xref: /openbmc/linux/arch/arm64/include/asm/kvm_host.h (revision addee42a)
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/types.h>
26 #include <linux/kvm_types.h>
27 #include <asm/cpufeature.h>
28 #include <asm/daifflags.h>
29 #include <asm/fpsimd.h>
30 #include <asm/kvm.h>
31 #include <asm/kvm_asm.h>
32 #include <asm/kvm_mmio.h>
33 
34 #define __KVM_HAVE_ARCH_INTC_INITIALIZED
35 
36 #define KVM_USER_MEM_SLOTS 512
37 #define KVM_HALT_POLL_NS_DEFAULT 500000
38 
39 #include <kvm/arm_vgic.h>
40 #include <kvm/arm_arch_timer.h>
41 #include <kvm/arm_pmu.h>
42 
43 #define KVM_MAX_VCPUS VGIC_V3_MAX_CPUS
44 
45 #define KVM_VCPU_MAX_FEATURES 4
46 
47 #define KVM_REQ_SLEEP \
48 	KVM_ARCH_REQ_FLAGS(0, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
49 #define KVM_REQ_IRQ_PENDING	KVM_ARCH_REQ(1)
50 
51 DECLARE_STATIC_KEY_FALSE(userspace_irqchip_in_use);
52 
53 int __attribute_const__ kvm_target_cpu(void);
54 int kvm_reset_vcpu(struct kvm_vcpu *vcpu);
55 int kvm_arch_dev_ioctl_check_extension(struct kvm *kvm, long ext);
56 void __extended_idmap_trampoline(phys_addr_t boot_pgd, phys_addr_t idmap_start);
57 
58 struct kvm_arch {
59 	/* The VMID generation used for the virt. memory system */
60 	u64    vmid_gen;
61 	u32    vmid;
62 
63 	/* 1-level 2nd stage table and lock */
64 	spinlock_t pgd_lock;
65 	pgd_t *pgd;
66 
67 	/* VTTBR value associated with above pgd and vmid */
68 	u64    vttbr;
69 
70 	/* The last vcpu id that ran on each physical CPU */
71 	int __percpu *last_vcpu_ran;
72 
73 	/* The maximum number of vCPUs depends on the used GIC model */
74 	int max_vcpus;
75 
76 	/* Interrupt controller */
77 	struct vgic_dist	vgic;
78 
79 	/* Mandated version of PSCI */
80 	u32 psci_version;
81 };
82 
83 #define KVM_NR_MEM_OBJS     40
84 
85 /*
86  * We don't want allocation failures within the mmu code, so we preallocate
87  * enough memory for a single page fault in a cache.
88  */
89 struct kvm_mmu_memory_cache {
90 	int nobjs;
91 	void *objects[KVM_NR_MEM_OBJS];
92 };
93 
94 struct kvm_vcpu_fault_info {
95 	u32 esr_el2;		/* Hyp Syndrom Register */
96 	u64 far_el2;		/* Hyp Fault Address Register */
97 	u64 hpfar_el2;		/* Hyp IPA Fault Address Register */
98 	u64 disr_el1;		/* Deferred [SError] Status Register */
99 };
100 
101 /*
102  * 0 is reserved as an invalid value.
103  * Order should be kept in sync with the save/restore code.
104  */
105 enum vcpu_sysreg {
106 	__INVALID_SYSREG__,
107 	MPIDR_EL1,	/* MultiProcessor Affinity Register */
108 	CSSELR_EL1,	/* Cache Size Selection Register */
109 	SCTLR_EL1,	/* System Control Register */
110 	ACTLR_EL1,	/* Auxiliary Control Register */
111 	CPACR_EL1,	/* Coprocessor Access Control */
112 	TTBR0_EL1,	/* Translation Table Base Register 0 */
113 	TTBR1_EL1,	/* Translation Table Base Register 1 */
114 	TCR_EL1,	/* Translation Control Register */
115 	ESR_EL1,	/* Exception Syndrome Register */
116 	AFSR0_EL1,	/* Auxiliary Fault Status Register 0 */
117 	AFSR1_EL1,	/* Auxiliary Fault Status Register 1 */
118 	FAR_EL1,	/* Fault Address Register */
119 	MAIR_EL1,	/* Memory Attribute Indirection Register */
120 	VBAR_EL1,	/* Vector Base Address Register */
121 	CONTEXTIDR_EL1,	/* Context ID Register */
122 	TPIDR_EL0,	/* Thread ID, User R/W */
123 	TPIDRRO_EL0,	/* Thread ID, User R/O */
124 	TPIDR_EL1,	/* Thread ID, Privileged */
125 	AMAIR_EL1,	/* Aux Memory Attribute Indirection Register */
126 	CNTKCTL_EL1,	/* Timer Control Register (EL1) */
127 	PAR_EL1,	/* Physical Address Register */
128 	MDSCR_EL1,	/* Monitor Debug System Control Register */
129 	MDCCINT_EL1,	/* Monitor Debug Comms Channel Interrupt Enable Reg */
130 	DISR_EL1,	/* Deferred Interrupt Status Register */
131 
132 	/* Performance Monitors Registers */
133 	PMCR_EL0,	/* Control Register */
134 	PMSELR_EL0,	/* Event Counter Selection Register */
135 	PMEVCNTR0_EL0,	/* Event Counter Register (0-30) */
136 	PMEVCNTR30_EL0 = PMEVCNTR0_EL0 + 30,
137 	PMCCNTR_EL0,	/* Cycle Counter Register */
138 	PMEVTYPER0_EL0,	/* Event Type Register (0-30) */
139 	PMEVTYPER30_EL0 = PMEVTYPER0_EL0 + 30,
140 	PMCCFILTR_EL0,	/* Cycle Count Filter Register */
141 	PMCNTENSET_EL0,	/* Count Enable Set Register */
142 	PMINTENSET_EL1,	/* Interrupt Enable Set Register */
143 	PMOVSSET_EL0,	/* Overflow Flag Status Set Register */
144 	PMSWINC_EL0,	/* Software Increment Register */
145 	PMUSERENR_EL0,	/* User Enable Register */
146 
147 	/* 32bit specific registers. Keep them at the end of the range */
148 	DACR32_EL2,	/* Domain Access Control Register */
149 	IFSR32_EL2,	/* Instruction Fault Status Register */
150 	FPEXC32_EL2,	/* Floating-Point Exception Control Register */
151 	DBGVCR32_EL2,	/* Debug Vector Catch Register */
152 
153 	NR_SYS_REGS	/* Nothing after this line! */
154 };
155 
156 /* 32bit mapping */
157 #define c0_MPIDR	(MPIDR_EL1 * 2)	/* MultiProcessor ID Register */
158 #define c0_CSSELR	(CSSELR_EL1 * 2)/* Cache Size Selection Register */
159 #define c1_SCTLR	(SCTLR_EL1 * 2)	/* System Control Register */
160 #define c1_ACTLR	(ACTLR_EL1 * 2)	/* Auxiliary Control Register */
161 #define c1_CPACR	(CPACR_EL1 * 2)	/* Coprocessor Access Control */
162 #define c2_TTBR0	(TTBR0_EL1 * 2)	/* Translation Table Base Register 0 */
163 #define c2_TTBR0_high	(c2_TTBR0 + 1)	/* TTBR0 top 32 bits */
164 #define c2_TTBR1	(TTBR1_EL1 * 2)	/* Translation Table Base Register 1 */
165 #define c2_TTBR1_high	(c2_TTBR1 + 1)	/* TTBR1 top 32 bits */
166 #define c2_TTBCR	(TCR_EL1 * 2)	/* Translation Table Base Control R. */
167 #define c3_DACR		(DACR32_EL2 * 2)/* Domain Access Control Register */
168 #define c5_DFSR		(ESR_EL1 * 2)	/* Data Fault Status Register */
169 #define c5_IFSR		(IFSR32_EL2 * 2)/* Instruction Fault Status Register */
170 #define c5_ADFSR	(AFSR0_EL1 * 2)	/* Auxiliary Data Fault Status R */
171 #define c5_AIFSR	(AFSR1_EL1 * 2)	/* Auxiliary Instr Fault Status R */
172 #define c6_DFAR		(FAR_EL1 * 2)	/* Data Fault Address Register */
173 #define c6_IFAR		(c6_DFAR + 1)	/* Instruction Fault Address Register */
174 #define c7_PAR		(PAR_EL1 * 2)	/* Physical Address Register */
175 #define c7_PAR_high	(c7_PAR + 1)	/* PAR top 32 bits */
176 #define c10_PRRR	(MAIR_EL1 * 2)	/* Primary Region Remap Register */
177 #define c10_NMRR	(c10_PRRR + 1)	/* Normal Memory Remap Register */
178 #define c12_VBAR	(VBAR_EL1 * 2)	/* Vector Base Address Register */
179 #define c13_CID		(CONTEXTIDR_EL1 * 2)	/* Context ID Register */
180 #define c13_TID_URW	(TPIDR_EL0 * 2)	/* Thread ID, User R/W */
181 #define c13_TID_URO	(TPIDRRO_EL0 * 2)/* Thread ID, User R/O */
182 #define c13_TID_PRIV	(TPIDR_EL1 * 2)	/* Thread ID, Privileged */
183 #define c10_AMAIR0	(AMAIR_EL1 * 2)	/* Aux Memory Attr Indirection Reg */
184 #define c10_AMAIR1	(c10_AMAIR0 + 1)/* Aux Memory Attr Indirection Reg */
185 #define c14_CNTKCTL	(CNTKCTL_EL1 * 2) /* Timer Control Register (PL1) */
186 
187 #define cp14_DBGDSCRext	(MDSCR_EL1 * 2)
188 #define cp14_DBGBCR0	(DBGBCR0_EL1 * 2)
189 #define cp14_DBGBVR0	(DBGBVR0_EL1 * 2)
190 #define cp14_DBGBXVR0	(cp14_DBGBVR0 + 1)
191 #define cp14_DBGWCR0	(DBGWCR0_EL1 * 2)
192 #define cp14_DBGWVR0	(DBGWVR0_EL1 * 2)
193 #define cp14_DBGDCCINT	(MDCCINT_EL1 * 2)
194 
195 #define NR_COPRO_REGS	(NR_SYS_REGS * 2)
196 
197 struct kvm_cpu_context {
198 	struct kvm_regs	gp_regs;
199 	union {
200 		u64 sys_regs[NR_SYS_REGS];
201 		u32 copro[NR_COPRO_REGS];
202 	};
203 
204 	struct kvm_vcpu *__hyp_running_vcpu;
205 };
206 
207 typedef struct kvm_cpu_context kvm_cpu_context_t;
208 
209 struct kvm_vcpu_arch {
210 	struct kvm_cpu_context ctxt;
211 
212 	/* HYP configuration */
213 	u64 hcr_el2;
214 	u32 mdcr_el2;
215 
216 	/* Exception Information */
217 	struct kvm_vcpu_fault_info fault;
218 
219 	/* Guest debug state */
220 	u64 debug_flags;
221 
222 	/*
223 	 * We maintain more than a single set of debug registers to support
224 	 * debugging the guest from the host and to maintain separate host and
225 	 * guest state during world switches. vcpu_debug_state are the debug
226 	 * registers of the vcpu as the guest sees them.  host_debug_state are
227 	 * the host registers which are saved and restored during
228 	 * world switches. external_debug_state contains the debug
229 	 * values we want to debug the guest. This is set via the
230 	 * KVM_SET_GUEST_DEBUG ioctl.
231 	 *
232 	 * debug_ptr points to the set of debug registers that should be loaded
233 	 * onto the hardware when running the guest.
234 	 */
235 	struct kvm_guest_debug_arch *debug_ptr;
236 	struct kvm_guest_debug_arch vcpu_debug_state;
237 	struct kvm_guest_debug_arch external_debug_state;
238 
239 	/* Pointer to host CPU context */
240 	kvm_cpu_context_t *host_cpu_context;
241 	struct {
242 		/* {Break,watch}point registers */
243 		struct kvm_guest_debug_arch regs;
244 		/* Statistical profiling extension */
245 		u64 pmscr_el1;
246 	} host_debug_state;
247 
248 	/* VGIC state */
249 	struct vgic_cpu vgic_cpu;
250 	struct arch_timer_cpu timer_cpu;
251 	struct kvm_pmu pmu;
252 
253 	/*
254 	 * Anything that is not used directly from assembly code goes
255 	 * here.
256 	 */
257 
258 	/*
259 	 * Guest registers we preserve during guest debugging.
260 	 *
261 	 * These shadow registers are updated by the kvm_handle_sys_reg
262 	 * trap handler if the guest accesses or updates them while we
263 	 * are using guest debug.
264 	 */
265 	struct {
266 		u32	mdscr_el1;
267 	} guest_debug_preserved;
268 
269 	/* vcpu power-off state */
270 	bool power_off;
271 
272 	/* Don't run the guest (internal implementation need) */
273 	bool pause;
274 
275 	/* IO related fields */
276 	struct kvm_decode mmio_decode;
277 
278 	/* Cache some mmu pages needed inside spinlock regions */
279 	struct kvm_mmu_memory_cache mmu_page_cache;
280 
281 	/* Target CPU and feature flags */
282 	int target;
283 	DECLARE_BITMAP(features, KVM_VCPU_MAX_FEATURES);
284 
285 	/* Detect first run of a vcpu */
286 	bool has_run_once;
287 
288 	/* Virtual SError ESR to restore when HCR_EL2.VSE is set */
289 	u64 vsesr_el2;
290 
291 	/* True when deferrable sysregs are loaded on the physical CPU,
292 	 * see kvm_vcpu_load_sysregs and kvm_vcpu_put_sysregs. */
293 	bool sysregs_loaded_on_cpu;
294 };
295 
296 #define vcpu_gp_regs(v)		(&(v)->arch.ctxt.gp_regs)
297 
298 /*
299  * Only use __vcpu_sys_reg if you know you want the memory backed version of a
300  * register, and not the one most recently accessed by a running VCPU.  For
301  * example, for userspace access or for system registers that are never context
302  * switched, but only emulated.
303  */
304 #define __vcpu_sys_reg(v,r)	((v)->arch.ctxt.sys_regs[(r)])
305 
306 u64 vcpu_read_sys_reg(struct kvm_vcpu *vcpu, int reg);
307 void vcpu_write_sys_reg(struct kvm_vcpu *vcpu, u64 val, int reg);
308 
309 /*
310  * CP14 and CP15 live in the same array, as they are backed by the
311  * same system registers.
312  */
313 #define vcpu_cp14(v,r)		((v)->arch.ctxt.copro[(r)])
314 #define vcpu_cp15(v,r)		((v)->arch.ctxt.copro[(r)])
315 
316 struct kvm_vm_stat {
317 	ulong remote_tlb_flush;
318 };
319 
320 struct kvm_vcpu_stat {
321 	u64 halt_successful_poll;
322 	u64 halt_attempted_poll;
323 	u64 halt_poll_invalid;
324 	u64 halt_wakeup;
325 	u64 hvc_exit_stat;
326 	u64 wfe_exit_stat;
327 	u64 wfi_exit_stat;
328 	u64 mmio_exit_user;
329 	u64 mmio_exit_kernel;
330 	u64 exits;
331 };
332 
333 int kvm_vcpu_preferred_target(struct kvm_vcpu_init *init);
334 unsigned long kvm_arm_num_regs(struct kvm_vcpu *vcpu);
335 int kvm_arm_copy_reg_indices(struct kvm_vcpu *vcpu, u64 __user *indices);
336 int kvm_arm_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
337 int kvm_arm_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
338 
339 #define KVM_ARCH_WANT_MMU_NOTIFIER
340 int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
341 int kvm_unmap_hva_range(struct kvm *kvm,
342 			unsigned long start, unsigned long end);
343 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
344 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
345 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
346 
347 struct kvm_vcpu *kvm_arm_get_running_vcpu(void);
348 struct kvm_vcpu * __percpu *kvm_get_running_vcpus(void);
349 void kvm_arm_halt_guest(struct kvm *kvm);
350 void kvm_arm_resume_guest(struct kvm *kvm);
351 
352 u64 __kvm_call_hyp(void *hypfn, ...);
353 #define kvm_call_hyp(f, ...) __kvm_call_hyp(kvm_ksym_ref(f), ##__VA_ARGS__)
354 
355 void force_vm_exit(const cpumask_t *mask);
356 void kvm_mmu_wp_memory_region(struct kvm *kvm, int slot);
357 
358 int handle_exit(struct kvm_vcpu *vcpu, struct kvm_run *run,
359 		int exception_index);
360 void handle_exit_early(struct kvm_vcpu *vcpu, struct kvm_run *run,
361 		       int exception_index);
362 
363 int kvm_perf_init(void);
364 int kvm_perf_teardown(void);
365 
366 struct kvm_vcpu *kvm_mpidr_to_vcpu(struct kvm *kvm, unsigned long mpidr);
367 
368 void __kvm_set_tpidr_el2(u64 tpidr_el2);
369 DECLARE_PER_CPU(kvm_cpu_context_t, kvm_host_cpu_state);
370 
371 static inline void __cpu_init_hyp_mode(phys_addr_t pgd_ptr,
372 				       unsigned long hyp_stack_ptr,
373 				       unsigned long vector_ptr)
374 {
375 	u64 tpidr_el2;
376 
377 	/*
378 	 * Call initialization code, and switch to the full blown HYP code.
379 	 * If the cpucaps haven't been finalized yet, something has gone very
380 	 * wrong, and hyp will crash and burn when it uses any
381 	 * cpus_have_const_cap() wrapper.
382 	 */
383 	BUG_ON(!static_branch_likely(&arm64_const_caps_ready));
384 	__kvm_call_hyp((void *)pgd_ptr, hyp_stack_ptr, vector_ptr);
385 
386 	/*
387 	 * Calculate the raw per-cpu offset without a translation from the
388 	 * kernel's mapping to the linear mapping, and store it in tpidr_el2
389 	 * so that we can use adr_l to access per-cpu variables in EL2.
390 	 */
391 	tpidr_el2 = (u64)this_cpu_ptr(&kvm_host_cpu_state)
392 		- (u64)kvm_ksym_ref(kvm_host_cpu_state);
393 
394 	kvm_call_hyp(__kvm_set_tpidr_el2, tpidr_el2);
395 }
396 
397 static inline void kvm_arch_hardware_unsetup(void) {}
398 static inline void kvm_arch_sync_events(struct kvm *kvm) {}
399 static inline void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu) {}
400 static inline void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu) {}
401 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
402 
403 void kvm_arm_init_debug(void);
404 void kvm_arm_setup_debug(struct kvm_vcpu *vcpu);
405 void kvm_arm_clear_debug(struct kvm_vcpu *vcpu);
406 void kvm_arm_reset_debug_ptr(struct kvm_vcpu *vcpu);
407 bool kvm_arm_handle_step_debug(struct kvm_vcpu *vcpu, struct kvm_run *run);
408 int kvm_arm_vcpu_arch_set_attr(struct kvm_vcpu *vcpu,
409 			       struct kvm_device_attr *attr);
410 int kvm_arm_vcpu_arch_get_attr(struct kvm_vcpu *vcpu,
411 			       struct kvm_device_attr *attr);
412 int kvm_arm_vcpu_arch_has_attr(struct kvm_vcpu *vcpu,
413 			       struct kvm_device_attr *attr);
414 
415 static inline void __cpu_init_stage2(void)
416 {
417 	u32 parange = kvm_call_hyp(__init_stage2_translation);
418 
419 	WARN_ONCE(parange < 40,
420 		  "PARange is %d bits, unsupported configuration!", parange);
421 }
422 
423 /*
424  * All host FP/SIMD state is restored on guest exit, so nothing needs
425  * doing here except in the SVE case:
426 */
427 static inline void kvm_fpsimd_flush_cpu_state(void)
428 {
429 	if (system_supports_sve())
430 		sve_flush_cpu_state();
431 }
432 
433 static inline void kvm_arm_vhe_guest_enter(void)
434 {
435 	local_daif_mask();
436 }
437 
438 static inline void kvm_arm_vhe_guest_exit(void)
439 {
440 	local_daif_restore(DAIF_PROCCTX_NOIRQ);
441 
442 	/*
443 	 * When we exit from the guest we change a number of CPU configuration
444 	 * parameters, such as traps.  Make sure these changes take effect
445 	 * before running the host or additional guests.
446 	 */
447 	isb();
448 }
449 
450 static inline bool kvm_arm_harden_branch_predictor(void)
451 {
452 	return cpus_have_const_cap(ARM64_HARDEN_BRANCH_PREDICTOR);
453 }
454 
455 void kvm_vcpu_load_sysregs(struct kvm_vcpu *vcpu);
456 void kvm_vcpu_put_sysregs(struct kvm_vcpu *vcpu);
457 
458 #endif /* __ARM64_KVM_HOST_H__ */
459