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_L_MODE_RESERVED_BITS 0xFFFFFF0000000000ULL 54 #define CR4_RESERVED_BITS \ 55 (~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\ 56 | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE \ 57 | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR | X86_CR4_PCIDE \ 58 | X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_FSGSBASE \ 59 | X86_CR4_OSXMMEXCPT | X86_CR4_VMXE | X86_CR4_SMAP)) 60 61 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR) 62 63 64 65 #define INVALID_PAGE (~(hpa_t)0) 66 #define VALID_PAGE(x) ((x) != INVALID_PAGE) 67 68 #define UNMAPPED_GVA (~(gpa_t)0) 69 70 /* KVM Hugepage definitions for x86 */ 71 #define KVM_NR_PAGE_SIZES 3 72 #define KVM_HPAGE_GFN_SHIFT(x) (((x) - 1) * 9) 73 #define KVM_HPAGE_SHIFT(x) (PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x)) 74 #define KVM_HPAGE_SIZE(x) (1UL << KVM_HPAGE_SHIFT(x)) 75 #define KVM_HPAGE_MASK(x) (~(KVM_HPAGE_SIZE(x) - 1)) 76 #define KVM_PAGES_PER_HPAGE(x) (KVM_HPAGE_SIZE(x) / PAGE_SIZE) 77 78 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level) 79 { 80 /* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */ 81 return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) - 82 (base_gfn >> KVM_HPAGE_GFN_SHIFT(level)); 83 } 84 85 #define SELECTOR_TI_MASK (1 << 2) 86 #define SELECTOR_RPL_MASK 0x03 87 88 #define IOPL_SHIFT 12 89 90 #define KVM_PERMILLE_MMU_PAGES 20 91 #define KVM_MIN_ALLOC_MMU_PAGES 64 92 #define KVM_MMU_HASH_SHIFT 10 93 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT) 94 #define KVM_MIN_FREE_MMU_PAGES 5 95 #define KVM_REFILL_PAGES 25 96 #define KVM_MAX_CPUID_ENTRIES 80 97 #define KVM_NR_FIXED_MTRR_REGION 88 98 #define KVM_NR_VAR_MTRR 8 99 100 #define ASYNC_PF_PER_VCPU 64 101 102 enum kvm_reg { 103 VCPU_REGS_RAX = 0, 104 VCPU_REGS_RCX = 1, 105 VCPU_REGS_RDX = 2, 106 VCPU_REGS_RBX = 3, 107 VCPU_REGS_RSP = 4, 108 VCPU_REGS_RBP = 5, 109 VCPU_REGS_RSI = 6, 110 VCPU_REGS_RDI = 7, 111 #ifdef CONFIG_X86_64 112 VCPU_REGS_R8 = 8, 113 VCPU_REGS_R9 = 9, 114 VCPU_REGS_R10 = 10, 115 VCPU_REGS_R11 = 11, 116 VCPU_REGS_R12 = 12, 117 VCPU_REGS_R13 = 13, 118 VCPU_REGS_R14 = 14, 119 VCPU_REGS_R15 = 15, 120 #endif 121 VCPU_REGS_RIP, 122 NR_VCPU_REGS 123 }; 124 125 enum kvm_reg_ex { 126 VCPU_EXREG_PDPTR = NR_VCPU_REGS, 127 VCPU_EXREG_CR3, 128 VCPU_EXREG_RFLAGS, 129 VCPU_EXREG_SEGMENTS, 130 }; 131 132 enum { 133 VCPU_SREG_ES, 134 VCPU_SREG_CS, 135 VCPU_SREG_SS, 136 VCPU_SREG_DS, 137 VCPU_SREG_FS, 138 VCPU_SREG_GS, 139 VCPU_SREG_TR, 140 VCPU_SREG_LDTR, 141 }; 142 143 #include <asm/kvm_emulate.h> 144 145 #define KVM_NR_MEM_OBJS 40 146 147 #define KVM_NR_DB_REGS 4 148 149 #define DR6_BD (1 << 13) 150 #define DR6_BS (1 << 14) 151 #define DR6_RTM (1 << 16) 152 #define DR6_FIXED_1 0xfffe0ff0 153 #define DR6_INIT 0xffff0ff0 154 #define DR6_VOLATILE 0x0001e00f 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 0xffff2bff 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 (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long root); 257 unsigned long (*get_cr3)(struct kvm_vcpu *vcpu); 258 u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index); 259 int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err, 260 bool prefault); 261 void (*inject_page_fault)(struct kvm_vcpu *vcpu, 262 struct x86_exception *fault); 263 gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva, u32 access, 264 struct x86_exception *exception); 265 gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access, 266 struct x86_exception *exception); 267 int (*sync_page)(struct kvm_vcpu *vcpu, 268 struct kvm_mmu_page *sp); 269 void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva); 270 void (*update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp, 271 u64 *spte, const void *pte); 272 hpa_t root_hpa; 273 int root_level; 274 int shadow_root_level; 275 union kvm_mmu_page_role base_role; 276 bool direct_map; 277 278 /* 279 * Bitmap; bit set = permission fault 280 * Byte index: page fault error code [4:1] 281 * Bit index: pte permissions in ACC_* format 282 */ 283 u8 permissions[16]; 284 285 u64 *pae_root; 286 u64 *lm_root; 287 u64 rsvd_bits_mask[2][4]; 288 u64 bad_mt_xwr; 289 290 /* 291 * Bitmap: bit set = last pte in walk 292 * index[0:1]: level (zero-based) 293 * index[2]: pte.ps 294 */ 295 u8 last_pte_bitmap; 296 297 bool nx; 298 299 u64 pdptrs[4]; /* pae */ 300 }; 301 302 enum pmc_type { 303 KVM_PMC_GP = 0, 304 KVM_PMC_FIXED, 305 }; 306 307 struct kvm_pmc { 308 enum pmc_type type; 309 u8 idx; 310 u64 counter; 311 u64 eventsel; 312 struct perf_event *perf_event; 313 struct kvm_vcpu *vcpu; 314 }; 315 316 struct kvm_pmu { 317 unsigned nr_arch_gp_counters; 318 unsigned nr_arch_fixed_counters; 319 unsigned available_event_types; 320 u64 fixed_ctr_ctrl; 321 u64 global_ctrl; 322 u64 global_status; 323 u64 global_ovf_ctrl; 324 u64 counter_bitmask[2]; 325 u64 global_ctrl_mask; 326 u64 reserved_bits; 327 u8 version; 328 struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC]; 329 struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED]; 330 struct irq_work irq_work; 331 u64 reprogram_pmi; 332 }; 333 334 enum { 335 KVM_DEBUGREG_BP_ENABLED = 1, 336 KVM_DEBUGREG_WONT_EXIT = 2, 337 }; 338 339 struct kvm_vcpu_arch { 340 /* 341 * rip and regs accesses must go through 342 * kvm_{register,rip}_{read,write} functions. 343 */ 344 unsigned long regs[NR_VCPU_REGS]; 345 u32 regs_avail; 346 u32 regs_dirty; 347 348 unsigned long cr0; 349 unsigned long cr0_guest_owned_bits; 350 unsigned long cr2; 351 unsigned long cr3; 352 unsigned long cr4; 353 unsigned long cr4_guest_owned_bits; 354 unsigned long cr8; 355 u32 hflags; 356 u64 efer; 357 u64 apic_base; 358 struct kvm_lapic *apic; /* kernel irqchip context */ 359 unsigned long apic_attention; 360 int32_t apic_arb_prio; 361 int mp_state; 362 u64 ia32_misc_enable_msr; 363 bool tpr_access_reporting; 364 365 /* 366 * Paging state of the vcpu 367 * 368 * If the vcpu runs in guest mode with two level paging this still saves 369 * the paging mode of the l1 guest. This context is always used to 370 * handle faults. 371 */ 372 struct kvm_mmu mmu; 373 374 /* 375 * Paging state of an L2 guest (used for nested npt) 376 * 377 * This context will save all necessary information to walk page tables 378 * of the an L2 guest. This context is only initialized for page table 379 * walking and not for faulting since we never handle l2 page faults on 380 * the host. 381 */ 382 struct kvm_mmu nested_mmu; 383 384 /* 385 * Pointer to the mmu context currently used for 386 * gva_to_gpa translations. 387 */ 388 struct kvm_mmu *walk_mmu; 389 390 struct kvm_mmu_memory_cache mmu_pte_list_desc_cache; 391 struct kvm_mmu_memory_cache mmu_page_cache; 392 struct kvm_mmu_memory_cache mmu_page_header_cache; 393 394 struct fpu guest_fpu; 395 u64 xcr0; 396 u64 guest_supported_xcr0; 397 u32 guest_xstate_size; 398 399 struct kvm_pio_request pio; 400 void *pio_data; 401 402 u8 event_exit_inst_len; 403 404 struct kvm_queued_exception { 405 bool pending; 406 bool has_error_code; 407 bool reinject; 408 u8 nr; 409 u32 error_code; 410 } exception; 411 412 struct kvm_queued_interrupt { 413 bool pending; 414 bool soft; 415 u8 nr; 416 } interrupt; 417 418 int halt_request; /* real mode on Intel only */ 419 420 int cpuid_nent; 421 struct kvm_cpuid_entry2 cpuid_entries[KVM_MAX_CPUID_ENTRIES]; 422 /* emulate context */ 423 424 struct x86_emulate_ctxt emulate_ctxt; 425 bool emulate_regs_need_sync_to_vcpu; 426 bool emulate_regs_need_sync_from_vcpu; 427 int (*complete_userspace_io)(struct kvm_vcpu *vcpu); 428 429 gpa_t time; 430 struct pvclock_vcpu_time_info hv_clock; 431 unsigned int hw_tsc_khz; 432 struct gfn_to_hva_cache pv_time; 433 bool pv_time_enabled; 434 /* set guest stopped flag in pvclock flags field */ 435 bool pvclock_set_guest_stopped_request; 436 437 struct { 438 u64 msr_val; 439 u64 last_steal; 440 u64 accum_steal; 441 struct gfn_to_hva_cache stime; 442 struct kvm_steal_time steal; 443 } st; 444 445 u64 last_guest_tsc; 446 u64 last_host_tsc; 447 u64 tsc_offset_adjustment; 448 u64 this_tsc_nsec; 449 u64 this_tsc_write; 450 u64 this_tsc_generation; 451 bool tsc_catchup; 452 bool tsc_always_catchup; 453 s8 virtual_tsc_shift; 454 u32 virtual_tsc_mult; 455 u32 virtual_tsc_khz; 456 s64 ia32_tsc_adjust_msr; 457 458 atomic_t nmi_queued; /* unprocessed asynchronous NMIs */ 459 unsigned nmi_pending; /* NMI queued after currently running handler */ 460 bool nmi_injected; /* Trying to inject an NMI this entry */ 461 462 struct mtrr_state_type mtrr_state; 463 u64 pat; 464 465 unsigned switch_db_regs; 466 unsigned long db[KVM_NR_DB_REGS]; 467 unsigned long dr6; 468 unsigned long dr7; 469 unsigned long eff_db[KVM_NR_DB_REGS]; 470 unsigned long guest_debug_dr7; 471 472 u64 mcg_cap; 473 u64 mcg_status; 474 u64 mcg_ctl; 475 u64 *mce_banks; 476 477 /* Cache MMIO info */ 478 u64 mmio_gva; 479 unsigned access; 480 gfn_t mmio_gfn; 481 u64 mmio_gen; 482 483 struct kvm_pmu pmu; 484 485 /* used for guest single stepping over the given code position */ 486 unsigned long singlestep_rip; 487 488 /* fields used by HYPER-V emulation */ 489 u64 hv_vapic; 490 491 cpumask_var_t wbinvd_dirty_mask; 492 493 unsigned long last_retry_eip; 494 unsigned long last_retry_addr; 495 496 struct { 497 bool halted; 498 gfn_t gfns[roundup_pow_of_two(ASYNC_PF_PER_VCPU)]; 499 struct gfn_to_hva_cache data; 500 u64 msr_val; 501 u32 id; 502 bool send_user_only; 503 } apf; 504 505 /* OSVW MSRs (AMD only) */ 506 struct { 507 u64 length; 508 u64 status; 509 } osvw; 510 511 struct { 512 u64 msr_val; 513 struct gfn_to_hva_cache data; 514 } pv_eoi; 515 516 /* 517 * Indicate whether the access faults on its page table in guest 518 * which is set when fix page fault and used to detect unhandeable 519 * instruction. 520 */ 521 bool write_fault_to_shadow_pgtable; 522 523 /* set at EPT violation at this point */ 524 unsigned long exit_qualification; 525 526 /* pv related host specific info */ 527 struct { 528 bool pv_unhalted; 529 } pv; 530 }; 531 532 struct kvm_lpage_info { 533 int write_count; 534 }; 535 536 struct kvm_arch_memory_slot { 537 unsigned long *rmap[KVM_NR_PAGE_SIZES]; 538 struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1]; 539 }; 540 541 struct kvm_apic_map { 542 struct rcu_head rcu; 543 u8 ldr_bits; 544 /* fields bellow are used to decode ldr values in different modes */ 545 u32 cid_shift, cid_mask, lid_mask; 546 struct kvm_lapic *phys_map[256]; 547 /* first index is cluster id second is cpu id in a cluster */ 548 struct kvm_lapic *logical_map[16][16]; 549 }; 550 551 struct kvm_arch { 552 unsigned int n_used_mmu_pages; 553 unsigned int n_requested_mmu_pages; 554 unsigned int n_max_mmu_pages; 555 unsigned int indirect_shadow_pages; 556 unsigned long mmu_valid_gen; 557 struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES]; 558 /* 559 * Hash table of struct kvm_mmu_page. 560 */ 561 struct list_head active_mmu_pages; 562 struct list_head zapped_obsolete_pages; 563 564 struct list_head assigned_dev_head; 565 struct iommu_domain *iommu_domain; 566 bool iommu_noncoherent; 567 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA 568 atomic_t noncoherent_dma_count; 569 struct kvm_pic *vpic; 570 struct kvm_ioapic *vioapic; 571 struct kvm_pit *vpit; 572 int vapics_in_nmi_mode; 573 struct mutex apic_map_lock; 574 struct kvm_apic_map *apic_map; 575 576 unsigned int tss_addr; 577 bool apic_access_page_done; 578 579 gpa_t wall_clock; 580 581 bool ept_identity_pagetable_done; 582 gpa_t ept_identity_map_addr; 583 584 unsigned long irq_sources_bitmap; 585 s64 kvmclock_offset; 586 raw_spinlock_t tsc_write_lock; 587 u64 last_tsc_nsec; 588 u64 last_tsc_write; 589 u32 last_tsc_khz; 590 u64 cur_tsc_nsec; 591 u64 cur_tsc_write; 592 u64 cur_tsc_offset; 593 u64 cur_tsc_generation; 594 int nr_vcpus_matched_tsc; 595 596 spinlock_t pvclock_gtod_sync_lock; 597 bool use_master_clock; 598 u64 master_kernel_ns; 599 cycle_t master_cycle_now; 600 struct delayed_work kvmclock_update_work; 601 struct delayed_work kvmclock_sync_work; 602 603 struct kvm_xen_hvm_config xen_hvm_config; 604 605 /* fields used by HYPER-V emulation */ 606 u64 hv_guest_os_id; 607 u64 hv_hypercall; 608 u64 hv_tsc_page; 609 610 #ifdef CONFIG_KVM_MMU_AUDIT 611 int audit_point; 612 #endif 613 }; 614 615 struct kvm_vm_stat { 616 u32 mmu_shadow_zapped; 617 u32 mmu_pte_write; 618 u32 mmu_pte_updated; 619 u32 mmu_pde_zapped; 620 u32 mmu_flooded; 621 u32 mmu_recycled; 622 u32 mmu_cache_miss; 623 u32 mmu_unsync; 624 u32 remote_tlb_flush; 625 u32 lpages; 626 }; 627 628 struct kvm_vcpu_stat { 629 u32 pf_fixed; 630 u32 pf_guest; 631 u32 tlb_flush; 632 u32 invlpg; 633 634 u32 exits; 635 u32 io_exits; 636 u32 mmio_exits; 637 u32 signal_exits; 638 u32 irq_window_exits; 639 u32 nmi_window_exits; 640 u32 halt_exits; 641 u32 halt_wakeup; 642 u32 request_irq_exits; 643 u32 irq_exits; 644 u32 host_state_reload; 645 u32 efer_reload; 646 u32 fpu_reload; 647 u32 insn_emulation; 648 u32 insn_emulation_fail; 649 u32 hypercalls; 650 u32 irq_injections; 651 u32 nmi_injections; 652 }; 653 654 struct x86_instruction_info; 655 656 struct msr_data { 657 bool host_initiated; 658 u32 index; 659 u64 data; 660 }; 661 662 struct kvm_x86_ops { 663 int (*cpu_has_kvm_support)(void); /* __init */ 664 int (*disabled_by_bios)(void); /* __init */ 665 int (*hardware_enable)(void); 666 void (*hardware_disable)(void); 667 void (*check_processor_compatibility)(void *rtn); 668 int (*hardware_setup)(void); /* __init */ 669 void (*hardware_unsetup)(void); /* __exit */ 670 bool (*cpu_has_accelerated_tpr)(void); 671 void (*cpuid_update)(struct kvm_vcpu *vcpu); 672 673 /* Create, but do not attach this VCPU */ 674 struct kvm_vcpu *(*vcpu_create)(struct kvm *kvm, unsigned id); 675 void (*vcpu_free)(struct kvm_vcpu *vcpu); 676 void (*vcpu_reset)(struct kvm_vcpu *vcpu); 677 678 void (*prepare_guest_switch)(struct kvm_vcpu *vcpu); 679 void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu); 680 void (*vcpu_put)(struct kvm_vcpu *vcpu); 681 682 void (*update_db_bp_intercept)(struct kvm_vcpu *vcpu); 683 int (*get_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata); 684 int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr); 685 u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg); 686 void (*get_segment)(struct kvm_vcpu *vcpu, 687 struct kvm_segment *var, int seg); 688 int (*get_cpl)(struct kvm_vcpu *vcpu); 689 void (*set_segment)(struct kvm_vcpu *vcpu, 690 struct kvm_segment *var, int seg); 691 void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l); 692 void (*decache_cr0_guest_bits)(struct kvm_vcpu *vcpu); 693 void (*decache_cr3)(struct kvm_vcpu *vcpu); 694 void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu); 695 void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0); 696 void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3); 697 int (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4); 698 void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer); 699 void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt); 700 void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt); 701 void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt); 702 void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt); 703 u64 (*get_dr6)(struct kvm_vcpu *vcpu); 704 void (*set_dr6)(struct kvm_vcpu *vcpu, unsigned long value); 705 void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu); 706 void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value); 707 void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg); 708 unsigned long (*get_rflags)(struct kvm_vcpu *vcpu); 709 void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags); 710 void (*fpu_deactivate)(struct kvm_vcpu *vcpu); 711 712 void (*tlb_flush)(struct kvm_vcpu *vcpu); 713 714 void (*run)(struct kvm_vcpu *vcpu); 715 int (*handle_exit)(struct kvm_vcpu *vcpu); 716 void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu); 717 void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask); 718 u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu); 719 void (*patch_hypercall)(struct kvm_vcpu *vcpu, 720 unsigned char *hypercall_addr); 721 void (*set_irq)(struct kvm_vcpu *vcpu); 722 void (*set_nmi)(struct kvm_vcpu *vcpu); 723 void (*queue_exception)(struct kvm_vcpu *vcpu, unsigned nr, 724 bool has_error_code, u32 error_code, 725 bool reinject); 726 void (*cancel_injection)(struct kvm_vcpu *vcpu); 727 int (*interrupt_allowed)(struct kvm_vcpu *vcpu); 728 int (*nmi_allowed)(struct kvm_vcpu *vcpu); 729 bool (*get_nmi_mask)(struct kvm_vcpu *vcpu); 730 void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked); 731 void (*enable_nmi_window)(struct kvm_vcpu *vcpu); 732 void (*enable_irq_window)(struct kvm_vcpu *vcpu); 733 void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr); 734 int (*vm_has_apicv)(struct kvm *kvm); 735 void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr); 736 void (*hwapic_isr_update)(struct kvm *kvm, int isr); 737 void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap); 738 void (*set_virtual_x2apic_mode)(struct kvm_vcpu *vcpu, bool set); 739 void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu, hpa_t hpa); 740 void (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector); 741 void (*sync_pir_to_irr)(struct kvm_vcpu *vcpu); 742 int (*set_tss_addr)(struct kvm *kvm, unsigned int addr); 743 int (*get_tdp_level)(void); 744 u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio); 745 int (*get_lpage_level)(void); 746 bool (*rdtscp_supported)(void); 747 bool (*invpcid_supported)(void); 748 void (*adjust_tsc_offset)(struct kvm_vcpu *vcpu, s64 adjustment, bool host); 749 750 void (*set_tdp_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3); 751 752 void (*set_supported_cpuid)(u32 func, struct kvm_cpuid_entry2 *entry); 753 754 bool (*has_wbinvd_exit)(void); 755 756 void (*set_tsc_khz)(struct kvm_vcpu *vcpu, u32 user_tsc_khz, bool scale); 757 u64 (*read_tsc_offset)(struct kvm_vcpu *vcpu); 758 void (*write_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset); 759 760 u64 (*compute_tsc_offset)(struct kvm_vcpu *vcpu, u64 target_tsc); 761 u64 (*read_l1_tsc)(struct kvm_vcpu *vcpu, u64 host_tsc); 762 763 void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2); 764 765 int (*check_intercept)(struct kvm_vcpu *vcpu, 766 struct x86_instruction_info *info, 767 enum x86_intercept_stage stage); 768 void (*handle_external_intr)(struct kvm_vcpu *vcpu); 769 bool (*mpx_supported)(void); 770 771 int (*check_nested_events)(struct kvm_vcpu *vcpu, bool external_intr); 772 773 void (*sched_in)(struct kvm_vcpu *kvm, int cpu); 774 }; 775 776 struct kvm_arch_async_pf { 777 u32 token; 778 gfn_t gfn; 779 unsigned long cr3; 780 bool direct_map; 781 }; 782 783 extern struct kvm_x86_ops *kvm_x86_ops; 784 785 static inline void adjust_tsc_offset_guest(struct kvm_vcpu *vcpu, 786 s64 adjustment) 787 { 788 kvm_x86_ops->adjust_tsc_offset(vcpu, adjustment, false); 789 } 790 791 static inline void adjust_tsc_offset_host(struct kvm_vcpu *vcpu, s64 adjustment) 792 { 793 kvm_x86_ops->adjust_tsc_offset(vcpu, adjustment, true); 794 } 795 796 int kvm_mmu_module_init(void); 797 void kvm_mmu_module_exit(void); 798 799 void kvm_mmu_destroy(struct kvm_vcpu *vcpu); 800 int kvm_mmu_create(struct kvm_vcpu *vcpu); 801 void kvm_mmu_setup(struct kvm_vcpu *vcpu); 802 void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask, 803 u64 dirty_mask, u64 nx_mask, u64 x_mask); 804 805 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu); 806 void kvm_mmu_slot_remove_write_access(struct kvm *kvm, int slot); 807 void kvm_mmu_write_protect_pt_masked(struct kvm *kvm, 808 struct kvm_memory_slot *slot, 809 gfn_t gfn_offset, unsigned long mask); 810 void kvm_mmu_zap_all(struct kvm *kvm); 811 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm); 812 unsigned int kvm_mmu_calculate_mmu_pages(struct kvm *kvm); 813 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int kvm_nr_mmu_pages); 814 815 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3); 816 817 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa, 818 const void *val, int bytes); 819 u8 kvm_get_guest_memory_type(struct kvm_vcpu *vcpu, gfn_t gfn); 820 821 extern bool tdp_enabled; 822 823 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu); 824 825 /* control of guest tsc rate supported? */ 826 extern bool kvm_has_tsc_control; 827 /* minimum supported tsc_khz for guests */ 828 extern u32 kvm_min_guest_tsc_khz; 829 /* maximum supported tsc_khz for guests */ 830 extern u32 kvm_max_guest_tsc_khz; 831 832 enum emulation_result { 833 EMULATE_DONE, /* no further processing */ 834 EMULATE_USER_EXIT, /* kvm_run ready for userspace exit */ 835 EMULATE_FAIL, /* can't emulate this instruction */ 836 }; 837 838 #define EMULTYPE_NO_DECODE (1 << 0) 839 #define EMULTYPE_TRAP_UD (1 << 1) 840 #define EMULTYPE_SKIP (1 << 2) 841 #define EMULTYPE_RETRY (1 << 3) 842 #define EMULTYPE_NO_REEXECUTE (1 << 4) 843 int x86_emulate_instruction(struct kvm_vcpu *vcpu, unsigned long cr2, 844 int emulation_type, void *insn, int insn_len); 845 846 static inline int emulate_instruction(struct kvm_vcpu *vcpu, 847 int emulation_type) 848 { 849 return x86_emulate_instruction(vcpu, 0, emulation_type, NULL, 0); 850 } 851 852 void kvm_enable_efer_bits(u64); 853 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer); 854 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 *data); 855 int kvm_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr); 856 857 struct x86_emulate_ctxt; 858 859 int kvm_fast_pio_out(struct kvm_vcpu *vcpu, int size, unsigned short port); 860 void kvm_emulate_cpuid(struct kvm_vcpu *vcpu); 861 int kvm_emulate_halt(struct kvm_vcpu *vcpu); 862 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu); 863 864 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg); 865 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg); 866 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, unsigned int vector); 867 868 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index, 869 int reason, bool has_error_code, u32 error_code); 870 871 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0); 872 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3); 873 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4); 874 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8); 875 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val); 876 int kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val); 877 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu); 878 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw); 879 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l); 880 int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr); 881 882 int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata); 883 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr); 884 885 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu); 886 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags); 887 bool kvm_rdpmc(struct kvm_vcpu *vcpu); 888 889 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr); 890 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code); 891 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr); 892 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code); 893 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault); 894 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, 895 gfn_t gfn, void *data, int offset, int len, 896 u32 access); 897 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl); 898 899 static inline int __kvm_irq_line_state(unsigned long *irq_state, 900 int irq_source_id, int level) 901 { 902 /* Logical OR for level trig interrupt */ 903 if (level) 904 __set_bit(irq_source_id, irq_state); 905 else 906 __clear_bit(irq_source_id, irq_state); 907 908 return !!(*irq_state); 909 } 910 911 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level); 912 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id); 913 914 void kvm_inject_nmi(struct kvm_vcpu *vcpu); 915 916 int fx_init(struct kvm_vcpu *vcpu); 917 918 void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa, 919 const u8 *new, int bytes); 920 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn); 921 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva); 922 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu); 923 int kvm_mmu_load(struct kvm_vcpu *vcpu); 924 void kvm_mmu_unload(struct kvm_vcpu *vcpu); 925 void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu); 926 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access, 927 struct x86_exception *exception); 928 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva, 929 struct x86_exception *exception); 930 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva, 931 struct x86_exception *exception); 932 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva, 933 struct x86_exception *exception); 934 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva, 935 struct x86_exception *exception); 936 937 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu); 938 939 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva, u32 error_code, 940 void *insn, int insn_len); 941 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva); 942 void kvm_mmu_new_cr3(struct kvm_vcpu *vcpu); 943 944 void kvm_enable_tdp(void); 945 void kvm_disable_tdp(void); 946 947 static inline gpa_t translate_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access, 948 struct x86_exception *exception) 949 { 950 return gpa; 951 } 952 953 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page) 954 { 955 struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT); 956 957 return (struct kvm_mmu_page *)page_private(page); 958 } 959 960 static inline u16 kvm_read_ldt(void) 961 { 962 u16 ldt; 963 asm("sldt %0" : "=g"(ldt)); 964 return ldt; 965 } 966 967 static inline void kvm_load_ldt(u16 sel) 968 { 969 asm("lldt %0" : : "rm"(sel)); 970 } 971 972 #ifdef CONFIG_X86_64 973 static inline unsigned long read_msr(unsigned long msr) 974 { 975 u64 value; 976 977 rdmsrl(msr, value); 978 return value; 979 } 980 #endif 981 982 static inline u32 get_rdx_init_val(void) 983 { 984 return 0x600; /* P6 family */ 985 } 986 987 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code) 988 { 989 kvm_queue_exception_e(vcpu, GP_VECTOR, error_code); 990 } 991 992 #define TSS_IOPB_BASE_OFFSET 0x66 993 #define TSS_BASE_SIZE 0x68 994 #define TSS_IOPB_SIZE (65536 / 8) 995 #define TSS_REDIRECTION_SIZE (256 / 8) 996 #define RMODE_TSS_SIZE \ 997 (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1) 998 999 enum { 1000 TASK_SWITCH_CALL = 0, 1001 TASK_SWITCH_IRET = 1, 1002 TASK_SWITCH_JMP = 2, 1003 TASK_SWITCH_GATE = 3, 1004 }; 1005 1006 #define HF_GIF_MASK (1 << 0) 1007 #define HF_HIF_MASK (1 << 1) 1008 #define HF_VINTR_MASK (1 << 2) 1009 #define HF_NMI_MASK (1 << 3) 1010 #define HF_IRET_MASK (1 << 4) 1011 #define HF_GUEST_MASK (1 << 5) /* VCPU is in guest-mode */ 1012 1013 /* 1014 * Hardware virtualization extension instructions may fault if a 1015 * reboot turns off virtualization while processes are running. 1016 * Trap the fault and ignore the instruction if that happens. 1017 */ 1018 asmlinkage void kvm_spurious_fault(void); 1019 1020 #define ____kvm_handle_fault_on_reboot(insn, cleanup_insn) \ 1021 "666: " insn "\n\t" \ 1022 "668: \n\t" \ 1023 ".pushsection .fixup, \"ax\" \n" \ 1024 "667: \n\t" \ 1025 cleanup_insn "\n\t" \ 1026 "cmpb $0, kvm_rebooting \n\t" \ 1027 "jne 668b \n\t" \ 1028 __ASM_SIZE(push) " $666b \n\t" \ 1029 "call kvm_spurious_fault \n\t" \ 1030 ".popsection \n\t" \ 1031 _ASM_EXTABLE(666b, 667b) 1032 1033 #define __kvm_handle_fault_on_reboot(insn) \ 1034 ____kvm_handle_fault_on_reboot(insn, "") 1035 1036 #define KVM_ARCH_WANT_MMU_NOTIFIER 1037 int kvm_unmap_hva(struct kvm *kvm, unsigned long hva); 1038 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end); 1039 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end); 1040 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva); 1041 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte); 1042 int cpuid_maxphyaddr(struct kvm_vcpu *vcpu); 1043 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v); 1044 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu); 1045 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu); 1046 int kvm_cpu_get_interrupt(struct kvm_vcpu *v); 1047 void kvm_vcpu_reset(struct kvm_vcpu *vcpu); 1048 void kvm_vcpu_reload_apic_access_page(struct kvm_vcpu *vcpu); 1049 void kvm_arch_mmu_notifier_invalidate_page(struct kvm *kvm, 1050 unsigned long address); 1051 1052 void kvm_define_shared_msr(unsigned index, u32 msr); 1053 void kvm_set_shared_msr(unsigned index, u64 val, u64 mask); 1054 1055 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip); 1056 1057 void kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu, 1058 struct kvm_async_pf *work); 1059 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu, 1060 struct kvm_async_pf *work); 1061 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu, 1062 struct kvm_async_pf *work); 1063 bool kvm_arch_can_inject_async_page_present(struct kvm_vcpu *vcpu); 1064 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn); 1065 1066 void kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err); 1067 1068 int kvm_is_in_guest(void); 1069 1070 void kvm_pmu_init(struct kvm_vcpu *vcpu); 1071 void kvm_pmu_destroy(struct kvm_vcpu *vcpu); 1072 void kvm_pmu_reset(struct kvm_vcpu *vcpu); 1073 void kvm_pmu_cpuid_update(struct kvm_vcpu *vcpu); 1074 bool kvm_pmu_msr(struct kvm_vcpu *vcpu, u32 msr); 1075 int kvm_pmu_get_msr(struct kvm_vcpu *vcpu, u32 msr, u64 *data); 1076 int kvm_pmu_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info); 1077 int kvm_pmu_check_pmc(struct kvm_vcpu *vcpu, unsigned pmc); 1078 int kvm_pmu_read_pmc(struct kvm_vcpu *vcpu, unsigned pmc, u64 *data); 1079 void kvm_handle_pmu_event(struct kvm_vcpu *vcpu); 1080 void kvm_deliver_pmi(struct kvm_vcpu *vcpu); 1081 1082 #endif /* _ASM_X86_KVM_HOST_H */ 1083