1 /* SPDX-License-Identifier: GPL-2.0-only */ 2 /* 3 * Copyright (C) 2012,2013 - ARM Ltd 4 * Author: Marc Zyngier <marc.zyngier@arm.com> 5 * 6 * Derived from arch/arm/include/asm/kvm_host.h: 7 * Copyright (C) 2012 - Virtual Open Systems and Columbia University 8 * Author: Christoffer Dall <c.dall@virtualopensystems.com> 9 */ 10 11 #ifndef __ARM64_KVM_HOST_H__ 12 #define __ARM64_KVM_HOST_H__ 13 14 #include <linux/bitmap.h> 15 #include <linux/types.h> 16 #include <linux/jump_label.h> 17 #include <linux/kvm_types.h> 18 #include <linux/percpu.h> 19 #include <asm/arch_gicv3.h> 20 #include <asm/barrier.h> 21 #include <asm/cpufeature.h> 22 #include <asm/cputype.h> 23 #include <asm/daifflags.h> 24 #include <asm/fpsimd.h> 25 #include <asm/kvm.h> 26 #include <asm/kvm_asm.h> 27 #include <asm/thread_info.h> 28 29 #define __KVM_HAVE_ARCH_INTC_INITIALIZED 30 31 #define KVM_USER_MEM_SLOTS 512 32 #define KVM_HALT_POLL_NS_DEFAULT 500000 33 34 #include <kvm/arm_vgic.h> 35 #include <kvm/arm_arch_timer.h> 36 #include <kvm/arm_pmu.h> 37 38 #define KVM_MAX_VCPUS VGIC_V3_MAX_CPUS 39 40 #define KVM_VCPU_MAX_FEATURES 7 41 42 #define KVM_REQ_SLEEP \ 43 KVM_ARCH_REQ_FLAGS(0, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP) 44 #define KVM_REQ_IRQ_PENDING KVM_ARCH_REQ(1) 45 #define KVM_REQ_VCPU_RESET KVM_ARCH_REQ(2) 46 #define KVM_REQ_RECORD_STEAL KVM_ARCH_REQ(3) 47 48 DECLARE_STATIC_KEY_FALSE(userspace_irqchip_in_use); 49 50 extern unsigned int kvm_sve_max_vl; 51 int kvm_arm_init_sve(void); 52 53 int __attribute_const__ kvm_target_cpu(void); 54 int kvm_reset_vcpu(struct kvm_vcpu *vcpu); 55 void kvm_arm_vcpu_destroy(struct kvm_vcpu *vcpu); 56 int kvm_arch_vm_ioctl_check_extension(struct kvm *kvm, long ext); 57 void __extended_idmap_trampoline(phys_addr_t boot_pgd, phys_addr_t idmap_start); 58 59 struct kvm_vmid { 60 /* The VMID generation used for the virt. memory system */ 61 u64 vmid_gen; 62 u32 vmid; 63 }; 64 65 struct kvm_arch { 66 struct kvm_vmid vmid; 67 68 /* stage2 entry level table */ 69 pgd_t *pgd; 70 phys_addr_t pgd_phys; 71 72 /* VTCR_EL2 value for this VM */ 73 u64 vtcr; 74 75 /* The last vcpu id that ran on each physical CPU */ 76 int __percpu *last_vcpu_ran; 77 78 /* The maximum number of vCPUs depends on the used GIC model */ 79 int max_vcpus; 80 81 /* Interrupt controller */ 82 struct vgic_dist vgic; 83 84 /* Mandated version of PSCI */ 85 u32 psci_version; 86 87 /* 88 * If we encounter a data abort without valid instruction syndrome 89 * information, report this to user space. User space can (and 90 * should) opt in to this feature if KVM_CAP_ARM_NISV_TO_USER is 91 * supported. 92 */ 93 bool return_nisv_io_abort_to_user; 94 }; 95 96 #define KVM_NR_MEM_OBJS 40 97 98 /* 99 * We don't want allocation failures within the mmu code, so we preallocate 100 * enough memory for a single page fault in a cache. 101 */ 102 struct kvm_mmu_memory_cache { 103 int nobjs; 104 void *objects[KVM_NR_MEM_OBJS]; 105 }; 106 107 struct kvm_vcpu_fault_info { 108 u32 esr_el2; /* Hyp Syndrom Register */ 109 u64 far_el2; /* Hyp Fault Address Register */ 110 u64 hpfar_el2; /* Hyp IPA Fault Address Register */ 111 u64 disr_el1; /* Deferred [SError] Status Register */ 112 }; 113 114 /* 115 * 0 is reserved as an invalid value. 116 * Order should be kept in sync with the save/restore code. 117 */ 118 enum vcpu_sysreg { 119 __INVALID_SYSREG__, 120 MPIDR_EL1, /* MultiProcessor Affinity Register */ 121 CSSELR_EL1, /* Cache Size Selection Register */ 122 SCTLR_EL1, /* System Control Register */ 123 ACTLR_EL1, /* Auxiliary Control Register */ 124 CPACR_EL1, /* Coprocessor Access Control */ 125 ZCR_EL1, /* SVE Control */ 126 TTBR0_EL1, /* Translation Table Base Register 0 */ 127 TTBR1_EL1, /* Translation Table Base Register 1 */ 128 TCR_EL1, /* Translation Control Register */ 129 ESR_EL1, /* Exception Syndrome Register */ 130 AFSR0_EL1, /* Auxiliary Fault Status Register 0 */ 131 AFSR1_EL1, /* Auxiliary Fault Status Register 1 */ 132 FAR_EL1, /* Fault Address Register */ 133 MAIR_EL1, /* Memory Attribute Indirection Register */ 134 VBAR_EL1, /* Vector Base Address Register */ 135 CONTEXTIDR_EL1, /* Context ID Register */ 136 TPIDR_EL0, /* Thread ID, User R/W */ 137 TPIDRRO_EL0, /* Thread ID, User R/O */ 138 TPIDR_EL1, /* Thread ID, Privileged */ 139 AMAIR_EL1, /* Aux Memory Attribute Indirection Register */ 140 CNTKCTL_EL1, /* Timer Control Register (EL1) */ 141 PAR_EL1, /* Physical Address Register */ 142 MDSCR_EL1, /* Monitor Debug System Control Register */ 143 MDCCINT_EL1, /* Monitor Debug Comms Channel Interrupt Enable Reg */ 144 DISR_EL1, /* Deferred Interrupt Status Register */ 145 146 /* Performance Monitors Registers */ 147 PMCR_EL0, /* Control Register */ 148 PMSELR_EL0, /* Event Counter Selection Register */ 149 PMEVCNTR0_EL0, /* Event Counter Register (0-30) */ 150 PMEVCNTR30_EL0 = PMEVCNTR0_EL0 + 30, 151 PMCCNTR_EL0, /* Cycle Counter Register */ 152 PMEVTYPER0_EL0, /* Event Type Register (0-30) */ 153 PMEVTYPER30_EL0 = PMEVTYPER0_EL0 + 30, 154 PMCCFILTR_EL0, /* Cycle Count Filter Register */ 155 PMCNTENSET_EL0, /* Count Enable Set Register */ 156 PMINTENSET_EL1, /* Interrupt Enable Set Register */ 157 PMOVSSET_EL0, /* Overflow Flag Status Set Register */ 158 PMSWINC_EL0, /* Software Increment Register */ 159 PMUSERENR_EL0, /* User Enable Register */ 160 161 /* Pointer Authentication Registers in a strict increasing order. */ 162 APIAKEYLO_EL1, 163 APIAKEYHI_EL1, 164 APIBKEYLO_EL1, 165 APIBKEYHI_EL1, 166 APDAKEYLO_EL1, 167 APDAKEYHI_EL1, 168 APDBKEYLO_EL1, 169 APDBKEYHI_EL1, 170 APGAKEYLO_EL1, 171 APGAKEYHI_EL1, 172 173 /* 32bit specific registers. Keep them at the end of the range */ 174 DACR32_EL2, /* Domain Access Control Register */ 175 IFSR32_EL2, /* Instruction Fault Status Register */ 176 FPEXC32_EL2, /* Floating-Point Exception Control Register */ 177 DBGVCR32_EL2, /* Debug Vector Catch Register */ 178 179 NR_SYS_REGS /* Nothing after this line! */ 180 }; 181 182 /* 32bit mapping */ 183 #define c0_MPIDR (MPIDR_EL1 * 2) /* MultiProcessor ID Register */ 184 #define c0_CSSELR (CSSELR_EL1 * 2)/* Cache Size Selection Register */ 185 #define c1_SCTLR (SCTLR_EL1 * 2) /* System Control Register */ 186 #define c1_ACTLR (ACTLR_EL1 * 2) /* Auxiliary Control Register */ 187 #define c1_CPACR (CPACR_EL1 * 2) /* Coprocessor Access Control */ 188 #define c2_TTBR0 (TTBR0_EL1 * 2) /* Translation Table Base Register 0 */ 189 #define c2_TTBR0_high (c2_TTBR0 + 1) /* TTBR0 top 32 bits */ 190 #define c2_TTBR1 (TTBR1_EL1 * 2) /* Translation Table Base Register 1 */ 191 #define c2_TTBR1_high (c2_TTBR1 + 1) /* TTBR1 top 32 bits */ 192 #define c2_TTBCR (TCR_EL1 * 2) /* Translation Table Base Control R. */ 193 #define c3_DACR (DACR32_EL2 * 2)/* Domain Access Control Register */ 194 #define c5_DFSR (ESR_EL1 * 2) /* Data Fault Status Register */ 195 #define c5_IFSR (IFSR32_EL2 * 2)/* Instruction Fault Status Register */ 196 #define c5_ADFSR (AFSR0_EL1 * 2) /* Auxiliary Data Fault Status R */ 197 #define c5_AIFSR (AFSR1_EL1 * 2) /* Auxiliary Instr Fault Status R */ 198 #define c6_DFAR (FAR_EL1 * 2) /* Data Fault Address Register */ 199 #define c6_IFAR (c6_DFAR + 1) /* Instruction Fault Address Register */ 200 #define c7_PAR (PAR_EL1 * 2) /* Physical Address Register */ 201 #define c7_PAR_high (c7_PAR + 1) /* PAR top 32 bits */ 202 #define c10_PRRR (MAIR_EL1 * 2) /* Primary Region Remap Register */ 203 #define c10_NMRR (c10_PRRR + 1) /* Normal Memory Remap Register */ 204 #define c12_VBAR (VBAR_EL1 * 2) /* Vector Base Address Register */ 205 #define c13_CID (CONTEXTIDR_EL1 * 2) /* Context ID Register */ 206 #define c13_TID_URW (TPIDR_EL0 * 2) /* Thread ID, User R/W */ 207 #define c13_TID_URO (TPIDRRO_EL0 * 2)/* Thread ID, User R/O */ 208 #define c13_TID_PRIV (TPIDR_EL1 * 2) /* Thread ID, Privileged */ 209 #define c10_AMAIR0 (AMAIR_EL1 * 2) /* Aux Memory Attr Indirection Reg */ 210 #define c10_AMAIR1 (c10_AMAIR0 + 1)/* Aux Memory Attr Indirection Reg */ 211 #define c14_CNTKCTL (CNTKCTL_EL1 * 2) /* Timer Control Register (PL1) */ 212 213 #define cp14_DBGDSCRext (MDSCR_EL1 * 2) 214 #define cp14_DBGBCR0 (DBGBCR0_EL1 * 2) 215 #define cp14_DBGBVR0 (DBGBVR0_EL1 * 2) 216 #define cp14_DBGBXVR0 (cp14_DBGBVR0 + 1) 217 #define cp14_DBGWCR0 (DBGWCR0_EL1 * 2) 218 #define cp14_DBGWVR0 (DBGWVR0_EL1 * 2) 219 #define cp14_DBGDCCINT (MDCCINT_EL1 * 2) 220 221 #define NR_COPRO_REGS (NR_SYS_REGS * 2) 222 223 struct kvm_cpu_context { 224 struct kvm_regs gp_regs; 225 union { 226 u64 sys_regs[NR_SYS_REGS]; 227 u32 copro[NR_COPRO_REGS]; 228 }; 229 230 struct kvm_vcpu *__hyp_running_vcpu; 231 }; 232 233 struct kvm_pmu_events { 234 u32 events_host; 235 u32 events_guest; 236 }; 237 238 struct kvm_host_data { 239 struct kvm_cpu_context host_ctxt; 240 struct kvm_pmu_events pmu_events; 241 }; 242 243 typedef struct kvm_host_data kvm_host_data_t; 244 245 struct vcpu_reset_state { 246 unsigned long pc; 247 unsigned long r0; 248 bool be; 249 bool reset; 250 }; 251 252 struct kvm_vcpu_arch { 253 struct kvm_cpu_context ctxt; 254 void *sve_state; 255 unsigned int sve_max_vl; 256 257 /* HYP configuration */ 258 u64 hcr_el2; 259 u32 mdcr_el2; 260 261 /* Exception Information */ 262 struct kvm_vcpu_fault_info fault; 263 264 /* State of various workarounds, see kvm_asm.h for bit assignment */ 265 u64 workaround_flags; 266 267 /* Miscellaneous vcpu state flags */ 268 u64 flags; 269 270 /* 271 * We maintain more than a single set of debug registers to support 272 * debugging the guest from the host and to maintain separate host and 273 * guest state during world switches. vcpu_debug_state are the debug 274 * registers of the vcpu as the guest sees them. host_debug_state are 275 * the host registers which are saved and restored during 276 * world switches. external_debug_state contains the debug 277 * values we want to debug the guest. This is set via the 278 * KVM_SET_GUEST_DEBUG ioctl. 279 * 280 * debug_ptr points to the set of debug registers that should be loaded 281 * onto the hardware when running the guest. 282 */ 283 struct kvm_guest_debug_arch *debug_ptr; 284 struct kvm_guest_debug_arch vcpu_debug_state; 285 struct kvm_guest_debug_arch external_debug_state; 286 287 /* Pointer to host CPU context */ 288 struct kvm_cpu_context *host_cpu_context; 289 290 struct thread_info *host_thread_info; /* hyp VA */ 291 struct user_fpsimd_state *host_fpsimd_state; /* hyp VA */ 292 293 struct { 294 /* {Break,watch}point registers */ 295 struct kvm_guest_debug_arch regs; 296 /* Statistical profiling extension */ 297 u64 pmscr_el1; 298 } host_debug_state; 299 300 /* VGIC state */ 301 struct vgic_cpu vgic_cpu; 302 struct arch_timer_cpu timer_cpu; 303 struct kvm_pmu pmu; 304 305 /* 306 * Anything that is not used directly from assembly code goes 307 * here. 308 */ 309 310 /* 311 * Guest registers we preserve during guest debugging. 312 * 313 * These shadow registers are updated by the kvm_handle_sys_reg 314 * trap handler if the guest accesses or updates them while we 315 * are using guest debug. 316 */ 317 struct { 318 u32 mdscr_el1; 319 } guest_debug_preserved; 320 321 /* vcpu power-off state */ 322 bool power_off; 323 324 /* Don't run the guest (internal implementation need) */ 325 bool pause; 326 327 /* Cache some mmu pages needed inside spinlock regions */ 328 struct kvm_mmu_memory_cache mmu_page_cache; 329 330 /* Target CPU and feature flags */ 331 int target; 332 DECLARE_BITMAP(features, KVM_VCPU_MAX_FEATURES); 333 334 /* Detect first run of a vcpu */ 335 bool has_run_once; 336 337 /* Virtual SError ESR to restore when HCR_EL2.VSE is set */ 338 u64 vsesr_el2; 339 340 /* Additional reset state */ 341 struct vcpu_reset_state reset_state; 342 343 /* True when deferrable sysregs are loaded on the physical CPU, 344 * see kvm_vcpu_load_sysregs and kvm_vcpu_put_sysregs. */ 345 bool sysregs_loaded_on_cpu; 346 347 /* Guest PV state */ 348 struct { 349 u64 steal; 350 u64 last_steal; 351 gpa_t base; 352 } steal; 353 }; 354 355 /* Pointer to the vcpu's SVE FFR for sve_{save,load}_state() */ 356 #define vcpu_sve_pffr(vcpu) ((void *)((char *)((vcpu)->arch.sve_state) + \ 357 sve_ffr_offset((vcpu)->arch.sve_max_vl))) 358 359 #define vcpu_sve_state_size(vcpu) ({ \ 360 size_t __size_ret; \ 361 unsigned int __vcpu_vq; \ 362 \ 363 if (WARN_ON(!sve_vl_valid((vcpu)->arch.sve_max_vl))) { \ 364 __size_ret = 0; \ 365 } else { \ 366 __vcpu_vq = sve_vq_from_vl((vcpu)->arch.sve_max_vl); \ 367 __size_ret = SVE_SIG_REGS_SIZE(__vcpu_vq); \ 368 } \ 369 \ 370 __size_ret; \ 371 }) 372 373 /* vcpu_arch flags field values: */ 374 #define KVM_ARM64_DEBUG_DIRTY (1 << 0) 375 #define KVM_ARM64_FP_ENABLED (1 << 1) /* guest FP regs loaded */ 376 #define KVM_ARM64_FP_HOST (1 << 2) /* host FP regs loaded */ 377 #define KVM_ARM64_HOST_SVE_IN_USE (1 << 3) /* backup for host TIF_SVE */ 378 #define KVM_ARM64_HOST_SVE_ENABLED (1 << 4) /* SVE enabled for EL0 */ 379 #define KVM_ARM64_GUEST_HAS_SVE (1 << 5) /* SVE exposed to guest */ 380 #define KVM_ARM64_VCPU_SVE_FINALIZED (1 << 6) /* SVE config completed */ 381 #define KVM_ARM64_GUEST_HAS_PTRAUTH (1 << 7) /* PTRAUTH exposed to guest */ 382 383 #define vcpu_has_sve(vcpu) (system_supports_sve() && \ 384 ((vcpu)->arch.flags & KVM_ARM64_GUEST_HAS_SVE)) 385 386 #define vcpu_has_ptrauth(vcpu) ((system_supports_address_auth() || \ 387 system_supports_generic_auth()) && \ 388 ((vcpu)->arch.flags & KVM_ARM64_GUEST_HAS_PTRAUTH)) 389 390 #define vcpu_gp_regs(v) (&(v)->arch.ctxt.gp_regs) 391 392 /* 393 * Only use __vcpu_sys_reg if you know you want the memory backed version of a 394 * register, and not the one most recently accessed by a running VCPU. For 395 * example, for userspace access or for system registers that are never context 396 * switched, but only emulated. 397 */ 398 #define __vcpu_sys_reg(v,r) ((v)->arch.ctxt.sys_regs[(r)]) 399 400 u64 vcpu_read_sys_reg(const struct kvm_vcpu *vcpu, int reg); 401 void vcpu_write_sys_reg(struct kvm_vcpu *vcpu, u64 val, int reg); 402 403 /* 404 * CP14 and CP15 live in the same array, as they are backed by the 405 * same system registers. 406 */ 407 #define vcpu_cp14(v,r) ((v)->arch.ctxt.copro[(r)]) 408 #define vcpu_cp15(v,r) ((v)->arch.ctxt.copro[(r)]) 409 410 struct kvm_vm_stat { 411 ulong remote_tlb_flush; 412 }; 413 414 struct kvm_vcpu_stat { 415 u64 halt_successful_poll; 416 u64 halt_attempted_poll; 417 u64 halt_poll_invalid; 418 u64 halt_wakeup; 419 u64 hvc_exit_stat; 420 u64 wfe_exit_stat; 421 u64 wfi_exit_stat; 422 u64 mmio_exit_user; 423 u64 mmio_exit_kernel; 424 u64 exits; 425 }; 426 427 int kvm_vcpu_preferred_target(struct kvm_vcpu_init *init); 428 unsigned long kvm_arm_num_regs(struct kvm_vcpu *vcpu); 429 int kvm_arm_copy_reg_indices(struct kvm_vcpu *vcpu, u64 __user *indices); 430 int kvm_arm_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg); 431 int kvm_arm_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg); 432 int __kvm_arm_vcpu_get_events(struct kvm_vcpu *vcpu, 433 struct kvm_vcpu_events *events); 434 435 int __kvm_arm_vcpu_set_events(struct kvm_vcpu *vcpu, 436 struct kvm_vcpu_events *events); 437 438 #define KVM_ARCH_WANT_MMU_NOTIFIER 439 int kvm_unmap_hva_range(struct kvm *kvm, 440 unsigned long start, unsigned long end); 441 int kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte); 442 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end); 443 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva); 444 445 void kvm_arm_halt_guest(struct kvm *kvm); 446 void kvm_arm_resume_guest(struct kvm *kvm); 447 448 u64 __kvm_call_hyp(void *hypfn, ...); 449 450 /* 451 * The couple of isb() below are there to guarantee the same behaviour 452 * on VHE as on !VHE, where the eret to EL1 acts as a context 453 * synchronization event. 454 */ 455 #define kvm_call_hyp(f, ...) \ 456 do { \ 457 if (has_vhe()) { \ 458 f(__VA_ARGS__); \ 459 isb(); \ 460 } else { \ 461 __kvm_call_hyp(kvm_ksym_ref(f), ##__VA_ARGS__); \ 462 } \ 463 } while(0) 464 465 #define kvm_call_hyp_ret(f, ...) \ 466 ({ \ 467 typeof(f(__VA_ARGS__)) ret; \ 468 \ 469 if (has_vhe()) { \ 470 ret = f(__VA_ARGS__); \ 471 isb(); \ 472 } else { \ 473 ret = __kvm_call_hyp(kvm_ksym_ref(f), \ 474 ##__VA_ARGS__); \ 475 } \ 476 \ 477 ret; \ 478 }) 479 480 void force_vm_exit(const cpumask_t *mask); 481 void kvm_mmu_wp_memory_region(struct kvm *kvm, int slot); 482 483 int handle_exit(struct kvm_vcpu *vcpu, struct kvm_run *run, 484 int exception_index); 485 void handle_exit_early(struct kvm_vcpu *vcpu, struct kvm_run *run, 486 int exception_index); 487 488 /* MMIO helpers */ 489 void kvm_mmio_write_buf(void *buf, unsigned int len, unsigned long data); 490 unsigned long kvm_mmio_read_buf(const void *buf, unsigned int len); 491 492 int kvm_handle_mmio_return(struct kvm_vcpu *vcpu, struct kvm_run *run); 493 int io_mem_abort(struct kvm_vcpu *vcpu, struct kvm_run *run, 494 phys_addr_t fault_ipa); 495 496 int kvm_perf_init(void); 497 int kvm_perf_teardown(void); 498 499 long kvm_hypercall_pv_features(struct kvm_vcpu *vcpu); 500 gpa_t kvm_init_stolen_time(struct kvm_vcpu *vcpu); 501 void kvm_update_stolen_time(struct kvm_vcpu *vcpu); 502 503 int kvm_arm_pvtime_set_attr(struct kvm_vcpu *vcpu, 504 struct kvm_device_attr *attr); 505 int kvm_arm_pvtime_get_attr(struct kvm_vcpu *vcpu, 506 struct kvm_device_attr *attr); 507 int kvm_arm_pvtime_has_attr(struct kvm_vcpu *vcpu, 508 struct kvm_device_attr *attr); 509 510 static inline void kvm_arm_pvtime_vcpu_init(struct kvm_vcpu_arch *vcpu_arch) 511 { 512 vcpu_arch->steal.base = GPA_INVALID; 513 } 514 515 static inline bool kvm_arm_is_pvtime_enabled(struct kvm_vcpu_arch *vcpu_arch) 516 { 517 return (vcpu_arch->steal.base != GPA_INVALID); 518 } 519 520 void kvm_set_sei_esr(struct kvm_vcpu *vcpu, u64 syndrome); 521 522 struct kvm_vcpu *kvm_mpidr_to_vcpu(struct kvm *kvm, unsigned long mpidr); 523 524 DECLARE_PER_CPU(kvm_host_data_t, kvm_host_data); 525 526 static inline void kvm_init_host_cpu_context(struct kvm_cpu_context *cpu_ctxt) 527 { 528 /* The host's MPIDR is immutable, so let's set it up at boot time */ 529 cpu_ctxt->sys_regs[MPIDR_EL1] = read_cpuid_mpidr(); 530 } 531 532 void __kvm_enable_ssbs(void); 533 534 static inline void __cpu_init_hyp_mode(phys_addr_t pgd_ptr, 535 unsigned long hyp_stack_ptr, 536 unsigned long vector_ptr) 537 { 538 /* 539 * Calculate the raw per-cpu offset without a translation from the 540 * kernel's mapping to the linear mapping, and store it in tpidr_el2 541 * so that we can use adr_l to access per-cpu variables in EL2. 542 */ 543 u64 tpidr_el2 = ((u64)this_cpu_ptr(&kvm_host_data) - 544 (u64)kvm_ksym_ref(kvm_host_data)); 545 546 /* 547 * Call initialization code, and switch to the full blown HYP code. 548 * If the cpucaps haven't been finalized yet, something has gone very 549 * wrong, and hyp will crash and burn when it uses any 550 * cpus_have_const_cap() wrapper. 551 */ 552 BUG_ON(!system_capabilities_finalized()); 553 __kvm_call_hyp((void *)pgd_ptr, hyp_stack_ptr, vector_ptr, tpidr_el2); 554 555 /* 556 * Disabling SSBD on a non-VHE system requires us to enable SSBS 557 * at EL2. 558 */ 559 if (!has_vhe() && this_cpu_has_cap(ARM64_SSBS) && 560 arm64_get_ssbd_state() == ARM64_SSBD_FORCE_DISABLE) { 561 kvm_call_hyp(__kvm_enable_ssbs); 562 } 563 } 564 565 static inline bool kvm_arch_requires_vhe(void) 566 { 567 /* 568 * The Arm architecture specifies that implementation of SVE 569 * requires VHE also to be implemented. The KVM code for arm64 570 * relies on this when SVE is present: 571 */ 572 if (system_supports_sve()) 573 return true; 574 575 /* Some implementations have defects that confine them to VHE */ 576 if (cpus_have_cap(ARM64_WORKAROUND_SPECULATIVE_AT_VHE)) 577 return true; 578 579 return false; 580 } 581 582 void kvm_arm_vcpu_ptrauth_trap(struct kvm_vcpu *vcpu); 583 584 static inline void kvm_arch_hardware_unsetup(void) {} 585 static inline void kvm_arch_sync_events(struct kvm *kvm) {} 586 static inline void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu) {} 587 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {} 588 589 void kvm_arm_init_debug(void); 590 void kvm_arm_setup_debug(struct kvm_vcpu *vcpu); 591 void kvm_arm_clear_debug(struct kvm_vcpu *vcpu); 592 void kvm_arm_reset_debug_ptr(struct kvm_vcpu *vcpu); 593 int kvm_arm_vcpu_arch_set_attr(struct kvm_vcpu *vcpu, 594 struct kvm_device_attr *attr); 595 int kvm_arm_vcpu_arch_get_attr(struct kvm_vcpu *vcpu, 596 struct kvm_device_attr *attr); 597 int kvm_arm_vcpu_arch_has_attr(struct kvm_vcpu *vcpu, 598 struct kvm_device_attr *attr); 599 600 static inline void __cpu_init_stage2(void) {} 601 602 /* Guest/host FPSIMD coordination helpers */ 603 int kvm_arch_vcpu_run_map_fp(struct kvm_vcpu *vcpu); 604 void kvm_arch_vcpu_load_fp(struct kvm_vcpu *vcpu); 605 void kvm_arch_vcpu_ctxsync_fp(struct kvm_vcpu *vcpu); 606 void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu); 607 608 static inline bool kvm_pmu_counter_deferred(struct perf_event_attr *attr) 609 { 610 return (!has_vhe() && attr->exclude_host); 611 } 612 613 #ifdef CONFIG_KVM /* Avoid conflicts with core headers if CONFIG_KVM=n */ 614 static inline int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu) 615 { 616 return kvm_arch_vcpu_run_map_fp(vcpu); 617 } 618 619 void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr); 620 void kvm_clr_pmu_events(u32 clr); 621 622 void kvm_vcpu_pmu_restore_guest(struct kvm_vcpu *vcpu); 623 void kvm_vcpu_pmu_restore_host(struct kvm_vcpu *vcpu); 624 #else 625 static inline void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr) {} 626 static inline void kvm_clr_pmu_events(u32 clr) {} 627 #endif 628 629 static inline void kvm_arm_vhe_guest_enter(void) 630 { 631 local_daif_mask(); 632 633 /* 634 * Having IRQs masked via PMR when entering the guest means the GIC 635 * will not signal the CPU of interrupts of lower priority, and the 636 * only way to get out will be via guest exceptions. 637 * Naturally, we want to avoid this. 638 * 639 * local_daif_mask() already sets GIC_PRIO_PSR_I_SET, we just need a 640 * dsb to ensure the redistributor is forwards EL2 IRQs to the CPU. 641 */ 642 pmr_sync(); 643 } 644 645 static inline void kvm_arm_vhe_guest_exit(void) 646 { 647 /* 648 * local_daif_restore() takes care to properly restore PSTATE.DAIF 649 * and the GIC PMR if the host is using IRQ priorities. 650 */ 651 local_daif_restore(DAIF_PROCCTX_NOIRQ); 652 653 /* 654 * When we exit from the guest we change a number of CPU configuration 655 * parameters, such as traps. Make sure these changes take effect 656 * before running the host or additional guests. 657 */ 658 isb(); 659 } 660 661 #define KVM_BP_HARDEN_UNKNOWN -1 662 #define KVM_BP_HARDEN_WA_NEEDED 0 663 #define KVM_BP_HARDEN_NOT_REQUIRED 1 664 665 static inline int kvm_arm_harden_branch_predictor(void) 666 { 667 switch (get_spectre_v2_workaround_state()) { 668 case ARM64_BP_HARDEN_WA_NEEDED: 669 return KVM_BP_HARDEN_WA_NEEDED; 670 case ARM64_BP_HARDEN_NOT_REQUIRED: 671 return KVM_BP_HARDEN_NOT_REQUIRED; 672 case ARM64_BP_HARDEN_UNKNOWN: 673 default: 674 return KVM_BP_HARDEN_UNKNOWN; 675 } 676 } 677 678 #define KVM_SSBD_UNKNOWN -1 679 #define KVM_SSBD_FORCE_DISABLE 0 680 #define KVM_SSBD_KERNEL 1 681 #define KVM_SSBD_FORCE_ENABLE 2 682 #define KVM_SSBD_MITIGATED 3 683 684 static inline int kvm_arm_have_ssbd(void) 685 { 686 switch (arm64_get_ssbd_state()) { 687 case ARM64_SSBD_FORCE_DISABLE: 688 return KVM_SSBD_FORCE_DISABLE; 689 case ARM64_SSBD_KERNEL: 690 return KVM_SSBD_KERNEL; 691 case ARM64_SSBD_FORCE_ENABLE: 692 return KVM_SSBD_FORCE_ENABLE; 693 case ARM64_SSBD_MITIGATED: 694 return KVM_SSBD_MITIGATED; 695 case ARM64_SSBD_UNKNOWN: 696 default: 697 return KVM_SSBD_UNKNOWN; 698 } 699 } 700 701 void kvm_vcpu_load_sysregs(struct kvm_vcpu *vcpu); 702 void kvm_vcpu_put_sysregs(struct kvm_vcpu *vcpu); 703 704 void kvm_set_ipa_limit(void); 705 706 #define __KVM_HAVE_ARCH_VM_ALLOC 707 struct kvm *kvm_arch_alloc_vm(void); 708 void kvm_arch_free_vm(struct kvm *kvm); 709 710 int kvm_arm_setup_stage2(struct kvm *kvm, unsigned long type); 711 712 int kvm_arm_vcpu_finalize(struct kvm_vcpu *vcpu, int feature); 713 bool kvm_arm_vcpu_is_finalized(struct kvm_vcpu *vcpu); 714 715 #define kvm_arm_vcpu_sve_finalized(vcpu) \ 716 ((vcpu)->arch.flags & KVM_ARM64_VCPU_SVE_FINALIZED) 717 718 #endif /* __ARM64_KVM_HOST_H__ */ 719