arch/arm64/include/asm/kvm_host.h

   1 /*
   2  * Copyright (C) 2012,2013 - ARM Ltd
   3  * Author: Marc Zyngier <marc.zyngier@arm.com>
   4  *
   5  * Derived from arch/arm/include/asm/kvm_host.h:
   6  * Copyright (C) 2012 - Virtual Open Systems and Columbia University
   7  * Author: Christoffer Dall <c.dall@virtualopensystems.com>
   8  *
   9  * This program is free software; you can redistribute it and/or modify
  10  * it under the terms of the GNU General Public License version 2 as
  11  * published by the Free Software Foundation.
  12  *
  13  * This program is distributed in the hope that it will be useful,
  14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  16  * GNU General Public License for more details.
  17  *
  18  * You should have received a copy of the GNU General Public License
  19  * along with this program.  If not, see <http://www.gnu.org/licenses/>.
  20  */
  21
  22 #ifndef __ARM64_KVM_HOST_H__
  23 #define __ARM64_KVM_HOST_H__
  24
  25 #include <linux/types.h>
  26 #include <linux/kvm_types.h>
  27 #include <asm/cpufeature.h>
  28 #include <asm/daifflags.h>
  29 #include <asm/fpsimd.h>
  30 #include <asm/kvm.h>
  31 #include <asm/kvm_asm.h>
  32 #include <asm/kvm_mmio.h>
  33
  34 #define __KVM_HAVE_ARCH_INTC_INITIALIZED
  35
  36 #define KVM_USER_MEM_SLOTS 512
  37 #define KVM_HALT_POLL_NS_DEFAULT 500000
  38
  39 #include <kvm/arm_vgic.h>
  40 #include <kvm/arm_arch_timer.h>
  41 #include <kvm/arm_pmu.h>
  42
  43 #define KVM_MAX_VCPUS VGIC_V3_MAX_CPUS
  44
  45 #define KVM_VCPU_MAX_FEATURES 4
  46
  47 #define KVM_REQ_SLEEP \
  48         KVM_ARCH_REQ_FLAGS(0, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
  49 #define KVM_REQ_IRQ_PENDING     KVM_ARCH_REQ(1)
  50
  51 DECLARE_STATIC_KEY_FALSE(userspace_irqchip_in_use);
  52
  53 int __attribute_const__ kvm_target_cpu(void);
  54 int kvm_reset_vcpu(struct kvm_vcpu *vcpu);
  55 int kvm_arch_dev_ioctl_check_extension(struct kvm *kvm, long ext);
  56 void __extended_idmap_trampoline(phys_addr_t boot_pgd, phys_addr_t idmap_start);
  57
  58 struct kvm_arch {
  59         /* The VMID generation used for the virt. memory system */
  60         u64    vmid_gen;
  61         u32    vmid;
  62
  63         /* 1-level 2nd stage table and lock */
  64         spinlock_t pgd_lock;
  65         pgd_t *pgd;
  66
  67         /* VTTBR value associated with above pgd and vmid */
  68         u64    vttbr;
  69
  70         /* The last vcpu id that ran on each physical CPU */
  71         int __percpu *last_vcpu_ran;
  72
  73         /* The maximum number of vCPUs depends on the used GIC model */
  74         int max_vcpus;
  75
  76         /* Interrupt controller */
  77         struct vgic_dist        vgic;
  78 };
  79
  80 #define KVM_NR_MEM_OBJS     40
  81
  82 /*
  83  * We don't want allocation failures within the mmu code, so we preallocate
  84  * enough memory for a single page fault in a cache.
  85  */
  86 struct kvm_mmu_memory_cache {
  87         int nobjs;
  88         void *objects[KVM_NR_MEM_OBJS];
  89 };
  90
  91 struct kvm_vcpu_fault_info {
  92         u32 esr_el2;            /* Hyp Syndrom Register */
  93         u64 far_el2;            /* Hyp Fault Address Register */
  94         u64 hpfar_el2;          /* Hyp IPA Fault Address Register */
  95         u64 disr_el1;           /* Deferred [SError] Status Register */
  96 };
  97
  98 /*
  99  * 0 is reserved as an invalid value.
 100  * Order should be kept in sync with the save/restore code.
 101  */
 102 enum vcpu_sysreg {
 103         __INVALID_SYSREG__,
 104         MPIDR_EL1,      /* MultiProcessor Affinity Register */
 105         CSSELR_EL1,     /* Cache Size Selection Register */
 106         SCTLR_EL1,      /* System Control Register */
 107         ACTLR_EL1,      /* Auxiliary Control Register */
 108         CPACR_EL1,      /* Coprocessor Access Control */
 109         TTBR0_EL1,      /* Translation Table Base Register 0 */
 110         TTBR1_EL1,      /* Translation Table Base Register 1 */
 111         TCR_EL1,        /* Translation Control Register */
 112         ESR_EL1,        /* Exception Syndrome Register */
 113         AFSR0_EL1,      /* Auxiliary Fault Status Register 0 */
 114         AFSR1_EL1,      /* Auxiliary Fault Status Register 1 */
 115         FAR_EL1,        /* Fault Address Register */
 116         MAIR_EL1,       /* Memory Attribute Indirection Register */
 117         VBAR_EL1,       /* Vector Base Address Register */
 118         CONTEXTIDR_EL1, /* Context ID Register */
 119         TPIDR_EL0,      /* Thread ID, User R/W */
 120         TPIDRRO_EL0,    /* Thread ID, User R/O */
 121         TPIDR_EL1,      /* Thread ID, Privileged */
 122         AMAIR_EL1,      /* Aux Memory Attribute Indirection Register */
 123         CNTKCTL_EL1,    /* Timer Control Register (EL1) */
 124         PAR_EL1,        /* Physical Address Register */
 125         MDSCR_EL1,      /* Monitor Debug System Control Register */
 126         MDCCINT_EL1,    /* Monitor Debug Comms Channel Interrupt Enable Reg */
 127         DISR_EL1,       /* Deferred Interrupt Status Register */
 128
 129         /* Performance Monitors Registers */
 130         PMCR_EL0,       /* Control Register */
 131         PMSELR_EL0,     /* Event Counter Selection Register */
 132         PMEVCNTR0_EL0,  /* Event Counter Register (0-30) */
 133         PMEVCNTR30_EL0 = PMEVCNTR0_EL0 + 30,
 134         PMCCNTR_EL0,    /* Cycle Counter Register */
 135         PMEVTYPER0_EL0, /* Event Type Register (0-30) */
 136         PMEVTYPER30_EL0 = PMEVTYPER0_EL0 + 30,
 137         PMCCFILTR_EL0,  /* Cycle Count Filter Register */
 138         PMCNTENSET_EL0, /* Count Enable Set Register */
 139         PMINTENSET_EL1, /* Interrupt Enable Set Register */
 140         PMOVSSET_EL0,   /* Overflow Flag Status Set Register */
 141         PMSWINC_EL0,    /* Software Increment Register */
 142         PMUSERENR_EL0,  /* User Enable Register */
 143
 144         /* 32bit specific registers. Keep them at the end of the range */
 145         DACR32_EL2,     /* Domain Access Control Register */
 146         IFSR32_EL2,     /* Instruction Fault Status Register */
 147         FPEXC32_EL2,    /* Floating-Point Exception Control Register */
 148         DBGVCR32_EL2,   /* Debug Vector Catch Register */
 149
 150         NR_SYS_REGS     /* Nothing after this line! */
 151 };
 152
 153 /* 32bit mapping */
 154 #define c0_MPIDR        (MPIDR_EL1 * 2) /* MultiProcessor ID Register */
 155 #define c0_CSSELR       (CSSELR_EL1 * 2)/* Cache Size Selection Register */
 156 #define c1_SCTLR        (SCTLR_EL1 * 2) /* System Control Register */
 157 #define c1_ACTLR        (ACTLR_EL1 * 2) /* Auxiliary Control Register */
 158 #define c1_CPACR        (CPACR_EL1 * 2) /* Coprocessor Access Control */
 159 #define c2_TTBR0        (TTBR0_EL1 * 2) /* Translation Table Base Register 0 */
 160 #define c2_TTBR0_high   (c2_TTBR0 + 1)  /* TTBR0 top 32 bits */
 161 #define c2_TTBR1        (TTBR1_EL1 * 2) /* Translation Table Base Register 1 */
 162 #define c2_TTBR1_high   (c2_TTBR1 + 1)  /* TTBR1 top 32 bits */
 163 #define c2_TTBCR        (TCR_EL1 * 2)   /* Translation Table Base Control R. */
 164 #define c3_DACR         (DACR32_EL2 * 2)/* Domain Access Control Register */
 165 #define c5_DFSR         (ESR_EL1 * 2)   /* Data Fault Status Register */
 166 #define c5_IFSR         (IFSR32_EL2 * 2)/* Instruction Fault Status Register */
 167 #define c5_ADFSR        (AFSR0_EL1 * 2) /* Auxiliary Data Fault Status R */
 168 #define c5_AIFSR        (AFSR1_EL1 * 2) /* Auxiliary Instr Fault Status R */
 169 #define c6_DFAR         (FAR_EL1 * 2)   /* Data Fault Address Register */
 170 #define c6_IFAR         (c6_DFAR + 1)   /* Instruction Fault Address Register */
 171 #define c7_PAR          (PAR_EL1 * 2)   /* Physical Address Register */
 172 #define c7_PAR_high     (c7_PAR + 1)    /* PAR top 32 bits */
 173 #define c10_PRRR        (MAIR_EL1 * 2)  /* Primary Region Remap Register */
 174 #define c10_NMRR        (c10_PRRR + 1)  /* Normal Memory Remap Register */
 175 #define c12_VBAR        (VBAR_EL1 * 2)  /* Vector Base Address Register */
 176 #define c13_CID         (CONTEXTIDR_EL1 * 2)    /* Context ID Register */
 177 #define c13_TID_URW     (TPIDR_EL0 * 2) /* Thread ID, User R/W */
 178 #define c13_TID_URO     (TPIDRRO_EL0 * 2)/* Thread ID, User R/O */
 179 #define c13_TID_PRIV    (TPIDR_EL1 * 2) /* Thread ID, Privileged */
 180 #define c10_AMAIR0      (AMAIR_EL1 * 2) /* Aux Memory Attr Indirection Reg */
 181 #define c10_AMAIR1      (c10_AMAIR0 + 1)/* Aux Memory Attr Indirection Reg */
 182 #define c14_CNTKCTL     (CNTKCTL_EL1 * 2) /* Timer Control Register (PL1) */
 183
 184 #define cp14_DBGDSCRext (MDSCR_EL1 * 2)
 185 #define cp14_DBGBCR0    (DBGBCR0_EL1 * 2)
 186 #define cp14_DBGBVR0    (DBGBVR0_EL1 * 2)
 187 #define cp14_DBGBXVR0   (cp14_DBGBVR0 + 1)
 188 #define cp14_DBGWCR0    (DBGWCR0_EL1 * 2)
 189 #define cp14_DBGWVR0    (DBGWVR0_EL1 * 2)
 190 #define cp14_DBGDCCINT  (MDCCINT_EL1 * 2)
 191
 192 #define NR_COPRO_REGS   (NR_SYS_REGS * 2)
 193
 194 struct kvm_cpu_context {
 195         struct kvm_regs gp_regs;
 196         union {
 197                 u64 sys_regs[NR_SYS_REGS];
 198                 u32 copro[NR_COPRO_REGS];
 199         };
 200
 201         struct kvm_vcpu *__hyp_running_vcpu;
 202 };
 203
 204 typedef struct kvm_cpu_context kvm_cpu_context_t;
 205
 206 struct kvm_vcpu_arch {
 207         struct kvm_cpu_context ctxt;
 208
 209         /* HYP configuration */
 210         u64 hcr_el2;
 211         u32 mdcr_el2;
 212
 213         /* Exception Information */
 214         struct kvm_vcpu_fault_info fault;
 215
 216         /* Guest debug state */
 217         u64 debug_flags;
 218
 219         /*
 220          * We maintain more than a single set of debug registers to support
 221          * debugging the guest from the host and to maintain separate host and
 222          * guest state during world switches. vcpu_debug_state are the debug
 223          * registers of the vcpu as the guest sees them.  host_debug_state are
 224          * the host registers which are saved and restored during
 225          * world switches. external_debug_state contains the debug
 226          * values we want to debug the guest. This is set via the
 227          * KVM_SET_GUEST_DEBUG ioctl.
 228          *
 229          * debug_ptr points to the set of debug registers that should be loaded
 230          * onto the hardware when running the guest.
 231          */
 232         struct kvm_guest_debug_arch *debug_ptr;
 233         struct kvm_guest_debug_arch vcpu_debug_state;
 234         struct kvm_guest_debug_arch external_debug_state;
 235
 236         /* Pointer to host CPU context */
 237         kvm_cpu_context_t *host_cpu_context;
 238         struct {
 239                 /* {Break,watch}point registers */
 240                 struct kvm_guest_debug_arch regs;
 241                 /* Statistical profiling extension */
 242                 u64 pmscr_el1;
 243         } host_debug_state;
 244
 245         /* VGIC state */
 246         struct vgic_cpu vgic_cpu;
 247         struct arch_timer_cpu timer_cpu;
 248         struct kvm_pmu pmu;
 249
 250         /*
 251          * Anything that is not used directly from assembly code goes
 252          * here.
 253          */
 254
 255         /*
 256          * Guest registers we preserve during guest debugging.
 257          *
 258          * These shadow registers are updated by the kvm_handle_sys_reg
 259          * trap handler if the guest accesses or updates them while we
 260          * are using guest debug.
 261          */
 262         struct {
 263                 u32     mdscr_el1;
 264         } guest_debug_preserved;
 265
 266         /* vcpu power-off state */
 267         bool power_off;
 268
 269         /* Don't run the guest (internal implementation need) */
 270         bool pause;
 271
 272         /* IO related fields */
 273         struct kvm_decode mmio_decode;
 274
 275         /* Interrupt related fields */
 276         u64 irq_lines;          /* IRQ and FIQ levels */
 277
 278         /* Cache some mmu pages needed inside spinlock regions */
 279         struct kvm_mmu_memory_cache mmu_page_cache;
 280
 281         /* Target CPU and feature flags */
 282         int target;
 283         DECLARE_BITMAP(features, KVM_VCPU_MAX_FEATURES);
 284
 285         /* Detect first run of a vcpu */
 286         bool has_run_once;
 287
 288         /* Virtual SError ESR to restore when HCR_EL2.VSE is set */
 289         u64 vsesr_el2;
 290 };
 291
 292 #define vcpu_gp_regs(v)         (&(v)->arch.ctxt.gp_regs)
 293 #define vcpu_sys_reg(v,r)       ((v)->arch.ctxt.sys_regs[(r)])
 294 /*
 295  * CP14 and CP15 live in the same array, as they are backed by the
 296  * same system registers.
 297  */
 298 #define vcpu_cp14(v,r)          ((v)->arch.ctxt.copro[(r)])
 299 #define vcpu_cp15(v,r)          ((v)->arch.ctxt.copro[(r)])
 300
 301 #ifdef CONFIG_CPU_BIG_ENDIAN
 302 #define vcpu_cp15_64_high(v,r)  vcpu_cp15((v),(r))
 303 #define vcpu_cp15_64_low(v,r)   vcpu_cp15((v),(r) + 1)
 304 #else
 305 #define vcpu_cp15_64_high(v,r)  vcpu_cp15((v),(r) + 1)
 306 #define vcpu_cp15_64_low(v,r)   vcpu_cp15((v),(r))
 307 #endif
 308
 309 struct kvm_vm_stat {
 310         ulong remote_tlb_flush;
 311 };
 312
 313 struct kvm_vcpu_stat {
 314         u64 halt_successful_poll;
 315         u64 halt_attempted_poll;
 316         u64 halt_poll_invalid;
 317         u64 halt_wakeup;
 318         u64 hvc_exit_stat;
 319         u64 wfe_exit_stat;
 320         u64 wfi_exit_stat;
 321         u64 mmio_exit_user;
 322         u64 mmio_exit_kernel;
 323         u64 exits;
 324 };
 325
 326 int kvm_vcpu_preferred_target(struct kvm_vcpu_init *init);
 327 unsigned long kvm_arm_num_regs(struct kvm_vcpu *vcpu);
 328 int kvm_arm_copy_reg_indices(struct kvm_vcpu *vcpu, u64 __user *indices);
 329 int kvm_arm_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
 330 int kvm_arm_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
 331
 332 #define KVM_ARCH_WANT_MMU_NOTIFIER
 333 int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
 334 int kvm_unmap_hva_range(struct kvm *kvm,
 335                         unsigned long start, unsigned long end);
 336 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
 337 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
 338 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
 339
 340 struct kvm_vcpu *kvm_arm_get_running_vcpu(void);
 341 struct kvm_vcpu * __percpu *kvm_get_running_vcpus(void);
 342 void kvm_arm_halt_guest(struct kvm *kvm);
 343 void kvm_arm_resume_guest(struct kvm *kvm);
 344
 345 u64 __kvm_call_hyp(void *hypfn, ...);
 346 #define kvm_call_hyp(f, ...) __kvm_call_hyp(kvm_ksym_ref(f), ##__VA_ARGS__)
 347
 348 void force_vm_exit(const cpumask_t *mask);
 349 void kvm_mmu_wp_memory_region(struct kvm *kvm, int slot);
 350
 351 int handle_exit(struct kvm_vcpu *vcpu, struct kvm_run *run,
 352                 int exception_index);
 353 void handle_exit_early(struct kvm_vcpu *vcpu, struct kvm_run *run,
 354                        int exception_index);
 355
 356 int kvm_perf_init(void);
 357 int kvm_perf_teardown(void);
 358
 359 struct kvm_vcpu *kvm_mpidr_to_vcpu(struct kvm *kvm, unsigned long mpidr);
 360
 361 static inline void __cpu_init_hyp_mode(phys_addr_t pgd_ptr,
 362                                        unsigned long hyp_stack_ptr,
 363                                        unsigned long vector_ptr)
 364 {
 365         /*
 366          * Call initialization code, and switch to the full blown HYP code.
 367          * If the cpucaps haven't been finalized yet, something has gone very
 368          * wrong, and hyp will crash and burn when it uses any
 369          * cpus_have_const_cap() wrapper.
 370          */
 371         BUG_ON(!static_branch_likely(&arm64_const_caps_ready));
 372         __kvm_call_hyp((void *)pgd_ptr, hyp_stack_ptr, vector_ptr);
 373 }
 374
 375 static inline void kvm_arch_hardware_unsetup(void) {}
 376 static inline void kvm_arch_sync_events(struct kvm *kvm) {}
 377 static inline void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu) {}
 378 static inline void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu) {}
 379 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
 380
 381 void kvm_arm_init_debug(void);
 382 void kvm_arm_setup_debug(struct kvm_vcpu *vcpu);
 383 void kvm_arm_clear_debug(struct kvm_vcpu *vcpu);
 384 void kvm_arm_reset_debug_ptr(struct kvm_vcpu *vcpu);
 385 bool kvm_arm_handle_step_debug(struct kvm_vcpu *vcpu, struct kvm_run *run);
 386 int kvm_arm_vcpu_arch_set_attr(struct kvm_vcpu *vcpu,
 387                                struct kvm_device_attr *attr);
 388 int kvm_arm_vcpu_arch_get_attr(struct kvm_vcpu *vcpu,
 389                                struct kvm_device_attr *attr);
 390 int kvm_arm_vcpu_arch_has_attr(struct kvm_vcpu *vcpu,
 391                                struct kvm_device_attr *attr);
 392
 393 static inline void __cpu_init_stage2(void)
 394 {
 395         u32 parange = kvm_call_hyp(__init_stage2_translation);
 396
 397         WARN_ONCE(parange < 40,
 398                   "PARange is %d bits, unsupported configuration!", parange);
 399 }
 400
 401 /*
 402  * All host FP/SIMD state is restored on guest exit, so nothing needs
 403  * doing here except in the SVE case:
 404 */
 405 static inline void kvm_fpsimd_flush_cpu_state(void)
 406 {
 407         if (system_supports_sve())
 408                 sve_flush_cpu_state();
 409 }
 410
 411 static inline void kvm_arm_vhe_guest_enter(void)
 412 {
 413         local_daif_mask();
 414 }
 415
 416 static inline void kvm_arm_vhe_guest_exit(void)
 417 {
 418         local_daif_restore(DAIF_PROCCTX_NOIRQ);
 419 }
 420
 421 static inline bool kvm_arm_harden_branch_predictor(void)
 422 {
 423         return cpus_have_const_cap(ARM64_HARDEN_BRANCH_PREDICTOR);
 424 }
 425
 426 #endif /* __ARM64_KVM_HOST_H__ */