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 #include <linux/irqbypass.h> 28 #include <linux/hyperv.h> 29 30 #include <asm/apic.h> 31 #include <asm/pvclock-abi.h> 32 #include <asm/desc.h> 33 #include <asm/mtrr.h> 34 #include <asm/msr-index.h> 35 #include <asm/asm.h> 36 #include <asm/kvm_page_track.h> 37 38 #define KVM_MAX_VCPUS 288 39 #define KVM_SOFT_MAX_VCPUS 240 40 #define KVM_MAX_VCPU_ID 1023 41 #define KVM_USER_MEM_SLOTS 509 42 /* memory slots that are not exposed to userspace */ 43 #define KVM_PRIVATE_MEM_SLOTS 3 44 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS) 45 46 #define KVM_PIO_PAGE_OFFSET 1 47 #define KVM_COALESCED_MMIO_PAGE_OFFSET 2 48 #define KVM_HALT_POLL_NS_DEFAULT 400000 49 50 #define KVM_IRQCHIP_NUM_PINS KVM_IOAPIC_NUM_PINS 51 52 /* x86-specific vcpu->requests bit members */ 53 #define KVM_REQ_MIGRATE_TIMER 8 54 #define KVM_REQ_REPORT_TPR_ACCESS 9 55 #define KVM_REQ_TRIPLE_FAULT 10 56 #define KVM_REQ_MMU_SYNC 11 57 #define KVM_REQ_CLOCK_UPDATE 12 58 #define KVM_REQ_DEACTIVATE_FPU 13 59 #define KVM_REQ_EVENT 14 60 #define KVM_REQ_APF_HALT 15 61 #define KVM_REQ_STEAL_UPDATE 16 62 #define KVM_REQ_NMI 17 63 #define KVM_REQ_PMU 18 64 #define KVM_REQ_PMI 19 65 #define KVM_REQ_SMI 20 66 #define KVM_REQ_MASTERCLOCK_UPDATE 21 67 #define KVM_REQ_MCLOCK_INPROGRESS 22 68 #define KVM_REQ_SCAN_IOAPIC 23 69 #define KVM_REQ_GLOBAL_CLOCK_UPDATE 24 70 #define KVM_REQ_APIC_PAGE_RELOAD 25 71 #define KVM_REQ_HV_CRASH 26 72 #define KVM_REQ_IOAPIC_EOI_EXIT 27 73 #define KVM_REQ_HV_RESET 28 74 #define KVM_REQ_HV_EXIT 29 75 #define KVM_REQ_HV_STIMER 30 76 77 #define CR0_RESERVED_BITS \ 78 (~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \ 79 | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \ 80 | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG)) 81 82 #define CR3_L_MODE_RESERVED_BITS 0xFFFFFF0000000000ULL 83 #define CR3_PCID_INVD BIT_64(63) 84 #define CR4_RESERVED_BITS \ 85 (~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\ 86 | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE \ 87 | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR | X86_CR4_PCIDE \ 88 | X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_FSGSBASE \ 89 | X86_CR4_OSXMMEXCPT | X86_CR4_VMXE | X86_CR4_SMAP \ 90 | X86_CR4_PKE)) 91 92 #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR) 93 94 95 96 #define INVALID_PAGE (~(hpa_t)0) 97 #define VALID_PAGE(x) ((x) != INVALID_PAGE) 98 99 #define UNMAPPED_GVA (~(gpa_t)0) 100 101 /* KVM Hugepage definitions for x86 */ 102 #define KVM_NR_PAGE_SIZES 3 103 #define KVM_HPAGE_GFN_SHIFT(x) (((x) - 1) * 9) 104 #define KVM_HPAGE_SHIFT(x) (PAGE_SHIFT + KVM_HPAGE_GFN_SHIFT(x)) 105 #define KVM_HPAGE_SIZE(x) (1UL << KVM_HPAGE_SHIFT(x)) 106 #define KVM_HPAGE_MASK(x) (~(KVM_HPAGE_SIZE(x) - 1)) 107 #define KVM_PAGES_PER_HPAGE(x) (KVM_HPAGE_SIZE(x) / PAGE_SIZE) 108 109 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level) 110 { 111 /* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */ 112 return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) - 113 (base_gfn >> KVM_HPAGE_GFN_SHIFT(level)); 114 } 115 116 #define KVM_PERMILLE_MMU_PAGES 20 117 #define KVM_MIN_ALLOC_MMU_PAGES 64 118 #define KVM_MMU_HASH_SHIFT 10 119 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT) 120 #define KVM_MIN_FREE_MMU_PAGES 5 121 #define KVM_REFILL_PAGES 25 122 #define KVM_MAX_CPUID_ENTRIES 80 123 #define KVM_NR_FIXED_MTRR_REGION 88 124 #define KVM_NR_VAR_MTRR 8 125 126 #define ASYNC_PF_PER_VCPU 64 127 128 enum kvm_reg { 129 VCPU_REGS_RAX = 0, 130 VCPU_REGS_RCX = 1, 131 VCPU_REGS_RDX = 2, 132 VCPU_REGS_RBX = 3, 133 VCPU_REGS_RSP = 4, 134 VCPU_REGS_RBP = 5, 135 VCPU_REGS_RSI = 6, 136 VCPU_REGS_RDI = 7, 137 #ifdef CONFIG_X86_64 138 VCPU_REGS_R8 = 8, 139 VCPU_REGS_R9 = 9, 140 VCPU_REGS_R10 = 10, 141 VCPU_REGS_R11 = 11, 142 VCPU_REGS_R12 = 12, 143 VCPU_REGS_R13 = 13, 144 VCPU_REGS_R14 = 14, 145 VCPU_REGS_R15 = 15, 146 #endif 147 VCPU_REGS_RIP, 148 NR_VCPU_REGS 149 }; 150 151 enum kvm_reg_ex { 152 VCPU_EXREG_PDPTR = NR_VCPU_REGS, 153 VCPU_EXREG_CR3, 154 VCPU_EXREG_RFLAGS, 155 VCPU_EXREG_SEGMENTS, 156 }; 157 158 enum { 159 VCPU_SREG_ES, 160 VCPU_SREG_CS, 161 VCPU_SREG_SS, 162 VCPU_SREG_DS, 163 VCPU_SREG_FS, 164 VCPU_SREG_GS, 165 VCPU_SREG_TR, 166 VCPU_SREG_LDTR, 167 }; 168 169 #include <asm/kvm_emulate.h> 170 171 #define KVM_NR_MEM_OBJS 40 172 173 #define KVM_NR_DB_REGS 4 174 175 #define DR6_BD (1 << 13) 176 #define DR6_BS (1 << 14) 177 #define DR6_RTM (1 << 16) 178 #define DR6_FIXED_1 0xfffe0ff0 179 #define DR6_INIT 0xffff0ff0 180 #define DR6_VOLATILE 0x0001e00f 181 182 #define DR7_BP_EN_MASK 0x000000ff 183 #define DR7_GE (1 << 9) 184 #define DR7_GD (1 << 13) 185 #define DR7_FIXED_1 0x00000400 186 #define DR7_VOLATILE 0xffff2bff 187 188 #define PFERR_PRESENT_BIT 0 189 #define PFERR_WRITE_BIT 1 190 #define PFERR_USER_BIT 2 191 #define PFERR_RSVD_BIT 3 192 #define PFERR_FETCH_BIT 4 193 #define PFERR_PK_BIT 5 194 195 #define PFERR_PRESENT_MASK (1U << PFERR_PRESENT_BIT) 196 #define PFERR_WRITE_MASK (1U << PFERR_WRITE_BIT) 197 #define PFERR_USER_MASK (1U << PFERR_USER_BIT) 198 #define PFERR_RSVD_MASK (1U << PFERR_RSVD_BIT) 199 #define PFERR_FETCH_MASK (1U << PFERR_FETCH_BIT) 200 #define PFERR_PK_MASK (1U << PFERR_PK_BIT) 201 202 /* apic attention bits */ 203 #define KVM_APIC_CHECK_VAPIC 0 204 /* 205 * The following bit is set with PV-EOI, unset on EOI. 206 * We detect PV-EOI changes by guest by comparing 207 * this bit with PV-EOI in guest memory. 208 * See the implementation in apic_update_pv_eoi. 209 */ 210 #define KVM_APIC_PV_EOI_PENDING 1 211 212 struct kvm_kernel_irq_routing_entry; 213 214 /* 215 * We don't want allocation failures within the mmu code, so we preallocate 216 * enough memory for a single page fault in a cache. 217 */ 218 struct kvm_mmu_memory_cache { 219 int nobjs; 220 void *objects[KVM_NR_MEM_OBJS]; 221 }; 222 223 /* 224 * the pages used as guest page table on soft mmu are tracked by 225 * kvm_memory_slot.arch.gfn_track which is 16 bits, so the role bits used 226 * by indirect shadow page can not be more than 15 bits. 227 * 228 * Currently, we used 14 bits that are @level, @cr4_pae, @quadrant, @access, 229 * @nxe, @cr0_wp, @smep_andnot_wp and @smap_andnot_wp. 230 */ 231 union kvm_mmu_page_role { 232 unsigned word; 233 struct { 234 unsigned level:4; 235 unsigned cr4_pae:1; 236 unsigned quadrant:2; 237 unsigned direct:1; 238 unsigned access:3; 239 unsigned invalid:1; 240 unsigned nxe:1; 241 unsigned cr0_wp:1; 242 unsigned smep_andnot_wp:1; 243 unsigned smap_andnot_wp:1; 244 unsigned :8; 245 246 /* 247 * This is left at the top of the word so that 248 * kvm_memslots_for_spte_role can extract it with a 249 * simple shift. While there is room, give it a whole 250 * byte so it is also faster to load it from memory. 251 */ 252 unsigned smm:8; 253 }; 254 }; 255 256 struct kvm_rmap_head { 257 unsigned long val; 258 }; 259 260 struct kvm_mmu_page { 261 struct list_head link; 262 struct hlist_node hash_link; 263 264 /* 265 * The following two entries are used to key the shadow page in the 266 * hash table. 267 */ 268 gfn_t gfn; 269 union kvm_mmu_page_role role; 270 271 u64 *spt; 272 /* hold the gfn of each spte inside spt */ 273 gfn_t *gfns; 274 bool unsync; 275 int root_count; /* Currently serving as active root */ 276 unsigned int unsync_children; 277 struct kvm_rmap_head parent_ptes; /* rmap pointers to parent sptes */ 278 279 /* The page is obsolete if mmu_valid_gen != kvm->arch.mmu_valid_gen. */ 280 unsigned long mmu_valid_gen; 281 282 DECLARE_BITMAP(unsync_child_bitmap, 512); 283 284 #ifdef CONFIG_X86_32 285 /* 286 * Used out of the mmu-lock to avoid reading spte values while an 287 * update is in progress; see the comments in __get_spte_lockless(). 288 */ 289 int clear_spte_count; 290 #endif 291 292 /* Number of writes since the last time traversal visited this page. */ 293 atomic_t write_flooding_count; 294 }; 295 296 struct kvm_pio_request { 297 unsigned long count; 298 int in; 299 int port; 300 int size; 301 }; 302 303 struct rsvd_bits_validate { 304 u64 rsvd_bits_mask[2][4]; 305 u64 bad_mt_xwr; 306 }; 307 308 /* 309 * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level 310 * 32-bit). The kvm_mmu structure abstracts the details of the current mmu 311 * mode. 312 */ 313 struct kvm_mmu { 314 void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long root); 315 unsigned long (*get_cr3)(struct kvm_vcpu *vcpu); 316 u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index); 317 int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err, 318 bool prefault); 319 void (*inject_page_fault)(struct kvm_vcpu *vcpu, 320 struct x86_exception *fault); 321 gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva, u32 access, 322 struct x86_exception *exception); 323 gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access, 324 struct x86_exception *exception); 325 int (*sync_page)(struct kvm_vcpu *vcpu, 326 struct kvm_mmu_page *sp); 327 void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva); 328 void (*update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp, 329 u64 *spte, const void *pte); 330 hpa_t root_hpa; 331 int root_level; 332 int shadow_root_level; 333 union kvm_mmu_page_role base_role; 334 bool direct_map; 335 336 /* 337 * Bitmap; bit set = permission fault 338 * Byte index: page fault error code [4:1] 339 * Bit index: pte permissions in ACC_* format 340 */ 341 u8 permissions[16]; 342 343 /* 344 * The pkru_mask indicates if protection key checks are needed. It 345 * consists of 16 domains indexed by page fault error code bits [4:1], 346 * with PFEC.RSVD replaced by ACC_USER_MASK from the page tables. 347 * Each domain has 2 bits which are ANDed with AD and WD from PKRU. 348 */ 349 u32 pkru_mask; 350 351 u64 *pae_root; 352 u64 *lm_root; 353 354 /* 355 * check zero bits on shadow page table entries, these 356 * bits include not only hardware reserved bits but also 357 * the bits spte never used. 358 */ 359 struct rsvd_bits_validate shadow_zero_check; 360 361 struct rsvd_bits_validate guest_rsvd_check; 362 363 /* Can have large pages at levels 2..last_nonleaf_level-1. */ 364 u8 last_nonleaf_level; 365 366 bool nx; 367 368 u64 pdptrs[4]; /* pae */ 369 }; 370 371 enum pmc_type { 372 KVM_PMC_GP = 0, 373 KVM_PMC_FIXED, 374 }; 375 376 struct kvm_pmc { 377 enum pmc_type type; 378 u8 idx; 379 u64 counter; 380 u64 eventsel; 381 struct perf_event *perf_event; 382 struct kvm_vcpu *vcpu; 383 }; 384 385 struct kvm_pmu { 386 unsigned nr_arch_gp_counters; 387 unsigned nr_arch_fixed_counters; 388 unsigned available_event_types; 389 u64 fixed_ctr_ctrl; 390 u64 global_ctrl; 391 u64 global_status; 392 u64 global_ovf_ctrl; 393 u64 counter_bitmask[2]; 394 u64 global_ctrl_mask; 395 u64 reserved_bits; 396 u8 version; 397 struct kvm_pmc gp_counters[INTEL_PMC_MAX_GENERIC]; 398 struct kvm_pmc fixed_counters[INTEL_PMC_MAX_FIXED]; 399 struct irq_work irq_work; 400 u64 reprogram_pmi; 401 }; 402 403 struct kvm_pmu_ops; 404 405 enum { 406 KVM_DEBUGREG_BP_ENABLED = 1, 407 KVM_DEBUGREG_WONT_EXIT = 2, 408 KVM_DEBUGREG_RELOAD = 4, 409 }; 410 411 struct kvm_mtrr_range { 412 u64 base; 413 u64 mask; 414 struct list_head node; 415 }; 416 417 struct kvm_mtrr { 418 struct kvm_mtrr_range var_ranges[KVM_NR_VAR_MTRR]; 419 mtrr_type fixed_ranges[KVM_NR_FIXED_MTRR_REGION]; 420 u64 deftype; 421 422 struct list_head head; 423 }; 424 425 /* Hyper-V SynIC timer */ 426 struct kvm_vcpu_hv_stimer { 427 struct hrtimer timer; 428 int index; 429 u64 config; 430 u64 count; 431 u64 exp_time; 432 struct hv_message msg; 433 bool msg_pending; 434 }; 435 436 /* Hyper-V synthetic interrupt controller (SynIC)*/ 437 struct kvm_vcpu_hv_synic { 438 u64 version; 439 u64 control; 440 u64 msg_page; 441 u64 evt_page; 442 atomic64_t sint[HV_SYNIC_SINT_COUNT]; 443 atomic_t sint_to_gsi[HV_SYNIC_SINT_COUNT]; 444 DECLARE_BITMAP(auto_eoi_bitmap, 256); 445 DECLARE_BITMAP(vec_bitmap, 256); 446 bool active; 447 }; 448 449 /* Hyper-V per vcpu emulation context */ 450 struct kvm_vcpu_hv { 451 u64 hv_vapic; 452 s64 runtime_offset; 453 struct kvm_vcpu_hv_synic synic; 454 struct kvm_hyperv_exit exit; 455 struct kvm_vcpu_hv_stimer stimer[HV_SYNIC_STIMER_COUNT]; 456 DECLARE_BITMAP(stimer_pending_bitmap, HV_SYNIC_STIMER_COUNT); 457 }; 458 459 struct kvm_vcpu_arch { 460 /* 461 * rip and regs accesses must go through 462 * kvm_{register,rip}_{read,write} functions. 463 */ 464 unsigned long regs[NR_VCPU_REGS]; 465 u32 regs_avail; 466 u32 regs_dirty; 467 468 unsigned long cr0; 469 unsigned long cr0_guest_owned_bits; 470 unsigned long cr2; 471 unsigned long cr3; 472 unsigned long cr4; 473 unsigned long cr4_guest_owned_bits; 474 unsigned long cr8; 475 u32 hflags; 476 u64 efer; 477 u64 apic_base; 478 struct kvm_lapic *apic; /* kernel irqchip context */ 479 bool apicv_active; 480 DECLARE_BITMAP(ioapic_handled_vectors, 256); 481 unsigned long apic_attention; 482 int32_t apic_arb_prio; 483 int mp_state; 484 u64 ia32_misc_enable_msr; 485 u64 smbase; 486 bool tpr_access_reporting; 487 u64 ia32_xss; 488 489 /* 490 * Paging state of the vcpu 491 * 492 * If the vcpu runs in guest mode with two level paging this still saves 493 * the paging mode of the l1 guest. This context is always used to 494 * handle faults. 495 */ 496 struct kvm_mmu mmu; 497 498 /* 499 * Paging state of an L2 guest (used for nested npt) 500 * 501 * This context will save all necessary information to walk page tables 502 * of the an L2 guest. This context is only initialized for page table 503 * walking and not for faulting since we never handle l2 page faults on 504 * the host. 505 */ 506 struct kvm_mmu nested_mmu; 507 508 /* 509 * Pointer to the mmu context currently used for 510 * gva_to_gpa translations. 511 */ 512 struct kvm_mmu *walk_mmu; 513 514 struct kvm_mmu_memory_cache mmu_pte_list_desc_cache; 515 struct kvm_mmu_memory_cache mmu_page_cache; 516 struct kvm_mmu_memory_cache mmu_page_header_cache; 517 518 struct fpu guest_fpu; 519 u64 xcr0; 520 u64 guest_supported_xcr0; 521 u32 guest_xstate_size; 522 523 struct kvm_pio_request pio; 524 void *pio_data; 525 526 u8 event_exit_inst_len; 527 528 struct kvm_queued_exception { 529 bool pending; 530 bool has_error_code; 531 bool reinject; 532 u8 nr; 533 u32 error_code; 534 } exception; 535 536 struct kvm_queued_interrupt { 537 bool pending; 538 bool soft; 539 u8 nr; 540 } interrupt; 541 542 int halt_request; /* real mode on Intel only */ 543 544 int cpuid_nent; 545 struct kvm_cpuid_entry2 cpuid_entries[KVM_MAX_CPUID_ENTRIES]; 546 547 int maxphyaddr; 548 549 /* emulate context */ 550 551 struct x86_emulate_ctxt emulate_ctxt; 552 bool emulate_regs_need_sync_to_vcpu; 553 bool emulate_regs_need_sync_from_vcpu; 554 int (*complete_userspace_io)(struct kvm_vcpu *vcpu); 555 556 gpa_t time; 557 struct pvclock_vcpu_time_info hv_clock; 558 unsigned int hw_tsc_khz; 559 struct gfn_to_hva_cache pv_time; 560 bool pv_time_enabled; 561 /* set guest stopped flag in pvclock flags field */ 562 bool pvclock_set_guest_stopped_request; 563 564 struct { 565 u64 msr_val; 566 u64 last_steal; 567 struct gfn_to_hva_cache stime; 568 struct kvm_steal_time steal; 569 } st; 570 571 u64 tsc_offset; 572 u64 last_guest_tsc; 573 u64 last_host_tsc; 574 u64 tsc_offset_adjustment; 575 u64 this_tsc_nsec; 576 u64 this_tsc_write; 577 u64 this_tsc_generation; 578 bool tsc_catchup; 579 bool tsc_always_catchup; 580 s8 virtual_tsc_shift; 581 u32 virtual_tsc_mult; 582 u32 virtual_tsc_khz; 583 s64 ia32_tsc_adjust_msr; 584 u64 tsc_scaling_ratio; 585 586 atomic_t nmi_queued; /* unprocessed asynchronous NMIs */ 587 unsigned nmi_pending; /* NMI queued after currently running handler */ 588 bool nmi_injected; /* Trying to inject an NMI this entry */ 589 bool smi_pending; /* SMI queued after currently running handler */ 590 591 struct kvm_mtrr mtrr_state; 592 u64 pat; 593 594 unsigned switch_db_regs; 595 unsigned long db[KVM_NR_DB_REGS]; 596 unsigned long dr6; 597 unsigned long dr7; 598 unsigned long eff_db[KVM_NR_DB_REGS]; 599 unsigned long guest_debug_dr7; 600 601 u64 mcg_cap; 602 u64 mcg_status; 603 u64 mcg_ctl; 604 u64 mcg_ext_ctl; 605 u64 *mce_banks; 606 607 /* Cache MMIO info */ 608 u64 mmio_gva; 609 unsigned access; 610 gfn_t mmio_gfn; 611 u64 mmio_gen; 612 613 struct kvm_pmu pmu; 614 615 /* used for guest single stepping over the given code position */ 616 unsigned long singlestep_rip; 617 618 struct kvm_vcpu_hv hyperv; 619 620 cpumask_var_t wbinvd_dirty_mask; 621 622 unsigned long last_retry_eip; 623 unsigned long last_retry_addr; 624 625 struct { 626 bool halted; 627 gfn_t gfns[roundup_pow_of_two(ASYNC_PF_PER_VCPU)]; 628 struct gfn_to_hva_cache data; 629 u64 msr_val; 630 u32 id; 631 bool send_user_only; 632 } apf; 633 634 /* OSVW MSRs (AMD only) */ 635 struct { 636 u64 length; 637 u64 status; 638 } osvw; 639 640 struct { 641 u64 msr_val; 642 struct gfn_to_hva_cache data; 643 } pv_eoi; 644 645 /* 646 * Indicate whether the access faults on its page table in guest 647 * which is set when fix page fault and used to detect unhandeable 648 * instruction. 649 */ 650 bool write_fault_to_shadow_pgtable; 651 652 /* set at EPT violation at this point */ 653 unsigned long exit_qualification; 654 655 /* pv related host specific info */ 656 struct { 657 bool pv_unhalted; 658 } pv; 659 660 int pending_ioapic_eoi; 661 int pending_external_vector; 662 }; 663 664 struct kvm_lpage_info { 665 int disallow_lpage; 666 }; 667 668 struct kvm_arch_memory_slot { 669 struct kvm_rmap_head *rmap[KVM_NR_PAGE_SIZES]; 670 struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1]; 671 unsigned short *gfn_track[KVM_PAGE_TRACK_MAX]; 672 }; 673 674 /* 675 * We use as the mode the number of bits allocated in the LDR for the 676 * logical processor ID. It happens that these are all powers of two. 677 * This makes it is very easy to detect cases where the APICs are 678 * configured for multiple modes; in that case, we cannot use the map and 679 * hence cannot use kvm_irq_delivery_to_apic_fast either. 680 */ 681 #define KVM_APIC_MODE_XAPIC_CLUSTER 4 682 #define KVM_APIC_MODE_XAPIC_FLAT 8 683 #define KVM_APIC_MODE_X2APIC 16 684 685 struct kvm_apic_map { 686 struct rcu_head rcu; 687 u8 mode; 688 u32 max_apic_id; 689 union { 690 struct kvm_lapic *xapic_flat_map[8]; 691 struct kvm_lapic *xapic_cluster_map[16][4]; 692 }; 693 struct kvm_lapic *phys_map[]; 694 }; 695 696 /* Hyper-V emulation context */ 697 struct kvm_hv { 698 u64 hv_guest_os_id; 699 u64 hv_hypercall; 700 u64 hv_tsc_page; 701 702 /* Hyper-v based guest crash (NT kernel bugcheck) parameters */ 703 u64 hv_crash_param[HV_X64_MSR_CRASH_PARAMS]; 704 u64 hv_crash_ctl; 705 706 HV_REFERENCE_TSC_PAGE tsc_ref; 707 }; 708 709 struct kvm_arch { 710 unsigned int n_used_mmu_pages; 711 unsigned int n_requested_mmu_pages; 712 unsigned int n_max_mmu_pages; 713 unsigned int indirect_shadow_pages; 714 unsigned long mmu_valid_gen; 715 struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES]; 716 /* 717 * Hash table of struct kvm_mmu_page. 718 */ 719 struct list_head active_mmu_pages; 720 struct list_head zapped_obsolete_pages; 721 struct kvm_page_track_notifier_node mmu_sp_tracker; 722 struct kvm_page_track_notifier_head track_notifier_head; 723 724 struct list_head assigned_dev_head; 725 struct iommu_domain *iommu_domain; 726 bool iommu_noncoherent; 727 #define __KVM_HAVE_ARCH_NONCOHERENT_DMA 728 atomic_t noncoherent_dma_count; 729 #define __KVM_HAVE_ARCH_ASSIGNED_DEVICE 730 atomic_t assigned_device_count; 731 struct kvm_pic *vpic; 732 struct kvm_ioapic *vioapic; 733 struct kvm_pit *vpit; 734 atomic_t vapics_in_nmi_mode; 735 struct mutex apic_map_lock; 736 struct kvm_apic_map *apic_map; 737 738 unsigned int tss_addr; 739 bool apic_access_page_done; 740 741 gpa_t wall_clock; 742 743 bool ept_identity_pagetable_done; 744 gpa_t ept_identity_map_addr; 745 746 unsigned long irq_sources_bitmap; 747 s64 kvmclock_offset; 748 raw_spinlock_t tsc_write_lock; 749 u64 last_tsc_nsec; 750 u64 last_tsc_write; 751 u32 last_tsc_khz; 752 u64 cur_tsc_nsec; 753 u64 cur_tsc_write; 754 u64 cur_tsc_offset; 755 u64 cur_tsc_generation; 756 int nr_vcpus_matched_tsc; 757 758 spinlock_t pvclock_gtod_sync_lock; 759 bool use_master_clock; 760 u64 master_kernel_ns; 761 cycle_t master_cycle_now; 762 struct delayed_work kvmclock_update_work; 763 struct delayed_work kvmclock_sync_work; 764 765 struct kvm_xen_hvm_config xen_hvm_config; 766 767 /* reads protected by irq_srcu, writes by irq_lock */ 768 struct hlist_head mask_notifier_list; 769 770 struct kvm_hv hyperv; 771 772 #ifdef CONFIG_KVM_MMU_AUDIT 773 int audit_point; 774 #endif 775 776 bool boot_vcpu_runs_old_kvmclock; 777 u32 bsp_vcpu_id; 778 779 u64 disabled_quirks; 780 781 bool irqchip_split; 782 u8 nr_reserved_ioapic_pins; 783 784 bool disabled_lapic_found; 785 786 /* Struct members for AVIC */ 787 u32 avic_vm_id; 788 u32 ldr_mode; 789 struct page *avic_logical_id_table_page; 790 struct page *avic_physical_id_table_page; 791 struct hlist_node hnode; 792 793 bool x2apic_format; 794 bool x2apic_broadcast_quirk_disabled; 795 }; 796 797 struct kvm_vm_stat { 798 ulong mmu_shadow_zapped; 799 ulong mmu_pte_write; 800 ulong mmu_pte_updated; 801 ulong mmu_pde_zapped; 802 ulong mmu_flooded; 803 ulong mmu_recycled; 804 ulong mmu_cache_miss; 805 ulong mmu_unsync; 806 ulong remote_tlb_flush; 807 ulong lpages; 808 }; 809 810 struct kvm_vcpu_stat { 811 u64 pf_fixed; 812 u64 pf_guest; 813 u64 tlb_flush; 814 u64 invlpg; 815 816 u64 exits; 817 u64 io_exits; 818 u64 mmio_exits; 819 u64 signal_exits; 820 u64 irq_window_exits; 821 u64 nmi_window_exits; 822 u64 halt_exits; 823 u64 halt_successful_poll; 824 u64 halt_attempted_poll; 825 u64 halt_poll_invalid; 826 u64 halt_wakeup; 827 u64 request_irq_exits; 828 u64 irq_exits; 829 u64 host_state_reload; 830 u64 efer_reload; 831 u64 fpu_reload; 832 u64 insn_emulation; 833 u64 insn_emulation_fail; 834 u64 hypercalls; 835 u64 irq_injections; 836 u64 nmi_injections; 837 }; 838 839 struct x86_instruction_info; 840 841 struct msr_data { 842 bool host_initiated; 843 u32 index; 844 u64 data; 845 }; 846 847 struct kvm_lapic_irq { 848 u32 vector; 849 u16 delivery_mode; 850 u16 dest_mode; 851 bool level; 852 u16 trig_mode; 853 u32 shorthand; 854 u32 dest_id; 855 bool msi_redir_hint; 856 }; 857 858 struct kvm_x86_ops { 859 int (*cpu_has_kvm_support)(void); /* __init */ 860 int (*disabled_by_bios)(void); /* __init */ 861 int (*hardware_enable)(void); 862 void (*hardware_disable)(void); 863 void (*check_processor_compatibility)(void *rtn); 864 int (*hardware_setup)(void); /* __init */ 865 void (*hardware_unsetup)(void); /* __exit */ 866 bool (*cpu_has_accelerated_tpr)(void); 867 bool (*cpu_has_high_real_mode_segbase)(void); 868 void (*cpuid_update)(struct kvm_vcpu *vcpu); 869 870 int (*vm_init)(struct kvm *kvm); 871 void (*vm_destroy)(struct kvm *kvm); 872 873 /* Create, but do not attach this VCPU */ 874 struct kvm_vcpu *(*vcpu_create)(struct kvm *kvm, unsigned id); 875 void (*vcpu_free)(struct kvm_vcpu *vcpu); 876 void (*vcpu_reset)(struct kvm_vcpu *vcpu, bool init_event); 877 878 void (*prepare_guest_switch)(struct kvm_vcpu *vcpu); 879 void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu); 880 void (*vcpu_put)(struct kvm_vcpu *vcpu); 881 882 void (*update_bp_intercept)(struct kvm_vcpu *vcpu); 883 int (*get_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr); 884 int (*set_msr)(struct kvm_vcpu *vcpu, struct msr_data *msr); 885 u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg); 886 void (*get_segment)(struct kvm_vcpu *vcpu, 887 struct kvm_segment *var, int seg); 888 int (*get_cpl)(struct kvm_vcpu *vcpu); 889 void (*set_segment)(struct kvm_vcpu *vcpu, 890 struct kvm_segment *var, int seg); 891 void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l); 892 void (*decache_cr0_guest_bits)(struct kvm_vcpu *vcpu); 893 void (*decache_cr3)(struct kvm_vcpu *vcpu); 894 void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu); 895 void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0); 896 void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3); 897 int (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4); 898 void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer); 899 void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt); 900 void (*set_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt); 901 void (*get_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt); 902 void (*set_gdt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt); 903 u64 (*get_dr6)(struct kvm_vcpu *vcpu); 904 void (*set_dr6)(struct kvm_vcpu *vcpu, unsigned long value); 905 void (*sync_dirty_debug_regs)(struct kvm_vcpu *vcpu); 906 void (*set_dr7)(struct kvm_vcpu *vcpu, unsigned long value); 907 void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg); 908 unsigned long (*get_rflags)(struct kvm_vcpu *vcpu); 909 void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags); 910 u32 (*get_pkru)(struct kvm_vcpu *vcpu); 911 void (*fpu_activate)(struct kvm_vcpu *vcpu); 912 void (*fpu_deactivate)(struct kvm_vcpu *vcpu); 913 914 void (*tlb_flush)(struct kvm_vcpu *vcpu); 915 916 void (*run)(struct kvm_vcpu *vcpu); 917 int (*handle_exit)(struct kvm_vcpu *vcpu); 918 void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu); 919 void (*set_interrupt_shadow)(struct kvm_vcpu *vcpu, int mask); 920 u32 (*get_interrupt_shadow)(struct kvm_vcpu *vcpu); 921 void (*patch_hypercall)(struct kvm_vcpu *vcpu, 922 unsigned char *hypercall_addr); 923 void (*set_irq)(struct kvm_vcpu *vcpu); 924 void (*set_nmi)(struct kvm_vcpu *vcpu); 925 void (*queue_exception)(struct kvm_vcpu *vcpu, unsigned nr, 926 bool has_error_code, u32 error_code, 927 bool reinject); 928 void (*cancel_injection)(struct kvm_vcpu *vcpu); 929 int (*interrupt_allowed)(struct kvm_vcpu *vcpu); 930 int (*nmi_allowed)(struct kvm_vcpu *vcpu); 931 bool (*get_nmi_mask)(struct kvm_vcpu *vcpu); 932 void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked); 933 void (*enable_nmi_window)(struct kvm_vcpu *vcpu); 934 void (*enable_irq_window)(struct kvm_vcpu *vcpu); 935 void (*update_cr8_intercept)(struct kvm_vcpu *vcpu, int tpr, int irr); 936 bool (*get_enable_apicv)(void); 937 void (*refresh_apicv_exec_ctrl)(struct kvm_vcpu *vcpu); 938 void (*hwapic_irr_update)(struct kvm_vcpu *vcpu, int max_irr); 939 void (*hwapic_isr_update)(struct kvm_vcpu *vcpu, int isr); 940 void (*load_eoi_exitmap)(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap); 941 void (*set_virtual_x2apic_mode)(struct kvm_vcpu *vcpu, bool set); 942 void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu, hpa_t hpa); 943 void (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector); 944 void (*sync_pir_to_irr)(struct kvm_vcpu *vcpu); 945 int (*set_tss_addr)(struct kvm *kvm, unsigned int addr); 946 int (*get_tdp_level)(void); 947 u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio); 948 int (*get_lpage_level)(void); 949 bool (*rdtscp_supported)(void); 950 bool (*invpcid_supported)(void); 951 void (*adjust_tsc_offset_guest)(struct kvm_vcpu *vcpu, s64 adjustment); 952 953 void (*set_tdp_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3); 954 955 void (*set_supported_cpuid)(u32 func, struct kvm_cpuid_entry2 *entry); 956 957 bool (*has_wbinvd_exit)(void); 958 959 void (*write_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset); 960 961 u64 (*read_l1_tsc)(struct kvm_vcpu *vcpu, u64 host_tsc); 962 963 void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2); 964 965 int (*check_intercept)(struct kvm_vcpu *vcpu, 966 struct x86_instruction_info *info, 967 enum x86_intercept_stage stage); 968 void (*handle_external_intr)(struct kvm_vcpu *vcpu); 969 bool (*mpx_supported)(void); 970 bool (*xsaves_supported)(void); 971 972 int (*check_nested_events)(struct kvm_vcpu *vcpu, bool external_intr); 973 974 void (*sched_in)(struct kvm_vcpu *kvm, int cpu); 975 976 /* 977 * Arch-specific dirty logging hooks. These hooks are only supposed to 978 * be valid if the specific arch has hardware-accelerated dirty logging 979 * mechanism. Currently only for PML on VMX. 980 * 981 * - slot_enable_log_dirty: 982 * called when enabling log dirty mode for the slot. 983 * - slot_disable_log_dirty: 984 * called when disabling log dirty mode for the slot. 985 * also called when slot is created with log dirty disabled. 986 * - flush_log_dirty: 987 * called before reporting dirty_bitmap to userspace. 988 * - enable_log_dirty_pt_masked: 989 * called when reenabling log dirty for the GFNs in the mask after 990 * corresponding bits are cleared in slot->dirty_bitmap. 991 */ 992 void (*slot_enable_log_dirty)(struct kvm *kvm, 993 struct kvm_memory_slot *slot); 994 void (*slot_disable_log_dirty)(struct kvm *kvm, 995 struct kvm_memory_slot *slot); 996 void (*flush_log_dirty)(struct kvm *kvm); 997 void (*enable_log_dirty_pt_masked)(struct kvm *kvm, 998 struct kvm_memory_slot *slot, 999 gfn_t offset, unsigned long mask); 1000 /* pmu operations of sub-arch */ 1001 const struct kvm_pmu_ops *pmu_ops; 1002 1003 /* 1004 * Architecture specific hooks for vCPU blocking due to 1005 * HLT instruction. 1006 * Returns for .pre_block(): 1007 * - 0 means continue to block the vCPU. 1008 * - 1 means we cannot block the vCPU since some event 1009 * happens during this period, such as, 'ON' bit in 1010 * posted-interrupts descriptor is set. 1011 */ 1012 int (*pre_block)(struct kvm_vcpu *vcpu); 1013 void (*post_block)(struct kvm_vcpu *vcpu); 1014 1015 void (*vcpu_blocking)(struct kvm_vcpu *vcpu); 1016 void (*vcpu_unblocking)(struct kvm_vcpu *vcpu); 1017 1018 int (*update_pi_irte)(struct kvm *kvm, unsigned int host_irq, 1019 uint32_t guest_irq, bool set); 1020 void (*apicv_post_state_restore)(struct kvm_vcpu *vcpu); 1021 1022 int (*set_hv_timer)(struct kvm_vcpu *vcpu, u64 guest_deadline_tsc); 1023 void (*cancel_hv_timer)(struct kvm_vcpu *vcpu); 1024 1025 void (*setup_mce)(struct kvm_vcpu *vcpu); 1026 }; 1027 1028 struct kvm_arch_async_pf { 1029 u32 token; 1030 gfn_t gfn; 1031 unsigned long cr3; 1032 bool direct_map; 1033 }; 1034 1035 extern struct kvm_x86_ops *kvm_x86_ops; 1036 1037 int kvm_mmu_module_init(void); 1038 void kvm_mmu_module_exit(void); 1039 1040 void kvm_mmu_destroy(struct kvm_vcpu *vcpu); 1041 int kvm_mmu_create(struct kvm_vcpu *vcpu); 1042 void kvm_mmu_setup(struct kvm_vcpu *vcpu); 1043 void kvm_mmu_init_vm(struct kvm *kvm); 1044 void kvm_mmu_uninit_vm(struct kvm *kvm); 1045 void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask, 1046 u64 dirty_mask, u64 nx_mask, u64 x_mask, u64 p_mask); 1047 1048 void kvm_mmu_reset_context(struct kvm_vcpu *vcpu); 1049 void kvm_mmu_slot_remove_write_access(struct kvm *kvm, 1050 struct kvm_memory_slot *memslot); 1051 void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm, 1052 const struct kvm_memory_slot *memslot); 1053 void kvm_mmu_slot_leaf_clear_dirty(struct kvm *kvm, 1054 struct kvm_memory_slot *memslot); 1055 void kvm_mmu_slot_largepage_remove_write_access(struct kvm *kvm, 1056 struct kvm_memory_slot *memslot); 1057 void kvm_mmu_slot_set_dirty(struct kvm *kvm, 1058 struct kvm_memory_slot *memslot); 1059 void kvm_mmu_clear_dirty_pt_masked(struct kvm *kvm, 1060 struct kvm_memory_slot *slot, 1061 gfn_t gfn_offset, unsigned long mask); 1062 void kvm_mmu_zap_all(struct kvm *kvm); 1063 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, struct kvm_memslots *slots); 1064 unsigned int kvm_mmu_calculate_mmu_pages(struct kvm *kvm); 1065 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int kvm_nr_mmu_pages); 1066 1067 int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3); 1068 1069 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa, 1070 const void *val, int bytes); 1071 1072 struct kvm_irq_mask_notifier { 1073 void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked); 1074 int irq; 1075 struct hlist_node link; 1076 }; 1077 1078 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq, 1079 struct kvm_irq_mask_notifier *kimn); 1080 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq, 1081 struct kvm_irq_mask_notifier *kimn); 1082 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin, 1083 bool mask); 1084 1085 extern bool tdp_enabled; 1086 1087 u64 vcpu_tsc_khz(struct kvm_vcpu *vcpu); 1088 1089 /* control of guest tsc rate supported? */ 1090 extern bool kvm_has_tsc_control; 1091 /* maximum supported tsc_khz for guests */ 1092 extern u32 kvm_max_guest_tsc_khz; 1093 /* number of bits of the fractional part of the TSC scaling ratio */ 1094 extern u8 kvm_tsc_scaling_ratio_frac_bits; 1095 /* maximum allowed value of TSC scaling ratio */ 1096 extern u64 kvm_max_tsc_scaling_ratio; 1097 /* 1ull << kvm_tsc_scaling_ratio_frac_bits */ 1098 extern u64 kvm_default_tsc_scaling_ratio; 1099 1100 extern u64 kvm_mce_cap_supported; 1101 1102 enum emulation_result { 1103 EMULATE_DONE, /* no further processing */ 1104 EMULATE_USER_EXIT, /* kvm_run ready for userspace exit */ 1105 EMULATE_FAIL, /* can't emulate this instruction */ 1106 }; 1107 1108 #define EMULTYPE_NO_DECODE (1 << 0) 1109 #define EMULTYPE_TRAP_UD (1 << 1) 1110 #define EMULTYPE_SKIP (1 << 2) 1111 #define EMULTYPE_RETRY (1 << 3) 1112 #define EMULTYPE_NO_REEXECUTE (1 << 4) 1113 int x86_emulate_instruction(struct kvm_vcpu *vcpu, unsigned long cr2, 1114 int emulation_type, void *insn, int insn_len); 1115 1116 static inline int emulate_instruction(struct kvm_vcpu *vcpu, 1117 int emulation_type) 1118 { 1119 return x86_emulate_instruction(vcpu, 0, emulation_type, NULL, 0); 1120 } 1121 1122 void kvm_enable_efer_bits(u64); 1123 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer); 1124 int kvm_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr); 1125 int kvm_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr); 1126 1127 struct x86_emulate_ctxt; 1128 1129 int kvm_fast_pio_out(struct kvm_vcpu *vcpu, int size, unsigned short port); 1130 void kvm_emulate_cpuid(struct kvm_vcpu *vcpu); 1131 int kvm_emulate_halt(struct kvm_vcpu *vcpu); 1132 int kvm_vcpu_halt(struct kvm_vcpu *vcpu); 1133 int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu); 1134 1135 void kvm_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg); 1136 int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int seg); 1137 void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector); 1138 1139 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index, 1140 int reason, bool has_error_code, u32 error_code); 1141 1142 int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0); 1143 int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3); 1144 int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4); 1145 int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8); 1146 int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val); 1147 int kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val); 1148 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu); 1149 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw); 1150 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l); 1151 int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr); 1152 1153 int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr); 1154 int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr); 1155 1156 unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu); 1157 void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags); 1158 bool kvm_rdpmc(struct kvm_vcpu *vcpu); 1159 1160 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr); 1161 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code); 1162 void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr); 1163 void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code); 1164 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault); 1165 int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, 1166 gfn_t gfn, void *data, int offset, int len, 1167 u32 access); 1168 bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl); 1169 bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr); 1170 1171 static inline int __kvm_irq_line_state(unsigned long *irq_state, 1172 int irq_source_id, int level) 1173 { 1174 /* Logical OR for level trig interrupt */ 1175 if (level) 1176 __set_bit(irq_source_id, irq_state); 1177 else 1178 __clear_bit(irq_source_id, irq_state); 1179 1180 return !!(*irq_state); 1181 } 1182 1183 int kvm_pic_set_irq(struct kvm_pic *pic, int irq, int irq_source_id, int level); 1184 void kvm_pic_clear_all(struct kvm_pic *pic, int irq_source_id); 1185 1186 void kvm_inject_nmi(struct kvm_vcpu *vcpu); 1187 1188 int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn); 1189 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva); 1190 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu); 1191 int kvm_mmu_load(struct kvm_vcpu *vcpu); 1192 void kvm_mmu_unload(struct kvm_vcpu *vcpu); 1193 void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu); 1194 gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access, 1195 struct x86_exception *exception); 1196 gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva, 1197 struct x86_exception *exception); 1198 gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva, 1199 struct x86_exception *exception); 1200 gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva, 1201 struct x86_exception *exception); 1202 gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva, 1203 struct x86_exception *exception); 1204 1205 void kvm_vcpu_deactivate_apicv(struct kvm_vcpu *vcpu); 1206 1207 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu); 1208 1209 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva, u32 error_code, 1210 void *insn, int insn_len); 1211 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva); 1212 void kvm_mmu_new_cr3(struct kvm_vcpu *vcpu); 1213 1214 void kvm_enable_tdp(void); 1215 void kvm_disable_tdp(void); 1216 1217 static inline gpa_t translate_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access, 1218 struct x86_exception *exception) 1219 { 1220 return gpa; 1221 } 1222 1223 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page) 1224 { 1225 struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT); 1226 1227 return (struct kvm_mmu_page *)page_private(page); 1228 } 1229 1230 static inline u16 kvm_read_ldt(void) 1231 { 1232 u16 ldt; 1233 asm("sldt %0" : "=g"(ldt)); 1234 return ldt; 1235 } 1236 1237 static inline void kvm_load_ldt(u16 sel) 1238 { 1239 asm("lldt %0" : : "rm"(sel)); 1240 } 1241 1242 #ifdef CONFIG_X86_64 1243 static inline unsigned long read_msr(unsigned long msr) 1244 { 1245 u64 value; 1246 1247 rdmsrl(msr, value); 1248 return value; 1249 } 1250 #endif 1251 1252 static inline u32 get_rdx_init_val(void) 1253 { 1254 return 0x600; /* P6 family */ 1255 } 1256 1257 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code) 1258 { 1259 kvm_queue_exception_e(vcpu, GP_VECTOR, error_code); 1260 } 1261 1262 static inline u64 get_canonical(u64 la) 1263 { 1264 return ((int64_t)la << 16) >> 16; 1265 } 1266 1267 static inline bool is_noncanonical_address(u64 la) 1268 { 1269 #ifdef CONFIG_X86_64 1270 return get_canonical(la) != la; 1271 #else 1272 return false; 1273 #endif 1274 } 1275 1276 #define TSS_IOPB_BASE_OFFSET 0x66 1277 #define TSS_BASE_SIZE 0x68 1278 #define TSS_IOPB_SIZE (65536 / 8) 1279 #define TSS_REDIRECTION_SIZE (256 / 8) 1280 #define RMODE_TSS_SIZE \ 1281 (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1) 1282 1283 enum { 1284 TASK_SWITCH_CALL = 0, 1285 TASK_SWITCH_IRET = 1, 1286 TASK_SWITCH_JMP = 2, 1287 TASK_SWITCH_GATE = 3, 1288 }; 1289 1290 #define HF_GIF_MASK (1 << 0) 1291 #define HF_HIF_MASK (1 << 1) 1292 #define HF_VINTR_MASK (1 << 2) 1293 #define HF_NMI_MASK (1 << 3) 1294 #define HF_IRET_MASK (1 << 4) 1295 #define HF_GUEST_MASK (1 << 5) /* VCPU is in guest-mode */ 1296 #define HF_SMM_MASK (1 << 6) 1297 #define HF_SMM_INSIDE_NMI_MASK (1 << 7) 1298 1299 #define __KVM_VCPU_MULTIPLE_ADDRESS_SPACE 1300 #define KVM_ADDRESS_SPACE_NUM 2 1301 1302 #define kvm_arch_vcpu_memslots_id(vcpu) ((vcpu)->arch.hflags & HF_SMM_MASK ? 1 : 0) 1303 #define kvm_memslots_for_spte_role(kvm, role) __kvm_memslots(kvm, (role).smm) 1304 1305 /* 1306 * Hardware virtualization extension instructions may fault if a 1307 * reboot turns off virtualization while processes are running. 1308 * Trap the fault and ignore the instruction if that happens. 1309 */ 1310 asmlinkage void kvm_spurious_fault(void); 1311 1312 #define ____kvm_handle_fault_on_reboot(insn, cleanup_insn) \ 1313 "666: " insn "\n\t" \ 1314 "668: \n\t" \ 1315 ".pushsection .fixup, \"ax\" \n" \ 1316 "667: \n\t" \ 1317 cleanup_insn "\n\t" \ 1318 "cmpb $0, kvm_rebooting \n\t" \ 1319 "jne 668b \n\t" \ 1320 __ASM_SIZE(push) " $666b \n\t" \ 1321 "call kvm_spurious_fault \n\t" \ 1322 ".popsection \n\t" \ 1323 _ASM_EXTABLE(666b, 667b) 1324 1325 #define __kvm_handle_fault_on_reboot(insn) \ 1326 ____kvm_handle_fault_on_reboot(insn, "") 1327 1328 #define KVM_ARCH_WANT_MMU_NOTIFIER 1329 int kvm_unmap_hva(struct kvm *kvm, unsigned long hva); 1330 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end); 1331 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end); 1332 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva); 1333 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte); 1334 int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v); 1335 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu); 1336 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu); 1337 int kvm_cpu_get_interrupt(struct kvm_vcpu *v); 1338 void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event); 1339 void kvm_vcpu_reload_apic_access_page(struct kvm_vcpu *vcpu); 1340 void kvm_arch_mmu_notifier_invalidate_page(struct kvm *kvm, 1341 unsigned long address); 1342 1343 void kvm_define_shared_msr(unsigned index, u32 msr); 1344 int kvm_set_shared_msr(unsigned index, u64 val, u64 mask); 1345 1346 u64 kvm_scale_tsc(struct kvm_vcpu *vcpu, u64 tsc); 1347 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc); 1348 1349 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu); 1350 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip); 1351 1352 void kvm_make_mclock_inprogress_request(struct kvm *kvm); 1353 void kvm_make_scan_ioapic_request(struct kvm *kvm); 1354 1355 void kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu, 1356 struct kvm_async_pf *work); 1357 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu, 1358 struct kvm_async_pf *work); 1359 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu, 1360 struct kvm_async_pf *work); 1361 bool kvm_arch_can_inject_async_page_present(struct kvm_vcpu *vcpu); 1362 extern bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn); 1363 1364 void kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err); 1365 1366 int kvm_is_in_guest(void); 1367 1368 int __x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa, u32 size); 1369 int x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa, u32 size); 1370 bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu); 1371 bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu); 1372 1373 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq, 1374 struct kvm_vcpu **dest_vcpu); 1375 1376 void kvm_set_msi_irq(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e, 1377 struct kvm_lapic_irq *irq); 1378 1379 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu) 1380 { 1381 if (kvm_x86_ops->vcpu_blocking) 1382 kvm_x86_ops->vcpu_blocking(vcpu); 1383 } 1384 1385 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu) 1386 { 1387 if (kvm_x86_ops->vcpu_unblocking) 1388 kvm_x86_ops->vcpu_unblocking(vcpu); 1389 } 1390 1391 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {} 1392 1393 static inline int kvm_cpu_get_apicid(int mps_cpu) 1394 { 1395 #ifdef CONFIG_X86_LOCAL_APIC 1396 return __default_cpu_present_to_apicid(mps_cpu); 1397 #else 1398 WARN_ON_ONCE(1); 1399 return BAD_APICID; 1400 #endif 1401 } 1402 1403 #endif /* _ASM_X86_KVM_HOST_H */ 1404