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_VMCS12_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)
91 #define CR0_RESERVED_BITS \
92 (~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \
93 | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \
94 | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG))
96 #define CR4_RESERVED_BITS \
97 (~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\
98 | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE \
99 | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR | X86_CR4_PCIDE \
100 | X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_FSGSBASE \
101 | X86_CR4_OSXMMEXCPT | X86_CR4_LA57 | X86_CR4_VMXE \
102 | X86_CR4_SMAP | X86_CR4_PKE | X86_CR4_UMIP))
104 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR)
108 #define INVALID_PAGE (~(hpa_t)0)
109 #define VALID_PAGE(x) ((x) != INVALID_PAGE)
111 #define UNMAPPED_GVA (~(gpa_t)0)
113 /* KVM Hugepage definitions for x86 */
114 #define KVM_MAX_HUGEPAGE_LEVEL PG_LEVEL_1G
115 #define KVM_NR_PAGE_SIZES (KVM_MAX_HUGEPAGE_LEVEL - PG_LEVEL_4K + 1)
116 #define KVM_HPAGE_GFN_SHIFT(x) (((x) - 1) * 9)
117 #define KVM_HPAGE_SHIFT(x) (PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
118 #define KVM_HPAGE_SIZE(x) (1UL << KVM_HPAGE_SHIFT(x))
119 #define KVM_HPAGE_MASK(x) (~(KVM_HPAGE_SIZE(x) - 1))
120 #define KVM_PAGES_PER_HPAGE(x) (KVM_HPAGE_SIZE(x) / PAGE_SIZE)
122 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
124 /* KVM_HPAGE_GFN_SHIFT(PG_LEVEL_4K) must be 0. */
125 return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
126 (base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
129 #define KVM_PERMILLE_MMU_PAGES 20
130 #define KVM_MIN_ALLOC_MMU_PAGES 64UL
131 #define KVM_MMU_HASH_SHIFT 12
132 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
133 #define KVM_MIN_FREE_MMU_PAGES 5
134 #define KVM_REFILL_PAGES 25
135 #define KVM_MAX_CPUID_ENTRIES 80
136 #define KVM_NR_FIXED_MTRR_REGION 88
137 #define KVM_NR_VAR_MTRR 8
139 #define ASYNC_PF_PER_VCPU 64
142 VCPU_REGS_RAX = __VCPU_REGS_RAX,
143 VCPU_REGS_RCX = __VCPU_REGS_RCX,
144 VCPU_REGS_RDX = __VCPU_REGS_RDX,
145 VCPU_REGS_RBX = __VCPU_REGS_RBX,
146 VCPU_REGS_RSP = __VCPU_REGS_RSP,
147 VCPU_REGS_RBP = __VCPU_REGS_RBP,
148 VCPU_REGS_RSI = __VCPU_REGS_RSI,
149 VCPU_REGS_RDI = __VCPU_REGS_RDI,
151 VCPU_REGS_R8 = __VCPU_REGS_R8,
152 VCPU_REGS_R9 = __VCPU_REGS_R9,
153 VCPU_REGS_R10 = __VCPU_REGS_R10,
154 VCPU_REGS_R11 = __VCPU_REGS_R11,
155 VCPU_REGS_R12 = __VCPU_REGS_R12,
156 VCPU_REGS_R13 = __VCPU_REGS_R13,
157 VCPU_REGS_R14 = __VCPU_REGS_R14,
158 VCPU_REGS_R15 = __VCPU_REGS_R15,
163 VCPU_EXREG_PDPTR = NR_VCPU_REGS,
169 VCPU_EXREG_EXIT_INFO_1,
170 VCPU_EXREG_EXIT_INFO_2,
184 enum exit_fastpath_completion {
186 EXIT_FASTPATH_REENTER_GUEST,
187 EXIT_FASTPATH_EXIT_HANDLED,
189 typedef enum exit_fastpath_completion fastpath_t;
191 struct x86_emulate_ctxt;
192 struct x86_exception;
194 enum x86_intercept_stage;
196 #define KVM_NR_DB_REGS 4
198 #define DR6_BD (1 << 13)
199 #define DR6_BS (1 << 14)
200 #define DR6_BT (1 << 15)
201 #define DR6_RTM (1 << 16)
202 #define DR6_FIXED_1 0xfffe0ff0
203 #define DR6_INIT 0xffff0ff0
204 #define DR6_VOLATILE 0x0001e00f
206 #define DR7_BP_EN_MASK 0x000000ff
207 #define DR7_GE (1 << 9)
208 #define DR7_GD (1 << 13)
209 #define DR7_FIXED_1 0x00000400
210 #define DR7_VOLATILE 0xffff2bff
212 #define PFERR_PRESENT_BIT 0
213 #define PFERR_WRITE_BIT 1
214 #define PFERR_USER_BIT 2
215 #define PFERR_RSVD_BIT 3
216 #define PFERR_FETCH_BIT 4
217 #define PFERR_PK_BIT 5
218 #define PFERR_GUEST_FINAL_BIT 32
219 #define PFERR_GUEST_PAGE_BIT 33
221 #define PFERR_PRESENT_MASK (1U << PFERR_PRESENT_BIT)
222 #define PFERR_WRITE_MASK (1U << PFERR_WRITE_BIT)
223 #define PFERR_USER_MASK (1U << PFERR_USER_BIT)
224 #define PFERR_RSVD_MASK (1U << PFERR_RSVD_BIT)
225 #define PFERR_FETCH_MASK (1U << PFERR_FETCH_BIT)
226 #define PFERR_PK_MASK (1U << PFERR_PK_BIT)
227 #define PFERR_GUEST_FINAL_MASK (1ULL << PFERR_GUEST_FINAL_BIT)
228 #define PFERR_GUEST_PAGE_MASK (1ULL << PFERR_GUEST_PAGE_BIT)
230 #define PFERR_NESTED_GUEST_PAGE (PFERR_GUEST_PAGE_MASK | \
234 /* apic attention bits */
235 #define KVM_APIC_CHECK_VAPIC 0
237 * The following bit is set with PV-EOI, unset on EOI.
238 * We detect PV-EOI changes by guest by comparing
239 * this bit with PV-EOI in guest memory.
240 * See the implementation in apic_update_pv_eoi.
242 #define KVM_APIC_PV_EOI_PENDING 1
244 struct kvm_kernel_irq_routing_entry;
247 * the pages used as guest page table on soft mmu are tracked by
248 * kvm_memory_slot.arch.gfn_track which is 16 bits, so the role bits used
249 * by indirect shadow page can not be more than 15 bits.
251 * Currently, we used 14 bits that are @level, @gpte_is_8_bytes, @quadrant, @access,
252 * @nxe, @cr0_wp, @smep_andnot_wp and @smap_andnot_wp.
254 union kvm_mmu_page_role {
258 unsigned gpte_is_8_bytes:1;
265 unsigned smep_andnot_wp:1;
266 unsigned smap_andnot_wp:1;
267 unsigned ad_disabled:1;
268 unsigned guest_mode:1;
272 * This is left at the top of the word so that
273 * kvm_memslots_for_spte_role can extract it with a
274 * simple shift. While there is room, give it a whole
275 * byte so it is also faster to load it from memory.
281 union kvm_mmu_extended_role {
283 * This structure complements kvm_mmu_page_role caching everything needed for
284 * MMU configuration. If nothing in both these structures changed, MMU
285 * re-configuration can be skipped. @valid bit is set on first usage so we don't
286 * treat all-zero structure as valid data.
290 unsigned int valid:1;
291 unsigned int execonly:1;
292 unsigned int cr0_pg:1;
293 unsigned int cr4_pae:1;
294 unsigned int cr4_pse:1;
295 unsigned int cr4_pke:1;
296 unsigned int cr4_smap:1;
297 unsigned int cr4_smep:1;
298 unsigned int maxphyaddr:6;
305 union kvm_mmu_page_role base;
306 union kvm_mmu_extended_role ext;
310 struct kvm_rmap_head {
314 struct kvm_pio_request {
315 unsigned long linear_rip;
322 #define PT64_ROOT_MAX_LEVEL 5
324 struct rsvd_bits_validate {
325 u64 rsvd_bits_mask[2][PT64_ROOT_MAX_LEVEL];
329 struct kvm_mmu_root_info {
334 #define KVM_MMU_ROOT_INFO_INVALID \
335 ((struct kvm_mmu_root_info) { .pgd = INVALID_PAGE, .hpa = INVALID_PAGE })
337 #define KVM_MMU_NUM_PREV_ROOTS 3
342 * x86 supports 4 paging modes (5-level 64-bit, 4-level 64-bit, 3-level 32-bit,
343 * and 2-level 32-bit). The kvm_mmu structure abstracts the details of the
347 unsigned long (*get_guest_pgd)(struct kvm_vcpu *vcpu);
348 u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
349 int (*page_fault)(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u32 err,
351 void (*inject_page_fault)(struct kvm_vcpu *vcpu,
352 struct x86_exception *fault);
353 gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gpa_t gva_or_gpa,
354 u32 access, struct x86_exception *exception);
355 gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
356 struct x86_exception *exception);
357 int (*sync_page)(struct kvm_vcpu *vcpu,
358 struct kvm_mmu_page *sp);
359 void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva, hpa_t root_hpa);
360 void (*update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
361 u64 *spte, const void *pte);
364 union kvm_mmu_role mmu_role;
366 u8 shadow_root_level;
369 struct kvm_mmu_root_info prev_roots[KVM_MMU_NUM_PREV_ROOTS];
372 * Bitmap; bit set = permission fault
373 * Byte index: page fault error code [4:1]
374 * Bit index: pte permissions in ACC_* format
379 * The pkru_mask indicates if protection key checks are needed. It
380 * consists of 16 domains indexed by page fault error code bits [4:1],
381 * with PFEC.RSVD replaced by ACC_USER_MASK from the page tables.
382 * Each domain has 2 bits which are ANDed with AD and WD from PKRU.
390 * check zero bits on shadow page table entries, these
391 * bits include not only hardware reserved bits but also
392 * the bits spte never used.
394 struct rsvd_bits_validate shadow_zero_check;
396 struct rsvd_bits_validate guest_rsvd_check;
398 /* Can have large pages at levels 2..last_nonleaf_level-1. */
399 u8 last_nonleaf_level;
403 u64 pdptrs[4]; /* pae */
406 struct kvm_tlb_range {
421 struct perf_event *perf_event;
422 struct kvm_vcpu *vcpu;
424 * eventsel value for general purpose counters,
425 * ctrl value for fixed counters.
431 unsigned nr_arch_gp_counters;
432 unsigned nr_arch_fixed_counters;
433 unsigned available_event_types;
438 u64 counter_bitmask[2];
439 u64 global_ctrl_mask;
440 u64 global_ovf_ctrl_mask;
443 struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
444 struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED];
445 struct irq_work irq_work;
446 DECLARE_BITMAP(reprogram_pmi, X86_PMC_IDX_MAX);
447 DECLARE_BITMAP(all_valid_pmc_idx, X86_PMC_IDX_MAX);
448 DECLARE_BITMAP(pmc_in_use, X86_PMC_IDX_MAX);
451 * The gate to release perf_events not marked in
452 * pmc_in_use only once in a vcpu time slice.
457 * The total number of programmed perf_events and it helps to avoid
458 * redundant check before cleanup if guest don't use vPMU at all.
466 KVM_DEBUGREG_BP_ENABLED = 1,
467 KVM_DEBUGREG_WONT_EXIT = 2,
468 KVM_DEBUGREG_RELOAD = 4,
471 struct kvm_mtrr_range {
474 struct list_head node;
478 struct kvm_mtrr_range var_ranges[KVM_NR_VAR_MTRR];
479 mtrr_type fixed_ranges[KVM_NR_FIXED_MTRR_REGION];
482 struct list_head head;
485 /* Hyper-V SynIC timer */
486 struct kvm_vcpu_hv_stimer {
487 struct hrtimer timer;
489 union hv_stimer_config config;
492 struct hv_message msg;
496 /* Hyper-V synthetic interrupt controller (SynIC)*/
497 struct kvm_vcpu_hv_synic {
502 atomic64_t sint[HV_SYNIC_SINT_COUNT];
503 atomic_t sint_to_gsi[HV_SYNIC_SINT_COUNT];
504 DECLARE_BITMAP(auto_eoi_bitmap, 256);
505 DECLARE_BITMAP(vec_bitmap, 256);
507 bool dont_zero_synic_pages;
510 /* Hyper-V per vcpu emulation context */
515 struct kvm_vcpu_hv_synic synic;
516 struct kvm_hyperv_exit exit;
517 struct kvm_vcpu_hv_stimer stimer[HV_SYNIC_STIMER_COUNT];
518 DECLARE_BITMAP(stimer_pending_bitmap, HV_SYNIC_STIMER_COUNT);
522 struct kvm_vcpu_arch {
524 * rip and regs accesses must go through
525 * kvm_{register,rip}_{read,write} functions.
527 unsigned long regs[NR_VCPU_REGS];
532 unsigned long cr0_guest_owned_bits;
536 unsigned long cr4_guest_owned_bits;
537 unsigned long cr4_guest_rsvd_bits;
544 struct kvm_lapic *apic; /* kernel irqchip context */
546 bool load_eoi_exitmap_pending;
547 DECLARE_BITMAP(ioapic_handled_vectors, 256);
548 unsigned long apic_attention;
549 int32_t apic_arb_prio;
551 u64 ia32_misc_enable_msr;
554 bool tpr_access_reporting;
557 u64 microcode_version;
558 u64 arch_capabilities;
559 u64 perf_capabilities;
562 * Paging state of the vcpu
564 * If the vcpu runs in guest mode with two level paging this still saves
565 * the paging mode of the l1 guest. This context is always used to
570 /* Non-nested MMU for L1 */
571 struct kvm_mmu root_mmu;
573 /* L1 MMU when running nested */
574 struct kvm_mmu guest_mmu;
577 * Paging state of an L2 guest (used for nested npt)
579 * This context will save all necessary information to walk page tables
580 * of an L2 guest. This context is only initialized for page table
581 * walking and not for faulting since we never handle l2 page faults on
584 struct kvm_mmu nested_mmu;
587 * Pointer to the mmu context currently used for
588 * gva_to_gpa translations.
590 struct kvm_mmu *walk_mmu;
592 struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
593 struct kvm_mmu_memory_cache mmu_shadow_page_cache;
594 struct kvm_mmu_memory_cache mmu_gfn_array_cache;
595 struct kvm_mmu_memory_cache mmu_page_header_cache;
598 * QEMU userspace and the guest each have their own FPU state.
599 * In vcpu_run, we switch between the user and guest FPU contexts.
600 * While running a VCPU, the VCPU thread will have the guest FPU
603 * Note that while the PKRU state lives inside the fpu registers,
604 * it is switched out separately at VMENTER and VMEXIT time. The
605 * "guest_fpu" state here contains the guest FPU context, with the
608 struct fpu *user_fpu;
609 struct fpu *guest_fpu;
612 u64 guest_supported_xcr0;
614 struct kvm_pio_request pio;
617 u8 event_exit_inst_len;
619 struct kvm_queued_exception {
625 unsigned long payload;
630 struct kvm_queued_interrupt {
636 int halt_request; /* real mode on Intel only */
639 struct kvm_cpuid_entry2 cpuid_entries[KVM_MAX_CPUID_ENTRIES];
644 /* emulate context */
646 struct x86_emulate_ctxt *emulate_ctxt;
647 bool emulate_regs_need_sync_to_vcpu;
648 bool emulate_regs_need_sync_from_vcpu;
649 int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
652 struct pvclock_vcpu_time_info hv_clock;
653 unsigned int hw_tsc_khz;
654 struct gfn_to_hva_cache pv_time;
655 bool pv_time_enabled;
656 /* set guest stopped flag in pvclock flags field */
657 bool pvclock_set_guest_stopped_request;
663 struct gfn_to_pfn_cache cache;
670 u64 tsc_offset_adjustment;
673 u64 this_tsc_generation;
675 bool tsc_always_catchup;
676 s8 virtual_tsc_shift;
677 u32 virtual_tsc_mult;
679 s64 ia32_tsc_adjust_msr;
680 u64 msr_ia32_power_ctl;
681 u64 tsc_scaling_ratio;
683 atomic_t nmi_queued; /* unprocessed asynchronous NMIs */
684 unsigned nmi_pending; /* NMI queued after currently running handler */
685 bool nmi_injected; /* Trying to inject an NMI this entry */
686 bool smi_pending; /* SMI queued after currently running handler */
688 struct kvm_mtrr mtrr_state;
691 unsigned switch_db_regs;
692 unsigned long db[KVM_NR_DB_REGS];
695 unsigned long eff_db[KVM_NR_DB_REGS];
696 unsigned long guest_debug_dr7;
697 u64 msr_platform_info;
698 u64 msr_misc_features_enables;
706 /* Cache MMIO info */
708 unsigned mmio_access;
714 /* used for guest single stepping over the given code position */
715 unsigned long singlestep_rip;
717 struct kvm_vcpu_hv hyperv;
719 cpumask_var_t wbinvd_dirty_mask;
721 unsigned long last_retry_eip;
722 unsigned long last_retry_addr;
726 gfn_t gfns[ASYNC_PF_PER_VCPU];
727 struct gfn_to_hva_cache data;
728 u64 msr_en_val; /* MSR_KVM_ASYNC_PF_EN */
729 u64 msr_int_val; /* MSR_KVM_ASYNC_PF_INT */
734 unsigned long nested_apf_token;
735 bool delivery_as_pf_vmexit;
736 bool pageready_pending;
739 /* OSVW MSRs (AMD only) */
747 struct gfn_to_hva_cache data;
750 u64 msr_kvm_poll_control;
753 * Indicates the guest is trying to write a gfn that contains one or
754 * more of the PTEs used to translate the write itself, i.e. the access
755 * is changing its own translation in the guest page tables. KVM exits
756 * to userspace if emulation of the faulting instruction fails and this
757 * flag is set, as KVM cannot make forward progress.
759 * If emulation fails for a write to guest page tables, KVM unprotects
760 * (zaps) the shadow page for the target gfn and resumes the guest to
761 * retry the non-emulatable instruction (on hardware). Unprotecting the
762 * gfn doesn't allow forward progress for a self-changing access because
763 * doing so also zaps the translation for the gfn, i.e. retrying the
764 * instruction will hit a !PRESENT fault, which results in a new shadow
765 * page and sends KVM back to square one.
767 bool write_fault_to_shadow_pgtable;
769 /* set at EPT violation at this point */
770 unsigned long exit_qualification;
772 /* pv related host specific info */
777 int pending_ioapic_eoi;
778 int pending_external_vector;
780 /* be preempted when it's in kernel-mode(cpl=0) */
781 bool preempted_in_kernel;
783 /* Flush the L1 Data cache for L1TF mitigation on VMENTER */
786 /* Host CPU on which VM-entry was most recently attempted */
787 unsigned int last_vmentry_cpu;
789 /* AMD MSRC001_0015 Hardware Configuration */
793 struct kvm_lpage_info {
797 struct kvm_arch_memory_slot {
798 struct kvm_rmap_head *rmap[KVM_NR_PAGE_SIZES];
799 struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
800 unsigned short *gfn_track[KVM_PAGE_TRACK_MAX];
804 * We use as the mode the number of bits allocated in the LDR for the
805 * logical processor ID. It happens that these are all powers of two.
806 * This makes it is very easy to detect cases where the APICs are
807 * configured for multiple modes; in that case, we cannot use the map and
808 * hence cannot use kvm_irq_delivery_to_apic_fast either.
810 #define KVM_APIC_MODE_XAPIC_CLUSTER 4
811 #define KVM_APIC_MODE_XAPIC_FLAT 8
812 #define KVM_APIC_MODE_X2APIC 16
814 struct kvm_apic_map {
819 struct kvm_lapic *xapic_flat_map[8];
820 struct kvm_lapic *xapic_cluster_map[16][4];
822 struct kvm_lapic *phys_map[];
825 /* Hyper-V synthetic debugger (SynDbg)*/
826 struct kvm_hv_syndbg {
837 /* Hyper-V emulation context */
839 struct mutex hv_lock;
844 /* Hyper-v based guest crash (NT kernel bugcheck) parameters */
845 u64 hv_crash_param[HV_X64_MSR_CRASH_PARAMS];
848 struct ms_hyperv_tsc_page tsc_ref;
850 struct idr conn_to_evt;
852 u64 hv_reenlightenment_control;
853 u64 hv_tsc_emulation_control;
854 u64 hv_tsc_emulation_status;
856 /* How many vCPUs have VP index != vCPU index */
857 atomic_t num_mismatched_vp_indexes;
859 struct hv_partition_assist_pg *hv_pa_pg;
860 struct kvm_hv_syndbg hv_syndbg;
863 enum kvm_irqchip_mode {
865 KVM_IRQCHIP_KERNEL, /* created with KVM_CREATE_IRQCHIP */
866 KVM_IRQCHIP_SPLIT, /* created with KVM_CAP_SPLIT_IRQCHIP */
869 #define APICV_INHIBIT_REASON_DISABLE 0
870 #define APICV_INHIBIT_REASON_HYPERV 1
871 #define APICV_INHIBIT_REASON_NESTED 2
872 #define APICV_INHIBIT_REASON_IRQWIN 3
873 #define APICV_INHIBIT_REASON_PIT_REINJ 4
874 #define APICV_INHIBIT_REASON_X2APIC 5
877 unsigned long n_used_mmu_pages;
878 unsigned long n_requested_mmu_pages;
879 unsigned long n_max_mmu_pages;
880 unsigned int indirect_shadow_pages;
882 struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
884 * Hash table of struct kvm_mmu_page.
886 struct list_head active_mmu_pages;
887 struct list_head zapped_obsolete_pages;
888 struct list_head lpage_disallowed_mmu_pages;
889 struct kvm_page_track_notifier_node mmu_sp_tracker;
890 struct kvm_page_track_notifier_head track_notifier_head;
892 struct list_head assigned_dev_head;
893 struct iommu_domain *iommu_domain;
894 bool iommu_noncoherent;
895 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA
896 atomic_t noncoherent_dma_count;
897 #define __KVM_HAVE_ARCH_ASSIGNED_DEVICE
898 atomic_t assigned_device_count;
899 struct kvm_pic *vpic;
900 struct kvm_ioapic *vioapic;
901 struct kvm_pit *vpit;
902 atomic_t vapics_in_nmi_mode;
903 struct mutex apic_map_lock;
904 struct kvm_apic_map *apic_map;
905 atomic_t apic_map_dirty;
907 bool apic_access_page_done;
908 unsigned long apicv_inhibit_reasons;
915 bool cstate_in_guest;
917 unsigned long irq_sources_bitmap;
919 raw_spinlock_t tsc_write_lock;
926 u64 cur_tsc_generation;
927 int nr_vcpus_matched_tsc;
929 spinlock_t pvclock_gtod_sync_lock;
930 bool use_master_clock;
931 u64 master_kernel_ns;
932 u64 master_cycle_now;
933 struct delayed_work kvmclock_update_work;
934 struct delayed_work kvmclock_sync_work;
936 struct kvm_xen_hvm_config xen_hvm_config;
938 /* reads protected by irq_srcu, writes by irq_lock */
939 struct hlist_head mask_notifier_list;
941 struct kvm_hv hyperv;
943 #ifdef CONFIG_KVM_MMU_AUDIT
947 bool backwards_tsc_observed;
948 bool boot_vcpu_runs_old_kvmclock;
953 enum kvm_irqchip_mode irqchip_mode;
954 u8 nr_reserved_ioapic_pins;
956 bool disabled_lapic_found;
959 bool x2apic_broadcast_quirk_disabled;
961 bool guest_can_read_msr_platform_info;
962 bool exception_payload_enabled;
964 struct kvm_pmu_event_filter *pmu_event_filter;
965 struct task_struct *nx_lpage_recovery_thread;
969 ulong mmu_shadow_zapped;
971 ulong mmu_pte_updated;
972 ulong mmu_pde_zapped;
975 ulong mmu_cache_miss;
977 ulong remote_tlb_flush;
979 ulong nx_lpage_splits;
980 ulong max_mmu_page_hash_collisions;
983 struct kvm_vcpu_stat {
993 u64 irq_window_exits;
994 u64 nmi_window_exits;
997 u64 halt_successful_poll;
998 u64 halt_attempted_poll;
999 u64 halt_poll_invalid;
1001 u64 request_irq_exits;
1003 u64 host_state_reload;
1006 u64 insn_emulation_fail;
1011 u64 halt_poll_success_ns;
1012 u64 halt_poll_fail_ns;
1015 struct x86_instruction_info;
1018 bool host_initiated;
1023 struct kvm_lapic_irq {
1031 bool msi_redir_hint;
1034 static inline u16 kvm_lapic_irq_dest_mode(bool dest_mode_logical)
1036 return dest_mode_logical ? APIC_DEST_LOGICAL : APIC_DEST_PHYSICAL;
1039 struct kvm_x86_ops {
1040 int (*hardware_enable)(void);
1041 void (*hardware_disable)(void);
1042 void (*hardware_unsetup)(void);
1043 bool (*cpu_has_accelerated_tpr)(void);
1044 bool (*has_emulated_msr)(u32 index);
1045 void (*vcpu_after_set_cpuid)(struct kvm_vcpu *vcpu);
1047 unsigned int vm_size;
1048 int (*vm_init)(struct kvm *kvm);
1049 void (*vm_destroy)(struct kvm *kvm);
1051 /* Create, but do not attach this VCPU */
1052 int (*vcpu_create)(struct kvm_vcpu *vcpu);
1053 void (*vcpu_free)(struct kvm_vcpu *vcpu);
1054 void (*vcpu_reset)(struct kvm_vcpu *vcpu, bool init_event);
1056 void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
1057 void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
1058 void (*vcpu_put)(struct kvm_vcpu *vcpu);
1060 void (*update_exception_bitmap)(struct kvm_vcpu *vcpu);
1061 int (*get_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1062 int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
1063 u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
1064 void (*get_segment)(struct kvm_vcpu *vcpu,
1065 struct kvm_segment *var, int seg);
1066 int (*get_cpl)(struct kvm_vcpu *vcpu);
1067 void (*set_segment)(struct kvm_vcpu *vcpu,
1068 struct kvm_segment *var, int seg);
1069 void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
1070 void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
1071 int (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
1072 void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
1073 void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1074 void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1075 void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1076 void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
1077 void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu);
1078 void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
1079 void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
1080 unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
1081 void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
1083 void (*tlb_flush_all)(struct kvm_vcpu *vcpu);
1084 void (*tlb_flush_current)(struct kvm_vcpu *vcpu);
1085 int (*tlb_remote_flush)(struct kvm *kvm);
1086 int (*tlb_remote_flush_with_range)(struct kvm *kvm,
1087 struct kvm_tlb_range *range);
1090 * Flush any TLB entries associated with the given GVA.
1091 * Does not need to flush GPA->HPA mappings.
1092 * Can potentially get non-canonical addresses through INVLPGs, which
1093 * the implementation may choose to ignore if appropriate.
1095 void (*tlb_flush_gva)(struct kvm_vcpu *vcpu, gva_t addr);
1098 * Flush any TLB entries created by the guest. Like tlb_flush_gva(),
1099 * does not need to flush GPA->HPA mappings.
1101 void (*tlb_flush_guest)(struct kvm_vcpu *vcpu);
1103 enum exit_fastpath_completion (*run)(struct kvm_vcpu *vcpu);
1104 int (*handle_exit)(struct kvm_vcpu *vcpu,
1105 enum exit_fastpath_completion exit_fastpath);
1106 int (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
1107 void (*update_emulated_instruction)(struct kvm_vcpu *vcpu);
1108 void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
1109 u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu);
1110 void (*patch_hypercall)(struct kvm_vcpu *vcpu,
1111 unsigned char *hypercall_addr);
1112 void (*set_irq)(struct kvm_vcpu *vcpu);
1113 void (*set_nmi)(struct kvm_vcpu *vcpu);
1114 void (*queue_exception)(struct kvm_vcpu *vcpu);
1115 void (*cancel_injection)(struct kvm_vcpu *vcpu);
1116 int (*interrupt_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1117 int (*nmi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1118 bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
1119 void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
1120 void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
1121 void (*enable_irq_window)(struct kvm_vcpu *vcpu);
1122 void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
1123 bool (*check_apicv_inhibit_reasons)(ulong bit);
1124 void (*pre_update_apicv_exec_ctrl)(struct kvm *kvm, bool activate);
1125 void (*refresh_apicv_exec_ctrl)(struct kvm_vcpu *vcpu);
1126 void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
1127 void (*hwapic_isr_update)(struct kvm_vcpu *vcpu, int isr);
1128 bool (*guest_apic_has_interrupt)(struct kvm_vcpu *vcpu);
1129 void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
1130 void (*set_virtual_apic_mode)(struct kvm_vcpu *vcpu);
1131 void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu);
1132 int (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
1133 int (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
1134 int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
1135 int (*set_identity_map_addr)(struct kvm *kvm, u64 ident_addr);
1136 u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
1138 void (*load_mmu_pgd)(struct kvm_vcpu *vcpu, unsigned long pgd,
1141 bool (*has_wbinvd_exit)(void);
1143 /* Returns actual tsc_offset set in active VMCS */
1144 u64 (*write_l1_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
1146 void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2);
1148 int (*check_intercept)(struct kvm_vcpu *vcpu,
1149 struct x86_instruction_info *info,
1150 enum x86_intercept_stage stage,
1151 struct x86_exception *exception);
1152 void (*handle_exit_irqoff)(struct kvm_vcpu *vcpu);
1154 void (*request_immediate_exit)(struct kvm_vcpu *vcpu);
1156 void (*sched_in)(struct kvm_vcpu *kvm, int cpu);
1159 * Arch-specific dirty logging hooks. These hooks are only supposed to
1160 * be valid if the specific arch has hardware-accelerated dirty logging
1161 * mechanism. Currently only for PML on VMX.
1163 * - slot_enable_log_dirty:
1164 * called when enabling log dirty mode for the slot.
1165 * - slot_disable_log_dirty:
1166 * called when disabling log dirty mode for the slot.
1167 * also called when slot is created with log dirty disabled.
1168 * - flush_log_dirty:
1169 * called before reporting dirty_bitmap to userspace.
1170 * - enable_log_dirty_pt_masked:
1171 * called when reenabling log dirty for the GFNs in the mask after
1172 * corresponding bits are cleared in slot->dirty_bitmap.
1174 void (*slot_enable_log_dirty)(struct kvm *kvm,
1175 struct kvm_memory_slot *slot);
1176 void (*slot_disable_log_dirty)(struct kvm *kvm,
1177 struct kvm_memory_slot *slot);
1178 void (*flush_log_dirty)(struct kvm *kvm);
1179 void (*enable_log_dirty_pt_masked)(struct kvm *kvm,
1180 struct kvm_memory_slot *slot,
1181 gfn_t offset, unsigned long mask);
1183 /* pmu operations of sub-arch */
1184 const struct kvm_pmu_ops *pmu_ops;
1185 const struct kvm_x86_nested_ops *nested_ops;
1188 * Architecture specific hooks for vCPU blocking due to
1190 * Returns for .pre_block():
1191 * - 0 means continue to block the vCPU.
1192 * - 1 means we cannot block the vCPU since some event
1193 * happens during this period, such as, 'ON' bit in
1194 * posted-interrupts descriptor is set.
1196 int (*pre_block)(struct kvm_vcpu *vcpu);
1197 void (*post_block)(struct kvm_vcpu *vcpu);
1199 void (*vcpu_blocking)(struct kvm_vcpu *vcpu);
1200 void (*vcpu_unblocking)(struct kvm_vcpu *vcpu);
1202 int (*update_pi_irte)(struct kvm *kvm, unsigned int host_irq,
1203 uint32_t guest_irq, bool set);
1204 void (*apicv_post_state_restore)(struct kvm_vcpu *vcpu);
1205 bool (*dy_apicv_has_pending_interrupt)(struct kvm_vcpu *vcpu);
1207 int (*set_hv_timer)(struct kvm_vcpu *vcpu, u64 guest_deadline_tsc,
1209 void (*cancel_hv_timer)(struct kvm_vcpu *vcpu);
1211 void (*setup_mce)(struct kvm_vcpu *vcpu);
1213 int (*smi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
1214 int (*pre_enter_smm)(struct kvm_vcpu *vcpu, char *smstate);
1215 int (*pre_leave_smm)(struct kvm_vcpu *vcpu, const char *smstate);
1216 void (*enable_smi_window)(struct kvm_vcpu *vcpu);
1218 int (*mem_enc_op)(struct kvm *kvm, void __user *argp);
1219 int (*mem_enc_reg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1220 int (*mem_enc_unreg_region)(struct kvm *kvm, struct kvm_enc_region *argp);
1222 int (*get_msr_feature)(struct kvm_msr_entry *entry);
1224 bool (*need_emulation_on_page_fault)(struct kvm_vcpu *vcpu);
1226 bool (*apic_init_signal_blocked)(struct kvm_vcpu *vcpu);
1227 int (*enable_direct_tlbflush)(struct kvm_vcpu *vcpu);
1229 void (*migrate_timers)(struct kvm_vcpu *vcpu);
1232 struct kvm_x86_nested_ops {
1233 int (*check_events)(struct kvm_vcpu *vcpu);
1234 bool (*hv_timer_pending)(struct kvm_vcpu *vcpu);
1235 int (*get_state)(struct kvm_vcpu *vcpu,
1236 struct kvm_nested_state __user *user_kvm_nested_state,
1237 unsigned user_data_size);
1238 int (*set_state)(struct kvm_vcpu *vcpu,
1239 struct kvm_nested_state __user *user_kvm_nested_state,
1240 struct kvm_nested_state *kvm_state);
1241 bool (*get_vmcs12_pages)(struct kvm_vcpu *vcpu);
1242 int (*write_log_dirty)(struct kvm_vcpu *vcpu, gpa_t l2_gpa);
1244 int (*enable_evmcs)(struct kvm_vcpu *vcpu,
1245 uint16_t *vmcs_version);
1246 uint16_t (*get_evmcs_version)(struct kvm_vcpu *vcpu);
1249 struct kvm_x86_init_ops {
1250 int (*cpu_has_kvm_support)(void);
1251 int (*disabled_by_bios)(void);
1252 int (*check_processor_compatibility)(void);
1253 int (*hardware_setup)(void);
1255 struct kvm_x86_ops *runtime_ops;
1258 struct kvm_arch_async_pf {
1265 extern u64 __read_mostly host_efer;
1266 extern bool __read_mostly allow_smaller_maxphyaddr;
1267 extern struct kvm_x86_ops kvm_x86_ops;
1269 #define __KVM_HAVE_ARCH_VM_ALLOC
1270 static inline struct kvm *kvm_arch_alloc_vm(void)
1272 return __vmalloc(kvm_x86_ops.vm_size, GFP_KERNEL_ACCOUNT | __GFP_ZERO);
1274 void kvm_arch_free_vm(struct kvm *kvm);
1276 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
1277 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
1279 if (kvm_x86_ops.tlb_remote_flush &&
1280 !kvm_x86_ops.tlb_remote_flush(kvm))
1286 int kvm_mmu_module_init(void);
1287 void kvm_mmu_module_exit(void);
1289 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
1290 int kvm_mmu_create(struct kvm_vcpu *vcpu);
1291 void kvm_mmu_init_vm(struct kvm *kvm);
1292 void kvm_mmu_uninit_vm(struct kvm *kvm);
1293 void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
1294 u64 dirty_mask, u64 nx_mask, u64 x_mask, u64 p_mask,
1295 u64 acc_track_mask, u64 me_mask);
1297 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
1298 void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
1299 struct kvm_memory_slot *memslot,
1301 void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
1302 const struct kvm_memory_slot *memslot);
1303 void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
1304 struct kvm_memory_slot *memslot);
1305 void kvm_mmu_slot_largepage_remove_write_access(struct kvm *kvm,
1306 struct kvm_memory_slot *memslot);
1307 void kvm_mmu_slot_set_dirty(struct kvm *kvm,
1308 struct kvm_memory_slot *memslot);
1309 void kvm_mmu_clear_dirty_pt_masked(struct kvm *kvm,
1310 struct kvm_memory_slot *slot,
1311 gfn_t gfn_offset, unsigned long mask);
1312 void kvm_mmu_zap_all(struct kvm *kvm);
1313 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, u64 gen);
1314 unsigned long kvm_mmu_calculate_default_mmu_pages(struct kvm *kvm);
1315 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned long kvm_nr_mmu_pages);
1317 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
1318 bool pdptrs_changed(struct kvm_vcpu *vcpu);
1320 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
1321 const void *val, int bytes);
1323 struct kvm_irq_mask_notifier {
1324 void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
1326 struct hlist_node link;
1329 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
1330 struct kvm_irq_mask_notifier *kimn);
1331 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
1332 struct kvm_irq_mask_notifier *kimn);
1333 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
1336 extern bool tdp_enabled;
1338 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
1340 /* control of guest tsc rate supported? */
1341 extern bool kvm_has_tsc_control;
1342 /* maximum supported tsc_khz for guests */
1343 extern u32 kvm_max_guest_tsc_khz;
1344 /* number of bits of the fractional part of the TSC scaling ratio */
1345 extern u8 kvm_tsc_scaling_ratio_frac_bits;
1346 /* maximum allowed value of TSC scaling ratio */
1347 extern u64 kvm_max_tsc_scaling_ratio;
1348 /* 1ull << kvm_tsc_scaling_ratio_frac_bits */
1349 extern u64 kvm_default_tsc_scaling_ratio;
1351 extern u64 kvm_mce_cap_supported;
1354 * EMULTYPE_NO_DECODE - Set when re-emulating an instruction (after completing
1355 * userspace I/O) to indicate that the emulation context
1356 * should be resued as is, i.e. skip initialization of
1357 * emulation context, instruction fetch and decode.
1359 * EMULTYPE_TRAP_UD - Set when emulating an intercepted #UD from hardware.
1360 * Indicates that only select instructions (tagged with
1361 * EmulateOnUD) should be emulated (to minimize the emulator
1362 * attack surface). See also EMULTYPE_TRAP_UD_FORCED.
1364 * EMULTYPE_SKIP - Set when emulating solely to skip an instruction, i.e. to
1365 * decode the instruction length. For use *only* by
1366 * kvm_x86_ops.skip_emulated_instruction() implementations.
1368 * EMULTYPE_ALLOW_RETRY_PF - Set when the emulator should resume the guest to
1369 * retry native execution under certain conditions,
1370 * Can only be set in conjunction with EMULTYPE_PF.
1372 * EMULTYPE_TRAP_UD_FORCED - Set when emulating an intercepted #UD that was
1373 * triggered by KVM's magic "force emulation" prefix,
1374 * which is opt in via module param (off by default).
1375 * Bypasses EmulateOnUD restriction despite emulating
1376 * due to an intercepted #UD (see EMULTYPE_TRAP_UD).
1377 * Used to test the full emulator from userspace.
1379 * EMULTYPE_VMWARE_GP - Set when emulating an intercepted #GP for VMware
1380 * backdoor emulation, which is opt in via module param.
1381 * VMware backoor emulation handles select instructions
1382 * and reinjects the #GP for all other cases.
1384 * EMULTYPE_PF - Set when emulating MMIO by way of an intercepted #PF, in which
1385 * case the CR2/GPA value pass on the stack is valid.
1387 #define EMULTYPE_NO_DECODE (1 << 0)
1388 #define EMULTYPE_TRAP_UD (1 << 1)
1389 #define EMULTYPE_SKIP (1 << 2)
1390 #define EMULTYPE_ALLOW_RETRY_PF (1 << 3)
1391 #define EMULTYPE_TRAP_UD_FORCED (1 << 4)
1392 #define EMULTYPE_VMWARE_GP (1 << 5)
1393 #define EMULTYPE_PF (1 << 6)
1395 int kvm_emulate_instruction(struct kvm_vcpu *vcpu, int emulation_type);
1396 int kvm_emulate_instruction_from_buffer(struct kvm_vcpu *vcpu,
1397 void *insn, int insn_len);
1399 void kvm_enable_efer_bits(u64);
1400 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
1401 int __kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data, bool host_initiated);
1402 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data);
1403 int kvm_set_msr(struct kvm_vcpu *vcpu, u32 index, u64 data);
1404 int kvm_emulate_rdmsr(struct kvm_vcpu *vcpu);
1405 int kvm_emulate_wrmsr(struct kvm_vcpu *vcpu);
1407 int kvm_fast_pio(struct kvm_vcpu *vcpu, int size, unsigned short port, int in);
1408 int kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
1409 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
1410 int kvm_vcpu_halt(struct kvm_vcpu *vcpu);
1411 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
1413 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
1414 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
1415 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector);
1417 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
1418 int reason, bool has_error_code, u32 error_code);
1420 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
1421 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
1422 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
1423 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
1424 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
1425 int kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
1426 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
1427 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
1428 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
1429 int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr);
1431 int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1432 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1434 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
1435 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
1436 bool kvm_rdpmc(struct kvm_vcpu *vcpu);
1438 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1439 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1440 void kvm_queue_exception_p(struct kvm_vcpu *vcpu, unsigned nr, unsigned long payload);
1441 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1442 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1443 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
1444 bool kvm_inject_emulated_page_fault(struct kvm_vcpu *vcpu,
1445 struct x86_exception *fault);
1446 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1447 gfn_t gfn, void *data, int offset, int len,
1449 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
1450 bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr);
1452 static inline int __kvm_irq_line_state(unsigned long *irq_state,
1453 int irq_source_id, int level)
1455 /* Logical OR for level trig interrupt */
1457 __set_bit(irq_source_id, irq_state);
1459 __clear_bit(irq_source_id, irq_state);
1461 return !!(*irq_state);
1464 #define KVM_MMU_ROOT_CURRENT BIT(0)
1465 #define KVM_MMU_ROOT_PREVIOUS(i) BIT(1+i)
1466 #define KVM_MMU_ROOTS_ALL (~0UL)
1468 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
1469 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
1471 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
1473 void kvm_update_dr7(struct kvm_vcpu *vcpu);
1475 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
1476 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
1477 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
1478 int kvm_mmu_load(struct kvm_vcpu *vcpu);
1479 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
1480 void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu);
1481 void kvm_mmu_free_roots(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1482 ulong roots_to_free);
1483 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1484 struct x86_exception *exception);
1485 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
1486 struct x86_exception *exception);
1487 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
1488 struct x86_exception *exception);
1489 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
1490 struct x86_exception *exception);
1491 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
1492 struct x86_exception *exception);
1494 bool kvm_apicv_activated(struct kvm *kvm);
1495 void kvm_apicv_init(struct kvm *kvm, bool enable);
1496 void kvm_vcpu_update_apicv(struct kvm_vcpu *vcpu);
1497 void kvm_request_apicv_update(struct kvm *kvm, bool activate,
1500 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
1502 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u64 error_code,
1503 void *insn, int insn_len);
1504 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
1505 void kvm_mmu_invalidate_gva(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1506 gva_t gva, hpa_t root_hpa);
1507 void kvm_mmu_invpcid_gva(struct kvm_vcpu *vcpu, gva_t gva, unsigned long pcid);
1508 void kvm_mmu_new_pgd(struct kvm_vcpu *vcpu, gpa_t new_pgd, bool skip_tlb_flush,
1509 bool skip_mmu_sync);
1511 void kvm_configure_mmu(bool enable_tdp, int tdp_max_root_level,
1512 int tdp_huge_page_level);
1514 static inline u16 kvm_read_ldt(void)
1517 asm("sldt %0" : "=g"(ldt));
1521 static inline void kvm_load_ldt(u16 sel)
1523 asm("lldt %0" : : "rm"(sel));
1526 #ifdef CONFIG_X86_64
1527 static inline unsigned long read_msr(unsigned long msr)
1536 static inline u32 get_rdx_init_val(void)
1538 return 0x600; /* P6 family */
1541 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
1543 kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
1546 #define TSS_IOPB_BASE_OFFSET 0x66
1547 #define TSS_BASE_SIZE 0x68
1548 #define TSS_IOPB_SIZE (65536 / 8)
1549 #define TSS_REDIRECTION_SIZE (256 / 8)
1550 #define RMODE_TSS_SIZE \
1551 (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
1554 TASK_SWITCH_CALL = 0,
1555 TASK_SWITCH_IRET = 1,
1556 TASK_SWITCH_JMP = 2,
1557 TASK_SWITCH_GATE = 3,
1560 #define HF_GIF_MASK (1 << 0)
1561 #define HF_NMI_MASK (1 << 3)
1562 #define HF_IRET_MASK (1 << 4)
1563 #define HF_GUEST_MASK (1 << 5) /* VCPU is in guest-mode */
1564 #define HF_SMM_MASK (1 << 6)
1565 #define HF_SMM_INSIDE_NMI_MASK (1 << 7)
1567 #define __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
1568 #define KVM_ADDRESS_SPACE_NUM 2
1570 #define kvm_arch_vcpu_memslots_id(vcpu) ((vcpu)->arch.hflags & HF_SMM_MASK ? 1 : 0)
1571 #define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, (role).smm)
1573 asmlinkage void kvm_spurious_fault(void);
1576 * Hardware virtualization extension instructions may fault if a
1577 * reboot turns off virtualization while processes are running.
1578 * Usually after catching the fault we just panic; during reboot
1579 * instead the instruction is ignored.
1581 #define __kvm_handle_fault_on_reboot(insn) \
1587 ".pushsection .discard.instr_begin \n\t" \
1588 ".long 1b - . \n\t" \
1589 ".popsection \n\t" \
1590 "call kvm_spurious_fault \n\t" \
1592 ".pushsection .discard.instr_end \n\t" \
1593 ".long 1b - . \n\t" \
1594 ".popsection \n\t" \
1596 _ASM_EXTABLE(666b, 667b)
1598 #define KVM_ARCH_WANT_MMU_NOTIFIER
1599 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end,
1601 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
1602 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
1603 int kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
1604 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1605 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1606 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1607 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1608 void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event);
1609 void kvm_vcpu_reload_apic_access_page(struct kvm_vcpu *vcpu);
1611 int kvm_pv_send_ipi(struct kvm *kvm, unsigned long ipi_bitmap_low,
1612 unsigned long ipi_bitmap_high, u32 min,
1613 unsigned long icr, int op_64_bit);
1615 void kvm_define_shared_msr(unsigned index, u32 msr);
1616 int kvm_set_shared_msr(unsigned index, u64 val, u64 mask);
1618 u64 kvm_scale_tsc(struct kvm_vcpu *vcpu, u64 tsc);
1619 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc);
1621 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu);
1622 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1624 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
1625 void kvm_make_scan_ioapic_request(struct kvm *kvm);
1626 void kvm_make_scan_ioapic_request_mask(struct kvm *kvm,
1627 unsigned long *vcpu_bitmap);
1629 bool kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1630 struct kvm_async_pf *work);
1631 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1632 struct kvm_async_pf *work);
1633 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1634 struct kvm_async_pf *work);
1635 void kvm_arch_async_page_present_queued(struct kvm_vcpu *vcpu);
1636 bool kvm_arch_can_dequeue_async_page_present(struct kvm_vcpu *vcpu);
1637 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1639 int kvm_skip_emulated_instruction(struct kvm_vcpu *vcpu);
1640 int kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1641 void __kvm_request_immediate_exit(struct kvm_vcpu *vcpu);
1643 int kvm_is_in_guest(void);
1645 int __x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa, u32 size);
1646 bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu);
1647 bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu);
1649 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
1650 struct kvm_vcpu **dest_vcpu);
1652 void kvm_set_msi_irq(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e,
1653 struct kvm_lapic_irq *irq);
1655 static inline bool kvm_irq_is_postable(struct kvm_lapic_irq *irq)
1657 /* We can only post Fixed and LowPrio IRQs */
1658 return (irq->delivery_mode == APIC_DM_FIXED ||
1659 irq->delivery_mode == APIC_DM_LOWEST);
1662 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu)
1664 if (kvm_x86_ops.vcpu_blocking)
1665 kvm_x86_ops.vcpu_blocking(vcpu);
1668 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu)
1670 if (kvm_x86_ops.vcpu_unblocking)
1671 kvm_x86_ops.vcpu_unblocking(vcpu);
1674 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
1676 static inline int kvm_cpu_get_apicid(int mps_cpu)
1678 #ifdef CONFIG_X86_LOCAL_APIC
1679 return default_cpu_present_to_apicid(mps_cpu);
1686 #define put_smstate(type, buf, offset, val) \
1687 *(type *)((buf) + (offset) - 0x7e00) = val
1689 #define GET_SMSTATE(type, buf, offset) \
1690 (*(type *)((buf) + (offset) - 0x7e00))
1692 #endif /* _ASM_X86_KVM_HOST_H */