xref: /linux/arch/arm64/include/asm/kvm_host.h (revision 6316366129d2885fae07c2774f4b7ae0a45fb55d)
1 /* SPDX-License-Identifier: GPL-2.0-only */
2 /*
3  * Copyright (C) 2012,2013 - ARM Ltd
4  * Author: Marc Zyngier <marc.zyngier@arm.com>
5  *
6  * Derived from arch/arm/include/asm/kvm_host.h:
7  * Copyright (C) 2012 - Virtual Open Systems and Columbia University
8  * Author: Christoffer Dall <c.dall@virtualopensystems.com>
9  */
10 
11 #ifndef __ARM64_KVM_HOST_H__
12 #define __ARM64_KVM_HOST_H__
13 
14 #include <linux/arm-smccc.h>
15 #include <linux/bitmap.h>
16 #include <linux/types.h>
17 #include <linux/jump_label.h>
18 #include <linux/kvm_types.h>
19 #include <linux/maple_tree.h>
20 #include <linux/percpu.h>
21 #include <linux/psci.h>
22 #include <asm/arch_gicv3.h>
23 #include <asm/barrier.h>
24 #include <asm/cpufeature.h>
25 #include <asm/cputype.h>
26 #include <asm/daifflags.h>
27 #include <asm/fpsimd.h>
28 #include <asm/kvm.h>
29 #include <asm/kvm_asm.h>
30 #include <asm/vncr_mapping.h>
31 
32 #define __KVM_HAVE_ARCH_INTC_INITIALIZED
33 
34 #define KVM_HALT_POLL_NS_DEFAULT 500000
35 
36 #include <kvm/arm_vgic.h>
37 #include <kvm/arm_arch_timer.h>
38 #include <kvm/arm_pmu.h>
39 
40 #define KVM_MAX_VCPUS VGIC_V3_MAX_CPUS
41 
42 #define KVM_VCPU_MAX_FEATURES 9
43 #define KVM_VCPU_VALID_FEATURES	(BIT(KVM_VCPU_MAX_FEATURES) - 1)
44 
45 #define KVM_REQ_SLEEP \
46 	KVM_ARCH_REQ_FLAGS(0, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
47 #define KVM_REQ_IRQ_PENDING		KVM_ARCH_REQ(1)
48 #define KVM_REQ_VCPU_RESET		KVM_ARCH_REQ(2)
49 #define KVM_REQ_RECORD_STEAL		KVM_ARCH_REQ(3)
50 #define KVM_REQ_RELOAD_GICv4		KVM_ARCH_REQ(4)
51 #define KVM_REQ_RELOAD_PMU		KVM_ARCH_REQ(5)
52 #define KVM_REQ_SUSPEND			KVM_ARCH_REQ(6)
53 #define KVM_REQ_RESYNC_PMU_EL0		KVM_ARCH_REQ(7)
54 #define KVM_REQ_NESTED_S2_UNMAP		KVM_ARCH_REQ(8)
55 #define KVM_REQ_GUEST_HYP_IRQ_PENDING	KVM_ARCH_REQ(9)
56 #define KVM_REQ_MAP_L1_VNCR_EL2		KVM_ARCH_REQ(10)
57 #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 	/*
221 	 * true when this MMU needs to be unmapped before being used for a new
222 	 * purpose.
223 	 */
224 	bool	pending_unmap;
225 
226 	/*
227 	 *  0: Nobody is currently using this, check vttbr for validity
228 	 * >0: Somebody is actively using this.
229 	 */
230 	atomic_t refcnt;
231 };
232 
233 struct kvm_arch_memory_slot {
234 };
235 
236 /**
237  * struct kvm_smccc_features: Descriptor of the hypercall services exposed to the guests
238  *
239  * @std_bmap: Bitmap of standard secure service calls
240  * @std_hyp_bmap: Bitmap of standard hypervisor service calls
241  * @vendor_hyp_bmap: Bitmap of vendor specific hypervisor service calls
242  */
243 struct kvm_smccc_features {
244 	unsigned long std_bmap;
245 	unsigned long std_hyp_bmap;
246 	unsigned long vendor_hyp_bmap; /* Function numbers 0-63 */
247 	unsigned long vendor_hyp_bmap_2; /* Function numbers 64-127 */
248 };
249 
250 typedef unsigned int pkvm_handle_t;
251 
252 struct kvm_protected_vm {
253 	pkvm_handle_t handle;
254 	struct kvm_hyp_memcache teardown_mc;
255 	struct kvm_hyp_memcache stage2_teardown_mc;
256 	bool is_protected;
257 	bool is_created;
258 };
259 
260 struct kvm_mpidr_data {
261 	u64			mpidr_mask;
262 	DECLARE_FLEX_ARRAY(u16, cmpidr_to_idx);
263 };
264 
265 static inline u16 kvm_mpidr_index(struct kvm_mpidr_data *data, u64 mpidr)
266 {
267 	unsigned long index = 0, mask = data->mpidr_mask;
268 	unsigned long aff = mpidr & MPIDR_HWID_BITMASK;
269 
270 	bitmap_gather(&index, &aff, &mask, fls(mask));
271 
272 	return index;
273 }
274 
275 struct kvm_sysreg_masks;
276 
277 enum fgt_group_id {
278 	__NO_FGT_GROUP__,
279 	HFGRTR_GROUP,
280 	HFGWTR_GROUP = HFGRTR_GROUP,
281 	HDFGRTR_GROUP,
282 	HDFGWTR_GROUP = HDFGRTR_GROUP,
283 	HFGITR_GROUP,
284 	HAFGRTR_GROUP,
285 	HFGRTR2_GROUP,
286 	HFGWTR2_GROUP = HFGRTR2_GROUP,
287 	HDFGRTR2_GROUP,
288 	HDFGWTR2_GROUP = HDFGRTR2_GROUP,
289 	HFGITR2_GROUP,
290 
291 	/* Must be last */
292 	__NR_FGT_GROUP_IDS__
293 };
294 
295 struct kvm_arch {
296 	struct kvm_s2_mmu mmu;
297 
298 	/*
299 	 * Fine-Grained UNDEF, mimicking the FGT layout defined by the
300 	 * architecture. We track them globally, as we present the
301 	 * same feature-set to all vcpus.
302 	 *
303 	 * Index 0 is currently spare.
304 	 */
305 	u64 fgu[__NR_FGT_GROUP_IDS__];
306 
307 	/*
308 	 * Stage 2 paging state for VMs with nested S2 using a virtual
309 	 * VMID.
310 	 */
311 	struct kvm_s2_mmu *nested_mmus;
312 	size_t nested_mmus_size;
313 	int nested_mmus_next;
314 
315 	/* Interrupt controller */
316 	struct vgic_dist	vgic;
317 
318 	/* Timers */
319 	struct arch_timer_vm_data timer_data;
320 
321 	/* Mandated version of PSCI */
322 	u32 psci_version;
323 
324 	/* Protects VM-scoped configuration data */
325 	struct mutex config_lock;
326 
327 	/*
328 	 * If we encounter a data abort without valid instruction syndrome
329 	 * information, report this to user space.  User space can (and
330 	 * should) opt in to this feature if KVM_CAP_ARM_NISV_TO_USER is
331 	 * supported.
332 	 */
333 #define KVM_ARCH_FLAG_RETURN_NISV_IO_ABORT_TO_USER	0
334 	/* Memory Tagging Extension enabled for the guest */
335 #define KVM_ARCH_FLAG_MTE_ENABLED			1
336 	/* At least one vCPU has ran in the VM */
337 #define KVM_ARCH_FLAG_HAS_RAN_ONCE			2
338 	/* The vCPU feature set for the VM is configured */
339 #define KVM_ARCH_FLAG_VCPU_FEATURES_CONFIGURED		3
340 	/* PSCI SYSTEM_SUSPEND enabled for the guest */
341 #define KVM_ARCH_FLAG_SYSTEM_SUSPEND_ENABLED		4
342 	/* VM counter offset */
343 #define KVM_ARCH_FLAG_VM_COUNTER_OFFSET			5
344 	/* Timer PPIs made immutable */
345 #define KVM_ARCH_FLAG_TIMER_PPIS_IMMUTABLE		6
346 	/* Initial ID reg values loaded */
347 #define KVM_ARCH_FLAG_ID_REGS_INITIALIZED		7
348 	/* Fine-Grained UNDEF initialised */
349 #define KVM_ARCH_FLAG_FGU_INITIALIZED			8
350 	/* SVE exposed to guest */
351 #define KVM_ARCH_FLAG_GUEST_HAS_SVE			9
352 	/* MIDR_EL1, REVIDR_EL1, and AIDR_EL1 are writable from userspace */
353 #define KVM_ARCH_FLAG_WRITABLE_IMP_ID_REGS		10
354 	/* Unhandled SEAs are taken to userspace */
355 #define KVM_ARCH_FLAG_EXIT_SEA				11
356 	unsigned long flags;
357 
358 	/* VM-wide vCPU feature set */
359 	DECLARE_BITMAP(vcpu_features, KVM_VCPU_MAX_FEATURES);
360 
361 	/* MPIDR to vcpu index mapping, optional */
362 	struct kvm_mpidr_data *mpidr_data;
363 
364 	/*
365 	 * VM-wide PMU filter, implemented as a bitmap and big enough for
366 	 * up to 2^10 events (ARMv8.0) or 2^16 events (ARMv8.1+).
367 	 */
368 	unsigned long *pmu_filter;
369 	struct arm_pmu *arm_pmu;
370 
371 	cpumask_var_t supported_cpus;
372 
373 	/* Maximum number of counters for the guest */
374 	u8 nr_pmu_counters;
375 
376 	/* Hypercall features firmware registers' descriptor */
377 	struct kvm_smccc_features smccc_feat;
378 	struct maple_tree smccc_filter;
379 
380 	/*
381 	 * Emulated CPU ID registers per VM
382 	 * (Op0, Op1, CRn, CRm, Op2) of the ID registers to be saved in it
383 	 * is (3, 0, 0, crm, op2), where 1<=crm<8, 0<=op2<8.
384 	 *
385 	 * These emulated idregs are VM-wide, but accessed from the context of a vCPU.
386 	 * Atomic access to multiple idregs are guarded by kvm_arch.config_lock.
387 	 */
388 #define IDREG_IDX(id)		(((sys_reg_CRm(id) - 1) << 3) | sys_reg_Op2(id))
389 #define KVM_ARM_ID_REG_NUM	(IDREG_IDX(sys_reg(3, 0, 0, 7, 7)) + 1)
390 	u64 id_regs[KVM_ARM_ID_REG_NUM];
391 
392 	u64 midr_el1;
393 	u64 revidr_el1;
394 	u64 aidr_el1;
395 	u64 ctr_el0;
396 
397 	/* Masks for VNCR-backed and general EL2 sysregs */
398 	struct kvm_sysreg_masks	*sysreg_masks;
399 
400 	/* Count the number of VNCR_EL2 currently mapped */
401 	atomic_t vncr_map_count;
402 
403 	/*
404 	 * For an untrusted host VM, 'pkvm.handle' is used to lookup
405 	 * the associated pKVM instance in the hypervisor.
406 	 */
407 	struct kvm_protected_vm pkvm;
408 };
409 
410 struct kvm_vcpu_fault_info {
411 	u64 esr_el2;		/* Hyp Syndrom Register */
412 	u64 far_el2;		/* Hyp Fault Address Register */
413 	u64 hpfar_el2;		/* Hyp IPA Fault Address Register */
414 	u64 disr_el1;		/* Deferred [SError] Status Register */
415 };
416 
417 /*
418  * VNCR() just places the VNCR_capable registers in the enum after
419  * __VNCR_START__, and the value (after correction) to be an 8-byte offset
420  * from the VNCR base. As we don't require the enum to be otherwise ordered,
421  * we need the terrible hack below to ensure that we correctly size the
422  * sys_regs array, no matter what.
423  *
424  * The __MAX__ macro has been lifted from Sean Eron Anderson's wonderful
425  * treasure trove of bit hacks:
426  * https://graphics.stanford.edu/~seander/bithacks.html#IntegerMinOrMax
427  */
428 #define __MAX__(x,y)	((x) ^ (((x) ^ (y)) & -((x) < (y))))
429 #define VNCR(r)						\
430 	__before_##r,					\
431 	r = __VNCR_START__ + ((VNCR_ ## r) / 8),	\
432 	__after_##r = __MAX__(__before_##r - 1, r)
433 
434 #define MARKER(m)				\
435 	m, __after_##m = m - 1
436 
437 enum vcpu_sysreg {
438 	__INVALID_SYSREG__,   /* 0 is reserved as an invalid value */
439 	MPIDR_EL1,	/* MultiProcessor Affinity Register */
440 	CLIDR_EL1,	/* Cache Level ID Register */
441 	CSSELR_EL1,	/* Cache Size Selection Register */
442 	TPIDR_EL0,	/* Thread ID, User R/W */
443 	TPIDRRO_EL0,	/* Thread ID, User R/O */
444 	TPIDR_EL1,	/* Thread ID, Privileged */
445 	CNTKCTL_EL1,	/* Timer Control Register (EL1) */
446 	PAR_EL1,	/* Physical Address Register */
447 	MDCCINT_EL1,	/* Monitor Debug Comms Channel Interrupt Enable Reg */
448 	OSLSR_EL1,	/* OS Lock Status Register */
449 	DISR_EL1,	/* Deferred Interrupt Status Register */
450 
451 	/* Performance Monitors Registers */
452 	PMCR_EL0,	/* Control Register */
453 	PMSELR_EL0,	/* Event Counter Selection Register */
454 	PMEVCNTR0_EL0,	/* Event Counter Register (0-30) */
455 	PMEVCNTR30_EL0 = PMEVCNTR0_EL0 + 30,
456 	PMCCNTR_EL0,	/* Cycle Counter Register */
457 	PMEVTYPER0_EL0,	/* Event Type Register (0-30) */
458 	PMEVTYPER30_EL0 = PMEVTYPER0_EL0 + 30,
459 	PMCCFILTR_EL0,	/* Cycle Count Filter Register */
460 	PMCNTENSET_EL0,	/* Count Enable Set Register */
461 	PMINTENSET_EL1,	/* Interrupt Enable Set Register */
462 	PMOVSSET_EL0,	/* Overflow Flag Status Set Register */
463 	PMUSERENR_EL0,	/* User Enable Register */
464 
465 	/* Pointer Authentication Registers in a strict increasing order. */
466 	APIAKEYLO_EL1,
467 	APIAKEYHI_EL1,
468 	APIBKEYLO_EL1,
469 	APIBKEYHI_EL1,
470 	APDAKEYLO_EL1,
471 	APDAKEYHI_EL1,
472 	APDBKEYLO_EL1,
473 	APDBKEYHI_EL1,
474 	APGAKEYLO_EL1,
475 	APGAKEYHI_EL1,
476 
477 	/* Memory Tagging Extension registers */
478 	RGSR_EL1,	/* Random Allocation Tag Seed Register */
479 	GCR_EL1,	/* Tag Control Register */
480 	TFSRE0_EL1,	/* Tag Fault Status Register (EL0) */
481 
482 	POR_EL0,	/* Permission Overlay Register 0 (EL0) */
483 
484 	/* FP/SIMD/SVE */
485 	SVCR,
486 	FPMR,
487 
488 	/* 32bit specific registers. */
489 	DACR32_EL2,	/* Domain Access Control Register */
490 	IFSR32_EL2,	/* Instruction Fault Status Register */
491 	FPEXC32_EL2,	/* Floating-Point Exception Control Register */
492 	DBGVCR32_EL2,	/* Debug Vector Catch Register */
493 
494 	/* EL2 registers */
495 	ACTLR_EL2,	/* Auxiliary Control Register (EL2) */
496 	CPTR_EL2,	/* Architectural Feature Trap Register (EL2) */
497 	HACR_EL2,	/* Hypervisor Auxiliary Control Register */
498 	ZCR_EL2,	/* SVE Control Register (EL2) */
499 	TTBR0_EL2,	/* Translation Table Base Register 0 (EL2) */
500 	TTBR1_EL2,	/* Translation Table Base Register 1 (EL2) */
501 	TCR_EL2,	/* Translation Control Register (EL2) */
502 	PIRE0_EL2,	/* Permission Indirection Register 0 (EL2) */
503 	PIR_EL2,	/* Permission Indirection Register 1 (EL2) */
504 	POR_EL2,	/* Permission Overlay Register 2 (EL2) */
505 	SPSR_EL2,	/* EL2 saved program status register */
506 	ELR_EL2,	/* EL2 exception link register */
507 	AFSR0_EL2,	/* Auxiliary Fault Status Register 0 (EL2) */
508 	AFSR1_EL2,	/* Auxiliary Fault Status Register 1 (EL2) */
509 	ESR_EL2,	/* Exception Syndrome Register (EL2) */
510 	FAR_EL2,	/* Fault Address Register (EL2) */
511 	HPFAR_EL2,	/* Hypervisor IPA Fault Address Register */
512 	MAIR_EL2,	/* Memory Attribute Indirection Register (EL2) */
513 	AMAIR_EL2,	/* Auxiliary Memory Attribute Indirection Register (EL2) */
514 	VBAR_EL2,	/* Vector Base Address Register (EL2) */
515 	RVBAR_EL2,	/* Reset Vector Base Address Register */
516 	CONTEXTIDR_EL2,	/* Context ID Register (EL2) */
517 	SP_EL2,		/* EL2 Stack Pointer */
518 	CNTHP_CTL_EL2,
519 	CNTHP_CVAL_EL2,
520 	CNTHV_CTL_EL2,
521 	CNTHV_CVAL_EL2,
522 
523 	/* Anything from this can be RES0/RES1 sanitised */
524 	MARKER(__SANITISED_REG_START__),
525 	SCTLR_EL2,	/* System Control Register (EL2) */
526 	TCR2_EL2,	/* Extended Translation Control Register (EL2) */
527 	SCTLR2_EL2,	/* System Control Register 2 (EL2) */
528 	MDCR_EL2,	/* Monitor Debug Configuration Register (EL2) */
529 	CNTHCTL_EL2,	/* Counter-timer Hypervisor Control register */
530 
531 	/* Any VNCR-capable reg goes after this point */
532 	MARKER(__VNCR_START__),
533 
534 	VNCR(SCTLR_EL1),/* System Control Register */
535 	VNCR(ACTLR_EL1),/* Auxiliary Control Register */
536 	VNCR(CPACR_EL1),/* Coprocessor Access Control */
537 	VNCR(ZCR_EL1),	/* SVE Control */
538 	VNCR(TTBR0_EL1),/* Translation Table Base Register 0 */
539 	VNCR(TTBR1_EL1),/* Translation Table Base Register 1 */
540 	VNCR(TCR_EL1),	/* Translation Control Register */
541 	VNCR(TCR2_EL1),	/* Extended Translation Control Register */
542 	VNCR(SCTLR2_EL1), /* System Control Register 2 */
543 	VNCR(ESR_EL1),	/* Exception Syndrome Register */
544 	VNCR(AFSR0_EL1),/* Auxiliary Fault Status Register 0 */
545 	VNCR(AFSR1_EL1),/* Auxiliary Fault Status Register 1 */
546 	VNCR(FAR_EL1),	/* Fault Address Register */
547 	VNCR(MAIR_EL1),	/* Memory Attribute Indirection Register */
548 	VNCR(VBAR_EL1),	/* Vector Base Address Register */
549 	VNCR(CONTEXTIDR_EL1),	/* Context ID Register */
550 	VNCR(AMAIR_EL1),/* Aux Memory Attribute Indirection Register */
551 	VNCR(MDSCR_EL1),/* Monitor Debug System Control Register */
552 	VNCR(ELR_EL1),
553 	VNCR(SP_EL1),
554 	VNCR(SPSR_EL1),
555 	VNCR(TFSR_EL1),	/* Tag Fault Status Register (EL1) */
556 	VNCR(VPIDR_EL2),/* Virtualization Processor ID Register */
557 	VNCR(VMPIDR_EL2),/* Virtualization Multiprocessor ID Register */
558 	VNCR(HCR_EL2),	/* Hypervisor Configuration Register */
559 	VNCR(HSTR_EL2),	/* Hypervisor System Trap Register */
560 	VNCR(VTTBR_EL2),/* Virtualization Translation Table Base Register */
561 	VNCR(VTCR_EL2),	/* Virtualization Translation Control Register */
562 	VNCR(TPIDR_EL2),/* EL2 Software Thread ID Register */
563 	VNCR(HCRX_EL2),	/* Extended Hypervisor Configuration Register */
564 
565 	/* Permission Indirection Extension registers */
566 	VNCR(PIR_EL1),	 /* Permission Indirection Register 1 (EL1) */
567 	VNCR(PIRE0_EL1), /*  Permission Indirection Register 0 (EL1) */
568 
569 	VNCR(POR_EL1),	/* Permission Overlay Register 1 (EL1) */
570 
571 	/* FEAT_RAS registers */
572 	VNCR(VDISR_EL2),
573 	VNCR(VSESR_EL2),
574 
575 	VNCR(HFGRTR_EL2),
576 	VNCR(HFGWTR_EL2),
577 	VNCR(HFGITR_EL2),
578 	VNCR(HDFGRTR_EL2),
579 	VNCR(HDFGWTR_EL2),
580 	VNCR(HAFGRTR_EL2),
581 	VNCR(HFGRTR2_EL2),
582 	VNCR(HFGWTR2_EL2),
583 	VNCR(HFGITR2_EL2),
584 	VNCR(HDFGRTR2_EL2),
585 	VNCR(HDFGWTR2_EL2),
586 
587 	VNCR(VNCR_EL2),
588 
589 	VNCR(CNTVOFF_EL2),
590 	VNCR(CNTV_CVAL_EL0),
591 	VNCR(CNTV_CTL_EL0),
592 	VNCR(CNTP_CVAL_EL0),
593 	VNCR(CNTP_CTL_EL0),
594 
595 	VNCR(ICH_LR0_EL2),
596 	VNCR(ICH_LR1_EL2),
597 	VNCR(ICH_LR2_EL2),
598 	VNCR(ICH_LR3_EL2),
599 	VNCR(ICH_LR4_EL2),
600 	VNCR(ICH_LR5_EL2),
601 	VNCR(ICH_LR6_EL2),
602 	VNCR(ICH_LR7_EL2),
603 	VNCR(ICH_LR8_EL2),
604 	VNCR(ICH_LR9_EL2),
605 	VNCR(ICH_LR10_EL2),
606 	VNCR(ICH_LR11_EL2),
607 	VNCR(ICH_LR12_EL2),
608 	VNCR(ICH_LR13_EL2),
609 	VNCR(ICH_LR14_EL2),
610 	VNCR(ICH_LR15_EL2),
611 
612 	VNCR(ICH_AP0R0_EL2),
613 	VNCR(ICH_AP0R1_EL2),
614 	VNCR(ICH_AP0R2_EL2),
615 	VNCR(ICH_AP0R3_EL2),
616 	VNCR(ICH_AP1R0_EL2),
617 	VNCR(ICH_AP1R1_EL2),
618 	VNCR(ICH_AP1R2_EL2),
619 	VNCR(ICH_AP1R3_EL2),
620 	VNCR(ICH_HCR_EL2),
621 	VNCR(ICH_VMCR_EL2),
622 
623 	NR_SYS_REGS	/* Nothing after this line! */
624 };
625 
626 struct resx {
627 	u64	res0;
628 	u64	res1;
629 };
630 
631 struct kvm_sysreg_masks {
632 	struct resx mask[NR_SYS_REGS - __SANITISED_REG_START__];
633 };
634 
635 static inline struct resx __kvm_get_sysreg_resx(struct kvm_arch *arch,
636 						enum vcpu_sysreg sr)
637 {
638 	struct kvm_sysreg_masks *masks;
639 
640 	masks = arch->sysreg_masks;
641 	if (likely(masks &&
642 		   sr >= __SANITISED_REG_START__ && sr < NR_SYS_REGS))
643 		return masks->mask[sr - __SANITISED_REG_START__];
644 
645 	return (struct resx){};
646 }
647 
648 #define kvm_get_sysreg_resx(k, sr) __kvm_get_sysreg_resx(&(k)->arch, (sr))
649 
650 static inline void __kvm_set_sysreg_resx(struct kvm_arch *arch,
651 					 enum vcpu_sysreg sr, struct resx resx)
652 {
653 	arch->sysreg_masks->mask[sr - __SANITISED_REG_START__] = resx;
654 }
655 
656 #define kvm_set_sysreg_resx(k, sr, resx)		\
657 	__kvm_set_sysreg_resx(&(k)->arch, (sr), (resx))
658 
659 struct fgt_masks {
660 	const char	*str;
661 	u64		mask;
662 	u64		nmask;
663 	u64		res0;
664 	u64		res1;
665 };
666 
667 extern struct fgt_masks hfgrtr_masks;
668 extern struct fgt_masks hfgwtr_masks;
669 extern struct fgt_masks hfgitr_masks;
670 extern struct fgt_masks hdfgrtr_masks;
671 extern struct fgt_masks hdfgwtr_masks;
672 extern struct fgt_masks hafgrtr_masks;
673 extern struct fgt_masks hfgrtr2_masks;
674 extern struct fgt_masks hfgwtr2_masks;
675 extern struct fgt_masks hfgitr2_masks;
676 extern struct fgt_masks hdfgrtr2_masks;
677 extern struct fgt_masks hdfgwtr2_masks;
678 
679 extern struct fgt_masks kvm_nvhe_sym(hfgrtr_masks);
680 extern struct fgt_masks kvm_nvhe_sym(hfgwtr_masks);
681 extern struct fgt_masks kvm_nvhe_sym(hfgitr_masks);
682 extern struct fgt_masks kvm_nvhe_sym(hdfgrtr_masks);
683 extern struct fgt_masks kvm_nvhe_sym(hdfgwtr_masks);
684 extern struct fgt_masks kvm_nvhe_sym(hafgrtr_masks);
685 extern struct fgt_masks kvm_nvhe_sym(hfgrtr2_masks);
686 extern struct fgt_masks kvm_nvhe_sym(hfgwtr2_masks);
687 extern struct fgt_masks kvm_nvhe_sym(hfgitr2_masks);
688 extern struct fgt_masks kvm_nvhe_sym(hdfgrtr2_masks);
689 extern struct fgt_masks kvm_nvhe_sym(hdfgwtr2_masks);
690 
691 struct kvm_cpu_context {
692 	struct user_pt_regs regs;	/* sp = sp_el0 */
693 
694 	u64	spsr_abt;
695 	u64	spsr_und;
696 	u64	spsr_irq;
697 	u64	spsr_fiq;
698 
699 	struct user_fpsimd_state fp_regs;
700 
701 	u64 sys_regs[NR_SYS_REGS];
702 
703 	struct kvm_vcpu *__hyp_running_vcpu;
704 
705 	/* This pointer has to be 4kB aligned. */
706 	u64 *vncr_array;
707 };
708 
709 struct cpu_sve_state {
710 	__u64 zcr_el1;
711 
712 	/*
713 	 * Ordering is important since __sve_save_state/__sve_restore_state
714 	 * relies on it.
715 	 */
716 	__u32 fpsr;
717 	__u32 fpcr;
718 
719 	/* Must be SVE_VQ_BYTES (128 bit) aligned. */
720 	__u8 sve_regs[];
721 };
722 
723 /*
724  * This structure is instantiated on a per-CPU basis, and contains
725  * data that is:
726  *
727  * - tied to a single physical CPU, and
728  * - either have a lifetime that does not extend past vcpu_put()
729  * - or is an invariant for the lifetime of the system
730  *
731  * Use host_data_ptr(field) as a way to access a pointer to such a
732  * field.
733  */
734 struct kvm_host_data {
735 #define KVM_HOST_DATA_FLAG_HAS_SPE			0
736 #define KVM_HOST_DATA_FLAG_HAS_TRBE			1
737 #define KVM_HOST_DATA_FLAG_TRBE_ENABLED			2
738 #define KVM_HOST_DATA_FLAG_EL1_TRACING_CONFIGURED	3
739 #define KVM_HOST_DATA_FLAG_VCPU_IN_HYP_CONTEXT		4
740 #define KVM_HOST_DATA_FLAG_L1_VNCR_MAPPED		5
741 #define KVM_HOST_DATA_FLAG_HAS_BRBE			6
742 	unsigned long flags;
743 
744 	struct kvm_cpu_context host_ctxt;
745 
746 	/*
747 	 * Hyp VA.
748 	 * sve_state is only used in pKVM and if system_supports_sve().
749 	 */
750 	struct cpu_sve_state *sve_state;
751 
752 	/* Used by pKVM only. */
753 	u64	fpmr;
754 
755 	/* Ownership of the FP regs */
756 	enum {
757 		FP_STATE_FREE,
758 		FP_STATE_HOST_OWNED,
759 		FP_STATE_GUEST_OWNED,
760 	} fp_owner;
761 
762 	/*
763 	 * host_debug_state contains the host registers which are
764 	 * saved and restored during world switches.
765 	 */
766 	struct {
767 		/* {Break,watch}point registers */
768 		struct kvm_guest_debug_arch regs;
769 		/* Statistical profiling extension */
770 		u64 pmscr_el1;
771 		/* Self-hosted trace */
772 		u64 trfcr_el1;
773 		/* Values of trap registers for the host before guest entry. */
774 		u64 mdcr_el2;
775 		u64 brbcr_el1;
776 	} host_debug_state;
777 
778 	/* Guest trace filter value */
779 	u64 trfcr_while_in_guest;
780 
781 	/* Number of programmable event counters (PMCR_EL0.N) for this CPU */
782 	unsigned int nr_event_counters;
783 
784 	/* Number of debug breakpoints/watchpoints for this CPU (minus 1) */
785 	unsigned int debug_brps;
786 	unsigned int debug_wrps;
787 };
788 
789 struct kvm_host_psci_config {
790 	/* PSCI version used by host. */
791 	u32 version;
792 	u32 smccc_version;
793 
794 	/* Function IDs used by host if version is v0.1. */
795 	struct psci_0_1_function_ids function_ids_0_1;
796 
797 	bool psci_0_1_cpu_suspend_implemented;
798 	bool psci_0_1_cpu_on_implemented;
799 	bool psci_0_1_cpu_off_implemented;
800 	bool psci_0_1_migrate_implemented;
801 };
802 
803 extern struct kvm_host_psci_config kvm_nvhe_sym(kvm_host_psci_config);
804 #define kvm_host_psci_config CHOOSE_NVHE_SYM(kvm_host_psci_config)
805 
806 extern s64 kvm_nvhe_sym(hyp_physvirt_offset);
807 #define hyp_physvirt_offset CHOOSE_NVHE_SYM(hyp_physvirt_offset)
808 
809 extern u64 kvm_nvhe_sym(hyp_cpu_logical_map)[NR_CPUS];
810 #define hyp_cpu_logical_map CHOOSE_NVHE_SYM(hyp_cpu_logical_map)
811 
812 struct vcpu_reset_state {
813 	unsigned long	pc;
814 	unsigned long	r0;
815 	bool		be;
816 	bool		reset;
817 };
818 
819 struct vncr_tlb;
820 
821 struct kvm_vcpu_arch {
822 	struct kvm_cpu_context ctxt;
823 
824 	/*
825 	 * Guest floating point state
826 	 *
827 	 * The architecture has two main floating point extensions,
828 	 * the original FPSIMD and SVE.  These have overlapping
829 	 * register views, with the FPSIMD V registers occupying the
830 	 * low 128 bits of the SVE Z registers.  When the core
831 	 * floating point code saves the register state of a task it
832 	 * records which view it saved in fp_type.
833 	 */
834 	void *sve_state;
835 	enum fp_type fp_type;
836 	unsigned int sve_max_vl;
837 
838 	/* Stage 2 paging state used by the hardware on next switch */
839 	struct kvm_s2_mmu *hw_mmu;
840 
841 	/* Values of trap registers for the guest. */
842 	u64 hcr_el2;
843 	u64 hcrx_el2;
844 	u64 mdcr_el2;
845 
846 	struct {
847 		u64 r;
848 		u64 w;
849 	} fgt[__NR_FGT_GROUP_IDS__];
850 
851 	/* Exception Information */
852 	struct kvm_vcpu_fault_info fault;
853 
854 	/* Configuration flags, set once and for all before the vcpu can run */
855 	u8 cflags;
856 
857 	/* Input flags to the hypervisor code, potentially cleared after use */
858 	u8 iflags;
859 
860 	/* State flags for kernel bookkeeping, unused by the hypervisor code */
861 	u16 sflags;
862 
863 	/*
864 	 * Don't run the guest (internal implementation need).
865 	 *
866 	 * Contrary to the flags above, this is set/cleared outside of
867 	 * a vcpu context, and thus cannot be mixed with the flags
868 	 * themselves (or the flag accesses need to be made atomic).
869 	 */
870 	bool pause;
871 
872 	/*
873 	 * We maintain more than a single set of debug registers to support
874 	 * debugging the guest from the host and to maintain separate host and
875 	 * guest state during world switches. vcpu_debug_state are the debug
876 	 * registers of the vcpu as the guest sees them.
877 	 *
878 	 * external_debug_state contains the debug values we want to debug the
879 	 * guest. This is set via the KVM_SET_GUEST_DEBUG ioctl.
880 	 */
881 	struct kvm_guest_debug_arch vcpu_debug_state;
882 	struct kvm_guest_debug_arch external_debug_state;
883 	u64 external_mdscr_el1;
884 
885 	enum {
886 		VCPU_DEBUG_FREE,
887 		VCPU_DEBUG_HOST_OWNED,
888 		VCPU_DEBUG_GUEST_OWNED,
889 	} debug_owner;
890 
891 	/* VGIC state */
892 	struct vgic_cpu vgic_cpu;
893 	struct arch_timer_cpu timer_cpu;
894 	struct kvm_pmu pmu;
895 
896 	/* vcpu power state */
897 	struct kvm_mp_state mp_state;
898 	spinlock_t mp_state_lock;
899 
900 	/* Cache some mmu pages needed inside spinlock regions */
901 	struct kvm_mmu_memory_cache mmu_page_cache;
902 
903 	/* Pages to top-up the pKVM/EL2 guest pool */
904 	struct kvm_hyp_memcache pkvm_memcache;
905 
906 	/* Virtual SError ESR to restore when HCR_EL2.VSE is set */
907 	u64 vsesr_el2;
908 
909 	/* Additional reset state */
910 	struct vcpu_reset_state	reset_state;
911 
912 	/* Guest PV state */
913 	struct {
914 		u64 last_steal;
915 		gpa_t base;
916 	} steal;
917 
918 	/* Per-vcpu CCSIDR override or NULL */
919 	u32 *ccsidr;
920 
921 	/* Per-vcpu TLB for VNCR_EL2 -- NULL when !NV */
922 	struct vncr_tlb	*vncr_tlb;
923 };
924 
925 /*
926  * Each 'flag' is composed of a comma-separated triplet:
927  *
928  * - the flag-set it belongs to in the vcpu->arch structure
929  * - the value for that flag
930  * - the mask for that flag
931  *
932  *  __vcpu_single_flag() builds such a triplet for a single-bit flag.
933  * unpack_vcpu_flag() extract the flag value from the triplet for
934  * direct use outside of the flag accessors.
935  */
936 #define __vcpu_single_flag(_set, _f)	_set, (_f), (_f)
937 
938 #define __unpack_flag(_set, _f, _m)	_f
939 #define unpack_vcpu_flag(...)		__unpack_flag(__VA_ARGS__)
940 
941 #define __build_check_flag(v, flagset, f, m)			\
942 	do {							\
943 		typeof(v->arch.flagset) *_fset;			\
944 								\
945 		/* Check that the flags fit in the mask */	\
946 		BUILD_BUG_ON(HWEIGHT(m) != HWEIGHT((f) | (m)));	\
947 		/* Check that the flags fit in the type */	\
948 		BUILD_BUG_ON((sizeof(*_fset) * 8) <= __fls(m));	\
949 	} while (0)
950 
951 #define __vcpu_get_flag(v, flagset, f, m)			\
952 	({							\
953 		__build_check_flag(v, flagset, f, m);		\
954 								\
955 		READ_ONCE(v->arch.flagset) & (m);		\
956 	})
957 
958 /*
959  * Note that the set/clear accessors must be preempt-safe in order to
960  * avoid nesting them with load/put which also manipulate flags...
961  */
962 #ifdef __KVM_NVHE_HYPERVISOR__
963 /* the nVHE hypervisor is always non-preemptible */
964 #define __vcpu_flags_preempt_disable()
965 #define __vcpu_flags_preempt_enable()
966 #else
967 #define __vcpu_flags_preempt_disable()	preempt_disable()
968 #define __vcpu_flags_preempt_enable()	preempt_enable()
969 #endif
970 
971 #define __vcpu_set_flag(v, flagset, f, m)			\
972 	do {							\
973 		typeof(v->arch.flagset) *fset;			\
974 								\
975 		__build_check_flag(v, flagset, f, m);		\
976 								\
977 		fset = &v->arch.flagset;			\
978 		__vcpu_flags_preempt_disable();			\
979 		if (HWEIGHT(m) > 1)				\
980 			*fset &= ~(m);				\
981 		*fset |= (f);					\
982 		__vcpu_flags_preempt_enable();			\
983 	} while (0)
984 
985 #define __vcpu_clear_flag(v, flagset, f, m)			\
986 	do {							\
987 		typeof(v->arch.flagset) *fset;			\
988 								\
989 		__build_check_flag(v, flagset, f, m);		\
990 								\
991 		fset = &v->arch.flagset;			\
992 		__vcpu_flags_preempt_disable();			\
993 		*fset &= ~(m);					\
994 		__vcpu_flags_preempt_enable();			\
995 	} while (0)
996 
997 #define __vcpu_test_and_clear_flag(v, flagset, f, m)		\
998 	({							\
999 		typeof(v->arch.flagset) set;			\
1000 								\
1001 		set = __vcpu_get_flag(v, flagset, f, m);	\
1002 		__vcpu_clear_flag(v, flagset, f, m);		\
1003 								\
1004 		set;						\
1005 	})
1006 
1007 #define vcpu_get_flag(v, ...)	__vcpu_get_flag((v), __VA_ARGS__)
1008 #define vcpu_set_flag(v, ...)	__vcpu_set_flag((v), __VA_ARGS__)
1009 #define vcpu_clear_flag(v, ...)	__vcpu_clear_flag((v), __VA_ARGS__)
1010 #define vcpu_test_and_clear_flag(v, ...)			\
1011 	__vcpu_test_and_clear_flag((v), __VA_ARGS__)
1012 
1013 /* KVM_ARM_VCPU_INIT completed */
1014 #define VCPU_INITIALIZED	__vcpu_single_flag(cflags, BIT(0))
1015 /* SVE config completed */
1016 #define VCPU_SVE_FINALIZED	__vcpu_single_flag(cflags, BIT(1))
1017 /* pKVM VCPU setup completed */
1018 #define VCPU_PKVM_FINALIZED	__vcpu_single_flag(cflags, BIT(2))
1019 
1020 /* Exception pending */
1021 #define PENDING_EXCEPTION	__vcpu_single_flag(iflags, BIT(0))
1022 /*
1023  * PC increment. Overlaps with EXCEPT_MASK on purpose so that it can't
1024  * be set together with an exception...
1025  */
1026 #define INCREMENT_PC		__vcpu_single_flag(iflags, BIT(1))
1027 /* Target EL/MODE (not a single flag, but let's abuse the macro) */
1028 #define EXCEPT_MASK		__vcpu_single_flag(iflags, GENMASK(3, 1))
1029 
1030 /* Helpers to encode exceptions with minimum fuss */
1031 #define __EXCEPT_MASK_VAL	unpack_vcpu_flag(EXCEPT_MASK)
1032 #define __EXCEPT_SHIFT		__builtin_ctzl(__EXCEPT_MASK_VAL)
1033 #define __vcpu_except_flags(_f)	iflags, (_f << __EXCEPT_SHIFT), __EXCEPT_MASK_VAL
1034 
1035 /*
1036  * When PENDING_EXCEPTION is set, EXCEPT_MASK can take the following
1037  * values:
1038  *
1039  * For AArch32 EL1:
1040  */
1041 #define EXCEPT_AA32_UND		__vcpu_except_flags(0)
1042 #define EXCEPT_AA32_IABT	__vcpu_except_flags(1)
1043 #define EXCEPT_AA32_DABT	__vcpu_except_flags(2)
1044 /* For AArch64: */
1045 #define EXCEPT_AA64_EL1_SYNC	__vcpu_except_flags(0)
1046 #define EXCEPT_AA64_EL1_IRQ	__vcpu_except_flags(1)
1047 #define EXCEPT_AA64_EL1_FIQ	__vcpu_except_flags(2)
1048 #define EXCEPT_AA64_EL1_SERR	__vcpu_except_flags(3)
1049 /* For AArch64 with NV: */
1050 #define EXCEPT_AA64_EL2_SYNC	__vcpu_except_flags(4)
1051 #define EXCEPT_AA64_EL2_IRQ	__vcpu_except_flags(5)
1052 #define EXCEPT_AA64_EL2_FIQ	__vcpu_except_flags(6)
1053 #define EXCEPT_AA64_EL2_SERR	__vcpu_except_flags(7)
1054 
1055 /* Physical CPU not in supported_cpus */
1056 #define ON_UNSUPPORTED_CPU	__vcpu_single_flag(sflags, BIT(0))
1057 /* WFIT instruction trapped */
1058 #define IN_WFIT			__vcpu_single_flag(sflags, BIT(1))
1059 /* vcpu system registers loaded on physical CPU */
1060 #define SYSREGS_ON_CPU		__vcpu_single_flag(sflags, BIT(2))
1061 /* Software step state is Active-pending for external debug */
1062 #define HOST_SS_ACTIVE_PENDING	__vcpu_single_flag(sflags, BIT(3))
1063 /* Software step state is Active pending for guest debug */
1064 #define GUEST_SS_ACTIVE_PENDING __vcpu_single_flag(sflags, BIT(4))
1065 /* PMUSERENR for the guest EL0 is on physical CPU */
1066 #define PMUSERENR_ON_CPU	__vcpu_single_flag(sflags, BIT(5))
1067 /* WFI instruction trapped */
1068 #define IN_WFI			__vcpu_single_flag(sflags, BIT(6))
1069 /* KVM is currently emulating a nested ERET */
1070 #define IN_NESTED_ERET		__vcpu_single_flag(sflags, BIT(7))
1071 /* SError pending for nested guest */
1072 #define NESTED_SERROR_PENDING	__vcpu_single_flag(sflags, BIT(8))
1073 
1074 
1075 /* Pointer to the vcpu's SVE FFR for sve_{save,load}_state() */
1076 #define vcpu_sve_pffr(vcpu) (kern_hyp_va((vcpu)->arch.sve_state) +	\
1077 			     sve_ffr_offset((vcpu)->arch.sve_max_vl))
1078 
1079 #define vcpu_sve_max_vq(vcpu)	sve_vq_from_vl((vcpu)->arch.sve_max_vl)
1080 
1081 #define vcpu_sve_zcr_elx(vcpu)						\
1082 	(unlikely(is_hyp_ctxt(vcpu)) ? ZCR_EL2 : ZCR_EL1)
1083 
1084 #define sve_state_size_from_vl(sve_max_vl) ({				\
1085 	size_t __size_ret;						\
1086 	unsigned int __vq;						\
1087 									\
1088 	if (WARN_ON(!sve_vl_valid(sve_max_vl))) {			\
1089 		__size_ret = 0;						\
1090 	} else {							\
1091 		__vq = sve_vq_from_vl(sve_max_vl);			\
1092 		__size_ret = SVE_SIG_REGS_SIZE(__vq);			\
1093 	}								\
1094 									\
1095 	__size_ret;							\
1096 })
1097 
1098 #define vcpu_sve_state_size(vcpu) sve_state_size_from_vl((vcpu)->arch.sve_max_vl)
1099 
1100 #define KVM_GUESTDBG_VALID_MASK (KVM_GUESTDBG_ENABLE | \
1101 				 KVM_GUESTDBG_USE_SW_BP | \
1102 				 KVM_GUESTDBG_USE_HW | \
1103 				 KVM_GUESTDBG_SINGLESTEP)
1104 
1105 #define kvm_has_sve(kvm)	(system_supports_sve() &&		\
1106 				 test_bit(KVM_ARCH_FLAG_GUEST_HAS_SVE, &(kvm)->arch.flags))
1107 
1108 #ifdef __KVM_NVHE_HYPERVISOR__
1109 #define vcpu_has_sve(vcpu)	kvm_has_sve(kern_hyp_va((vcpu)->kvm))
1110 #else
1111 #define vcpu_has_sve(vcpu)	kvm_has_sve((vcpu)->kvm)
1112 #endif
1113 
1114 #ifdef CONFIG_ARM64_PTR_AUTH
1115 #define vcpu_has_ptrauth(vcpu)						\
1116 	((cpus_have_final_cap(ARM64_HAS_ADDRESS_AUTH) ||		\
1117 	  cpus_have_final_cap(ARM64_HAS_GENERIC_AUTH)) &&		\
1118 	 (vcpu_has_feature(vcpu, KVM_ARM_VCPU_PTRAUTH_ADDRESS) ||       \
1119 	  vcpu_has_feature(vcpu, KVM_ARM_VCPU_PTRAUTH_GENERIC)))
1120 #else
1121 #define vcpu_has_ptrauth(vcpu)		false
1122 #endif
1123 
1124 #define vcpu_on_unsupported_cpu(vcpu)					\
1125 	vcpu_get_flag(vcpu, ON_UNSUPPORTED_CPU)
1126 
1127 #define vcpu_set_on_unsupported_cpu(vcpu)				\
1128 	vcpu_set_flag(vcpu, ON_UNSUPPORTED_CPU)
1129 
1130 #define vcpu_clear_on_unsupported_cpu(vcpu)				\
1131 	vcpu_clear_flag(vcpu, ON_UNSUPPORTED_CPU)
1132 
1133 #define vcpu_gp_regs(v)		(&(v)->arch.ctxt.regs)
1134 
1135 /*
1136  * Only use __vcpu_sys_reg/ctxt_sys_reg if you know you want the
1137  * memory backed version of a register, and not the one most recently
1138  * accessed by a running VCPU.  For example, for userspace access or
1139  * for system registers that are never context switched, but only
1140  * emulated.
1141  *
1142  * Don't bother with VNCR-based accesses in the nVHE code, it has no
1143  * business dealing with NV.
1144  */
1145 static inline u64 *___ctxt_sys_reg(const struct kvm_cpu_context *ctxt, int r)
1146 {
1147 #if !defined (__KVM_NVHE_HYPERVISOR__)
1148 	if (unlikely(cpus_have_final_cap(ARM64_HAS_NESTED_VIRT) &&
1149 		     r >= __VNCR_START__ && ctxt->vncr_array))
1150 		return &ctxt->vncr_array[r - __VNCR_START__];
1151 #endif
1152 	return (u64 *)&ctxt->sys_regs[r];
1153 }
1154 
1155 #define __ctxt_sys_reg(c,r)						\
1156 	({								\
1157 		BUILD_BUG_ON(__builtin_constant_p(r) &&			\
1158 			     (r) >= NR_SYS_REGS);			\
1159 		___ctxt_sys_reg(c, r);					\
1160 	})
1161 
1162 #define ctxt_sys_reg(c,r)	(*__ctxt_sys_reg(c,r))
1163 
1164 u64 kvm_vcpu_apply_reg_masks(const struct kvm_vcpu *, enum vcpu_sysreg, u64);
1165 
1166 #define __vcpu_assign_sys_reg(v, r, val)				\
1167 	do {								\
1168 		const struct kvm_cpu_context *ctxt = &(v)->arch.ctxt;	\
1169 		u64 __v = (val);					\
1170 		if (vcpu_has_nv((v)) && (r) >= __SANITISED_REG_START__)	\
1171 			__v = kvm_vcpu_apply_reg_masks((v), (r), __v);	\
1172 									\
1173 		ctxt_sys_reg(ctxt, (r)) = __v;				\
1174 	} while (0)
1175 
1176 #define __vcpu_rmw_sys_reg(v, r, op, val)				\
1177 	do {								\
1178 		const struct kvm_cpu_context *ctxt = &(v)->arch.ctxt;	\
1179 		u64 __v = ctxt_sys_reg(ctxt, (r));			\
1180 		__v op (val);						\
1181 		if (vcpu_has_nv((v)) && (r) >= __SANITISED_REG_START__)	\
1182 			__v = kvm_vcpu_apply_reg_masks((v), (r), __v);	\
1183 									\
1184 		ctxt_sys_reg(ctxt, (r)) = __v;				\
1185 	} while (0)
1186 
1187 #define __vcpu_sys_reg(v,r)						\
1188 	({								\
1189 		const struct kvm_cpu_context *ctxt = &(v)->arch.ctxt;	\
1190 		u64 __v = ctxt_sys_reg(ctxt, (r));			\
1191 		if (vcpu_has_nv((v)) && (r) >= __SANITISED_REG_START__)	\
1192 			__v = kvm_vcpu_apply_reg_masks((v), (r), __v);	\
1193 		__v;							\
1194 	})
1195 
1196 u64 vcpu_read_sys_reg(const struct kvm_vcpu *, enum vcpu_sysreg);
1197 void vcpu_write_sys_reg(struct kvm_vcpu *, u64, enum vcpu_sysreg);
1198 
1199 struct kvm_vm_stat {
1200 	struct kvm_vm_stat_generic generic;
1201 };
1202 
1203 struct kvm_vcpu_stat {
1204 	struct kvm_vcpu_stat_generic generic;
1205 	u64 hvc_exit_stat;
1206 	u64 wfe_exit_stat;
1207 	u64 wfi_exit_stat;
1208 	u64 mmio_exit_user;
1209 	u64 mmio_exit_kernel;
1210 	u64 signal_exits;
1211 	u64 exits;
1212 };
1213 
1214 unsigned long kvm_arm_num_regs(struct kvm_vcpu *vcpu);
1215 int kvm_arm_copy_reg_indices(struct kvm_vcpu *vcpu, u64 __user *indices);
1216 int kvm_arm_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
1217 int kvm_arm_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
1218 
1219 unsigned long kvm_arm_num_sys_reg_descs(struct kvm_vcpu *vcpu);
1220 int kvm_arm_copy_sys_reg_indices(struct kvm_vcpu *vcpu, u64 __user *uindices);
1221 
1222 int __kvm_arm_vcpu_get_events(struct kvm_vcpu *vcpu,
1223 			      struct kvm_vcpu_events *events);
1224 
1225 int __kvm_arm_vcpu_set_events(struct kvm_vcpu *vcpu,
1226 			      struct kvm_vcpu_events *events);
1227 
1228 void kvm_arm_halt_guest(struct kvm *kvm);
1229 void kvm_arm_resume_guest(struct kvm *kvm);
1230 
1231 #define vcpu_has_run_once(vcpu)	(!!READ_ONCE((vcpu)->pid))
1232 
1233 #ifndef __KVM_NVHE_HYPERVISOR__
1234 #define kvm_call_hyp_nvhe(f, ...)						\
1235 	({								\
1236 		struct arm_smccc_res res;				\
1237 									\
1238 		arm_smccc_1_1_hvc(KVM_HOST_SMCCC_FUNC(f),		\
1239 				  ##__VA_ARGS__, &res);			\
1240 		WARN_ON(res.a0 != SMCCC_RET_SUCCESS);			\
1241 									\
1242 		res.a1;							\
1243 	})
1244 
1245 /*
1246  * The isb() below is there to guarantee the same behaviour on VHE as on !VHE,
1247  * where the eret to EL1 acts as a context synchronization event.
1248  */
1249 #define kvm_call_hyp(f, ...)						\
1250 	do {								\
1251 		if (has_vhe()) {					\
1252 			f(__VA_ARGS__);					\
1253 			isb();						\
1254 		} else {						\
1255 			kvm_call_hyp_nvhe(f, ##__VA_ARGS__);		\
1256 		}							\
1257 	} while(0)
1258 
1259 #define kvm_call_hyp_ret(f, ...)					\
1260 	({								\
1261 		typeof(f(__VA_ARGS__)) ret;				\
1262 									\
1263 		if (has_vhe()) {					\
1264 			ret = f(__VA_ARGS__);				\
1265 		} else {						\
1266 			ret = kvm_call_hyp_nvhe(f, ##__VA_ARGS__);	\
1267 		}							\
1268 									\
1269 		ret;							\
1270 	})
1271 #else /* __KVM_NVHE_HYPERVISOR__ */
1272 #define kvm_call_hyp(f, ...) f(__VA_ARGS__)
1273 #define kvm_call_hyp_ret(f, ...) f(__VA_ARGS__)
1274 #define kvm_call_hyp_nvhe(f, ...) f(__VA_ARGS__)
1275 #endif /* __KVM_NVHE_HYPERVISOR__ */
1276 
1277 int handle_exit(struct kvm_vcpu *vcpu, int exception_index);
1278 void handle_exit_early(struct kvm_vcpu *vcpu, int exception_index);
1279 
1280 int kvm_handle_cp14_load_store(struct kvm_vcpu *vcpu);
1281 int kvm_handle_cp14_32(struct kvm_vcpu *vcpu);
1282 int kvm_handle_cp14_64(struct kvm_vcpu *vcpu);
1283 int kvm_handle_cp15_32(struct kvm_vcpu *vcpu);
1284 int kvm_handle_cp15_64(struct kvm_vcpu *vcpu);
1285 int kvm_handle_sys_reg(struct kvm_vcpu *vcpu);
1286 int kvm_handle_cp10_id(struct kvm_vcpu *vcpu);
1287 
1288 void kvm_sys_regs_create_debugfs(struct kvm *kvm);
1289 void kvm_reset_sys_regs(struct kvm_vcpu *vcpu);
1290 
1291 int __init kvm_sys_reg_table_init(void);
1292 struct sys_reg_desc;
1293 int __init populate_sysreg_config(const struct sys_reg_desc *sr,
1294 				  unsigned int idx);
1295 int __init populate_nv_trap_config(void);
1296 
1297 void kvm_calculate_traps(struct kvm_vcpu *vcpu);
1298 
1299 /* MMIO helpers */
1300 void kvm_mmio_write_buf(void *buf, unsigned int len, unsigned long data);
1301 unsigned long kvm_mmio_read_buf(const void *buf, unsigned int len);
1302 
1303 int kvm_handle_mmio_return(struct kvm_vcpu *vcpu);
1304 int io_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa);
1305 
1306 /*
1307  * Returns true if a Performance Monitoring Interrupt (PMI), a.k.a. perf event,
1308  * arrived in guest context.  For arm64, any event that arrives while a vCPU is
1309  * loaded is considered to be "in guest".
1310  */
1311 static inline bool kvm_arch_pmi_in_guest(struct kvm_vcpu *vcpu)
1312 {
1313 	return IS_ENABLED(CONFIG_GUEST_PERF_EVENTS) && !!vcpu;
1314 }
1315 
1316 long kvm_hypercall_pv_features(struct kvm_vcpu *vcpu);
1317 gpa_t kvm_init_stolen_time(struct kvm_vcpu *vcpu);
1318 void kvm_update_stolen_time(struct kvm_vcpu *vcpu);
1319 
1320 bool kvm_arm_pvtime_supported(void);
1321 int kvm_arm_pvtime_set_attr(struct kvm_vcpu *vcpu,
1322 			    struct kvm_device_attr *attr);
1323 int kvm_arm_pvtime_get_attr(struct kvm_vcpu *vcpu,
1324 			    struct kvm_device_attr *attr);
1325 int kvm_arm_pvtime_has_attr(struct kvm_vcpu *vcpu,
1326 			    struct kvm_device_attr *attr);
1327 
1328 extern unsigned int __ro_after_init kvm_arm_vmid_bits;
1329 int __init kvm_arm_vmid_alloc_init(void);
1330 void __init kvm_arm_vmid_alloc_free(void);
1331 void kvm_arm_vmid_update(struct kvm_vmid *kvm_vmid);
1332 void kvm_arm_vmid_clear_active(void);
1333 
1334 static inline void kvm_arm_pvtime_vcpu_init(struct kvm_vcpu_arch *vcpu_arch)
1335 {
1336 	vcpu_arch->steal.base = INVALID_GPA;
1337 }
1338 
1339 static inline bool kvm_arm_is_pvtime_enabled(struct kvm_vcpu_arch *vcpu_arch)
1340 {
1341 	return (vcpu_arch->steal.base != INVALID_GPA);
1342 }
1343 
1344 struct kvm_vcpu *kvm_mpidr_to_vcpu(struct kvm *kvm, unsigned long mpidr);
1345 
1346 DECLARE_KVM_HYP_PER_CPU(struct kvm_host_data, kvm_host_data);
1347 
1348 /*
1349  * How we access per-CPU host data depends on the where we access it from,
1350  * and the mode we're in:
1351  *
1352  * - VHE and nVHE hypervisor bits use their locally defined instance
1353  *
1354  * - the rest of the kernel use either the VHE or nVHE one, depending on
1355  *   the mode we're running in.
1356  *
1357  *   Unless we're in protected mode, fully deprivileged, and the nVHE
1358  *   per-CPU stuff is exclusively accessible to the protected EL2 code.
1359  *   In this case, the EL1 code uses the *VHE* data as its private state
1360  *   (which makes sense in a way as there shouldn't be any shared state
1361  *   between the host and the hypervisor).
1362  *
1363  * Yes, this is all totally trivial. Shoot me now.
1364  */
1365 #if defined(__KVM_NVHE_HYPERVISOR__) || defined(__KVM_VHE_HYPERVISOR__)
1366 #define host_data_ptr(f)	(&this_cpu_ptr(&kvm_host_data)->f)
1367 #else
1368 #define host_data_ptr(f)						\
1369 	(static_branch_unlikely(&kvm_protected_mode_initialized) ?	\
1370 	 &this_cpu_ptr(&kvm_host_data)->f :				\
1371 	 &this_cpu_ptr_hyp_sym(kvm_host_data)->f)
1372 #endif
1373 
1374 #define host_data_test_flag(flag)					\
1375 	(test_bit(KVM_HOST_DATA_FLAG_##flag, host_data_ptr(flags)))
1376 #define host_data_set_flag(flag)					\
1377 	set_bit(KVM_HOST_DATA_FLAG_##flag, host_data_ptr(flags))
1378 #define host_data_clear_flag(flag)					\
1379 	clear_bit(KVM_HOST_DATA_FLAG_##flag, host_data_ptr(flags))
1380 
1381 /* Check whether the FP regs are owned by the guest */
1382 static inline bool guest_owns_fp_regs(void)
1383 {
1384 	return *host_data_ptr(fp_owner) == FP_STATE_GUEST_OWNED;
1385 }
1386 
1387 /* Check whether the FP regs are owned by the host */
1388 static inline bool host_owns_fp_regs(void)
1389 {
1390 	return *host_data_ptr(fp_owner) == FP_STATE_HOST_OWNED;
1391 }
1392 
1393 static inline void kvm_init_host_cpu_context(struct kvm_cpu_context *cpu_ctxt)
1394 {
1395 	/* The host's MPIDR is immutable, so let's set it up at boot time */
1396 	ctxt_sys_reg(cpu_ctxt, MPIDR_EL1) = read_cpuid_mpidr();
1397 }
1398 
1399 static inline bool kvm_system_needs_idmapped_vectors(void)
1400 {
1401 	return cpus_have_final_cap(ARM64_SPECTRE_V3A);
1402 }
1403 
1404 void kvm_init_host_debug_data(void);
1405 void kvm_debug_init_vhe(void);
1406 void kvm_vcpu_load_debug(struct kvm_vcpu *vcpu);
1407 void kvm_vcpu_put_debug(struct kvm_vcpu *vcpu);
1408 void kvm_debug_set_guest_ownership(struct kvm_vcpu *vcpu);
1409 void kvm_debug_handle_oslar(struct kvm_vcpu *vcpu, u64 val);
1410 
1411 #define kvm_vcpu_os_lock_enabled(vcpu)		\
1412 	(!!(__vcpu_sys_reg(vcpu, OSLSR_EL1) & OSLSR_EL1_OSLK))
1413 
1414 #define kvm_debug_regs_in_use(vcpu)		\
1415 	((vcpu)->arch.debug_owner != VCPU_DEBUG_FREE)
1416 #define kvm_host_owns_debug_regs(vcpu)		\
1417 	((vcpu)->arch.debug_owner == VCPU_DEBUG_HOST_OWNED)
1418 #define kvm_guest_owns_debug_regs(vcpu)		\
1419 	((vcpu)->arch.debug_owner == VCPU_DEBUG_GUEST_OWNED)
1420 
1421 int kvm_arm_vcpu_arch_set_attr(struct kvm_vcpu *vcpu,
1422 			       struct kvm_device_attr *attr);
1423 int kvm_arm_vcpu_arch_get_attr(struct kvm_vcpu *vcpu,
1424 			       struct kvm_device_attr *attr);
1425 int kvm_arm_vcpu_arch_has_attr(struct kvm_vcpu *vcpu,
1426 			       struct kvm_device_attr *attr);
1427 
1428 int kvm_vm_ioctl_mte_copy_tags(struct kvm *kvm,
1429 			       struct kvm_arm_copy_mte_tags *copy_tags);
1430 int kvm_vm_ioctl_set_counter_offset(struct kvm *kvm,
1431 				    struct kvm_arm_counter_offset *offset);
1432 int kvm_vm_ioctl_get_reg_writable_masks(struct kvm *kvm,
1433 					struct reg_mask_range *range);
1434 
1435 /* Guest/host FPSIMD coordination helpers */
1436 void kvm_arch_vcpu_load_fp(struct kvm_vcpu *vcpu);
1437 void kvm_arch_vcpu_ctxflush_fp(struct kvm_vcpu *vcpu);
1438 void kvm_arch_vcpu_ctxsync_fp(struct kvm_vcpu *vcpu);
1439 void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu);
1440 
1441 static inline bool kvm_pmu_counter_deferred(struct perf_event_attr *attr)
1442 {
1443 	return (!has_vhe() && attr->exclude_host);
1444 }
1445 
1446 #ifdef CONFIG_KVM
1447 void kvm_set_pmu_events(u64 set, struct perf_event_attr *attr);
1448 void kvm_clr_pmu_events(u64 clr);
1449 bool kvm_set_pmuserenr(u64 val);
1450 void kvm_enable_trbe(void);
1451 void kvm_disable_trbe(void);
1452 void kvm_tracing_set_el1_configuration(u64 trfcr_while_in_guest);
1453 #else
1454 static inline void kvm_set_pmu_events(u64 set, struct perf_event_attr *attr) {}
1455 static inline void kvm_clr_pmu_events(u64 clr) {}
1456 static inline bool kvm_set_pmuserenr(u64 val)
1457 {
1458 	return false;
1459 }
1460 static inline void kvm_enable_trbe(void) {}
1461 static inline void kvm_disable_trbe(void) {}
1462 static inline void kvm_tracing_set_el1_configuration(u64 trfcr_while_in_guest) {}
1463 #endif
1464 
1465 void kvm_vcpu_load_vhe(struct kvm_vcpu *vcpu);
1466 void kvm_vcpu_put_vhe(struct kvm_vcpu *vcpu);
1467 
1468 int __init kvm_set_ipa_limit(void);
1469 u32 kvm_get_pa_bits(struct kvm *kvm);
1470 
1471 #define __KVM_HAVE_ARCH_VM_ALLOC
1472 struct kvm *kvm_arch_alloc_vm(void);
1473 
1474 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS
1475 
1476 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS_RANGE
1477 
1478 #define kvm_vm_is_protected(kvm)	(is_protected_kvm_enabled() && (kvm)->arch.pkvm.is_protected)
1479 
1480 #define vcpu_is_protected(vcpu)		kvm_vm_is_protected((vcpu)->kvm)
1481 
1482 int kvm_arm_vcpu_finalize(struct kvm_vcpu *vcpu, int feature);
1483 bool kvm_arm_vcpu_is_finalized(struct kvm_vcpu *vcpu);
1484 
1485 #define kvm_arm_vcpu_sve_finalized(vcpu) vcpu_get_flag(vcpu, VCPU_SVE_FINALIZED)
1486 
1487 #define kvm_has_mte(kvm)					\
1488 	(system_supports_mte() &&				\
1489 	 test_bit(KVM_ARCH_FLAG_MTE_ENABLED, &(kvm)->arch.flags))
1490 
1491 #define kvm_supports_32bit_el0()				\
1492 	(system_supports_32bit_el0() &&				\
1493 	 !static_branch_unlikely(&arm64_mismatched_32bit_el0))
1494 
1495 #define kvm_vm_has_ran_once(kvm)					\
1496 	(test_bit(KVM_ARCH_FLAG_HAS_RAN_ONCE, &(kvm)->arch.flags))
1497 
1498 static inline bool __vcpu_has_feature(const struct kvm_arch *ka, int feature)
1499 {
1500 	return test_bit(feature, ka->vcpu_features);
1501 }
1502 
1503 #define kvm_vcpu_has_feature(k, f)	__vcpu_has_feature(&(k)->arch, (f))
1504 #define vcpu_has_feature(v, f)	__vcpu_has_feature(&(v)->kvm->arch, (f))
1505 
1506 #define kvm_vcpu_initialized(v) vcpu_get_flag(vcpu, VCPU_INITIALIZED)
1507 
1508 int kvm_trng_call(struct kvm_vcpu *vcpu);
1509 #ifdef CONFIG_KVM
1510 extern phys_addr_t hyp_mem_base;
1511 extern phys_addr_t hyp_mem_size;
1512 void __init kvm_hyp_reserve(void);
1513 #else
1514 static inline void kvm_hyp_reserve(void) { }
1515 #endif
1516 
1517 void kvm_arm_vcpu_power_off(struct kvm_vcpu *vcpu);
1518 bool kvm_arm_vcpu_stopped(struct kvm_vcpu *vcpu);
1519 
1520 static inline u64 *__vm_id_reg(struct kvm_arch *ka, u32 reg)
1521 {
1522 	switch (reg) {
1523 	case sys_reg(3, 0, 0, 1, 0) ... sys_reg(3, 0, 0, 7, 7):
1524 		return &ka->id_regs[IDREG_IDX(reg)];
1525 	case SYS_CTR_EL0:
1526 		return &ka->ctr_el0;
1527 	case SYS_MIDR_EL1:
1528 		return &ka->midr_el1;
1529 	case SYS_REVIDR_EL1:
1530 		return &ka->revidr_el1;
1531 	case SYS_AIDR_EL1:
1532 		return &ka->aidr_el1;
1533 	default:
1534 		WARN_ON_ONCE(1);
1535 		return NULL;
1536 	}
1537 }
1538 
1539 #define kvm_read_vm_id_reg(kvm, reg)					\
1540 	({ u64 __val = *__vm_id_reg(&(kvm)->arch, reg); __val; })
1541 
1542 void kvm_set_vm_id_reg(struct kvm *kvm, u32 reg, u64 val);
1543 
1544 #define __expand_field_sign_unsigned(id, fld, val)			\
1545 	((u64)SYS_FIELD_VALUE(id, fld, val))
1546 
1547 #define __expand_field_sign_signed(id, fld, val)			\
1548 	({								\
1549 		u64 __val = SYS_FIELD_VALUE(id, fld, val);		\
1550 		sign_extend64(__val, id##_##fld##_WIDTH - 1);		\
1551 	})
1552 
1553 #define get_idreg_field_unsigned(kvm, id, fld)				\
1554 	({								\
1555 		u64 __val = kvm_read_vm_id_reg((kvm), SYS_##id);	\
1556 		FIELD_GET(id##_##fld##_MASK, __val);			\
1557 	})
1558 
1559 #define get_idreg_field_signed(kvm, id, fld)				\
1560 	({								\
1561 		u64 __val = get_idreg_field_unsigned(kvm, id, fld);	\
1562 		sign_extend64(__val, id##_##fld##_WIDTH - 1);		\
1563 	})
1564 
1565 #define get_idreg_field_enum(kvm, id, fld)				\
1566 	get_idreg_field_unsigned(kvm, id, fld)
1567 
1568 #define kvm_cmp_feat_signed(kvm, id, fld, op, limit)			\
1569 	(get_idreg_field_signed((kvm), id, fld) op __expand_field_sign_signed(id, fld, limit))
1570 
1571 #define kvm_cmp_feat_unsigned(kvm, id, fld, op, limit)			\
1572 	(get_idreg_field_unsigned((kvm), id, fld) op __expand_field_sign_unsigned(id, fld, limit))
1573 
1574 #define kvm_cmp_feat(kvm, id, fld, op, limit)				\
1575 	(id##_##fld##_SIGNED ?						\
1576 	 kvm_cmp_feat_signed(kvm, id, fld, op, limit) :			\
1577 	 kvm_cmp_feat_unsigned(kvm, id, fld, op, limit))
1578 
1579 #define __kvm_has_feat(kvm, id, fld, limit)				\
1580 	kvm_cmp_feat(kvm, id, fld, >=, limit)
1581 
1582 #define kvm_has_feat(kvm, ...) __kvm_has_feat(kvm, __VA_ARGS__)
1583 
1584 #define __kvm_has_feat_enum(kvm, id, fld, val)				\
1585 	kvm_cmp_feat_unsigned(kvm, id, fld, ==, val)
1586 
1587 #define kvm_has_feat_enum(kvm, ...) __kvm_has_feat_enum(kvm, __VA_ARGS__)
1588 
1589 #define kvm_has_feat_range(kvm, id, fld, min, max)			\
1590 	(kvm_cmp_feat(kvm, id, fld, >=, min) &&				\
1591 	kvm_cmp_feat(kvm, id, fld, <=, max))
1592 
1593 /* Check for a given level of PAuth support */
1594 #define kvm_has_pauth(k, l)						\
1595 	({								\
1596 		bool pa, pi, pa3;					\
1597 									\
1598 		pa  = kvm_has_feat((k), ID_AA64ISAR1_EL1, APA, l);	\
1599 		pa &= kvm_has_feat((k), ID_AA64ISAR1_EL1, GPA, IMP);	\
1600 		pi  = kvm_has_feat((k), ID_AA64ISAR1_EL1, API, l);	\
1601 		pi &= kvm_has_feat((k), ID_AA64ISAR1_EL1, GPI, IMP);	\
1602 		pa3  = kvm_has_feat((k), ID_AA64ISAR2_EL1, APA3, l);	\
1603 		pa3 &= kvm_has_feat((k), ID_AA64ISAR2_EL1, GPA3, IMP);	\
1604 									\
1605 		(pa + pi + pa3) == 1;					\
1606 	})
1607 
1608 #define kvm_has_fpmr(k)					\
1609 	(system_supports_fpmr() &&			\
1610 	 kvm_has_feat((k), ID_AA64PFR2_EL1, FPMR, IMP))
1611 
1612 #define kvm_has_tcr2(k)				\
1613 	(kvm_has_feat((k), ID_AA64MMFR3_EL1, TCRX, IMP))
1614 
1615 #define kvm_has_s1pie(k)				\
1616 	(kvm_has_feat((k), ID_AA64MMFR3_EL1, S1PIE, IMP))
1617 
1618 #define kvm_has_s1poe(k)				\
1619 	(kvm_has_feat((k), ID_AA64MMFR3_EL1, S1POE, IMP))
1620 
1621 #define kvm_has_ras(k)					\
1622 	(kvm_has_feat((k), ID_AA64PFR0_EL1, RAS, IMP))
1623 
1624 #define kvm_has_sctlr2(k)				\
1625 	(kvm_has_feat((k), ID_AA64MMFR3_EL1, SCTLRX, IMP))
1626 
1627 static inline bool kvm_arch_has_irq_bypass(void)
1628 {
1629 	return true;
1630 }
1631 
1632 void compute_fgu(struct kvm *kvm, enum fgt_group_id fgt);
1633 struct resx get_reg_fixed_bits(struct kvm *kvm, enum vcpu_sysreg reg);
1634 void check_feature_map(void);
1635 void kvm_vcpu_load_fgt(struct kvm_vcpu *vcpu);
1636 
1637 static __always_inline enum fgt_group_id __fgt_reg_to_group_id(enum vcpu_sysreg reg)
1638 {
1639 	switch (reg) {
1640 	case HFGRTR_EL2:
1641 	case HFGWTR_EL2:
1642 		return HFGRTR_GROUP;
1643 	case HFGITR_EL2:
1644 		return HFGITR_GROUP;
1645 	case HDFGRTR_EL2:
1646 	case HDFGWTR_EL2:
1647 		return HDFGRTR_GROUP;
1648 	case HAFGRTR_EL2:
1649 		return HAFGRTR_GROUP;
1650 	case HFGRTR2_EL2:
1651 	case HFGWTR2_EL2:
1652 		return HFGRTR2_GROUP;
1653 	case HFGITR2_EL2:
1654 		return HFGITR2_GROUP;
1655 	case HDFGRTR2_EL2:
1656 	case HDFGWTR2_EL2:
1657 		return HDFGRTR2_GROUP;
1658 	default:
1659 		BUILD_BUG_ON(1);
1660 	}
1661 }
1662 
1663 #define vcpu_fgt(vcpu, reg)						\
1664 	({								\
1665 		enum fgt_group_id id = __fgt_reg_to_group_id(reg);	\
1666 		u64 *p;							\
1667 		switch (reg) {						\
1668 		case HFGWTR_EL2:					\
1669 		case HDFGWTR_EL2:					\
1670 		case HFGWTR2_EL2:					\
1671 		case HDFGWTR2_EL2:					\
1672 			p = &(vcpu)->arch.fgt[id].w;			\
1673 			break;						\
1674 		default:						\
1675 			p = &(vcpu)->arch.fgt[id].r;			\
1676 			break;						\
1677 		}							\
1678 									\
1679 		p;							\
1680 	})
1681 
1682 long kvm_get_cap_for_kvm_ioctl(unsigned int ioctl, long *ext);
1683 
1684 #endif /* __ARM64_KVM_HOST_H__ */
1685