xref: /openbmc/linux/arch/arm64/include/asm/kvm_host.h (revision e6486939)
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 	/* The maximum number of vCPUs depends on the used GIC model */
111 	int max_vcpus;
112 
113 	/* Interrupt controller */
114 	struct vgic_dist	vgic;
115 
116 	/* Mandated version of PSCI */
117 	u32 psci_version;
118 
119 	/*
120 	 * If we encounter a data abort without valid instruction syndrome
121 	 * information, report this to user space.  User space can (and
122 	 * should) opt in to this feature if KVM_CAP_ARM_NISV_TO_USER is
123 	 * supported.
124 	 */
125 #define KVM_ARCH_FLAG_RETURN_NISV_IO_ABORT_TO_USER	0
126 	/* Memory Tagging Extension enabled for the guest */
127 #define KVM_ARCH_FLAG_MTE_ENABLED			1
128 	/* At least one vCPU has ran in the VM */
129 #define KVM_ARCH_FLAG_HAS_RAN_ONCE			2
130 	unsigned long flags;
131 
132 	/*
133 	 * VM-wide PMU filter, implemented as a bitmap and big enough for
134 	 * up to 2^10 events (ARMv8.0) or 2^16 events (ARMv8.1+).
135 	 */
136 	unsigned long *pmu_filter;
137 	struct arm_pmu *arm_pmu;
138 
139 	cpumask_var_t supported_cpus;
140 
141 	u8 pfr0_csv2;
142 	u8 pfr0_csv3;
143 };
144 
145 struct kvm_vcpu_fault_info {
146 	u32 esr_el2;		/* Hyp Syndrom Register */
147 	u64 far_el2;		/* Hyp Fault Address Register */
148 	u64 hpfar_el2;		/* Hyp IPA Fault Address Register */
149 	u64 disr_el1;		/* Deferred [SError] Status Register */
150 };
151 
152 enum vcpu_sysreg {
153 	__INVALID_SYSREG__,   /* 0 is reserved as an invalid value */
154 	MPIDR_EL1,	/* MultiProcessor Affinity Register */
155 	CSSELR_EL1,	/* Cache Size Selection Register */
156 	SCTLR_EL1,	/* System Control Register */
157 	ACTLR_EL1,	/* Auxiliary Control Register */
158 	CPACR_EL1,	/* Coprocessor Access Control */
159 	ZCR_EL1,	/* SVE Control */
160 	TTBR0_EL1,	/* Translation Table Base Register 0 */
161 	TTBR1_EL1,	/* Translation Table Base Register 1 */
162 	TCR_EL1,	/* Translation Control Register */
163 	ESR_EL1,	/* Exception Syndrome Register */
164 	AFSR0_EL1,	/* Auxiliary Fault Status Register 0 */
165 	AFSR1_EL1,	/* Auxiliary Fault Status Register 1 */
166 	FAR_EL1,	/* Fault Address Register */
167 	MAIR_EL1,	/* Memory Attribute Indirection Register */
168 	VBAR_EL1,	/* Vector Base Address Register */
169 	CONTEXTIDR_EL1,	/* Context ID Register */
170 	TPIDR_EL0,	/* Thread ID, User R/W */
171 	TPIDRRO_EL0,	/* Thread ID, User R/O */
172 	TPIDR_EL1,	/* Thread ID, Privileged */
173 	AMAIR_EL1,	/* Aux Memory Attribute Indirection Register */
174 	CNTKCTL_EL1,	/* Timer Control Register (EL1) */
175 	PAR_EL1,	/* Physical Address Register */
176 	MDSCR_EL1,	/* Monitor Debug System Control Register */
177 	MDCCINT_EL1,	/* Monitor Debug Comms Channel Interrupt Enable Reg */
178 	OSLSR_EL1,	/* OS Lock Status Register */
179 	DISR_EL1,	/* Deferred Interrupt Status Register */
180 
181 	/* Performance Monitors Registers */
182 	PMCR_EL0,	/* Control Register */
183 	PMSELR_EL0,	/* Event Counter Selection Register */
184 	PMEVCNTR0_EL0,	/* Event Counter Register (0-30) */
185 	PMEVCNTR30_EL0 = PMEVCNTR0_EL0 + 30,
186 	PMCCNTR_EL0,	/* Cycle Counter Register */
187 	PMEVTYPER0_EL0,	/* Event Type Register (0-30) */
188 	PMEVTYPER30_EL0 = PMEVTYPER0_EL0 + 30,
189 	PMCCFILTR_EL0,	/* Cycle Count Filter Register */
190 	PMCNTENSET_EL0,	/* Count Enable Set Register */
191 	PMINTENSET_EL1,	/* Interrupt Enable Set Register */
192 	PMOVSSET_EL0,	/* Overflow Flag Status Set Register */
193 	PMUSERENR_EL0,	/* User Enable Register */
194 
195 	/* Pointer Authentication Registers in a strict increasing order. */
196 	APIAKEYLO_EL1,
197 	APIAKEYHI_EL1,
198 	APIBKEYLO_EL1,
199 	APIBKEYHI_EL1,
200 	APDAKEYLO_EL1,
201 	APDAKEYHI_EL1,
202 	APDBKEYLO_EL1,
203 	APDBKEYHI_EL1,
204 	APGAKEYLO_EL1,
205 	APGAKEYHI_EL1,
206 
207 	ELR_EL1,
208 	SP_EL1,
209 	SPSR_EL1,
210 
211 	CNTVOFF_EL2,
212 	CNTV_CVAL_EL0,
213 	CNTV_CTL_EL0,
214 	CNTP_CVAL_EL0,
215 	CNTP_CTL_EL0,
216 
217 	/* Memory Tagging Extension registers */
218 	RGSR_EL1,	/* Random Allocation Tag Seed Register */
219 	GCR_EL1,	/* Tag Control Register */
220 	TFSR_EL1,	/* Tag Fault Status Register (EL1) */
221 	TFSRE0_EL1,	/* Tag Fault Status Register (EL0) */
222 
223 	/* 32bit specific registers. Keep them at the end of the range */
224 	DACR32_EL2,	/* Domain Access Control Register */
225 	IFSR32_EL2,	/* Instruction Fault Status Register */
226 	FPEXC32_EL2,	/* Floating-Point Exception Control Register */
227 	DBGVCR32_EL2,	/* Debug Vector Catch Register */
228 
229 	NR_SYS_REGS	/* Nothing after this line! */
230 };
231 
232 struct kvm_cpu_context {
233 	struct user_pt_regs regs;	/* sp = sp_el0 */
234 
235 	u64	spsr_abt;
236 	u64	spsr_und;
237 	u64	spsr_irq;
238 	u64	spsr_fiq;
239 
240 	struct user_fpsimd_state fp_regs;
241 
242 	u64 sys_regs[NR_SYS_REGS];
243 
244 	struct kvm_vcpu *__hyp_running_vcpu;
245 };
246 
247 struct kvm_pmu_events {
248 	u32 events_host;
249 	u32 events_guest;
250 };
251 
252 struct kvm_host_data {
253 	struct kvm_cpu_context host_ctxt;
254 	struct kvm_pmu_events pmu_events;
255 };
256 
257 struct kvm_host_psci_config {
258 	/* PSCI version used by host. */
259 	u32 version;
260 
261 	/* Function IDs used by host if version is v0.1. */
262 	struct psci_0_1_function_ids function_ids_0_1;
263 
264 	bool psci_0_1_cpu_suspend_implemented;
265 	bool psci_0_1_cpu_on_implemented;
266 	bool psci_0_1_cpu_off_implemented;
267 	bool psci_0_1_migrate_implemented;
268 };
269 
270 extern struct kvm_host_psci_config kvm_nvhe_sym(kvm_host_psci_config);
271 #define kvm_host_psci_config CHOOSE_NVHE_SYM(kvm_host_psci_config)
272 
273 extern s64 kvm_nvhe_sym(hyp_physvirt_offset);
274 #define hyp_physvirt_offset CHOOSE_NVHE_SYM(hyp_physvirt_offset)
275 
276 extern u64 kvm_nvhe_sym(hyp_cpu_logical_map)[NR_CPUS];
277 #define hyp_cpu_logical_map CHOOSE_NVHE_SYM(hyp_cpu_logical_map)
278 
279 struct vcpu_reset_state {
280 	unsigned long	pc;
281 	unsigned long	r0;
282 	bool		be;
283 	bool		reset;
284 };
285 
286 struct kvm_vcpu_arch {
287 	struct kvm_cpu_context ctxt;
288 	void *sve_state;
289 	unsigned int sve_max_vl;
290 
291 	/* Stage 2 paging state used by the hardware on next switch */
292 	struct kvm_s2_mmu *hw_mmu;
293 
294 	/* Values of trap registers for the guest. */
295 	u64 hcr_el2;
296 	u64 mdcr_el2;
297 	u64 cptr_el2;
298 
299 	/* Values of trap registers for the host before guest entry. */
300 	u64 mdcr_el2_host;
301 
302 	/* Exception Information */
303 	struct kvm_vcpu_fault_info fault;
304 
305 	/* Miscellaneous vcpu state flags */
306 	u64 flags;
307 
308 	/*
309 	 * We maintain more than a single set of debug registers to support
310 	 * debugging the guest from the host and to maintain separate host and
311 	 * guest state during world switches. vcpu_debug_state are the debug
312 	 * registers of the vcpu as the guest sees them.  host_debug_state are
313 	 * the host registers which are saved and restored during
314 	 * world switches. external_debug_state contains the debug
315 	 * values we want to debug the guest. This is set via the
316 	 * KVM_SET_GUEST_DEBUG ioctl.
317 	 *
318 	 * debug_ptr points to the set of debug registers that should be loaded
319 	 * onto the hardware when running the guest.
320 	 */
321 	struct kvm_guest_debug_arch *debug_ptr;
322 	struct kvm_guest_debug_arch vcpu_debug_state;
323 	struct kvm_guest_debug_arch external_debug_state;
324 
325 	struct user_fpsimd_state *host_fpsimd_state;	/* hyp VA */
326 	struct task_struct *parent_task;
327 
328 	struct {
329 		/* {Break,watch}point registers */
330 		struct kvm_guest_debug_arch regs;
331 		/* Statistical profiling extension */
332 		u64 pmscr_el1;
333 		/* Self-hosted trace */
334 		u64 trfcr_el1;
335 	} host_debug_state;
336 
337 	/* VGIC state */
338 	struct vgic_cpu vgic_cpu;
339 	struct arch_timer_cpu timer_cpu;
340 	struct kvm_pmu pmu;
341 
342 	/*
343 	 * Anything that is not used directly from assembly code goes
344 	 * here.
345 	 */
346 
347 	/*
348 	 * Guest registers we preserve during guest debugging.
349 	 *
350 	 * These shadow registers are updated by the kvm_handle_sys_reg
351 	 * trap handler if the guest accesses or updates them while we
352 	 * are using guest debug.
353 	 */
354 	struct {
355 		u32	mdscr_el1;
356 	} guest_debug_preserved;
357 
358 	/* vcpu power-off state */
359 	bool power_off;
360 
361 	/* Don't run the guest (internal implementation need) */
362 	bool pause;
363 
364 	/* Cache some mmu pages needed inside spinlock regions */
365 	struct kvm_mmu_memory_cache mmu_page_cache;
366 
367 	/* Target CPU and feature flags */
368 	int target;
369 	DECLARE_BITMAP(features, KVM_VCPU_MAX_FEATURES);
370 
371 	/* Virtual SError ESR to restore when HCR_EL2.VSE is set */
372 	u64 vsesr_el2;
373 
374 	/* Additional reset state */
375 	struct vcpu_reset_state	reset_state;
376 
377 	/* True when deferrable sysregs are loaded on the physical CPU,
378 	 * see kvm_vcpu_load_sysregs_vhe and kvm_vcpu_put_sysregs_vhe. */
379 	bool sysregs_loaded_on_cpu;
380 
381 	/* Guest PV state */
382 	struct {
383 		u64 last_steal;
384 		gpa_t base;
385 	} steal;
386 };
387 
388 /* Pointer to the vcpu's SVE FFR for sve_{save,load}_state() */
389 #define vcpu_sve_pffr(vcpu) (kern_hyp_va((vcpu)->arch.sve_state) +	\
390 			     sve_ffr_offset((vcpu)->arch.sve_max_vl))
391 
392 #define vcpu_sve_max_vq(vcpu)	sve_vq_from_vl((vcpu)->arch.sve_max_vl)
393 
394 #define vcpu_sve_state_size(vcpu) ({					\
395 	size_t __size_ret;						\
396 	unsigned int __vcpu_vq;						\
397 									\
398 	if (WARN_ON(!sve_vl_valid((vcpu)->arch.sve_max_vl))) {		\
399 		__size_ret = 0;						\
400 	} else {							\
401 		__vcpu_vq = vcpu_sve_max_vq(vcpu);			\
402 		__size_ret = SVE_SIG_REGS_SIZE(__vcpu_vq);		\
403 	}								\
404 									\
405 	__size_ret;							\
406 })
407 
408 /* vcpu_arch flags field values: */
409 #define KVM_ARM64_DEBUG_DIRTY		(1 << 0)
410 #define KVM_ARM64_FP_ENABLED		(1 << 1) /* guest FP regs loaded */
411 #define KVM_ARM64_FP_HOST		(1 << 2) /* host FP regs loaded */
412 #define KVM_ARM64_HOST_SVE_ENABLED	(1 << 4) /* SVE enabled for EL0 */
413 #define KVM_ARM64_GUEST_HAS_SVE		(1 << 5) /* SVE exposed to guest */
414 #define KVM_ARM64_VCPU_SVE_FINALIZED	(1 << 6) /* SVE config completed */
415 #define KVM_ARM64_GUEST_HAS_PTRAUTH	(1 << 7) /* PTRAUTH exposed to guest */
416 #define KVM_ARM64_PENDING_EXCEPTION	(1 << 8) /* Exception pending */
417 /*
418  * Overlaps with KVM_ARM64_EXCEPT_MASK on purpose so that it can't be
419  * set together with an exception...
420  */
421 #define KVM_ARM64_INCREMENT_PC		(1 << 9) /* Increment PC */
422 #define KVM_ARM64_EXCEPT_MASK		(7 << 9) /* Target EL/MODE */
423 /*
424  * When KVM_ARM64_PENDING_EXCEPTION is set, KVM_ARM64_EXCEPT_MASK can
425  * take the following values:
426  *
427  * For AArch32 EL1:
428  */
429 #define KVM_ARM64_EXCEPT_AA32_UND	(0 << 9)
430 #define KVM_ARM64_EXCEPT_AA32_IABT	(1 << 9)
431 #define KVM_ARM64_EXCEPT_AA32_DABT	(2 << 9)
432 /* For AArch64: */
433 #define KVM_ARM64_EXCEPT_AA64_ELx_SYNC	(0 << 9)
434 #define KVM_ARM64_EXCEPT_AA64_ELx_IRQ	(1 << 9)
435 #define KVM_ARM64_EXCEPT_AA64_ELx_FIQ	(2 << 9)
436 #define KVM_ARM64_EXCEPT_AA64_ELx_SERR	(3 << 9)
437 #define KVM_ARM64_EXCEPT_AA64_EL1	(0 << 11)
438 #define KVM_ARM64_EXCEPT_AA64_EL2	(1 << 11)
439 
440 #define KVM_ARM64_DEBUG_STATE_SAVE_SPE	(1 << 12) /* Save SPE context if active  */
441 #define KVM_ARM64_DEBUG_STATE_SAVE_TRBE	(1 << 13) /* Save TRBE context if active  */
442 #define KVM_ARM64_FP_FOREIGN_FPSTATE	(1 << 14)
443 #define KVM_ARM64_ON_UNSUPPORTED_CPU	(1 << 15) /* Physical CPU not in supported_cpus */
444 
445 #define KVM_GUESTDBG_VALID_MASK (KVM_GUESTDBG_ENABLE | \
446 				 KVM_GUESTDBG_USE_SW_BP | \
447 				 KVM_GUESTDBG_USE_HW | \
448 				 KVM_GUESTDBG_SINGLESTEP)
449 
450 #define vcpu_has_sve(vcpu) (system_supports_sve() &&			\
451 			    ((vcpu)->arch.flags & KVM_ARM64_GUEST_HAS_SVE))
452 
453 #ifdef CONFIG_ARM64_PTR_AUTH
454 #define vcpu_has_ptrauth(vcpu)						\
455 	((cpus_have_final_cap(ARM64_HAS_ADDRESS_AUTH) ||		\
456 	  cpus_have_final_cap(ARM64_HAS_GENERIC_AUTH)) &&		\
457 	 (vcpu)->arch.flags & KVM_ARM64_GUEST_HAS_PTRAUTH)
458 #else
459 #define vcpu_has_ptrauth(vcpu)		false
460 #endif
461 
462 #define vcpu_on_unsupported_cpu(vcpu)					\
463 	((vcpu)->arch.flags & KVM_ARM64_ON_UNSUPPORTED_CPU)
464 
465 #define vcpu_set_on_unsupported_cpu(vcpu)				\
466 	((vcpu)->arch.flags |= KVM_ARM64_ON_UNSUPPORTED_CPU)
467 
468 #define vcpu_clear_on_unsupported_cpu(vcpu)				\
469 	((vcpu)->arch.flags &= ~KVM_ARM64_ON_UNSUPPORTED_CPU)
470 
471 #define vcpu_gp_regs(v)		(&(v)->arch.ctxt.regs)
472 
473 /*
474  * Only use __vcpu_sys_reg/ctxt_sys_reg if you know you want the
475  * memory backed version of a register, and not the one most recently
476  * accessed by a running VCPU.  For example, for userspace access or
477  * for system registers that are never context switched, but only
478  * emulated.
479  */
480 #define __ctxt_sys_reg(c,r)	(&(c)->sys_regs[(r)])
481 
482 #define ctxt_sys_reg(c,r)	(*__ctxt_sys_reg(c,r))
483 
484 #define __vcpu_sys_reg(v,r)	(ctxt_sys_reg(&(v)->arch.ctxt, (r)))
485 
486 u64 vcpu_read_sys_reg(const struct kvm_vcpu *vcpu, int reg);
487 void vcpu_write_sys_reg(struct kvm_vcpu *vcpu, u64 val, int reg);
488 
489 static inline bool __vcpu_read_sys_reg_from_cpu(int reg, u64 *val)
490 {
491 	/*
492 	 * *** VHE ONLY ***
493 	 *
494 	 * System registers listed in the switch are not saved on every
495 	 * exit from the guest but are only saved on vcpu_put.
496 	 *
497 	 * Note that MPIDR_EL1 for the guest is set by KVM via VMPIDR_EL2 but
498 	 * should never be listed below, because the guest cannot modify its
499 	 * own MPIDR_EL1 and MPIDR_EL1 is accessed for VCPU A from VCPU B's
500 	 * thread when emulating cross-VCPU communication.
501 	 */
502 	if (!has_vhe())
503 		return false;
504 
505 	switch (reg) {
506 	case CSSELR_EL1:	*val = read_sysreg_s(SYS_CSSELR_EL1);	break;
507 	case SCTLR_EL1:		*val = read_sysreg_s(SYS_SCTLR_EL12);	break;
508 	case CPACR_EL1:		*val = read_sysreg_s(SYS_CPACR_EL12);	break;
509 	case TTBR0_EL1:		*val = read_sysreg_s(SYS_TTBR0_EL12);	break;
510 	case TTBR1_EL1:		*val = read_sysreg_s(SYS_TTBR1_EL12);	break;
511 	case TCR_EL1:		*val = read_sysreg_s(SYS_TCR_EL12);	break;
512 	case ESR_EL1:		*val = read_sysreg_s(SYS_ESR_EL12);	break;
513 	case AFSR0_EL1:		*val = read_sysreg_s(SYS_AFSR0_EL12);	break;
514 	case AFSR1_EL1:		*val = read_sysreg_s(SYS_AFSR1_EL12);	break;
515 	case FAR_EL1:		*val = read_sysreg_s(SYS_FAR_EL12);	break;
516 	case MAIR_EL1:		*val = read_sysreg_s(SYS_MAIR_EL12);	break;
517 	case VBAR_EL1:		*val = read_sysreg_s(SYS_VBAR_EL12);	break;
518 	case CONTEXTIDR_EL1:	*val = read_sysreg_s(SYS_CONTEXTIDR_EL12);break;
519 	case TPIDR_EL0:		*val = read_sysreg_s(SYS_TPIDR_EL0);	break;
520 	case TPIDRRO_EL0:	*val = read_sysreg_s(SYS_TPIDRRO_EL0);	break;
521 	case TPIDR_EL1:		*val = read_sysreg_s(SYS_TPIDR_EL1);	break;
522 	case AMAIR_EL1:		*val = read_sysreg_s(SYS_AMAIR_EL12);	break;
523 	case CNTKCTL_EL1:	*val = read_sysreg_s(SYS_CNTKCTL_EL12);	break;
524 	case ELR_EL1:		*val = read_sysreg_s(SYS_ELR_EL12);	break;
525 	case PAR_EL1:		*val = read_sysreg_par();		break;
526 	case DACR32_EL2:	*val = read_sysreg_s(SYS_DACR32_EL2);	break;
527 	case IFSR32_EL2:	*val = read_sysreg_s(SYS_IFSR32_EL2);	break;
528 	case DBGVCR32_EL2:	*val = read_sysreg_s(SYS_DBGVCR32_EL2);	break;
529 	default:		return false;
530 	}
531 
532 	return true;
533 }
534 
535 static inline bool __vcpu_write_sys_reg_to_cpu(u64 val, int reg)
536 {
537 	/*
538 	 * *** VHE ONLY ***
539 	 *
540 	 * System registers listed in the switch are not restored on every
541 	 * entry to the guest but are only restored on vcpu_load.
542 	 *
543 	 * Note that MPIDR_EL1 for the guest is set by KVM via VMPIDR_EL2 but
544 	 * should never be listed below, because the MPIDR should only be set
545 	 * once, before running the VCPU, and never changed later.
546 	 */
547 	if (!has_vhe())
548 		return false;
549 
550 	switch (reg) {
551 	case CSSELR_EL1:	write_sysreg_s(val, SYS_CSSELR_EL1);	break;
552 	case SCTLR_EL1:		write_sysreg_s(val, SYS_SCTLR_EL12);	break;
553 	case CPACR_EL1:		write_sysreg_s(val, SYS_CPACR_EL12);	break;
554 	case TTBR0_EL1:		write_sysreg_s(val, SYS_TTBR0_EL12);	break;
555 	case TTBR1_EL1:		write_sysreg_s(val, SYS_TTBR1_EL12);	break;
556 	case TCR_EL1:		write_sysreg_s(val, SYS_TCR_EL12);	break;
557 	case ESR_EL1:		write_sysreg_s(val, SYS_ESR_EL12);	break;
558 	case AFSR0_EL1:		write_sysreg_s(val, SYS_AFSR0_EL12);	break;
559 	case AFSR1_EL1:		write_sysreg_s(val, SYS_AFSR1_EL12);	break;
560 	case FAR_EL1:		write_sysreg_s(val, SYS_FAR_EL12);	break;
561 	case MAIR_EL1:		write_sysreg_s(val, SYS_MAIR_EL12);	break;
562 	case VBAR_EL1:		write_sysreg_s(val, SYS_VBAR_EL12);	break;
563 	case CONTEXTIDR_EL1:	write_sysreg_s(val, SYS_CONTEXTIDR_EL12);break;
564 	case TPIDR_EL0:		write_sysreg_s(val, SYS_TPIDR_EL0);	break;
565 	case TPIDRRO_EL0:	write_sysreg_s(val, SYS_TPIDRRO_EL0);	break;
566 	case TPIDR_EL1:		write_sysreg_s(val, SYS_TPIDR_EL1);	break;
567 	case AMAIR_EL1:		write_sysreg_s(val, SYS_AMAIR_EL12);	break;
568 	case CNTKCTL_EL1:	write_sysreg_s(val, SYS_CNTKCTL_EL12);	break;
569 	case ELR_EL1:		write_sysreg_s(val, SYS_ELR_EL12);	break;
570 	case PAR_EL1:		write_sysreg_s(val, SYS_PAR_EL1);	break;
571 	case DACR32_EL2:	write_sysreg_s(val, SYS_DACR32_EL2);	break;
572 	case IFSR32_EL2:	write_sysreg_s(val, SYS_IFSR32_EL2);	break;
573 	case DBGVCR32_EL2:	write_sysreg_s(val, SYS_DBGVCR32_EL2);	break;
574 	default:		return false;
575 	}
576 
577 	return true;
578 }
579 
580 struct kvm_vm_stat {
581 	struct kvm_vm_stat_generic generic;
582 };
583 
584 struct kvm_vcpu_stat {
585 	struct kvm_vcpu_stat_generic generic;
586 	u64 hvc_exit_stat;
587 	u64 wfe_exit_stat;
588 	u64 wfi_exit_stat;
589 	u64 mmio_exit_user;
590 	u64 mmio_exit_kernel;
591 	u64 signal_exits;
592 	u64 exits;
593 };
594 
595 void kvm_vcpu_preferred_target(struct kvm_vcpu_init *init);
596 unsigned long kvm_arm_num_regs(struct kvm_vcpu *vcpu);
597 int kvm_arm_copy_reg_indices(struct kvm_vcpu *vcpu, u64 __user *indices);
598 int kvm_arm_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
599 int kvm_arm_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
600 
601 unsigned long kvm_arm_num_sys_reg_descs(struct kvm_vcpu *vcpu);
602 int kvm_arm_copy_sys_reg_indices(struct kvm_vcpu *vcpu, u64 __user *uindices);
603 int kvm_arm_sys_reg_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *);
604 int kvm_arm_sys_reg_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *);
605 
606 int __kvm_arm_vcpu_get_events(struct kvm_vcpu *vcpu,
607 			      struct kvm_vcpu_events *events);
608 
609 int __kvm_arm_vcpu_set_events(struct kvm_vcpu *vcpu,
610 			      struct kvm_vcpu_events *events);
611 
612 #define KVM_ARCH_WANT_MMU_NOTIFIER
613 
614 void kvm_arm_halt_guest(struct kvm *kvm);
615 void kvm_arm_resume_guest(struct kvm *kvm);
616 
617 #define vcpu_has_run_once(vcpu)	!!rcu_access_pointer((vcpu)->pid)
618 
619 #ifndef __KVM_NVHE_HYPERVISOR__
620 #define kvm_call_hyp_nvhe(f, ...)						\
621 	({								\
622 		struct arm_smccc_res res;				\
623 									\
624 		arm_smccc_1_1_hvc(KVM_HOST_SMCCC_FUNC(f),		\
625 				  ##__VA_ARGS__, &res);			\
626 		WARN_ON(res.a0 != SMCCC_RET_SUCCESS);			\
627 									\
628 		res.a1;							\
629 	})
630 
631 /*
632  * The couple of isb() below are there to guarantee the same behaviour
633  * on VHE as on !VHE, where the eret to EL1 acts as a context
634  * synchronization event.
635  */
636 #define kvm_call_hyp(f, ...)						\
637 	do {								\
638 		if (has_vhe()) {					\
639 			f(__VA_ARGS__);					\
640 			isb();						\
641 		} else {						\
642 			kvm_call_hyp_nvhe(f, ##__VA_ARGS__);		\
643 		}							\
644 	} while(0)
645 
646 #define kvm_call_hyp_ret(f, ...)					\
647 	({								\
648 		typeof(f(__VA_ARGS__)) ret;				\
649 									\
650 		if (has_vhe()) {					\
651 			ret = f(__VA_ARGS__);				\
652 			isb();						\
653 		} else {						\
654 			ret = kvm_call_hyp_nvhe(f, ##__VA_ARGS__);	\
655 		}							\
656 									\
657 		ret;							\
658 	})
659 #else /* __KVM_NVHE_HYPERVISOR__ */
660 #define kvm_call_hyp(f, ...) f(__VA_ARGS__)
661 #define kvm_call_hyp_ret(f, ...) f(__VA_ARGS__)
662 #define kvm_call_hyp_nvhe(f, ...) f(__VA_ARGS__)
663 #endif /* __KVM_NVHE_HYPERVISOR__ */
664 
665 void force_vm_exit(const cpumask_t *mask);
666 
667 int handle_exit(struct kvm_vcpu *vcpu, int exception_index);
668 void handle_exit_early(struct kvm_vcpu *vcpu, int exception_index);
669 
670 int kvm_handle_cp14_load_store(struct kvm_vcpu *vcpu);
671 int kvm_handle_cp14_32(struct kvm_vcpu *vcpu);
672 int kvm_handle_cp14_64(struct kvm_vcpu *vcpu);
673 int kvm_handle_cp15_32(struct kvm_vcpu *vcpu);
674 int kvm_handle_cp15_64(struct kvm_vcpu *vcpu);
675 int kvm_handle_sys_reg(struct kvm_vcpu *vcpu);
676 
677 void kvm_reset_sys_regs(struct kvm_vcpu *vcpu);
678 
679 void kvm_sys_reg_table_init(void);
680 
681 /* MMIO helpers */
682 void kvm_mmio_write_buf(void *buf, unsigned int len, unsigned long data);
683 unsigned long kvm_mmio_read_buf(const void *buf, unsigned int len);
684 
685 int kvm_handle_mmio_return(struct kvm_vcpu *vcpu);
686 int io_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa);
687 
688 /*
689  * Returns true if a Performance Monitoring Interrupt (PMI), a.k.a. perf event,
690  * arrived in guest context.  For arm64, any event that arrives while a vCPU is
691  * loaded is considered to be "in guest".
692  */
693 static inline bool kvm_arch_pmi_in_guest(struct kvm_vcpu *vcpu)
694 {
695 	return IS_ENABLED(CONFIG_GUEST_PERF_EVENTS) && !!vcpu;
696 }
697 
698 long kvm_hypercall_pv_features(struct kvm_vcpu *vcpu);
699 gpa_t kvm_init_stolen_time(struct kvm_vcpu *vcpu);
700 void kvm_update_stolen_time(struct kvm_vcpu *vcpu);
701 
702 bool kvm_arm_pvtime_supported(void);
703 int kvm_arm_pvtime_set_attr(struct kvm_vcpu *vcpu,
704 			    struct kvm_device_attr *attr);
705 int kvm_arm_pvtime_get_attr(struct kvm_vcpu *vcpu,
706 			    struct kvm_device_attr *attr);
707 int kvm_arm_pvtime_has_attr(struct kvm_vcpu *vcpu,
708 			    struct kvm_device_attr *attr);
709 
710 extern unsigned int kvm_arm_vmid_bits;
711 int kvm_arm_vmid_alloc_init(void);
712 void kvm_arm_vmid_alloc_free(void);
713 void kvm_arm_vmid_update(struct kvm_vmid *kvm_vmid);
714 void kvm_arm_vmid_clear_active(void);
715 
716 static inline void kvm_arm_pvtime_vcpu_init(struct kvm_vcpu_arch *vcpu_arch)
717 {
718 	vcpu_arch->steal.base = GPA_INVALID;
719 }
720 
721 static inline bool kvm_arm_is_pvtime_enabled(struct kvm_vcpu_arch *vcpu_arch)
722 {
723 	return (vcpu_arch->steal.base != GPA_INVALID);
724 }
725 
726 void kvm_set_sei_esr(struct kvm_vcpu *vcpu, u64 syndrome);
727 
728 struct kvm_vcpu *kvm_mpidr_to_vcpu(struct kvm *kvm, unsigned long mpidr);
729 
730 DECLARE_KVM_HYP_PER_CPU(struct kvm_host_data, kvm_host_data);
731 
732 static inline void kvm_init_host_cpu_context(struct kvm_cpu_context *cpu_ctxt)
733 {
734 	/* The host's MPIDR is immutable, so let's set it up at boot time */
735 	ctxt_sys_reg(cpu_ctxt, MPIDR_EL1) = read_cpuid_mpidr();
736 }
737 
738 static inline bool kvm_system_needs_idmapped_vectors(void)
739 {
740 	return cpus_have_const_cap(ARM64_SPECTRE_V3A);
741 }
742 
743 void kvm_arm_vcpu_ptrauth_trap(struct kvm_vcpu *vcpu);
744 
745 static inline void kvm_arch_hardware_unsetup(void) {}
746 static inline void kvm_arch_sync_events(struct kvm *kvm) {}
747 static inline void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu) {}
748 
749 void kvm_arm_init_debug(void);
750 void kvm_arm_vcpu_init_debug(struct kvm_vcpu *vcpu);
751 void kvm_arm_setup_debug(struct kvm_vcpu *vcpu);
752 void kvm_arm_clear_debug(struct kvm_vcpu *vcpu);
753 void kvm_arm_reset_debug_ptr(struct kvm_vcpu *vcpu);
754 
755 #define kvm_vcpu_os_lock_enabled(vcpu)		\
756 	(!!(__vcpu_sys_reg(vcpu, OSLSR_EL1) & SYS_OSLSR_OSLK))
757 
758 int kvm_arm_vcpu_arch_set_attr(struct kvm_vcpu *vcpu,
759 			       struct kvm_device_attr *attr);
760 int kvm_arm_vcpu_arch_get_attr(struct kvm_vcpu *vcpu,
761 			       struct kvm_device_attr *attr);
762 int kvm_arm_vcpu_arch_has_attr(struct kvm_vcpu *vcpu,
763 			       struct kvm_device_attr *attr);
764 
765 long kvm_vm_ioctl_mte_copy_tags(struct kvm *kvm,
766 				struct kvm_arm_copy_mte_tags *copy_tags);
767 
768 /* Guest/host FPSIMD coordination helpers */
769 int kvm_arch_vcpu_run_map_fp(struct kvm_vcpu *vcpu);
770 void kvm_arch_vcpu_load_fp(struct kvm_vcpu *vcpu);
771 void kvm_arch_vcpu_ctxflush_fp(struct kvm_vcpu *vcpu);
772 void kvm_arch_vcpu_ctxsync_fp(struct kvm_vcpu *vcpu);
773 void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu);
774 void kvm_vcpu_unshare_task_fp(struct kvm_vcpu *vcpu);
775 
776 static inline bool kvm_pmu_counter_deferred(struct perf_event_attr *attr)
777 {
778 	return (!has_vhe() && attr->exclude_host);
779 }
780 
781 /* Flags for host debug state */
782 void kvm_arch_vcpu_load_debug_state_flags(struct kvm_vcpu *vcpu);
783 void kvm_arch_vcpu_put_debug_state_flags(struct kvm_vcpu *vcpu);
784 
785 #ifdef CONFIG_KVM
786 void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr);
787 void kvm_clr_pmu_events(u32 clr);
788 
789 void kvm_vcpu_pmu_restore_guest(struct kvm_vcpu *vcpu);
790 void kvm_vcpu_pmu_restore_host(struct kvm_vcpu *vcpu);
791 #else
792 static inline void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr) {}
793 static inline void kvm_clr_pmu_events(u32 clr) {}
794 #endif
795 
796 void kvm_vcpu_load_sysregs_vhe(struct kvm_vcpu *vcpu);
797 void kvm_vcpu_put_sysregs_vhe(struct kvm_vcpu *vcpu);
798 
799 int kvm_set_ipa_limit(void);
800 
801 #define __KVM_HAVE_ARCH_VM_ALLOC
802 struct kvm *kvm_arch_alloc_vm(void);
803 
804 int kvm_arm_setup_stage2(struct kvm *kvm, unsigned long type);
805 
806 static inline bool kvm_vm_is_protected(struct kvm *kvm)
807 {
808 	return false;
809 }
810 
811 void kvm_init_protected_traps(struct kvm_vcpu *vcpu);
812 
813 int kvm_arm_vcpu_finalize(struct kvm_vcpu *vcpu, int feature);
814 bool kvm_arm_vcpu_is_finalized(struct kvm_vcpu *vcpu);
815 
816 #define kvm_arm_vcpu_sve_finalized(vcpu) \
817 	((vcpu)->arch.flags & KVM_ARM64_VCPU_SVE_FINALIZED)
818 
819 #define kvm_has_mte(kvm)					\
820 	(system_supports_mte() &&				\
821 	 test_bit(KVM_ARCH_FLAG_MTE_ENABLED, &(kvm)->arch.flags))
822 #define kvm_vcpu_has_pmu(vcpu)					\
823 	(test_bit(KVM_ARM_VCPU_PMU_V3, (vcpu)->arch.features))
824 
825 int kvm_trng_call(struct kvm_vcpu *vcpu);
826 #ifdef CONFIG_KVM
827 extern phys_addr_t hyp_mem_base;
828 extern phys_addr_t hyp_mem_size;
829 void __init kvm_hyp_reserve(void);
830 #else
831 static inline void kvm_hyp_reserve(void) { }
832 #endif
833 
834 #endif /* __ARM64_KVM_HOST_H__ */
835