xref: /linux/arch/arm64/include/asm/kvm_host.h (revision 7c31c06e2d2d75859d773ba940e56d1db2bd1fcd)
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 	ICH_HFGRTR_GROUP,
291 	ICH_HFGWTR_GROUP = ICH_HFGRTR_GROUP,
292 	ICH_HFGITR_GROUP,
293 
294 	/* Must be last */
295 	__NR_FGT_GROUP_IDS__
296 };
297 
298 struct kvm_arch {
299 	struct kvm_s2_mmu mmu;
300 
301 	/*
302 	 * Fine-Grained UNDEF, mimicking the FGT layout defined by the
303 	 * architecture. We track them globally, as we present the
304 	 * same feature-set to all vcpus.
305 	 *
306 	 * Index 0 is currently spare.
307 	 */
308 	u64 fgu[__NR_FGT_GROUP_IDS__];
309 
310 	/*
311 	 * Stage 2 paging state for VMs with nested S2 using a virtual
312 	 * VMID.
313 	 */
314 	struct kvm_s2_mmu *nested_mmus;
315 	size_t nested_mmus_size;
316 	int nested_mmus_next;
317 
318 	/* Interrupt controller */
319 	struct vgic_dist	vgic;
320 
321 	/* Timers */
322 	struct arch_timer_vm_data timer_data;
323 
324 	/* Mandated version of PSCI */
325 	u32 psci_version;
326 
327 	/* Protects VM-scoped configuration data */
328 	struct mutex config_lock;
329 
330 	/*
331 	 * If we encounter a data abort without valid instruction syndrome
332 	 * information, report this to user space.  User space can (and
333 	 * should) opt in to this feature if KVM_CAP_ARM_NISV_TO_USER is
334 	 * supported.
335 	 */
336 #define KVM_ARCH_FLAG_RETURN_NISV_IO_ABORT_TO_USER	0
337 	/* Memory Tagging Extension enabled for the guest */
338 #define KVM_ARCH_FLAG_MTE_ENABLED			1
339 	/* At least one vCPU has ran in the VM */
340 #define KVM_ARCH_FLAG_HAS_RAN_ONCE			2
341 	/* The vCPU feature set for the VM is configured */
342 #define KVM_ARCH_FLAG_VCPU_FEATURES_CONFIGURED		3
343 	/* PSCI SYSTEM_SUSPEND enabled for the guest */
344 #define KVM_ARCH_FLAG_SYSTEM_SUSPEND_ENABLED		4
345 	/* VM counter offset */
346 #define KVM_ARCH_FLAG_VM_COUNTER_OFFSET			5
347 	/* Timer PPIs made immutable */
348 #define KVM_ARCH_FLAG_TIMER_PPIS_IMMUTABLE		6
349 	/* Initial ID reg values loaded */
350 #define KVM_ARCH_FLAG_ID_REGS_INITIALIZED		7
351 	/* Fine-Grained UNDEF initialised */
352 #define KVM_ARCH_FLAG_FGU_INITIALIZED			8
353 	/* SVE exposed to guest */
354 #define KVM_ARCH_FLAG_GUEST_HAS_SVE			9
355 	/* MIDR_EL1, REVIDR_EL1, and AIDR_EL1 are writable from userspace */
356 #define KVM_ARCH_FLAG_WRITABLE_IMP_ID_REGS		10
357 	/* Unhandled SEAs are taken to userspace */
358 #define KVM_ARCH_FLAG_EXIT_SEA				11
359 	unsigned long flags;
360 
361 	/* VM-wide vCPU feature set */
362 	DECLARE_BITMAP(vcpu_features, KVM_VCPU_MAX_FEATURES);
363 
364 	/* MPIDR to vcpu index mapping, optional */
365 	struct kvm_mpidr_data *mpidr_data;
366 
367 	/*
368 	 * VM-wide PMU filter, implemented as a bitmap and big enough for
369 	 * up to 2^10 events (ARMv8.0) or 2^16 events (ARMv8.1+).
370 	 */
371 	unsigned long *pmu_filter;
372 	struct arm_pmu *arm_pmu;
373 
374 	cpumask_var_t supported_cpus;
375 
376 	/* Maximum number of counters for the guest */
377 	u8 nr_pmu_counters;
378 
379 	/* Hypercall features firmware registers' descriptor */
380 	struct kvm_smccc_features smccc_feat;
381 	struct maple_tree smccc_filter;
382 
383 	/*
384 	 * Emulated CPU ID registers per VM
385 	 * (Op0, Op1, CRn, CRm, Op2) of the ID registers to be saved in it
386 	 * is (3, 0, 0, crm, op2), where 1<=crm<8, 0<=op2<8.
387 	 *
388 	 * These emulated idregs are VM-wide, but accessed from the context of a vCPU.
389 	 * Atomic access to multiple idregs are guarded by kvm_arch.config_lock.
390 	 */
391 #define IDREG_IDX(id)		(((sys_reg_CRm(id) - 1) << 3) | sys_reg_Op2(id))
392 #define KVM_ARM_ID_REG_NUM	(IDREG_IDX(sys_reg(3, 0, 0, 7, 7)) + 1)
393 	u64 id_regs[KVM_ARM_ID_REG_NUM];
394 
395 	u64 midr_el1;
396 	u64 revidr_el1;
397 	u64 aidr_el1;
398 	u64 ctr_el0;
399 
400 	/* Masks for VNCR-backed and general EL2 sysregs */
401 	struct kvm_sysreg_masks	*sysreg_masks;
402 
403 	/* Count the number of VNCR_EL2 currently mapped */
404 	atomic_t vncr_map_count;
405 
406 	/*
407 	 * For an untrusted host VM, 'pkvm.handle' is used to lookup
408 	 * the associated pKVM instance in the hypervisor.
409 	 */
410 	struct kvm_protected_vm pkvm;
411 };
412 
413 struct kvm_vcpu_fault_info {
414 	u64 esr_el2;		/* Hyp Syndrom Register */
415 	u64 far_el2;		/* Hyp Fault Address Register */
416 	u64 hpfar_el2;		/* Hyp IPA Fault Address Register */
417 	u64 disr_el1;		/* Deferred [SError] Status Register */
418 };
419 
420 /*
421  * VNCR() just places the VNCR_capable registers in the enum after
422  * __VNCR_START__, and the value (after correction) to be an 8-byte offset
423  * from the VNCR base. As we don't require the enum to be otherwise ordered,
424  * we need the terrible hack below to ensure that we correctly size the
425  * sys_regs array, no matter what.
426  *
427  * The __MAX__ macro has been lifted from Sean Eron Anderson's wonderful
428  * treasure trove of bit hacks:
429  * https://graphics.stanford.edu/~seander/bithacks.html#IntegerMinOrMax
430  */
431 #define __MAX__(x,y)	((x) ^ (((x) ^ (y)) & -((x) < (y))))
432 #define VNCR(r)						\
433 	__before_##r,					\
434 	r = __VNCR_START__ + ((VNCR_ ## r) / 8),	\
435 	__after_##r = __MAX__(__before_##r - 1, r)
436 
437 #define MARKER(m)				\
438 	m, __after_##m = m - 1
439 
440 enum vcpu_sysreg {
441 	__INVALID_SYSREG__,   /* 0 is reserved as an invalid value */
442 	MPIDR_EL1,	/* MultiProcessor Affinity Register */
443 	CLIDR_EL1,	/* Cache Level ID Register */
444 	CSSELR_EL1,	/* Cache Size Selection Register */
445 	TPIDR_EL0,	/* Thread ID, User R/W */
446 	TPIDRRO_EL0,	/* Thread ID, User R/O */
447 	TPIDR_EL1,	/* Thread ID, Privileged */
448 	CNTKCTL_EL1,	/* Timer Control Register (EL1) */
449 	PAR_EL1,	/* Physical Address Register */
450 	MDCCINT_EL1,	/* Monitor Debug Comms Channel Interrupt Enable Reg */
451 	OSLSR_EL1,	/* OS Lock Status Register */
452 	DISR_EL1,	/* Deferred Interrupt Status Register */
453 
454 	/* Performance Monitors Registers */
455 	PMCR_EL0,	/* Control Register */
456 	PMSELR_EL0,	/* Event Counter Selection Register */
457 	PMEVCNTR0_EL0,	/* Event Counter Register (0-30) */
458 	PMEVCNTR30_EL0 = PMEVCNTR0_EL0 + 30,
459 	PMCCNTR_EL0,	/* Cycle Counter Register */
460 	PMEVTYPER0_EL0,	/* Event Type Register (0-30) */
461 	PMEVTYPER30_EL0 = PMEVTYPER0_EL0 + 30,
462 	PMCCFILTR_EL0,	/* Cycle Count Filter Register */
463 	PMCNTENSET_EL0,	/* Count Enable Set Register */
464 	PMINTENSET_EL1,	/* Interrupt Enable Set Register */
465 	PMOVSSET_EL0,	/* Overflow Flag Status Set Register */
466 	PMUSERENR_EL0,	/* User Enable Register */
467 
468 	/* Pointer Authentication Registers in a strict increasing order. */
469 	APIAKEYLO_EL1,
470 	APIAKEYHI_EL1,
471 	APIBKEYLO_EL1,
472 	APIBKEYHI_EL1,
473 	APDAKEYLO_EL1,
474 	APDAKEYHI_EL1,
475 	APDBKEYLO_EL1,
476 	APDBKEYHI_EL1,
477 	APGAKEYLO_EL1,
478 	APGAKEYHI_EL1,
479 
480 	/* Memory Tagging Extension registers */
481 	RGSR_EL1,	/* Random Allocation Tag Seed Register */
482 	GCR_EL1,	/* Tag Control Register */
483 	TFSRE0_EL1,	/* Tag Fault Status Register (EL0) */
484 
485 	POR_EL0,	/* Permission Overlay Register 0 (EL0) */
486 
487 	/* FP/SIMD/SVE */
488 	SVCR,
489 	FPMR,
490 
491 	/* 32bit specific registers. */
492 	DACR32_EL2,	/* Domain Access Control Register */
493 	IFSR32_EL2,	/* Instruction Fault Status Register */
494 	FPEXC32_EL2,	/* Floating-Point Exception Control Register */
495 	DBGVCR32_EL2,	/* Debug Vector Catch Register */
496 
497 	/* EL2 registers */
498 	ACTLR_EL2,	/* Auxiliary Control Register (EL2) */
499 	CPTR_EL2,	/* Architectural Feature Trap Register (EL2) */
500 	HACR_EL2,	/* Hypervisor Auxiliary Control Register */
501 	ZCR_EL2,	/* SVE Control Register (EL2) */
502 	TTBR0_EL2,	/* Translation Table Base Register 0 (EL2) */
503 	TTBR1_EL2,	/* Translation Table Base Register 1 (EL2) */
504 	TCR_EL2,	/* Translation Control Register (EL2) */
505 	PIRE0_EL2,	/* Permission Indirection Register 0 (EL2) */
506 	PIR_EL2,	/* Permission Indirection Register 1 (EL2) */
507 	POR_EL2,	/* Permission Overlay Register 2 (EL2) */
508 	SPSR_EL2,	/* EL2 saved program status register */
509 	ELR_EL2,	/* EL2 exception link register */
510 	AFSR0_EL2,	/* Auxiliary Fault Status Register 0 (EL2) */
511 	AFSR1_EL2,	/* Auxiliary Fault Status Register 1 (EL2) */
512 	ESR_EL2,	/* Exception Syndrome Register (EL2) */
513 	FAR_EL2,	/* Fault Address Register (EL2) */
514 	HPFAR_EL2,	/* Hypervisor IPA Fault Address Register */
515 	MAIR_EL2,	/* Memory Attribute Indirection Register (EL2) */
516 	AMAIR_EL2,	/* Auxiliary Memory Attribute Indirection Register (EL2) */
517 	VBAR_EL2,	/* Vector Base Address Register (EL2) */
518 	RVBAR_EL2,	/* Reset Vector Base Address Register */
519 	CONTEXTIDR_EL2,	/* Context ID Register (EL2) */
520 	SP_EL2,		/* EL2 Stack Pointer */
521 	CNTHP_CTL_EL2,
522 	CNTHP_CVAL_EL2,
523 	CNTHV_CTL_EL2,
524 	CNTHV_CVAL_EL2,
525 
526 	/* Anything from this can be RES0/RES1 sanitised */
527 	MARKER(__SANITISED_REG_START__),
528 	SCTLR_EL2,	/* System Control Register (EL2) */
529 	TCR2_EL2,	/* Extended Translation Control Register (EL2) */
530 	SCTLR2_EL2,	/* System Control Register 2 (EL2) */
531 	MDCR_EL2,	/* Monitor Debug Configuration Register (EL2) */
532 	CNTHCTL_EL2,	/* Counter-timer Hypervisor Control register */
533 
534 	/* Any VNCR-capable reg goes after this point */
535 	MARKER(__VNCR_START__),
536 
537 	VNCR(SCTLR_EL1),/* System Control Register */
538 	VNCR(ACTLR_EL1),/* Auxiliary Control Register */
539 	VNCR(CPACR_EL1),/* Coprocessor Access Control */
540 	VNCR(ZCR_EL1),	/* SVE Control */
541 	VNCR(TTBR0_EL1),/* Translation Table Base Register 0 */
542 	VNCR(TTBR1_EL1),/* Translation Table Base Register 1 */
543 	VNCR(TCR_EL1),	/* Translation Control Register */
544 	VNCR(TCR2_EL1),	/* Extended Translation Control Register */
545 	VNCR(SCTLR2_EL1), /* System Control Register 2 */
546 	VNCR(ESR_EL1),	/* Exception Syndrome Register */
547 	VNCR(AFSR0_EL1),/* Auxiliary Fault Status Register 0 */
548 	VNCR(AFSR1_EL1),/* Auxiliary Fault Status Register 1 */
549 	VNCR(FAR_EL1),	/* Fault Address Register */
550 	VNCR(MAIR_EL1),	/* Memory Attribute Indirection Register */
551 	VNCR(VBAR_EL1),	/* Vector Base Address Register */
552 	VNCR(CONTEXTIDR_EL1),	/* Context ID Register */
553 	VNCR(AMAIR_EL1),/* Aux Memory Attribute Indirection Register */
554 	VNCR(MDSCR_EL1),/* Monitor Debug System Control Register */
555 	VNCR(ELR_EL1),
556 	VNCR(SP_EL1),
557 	VNCR(SPSR_EL1),
558 	VNCR(TFSR_EL1),	/* Tag Fault Status Register (EL1) */
559 	VNCR(VPIDR_EL2),/* Virtualization Processor ID Register */
560 	VNCR(VMPIDR_EL2),/* Virtualization Multiprocessor ID Register */
561 	VNCR(HCR_EL2),	/* Hypervisor Configuration Register */
562 	VNCR(HSTR_EL2),	/* Hypervisor System Trap Register */
563 	VNCR(VTTBR_EL2),/* Virtualization Translation Table Base Register */
564 	VNCR(VTCR_EL2),	/* Virtualization Translation Control Register */
565 	VNCR(TPIDR_EL2),/* EL2 Software Thread ID Register */
566 	VNCR(HCRX_EL2),	/* Extended Hypervisor Configuration Register */
567 
568 	/* Permission Indirection Extension registers */
569 	VNCR(PIR_EL1),	 /* Permission Indirection Register 1 (EL1) */
570 	VNCR(PIRE0_EL1), /*  Permission Indirection Register 0 (EL1) */
571 
572 	VNCR(POR_EL1),	/* Permission Overlay Register 1 (EL1) */
573 
574 	/* FEAT_RAS registers */
575 	VNCR(VDISR_EL2),
576 	VNCR(VSESR_EL2),
577 
578 	VNCR(HFGRTR_EL2),
579 	VNCR(HFGWTR_EL2),
580 	VNCR(HFGITR_EL2),
581 	VNCR(HDFGRTR_EL2),
582 	VNCR(HDFGWTR_EL2),
583 	VNCR(HAFGRTR_EL2),
584 	VNCR(HFGRTR2_EL2),
585 	VNCR(HFGWTR2_EL2),
586 	VNCR(HFGITR2_EL2),
587 	VNCR(HDFGRTR2_EL2),
588 	VNCR(HDFGWTR2_EL2),
589 
590 	VNCR(VNCR_EL2),
591 
592 	VNCR(CNTVOFF_EL2),
593 	VNCR(CNTV_CVAL_EL0),
594 	VNCR(CNTV_CTL_EL0),
595 	VNCR(CNTP_CVAL_EL0),
596 	VNCR(CNTP_CTL_EL0),
597 
598 	VNCR(ICH_LR0_EL2),
599 	VNCR(ICH_LR1_EL2),
600 	VNCR(ICH_LR2_EL2),
601 	VNCR(ICH_LR3_EL2),
602 	VNCR(ICH_LR4_EL2),
603 	VNCR(ICH_LR5_EL2),
604 	VNCR(ICH_LR6_EL2),
605 	VNCR(ICH_LR7_EL2),
606 	VNCR(ICH_LR8_EL2),
607 	VNCR(ICH_LR9_EL2),
608 	VNCR(ICH_LR10_EL2),
609 	VNCR(ICH_LR11_EL2),
610 	VNCR(ICH_LR12_EL2),
611 	VNCR(ICH_LR13_EL2),
612 	VNCR(ICH_LR14_EL2),
613 	VNCR(ICH_LR15_EL2),
614 
615 	VNCR(ICH_AP0R0_EL2),
616 	VNCR(ICH_AP0R1_EL2),
617 	VNCR(ICH_AP0R2_EL2),
618 	VNCR(ICH_AP0R3_EL2),
619 	VNCR(ICH_AP1R0_EL2),
620 	VNCR(ICH_AP1R1_EL2),
621 	VNCR(ICH_AP1R2_EL2),
622 	VNCR(ICH_AP1R3_EL2),
623 	VNCR(ICH_HCR_EL2),
624 	VNCR(ICH_VMCR_EL2),
625 
626 	VNCR(ICH_HFGRTR_EL2),
627 	VNCR(ICH_HFGWTR_EL2),
628 	VNCR(ICH_HFGITR_EL2),
629 
630 	NR_SYS_REGS	/* Nothing after this line! */
631 };
632 
633 struct resx {
634 	u64	res0;
635 	u64	res1;
636 };
637 
638 struct kvm_sysreg_masks {
639 	struct resx mask[NR_SYS_REGS - __SANITISED_REG_START__];
640 };
641 
642 static inline struct resx __kvm_get_sysreg_resx(struct kvm_arch *arch,
643 						enum vcpu_sysreg sr)
644 {
645 	struct kvm_sysreg_masks *masks;
646 
647 	masks = arch->sysreg_masks;
648 	if (likely(masks &&
649 		   sr >= __SANITISED_REG_START__ && sr < NR_SYS_REGS))
650 		return masks->mask[sr - __SANITISED_REG_START__];
651 
652 	return (struct resx){};
653 }
654 
655 #define kvm_get_sysreg_resx(k, sr) __kvm_get_sysreg_resx(&(k)->arch, (sr))
656 
657 static inline void __kvm_set_sysreg_resx(struct kvm_arch *arch,
658 					 enum vcpu_sysreg sr, struct resx resx)
659 {
660 	arch->sysreg_masks->mask[sr - __SANITISED_REG_START__] = resx;
661 }
662 
663 #define kvm_set_sysreg_resx(k, sr, resx)		\
664 	__kvm_set_sysreg_resx(&(k)->arch, (sr), (resx))
665 
666 struct fgt_masks {
667 	const char	*str;
668 	u64		mask;
669 	u64		nmask;
670 	u64		res0;
671 	u64		res1;
672 };
673 
674 extern struct fgt_masks hfgrtr_masks;
675 extern struct fgt_masks hfgwtr_masks;
676 extern struct fgt_masks hfgitr_masks;
677 extern struct fgt_masks hdfgrtr_masks;
678 extern struct fgt_masks hdfgwtr_masks;
679 extern struct fgt_masks hafgrtr_masks;
680 extern struct fgt_masks hfgrtr2_masks;
681 extern struct fgt_masks hfgwtr2_masks;
682 extern struct fgt_masks hfgitr2_masks;
683 extern struct fgt_masks hdfgrtr2_masks;
684 extern struct fgt_masks hdfgwtr2_masks;
685 extern struct fgt_masks ich_hfgrtr_masks;
686 extern struct fgt_masks ich_hfgwtr_masks;
687 extern struct fgt_masks ich_hfgitr_masks;
688 
689 extern struct fgt_masks kvm_nvhe_sym(hfgrtr_masks);
690 extern struct fgt_masks kvm_nvhe_sym(hfgwtr_masks);
691 extern struct fgt_masks kvm_nvhe_sym(hfgitr_masks);
692 extern struct fgt_masks kvm_nvhe_sym(hdfgrtr_masks);
693 extern struct fgt_masks kvm_nvhe_sym(hdfgwtr_masks);
694 extern struct fgt_masks kvm_nvhe_sym(hafgrtr_masks);
695 extern struct fgt_masks kvm_nvhe_sym(hfgrtr2_masks);
696 extern struct fgt_masks kvm_nvhe_sym(hfgwtr2_masks);
697 extern struct fgt_masks kvm_nvhe_sym(hfgitr2_masks);
698 extern struct fgt_masks kvm_nvhe_sym(hdfgrtr2_masks);
699 extern struct fgt_masks kvm_nvhe_sym(hdfgwtr2_masks);
700 extern struct fgt_masks kvm_nvhe_sym(ich_hfgrtr_masks);
701 extern struct fgt_masks kvm_nvhe_sym(ich_hfgwtr_masks);
702 extern struct fgt_masks kvm_nvhe_sym(ich_hfgitr_masks);
703 
704 struct kvm_cpu_context {
705 	struct user_pt_regs regs;	/* sp = sp_el0 */
706 
707 	u64	spsr_abt;
708 	u64	spsr_und;
709 	u64	spsr_irq;
710 	u64	spsr_fiq;
711 
712 	struct user_fpsimd_state fp_regs;
713 
714 	u64 sys_regs[NR_SYS_REGS];
715 
716 	struct kvm_vcpu *__hyp_running_vcpu;
717 
718 	/* This pointer has to be 4kB aligned. */
719 	u64 *vncr_array;
720 };
721 
722 struct cpu_sve_state {
723 	__u64 zcr_el1;
724 
725 	/*
726 	 * Ordering is important since __sve_save_state/__sve_restore_state
727 	 * relies on it.
728 	 */
729 	__u32 fpsr;
730 	__u32 fpcr;
731 
732 	/* Must be SVE_VQ_BYTES (128 bit) aligned. */
733 	__u8 sve_regs[];
734 };
735 
736 /*
737  * This structure is instantiated on a per-CPU basis, and contains
738  * data that is:
739  *
740  * - tied to a single physical CPU, and
741  * - either have a lifetime that does not extend past vcpu_put()
742  * - or is an invariant for the lifetime of the system
743  *
744  * Use host_data_ptr(field) as a way to access a pointer to such a
745  * field.
746  */
747 struct kvm_host_data {
748 #define KVM_HOST_DATA_FLAG_HAS_SPE			0
749 #define KVM_HOST_DATA_FLAG_HAS_TRBE			1
750 #define KVM_HOST_DATA_FLAG_TRBE_ENABLED			2
751 #define KVM_HOST_DATA_FLAG_EL1_TRACING_CONFIGURED	3
752 #define KVM_HOST_DATA_FLAG_VCPU_IN_HYP_CONTEXT		4
753 #define KVM_HOST_DATA_FLAG_L1_VNCR_MAPPED		5
754 #define KVM_HOST_DATA_FLAG_HAS_BRBE			6
755 	unsigned long flags;
756 
757 	struct kvm_cpu_context host_ctxt;
758 
759 	/*
760 	 * Hyp VA.
761 	 * sve_state is only used in pKVM and if system_supports_sve().
762 	 */
763 	struct cpu_sve_state *sve_state;
764 
765 	/* Used by pKVM only. */
766 	u64	fpmr;
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 		/* Self-hosted trace */
785 		u64 trfcr_el1;
786 		/* Values of trap registers for the host before guest entry. */
787 		u64 mdcr_el2;
788 		u64 brbcr_el1;
789 	} host_debug_state;
790 
791 	/* Guest trace filter value */
792 	u64 trfcr_while_in_guest;
793 
794 	/* Number of programmable event counters (PMCR_EL0.N) for this CPU */
795 	unsigned int nr_event_counters;
796 
797 	/* Number of debug breakpoints/watchpoints for this CPU (minus 1) */
798 	unsigned int debug_brps;
799 	unsigned int debug_wrps;
800 
801 	/* Last vgic_irq part of the AP list recorded in an LR */
802 	struct vgic_irq *last_lr_irq;
803 
804 	/* PPI state tracking for GICv5-based guests */
805 	struct {
806 		/*
807 		 * For tracking the PPI pending state, we need both the entry
808 		 * state and exit state to correctly detect edges as it is
809 		 * possible that an interrupt has been injected in software in
810 		 * the interim.
811 		 */
812 		DECLARE_BITMAP(pendr_entry, VGIC_V5_NR_PRIVATE_IRQS);
813 		DECLARE_BITMAP(pendr_exit, VGIC_V5_NR_PRIVATE_IRQS);
814 
815 		/* The saved state of the regs when leaving the guest */
816 		DECLARE_BITMAP(activer_exit, VGIC_V5_NR_PRIVATE_IRQS);
817 	} vgic_v5_ppi_state;
818 };
819 
820 struct kvm_host_psci_config {
821 	/* PSCI version used by host. */
822 	u32 version;
823 	u32 smccc_version;
824 
825 	/* Function IDs used by host if version is v0.1. */
826 	struct psci_0_1_function_ids function_ids_0_1;
827 
828 	bool psci_0_1_cpu_suspend_implemented;
829 	bool psci_0_1_cpu_on_implemented;
830 	bool psci_0_1_cpu_off_implemented;
831 	bool psci_0_1_migrate_implemented;
832 };
833 
834 extern struct kvm_host_psci_config kvm_nvhe_sym(kvm_host_psci_config);
835 #define kvm_host_psci_config CHOOSE_NVHE_SYM(kvm_host_psci_config)
836 
837 extern s64 kvm_nvhe_sym(hyp_physvirt_offset);
838 #define hyp_physvirt_offset CHOOSE_NVHE_SYM(hyp_physvirt_offset)
839 
840 extern u64 kvm_nvhe_sym(hyp_cpu_logical_map)[NR_CPUS];
841 #define hyp_cpu_logical_map CHOOSE_NVHE_SYM(hyp_cpu_logical_map)
842 
843 struct vcpu_reset_state {
844 	unsigned long	pc;
845 	unsigned long	r0;
846 	bool		be;
847 	bool		reset;
848 };
849 
850 struct vncr_tlb;
851 
852 struct kvm_vcpu_arch {
853 	struct kvm_cpu_context ctxt;
854 
855 	/*
856 	 * Guest floating point state
857 	 *
858 	 * The architecture has two main floating point extensions,
859 	 * the original FPSIMD and SVE.  These have overlapping
860 	 * register views, with the FPSIMD V registers occupying the
861 	 * low 128 bits of the SVE Z registers.  When the core
862 	 * floating point code saves the register state of a task it
863 	 * records which view it saved in fp_type.
864 	 */
865 	void *sve_state;
866 	enum fp_type fp_type;
867 	unsigned int sve_max_vl;
868 
869 	/* Stage 2 paging state used by the hardware on next switch */
870 	struct kvm_s2_mmu *hw_mmu;
871 
872 	/* Values of trap registers for the guest. */
873 	u64 hcr_el2;
874 	u64 hcrx_el2;
875 	u64 mdcr_el2;
876 
877 	struct {
878 		u64 r;
879 		u64 w;
880 	} fgt[__NR_FGT_GROUP_IDS__];
881 
882 	/* Exception Information */
883 	struct kvm_vcpu_fault_info fault;
884 
885 	/* Configuration flags, set once and for all before the vcpu can run */
886 	u8 cflags;
887 
888 	/* Input flags to the hypervisor code, potentially cleared after use */
889 	u8 iflags;
890 
891 	/* State flags for kernel bookkeeping, unused by the hypervisor code */
892 	u16 sflags;
893 
894 	/*
895 	 * Don't run the guest (internal implementation need).
896 	 *
897 	 * Contrary to the flags above, this is set/cleared outside of
898 	 * a vcpu context, and thus cannot be mixed with the flags
899 	 * themselves (or the flag accesses need to be made atomic).
900 	 */
901 	bool pause;
902 
903 	/*
904 	 * We maintain more than a single set of debug registers to support
905 	 * debugging the guest from the host and to maintain separate host and
906 	 * guest state during world switches. vcpu_debug_state are the debug
907 	 * registers of the vcpu as the guest sees them.
908 	 *
909 	 * external_debug_state contains the debug values we want to debug the
910 	 * guest. This is set via the KVM_SET_GUEST_DEBUG ioctl.
911 	 */
912 	struct kvm_guest_debug_arch vcpu_debug_state;
913 	struct kvm_guest_debug_arch external_debug_state;
914 	u64 external_mdscr_el1;
915 
916 	enum {
917 		VCPU_DEBUG_FREE,
918 		VCPU_DEBUG_HOST_OWNED,
919 		VCPU_DEBUG_GUEST_OWNED,
920 	} debug_owner;
921 
922 	/* VGIC state */
923 	struct vgic_cpu vgic_cpu;
924 	struct arch_timer_cpu timer_cpu;
925 	struct kvm_pmu pmu;
926 
927 	/* vcpu power state */
928 	struct kvm_mp_state mp_state;
929 	spinlock_t mp_state_lock;
930 
931 	/* Cache some mmu pages needed inside spinlock regions */
932 	struct kvm_mmu_memory_cache mmu_page_cache;
933 
934 	/* Pages to top-up the pKVM/EL2 guest pool */
935 	struct kvm_hyp_memcache pkvm_memcache;
936 
937 	/* Virtual SError ESR to restore when HCR_EL2.VSE is set */
938 	u64 vsesr_el2;
939 
940 	/* Additional reset state */
941 	struct vcpu_reset_state	reset_state;
942 
943 	/* Guest PV state */
944 	struct {
945 		u64 last_steal;
946 		gpa_t base;
947 	} steal;
948 
949 	/* Per-vcpu CCSIDR override or NULL */
950 	u32 *ccsidr;
951 
952 	/* Per-vcpu TLB for VNCR_EL2 -- NULL when !NV */
953 	struct vncr_tlb	*vncr_tlb;
954 };
955 
956 /*
957  * Each 'flag' is composed of a comma-separated triplet:
958  *
959  * - the flag-set it belongs to in the vcpu->arch structure
960  * - the value for that flag
961  * - the mask for that flag
962  *
963  *  __vcpu_single_flag() builds such a triplet for a single-bit flag.
964  * unpack_vcpu_flag() extract the flag value from the triplet for
965  * direct use outside of the flag accessors.
966  */
967 #define __vcpu_single_flag(_set, _f)	_set, (_f), (_f)
968 
969 #define __unpack_flag(_set, _f, _m)	_f
970 #define unpack_vcpu_flag(...)		__unpack_flag(__VA_ARGS__)
971 
972 #define __build_check_flag(v, flagset, f, m)			\
973 	do {							\
974 		typeof(v->arch.flagset) *_fset;			\
975 								\
976 		/* Check that the flags fit in the mask */	\
977 		BUILD_BUG_ON(HWEIGHT(m) != HWEIGHT((f) | (m)));	\
978 		/* Check that the flags fit in the type */	\
979 		BUILD_BUG_ON((sizeof(*_fset) * 8) <= __fls(m));	\
980 	} while (0)
981 
982 #define __vcpu_get_flag(v, flagset, f, m)			\
983 	({							\
984 		__build_check_flag(v, flagset, f, m);		\
985 								\
986 		READ_ONCE(v->arch.flagset) & (m);		\
987 	})
988 
989 /*
990  * Note that the set/clear accessors must be preempt-safe in order to
991  * avoid nesting them with load/put which also manipulate flags...
992  */
993 #ifdef __KVM_NVHE_HYPERVISOR__
994 /* the nVHE hypervisor is always non-preemptible */
995 #define __vcpu_flags_preempt_disable()
996 #define __vcpu_flags_preempt_enable()
997 #else
998 #define __vcpu_flags_preempt_disable()	preempt_disable()
999 #define __vcpu_flags_preempt_enable()	preempt_enable()
1000 #endif
1001 
1002 #define __vcpu_set_flag(v, flagset, f, m)			\
1003 	do {							\
1004 		typeof(v->arch.flagset) *fset;			\
1005 								\
1006 		__build_check_flag(v, flagset, f, m);		\
1007 								\
1008 		fset = &v->arch.flagset;			\
1009 		__vcpu_flags_preempt_disable();			\
1010 		if (HWEIGHT(m) > 1)				\
1011 			*fset &= ~(m);				\
1012 		*fset |= (f);					\
1013 		__vcpu_flags_preempt_enable();			\
1014 	} while (0)
1015 
1016 #define __vcpu_clear_flag(v, flagset, f, m)			\
1017 	do {							\
1018 		typeof(v->arch.flagset) *fset;			\
1019 								\
1020 		__build_check_flag(v, flagset, f, m);		\
1021 								\
1022 		fset = &v->arch.flagset;			\
1023 		__vcpu_flags_preempt_disable();			\
1024 		*fset &= ~(m);					\
1025 		__vcpu_flags_preempt_enable();			\
1026 	} while (0)
1027 
1028 #define __vcpu_test_and_clear_flag(v, flagset, f, m)		\
1029 	({							\
1030 		typeof(v->arch.flagset) set;			\
1031 								\
1032 		set = __vcpu_get_flag(v, flagset, f, m);	\
1033 		__vcpu_clear_flag(v, flagset, f, m);		\
1034 								\
1035 		set;						\
1036 	})
1037 
1038 #define vcpu_get_flag(v, ...)	__vcpu_get_flag((v), __VA_ARGS__)
1039 #define vcpu_set_flag(v, ...)	__vcpu_set_flag((v), __VA_ARGS__)
1040 #define vcpu_clear_flag(v, ...)	__vcpu_clear_flag((v), __VA_ARGS__)
1041 #define vcpu_test_and_clear_flag(v, ...)			\
1042 	__vcpu_test_and_clear_flag((v), __VA_ARGS__)
1043 
1044 /* KVM_ARM_VCPU_INIT completed */
1045 #define VCPU_INITIALIZED	__vcpu_single_flag(cflags, BIT(0))
1046 /* SVE config completed */
1047 #define VCPU_SVE_FINALIZED	__vcpu_single_flag(cflags, BIT(1))
1048 /* pKVM VCPU setup completed */
1049 #define VCPU_PKVM_FINALIZED	__vcpu_single_flag(cflags, BIT(2))
1050 
1051 /* Exception pending */
1052 #define PENDING_EXCEPTION	__vcpu_single_flag(iflags, BIT(0))
1053 /*
1054  * PC increment. Overlaps with EXCEPT_MASK on purpose so that it can't
1055  * be set together with an exception...
1056  */
1057 #define INCREMENT_PC		__vcpu_single_flag(iflags, BIT(1))
1058 /* Target EL/MODE (not a single flag, but let's abuse the macro) */
1059 #define EXCEPT_MASK		__vcpu_single_flag(iflags, GENMASK(3, 1))
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 
1105 
1106 /* Pointer to the vcpu's SVE FFR for sve_{save,load}_state() */
1107 #define vcpu_sve_pffr(vcpu) (kern_hyp_va((vcpu)->arch.sve_state) +	\
1108 			     sve_ffr_offset((vcpu)->arch.sve_max_vl))
1109 
1110 #define vcpu_sve_max_vq(vcpu)	sve_vq_from_vl((vcpu)->arch.sve_max_vl)
1111 
1112 #define vcpu_sve_zcr_elx(vcpu)						\
1113 	(unlikely(is_hyp_ctxt(vcpu)) ? ZCR_EL2 : ZCR_EL1)
1114 
1115 #define sve_state_size_from_vl(sve_max_vl) ({				\
1116 	size_t __size_ret;						\
1117 	unsigned int __vq;						\
1118 									\
1119 	if (WARN_ON(!sve_vl_valid(sve_max_vl))) {			\
1120 		__size_ret = 0;						\
1121 	} else {							\
1122 		__vq = sve_vq_from_vl(sve_max_vl);			\
1123 		__size_ret = SVE_SIG_REGS_SIZE(__vq);			\
1124 	}								\
1125 									\
1126 	__size_ret;							\
1127 })
1128 
1129 #define vcpu_sve_state_size(vcpu) sve_state_size_from_vl((vcpu)->arch.sve_max_vl)
1130 
1131 #define KVM_GUESTDBG_VALID_MASK (KVM_GUESTDBG_ENABLE | \
1132 				 KVM_GUESTDBG_USE_SW_BP | \
1133 				 KVM_GUESTDBG_USE_HW | \
1134 				 KVM_GUESTDBG_SINGLESTEP)
1135 
1136 #define kvm_has_sve(kvm)	(system_supports_sve() &&		\
1137 				 test_bit(KVM_ARCH_FLAG_GUEST_HAS_SVE, &(kvm)->arch.flags))
1138 
1139 #ifdef __KVM_NVHE_HYPERVISOR__
1140 #define vcpu_has_sve(vcpu)	kvm_has_sve(kern_hyp_va((vcpu)->kvm))
1141 #else
1142 #define vcpu_has_sve(vcpu)	kvm_has_sve((vcpu)->kvm)
1143 #endif
1144 
1145 #ifdef CONFIG_ARM64_PTR_AUTH
1146 #define vcpu_has_ptrauth(vcpu)						\
1147 	((cpus_have_final_cap(ARM64_HAS_ADDRESS_AUTH) ||		\
1148 	  cpus_have_final_cap(ARM64_HAS_GENERIC_AUTH)) &&		\
1149 	 (vcpu_has_feature(vcpu, KVM_ARM_VCPU_PTRAUTH_ADDRESS) ||       \
1150 	  vcpu_has_feature(vcpu, KVM_ARM_VCPU_PTRAUTH_GENERIC)))
1151 #else
1152 #define vcpu_has_ptrauth(vcpu)		false
1153 #endif
1154 
1155 #define vcpu_on_unsupported_cpu(vcpu)					\
1156 	vcpu_get_flag(vcpu, ON_UNSUPPORTED_CPU)
1157 
1158 #define vcpu_set_on_unsupported_cpu(vcpu)				\
1159 	vcpu_set_flag(vcpu, ON_UNSUPPORTED_CPU)
1160 
1161 #define vcpu_clear_on_unsupported_cpu(vcpu)				\
1162 	vcpu_clear_flag(vcpu, ON_UNSUPPORTED_CPU)
1163 
1164 #define vcpu_gp_regs(v)		(&(v)->arch.ctxt.regs)
1165 
1166 /*
1167  * Only use __vcpu_sys_reg/ctxt_sys_reg if you know you want the
1168  * memory backed version of a register, and not the one most recently
1169  * accessed by a running VCPU.  For example, for userspace access or
1170  * for system registers that are never context switched, but only
1171  * emulated.
1172  *
1173  * Don't bother with VNCR-based accesses in the nVHE code, it has no
1174  * business dealing with NV.
1175  */
1176 static inline u64 *___ctxt_sys_reg(const struct kvm_cpu_context *ctxt, int r)
1177 {
1178 #if !defined (__KVM_NVHE_HYPERVISOR__)
1179 	if (unlikely(cpus_have_final_cap(ARM64_HAS_NESTED_VIRT) &&
1180 		     r >= __VNCR_START__ && ctxt->vncr_array))
1181 		return &ctxt->vncr_array[r - __VNCR_START__];
1182 #endif
1183 	return (u64 *)&ctxt->sys_regs[r];
1184 }
1185 
1186 #define __ctxt_sys_reg(c,r)						\
1187 	({								\
1188 		BUILD_BUG_ON(__builtin_constant_p(r) &&			\
1189 			     (r) >= NR_SYS_REGS);			\
1190 		___ctxt_sys_reg(c, r);					\
1191 	})
1192 
1193 #define ctxt_sys_reg(c,r)	(*__ctxt_sys_reg(c,r))
1194 
1195 u64 kvm_vcpu_apply_reg_masks(const struct kvm_vcpu *, enum vcpu_sysreg, u64);
1196 
1197 #define __vcpu_assign_sys_reg(v, r, val)				\
1198 	do {								\
1199 		const struct kvm_cpu_context *ctxt = &(v)->arch.ctxt;	\
1200 		u64 __v = (val);					\
1201 		if (vcpu_has_nv((v)) && (r) >= __SANITISED_REG_START__)	\
1202 			__v = kvm_vcpu_apply_reg_masks((v), (r), __v);	\
1203 									\
1204 		ctxt_sys_reg(ctxt, (r)) = __v;				\
1205 	} while (0)
1206 
1207 #define __vcpu_rmw_sys_reg(v, r, op, val)				\
1208 	do {								\
1209 		const struct kvm_cpu_context *ctxt = &(v)->arch.ctxt;	\
1210 		u64 __v = ctxt_sys_reg(ctxt, (r));			\
1211 		__v op (val);						\
1212 		if (vcpu_has_nv((v)) && (r) >= __SANITISED_REG_START__)	\
1213 			__v = kvm_vcpu_apply_reg_masks((v), (r), __v);	\
1214 									\
1215 		ctxt_sys_reg(ctxt, (r)) = __v;				\
1216 	} while (0)
1217 
1218 #define __vcpu_sys_reg(v,r)						\
1219 	({								\
1220 		const struct kvm_cpu_context *ctxt = &(v)->arch.ctxt;	\
1221 		u64 __v = ctxt_sys_reg(ctxt, (r));			\
1222 		if (vcpu_has_nv((v)) && (r) >= __SANITISED_REG_START__)	\
1223 			__v = kvm_vcpu_apply_reg_masks((v), (r), __v);	\
1224 		__v;							\
1225 	})
1226 
1227 u64 vcpu_read_sys_reg(const struct kvm_vcpu *, enum vcpu_sysreg);
1228 void vcpu_write_sys_reg(struct kvm_vcpu *, u64, enum vcpu_sysreg);
1229 
1230 struct kvm_vm_stat {
1231 	struct kvm_vm_stat_generic generic;
1232 };
1233 
1234 struct kvm_vcpu_stat {
1235 	struct kvm_vcpu_stat_generic generic;
1236 	u64 hvc_exit_stat;
1237 	u64 wfe_exit_stat;
1238 	u64 wfi_exit_stat;
1239 	u64 mmio_exit_user;
1240 	u64 mmio_exit_kernel;
1241 	u64 signal_exits;
1242 	u64 exits;
1243 };
1244 
1245 unsigned long kvm_arm_num_regs(struct kvm_vcpu *vcpu);
1246 int kvm_arm_copy_reg_indices(struct kvm_vcpu *vcpu, u64 __user *indices);
1247 int kvm_arm_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
1248 int kvm_arm_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
1249 
1250 unsigned long kvm_arm_num_sys_reg_descs(struct kvm_vcpu *vcpu);
1251 int kvm_arm_copy_sys_reg_indices(struct kvm_vcpu *vcpu, u64 __user *uindices);
1252 
1253 int __kvm_arm_vcpu_get_events(struct kvm_vcpu *vcpu,
1254 			      struct kvm_vcpu_events *events);
1255 
1256 int __kvm_arm_vcpu_set_events(struct kvm_vcpu *vcpu,
1257 			      struct kvm_vcpu_events *events);
1258 
1259 void kvm_arm_halt_guest(struct kvm *kvm);
1260 void kvm_arm_resume_guest(struct kvm *kvm);
1261 
1262 #define vcpu_has_run_once(vcpu)	(!!READ_ONCE((vcpu)->pid))
1263 
1264 #ifndef __KVM_NVHE_HYPERVISOR__
1265 #define kvm_call_hyp_nvhe(f, ...)						\
1266 	({								\
1267 		struct arm_smccc_res res;				\
1268 									\
1269 		arm_smccc_1_1_hvc(KVM_HOST_SMCCC_FUNC(f),		\
1270 				  ##__VA_ARGS__, &res);			\
1271 		WARN_ON(res.a0 != SMCCC_RET_SUCCESS);			\
1272 									\
1273 		res.a1;							\
1274 	})
1275 
1276 /*
1277  * The isb() below is there to guarantee the same behaviour on VHE as on !VHE,
1278  * where the eret to EL1 acts as a context synchronization event.
1279  */
1280 #define kvm_call_hyp(f, ...)						\
1281 	do {								\
1282 		if (has_vhe()) {					\
1283 			f(__VA_ARGS__);					\
1284 			isb();						\
1285 		} else {						\
1286 			kvm_call_hyp_nvhe(f, ##__VA_ARGS__);		\
1287 		}							\
1288 	} while(0)
1289 
1290 #define kvm_call_hyp_ret(f, ...)					\
1291 	({								\
1292 		typeof(f(__VA_ARGS__)) ret;				\
1293 									\
1294 		if (has_vhe()) {					\
1295 			ret = f(__VA_ARGS__);				\
1296 		} else {						\
1297 			ret = kvm_call_hyp_nvhe(f, ##__VA_ARGS__);	\
1298 		}							\
1299 									\
1300 		ret;							\
1301 	})
1302 #else /* __KVM_NVHE_HYPERVISOR__ */
1303 #define kvm_call_hyp(f, ...) f(__VA_ARGS__)
1304 #define kvm_call_hyp_ret(f, ...) f(__VA_ARGS__)
1305 #define kvm_call_hyp_nvhe(f, ...) f(__VA_ARGS__)
1306 #endif /* __KVM_NVHE_HYPERVISOR__ */
1307 
1308 int handle_exit(struct kvm_vcpu *vcpu, int exception_index);
1309 void handle_exit_early(struct kvm_vcpu *vcpu, int exception_index);
1310 
1311 int kvm_handle_cp14_load_store(struct kvm_vcpu *vcpu);
1312 int kvm_handle_cp14_32(struct kvm_vcpu *vcpu);
1313 int kvm_handle_cp14_64(struct kvm_vcpu *vcpu);
1314 int kvm_handle_cp15_32(struct kvm_vcpu *vcpu);
1315 int kvm_handle_cp15_64(struct kvm_vcpu *vcpu);
1316 int kvm_handle_sys_reg(struct kvm_vcpu *vcpu);
1317 int kvm_handle_cp10_id(struct kvm_vcpu *vcpu);
1318 
1319 void kvm_sys_regs_create_debugfs(struct kvm *kvm);
1320 void kvm_reset_sys_regs(struct kvm_vcpu *vcpu);
1321 
1322 int __init kvm_sys_reg_table_init(void);
1323 struct sys_reg_desc;
1324 int __init populate_sysreg_config(const struct sys_reg_desc *sr,
1325 				  unsigned int idx);
1326 int __init populate_nv_trap_config(void);
1327 
1328 void kvm_calculate_traps(struct kvm_vcpu *vcpu);
1329 
1330 /* MMIO helpers */
1331 void kvm_mmio_write_buf(void *buf, unsigned int len, unsigned long data);
1332 unsigned long kvm_mmio_read_buf(const void *buf, unsigned int len);
1333 
1334 int kvm_handle_mmio_return(struct kvm_vcpu *vcpu);
1335 int io_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa);
1336 
1337 /*
1338  * Returns true if a Performance Monitoring Interrupt (PMI), a.k.a. perf event,
1339  * arrived in guest context.  For arm64, any event that arrives while a vCPU is
1340  * loaded is considered to be "in guest".
1341  */
1342 static inline bool kvm_arch_pmi_in_guest(struct kvm_vcpu *vcpu)
1343 {
1344 	return IS_ENABLED(CONFIG_GUEST_PERF_EVENTS) && !!vcpu;
1345 }
1346 
1347 long kvm_hypercall_pv_features(struct kvm_vcpu *vcpu);
1348 gpa_t kvm_init_stolen_time(struct kvm_vcpu *vcpu);
1349 void kvm_update_stolen_time(struct kvm_vcpu *vcpu);
1350 
1351 bool kvm_arm_pvtime_supported(void);
1352 int kvm_arm_pvtime_set_attr(struct kvm_vcpu *vcpu,
1353 			    struct kvm_device_attr *attr);
1354 int kvm_arm_pvtime_get_attr(struct kvm_vcpu *vcpu,
1355 			    struct kvm_device_attr *attr);
1356 int kvm_arm_pvtime_has_attr(struct kvm_vcpu *vcpu,
1357 			    struct kvm_device_attr *attr);
1358 
1359 extern unsigned int __ro_after_init kvm_arm_vmid_bits;
1360 int __init kvm_arm_vmid_alloc_init(void);
1361 void __init kvm_arm_vmid_alloc_free(void);
1362 void kvm_arm_vmid_update(struct kvm_vmid *kvm_vmid);
1363 void kvm_arm_vmid_clear_active(void);
1364 
1365 static inline void kvm_arm_pvtime_vcpu_init(struct kvm_vcpu_arch *vcpu_arch)
1366 {
1367 	vcpu_arch->steal.base = INVALID_GPA;
1368 }
1369 
1370 static inline bool kvm_arm_is_pvtime_enabled(struct kvm_vcpu_arch *vcpu_arch)
1371 {
1372 	return (vcpu_arch->steal.base != INVALID_GPA);
1373 }
1374 
1375 struct kvm_vcpu *kvm_mpidr_to_vcpu(struct kvm *kvm, unsigned long mpidr);
1376 
1377 DECLARE_KVM_HYP_PER_CPU(struct kvm_host_data, kvm_host_data);
1378 
1379 /*
1380  * How we access per-CPU host data depends on the where we access it from,
1381  * and the mode we're in:
1382  *
1383  * - VHE and nVHE hypervisor bits use their locally defined instance
1384  *
1385  * - the rest of the kernel use either the VHE or nVHE one, depending on
1386  *   the mode we're running in.
1387  *
1388  *   Unless we're in protected mode, fully deprivileged, and the nVHE
1389  *   per-CPU stuff is exclusively accessible to the protected EL2 code.
1390  *   In this case, the EL1 code uses the *VHE* data as its private state
1391  *   (which makes sense in a way as there shouldn't be any shared state
1392  *   between the host and the hypervisor).
1393  *
1394  * Yes, this is all totally trivial. Shoot me now.
1395  */
1396 #if defined(__KVM_NVHE_HYPERVISOR__) || defined(__KVM_VHE_HYPERVISOR__)
1397 #define host_data_ptr(f)	(&this_cpu_ptr(&kvm_host_data)->f)
1398 #else
1399 #define host_data_ptr(f)						\
1400 	(static_branch_unlikely(&kvm_protected_mode_initialized) ?	\
1401 	 &this_cpu_ptr(&kvm_host_data)->f :				\
1402 	 &this_cpu_ptr_hyp_sym(kvm_host_data)->f)
1403 #endif
1404 
1405 #define host_data_test_flag(flag)					\
1406 	(test_bit(KVM_HOST_DATA_FLAG_##flag, host_data_ptr(flags)))
1407 #define host_data_set_flag(flag)					\
1408 	set_bit(KVM_HOST_DATA_FLAG_##flag, host_data_ptr(flags))
1409 #define host_data_clear_flag(flag)					\
1410 	clear_bit(KVM_HOST_DATA_FLAG_##flag, host_data_ptr(flags))
1411 
1412 /* Check whether the FP regs are owned by the guest */
1413 static inline bool guest_owns_fp_regs(void)
1414 {
1415 	return *host_data_ptr(fp_owner) == FP_STATE_GUEST_OWNED;
1416 }
1417 
1418 /* Check whether the FP regs are owned by the host */
1419 static inline bool host_owns_fp_regs(void)
1420 {
1421 	return *host_data_ptr(fp_owner) == FP_STATE_HOST_OWNED;
1422 }
1423 
1424 static inline void kvm_init_host_cpu_context(struct kvm_cpu_context *cpu_ctxt)
1425 {
1426 	/* The host's MPIDR is immutable, so let's set it up at boot time */
1427 	ctxt_sys_reg(cpu_ctxt, MPIDR_EL1) = read_cpuid_mpidr();
1428 }
1429 
1430 static inline bool kvm_system_needs_idmapped_vectors(void)
1431 {
1432 	return cpus_have_final_cap(ARM64_SPECTRE_V3A);
1433 }
1434 
1435 void kvm_init_host_debug_data(void);
1436 void kvm_debug_init_vhe(void);
1437 void kvm_vcpu_load_debug(struct kvm_vcpu *vcpu);
1438 void kvm_vcpu_put_debug(struct kvm_vcpu *vcpu);
1439 void kvm_debug_set_guest_ownership(struct kvm_vcpu *vcpu);
1440 void kvm_debug_handle_oslar(struct kvm_vcpu *vcpu, u64 val);
1441 
1442 #define kvm_vcpu_os_lock_enabled(vcpu)		\
1443 	(!!(__vcpu_sys_reg(vcpu, OSLSR_EL1) & OSLSR_EL1_OSLK))
1444 
1445 #define kvm_debug_regs_in_use(vcpu)		\
1446 	((vcpu)->arch.debug_owner != VCPU_DEBUG_FREE)
1447 #define kvm_host_owns_debug_regs(vcpu)		\
1448 	((vcpu)->arch.debug_owner == VCPU_DEBUG_HOST_OWNED)
1449 #define kvm_guest_owns_debug_regs(vcpu)		\
1450 	((vcpu)->arch.debug_owner == VCPU_DEBUG_GUEST_OWNED)
1451 
1452 int kvm_arm_vcpu_arch_set_attr(struct kvm_vcpu *vcpu,
1453 			       struct kvm_device_attr *attr);
1454 int kvm_arm_vcpu_arch_get_attr(struct kvm_vcpu *vcpu,
1455 			       struct kvm_device_attr *attr);
1456 int kvm_arm_vcpu_arch_has_attr(struct kvm_vcpu *vcpu,
1457 			       struct kvm_device_attr *attr);
1458 
1459 int kvm_vm_ioctl_mte_copy_tags(struct kvm *kvm,
1460 			       struct kvm_arm_copy_mte_tags *copy_tags);
1461 int kvm_vm_ioctl_set_counter_offset(struct kvm *kvm,
1462 				    struct kvm_arm_counter_offset *offset);
1463 int kvm_vm_ioctl_get_reg_writable_masks(struct kvm *kvm,
1464 					struct reg_mask_range *range);
1465 
1466 /* Guest/host FPSIMD coordination helpers */
1467 void kvm_arch_vcpu_load_fp(struct kvm_vcpu *vcpu);
1468 void kvm_arch_vcpu_ctxflush_fp(struct kvm_vcpu *vcpu);
1469 void kvm_arch_vcpu_ctxsync_fp(struct kvm_vcpu *vcpu);
1470 void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu);
1471 
1472 static inline bool kvm_pmu_counter_deferred(struct perf_event_attr *attr)
1473 {
1474 	return (!has_vhe() && attr->exclude_host);
1475 }
1476 
1477 #ifdef CONFIG_KVM
1478 void kvm_set_pmu_events(u64 set, struct perf_event_attr *attr);
1479 void kvm_clr_pmu_events(u64 clr);
1480 bool kvm_set_pmuserenr(u64 val);
1481 void kvm_enable_trbe(void);
1482 void kvm_disable_trbe(void);
1483 void kvm_tracing_set_el1_configuration(u64 trfcr_while_in_guest);
1484 #else
1485 static inline void kvm_set_pmu_events(u64 set, struct perf_event_attr *attr) {}
1486 static inline void kvm_clr_pmu_events(u64 clr) {}
1487 static inline bool kvm_set_pmuserenr(u64 val)
1488 {
1489 	return false;
1490 }
1491 static inline void kvm_enable_trbe(void) {}
1492 static inline void kvm_disable_trbe(void) {}
1493 static inline void kvm_tracing_set_el1_configuration(u64 trfcr_while_in_guest) {}
1494 #endif
1495 
1496 void kvm_vcpu_load_vhe(struct kvm_vcpu *vcpu);
1497 void kvm_vcpu_put_vhe(struct kvm_vcpu *vcpu);
1498 
1499 int __init kvm_set_ipa_limit(void);
1500 u32 kvm_get_pa_bits(struct kvm *kvm);
1501 
1502 #define __KVM_HAVE_ARCH_VM_ALLOC
1503 struct kvm *kvm_arch_alloc_vm(void);
1504 
1505 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS
1506 
1507 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS_RANGE
1508 
1509 #define kvm_vm_is_protected(kvm)	(is_protected_kvm_enabled() && (kvm)->arch.pkvm.is_protected)
1510 
1511 #define vcpu_is_protected(vcpu)		kvm_vm_is_protected((vcpu)->kvm)
1512 
1513 int kvm_arm_vcpu_finalize(struct kvm_vcpu *vcpu, int feature);
1514 bool kvm_arm_vcpu_is_finalized(struct kvm_vcpu *vcpu);
1515 
1516 #define kvm_arm_vcpu_sve_finalized(vcpu) vcpu_get_flag(vcpu, VCPU_SVE_FINALIZED)
1517 
1518 #define kvm_has_mte(kvm)					\
1519 	(system_supports_mte() &&				\
1520 	 test_bit(KVM_ARCH_FLAG_MTE_ENABLED, &(kvm)->arch.flags))
1521 
1522 #define kvm_supports_32bit_el0()				\
1523 	(system_supports_32bit_el0() &&				\
1524 	 !static_branch_unlikely(&arm64_mismatched_32bit_el0))
1525 
1526 #define kvm_vm_has_ran_once(kvm)					\
1527 	(test_bit(KVM_ARCH_FLAG_HAS_RAN_ONCE, &(kvm)->arch.flags))
1528 
1529 static inline bool __vcpu_has_feature(const struct kvm_arch *ka, int feature)
1530 {
1531 	return test_bit(feature, ka->vcpu_features);
1532 }
1533 
1534 #define kvm_vcpu_has_feature(k, f)	__vcpu_has_feature(&(k)->arch, (f))
1535 #define vcpu_has_feature(v, f)	__vcpu_has_feature(&(v)->kvm->arch, (f))
1536 
1537 #define kvm_vcpu_initialized(v) vcpu_get_flag(vcpu, VCPU_INITIALIZED)
1538 
1539 int kvm_trng_call(struct kvm_vcpu *vcpu);
1540 #ifdef CONFIG_KVM
1541 extern phys_addr_t hyp_mem_base;
1542 extern phys_addr_t hyp_mem_size;
1543 void __init kvm_hyp_reserve(void);
1544 #else
1545 static inline void kvm_hyp_reserve(void) { }
1546 #endif
1547 
1548 void kvm_arm_vcpu_power_off(struct kvm_vcpu *vcpu);
1549 bool kvm_arm_vcpu_stopped(struct kvm_vcpu *vcpu);
1550 
1551 static inline u64 *__vm_id_reg(struct kvm_arch *ka, u32 reg)
1552 {
1553 	switch (reg) {
1554 	case sys_reg(3, 0, 0, 1, 0) ... sys_reg(3, 0, 0, 7, 7):
1555 		return &ka->id_regs[IDREG_IDX(reg)];
1556 	case SYS_CTR_EL0:
1557 		return &ka->ctr_el0;
1558 	case SYS_MIDR_EL1:
1559 		return &ka->midr_el1;
1560 	case SYS_REVIDR_EL1:
1561 		return &ka->revidr_el1;
1562 	case SYS_AIDR_EL1:
1563 		return &ka->aidr_el1;
1564 	default:
1565 		WARN_ON_ONCE(1);
1566 		return NULL;
1567 	}
1568 }
1569 
1570 #define kvm_read_vm_id_reg(kvm, reg)					\
1571 	({ u64 __val = *__vm_id_reg(&(kvm)->arch, reg); __val; })
1572 
1573 void kvm_set_vm_id_reg(struct kvm *kvm, u32 reg, u64 val);
1574 
1575 #define __expand_field_sign_unsigned(id, fld, val)			\
1576 	((u64)SYS_FIELD_VALUE(id, fld, val))
1577 
1578 #define __expand_field_sign_signed(id, fld, val)			\
1579 	({								\
1580 		u64 __val = SYS_FIELD_VALUE(id, fld, val);		\
1581 		sign_extend64(__val, id##_##fld##_WIDTH - 1);		\
1582 	})
1583 
1584 #define get_idreg_field_unsigned(kvm, id, fld)				\
1585 	({								\
1586 		u64 __val = kvm_read_vm_id_reg((kvm), SYS_##id);	\
1587 		FIELD_GET(id##_##fld##_MASK, __val);			\
1588 	})
1589 
1590 #define get_idreg_field_signed(kvm, id, fld)				\
1591 	({								\
1592 		u64 __val = get_idreg_field_unsigned(kvm, id, fld);	\
1593 		sign_extend64(__val, id##_##fld##_WIDTH - 1);		\
1594 	})
1595 
1596 #define get_idreg_field_enum(kvm, id, fld)				\
1597 	get_idreg_field_unsigned(kvm, id, fld)
1598 
1599 #define kvm_cmp_feat_signed(kvm, id, fld, op, limit)			\
1600 	(get_idreg_field_signed((kvm), id, fld) op __expand_field_sign_signed(id, fld, limit))
1601 
1602 #define kvm_cmp_feat_unsigned(kvm, id, fld, op, limit)			\
1603 	(get_idreg_field_unsigned((kvm), id, fld) op __expand_field_sign_unsigned(id, fld, limit))
1604 
1605 #define kvm_cmp_feat(kvm, id, fld, op, limit)				\
1606 	(id##_##fld##_SIGNED ?						\
1607 	 kvm_cmp_feat_signed(kvm, id, fld, op, limit) :			\
1608 	 kvm_cmp_feat_unsigned(kvm, id, fld, op, limit))
1609 
1610 #define __kvm_has_feat(kvm, id, fld, limit)				\
1611 	kvm_cmp_feat(kvm, id, fld, >=, limit)
1612 
1613 #define kvm_has_feat(kvm, ...) __kvm_has_feat(kvm, __VA_ARGS__)
1614 
1615 #define __kvm_has_feat_enum(kvm, id, fld, val)				\
1616 	kvm_cmp_feat_unsigned(kvm, id, fld, ==, val)
1617 
1618 #define kvm_has_feat_enum(kvm, ...) __kvm_has_feat_enum(kvm, __VA_ARGS__)
1619 
1620 #define kvm_has_feat_range(kvm, id, fld, min, max)			\
1621 	(kvm_cmp_feat(kvm, id, fld, >=, min) &&				\
1622 	kvm_cmp_feat(kvm, id, fld, <=, max))
1623 
1624 /* Check for a given level of PAuth support */
1625 #define kvm_has_pauth(k, l)						\
1626 	({								\
1627 		bool pa, pi, pa3;					\
1628 									\
1629 		pa  = kvm_has_feat((k), ID_AA64ISAR1_EL1, APA, l);	\
1630 		pa &= kvm_has_feat((k), ID_AA64ISAR1_EL1, GPA, IMP);	\
1631 		pi  = kvm_has_feat((k), ID_AA64ISAR1_EL1, API, l);	\
1632 		pi &= kvm_has_feat((k), ID_AA64ISAR1_EL1, GPI, IMP);	\
1633 		pa3  = kvm_has_feat((k), ID_AA64ISAR2_EL1, APA3, l);	\
1634 		pa3 &= kvm_has_feat((k), ID_AA64ISAR2_EL1, GPA3, IMP);	\
1635 									\
1636 		(pa + pi + pa3) == 1;					\
1637 	})
1638 
1639 #define kvm_has_fpmr(k)					\
1640 	(system_supports_fpmr() &&			\
1641 	 kvm_has_feat((k), ID_AA64PFR2_EL1, FPMR, IMP))
1642 
1643 #define kvm_has_tcr2(k)				\
1644 	(kvm_has_feat((k), ID_AA64MMFR3_EL1, TCRX, IMP))
1645 
1646 #define kvm_has_s1pie(k)				\
1647 	(kvm_has_feat((k), ID_AA64MMFR3_EL1, S1PIE, IMP))
1648 
1649 #define kvm_has_s1poe(k)				\
1650 	(system_supports_poe() &&			\
1651 	 kvm_has_feat((k), ID_AA64MMFR3_EL1, S1POE, IMP))
1652 
1653 #define kvm_has_ras(k)					\
1654 	(kvm_has_feat((k), ID_AA64PFR0_EL1, RAS, IMP))
1655 
1656 #define kvm_has_sctlr2(k)				\
1657 	(kvm_has_feat((k), ID_AA64MMFR3_EL1, SCTLRX, IMP))
1658 
1659 static inline bool kvm_arch_has_irq_bypass(void)
1660 {
1661 	return true;
1662 }
1663 
1664 void compute_fgu(struct kvm *kvm, enum fgt_group_id fgt);
1665 struct resx get_reg_fixed_bits(struct kvm *kvm, enum vcpu_sysreg reg);
1666 void check_feature_map(void);
1667 void kvm_vcpu_load_fgt(struct kvm_vcpu *vcpu);
1668 
1669 static __always_inline enum fgt_group_id __fgt_reg_to_group_id(enum vcpu_sysreg reg)
1670 {
1671 	switch (reg) {
1672 	case HFGRTR_EL2:
1673 	case HFGWTR_EL2:
1674 		return HFGRTR_GROUP;
1675 	case HFGITR_EL2:
1676 		return HFGITR_GROUP;
1677 	case HDFGRTR_EL2:
1678 	case HDFGWTR_EL2:
1679 		return HDFGRTR_GROUP;
1680 	case HAFGRTR_EL2:
1681 		return HAFGRTR_GROUP;
1682 	case HFGRTR2_EL2:
1683 	case HFGWTR2_EL2:
1684 		return HFGRTR2_GROUP;
1685 	case HFGITR2_EL2:
1686 		return HFGITR2_GROUP;
1687 	case HDFGRTR2_EL2:
1688 	case HDFGWTR2_EL2:
1689 		return HDFGRTR2_GROUP;
1690 	case ICH_HFGRTR_EL2:
1691 	case ICH_HFGWTR_EL2:
1692 		return ICH_HFGRTR_GROUP;
1693 	case ICH_HFGITR_EL2:
1694 		return ICH_HFGITR_GROUP;
1695 	default:
1696 		BUILD_BUG_ON(1);
1697 	}
1698 }
1699 
1700 #define vcpu_fgt(vcpu, reg)						\
1701 	({								\
1702 		enum fgt_group_id id = __fgt_reg_to_group_id(reg);	\
1703 		u64 *p;							\
1704 		switch (reg) {						\
1705 		case HFGWTR_EL2:					\
1706 		case HDFGWTR_EL2:					\
1707 		case HFGWTR2_EL2:					\
1708 		case HDFGWTR2_EL2:					\
1709 		case ICH_HFGWTR_EL2:					\
1710 			p = &(vcpu)->arch.fgt[id].w;			\
1711 			break;						\
1712 		default:						\
1713 			p = &(vcpu)->arch.fgt[id].r;			\
1714 			break;						\
1715 		}							\
1716 									\
1717 		p;							\
1718 	})
1719 
1720 long kvm_get_cap_for_kvm_ioctl(unsigned int ioctl, long *ext);
1721 
1722 #endif /* __ARM64_KVM_HOST_H__ */
1723