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
42 #define KVM_VCPU_VALID_FEATURES (BIT(KVM_VCPU_MAX_FEATURES) - 1)
44 #define KVM_REQ_SLEEP \
45 KVM_ARCH_REQ_FLAGS(0, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
46 #define KVM_REQ_IRQ_PENDING KVM_ARCH_REQ(1)
47 #define KVM_REQ_VCPU_RESET KVM_ARCH_REQ(2)
48 #define KVM_REQ_RECORD_STEAL KVM_ARCH_REQ(3)
49 #define KVM_REQ_RELOAD_GICv4 KVM_ARCH_REQ(4)
50 #define KVM_REQ_RELOAD_PMU KVM_ARCH_REQ(5)
51 #define KVM_REQ_SUSPEND KVM_ARCH_REQ(6)
53 #define KVM_DIRTY_LOG_MANUAL_CAPS (KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE | \
54 KVM_DIRTY_LOG_INITIALLY_SET)
56 #define KVM_HAVE_MMU_RWLOCK
59 * Mode of operation configurable with kvm-arm.mode early param.
60 * See Documentation/admin-guide/kernel-parameters.txt for more information.
69 enum kvm_mode kvm_get_mode(void);
71 static inline enum kvm_mode kvm_get_mode(void) { return KVM_MODE_NONE; };
74 DECLARE_STATIC_KEY_FALSE(userspace_irqchip_in_use);
76 extern unsigned int __ro_after_init kvm_sve_max_vl;
77 int __init kvm_arm_init_sve(void);
79 u32 __attribute_const__ kvm_target_cpu(void);
80 int kvm_reset_vcpu(struct kvm_vcpu *vcpu);
81 void kvm_arm_vcpu_destroy(struct kvm_vcpu *vcpu);
83 struct kvm_hyp_memcache {
85 unsigned long nr_pages;
88 static inline void push_hyp_memcache(struct kvm_hyp_memcache *mc,
90 phys_addr_t (*to_pa)(void *virt))
97 static inline void *pop_hyp_memcache(struct kvm_hyp_memcache *mc,
98 void *(*to_va)(phys_addr_t phys))
100 phys_addr_t *p = to_va(mc->head);
111 static inline int __topup_hyp_memcache(struct kvm_hyp_memcache *mc,
112 unsigned long min_pages,
113 void *(*alloc_fn)(void *arg),
114 phys_addr_t (*to_pa)(void *virt),
117 while (mc->nr_pages < min_pages) {
118 phys_addr_t *p = alloc_fn(arg);
122 push_hyp_memcache(mc, p, to_pa);
128 static inline void __free_hyp_memcache(struct kvm_hyp_memcache *mc,
129 void (*free_fn)(void *virt, void *arg),
130 void *(*to_va)(phys_addr_t phys),
134 free_fn(pop_hyp_memcache(mc, to_va), arg);
137 void free_hyp_memcache(struct kvm_hyp_memcache *mc);
138 int topup_hyp_memcache(struct kvm_hyp_memcache *mc, unsigned long min_pages);
145 struct kvm_vmid vmid;
148 * stage2 entry level table
150 * Two kvm_s2_mmu structures in the same VM can point to the same
151 * pgd here. This happens when running a guest using a
152 * translation regime that isn't affected by its own stage-2
153 * translation, such as a non-VHE hypervisor running at vEL2, or
154 * for vEL1/EL0 with vHCR_EL2.VM == 0. In that case, we use the
155 * canonical stage-2 page tables.
157 phys_addr_t pgd_phys;
158 struct kvm_pgtable *pgt;
160 /* The last vcpu id that ran on each physical CPU */
161 int __percpu *last_vcpu_ran;
163 struct kvm_arch *arch;
166 struct kvm_arch_memory_slot {
170 * struct kvm_smccc_features: Descriptor of the hypercall services exposed to the guests
172 * @std_bmap: Bitmap of standard secure service calls
173 * @std_hyp_bmap: Bitmap of standard hypervisor service calls
174 * @vendor_hyp_bmap: Bitmap of vendor specific hypervisor service calls
176 struct kvm_smccc_features {
177 unsigned long std_bmap;
178 unsigned long std_hyp_bmap;
179 unsigned long vendor_hyp_bmap;
182 typedef unsigned int pkvm_handle_t;
184 struct kvm_protected_vm {
185 pkvm_handle_t handle;
186 struct kvm_hyp_memcache teardown_mc;
190 struct kvm_s2_mmu mmu;
192 /* VTCR_EL2 value for this VM */
195 /* Interrupt controller */
196 struct vgic_dist vgic;
199 struct arch_timer_vm_data timer_data;
201 /* Mandated version of PSCI */
204 /* Protects VM-scoped configuration data */
205 struct mutex config_lock;
208 * If we encounter a data abort without valid instruction syndrome
209 * information, report this to user space. User space can (and
210 * should) opt in to this feature if KVM_CAP_ARM_NISV_TO_USER is
213 #define KVM_ARCH_FLAG_RETURN_NISV_IO_ABORT_TO_USER 0
214 /* Memory Tagging Extension enabled for the guest */
215 #define KVM_ARCH_FLAG_MTE_ENABLED 1
216 /* At least one vCPU has ran in the VM */
217 #define KVM_ARCH_FLAG_HAS_RAN_ONCE 2
218 /* The vCPU feature set for the VM is configured */
219 #define KVM_ARCH_FLAG_VCPU_FEATURES_CONFIGURED 3
220 /* PSCI SYSTEM_SUSPEND enabled for the guest */
221 #define KVM_ARCH_FLAG_SYSTEM_SUSPEND_ENABLED 4
222 /* VM counter offset */
223 #define KVM_ARCH_FLAG_VM_COUNTER_OFFSET 5
224 /* Timer PPIs made immutable */
225 #define KVM_ARCH_FLAG_TIMER_PPIS_IMMUTABLE 6
226 /* SMCCC filter initialized for the VM */
227 #define KVM_ARCH_FLAG_SMCCC_FILTER_CONFIGURED 7
230 /* VM-wide vCPU feature set */
231 DECLARE_BITMAP(vcpu_features, KVM_VCPU_MAX_FEATURES);
234 * VM-wide PMU filter, implemented as a bitmap and big enough for
235 * up to 2^10 events (ARMv8.0) or 2^16 events (ARMv8.1+).
237 unsigned long *pmu_filter;
238 struct arm_pmu *arm_pmu;
240 cpumask_var_t supported_cpus;
249 /* Hypercall features firmware registers' descriptor */
250 struct kvm_smccc_features smccc_feat;
251 struct maple_tree smccc_filter;
254 * For an untrusted host VM, 'pkvm.handle' is used to lookup
255 * the associated pKVM instance in the hypervisor.
257 struct kvm_protected_vm pkvm;
260 struct kvm_vcpu_fault_info {
261 u64 esr_el2; /* Hyp Syndrom Register */
262 u64 far_el2; /* Hyp Fault Address Register */
263 u64 hpfar_el2; /* Hyp IPA Fault Address Register */
264 u64 disr_el1; /* Deferred [SError] Status Register */
268 __INVALID_SYSREG__, /* 0 is reserved as an invalid value */
269 MPIDR_EL1, /* MultiProcessor Affinity Register */
270 CLIDR_EL1, /* Cache Level ID Register */
271 CSSELR_EL1, /* Cache Size Selection Register */
272 SCTLR_EL1, /* System Control Register */
273 ACTLR_EL1, /* Auxiliary Control Register */
274 CPACR_EL1, /* Coprocessor Access Control */
275 ZCR_EL1, /* SVE Control */
276 TTBR0_EL1, /* Translation Table Base Register 0 */
277 TTBR1_EL1, /* Translation Table Base Register 1 */
278 TCR_EL1, /* Translation Control Register */
279 ESR_EL1, /* Exception Syndrome Register */
280 AFSR0_EL1, /* Auxiliary Fault Status Register 0 */
281 AFSR1_EL1, /* Auxiliary Fault Status Register 1 */
282 FAR_EL1, /* Fault Address Register */
283 MAIR_EL1, /* Memory Attribute Indirection Register */
284 VBAR_EL1, /* Vector Base Address Register */
285 CONTEXTIDR_EL1, /* Context ID Register */
286 TPIDR_EL0, /* Thread ID, User R/W */
287 TPIDRRO_EL0, /* Thread ID, User R/O */
288 TPIDR_EL1, /* Thread ID, Privileged */
289 AMAIR_EL1, /* Aux Memory Attribute Indirection Register */
290 CNTKCTL_EL1, /* Timer Control Register (EL1) */
291 PAR_EL1, /* Physical Address Register */
292 MDSCR_EL1, /* Monitor Debug System Control Register */
293 MDCCINT_EL1, /* Monitor Debug Comms Channel Interrupt Enable Reg */
294 OSLSR_EL1, /* OS Lock Status Register */
295 DISR_EL1, /* Deferred Interrupt Status Register */
297 /* Performance Monitors Registers */
298 PMCR_EL0, /* Control Register */
299 PMSELR_EL0, /* Event Counter Selection Register */
300 PMEVCNTR0_EL0, /* Event Counter Register (0-30) */
301 PMEVCNTR30_EL0 = PMEVCNTR0_EL0 + 30,
302 PMCCNTR_EL0, /* Cycle Counter Register */
303 PMEVTYPER0_EL0, /* Event Type Register (0-30) */
304 PMEVTYPER30_EL0 = PMEVTYPER0_EL0 + 30,
305 PMCCFILTR_EL0, /* Cycle Count Filter Register */
306 PMCNTENSET_EL0, /* Count Enable Set Register */
307 PMINTENSET_EL1, /* Interrupt Enable Set Register */
308 PMOVSSET_EL0, /* Overflow Flag Status Set Register */
309 PMUSERENR_EL0, /* User Enable Register */
311 /* Pointer Authentication Registers in a strict increasing order. */
333 /* Memory Tagging Extension registers */
334 RGSR_EL1, /* Random Allocation Tag Seed Register */
335 GCR_EL1, /* Tag Control Register */
336 TFSR_EL1, /* Tag Fault Status Register (EL1) */
337 TFSRE0_EL1, /* Tag Fault Status Register (EL0) */
339 /* 32bit specific registers. */
340 DACR32_EL2, /* Domain Access Control Register */
341 IFSR32_EL2, /* Instruction Fault Status Register */
342 FPEXC32_EL2, /* Floating-Point Exception Control Register */
343 DBGVCR32_EL2, /* Debug Vector Catch Register */
346 VPIDR_EL2, /* Virtualization Processor ID Register */
347 VMPIDR_EL2, /* Virtualization Multiprocessor ID Register */
348 SCTLR_EL2, /* System Control Register (EL2) */
349 ACTLR_EL2, /* Auxiliary Control Register (EL2) */
350 HCR_EL2, /* Hypervisor Configuration Register */
351 MDCR_EL2, /* Monitor Debug Configuration Register (EL2) */
352 CPTR_EL2, /* Architectural Feature Trap Register (EL2) */
353 HSTR_EL2, /* Hypervisor System Trap Register */
354 HACR_EL2, /* Hypervisor Auxiliary Control Register */
355 TTBR0_EL2, /* Translation Table Base Register 0 (EL2) */
356 TTBR1_EL2, /* Translation Table Base Register 1 (EL2) */
357 TCR_EL2, /* Translation Control Register (EL2) */
358 VTTBR_EL2, /* Virtualization Translation Table Base Register */
359 VTCR_EL2, /* Virtualization Translation Control Register */
360 SPSR_EL2, /* EL2 saved program status register */
361 ELR_EL2, /* EL2 exception link register */
362 AFSR0_EL2, /* Auxiliary Fault Status Register 0 (EL2) */
363 AFSR1_EL2, /* Auxiliary Fault Status Register 1 (EL2) */
364 ESR_EL2, /* Exception Syndrome Register (EL2) */
365 FAR_EL2, /* Fault Address Register (EL2) */
366 HPFAR_EL2, /* Hypervisor IPA Fault Address Register */
367 MAIR_EL2, /* Memory Attribute Indirection Register (EL2) */
368 AMAIR_EL2, /* Auxiliary Memory Attribute Indirection Register (EL2) */
369 VBAR_EL2, /* Vector Base Address Register (EL2) */
370 RVBAR_EL2, /* Reset Vector Base Address Register */
371 CONTEXTIDR_EL2, /* Context ID Register (EL2) */
372 TPIDR_EL2, /* EL2 Software Thread ID Register */
373 CNTHCTL_EL2, /* Counter-timer Hypervisor Control register */
374 SP_EL2, /* EL2 Stack Pointer */
380 NR_SYS_REGS /* Nothing after this line! */
383 struct kvm_cpu_context {
384 struct user_pt_regs regs; /* sp = sp_el0 */
391 struct user_fpsimd_state fp_regs;
393 u64 sys_regs[NR_SYS_REGS];
395 struct kvm_vcpu *__hyp_running_vcpu;
398 struct kvm_host_data {
399 struct kvm_cpu_context host_ctxt;
402 struct kvm_host_psci_config {
403 /* PSCI version used by host. */
406 /* Function IDs used by host if version is v0.1. */
407 struct psci_0_1_function_ids function_ids_0_1;
409 bool psci_0_1_cpu_suspend_implemented;
410 bool psci_0_1_cpu_on_implemented;
411 bool psci_0_1_cpu_off_implemented;
412 bool psci_0_1_migrate_implemented;
415 extern struct kvm_host_psci_config kvm_nvhe_sym(kvm_host_psci_config);
416 #define kvm_host_psci_config CHOOSE_NVHE_SYM(kvm_host_psci_config)
418 extern s64 kvm_nvhe_sym(hyp_physvirt_offset);
419 #define hyp_physvirt_offset CHOOSE_NVHE_SYM(hyp_physvirt_offset)
421 extern u64 kvm_nvhe_sym(hyp_cpu_logical_map)[NR_CPUS];
422 #define hyp_cpu_logical_map CHOOSE_NVHE_SYM(hyp_cpu_logical_map)
424 struct vcpu_reset_state {
431 struct kvm_vcpu_arch {
432 struct kvm_cpu_context ctxt;
435 * Guest floating point state
437 * The architecture has two main floating point extensions,
438 * the original FPSIMD and SVE. These have overlapping
439 * register views, with the FPSIMD V registers occupying the
440 * low 128 bits of the SVE Z registers. When the core
441 * floating point code saves the register state of a task it
442 * records which view it saved in fp_type.
445 enum fp_type fp_type;
446 unsigned int sve_max_vl;
449 /* Stage 2 paging state used by the hardware on next switch */
450 struct kvm_s2_mmu *hw_mmu;
452 /* Values of trap registers for the guest. */
457 /* Values of trap registers for the host before guest entry. */
460 /* Exception Information */
461 struct kvm_vcpu_fault_info fault;
463 /* Ownership of the FP regs */
467 FP_STATE_GUEST_OWNED,
470 /* Configuration flags, set once and for all before the vcpu can run */
473 /* Input flags to the hypervisor code, potentially cleared after use */
476 /* State flags for kernel bookkeeping, unused by the hypervisor code */
480 * Don't run the guest (internal implementation need).
482 * Contrary to the flags above, this is set/cleared outside of
483 * a vcpu context, and thus cannot be mixed with the flags
484 * themselves (or the flag accesses need to be made atomic).
489 * We maintain more than a single set of debug registers to support
490 * debugging the guest from the host and to maintain separate host and
491 * guest state during world switches. vcpu_debug_state are the debug
492 * registers of the vcpu as the guest sees them. host_debug_state are
493 * the host registers which are saved and restored during
494 * world switches. external_debug_state contains the debug
495 * values we want to debug the guest. This is set via the
496 * KVM_SET_GUEST_DEBUG ioctl.
498 * debug_ptr points to the set of debug registers that should be loaded
499 * onto the hardware when running the guest.
501 struct kvm_guest_debug_arch *debug_ptr;
502 struct kvm_guest_debug_arch vcpu_debug_state;
503 struct kvm_guest_debug_arch external_debug_state;
505 struct user_fpsimd_state *host_fpsimd_state; /* hyp VA */
506 struct task_struct *parent_task;
509 /* {Break,watch}point registers */
510 struct kvm_guest_debug_arch regs;
511 /* Statistical profiling extension */
513 /* Self-hosted trace */
518 struct vgic_cpu vgic_cpu;
519 struct arch_timer_cpu timer_cpu;
523 * Guest registers we preserve during guest debugging.
525 * These shadow registers are updated by the kvm_handle_sys_reg
526 * trap handler if the guest accesses or updates them while we
527 * are using guest debug.
532 } guest_debug_preserved;
534 /* vcpu power state */
535 struct kvm_mp_state mp_state;
536 spinlock_t mp_state_lock;
538 /* Cache some mmu pages needed inside spinlock regions */
539 struct kvm_mmu_memory_cache mmu_page_cache;
541 /* Target CPU and feature flags */
543 DECLARE_BITMAP(features, KVM_VCPU_MAX_FEATURES);
545 /* Virtual SError ESR to restore when HCR_EL2.VSE is set */
548 /* Additional reset state */
549 struct vcpu_reset_state reset_state;
557 /* Per-vcpu CCSIDR override or NULL */
562 * Each 'flag' is composed of a comma-separated triplet:
564 * - the flag-set it belongs to in the vcpu->arch structure
565 * - the value for that flag
566 * - the mask for that flag
568 * __vcpu_single_flag() builds such a triplet for a single-bit flag.
569 * unpack_vcpu_flag() extract the flag value from the triplet for
570 * direct use outside of the flag accessors.
572 #define __vcpu_single_flag(_set, _f) _set, (_f), (_f)
574 #define __unpack_flag(_set, _f, _m) _f
575 #define unpack_vcpu_flag(...) __unpack_flag(__VA_ARGS__)
577 #define __build_check_flag(v, flagset, f, m) \
579 typeof(v->arch.flagset) *_fset; \
581 /* Check that the flags fit in the mask */ \
582 BUILD_BUG_ON(HWEIGHT(m) != HWEIGHT((f) | (m))); \
583 /* Check that the flags fit in the type */ \
584 BUILD_BUG_ON((sizeof(*_fset) * 8) <= __fls(m)); \
587 #define __vcpu_get_flag(v, flagset, f, m) \
589 __build_check_flag(v, flagset, f, m); \
591 READ_ONCE(v->arch.flagset) & (m); \
595 * Note that the set/clear accessors must be preempt-safe in order to
596 * avoid nesting them with load/put which also manipulate flags...
598 #ifdef __KVM_NVHE_HYPERVISOR__
599 /* the nVHE hypervisor is always non-preemptible */
600 #define __vcpu_flags_preempt_disable()
601 #define __vcpu_flags_preempt_enable()
603 #define __vcpu_flags_preempt_disable() preempt_disable()
604 #define __vcpu_flags_preempt_enable() preempt_enable()
607 #define __vcpu_set_flag(v, flagset, f, m) \
609 typeof(v->arch.flagset) *fset; \
611 __build_check_flag(v, flagset, f, m); \
613 fset = &v->arch.flagset; \
614 __vcpu_flags_preempt_disable(); \
615 if (HWEIGHT(m) > 1) \
618 __vcpu_flags_preempt_enable(); \
621 #define __vcpu_clear_flag(v, flagset, f, m) \
623 typeof(v->arch.flagset) *fset; \
625 __build_check_flag(v, flagset, f, m); \
627 fset = &v->arch.flagset; \
628 __vcpu_flags_preempt_disable(); \
630 __vcpu_flags_preempt_enable(); \
633 #define vcpu_get_flag(v, ...) __vcpu_get_flag((v), __VA_ARGS__)
634 #define vcpu_set_flag(v, ...) __vcpu_set_flag((v), __VA_ARGS__)
635 #define vcpu_clear_flag(v, ...) __vcpu_clear_flag((v), __VA_ARGS__)
637 /* SVE exposed to guest */
638 #define GUEST_HAS_SVE __vcpu_single_flag(cflags, BIT(0))
639 /* SVE config completed */
640 #define VCPU_SVE_FINALIZED __vcpu_single_flag(cflags, BIT(1))
641 /* PTRAUTH exposed to guest */
642 #define GUEST_HAS_PTRAUTH __vcpu_single_flag(cflags, BIT(2))
644 /* Exception pending */
645 #define PENDING_EXCEPTION __vcpu_single_flag(iflags, BIT(0))
647 * PC increment. Overlaps with EXCEPT_MASK on purpose so that it can't
648 * be set together with an exception...
650 #define INCREMENT_PC __vcpu_single_flag(iflags, BIT(1))
651 /* Target EL/MODE (not a single flag, but let's abuse the macro) */
652 #define EXCEPT_MASK __vcpu_single_flag(iflags, GENMASK(3, 1))
654 /* Helpers to encode exceptions with minimum fuss */
655 #define __EXCEPT_MASK_VAL unpack_vcpu_flag(EXCEPT_MASK)
656 #define __EXCEPT_SHIFT __builtin_ctzl(__EXCEPT_MASK_VAL)
657 #define __vcpu_except_flags(_f) iflags, (_f << __EXCEPT_SHIFT), __EXCEPT_MASK_VAL
660 * When PENDING_EXCEPTION is set, EXCEPT_MASK can take the following
665 #define EXCEPT_AA32_UND __vcpu_except_flags(0)
666 #define EXCEPT_AA32_IABT __vcpu_except_flags(1)
667 #define EXCEPT_AA32_DABT __vcpu_except_flags(2)
669 #define EXCEPT_AA64_EL1_SYNC __vcpu_except_flags(0)
670 #define EXCEPT_AA64_EL1_IRQ __vcpu_except_flags(1)
671 #define EXCEPT_AA64_EL1_FIQ __vcpu_except_flags(2)
672 #define EXCEPT_AA64_EL1_SERR __vcpu_except_flags(3)
673 /* For AArch64 with NV: */
674 #define EXCEPT_AA64_EL2_SYNC __vcpu_except_flags(4)
675 #define EXCEPT_AA64_EL2_IRQ __vcpu_except_flags(5)
676 #define EXCEPT_AA64_EL2_FIQ __vcpu_except_flags(6)
677 #define EXCEPT_AA64_EL2_SERR __vcpu_except_flags(7)
678 /* Guest debug is live */
679 #define DEBUG_DIRTY __vcpu_single_flag(iflags, BIT(4))
680 /* Save SPE context if active */
681 #define DEBUG_STATE_SAVE_SPE __vcpu_single_flag(iflags, BIT(5))
682 /* Save TRBE context if active */
683 #define DEBUG_STATE_SAVE_TRBE __vcpu_single_flag(iflags, BIT(6))
684 /* vcpu running in HYP context */
685 #define VCPU_HYP_CONTEXT __vcpu_single_flag(iflags, BIT(7))
687 /* SVE enabled for host EL0 */
688 #define HOST_SVE_ENABLED __vcpu_single_flag(sflags, BIT(0))
689 /* SME enabled for EL0 */
690 #define HOST_SME_ENABLED __vcpu_single_flag(sflags, BIT(1))
691 /* Physical CPU not in supported_cpus */
692 #define ON_UNSUPPORTED_CPU __vcpu_single_flag(sflags, BIT(2))
693 /* WFIT instruction trapped */
694 #define IN_WFIT __vcpu_single_flag(sflags, BIT(3))
695 /* vcpu system registers loaded on physical CPU */
696 #define SYSREGS_ON_CPU __vcpu_single_flag(sflags, BIT(4))
697 /* Software step state is Active-pending */
698 #define DBG_SS_ACTIVE_PENDING __vcpu_single_flag(sflags, BIT(5))
701 /* Pointer to the vcpu's SVE FFR for sve_{save,load}_state() */
702 #define vcpu_sve_pffr(vcpu) (kern_hyp_va((vcpu)->arch.sve_state) + \
703 sve_ffr_offset((vcpu)->arch.sve_max_vl))
705 #define vcpu_sve_max_vq(vcpu) sve_vq_from_vl((vcpu)->arch.sve_max_vl)
707 #define vcpu_sve_state_size(vcpu) ({ \
709 unsigned int __vcpu_vq; \
711 if (WARN_ON(!sve_vl_valid((vcpu)->arch.sve_max_vl))) { \
714 __vcpu_vq = vcpu_sve_max_vq(vcpu); \
715 __size_ret = SVE_SIG_REGS_SIZE(__vcpu_vq); \
721 #define KVM_GUESTDBG_VALID_MASK (KVM_GUESTDBG_ENABLE | \
722 KVM_GUESTDBG_USE_SW_BP | \
723 KVM_GUESTDBG_USE_HW | \
724 KVM_GUESTDBG_SINGLESTEP)
726 #define vcpu_has_sve(vcpu) (system_supports_sve() && \
727 vcpu_get_flag(vcpu, GUEST_HAS_SVE))
729 #ifdef CONFIG_ARM64_PTR_AUTH
730 #define vcpu_has_ptrauth(vcpu) \
731 ((cpus_have_final_cap(ARM64_HAS_ADDRESS_AUTH) || \
732 cpus_have_final_cap(ARM64_HAS_GENERIC_AUTH)) && \
733 vcpu_get_flag(vcpu, GUEST_HAS_PTRAUTH))
735 #define vcpu_has_ptrauth(vcpu) false
738 #define vcpu_on_unsupported_cpu(vcpu) \
739 vcpu_get_flag(vcpu, ON_UNSUPPORTED_CPU)
741 #define vcpu_set_on_unsupported_cpu(vcpu) \
742 vcpu_set_flag(vcpu, ON_UNSUPPORTED_CPU)
744 #define vcpu_clear_on_unsupported_cpu(vcpu) \
745 vcpu_clear_flag(vcpu, ON_UNSUPPORTED_CPU)
747 #define vcpu_gp_regs(v) (&(v)->arch.ctxt.regs)
750 * Only use __vcpu_sys_reg/ctxt_sys_reg if you know you want the
751 * memory backed version of a register, and not the one most recently
752 * accessed by a running VCPU. For example, for userspace access or
753 * for system registers that are never context switched, but only
756 #define __ctxt_sys_reg(c,r) (&(c)->sys_regs[(r)])
758 #define ctxt_sys_reg(c,r) (*__ctxt_sys_reg(c,r))
760 #define __vcpu_sys_reg(v,r) (ctxt_sys_reg(&(v)->arch.ctxt, (r)))
762 u64 vcpu_read_sys_reg(const struct kvm_vcpu *vcpu, int reg);
763 void vcpu_write_sys_reg(struct kvm_vcpu *vcpu, u64 val, int reg);
765 static inline bool __vcpu_read_sys_reg_from_cpu(int reg, u64 *val)
770 * System registers listed in the switch are not saved on every
771 * exit from the guest but are only saved on vcpu_put.
773 * Note that MPIDR_EL1 for the guest is set by KVM via VMPIDR_EL2 but
774 * should never be listed below, because the guest cannot modify its
775 * own MPIDR_EL1 and MPIDR_EL1 is accessed for VCPU A from VCPU B's
776 * thread when emulating cross-VCPU communication.
782 case SCTLR_EL1: *val = read_sysreg_s(SYS_SCTLR_EL12); break;
783 case CPACR_EL1: *val = read_sysreg_s(SYS_CPACR_EL12); break;
784 case TTBR0_EL1: *val = read_sysreg_s(SYS_TTBR0_EL12); break;
785 case TTBR1_EL1: *val = read_sysreg_s(SYS_TTBR1_EL12); break;
786 case TCR_EL1: *val = read_sysreg_s(SYS_TCR_EL12); break;
787 case ESR_EL1: *val = read_sysreg_s(SYS_ESR_EL12); break;
788 case AFSR0_EL1: *val = read_sysreg_s(SYS_AFSR0_EL12); break;
789 case AFSR1_EL1: *val = read_sysreg_s(SYS_AFSR1_EL12); break;
790 case FAR_EL1: *val = read_sysreg_s(SYS_FAR_EL12); break;
791 case MAIR_EL1: *val = read_sysreg_s(SYS_MAIR_EL12); break;
792 case VBAR_EL1: *val = read_sysreg_s(SYS_VBAR_EL12); break;
793 case CONTEXTIDR_EL1: *val = read_sysreg_s(SYS_CONTEXTIDR_EL12);break;
794 case TPIDR_EL0: *val = read_sysreg_s(SYS_TPIDR_EL0); break;
795 case TPIDRRO_EL0: *val = read_sysreg_s(SYS_TPIDRRO_EL0); break;
796 case TPIDR_EL1: *val = read_sysreg_s(SYS_TPIDR_EL1); break;
797 case AMAIR_EL1: *val = read_sysreg_s(SYS_AMAIR_EL12); break;
798 case CNTKCTL_EL1: *val = read_sysreg_s(SYS_CNTKCTL_EL12); break;
799 case ELR_EL1: *val = read_sysreg_s(SYS_ELR_EL12); break;
800 case PAR_EL1: *val = read_sysreg_par(); break;
801 case DACR32_EL2: *val = read_sysreg_s(SYS_DACR32_EL2); break;
802 case IFSR32_EL2: *val = read_sysreg_s(SYS_IFSR32_EL2); break;
803 case DBGVCR32_EL2: *val = read_sysreg_s(SYS_DBGVCR32_EL2); break;
804 default: return false;
810 static inline bool __vcpu_write_sys_reg_to_cpu(u64 val, int reg)
815 * System registers listed in the switch are not restored on every
816 * entry to the guest but are only restored on vcpu_load.
818 * Note that MPIDR_EL1 for the guest is set by KVM via VMPIDR_EL2 but
819 * should never be listed below, because the MPIDR should only be set
820 * once, before running the VCPU, and never changed later.
826 case SCTLR_EL1: write_sysreg_s(val, SYS_SCTLR_EL12); break;
827 case CPACR_EL1: write_sysreg_s(val, SYS_CPACR_EL12); break;
828 case TTBR0_EL1: write_sysreg_s(val, SYS_TTBR0_EL12); break;
829 case TTBR1_EL1: write_sysreg_s(val, SYS_TTBR1_EL12); break;
830 case TCR_EL1: write_sysreg_s(val, SYS_TCR_EL12); break;
831 case ESR_EL1: write_sysreg_s(val, SYS_ESR_EL12); break;
832 case AFSR0_EL1: write_sysreg_s(val, SYS_AFSR0_EL12); break;
833 case AFSR1_EL1: write_sysreg_s(val, SYS_AFSR1_EL12); break;
834 case FAR_EL1: write_sysreg_s(val, SYS_FAR_EL12); break;
835 case MAIR_EL1: write_sysreg_s(val, SYS_MAIR_EL12); break;
836 case VBAR_EL1: write_sysreg_s(val, SYS_VBAR_EL12); break;
837 case CONTEXTIDR_EL1: write_sysreg_s(val, SYS_CONTEXTIDR_EL12);break;
838 case TPIDR_EL0: write_sysreg_s(val, SYS_TPIDR_EL0); break;
839 case TPIDRRO_EL0: write_sysreg_s(val, SYS_TPIDRRO_EL0); break;
840 case TPIDR_EL1: write_sysreg_s(val, SYS_TPIDR_EL1); break;
841 case AMAIR_EL1: write_sysreg_s(val, SYS_AMAIR_EL12); break;
842 case CNTKCTL_EL1: write_sysreg_s(val, SYS_CNTKCTL_EL12); break;
843 case ELR_EL1: write_sysreg_s(val, SYS_ELR_EL12); break;
844 case PAR_EL1: write_sysreg_s(val, SYS_PAR_EL1); break;
845 case DACR32_EL2: write_sysreg_s(val, SYS_DACR32_EL2); break;
846 case IFSR32_EL2: write_sysreg_s(val, SYS_IFSR32_EL2); break;
847 case DBGVCR32_EL2: write_sysreg_s(val, SYS_DBGVCR32_EL2); break;
848 default: return false;
855 struct kvm_vm_stat_generic generic;
858 struct kvm_vcpu_stat {
859 struct kvm_vcpu_stat_generic generic;
864 u64 mmio_exit_kernel;
869 void kvm_vcpu_preferred_target(struct kvm_vcpu_init *init);
870 unsigned long kvm_arm_num_regs(struct kvm_vcpu *vcpu);
871 int kvm_arm_copy_reg_indices(struct kvm_vcpu *vcpu, u64 __user *indices);
872 int kvm_arm_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
873 int kvm_arm_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
875 unsigned long kvm_arm_num_sys_reg_descs(struct kvm_vcpu *vcpu);
876 int kvm_arm_copy_sys_reg_indices(struct kvm_vcpu *vcpu, u64 __user *uindices);
878 int __kvm_arm_vcpu_get_events(struct kvm_vcpu *vcpu,
879 struct kvm_vcpu_events *events);
881 int __kvm_arm_vcpu_set_events(struct kvm_vcpu *vcpu,
882 struct kvm_vcpu_events *events);
884 #define KVM_ARCH_WANT_MMU_NOTIFIER
886 void kvm_arm_halt_guest(struct kvm *kvm);
887 void kvm_arm_resume_guest(struct kvm *kvm);
889 #define vcpu_has_run_once(vcpu) !!rcu_access_pointer((vcpu)->pid)
891 #ifndef __KVM_NVHE_HYPERVISOR__
892 #define kvm_call_hyp_nvhe(f, ...) \
894 struct arm_smccc_res res; \
896 arm_smccc_1_1_hvc(KVM_HOST_SMCCC_FUNC(f), \
897 ##__VA_ARGS__, &res); \
898 WARN_ON(res.a0 != SMCCC_RET_SUCCESS); \
904 * The couple of isb() below are there to guarantee the same behaviour
905 * on VHE as on !VHE, where the eret to EL1 acts as a context
906 * synchronization event.
908 #define kvm_call_hyp(f, ...) \
914 kvm_call_hyp_nvhe(f, ##__VA_ARGS__); \
918 #define kvm_call_hyp_ret(f, ...) \
920 typeof(f(__VA_ARGS__)) ret; \
923 ret = f(__VA_ARGS__); \
926 ret = kvm_call_hyp_nvhe(f, ##__VA_ARGS__); \
931 #else /* __KVM_NVHE_HYPERVISOR__ */
932 #define kvm_call_hyp(f, ...) f(__VA_ARGS__)
933 #define kvm_call_hyp_ret(f, ...) f(__VA_ARGS__)
934 #define kvm_call_hyp_nvhe(f, ...) f(__VA_ARGS__)
935 #endif /* __KVM_NVHE_HYPERVISOR__ */
937 void force_vm_exit(const cpumask_t *mask);
939 int handle_exit(struct kvm_vcpu *vcpu, int exception_index);
940 void handle_exit_early(struct kvm_vcpu *vcpu, int exception_index);
942 int kvm_handle_cp14_load_store(struct kvm_vcpu *vcpu);
943 int kvm_handle_cp14_32(struct kvm_vcpu *vcpu);
944 int kvm_handle_cp14_64(struct kvm_vcpu *vcpu);
945 int kvm_handle_cp15_32(struct kvm_vcpu *vcpu);
946 int kvm_handle_cp15_64(struct kvm_vcpu *vcpu);
947 int kvm_handle_sys_reg(struct kvm_vcpu *vcpu);
948 int kvm_handle_cp10_id(struct kvm_vcpu *vcpu);
950 void kvm_reset_sys_regs(struct kvm_vcpu *vcpu);
952 int __init kvm_sys_reg_table_init(void);
954 bool lock_all_vcpus(struct kvm *kvm);
955 void unlock_all_vcpus(struct kvm *kvm);
958 void kvm_mmio_write_buf(void *buf, unsigned int len, unsigned long data);
959 unsigned long kvm_mmio_read_buf(const void *buf, unsigned int len);
961 int kvm_handle_mmio_return(struct kvm_vcpu *vcpu);
962 int io_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa);
965 * Returns true if a Performance Monitoring Interrupt (PMI), a.k.a. perf event,
966 * arrived in guest context. For arm64, any event that arrives while a vCPU is
967 * loaded is considered to be "in guest".
969 static inline bool kvm_arch_pmi_in_guest(struct kvm_vcpu *vcpu)
971 return IS_ENABLED(CONFIG_GUEST_PERF_EVENTS) && !!vcpu;
974 long kvm_hypercall_pv_features(struct kvm_vcpu *vcpu);
975 gpa_t kvm_init_stolen_time(struct kvm_vcpu *vcpu);
976 void kvm_update_stolen_time(struct kvm_vcpu *vcpu);
978 bool kvm_arm_pvtime_supported(void);
979 int kvm_arm_pvtime_set_attr(struct kvm_vcpu *vcpu,
980 struct kvm_device_attr *attr);
981 int kvm_arm_pvtime_get_attr(struct kvm_vcpu *vcpu,
982 struct kvm_device_attr *attr);
983 int kvm_arm_pvtime_has_attr(struct kvm_vcpu *vcpu,
984 struct kvm_device_attr *attr);
986 extern unsigned int __ro_after_init kvm_arm_vmid_bits;
987 int __init kvm_arm_vmid_alloc_init(void);
988 void __init kvm_arm_vmid_alloc_free(void);
989 void kvm_arm_vmid_update(struct kvm_vmid *kvm_vmid);
990 void kvm_arm_vmid_clear_active(void);
992 static inline void kvm_arm_pvtime_vcpu_init(struct kvm_vcpu_arch *vcpu_arch)
994 vcpu_arch->steal.base = INVALID_GPA;
997 static inline bool kvm_arm_is_pvtime_enabled(struct kvm_vcpu_arch *vcpu_arch)
999 return (vcpu_arch->steal.base != INVALID_GPA);
1002 void kvm_set_sei_esr(struct kvm_vcpu *vcpu, u64 syndrome);
1004 struct kvm_vcpu *kvm_mpidr_to_vcpu(struct kvm *kvm, unsigned long mpidr);
1006 DECLARE_KVM_HYP_PER_CPU(struct kvm_host_data, kvm_host_data);
1008 static inline void kvm_init_host_cpu_context(struct kvm_cpu_context *cpu_ctxt)
1010 /* The host's MPIDR is immutable, so let's set it up at boot time */
1011 ctxt_sys_reg(cpu_ctxt, MPIDR_EL1) = read_cpuid_mpidr();
1014 static inline bool kvm_system_needs_idmapped_vectors(void)
1016 return cpus_have_const_cap(ARM64_SPECTRE_V3A);
1019 void kvm_arm_vcpu_ptrauth_trap(struct kvm_vcpu *vcpu);
1021 static inline void kvm_arch_sync_events(struct kvm *kvm) {}
1022 static inline void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu) {}
1024 void kvm_arm_init_debug(void);
1025 void kvm_arm_vcpu_init_debug(struct kvm_vcpu *vcpu);
1026 void kvm_arm_setup_debug(struct kvm_vcpu *vcpu);
1027 void kvm_arm_clear_debug(struct kvm_vcpu *vcpu);
1028 void kvm_arm_reset_debug_ptr(struct kvm_vcpu *vcpu);
1030 #define kvm_vcpu_os_lock_enabled(vcpu) \
1031 (!!(__vcpu_sys_reg(vcpu, OSLSR_EL1) & SYS_OSLSR_OSLK))
1033 int kvm_arm_vcpu_arch_set_attr(struct kvm_vcpu *vcpu,
1034 struct kvm_device_attr *attr);
1035 int kvm_arm_vcpu_arch_get_attr(struct kvm_vcpu *vcpu,
1036 struct kvm_device_attr *attr);
1037 int kvm_arm_vcpu_arch_has_attr(struct kvm_vcpu *vcpu,
1038 struct kvm_device_attr *attr);
1040 int kvm_vm_ioctl_mte_copy_tags(struct kvm *kvm,
1041 struct kvm_arm_copy_mte_tags *copy_tags);
1042 int kvm_vm_ioctl_set_counter_offset(struct kvm *kvm,
1043 struct kvm_arm_counter_offset *offset);
1045 /* Guest/host FPSIMD coordination helpers */
1046 int kvm_arch_vcpu_run_map_fp(struct kvm_vcpu *vcpu);
1047 void kvm_arch_vcpu_load_fp(struct kvm_vcpu *vcpu);
1048 void kvm_arch_vcpu_ctxflush_fp(struct kvm_vcpu *vcpu);
1049 void kvm_arch_vcpu_ctxsync_fp(struct kvm_vcpu *vcpu);
1050 void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu);
1051 void kvm_vcpu_unshare_task_fp(struct kvm_vcpu *vcpu);
1053 static inline bool kvm_pmu_counter_deferred(struct perf_event_attr *attr)
1055 return (!has_vhe() && attr->exclude_host);
1058 /* Flags for host debug state */
1059 void kvm_arch_vcpu_load_debug_state_flags(struct kvm_vcpu *vcpu);
1060 void kvm_arch_vcpu_put_debug_state_flags(struct kvm_vcpu *vcpu);
1063 void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr);
1064 void kvm_clr_pmu_events(u32 clr);
1066 static inline void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr) {}
1067 static inline void kvm_clr_pmu_events(u32 clr) {}
1070 void kvm_vcpu_load_sysregs_vhe(struct kvm_vcpu *vcpu);
1071 void kvm_vcpu_put_sysregs_vhe(struct kvm_vcpu *vcpu);
1073 int __init kvm_set_ipa_limit(void);
1075 #define __KVM_HAVE_ARCH_VM_ALLOC
1076 struct kvm *kvm_arch_alloc_vm(void);
1078 static inline bool kvm_vm_is_protected(struct kvm *kvm)
1083 void kvm_init_protected_traps(struct kvm_vcpu *vcpu);
1085 int kvm_arm_vcpu_finalize(struct kvm_vcpu *vcpu, int feature);
1086 bool kvm_arm_vcpu_is_finalized(struct kvm_vcpu *vcpu);
1088 #define kvm_arm_vcpu_sve_finalized(vcpu) vcpu_get_flag(vcpu, VCPU_SVE_FINALIZED)
1090 #define kvm_has_mte(kvm) \
1091 (system_supports_mte() && \
1092 test_bit(KVM_ARCH_FLAG_MTE_ENABLED, &(kvm)->arch.flags))
1094 #define kvm_supports_32bit_el0() \
1095 (system_supports_32bit_el0() && \
1096 !static_branch_unlikely(&arm64_mismatched_32bit_el0))
1098 #define kvm_vm_has_ran_once(kvm) \
1099 (test_bit(KVM_ARCH_FLAG_HAS_RAN_ONCE, &(kvm)->arch.flags))
1101 int kvm_trng_call(struct kvm_vcpu *vcpu);
1103 extern phys_addr_t hyp_mem_base;
1104 extern phys_addr_t hyp_mem_size;
1105 void __init kvm_hyp_reserve(void);
1107 static inline void kvm_hyp_reserve(void) { }
1110 void kvm_arm_vcpu_power_off(struct kvm_vcpu *vcpu);
1111 bool kvm_arm_vcpu_stopped(struct kvm_vcpu *vcpu);
1113 #endif /* __ARM64_KVM_HOST_H__ */