xref: /openbmc/linux/arch/arm64/include/asm/kvm_host.h (revision 46b18782)
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 /*
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 	KVM_MODE_NONE,
61 };
62 enum kvm_mode kvm_get_mode(void);
63 
64 DECLARE_STATIC_KEY_FALSE(userspace_irqchip_in_use);
65 
66 extern unsigned int kvm_sve_max_vl;
67 int kvm_arm_init_sve(void);
68 
69 u32 __attribute_const__ kvm_target_cpu(void);
70 int kvm_reset_vcpu(struct kvm_vcpu *vcpu);
71 void kvm_arm_vcpu_destroy(struct kvm_vcpu *vcpu);
72 
73 struct kvm_vmid {
74 	/* The VMID generation used for the virt. memory system */
75 	u64    vmid_gen;
76 	u32    vmid;
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 	bool return_nisv_io_abort_to_user;
126 
127 	/*
128 	 * VM-wide PMU filter, implemented as a bitmap and big enough for
129 	 * up to 2^10 events (ARMv8.0) or 2^16 events (ARMv8.1+).
130 	 */
131 	unsigned long *pmu_filter;
132 	struct arm_pmu *arm_pmu;
133 
134 	u8 pfr0_csv2;
135 	u8 pfr0_csv3;
136 
137 	/* Memory Tagging Extension enabled for the guest */
138 	bool mte_enabled;
139 	bool ran_once;
140 };
141 
142 struct kvm_vcpu_fault_info {
143 	u32 esr_el2;		/* Hyp Syndrom Register */
144 	u64 far_el2;		/* Hyp Fault Address Register */
145 	u64 hpfar_el2;		/* Hyp IPA Fault Address Register */
146 	u64 disr_el1;		/* Deferred [SError] Status Register */
147 };
148 
149 enum vcpu_sysreg {
150 	__INVALID_SYSREG__,   /* 0 is reserved as an invalid value */
151 	MPIDR_EL1,	/* MultiProcessor Affinity Register */
152 	CSSELR_EL1,	/* Cache Size Selection Register */
153 	SCTLR_EL1,	/* System Control Register */
154 	ACTLR_EL1,	/* Auxiliary Control Register */
155 	CPACR_EL1,	/* Coprocessor Access Control */
156 	ZCR_EL1,	/* SVE Control */
157 	TTBR0_EL1,	/* Translation Table Base Register 0 */
158 	TTBR1_EL1,	/* Translation Table Base Register 1 */
159 	TCR_EL1,	/* Translation Control Register */
160 	ESR_EL1,	/* Exception Syndrome Register */
161 	AFSR0_EL1,	/* Auxiliary Fault Status Register 0 */
162 	AFSR1_EL1,	/* Auxiliary Fault Status Register 1 */
163 	FAR_EL1,	/* Fault Address Register */
164 	MAIR_EL1,	/* Memory Attribute Indirection Register */
165 	VBAR_EL1,	/* Vector Base Address Register */
166 	CONTEXTIDR_EL1,	/* Context ID Register */
167 	TPIDR_EL0,	/* Thread ID, User R/W */
168 	TPIDRRO_EL0,	/* Thread ID, User R/O */
169 	TPIDR_EL1,	/* Thread ID, Privileged */
170 	AMAIR_EL1,	/* Aux Memory Attribute Indirection Register */
171 	CNTKCTL_EL1,	/* Timer Control Register (EL1) */
172 	PAR_EL1,	/* Physical Address Register */
173 	MDSCR_EL1,	/* Monitor Debug System Control Register */
174 	MDCCINT_EL1,	/* Monitor Debug Comms Channel Interrupt Enable Reg */
175 	DISR_EL1,	/* Deferred Interrupt Status Register */
176 
177 	/* Performance Monitors Registers */
178 	PMCR_EL0,	/* Control Register */
179 	PMSELR_EL0,	/* Event Counter Selection Register */
180 	PMEVCNTR0_EL0,	/* Event Counter Register (0-30) */
181 	PMEVCNTR30_EL0 = PMEVCNTR0_EL0 + 30,
182 	PMCCNTR_EL0,	/* Cycle Counter Register */
183 	PMEVTYPER0_EL0,	/* Event Type Register (0-30) */
184 	PMEVTYPER30_EL0 = PMEVTYPER0_EL0 + 30,
185 	PMCCFILTR_EL0,	/* Cycle Count Filter Register */
186 	PMCNTENSET_EL0,	/* Count Enable Set Register */
187 	PMINTENSET_EL1,	/* Interrupt Enable Set Register */
188 	PMOVSSET_EL0,	/* Overflow Flag Status Set Register */
189 	PMUSERENR_EL0,	/* User Enable Register */
190 
191 	/* Pointer Authentication Registers in a strict increasing order. */
192 	APIAKEYLO_EL1,
193 	APIAKEYHI_EL1,
194 	APIBKEYLO_EL1,
195 	APIBKEYHI_EL1,
196 	APDAKEYLO_EL1,
197 	APDAKEYHI_EL1,
198 	APDBKEYLO_EL1,
199 	APDBKEYHI_EL1,
200 	APGAKEYLO_EL1,
201 	APGAKEYHI_EL1,
202 
203 	ELR_EL1,
204 	SP_EL1,
205 	SPSR_EL1,
206 
207 	CNTVOFF_EL2,
208 	CNTV_CVAL_EL0,
209 	CNTV_CTL_EL0,
210 	CNTP_CVAL_EL0,
211 	CNTP_CTL_EL0,
212 
213 	/* Memory Tagging Extension registers */
214 	RGSR_EL1,	/* Random Allocation Tag Seed Register */
215 	GCR_EL1,	/* Tag Control Register */
216 	TFSR_EL1,	/* Tag Fault Status Register (EL1) */
217 	TFSRE0_EL1,	/* Tag Fault Status Register (EL0) */
218 
219 	/* 32bit specific registers. Keep them at the end of the range */
220 	DACR32_EL2,	/* Domain Access Control Register */
221 	IFSR32_EL2,	/* Instruction Fault Status Register */
222 	FPEXC32_EL2,	/* Floating-Point Exception Control Register */
223 	DBGVCR32_EL2,	/* Debug Vector Catch Register */
224 
225 	NR_SYS_REGS	/* Nothing after this line! */
226 };
227 
228 struct kvm_cpu_context {
229 	struct user_pt_regs regs;	/* sp = sp_el0 */
230 
231 	u64	spsr_abt;
232 	u64	spsr_und;
233 	u64	spsr_irq;
234 	u64	spsr_fiq;
235 
236 	struct user_fpsimd_state fp_regs;
237 
238 	u64 sys_regs[NR_SYS_REGS];
239 
240 	struct kvm_vcpu *__hyp_running_vcpu;
241 };
242 
243 struct kvm_pmu_events {
244 	u32 events_host;
245 	u32 events_guest;
246 };
247 
248 struct kvm_host_data {
249 	struct kvm_cpu_context host_ctxt;
250 	struct kvm_pmu_events pmu_events;
251 };
252 
253 struct kvm_host_psci_config {
254 	/* PSCI version used by host. */
255 	u32 version;
256 
257 	/* Function IDs used by host if version is v0.1. */
258 	struct psci_0_1_function_ids function_ids_0_1;
259 
260 	bool psci_0_1_cpu_suspend_implemented;
261 	bool psci_0_1_cpu_on_implemented;
262 	bool psci_0_1_cpu_off_implemented;
263 	bool psci_0_1_migrate_implemented;
264 };
265 
266 extern struct kvm_host_psci_config kvm_nvhe_sym(kvm_host_psci_config);
267 #define kvm_host_psci_config CHOOSE_NVHE_SYM(kvm_host_psci_config)
268 
269 extern s64 kvm_nvhe_sym(hyp_physvirt_offset);
270 #define hyp_physvirt_offset CHOOSE_NVHE_SYM(hyp_physvirt_offset)
271 
272 extern u64 kvm_nvhe_sym(hyp_cpu_logical_map)[NR_CPUS];
273 #define hyp_cpu_logical_map CHOOSE_NVHE_SYM(hyp_cpu_logical_map)
274 
275 struct vcpu_reset_state {
276 	unsigned long	pc;
277 	unsigned long	r0;
278 	bool		be;
279 	bool		reset;
280 };
281 
282 struct kvm_vcpu_arch {
283 	struct kvm_cpu_context ctxt;
284 	void *sve_state;
285 	unsigned int sve_max_vl;
286 
287 	/* Stage 2 paging state used by the hardware on next switch */
288 	struct kvm_s2_mmu *hw_mmu;
289 
290 	/* Values of trap registers for the guest. */
291 	u64 hcr_el2;
292 	u64 mdcr_el2;
293 	u64 cptr_el2;
294 
295 	/* Values of trap registers for the host before guest entry. */
296 	u64 mdcr_el2_host;
297 
298 	/* Exception Information */
299 	struct kvm_vcpu_fault_info fault;
300 
301 	/* Miscellaneous vcpu state flags */
302 	u64 flags;
303 
304 	/*
305 	 * We maintain more than a single set of debug registers to support
306 	 * debugging the guest from the host and to maintain separate host and
307 	 * guest state during world switches. vcpu_debug_state are the debug
308 	 * registers of the vcpu as the guest sees them.  host_debug_state are
309 	 * the host registers which are saved and restored during
310 	 * world switches. external_debug_state contains the debug
311 	 * values we want to debug the guest. This is set via the
312 	 * KVM_SET_GUEST_DEBUG ioctl.
313 	 *
314 	 * debug_ptr points to the set of debug registers that should be loaded
315 	 * onto the hardware when running the guest.
316 	 */
317 	struct kvm_guest_debug_arch *debug_ptr;
318 	struct kvm_guest_debug_arch vcpu_debug_state;
319 	struct kvm_guest_debug_arch external_debug_state;
320 
321 	struct user_fpsimd_state *host_fpsimd_state;	/* hyp VA */
322 	struct task_struct *parent_task;
323 
324 	struct {
325 		/* {Break,watch}point registers */
326 		struct kvm_guest_debug_arch regs;
327 		/* Statistical profiling extension */
328 		u64 pmscr_el1;
329 		/* Self-hosted trace */
330 		u64 trfcr_el1;
331 	} host_debug_state;
332 
333 	/* VGIC state */
334 	struct vgic_cpu vgic_cpu;
335 	struct arch_timer_cpu timer_cpu;
336 	struct kvm_pmu pmu;
337 
338 	/*
339 	 * Anything that is not used directly from assembly code goes
340 	 * here.
341 	 */
342 
343 	/*
344 	 * Guest registers we preserve during guest debugging.
345 	 *
346 	 * These shadow registers are updated by the kvm_handle_sys_reg
347 	 * trap handler if the guest accesses or updates them while we
348 	 * are using guest debug.
349 	 */
350 	struct {
351 		u32	mdscr_el1;
352 	} guest_debug_preserved;
353 
354 	/* vcpu power-off state */
355 	bool power_off;
356 
357 	/* Don't run the guest (internal implementation need) */
358 	bool pause;
359 
360 	/* Cache some mmu pages needed inside spinlock regions */
361 	struct kvm_mmu_memory_cache mmu_page_cache;
362 
363 	/* Target CPU and feature flags */
364 	int target;
365 	DECLARE_BITMAP(features, KVM_VCPU_MAX_FEATURES);
366 
367 	/* Virtual SError ESR to restore when HCR_EL2.VSE is set */
368 	u64 vsesr_el2;
369 
370 	/* Additional reset state */
371 	struct vcpu_reset_state	reset_state;
372 
373 	/* True when deferrable sysregs are loaded on the physical CPU,
374 	 * see kvm_vcpu_load_sysregs_vhe and kvm_vcpu_put_sysregs_vhe. */
375 	bool sysregs_loaded_on_cpu;
376 
377 	/* Guest PV state */
378 	struct {
379 		u64 last_steal;
380 		gpa_t base;
381 	} steal;
382 };
383 
384 /* Pointer to the vcpu's SVE FFR for sve_{save,load}_state() */
385 #define vcpu_sve_pffr(vcpu) (kern_hyp_va((vcpu)->arch.sve_state) +	\
386 			     sve_ffr_offset((vcpu)->arch.sve_max_vl))
387 
388 #define vcpu_sve_max_vq(vcpu)	sve_vq_from_vl((vcpu)->arch.sve_max_vl)
389 
390 #define vcpu_sve_state_size(vcpu) ({					\
391 	size_t __size_ret;						\
392 	unsigned int __vcpu_vq;						\
393 									\
394 	if (WARN_ON(!sve_vl_valid((vcpu)->arch.sve_max_vl))) {		\
395 		__size_ret = 0;						\
396 	} else {							\
397 		__vcpu_vq = vcpu_sve_max_vq(vcpu);			\
398 		__size_ret = SVE_SIG_REGS_SIZE(__vcpu_vq);		\
399 	}								\
400 									\
401 	__size_ret;							\
402 })
403 
404 /* vcpu_arch flags field values: */
405 #define KVM_ARM64_DEBUG_DIRTY		(1 << 0)
406 #define KVM_ARM64_FP_ENABLED		(1 << 1) /* guest FP regs loaded */
407 #define KVM_ARM64_FP_HOST		(1 << 2) /* host FP regs loaded */
408 #define KVM_ARM64_HOST_SVE_ENABLED	(1 << 4) /* SVE enabled for EL0 */
409 #define KVM_ARM64_GUEST_HAS_SVE		(1 << 5) /* SVE exposed to guest */
410 #define KVM_ARM64_VCPU_SVE_FINALIZED	(1 << 6) /* SVE config completed */
411 #define KVM_ARM64_GUEST_HAS_PTRAUTH	(1 << 7) /* PTRAUTH exposed to guest */
412 #define KVM_ARM64_PENDING_EXCEPTION	(1 << 8) /* Exception pending */
413 /*
414  * Overlaps with KVM_ARM64_EXCEPT_MASK on purpose so that it can't be
415  * set together with an exception...
416  */
417 #define KVM_ARM64_INCREMENT_PC		(1 << 9) /* Increment PC */
418 #define KVM_ARM64_EXCEPT_MASK		(7 << 9) /* Target EL/MODE */
419 /*
420  * When KVM_ARM64_PENDING_EXCEPTION is set, KVM_ARM64_EXCEPT_MASK can
421  * take the following values:
422  *
423  * For AArch32 EL1:
424  */
425 #define KVM_ARM64_EXCEPT_AA32_UND	(0 << 9)
426 #define KVM_ARM64_EXCEPT_AA32_IABT	(1 << 9)
427 #define KVM_ARM64_EXCEPT_AA32_DABT	(2 << 9)
428 /* For AArch64: */
429 #define KVM_ARM64_EXCEPT_AA64_ELx_SYNC	(0 << 9)
430 #define KVM_ARM64_EXCEPT_AA64_ELx_IRQ	(1 << 9)
431 #define KVM_ARM64_EXCEPT_AA64_ELx_FIQ	(2 << 9)
432 #define KVM_ARM64_EXCEPT_AA64_ELx_SERR	(3 << 9)
433 #define KVM_ARM64_EXCEPT_AA64_EL1	(0 << 11)
434 #define KVM_ARM64_EXCEPT_AA64_EL2	(1 << 11)
435 
436 #define KVM_ARM64_DEBUG_STATE_SAVE_SPE	(1 << 12) /* Save SPE context if active  */
437 #define KVM_ARM64_DEBUG_STATE_SAVE_TRBE	(1 << 13) /* Save TRBE context if active  */
438 #define KVM_ARM64_FP_FOREIGN_FPSTATE	(1 << 14)
439 
440 #define KVM_GUESTDBG_VALID_MASK (KVM_GUESTDBG_ENABLE | \
441 				 KVM_GUESTDBG_USE_SW_BP | \
442 				 KVM_GUESTDBG_USE_HW | \
443 				 KVM_GUESTDBG_SINGLESTEP)
444 
445 #define vcpu_has_sve(vcpu) (system_supports_sve() &&			\
446 			    ((vcpu)->arch.flags & KVM_ARM64_GUEST_HAS_SVE))
447 
448 #ifdef CONFIG_ARM64_PTR_AUTH
449 #define vcpu_has_ptrauth(vcpu)						\
450 	((cpus_have_final_cap(ARM64_HAS_ADDRESS_AUTH) ||		\
451 	  cpus_have_final_cap(ARM64_HAS_GENERIC_AUTH)) &&		\
452 	 (vcpu)->arch.flags & KVM_ARM64_GUEST_HAS_PTRAUTH)
453 #else
454 #define vcpu_has_ptrauth(vcpu)		false
455 #endif
456 
457 #define vcpu_gp_regs(v)		(&(v)->arch.ctxt.regs)
458 
459 /*
460  * Only use __vcpu_sys_reg/ctxt_sys_reg if you know you want the
461  * memory backed version of a register, and not the one most recently
462  * accessed by a running VCPU.  For example, for userspace access or
463  * for system registers that are never context switched, but only
464  * emulated.
465  */
466 #define __ctxt_sys_reg(c,r)	(&(c)->sys_regs[(r)])
467 
468 #define ctxt_sys_reg(c,r)	(*__ctxt_sys_reg(c,r))
469 
470 #define __vcpu_sys_reg(v,r)	(ctxt_sys_reg(&(v)->arch.ctxt, (r)))
471 
472 u64 vcpu_read_sys_reg(const struct kvm_vcpu *vcpu, int reg);
473 void vcpu_write_sys_reg(struct kvm_vcpu *vcpu, u64 val, int reg);
474 
475 static inline bool __vcpu_read_sys_reg_from_cpu(int reg, u64 *val)
476 {
477 	/*
478 	 * *** VHE ONLY ***
479 	 *
480 	 * System registers listed in the switch are not saved on every
481 	 * exit from the guest but are only saved on vcpu_put.
482 	 *
483 	 * Note that MPIDR_EL1 for the guest is set by KVM via VMPIDR_EL2 but
484 	 * should never be listed below, because the guest cannot modify its
485 	 * own MPIDR_EL1 and MPIDR_EL1 is accessed for VCPU A from VCPU B's
486 	 * thread when emulating cross-VCPU communication.
487 	 */
488 	if (!has_vhe())
489 		return false;
490 
491 	switch (reg) {
492 	case CSSELR_EL1:	*val = read_sysreg_s(SYS_CSSELR_EL1);	break;
493 	case SCTLR_EL1:		*val = read_sysreg_s(SYS_SCTLR_EL12);	break;
494 	case CPACR_EL1:		*val = read_sysreg_s(SYS_CPACR_EL12);	break;
495 	case TTBR0_EL1:		*val = read_sysreg_s(SYS_TTBR0_EL12);	break;
496 	case TTBR1_EL1:		*val = read_sysreg_s(SYS_TTBR1_EL12);	break;
497 	case TCR_EL1:		*val = read_sysreg_s(SYS_TCR_EL12);	break;
498 	case ESR_EL1:		*val = read_sysreg_s(SYS_ESR_EL12);	break;
499 	case AFSR0_EL1:		*val = read_sysreg_s(SYS_AFSR0_EL12);	break;
500 	case AFSR1_EL1:		*val = read_sysreg_s(SYS_AFSR1_EL12);	break;
501 	case FAR_EL1:		*val = read_sysreg_s(SYS_FAR_EL12);	break;
502 	case MAIR_EL1:		*val = read_sysreg_s(SYS_MAIR_EL12);	break;
503 	case VBAR_EL1:		*val = read_sysreg_s(SYS_VBAR_EL12);	break;
504 	case CONTEXTIDR_EL1:	*val = read_sysreg_s(SYS_CONTEXTIDR_EL12);break;
505 	case TPIDR_EL0:		*val = read_sysreg_s(SYS_TPIDR_EL0);	break;
506 	case TPIDRRO_EL0:	*val = read_sysreg_s(SYS_TPIDRRO_EL0);	break;
507 	case TPIDR_EL1:		*val = read_sysreg_s(SYS_TPIDR_EL1);	break;
508 	case AMAIR_EL1:		*val = read_sysreg_s(SYS_AMAIR_EL12);	break;
509 	case CNTKCTL_EL1:	*val = read_sysreg_s(SYS_CNTKCTL_EL12);	break;
510 	case ELR_EL1:		*val = read_sysreg_s(SYS_ELR_EL12);	break;
511 	case PAR_EL1:		*val = read_sysreg_par();		break;
512 	case DACR32_EL2:	*val = read_sysreg_s(SYS_DACR32_EL2);	break;
513 	case IFSR32_EL2:	*val = read_sysreg_s(SYS_IFSR32_EL2);	break;
514 	case DBGVCR32_EL2:	*val = read_sysreg_s(SYS_DBGVCR32_EL2);	break;
515 	default:		return false;
516 	}
517 
518 	return true;
519 }
520 
521 static inline bool __vcpu_write_sys_reg_to_cpu(u64 val, int reg)
522 {
523 	/*
524 	 * *** VHE ONLY ***
525 	 *
526 	 * System registers listed in the switch are not restored on every
527 	 * entry to the guest but are only restored on vcpu_load.
528 	 *
529 	 * Note that MPIDR_EL1 for the guest is set by KVM via VMPIDR_EL2 but
530 	 * should never be listed below, because the MPIDR should only be set
531 	 * once, before running the VCPU, and never changed later.
532 	 */
533 	if (!has_vhe())
534 		return false;
535 
536 	switch (reg) {
537 	case CSSELR_EL1:	write_sysreg_s(val, SYS_CSSELR_EL1);	break;
538 	case SCTLR_EL1:		write_sysreg_s(val, SYS_SCTLR_EL12);	break;
539 	case CPACR_EL1:		write_sysreg_s(val, SYS_CPACR_EL12);	break;
540 	case TTBR0_EL1:		write_sysreg_s(val, SYS_TTBR0_EL12);	break;
541 	case TTBR1_EL1:		write_sysreg_s(val, SYS_TTBR1_EL12);	break;
542 	case TCR_EL1:		write_sysreg_s(val, SYS_TCR_EL12);	break;
543 	case ESR_EL1:		write_sysreg_s(val, SYS_ESR_EL12);	break;
544 	case AFSR0_EL1:		write_sysreg_s(val, SYS_AFSR0_EL12);	break;
545 	case AFSR1_EL1:		write_sysreg_s(val, SYS_AFSR1_EL12);	break;
546 	case FAR_EL1:		write_sysreg_s(val, SYS_FAR_EL12);	break;
547 	case MAIR_EL1:		write_sysreg_s(val, SYS_MAIR_EL12);	break;
548 	case VBAR_EL1:		write_sysreg_s(val, SYS_VBAR_EL12);	break;
549 	case CONTEXTIDR_EL1:	write_sysreg_s(val, SYS_CONTEXTIDR_EL12);break;
550 	case TPIDR_EL0:		write_sysreg_s(val, SYS_TPIDR_EL0);	break;
551 	case TPIDRRO_EL0:	write_sysreg_s(val, SYS_TPIDRRO_EL0);	break;
552 	case TPIDR_EL1:		write_sysreg_s(val, SYS_TPIDR_EL1);	break;
553 	case AMAIR_EL1:		write_sysreg_s(val, SYS_AMAIR_EL12);	break;
554 	case CNTKCTL_EL1:	write_sysreg_s(val, SYS_CNTKCTL_EL12);	break;
555 	case ELR_EL1:		write_sysreg_s(val, SYS_ELR_EL12);	break;
556 	case PAR_EL1:		write_sysreg_s(val, SYS_PAR_EL1);	break;
557 	case DACR32_EL2:	write_sysreg_s(val, SYS_DACR32_EL2);	break;
558 	case IFSR32_EL2:	write_sysreg_s(val, SYS_IFSR32_EL2);	break;
559 	case DBGVCR32_EL2:	write_sysreg_s(val, SYS_DBGVCR32_EL2);	break;
560 	default:		return false;
561 	}
562 
563 	return true;
564 }
565 
566 struct kvm_vm_stat {
567 	struct kvm_vm_stat_generic generic;
568 };
569 
570 struct kvm_vcpu_stat {
571 	struct kvm_vcpu_stat_generic generic;
572 	u64 hvc_exit_stat;
573 	u64 wfe_exit_stat;
574 	u64 wfi_exit_stat;
575 	u64 mmio_exit_user;
576 	u64 mmio_exit_kernel;
577 	u64 signal_exits;
578 	u64 exits;
579 };
580 
581 void kvm_vcpu_preferred_target(struct kvm_vcpu_init *init);
582 unsigned long kvm_arm_num_regs(struct kvm_vcpu *vcpu);
583 int kvm_arm_copy_reg_indices(struct kvm_vcpu *vcpu, u64 __user *indices);
584 int kvm_arm_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
585 int kvm_arm_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
586 
587 unsigned long kvm_arm_num_sys_reg_descs(struct kvm_vcpu *vcpu);
588 int kvm_arm_copy_sys_reg_indices(struct kvm_vcpu *vcpu, u64 __user *uindices);
589 int kvm_arm_sys_reg_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *);
590 int kvm_arm_sys_reg_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *);
591 
592 int __kvm_arm_vcpu_get_events(struct kvm_vcpu *vcpu,
593 			      struct kvm_vcpu_events *events);
594 
595 int __kvm_arm_vcpu_set_events(struct kvm_vcpu *vcpu,
596 			      struct kvm_vcpu_events *events);
597 
598 #define KVM_ARCH_WANT_MMU_NOTIFIER
599 
600 void kvm_arm_halt_guest(struct kvm *kvm);
601 void kvm_arm_resume_guest(struct kvm *kvm);
602 
603 #define vcpu_has_run_once(vcpu)	!!rcu_access_pointer((vcpu)->pid)
604 
605 #ifndef __KVM_NVHE_HYPERVISOR__
606 #define kvm_call_hyp_nvhe(f, ...)						\
607 	({								\
608 		struct arm_smccc_res res;				\
609 									\
610 		arm_smccc_1_1_hvc(KVM_HOST_SMCCC_FUNC(f),		\
611 				  ##__VA_ARGS__, &res);			\
612 		WARN_ON(res.a0 != SMCCC_RET_SUCCESS);			\
613 									\
614 		res.a1;							\
615 	})
616 
617 /*
618  * The couple of isb() below are there to guarantee the same behaviour
619  * on VHE as on !VHE, where the eret to EL1 acts as a context
620  * synchronization event.
621  */
622 #define kvm_call_hyp(f, ...)						\
623 	do {								\
624 		if (has_vhe()) {					\
625 			f(__VA_ARGS__);					\
626 			isb();						\
627 		} else {						\
628 			kvm_call_hyp_nvhe(f, ##__VA_ARGS__);		\
629 		}							\
630 	} while(0)
631 
632 #define kvm_call_hyp_ret(f, ...)					\
633 	({								\
634 		typeof(f(__VA_ARGS__)) ret;				\
635 									\
636 		if (has_vhe()) {					\
637 			ret = f(__VA_ARGS__);				\
638 			isb();						\
639 		} else {						\
640 			ret = kvm_call_hyp_nvhe(f, ##__VA_ARGS__);	\
641 		}							\
642 									\
643 		ret;							\
644 	})
645 #else /* __KVM_NVHE_HYPERVISOR__ */
646 #define kvm_call_hyp(f, ...) f(__VA_ARGS__)
647 #define kvm_call_hyp_ret(f, ...) f(__VA_ARGS__)
648 #define kvm_call_hyp_nvhe(f, ...) f(__VA_ARGS__)
649 #endif /* __KVM_NVHE_HYPERVISOR__ */
650 
651 void force_vm_exit(const cpumask_t *mask);
652 
653 int handle_exit(struct kvm_vcpu *vcpu, int exception_index);
654 void handle_exit_early(struct kvm_vcpu *vcpu, int exception_index);
655 
656 int kvm_handle_cp14_load_store(struct kvm_vcpu *vcpu);
657 int kvm_handle_cp14_32(struct kvm_vcpu *vcpu);
658 int kvm_handle_cp14_64(struct kvm_vcpu *vcpu);
659 int kvm_handle_cp15_32(struct kvm_vcpu *vcpu);
660 int kvm_handle_cp15_64(struct kvm_vcpu *vcpu);
661 int kvm_handle_sys_reg(struct kvm_vcpu *vcpu);
662 
663 void kvm_reset_sys_regs(struct kvm_vcpu *vcpu);
664 
665 void kvm_sys_reg_table_init(void);
666 
667 /* MMIO helpers */
668 void kvm_mmio_write_buf(void *buf, unsigned int len, unsigned long data);
669 unsigned long kvm_mmio_read_buf(const void *buf, unsigned int len);
670 
671 int kvm_handle_mmio_return(struct kvm_vcpu *vcpu);
672 int io_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa);
673 
674 /*
675  * Returns true if a Performance Monitoring Interrupt (PMI), a.k.a. perf event,
676  * arrived in guest context.  For arm64, any event that arrives while a vCPU is
677  * loaded is considered to be "in guest".
678  */
679 static inline bool kvm_arch_pmi_in_guest(struct kvm_vcpu *vcpu)
680 {
681 	return IS_ENABLED(CONFIG_GUEST_PERF_EVENTS) && !!vcpu;
682 }
683 
684 long kvm_hypercall_pv_features(struct kvm_vcpu *vcpu);
685 gpa_t kvm_init_stolen_time(struct kvm_vcpu *vcpu);
686 void kvm_update_stolen_time(struct kvm_vcpu *vcpu);
687 
688 bool kvm_arm_pvtime_supported(void);
689 int kvm_arm_pvtime_set_attr(struct kvm_vcpu *vcpu,
690 			    struct kvm_device_attr *attr);
691 int kvm_arm_pvtime_get_attr(struct kvm_vcpu *vcpu,
692 			    struct kvm_device_attr *attr);
693 int kvm_arm_pvtime_has_attr(struct kvm_vcpu *vcpu,
694 			    struct kvm_device_attr *attr);
695 
696 static inline void kvm_arm_pvtime_vcpu_init(struct kvm_vcpu_arch *vcpu_arch)
697 {
698 	vcpu_arch->steal.base = GPA_INVALID;
699 }
700 
701 static inline bool kvm_arm_is_pvtime_enabled(struct kvm_vcpu_arch *vcpu_arch)
702 {
703 	return (vcpu_arch->steal.base != GPA_INVALID);
704 }
705 
706 void kvm_set_sei_esr(struct kvm_vcpu *vcpu, u64 syndrome);
707 
708 struct kvm_vcpu *kvm_mpidr_to_vcpu(struct kvm *kvm, unsigned long mpidr);
709 
710 DECLARE_KVM_HYP_PER_CPU(struct kvm_host_data, kvm_host_data);
711 
712 static inline void kvm_init_host_cpu_context(struct kvm_cpu_context *cpu_ctxt)
713 {
714 	/* The host's MPIDR is immutable, so let's set it up at boot time */
715 	ctxt_sys_reg(cpu_ctxt, MPIDR_EL1) = read_cpuid_mpidr();
716 }
717 
718 void kvm_arm_vcpu_ptrauth_trap(struct kvm_vcpu *vcpu);
719 
720 static inline void kvm_arch_hardware_unsetup(void) {}
721 static inline void kvm_arch_sync_events(struct kvm *kvm) {}
722 static inline void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu) {}
723 
724 void kvm_arm_init_debug(void);
725 void kvm_arm_vcpu_init_debug(struct kvm_vcpu *vcpu);
726 void kvm_arm_setup_debug(struct kvm_vcpu *vcpu);
727 void kvm_arm_clear_debug(struct kvm_vcpu *vcpu);
728 void kvm_arm_reset_debug_ptr(struct kvm_vcpu *vcpu);
729 int kvm_arm_vcpu_arch_set_attr(struct kvm_vcpu *vcpu,
730 			       struct kvm_device_attr *attr);
731 int kvm_arm_vcpu_arch_get_attr(struct kvm_vcpu *vcpu,
732 			       struct kvm_device_attr *attr);
733 int kvm_arm_vcpu_arch_has_attr(struct kvm_vcpu *vcpu,
734 			       struct kvm_device_attr *attr);
735 
736 long kvm_vm_ioctl_mte_copy_tags(struct kvm *kvm,
737 				struct kvm_arm_copy_mte_tags *copy_tags);
738 
739 /* Guest/host FPSIMD coordination helpers */
740 int kvm_arch_vcpu_run_map_fp(struct kvm_vcpu *vcpu);
741 void kvm_arch_vcpu_load_fp(struct kvm_vcpu *vcpu);
742 void kvm_arch_vcpu_ctxflush_fp(struct kvm_vcpu *vcpu);
743 void kvm_arch_vcpu_ctxsync_fp(struct kvm_vcpu *vcpu);
744 void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu);
745 void kvm_vcpu_unshare_task_fp(struct kvm_vcpu *vcpu);
746 
747 static inline bool kvm_pmu_counter_deferred(struct perf_event_attr *attr)
748 {
749 	return (!has_vhe() && attr->exclude_host);
750 }
751 
752 /* Flags for host debug state */
753 void kvm_arch_vcpu_load_debug_state_flags(struct kvm_vcpu *vcpu);
754 void kvm_arch_vcpu_put_debug_state_flags(struct kvm_vcpu *vcpu);
755 
756 #ifdef CONFIG_KVM
757 void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr);
758 void kvm_clr_pmu_events(u32 clr);
759 
760 void kvm_vcpu_pmu_restore_guest(struct kvm_vcpu *vcpu);
761 void kvm_vcpu_pmu_restore_host(struct kvm_vcpu *vcpu);
762 #else
763 static inline void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr) {}
764 static inline void kvm_clr_pmu_events(u32 clr) {}
765 #endif
766 
767 void kvm_vcpu_load_sysregs_vhe(struct kvm_vcpu *vcpu);
768 void kvm_vcpu_put_sysregs_vhe(struct kvm_vcpu *vcpu);
769 
770 int kvm_set_ipa_limit(void);
771 
772 #define __KVM_HAVE_ARCH_VM_ALLOC
773 struct kvm *kvm_arch_alloc_vm(void);
774 
775 int kvm_arm_setup_stage2(struct kvm *kvm, unsigned long type);
776 
777 static inline bool kvm_vm_is_protected(struct kvm *kvm)
778 {
779 	return false;
780 }
781 
782 void kvm_init_protected_traps(struct kvm_vcpu *vcpu);
783 
784 int kvm_arm_vcpu_finalize(struct kvm_vcpu *vcpu, int feature);
785 bool kvm_arm_vcpu_is_finalized(struct kvm_vcpu *vcpu);
786 
787 #define kvm_arm_vcpu_sve_finalized(vcpu) \
788 	((vcpu)->arch.flags & KVM_ARM64_VCPU_SVE_FINALIZED)
789 
790 #define kvm_has_mte(kvm) (system_supports_mte() && (kvm)->arch.mte_enabled)
791 #define kvm_vcpu_has_pmu(vcpu)					\
792 	(test_bit(KVM_ARM_VCPU_PMU_V3, (vcpu)->arch.features))
793 
794 int kvm_trng_call(struct kvm_vcpu *vcpu);
795 #ifdef CONFIG_KVM
796 extern phys_addr_t hyp_mem_base;
797 extern phys_addr_t hyp_mem_size;
798 void __init kvm_hyp_reserve(void);
799 #else
800 static inline void kvm_hyp_reserve(void) { }
801 #endif
802 
803 #endif /* __ARM64_KVM_HOST_H__ */
804