xref: /openbmc/linux/arch/arm64/include/asm/kvm_host.h (revision f05643a0)
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/percpu.h>
20 #include <linux/psci.h>
21 #include <asm/arch_gicv3.h>
22 #include <asm/barrier.h>
23 #include <asm/cpufeature.h>
24 #include <asm/cputype.h>
25 #include <asm/daifflags.h>
26 #include <asm/fpsimd.h>
27 #include <asm/kvm.h>
28 #include <asm/kvm_asm.h>
29 
30 #define __KVM_HAVE_ARCH_INTC_INITIALIZED
31 
32 #define KVM_HALT_POLL_NS_DEFAULT 500000
33 
34 #include <kvm/arm_vgic.h>
35 #include <kvm/arm_arch_timer.h>
36 #include <kvm/arm_pmu.h>
37 
38 #define KVM_MAX_VCPUS VGIC_V3_MAX_CPUS
39 
40 #define KVM_VCPU_MAX_FEATURES 7
41 
42 #define KVM_REQ_SLEEP \
43 	KVM_ARCH_REQ_FLAGS(0, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
44 #define KVM_REQ_IRQ_PENDING	KVM_ARCH_REQ(1)
45 #define KVM_REQ_VCPU_RESET	KVM_ARCH_REQ(2)
46 #define KVM_REQ_RECORD_STEAL	KVM_ARCH_REQ(3)
47 #define KVM_REQ_RELOAD_GICv4	KVM_ARCH_REQ(4)
48 #define KVM_REQ_RELOAD_PMU	KVM_ARCH_REQ(5)
49 #define KVM_REQ_SUSPEND		KVM_ARCH_REQ(6)
50 
51 #define KVM_DIRTY_LOG_MANUAL_CAPS   (KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE | \
52 				     KVM_DIRTY_LOG_INITIALLY_SET)
53 
54 #define KVM_HAVE_MMU_RWLOCK
55 
56 /*
57  * Mode of operation configurable with kvm-arm.mode early param.
58  * See Documentation/admin-guide/kernel-parameters.txt for more information.
59  */
60 enum kvm_mode {
61 	KVM_MODE_DEFAULT,
62 	KVM_MODE_PROTECTED,
63 	KVM_MODE_NONE,
64 };
65 enum kvm_mode kvm_get_mode(void);
66 
67 DECLARE_STATIC_KEY_FALSE(userspace_irqchip_in_use);
68 
69 extern unsigned int kvm_sve_max_vl;
70 int kvm_arm_init_sve(void);
71 
72 u32 __attribute_const__ kvm_target_cpu(void);
73 int kvm_reset_vcpu(struct kvm_vcpu *vcpu);
74 void kvm_arm_vcpu_destroy(struct kvm_vcpu *vcpu);
75 
76 struct kvm_vmid {
77 	atomic64_t id;
78 };
79 
80 struct kvm_s2_mmu {
81 	struct kvm_vmid vmid;
82 
83 	/*
84 	 * stage2 entry level table
85 	 *
86 	 * Two kvm_s2_mmu structures in the same VM can point to the same
87 	 * pgd here.  This happens when running a guest using a
88 	 * translation regime that isn't affected by its own stage-2
89 	 * translation, such as a non-VHE hypervisor running at vEL2, or
90 	 * for vEL1/EL0 with vHCR_EL2.VM == 0.  In that case, we use the
91 	 * canonical stage-2 page tables.
92 	 */
93 	phys_addr_t	pgd_phys;
94 	struct kvm_pgtable *pgt;
95 
96 	/* The last vcpu id that ran on each physical CPU */
97 	int __percpu *last_vcpu_ran;
98 
99 	struct kvm_arch *arch;
100 };
101 
102 struct kvm_arch_memory_slot {
103 };
104 
105 /**
106  * struct kvm_smccc_features: Descriptor of the hypercall services exposed to the guests
107  *
108  * @std_bmap: Bitmap of standard secure service calls
109  * @std_hyp_bmap: Bitmap of standard hypervisor service calls
110  * @vendor_hyp_bmap: Bitmap of vendor specific hypervisor service calls
111  */
112 struct kvm_smccc_features {
113 	unsigned long std_bmap;
114 	unsigned long std_hyp_bmap;
115 	unsigned long vendor_hyp_bmap;
116 };
117 
118 struct kvm_arch {
119 	struct kvm_s2_mmu mmu;
120 
121 	/* VTCR_EL2 value for this VM */
122 	u64    vtcr;
123 
124 	/* Interrupt controller */
125 	struct vgic_dist	vgic;
126 
127 	/* Mandated version of PSCI */
128 	u32 psci_version;
129 
130 	/*
131 	 * If we encounter a data abort without valid instruction syndrome
132 	 * information, report this to user space.  User space can (and
133 	 * should) opt in to this feature if KVM_CAP_ARM_NISV_TO_USER is
134 	 * supported.
135 	 */
136 #define KVM_ARCH_FLAG_RETURN_NISV_IO_ABORT_TO_USER	0
137 	/* Memory Tagging Extension enabled for the guest */
138 #define KVM_ARCH_FLAG_MTE_ENABLED			1
139 	/* At least one vCPU has ran in the VM */
140 #define KVM_ARCH_FLAG_HAS_RAN_ONCE			2
141 	/*
142 	 * The following two bits are used to indicate the guest's EL1
143 	 * register width configuration. A value of KVM_ARCH_FLAG_EL1_32BIT
144 	 * bit is valid only when KVM_ARCH_FLAG_REG_WIDTH_CONFIGURED is set.
145 	 * Otherwise, the guest's EL1 register width has not yet been
146 	 * determined yet.
147 	 */
148 #define KVM_ARCH_FLAG_REG_WIDTH_CONFIGURED		3
149 #define KVM_ARCH_FLAG_EL1_32BIT				4
150 	/* PSCI SYSTEM_SUSPEND enabled for the guest */
151 #define KVM_ARCH_FLAG_SYSTEM_SUSPEND_ENABLED		5
152 
153 	unsigned long flags;
154 
155 	/*
156 	 * VM-wide PMU filter, implemented as a bitmap and big enough for
157 	 * up to 2^10 events (ARMv8.0) or 2^16 events (ARMv8.1+).
158 	 */
159 	unsigned long *pmu_filter;
160 	struct arm_pmu *arm_pmu;
161 
162 	cpumask_var_t supported_cpus;
163 
164 	u8 pfr0_csv2;
165 	u8 pfr0_csv3;
166 
167 	/* Hypercall features firmware registers' descriptor */
168 	struct kvm_smccc_features smccc_feat;
169 };
170 
171 struct kvm_vcpu_fault_info {
172 	u64 esr_el2;		/* Hyp Syndrom Register */
173 	u64 far_el2;		/* Hyp Fault Address Register */
174 	u64 hpfar_el2;		/* Hyp IPA Fault Address Register */
175 	u64 disr_el1;		/* Deferred [SError] Status Register */
176 };
177 
178 enum vcpu_sysreg {
179 	__INVALID_SYSREG__,   /* 0 is reserved as an invalid value */
180 	MPIDR_EL1,	/* MultiProcessor Affinity Register */
181 	CSSELR_EL1,	/* Cache Size Selection Register */
182 	SCTLR_EL1,	/* System Control Register */
183 	ACTLR_EL1,	/* Auxiliary Control Register */
184 	CPACR_EL1,	/* Coprocessor Access Control */
185 	ZCR_EL1,	/* SVE Control */
186 	TTBR0_EL1,	/* Translation Table Base Register 0 */
187 	TTBR1_EL1,	/* Translation Table Base Register 1 */
188 	TCR_EL1,	/* Translation Control Register */
189 	ESR_EL1,	/* Exception Syndrome Register */
190 	AFSR0_EL1,	/* Auxiliary Fault Status Register 0 */
191 	AFSR1_EL1,	/* Auxiliary Fault Status Register 1 */
192 	FAR_EL1,	/* Fault Address Register */
193 	MAIR_EL1,	/* Memory Attribute Indirection Register */
194 	VBAR_EL1,	/* Vector Base Address Register */
195 	CONTEXTIDR_EL1,	/* Context ID Register */
196 	TPIDR_EL0,	/* Thread ID, User R/W */
197 	TPIDRRO_EL0,	/* Thread ID, User R/O */
198 	TPIDR_EL1,	/* Thread ID, Privileged */
199 	AMAIR_EL1,	/* Aux Memory Attribute Indirection Register */
200 	CNTKCTL_EL1,	/* Timer Control Register (EL1) */
201 	PAR_EL1,	/* Physical Address Register */
202 	MDSCR_EL1,	/* Monitor Debug System Control Register */
203 	MDCCINT_EL1,	/* Monitor Debug Comms Channel Interrupt Enable Reg */
204 	OSLSR_EL1,	/* OS Lock Status Register */
205 	DISR_EL1,	/* Deferred Interrupt Status Register */
206 
207 	/* Performance Monitors Registers */
208 	PMCR_EL0,	/* Control Register */
209 	PMSELR_EL0,	/* Event Counter Selection Register */
210 	PMEVCNTR0_EL0,	/* Event Counter Register (0-30) */
211 	PMEVCNTR30_EL0 = PMEVCNTR0_EL0 + 30,
212 	PMCCNTR_EL0,	/* Cycle Counter Register */
213 	PMEVTYPER0_EL0,	/* Event Type Register (0-30) */
214 	PMEVTYPER30_EL0 = PMEVTYPER0_EL0 + 30,
215 	PMCCFILTR_EL0,	/* Cycle Count Filter Register */
216 	PMCNTENSET_EL0,	/* Count Enable Set Register */
217 	PMINTENSET_EL1,	/* Interrupt Enable Set Register */
218 	PMOVSSET_EL0,	/* Overflow Flag Status Set Register */
219 	PMUSERENR_EL0,	/* User Enable Register */
220 
221 	/* Pointer Authentication Registers in a strict increasing order. */
222 	APIAKEYLO_EL1,
223 	APIAKEYHI_EL1,
224 	APIBKEYLO_EL1,
225 	APIBKEYHI_EL1,
226 	APDAKEYLO_EL1,
227 	APDAKEYHI_EL1,
228 	APDBKEYLO_EL1,
229 	APDBKEYHI_EL1,
230 	APGAKEYLO_EL1,
231 	APGAKEYHI_EL1,
232 
233 	ELR_EL1,
234 	SP_EL1,
235 	SPSR_EL1,
236 
237 	CNTVOFF_EL2,
238 	CNTV_CVAL_EL0,
239 	CNTV_CTL_EL0,
240 	CNTP_CVAL_EL0,
241 	CNTP_CTL_EL0,
242 
243 	/* Memory Tagging Extension registers */
244 	RGSR_EL1,	/* Random Allocation Tag Seed Register */
245 	GCR_EL1,	/* Tag Control Register */
246 	TFSR_EL1,	/* Tag Fault Status Register (EL1) */
247 	TFSRE0_EL1,	/* Tag Fault Status Register (EL0) */
248 
249 	/* 32bit specific registers. Keep them at the end of the range */
250 	DACR32_EL2,	/* Domain Access Control Register */
251 	IFSR32_EL2,	/* Instruction Fault Status Register */
252 	FPEXC32_EL2,	/* Floating-Point Exception Control Register */
253 	DBGVCR32_EL2,	/* Debug Vector Catch Register */
254 
255 	NR_SYS_REGS	/* Nothing after this line! */
256 };
257 
258 struct kvm_cpu_context {
259 	struct user_pt_regs regs;	/* sp = sp_el0 */
260 
261 	u64	spsr_abt;
262 	u64	spsr_und;
263 	u64	spsr_irq;
264 	u64	spsr_fiq;
265 
266 	struct user_fpsimd_state fp_regs;
267 
268 	u64 sys_regs[NR_SYS_REGS];
269 
270 	struct kvm_vcpu *__hyp_running_vcpu;
271 };
272 
273 struct kvm_host_data {
274 	struct kvm_cpu_context host_ctxt;
275 };
276 
277 struct kvm_host_psci_config {
278 	/* PSCI version used by host. */
279 	u32 version;
280 
281 	/* Function IDs used by host if version is v0.1. */
282 	struct psci_0_1_function_ids function_ids_0_1;
283 
284 	bool psci_0_1_cpu_suspend_implemented;
285 	bool psci_0_1_cpu_on_implemented;
286 	bool psci_0_1_cpu_off_implemented;
287 	bool psci_0_1_migrate_implemented;
288 };
289 
290 extern struct kvm_host_psci_config kvm_nvhe_sym(kvm_host_psci_config);
291 #define kvm_host_psci_config CHOOSE_NVHE_SYM(kvm_host_psci_config)
292 
293 extern s64 kvm_nvhe_sym(hyp_physvirt_offset);
294 #define hyp_physvirt_offset CHOOSE_NVHE_SYM(hyp_physvirt_offset)
295 
296 extern u64 kvm_nvhe_sym(hyp_cpu_logical_map)[NR_CPUS];
297 #define hyp_cpu_logical_map CHOOSE_NVHE_SYM(hyp_cpu_logical_map)
298 
299 struct vcpu_reset_state {
300 	unsigned long	pc;
301 	unsigned long	r0;
302 	bool		be;
303 	bool		reset;
304 };
305 
306 struct kvm_vcpu_arch {
307 	struct kvm_cpu_context ctxt;
308 
309 	/* Guest floating point state */
310 	void *sve_state;
311 	unsigned int sve_max_vl;
312 	u64 svcr;
313 
314 	/* Stage 2 paging state used by the hardware on next switch */
315 	struct kvm_s2_mmu *hw_mmu;
316 
317 	/* Values of trap registers for the guest. */
318 	u64 hcr_el2;
319 	u64 mdcr_el2;
320 	u64 cptr_el2;
321 
322 	/* Values of trap registers for the host before guest entry. */
323 	u64 mdcr_el2_host;
324 
325 	/* Exception Information */
326 	struct kvm_vcpu_fault_info fault;
327 
328 	/* Miscellaneous vcpu state flags */
329 	u64 flags;
330 
331 	/*
332 	 * We maintain more than a single set of debug registers to support
333 	 * debugging the guest from the host and to maintain separate host and
334 	 * guest state during world switches. vcpu_debug_state are the debug
335 	 * registers of the vcpu as the guest sees them.  host_debug_state are
336 	 * the host registers which are saved and restored during
337 	 * world switches. external_debug_state contains the debug
338 	 * values we want to debug the guest. This is set via the
339 	 * KVM_SET_GUEST_DEBUG ioctl.
340 	 *
341 	 * debug_ptr points to the set of debug registers that should be loaded
342 	 * onto the hardware when running the guest.
343 	 */
344 	struct kvm_guest_debug_arch *debug_ptr;
345 	struct kvm_guest_debug_arch vcpu_debug_state;
346 	struct kvm_guest_debug_arch external_debug_state;
347 
348 	struct user_fpsimd_state *host_fpsimd_state;	/* hyp VA */
349 	struct task_struct *parent_task;
350 
351 	struct {
352 		/* {Break,watch}point registers */
353 		struct kvm_guest_debug_arch regs;
354 		/* Statistical profiling extension */
355 		u64 pmscr_el1;
356 		/* Self-hosted trace */
357 		u64 trfcr_el1;
358 	} host_debug_state;
359 
360 	/* VGIC state */
361 	struct vgic_cpu vgic_cpu;
362 	struct arch_timer_cpu timer_cpu;
363 	struct kvm_pmu pmu;
364 
365 	/*
366 	 * Guest registers we preserve during guest debugging.
367 	 *
368 	 * These shadow registers are updated by the kvm_handle_sys_reg
369 	 * trap handler if the guest accesses or updates them while we
370 	 * are using guest debug.
371 	 */
372 	struct {
373 		u32	mdscr_el1;
374 	} guest_debug_preserved;
375 
376 	/* vcpu power state */
377 	struct kvm_mp_state mp_state;
378 
379 	/* Don't run the guest (internal implementation need) */
380 	bool pause;
381 
382 	/* Cache some mmu pages needed inside spinlock regions */
383 	struct kvm_mmu_memory_cache mmu_page_cache;
384 
385 	/* Target CPU and feature flags */
386 	int target;
387 	DECLARE_BITMAP(features, KVM_VCPU_MAX_FEATURES);
388 
389 	/* Virtual SError ESR to restore when HCR_EL2.VSE is set */
390 	u64 vsesr_el2;
391 
392 	/* Additional reset state */
393 	struct vcpu_reset_state	reset_state;
394 
395 	/* True when deferrable sysregs are loaded on the physical CPU,
396 	 * see kvm_vcpu_load_sysregs_vhe and kvm_vcpu_put_sysregs_vhe. */
397 	bool sysregs_loaded_on_cpu;
398 
399 	/* Guest PV state */
400 	struct {
401 		u64 last_steal;
402 		gpa_t base;
403 	} steal;
404 };
405 
406 /* Pointer to the vcpu's SVE FFR for sve_{save,load}_state() */
407 #define vcpu_sve_pffr(vcpu) (kern_hyp_va((vcpu)->arch.sve_state) +	\
408 			     sve_ffr_offset((vcpu)->arch.sve_max_vl))
409 
410 #define vcpu_sve_max_vq(vcpu)	sve_vq_from_vl((vcpu)->arch.sve_max_vl)
411 
412 #define vcpu_sve_state_size(vcpu) ({					\
413 	size_t __size_ret;						\
414 	unsigned int __vcpu_vq;						\
415 									\
416 	if (WARN_ON(!sve_vl_valid((vcpu)->arch.sve_max_vl))) {		\
417 		__size_ret = 0;						\
418 	} else {							\
419 		__vcpu_vq = vcpu_sve_max_vq(vcpu);			\
420 		__size_ret = SVE_SIG_REGS_SIZE(__vcpu_vq);		\
421 	}								\
422 									\
423 	__size_ret;							\
424 })
425 
426 /* vcpu_arch flags field values: */
427 #define KVM_ARM64_DEBUG_DIRTY		(1 << 0)
428 #define KVM_ARM64_FP_ENABLED		(1 << 1) /* guest FP regs loaded */
429 #define KVM_ARM64_FP_HOST		(1 << 2) /* host FP regs loaded */
430 #define KVM_ARM64_HOST_SVE_ENABLED	(1 << 4) /* SVE enabled for EL0 */
431 #define KVM_ARM64_GUEST_HAS_SVE		(1 << 5) /* SVE exposed to guest */
432 #define KVM_ARM64_VCPU_SVE_FINALIZED	(1 << 6) /* SVE config completed */
433 #define KVM_ARM64_GUEST_HAS_PTRAUTH	(1 << 7) /* PTRAUTH exposed to guest */
434 #define KVM_ARM64_PENDING_EXCEPTION	(1 << 8) /* Exception pending */
435 /*
436  * Overlaps with KVM_ARM64_EXCEPT_MASK on purpose so that it can't be
437  * set together with an exception...
438  */
439 #define KVM_ARM64_INCREMENT_PC		(1 << 9) /* Increment PC */
440 #define KVM_ARM64_EXCEPT_MASK		(7 << 9) /* Target EL/MODE */
441 /*
442  * When KVM_ARM64_PENDING_EXCEPTION is set, KVM_ARM64_EXCEPT_MASK can
443  * take the following values:
444  *
445  * For AArch32 EL1:
446  */
447 #define KVM_ARM64_EXCEPT_AA32_UND	(0 << 9)
448 #define KVM_ARM64_EXCEPT_AA32_IABT	(1 << 9)
449 #define KVM_ARM64_EXCEPT_AA32_DABT	(2 << 9)
450 /* For AArch64: */
451 #define KVM_ARM64_EXCEPT_AA64_ELx_SYNC	(0 << 9)
452 #define KVM_ARM64_EXCEPT_AA64_ELx_IRQ	(1 << 9)
453 #define KVM_ARM64_EXCEPT_AA64_ELx_FIQ	(2 << 9)
454 #define KVM_ARM64_EXCEPT_AA64_ELx_SERR	(3 << 9)
455 #define KVM_ARM64_EXCEPT_AA64_EL1	(0 << 11)
456 #define KVM_ARM64_EXCEPT_AA64_EL2	(1 << 11)
457 
458 #define KVM_ARM64_DEBUG_STATE_SAVE_SPE	(1 << 12) /* Save SPE context if active  */
459 #define KVM_ARM64_DEBUG_STATE_SAVE_TRBE	(1 << 13) /* Save TRBE context if active  */
460 #define KVM_ARM64_FP_FOREIGN_FPSTATE	(1 << 14)
461 #define KVM_ARM64_ON_UNSUPPORTED_CPU	(1 << 15) /* Physical CPU not in supported_cpus */
462 #define KVM_ARM64_HOST_SME_ENABLED	(1 << 16) /* SME enabled for EL0 */
463 #define KVM_ARM64_WFIT			(1 << 17) /* WFIT instruction trapped */
464 
465 #define KVM_GUESTDBG_VALID_MASK (KVM_GUESTDBG_ENABLE | \
466 				 KVM_GUESTDBG_USE_SW_BP | \
467 				 KVM_GUESTDBG_USE_HW | \
468 				 KVM_GUESTDBG_SINGLESTEP)
469 
470 #define vcpu_has_sve(vcpu) (system_supports_sve() &&			\
471 			    ((vcpu)->arch.flags & KVM_ARM64_GUEST_HAS_SVE))
472 
473 #ifdef CONFIG_ARM64_PTR_AUTH
474 #define vcpu_has_ptrauth(vcpu)						\
475 	((cpus_have_final_cap(ARM64_HAS_ADDRESS_AUTH) ||		\
476 	  cpus_have_final_cap(ARM64_HAS_GENERIC_AUTH)) &&		\
477 	 (vcpu)->arch.flags & KVM_ARM64_GUEST_HAS_PTRAUTH)
478 #else
479 #define vcpu_has_ptrauth(vcpu)		false
480 #endif
481 
482 #define vcpu_on_unsupported_cpu(vcpu)					\
483 	((vcpu)->arch.flags & KVM_ARM64_ON_UNSUPPORTED_CPU)
484 
485 #define vcpu_set_on_unsupported_cpu(vcpu)				\
486 	((vcpu)->arch.flags |= KVM_ARM64_ON_UNSUPPORTED_CPU)
487 
488 #define vcpu_clear_on_unsupported_cpu(vcpu)				\
489 	((vcpu)->arch.flags &= ~KVM_ARM64_ON_UNSUPPORTED_CPU)
490 
491 #define vcpu_gp_regs(v)		(&(v)->arch.ctxt.regs)
492 
493 /*
494  * Only use __vcpu_sys_reg/ctxt_sys_reg if you know you want the
495  * memory backed version of a register, and not the one most recently
496  * accessed by a running VCPU.  For example, for userspace access or
497  * for system registers that are never context switched, but only
498  * emulated.
499  */
500 #define __ctxt_sys_reg(c,r)	(&(c)->sys_regs[(r)])
501 
502 #define ctxt_sys_reg(c,r)	(*__ctxt_sys_reg(c,r))
503 
504 #define __vcpu_sys_reg(v,r)	(ctxt_sys_reg(&(v)->arch.ctxt, (r)))
505 
506 u64 vcpu_read_sys_reg(const struct kvm_vcpu *vcpu, int reg);
507 void vcpu_write_sys_reg(struct kvm_vcpu *vcpu, u64 val, int reg);
508 
509 static inline bool __vcpu_read_sys_reg_from_cpu(int reg, u64 *val)
510 {
511 	/*
512 	 * *** VHE ONLY ***
513 	 *
514 	 * System registers listed in the switch are not saved on every
515 	 * exit from the guest but are only saved on vcpu_put.
516 	 *
517 	 * Note that MPIDR_EL1 for the guest is set by KVM via VMPIDR_EL2 but
518 	 * should never be listed below, because the guest cannot modify its
519 	 * own MPIDR_EL1 and MPIDR_EL1 is accessed for VCPU A from VCPU B's
520 	 * thread when emulating cross-VCPU communication.
521 	 */
522 	if (!has_vhe())
523 		return false;
524 
525 	switch (reg) {
526 	case CSSELR_EL1:	*val = read_sysreg_s(SYS_CSSELR_EL1);	break;
527 	case SCTLR_EL1:		*val = read_sysreg_s(SYS_SCTLR_EL12);	break;
528 	case CPACR_EL1:		*val = read_sysreg_s(SYS_CPACR_EL12);	break;
529 	case TTBR0_EL1:		*val = read_sysreg_s(SYS_TTBR0_EL12);	break;
530 	case TTBR1_EL1:		*val = read_sysreg_s(SYS_TTBR1_EL12);	break;
531 	case TCR_EL1:		*val = read_sysreg_s(SYS_TCR_EL12);	break;
532 	case ESR_EL1:		*val = read_sysreg_s(SYS_ESR_EL12);	break;
533 	case AFSR0_EL1:		*val = read_sysreg_s(SYS_AFSR0_EL12);	break;
534 	case AFSR1_EL1:		*val = read_sysreg_s(SYS_AFSR1_EL12);	break;
535 	case FAR_EL1:		*val = read_sysreg_s(SYS_FAR_EL12);	break;
536 	case MAIR_EL1:		*val = read_sysreg_s(SYS_MAIR_EL12);	break;
537 	case VBAR_EL1:		*val = read_sysreg_s(SYS_VBAR_EL12);	break;
538 	case CONTEXTIDR_EL1:	*val = read_sysreg_s(SYS_CONTEXTIDR_EL12);break;
539 	case TPIDR_EL0:		*val = read_sysreg_s(SYS_TPIDR_EL0);	break;
540 	case TPIDRRO_EL0:	*val = read_sysreg_s(SYS_TPIDRRO_EL0);	break;
541 	case TPIDR_EL1:		*val = read_sysreg_s(SYS_TPIDR_EL1);	break;
542 	case AMAIR_EL1:		*val = read_sysreg_s(SYS_AMAIR_EL12);	break;
543 	case CNTKCTL_EL1:	*val = read_sysreg_s(SYS_CNTKCTL_EL12);	break;
544 	case ELR_EL1:		*val = read_sysreg_s(SYS_ELR_EL12);	break;
545 	case PAR_EL1:		*val = read_sysreg_par();		break;
546 	case DACR32_EL2:	*val = read_sysreg_s(SYS_DACR32_EL2);	break;
547 	case IFSR32_EL2:	*val = read_sysreg_s(SYS_IFSR32_EL2);	break;
548 	case DBGVCR32_EL2:	*val = read_sysreg_s(SYS_DBGVCR32_EL2);	break;
549 	default:		return false;
550 	}
551 
552 	return true;
553 }
554 
555 static inline bool __vcpu_write_sys_reg_to_cpu(u64 val, int reg)
556 {
557 	/*
558 	 * *** VHE ONLY ***
559 	 *
560 	 * System registers listed in the switch are not restored on every
561 	 * entry to the guest but are only restored on vcpu_load.
562 	 *
563 	 * Note that MPIDR_EL1 for the guest is set by KVM via VMPIDR_EL2 but
564 	 * should never be listed below, because the MPIDR should only be set
565 	 * once, before running the VCPU, and never changed later.
566 	 */
567 	if (!has_vhe())
568 		return false;
569 
570 	switch (reg) {
571 	case CSSELR_EL1:	write_sysreg_s(val, SYS_CSSELR_EL1);	break;
572 	case SCTLR_EL1:		write_sysreg_s(val, SYS_SCTLR_EL12);	break;
573 	case CPACR_EL1:		write_sysreg_s(val, SYS_CPACR_EL12);	break;
574 	case TTBR0_EL1:		write_sysreg_s(val, SYS_TTBR0_EL12);	break;
575 	case TTBR1_EL1:		write_sysreg_s(val, SYS_TTBR1_EL12);	break;
576 	case TCR_EL1:		write_sysreg_s(val, SYS_TCR_EL12);	break;
577 	case ESR_EL1:		write_sysreg_s(val, SYS_ESR_EL12);	break;
578 	case AFSR0_EL1:		write_sysreg_s(val, SYS_AFSR0_EL12);	break;
579 	case AFSR1_EL1:		write_sysreg_s(val, SYS_AFSR1_EL12);	break;
580 	case FAR_EL1:		write_sysreg_s(val, SYS_FAR_EL12);	break;
581 	case MAIR_EL1:		write_sysreg_s(val, SYS_MAIR_EL12);	break;
582 	case VBAR_EL1:		write_sysreg_s(val, SYS_VBAR_EL12);	break;
583 	case CONTEXTIDR_EL1:	write_sysreg_s(val, SYS_CONTEXTIDR_EL12);break;
584 	case TPIDR_EL0:		write_sysreg_s(val, SYS_TPIDR_EL0);	break;
585 	case TPIDRRO_EL0:	write_sysreg_s(val, SYS_TPIDRRO_EL0);	break;
586 	case TPIDR_EL1:		write_sysreg_s(val, SYS_TPIDR_EL1);	break;
587 	case AMAIR_EL1:		write_sysreg_s(val, SYS_AMAIR_EL12);	break;
588 	case CNTKCTL_EL1:	write_sysreg_s(val, SYS_CNTKCTL_EL12);	break;
589 	case ELR_EL1:		write_sysreg_s(val, SYS_ELR_EL12);	break;
590 	case PAR_EL1:		write_sysreg_s(val, SYS_PAR_EL1);	break;
591 	case DACR32_EL2:	write_sysreg_s(val, SYS_DACR32_EL2);	break;
592 	case IFSR32_EL2:	write_sysreg_s(val, SYS_IFSR32_EL2);	break;
593 	case DBGVCR32_EL2:	write_sysreg_s(val, SYS_DBGVCR32_EL2);	break;
594 	default:		return false;
595 	}
596 
597 	return true;
598 }
599 
600 struct kvm_vm_stat {
601 	struct kvm_vm_stat_generic generic;
602 };
603 
604 struct kvm_vcpu_stat {
605 	struct kvm_vcpu_stat_generic generic;
606 	u64 hvc_exit_stat;
607 	u64 wfe_exit_stat;
608 	u64 wfi_exit_stat;
609 	u64 mmio_exit_user;
610 	u64 mmio_exit_kernel;
611 	u64 signal_exits;
612 	u64 exits;
613 };
614 
615 void kvm_vcpu_preferred_target(struct kvm_vcpu_init *init);
616 unsigned long kvm_arm_num_regs(struct kvm_vcpu *vcpu);
617 int kvm_arm_copy_reg_indices(struct kvm_vcpu *vcpu, u64 __user *indices);
618 int kvm_arm_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
619 int kvm_arm_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
620 
621 unsigned long kvm_arm_num_sys_reg_descs(struct kvm_vcpu *vcpu);
622 int kvm_arm_copy_sys_reg_indices(struct kvm_vcpu *vcpu, u64 __user *uindices);
623 int kvm_arm_sys_reg_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *);
624 int kvm_arm_sys_reg_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *);
625 
626 int __kvm_arm_vcpu_get_events(struct kvm_vcpu *vcpu,
627 			      struct kvm_vcpu_events *events);
628 
629 int __kvm_arm_vcpu_set_events(struct kvm_vcpu *vcpu,
630 			      struct kvm_vcpu_events *events);
631 
632 #define KVM_ARCH_WANT_MMU_NOTIFIER
633 
634 void kvm_arm_halt_guest(struct kvm *kvm);
635 void kvm_arm_resume_guest(struct kvm *kvm);
636 
637 #define vcpu_has_run_once(vcpu)	!!rcu_access_pointer((vcpu)->pid)
638 
639 #ifndef __KVM_NVHE_HYPERVISOR__
640 #define kvm_call_hyp_nvhe(f, ...)						\
641 	({								\
642 		struct arm_smccc_res res;				\
643 									\
644 		arm_smccc_1_1_hvc(KVM_HOST_SMCCC_FUNC(f),		\
645 				  ##__VA_ARGS__, &res);			\
646 		WARN_ON(res.a0 != SMCCC_RET_SUCCESS);			\
647 									\
648 		res.a1;							\
649 	})
650 
651 /*
652  * The couple of isb() below are there to guarantee the same behaviour
653  * on VHE as on !VHE, where the eret to EL1 acts as a context
654  * synchronization event.
655  */
656 #define kvm_call_hyp(f, ...)						\
657 	do {								\
658 		if (has_vhe()) {					\
659 			f(__VA_ARGS__);					\
660 			isb();						\
661 		} else {						\
662 			kvm_call_hyp_nvhe(f, ##__VA_ARGS__);		\
663 		}							\
664 	} while(0)
665 
666 #define kvm_call_hyp_ret(f, ...)					\
667 	({								\
668 		typeof(f(__VA_ARGS__)) ret;				\
669 									\
670 		if (has_vhe()) {					\
671 			ret = f(__VA_ARGS__);				\
672 			isb();						\
673 		} else {						\
674 			ret = kvm_call_hyp_nvhe(f, ##__VA_ARGS__);	\
675 		}							\
676 									\
677 		ret;							\
678 	})
679 #else /* __KVM_NVHE_HYPERVISOR__ */
680 #define kvm_call_hyp(f, ...) f(__VA_ARGS__)
681 #define kvm_call_hyp_ret(f, ...) f(__VA_ARGS__)
682 #define kvm_call_hyp_nvhe(f, ...) f(__VA_ARGS__)
683 #endif /* __KVM_NVHE_HYPERVISOR__ */
684 
685 void force_vm_exit(const cpumask_t *mask);
686 
687 int handle_exit(struct kvm_vcpu *vcpu, int exception_index);
688 void handle_exit_early(struct kvm_vcpu *vcpu, int exception_index);
689 
690 int kvm_handle_cp14_load_store(struct kvm_vcpu *vcpu);
691 int kvm_handle_cp14_32(struct kvm_vcpu *vcpu);
692 int kvm_handle_cp14_64(struct kvm_vcpu *vcpu);
693 int kvm_handle_cp15_32(struct kvm_vcpu *vcpu);
694 int kvm_handle_cp15_64(struct kvm_vcpu *vcpu);
695 int kvm_handle_sys_reg(struct kvm_vcpu *vcpu);
696 int kvm_handle_cp10_id(struct kvm_vcpu *vcpu);
697 
698 void kvm_reset_sys_regs(struct kvm_vcpu *vcpu);
699 
700 int kvm_sys_reg_table_init(void);
701 
702 /* MMIO helpers */
703 void kvm_mmio_write_buf(void *buf, unsigned int len, unsigned long data);
704 unsigned long kvm_mmio_read_buf(const void *buf, unsigned int len);
705 
706 int kvm_handle_mmio_return(struct kvm_vcpu *vcpu);
707 int io_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa);
708 
709 /*
710  * Returns true if a Performance Monitoring Interrupt (PMI), a.k.a. perf event,
711  * arrived in guest context.  For arm64, any event that arrives while a vCPU is
712  * loaded is considered to be "in guest".
713  */
714 static inline bool kvm_arch_pmi_in_guest(struct kvm_vcpu *vcpu)
715 {
716 	return IS_ENABLED(CONFIG_GUEST_PERF_EVENTS) && !!vcpu;
717 }
718 
719 long kvm_hypercall_pv_features(struct kvm_vcpu *vcpu);
720 gpa_t kvm_init_stolen_time(struct kvm_vcpu *vcpu);
721 void kvm_update_stolen_time(struct kvm_vcpu *vcpu);
722 
723 bool kvm_arm_pvtime_supported(void);
724 int kvm_arm_pvtime_set_attr(struct kvm_vcpu *vcpu,
725 			    struct kvm_device_attr *attr);
726 int kvm_arm_pvtime_get_attr(struct kvm_vcpu *vcpu,
727 			    struct kvm_device_attr *attr);
728 int kvm_arm_pvtime_has_attr(struct kvm_vcpu *vcpu,
729 			    struct kvm_device_attr *attr);
730 
731 extern unsigned int kvm_arm_vmid_bits;
732 int kvm_arm_vmid_alloc_init(void);
733 void kvm_arm_vmid_alloc_free(void);
734 void kvm_arm_vmid_update(struct kvm_vmid *kvm_vmid);
735 void kvm_arm_vmid_clear_active(void);
736 
737 static inline void kvm_arm_pvtime_vcpu_init(struct kvm_vcpu_arch *vcpu_arch)
738 {
739 	vcpu_arch->steal.base = GPA_INVALID;
740 }
741 
742 static inline bool kvm_arm_is_pvtime_enabled(struct kvm_vcpu_arch *vcpu_arch)
743 {
744 	return (vcpu_arch->steal.base != GPA_INVALID);
745 }
746 
747 void kvm_set_sei_esr(struct kvm_vcpu *vcpu, u64 syndrome);
748 
749 struct kvm_vcpu *kvm_mpidr_to_vcpu(struct kvm *kvm, unsigned long mpidr);
750 
751 DECLARE_KVM_HYP_PER_CPU(struct kvm_host_data, kvm_host_data);
752 
753 static inline void kvm_init_host_cpu_context(struct kvm_cpu_context *cpu_ctxt)
754 {
755 	/* The host's MPIDR is immutable, so let's set it up at boot time */
756 	ctxt_sys_reg(cpu_ctxt, MPIDR_EL1) = read_cpuid_mpidr();
757 }
758 
759 static inline bool kvm_system_needs_idmapped_vectors(void)
760 {
761 	return cpus_have_const_cap(ARM64_SPECTRE_V3A);
762 }
763 
764 void kvm_arm_vcpu_ptrauth_trap(struct kvm_vcpu *vcpu);
765 
766 static inline void kvm_arch_hardware_unsetup(void) {}
767 static inline void kvm_arch_sync_events(struct kvm *kvm) {}
768 static inline void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu) {}
769 
770 void kvm_arm_init_debug(void);
771 void kvm_arm_vcpu_init_debug(struct kvm_vcpu *vcpu);
772 void kvm_arm_setup_debug(struct kvm_vcpu *vcpu);
773 void kvm_arm_clear_debug(struct kvm_vcpu *vcpu);
774 void kvm_arm_reset_debug_ptr(struct kvm_vcpu *vcpu);
775 
776 #define kvm_vcpu_os_lock_enabled(vcpu)		\
777 	(!!(__vcpu_sys_reg(vcpu, OSLSR_EL1) & SYS_OSLSR_OSLK))
778 
779 int kvm_arm_vcpu_arch_set_attr(struct kvm_vcpu *vcpu,
780 			       struct kvm_device_attr *attr);
781 int kvm_arm_vcpu_arch_get_attr(struct kvm_vcpu *vcpu,
782 			       struct kvm_device_attr *attr);
783 int kvm_arm_vcpu_arch_has_attr(struct kvm_vcpu *vcpu,
784 			       struct kvm_device_attr *attr);
785 
786 long kvm_vm_ioctl_mte_copy_tags(struct kvm *kvm,
787 				struct kvm_arm_copy_mte_tags *copy_tags);
788 
789 /* Guest/host FPSIMD coordination helpers */
790 int kvm_arch_vcpu_run_map_fp(struct kvm_vcpu *vcpu);
791 void kvm_arch_vcpu_load_fp(struct kvm_vcpu *vcpu);
792 void kvm_arch_vcpu_ctxflush_fp(struct kvm_vcpu *vcpu);
793 void kvm_arch_vcpu_ctxsync_fp(struct kvm_vcpu *vcpu);
794 void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu);
795 void kvm_vcpu_unshare_task_fp(struct kvm_vcpu *vcpu);
796 
797 static inline bool kvm_pmu_counter_deferred(struct perf_event_attr *attr)
798 {
799 	return (!has_vhe() && attr->exclude_host);
800 }
801 
802 /* Flags for host debug state */
803 void kvm_arch_vcpu_load_debug_state_flags(struct kvm_vcpu *vcpu);
804 void kvm_arch_vcpu_put_debug_state_flags(struct kvm_vcpu *vcpu);
805 
806 #ifdef CONFIG_KVM
807 void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr);
808 void kvm_clr_pmu_events(u32 clr);
809 #else
810 static inline void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr) {}
811 static inline void kvm_clr_pmu_events(u32 clr) {}
812 #endif
813 
814 void kvm_vcpu_load_sysregs_vhe(struct kvm_vcpu *vcpu);
815 void kvm_vcpu_put_sysregs_vhe(struct kvm_vcpu *vcpu);
816 
817 int kvm_set_ipa_limit(void);
818 
819 #define __KVM_HAVE_ARCH_VM_ALLOC
820 struct kvm *kvm_arch_alloc_vm(void);
821 
822 int kvm_arm_setup_stage2(struct kvm *kvm, unsigned long type);
823 
824 static inline bool kvm_vm_is_protected(struct kvm *kvm)
825 {
826 	return false;
827 }
828 
829 void kvm_init_protected_traps(struct kvm_vcpu *vcpu);
830 
831 int kvm_arm_vcpu_finalize(struct kvm_vcpu *vcpu, int feature);
832 bool kvm_arm_vcpu_is_finalized(struct kvm_vcpu *vcpu);
833 
834 #define kvm_arm_vcpu_sve_finalized(vcpu) \
835 	((vcpu)->arch.flags & KVM_ARM64_VCPU_SVE_FINALIZED)
836 
837 #define kvm_has_mte(kvm)					\
838 	(system_supports_mte() &&				\
839 	 test_bit(KVM_ARCH_FLAG_MTE_ENABLED, &(kvm)->arch.flags))
840 
841 int kvm_trng_call(struct kvm_vcpu *vcpu);
842 #ifdef CONFIG_KVM
843 extern phys_addr_t hyp_mem_base;
844 extern phys_addr_t hyp_mem_size;
845 void __init kvm_hyp_reserve(void);
846 #else
847 static inline void kvm_hyp_reserve(void) { }
848 #endif
849 
850 void kvm_arm_vcpu_power_off(struct kvm_vcpu *vcpu);
851 bool kvm_arm_vcpu_stopped(struct kvm_vcpu *vcpu);
852 
853 #endif /* __ARM64_KVM_HOST_H__ */
854