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/maple_tree.h> 20 #include <linux/percpu.h> 21 #include <linux/psci.h> 22 #include <asm/arch_gicv3.h> 23 #include <asm/barrier.h> 24 #include <asm/cpufeature.h> 25 #include <asm/cputype.h> 26 #include <asm/daifflags.h> 27 #include <asm/fpsimd.h> 28 #include <asm/kvm.h> 29 #include <asm/kvm_asm.h> 30 #include <asm/vncr_mapping.h> 31 32 #define __KVM_HAVE_ARCH_INTC_INITIALIZED 33 34 #define KVM_HALT_POLL_NS_DEFAULT 500000 35 36 #include <kvm/arm_vgic.h> 37 #include <kvm/arm_arch_timer.h> 38 #include <kvm/arm_pmu.h> 39 40 #define KVM_MAX_VCPUS VGIC_V3_MAX_CPUS 41 42 #define KVM_VCPU_MAX_FEATURES 9 43 #define KVM_VCPU_VALID_FEATURES (BIT(KVM_VCPU_MAX_FEATURES) - 1) 44 45 #define KVM_REQ_SLEEP \ 46 KVM_ARCH_REQ_FLAGS(0, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP) 47 #define KVM_REQ_IRQ_PENDING KVM_ARCH_REQ(1) 48 #define KVM_REQ_VCPU_RESET KVM_ARCH_REQ(2) 49 #define KVM_REQ_RECORD_STEAL KVM_ARCH_REQ(3) 50 #define KVM_REQ_RELOAD_GICv4 KVM_ARCH_REQ(4) 51 #define KVM_REQ_RELOAD_PMU KVM_ARCH_REQ(5) 52 #define KVM_REQ_SUSPEND KVM_ARCH_REQ(6) 53 #define KVM_REQ_RESYNC_PMU_EL0 KVM_ARCH_REQ(7) 54 #define KVM_REQ_NESTED_S2_UNMAP KVM_ARCH_REQ(8) 55 #define KVM_REQ_GUEST_HYP_IRQ_PENDING KVM_ARCH_REQ(9) 56 #define KVM_REQ_MAP_L1_VNCR_EL2 KVM_ARCH_REQ(10) 57 58 #define KVM_DIRTY_LOG_MANUAL_CAPS (KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE | \ 59 KVM_DIRTY_LOG_INITIALLY_SET) 60 61 #define KVM_HAVE_MMU_RWLOCK 62 63 /* 64 * Mode of operation configurable with kvm-arm.mode early param. 65 * See Documentation/admin-guide/kernel-parameters.txt for more information. 66 */ 67 enum kvm_mode { 68 KVM_MODE_DEFAULT, 69 KVM_MODE_PROTECTED, 70 KVM_MODE_NV, 71 KVM_MODE_NONE, 72 }; 73 #ifdef CONFIG_KVM 74 enum kvm_mode kvm_get_mode(void); 75 #else 76 static inline enum kvm_mode kvm_get_mode(void) { return KVM_MODE_NONE; }; 77 #endif 78 79 extern unsigned int __ro_after_init kvm_sve_max_vl; 80 extern unsigned int __ro_after_init kvm_host_sve_max_vl; 81 int __init kvm_arm_init_sve(void); 82 83 u32 __attribute_const__ kvm_target_cpu(void); 84 void kvm_reset_vcpu(struct kvm_vcpu *vcpu); 85 void kvm_arm_vcpu_destroy(struct kvm_vcpu *vcpu); 86 87 struct kvm_hyp_memcache { 88 phys_addr_t head; 89 unsigned long nr_pages; 90 struct pkvm_mapping *mapping; /* only used from EL1 */ 91 92 #define HYP_MEMCACHE_ACCOUNT_STAGE2 BIT(1) 93 unsigned long flags; 94 }; 95 96 static inline void push_hyp_memcache(struct kvm_hyp_memcache *mc, 97 phys_addr_t *p, 98 phys_addr_t (*to_pa)(void *virt)) 99 { 100 *p = mc->head; 101 mc->head = to_pa(p); 102 mc->nr_pages++; 103 } 104 105 static inline void *pop_hyp_memcache(struct kvm_hyp_memcache *mc, 106 void *(*to_va)(phys_addr_t phys)) 107 { 108 phys_addr_t *p = to_va(mc->head & PAGE_MASK); 109 110 if (!mc->nr_pages) 111 return NULL; 112 113 mc->head = *p; 114 mc->nr_pages--; 115 116 return p; 117 } 118 119 static inline int __topup_hyp_memcache(struct kvm_hyp_memcache *mc, 120 unsigned long min_pages, 121 void *(*alloc_fn)(void *arg), 122 phys_addr_t (*to_pa)(void *virt), 123 void *arg) 124 { 125 while (mc->nr_pages < min_pages) { 126 phys_addr_t *p = alloc_fn(arg); 127 128 if (!p) 129 return -ENOMEM; 130 push_hyp_memcache(mc, p, to_pa); 131 } 132 133 return 0; 134 } 135 136 static inline void __free_hyp_memcache(struct kvm_hyp_memcache *mc, 137 void (*free_fn)(void *virt, void *arg), 138 void *(*to_va)(phys_addr_t phys), 139 void *arg) 140 { 141 while (mc->nr_pages) 142 free_fn(pop_hyp_memcache(mc, to_va), arg); 143 } 144 145 void free_hyp_memcache(struct kvm_hyp_memcache *mc); 146 int topup_hyp_memcache(struct kvm_hyp_memcache *mc, unsigned long min_pages); 147 148 struct kvm_vmid { 149 atomic64_t id; 150 }; 151 152 struct kvm_s2_mmu { 153 struct kvm_vmid vmid; 154 155 /* 156 * stage2 entry level table 157 * 158 * Two kvm_s2_mmu structures in the same VM can point to the same 159 * pgd here. This happens when running a guest using a 160 * translation regime that isn't affected by its own stage-2 161 * translation, such as a non-VHE hypervisor running at vEL2, or 162 * for vEL1/EL0 with vHCR_EL2.VM == 0. In that case, we use the 163 * canonical stage-2 page tables. 164 */ 165 phys_addr_t pgd_phys; 166 struct kvm_pgtable *pgt; 167 168 /* 169 * VTCR value used on the host. For a non-NV guest (or a NV 170 * guest that runs in a context where its own S2 doesn't 171 * apply), its T0SZ value reflects that of the IPA size. 172 * 173 * For a shadow S2 MMU, T0SZ reflects the PARange exposed to 174 * the guest. 175 */ 176 u64 vtcr; 177 178 /* The last vcpu id that ran on each physical CPU */ 179 int __percpu *last_vcpu_ran; 180 181 #define KVM_ARM_EAGER_SPLIT_CHUNK_SIZE_DEFAULT 0 182 /* 183 * Memory cache used to split 184 * KVM_CAP_ARM_EAGER_SPLIT_CHUNK_SIZE worth of huge pages. It 185 * is used to allocate stage2 page tables while splitting huge 186 * pages. The choice of KVM_CAP_ARM_EAGER_SPLIT_CHUNK_SIZE 187 * influences both the capacity of the split page cache, and 188 * how often KVM reschedules. Be wary of raising CHUNK_SIZE 189 * too high. 190 * 191 * Protected by kvm->slots_lock. 192 */ 193 struct kvm_mmu_memory_cache split_page_cache; 194 uint64_t split_page_chunk_size; 195 196 struct kvm_arch *arch; 197 198 /* 199 * For a shadow stage-2 MMU, the virtual vttbr used by the 200 * host to parse the guest S2. 201 * This either contains: 202 * - the virtual VTTBR programmed by the guest hypervisor with 203 * CnP cleared 204 * - The value 1 (VMID=0, BADDR=0, CnP=1) if invalid 205 * 206 * We also cache the full VTCR which gets used for TLB invalidation, 207 * taking the ARM ARM's "Any of the bits in VTCR_EL2 are permitted 208 * to be cached in a TLB" to the letter. 209 */ 210 u64 tlb_vttbr; 211 u64 tlb_vtcr; 212 213 /* 214 * true when this represents a nested context where virtual 215 * HCR_EL2.VM == 1 216 */ 217 bool nested_stage2_enabled; 218 219 /* 220 * true when this MMU needs to be unmapped before being used for a new 221 * purpose. 222 */ 223 bool pending_unmap; 224 225 /* 226 * 0: Nobody is currently using this, check vttbr for validity 227 * >0: Somebody is actively using this. 228 */ 229 atomic_t refcnt; 230 }; 231 232 struct kvm_arch_memory_slot { 233 }; 234 235 /** 236 * struct kvm_smccc_features: Descriptor of the hypercall services exposed to the guests 237 * 238 * @std_bmap: Bitmap of standard secure service calls 239 * @std_hyp_bmap: Bitmap of standard hypervisor service calls 240 * @vendor_hyp_bmap: Bitmap of vendor specific hypervisor service calls 241 */ 242 struct kvm_smccc_features { 243 unsigned long std_bmap; 244 unsigned long std_hyp_bmap; 245 unsigned long vendor_hyp_bmap; /* Function numbers 0-63 */ 246 unsigned long vendor_hyp_bmap_2; /* Function numbers 64-127 */ 247 }; 248 249 typedef unsigned int pkvm_handle_t; 250 251 struct kvm_protected_vm { 252 pkvm_handle_t handle; 253 struct kvm_hyp_memcache teardown_mc; 254 struct kvm_hyp_memcache stage2_teardown_mc; 255 bool enabled; 256 }; 257 258 struct kvm_mpidr_data { 259 u64 mpidr_mask; 260 DECLARE_FLEX_ARRAY(u16, cmpidr_to_idx); 261 }; 262 263 static inline u16 kvm_mpidr_index(struct kvm_mpidr_data *data, u64 mpidr) 264 { 265 unsigned long index = 0, mask = data->mpidr_mask; 266 unsigned long aff = mpidr & MPIDR_HWID_BITMASK; 267 268 bitmap_gather(&index, &aff, &mask, fls(mask)); 269 270 return index; 271 } 272 273 struct kvm_sysreg_masks; 274 275 enum fgt_group_id { 276 __NO_FGT_GROUP__, 277 HFGRTR_GROUP, 278 HFGWTR_GROUP = HFGRTR_GROUP, 279 HDFGRTR_GROUP, 280 HDFGWTR_GROUP = HDFGRTR_GROUP, 281 HFGITR_GROUP, 282 HAFGRTR_GROUP, 283 HFGRTR2_GROUP, 284 HFGWTR2_GROUP = HFGRTR2_GROUP, 285 HDFGRTR2_GROUP, 286 HDFGWTR2_GROUP = HDFGRTR2_GROUP, 287 HFGITR2_GROUP, 288 289 /* Must be last */ 290 __NR_FGT_GROUP_IDS__ 291 }; 292 293 struct kvm_arch { 294 struct kvm_s2_mmu mmu; 295 296 /* 297 * Fine-Grained UNDEF, mimicking the FGT layout defined by the 298 * architecture. We track them globally, as we present the 299 * same feature-set to all vcpus. 300 * 301 * Index 0 is currently spare. 302 */ 303 u64 fgu[__NR_FGT_GROUP_IDS__]; 304 305 /* 306 * Stage 2 paging state for VMs with nested S2 using a virtual 307 * VMID. 308 */ 309 struct kvm_s2_mmu *nested_mmus; 310 size_t nested_mmus_size; 311 int nested_mmus_next; 312 313 /* Interrupt controller */ 314 struct vgic_dist vgic; 315 316 /* Timers */ 317 struct arch_timer_vm_data timer_data; 318 319 /* Mandated version of PSCI */ 320 u32 psci_version; 321 322 /* Protects VM-scoped configuration data */ 323 struct mutex config_lock; 324 325 /* 326 * If we encounter a data abort without valid instruction syndrome 327 * information, report this to user space. User space can (and 328 * should) opt in to this feature if KVM_CAP_ARM_NISV_TO_USER is 329 * supported. 330 */ 331 #define KVM_ARCH_FLAG_RETURN_NISV_IO_ABORT_TO_USER 0 332 /* Memory Tagging Extension enabled for the guest */ 333 #define KVM_ARCH_FLAG_MTE_ENABLED 1 334 /* At least one vCPU has ran in the VM */ 335 #define KVM_ARCH_FLAG_HAS_RAN_ONCE 2 336 /* The vCPU feature set for the VM is configured */ 337 #define KVM_ARCH_FLAG_VCPU_FEATURES_CONFIGURED 3 338 /* PSCI SYSTEM_SUSPEND enabled for the guest */ 339 #define KVM_ARCH_FLAG_SYSTEM_SUSPEND_ENABLED 4 340 /* VM counter offset */ 341 #define KVM_ARCH_FLAG_VM_COUNTER_OFFSET 5 342 /* Timer PPIs made immutable */ 343 #define KVM_ARCH_FLAG_TIMER_PPIS_IMMUTABLE 6 344 /* Initial ID reg values loaded */ 345 #define KVM_ARCH_FLAG_ID_REGS_INITIALIZED 7 346 /* Fine-Grained UNDEF initialised */ 347 #define KVM_ARCH_FLAG_FGU_INITIALIZED 8 348 /* SVE exposed to guest */ 349 #define KVM_ARCH_FLAG_GUEST_HAS_SVE 9 350 /* MIDR_EL1, REVIDR_EL1, and AIDR_EL1 are writable from userspace */ 351 #define KVM_ARCH_FLAG_WRITABLE_IMP_ID_REGS 10 352 unsigned long flags; 353 354 /* VM-wide vCPU feature set */ 355 DECLARE_BITMAP(vcpu_features, KVM_VCPU_MAX_FEATURES); 356 357 /* MPIDR to vcpu index mapping, optional */ 358 struct kvm_mpidr_data *mpidr_data; 359 360 /* 361 * VM-wide PMU filter, implemented as a bitmap and big enough for 362 * up to 2^10 events (ARMv8.0) or 2^16 events (ARMv8.1+). 363 */ 364 unsigned long *pmu_filter; 365 struct arm_pmu *arm_pmu; 366 367 cpumask_var_t supported_cpus; 368 369 /* Maximum number of counters for the guest */ 370 u8 nr_pmu_counters; 371 372 /* Iterator for idreg debugfs */ 373 u8 idreg_debugfs_iter; 374 375 /* Hypercall features firmware registers' descriptor */ 376 struct kvm_smccc_features smccc_feat; 377 struct maple_tree smccc_filter; 378 379 /* 380 * Emulated CPU ID registers per VM 381 * (Op0, Op1, CRn, CRm, Op2) of the ID registers to be saved in it 382 * is (3, 0, 0, crm, op2), where 1<=crm<8, 0<=op2<8. 383 * 384 * These emulated idregs are VM-wide, but accessed from the context of a vCPU. 385 * Atomic access to multiple idregs are guarded by kvm_arch.config_lock. 386 */ 387 #define IDREG_IDX(id) (((sys_reg_CRm(id) - 1) << 3) | sys_reg_Op2(id)) 388 #define KVM_ARM_ID_REG_NUM (IDREG_IDX(sys_reg(3, 0, 0, 7, 7)) + 1) 389 u64 id_regs[KVM_ARM_ID_REG_NUM]; 390 391 u64 midr_el1; 392 u64 revidr_el1; 393 u64 aidr_el1; 394 u64 ctr_el0; 395 396 /* Masks for VNCR-backed and general EL2 sysregs */ 397 struct kvm_sysreg_masks *sysreg_masks; 398 399 /* Count the number of VNCR_EL2 currently mapped */ 400 atomic_t vncr_map_count; 401 402 /* 403 * For an untrusted host VM, 'pkvm.handle' is used to lookup 404 * the associated pKVM instance in the hypervisor. 405 */ 406 struct kvm_protected_vm pkvm; 407 }; 408 409 struct kvm_vcpu_fault_info { 410 u64 esr_el2; /* Hyp Syndrom Register */ 411 u64 far_el2; /* Hyp Fault Address Register */ 412 u64 hpfar_el2; /* Hyp IPA Fault Address Register */ 413 u64 disr_el1; /* Deferred [SError] Status Register */ 414 }; 415 416 /* 417 * VNCR() just places the VNCR_capable registers in the enum after 418 * __VNCR_START__, and the value (after correction) to be an 8-byte offset 419 * from the VNCR base. As we don't require the enum to be otherwise ordered, 420 * we need the terrible hack below to ensure that we correctly size the 421 * sys_regs array, no matter what. 422 * 423 * The __MAX__ macro has been lifted from Sean Eron Anderson's wonderful 424 * treasure trove of bit hacks: 425 * https://graphics.stanford.edu/~seander/bithacks.html#IntegerMinOrMax 426 */ 427 #define __MAX__(x,y) ((x) ^ (((x) ^ (y)) & -((x) < (y)))) 428 #define VNCR(r) \ 429 __before_##r, \ 430 r = __VNCR_START__ + ((VNCR_ ## r) / 8), \ 431 __after_##r = __MAX__(__before_##r - 1, r) 432 433 #define MARKER(m) \ 434 m, __after_##m = m - 1 435 436 enum vcpu_sysreg { 437 __INVALID_SYSREG__, /* 0 is reserved as an invalid value */ 438 MPIDR_EL1, /* MultiProcessor Affinity Register */ 439 CLIDR_EL1, /* Cache Level ID Register */ 440 CSSELR_EL1, /* Cache Size Selection Register */ 441 TPIDR_EL0, /* Thread ID, User R/W */ 442 TPIDRRO_EL0, /* Thread ID, User R/O */ 443 TPIDR_EL1, /* Thread ID, Privileged */ 444 CNTKCTL_EL1, /* Timer Control Register (EL1) */ 445 PAR_EL1, /* Physical Address Register */ 446 MDCCINT_EL1, /* Monitor Debug Comms Channel Interrupt Enable Reg */ 447 OSLSR_EL1, /* OS Lock Status Register */ 448 DISR_EL1, /* Deferred Interrupt Status Register */ 449 450 /* Performance Monitors Registers */ 451 PMCR_EL0, /* Control Register */ 452 PMSELR_EL0, /* Event Counter Selection Register */ 453 PMEVCNTR0_EL0, /* Event Counter Register (0-30) */ 454 PMEVCNTR30_EL0 = PMEVCNTR0_EL0 + 30, 455 PMCCNTR_EL0, /* Cycle Counter Register */ 456 PMEVTYPER0_EL0, /* Event Type Register (0-30) */ 457 PMEVTYPER30_EL0 = PMEVTYPER0_EL0 + 30, 458 PMCCFILTR_EL0, /* Cycle Count Filter Register */ 459 PMCNTENSET_EL0, /* Count Enable Set Register */ 460 PMINTENSET_EL1, /* Interrupt Enable Set Register */ 461 PMOVSSET_EL0, /* Overflow Flag Status Set Register */ 462 PMUSERENR_EL0, /* User Enable Register */ 463 464 /* Pointer Authentication Registers in a strict increasing order. */ 465 APIAKEYLO_EL1, 466 APIAKEYHI_EL1, 467 APIBKEYLO_EL1, 468 APIBKEYHI_EL1, 469 APDAKEYLO_EL1, 470 APDAKEYHI_EL1, 471 APDBKEYLO_EL1, 472 APDBKEYHI_EL1, 473 APGAKEYLO_EL1, 474 APGAKEYHI_EL1, 475 476 /* Memory Tagging Extension registers */ 477 RGSR_EL1, /* Random Allocation Tag Seed Register */ 478 GCR_EL1, /* Tag Control Register */ 479 TFSRE0_EL1, /* Tag Fault Status Register (EL0) */ 480 481 POR_EL0, /* Permission Overlay Register 0 (EL0) */ 482 483 /* FP/SIMD/SVE */ 484 SVCR, 485 FPMR, 486 487 /* 32bit specific registers. */ 488 DACR32_EL2, /* Domain Access Control Register */ 489 IFSR32_EL2, /* Instruction Fault Status Register */ 490 FPEXC32_EL2, /* Floating-Point Exception Control Register */ 491 DBGVCR32_EL2, /* Debug Vector Catch Register */ 492 493 /* EL2 registers */ 494 SCTLR_EL2, /* System Control Register (EL2) */ 495 ACTLR_EL2, /* Auxiliary Control Register (EL2) */ 496 CPTR_EL2, /* Architectural Feature Trap Register (EL2) */ 497 HACR_EL2, /* Hypervisor Auxiliary Control Register */ 498 ZCR_EL2, /* SVE Control Register (EL2) */ 499 TTBR0_EL2, /* Translation Table Base Register 0 (EL2) */ 500 TTBR1_EL2, /* Translation Table Base Register 1 (EL2) */ 501 TCR_EL2, /* Translation Control Register (EL2) */ 502 PIRE0_EL2, /* Permission Indirection Register 0 (EL2) */ 503 PIR_EL2, /* Permission Indirection Register 1 (EL2) */ 504 POR_EL2, /* Permission Overlay Register 2 (EL2) */ 505 SPSR_EL2, /* EL2 saved program status register */ 506 ELR_EL2, /* EL2 exception link register */ 507 AFSR0_EL2, /* Auxiliary Fault Status Register 0 (EL2) */ 508 AFSR1_EL2, /* Auxiliary Fault Status Register 1 (EL2) */ 509 ESR_EL2, /* Exception Syndrome Register (EL2) */ 510 FAR_EL2, /* Fault Address Register (EL2) */ 511 HPFAR_EL2, /* Hypervisor IPA Fault Address Register */ 512 MAIR_EL2, /* Memory Attribute Indirection Register (EL2) */ 513 AMAIR_EL2, /* Auxiliary Memory Attribute Indirection Register (EL2) */ 514 VBAR_EL2, /* Vector Base Address Register (EL2) */ 515 RVBAR_EL2, /* Reset Vector Base Address Register */ 516 CONTEXTIDR_EL2, /* Context ID Register (EL2) */ 517 SP_EL2, /* EL2 Stack Pointer */ 518 CNTHP_CTL_EL2, 519 CNTHP_CVAL_EL2, 520 CNTHV_CTL_EL2, 521 CNTHV_CVAL_EL2, 522 523 /* Anything from this can be RES0/RES1 sanitised */ 524 MARKER(__SANITISED_REG_START__), 525 TCR2_EL2, /* Extended Translation Control Register (EL2) */ 526 MDCR_EL2, /* Monitor Debug Configuration Register (EL2) */ 527 CNTHCTL_EL2, /* Counter-timer Hypervisor Control register */ 528 529 /* Any VNCR-capable reg goes after this point */ 530 MARKER(__VNCR_START__), 531 532 VNCR(SCTLR_EL1),/* System Control Register */ 533 VNCR(ACTLR_EL1),/* Auxiliary Control Register */ 534 VNCR(CPACR_EL1),/* Coprocessor Access Control */ 535 VNCR(ZCR_EL1), /* SVE Control */ 536 VNCR(TTBR0_EL1),/* Translation Table Base Register 0 */ 537 VNCR(TTBR1_EL1),/* Translation Table Base Register 1 */ 538 VNCR(TCR_EL1), /* Translation Control Register */ 539 VNCR(TCR2_EL1), /* Extended Translation Control Register */ 540 VNCR(ESR_EL1), /* Exception Syndrome Register */ 541 VNCR(AFSR0_EL1),/* Auxiliary Fault Status Register 0 */ 542 VNCR(AFSR1_EL1),/* Auxiliary Fault Status Register 1 */ 543 VNCR(FAR_EL1), /* Fault Address Register */ 544 VNCR(MAIR_EL1), /* Memory Attribute Indirection Register */ 545 VNCR(VBAR_EL1), /* Vector Base Address Register */ 546 VNCR(CONTEXTIDR_EL1), /* Context ID Register */ 547 VNCR(AMAIR_EL1),/* Aux Memory Attribute Indirection Register */ 548 VNCR(MDSCR_EL1),/* Monitor Debug System Control Register */ 549 VNCR(ELR_EL1), 550 VNCR(SP_EL1), 551 VNCR(SPSR_EL1), 552 VNCR(TFSR_EL1), /* Tag Fault Status Register (EL1) */ 553 VNCR(VPIDR_EL2),/* Virtualization Processor ID Register */ 554 VNCR(VMPIDR_EL2),/* Virtualization Multiprocessor ID Register */ 555 VNCR(HCR_EL2), /* Hypervisor Configuration Register */ 556 VNCR(HSTR_EL2), /* Hypervisor System Trap Register */ 557 VNCR(VTTBR_EL2),/* Virtualization Translation Table Base Register */ 558 VNCR(VTCR_EL2), /* Virtualization Translation Control Register */ 559 VNCR(TPIDR_EL2),/* EL2 Software Thread ID Register */ 560 VNCR(HCRX_EL2), /* Extended Hypervisor Configuration Register */ 561 562 /* Permission Indirection Extension registers */ 563 VNCR(PIR_EL1), /* Permission Indirection Register 1 (EL1) */ 564 VNCR(PIRE0_EL1), /* Permission Indirection Register 0 (EL1) */ 565 566 VNCR(POR_EL1), /* Permission Overlay Register 1 (EL1) */ 567 568 VNCR(HFGRTR_EL2), 569 VNCR(HFGWTR_EL2), 570 VNCR(HFGITR_EL2), 571 VNCR(HDFGRTR_EL2), 572 VNCR(HDFGWTR_EL2), 573 VNCR(HAFGRTR_EL2), 574 VNCR(HFGRTR2_EL2), 575 VNCR(HFGWTR2_EL2), 576 VNCR(HFGITR2_EL2), 577 VNCR(HDFGRTR2_EL2), 578 VNCR(HDFGWTR2_EL2), 579 580 VNCR(VNCR_EL2), 581 582 VNCR(CNTVOFF_EL2), 583 VNCR(CNTV_CVAL_EL0), 584 VNCR(CNTV_CTL_EL0), 585 VNCR(CNTP_CVAL_EL0), 586 VNCR(CNTP_CTL_EL0), 587 588 VNCR(ICH_LR0_EL2), 589 VNCR(ICH_LR1_EL2), 590 VNCR(ICH_LR2_EL2), 591 VNCR(ICH_LR3_EL2), 592 VNCR(ICH_LR4_EL2), 593 VNCR(ICH_LR5_EL2), 594 VNCR(ICH_LR6_EL2), 595 VNCR(ICH_LR7_EL2), 596 VNCR(ICH_LR8_EL2), 597 VNCR(ICH_LR9_EL2), 598 VNCR(ICH_LR10_EL2), 599 VNCR(ICH_LR11_EL2), 600 VNCR(ICH_LR12_EL2), 601 VNCR(ICH_LR13_EL2), 602 VNCR(ICH_LR14_EL2), 603 VNCR(ICH_LR15_EL2), 604 605 VNCR(ICH_AP0R0_EL2), 606 VNCR(ICH_AP0R1_EL2), 607 VNCR(ICH_AP0R2_EL2), 608 VNCR(ICH_AP0R3_EL2), 609 VNCR(ICH_AP1R0_EL2), 610 VNCR(ICH_AP1R1_EL2), 611 VNCR(ICH_AP1R2_EL2), 612 VNCR(ICH_AP1R3_EL2), 613 VNCR(ICH_HCR_EL2), 614 VNCR(ICH_VMCR_EL2), 615 616 NR_SYS_REGS /* Nothing after this line! */ 617 }; 618 619 struct kvm_sysreg_masks { 620 struct { 621 u64 res0; 622 u64 res1; 623 } mask[NR_SYS_REGS - __SANITISED_REG_START__]; 624 }; 625 626 struct fgt_masks { 627 const char *str; 628 u64 mask; 629 u64 nmask; 630 u64 res0; 631 }; 632 633 extern struct fgt_masks hfgrtr_masks; 634 extern struct fgt_masks hfgwtr_masks; 635 extern struct fgt_masks hfgitr_masks; 636 extern struct fgt_masks hdfgrtr_masks; 637 extern struct fgt_masks hdfgwtr_masks; 638 extern struct fgt_masks hafgrtr_masks; 639 extern struct fgt_masks hfgrtr2_masks; 640 extern struct fgt_masks hfgwtr2_masks; 641 extern struct fgt_masks hfgitr2_masks; 642 extern struct fgt_masks hdfgrtr2_masks; 643 extern struct fgt_masks hdfgwtr2_masks; 644 645 extern struct fgt_masks kvm_nvhe_sym(hfgrtr_masks); 646 extern struct fgt_masks kvm_nvhe_sym(hfgwtr_masks); 647 extern struct fgt_masks kvm_nvhe_sym(hfgitr_masks); 648 extern struct fgt_masks kvm_nvhe_sym(hdfgrtr_masks); 649 extern struct fgt_masks kvm_nvhe_sym(hdfgwtr_masks); 650 extern struct fgt_masks kvm_nvhe_sym(hafgrtr_masks); 651 extern struct fgt_masks kvm_nvhe_sym(hfgrtr2_masks); 652 extern struct fgt_masks kvm_nvhe_sym(hfgwtr2_masks); 653 extern struct fgt_masks kvm_nvhe_sym(hfgitr2_masks); 654 extern struct fgt_masks kvm_nvhe_sym(hdfgrtr2_masks); 655 extern struct fgt_masks kvm_nvhe_sym(hdfgwtr2_masks); 656 657 struct kvm_cpu_context { 658 struct user_pt_regs regs; /* sp = sp_el0 */ 659 660 u64 spsr_abt; 661 u64 spsr_und; 662 u64 spsr_irq; 663 u64 spsr_fiq; 664 665 struct user_fpsimd_state fp_regs; 666 667 u64 sys_regs[NR_SYS_REGS]; 668 669 struct kvm_vcpu *__hyp_running_vcpu; 670 671 /* This pointer has to be 4kB aligned. */ 672 u64 *vncr_array; 673 }; 674 675 struct cpu_sve_state { 676 __u64 zcr_el1; 677 678 /* 679 * Ordering is important since __sve_save_state/__sve_restore_state 680 * relies on it. 681 */ 682 __u32 fpsr; 683 __u32 fpcr; 684 685 /* Must be SVE_VQ_BYTES (128 bit) aligned. */ 686 __u8 sve_regs[]; 687 }; 688 689 /* 690 * This structure is instantiated on a per-CPU basis, and contains 691 * data that is: 692 * 693 * - tied to a single physical CPU, and 694 * - either have a lifetime that does not extend past vcpu_put() 695 * - or is an invariant for the lifetime of the system 696 * 697 * Use host_data_ptr(field) as a way to access a pointer to such a 698 * field. 699 */ 700 struct kvm_host_data { 701 #define KVM_HOST_DATA_FLAG_HAS_SPE 0 702 #define KVM_HOST_DATA_FLAG_HAS_TRBE 1 703 #define KVM_HOST_DATA_FLAG_TRBE_ENABLED 4 704 #define KVM_HOST_DATA_FLAG_EL1_TRACING_CONFIGURED 5 705 #define KVM_HOST_DATA_FLAG_VCPU_IN_HYP_CONTEXT 6 706 #define KVM_HOST_DATA_FLAG_L1_VNCR_MAPPED 7 707 #define KVM_HOST_DATA_FLAG_HAS_BRBE 8 708 unsigned long flags; 709 710 struct kvm_cpu_context host_ctxt; 711 712 /* 713 * Hyp VA. 714 * sve_state is only used in pKVM and if system_supports_sve(). 715 */ 716 struct cpu_sve_state *sve_state; 717 718 /* Used by pKVM only. */ 719 u64 fpmr; 720 721 /* Ownership of the FP regs */ 722 enum { 723 FP_STATE_FREE, 724 FP_STATE_HOST_OWNED, 725 FP_STATE_GUEST_OWNED, 726 } fp_owner; 727 728 /* 729 * host_debug_state contains the host registers which are 730 * saved and restored during world switches. 731 */ 732 struct { 733 /* {Break,watch}point registers */ 734 struct kvm_guest_debug_arch regs; 735 /* Statistical profiling extension */ 736 u64 pmscr_el1; 737 /* Self-hosted trace */ 738 u64 trfcr_el1; 739 /* Values of trap registers for the host before guest entry. */ 740 u64 mdcr_el2; 741 u64 brbcr_el1; 742 } host_debug_state; 743 744 /* Guest trace filter value */ 745 u64 trfcr_while_in_guest; 746 747 /* Number of programmable event counters (PMCR_EL0.N) for this CPU */ 748 unsigned int nr_event_counters; 749 750 /* Number of debug breakpoints/watchpoints for this CPU (minus 1) */ 751 unsigned int debug_brps; 752 unsigned int debug_wrps; 753 }; 754 755 struct kvm_host_psci_config { 756 /* PSCI version used by host. */ 757 u32 version; 758 u32 smccc_version; 759 760 /* Function IDs used by host if version is v0.1. */ 761 struct psci_0_1_function_ids function_ids_0_1; 762 763 bool psci_0_1_cpu_suspend_implemented; 764 bool psci_0_1_cpu_on_implemented; 765 bool psci_0_1_cpu_off_implemented; 766 bool psci_0_1_migrate_implemented; 767 }; 768 769 extern struct kvm_host_psci_config kvm_nvhe_sym(kvm_host_psci_config); 770 #define kvm_host_psci_config CHOOSE_NVHE_SYM(kvm_host_psci_config) 771 772 extern s64 kvm_nvhe_sym(hyp_physvirt_offset); 773 #define hyp_physvirt_offset CHOOSE_NVHE_SYM(hyp_physvirt_offset) 774 775 extern u64 kvm_nvhe_sym(hyp_cpu_logical_map)[NR_CPUS]; 776 #define hyp_cpu_logical_map CHOOSE_NVHE_SYM(hyp_cpu_logical_map) 777 778 struct vcpu_reset_state { 779 unsigned long pc; 780 unsigned long r0; 781 bool be; 782 bool reset; 783 }; 784 785 struct vncr_tlb; 786 787 struct kvm_vcpu_arch { 788 struct kvm_cpu_context ctxt; 789 790 /* 791 * Guest floating point state 792 * 793 * The architecture has two main floating point extensions, 794 * the original FPSIMD and SVE. These have overlapping 795 * register views, with the FPSIMD V registers occupying the 796 * low 128 bits of the SVE Z registers. When the core 797 * floating point code saves the register state of a task it 798 * records which view it saved in fp_type. 799 */ 800 void *sve_state; 801 enum fp_type fp_type; 802 unsigned int sve_max_vl; 803 804 /* Stage 2 paging state used by the hardware on next switch */ 805 struct kvm_s2_mmu *hw_mmu; 806 807 /* Values of trap registers for the guest. */ 808 u64 hcr_el2; 809 u64 hcrx_el2; 810 u64 mdcr_el2; 811 812 /* Exception Information */ 813 struct kvm_vcpu_fault_info fault; 814 815 /* Configuration flags, set once and for all before the vcpu can run */ 816 u8 cflags; 817 818 /* Input flags to the hypervisor code, potentially cleared after use */ 819 u8 iflags; 820 821 /* State flags for kernel bookkeeping, unused by the hypervisor code */ 822 u8 sflags; 823 824 /* 825 * Don't run the guest (internal implementation need). 826 * 827 * Contrary to the flags above, this is set/cleared outside of 828 * a vcpu context, and thus cannot be mixed with the flags 829 * themselves (or the flag accesses need to be made atomic). 830 */ 831 bool pause; 832 833 /* 834 * We maintain more than a single set of debug registers to support 835 * debugging the guest from the host and to maintain separate host and 836 * guest state during world switches. vcpu_debug_state are the debug 837 * registers of the vcpu as the guest sees them. 838 * 839 * external_debug_state contains the debug values we want to debug the 840 * guest. This is set via the KVM_SET_GUEST_DEBUG ioctl. 841 */ 842 struct kvm_guest_debug_arch vcpu_debug_state; 843 struct kvm_guest_debug_arch external_debug_state; 844 u64 external_mdscr_el1; 845 846 enum { 847 VCPU_DEBUG_FREE, 848 VCPU_DEBUG_HOST_OWNED, 849 VCPU_DEBUG_GUEST_OWNED, 850 } debug_owner; 851 852 /* VGIC state */ 853 struct vgic_cpu vgic_cpu; 854 struct arch_timer_cpu timer_cpu; 855 struct kvm_pmu pmu; 856 857 /* vcpu power state */ 858 struct kvm_mp_state mp_state; 859 spinlock_t mp_state_lock; 860 861 /* Cache some mmu pages needed inside spinlock regions */ 862 struct kvm_mmu_memory_cache mmu_page_cache; 863 864 /* Pages to top-up the pKVM/EL2 guest pool */ 865 struct kvm_hyp_memcache pkvm_memcache; 866 867 /* Virtual SError ESR to restore when HCR_EL2.VSE is set */ 868 u64 vsesr_el2; 869 870 /* Additional reset state */ 871 struct vcpu_reset_state reset_state; 872 873 /* Guest PV state */ 874 struct { 875 u64 last_steal; 876 gpa_t base; 877 } steal; 878 879 /* Per-vcpu CCSIDR override or NULL */ 880 u32 *ccsidr; 881 882 /* Per-vcpu TLB for VNCR_EL2 -- NULL when !NV */ 883 struct vncr_tlb *vncr_tlb; 884 }; 885 886 /* 887 * Each 'flag' is composed of a comma-separated triplet: 888 * 889 * - the flag-set it belongs to in the vcpu->arch structure 890 * - the value for that flag 891 * - the mask for that flag 892 * 893 * __vcpu_single_flag() builds such a triplet for a single-bit flag. 894 * unpack_vcpu_flag() extract the flag value from the triplet for 895 * direct use outside of the flag accessors. 896 */ 897 #define __vcpu_single_flag(_set, _f) _set, (_f), (_f) 898 899 #define __unpack_flag(_set, _f, _m) _f 900 #define unpack_vcpu_flag(...) __unpack_flag(__VA_ARGS__) 901 902 #define __build_check_flag(v, flagset, f, m) \ 903 do { \ 904 typeof(v->arch.flagset) *_fset; \ 905 \ 906 /* Check that the flags fit in the mask */ \ 907 BUILD_BUG_ON(HWEIGHT(m) != HWEIGHT((f) | (m))); \ 908 /* Check that the flags fit in the type */ \ 909 BUILD_BUG_ON((sizeof(*_fset) * 8) <= __fls(m)); \ 910 } while (0) 911 912 #define __vcpu_get_flag(v, flagset, f, m) \ 913 ({ \ 914 __build_check_flag(v, flagset, f, m); \ 915 \ 916 READ_ONCE(v->arch.flagset) & (m); \ 917 }) 918 919 /* 920 * Note that the set/clear accessors must be preempt-safe in order to 921 * avoid nesting them with load/put which also manipulate flags... 922 */ 923 #ifdef __KVM_NVHE_HYPERVISOR__ 924 /* the nVHE hypervisor is always non-preemptible */ 925 #define __vcpu_flags_preempt_disable() 926 #define __vcpu_flags_preempt_enable() 927 #else 928 #define __vcpu_flags_preempt_disable() preempt_disable() 929 #define __vcpu_flags_preempt_enable() preempt_enable() 930 #endif 931 932 #define __vcpu_set_flag(v, flagset, f, m) \ 933 do { \ 934 typeof(v->arch.flagset) *fset; \ 935 \ 936 __build_check_flag(v, flagset, f, m); \ 937 \ 938 fset = &v->arch.flagset; \ 939 __vcpu_flags_preempt_disable(); \ 940 if (HWEIGHT(m) > 1) \ 941 *fset &= ~(m); \ 942 *fset |= (f); \ 943 __vcpu_flags_preempt_enable(); \ 944 } while (0) 945 946 #define __vcpu_clear_flag(v, flagset, f, m) \ 947 do { \ 948 typeof(v->arch.flagset) *fset; \ 949 \ 950 __build_check_flag(v, flagset, f, m); \ 951 \ 952 fset = &v->arch.flagset; \ 953 __vcpu_flags_preempt_disable(); \ 954 *fset &= ~(m); \ 955 __vcpu_flags_preempt_enable(); \ 956 } while (0) 957 958 #define vcpu_get_flag(v, ...) __vcpu_get_flag((v), __VA_ARGS__) 959 #define vcpu_set_flag(v, ...) __vcpu_set_flag((v), __VA_ARGS__) 960 #define vcpu_clear_flag(v, ...) __vcpu_clear_flag((v), __VA_ARGS__) 961 962 /* KVM_ARM_VCPU_INIT completed */ 963 #define VCPU_INITIALIZED __vcpu_single_flag(cflags, BIT(0)) 964 /* SVE config completed */ 965 #define VCPU_SVE_FINALIZED __vcpu_single_flag(cflags, BIT(1)) 966 /* pKVM VCPU setup completed */ 967 #define VCPU_PKVM_FINALIZED __vcpu_single_flag(cflags, BIT(2)) 968 969 /* Exception pending */ 970 #define PENDING_EXCEPTION __vcpu_single_flag(iflags, BIT(0)) 971 /* 972 * PC increment. Overlaps with EXCEPT_MASK on purpose so that it can't 973 * be set together with an exception... 974 */ 975 #define INCREMENT_PC __vcpu_single_flag(iflags, BIT(1)) 976 /* Target EL/MODE (not a single flag, but let's abuse the macro) */ 977 #define EXCEPT_MASK __vcpu_single_flag(iflags, GENMASK(3, 1)) 978 979 /* Helpers to encode exceptions with minimum fuss */ 980 #define __EXCEPT_MASK_VAL unpack_vcpu_flag(EXCEPT_MASK) 981 #define __EXCEPT_SHIFT __builtin_ctzl(__EXCEPT_MASK_VAL) 982 #define __vcpu_except_flags(_f) iflags, (_f << __EXCEPT_SHIFT), __EXCEPT_MASK_VAL 983 984 /* 985 * When PENDING_EXCEPTION is set, EXCEPT_MASK can take the following 986 * values: 987 * 988 * For AArch32 EL1: 989 */ 990 #define EXCEPT_AA32_UND __vcpu_except_flags(0) 991 #define EXCEPT_AA32_IABT __vcpu_except_flags(1) 992 #define EXCEPT_AA32_DABT __vcpu_except_flags(2) 993 /* For AArch64: */ 994 #define EXCEPT_AA64_EL1_SYNC __vcpu_except_flags(0) 995 #define EXCEPT_AA64_EL1_IRQ __vcpu_except_flags(1) 996 #define EXCEPT_AA64_EL1_FIQ __vcpu_except_flags(2) 997 #define EXCEPT_AA64_EL1_SERR __vcpu_except_flags(3) 998 /* For AArch64 with NV: */ 999 #define EXCEPT_AA64_EL2_SYNC __vcpu_except_flags(4) 1000 #define EXCEPT_AA64_EL2_IRQ __vcpu_except_flags(5) 1001 #define EXCEPT_AA64_EL2_FIQ __vcpu_except_flags(6) 1002 #define EXCEPT_AA64_EL2_SERR __vcpu_except_flags(7) 1003 1004 /* Physical CPU not in supported_cpus */ 1005 #define ON_UNSUPPORTED_CPU __vcpu_single_flag(sflags, BIT(0)) 1006 /* WFIT instruction trapped */ 1007 #define IN_WFIT __vcpu_single_flag(sflags, BIT(1)) 1008 /* vcpu system registers loaded on physical CPU */ 1009 #define SYSREGS_ON_CPU __vcpu_single_flag(sflags, BIT(2)) 1010 /* Software step state is Active-pending for external debug */ 1011 #define HOST_SS_ACTIVE_PENDING __vcpu_single_flag(sflags, BIT(3)) 1012 /* Software step state is Active pending for guest debug */ 1013 #define GUEST_SS_ACTIVE_PENDING __vcpu_single_flag(sflags, BIT(4)) 1014 /* PMUSERENR for the guest EL0 is on physical CPU */ 1015 #define PMUSERENR_ON_CPU __vcpu_single_flag(sflags, BIT(5)) 1016 /* WFI instruction trapped */ 1017 #define IN_WFI __vcpu_single_flag(sflags, BIT(6)) 1018 /* KVM is currently emulating a nested ERET */ 1019 #define IN_NESTED_ERET __vcpu_single_flag(sflags, BIT(7)) 1020 1021 1022 /* Pointer to the vcpu's SVE FFR for sve_{save,load}_state() */ 1023 #define vcpu_sve_pffr(vcpu) (kern_hyp_va((vcpu)->arch.sve_state) + \ 1024 sve_ffr_offset((vcpu)->arch.sve_max_vl)) 1025 1026 #define vcpu_sve_max_vq(vcpu) sve_vq_from_vl((vcpu)->arch.sve_max_vl) 1027 1028 #define vcpu_sve_zcr_elx(vcpu) \ 1029 (unlikely(is_hyp_ctxt(vcpu)) ? ZCR_EL2 : ZCR_EL1) 1030 1031 #define sve_state_size_from_vl(sve_max_vl) ({ \ 1032 size_t __size_ret; \ 1033 unsigned int __vq; \ 1034 \ 1035 if (WARN_ON(!sve_vl_valid(sve_max_vl))) { \ 1036 __size_ret = 0; \ 1037 } else { \ 1038 __vq = sve_vq_from_vl(sve_max_vl); \ 1039 __size_ret = SVE_SIG_REGS_SIZE(__vq); \ 1040 } \ 1041 \ 1042 __size_ret; \ 1043 }) 1044 1045 #define vcpu_sve_state_size(vcpu) sve_state_size_from_vl((vcpu)->arch.sve_max_vl) 1046 1047 #define KVM_GUESTDBG_VALID_MASK (KVM_GUESTDBG_ENABLE | \ 1048 KVM_GUESTDBG_USE_SW_BP | \ 1049 KVM_GUESTDBG_USE_HW | \ 1050 KVM_GUESTDBG_SINGLESTEP) 1051 1052 #define kvm_has_sve(kvm) (system_supports_sve() && \ 1053 test_bit(KVM_ARCH_FLAG_GUEST_HAS_SVE, &(kvm)->arch.flags)) 1054 1055 #ifdef __KVM_NVHE_HYPERVISOR__ 1056 #define vcpu_has_sve(vcpu) kvm_has_sve(kern_hyp_va((vcpu)->kvm)) 1057 #else 1058 #define vcpu_has_sve(vcpu) kvm_has_sve((vcpu)->kvm) 1059 #endif 1060 1061 #ifdef CONFIG_ARM64_PTR_AUTH 1062 #define vcpu_has_ptrauth(vcpu) \ 1063 ((cpus_have_final_cap(ARM64_HAS_ADDRESS_AUTH) || \ 1064 cpus_have_final_cap(ARM64_HAS_GENERIC_AUTH)) && \ 1065 (vcpu_has_feature(vcpu, KVM_ARM_VCPU_PTRAUTH_ADDRESS) || \ 1066 vcpu_has_feature(vcpu, KVM_ARM_VCPU_PTRAUTH_GENERIC))) 1067 #else 1068 #define vcpu_has_ptrauth(vcpu) false 1069 #endif 1070 1071 #define vcpu_on_unsupported_cpu(vcpu) \ 1072 vcpu_get_flag(vcpu, ON_UNSUPPORTED_CPU) 1073 1074 #define vcpu_set_on_unsupported_cpu(vcpu) \ 1075 vcpu_set_flag(vcpu, ON_UNSUPPORTED_CPU) 1076 1077 #define vcpu_clear_on_unsupported_cpu(vcpu) \ 1078 vcpu_clear_flag(vcpu, ON_UNSUPPORTED_CPU) 1079 1080 #define vcpu_gp_regs(v) (&(v)->arch.ctxt.regs) 1081 1082 /* 1083 * Only use __vcpu_sys_reg/ctxt_sys_reg if you know you want the 1084 * memory backed version of a register, and not the one most recently 1085 * accessed by a running VCPU. For example, for userspace access or 1086 * for system registers that are never context switched, but only 1087 * emulated. 1088 * 1089 * Don't bother with VNCR-based accesses in the nVHE code, it has no 1090 * business dealing with NV. 1091 */ 1092 static inline u64 *___ctxt_sys_reg(const struct kvm_cpu_context *ctxt, int r) 1093 { 1094 #if !defined (__KVM_NVHE_HYPERVISOR__) 1095 if (unlikely(cpus_have_final_cap(ARM64_HAS_NESTED_VIRT) && 1096 r >= __VNCR_START__ && ctxt->vncr_array)) 1097 return &ctxt->vncr_array[r - __VNCR_START__]; 1098 #endif 1099 return (u64 *)&ctxt->sys_regs[r]; 1100 } 1101 1102 #define __ctxt_sys_reg(c,r) \ 1103 ({ \ 1104 BUILD_BUG_ON(__builtin_constant_p(r) && \ 1105 (r) >= NR_SYS_REGS); \ 1106 ___ctxt_sys_reg(c, r); \ 1107 }) 1108 1109 #define ctxt_sys_reg(c,r) (*__ctxt_sys_reg(c,r)) 1110 1111 u64 kvm_vcpu_apply_reg_masks(const struct kvm_vcpu *, enum vcpu_sysreg, u64); 1112 #define __vcpu_sys_reg(v,r) \ 1113 (*({ \ 1114 const struct kvm_cpu_context *ctxt = &(v)->arch.ctxt; \ 1115 u64 *__r = __ctxt_sys_reg(ctxt, (r)); \ 1116 if (vcpu_has_nv((v)) && (r) >= __SANITISED_REG_START__) \ 1117 *__r = kvm_vcpu_apply_reg_masks((v), (r), *__r);\ 1118 __r; \ 1119 })) 1120 1121 u64 vcpu_read_sys_reg(const struct kvm_vcpu *vcpu, int reg); 1122 void vcpu_write_sys_reg(struct kvm_vcpu *vcpu, u64 val, int reg); 1123 1124 static inline bool __vcpu_read_sys_reg_from_cpu(int reg, u64 *val) 1125 { 1126 /* 1127 * *** VHE ONLY *** 1128 * 1129 * System registers listed in the switch are not saved on every 1130 * exit from the guest but are only saved on vcpu_put. 1131 * 1132 * Note that MPIDR_EL1 for the guest is set by KVM via VMPIDR_EL2 but 1133 * should never be listed below, because the guest cannot modify its 1134 * own MPIDR_EL1 and MPIDR_EL1 is accessed for VCPU A from VCPU B's 1135 * thread when emulating cross-VCPU communication. 1136 */ 1137 if (!has_vhe()) 1138 return false; 1139 1140 switch (reg) { 1141 case SCTLR_EL1: *val = read_sysreg_s(SYS_SCTLR_EL12); break; 1142 case CPACR_EL1: *val = read_sysreg_s(SYS_CPACR_EL12); break; 1143 case TTBR0_EL1: *val = read_sysreg_s(SYS_TTBR0_EL12); break; 1144 case TTBR1_EL1: *val = read_sysreg_s(SYS_TTBR1_EL12); break; 1145 case TCR_EL1: *val = read_sysreg_s(SYS_TCR_EL12); break; 1146 case TCR2_EL1: *val = read_sysreg_s(SYS_TCR2_EL12); break; 1147 case PIR_EL1: *val = read_sysreg_s(SYS_PIR_EL12); break; 1148 case PIRE0_EL1: *val = read_sysreg_s(SYS_PIRE0_EL12); break; 1149 case POR_EL1: *val = read_sysreg_s(SYS_POR_EL12); break; 1150 case ESR_EL1: *val = read_sysreg_s(SYS_ESR_EL12); break; 1151 case AFSR0_EL1: *val = read_sysreg_s(SYS_AFSR0_EL12); break; 1152 case AFSR1_EL1: *val = read_sysreg_s(SYS_AFSR1_EL12); break; 1153 case FAR_EL1: *val = read_sysreg_s(SYS_FAR_EL12); break; 1154 case MAIR_EL1: *val = read_sysreg_s(SYS_MAIR_EL12); break; 1155 case VBAR_EL1: *val = read_sysreg_s(SYS_VBAR_EL12); break; 1156 case CONTEXTIDR_EL1: *val = read_sysreg_s(SYS_CONTEXTIDR_EL12);break; 1157 case TPIDR_EL0: *val = read_sysreg_s(SYS_TPIDR_EL0); break; 1158 case TPIDRRO_EL0: *val = read_sysreg_s(SYS_TPIDRRO_EL0); break; 1159 case TPIDR_EL1: *val = read_sysreg_s(SYS_TPIDR_EL1); break; 1160 case AMAIR_EL1: *val = read_sysreg_s(SYS_AMAIR_EL12); break; 1161 case CNTKCTL_EL1: *val = read_sysreg_s(SYS_CNTKCTL_EL12); break; 1162 case ELR_EL1: *val = read_sysreg_s(SYS_ELR_EL12); break; 1163 case SPSR_EL1: *val = read_sysreg_s(SYS_SPSR_EL12); break; 1164 case PAR_EL1: *val = read_sysreg_par(); break; 1165 case DACR32_EL2: *val = read_sysreg_s(SYS_DACR32_EL2); break; 1166 case IFSR32_EL2: *val = read_sysreg_s(SYS_IFSR32_EL2); break; 1167 case DBGVCR32_EL2: *val = read_sysreg_s(SYS_DBGVCR32_EL2); break; 1168 case ZCR_EL1: *val = read_sysreg_s(SYS_ZCR_EL12); break; 1169 default: return false; 1170 } 1171 1172 return true; 1173 } 1174 1175 static inline bool __vcpu_write_sys_reg_to_cpu(u64 val, int reg) 1176 { 1177 /* 1178 * *** VHE ONLY *** 1179 * 1180 * System registers listed in the switch are not restored on every 1181 * entry to the guest but are only restored on vcpu_load. 1182 * 1183 * Note that MPIDR_EL1 for the guest is set by KVM via VMPIDR_EL2 but 1184 * should never be listed below, because the MPIDR should only be set 1185 * once, before running the VCPU, and never changed later. 1186 */ 1187 if (!has_vhe()) 1188 return false; 1189 1190 switch (reg) { 1191 case SCTLR_EL1: write_sysreg_s(val, SYS_SCTLR_EL12); break; 1192 case CPACR_EL1: write_sysreg_s(val, SYS_CPACR_EL12); break; 1193 case TTBR0_EL1: write_sysreg_s(val, SYS_TTBR0_EL12); break; 1194 case TTBR1_EL1: write_sysreg_s(val, SYS_TTBR1_EL12); break; 1195 case TCR_EL1: write_sysreg_s(val, SYS_TCR_EL12); break; 1196 case TCR2_EL1: write_sysreg_s(val, SYS_TCR2_EL12); break; 1197 case PIR_EL1: write_sysreg_s(val, SYS_PIR_EL12); break; 1198 case PIRE0_EL1: write_sysreg_s(val, SYS_PIRE0_EL12); break; 1199 case POR_EL1: write_sysreg_s(val, SYS_POR_EL12); break; 1200 case ESR_EL1: write_sysreg_s(val, SYS_ESR_EL12); break; 1201 case AFSR0_EL1: write_sysreg_s(val, SYS_AFSR0_EL12); break; 1202 case AFSR1_EL1: write_sysreg_s(val, SYS_AFSR1_EL12); break; 1203 case FAR_EL1: write_sysreg_s(val, SYS_FAR_EL12); break; 1204 case MAIR_EL1: write_sysreg_s(val, SYS_MAIR_EL12); break; 1205 case VBAR_EL1: write_sysreg_s(val, SYS_VBAR_EL12); break; 1206 case CONTEXTIDR_EL1: write_sysreg_s(val, SYS_CONTEXTIDR_EL12);break; 1207 case TPIDR_EL0: write_sysreg_s(val, SYS_TPIDR_EL0); break; 1208 case TPIDRRO_EL0: write_sysreg_s(val, SYS_TPIDRRO_EL0); break; 1209 case TPIDR_EL1: write_sysreg_s(val, SYS_TPIDR_EL1); break; 1210 case AMAIR_EL1: write_sysreg_s(val, SYS_AMAIR_EL12); break; 1211 case CNTKCTL_EL1: write_sysreg_s(val, SYS_CNTKCTL_EL12); break; 1212 case ELR_EL1: write_sysreg_s(val, SYS_ELR_EL12); break; 1213 case SPSR_EL1: write_sysreg_s(val, SYS_SPSR_EL12); break; 1214 case PAR_EL1: write_sysreg_s(val, SYS_PAR_EL1); break; 1215 case DACR32_EL2: write_sysreg_s(val, SYS_DACR32_EL2); break; 1216 case IFSR32_EL2: write_sysreg_s(val, SYS_IFSR32_EL2); break; 1217 case DBGVCR32_EL2: write_sysreg_s(val, SYS_DBGVCR32_EL2); break; 1218 case ZCR_EL1: write_sysreg_s(val, SYS_ZCR_EL12); break; 1219 default: return false; 1220 } 1221 1222 return true; 1223 } 1224 1225 struct kvm_vm_stat { 1226 struct kvm_vm_stat_generic generic; 1227 }; 1228 1229 struct kvm_vcpu_stat { 1230 struct kvm_vcpu_stat_generic generic; 1231 u64 hvc_exit_stat; 1232 u64 wfe_exit_stat; 1233 u64 wfi_exit_stat; 1234 u64 mmio_exit_user; 1235 u64 mmio_exit_kernel; 1236 u64 signal_exits; 1237 u64 exits; 1238 }; 1239 1240 unsigned long kvm_arm_num_regs(struct kvm_vcpu *vcpu); 1241 int kvm_arm_copy_reg_indices(struct kvm_vcpu *vcpu, u64 __user *indices); 1242 int kvm_arm_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg); 1243 int kvm_arm_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg); 1244 1245 unsigned long kvm_arm_num_sys_reg_descs(struct kvm_vcpu *vcpu); 1246 int kvm_arm_copy_sys_reg_indices(struct kvm_vcpu *vcpu, u64 __user *uindices); 1247 1248 int __kvm_arm_vcpu_get_events(struct kvm_vcpu *vcpu, 1249 struct kvm_vcpu_events *events); 1250 1251 int __kvm_arm_vcpu_set_events(struct kvm_vcpu *vcpu, 1252 struct kvm_vcpu_events *events); 1253 1254 void kvm_arm_halt_guest(struct kvm *kvm); 1255 void kvm_arm_resume_guest(struct kvm *kvm); 1256 1257 #define vcpu_has_run_once(vcpu) (!!READ_ONCE((vcpu)->pid)) 1258 1259 #ifndef __KVM_NVHE_HYPERVISOR__ 1260 #define kvm_call_hyp_nvhe(f, ...) \ 1261 ({ \ 1262 struct arm_smccc_res res; \ 1263 \ 1264 arm_smccc_1_1_hvc(KVM_HOST_SMCCC_FUNC(f), \ 1265 ##__VA_ARGS__, &res); \ 1266 WARN_ON(res.a0 != SMCCC_RET_SUCCESS); \ 1267 \ 1268 res.a1; \ 1269 }) 1270 1271 /* 1272 * The couple of isb() below are there to guarantee the same behaviour 1273 * on VHE as on !VHE, where the eret to EL1 acts as a context 1274 * synchronization event. 1275 */ 1276 #define kvm_call_hyp(f, ...) \ 1277 do { \ 1278 if (has_vhe()) { \ 1279 f(__VA_ARGS__); \ 1280 isb(); \ 1281 } else { \ 1282 kvm_call_hyp_nvhe(f, ##__VA_ARGS__); \ 1283 } \ 1284 } while(0) 1285 1286 #define kvm_call_hyp_ret(f, ...) \ 1287 ({ \ 1288 typeof(f(__VA_ARGS__)) ret; \ 1289 \ 1290 if (has_vhe()) { \ 1291 ret = f(__VA_ARGS__); \ 1292 isb(); \ 1293 } else { \ 1294 ret = kvm_call_hyp_nvhe(f, ##__VA_ARGS__); \ 1295 } \ 1296 \ 1297 ret; \ 1298 }) 1299 #else /* __KVM_NVHE_HYPERVISOR__ */ 1300 #define kvm_call_hyp(f, ...) f(__VA_ARGS__) 1301 #define kvm_call_hyp_ret(f, ...) f(__VA_ARGS__) 1302 #define kvm_call_hyp_nvhe(f, ...) f(__VA_ARGS__) 1303 #endif /* __KVM_NVHE_HYPERVISOR__ */ 1304 1305 int handle_exit(struct kvm_vcpu *vcpu, int exception_index); 1306 void handle_exit_early(struct kvm_vcpu *vcpu, int exception_index); 1307 1308 int kvm_handle_cp14_load_store(struct kvm_vcpu *vcpu); 1309 int kvm_handle_cp14_32(struct kvm_vcpu *vcpu); 1310 int kvm_handle_cp14_64(struct kvm_vcpu *vcpu); 1311 int kvm_handle_cp15_32(struct kvm_vcpu *vcpu); 1312 int kvm_handle_cp15_64(struct kvm_vcpu *vcpu); 1313 int kvm_handle_sys_reg(struct kvm_vcpu *vcpu); 1314 int kvm_handle_cp10_id(struct kvm_vcpu *vcpu); 1315 1316 void kvm_sys_regs_create_debugfs(struct kvm *kvm); 1317 void kvm_reset_sys_regs(struct kvm_vcpu *vcpu); 1318 1319 int __init kvm_sys_reg_table_init(void); 1320 struct sys_reg_desc; 1321 int __init populate_sysreg_config(const struct sys_reg_desc *sr, 1322 unsigned int idx); 1323 int __init populate_nv_trap_config(void); 1324 1325 void kvm_calculate_traps(struct kvm_vcpu *vcpu); 1326 1327 /* MMIO helpers */ 1328 void kvm_mmio_write_buf(void *buf, unsigned int len, unsigned long data); 1329 unsigned long kvm_mmio_read_buf(const void *buf, unsigned int len); 1330 1331 int kvm_handle_mmio_return(struct kvm_vcpu *vcpu); 1332 int io_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa); 1333 1334 /* 1335 * Returns true if a Performance Monitoring Interrupt (PMI), a.k.a. perf event, 1336 * arrived in guest context. For arm64, any event that arrives while a vCPU is 1337 * loaded is considered to be "in guest". 1338 */ 1339 static inline bool kvm_arch_pmi_in_guest(struct kvm_vcpu *vcpu) 1340 { 1341 return IS_ENABLED(CONFIG_GUEST_PERF_EVENTS) && !!vcpu; 1342 } 1343 1344 long kvm_hypercall_pv_features(struct kvm_vcpu *vcpu); 1345 gpa_t kvm_init_stolen_time(struct kvm_vcpu *vcpu); 1346 void kvm_update_stolen_time(struct kvm_vcpu *vcpu); 1347 1348 bool kvm_arm_pvtime_supported(void); 1349 int kvm_arm_pvtime_set_attr(struct kvm_vcpu *vcpu, 1350 struct kvm_device_attr *attr); 1351 int kvm_arm_pvtime_get_attr(struct kvm_vcpu *vcpu, 1352 struct kvm_device_attr *attr); 1353 int kvm_arm_pvtime_has_attr(struct kvm_vcpu *vcpu, 1354 struct kvm_device_attr *attr); 1355 1356 extern unsigned int __ro_after_init kvm_arm_vmid_bits; 1357 int __init kvm_arm_vmid_alloc_init(void); 1358 void __init kvm_arm_vmid_alloc_free(void); 1359 void kvm_arm_vmid_update(struct kvm_vmid *kvm_vmid); 1360 void kvm_arm_vmid_clear_active(void); 1361 1362 static inline void kvm_arm_pvtime_vcpu_init(struct kvm_vcpu_arch *vcpu_arch) 1363 { 1364 vcpu_arch->steal.base = INVALID_GPA; 1365 } 1366 1367 static inline bool kvm_arm_is_pvtime_enabled(struct kvm_vcpu_arch *vcpu_arch) 1368 { 1369 return (vcpu_arch->steal.base != INVALID_GPA); 1370 } 1371 1372 void kvm_set_sei_esr(struct kvm_vcpu *vcpu, u64 syndrome); 1373 1374 struct kvm_vcpu *kvm_mpidr_to_vcpu(struct kvm *kvm, unsigned long mpidr); 1375 1376 DECLARE_KVM_HYP_PER_CPU(struct kvm_host_data, kvm_host_data); 1377 1378 /* 1379 * How we access per-CPU host data depends on the where we access it from, 1380 * and the mode we're in: 1381 * 1382 * - VHE and nVHE hypervisor bits use their locally defined instance 1383 * 1384 * - the rest of the kernel use either the VHE or nVHE one, depending on 1385 * the mode we're running in. 1386 * 1387 * Unless we're in protected mode, fully deprivileged, and the nVHE 1388 * per-CPU stuff is exclusively accessible to the protected EL2 code. 1389 * In this case, the EL1 code uses the *VHE* data as its private state 1390 * (which makes sense in a way as there shouldn't be any shared state 1391 * between the host and the hypervisor). 1392 * 1393 * Yes, this is all totally trivial. Shoot me now. 1394 */ 1395 #if defined(__KVM_NVHE_HYPERVISOR__) || defined(__KVM_VHE_HYPERVISOR__) 1396 #define host_data_ptr(f) (&this_cpu_ptr(&kvm_host_data)->f) 1397 #else 1398 #define host_data_ptr(f) \ 1399 (static_branch_unlikely(&kvm_protected_mode_initialized) ? \ 1400 &this_cpu_ptr(&kvm_host_data)->f : \ 1401 &this_cpu_ptr_hyp_sym(kvm_host_data)->f) 1402 #endif 1403 1404 #define host_data_test_flag(flag) \ 1405 (test_bit(KVM_HOST_DATA_FLAG_##flag, host_data_ptr(flags))) 1406 #define host_data_set_flag(flag) \ 1407 set_bit(KVM_HOST_DATA_FLAG_##flag, host_data_ptr(flags)) 1408 #define host_data_clear_flag(flag) \ 1409 clear_bit(KVM_HOST_DATA_FLAG_##flag, host_data_ptr(flags)) 1410 1411 /* Check whether the FP regs are owned by the guest */ 1412 static inline bool guest_owns_fp_regs(void) 1413 { 1414 return *host_data_ptr(fp_owner) == FP_STATE_GUEST_OWNED; 1415 } 1416 1417 /* Check whether the FP regs are owned by the host */ 1418 static inline bool host_owns_fp_regs(void) 1419 { 1420 return *host_data_ptr(fp_owner) == FP_STATE_HOST_OWNED; 1421 } 1422 1423 static inline void kvm_init_host_cpu_context(struct kvm_cpu_context *cpu_ctxt) 1424 { 1425 /* The host's MPIDR is immutable, so let's set it up at boot time */ 1426 ctxt_sys_reg(cpu_ctxt, MPIDR_EL1) = read_cpuid_mpidr(); 1427 } 1428 1429 static inline bool kvm_system_needs_idmapped_vectors(void) 1430 { 1431 return cpus_have_final_cap(ARM64_SPECTRE_V3A); 1432 } 1433 1434 void kvm_init_host_debug_data(void); 1435 void kvm_vcpu_load_debug(struct kvm_vcpu *vcpu); 1436 void kvm_vcpu_put_debug(struct kvm_vcpu *vcpu); 1437 void kvm_debug_set_guest_ownership(struct kvm_vcpu *vcpu); 1438 void kvm_debug_handle_oslar(struct kvm_vcpu *vcpu, u64 val); 1439 1440 #define kvm_vcpu_os_lock_enabled(vcpu) \ 1441 (!!(__vcpu_sys_reg(vcpu, OSLSR_EL1) & OSLSR_EL1_OSLK)) 1442 1443 #define kvm_debug_regs_in_use(vcpu) \ 1444 ((vcpu)->arch.debug_owner != VCPU_DEBUG_FREE) 1445 #define kvm_host_owns_debug_regs(vcpu) \ 1446 ((vcpu)->arch.debug_owner == VCPU_DEBUG_HOST_OWNED) 1447 #define kvm_guest_owns_debug_regs(vcpu) \ 1448 ((vcpu)->arch.debug_owner == VCPU_DEBUG_GUEST_OWNED) 1449 1450 int kvm_arm_vcpu_arch_set_attr(struct kvm_vcpu *vcpu, 1451 struct kvm_device_attr *attr); 1452 int kvm_arm_vcpu_arch_get_attr(struct kvm_vcpu *vcpu, 1453 struct kvm_device_attr *attr); 1454 int kvm_arm_vcpu_arch_has_attr(struct kvm_vcpu *vcpu, 1455 struct kvm_device_attr *attr); 1456 1457 int kvm_vm_ioctl_mte_copy_tags(struct kvm *kvm, 1458 struct kvm_arm_copy_mte_tags *copy_tags); 1459 int kvm_vm_ioctl_set_counter_offset(struct kvm *kvm, 1460 struct kvm_arm_counter_offset *offset); 1461 int kvm_vm_ioctl_get_reg_writable_masks(struct kvm *kvm, 1462 struct reg_mask_range *range); 1463 1464 /* Guest/host FPSIMD coordination helpers */ 1465 int kvm_arch_vcpu_run_map_fp(struct kvm_vcpu *vcpu); 1466 void kvm_arch_vcpu_load_fp(struct kvm_vcpu *vcpu); 1467 void kvm_arch_vcpu_ctxflush_fp(struct kvm_vcpu *vcpu); 1468 void kvm_arch_vcpu_ctxsync_fp(struct kvm_vcpu *vcpu); 1469 void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu); 1470 1471 static inline bool kvm_pmu_counter_deferred(struct perf_event_attr *attr) 1472 { 1473 return (!has_vhe() && attr->exclude_host); 1474 } 1475 1476 #ifdef CONFIG_KVM 1477 void kvm_set_pmu_events(u64 set, struct perf_event_attr *attr); 1478 void kvm_clr_pmu_events(u64 clr); 1479 bool kvm_set_pmuserenr(u64 val); 1480 void kvm_enable_trbe(void); 1481 void kvm_disable_trbe(void); 1482 void kvm_tracing_set_el1_configuration(u64 trfcr_while_in_guest); 1483 #else 1484 static inline void kvm_set_pmu_events(u64 set, struct perf_event_attr *attr) {} 1485 static inline void kvm_clr_pmu_events(u64 clr) {} 1486 static inline bool kvm_set_pmuserenr(u64 val) 1487 { 1488 return false; 1489 } 1490 static inline void kvm_enable_trbe(void) {} 1491 static inline void kvm_disable_trbe(void) {} 1492 static inline void kvm_tracing_set_el1_configuration(u64 trfcr_while_in_guest) {} 1493 #endif 1494 1495 void kvm_vcpu_load_vhe(struct kvm_vcpu *vcpu); 1496 void kvm_vcpu_put_vhe(struct kvm_vcpu *vcpu); 1497 1498 int __init kvm_set_ipa_limit(void); 1499 u32 kvm_get_pa_bits(struct kvm *kvm); 1500 1501 #define __KVM_HAVE_ARCH_VM_ALLOC 1502 struct kvm *kvm_arch_alloc_vm(void); 1503 1504 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS 1505 1506 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS_RANGE 1507 1508 #define kvm_vm_is_protected(kvm) (is_protected_kvm_enabled() && (kvm)->arch.pkvm.enabled) 1509 1510 #define vcpu_is_protected(vcpu) kvm_vm_is_protected((vcpu)->kvm) 1511 1512 int kvm_arm_vcpu_finalize(struct kvm_vcpu *vcpu, int feature); 1513 bool kvm_arm_vcpu_is_finalized(struct kvm_vcpu *vcpu); 1514 1515 #define kvm_arm_vcpu_sve_finalized(vcpu) vcpu_get_flag(vcpu, VCPU_SVE_FINALIZED) 1516 1517 #define kvm_has_mte(kvm) \ 1518 (system_supports_mte() && \ 1519 test_bit(KVM_ARCH_FLAG_MTE_ENABLED, &(kvm)->arch.flags)) 1520 1521 #define kvm_supports_32bit_el0() \ 1522 (system_supports_32bit_el0() && \ 1523 !static_branch_unlikely(&arm64_mismatched_32bit_el0)) 1524 1525 #define kvm_vm_has_ran_once(kvm) \ 1526 (test_bit(KVM_ARCH_FLAG_HAS_RAN_ONCE, &(kvm)->arch.flags)) 1527 1528 static inline bool __vcpu_has_feature(const struct kvm_arch *ka, int feature) 1529 { 1530 return test_bit(feature, ka->vcpu_features); 1531 } 1532 1533 #define kvm_vcpu_has_feature(k, f) __vcpu_has_feature(&(k)->arch, (f)) 1534 #define vcpu_has_feature(v, f) __vcpu_has_feature(&(v)->kvm->arch, (f)) 1535 1536 #define kvm_vcpu_initialized(v) vcpu_get_flag(vcpu, VCPU_INITIALIZED) 1537 1538 int kvm_trng_call(struct kvm_vcpu *vcpu); 1539 #ifdef CONFIG_KVM 1540 extern phys_addr_t hyp_mem_base; 1541 extern phys_addr_t hyp_mem_size; 1542 void __init kvm_hyp_reserve(void); 1543 #else 1544 static inline void kvm_hyp_reserve(void) { } 1545 #endif 1546 1547 void kvm_arm_vcpu_power_off(struct kvm_vcpu *vcpu); 1548 bool kvm_arm_vcpu_stopped(struct kvm_vcpu *vcpu); 1549 1550 static inline u64 *__vm_id_reg(struct kvm_arch *ka, u32 reg) 1551 { 1552 switch (reg) { 1553 case sys_reg(3, 0, 0, 1, 0) ... sys_reg(3, 0, 0, 7, 7): 1554 return &ka->id_regs[IDREG_IDX(reg)]; 1555 case SYS_CTR_EL0: 1556 return &ka->ctr_el0; 1557 case SYS_MIDR_EL1: 1558 return &ka->midr_el1; 1559 case SYS_REVIDR_EL1: 1560 return &ka->revidr_el1; 1561 case SYS_AIDR_EL1: 1562 return &ka->aidr_el1; 1563 default: 1564 WARN_ON_ONCE(1); 1565 return NULL; 1566 } 1567 } 1568 1569 #define kvm_read_vm_id_reg(kvm, reg) \ 1570 ({ u64 __val = *__vm_id_reg(&(kvm)->arch, reg); __val; }) 1571 1572 void kvm_set_vm_id_reg(struct kvm *kvm, u32 reg, u64 val); 1573 1574 #define __expand_field_sign_unsigned(id, fld, val) \ 1575 ((u64)SYS_FIELD_VALUE(id, fld, val)) 1576 1577 #define __expand_field_sign_signed(id, fld, val) \ 1578 ({ \ 1579 u64 __val = SYS_FIELD_VALUE(id, fld, val); \ 1580 sign_extend64(__val, id##_##fld##_WIDTH - 1); \ 1581 }) 1582 1583 #define get_idreg_field_unsigned(kvm, id, fld) \ 1584 ({ \ 1585 u64 __val = kvm_read_vm_id_reg((kvm), SYS_##id); \ 1586 FIELD_GET(id##_##fld##_MASK, __val); \ 1587 }) 1588 1589 #define get_idreg_field_signed(kvm, id, fld) \ 1590 ({ \ 1591 u64 __val = get_idreg_field_unsigned(kvm, id, fld); \ 1592 sign_extend64(__val, id##_##fld##_WIDTH - 1); \ 1593 }) 1594 1595 #define get_idreg_field_enum(kvm, id, fld) \ 1596 get_idreg_field_unsigned(kvm, id, fld) 1597 1598 #define kvm_cmp_feat_signed(kvm, id, fld, op, limit) \ 1599 (get_idreg_field_signed((kvm), id, fld) op __expand_field_sign_signed(id, fld, limit)) 1600 1601 #define kvm_cmp_feat_unsigned(kvm, id, fld, op, limit) \ 1602 (get_idreg_field_unsigned((kvm), id, fld) op __expand_field_sign_unsigned(id, fld, limit)) 1603 1604 #define kvm_cmp_feat(kvm, id, fld, op, limit) \ 1605 (id##_##fld##_SIGNED ? \ 1606 kvm_cmp_feat_signed(kvm, id, fld, op, limit) : \ 1607 kvm_cmp_feat_unsigned(kvm, id, fld, op, limit)) 1608 1609 #define __kvm_has_feat(kvm, id, fld, limit) \ 1610 kvm_cmp_feat(kvm, id, fld, >=, limit) 1611 1612 #define kvm_has_feat(kvm, ...) __kvm_has_feat(kvm, __VA_ARGS__) 1613 1614 #define __kvm_has_feat_enum(kvm, id, fld, val) \ 1615 kvm_cmp_feat_unsigned(kvm, id, fld, ==, val) 1616 1617 #define kvm_has_feat_enum(kvm, ...) __kvm_has_feat_enum(kvm, __VA_ARGS__) 1618 1619 #define kvm_has_feat_range(kvm, id, fld, min, max) \ 1620 (kvm_cmp_feat(kvm, id, fld, >=, min) && \ 1621 kvm_cmp_feat(kvm, id, fld, <=, max)) 1622 1623 /* Check for a given level of PAuth support */ 1624 #define kvm_has_pauth(k, l) \ 1625 ({ \ 1626 bool pa, pi, pa3; \ 1627 \ 1628 pa = kvm_has_feat((k), ID_AA64ISAR1_EL1, APA, l); \ 1629 pa &= kvm_has_feat((k), ID_AA64ISAR1_EL1, GPA, IMP); \ 1630 pi = kvm_has_feat((k), ID_AA64ISAR1_EL1, API, l); \ 1631 pi &= kvm_has_feat((k), ID_AA64ISAR1_EL1, GPI, IMP); \ 1632 pa3 = kvm_has_feat((k), ID_AA64ISAR2_EL1, APA3, l); \ 1633 pa3 &= kvm_has_feat((k), ID_AA64ISAR2_EL1, GPA3, IMP); \ 1634 \ 1635 (pa + pi + pa3) == 1; \ 1636 }) 1637 1638 #define kvm_has_fpmr(k) \ 1639 (system_supports_fpmr() && \ 1640 kvm_has_feat((k), ID_AA64PFR2_EL1, FPMR, IMP)) 1641 1642 #define kvm_has_tcr2(k) \ 1643 (kvm_has_feat((k), ID_AA64MMFR3_EL1, TCRX, IMP)) 1644 1645 #define kvm_has_s1pie(k) \ 1646 (kvm_has_feat((k), ID_AA64MMFR3_EL1, S1PIE, IMP)) 1647 1648 #define kvm_has_s1poe(k) \ 1649 (kvm_has_feat((k), ID_AA64MMFR3_EL1, S1POE, IMP)) 1650 1651 static inline bool kvm_arch_has_irq_bypass(void) 1652 { 1653 return true; 1654 } 1655 1656 void compute_fgu(struct kvm *kvm, enum fgt_group_id fgt); 1657 void get_reg_fixed_bits(struct kvm *kvm, enum vcpu_sysreg reg, u64 *res0, u64 *res1); 1658 void check_feature_map(void); 1659 1660 1661 #endif /* __ARM64_KVM_HOST_H__ */ 1662