2 * Kernel-based Virtual Machine driver for Linux
4 * This header defines architecture specific interfaces, x86 version
6 * This work is licensed under the terms of the GNU GPL, version 2. See
7 * the COPYING file in the top-level directory.
11 #ifndef _ASM_X86_KVM_HOST_H
12 #define _ASM_X86_KVM_HOST_H
14 #include <linux/types.h>
16 #include <linux/mmu_notifier.h>
17 #include <linux/tracepoint.h>
18 #include <linux/cpumask.h>
19 #include <linux/irq_work.h>
21 #include <linux/kvm.h>
22 #include <linux/kvm_para.h>
23 #include <linux/kvm_types.h>
24 #include <linux/perf_event.h>
25 #include <linux/pvclock_gtod.h>
26 #include <linux/clocksource.h>
27 #include <linux/irqbypass.h>
28 #include <linux/hyperv.h>
30 #include <asm/pvclock-abi.h>
33 #include <asm/msr-index.h>
35 #include <asm/kvm_page_track.h>
37 #define KVM_MAX_VCPUS 255
38 #define KVM_SOFT_MAX_VCPUS 160
39 #define KVM_USER_MEM_SLOTS 509
40 /* memory slots that are not exposed to userspace */
41 #define KVM_PRIVATE_MEM_SLOTS 3
42 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
44 #define KVM_PIO_PAGE_OFFSET 1
45 #define KVM_COALESCED_MMIO_PAGE_OFFSET 2
46 #define KVM_HALT_POLL_NS_DEFAULT 500000
48 #define KVM_IRQCHIP_NUM_PINS KVM_IOAPIC_NUM_PINS
50 /* x86-specific vcpu->requests bit members */
51 #define KVM_REQ_MIGRATE_TIMER 8
52 #define KVM_REQ_REPORT_TPR_ACCESS 9
53 #define KVM_REQ_TRIPLE_FAULT 10
54 #define KVM_REQ_MMU_SYNC 11
55 #define KVM_REQ_CLOCK_UPDATE 12
56 #define KVM_REQ_DEACTIVATE_FPU 13
57 #define KVM_REQ_EVENT 14
58 #define KVM_REQ_APF_HALT 15
59 #define KVM_REQ_STEAL_UPDATE 16
60 #define KVM_REQ_NMI 17
61 #define KVM_REQ_PMU 18
62 #define KVM_REQ_PMI 19
63 #define KVM_REQ_SMI 20
64 #define KVM_REQ_MASTERCLOCK_UPDATE 21
65 #define KVM_REQ_MCLOCK_INPROGRESS 22
66 #define KVM_REQ_SCAN_IOAPIC 23
67 #define KVM_REQ_GLOBAL_CLOCK_UPDATE 24
68 #define KVM_REQ_APIC_PAGE_RELOAD 25
69 #define KVM_REQ_HV_CRASH 26
70 #define KVM_REQ_IOAPIC_EOI_EXIT 27
71 #define KVM_REQ_HV_RESET 28
72 #define KVM_REQ_HV_EXIT 29
73 #define KVM_REQ_HV_STIMER 30
75 #define CR0_RESERVED_BITS \
76 (~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \
77 | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \
78 | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG))
80 #define CR3_L_MODE_RESERVED_BITS 0xFFFFFF0000000000ULL
81 #define CR3_PCID_INVD BIT_64(63)
82 #define CR4_RESERVED_BITS \
83 (~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\
84 | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE \
85 | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR | X86_CR4_PCIDE \
86 | X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_FSGSBASE \
87 | X86_CR4_OSXMMEXCPT | X86_CR4_VMXE | X86_CR4_SMAP))
89 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR)
93 #define INVALID_PAGE (~(hpa_t)0)
94 #define VALID_PAGE(x) ((x) != INVALID_PAGE)
96 #define UNMAPPED_GVA (~(gpa_t)0)
98 /* KVM Hugepage definitions for x86 */
99 #define KVM_NR_PAGE_SIZES 3
100 #define KVM_HPAGE_GFN_SHIFT(x) (((x) - 1) * 9)
101 #define KVM_HPAGE_SHIFT(x) (PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
102 #define KVM_HPAGE_SIZE(x) (1UL << KVM_HPAGE_SHIFT(x))
103 #define KVM_HPAGE_MASK(x) (~(KVM_HPAGE_SIZE(x) - 1))
104 #define KVM_PAGES_PER_HPAGE(x) (KVM_HPAGE_SIZE(x) / PAGE_SIZE)
106 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
108 /* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */
109 return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
110 (base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
113 #define KVM_PERMILLE_MMU_PAGES 20
114 #define KVM_MIN_ALLOC_MMU_PAGES 64
115 #define KVM_MMU_HASH_SHIFT 10
116 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
117 #define KVM_MIN_FREE_MMU_PAGES 5
118 #define KVM_REFILL_PAGES 25
119 #define KVM_MAX_CPUID_ENTRIES 80
120 #define KVM_NR_FIXED_MTRR_REGION 88
121 #define KVM_NR_VAR_MTRR 8
123 #define ASYNC_PF_PER_VCPU 64
149 VCPU_EXREG_PDPTR = NR_VCPU_REGS,
166 #include <asm/kvm_emulate.h>
168 #define KVM_NR_MEM_OBJS 40
170 #define KVM_NR_DB_REGS 4
172 #define DR6_BD (1 << 13)
173 #define DR6_BS (1 << 14)
174 #define DR6_RTM (1 << 16)
175 #define DR6_FIXED_1 0xfffe0ff0
176 #define DR6_INIT 0xffff0ff0
177 #define DR6_VOLATILE 0x0001e00f
179 #define DR7_BP_EN_MASK 0x000000ff
180 #define DR7_GE (1 << 9)
181 #define DR7_GD (1 << 13)
182 #define DR7_FIXED_1 0x00000400
183 #define DR7_VOLATILE 0xffff2bff
185 #define PFERR_PRESENT_BIT 0
186 #define PFERR_WRITE_BIT 1
187 #define PFERR_USER_BIT 2
188 #define PFERR_RSVD_BIT 3
189 #define PFERR_FETCH_BIT 4
191 #define PFERR_PRESENT_MASK (1U << PFERR_PRESENT_BIT)
192 #define PFERR_WRITE_MASK (1U << PFERR_WRITE_BIT)
193 #define PFERR_USER_MASK (1U << PFERR_USER_BIT)
194 #define PFERR_RSVD_MASK (1U << PFERR_RSVD_BIT)
195 #define PFERR_FETCH_MASK (1U << PFERR_FETCH_BIT)
197 /* apic attention bits */
198 #define KVM_APIC_CHECK_VAPIC 0
200 * The following bit is set with PV-EOI, unset on EOI.
201 * We detect PV-EOI changes by guest by comparing
202 * this bit with PV-EOI in guest memory.
203 * See the implementation in apic_update_pv_eoi.
205 #define KVM_APIC_PV_EOI_PENDING 1
207 struct kvm_kernel_irq_routing_entry;
210 * We don't want allocation failures within the mmu code, so we preallocate
211 * enough memory for a single page fault in a cache.
213 struct kvm_mmu_memory_cache {
215 void *objects[KVM_NR_MEM_OBJS];
219 * the pages used as guest page table on soft mmu are tracked by
220 * kvm_memory_slot.arch.gfn_track which is 16 bits, so the role bits used
221 * by indirect shadow page can not be more than 15 bits.
223 * Currently, we used 14 bits that are @level, @cr4_pae, @quadrant, @access,
224 * @nxe, @cr0_wp, @smep_andnot_wp and @smap_andnot_wp.
226 union kvm_mmu_page_role {
237 unsigned smep_andnot_wp:1;
238 unsigned smap_andnot_wp:1;
242 * This is left at the top of the word so that
243 * kvm_memslots_for_spte_role can extract it with a
244 * simple shift. While there is room, give it a whole
245 * byte so it is also faster to load it from memory.
251 struct kvm_rmap_head {
255 struct kvm_mmu_page {
256 struct list_head link;
257 struct hlist_node hash_link;
260 * The following two entries are used to key the shadow page in the
264 union kvm_mmu_page_role role;
267 /* hold the gfn of each spte inside spt */
270 int root_count; /* Currently serving as active root */
271 unsigned int unsync_children;
272 struct kvm_rmap_head parent_ptes; /* rmap pointers to parent sptes */
274 /* The page is obsolete if mmu_valid_gen != kvm->arch.mmu_valid_gen. */
275 unsigned long mmu_valid_gen;
277 DECLARE_BITMAP(unsync_child_bitmap, 512);
281 * Used out of the mmu-lock to avoid reading spte values while an
282 * update is in progress; see the comments in __get_spte_lockless().
284 int clear_spte_count;
287 /* Number of writes since the last time traversal visited this page. */
288 atomic_t write_flooding_count;
291 struct kvm_pio_request {
298 struct rsvd_bits_validate {
299 u64 rsvd_bits_mask[2][4];
304 * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
305 * 32-bit). The kvm_mmu structure abstracts the details of the current mmu
309 void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long root);
310 unsigned long (*get_cr3)(struct kvm_vcpu *vcpu);
311 u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
312 int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err,
314 void (*inject_page_fault)(struct kvm_vcpu *vcpu,
315 struct x86_exception *fault);
316 gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva, u32 access,
317 struct x86_exception *exception);
318 gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
319 struct x86_exception *exception);
320 int (*sync_page)(struct kvm_vcpu *vcpu,
321 struct kvm_mmu_page *sp);
322 void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva);
323 void (*update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
324 u64 *spte, const void *pte);
327 int shadow_root_level;
328 union kvm_mmu_page_role base_role;
332 * Bitmap; bit set = permission fault
333 * Byte index: page fault error code [4:1]
334 * Bit index: pte permissions in ACC_* format
342 * check zero bits on shadow page table entries, these
343 * bits include not only hardware reserved bits but also
344 * the bits spte never used.
346 struct rsvd_bits_validate shadow_zero_check;
348 struct rsvd_bits_validate guest_rsvd_check;
350 /* Can have large pages at levels 2..last_nonleaf_level-1. */
351 u8 last_nonleaf_level;
355 u64 pdptrs[4]; /* pae */
368 struct perf_event *perf_event;
369 struct kvm_vcpu *vcpu;
373 unsigned nr_arch_gp_counters;
374 unsigned nr_arch_fixed_counters;
375 unsigned available_event_types;
380 u64 counter_bitmask[2];
381 u64 global_ctrl_mask;
384 struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
385 struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED];
386 struct irq_work irq_work;
393 KVM_DEBUGREG_BP_ENABLED = 1,
394 KVM_DEBUGREG_WONT_EXIT = 2,
395 KVM_DEBUGREG_RELOAD = 4,
398 struct kvm_mtrr_range {
401 struct list_head node;
405 struct kvm_mtrr_range var_ranges[KVM_NR_VAR_MTRR];
406 mtrr_type fixed_ranges[KVM_NR_FIXED_MTRR_REGION];
409 struct list_head head;
412 /* Hyper-V SynIC timer */
413 struct kvm_vcpu_hv_stimer {
414 struct hrtimer timer;
419 struct hv_message msg;
423 /* Hyper-V synthetic interrupt controller (SynIC)*/
424 struct kvm_vcpu_hv_synic {
429 atomic64_t sint[HV_SYNIC_SINT_COUNT];
430 atomic_t sint_to_gsi[HV_SYNIC_SINT_COUNT];
431 DECLARE_BITMAP(auto_eoi_bitmap, 256);
432 DECLARE_BITMAP(vec_bitmap, 256);
436 /* Hyper-V per vcpu emulation context */
440 struct kvm_vcpu_hv_synic synic;
441 struct kvm_hyperv_exit exit;
442 struct kvm_vcpu_hv_stimer stimer[HV_SYNIC_STIMER_COUNT];
443 DECLARE_BITMAP(stimer_pending_bitmap, HV_SYNIC_STIMER_COUNT);
446 struct kvm_vcpu_arch {
448 * rip and regs accesses must go through
449 * kvm_{register,rip}_{read,write} functions.
451 unsigned long regs[NR_VCPU_REGS];
456 unsigned long cr0_guest_owned_bits;
460 unsigned long cr4_guest_owned_bits;
465 struct kvm_lapic *apic; /* kernel irqchip context */
467 DECLARE_BITMAP(ioapic_handled_vectors, 256);
468 unsigned long apic_attention;
469 int32_t apic_arb_prio;
471 u64 ia32_misc_enable_msr;
473 bool tpr_access_reporting;
477 * Paging state of the vcpu
479 * If the vcpu runs in guest mode with two level paging this still saves
480 * the paging mode of the l1 guest. This context is always used to
486 * Paging state of an L2 guest (used for nested npt)
488 * This context will save all necessary information to walk page tables
489 * of the an L2 guest. This context is only initialized for page table
490 * walking and not for faulting since we never handle l2 page faults on
493 struct kvm_mmu nested_mmu;
496 * Pointer to the mmu context currently used for
497 * gva_to_gpa translations.
499 struct kvm_mmu *walk_mmu;
501 struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
502 struct kvm_mmu_memory_cache mmu_page_cache;
503 struct kvm_mmu_memory_cache mmu_page_header_cache;
505 struct fpu guest_fpu;
507 u64 guest_supported_xcr0;
508 u32 guest_xstate_size;
510 struct kvm_pio_request pio;
513 u8 event_exit_inst_len;
515 struct kvm_queued_exception {
523 struct kvm_queued_interrupt {
529 int halt_request; /* real mode on Intel only */
532 struct kvm_cpuid_entry2 cpuid_entries[KVM_MAX_CPUID_ENTRIES];
536 /* emulate context */
538 struct x86_emulate_ctxt emulate_ctxt;
539 bool emulate_regs_need_sync_to_vcpu;
540 bool emulate_regs_need_sync_from_vcpu;
541 int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
544 struct pvclock_vcpu_time_info hv_clock;
545 unsigned int hw_tsc_khz;
546 struct gfn_to_hva_cache pv_time;
547 bool pv_time_enabled;
548 /* set guest stopped flag in pvclock flags field */
549 bool pvclock_set_guest_stopped_request;
555 struct gfn_to_hva_cache stime;
556 struct kvm_steal_time steal;
561 u64 tsc_offset_adjustment;
564 u64 this_tsc_generation;
566 bool tsc_always_catchup;
567 s8 virtual_tsc_shift;
568 u32 virtual_tsc_mult;
570 s64 ia32_tsc_adjust_msr;
571 u64 tsc_scaling_ratio;
573 atomic_t nmi_queued; /* unprocessed asynchronous NMIs */
574 unsigned nmi_pending; /* NMI queued after currently running handler */
575 bool nmi_injected; /* Trying to inject an NMI this entry */
576 bool smi_pending; /* SMI queued after currently running handler */
578 struct kvm_mtrr mtrr_state;
581 unsigned switch_db_regs;
582 unsigned long db[KVM_NR_DB_REGS];
585 unsigned long eff_db[KVM_NR_DB_REGS];
586 unsigned long guest_debug_dr7;
593 /* Cache MMIO info */
601 /* used for guest single stepping over the given code position */
602 unsigned long singlestep_rip;
604 struct kvm_vcpu_hv hyperv;
606 cpumask_var_t wbinvd_dirty_mask;
608 unsigned long last_retry_eip;
609 unsigned long last_retry_addr;
613 gfn_t gfns[roundup_pow_of_two(ASYNC_PF_PER_VCPU)];
614 struct gfn_to_hva_cache data;
620 /* OSVW MSRs (AMD only) */
628 struct gfn_to_hva_cache data;
632 * Indicate whether the access faults on its page table in guest
633 * which is set when fix page fault and used to detect unhandeable
636 bool write_fault_to_shadow_pgtable;
638 /* set at EPT violation at this point */
639 unsigned long exit_qualification;
641 /* pv related host specific info */
646 int pending_ioapic_eoi;
647 int pending_external_vector;
650 struct kvm_lpage_info {
654 struct kvm_arch_memory_slot {
655 struct kvm_rmap_head *rmap[KVM_NR_PAGE_SIZES];
656 struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
657 unsigned short *gfn_track[KVM_PAGE_TRACK_MAX];
661 * We use as the mode the number of bits allocated in the LDR for the
662 * logical processor ID. It happens that these are all powers of two.
663 * This makes it is very easy to detect cases where the APICs are
664 * configured for multiple modes; in that case, we cannot use the map and
665 * hence cannot use kvm_irq_delivery_to_apic_fast either.
667 #define KVM_APIC_MODE_XAPIC_CLUSTER 4
668 #define KVM_APIC_MODE_XAPIC_FLAT 8
669 #define KVM_APIC_MODE_X2APIC 16
671 struct kvm_apic_map {
674 struct kvm_lapic *phys_map[256];
675 /* first index is cluster id second is cpu id in a cluster */
676 struct kvm_lapic *logical_map[16][16];
679 /* Hyper-V emulation context */
685 /* Hyper-v based guest crash (NT kernel bugcheck) parameters */
686 u64 hv_crash_param[HV_X64_MSR_CRASH_PARAMS];
691 unsigned int n_used_mmu_pages;
692 unsigned int n_requested_mmu_pages;
693 unsigned int n_max_mmu_pages;
694 unsigned int indirect_shadow_pages;
695 unsigned long mmu_valid_gen;
696 struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
698 * Hash table of struct kvm_mmu_page.
700 struct list_head active_mmu_pages;
701 struct list_head zapped_obsolete_pages;
702 struct kvm_page_track_notifier_node mmu_sp_tracker;
703 struct kvm_page_track_notifier_head track_notifier_head;
705 struct list_head assigned_dev_head;
706 struct iommu_domain *iommu_domain;
707 bool iommu_noncoherent;
708 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA
709 atomic_t noncoherent_dma_count;
710 #define __KVM_HAVE_ARCH_ASSIGNED_DEVICE
711 atomic_t assigned_device_count;
712 struct kvm_pic *vpic;
713 struct kvm_ioapic *vioapic;
714 struct kvm_pit *vpit;
715 atomic_t vapics_in_nmi_mode;
716 struct mutex apic_map_lock;
717 struct kvm_apic_map *apic_map;
719 unsigned int tss_addr;
720 bool apic_access_page_done;
724 bool ept_identity_pagetable_done;
725 gpa_t ept_identity_map_addr;
727 unsigned long irq_sources_bitmap;
729 raw_spinlock_t tsc_write_lock;
736 u64 cur_tsc_generation;
737 int nr_vcpus_matched_tsc;
739 spinlock_t pvclock_gtod_sync_lock;
740 bool use_master_clock;
741 u64 master_kernel_ns;
742 cycle_t master_cycle_now;
743 struct delayed_work kvmclock_update_work;
744 struct delayed_work kvmclock_sync_work;
746 struct kvm_xen_hvm_config xen_hvm_config;
748 /* reads protected by irq_srcu, writes by irq_lock */
749 struct hlist_head mask_notifier_list;
751 struct kvm_hv hyperv;
753 #ifdef CONFIG_KVM_MMU_AUDIT
757 bool boot_vcpu_runs_old_kvmclock;
763 u8 nr_reserved_ioapic_pins;
765 bool disabled_lapic_found;
769 u32 mmu_shadow_zapped;
777 u32 remote_tlb_flush;
781 struct kvm_vcpu_stat {
791 u32 irq_window_exits;
792 u32 nmi_window_exits;
794 u32 halt_successful_poll;
795 u32 halt_attempted_poll;
797 u32 request_irq_exits;
799 u32 host_state_reload;
803 u32 insn_emulation_fail;
809 struct x86_instruction_info;
817 struct kvm_lapic_irq {
829 int (*cpu_has_kvm_support)(void); /* __init */
830 int (*disabled_by_bios)(void); /* __init */
831 int (*hardware_enable)(void);
832 void (*hardware_disable)(void);
833 void (*check_processor_compatibility)(void *rtn);
834 int (*hardware_setup)(void); /* __init */
835 void (*hardware_unsetup)(void); /* __exit */
836 bool (*cpu_has_accelerated_tpr)(void);
837 bool (*cpu_has_high_real_mode_segbase)(void);
838 void (*cpuid_update)(struct kvm_vcpu *vcpu);
840 /* Create, but do not attach this VCPU */
841 struct kvm_vcpu *(*vcpu_create)(struct kvm *kvm, unsigned id);
842 void (*vcpu_free)(struct kvm_vcpu *vcpu);
843 void (*vcpu_reset)(struct kvm_vcpu *vcpu, bool init_event);
845 void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
846 void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
847 void (*vcpu_put)(struct kvm_vcpu *vcpu);
849 void (*update_bp_intercept)(struct kvm_vcpu *vcpu);
850 int (*get_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
851 int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
852 u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
853 void (*get_segment)(struct kvm_vcpu *vcpu,
854 struct kvm_segment *var, int seg);
855 int (*get_cpl)(struct kvm_vcpu *vcpu);
856 void (*set_segment)(struct kvm_vcpu *vcpu,
857 struct kvm_segment *var, int seg);
858 void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
859 void (*decache_cr0_guest_bits)(struct kvm_vcpu *vcpu);
860 void (*decache_cr3)(struct kvm_vcpu *vcpu);
861 void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu);
862 void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
863 void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
864 int (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
865 void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
866 void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
867 void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
868 void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
869 void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
870 u64 (*get_dr6)(struct kvm_vcpu *vcpu);
871 void (*set_dr6)(struct kvm_vcpu *vcpu, unsigned long value);
872 void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu);
873 void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
874 void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
875 unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
876 void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
877 void (*fpu_activate)(struct kvm_vcpu *vcpu);
878 void (*fpu_deactivate)(struct kvm_vcpu *vcpu);
880 void (*tlb_flush)(struct kvm_vcpu *vcpu);
882 void (*run)(struct kvm_vcpu *vcpu);
883 int (*handle_exit)(struct kvm_vcpu *vcpu);
884 void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
885 void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
886 u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu);
887 void (*patch_hypercall)(struct kvm_vcpu *vcpu,
888 unsigned char *hypercall_addr);
889 void (*set_irq)(struct kvm_vcpu *vcpu);
890 void (*set_nmi)(struct kvm_vcpu *vcpu);
891 void (*queue_exception)(struct kvm_vcpu *vcpu, unsigned nr,
892 bool has_error_code, u32 error_code,
894 void (*cancel_injection)(struct kvm_vcpu *vcpu);
895 int (*interrupt_allowed)(struct kvm_vcpu *vcpu);
896 int (*nmi_allowed)(struct kvm_vcpu *vcpu);
897 bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
898 void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
899 void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
900 void (*enable_irq_window)(struct kvm_vcpu *vcpu);
901 void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
902 bool (*get_enable_apicv)(void);
903 void (*refresh_apicv_exec_ctrl)(struct kvm_vcpu *vcpu);
904 void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
905 void (*hwapic_isr_update)(struct kvm *kvm, int isr);
906 void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
907 void (*set_virtual_x2apic_mode)(struct kvm_vcpu *vcpu, bool set);
908 void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu, hpa_t hpa);
909 void (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
910 void (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
911 int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
912 int (*get_tdp_level)(void);
913 u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
914 int (*get_lpage_level)(void);
915 bool (*rdtscp_supported)(void);
916 bool (*invpcid_supported)(void);
917 void (*adjust_tsc_offset_guest)(struct kvm_vcpu *vcpu, s64 adjustment);
919 void (*set_tdp_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
921 void (*set_supported_cpuid)(u32 func, struct kvm_cpuid_entry2 *entry);
923 bool (*has_wbinvd_exit)(void);
925 u64 (*read_tsc_offset)(struct kvm_vcpu *vcpu);
926 void (*write_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
928 u64 (*read_l1_tsc)(struct kvm_vcpu *vcpu, u64 host_tsc);
930 void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2);
932 int (*check_intercept)(struct kvm_vcpu *vcpu,
933 struct x86_instruction_info *info,
934 enum x86_intercept_stage stage);
935 void (*handle_external_intr)(struct kvm_vcpu *vcpu);
936 bool (*mpx_supported)(void);
937 bool (*xsaves_supported)(void);
939 int (*check_nested_events)(struct kvm_vcpu *vcpu, bool external_intr);
941 void (*sched_in)(struct kvm_vcpu *kvm, int cpu);
944 * Arch-specific dirty logging hooks. These hooks are only supposed to
945 * be valid if the specific arch has hardware-accelerated dirty logging
946 * mechanism. Currently only for PML on VMX.
948 * - slot_enable_log_dirty:
949 * called when enabling log dirty mode for the slot.
950 * - slot_disable_log_dirty:
951 * called when disabling log dirty mode for the slot.
952 * also called when slot is created with log dirty disabled.
954 * called before reporting dirty_bitmap to userspace.
955 * - enable_log_dirty_pt_masked:
956 * called when reenabling log dirty for the GFNs in the mask after
957 * corresponding bits are cleared in slot->dirty_bitmap.
959 void (*slot_enable_log_dirty)(struct kvm *kvm,
960 struct kvm_memory_slot *slot);
961 void (*slot_disable_log_dirty)(struct kvm *kvm,
962 struct kvm_memory_slot *slot);
963 void (*flush_log_dirty)(struct kvm *kvm);
964 void (*enable_log_dirty_pt_masked)(struct kvm *kvm,
965 struct kvm_memory_slot *slot,
966 gfn_t offset, unsigned long mask);
967 /* pmu operations of sub-arch */
968 const struct kvm_pmu_ops *pmu_ops;
971 * Architecture specific hooks for vCPU blocking due to
973 * Returns for .pre_block():
974 * - 0 means continue to block the vCPU.
975 * - 1 means we cannot block the vCPU since some event
976 * happens during this period, such as, 'ON' bit in
977 * posted-interrupts descriptor is set.
979 int (*pre_block)(struct kvm_vcpu *vcpu);
980 void (*post_block)(struct kvm_vcpu *vcpu);
981 int (*update_pi_irte)(struct kvm *kvm, unsigned int host_irq,
982 uint32_t guest_irq, bool set);
985 struct kvm_arch_async_pf {
992 extern struct kvm_x86_ops *kvm_x86_ops;
994 int kvm_mmu_module_init(void);
995 void kvm_mmu_module_exit(void);
997 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
998 int kvm_mmu_create(struct kvm_vcpu *vcpu);
999 void kvm_mmu_setup(struct kvm_vcpu *vcpu);
1000 void kvm_mmu_init_vm(struct kvm *kvm);
1001 void kvm_mmu_uninit_vm(struct kvm *kvm);
1002 void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
1003 u64 dirty_mask, u64 nx_mask, u64 x_mask);
1005 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
1006 void kvm_mmu_slot_remove_write_access(struct kvm *kvm,
1007 struct kvm_memory_slot *memslot);
1008 void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
1009 const struct kvm_memory_slot *memslot);
1010 void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm,
1011 struct kvm_memory_slot *memslot);
1012 void kvm_mmu_slot_largepage_remove_write_access(struct kvm *kvm,
1013 struct kvm_memory_slot *memslot);
1014 void kvm_mmu_slot_set_dirty(struct kvm *kvm,
1015 struct kvm_memory_slot *memslot);
1016 void kvm_mmu_clear_dirty_pt_masked(struct kvm *kvm,
1017 struct kvm_memory_slot *slot,
1018 gfn_t gfn_offset, unsigned long mask);
1019 void kvm_mmu_zap_all(struct kvm *kvm);
1020 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, struct kvm_memslots *slots);
1021 unsigned int kvm_mmu_calculate_mmu_pages(struct kvm *kvm);
1022 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int kvm_nr_mmu_pages);
1024 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
1026 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
1027 const void *val, int bytes);
1029 struct kvm_irq_mask_notifier {
1030 void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
1032 struct hlist_node link;
1035 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
1036 struct kvm_irq_mask_notifier *kimn);
1037 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
1038 struct kvm_irq_mask_notifier *kimn);
1039 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
1042 extern bool tdp_enabled;
1044 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
1046 /* control of guest tsc rate supported? */
1047 extern bool kvm_has_tsc_control;
1048 /* maximum supported tsc_khz for guests */
1049 extern u32 kvm_max_guest_tsc_khz;
1050 /* number of bits of the fractional part of the TSC scaling ratio */
1051 extern u8 kvm_tsc_scaling_ratio_frac_bits;
1052 /* maximum allowed value of TSC scaling ratio */
1053 extern u64 kvm_max_tsc_scaling_ratio;
1055 enum emulation_result {
1056 EMULATE_DONE, /* no further processing */
1057 EMULATE_USER_EXIT, /* kvm_run ready for userspace exit */
1058 EMULATE_FAIL, /* can't emulate this instruction */
1061 #define EMULTYPE_NO_DECODE (1 << 0)
1062 #define EMULTYPE_TRAP_UD (1 << 1)
1063 #define EMULTYPE_SKIP (1 << 2)
1064 #define EMULTYPE_RETRY (1 << 3)
1065 #define EMULTYPE_NO_REEXECUTE (1 << 4)
1066 int x86_emulate_instruction(struct kvm_vcpu *vcpu, unsigned long cr2,
1067 int emulation_type, void *insn, int insn_len);
1069 static inline int emulate_instruction(struct kvm_vcpu *vcpu,
1072 return x86_emulate_instruction(vcpu, 0, emulation_type, NULL, 0);
1075 void kvm_enable_efer_bits(u64);
1076 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
1077 int kvm_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr);
1078 int kvm_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr);
1080 struct x86_emulate_ctxt;
1082 int kvm_fast_pio_out(struct kvm_vcpu *vcpu, int size, unsigned short port);
1083 void kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
1084 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
1085 int kvm_vcpu_halt(struct kvm_vcpu *vcpu);
1086 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
1088 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
1089 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
1090 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector);
1092 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
1093 int reason, bool has_error_code, u32 error_code);
1095 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
1096 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
1097 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
1098 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
1099 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
1100 int kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
1101 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
1102 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
1103 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
1104 int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr);
1106 int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1107 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
1109 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
1110 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
1111 bool kvm_rdpmc(struct kvm_vcpu *vcpu);
1113 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1114 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1115 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
1116 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
1117 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
1118 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
1119 gfn_t gfn, void *data, int offset, int len,
1121 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
1122 bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr);
1124 static inline int __kvm_irq_line_state(unsigned long *irq_state,
1125 int irq_source_id, int level)
1127 /* Logical OR for level trig interrupt */
1129 __set_bit(irq_source_id, irq_state);
1131 __clear_bit(irq_source_id, irq_state);
1133 return !!(*irq_state);
1136 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
1137 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
1139 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
1141 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
1142 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
1143 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
1144 int kvm_mmu_load(struct kvm_vcpu *vcpu);
1145 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
1146 void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu);
1147 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1148 struct x86_exception *exception);
1149 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
1150 struct x86_exception *exception);
1151 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
1152 struct x86_exception *exception);
1153 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
1154 struct x86_exception *exception);
1155 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
1156 struct x86_exception *exception);
1158 void kvm_vcpu_deactivate_apicv(struct kvm_vcpu *vcpu);
1160 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
1162 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva, u32 error_code,
1163 void *insn, int insn_len);
1164 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
1165 void kvm_mmu_new_cr3(struct kvm_vcpu *vcpu);
1167 void kvm_enable_tdp(void);
1168 void kvm_disable_tdp(void);
1170 static inline gpa_t translate_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
1171 struct x86_exception *exception)
1176 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
1178 struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
1180 return (struct kvm_mmu_page *)page_private(page);
1183 static inline u16 kvm_read_ldt(void)
1186 asm("sldt %0" : "=g"(ldt));
1190 static inline void kvm_load_ldt(u16 sel)
1192 asm("lldt %0" : : "rm"(sel));
1195 #ifdef CONFIG_X86_64
1196 static inline unsigned long read_msr(unsigned long msr)
1205 static inline u32 get_rdx_init_val(void)
1207 return 0x600; /* P6 family */
1210 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
1212 kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
1215 static inline u64 get_canonical(u64 la)
1217 return ((int64_t)la << 16) >> 16;
1220 static inline bool is_noncanonical_address(u64 la)
1222 #ifdef CONFIG_X86_64
1223 return get_canonical(la) != la;
1229 #define TSS_IOPB_BASE_OFFSET 0x66
1230 #define TSS_BASE_SIZE 0x68
1231 #define TSS_IOPB_SIZE (65536 / 8)
1232 #define TSS_REDIRECTION_SIZE (256 / 8)
1233 #define RMODE_TSS_SIZE \
1234 (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
1237 TASK_SWITCH_CALL = 0,
1238 TASK_SWITCH_IRET = 1,
1239 TASK_SWITCH_JMP = 2,
1240 TASK_SWITCH_GATE = 3,
1243 #define HF_GIF_MASK (1 << 0)
1244 #define HF_HIF_MASK (1 << 1)
1245 #define HF_VINTR_MASK (1 << 2)
1246 #define HF_NMI_MASK (1 << 3)
1247 #define HF_IRET_MASK (1 << 4)
1248 #define HF_GUEST_MASK (1 << 5) /* VCPU is in guest-mode */
1249 #define HF_SMM_MASK (1 << 6)
1250 #define HF_SMM_INSIDE_NMI_MASK (1 << 7)
1252 #define __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
1253 #define KVM_ADDRESS_SPACE_NUM 2
1255 #define kvm_arch_vcpu_memslots_id(vcpu) ((vcpu)->arch.hflags & HF_SMM_MASK ? 1 : 0)
1256 #define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, (role).smm)
1259 * Hardware virtualization extension instructions may fault if a
1260 * reboot turns off virtualization while processes are running.
1261 * Trap the fault and ignore the instruction if that happens.
1263 asmlinkage void kvm_spurious_fault(void);
1265 #define ____kvm_handle_fault_on_reboot(insn, cleanup_insn) \
1266 "666: " insn "\n\t" \
1268 ".pushsection .fixup, \"ax\" \n" \
1270 cleanup_insn "\n\t" \
1271 "cmpb $0, kvm_rebooting \n\t" \
1273 __ASM_SIZE(push) " $666b \n\t" \
1274 "call kvm_spurious_fault \n\t" \
1275 ".popsection \n\t" \
1276 _ASM_EXTABLE(666b, 667b)
1278 #define __kvm_handle_fault_on_reboot(insn) \
1279 ____kvm_handle_fault_on_reboot(insn, "")
1281 #define KVM_ARCH_WANT_MMU_NOTIFIER
1282 int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
1283 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end);
1284 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
1285 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
1286 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
1287 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1288 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1289 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1290 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1291 void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event);
1292 void kvm_vcpu_reload_apic_access_page(struct kvm_vcpu *vcpu);
1293 void kvm_arch_mmu_notifier_invalidate_page(struct kvm *kvm,
1294 unsigned long address);
1296 void kvm_define_shared_msr(unsigned index, u32 msr);
1297 int kvm_set_shared_msr(unsigned index, u64 val, u64 mask);
1299 u64 kvm_scale_tsc(struct kvm_vcpu *vcpu, u64 tsc);
1300 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc);
1302 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu);
1303 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1305 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
1306 void kvm_make_scan_ioapic_request(struct kvm *kvm);
1308 void kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1309 struct kvm_async_pf *work);
1310 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1311 struct kvm_async_pf *work);
1312 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1313 struct kvm_async_pf *work);
1314 bool kvm_arch_can_inject_async_page_present(struct kvm_vcpu *vcpu);
1315 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1317 void kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1319 int kvm_is_in_guest(void);
1321 int __x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa, u32 size);
1322 int x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa, u32 size);
1323 bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu);
1324 bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu);
1326 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
1327 struct kvm_vcpu **dest_vcpu);
1329 void kvm_set_msi_irq(struct kvm_kernel_irq_routing_entry *e,
1330 struct kvm_lapic_irq *irq);
1332 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu) {}
1333 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu) {}
1335 #endif /* _ASM_X86_KVM_HOST_H */