arch/x86/include/asm/kvm_host.h

   1 /*
   2  * Kernel-based Virtual Machine driver for Linux
   3  *
   4  * This header defines architecture specific interfaces, x86 version
   5  *
   6  * This work is licensed under the terms of the GNU GPL, version 2.  See
   7  * the COPYING file in the top-level directory.
   8  *
   9  */
  10
  11 #ifndef _ASM_X86_KVM_HOST_H
  12 #define _ASM_X86_KVM_HOST_H
  13
  14 #include <linux/types.h>
  15 #include <linux/mm.h>
  16 #include <linux/mmu_notifier.h>
  17 #include <linux/tracepoint.h>
  18 #include <linux/cpumask.h>
  19 #include <linux/irq_work.h>
  20
  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
  28 #include <asm/pvclock-abi.h>
  29 #include <asm/desc.h>
  30 #include <asm/mtrr.h>
  31 #include <asm/msr-index.h>
  32 #include <asm/asm.h>
  33
  34 #define KVM_MAX_VCPUS 255
  35 #define KVM_SOFT_MAX_VCPUS 160
  36 #define KVM_USER_MEM_SLOTS 125
  37 /* memory slots that are not exposed to userspace */
  38 #define KVM_PRIVATE_MEM_SLOTS 3
  39 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
  40
  41 #define KVM_MMIO_SIZE 16
  42
  43 #define KVM_PIO_PAGE_OFFSET 1
  44 #define KVM_COALESCED_MMIO_PAGE_OFFSET 2
  45
  46 #define KVM_IRQCHIP_NUM_PINS  KVM_IOAPIC_NUM_PINS
  47
  48 #define CR0_RESERVED_BITS                                               \
  49         (~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \
  50                           | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \
  51                           | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG))
  52
  53 #define CR3_PAE_RESERVED_BITS ((X86_CR3_PWT | X86_CR3_PCD) - 1)
  54 #define CR3_NONPAE_RESERVED_BITS ((PAGE_SIZE-1) & ~(X86_CR3_PWT | X86_CR3_PCD))
  55 #define CR3_PCID_ENABLED_RESERVED_BITS 0xFFFFFF0000000000ULL
  56 #define CR3_L_MODE_RESERVED_BITS (CR3_NONPAE_RESERVED_BITS |    \
  57                                   0xFFFFFF0000000000ULL)
  58 #define CR4_RESERVED_BITS                                               \
  59         (~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\
  60                           | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE     \
  61                           | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR | X86_CR4_PCIDE \
  62                           | X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_FSGSBASE \
  63                           | X86_CR4_OSXMMEXCPT | X86_CR4_VMXE))
  64
  65 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR)
  66
  67
  68
  69 #define INVALID_PAGE (~(hpa_t)0)
  70 #define VALID_PAGE(x) ((x) != INVALID_PAGE)
  71
  72 #define UNMAPPED_GVA (~(gpa_t)0)
  73
  74 /* KVM Hugepage definitions for x86 */
  75 #define KVM_NR_PAGE_SIZES       3
  76 #define KVM_HPAGE_GFN_SHIFT(x)  (((x) - 1) * 9)
  77 #define KVM_HPAGE_SHIFT(x)      (PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x))
  78 #define KVM_HPAGE_SIZE(x)       (1UL << KVM_HPAGE_SHIFT(x))
  79 #define KVM_HPAGE_MASK(x)       (~(KVM_HPAGE_SIZE(x) - 1))
  80 #define KVM_PAGES_PER_HPAGE(x)  (KVM_HPAGE_SIZE(x) / PAGE_SIZE)
  81
  82 #define SELECTOR_TI_MASK (1 << 2)
  83 #define SELECTOR_RPL_MASK 0x03
  84
  85 #define IOPL_SHIFT 12
  86
  87 #define KVM_PERMILLE_MMU_PAGES 20
  88 #define KVM_MIN_ALLOC_MMU_PAGES 64
  89 #define KVM_MMU_HASH_SHIFT 10
  90 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
  91 #define KVM_MIN_FREE_MMU_PAGES 5
  92 #define KVM_REFILL_PAGES 25
  93 #define KVM_MAX_CPUID_ENTRIES 80
  94 #define KVM_NR_FIXED_MTRR_REGION 88
  95 #define KVM_NR_VAR_MTRR 8
  96
  97 #define ASYNC_PF_PER_VCPU 64
  98
  99 struct kvm_vcpu;
 100 struct kvm;
 101 struct kvm_async_pf;
 102
 103 enum kvm_reg {
 104         VCPU_REGS_RAX = 0,
 105         VCPU_REGS_RCX = 1,
 106         VCPU_REGS_RDX = 2,
 107         VCPU_REGS_RBX = 3,
 108         VCPU_REGS_RSP = 4,
 109         VCPU_REGS_RBP = 5,
 110         VCPU_REGS_RSI = 6,
 111         VCPU_REGS_RDI = 7,
 112 #ifdef CONFIG_X86_64
 113         VCPU_REGS_R8 = 8,
 114         VCPU_REGS_R9 = 9,
 115         VCPU_REGS_R10 = 10,
 116         VCPU_REGS_R11 = 11,
 117         VCPU_REGS_R12 = 12,
 118         VCPU_REGS_R13 = 13,
 119         VCPU_REGS_R14 = 14,
 120         VCPU_REGS_R15 = 15,
 121 #endif
 122         VCPU_REGS_RIP,
 123         NR_VCPU_REGS
 124 };
 125
 126 enum kvm_reg_ex {
 127         VCPU_EXREG_PDPTR = NR_VCPU_REGS,
 128         VCPU_EXREG_CR3,
 129         VCPU_EXREG_RFLAGS,
 130         VCPU_EXREG_CPL,
 131         VCPU_EXREG_SEGMENTS,
 132 };
 133
 134 enum {
 135         VCPU_SREG_ES,
 136         VCPU_SREG_CS,
 137         VCPU_SREG_SS,
 138         VCPU_SREG_DS,
 139         VCPU_SREG_FS,
 140         VCPU_SREG_GS,
 141         VCPU_SREG_TR,
 142         VCPU_SREG_LDTR,
 143 };
 144
 145 #include <asm/kvm_emulate.h>
 146
 147 #define KVM_NR_MEM_OBJS 40
 148
 149 #define KVM_NR_DB_REGS  4
 150
 151 #define DR6_BD          (1 << 13)
 152 #define DR6_BS          (1 << 14)
 153 #define DR6_FIXED_1     0xffff0ff0
 154 #define DR6_VOLATILE    0x0000e00f
 155
 156 #define DR7_BP_EN_MASK  0x000000ff
 157 #define DR7_GE          (1 << 9)
 158 #define DR7_GD          (1 << 13)
 159 #define DR7_FIXED_1     0x00000400
 160 #define DR7_VOLATILE    0xffff23ff
 161
 162 /* apic attention bits */
 163 #define KVM_APIC_CHECK_VAPIC    0
 164 /*
 165  * The following bit is set with PV-EOI, unset on EOI.
 166  * We detect PV-EOI changes by guest by comparing
 167  * this bit with PV-EOI in guest memory.
 168  * See the implementation in apic_update_pv_eoi.
 169  */
 170 #define KVM_APIC_PV_EOI_PENDING 1
 171
 172 /*
 173  * We don't want allocation failures within the mmu code, so we preallocate
 174  * enough memory for a single page fault in a cache.
 175  */
 176 struct kvm_mmu_memory_cache {
 177         int nobjs;
 178         void *objects[KVM_NR_MEM_OBJS];
 179 };
 180
 181 /*
 182  * kvm_mmu_page_role, below, is defined as:
 183  *
 184  *   bits 0:3 - total guest paging levels (2-4, or zero for real mode)
 185  *   bits 4:7 - page table level for this shadow (1-4)
 186  *   bits 8:9 - page table quadrant for 2-level guests
 187  *   bit   16 - direct mapping of virtual to physical mapping at gfn
 188  *              used for real mode and two-dimensional paging
 189  *   bits 17:19 - common access permissions for all ptes in this shadow page
 190  */
 191 union kvm_mmu_page_role {
 192         unsigned word;
 193         struct {
 194                 unsigned level:4;
 195                 unsigned cr4_pae:1;
 196                 unsigned quadrant:2;
 197                 unsigned pad_for_nice_hex_output:6;
 198                 unsigned direct:1;
 199                 unsigned access:3;
 200                 unsigned invalid:1;
 201                 unsigned nxe:1;
 202                 unsigned cr0_wp:1;
 203                 unsigned smep_andnot_wp:1;
 204         };
 205 };
 206
 207 struct kvm_mmu_page {
 208         struct list_head link;
 209         struct hlist_node hash_link;
 210
 211         /*
 212          * The following two entries are used to key the shadow page in the
 213          * hash table.
 214          */
 215         gfn_t gfn;
 216         union kvm_mmu_page_role role;
 217
 218         u64 *spt;
 219         /* hold the gfn of each spte inside spt */
 220         gfn_t *gfns;
 221         bool unsync;
 222         int root_count;          /* Currently serving as active root */
 223         unsigned int unsync_children;
 224         unsigned long parent_ptes;      /* Reverse mapping for parent_pte */
 225
 226         /* The page is obsolete if mmu_valid_gen != kvm->arch.mmu_valid_gen.  */
 227         unsigned long mmu_valid_gen;
 228
 229         DECLARE_BITMAP(unsync_child_bitmap, 512);
 230
 231 #ifdef CONFIG_X86_32
 232         /*
 233          * Used out of the mmu-lock to avoid reading spte values while an
 234          * update is in progress; see the comments in __get_spte_lockless().
 235          */
 236         int clear_spte_count;
 237 #endif
 238
 239         /* Number of writes since the last time traversal visited this page.  */
 240         int write_flooding_count;
 241 };
 242
 243 struct kvm_pio_request {
 244         unsigned long count;
 245         int in;
 246         int port;
 247         int size;
 248 };
 249
 250 /*
 251  * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
 252  * 32-bit).  The kvm_mmu structure abstracts the details of the current mmu
 253  * mode.
 254  */
 255 struct kvm_mmu {
 256         void (*new_cr3)(struct kvm_vcpu *vcpu);
 257         void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long root);
 258         unsigned long (*get_cr3)(struct kvm_vcpu *vcpu);
 259         u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index);
 260         int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err,
 261                           bool prefault);
 262         void (*inject_page_fault)(struct kvm_vcpu *vcpu,
 263                                   struct x86_exception *fault);
 264         void (*free)(struct kvm_vcpu *vcpu);
 265         gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva, u32 access,
 266                             struct x86_exception *exception);
 267         gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access);
 268         int (*sync_page)(struct kvm_vcpu *vcpu,
 269                          struct kvm_mmu_page *sp);
 270         void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva);
 271         void (*update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
 272                            u64 *spte, const void *pte);
 273         hpa_t root_hpa;
 274         int root_level;
 275         int shadow_root_level;
 276         union kvm_mmu_page_role base_role;
 277         bool direct_map;
 278
 279         /*
 280          * Bitmap; bit set = permission fault
 281          * Byte index: page fault error code [4:1]
 282          * Bit index: pte permissions in ACC_* format
 283          */
 284         u8 permissions[16];
 285
 286         u64 *pae_root;
 287         u64 *lm_root;
 288         u64 rsvd_bits_mask[2][4];
 289         u64 bad_mt_xwr;
 290
 291         /*
 292          * Bitmap: bit set = last pte in walk
 293          * index[0:1]: level (zero-based)
 294          * index[2]: pte.ps
 295          */
 296         u8 last_pte_bitmap;
 297
 298         bool nx;
 299
 300         u64 pdptrs[4]; /* pae */
 301 };
 302
 303 enum pmc_type {
 304         KVM_PMC_GP = 0,
 305         KVM_PMC_FIXED,
 306 };
 307
 308 struct kvm_pmc {
 309         enum pmc_type type;
 310         u8 idx;
 311         u64 counter;
 312         u64 eventsel;
 313         struct perf_event *perf_event;
 314         struct kvm_vcpu *vcpu;
 315 };
 316
 317 struct kvm_pmu {
 318         unsigned nr_arch_gp_counters;
 319         unsigned nr_arch_fixed_counters;
 320         unsigned available_event_types;
 321         u64 fixed_ctr_ctrl;
 322         u64 global_ctrl;
 323         u64 global_status;
 324         u64 global_ovf_ctrl;
 325         u64 counter_bitmask[2];
 326         u64 global_ctrl_mask;
 327         u64 reserved_bits;
 328         u8 version;
 329         struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC];
 330         struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED];
 331         struct irq_work irq_work;
 332         u64 reprogram_pmi;
 333 };
 334
 335 struct kvm_vcpu_arch {
 336         /*
 337          * rip and regs accesses must go through
 338          * kvm_{register,rip}_{read,write} functions.
 339          */
 340         unsigned long regs[NR_VCPU_REGS];
 341         u32 regs_avail;
 342         u32 regs_dirty;
 343
 344         unsigned long cr0;
 345         unsigned long cr0_guest_owned_bits;
 346         unsigned long cr2;
 347         unsigned long cr3;
 348         unsigned long cr4;
 349         unsigned long cr4_guest_owned_bits;
 350         unsigned long cr8;
 351         u32 hflags;
 352         u64 efer;
 353         u64 apic_base;
 354         struct kvm_lapic *apic;    /* kernel irqchip context */
 355         unsigned long apic_attention;
 356         int32_t apic_arb_prio;
 357         int mp_state;
 358         u64 ia32_misc_enable_msr;
 359         bool tpr_access_reporting;
 360
 361         /*
 362          * Paging state of the vcpu
 363          *
 364          * If the vcpu runs in guest mode with two level paging this still saves
 365          * the paging mode of the l1 guest. This context is always used to
 366          * handle faults.
 367          */
 368         struct kvm_mmu mmu;
 369
 370         /*
 371          * Paging state of an L2 guest (used for nested npt)
 372          *
 373          * This context will save all necessary information to walk page tables
 374          * of the an L2 guest. This context is only initialized for page table
 375          * walking and not for faulting since we never handle l2 page faults on
 376          * the host.
 377          */
 378         struct kvm_mmu nested_mmu;
 379
 380         /*
 381          * Pointer to the mmu context currently used for
 382          * gva_to_gpa translations.
 383          */
 384         struct kvm_mmu *walk_mmu;
 385
 386         struct kvm_mmu_memory_cache mmu_pte_list_desc_cache;
 387         struct kvm_mmu_memory_cache mmu_page_cache;
 388         struct kvm_mmu_memory_cache mmu_page_header_cache;
 389
 390         struct fpu guest_fpu;
 391         u64 xcr0;
 392
 393         struct kvm_pio_request pio;
 394         void *pio_data;
 395
 396         u8 event_exit_inst_len;
 397
 398         struct kvm_queued_exception {
 399                 bool pending;
 400                 bool has_error_code;
 401                 bool reinject;
 402                 u8 nr;
 403                 u32 error_code;
 404         } exception;
 405
 406         struct kvm_queued_interrupt {
 407                 bool pending;
 408                 bool soft;
 409                 u8 nr;
 410         } interrupt;
 411
 412         int halt_request; /* real mode on Intel only */
 413
 414         int cpuid_nent;
 415         struct kvm_cpuid_entry2 cpuid_entries[KVM_MAX_CPUID_ENTRIES];
 416         /* emulate context */
 417
 418         struct x86_emulate_ctxt emulate_ctxt;
 419         bool emulate_regs_need_sync_to_vcpu;
 420         bool emulate_regs_need_sync_from_vcpu;
 421         int (*complete_userspace_io)(struct kvm_vcpu *vcpu);
 422
 423         gpa_t time;
 424         struct pvclock_vcpu_time_info hv_clock;
 425         unsigned int hw_tsc_khz;
 426         struct gfn_to_hva_cache pv_time;
 427         bool pv_time_enabled;
 428         /* set guest stopped flag in pvclock flags field */
 429         bool pvclock_set_guest_stopped_request;
 430
 431         struct {
 432                 u64 msr_val;
 433                 u64 last_steal;
 434                 u64 accum_steal;
 435                 struct gfn_to_hva_cache stime;
 436                 struct kvm_steal_time steal;
 437         } st;
 438
 439         u64 last_guest_tsc;
 440         u64 last_kernel_ns;
 441         u64 last_host_tsc;
 442         u64 tsc_offset_adjustment;
 443         u64 this_tsc_nsec;
 444         u64 this_tsc_write;
 445         u8  this_tsc_generation;
 446         bool tsc_catchup;
 447         bool tsc_always_catchup;
 448         s8 virtual_tsc_shift;
 449         u32 virtual_tsc_mult;
 450         u32 virtual_tsc_khz;
 451         s64 ia32_tsc_adjust_msr;
 452
 453         atomic_t nmi_queued;  /* unprocessed asynchronous NMIs */
 454         unsigned nmi_pending; /* NMI queued after currently running handler */
 455         bool nmi_injected;    /* Trying to inject an NMI this entry */
 456
 457         struct mtrr_state_type mtrr_state;
 458         u64 pat;
 459
 460         int switch_db_regs;
 461         unsigned long db[KVM_NR_DB_REGS];
 462         unsigned long dr6;
 463         unsigned long dr7;
 464         unsigned long eff_db[KVM_NR_DB_REGS];
 465         unsigned long guest_debug_dr7;
 466
 467         u64 mcg_cap;
 468         u64 mcg_status;
 469         u64 mcg_ctl;
 470         u64 *mce_banks;
 471
 472         /* Cache MMIO info */
 473         u64 mmio_gva;
 474         unsigned access;
 475         gfn_t mmio_gfn;
 476
 477         struct kvm_pmu pmu;
 478
 479         /* used for guest single stepping over the given code position */
 480         unsigned long singlestep_rip;
 481
 482         /* fields used by HYPER-V emulation */
 483         u64 hv_vapic;
 484
 485         cpumask_var_t wbinvd_dirty_mask;
 486
 487         unsigned long last_retry_eip;
 488         unsigned long last_retry_addr;
 489
 490         struct {
 491                 bool halted;
 492                 gfn_t gfns[roundup_pow_of_two(ASYNC_PF_PER_VCPU)];
 493                 struct gfn_to_hva_cache data;
 494                 u64 msr_val;
 495                 u32 id;
 496                 bool send_user_only;
 497         } apf;
 498
 499         /* OSVW MSRs (AMD only) */
 500         struct {
 501                 u64 length;
 502                 u64 status;
 503         } osvw;
 504
 505         struct {
 506                 u64 msr_val;
 507                 struct gfn_to_hva_cache data;
 508         } pv_eoi;
 509
 510         /*
 511          * Indicate whether the access faults on its page table in guest
 512          * which is set when fix page fault and used to detect unhandeable
 513          * instruction.
 514          */
 515         bool write_fault_to_shadow_pgtable;
 516
 517         /* set at EPT violation at this point */
 518         unsigned long exit_qualification;
 519
 520         /* pv related host specific info */
 521         struct {
 522                 bool pv_unhalted;
 523         } pv;
 524 };
 525
 526 struct kvm_lpage_info {
 527         int write_count;
 528 };
 529
 530 struct kvm_arch_memory_slot {
 531         unsigned long *rmap[KVM_NR_PAGE_SIZES];
 532         struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
 533 };
 534
 535 struct kvm_apic_map {
 536         struct rcu_head rcu;
 537         u8 ldr_bits;
 538         /* fields bellow are used to decode ldr values in different modes */
 539         u32 cid_shift, cid_mask, lid_mask;
 540         struct kvm_lapic *phys_map[256];
 541         /* first index is cluster id second is cpu id in a cluster */
 542         struct kvm_lapic *logical_map[16][16];
 543 };
 544
 545 struct kvm_arch {
 546         unsigned int n_used_mmu_pages;
 547         unsigned int n_requested_mmu_pages;
 548         unsigned int n_max_mmu_pages;
 549         unsigned int indirect_shadow_pages;
 550         unsigned long mmu_valid_gen;
 551         struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
 552         /*
 553          * Hash table of struct kvm_mmu_page.
 554          */
 555         struct list_head active_mmu_pages;
 556         struct list_head zapped_obsolete_pages;
 557
 558         struct list_head assigned_dev_head;
 559         struct iommu_domain *iommu_domain;
 560         int iommu_flags;
 561         struct kvm_pic *vpic;
 562         struct kvm_ioapic *vioapic;
 563         struct kvm_pit *vpit;
 564         int vapics_in_nmi_mode;
 565         struct mutex apic_map_lock;
 566         struct kvm_apic_map *apic_map;
 567
 568         unsigned int tss_addr;
 569         struct page *apic_access_page;
 570
 571         gpa_t wall_clock;
 572
 573         struct page *ept_identity_pagetable;
 574         bool ept_identity_pagetable_done;
 575         gpa_t ept_identity_map_addr;
 576
 577         unsigned long irq_sources_bitmap;
 578         s64 kvmclock_offset;
 579         raw_spinlock_t tsc_write_lock;
 580         u64 last_tsc_nsec;
 581         u64 last_tsc_write;
 582         u32 last_tsc_khz;
 583         u64 cur_tsc_nsec;
 584         u64 cur_tsc_write;
 585         u64 cur_tsc_offset;
 586         u8  cur_tsc_generation;
 587         int nr_vcpus_matched_tsc;
 588
 589         spinlock_t pvclock_gtod_sync_lock;
 590         bool use_master_clock;
 591         u64 master_kernel_ns;
 592         cycle_t master_cycle_now;
 593
 594         struct kvm_xen_hvm_config xen_hvm_config;
 595
 596         /* fields used by HYPER-V emulation */
 597         u64 hv_guest_os_id;
 598         u64 hv_hypercall;
 599
 600         #ifdef CONFIG_KVM_MMU_AUDIT
 601         int audit_point;
 602         #endif
 603 };
 604
 605 struct kvm_vm_stat {
 606         u32 mmu_shadow_zapped;
 607         u32 mmu_pte_write;
 608         u32 mmu_pte_updated;
 609         u32 mmu_pde_zapped;
 610         u32 mmu_flooded;
 611         u32 mmu_recycled;
 612         u32 mmu_cache_miss;
 613         u32 mmu_unsync;
 614         u32 remote_tlb_flush;
 615         u32 lpages;
 616 };
 617
 618 struct kvm_vcpu_stat {
 619         u32 pf_fixed;
 620         u32 pf_guest;
 621         u32 tlb_flush;
 622         u32 invlpg;
 623
 624         u32 exits;
 625         u32 io_exits;
 626         u32 mmio_exits;
 627         u32 signal_exits;
 628         u32 irq_window_exits;
 629         u32 nmi_window_exits;
 630         u32 halt_exits;
 631         u32 halt_wakeup;
 632         u32 request_irq_exits;
 633         u32 irq_exits;
 634         u32 host_state_reload;
 635         u32 efer_reload;
 636         u32 fpu_reload;
 637         u32 insn_emulation;
 638         u32 insn_emulation_fail;
 639         u32 hypercalls;
 640         u32 irq_injections;
 641         u32 nmi_injections;
 642 };
 643
 644 struct x86_instruction_info;
 645
 646 struct msr_data {
 647         bool host_initiated;
 648         u32 index;
 649         u64 data;
 650 };
 651
 652 struct kvm_x86_ops {
 653         int (*cpu_has_kvm_support)(void);          /* __init */
 654         int (*disabled_by_bios)(void);             /* __init */
 655         int (*hardware_enable)(void *dummy);
 656         void (*hardware_disable)(void *dummy);
 657         void (*check_processor_compatibility)(void *rtn);
 658         int (*hardware_setup)(void);               /* __init */
 659         void (*hardware_unsetup)(void);            /* __exit */
 660         bool (*cpu_has_accelerated_tpr)(void);
 661         void (*cpuid_update)(struct kvm_vcpu *vcpu);
 662
 663         /* Create, but do not attach this VCPU */
 664         struct kvm_vcpu *(*vcpu_create)(struct kvm *kvm, unsigned id);
 665         void (*vcpu_free)(struct kvm_vcpu *vcpu);
 666         void (*vcpu_reset)(struct kvm_vcpu *vcpu);
 667
 668         void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
 669         void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
 670         void (*vcpu_put)(struct kvm_vcpu *vcpu);
 671
 672         void (*update_db_bp_intercept)(struct kvm_vcpu *vcpu);
 673         int (*get_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata);
 674         int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr);
 675         u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
 676         void (*get_segment)(struct kvm_vcpu *vcpu,
 677                             struct kvm_segment *var, int seg);
 678         int (*get_cpl)(struct kvm_vcpu *vcpu);
 679         void (*set_segment)(struct kvm_vcpu *vcpu,
 680                             struct kvm_segment *var, int seg);
 681         void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
 682         void (*decache_cr0_guest_bits)(struct kvm_vcpu *vcpu);
 683         void (*decache_cr3)(struct kvm_vcpu *vcpu);
 684         void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu);
 685         void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
 686         void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
 687         int (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
 688         void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
 689         void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
 690         void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
 691         void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
 692         void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);
 693         void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value);
 694         void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
 695         unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
 696         void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
 697         void (*fpu_activate)(struct kvm_vcpu *vcpu);
 698         void (*fpu_deactivate)(struct kvm_vcpu *vcpu);
 699
 700         void (*tlb_flush)(struct kvm_vcpu *vcpu);
 701
 702         void (*run)(struct kvm_vcpu *vcpu);
 703         int (*handle_exit)(struct kvm_vcpu *vcpu);
 704         void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
 705         void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
 706         u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask);
 707         void (*patch_hypercall)(struct kvm_vcpu *vcpu,
 708                                 unsigned char *hypercall_addr);
 709         void (*set_irq)(struct kvm_vcpu *vcpu);
 710         void (*set_nmi)(struct kvm_vcpu *vcpu);
 711         void (*queue_exception)(struct kvm_vcpu *vcpu, unsigned nr,
 712                                 bool has_error_code, u32 error_code,
 713                                 bool reinject);
 714         void (*cancel_injection)(struct kvm_vcpu *vcpu);
 715         int (*interrupt_allowed)(struct kvm_vcpu *vcpu);
 716         int (*nmi_allowed)(struct kvm_vcpu *vcpu);
 717         bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
 718         void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
 719         int (*enable_nmi_window)(struct kvm_vcpu *vcpu);
 720         int (*enable_irq_window)(struct kvm_vcpu *vcpu);
 721         void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr);
 722         int (*vm_has_apicv)(struct kvm *kvm);
 723         void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr);
 724         void (*hwapic_isr_update)(struct kvm *kvm, int isr);
 725         void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap);
 726         void (*set_virtual_x2apic_mode)(struct kvm_vcpu *vcpu, bool set);
 727         void (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
 728         void (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
 729         int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
 730         int (*get_tdp_level)(void);
 731         u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
 732         int (*get_lpage_level)(void);
 733         bool (*rdtscp_supported)(void);
 734         bool (*invpcid_supported)(void);
 735         void (*adjust_tsc_offset)(struct kvm_vcpu *vcpu, s64 adjustment, bool host);
 736
 737         void (*set_tdp_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
 738
 739         void (*set_supported_cpuid)(u32 func, struct kvm_cpuid_entry2 *entry);
 740
 741         bool (*has_wbinvd_exit)(void);
 742
 743         void (*set_tsc_khz)(struct kvm_vcpu *vcpu, u32 user_tsc_khz, bool scale);
 744         u64 (*read_tsc_offset)(struct kvm_vcpu *vcpu);
 745         void (*write_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
 746
 747         u64 (*compute_tsc_offset)(struct kvm_vcpu *vcpu, u64 target_tsc);
 748         u64 (*read_l1_tsc)(struct kvm_vcpu *vcpu, u64 host_tsc);
 749
 750         void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2);
 751
 752         int (*check_intercept)(struct kvm_vcpu *vcpu,
 753                                struct x86_instruction_info *info,
 754                                enum x86_intercept_stage stage);
 755         void (*handle_external_intr)(struct kvm_vcpu *vcpu);
 756 };
 757
 758 struct kvm_arch_async_pf {
 759         u32 token;
 760         gfn_t gfn;
 761         unsigned long cr3;
 762         bool direct_map;
 763 };
 764
 765 extern struct kvm_x86_ops *kvm_x86_ops;
 766
 767 static inline void adjust_tsc_offset_guest(struct kvm_vcpu *vcpu,
 768                                            s64 adjustment)
 769 {
 770         kvm_x86_ops->adjust_tsc_offset(vcpu, adjustment, false);
 771 }
 772
 773 static inline void adjust_tsc_offset_host(struct kvm_vcpu *vcpu, s64 adjustment)
 774 {
 775         kvm_x86_ops->adjust_tsc_offset(vcpu, adjustment, true);
 776 }
 777
 778 int kvm_mmu_module_init(void);
 779 void kvm_mmu_module_exit(void);
 780
 781 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
 782 int kvm_mmu_create(struct kvm_vcpu *vcpu);
 783 int kvm_mmu_setup(struct kvm_vcpu *vcpu);
 784 void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
 785                 u64 dirty_mask, u64 nx_mask, u64 x_mask);
 786
 787 int kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
 788 void kvm_mmu_slot_remove_write_access(struct kvm *kvm, int slot);
 789 void kvm_mmu_write_protect_pt_masked(struct kvm *kvm,
 790                                      struct kvm_memory_slot *slot,
 791                                      gfn_t gfn_offset, unsigned long mask);
 792 void kvm_mmu_zap_all(struct kvm *kvm);
 793 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm);
 794 unsigned int kvm_mmu_calculate_mmu_pages(struct kvm *kvm);
 795 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int kvm_nr_mmu_pages);
 796
 797 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3);
 798
 799 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
 800                           const void *val, int bytes);
 801 u8 kvm_get_guest_memory_type(struct kvm_vcpu *vcpu, gfn_t gfn);
 802
 803 extern bool tdp_enabled;
 804
 805 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu);
 806
 807 /* control of guest tsc rate supported? */
 808 extern bool kvm_has_tsc_control;
 809 /* minimum supported tsc_khz for guests */
 810 extern u32  kvm_min_guest_tsc_khz;
 811 /* maximum supported tsc_khz for guests */
 812 extern u32  kvm_max_guest_tsc_khz;
 813
 814 enum emulation_result {
 815         EMULATE_DONE,         /* no further processing */
 816         EMULATE_USER_EXIT,    /* kvm_run ready for userspace exit */
 817         EMULATE_FAIL,         /* can't emulate this instruction */
 818 };
 819
 820 #define EMULTYPE_NO_DECODE          (1 << 0)
 821 #define EMULTYPE_TRAP_UD            (1 << 1)
 822 #define EMULTYPE_SKIP               (1 << 2)
 823 #define EMULTYPE_RETRY              (1 << 3)
 824 #define EMULTYPE_NO_REEXECUTE       (1 << 4)
 825 int x86_emulate_instruction(struct kvm_vcpu *vcpu, unsigned long cr2,
 826                             int emulation_type, void *insn, int insn_len);
 827
 828 static inline int emulate_instruction(struct kvm_vcpu *vcpu,
 829                         int emulation_type)
 830 {
 831         return x86_emulate_instruction(vcpu, 0, emulation_type, NULL, 0);
 832 }
 833
 834 void kvm_enable_efer_bits(u64);
 835 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer);
 836 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 *data);
 837 int kvm_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr);
 838
 839 struct x86_emulate_ctxt;
 840
 841 int kvm_fast_pio_out(struct kvm_vcpu *vcpu, int size, unsigned short port);
 842 void kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
 843 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
 844 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu);
 845
 846 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg);
 847 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg);
 848 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, unsigned int vector);
 849
 850 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
 851                     int reason, bool has_error_code, u32 error_code);
 852
 853 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
 854 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
 855 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
 856 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
 857 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val);
 858 int kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val);
 859 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
 860 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
 861 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
 862 int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr);
 863
 864 int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata);
 865 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr);
 866
 867 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu);
 868 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
 869 bool kvm_rdpmc(struct kvm_vcpu *vcpu);
 870
 871 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
 872 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
 873 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr);
 874 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
 875 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
 876 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
 877                             gfn_t gfn, void *data, int offset, int len,
 878                             u32 access);
 879 void kvm_propagate_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault);
 880 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl);
 881
 882 static inline int __kvm_irq_line_state(unsigned long *irq_state,
 883                                        int irq_source_id, int level)
 884 {
 885         /* Logical OR for level trig interrupt */
 886         if (level)
 887                 __set_bit(irq_source_id, irq_state);
 888         else
 889                 __clear_bit(irq_source_id, irq_state);
 890
 891         return !!(*irq_state);
 892 }
 893
 894 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level);
 895 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id);
 896
 897 void kvm_inject_nmi(struct kvm_vcpu *vcpu);
 898
 899 int fx_init(struct kvm_vcpu *vcpu);
 900
 901 void kvm_mmu_flush_tlb(struct kvm_vcpu *vcpu);
 902 void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
 903                        const u8 *new, int bytes);
 904 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn);
 905 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
 906 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
 907 int kvm_mmu_load(struct kvm_vcpu *vcpu);
 908 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
 909 void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu);
 910 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access);
 911 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
 912                               struct x86_exception *exception);
 913 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
 914                                struct x86_exception *exception);
 915 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
 916                                struct x86_exception *exception);
 917 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
 918                                 struct x86_exception *exception);
 919
 920 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
 921
 922 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva, u32 error_code,
 923                        void *insn, int insn_len);
 924 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
 925
 926 void kvm_enable_tdp(void);
 927 void kvm_disable_tdp(void);
 928
 929 int complete_pio(struct kvm_vcpu *vcpu);
 930 bool kvm_check_iopl(struct kvm_vcpu *vcpu);
 931
 932 static inline gpa_t translate_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access)
 933 {
 934         return gpa;
 935 }
 936
 937 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
 938 {
 939         struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
 940
 941         return (struct kvm_mmu_page *)page_private(page);
 942 }
 943
 944 static inline u16 kvm_read_ldt(void)
 945 {
 946         u16 ldt;
 947         asm("sldt %0" : "=g"(ldt));
 948         return ldt;
 949 }
 950
 951 static inline void kvm_load_ldt(u16 sel)
 952 {
 953         asm("lldt %0" : : "rm"(sel));
 954 }
 955
 956 #ifdef CONFIG_X86_64
 957 static inline unsigned long read_msr(unsigned long msr)
 958 {
 959         u64 value;
 960
 961         rdmsrl(msr, value);
 962         return value;
 963 }
 964 #endif
 965
 966 static inline u32 get_rdx_init_val(void)
 967 {
 968         return 0x600; /* P6 family */
 969 }
 970
 971 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
 972 {
 973         kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
 974 }
 975
 976 #define TSS_IOPB_BASE_OFFSET 0x66
 977 #define TSS_BASE_SIZE 0x68
 978 #define TSS_IOPB_SIZE (65536 / 8)
 979 #define TSS_REDIRECTION_SIZE (256 / 8)
 980 #define RMODE_TSS_SIZE                                                  \
 981         (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
 982
 983 enum {
 984         TASK_SWITCH_CALL = 0,
 985         TASK_SWITCH_IRET = 1,
 986         TASK_SWITCH_JMP = 2,
 987         TASK_SWITCH_GATE = 3,
 988 };
 989
 990 #define HF_GIF_MASK             (1 << 0)
 991 #define HF_HIF_MASK             (1 << 1)
 992 #define HF_VINTR_MASK           (1 << 2)
 993 #define HF_NMI_MASK             (1 << 3)
 994 #define HF_IRET_MASK            (1 << 4)
 995 #define HF_GUEST_MASK           (1 << 5) /* VCPU is in guest-mode */
 996
 997 /*
 998  * Hardware virtualization extension instructions may fault if a
 999  * reboot turns off virtualization while processes are running.
1000  * Trap the fault and ignore the instruction if that happens.
1001  */
1002 asmlinkage void kvm_spurious_fault(void);
1003
1004 #define ____kvm_handle_fault_on_reboot(insn, cleanup_insn)      \
1005         "666: " insn "\n\t" \
1006         "668: \n\t"                           \
1007         ".pushsection .fixup, \"ax\" \n" \
1008         "667: \n\t" \
1009         cleanup_insn "\n\t"                   \
1010         "cmpb $0, kvm_rebooting \n\t"         \
1011         "jne 668b \n\t"                       \
1012         __ASM_SIZE(push) " $666b \n\t"        \
1013         "call kvm_spurious_fault \n\t"        \
1014         ".popsection \n\t" \
1015         _ASM_EXTABLE(666b, 667b)
1016
1017 #define __kvm_handle_fault_on_reboot(insn)              \
1018         ____kvm_handle_fault_on_reboot(insn, "")
1019
1020 #define KVM_ARCH_WANT_MMU_NOTIFIER
1021 int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
1022 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end);
1023 int kvm_age_hva(struct kvm *kvm, unsigned long hva);
1024 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
1025 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
1026 int cpuid_maxphyaddr(struct kvm_vcpu *vcpu);
1027 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v);
1028 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu);
1029 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu);
1030 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
1031 void kvm_vcpu_reset(struct kvm_vcpu *vcpu);
1032
1033 void kvm_define_shared_msr(unsigned index, u32 msr);
1034 void kvm_set_shared_msr(unsigned index, u64 val, u64 mask);
1035
1036 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
1037
1038 void kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
1039                                      struct kvm_async_pf *work);
1040 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
1041                                  struct kvm_async_pf *work);
1042 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
1043                                struct kvm_async_pf *work);
1044 bool kvm_arch_can_inject_async_page_present(struct kvm_vcpu *vcpu);
1045 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1046
1047 void kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err);
1048
1049 int kvm_is_in_guest(void);
1050
1051 void kvm_pmu_init(struct kvm_vcpu *vcpu);
1052 void kvm_pmu_destroy(struct kvm_vcpu *vcpu);
1053 void kvm_pmu_reset(struct kvm_vcpu *vcpu);
1054 void kvm_pmu_cpuid_update(struct kvm_vcpu *vcpu);
1055 bool kvm_pmu_msr(struct kvm_vcpu *vcpu, u32 msr);
1056 int kvm_pmu_get_msr(struct kvm_vcpu *vcpu, u32 msr, u64 *data);
1057 int kvm_pmu_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info);
1058 int kvm_pmu_read_pmc(struct kvm_vcpu *vcpu, unsigned pmc, u64 *data);
1059 void kvm_handle_pmu_event(struct kvm_vcpu *vcpu);
1060 void kvm_deliver_pmi(struct kvm_vcpu *vcpu);
1061
1062 #endif /* _ASM_X86_KVM_HOST_H */