xref: /openbmc/linux/arch/arm64/include/asm/kvm_host.h (revision 3248136b)
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 	atomic64_t id;
75 };
76 
77 struct kvm_s2_mmu {
78 	struct kvm_vmid vmid;
79 
80 	/*
81 	 * stage2 entry level table
82 	 *
83 	 * Two kvm_s2_mmu structures in the same VM can point to the same
84 	 * pgd here.  This happens when running a guest using a
85 	 * translation regime that isn't affected by its own stage-2
86 	 * translation, such as a non-VHE hypervisor running at vEL2, or
87 	 * for vEL1/EL0 with vHCR_EL2.VM == 0.  In that case, we use the
88 	 * canonical stage-2 page tables.
89 	 */
90 	phys_addr_t	pgd_phys;
91 	struct kvm_pgtable *pgt;
92 
93 	/* The last vcpu id that ran on each physical CPU */
94 	int __percpu *last_vcpu_ran;
95 
96 	struct kvm_arch *arch;
97 };
98 
99 struct kvm_arch_memory_slot {
100 };
101 
102 struct kvm_arch {
103 	struct kvm_s2_mmu mmu;
104 
105 	/* VTCR_EL2 value for this VM */
106 	u64    vtcr;
107 
108 	/* The maximum number of vCPUs depends on the used GIC model */
109 	int max_vcpus;
110 
111 	/* Interrupt controller */
112 	struct vgic_dist	vgic;
113 
114 	/* Mandated version of PSCI */
115 	u32 psci_version;
116 
117 	/*
118 	 * If we encounter a data abort without valid instruction syndrome
119 	 * information, report this to user space.  User space can (and
120 	 * should) opt in to this feature if KVM_CAP_ARM_NISV_TO_USER is
121 	 * supported.
122 	 */
123 	bool return_nisv_io_abort_to_user;
124 
125 	/*
126 	 * VM-wide PMU filter, implemented as a bitmap and big enough for
127 	 * up to 2^10 events (ARMv8.0) or 2^16 events (ARMv8.1+).
128 	 */
129 	unsigned long *pmu_filter;
130 	unsigned int pmuver;
131 
132 	u8 pfr0_csv2;
133 	u8 pfr0_csv3;
134 
135 	/* Memory Tagging Extension enabled for the guest */
136 	bool mte_enabled;
137 };
138 
139 struct kvm_vcpu_fault_info {
140 	u32 esr_el2;		/* Hyp Syndrom Register */
141 	u64 far_el2;		/* Hyp Fault Address Register */
142 	u64 hpfar_el2;		/* Hyp IPA Fault Address Register */
143 	u64 disr_el1;		/* Deferred [SError] Status Register */
144 };
145 
146 enum vcpu_sysreg {
147 	__INVALID_SYSREG__,   /* 0 is reserved as an invalid value */
148 	MPIDR_EL1,	/* MultiProcessor Affinity Register */
149 	CSSELR_EL1,	/* Cache Size Selection Register */
150 	SCTLR_EL1,	/* System Control Register */
151 	ACTLR_EL1,	/* Auxiliary Control Register */
152 	CPACR_EL1,	/* Coprocessor Access Control */
153 	ZCR_EL1,	/* SVE Control */
154 	TTBR0_EL1,	/* Translation Table Base Register 0 */
155 	TTBR1_EL1,	/* Translation Table Base Register 1 */
156 	TCR_EL1,	/* Translation Control Register */
157 	ESR_EL1,	/* Exception Syndrome Register */
158 	AFSR0_EL1,	/* Auxiliary Fault Status Register 0 */
159 	AFSR1_EL1,	/* Auxiliary Fault Status Register 1 */
160 	FAR_EL1,	/* Fault Address Register */
161 	MAIR_EL1,	/* Memory Attribute Indirection Register */
162 	VBAR_EL1,	/* Vector Base Address Register */
163 	CONTEXTIDR_EL1,	/* Context ID Register */
164 	TPIDR_EL0,	/* Thread ID, User R/W */
165 	TPIDRRO_EL0,	/* Thread ID, User R/O */
166 	TPIDR_EL1,	/* Thread ID, Privileged */
167 	AMAIR_EL1,	/* Aux Memory Attribute Indirection Register */
168 	CNTKCTL_EL1,	/* Timer Control Register (EL1) */
169 	PAR_EL1,	/* Physical Address Register */
170 	MDSCR_EL1,	/* Monitor Debug System Control Register */
171 	MDCCINT_EL1,	/* Monitor Debug Comms Channel Interrupt Enable Reg */
172 	DISR_EL1,	/* Deferred Interrupt Status Register */
173 
174 	/* Performance Monitors Registers */
175 	PMCR_EL0,	/* Control Register */
176 	PMSELR_EL0,	/* Event Counter Selection Register */
177 	PMEVCNTR0_EL0,	/* Event Counter Register (0-30) */
178 	PMEVCNTR30_EL0 = PMEVCNTR0_EL0 + 30,
179 	PMCCNTR_EL0,	/* Cycle Counter Register */
180 	PMEVTYPER0_EL0,	/* Event Type Register (0-30) */
181 	PMEVTYPER30_EL0 = PMEVTYPER0_EL0 + 30,
182 	PMCCFILTR_EL0,	/* Cycle Count Filter Register */
183 	PMCNTENSET_EL0,	/* Count Enable Set Register */
184 	PMINTENSET_EL1,	/* Interrupt Enable Set Register */
185 	PMOVSSET_EL0,	/* Overflow Flag Status Set Register */
186 	PMUSERENR_EL0,	/* User Enable Register */
187 
188 	/* Pointer Authentication Registers in a strict increasing order. */
189 	APIAKEYLO_EL1,
190 	APIAKEYHI_EL1,
191 	APIBKEYLO_EL1,
192 	APIBKEYHI_EL1,
193 	APDAKEYLO_EL1,
194 	APDAKEYHI_EL1,
195 	APDBKEYLO_EL1,
196 	APDBKEYHI_EL1,
197 	APGAKEYLO_EL1,
198 	APGAKEYHI_EL1,
199 
200 	ELR_EL1,
201 	SP_EL1,
202 	SPSR_EL1,
203 
204 	CNTVOFF_EL2,
205 	CNTV_CVAL_EL0,
206 	CNTV_CTL_EL0,
207 	CNTP_CVAL_EL0,
208 	CNTP_CTL_EL0,
209 
210 	/* Memory Tagging Extension registers */
211 	RGSR_EL1,	/* Random Allocation Tag Seed Register */
212 	GCR_EL1,	/* Tag Control Register */
213 	TFSR_EL1,	/* Tag Fault Status Register (EL1) */
214 	TFSRE0_EL1,	/* Tag Fault Status Register (EL0) */
215 
216 	/* 32bit specific registers. Keep them at the end of the range */
217 	DACR32_EL2,	/* Domain Access Control Register */
218 	IFSR32_EL2,	/* Instruction Fault Status Register */
219 	FPEXC32_EL2,	/* Floating-Point Exception Control Register */
220 	DBGVCR32_EL2,	/* Debug Vector Catch Register */
221 
222 	NR_SYS_REGS	/* Nothing after this line! */
223 };
224 
225 struct kvm_cpu_context {
226 	struct user_pt_regs regs;	/* sp = sp_el0 */
227 
228 	u64	spsr_abt;
229 	u64	spsr_und;
230 	u64	spsr_irq;
231 	u64	spsr_fiq;
232 
233 	struct user_fpsimd_state fp_regs;
234 
235 	u64 sys_regs[NR_SYS_REGS];
236 
237 	struct kvm_vcpu *__hyp_running_vcpu;
238 };
239 
240 struct kvm_pmu_events {
241 	u32 events_host;
242 	u32 events_guest;
243 };
244 
245 struct kvm_host_data {
246 	struct kvm_cpu_context host_ctxt;
247 	struct kvm_pmu_events pmu_events;
248 };
249 
250 struct kvm_host_psci_config {
251 	/* PSCI version used by host. */
252 	u32 version;
253 
254 	/* Function IDs used by host if version is v0.1. */
255 	struct psci_0_1_function_ids function_ids_0_1;
256 
257 	bool psci_0_1_cpu_suspend_implemented;
258 	bool psci_0_1_cpu_on_implemented;
259 	bool psci_0_1_cpu_off_implemented;
260 	bool psci_0_1_migrate_implemented;
261 };
262 
263 extern struct kvm_host_psci_config kvm_nvhe_sym(kvm_host_psci_config);
264 #define kvm_host_psci_config CHOOSE_NVHE_SYM(kvm_host_psci_config)
265 
266 extern s64 kvm_nvhe_sym(hyp_physvirt_offset);
267 #define hyp_physvirt_offset CHOOSE_NVHE_SYM(hyp_physvirt_offset)
268 
269 extern u64 kvm_nvhe_sym(hyp_cpu_logical_map)[NR_CPUS];
270 #define hyp_cpu_logical_map CHOOSE_NVHE_SYM(hyp_cpu_logical_map)
271 
272 struct vcpu_reset_state {
273 	unsigned long	pc;
274 	unsigned long	r0;
275 	bool		be;
276 	bool		reset;
277 };
278 
279 struct kvm_vcpu_arch {
280 	struct kvm_cpu_context ctxt;
281 	void *sve_state;
282 	unsigned int sve_max_vl;
283 
284 	/* Stage 2 paging state used by the hardware on next switch */
285 	struct kvm_s2_mmu *hw_mmu;
286 
287 	/* Values of trap registers for the guest. */
288 	u64 hcr_el2;
289 	u64 mdcr_el2;
290 	u64 cptr_el2;
291 
292 	/* Values of trap registers for the host before guest entry. */
293 	u64 mdcr_el2_host;
294 
295 	/* Exception Information */
296 	struct kvm_vcpu_fault_info fault;
297 
298 	/* Miscellaneous vcpu state flags */
299 	u64 flags;
300 
301 	/*
302 	 * We maintain more than a single set of debug registers to support
303 	 * debugging the guest from the host and to maintain separate host and
304 	 * guest state during world switches. vcpu_debug_state are the debug
305 	 * registers of the vcpu as the guest sees them.  host_debug_state are
306 	 * the host registers which are saved and restored during
307 	 * world switches. external_debug_state contains the debug
308 	 * values we want to debug the guest. This is set via the
309 	 * KVM_SET_GUEST_DEBUG ioctl.
310 	 *
311 	 * debug_ptr points to the set of debug registers that should be loaded
312 	 * onto the hardware when running the guest.
313 	 */
314 	struct kvm_guest_debug_arch *debug_ptr;
315 	struct kvm_guest_debug_arch vcpu_debug_state;
316 	struct kvm_guest_debug_arch external_debug_state;
317 
318 	struct user_fpsimd_state *host_fpsimd_state;	/* hyp VA */
319 	struct task_struct *parent_task;
320 
321 	struct {
322 		/* {Break,watch}point registers */
323 		struct kvm_guest_debug_arch regs;
324 		/* Statistical profiling extension */
325 		u64 pmscr_el1;
326 		/* Self-hosted trace */
327 		u64 trfcr_el1;
328 	} host_debug_state;
329 
330 	/* VGIC state */
331 	struct vgic_cpu vgic_cpu;
332 	struct arch_timer_cpu timer_cpu;
333 	struct kvm_pmu pmu;
334 
335 	/*
336 	 * Anything that is not used directly from assembly code goes
337 	 * here.
338 	 */
339 
340 	/*
341 	 * Guest registers we preserve during guest debugging.
342 	 *
343 	 * These shadow registers are updated by the kvm_handle_sys_reg
344 	 * trap handler if the guest accesses or updates them while we
345 	 * are using guest debug.
346 	 */
347 	struct {
348 		u32	mdscr_el1;
349 	} guest_debug_preserved;
350 
351 	/* vcpu power-off state */
352 	bool power_off;
353 
354 	/* Don't run the guest (internal implementation need) */
355 	bool pause;
356 
357 	/* Cache some mmu pages needed inside spinlock regions */
358 	struct kvm_mmu_memory_cache mmu_page_cache;
359 
360 	/* Target CPU and feature flags */
361 	int target;
362 	DECLARE_BITMAP(features, KVM_VCPU_MAX_FEATURES);
363 
364 	/* Virtual SError ESR to restore when HCR_EL2.VSE is set */
365 	u64 vsesr_el2;
366 
367 	/* Additional reset state */
368 	struct vcpu_reset_state	reset_state;
369 
370 	/* True when deferrable sysregs are loaded on the physical CPU,
371 	 * see kvm_vcpu_load_sysregs_vhe and kvm_vcpu_put_sysregs_vhe. */
372 	bool sysregs_loaded_on_cpu;
373 
374 	/* Guest PV state */
375 	struct {
376 		u64 last_steal;
377 		gpa_t base;
378 	} steal;
379 };
380 
381 /* Pointer to the vcpu's SVE FFR for sve_{save,load}_state() */
382 #define vcpu_sve_pffr(vcpu) (kern_hyp_va((vcpu)->arch.sve_state) +	\
383 			     sve_ffr_offset((vcpu)->arch.sve_max_vl))
384 
385 #define vcpu_sve_max_vq(vcpu)	sve_vq_from_vl((vcpu)->arch.sve_max_vl)
386 
387 #define vcpu_sve_state_size(vcpu) ({					\
388 	size_t __size_ret;						\
389 	unsigned int __vcpu_vq;						\
390 									\
391 	if (WARN_ON(!sve_vl_valid((vcpu)->arch.sve_max_vl))) {		\
392 		__size_ret = 0;						\
393 	} else {							\
394 		__vcpu_vq = vcpu_sve_max_vq(vcpu);			\
395 		__size_ret = SVE_SIG_REGS_SIZE(__vcpu_vq);		\
396 	}								\
397 									\
398 	__size_ret;							\
399 })
400 
401 /* vcpu_arch flags field values: */
402 #define KVM_ARM64_DEBUG_DIRTY		(1 << 0)
403 #define KVM_ARM64_FP_ENABLED		(1 << 1) /* guest FP regs loaded */
404 #define KVM_ARM64_FP_HOST		(1 << 2) /* host FP regs loaded */
405 #define KVM_ARM64_HOST_SVE_ENABLED	(1 << 4) /* SVE enabled for EL0 */
406 #define KVM_ARM64_GUEST_HAS_SVE		(1 << 5) /* SVE exposed to guest */
407 #define KVM_ARM64_VCPU_SVE_FINALIZED	(1 << 6) /* SVE config completed */
408 #define KVM_ARM64_GUEST_HAS_PTRAUTH	(1 << 7) /* PTRAUTH exposed to guest */
409 #define KVM_ARM64_PENDING_EXCEPTION	(1 << 8) /* Exception pending */
410 /*
411  * Overlaps with KVM_ARM64_EXCEPT_MASK on purpose so that it can't be
412  * set together with an exception...
413  */
414 #define KVM_ARM64_INCREMENT_PC		(1 << 9) /* Increment PC */
415 #define KVM_ARM64_EXCEPT_MASK		(7 << 9) /* Target EL/MODE */
416 /*
417  * When KVM_ARM64_PENDING_EXCEPTION is set, KVM_ARM64_EXCEPT_MASK can
418  * take the following values:
419  *
420  * For AArch32 EL1:
421  */
422 #define KVM_ARM64_EXCEPT_AA32_UND	(0 << 9)
423 #define KVM_ARM64_EXCEPT_AA32_IABT	(1 << 9)
424 #define KVM_ARM64_EXCEPT_AA32_DABT	(2 << 9)
425 /* For AArch64: */
426 #define KVM_ARM64_EXCEPT_AA64_ELx_SYNC	(0 << 9)
427 #define KVM_ARM64_EXCEPT_AA64_ELx_IRQ	(1 << 9)
428 #define KVM_ARM64_EXCEPT_AA64_ELx_FIQ	(2 << 9)
429 #define KVM_ARM64_EXCEPT_AA64_ELx_SERR	(3 << 9)
430 #define KVM_ARM64_EXCEPT_AA64_EL1	(0 << 11)
431 #define KVM_ARM64_EXCEPT_AA64_EL2	(1 << 11)
432 
433 #define KVM_ARM64_DEBUG_STATE_SAVE_SPE	(1 << 12) /* Save SPE context if active  */
434 #define KVM_ARM64_DEBUG_STATE_SAVE_TRBE	(1 << 13) /* Save TRBE context if active  */
435 #define KVM_ARM64_FP_FOREIGN_FPSTATE	(1 << 14)
436 
437 #define KVM_GUESTDBG_VALID_MASK (KVM_GUESTDBG_ENABLE | \
438 				 KVM_GUESTDBG_USE_SW_BP | \
439 				 KVM_GUESTDBG_USE_HW | \
440 				 KVM_GUESTDBG_SINGLESTEP)
441 
442 #define vcpu_has_sve(vcpu) (system_supports_sve() &&			\
443 			    ((vcpu)->arch.flags & KVM_ARM64_GUEST_HAS_SVE))
444 
445 #ifdef CONFIG_ARM64_PTR_AUTH
446 #define vcpu_has_ptrauth(vcpu)						\
447 	((cpus_have_final_cap(ARM64_HAS_ADDRESS_AUTH) ||		\
448 	  cpus_have_final_cap(ARM64_HAS_GENERIC_AUTH)) &&		\
449 	 (vcpu)->arch.flags & KVM_ARM64_GUEST_HAS_PTRAUTH)
450 #else
451 #define vcpu_has_ptrauth(vcpu)		false
452 #endif
453 
454 #define vcpu_gp_regs(v)		(&(v)->arch.ctxt.regs)
455 
456 /*
457  * Only use __vcpu_sys_reg/ctxt_sys_reg if you know you want the
458  * memory backed version of a register, and not the one most recently
459  * accessed by a running VCPU.  For example, for userspace access or
460  * for system registers that are never context switched, but only
461  * emulated.
462  */
463 #define __ctxt_sys_reg(c,r)	(&(c)->sys_regs[(r)])
464 
465 #define ctxt_sys_reg(c,r)	(*__ctxt_sys_reg(c,r))
466 
467 #define __vcpu_sys_reg(v,r)	(ctxt_sys_reg(&(v)->arch.ctxt, (r)))
468 
469 u64 vcpu_read_sys_reg(const struct kvm_vcpu *vcpu, int reg);
470 void vcpu_write_sys_reg(struct kvm_vcpu *vcpu, u64 val, int reg);
471 
472 static inline bool __vcpu_read_sys_reg_from_cpu(int reg, u64 *val)
473 {
474 	/*
475 	 * *** VHE ONLY ***
476 	 *
477 	 * System registers listed in the switch are not saved on every
478 	 * exit from the guest but are only saved on vcpu_put.
479 	 *
480 	 * Note that MPIDR_EL1 for the guest is set by KVM via VMPIDR_EL2 but
481 	 * should never be listed below, because the guest cannot modify its
482 	 * own MPIDR_EL1 and MPIDR_EL1 is accessed for VCPU A from VCPU B's
483 	 * thread when emulating cross-VCPU communication.
484 	 */
485 	if (!has_vhe())
486 		return false;
487 
488 	switch (reg) {
489 	case CSSELR_EL1:	*val = read_sysreg_s(SYS_CSSELR_EL1);	break;
490 	case SCTLR_EL1:		*val = read_sysreg_s(SYS_SCTLR_EL12);	break;
491 	case CPACR_EL1:		*val = read_sysreg_s(SYS_CPACR_EL12);	break;
492 	case TTBR0_EL1:		*val = read_sysreg_s(SYS_TTBR0_EL12);	break;
493 	case TTBR1_EL1:		*val = read_sysreg_s(SYS_TTBR1_EL12);	break;
494 	case TCR_EL1:		*val = read_sysreg_s(SYS_TCR_EL12);	break;
495 	case ESR_EL1:		*val = read_sysreg_s(SYS_ESR_EL12);	break;
496 	case AFSR0_EL1:		*val = read_sysreg_s(SYS_AFSR0_EL12);	break;
497 	case AFSR1_EL1:		*val = read_sysreg_s(SYS_AFSR1_EL12);	break;
498 	case FAR_EL1:		*val = read_sysreg_s(SYS_FAR_EL12);	break;
499 	case MAIR_EL1:		*val = read_sysreg_s(SYS_MAIR_EL12);	break;
500 	case VBAR_EL1:		*val = read_sysreg_s(SYS_VBAR_EL12);	break;
501 	case CONTEXTIDR_EL1:	*val = read_sysreg_s(SYS_CONTEXTIDR_EL12);break;
502 	case TPIDR_EL0:		*val = read_sysreg_s(SYS_TPIDR_EL0);	break;
503 	case TPIDRRO_EL0:	*val = read_sysreg_s(SYS_TPIDRRO_EL0);	break;
504 	case TPIDR_EL1:		*val = read_sysreg_s(SYS_TPIDR_EL1);	break;
505 	case AMAIR_EL1:		*val = read_sysreg_s(SYS_AMAIR_EL12);	break;
506 	case CNTKCTL_EL1:	*val = read_sysreg_s(SYS_CNTKCTL_EL12);	break;
507 	case ELR_EL1:		*val = read_sysreg_s(SYS_ELR_EL12);	break;
508 	case PAR_EL1:		*val = read_sysreg_par();		break;
509 	case DACR32_EL2:	*val = read_sysreg_s(SYS_DACR32_EL2);	break;
510 	case IFSR32_EL2:	*val = read_sysreg_s(SYS_IFSR32_EL2);	break;
511 	case DBGVCR32_EL2:	*val = read_sysreg_s(SYS_DBGVCR32_EL2);	break;
512 	default:		return false;
513 	}
514 
515 	return true;
516 }
517 
518 static inline bool __vcpu_write_sys_reg_to_cpu(u64 val, int reg)
519 {
520 	/*
521 	 * *** VHE ONLY ***
522 	 *
523 	 * System registers listed in the switch are not restored on every
524 	 * entry to the guest but are only restored on vcpu_load.
525 	 *
526 	 * Note that MPIDR_EL1 for the guest is set by KVM via VMPIDR_EL2 but
527 	 * should never be listed below, because the MPIDR should only be set
528 	 * once, before running the VCPU, and never changed later.
529 	 */
530 	if (!has_vhe())
531 		return false;
532 
533 	switch (reg) {
534 	case CSSELR_EL1:	write_sysreg_s(val, SYS_CSSELR_EL1);	break;
535 	case SCTLR_EL1:		write_sysreg_s(val, SYS_SCTLR_EL12);	break;
536 	case CPACR_EL1:		write_sysreg_s(val, SYS_CPACR_EL12);	break;
537 	case TTBR0_EL1:		write_sysreg_s(val, SYS_TTBR0_EL12);	break;
538 	case TTBR1_EL1:		write_sysreg_s(val, SYS_TTBR1_EL12);	break;
539 	case TCR_EL1:		write_sysreg_s(val, SYS_TCR_EL12);	break;
540 	case ESR_EL1:		write_sysreg_s(val, SYS_ESR_EL12);	break;
541 	case AFSR0_EL1:		write_sysreg_s(val, SYS_AFSR0_EL12);	break;
542 	case AFSR1_EL1:		write_sysreg_s(val, SYS_AFSR1_EL12);	break;
543 	case FAR_EL1:		write_sysreg_s(val, SYS_FAR_EL12);	break;
544 	case MAIR_EL1:		write_sysreg_s(val, SYS_MAIR_EL12);	break;
545 	case VBAR_EL1:		write_sysreg_s(val, SYS_VBAR_EL12);	break;
546 	case CONTEXTIDR_EL1:	write_sysreg_s(val, SYS_CONTEXTIDR_EL12);break;
547 	case TPIDR_EL0:		write_sysreg_s(val, SYS_TPIDR_EL0);	break;
548 	case TPIDRRO_EL0:	write_sysreg_s(val, SYS_TPIDRRO_EL0);	break;
549 	case TPIDR_EL1:		write_sysreg_s(val, SYS_TPIDR_EL1);	break;
550 	case AMAIR_EL1:		write_sysreg_s(val, SYS_AMAIR_EL12);	break;
551 	case CNTKCTL_EL1:	write_sysreg_s(val, SYS_CNTKCTL_EL12);	break;
552 	case ELR_EL1:		write_sysreg_s(val, SYS_ELR_EL12);	break;
553 	case PAR_EL1:		write_sysreg_s(val, SYS_PAR_EL1);	break;
554 	case DACR32_EL2:	write_sysreg_s(val, SYS_DACR32_EL2);	break;
555 	case IFSR32_EL2:	write_sysreg_s(val, SYS_IFSR32_EL2);	break;
556 	case DBGVCR32_EL2:	write_sysreg_s(val, SYS_DBGVCR32_EL2);	break;
557 	default:		return false;
558 	}
559 
560 	return true;
561 }
562 
563 struct kvm_vm_stat {
564 	struct kvm_vm_stat_generic generic;
565 };
566 
567 struct kvm_vcpu_stat {
568 	struct kvm_vcpu_stat_generic generic;
569 	u64 hvc_exit_stat;
570 	u64 wfe_exit_stat;
571 	u64 wfi_exit_stat;
572 	u64 mmio_exit_user;
573 	u64 mmio_exit_kernel;
574 	u64 signal_exits;
575 	u64 exits;
576 };
577 
578 void kvm_vcpu_preferred_target(struct kvm_vcpu_init *init);
579 unsigned long kvm_arm_num_regs(struct kvm_vcpu *vcpu);
580 int kvm_arm_copy_reg_indices(struct kvm_vcpu *vcpu, u64 __user *indices);
581 int kvm_arm_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
582 int kvm_arm_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
583 
584 unsigned long kvm_arm_num_sys_reg_descs(struct kvm_vcpu *vcpu);
585 int kvm_arm_copy_sys_reg_indices(struct kvm_vcpu *vcpu, u64 __user *uindices);
586 int kvm_arm_sys_reg_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *);
587 int kvm_arm_sys_reg_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *);
588 
589 int __kvm_arm_vcpu_get_events(struct kvm_vcpu *vcpu,
590 			      struct kvm_vcpu_events *events);
591 
592 int __kvm_arm_vcpu_set_events(struct kvm_vcpu *vcpu,
593 			      struct kvm_vcpu_events *events);
594 
595 #define KVM_ARCH_WANT_MMU_NOTIFIER
596 
597 void kvm_arm_halt_guest(struct kvm *kvm);
598 void kvm_arm_resume_guest(struct kvm *kvm);
599 
600 #define vcpu_has_run_once(vcpu)	!!rcu_access_pointer((vcpu)->pid)
601 
602 #ifndef __KVM_NVHE_HYPERVISOR__
603 #define kvm_call_hyp_nvhe(f, ...)						\
604 	({								\
605 		struct arm_smccc_res res;				\
606 									\
607 		arm_smccc_1_1_hvc(KVM_HOST_SMCCC_FUNC(f),		\
608 				  ##__VA_ARGS__, &res);			\
609 		WARN_ON(res.a0 != SMCCC_RET_SUCCESS);			\
610 									\
611 		res.a1;							\
612 	})
613 
614 /*
615  * The couple of isb() below are there to guarantee the same behaviour
616  * on VHE as on !VHE, where the eret to EL1 acts as a context
617  * synchronization event.
618  */
619 #define kvm_call_hyp(f, ...)						\
620 	do {								\
621 		if (has_vhe()) {					\
622 			f(__VA_ARGS__);					\
623 			isb();						\
624 		} else {						\
625 			kvm_call_hyp_nvhe(f, ##__VA_ARGS__);		\
626 		}							\
627 	} while(0)
628 
629 #define kvm_call_hyp_ret(f, ...)					\
630 	({								\
631 		typeof(f(__VA_ARGS__)) ret;				\
632 									\
633 		if (has_vhe()) {					\
634 			ret = f(__VA_ARGS__);				\
635 			isb();						\
636 		} else {						\
637 			ret = kvm_call_hyp_nvhe(f, ##__VA_ARGS__);	\
638 		}							\
639 									\
640 		ret;							\
641 	})
642 #else /* __KVM_NVHE_HYPERVISOR__ */
643 #define kvm_call_hyp(f, ...) f(__VA_ARGS__)
644 #define kvm_call_hyp_ret(f, ...) f(__VA_ARGS__)
645 #define kvm_call_hyp_nvhe(f, ...) f(__VA_ARGS__)
646 #endif /* __KVM_NVHE_HYPERVISOR__ */
647 
648 void force_vm_exit(const cpumask_t *mask);
649 
650 int handle_exit(struct kvm_vcpu *vcpu, int exception_index);
651 void handle_exit_early(struct kvm_vcpu *vcpu, int exception_index);
652 
653 int kvm_handle_cp14_load_store(struct kvm_vcpu *vcpu);
654 int kvm_handle_cp14_32(struct kvm_vcpu *vcpu);
655 int kvm_handle_cp14_64(struct kvm_vcpu *vcpu);
656 int kvm_handle_cp15_32(struct kvm_vcpu *vcpu);
657 int kvm_handle_cp15_64(struct kvm_vcpu *vcpu);
658 int kvm_handle_sys_reg(struct kvm_vcpu *vcpu);
659 
660 void kvm_reset_sys_regs(struct kvm_vcpu *vcpu);
661 
662 void kvm_sys_reg_table_init(void);
663 
664 /* MMIO helpers */
665 void kvm_mmio_write_buf(void *buf, unsigned int len, unsigned long data);
666 unsigned long kvm_mmio_read_buf(const void *buf, unsigned int len);
667 
668 int kvm_handle_mmio_return(struct kvm_vcpu *vcpu);
669 int io_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa);
670 
671 /*
672  * Returns true if a Performance Monitoring Interrupt (PMI), a.k.a. perf event,
673  * arrived in guest context.  For arm64, any event that arrives while a vCPU is
674  * loaded is considered to be "in guest".
675  */
676 static inline bool kvm_arch_pmi_in_guest(struct kvm_vcpu *vcpu)
677 {
678 	return IS_ENABLED(CONFIG_GUEST_PERF_EVENTS) && !!vcpu;
679 }
680 
681 long kvm_hypercall_pv_features(struct kvm_vcpu *vcpu);
682 gpa_t kvm_init_stolen_time(struct kvm_vcpu *vcpu);
683 void kvm_update_stolen_time(struct kvm_vcpu *vcpu);
684 
685 bool kvm_arm_pvtime_supported(void);
686 int kvm_arm_pvtime_set_attr(struct kvm_vcpu *vcpu,
687 			    struct kvm_device_attr *attr);
688 int kvm_arm_pvtime_get_attr(struct kvm_vcpu *vcpu,
689 			    struct kvm_device_attr *attr);
690 int kvm_arm_pvtime_has_attr(struct kvm_vcpu *vcpu,
691 			    struct kvm_device_attr *attr);
692 
693 extern unsigned int kvm_arm_vmid_bits;
694 int kvm_arm_vmid_alloc_init(void);
695 void kvm_arm_vmid_alloc_free(void);
696 void kvm_arm_vmid_update(struct kvm_vmid *kvm_vmid);
697 
698 static inline void kvm_arm_pvtime_vcpu_init(struct kvm_vcpu_arch *vcpu_arch)
699 {
700 	vcpu_arch->steal.base = GPA_INVALID;
701 }
702 
703 static inline bool kvm_arm_is_pvtime_enabled(struct kvm_vcpu_arch *vcpu_arch)
704 {
705 	return (vcpu_arch->steal.base != GPA_INVALID);
706 }
707 
708 void kvm_set_sei_esr(struct kvm_vcpu *vcpu, u64 syndrome);
709 
710 struct kvm_vcpu *kvm_mpidr_to_vcpu(struct kvm *kvm, unsigned long mpidr);
711 
712 DECLARE_KVM_HYP_PER_CPU(struct kvm_host_data, kvm_host_data);
713 
714 static inline void kvm_init_host_cpu_context(struct kvm_cpu_context *cpu_ctxt)
715 {
716 	/* The host's MPIDR is immutable, so let's set it up at boot time */
717 	ctxt_sys_reg(cpu_ctxt, MPIDR_EL1) = read_cpuid_mpidr();
718 }
719 
720 void kvm_arm_vcpu_ptrauth_trap(struct kvm_vcpu *vcpu);
721 
722 static inline void kvm_arch_hardware_unsetup(void) {}
723 static inline void kvm_arch_sync_events(struct kvm *kvm) {}
724 static inline void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu) {}
725 
726 void kvm_arm_init_debug(void);
727 void kvm_arm_vcpu_init_debug(struct kvm_vcpu *vcpu);
728 void kvm_arm_setup_debug(struct kvm_vcpu *vcpu);
729 void kvm_arm_clear_debug(struct kvm_vcpu *vcpu);
730 void kvm_arm_reset_debug_ptr(struct kvm_vcpu *vcpu);
731 int kvm_arm_vcpu_arch_set_attr(struct kvm_vcpu *vcpu,
732 			       struct kvm_device_attr *attr);
733 int kvm_arm_vcpu_arch_get_attr(struct kvm_vcpu *vcpu,
734 			       struct kvm_device_attr *attr);
735 int kvm_arm_vcpu_arch_has_attr(struct kvm_vcpu *vcpu,
736 			       struct kvm_device_attr *attr);
737 
738 long kvm_vm_ioctl_mte_copy_tags(struct kvm *kvm,
739 				struct kvm_arm_copy_mte_tags *copy_tags);
740 
741 /* Guest/host FPSIMD coordination helpers */
742 int kvm_arch_vcpu_run_map_fp(struct kvm_vcpu *vcpu);
743 void kvm_arch_vcpu_load_fp(struct kvm_vcpu *vcpu);
744 void kvm_arch_vcpu_ctxflush_fp(struct kvm_vcpu *vcpu);
745 void kvm_arch_vcpu_ctxsync_fp(struct kvm_vcpu *vcpu);
746 void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu);
747 void kvm_vcpu_unshare_task_fp(struct kvm_vcpu *vcpu);
748 
749 static inline bool kvm_pmu_counter_deferred(struct perf_event_attr *attr)
750 {
751 	return (!has_vhe() && attr->exclude_host);
752 }
753 
754 /* Flags for host debug state */
755 void kvm_arch_vcpu_load_debug_state_flags(struct kvm_vcpu *vcpu);
756 void kvm_arch_vcpu_put_debug_state_flags(struct kvm_vcpu *vcpu);
757 
758 #ifdef CONFIG_KVM
759 void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr);
760 void kvm_clr_pmu_events(u32 clr);
761 
762 void kvm_vcpu_pmu_restore_guest(struct kvm_vcpu *vcpu);
763 void kvm_vcpu_pmu_restore_host(struct kvm_vcpu *vcpu);
764 #else
765 static inline void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr) {}
766 static inline void kvm_clr_pmu_events(u32 clr) {}
767 #endif
768 
769 void kvm_vcpu_load_sysregs_vhe(struct kvm_vcpu *vcpu);
770 void kvm_vcpu_put_sysregs_vhe(struct kvm_vcpu *vcpu);
771 
772 int kvm_set_ipa_limit(void);
773 
774 #define __KVM_HAVE_ARCH_VM_ALLOC
775 struct kvm *kvm_arch_alloc_vm(void);
776 
777 int kvm_arm_setup_stage2(struct kvm *kvm, unsigned long type);
778 
779 static inline bool kvm_vm_is_protected(struct kvm *kvm)
780 {
781 	return false;
782 }
783 
784 void kvm_init_protected_traps(struct kvm_vcpu *vcpu);
785 
786 int kvm_arm_vcpu_finalize(struct kvm_vcpu *vcpu, int feature);
787 bool kvm_arm_vcpu_is_finalized(struct kvm_vcpu *vcpu);
788 
789 #define kvm_arm_vcpu_sve_finalized(vcpu) \
790 	((vcpu)->arch.flags & KVM_ARM64_VCPU_SVE_FINALIZED)
791 
792 #define kvm_has_mte(kvm) (system_supports_mte() && (kvm)->arch.mte_enabled)
793 #define kvm_vcpu_has_pmu(vcpu)					\
794 	(test_bit(KVM_ARM_VCPU_PMU_V3, (vcpu)->arch.features))
795 
796 int kvm_trng_call(struct kvm_vcpu *vcpu);
797 #ifdef CONFIG_KVM
798 extern phys_addr_t hyp_mem_base;
799 extern phys_addr_t hyp_mem_size;
800 void __init kvm_hyp_reserve(void);
801 #else
802 static inline void kvm_hyp_reserve(void) { }
803 #endif
804 
805 #endif /* __ARM64_KVM_HOST_H__ */
806