| 1 | /* SPDX-License-Identifier: GPL-2.0 */ |
| 2 | #ifndef ARCH_X86_KVM_X86_H |
| 3 | #define ARCH_X86_KVM_X86_H |
| 4 | |
| 5 | #include <linux/kvm_host.h> |
| 6 | #include <asm/fpu/xstate.h> |
| 7 | #include <asm/mce.h> |
| 8 | #include <asm/pvclock.h> |
| 9 | #include "kvm_cache_regs.h" |
| 10 | #include "kvm_emulate.h" |
| 11 | #include "cpuid.h" |
| 12 | |
| 13 | #define KVM_MAX_MCE_BANKS 32 |
| 14 | |
| 15 | struct kvm_caps { |
| 16 | /* control of guest tsc rate supported? */ |
| 17 | bool has_tsc_control; |
| 18 | /* maximum supported tsc_khz for guests */ |
| 19 | u32 max_guest_tsc_khz; |
| 20 | /* number of bits of the fractional part of the TSC scaling ratio */ |
| 21 | u8 tsc_scaling_ratio_frac_bits; |
| 22 | /* maximum allowed value of TSC scaling ratio */ |
| 23 | u64 max_tsc_scaling_ratio; |
| 24 | /* 1ull << kvm_caps.tsc_scaling_ratio_frac_bits */ |
| 25 | u64 default_tsc_scaling_ratio; |
| 26 | /* bus lock detection supported? */ |
| 27 | bool has_bus_lock_exit; |
| 28 | /* notify VM exit supported? */ |
| 29 | bool has_notify_vmexit; |
| 30 | /* bit mask of VM types */ |
| 31 | u32 supported_vm_types; |
| 32 | |
| 33 | u64 supported_mce_cap; |
| 34 | u64 supported_xcr0; |
| 35 | u64 supported_xss; |
| 36 | u64 supported_perf_cap; |
| 37 | |
| 38 | u64 supported_quirks; |
| 39 | u64 inapplicable_quirks; |
| 40 | }; |
| 41 | |
| 42 | struct kvm_host_values { |
| 43 | /* |
| 44 | * The host's raw MAXPHYADDR, i.e. the number of non-reserved physical |
| 45 | * address bits irrespective of features that repurpose legal bits, |
| 46 | * e.g. MKTME. |
| 47 | */ |
| 48 | u8 maxphyaddr; |
| 49 | |
| 50 | u64 efer; |
| 51 | u64 xcr0; |
| 52 | u64 xss; |
| 53 | u64 s_cet; |
| 54 | u64 arch_capabilities; |
| 55 | }; |
| 56 | |
| 57 | void kvm_spurious_fault(void); |
| 58 | |
| 59 | #define SIZE_OF_MEMSLOTS_HASHTABLE \ |
| 60 | (sizeof(((struct kvm_memslots *)0)->id_hash) * 2 * KVM_MAX_NR_ADDRESS_SPACES) |
| 61 | |
| 62 | /* Sanity check the size of the memslot hash tables. */ |
| 63 | static_assert(SIZE_OF_MEMSLOTS_HASHTABLE == |
| 64 | (1024 * (1 + IS_ENABLED(CONFIG_X86_64)) * (1 + IS_ENABLED(CONFIG_KVM_SMM)))); |
| 65 | |
| 66 | /* |
| 67 | * Assert that "struct kvm_{svm,vmx,tdx}" is an order-0 or order-1 allocation. |
| 68 | * Spilling over to an order-2 allocation isn't fundamentally problematic, but |
| 69 | * isn't expected to happen in the foreseeable future (O(years)). Assert that |
| 70 | * the size is an order-0 allocation when ignoring the memslot hash tables, to |
| 71 | * help detect and debug unexpected size increases. |
| 72 | */ |
| 73 | #define KVM_SANITY_CHECK_VM_STRUCT_SIZE(x) \ |
| 74 | do { \ |
| 75 | BUILD_BUG_ON(get_order(sizeof(struct x) - SIZE_OF_MEMSLOTS_HASHTABLE) && \ |
| 76 | !IS_ENABLED(CONFIG_DEBUG_KERNEL) && !IS_ENABLED(CONFIG_KASAN)); \ |
| 77 | BUILD_BUG_ON(get_order(sizeof(struct x)) > 1 && \ |
| 78 | !IS_ENABLED(CONFIG_DEBUG_KERNEL) && !IS_ENABLED(CONFIG_KASAN)); \ |
| 79 | } while (0) |
| 80 | |
| 81 | #define KVM_NESTED_VMENTER_CONSISTENCY_CHECK(consistency_check) \ |
| 82 | ({ \ |
| 83 | bool failed = (consistency_check); \ |
| 84 | if (failed) \ |
| 85 | trace_kvm_nested_vmenter_failed(#consistency_check, 0); \ |
| 86 | failed; \ |
| 87 | }) |
| 88 | |
| 89 | /* |
| 90 | * The first...last VMX feature MSRs that are emulated by KVM. This may or may |
| 91 | * not cover all known VMX MSRs, as KVM doesn't emulate an MSR until there's an |
| 92 | * associated feature that KVM supports for nested virtualization. |
| 93 | */ |
| 94 | #define KVM_FIRST_EMULATED_VMX_MSR MSR_IA32_VMX_BASIC |
| 95 | #define KVM_LAST_EMULATED_VMX_MSR MSR_IA32_VMX_VMFUNC |
| 96 | |
| 97 | #define KVM_DEFAULT_PLE_GAP 128 |
| 98 | #define KVM_VMX_DEFAULT_PLE_WINDOW 4096 |
| 99 | #define KVM_DEFAULT_PLE_WINDOW_GROW 2 |
| 100 | #define KVM_DEFAULT_PLE_WINDOW_SHRINK 0 |
| 101 | #define KVM_VMX_DEFAULT_PLE_WINDOW_MAX UINT_MAX |
| 102 | #define KVM_SVM_DEFAULT_PLE_WINDOW_MAX USHRT_MAX |
| 103 | #define KVM_SVM_DEFAULT_PLE_WINDOW 3000 |
| 104 | |
| 105 | /* |
| 106 | * KVM's internal, non-ABI indices for synthetic MSRs. The values themselves |
| 107 | * are arbitrary and have no meaning, the only requirement is that they don't |
| 108 | * conflict with "real" MSRs that KVM supports. Use values at the upper end |
| 109 | * of KVM's reserved paravirtual MSR range to minimize churn, i.e. these values |
| 110 | * will be usable until KVM exhausts its supply of paravirtual MSR indices. |
| 111 | */ |
| 112 | |
| 113 | #define MSR_KVM_INTERNAL_GUEST_SSP 0x4b564dff |
| 114 | |
| 115 | static inline unsigned int __grow_ple_window(unsigned int val, |
| 116 | unsigned int base, unsigned int modifier, unsigned int max) |
| 117 | { |
| 118 | u64 ret = val; |
| 119 | |
| 120 | if (modifier < 1) |
| 121 | return base; |
| 122 | |
| 123 | if (modifier < base) |
| 124 | ret *= modifier; |
| 125 | else |
| 126 | ret += modifier; |
| 127 | |
| 128 | return min(ret, (u64)max); |
| 129 | } |
| 130 | |
| 131 | static inline unsigned int __shrink_ple_window(unsigned int val, |
| 132 | unsigned int base, unsigned int modifier, unsigned int min) |
| 133 | { |
| 134 | if (modifier < 1) |
| 135 | return base; |
| 136 | |
| 137 | if (modifier < base) |
| 138 | val /= modifier; |
| 139 | else |
| 140 | val -= modifier; |
| 141 | |
| 142 | return max(val, min); |
| 143 | } |
| 144 | |
| 145 | #define MSR_IA32_CR_PAT_DEFAULT \ |
| 146 | PAT_VALUE(WB, WT, UC_MINUS, UC, WB, WT, UC_MINUS, UC) |
| 147 | |
| 148 | void kvm_service_local_tlb_flush_requests(struct kvm_vcpu *vcpu); |
| 149 | int kvm_check_nested_events(struct kvm_vcpu *vcpu); |
| 150 | |
| 151 | /* Forcibly leave the nested mode in cases like a vCPU reset */ |
| 152 | static inline void kvm_leave_nested(struct kvm_vcpu *vcpu) |
| 153 | { |
| 154 | kvm_x86_ops.nested_ops->leave_nested(vcpu); |
| 155 | } |
| 156 | |
| 157 | /* |
| 158 | * If IBRS is advertised to the vCPU, KVM must flush the indirect branch |
| 159 | * predictors when transitioning from L2 to L1, as L1 expects hardware (KVM in |
| 160 | * this case) to provide separate predictor modes. Bare metal isolates the host |
| 161 | * from the guest, but doesn't isolate different guests from one another (in |
| 162 | * this case L1 and L2). The exception is if bare metal supports same mode IBRS, |
| 163 | * which offers protection within the same mode, and hence protects L1 from L2. |
| 164 | */ |
| 165 | static inline void kvm_nested_vmexit_handle_ibrs(struct kvm_vcpu *vcpu) |
| 166 | { |
| 167 | if (cpu_feature_enabled(X86_FEATURE_AMD_IBRS_SAME_MODE)) |
| 168 | return; |
| 169 | |
| 170 | if (guest_cpu_cap_has(vcpu, X86_FEATURE_SPEC_CTRL) || |
| 171 | guest_cpu_cap_has(vcpu, X86_FEATURE_AMD_IBRS)) |
| 172 | indirect_branch_prediction_barrier(); |
| 173 | } |
| 174 | |
| 175 | static inline bool kvm_vcpu_has_run(struct kvm_vcpu *vcpu) |
| 176 | { |
| 177 | return vcpu->arch.last_vmentry_cpu != -1; |
| 178 | } |
| 179 | |
| 180 | static inline void kvm_set_mp_state(struct kvm_vcpu *vcpu, int mp_state) |
| 181 | { |
| 182 | vcpu->arch.mp_state = mp_state; |
| 183 | if (mp_state == KVM_MP_STATE_RUNNABLE) |
| 184 | vcpu->arch.pv.pv_unhalted = false; |
| 185 | } |
| 186 | |
| 187 | static inline bool kvm_is_exception_pending(struct kvm_vcpu *vcpu) |
| 188 | { |
| 189 | return vcpu->arch.exception.pending || |
| 190 | vcpu->arch.exception_vmexit.pending || |
| 191 | kvm_test_request(KVM_REQ_TRIPLE_FAULT, vcpu); |
| 192 | } |
| 193 | |
| 194 | static inline void kvm_clear_exception_queue(struct kvm_vcpu *vcpu) |
| 195 | { |
| 196 | vcpu->arch.exception.pending = false; |
| 197 | vcpu->arch.exception.injected = false; |
| 198 | vcpu->arch.exception_vmexit.pending = false; |
| 199 | } |
| 200 | |
| 201 | static inline void kvm_queue_interrupt(struct kvm_vcpu *vcpu, u8 vector, |
| 202 | bool soft) |
| 203 | { |
| 204 | vcpu->arch.interrupt.injected = true; |
| 205 | vcpu->arch.interrupt.soft = soft; |
| 206 | vcpu->arch.interrupt.nr = vector; |
| 207 | } |
| 208 | |
| 209 | static inline void kvm_clear_interrupt_queue(struct kvm_vcpu *vcpu) |
| 210 | { |
| 211 | vcpu->arch.interrupt.injected = false; |
| 212 | } |
| 213 | |
| 214 | static inline bool kvm_event_needs_reinjection(struct kvm_vcpu *vcpu) |
| 215 | { |
| 216 | return vcpu->arch.exception.injected || vcpu->arch.interrupt.injected || |
| 217 | vcpu->arch.nmi_injected; |
| 218 | } |
| 219 | |
| 220 | static inline bool kvm_exception_is_soft(unsigned int nr) |
| 221 | { |
| 222 | return (nr == BP_VECTOR) || (nr == OF_VECTOR); |
| 223 | } |
| 224 | |
| 225 | static inline bool is_protmode(struct kvm_vcpu *vcpu) |
| 226 | { |
| 227 | return kvm_is_cr0_bit_set(vcpu, X86_CR0_PE); |
| 228 | } |
| 229 | |
| 230 | static inline bool is_long_mode(struct kvm_vcpu *vcpu) |
| 231 | { |
| 232 | #ifdef CONFIG_X86_64 |
| 233 | return !!(vcpu->arch.efer & EFER_LMA); |
| 234 | #else |
| 235 | return false; |
| 236 | #endif |
| 237 | } |
| 238 | |
| 239 | static inline bool is_64_bit_mode(struct kvm_vcpu *vcpu) |
| 240 | { |
| 241 | int cs_db, cs_l; |
| 242 | |
| 243 | WARN_ON_ONCE(vcpu->arch.guest_state_protected); |
| 244 | |
| 245 | if (!is_long_mode(vcpu)) |
| 246 | return false; |
| 247 | kvm_x86_call(get_cs_db_l_bits)(vcpu, &cs_db, &cs_l); |
| 248 | return cs_l; |
| 249 | } |
| 250 | |
| 251 | static inline bool is_64_bit_hypercall(struct kvm_vcpu *vcpu) |
| 252 | { |
| 253 | /* |
| 254 | * If running with protected guest state, the CS register is not |
| 255 | * accessible. The hypercall register values will have had to been |
| 256 | * provided in 64-bit mode, so assume the guest is in 64-bit. |
| 257 | */ |
| 258 | return vcpu->arch.guest_state_protected || is_64_bit_mode(vcpu); |
| 259 | } |
| 260 | |
| 261 | static inline bool x86_exception_has_error_code(unsigned int vector) |
| 262 | { |
| 263 | static u32 exception_has_error_code = BIT(DF_VECTOR) | BIT(TS_VECTOR) | |
| 264 | BIT(NP_VECTOR) | BIT(SS_VECTOR) | BIT(GP_VECTOR) | |
| 265 | BIT(PF_VECTOR) | BIT(AC_VECTOR); |
| 266 | |
| 267 | return (1U << vector) & exception_has_error_code; |
| 268 | } |
| 269 | |
| 270 | static inline bool mmu_is_nested(struct kvm_vcpu *vcpu) |
| 271 | { |
| 272 | return vcpu->arch.walk_mmu == &vcpu->arch.nested_mmu; |
| 273 | } |
| 274 | |
| 275 | static inline bool is_pae(struct kvm_vcpu *vcpu) |
| 276 | { |
| 277 | return kvm_is_cr4_bit_set(vcpu, X86_CR4_PAE); |
| 278 | } |
| 279 | |
| 280 | static inline bool is_pse(struct kvm_vcpu *vcpu) |
| 281 | { |
| 282 | return kvm_is_cr4_bit_set(vcpu, X86_CR4_PSE); |
| 283 | } |
| 284 | |
| 285 | static inline bool is_paging(struct kvm_vcpu *vcpu) |
| 286 | { |
| 287 | return likely(kvm_is_cr0_bit_set(vcpu, X86_CR0_PG)); |
| 288 | } |
| 289 | |
| 290 | static inline bool is_pae_paging(struct kvm_vcpu *vcpu) |
| 291 | { |
| 292 | return !is_long_mode(vcpu) && is_pae(vcpu) && is_paging(vcpu); |
| 293 | } |
| 294 | |
| 295 | static inline u8 vcpu_virt_addr_bits(struct kvm_vcpu *vcpu) |
| 296 | { |
| 297 | return kvm_is_cr4_bit_set(vcpu, X86_CR4_LA57) ? 57 : 48; |
| 298 | } |
| 299 | |
| 300 | static inline u8 max_host_virt_addr_bits(void) |
| 301 | { |
| 302 | return kvm_cpu_cap_has(X86_FEATURE_LA57) ? 57 : 48; |
| 303 | } |
| 304 | |
| 305 | /* |
| 306 | * x86 MSRs which contain linear addresses, x86 hidden segment bases, and |
| 307 | * IDT/GDT bases have static canonicality checks, the size of which depends |
| 308 | * only on the CPU's support for 5-level paging, rather than on the state of |
| 309 | * CR4.LA57. This applies to both WRMSR and to other instructions that set |
| 310 | * their values, e.g. SGDT. |
| 311 | * |
| 312 | * KVM passes through most of these MSRS and also doesn't intercept the |
| 313 | * instructions that set the hidden segment bases. |
| 314 | * |
| 315 | * Because of this, to be consistent with hardware, even if the guest doesn't |
| 316 | * have LA57 enabled in its CPUID, perform canonicality checks based on *host* |
| 317 | * support for 5 level paging. |
| 318 | * |
| 319 | * Finally, instructions which are related to MMU invalidation of a given |
| 320 | * linear address, also have a similar static canonical check on address. |
| 321 | * This allows for example to invalidate 5-level addresses of a guest from a |
| 322 | * host which uses 4-level paging. |
| 323 | */ |
| 324 | static inline bool is_noncanonical_address(u64 la, struct kvm_vcpu *vcpu, |
| 325 | unsigned int flags) |
| 326 | { |
| 327 | if (flags & (X86EMUL_F_INVLPG | X86EMUL_F_MSR | X86EMUL_F_DT_LOAD)) |
| 328 | return !__is_canonical_address(vaddr: la, vaddr_bits: max_host_virt_addr_bits()); |
| 329 | else |
| 330 | return !__is_canonical_address(vaddr: la, vaddr_bits: vcpu_virt_addr_bits(vcpu)); |
| 331 | } |
| 332 | |
| 333 | static inline bool is_noncanonical_msr_address(u64 la, struct kvm_vcpu *vcpu) |
| 334 | { |
| 335 | return is_noncanonical_address(la, vcpu, X86EMUL_F_MSR); |
| 336 | } |
| 337 | |
| 338 | static inline bool is_noncanonical_base_address(u64 la, struct kvm_vcpu *vcpu) |
| 339 | { |
| 340 | return is_noncanonical_address(la, vcpu, X86EMUL_F_DT_LOAD); |
| 341 | } |
| 342 | |
| 343 | static inline bool is_noncanonical_invlpg_address(u64 la, struct kvm_vcpu *vcpu) |
| 344 | { |
| 345 | return is_noncanonical_address(la, vcpu, X86EMUL_F_INVLPG); |
| 346 | } |
| 347 | |
| 348 | static inline void vcpu_cache_mmio_info(struct kvm_vcpu *vcpu, |
| 349 | gva_t gva, gfn_t gfn, unsigned access) |
| 350 | { |
| 351 | u64 gen = kvm_memslots(kvm: vcpu->kvm)->generation; |
| 352 | |
| 353 | if (unlikely(gen & KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS)) |
| 354 | return; |
| 355 | |
| 356 | /* |
| 357 | * If this is a shadow nested page table, the "GVA" is |
| 358 | * actually a nGPA. |
| 359 | */ |
| 360 | vcpu->arch.mmio_gva = mmu_is_nested(vcpu) ? 0 : gva & PAGE_MASK; |
| 361 | vcpu->arch.mmio_access = access; |
| 362 | vcpu->arch.mmio_gfn = gfn; |
| 363 | vcpu->arch.mmio_gen = gen; |
| 364 | } |
| 365 | |
| 366 | static inline bool vcpu_match_mmio_gen(struct kvm_vcpu *vcpu) |
| 367 | { |
| 368 | return vcpu->arch.mmio_gen == kvm_memslots(kvm: vcpu->kvm)->generation; |
| 369 | } |
| 370 | |
| 371 | /* |
| 372 | * Clear the mmio cache info for the given gva. If gva is MMIO_GVA_ANY, we |
| 373 | * clear all mmio cache info. |
| 374 | */ |
| 375 | #define MMIO_GVA_ANY (~(gva_t)0) |
| 376 | |
| 377 | static inline void vcpu_clear_mmio_info(struct kvm_vcpu *vcpu, gva_t gva) |
| 378 | { |
| 379 | if (gva != MMIO_GVA_ANY && vcpu->arch.mmio_gva != (gva & PAGE_MASK)) |
| 380 | return; |
| 381 | |
| 382 | vcpu->arch.mmio_gva = 0; |
| 383 | } |
| 384 | |
| 385 | static inline bool vcpu_match_mmio_gva(struct kvm_vcpu *vcpu, unsigned long gva) |
| 386 | { |
| 387 | if (vcpu_match_mmio_gen(vcpu) && vcpu->arch.mmio_gva && |
| 388 | vcpu->arch.mmio_gva == (gva & PAGE_MASK)) |
| 389 | return true; |
| 390 | |
| 391 | return false; |
| 392 | } |
| 393 | |
| 394 | static inline bool vcpu_match_mmio_gpa(struct kvm_vcpu *vcpu, gpa_t gpa) |
| 395 | { |
| 396 | if (vcpu_match_mmio_gen(vcpu) && vcpu->arch.mmio_gfn && |
| 397 | vcpu->arch.mmio_gfn == gpa >> PAGE_SHIFT) |
| 398 | return true; |
| 399 | |
| 400 | return false; |
| 401 | } |
| 402 | |
| 403 | static inline unsigned long kvm_register_read(struct kvm_vcpu *vcpu, int reg) |
| 404 | { |
| 405 | unsigned long val = kvm_register_read_raw(vcpu, reg); |
| 406 | |
| 407 | return is_64_bit_mode(vcpu) ? val : (u32)val; |
| 408 | } |
| 409 | |
| 410 | static inline void kvm_register_write(struct kvm_vcpu *vcpu, |
| 411 | int reg, unsigned long val) |
| 412 | { |
| 413 | if (!is_64_bit_mode(vcpu)) |
| 414 | val = (u32)val; |
| 415 | return kvm_register_write_raw(vcpu, reg, val); |
| 416 | } |
| 417 | |
| 418 | static inline bool kvm_check_has_quirk(struct kvm *kvm, u64 quirk) |
| 419 | { |
| 420 | return !(kvm->arch.disabled_quirks & quirk); |
| 421 | } |
| 422 | |
| 423 | static __always_inline void kvm_request_l1tf_flush_l1d(void) |
| 424 | { |
| 425 | #if IS_ENABLED(CONFIG_CPU_MITIGATIONS) && IS_ENABLED(CONFIG_KVM_INTEL) |
| 426 | /* |
| 427 | * Use a raw write to set the per-CPU flag, as KVM will ensure a flush |
| 428 | * even if preemption is currently enabled.. If the current vCPU task |
| 429 | * is migrated to a different CPU (or userspace runs the vCPU on a |
| 430 | * different task) before the next VM-Entry, then kvm_arch_vcpu_load() |
| 431 | * will request a flush on the new CPU. |
| 432 | */ |
| 433 | raw_cpu_write(irq_stat.kvm_cpu_l1tf_flush_l1d, 1); |
| 434 | #endif |
| 435 | } |
| 436 | |
| 437 | void kvm_inject_realmode_interrupt(struct kvm_vcpu *vcpu, int irq, int inc_eip); |
| 438 | |
| 439 | u64 get_kvmclock_ns(struct kvm *kvm); |
| 440 | uint64_t kvm_get_wall_clock_epoch(struct kvm *kvm); |
| 441 | bool kvm_get_monotonic_and_clockread(s64 *kernel_ns, u64 *tsc_timestamp); |
| 442 | int kvm_guest_time_update(struct kvm_vcpu *v); |
| 443 | |
| 444 | int kvm_read_guest_virt(struct kvm_vcpu *vcpu, |
| 445 | gva_t addr, void *val, unsigned int bytes, |
| 446 | struct x86_exception *exception); |
| 447 | |
| 448 | int kvm_write_guest_virt_system(struct kvm_vcpu *vcpu, |
| 449 | gva_t addr, void *val, unsigned int bytes, |
| 450 | struct x86_exception *exception); |
| 451 | |
| 452 | int handle_ud(struct kvm_vcpu *vcpu); |
| 453 | |
| 454 | void kvm_deliver_exception_payload(struct kvm_vcpu *vcpu, |
| 455 | struct kvm_queued_exception *ex); |
| 456 | |
| 457 | int kvm_mtrr_set_msr(struct kvm_vcpu *vcpu, u32 msr, u64 data); |
| 458 | int kvm_mtrr_get_msr(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata); |
| 459 | void kvm_fixup_and_inject_pf_error(struct kvm_vcpu *vcpu, gva_t gva, u16 error_code); |
| 460 | int x86_decode_emulated_instruction(struct kvm_vcpu *vcpu, int emulation_type, |
| 461 | void *insn, int insn_len); |
| 462 | int x86_emulate_instruction(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, |
| 463 | int emulation_type, void *insn, int insn_len); |
| 464 | fastpath_t handle_fastpath_wrmsr(struct kvm_vcpu *vcpu); |
| 465 | fastpath_t handle_fastpath_wrmsr_imm(struct kvm_vcpu *vcpu, u32 msr, int reg); |
| 466 | fastpath_t handle_fastpath_hlt(struct kvm_vcpu *vcpu); |
| 467 | fastpath_t handle_fastpath_invd(struct kvm_vcpu *vcpu); |
| 468 | |
| 469 | extern struct kvm_caps kvm_caps; |
| 470 | extern struct kvm_host_values kvm_host; |
| 471 | |
| 472 | extern bool enable_pmu; |
| 473 | |
| 474 | void kvm_setup_xss_caps(void); |
| 475 | |
| 476 | /* |
| 477 | * Get a filtered version of KVM's supported XCR0 that strips out dynamic |
| 478 | * features for which the current process doesn't (yet) have permission to use. |
| 479 | * This is intended to be used only when enumerating support to userspace, |
| 480 | * e.g. in KVM_GET_SUPPORTED_CPUID and KVM_CAP_XSAVE2, it does NOT need to be |
| 481 | * used to check/restrict guest behavior as KVM rejects KVM_SET_CPUID{2} if |
| 482 | * userspace attempts to enable unpermitted features. |
| 483 | */ |
| 484 | static inline u64 kvm_get_filtered_xcr0(void) |
| 485 | { |
| 486 | u64 permitted_xcr0 = kvm_caps.supported_xcr0; |
| 487 | |
| 488 | BUILD_BUG_ON(XFEATURE_MASK_USER_DYNAMIC != XFEATURE_MASK_XTILE_DATA); |
| 489 | |
| 490 | if (permitted_xcr0 & XFEATURE_MASK_USER_DYNAMIC) { |
| 491 | permitted_xcr0 &= xstate_get_guest_group_perm(); |
| 492 | |
| 493 | /* |
| 494 | * Treat XTILE_CFG as unsupported if the current process isn't |
| 495 | * allowed to use XTILE_DATA, as attempting to set XTILE_CFG in |
| 496 | * XCR0 without setting XTILE_DATA is architecturally illegal. |
| 497 | */ |
| 498 | if (!(permitted_xcr0 & XFEATURE_MASK_XTILE_DATA)) |
| 499 | permitted_xcr0 &= ~XFEATURE_MASK_XTILE_CFG; |
| 500 | } |
| 501 | return permitted_xcr0; |
| 502 | } |
| 503 | |
| 504 | static inline bool kvm_mpx_supported(void) |
| 505 | { |
| 506 | return (kvm_caps.supported_xcr0 & (XFEATURE_MASK_BNDREGS | XFEATURE_MASK_BNDCSR)) |
| 507 | == (XFEATURE_MASK_BNDREGS | XFEATURE_MASK_BNDCSR); |
| 508 | } |
| 509 | |
| 510 | extern unsigned int min_timer_period_us; |
| 511 | |
| 512 | extern bool enable_vmware_backdoor; |
| 513 | |
| 514 | extern int pi_inject_timer; |
| 515 | |
| 516 | extern bool report_ignored_msrs; |
| 517 | |
| 518 | extern bool eager_page_split; |
| 519 | |
| 520 | static inline void kvm_pr_unimpl_wrmsr(struct kvm_vcpu *vcpu, u32 msr, u64 data) |
| 521 | { |
| 522 | if (report_ignored_msrs) |
| 523 | vcpu_unimpl(vcpu, "Unhandled WRMSR(0x%x) = 0x%llx\n" , msr, data); |
| 524 | } |
| 525 | |
| 526 | static inline void kvm_pr_unimpl_rdmsr(struct kvm_vcpu *vcpu, u32 msr) |
| 527 | { |
| 528 | if (report_ignored_msrs) |
| 529 | vcpu_unimpl(vcpu, "Unhandled RDMSR(0x%x)\n" , msr); |
| 530 | } |
| 531 | |
| 532 | static inline u64 nsec_to_cycles(struct kvm_vcpu *vcpu, u64 nsec) |
| 533 | { |
| 534 | return pvclock_scale_delta(delta: nsec, mul_frac: vcpu->arch.virtual_tsc_mult, |
| 535 | shift: vcpu->arch.virtual_tsc_shift); |
| 536 | } |
| 537 | |
| 538 | /* Same "calling convention" as do_div: |
| 539 | * - divide (n << 32) by base |
| 540 | * - put result in n |
| 541 | * - return remainder |
| 542 | */ |
| 543 | #define do_shl32_div32(n, base) \ |
| 544 | ({ \ |
| 545 | u32 __quot, __rem; \ |
| 546 | asm("divl %2" : "=a" (__quot), "=d" (__rem) \ |
| 547 | : "rm" (base), "0" (0), "1" ((u32) n)); \ |
| 548 | n = __quot; \ |
| 549 | __rem; \ |
| 550 | }) |
| 551 | |
| 552 | static inline void kvm_disable_exits(struct kvm *kvm, u64 mask) |
| 553 | { |
| 554 | kvm->arch.disabled_exits |= mask; |
| 555 | } |
| 556 | |
| 557 | static inline bool kvm_mwait_in_guest(struct kvm *kvm) |
| 558 | { |
| 559 | return kvm->arch.disabled_exits & KVM_X86_DISABLE_EXITS_MWAIT; |
| 560 | } |
| 561 | |
| 562 | static inline bool kvm_hlt_in_guest(struct kvm *kvm) |
| 563 | { |
| 564 | return kvm->arch.disabled_exits & KVM_X86_DISABLE_EXITS_HLT; |
| 565 | } |
| 566 | |
| 567 | static inline bool kvm_pause_in_guest(struct kvm *kvm) |
| 568 | { |
| 569 | return kvm->arch.disabled_exits & KVM_X86_DISABLE_EXITS_PAUSE; |
| 570 | } |
| 571 | |
| 572 | static inline bool kvm_cstate_in_guest(struct kvm *kvm) |
| 573 | { |
| 574 | return kvm->arch.disabled_exits & KVM_X86_DISABLE_EXITS_CSTATE; |
| 575 | } |
| 576 | |
| 577 | static inline bool kvm_aperfmperf_in_guest(struct kvm *kvm) |
| 578 | { |
| 579 | return kvm->arch.disabled_exits & KVM_X86_DISABLE_EXITS_APERFMPERF; |
| 580 | } |
| 581 | |
| 582 | static inline bool kvm_notify_vmexit_enabled(struct kvm *kvm) |
| 583 | { |
| 584 | return kvm->arch.notify_vmexit_flags & KVM_X86_NOTIFY_VMEXIT_ENABLED; |
| 585 | } |
| 586 | |
| 587 | static __always_inline void kvm_before_interrupt(struct kvm_vcpu *vcpu, |
| 588 | enum kvm_intr_type intr) |
| 589 | { |
| 590 | WRITE_ONCE(vcpu->arch.handling_intr_from_guest, (u8)intr); |
| 591 | } |
| 592 | |
| 593 | static __always_inline void kvm_after_interrupt(struct kvm_vcpu *vcpu) |
| 594 | { |
| 595 | WRITE_ONCE(vcpu->arch.handling_intr_from_guest, 0); |
| 596 | } |
| 597 | |
| 598 | static inline bool kvm_handling_nmi_from_guest(struct kvm_vcpu *vcpu) |
| 599 | { |
| 600 | return vcpu->arch.handling_intr_from_guest == KVM_HANDLING_NMI; |
| 601 | } |
| 602 | |
| 603 | static inline bool kvm_pat_valid(u64 data) |
| 604 | { |
| 605 | if (data & 0xF8F8F8F8F8F8F8F8ull) |
| 606 | return false; |
| 607 | /* 0, 1, 4, 5, 6, 7 are valid values. */ |
| 608 | return (data | ((data & 0x0202020202020202ull) << 1)) == data; |
| 609 | } |
| 610 | |
| 611 | static inline bool kvm_dr7_valid(u64 data) |
| 612 | { |
| 613 | /* Bits [63:32] are reserved */ |
| 614 | return !(data >> 32); |
| 615 | } |
| 616 | static inline bool kvm_dr6_valid(u64 data) |
| 617 | { |
| 618 | /* Bits [63:32] are reserved */ |
| 619 | return !(data >> 32); |
| 620 | } |
| 621 | |
| 622 | /* |
| 623 | * Trigger machine check on the host. We assume all the MSRs are already set up |
| 624 | * by the CPU and that we still run on the same CPU as the MCE occurred on. |
| 625 | * We pass a fake environment to the machine check handler because we want |
| 626 | * the guest to be always treated like user space, no matter what context |
| 627 | * it used internally. |
| 628 | */ |
| 629 | static inline void kvm_machine_check(void) |
| 630 | { |
| 631 | #if defined(CONFIG_X86_MCE) |
| 632 | struct pt_regs regs = { |
| 633 | .cs = 3, /* Fake ring 3 no matter what the guest ran on */ |
| 634 | .flags = X86_EFLAGS_IF, |
| 635 | }; |
| 636 | |
| 637 | do_machine_check(pt_regs: ®s); |
| 638 | #endif |
| 639 | } |
| 640 | |
| 641 | int kvm_spec_ctrl_test_value(u64 value); |
| 642 | int kvm_handle_memory_failure(struct kvm_vcpu *vcpu, int r, |
| 643 | struct x86_exception *e); |
| 644 | int kvm_handle_invpcid(struct kvm_vcpu *vcpu, unsigned long type, gva_t gva); |
| 645 | bool kvm_msr_allowed(struct kvm_vcpu *vcpu, u32 index, u32 type); |
| 646 | |
| 647 | enum kvm_msr_access { |
| 648 | MSR_TYPE_R = BIT(0), |
| 649 | MSR_TYPE_W = BIT(1), |
| 650 | MSR_TYPE_RW = MSR_TYPE_R | MSR_TYPE_W, |
| 651 | }; |
| 652 | |
| 653 | /* |
| 654 | * Internal error codes that are used to indicate that MSR emulation encountered |
| 655 | * an error that should result in #GP in the guest, unless userspace handles it. |
| 656 | * Note, '1', '0', and negative numbers are off limits, as they are used by KVM |
| 657 | * as part of KVM's lightly documented internal KVM_RUN return codes. |
| 658 | * |
| 659 | * UNSUPPORTED - The MSR isn't supported, either because it is completely |
| 660 | * unknown to KVM, or because the MSR should not exist according |
| 661 | * to the vCPU model. |
| 662 | * |
| 663 | * FILTERED - Access to the MSR is denied by a userspace MSR filter. |
| 664 | */ |
| 665 | #define KVM_MSR_RET_UNSUPPORTED 2 |
| 666 | #define KVM_MSR_RET_FILTERED 3 |
| 667 | |
| 668 | static inline bool __kvm_is_valid_cr4(struct kvm_vcpu *vcpu, unsigned long cr4) |
| 669 | { |
| 670 | return !(cr4 & vcpu->arch.cr4_guest_rsvd_bits); |
| 671 | } |
| 672 | |
| 673 | #define __cr4_reserved_bits(__cpu_has, __c) \ |
| 674 | ({ \ |
| 675 | u64 __reserved_bits = CR4_RESERVED_BITS; \ |
| 676 | \ |
| 677 | if (!__cpu_has(__c, X86_FEATURE_XSAVE)) \ |
| 678 | __reserved_bits |= X86_CR4_OSXSAVE; \ |
| 679 | if (!__cpu_has(__c, X86_FEATURE_SMEP)) \ |
| 680 | __reserved_bits |= X86_CR4_SMEP; \ |
| 681 | if (!__cpu_has(__c, X86_FEATURE_SMAP)) \ |
| 682 | __reserved_bits |= X86_CR4_SMAP; \ |
| 683 | if (!__cpu_has(__c, X86_FEATURE_FSGSBASE)) \ |
| 684 | __reserved_bits |= X86_CR4_FSGSBASE; \ |
| 685 | if (!__cpu_has(__c, X86_FEATURE_PKU)) \ |
| 686 | __reserved_bits |= X86_CR4_PKE; \ |
| 687 | if (!__cpu_has(__c, X86_FEATURE_LA57)) \ |
| 688 | __reserved_bits |= X86_CR4_LA57; \ |
| 689 | if (!__cpu_has(__c, X86_FEATURE_UMIP)) \ |
| 690 | __reserved_bits |= X86_CR4_UMIP; \ |
| 691 | if (!__cpu_has(__c, X86_FEATURE_VMX)) \ |
| 692 | __reserved_bits |= X86_CR4_VMXE; \ |
| 693 | if (!__cpu_has(__c, X86_FEATURE_PCID)) \ |
| 694 | __reserved_bits |= X86_CR4_PCIDE; \ |
| 695 | if (!__cpu_has(__c, X86_FEATURE_LAM)) \ |
| 696 | __reserved_bits |= X86_CR4_LAM_SUP; \ |
| 697 | if (!__cpu_has(__c, X86_FEATURE_SHSTK) && \ |
| 698 | !__cpu_has(__c, X86_FEATURE_IBT)) \ |
| 699 | __reserved_bits |= X86_CR4_CET; \ |
| 700 | __reserved_bits; \ |
| 701 | }) |
| 702 | |
| 703 | int kvm_sev_es_mmio_write(struct kvm_vcpu *vcpu, gpa_t src, unsigned int bytes, |
| 704 | void *dst); |
| 705 | int kvm_sev_es_mmio_read(struct kvm_vcpu *vcpu, gpa_t src, unsigned int bytes, |
| 706 | void *dst); |
| 707 | int kvm_sev_es_string_io(struct kvm_vcpu *vcpu, unsigned int size, |
| 708 | unsigned int port, void *data, unsigned int count, |
| 709 | int in); |
| 710 | |
| 711 | static inline bool user_exit_on_hypercall(struct kvm *kvm, unsigned long hc_nr) |
| 712 | { |
| 713 | return kvm->arch.hypercall_exit_enabled & BIT(hc_nr); |
| 714 | } |
| 715 | |
| 716 | int ____kvm_emulate_hypercall(struct kvm_vcpu *vcpu, int cpl, |
| 717 | int (*complete_hypercall)(struct kvm_vcpu *)); |
| 718 | |
| 719 | #define __kvm_emulate_hypercall(_vcpu, cpl, complete_hypercall) \ |
| 720 | ({ \ |
| 721 | int __ret; \ |
| 722 | __ret = ____kvm_emulate_hypercall(_vcpu, cpl, complete_hypercall); \ |
| 723 | \ |
| 724 | if (__ret > 0) \ |
| 725 | __ret = complete_hypercall(_vcpu); \ |
| 726 | __ret; \ |
| 727 | }) |
| 728 | |
| 729 | int kvm_emulate_hypercall(struct kvm_vcpu *vcpu); |
| 730 | |
| 731 | #define CET_US_RESERVED_BITS GENMASK(9, 6) |
| 732 | #define CET_US_SHSTK_MASK_BITS GENMASK(1, 0) |
| 733 | #define CET_US_IBT_MASK_BITS (GENMASK_ULL(5, 2) | GENMASK_ULL(63, 10)) |
| 734 | #define CET_US_LEGACY_BITMAP_BASE(data) ((data) >> 12) |
| 735 | |
| 736 | static inline bool kvm_is_valid_u_s_cet(struct kvm_vcpu *vcpu, u64 data) |
| 737 | { |
| 738 | if (data & CET_US_RESERVED_BITS) |
| 739 | return false; |
| 740 | if (!guest_cpu_cap_has(vcpu, X86_FEATURE_SHSTK) && |
| 741 | (data & CET_US_SHSTK_MASK_BITS)) |
| 742 | return false; |
| 743 | if (!guest_cpu_cap_has(vcpu, X86_FEATURE_IBT) && |
| 744 | (data & CET_US_IBT_MASK_BITS)) |
| 745 | return false; |
| 746 | if (!IS_ALIGNED(CET_US_LEGACY_BITMAP_BASE(data), 4)) |
| 747 | return false; |
| 748 | /* IBT can be suppressed iff the TRACKER isn't WAIT_ENDBR. */ |
| 749 | if ((data & CET_SUPPRESS) && (data & CET_WAIT_ENDBR)) |
| 750 | return false; |
| 751 | |
| 752 | return true; |
| 753 | } |
| 754 | #endif |
| 755 | |