xref: /openbmc/linux/arch/arm64/include/asm/kvm_host.h (revision 151f4e2b)
1 /*
2  * Copyright (C) 2012,2013 - ARM Ltd
3  * Author: Marc Zyngier <marc.zyngier@arm.com>
4  *
5  * Derived from arch/arm/include/asm/kvm_host.h:
6  * Copyright (C) 2012 - Virtual Open Systems and Columbia University
7  * Author: Christoffer Dall <c.dall@virtualopensystems.com>
8  *
9  * This program is free software; you can redistribute it and/or modify
10  * it under the terms of the GNU General Public License version 2 as
11  * published by the Free Software Foundation.
12  *
13  * This program is distributed in the hope that it will be useful,
14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
16  * GNU General Public License for more details.
17  *
18  * You should have received a copy of the GNU General Public License
19  * along with this program.  If not, see <http://www.gnu.org/licenses/>.
20  */
21 
22 #ifndef __ARM64_KVM_HOST_H__
23 #define __ARM64_KVM_HOST_H__
24 
25 #include <linux/bitmap.h>
26 #include <linux/types.h>
27 #include <linux/jump_label.h>
28 #include <linux/kvm_types.h>
29 #include <linux/percpu.h>
30 #include <asm/arch_gicv3.h>
31 #include <asm/barrier.h>
32 #include <asm/cpufeature.h>
33 #include <asm/daifflags.h>
34 #include <asm/fpsimd.h>
35 #include <asm/kvm.h>
36 #include <asm/kvm_asm.h>
37 #include <asm/kvm_mmio.h>
38 #include <asm/smp_plat.h>
39 #include <asm/thread_info.h>
40 
41 #define __KVM_HAVE_ARCH_INTC_INITIALIZED
42 
43 #define KVM_USER_MEM_SLOTS 512
44 #define KVM_HALT_POLL_NS_DEFAULT 500000
45 
46 #include <kvm/arm_vgic.h>
47 #include <kvm/arm_arch_timer.h>
48 #include <kvm/arm_pmu.h>
49 
50 #define KVM_MAX_VCPUS VGIC_V3_MAX_CPUS
51 
52 #define KVM_VCPU_MAX_FEATURES 7
53 
54 #define KVM_REQ_SLEEP \
55 	KVM_ARCH_REQ_FLAGS(0, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
56 #define KVM_REQ_IRQ_PENDING	KVM_ARCH_REQ(1)
57 #define KVM_REQ_VCPU_RESET	KVM_ARCH_REQ(2)
58 
59 DECLARE_STATIC_KEY_FALSE(userspace_irqchip_in_use);
60 
61 extern unsigned int kvm_sve_max_vl;
62 int kvm_arm_init_sve(void);
63 
64 int __attribute_const__ kvm_target_cpu(void);
65 int kvm_reset_vcpu(struct kvm_vcpu *vcpu);
66 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
67 int kvm_arch_vm_ioctl_check_extension(struct kvm *kvm, long ext);
68 void __extended_idmap_trampoline(phys_addr_t boot_pgd, phys_addr_t idmap_start);
69 
70 struct kvm_vmid {
71 	/* The VMID generation used for the virt. memory system */
72 	u64    vmid_gen;
73 	u32    vmid;
74 };
75 
76 struct kvm_arch {
77 	struct kvm_vmid vmid;
78 
79 	/* stage2 entry level table */
80 	pgd_t *pgd;
81 	phys_addr_t pgd_phys;
82 
83 	/* VTCR_EL2 value for this VM */
84 	u64    vtcr;
85 
86 	/* The last vcpu id that ran on each physical CPU */
87 	int __percpu *last_vcpu_ran;
88 
89 	/* The maximum number of vCPUs depends on the used GIC model */
90 	int max_vcpus;
91 
92 	/* Interrupt controller */
93 	struct vgic_dist	vgic;
94 
95 	/* Mandated version of PSCI */
96 	u32 psci_version;
97 };
98 
99 #define KVM_NR_MEM_OBJS     40
100 
101 /*
102  * We don't want allocation failures within the mmu code, so we preallocate
103  * enough memory for a single page fault in a cache.
104  */
105 struct kvm_mmu_memory_cache {
106 	int nobjs;
107 	void *objects[KVM_NR_MEM_OBJS];
108 };
109 
110 struct kvm_vcpu_fault_info {
111 	u32 esr_el2;		/* Hyp Syndrom Register */
112 	u64 far_el2;		/* Hyp Fault Address Register */
113 	u64 hpfar_el2;		/* Hyp IPA Fault Address Register */
114 	u64 disr_el1;		/* Deferred [SError] Status Register */
115 };
116 
117 /*
118  * 0 is reserved as an invalid value.
119  * Order should be kept in sync with the save/restore code.
120  */
121 enum vcpu_sysreg {
122 	__INVALID_SYSREG__,
123 	MPIDR_EL1,	/* MultiProcessor Affinity Register */
124 	CSSELR_EL1,	/* Cache Size Selection Register */
125 	SCTLR_EL1,	/* System Control Register */
126 	ACTLR_EL1,	/* Auxiliary Control Register */
127 	CPACR_EL1,	/* Coprocessor Access Control */
128 	ZCR_EL1,	/* SVE Control */
129 	TTBR0_EL1,	/* Translation Table Base Register 0 */
130 	TTBR1_EL1,	/* Translation Table Base Register 1 */
131 	TCR_EL1,	/* Translation Control Register */
132 	ESR_EL1,	/* Exception Syndrome Register */
133 	AFSR0_EL1,	/* Auxiliary Fault Status Register 0 */
134 	AFSR1_EL1,	/* Auxiliary Fault Status Register 1 */
135 	FAR_EL1,	/* Fault Address Register */
136 	MAIR_EL1,	/* Memory Attribute Indirection Register */
137 	VBAR_EL1,	/* Vector Base Address Register */
138 	CONTEXTIDR_EL1,	/* Context ID Register */
139 	TPIDR_EL0,	/* Thread ID, User R/W */
140 	TPIDRRO_EL0,	/* Thread ID, User R/O */
141 	TPIDR_EL1,	/* Thread ID, Privileged */
142 	AMAIR_EL1,	/* Aux Memory Attribute Indirection Register */
143 	CNTKCTL_EL1,	/* Timer Control Register (EL1) */
144 	PAR_EL1,	/* Physical Address Register */
145 	MDSCR_EL1,	/* Monitor Debug System Control Register */
146 	MDCCINT_EL1,	/* Monitor Debug Comms Channel Interrupt Enable Reg */
147 	DISR_EL1,	/* Deferred Interrupt Status Register */
148 
149 	/* Performance Monitors Registers */
150 	PMCR_EL0,	/* Control Register */
151 	PMSELR_EL0,	/* Event Counter Selection Register */
152 	PMEVCNTR0_EL0,	/* Event Counter Register (0-30) */
153 	PMEVCNTR30_EL0 = PMEVCNTR0_EL0 + 30,
154 	PMCCNTR_EL0,	/* Cycle Counter Register */
155 	PMEVTYPER0_EL0,	/* Event Type Register (0-30) */
156 	PMEVTYPER30_EL0 = PMEVTYPER0_EL0 + 30,
157 	PMCCFILTR_EL0,	/* Cycle Count Filter Register */
158 	PMCNTENSET_EL0,	/* Count Enable Set Register */
159 	PMINTENSET_EL1,	/* Interrupt Enable Set Register */
160 	PMOVSSET_EL0,	/* Overflow Flag Status Set Register */
161 	PMSWINC_EL0,	/* Software Increment Register */
162 	PMUSERENR_EL0,	/* User Enable Register */
163 
164 	/* Pointer Authentication Registers in a strict increasing order. */
165 	APIAKEYLO_EL1,
166 	APIAKEYHI_EL1,
167 	APIBKEYLO_EL1,
168 	APIBKEYHI_EL1,
169 	APDAKEYLO_EL1,
170 	APDAKEYHI_EL1,
171 	APDBKEYLO_EL1,
172 	APDBKEYHI_EL1,
173 	APGAKEYLO_EL1,
174 	APGAKEYHI_EL1,
175 
176 	/* 32bit specific registers. Keep them at the end of the range */
177 	DACR32_EL2,	/* Domain Access Control Register */
178 	IFSR32_EL2,	/* Instruction Fault Status Register */
179 	FPEXC32_EL2,	/* Floating-Point Exception Control Register */
180 	DBGVCR32_EL2,	/* Debug Vector Catch Register */
181 
182 	NR_SYS_REGS	/* Nothing after this line! */
183 };
184 
185 /* 32bit mapping */
186 #define c0_MPIDR	(MPIDR_EL1 * 2)	/* MultiProcessor ID Register */
187 #define c0_CSSELR	(CSSELR_EL1 * 2)/* Cache Size Selection Register */
188 #define c1_SCTLR	(SCTLR_EL1 * 2)	/* System Control Register */
189 #define c1_ACTLR	(ACTLR_EL1 * 2)	/* Auxiliary Control Register */
190 #define c1_CPACR	(CPACR_EL1 * 2)	/* Coprocessor Access Control */
191 #define c2_TTBR0	(TTBR0_EL1 * 2)	/* Translation Table Base Register 0 */
192 #define c2_TTBR0_high	(c2_TTBR0 + 1)	/* TTBR0 top 32 bits */
193 #define c2_TTBR1	(TTBR1_EL1 * 2)	/* Translation Table Base Register 1 */
194 #define c2_TTBR1_high	(c2_TTBR1 + 1)	/* TTBR1 top 32 bits */
195 #define c2_TTBCR	(TCR_EL1 * 2)	/* Translation Table Base Control R. */
196 #define c3_DACR		(DACR32_EL2 * 2)/* Domain Access Control Register */
197 #define c5_DFSR		(ESR_EL1 * 2)	/* Data Fault Status Register */
198 #define c5_IFSR		(IFSR32_EL2 * 2)/* Instruction Fault Status Register */
199 #define c5_ADFSR	(AFSR0_EL1 * 2)	/* Auxiliary Data Fault Status R */
200 #define c5_AIFSR	(AFSR1_EL1 * 2)	/* Auxiliary Instr Fault Status R */
201 #define c6_DFAR		(FAR_EL1 * 2)	/* Data Fault Address Register */
202 #define c6_IFAR		(c6_DFAR + 1)	/* Instruction Fault Address Register */
203 #define c7_PAR		(PAR_EL1 * 2)	/* Physical Address Register */
204 #define c7_PAR_high	(c7_PAR + 1)	/* PAR top 32 bits */
205 #define c10_PRRR	(MAIR_EL1 * 2)	/* Primary Region Remap Register */
206 #define c10_NMRR	(c10_PRRR + 1)	/* Normal Memory Remap Register */
207 #define c12_VBAR	(VBAR_EL1 * 2)	/* Vector Base Address Register */
208 #define c13_CID		(CONTEXTIDR_EL1 * 2)	/* Context ID Register */
209 #define c13_TID_URW	(TPIDR_EL0 * 2)	/* Thread ID, User R/W */
210 #define c13_TID_URO	(TPIDRRO_EL0 * 2)/* Thread ID, User R/O */
211 #define c13_TID_PRIV	(TPIDR_EL1 * 2)	/* Thread ID, Privileged */
212 #define c10_AMAIR0	(AMAIR_EL1 * 2)	/* Aux Memory Attr Indirection Reg */
213 #define c10_AMAIR1	(c10_AMAIR0 + 1)/* Aux Memory Attr Indirection Reg */
214 #define c14_CNTKCTL	(CNTKCTL_EL1 * 2) /* Timer Control Register (PL1) */
215 
216 #define cp14_DBGDSCRext	(MDSCR_EL1 * 2)
217 #define cp14_DBGBCR0	(DBGBCR0_EL1 * 2)
218 #define cp14_DBGBVR0	(DBGBVR0_EL1 * 2)
219 #define cp14_DBGBXVR0	(cp14_DBGBVR0 + 1)
220 #define cp14_DBGWCR0	(DBGWCR0_EL1 * 2)
221 #define cp14_DBGWVR0	(DBGWVR0_EL1 * 2)
222 #define cp14_DBGDCCINT	(MDCCINT_EL1 * 2)
223 
224 #define NR_COPRO_REGS	(NR_SYS_REGS * 2)
225 
226 struct kvm_cpu_context {
227 	struct kvm_regs	gp_regs;
228 	union {
229 		u64 sys_regs[NR_SYS_REGS];
230 		u32 copro[NR_COPRO_REGS];
231 	};
232 
233 	struct kvm_vcpu *__hyp_running_vcpu;
234 };
235 
236 struct kvm_pmu_events {
237 	u32 events_host;
238 	u32 events_guest;
239 };
240 
241 struct kvm_host_data {
242 	struct kvm_cpu_context host_ctxt;
243 	struct kvm_pmu_events pmu_events;
244 };
245 
246 typedef struct kvm_host_data kvm_host_data_t;
247 
248 struct vcpu_reset_state {
249 	unsigned long	pc;
250 	unsigned long	r0;
251 	bool		be;
252 	bool		reset;
253 };
254 
255 struct kvm_vcpu_arch {
256 	struct kvm_cpu_context ctxt;
257 	void *sve_state;
258 	unsigned int sve_max_vl;
259 
260 	/* HYP configuration */
261 	u64 hcr_el2;
262 	u32 mdcr_el2;
263 
264 	/* Exception Information */
265 	struct kvm_vcpu_fault_info fault;
266 
267 	/* State of various workarounds, see kvm_asm.h for bit assignment */
268 	u64 workaround_flags;
269 
270 	/* Miscellaneous vcpu state flags */
271 	u64 flags;
272 
273 	/*
274 	 * We maintain more than a single set of debug registers to support
275 	 * debugging the guest from the host and to maintain separate host and
276 	 * guest state during world switches. vcpu_debug_state are the debug
277 	 * registers of the vcpu as the guest sees them.  host_debug_state are
278 	 * the host registers which are saved and restored during
279 	 * world switches. external_debug_state contains the debug
280 	 * values we want to debug the guest. This is set via the
281 	 * KVM_SET_GUEST_DEBUG ioctl.
282 	 *
283 	 * debug_ptr points to the set of debug registers that should be loaded
284 	 * onto the hardware when running the guest.
285 	 */
286 	struct kvm_guest_debug_arch *debug_ptr;
287 	struct kvm_guest_debug_arch vcpu_debug_state;
288 	struct kvm_guest_debug_arch external_debug_state;
289 
290 	/* Pointer to host CPU context */
291 	struct kvm_cpu_context *host_cpu_context;
292 
293 	struct thread_info *host_thread_info;	/* hyp VA */
294 	struct user_fpsimd_state *host_fpsimd_state;	/* hyp VA */
295 
296 	struct {
297 		/* {Break,watch}point registers */
298 		struct kvm_guest_debug_arch regs;
299 		/* Statistical profiling extension */
300 		u64 pmscr_el1;
301 	} host_debug_state;
302 
303 	/* VGIC state */
304 	struct vgic_cpu vgic_cpu;
305 	struct arch_timer_cpu timer_cpu;
306 	struct kvm_pmu pmu;
307 
308 	/*
309 	 * Anything that is not used directly from assembly code goes
310 	 * here.
311 	 */
312 
313 	/*
314 	 * Guest registers we preserve during guest debugging.
315 	 *
316 	 * These shadow registers are updated by the kvm_handle_sys_reg
317 	 * trap handler if the guest accesses or updates them while we
318 	 * are using guest debug.
319 	 */
320 	struct {
321 		u32	mdscr_el1;
322 	} guest_debug_preserved;
323 
324 	/* vcpu power-off state */
325 	bool power_off;
326 
327 	/* Don't run the guest (internal implementation need) */
328 	bool pause;
329 
330 	/* IO related fields */
331 	struct kvm_decode mmio_decode;
332 
333 	/* Cache some mmu pages needed inside spinlock regions */
334 	struct kvm_mmu_memory_cache mmu_page_cache;
335 
336 	/* Target CPU and feature flags */
337 	int target;
338 	DECLARE_BITMAP(features, KVM_VCPU_MAX_FEATURES);
339 
340 	/* Detect first run of a vcpu */
341 	bool has_run_once;
342 
343 	/* Virtual SError ESR to restore when HCR_EL2.VSE is set */
344 	u64 vsesr_el2;
345 
346 	/* Additional reset state */
347 	struct vcpu_reset_state	reset_state;
348 
349 	/* True when deferrable sysregs are loaded on the physical CPU,
350 	 * see kvm_vcpu_load_sysregs and kvm_vcpu_put_sysregs. */
351 	bool sysregs_loaded_on_cpu;
352 };
353 
354 /* Pointer to the vcpu's SVE FFR for sve_{save,load}_state() */
355 #define vcpu_sve_pffr(vcpu) ((void *)((char *)((vcpu)->arch.sve_state) + \
356 				      sve_ffr_offset((vcpu)->arch.sve_max_vl)))
357 
358 #define vcpu_sve_state_size(vcpu) ({					\
359 	size_t __size_ret;						\
360 	unsigned int __vcpu_vq;						\
361 									\
362 	if (WARN_ON(!sve_vl_valid((vcpu)->arch.sve_max_vl))) {		\
363 		__size_ret = 0;						\
364 	} else {							\
365 		__vcpu_vq = sve_vq_from_vl((vcpu)->arch.sve_max_vl);	\
366 		__size_ret = SVE_SIG_REGS_SIZE(__vcpu_vq);		\
367 	}								\
368 									\
369 	__size_ret;							\
370 })
371 
372 /* vcpu_arch flags field values: */
373 #define KVM_ARM64_DEBUG_DIRTY		(1 << 0)
374 #define KVM_ARM64_FP_ENABLED		(1 << 1) /* guest FP regs loaded */
375 #define KVM_ARM64_FP_HOST		(1 << 2) /* host FP regs loaded */
376 #define KVM_ARM64_HOST_SVE_IN_USE	(1 << 3) /* backup for host TIF_SVE */
377 #define KVM_ARM64_HOST_SVE_ENABLED	(1 << 4) /* SVE enabled for EL0 */
378 #define KVM_ARM64_GUEST_HAS_SVE		(1 << 5) /* SVE exposed to guest */
379 #define KVM_ARM64_VCPU_SVE_FINALIZED	(1 << 6) /* SVE config completed */
380 #define KVM_ARM64_GUEST_HAS_PTRAUTH	(1 << 7) /* PTRAUTH exposed to guest */
381 
382 #define vcpu_has_sve(vcpu) (system_supports_sve() && \
383 			    ((vcpu)->arch.flags & KVM_ARM64_GUEST_HAS_SVE))
384 
385 #define vcpu_has_ptrauth(vcpu)	((system_supports_address_auth() || \
386 				  system_supports_generic_auth()) && \
387 				 ((vcpu)->arch.flags & KVM_ARM64_GUEST_HAS_PTRAUTH))
388 
389 #define vcpu_gp_regs(v)		(&(v)->arch.ctxt.gp_regs)
390 
391 /*
392  * Only use __vcpu_sys_reg if you know you want the memory backed version of a
393  * register, and not the one most recently accessed by a running VCPU.  For
394  * example, for userspace access or for system registers that are never context
395  * switched, but only emulated.
396  */
397 #define __vcpu_sys_reg(v,r)	((v)->arch.ctxt.sys_regs[(r)])
398 
399 u64 vcpu_read_sys_reg(const struct kvm_vcpu *vcpu, int reg);
400 void vcpu_write_sys_reg(struct kvm_vcpu *vcpu, u64 val, int reg);
401 
402 /*
403  * CP14 and CP15 live in the same array, as they are backed by the
404  * same system registers.
405  */
406 #define vcpu_cp14(v,r)		((v)->arch.ctxt.copro[(r)])
407 #define vcpu_cp15(v,r)		((v)->arch.ctxt.copro[(r)])
408 
409 struct kvm_vm_stat {
410 	ulong remote_tlb_flush;
411 };
412 
413 struct kvm_vcpu_stat {
414 	u64 halt_successful_poll;
415 	u64 halt_attempted_poll;
416 	u64 halt_poll_invalid;
417 	u64 halt_wakeup;
418 	u64 hvc_exit_stat;
419 	u64 wfe_exit_stat;
420 	u64 wfi_exit_stat;
421 	u64 mmio_exit_user;
422 	u64 mmio_exit_kernel;
423 	u64 exits;
424 };
425 
426 int kvm_vcpu_preferred_target(struct kvm_vcpu_init *init);
427 unsigned long kvm_arm_num_regs(struct kvm_vcpu *vcpu);
428 int kvm_arm_copy_reg_indices(struct kvm_vcpu *vcpu, u64 __user *indices);
429 int kvm_arm_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
430 int kvm_arm_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
431 int __kvm_arm_vcpu_get_events(struct kvm_vcpu *vcpu,
432 			      struct kvm_vcpu_events *events);
433 
434 int __kvm_arm_vcpu_set_events(struct kvm_vcpu *vcpu,
435 			      struct kvm_vcpu_events *events);
436 
437 #define KVM_ARCH_WANT_MMU_NOTIFIER
438 int kvm_unmap_hva_range(struct kvm *kvm,
439 			unsigned long start, unsigned long end);
440 int kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
441 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
442 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
443 
444 struct kvm_vcpu *kvm_arm_get_running_vcpu(void);
445 struct kvm_vcpu * __percpu *kvm_get_running_vcpus(void);
446 void kvm_arm_halt_guest(struct kvm *kvm);
447 void kvm_arm_resume_guest(struct kvm *kvm);
448 
449 u64 __kvm_call_hyp(void *hypfn, ...);
450 
451 /*
452  * The couple of isb() below are there to guarantee the same behaviour
453  * on VHE as on !VHE, where the eret to EL1 acts as a context
454  * synchronization event.
455  */
456 #define kvm_call_hyp(f, ...)						\
457 	do {								\
458 		if (has_vhe()) {					\
459 			f(__VA_ARGS__);					\
460 			isb();						\
461 		} else {						\
462 			__kvm_call_hyp(kvm_ksym_ref(f), ##__VA_ARGS__); \
463 		}							\
464 	} while(0)
465 
466 #define kvm_call_hyp_ret(f, ...)					\
467 	({								\
468 		typeof(f(__VA_ARGS__)) ret;				\
469 									\
470 		if (has_vhe()) {					\
471 			ret = f(__VA_ARGS__);				\
472 			isb();						\
473 		} else {						\
474 			ret = __kvm_call_hyp(kvm_ksym_ref(f),		\
475 					     ##__VA_ARGS__);		\
476 		}							\
477 									\
478 		ret;							\
479 	})
480 
481 void force_vm_exit(const cpumask_t *mask);
482 void kvm_mmu_wp_memory_region(struct kvm *kvm, int slot);
483 
484 int handle_exit(struct kvm_vcpu *vcpu, struct kvm_run *run,
485 		int exception_index);
486 void handle_exit_early(struct kvm_vcpu *vcpu, struct kvm_run *run,
487 		       int exception_index);
488 
489 int kvm_perf_init(void);
490 int kvm_perf_teardown(void);
491 
492 void kvm_set_sei_esr(struct kvm_vcpu *vcpu, u64 syndrome);
493 
494 struct kvm_vcpu *kvm_mpidr_to_vcpu(struct kvm *kvm, unsigned long mpidr);
495 
496 DECLARE_PER_CPU(kvm_host_data_t, kvm_host_data);
497 
498 static inline void kvm_init_host_cpu_context(struct kvm_cpu_context *cpu_ctxt,
499 					     int cpu)
500 {
501 	/* The host's MPIDR is immutable, so let's set it up at boot time */
502 	cpu_ctxt->sys_regs[MPIDR_EL1] = cpu_logical_map(cpu);
503 }
504 
505 void __kvm_enable_ssbs(void);
506 
507 static inline void __cpu_init_hyp_mode(phys_addr_t pgd_ptr,
508 				       unsigned long hyp_stack_ptr,
509 				       unsigned long vector_ptr)
510 {
511 	/*
512 	 * Calculate the raw per-cpu offset without a translation from the
513 	 * kernel's mapping to the linear mapping, and store it in tpidr_el2
514 	 * so that we can use adr_l to access per-cpu variables in EL2.
515 	 */
516 	u64 tpidr_el2 = ((u64)this_cpu_ptr(&kvm_host_data) -
517 			 (u64)kvm_ksym_ref(kvm_host_data));
518 
519 	/*
520 	 * Call initialization code, and switch to the full blown HYP code.
521 	 * If the cpucaps haven't been finalized yet, something has gone very
522 	 * wrong, and hyp will crash and burn when it uses any
523 	 * cpus_have_const_cap() wrapper.
524 	 */
525 	BUG_ON(!static_branch_likely(&arm64_const_caps_ready));
526 	__kvm_call_hyp((void *)pgd_ptr, hyp_stack_ptr, vector_ptr, tpidr_el2);
527 
528 	/*
529 	 * Disabling SSBD on a non-VHE system requires us to enable SSBS
530 	 * at EL2.
531 	 */
532 	if (!has_vhe() && this_cpu_has_cap(ARM64_SSBS) &&
533 	    arm64_get_ssbd_state() == ARM64_SSBD_FORCE_DISABLE) {
534 		kvm_call_hyp(__kvm_enable_ssbs);
535 	}
536 }
537 
538 static inline bool kvm_arch_requires_vhe(void)
539 {
540 	/*
541 	 * The Arm architecture specifies that implementation of SVE
542 	 * requires VHE also to be implemented.  The KVM code for arm64
543 	 * relies on this when SVE is present:
544 	 */
545 	if (system_supports_sve())
546 		return true;
547 
548 	/* Some implementations have defects that confine them to VHE */
549 	if (cpus_have_cap(ARM64_WORKAROUND_1165522))
550 		return true;
551 
552 	return false;
553 }
554 
555 void kvm_arm_vcpu_ptrauth_trap(struct kvm_vcpu *vcpu);
556 
557 static inline void kvm_arch_hardware_unsetup(void) {}
558 static inline void kvm_arch_sync_events(struct kvm *kvm) {}
559 static inline void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu) {}
560 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
561 
562 void kvm_arm_init_debug(void);
563 void kvm_arm_setup_debug(struct kvm_vcpu *vcpu);
564 void kvm_arm_clear_debug(struct kvm_vcpu *vcpu);
565 void kvm_arm_reset_debug_ptr(struct kvm_vcpu *vcpu);
566 int kvm_arm_vcpu_arch_set_attr(struct kvm_vcpu *vcpu,
567 			       struct kvm_device_attr *attr);
568 int kvm_arm_vcpu_arch_get_attr(struct kvm_vcpu *vcpu,
569 			       struct kvm_device_attr *attr);
570 int kvm_arm_vcpu_arch_has_attr(struct kvm_vcpu *vcpu,
571 			       struct kvm_device_attr *attr);
572 
573 static inline void __cpu_init_stage2(void) {}
574 
575 /* Guest/host FPSIMD coordination helpers */
576 int kvm_arch_vcpu_run_map_fp(struct kvm_vcpu *vcpu);
577 void kvm_arch_vcpu_load_fp(struct kvm_vcpu *vcpu);
578 void kvm_arch_vcpu_ctxsync_fp(struct kvm_vcpu *vcpu);
579 void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu);
580 
581 static inline bool kvm_pmu_counter_deferred(struct perf_event_attr *attr)
582 {
583 	return (!has_vhe() && attr->exclude_host);
584 }
585 
586 #ifdef CONFIG_KVM /* Avoid conflicts with core headers if CONFIG_KVM=n */
587 static inline int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu)
588 {
589 	return kvm_arch_vcpu_run_map_fp(vcpu);
590 }
591 
592 void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr);
593 void kvm_clr_pmu_events(u32 clr);
594 
595 void __pmu_switch_to_host(struct kvm_cpu_context *host_ctxt);
596 bool __pmu_switch_to_guest(struct kvm_cpu_context *host_ctxt);
597 
598 void kvm_vcpu_pmu_restore_guest(struct kvm_vcpu *vcpu);
599 void kvm_vcpu_pmu_restore_host(struct kvm_vcpu *vcpu);
600 #else
601 static inline void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr) {}
602 static inline void kvm_clr_pmu_events(u32 clr) {}
603 #endif
604 
605 static inline void kvm_arm_vhe_guest_enter(void)
606 {
607 	local_daif_mask();
608 
609 	/*
610 	 * Having IRQs masked via PMR when entering the guest means the GIC
611 	 * will not signal the CPU of interrupts of lower priority, and the
612 	 * only way to get out will be via guest exceptions.
613 	 * Naturally, we want to avoid this.
614 	 */
615 	if (system_uses_irq_prio_masking()) {
616 		gic_write_pmr(GIC_PRIO_IRQON);
617 		dsb(sy);
618 	}
619 }
620 
621 static inline void kvm_arm_vhe_guest_exit(void)
622 {
623 	/*
624 	 * local_daif_restore() takes care to properly restore PSTATE.DAIF
625 	 * and the GIC PMR if the host is using IRQ priorities.
626 	 */
627 	local_daif_restore(DAIF_PROCCTX_NOIRQ);
628 
629 	/*
630 	 * When we exit from the guest we change a number of CPU configuration
631 	 * parameters, such as traps.  Make sure these changes take effect
632 	 * before running the host or additional guests.
633 	 */
634 	isb();
635 }
636 
637 static inline bool kvm_arm_harden_branch_predictor(void)
638 {
639 	return cpus_have_const_cap(ARM64_HARDEN_BRANCH_PREDICTOR);
640 }
641 
642 #define KVM_SSBD_UNKNOWN		-1
643 #define KVM_SSBD_FORCE_DISABLE		0
644 #define KVM_SSBD_KERNEL		1
645 #define KVM_SSBD_FORCE_ENABLE		2
646 #define KVM_SSBD_MITIGATED		3
647 
648 static inline int kvm_arm_have_ssbd(void)
649 {
650 	switch (arm64_get_ssbd_state()) {
651 	case ARM64_SSBD_FORCE_DISABLE:
652 		return KVM_SSBD_FORCE_DISABLE;
653 	case ARM64_SSBD_KERNEL:
654 		return KVM_SSBD_KERNEL;
655 	case ARM64_SSBD_FORCE_ENABLE:
656 		return KVM_SSBD_FORCE_ENABLE;
657 	case ARM64_SSBD_MITIGATED:
658 		return KVM_SSBD_MITIGATED;
659 	case ARM64_SSBD_UNKNOWN:
660 	default:
661 		return KVM_SSBD_UNKNOWN;
662 	}
663 }
664 
665 void kvm_vcpu_load_sysregs(struct kvm_vcpu *vcpu);
666 void kvm_vcpu_put_sysregs(struct kvm_vcpu *vcpu);
667 
668 void kvm_set_ipa_limit(void);
669 
670 #define __KVM_HAVE_ARCH_VM_ALLOC
671 struct kvm *kvm_arch_alloc_vm(void);
672 void kvm_arch_free_vm(struct kvm *kvm);
673 
674 int kvm_arm_setup_stage2(struct kvm *kvm, unsigned long type);
675 
676 int kvm_arm_vcpu_finalize(struct kvm_vcpu *vcpu, int feature);
677 bool kvm_arm_vcpu_is_finalized(struct kvm_vcpu *vcpu);
678 
679 #define kvm_arm_vcpu_sve_finalized(vcpu) \
680 	((vcpu)->arch.flags & KVM_ARM64_VCPU_SVE_FINALIZED)
681 
682 #endif /* __ARM64_KVM_HOST_H__ */
683