xref: /openbmc/linux/arch/arm64/include/asm/kvm_host.h (revision b1f778a2)
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 
31 #define __KVM_HAVE_ARCH_INTC_INITIALIZED
32 
33 #define KVM_HALT_POLL_NS_DEFAULT 500000
34 
35 #include <kvm/arm_vgic.h>
36 #include <kvm/arm_arch_timer.h>
37 #include <kvm/arm_pmu.h>
38 
39 #define KVM_MAX_VCPUS VGIC_V3_MAX_CPUS
40 
41 #define KVM_VCPU_MAX_FEATURES 7
42 #define KVM_VCPU_VALID_FEATURES	(BIT(KVM_VCPU_MAX_FEATURES) - 1)
43 
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)
52 #define KVM_REQ_RESYNC_PMU_EL0	KVM_ARCH_REQ(7)
53 
54 #define KVM_DIRTY_LOG_MANUAL_CAPS   (KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE | \
55 				     KVM_DIRTY_LOG_INITIALLY_SET)
56 
57 #define KVM_HAVE_MMU_RWLOCK
58 
59 /*
60  * Mode of operation configurable with kvm-arm.mode early param.
61  * See Documentation/admin-guide/kernel-parameters.txt for more information.
62  */
63 enum kvm_mode {
64 	KVM_MODE_DEFAULT,
65 	KVM_MODE_PROTECTED,
66 	KVM_MODE_NV,
67 	KVM_MODE_NONE,
68 };
69 #ifdef CONFIG_KVM
70 enum kvm_mode kvm_get_mode(void);
71 #else
72 static inline enum kvm_mode kvm_get_mode(void) { return KVM_MODE_NONE; };
73 #endif
74 
75 DECLARE_STATIC_KEY_FALSE(userspace_irqchip_in_use);
76 
77 extern unsigned int __ro_after_init kvm_sve_max_vl;
78 int __init kvm_arm_init_sve(void);
79 
80 u32 __attribute_const__ kvm_target_cpu(void);
81 int kvm_reset_vcpu(struct kvm_vcpu *vcpu);
82 void kvm_arm_vcpu_destroy(struct kvm_vcpu *vcpu);
83 
84 struct kvm_hyp_memcache {
85 	phys_addr_t head;
86 	unsigned long nr_pages;
87 };
88 
89 static inline void push_hyp_memcache(struct kvm_hyp_memcache *mc,
90 				     phys_addr_t *p,
91 				     phys_addr_t (*to_pa)(void *virt))
92 {
93 	*p = mc->head;
94 	mc->head = to_pa(p);
95 	mc->nr_pages++;
96 }
97 
98 static inline void *pop_hyp_memcache(struct kvm_hyp_memcache *mc,
99 				     void *(*to_va)(phys_addr_t phys))
100 {
101 	phys_addr_t *p = to_va(mc->head);
102 
103 	if (!mc->nr_pages)
104 		return NULL;
105 
106 	mc->head = *p;
107 	mc->nr_pages--;
108 
109 	return p;
110 }
111 
112 static inline int __topup_hyp_memcache(struct kvm_hyp_memcache *mc,
113 				       unsigned long min_pages,
114 				       void *(*alloc_fn)(void *arg),
115 				       phys_addr_t (*to_pa)(void *virt),
116 				       void *arg)
117 {
118 	while (mc->nr_pages < min_pages) {
119 		phys_addr_t *p = alloc_fn(arg);
120 
121 		if (!p)
122 			return -ENOMEM;
123 		push_hyp_memcache(mc, p, to_pa);
124 	}
125 
126 	return 0;
127 }
128 
129 static inline void __free_hyp_memcache(struct kvm_hyp_memcache *mc,
130 				       void (*free_fn)(void *virt, void *arg),
131 				       void *(*to_va)(phys_addr_t phys),
132 				       void *arg)
133 {
134 	while (mc->nr_pages)
135 		free_fn(pop_hyp_memcache(mc, to_va), arg);
136 }
137 
138 void free_hyp_memcache(struct kvm_hyp_memcache *mc);
139 int topup_hyp_memcache(struct kvm_hyp_memcache *mc, unsigned long min_pages);
140 
141 struct kvm_vmid {
142 	atomic64_t id;
143 };
144 
145 struct kvm_s2_mmu {
146 	struct kvm_vmid vmid;
147 
148 	/*
149 	 * stage2 entry level table
150 	 *
151 	 * Two kvm_s2_mmu structures in the same VM can point to the same
152 	 * pgd here.  This happens when running a guest using a
153 	 * translation regime that isn't affected by its own stage-2
154 	 * translation, such as a non-VHE hypervisor running at vEL2, or
155 	 * for vEL1/EL0 with vHCR_EL2.VM == 0.  In that case, we use the
156 	 * canonical stage-2 page tables.
157 	 */
158 	phys_addr_t	pgd_phys;
159 	struct kvm_pgtable *pgt;
160 
161 	/* The last vcpu id that ran on each physical CPU */
162 	int __percpu *last_vcpu_ran;
163 
164 #define KVM_ARM_EAGER_SPLIT_CHUNK_SIZE_DEFAULT 0
165 	/*
166 	 * Memory cache used to split
167 	 * KVM_CAP_ARM_EAGER_SPLIT_CHUNK_SIZE worth of huge pages. It
168 	 * is used to allocate stage2 page tables while splitting huge
169 	 * pages. The choice of KVM_CAP_ARM_EAGER_SPLIT_CHUNK_SIZE
170 	 * influences both the capacity of the split page cache, and
171 	 * how often KVM reschedules. Be wary of raising CHUNK_SIZE
172 	 * too high.
173 	 *
174 	 * Protected by kvm->slots_lock.
175 	 */
176 	struct kvm_mmu_memory_cache split_page_cache;
177 	uint64_t split_page_chunk_size;
178 
179 	struct kvm_arch *arch;
180 };
181 
182 struct kvm_arch_memory_slot {
183 };
184 
185 /**
186  * struct kvm_smccc_features: Descriptor of the hypercall services exposed to the guests
187  *
188  * @std_bmap: Bitmap of standard secure service calls
189  * @std_hyp_bmap: Bitmap of standard hypervisor service calls
190  * @vendor_hyp_bmap: Bitmap of vendor specific hypervisor service calls
191  */
192 struct kvm_smccc_features {
193 	unsigned long std_bmap;
194 	unsigned long std_hyp_bmap;
195 	unsigned long vendor_hyp_bmap;
196 };
197 
198 typedef unsigned int pkvm_handle_t;
199 
200 struct kvm_protected_vm {
201 	pkvm_handle_t handle;
202 	struct kvm_hyp_memcache teardown_mc;
203 };
204 
205 struct kvm_arch {
206 	struct kvm_s2_mmu mmu;
207 
208 	/* VTCR_EL2 value for this VM */
209 	u64    vtcr;
210 
211 	/* Interrupt controller */
212 	struct vgic_dist	vgic;
213 
214 	/* Timers */
215 	struct arch_timer_vm_data timer_data;
216 
217 	/* Mandated version of PSCI */
218 	u32 psci_version;
219 
220 	/* Protects VM-scoped configuration data */
221 	struct mutex config_lock;
222 
223 	/*
224 	 * If we encounter a data abort without valid instruction syndrome
225 	 * information, report this to user space.  User space can (and
226 	 * should) opt in to this feature if KVM_CAP_ARM_NISV_TO_USER is
227 	 * supported.
228 	 */
229 #define KVM_ARCH_FLAG_RETURN_NISV_IO_ABORT_TO_USER	0
230 	/* Memory Tagging Extension enabled for the guest */
231 #define KVM_ARCH_FLAG_MTE_ENABLED			1
232 	/* At least one vCPU has ran in the VM */
233 #define KVM_ARCH_FLAG_HAS_RAN_ONCE			2
234 	/* The vCPU feature set for the VM is configured */
235 #define KVM_ARCH_FLAG_VCPU_FEATURES_CONFIGURED		3
236 	/* PSCI SYSTEM_SUSPEND enabled for the guest */
237 #define KVM_ARCH_FLAG_SYSTEM_SUSPEND_ENABLED		4
238 	/* VM counter offset */
239 #define KVM_ARCH_FLAG_VM_COUNTER_OFFSET			5
240 	/* Timer PPIs made immutable */
241 #define KVM_ARCH_FLAG_TIMER_PPIS_IMMUTABLE		6
242 	/* SMCCC filter initialized for the VM */
243 #define KVM_ARCH_FLAG_SMCCC_FILTER_CONFIGURED		7
244 	/* Initial ID reg values loaded */
245 #define KVM_ARCH_FLAG_ID_REGS_INITIALIZED		8
246 	unsigned long flags;
247 
248 	/* VM-wide vCPU feature set */
249 	DECLARE_BITMAP(vcpu_features, KVM_VCPU_MAX_FEATURES);
250 
251 	/*
252 	 * VM-wide PMU filter, implemented as a bitmap and big enough for
253 	 * up to 2^10 events (ARMv8.0) or 2^16 events (ARMv8.1+).
254 	 */
255 	unsigned long *pmu_filter;
256 	struct arm_pmu *arm_pmu;
257 
258 	cpumask_var_t supported_cpus;
259 
260 	/* Hypercall features firmware registers' descriptor */
261 	struct kvm_smccc_features smccc_feat;
262 	struct maple_tree smccc_filter;
263 
264 	/*
265 	 * Emulated CPU ID registers per VM
266 	 * (Op0, Op1, CRn, CRm, Op2) of the ID registers to be saved in it
267 	 * is (3, 0, 0, crm, op2), where 1<=crm<8, 0<=op2<8.
268 	 *
269 	 * These emulated idregs are VM-wide, but accessed from the context of a vCPU.
270 	 * Atomic access to multiple idregs are guarded by kvm_arch.config_lock.
271 	 */
272 #define IDREG_IDX(id)		(((sys_reg_CRm(id) - 1) << 3) | sys_reg_Op2(id))
273 #define IDREG(kvm, id)		((kvm)->arch.id_regs[IDREG_IDX(id)])
274 #define KVM_ARM_ID_REG_NUM	(IDREG_IDX(sys_reg(3, 0, 0, 7, 7)) + 1)
275 	u64 id_regs[KVM_ARM_ID_REG_NUM];
276 
277 	/*
278 	 * For an untrusted host VM, 'pkvm.handle' is used to lookup
279 	 * the associated pKVM instance in the hypervisor.
280 	 */
281 	struct kvm_protected_vm pkvm;
282 };
283 
284 struct kvm_vcpu_fault_info {
285 	u64 esr_el2;		/* Hyp Syndrom Register */
286 	u64 far_el2;		/* Hyp Fault Address Register */
287 	u64 hpfar_el2;		/* Hyp IPA Fault Address Register */
288 	u64 disr_el1;		/* Deferred [SError] Status Register */
289 };
290 
291 enum vcpu_sysreg {
292 	__INVALID_SYSREG__,   /* 0 is reserved as an invalid value */
293 	MPIDR_EL1,	/* MultiProcessor Affinity Register */
294 	CLIDR_EL1,	/* Cache Level ID Register */
295 	CSSELR_EL1,	/* Cache Size Selection Register */
296 	SCTLR_EL1,	/* System Control Register */
297 	ACTLR_EL1,	/* Auxiliary Control Register */
298 	CPACR_EL1,	/* Coprocessor Access Control */
299 	ZCR_EL1,	/* SVE Control */
300 	TTBR0_EL1,	/* Translation Table Base Register 0 */
301 	TTBR1_EL1,	/* Translation Table Base Register 1 */
302 	TCR_EL1,	/* Translation Control Register */
303 	TCR2_EL1,	/* Extended Translation Control Register */
304 	ESR_EL1,	/* Exception Syndrome Register */
305 	AFSR0_EL1,	/* Auxiliary Fault Status Register 0 */
306 	AFSR1_EL1,	/* Auxiliary Fault Status Register 1 */
307 	FAR_EL1,	/* Fault Address Register */
308 	MAIR_EL1,	/* Memory Attribute Indirection Register */
309 	VBAR_EL1,	/* Vector Base Address Register */
310 	CONTEXTIDR_EL1,	/* Context ID Register */
311 	TPIDR_EL0,	/* Thread ID, User R/W */
312 	TPIDRRO_EL0,	/* Thread ID, User R/O */
313 	TPIDR_EL1,	/* Thread ID, Privileged */
314 	AMAIR_EL1,	/* Aux Memory Attribute Indirection Register */
315 	CNTKCTL_EL1,	/* Timer Control Register (EL1) */
316 	PAR_EL1,	/* Physical Address Register */
317 	MDSCR_EL1,	/* Monitor Debug System Control Register */
318 	MDCCINT_EL1,	/* Monitor Debug Comms Channel Interrupt Enable Reg */
319 	OSLSR_EL1,	/* OS Lock Status Register */
320 	DISR_EL1,	/* Deferred Interrupt Status Register */
321 
322 	/* Performance Monitors Registers */
323 	PMCR_EL0,	/* Control Register */
324 	PMSELR_EL0,	/* Event Counter Selection Register */
325 	PMEVCNTR0_EL0,	/* Event Counter Register (0-30) */
326 	PMEVCNTR30_EL0 = PMEVCNTR0_EL0 + 30,
327 	PMCCNTR_EL0,	/* Cycle Counter Register */
328 	PMEVTYPER0_EL0,	/* Event Type Register (0-30) */
329 	PMEVTYPER30_EL0 = PMEVTYPER0_EL0 + 30,
330 	PMCCFILTR_EL0,	/* Cycle Count Filter Register */
331 	PMCNTENSET_EL0,	/* Count Enable Set Register */
332 	PMINTENSET_EL1,	/* Interrupt Enable Set Register */
333 	PMOVSSET_EL0,	/* Overflow Flag Status Set Register */
334 	PMUSERENR_EL0,	/* User Enable Register */
335 
336 	/* Pointer Authentication Registers in a strict increasing order. */
337 	APIAKEYLO_EL1,
338 	APIAKEYHI_EL1,
339 	APIBKEYLO_EL1,
340 	APIBKEYHI_EL1,
341 	APDAKEYLO_EL1,
342 	APDAKEYHI_EL1,
343 	APDBKEYLO_EL1,
344 	APDBKEYHI_EL1,
345 	APGAKEYLO_EL1,
346 	APGAKEYHI_EL1,
347 
348 	ELR_EL1,
349 	SP_EL1,
350 	SPSR_EL1,
351 
352 	CNTVOFF_EL2,
353 	CNTV_CVAL_EL0,
354 	CNTV_CTL_EL0,
355 	CNTP_CVAL_EL0,
356 	CNTP_CTL_EL0,
357 
358 	/* Memory Tagging Extension registers */
359 	RGSR_EL1,	/* Random Allocation Tag Seed Register */
360 	GCR_EL1,	/* Tag Control Register */
361 	TFSR_EL1,	/* Tag Fault Status Register (EL1) */
362 	TFSRE0_EL1,	/* Tag Fault Status Register (EL0) */
363 
364 	/* Permission Indirection Extension registers */
365 	PIR_EL1,       /* Permission Indirection Register 1 (EL1) */
366 	PIRE0_EL1,     /*  Permission Indirection Register 0 (EL1) */
367 
368 	/* 32bit specific registers. */
369 	DACR32_EL2,	/* Domain Access Control Register */
370 	IFSR32_EL2,	/* Instruction Fault Status Register */
371 	FPEXC32_EL2,	/* Floating-Point Exception Control Register */
372 	DBGVCR32_EL2,	/* Debug Vector Catch Register */
373 
374 	/* EL2 registers */
375 	VPIDR_EL2,	/* Virtualization Processor ID Register */
376 	VMPIDR_EL2,	/* Virtualization Multiprocessor ID Register */
377 	SCTLR_EL2,	/* System Control Register (EL2) */
378 	ACTLR_EL2,	/* Auxiliary Control Register (EL2) */
379 	HCR_EL2,	/* Hypervisor Configuration Register */
380 	MDCR_EL2,	/* Monitor Debug Configuration Register (EL2) */
381 	CPTR_EL2,	/* Architectural Feature Trap Register (EL2) */
382 	HSTR_EL2,	/* Hypervisor System Trap Register */
383 	HACR_EL2,	/* Hypervisor Auxiliary Control Register */
384 	TTBR0_EL2,	/* Translation Table Base Register 0 (EL2) */
385 	TTBR1_EL2,	/* Translation Table Base Register 1 (EL2) */
386 	TCR_EL2,	/* Translation Control Register (EL2) */
387 	VTTBR_EL2,	/* Virtualization Translation Table Base Register */
388 	VTCR_EL2,	/* Virtualization Translation Control Register */
389 	SPSR_EL2,	/* EL2 saved program status register */
390 	ELR_EL2,	/* EL2 exception link register */
391 	AFSR0_EL2,	/* Auxiliary Fault Status Register 0 (EL2) */
392 	AFSR1_EL2,	/* Auxiliary Fault Status Register 1 (EL2) */
393 	ESR_EL2,	/* Exception Syndrome Register (EL2) */
394 	FAR_EL2,	/* Fault Address Register (EL2) */
395 	HPFAR_EL2,	/* Hypervisor IPA Fault Address Register */
396 	MAIR_EL2,	/* Memory Attribute Indirection Register (EL2) */
397 	AMAIR_EL2,	/* Auxiliary Memory Attribute Indirection Register (EL2) */
398 	VBAR_EL2,	/* Vector Base Address Register (EL2) */
399 	RVBAR_EL2,	/* Reset Vector Base Address Register */
400 	CONTEXTIDR_EL2,	/* Context ID Register (EL2) */
401 	TPIDR_EL2,	/* EL2 Software Thread ID Register */
402 	CNTHCTL_EL2,	/* Counter-timer Hypervisor Control register */
403 	SP_EL2,		/* EL2 Stack Pointer */
404 	CNTHP_CTL_EL2,
405 	CNTHP_CVAL_EL2,
406 	CNTHV_CTL_EL2,
407 	CNTHV_CVAL_EL2,
408 
409 	NR_SYS_REGS	/* Nothing after this line! */
410 };
411 
412 struct kvm_cpu_context {
413 	struct user_pt_regs regs;	/* sp = sp_el0 */
414 
415 	u64	spsr_abt;
416 	u64	spsr_und;
417 	u64	spsr_irq;
418 	u64	spsr_fiq;
419 
420 	struct user_fpsimd_state fp_regs;
421 
422 	u64 sys_regs[NR_SYS_REGS];
423 
424 	struct kvm_vcpu *__hyp_running_vcpu;
425 };
426 
427 struct kvm_host_data {
428 	struct kvm_cpu_context host_ctxt;
429 };
430 
431 struct kvm_host_psci_config {
432 	/* PSCI version used by host. */
433 	u32 version;
434 	u32 smccc_version;
435 
436 	/* Function IDs used by host if version is v0.1. */
437 	struct psci_0_1_function_ids function_ids_0_1;
438 
439 	bool psci_0_1_cpu_suspend_implemented;
440 	bool psci_0_1_cpu_on_implemented;
441 	bool psci_0_1_cpu_off_implemented;
442 	bool psci_0_1_migrate_implemented;
443 };
444 
445 extern struct kvm_host_psci_config kvm_nvhe_sym(kvm_host_psci_config);
446 #define kvm_host_psci_config CHOOSE_NVHE_SYM(kvm_host_psci_config)
447 
448 extern s64 kvm_nvhe_sym(hyp_physvirt_offset);
449 #define hyp_physvirt_offset CHOOSE_NVHE_SYM(hyp_physvirt_offset)
450 
451 extern u64 kvm_nvhe_sym(hyp_cpu_logical_map)[NR_CPUS];
452 #define hyp_cpu_logical_map CHOOSE_NVHE_SYM(hyp_cpu_logical_map)
453 
454 struct vcpu_reset_state {
455 	unsigned long	pc;
456 	unsigned long	r0;
457 	bool		be;
458 	bool		reset;
459 };
460 
461 struct kvm_vcpu_arch {
462 	struct kvm_cpu_context ctxt;
463 
464 	/*
465 	 * Guest floating point state
466 	 *
467 	 * The architecture has two main floating point extensions,
468 	 * the original FPSIMD and SVE.  These have overlapping
469 	 * register views, with the FPSIMD V registers occupying the
470 	 * low 128 bits of the SVE Z registers.  When the core
471 	 * floating point code saves the register state of a task it
472 	 * records which view it saved in fp_type.
473 	 */
474 	void *sve_state;
475 	enum fp_type fp_type;
476 	unsigned int sve_max_vl;
477 	u64 svcr;
478 
479 	/* Stage 2 paging state used by the hardware on next switch */
480 	struct kvm_s2_mmu *hw_mmu;
481 
482 	/* Values of trap registers for the guest. */
483 	u64 hcr_el2;
484 	u64 mdcr_el2;
485 	u64 cptr_el2;
486 
487 	/* Values of trap registers for the host before guest entry. */
488 	u64 mdcr_el2_host;
489 
490 	/* Exception Information */
491 	struct kvm_vcpu_fault_info fault;
492 
493 	/* Ownership of the FP regs */
494 	enum {
495 		FP_STATE_FREE,
496 		FP_STATE_HOST_OWNED,
497 		FP_STATE_GUEST_OWNED,
498 	} fp_state;
499 
500 	/* Configuration flags, set once and for all before the vcpu can run */
501 	u8 cflags;
502 
503 	/* Input flags to the hypervisor code, potentially cleared after use */
504 	u8 iflags;
505 
506 	/* State flags for kernel bookkeeping, unused by the hypervisor code */
507 	u8 sflags;
508 
509 	/*
510 	 * Don't run the guest (internal implementation need).
511 	 *
512 	 * Contrary to the flags above, this is set/cleared outside of
513 	 * a vcpu context, and thus cannot be mixed with the flags
514 	 * themselves (or the flag accesses need to be made atomic).
515 	 */
516 	bool pause;
517 
518 	/*
519 	 * We maintain more than a single set of debug registers to support
520 	 * debugging the guest from the host and to maintain separate host and
521 	 * guest state during world switches. vcpu_debug_state are the debug
522 	 * registers of the vcpu as the guest sees them.  host_debug_state are
523 	 * the host registers which are saved and restored during
524 	 * world switches. external_debug_state contains the debug
525 	 * values we want to debug the guest. This is set via the
526 	 * KVM_SET_GUEST_DEBUG ioctl.
527 	 *
528 	 * debug_ptr points to the set of debug registers that should be loaded
529 	 * onto the hardware when running the guest.
530 	 */
531 	struct kvm_guest_debug_arch *debug_ptr;
532 	struct kvm_guest_debug_arch vcpu_debug_state;
533 	struct kvm_guest_debug_arch external_debug_state;
534 
535 	struct user_fpsimd_state *host_fpsimd_state;	/* hyp VA */
536 	struct task_struct *parent_task;
537 
538 	struct {
539 		/* {Break,watch}point registers */
540 		struct kvm_guest_debug_arch regs;
541 		/* Statistical profiling extension */
542 		u64 pmscr_el1;
543 		/* Self-hosted trace */
544 		u64 trfcr_el1;
545 	} host_debug_state;
546 
547 	/* VGIC state */
548 	struct vgic_cpu vgic_cpu;
549 	struct arch_timer_cpu timer_cpu;
550 	struct kvm_pmu pmu;
551 
552 	/*
553 	 * Guest registers we preserve during guest debugging.
554 	 *
555 	 * These shadow registers are updated by the kvm_handle_sys_reg
556 	 * trap handler if the guest accesses or updates them while we
557 	 * are using guest debug.
558 	 */
559 	struct {
560 		u32	mdscr_el1;
561 		bool	pstate_ss;
562 	} guest_debug_preserved;
563 
564 	/* vcpu power state */
565 	struct kvm_mp_state mp_state;
566 	spinlock_t mp_state_lock;
567 
568 	/* Cache some mmu pages needed inside spinlock regions */
569 	struct kvm_mmu_memory_cache mmu_page_cache;
570 
571 	/* Target CPU and feature flags */
572 	int target;
573 	DECLARE_BITMAP(features, KVM_VCPU_MAX_FEATURES);
574 
575 	/* Virtual SError ESR to restore when HCR_EL2.VSE is set */
576 	u64 vsesr_el2;
577 
578 	/* Additional reset state */
579 	struct vcpu_reset_state	reset_state;
580 
581 	/* Guest PV state */
582 	struct {
583 		u64 last_steal;
584 		gpa_t base;
585 	} steal;
586 
587 	/* Per-vcpu CCSIDR override or NULL */
588 	u32 *ccsidr;
589 };
590 
591 /*
592  * Each 'flag' is composed of a comma-separated triplet:
593  *
594  * - the flag-set it belongs to in the vcpu->arch structure
595  * - the value for that flag
596  * - the mask for that flag
597  *
598  *  __vcpu_single_flag() builds such a triplet for a single-bit flag.
599  * unpack_vcpu_flag() extract the flag value from the triplet for
600  * direct use outside of the flag accessors.
601  */
602 #define __vcpu_single_flag(_set, _f)	_set, (_f), (_f)
603 
604 #define __unpack_flag(_set, _f, _m)	_f
605 #define unpack_vcpu_flag(...)		__unpack_flag(__VA_ARGS__)
606 
607 #define __build_check_flag(v, flagset, f, m)			\
608 	do {							\
609 		typeof(v->arch.flagset) *_fset;			\
610 								\
611 		/* Check that the flags fit in the mask */	\
612 		BUILD_BUG_ON(HWEIGHT(m) != HWEIGHT((f) | (m)));	\
613 		/* Check that the flags fit in the type */	\
614 		BUILD_BUG_ON((sizeof(*_fset) * 8) <= __fls(m));	\
615 	} while (0)
616 
617 #define __vcpu_get_flag(v, flagset, f, m)			\
618 	({							\
619 		__build_check_flag(v, flagset, f, m);		\
620 								\
621 		READ_ONCE(v->arch.flagset) & (m);		\
622 	})
623 
624 /*
625  * Note that the set/clear accessors must be preempt-safe in order to
626  * avoid nesting them with load/put which also manipulate flags...
627  */
628 #ifdef __KVM_NVHE_HYPERVISOR__
629 /* the nVHE hypervisor is always non-preemptible */
630 #define __vcpu_flags_preempt_disable()
631 #define __vcpu_flags_preempt_enable()
632 #else
633 #define __vcpu_flags_preempt_disable()	preempt_disable()
634 #define __vcpu_flags_preempt_enable()	preempt_enable()
635 #endif
636 
637 #define __vcpu_set_flag(v, flagset, f, m)			\
638 	do {							\
639 		typeof(v->arch.flagset) *fset;			\
640 								\
641 		__build_check_flag(v, flagset, f, m);		\
642 								\
643 		fset = &v->arch.flagset;			\
644 		__vcpu_flags_preempt_disable();			\
645 		if (HWEIGHT(m) > 1)				\
646 			*fset &= ~(m);				\
647 		*fset |= (f);					\
648 		__vcpu_flags_preempt_enable();			\
649 	} while (0)
650 
651 #define __vcpu_clear_flag(v, flagset, f, m)			\
652 	do {							\
653 		typeof(v->arch.flagset) *fset;			\
654 								\
655 		__build_check_flag(v, flagset, f, m);		\
656 								\
657 		fset = &v->arch.flagset;			\
658 		__vcpu_flags_preempt_disable();			\
659 		*fset &= ~(m);					\
660 		__vcpu_flags_preempt_enable();			\
661 	} while (0)
662 
663 #define vcpu_get_flag(v, ...)	__vcpu_get_flag((v), __VA_ARGS__)
664 #define vcpu_set_flag(v, ...)	__vcpu_set_flag((v), __VA_ARGS__)
665 #define vcpu_clear_flag(v, ...)	__vcpu_clear_flag((v), __VA_ARGS__)
666 
667 /* SVE exposed to guest */
668 #define GUEST_HAS_SVE		__vcpu_single_flag(cflags, BIT(0))
669 /* SVE config completed */
670 #define VCPU_SVE_FINALIZED	__vcpu_single_flag(cflags, BIT(1))
671 /* PTRAUTH exposed to guest */
672 #define GUEST_HAS_PTRAUTH	__vcpu_single_flag(cflags, BIT(2))
673 
674 /* Exception pending */
675 #define PENDING_EXCEPTION	__vcpu_single_flag(iflags, BIT(0))
676 /*
677  * PC increment. Overlaps with EXCEPT_MASK on purpose so that it can't
678  * be set together with an exception...
679  */
680 #define INCREMENT_PC		__vcpu_single_flag(iflags, BIT(1))
681 /* Target EL/MODE (not a single flag, but let's abuse the macro) */
682 #define EXCEPT_MASK		__vcpu_single_flag(iflags, GENMASK(3, 1))
683 
684 /* Helpers to encode exceptions with minimum fuss */
685 #define __EXCEPT_MASK_VAL	unpack_vcpu_flag(EXCEPT_MASK)
686 #define __EXCEPT_SHIFT		__builtin_ctzl(__EXCEPT_MASK_VAL)
687 #define __vcpu_except_flags(_f)	iflags, (_f << __EXCEPT_SHIFT), __EXCEPT_MASK_VAL
688 
689 /*
690  * When PENDING_EXCEPTION is set, EXCEPT_MASK can take the following
691  * values:
692  *
693  * For AArch32 EL1:
694  */
695 #define EXCEPT_AA32_UND		__vcpu_except_flags(0)
696 #define EXCEPT_AA32_IABT	__vcpu_except_flags(1)
697 #define EXCEPT_AA32_DABT	__vcpu_except_flags(2)
698 /* For AArch64: */
699 #define EXCEPT_AA64_EL1_SYNC	__vcpu_except_flags(0)
700 #define EXCEPT_AA64_EL1_IRQ	__vcpu_except_flags(1)
701 #define EXCEPT_AA64_EL1_FIQ	__vcpu_except_flags(2)
702 #define EXCEPT_AA64_EL1_SERR	__vcpu_except_flags(3)
703 /* For AArch64 with NV: */
704 #define EXCEPT_AA64_EL2_SYNC	__vcpu_except_flags(4)
705 #define EXCEPT_AA64_EL2_IRQ	__vcpu_except_flags(5)
706 #define EXCEPT_AA64_EL2_FIQ	__vcpu_except_flags(6)
707 #define EXCEPT_AA64_EL2_SERR	__vcpu_except_flags(7)
708 /* Guest debug is live */
709 #define DEBUG_DIRTY		__vcpu_single_flag(iflags, BIT(4))
710 /* Save SPE context if active  */
711 #define DEBUG_STATE_SAVE_SPE	__vcpu_single_flag(iflags, BIT(5))
712 /* Save TRBE context if active  */
713 #define DEBUG_STATE_SAVE_TRBE	__vcpu_single_flag(iflags, BIT(6))
714 /* vcpu running in HYP context */
715 #define VCPU_HYP_CONTEXT	__vcpu_single_flag(iflags, BIT(7))
716 
717 /* SVE enabled for host EL0 */
718 #define HOST_SVE_ENABLED	__vcpu_single_flag(sflags, BIT(0))
719 /* SME enabled for EL0 */
720 #define HOST_SME_ENABLED	__vcpu_single_flag(sflags, BIT(1))
721 /* Physical CPU not in supported_cpus */
722 #define ON_UNSUPPORTED_CPU	__vcpu_single_flag(sflags, BIT(2))
723 /* WFIT instruction trapped */
724 #define IN_WFIT			__vcpu_single_flag(sflags, BIT(3))
725 /* vcpu system registers loaded on physical CPU */
726 #define SYSREGS_ON_CPU		__vcpu_single_flag(sflags, BIT(4))
727 /* Software step state is Active-pending */
728 #define DBG_SS_ACTIVE_PENDING	__vcpu_single_flag(sflags, BIT(5))
729 /* PMUSERENR for the guest EL0 is on physical CPU */
730 #define PMUSERENR_ON_CPU	__vcpu_single_flag(sflags, BIT(6))
731 /* WFI instruction trapped */
732 #define IN_WFI			__vcpu_single_flag(sflags, BIT(7))
733 
734 
735 /* Pointer to the vcpu's SVE FFR for sve_{save,load}_state() */
736 #define vcpu_sve_pffr(vcpu) (kern_hyp_va((vcpu)->arch.sve_state) +	\
737 			     sve_ffr_offset((vcpu)->arch.sve_max_vl))
738 
739 #define vcpu_sve_max_vq(vcpu)	sve_vq_from_vl((vcpu)->arch.sve_max_vl)
740 
741 #define vcpu_sve_state_size(vcpu) ({					\
742 	size_t __size_ret;						\
743 	unsigned int __vcpu_vq;						\
744 									\
745 	if (WARN_ON(!sve_vl_valid((vcpu)->arch.sve_max_vl))) {		\
746 		__size_ret = 0;						\
747 	} else {							\
748 		__vcpu_vq = vcpu_sve_max_vq(vcpu);			\
749 		__size_ret = SVE_SIG_REGS_SIZE(__vcpu_vq);		\
750 	}								\
751 									\
752 	__size_ret;							\
753 })
754 
755 #define KVM_GUESTDBG_VALID_MASK (KVM_GUESTDBG_ENABLE | \
756 				 KVM_GUESTDBG_USE_SW_BP | \
757 				 KVM_GUESTDBG_USE_HW | \
758 				 KVM_GUESTDBG_SINGLESTEP)
759 
760 #define vcpu_has_sve(vcpu) (system_supports_sve() &&			\
761 			    vcpu_get_flag(vcpu, GUEST_HAS_SVE))
762 
763 #ifdef CONFIG_ARM64_PTR_AUTH
764 #define vcpu_has_ptrauth(vcpu)						\
765 	((cpus_have_final_cap(ARM64_HAS_ADDRESS_AUTH) ||		\
766 	  cpus_have_final_cap(ARM64_HAS_GENERIC_AUTH)) &&		\
767 	  vcpu_get_flag(vcpu, GUEST_HAS_PTRAUTH))
768 #else
769 #define vcpu_has_ptrauth(vcpu)		false
770 #endif
771 
772 #define vcpu_on_unsupported_cpu(vcpu)					\
773 	vcpu_get_flag(vcpu, ON_UNSUPPORTED_CPU)
774 
775 #define vcpu_set_on_unsupported_cpu(vcpu)				\
776 	vcpu_set_flag(vcpu, ON_UNSUPPORTED_CPU)
777 
778 #define vcpu_clear_on_unsupported_cpu(vcpu)				\
779 	vcpu_clear_flag(vcpu, ON_UNSUPPORTED_CPU)
780 
781 #define vcpu_gp_regs(v)		(&(v)->arch.ctxt.regs)
782 
783 /*
784  * Only use __vcpu_sys_reg/ctxt_sys_reg if you know you want the
785  * memory backed version of a register, and not the one most recently
786  * accessed by a running VCPU.  For example, for userspace access or
787  * for system registers that are never context switched, but only
788  * emulated.
789  */
790 #define __ctxt_sys_reg(c,r)	(&(c)->sys_regs[(r)])
791 
792 #define ctxt_sys_reg(c,r)	(*__ctxt_sys_reg(c,r))
793 
794 #define __vcpu_sys_reg(v,r)	(ctxt_sys_reg(&(v)->arch.ctxt, (r)))
795 
796 u64 vcpu_read_sys_reg(const struct kvm_vcpu *vcpu, int reg);
797 void vcpu_write_sys_reg(struct kvm_vcpu *vcpu, u64 val, int reg);
798 
799 static inline bool __vcpu_read_sys_reg_from_cpu(int reg, u64 *val)
800 {
801 	/*
802 	 * *** VHE ONLY ***
803 	 *
804 	 * System registers listed in the switch are not saved on every
805 	 * exit from the guest but are only saved on vcpu_put.
806 	 *
807 	 * Note that MPIDR_EL1 for the guest is set by KVM via VMPIDR_EL2 but
808 	 * should never be listed below, because the guest cannot modify its
809 	 * own MPIDR_EL1 and MPIDR_EL1 is accessed for VCPU A from VCPU B's
810 	 * thread when emulating cross-VCPU communication.
811 	 */
812 	if (!has_vhe())
813 		return false;
814 
815 	switch (reg) {
816 	case SCTLR_EL1:		*val = read_sysreg_s(SYS_SCTLR_EL12);	break;
817 	case CPACR_EL1:		*val = read_sysreg_s(SYS_CPACR_EL12);	break;
818 	case TTBR0_EL1:		*val = read_sysreg_s(SYS_TTBR0_EL12);	break;
819 	case TTBR1_EL1:		*val = read_sysreg_s(SYS_TTBR1_EL12);	break;
820 	case TCR_EL1:		*val = read_sysreg_s(SYS_TCR_EL12);	break;
821 	case ESR_EL1:		*val = read_sysreg_s(SYS_ESR_EL12);	break;
822 	case AFSR0_EL1:		*val = read_sysreg_s(SYS_AFSR0_EL12);	break;
823 	case AFSR1_EL1:		*val = read_sysreg_s(SYS_AFSR1_EL12);	break;
824 	case FAR_EL1:		*val = read_sysreg_s(SYS_FAR_EL12);	break;
825 	case MAIR_EL1:		*val = read_sysreg_s(SYS_MAIR_EL12);	break;
826 	case VBAR_EL1:		*val = read_sysreg_s(SYS_VBAR_EL12);	break;
827 	case CONTEXTIDR_EL1:	*val = read_sysreg_s(SYS_CONTEXTIDR_EL12);break;
828 	case TPIDR_EL0:		*val = read_sysreg_s(SYS_TPIDR_EL0);	break;
829 	case TPIDRRO_EL0:	*val = read_sysreg_s(SYS_TPIDRRO_EL0);	break;
830 	case TPIDR_EL1:		*val = read_sysreg_s(SYS_TPIDR_EL1);	break;
831 	case AMAIR_EL1:		*val = read_sysreg_s(SYS_AMAIR_EL12);	break;
832 	case CNTKCTL_EL1:	*val = read_sysreg_s(SYS_CNTKCTL_EL12);	break;
833 	case ELR_EL1:		*val = read_sysreg_s(SYS_ELR_EL12);	break;
834 	case PAR_EL1:		*val = read_sysreg_par();		break;
835 	case DACR32_EL2:	*val = read_sysreg_s(SYS_DACR32_EL2);	break;
836 	case IFSR32_EL2:	*val = read_sysreg_s(SYS_IFSR32_EL2);	break;
837 	case DBGVCR32_EL2:	*val = read_sysreg_s(SYS_DBGVCR32_EL2);	break;
838 	default:		return false;
839 	}
840 
841 	return true;
842 }
843 
844 static inline bool __vcpu_write_sys_reg_to_cpu(u64 val, int reg)
845 {
846 	/*
847 	 * *** VHE ONLY ***
848 	 *
849 	 * System registers listed in the switch are not restored on every
850 	 * entry to the guest but are only restored on vcpu_load.
851 	 *
852 	 * Note that MPIDR_EL1 for the guest is set by KVM via VMPIDR_EL2 but
853 	 * should never be listed below, because the MPIDR should only be set
854 	 * once, before running the VCPU, and never changed later.
855 	 */
856 	if (!has_vhe())
857 		return false;
858 
859 	switch (reg) {
860 	case SCTLR_EL1:		write_sysreg_s(val, SYS_SCTLR_EL12);	break;
861 	case CPACR_EL1:		write_sysreg_s(val, SYS_CPACR_EL12);	break;
862 	case TTBR0_EL1:		write_sysreg_s(val, SYS_TTBR0_EL12);	break;
863 	case TTBR1_EL1:		write_sysreg_s(val, SYS_TTBR1_EL12);	break;
864 	case TCR_EL1:		write_sysreg_s(val, SYS_TCR_EL12);	break;
865 	case ESR_EL1:		write_sysreg_s(val, SYS_ESR_EL12);	break;
866 	case AFSR0_EL1:		write_sysreg_s(val, SYS_AFSR0_EL12);	break;
867 	case AFSR1_EL1:		write_sysreg_s(val, SYS_AFSR1_EL12);	break;
868 	case FAR_EL1:		write_sysreg_s(val, SYS_FAR_EL12);	break;
869 	case MAIR_EL1:		write_sysreg_s(val, SYS_MAIR_EL12);	break;
870 	case VBAR_EL1:		write_sysreg_s(val, SYS_VBAR_EL12);	break;
871 	case CONTEXTIDR_EL1:	write_sysreg_s(val, SYS_CONTEXTIDR_EL12);break;
872 	case TPIDR_EL0:		write_sysreg_s(val, SYS_TPIDR_EL0);	break;
873 	case TPIDRRO_EL0:	write_sysreg_s(val, SYS_TPIDRRO_EL0);	break;
874 	case TPIDR_EL1:		write_sysreg_s(val, SYS_TPIDR_EL1);	break;
875 	case AMAIR_EL1:		write_sysreg_s(val, SYS_AMAIR_EL12);	break;
876 	case CNTKCTL_EL1:	write_sysreg_s(val, SYS_CNTKCTL_EL12);	break;
877 	case ELR_EL1:		write_sysreg_s(val, SYS_ELR_EL12);	break;
878 	case PAR_EL1:		write_sysreg_s(val, SYS_PAR_EL1);	break;
879 	case DACR32_EL2:	write_sysreg_s(val, SYS_DACR32_EL2);	break;
880 	case IFSR32_EL2:	write_sysreg_s(val, SYS_IFSR32_EL2);	break;
881 	case DBGVCR32_EL2:	write_sysreg_s(val, SYS_DBGVCR32_EL2);	break;
882 	default:		return false;
883 	}
884 
885 	return true;
886 }
887 
888 struct kvm_vm_stat {
889 	struct kvm_vm_stat_generic generic;
890 };
891 
892 struct kvm_vcpu_stat {
893 	struct kvm_vcpu_stat_generic generic;
894 	u64 hvc_exit_stat;
895 	u64 wfe_exit_stat;
896 	u64 wfi_exit_stat;
897 	u64 mmio_exit_user;
898 	u64 mmio_exit_kernel;
899 	u64 signal_exits;
900 	u64 exits;
901 };
902 
903 void kvm_vcpu_preferred_target(struct kvm_vcpu_init *init);
904 unsigned long kvm_arm_num_regs(struct kvm_vcpu *vcpu);
905 int kvm_arm_copy_reg_indices(struct kvm_vcpu *vcpu, u64 __user *indices);
906 int kvm_arm_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
907 int kvm_arm_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
908 
909 unsigned long kvm_arm_num_sys_reg_descs(struct kvm_vcpu *vcpu);
910 int kvm_arm_copy_sys_reg_indices(struct kvm_vcpu *vcpu, u64 __user *uindices);
911 
912 int __kvm_arm_vcpu_get_events(struct kvm_vcpu *vcpu,
913 			      struct kvm_vcpu_events *events);
914 
915 int __kvm_arm_vcpu_set_events(struct kvm_vcpu *vcpu,
916 			      struct kvm_vcpu_events *events);
917 
918 #define KVM_ARCH_WANT_MMU_NOTIFIER
919 
920 void kvm_arm_halt_guest(struct kvm *kvm);
921 void kvm_arm_resume_guest(struct kvm *kvm);
922 
923 #define vcpu_has_run_once(vcpu)	!!rcu_access_pointer((vcpu)->pid)
924 
925 #ifndef __KVM_NVHE_HYPERVISOR__
926 #define kvm_call_hyp_nvhe(f, ...)						\
927 	({								\
928 		struct arm_smccc_res res;				\
929 									\
930 		arm_smccc_1_1_hvc(KVM_HOST_SMCCC_FUNC(f),		\
931 				  ##__VA_ARGS__, &res);			\
932 		WARN_ON(res.a0 != SMCCC_RET_SUCCESS);			\
933 									\
934 		res.a1;							\
935 	})
936 
937 /*
938  * The couple of isb() below are there to guarantee the same behaviour
939  * on VHE as on !VHE, where the eret to EL1 acts as a context
940  * synchronization event.
941  */
942 #define kvm_call_hyp(f, ...)						\
943 	do {								\
944 		if (has_vhe()) {					\
945 			f(__VA_ARGS__);					\
946 			isb();						\
947 		} else {						\
948 			kvm_call_hyp_nvhe(f, ##__VA_ARGS__);		\
949 		}							\
950 	} while(0)
951 
952 #define kvm_call_hyp_ret(f, ...)					\
953 	({								\
954 		typeof(f(__VA_ARGS__)) ret;				\
955 									\
956 		if (has_vhe()) {					\
957 			ret = f(__VA_ARGS__);				\
958 			isb();						\
959 		} else {						\
960 			ret = kvm_call_hyp_nvhe(f, ##__VA_ARGS__);	\
961 		}							\
962 									\
963 		ret;							\
964 	})
965 #else /* __KVM_NVHE_HYPERVISOR__ */
966 #define kvm_call_hyp(f, ...) f(__VA_ARGS__)
967 #define kvm_call_hyp_ret(f, ...) f(__VA_ARGS__)
968 #define kvm_call_hyp_nvhe(f, ...) f(__VA_ARGS__)
969 #endif /* __KVM_NVHE_HYPERVISOR__ */
970 
971 void force_vm_exit(const cpumask_t *mask);
972 
973 int handle_exit(struct kvm_vcpu *vcpu, int exception_index);
974 void handle_exit_early(struct kvm_vcpu *vcpu, int exception_index);
975 
976 int kvm_handle_cp14_load_store(struct kvm_vcpu *vcpu);
977 int kvm_handle_cp14_32(struct kvm_vcpu *vcpu);
978 int kvm_handle_cp14_64(struct kvm_vcpu *vcpu);
979 int kvm_handle_cp15_32(struct kvm_vcpu *vcpu);
980 int kvm_handle_cp15_64(struct kvm_vcpu *vcpu);
981 int kvm_handle_sys_reg(struct kvm_vcpu *vcpu);
982 int kvm_handle_cp10_id(struct kvm_vcpu *vcpu);
983 
984 void kvm_reset_sys_regs(struct kvm_vcpu *vcpu);
985 
986 int __init kvm_sys_reg_table_init(void);
987 
988 bool lock_all_vcpus(struct kvm *kvm);
989 void unlock_all_vcpus(struct kvm *kvm);
990 
991 /* MMIO helpers */
992 void kvm_mmio_write_buf(void *buf, unsigned int len, unsigned long data);
993 unsigned long kvm_mmio_read_buf(const void *buf, unsigned int len);
994 
995 int kvm_handle_mmio_return(struct kvm_vcpu *vcpu);
996 int io_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa);
997 
998 /*
999  * Returns true if a Performance Monitoring Interrupt (PMI), a.k.a. perf event,
1000  * arrived in guest context.  For arm64, any event that arrives while a vCPU is
1001  * loaded is considered to be "in guest".
1002  */
1003 static inline bool kvm_arch_pmi_in_guest(struct kvm_vcpu *vcpu)
1004 {
1005 	return IS_ENABLED(CONFIG_GUEST_PERF_EVENTS) && !!vcpu;
1006 }
1007 
1008 long kvm_hypercall_pv_features(struct kvm_vcpu *vcpu);
1009 gpa_t kvm_init_stolen_time(struct kvm_vcpu *vcpu);
1010 void kvm_update_stolen_time(struct kvm_vcpu *vcpu);
1011 
1012 bool kvm_arm_pvtime_supported(void);
1013 int kvm_arm_pvtime_set_attr(struct kvm_vcpu *vcpu,
1014 			    struct kvm_device_attr *attr);
1015 int kvm_arm_pvtime_get_attr(struct kvm_vcpu *vcpu,
1016 			    struct kvm_device_attr *attr);
1017 int kvm_arm_pvtime_has_attr(struct kvm_vcpu *vcpu,
1018 			    struct kvm_device_attr *attr);
1019 
1020 extern unsigned int __ro_after_init kvm_arm_vmid_bits;
1021 int __init kvm_arm_vmid_alloc_init(void);
1022 void __init kvm_arm_vmid_alloc_free(void);
1023 void kvm_arm_vmid_update(struct kvm_vmid *kvm_vmid);
1024 void kvm_arm_vmid_clear_active(void);
1025 
1026 static inline void kvm_arm_pvtime_vcpu_init(struct kvm_vcpu_arch *vcpu_arch)
1027 {
1028 	vcpu_arch->steal.base = INVALID_GPA;
1029 }
1030 
1031 static inline bool kvm_arm_is_pvtime_enabled(struct kvm_vcpu_arch *vcpu_arch)
1032 {
1033 	return (vcpu_arch->steal.base != INVALID_GPA);
1034 }
1035 
1036 void kvm_set_sei_esr(struct kvm_vcpu *vcpu, u64 syndrome);
1037 
1038 struct kvm_vcpu *kvm_mpidr_to_vcpu(struct kvm *kvm, unsigned long mpidr);
1039 
1040 DECLARE_KVM_HYP_PER_CPU(struct kvm_host_data, kvm_host_data);
1041 
1042 static inline void kvm_init_host_cpu_context(struct kvm_cpu_context *cpu_ctxt)
1043 {
1044 	/* The host's MPIDR is immutable, so let's set it up at boot time */
1045 	ctxt_sys_reg(cpu_ctxt, MPIDR_EL1) = read_cpuid_mpidr();
1046 }
1047 
1048 static inline bool kvm_system_needs_idmapped_vectors(void)
1049 {
1050 	return cpus_have_const_cap(ARM64_SPECTRE_V3A);
1051 }
1052 
1053 void kvm_arm_vcpu_ptrauth_trap(struct kvm_vcpu *vcpu);
1054 
1055 static inline void kvm_arch_sync_events(struct kvm *kvm) {}
1056 static inline void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu) {}
1057 
1058 void kvm_arm_init_debug(void);
1059 void kvm_arm_vcpu_init_debug(struct kvm_vcpu *vcpu);
1060 void kvm_arm_setup_debug(struct kvm_vcpu *vcpu);
1061 void kvm_arm_clear_debug(struct kvm_vcpu *vcpu);
1062 void kvm_arm_reset_debug_ptr(struct kvm_vcpu *vcpu);
1063 
1064 #define kvm_vcpu_os_lock_enabled(vcpu)		\
1065 	(!!(__vcpu_sys_reg(vcpu, OSLSR_EL1) & OSLSR_EL1_OSLK))
1066 
1067 int kvm_arm_vcpu_arch_set_attr(struct kvm_vcpu *vcpu,
1068 			       struct kvm_device_attr *attr);
1069 int kvm_arm_vcpu_arch_get_attr(struct kvm_vcpu *vcpu,
1070 			       struct kvm_device_attr *attr);
1071 int kvm_arm_vcpu_arch_has_attr(struct kvm_vcpu *vcpu,
1072 			       struct kvm_device_attr *attr);
1073 
1074 int kvm_vm_ioctl_mte_copy_tags(struct kvm *kvm,
1075 			       struct kvm_arm_copy_mte_tags *copy_tags);
1076 int kvm_vm_ioctl_set_counter_offset(struct kvm *kvm,
1077 				    struct kvm_arm_counter_offset *offset);
1078 
1079 /* Guest/host FPSIMD coordination helpers */
1080 int kvm_arch_vcpu_run_map_fp(struct kvm_vcpu *vcpu);
1081 void kvm_arch_vcpu_load_fp(struct kvm_vcpu *vcpu);
1082 void kvm_arch_vcpu_ctxflush_fp(struct kvm_vcpu *vcpu);
1083 void kvm_arch_vcpu_ctxsync_fp(struct kvm_vcpu *vcpu);
1084 void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu);
1085 void kvm_vcpu_unshare_task_fp(struct kvm_vcpu *vcpu);
1086 
1087 static inline bool kvm_pmu_counter_deferred(struct perf_event_attr *attr)
1088 {
1089 	return (!has_vhe() && attr->exclude_host);
1090 }
1091 
1092 /* Flags for host debug state */
1093 void kvm_arch_vcpu_load_debug_state_flags(struct kvm_vcpu *vcpu);
1094 void kvm_arch_vcpu_put_debug_state_flags(struct kvm_vcpu *vcpu);
1095 
1096 #ifdef CONFIG_KVM
1097 void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr);
1098 void kvm_clr_pmu_events(u32 clr);
1099 bool kvm_set_pmuserenr(u64 val);
1100 #else
1101 static inline void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr) {}
1102 static inline void kvm_clr_pmu_events(u32 clr) {}
1103 static inline bool kvm_set_pmuserenr(u64 val)
1104 {
1105 	return false;
1106 }
1107 #endif
1108 
1109 void kvm_vcpu_load_sysregs_vhe(struct kvm_vcpu *vcpu);
1110 void kvm_vcpu_put_sysregs_vhe(struct kvm_vcpu *vcpu);
1111 
1112 int __init kvm_set_ipa_limit(void);
1113 
1114 #define __KVM_HAVE_ARCH_VM_ALLOC
1115 struct kvm *kvm_arch_alloc_vm(void);
1116 
1117 static inline bool kvm_vm_is_protected(struct kvm *kvm)
1118 {
1119 	return false;
1120 }
1121 
1122 void kvm_init_protected_traps(struct kvm_vcpu *vcpu);
1123 
1124 int kvm_arm_vcpu_finalize(struct kvm_vcpu *vcpu, int feature);
1125 bool kvm_arm_vcpu_is_finalized(struct kvm_vcpu *vcpu);
1126 
1127 #define kvm_arm_vcpu_sve_finalized(vcpu) vcpu_get_flag(vcpu, VCPU_SVE_FINALIZED)
1128 
1129 #define kvm_has_mte(kvm)					\
1130 	(system_supports_mte() &&				\
1131 	 test_bit(KVM_ARCH_FLAG_MTE_ENABLED, &(kvm)->arch.flags))
1132 
1133 #define kvm_supports_32bit_el0()				\
1134 	(system_supports_32bit_el0() &&				\
1135 	 !static_branch_unlikely(&arm64_mismatched_32bit_el0))
1136 
1137 #define kvm_vm_has_ran_once(kvm)					\
1138 	(test_bit(KVM_ARCH_FLAG_HAS_RAN_ONCE, &(kvm)->arch.flags))
1139 
1140 int kvm_trng_call(struct kvm_vcpu *vcpu);
1141 #ifdef CONFIG_KVM
1142 extern phys_addr_t hyp_mem_base;
1143 extern phys_addr_t hyp_mem_size;
1144 void __init kvm_hyp_reserve(void);
1145 #else
1146 static inline void kvm_hyp_reserve(void) { }
1147 #endif
1148 
1149 void kvm_arm_vcpu_power_off(struct kvm_vcpu *vcpu);
1150 bool kvm_arm_vcpu_stopped(struct kvm_vcpu *vcpu);
1151 
1152 #endif /* __ARM64_KVM_HOST_H__ */
1153