xref: /openbmc/linux/arch/arm64/include/asm/kvm_host.h (revision 7ae5c03a)
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 #define KVM_REQ_SUSPEND		KVM_ARCH_REQ(6)
50 
51 #define KVM_DIRTY_LOG_MANUAL_CAPS   (KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE | \
52 				     KVM_DIRTY_LOG_INITIALLY_SET)
53 
54 #define KVM_HAVE_MMU_RWLOCK
55 
56 /*
57  * Mode of operation configurable with kvm-arm.mode early param.
58  * See Documentation/admin-guide/kernel-parameters.txt for more information.
59  */
60 enum kvm_mode {
61 	KVM_MODE_DEFAULT,
62 	KVM_MODE_PROTECTED,
63 	KVM_MODE_NONE,
64 };
65 enum kvm_mode kvm_get_mode(void);
66 
67 DECLARE_STATIC_KEY_FALSE(userspace_irqchip_in_use);
68 
69 extern unsigned int kvm_sve_max_vl;
70 int kvm_arm_init_sve(void);
71 
72 u32 __attribute_const__ kvm_target_cpu(void);
73 int kvm_reset_vcpu(struct kvm_vcpu *vcpu);
74 void kvm_arm_vcpu_destroy(struct kvm_vcpu *vcpu);
75 
76 struct kvm_vmid {
77 	atomic64_t id;
78 };
79 
80 struct kvm_s2_mmu {
81 	struct kvm_vmid vmid;
82 
83 	/*
84 	 * stage2 entry level table
85 	 *
86 	 * Two kvm_s2_mmu structures in the same VM can point to the same
87 	 * pgd here.  This happens when running a guest using a
88 	 * translation regime that isn't affected by its own stage-2
89 	 * translation, such as a non-VHE hypervisor running at vEL2, or
90 	 * for vEL1/EL0 with vHCR_EL2.VM == 0.  In that case, we use the
91 	 * canonical stage-2 page tables.
92 	 */
93 	phys_addr_t	pgd_phys;
94 	struct kvm_pgtable *pgt;
95 
96 	/* The last vcpu id that ran on each physical CPU */
97 	int __percpu *last_vcpu_ran;
98 
99 	struct kvm_arch *arch;
100 };
101 
102 struct kvm_arch_memory_slot {
103 };
104 
105 /**
106  * struct kvm_smccc_features: Descriptor of the hypercall services exposed to the guests
107  *
108  * @std_bmap: Bitmap of standard secure service calls
109  * @std_hyp_bmap: Bitmap of standard hypervisor service calls
110  * @vendor_hyp_bmap: Bitmap of vendor specific hypervisor service calls
111  */
112 struct kvm_smccc_features {
113 	unsigned long std_bmap;
114 	unsigned long std_hyp_bmap;
115 	unsigned long vendor_hyp_bmap;
116 };
117 
118 struct kvm_arch {
119 	struct kvm_s2_mmu mmu;
120 
121 	/* VTCR_EL2 value for this VM */
122 	u64    vtcr;
123 
124 	/* Interrupt controller */
125 	struct vgic_dist	vgic;
126 
127 	/* Mandated version of PSCI */
128 	u32 psci_version;
129 
130 	/*
131 	 * If we encounter a data abort without valid instruction syndrome
132 	 * information, report this to user space.  User space can (and
133 	 * should) opt in to this feature if KVM_CAP_ARM_NISV_TO_USER is
134 	 * supported.
135 	 */
136 #define KVM_ARCH_FLAG_RETURN_NISV_IO_ABORT_TO_USER	0
137 	/* Memory Tagging Extension enabled for the guest */
138 #define KVM_ARCH_FLAG_MTE_ENABLED			1
139 	/* At least one vCPU has ran in the VM */
140 #define KVM_ARCH_FLAG_HAS_RAN_ONCE			2
141 	/*
142 	 * The following two bits are used to indicate the guest's EL1
143 	 * register width configuration. A value of KVM_ARCH_FLAG_EL1_32BIT
144 	 * bit is valid only when KVM_ARCH_FLAG_REG_WIDTH_CONFIGURED is set.
145 	 * Otherwise, the guest's EL1 register width has not yet been
146 	 * determined yet.
147 	 */
148 #define KVM_ARCH_FLAG_REG_WIDTH_CONFIGURED		3
149 #define KVM_ARCH_FLAG_EL1_32BIT				4
150 	/* PSCI SYSTEM_SUSPEND enabled for the guest */
151 #define KVM_ARCH_FLAG_SYSTEM_SUSPEND_ENABLED		5
152 
153 	unsigned long flags;
154 
155 	/*
156 	 * VM-wide PMU filter, implemented as a bitmap and big enough for
157 	 * up to 2^10 events (ARMv8.0) or 2^16 events (ARMv8.1+).
158 	 */
159 	unsigned long *pmu_filter;
160 	struct arm_pmu *arm_pmu;
161 
162 	cpumask_var_t supported_cpus;
163 
164 	u8 pfr0_csv2;
165 	u8 pfr0_csv3;
166 
167 	/* Hypercall features firmware registers' descriptor */
168 	struct kvm_smccc_features smccc_feat;
169 };
170 
171 struct kvm_vcpu_fault_info {
172 	u64 esr_el2;		/* Hyp Syndrom Register */
173 	u64 far_el2;		/* Hyp Fault Address Register */
174 	u64 hpfar_el2;		/* Hyp IPA Fault Address Register */
175 	u64 disr_el1;		/* Deferred [SError] Status Register */
176 };
177 
178 enum vcpu_sysreg {
179 	__INVALID_SYSREG__,   /* 0 is reserved as an invalid value */
180 	MPIDR_EL1,	/* MultiProcessor Affinity Register */
181 	CSSELR_EL1,	/* Cache Size Selection Register */
182 	SCTLR_EL1,	/* System Control Register */
183 	ACTLR_EL1,	/* Auxiliary Control Register */
184 	CPACR_EL1,	/* Coprocessor Access Control */
185 	ZCR_EL1,	/* SVE Control */
186 	TTBR0_EL1,	/* Translation Table Base Register 0 */
187 	TTBR1_EL1,	/* Translation Table Base Register 1 */
188 	TCR_EL1,	/* Translation Control Register */
189 	ESR_EL1,	/* Exception Syndrome Register */
190 	AFSR0_EL1,	/* Auxiliary Fault Status Register 0 */
191 	AFSR1_EL1,	/* Auxiliary Fault Status Register 1 */
192 	FAR_EL1,	/* Fault Address Register */
193 	MAIR_EL1,	/* Memory Attribute Indirection Register */
194 	VBAR_EL1,	/* Vector Base Address Register */
195 	CONTEXTIDR_EL1,	/* Context ID Register */
196 	TPIDR_EL0,	/* Thread ID, User R/W */
197 	TPIDRRO_EL0,	/* Thread ID, User R/O */
198 	TPIDR_EL1,	/* Thread ID, Privileged */
199 	AMAIR_EL1,	/* Aux Memory Attribute Indirection Register */
200 	CNTKCTL_EL1,	/* Timer Control Register (EL1) */
201 	PAR_EL1,	/* Physical Address Register */
202 	MDSCR_EL1,	/* Monitor Debug System Control Register */
203 	MDCCINT_EL1,	/* Monitor Debug Comms Channel Interrupt Enable Reg */
204 	OSLSR_EL1,	/* OS Lock Status Register */
205 	DISR_EL1,	/* Deferred Interrupt Status Register */
206 
207 	/* Performance Monitors Registers */
208 	PMCR_EL0,	/* Control Register */
209 	PMSELR_EL0,	/* Event Counter Selection Register */
210 	PMEVCNTR0_EL0,	/* Event Counter Register (0-30) */
211 	PMEVCNTR30_EL0 = PMEVCNTR0_EL0 + 30,
212 	PMCCNTR_EL0,	/* Cycle Counter Register */
213 	PMEVTYPER0_EL0,	/* Event Type Register (0-30) */
214 	PMEVTYPER30_EL0 = PMEVTYPER0_EL0 + 30,
215 	PMCCFILTR_EL0,	/* Cycle Count Filter Register */
216 	PMCNTENSET_EL0,	/* Count Enable Set Register */
217 	PMINTENSET_EL1,	/* Interrupt Enable Set Register */
218 	PMOVSSET_EL0,	/* Overflow Flag Status Set Register */
219 	PMUSERENR_EL0,	/* User Enable Register */
220 
221 	/* Pointer Authentication Registers in a strict increasing order. */
222 	APIAKEYLO_EL1,
223 	APIAKEYHI_EL1,
224 	APIBKEYLO_EL1,
225 	APIBKEYHI_EL1,
226 	APDAKEYLO_EL1,
227 	APDAKEYHI_EL1,
228 	APDBKEYLO_EL1,
229 	APDBKEYHI_EL1,
230 	APGAKEYLO_EL1,
231 	APGAKEYHI_EL1,
232 
233 	ELR_EL1,
234 	SP_EL1,
235 	SPSR_EL1,
236 
237 	CNTVOFF_EL2,
238 	CNTV_CVAL_EL0,
239 	CNTV_CTL_EL0,
240 	CNTP_CVAL_EL0,
241 	CNTP_CTL_EL0,
242 
243 	/* Memory Tagging Extension registers */
244 	RGSR_EL1,	/* Random Allocation Tag Seed Register */
245 	GCR_EL1,	/* Tag Control Register */
246 	TFSR_EL1,	/* Tag Fault Status Register (EL1) */
247 	TFSRE0_EL1,	/* Tag Fault Status Register (EL0) */
248 
249 	/* 32bit specific registers. Keep them at the end of the range */
250 	DACR32_EL2,	/* Domain Access Control Register */
251 	IFSR32_EL2,	/* Instruction Fault Status Register */
252 	FPEXC32_EL2,	/* Floating-Point Exception Control Register */
253 	DBGVCR32_EL2,	/* Debug Vector Catch Register */
254 
255 	NR_SYS_REGS	/* Nothing after this line! */
256 };
257 
258 struct kvm_cpu_context {
259 	struct user_pt_regs regs;	/* sp = sp_el0 */
260 
261 	u64	spsr_abt;
262 	u64	spsr_und;
263 	u64	spsr_irq;
264 	u64	spsr_fiq;
265 
266 	struct user_fpsimd_state fp_regs;
267 
268 	u64 sys_regs[NR_SYS_REGS];
269 
270 	struct kvm_vcpu *__hyp_running_vcpu;
271 };
272 
273 struct kvm_host_data {
274 	struct kvm_cpu_context host_ctxt;
275 };
276 
277 struct kvm_host_psci_config {
278 	/* PSCI version used by host. */
279 	u32 version;
280 
281 	/* Function IDs used by host if version is v0.1. */
282 	struct psci_0_1_function_ids function_ids_0_1;
283 
284 	bool psci_0_1_cpu_suspend_implemented;
285 	bool psci_0_1_cpu_on_implemented;
286 	bool psci_0_1_cpu_off_implemented;
287 	bool psci_0_1_migrate_implemented;
288 };
289 
290 extern struct kvm_host_psci_config kvm_nvhe_sym(kvm_host_psci_config);
291 #define kvm_host_psci_config CHOOSE_NVHE_SYM(kvm_host_psci_config)
292 
293 extern s64 kvm_nvhe_sym(hyp_physvirt_offset);
294 #define hyp_physvirt_offset CHOOSE_NVHE_SYM(hyp_physvirt_offset)
295 
296 extern u64 kvm_nvhe_sym(hyp_cpu_logical_map)[NR_CPUS];
297 #define hyp_cpu_logical_map CHOOSE_NVHE_SYM(hyp_cpu_logical_map)
298 
299 struct vcpu_reset_state {
300 	unsigned long	pc;
301 	unsigned long	r0;
302 	bool		be;
303 	bool		reset;
304 };
305 
306 struct kvm_vcpu_arch {
307 	struct kvm_cpu_context ctxt;
308 
309 	/* Guest floating point state */
310 	void *sve_state;
311 	unsigned int sve_max_vl;
312 	u64 svcr;
313 
314 	/* Stage 2 paging state used by the hardware on next switch */
315 	struct kvm_s2_mmu *hw_mmu;
316 
317 	/* Values of trap registers for the guest. */
318 	u64 hcr_el2;
319 	u64 mdcr_el2;
320 	u64 cptr_el2;
321 
322 	/* Values of trap registers for the host before guest entry. */
323 	u64 mdcr_el2_host;
324 
325 	/* Exception Information */
326 	struct kvm_vcpu_fault_info fault;
327 
328 	/* Ownership of the FP regs */
329 	enum {
330 		FP_STATE_FREE,
331 		FP_STATE_HOST_OWNED,
332 		FP_STATE_GUEST_OWNED,
333 	} fp_state;
334 
335 	/* Configuration flags, set once and for all before the vcpu can run */
336 	u8 cflags;
337 
338 	/* Input flags to the hypervisor code, potentially cleared after use */
339 	u8 iflags;
340 
341 	/* State flags for kernel bookkeeping, unused by the hypervisor code */
342 	u8 sflags;
343 
344 	/*
345 	 * Don't run the guest (internal implementation need).
346 	 *
347 	 * Contrary to the flags above, this is set/cleared outside of
348 	 * a vcpu context, and thus cannot be mixed with the flags
349 	 * themselves (or the flag accesses need to be made atomic).
350 	 */
351 	bool pause;
352 
353 	/*
354 	 * We maintain more than a single set of debug registers to support
355 	 * debugging the guest from the host and to maintain separate host and
356 	 * guest state during world switches. vcpu_debug_state are the debug
357 	 * registers of the vcpu as the guest sees them.  host_debug_state are
358 	 * the host registers which are saved and restored during
359 	 * world switches. external_debug_state contains the debug
360 	 * values we want to debug the guest. This is set via the
361 	 * KVM_SET_GUEST_DEBUG ioctl.
362 	 *
363 	 * debug_ptr points to the set of debug registers that should be loaded
364 	 * onto the hardware when running the guest.
365 	 */
366 	struct kvm_guest_debug_arch *debug_ptr;
367 	struct kvm_guest_debug_arch vcpu_debug_state;
368 	struct kvm_guest_debug_arch external_debug_state;
369 
370 	struct user_fpsimd_state *host_fpsimd_state;	/* hyp VA */
371 	struct task_struct *parent_task;
372 
373 	struct {
374 		/* {Break,watch}point registers */
375 		struct kvm_guest_debug_arch regs;
376 		/* Statistical profiling extension */
377 		u64 pmscr_el1;
378 		/* Self-hosted trace */
379 		u64 trfcr_el1;
380 	} host_debug_state;
381 
382 	/* VGIC state */
383 	struct vgic_cpu vgic_cpu;
384 	struct arch_timer_cpu timer_cpu;
385 	struct kvm_pmu pmu;
386 
387 	/*
388 	 * Guest registers we preserve during guest debugging.
389 	 *
390 	 * These shadow registers are updated by the kvm_handle_sys_reg
391 	 * trap handler if the guest accesses or updates them while we
392 	 * are using guest debug.
393 	 */
394 	struct {
395 		u32	mdscr_el1;
396 	} guest_debug_preserved;
397 
398 	/* vcpu power state */
399 	struct kvm_mp_state mp_state;
400 
401 	/* Cache some mmu pages needed inside spinlock regions */
402 	struct kvm_mmu_memory_cache mmu_page_cache;
403 
404 	/* Target CPU and feature flags */
405 	int target;
406 	DECLARE_BITMAP(features, KVM_VCPU_MAX_FEATURES);
407 
408 	/* Virtual SError ESR to restore when HCR_EL2.VSE is set */
409 	u64 vsesr_el2;
410 
411 	/* Additional reset state */
412 	struct vcpu_reset_state	reset_state;
413 
414 	/* Guest PV state */
415 	struct {
416 		u64 last_steal;
417 		gpa_t base;
418 	} steal;
419 };
420 
421 /*
422  * Each 'flag' is composed of a comma-separated triplet:
423  *
424  * - the flag-set it belongs to in the vcpu->arch structure
425  * - the value for that flag
426  * - the mask for that flag
427  *
428  *  __vcpu_single_flag() builds such a triplet for a single-bit flag.
429  * unpack_vcpu_flag() extract the flag value from the triplet for
430  * direct use outside of the flag accessors.
431  */
432 #define __vcpu_single_flag(_set, _f)	_set, (_f), (_f)
433 
434 #define __unpack_flag(_set, _f, _m)	_f
435 #define unpack_vcpu_flag(...)		__unpack_flag(__VA_ARGS__)
436 
437 #define __build_check_flag(v, flagset, f, m)			\
438 	do {							\
439 		typeof(v->arch.flagset) *_fset;			\
440 								\
441 		/* Check that the flags fit in the mask */	\
442 		BUILD_BUG_ON(HWEIGHT(m) != HWEIGHT((f) | (m)));	\
443 		/* Check that the flags fit in the type */	\
444 		BUILD_BUG_ON((sizeof(*_fset) * 8) <= __fls(m));	\
445 	} while (0)
446 
447 #define __vcpu_get_flag(v, flagset, f, m)			\
448 	({							\
449 		__build_check_flag(v, flagset, f, m);		\
450 								\
451 		v->arch.flagset & (m);				\
452 	})
453 
454 #define __vcpu_set_flag(v, flagset, f, m)			\
455 	do {							\
456 		typeof(v->arch.flagset) *fset;			\
457 								\
458 		__build_check_flag(v, flagset, f, m);		\
459 								\
460 		fset = &v->arch.flagset;			\
461 		if (HWEIGHT(m) > 1)				\
462 			*fset &= ~(m);				\
463 		*fset |= (f);					\
464 	} while (0)
465 
466 #define __vcpu_clear_flag(v, flagset, f, m)			\
467 	do {							\
468 		typeof(v->arch.flagset) *fset;			\
469 								\
470 		__build_check_flag(v, flagset, f, m);		\
471 								\
472 		fset = &v->arch.flagset;			\
473 		*fset &= ~(m);					\
474 	} while (0)
475 
476 #define vcpu_get_flag(v, ...)	__vcpu_get_flag((v), __VA_ARGS__)
477 #define vcpu_set_flag(v, ...)	__vcpu_set_flag((v), __VA_ARGS__)
478 #define vcpu_clear_flag(v, ...)	__vcpu_clear_flag((v), __VA_ARGS__)
479 
480 /* SVE exposed to guest */
481 #define GUEST_HAS_SVE		__vcpu_single_flag(cflags, BIT(0))
482 /* SVE config completed */
483 #define VCPU_SVE_FINALIZED	__vcpu_single_flag(cflags, BIT(1))
484 /* PTRAUTH exposed to guest */
485 #define GUEST_HAS_PTRAUTH	__vcpu_single_flag(cflags, BIT(2))
486 
487 /* Exception pending */
488 #define PENDING_EXCEPTION	__vcpu_single_flag(iflags, BIT(0))
489 /*
490  * PC increment. Overlaps with EXCEPT_MASK on purpose so that it can't
491  * be set together with an exception...
492  */
493 #define INCREMENT_PC		__vcpu_single_flag(iflags, BIT(1))
494 /* Target EL/MODE (not a single flag, but let's abuse the macro) */
495 #define EXCEPT_MASK		__vcpu_single_flag(iflags, GENMASK(3, 1))
496 
497 /* Helpers to encode exceptions with minimum fuss */
498 #define __EXCEPT_MASK_VAL	unpack_vcpu_flag(EXCEPT_MASK)
499 #define __EXCEPT_SHIFT		__builtin_ctzl(__EXCEPT_MASK_VAL)
500 #define __vcpu_except_flags(_f)	iflags, (_f << __EXCEPT_SHIFT), __EXCEPT_MASK_VAL
501 
502 /*
503  * When PENDING_EXCEPTION is set, EXCEPT_MASK can take the following
504  * values:
505  *
506  * For AArch32 EL1:
507  */
508 #define EXCEPT_AA32_UND		__vcpu_except_flags(0)
509 #define EXCEPT_AA32_IABT	__vcpu_except_flags(1)
510 #define EXCEPT_AA32_DABT	__vcpu_except_flags(2)
511 /* For AArch64: */
512 #define EXCEPT_AA64_EL1_SYNC	__vcpu_except_flags(0)
513 #define EXCEPT_AA64_EL1_IRQ	__vcpu_except_flags(1)
514 #define EXCEPT_AA64_EL1_FIQ	__vcpu_except_flags(2)
515 #define EXCEPT_AA64_EL1_SERR	__vcpu_except_flags(3)
516 /* For AArch64 with NV (one day): */
517 #define EXCEPT_AA64_EL2_SYNC	__vcpu_except_flags(4)
518 #define EXCEPT_AA64_EL2_IRQ	__vcpu_except_flags(5)
519 #define EXCEPT_AA64_EL2_FIQ	__vcpu_except_flags(6)
520 #define EXCEPT_AA64_EL2_SERR	__vcpu_except_flags(7)
521 /* Guest debug is live */
522 #define DEBUG_DIRTY		__vcpu_single_flag(iflags, BIT(4))
523 /* Save SPE context if active  */
524 #define DEBUG_STATE_SAVE_SPE	__vcpu_single_flag(iflags, BIT(5))
525 /* Save TRBE context if active  */
526 #define DEBUG_STATE_SAVE_TRBE	__vcpu_single_flag(iflags, BIT(6))
527 
528 /* SVE enabled for host EL0 */
529 #define HOST_SVE_ENABLED	__vcpu_single_flag(sflags, BIT(0))
530 /* SME enabled for EL0 */
531 #define HOST_SME_ENABLED	__vcpu_single_flag(sflags, BIT(1))
532 /* Physical CPU not in supported_cpus */
533 #define ON_UNSUPPORTED_CPU	__vcpu_single_flag(sflags, BIT(2))
534 /* WFIT instruction trapped */
535 #define IN_WFIT			__vcpu_single_flag(sflags, BIT(3))
536 /* vcpu system registers loaded on physical CPU */
537 #define SYSREGS_ON_CPU		__vcpu_single_flag(sflags, BIT(4))
538 
539 /* Pointer to the vcpu's SVE FFR for sve_{save,load}_state() */
540 #define vcpu_sve_pffr(vcpu) (kern_hyp_va((vcpu)->arch.sve_state) +	\
541 			     sve_ffr_offset((vcpu)->arch.sve_max_vl))
542 
543 #define vcpu_sve_max_vq(vcpu)	sve_vq_from_vl((vcpu)->arch.sve_max_vl)
544 
545 #define vcpu_sve_state_size(vcpu) ({					\
546 	size_t __size_ret;						\
547 	unsigned int __vcpu_vq;						\
548 									\
549 	if (WARN_ON(!sve_vl_valid((vcpu)->arch.sve_max_vl))) {		\
550 		__size_ret = 0;						\
551 	} else {							\
552 		__vcpu_vq = vcpu_sve_max_vq(vcpu);			\
553 		__size_ret = SVE_SIG_REGS_SIZE(__vcpu_vq);		\
554 	}								\
555 									\
556 	__size_ret;							\
557 })
558 
559 #define KVM_GUESTDBG_VALID_MASK (KVM_GUESTDBG_ENABLE | \
560 				 KVM_GUESTDBG_USE_SW_BP | \
561 				 KVM_GUESTDBG_USE_HW | \
562 				 KVM_GUESTDBG_SINGLESTEP)
563 
564 #define vcpu_has_sve(vcpu) (system_supports_sve() &&			\
565 			    vcpu_get_flag(vcpu, GUEST_HAS_SVE))
566 
567 #ifdef CONFIG_ARM64_PTR_AUTH
568 #define vcpu_has_ptrauth(vcpu)						\
569 	((cpus_have_final_cap(ARM64_HAS_ADDRESS_AUTH) ||		\
570 	  cpus_have_final_cap(ARM64_HAS_GENERIC_AUTH)) &&		\
571 	  vcpu_get_flag(vcpu, GUEST_HAS_PTRAUTH))
572 #else
573 #define vcpu_has_ptrauth(vcpu)		false
574 #endif
575 
576 #define vcpu_on_unsupported_cpu(vcpu)					\
577 	vcpu_get_flag(vcpu, ON_UNSUPPORTED_CPU)
578 
579 #define vcpu_set_on_unsupported_cpu(vcpu)				\
580 	vcpu_set_flag(vcpu, ON_UNSUPPORTED_CPU)
581 
582 #define vcpu_clear_on_unsupported_cpu(vcpu)				\
583 	vcpu_clear_flag(vcpu, ON_UNSUPPORTED_CPU)
584 
585 #define vcpu_gp_regs(v)		(&(v)->arch.ctxt.regs)
586 
587 /*
588  * Only use __vcpu_sys_reg/ctxt_sys_reg if you know you want the
589  * memory backed version of a register, and not the one most recently
590  * accessed by a running VCPU.  For example, for userspace access or
591  * for system registers that are never context switched, but only
592  * emulated.
593  */
594 #define __ctxt_sys_reg(c,r)	(&(c)->sys_regs[(r)])
595 
596 #define ctxt_sys_reg(c,r)	(*__ctxt_sys_reg(c,r))
597 
598 #define __vcpu_sys_reg(v,r)	(ctxt_sys_reg(&(v)->arch.ctxt, (r)))
599 
600 u64 vcpu_read_sys_reg(const struct kvm_vcpu *vcpu, int reg);
601 void vcpu_write_sys_reg(struct kvm_vcpu *vcpu, u64 val, int reg);
602 
603 static inline bool __vcpu_read_sys_reg_from_cpu(int reg, u64 *val)
604 {
605 	/*
606 	 * *** VHE ONLY ***
607 	 *
608 	 * System registers listed in the switch are not saved on every
609 	 * exit from the guest but are only saved on vcpu_put.
610 	 *
611 	 * Note that MPIDR_EL1 for the guest is set by KVM via VMPIDR_EL2 but
612 	 * should never be listed below, because the guest cannot modify its
613 	 * own MPIDR_EL1 and MPIDR_EL1 is accessed for VCPU A from VCPU B's
614 	 * thread when emulating cross-VCPU communication.
615 	 */
616 	if (!has_vhe())
617 		return false;
618 
619 	switch (reg) {
620 	case CSSELR_EL1:	*val = read_sysreg_s(SYS_CSSELR_EL1);	break;
621 	case SCTLR_EL1:		*val = read_sysreg_s(SYS_SCTLR_EL12);	break;
622 	case CPACR_EL1:		*val = read_sysreg_s(SYS_CPACR_EL12);	break;
623 	case TTBR0_EL1:		*val = read_sysreg_s(SYS_TTBR0_EL12);	break;
624 	case TTBR1_EL1:		*val = read_sysreg_s(SYS_TTBR1_EL12);	break;
625 	case TCR_EL1:		*val = read_sysreg_s(SYS_TCR_EL12);	break;
626 	case ESR_EL1:		*val = read_sysreg_s(SYS_ESR_EL12);	break;
627 	case AFSR0_EL1:		*val = read_sysreg_s(SYS_AFSR0_EL12);	break;
628 	case AFSR1_EL1:		*val = read_sysreg_s(SYS_AFSR1_EL12);	break;
629 	case FAR_EL1:		*val = read_sysreg_s(SYS_FAR_EL12);	break;
630 	case MAIR_EL1:		*val = read_sysreg_s(SYS_MAIR_EL12);	break;
631 	case VBAR_EL1:		*val = read_sysreg_s(SYS_VBAR_EL12);	break;
632 	case CONTEXTIDR_EL1:	*val = read_sysreg_s(SYS_CONTEXTIDR_EL12);break;
633 	case TPIDR_EL0:		*val = read_sysreg_s(SYS_TPIDR_EL0);	break;
634 	case TPIDRRO_EL0:	*val = read_sysreg_s(SYS_TPIDRRO_EL0);	break;
635 	case TPIDR_EL1:		*val = read_sysreg_s(SYS_TPIDR_EL1);	break;
636 	case AMAIR_EL1:		*val = read_sysreg_s(SYS_AMAIR_EL12);	break;
637 	case CNTKCTL_EL1:	*val = read_sysreg_s(SYS_CNTKCTL_EL12);	break;
638 	case ELR_EL1:		*val = read_sysreg_s(SYS_ELR_EL12);	break;
639 	case PAR_EL1:		*val = read_sysreg_par();		break;
640 	case DACR32_EL2:	*val = read_sysreg_s(SYS_DACR32_EL2);	break;
641 	case IFSR32_EL2:	*val = read_sysreg_s(SYS_IFSR32_EL2);	break;
642 	case DBGVCR32_EL2:	*val = read_sysreg_s(SYS_DBGVCR32_EL2);	break;
643 	default:		return false;
644 	}
645 
646 	return true;
647 }
648 
649 static inline bool __vcpu_write_sys_reg_to_cpu(u64 val, int reg)
650 {
651 	/*
652 	 * *** VHE ONLY ***
653 	 *
654 	 * System registers listed in the switch are not restored on every
655 	 * entry to the guest but are only restored on vcpu_load.
656 	 *
657 	 * Note that MPIDR_EL1 for the guest is set by KVM via VMPIDR_EL2 but
658 	 * should never be listed below, because the MPIDR should only be set
659 	 * once, before running the VCPU, and never changed later.
660 	 */
661 	if (!has_vhe())
662 		return false;
663 
664 	switch (reg) {
665 	case CSSELR_EL1:	write_sysreg_s(val, SYS_CSSELR_EL1);	break;
666 	case SCTLR_EL1:		write_sysreg_s(val, SYS_SCTLR_EL12);	break;
667 	case CPACR_EL1:		write_sysreg_s(val, SYS_CPACR_EL12);	break;
668 	case TTBR0_EL1:		write_sysreg_s(val, SYS_TTBR0_EL12);	break;
669 	case TTBR1_EL1:		write_sysreg_s(val, SYS_TTBR1_EL12);	break;
670 	case TCR_EL1:		write_sysreg_s(val, SYS_TCR_EL12);	break;
671 	case ESR_EL1:		write_sysreg_s(val, SYS_ESR_EL12);	break;
672 	case AFSR0_EL1:		write_sysreg_s(val, SYS_AFSR0_EL12);	break;
673 	case AFSR1_EL1:		write_sysreg_s(val, SYS_AFSR1_EL12);	break;
674 	case FAR_EL1:		write_sysreg_s(val, SYS_FAR_EL12);	break;
675 	case MAIR_EL1:		write_sysreg_s(val, SYS_MAIR_EL12);	break;
676 	case VBAR_EL1:		write_sysreg_s(val, SYS_VBAR_EL12);	break;
677 	case CONTEXTIDR_EL1:	write_sysreg_s(val, SYS_CONTEXTIDR_EL12);break;
678 	case TPIDR_EL0:		write_sysreg_s(val, SYS_TPIDR_EL0);	break;
679 	case TPIDRRO_EL0:	write_sysreg_s(val, SYS_TPIDRRO_EL0);	break;
680 	case TPIDR_EL1:		write_sysreg_s(val, SYS_TPIDR_EL1);	break;
681 	case AMAIR_EL1:		write_sysreg_s(val, SYS_AMAIR_EL12);	break;
682 	case CNTKCTL_EL1:	write_sysreg_s(val, SYS_CNTKCTL_EL12);	break;
683 	case ELR_EL1:		write_sysreg_s(val, SYS_ELR_EL12);	break;
684 	case PAR_EL1:		write_sysreg_s(val, SYS_PAR_EL1);	break;
685 	case DACR32_EL2:	write_sysreg_s(val, SYS_DACR32_EL2);	break;
686 	case IFSR32_EL2:	write_sysreg_s(val, SYS_IFSR32_EL2);	break;
687 	case DBGVCR32_EL2:	write_sysreg_s(val, SYS_DBGVCR32_EL2);	break;
688 	default:		return false;
689 	}
690 
691 	return true;
692 }
693 
694 struct kvm_vm_stat {
695 	struct kvm_vm_stat_generic generic;
696 };
697 
698 struct kvm_vcpu_stat {
699 	struct kvm_vcpu_stat_generic generic;
700 	u64 hvc_exit_stat;
701 	u64 wfe_exit_stat;
702 	u64 wfi_exit_stat;
703 	u64 mmio_exit_user;
704 	u64 mmio_exit_kernel;
705 	u64 signal_exits;
706 	u64 exits;
707 };
708 
709 void kvm_vcpu_preferred_target(struct kvm_vcpu_init *init);
710 unsigned long kvm_arm_num_regs(struct kvm_vcpu *vcpu);
711 int kvm_arm_copy_reg_indices(struct kvm_vcpu *vcpu, u64 __user *indices);
712 int kvm_arm_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
713 int kvm_arm_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
714 
715 unsigned long kvm_arm_num_sys_reg_descs(struct kvm_vcpu *vcpu);
716 int kvm_arm_copy_sys_reg_indices(struct kvm_vcpu *vcpu, u64 __user *uindices);
717 
718 int __kvm_arm_vcpu_get_events(struct kvm_vcpu *vcpu,
719 			      struct kvm_vcpu_events *events);
720 
721 int __kvm_arm_vcpu_set_events(struct kvm_vcpu *vcpu,
722 			      struct kvm_vcpu_events *events);
723 
724 #define KVM_ARCH_WANT_MMU_NOTIFIER
725 
726 void kvm_arm_halt_guest(struct kvm *kvm);
727 void kvm_arm_resume_guest(struct kvm *kvm);
728 
729 #define vcpu_has_run_once(vcpu)	!!rcu_access_pointer((vcpu)->pid)
730 
731 #ifndef __KVM_NVHE_HYPERVISOR__
732 #define kvm_call_hyp_nvhe(f, ...)						\
733 	({								\
734 		struct arm_smccc_res res;				\
735 									\
736 		arm_smccc_1_1_hvc(KVM_HOST_SMCCC_FUNC(f),		\
737 				  ##__VA_ARGS__, &res);			\
738 		WARN_ON(res.a0 != SMCCC_RET_SUCCESS);			\
739 									\
740 		res.a1;							\
741 	})
742 
743 /*
744  * The couple of isb() below are there to guarantee the same behaviour
745  * on VHE as on !VHE, where the eret to EL1 acts as a context
746  * synchronization event.
747  */
748 #define kvm_call_hyp(f, ...)						\
749 	do {								\
750 		if (has_vhe()) {					\
751 			f(__VA_ARGS__);					\
752 			isb();						\
753 		} else {						\
754 			kvm_call_hyp_nvhe(f, ##__VA_ARGS__);		\
755 		}							\
756 	} while(0)
757 
758 #define kvm_call_hyp_ret(f, ...)					\
759 	({								\
760 		typeof(f(__VA_ARGS__)) ret;				\
761 									\
762 		if (has_vhe()) {					\
763 			ret = f(__VA_ARGS__);				\
764 			isb();						\
765 		} else {						\
766 			ret = kvm_call_hyp_nvhe(f, ##__VA_ARGS__);	\
767 		}							\
768 									\
769 		ret;							\
770 	})
771 #else /* __KVM_NVHE_HYPERVISOR__ */
772 #define kvm_call_hyp(f, ...) f(__VA_ARGS__)
773 #define kvm_call_hyp_ret(f, ...) f(__VA_ARGS__)
774 #define kvm_call_hyp_nvhe(f, ...) f(__VA_ARGS__)
775 #endif /* __KVM_NVHE_HYPERVISOR__ */
776 
777 void force_vm_exit(const cpumask_t *mask);
778 
779 int handle_exit(struct kvm_vcpu *vcpu, int exception_index);
780 void handle_exit_early(struct kvm_vcpu *vcpu, int exception_index);
781 
782 int kvm_handle_cp14_load_store(struct kvm_vcpu *vcpu);
783 int kvm_handle_cp14_32(struct kvm_vcpu *vcpu);
784 int kvm_handle_cp14_64(struct kvm_vcpu *vcpu);
785 int kvm_handle_cp15_32(struct kvm_vcpu *vcpu);
786 int kvm_handle_cp15_64(struct kvm_vcpu *vcpu);
787 int kvm_handle_sys_reg(struct kvm_vcpu *vcpu);
788 int kvm_handle_cp10_id(struct kvm_vcpu *vcpu);
789 
790 void kvm_reset_sys_regs(struct kvm_vcpu *vcpu);
791 
792 int kvm_sys_reg_table_init(void);
793 
794 /* MMIO helpers */
795 void kvm_mmio_write_buf(void *buf, unsigned int len, unsigned long data);
796 unsigned long kvm_mmio_read_buf(const void *buf, unsigned int len);
797 
798 int kvm_handle_mmio_return(struct kvm_vcpu *vcpu);
799 int io_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa);
800 
801 /*
802  * Returns true if a Performance Monitoring Interrupt (PMI), a.k.a. perf event,
803  * arrived in guest context.  For arm64, any event that arrives while a vCPU is
804  * loaded is considered to be "in guest".
805  */
806 static inline bool kvm_arch_pmi_in_guest(struct kvm_vcpu *vcpu)
807 {
808 	return IS_ENABLED(CONFIG_GUEST_PERF_EVENTS) && !!vcpu;
809 }
810 
811 long kvm_hypercall_pv_features(struct kvm_vcpu *vcpu);
812 gpa_t kvm_init_stolen_time(struct kvm_vcpu *vcpu);
813 void kvm_update_stolen_time(struct kvm_vcpu *vcpu);
814 
815 bool kvm_arm_pvtime_supported(void);
816 int kvm_arm_pvtime_set_attr(struct kvm_vcpu *vcpu,
817 			    struct kvm_device_attr *attr);
818 int kvm_arm_pvtime_get_attr(struct kvm_vcpu *vcpu,
819 			    struct kvm_device_attr *attr);
820 int kvm_arm_pvtime_has_attr(struct kvm_vcpu *vcpu,
821 			    struct kvm_device_attr *attr);
822 
823 extern unsigned int kvm_arm_vmid_bits;
824 int kvm_arm_vmid_alloc_init(void);
825 void kvm_arm_vmid_alloc_free(void);
826 void kvm_arm_vmid_update(struct kvm_vmid *kvm_vmid);
827 void kvm_arm_vmid_clear_active(void);
828 
829 static inline void kvm_arm_pvtime_vcpu_init(struct kvm_vcpu_arch *vcpu_arch)
830 {
831 	vcpu_arch->steal.base = GPA_INVALID;
832 }
833 
834 static inline bool kvm_arm_is_pvtime_enabled(struct kvm_vcpu_arch *vcpu_arch)
835 {
836 	return (vcpu_arch->steal.base != GPA_INVALID);
837 }
838 
839 void kvm_set_sei_esr(struct kvm_vcpu *vcpu, u64 syndrome);
840 
841 struct kvm_vcpu *kvm_mpidr_to_vcpu(struct kvm *kvm, unsigned long mpidr);
842 
843 DECLARE_KVM_HYP_PER_CPU(struct kvm_host_data, kvm_host_data);
844 
845 static inline void kvm_init_host_cpu_context(struct kvm_cpu_context *cpu_ctxt)
846 {
847 	/* The host's MPIDR is immutable, so let's set it up at boot time */
848 	ctxt_sys_reg(cpu_ctxt, MPIDR_EL1) = read_cpuid_mpidr();
849 }
850 
851 static inline bool kvm_system_needs_idmapped_vectors(void)
852 {
853 	return cpus_have_const_cap(ARM64_SPECTRE_V3A);
854 }
855 
856 void kvm_arm_vcpu_ptrauth_trap(struct kvm_vcpu *vcpu);
857 
858 static inline void kvm_arch_hardware_unsetup(void) {}
859 static inline void kvm_arch_sync_events(struct kvm *kvm) {}
860 static inline void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu) {}
861 
862 void kvm_arm_init_debug(void);
863 void kvm_arm_vcpu_init_debug(struct kvm_vcpu *vcpu);
864 void kvm_arm_setup_debug(struct kvm_vcpu *vcpu);
865 void kvm_arm_clear_debug(struct kvm_vcpu *vcpu);
866 void kvm_arm_reset_debug_ptr(struct kvm_vcpu *vcpu);
867 
868 #define kvm_vcpu_os_lock_enabled(vcpu)		\
869 	(!!(__vcpu_sys_reg(vcpu, OSLSR_EL1) & SYS_OSLSR_OSLK))
870 
871 int kvm_arm_vcpu_arch_set_attr(struct kvm_vcpu *vcpu,
872 			       struct kvm_device_attr *attr);
873 int kvm_arm_vcpu_arch_get_attr(struct kvm_vcpu *vcpu,
874 			       struct kvm_device_attr *attr);
875 int kvm_arm_vcpu_arch_has_attr(struct kvm_vcpu *vcpu,
876 			       struct kvm_device_attr *attr);
877 
878 long kvm_vm_ioctl_mte_copy_tags(struct kvm *kvm,
879 				struct kvm_arm_copy_mte_tags *copy_tags);
880 
881 /* Guest/host FPSIMD coordination helpers */
882 int kvm_arch_vcpu_run_map_fp(struct kvm_vcpu *vcpu);
883 void kvm_arch_vcpu_load_fp(struct kvm_vcpu *vcpu);
884 void kvm_arch_vcpu_ctxflush_fp(struct kvm_vcpu *vcpu);
885 void kvm_arch_vcpu_ctxsync_fp(struct kvm_vcpu *vcpu);
886 void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu);
887 void kvm_vcpu_unshare_task_fp(struct kvm_vcpu *vcpu);
888 
889 static inline bool kvm_pmu_counter_deferred(struct perf_event_attr *attr)
890 {
891 	return (!has_vhe() && attr->exclude_host);
892 }
893 
894 /* Flags for host debug state */
895 void kvm_arch_vcpu_load_debug_state_flags(struct kvm_vcpu *vcpu);
896 void kvm_arch_vcpu_put_debug_state_flags(struct kvm_vcpu *vcpu);
897 
898 #ifdef CONFIG_KVM
899 void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr);
900 void kvm_clr_pmu_events(u32 clr);
901 #else
902 static inline void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr) {}
903 static inline void kvm_clr_pmu_events(u32 clr) {}
904 #endif
905 
906 void kvm_vcpu_load_sysregs_vhe(struct kvm_vcpu *vcpu);
907 void kvm_vcpu_put_sysregs_vhe(struct kvm_vcpu *vcpu);
908 
909 int kvm_set_ipa_limit(void);
910 
911 #define __KVM_HAVE_ARCH_VM_ALLOC
912 struct kvm *kvm_arch_alloc_vm(void);
913 
914 int kvm_arm_setup_stage2(struct kvm *kvm, unsigned long type);
915 
916 static inline bool kvm_vm_is_protected(struct kvm *kvm)
917 {
918 	return false;
919 }
920 
921 void kvm_init_protected_traps(struct kvm_vcpu *vcpu);
922 
923 int kvm_arm_vcpu_finalize(struct kvm_vcpu *vcpu, int feature);
924 bool kvm_arm_vcpu_is_finalized(struct kvm_vcpu *vcpu);
925 
926 #define kvm_arm_vcpu_sve_finalized(vcpu) vcpu_get_flag(vcpu, VCPU_SVE_FINALIZED)
927 
928 #define kvm_has_mte(kvm)					\
929 	(system_supports_mte() &&				\
930 	 test_bit(KVM_ARCH_FLAG_MTE_ENABLED, &(kvm)->arch.flags))
931 
932 int kvm_trng_call(struct kvm_vcpu *vcpu);
933 #ifdef CONFIG_KVM
934 extern phys_addr_t hyp_mem_base;
935 extern phys_addr_t hyp_mem_size;
936 void __init kvm_hyp_reserve(void);
937 #else
938 static inline void kvm_hyp_reserve(void) { }
939 #endif
940 
941 void kvm_arm_vcpu_power_off(struct kvm_vcpu *vcpu);
942 bool kvm_arm_vcpu_stopped(struct kvm_vcpu *vcpu);
943 
944 #endif /* __ARM64_KVM_HOST_H__ */
945