1 /* SPDX-License-Identifier: GPL-2.0-only */
3 * Copyright (C) 2012,2013 - ARM Ltd
4 * Author: Marc Zyngier <marc.zyngier@arm.com>
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>
11 #ifndef __ARM64_KVM_HOST_H__
12 #define __ARM64_KVM_HOST_H__
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>
29 #include <asm/kvm_asm.h>
31 #define __KVM_HAVE_ARCH_INTC_INITIALIZED
33 #define KVM_HALT_POLL_NS_DEFAULT 500000
35 #include <kvm/arm_vgic.h>
36 #include <kvm/arm_arch_timer.h>
37 #include <kvm/arm_pmu.h>
39 #define KVM_MAX_VCPUS VGIC_V3_MAX_CPUS
41 #define KVM_VCPU_MAX_FEATURES 7
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 #define KVM_REQ_SUSPEND KVM_ARCH_REQ(6)
52 #define KVM_DIRTY_LOG_MANUAL_CAPS (KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE | \
53 KVM_DIRTY_LOG_INITIALLY_SET)
55 #define KVM_HAVE_MMU_RWLOCK
58 * Mode of operation configurable with kvm-arm.mode early param.
59 * See Documentation/admin-guide/kernel-parameters.txt for more information.
68 enum kvm_mode kvm_get_mode(void);
70 static inline enum kvm_mode kvm_get_mode(void) { return KVM_MODE_NONE; };
73 DECLARE_STATIC_KEY_FALSE(userspace_irqchip_in_use);
75 extern unsigned int __ro_after_init kvm_sve_max_vl;
76 int __init kvm_arm_init_sve(void);
78 u32 __attribute_const__ kvm_target_cpu(void);
79 int kvm_reset_vcpu(struct kvm_vcpu *vcpu);
80 void kvm_arm_vcpu_destroy(struct kvm_vcpu *vcpu);
82 struct kvm_hyp_memcache {
84 unsigned long nr_pages;
87 static inline void push_hyp_memcache(struct kvm_hyp_memcache *mc,
89 phys_addr_t (*to_pa)(void *virt))
96 static inline void *pop_hyp_memcache(struct kvm_hyp_memcache *mc,
97 void *(*to_va)(phys_addr_t phys))
99 phys_addr_t *p = to_va(mc->head);
110 static inline int __topup_hyp_memcache(struct kvm_hyp_memcache *mc,
111 unsigned long min_pages,
112 void *(*alloc_fn)(void *arg),
113 phys_addr_t (*to_pa)(void *virt),
116 while (mc->nr_pages < min_pages) {
117 phys_addr_t *p = alloc_fn(arg);
121 push_hyp_memcache(mc, p, to_pa);
127 static inline void __free_hyp_memcache(struct kvm_hyp_memcache *mc,
128 void (*free_fn)(void *virt, void *arg),
129 void *(*to_va)(phys_addr_t phys),
133 free_fn(pop_hyp_memcache(mc, to_va), arg);
136 void free_hyp_memcache(struct kvm_hyp_memcache *mc);
137 int topup_hyp_memcache(struct kvm_hyp_memcache *mc, unsigned long min_pages);
144 struct kvm_vmid vmid;
147 * stage2 entry level table
149 * Two kvm_s2_mmu structures in the same VM can point to the same
150 * pgd here. This happens when running a guest using a
151 * translation regime that isn't affected by its own stage-2
152 * translation, such as a non-VHE hypervisor running at vEL2, or
153 * for vEL1/EL0 with vHCR_EL2.VM == 0. In that case, we use the
154 * canonical stage-2 page tables.
156 phys_addr_t pgd_phys;
157 struct kvm_pgtable *pgt;
159 /* The last vcpu id that ran on each physical CPU */
160 int __percpu *last_vcpu_ran;
162 struct kvm_arch *arch;
165 struct kvm_arch_memory_slot {
169 * struct kvm_smccc_features: Descriptor of the hypercall services exposed to the guests
171 * @std_bmap: Bitmap of standard secure service calls
172 * @std_hyp_bmap: Bitmap of standard hypervisor service calls
173 * @vendor_hyp_bmap: Bitmap of vendor specific hypervisor service calls
175 struct kvm_smccc_features {
176 unsigned long std_bmap;
177 unsigned long std_hyp_bmap;
178 unsigned long vendor_hyp_bmap;
181 typedef unsigned int pkvm_handle_t;
183 struct kvm_protected_vm {
184 pkvm_handle_t handle;
185 struct kvm_hyp_memcache teardown_mc;
189 struct kvm_s2_mmu mmu;
191 /* VTCR_EL2 value for this VM */
194 /* Interrupt controller */
195 struct vgic_dist vgic;
198 struct arch_timer_vm_data timer_data;
200 /* Mandated version of PSCI */
203 /* Protects VM-scoped configuration data */
204 struct mutex config_lock;
207 * If we encounter a data abort without valid instruction syndrome
208 * information, report this to user space. User space can (and
209 * should) opt in to this feature if KVM_CAP_ARM_NISV_TO_USER is
212 #define KVM_ARCH_FLAG_RETURN_NISV_IO_ABORT_TO_USER 0
213 /* Memory Tagging Extension enabled for the guest */
214 #define KVM_ARCH_FLAG_MTE_ENABLED 1
215 /* At least one vCPU has ran in the VM */
216 #define KVM_ARCH_FLAG_HAS_RAN_ONCE 2
218 * The following two bits are used to indicate the guest's EL1
219 * register width configuration. A value of KVM_ARCH_FLAG_EL1_32BIT
220 * bit is valid only when KVM_ARCH_FLAG_REG_WIDTH_CONFIGURED is set.
221 * Otherwise, the guest's EL1 register width has not yet been
224 #define KVM_ARCH_FLAG_REG_WIDTH_CONFIGURED 3
225 #define KVM_ARCH_FLAG_EL1_32BIT 4
226 /* PSCI SYSTEM_SUSPEND enabled for the guest */
227 #define KVM_ARCH_FLAG_SYSTEM_SUSPEND_ENABLED 5
228 /* VM counter offset */
229 #define KVM_ARCH_FLAG_VM_COUNTER_OFFSET 6
230 /* Timer PPIs made immutable */
231 #define KVM_ARCH_FLAG_TIMER_PPIS_IMMUTABLE 7
232 /* SMCCC filter initialized for the VM */
233 #define KVM_ARCH_FLAG_SMCCC_FILTER_CONFIGURED 8
237 * VM-wide PMU filter, implemented as a bitmap and big enough for
238 * up to 2^10 events (ARMv8.0) or 2^16 events (ARMv8.1+).
240 unsigned long *pmu_filter;
241 struct arm_pmu *arm_pmu;
243 cpumask_var_t supported_cpus;
252 /* Hypercall features firmware registers' descriptor */
253 struct kvm_smccc_features smccc_feat;
254 struct maple_tree smccc_filter;
257 * For an untrusted host VM, 'pkvm.handle' is used to lookup
258 * the associated pKVM instance in the hypervisor.
260 struct kvm_protected_vm pkvm;
263 struct kvm_vcpu_fault_info {
264 u64 esr_el2; /* Hyp Syndrom Register */
265 u64 far_el2; /* Hyp Fault Address Register */
266 u64 hpfar_el2; /* Hyp IPA Fault Address Register */
267 u64 disr_el1; /* Deferred [SError] Status Register */
271 __INVALID_SYSREG__, /* 0 is reserved as an invalid value */
272 MPIDR_EL1, /* MultiProcessor Affinity Register */
273 CLIDR_EL1, /* Cache Level ID Register */
274 CSSELR_EL1, /* Cache Size Selection Register */
275 SCTLR_EL1, /* System Control Register */
276 ACTLR_EL1, /* Auxiliary Control Register */
277 CPACR_EL1, /* Coprocessor Access Control */
278 ZCR_EL1, /* SVE Control */
279 TTBR0_EL1, /* Translation Table Base Register 0 */
280 TTBR1_EL1, /* Translation Table Base Register 1 */
281 TCR_EL1, /* Translation Control Register */
282 TCR2_EL1, /* Extended Translation Control Register */
283 ESR_EL1, /* Exception Syndrome Register */
284 AFSR0_EL1, /* Auxiliary Fault Status Register 0 */
285 AFSR1_EL1, /* Auxiliary Fault Status Register 1 */
286 FAR_EL1, /* Fault Address Register */
287 MAIR_EL1, /* Memory Attribute Indirection Register */
288 VBAR_EL1, /* Vector Base Address Register */
289 CONTEXTIDR_EL1, /* Context ID Register */
290 TPIDR_EL0, /* Thread ID, User R/W */
291 TPIDRRO_EL0, /* Thread ID, User R/O */
292 TPIDR_EL1, /* Thread ID, Privileged */
293 AMAIR_EL1, /* Aux Memory Attribute Indirection Register */
294 CNTKCTL_EL1, /* Timer Control Register (EL1) */
295 PAR_EL1, /* Physical Address Register */
296 MDSCR_EL1, /* Monitor Debug System Control Register */
297 MDCCINT_EL1, /* Monitor Debug Comms Channel Interrupt Enable Reg */
298 OSLSR_EL1, /* OS Lock Status Register */
299 DISR_EL1, /* Deferred Interrupt Status Register */
301 /* Performance Monitors Registers */
302 PMCR_EL0, /* Control Register */
303 PMSELR_EL0, /* Event Counter Selection Register */
304 PMEVCNTR0_EL0, /* Event Counter Register (0-30) */
305 PMEVCNTR30_EL0 = PMEVCNTR0_EL0 + 30,
306 PMCCNTR_EL0, /* Cycle Counter Register */
307 PMEVTYPER0_EL0, /* Event Type Register (0-30) */
308 PMEVTYPER30_EL0 = PMEVTYPER0_EL0 + 30,
309 PMCCFILTR_EL0, /* Cycle Count Filter Register */
310 PMCNTENSET_EL0, /* Count Enable Set Register */
311 PMINTENSET_EL1, /* Interrupt Enable Set Register */
312 PMOVSSET_EL0, /* Overflow Flag Status Set Register */
313 PMUSERENR_EL0, /* User Enable Register */
315 /* Pointer Authentication Registers in a strict increasing order. */
337 /* Memory Tagging Extension registers */
338 RGSR_EL1, /* Random Allocation Tag Seed Register */
339 GCR_EL1, /* Tag Control Register */
340 TFSR_EL1, /* Tag Fault Status Register (EL1) */
341 TFSRE0_EL1, /* Tag Fault Status Register (EL0) */
343 /* Permission Indirection Extension registers */
344 PIR_EL1, /* Permission Indirection Register 1 (EL1) */
345 PIRE0_EL1, /* Permission Indirection Register 0 (EL1) */
347 /* 32bit specific registers. */
348 DACR32_EL2, /* Domain Access Control Register */
349 IFSR32_EL2, /* Instruction Fault Status Register */
350 FPEXC32_EL2, /* Floating-Point Exception Control Register */
351 DBGVCR32_EL2, /* Debug Vector Catch Register */
354 VPIDR_EL2, /* Virtualization Processor ID Register */
355 VMPIDR_EL2, /* Virtualization Multiprocessor ID Register */
356 SCTLR_EL2, /* System Control Register (EL2) */
357 ACTLR_EL2, /* Auxiliary Control Register (EL2) */
358 HCR_EL2, /* Hypervisor Configuration Register */
359 MDCR_EL2, /* Monitor Debug Configuration Register (EL2) */
360 CPTR_EL2, /* Architectural Feature Trap Register (EL2) */
361 HSTR_EL2, /* Hypervisor System Trap Register */
362 HACR_EL2, /* Hypervisor Auxiliary Control Register */
363 TTBR0_EL2, /* Translation Table Base Register 0 (EL2) */
364 TTBR1_EL2, /* Translation Table Base Register 1 (EL2) */
365 TCR_EL2, /* Translation Control Register (EL2) */
366 VTTBR_EL2, /* Virtualization Translation Table Base Register */
367 VTCR_EL2, /* Virtualization Translation Control Register */
368 SPSR_EL2, /* EL2 saved program status register */
369 ELR_EL2, /* EL2 exception link register */
370 AFSR0_EL2, /* Auxiliary Fault Status Register 0 (EL2) */
371 AFSR1_EL2, /* Auxiliary Fault Status Register 1 (EL2) */
372 ESR_EL2, /* Exception Syndrome Register (EL2) */
373 FAR_EL2, /* Fault Address Register (EL2) */
374 HPFAR_EL2, /* Hypervisor IPA Fault Address Register */
375 MAIR_EL2, /* Memory Attribute Indirection Register (EL2) */
376 AMAIR_EL2, /* Auxiliary Memory Attribute Indirection Register (EL2) */
377 VBAR_EL2, /* Vector Base Address Register (EL2) */
378 RVBAR_EL2, /* Reset Vector Base Address Register */
379 CONTEXTIDR_EL2, /* Context ID Register (EL2) */
380 TPIDR_EL2, /* EL2 Software Thread ID Register */
381 CNTHCTL_EL2, /* Counter-timer Hypervisor Control register */
382 SP_EL2, /* EL2 Stack Pointer */
388 NR_SYS_REGS /* Nothing after this line! */
391 struct kvm_cpu_context {
392 struct user_pt_regs regs; /* sp = sp_el0 */
399 struct user_fpsimd_state fp_regs;
401 u64 sys_regs[NR_SYS_REGS];
403 struct kvm_vcpu *__hyp_running_vcpu;
406 struct kvm_host_data {
407 struct kvm_cpu_context host_ctxt;
410 struct kvm_host_psci_config {
411 /* PSCI version used by host. */
414 /* Function IDs used by host if version is v0.1. */
415 struct psci_0_1_function_ids function_ids_0_1;
417 bool psci_0_1_cpu_suspend_implemented;
418 bool psci_0_1_cpu_on_implemented;
419 bool psci_0_1_cpu_off_implemented;
420 bool psci_0_1_migrate_implemented;
423 extern struct kvm_host_psci_config kvm_nvhe_sym(kvm_host_psci_config);
424 #define kvm_host_psci_config CHOOSE_NVHE_SYM(kvm_host_psci_config)
426 extern s64 kvm_nvhe_sym(hyp_physvirt_offset);
427 #define hyp_physvirt_offset CHOOSE_NVHE_SYM(hyp_physvirt_offset)
429 extern u64 kvm_nvhe_sym(hyp_cpu_logical_map)[NR_CPUS];
430 #define hyp_cpu_logical_map CHOOSE_NVHE_SYM(hyp_cpu_logical_map)
432 struct vcpu_reset_state {
439 struct kvm_vcpu_arch {
440 struct kvm_cpu_context ctxt;
443 * Guest floating point state
445 * The architecture has two main floating point extensions,
446 * the original FPSIMD and SVE. These have overlapping
447 * register views, with the FPSIMD V registers occupying the
448 * low 128 bits of the SVE Z registers. When the core
449 * floating point code saves the register state of a task it
450 * records which view it saved in fp_type.
453 enum fp_type fp_type;
454 unsigned int sve_max_vl;
457 /* Stage 2 paging state used by the hardware on next switch */
458 struct kvm_s2_mmu *hw_mmu;
460 /* Values of trap registers for the guest. */
465 /* Values of trap registers for the host before guest entry. */
468 /* Exception Information */
469 struct kvm_vcpu_fault_info fault;
471 /* Ownership of the FP regs */
475 FP_STATE_GUEST_OWNED,
478 /* Configuration flags, set once and for all before the vcpu can run */
481 /* Input flags to the hypervisor code, potentially cleared after use */
484 /* State flags for kernel bookkeeping, unused by the hypervisor code */
488 * Don't run the guest (internal implementation need).
490 * Contrary to the flags above, this is set/cleared outside of
491 * a vcpu context, and thus cannot be mixed with the flags
492 * themselves (or the flag accesses need to be made atomic).
497 * We maintain more than a single set of debug registers to support
498 * debugging the guest from the host and to maintain separate host and
499 * guest state during world switches. vcpu_debug_state are the debug
500 * registers of the vcpu as the guest sees them. host_debug_state are
501 * the host registers which are saved and restored during
502 * world switches. external_debug_state contains the debug
503 * values we want to debug the guest. This is set via the
504 * KVM_SET_GUEST_DEBUG ioctl.
506 * debug_ptr points to the set of debug registers that should be loaded
507 * onto the hardware when running the guest.
509 struct kvm_guest_debug_arch *debug_ptr;
510 struct kvm_guest_debug_arch vcpu_debug_state;
511 struct kvm_guest_debug_arch external_debug_state;
513 struct user_fpsimd_state *host_fpsimd_state; /* hyp VA */
514 struct task_struct *parent_task;
517 /* {Break,watch}point registers */
518 struct kvm_guest_debug_arch regs;
519 /* Statistical profiling extension */
521 /* Self-hosted trace */
526 struct vgic_cpu vgic_cpu;
527 struct arch_timer_cpu timer_cpu;
531 * Guest registers we preserve during guest debugging.
533 * These shadow registers are updated by the kvm_handle_sys_reg
534 * trap handler if the guest accesses or updates them while we
535 * are using guest debug.
540 } guest_debug_preserved;
542 /* vcpu power state */
543 struct kvm_mp_state mp_state;
544 spinlock_t mp_state_lock;
546 /* Cache some mmu pages needed inside spinlock regions */
547 struct kvm_mmu_memory_cache mmu_page_cache;
549 /* Target CPU and feature flags */
551 DECLARE_BITMAP(features, KVM_VCPU_MAX_FEATURES);
553 /* Virtual SError ESR to restore when HCR_EL2.VSE is set */
556 /* Additional reset state */
557 struct vcpu_reset_state reset_state;
565 /* Per-vcpu CCSIDR override or NULL */
570 * Each 'flag' is composed of a comma-separated triplet:
572 * - the flag-set it belongs to in the vcpu->arch structure
573 * - the value for that flag
574 * - the mask for that flag
576 * __vcpu_single_flag() builds such a triplet for a single-bit flag.
577 * unpack_vcpu_flag() extract the flag value from the triplet for
578 * direct use outside of the flag accessors.
580 #define __vcpu_single_flag(_set, _f) _set, (_f), (_f)
582 #define __unpack_flag(_set, _f, _m) _f
583 #define unpack_vcpu_flag(...) __unpack_flag(__VA_ARGS__)
585 #define __build_check_flag(v, flagset, f, m) \
587 typeof(v->arch.flagset) *_fset; \
589 /* Check that the flags fit in the mask */ \
590 BUILD_BUG_ON(HWEIGHT(m) != HWEIGHT((f) | (m))); \
591 /* Check that the flags fit in the type */ \
592 BUILD_BUG_ON((sizeof(*_fset) * 8) <= __fls(m)); \
595 #define __vcpu_get_flag(v, flagset, f, m) \
597 __build_check_flag(v, flagset, f, m); \
599 READ_ONCE(v->arch.flagset) & (m); \
603 * Note that the set/clear accessors must be preempt-safe in order to
604 * avoid nesting them with load/put which also manipulate flags...
606 #ifdef __KVM_NVHE_HYPERVISOR__
607 /* the nVHE hypervisor is always non-preemptible */
608 #define __vcpu_flags_preempt_disable()
609 #define __vcpu_flags_preempt_enable()
611 #define __vcpu_flags_preempt_disable() preempt_disable()
612 #define __vcpu_flags_preempt_enable() preempt_enable()
615 #define __vcpu_set_flag(v, flagset, f, m) \
617 typeof(v->arch.flagset) *fset; \
619 __build_check_flag(v, flagset, f, m); \
621 fset = &v->arch.flagset; \
622 __vcpu_flags_preempt_disable(); \
623 if (HWEIGHT(m) > 1) \
626 __vcpu_flags_preempt_enable(); \
629 #define __vcpu_clear_flag(v, flagset, f, m) \
631 typeof(v->arch.flagset) *fset; \
633 __build_check_flag(v, flagset, f, m); \
635 fset = &v->arch.flagset; \
636 __vcpu_flags_preempt_disable(); \
638 __vcpu_flags_preempt_enable(); \
641 #define vcpu_get_flag(v, ...) __vcpu_get_flag((v), __VA_ARGS__)
642 #define vcpu_set_flag(v, ...) __vcpu_set_flag((v), __VA_ARGS__)
643 #define vcpu_clear_flag(v, ...) __vcpu_clear_flag((v), __VA_ARGS__)
645 /* SVE exposed to guest */
646 #define GUEST_HAS_SVE __vcpu_single_flag(cflags, BIT(0))
647 /* SVE config completed */
648 #define VCPU_SVE_FINALIZED __vcpu_single_flag(cflags, BIT(1))
649 /* PTRAUTH exposed to guest */
650 #define GUEST_HAS_PTRAUTH __vcpu_single_flag(cflags, BIT(2))
652 /* Exception pending */
653 #define PENDING_EXCEPTION __vcpu_single_flag(iflags, BIT(0))
655 * PC increment. Overlaps with EXCEPT_MASK on purpose so that it can't
656 * be set together with an exception...
658 #define INCREMENT_PC __vcpu_single_flag(iflags, BIT(1))
659 /* Target EL/MODE (not a single flag, but let's abuse the macro) */
660 #define EXCEPT_MASK __vcpu_single_flag(iflags, GENMASK(3, 1))
662 /* Helpers to encode exceptions with minimum fuss */
663 #define __EXCEPT_MASK_VAL unpack_vcpu_flag(EXCEPT_MASK)
664 #define __EXCEPT_SHIFT __builtin_ctzl(__EXCEPT_MASK_VAL)
665 #define __vcpu_except_flags(_f) iflags, (_f << __EXCEPT_SHIFT), __EXCEPT_MASK_VAL
668 * When PENDING_EXCEPTION is set, EXCEPT_MASK can take the following
673 #define EXCEPT_AA32_UND __vcpu_except_flags(0)
674 #define EXCEPT_AA32_IABT __vcpu_except_flags(1)
675 #define EXCEPT_AA32_DABT __vcpu_except_flags(2)
677 #define EXCEPT_AA64_EL1_SYNC __vcpu_except_flags(0)
678 #define EXCEPT_AA64_EL1_IRQ __vcpu_except_flags(1)
679 #define EXCEPT_AA64_EL1_FIQ __vcpu_except_flags(2)
680 #define EXCEPT_AA64_EL1_SERR __vcpu_except_flags(3)
681 /* For AArch64 with NV: */
682 #define EXCEPT_AA64_EL2_SYNC __vcpu_except_flags(4)
683 #define EXCEPT_AA64_EL2_IRQ __vcpu_except_flags(5)
684 #define EXCEPT_AA64_EL2_FIQ __vcpu_except_flags(6)
685 #define EXCEPT_AA64_EL2_SERR __vcpu_except_flags(7)
686 /* Guest debug is live */
687 #define DEBUG_DIRTY __vcpu_single_flag(iflags, BIT(4))
688 /* Save SPE context if active */
689 #define DEBUG_STATE_SAVE_SPE __vcpu_single_flag(iflags, BIT(5))
690 /* Save TRBE context if active */
691 #define DEBUG_STATE_SAVE_TRBE __vcpu_single_flag(iflags, BIT(6))
692 /* vcpu running in HYP context */
693 #define VCPU_HYP_CONTEXT __vcpu_single_flag(iflags, BIT(7))
695 /* SVE enabled for host EL0 */
696 #define HOST_SVE_ENABLED __vcpu_single_flag(sflags, BIT(0))
697 /* SME enabled for EL0 */
698 #define HOST_SME_ENABLED __vcpu_single_flag(sflags, BIT(1))
699 /* Physical CPU not in supported_cpus */
700 #define ON_UNSUPPORTED_CPU __vcpu_single_flag(sflags, BIT(2))
701 /* WFIT instruction trapped */
702 #define IN_WFIT __vcpu_single_flag(sflags, BIT(3))
703 /* vcpu system registers loaded on physical CPU */
704 #define SYSREGS_ON_CPU __vcpu_single_flag(sflags, BIT(4))
705 /* Software step state is Active-pending */
706 #define DBG_SS_ACTIVE_PENDING __vcpu_single_flag(sflags, BIT(5))
707 /* PMUSERENR for the guest EL0 is on physical CPU */
708 #define PMUSERENR_ON_CPU __vcpu_single_flag(sflags, BIT(6))
711 /* Pointer to the vcpu's SVE FFR for sve_{save,load}_state() */
712 #define vcpu_sve_pffr(vcpu) (kern_hyp_va((vcpu)->arch.sve_state) + \
713 sve_ffr_offset((vcpu)->arch.sve_max_vl))
715 #define vcpu_sve_max_vq(vcpu) sve_vq_from_vl((vcpu)->arch.sve_max_vl)
717 #define vcpu_sve_state_size(vcpu) ({ \
719 unsigned int __vcpu_vq; \
721 if (WARN_ON(!sve_vl_valid((vcpu)->arch.sve_max_vl))) { \
724 __vcpu_vq = vcpu_sve_max_vq(vcpu); \
725 __size_ret = SVE_SIG_REGS_SIZE(__vcpu_vq); \
731 #define KVM_GUESTDBG_VALID_MASK (KVM_GUESTDBG_ENABLE | \
732 KVM_GUESTDBG_USE_SW_BP | \
733 KVM_GUESTDBG_USE_HW | \
734 KVM_GUESTDBG_SINGLESTEP)
736 #define vcpu_has_sve(vcpu) (system_supports_sve() && \
737 vcpu_get_flag(vcpu, GUEST_HAS_SVE))
739 #ifdef CONFIG_ARM64_PTR_AUTH
740 #define vcpu_has_ptrauth(vcpu) \
741 ((cpus_have_final_cap(ARM64_HAS_ADDRESS_AUTH) || \
742 cpus_have_final_cap(ARM64_HAS_GENERIC_AUTH)) && \
743 vcpu_get_flag(vcpu, GUEST_HAS_PTRAUTH))
745 #define vcpu_has_ptrauth(vcpu) false
748 #define vcpu_on_unsupported_cpu(vcpu) \
749 vcpu_get_flag(vcpu, ON_UNSUPPORTED_CPU)
751 #define vcpu_set_on_unsupported_cpu(vcpu) \
752 vcpu_set_flag(vcpu, ON_UNSUPPORTED_CPU)
754 #define vcpu_clear_on_unsupported_cpu(vcpu) \
755 vcpu_clear_flag(vcpu, ON_UNSUPPORTED_CPU)
757 #define vcpu_gp_regs(v) (&(v)->arch.ctxt.regs)
760 * Only use __vcpu_sys_reg/ctxt_sys_reg if you know you want the
761 * memory backed version of a register, and not the one most recently
762 * accessed by a running VCPU. For example, for userspace access or
763 * for system registers that are never context switched, but only
766 #define __ctxt_sys_reg(c,r) (&(c)->sys_regs[(r)])
768 #define ctxt_sys_reg(c,r) (*__ctxt_sys_reg(c,r))
770 #define __vcpu_sys_reg(v,r) (ctxt_sys_reg(&(v)->arch.ctxt, (r)))
772 u64 vcpu_read_sys_reg(const struct kvm_vcpu *vcpu, int reg);
773 void vcpu_write_sys_reg(struct kvm_vcpu *vcpu, u64 val, int reg);
775 static inline bool __vcpu_read_sys_reg_from_cpu(int reg, u64 *val)
780 * System registers listed in the switch are not saved on every
781 * exit from the guest but are only saved on vcpu_put.
783 * Note that MPIDR_EL1 for the guest is set by KVM via VMPIDR_EL2 but
784 * should never be listed below, because the guest cannot modify its
785 * own MPIDR_EL1 and MPIDR_EL1 is accessed for VCPU A from VCPU B's
786 * thread when emulating cross-VCPU communication.
792 case SCTLR_EL1: *val = read_sysreg_s(SYS_SCTLR_EL12); break;
793 case CPACR_EL1: *val = read_sysreg_s(SYS_CPACR_EL12); break;
794 case TTBR0_EL1: *val = read_sysreg_s(SYS_TTBR0_EL12); break;
795 case TTBR1_EL1: *val = read_sysreg_s(SYS_TTBR1_EL12); break;
796 case TCR_EL1: *val = read_sysreg_s(SYS_TCR_EL12); break;
797 case ESR_EL1: *val = read_sysreg_s(SYS_ESR_EL12); break;
798 case AFSR0_EL1: *val = read_sysreg_s(SYS_AFSR0_EL12); break;
799 case AFSR1_EL1: *val = read_sysreg_s(SYS_AFSR1_EL12); break;
800 case FAR_EL1: *val = read_sysreg_s(SYS_FAR_EL12); break;
801 case MAIR_EL1: *val = read_sysreg_s(SYS_MAIR_EL12); break;
802 case VBAR_EL1: *val = read_sysreg_s(SYS_VBAR_EL12); break;
803 case CONTEXTIDR_EL1: *val = read_sysreg_s(SYS_CONTEXTIDR_EL12);break;
804 case TPIDR_EL0: *val = read_sysreg_s(SYS_TPIDR_EL0); break;
805 case TPIDRRO_EL0: *val = read_sysreg_s(SYS_TPIDRRO_EL0); break;
806 case TPIDR_EL1: *val = read_sysreg_s(SYS_TPIDR_EL1); break;
807 case AMAIR_EL1: *val = read_sysreg_s(SYS_AMAIR_EL12); break;
808 case CNTKCTL_EL1: *val = read_sysreg_s(SYS_CNTKCTL_EL12); break;
809 case ELR_EL1: *val = read_sysreg_s(SYS_ELR_EL12); break;
810 case PAR_EL1: *val = read_sysreg_par(); break;
811 case DACR32_EL2: *val = read_sysreg_s(SYS_DACR32_EL2); break;
812 case IFSR32_EL2: *val = read_sysreg_s(SYS_IFSR32_EL2); break;
813 case DBGVCR32_EL2: *val = read_sysreg_s(SYS_DBGVCR32_EL2); break;
814 default: return false;
820 static inline bool __vcpu_write_sys_reg_to_cpu(u64 val, int reg)
825 * System registers listed in the switch are not restored on every
826 * entry to the guest but are only restored on vcpu_load.
828 * Note that MPIDR_EL1 for the guest is set by KVM via VMPIDR_EL2 but
829 * should never be listed below, because the MPIDR should only be set
830 * once, before running the VCPU, and never changed later.
836 case SCTLR_EL1: write_sysreg_s(val, SYS_SCTLR_EL12); break;
837 case CPACR_EL1: write_sysreg_s(val, SYS_CPACR_EL12); break;
838 case TTBR0_EL1: write_sysreg_s(val, SYS_TTBR0_EL12); break;
839 case TTBR1_EL1: write_sysreg_s(val, SYS_TTBR1_EL12); break;
840 case TCR_EL1: write_sysreg_s(val, SYS_TCR_EL12); break;
841 case ESR_EL1: write_sysreg_s(val, SYS_ESR_EL12); break;
842 case AFSR0_EL1: write_sysreg_s(val, SYS_AFSR0_EL12); break;
843 case AFSR1_EL1: write_sysreg_s(val, SYS_AFSR1_EL12); break;
844 case FAR_EL1: write_sysreg_s(val, SYS_FAR_EL12); break;
845 case MAIR_EL1: write_sysreg_s(val, SYS_MAIR_EL12); break;
846 case VBAR_EL1: write_sysreg_s(val, SYS_VBAR_EL12); break;
847 case CONTEXTIDR_EL1: write_sysreg_s(val, SYS_CONTEXTIDR_EL12);break;
848 case TPIDR_EL0: write_sysreg_s(val, SYS_TPIDR_EL0); break;
849 case TPIDRRO_EL0: write_sysreg_s(val, SYS_TPIDRRO_EL0); break;
850 case TPIDR_EL1: write_sysreg_s(val, SYS_TPIDR_EL1); break;
851 case AMAIR_EL1: write_sysreg_s(val, SYS_AMAIR_EL12); break;
852 case CNTKCTL_EL1: write_sysreg_s(val, SYS_CNTKCTL_EL12); break;
853 case ELR_EL1: write_sysreg_s(val, SYS_ELR_EL12); break;
854 case PAR_EL1: write_sysreg_s(val, SYS_PAR_EL1); break;
855 case DACR32_EL2: write_sysreg_s(val, SYS_DACR32_EL2); break;
856 case IFSR32_EL2: write_sysreg_s(val, SYS_IFSR32_EL2); break;
857 case DBGVCR32_EL2: write_sysreg_s(val, SYS_DBGVCR32_EL2); break;
858 default: return false;
865 struct kvm_vm_stat_generic generic;
868 struct kvm_vcpu_stat {
869 struct kvm_vcpu_stat_generic generic;
874 u64 mmio_exit_kernel;
879 void kvm_vcpu_preferred_target(struct kvm_vcpu_init *init);
880 unsigned long kvm_arm_num_regs(struct kvm_vcpu *vcpu);
881 int kvm_arm_copy_reg_indices(struct kvm_vcpu *vcpu, u64 __user *indices);
882 int kvm_arm_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
883 int kvm_arm_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
885 unsigned long kvm_arm_num_sys_reg_descs(struct kvm_vcpu *vcpu);
886 int kvm_arm_copy_sys_reg_indices(struct kvm_vcpu *vcpu, u64 __user *uindices);
888 int __kvm_arm_vcpu_get_events(struct kvm_vcpu *vcpu,
889 struct kvm_vcpu_events *events);
891 int __kvm_arm_vcpu_set_events(struct kvm_vcpu *vcpu,
892 struct kvm_vcpu_events *events);
894 #define KVM_ARCH_WANT_MMU_NOTIFIER
896 void kvm_arm_halt_guest(struct kvm *kvm);
897 void kvm_arm_resume_guest(struct kvm *kvm);
899 #define vcpu_has_run_once(vcpu) !!rcu_access_pointer((vcpu)->pid)
901 #ifndef __KVM_NVHE_HYPERVISOR__
902 #define kvm_call_hyp_nvhe(f, ...) \
904 struct arm_smccc_res res; \
906 arm_smccc_1_1_hvc(KVM_HOST_SMCCC_FUNC(f), \
907 ##__VA_ARGS__, &res); \
908 WARN_ON(res.a0 != SMCCC_RET_SUCCESS); \
914 * The couple of isb() below are there to guarantee the same behaviour
915 * on VHE as on !VHE, where the eret to EL1 acts as a context
916 * synchronization event.
918 #define kvm_call_hyp(f, ...) \
924 kvm_call_hyp_nvhe(f, ##__VA_ARGS__); \
928 #define kvm_call_hyp_ret(f, ...) \
930 typeof(f(__VA_ARGS__)) ret; \
933 ret = f(__VA_ARGS__); \
936 ret = kvm_call_hyp_nvhe(f, ##__VA_ARGS__); \
941 #else /* __KVM_NVHE_HYPERVISOR__ */
942 #define kvm_call_hyp(f, ...) f(__VA_ARGS__)
943 #define kvm_call_hyp_ret(f, ...) f(__VA_ARGS__)
944 #define kvm_call_hyp_nvhe(f, ...) f(__VA_ARGS__)
945 #endif /* __KVM_NVHE_HYPERVISOR__ */
947 void force_vm_exit(const cpumask_t *mask);
949 int handle_exit(struct kvm_vcpu *vcpu, int exception_index);
950 void handle_exit_early(struct kvm_vcpu *vcpu, int exception_index);
952 int kvm_handle_cp14_load_store(struct kvm_vcpu *vcpu);
953 int kvm_handle_cp14_32(struct kvm_vcpu *vcpu);
954 int kvm_handle_cp14_64(struct kvm_vcpu *vcpu);
955 int kvm_handle_cp15_32(struct kvm_vcpu *vcpu);
956 int kvm_handle_cp15_64(struct kvm_vcpu *vcpu);
957 int kvm_handle_sys_reg(struct kvm_vcpu *vcpu);
958 int kvm_handle_cp10_id(struct kvm_vcpu *vcpu);
960 void kvm_reset_sys_regs(struct kvm_vcpu *vcpu);
962 int __init kvm_sys_reg_table_init(void);
964 bool lock_all_vcpus(struct kvm *kvm);
965 void unlock_all_vcpus(struct kvm *kvm);
968 void kvm_mmio_write_buf(void *buf, unsigned int len, unsigned long data);
969 unsigned long kvm_mmio_read_buf(const void *buf, unsigned int len);
971 int kvm_handle_mmio_return(struct kvm_vcpu *vcpu);
972 int io_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa);
975 * Returns true if a Performance Monitoring Interrupt (PMI), a.k.a. perf event,
976 * arrived in guest context. For arm64, any event that arrives while a vCPU is
977 * loaded is considered to be "in guest".
979 static inline bool kvm_arch_pmi_in_guest(struct kvm_vcpu *vcpu)
981 return IS_ENABLED(CONFIG_GUEST_PERF_EVENTS) && !!vcpu;
984 long kvm_hypercall_pv_features(struct kvm_vcpu *vcpu);
985 gpa_t kvm_init_stolen_time(struct kvm_vcpu *vcpu);
986 void kvm_update_stolen_time(struct kvm_vcpu *vcpu);
988 bool kvm_arm_pvtime_supported(void);
989 int kvm_arm_pvtime_set_attr(struct kvm_vcpu *vcpu,
990 struct kvm_device_attr *attr);
991 int kvm_arm_pvtime_get_attr(struct kvm_vcpu *vcpu,
992 struct kvm_device_attr *attr);
993 int kvm_arm_pvtime_has_attr(struct kvm_vcpu *vcpu,
994 struct kvm_device_attr *attr);
996 extern unsigned int __ro_after_init kvm_arm_vmid_bits;
997 int __init kvm_arm_vmid_alloc_init(void);
998 void __init kvm_arm_vmid_alloc_free(void);
999 void kvm_arm_vmid_update(struct kvm_vmid *kvm_vmid);
1000 void kvm_arm_vmid_clear_active(void);
1002 static inline void kvm_arm_pvtime_vcpu_init(struct kvm_vcpu_arch *vcpu_arch)
1004 vcpu_arch->steal.base = INVALID_GPA;
1007 static inline bool kvm_arm_is_pvtime_enabled(struct kvm_vcpu_arch *vcpu_arch)
1009 return (vcpu_arch->steal.base != INVALID_GPA);
1012 void kvm_set_sei_esr(struct kvm_vcpu *vcpu, u64 syndrome);
1014 struct kvm_vcpu *kvm_mpidr_to_vcpu(struct kvm *kvm, unsigned long mpidr);
1016 DECLARE_KVM_HYP_PER_CPU(struct kvm_host_data, kvm_host_data);
1018 static inline void kvm_init_host_cpu_context(struct kvm_cpu_context *cpu_ctxt)
1020 /* The host's MPIDR is immutable, so let's set it up at boot time */
1021 ctxt_sys_reg(cpu_ctxt, MPIDR_EL1) = read_cpuid_mpidr();
1024 static inline bool kvm_system_needs_idmapped_vectors(void)
1026 return cpus_have_const_cap(ARM64_SPECTRE_V3A);
1029 void kvm_arm_vcpu_ptrauth_trap(struct kvm_vcpu *vcpu);
1031 static inline void kvm_arch_sync_events(struct kvm *kvm) {}
1032 static inline void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu) {}
1034 void kvm_arm_init_debug(void);
1035 void kvm_arm_vcpu_init_debug(struct kvm_vcpu *vcpu);
1036 void kvm_arm_setup_debug(struct kvm_vcpu *vcpu);
1037 void kvm_arm_clear_debug(struct kvm_vcpu *vcpu);
1038 void kvm_arm_reset_debug_ptr(struct kvm_vcpu *vcpu);
1040 #define kvm_vcpu_os_lock_enabled(vcpu) \
1041 (!!(__vcpu_sys_reg(vcpu, OSLSR_EL1) & OSLSR_EL1_OSLK))
1043 int kvm_arm_vcpu_arch_set_attr(struct kvm_vcpu *vcpu,
1044 struct kvm_device_attr *attr);
1045 int kvm_arm_vcpu_arch_get_attr(struct kvm_vcpu *vcpu,
1046 struct kvm_device_attr *attr);
1047 int kvm_arm_vcpu_arch_has_attr(struct kvm_vcpu *vcpu,
1048 struct kvm_device_attr *attr);
1050 int kvm_vm_ioctl_mte_copy_tags(struct kvm *kvm,
1051 struct kvm_arm_copy_mte_tags *copy_tags);
1052 int kvm_vm_ioctl_set_counter_offset(struct kvm *kvm,
1053 struct kvm_arm_counter_offset *offset);
1055 /* Guest/host FPSIMD coordination helpers */
1056 int kvm_arch_vcpu_run_map_fp(struct kvm_vcpu *vcpu);
1057 void kvm_arch_vcpu_load_fp(struct kvm_vcpu *vcpu);
1058 void kvm_arch_vcpu_ctxflush_fp(struct kvm_vcpu *vcpu);
1059 void kvm_arch_vcpu_ctxsync_fp(struct kvm_vcpu *vcpu);
1060 void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu);
1061 void kvm_vcpu_unshare_task_fp(struct kvm_vcpu *vcpu);
1063 static inline bool kvm_pmu_counter_deferred(struct perf_event_attr *attr)
1065 return (!has_vhe() && attr->exclude_host);
1068 /* Flags for host debug state */
1069 void kvm_arch_vcpu_load_debug_state_flags(struct kvm_vcpu *vcpu);
1070 void kvm_arch_vcpu_put_debug_state_flags(struct kvm_vcpu *vcpu);
1073 void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr);
1074 void kvm_clr_pmu_events(u32 clr);
1075 bool kvm_set_pmuserenr(u64 val);
1077 static inline void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr) {}
1078 static inline void kvm_clr_pmu_events(u32 clr) {}
1079 static inline bool kvm_set_pmuserenr(u64 val)
1085 void kvm_vcpu_load_sysregs_vhe(struct kvm_vcpu *vcpu);
1086 void kvm_vcpu_put_sysregs_vhe(struct kvm_vcpu *vcpu);
1088 int __init kvm_set_ipa_limit(void);
1090 #define __KVM_HAVE_ARCH_VM_ALLOC
1091 struct kvm *kvm_arch_alloc_vm(void);
1093 static inline bool kvm_vm_is_protected(struct kvm *kvm)
1098 void kvm_init_protected_traps(struct kvm_vcpu *vcpu);
1100 int kvm_arm_vcpu_finalize(struct kvm_vcpu *vcpu, int feature);
1101 bool kvm_arm_vcpu_is_finalized(struct kvm_vcpu *vcpu);
1103 #define kvm_arm_vcpu_sve_finalized(vcpu) vcpu_get_flag(vcpu, VCPU_SVE_FINALIZED)
1105 #define kvm_has_mte(kvm) \
1106 (system_supports_mte() && \
1107 test_bit(KVM_ARCH_FLAG_MTE_ENABLED, &(kvm)->arch.flags))
1109 #define kvm_supports_32bit_el0() \
1110 (system_supports_32bit_el0() && \
1111 !static_branch_unlikely(&arm64_mismatched_32bit_el0))
1113 #define kvm_vm_has_ran_once(kvm) \
1114 (test_bit(KVM_ARCH_FLAG_HAS_RAN_ONCE, &(kvm)->arch.flags))
1116 int kvm_trng_call(struct kvm_vcpu *vcpu);
1118 extern phys_addr_t hyp_mem_base;
1119 extern phys_addr_t hyp_mem_size;
1120 void __init kvm_hyp_reserve(void);
1122 static inline void kvm_hyp_reserve(void) { }
1125 void kvm_arm_vcpu_power_off(struct kvm_vcpu *vcpu);
1126 bool kvm_arm_vcpu_stopped(struct kvm_vcpu *vcpu);
1128 #endif /* __ARM64_KVM_HOST_H__ */