xref: /linux/arch/arm64/include/asm/kvm_host.h (revision 55f3538c4923e9dfca132e99ebec370e8094afda)
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 int __attribute_const__ kvm_target_cpu(void);
52 int kvm_reset_vcpu(struct kvm_vcpu *vcpu);
53 int kvm_arch_dev_ioctl_check_extension(struct kvm *kvm, long ext);
54 void __extended_idmap_trampoline(phys_addr_t boot_pgd, phys_addr_t idmap_start);
55 
56 struct kvm_arch {
57 	/* The VMID generation used for the virt. memory system */
58 	u64    vmid_gen;
59 	u32    vmid;
60 
61 	/* 1-level 2nd stage table and lock */
62 	spinlock_t pgd_lock;
63 	pgd_t *pgd;
64 
65 	/* VTTBR value associated with above pgd and vmid */
66 	u64    vttbr;
67 
68 	/* The last vcpu id that ran on each physical CPU */
69 	int __percpu *last_vcpu_ran;
70 
71 	/* The maximum number of vCPUs depends on the used GIC model */
72 	int max_vcpus;
73 
74 	/* Interrupt controller */
75 	struct vgic_dist	vgic;
76 };
77 
78 #define KVM_NR_MEM_OBJS     40
79 
80 /*
81  * We don't want allocation failures within the mmu code, so we preallocate
82  * enough memory for a single page fault in a cache.
83  */
84 struct kvm_mmu_memory_cache {
85 	int nobjs;
86 	void *objects[KVM_NR_MEM_OBJS];
87 };
88 
89 struct kvm_vcpu_fault_info {
90 	u32 esr_el2;		/* Hyp Syndrom Register */
91 	u64 far_el2;		/* Hyp Fault Address Register */
92 	u64 hpfar_el2;		/* Hyp IPA Fault Address Register */
93 	u64 disr_el1;		/* Deferred [SError] Status Register */
94 };
95 
96 /*
97  * 0 is reserved as an invalid value.
98  * Order should be kept in sync with the save/restore code.
99  */
100 enum vcpu_sysreg {
101 	__INVALID_SYSREG__,
102 	MPIDR_EL1,	/* MultiProcessor Affinity Register */
103 	CSSELR_EL1,	/* Cache Size Selection Register */
104 	SCTLR_EL1,	/* System Control Register */
105 	ACTLR_EL1,	/* Auxiliary Control Register */
106 	CPACR_EL1,	/* Coprocessor Access Control */
107 	TTBR0_EL1,	/* Translation Table Base Register 0 */
108 	TTBR1_EL1,	/* Translation Table Base Register 1 */
109 	TCR_EL1,	/* Translation Control Register */
110 	ESR_EL1,	/* Exception Syndrome Register */
111 	AFSR0_EL1,	/* Auxiliary Fault Status Register 0 */
112 	AFSR1_EL1,	/* Auxiliary Fault Status Register 1 */
113 	FAR_EL1,	/* Fault Address Register */
114 	MAIR_EL1,	/* Memory Attribute Indirection Register */
115 	VBAR_EL1,	/* Vector Base Address Register */
116 	CONTEXTIDR_EL1,	/* Context ID Register */
117 	TPIDR_EL0,	/* Thread ID, User R/W */
118 	TPIDRRO_EL0,	/* Thread ID, User R/O */
119 	TPIDR_EL1,	/* Thread ID, Privileged */
120 	AMAIR_EL1,	/* Aux Memory Attribute Indirection Register */
121 	CNTKCTL_EL1,	/* Timer Control Register (EL1) */
122 	PAR_EL1,	/* Physical Address Register */
123 	MDSCR_EL1,	/* Monitor Debug System Control Register */
124 	MDCCINT_EL1,	/* Monitor Debug Comms Channel Interrupt Enable Reg */
125 	DISR_EL1,	/* Deferred Interrupt Status Register */
126 
127 	/* Performance Monitors Registers */
128 	PMCR_EL0,	/* Control Register */
129 	PMSELR_EL0,	/* Event Counter Selection Register */
130 	PMEVCNTR0_EL0,	/* Event Counter Register (0-30) */
131 	PMEVCNTR30_EL0 = PMEVCNTR0_EL0 + 30,
132 	PMCCNTR_EL0,	/* Cycle Counter Register */
133 	PMEVTYPER0_EL0,	/* Event Type Register (0-30) */
134 	PMEVTYPER30_EL0 = PMEVTYPER0_EL0 + 30,
135 	PMCCFILTR_EL0,	/* Cycle Count Filter Register */
136 	PMCNTENSET_EL0,	/* Count Enable Set Register */
137 	PMINTENSET_EL1,	/* Interrupt Enable Set Register */
138 	PMOVSSET_EL0,	/* Overflow Flag Status Set Register */
139 	PMSWINC_EL0,	/* Software Increment Register */
140 	PMUSERENR_EL0,	/* User Enable Register */
141 
142 	/* 32bit specific registers. Keep them at the end of the range */
143 	DACR32_EL2,	/* Domain Access Control Register */
144 	IFSR32_EL2,	/* Instruction Fault Status Register */
145 	FPEXC32_EL2,	/* Floating-Point Exception Control Register */
146 	DBGVCR32_EL2,	/* Debug Vector Catch Register */
147 
148 	NR_SYS_REGS	/* Nothing after this line! */
149 };
150 
151 /* 32bit mapping */
152 #define c0_MPIDR	(MPIDR_EL1 * 2)	/* MultiProcessor ID Register */
153 #define c0_CSSELR	(CSSELR_EL1 * 2)/* Cache Size Selection Register */
154 #define c1_SCTLR	(SCTLR_EL1 * 2)	/* System Control Register */
155 #define c1_ACTLR	(ACTLR_EL1 * 2)	/* Auxiliary Control Register */
156 #define c1_CPACR	(CPACR_EL1 * 2)	/* Coprocessor Access Control */
157 #define c2_TTBR0	(TTBR0_EL1 * 2)	/* Translation Table Base Register 0 */
158 #define c2_TTBR0_high	(c2_TTBR0 + 1)	/* TTBR0 top 32 bits */
159 #define c2_TTBR1	(TTBR1_EL1 * 2)	/* Translation Table Base Register 1 */
160 #define c2_TTBR1_high	(c2_TTBR1 + 1)	/* TTBR1 top 32 bits */
161 #define c2_TTBCR	(TCR_EL1 * 2)	/* Translation Table Base Control R. */
162 #define c3_DACR		(DACR32_EL2 * 2)/* Domain Access Control Register */
163 #define c5_DFSR		(ESR_EL1 * 2)	/* Data Fault Status Register */
164 #define c5_IFSR		(IFSR32_EL2 * 2)/* Instruction Fault Status Register */
165 #define c5_ADFSR	(AFSR0_EL1 * 2)	/* Auxiliary Data Fault Status R */
166 #define c5_AIFSR	(AFSR1_EL1 * 2)	/* Auxiliary Instr Fault Status R */
167 #define c6_DFAR		(FAR_EL1 * 2)	/* Data Fault Address Register */
168 #define c6_IFAR		(c6_DFAR + 1)	/* Instruction Fault Address Register */
169 #define c7_PAR		(PAR_EL1 * 2)	/* Physical Address Register */
170 #define c7_PAR_high	(c7_PAR + 1)	/* PAR top 32 bits */
171 #define c10_PRRR	(MAIR_EL1 * 2)	/* Primary Region Remap Register */
172 #define c10_NMRR	(c10_PRRR + 1)	/* Normal Memory Remap Register */
173 #define c12_VBAR	(VBAR_EL1 * 2)	/* Vector Base Address Register */
174 #define c13_CID		(CONTEXTIDR_EL1 * 2)	/* Context ID Register */
175 #define c13_TID_URW	(TPIDR_EL0 * 2)	/* Thread ID, User R/W */
176 #define c13_TID_URO	(TPIDRRO_EL0 * 2)/* Thread ID, User R/O */
177 #define c13_TID_PRIV	(TPIDR_EL1 * 2)	/* Thread ID, Privileged */
178 #define c10_AMAIR0	(AMAIR_EL1 * 2)	/* Aux Memory Attr Indirection Reg */
179 #define c10_AMAIR1	(c10_AMAIR0 + 1)/* Aux Memory Attr Indirection Reg */
180 #define c14_CNTKCTL	(CNTKCTL_EL1 * 2) /* Timer Control Register (PL1) */
181 
182 #define cp14_DBGDSCRext	(MDSCR_EL1 * 2)
183 #define cp14_DBGBCR0	(DBGBCR0_EL1 * 2)
184 #define cp14_DBGBVR0	(DBGBVR0_EL1 * 2)
185 #define cp14_DBGBXVR0	(cp14_DBGBVR0 + 1)
186 #define cp14_DBGWCR0	(DBGWCR0_EL1 * 2)
187 #define cp14_DBGWVR0	(DBGWVR0_EL1 * 2)
188 #define cp14_DBGDCCINT	(MDCCINT_EL1 * 2)
189 
190 #define NR_COPRO_REGS	(NR_SYS_REGS * 2)
191 
192 struct kvm_cpu_context {
193 	struct kvm_regs	gp_regs;
194 	union {
195 		u64 sys_regs[NR_SYS_REGS];
196 		u32 copro[NR_COPRO_REGS];
197 	};
198 
199 	struct kvm_vcpu *__hyp_running_vcpu;
200 };
201 
202 typedef struct kvm_cpu_context kvm_cpu_context_t;
203 
204 struct kvm_vcpu_arch {
205 	struct kvm_cpu_context ctxt;
206 
207 	/* HYP configuration */
208 	u64 hcr_el2;
209 	u32 mdcr_el2;
210 
211 	/* Exception Information */
212 	struct kvm_vcpu_fault_info fault;
213 
214 	/* Guest debug state */
215 	u64 debug_flags;
216 
217 	/*
218 	 * We maintain more than a single set of debug registers to support
219 	 * debugging the guest from the host and to maintain separate host and
220 	 * guest state during world switches. vcpu_debug_state are the debug
221 	 * registers of the vcpu as the guest sees them.  host_debug_state are
222 	 * the host registers which are saved and restored during
223 	 * world switches. external_debug_state contains the debug
224 	 * values we want to debug the guest. This is set via the
225 	 * KVM_SET_GUEST_DEBUG ioctl.
226 	 *
227 	 * debug_ptr points to the set of debug registers that should be loaded
228 	 * onto the hardware when running the guest.
229 	 */
230 	struct kvm_guest_debug_arch *debug_ptr;
231 	struct kvm_guest_debug_arch vcpu_debug_state;
232 	struct kvm_guest_debug_arch external_debug_state;
233 
234 	/* Pointer to host CPU context */
235 	kvm_cpu_context_t *host_cpu_context;
236 	struct {
237 		/* {Break,watch}point registers */
238 		struct kvm_guest_debug_arch regs;
239 		/* Statistical profiling extension */
240 		u64 pmscr_el1;
241 	} host_debug_state;
242 
243 	/* VGIC state */
244 	struct vgic_cpu vgic_cpu;
245 	struct arch_timer_cpu timer_cpu;
246 	struct kvm_pmu pmu;
247 
248 	/*
249 	 * Anything that is not used directly from assembly code goes
250 	 * here.
251 	 */
252 
253 	/*
254 	 * Guest registers we preserve during guest debugging.
255 	 *
256 	 * These shadow registers are updated by the kvm_handle_sys_reg
257 	 * trap handler if the guest accesses or updates them while we
258 	 * are using guest debug.
259 	 */
260 	struct {
261 		u32	mdscr_el1;
262 	} guest_debug_preserved;
263 
264 	/* vcpu power-off state */
265 	bool power_off;
266 
267 	/* Don't run the guest (internal implementation need) */
268 	bool pause;
269 
270 	/* IO related fields */
271 	struct kvm_decode mmio_decode;
272 
273 	/* Interrupt related fields */
274 	u64 irq_lines;		/* IRQ and FIQ levels */
275 
276 	/* Cache some mmu pages needed inside spinlock regions */
277 	struct kvm_mmu_memory_cache mmu_page_cache;
278 
279 	/* Target CPU and feature flags */
280 	int target;
281 	DECLARE_BITMAP(features, KVM_VCPU_MAX_FEATURES);
282 
283 	/* Detect first run of a vcpu */
284 	bool has_run_once;
285 
286 	/* Virtual SError ESR to restore when HCR_EL2.VSE is set */
287 	u64 vsesr_el2;
288 };
289 
290 #define vcpu_gp_regs(v)		(&(v)->arch.ctxt.gp_regs)
291 #define vcpu_sys_reg(v,r)	((v)->arch.ctxt.sys_regs[(r)])
292 /*
293  * CP14 and CP15 live in the same array, as they are backed by the
294  * same system registers.
295  */
296 #define vcpu_cp14(v,r)		((v)->arch.ctxt.copro[(r)])
297 #define vcpu_cp15(v,r)		((v)->arch.ctxt.copro[(r)])
298 
299 #ifdef CONFIG_CPU_BIG_ENDIAN
300 #define vcpu_cp15_64_high(v,r)	vcpu_cp15((v),(r))
301 #define vcpu_cp15_64_low(v,r)	vcpu_cp15((v),(r) + 1)
302 #else
303 #define vcpu_cp15_64_high(v,r)	vcpu_cp15((v),(r) + 1)
304 #define vcpu_cp15_64_low(v,r)	vcpu_cp15((v),(r))
305 #endif
306 
307 struct kvm_vm_stat {
308 	ulong remote_tlb_flush;
309 };
310 
311 struct kvm_vcpu_stat {
312 	u64 halt_successful_poll;
313 	u64 halt_attempted_poll;
314 	u64 halt_poll_invalid;
315 	u64 halt_wakeup;
316 	u64 hvc_exit_stat;
317 	u64 wfe_exit_stat;
318 	u64 wfi_exit_stat;
319 	u64 mmio_exit_user;
320 	u64 mmio_exit_kernel;
321 	u64 exits;
322 };
323 
324 int kvm_vcpu_preferred_target(struct kvm_vcpu_init *init);
325 unsigned long kvm_arm_num_regs(struct kvm_vcpu *vcpu);
326 int kvm_arm_copy_reg_indices(struct kvm_vcpu *vcpu, u64 __user *indices);
327 int kvm_arm_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
328 int kvm_arm_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
329 
330 #define KVM_ARCH_WANT_MMU_NOTIFIER
331 int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
332 int kvm_unmap_hva_range(struct kvm *kvm,
333 			unsigned long start, unsigned long end);
334 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
335 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
336 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
337 
338 struct kvm_vcpu *kvm_arm_get_running_vcpu(void);
339 struct kvm_vcpu * __percpu *kvm_get_running_vcpus(void);
340 void kvm_arm_halt_guest(struct kvm *kvm);
341 void kvm_arm_resume_guest(struct kvm *kvm);
342 
343 u64 __kvm_call_hyp(void *hypfn, ...);
344 #define kvm_call_hyp(f, ...) __kvm_call_hyp(kvm_ksym_ref(f), ##__VA_ARGS__)
345 
346 void force_vm_exit(const cpumask_t *mask);
347 void kvm_mmu_wp_memory_region(struct kvm *kvm, int slot);
348 
349 int handle_exit(struct kvm_vcpu *vcpu, struct kvm_run *run,
350 		int exception_index);
351 void handle_exit_early(struct kvm_vcpu *vcpu, struct kvm_run *run,
352 		       int exception_index);
353 
354 int kvm_perf_init(void);
355 int kvm_perf_teardown(void);
356 
357 struct kvm_vcpu *kvm_mpidr_to_vcpu(struct kvm *kvm, unsigned long mpidr);
358 
359 static inline void __cpu_init_hyp_mode(phys_addr_t pgd_ptr,
360 				       unsigned long hyp_stack_ptr,
361 				       unsigned long vector_ptr)
362 {
363 	/*
364 	 * Call initialization code, and switch to the full blown HYP code.
365 	 * If the cpucaps haven't been finalized yet, something has gone very
366 	 * wrong, and hyp will crash and burn when it uses any
367 	 * cpus_have_const_cap() wrapper.
368 	 */
369 	BUG_ON(!static_branch_likely(&arm64_const_caps_ready));
370 	__kvm_call_hyp((void *)pgd_ptr, hyp_stack_ptr, vector_ptr);
371 }
372 
373 static inline void kvm_arch_hardware_unsetup(void) {}
374 static inline void kvm_arch_sync_events(struct kvm *kvm) {}
375 static inline void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu) {}
376 static inline void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu) {}
377 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
378 
379 void kvm_arm_init_debug(void);
380 void kvm_arm_setup_debug(struct kvm_vcpu *vcpu);
381 void kvm_arm_clear_debug(struct kvm_vcpu *vcpu);
382 void kvm_arm_reset_debug_ptr(struct kvm_vcpu *vcpu);
383 bool kvm_arm_handle_step_debug(struct kvm_vcpu *vcpu, struct kvm_run *run);
384 int kvm_arm_vcpu_arch_set_attr(struct kvm_vcpu *vcpu,
385 			       struct kvm_device_attr *attr);
386 int kvm_arm_vcpu_arch_get_attr(struct kvm_vcpu *vcpu,
387 			       struct kvm_device_attr *attr);
388 int kvm_arm_vcpu_arch_has_attr(struct kvm_vcpu *vcpu,
389 			       struct kvm_device_attr *attr);
390 
391 static inline void __cpu_init_stage2(void)
392 {
393 	u32 parange = kvm_call_hyp(__init_stage2_translation);
394 
395 	WARN_ONCE(parange < 40,
396 		  "PARange is %d bits, unsupported configuration!", parange);
397 }
398 
399 /*
400  * All host FP/SIMD state is restored on guest exit, so nothing needs
401  * doing here except in the SVE case:
402 */
403 static inline void kvm_fpsimd_flush_cpu_state(void)
404 {
405 	if (system_supports_sve())
406 		sve_flush_cpu_state();
407 }
408 
409 static inline void kvm_arm_vhe_guest_enter(void)
410 {
411 	local_daif_mask();
412 }
413 
414 static inline void kvm_arm_vhe_guest_exit(void)
415 {
416 	local_daif_restore(DAIF_PROCCTX_NOIRQ);
417 }
418 #endif /* __ARM64_KVM_HOST_H__ */
419