linux/arch/arm64/include/asm/kvm_host.h
<<
>>
Prefs
   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#include <asm/thread_info.h>
  30
  31#define __KVM_HAVE_ARCH_INTC_INITIALIZED
  32
  33#define KVM_HALT_POLL_NS_DEFAULT 500000
  34
  35#include <kvm/arm_vgic.h>
  36#include <kvm/arm_arch_timer.h>
  37#include <kvm/arm_pmu.h>
  38
  39#define KVM_MAX_VCPUS VGIC_V3_MAX_CPUS
  40
  41#define KVM_VCPU_MAX_FEATURES 7
  42
  43#define KVM_REQ_SLEEP \
  44        KVM_ARCH_REQ_FLAGS(0, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
  45#define KVM_REQ_IRQ_PENDING     KVM_ARCH_REQ(1)
  46#define KVM_REQ_VCPU_RESET      KVM_ARCH_REQ(2)
  47#define KVM_REQ_RECORD_STEAL    KVM_ARCH_REQ(3)
  48#define KVM_REQ_RELOAD_GICv4    KVM_ARCH_REQ(4)
  49#define KVM_REQ_RELOAD_PMU      KVM_ARCH_REQ(5)
  50
  51#define KVM_DIRTY_LOG_MANUAL_CAPS   (KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE | \
  52                                     KVM_DIRTY_LOG_INITIALLY_SET)
  53
  54/*
  55 * Mode of operation configurable with kvm-arm.mode early param.
  56 * See Documentation/admin-guide/kernel-parameters.txt for more information.
  57 */
  58enum kvm_mode {
  59        KVM_MODE_DEFAULT,
  60        KVM_MODE_PROTECTED,
  61};
  62enum kvm_mode kvm_get_mode(void);
  63
  64DECLARE_STATIC_KEY_FALSE(userspace_irqchip_in_use);
  65
  66extern unsigned int kvm_sve_max_vl;
  67int kvm_arm_init_sve(void);
  68
  69u32 __attribute_const__ kvm_target_cpu(void);
  70int kvm_reset_vcpu(struct kvm_vcpu *vcpu);
  71void kvm_arm_vcpu_destroy(struct kvm_vcpu *vcpu);
  72
  73struct kvm_vmid {
  74        /* The VMID generation used for the virt. memory system */
  75        u64    vmid_gen;
  76        u32    vmid;
  77};
  78
  79struct 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
 101struct kvm_arch_memory_slot {
 102};
 103
 104struct 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        bool return_nisv_io_abort_to_user;
 126
 127        /*
 128         * VM-wide PMU filter, implemented as a bitmap and big enough for
 129         * up to 2^10 events (ARMv8.0) or 2^16 events (ARMv8.1+).
 130         */
 131        unsigned long *pmu_filter;
 132        unsigned int pmuver;
 133
 134        u8 pfr0_csv2;
 135        u8 pfr0_csv3;
 136
 137        /* Memory Tagging Extension enabled for the guest */
 138        bool mte_enabled;
 139};
 140
 141struct kvm_vcpu_fault_info {
 142        u32 esr_el2;            /* Hyp Syndrom Register */
 143        u64 far_el2;            /* Hyp Fault Address Register */
 144        u64 hpfar_el2;          /* Hyp IPA Fault Address Register */
 145        u64 disr_el1;           /* Deferred [SError] Status Register */
 146};
 147
 148enum vcpu_sysreg {
 149        __INVALID_SYSREG__,   /* 0 is reserved as an invalid value */
 150        MPIDR_EL1,      /* MultiProcessor Affinity Register */
 151        CSSELR_EL1,     /* Cache Size Selection Register */
 152        SCTLR_EL1,      /* System Control Register */
 153        ACTLR_EL1,      /* Auxiliary Control Register */
 154        CPACR_EL1,      /* Coprocessor Access Control */
 155        ZCR_EL1,        /* SVE Control */
 156        TTBR0_EL1,      /* Translation Table Base Register 0 */
 157        TTBR1_EL1,      /* Translation Table Base Register 1 */
 158        TCR_EL1,        /* Translation Control Register */
 159        ESR_EL1,        /* Exception Syndrome Register */
 160        AFSR0_EL1,      /* Auxiliary Fault Status Register 0 */
 161        AFSR1_EL1,      /* Auxiliary Fault Status Register 1 */
 162        FAR_EL1,        /* Fault Address Register */
 163        MAIR_EL1,       /* Memory Attribute Indirection Register */
 164        VBAR_EL1,       /* Vector Base Address Register */
 165        CONTEXTIDR_EL1, /* Context ID Register */
 166        TPIDR_EL0,      /* Thread ID, User R/W */
 167        TPIDRRO_EL0,    /* Thread ID, User R/O */
 168        TPIDR_EL1,      /* Thread ID, Privileged */
 169        AMAIR_EL1,      /* Aux Memory Attribute Indirection Register */
 170        CNTKCTL_EL1,    /* Timer Control Register (EL1) */
 171        PAR_EL1,        /* Physical Address Register */
 172        MDSCR_EL1,      /* Monitor Debug System Control Register */
 173        MDCCINT_EL1,    /* Monitor Debug Comms Channel Interrupt Enable Reg */
 174        DISR_EL1,       /* Deferred Interrupt Status Register */
 175
 176        /* Performance Monitors Registers */
 177        PMCR_EL0,       /* Control Register */
 178        PMSELR_EL0,     /* Event Counter Selection Register */
 179        PMEVCNTR0_EL0,  /* Event Counter Register (0-30) */
 180        PMEVCNTR30_EL0 = PMEVCNTR0_EL0 + 30,
 181        PMCCNTR_EL0,    /* Cycle Counter Register */
 182        PMEVTYPER0_EL0, /* Event Type Register (0-30) */
 183        PMEVTYPER30_EL0 = PMEVTYPER0_EL0 + 30,
 184        PMCCFILTR_EL0,  /* Cycle Count Filter Register */
 185        PMCNTENSET_EL0, /* Count Enable Set Register */
 186        PMINTENSET_EL1, /* Interrupt Enable Set Register */
 187        PMOVSSET_EL0,   /* Overflow Flag Status Set Register */
 188        PMUSERENR_EL0,  /* User Enable Register */
 189
 190        /* Pointer Authentication Registers in a strict increasing order. */
 191        APIAKEYLO_EL1,
 192        APIAKEYHI_EL1,
 193        APIBKEYLO_EL1,
 194        APIBKEYHI_EL1,
 195        APDAKEYLO_EL1,
 196        APDAKEYHI_EL1,
 197        APDBKEYLO_EL1,
 198        APDBKEYHI_EL1,
 199        APGAKEYLO_EL1,
 200        APGAKEYHI_EL1,
 201
 202        ELR_EL1,
 203        SP_EL1,
 204        SPSR_EL1,
 205
 206        CNTVOFF_EL2,
 207        CNTV_CVAL_EL0,
 208        CNTV_CTL_EL0,
 209        CNTP_CVAL_EL0,
 210        CNTP_CTL_EL0,
 211
 212        /* Memory Tagging Extension registers */
 213        RGSR_EL1,       /* Random Allocation Tag Seed Register */
 214        GCR_EL1,        /* Tag Control Register */
 215        TFSR_EL1,       /* Tag Fault Status Register (EL1) */
 216        TFSRE0_EL1,     /* Tag Fault Status Register (EL0) */
 217
 218        /* 32bit specific registers. Keep them at the end of the range */
 219        DACR32_EL2,     /* Domain Access Control Register */
 220        IFSR32_EL2,     /* Instruction Fault Status Register */
 221        FPEXC32_EL2,    /* Floating-Point Exception Control Register */
 222        DBGVCR32_EL2,   /* Debug Vector Catch Register */
 223
 224        NR_SYS_REGS     /* Nothing after this line! */
 225};
 226
 227struct kvm_cpu_context {
 228        struct user_pt_regs regs;       /* sp = sp_el0 */
 229
 230        u64     spsr_abt;
 231        u64     spsr_und;
 232        u64     spsr_irq;
 233        u64     spsr_fiq;
 234
 235        struct user_fpsimd_state fp_regs;
 236
 237        u64 sys_regs[NR_SYS_REGS];
 238
 239        struct kvm_vcpu *__hyp_running_vcpu;
 240};
 241
 242struct kvm_pmu_events {
 243        u32 events_host;
 244        u32 events_guest;
 245};
 246
 247struct kvm_host_data {
 248        struct kvm_cpu_context host_ctxt;
 249        struct kvm_pmu_events pmu_events;
 250};
 251
 252struct kvm_host_psci_config {
 253        /* PSCI version used by host. */
 254        u32 version;
 255
 256        /* Function IDs used by host if version is v0.1. */
 257        struct psci_0_1_function_ids function_ids_0_1;
 258
 259        bool psci_0_1_cpu_suspend_implemented;
 260        bool psci_0_1_cpu_on_implemented;
 261        bool psci_0_1_cpu_off_implemented;
 262        bool psci_0_1_migrate_implemented;
 263};
 264
 265extern struct kvm_host_psci_config kvm_nvhe_sym(kvm_host_psci_config);
 266#define kvm_host_psci_config CHOOSE_NVHE_SYM(kvm_host_psci_config)
 267
 268extern s64 kvm_nvhe_sym(hyp_physvirt_offset);
 269#define hyp_physvirt_offset CHOOSE_NVHE_SYM(hyp_physvirt_offset)
 270
 271extern u64 kvm_nvhe_sym(hyp_cpu_logical_map)[NR_CPUS];
 272#define hyp_cpu_logical_map CHOOSE_NVHE_SYM(hyp_cpu_logical_map)
 273
 274struct vcpu_reset_state {
 275        unsigned long   pc;
 276        unsigned long   r0;
 277        bool            be;
 278        bool            reset;
 279};
 280
 281struct kvm_vcpu_arch {
 282        struct kvm_cpu_context ctxt;
 283        void *sve_state;
 284        unsigned int sve_max_vl;
 285
 286        /* Stage 2 paging state used by the hardware on next switch */
 287        struct kvm_s2_mmu *hw_mmu;
 288
 289        /* Values of trap registers for the guest. */
 290        u64 hcr_el2;
 291        u64 mdcr_el2;
 292        u64 cptr_el2;
 293
 294        /* Values of trap registers for the host before guest entry. */
 295        u64 mdcr_el2_host;
 296
 297        /* Exception Information */
 298        struct kvm_vcpu_fault_info fault;
 299
 300        /* State of various workarounds, see kvm_asm.h for bit assignment */
 301        u64 workaround_flags;
 302
 303        /* Miscellaneous vcpu state flags */
 304        u64 flags;
 305
 306        /*
 307         * We maintain more than a single set of debug registers to support
 308         * debugging the guest from the host and to maintain separate host and
 309         * guest state during world switches. vcpu_debug_state are the debug
 310         * registers of the vcpu as the guest sees them.  host_debug_state are
 311         * the host registers which are saved and restored during
 312         * world switches. external_debug_state contains the debug
 313         * values we want to debug the guest. This is set via the
 314         * KVM_SET_GUEST_DEBUG ioctl.
 315         *
 316         * debug_ptr points to the set of debug registers that should be loaded
 317         * onto the hardware when running the guest.
 318         */
 319        struct kvm_guest_debug_arch *debug_ptr;
 320        struct kvm_guest_debug_arch vcpu_debug_state;
 321        struct kvm_guest_debug_arch external_debug_state;
 322
 323        struct thread_info *host_thread_info;   /* hyp VA */
 324        struct user_fpsimd_state *host_fpsimd_state;    /* hyp VA */
 325
 326        struct {
 327                /* {Break,watch}point registers */
 328                struct kvm_guest_debug_arch regs;
 329                /* Statistical profiling extension */
 330                u64 pmscr_el1;
 331                /* Self-hosted trace */
 332                u64 trfcr_el1;
 333        } host_debug_state;
 334
 335        /* VGIC state */
 336        struct vgic_cpu vgic_cpu;
 337        struct arch_timer_cpu timer_cpu;
 338        struct kvm_pmu pmu;
 339
 340        /*
 341         * Anything that is not used directly from assembly code goes
 342         * here.
 343         */
 344
 345        /*
 346         * Guest registers we preserve during guest debugging.
 347         *
 348         * These shadow registers are updated by the kvm_handle_sys_reg
 349         * trap handler if the guest accesses or updates them while we
 350         * are using guest debug.
 351         */
 352        struct {
 353                u32     mdscr_el1;
 354        } guest_debug_preserved;
 355
 356        /* vcpu power-off state */
 357        bool power_off;
 358
 359        /* Don't run the guest (internal implementation need) */
 360        bool pause;
 361
 362        /* Cache some mmu pages needed inside spinlock regions */
 363        struct kvm_mmu_memory_cache mmu_page_cache;
 364
 365        /* Target CPU and feature flags */
 366        int target;
 367        DECLARE_BITMAP(features, KVM_VCPU_MAX_FEATURES);
 368
 369        /* Detect first run of a vcpu */
 370        bool has_run_once;
 371
 372        /* Virtual SError ESR to restore when HCR_EL2.VSE is set */
 373        u64 vsesr_el2;
 374
 375        /* Additional reset state */
 376        struct vcpu_reset_state reset_state;
 377
 378        /* True when deferrable sysregs are loaded on the physical CPU,
 379         * see kvm_vcpu_load_sysregs_vhe and kvm_vcpu_put_sysregs_vhe. */
 380        bool sysregs_loaded_on_cpu;
 381
 382        /* Guest PV state */
 383        struct {
 384                u64 last_steal;
 385                gpa_t base;
 386        } steal;
 387};
 388
 389/* Pointer to the vcpu's SVE FFR for sve_{save,load}_state() */
 390#define vcpu_sve_pffr(vcpu) (kern_hyp_va((vcpu)->arch.sve_state) +      \
 391                             sve_ffr_offset((vcpu)->arch.sve_max_vl))
 392
 393#define vcpu_sve_max_vq(vcpu)   sve_vq_from_vl((vcpu)->arch.sve_max_vl)
 394
 395#define vcpu_sve_state_size(vcpu) ({                                    \
 396        size_t __size_ret;                                              \
 397        unsigned int __vcpu_vq;                                         \
 398                                                                        \
 399        if (WARN_ON(!sve_vl_valid((vcpu)->arch.sve_max_vl))) {          \
 400                __size_ret = 0;                                         \
 401        } else {                                                        \
 402                __vcpu_vq = vcpu_sve_max_vq(vcpu);                      \
 403                __size_ret = SVE_SIG_REGS_SIZE(__vcpu_vq);              \
 404        }                                                               \
 405                                                                        \
 406        __size_ret;                                                     \
 407})
 408
 409/* vcpu_arch flags field values: */
 410#define KVM_ARM64_DEBUG_DIRTY           (1 << 0)
 411#define KVM_ARM64_FP_ENABLED            (1 << 1) /* guest FP regs loaded */
 412#define KVM_ARM64_FP_HOST               (1 << 2) /* host FP regs loaded */
 413#define KVM_ARM64_HOST_SVE_IN_USE       (1 << 3) /* backup for host TIF_SVE */
 414#define KVM_ARM64_HOST_SVE_ENABLED      (1 << 4) /* SVE enabled for EL0 */
 415#define KVM_ARM64_GUEST_HAS_SVE         (1 << 5) /* SVE exposed to guest */
 416#define KVM_ARM64_VCPU_SVE_FINALIZED    (1 << 6) /* SVE config completed */
 417#define KVM_ARM64_GUEST_HAS_PTRAUTH     (1 << 7) /* PTRAUTH exposed to guest */
 418#define KVM_ARM64_PENDING_EXCEPTION     (1 << 8) /* Exception pending */
 419#define KVM_ARM64_EXCEPT_MASK           (7 << 9) /* Target EL/MODE */
 420#define KVM_ARM64_DEBUG_STATE_SAVE_SPE  (1 << 12) /* Save SPE context if active  */
 421#define KVM_ARM64_DEBUG_STATE_SAVE_TRBE (1 << 13) /* Save TRBE context if active  */
 422
 423#define KVM_GUESTDBG_VALID_MASK (KVM_GUESTDBG_ENABLE | \
 424                                 KVM_GUESTDBG_USE_SW_BP | \
 425                                 KVM_GUESTDBG_USE_HW | \
 426                                 KVM_GUESTDBG_SINGLESTEP)
 427/*
 428 * When KVM_ARM64_PENDING_EXCEPTION is set, KVM_ARM64_EXCEPT_MASK can
 429 * take the following values:
 430 *
 431 * For AArch32 EL1:
 432 */
 433#define KVM_ARM64_EXCEPT_AA32_UND       (0 << 9)
 434#define KVM_ARM64_EXCEPT_AA32_IABT      (1 << 9)
 435#define KVM_ARM64_EXCEPT_AA32_DABT      (2 << 9)
 436/* For AArch64: */
 437#define KVM_ARM64_EXCEPT_AA64_ELx_SYNC  (0 << 9)
 438#define KVM_ARM64_EXCEPT_AA64_ELx_IRQ   (1 << 9)
 439#define KVM_ARM64_EXCEPT_AA64_ELx_FIQ   (2 << 9)
 440#define KVM_ARM64_EXCEPT_AA64_ELx_SERR  (3 << 9)
 441#define KVM_ARM64_EXCEPT_AA64_EL1       (0 << 11)
 442#define KVM_ARM64_EXCEPT_AA64_EL2       (1 << 11)
 443
 444/*
 445 * Overlaps with KVM_ARM64_EXCEPT_MASK on purpose so that it can't be
 446 * set together with an exception...
 447 */
 448#define KVM_ARM64_INCREMENT_PC          (1 << 9) /* Increment PC */
 449
 450#define vcpu_has_sve(vcpu) (system_supports_sve() &&                    \
 451                            ((vcpu)->arch.flags & KVM_ARM64_GUEST_HAS_SVE))
 452
 453#ifdef CONFIG_ARM64_PTR_AUTH
 454#define vcpu_has_ptrauth(vcpu)                                          \
 455        ((cpus_have_final_cap(ARM64_HAS_ADDRESS_AUTH) ||                \
 456          cpus_have_final_cap(ARM64_HAS_GENERIC_AUTH)) &&               \
 457         (vcpu)->arch.flags & KVM_ARM64_GUEST_HAS_PTRAUTH)
 458#else
 459#define vcpu_has_ptrauth(vcpu)          false
 460#endif
 461
 462#define vcpu_gp_regs(v)         (&(v)->arch.ctxt.regs)
 463
 464/*
 465 * Only use __vcpu_sys_reg/ctxt_sys_reg if you know you want the
 466 * memory backed version of a register, and not the one most recently
 467 * accessed by a running VCPU.  For example, for userspace access or
 468 * for system registers that are never context switched, but only
 469 * emulated.
 470 */
 471#define __ctxt_sys_reg(c,r)     (&(c)->sys_regs[(r)])
 472
 473#define ctxt_sys_reg(c,r)       (*__ctxt_sys_reg(c,r))
 474
 475#define __vcpu_sys_reg(v,r)     (ctxt_sys_reg(&(v)->arch.ctxt, (r)))
 476
 477u64 vcpu_read_sys_reg(const struct kvm_vcpu *vcpu, int reg);
 478void vcpu_write_sys_reg(struct kvm_vcpu *vcpu, u64 val, int reg);
 479
 480static inline bool __vcpu_read_sys_reg_from_cpu(int reg, u64 *val)
 481{
 482        /*
 483         * *** VHE ONLY ***
 484         *
 485         * System registers listed in the switch are not saved on every
 486         * exit from the guest but are only saved on vcpu_put.
 487         *
 488         * Note that MPIDR_EL1 for the guest is set by KVM via VMPIDR_EL2 but
 489         * should never be listed below, because the guest cannot modify its
 490         * own MPIDR_EL1 and MPIDR_EL1 is accessed for VCPU A from VCPU B's
 491         * thread when emulating cross-VCPU communication.
 492         */
 493        if (!has_vhe())
 494                return false;
 495
 496        switch (reg) {
 497        case CSSELR_EL1:        *val = read_sysreg_s(SYS_CSSELR_EL1);   break;
 498        case SCTLR_EL1:         *val = read_sysreg_s(SYS_SCTLR_EL12);   break;
 499        case CPACR_EL1:         *val = read_sysreg_s(SYS_CPACR_EL12);   break;
 500        case TTBR0_EL1:         *val = read_sysreg_s(SYS_TTBR0_EL12);   break;
 501        case TTBR1_EL1:         *val = read_sysreg_s(SYS_TTBR1_EL12);   break;
 502        case TCR_EL1:           *val = read_sysreg_s(SYS_TCR_EL12);     break;
 503        case ESR_EL1:           *val = read_sysreg_s(SYS_ESR_EL12);     break;
 504        case AFSR0_EL1:         *val = read_sysreg_s(SYS_AFSR0_EL12);   break;
 505        case AFSR1_EL1:         *val = read_sysreg_s(SYS_AFSR1_EL12);   break;
 506        case FAR_EL1:           *val = read_sysreg_s(SYS_FAR_EL12);     break;
 507        case MAIR_EL1:          *val = read_sysreg_s(SYS_MAIR_EL12);    break;
 508        case VBAR_EL1:          *val = read_sysreg_s(SYS_VBAR_EL12);    break;
 509        case CONTEXTIDR_EL1:    *val = read_sysreg_s(SYS_CONTEXTIDR_EL12);break;
 510        case TPIDR_EL0:         *val = read_sysreg_s(SYS_TPIDR_EL0);    break;
 511        case TPIDRRO_EL0:       *val = read_sysreg_s(SYS_TPIDRRO_EL0);  break;
 512        case TPIDR_EL1:         *val = read_sysreg_s(SYS_TPIDR_EL1);    break;
 513        case AMAIR_EL1:         *val = read_sysreg_s(SYS_AMAIR_EL12);   break;
 514        case CNTKCTL_EL1:       *val = read_sysreg_s(SYS_CNTKCTL_EL12); break;
 515        case ELR_EL1:           *val = read_sysreg_s(SYS_ELR_EL12);     break;
 516        case PAR_EL1:           *val = read_sysreg_par();               break;
 517        case DACR32_EL2:        *val = read_sysreg_s(SYS_DACR32_EL2);   break;
 518        case IFSR32_EL2:        *val = read_sysreg_s(SYS_IFSR32_EL2);   break;
 519        case DBGVCR32_EL2:      *val = read_sysreg_s(SYS_DBGVCR32_EL2); break;
 520        default:                return false;
 521        }
 522
 523        return true;
 524}
 525
 526static inline bool __vcpu_write_sys_reg_to_cpu(u64 val, int reg)
 527{
 528        /*
 529         * *** VHE ONLY ***
 530         *
 531         * System registers listed in the switch are not restored on every
 532         * entry to the guest but are only restored on vcpu_load.
 533         *
 534         * Note that MPIDR_EL1 for the guest is set by KVM via VMPIDR_EL2 but
 535         * should never be listed below, because the MPIDR should only be set
 536         * once, before running the VCPU, and never changed later.
 537         */
 538        if (!has_vhe())
 539                return false;
 540
 541        switch (reg) {
 542        case CSSELR_EL1:        write_sysreg_s(val, SYS_CSSELR_EL1);    break;
 543        case SCTLR_EL1:         write_sysreg_s(val, SYS_SCTLR_EL12);    break;
 544        case CPACR_EL1:         write_sysreg_s(val, SYS_CPACR_EL12);    break;
 545        case TTBR0_EL1:         write_sysreg_s(val, SYS_TTBR0_EL12);    break;
 546        case TTBR1_EL1:         write_sysreg_s(val, SYS_TTBR1_EL12);    break;
 547        case TCR_EL1:           write_sysreg_s(val, SYS_TCR_EL12);      break;
 548        case ESR_EL1:           write_sysreg_s(val, SYS_ESR_EL12);      break;
 549        case AFSR0_EL1:         write_sysreg_s(val, SYS_AFSR0_EL12);    break;
 550        case AFSR1_EL1:         write_sysreg_s(val, SYS_AFSR1_EL12);    break;
 551        case FAR_EL1:           write_sysreg_s(val, SYS_FAR_EL12);      break;
 552        case MAIR_EL1:          write_sysreg_s(val, SYS_MAIR_EL12);     break;
 553        case VBAR_EL1:          write_sysreg_s(val, SYS_VBAR_EL12);     break;
 554        case CONTEXTIDR_EL1:    write_sysreg_s(val, SYS_CONTEXTIDR_EL12);break;
 555        case TPIDR_EL0:         write_sysreg_s(val, SYS_TPIDR_EL0);     break;
 556        case TPIDRRO_EL0:       write_sysreg_s(val, SYS_TPIDRRO_EL0);   break;
 557        case TPIDR_EL1:         write_sysreg_s(val, SYS_TPIDR_EL1);     break;
 558        case AMAIR_EL1:         write_sysreg_s(val, SYS_AMAIR_EL12);    break;
 559        case CNTKCTL_EL1:       write_sysreg_s(val, SYS_CNTKCTL_EL12);  break;
 560        case ELR_EL1:           write_sysreg_s(val, SYS_ELR_EL12);      break;
 561        case PAR_EL1:           write_sysreg_s(val, SYS_PAR_EL1);       break;
 562        case DACR32_EL2:        write_sysreg_s(val, SYS_DACR32_EL2);    break;
 563        case IFSR32_EL2:        write_sysreg_s(val, SYS_IFSR32_EL2);    break;
 564        case DBGVCR32_EL2:      write_sysreg_s(val, SYS_DBGVCR32_EL2);  break;
 565        default:                return false;
 566        }
 567
 568        return true;
 569}
 570
 571struct kvm_vm_stat {
 572        struct kvm_vm_stat_generic generic;
 573};
 574
 575struct kvm_vcpu_stat {
 576        struct kvm_vcpu_stat_generic generic;
 577        u64 hvc_exit_stat;
 578        u64 wfe_exit_stat;
 579        u64 wfi_exit_stat;
 580        u64 mmio_exit_user;
 581        u64 mmio_exit_kernel;
 582        u64 signal_exits;
 583        u64 exits;
 584};
 585
 586int kvm_vcpu_preferred_target(struct kvm_vcpu_init *init);
 587unsigned long kvm_arm_num_regs(struct kvm_vcpu *vcpu);
 588int kvm_arm_copy_reg_indices(struct kvm_vcpu *vcpu, u64 __user *indices);
 589int kvm_arm_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
 590int kvm_arm_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
 591
 592unsigned long kvm_arm_num_sys_reg_descs(struct kvm_vcpu *vcpu);
 593int kvm_arm_copy_sys_reg_indices(struct kvm_vcpu *vcpu, u64 __user *uindices);
 594int kvm_arm_sys_reg_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *);
 595int kvm_arm_sys_reg_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *);
 596
 597int __kvm_arm_vcpu_get_events(struct kvm_vcpu *vcpu,
 598                              struct kvm_vcpu_events *events);
 599
 600int __kvm_arm_vcpu_set_events(struct kvm_vcpu *vcpu,
 601                              struct kvm_vcpu_events *events);
 602
 603#define KVM_ARCH_WANT_MMU_NOTIFIER
 604
 605void kvm_arm_halt_guest(struct kvm *kvm);
 606void kvm_arm_resume_guest(struct kvm *kvm);
 607
 608#ifndef __KVM_NVHE_HYPERVISOR__
 609#define kvm_call_hyp_nvhe(f, ...)                                               \
 610        ({                                                              \
 611                struct arm_smccc_res res;                               \
 612                                                                        \
 613                arm_smccc_1_1_hvc(KVM_HOST_SMCCC_FUNC(f),               \
 614                                  ##__VA_ARGS__, &res);                 \
 615                WARN_ON(res.a0 != SMCCC_RET_SUCCESS);                   \
 616                                                                        \
 617                res.a1;                                                 \
 618        })
 619
 620/*
 621 * The couple of isb() below are there to guarantee the same behaviour
 622 * on VHE as on !VHE, where the eret to EL1 acts as a context
 623 * synchronization event.
 624 */
 625#define kvm_call_hyp(f, ...)                                            \
 626        do {                                                            \
 627                if (has_vhe()) {                                        \
 628                        f(__VA_ARGS__);                                 \
 629                        isb();                                          \
 630                } else {                                                \
 631                        kvm_call_hyp_nvhe(f, ##__VA_ARGS__);            \
 632                }                                                       \
 633        } while(0)
 634
 635#define kvm_call_hyp_ret(f, ...)                                        \
 636        ({                                                              \
 637                typeof(f(__VA_ARGS__)) ret;                             \
 638                                                                        \
 639                if (has_vhe()) {                                        \
 640                        ret = f(__VA_ARGS__);                           \
 641                        isb();                                          \
 642                } else {                                                \
 643                        ret = kvm_call_hyp_nvhe(f, ##__VA_ARGS__);      \
 644                }                                                       \
 645                                                                        \
 646                ret;                                                    \
 647        })
 648#else /* __KVM_NVHE_HYPERVISOR__ */
 649#define kvm_call_hyp(f, ...) f(__VA_ARGS__)
 650#define kvm_call_hyp_ret(f, ...) f(__VA_ARGS__)
 651#define kvm_call_hyp_nvhe(f, ...) f(__VA_ARGS__)
 652#endif /* __KVM_NVHE_HYPERVISOR__ */
 653
 654void force_vm_exit(const cpumask_t *mask);
 655
 656int handle_exit(struct kvm_vcpu *vcpu, int exception_index);
 657void handle_exit_early(struct kvm_vcpu *vcpu, int exception_index);
 658
 659int kvm_handle_cp14_load_store(struct kvm_vcpu *vcpu);
 660int kvm_handle_cp14_32(struct kvm_vcpu *vcpu);
 661int kvm_handle_cp14_64(struct kvm_vcpu *vcpu);
 662int kvm_handle_cp15_32(struct kvm_vcpu *vcpu);
 663int kvm_handle_cp15_64(struct kvm_vcpu *vcpu);
 664int kvm_handle_sys_reg(struct kvm_vcpu *vcpu);
 665
 666void kvm_reset_sys_regs(struct kvm_vcpu *vcpu);
 667
 668void kvm_sys_reg_table_init(void);
 669
 670/* MMIO helpers */
 671void kvm_mmio_write_buf(void *buf, unsigned int len, unsigned long data);
 672unsigned long kvm_mmio_read_buf(const void *buf, unsigned int len);
 673
 674int kvm_handle_mmio_return(struct kvm_vcpu *vcpu);
 675int io_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa);
 676
 677int kvm_perf_init(void);
 678int kvm_perf_teardown(void);
 679
 680long kvm_hypercall_pv_features(struct kvm_vcpu *vcpu);
 681gpa_t kvm_init_stolen_time(struct kvm_vcpu *vcpu);
 682void kvm_update_stolen_time(struct kvm_vcpu *vcpu);
 683
 684bool kvm_arm_pvtime_supported(void);
 685int kvm_arm_pvtime_set_attr(struct kvm_vcpu *vcpu,
 686                            struct kvm_device_attr *attr);
 687int kvm_arm_pvtime_get_attr(struct kvm_vcpu *vcpu,
 688                            struct kvm_device_attr *attr);
 689int kvm_arm_pvtime_has_attr(struct kvm_vcpu *vcpu,
 690                            struct kvm_device_attr *attr);
 691
 692static inline void kvm_arm_pvtime_vcpu_init(struct kvm_vcpu_arch *vcpu_arch)
 693{
 694        vcpu_arch->steal.base = GPA_INVALID;
 695}
 696
 697static inline bool kvm_arm_is_pvtime_enabled(struct kvm_vcpu_arch *vcpu_arch)
 698{
 699        return (vcpu_arch->steal.base != GPA_INVALID);
 700}
 701
 702void kvm_set_sei_esr(struct kvm_vcpu *vcpu, u64 syndrome);
 703
 704struct kvm_vcpu *kvm_mpidr_to_vcpu(struct kvm *kvm, unsigned long mpidr);
 705
 706DECLARE_KVM_HYP_PER_CPU(struct kvm_host_data, kvm_host_data);
 707
 708static inline void kvm_init_host_cpu_context(struct kvm_cpu_context *cpu_ctxt)
 709{
 710        /* The host's MPIDR is immutable, so let's set it up at boot time */
 711        ctxt_sys_reg(cpu_ctxt, MPIDR_EL1) = read_cpuid_mpidr();
 712}
 713
 714void kvm_arm_vcpu_ptrauth_trap(struct kvm_vcpu *vcpu);
 715
 716static inline void kvm_arch_hardware_unsetup(void) {}
 717static inline void kvm_arch_sync_events(struct kvm *kvm) {}
 718static inline void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu) {}
 719static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
 720
 721void kvm_arm_init_debug(void);
 722void kvm_arm_vcpu_init_debug(struct kvm_vcpu *vcpu);
 723void kvm_arm_setup_debug(struct kvm_vcpu *vcpu);
 724void kvm_arm_clear_debug(struct kvm_vcpu *vcpu);
 725void kvm_arm_reset_debug_ptr(struct kvm_vcpu *vcpu);
 726int kvm_arm_vcpu_arch_set_attr(struct kvm_vcpu *vcpu,
 727                               struct kvm_device_attr *attr);
 728int kvm_arm_vcpu_arch_get_attr(struct kvm_vcpu *vcpu,
 729                               struct kvm_device_attr *attr);
 730int kvm_arm_vcpu_arch_has_attr(struct kvm_vcpu *vcpu,
 731                               struct kvm_device_attr *attr);
 732
 733long kvm_vm_ioctl_mte_copy_tags(struct kvm *kvm,
 734                                struct kvm_arm_copy_mte_tags *copy_tags);
 735
 736/* Guest/host FPSIMD coordination helpers */
 737int kvm_arch_vcpu_run_map_fp(struct kvm_vcpu *vcpu);
 738void kvm_arch_vcpu_load_fp(struct kvm_vcpu *vcpu);
 739void kvm_arch_vcpu_ctxsync_fp(struct kvm_vcpu *vcpu);
 740void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu);
 741
 742static inline bool kvm_pmu_counter_deferred(struct perf_event_attr *attr)
 743{
 744        return (!has_vhe() && attr->exclude_host);
 745}
 746
 747/* Flags for host debug state */
 748void kvm_arch_vcpu_load_debug_state_flags(struct kvm_vcpu *vcpu);
 749void kvm_arch_vcpu_put_debug_state_flags(struct kvm_vcpu *vcpu);
 750
 751#ifdef CONFIG_KVM /* Avoid conflicts with core headers if CONFIG_KVM=n */
 752static inline int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu)
 753{
 754        return kvm_arch_vcpu_run_map_fp(vcpu);
 755}
 756
 757void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr);
 758void kvm_clr_pmu_events(u32 clr);
 759
 760void kvm_vcpu_pmu_restore_guest(struct kvm_vcpu *vcpu);
 761void kvm_vcpu_pmu_restore_host(struct kvm_vcpu *vcpu);
 762#else
 763static inline void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr) {}
 764static inline void kvm_clr_pmu_events(u32 clr) {}
 765#endif
 766
 767void kvm_vcpu_load_sysregs_vhe(struct kvm_vcpu *vcpu);
 768void kvm_vcpu_put_sysregs_vhe(struct kvm_vcpu *vcpu);
 769
 770int kvm_set_ipa_limit(void);
 771
 772#define __KVM_HAVE_ARCH_VM_ALLOC
 773struct kvm *kvm_arch_alloc_vm(void);
 774void kvm_arch_free_vm(struct kvm *kvm);
 775
 776int kvm_arm_setup_stage2(struct kvm *kvm, unsigned long type);
 777
 778static inline bool kvm_vm_is_protected(struct kvm *kvm)
 779{
 780        return false;
 781}
 782
 783int kvm_arm_vcpu_finalize(struct kvm_vcpu *vcpu, int feature);
 784bool kvm_arm_vcpu_is_finalized(struct kvm_vcpu *vcpu);
 785
 786#define kvm_arm_vcpu_sve_finalized(vcpu) \
 787        ((vcpu)->arch.flags & KVM_ARM64_VCPU_SVE_FINALIZED)
 788
 789#define kvm_has_mte(kvm) (system_supports_mte() && (kvm)->arch.mte_enabled)
 790#define kvm_vcpu_has_pmu(vcpu)                                  \
 791        (test_bit(KVM_ARM_VCPU_PMU_V3, (vcpu)->arch.features))
 792
 793int kvm_trng_call(struct kvm_vcpu *vcpu);
 794#ifdef CONFIG_KVM
 795extern phys_addr_t hyp_mem_base;
 796extern phys_addr_t hyp_mem_size;
 797void __init kvm_hyp_reserve(void);
 798#else
 799static inline void kvm_hyp_reserve(void) { }
 800#endif
 801
 802#endif /* __ARM64_KVM_HOST_H__ */
 803