xref: /openbmc/linux/arch/arm64/include/asm/kvm_host.h (revision f502cc56)
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 
50 #define KVM_DIRTY_LOG_MANUAL_CAPS   (KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE | \
51 				     KVM_DIRTY_LOG_INITIALLY_SET)
52 
53 #define KVM_HAVE_MMU_RWLOCK
54 
55 /*
56  * Mode of operation configurable with kvm-arm.mode early param.
57  * See Documentation/admin-guide/kernel-parameters.txt for more information.
58  */
59 enum kvm_mode {
60 	KVM_MODE_DEFAULT,
61 	KVM_MODE_PROTECTED,
62 	KVM_MODE_NONE,
63 };
64 enum kvm_mode kvm_get_mode(void);
65 
66 DECLARE_STATIC_KEY_FALSE(userspace_irqchip_in_use);
67 
68 extern unsigned int kvm_sve_max_vl;
69 int kvm_arm_init_sve(void);
70 
71 u32 __attribute_const__ kvm_target_cpu(void);
72 int kvm_reset_vcpu(struct kvm_vcpu *vcpu);
73 void kvm_arm_vcpu_destroy(struct kvm_vcpu *vcpu);
74 
75 struct kvm_vmid {
76 	atomic64_t id;
77 };
78 
79 struct kvm_s2_mmu {
80 	struct kvm_vmid vmid;
81 
82 	/*
83 	 * stage2 entry level table
84 	 *
85 	 * Two kvm_s2_mmu structures in the same VM can point to the same
86 	 * pgd here.  This happens when running a guest using a
87 	 * translation regime that isn't affected by its own stage-2
88 	 * translation, such as a non-VHE hypervisor running at vEL2, or
89 	 * for vEL1/EL0 with vHCR_EL2.VM == 0.  In that case, we use the
90 	 * canonical stage-2 page tables.
91 	 */
92 	phys_addr_t	pgd_phys;
93 	struct kvm_pgtable *pgt;
94 
95 	/* The last vcpu id that ran on each physical CPU */
96 	int __percpu *last_vcpu_ran;
97 
98 	struct kvm_arch *arch;
99 };
100 
101 struct kvm_arch_memory_slot {
102 };
103 
104 struct kvm_arch {
105 	struct kvm_s2_mmu mmu;
106 
107 	/* VTCR_EL2 value for this VM */
108 	u64    vtcr;
109 
110 	/* Interrupt controller */
111 	struct vgic_dist	vgic;
112 
113 	/* Mandated version of PSCI */
114 	u32 psci_version;
115 
116 	/*
117 	 * If we encounter a data abort without valid instruction syndrome
118 	 * information, report this to user space.  User space can (and
119 	 * should) opt in to this feature if KVM_CAP_ARM_NISV_TO_USER is
120 	 * supported.
121 	 */
122 #define KVM_ARCH_FLAG_RETURN_NISV_IO_ABORT_TO_USER	0
123 	/* Memory Tagging Extension enabled for the guest */
124 #define KVM_ARCH_FLAG_MTE_ENABLED			1
125 	/* At least one vCPU has ran in the VM */
126 #define KVM_ARCH_FLAG_HAS_RAN_ONCE			2
127 	/*
128 	 * The following two bits are used to indicate the guest's EL1
129 	 * register width configuration. A value of KVM_ARCH_FLAG_EL1_32BIT
130 	 * bit is valid only when KVM_ARCH_FLAG_REG_WIDTH_CONFIGURED is set.
131 	 * Otherwise, the guest's EL1 register width has not yet been
132 	 * determined yet.
133 	 */
134 #define KVM_ARCH_FLAG_REG_WIDTH_CONFIGURED		3
135 #define KVM_ARCH_FLAG_EL1_32BIT				4
136 
137 	unsigned long flags;
138 
139 	/*
140 	 * VM-wide PMU filter, implemented as a bitmap and big enough for
141 	 * up to 2^10 events (ARMv8.0) or 2^16 events (ARMv8.1+).
142 	 */
143 	unsigned long *pmu_filter;
144 	struct arm_pmu *arm_pmu;
145 
146 	cpumask_var_t supported_cpus;
147 
148 	u8 pfr0_csv2;
149 	u8 pfr0_csv3;
150 };
151 
152 struct kvm_vcpu_fault_info {
153 	u32 esr_el2;		/* Hyp Syndrom Register */
154 	u64 far_el2;		/* Hyp Fault Address Register */
155 	u64 hpfar_el2;		/* Hyp IPA Fault Address Register */
156 	u64 disr_el1;		/* Deferred [SError] Status Register */
157 };
158 
159 enum vcpu_sysreg {
160 	__INVALID_SYSREG__,   /* 0 is reserved as an invalid value */
161 	MPIDR_EL1,	/* MultiProcessor Affinity Register */
162 	CSSELR_EL1,	/* Cache Size Selection Register */
163 	SCTLR_EL1,	/* System Control Register */
164 	ACTLR_EL1,	/* Auxiliary Control Register */
165 	CPACR_EL1,	/* Coprocessor Access Control */
166 	ZCR_EL1,	/* SVE Control */
167 	TTBR0_EL1,	/* Translation Table Base Register 0 */
168 	TTBR1_EL1,	/* Translation Table Base Register 1 */
169 	TCR_EL1,	/* Translation Control Register */
170 	ESR_EL1,	/* Exception Syndrome Register */
171 	AFSR0_EL1,	/* Auxiliary Fault Status Register 0 */
172 	AFSR1_EL1,	/* Auxiliary Fault Status Register 1 */
173 	FAR_EL1,	/* Fault Address Register */
174 	MAIR_EL1,	/* Memory Attribute Indirection Register */
175 	VBAR_EL1,	/* Vector Base Address Register */
176 	CONTEXTIDR_EL1,	/* Context ID Register */
177 	TPIDR_EL0,	/* Thread ID, User R/W */
178 	TPIDRRO_EL0,	/* Thread ID, User R/O */
179 	TPIDR_EL1,	/* Thread ID, Privileged */
180 	AMAIR_EL1,	/* Aux Memory Attribute Indirection Register */
181 	CNTKCTL_EL1,	/* Timer Control Register (EL1) */
182 	PAR_EL1,	/* Physical Address Register */
183 	MDSCR_EL1,	/* Monitor Debug System Control Register */
184 	MDCCINT_EL1,	/* Monitor Debug Comms Channel Interrupt Enable Reg */
185 	OSLSR_EL1,	/* OS Lock Status Register */
186 	DISR_EL1,	/* Deferred Interrupt Status Register */
187 
188 	/* Performance Monitors Registers */
189 	PMCR_EL0,	/* Control Register */
190 	PMSELR_EL0,	/* Event Counter Selection Register */
191 	PMEVCNTR0_EL0,	/* Event Counter Register (0-30) */
192 	PMEVCNTR30_EL0 = PMEVCNTR0_EL0 + 30,
193 	PMCCNTR_EL0,	/* Cycle Counter Register */
194 	PMEVTYPER0_EL0,	/* Event Type Register (0-30) */
195 	PMEVTYPER30_EL0 = PMEVTYPER0_EL0 + 30,
196 	PMCCFILTR_EL0,	/* Cycle Count Filter Register */
197 	PMCNTENSET_EL0,	/* Count Enable Set Register */
198 	PMINTENSET_EL1,	/* Interrupt Enable Set Register */
199 	PMOVSSET_EL0,	/* Overflow Flag Status Set Register */
200 	PMUSERENR_EL0,	/* User Enable Register */
201 
202 	/* Pointer Authentication Registers in a strict increasing order. */
203 	APIAKEYLO_EL1,
204 	APIAKEYHI_EL1,
205 	APIBKEYLO_EL1,
206 	APIBKEYHI_EL1,
207 	APDAKEYLO_EL1,
208 	APDAKEYHI_EL1,
209 	APDBKEYLO_EL1,
210 	APDBKEYHI_EL1,
211 	APGAKEYLO_EL1,
212 	APGAKEYHI_EL1,
213 
214 	ELR_EL1,
215 	SP_EL1,
216 	SPSR_EL1,
217 
218 	CNTVOFF_EL2,
219 	CNTV_CVAL_EL0,
220 	CNTV_CTL_EL0,
221 	CNTP_CVAL_EL0,
222 	CNTP_CTL_EL0,
223 
224 	/* Memory Tagging Extension registers */
225 	RGSR_EL1,	/* Random Allocation Tag Seed Register */
226 	GCR_EL1,	/* Tag Control Register */
227 	TFSR_EL1,	/* Tag Fault Status Register (EL1) */
228 	TFSRE0_EL1,	/* Tag Fault Status Register (EL0) */
229 
230 	/* 32bit specific registers. Keep them at the end of the range */
231 	DACR32_EL2,	/* Domain Access Control Register */
232 	IFSR32_EL2,	/* Instruction Fault Status Register */
233 	FPEXC32_EL2,	/* Floating-Point Exception Control Register */
234 	DBGVCR32_EL2,	/* Debug Vector Catch Register */
235 
236 	NR_SYS_REGS	/* Nothing after this line! */
237 };
238 
239 struct kvm_cpu_context {
240 	struct user_pt_regs regs;	/* sp = sp_el0 */
241 
242 	u64	spsr_abt;
243 	u64	spsr_und;
244 	u64	spsr_irq;
245 	u64	spsr_fiq;
246 
247 	struct user_fpsimd_state fp_regs;
248 
249 	u64 sys_regs[NR_SYS_REGS];
250 
251 	struct kvm_vcpu *__hyp_running_vcpu;
252 };
253 
254 struct kvm_pmu_events {
255 	u32 events_host;
256 	u32 events_guest;
257 };
258 
259 struct kvm_host_data {
260 	struct kvm_cpu_context host_ctxt;
261 	struct kvm_pmu_events pmu_events;
262 };
263 
264 struct kvm_host_psci_config {
265 	/* PSCI version used by host. */
266 	u32 version;
267 
268 	/* Function IDs used by host if version is v0.1. */
269 	struct psci_0_1_function_ids function_ids_0_1;
270 
271 	bool psci_0_1_cpu_suspend_implemented;
272 	bool psci_0_1_cpu_on_implemented;
273 	bool psci_0_1_cpu_off_implemented;
274 	bool psci_0_1_migrate_implemented;
275 };
276 
277 extern struct kvm_host_psci_config kvm_nvhe_sym(kvm_host_psci_config);
278 #define kvm_host_psci_config CHOOSE_NVHE_SYM(kvm_host_psci_config)
279 
280 extern s64 kvm_nvhe_sym(hyp_physvirt_offset);
281 #define hyp_physvirt_offset CHOOSE_NVHE_SYM(hyp_physvirt_offset)
282 
283 extern u64 kvm_nvhe_sym(hyp_cpu_logical_map)[NR_CPUS];
284 #define hyp_cpu_logical_map CHOOSE_NVHE_SYM(hyp_cpu_logical_map)
285 
286 struct vcpu_reset_state {
287 	unsigned long	pc;
288 	unsigned long	r0;
289 	bool		be;
290 	bool		reset;
291 };
292 
293 struct kvm_vcpu_arch {
294 	struct kvm_cpu_context ctxt;
295 	void *sve_state;
296 	unsigned int sve_max_vl;
297 
298 	/* Stage 2 paging state used by the hardware on next switch */
299 	struct kvm_s2_mmu *hw_mmu;
300 
301 	/* Values of trap registers for the guest. */
302 	u64 hcr_el2;
303 	u64 mdcr_el2;
304 	u64 cptr_el2;
305 
306 	/* Values of trap registers for the host before guest entry. */
307 	u64 mdcr_el2_host;
308 
309 	/* Exception Information */
310 	struct kvm_vcpu_fault_info fault;
311 
312 	/* Miscellaneous vcpu state flags */
313 	u64 flags;
314 
315 	/*
316 	 * We maintain more than a single set of debug registers to support
317 	 * debugging the guest from the host and to maintain separate host and
318 	 * guest state during world switches. vcpu_debug_state are the debug
319 	 * registers of the vcpu as the guest sees them.  host_debug_state are
320 	 * the host registers which are saved and restored during
321 	 * world switches. external_debug_state contains the debug
322 	 * values we want to debug the guest. This is set via the
323 	 * KVM_SET_GUEST_DEBUG ioctl.
324 	 *
325 	 * debug_ptr points to the set of debug registers that should be loaded
326 	 * onto the hardware when running the guest.
327 	 */
328 	struct kvm_guest_debug_arch *debug_ptr;
329 	struct kvm_guest_debug_arch vcpu_debug_state;
330 	struct kvm_guest_debug_arch external_debug_state;
331 
332 	struct user_fpsimd_state *host_fpsimd_state;	/* hyp VA */
333 	struct task_struct *parent_task;
334 
335 	struct {
336 		/* {Break,watch}point registers */
337 		struct kvm_guest_debug_arch regs;
338 		/* Statistical profiling extension */
339 		u64 pmscr_el1;
340 		/* Self-hosted trace */
341 		u64 trfcr_el1;
342 	} host_debug_state;
343 
344 	/* VGIC state */
345 	struct vgic_cpu vgic_cpu;
346 	struct arch_timer_cpu timer_cpu;
347 	struct kvm_pmu pmu;
348 
349 	/*
350 	 * Anything that is not used directly from assembly code goes
351 	 * here.
352 	 */
353 
354 	/*
355 	 * Guest registers we preserve during guest debugging.
356 	 *
357 	 * These shadow registers are updated by the kvm_handle_sys_reg
358 	 * trap handler if the guest accesses or updates them while we
359 	 * are using guest debug.
360 	 */
361 	struct {
362 		u32	mdscr_el1;
363 	} guest_debug_preserved;
364 
365 	/* vcpu power-off state */
366 	bool power_off;
367 
368 	/* Don't run the guest (internal implementation need) */
369 	bool pause;
370 
371 	/* Cache some mmu pages needed inside spinlock regions */
372 	struct kvm_mmu_memory_cache mmu_page_cache;
373 
374 	/* Target CPU and feature flags */
375 	int target;
376 	DECLARE_BITMAP(features, KVM_VCPU_MAX_FEATURES);
377 
378 	/* Virtual SError ESR to restore when HCR_EL2.VSE is set */
379 	u64 vsesr_el2;
380 
381 	/* Additional reset state */
382 	struct vcpu_reset_state	reset_state;
383 
384 	/* True when deferrable sysregs are loaded on the physical CPU,
385 	 * see kvm_vcpu_load_sysregs_vhe and kvm_vcpu_put_sysregs_vhe. */
386 	bool sysregs_loaded_on_cpu;
387 
388 	/* Guest PV state */
389 	struct {
390 		u64 last_steal;
391 		gpa_t base;
392 	} steal;
393 };
394 
395 /* Pointer to the vcpu's SVE FFR for sve_{save,load}_state() */
396 #define vcpu_sve_pffr(vcpu) (kern_hyp_va((vcpu)->arch.sve_state) +	\
397 			     sve_ffr_offset((vcpu)->arch.sve_max_vl))
398 
399 #define vcpu_sve_max_vq(vcpu)	sve_vq_from_vl((vcpu)->arch.sve_max_vl)
400 
401 #define vcpu_sve_state_size(vcpu) ({					\
402 	size_t __size_ret;						\
403 	unsigned int __vcpu_vq;						\
404 									\
405 	if (WARN_ON(!sve_vl_valid((vcpu)->arch.sve_max_vl))) {		\
406 		__size_ret = 0;						\
407 	} else {							\
408 		__vcpu_vq = vcpu_sve_max_vq(vcpu);			\
409 		__size_ret = SVE_SIG_REGS_SIZE(__vcpu_vq);		\
410 	}								\
411 									\
412 	__size_ret;							\
413 })
414 
415 /* vcpu_arch flags field values: */
416 #define KVM_ARM64_DEBUG_DIRTY		(1 << 0)
417 #define KVM_ARM64_FP_ENABLED		(1 << 1) /* guest FP regs loaded */
418 #define KVM_ARM64_FP_HOST		(1 << 2) /* host FP regs loaded */
419 #define KVM_ARM64_HOST_SVE_ENABLED	(1 << 4) /* SVE enabled for EL0 */
420 #define KVM_ARM64_GUEST_HAS_SVE		(1 << 5) /* SVE exposed to guest */
421 #define KVM_ARM64_VCPU_SVE_FINALIZED	(1 << 6) /* SVE config completed */
422 #define KVM_ARM64_GUEST_HAS_PTRAUTH	(1 << 7) /* PTRAUTH exposed to guest */
423 #define KVM_ARM64_PENDING_EXCEPTION	(1 << 8) /* Exception pending */
424 /*
425  * Overlaps with KVM_ARM64_EXCEPT_MASK on purpose so that it can't be
426  * set together with an exception...
427  */
428 #define KVM_ARM64_INCREMENT_PC		(1 << 9) /* Increment PC */
429 #define KVM_ARM64_EXCEPT_MASK		(7 << 9) /* Target EL/MODE */
430 /*
431  * When KVM_ARM64_PENDING_EXCEPTION is set, KVM_ARM64_EXCEPT_MASK can
432  * take the following values:
433  *
434  * For AArch32 EL1:
435  */
436 #define KVM_ARM64_EXCEPT_AA32_UND	(0 << 9)
437 #define KVM_ARM64_EXCEPT_AA32_IABT	(1 << 9)
438 #define KVM_ARM64_EXCEPT_AA32_DABT	(2 << 9)
439 /* For AArch64: */
440 #define KVM_ARM64_EXCEPT_AA64_ELx_SYNC	(0 << 9)
441 #define KVM_ARM64_EXCEPT_AA64_ELx_IRQ	(1 << 9)
442 #define KVM_ARM64_EXCEPT_AA64_ELx_FIQ	(2 << 9)
443 #define KVM_ARM64_EXCEPT_AA64_ELx_SERR	(3 << 9)
444 #define KVM_ARM64_EXCEPT_AA64_EL1	(0 << 11)
445 #define KVM_ARM64_EXCEPT_AA64_EL2	(1 << 11)
446 
447 #define KVM_ARM64_DEBUG_STATE_SAVE_SPE	(1 << 12) /* Save SPE context if active  */
448 #define KVM_ARM64_DEBUG_STATE_SAVE_TRBE	(1 << 13) /* Save TRBE context if active  */
449 #define KVM_ARM64_FP_FOREIGN_FPSTATE	(1 << 14)
450 #define KVM_ARM64_ON_UNSUPPORTED_CPU	(1 << 15) /* Physical CPU not in supported_cpus */
451 
452 #define KVM_GUESTDBG_VALID_MASK (KVM_GUESTDBG_ENABLE | \
453 				 KVM_GUESTDBG_USE_SW_BP | \
454 				 KVM_GUESTDBG_USE_HW | \
455 				 KVM_GUESTDBG_SINGLESTEP)
456 
457 #define vcpu_has_sve(vcpu) (system_supports_sve() &&			\
458 			    ((vcpu)->arch.flags & KVM_ARM64_GUEST_HAS_SVE))
459 
460 #ifdef CONFIG_ARM64_PTR_AUTH
461 #define vcpu_has_ptrauth(vcpu)						\
462 	((cpus_have_final_cap(ARM64_HAS_ADDRESS_AUTH) ||		\
463 	  cpus_have_final_cap(ARM64_HAS_GENERIC_AUTH)) &&		\
464 	 (vcpu)->arch.flags & KVM_ARM64_GUEST_HAS_PTRAUTH)
465 #else
466 #define vcpu_has_ptrauth(vcpu)		false
467 #endif
468 
469 #define vcpu_on_unsupported_cpu(vcpu)					\
470 	((vcpu)->arch.flags & KVM_ARM64_ON_UNSUPPORTED_CPU)
471 
472 #define vcpu_set_on_unsupported_cpu(vcpu)				\
473 	((vcpu)->arch.flags |= KVM_ARM64_ON_UNSUPPORTED_CPU)
474 
475 #define vcpu_clear_on_unsupported_cpu(vcpu)				\
476 	((vcpu)->arch.flags &= ~KVM_ARM64_ON_UNSUPPORTED_CPU)
477 
478 #define vcpu_gp_regs(v)		(&(v)->arch.ctxt.regs)
479 
480 /*
481  * Only use __vcpu_sys_reg/ctxt_sys_reg if you know you want the
482  * memory backed version of a register, and not the one most recently
483  * accessed by a running VCPU.  For example, for userspace access or
484  * for system registers that are never context switched, but only
485  * emulated.
486  */
487 #define __ctxt_sys_reg(c,r)	(&(c)->sys_regs[(r)])
488 
489 #define ctxt_sys_reg(c,r)	(*__ctxt_sys_reg(c,r))
490 
491 #define __vcpu_sys_reg(v,r)	(ctxt_sys_reg(&(v)->arch.ctxt, (r)))
492 
493 u64 vcpu_read_sys_reg(const struct kvm_vcpu *vcpu, int reg);
494 void vcpu_write_sys_reg(struct kvm_vcpu *vcpu, u64 val, int reg);
495 
496 static inline bool __vcpu_read_sys_reg_from_cpu(int reg, u64 *val)
497 {
498 	/*
499 	 * *** VHE ONLY ***
500 	 *
501 	 * System registers listed in the switch are not saved on every
502 	 * exit from the guest but are only saved on vcpu_put.
503 	 *
504 	 * Note that MPIDR_EL1 for the guest is set by KVM via VMPIDR_EL2 but
505 	 * should never be listed below, because the guest cannot modify its
506 	 * own MPIDR_EL1 and MPIDR_EL1 is accessed for VCPU A from VCPU B's
507 	 * thread when emulating cross-VCPU communication.
508 	 */
509 	if (!has_vhe())
510 		return false;
511 
512 	switch (reg) {
513 	case CSSELR_EL1:	*val = read_sysreg_s(SYS_CSSELR_EL1);	break;
514 	case SCTLR_EL1:		*val = read_sysreg_s(SYS_SCTLR_EL12);	break;
515 	case CPACR_EL1:		*val = read_sysreg_s(SYS_CPACR_EL12);	break;
516 	case TTBR0_EL1:		*val = read_sysreg_s(SYS_TTBR0_EL12);	break;
517 	case TTBR1_EL1:		*val = read_sysreg_s(SYS_TTBR1_EL12);	break;
518 	case TCR_EL1:		*val = read_sysreg_s(SYS_TCR_EL12);	break;
519 	case ESR_EL1:		*val = read_sysreg_s(SYS_ESR_EL12);	break;
520 	case AFSR0_EL1:		*val = read_sysreg_s(SYS_AFSR0_EL12);	break;
521 	case AFSR1_EL1:		*val = read_sysreg_s(SYS_AFSR1_EL12);	break;
522 	case FAR_EL1:		*val = read_sysreg_s(SYS_FAR_EL12);	break;
523 	case MAIR_EL1:		*val = read_sysreg_s(SYS_MAIR_EL12);	break;
524 	case VBAR_EL1:		*val = read_sysreg_s(SYS_VBAR_EL12);	break;
525 	case CONTEXTIDR_EL1:	*val = read_sysreg_s(SYS_CONTEXTIDR_EL12);break;
526 	case TPIDR_EL0:		*val = read_sysreg_s(SYS_TPIDR_EL0);	break;
527 	case TPIDRRO_EL0:	*val = read_sysreg_s(SYS_TPIDRRO_EL0);	break;
528 	case TPIDR_EL1:		*val = read_sysreg_s(SYS_TPIDR_EL1);	break;
529 	case AMAIR_EL1:		*val = read_sysreg_s(SYS_AMAIR_EL12);	break;
530 	case CNTKCTL_EL1:	*val = read_sysreg_s(SYS_CNTKCTL_EL12);	break;
531 	case ELR_EL1:		*val = read_sysreg_s(SYS_ELR_EL12);	break;
532 	case PAR_EL1:		*val = read_sysreg_par();		break;
533 	case DACR32_EL2:	*val = read_sysreg_s(SYS_DACR32_EL2);	break;
534 	case IFSR32_EL2:	*val = read_sysreg_s(SYS_IFSR32_EL2);	break;
535 	case DBGVCR32_EL2:	*val = read_sysreg_s(SYS_DBGVCR32_EL2);	break;
536 	default:		return false;
537 	}
538 
539 	return true;
540 }
541 
542 static inline bool __vcpu_write_sys_reg_to_cpu(u64 val, int reg)
543 {
544 	/*
545 	 * *** VHE ONLY ***
546 	 *
547 	 * System registers listed in the switch are not restored on every
548 	 * entry to the guest but are only restored on vcpu_load.
549 	 *
550 	 * Note that MPIDR_EL1 for the guest is set by KVM via VMPIDR_EL2 but
551 	 * should never be listed below, because the MPIDR should only be set
552 	 * once, before running the VCPU, and never changed later.
553 	 */
554 	if (!has_vhe())
555 		return false;
556 
557 	switch (reg) {
558 	case CSSELR_EL1:	write_sysreg_s(val, SYS_CSSELR_EL1);	break;
559 	case SCTLR_EL1:		write_sysreg_s(val, SYS_SCTLR_EL12);	break;
560 	case CPACR_EL1:		write_sysreg_s(val, SYS_CPACR_EL12);	break;
561 	case TTBR0_EL1:		write_sysreg_s(val, SYS_TTBR0_EL12);	break;
562 	case TTBR1_EL1:		write_sysreg_s(val, SYS_TTBR1_EL12);	break;
563 	case TCR_EL1:		write_sysreg_s(val, SYS_TCR_EL12);	break;
564 	case ESR_EL1:		write_sysreg_s(val, SYS_ESR_EL12);	break;
565 	case AFSR0_EL1:		write_sysreg_s(val, SYS_AFSR0_EL12);	break;
566 	case AFSR1_EL1:		write_sysreg_s(val, SYS_AFSR1_EL12);	break;
567 	case FAR_EL1:		write_sysreg_s(val, SYS_FAR_EL12);	break;
568 	case MAIR_EL1:		write_sysreg_s(val, SYS_MAIR_EL12);	break;
569 	case VBAR_EL1:		write_sysreg_s(val, SYS_VBAR_EL12);	break;
570 	case CONTEXTIDR_EL1:	write_sysreg_s(val, SYS_CONTEXTIDR_EL12);break;
571 	case TPIDR_EL0:		write_sysreg_s(val, SYS_TPIDR_EL0);	break;
572 	case TPIDRRO_EL0:	write_sysreg_s(val, SYS_TPIDRRO_EL0);	break;
573 	case TPIDR_EL1:		write_sysreg_s(val, SYS_TPIDR_EL1);	break;
574 	case AMAIR_EL1:		write_sysreg_s(val, SYS_AMAIR_EL12);	break;
575 	case CNTKCTL_EL1:	write_sysreg_s(val, SYS_CNTKCTL_EL12);	break;
576 	case ELR_EL1:		write_sysreg_s(val, SYS_ELR_EL12);	break;
577 	case PAR_EL1:		write_sysreg_s(val, SYS_PAR_EL1);	break;
578 	case DACR32_EL2:	write_sysreg_s(val, SYS_DACR32_EL2);	break;
579 	case IFSR32_EL2:	write_sysreg_s(val, SYS_IFSR32_EL2);	break;
580 	case DBGVCR32_EL2:	write_sysreg_s(val, SYS_DBGVCR32_EL2);	break;
581 	default:		return false;
582 	}
583 
584 	return true;
585 }
586 
587 struct kvm_vm_stat {
588 	struct kvm_vm_stat_generic generic;
589 };
590 
591 struct kvm_vcpu_stat {
592 	struct kvm_vcpu_stat_generic generic;
593 	u64 hvc_exit_stat;
594 	u64 wfe_exit_stat;
595 	u64 wfi_exit_stat;
596 	u64 mmio_exit_user;
597 	u64 mmio_exit_kernel;
598 	u64 signal_exits;
599 	u64 exits;
600 };
601 
602 void kvm_vcpu_preferred_target(struct kvm_vcpu_init *init);
603 unsigned long kvm_arm_num_regs(struct kvm_vcpu *vcpu);
604 int kvm_arm_copy_reg_indices(struct kvm_vcpu *vcpu, u64 __user *indices);
605 int kvm_arm_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
606 int kvm_arm_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
607 
608 unsigned long kvm_arm_num_sys_reg_descs(struct kvm_vcpu *vcpu);
609 int kvm_arm_copy_sys_reg_indices(struct kvm_vcpu *vcpu, u64 __user *uindices);
610 int kvm_arm_sys_reg_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *);
611 int kvm_arm_sys_reg_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *);
612 
613 int __kvm_arm_vcpu_get_events(struct kvm_vcpu *vcpu,
614 			      struct kvm_vcpu_events *events);
615 
616 int __kvm_arm_vcpu_set_events(struct kvm_vcpu *vcpu,
617 			      struct kvm_vcpu_events *events);
618 
619 #define KVM_ARCH_WANT_MMU_NOTIFIER
620 
621 void kvm_arm_halt_guest(struct kvm *kvm);
622 void kvm_arm_resume_guest(struct kvm *kvm);
623 
624 #define vcpu_has_run_once(vcpu)	!!rcu_access_pointer((vcpu)->pid)
625 
626 #ifndef __KVM_NVHE_HYPERVISOR__
627 #define kvm_call_hyp_nvhe(f, ...)						\
628 	({								\
629 		struct arm_smccc_res res;				\
630 									\
631 		arm_smccc_1_1_hvc(KVM_HOST_SMCCC_FUNC(f),		\
632 				  ##__VA_ARGS__, &res);			\
633 		WARN_ON(res.a0 != SMCCC_RET_SUCCESS);			\
634 									\
635 		res.a1;							\
636 	})
637 
638 /*
639  * The couple of isb() below are there to guarantee the same behaviour
640  * on VHE as on !VHE, where the eret to EL1 acts as a context
641  * synchronization event.
642  */
643 #define kvm_call_hyp(f, ...)						\
644 	do {								\
645 		if (has_vhe()) {					\
646 			f(__VA_ARGS__);					\
647 			isb();						\
648 		} else {						\
649 			kvm_call_hyp_nvhe(f, ##__VA_ARGS__);		\
650 		}							\
651 	} while(0)
652 
653 #define kvm_call_hyp_ret(f, ...)					\
654 	({								\
655 		typeof(f(__VA_ARGS__)) ret;				\
656 									\
657 		if (has_vhe()) {					\
658 			ret = f(__VA_ARGS__);				\
659 			isb();						\
660 		} else {						\
661 			ret = kvm_call_hyp_nvhe(f, ##__VA_ARGS__);	\
662 		}							\
663 									\
664 		ret;							\
665 	})
666 #else /* __KVM_NVHE_HYPERVISOR__ */
667 #define kvm_call_hyp(f, ...) f(__VA_ARGS__)
668 #define kvm_call_hyp_ret(f, ...) f(__VA_ARGS__)
669 #define kvm_call_hyp_nvhe(f, ...) f(__VA_ARGS__)
670 #endif /* __KVM_NVHE_HYPERVISOR__ */
671 
672 void force_vm_exit(const cpumask_t *mask);
673 
674 int handle_exit(struct kvm_vcpu *vcpu, int exception_index);
675 void handle_exit_early(struct kvm_vcpu *vcpu, int exception_index);
676 
677 int kvm_handle_cp14_load_store(struct kvm_vcpu *vcpu);
678 int kvm_handle_cp14_32(struct kvm_vcpu *vcpu);
679 int kvm_handle_cp14_64(struct kvm_vcpu *vcpu);
680 int kvm_handle_cp15_32(struct kvm_vcpu *vcpu);
681 int kvm_handle_cp15_64(struct kvm_vcpu *vcpu);
682 int kvm_handle_sys_reg(struct kvm_vcpu *vcpu);
683 
684 void kvm_reset_sys_regs(struct kvm_vcpu *vcpu);
685 
686 void kvm_sys_reg_table_init(void);
687 
688 /* MMIO helpers */
689 void kvm_mmio_write_buf(void *buf, unsigned int len, unsigned long data);
690 unsigned long kvm_mmio_read_buf(const void *buf, unsigned int len);
691 
692 int kvm_handle_mmio_return(struct kvm_vcpu *vcpu);
693 int io_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa);
694 
695 /*
696  * Returns true if a Performance Monitoring Interrupt (PMI), a.k.a. perf event,
697  * arrived in guest context.  For arm64, any event that arrives while a vCPU is
698  * loaded is considered to be "in guest".
699  */
700 static inline bool kvm_arch_pmi_in_guest(struct kvm_vcpu *vcpu)
701 {
702 	return IS_ENABLED(CONFIG_GUEST_PERF_EVENTS) && !!vcpu;
703 }
704 
705 long kvm_hypercall_pv_features(struct kvm_vcpu *vcpu);
706 gpa_t kvm_init_stolen_time(struct kvm_vcpu *vcpu);
707 void kvm_update_stolen_time(struct kvm_vcpu *vcpu);
708 
709 bool kvm_arm_pvtime_supported(void);
710 int kvm_arm_pvtime_set_attr(struct kvm_vcpu *vcpu,
711 			    struct kvm_device_attr *attr);
712 int kvm_arm_pvtime_get_attr(struct kvm_vcpu *vcpu,
713 			    struct kvm_device_attr *attr);
714 int kvm_arm_pvtime_has_attr(struct kvm_vcpu *vcpu,
715 			    struct kvm_device_attr *attr);
716 
717 extern unsigned int kvm_arm_vmid_bits;
718 int kvm_arm_vmid_alloc_init(void);
719 void kvm_arm_vmid_alloc_free(void);
720 void kvm_arm_vmid_update(struct kvm_vmid *kvm_vmid);
721 void kvm_arm_vmid_clear_active(void);
722 
723 static inline void kvm_arm_pvtime_vcpu_init(struct kvm_vcpu_arch *vcpu_arch)
724 {
725 	vcpu_arch->steal.base = GPA_INVALID;
726 }
727 
728 static inline bool kvm_arm_is_pvtime_enabled(struct kvm_vcpu_arch *vcpu_arch)
729 {
730 	return (vcpu_arch->steal.base != GPA_INVALID);
731 }
732 
733 void kvm_set_sei_esr(struct kvm_vcpu *vcpu, u64 syndrome);
734 
735 struct kvm_vcpu *kvm_mpidr_to_vcpu(struct kvm *kvm, unsigned long mpidr);
736 
737 DECLARE_KVM_HYP_PER_CPU(struct kvm_host_data, kvm_host_data);
738 
739 static inline void kvm_init_host_cpu_context(struct kvm_cpu_context *cpu_ctxt)
740 {
741 	/* The host's MPIDR is immutable, so let's set it up at boot time */
742 	ctxt_sys_reg(cpu_ctxt, MPIDR_EL1) = read_cpuid_mpidr();
743 }
744 
745 static inline bool kvm_system_needs_idmapped_vectors(void)
746 {
747 	return cpus_have_const_cap(ARM64_SPECTRE_V3A);
748 }
749 
750 void kvm_arm_vcpu_ptrauth_trap(struct kvm_vcpu *vcpu);
751 
752 static inline void kvm_arch_hardware_unsetup(void) {}
753 static inline void kvm_arch_sync_events(struct kvm *kvm) {}
754 static inline void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu) {}
755 
756 void kvm_arm_init_debug(void);
757 void kvm_arm_vcpu_init_debug(struct kvm_vcpu *vcpu);
758 void kvm_arm_setup_debug(struct kvm_vcpu *vcpu);
759 void kvm_arm_clear_debug(struct kvm_vcpu *vcpu);
760 void kvm_arm_reset_debug_ptr(struct kvm_vcpu *vcpu);
761 
762 #define kvm_vcpu_os_lock_enabled(vcpu)		\
763 	(!!(__vcpu_sys_reg(vcpu, OSLSR_EL1) & SYS_OSLSR_OSLK))
764 
765 int kvm_arm_vcpu_arch_set_attr(struct kvm_vcpu *vcpu,
766 			       struct kvm_device_attr *attr);
767 int kvm_arm_vcpu_arch_get_attr(struct kvm_vcpu *vcpu,
768 			       struct kvm_device_attr *attr);
769 int kvm_arm_vcpu_arch_has_attr(struct kvm_vcpu *vcpu,
770 			       struct kvm_device_attr *attr);
771 
772 long kvm_vm_ioctl_mte_copy_tags(struct kvm *kvm,
773 				struct kvm_arm_copy_mte_tags *copy_tags);
774 
775 /* Guest/host FPSIMD coordination helpers */
776 int kvm_arch_vcpu_run_map_fp(struct kvm_vcpu *vcpu);
777 void kvm_arch_vcpu_load_fp(struct kvm_vcpu *vcpu);
778 void kvm_arch_vcpu_ctxflush_fp(struct kvm_vcpu *vcpu);
779 void kvm_arch_vcpu_ctxsync_fp(struct kvm_vcpu *vcpu);
780 void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu);
781 void kvm_vcpu_unshare_task_fp(struct kvm_vcpu *vcpu);
782 
783 static inline bool kvm_pmu_counter_deferred(struct perf_event_attr *attr)
784 {
785 	return (!has_vhe() && attr->exclude_host);
786 }
787 
788 /* Flags for host debug state */
789 void kvm_arch_vcpu_load_debug_state_flags(struct kvm_vcpu *vcpu);
790 void kvm_arch_vcpu_put_debug_state_flags(struct kvm_vcpu *vcpu);
791 
792 #ifdef CONFIG_KVM
793 void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr);
794 void kvm_clr_pmu_events(u32 clr);
795 
796 void kvm_vcpu_pmu_restore_guest(struct kvm_vcpu *vcpu);
797 void kvm_vcpu_pmu_restore_host(struct kvm_vcpu *vcpu);
798 #else
799 static inline void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr) {}
800 static inline void kvm_clr_pmu_events(u32 clr) {}
801 #endif
802 
803 void kvm_vcpu_load_sysregs_vhe(struct kvm_vcpu *vcpu);
804 void kvm_vcpu_put_sysregs_vhe(struct kvm_vcpu *vcpu);
805 
806 int kvm_set_ipa_limit(void);
807 
808 #define __KVM_HAVE_ARCH_VM_ALLOC
809 struct kvm *kvm_arch_alloc_vm(void);
810 
811 int kvm_arm_setup_stage2(struct kvm *kvm, unsigned long type);
812 
813 static inline bool kvm_vm_is_protected(struct kvm *kvm)
814 {
815 	return false;
816 }
817 
818 void kvm_init_protected_traps(struct kvm_vcpu *vcpu);
819 
820 int kvm_arm_vcpu_finalize(struct kvm_vcpu *vcpu, int feature);
821 bool kvm_arm_vcpu_is_finalized(struct kvm_vcpu *vcpu);
822 
823 #define kvm_arm_vcpu_sve_finalized(vcpu) \
824 	((vcpu)->arch.flags & KVM_ARM64_VCPU_SVE_FINALIZED)
825 
826 #define kvm_has_mte(kvm)					\
827 	(system_supports_mte() &&				\
828 	 test_bit(KVM_ARCH_FLAG_MTE_ENABLED, &(kvm)->arch.flags))
829 #define kvm_vcpu_has_pmu(vcpu)					\
830 	(test_bit(KVM_ARM_VCPU_PMU_V3, (vcpu)->arch.features))
831 
832 int kvm_trng_call(struct kvm_vcpu *vcpu);
833 #ifdef CONFIG_KVM
834 extern phys_addr_t hyp_mem_base;
835 extern phys_addr_t hyp_mem_size;
836 void __init kvm_hyp_reserve(void);
837 #else
838 static inline void kvm_hyp_reserve(void) { }
839 #endif
840 
841 #endif /* __ARM64_KVM_HOST_H__ */
842