xref: /linux/arch/arm64/include/asm/kvm_host.h (revision 63eb28bb1402891b1ad2be02a530f29a9dd7f1cd)
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 
58 #define KVM_DIRTY_LOG_MANUAL_CAPS   (KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE | \
59 				     KVM_DIRTY_LOG_INITIALLY_SET)
60 
61 #define KVM_HAVE_MMU_RWLOCK
62 
63 /*
64  * Mode of operation configurable with kvm-arm.mode early param.
65  * See Documentation/admin-guide/kernel-parameters.txt for more information.
66  */
67 enum kvm_mode {
68 	KVM_MODE_DEFAULT,
69 	KVM_MODE_PROTECTED,
70 	KVM_MODE_NV,
71 	KVM_MODE_NONE,
72 };
73 #ifdef CONFIG_KVM
74 enum kvm_mode kvm_get_mode(void);
75 #else
kvm_get_mode(void)76 static inline enum kvm_mode kvm_get_mode(void) { return KVM_MODE_NONE; };
77 #endif
78 
79 extern unsigned int __ro_after_init kvm_sve_max_vl;
80 extern unsigned int __ro_after_init kvm_host_sve_max_vl;
81 int __init kvm_arm_init_sve(void);
82 
83 u32 __attribute_const__ kvm_target_cpu(void);
84 void kvm_reset_vcpu(struct kvm_vcpu *vcpu);
85 void kvm_arm_vcpu_destroy(struct kvm_vcpu *vcpu);
86 
87 struct kvm_hyp_memcache {
88 	phys_addr_t head;
89 	unsigned long nr_pages;
90 	struct pkvm_mapping *mapping; /* only used from EL1 */
91 
92 #define	HYP_MEMCACHE_ACCOUNT_STAGE2	BIT(1)
93 	unsigned long flags;
94 };
95 
push_hyp_memcache(struct kvm_hyp_memcache * mc,phys_addr_t * p,phys_addr_t (* to_pa)(void * virt))96 static inline void push_hyp_memcache(struct kvm_hyp_memcache *mc,
97 				     phys_addr_t *p,
98 				     phys_addr_t (*to_pa)(void *virt))
99 {
100 	*p = mc->head;
101 	mc->head = to_pa(p);
102 	mc->nr_pages++;
103 }
104 
pop_hyp_memcache(struct kvm_hyp_memcache * mc,void * (* to_va)(phys_addr_t phys))105 static inline void *pop_hyp_memcache(struct kvm_hyp_memcache *mc,
106 				     void *(*to_va)(phys_addr_t phys))
107 {
108 	phys_addr_t *p = to_va(mc->head & PAGE_MASK);
109 
110 	if (!mc->nr_pages)
111 		return NULL;
112 
113 	mc->head = *p;
114 	mc->nr_pages--;
115 
116 	return p;
117 }
118 
__topup_hyp_memcache(struct kvm_hyp_memcache * mc,unsigned long min_pages,void * (* alloc_fn)(void * arg),phys_addr_t (* to_pa)(void * virt),void * arg)119 static inline int __topup_hyp_memcache(struct kvm_hyp_memcache *mc,
120 				       unsigned long min_pages,
121 				       void *(*alloc_fn)(void *arg),
122 				       phys_addr_t (*to_pa)(void *virt),
123 				       void *arg)
124 {
125 	while (mc->nr_pages < min_pages) {
126 		phys_addr_t *p = alloc_fn(arg);
127 
128 		if (!p)
129 			return -ENOMEM;
130 		push_hyp_memcache(mc, p, to_pa);
131 	}
132 
133 	return 0;
134 }
135 
__free_hyp_memcache(struct kvm_hyp_memcache * mc,void (* free_fn)(void * virt,void * arg),void * (* to_va)(phys_addr_t phys),void * arg)136 static inline void __free_hyp_memcache(struct kvm_hyp_memcache *mc,
137 				       void (*free_fn)(void *virt, void *arg),
138 				       void *(*to_va)(phys_addr_t phys),
139 				       void *arg)
140 {
141 	while (mc->nr_pages)
142 		free_fn(pop_hyp_memcache(mc, to_va), arg);
143 }
144 
145 void free_hyp_memcache(struct kvm_hyp_memcache *mc);
146 int topup_hyp_memcache(struct kvm_hyp_memcache *mc, unsigned long min_pages);
147 
148 struct kvm_vmid {
149 	atomic64_t id;
150 };
151 
152 struct kvm_s2_mmu {
153 	struct kvm_vmid vmid;
154 
155 	/*
156 	 * stage2 entry level table
157 	 *
158 	 * Two kvm_s2_mmu structures in the same VM can point to the same
159 	 * pgd here.  This happens when running a guest using a
160 	 * translation regime that isn't affected by its own stage-2
161 	 * translation, such as a non-VHE hypervisor running at vEL2, or
162 	 * for vEL1/EL0 with vHCR_EL2.VM == 0.  In that case, we use the
163 	 * canonical stage-2 page tables.
164 	 */
165 	phys_addr_t	pgd_phys;
166 	struct kvm_pgtable *pgt;
167 
168 	/*
169 	 * VTCR value used on the host. For a non-NV guest (or a NV
170 	 * guest that runs in a context where its own S2 doesn't
171 	 * apply), its T0SZ value reflects that of the IPA size.
172 	 *
173 	 * For a shadow S2 MMU, T0SZ reflects the PARange exposed to
174 	 * the guest.
175 	 */
176 	u64	vtcr;
177 
178 	/* The last vcpu id that ran on each physical CPU */
179 	int __percpu *last_vcpu_ran;
180 
181 #define KVM_ARM_EAGER_SPLIT_CHUNK_SIZE_DEFAULT 0
182 	/*
183 	 * Memory cache used to split
184 	 * KVM_CAP_ARM_EAGER_SPLIT_CHUNK_SIZE worth of huge pages. It
185 	 * is used to allocate stage2 page tables while splitting huge
186 	 * pages. The choice of KVM_CAP_ARM_EAGER_SPLIT_CHUNK_SIZE
187 	 * influences both the capacity of the split page cache, and
188 	 * how often KVM reschedules. Be wary of raising CHUNK_SIZE
189 	 * too high.
190 	 *
191 	 * Protected by kvm->slots_lock.
192 	 */
193 	struct kvm_mmu_memory_cache split_page_cache;
194 	uint64_t split_page_chunk_size;
195 
196 	struct kvm_arch *arch;
197 
198 	/*
199 	 * For a shadow stage-2 MMU, the virtual vttbr used by the
200 	 * host to parse the guest S2.
201 	 * This either contains:
202 	 * - the virtual VTTBR programmed by the guest hypervisor with
203          *   CnP cleared
204 	 * - The value 1 (VMID=0, BADDR=0, CnP=1) if invalid
205 	 *
206 	 * We also cache the full VTCR which gets used for TLB invalidation,
207 	 * taking the ARM ARM's "Any of the bits in VTCR_EL2 are permitted
208 	 * to be cached in a TLB" to the letter.
209 	 */
210 	u64	tlb_vttbr;
211 	u64	tlb_vtcr;
212 
213 	/*
214 	 * true when this represents a nested context where virtual
215 	 * HCR_EL2.VM == 1
216 	 */
217 	bool	nested_stage2_enabled;
218 
219 	/*
220 	 * true when this MMU needs to be unmapped before being used for a new
221 	 * purpose.
222 	 */
223 	bool	pending_unmap;
224 
225 	/*
226 	 *  0: Nobody is currently using this, check vttbr for validity
227 	 * >0: Somebody is actively using this.
228 	 */
229 	atomic_t refcnt;
230 };
231 
232 struct kvm_arch_memory_slot {
233 };
234 
235 /**
236  * struct kvm_smccc_features: Descriptor of the hypercall services exposed to the guests
237  *
238  * @std_bmap: Bitmap of standard secure service calls
239  * @std_hyp_bmap: Bitmap of standard hypervisor service calls
240  * @vendor_hyp_bmap: Bitmap of vendor specific hypervisor service calls
241  */
242 struct kvm_smccc_features {
243 	unsigned long std_bmap;
244 	unsigned long std_hyp_bmap;
245 	unsigned long vendor_hyp_bmap; /* Function numbers 0-63 */
246 	unsigned long vendor_hyp_bmap_2; /* Function numbers 64-127 */
247 };
248 
249 typedef unsigned int pkvm_handle_t;
250 
251 struct kvm_protected_vm {
252 	pkvm_handle_t handle;
253 	struct kvm_hyp_memcache teardown_mc;
254 	struct kvm_hyp_memcache stage2_teardown_mc;
255 	bool enabled;
256 };
257 
258 struct kvm_mpidr_data {
259 	u64			mpidr_mask;
260 	DECLARE_FLEX_ARRAY(u16, cmpidr_to_idx);
261 };
262 
kvm_mpidr_index(struct kvm_mpidr_data * data,u64 mpidr)263 static inline u16 kvm_mpidr_index(struct kvm_mpidr_data *data, u64 mpidr)
264 {
265 	unsigned long index = 0, mask = data->mpidr_mask;
266 	unsigned long aff = mpidr & MPIDR_HWID_BITMASK;
267 
268 	bitmap_gather(&index, &aff, &mask, fls(mask));
269 
270 	return index;
271 }
272 
273 struct kvm_sysreg_masks;
274 
275 enum fgt_group_id {
276 	__NO_FGT_GROUP__,
277 	HFGRTR_GROUP,
278 	HFGWTR_GROUP = HFGRTR_GROUP,
279 	HDFGRTR_GROUP,
280 	HDFGWTR_GROUP = HDFGRTR_GROUP,
281 	HFGITR_GROUP,
282 	HAFGRTR_GROUP,
283 	HFGRTR2_GROUP,
284 	HFGWTR2_GROUP = HFGRTR2_GROUP,
285 	HDFGRTR2_GROUP,
286 	HDFGWTR2_GROUP = HDFGRTR2_GROUP,
287 	HFGITR2_GROUP,
288 
289 	/* Must be last */
290 	__NR_FGT_GROUP_IDS__
291 };
292 
293 struct kvm_arch {
294 	struct kvm_s2_mmu mmu;
295 
296 	/*
297 	 * Fine-Grained UNDEF, mimicking the FGT layout defined by the
298 	 * architecture. We track them globally, as we present the
299 	 * same feature-set to all vcpus.
300 	 *
301 	 * Index 0 is currently spare.
302 	 */
303 	u64 fgu[__NR_FGT_GROUP_IDS__];
304 
305 	/*
306 	 * Stage 2 paging state for VMs with nested S2 using a virtual
307 	 * VMID.
308 	 */
309 	struct kvm_s2_mmu *nested_mmus;
310 	size_t nested_mmus_size;
311 	int nested_mmus_next;
312 
313 	/* Interrupt controller */
314 	struct vgic_dist	vgic;
315 
316 	/* Timers */
317 	struct arch_timer_vm_data timer_data;
318 
319 	/* Mandated version of PSCI */
320 	u32 psci_version;
321 
322 	/* Protects VM-scoped configuration data */
323 	struct mutex config_lock;
324 
325 	/*
326 	 * If we encounter a data abort without valid instruction syndrome
327 	 * information, report this to user space.  User space can (and
328 	 * should) opt in to this feature if KVM_CAP_ARM_NISV_TO_USER is
329 	 * supported.
330 	 */
331 #define KVM_ARCH_FLAG_RETURN_NISV_IO_ABORT_TO_USER	0
332 	/* Memory Tagging Extension enabled for the guest */
333 #define KVM_ARCH_FLAG_MTE_ENABLED			1
334 	/* At least one vCPU has ran in the VM */
335 #define KVM_ARCH_FLAG_HAS_RAN_ONCE			2
336 	/* The vCPU feature set for the VM is configured */
337 #define KVM_ARCH_FLAG_VCPU_FEATURES_CONFIGURED		3
338 	/* PSCI SYSTEM_SUSPEND enabled for the guest */
339 #define KVM_ARCH_FLAG_SYSTEM_SUSPEND_ENABLED		4
340 	/* VM counter offset */
341 #define KVM_ARCH_FLAG_VM_COUNTER_OFFSET			5
342 	/* Timer PPIs made immutable */
343 #define KVM_ARCH_FLAG_TIMER_PPIS_IMMUTABLE		6
344 	/* Initial ID reg values loaded */
345 #define KVM_ARCH_FLAG_ID_REGS_INITIALIZED		7
346 	/* Fine-Grained UNDEF initialised */
347 #define KVM_ARCH_FLAG_FGU_INITIALIZED			8
348 	/* SVE exposed to guest */
349 #define KVM_ARCH_FLAG_GUEST_HAS_SVE			9
350 	/* MIDR_EL1, REVIDR_EL1, and AIDR_EL1 are writable from userspace */
351 #define KVM_ARCH_FLAG_WRITABLE_IMP_ID_REGS		10
352 	unsigned long flags;
353 
354 	/* VM-wide vCPU feature set */
355 	DECLARE_BITMAP(vcpu_features, KVM_VCPU_MAX_FEATURES);
356 
357 	/* MPIDR to vcpu index mapping, optional */
358 	struct kvm_mpidr_data *mpidr_data;
359 
360 	/*
361 	 * VM-wide PMU filter, implemented as a bitmap and big enough for
362 	 * up to 2^10 events (ARMv8.0) or 2^16 events (ARMv8.1+).
363 	 */
364 	unsigned long *pmu_filter;
365 	struct arm_pmu *arm_pmu;
366 
367 	cpumask_var_t supported_cpus;
368 
369 	/* Maximum number of counters for the guest */
370 	u8 nr_pmu_counters;
371 
372 	/* Iterator for idreg debugfs */
373 	u8	idreg_debugfs_iter;
374 
375 	/* Hypercall features firmware registers' descriptor */
376 	struct kvm_smccc_features smccc_feat;
377 	struct maple_tree smccc_filter;
378 
379 	/*
380 	 * Emulated CPU ID registers per VM
381 	 * (Op0, Op1, CRn, CRm, Op2) of the ID registers to be saved in it
382 	 * is (3, 0, 0, crm, op2), where 1<=crm<8, 0<=op2<8.
383 	 *
384 	 * These emulated idregs are VM-wide, but accessed from the context of a vCPU.
385 	 * Atomic access to multiple idregs are guarded by kvm_arch.config_lock.
386 	 */
387 #define IDREG_IDX(id)		(((sys_reg_CRm(id) - 1) << 3) | sys_reg_Op2(id))
388 #define KVM_ARM_ID_REG_NUM	(IDREG_IDX(sys_reg(3, 0, 0, 7, 7)) + 1)
389 	u64 id_regs[KVM_ARM_ID_REG_NUM];
390 
391 	u64 midr_el1;
392 	u64 revidr_el1;
393 	u64 aidr_el1;
394 	u64 ctr_el0;
395 
396 	/* Masks for VNCR-backed and general EL2 sysregs */
397 	struct kvm_sysreg_masks	*sysreg_masks;
398 
399 	/* Count the number of VNCR_EL2 currently mapped */
400 	atomic_t vncr_map_count;
401 
402 	/*
403 	 * For an untrusted host VM, 'pkvm.handle' is used to lookup
404 	 * the associated pKVM instance in the hypervisor.
405 	 */
406 	struct kvm_protected_vm pkvm;
407 };
408 
409 struct kvm_vcpu_fault_info {
410 	u64 esr_el2;		/* Hyp Syndrom Register */
411 	u64 far_el2;		/* Hyp Fault Address Register */
412 	u64 hpfar_el2;		/* Hyp IPA Fault Address Register */
413 	u64 disr_el1;		/* Deferred [SError] Status Register */
414 };
415 
416 /*
417  * VNCR() just places the VNCR_capable registers in the enum after
418  * __VNCR_START__, and the value (after correction) to be an 8-byte offset
419  * from the VNCR base. As we don't require the enum to be otherwise ordered,
420  * we need the terrible hack below to ensure that we correctly size the
421  * sys_regs array, no matter what.
422  *
423  * The __MAX__ macro has been lifted from Sean Eron Anderson's wonderful
424  * treasure trove of bit hacks:
425  * https://graphics.stanford.edu/~seander/bithacks.html#IntegerMinOrMax
426  */
427 #define __MAX__(x,y)	((x) ^ (((x) ^ (y)) & -((x) < (y))))
428 #define VNCR(r)						\
429 	__before_##r,					\
430 	r = __VNCR_START__ + ((VNCR_ ## r) / 8),	\
431 	__after_##r = __MAX__(__before_##r - 1, r)
432 
433 #define MARKER(m)				\
434 	m, __after_##m = m - 1
435 
436 enum vcpu_sysreg {
437 	__INVALID_SYSREG__,   /* 0 is reserved as an invalid value */
438 	MPIDR_EL1,	/* MultiProcessor Affinity Register */
439 	CLIDR_EL1,	/* Cache Level ID Register */
440 	CSSELR_EL1,	/* Cache Size Selection Register */
441 	TPIDR_EL0,	/* Thread ID, User R/W */
442 	TPIDRRO_EL0,	/* Thread ID, User R/O */
443 	TPIDR_EL1,	/* Thread ID, Privileged */
444 	CNTKCTL_EL1,	/* Timer Control Register (EL1) */
445 	PAR_EL1,	/* Physical Address Register */
446 	MDCCINT_EL1,	/* Monitor Debug Comms Channel Interrupt Enable Reg */
447 	OSLSR_EL1,	/* OS Lock Status Register */
448 	DISR_EL1,	/* Deferred Interrupt Status Register */
449 
450 	/* Performance Monitors Registers */
451 	PMCR_EL0,	/* Control Register */
452 	PMSELR_EL0,	/* Event Counter Selection Register */
453 	PMEVCNTR0_EL0,	/* Event Counter Register (0-30) */
454 	PMEVCNTR30_EL0 = PMEVCNTR0_EL0 + 30,
455 	PMCCNTR_EL0,	/* Cycle Counter Register */
456 	PMEVTYPER0_EL0,	/* Event Type Register (0-30) */
457 	PMEVTYPER30_EL0 = PMEVTYPER0_EL0 + 30,
458 	PMCCFILTR_EL0,	/* Cycle Count Filter Register */
459 	PMCNTENSET_EL0,	/* Count Enable Set Register */
460 	PMINTENSET_EL1,	/* Interrupt Enable Set Register */
461 	PMOVSSET_EL0,	/* Overflow Flag Status Set Register */
462 	PMUSERENR_EL0,	/* User Enable Register */
463 
464 	/* Pointer Authentication Registers in a strict increasing order. */
465 	APIAKEYLO_EL1,
466 	APIAKEYHI_EL1,
467 	APIBKEYLO_EL1,
468 	APIBKEYHI_EL1,
469 	APDAKEYLO_EL1,
470 	APDAKEYHI_EL1,
471 	APDBKEYLO_EL1,
472 	APDBKEYHI_EL1,
473 	APGAKEYLO_EL1,
474 	APGAKEYHI_EL1,
475 
476 	/* Memory Tagging Extension registers */
477 	RGSR_EL1,	/* Random Allocation Tag Seed Register */
478 	GCR_EL1,	/* Tag Control Register */
479 	TFSRE0_EL1,	/* Tag Fault Status Register (EL0) */
480 
481 	POR_EL0,	/* Permission Overlay Register 0 (EL0) */
482 
483 	/* FP/SIMD/SVE */
484 	SVCR,
485 	FPMR,
486 
487 	/* 32bit specific registers. */
488 	DACR32_EL2,	/* Domain Access Control Register */
489 	IFSR32_EL2,	/* Instruction Fault Status Register */
490 	FPEXC32_EL2,	/* Floating-Point Exception Control Register */
491 	DBGVCR32_EL2,	/* Debug Vector Catch Register */
492 
493 	/* EL2 registers */
494 	SCTLR_EL2,	/* System Control Register (EL2) */
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 	TCR2_EL2,	/* Extended Translation Control Register (EL2) */
526 	SCTLR2_EL2,	/* System Control Register 2 (EL2) */
527 	MDCR_EL2,	/* Monitor Debug Configuration Register (EL2) */
528 	CNTHCTL_EL2,	/* Counter-timer Hypervisor Control register */
529 
530 	/* Any VNCR-capable reg goes after this point */
531 	MARKER(__VNCR_START__),
532 
533 	VNCR(SCTLR_EL1),/* System Control Register */
534 	VNCR(ACTLR_EL1),/* Auxiliary Control Register */
535 	VNCR(CPACR_EL1),/* Coprocessor Access Control */
536 	VNCR(ZCR_EL1),	/* SVE Control */
537 	VNCR(TTBR0_EL1),/* Translation Table Base Register 0 */
538 	VNCR(TTBR1_EL1),/* Translation Table Base Register 1 */
539 	VNCR(TCR_EL1),	/* Translation Control Register */
540 	VNCR(TCR2_EL1),	/* Extended Translation Control Register */
541 	VNCR(SCTLR2_EL1), /* System Control Register 2 */
542 	VNCR(ESR_EL1),	/* Exception Syndrome Register */
543 	VNCR(AFSR0_EL1),/* Auxiliary Fault Status Register 0 */
544 	VNCR(AFSR1_EL1),/* Auxiliary Fault Status Register 1 */
545 	VNCR(FAR_EL1),	/* Fault Address Register */
546 	VNCR(MAIR_EL1),	/* Memory Attribute Indirection Register */
547 	VNCR(VBAR_EL1),	/* Vector Base Address Register */
548 	VNCR(CONTEXTIDR_EL1),	/* Context ID Register */
549 	VNCR(AMAIR_EL1),/* Aux Memory Attribute Indirection Register */
550 	VNCR(MDSCR_EL1),/* Monitor Debug System Control Register */
551 	VNCR(ELR_EL1),
552 	VNCR(SP_EL1),
553 	VNCR(SPSR_EL1),
554 	VNCR(TFSR_EL1),	/* Tag Fault Status Register (EL1) */
555 	VNCR(VPIDR_EL2),/* Virtualization Processor ID Register */
556 	VNCR(VMPIDR_EL2),/* Virtualization Multiprocessor ID Register */
557 	VNCR(HCR_EL2),	/* Hypervisor Configuration Register */
558 	VNCR(HSTR_EL2),	/* Hypervisor System Trap Register */
559 	VNCR(VTTBR_EL2),/* Virtualization Translation Table Base Register */
560 	VNCR(VTCR_EL2),	/* Virtualization Translation Control Register */
561 	VNCR(TPIDR_EL2),/* EL2 Software Thread ID Register */
562 	VNCR(HCRX_EL2),	/* Extended Hypervisor Configuration Register */
563 
564 	/* Permission Indirection Extension registers */
565 	VNCR(PIR_EL1),	 /* Permission Indirection Register 1 (EL1) */
566 	VNCR(PIRE0_EL1), /*  Permission Indirection Register 0 (EL1) */
567 
568 	VNCR(POR_EL1),	/* Permission Overlay Register 1 (EL1) */
569 
570 	/* FEAT_RAS registers */
571 	VNCR(VDISR_EL2),
572 	VNCR(VSESR_EL2),
573 
574 	VNCR(HFGRTR_EL2),
575 	VNCR(HFGWTR_EL2),
576 	VNCR(HFGITR_EL2),
577 	VNCR(HDFGRTR_EL2),
578 	VNCR(HDFGWTR_EL2),
579 	VNCR(HAFGRTR_EL2),
580 	VNCR(HFGRTR2_EL2),
581 	VNCR(HFGWTR2_EL2),
582 	VNCR(HFGITR2_EL2),
583 	VNCR(HDFGRTR2_EL2),
584 	VNCR(HDFGWTR2_EL2),
585 
586 	VNCR(VNCR_EL2),
587 
588 	VNCR(CNTVOFF_EL2),
589 	VNCR(CNTV_CVAL_EL0),
590 	VNCR(CNTV_CTL_EL0),
591 	VNCR(CNTP_CVAL_EL0),
592 	VNCR(CNTP_CTL_EL0),
593 
594 	VNCR(ICH_LR0_EL2),
595 	VNCR(ICH_LR1_EL2),
596 	VNCR(ICH_LR2_EL2),
597 	VNCR(ICH_LR3_EL2),
598 	VNCR(ICH_LR4_EL2),
599 	VNCR(ICH_LR5_EL2),
600 	VNCR(ICH_LR6_EL2),
601 	VNCR(ICH_LR7_EL2),
602 	VNCR(ICH_LR8_EL2),
603 	VNCR(ICH_LR9_EL2),
604 	VNCR(ICH_LR10_EL2),
605 	VNCR(ICH_LR11_EL2),
606 	VNCR(ICH_LR12_EL2),
607 	VNCR(ICH_LR13_EL2),
608 	VNCR(ICH_LR14_EL2),
609 	VNCR(ICH_LR15_EL2),
610 
611 	VNCR(ICH_AP0R0_EL2),
612 	VNCR(ICH_AP0R1_EL2),
613 	VNCR(ICH_AP0R2_EL2),
614 	VNCR(ICH_AP0R3_EL2),
615 	VNCR(ICH_AP1R0_EL2),
616 	VNCR(ICH_AP1R1_EL2),
617 	VNCR(ICH_AP1R2_EL2),
618 	VNCR(ICH_AP1R3_EL2),
619 	VNCR(ICH_HCR_EL2),
620 	VNCR(ICH_VMCR_EL2),
621 
622 	NR_SYS_REGS	/* Nothing after this line! */
623 };
624 
625 struct kvm_sysreg_masks {
626 	struct {
627 		u64	res0;
628 		u64	res1;
629 	} mask[NR_SYS_REGS - __SANITISED_REG_START__];
630 };
631 
632 struct fgt_masks {
633 	const char	*str;
634 	u64		mask;
635 	u64		nmask;
636 	u64		res0;
637 };
638 
639 extern struct fgt_masks hfgrtr_masks;
640 extern struct fgt_masks hfgwtr_masks;
641 extern struct fgt_masks hfgitr_masks;
642 extern struct fgt_masks hdfgrtr_masks;
643 extern struct fgt_masks hdfgwtr_masks;
644 extern struct fgt_masks hafgrtr_masks;
645 extern struct fgt_masks hfgrtr2_masks;
646 extern struct fgt_masks hfgwtr2_masks;
647 extern struct fgt_masks hfgitr2_masks;
648 extern struct fgt_masks hdfgrtr2_masks;
649 extern struct fgt_masks hdfgwtr2_masks;
650 
651 extern struct fgt_masks kvm_nvhe_sym(hfgrtr_masks);
652 extern struct fgt_masks kvm_nvhe_sym(hfgwtr_masks);
653 extern struct fgt_masks kvm_nvhe_sym(hfgitr_masks);
654 extern struct fgt_masks kvm_nvhe_sym(hdfgrtr_masks);
655 extern struct fgt_masks kvm_nvhe_sym(hdfgwtr_masks);
656 extern struct fgt_masks kvm_nvhe_sym(hafgrtr_masks);
657 extern struct fgt_masks kvm_nvhe_sym(hfgrtr2_masks);
658 extern struct fgt_masks kvm_nvhe_sym(hfgwtr2_masks);
659 extern struct fgt_masks kvm_nvhe_sym(hfgitr2_masks);
660 extern struct fgt_masks kvm_nvhe_sym(hdfgrtr2_masks);
661 extern struct fgt_masks kvm_nvhe_sym(hdfgwtr2_masks);
662 
663 struct kvm_cpu_context {
664 	struct user_pt_regs regs;	/* sp = sp_el0 */
665 
666 	u64	spsr_abt;
667 	u64	spsr_und;
668 	u64	spsr_irq;
669 	u64	spsr_fiq;
670 
671 	struct user_fpsimd_state fp_regs;
672 
673 	u64 sys_regs[NR_SYS_REGS];
674 
675 	struct kvm_vcpu *__hyp_running_vcpu;
676 
677 	/* This pointer has to be 4kB aligned. */
678 	u64 *vncr_array;
679 };
680 
681 struct cpu_sve_state {
682 	__u64 zcr_el1;
683 
684 	/*
685 	 * Ordering is important since __sve_save_state/__sve_restore_state
686 	 * relies on it.
687 	 */
688 	__u32 fpsr;
689 	__u32 fpcr;
690 
691 	/* Must be SVE_VQ_BYTES (128 bit) aligned. */
692 	__u8 sve_regs[];
693 };
694 
695 /*
696  * This structure is instantiated on a per-CPU basis, and contains
697  * data that is:
698  *
699  * - tied to a single physical CPU, and
700  * - either have a lifetime that does not extend past vcpu_put()
701  * - or is an invariant for the lifetime of the system
702  *
703  * Use host_data_ptr(field) as a way to access a pointer to such a
704  * field.
705  */
706 struct kvm_host_data {
707 #define KVM_HOST_DATA_FLAG_HAS_SPE			0
708 #define KVM_HOST_DATA_FLAG_HAS_TRBE			1
709 #define KVM_HOST_DATA_FLAG_TRBE_ENABLED			4
710 #define KVM_HOST_DATA_FLAG_EL1_TRACING_CONFIGURED	5
711 #define KVM_HOST_DATA_FLAG_VCPU_IN_HYP_CONTEXT		6
712 #define KVM_HOST_DATA_FLAG_L1_VNCR_MAPPED		7
713 #define KVM_HOST_DATA_FLAG_HAS_BRBE			8
714 	unsigned long flags;
715 
716 	struct kvm_cpu_context host_ctxt;
717 
718 	/*
719 	 * Hyp VA.
720 	 * sve_state is only used in pKVM and if system_supports_sve().
721 	 */
722 	struct cpu_sve_state *sve_state;
723 
724 	/* Used by pKVM only. */
725 	u64	fpmr;
726 
727 	/* Ownership of the FP regs */
728 	enum {
729 		FP_STATE_FREE,
730 		FP_STATE_HOST_OWNED,
731 		FP_STATE_GUEST_OWNED,
732 	} fp_owner;
733 
734 	/*
735 	 * host_debug_state contains the host registers which are
736 	 * saved and restored during world switches.
737 	 */
738 	struct {
739 		/* {Break,watch}point registers */
740 		struct kvm_guest_debug_arch regs;
741 		/* Statistical profiling extension */
742 		u64 pmscr_el1;
743 		/* Self-hosted trace */
744 		u64 trfcr_el1;
745 		/* Values of trap registers for the host before guest entry. */
746 		u64 mdcr_el2;
747 		u64 brbcr_el1;
748 	} host_debug_state;
749 
750 	/* Guest trace filter value */
751 	u64 trfcr_while_in_guest;
752 
753 	/* Number of programmable event counters (PMCR_EL0.N) for this CPU */
754 	unsigned int nr_event_counters;
755 
756 	/* Number of debug breakpoints/watchpoints for this CPU (minus 1) */
757 	unsigned int debug_brps;
758 	unsigned int debug_wrps;
759 };
760 
761 struct kvm_host_psci_config {
762 	/* PSCI version used by host. */
763 	u32 version;
764 	u32 smccc_version;
765 
766 	/* Function IDs used by host if version is v0.1. */
767 	struct psci_0_1_function_ids function_ids_0_1;
768 
769 	bool psci_0_1_cpu_suspend_implemented;
770 	bool psci_0_1_cpu_on_implemented;
771 	bool psci_0_1_cpu_off_implemented;
772 	bool psci_0_1_migrate_implemented;
773 };
774 
775 extern struct kvm_host_psci_config kvm_nvhe_sym(kvm_host_psci_config);
776 #define kvm_host_psci_config CHOOSE_NVHE_SYM(kvm_host_psci_config)
777 
778 extern s64 kvm_nvhe_sym(hyp_physvirt_offset);
779 #define hyp_physvirt_offset CHOOSE_NVHE_SYM(hyp_physvirt_offset)
780 
781 extern u64 kvm_nvhe_sym(hyp_cpu_logical_map)[NR_CPUS];
782 #define hyp_cpu_logical_map CHOOSE_NVHE_SYM(hyp_cpu_logical_map)
783 
784 struct vcpu_reset_state {
785 	unsigned long	pc;
786 	unsigned long	r0;
787 	bool		be;
788 	bool		reset;
789 };
790 
791 struct vncr_tlb;
792 
793 struct kvm_vcpu_arch {
794 	struct kvm_cpu_context ctxt;
795 
796 	/*
797 	 * Guest floating point state
798 	 *
799 	 * The architecture has two main floating point extensions,
800 	 * the original FPSIMD and SVE.  These have overlapping
801 	 * register views, with the FPSIMD V registers occupying the
802 	 * low 128 bits of the SVE Z registers.  When the core
803 	 * floating point code saves the register state of a task it
804 	 * records which view it saved in fp_type.
805 	 */
806 	void *sve_state;
807 	enum fp_type fp_type;
808 	unsigned int sve_max_vl;
809 
810 	/* Stage 2 paging state used by the hardware on next switch */
811 	struct kvm_s2_mmu *hw_mmu;
812 
813 	/* Values of trap registers for the guest. */
814 	u64 hcr_el2;
815 	u64 hcrx_el2;
816 	u64 mdcr_el2;
817 
818 	/* Exception Information */
819 	struct kvm_vcpu_fault_info fault;
820 
821 	/* Configuration flags, set once and for all before the vcpu can run */
822 	u8 cflags;
823 
824 	/* Input flags to the hypervisor code, potentially cleared after use */
825 	u8 iflags;
826 
827 	/* State flags for kernel bookkeeping, unused by the hypervisor code */
828 	u16 sflags;
829 
830 	/*
831 	 * Don't run the guest (internal implementation need).
832 	 *
833 	 * Contrary to the flags above, this is set/cleared outside of
834 	 * a vcpu context, and thus cannot be mixed with the flags
835 	 * themselves (or the flag accesses need to be made atomic).
836 	 */
837 	bool pause;
838 
839 	/*
840 	 * We maintain more than a single set of debug registers to support
841 	 * debugging the guest from the host and to maintain separate host and
842 	 * guest state during world switches. vcpu_debug_state are the debug
843 	 * registers of the vcpu as the guest sees them.
844 	 *
845 	 * external_debug_state contains the debug values we want to debug the
846 	 * guest. This is set via the KVM_SET_GUEST_DEBUG ioctl.
847 	 */
848 	struct kvm_guest_debug_arch vcpu_debug_state;
849 	struct kvm_guest_debug_arch external_debug_state;
850 	u64 external_mdscr_el1;
851 
852 	enum {
853 		VCPU_DEBUG_FREE,
854 		VCPU_DEBUG_HOST_OWNED,
855 		VCPU_DEBUG_GUEST_OWNED,
856 	} debug_owner;
857 
858 	/* VGIC state */
859 	struct vgic_cpu vgic_cpu;
860 	struct arch_timer_cpu timer_cpu;
861 	struct kvm_pmu pmu;
862 
863 	/* vcpu power state */
864 	struct kvm_mp_state mp_state;
865 	spinlock_t mp_state_lock;
866 
867 	/* Cache some mmu pages needed inside spinlock regions */
868 	struct kvm_mmu_memory_cache mmu_page_cache;
869 
870 	/* Pages to top-up the pKVM/EL2 guest pool */
871 	struct kvm_hyp_memcache pkvm_memcache;
872 
873 	/* Virtual SError ESR to restore when HCR_EL2.VSE is set */
874 	u64 vsesr_el2;
875 
876 	/* Additional reset state */
877 	struct vcpu_reset_state	reset_state;
878 
879 	/* Guest PV state */
880 	struct {
881 		u64 last_steal;
882 		gpa_t base;
883 	} steal;
884 
885 	/* Per-vcpu CCSIDR override or NULL */
886 	u32 *ccsidr;
887 
888 	/* Per-vcpu TLB for VNCR_EL2 -- NULL when !NV */
889 	struct vncr_tlb	*vncr_tlb;
890 };
891 
892 /*
893  * Each 'flag' is composed of a comma-separated triplet:
894  *
895  * - the flag-set it belongs to in the vcpu->arch structure
896  * - the value for that flag
897  * - the mask for that flag
898  *
899  *  __vcpu_single_flag() builds such a triplet for a single-bit flag.
900  * unpack_vcpu_flag() extract the flag value from the triplet for
901  * direct use outside of the flag accessors.
902  */
903 #define __vcpu_single_flag(_set, _f)	_set, (_f), (_f)
904 
905 #define __unpack_flag(_set, _f, _m)	_f
906 #define unpack_vcpu_flag(...)		__unpack_flag(__VA_ARGS__)
907 
908 #define __build_check_flag(v, flagset, f, m)			\
909 	do {							\
910 		typeof(v->arch.flagset) *_fset;			\
911 								\
912 		/* Check that the flags fit in the mask */	\
913 		BUILD_BUG_ON(HWEIGHT(m) != HWEIGHT((f) | (m)));	\
914 		/* Check that the flags fit in the type */	\
915 		BUILD_BUG_ON((sizeof(*_fset) * 8) <= __fls(m));	\
916 	} while (0)
917 
918 #define __vcpu_get_flag(v, flagset, f, m)			\
919 	({							\
920 		__build_check_flag(v, flagset, f, m);		\
921 								\
922 		READ_ONCE(v->arch.flagset) & (m);		\
923 	})
924 
925 /*
926  * Note that the set/clear accessors must be preempt-safe in order to
927  * avoid nesting them with load/put which also manipulate flags...
928  */
929 #ifdef __KVM_NVHE_HYPERVISOR__
930 /* the nVHE hypervisor is always non-preemptible */
931 #define __vcpu_flags_preempt_disable()
932 #define __vcpu_flags_preempt_enable()
933 #else
934 #define __vcpu_flags_preempt_disable()	preempt_disable()
935 #define __vcpu_flags_preempt_enable()	preempt_enable()
936 #endif
937 
938 #define __vcpu_set_flag(v, flagset, f, m)			\
939 	do {							\
940 		typeof(v->arch.flagset) *fset;			\
941 								\
942 		__build_check_flag(v, flagset, f, m);		\
943 								\
944 		fset = &v->arch.flagset;			\
945 		__vcpu_flags_preempt_disable();			\
946 		if (HWEIGHT(m) > 1)				\
947 			*fset &= ~(m);				\
948 		*fset |= (f);					\
949 		__vcpu_flags_preempt_enable();			\
950 	} while (0)
951 
952 #define __vcpu_clear_flag(v, flagset, f, m)			\
953 	do {							\
954 		typeof(v->arch.flagset) *fset;			\
955 								\
956 		__build_check_flag(v, flagset, f, m);		\
957 								\
958 		fset = &v->arch.flagset;			\
959 		__vcpu_flags_preempt_disable();			\
960 		*fset &= ~(m);					\
961 		__vcpu_flags_preempt_enable();			\
962 	} while (0)
963 
964 #define __vcpu_test_and_clear_flag(v, flagset, f, m)		\
965 	({							\
966 		typeof(v->arch.flagset) set;			\
967 								\
968 		set = __vcpu_get_flag(v, flagset, f, m);	\
969 		__vcpu_clear_flag(v, flagset, f, m);		\
970 								\
971 		set;						\
972 	})
973 
974 #define vcpu_get_flag(v, ...)	__vcpu_get_flag((v), __VA_ARGS__)
975 #define vcpu_set_flag(v, ...)	__vcpu_set_flag((v), __VA_ARGS__)
976 #define vcpu_clear_flag(v, ...)	__vcpu_clear_flag((v), __VA_ARGS__)
977 #define vcpu_test_and_clear_flag(v, ...)			\
978 	__vcpu_test_and_clear_flag((v), __VA_ARGS__)
979 
980 /* KVM_ARM_VCPU_INIT completed */
981 #define VCPU_INITIALIZED	__vcpu_single_flag(cflags, BIT(0))
982 /* SVE config completed */
983 #define VCPU_SVE_FINALIZED	__vcpu_single_flag(cflags, BIT(1))
984 /* pKVM VCPU setup completed */
985 #define VCPU_PKVM_FINALIZED	__vcpu_single_flag(cflags, BIT(2))
986 
987 /* Exception pending */
988 #define PENDING_EXCEPTION	__vcpu_single_flag(iflags, BIT(0))
989 /*
990  * PC increment. Overlaps with EXCEPT_MASK on purpose so that it can't
991  * be set together with an exception...
992  */
993 #define INCREMENT_PC		__vcpu_single_flag(iflags, BIT(1))
994 /* Target EL/MODE (not a single flag, but let's abuse the macro) */
995 #define EXCEPT_MASK		__vcpu_single_flag(iflags, GENMASK(3, 1))
996 
997 /* Helpers to encode exceptions with minimum fuss */
998 #define __EXCEPT_MASK_VAL	unpack_vcpu_flag(EXCEPT_MASK)
999 #define __EXCEPT_SHIFT		__builtin_ctzl(__EXCEPT_MASK_VAL)
1000 #define __vcpu_except_flags(_f)	iflags, (_f << __EXCEPT_SHIFT), __EXCEPT_MASK_VAL
1001 
1002 /*
1003  * When PENDING_EXCEPTION is set, EXCEPT_MASK can take the following
1004  * values:
1005  *
1006  * For AArch32 EL1:
1007  */
1008 #define EXCEPT_AA32_UND		__vcpu_except_flags(0)
1009 #define EXCEPT_AA32_IABT	__vcpu_except_flags(1)
1010 #define EXCEPT_AA32_DABT	__vcpu_except_flags(2)
1011 /* For AArch64: */
1012 #define EXCEPT_AA64_EL1_SYNC	__vcpu_except_flags(0)
1013 #define EXCEPT_AA64_EL1_IRQ	__vcpu_except_flags(1)
1014 #define EXCEPT_AA64_EL1_FIQ	__vcpu_except_flags(2)
1015 #define EXCEPT_AA64_EL1_SERR	__vcpu_except_flags(3)
1016 /* For AArch64 with NV: */
1017 #define EXCEPT_AA64_EL2_SYNC	__vcpu_except_flags(4)
1018 #define EXCEPT_AA64_EL2_IRQ	__vcpu_except_flags(5)
1019 #define EXCEPT_AA64_EL2_FIQ	__vcpu_except_flags(6)
1020 #define EXCEPT_AA64_EL2_SERR	__vcpu_except_flags(7)
1021 
1022 /* Physical CPU not in supported_cpus */
1023 #define ON_UNSUPPORTED_CPU	__vcpu_single_flag(sflags, BIT(0))
1024 /* WFIT instruction trapped */
1025 #define IN_WFIT			__vcpu_single_flag(sflags, BIT(1))
1026 /* vcpu system registers loaded on physical CPU */
1027 #define SYSREGS_ON_CPU		__vcpu_single_flag(sflags, BIT(2))
1028 /* Software step state is Active-pending for external debug */
1029 #define HOST_SS_ACTIVE_PENDING	__vcpu_single_flag(sflags, BIT(3))
1030 /* Software step state is Active pending for guest debug */
1031 #define GUEST_SS_ACTIVE_PENDING __vcpu_single_flag(sflags, BIT(4))
1032 /* PMUSERENR for the guest EL0 is on physical CPU */
1033 #define PMUSERENR_ON_CPU	__vcpu_single_flag(sflags, BIT(5))
1034 /* WFI instruction trapped */
1035 #define IN_WFI			__vcpu_single_flag(sflags, BIT(6))
1036 /* KVM is currently emulating a nested ERET */
1037 #define IN_NESTED_ERET		__vcpu_single_flag(sflags, BIT(7))
1038 /* SError pending for nested guest */
1039 #define NESTED_SERROR_PENDING	__vcpu_single_flag(sflags, BIT(8))
1040 
1041 
1042 /* Pointer to the vcpu's SVE FFR for sve_{save,load}_state() */
1043 #define vcpu_sve_pffr(vcpu) (kern_hyp_va((vcpu)->arch.sve_state) +	\
1044 			     sve_ffr_offset((vcpu)->arch.sve_max_vl))
1045 
1046 #define vcpu_sve_max_vq(vcpu)	sve_vq_from_vl((vcpu)->arch.sve_max_vl)
1047 
1048 #define vcpu_sve_zcr_elx(vcpu)						\
1049 	(unlikely(is_hyp_ctxt(vcpu)) ? ZCR_EL2 : ZCR_EL1)
1050 
1051 #define sve_state_size_from_vl(sve_max_vl) ({				\
1052 	size_t __size_ret;						\
1053 	unsigned int __vq;						\
1054 									\
1055 	if (WARN_ON(!sve_vl_valid(sve_max_vl))) {			\
1056 		__size_ret = 0;						\
1057 	} else {							\
1058 		__vq = sve_vq_from_vl(sve_max_vl);			\
1059 		__size_ret = SVE_SIG_REGS_SIZE(__vq);			\
1060 	}								\
1061 									\
1062 	__size_ret;							\
1063 })
1064 
1065 #define vcpu_sve_state_size(vcpu) sve_state_size_from_vl((vcpu)->arch.sve_max_vl)
1066 
1067 #define KVM_GUESTDBG_VALID_MASK (KVM_GUESTDBG_ENABLE | \
1068 				 KVM_GUESTDBG_USE_SW_BP | \
1069 				 KVM_GUESTDBG_USE_HW | \
1070 				 KVM_GUESTDBG_SINGLESTEP)
1071 
1072 #define kvm_has_sve(kvm)	(system_supports_sve() &&		\
1073 				 test_bit(KVM_ARCH_FLAG_GUEST_HAS_SVE, &(kvm)->arch.flags))
1074 
1075 #ifdef __KVM_NVHE_HYPERVISOR__
1076 #define vcpu_has_sve(vcpu)	kvm_has_sve(kern_hyp_va((vcpu)->kvm))
1077 #else
1078 #define vcpu_has_sve(vcpu)	kvm_has_sve((vcpu)->kvm)
1079 #endif
1080 
1081 #ifdef CONFIG_ARM64_PTR_AUTH
1082 #define vcpu_has_ptrauth(vcpu)						\
1083 	((cpus_have_final_cap(ARM64_HAS_ADDRESS_AUTH) ||		\
1084 	  cpus_have_final_cap(ARM64_HAS_GENERIC_AUTH)) &&		\
1085 	 (vcpu_has_feature(vcpu, KVM_ARM_VCPU_PTRAUTH_ADDRESS) ||       \
1086 	  vcpu_has_feature(vcpu, KVM_ARM_VCPU_PTRAUTH_GENERIC)))
1087 #else
1088 #define vcpu_has_ptrauth(vcpu)		false
1089 #endif
1090 
1091 #define vcpu_on_unsupported_cpu(vcpu)					\
1092 	vcpu_get_flag(vcpu, ON_UNSUPPORTED_CPU)
1093 
1094 #define vcpu_set_on_unsupported_cpu(vcpu)				\
1095 	vcpu_set_flag(vcpu, ON_UNSUPPORTED_CPU)
1096 
1097 #define vcpu_clear_on_unsupported_cpu(vcpu)				\
1098 	vcpu_clear_flag(vcpu, ON_UNSUPPORTED_CPU)
1099 
1100 #define vcpu_gp_regs(v)		(&(v)->arch.ctxt.regs)
1101 
1102 /*
1103  * Only use __vcpu_sys_reg/ctxt_sys_reg if you know you want the
1104  * memory backed version of a register, and not the one most recently
1105  * accessed by a running VCPU.  For example, for userspace access or
1106  * for system registers that are never context switched, but only
1107  * emulated.
1108  *
1109  * Don't bother with VNCR-based accesses in the nVHE code, it has no
1110  * business dealing with NV.
1111  */
___ctxt_sys_reg(const struct kvm_cpu_context * ctxt,int r)1112 static inline u64 *___ctxt_sys_reg(const struct kvm_cpu_context *ctxt, int r)
1113 {
1114 #if !defined (__KVM_NVHE_HYPERVISOR__)
1115 	if (unlikely(cpus_have_final_cap(ARM64_HAS_NESTED_VIRT) &&
1116 		     r >= __VNCR_START__ && ctxt->vncr_array))
1117 		return &ctxt->vncr_array[r - __VNCR_START__];
1118 #endif
1119 	return (u64 *)&ctxt->sys_regs[r];
1120 }
1121 
1122 #define __ctxt_sys_reg(c,r)						\
1123 	({								\
1124 		BUILD_BUG_ON(__builtin_constant_p(r) &&			\
1125 			     (r) >= NR_SYS_REGS);			\
1126 		___ctxt_sys_reg(c, r);					\
1127 	})
1128 
1129 #define ctxt_sys_reg(c,r)	(*__ctxt_sys_reg(c,r))
1130 
1131 u64 kvm_vcpu_apply_reg_masks(const struct kvm_vcpu *, enum vcpu_sysreg, u64);
1132 
1133 #define __vcpu_assign_sys_reg(v, r, val)				\
1134 	do {								\
1135 		const struct kvm_cpu_context *ctxt = &(v)->arch.ctxt;	\
1136 		u64 __v = (val);					\
1137 		if (vcpu_has_nv((v)) && (r) >= __SANITISED_REG_START__)	\
1138 			__v = kvm_vcpu_apply_reg_masks((v), (r), __v);	\
1139 									\
1140 		ctxt_sys_reg(ctxt, (r)) = __v;				\
1141 	} while (0)
1142 
1143 #define __vcpu_rmw_sys_reg(v, r, op, val)				\
1144 	do {								\
1145 		const struct kvm_cpu_context *ctxt = &(v)->arch.ctxt;	\
1146 		u64 __v = ctxt_sys_reg(ctxt, (r));			\
1147 		__v op (val);						\
1148 		if (vcpu_has_nv((v)) && (r) >= __SANITISED_REG_START__)	\
1149 			__v = kvm_vcpu_apply_reg_masks((v), (r), __v);	\
1150 									\
1151 		ctxt_sys_reg(ctxt, (r)) = __v;				\
1152 	} while (0)
1153 
1154 #define __vcpu_sys_reg(v,r)						\
1155 	({								\
1156 		const struct kvm_cpu_context *ctxt = &(v)->arch.ctxt;	\
1157 		u64 __v = ctxt_sys_reg(ctxt, (r));			\
1158 		if (vcpu_has_nv((v)) && (r) >= __SANITISED_REG_START__)	\
1159 			__v = kvm_vcpu_apply_reg_masks((v), (r), __v);	\
1160 		__v;							\
1161 	})
1162 
1163 u64 vcpu_read_sys_reg(const struct kvm_vcpu *vcpu, int reg);
1164 void vcpu_write_sys_reg(struct kvm_vcpu *vcpu, u64 val, int reg);
1165 
__vcpu_read_sys_reg_from_cpu(int reg,u64 * val)1166 static inline bool __vcpu_read_sys_reg_from_cpu(int reg, u64 *val)
1167 {
1168 	/*
1169 	 * *** VHE ONLY ***
1170 	 *
1171 	 * System registers listed in the switch are not saved on every
1172 	 * exit from the guest but are only saved on vcpu_put.
1173 	 *
1174 	 * SYSREGS_ON_CPU *MUST* be checked before using this helper.
1175 	 *
1176 	 * Note that MPIDR_EL1 for the guest is set by KVM via VMPIDR_EL2 but
1177 	 * should never be listed below, because the guest cannot modify its
1178 	 * own MPIDR_EL1 and MPIDR_EL1 is accessed for VCPU A from VCPU B's
1179 	 * thread when emulating cross-VCPU communication.
1180 	 */
1181 	if (!has_vhe())
1182 		return false;
1183 
1184 	switch (reg) {
1185 	case SCTLR_EL1:		*val = read_sysreg_s(SYS_SCTLR_EL12);	break;
1186 	case CPACR_EL1:		*val = read_sysreg_s(SYS_CPACR_EL12);	break;
1187 	case TTBR0_EL1:		*val = read_sysreg_s(SYS_TTBR0_EL12);	break;
1188 	case TTBR1_EL1:		*val = read_sysreg_s(SYS_TTBR1_EL12);	break;
1189 	case TCR_EL1:		*val = read_sysreg_s(SYS_TCR_EL12);	break;
1190 	case TCR2_EL1:		*val = read_sysreg_s(SYS_TCR2_EL12);	break;
1191 	case PIR_EL1:		*val = read_sysreg_s(SYS_PIR_EL12);	break;
1192 	case PIRE0_EL1:		*val = read_sysreg_s(SYS_PIRE0_EL12);	break;
1193 	case POR_EL1:		*val = read_sysreg_s(SYS_POR_EL12);	break;
1194 	case ESR_EL1:		*val = read_sysreg_s(SYS_ESR_EL12);	break;
1195 	case AFSR0_EL1:		*val = read_sysreg_s(SYS_AFSR0_EL12);	break;
1196 	case AFSR1_EL1:		*val = read_sysreg_s(SYS_AFSR1_EL12);	break;
1197 	case FAR_EL1:		*val = read_sysreg_s(SYS_FAR_EL12);	break;
1198 	case MAIR_EL1:		*val = read_sysreg_s(SYS_MAIR_EL12);	break;
1199 	case VBAR_EL1:		*val = read_sysreg_s(SYS_VBAR_EL12);	break;
1200 	case CONTEXTIDR_EL1:	*val = read_sysreg_s(SYS_CONTEXTIDR_EL12);break;
1201 	case TPIDR_EL0:		*val = read_sysreg_s(SYS_TPIDR_EL0);	break;
1202 	case TPIDRRO_EL0:	*val = read_sysreg_s(SYS_TPIDRRO_EL0);	break;
1203 	case TPIDR_EL1:		*val = read_sysreg_s(SYS_TPIDR_EL1);	break;
1204 	case AMAIR_EL1:		*val = read_sysreg_s(SYS_AMAIR_EL12);	break;
1205 	case CNTKCTL_EL1:	*val = read_sysreg_s(SYS_CNTKCTL_EL12);	break;
1206 	case ELR_EL1:		*val = read_sysreg_s(SYS_ELR_EL12);	break;
1207 	case SPSR_EL1:		*val = read_sysreg_s(SYS_SPSR_EL12);	break;
1208 	case PAR_EL1:		*val = read_sysreg_par();		break;
1209 	case DACR32_EL2:	*val = read_sysreg_s(SYS_DACR32_EL2);	break;
1210 	case IFSR32_EL2:	*val = read_sysreg_s(SYS_IFSR32_EL2);	break;
1211 	case DBGVCR32_EL2:	*val = read_sysreg_s(SYS_DBGVCR32_EL2);	break;
1212 	case ZCR_EL1:		*val = read_sysreg_s(SYS_ZCR_EL12);	break;
1213 	case SCTLR2_EL1:	*val = read_sysreg_s(SYS_SCTLR2_EL12);	break;
1214 	default:		return false;
1215 	}
1216 
1217 	return true;
1218 }
1219 
__vcpu_write_sys_reg_to_cpu(u64 val,int reg)1220 static inline bool __vcpu_write_sys_reg_to_cpu(u64 val, int reg)
1221 {
1222 	/*
1223 	 * *** VHE ONLY ***
1224 	 *
1225 	 * System registers listed in the switch are not restored on every
1226 	 * entry to the guest but are only restored on vcpu_load.
1227 	 *
1228 	 * SYSREGS_ON_CPU *MUST* be checked before using this helper.
1229 	 *
1230 	 * Note that MPIDR_EL1 for the guest is set by KVM via VMPIDR_EL2 but
1231 	 * should never be listed below, because the MPIDR should only be set
1232 	 * once, before running the VCPU, and never changed later.
1233 	 */
1234 	if (!has_vhe())
1235 		return false;
1236 
1237 	switch (reg) {
1238 	case SCTLR_EL1:		write_sysreg_s(val, SYS_SCTLR_EL12);	break;
1239 	case CPACR_EL1:		write_sysreg_s(val, SYS_CPACR_EL12);	break;
1240 	case TTBR0_EL1:		write_sysreg_s(val, SYS_TTBR0_EL12);	break;
1241 	case TTBR1_EL1:		write_sysreg_s(val, SYS_TTBR1_EL12);	break;
1242 	case TCR_EL1:		write_sysreg_s(val, SYS_TCR_EL12);	break;
1243 	case TCR2_EL1:		write_sysreg_s(val, SYS_TCR2_EL12);	break;
1244 	case PIR_EL1:		write_sysreg_s(val, SYS_PIR_EL12);	break;
1245 	case PIRE0_EL1:		write_sysreg_s(val, SYS_PIRE0_EL12);	break;
1246 	case POR_EL1:		write_sysreg_s(val, SYS_POR_EL12);	break;
1247 	case ESR_EL1:		write_sysreg_s(val, SYS_ESR_EL12);	break;
1248 	case AFSR0_EL1:		write_sysreg_s(val, SYS_AFSR0_EL12);	break;
1249 	case AFSR1_EL1:		write_sysreg_s(val, SYS_AFSR1_EL12);	break;
1250 	case FAR_EL1:		write_sysreg_s(val, SYS_FAR_EL12);	break;
1251 	case MAIR_EL1:		write_sysreg_s(val, SYS_MAIR_EL12);	break;
1252 	case VBAR_EL1:		write_sysreg_s(val, SYS_VBAR_EL12);	break;
1253 	case CONTEXTIDR_EL1:	write_sysreg_s(val, SYS_CONTEXTIDR_EL12);break;
1254 	case TPIDR_EL0:		write_sysreg_s(val, SYS_TPIDR_EL0);	break;
1255 	case TPIDRRO_EL0:	write_sysreg_s(val, SYS_TPIDRRO_EL0);	break;
1256 	case TPIDR_EL1:		write_sysreg_s(val, SYS_TPIDR_EL1);	break;
1257 	case AMAIR_EL1:		write_sysreg_s(val, SYS_AMAIR_EL12);	break;
1258 	case CNTKCTL_EL1:	write_sysreg_s(val, SYS_CNTKCTL_EL12);	break;
1259 	case ELR_EL1:		write_sysreg_s(val, SYS_ELR_EL12);	break;
1260 	case SPSR_EL1:		write_sysreg_s(val, SYS_SPSR_EL12);	break;
1261 	case PAR_EL1:		write_sysreg_s(val, SYS_PAR_EL1);	break;
1262 	case DACR32_EL2:	write_sysreg_s(val, SYS_DACR32_EL2);	break;
1263 	case IFSR32_EL2:	write_sysreg_s(val, SYS_IFSR32_EL2);	break;
1264 	case DBGVCR32_EL2:	write_sysreg_s(val, SYS_DBGVCR32_EL2);	break;
1265 	case ZCR_EL1:		write_sysreg_s(val, SYS_ZCR_EL12);	break;
1266 	case SCTLR2_EL1:	write_sysreg_s(val, SYS_SCTLR2_EL12);	break;
1267 	default:		return false;
1268 	}
1269 
1270 	return true;
1271 }
1272 
1273 struct kvm_vm_stat {
1274 	struct kvm_vm_stat_generic generic;
1275 };
1276 
1277 struct kvm_vcpu_stat {
1278 	struct kvm_vcpu_stat_generic generic;
1279 	u64 hvc_exit_stat;
1280 	u64 wfe_exit_stat;
1281 	u64 wfi_exit_stat;
1282 	u64 mmio_exit_user;
1283 	u64 mmio_exit_kernel;
1284 	u64 signal_exits;
1285 	u64 exits;
1286 };
1287 
1288 unsigned long kvm_arm_num_regs(struct kvm_vcpu *vcpu);
1289 int kvm_arm_copy_reg_indices(struct kvm_vcpu *vcpu, u64 __user *indices);
1290 int kvm_arm_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
1291 int kvm_arm_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
1292 
1293 unsigned long kvm_arm_num_sys_reg_descs(struct kvm_vcpu *vcpu);
1294 int kvm_arm_copy_sys_reg_indices(struct kvm_vcpu *vcpu, u64 __user *uindices);
1295 
1296 int __kvm_arm_vcpu_get_events(struct kvm_vcpu *vcpu,
1297 			      struct kvm_vcpu_events *events);
1298 
1299 int __kvm_arm_vcpu_set_events(struct kvm_vcpu *vcpu,
1300 			      struct kvm_vcpu_events *events);
1301 
1302 void kvm_arm_halt_guest(struct kvm *kvm);
1303 void kvm_arm_resume_guest(struct kvm *kvm);
1304 
1305 #define vcpu_has_run_once(vcpu)	(!!READ_ONCE((vcpu)->pid))
1306 
1307 #ifndef __KVM_NVHE_HYPERVISOR__
1308 #define kvm_call_hyp_nvhe(f, ...)						\
1309 	({								\
1310 		struct arm_smccc_res res;				\
1311 									\
1312 		arm_smccc_1_1_hvc(KVM_HOST_SMCCC_FUNC(f),		\
1313 				  ##__VA_ARGS__, &res);			\
1314 		WARN_ON(res.a0 != SMCCC_RET_SUCCESS);			\
1315 									\
1316 		res.a1;							\
1317 	})
1318 
1319 /*
1320  * The isb() below is there to guarantee the same behaviour on VHE as on !VHE,
1321  * where the eret to EL1 acts as a context synchronization event.
1322  */
1323 #define kvm_call_hyp(f, ...)						\
1324 	do {								\
1325 		if (has_vhe()) {					\
1326 			f(__VA_ARGS__);					\
1327 			isb();						\
1328 		} else {						\
1329 			kvm_call_hyp_nvhe(f, ##__VA_ARGS__);		\
1330 		}							\
1331 	} while(0)
1332 
1333 #define kvm_call_hyp_ret(f, ...)					\
1334 	({								\
1335 		typeof(f(__VA_ARGS__)) ret;				\
1336 									\
1337 		if (has_vhe()) {					\
1338 			ret = f(__VA_ARGS__);				\
1339 		} else {						\
1340 			ret = kvm_call_hyp_nvhe(f, ##__VA_ARGS__);	\
1341 		}							\
1342 									\
1343 		ret;							\
1344 	})
1345 #else /* __KVM_NVHE_HYPERVISOR__ */
1346 #define kvm_call_hyp(f, ...) f(__VA_ARGS__)
1347 #define kvm_call_hyp_ret(f, ...) f(__VA_ARGS__)
1348 #define kvm_call_hyp_nvhe(f, ...) f(__VA_ARGS__)
1349 #endif /* __KVM_NVHE_HYPERVISOR__ */
1350 
1351 int handle_exit(struct kvm_vcpu *vcpu, int exception_index);
1352 void handle_exit_early(struct kvm_vcpu *vcpu, int exception_index);
1353 
1354 int kvm_handle_cp14_load_store(struct kvm_vcpu *vcpu);
1355 int kvm_handle_cp14_32(struct kvm_vcpu *vcpu);
1356 int kvm_handle_cp14_64(struct kvm_vcpu *vcpu);
1357 int kvm_handle_cp15_32(struct kvm_vcpu *vcpu);
1358 int kvm_handle_cp15_64(struct kvm_vcpu *vcpu);
1359 int kvm_handle_sys_reg(struct kvm_vcpu *vcpu);
1360 int kvm_handle_cp10_id(struct kvm_vcpu *vcpu);
1361 
1362 void kvm_sys_regs_create_debugfs(struct kvm *kvm);
1363 void kvm_reset_sys_regs(struct kvm_vcpu *vcpu);
1364 
1365 int __init kvm_sys_reg_table_init(void);
1366 struct sys_reg_desc;
1367 int __init populate_sysreg_config(const struct sys_reg_desc *sr,
1368 				  unsigned int idx);
1369 int __init populate_nv_trap_config(void);
1370 
1371 void kvm_calculate_traps(struct kvm_vcpu *vcpu);
1372 
1373 /* MMIO helpers */
1374 void kvm_mmio_write_buf(void *buf, unsigned int len, unsigned long data);
1375 unsigned long kvm_mmio_read_buf(const void *buf, unsigned int len);
1376 
1377 int kvm_handle_mmio_return(struct kvm_vcpu *vcpu);
1378 int io_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa);
1379 
1380 /*
1381  * Returns true if a Performance Monitoring Interrupt (PMI), a.k.a. perf event,
1382  * arrived in guest context.  For arm64, any event that arrives while a vCPU is
1383  * loaded is considered to be "in guest".
1384  */
kvm_arch_pmi_in_guest(struct kvm_vcpu * vcpu)1385 static inline bool kvm_arch_pmi_in_guest(struct kvm_vcpu *vcpu)
1386 {
1387 	return IS_ENABLED(CONFIG_GUEST_PERF_EVENTS) && !!vcpu;
1388 }
1389 
1390 long kvm_hypercall_pv_features(struct kvm_vcpu *vcpu);
1391 gpa_t kvm_init_stolen_time(struct kvm_vcpu *vcpu);
1392 void kvm_update_stolen_time(struct kvm_vcpu *vcpu);
1393 
1394 bool kvm_arm_pvtime_supported(void);
1395 int kvm_arm_pvtime_set_attr(struct kvm_vcpu *vcpu,
1396 			    struct kvm_device_attr *attr);
1397 int kvm_arm_pvtime_get_attr(struct kvm_vcpu *vcpu,
1398 			    struct kvm_device_attr *attr);
1399 int kvm_arm_pvtime_has_attr(struct kvm_vcpu *vcpu,
1400 			    struct kvm_device_attr *attr);
1401 
1402 extern unsigned int __ro_after_init kvm_arm_vmid_bits;
1403 int __init kvm_arm_vmid_alloc_init(void);
1404 void __init kvm_arm_vmid_alloc_free(void);
1405 void kvm_arm_vmid_update(struct kvm_vmid *kvm_vmid);
1406 void kvm_arm_vmid_clear_active(void);
1407 
kvm_arm_pvtime_vcpu_init(struct kvm_vcpu_arch * vcpu_arch)1408 static inline void kvm_arm_pvtime_vcpu_init(struct kvm_vcpu_arch *vcpu_arch)
1409 {
1410 	vcpu_arch->steal.base = INVALID_GPA;
1411 }
1412 
kvm_arm_is_pvtime_enabled(struct kvm_vcpu_arch * vcpu_arch)1413 static inline bool kvm_arm_is_pvtime_enabled(struct kvm_vcpu_arch *vcpu_arch)
1414 {
1415 	return (vcpu_arch->steal.base != INVALID_GPA);
1416 }
1417 
1418 struct kvm_vcpu *kvm_mpidr_to_vcpu(struct kvm *kvm, unsigned long mpidr);
1419 
1420 DECLARE_KVM_HYP_PER_CPU(struct kvm_host_data, kvm_host_data);
1421 
1422 /*
1423  * How we access per-CPU host data depends on the where we access it from,
1424  * and the mode we're in:
1425  *
1426  * - VHE and nVHE hypervisor bits use their locally defined instance
1427  *
1428  * - the rest of the kernel use either the VHE or nVHE one, depending on
1429  *   the mode we're running in.
1430  *
1431  *   Unless we're in protected mode, fully deprivileged, and the nVHE
1432  *   per-CPU stuff is exclusively accessible to the protected EL2 code.
1433  *   In this case, the EL1 code uses the *VHE* data as its private state
1434  *   (which makes sense in a way as there shouldn't be any shared state
1435  *   between the host and the hypervisor).
1436  *
1437  * Yes, this is all totally trivial. Shoot me now.
1438  */
1439 #if defined(__KVM_NVHE_HYPERVISOR__) || defined(__KVM_VHE_HYPERVISOR__)
1440 #define host_data_ptr(f)	(&this_cpu_ptr(&kvm_host_data)->f)
1441 #else
1442 #define host_data_ptr(f)						\
1443 	(static_branch_unlikely(&kvm_protected_mode_initialized) ?	\
1444 	 &this_cpu_ptr(&kvm_host_data)->f :				\
1445 	 &this_cpu_ptr_hyp_sym(kvm_host_data)->f)
1446 #endif
1447 
1448 #define host_data_test_flag(flag)					\
1449 	(test_bit(KVM_HOST_DATA_FLAG_##flag, host_data_ptr(flags)))
1450 #define host_data_set_flag(flag)					\
1451 	set_bit(KVM_HOST_DATA_FLAG_##flag, host_data_ptr(flags))
1452 #define host_data_clear_flag(flag)					\
1453 	clear_bit(KVM_HOST_DATA_FLAG_##flag, host_data_ptr(flags))
1454 
1455 /* Check whether the FP regs are owned by the guest */
guest_owns_fp_regs(void)1456 static inline bool guest_owns_fp_regs(void)
1457 {
1458 	return *host_data_ptr(fp_owner) == FP_STATE_GUEST_OWNED;
1459 }
1460 
1461 /* Check whether the FP regs are owned by the host */
host_owns_fp_regs(void)1462 static inline bool host_owns_fp_regs(void)
1463 {
1464 	return *host_data_ptr(fp_owner) == FP_STATE_HOST_OWNED;
1465 }
1466 
kvm_init_host_cpu_context(struct kvm_cpu_context * cpu_ctxt)1467 static inline void kvm_init_host_cpu_context(struct kvm_cpu_context *cpu_ctxt)
1468 {
1469 	/* The host's MPIDR is immutable, so let's set it up at boot time */
1470 	ctxt_sys_reg(cpu_ctxt, MPIDR_EL1) = read_cpuid_mpidr();
1471 }
1472 
kvm_system_needs_idmapped_vectors(void)1473 static inline bool kvm_system_needs_idmapped_vectors(void)
1474 {
1475 	return cpus_have_final_cap(ARM64_SPECTRE_V3A);
1476 }
1477 
1478 void kvm_init_host_debug_data(void);
1479 void kvm_vcpu_load_debug(struct kvm_vcpu *vcpu);
1480 void kvm_vcpu_put_debug(struct kvm_vcpu *vcpu);
1481 void kvm_debug_set_guest_ownership(struct kvm_vcpu *vcpu);
1482 void kvm_debug_handle_oslar(struct kvm_vcpu *vcpu, u64 val);
1483 
1484 #define kvm_vcpu_os_lock_enabled(vcpu)		\
1485 	(!!(__vcpu_sys_reg(vcpu, OSLSR_EL1) & OSLSR_EL1_OSLK))
1486 
1487 #define kvm_debug_regs_in_use(vcpu)		\
1488 	((vcpu)->arch.debug_owner != VCPU_DEBUG_FREE)
1489 #define kvm_host_owns_debug_regs(vcpu)		\
1490 	((vcpu)->arch.debug_owner == VCPU_DEBUG_HOST_OWNED)
1491 #define kvm_guest_owns_debug_regs(vcpu)		\
1492 	((vcpu)->arch.debug_owner == VCPU_DEBUG_GUEST_OWNED)
1493 
1494 int kvm_arm_vcpu_arch_set_attr(struct kvm_vcpu *vcpu,
1495 			       struct kvm_device_attr *attr);
1496 int kvm_arm_vcpu_arch_get_attr(struct kvm_vcpu *vcpu,
1497 			       struct kvm_device_attr *attr);
1498 int kvm_arm_vcpu_arch_has_attr(struct kvm_vcpu *vcpu,
1499 			       struct kvm_device_attr *attr);
1500 
1501 int kvm_vm_ioctl_mte_copy_tags(struct kvm *kvm,
1502 			       struct kvm_arm_copy_mte_tags *copy_tags);
1503 int kvm_vm_ioctl_set_counter_offset(struct kvm *kvm,
1504 				    struct kvm_arm_counter_offset *offset);
1505 int kvm_vm_ioctl_get_reg_writable_masks(struct kvm *kvm,
1506 					struct reg_mask_range *range);
1507 
1508 /* Guest/host FPSIMD coordination helpers */
1509 void kvm_arch_vcpu_load_fp(struct kvm_vcpu *vcpu);
1510 void kvm_arch_vcpu_ctxflush_fp(struct kvm_vcpu *vcpu);
1511 void kvm_arch_vcpu_ctxsync_fp(struct kvm_vcpu *vcpu);
1512 void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu);
1513 
kvm_pmu_counter_deferred(struct perf_event_attr * attr)1514 static inline bool kvm_pmu_counter_deferred(struct perf_event_attr *attr)
1515 {
1516 	return (!has_vhe() && attr->exclude_host);
1517 }
1518 
1519 #ifdef CONFIG_KVM
1520 void kvm_set_pmu_events(u64 set, struct perf_event_attr *attr);
1521 void kvm_clr_pmu_events(u64 clr);
1522 bool kvm_set_pmuserenr(u64 val);
1523 void kvm_enable_trbe(void);
1524 void kvm_disable_trbe(void);
1525 void kvm_tracing_set_el1_configuration(u64 trfcr_while_in_guest);
1526 #else
kvm_set_pmu_events(u64 set,struct perf_event_attr * attr)1527 static inline void kvm_set_pmu_events(u64 set, struct perf_event_attr *attr) {}
kvm_clr_pmu_events(u64 clr)1528 static inline void kvm_clr_pmu_events(u64 clr) {}
kvm_set_pmuserenr(u64 val)1529 static inline bool kvm_set_pmuserenr(u64 val)
1530 {
1531 	return false;
1532 }
kvm_enable_trbe(void)1533 static inline void kvm_enable_trbe(void) {}
kvm_disable_trbe(void)1534 static inline void kvm_disable_trbe(void) {}
kvm_tracing_set_el1_configuration(u64 trfcr_while_in_guest)1535 static inline void kvm_tracing_set_el1_configuration(u64 trfcr_while_in_guest) {}
1536 #endif
1537 
1538 void kvm_vcpu_load_vhe(struct kvm_vcpu *vcpu);
1539 void kvm_vcpu_put_vhe(struct kvm_vcpu *vcpu);
1540 
1541 int __init kvm_set_ipa_limit(void);
1542 u32 kvm_get_pa_bits(struct kvm *kvm);
1543 
1544 #define __KVM_HAVE_ARCH_VM_ALLOC
1545 struct kvm *kvm_arch_alloc_vm(void);
1546 
1547 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS
1548 
1549 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS_RANGE
1550 
1551 #define kvm_vm_is_protected(kvm)	(is_protected_kvm_enabled() && (kvm)->arch.pkvm.enabled)
1552 
1553 #define vcpu_is_protected(vcpu)		kvm_vm_is_protected((vcpu)->kvm)
1554 
1555 int kvm_arm_vcpu_finalize(struct kvm_vcpu *vcpu, int feature);
1556 bool kvm_arm_vcpu_is_finalized(struct kvm_vcpu *vcpu);
1557 
1558 #define kvm_arm_vcpu_sve_finalized(vcpu) vcpu_get_flag(vcpu, VCPU_SVE_FINALIZED)
1559 
1560 #define kvm_has_mte(kvm)					\
1561 	(system_supports_mte() &&				\
1562 	 test_bit(KVM_ARCH_FLAG_MTE_ENABLED, &(kvm)->arch.flags))
1563 
1564 #define kvm_supports_32bit_el0()				\
1565 	(system_supports_32bit_el0() &&				\
1566 	 !static_branch_unlikely(&arm64_mismatched_32bit_el0))
1567 
1568 #define kvm_vm_has_ran_once(kvm)					\
1569 	(test_bit(KVM_ARCH_FLAG_HAS_RAN_ONCE, &(kvm)->arch.flags))
1570 
__vcpu_has_feature(const struct kvm_arch * ka,int feature)1571 static inline bool __vcpu_has_feature(const struct kvm_arch *ka, int feature)
1572 {
1573 	return test_bit(feature, ka->vcpu_features);
1574 }
1575 
1576 #define kvm_vcpu_has_feature(k, f)	__vcpu_has_feature(&(k)->arch, (f))
1577 #define vcpu_has_feature(v, f)	__vcpu_has_feature(&(v)->kvm->arch, (f))
1578 
1579 #define kvm_vcpu_initialized(v) vcpu_get_flag(vcpu, VCPU_INITIALIZED)
1580 
1581 int kvm_trng_call(struct kvm_vcpu *vcpu);
1582 #ifdef CONFIG_KVM
1583 extern phys_addr_t hyp_mem_base;
1584 extern phys_addr_t hyp_mem_size;
1585 void __init kvm_hyp_reserve(void);
1586 #else
kvm_hyp_reserve(void)1587 static inline void kvm_hyp_reserve(void) { }
1588 #endif
1589 
1590 void kvm_arm_vcpu_power_off(struct kvm_vcpu *vcpu);
1591 bool kvm_arm_vcpu_stopped(struct kvm_vcpu *vcpu);
1592 
__vm_id_reg(struct kvm_arch * ka,u32 reg)1593 static inline u64 *__vm_id_reg(struct kvm_arch *ka, u32 reg)
1594 {
1595 	switch (reg) {
1596 	case sys_reg(3, 0, 0, 1, 0) ... sys_reg(3, 0, 0, 7, 7):
1597 		return &ka->id_regs[IDREG_IDX(reg)];
1598 	case SYS_CTR_EL0:
1599 		return &ka->ctr_el0;
1600 	case SYS_MIDR_EL1:
1601 		return &ka->midr_el1;
1602 	case SYS_REVIDR_EL1:
1603 		return &ka->revidr_el1;
1604 	case SYS_AIDR_EL1:
1605 		return &ka->aidr_el1;
1606 	default:
1607 		WARN_ON_ONCE(1);
1608 		return NULL;
1609 	}
1610 }
1611 
1612 #define kvm_read_vm_id_reg(kvm, reg)					\
1613 	({ u64 __val = *__vm_id_reg(&(kvm)->arch, reg); __val; })
1614 
1615 void kvm_set_vm_id_reg(struct kvm *kvm, u32 reg, u64 val);
1616 
1617 #define __expand_field_sign_unsigned(id, fld, val)			\
1618 	((u64)SYS_FIELD_VALUE(id, fld, val))
1619 
1620 #define __expand_field_sign_signed(id, fld, val)			\
1621 	({								\
1622 		u64 __val = SYS_FIELD_VALUE(id, fld, val);		\
1623 		sign_extend64(__val, id##_##fld##_WIDTH - 1);		\
1624 	})
1625 
1626 #define get_idreg_field_unsigned(kvm, id, fld)				\
1627 	({								\
1628 		u64 __val = kvm_read_vm_id_reg((kvm), SYS_##id);	\
1629 		FIELD_GET(id##_##fld##_MASK, __val);			\
1630 	})
1631 
1632 #define get_idreg_field_signed(kvm, id, fld)				\
1633 	({								\
1634 		u64 __val = get_idreg_field_unsigned(kvm, id, fld);	\
1635 		sign_extend64(__val, id##_##fld##_WIDTH - 1);		\
1636 	})
1637 
1638 #define get_idreg_field_enum(kvm, id, fld)				\
1639 	get_idreg_field_unsigned(kvm, id, fld)
1640 
1641 #define kvm_cmp_feat_signed(kvm, id, fld, op, limit)			\
1642 	(get_idreg_field_signed((kvm), id, fld) op __expand_field_sign_signed(id, fld, limit))
1643 
1644 #define kvm_cmp_feat_unsigned(kvm, id, fld, op, limit)			\
1645 	(get_idreg_field_unsigned((kvm), id, fld) op __expand_field_sign_unsigned(id, fld, limit))
1646 
1647 #define kvm_cmp_feat(kvm, id, fld, op, limit)				\
1648 	(id##_##fld##_SIGNED ?						\
1649 	 kvm_cmp_feat_signed(kvm, id, fld, op, limit) :			\
1650 	 kvm_cmp_feat_unsigned(kvm, id, fld, op, limit))
1651 
1652 #define __kvm_has_feat(kvm, id, fld, limit)				\
1653 	kvm_cmp_feat(kvm, id, fld, >=, limit)
1654 
1655 #define kvm_has_feat(kvm, ...) __kvm_has_feat(kvm, __VA_ARGS__)
1656 
1657 #define __kvm_has_feat_enum(kvm, id, fld, val)				\
1658 	kvm_cmp_feat_unsigned(kvm, id, fld, ==, val)
1659 
1660 #define kvm_has_feat_enum(kvm, ...) __kvm_has_feat_enum(kvm, __VA_ARGS__)
1661 
1662 #define kvm_has_feat_range(kvm, id, fld, min, max)			\
1663 	(kvm_cmp_feat(kvm, id, fld, >=, min) &&				\
1664 	kvm_cmp_feat(kvm, id, fld, <=, max))
1665 
1666 /* Check for a given level of PAuth support */
1667 #define kvm_has_pauth(k, l)						\
1668 	({								\
1669 		bool pa, pi, pa3;					\
1670 									\
1671 		pa  = kvm_has_feat((k), ID_AA64ISAR1_EL1, APA, l);	\
1672 		pa &= kvm_has_feat((k), ID_AA64ISAR1_EL1, GPA, IMP);	\
1673 		pi  = kvm_has_feat((k), ID_AA64ISAR1_EL1, API, l);	\
1674 		pi &= kvm_has_feat((k), ID_AA64ISAR1_EL1, GPI, IMP);	\
1675 		pa3  = kvm_has_feat((k), ID_AA64ISAR2_EL1, APA3, l);	\
1676 		pa3 &= kvm_has_feat((k), ID_AA64ISAR2_EL1, GPA3, IMP);	\
1677 									\
1678 		(pa + pi + pa3) == 1;					\
1679 	})
1680 
1681 #define kvm_has_fpmr(k)					\
1682 	(system_supports_fpmr() &&			\
1683 	 kvm_has_feat((k), ID_AA64PFR2_EL1, FPMR, IMP))
1684 
1685 #define kvm_has_tcr2(k)				\
1686 	(kvm_has_feat((k), ID_AA64MMFR3_EL1, TCRX, IMP))
1687 
1688 #define kvm_has_s1pie(k)				\
1689 	(kvm_has_feat((k), ID_AA64MMFR3_EL1, S1PIE, IMP))
1690 
1691 #define kvm_has_s1poe(k)				\
1692 	(kvm_has_feat((k), ID_AA64MMFR3_EL1, S1POE, IMP))
1693 
1694 #define kvm_has_ras(k)					\
1695 	(kvm_has_feat((k), ID_AA64PFR0_EL1, RAS, IMP))
1696 
1697 #define kvm_has_sctlr2(k)				\
1698 	(kvm_has_feat((k), ID_AA64MMFR3_EL1, SCTLRX, IMP))
1699 
kvm_arch_has_irq_bypass(void)1700 static inline bool kvm_arch_has_irq_bypass(void)
1701 {
1702 	return true;
1703 }
1704 
1705 void compute_fgu(struct kvm *kvm, enum fgt_group_id fgt);
1706 void get_reg_fixed_bits(struct kvm *kvm, enum vcpu_sysreg reg, u64 *res0, u64 *res1);
1707 void check_feature_map(void);
1708 
1709 
1710 #endif /* __ARM64_KVM_HOST_H__ */
1711