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/arm-smccc.h> 15 #include <linux/bitmap.h> 16 #include <linux/types.h> 17 #include <linux/jump_label.h> 18 #include <linux/kvm_types.h> 19 #include <linux/percpu.h> 20 #include <linux/psci.h> 21 #include <asm/arch_gicv3.h> 22 #include <asm/barrier.h> 23 #include <asm/cpufeature.h> 24 #include <asm/cputype.h> 25 #include <asm/daifflags.h> 26 #include <asm/fpsimd.h> 27 #include <asm/kvm.h> 28 #include <asm/kvm_asm.h> 29 30 #define __KVM_HAVE_ARCH_INTC_INITIALIZED 31 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 #define KVM_REQ_RELOAD_GICv4 KVM_ARCH_REQ(4) 48 #define KVM_REQ_RELOAD_PMU KVM_ARCH_REQ(5) 49 50 #define KVM_DIRTY_LOG_MANUAL_CAPS (KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE | \ 51 KVM_DIRTY_LOG_INITIALLY_SET) 52 53 #define KVM_HAVE_MMU_RWLOCK 54 55 /* 56 * Mode of operation configurable with kvm-arm.mode early param. 57 * See Documentation/admin-guide/kernel-parameters.txt for more information. 58 */ 59 enum kvm_mode { 60 KVM_MODE_DEFAULT, 61 KVM_MODE_PROTECTED, 62 KVM_MODE_NONE, 63 }; 64 enum kvm_mode kvm_get_mode(void); 65 66 DECLARE_STATIC_KEY_FALSE(userspace_irqchip_in_use); 67 68 extern unsigned int kvm_sve_max_vl; 69 int kvm_arm_init_sve(void); 70 71 u32 __attribute_const__ kvm_target_cpu(void); 72 int kvm_reset_vcpu(struct kvm_vcpu *vcpu); 73 void kvm_arm_vcpu_destroy(struct kvm_vcpu *vcpu); 74 75 struct kvm_vmid { 76 atomic64_t id; 77 }; 78 79 struct kvm_s2_mmu { 80 struct kvm_vmid vmid; 81 82 /* 83 * stage2 entry level table 84 * 85 * Two kvm_s2_mmu structures in the same VM can point to the same 86 * pgd here. This happens when running a guest using a 87 * translation regime that isn't affected by its own stage-2 88 * translation, such as a non-VHE hypervisor running at vEL2, or 89 * for vEL1/EL0 with vHCR_EL2.VM == 0. In that case, we use the 90 * canonical stage-2 page tables. 91 */ 92 phys_addr_t pgd_phys; 93 struct kvm_pgtable *pgt; 94 95 /* The last vcpu id that ran on each physical CPU */ 96 int __percpu *last_vcpu_ran; 97 98 struct kvm_arch *arch; 99 }; 100 101 struct kvm_arch_memory_slot { 102 }; 103 104 struct kvm_arch { 105 struct kvm_s2_mmu mmu; 106 107 /* VTCR_EL2 value for this VM */ 108 u64 vtcr; 109 110 /* The maximum number of vCPUs depends on the used GIC model */ 111 int max_vcpus; 112 113 /* Interrupt controller */ 114 struct vgic_dist vgic; 115 116 /* Mandated version of PSCI */ 117 u32 psci_version; 118 119 /* 120 * If we encounter a data abort without valid instruction syndrome 121 * information, report this to user space. User space can (and 122 * should) opt in to this feature if KVM_CAP_ARM_NISV_TO_USER is 123 * supported. 124 */ 125 #define KVM_ARCH_FLAG_RETURN_NISV_IO_ABORT_TO_USER 0 126 /* Memory Tagging Extension enabled for the guest */ 127 #define KVM_ARCH_FLAG_MTE_ENABLED 1 128 /* At least one vCPU has ran in the VM */ 129 #define KVM_ARCH_FLAG_HAS_RAN_ONCE 2 130 /* 131 * The following two bits are used to indicate the guest's EL1 132 * register width configuration. A value of KVM_ARCH_FLAG_EL1_32BIT 133 * bit is valid only when KVM_ARCH_FLAG_REG_WIDTH_CONFIGURED is set. 134 * Otherwise, the guest's EL1 register width has not yet been 135 * determined yet. 136 */ 137 #define KVM_ARCH_FLAG_REG_WIDTH_CONFIGURED 3 138 #define KVM_ARCH_FLAG_EL1_32BIT 4 139 140 unsigned long flags; 141 142 /* 143 * VM-wide PMU filter, implemented as a bitmap and big enough for 144 * up to 2^10 events (ARMv8.0) or 2^16 events (ARMv8.1+). 145 */ 146 unsigned long *pmu_filter; 147 struct arm_pmu *arm_pmu; 148 149 cpumask_var_t supported_cpus; 150 151 u8 pfr0_csv2; 152 u8 pfr0_csv3; 153 }; 154 155 struct kvm_vcpu_fault_info { 156 u64 esr_el2; /* Hyp Syndrom Register */ 157 u64 far_el2; /* Hyp Fault Address Register */ 158 u64 hpfar_el2; /* Hyp IPA Fault Address Register */ 159 u64 disr_el1; /* Deferred [SError] Status Register */ 160 }; 161 162 enum vcpu_sysreg { 163 __INVALID_SYSREG__, /* 0 is reserved as an invalid value */ 164 MPIDR_EL1, /* MultiProcessor Affinity Register */ 165 CSSELR_EL1, /* Cache Size Selection Register */ 166 SCTLR_EL1, /* System Control Register */ 167 ACTLR_EL1, /* Auxiliary Control Register */ 168 CPACR_EL1, /* Coprocessor Access Control */ 169 ZCR_EL1, /* SVE Control */ 170 TTBR0_EL1, /* Translation Table Base Register 0 */ 171 TTBR1_EL1, /* Translation Table Base Register 1 */ 172 TCR_EL1, /* Translation Control Register */ 173 ESR_EL1, /* Exception Syndrome Register */ 174 AFSR0_EL1, /* Auxiliary Fault Status Register 0 */ 175 AFSR1_EL1, /* Auxiliary Fault Status Register 1 */ 176 FAR_EL1, /* Fault Address Register */ 177 MAIR_EL1, /* Memory Attribute Indirection Register */ 178 VBAR_EL1, /* Vector Base Address Register */ 179 CONTEXTIDR_EL1, /* Context ID Register */ 180 TPIDR_EL0, /* Thread ID, User R/W */ 181 TPIDRRO_EL0, /* Thread ID, User R/O */ 182 TPIDR_EL1, /* Thread ID, Privileged */ 183 AMAIR_EL1, /* Aux Memory Attribute Indirection Register */ 184 CNTKCTL_EL1, /* Timer Control Register (EL1) */ 185 PAR_EL1, /* Physical Address Register */ 186 MDSCR_EL1, /* Monitor Debug System Control Register */ 187 MDCCINT_EL1, /* Monitor Debug Comms Channel Interrupt Enable Reg */ 188 OSLSR_EL1, /* OS Lock Status Register */ 189 DISR_EL1, /* Deferred Interrupt Status Register */ 190 191 /* Performance Monitors Registers */ 192 PMCR_EL0, /* Control Register */ 193 PMSELR_EL0, /* Event Counter Selection Register */ 194 PMEVCNTR0_EL0, /* Event Counter Register (0-30) */ 195 PMEVCNTR30_EL0 = PMEVCNTR0_EL0 + 30, 196 PMCCNTR_EL0, /* Cycle Counter Register */ 197 PMEVTYPER0_EL0, /* Event Type Register (0-30) */ 198 PMEVTYPER30_EL0 = PMEVTYPER0_EL0 + 30, 199 PMCCFILTR_EL0, /* Cycle Count Filter Register */ 200 PMCNTENSET_EL0, /* Count Enable Set Register */ 201 PMINTENSET_EL1, /* Interrupt Enable Set Register */ 202 PMOVSSET_EL0, /* Overflow Flag Status Set Register */ 203 PMUSERENR_EL0, /* User Enable Register */ 204 205 /* Pointer Authentication Registers in a strict increasing order. */ 206 APIAKEYLO_EL1, 207 APIAKEYHI_EL1, 208 APIBKEYLO_EL1, 209 APIBKEYHI_EL1, 210 APDAKEYLO_EL1, 211 APDAKEYHI_EL1, 212 APDBKEYLO_EL1, 213 APDBKEYHI_EL1, 214 APGAKEYLO_EL1, 215 APGAKEYHI_EL1, 216 217 ELR_EL1, 218 SP_EL1, 219 SPSR_EL1, 220 221 CNTVOFF_EL2, 222 CNTV_CVAL_EL0, 223 CNTV_CTL_EL0, 224 CNTP_CVAL_EL0, 225 CNTP_CTL_EL0, 226 227 /* Memory Tagging Extension registers */ 228 RGSR_EL1, /* Random Allocation Tag Seed Register */ 229 GCR_EL1, /* Tag Control Register */ 230 TFSR_EL1, /* Tag Fault Status Register (EL1) */ 231 TFSRE0_EL1, /* Tag Fault Status Register (EL0) */ 232 233 /* 32bit specific registers. Keep them at the end of the range */ 234 DACR32_EL2, /* Domain Access Control Register */ 235 IFSR32_EL2, /* Instruction Fault Status Register */ 236 FPEXC32_EL2, /* Floating-Point Exception Control Register */ 237 DBGVCR32_EL2, /* Debug Vector Catch Register */ 238 239 NR_SYS_REGS /* Nothing after this line! */ 240 }; 241 242 struct kvm_cpu_context { 243 struct user_pt_regs regs; /* sp = sp_el0 */ 244 245 u64 spsr_abt; 246 u64 spsr_und; 247 u64 spsr_irq; 248 u64 spsr_fiq; 249 250 struct user_fpsimd_state fp_regs; 251 252 u64 sys_regs[NR_SYS_REGS]; 253 254 struct kvm_vcpu *__hyp_running_vcpu; 255 }; 256 257 struct kvm_pmu_events { 258 u32 events_host; 259 u32 events_guest; 260 }; 261 262 struct kvm_host_data { 263 struct kvm_cpu_context host_ctxt; 264 struct kvm_pmu_events pmu_events; 265 }; 266 267 struct kvm_host_psci_config { 268 /* PSCI version used by host. */ 269 u32 version; 270 271 /* Function IDs used by host if version is v0.1. */ 272 struct psci_0_1_function_ids function_ids_0_1; 273 274 bool psci_0_1_cpu_suspend_implemented; 275 bool psci_0_1_cpu_on_implemented; 276 bool psci_0_1_cpu_off_implemented; 277 bool psci_0_1_migrate_implemented; 278 }; 279 280 extern struct kvm_host_psci_config kvm_nvhe_sym(kvm_host_psci_config); 281 #define kvm_host_psci_config CHOOSE_NVHE_SYM(kvm_host_psci_config) 282 283 extern s64 kvm_nvhe_sym(hyp_physvirt_offset); 284 #define hyp_physvirt_offset CHOOSE_NVHE_SYM(hyp_physvirt_offset) 285 286 extern u64 kvm_nvhe_sym(hyp_cpu_logical_map)[NR_CPUS]; 287 #define hyp_cpu_logical_map CHOOSE_NVHE_SYM(hyp_cpu_logical_map) 288 289 struct vcpu_reset_state { 290 unsigned long pc; 291 unsigned long r0; 292 bool be; 293 bool reset; 294 }; 295 296 struct kvm_vcpu_arch { 297 struct kvm_cpu_context ctxt; 298 299 /* Guest floating point state */ 300 void *sve_state; 301 unsigned int sve_max_vl; 302 u64 svcr; 303 304 /* Stage 2 paging state used by the hardware on next switch */ 305 struct kvm_s2_mmu *hw_mmu; 306 307 /* Values of trap registers for the guest. */ 308 u64 hcr_el2; 309 u64 mdcr_el2; 310 u64 cptr_el2; 311 312 /* Values of trap registers for the host before guest entry. */ 313 u64 mdcr_el2_host; 314 315 /* Exception Information */ 316 struct kvm_vcpu_fault_info fault; 317 318 /* Miscellaneous vcpu state flags */ 319 u64 flags; 320 321 /* 322 * We maintain more than a single set of debug registers to support 323 * debugging the guest from the host and to maintain separate host and 324 * guest state during world switches. vcpu_debug_state are the debug 325 * registers of the vcpu as the guest sees them. host_debug_state are 326 * the host registers which are saved and restored during 327 * world switches. external_debug_state contains the debug 328 * values we want to debug the guest. This is set via the 329 * KVM_SET_GUEST_DEBUG ioctl. 330 * 331 * debug_ptr points to the set of debug registers that should be loaded 332 * onto the hardware when running the guest. 333 */ 334 struct kvm_guest_debug_arch *debug_ptr; 335 struct kvm_guest_debug_arch vcpu_debug_state; 336 struct kvm_guest_debug_arch external_debug_state; 337 338 struct user_fpsimd_state *host_fpsimd_state; /* hyp VA */ 339 struct task_struct *parent_task; 340 341 struct { 342 /* {Break,watch}point registers */ 343 struct kvm_guest_debug_arch regs; 344 /* Statistical profiling extension */ 345 u64 pmscr_el1; 346 /* Self-hosted trace */ 347 u64 trfcr_el1; 348 } host_debug_state; 349 350 /* VGIC state */ 351 struct vgic_cpu vgic_cpu; 352 struct arch_timer_cpu timer_cpu; 353 struct kvm_pmu pmu; 354 355 /* 356 * Anything that is not used directly from assembly code goes 357 * here. 358 */ 359 360 /* 361 * Guest registers we preserve during guest debugging. 362 * 363 * These shadow registers are updated by the kvm_handle_sys_reg 364 * trap handler if the guest accesses or updates them while we 365 * are using guest debug. 366 */ 367 struct { 368 u32 mdscr_el1; 369 } guest_debug_preserved; 370 371 /* vcpu power-off state */ 372 bool power_off; 373 374 /* Don't run the guest (internal implementation need) */ 375 bool pause; 376 377 /* Cache some mmu pages needed inside spinlock regions */ 378 struct kvm_mmu_memory_cache mmu_page_cache; 379 380 /* Target CPU and feature flags */ 381 int target; 382 DECLARE_BITMAP(features, KVM_VCPU_MAX_FEATURES); 383 384 /* Virtual SError ESR to restore when HCR_EL2.VSE is set */ 385 u64 vsesr_el2; 386 387 /* Additional reset state */ 388 struct vcpu_reset_state reset_state; 389 390 /* True when deferrable sysregs are loaded on the physical CPU, 391 * see kvm_vcpu_load_sysregs_vhe and kvm_vcpu_put_sysregs_vhe. */ 392 bool sysregs_loaded_on_cpu; 393 394 /* Guest PV state */ 395 struct { 396 u64 last_steal; 397 gpa_t base; 398 } steal; 399 }; 400 401 /* Pointer to the vcpu's SVE FFR for sve_{save,load}_state() */ 402 #define vcpu_sve_pffr(vcpu) (kern_hyp_va((vcpu)->arch.sve_state) + \ 403 sve_ffr_offset((vcpu)->arch.sve_max_vl)) 404 405 #define vcpu_sve_max_vq(vcpu) sve_vq_from_vl((vcpu)->arch.sve_max_vl) 406 407 #define vcpu_sve_state_size(vcpu) ({ \ 408 size_t __size_ret; \ 409 unsigned int __vcpu_vq; \ 410 \ 411 if (WARN_ON(!sve_vl_valid((vcpu)->arch.sve_max_vl))) { \ 412 __size_ret = 0; \ 413 } else { \ 414 __vcpu_vq = vcpu_sve_max_vq(vcpu); \ 415 __size_ret = SVE_SIG_REGS_SIZE(__vcpu_vq); \ 416 } \ 417 \ 418 __size_ret; \ 419 }) 420 421 /* vcpu_arch flags field values: */ 422 #define KVM_ARM64_DEBUG_DIRTY (1 << 0) 423 #define KVM_ARM64_FP_ENABLED (1 << 1) /* guest FP regs loaded */ 424 #define KVM_ARM64_FP_HOST (1 << 2) /* host FP regs loaded */ 425 #define KVM_ARM64_HOST_SVE_ENABLED (1 << 4) /* SVE enabled for EL0 */ 426 #define KVM_ARM64_GUEST_HAS_SVE (1 << 5) /* SVE exposed to guest */ 427 #define KVM_ARM64_VCPU_SVE_FINALIZED (1 << 6) /* SVE config completed */ 428 #define KVM_ARM64_GUEST_HAS_PTRAUTH (1 << 7) /* PTRAUTH exposed to guest */ 429 #define KVM_ARM64_PENDING_EXCEPTION (1 << 8) /* Exception pending */ 430 /* 431 * Overlaps with KVM_ARM64_EXCEPT_MASK on purpose so that it can't be 432 * set together with an exception... 433 */ 434 #define KVM_ARM64_INCREMENT_PC (1 << 9) /* Increment PC */ 435 #define KVM_ARM64_EXCEPT_MASK (7 << 9) /* Target EL/MODE */ 436 /* 437 * When KVM_ARM64_PENDING_EXCEPTION is set, KVM_ARM64_EXCEPT_MASK can 438 * take the following values: 439 * 440 * For AArch32 EL1: 441 */ 442 #define KVM_ARM64_EXCEPT_AA32_UND (0 << 9) 443 #define KVM_ARM64_EXCEPT_AA32_IABT (1 << 9) 444 #define KVM_ARM64_EXCEPT_AA32_DABT (2 << 9) 445 /* For AArch64: */ 446 #define KVM_ARM64_EXCEPT_AA64_ELx_SYNC (0 << 9) 447 #define KVM_ARM64_EXCEPT_AA64_ELx_IRQ (1 << 9) 448 #define KVM_ARM64_EXCEPT_AA64_ELx_FIQ (2 << 9) 449 #define KVM_ARM64_EXCEPT_AA64_ELx_SERR (3 << 9) 450 #define KVM_ARM64_EXCEPT_AA64_EL1 (0 << 11) 451 #define KVM_ARM64_EXCEPT_AA64_EL2 (1 << 11) 452 453 #define KVM_ARM64_DEBUG_STATE_SAVE_SPE (1 << 12) /* Save SPE context if active */ 454 #define KVM_ARM64_DEBUG_STATE_SAVE_TRBE (1 << 13) /* Save TRBE context if active */ 455 #define KVM_ARM64_FP_FOREIGN_FPSTATE (1 << 14) 456 #define KVM_ARM64_ON_UNSUPPORTED_CPU (1 << 15) /* Physical CPU not in supported_cpus */ 457 #define KVM_ARM64_HOST_SME_ENABLED (1 << 16) /* SME enabled for EL0 */ 458 459 #define KVM_GUESTDBG_VALID_MASK (KVM_GUESTDBG_ENABLE | \ 460 KVM_GUESTDBG_USE_SW_BP | \ 461 KVM_GUESTDBG_USE_HW | \ 462 KVM_GUESTDBG_SINGLESTEP) 463 464 #define vcpu_has_sve(vcpu) (system_supports_sve() && \ 465 ((vcpu)->arch.flags & KVM_ARM64_GUEST_HAS_SVE)) 466 467 #ifdef CONFIG_ARM64_PTR_AUTH 468 #define vcpu_has_ptrauth(vcpu) \ 469 ((cpus_have_final_cap(ARM64_HAS_ADDRESS_AUTH) || \ 470 cpus_have_final_cap(ARM64_HAS_GENERIC_AUTH)) && \ 471 (vcpu)->arch.flags & KVM_ARM64_GUEST_HAS_PTRAUTH) 472 #else 473 #define vcpu_has_ptrauth(vcpu) false 474 #endif 475 476 #define vcpu_on_unsupported_cpu(vcpu) \ 477 ((vcpu)->arch.flags & KVM_ARM64_ON_UNSUPPORTED_CPU) 478 479 #define vcpu_set_on_unsupported_cpu(vcpu) \ 480 ((vcpu)->arch.flags |= KVM_ARM64_ON_UNSUPPORTED_CPU) 481 482 #define vcpu_clear_on_unsupported_cpu(vcpu) \ 483 ((vcpu)->arch.flags &= ~KVM_ARM64_ON_UNSUPPORTED_CPU) 484 485 #define vcpu_gp_regs(v) (&(v)->arch.ctxt.regs) 486 487 /* 488 * Only use __vcpu_sys_reg/ctxt_sys_reg if you know you want the 489 * memory backed version of a register, and not the one most recently 490 * accessed by a running VCPU. For example, for userspace access or 491 * for system registers that are never context switched, but only 492 * emulated. 493 */ 494 #define __ctxt_sys_reg(c,r) (&(c)->sys_regs[(r)]) 495 496 #define ctxt_sys_reg(c,r) (*__ctxt_sys_reg(c,r)) 497 498 #define __vcpu_sys_reg(v,r) (ctxt_sys_reg(&(v)->arch.ctxt, (r))) 499 500 u64 vcpu_read_sys_reg(const struct kvm_vcpu *vcpu, int reg); 501 void vcpu_write_sys_reg(struct kvm_vcpu *vcpu, u64 val, int reg); 502 503 static inline bool __vcpu_read_sys_reg_from_cpu(int reg, u64 *val) 504 { 505 /* 506 * *** VHE ONLY *** 507 * 508 * System registers listed in the switch are not saved on every 509 * exit from the guest but are only saved on vcpu_put. 510 * 511 * Note that MPIDR_EL1 for the guest is set by KVM via VMPIDR_EL2 but 512 * should never be listed below, because the guest cannot modify its 513 * own MPIDR_EL1 and MPIDR_EL1 is accessed for VCPU A from VCPU B's 514 * thread when emulating cross-VCPU communication. 515 */ 516 if (!has_vhe()) 517 return false; 518 519 switch (reg) { 520 case CSSELR_EL1: *val = read_sysreg_s(SYS_CSSELR_EL1); break; 521 case SCTLR_EL1: *val = read_sysreg_s(SYS_SCTLR_EL12); break; 522 case CPACR_EL1: *val = read_sysreg_s(SYS_CPACR_EL12); break; 523 case TTBR0_EL1: *val = read_sysreg_s(SYS_TTBR0_EL12); break; 524 case TTBR1_EL1: *val = read_sysreg_s(SYS_TTBR1_EL12); break; 525 case TCR_EL1: *val = read_sysreg_s(SYS_TCR_EL12); break; 526 case ESR_EL1: *val = read_sysreg_s(SYS_ESR_EL12); break; 527 case AFSR0_EL1: *val = read_sysreg_s(SYS_AFSR0_EL12); break; 528 case AFSR1_EL1: *val = read_sysreg_s(SYS_AFSR1_EL12); break; 529 case FAR_EL1: *val = read_sysreg_s(SYS_FAR_EL12); break; 530 case MAIR_EL1: *val = read_sysreg_s(SYS_MAIR_EL12); break; 531 case VBAR_EL1: *val = read_sysreg_s(SYS_VBAR_EL12); break; 532 case CONTEXTIDR_EL1: *val = read_sysreg_s(SYS_CONTEXTIDR_EL12);break; 533 case TPIDR_EL0: *val = read_sysreg_s(SYS_TPIDR_EL0); break; 534 case TPIDRRO_EL0: *val = read_sysreg_s(SYS_TPIDRRO_EL0); break; 535 case TPIDR_EL1: *val = read_sysreg_s(SYS_TPIDR_EL1); break; 536 case AMAIR_EL1: *val = read_sysreg_s(SYS_AMAIR_EL12); break; 537 case CNTKCTL_EL1: *val = read_sysreg_s(SYS_CNTKCTL_EL12); break; 538 case ELR_EL1: *val = read_sysreg_s(SYS_ELR_EL12); break; 539 case PAR_EL1: *val = read_sysreg_par(); break; 540 case DACR32_EL2: *val = read_sysreg_s(SYS_DACR32_EL2); break; 541 case IFSR32_EL2: *val = read_sysreg_s(SYS_IFSR32_EL2); break; 542 case DBGVCR32_EL2: *val = read_sysreg_s(SYS_DBGVCR32_EL2); break; 543 default: return false; 544 } 545 546 return true; 547 } 548 549 static inline bool __vcpu_write_sys_reg_to_cpu(u64 val, int reg) 550 { 551 /* 552 * *** VHE ONLY *** 553 * 554 * System registers listed in the switch are not restored on every 555 * entry to the guest but are only restored on vcpu_load. 556 * 557 * Note that MPIDR_EL1 for the guest is set by KVM via VMPIDR_EL2 but 558 * should never be listed below, because the MPIDR should only be set 559 * once, before running the VCPU, and never changed later. 560 */ 561 if (!has_vhe()) 562 return false; 563 564 switch (reg) { 565 case CSSELR_EL1: write_sysreg_s(val, SYS_CSSELR_EL1); break; 566 case SCTLR_EL1: write_sysreg_s(val, SYS_SCTLR_EL12); break; 567 case CPACR_EL1: write_sysreg_s(val, SYS_CPACR_EL12); break; 568 case TTBR0_EL1: write_sysreg_s(val, SYS_TTBR0_EL12); break; 569 case TTBR1_EL1: write_sysreg_s(val, SYS_TTBR1_EL12); break; 570 case TCR_EL1: write_sysreg_s(val, SYS_TCR_EL12); break; 571 case ESR_EL1: write_sysreg_s(val, SYS_ESR_EL12); break; 572 case AFSR0_EL1: write_sysreg_s(val, SYS_AFSR0_EL12); break; 573 case AFSR1_EL1: write_sysreg_s(val, SYS_AFSR1_EL12); break; 574 case FAR_EL1: write_sysreg_s(val, SYS_FAR_EL12); break; 575 case MAIR_EL1: write_sysreg_s(val, SYS_MAIR_EL12); break; 576 case VBAR_EL1: write_sysreg_s(val, SYS_VBAR_EL12); break; 577 case CONTEXTIDR_EL1: write_sysreg_s(val, SYS_CONTEXTIDR_EL12);break; 578 case TPIDR_EL0: write_sysreg_s(val, SYS_TPIDR_EL0); break; 579 case TPIDRRO_EL0: write_sysreg_s(val, SYS_TPIDRRO_EL0); break; 580 case TPIDR_EL1: write_sysreg_s(val, SYS_TPIDR_EL1); break; 581 case AMAIR_EL1: write_sysreg_s(val, SYS_AMAIR_EL12); break; 582 case CNTKCTL_EL1: write_sysreg_s(val, SYS_CNTKCTL_EL12); break; 583 case ELR_EL1: write_sysreg_s(val, SYS_ELR_EL12); break; 584 case PAR_EL1: write_sysreg_s(val, SYS_PAR_EL1); break; 585 case DACR32_EL2: write_sysreg_s(val, SYS_DACR32_EL2); break; 586 case IFSR32_EL2: write_sysreg_s(val, SYS_IFSR32_EL2); break; 587 case DBGVCR32_EL2: write_sysreg_s(val, SYS_DBGVCR32_EL2); break; 588 default: return false; 589 } 590 591 return true; 592 } 593 594 struct kvm_vm_stat { 595 struct kvm_vm_stat_generic generic; 596 }; 597 598 struct kvm_vcpu_stat { 599 struct kvm_vcpu_stat_generic generic; 600 u64 hvc_exit_stat; 601 u64 wfe_exit_stat; 602 u64 wfi_exit_stat; 603 u64 mmio_exit_user; 604 u64 mmio_exit_kernel; 605 u64 signal_exits; 606 u64 exits; 607 }; 608 609 void kvm_vcpu_preferred_target(struct kvm_vcpu_init *init); 610 unsigned long kvm_arm_num_regs(struct kvm_vcpu *vcpu); 611 int kvm_arm_copy_reg_indices(struct kvm_vcpu *vcpu, u64 __user *indices); 612 int kvm_arm_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg); 613 int kvm_arm_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg); 614 615 unsigned long kvm_arm_num_sys_reg_descs(struct kvm_vcpu *vcpu); 616 int kvm_arm_copy_sys_reg_indices(struct kvm_vcpu *vcpu, u64 __user *uindices); 617 int kvm_arm_sys_reg_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *); 618 int kvm_arm_sys_reg_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *); 619 620 int __kvm_arm_vcpu_get_events(struct kvm_vcpu *vcpu, 621 struct kvm_vcpu_events *events); 622 623 int __kvm_arm_vcpu_set_events(struct kvm_vcpu *vcpu, 624 struct kvm_vcpu_events *events); 625 626 #define KVM_ARCH_WANT_MMU_NOTIFIER 627 628 void kvm_arm_halt_guest(struct kvm *kvm); 629 void kvm_arm_resume_guest(struct kvm *kvm); 630 631 #define vcpu_has_run_once(vcpu) !!rcu_access_pointer((vcpu)->pid) 632 633 #ifndef __KVM_NVHE_HYPERVISOR__ 634 #define kvm_call_hyp_nvhe(f, ...) \ 635 ({ \ 636 struct arm_smccc_res res; \ 637 \ 638 arm_smccc_1_1_hvc(KVM_HOST_SMCCC_FUNC(f), \ 639 ##__VA_ARGS__, &res); \ 640 WARN_ON(res.a0 != SMCCC_RET_SUCCESS); \ 641 \ 642 res.a1; \ 643 }) 644 645 /* 646 * The couple of isb() below are there to guarantee the same behaviour 647 * on VHE as on !VHE, where the eret to EL1 acts as a context 648 * synchronization event. 649 */ 650 #define kvm_call_hyp(f, ...) \ 651 do { \ 652 if (has_vhe()) { \ 653 f(__VA_ARGS__); \ 654 isb(); \ 655 } else { \ 656 kvm_call_hyp_nvhe(f, ##__VA_ARGS__); \ 657 } \ 658 } while(0) 659 660 #define kvm_call_hyp_ret(f, ...) \ 661 ({ \ 662 typeof(f(__VA_ARGS__)) ret; \ 663 \ 664 if (has_vhe()) { \ 665 ret = f(__VA_ARGS__); \ 666 isb(); \ 667 } else { \ 668 ret = kvm_call_hyp_nvhe(f, ##__VA_ARGS__); \ 669 } \ 670 \ 671 ret; \ 672 }) 673 #else /* __KVM_NVHE_HYPERVISOR__ */ 674 #define kvm_call_hyp(f, ...) f(__VA_ARGS__) 675 #define kvm_call_hyp_ret(f, ...) f(__VA_ARGS__) 676 #define kvm_call_hyp_nvhe(f, ...) f(__VA_ARGS__) 677 #endif /* __KVM_NVHE_HYPERVISOR__ */ 678 679 void force_vm_exit(const cpumask_t *mask); 680 681 int handle_exit(struct kvm_vcpu *vcpu, int exception_index); 682 void handle_exit_early(struct kvm_vcpu *vcpu, int exception_index); 683 684 int kvm_handle_cp14_load_store(struct kvm_vcpu *vcpu); 685 int kvm_handle_cp14_32(struct kvm_vcpu *vcpu); 686 int kvm_handle_cp14_64(struct kvm_vcpu *vcpu); 687 int kvm_handle_cp15_32(struct kvm_vcpu *vcpu); 688 int kvm_handle_cp15_64(struct kvm_vcpu *vcpu); 689 int kvm_handle_sys_reg(struct kvm_vcpu *vcpu); 690 691 void kvm_reset_sys_regs(struct kvm_vcpu *vcpu); 692 693 void kvm_sys_reg_table_init(void); 694 695 /* MMIO helpers */ 696 void kvm_mmio_write_buf(void *buf, unsigned int len, unsigned long data); 697 unsigned long kvm_mmio_read_buf(const void *buf, unsigned int len); 698 699 int kvm_handle_mmio_return(struct kvm_vcpu *vcpu); 700 int io_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa); 701 702 /* 703 * Returns true if a Performance Monitoring Interrupt (PMI), a.k.a. perf event, 704 * arrived in guest context. For arm64, any event that arrives while a vCPU is 705 * loaded is considered to be "in guest". 706 */ 707 static inline bool kvm_arch_pmi_in_guest(struct kvm_vcpu *vcpu) 708 { 709 return IS_ENABLED(CONFIG_GUEST_PERF_EVENTS) && !!vcpu; 710 } 711 712 long kvm_hypercall_pv_features(struct kvm_vcpu *vcpu); 713 gpa_t kvm_init_stolen_time(struct kvm_vcpu *vcpu); 714 void kvm_update_stolen_time(struct kvm_vcpu *vcpu); 715 716 bool kvm_arm_pvtime_supported(void); 717 int kvm_arm_pvtime_set_attr(struct kvm_vcpu *vcpu, 718 struct kvm_device_attr *attr); 719 int kvm_arm_pvtime_get_attr(struct kvm_vcpu *vcpu, 720 struct kvm_device_attr *attr); 721 int kvm_arm_pvtime_has_attr(struct kvm_vcpu *vcpu, 722 struct kvm_device_attr *attr); 723 724 extern unsigned int kvm_arm_vmid_bits; 725 int kvm_arm_vmid_alloc_init(void); 726 void kvm_arm_vmid_alloc_free(void); 727 void kvm_arm_vmid_update(struct kvm_vmid *kvm_vmid); 728 void kvm_arm_vmid_clear_active(void); 729 730 static inline void kvm_arm_pvtime_vcpu_init(struct kvm_vcpu_arch *vcpu_arch) 731 { 732 vcpu_arch->steal.base = GPA_INVALID; 733 } 734 735 static inline bool kvm_arm_is_pvtime_enabled(struct kvm_vcpu_arch *vcpu_arch) 736 { 737 return (vcpu_arch->steal.base != GPA_INVALID); 738 } 739 740 void kvm_set_sei_esr(struct kvm_vcpu *vcpu, u64 syndrome); 741 742 struct kvm_vcpu *kvm_mpidr_to_vcpu(struct kvm *kvm, unsigned long mpidr); 743 744 DECLARE_KVM_HYP_PER_CPU(struct kvm_host_data, kvm_host_data); 745 746 static inline void kvm_init_host_cpu_context(struct kvm_cpu_context *cpu_ctxt) 747 { 748 /* The host's MPIDR is immutable, so let's set it up at boot time */ 749 ctxt_sys_reg(cpu_ctxt, MPIDR_EL1) = read_cpuid_mpidr(); 750 } 751 752 static inline bool kvm_system_needs_idmapped_vectors(void) 753 { 754 return cpus_have_const_cap(ARM64_SPECTRE_V3A); 755 } 756 757 void kvm_arm_vcpu_ptrauth_trap(struct kvm_vcpu *vcpu); 758 759 static inline void kvm_arch_hardware_unsetup(void) {} 760 static inline void kvm_arch_sync_events(struct kvm *kvm) {} 761 static inline void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu) {} 762 763 void kvm_arm_init_debug(void); 764 void kvm_arm_vcpu_init_debug(struct kvm_vcpu *vcpu); 765 void kvm_arm_setup_debug(struct kvm_vcpu *vcpu); 766 void kvm_arm_clear_debug(struct kvm_vcpu *vcpu); 767 void kvm_arm_reset_debug_ptr(struct kvm_vcpu *vcpu); 768 769 #define kvm_vcpu_os_lock_enabled(vcpu) \ 770 (!!(__vcpu_sys_reg(vcpu, OSLSR_EL1) & SYS_OSLSR_OSLK)) 771 772 int kvm_arm_vcpu_arch_set_attr(struct kvm_vcpu *vcpu, 773 struct kvm_device_attr *attr); 774 int kvm_arm_vcpu_arch_get_attr(struct kvm_vcpu *vcpu, 775 struct kvm_device_attr *attr); 776 int kvm_arm_vcpu_arch_has_attr(struct kvm_vcpu *vcpu, 777 struct kvm_device_attr *attr); 778 779 long kvm_vm_ioctl_mte_copy_tags(struct kvm *kvm, 780 struct kvm_arm_copy_mte_tags *copy_tags); 781 782 /* Guest/host FPSIMD coordination helpers */ 783 int kvm_arch_vcpu_run_map_fp(struct kvm_vcpu *vcpu); 784 void kvm_arch_vcpu_load_fp(struct kvm_vcpu *vcpu); 785 void kvm_arch_vcpu_ctxflush_fp(struct kvm_vcpu *vcpu); 786 void kvm_arch_vcpu_ctxsync_fp(struct kvm_vcpu *vcpu); 787 void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu); 788 void kvm_vcpu_unshare_task_fp(struct kvm_vcpu *vcpu); 789 790 static inline bool kvm_pmu_counter_deferred(struct perf_event_attr *attr) 791 { 792 return (!has_vhe() && attr->exclude_host); 793 } 794 795 /* Flags for host debug state */ 796 void kvm_arch_vcpu_load_debug_state_flags(struct kvm_vcpu *vcpu); 797 void kvm_arch_vcpu_put_debug_state_flags(struct kvm_vcpu *vcpu); 798 799 #ifdef CONFIG_KVM 800 void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr); 801 void kvm_clr_pmu_events(u32 clr); 802 803 void kvm_vcpu_pmu_restore_guest(struct kvm_vcpu *vcpu); 804 void kvm_vcpu_pmu_restore_host(struct kvm_vcpu *vcpu); 805 #else 806 static inline void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr) {} 807 static inline void kvm_clr_pmu_events(u32 clr) {} 808 #endif 809 810 void kvm_vcpu_load_sysregs_vhe(struct kvm_vcpu *vcpu); 811 void kvm_vcpu_put_sysregs_vhe(struct kvm_vcpu *vcpu); 812 813 int kvm_set_ipa_limit(void); 814 815 #define __KVM_HAVE_ARCH_VM_ALLOC 816 struct kvm *kvm_arch_alloc_vm(void); 817 818 int kvm_arm_setup_stage2(struct kvm *kvm, unsigned long type); 819 820 static inline bool kvm_vm_is_protected(struct kvm *kvm) 821 { 822 return false; 823 } 824 825 void kvm_init_protected_traps(struct kvm_vcpu *vcpu); 826 827 int kvm_arm_vcpu_finalize(struct kvm_vcpu *vcpu, int feature); 828 bool kvm_arm_vcpu_is_finalized(struct kvm_vcpu *vcpu); 829 830 #define kvm_arm_vcpu_sve_finalized(vcpu) \ 831 ((vcpu)->arch.flags & KVM_ARM64_VCPU_SVE_FINALIZED) 832 833 #define kvm_has_mte(kvm) \ 834 (system_supports_mte() && \ 835 test_bit(KVM_ARCH_FLAG_MTE_ENABLED, &(kvm)->arch.flags)) 836 #define kvm_vcpu_has_pmu(vcpu) \ 837 (test_bit(KVM_ARM_VCPU_PMU_V3, (vcpu)->arch.features)) 838 839 int kvm_trng_call(struct kvm_vcpu *vcpu); 840 #ifdef CONFIG_KVM 841 extern phys_addr_t hyp_mem_base; 842 extern phys_addr_t hyp_mem_size; 843 void __init kvm_hyp_reserve(void); 844 #else 845 static inline void kvm_hyp_reserve(void) { } 846 #endif 847 848 #endif /* __ARM64_KVM_HOST_H__ */ 849