAGE_TRACK_MAX];
805 * We use as the mode the number of bits allocated in the LDR for the
806 * logical processor ID. It happens that these are all powers of two.
807 * This makes it is very easy to detect cases where the APICs are
808 * configured for multiple modes; in that case, we cannot use the map and
809 * hence cannot use kvm_irq_delivery_to_apic_fast either.
811 #define KVM_APIC_MODE_XAPIC_CLUSTER 4
812 #define KVM_APIC_MODE_XAPIC_FLAT 8
813 #define KVM_APIC_MODE_X2APIC 16
815 struct kvm_apic_map {
820 struct kvm_lapic *xapic_flat_map[8];
821 struct kvm_lapic *xapic_cluster_map[16][4];
823 struct kvm_lapic *phys_map[];
826 /* Hyper-V synthetic debugger (SynDbg)*/
827 struct kvm_hv_syndbg {
838 /* Hyper-V emulation context */
840 struct mutex hv_lock;
845 /* Hyper-v based guest crash (NT kernel bugcheck) parameters */
846 u64 hv_crash_param[HV_X64_MSR_CRASH_PARAMS];
849 struct ms_hyperv_tsc_page tsc_ref;
851 struct idr conn_to_evt;
853 u64 hv_reenlightenment_control;
854 u64 hv_tsc_emulation_control;
855 u64 hv_tsc_emulation_status;
857 /* How many vCPUs have VP index != vCPU index */
858 atomic_t num_mismatched_vp_indexes;
860 struct hv_partition_assist_pg *hv_pa_pg;
861 struct kvm_hv_syndbg hv_syndbg;
864 struct msr_bitmap_range {
868 unsigned long *bitmap;
871 enum kvm_irqchip_mode {
873 KVM_IRQCHIP_KERNEL, /* created with KVM_CREATE_IRQCHIP */
874 KVM_IRQCHIP_SPLIT, /* created with KVM_CAP_SPLIT_IRQCHIP */
877 #define APICV_INHIBIT_REASON_DISABLE 0
878 #define APICV_INHIBIT_REASON_HYPERV 1
879 #define APICV_INHIBIT_REASON_NESTED 2
880 #define APICV_INHIBIT_REASON_IRQWIN 3
881 #define APICV_INHIBIT_REASON_PIT_REINJ 4
882 #define APICV_INHIBIT_REASON_X2APIC 5
885 unsigned long n_used_mmu_pages;
886 unsigned long n_requested_mmu_pages;
887 unsigned long n_max_mmu_pages;
888 unsigned int indirect_shadow_pages;
890 struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
892 * Hash table of struct kvm_mmu_page.
894 struct list_head active_mmu_pages;
895 struct list_head zapped_obsolete_pages;
896 struct list_head lpage_disallowed_mmu_pages;
897 struct kvm_page_track_notifier_node mmu_sp_tracker;
898 struct kvm_page_track_notifier_head track_notifier_head;
900 struct list_head assigned_dev_head;
901 struct iommu_domain *iommu_domain;
902 bool iommu_noncoherent;
903 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA
904 atomic_t noncoherent_dma_count;
905 #define __KVM_HAVE_ARCH_ASSIGNED_DEVICE
906 atomic_t assigned_device_count;
907 struct kvm_pic *vpic;
908 struct kvm_ioapic *vioapic;
909 struct kvm_pit *vpit;
910 atomic_t vapics_in_nmi_mode;
911 struct mutex apic_map_lock;
912 struct kvm_apic_map *apic_map;
913 atomic_t apic_map_dirty;
915 bool apic_access_page_done;
916 unsigned long apicv_inhibit_reasons;
923 bool cstate_in_guest;
925 unsigned long irq_sources_bitmap;
927 raw_spinlock_t tsc_write_lock;
934 u64 cur_tsc_generation;
935 int nr_vcpus_matched_tsc;
937 spinlock_t pvclock_gtod_sync_lock;
938 bool use_master_clock;
939 u64 master_kernel_ns;
940 u64 master_cycle_now;
941 struct delayed_work kvmclock_update_work;
942 struct delayed_work kvmclock_sync_work;
944 struct kvm_xen_hvm_config xen_hvm_config;
946 /* reads protected by irq_srcu, writes by irq_lock */
947 struct hlist_head mask_notifier_list;
949 struct kvm_hv hyperv;
951 #ifdef CONFIG_KVM_MMU_AUDIT
955 bool backwards_tsc_observed;
956 bool boot_vcpu_runs_old_kvmclock;
961 enum kvm_irqchip_mode irqchip_mode;
962 u8 nr_reserved_ioapic_pins;
964 bool disabled_lapic_found;
967 bool x2apic_broadcast_quirk_disabled;
969 bool guest_can_read_msr_platform_info;
970 bool exception_payload_enabled;
972 /* Deflect RDMSR and WRMSR to user space when they trigger a #GP */
973 u32 user_space_msr_mask;
977 bool default_allow:1;
978 struct msr_bitmap_range ranges[16];
981 struct kvm_pmu_event_filter *pmu_event_filter;
982 struct task_struct *nx_lpage_recovery_thread;
986 ulong mmu_shadow_zapped;
988 ulong mmu_pte_updated;
989 ulong mmu_pde_zapped;
992 ulong mmu_cache_miss;
994 ulong remote_tlb_flush;
996 ulong nx_lpage_splits;
997 ulong max_mmu_page_hash_collisions;
1000 struct kvm_vcpu_stat {
1010 u64 irq_window_exits;
1011 u64 nmi_window_exits;
1014 u64 halt_successful_poll;
1015 u64 halt_attempted_poll;
1016 u64 halt_poll_invalid;
1018 u64 request_irq_exits;
1020 u64 host_state_reload;
1023 u64 insn_emulation_fail;
1028 u64 halt_poll_success_ns;
1029 u64 halt_poll_fail_ns;
1032 struct x86_instruction_info;
1035 bool host_initiated;
1040 struct kvm_lapic_irq {
1048 bool msi_redir_hint;
1051 static inline u16 kvm_lapic_irq_dest_mode(bool dest_mode_logical)
1053 return dest_mode_logical ? APIC_DEST_LOGICAL : APIC_DEST_PHYSICAL;
1056 struct kvm_x86_ops {
1057 int (*hardware_enable)(void);
1058 void (*hardware_disable)(void);
1059 void (*hardware_unsetup)(void);
1060 bool (*cpu_has_accelerated_tpr)(void);
1061 bool (*has_emulated_msr)(u32 index);
1062 void (*vcpu_after_set_cpuid)(struct kvm_vcpu *vcpu);
1064 unsigned int vm_size;
1065 int (*vm_init)(struct kvm *kvm);
1066 void (*vm_destroy)(struct kvm *kvm);
1068 /* Create, but do not attach this VCPU */
1069 int (*vcpu_create)(struct kvm_vcpu *vcpu);
1070 void (*vcpu_free)(struct kvm_vcpu *vcpu);
1071 void (*vcpu_reset)(struct kvm_vcpu *vcpu, bool init_event);
1073 void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
1074 void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
1075 void (*vcpu_put)(struct kvm_vcpu *vcpu);
1077 void (*update_exception_bitmap)(struct kvm_vcpu *vcpu);
1078 int (*get_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1079 int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1080 u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
1081 void (*get_segment)(struct kvm_vcpu *vcpu,
1082 struct kvm_segment *var, int seg);
1083 int (*get_cpl)(struct kvm_vcpu *vcpu);
1084 void (*set_segment)(struct kvm_vcpu *vcpu,
1085 struct kvm_segment *var, int seg);
1086 void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
1087 void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
1088 int (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
1089 void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
1090 void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1091 void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1092 void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1093 void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1094 void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu);
1095 void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
1096 void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
1097 unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
1098 void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
1100 void (*tlb_flush_all)(struct kvm_vcpu *vcpu);
1101 void (*tlb_flush_current)(struct kvm_vcpu *vcpu);
1102 int (*tlb_remote_flush)(struct kvm *kvm);
1103 int (*tlb_remote_flush_with_range)(struct kvm *kvm,
1104 struct kvm_tlb_range *range);
1107 * Flush any TLB entries associated with the given GVA.
1108 * Does not need to flush GPA->HPA mappings.
1109 * Can potentially get non-canonical addresses through INVLPGs, which
1110 * the implementation may choose to ignore if appropriate.
1112 void (*tlb_flush_gva)(struct kvm_vcpu *vcpu, gva_t addr);
1115 * Flush any TLB entries created by the guest. Like tlb_flush_gva(),
1116 * does not need to flush GPA->HPA mappings.
1118 void (*tlb_flush_guest)(struct kvm_vcpu *vcpu);
1120 enum exit_fastpath_completion (*run)(struct kvm_vcpu *vcpu);
1121 int (*handle_exit)(struct kvm_vcpu *vcpu,
1122 enum exit_fastpath_completion exit_fastpath);
1123 int (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
1124 void (*update_emulated_instruction)(struct kvm_vcpu *vcpu);
1125 void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
1126 u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu);
1127 void (*patch_hypercall)(struct kvm_vcpu *vcpu,
1128 unsigned char *hypercall_addr);
1129 void (*set_irq)(struct kvm_vcpu *vcpu);
1130 void (*set_nmi)(struct kvm_vcpu *vcpu);
1131 void (*queue_exception)(struct kvm_vcpu *vcpu);
1132 void (*cancel_injection)(struct kvm_vcpu *vcpu);
1133 int (*interrupt_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1134 int (*nmi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1135 bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
1136 void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
1137 void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
1138 void (*enable_irq_window)(struct kvm_vcpu *vcpu);
1139 void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
1140 bool (*check_apicv_inhibit_reasons)(ulong bit);
1141 void (*pre_update_apicv_exec_ctrl)(struct kvm *kvm, bool activate);
1142 void (*refresh_apicv_exec_ctrl)(struct kvm_vcpu *vcpu);
1143 void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
1144 void (*hwapic_isr_update)(struct kvm_vcpu *vcpu, int isr);
1145 bool (*guest_apic_has_interrupt)(struct kvm_vcpu *vcpu);
1146 void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
1147 void (*set_virtual_apic_mode)(struct kvm_vcpu *vcpu);
1148 void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu);
1149 int (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
1150 int (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
1151 int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
1152 int (*set_identity_map_addr)(struct kvm *kvm, u64 ident_addr);
1153 u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
1155 void (*load_mmu_pgd)(struct kvm_vcpu *vcpu, unsigned long pgd,
1158 bool (*has_wbinvd_exit)(void);
1160 /* Returns actual tsc_offset set in active VMCS */
1161 u64 (*write_l1_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
1164 * Retrieve somewhat arbitrary exit information. Intended to be used
1165 * only from within tracepoints to avoid VMREADs when tracing is off.
1167 void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2,
1168 u32 *exit_int_info, u32 *exit_int_info_err_code);
1170 int (*check_intercept)(struct kvm_vcpu *vcpu,
1171 struct x86_instruction_info *info,
1172 enum x86_intercept_stage stage,
1173 struct x86_exception *exception);
1174 void (*handle_exit_irqoff)(struct kvm_vcpu *vcpu);
1176 void (*request_immediate_exit)(struct kvm_vcpu *vcpu);
1178 void (*sched_in)(struct kvm_vcpu *kvm, int cpu);
1181 * Arch-specific dirty logging hooks. These hooks are only supposed to
1182 * be valid if the specific arch has hardware-accelerated dirty logging
1183 * mechanism. Currently only for PML on VMX.
1185 * - slot_enable_log_dirty:
1186 * called when enabling log dirty mode for the slot.
1187 * - slot_disable_log_dirty:
1188 * called when disabling log dirty mode for the slot.
1189 * also called when slot is created with log dirty disabled.
1190 * - flush_log_dirty:
1191 * called before reporting dirty_bitmap to userspace.
1192 * - enable_log_dirty_pt_masked:
1193 * called when reenabling log dirty for the GFNs in the mask after
1194 * corresponding bits are cleared in slot->dirty_bitmap.
1196 void (*slot_enable_log_dirty)(struct kvm *kvm,
1197 struct kvm_memory_slot *slot);
1198 void (*slot_disable_log_dirty)(struct kvm *kvm,
1199 struct kvm_memory_slot *slot);
1200 void (*flush_log_dirty)(struct kvm *kvm);
1201 void (*enable_log_dirty_pt_masked)(struct kvm *kvm,
1202 struct kvm_memory_slot *slot,
1203 gfn_t offset, unsigned long mask);
1205 /* pmu operations of sub-arch */
1206 const struct kvm_pmu_ops *pmu_ops;
1207 const struct kvm_x86_nested_ops *nested_ops;
1210 * Architecture specific hooks for vCPU blocking due to
1212 * Returns for .pre_block():
1213 * - 0 means continue to block the vCPU.
1214 * - 1 means we cannot block the vCPU since some event
1215 * happens during this period, such as, 'ON' bit in
1216 * posted-interrupts descriptor is set.
1218 int (*pre_block)(struct kvm_vcpu *vcpu);
1219 void (*post_block)(struct kvm_vcpu *vcpu);
1221 void (*vcpu_blocking)(struct kvm_vcpu *vcpu);
1222 void (*vcpu_unblocking)(struct kvm_vcpu *vcpu);
1224 int (*update_pi_irte)(struct kvm *kvm, unsigned int host_irq,
1225 uint32_t guest_irq, bool set);
1226 void (*apicv_post_state_restore)(struct kvm_vcpu *vcpu);
1227 bool (*dy_apicv_has_pending_interrupt)(struct kvm_vcpu *vcpu);
1229 int (*set_hv_timer)(struct kvm_vcpu *vcpu, u64 guest_deadline_tsc,
1231 void (*cancel_hv_timer)(struct kvm_vcpu *vcpu);
1233 void (*setup_mce)(struct kvm_vcpu *vcpu);
1235 int (*smi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1236 int (*pre_enter_smm)(struct kvm_vcpu *vcpu, char *smstate);
1237 int (*pre_leave_smm)(struct kvm_vcpu *vcpu, const char *smstate);
1238 void (*enable_smi_window)(struct kvm_vcpu *vcpu);
1240 int (*mem_enc_op)(struct kvm *kvm, void __user *argp);
1241 int (*mem_enc_reg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1242 int (*mem_enc_unreg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1244 int (*get_msr_feature)(struct kvm_msr_entry *entry);
1246 bool (*can_emulate_instruction)(struct kvm_vcpu *vcpu, void *insn, int insn_len);
1248 bool (*apic_init_signal_blocked)(struct kvm_vcpu *vcpu);
1249 int (*enable_direct_tlbflush)(struct kvm_vcpu *vcpu);
1251 void (*migrate_timers)(struct kvm_vcpu *vcpu);
1252 void (*msr_filter_changed)(struct kvm_vcpu *vcpu);
1255 struct kvm_x86_nested_ops {
1256 int (*check_events)(struct kvm_vcpu *vcpu);
1257 bool (*hv_timer_pending)(struct kvm_vcpu *vcpu);
1258 int (*get_state)(struct kvm_vcpu *vcpu,
1259 struct kvm_nested_state __user *user_kvm_nested_state,
1260 unsigned user_data_size);
1261 int (*set_state)(struct kvm_vcpu *vcpu,
1262 struct kvm_nested_state __user *user_kvm_nested_state,
1263 struct kvm_nested_state *kvm_state);
1264 bool (*get_nested_state_pages)(struct kvm_vcpu *vcpu);
1265 int (*write_log_dirty)(struct kvm_vcpu *vcpu, gpa_t l2_gpa);
1267 int (*enable_evmcs)(struct kvm_vcpu *vcpu,
1268 uint16_t *vmcs_version);
1269 uint16_t (*get_evmcs_version)(struct kvm_vcpu *vcpu);
1272 struct kvm_x86_init_ops {
1273 int (*cpu_has_kvm_support)(void);
1274 int (*disabled_by_bios)(void);
1275 int (*check_processor_compatibility)(void);
1276 int (*hardware_setup)(void);
1278 struct kvm_x86_ops *runtime_ops;
1281 struct kvm_arch_async_pf {
1288 extern u64 __read_mostly host_efer;
1289 extern bool __read_mostly allow_smaller_maxphyaddr;
1290 extern struct kvm_x86_ops kvm_x86_ops;
1292 #define __KVM_HAVE_ARCH_VM_ALLOC
1293 static inline struct kvm *kvm_arch_alloc_vm(void)
1295 return __vmalloc(kvm_x86_ops.vm_size, GFP_KERNEL_ACCOUNT | __GFP_ZERO);
1297 void kvm_arch_free_vm(struct kvm *kvm);
1299 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
1300 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
1302 if (kvm_x86_ops.tlb_remote_flush &&
1303 !kvm_x86_ops.tlb_remote_flush(kvm))
1309 int kvm_mmu_module_init(void);
1310 void kvm_mmu_module_exit(void);
1312 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
1313 int kvm_mmu_create(struct kvm_vcpu *vcpu);
1314 void kvm_mmu_init_vm(struct kvm *kvm);
1315 void kvm_mmu_uninit_vm(struct kvm *kvm);
1316 void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
1317 u64 dirty_mask, u64 nx_mask, u64 x_mask, u64 p_mask,
1318 u64 acc_track_mask, u64 me_mask);
1320 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
1321 void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
1322 struct kvm_memory_slot *memslot,
1324 void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
1325 const struct kvm_memory_slot *memslot);
1326 void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
1327 struct kvm_memory_slot *memslot);
1328 void kvm_mmu_slot_largepage_remove_write_access(struct kvm *kvm,
1329 struct kvm_memory_slot *memslot);
1330 void kvm_mmu_slot_set_dirty(struct kvm *kvm,
1331 struct kvm_memory_slot *memslot);
1332 void kvm_mmu_clear_dirty_pt_masked(struct kvm *kvm,
1333 struct kvm_memory_slot *slot,
1334 gfn_t gfn_offset, unsigned long mask);
1335 void kvm_mmu_zap_all(struct kvm *kvm);
1336 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, u64 gen);
1337 unsigned long kvm_mmu_calculate_default_mmu_pages(struct kvm *kvm);
1338 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned long kvm_nr_mmu_pages);
1340 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
1341 bool pdptrs_changed(struct kvm_vcpu *vcpu);
1343 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
1344 const void *val, int bytes);
1346 struct kvm_irq_mask_notifier {
1347 void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
1349 struct hlist_node link;
1352 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
1353 struct kvm_irq_mask_notifier *kimn);
1354 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
1355 struct kvm_irq_mask_notifier *kimn);
1356 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
1359 extern bool tdp_enabled;
1361 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
1363 /* control of guest tsc rate supported? */
1364 extern bool kvm_has_tsc_control;
1365 /* maximum supported tsc_khz for guests */
1366 extern u32 kvm_max_guest_tsc_khz;
1367 /* number of bits of the fractional part of the TSC scaling ratio */
1368 extern u8 kvm_tsc_scaling_ratio_frac_bits;
1369 /* maximum allowed value of TSC scaling ratio */
1370 extern u64 kvm_max_tsc_scaling_ratio;
1371 /* 1ull << kvm_tsc_scaling_ratio_frac_bits */
1372 extern u64 kvm_default_tsc_scaling_ratio;
1374 extern u64 kvm_mce_cap_supported;
1377 * EMULTYPE_NO_DECODE - Set when re-emulating an instruction (after completing
1378 * userspace I/O) to indicate that the emulation context
1379 * should be resued as is, i.e. skip initialization of
1380 * emulation context, instruction fetch and decode.
1382 * EMULTYPE_TRAP_UD - Set when emulating an intercepted #UD from hardware.
1383 * Indicates that only select instructions (tagged with
1384 * EmulateOnUD) should be emulated (to minimize the emulator
1385 * attack surface). See also EMULTYPE_TRAP_UD_FORCED.
1387 * EMULTYPE_SKIP - Set when emulating solely to skip an instruction, i.e. to
1388 * decode the instruction length. For use *only* by
1389 * kvm_x86_ops.skip_emulated_instruction() implementations.
1391 * EMULTYPE_ALLOW_RETRY_PF - Set when the emulator should resume the guest to
1392 * retry native execution under certain conditions,
1393 * Can only be set in conjunction with EMULTYPE_PF.
1395 * EMULTYPE_TRAP_UD_FORCED - Set when emulating an intercepted #UD that was
1396 * triggered by KVM's magic "force emulation" prefix,
1397 * which is opt in via module param (off by default).
1398 * Bypasses EmulateOnUD restriction despite emulating
1399 * due to an intercepted #UD (see EMULTYPE_TRAP_UD).
1400 * Used to test the full emulator from userspace.
1402 * EMULTYPE_VMWARE_GP - Set when emulating an intercepted #GP for VMware
1403 * backdoor emulation, which is opt in via module param.
1404 * VMware backoor emulation handles select instructions
1405 * and reinjects the #GP for all other cases.
1407 * EMULTYPE_PF - Set when emulating MMIO by way of an intercepted #PF, in which
1408 * case the CR2/GPA value pass on the stack is valid.
1410 #define EMULTYPE_NO_DECODE (1 << 0)
1411 #define EMULTYPE_TRAP_UD (1 << 1)
1412 #define EMULTYPE_SKIP (1 << 2)
1413 #define EMULTYPE_ALLOW_RETRY_PF (1 << 3)
1414 #define EMULTYPE_TRAP_UD_FORCED (1 << 4)
1415 #define EMULTYPE_VMWARE_GP (1 << 5)
1416 #define EMULTYPE_PF (1 << 6)
1418 int kvm_emulate_instruction(struct kvm_vcpu *vcpu, int emulation_type);
1419 int kvm_emulate_instruction_from_buffer(struct kvm_vcpu *vcpu,
1420 void *insn, int insn_len);
1422 void kvm_enable_efer_bits(u64);
1423 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
1424 int __kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data, bool host_initiated);
1425 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data);
1426 int kvm_set_msr(struct kvm_vcpu *vcpu, u32 index, u64 data);
1427 int kvm_emulate_rdmsr(struct kvm_vcpu *vcpu);
1428 int kvm_emulate_wrmsr(struct kvm_vcpu *vcpu);
1430 int kvm_fast_pio(struct kvm_vcpu *vcpu, int size, unsigned short port, int in);
1431 int kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
1432 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
1433 int kvm_vcpu_halt(struct kvm_vcpu *vcpu);
1434 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
1436 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
1437 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
1438 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector);
1440 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
1441 int reason, bool has_error_code, u32 error_code);
1443 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
1444 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
1445 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
1446 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
1447 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
1448 int kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
1449 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
1450 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
1451 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
1452 int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr);
1454 int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1455 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1457 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
1458 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
1459 bool kvm_rdpmc(struct kvm_vcpu *vcpu);
1461 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1462 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1463 void kvm_queue_exception_p(struct kvm_vcpu *vcpu, unsigned nr, unsigned long payload);
1464 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1465 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1466 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
1467 bool kvm_inject_emulated_page_fault(struct kvm_vcpu *vcpu,
1468 struct x86_exception *fault);
1469 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1470 gfn_t gfn, void *data, int offset, int len,
1472 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
1473 bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr);
1475 static inline int __kvm_irq_line_state(unsigned long *irq_state,
1476 int irq_source_id, int level)
1478 /* Logical OR for level trig interrupt */
1480 __set_bit(irq_source_id, irq_state);
1482 __clear_bit(irq_source_id, irq_state);
1484 return !!(*irq_state);
1487 #define KVM_MMU_ROOT_CURRENT BIT(0)
1488 #define KVM_MMU_ROOT_PREVIOUS(i) BIT(1+i)
1489 #define KVM_MMU_ROOTS_ALL (~0UL)
1491 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
1492 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
1494 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
1496 void kvm_update_dr7(struct kvm_vcpu *vcpu);
1498 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
1499 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
1500 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
1501 int kvm_mmu_load(struct kvm_vcpu *vcpu);
1502 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
1503 void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu);
1504 void kvm_mmu_free_roots(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1505 ulong roots_to_free);
1506 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1507 struct x86_exception *exception);
1508 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
1509 struct x86_exception *exception);
1510 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
1511 struct x86_exception *exception);
1512 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
1513 struct x86_exception *exception);
1514 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
1515 struct x86_exception *exception);
1517 bool kvm_apicv_activated(struct kvm *kvm);
1518 void kvm_apicv_init(struct kvm *kvm, bool enable);
1519 void kvm_vcpu_update_apicv(struct kvm_vcpu *vcpu);
1520 void kvm_request_apicv_update(struct kvm *kvm, bool activate,
1523 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
1525 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u64 error_code,
1526 void *insn, int insn_len);
1527 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
1528 void kvm_mmu_invalidate_gva(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1529 gva_t gva, hpa_t root_hpa);
1530 void kvm_mmu_invpcid_gva(struct kvm_vcpu *vcpu, gva_t gva, unsigned long pcid);
1531 void kvm_mmu_new_pgd(struct kvm_vcpu *vcpu, gpa_t new_pgd, bool skip_tlb_flush,
1532 bool skip_mmu_sync);
1534 void kvm_configure_mmu(bool enable_tdp, int tdp_max_root_level,
1535 int tdp_huge_page_level);
1537 static inline u16 kvm_read_ldt(void)
1540 asm("sldt %0" : "=g"(ldt));
1544 static inline void kvm_load_ldt(u16 sel)
1546 asm("lldt %0" : : "rm"(sel));
1549 #ifdef CONFIG_X86_64
1550 static inline unsigned long read_msr(unsigned long msr)
1559 static inline u32 get_rdx_init_val(void)
1561 return 0x600; /* P6 family */
1564 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
1566 kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
1569 #define TSS_IOPB_BASE_OFFSET 0x66
1570 #define TSS_BASE_SIZE 0x68
1571 #define TSS_IOPB_SIZE (65536 / 8)
1572 #define TSS_REDIRECTION_SIZE (256 / 8)
1573 #define RMODE_TSS_SIZE \
1574 (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
1577 TASK_SWITCH_CALL = 0,
1578 TASK_SWITCH_IRET = 1,
1579 TASK_SWITCH_JMP = 2,
1580 TASK_SWITCH_GATE = 3,
1583 #define HF_GIF_MASK (1 << 0)
1584 #define HF_NMI_MASK (1 << 3)
1585 #define HF_IRET_MASK (1 << 4)
1586 #define HF_GUEST_MASK (1 << 5) /* VCPU is in guest-mode */
1587 #define HF_SMM_MASK (1 << 6)
1588 #define HF_SMM_INSIDE_NMI_MASK (1 << 7)
1590 #define __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
1591 #define KVM_ADDRESS_SPACE_NUM 2
1593 #define kvm_arch_vcpu_memslots_id(vcpu) ((vcpu)->arch.hflags & HF_SMM_MASK ? 1 : 0)
1594 #define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, (role).smm)
1596 asmlinkage void kvm_spurious_fault(void);
1599 * Hardware virtualization extension instructions may fault if a
1600 * reboot turns off virtualization while processes are running.
1601 * Usually after catching the fault we just panic; during reboot
1602 * instead the instruction is ignored.
1604 #define __kvm_handle_fault_on_reboot(insn) \
1610 ".pushsection .discard.instr_begin \n\t" \
1611 ".long 1b - . \n\t" \
1612 ".popsection \n\t" \
1613 "call kvm_spurious_fault \n\t" \
1615 ".pushsection .discard.instr_end \n\t" \
1616 ".long 1b - . \n\t" \
1617 ".popsection \n\t" \
1619 _ASM_EXTABLE(666b, 667b)
1621 #define KVM_ARCH_WANT_MMU_NOTIFIER
1622 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end,
1624 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
1625 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
1626 int kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
1627 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1628 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1629 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1630 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1631 void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event);
1632 void kvm_vcpu_reload_apic_access_page(struct kvm_vcpu *vcpu);
1634 int kvm_pv_send_ipi(struct kvm *kvm, unsigned long ipi_bitmap_low,
1635 unsigned long ipi_bitmap_high, u32 min,
1636 unsigned long icr, int op_64_bit);
1638 void kvm_define_user_return_msr(unsigned index, u32 msr);
1639 int kvm_set_user_return_msr(unsigned index, u64 val, u64 mask);
1641 u64 kvm_scale_tsc(struct kvm_vcpu *vcpu, u64 tsc);
1642 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc);
1644 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu);
1645 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1647 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
1648 void kvm_make_scan_ioapic_request(struct kvm *kvm);
1649 void kvm_make_scan_ioapic_request_mask(struct kvm *kvm,
1650 unsigned long *vcpu_bitmap);
1652 bool kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1653 struct kvm_async_pf *work);
1654 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1655 struct kvm_async_pf *work);
1656 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1657 struct kvm_async_pf *work);
1658 void kvm_arch_async_page_present_queued(struct kvm_vcpu *vcpu);
1659 bool kvm_arch_can_dequeue_async_page_present(struct kvm_vcpu *vcpu);
1660 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1662 int kvm_skip_emulated_instruction(struct kvm_vcpu *vcpu);
1663 int kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1664 void __kvm_request_immediate_exit(struct kvm_vcpu *vcpu);
1666 int kvm_is_in_guest(void);
1668 int __x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa, u32 size);
1669 bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu);
1670 bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu);
1672 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
1673 struct kvm_vcpu **dest_vcpu);
1675 void kvm_set_msi_irq(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e,
1676 struct kvm_lapic_irq *irq);
1678 static inline bool kvm_irq_is_postable(struct kvm_lapic_irq *irq)
1680 /* We can only post Fixed and LowPrio IRQs */
1681 return (irq->delivery_mode == APIC_DM_FIXED ||
1682 irq->delivery_mode == APIC_DM_LOWEST);
1685 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu)
1687 if (kvm_x86_ops.vcpu_blocking)
1688 kvm_x86_ops.vcpu_blocking(vcpu);
1691 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu)
1693 if (kvm_x86_ops.vcpu_unblocking)
1694 kvm_x86_ops.vcpu_unblocking(vcpu);
1697 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
1699 static inline int kvm_cpu_get_apicid(int mps_cpu)
1701 #ifdef CONFIG_X86_LOCAL_APIC
1702 return default_cpu_present_to_apicid(mps_cpu);
1709 #define put_smstate(type, buf, offset, val) \
1710 *(type *)((buf) + (offset) - 0x7e00) = val
1712 #define GET_SMSTATE(type, buf, offset) \
1713 (*(type *)((buf) + (offset) - 0x7e00))
1715 #endif /* _ASM_X86_KVM_HOST_H */