xref: /linux/arch/arm64/include/asm/kvm_host.h (revision 3a2c4d55e32ad65efebdb6de44eef3bfa08bb49d)
1 /* SPDX-License-Identifier: GPL-2.0-only */
2 /*
3  * Copyright (C) 2012,2013 - ARM Ltd
4  * Author: Marc Zyngier <marc.zyngier@arm.com>
5  *
6  * Derived from arch/arm/include/asm/kvm_host.h:
7  * Copyright (C) 2012 - Virtual Open Systems and Columbia University
8  * Author: Christoffer Dall <c.dall@virtualopensystems.com>
9  */
10 
11 #ifndef __ARM64_KVM_HOST_H__
12 #define __ARM64_KVM_HOST_H__
13 
14 #include <linux/arm-smccc.h>
15 #include <linux/bitmap.h>
16 #include <linux/types.h>
17 #include <linux/jump_label.h>
18 #include <linux/kvm_types.h>
19 #include <linux/maple_tree.h>
20 #include <linux/percpu.h>
21 #include <linux/psci.h>
22 #include <asm/arch_gicv3.h>
23 #include <asm/barrier.h>
24 #include <asm/cpufeature.h>
25 #include <asm/cputype.h>
26 #include <asm/daifflags.h>
27 #include <asm/fpsimd.h>
28 #include <asm/kvm.h>
29 #include <asm/kvm_asm.h>
30 #include <asm/vncr_mapping.h>
31 
32 #define __KVM_HAVE_ARCH_INTC_INITIALIZED
33 
34 #define KVM_HALT_POLL_NS_DEFAULT 500000
35 
36 #include <kvm/arm_vgic.h>
37 #include <kvm/arm_arch_timer.h>
38 #include <kvm/arm_pmu.h>
39 
40 #define KVM_MAX_VCPUS VGIC_V3_MAX_CPUS
41 
42 #define KVM_VCPU_MAX_FEATURES 10
43 #define KVM_VCPU_VALID_FEATURES	(BIT(KVM_VCPU_MAX_FEATURES) - 1)
44 
45 #define KVM_REQ_SLEEP \
46 	KVM_ARCH_REQ_FLAGS(0, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
47 #define KVM_REQ_IRQ_PENDING		KVM_ARCH_REQ(1)
48 #define KVM_REQ_VCPU_RESET		KVM_ARCH_REQ(2)
49 #define KVM_REQ_RECORD_STEAL		KVM_ARCH_REQ(3)
50 #define KVM_REQ_RELOAD_GICv4		KVM_ARCH_REQ(4)
51 #define KVM_REQ_RELOAD_PMU		KVM_ARCH_REQ(5)
52 #define KVM_REQ_SUSPEND			KVM_ARCH_REQ(6)
53 #define KVM_REQ_RESYNC_PMU_EL0		KVM_ARCH_REQ(7)
54 #define KVM_REQ_NESTED_S2_UNMAP		KVM_ARCH_REQ(8)
55 #define KVM_REQ_GUEST_HYP_IRQ_PENDING	KVM_ARCH_REQ(9)
56 #define KVM_REQ_MAP_L1_VNCR_EL2		KVM_ARCH_REQ(10)
57 #define KVM_REQ_VGIC_PROCESS_UPDATE	KVM_ARCH_REQ(11)
58 
59 #define KVM_DIRTY_LOG_MANUAL_CAPS   (KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE | \
60 				     KVM_DIRTY_LOG_INITIALLY_SET)
61 
62 #define KVM_HAVE_MMU_RWLOCK
63 
64 /*
65  * Mode of operation configurable with kvm-arm.mode early param.
66  * See Documentation/admin-guide/kernel-parameters.txt for more information.
67  */
68 enum kvm_mode {
69 	KVM_MODE_DEFAULT,
70 	KVM_MODE_PROTECTED,
71 	KVM_MODE_NV,
72 	KVM_MODE_NONE,
73 };
74 #ifdef CONFIG_KVM
75 enum kvm_mode kvm_get_mode(void);
76 #else
77 static inline enum kvm_mode kvm_get_mode(void) { return KVM_MODE_NONE; };
78 #endif
79 
80 extern unsigned int __ro_after_init kvm_sve_max_vl;
81 extern unsigned int __ro_after_init kvm_host_sve_max_vl;
82 int __init kvm_arm_init_sve(void);
83 
84 u32 __attribute_const__ kvm_target_cpu(void);
85 void kvm_reset_vcpu(struct kvm_vcpu *vcpu);
86 void kvm_arm_vcpu_destroy(struct kvm_vcpu *vcpu);
87 
88 struct kvm_hyp_memcache {
89 	phys_addr_t head;
90 	unsigned long nr_pages;
91 	struct pkvm_mapping *mapping; /* only used from EL1 */
92 
93 #define	HYP_MEMCACHE_ACCOUNT_STAGE2	BIT(1)
94 	unsigned long flags;
95 };
96 
97 static inline void push_hyp_memcache(struct kvm_hyp_memcache *mc,
98 				     phys_addr_t *p,
99 				     phys_addr_t (*to_pa)(void *virt))
100 {
101 	*p = mc->head;
102 	mc->head = to_pa(p);
103 	mc->nr_pages++;
104 }
105 
106 static inline void *pop_hyp_memcache(struct kvm_hyp_memcache *mc,
107 				     void *(*to_va)(phys_addr_t phys))
108 {
109 	phys_addr_t *p = to_va(mc->head & PAGE_MASK);
110 
111 	if (!mc->nr_pages)
112 		return NULL;
113 
114 	mc->head = *p;
115 	mc->nr_pages--;
116 
117 	return p;
118 }
119 
120 static inline int __topup_hyp_memcache(struct kvm_hyp_memcache *mc,
121 				       unsigned long min_pages,
122 				       void *(*alloc_fn)(void *arg),
123 				       phys_addr_t (*to_pa)(void *virt),
124 				       void *arg)
125 {
126 	while (mc->nr_pages < min_pages) {
127 		phys_addr_t *p = alloc_fn(arg);
128 
129 		if (!p)
130 			return -ENOMEM;
131 		push_hyp_memcache(mc, p, to_pa);
132 	}
133 
134 	return 0;
135 }
136 
137 static inline void __free_hyp_memcache(struct kvm_hyp_memcache *mc,
138 				       void (*free_fn)(void *virt, void *arg),
139 				       void *(*to_va)(phys_addr_t phys),
140 				       void *arg)
141 {
142 	while (mc->nr_pages)
143 		free_fn(pop_hyp_memcache(mc, to_va), arg);
144 }
145 
146 void free_hyp_memcache(struct kvm_hyp_memcache *mc);
147 int topup_hyp_memcache(struct kvm_hyp_memcache *mc, unsigned long min_pages);
148 
149 struct kvm_vmid {
150 	atomic64_t id;
151 };
152 
153 struct kvm_s2_mmu {
154 	struct kvm_vmid vmid;
155 
156 	/*
157 	 * stage2 entry level table
158 	 *
159 	 * Two kvm_s2_mmu structures in the same VM can point to the same
160 	 * pgd here.  This happens when running a guest using a
161 	 * translation regime that isn't affected by its own stage-2
162 	 * translation, such as a non-VHE hypervisor running at vEL2, or
163 	 * for vEL1/EL0 with vHCR_EL2.VM == 0.  In that case, we use the
164 	 * canonical stage-2 page tables.
165 	 */
166 	phys_addr_t	pgd_phys;
167 	struct kvm_pgtable *pgt;
168 
169 	/*
170 	 * VTCR value used on the host. For a non-NV guest (or a NV
171 	 * guest that runs in a context where its own S2 doesn't
172 	 * apply), its T0SZ value reflects that of the IPA size.
173 	 *
174 	 * For a shadow S2 MMU, T0SZ reflects the PARange exposed to
175 	 * the guest.
176 	 */
177 	u64	vtcr;
178 
179 	/* The last vcpu id that ran on each physical CPU */
180 	int __percpu *last_vcpu_ran;
181 
182 #define KVM_ARM_EAGER_SPLIT_CHUNK_SIZE_DEFAULT 0
183 	/*
184 	 * Memory cache used to split
185 	 * KVM_CAP_ARM_EAGER_SPLIT_CHUNK_SIZE worth of huge pages. It
186 	 * is used to allocate stage2 page tables while splitting huge
187 	 * pages. The choice of KVM_CAP_ARM_EAGER_SPLIT_CHUNK_SIZE
188 	 * influences both the capacity of the split page cache, and
189 	 * how often KVM reschedules. Be wary of raising CHUNK_SIZE
190 	 * too high.
191 	 *
192 	 * Protected by kvm->slots_lock.
193 	 */
194 	struct kvm_mmu_memory_cache split_page_cache;
195 	uint64_t split_page_chunk_size;
196 
197 	struct kvm_arch *arch;
198 
199 	/*
200 	 * For a shadow stage-2 MMU, the virtual vttbr used by the
201 	 * host to parse the guest S2.
202 	 * This either contains:
203 	 * - the virtual VTTBR programmed by the guest hypervisor with
204 	 *   CnP cleared
205 	 * - The value 1 (VMID=0, BADDR=0, CnP=1) if invalid
206 	 *
207 	 * We also cache the full VTCR which gets used for TLB invalidation,
208 	 * taking the ARM ARM's "Any of the bits in VTCR_EL2 are permitted
209 	 * to be cached in a TLB" to the letter.
210 	 */
211 	u64	tlb_vttbr;
212 	u64	tlb_vtcr;
213 
214 	/*
215 	 * true when this represents a nested context where virtual
216 	 * HCR_EL2.VM == 1
217 	 */
218 	bool	nested_stage2_enabled;
219 
220 #ifdef CONFIG_PTDUMP_STAGE2_DEBUGFS
221 	struct dentry *shadow_pt_debugfs_dentry;
222 #endif
223 
224 	/*
225 	 * true when this MMU needs to be unmapped before being used for a new
226 	 * purpose.
227 	 */
228 	bool	pending_unmap;
229 
230 	/*
231 	 *  0: Nobody is currently using this, check vttbr for validity
232 	 * >0: Somebody is actively using this.
233 	 */
234 	atomic_t refcnt;
235 };
236 
237 struct kvm_arch_memory_slot {
238 };
239 
240 /**
241  * struct kvm_smccc_features: Descriptor of the hypercall services exposed to the guests
242  *
243  * @std_bmap: Bitmap of standard secure service calls
244  * @std_hyp_bmap: Bitmap of standard hypervisor service calls
245  * @vendor_hyp_bmap: Bitmap of vendor specific hypervisor service calls
246  */
247 struct kvm_smccc_features {
248 	unsigned long std_bmap;
249 	unsigned long std_hyp_bmap;
250 	unsigned long vendor_hyp_bmap; /* Function numbers 0-63 */
251 	unsigned long vendor_hyp_bmap_2; /* Function numbers 64-127 */
252 };
253 
254 typedef u16 pkvm_handle_t;
255 
256 struct kvm_protected_vm {
257 	pkvm_handle_t handle;
258 	struct kvm_hyp_memcache teardown_mc;
259 	struct kvm_hyp_memcache stage2_teardown_mc;
260 	bool is_protected;
261 	bool is_created;
262 
263 	/*
264 	 * True when the guest is being torn down. When in this state, the
265 	 * guest's vCPUs can't be loaded anymore, but its pages can be
266 	 * reclaimed by the host.
267 	 */
268 	bool is_dying;
269 };
270 
271 struct kvm_mpidr_data {
272 	u64			mpidr_mask;
273 	DECLARE_FLEX_ARRAY(u16, cmpidr_to_idx);
274 };
275 
276 static inline u16 kvm_mpidr_index(struct kvm_mpidr_data *data, u64 mpidr)
277 {
278 	unsigned long index = 0, mask = data->mpidr_mask;
279 	unsigned long aff = mpidr & MPIDR_HWID_BITMASK;
280 
281 	bitmap_gather(&index, &aff, &mask, fls(mask));
282 
283 	return index;
284 }
285 
286 struct kvm_sysreg_masks;
287 
288 enum fgt_group_id {
289 	__NO_FGT_GROUP__,
290 	HFGRTR_GROUP,
291 	HFGWTR_GROUP = HFGRTR_GROUP,
292 	HDFGRTR_GROUP,
293 	HDFGWTR_GROUP = HDFGRTR_GROUP,
294 	HFGITR_GROUP,
295 	HAFGRTR_GROUP,
296 	HFGRTR2_GROUP,
297 	HFGWTR2_GROUP = HFGRTR2_GROUP,
298 	HDFGRTR2_GROUP,
299 	HDFGWTR2_GROUP = HDFGRTR2_GROUP,
300 	HFGITR2_GROUP,
301 	ICH_HFGRTR_GROUP,
302 	ICH_HFGWTR_GROUP = ICH_HFGRTR_GROUP,
303 	ICH_HFGITR_GROUP,
304 
305 	/* Must be last */
306 	__NR_FGT_GROUP_IDS__
307 };
308 
309 struct kvm_arch {
310 	struct kvm_s2_mmu mmu;
311 
312 	/*
313 	 * Fine-Grained UNDEF, mimicking the FGT layout defined by the
314 	 * architecture. We track them globally, as we present the
315 	 * same feature-set to all vcpus.
316 	 *
317 	 * Index 0 is currently spare.
318 	 */
319 	u64 fgu[__NR_FGT_GROUP_IDS__];
320 
321 	/*
322 	 * Stage 2 paging state for VMs with nested S2 using a virtual
323 	 * VMID.
324 	 */
325 	struct kvm_s2_mmu *nested_mmus;
326 	size_t nested_mmus_size;
327 	int nested_mmus_next;
328 
329 	/* Interrupt controller */
330 	struct vgic_dist	vgic;
331 
332 	/* Timers */
333 	struct arch_timer_vm_data timer_data;
334 
335 	/* Mandated version of PSCI */
336 	u32 psci_version;
337 
338 	/* Protects VM-scoped configuration data */
339 	struct mutex config_lock;
340 
341 	/*
342 	 * If we encounter a data abort without valid instruction syndrome
343 	 * information, report this to user space.  User space can (and
344 	 * should) opt in to this feature if KVM_CAP_ARM_NISV_TO_USER is
345 	 * supported.
346 	 */
347 #define KVM_ARCH_FLAG_RETURN_NISV_IO_ABORT_TO_USER	0
348 	/* Memory Tagging Extension enabled for the guest */
349 #define KVM_ARCH_FLAG_MTE_ENABLED			1
350 	/* At least one vCPU has ran in the VM */
351 #define KVM_ARCH_FLAG_HAS_RAN_ONCE			2
352 	/* The vCPU feature set for the VM is configured */
353 #define KVM_ARCH_FLAG_VCPU_FEATURES_CONFIGURED		3
354 	/* PSCI SYSTEM_SUSPEND enabled for the guest */
355 #define KVM_ARCH_FLAG_SYSTEM_SUSPEND_ENABLED		4
356 	/* VM counter offset */
357 #define KVM_ARCH_FLAG_VM_COUNTER_OFFSET			5
358 	/* Timer PPIs made immutable */
359 #define KVM_ARCH_FLAG_TIMER_PPIS_IMMUTABLE		6
360 	/* Initial ID reg values loaded */
361 #define KVM_ARCH_FLAG_ID_REGS_INITIALIZED		7
362 	/* Fine-Grained UNDEF initialised */
363 #define KVM_ARCH_FLAG_FGU_INITIALIZED			8
364 	/* SVE exposed to guest */
365 #define KVM_ARCH_FLAG_GUEST_HAS_SVE			9
366 	/* MIDR_EL1, REVIDR_EL1, and AIDR_EL1 are writable from userspace */
367 #define KVM_ARCH_FLAG_WRITABLE_IMP_ID_REGS		10
368 	/* Unhandled SEAs are taken to userspace */
369 #define KVM_ARCH_FLAG_EXIT_SEA				11
370 	unsigned long flags;
371 
372 	/* VM-wide vCPU feature set */
373 	DECLARE_BITMAP(vcpu_features, KVM_VCPU_MAX_FEATURES);
374 
375 	/* MPIDR to vcpu index mapping, optional */
376 	struct kvm_mpidr_data *mpidr_data;
377 
378 	/*
379 	 * VM-wide PMU filter, implemented as a bitmap and big enough for
380 	 * up to 2^10 events (ARMv8.0) or 2^16 events (ARMv8.1+).
381 	 */
382 	unsigned long *pmu_filter;
383 	struct arm_pmu *arm_pmu;
384 
385 	cpumask_var_t supported_cpus;
386 
387 	/* Maximum number of counters for the guest */
388 	u8 nr_pmu_counters;
389 
390 	/* PMMIR_EL1.SLOTS value exposed to the guest. */
391 	u8 pmmir_slots;
392 
393 	/* Hypercall features firmware registers' descriptor */
394 	struct kvm_smccc_features smccc_feat;
395 	struct maple_tree smccc_filter;
396 
397 	/*
398 	 * Emulated CPU ID registers per VM
399 	 * (Op0, Op1, CRn, CRm, Op2) of the ID registers to be saved in it
400 	 * is (3, 0, 0, crm, op2), where 1<=crm<8, 0<=op2<8.
401 	 *
402 	 * These emulated idregs are VM-wide, but accessed from the context of a vCPU.
403 	 * Atomic access to multiple idregs are guarded by kvm_arch.config_lock.
404 	 */
405 #define IDREG_IDX(id)		(((sys_reg_CRm(id) - 1) << 3) | sys_reg_Op2(id))
406 #define KVM_ARM_ID_REG_NUM	(IDREG_IDX(sys_reg(3, 0, 0, 7, 7)) + 1)
407 	u64 id_regs[KVM_ARM_ID_REG_NUM];
408 
409 	u64 midr_el1;
410 	u64 revidr_el1;
411 	u64 aidr_el1;
412 	u64 ctr_el0;
413 
414 	/* Masks for VNCR-backed and general EL2 sysregs */
415 	struct kvm_sysreg_masks	*sysreg_masks;
416 
417 	/* Count the number of VNCR_EL2 TLBs */
418 	atomic_t vncr_tlb_count;
419 
420 	/*
421 	 * For an untrusted host VM, 'pkvm.handle' is used to lookup
422 	 * the associated pKVM instance in the hypervisor.
423 	 */
424 	struct kvm_protected_vm pkvm;
425 
426 #ifdef CONFIG_PTDUMP_STAGE2_DEBUGFS
427 	/* Nested virtualization info */
428 	struct dentry *debugfs_nv_dentry;
429 #endif
430 };
431 
432 struct kvm_vcpu_fault_info {
433 	u64 esr_el2;		/* Hyp Syndrom Register */
434 	u64 far_el2;		/* Hyp Fault Address Register */
435 	u64 hpfar_el2;		/* Hyp IPA Fault Address Register */
436 	u64 disr_el1;		/* Deferred [SError] Status Register */
437 };
438 
439 /*
440  * VNCR() just places the VNCR_capable registers in the enum after
441  * __VNCR_START__, and the value (after correction) to be an 8-byte offset
442  * from the VNCR base. As we don't require the enum to be otherwise ordered,
443  * we need the terrible hack below to ensure that we correctly size the
444  * sys_regs array, no matter what.
445  *
446  * The __MAX__ macro has been lifted from Sean Eron Anderson's wonderful
447  * treasure trove of bit hacks:
448  * https://graphics.stanford.edu/~seander/bithacks.html#IntegerMinOrMax
449  */
450 #define __MAX__(x,y)	((x) ^ (((x) ^ (y)) & -((x) < (y))))
451 #define VNCR(r)						\
452 	__before_##r,					\
453 	r = __VNCR_START__ + ((VNCR_ ## r) / 8),	\
454 	__after_##r = __MAX__(__before_##r - 1, r)
455 
456 enum vcpu_sysreg {
457 	__INVALID_SYSREG__,   /* 0 is reserved as an invalid value */
458 	MPIDR_EL1,	/* MultiProcessor Affinity Register */
459 	CLIDR_EL1,	/* Cache Level ID Register */
460 	CSSELR_EL1,	/* Cache Size Selection Register */
461 	TPIDR_EL0,	/* Thread ID, User R/W */
462 	TPIDRRO_EL0,	/* Thread ID, User R/O */
463 	TPIDR_EL1,	/* Thread ID, Privileged */
464 	CNTKCTL_EL1,	/* Timer Control Register (EL1) */
465 	PAR_EL1,	/* Physical Address Register */
466 	MDCCINT_EL1,	/* Monitor Debug Comms Channel Interrupt Enable Reg */
467 	OSLSR_EL1,	/* OS Lock Status Register */
468 	DISR_EL1,	/* Deferred Interrupt Status Register */
469 
470 	/* Performance Monitors Registers */
471 	PMCR_EL0,	/* Control Register */
472 	PMSELR_EL0,	/* Event Counter Selection Register */
473 	PMEVCNTR0_EL0,	/* Event Counter Register (0-30) */
474 	PMEVCNTR30_EL0 = PMEVCNTR0_EL0 + 30,
475 	PMCCNTR_EL0,	/* Cycle Counter Register */
476 	PMEVTYPER0_EL0,	/* Event Type Register (0-30) */
477 	PMEVTYPER30_EL0 = PMEVTYPER0_EL0 + 30,
478 	PMCCFILTR_EL0,	/* Cycle Count Filter Register */
479 	PMCNTENSET_EL0,	/* Count Enable Set Register */
480 	PMINTENSET_EL1,	/* Interrupt Enable Set Register */
481 	PMOVSSET_EL0,	/* Overflow Flag Status Set Register */
482 	PMUSERENR_EL0,	/* User Enable Register */
483 
484 	/* Pointer Authentication Registers in a strict increasing order. */
485 	APIAKEYLO_EL1,
486 	APIAKEYHI_EL1,
487 	APIBKEYLO_EL1,
488 	APIBKEYHI_EL1,
489 	APDAKEYLO_EL1,
490 	APDAKEYHI_EL1,
491 	APDBKEYLO_EL1,
492 	APDBKEYHI_EL1,
493 	APGAKEYLO_EL1,
494 	APGAKEYHI_EL1,
495 
496 	/* Memory Tagging Extension registers */
497 	RGSR_EL1,	/* Random Allocation Tag Seed Register */
498 	GCR_EL1,	/* Tag Control Register */
499 	TFSRE0_EL1,	/* Tag Fault Status Register (EL0) */
500 
501 	POR_EL0,	/* Permission Overlay Register 0 (EL0) */
502 
503 	/* FP/SIMD/SVE */
504 	SVCR,
505 	FPMR,
506 
507 	/* 32bit specific registers. */
508 	DACR32_EL2,	/* Domain Access Control Register */
509 	IFSR32_EL2,	/* Instruction Fault Status Register */
510 	FPEXC32_EL2,	/* Floating-Point Exception Control Register */
511 	DBGVCR32_EL2,	/* Debug Vector Catch Register */
512 
513 	/* EL2 registers */
514 	ACTLR_EL2,	/* Auxiliary Control Register (EL2) */
515 	CPTR_EL2,	/* Architectural Feature Trap Register (EL2) */
516 	HACR_EL2,	/* Hypervisor Auxiliary Control Register */
517 	TTBR0_EL2,	/* Translation Table Base Register 0 (EL2) */
518 	TTBR1_EL2,	/* Translation Table Base Register 1 (EL2) */
519 	TCR_EL2,	/* Translation Control Register (EL2) */
520 	PIRE0_EL2,	/* Permission Indirection Register 0 (EL2) */
521 	PIR_EL2,	/* Permission Indirection Register 1 (EL2) */
522 	POR_EL2,	/* Permission Overlay Register 2 (EL2) */
523 	SPSR_EL2,	/* EL2 saved program status register */
524 	ELR_EL2,	/* EL2 exception link register */
525 	AFSR0_EL2,	/* Auxiliary Fault Status Register 0 (EL2) */
526 	AFSR1_EL2,	/* Auxiliary Fault Status Register 1 (EL2) */
527 	ESR_EL2,	/* Exception Syndrome Register (EL2) */
528 	FAR_EL2,	/* Fault Address Register (EL2) */
529 	HPFAR_EL2,	/* Hypervisor IPA Fault Address Register */
530 	MAIR_EL2,	/* Memory Attribute Indirection Register (EL2) */
531 	AMAIR_EL2,	/* Auxiliary Memory Attribute Indirection Register (EL2) */
532 	VBAR_EL2,	/* Vector Base Address Register (EL2) */
533 	RVBAR_EL2,	/* Reset Vector Base Address Register */
534 	CONTEXTIDR_EL2,	/* Context ID Register (EL2) */
535 	SP_EL2,		/* EL2 Stack Pointer */
536 	CNTHP_CTL_EL2,
537 	CNTHP_CVAL_EL2,
538 	CNTHV_CTL_EL2,
539 	CNTHV_CVAL_EL2,
540 
541 	/* Anything from this can be RES0/RES1 sanitised */
542 	MARKER(__SANITISED_REG_START__),
543 	SCTLR_EL2,	/* System Control Register (EL2) */
544 	TCR2_EL2,	/* Extended Translation Control Register (EL2) */
545 	SCTLR2_EL2,	/* System Control Register 2 (EL2) */
546 	MDCR_EL2,	/* Monitor Debug Configuration Register (EL2) */
547 	CNTHCTL_EL2,	/* Counter-timer Hypervisor Control register */
548 	ZCR_EL2,	/* SVE Control Register (EL2) */
549 	HCR_EL2,	/* Hypervisor Control Register */
550 
551 	/* Any VNCR-capable reg goes after this point */
552 	MARKER(__VNCR_START__),
553 
554 	VNCR(SCTLR_EL1),/* System Control Register */
555 	VNCR(ACTLR_EL1),/* Auxiliary Control Register */
556 	VNCR(CPACR_EL1),/* Coprocessor Access Control */
557 	VNCR(ZCR_EL1),	/* SVE Control */
558 	VNCR(TTBR0_EL1),/* Translation Table Base Register 0 */
559 	VNCR(TTBR1_EL1),/* Translation Table Base Register 1 */
560 	VNCR(TCR_EL1),	/* Translation Control Register */
561 	VNCR(TCR2_EL1),	/* Extended Translation Control Register */
562 	VNCR(SCTLR2_EL1), /* System Control Register 2 */
563 	VNCR(ESR_EL1),	/* Exception Syndrome Register */
564 	VNCR(AFSR0_EL1),/* Auxiliary Fault Status Register 0 */
565 	VNCR(AFSR1_EL1),/* Auxiliary Fault Status Register 1 */
566 	VNCR(FAR_EL1),	/* Fault Address Register */
567 	VNCR(MAIR_EL1),	/* Memory Attribute Indirection Register */
568 	VNCR(VBAR_EL1),	/* Vector Base Address Register */
569 	VNCR(CONTEXTIDR_EL1),	/* Context ID Register */
570 	VNCR(AMAIR_EL1),/* Aux Memory Attribute Indirection Register */
571 	VNCR(MDSCR_EL1),/* Monitor Debug System Control Register */
572 	VNCR(ELR_EL1),
573 	VNCR(SP_EL1),
574 	VNCR(SPSR_EL1),
575 	VNCR(TFSR_EL1),	/* Tag Fault Status Register (EL1) */
576 	VNCR(VPIDR_EL2),/* Virtualization Processor ID Register */
577 	VNCR(VMPIDR_EL2),/* Virtualization Multiprocessor ID Register */
578 	VNCR(NVHCR_EL2),/* NV Hypervisor Configuration Register */
579 	VNCR(HSTR_EL2),	/* Hypervisor System Trap Register */
580 	VNCR(VTTBR_EL2),/* Virtualization Translation Table Base Register */
581 	VNCR(VTCR_EL2),	/* Virtualization Translation Control Register */
582 	VNCR(TPIDR_EL2),/* EL2 Software Thread ID Register */
583 	VNCR(HCRX_EL2),	/* Extended Hypervisor Configuration Register */
584 
585 	/* Permission Indirection Extension registers */
586 	VNCR(PIR_EL1),	 /* Permission Indirection Register 1 (EL1) */
587 	VNCR(PIRE0_EL1), /*  Permission Indirection Register 0 (EL1) */
588 
589 	VNCR(POR_EL1),	/* Permission Overlay Register 1 (EL1) */
590 
591 	/* FEAT_RAS registers */
592 	VNCR(VDISR_EL2),
593 	VNCR(VSESR_EL2),
594 
595 	VNCR(HFGRTR_EL2),
596 	VNCR(HFGWTR_EL2),
597 	VNCR(HFGITR_EL2),
598 	VNCR(HDFGRTR_EL2),
599 	VNCR(HDFGWTR_EL2),
600 	VNCR(HAFGRTR_EL2),
601 	VNCR(HFGRTR2_EL2),
602 	VNCR(HFGWTR2_EL2),
603 	VNCR(HFGITR2_EL2),
604 	VNCR(HDFGRTR2_EL2),
605 	VNCR(HDFGWTR2_EL2),
606 
607 	VNCR(VNCR_EL2),
608 
609 	VNCR(CNTVOFF_EL2),
610 	VNCR(CNTV_CVAL_EL0),
611 	VNCR(CNTV_CTL_EL0),
612 	VNCR(CNTP_CVAL_EL0),
613 	VNCR(CNTP_CTL_EL0),
614 
615 	VNCR(ICH_LR0_EL2),
616 	VNCR(ICH_LR1_EL2),
617 	VNCR(ICH_LR2_EL2),
618 	VNCR(ICH_LR3_EL2),
619 	VNCR(ICH_LR4_EL2),
620 	VNCR(ICH_LR5_EL2),
621 	VNCR(ICH_LR6_EL2),
622 	VNCR(ICH_LR7_EL2),
623 	VNCR(ICH_LR8_EL2),
624 	VNCR(ICH_LR9_EL2),
625 	VNCR(ICH_LR10_EL2),
626 	VNCR(ICH_LR11_EL2),
627 	VNCR(ICH_LR12_EL2),
628 	VNCR(ICH_LR13_EL2),
629 	VNCR(ICH_LR14_EL2),
630 	VNCR(ICH_LR15_EL2),
631 
632 	VNCR(ICH_AP0R0_EL2),
633 	VNCR(ICH_AP0R1_EL2),
634 	VNCR(ICH_AP0R2_EL2),
635 	VNCR(ICH_AP0R3_EL2),
636 	VNCR(ICH_AP1R0_EL2),
637 	VNCR(ICH_AP1R1_EL2),
638 	VNCR(ICH_AP1R2_EL2),
639 	VNCR(ICH_AP1R3_EL2),
640 	VNCR(ICH_HCR_EL2),
641 	VNCR(ICH_VMCR_EL2),
642 
643 	VNCR(ICH_HFGRTR_EL2),
644 	VNCR(ICH_HFGWTR_EL2),
645 	VNCR(ICH_HFGITR_EL2),
646 
647 	NR_SYS_REGS	/* Nothing after this line! */
648 };
649 
650 struct resx {
651 	u64	res0;
652 	u64	res1;
653 };
654 
655 struct kvm_sysreg_masks {
656 	struct resx mask[NR_SYS_REGS - __SANITISED_REG_START__];
657 };
658 
659 static inline struct resx __kvm_get_sysreg_resx(struct kvm_arch *arch,
660 						enum vcpu_sysreg sr)
661 {
662 	struct kvm_sysreg_masks *masks;
663 
664 	masks = arch->sysreg_masks;
665 	if (likely(masks &&
666 		   sr >= __SANITISED_REG_START__ && sr < NR_SYS_REGS))
667 		return masks->mask[sr - __SANITISED_REG_START__];
668 
669 	return (struct resx){};
670 }
671 
672 #define kvm_get_sysreg_resx(k, sr) __kvm_get_sysreg_resx(&(k)->arch, (sr))
673 
674 static inline void __kvm_set_sysreg_resx(struct kvm_arch *arch,
675 					 enum vcpu_sysreg sr, struct resx resx)
676 {
677 	arch->sysreg_masks->mask[sr - __SANITISED_REG_START__] = resx;
678 }
679 
680 #define kvm_set_sysreg_resx(k, sr, resx)		\
681 	__kvm_set_sysreg_resx(&(k)->arch, (sr), (resx))
682 
683 struct fgt_masks {
684 	const char	*str;
685 	u64		mask;
686 	u64		nmask;
687 	u64		res0;
688 	u64		res1;
689 };
690 
691 extern struct fgt_masks hfgrtr_masks;
692 extern struct fgt_masks hfgwtr_masks;
693 extern struct fgt_masks hfgitr_masks;
694 extern struct fgt_masks hdfgrtr_masks;
695 extern struct fgt_masks hdfgwtr_masks;
696 extern struct fgt_masks hafgrtr_masks;
697 extern struct fgt_masks hfgrtr2_masks;
698 extern struct fgt_masks hfgwtr2_masks;
699 extern struct fgt_masks hfgitr2_masks;
700 extern struct fgt_masks hdfgrtr2_masks;
701 extern struct fgt_masks hdfgwtr2_masks;
702 extern struct fgt_masks ich_hfgrtr_masks;
703 extern struct fgt_masks ich_hfgwtr_masks;
704 extern struct fgt_masks ich_hfgitr_masks;
705 
706 extern struct fgt_masks kvm_nvhe_sym(hfgrtr_masks);
707 extern struct fgt_masks kvm_nvhe_sym(hfgwtr_masks);
708 extern struct fgt_masks kvm_nvhe_sym(hfgitr_masks);
709 extern struct fgt_masks kvm_nvhe_sym(hdfgrtr_masks);
710 extern struct fgt_masks kvm_nvhe_sym(hdfgwtr_masks);
711 extern struct fgt_masks kvm_nvhe_sym(hafgrtr_masks);
712 extern struct fgt_masks kvm_nvhe_sym(hfgrtr2_masks);
713 extern struct fgt_masks kvm_nvhe_sym(hfgwtr2_masks);
714 extern struct fgt_masks kvm_nvhe_sym(hfgitr2_masks);
715 extern struct fgt_masks kvm_nvhe_sym(hdfgrtr2_masks);
716 extern struct fgt_masks kvm_nvhe_sym(hdfgwtr2_masks);
717 extern struct fgt_masks kvm_nvhe_sym(ich_hfgrtr_masks);
718 extern struct fgt_masks kvm_nvhe_sym(ich_hfgwtr_masks);
719 extern struct fgt_masks kvm_nvhe_sym(ich_hfgitr_masks);
720 
721 struct kvm_cpu_context {
722 	struct user_pt_regs regs;	/* sp = sp_el0 */
723 
724 	u64	spsr_abt;
725 	u64	spsr_und;
726 	u64	spsr_irq;
727 	u64	spsr_fiq;
728 
729 	struct user_fpsimd_state fp_regs;
730 
731 	u64 sys_regs[NR_SYS_REGS];
732 
733 	struct kvm_vcpu *__hyp_running_vcpu;
734 
735 	/* This pointer has to be 4kB aligned. */
736 	u64 *vncr_array;
737 };
738 
739 /*
740  * This structure is instantiated on a per-CPU basis, and contains
741  * data that is:
742  *
743  * - tied to a single physical CPU, and
744  * - either have a lifetime that does not extend past vcpu_put()
745  * - or is an invariant for the lifetime of the system
746  *
747  * Use host_data_ptr(field) as a way to access a pointer to such a
748  * field.
749  */
750 struct kvm_host_data {
751 #define KVM_HOST_DATA_FLAG_HAS_SPE			0
752 #define KVM_HOST_DATA_FLAG_HAS_TRBE			1
753 #define KVM_HOST_DATA_FLAG_TRBE_ENABLED			2
754 #define KVM_HOST_DATA_FLAG_EL1_TRACING_CONFIGURED	3
755 #define KVM_HOST_DATA_FLAG_VCPU_IN_HYP_CONTEXT		4
756 #define KVM_HOST_DATA_FLAG_L1_VNCR_MAPPED		5
757 #define KVM_HOST_DATA_FLAG_HAS_BRBE			6
758 	unsigned long flags;
759 
760 	struct kvm_cpu_context host_ctxt;
761 
762 	/*
763 	 * Hyp VA.
764 	 * sve_regs is only used in pKVM and if system_supports_sve().
765 	 */
766 	struct arm64_sve_state *sve_regs;
767 
768 	/* Ownership of the FP regs */
769 	enum {
770 		FP_STATE_FREE,
771 		FP_STATE_HOST_OWNED,
772 		FP_STATE_GUEST_OWNED,
773 	} fp_owner;
774 
775 	/*
776 	 * host_debug_state contains the host registers which are
777 	 * saved and restored during world switches.
778 	 */
779 	struct {
780 		/* {Break,watch}point registers */
781 		struct kvm_guest_debug_arch regs;
782 		/* Statistical profiling extension */
783 		u64 pmscr_el1;
784 		u64 pmblimitr_el1;
785 		/* Self-hosted trace */
786 		u64 trfcr_el1;
787 		u64 trblimitr_el1;
788 		/* Values of trap registers for the host before guest entry. */
789 		u64 mdcr_el2;
790 		u64 brbcr_el1;
791 	} host_debug_state;
792 
793 	/* Guest trace filter value */
794 	u64 trfcr_while_in_guest;
795 
796 	/* Number of programmable event counters (PMCR_EL0.N) for this CPU */
797 	unsigned int nr_event_counters;
798 
799 	/* Number of debug breakpoints/watchpoints for this CPU (minus 1) */
800 	unsigned int debug_brps;
801 	unsigned int debug_wrps;
802 
803 	/* Last vgic_irq part of the AP list recorded in an LR */
804 	struct vgic_irq *last_lr_irq;
805 
806 	/* PPI state tracking for GICv5-based guests */
807 	struct {
808 		DECLARE_BITMAP(pendr, VGIC_V5_NR_PRIVATE_IRQS);
809 
810 		/* The saved state of the regs when leaving the guest */
811 		DECLARE_BITMAP(activer_exit, VGIC_V5_NR_PRIVATE_IRQS);
812 	} vgic_v5_ppi_state;
813 };
814 
815 struct kvm_host_psci_config {
816 	/* PSCI version used by host. */
817 	u32 version;
818 	u32 smccc_version;
819 
820 	/* Function IDs used by host if version is v0.1. */
821 	struct psci_0_1_function_ids function_ids_0_1;
822 
823 	bool psci_0_1_cpu_suspend_implemented;
824 	bool psci_0_1_cpu_on_implemented;
825 	bool psci_0_1_cpu_off_implemented;
826 	bool psci_0_1_migrate_implemented;
827 };
828 
829 extern struct kvm_host_psci_config kvm_nvhe_sym(kvm_host_psci_config);
830 #define kvm_host_psci_config CHOOSE_NVHE_SYM(kvm_host_psci_config)
831 
832 extern s64 kvm_nvhe_sym(hyp_physvirt_offset);
833 #define hyp_physvirt_offset CHOOSE_NVHE_SYM(hyp_physvirt_offset)
834 
835 extern u64 kvm_nvhe_sym(hyp_cpu_logical_map)[NR_CPUS];
836 #define hyp_cpu_logical_map CHOOSE_NVHE_SYM(hyp_cpu_logical_map)
837 
838 struct vcpu_reset_state {
839 	unsigned long	pc;
840 	unsigned long	r0;
841 	bool		be;
842 	bool		reset;
843 };
844 
845 struct vncr_tlb;
846 
847 struct kvm_vcpu_arch {
848 	struct kvm_cpu_context ctxt;
849 
850 	/*
851 	 * Guest floating point state
852 	 *
853 	 * The architecture has two main floating point extensions,
854 	 * the original FPSIMD and SVE.  These have overlapping
855 	 * register views, with the FPSIMD V registers occupying the
856 	 * low 128 bits of the SVE Z registers.  When the core
857 	 * floating point code saves the register state of a task it
858 	 * records which view it saved in fp_type.
859 	 */
860 	struct arm64_sve_state *sve_state;
861 	enum fp_type fp_type;
862 	unsigned int sve_max_vl;
863 
864 	/* Stage 2 paging state used by the hardware on next switch */
865 	struct kvm_s2_mmu *hw_mmu;
866 
867 	/* Values of trap registers for the guest. */
868 	u64 hcr_el2;
869 	u64 hcrx_el2;
870 	u64 mdcr_el2;
871 
872 	struct {
873 		u64 r;
874 		u64 w;
875 	} fgt[__NR_FGT_GROUP_IDS__];
876 
877 	/* Exception Information */
878 	struct kvm_vcpu_fault_info fault;
879 
880 	/* Configuration flags, set once and for all before the vcpu can run */
881 	u8 cflags;
882 
883 	/* Input flags to the hypervisor code, potentially cleared after use */
884 	u8 iflags;
885 
886 	/* State flags for kernel bookkeeping, unused by the hypervisor code */
887 	u16 sflags;
888 
889 	/*
890 	 * Don't run the guest (internal implementation need).
891 	 *
892 	 * Contrary to the flags above, this is set/cleared outside of
893 	 * a vcpu context, and thus cannot be mixed with the flags
894 	 * themselves (or the flag accesses need to be made atomic).
895 	 */
896 	bool pause;
897 
898 	/*
899 	 * We maintain more than a single set of debug registers to support
900 	 * debugging the guest from the host and to maintain separate host and
901 	 * guest state during world switches. vcpu_debug_state are the debug
902 	 * registers of the vcpu as the guest sees them.
903 	 *
904 	 * external_debug_state contains the debug values we want to debug the
905 	 * guest. This is set via the KVM_SET_GUEST_DEBUG ioctl.
906 	 */
907 	struct kvm_guest_debug_arch vcpu_debug_state;
908 	struct kvm_guest_debug_arch external_debug_state;
909 	u64 external_mdscr_el1;
910 
911 	enum {
912 		VCPU_DEBUG_FREE,
913 		VCPU_DEBUG_HOST_OWNED,
914 		VCPU_DEBUG_GUEST_OWNED,
915 	} debug_owner;
916 
917 	/* VGIC state */
918 	struct vgic_cpu vgic_cpu;
919 	struct arch_timer_cpu timer_cpu;
920 	struct kvm_pmu pmu;
921 
922 	/* vcpu power state */
923 	struct kvm_mp_state mp_state;
924 	spinlock_t mp_state_lock;
925 
926 	/* Cache some mmu pages needed inside spinlock regions */
927 	struct kvm_mmu_memory_cache mmu_page_cache;
928 
929 	/* Pages to top-up the pKVM/EL2 guest pool */
930 	struct kvm_hyp_memcache pkvm_memcache;
931 
932 	/* Virtual SError ESR to restore when HCR_EL2.VSE is set */
933 	u64 vsesr_el2;
934 
935 	/* Additional reset state */
936 	struct vcpu_reset_state	reset_state;
937 
938 	/* Guest PV state */
939 	struct {
940 		u64 last_steal;
941 		gpa_t base;
942 	} steal;
943 
944 	/* Per-vcpu CCSIDR override or NULL */
945 	u32 *ccsidr;
946 
947 	/* Per-vcpu TLB for VNCR_EL2 -- NULL when !NV */
948 	struct vncr_tlb	*vncr_tlb;
949 
950 	/* Hyp-readable copy of kvm_vcpu::pid */
951 	pid_t pid;
952 };
953 
954 /*
955  * Each 'flag' is composed of a comma-separated triplet:
956  *
957  * - the flag-set it belongs to in the vcpu->arch structure
958  * - the value for that flag
959  * - the mask for that flag
960  *
961  *  __vcpu_single_flag() builds such a triplet for a single-bit flag.
962  * unpack_vcpu_flag() extract the flag value from the triplet for
963  * direct use outside of the flag accessors.
964  */
965 #define __vcpu_single_flag(_set, _f)	_set, (_f), (_f)
966 
967 #define __unpack_flag(_set, _f, _m)	_f
968 #define unpack_vcpu_flag(...)		__unpack_flag(__VA_ARGS__)
969 
970 #define __build_check_flag(v, flagset, f, m)			\
971 	do {							\
972 		typeof(v->arch.flagset) *_fset;			\
973 								\
974 		/* Check that the flags fit in the mask */	\
975 		BUILD_BUG_ON(HWEIGHT(m) != HWEIGHT((f) | (m)));	\
976 		/* Check that the flags fit in the type */	\
977 		BUILD_BUG_ON((sizeof(*_fset) * 8) <= __fls(m));	\
978 	} while (0)
979 
980 #define __vcpu_get_flag(v, flagset, f, m)			\
981 	({							\
982 		__build_check_flag(v, flagset, f, m);		\
983 								\
984 		READ_ONCE(v->arch.flagset) & (m);		\
985 	})
986 
987 /*
988  * Note that the set/clear accessors must be preempt-safe in order to
989  * avoid nesting them with load/put which also manipulate flags...
990  */
991 #ifdef __KVM_NVHE_HYPERVISOR__
992 /* the nVHE hypervisor is always non-preemptible */
993 #define __vcpu_flags_preempt_disable()
994 #define __vcpu_flags_preempt_enable()
995 #else
996 #define __vcpu_flags_preempt_disable()	preempt_disable()
997 #define __vcpu_flags_preempt_enable()	preempt_enable()
998 #endif
999 
1000 #define __vcpu_set_flag(v, flagset, f, m)			\
1001 	do {							\
1002 		typeof(v->arch.flagset) *fset;			\
1003 								\
1004 		__build_check_flag(v, flagset, f, m);		\
1005 								\
1006 		fset = &v->arch.flagset;			\
1007 		__vcpu_flags_preempt_disable();			\
1008 		if (HWEIGHT(m) > 1)				\
1009 			*fset &= ~(m);				\
1010 		*fset |= (f);					\
1011 		__vcpu_flags_preempt_enable();			\
1012 	} while (0)
1013 
1014 #define __vcpu_clear_flag(v, flagset, f, m)			\
1015 	do {							\
1016 		typeof(v->arch.flagset) *fset;			\
1017 								\
1018 		__build_check_flag(v, flagset, f, m);		\
1019 								\
1020 		fset = &v->arch.flagset;			\
1021 		__vcpu_flags_preempt_disable();			\
1022 		*fset &= ~(m);					\
1023 		__vcpu_flags_preempt_enable();			\
1024 	} while (0)
1025 
1026 #define __vcpu_test_and_clear_flag(v, flagset, f, m)		\
1027 	({							\
1028 		typeof(v->arch.flagset) set;			\
1029 								\
1030 		set = __vcpu_get_flag(v, flagset, f, m);	\
1031 		__vcpu_clear_flag(v, flagset, f, m);		\
1032 								\
1033 		set;						\
1034 	})
1035 
1036 #define vcpu_get_flag(v, ...)	__vcpu_get_flag((v), __VA_ARGS__)
1037 #define vcpu_set_flag(v, ...)	__vcpu_set_flag((v), __VA_ARGS__)
1038 #define vcpu_clear_flag(v, ...)	__vcpu_clear_flag((v), __VA_ARGS__)
1039 #define vcpu_test_and_clear_flag(v, ...)			\
1040 	__vcpu_test_and_clear_flag((v), __VA_ARGS__)
1041 
1042 /* KVM_ARM_VCPU_INIT completed */
1043 #define VCPU_INITIALIZED	__vcpu_single_flag(cflags, BIT(0))
1044 /* SVE config completed */
1045 #define VCPU_SVE_FINALIZED	__vcpu_single_flag(cflags, BIT(1))
1046 /* pKVM VCPU setup completed */
1047 #define VCPU_PKVM_FINALIZED	__vcpu_single_flag(cflags, BIT(2))
1048 
1049 /* Exception pending */
1050 #define PENDING_EXCEPTION	__vcpu_single_flag(iflags, BIT(0))
1051 /*
1052  * PC increment. Overlaps with EXCEPT_MASK on purpose so that it can't
1053  * be set together with an exception...
1054  */
1055 #define INCREMENT_PC		__vcpu_single_flag(iflags, BIT(1))
1056 /* Target EL/MODE (not a single flag, but let's abuse the macro) */
1057 #define EXCEPT_MASK		__vcpu_single_flag(iflags, GENMASK(3, 1))
1058 /* Host-set: the hyp flushes the non-protected vCPU state in on entry */
1059 #define PKVM_HOST_STATE_DIRTY	__vcpu_single_flag(iflags, BIT(4))
1060 
1061 /* Helpers to encode exceptions with minimum fuss */
1062 #define __EXCEPT_MASK_VAL	unpack_vcpu_flag(EXCEPT_MASK)
1063 #define __EXCEPT_SHIFT		__builtin_ctzl(__EXCEPT_MASK_VAL)
1064 #define __vcpu_except_flags(_f)	iflags, (_f << __EXCEPT_SHIFT), __EXCEPT_MASK_VAL
1065 
1066 /*
1067  * When PENDING_EXCEPTION is set, EXCEPT_MASK can take the following
1068  * values:
1069  *
1070  * For AArch32 EL1:
1071  */
1072 #define EXCEPT_AA32_UND		__vcpu_except_flags(0)
1073 #define EXCEPT_AA32_IABT	__vcpu_except_flags(1)
1074 #define EXCEPT_AA32_DABT	__vcpu_except_flags(2)
1075 /* For AArch64: */
1076 #define EXCEPT_AA64_EL1_SYNC	__vcpu_except_flags(0)
1077 #define EXCEPT_AA64_EL1_IRQ	__vcpu_except_flags(1)
1078 #define EXCEPT_AA64_EL1_FIQ	__vcpu_except_flags(2)
1079 #define EXCEPT_AA64_EL1_SERR	__vcpu_except_flags(3)
1080 /* For AArch64 with NV: */
1081 #define EXCEPT_AA64_EL2_SYNC	__vcpu_except_flags(4)
1082 #define EXCEPT_AA64_EL2_IRQ	__vcpu_except_flags(5)
1083 #define EXCEPT_AA64_EL2_FIQ	__vcpu_except_flags(6)
1084 #define EXCEPT_AA64_EL2_SERR	__vcpu_except_flags(7)
1085 
1086 /* Physical CPU not in supported_cpus */
1087 #define ON_UNSUPPORTED_CPU	__vcpu_single_flag(sflags, BIT(0))
1088 /* WFIT instruction trapped */
1089 #define IN_WFIT			__vcpu_single_flag(sflags, BIT(1))
1090 /* vcpu system registers loaded on physical CPU */
1091 #define SYSREGS_ON_CPU		__vcpu_single_flag(sflags, BIT(2))
1092 /* Software step state is Active-pending for external debug */
1093 #define HOST_SS_ACTIVE_PENDING	__vcpu_single_flag(sflags, BIT(3))
1094 /* Software step state is Active pending for guest debug */
1095 #define GUEST_SS_ACTIVE_PENDING __vcpu_single_flag(sflags, BIT(4))
1096 /* PMUSERENR for the guest EL0 is on physical CPU */
1097 #define PMUSERENR_ON_CPU	__vcpu_single_flag(sflags, BIT(5))
1098 /* WFI instruction trapped */
1099 #define IN_WFI			__vcpu_single_flag(sflags, BIT(6))
1100 /* KVM is currently emulating a nested ERET */
1101 #define IN_NESTED_ERET		__vcpu_single_flag(sflags, BIT(7))
1102 /* SError pending for nested guest */
1103 #define NESTED_SERROR_PENDING	__vcpu_single_flag(sflags, BIT(8))
1104 /* KVM is currently emulating an L2 to L1 exception */
1105 #define IN_NESTED_EXCEPTION	__vcpu_single_flag(sflags, BIT(9))
1106 
1107 #define vcpu_sve_max_vq(vcpu)	sve_vq_from_vl((vcpu)->arch.sve_max_vl)
1108 
1109 #define vcpu_sve_zcr_elx(vcpu)						\
1110 	(unlikely(is_hyp_ctxt(vcpu)) ? ZCR_EL2 : ZCR_EL1)
1111 
1112 #define sve_state_size_from_vl(sve_max_vl) ({				\
1113 	size_t __size_ret;						\
1114 	unsigned int __vq;						\
1115 									\
1116 	if (WARN_ON(!sve_vl_valid(sve_max_vl))) {			\
1117 		__size_ret = 0;						\
1118 	} else {							\
1119 		__vq = sve_vq_from_vl(sve_max_vl);			\
1120 		__size_ret = SVE_SIG_REGS_SIZE(__vq);			\
1121 	}								\
1122 									\
1123 	__size_ret;							\
1124 })
1125 
1126 #define vcpu_sve_state_size(vcpu) sve_state_size_from_vl((vcpu)->arch.sve_max_vl)
1127 
1128 #define KVM_GUESTDBG_VALID_MASK (KVM_GUESTDBG_ENABLE | \
1129 				 KVM_GUESTDBG_USE_SW_BP | \
1130 				 KVM_GUESTDBG_USE_HW | \
1131 				 KVM_GUESTDBG_SINGLESTEP)
1132 
1133 #define kvm_has_sve(kvm)	(system_supports_sve() &&		\
1134 				 test_bit(KVM_ARCH_FLAG_GUEST_HAS_SVE, &(kvm)->arch.flags))
1135 
1136 #ifdef __KVM_NVHE_HYPERVISOR__
1137 #define vcpu_has_sve(vcpu)	kvm_has_sve(kern_hyp_va((vcpu)->kvm))
1138 #else
1139 #define vcpu_has_sve(vcpu)	kvm_has_sve((vcpu)->kvm)
1140 #endif
1141 
1142 #ifdef CONFIG_ARM64_PTR_AUTH
1143 #define vcpu_has_ptrauth(vcpu)						\
1144 	((cpus_have_final_cap(ARM64_HAS_ADDRESS_AUTH) ||		\
1145 	  cpus_have_final_cap(ARM64_HAS_GENERIC_AUTH)) &&		\
1146 	 (vcpu_has_feature(vcpu, KVM_ARM_VCPU_PTRAUTH_ADDRESS) ||       \
1147 	  vcpu_has_feature(vcpu, KVM_ARM_VCPU_PTRAUTH_GENERIC)))
1148 #else
1149 #define vcpu_has_ptrauth(vcpu)		false
1150 #endif
1151 
1152 #define vcpu_on_unsupported_cpu(vcpu)					\
1153 	vcpu_get_flag(vcpu, ON_UNSUPPORTED_CPU)
1154 
1155 #define vcpu_set_on_unsupported_cpu(vcpu)				\
1156 	vcpu_set_flag(vcpu, ON_UNSUPPORTED_CPU)
1157 
1158 #define vcpu_clear_on_unsupported_cpu(vcpu)				\
1159 	vcpu_clear_flag(vcpu, ON_UNSUPPORTED_CPU)
1160 
1161 #define vcpu_gp_regs(v)		(&(v)->arch.ctxt.regs)
1162 
1163 /*
1164  * Only use __vcpu_sys_reg/ctxt_sys_reg if you know you want the
1165  * memory backed version of a register, and not the one most recently
1166  * accessed by a running VCPU.  For example, for userspace access or
1167  * for system registers that are never context switched, but only
1168  * emulated.
1169  *
1170  * Don't bother with VNCR-based accesses in the nVHE code, it has no
1171  * business dealing with NV.
1172  */
1173 static inline u64 *___ctxt_sys_reg(const struct kvm_cpu_context *ctxt, int r)
1174 {
1175 #if !defined (__KVM_NVHE_HYPERVISOR__)
1176 	if (unlikely(cpus_have_final_cap(ARM64_HAS_NESTED_VIRT) &&
1177 		     r >= __VNCR_START__ && ctxt->vncr_array))
1178 		return &ctxt->vncr_array[r - __VNCR_START__];
1179 #endif
1180 	return (u64 *)&ctxt->sys_regs[r];
1181 }
1182 
1183 #define __ctxt_sys_reg(c,r)						\
1184 	({								\
1185 		BUILD_BUG_ON(__builtin_constant_p(r) &&			\
1186 			     (r) >= NR_SYS_REGS);			\
1187 		___ctxt_sys_reg(c, r);					\
1188 	})
1189 
1190 #define ctxt_sys_reg(c,r)	(*__ctxt_sys_reg(c,r))
1191 
1192 u64 kvm_vcpu_apply_reg_masks(const struct kvm_vcpu *, enum vcpu_sysreg, u64);
1193 
1194 #define __vcpu_assign_sys_reg(v, r, val)				\
1195 	do {								\
1196 		const struct kvm_cpu_context *ctxt = &(v)->arch.ctxt;	\
1197 		u64 __v = (val);					\
1198 		if (vcpu_has_nv((v)) && (r) >= __SANITISED_REG_START__)	\
1199 			__v = kvm_vcpu_apply_reg_masks((v), (r), __v);	\
1200 									\
1201 		ctxt_sys_reg(ctxt, (r)) = __v;				\
1202 	} while (0)
1203 
1204 #define __vcpu_rmw_sys_reg(v, r, op, val)				\
1205 	do {								\
1206 		const struct kvm_cpu_context *ctxt = &(v)->arch.ctxt;	\
1207 		u64 __v = ctxt_sys_reg(ctxt, (r));			\
1208 		__v op (val);						\
1209 		if (vcpu_has_nv((v)) && (r) >= __SANITISED_REG_START__)	\
1210 			__v = kvm_vcpu_apply_reg_masks((v), (r), __v);	\
1211 									\
1212 		ctxt_sys_reg(ctxt, (r)) = __v;				\
1213 	} while (0)
1214 
1215 #define __vcpu_sys_reg(v,r)						\
1216 	({								\
1217 		const struct kvm_cpu_context *ctxt = &(v)->arch.ctxt;	\
1218 		u64 __v = ctxt_sys_reg(ctxt, (r));			\
1219 		if (vcpu_has_nv((v)) && (r) >= __SANITISED_REG_START__)	\
1220 			__v = kvm_vcpu_apply_reg_masks((v), (r), __v);	\
1221 		__v;							\
1222 	})
1223 
1224 u64 vcpu_read_sys_reg(const struct kvm_vcpu *, enum vcpu_sysreg);
1225 void vcpu_write_sys_reg(struct kvm_vcpu *, u64, enum vcpu_sysreg);
1226 
1227 struct kvm_vm_stat {
1228 	struct kvm_vm_stat_generic generic;
1229 };
1230 
1231 struct kvm_vcpu_stat {
1232 	struct kvm_vcpu_stat_generic generic;
1233 	u64 hvc_exit_stat;
1234 	u64 wfe_exit_stat;
1235 	u64 wfi_exit_stat;
1236 	u64 mmio_exit_user;
1237 	u64 mmio_exit_kernel;
1238 	u64 signal_exits;
1239 	u64 exits;
1240 };
1241 
1242 unsigned long kvm_arm_num_regs(struct kvm_vcpu *vcpu);
1243 int kvm_arm_copy_reg_indices(struct kvm_vcpu *vcpu, u64 __user *indices);
1244 int kvm_arm_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
1245 int kvm_arm_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
1246 
1247 unsigned long kvm_arm_num_sys_reg_descs(struct kvm_vcpu *vcpu);
1248 int kvm_arm_copy_sys_reg_indices(struct kvm_vcpu *vcpu, u64 __user *uindices);
1249 
1250 int __kvm_arm_vcpu_get_events(struct kvm_vcpu *vcpu,
1251 			      struct kvm_vcpu_events *events);
1252 
1253 int __kvm_arm_vcpu_set_events(struct kvm_vcpu *vcpu,
1254 			      struct kvm_vcpu_events *events);
1255 
1256 void kvm_arm_halt_guest(struct kvm *kvm);
1257 void kvm_arm_resume_guest(struct kvm *kvm);
1258 
1259 #define vcpu_has_run_once(vcpu)	(!!READ_ONCE((vcpu)->pid))
1260 
1261 #ifndef __KVM_NVHE_HYPERVISOR__
1262 #define kvm_call_hyp_nvhe(f, ...)					\
1263 	({								\
1264 		struct arm_smccc_res res;				\
1265 									\
1266 		arm_smccc_1_1_hvc(KVM_HOST_SMCCC_FUNC(f),		\
1267 				  ##__VA_ARGS__, &res);			\
1268 		if (WARN_ON(res.a0 != SMCCC_RET_SUCCESS))		\
1269 			res.a1 = -EOPNOTSUPP;				\
1270 									\
1271 		res.a1;							\
1272 	})
1273 
1274 /*
1275  * The isb() below is there to guarantee the same behaviour on VHE as on !VHE,
1276  * where the eret to EL1 acts as a context synchronization event.
1277  */
1278 #define kvm_call_hyp(f, ...)						\
1279 	do {								\
1280 		if (has_vhe()) {					\
1281 			f(__VA_ARGS__);					\
1282 			isb();						\
1283 		} else {						\
1284 			kvm_call_hyp_nvhe(f, ##__VA_ARGS__);		\
1285 		}							\
1286 	} while(0)
1287 
1288 #define kvm_call_hyp_ret(f, ...)					\
1289 	({								\
1290 		typeof(f(__VA_ARGS__)) ret;				\
1291 									\
1292 		if (has_vhe()) {					\
1293 			ret = f(__VA_ARGS__);				\
1294 		} else {						\
1295 			ret = kvm_call_hyp_nvhe(f, ##__VA_ARGS__);	\
1296 		}							\
1297 									\
1298 		ret;							\
1299 	})
1300 #else /* __KVM_NVHE_HYPERVISOR__ */
1301 #define kvm_call_hyp(f, ...) f(__VA_ARGS__)
1302 #define kvm_call_hyp_ret(f, ...) f(__VA_ARGS__)
1303 #define kvm_call_hyp_nvhe(f, ...) f(__VA_ARGS__)
1304 #endif /* __KVM_NVHE_HYPERVISOR__ */
1305 
1306 int handle_exit(struct kvm_vcpu *vcpu, int exception_index);
1307 void handle_exit_early(struct kvm_vcpu *vcpu, int exception_index);
1308 
1309 int kvm_handle_cp14_load_store(struct kvm_vcpu *vcpu);
1310 int kvm_handle_cp14_32(struct kvm_vcpu *vcpu);
1311 int kvm_handle_cp14_64(struct kvm_vcpu *vcpu);
1312 int kvm_handle_cp15_32(struct kvm_vcpu *vcpu);
1313 int kvm_handle_cp15_64(struct kvm_vcpu *vcpu);
1314 int kvm_handle_sys_reg(struct kvm_vcpu *vcpu);
1315 int kvm_handle_cp10_id(struct kvm_vcpu *vcpu);
1316 
1317 void kvm_sys_regs_create_debugfs(struct kvm *kvm);
1318 void kvm_reset_sys_regs(struct kvm_vcpu *vcpu);
1319 
1320 int __init kvm_sys_reg_table_init(void);
1321 struct sys_reg_desc;
1322 int __init populate_sysreg_config(const struct sys_reg_desc *sr,
1323 				  unsigned int idx);
1324 int __init populate_nv_trap_config(void);
1325 
1326 void kvm_calculate_traps(struct kvm_vcpu *vcpu);
1327 
1328 /* MMIO helpers */
1329 void kvm_mmio_write_buf(void *buf, unsigned int len, unsigned long data);
1330 unsigned long kvm_mmio_read_buf(const void *buf, unsigned int len);
1331 
1332 int kvm_handle_mmio_return(struct kvm_vcpu *vcpu);
1333 int io_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa);
1334 
1335 /*
1336  * Returns true if a Performance Monitoring Interrupt (PMI), a.k.a. perf event,
1337  * arrived in guest context.  For arm64, any event that arrives while a vCPU is
1338  * loaded is considered to be "in guest".
1339  */
1340 static inline bool kvm_arch_pmi_in_guest(struct kvm_vcpu *vcpu)
1341 {
1342 	return IS_ENABLED(CONFIG_GUEST_PERF_EVENTS) && !!vcpu;
1343 }
1344 
1345 long kvm_hypercall_pv_features(struct kvm_vcpu *vcpu);
1346 gpa_t kvm_init_stolen_time(struct kvm_vcpu *vcpu);
1347 void kvm_update_stolen_time(struct kvm_vcpu *vcpu);
1348 
1349 bool kvm_arm_pvtime_supported(void);
1350 int kvm_arm_pvtime_set_attr(struct kvm_vcpu *vcpu,
1351 			    struct kvm_device_attr *attr);
1352 int kvm_arm_pvtime_get_attr(struct kvm_vcpu *vcpu,
1353 			    struct kvm_device_attr *attr);
1354 int kvm_arm_pvtime_has_attr(struct kvm_vcpu *vcpu,
1355 			    struct kvm_device_attr *attr);
1356 
1357 extern unsigned int __ro_after_init kvm_arm_vmid_bits;
1358 int __init kvm_arm_vmid_alloc_init(void);
1359 void __init kvm_arm_vmid_alloc_free(void);
1360 void kvm_arm_vmid_update(struct kvm_vmid *kvm_vmid);
1361 void kvm_arm_vmid_clear_active(void);
1362 
1363 static inline void kvm_arm_pvtime_vcpu_init(struct kvm_vcpu_arch *vcpu_arch)
1364 {
1365 	vcpu_arch->steal.base = INVALID_GPA;
1366 }
1367 
1368 static inline bool kvm_arm_is_pvtime_enabled(struct kvm_vcpu_arch *vcpu_arch)
1369 {
1370 	return (vcpu_arch->steal.base != INVALID_GPA);
1371 }
1372 
1373 struct kvm_vcpu *kvm_mpidr_to_vcpu(struct kvm *kvm, unsigned long mpidr);
1374 
1375 DECLARE_KVM_HYP_PER_CPU(struct kvm_host_data, kvm_host_data);
1376 
1377 /*
1378  * How we access per-CPU host data depends on the where we access it from,
1379  * and the mode we're in:
1380  *
1381  * - VHE and nVHE hypervisor bits use their locally defined instance
1382  *
1383  * - the rest of the kernel use either the VHE or nVHE one, depending on
1384  *   the mode we're running in.
1385  *
1386  *   Unless we're in protected mode, fully deprivileged, and the nVHE
1387  *   per-CPU stuff is exclusively accessible to the protected EL2 code.
1388  *   In this case, the EL1 code uses the *VHE* data as its private state
1389  *   (which makes sense in a way as there shouldn't be any shared state
1390  *   between the host and the hypervisor).
1391  *
1392  * Yes, this is all totally trivial. Shoot me now.
1393  */
1394 #if defined(__KVM_NVHE_HYPERVISOR__) || defined(__KVM_VHE_HYPERVISOR__)
1395 #define host_data_ptr(f)	(&this_cpu_ptr(&kvm_host_data)->f)
1396 #else
1397 #define host_data_ptr(f)						\
1398 	(static_branch_unlikely(&kvm_protected_mode_initialized) ?	\
1399 	 &this_cpu_ptr(&kvm_host_data)->f :				\
1400 	 &this_cpu_ptr_hyp_sym(kvm_host_data)->f)
1401 #endif
1402 
1403 #define host_data_test_flag(flag)					\
1404 	(test_bit(KVM_HOST_DATA_FLAG_##flag, host_data_ptr(flags)))
1405 #define host_data_set_flag(flag)					\
1406 	set_bit(KVM_HOST_DATA_FLAG_##flag, host_data_ptr(flags))
1407 #define host_data_clear_flag(flag)					\
1408 	clear_bit(KVM_HOST_DATA_FLAG_##flag, host_data_ptr(flags))
1409 
1410 /* Check whether the FP regs are owned by the guest */
1411 static inline bool guest_owns_fp_regs(void)
1412 {
1413 	return *host_data_ptr(fp_owner) == FP_STATE_GUEST_OWNED;
1414 }
1415 
1416 /* Check whether the FP regs are owned by the host */
1417 static inline bool host_owns_fp_regs(void)
1418 {
1419 	return *host_data_ptr(fp_owner) == FP_STATE_HOST_OWNED;
1420 }
1421 
1422 static inline void kvm_init_host_cpu_context(struct kvm_cpu_context *cpu_ctxt)
1423 {
1424 	/* The host's MPIDR is immutable, so let's set it up at boot time */
1425 	ctxt_sys_reg(cpu_ctxt, MPIDR_EL1) = read_cpuid_mpidr();
1426 }
1427 
1428 static inline bool kvm_system_needs_idmapped_vectors(void)
1429 {
1430 	return cpus_have_final_cap(ARM64_SPECTRE_V3A);
1431 }
1432 
1433 void kvm_init_host_debug_data(void);
1434 void kvm_debug_init_vhe(void);
1435 void kvm_vcpu_load_debug(struct kvm_vcpu *vcpu);
1436 void kvm_vcpu_put_debug(struct kvm_vcpu *vcpu);
1437 void kvm_debug_set_guest_ownership(struct kvm_vcpu *vcpu);
1438 void kvm_debug_handle_oslar(struct kvm_vcpu *vcpu, u64 val);
1439 
1440 #define kvm_vcpu_os_lock_enabled(vcpu)		\
1441 	(!!(__vcpu_sys_reg(vcpu, OSLSR_EL1) & OSLSR_EL1_OSLK))
1442 
1443 #define kvm_debug_regs_in_use(vcpu)		\
1444 	((vcpu)->arch.debug_owner != VCPU_DEBUG_FREE)
1445 #define kvm_host_owns_debug_regs(vcpu)		\
1446 	((vcpu)->arch.debug_owner == VCPU_DEBUG_HOST_OWNED)
1447 #define kvm_guest_owns_debug_regs(vcpu)		\
1448 	((vcpu)->arch.debug_owner == VCPU_DEBUG_GUEST_OWNED)
1449 
1450 int kvm_arm_vcpu_arch_set_attr(struct kvm_vcpu *vcpu,
1451 			       struct kvm_device_attr *attr);
1452 int kvm_arm_vcpu_arch_get_attr(struct kvm_vcpu *vcpu,
1453 			       struct kvm_device_attr *attr);
1454 int kvm_arm_vcpu_arch_has_attr(struct kvm_vcpu *vcpu,
1455 			       struct kvm_device_attr *attr);
1456 
1457 int kvm_vm_ioctl_mte_copy_tags(struct kvm *kvm,
1458 			       struct kvm_arm_copy_mte_tags *copy_tags);
1459 int kvm_vm_ioctl_set_counter_offset(struct kvm *kvm,
1460 				    struct kvm_arm_counter_offset *offset);
1461 int kvm_vm_ioctl_get_reg_writable_masks(struct kvm *kvm,
1462 					struct reg_mask_range *range);
1463 
1464 /* Guest/host FPSIMD coordination helpers */
1465 void kvm_arch_vcpu_load_fp(struct kvm_vcpu *vcpu);
1466 void kvm_arch_vcpu_ctxflush_fp(struct kvm_vcpu *vcpu);
1467 void kvm_arch_vcpu_ctxsync_fp(struct kvm_vcpu *vcpu);
1468 void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu);
1469 
1470 static inline bool kvm_pmu_counter_deferred(struct perf_event_attr *attr)
1471 {
1472 	return (!has_vhe() && attr->exclude_host);
1473 }
1474 
1475 #ifdef CONFIG_KVM
1476 void kvm_set_pmu_events(u64 set, struct perf_event_attr *attr);
1477 void kvm_clr_pmu_events(u64 clr);
1478 bool kvm_set_pmuserenr(u64 val);
1479 void kvm_enable_trbe(void);
1480 void kvm_disable_trbe(void);
1481 void kvm_tracing_set_el1_configuration(u64 trfcr_while_in_guest);
1482 #else
1483 static inline void kvm_set_pmu_events(u64 set, struct perf_event_attr *attr) {}
1484 static inline void kvm_clr_pmu_events(u64 clr) {}
1485 static inline bool kvm_set_pmuserenr(u64 val)
1486 {
1487 	return false;
1488 }
1489 static inline void kvm_enable_trbe(void) {}
1490 static inline void kvm_disable_trbe(void) {}
1491 static inline void kvm_tracing_set_el1_configuration(u64 trfcr_while_in_guest) {}
1492 #endif
1493 
1494 void kvm_vcpu_load_vhe(struct kvm_vcpu *vcpu);
1495 void kvm_vcpu_put_vhe(struct kvm_vcpu *vcpu);
1496 
1497 int __init kvm_set_ipa_limit(void);
1498 u32 kvm_get_pa_bits(struct kvm *kvm);
1499 
1500 #define __KVM_HAVE_ARCH_VM_ALLOC
1501 struct kvm *kvm_arch_alloc_vm(void);
1502 
1503 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS
1504 
1505 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS_RANGE
1506 
1507 #define kvm_vm_is_protected(kvm)	(is_protected_kvm_enabled() && (kvm)->arch.pkvm.is_protected)
1508 
1509 #define vcpu_is_protected(vcpu)		kvm_vm_is_protected((vcpu)->kvm)
1510 
1511 int kvm_arm_vcpu_finalize(struct kvm_vcpu *vcpu, int feature);
1512 bool kvm_arm_vcpu_is_finalized(struct kvm_vcpu *vcpu);
1513 
1514 #define kvm_arm_vcpu_sve_finalized(vcpu) vcpu_get_flag(vcpu, VCPU_SVE_FINALIZED)
1515 
1516 #define kvm_has_mte(kvm)					\
1517 	(system_supports_mte() &&				\
1518 	 test_bit(KVM_ARCH_FLAG_MTE_ENABLED, &(kvm)->arch.flags))
1519 
1520 #define kvm_supports_32bit_el0()				\
1521 	(system_supports_32bit_el0() &&				\
1522 	 !static_branch_unlikely(&arm64_mismatched_32bit_el0))
1523 
1524 #define kvm_vm_has_ran_once(kvm)					\
1525 	(test_bit(KVM_ARCH_FLAG_HAS_RAN_ONCE, &(kvm)->arch.flags))
1526 
1527 static inline bool __vcpu_has_feature(const struct kvm_arch *ka, int feature)
1528 {
1529 	return test_bit(feature, ka->vcpu_features);
1530 }
1531 
1532 #define kvm_vcpu_has_feature(k, f)	__vcpu_has_feature(&(k)->arch, (f))
1533 #define vcpu_has_feature(v, f)	__vcpu_has_feature(&(v)->kvm->arch, (f))
1534 
1535 #define kvm_vcpu_initialized(v) vcpu_get_flag(v, VCPU_INITIALIZED)
1536 
1537 int kvm_trng_call(struct kvm_vcpu *vcpu);
1538 #ifdef CONFIG_KVM
1539 extern phys_addr_t hyp_mem_base;
1540 extern phys_addr_t hyp_mem_size;
1541 void __init kvm_hyp_reserve(void);
1542 #else
1543 static inline void kvm_hyp_reserve(void) { }
1544 #endif
1545 
1546 void kvm_arm_vcpu_power_off(struct kvm_vcpu *vcpu);
1547 bool kvm_arm_vcpu_stopped(struct kvm_vcpu *vcpu);
1548 
1549 static inline u64 *__vm_id_reg(struct kvm_arch *ka, u32 reg)
1550 {
1551 	switch (reg) {
1552 	case sys_reg(3, 0, 0, 1, 0) ... sys_reg(3, 0, 0, 7, 7):
1553 		return &ka->id_regs[IDREG_IDX(reg)];
1554 	case SYS_CTR_EL0:
1555 		return &ka->ctr_el0;
1556 	case SYS_MIDR_EL1:
1557 		return &ka->midr_el1;
1558 	case SYS_REVIDR_EL1:
1559 		return &ka->revidr_el1;
1560 	case SYS_AIDR_EL1:
1561 		return &ka->aidr_el1;
1562 	default:
1563 		WARN_ON_ONCE(1);
1564 		return NULL;
1565 	}
1566 }
1567 
1568 #define kvm_read_vm_id_reg(kvm, reg)					\
1569 	({ u64 __val = *__vm_id_reg(&(kvm)->arch, reg); __val; })
1570 
1571 void kvm_set_vm_id_reg(struct kvm *kvm, u32 reg, u64 val);
1572 
1573 #define __expand_field_sign_unsigned(id, fld, val)			\
1574 	((u64)SYS_FIELD_VALUE(id, fld, val))
1575 
1576 #define __expand_field_sign_signed(id, fld, val)			\
1577 	({								\
1578 		u64 __val = SYS_FIELD_VALUE(id, fld, val);		\
1579 		sign_extend64(__val, id##_##fld##_WIDTH - 1);		\
1580 	})
1581 
1582 #define get_idreg_field_unsigned(kvm, id, fld)				\
1583 	({								\
1584 		u64 __val = kvm_read_vm_id_reg((kvm), SYS_##id);	\
1585 		FIELD_GET(id##_##fld##_MASK, __val);			\
1586 	})
1587 
1588 #define get_idreg_field_signed(kvm, id, fld)				\
1589 	({								\
1590 		u64 __val = get_idreg_field_unsigned(kvm, id, fld);	\
1591 		sign_extend64(__val, id##_##fld##_WIDTH - 1);		\
1592 	})
1593 
1594 #define get_idreg_field_enum(kvm, id, fld)				\
1595 	get_idreg_field_unsigned(kvm, id, fld)
1596 
1597 #define kvm_cmp_feat_signed(kvm, id, fld, op, limit)			\
1598 	(get_idreg_field_signed((kvm), id, fld) op __expand_field_sign_signed(id, fld, limit))
1599 
1600 #define kvm_cmp_feat_unsigned(kvm, id, fld, op, limit)			\
1601 	(get_idreg_field_unsigned((kvm), id, fld) op __expand_field_sign_unsigned(id, fld, limit))
1602 
1603 #define kvm_cmp_feat(kvm, id, fld, op, limit)				\
1604 	(id##_##fld##_SIGNED ?						\
1605 	 kvm_cmp_feat_signed(kvm, id, fld, op, limit) :			\
1606 	 kvm_cmp_feat_unsigned(kvm, id, fld, op, limit))
1607 
1608 #define __kvm_has_feat(kvm, id, fld, limit)				\
1609 	kvm_cmp_feat(kvm, id, fld, >=, limit)
1610 
1611 #define kvm_has_feat(kvm, ...) __kvm_has_feat(kvm, __VA_ARGS__)
1612 
1613 #define __kvm_has_feat_enum(kvm, id, fld, val)				\
1614 	kvm_cmp_feat_unsigned(kvm, id, fld, ==, val)
1615 
1616 #define kvm_has_feat_enum(kvm, ...) __kvm_has_feat_enum(kvm, __VA_ARGS__)
1617 
1618 #define kvm_has_feat_range(kvm, id, fld, min, max)			\
1619 	(kvm_cmp_feat(kvm, id, fld, >=, min) &&				\
1620 	kvm_cmp_feat(kvm, id, fld, <=, max))
1621 
1622 /* Check for a given level of PAuth support */
1623 #define kvm_has_pauth(k, l)						\
1624 	({								\
1625 		bool pa, pi, pa3;					\
1626 									\
1627 		pa  = kvm_has_feat((k), ID_AA64ISAR1_EL1, APA, l);	\
1628 		pa &= kvm_has_feat((k), ID_AA64ISAR1_EL1, GPA, IMP);	\
1629 		pi  = kvm_has_feat((k), ID_AA64ISAR1_EL1, API, l);	\
1630 		pi &= kvm_has_feat((k), ID_AA64ISAR1_EL1, GPI, IMP);	\
1631 		pa3  = kvm_has_feat((k), ID_AA64ISAR2_EL1, APA3, l);	\
1632 		pa3 &= kvm_has_feat((k), ID_AA64ISAR2_EL1, GPA3, IMP);	\
1633 									\
1634 		(pa + pi + pa3) == 1;					\
1635 	})
1636 
1637 #define kvm_has_fpmr(k)					\
1638 	(system_supports_fpmr() &&			\
1639 	 kvm_has_feat((k), ID_AA64PFR2_EL1, FPMR, IMP))
1640 
1641 #define kvm_has_tcr2(k)				\
1642 	(kvm_has_feat((k), ID_AA64MMFR3_EL1, TCRX, IMP))
1643 
1644 #define kvm_has_s1pie(k)				\
1645 	(kvm_has_feat((k), ID_AA64MMFR3_EL1, S1PIE, IMP))
1646 
1647 #define kvm_has_s1poe(k)				\
1648 	(system_supports_poe() &&			\
1649 	 kvm_has_feat((k), ID_AA64MMFR3_EL1, S1POE, IMP))
1650 
1651 #define kvm_has_ras(k)					\
1652 	(kvm_has_feat((k), ID_AA64PFR0_EL1, RAS, IMP))
1653 
1654 #define kvm_has_sctlr2(k)				\
1655 	(kvm_has_feat((k), ID_AA64MMFR3_EL1, SCTLRX, IMP))
1656 
1657 static inline bool kvm_arch_has_irq_bypass(void)
1658 {
1659 	return true;
1660 }
1661 
1662 void compute_fgu(struct kvm *kvm, enum fgt_group_id fgt);
1663 struct resx get_reg_fixed_bits(struct kvm *kvm, enum vcpu_sysreg reg);
1664 void check_feature_map(void);
1665 void kvm_vcpu_load_fgt(struct kvm_vcpu *vcpu);
1666 
1667 static __always_inline enum fgt_group_id __fgt_reg_to_group_id(enum vcpu_sysreg reg)
1668 {
1669 	switch (reg) {
1670 	case HFGRTR_EL2:
1671 	case HFGWTR_EL2:
1672 		return HFGRTR_GROUP;
1673 	case HFGITR_EL2:
1674 		return HFGITR_GROUP;
1675 	case HDFGRTR_EL2:
1676 	case HDFGWTR_EL2:
1677 		return HDFGRTR_GROUP;
1678 	case HAFGRTR_EL2:
1679 		return HAFGRTR_GROUP;
1680 	case HFGRTR2_EL2:
1681 	case HFGWTR2_EL2:
1682 		return HFGRTR2_GROUP;
1683 	case HFGITR2_EL2:
1684 		return HFGITR2_GROUP;
1685 	case HDFGRTR2_EL2:
1686 	case HDFGWTR2_EL2:
1687 		return HDFGRTR2_GROUP;
1688 	case ICH_HFGRTR_EL2:
1689 	case ICH_HFGWTR_EL2:
1690 		return ICH_HFGRTR_GROUP;
1691 	case ICH_HFGITR_EL2:
1692 		return ICH_HFGITR_GROUP;
1693 	default:
1694 		BUILD_BUG_ON(1);
1695 	}
1696 }
1697 
1698 #define vcpu_fgt(vcpu, reg)						\
1699 	({								\
1700 		enum fgt_group_id id = __fgt_reg_to_group_id(reg);	\
1701 		u64 *p;							\
1702 		switch (reg) {						\
1703 		case HFGWTR_EL2:					\
1704 		case HDFGWTR_EL2:					\
1705 		case HFGWTR2_EL2:					\
1706 		case HDFGWTR2_EL2:					\
1707 		case ICH_HFGWTR_EL2:					\
1708 			p = &(vcpu)->arch.fgt[id].w;			\
1709 			break;						\
1710 		default:						\
1711 			p = &(vcpu)->arch.fgt[id].r;			\
1712 			break;						\
1713 		}							\
1714 									\
1715 		p;							\
1716 	})
1717 
1718 long kvm_get_cap_for_kvm_ioctl(unsigned int ioctl, long *ext);
1719 
1720 #endif /* __ARM64_KVM_HOST_H__ */
1721