1 /* SPDX-License-Identifier: GPL-2.0 */ 2 #ifndef __KVM_X86_VMX_H 3 #define __KVM_X86_VMX_H 4 5 #include <linux/kvm_host.h> 6 7 #include <asm/kvm.h> 8 #include <asm/intel_pt.h> 9 #include <asm/perf_event.h> 10 11 #include "capabilities.h" 12 #include "../kvm_cache_regs.h" 13 #include "posted_intr.h" 14 #include "vmcs.h" 15 #include "vmx_ops.h" 16 #include "../cpuid.h" 17 #include "run_flags.h" 18 19 #define MSR_TYPE_R 1 20 #define MSR_TYPE_W 2 21 #define MSR_TYPE_RW 3 22 23 #define X2APIC_MSR(r) (APIC_BASE_MSR + ((r) >> 4)) 24 25 #ifdef CONFIG_X86_64 26 #define MAX_NR_USER_RETURN_MSRS 7 27 #else 28 #define MAX_NR_USER_RETURN_MSRS 4 29 #endif 30 31 #define MAX_NR_LOADSTORE_MSRS 8 32 33 struct vmx_msrs { 34 unsigned int nr; 35 struct vmx_msr_entry val[MAX_NR_LOADSTORE_MSRS]; 36 }; 37 38 struct vmx_uret_msr { 39 bool load_into_hardware; 40 u64 data; 41 u64 mask; 42 }; 43 44 enum segment_cache_field { 45 SEG_FIELD_SEL = 0, 46 SEG_FIELD_BASE = 1, 47 SEG_FIELD_LIMIT = 2, 48 SEG_FIELD_AR = 3, 49 50 SEG_FIELD_NR = 4 51 }; 52 53 #define RTIT_ADDR_RANGE 4 54 55 struct pt_ctx { 56 u64 ctl; 57 u64 status; 58 u64 output_base; 59 u64 output_mask; 60 u64 cr3_match; 61 u64 addr_a[RTIT_ADDR_RANGE]; 62 u64 addr_b[RTIT_ADDR_RANGE]; 63 }; 64 65 struct pt_desc { 66 u64 ctl_bitmask; 67 u32 num_address_ranges; 68 u32 caps[PT_CPUID_REGS_NUM * PT_CPUID_LEAVES]; 69 struct pt_ctx host; 70 struct pt_ctx guest; 71 }; 72 73 union vmx_exit_reason { 74 struct { 75 u32 basic : 16; 76 u32 reserved16 : 1; 77 u32 reserved17 : 1; 78 u32 reserved18 : 1; 79 u32 reserved19 : 1; 80 u32 reserved20 : 1; 81 u32 reserved21 : 1; 82 u32 reserved22 : 1; 83 u32 reserved23 : 1; 84 u32 reserved24 : 1; 85 u32 reserved25 : 1; 86 u32 bus_lock_detected : 1; 87 u32 enclave_mode : 1; 88 u32 smi_pending_mtf : 1; 89 u32 smi_from_vmx_root : 1; 90 u32 reserved30 : 1; 91 u32 failed_vmentry : 1; 92 }; 93 u32 full; 94 }; 95 96 struct lbr_desc { 97 /* Basic info about guest LBR records. */ 98 struct x86_pmu_lbr records; 99 100 /* 101 * Emulate LBR feature via passthrough LBR registers when the 102 * per-vcpu guest LBR event is scheduled on the current pcpu. 103 * 104 * The records may be inaccurate if the host reclaims the LBR. 105 */ 106 struct perf_event *event; 107 108 /* True if LBRs are marked as not intercepted in the MSR bitmap */ 109 bool msr_passthrough; 110 }; 111 112 /* 113 * The nested_vmx structure is part of vcpu_vmx, and holds information we need 114 * for correct emulation of VMX (i.e., nested VMX) on this vcpu. 115 */ 116 struct nested_vmx { 117 /* Has the level1 guest done vmxon? */ 118 bool vmxon; 119 gpa_t vmxon_ptr; 120 bool pml_full; 121 122 /* The guest-physical address of the current VMCS L1 keeps for L2 */ 123 gpa_t current_vmptr; 124 /* 125 * Cache of the guest's VMCS, existing outside of guest memory. 126 * Loaded from guest memory during VMPTRLD. Flushed to guest 127 * memory during VMCLEAR and VMPTRLD. 128 */ 129 struct vmcs12 *cached_vmcs12; 130 /* 131 * Cache of the guest's shadow VMCS, existing outside of guest 132 * memory. Loaded from guest memory during VM entry. Flushed 133 * to guest memory during VM exit. 134 */ 135 struct vmcs12 *cached_shadow_vmcs12; 136 137 /* 138 * GPA to HVA cache for accessing vmcs12->vmcs_link_pointer 139 */ 140 struct gfn_to_hva_cache shadow_vmcs12_cache; 141 142 /* 143 * GPA to HVA cache for VMCS12 144 */ 145 struct gfn_to_hva_cache vmcs12_cache; 146 147 /* 148 * Indicates if the shadow vmcs or enlightened vmcs must be updated 149 * with the data held by struct vmcs12. 150 */ 151 bool need_vmcs12_to_shadow_sync; 152 bool dirty_vmcs12; 153 154 /* 155 * Indicates whether MSR bitmap for L2 needs to be rebuilt due to 156 * changes in MSR bitmap for L1 or switching to a different L2. Note, 157 * this flag can only be used reliably in conjunction with a paravirt L1 158 * which informs L0 whether any changes to MSR bitmap for L2 were done 159 * on its side. 160 */ 161 bool force_msr_bitmap_recalc; 162 163 /* 164 * Indicates lazily loaded guest state has not yet been decached from 165 * vmcs02. 166 */ 167 bool need_sync_vmcs02_to_vmcs12_rare; 168 169 /* 170 * vmcs02 has been initialized, i.e. state that is constant for 171 * vmcs02 has been written to the backing VMCS. Initialization 172 * is delayed until L1 actually attempts to run a nested VM. 173 */ 174 bool vmcs02_initialized; 175 176 bool change_vmcs01_virtual_apic_mode; 177 bool reload_vmcs01_apic_access_page; 178 bool update_vmcs01_cpu_dirty_logging; 179 bool update_vmcs01_apicv_status; 180 181 /* 182 * Enlightened VMCS has been enabled. It does not mean that L1 has to 183 * use it. However, VMX features available to L1 will be limited based 184 * on what the enlightened VMCS supports. 185 */ 186 bool enlightened_vmcs_enabled; 187 188 /* L2 must run next, and mustn't decide to exit to L1. */ 189 bool nested_run_pending; 190 191 /* Pending MTF VM-exit into L1. */ 192 bool mtf_pending; 193 194 struct loaded_vmcs vmcs02; 195 196 /* 197 * Guest pages referred to in the vmcs02 with host-physical 198 * pointers, so we must keep them pinned while L2 runs. 199 */ 200 struct kvm_host_map apic_access_page_map; 201 struct kvm_host_map virtual_apic_map; 202 struct kvm_host_map pi_desc_map; 203 204 struct kvm_host_map msr_bitmap_map; 205 206 struct pi_desc *pi_desc; 207 bool pi_pending; 208 u16 posted_intr_nv; 209 210 struct hrtimer preemption_timer; 211 u64 preemption_timer_deadline; 212 bool has_preemption_timer_deadline; 213 bool preemption_timer_expired; 214 215 /* 216 * Used to snapshot MSRs that are conditionally loaded on VM-Enter in 217 * order to propagate the guest's pre-VM-Enter value into vmcs02. For 218 * emulation of VMLAUNCH/VMRESUME, the snapshot will be of L1's value. 219 * For KVM_SET_NESTED_STATE, the snapshot is of L2's value, _if_ 220 * userspace restores MSRs before nested state. If userspace restores 221 * MSRs after nested state, the snapshot holds garbage, but KVM can't 222 * detect that, and the garbage value in vmcs02 will be overwritten by 223 * MSR restoration in any case. 224 */ 225 u64 pre_vmenter_debugctl; 226 u64 pre_vmenter_bndcfgs; 227 228 /* to migrate it to L1 if L2 writes to L1's CR8 directly */ 229 int l1_tpr_threshold; 230 231 u16 vpid02; 232 u16 last_vpid; 233 234 struct nested_vmx_msrs msrs; 235 236 /* SMM related state */ 237 struct { 238 /* in VMX operation on SMM entry? */ 239 bool vmxon; 240 /* in guest mode on SMM entry? */ 241 bool guest_mode; 242 } smm; 243 244 #ifdef CONFIG_KVM_HYPERV 245 gpa_t hv_evmcs_vmptr; 246 struct kvm_host_map hv_evmcs_map; 247 struct hv_enlightened_vmcs *hv_evmcs; 248 #endif 249 }; 250 251 struct vcpu_vmx { 252 struct kvm_vcpu vcpu; 253 u8 fail; 254 u8 x2apic_msr_bitmap_mode; 255 256 /* 257 * If true, host state has been stored in vmx->loaded_vmcs for 258 * the CPU registers that only need to be switched when transitioning 259 * to/from the kernel, and the registers have been loaded with guest 260 * values. If false, host state is loaded in the CPU registers 261 * and vmx->loaded_vmcs->host_state is invalid. 262 */ 263 bool guest_state_loaded; 264 265 unsigned long exit_qualification; 266 u32 exit_intr_info; 267 u32 idt_vectoring_info; 268 ulong rflags; 269 270 /* 271 * User return MSRs are always emulated when enabled in the guest, but 272 * only loaded into hardware when necessary, e.g. SYSCALL #UDs outside 273 * of 64-bit mode or if EFER.SCE=1, thus the SYSCALL MSRs don't need to 274 * be loaded into hardware if those conditions aren't met. 275 */ 276 struct vmx_uret_msr guest_uret_msrs[MAX_NR_USER_RETURN_MSRS]; 277 bool guest_uret_msrs_loaded; 278 #ifdef CONFIG_X86_64 279 u64 msr_host_kernel_gs_base; 280 u64 msr_guest_kernel_gs_base; 281 #endif 282 283 u64 spec_ctrl; 284 u32 msr_ia32_umwait_control; 285 286 /* 287 * loaded_vmcs points to the VMCS currently used in this vcpu. For a 288 * non-nested (L1) guest, it always points to vmcs01. For a nested 289 * guest (L2), it points to a different VMCS. 290 */ 291 struct loaded_vmcs vmcs01; 292 struct loaded_vmcs *loaded_vmcs; 293 294 struct msr_autoload { 295 struct vmx_msrs guest; 296 struct vmx_msrs host; 297 } msr_autoload; 298 299 struct msr_autostore { 300 struct vmx_msrs guest; 301 } msr_autostore; 302 303 struct { 304 int vm86_active; 305 ulong save_rflags; 306 struct kvm_segment segs[8]; 307 } rmode; 308 struct { 309 u32 bitmask; /* 4 bits per segment (1 bit per field) */ 310 struct kvm_save_segment { 311 u16 selector; 312 unsigned long base; 313 u32 limit; 314 u32 ar; 315 } seg[8]; 316 } segment_cache; 317 int vpid; 318 bool emulation_required; 319 320 union vmx_exit_reason exit_reason; 321 322 /* Posted interrupt descriptor */ 323 struct pi_desc pi_desc; 324 325 /* Used if this vCPU is waiting for PI notification wakeup. */ 326 struct list_head pi_wakeup_list; 327 328 /* Support for a guest hypervisor (nested VMX) */ 329 struct nested_vmx nested; 330 331 /* Dynamic PLE window. */ 332 unsigned int ple_window; 333 bool ple_window_dirty; 334 335 /* Support for PML */ 336 #define PML_ENTITY_NUM 512 337 struct page *pml_pg; 338 339 /* apic deadline value in host tsc */ 340 u64 hv_deadline_tsc; 341 342 unsigned long host_debugctlmsr; 343 344 /* 345 * Only bits masked by msr_ia32_feature_control_valid_bits can be set in 346 * msr_ia32_feature_control. FEAT_CTL_LOCKED is always included 347 * in msr_ia32_feature_control_valid_bits. 348 */ 349 u64 msr_ia32_feature_control; 350 u64 msr_ia32_feature_control_valid_bits; 351 /* SGX Launch Control public key hash */ 352 u64 msr_ia32_sgxlepubkeyhash[4]; 353 u64 msr_ia32_mcu_opt_ctrl; 354 bool disable_fb_clear; 355 356 struct pt_desc pt_desc; 357 struct lbr_desc lbr_desc; 358 359 /* Save desired MSR intercept (read: pass-through) state */ 360 #define MAX_POSSIBLE_PASSTHROUGH_MSRS 16 361 struct { 362 DECLARE_BITMAP(read, MAX_POSSIBLE_PASSTHROUGH_MSRS); 363 DECLARE_BITMAP(write, MAX_POSSIBLE_PASSTHROUGH_MSRS); 364 } shadow_msr_intercept; 365 }; 366 367 struct kvm_vmx { 368 struct kvm kvm; 369 370 unsigned int tss_addr; 371 bool ept_identity_pagetable_done; 372 gpa_t ept_identity_map_addr; 373 /* Posted Interrupt Descriptor (PID) table for IPI virtualization */ 374 u64 *pid_table; 375 }; 376 377 void vmx_vcpu_load_vmcs(struct kvm_vcpu *vcpu, int cpu, 378 struct loaded_vmcs *buddy); 379 int allocate_vpid(void); 380 void free_vpid(int vpid); 381 void vmx_set_constant_host_state(struct vcpu_vmx *vmx); 382 void vmx_prepare_switch_to_guest(struct kvm_vcpu *vcpu); 383 void vmx_set_host_fs_gs(struct vmcs_host_state *host, u16 fs_sel, u16 gs_sel, 384 unsigned long fs_base, unsigned long gs_base); 385 int vmx_get_cpl(struct kvm_vcpu *vcpu); 386 bool vmx_emulation_required(struct kvm_vcpu *vcpu); 387 unsigned long vmx_get_rflags(struct kvm_vcpu *vcpu); 388 void vmx_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags); 389 u32 vmx_get_interrupt_shadow(struct kvm_vcpu *vcpu); 390 void vmx_set_interrupt_shadow(struct kvm_vcpu *vcpu, int mask); 391 int vmx_set_efer(struct kvm_vcpu *vcpu, u64 efer); 392 void vmx_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0); 393 void vmx_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4); 394 void set_cr4_guest_host_mask(struct vcpu_vmx *vmx); 395 void ept_save_pdptrs(struct kvm_vcpu *vcpu); 396 void vmx_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg); 397 void __vmx_set_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg); 398 u64 construct_eptp(struct kvm_vcpu *vcpu, hpa_t root_hpa, int root_level); 399 400 bool vmx_guest_inject_ac(struct kvm_vcpu *vcpu); 401 void vmx_update_exception_bitmap(struct kvm_vcpu *vcpu); 402 bool vmx_nmi_blocked(struct kvm_vcpu *vcpu); 403 bool vmx_interrupt_blocked(struct kvm_vcpu *vcpu); 404 bool vmx_get_nmi_mask(struct kvm_vcpu *vcpu); 405 void vmx_set_nmi_mask(struct kvm_vcpu *vcpu, bool masked); 406 void vmx_set_virtual_apic_mode(struct kvm_vcpu *vcpu); 407 struct vmx_uret_msr *vmx_find_uret_msr(struct vcpu_vmx *vmx, u32 msr); 408 void pt_update_intercept_for_msr(struct kvm_vcpu *vcpu); 409 void vmx_update_host_rsp(struct vcpu_vmx *vmx, unsigned long host_rsp); 410 void vmx_spec_ctrl_restore_host(struct vcpu_vmx *vmx, unsigned int flags); 411 unsigned int __vmx_vcpu_run_flags(struct vcpu_vmx *vmx); 412 bool __vmx_vcpu_run(struct vcpu_vmx *vmx, unsigned long *regs, 413 unsigned int flags); 414 int vmx_find_loadstore_msr_slot(struct vmx_msrs *m, u32 msr); 415 void vmx_ept_load_pdptrs(struct kvm_vcpu *vcpu); 416 417 void vmx_disable_intercept_for_msr(struct kvm_vcpu *vcpu, u32 msr, int type); 418 void vmx_enable_intercept_for_msr(struct kvm_vcpu *vcpu, u32 msr, int type); 419 420 u64 vmx_get_l2_tsc_offset(struct kvm_vcpu *vcpu); 421 u64 vmx_get_l2_tsc_multiplier(struct kvm_vcpu *vcpu); 422 423 gva_t vmx_get_untagged_addr(struct kvm_vcpu *vcpu, gva_t gva, unsigned int flags); 424 425 static inline void vmx_set_intercept_for_msr(struct kvm_vcpu *vcpu, u32 msr, 426 int type, bool value) 427 { 428 if (value) 429 vmx_enable_intercept_for_msr(vcpu, msr, type); 430 else 431 vmx_disable_intercept_for_msr(vcpu, msr, type); 432 } 433 434 void vmx_update_cpu_dirty_logging(struct kvm_vcpu *vcpu); 435 436 /* 437 * Note, early Intel manuals have the write-low and read-high bitmap offsets 438 * the wrong way round. The bitmaps control MSRs 0x00000000-0x00001fff and 439 * 0xc0000000-0xc0001fff. The former (low) uses bytes 0-0x3ff for reads and 440 * 0x800-0xbff for writes. The latter (high) uses 0x400-0x7ff for reads and 441 * 0xc00-0xfff for writes. MSRs not covered by either of the ranges always 442 * VM-Exit. 443 */ 444 #define __BUILD_VMX_MSR_BITMAP_HELPER(rtype, action, bitop, access, base) \ 445 static inline rtype vmx_##action##_msr_bitmap_##access(unsigned long *bitmap, \ 446 u32 msr) \ 447 { \ 448 int f = sizeof(unsigned long); \ 449 \ 450 if (msr <= 0x1fff) \ 451 return bitop##_bit(msr, bitmap + base / f); \ 452 else if ((msr >= 0xc0000000) && (msr <= 0xc0001fff)) \ 453 return bitop##_bit(msr & 0x1fff, bitmap + (base + 0x400) / f); \ 454 return (rtype)true; \ 455 } 456 #define BUILD_VMX_MSR_BITMAP_HELPERS(ret_type, action, bitop) \ 457 __BUILD_VMX_MSR_BITMAP_HELPER(ret_type, action, bitop, read, 0x0) \ 458 __BUILD_VMX_MSR_BITMAP_HELPER(ret_type, action, bitop, write, 0x800) 459 460 BUILD_VMX_MSR_BITMAP_HELPERS(bool, test, test) 461 BUILD_VMX_MSR_BITMAP_HELPERS(void, clear, __clear) 462 BUILD_VMX_MSR_BITMAP_HELPERS(void, set, __set) 463 464 static inline u8 vmx_get_rvi(void) 465 { 466 return vmcs_read16(GUEST_INTR_STATUS) & 0xff; 467 } 468 469 #define __KVM_REQUIRED_VMX_VM_ENTRY_CONTROLS \ 470 (VM_ENTRY_LOAD_DEBUG_CONTROLS) 471 #ifdef CONFIG_X86_64 472 #define KVM_REQUIRED_VMX_VM_ENTRY_CONTROLS \ 473 (__KVM_REQUIRED_VMX_VM_ENTRY_CONTROLS | \ 474 VM_ENTRY_IA32E_MODE) 475 #else 476 #define KVM_REQUIRED_VMX_VM_ENTRY_CONTROLS \ 477 __KVM_REQUIRED_VMX_VM_ENTRY_CONTROLS 478 #endif 479 #define KVM_OPTIONAL_VMX_VM_ENTRY_CONTROLS \ 480 (VM_ENTRY_LOAD_IA32_PERF_GLOBAL_CTRL | \ 481 VM_ENTRY_LOAD_IA32_PAT | \ 482 VM_ENTRY_LOAD_IA32_EFER | \ 483 VM_ENTRY_LOAD_BNDCFGS | \ 484 VM_ENTRY_PT_CONCEAL_PIP | \ 485 VM_ENTRY_LOAD_IA32_RTIT_CTL) 486 487 #define __KVM_REQUIRED_VMX_VM_EXIT_CONTROLS \ 488 (VM_EXIT_SAVE_DEBUG_CONTROLS | \ 489 VM_EXIT_ACK_INTR_ON_EXIT) 490 #ifdef CONFIG_X86_64 491 #define KVM_REQUIRED_VMX_VM_EXIT_CONTROLS \ 492 (__KVM_REQUIRED_VMX_VM_EXIT_CONTROLS | \ 493 VM_EXIT_HOST_ADDR_SPACE_SIZE) 494 #else 495 #define KVM_REQUIRED_VMX_VM_EXIT_CONTROLS \ 496 __KVM_REQUIRED_VMX_VM_EXIT_CONTROLS 497 #endif 498 #define KVM_OPTIONAL_VMX_VM_EXIT_CONTROLS \ 499 (VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL | \ 500 VM_EXIT_SAVE_IA32_PAT | \ 501 VM_EXIT_LOAD_IA32_PAT | \ 502 VM_EXIT_SAVE_IA32_EFER | \ 503 VM_EXIT_SAVE_VMX_PREEMPTION_TIMER | \ 504 VM_EXIT_LOAD_IA32_EFER | \ 505 VM_EXIT_CLEAR_BNDCFGS | \ 506 VM_EXIT_PT_CONCEAL_PIP | \ 507 VM_EXIT_CLEAR_IA32_RTIT_CTL) 508 509 #define KVM_REQUIRED_VMX_PIN_BASED_VM_EXEC_CONTROL \ 510 (PIN_BASED_EXT_INTR_MASK | \ 511 PIN_BASED_NMI_EXITING) 512 #define KVM_OPTIONAL_VMX_PIN_BASED_VM_EXEC_CONTROL \ 513 (PIN_BASED_VIRTUAL_NMIS | \ 514 PIN_BASED_POSTED_INTR | \ 515 PIN_BASED_VMX_PREEMPTION_TIMER) 516 517 #define __KVM_REQUIRED_VMX_CPU_BASED_VM_EXEC_CONTROL \ 518 (CPU_BASED_HLT_EXITING | \ 519 CPU_BASED_CR3_LOAD_EXITING | \ 520 CPU_BASED_CR3_STORE_EXITING | \ 521 CPU_BASED_UNCOND_IO_EXITING | \ 522 CPU_BASED_MOV_DR_EXITING | \ 523 CPU_BASED_USE_TSC_OFFSETTING | \ 524 CPU_BASED_MWAIT_EXITING | \ 525 CPU_BASED_MONITOR_EXITING | \ 526 CPU_BASED_INVLPG_EXITING | \ 527 CPU_BASED_RDPMC_EXITING | \ 528 CPU_BASED_INTR_WINDOW_EXITING) 529 530 #ifdef CONFIG_X86_64 531 #define KVM_REQUIRED_VMX_CPU_BASED_VM_EXEC_CONTROL \ 532 (__KVM_REQUIRED_VMX_CPU_BASED_VM_EXEC_CONTROL | \ 533 CPU_BASED_CR8_LOAD_EXITING | \ 534 CPU_BASED_CR8_STORE_EXITING) 535 #else 536 #define KVM_REQUIRED_VMX_CPU_BASED_VM_EXEC_CONTROL \ 537 __KVM_REQUIRED_VMX_CPU_BASED_VM_EXEC_CONTROL 538 #endif 539 540 #define KVM_OPTIONAL_VMX_CPU_BASED_VM_EXEC_CONTROL \ 541 (CPU_BASED_RDTSC_EXITING | \ 542 CPU_BASED_TPR_SHADOW | \ 543 CPU_BASED_USE_IO_BITMAPS | \ 544 CPU_BASED_MONITOR_TRAP_FLAG | \ 545 CPU_BASED_USE_MSR_BITMAPS | \ 546 CPU_BASED_NMI_WINDOW_EXITING | \ 547 CPU_BASED_PAUSE_EXITING | \ 548 CPU_BASED_ACTIVATE_SECONDARY_CONTROLS | \ 549 CPU_BASED_ACTIVATE_TERTIARY_CONTROLS) 550 551 #define KVM_REQUIRED_VMX_SECONDARY_VM_EXEC_CONTROL 0 552 #define KVM_OPTIONAL_VMX_SECONDARY_VM_EXEC_CONTROL \ 553 (SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES | \ 554 SECONDARY_EXEC_VIRTUALIZE_X2APIC_MODE | \ 555 SECONDARY_EXEC_WBINVD_EXITING | \ 556 SECONDARY_EXEC_ENABLE_VPID | \ 557 SECONDARY_EXEC_ENABLE_EPT | \ 558 SECONDARY_EXEC_UNRESTRICTED_GUEST | \ 559 SECONDARY_EXEC_PAUSE_LOOP_EXITING | \ 560 SECONDARY_EXEC_DESC | \ 561 SECONDARY_EXEC_ENABLE_RDTSCP | \ 562 SECONDARY_EXEC_ENABLE_INVPCID | \ 563 SECONDARY_EXEC_APIC_REGISTER_VIRT | \ 564 SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY | \ 565 SECONDARY_EXEC_SHADOW_VMCS | \ 566 SECONDARY_EXEC_ENABLE_XSAVES | \ 567 SECONDARY_EXEC_RDSEED_EXITING | \ 568 SECONDARY_EXEC_RDRAND_EXITING | \ 569 SECONDARY_EXEC_ENABLE_PML | \ 570 SECONDARY_EXEC_TSC_SCALING | \ 571 SECONDARY_EXEC_ENABLE_USR_WAIT_PAUSE | \ 572 SECONDARY_EXEC_PT_USE_GPA | \ 573 SECONDARY_EXEC_PT_CONCEAL_VMX | \ 574 SECONDARY_EXEC_ENABLE_VMFUNC | \ 575 SECONDARY_EXEC_BUS_LOCK_DETECTION | \ 576 SECONDARY_EXEC_NOTIFY_VM_EXITING | \ 577 SECONDARY_EXEC_ENCLS_EXITING) 578 579 #define KVM_REQUIRED_VMX_TERTIARY_VM_EXEC_CONTROL 0 580 #define KVM_OPTIONAL_VMX_TERTIARY_VM_EXEC_CONTROL \ 581 (TERTIARY_EXEC_IPI_VIRT) 582 583 #define BUILD_CONTROLS_SHADOW(lname, uname, bits) \ 584 static inline void lname##_controls_set(struct vcpu_vmx *vmx, u##bits val) \ 585 { \ 586 if (vmx->loaded_vmcs->controls_shadow.lname != val) { \ 587 vmcs_write##bits(uname, val); \ 588 vmx->loaded_vmcs->controls_shadow.lname = val; \ 589 } \ 590 } \ 591 static inline u##bits __##lname##_controls_get(struct loaded_vmcs *vmcs) \ 592 { \ 593 return vmcs->controls_shadow.lname; \ 594 } \ 595 static inline u##bits lname##_controls_get(struct vcpu_vmx *vmx) \ 596 { \ 597 return __##lname##_controls_get(vmx->loaded_vmcs); \ 598 } \ 599 static __always_inline void lname##_controls_setbit(struct vcpu_vmx *vmx, u##bits val) \ 600 { \ 601 BUILD_BUG_ON(!(val & (KVM_REQUIRED_VMX_##uname | KVM_OPTIONAL_VMX_##uname))); \ 602 lname##_controls_set(vmx, lname##_controls_get(vmx) | val); \ 603 } \ 604 static __always_inline void lname##_controls_clearbit(struct vcpu_vmx *vmx, u##bits val) \ 605 { \ 606 BUILD_BUG_ON(!(val & (KVM_REQUIRED_VMX_##uname | KVM_OPTIONAL_VMX_##uname))); \ 607 lname##_controls_set(vmx, lname##_controls_get(vmx) & ~val); \ 608 } 609 BUILD_CONTROLS_SHADOW(vm_entry, VM_ENTRY_CONTROLS, 32) 610 BUILD_CONTROLS_SHADOW(vm_exit, VM_EXIT_CONTROLS, 32) 611 BUILD_CONTROLS_SHADOW(pin, PIN_BASED_VM_EXEC_CONTROL, 32) 612 BUILD_CONTROLS_SHADOW(exec, CPU_BASED_VM_EXEC_CONTROL, 32) 613 BUILD_CONTROLS_SHADOW(secondary_exec, SECONDARY_VM_EXEC_CONTROL, 32) 614 BUILD_CONTROLS_SHADOW(tertiary_exec, TERTIARY_VM_EXEC_CONTROL, 64) 615 616 /* 617 * VMX_REGS_LAZY_LOAD_SET - The set of registers that will be updated in the 618 * cache on demand. Other registers not listed here are synced to 619 * the cache immediately after VM-Exit. 620 */ 621 #define VMX_REGS_LAZY_LOAD_SET ((1 << VCPU_REGS_RIP) | \ 622 (1 << VCPU_REGS_RSP) | \ 623 (1 << VCPU_EXREG_RFLAGS) | \ 624 (1 << VCPU_EXREG_PDPTR) | \ 625 (1 << VCPU_EXREG_SEGMENTS) | \ 626 (1 << VCPU_EXREG_CR0) | \ 627 (1 << VCPU_EXREG_CR3) | \ 628 (1 << VCPU_EXREG_CR4) | \ 629 (1 << VCPU_EXREG_EXIT_INFO_1) | \ 630 (1 << VCPU_EXREG_EXIT_INFO_2)) 631 632 static inline unsigned long vmx_l1_guest_owned_cr0_bits(void) 633 { 634 unsigned long bits = KVM_POSSIBLE_CR0_GUEST_BITS; 635 636 /* 637 * CR0.WP needs to be intercepted when KVM is shadowing legacy paging 638 * in order to construct shadow PTEs with the correct protections. 639 * Note! CR0.WP technically can be passed through to the guest if 640 * paging is disabled, but checking CR0.PG would generate a cyclical 641 * dependency of sorts due to forcing the caller to ensure CR0 holds 642 * the correct value prior to determining which CR0 bits can be owned 643 * by L1. Keep it simple and limit the optimization to EPT. 644 */ 645 if (!enable_ept) 646 bits &= ~X86_CR0_WP; 647 return bits; 648 } 649 650 static __always_inline struct kvm_vmx *to_kvm_vmx(struct kvm *kvm) 651 { 652 return container_of(kvm, struct kvm_vmx, kvm); 653 } 654 655 static __always_inline struct vcpu_vmx *to_vmx(struct kvm_vcpu *vcpu) 656 { 657 return container_of(vcpu, struct vcpu_vmx, vcpu); 658 } 659 660 static inline struct lbr_desc *vcpu_to_lbr_desc(struct kvm_vcpu *vcpu) 661 { 662 return &to_vmx(vcpu)->lbr_desc; 663 } 664 665 static inline struct x86_pmu_lbr *vcpu_to_lbr_records(struct kvm_vcpu *vcpu) 666 { 667 return &vcpu_to_lbr_desc(vcpu)->records; 668 } 669 670 static inline bool intel_pmu_lbr_is_enabled(struct kvm_vcpu *vcpu) 671 { 672 return !!vcpu_to_lbr_records(vcpu)->nr; 673 } 674 675 void intel_pmu_cross_mapped_check(struct kvm_pmu *pmu); 676 int intel_pmu_create_guest_lbr_event(struct kvm_vcpu *vcpu); 677 void vmx_passthrough_lbr_msrs(struct kvm_vcpu *vcpu); 678 679 static __always_inline unsigned long vmx_get_exit_qual(struct kvm_vcpu *vcpu) 680 { 681 struct vcpu_vmx *vmx = to_vmx(vcpu); 682 683 if (!kvm_register_test_and_mark_available(vcpu, VCPU_EXREG_EXIT_INFO_1)) 684 vmx->exit_qualification = vmcs_readl(EXIT_QUALIFICATION); 685 686 return vmx->exit_qualification; 687 } 688 689 static __always_inline u32 vmx_get_intr_info(struct kvm_vcpu *vcpu) 690 { 691 struct vcpu_vmx *vmx = to_vmx(vcpu); 692 693 if (!kvm_register_test_and_mark_available(vcpu, VCPU_EXREG_EXIT_INFO_2)) 694 vmx->exit_intr_info = vmcs_read32(VM_EXIT_INTR_INFO); 695 696 return vmx->exit_intr_info; 697 } 698 699 struct vmcs *alloc_vmcs_cpu(bool shadow, int cpu, gfp_t flags); 700 void free_vmcs(struct vmcs *vmcs); 701 int alloc_loaded_vmcs(struct loaded_vmcs *loaded_vmcs); 702 void free_loaded_vmcs(struct loaded_vmcs *loaded_vmcs); 703 void loaded_vmcs_clear(struct loaded_vmcs *loaded_vmcs); 704 705 static inline struct vmcs *alloc_vmcs(bool shadow) 706 { 707 return alloc_vmcs_cpu(shadow, raw_smp_processor_id(), 708 GFP_KERNEL_ACCOUNT); 709 } 710 711 static inline bool vmx_has_waitpkg(struct vcpu_vmx *vmx) 712 { 713 return secondary_exec_controls_get(vmx) & 714 SECONDARY_EXEC_ENABLE_USR_WAIT_PAUSE; 715 } 716 717 static inline bool vmx_need_pf_intercept(struct kvm_vcpu *vcpu) 718 { 719 if (!enable_ept) 720 return true; 721 722 return allow_smaller_maxphyaddr && cpuid_maxphyaddr(vcpu) < boot_cpu_data.x86_phys_bits; 723 } 724 725 static inline bool is_unrestricted_guest(struct kvm_vcpu *vcpu) 726 { 727 return enable_unrestricted_guest && (!is_guest_mode(vcpu) || 728 (secondary_exec_controls_get(to_vmx(vcpu)) & 729 SECONDARY_EXEC_UNRESTRICTED_GUEST)); 730 } 731 732 bool __vmx_guest_state_valid(struct kvm_vcpu *vcpu); 733 static inline bool vmx_guest_state_valid(struct kvm_vcpu *vcpu) 734 { 735 return is_unrestricted_guest(vcpu) || __vmx_guest_state_valid(vcpu); 736 } 737 738 void dump_vmcs(struct kvm_vcpu *vcpu); 739 740 static inline int vmx_get_instr_info_reg2(u32 vmx_instr_info) 741 { 742 return (vmx_instr_info >> 28) & 0xf; 743 } 744 745 static inline bool vmx_can_use_ipiv(struct kvm_vcpu *vcpu) 746 { 747 return lapic_in_kernel(vcpu) && enable_ipiv; 748 } 749 750 #endif /* __KVM_X86_VMX_H */ 751