xref: /openbmc/linux/arch/arm64/include/asm/kvm_host.h (revision cfb1a98d)
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 #include <asm/thread_info.h>
30 
31 #define __KVM_HAVE_ARCH_INTC_INITIALIZED
32 
33 #define KVM_HALT_POLL_NS_DEFAULT 500000
34 
35 #include <kvm/arm_vgic.h>
36 #include <kvm/arm_arch_timer.h>
37 #include <kvm/arm_pmu.h>
38 
39 #define KVM_MAX_VCPUS VGIC_V3_MAX_CPUS
40 
41 #define KVM_VCPU_MAX_FEATURES 7
42 
43 #define KVM_REQ_SLEEP \
44 	KVM_ARCH_REQ_FLAGS(0, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
45 #define KVM_REQ_IRQ_PENDING	KVM_ARCH_REQ(1)
46 #define KVM_REQ_VCPU_RESET	KVM_ARCH_REQ(2)
47 #define KVM_REQ_RECORD_STEAL	KVM_ARCH_REQ(3)
48 #define KVM_REQ_RELOAD_GICv4	KVM_ARCH_REQ(4)
49 
50 #define KVM_DIRTY_LOG_MANUAL_CAPS   (KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE | \
51 				     KVM_DIRTY_LOG_INITIALLY_SET)
52 
53 /*
54  * Mode of operation configurable with kvm-arm.mode early param.
55  * See Documentation/admin-guide/kernel-parameters.txt for more information.
56  */
57 enum kvm_mode {
58 	KVM_MODE_DEFAULT,
59 	KVM_MODE_PROTECTED,
60 };
61 enum kvm_mode kvm_get_mode(void);
62 
63 DECLARE_STATIC_KEY_FALSE(userspace_irqchip_in_use);
64 
65 extern unsigned int kvm_sve_max_vl;
66 int kvm_arm_init_sve(void);
67 
68 int __attribute_const__ kvm_target_cpu(void);
69 int kvm_reset_vcpu(struct kvm_vcpu *vcpu);
70 void kvm_arm_vcpu_destroy(struct kvm_vcpu *vcpu);
71 
72 struct kvm_vmid {
73 	/* The VMID generation used for the virt. memory system */
74 	u64    vmid_gen;
75 	u32    vmid;
76 };
77 
78 struct kvm_s2_mmu {
79 	struct kvm_vmid vmid;
80 
81 	/*
82 	 * stage2 entry level table
83 	 *
84 	 * Two kvm_s2_mmu structures in the same VM can point to the same
85 	 * pgd here.  This happens when running a guest using a
86 	 * translation regime that isn't affected by its own stage-2
87 	 * translation, such as a non-VHE hypervisor running at vEL2, or
88 	 * for vEL1/EL0 with vHCR_EL2.VM == 0.  In that case, we use the
89 	 * canonical stage-2 page tables.
90 	 */
91 	phys_addr_t	pgd_phys;
92 	struct kvm_pgtable *pgt;
93 
94 	/* The last vcpu id that ran on each physical CPU */
95 	int __percpu *last_vcpu_ran;
96 
97 	struct kvm_arch *arch;
98 };
99 
100 struct kvm_arch_memory_slot {
101 };
102 
103 struct kvm_arch {
104 	struct kvm_s2_mmu mmu;
105 
106 	/* VTCR_EL2 value for this VM */
107 	u64    vtcr;
108 
109 	/* The maximum number of vCPUs depends on the used GIC model */
110 	int max_vcpus;
111 
112 	/* Interrupt controller */
113 	struct vgic_dist	vgic;
114 
115 	/* Mandated version of PSCI */
116 	u32 psci_version;
117 
118 	/*
119 	 * If we encounter a data abort without valid instruction syndrome
120 	 * information, report this to user space.  User space can (and
121 	 * should) opt in to this feature if KVM_CAP_ARM_NISV_TO_USER is
122 	 * supported.
123 	 */
124 	bool return_nisv_io_abort_to_user;
125 
126 	/*
127 	 * VM-wide PMU filter, implemented as a bitmap and big enough for
128 	 * up to 2^10 events (ARMv8.0) or 2^16 events (ARMv8.1+).
129 	 */
130 	unsigned long *pmu_filter;
131 	unsigned int pmuver;
132 
133 	u8 pfr0_csv2;
134 	u8 pfr0_csv3;
135 };
136 
137 struct kvm_vcpu_fault_info {
138 	u32 esr_el2;		/* Hyp Syndrom Register */
139 	u64 far_el2;		/* Hyp Fault Address Register */
140 	u64 hpfar_el2;		/* Hyp IPA Fault Address Register */
141 	u64 disr_el1;		/* Deferred [SError] Status Register */
142 };
143 
144 enum vcpu_sysreg {
145 	__INVALID_SYSREG__,   /* 0 is reserved as an invalid value */
146 	MPIDR_EL1,	/* MultiProcessor Affinity Register */
147 	CSSELR_EL1,	/* Cache Size Selection Register */
148 	SCTLR_EL1,	/* System Control Register */
149 	ACTLR_EL1,	/* Auxiliary Control Register */
150 	CPACR_EL1,	/* Coprocessor Access Control */
151 	ZCR_EL1,	/* SVE Control */
152 	TTBR0_EL1,	/* Translation Table Base Register 0 */
153 	TTBR1_EL1,	/* Translation Table Base Register 1 */
154 	TCR_EL1,	/* Translation Control Register */
155 	ESR_EL1,	/* Exception Syndrome Register */
156 	AFSR0_EL1,	/* Auxiliary Fault Status Register 0 */
157 	AFSR1_EL1,	/* Auxiliary Fault Status Register 1 */
158 	FAR_EL1,	/* Fault Address Register */
159 	MAIR_EL1,	/* Memory Attribute Indirection Register */
160 	VBAR_EL1,	/* Vector Base Address Register */
161 	CONTEXTIDR_EL1,	/* Context ID Register */
162 	TPIDR_EL0,	/* Thread ID, User R/W */
163 	TPIDRRO_EL0,	/* Thread ID, User R/O */
164 	TPIDR_EL1,	/* Thread ID, Privileged */
165 	AMAIR_EL1,	/* Aux Memory Attribute Indirection Register */
166 	CNTKCTL_EL1,	/* Timer Control Register (EL1) */
167 	PAR_EL1,	/* Physical Address Register */
168 	MDSCR_EL1,	/* Monitor Debug System Control Register */
169 	MDCCINT_EL1,	/* Monitor Debug Comms Channel Interrupt Enable Reg */
170 	DISR_EL1,	/* Deferred Interrupt Status Register */
171 
172 	/* Performance Monitors Registers */
173 	PMCR_EL0,	/* Control Register */
174 	PMSELR_EL0,	/* Event Counter Selection Register */
175 	PMEVCNTR0_EL0,	/* Event Counter Register (0-30) */
176 	PMEVCNTR30_EL0 = PMEVCNTR0_EL0 + 30,
177 	PMCCNTR_EL0,	/* Cycle Counter Register */
178 	PMEVTYPER0_EL0,	/* Event Type Register (0-30) */
179 	PMEVTYPER30_EL0 = PMEVTYPER0_EL0 + 30,
180 	PMCCFILTR_EL0,	/* Cycle Count Filter Register */
181 	PMCNTENSET_EL0,	/* Count Enable Set Register */
182 	PMINTENSET_EL1,	/* Interrupt Enable Set Register */
183 	PMOVSSET_EL0,	/* Overflow Flag Status Set Register */
184 	PMSWINC_EL0,	/* Software Increment Register */
185 	PMUSERENR_EL0,	/* User Enable Register */
186 
187 	/* Pointer Authentication Registers in a strict increasing order. */
188 	APIAKEYLO_EL1,
189 	APIAKEYHI_EL1,
190 	APIBKEYLO_EL1,
191 	APIBKEYHI_EL1,
192 	APDAKEYLO_EL1,
193 	APDAKEYHI_EL1,
194 	APDBKEYLO_EL1,
195 	APDBKEYHI_EL1,
196 	APGAKEYLO_EL1,
197 	APGAKEYHI_EL1,
198 
199 	ELR_EL1,
200 	SP_EL1,
201 	SPSR_EL1,
202 
203 	CNTVOFF_EL2,
204 	CNTV_CVAL_EL0,
205 	CNTV_CTL_EL0,
206 	CNTP_CVAL_EL0,
207 	CNTP_CTL_EL0,
208 
209 	/* 32bit specific registers. Keep them at the end of the range */
210 	DACR32_EL2,	/* Domain Access Control Register */
211 	IFSR32_EL2,	/* Instruction Fault Status Register */
212 	FPEXC32_EL2,	/* Floating-Point Exception Control Register */
213 	DBGVCR32_EL2,	/* Debug Vector Catch Register */
214 
215 	NR_SYS_REGS	/* Nothing after this line! */
216 };
217 
218 struct kvm_cpu_context {
219 	struct user_pt_regs regs;	/* sp = sp_el0 */
220 
221 	u64	spsr_abt;
222 	u64	spsr_und;
223 	u64	spsr_irq;
224 	u64	spsr_fiq;
225 
226 	struct user_fpsimd_state fp_regs;
227 
228 	u64 sys_regs[NR_SYS_REGS];
229 
230 	struct kvm_vcpu *__hyp_running_vcpu;
231 };
232 
233 struct kvm_pmu_events {
234 	u32 events_host;
235 	u32 events_guest;
236 };
237 
238 struct kvm_host_data {
239 	struct kvm_cpu_context host_ctxt;
240 	struct kvm_pmu_events pmu_events;
241 };
242 
243 struct kvm_host_psci_config {
244 	/* PSCI version used by host. */
245 	u32 version;
246 
247 	/* Function IDs used by host if version is v0.1. */
248 	struct psci_0_1_function_ids function_ids_0_1;
249 
250 	bool psci_0_1_cpu_suspend_implemented;
251 	bool psci_0_1_cpu_on_implemented;
252 	bool psci_0_1_cpu_off_implemented;
253 	bool psci_0_1_migrate_implemented;
254 };
255 
256 extern struct kvm_host_psci_config kvm_nvhe_sym(kvm_host_psci_config);
257 #define kvm_host_psci_config CHOOSE_NVHE_SYM(kvm_host_psci_config)
258 
259 extern s64 kvm_nvhe_sym(hyp_physvirt_offset);
260 #define hyp_physvirt_offset CHOOSE_NVHE_SYM(hyp_physvirt_offset)
261 
262 extern u64 kvm_nvhe_sym(hyp_cpu_logical_map)[NR_CPUS];
263 #define hyp_cpu_logical_map CHOOSE_NVHE_SYM(hyp_cpu_logical_map)
264 
265 struct vcpu_reset_state {
266 	unsigned long	pc;
267 	unsigned long	r0;
268 	bool		be;
269 	bool		reset;
270 };
271 
272 struct kvm_vcpu_arch {
273 	struct kvm_cpu_context ctxt;
274 	void *sve_state;
275 	unsigned int sve_max_vl;
276 
277 	/* Stage 2 paging state used by the hardware on next switch */
278 	struct kvm_s2_mmu *hw_mmu;
279 
280 	/* HYP configuration */
281 	u64 hcr_el2;
282 	u32 mdcr_el2;
283 
284 	/* Exception Information */
285 	struct kvm_vcpu_fault_info fault;
286 
287 	/* State of various workarounds, see kvm_asm.h for bit assignment */
288 	u64 workaround_flags;
289 
290 	/* Miscellaneous vcpu state flags */
291 	u64 flags;
292 
293 	/*
294 	 * We maintain more than a single set of debug registers to support
295 	 * debugging the guest from the host and to maintain separate host and
296 	 * guest state during world switches. vcpu_debug_state are the debug
297 	 * registers of the vcpu as the guest sees them.  host_debug_state are
298 	 * the host registers which are saved and restored during
299 	 * world switches. external_debug_state contains the debug
300 	 * values we want to debug the guest. This is set via the
301 	 * KVM_SET_GUEST_DEBUG ioctl.
302 	 *
303 	 * debug_ptr points to the set of debug registers that should be loaded
304 	 * onto the hardware when running the guest.
305 	 */
306 	struct kvm_guest_debug_arch *debug_ptr;
307 	struct kvm_guest_debug_arch vcpu_debug_state;
308 	struct kvm_guest_debug_arch external_debug_state;
309 
310 	struct thread_info *host_thread_info;	/* hyp VA */
311 	struct user_fpsimd_state *host_fpsimd_state;	/* hyp VA */
312 
313 	struct {
314 		/* {Break,watch}point registers */
315 		struct kvm_guest_debug_arch regs;
316 		/* Statistical profiling extension */
317 		u64 pmscr_el1;
318 	} host_debug_state;
319 
320 	/* VGIC state */
321 	struct vgic_cpu vgic_cpu;
322 	struct arch_timer_cpu timer_cpu;
323 	struct kvm_pmu pmu;
324 
325 	/*
326 	 * Anything that is not used directly from assembly code goes
327 	 * here.
328 	 */
329 
330 	/*
331 	 * Guest registers we preserve during guest debugging.
332 	 *
333 	 * These shadow registers are updated by the kvm_handle_sys_reg
334 	 * trap handler if the guest accesses or updates them while we
335 	 * are using guest debug.
336 	 */
337 	struct {
338 		u32	mdscr_el1;
339 	} guest_debug_preserved;
340 
341 	/* vcpu power-off state */
342 	bool power_off;
343 
344 	/* Don't run the guest (internal implementation need) */
345 	bool pause;
346 
347 	/* Cache some mmu pages needed inside spinlock regions */
348 	struct kvm_mmu_memory_cache mmu_page_cache;
349 
350 	/* Target CPU and feature flags */
351 	int target;
352 	DECLARE_BITMAP(features, KVM_VCPU_MAX_FEATURES);
353 
354 	/* Detect first run of a vcpu */
355 	bool has_run_once;
356 
357 	/* Virtual SError ESR to restore when HCR_EL2.VSE is set */
358 	u64 vsesr_el2;
359 
360 	/* Additional reset state */
361 	struct vcpu_reset_state	reset_state;
362 
363 	/* True when deferrable sysregs are loaded on the physical CPU,
364 	 * see kvm_vcpu_load_sysregs_vhe and kvm_vcpu_put_sysregs_vhe. */
365 	bool sysregs_loaded_on_cpu;
366 
367 	/* Guest PV state */
368 	struct {
369 		u64 last_steal;
370 		gpa_t base;
371 	} steal;
372 };
373 
374 /* Pointer to the vcpu's SVE FFR for sve_{save,load}_state() */
375 #define vcpu_sve_pffr(vcpu) (kern_hyp_va((vcpu)->arch.sve_state) +	\
376 			     sve_ffr_offset((vcpu)->arch.sve_max_vl))
377 
378 #define vcpu_sve_max_vq(vcpu)	sve_vq_from_vl((vcpu)->arch.sve_max_vl)
379 
380 #define vcpu_sve_state_size(vcpu) ({					\
381 	size_t __size_ret;						\
382 	unsigned int __vcpu_vq;						\
383 									\
384 	if (WARN_ON(!sve_vl_valid((vcpu)->arch.sve_max_vl))) {		\
385 		__size_ret = 0;						\
386 	} else {							\
387 		__vcpu_vq = vcpu_sve_max_vq(vcpu);			\
388 		__size_ret = SVE_SIG_REGS_SIZE(__vcpu_vq);		\
389 	}								\
390 									\
391 	__size_ret;							\
392 })
393 
394 /* vcpu_arch flags field values: */
395 #define KVM_ARM64_DEBUG_DIRTY		(1 << 0)
396 #define KVM_ARM64_FP_ENABLED		(1 << 1) /* guest FP regs loaded */
397 #define KVM_ARM64_FP_HOST		(1 << 2) /* host FP regs loaded */
398 #define KVM_ARM64_HOST_SVE_IN_USE	(1 << 3) /* backup for host TIF_SVE */
399 #define KVM_ARM64_HOST_SVE_ENABLED	(1 << 4) /* SVE enabled for EL0 */
400 #define KVM_ARM64_GUEST_HAS_SVE		(1 << 5) /* SVE exposed to guest */
401 #define KVM_ARM64_VCPU_SVE_FINALIZED	(1 << 6) /* SVE config completed */
402 #define KVM_ARM64_GUEST_HAS_PTRAUTH	(1 << 7) /* PTRAUTH exposed to guest */
403 #define KVM_ARM64_PENDING_EXCEPTION	(1 << 8) /* Exception pending */
404 #define KVM_ARM64_EXCEPT_MASK		(7 << 9) /* Target EL/MODE */
405 
406 /*
407  * When KVM_ARM64_PENDING_EXCEPTION is set, KVM_ARM64_EXCEPT_MASK can
408  * take the following values:
409  *
410  * For AArch32 EL1:
411  */
412 #define KVM_ARM64_EXCEPT_AA32_UND	(0 << 9)
413 #define KVM_ARM64_EXCEPT_AA32_IABT	(1 << 9)
414 #define KVM_ARM64_EXCEPT_AA32_DABT	(2 << 9)
415 /* For AArch64: */
416 #define KVM_ARM64_EXCEPT_AA64_ELx_SYNC	(0 << 9)
417 #define KVM_ARM64_EXCEPT_AA64_ELx_IRQ	(1 << 9)
418 #define KVM_ARM64_EXCEPT_AA64_ELx_FIQ	(2 << 9)
419 #define KVM_ARM64_EXCEPT_AA64_ELx_SERR	(3 << 9)
420 #define KVM_ARM64_EXCEPT_AA64_EL1	(0 << 11)
421 #define KVM_ARM64_EXCEPT_AA64_EL2	(1 << 11)
422 
423 /*
424  * Overlaps with KVM_ARM64_EXCEPT_MASK on purpose so that it can't be
425  * set together with an exception...
426  */
427 #define KVM_ARM64_INCREMENT_PC		(1 << 9) /* Increment PC */
428 
429 #define vcpu_has_sve(vcpu) (system_supports_sve() &&			\
430 			    ((vcpu)->arch.flags & KVM_ARM64_GUEST_HAS_SVE))
431 
432 #ifdef CONFIG_ARM64_PTR_AUTH
433 #define vcpu_has_ptrauth(vcpu)						\
434 	((cpus_have_final_cap(ARM64_HAS_ADDRESS_AUTH) ||		\
435 	  cpus_have_final_cap(ARM64_HAS_GENERIC_AUTH)) &&		\
436 	 (vcpu)->arch.flags & KVM_ARM64_GUEST_HAS_PTRAUTH)
437 #else
438 #define vcpu_has_ptrauth(vcpu)		false
439 #endif
440 
441 #define vcpu_gp_regs(v)		(&(v)->arch.ctxt.regs)
442 
443 /*
444  * Only use __vcpu_sys_reg/ctxt_sys_reg if you know you want the
445  * memory backed version of a register, and not the one most recently
446  * accessed by a running VCPU.  For example, for userspace access or
447  * for system registers that are never context switched, but only
448  * emulated.
449  */
450 #define __ctxt_sys_reg(c,r)	(&(c)->sys_regs[(r)])
451 
452 #define ctxt_sys_reg(c,r)	(*__ctxt_sys_reg(c,r))
453 
454 #define __vcpu_sys_reg(v,r)	(ctxt_sys_reg(&(v)->arch.ctxt, (r)))
455 
456 u64 vcpu_read_sys_reg(const struct kvm_vcpu *vcpu, int reg);
457 void vcpu_write_sys_reg(struct kvm_vcpu *vcpu, u64 val, int reg);
458 
459 static inline bool __vcpu_read_sys_reg_from_cpu(int reg, u64 *val)
460 {
461 	/*
462 	 * *** VHE ONLY ***
463 	 *
464 	 * System registers listed in the switch are not saved on every
465 	 * exit from the guest but are only saved on vcpu_put.
466 	 *
467 	 * Note that MPIDR_EL1 for the guest is set by KVM via VMPIDR_EL2 but
468 	 * should never be listed below, because the guest cannot modify its
469 	 * own MPIDR_EL1 and MPIDR_EL1 is accessed for VCPU A from VCPU B's
470 	 * thread when emulating cross-VCPU communication.
471 	 */
472 	if (!has_vhe())
473 		return false;
474 
475 	switch (reg) {
476 	case CSSELR_EL1:	*val = read_sysreg_s(SYS_CSSELR_EL1);	break;
477 	case SCTLR_EL1:		*val = read_sysreg_s(SYS_SCTLR_EL12);	break;
478 	case CPACR_EL1:		*val = read_sysreg_s(SYS_CPACR_EL12);	break;
479 	case TTBR0_EL1:		*val = read_sysreg_s(SYS_TTBR0_EL12);	break;
480 	case TTBR1_EL1:		*val = read_sysreg_s(SYS_TTBR1_EL12);	break;
481 	case TCR_EL1:		*val = read_sysreg_s(SYS_TCR_EL12);	break;
482 	case ESR_EL1:		*val = read_sysreg_s(SYS_ESR_EL12);	break;
483 	case AFSR0_EL1:		*val = read_sysreg_s(SYS_AFSR0_EL12);	break;
484 	case AFSR1_EL1:		*val = read_sysreg_s(SYS_AFSR1_EL12);	break;
485 	case FAR_EL1:		*val = read_sysreg_s(SYS_FAR_EL12);	break;
486 	case MAIR_EL1:		*val = read_sysreg_s(SYS_MAIR_EL12);	break;
487 	case VBAR_EL1:		*val = read_sysreg_s(SYS_VBAR_EL12);	break;
488 	case CONTEXTIDR_EL1:	*val = read_sysreg_s(SYS_CONTEXTIDR_EL12);break;
489 	case TPIDR_EL0:		*val = read_sysreg_s(SYS_TPIDR_EL0);	break;
490 	case TPIDRRO_EL0:	*val = read_sysreg_s(SYS_TPIDRRO_EL0);	break;
491 	case TPIDR_EL1:		*val = read_sysreg_s(SYS_TPIDR_EL1);	break;
492 	case AMAIR_EL1:		*val = read_sysreg_s(SYS_AMAIR_EL12);	break;
493 	case CNTKCTL_EL1:	*val = read_sysreg_s(SYS_CNTKCTL_EL12);	break;
494 	case ELR_EL1:		*val = read_sysreg_s(SYS_ELR_EL12);	break;
495 	case PAR_EL1:		*val = read_sysreg_par();		break;
496 	case DACR32_EL2:	*val = read_sysreg_s(SYS_DACR32_EL2);	break;
497 	case IFSR32_EL2:	*val = read_sysreg_s(SYS_IFSR32_EL2);	break;
498 	case DBGVCR32_EL2:	*val = read_sysreg_s(SYS_DBGVCR32_EL2);	break;
499 	default:		return false;
500 	}
501 
502 	return true;
503 }
504 
505 static inline bool __vcpu_write_sys_reg_to_cpu(u64 val, int reg)
506 {
507 	/*
508 	 * *** VHE ONLY ***
509 	 *
510 	 * System registers listed in the switch are not restored on every
511 	 * entry to the guest but are only restored on vcpu_load.
512 	 *
513 	 * Note that MPIDR_EL1 for the guest is set by KVM via VMPIDR_EL2 but
514 	 * should never be listed below, because the MPIDR should only be set
515 	 * once, before running the VCPU, and never changed later.
516 	 */
517 	if (!has_vhe())
518 		return false;
519 
520 	switch (reg) {
521 	case CSSELR_EL1:	write_sysreg_s(val, SYS_CSSELR_EL1);	break;
522 	case SCTLR_EL1:		write_sysreg_s(val, SYS_SCTLR_EL12);	break;
523 	case CPACR_EL1:		write_sysreg_s(val, SYS_CPACR_EL12);	break;
524 	case TTBR0_EL1:		write_sysreg_s(val, SYS_TTBR0_EL12);	break;
525 	case TTBR1_EL1:		write_sysreg_s(val, SYS_TTBR1_EL12);	break;
526 	case TCR_EL1:		write_sysreg_s(val, SYS_TCR_EL12);	break;
527 	case ESR_EL1:		write_sysreg_s(val, SYS_ESR_EL12);	break;
528 	case AFSR0_EL1:		write_sysreg_s(val, SYS_AFSR0_EL12);	break;
529 	case AFSR1_EL1:		write_sysreg_s(val, SYS_AFSR1_EL12);	break;
530 	case FAR_EL1:		write_sysreg_s(val, SYS_FAR_EL12);	break;
531 	case MAIR_EL1:		write_sysreg_s(val, SYS_MAIR_EL12);	break;
532 	case VBAR_EL1:		write_sysreg_s(val, SYS_VBAR_EL12);	break;
533 	case CONTEXTIDR_EL1:	write_sysreg_s(val, SYS_CONTEXTIDR_EL12);break;
534 	case TPIDR_EL0:		write_sysreg_s(val, SYS_TPIDR_EL0);	break;
535 	case TPIDRRO_EL0:	write_sysreg_s(val, SYS_TPIDRRO_EL0);	break;
536 	case TPIDR_EL1:		write_sysreg_s(val, SYS_TPIDR_EL1);	break;
537 	case AMAIR_EL1:		write_sysreg_s(val, SYS_AMAIR_EL12);	break;
538 	case CNTKCTL_EL1:	write_sysreg_s(val, SYS_CNTKCTL_EL12);	break;
539 	case ELR_EL1:		write_sysreg_s(val, SYS_ELR_EL12);	break;
540 	case PAR_EL1:		write_sysreg_s(val, SYS_PAR_EL1);	break;
541 	case DACR32_EL2:	write_sysreg_s(val, SYS_DACR32_EL2);	break;
542 	case IFSR32_EL2:	write_sysreg_s(val, SYS_IFSR32_EL2);	break;
543 	case DBGVCR32_EL2:	write_sysreg_s(val, SYS_DBGVCR32_EL2);	break;
544 	default:		return false;
545 	}
546 
547 	return true;
548 }
549 
550 struct kvm_vm_stat {
551 	ulong remote_tlb_flush;
552 };
553 
554 struct kvm_vcpu_stat {
555 	u64 halt_successful_poll;
556 	u64 halt_attempted_poll;
557 	u64 halt_poll_success_ns;
558 	u64 halt_poll_fail_ns;
559 	u64 halt_poll_invalid;
560 	u64 halt_wakeup;
561 	u64 hvc_exit_stat;
562 	u64 wfe_exit_stat;
563 	u64 wfi_exit_stat;
564 	u64 mmio_exit_user;
565 	u64 mmio_exit_kernel;
566 	u64 exits;
567 };
568 
569 int kvm_vcpu_preferred_target(struct kvm_vcpu_init *init);
570 unsigned long kvm_arm_num_regs(struct kvm_vcpu *vcpu);
571 int kvm_arm_copy_reg_indices(struct kvm_vcpu *vcpu, u64 __user *indices);
572 int kvm_arm_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
573 int kvm_arm_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
574 
575 unsigned long kvm_arm_num_sys_reg_descs(struct kvm_vcpu *vcpu);
576 int kvm_arm_copy_sys_reg_indices(struct kvm_vcpu *vcpu, u64 __user *uindices);
577 int kvm_arm_sys_reg_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *);
578 int kvm_arm_sys_reg_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *);
579 
580 int __kvm_arm_vcpu_get_events(struct kvm_vcpu *vcpu,
581 			      struct kvm_vcpu_events *events);
582 
583 int __kvm_arm_vcpu_set_events(struct kvm_vcpu *vcpu,
584 			      struct kvm_vcpu_events *events);
585 
586 #define KVM_ARCH_WANT_MMU_NOTIFIER
587 int kvm_unmap_hva_range(struct kvm *kvm,
588 			unsigned long start, unsigned long end, unsigned flags);
589 int kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
590 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
591 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
592 
593 void kvm_arm_halt_guest(struct kvm *kvm);
594 void kvm_arm_resume_guest(struct kvm *kvm);
595 
596 #ifndef __KVM_NVHE_HYPERVISOR__
597 #define kvm_call_hyp_nvhe(f, ...)						\
598 	({								\
599 		struct arm_smccc_res res;				\
600 									\
601 		arm_smccc_1_1_hvc(KVM_HOST_SMCCC_FUNC(f),		\
602 				  ##__VA_ARGS__, &res);			\
603 		WARN_ON(res.a0 != SMCCC_RET_SUCCESS);			\
604 									\
605 		res.a1;							\
606 	})
607 
608 /*
609  * The couple of isb() below are there to guarantee the same behaviour
610  * on VHE as on !VHE, where the eret to EL1 acts as a context
611  * synchronization event.
612  */
613 #define kvm_call_hyp(f, ...)						\
614 	do {								\
615 		if (has_vhe()) {					\
616 			f(__VA_ARGS__);					\
617 			isb();						\
618 		} else {						\
619 			kvm_call_hyp_nvhe(f, ##__VA_ARGS__);		\
620 		}							\
621 	} while(0)
622 
623 #define kvm_call_hyp_ret(f, ...)					\
624 	({								\
625 		typeof(f(__VA_ARGS__)) ret;				\
626 									\
627 		if (has_vhe()) {					\
628 			ret = f(__VA_ARGS__);				\
629 			isb();						\
630 		} else {						\
631 			ret = kvm_call_hyp_nvhe(f, ##__VA_ARGS__);	\
632 		}							\
633 									\
634 		ret;							\
635 	})
636 #else /* __KVM_NVHE_HYPERVISOR__ */
637 #define kvm_call_hyp(f, ...) f(__VA_ARGS__)
638 #define kvm_call_hyp_ret(f, ...) f(__VA_ARGS__)
639 #define kvm_call_hyp_nvhe(f, ...) f(__VA_ARGS__)
640 #endif /* __KVM_NVHE_HYPERVISOR__ */
641 
642 void force_vm_exit(const cpumask_t *mask);
643 void kvm_mmu_wp_memory_region(struct kvm *kvm, int slot);
644 
645 int handle_exit(struct kvm_vcpu *vcpu, int exception_index);
646 void handle_exit_early(struct kvm_vcpu *vcpu, int exception_index);
647 
648 int kvm_handle_cp14_load_store(struct kvm_vcpu *vcpu);
649 int kvm_handle_cp14_32(struct kvm_vcpu *vcpu);
650 int kvm_handle_cp14_64(struct kvm_vcpu *vcpu);
651 int kvm_handle_cp15_32(struct kvm_vcpu *vcpu);
652 int kvm_handle_cp15_64(struct kvm_vcpu *vcpu);
653 int kvm_handle_sys_reg(struct kvm_vcpu *vcpu);
654 
655 void kvm_reset_sys_regs(struct kvm_vcpu *vcpu);
656 
657 void kvm_sys_reg_table_init(void);
658 
659 /* MMIO helpers */
660 void kvm_mmio_write_buf(void *buf, unsigned int len, unsigned long data);
661 unsigned long kvm_mmio_read_buf(const void *buf, unsigned int len);
662 
663 int kvm_handle_mmio_return(struct kvm_vcpu *vcpu);
664 int io_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa);
665 
666 int kvm_perf_init(void);
667 int kvm_perf_teardown(void);
668 
669 long kvm_hypercall_pv_features(struct kvm_vcpu *vcpu);
670 gpa_t kvm_init_stolen_time(struct kvm_vcpu *vcpu);
671 void kvm_update_stolen_time(struct kvm_vcpu *vcpu);
672 
673 bool kvm_arm_pvtime_supported(void);
674 int kvm_arm_pvtime_set_attr(struct kvm_vcpu *vcpu,
675 			    struct kvm_device_attr *attr);
676 int kvm_arm_pvtime_get_attr(struct kvm_vcpu *vcpu,
677 			    struct kvm_device_attr *attr);
678 int kvm_arm_pvtime_has_attr(struct kvm_vcpu *vcpu,
679 			    struct kvm_device_attr *attr);
680 
681 static inline void kvm_arm_pvtime_vcpu_init(struct kvm_vcpu_arch *vcpu_arch)
682 {
683 	vcpu_arch->steal.base = GPA_INVALID;
684 }
685 
686 static inline bool kvm_arm_is_pvtime_enabled(struct kvm_vcpu_arch *vcpu_arch)
687 {
688 	return (vcpu_arch->steal.base != GPA_INVALID);
689 }
690 
691 void kvm_set_sei_esr(struct kvm_vcpu *vcpu, u64 syndrome);
692 
693 struct kvm_vcpu *kvm_mpidr_to_vcpu(struct kvm *kvm, unsigned long mpidr);
694 
695 DECLARE_KVM_HYP_PER_CPU(struct kvm_host_data, kvm_host_data);
696 
697 static inline void kvm_init_host_cpu_context(struct kvm_cpu_context *cpu_ctxt)
698 {
699 	/* The host's MPIDR is immutable, so let's set it up at boot time */
700 	ctxt_sys_reg(cpu_ctxt, MPIDR_EL1) = read_cpuid_mpidr();
701 }
702 
703 void kvm_arm_vcpu_ptrauth_trap(struct kvm_vcpu *vcpu);
704 
705 static inline void kvm_arch_hardware_unsetup(void) {}
706 static inline void kvm_arch_sync_events(struct kvm *kvm) {}
707 static inline void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu) {}
708 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
709 
710 void kvm_arm_init_debug(void);
711 void kvm_arm_setup_debug(struct kvm_vcpu *vcpu);
712 void kvm_arm_clear_debug(struct kvm_vcpu *vcpu);
713 void kvm_arm_reset_debug_ptr(struct kvm_vcpu *vcpu);
714 int kvm_arm_vcpu_arch_set_attr(struct kvm_vcpu *vcpu,
715 			       struct kvm_device_attr *attr);
716 int kvm_arm_vcpu_arch_get_attr(struct kvm_vcpu *vcpu,
717 			       struct kvm_device_attr *attr);
718 int kvm_arm_vcpu_arch_has_attr(struct kvm_vcpu *vcpu,
719 			       struct kvm_device_attr *attr);
720 
721 /* Guest/host FPSIMD coordination helpers */
722 int kvm_arch_vcpu_run_map_fp(struct kvm_vcpu *vcpu);
723 void kvm_arch_vcpu_load_fp(struct kvm_vcpu *vcpu);
724 void kvm_arch_vcpu_ctxsync_fp(struct kvm_vcpu *vcpu);
725 void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu);
726 
727 static inline bool kvm_pmu_counter_deferred(struct perf_event_attr *attr)
728 {
729 	return (!has_vhe() && attr->exclude_host);
730 }
731 
732 #ifdef CONFIG_KVM /* Avoid conflicts with core headers if CONFIG_KVM=n */
733 static inline int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu)
734 {
735 	return kvm_arch_vcpu_run_map_fp(vcpu);
736 }
737 
738 void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr);
739 void kvm_clr_pmu_events(u32 clr);
740 
741 void kvm_vcpu_pmu_restore_guest(struct kvm_vcpu *vcpu);
742 void kvm_vcpu_pmu_restore_host(struct kvm_vcpu *vcpu);
743 
744 void setup_kvm_el2_caps(void);
745 #else
746 static inline void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr) {}
747 static inline void kvm_clr_pmu_events(u32 clr) {}
748 
749 static inline void setup_kvm_el2_caps(void) {}
750 #endif
751 
752 void kvm_vcpu_load_sysregs_vhe(struct kvm_vcpu *vcpu);
753 void kvm_vcpu_put_sysregs_vhe(struct kvm_vcpu *vcpu);
754 
755 int kvm_set_ipa_limit(void);
756 
757 #define __KVM_HAVE_ARCH_VM_ALLOC
758 struct kvm *kvm_arch_alloc_vm(void);
759 void kvm_arch_free_vm(struct kvm *kvm);
760 
761 int kvm_arm_setup_stage2(struct kvm *kvm, unsigned long type);
762 
763 int kvm_arm_vcpu_finalize(struct kvm_vcpu *vcpu, int feature);
764 bool kvm_arm_vcpu_is_finalized(struct kvm_vcpu *vcpu);
765 
766 #define kvm_arm_vcpu_sve_finalized(vcpu) \
767 	((vcpu)->arch.flags & KVM_ARM64_VCPU_SVE_FINALIZED)
768 
769 #define kvm_vcpu_has_pmu(vcpu)					\
770 	(test_bit(KVM_ARM_VCPU_PMU_V3, (vcpu)->arch.features))
771 
772 int kvm_trng_call(struct kvm_vcpu *vcpu);
773 #ifdef CONFIG_KVM
774 extern phys_addr_t hyp_mem_base;
775 extern phys_addr_t hyp_mem_size;
776 void __init kvm_hyp_reserve(void);
777 #else
778 static inline void kvm_hyp_reserve(void) { }
779 #endif
780 
781 #endif /* __ARM64_KVM_HOST_H__ */
782