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 10 99 100 #define ASYNC_PF_PER_VCPU 64 101 102 struct kvm_vcpu; 103 struct kvm; 104 struct kvm_async_pf; 105 106 enum kvm_reg { 107 VCPU_REGS_RAX = 0, 108 VCPU_REGS_RCX = 1, 109 VCPU_REGS_RDX = 2, 110 VCPU_REGS_RBX = 3, 111 VCPU_REGS_RSP = 4, 112 VCPU_REGS_RBP = 5, 113 VCPU_REGS_RSI = 6, 114 VCPU_REGS_RDI = 7, 115 #ifdef CONFIG_X86_64 116 VCPU_REGS_R8 = 8, 117 VCPU_REGS_R9 = 9, 118 VCPU_REGS_R10 = 10, 119 VCPU_REGS_R11 = 11, 120 VCPU_REGS_R12 = 12, 121 VCPU_REGS_R13 = 13, 122 VCPU_REGS_R14 = 14, 123 VCPU_REGS_R15 = 15, 124 #endif 125 VCPU_REGS_RIP, 126 NR_VCPU_REGS 127 }; 128 129 enum kvm_reg_ex { 130 VCPU_EXREG_PDPTR = NR_VCPU_REGS, 131 VCPU_EXREG_CR3, 132 VCPU_EXREG_RFLAGS, 133 VCPU_EXREG_SEGMENTS, 134 }; 135 136 enum { 137 VCPU_SREG_ES, 138 VCPU_SREG_CS, 139 VCPU_SREG_SS, 140 VCPU_SREG_DS, 141 VCPU_SREG_FS, 142 VCPU_SREG_GS, 143 VCPU_SREG_TR, 144 VCPU_SREG_LDTR, 145 }; 146 147 #include <asm/kvm_emulate.h> 148 149 #define KVM_NR_MEM_OBJS 40 150 151 #define KVM_NR_DB_REGS 4 152 153 #define DR6_BD (1 << 13) 154 #define DR6_BS (1 << 14) 155 #define DR6_RTM (1 << 16) 156 #define DR6_FIXED_1 0xfffe0ff0 157 #define DR6_INIT 0xffff0ff0 158 #define DR6_VOLATILE 0x0001e00f 159 160 #define DR7_BP_EN_MASK 0x000000ff 161 #define DR7_GE (1 << 9) 162 #define DR7_GD (1 << 13) 163 #define DR7_FIXED_1 0x00000400 164 #define DR7_VOLATILE 0xffff2bff 165 166 /* apic attention bits */ 167 #define KVM_APIC_CHECK_VAPIC 0 168 /* 169 * The following bit is set with PV-EOI, unset on EOI. 170 * We detect PV-EOI changes by guest by comparing 171 * this bit with PV-EOI in guest memory. 172 * See the implementation in apic_update_pv_eoi. 173 */ 174 #define KVM_APIC_PV_EOI_PENDING 1 175 176 /* 177 * We don't want allocation failures within the mmu code, so we preallocate 178 * enough memory for a single page fault in a cache. 179 */ 180 struct kvm_mmu_memory_cache { 181 int nobjs; 182 void *objects[KVM_NR_MEM_OBJS]; 183 }; 184 185 /* 186 * kvm_mmu_page_role, below, is defined as: 187 * 188 * bits 0:3 - total guest paging levels (2-4, or zero for real mode) 189 * bits 4:7 - page table level for this shadow (1-4) 190 * bits 8:9 - page table quadrant for 2-level guests 191 * bit 16 - direct mapping of virtual to physical mapping at gfn 192 * used for real mode and two-dimensional paging 193 * bits 17:19 - common access permissions for all ptes in this shadow page 194 */ 195 union kvm_mmu_page_role { 196 unsigned word; 197 struct { 198 unsigned level:4; 199 unsigned cr4_pae:1; 200 unsigned quadrant:2; 201 unsigned pad_for_nice_hex_output:6; 202 unsigned direct:1; 203 unsigned access:3; 204 unsigned invalid:1; 205 unsigned nxe:1; 206 unsigned cr0_wp:1; 207 unsigned smep_andnot_wp:1; 208 }; 209 }; 210 211 struct kvm_mmu_page { 212 struct list_head link; 213 struct hlist_node hash_link; 214 215 /* 216 * The following two entries are used to key the shadow page in the 217 * hash table. 218 */ 219 gfn_t gfn; 220 union kvm_mmu_page_role role; 221 222 u64 *spt; 223 /* hold the gfn of each spte inside spt */ 224 gfn_t *gfns; 225 bool unsync; 226 int root_count; /* Currently serving as active root */ 227 unsigned int unsync_children; 228 unsigned long parent_ptes; /* Reverse mapping for parent_pte */ 229 230 /* The page is obsolete if mmu_valid_gen != kvm->arch.mmu_valid_gen. */ 231 unsigned long mmu_valid_gen; 232 233 DECLARE_BITMAP(unsync_child_bitmap, 512); 234 235 #ifdef CONFIG_X86_32 236 /* 237 * Used out of the mmu-lock to avoid reading spte values while an 238 * update is in progress; see the comments in __get_spte_lockless(). 239 */ 240 int clear_spte_count; 241 #endif 242 243 /* Number of writes since the last time traversal visited this page. */ 244 int write_flooding_count; 245 }; 246 247 struct kvm_pio_request { 248 unsigned long count; 249 int in; 250 int port; 251 int size; 252 }; 253 254 /* 255 * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level 256 * 32-bit). The kvm_mmu structure abstracts the details of the current mmu 257 * mode. 258 */ 259 struct kvm_mmu { 260 void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long root); 261 unsigned long (*get_cr3)(struct kvm_vcpu *vcpu); 262 u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index); 263 int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err, 264 bool prefault); 265 void (*inject_page_fault)(struct kvm_vcpu *vcpu, 266 struct x86_exception *fault); 267 gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva, u32 access, 268 struct x86_exception *exception); 269 gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access); 270 int (*sync_page)(struct kvm_vcpu *vcpu, 271 struct kvm_mmu_page *sp); 272 void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva); 273 void (*update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp, 274 u64 *spte, const void *pte); 275 hpa_t root_hpa; 276 int root_level; 277 int shadow_root_level; 278 union kvm_mmu_page_role base_role; 279 bool direct_map; 280 281 /* 282 * Bitmap; bit set = permission fault 283 * Byte index: page fault error code [4:1] 284 * Bit index: pte permissions in ACC_* format 285 */ 286 u8 permissions[16]; 287 288 u64 *pae_root; 289 u64 *lm_root; 290 u64 rsvd_bits_mask[2][4]; 291 u64 bad_mt_xwr; 292 293 /* 294 * Bitmap: bit set = last pte in walk 295 * index[0:1]: level (zero-based) 296 * index[2]: pte.ps 297 */ 298 u8 last_pte_bitmap; 299 300 bool nx; 301 302 u64 pdptrs[4]; /* pae */ 303 }; 304 305 enum pmc_type { 306 KVM_PMC_GP = 0, 307 KVM_PMC_FIXED, 308 }; 309 310 struct kvm_pmc { 311 enum pmc_type type; 312 u8 idx; 313 u64 counter; 314 u64 eventsel; 315 struct perf_event *perf_event; 316 struct kvm_vcpu *vcpu; 317 }; 318 319 struct kvm_pmu { 320 unsigned nr_arch_gp_counters; 321 unsigned nr_arch_fixed_counters; 322 unsigned available_event_types; 323 u64 fixed_ctr_ctrl; 324 u64 global_ctrl; 325 u64 global_status; 326 u64 global_ovf_ctrl; 327 u64 counter_bitmask[2]; 328 u64 global_ctrl_mask; 329 u64 reserved_bits; 330 u8 version; 331 struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC]; 332 struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED]; 333 struct irq_work irq_work; 334 u64 reprogram_pmi; 335 }; 336 337 enum { 338 KVM_DEBUGREG_BP_ENABLED = 1, 339 KVM_DEBUGREG_WONT_EXIT = 2, 340 }; 341 342 struct kvm_vcpu_arch { 343 /* 344 * rip and regs accesses must go through 345 * kvm_{register,rip}_{read,write} functions. 346 */ 347 unsigned long regs[NR_VCPU_REGS]; 348 u32 regs_avail; 349 u32 regs_dirty; 350 351 unsigned long cr0; 352 unsigned long cr0_guest_owned_bits; 353 unsigned long cr2; 354 unsigned long cr3; 355 unsigned long cr4; 356 unsigned long cr4_guest_owned_bits; 357 unsigned long cr8; 358 u32 hflags; 359 u64 efer; 360 u64 apic_base; 361 struct kvm_lapic *apic; /* kernel irqchip context */ 362 unsigned long apic_attention; 363 int32_t apic_arb_prio; 364 int mp_state; 365 u64 ia32_misc_enable_msr; 366 bool tpr_access_reporting; 367 368 /* 369 * Paging state of the vcpu 370 * 371 * If the vcpu runs in guest mode with two level paging this still saves 372 * the paging mode of the l1 guest. This context is always used to 373 * handle faults. 374 */ 375 struct kvm_mmu mmu; 376 377 /* 378 * Paging state of an L2 guest (used for nested npt) 379 * 380 * This context will save all necessary information to walk page tables 381 * of the an L2 guest. This context is only initialized for page table 382 * walking and not for faulting since we never handle l2 page faults on 383 * the host. 384 */ 385 struct kvm_mmu nested_mmu; 386 387 /* 388 * Pointer to the mmu context currently used for 389 * gva_to_gpa translations. 390 */ 391 struct kvm_mmu *walk_mmu; 392 393 struct kvm_mmu_memory_cache mmu_pte_list_desc_cache; 394 struct kvm_mmu_memory_cache mmu_page_cache; 395 struct kvm_mmu_memory_cache mmu_page_header_cache; 396 397 struct fpu guest_fpu; 398 u64 xcr0; 399 u64 guest_supported_xcr0; 400 u32 guest_xstate_size; 401 402 struct kvm_pio_request pio; 403 void *pio_data; 404 405 u8 event_exit_inst_len; 406 407 struct kvm_queued_exception { 408 bool pending; 409 bool has_error_code; 410 bool reinject; 411 u8 nr; 412 u32 error_code; 413 } exception; 414 415 struct kvm_queued_interrupt { 416 bool pending; 417 bool soft; 418 u8 nr; 419 } interrupt; 420 421 int halt_request; /* real mode on Intel only */ 422 423 int cpuid_nent; 424 struct kvm_cpuid_entry2 cpuid_entries[KVM_MAX_CPUID_ENTRIES]; 425 /* emulate context */ 426 427 struct x86_emulate_ctxt emulate_ctxt; 428 bool emulate_regs_need_sync_to_vcpu; 429 bool emulate_regs_need_sync_from_vcpu; 430 int (*complete_userspace_io)(struct kvm_vcpu *vcpu); 431 432 gpa_t time; 433 struct pvclock_vcpu_time_info hv_clock; 434 unsigned int hw_tsc_khz; 435 struct gfn_to_hva_cache pv_time; 436 bool pv_time_enabled; 437 /* set guest stopped flag in pvclock flags field */ 438 bool pvclock_set_guest_stopped_request; 439 440 struct { 441 u64 msr_val; 442 u64 last_steal; 443 u64 accum_steal; 444 struct gfn_to_hva_cache stime; 445 struct kvm_steal_time steal; 446 } st; 447 448 u64 last_guest_tsc; 449 u64 last_host_tsc; 450 u64 tsc_offset_adjustment; 451 u64 this_tsc_nsec; 452 u64 this_tsc_write; 453 u64 this_tsc_generation; 454 bool tsc_catchup; 455 bool tsc_always_catchup; 456 s8 virtual_tsc_shift; 457 u32 virtual_tsc_mult; 458 u32 virtual_tsc_khz; 459 s64 ia32_tsc_adjust_msr; 460 461 atomic_t nmi_queued; /* unprocessed asynchronous NMIs */ 462 unsigned nmi_pending; /* NMI queued after currently running handler */ 463 bool nmi_injected; /* Trying to inject an NMI this entry */ 464 465 struct mtrr_state_type mtrr_state; 466 u64 pat; 467 468 unsigned switch_db_regs; 469 unsigned long db[KVM_NR_DB_REGS]; 470 unsigned long dr6; 471 unsigned long dr7; 472 unsigned long eff_db[KVM_NR_DB_REGS]; 473 unsigned long guest_debug_dr7; 474 475 u64 mcg_cap; 476 u64 mcg_status; 477 u64 mcg_ctl; 478 u64 *mce_banks; 479 480 /* Cache MMIO info */ 481 u64 mmio_gva; 482 unsigned access; 483 gfn_t mmio_gfn; 484 485 struct kvm_pmu pmu; 486 487 /* used for guest single stepping over the given code position */ 488 unsigned long singlestep_rip; 489 490 /* fields used by HYPER-V emulation */ 491 u64 hv_vapic; 492 493 cpumask_var_t wbinvd_dirty_mask; 494 495 unsigned long last_retry_eip; 496 unsigned long last_retry_addr; 497 498 struct { 499 bool halted; 500 gfn_t gfns[roundup_pow_of_two(ASYNC_PF_PER_VCPU)]; 501 struct gfn_to_hva_cache data; 502 u64 msr_val; 503 u32 id; 504 bool send_user_only; 505 } apf; 506 507 /* OSVW MSRs (AMD only) */ 508 struct { 509 u64 length; 510 u64 status; 511 } osvw; 512 513 struct { 514 u64 msr_val; 515 struct gfn_to_hva_cache data; 516 } pv_eoi; 517 518 /* 519 * Indicate whether the access faults on its page table in guest 520 * which is set when fix page fault and used to detect unhandeable 521 * instruction. 522 */ 523 bool write_fault_to_shadow_pgtable; 524 525 /* set at EPT violation at this point */ 526 unsigned long exit_qualification; 527 528 /* pv related host specific info */ 529 struct { 530 bool pv_unhalted; 531 } pv; 532 }; 533 534 struct kvm_lpage_info { 535 int write_count; 536 }; 537 538 struct kvm_arch_memory_slot { 539 unsigned long *rmap[KVM_NR_PAGE_SIZES]; 540 struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1]; 541 }; 542 543 struct kvm_apic_map { 544 struct rcu_head rcu; 545 u8 ldr_bits; 546 /* fields bellow are used to decode ldr values in different modes */ 547 u32 cid_shift, cid_mask, lid_mask; 548 struct kvm_lapic *phys_map[256]; 549 /* first index is cluster id second is cpu id in a cluster */ 550 struct kvm_lapic *logical_map[16][16]; 551 }; 552 553 struct kvm_arch { 554 unsigned int n_used_mmu_pages; 555 unsigned int n_requested_mmu_pages; 556 unsigned int n_max_mmu_pages; 557 unsigned int indirect_shadow_pages; 558 unsigned long mmu_valid_gen; 559 struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES]; 560 /* 561 * Hash table of struct kvm_mmu_page. 562 */ 563 struct list_head active_mmu_pages; 564 struct list_head zapped_obsolete_pages; 565 566 struct list_head assigned_dev_head; 567 struct iommu_domain *iommu_domain; 568 bool iommu_noncoherent; 569 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA 570 atomic_t noncoherent_dma_count; 571 struct kvm_pic *vpic; 572 struct kvm_ioapic *vioapic; 573 struct kvm_pit *vpit; 574 int vapics_in_nmi_mode; 575 struct mutex apic_map_lock; 576 struct kvm_apic_map *apic_map; 577 578 unsigned int tss_addr; 579 struct page *apic_access_page; 580 581 gpa_t wall_clock; 582 583 struct page *ept_identity_pagetable; 584 bool ept_identity_pagetable_done; 585 gpa_t ept_identity_map_addr; 586 587 unsigned long irq_sources_bitmap; 588 s64 kvmclock_offset; 589 raw_spinlock_t tsc_write_lock; 590 u64 last_tsc_nsec; 591 u64 last_tsc_write; 592 u32 last_tsc_khz; 593 u64 cur_tsc_nsec; 594 u64 cur_tsc_write; 595 u64 cur_tsc_offset; 596 u64 cur_tsc_generation; 597 int nr_vcpus_matched_tsc; 598 599 spinlock_t pvclock_gtod_sync_lock; 600 bool use_master_clock; 601 u64 master_kernel_ns; 602 cycle_t master_cycle_now; 603 struct delayed_work kvmclock_update_work; 604 struct delayed_work kvmclock_sync_work; 605 606 struct kvm_xen_hvm_config xen_hvm_config; 607 608 /* fields used by HYPER-V emulation */ 609 u64 hv_guest_os_id; 610 u64 hv_hypercall; 611 u64 hv_tsc_page; 612 613 #ifdef CONFIG_KVM_MMU_AUDIT 614 int audit_point; 615 #endif 616 }; 617 618 struct kvm_vm_stat { 619 u32 mmu_shadow_zapped; 620 u32 mmu_pte_write; 621 u32 mmu_pte_updated; 622 u32 mmu_pde_zapped; 623 u32 mmu_flooded; 624 u32 mmu_recycled; 625 u32 mmu_cache_miss; 626 u32 mmu_unsync; 627 u32 remote_tlb_flush; 628 u32 lpages; 629 }; 630 631 struct kvm_vcpu_stat { 632 u32 pf_fixed; 633 u32 pf_guest; 634 u32 tlb_flush; 635 u32 invlpg; 636 637 u32 exits; 638 u32 io_exits; 639 u32 mmio_exits; 640 u32 signal_exits; 641 u32 irq_window_exits; 642 u32 nmi_window_exits; 643 u32 halt_exits; 644 u32 halt_wakeup; 645 u32 request_irq_exits; 646 u32 irq_exits; 647 u32 host_state_reload; 648 u32 efer_reload; 649 u32 fpu_reload; 650 u32 insn_emulation; 651 u32 insn_emulation_fail; 652 u32 hypercalls; 653 u32 irq_injections; 654 u32 nmi_injections; 655 }; 656 657 struct x86_instruction_info; 658 659 struct msr_data { 660 bool host_initiated; 661 u32 index; 662 u64 data; 663 }; 664 665 struct kvm_x86_ops { 666 int (*cpu_has_kvm_support)(void); /* __init */ 667 int (*disabled_by_bios)(void); /* __init */ 668 int (*hardware_enable)(void *dummy); 669 void (*hardware_disable)(void *dummy); 670 void (*check_processor_compatibility)(void *rtn); 671 int (*hardware_setup)(void); /* __init */ 672 void (*hardware_unsetup)(void); /* __exit */ 673 bool (*cpu_has_accelerated_tpr)(void); 674 void (*cpuid_update)(struct kvm_vcpu *vcpu); 675 676 /* Create, but do not attach this VCPU */ 677 struct kvm_vcpu *(*vcpu_create)(struct kvm *kvm, unsigned id); 678 void (*vcpu_free)(struct kvm_vcpu *vcpu); 679 void (*vcpu_reset)(struct kvm_vcpu *vcpu); 680 681 void (*prepare_guest_switch)(struct kvm_vcpu *vcpu); 682 void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu); 683 void (*vcpu_put)(struct kvm_vcpu *vcpu); 684 685 void (*update_db_bp_intercept)(struct kvm_vcpu *vcpu); 686 int (*get_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata); 687 int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr); 688 u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg); 689 void (*get_segment)(struct kvm_vcpu *vcpu, 690 struct kvm_segment *var, int seg); 691 int (*get_cpl)(struct kvm_vcpu *vcpu); 692 void (*set_segment)(struct kvm_vcpu *vcpu, 693 struct kvm_segment *var, int seg); 694 void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l); 695 void (*decache_cr0_guest_bits)(struct kvm_vcpu *vcpu); 696 void (*decache_cr3)(struct kvm_vcpu *vcpu); 697 void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu); 698 void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0); 699 void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3); 700 int (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4); 701 void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer); 702 void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt); 703 void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt); 704 void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt); 705 void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt); 706 u64 (*get_dr6)(struct kvm_vcpu *vcpu); 707 void (*set_dr6)(struct kvm_vcpu *vcpu, unsigned long value); 708 void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu); 709 void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value); 710 void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg); 711 unsigned long (*get_rflags)(struct kvm_vcpu *vcpu); 712 void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags); 713 void (*fpu_activate)(struct kvm_vcpu *vcpu); 714 void (*fpu_deactivate)(struct kvm_vcpu *vcpu); 715 716 void (*tlb_flush)(struct kvm_vcpu *vcpu); 717 718 void (*run)(struct kvm_vcpu *vcpu); 719 int (*handle_exit)(struct kvm_vcpu *vcpu); 720 void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu); 721 void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask); 722 u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu); 723 void (*patch_hypercall)(struct kvm_vcpu *vcpu, 724 unsigned char *hypercall_addr); 725 void (*set_irq)(struct kvm_vcpu *vcpu); 726 void (*set_nmi)(struct kvm_vcpu *vcpu); 727 void (*queue_exception)(struct kvm_vcpu *vcpu, unsigned nr, 728 bool has_error_code, u32 error_code, 729 bool reinject); 730 void (*cancel_injection)(struct kvm_vcpu *vcpu); 731 int (*interrupt_allowed)(struct kvm_vcpu *vcpu); 732 int (*nmi_allowed)(struct kvm_vcpu *vcpu); 733 bool (*get_nmi_mask)(struct kvm_vcpu *vcpu); 734 void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked); 735 void (*enable_nmi_window)(struct kvm_vcpu *vcpu); 736 void (*enable_irq_window)(struct kvm_vcpu *vcpu); 737 void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr); 738 int (*vm_has_apicv)(struct kvm *kvm); 739 void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr); 740 void (*hwapic_isr_update)(struct kvm *kvm, int isr); 741 void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap); 742 void (*set_virtual_x2apic_mode)(struct kvm_vcpu *vcpu, bool set); 743 void (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector); 744 void (*sync_pir_to_irr)(struct kvm_vcpu *vcpu); 745 int (*set_tss_addr)(struct kvm *kvm, unsigned int addr); 746 int (*get_tdp_level)(void); 747 u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio); 748 int (*get_lpage_level)(void); 749 bool (*rdtscp_supported)(void); 750 bool (*invpcid_supported)(void); 751 void (*adjust_tsc_offset)(struct kvm_vcpu *vcpu, s64 adjustment, bool host); 752 753 void (*set_tdp_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3); 754 755 void (*set_supported_cpuid)(u32 func, struct kvm_cpuid_entry2 *entry); 756 757 bool (*has_wbinvd_exit)(void); 758 759 void (*set_tsc_khz)(struct kvm_vcpu *vcpu, u32 user_tsc_khz, bool scale); 760 u64 (*read_tsc_offset)(struct kvm_vcpu *vcpu); 761 void (*write_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset); 762 763 u64 (*compute_tsc_offset)(struct kvm_vcpu *vcpu, u64 target_tsc); 764 u64 (*read_l1_tsc)(struct kvm_vcpu *vcpu, u64 host_tsc); 765 766 void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2); 767 768 int (*check_intercept)(struct kvm_vcpu *vcpu, 769 struct x86_instruction_info *info, 770 enum x86_intercept_stage stage); 771 void (*handle_external_intr)(struct kvm_vcpu *vcpu); 772 bool (*mpx_supported)(void); 773 774 int (*check_nested_events)(struct kvm_vcpu *vcpu, bool external_intr); 775 }; 776 777 struct kvm_arch_async_pf { 778 u32 token; 779 gfn_t gfn; 780 unsigned long cr3; 781 bool direct_map; 782 }; 783 784 extern struct kvm_x86_ops *kvm_x86_ops; 785 786 static inline void adjust_tsc_offset_guest(struct kvm_vcpu *vcpu, 787 s64 adjustment) 788 { 789 kvm_x86_ops->adjust_tsc_offset(vcpu, adjustment, false); 790 } 791 792 static inline void adjust_tsc_offset_host(struct kvm_vcpu *vcpu, s64 adjustment) 793 { 794 kvm_x86_ops->adjust_tsc_offset(vcpu, adjustment, true); 795 } 796 797 int kvm_mmu_module_init(void); 798 void kvm_mmu_module_exit(void); 799 800 void kvm_mmu_destroy(struct kvm_vcpu *vcpu); 801 int kvm_mmu_create(struct kvm_vcpu *vcpu); 802 void kvm_mmu_setup(struct kvm_vcpu *vcpu); 803 void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask, 804 u64 dirty_mask, u64 nx_mask, u64 x_mask); 805 806 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu); 807 void kvm_mmu_slot_remove_write_access(struct kvm *kvm, int slot); 808 void kvm_mmu_write_protect_pt_masked(struct kvm *kvm, 809 struct kvm_memory_slot *slot, 810 gfn_t gfn_offset, unsigned long mask); 811 void kvm_mmu_zap_all(struct kvm *kvm); 812 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm); 813 unsigned int kvm_mmu_calculate_mmu_pages(struct kvm *kvm); 814 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int kvm_nr_mmu_pages); 815 816 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3); 817 818 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa, 819 const void *val, int bytes); 820 u8 kvm_get_guest_memory_type(struct kvm_vcpu *vcpu, gfn_t gfn); 821 822 extern bool tdp_enabled; 823 824 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu); 825 826 /* control of guest tsc rate supported? */ 827 extern bool kvm_has_tsc_control; 828 /* minimum supported tsc_khz for guests */ 829 extern u32 kvm_min_guest_tsc_khz; 830 /* maximum supported tsc_khz for guests */ 831 extern u32 kvm_max_guest_tsc_khz; 832 833 enum emulation_result { 834 EMULATE_DONE, /* no further processing */ 835 EMULATE_USER_EXIT, /* kvm_run ready for userspace exit */ 836 EMULATE_FAIL, /* can't emulate this instruction */ 837 }; 838 839 #define EMULTYPE_NO_DECODE (1 << 0) 840 #define EMULTYPE_TRAP_UD (1 << 1) 841 #define EMULTYPE_SKIP (1 << 2) 842 #define EMULTYPE_RETRY (1 << 3) 843 #define EMULTYPE_NO_REEXECUTE (1 << 4) 844 int x86_emulate_instruction(struct kvm_vcpu *vcpu, unsigned long cr2, 845 int emulation_type, void *insn, int insn_len); 846 847 static inline int emulate_instruction(struct kvm_vcpu *vcpu, 848 int emulation_type) 849 { 850 return x86_emulate_instruction(vcpu, 0, emulation_type, NULL, 0); 851 } 852 853 void kvm_enable_efer_bits(u64); 854 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer); 855 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 *data); 856 int kvm_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr); 857 858 struct x86_emulate_ctxt; 859 860 int kvm_fast_pio_out(struct kvm_vcpu *vcpu, int size, unsigned short port); 861 void kvm_emulate_cpuid(struct kvm_vcpu *vcpu); 862 int kvm_emulate_halt(struct kvm_vcpu *vcpu); 863 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu); 864 865 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg); 866 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg); 867 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, unsigned int vector); 868 869 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index, 870 int reason, bool has_error_code, u32 error_code); 871 872 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0); 873 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3); 874 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4); 875 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8); 876 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val); 877 int kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val); 878 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu); 879 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw); 880 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l); 881 int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr); 882 883 int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata); 884 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr); 885 886 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu); 887 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags); 888 bool kvm_rdpmc(struct kvm_vcpu *vcpu); 889 890 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr); 891 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code); 892 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr); 893 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code); 894 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault); 895 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, 896 gfn_t gfn, void *data, int offset, int len, 897 u32 access); 898 void kvm_propagate_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault); 899 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl); 900 901 static inline int __kvm_irq_line_state(unsigned long *irq_state, 902 int irq_source_id, int level) 903 { 904 /* Logical OR for level trig interrupt */ 905 if (level) 906 __set_bit(irq_source_id, irq_state); 907 else 908 __clear_bit(irq_source_id, irq_state); 909 910 return !!(*irq_state); 911 } 912 913 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level); 914 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id); 915 916 void kvm_inject_nmi(struct kvm_vcpu *vcpu); 917 918 int fx_init(struct kvm_vcpu *vcpu); 919 920 void kvm_mmu_flush_tlb(struct kvm_vcpu *vcpu); 921 void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa, 922 const u8 *new, int bytes); 923 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn); 924 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva); 925 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu); 926 int kvm_mmu_load(struct kvm_vcpu *vcpu); 927 void kvm_mmu_unload(struct kvm_vcpu *vcpu); 928 void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu); 929 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access); 930 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva, 931 struct x86_exception *exception); 932 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva, 933 struct x86_exception *exception); 934 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva, 935 struct x86_exception *exception); 936 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva, 937 struct x86_exception *exception); 938 939 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu); 940 941 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva, u32 error_code, 942 void *insn, int insn_len); 943 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva); 944 void kvm_mmu_new_cr3(struct kvm_vcpu *vcpu); 945 946 void kvm_enable_tdp(void); 947 void kvm_disable_tdp(void); 948 949 static inline gpa_t translate_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access) 950 { 951 return gpa; 952 } 953 954 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page) 955 { 956 struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT); 957 958 return (struct kvm_mmu_page *)page_private(page); 959 } 960 961 static inline u16 kvm_read_ldt(void) 962 { 963 u16 ldt; 964 asm("sldt %0" : "=g"(ldt)); 965 return ldt; 966 } 967 968 static inline void kvm_load_ldt(u16 sel) 969 { 970 asm("lldt %0" : : "rm"(sel)); 971 } 972 973 #ifdef CONFIG_X86_64 974 static inline unsigned long read_msr(unsigned long msr) 975 { 976 u64 value; 977 978 rdmsrl(msr, value); 979 return value; 980 } 981 #endif 982 983 static inline u32 get_rdx_init_val(void) 984 { 985 return 0x600; /* P6 family */ 986 } 987 988 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code) 989 { 990 kvm_queue_exception_e(vcpu, GP_VECTOR, error_code); 991 } 992 993 #define TSS_IOPB_BASE_OFFSET 0x66 994 #define TSS_BASE_SIZE 0x68 995 #define TSS_IOPB_SIZE (65536 / 8) 996 #define TSS_REDIRECTION_SIZE (256 / 8) 997 #define RMODE_TSS_SIZE \ 998 (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1) 999 1000 enum { 1001 TASK_SWITCH_CALL = 0, 1002 TASK_SWITCH_IRET = 1, 1003 TASK_SWITCH_JMP = 2, 1004 TASK_SWITCH_GATE = 3, 1005 }; 1006 1007 #define HF_GIF_MASK (1 << 0) 1008 #define HF_HIF_MASK (1 << 1) 1009 #define HF_VINTR_MASK (1 << 2) 1010 #define HF_NMI_MASK (1 << 3) 1011 #define HF_IRET_MASK (1 << 4) 1012 #define HF_GUEST_MASK (1 << 5) /* VCPU is in guest-mode */ 1013 1014 /* 1015 * Hardware virtualization extension instructions may fault if a 1016 * reboot turns off virtualization while processes are running. 1017 * Trap the fault and ignore the instruction if that happens. 1018 */ 1019 asmlinkage void kvm_spurious_fault(void); 1020 1021 #define ____kvm_handle_fault_on_reboot(insn, cleanup_insn) \ 1022 "666: " insn "\n\t" \ 1023 "668: \n\t" \ 1024 ".pushsection .fixup, \"ax\" \n" \ 1025 "667: \n\t" \ 1026 cleanup_insn "\n\t" \ 1027 "cmpb $0, kvm_rebooting \n\t" \ 1028 "jne 668b \n\t" \ 1029 __ASM_SIZE(push) " $666b \n\t" \ 1030 "call kvm_spurious_fault \n\t" \ 1031 ".popsection \n\t" \ 1032 _ASM_EXTABLE(666b, 667b) 1033 1034 #define __kvm_handle_fault_on_reboot(insn) \ 1035 ____kvm_handle_fault_on_reboot(insn, "") 1036 1037 #define KVM_ARCH_WANT_MMU_NOTIFIER 1038 int kvm_unmap_hva(struct kvm *kvm, unsigned long hva); 1039 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end); 1040 int kvm_age_hva(struct kvm *kvm, unsigned long hva); 1041 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva); 1042 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte); 1043 int cpuid_maxphyaddr(struct kvm_vcpu *vcpu); 1044 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v); 1045 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu); 1046 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu); 1047 int kvm_cpu_get_interrupt(struct kvm_vcpu *v); 1048 void kvm_vcpu_reset(struct kvm_vcpu *vcpu); 1049 1050 void kvm_define_shared_msr(unsigned index, u32 msr); 1051 void kvm_set_shared_msr(unsigned index, u64 val, u64 mask); 1052 1053 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip); 1054 1055 void kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu, 1056 struct kvm_async_pf *work); 1057 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu, 1058 struct kvm_async_pf *work); 1059 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu, 1060 struct kvm_async_pf *work); 1061 bool kvm_arch_can_inject_async_page_present(struct kvm_vcpu *vcpu); 1062 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn); 1063 1064 void kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err); 1065 1066 int kvm_is_in_guest(void); 1067 1068 void kvm_pmu_init(struct kvm_vcpu *vcpu); 1069 void kvm_pmu_destroy(struct kvm_vcpu *vcpu); 1070 void kvm_pmu_reset(struct kvm_vcpu *vcpu); 1071 void kvm_pmu_cpuid_update(struct kvm_vcpu *vcpu); 1072 bool kvm_pmu_msr(struct kvm_vcpu *vcpu, u32 msr); 1073 int kvm_pmu_get_msr(struct kvm_vcpu *vcpu, u32 msr, u64 *data); 1074 int kvm_pmu_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info); 1075 int kvm_pmu_check_pmc(struct kvm_vcpu *vcpu, unsigned pmc); 1076 int kvm_pmu_read_pmc(struct kvm_vcpu *vcpu, unsigned pmc, u64 *data); 1077 void kvm_handle_pmu_event(struct kvm_vcpu *vcpu); 1078 void kvm_deliver_pmi(struct kvm_vcpu *vcpu); 1079 1080 #endif /* _ASM_X86_KVM_HOST_H */ 1081